xref: /openbmc/linux/drivers/net/ethernet/sfc/net_driver.h (revision 179dd8c0348af75b02c7d72eaaf1cb179f1721ef)
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/rwsem.h>
29 #include <linux/vmalloc.h>
30 #include <linux/i2c.h>
31 #include <linux/mtd/mtd.h>
32 #include <net/busy_poll.h>
33 
34 #include "enum.h"
35 #include "bitfield.h"
36 #include "filter.h"
37 
38 /**************************************************************************
39  *
40  * Build definitions
41  *
42  **************************************************************************/
43 
44 #define EFX_DRIVER_VERSION	"4.0"
45 
46 #ifdef DEBUG
47 #define EFX_BUG_ON_PARANOID(x) BUG_ON(x)
48 #define EFX_WARN_ON_PARANOID(x) WARN_ON(x)
49 #else
50 #define EFX_BUG_ON_PARANOID(x) do {} while (0)
51 #define EFX_WARN_ON_PARANOID(x) do {} while (0)
52 #endif
53 
54 /**************************************************************************
55  *
56  * Efx data structures
57  *
58  **************************************************************************/
59 
60 #define EFX_MAX_CHANNELS 32U
61 #define EFX_MAX_RX_QUEUES EFX_MAX_CHANNELS
62 #define EFX_EXTRA_CHANNEL_IOV	0
63 #define EFX_EXTRA_CHANNEL_PTP	1
64 #define EFX_MAX_EXTRA_CHANNELS	2U
65 
66 /* Checksum generation is a per-queue option in hardware, so each
67  * queue visible to the networking core is backed by two hardware TX
68  * queues. */
69 #define EFX_MAX_TX_TC		2
70 #define EFX_MAX_CORE_TX_QUEUES	(EFX_MAX_TX_TC * EFX_MAX_CHANNELS)
71 #define EFX_TXQ_TYPE_OFFLOAD	1	/* flag */
72 #define EFX_TXQ_TYPE_HIGHPRI	2	/* flag */
73 #define EFX_TXQ_TYPES		4
74 #define EFX_MAX_TX_QUEUES	(EFX_TXQ_TYPES * EFX_MAX_CHANNELS)
75 
76 /* Maximum possible MTU the driver supports */
77 #define EFX_MAX_MTU (9 * 1024)
78 
79 /* Size of an RX scatter buffer.  Small enough to pack 2 into a 4K page,
80  * and should be a multiple of the cache line size.
81  */
82 #define EFX_RX_USR_BUF_SIZE	(2048 - 256)
83 
84 /* If possible, we should ensure cache line alignment at start and end
85  * of every buffer.  Otherwise, we just need to ensure 4-byte
86  * alignment of the network header.
87  */
88 #if NET_IP_ALIGN == 0
89 #define EFX_RX_BUF_ALIGNMENT	L1_CACHE_BYTES
90 #else
91 #define EFX_RX_BUF_ALIGNMENT	4
92 #endif
93 
94 /* Forward declare Precision Time Protocol (PTP) support structure. */
95 struct efx_ptp_data;
96 struct hwtstamp_config;
97 
98 struct efx_self_tests;
99 
100 /**
101  * struct efx_buffer - A general-purpose DMA buffer
102  * @addr: host base address of the buffer
103  * @dma_addr: DMA base address of the buffer
104  * @len: Buffer length, in bytes
105  *
106  * The NIC uses these buffers for its interrupt status registers and
107  * MAC stats dumps.
108  */
109 struct efx_buffer {
110 	void *addr;
111 	dma_addr_t dma_addr;
112 	unsigned int len;
113 };
114 
115 /**
116  * struct efx_special_buffer - DMA buffer entered into buffer table
117  * @buf: Standard &struct efx_buffer
118  * @index: Buffer index within controller;s buffer table
119  * @entries: Number of buffer table entries
120  *
121  * The NIC has a buffer table that maps buffers of size %EFX_BUF_SIZE.
122  * Event and descriptor rings are addressed via one or more buffer
123  * table entries (and so can be physically non-contiguous, although we
124  * currently do not take advantage of that).  On Falcon and Siena we
125  * have to take care of allocating and initialising the entries
126  * ourselves.  On later hardware this is managed by the firmware and
127  * @index and @entries are left as 0.
128  */
129 struct efx_special_buffer {
130 	struct efx_buffer buf;
131 	unsigned int index;
132 	unsigned int entries;
133 };
134 
135 /**
136  * struct efx_tx_buffer - buffer state for a TX descriptor
137  * @skb: When @flags & %EFX_TX_BUF_SKB, the associated socket buffer to be
138  *	freed when descriptor completes
139  * @heap_buf: When @flags & %EFX_TX_BUF_HEAP, the associated heap buffer to be
140  *	freed when descriptor completes.
141  * @option: When @flags & %EFX_TX_BUF_OPTION, a NIC-specific option descriptor.
142  * @dma_addr: DMA address of the fragment.
143  * @flags: Flags for allocation and DMA mapping type
144  * @len: Length of this fragment.
145  *	This field is zero when the queue slot is empty.
146  * @unmap_len: Length of this fragment to unmap
147  * @dma_offset: Offset of @dma_addr from the address of the backing DMA mapping.
148  * Only valid if @unmap_len != 0.
149  */
150 struct efx_tx_buffer {
151 	union {
152 		const struct sk_buff *skb;
153 		void *heap_buf;
154 	};
155 	union {
156 		efx_qword_t option;
157 		dma_addr_t dma_addr;
158 	};
159 	unsigned short flags;
160 	unsigned short len;
161 	unsigned short unmap_len;
162 	unsigned short dma_offset;
163 };
164 #define EFX_TX_BUF_CONT		1	/* not last descriptor of packet */
165 #define EFX_TX_BUF_SKB		2	/* buffer is last part of skb */
166 #define EFX_TX_BUF_HEAP		4	/* buffer was allocated with kmalloc() */
167 #define EFX_TX_BUF_MAP_SINGLE	8	/* buffer was mapped with dma_map_single() */
168 #define EFX_TX_BUF_OPTION	0x10	/* empty buffer for option descriptor */
169 
170 /**
171  * struct efx_tx_queue - An Efx TX queue
172  *
173  * This is a ring buffer of TX fragments.
174  * Since the TX completion path always executes on the same
175  * CPU and the xmit path can operate on different CPUs,
176  * performance is increased by ensuring that the completion
177  * path and the xmit path operate on different cache lines.
178  * This is particularly important if the xmit path is always
179  * executing on one CPU which is different from the completion
180  * path.  There is also a cache line for members which are
181  * read but not written on the fast path.
182  *
183  * @efx: The associated Efx NIC
184  * @queue: DMA queue number
185  * @channel: The associated channel
186  * @core_txq: The networking core TX queue structure
187  * @buffer: The software buffer ring
188  * @tsoh_page: Array of pages of TSO header buffers
189  * @txd: The hardware descriptor ring
190  * @ptr_mask: The size of the ring minus 1.
191  * @piobuf: PIO buffer region for this TX queue (shared with its partner).
192  *	Size of the region is efx_piobuf_size.
193  * @piobuf_offset: Buffer offset to be specified in PIO descriptors
194  * @initialised: Has hardware queue been initialised?
195  * @read_count: Current read pointer.
196  *	This is the number of buffers that have been removed from both rings.
197  * @old_write_count: The value of @write_count when last checked.
198  *	This is here for performance reasons.  The xmit path will
199  *	only get the up-to-date value of @write_count if this
200  *	variable indicates that the queue is empty.  This is to
201  *	avoid cache-line ping-pong between the xmit path and the
202  *	completion path.
203  * @merge_events: Number of TX merged completion events
204  * @insert_count: Current insert pointer
205  *	This is the number of buffers that have been added to the
206  *	software ring.
207  * @write_count: Current write pointer
208  *	This is the number of buffers that have been added to the
209  *	hardware ring.
210  * @old_read_count: The value of read_count when last checked.
211  *	This is here for performance reasons.  The xmit path will
212  *	only get the up-to-date value of read_count if this
213  *	variable indicates that the queue is full.  This is to
214  *	avoid cache-line ping-pong between the xmit path and the
215  *	completion path.
216  * @tso_bursts: Number of times TSO xmit invoked by kernel
217  * @tso_long_headers: Number of packets with headers too long for standard
218  *	blocks
219  * @tso_packets: Number of packets via the TSO xmit path
220  * @pushes: Number of times the TX push feature has been used
221  * @pio_packets: Number of times the TX PIO feature has been used
222  * @empty_read_count: If the completion path has seen the queue as empty
223  *	and the transmission path has not yet checked this, the value of
224  *	@read_count bitwise-added to %EFX_EMPTY_COUNT_VALID; otherwise 0.
225  */
226 struct efx_tx_queue {
227 	/* Members which don't change on the fast path */
228 	struct efx_nic *efx ____cacheline_aligned_in_smp;
229 	unsigned queue;
230 	struct efx_channel *channel;
231 	struct netdev_queue *core_txq;
232 	struct efx_tx_buffer *buffer;
233 	struct efx_buffer *tsoh_page;
234 	struct efx_special_buffer txd;
235 	unsigned int ptr_mask;
236 	void __iomem *piobuf;
237 	unsigned int piobuf_offset;
238 	bool initialised;
239 
240 	/* Members used mainly on the completion path */
241 	unsigned int read_count ____cacheline_aligned_in_smp;
242 	unsigned int old_write_count;
243 	unsigned int merge_events;
244 
245 	/* Members used only on the xmit path */
246 	unsigned int insert_count ____cacheline_aligned_in_smp;
247 	unsigned int write_count;
248 	unsigned int old_read_count;
249 	unsigned int tso_bursts;
250 	unsigned int tso_long_headers;
251 	unsigned int tso_packets;
252 	unsigned int pushes;
253 	unsigned int pio_packets;
254 	/* Statistics to supplement MAC stats */
255 	unsigned long tx_packets;
256 
257 	/* Members shared between paths and sometimes updated */
258 	unsigned int empty_read_count ____cacheline_aligned_in_smp;
259 #define EFX_EMPTY_COUNT_VALID 0x80000000
260 	atomic_t flush_outstanding;
261 };
262 
263 /**
264  * struct efx_rx_buffer - An Efx RX data buffer
265  * @dma_addr: DMA base address of the buffer
266  * @page: The associated page buffer.
267  *	Will be %NULL if the buffer slot is currently free.
268  * @page_offset: If pending: offset in @page of DMA base address.
269  *	If completed: offset in @page of Ethernet header.
270  * @len: If pending: length for DMA descriptor.
271  *	If completed: received length, excluding hash prefix.
272  * @flags: Flags for buffer and packet state.  These are only set on the
273  *	first buffer of a scattered packet.
274  */
275 struct efx_rx_buffer {
276 	dma_addr_t dma_addr;
277 	struct page *page;
278 	u16 page_offset;
279 	u16 len;
280 	u16 flags;
281 };
282 #define EFX_RX_BUF_LAST_IN_PAGE	0x0001
283 #define EFX_RX_PKT_CSUMMED	0x0002
284 #define EFX_RX_PKT_DISCARD	0x0004
285 #define EFX_RX_PKT_TCP		0x0040
286 #define EFX_RX_PKT_PREFIX_LEN	0x0080	/* length is in prefix only */
287 
288 /**
289  * struct efx_rx_page_state - Page-based rx buffer state
290  *
291  * Inserted at the start of every page allocated for receive buffers.
292  * Used to facilitate sharing dma mappings between recycled rx buffers
293  * and those passed up to the kernel.
294  *
295  * @dma_addr: The dma address of this page.
296  */
297 struct efx_rx_page_state {
298 	dma_addr_t dma_addr;
299 
300 	unsigned int __pad[0] ____cacheline_aligned;
301 };
302 
303 /**
304  * struct efx_rx_queue - An Efx RX queue
305  * @efx: The associated Efx NIC
306  * @core_index:  Index of network core RX queue.  Will be >= 0 iff this
307  *	is associated with a real RX queue.
308  * @buffer: The software buffer ring
309  * @rxd: The hardware descriptor ring
310  * @ptr_mask: The size of the ring minus 1.
311  * @refill_enabled: Enable refill whenever fill level is low
312  * @flush_pending: Set when a RX flush is pending. Has the same lifetime as
313  *	@rxq_flush_pending.
314  * @added_count: Number of buffers added to the receive queue.
315  * @notified_count: Number of buffers given to NIC (<= @added_count).
316  * @removed_count: Number of buffers removed from the receive queue.
317  * @scatter_n: Used by NIC specific receive code.
318  * @scatter_len: Used by NIC specific receive code.
319  * @page_ring: The ring to store DMA mapped pages for reuse.
320  * @page_add: Counter to calculate the write pointer for the recycle ring.
321  * @page_remove: Counter to calculate the read pointer for the recycle ring.
322  * @page_recycle_count: The number of pages that have been recycled.
323  * @page_recycle_failed: The number of pages that couldn't be recycled because
324  *      the kernel still held a reference to them.
325  * @page_recycle_full: The number of pages that were released because the
326  *      recycle ring was full.
327  * @page_ptr_mask: The number of pages in the RX recycle ring minus 1.
328  * @max_fill: RX descriptor maximum fill level (<= ring size)
329  * @fast_fill_trigger: RX descriptor fill level that will trigger a fast fill
330  *	(<= @max_fill)
331  * @min_fill: RX descriptor minimum non-zero fill level.
332  *	This records the minimum fill level observed when a ring
333  *	refill was triggered.
334  * @recycle_count: RX buffer recycle counter.
335  * @slow_fill: Timer used to defer efx_nic_generate_fill_event().
336  */
337 struct efx_rx_queue {
338 	struct efx_nic *efx;
339 	int core_index;
340 	struct efx_rx_buffer *buffer;
341 	struct efx_special_buffer rxd;
342 	unsigned int ptr_mask;
343 	bool refill_enabled;
344 	bool flush_pending;
345 
346 	unsigned int added_count;
347 	unsigned int notified_count;
348 	unsigned int removed_count;
349 	unsigned int scatter_n;
350 	unsigned int scatter_len;
351 	struct page **page_ring;
352 	unsigned int page_add;
353 	unsigned int page_remove;
354 	unsigned int page_recycle_count;
355 	unsigned int page_recycle_failed;
356 	unsigned int page_recycle_full;
357 	unsigned int page_ptr_mask;
358 	unsigned int max_fill;
359 	unsigned int fast_fill_trigger;
360 	unsigned int min_fill;
361 	unsigned int min_overfill;
362 	unsigned int recycle_count;
363 	struct timer_list slow_fill;
364 	unsigned int slow_fill_count;
365 	/* Statistics to supplement MAC stats */
366 	unsigned long rx_packets;
367 };
368 
369 enum efx_sync_events_state {
370 	SYNC_EVENTS_DISABLED = 0,
371 	SYNC_EVENTS_QUIESCENT,
372 	SYNC_EVENTS_REQUESTED,
373 	SYNC_EVENTS_VALID,
374 };
375 
376 /**
377  * struct efx_channel - An Efx channel
378  *
379  * A channel comprises an event queue, at least one TX queue, at least
380  * one RX queue, and an associated tasklet for processing the event
381  * queue.
382  *
383  * @efx: Associated Efx NIC
384  * @channel: Channel instance number
385  * @type: Channel type definition
386  * @eventq_init: Event queue initialised flag
387  * @enabled: Channel enabled indicator
388  * @irq: IRQ number (MSI and MSI-X only)
389  * @irq_moderation: IRQ moderation value (in hardware ticks)
390  * @napi_dev: Net device used with NAPI
391  * @napi_str: NAPI control structure
392  * @state: state for NAPI vs busy polling
393  * @state_lock: lock protecting @state
394  * @eventq: Event queue buffer
395  * @eventq_mask: Event queue pointer mask
396  * @eventq_read_ptr: Event queue read pointer
397  * @event_test_cpu: Last CPU to handle interrupt or test event for this channel
398  * @irq_count: Number of IRQs since last adaptive moderation decision
399  * @irq_mod_score: IRQ moderation score
400  * @n_rx_tobe_disc: Count of RX_TOBE_DISC errors
401  * @n_rx_ip_hdr_chksum_err: Count of RX IP header checksum errors
402  * @n_rx_tcp_udp_chksum_err: Count of RX TCP and UDP checksum errors
403  * @n_rx_mcast_mismatch: Count of unmatched multicast frames
404  * @n_rx_frm_trunc: Count of RX_FRM_TRUNC errors
405  * @n_rx_overlength: Count of RX_OVERLENGTH errors
406  * @n_skbuff_leaks: Count of skbuffs leaked due to RX overrun
407  * @n_rx_nodesc_trunc: Number of RX packets truncated and then dropped due to
408  *	lack of descriptors
409  * @n_rx_merge_events: Number of RX merged completion events
410  * @n_rx_merge_packets: Number of RX packets completed by merged events
411  * @rx_pkt_n_frags: Number of fragments in next packet to be delivered by
412  *	__efx_rx_packet(), or zero if there is none
413  * @rx_pkt_index: Ring index of first buffer for next packet to be delivered
414  *	by __efx_rx_packet(), if @rx_pkt_n_frags != 0
415  * @rx_queue: RX queue for this channel
416  * @tx_queue: TX queues for this channel
417  * @sync_events_state: Current state of sync events on this channel
418  * @sync_timestamp_major: Major part of the last ptp sync event
419  * @sync_timestamp_minor: Minor part of the last ptp sync event
420  */
421 struct efx_channel {
422 	struct efx_nic *efx;
423 	int channel;
424 	const struct efx_channel_type *type;
425 	bool eventq_init;
426 	bool enabled;
427 	int irq;
428 	unsigned int irq_moderation;
429 	struct net_device *napi_dev;
430 	struct napi_struct napi_str;
431 #ifdef CONFIG_NET_RX_BUSY_POLL
432 	unsigned int state;
433 	spinlock_t state_lock;
434 #define EFX_CHANNEL_STATE_IDLE		0
435 #define EFX_CHANNEL_STATE_NAPI		(1 << 0)  /* NAPI owns this channel */
436 #define EFX_CHANNEL_STATE_POLL		(1 << 1)  /* poll owns this channel */
437 #define EFX_CHANNEL_STATE_DISABLED	(1 << 2)  /* channel is disabled */
438 #define EFX_CHANNEL_STATE_NAPI_YIELD	(1 << 3)  /* NAPI yielded this channel */
439 #define EFX_CHANNEL_STATE_POLL_YIELD	(1 << 4)  /* poll yielded this channel */
440 #define EFX_CHANNEL_OWNED \
441 	(EFX_CHANNEL_STATE_NAPI | EFX_CHANNEL_STATE_POLL)
442 #define EFX_CHANNEL_LOCKED \
443 	(EFX_CHANNEL_OWNED | EFX_CHANNEL_STATE_DISABLED)
444 #define EFX_CHANNEL_USER_PEND \
445 	(EFX_CHANNEL_STATE_POLL | EFX_CHANNEL_STATE_POLL_YIELD)
446 #endif /* CONFIG_NET_RX_BUSY_POLL */
447 	struct efx_special_buffer eventq;
448 	unsigned int eventq_mask;
449 	unsigned int eventq_read_ptr;
450 	int event_test_cpu;
451 
452 	unsigned int irq_count;
453 	unsigned int irq_mod_score;
454 #ifdef CONFIG_RFS_ACCEL
455 	unsigned int rfs_filters_added;
456 #endif
457 
458 	unsigned n_rx_tobe_disc;
459 	unsigned n_rx_ip_hdr_chksum_err;
460 	unsigned n_rx_tcp_udp_chksum_err;
461 	unsigned n_rx_mcast_mismatch;
462 	unsigned n_rx_frm_trunc;
463 	unsigned n_rx_overlength;
464 	unsigned n_skbuff_leaks;
465 	unsigned int n_rx_nodesc_trunc;
466 	unsigned int n_rx_merge_events;
467 	unsigned int n_rx_merge_packets;
468 
469 	unsigned int rx_pkt_n_frags;
470 	unsigned int rx_pkt_index;
471 
472 	struct efx_rx_queue rx_queue;
473 	struct efx_tx_queue tx_queue[EFX_TXQ_TYPES];
474 
475 	enum efx_sync_events_state sync_events_state;
476 	u32 sync_timestamp_major;
477 	u32 sync_timestamp_minor;
478 };
479 
480 #ifdef CONFIG_NET_RX_BUSY_POLL
481 static inline void efx_channel_init_lock(struct efx_channel *channel)
482 {
483 	spin_lock_init(&channel->state_lock);
484 }
485 
486 /* Called from the device poll routine to get ownership of a channel. */
487 static inline bool efx_channel_lock_napi(struct efx_channel *channel)
488 {
489 	bool rc = true;
490 
491 	spin_lock_bh(&channel->state_lock);
492 	if (channel->state & EFX_CHANNEL_LOCKED) {
493 		WARN_ON(channel->state & EFX_CHANNEL_STATE_NAPI);
494 		channel->state |= EFX_CHANNEL_STATE_NAPI_YIELD;
495 		rc = false;
496 	} else {
497 		/* we don't care if someone yielded */
498 		channel->state = EFX_CHANNEL_STATE_NAPI;
499 	}
500 	spin_unlock_bh(&channel->state_lock);
501 	return rc;
502 }
503 
504 static inline void efx_channel_unlock_napi(struct efx_channel *channel)
505 {
506 	spin_lock_bh(&channel->state_lock);
507 	WARN_ON(channel->state &
508 		(EFX_CHANNEL_STATE_POLL | EFX_CHANNEL_STATE_NAPI_YIELD));
509 
510 	channel->state &= EFX_CHANNEL_STATE_DISABLED;
511 	spin_unlock_bh(&channel->state_lock);
512 }
513 
514 /* Called from efx_busy_poll(). */
515 static inline bool efx_channel_lock_poll(struct efx_channel *channel)
516 {
517 	bool rc = true;
518 
519 	spin_lock_bh(&channel->state_lock);
520 	if ((channel->state & EFX_CHANNEL_LOCKED)) {
521 		channel->state |= EFX_CHANNEL_STATE_POLL_YIELD;
522 		rc = false;
523 	} else {
524 		/* preserve yield marks */
525 		channel->state |= EFX_CHANNEL_STATE_POLL;
526 	}
527 	spin_unlock_bh(&channel->state_lock);
528 	return rc;
529 }
530 
531 /* Returns true if NAPI tried to get the channel while it was locked. */
532 static inline void efx_channel_unlock_poll(struct efx_channel *channel)
533 {
534 	spin_lock_bh(&channel->state_lock);
535 	WARN_ON(channel->state & EFX_CHANNEL_STATE_NAPI);
536 
537 	/* will reset state to idle, unless channel is disabled */
538 	channel->state &= EFX_CHANNEL_STATE_DISABLED;
539 	spin_unlock_bh(&channel->state_lock);
540 }
541 
542 /* True if a socket is polling, even if it did not get the lock. */
543 static inline bool efx_channel_busy_polling(struct efx_channel *channel)
544 {
545 	WARN_ON(!(channel->state & EFX_CHANNEL_OWNED));
546 	return channel->state & EFX_CHANNEL_USER_PEND;
547 }
548 
549 static inline void efx_channel_enable(struct efx_channel *channel)
550 {
551 	spin_lock_bh(&channel->state_lock);
552 	channel->state = EFX_CHANNEL_STATE_IDLE;
553 	spin_unlock_bh(&channel->state_lock);
554 }
555 
556 /* False if the channel is currently owned. */
557 static inline bool efx_channel_disable(struct efx_channel *channel)
558 {
559 	bool rc = true;
560 
561 	spin_lock_bh(&channel->state_lock);
562 	if (channel->state & EFX_CHANNEL_OWNED)
563 		rc = false;
564 	channel->state |= EFX_CHANNEL_STATE_DISABLED;
565 	spin_unlock_bh(&channel->state_lock);
566 
567 	return rc;
568 }
569 
570 #else /* CONFIG_NET_RX_BUSY_POLL */
571 
572 static inline void efx_channel_init_lock(struct efx_channel *channel)
573 {
574 }
575 
576 static inline bool efx_channel_lock_napi(struct efx_channel *channel)
577 {
578 	return true;
579 }
580 
581 static inline void efx_channel_unlock_napi(struct efx_channel *channel)
582 {
583 }
584 
585 static inline bool efx_channel_lock_poll(struct efx_channel *channel)
586 {
587 	return false;
588 }
589 
590 static inline void efx_channel_unlock_poll(struct efx_channel *channel)
591 {
592 }
593 
594 static inline bool efx_channel_busy_polling(struct efx_channel *channel)
595 {
596 	return false;
597 }
598 
599 static inline void efx_channel_enable(struct efx_channel *channel)
600 {
601 }
602 
603 static inline bool efx_channel_disable(struct efx_channel *channel)
604 {
605 	return true;
606 }
607 #endif /* CONFIG_NET_RX_BUSY_POLL */
608 
609 /**
610  * struct efx_msi_context - Context for each MSI
611  * @efx: The associated NIC
612  * @index: Index of the channel/IRQ
613  * @name: Name of the channel/IRQ
614  *
615  * Unlike &struct efx_channel, this is never reallocated and is always
616  * safe for the IRQ handler to access.
617  */
618 struct efx_msi_context {
619 	struct efx_nic *efx;
620 	unsigned int index;
621 	char name[IFNAMSIZ + 6];
622 };
623 
624 /**
625  * struct efx_channel_type - distinguishes traffic and extra channels
626  * @handle_no_channel: Handle failure to allocate an extra channel
627  * @pre_probe: Set up extra state prior to initialisation
628  * @post_remove: Tear down extra state after finalisation, if allocated.
629  *	May be called on channels that have not been probed.
630  * @get_name: Generate the channel's name (used for its IRQ handler)
631  * @copy: Copy the channel state prior to reallocation.  May be %NULL if
632  *	reallocation is not supported.
633  * @receive_skb: Handle an skb ready to be passed to netif_receive_skb()
634  * @keep_eventq: Flag for whether event queue should be kept initialised
635  *	while the device is stopped
636  */
637 struct efx_channel_type {
638 	void (*handle_no_channel)(struct efx_nic *);
639 	int (*pre_probe)(struct efx_channel *);
640 	void (*post_remove)(struct efx_channel *);
641 	void (*get_name)(struct efx_channel *, char *buf, size_t len);
642 	struct efx_channel *(*copy)(const struct efx_channel *);
643 	bool (*receive_skb)(struct efx_channel *, struct sk_buff *);
644 	bool keep_eventq;
645 };
646 
647 enum efx_led_mode {
648 	EFX_LED_OFF	= 0,
649 	EFX_LED_ON	= 1,
650 	EFX_LED_DEFAULT	= 2
651 };
652 
653 #define STRING_TABLE_LOOKUP(val, member) \
654 	((val) < member ## _max) ? member ## _names[val] : "(invalid)"
655 
656 extern const char *const efx_loopback_mode_names[];
657 extern const unsigned int efx_loopback_mode_max;
658 #define LOOPBACK_MODE(efx) \
659 	STRING_TABLE_LOOKUP((efx)->loopback_mode, efx_loopback_mode)
660 
661 extern const char *const efx_reset_type_names[];
662 extern const unsigned int efx_reset_type_max;
663 #define RESET_TYPE(type) \
664 	STRING_TABLE_LOOKUP(type, efx_reset_type)
665 
666 enum efx_int_mode {
667 	/* Be careful if altering to correct macro below */
668 	EFX_INT_MODE_MSIX = 0,
669 	EFX_INT_MODE_MSI = 1,
670 	EFX_INT_MODE_LEGACY = 2,
671 	EFX_INT_MODE_MAX	/* Insert any new items before this */
672 };
673 #define EFX_INT_MODE_USE_MSI(x) (((x)->interrupt_mode) <= EFX_INT_MODE_MSI)
674 
675 enum nic_state {
676 	STATE_UNINIT = 0,	/* device being probed/removed or is frozen */
677 	STATE_READY = 1,	/* hardware ready and netdev registered */
678 	STATE_DISABLED = 2,	/* device disabled due to hardware errors */
679 	STATE_RECOVERY = 3,	/* device recovering from PCI error */
680 };
681 
682 /* Forward declaration */
683 struct efx_nic;
684 
685 /* Pseudo bit-mask flow control field */
686 #define EFX_FC_RX	FLOW_CTRL_RX
687 #define EFX_FC_TX	FLOW_CTRL_TX
688 #define EFX_FC_AUTO	4
689 
690 /**
691  * struct efx_link_state - Current state of the link
692  * @up: Link is up
693  * @fd: Link is full-duplex
694  * @fc: Actual flow control flags
695  * @speed: Link speed (Mbps)
696  */
697 struct efx_link_state {
698 	bool up;
699 	bool fd;
700 	u8 fc;
701 	unsigned int speed;
702 };
703 
704 static inline bool efx_link_state_equal(const struct efx_link_state *left,
705 					const struct efx_link_state *right)
706 {
707 	return left->up == right->up && left->fd == right->fd &&
708 		left->fc == right->fc && left->speed == right->speed;
709 }
710 
711 /**
712  * struct efx_phy_operations - Efx PHY operations table
713  * @probe: Probe PHY and initialise efx->mdio.mode_support, efx->mdio.mmds,
714  *	efx->loopback_modes.
715  * @init: Initialise PHY
716  * @fini: Shut down PHY
717  * @reconfigure: Reconfigure PHY (e.g. for new link parameters)
718  * @poll: Update @link_state and report whether it changed.
719  *	Serialised by the mac_lock.
720  * @get_settings: Get ethtool settings. Serialised by the mac_lock.
721  * @set_settings: Set ethtool settings. Serialised by the mac_lock.
722  * @set_npage_adv: Set abilities advertised in (Extended) Next Page
723  *	(only needed where AN bit is set in mmds)
724  * @test_alive: Test that PHY is 'alive' (online)
725  * @test_name: Get the name of a PHY-specific test/result
726  * @run_tests: Run tests and record results as appropriate (offline).
727  *	Flags are the ethtool tests flags.
728  */
729 struct efx_phy_operations {
730 	int (*probe) (struct efx_nic *efx);
731 	int (*init) (struct efx_nic *efx);
732 	void (*fini) (struct efx_nic *efx);
733 	void (*remove) (struct efx_nic *efx);
734 	int (*reconfigure) (struct efx_nic *efx);
735 	bool (*poll) (struct efx_nic *efx);
736 	void (*get_settings) (struct efx_nic *efx,
737 			      struct ethtool_cmd *ecmd);
738 	int (*set_settings) (struct efx_nic *efx,
739 			     struct ethtool_cmd *ecmd);
740 	void (*set_npage_adv) (struct efx_nic *efx, u32);
741 	int (*test_alive) (struct efx_nic *efx);
742 	const char *(*test_name) (struct efx_nic *efx, unsigned int index);
743 	int (*run_tests) (struct efx_nic *efx, int *results, unsigned flags);
744 	int (*get_module_eeprom) (struct efx_nic *efx,
745 			       struct ethtool_eeprom *ee,
746 			       u8 *data);
747 	int (*get_module_info) (struct efx_nic *efx,
748 				struct ethtool_modinfo *modinfo);
749 };
750 
751 /**
752  * enum efx_phy_mode - PHY operating mode flags
753  * @PHY_MODE_NORMAL: on and should pass traffic
754  * @PHY_MODE_TX_DISABLED: on with TX disabled
755  * @PHY_MODE_LOW_POWER: set to low power through MDIO
756  * @PHY_MODE_OFF: switched off through external control
757  * @PHY_MODE_SPECIAL: on but will not pass traffic
758  */
759 enum efx_phy_mode {
760 	PHY_MODE_NORMAL		= 0,
761 	PHY_MODE_TX_DISABLED	= 1,
762 	PHY_MODE_LOW_POWER	= 2,
763 	PHY_MODE_OFF		= 4,
764 	PHY_MODE_SPECIAL	= 8,
765 };
766 
767 static inline bool efx_phy_mode_disabled(enum efx_phy_mode mode)
768 {
769 	return !!(mode & ~PHY_MODE_TX_DISABLED);
770 }
771 
772 /**
773  * struct efx_hw_stat_desc - Description of a hardware statistic
774  * @name: Name of the statistic as visible through ethtool, or %NULL if
775  *	it should not be exposed
776  * @dma_width: Width in bits (0 for non-DMA statistics)
777  * @offset: Offset within stats (ignored for non-DMA statistics)
778  */
779 struct efx_hw_stat_desc {
780 	const char *name;
781 	u16 dma_width;
782 	u16 offset;
783 };
784 
785 /* Number of bits used in a multicast filter hash address */
786 #define EFX_MCAST_HASH_BITS 8
787 
788 /* Number of (single-bit) entries in a multicast filter hash */
789 #define EFX_MCAST_HASH_ENTRIES (1 << EFX_MCAST_HASH_BITS)
790 
791 /* An Efx multicast filter hash */
792 union efx_multicast_hash {
793 	u8 byte[EFX_MCAST_HASH_ENTRIES / 8];
794 	efx_oword_t oword[EFX_MCAST_HASH_ENTRIES / sizeof(efx_oword_t) / 8];
795 };
796 
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_sem: Filter table rw_semaphore, for freeing the table
901  * @filter_lock: Filter table lock, for mere content changes
902  * @filter_state: Architecture-dependent filter table state
903  * @rps_flow_id: Flow IDs of filters allocated for accelerated RFS,
904  *	indexed by filter ID
905  * @rps_expire_index: Next index to check for expiry in @rps_flow_id
906  * @active_queues: Count of RX and TX queues that haven't been flushed and drained.
907  * @rxq_flush_pending: Count of number of receive queues that need to be flushed.
908  *	Decremented when the efx_flush_rx_queue() is called.
909  * @rxq_flush_outstanding: Count of number of RX flushes started but not yet
910  *	completed (either success or failure). Not used when MCDI is used to
911  *	flush receive queues.
912  * @flush_wq: wait queue used by efx_nic_flush_queues() to wait for flush completions.
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  * @ptp_data: PTP state data
917  * @vpd_sn: Serial number read from VPD
918  * @monitor_work: Hardware monitor workitem
919  * @biu_lock: BIU (bus interface unit) lock
920  * @last_irq_cpu: Last CPU to handle a possible test interrupt.  This
921  *	field is used by efx_test_interrupts() to verify that an
922  *	interrupt has occurred.
923  * @stats_lock: Statistics update lock. Must be held when calling
924  *	efx_nic_type::{update,start,stop}_stats.
925  * @n_rx_noskb_drops: Count of RX packets dropped due to failure to allocate an skb
926  *
927  * This is stored in the private area of the &struct net_device.
928  */
929 struct efx_nic {
930 	/* The following fields should be written very rarely */
931 
932 	char name[IFNAMSIZ];
933 	struct list_head node;
934 	struct efx_nic *primary;
935 	struct list_head secondary_list;
936 	struct pci_dev *pci_dev;
937 	unsigned int port_num;
938 	const struct efx_nic_type *type;
939 	int legacy_irq;
940 	bool eeh_disabled_legacy_irq;
941 	struct workqueue_struct *workqueue;
942 	char workqueue_name[16];
943 	struct work_struct reset_work;
944 	resource_size_t membase_phys;
945 	void __iomem *membase;
946 
947 	enum efx_int_mode interrupt_mode;
948 	unsigned int timer_quantum_ns;
949 	bool irq_rx_adaptive;
950 	unsigned int irq_rx_moderation;
951 	u32 msg_enable;
952 
953 	enum nic_state state;
954 	unsigned long reset_pending;
955 
956 	struct efx_channel *channel[EFX_MAX_CHANNELS];
957 	struct efx_msi_context msi_context[EFX_MAX_CHANNELS];
958 	const struct efx_channel_type *
959 	extra_channel_type[EFX_MAX_EXTRA_CHANNELS];
960 
961 	unsigned rxq_entries;
962 	unsigned txq_entries;
963 	unsigned int txq_stop_thresh;
964 	unsigned int txq_wake_thresh;
965 
966 	unsigned tx_dc_base;
967 	unsigned rx_dc_base;
968 	unsigned sram_lim_qw;
969 	unsigned next_buffer_table;
970 
971 	unsigned int max_channels;
972 	unsigned n_channels;
973 	unsigned n_rx_channels;
974 	unsigned rss_spread;
975 	unsigned tx_channel_offset;
976 	unsigned n_tx_channels;
977 	unsigned int rx_ip_align;
978 	unsigned int rx_dma_len;
979 	unsigned int rx_buffer_order;
980 	unsigned int rx_buffer_truesize;
981 	unsigned int rx_page_buf_step;
982 	unsigned int rx_bufs_per_page;
983 	unsigned int rx_pages_per_batch;
984 	unsigned int rx_prefix_size;
985 	int rx_packet_hash_offset;
986 	int rx_packet_len_offset;
987 	int rx_packet_ts_offset;
988 	u8 rx_hash_key[40];
989 	u32 rx_indir_table[128];
990 	bool rx_scatter;
991 
992 	unsigned int_error_count;
993 	unsigned long int_error_expire;
994 
995 	bool irq_soft_enabled;
996 	struct efx_buffer irq_status;
997 	unsigned irq_zero_count;
998 	unsigned irq_level;
999 	struct delayed_work selftest_work;
1000 
1001 #ifdef CONFIG_SFC_MTD
1002 	struct list_head mtd_list;
1003 #endif
1004 
1005 	void *nic_data;
1006 	struct efx_mcdi_data *mcdi;
1007 
1008 	struct mutex mac_lock;
1009 	struct work_struct mac_work;
1010 	bool port_enabled;
1011 
1012 	bool mc_bist_for_other_fn;
1013 	bool port_initialized;
1014 	struct net_device *net_dev;
1015 
1016 	struct efx_buffer stats_buffer;
1017 	u64 rx_nodesc_drops_total;
1018 	u64 rx_nodesc_drops_while_down;
1019 	bool rx_nodesc_drops_prev_state;
1020 
1021 	unsigned int phy_type;
1022 	const struct efx_phy_operations *phy_op;
1023 	void *phy_data;
1024 	struct mdio_if_info mdio;
1025 	unsigned int mdio_bus;
1026 	enum efx_phy_mode phy_mode;
1027 
1028 	u32 link_advertising;
1029 	struct efx_link_state link_state;
1030 	unsigned int n_link_state_changes;
1031 
1032 	bool unicast_filter;
1033 	union efx_multicast_hash multicast_hash;
1034 	u8 wanted_fc;
1035 	unsigned fc_disable;
1036 
1037 	atomic_t rx_reset;
1038 	enum efx_loopback_mode loopback_mode;
1039 	u64 loopback_modes;
1040 
1041 	void *loopback_selftest;
1042 
1043 	struct rw_semaphore filter_sem;
1044 	spinlock_t filter_lock;
1045 	void *filter_state;
1046 #ifdef CONFIG_RFS_ACCEL
1047 	u32 *rps_flow_id;
1048 	unsigned int rps_expire_index;
1049 #endif
1050 
1051 	atomic_t active_queues;
1052 	atomic_t rxq_flush_pending;
1053 	atomic_t rxq_flush_outstanding;
1054 	wait_queue_head_t flush_wq;
1055 
1056 #ifdef CONFIG_SFC_SRIOV
1057 	unsigned vf_count;
1058 	unsigned vf_init_count;
1059 	unsigned vi_scale;
1060 #endif
1061 
1062 	struct efx_ptp_data *ptp_data;
1063 
1064 	char *vpd_sn;
1065 
1066 	/* The following fields may be written more often */
1067 
1068 	struct delayed_work monitor_work ____cacheline_aligned_in_smp;
1069 	spinlock_t biu_lock;
1070 	int last_irq_cpu;
1071 	spinlock_t stats_lock;
1072 	atomic_t n_rx_noskb_drops;
1073 };
1074 
1075 static inline int efx_dev_registered(struct efx_nic *efx)
1076 {
1077 	return efx->net_dev->reg_state == NETREG_REGISTERED;
1078 }
1079 
1080 static inline unsigned int efx_port_num(struct efx_nic *efx)
1081 {
1082 	return efx->port_num;
1083 }
1084 
1085 struct efx_mtd_partition {
1086 	struct list_head node;
1087 	struct mtd_info mtd;
1088 	const char *dev_type_name;
1089 	const char *type_name;
1090 	char name[IFNAMSIZ + 20];
1091 };
1092 
1093 /**
1094  * struct efx_nic_type - Efx device type definition
1095  * @mem_bar: Get the memory BAR
1096  * @mem_map_size: Get memory BAR mapped size
1097  * @probe: Probe the controller
1098  * @remove: Free resources allocated by probe()
1099  * @init: Initialise the controller
1100  * @dimension_resources: Dimension controller resources (buffer table,
1101  *	and VIs once the available interrupt resources are clear)
1102  * @fini: Shut down the controller
1103  * @monitor: Periodic function for polling link state and hardware monitor
1104  * @map_reset_reason: Map ethtool reset reason to a reset method
1105  * @map_reset_flags: Map ethtool reset flags to a reset method, if possible
1106  * @reset: Reset the controller hardware and possibly the PHY.  This will
1107  *	be called while the controller is uninitialised.
1108  * @probe_port: Probe the MAC and PHY
1109  * @remove_port: Free resources allocated by probe_port()
1110  * @handle_global_event: Handle a "global" event (may be %NULL)
1111  * @fini_dmaq: Flush and finalise DMA queues (RX and TX queues)
1112  * @prepare_flush: Prepare the hardware for flushing the DMA queues
1113  *	(for Falcon architecture)
1114  * @finish_flush: Clean up after flushing the DMA queues (for Falcon
1115  *	architecture)
1116  * @prepare_flr: Prepare for an FLR
1117  * @finish_flr: Clean up after an FLR
1118  * @describe_stats: Describe statistics for ethtool
1119  * @update_stats: Update statistics not provided by event handling.
1120  *	Either argument may be %NULL.
1121  * @start_stats: Start the regular fetching of statistics
1122  * @pull_stats: Pull stats from the NIC and wait until they arrive.
1123  * @stop_stats: Stop the regular fetching of statistics
1124  * @set_id_led: Set state of identifying LED or revert to automatic function
1125  * @push_irq_moderation: Apply interrupt moderation value
1126  * @reconfigure_port: Push loopback/power/txdis changes to the MAC and PHY
1127  * @prepare_enable_fc_tx: Prepare MAC to enable pause frame TX (may be %NULL)
1128  * @reconfigure_mac: Push MAC address, MTU, flow control and filter settings
1129  *	to the hardware.  Serialised by the mac_lock.
1130  * @check_mac_fault: Check MAC fault state. True if fault present.
1131  * @get_wol: Get WoL configuration from driver state
1132  * @set_wol: Push WoL configuration to the NIC
1133  * @resume_wol: Synchronise WoL state between driver and MC (e.g. after resume)
1134  * @test_chip: Test registers.  May use efx_farch_test_registers(), and is
1135  *	expected to reset the NIC.
1136  * @test_nvram: Test validity of NVRAM contents
1137  * @mcdi_request: Send an MCDI request with the given header and SDU.
1138  *	The SDU length may be any value from 0 up to the protocol-
1139  *	defined maximum, but its buffer will be padded to a multiple
1140  *	of 4 bytes.
1141  * @mcdi_poll_response: Test whether an MCDI response is available.
1142  * @mcdi_read_response: Read the MCDI response PDU.  The offset will
1143  *	be a multiple of 4.  The length may not be, but the buffer
1144  *	will be padded so it is safe to round up.
1145  * @mcdi_poll_reboot: Test whether the MCDI has rebooted.  If so,
1146  *	return an appropriate error code for aborting any current
1147  *	request; otherwise return 0.
1148  * @irq_enable_master: Enable IRQs on the NIC.  Each event queue must
1149  *	be separately enabled after this.
1150  * @irq_test_generate: Generate a test IRQ
1151  * @irq_disable_non_ev: Disable non-event IRQs on the NIC.  Each event
1152  *	queue must be separately disabled before this.
1153  * @irq_handle_msi: Handle MSI for a channel.  The @dev_id argument is
1154  *	a pointer to the &struct efx_msi_context for the channel.
1155  * @irq_handle_legacy: Handle legacy interrupt.  The @dev_id argument
1156  *	is a pointer to the &struct efx_nic.
1157  * @tx_probe: Allocate resources for TX queue
1158  * @tx_init: Initialise TX queue on the NIC
1159  * @tx_remove: Free resources for TX queue
1160  * @tx_write: Write TX descriptors and doorbell
1161  * @rx_push_rss_config: Write RSS hash key and indirection table to the NIC
1162  * @rx_probe: Allocate resources for RX queue
1163  * @rx_init: Initialise RX queue on the NIC
1164  * @rx_remove: Free resources for RX queue
1165  * @rx_write: Write RX descriptors and doorbell
1166  * @rx_defer_refill: Generate a refill reminder event
1167  * @ev_probe: Allocate resources for event queue
1168  * @ev_init: Initialise event queue on the NIC
1169  * @ev_fini: Deinitialise event queue on the NIC
1170  * @ev_remove: Free resources for event queue
1171  * @ev_process: Process events for a queue, up to the given NAPI quota
1172  * @ev_read_ack: Acknowledge read events on a queue, rearming its IRQ
1173  * @ev_test_generate: Generate a test event
1174  * @filter_table_probe: Probe filter capabilities and set up filter software state
1175  * @filter_table_restore: Restore filters removed from hardware
1176  * @filter_table_remove: Remove filters from hardware and tear down software state
1177  * @filter_update_rx_scatter: Update filters after change to rx scatter setting
1178  * @filter_insert: add or replace a filter
1179  * @filter_remove_safe: remove a filter by ID, carefully
1180  * @filter_get_safe: retrieve a filter by ID, carefully
1181  * @filter_clear_rx: Remove all RX filters whose priority is less than or
1182  *	equal to the given priority and is not %EFX_FILTER_PRI_AUTO
1183  * @filter_count_rx_used: Get the number of filters in use at a given priority
1184  * @filter_get_rx_id_limit: Get maximum value of a filter id, plus 1
1185  * @filter_get_rx_ids: Get list of RX filters at a given priority
1186  * @filter_rfs_insert: Add or replace a filter for RFS.  This must be
1187  *	atomic.  The hardware change may be asynchronous but should
1188  *	not be delayed for long.  It may fail if this can't be done
1189  *	atomically.
1190  * @filter_rfs_expire_one: Consider expiring a filter inserted for RFS.
1191  *	This must check whether the specified table entry is used by RFS
1192  *	and that rps_may_expire_flow() returns true for it.
1193  * @mtd_probe: Probe and add MTD partitions associated with this net device,
1194  *	 using efx_mtd_add()
1195  * @mtd_rename: Set an MTD partition name using the net device name
1196  * @mtd_read: Read from an MTD partition
1197  * @mtd_erase: Erase part of an MTD partition
1198  * @mtd_write: Write to an MTD partition
1199  * @mtd_sync: Wait for write-back to complete on MTD partition.  This
1200  *	also notifies the driver that a writer has finished using this
1201  *	partition.
1202  * @ptp_write_host_time: Send host time to MC as part of sync protocol
1203  * @ptp_set_ts_sync_events: Enable or disable sync events for inline RX
1204  *	timestamping, possibly only temporarily for the purposes of a reset.
1205  * @ptp_set_ts_config: Set hardware timestamp configuration.  The flags
1206  *	and tx_type will already have been validated but this operation
1207  *	must validate and update rx_filter.
1208  * @set_mac_address: Set the MAC address of the device
1209  * @revision: Hardware architecture revision
1210  * @txd_ptr_tbl_base: TX descriptor ring base address
1211  * @rxd_ptr_tbl_base: RX descriptor ring base address
1212  * @buf_tbl_base: Buffer table base address
1213  * @evq_ptr_tbl_base: Event queue pointer table base address
1214  * @evq_rptr_tbl_base: Event queue read-pointer table base address
1215  * @max_dma_mask: Maximum possible DMA mask
1216  * @rx_prefix_size: Size of RX prefix before packet data
1217  * @rx_hash_offset: Offset of RX flow hash within prefix
1218  * @rx_ts_offset: Offset of timestamp within prefix
1219  * @rx_buffer_padding: Size of padding at end of RX packet
1220  * @can_rx_scatter: NIC is able to scatter packets to multiple buffers
1221  * @always_rx_scatter: NIC will always scatter packets to multiple buffers
1222  * @max_interrupt_mode: Highest capability interrupt mode supported
1223  *	from &enum efx_init_mode.
1224  * @timer_period_max: Maximum period of interrupt timer (in ticks)
1225  * @offload_features: net_device feature flags for protocol offload
1226  *	features implemented in hardware
1227  * @mcdi_max_ver: Maximum MCDI version supported
1228  * @hwtstamp_filters: Mask of hardware timestamp filter types supported
1229  */
1230 struct efx_nic_type {
1231 	bool is_vf;
1232 	unsigned int mem_bar;
1233 	unsigned int (*mem_map_size)(struct efx_nic *efx);
1234 	int (*probe)(struct efx_nic *efx);
1235 	void (*remove)(struct efx_nic *efx);
1236 	int (*init)(struct efx_nic *efx);
1237 	int (*dimension_resources)(struct efx_nic *efx);
1238 	void (*fini)(struct efx_nic *efx);
1239 	void (*monitor)(struct efx_nic *efx);
1240 	enum reset_type (*map_reset_reason)(enum reset_type reason);
1241 	int (*map_reset_flags)(u32 *flags);
1242 	int (*reset)(struct efx_nic *efx, enum reset_type method);
1243 	int (*probe_port)(struct efx_nic *efx);
1244 	void (*remove_port)(struct efx_nic *efx);
1245 	bool (*handle_global_event)(struct efx_channel *channel, efx_qword_t *);
1246 	int (*fini_dmaq)(struct efx_nic *efx);
1247 	void (*prepare_flush)(struct efx_nic *efx);
1248 	void (*finish_flush)(struct efx_nic *efx);
1249 	void (*prepare_flr)(struct efx_nic *efx);
1250 	void (*finish_flr)(struct efx_nic *efx);
1251 	size_t (*describe_stats)(struct efx_nic *efx, u8 *names);
1252 	size_t (*update_stats)(struct efx_nic *efx, u64 *full_stats,
1253 			       struct rtnl_link_stats64 *core_stats);
1254 	void (*start_stats)(struct efx_nic *efx);
1255 	void (*pull_stats)(struct efx_nic *efx);
1256 	void (*stop_stats)(struct efx_nic *efx);
1257 	void (*set_id_led)(struct efx_nic *efx, enum efx_led_mode mode);
1258 	void (*push_irq_moderation)(struct efx_channel *channel);
1259 	int (*reconfigure_port)(struct efx_nic *efx);
1260 	void (*prepare_enable_fc_tx)(struct efx_nic *efx);
1261 	int (*reconfigure_mac)(struct efx_nic *efx);
1262 	bool (*check_mac_fault)(struct efx_nic *efx);
1263 	void (*get_wol)(struct efx_nic *efx, struct ethtool_wolinfo *wol);
1264 	int (*set_wol)(struct efx_nic *efx, u32 type);
1265 	void (*resume_wol)(struct efx_nic *efx);
1266 	int (*test_chip)(struct efx_nic *efx, struct efx_self_tests *tests);
1267 	int (*test_nvram)(struct efx_nic *efx);
1268 	void (*mcdi_request)(struct efx_nic *efx,
1269 			     const efx_dword_t *hdr, size_t hdr_len,
1270 			     const efx_dword_t *sdu, size_t sdu_len);
1271 	bool (*mcdi_poll_response)(struct efx_nic *efx);
1272 	void (*mcdi_read_response)(struct efx_nic *efx, efx_dword_t *pdu,
1273 				   size_t pdu_offset, size_t pdu_len);
1274 	int (*mcdi_poll_reboot)(struct efx_nic *efx);
1275 	void (*irq_enable_master)(struct efx_nic *efx);
1276 	void (*irq_test_generate)(struct efx_nic *efx);
1277 	void (*irq_disable_non_ev)(struct efx_nic *efx);
1278 	irqreturn_t (*irq_handle_msi)(int irq, void *dev_id);
1279 	irqreturn_t (*irq_handle_legacy)(int irq, void *dev_id);
1280 	int (*tx_probe)(struct efx_tx_queue *tx_queue);
1281 	void (*tx_init)(struct efx_tx_queue *tx_queue);
1282 	void (*tx_remove)(struct efx_tx_queue *tx_queue);
1283 	void (*tx_write)(struct efx_tx_queue *tx_queue);
1284 	int (*rx_push_rss_config)(struct efx_nic *efx, bool user,
1285 				  const u32 *rx_indir_table);
1286 	int (*rx_probe)(struct efx_rx_queue *rx_queue);
1287 	void (*rx_init)(struct efx_rx_queue *rx_queue);
1288 	void (*rx_remove)(struct efx_rx_queue *rx_queue);
1289 	void (*rx_write)(struct efx_rx_queue *rx_queue);
1290 	void (*rx_defer_refill)(struct efx_rx_queue *rx_queue);
1291 	int (*ev_probe)(struct efx_channel *channel);
1292 	int (*ev_init)(struct efx_channel *channel);
1293 	void (*ev_fini)(struct efx_channel *channel);
1294 	void (*ev_remove)(struct efx_channel *channel);
1295 	int (*ev_process)(struct efx_channel *channel, int quota);
1296 	void (*ev_read_ack)(struct efx_channel *channel);
1297 	void (*ev_test_generate)(struct efx_channel *channel);
1298 	int (*filter_table_probe)(struct efx_nic *efx);
1299 	void (*filter_table_restore)(struct efx_nic *efx);
1300 	void (*filter_table_remove)(struct efx_nic *efx);
1301 	void (*filter_update_rx_scatter)(struct efx_nic *efx);
1302 	s32 (*filter_insert)(struct efx_nic *efx,
1303 			     struct efx_filter_spec *spec, bool replace);
1304 	int (*filter_remove_safe)(struct efx_nic *efx,
1305 				  enum efx_filter_priority priority,
1306 				  u32 filter_id);
1307 	int (*filter_get_safe)(struct efx_nic *efx,
1308 			       enum efx_filter_priority priority,
1309 			       u32 filter_id, struct efx_filter_spec *);
1310 	int (*filter_clear_rx)(struct efx_nic *efx,
1311 			       enum efx_filter_priority priority);
1312 	u32 (*filter_count_rx_used)(struct efx_nic *efx,
1313 				    enum efx_filter_priority priority);
1314 	u32 (*filter_get_rx_id_limit)(struct efx_nic *efx);
1315 	s32 (*filter_get_rx_ids)(struct efx_nic *efx,
1316 				 enum efx_filter_priority priority,
1317 				 u32 *buf, u32 size);
1318 #ifdef CONFIG_RFS_ACCEL
1319 	s32 (*filter_rfs_insert)(struct efx_nic *efx,
1320 				 struct efx_filter_spec *spec);
1321 	bool (*filter_rfs_expire_one)(struct efx_nic *efx, u32 flow_id,
1322 				      unsigned int index);
1323 #endif
1324 #ifdef CONFIG_SFC_MTD
1325 	int (*mtd_probe)(struct efx_nic *efx);
1326 	void (*mtd_rename)(struct efx_mtd_partition *part);
1327 	int (*mtd_read)(struct mtd_info *mtd, loff_t start, size_t len,
1328 			size_t *retlen, u8 *buffer);
1329 	int (*mtd_erase)(struct mtd_info *mtd, loff_t start, size_t len);
1330 	int (*mtd_write)(struct mtd_info *mtd, loff_t start, size_t len,
1331 			 size_t *retlen, const u8 *buffer);
1332 	int (*mtd_sync)(struct mtd_info *mtd);
1333 #endif
1334 	void (*ptp_write_host_time)(struct efx_nic *efx, u32 host_time);
1335 	int (*ptp_set_ts_sync_events)(struct efx_nic *efx, bool en, bool temp);
1336 	int (*ptp_set_ts_config)(struct efx_nic *efx,
1337 				 struct hwtstamp_config *init);
1338 	int (*sriov_configure)(struct efx_nic *efx, int num_vfs);
1339 	int (*sriov_init)(struct efx_nic *efx);
1340 	void (*sriov_fini)(struct efx_nic *efx);
1341 	bool (*sriov_wanted)(struct efx_nic *efx);
1342 	void (*sriov_reset)(struct efx_nic *efx);
1343 	void (*sriov_flr)(struct efx_nic *efx, unsigned vf_i);
1344 	int (*sriov_set_vf_mac)(struct efx_nic *efx, int vf_i, u8 *mac);
1345 	int (*sriov_set_vf_vlan)(struct efx_nic *efx, int vf_i, u16 vlan,
1346 				 u8 qos);
1347 	int (*sriov_set_vf_spoofchk)(struct efx_nic *efx, int vf_i,
1348 				     bool spoofchk);
1349 	int (*sriov_get_vf_config)(struct efx_nic *efx, int vf_i,
1350 				   struct ifla_vf_info *ivi);
1351 	int (*sriov_set_vf_link_state)(struct efx_nic *efx, int vf_i,
1352 				       int link_state);
1353 	int (*sriov_get_phys_port_id)(struct efx_nic *efx,
1354 				      struct netdev_phys_item_id *ppid);
1355 	int (*vswitching_probe)(struct efx_nic *efx);
1356 	int (*vswitching_restore)(struct efx_nic *efx);
1357 	void (*vswitching_remove)(struct efx_nic *efx);
1358 	int (*get_mac_address)(struct efx_nic *efx, unsigned char *perm_addr);
1359 	int (*set_mac_address)(struct efx_nic *efx);
1360 
1361 	int revision;
1362 	unsigned int txd_ptr_tbl_base;
1363 	unsigned int rxd_ptr_tbl_base;
1364 	unsigned int buf_tbl_base;
1365 	unsigned int evq_ptr_tbl_base;
1366 	unsigned int evq_rptr_tbl_base;
1367 	u64 max_dma_mask;
1368 	unsigned int rx_prefix_size;
1369 	unsigned int rx_hash_offset;
1370 	unsigned int rx_ts_offset;
1371 	unsigned int rx_buffer_padding;
1372 	bool can_rx_scatter;
1373 	bool always_rx_scatter;
1374 	unsigned int max_interrupt_mode;
1375 	unsigned int timer_period_max;
1376 	netdev_features_t offload_features;
1377 	int mcdi_max_ver;
1378 	unsigned int max_rx_ip_filters;
1379 	u32 hwtstamp_filters;
1380 };
1381 
1382 /**************************************************************************
1383  *
1384  * Prototypes and inline functions
1385  *
1386  *************************************************************************/
1387 
1388 static inline struct efx_channel *
1389 efx_get_channel(struct efx_nic *efx, unsigned index)
1390 {
1391 	EFX_BUG_ON_PARANOID(index >= efx->n_channels);
1392 	return efx->channel[index];
1393 }
1394 
1395 /* Iterate over all used channels */
1396 #define efx_for_each_channel(_channel, _efx)				\
1397 	for (_channel = (_efx)->channel[0];				\
1398 	     _channel;							\
1399 	     _channel = (_channel->channel + 1 < (_efx)->n_channels) ?	\
1400 		     (_efx)->channel[_channel->channel + 1] : NULL)
1401 
1402 /* Iterate over all used channels in reverse */
1403 #define efx_for_each_channel_rev(_channel, _efx)			\
1404 	for (_channel = (_efx)->channel[(_efx)->n_channels - 1];	\
1405 	     _channel;							\
1406 	     _channel = _channel->channel ?				\
1407 		     (_efx)->channel[_channel->channel - 1] : NULL)
1408 
1409 static inline struct efx_tx_queue *
1410 efx_get_tx_queue(struct efx_nic *efx, unsigned index, unsigned type)
1411 {
1412 	EFX_BUG_ON_PARANOID(index >= efx->n_tx_channels ||
1413 			    type >= EFX_TXQ_TYPES);
1414 	return &efx->channel[efx->tx_channel_offset + index]->tx_queue[type];
1415 }
1416 
1417 static inline bool efx_channel_has_tx_queues(struct efx_channel *channel)
1418 {
1419 	return channel->channel - channel->efx->tx_channel_offset <
1420 		channel->efx->n_tx_channels;
1421 }
1422 
1423 static inline struct efx_tx_queue *
1424 efx_channel_get_tx_queue(struct efx_channel *channel, unsigned type)
1425 {
1426 	EFX_BUG_ON_PARANOID(!efx_channel_has_tx_queues(channel) ||
1427 			    type >= EFX_TXQ_TYPES);
1428 	return &channel->tx_queue[type];
1429 }
1430 
1431 static inline bool efx_tx_queue_used(struct efx_tx_queue *tx_queue)
1432 {
1433 	return !(tx_queue->efx->net_dev->num_tc < 2 &&
1434 		 tx_queue->queue & EFX_TXQ_TYPE_HIGHPRI);
1435 }
1436 
1437 /* Iterate over all TX queues belonging to a channel */
1438 #define efx_for_each_channel_tx_queue(_tx_queue, _channel)		\
1439 	if (!efx_channel_has_tx_queues(_channel))			\
1440 		;							\
1441 	else								\
1442 		for (_tx_queue = (_channel)->tx_queue;			\
1443 		     _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES && \
1444 			     efx_tx_queue_used(_tx_queue);		\
1445 		     _tx_queue++)
1446 
1447 /* Iterate over all possible TX queues belonging to a channel */
1448 #define efx_for_each_possible_channel_tx_queue(_tx_queue, _channel)	\
1449 	if (!efx_channel_has_tx_queues(_channel))			\
1450 		;							\
1451 	else								\
1452 		for (_tx_queue = (_channel)->tx_queue;			\
1453 		     _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES;	\
1454 		     _tx_queue++)
1455 
1456 static inline bool efx_channel_has_rx_queue(struct efx_channel *channel)
1457 {
1458 	return channel->rx_queue.core_index >= 0;
1459 }
1460 
1461 static inline struct efx_rx_queue *
1462 efx_channel_get_rx_queue(struct efx_channel *channel)
1463 {
1464 	EFX_BUG_ON_PARANOID(!efx_channel_has_rx_queue(channel));
1465 	return &channel->rx_queue;
1466 }
1467 
1468 /* Iterate over all RX queues belonging to a channel */
1469 #define efx_for_each_channel_rx_queue(_rx_queue, _channel)		\
1470 	if (!efx_channel_has_rx_queue(_channel))			\
1471 		;							\
1472 	else								\
1473 		for (_rx_queue = &(_channel)->rx_queue;			\
1474 		     _rx_queue;						\
1475 		     _rx_queue = NULL)
1476 
1477 static inline struct efx_channel *
1478 efx_rx_queue_channel(struct efx_rx_queue *rx_queue)
1479 {
1480 	return container_of(rx_queue, struct efx_channel, rx_queue);
1481 }
1482 
1483 static inline int efx_rx_queue_index(struct efx_rx_queue *rx_queue)
1484 {
1485 	return efx_rx_queue_channel(rx_queue)->channel;
1486 }
1487 
1488 /* Returns a pointer to the specified receive buffer in the RX
1489  * descriptor queue.
1490  */
1491 static inline struct efx_rx_buffer *efx_rx_buffer(struct efx_rx_queue *rx_queue,
1492 						  unsigned int index)
1493 {
1494 	return &rx_queue->buffer[index];
1495 }
1496 
1497 /**
1498  * EFX_MAX_FRAME_LEN - calculate maximum frame length
1499  *
1500  * This calculates the maximum frame length that will be used for a
1501  * given MTU.  The frame length will be equal to the MTU plus a
1502  * constant amount of header space and padding.  This is the quantity
1503  * that the net driver will program into the MAC as the maximum frame
1504  * length.
1505  *
1506  * The 10G MAC requires 8-byte alignment on the frame
1507  * length, so we round up to the nearest 8.
1508  *
1509  * Re-clocking by the XGXS on RX can reduce an IPG to 32 bits (half an
1510  * XGMII cycle).  If the frame length reaches the maximum value in the
1511  * same cycle, the XMAC can miss the IPG altogether.  We work around
1512  * this by adding a further 16 bytes.
1513  */
1514 #define EFX_MAX_FRAME_LEN(mtu) \
1515 	((((mtu) + ETH_HLEN + VLAN_HLEN + 4/* FCS */ + 7) & ~7) + 16)
1516 
1517 static inline bool efx_xmit_with_hwtstamp(struct sk_buff *skb)
1518 {
1519 	return skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP;
1520 }
1521 static inline void efx_xmit_hwtstamp_pending(struct sk_buff *skb)
1522 {
1523 	skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
1524 }
1525 
1526 #endif /* EFX_NET_DRIVER_H */
1527