xref: /linux/drivers/net/ethernet/sfc/net_driver.h (revision 7f71507851fc7764b36a3221839607d3a45c2025)
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /****************************************************************************
3  * Driver for Solarflare network controllers and boards
4  * Copyright 2005-2006 Fen Systems Ltd.
5  * Copyright 2005-2013 Solarflare Communications Inc.
6  */
7 
8 /* Common definitions for all Efx net driver code */
9 
10 #ifndef EFX_NET_DRIVER_H
11 #define EFX_NET_DRIVER_H
12 
13 #include <linux/netdevice.h>
14 #include <linux/etherdevice.h>
15 #include <linux/ethtool.h>
16 #include <linux/if_vlan.h>
17 #include <linux/timer.h>
18 #include <linux/mdio.h>
19 #include <linux/list.h>
20 #include <linux/pci.h>
21 #include <linux/device.h>
22 #include <linux/highmem.h>
23 #include <linux/workqueue.h>
24 #include <linux/mutex.h>
25 #include <linux/rwsem.h>
26 #include <linux/vmalloc.h>
27 #include <linux/mtd/mtd.h>
28 #include <net/busy_poll.h>
29 #include <net/xdp.h>
30 #include <net/netevent.h>
31 
32 #include "enum.h"
33 #include "bitfield.h"
34 #include "filter.h"
35 
36 /**************************************************************************
37  *
38  * Build definitions
39  *
40  **************************************************************************/
41 
42 #ifdef DEBUG
43 #define EFX_WARN_ON_ONCE_PARANOID(x) WARN_ON_ONCE(x)
44 #define EFX_WARN_ON_PARANOID(x) WARN_ON(x)
45 #else
46 #define EFX_WARN_ON_ONCE_PARANOID(x) do {} while (0)
47 #define EFX_WARN_ON_PARANOID(x) do {} while (0)
48 #endif
49 
50 /**************************************************************************
51  *
52  * Efx data structures
53  *
54  **************************************************************************/
55 
56 #define EFX_MAX_CHANNELS 32U
57 #define EFX_MAX_RX_QUEUES EFX_MAX_CHANNELS
58 #define EFX_EXTRA_CHANNEL_IOV	0
59 #define EFX_EXTRA_CHANNEL_PTP	1
60 #define EFX_EXTRA_CHANNEL_TC	2
61 #define EFX_MAX_EXTRA_CHANNELS	3U
62 
63 /* Checksum generation is a per-queue option in hardware, so each
64  * queue visible to the networking core is backed by two hardware TX
65  * queues. */
66 #define EFX_MAX_TX_TC		2
67 #define EFX_MAX_CORE_TX_QUEUES	(EFX_MAX_TX_TC * EFX_MAX_CHANNELS)
68 #define EFX_TXQ_TYPE_OUTER_CSUM	1	/* Outer checksum offload */
69 #define EFX_TXQ_TYPE_INNER_CSUM	2	/* Inner checksum offload */
70 #define EFX_TXQ_TYPES		4
71 #define EFX_MAX_TXQ_PER_CHANNEL	4
72 #define EFX_MAX_TX_QUEUES	(EFX_MAX_TXQ_PER_CHANNEL * EFX_MAX_CHANNELS)
73 
74 /* Maximum possible MTU the driver supports */
75 #define EFX_MAX_MTU (9 * 1024)
76 
77 /* Minimum MTU, from RFC791 (IP) */
78 #define EFX_MIN_MTU 68
79 
80 /* Maximum total header length for TSOv2 */
81 #define EFX_TSO2_MAX_HDRLEN	208
82 
83 /* Size of an RX scatter buffer.  Small enough to pack 2 into a 4K page,
84  * and should be a multiple of the cache line size.
85  */
86 #define EFX_RX_USR_BUF_SIZE	(2048 - 256)
87 
88 /* If possible, we should ensure cache line alignment at start and end
89  * of every buffer.  Otherwise, we just need to ensure 4-byte
90  * alignment of the network header.
91  */
92 #if NET_IP_ALIGN == 0
93 #define EFX_RX_BUF_ALIGNMENT	L1_CACHE_BYTES
94 #else
95 #define EFX_RX_BUF_ALIGNMENT	4
96 #endif
97 
98 /* Non-standard XDP_PACKET_HEADROOM and tailroom to satisfy XDP_REDIRECT and
99  * still fit two standard MTU size packets into a single 4K page.
100  */
101 #define EFX_XDP_HEADROOM	128
102 #define EFX_XDP_TAILROOM	SKB_DATA_ALIGN(sizeof(struct skb_shared_info))
103 
104 /* Forward declare Precision Time Protocol (PTP) support structure. */
105 struct efx_ptp_data;
106 struct hwtstamp_config;
107 
108 struct efx_self_tests;
109 
110 /**
111  * struct efx_buffer - A general-purpose DMA buffer
112  * @addr: host base address of the buffer
113  * @dma_addr: DMA base address of the buffer
114  * @len: Buffer length, in bytes
115  *
116  * The NIC uses these buffers for its interrupt status registers and
117  * MAC stats dumps.
118  */
119 struct efx_buffer {
120 	void *addr;
121 	dma_addr_t dma_addr;
122 	unsigned int len;
123 };
124 
125 /**
126  * struct efx_tx_buffer - buffer state for a TX descriptor
127  * @skb: When @flags & %EFX_TX_BUF_SKB, the associated socket buffer to be
128  *	freed when descriptor completes
129  * @xdpf: When @flags & %EFX_TX_BUF_XDP, the XDP frame information; its @data
130  *	member is the associated buffer to drop a page reference on.
131  * @option: When @flags & %EFX_TX_BUF_OPTION, an EF10-specific option
132  *	descriptor.
133  * @dma_addr: DMA address of the fragment.
134  * @flags: Flags for allocation and DMA mapping type
135  * @len: Length of this fragment.
136  *	This field is zero when the queue slot is empty.
137  * @unmap_len: Length of this fragment to unmap
138  * @dma_offset: Offset of @dma_addr from the address of the backing DMA mapping.
139  * Only valid if @unmap_len != 0.
140  */
141 struct efx_tx_buffer {
142 	union {
143 		const struct sk_buff *skb;
144 		struct xdp_frame *xdpf;
145 	};
146 	union {
147 		efx_qword_t option;    /* EF10 */
148 		dma_addr_t dma_addr;
149 	};
150 	unsigned short flags;
151 	unsigned short len;
152 	unsigned short unmap_len;
153 	unsigned short dma_offset;
154 };
155 #define EFX_TX_BUF_CONT		1	/* not last descriptor of packet */
156 #define EFX_TX_BUF_SKB		2	/* buffer is last part of skb */
157 #define EFX_TX_BUF_MAP_SINGLE	8	/* buffer was mapped with dma_map_single() */
158 #define EFX_TX_BUF_OPTION	0x10	/* empty buffer for option descriptor */
159 #define EFX_TX_BUF_XDP		0x20	/* buffer was sent with XDP */
160 #define EFX_TX_BUF_TSO_V3	0x40	/* empty buffer for a TSO_V3 descriptor */
161 #define EFX_TX_BUF_EFV		0x100	/* buffer was sent from representor */
162 
163 /**
164  * struct efx_tx_queue - An Efx TX queue
165  *
166  * This is a ring buffer of TX fragments.
167  * Since the TX completion path always executes on the same
168  * CPU and the xmit path can operate on different CPUs,
169  * performance is increased by ensuring that the completion
170  * path and the xmit path operate on different cache lines.
171  * This is particularly important if the xmit path is always
172  * executing on one CPU which is different from the completion
173  * path.  There is also a cache line for members which are
174  * read but not written on the fast path.
175  *
176  * @efx: The associated Efx NIC
177  * @queue: DMA queue number
178  * @label: Label for TX completion events.
179  *	Is our index within @channel->tx_queue array.
180  * @type: configuration type of this TX queue.  A bitmask of %EFX_TXQ_TYPE_* flags.
181  * @tso_version: Version of TSO in use for this queue.
182  * @tso_encap: Is encapsulated TSO supported? Supported in TSOv2 on 8000 series.
183  * @channel: The associated channel
184  * @core_txq: The networking core TX queue structure
185  * @buffer: The software buffer ring
186  * @cb_page: Array of pages of copy buffers.  Carved up according to
187  *	%EFX_TX_CB_ORDER into %EFX_TX_CB_SIZE-sized chunks.
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  * @timestamping: Is timestamping enabled for this channel?
195  * @xdp_tx: Is this an XDP tx queue?
196  * @old_complete_packets: Value of @complete_packets as of last
197  *	efx_init_tx_queue()
198  * @old_complete_bytes: Value of @complete_bytes as of last
199  *	efx_init_tx_queue()
200  * @old_tso_bursts: Value of @tso_bursts as of last efx_init_tx_queue()
201  * @old_tso_packets: Value of @tso_packets as of last efx_init_tx_queue()
202  * @read_count: Current read pointer.
203  *	This is the number of buffers that have been removed from both rings.
204  * @old_write_count: The value of @write_count when last checked.
205  *	This is here for performance reasons.  The xmit path will
206  *	only get the up-to-date value of @write_count if this
207  *	variable indicates that the queue is empty.  This is to
208  *	avoid cache-line ping-pong between the xmit path and the
209  *	completion path.
210  * @merge_events: Number of TX merged completion events
211  * @bytes_compl: Number of bytes completed during this NAPI poll
212  *	(efx_process_channel()).  For BQL.
213  * @pkts_compl: Number of packets completed during this NAPI poll.
214  * @complete_packets: Number of packets completed since this struct was
215  *	created.  Only counts SKB packets, not XDP TX (it accumulates
216  *	the same values that are reported to BQL).
217  * @complete_bytes: Number of bytes completed since this struct was
218  *	created.  For TSO, counts the superframe size, not the sizes of
219  *	generated frames on the wire (i.e. the headers are only counted
220  *	once)
221  * @complete_xdp_packets: Number of XDP TX packets completed since this
222  *	struct was created.
223  * @complete_xdp_bytes: Number of XDP TX bytes completed since this
224  *	struct was created.
225  * @completed_timestamp_major: Top part of the most recent tx timestamp.
226  * @completed_timestamp_minor: Low part of the most recent tx timestamp.
227  * @insert_count: Current insert pointer
228  *	This is the number of buffers that have been added to the
229  *	software ring.
230  * @write_count: Current write pointer
231  *	This is the number of buffers that have been added to the
232  *	hardware ring.
233  * @packet_write_count: Completable write pointer
234  *	This is the write pointer of the last packet written.
235  *	Normally this will equal @write_count, but as option descriptors
236  *	don't produce completion events, they won't update this.
237  *	Filled in iff @efx->type->option_descriptors; only used for PIO.
238  *	Thus, this is only written and used on EF10.
239  * @old_read_count: The value of read_count when last checked.
240  *	This is here for performance reasons.  The xmit path will
241  *	only get the up-to-date value of read_count if this
242  *	variable indicates that the queue is full.  This is to
243  *	avoid cache-line ping-pong between the xmit path and the
244  *	completion path.
245  * @tso_bursts: Number of times TSO xmit invoked by kernel
246  * @tso_long_headers: Number of packets with headers too long for standard
247  *	blocks
248  * @tso_packets: Number of packets via the TSO xmit path
249  * @tso_fallbacks: Number of times TSO fallback used
250  * @pushes: Number of times the TX push feature has been used
251  * @pio_packets: Number of times the TX PIO feature has been used
252  * @xmit_pending: Are any packets waiting to be pushed to the NIC
253  * @cb_packets: Number of times the TX copybreak feature has been used
254  * @notify_count: Count of notified descriptors to the NIC
255  * @tx_packets: Number of packets sent since this struct was created
256  * @empty_read_count: If the completion path has seen the queue as empty
257  *	and the transmission path has not yet checked this, the value of
258  *	@read_count bitwise-added to %EFX_EMPTY_COUNT_VALID; otherwise 0.
259  */
260 struct efx_tx_queue {
261 	/* Members which don't change on the fast path */
262 	struct efx_nic *efx ____cacheline_aligned_in_smp;
263 	unsigned int queue;
264 	unsigned int label;
265 	unsigned int type;
266 	unsigned int tso_version;
267 	bool tso_encap;
268 	struct efx_channel *channel;
269 	struct netdev_queue *core_txq;
270 	struct efx_tx_buffer *buffer;
271 	struct efx_buffer *cb_page;
272 	struct efx_buffer txd;
273 	unsigned int ptr_mask;
274 	void __iomem *piobuf;
275 	unsigned int piobuf_offset;
276 	bool initialised;
277 	bool timestamping;
278 	bool xdp_tx;
279 	unsigned long old_complete_packets;
280 	unsigned long old_complete_bytes;
281 	unsigned int old_tso_bursts;
282 	unsigned int old_tso_packets;
283 
284 	/* Members used mainly on the completion path */
285 	unsigned int read_count ____cacheline_aligned_in_smp;
286 	unsigned int old_write_count;
287 	unsigned int merge_events;
288 	unsigned int bytes_compl;
289 	unsigned int pkts_compl;
290 	unsigned long complete_packets;
291 	unsigned long complete_bytes;
292 	unsigned long complete_xdp_packets;
293 	unsigned long complete_xdp_bytes;
294 	u32 completed_timestamp_major;
295 	u32 completed_timestamp_minor;
296 
297 	/* Members used only on the xmit path */
298 	unsigned int insert_count ____cacheline_aligned_in_smp;
299 	unsigned int write_count;
300 	unsigned int packet_write_count;
301 	unsigned int old_read_count;
302 	unsigned int tso_bursts;
303 	unsigned int tso_long_headers;
304 	unsigned int tso_packets;
305 	unsigned int tso_fallbacks;
306 	unsigned int pushes;
307 	unsigned int pio_packets;
308 	bool xmit_pending;
309 	unsigned int cb_packets;
310 	unsigned int notify_count;
311 	/* Statistics to supplement MAC stats */
312 	unsigned long tx_packets;
313 
314 	/* Members shared between paths and sometimes updated */
315 	unsigned int empty_read_count ____cacheline_aligned_in_smp;
316 #define EFX_EMPTY_COUNT_VALID 0x80000000
317 	atomic_t flush_outstanding;
318 };
319 
320 #define EFX_TX_CB_ORDER	7
321 #define EFX_TX_CB_SIZE	(1 << EFX_TX_CB_ORDER) - NET_IP_ALIGN
322 
323 /**
324  * struct efx_rx_buffer - An Efx RX data buffer
325  * @dma_addr: DMA base address of the buffer
326  * @page: The associated page buffer.
327  *	Will be %NULL if the buffer slot is currently free.
328  * @page_offset: If pending: offset in @page of DMA base address.
329  *	If completed: offset in @page of Ethernet header.
330  * @len: If pending: length for DMA descriptor.
331  *	If completed: received length, excluding hash prefix.
332  * @flags: Flags for buffer and packet state.  These are only set on the
333  *	first buffer of a scattered packet.
334  */
335 struct efx_rx_buffer {
336 	dma_addr_t dma_addr;
337 	struct page *page;
338 	u16 page_offset;
339 	u16 len;
340 	u16 flags;
341 };
342 #define EFX_RX_BUF_LAST_IN_PAGE	0x0001
343 #define EFX_RX_PKT_CSUMMED	0x0002
344 #define EFX_RX_PKT_DISCARD	0x0004
345 #define EFX_RX_PKT_TCP		0x0040
346 #define EFX_RX_PKT_PREFIX_LEN	0x0080	/* length is in prefix only */
347 #define EFX_RX_PKT_CSUM_LEVEL	0x0200
348 
349 /**
350  * struct efx_rx_page_state - Page-based rx buffer state
351  *
352  * Inserted at the start of every page allocated for receive buffers.
353  * Used to facilitate sharing dma mappings between recycled rx buffers
354  * and those passed up to the kernel.
355  *
356  * @dma_addr: The dma address of this page.
357  */
358 struct efx_rx_page_state {
359 	dma_addr_t dma_addr;
360 
361 	unsigned int __pad[] ____cacheline_aligned;
362 };
363 
364 /**
365  * struct efx_rx_queue - An Efx RX queue
366  * @efx: The associated Efx NIC
367  * @core_index:  Index of network core RX queue.  Will be >= 0 iff this
368  *	is associated with a real RX queue.
369  * @buffer: The software buffer ring
370  * @rxd: The hardware descriptor ring
371  * @ptr_mask: The size of the ring minus 1.
372  * @refill_enabled: Enable refill whenever fill level is low
373  * @flush_pending: Set when a RX flush is pending. Has the same lifetime as
374  *	@rxq_flush_pending.
375  * @grant_credits: Posted RX descriptors need to be granted to the MAE with
376  *	%MC_CMD_MAE_COUNTERS_STREAM_GIVE_CREDITS.  For %EFX_EXTRA_CHANNEL_TC,
377  *	and only supported on EF100.
378  * @added_count: Number of buffers added to the receive queue.
379  * @notified_count: Number of buffers given to NIC (<= @added_count).
380  * @granted_count: Number of buffers granted to the MAE (<= @notified_count).
381  * @removed_count: Number of buffers removed from the receive queue.
382  * @scatter_n: Used by NIC specific receive code.
383  * @scatter_len: Used by NIC specific receive code.
384  * @page_ring: The ring to store DMA mapped pages for reuse.
385  * @page_add: Counter to calculate the write pointer for the recycle ring.
386  * @page_remove: Counter to calculate the read pointer for the recycle ring.
387  * @page_recycle_count: The number of pages that have been recycled.
388  * @page_recycle_failed: The number of pages that couldn't be recycled because
389  *      the kernel still held a reference to them.
390  * @page_recycle_full: The number of pages that were released because the
391  *      recycle ring was full.
392  * @page_ptr_mask: The number of pages in the RX recycle ring minus 1.
393  * @max_fill: RX descriptor maximum fill level (<= ring size)
394  * @fast_fill_trigger: RX descriptor fill level that will trigger a fast fill
395  *	(<= @max_fill)
396  * @min_fill: RX descriptor minimum non-zero fill level.
397  *	This records the minimum fill level observed when a ring
398  *	refill was triggered.
399  * @recycle_count: RX buffer recycle counter.
400  * @slow_fill: Timer used to defer efx_nic_generate_fill_event().
401  * @grant_work: workitem used to grant credits to the MAE if @grant_credits
402  * @rx_packets: Number of packets received since this struct was created
403  * @rx_bytes: Number of bytes received since this struct was created
404  * @old_rx_packets: Value of @rx_packets as of last efx_init_rx_queue()
405  * @old_rx_bytes: Value of @rx_bytes as of last efx_init_rx_queue()
406  * @xdp_rxq_info: XDP specific RX queue information.
407  * @xdp_rxq_info_valid: Is xdp_rxq_info valid data?.
408  */
409 struct efx_rx_queue {
410 	struct efx_nic *efx;
411 	int core_index;
412 	struct efx_rx_buffer *buffer;
413 	struct efx_buffer rxd;
414 	unsigned int ptr_mask;
415 	bool refill_enabled;
416 	bool flush_pending;
417 	bool grant_credits;
418 
419 	unsigned int added_count;
420 	unsigned int notified_count;
421 	unsigned int granted_count;
422 	unsigned int removed_count;
423 	unsigned int scatter_n;
424 	unsigned int scatter_len;
425 	struct page **page_ring;
426 	unsigned int page_add;
427 	unsigned int page_remove;
428 	unsigned int page_recycle_count;
429 	unsigned int page_recycle_failed;
430 	unsigned int page_recycle_full;
431 	unsigned int page_ptr_mask;
432 	unsigned int max_fill;
433 	unsigned int fast_fill_trigger;
434 	unsigned int min_fill;
435 	unsigned int min_overfill;
436 	unsigned int recycle_count;
437 	struct timer_list slow_fill;
438 	unsigned int slow_fill_count;
439 	struct work_struct grant_work;
440 	/* Statistics to supplement MAC stats */
441 	unsigned long rx_packets;
442 	unsigned long rx_bytes;
443 	unsigned long old_rx_packets;
444 	unsigned long old_rx_bytes;
445 	struct xdp_rxq_info xdp_rxq_info;
446 	bool xdp_rxq_info_valid;
447 };
448 
449 enum efx_sync_events_state {
450 	SYNC_EVENTS_DISABLED = 0,
451 	SYNC_EVENTS_QUIESCENT,
452 	SYNC_EVENTS_REQUESTED,
453 	SYNC_EVENTS_VALID,
454 };
455 
456 /**
457  * struct efx_channel - An Efx channel
458  *
459  * A channel comprises an event queue, at least one TX queue, at least
460  * one RX queue, and an associated tasklet for processing the event
461  * queue.
462  *
463  * @efx: Associated Efx NIC
464  * @channel: Channel instance number
465  * @type: Channel type definition
466  * @eventq_init: Event queue initialised flag
467  * @enabled: Channel enabled indicator
468  * @irq: IRQ number (MSI and MSI-X only)
469  * @irq_moderation_us: IRQ moderation value (in microseconds)
470  * @napi_dev: Net device used with NAPI
471  * @napi_str: NAPI control structure
472  * @state: state for NAPI vs busy polling
473  * @state_lock: lock protecting @state
474  * @eventq: Event queue buffer
475  * @eventq_mask: Event queue pointer mask
476  * @eventq_read_ptr: Event queue read pointer
477  * @event_test_cpu: Last CPU to handle interrupt or test event for this channel
478  * @irq_count: Number of IRQs since last adaptive moderation decision
479  * @irq_mod_score: IRQ moderation score
480  * @rfs_filter_count: number of accelerated RFS filters currently in place;
481  *	equals the count of @rps_flow_id slots filled
482  * @rfs_last_expiry: value of jiffies last time some accelerated RFS filters
483  *	were checked for expiry
484  * @rfs_expire_index: next accelerated RFS filter ID to check for expiry
485  * @n_rfs_succeeded: number of successful accelerated RFS filter insertions
486  * @n_rfs_failed: number of failed accelerated RFS filter insertions
487  * @filter_work: Work item for efx_filter_rfs_expire()
488  * @rps_flow_id: Flow IDs of filters allocated for accelerated RFS,
489  *      indexed by filter ID
490  * @n_rx_ip_hdr_chksum_err: Count of RX IP header checksum errors
491  * @n_rx_tcp_udp_chksum_err: Count of RX TCP and UDP checksum errors
492  * @n_rx_frm_trunc: Count of RX_FRM_TRUNC errors
493  * @n_rx_overlength: Count of RX_OVERLENGTH errors
494  * @n_skbuff_leaks: Count of skbuffs leaked due to RX overrun
495  * @n_rx_nodesc_trunc: Number of RX packets truncated and then dropped due to
496  *	lack of descriptors
497  * @n_rx_merge_events: Number of RX merged completion events
498  * @n_rx_merge_packets: Number of RX packets completed by merged events
499  * @n_rx_xdp_drops: Count of RX packets intentionally dropped due to XDP
500  * @n_rx_xdp_bad_drops: Count of RX packets dropped due to XDP errors
501  * @n_rx_xdp_tx: Count of RX packets retransmitted due to XDP
502  * @n_rx_xdp_redirect: Count of RX packets redirected to a different NIC by XDP
503  * @n_rx_mport_bad: Count of RX packets dropped because their ingress mport was
504  *	not recognised
505  * @old_n_rx_hw_drops: Count of all RX packets dropped for any reason as of last
506  *	efx_start_channels()
507  * @old_n_rx_hw_drop_overruns: Value of @n_rx_nodesc_trunc as of last
508  *	efx_start_channels()
509  * @rx_pkt_n_frags: Number of fragments in next packet to be delivered by
510  *	__efx_rx_packet(), or zero if there is none
511  * @rx_pkt_index: Ring index of first buffer for next packet to be delivered
512  *	by __efx_rx_packet(), if @rx_pkt_n_frags != 0
513  * @rx_list: list of SKBs from current RX, awaiting processing
514  * @rx_queue: RX queue for this channel
515  * @tx_queue: TX queues for this channel
516  * @tx_queue_by_type: pointers into @tx_queue, or %NULL, indexed by txq type
517  * @sync_events_state: Current state of sync events on this channel
518  * @sync_timestamp_major: Major part of the last ptp sync event
519  * @sync_timestamp_minor: Minor part of the last ptp sync event
520  */
521 struct efx_channel {
522 	struct efx_nic *efx;
523 	int channel;
524 	const struct efx_channel_type *type;
525 	bool eventq_init;
526 	bool enabled;
527 	int irq;
528 	unsigned int irq_moderation_us;
529 	struct net_device *napi_dev;
530 	struct napi_struct napi_str;
531 #ifdef CONFIG_NET_RX_BUSY_POLL
532 	unsigned long busy_poll_state;
533 #endif
534 	struct efx_buffer eventq;
535 	unsigned int eventq_mask;
536 	unsigned int eventq_read_ptr;
537 	int event_test_cpu;
538 
539 	unsigned int irq_count;
540 	unsigned int irq_mod_score;
541 #ifdef CONFIG_RFS_ACCEL
542 	unsigned int rfs_filter_count;
543 	unsigned int rfs_last_expiry;
544 	unsigned int rfs_expire_index;
545 	unsigned int n_rfs_succeeded;
546 	unsigned int n_rfs_failed;
547 	struct delayed_work filter_work;
548 #define RPS_FLOW_ID_INVALID 0xFFFFFFFF
549 	u32 *rps_flow_id;
550 #endif
551 
552 	unsigned int n_rx_ip_hdr_chksum_err;
553 	unsigned int n_rx_tcp_udp_chksum_err;
554 	unsigned int n_rx_outer_ip_hdr_chksum_err;
555 	unsigned int n_rx_outer_tcp_udp_chksum_err;
556 	unsigned int n_rx_inner_ip_hdr_chksum_err;
557 	unsigned int n_rx_inner_tcp_udp_chksum_err;
558 	unsigned int n_rx_eth_crc_err;
559 	unsigned int n_rx_frm_trunc;
560 	unsigned int n_rx_overlength;
561 	unsigned int n_skbuff_leaks;
562 	unsigned int n_rx_nodesc_trunc;
563 	unsigned int n_rx_merge_events;
564 	unsigned int n_rx_merge_packets;
565 	unsigned int n_rx_xdp_drops;
566 	unsigned int n_rx_xdp_bad_drops;
567 	unsigned int n_rx_xdp_tx;
568 	unsigned int n_rx_xdp_redirect;
569 	unsigned int n_rx_mport_bad;
570 
571 	unsigned int old_n_rx_hw_drops;
572 	unsigned int old_n_rx_hw_drop_overruns;
573 
574 	unsigned int rx_pkt_n_frags;
575 	unsigned int rx_pkt_index;
576 
577 	struct list_head *rx_list;
578 
579 	struct efx_rx_queue rx_queue;
580 	struct efx_tx_queue tx_queue[EFX_MAX_TXQ_PER_CHANNEL];
581 	struct efx_tx_queue *tx_queue_by_type[EFX_TXQ_TYPES];
582 
583 	enum efx_sync_events_state sync_events_state;
584 	u32 sync_timestamp_major;
585 	u32 sync_timestamp_minor;
586 };
587 
588 /**
589  * struct efx_msi_context - Context for each MSI
590  * @efx: The associated NIC
591  * @index: Index of the channel/IRQ
592  * @name: Name of the channel/IRQ
593  *
594  * Unlike &struct efx_channel, this is never reallocated and is always
595  * safe for the IRQ handler to access.
596  */
597 struct efx_msi_context {
598 	struct efx_nic *efx;
599 	unsigned int index;
600 	char name[IFNAMSIZ + 6];
601 };
602 
603 /**
604  * struct efx_channel_type - distinguishes traffic and extra channels
605  * @handle_no_channel: Handle failure to allocate an extra channel
606  * @pre_probe: Set up extra state prior to initialisation
607  * @start: called early in efx_start_channels()
608  * @stop: called early in efx_stop_channels()
609  * @post_remove: Tear down extra state after finalisation, if allocated.
610  *	May be called on channels that have not been probed.
611  * @get_name: Generate the channel's name (used for its IRQ handler)
612  * @copy: Copy the channel state prior to reallocation.  May be %NULL if
613  *	reallocation is not supported.
614  * @receive_skb: Handle an skb ready to be passed to netif_receive_skb()
615  * @receive_raw: Handle an RX buffer ready to be passed to __efx_rx_packet()
616  * @want_txqs: Determine whether this channel should have TX queues
617  *	created.  If %NULL, TX queues are not created.
618  * @keep_eventq: Flag for whether event queue should be kept initialised
619  *	while the device is stopped
620  * @want_pio: Flag for whether PIO buffers should be linked to this
621  *	channel's TX queues.
622  */
623 struct efx_channel_type {
624 	void (*handle_no_channel)(struct efx_nic *);
625 	int (*pre_probe)(struct efx_channel *);
626 	int (*start)(struct efx_channel *);
627 	void (*stop)(struct efx_channel *);
628 	void (*post_remove)(struct efx_channel *);
629 	void (*get_name)(struct efx_channel *, char *buf, size_t len);
630 	struct efx_channel *(*copy)(const struct efx_channel *);
631 	bool (*receive_skb)(struct efx_channel *, struct sk_buff *);
632 	bool (*receive_raw)(struct efx_rx_queue *, u32);
633 	bool (*want_txqs)(struct efx_channel *);
634 	bool keep_eventq;
635 	bool want_pio;
636 };
637 
638 enum efx_led_mode {
639 	EFX_LED_OFF	= 0,
640 	EFX_LED_ON	= 1,
641 	EFX_LED_DEFAULT	= 2
642 };
643 
644 #define STRING_TABLE_LOOKUP(val, member) \
645 	((val) < member ## _max) ? member ## _names[val] : "(invalid)"
646 
647 extern const char *const efx_loopback_mode_names[];
648 extern const unsigned int efx_loopback_mode_max;
649 #define LOOPBACK_MODE(efx) \
650 	STRING_TABLE_LOOKUP((efx)->loopback_mode, efx_loopback_mode)
651 
652 enum efx_int_mode {
653 	/* Be careful if altering to correct macro below */
654 	EFX_INT_MODE_MSIX = 0,
655 	EFX_INT_MODE_MSI = 1,
656 	EFX_INT_MODE_LEGACY = 2,
657 	EFX_INT_MODE_MAX	/* Insert any new items before this */
658 };
659 #define EFX_INT_MODE_USE_MSI(x) (((x)->interrupt_mode) <= EFX_INT_MODE_MSI)
660 
661 enum nic_state {
662 	STATE_UNINIT = 0,	/* device being probed/removed */
663 	STATE_PROBED,		/* hardware probed */
664 	STATE_NET_DOWN,		/* netdev registered */
665 	STATE_NET_UP,		/* ready for traffic */
666 	STATE_DISABLED,		/* device disabled due to hardware errors */
667 
668 	STATE_RECOVERY = 0x100,/* recovering from PCI error */
669 	STATE_FROZEN = 0x200,	/* frozen by power management */
670 };
671 
672 static inline bool efx_net_active(enum nic_state state)
673 {
674 	return state == STATE_NET_DOWN || state == STATE_NET_UP;
675 }
676 
677 static inline bool efx_frozen(enum nic_state state)
678 {
679 	return state & STATE_FROZEN;
680 }
681 
682 static inline bool efx_recovering(enum nic_state state)
683 {
684 	return state & STATE_RECOVERY;
685 }
686 
687 static inline enum nic_state efx_freeze(enum nic_state state)
688 {
689 	WARN_ON(!efx_net_active(state));
690 	return state | STATE_FROZEN;
691 }
692 
693 static inline enum nic_state efx_thaw(enum nic_state state)
694 {
695 	WARN_ON(!efx_frozen(state));
696 	return state & ~STATE_FROZEN;
697 }
698 
699 static inline enum nic_state efx_recover(enum nic_state state)
700 {
701 	WARN_ON(!efx_net_active(state));
702 	return state | STATE_RECOVERY;
703 }
704 
705 static inline enum nic_state efx_recovered(enum nic_state state)
706 {
707 	WARN_ON(!efx_recovering(state));
708 	return state & ~STATE_RECOVERY;
709 }
710 
711 /* Forward declaration */
712 struct efx_nic;
713 
714 /* Pseudo bit-mask flow control field */
715 #define EFX_FC_RX	FLOW_CTRL_RX
716 #define EFX_FC_TX	FLOW_CTRL_TX
717 #define EFX_FC_AUTO	4
718 
719 /**
720  * struct efx_link_state - Current state of the link
721  * @up: Link is up
722  * @fd: Link is full-duplex
723  * @fc: Actual flow control flags
724  * @speed: Link speed (Mbps)
725  */
726 struct efx_link_state {
727 	bool up;
728 	bool fd;
729 	u8 fc;
730 	unsigned int speed;
731 };
732 
733 static inline bool efx_link_state_equal(const struct efx_link_state *left,
734 					const struct efx_link_state *right)
735 {
736 	return left->up == right->up && left->fd == right->fd &&
737 		left->fc == right->fc && left->speed == right->speed;
738 }
739 
740 /**
741  * enum efx_phy_mode - PHY operating mode flags
742  * @PHY_MODE_NORMAL: on and should pass traffic
743  * @PHY_MODE_TX_DISABLED: on with TX disabled
744  * @PHY_MODE_LOW_POWER: set to low power through MDIO
745  * @PHY_MODE_OFF: switched off through external control
746  * @PHY_MODE_SPECIAL: on but will not pass traffic
747  */
748 enum efx_phy_mode {
749 	PHY_MODE_NORMAL		= 0,
750 	PHY_MODE_TX_DISABLED	= 1,
751 	PHY_MODE_LOW_POWER	= 2,
752 	PHY_MODE_OFF		= 4,
753 	PHY_MODE_SPECIAL	= 8,
754 };
755 
756 static inline bool efx_phy_mode_disabled(enum efx_phy_mode mode)
757 {
758 	return !!(mode & ~PHY_MODE_TX_DISABLED);
759 }
760 
761 /**
762  * struct efx_hw_stat_desc - Description of a hardware statistic
763  * @name: Name of the statistic as visible through ethtool, or %NULL if
764  *	it should not be exposed
765  * @dma_width: Width in bits (0 for non-DMA statistics)
766  * @offset: Offset within stats (ignored for non-DMA statistics)
767  */
768 struct efx_hw_stat_desc {
769 	const char *name;
770 	u16 dma_width;
771 	u16 offset;
772 };
773 
774 struct vfdi_status;
775 
776 /* The reserved RSS context value */
777 #define EFX_MCDI_RSS_CONTEXT_INVALID	0xffffffff
778 /**
779  * struct efx_rss_context_priv - driver private data for an RSS context
780  * @context_id: the RSS_CONTEXT_ID returned by MC firmware, or
781  *	%EFX_MCDI_RSS_CONTEXT_INVALID if this context is not present on the NIC.
782  * @rx_hash_udp_4tuple: UDP 4-tuple hashing enabled
783  */
784 struct efx_rss_context_priv {
785 	u32 context_id;
786 	bool rx_hash_udp_4tuple;
787 };
788 
789 /**
790  * struct efx_rss_context - an RSS context
791  * @priv: hardware-specific state
792  * @rx_hash_key: Toeplitz hash key for this RSS context
793  * @indir_table: Indirection table for this RSS context
794  */
795 struct efx_rss_context {
796 	struct efx_rss_context_priv priv;
797 	u8 rx_hash_key[40];
798 	u32 rx_indir_table[128];
799 };
800 
801 #ifdef CONFIG_RFS_ACCEL
802 /* Order of these is important, since filter_id >= %EFX_ARFS_FILTER_ID_PENDING
803  * is used to test if filter does or will exist.
804  */
805 #define EFX_ARFS_FILTER_ID_PENDING	-1
806 #define EFX_ARFS_FILTER_ID_ERROR	-2
807 #define EFX_ARFS_FILTER_ID_REMOVING	-3
808 /**
809  * struct efx_arfs_rule - record of an ARFS filter and its IDs
810  * @node: linkage into hash table
811  * @spec: details of the filter (used as key for hash table).  Use efx->type to
812  *	determine which member to use.
813  * @rxq_index: channel to which the filter will steer traffic.
814  * @arfs_id: filter ID which was returned to ARFS
815  * @filter_id: index in software filter table.  May be
816  *	%EFX_ARFS_FILTER_ID_PENDING if filter was not inserted yet,
817  *	%EFX_ARFS_FILTER_ID_ERROR if filter insertion failed, or
818  *	%EFX_ARFS_FILTER_ID_REMOVING if expiry is currently removing the filter.
819  */
820 struct efx_arfs_rule {
821 	struct hlist_node node;
822 	struct efx_filter_spec spec;
823 	u16 rxq_index;
824 	u16 arfs_id;
825 	s32 filter_id;
826 };
827 
828 /* Size chosen so that the table is one page (4kB) */
829 #define EFX_ARFS_HASH_TABLE_SIZE	512
830 
831 /**
832  * struct efx_async_filter_insertion - Request to asynchronously insert a filter
833  * @net_dev: Reference to the netdevice
834  * @spec: The filter to insert
835  * @work: Workitem for this request
836  * @rxq_index: Identifies the channel for which this request was made
837  * @flow_id: Identifies the kernel-side flow for which this request was made
838  */
839 struct efx_async_filter_insertion {
840 	struct net_device *net_dev;
841 	struct efx_filter_spec spec;
842 	struct work_struct work;
843 	u16 rxq_index;
844 	u32 flow_id;
845 };
846 
847 /* Maximum number of ARFS workitems that may be in flight on an efx_nic */
848 #define EFX_RPS_MAX_IN_FLIGHT	8
849 #endif /* CONFIG_RFS_ACCEL */
850 
851 enum efx_xdp_tx_queues_mode {
852 	EFX_XDP_TX_QUEUES_DEDICATED,	/* one queue per core, locking not needed */
853 	EFX_XDP_TX_QUEUES_SHARED,	/* each queue used by more than 1 core */
854 	EFX_XDP_TX_QUEUES_BORROWED	/* queues borrowed from net stack */
855 };
856 
857 struct efx_mae;
858 
859 /**
860  * struct efx_nic - an Efx NIC
861  * @name: Device name (net device name or bus id before net device registered)
862  * @pci_dev: The PCI device
863  * @node: List node for maintaning primary/secondary function lists
864  * @primary: &struct efx_nic instance for the primary function of this
865  *	controller.  May be the same structure, and may be %NULL if no
866  *	primary function is bound.  Serialised by rtnl_lock.
867  * @secondary_list: List of &struct efx_nic instances for the secondary PCI
868  *	functions of the controller, if this is for the primary function.
869  *	Serialised by rtnl_lock.
870  * @type: Controller type attributes
871  * @legacy_irq: IRQ number
872  * @workqueue: Workqueue for port reconfigures and the HW monitor.
873  *	Work items do not hold and must not acquire RTNL.
874  * @workqueue_name: Name of workqueue
875  * @reset_work: Scheduled reset workitem
876  * @membase_phys: Memory BAR value as physical address
877  * @membase: Memory BAR value
878  * @vi_stride: step between per-VI registers / memory regions
879  * @interrupt_mode: Interrupt mode
880  * @timer_quantum_ns: Interrupt timer quantum, in nanoseconds
881  * @timer_max_ns: Interrupt timer maximum value, in nanoseconds
882  * @irq_rx_adaptive: Adaptive IRQ moderation enabled for RX event queues
883  * @irqs_hooked: Channel interrupts are hooked
884  * @irq_rx_mod_step_us: Step size for IRQ moderation for RX event queues
885  * @irq_rx_moderation_us: IRQ moderation time for RX event queues
886  * @msg_enable: Log message enable flags
887  * @state: Device state number (%STATE_*). Serialised by the rtnl_lock.
888  * @reset_pending: Bitmask for pending resets
889  * @tx_queue: TX DMA queues
890  * @rx_queue: RX DMA queues
891  * @channel: Channels
892  * @msi_context: Context for each MSI
893  * @extra_channel_types: Types of extra (non-traffic) channels that
894  *	should be allocated for this NIC
895  * @mae: Details of the Match Action Engine
896  * @xdp_tx_queue_count: Number of entries in %xdp_tx_queues.
897  * @xdp_tx_queues: Array of pointers to tx queues used for XDP transmit.
898  * @xdp_txq_queues_mode: XDP TX queues sharing strategy.
899  * @rxq_entries: Size of receive queues requested by user.
900  * @txq_entries: Size of transmit queues requested by user.
901  * @txq_stop_thresh: TX queue fill level at or above which we stop it.
902  * @txq_wake_thresh: TX queue fill level at or below which we wake it.
903  * @tx_dc_base: Base qword address in SRAM of TX queue descriptor caches
904  * @rx_dc_base: Base qword address in SRAM of RX queue descriptor caches
905  * @sram_lim_qw: Qword address limit of SRAM
906  * @n_channels: Number of channels in use
907  * @n_rx_channels: Number of channels used for RX (= number of RX queues)
908  * @n_tx_channels: Number of channels used for TX
909  * @n_extra_tx_channels: Number of extra channels with TX queues
910  * @tx_queues_per_channel: number of TX queues probed on each channel
911  * @n_xdp_channels: Number of channels used for XDP TX
912  * @xdp_channel_offset: Offset of zeroth channel used for XPD TX.
913  * @xdp_tx_per_channel: Max number of TX queues on an XDP TX channel.
914  * @rx_ip_align: RX DMA address offset to have IP header aligned in
915  *	in accordance with NET_IP_ALIGN
916  * @rx_dma_len: Current maximum RX DMA length
917  * @rx_buffer_order: Order (log2) of number of pages for each RX buffer
918  * @rx_buffer_truesize: Amortised allocation size of an RX buffer,
919  *	for use in sk_buff::truesize
920  * @rx_prefix_size: Size of RX prefix before packet data
921  * @rx_packet_hash_offset: Offset of RX flow hash from start of packet data
922  *	(valid only if @rx_prefix_size != 0; always negative)
923  * @rx_packet_len_offset: Offset of RX packet length from start of packet data
924  *	(valid only for NICs that set %EFX_RX_PKT_PREFIX_LEN; always negative)
925  * @rx_packet_ts_offset: Offset of timestamp from start of packet data
926  *	(valid only if channel->sync_timestamps_enabled; always negative)
927  * @rx_scatter: Scatter mode enabled for receives
928  * @rss_context: Main RSS context.
929  * @vport_id: The function's vport ID, only relevant for PFs
930  * @int_error_count: Number of internal errors seen recently
931  * @int_error_expire: Time at which error count will be expired
932  * @must_realloc_vis: Flag: VIs have yet to be reallocated after MC reboot
933  * @irq_soft_enabled: Are IRQs soft-enabled? If not, IRQ handler will
934  *	acknowledge but do nothing else.
935  * @irq_status: Interrupt status buffer
936  * @irq_zero_count: Number of legacy IRQs seen with queue flags == 0
937  * @irq_level: IRQ level/index for IRQs not triggered by an event queue
938  * @selftest_work: Work item for asynchronous self-test
939  * @mtd_list: List of MTDs attached to the NIC
940  * @nic_data: Hardware dependent state
941  * @mcdi: Management-Controller-to-Driver Interface state
942  * @mac_lock: MAC access lock. Protects @port_enabled, @phy_mode,
943  *	efx_monitor() and efx_reconfigure_port()
944  * @port_enabled: Port enabled indicator.
945  *	Serialises efx_stop_all(), efx_start_all(), efx_monitor() and
946  *	efx_mac_work() with kernel interfaces. Safe to read under any
947  *	one of the rtnl_lock, mac_lock, or netif_tx_lock, but all three must
948  *	be held to modify it.
949  * @port_initialized: Port initialized?
950  * @net_dev: Operating system network device. Consider holding the rtnl lock
951  * @fixed_features: Features which cannot be turned off
952  * @num_mac_stats: Number of MAC stats reported by firmware (MAC_STATS_NUM_STATS
953  *	field of %MC_CMD_GET_CAPABILITIES_V4 response, or %MC_CMD_MAC_NSTATS)
954  * @stats_buffer: DMA buffer for statistics
955  * @phy_type: PHY type
956  * @phy_data: PHY private data (including PHY-specific stats)
957  * @mdio: PHY MDIO interface
958  * @mdio_bus: PHY MDIO bus ID (only used by Siena)
959  * @phy_mode: PHY operating mode. Serialised by @mac_lock.
960  * @link_advertising: Autonegotiation advertising flags
961  * @fec_config: Forward Error Correction configuration flags.  For bit positions
962  *	see &enum ethtool_fec_config_bits.
963  * @link_state: Current state of the link
964  * @n_link_state_changes: Number of times the link has changed state
965  * @wanted_fc: Wanted flow control flags
966  * @fc_disable: When non-zero flow control is disabled. Typically used to
967  *	ensure that network back pressure doesn't delay dma queue flushes.
968  *	Serialised by the rtnl lock.
969  * @mac_work: Work item for changing MAC promiscuity and multicast hash
970  * @loopback_mode: Loopback status
971  * @loopback_modes: Supported loopback mode bitmask
972  * @loopback_selftest: Offline self-test private state
973  * @xdp_prog: Current XDP programme for this interface
974  * @filter_sem: Filter table rw_semaphore, protects existence of @filter_state
975  * @filter_state: Architecture-dependent filter table state
976  * @rps_mutex: Protects RPS state of all channels
977  * @rps_slot_map: bitmap of in-flight entries in @rps_slot
978  * @rps_slot: array of ARFS insertion requests for efx_filter_rfs_work()
979  * @rps_hash_lock: Protects ARFS filter mapping state (@rps_hash_table and
980  *	@rps_next_id).
981  * @rps_hash_table: Mapping between ARFS filters and their various IDs
982  * @rps_next_id: next arfs_id for an ARFS filter
983  * @active_queues: Count of RX and TX queues that haven't been flushed and drained.
984  * @rxq_flush_pending: Count of number of receive queues that need to be flushed.
985  *	Decremented when the efx_flush_rx_queue() is called.
986  * @rxq_flush_outstanding: Count of number of RX flushes started but not yet
987  *	completed (either success or failure). Not used when MCDI is used to
988  *	flush receive queues.
989  * @flush_wq: wait queue used by efx_nic_flush_queues() to wait for flush completions.
990  * @vf_count: Number of VFs intended to be enabled.
991  * @vf_init_count: Number of VFs that have been fully initialised.
992  * @vi_scale: log2 number of vnics per VF.
993  * @vf_reps_lock: Protects vf_reps list
994  * @vf_reps: local VF reps
995  * @ptp_data: PTP state data
996  * @ptp_warned: has this NIC seen and warned about unexpected PTP events?
997  * @vpd_sn: Serial number read from VPD
998  * @xdp_rxq_info_failed: Have any of the rx queues failed to initialise their
999  *      xdp_rxq_info structures?
1000  * @netdev_notifier: Netdevice notifier.
1001  * @netevent_notifier: Netevent notifier (for neighbour updates).
1002  * @tc: state for TC offload (EF100).
1003  * @devlink: reference to devlink structure owned by this device
1004  * @dl_port: devlink port associated with the PF
1005  * @mem_bar: The BAR that is mapped into membase.
1006  * @reg_base: Offset from the start of the bar to the function control window.
1007  * @monitor_work: Hardware monitor workitem
1008  * @biu_lock: BIU (bus interface unit) lock
1009  * @last_irq_cpu: Last CPU to handle a possible test interrupt.  This
1010  *	field is used by efx_test_interrupts() to verify that an
1011  *	interrupt has occurred.
1012  * @stats_lock: Statistics update lock. Must be held when calling
1013  *	efx_nic_type::{update,start,stop}_stats.
1014  * @n_rx_noskb_drops: Count of RX packets dropped due to failure to allocate an skb
1015  *
1016  * This is stored in the private area of the &struct net_device.
1017  */
1018 struct efx_nic {
1019 	/* The following fields should be written very rarely */
1020 
1021 	char name[IFNAMSIZ];
1022 	struct list_head node;
1023 	struct efx_nic *primary;
1024 	struct list_head secondary_list;
1025 	struct pci_dev *pci_dev;
1026 	unsigned int port_num;
1027 	const struct efx_nic_type *type;
1028 	int legacy_irq;
1029 	bool eeh_disabled_legacy_irq;
1030 	struct workqueue_struct *workqueue;
1031 	char workqueue_name[16];
1032 	struct work_struct reset_work;
1033 	resource_size_t membase_phys;
1034 	void __iomem *membase;
1035 
1036 	unsigned int vi_stride;
1037 
1038 	enum efx_int_mode interrupt_mode;
1039 	unsigned int timer_quantum_ns;
1040 	unsigned int timer_max_ns;
1041 	bool irq_rx_adaptive;
1042 	bool irqs_hooked;
1043 	unsigned int irq_mod_step_us;
1044 	unsigned int irq_rx_moderation_us;
1045 	u32 msg_enable;
1046 
1047 	enum nic_state state;
1048 	unsigned long reset_pending;
1049 
1050 	struct efx_channel *channel[EFX_MAX_CHANNELS];
1051 	struct efx_msi_context msi_context[EFX_MAX_CHANNELS];
1052 	const struct efx_channel_type *
1053 	extra_channel_type[EFX_MAX_EXTRA_CHANNELS];
1054 	struct efx_mae *mae;
1055 
1056 	unsigned int xdp_tx_queue_count;
1057 	struct efx_tx_queue **xdp_tx_queues;
1058 	enum efx_xdp_tx_queues_mode xdp_txq_queues_mode;
1059 
1060 	unsigned rxq_entries;
1061 	unsigned txq_entries;
1062 	unsigned int txq_stop_thresh;
1063 	unsigned int txq_wake_thresh;
1064 
1065 	unsigned tx_dc_base;
1066 	unsigned rx_dc_base;
1067 	unsigned sram_lim_qw;
1068 
1069 	unsigned int max_channels;
1070 	unsigned int max_vis;
1071 	unsigned int max_tx_channels;
1072 	unsigned n_channels;
1073 	unsigned n_rx_channels;
1074 	unsigned rss_spread;
1075 	unsigned tx_channel_offset;
1076 	unsigned n_tx_channels;
1077 	unsigned n_extra_tx_channels;
1078 	unsigned int tx_queues_per_channel;
1079 	unsigned int n_xdp_channels;
1080 	unsigned int xdp_channel_offset;
1081 	unsigned int xdp_tx_per_channel;
1082 	unsigned int rx_ip_align;
1083 	unsigned int rx_dma_len;
1084 	unsigned int rx_buffer_order;
1085 	unsigned int rx_buffer_truesize;
1086 	unsigned int rx_page_buf_step;
1087 	unsigned int rx_bufs_per_page;
1088 	unsigned int rx_pages_per_batch;
1089 	unsigned int rx_prefix_size;
1090 	int rx_packet_hash_offset;
1091 	int rx_packet_len_offset;
1092 	int rx_packet_ts_offset;
1093 	bool rx_scatter;
1094 	struct efx_rss_context rss_context;
1095 	u32 vport_id;
1096 
1097 	unsigned int_error_count;
1098 	unsigned long int_error_expire;
1099 
1100 	bool must_realloc_vis;
1101 	bool irq_soft_enabled;
1102 	struct efx_buffer irq_status;
1103 	unsigned irq_zero_count;
1104 	unsigned irq_level;
1105 	struct delayed_work selftest_work;
1106 
1107 #ifdef CONFIG_SFC_MTD
1108 	struct list_head mtd_list;
1109 #endif
1110 
1111 	void *nic_data;
1112 	struct efx_mcdi_data *mcdi;
1113 
1114 	struct mutex mac_lock;
1115 	struct work_struct mac_work;
1116 	bool port_enabled;
1117 
1118 	bool mc_bist_for_other_fn;
1119 	bool port_initialized;
1120 	struct net_device *net_dev;
1121 
1122 	netdev_features_t fixed_features;
1123 
1124 	u16 num_mac_stats;
1125 	struct efx_buffer stats_buffer;
1126 	u64 rx_nodesc_drops_total;
1127 	u64 rx_nodesc_drops_while_down;
1128 	bool rx_nodesc_drops_prev_state;
1129 
1130 	unsigned int phy_type;
1131 	void *phy_data;
1132 	struct mdio_if_info mdio;
1133 	unsigned int mdio_bus;
1134 	enum efx_phy_mode phy_mode;
1135 
1136 	__ETHTOOL_DECLARE_LINK_MODE_MASK(link_advertising);
1137 	u32 fec_config;
1138 	struct efx_link_state link_state;
1139 	unsigned int n_link_state_changes;
1140 
1141 	u8 wanted_fc;
1142 	unsigned fc_disable;
1143 
1144 	atomic_t rx_reset;
1145 	enum efx_loopback_mode loopback_mode;
1146 	u64 loopback_modes;
1147 
1148 	void *loopback_selftest;
1149 	/* We access loopback_selftest immediately before running XDP,
1150 	 * so we want them next to each other.
1151 	 */
1152 	struct bpf_prog __rcu *xdp_prog;
1153 
1154 	struct rw_semaphore filter_sem;
1155 	void *filter_state;
1156 #ifdef CONFIG_RFS_ACCEL
1157 	struct mutex rps_mutex;
1158 	unsigned long rps_slot_map;
1159 	struct efx_async_filter_insertion rps_slot[EFX_RPS_MAX_IN_FLIGHT];
1160 	spinlock_t rps_hash_lock;
1161 	struct hlist_head *rps_hash_table;
1162 	u32 rps_next_id;
1163 #endif
1164 
1165 	atomic_t active_queues;
1166 	atomic_t rxq_flush_pending;
1167 	atomic_t rxq_flush_outstanding;
1168 	wait_queue_head_t flush_wq;
1169 
1170 #ifdef CONFIG_SFC_SRIOV
1171 	unsigned vf_count;
1172 	unsigned vf_init_count;
1173 	unsigned vi_scale;
1174 #endif
1175 	spinlock_t vf_reps_lock;
1176 	struct list_head vf_reps;
1177 
1178 	struct efx_ptp_data *ptp_data;
1179 	bool ptp_warned;
1180 
1181 	char *vpd_sn;
1182 	bool xdp_rxq_info_failed;
1183 
1184 	struct notifier_block netdev_notifier;
1185 	struct notifier_block netevent_notifier;
1186 	struct efx_tc_state *tc;
1187 
1188 	struct devlink *devlink;
1189 	struct devlink_port *dl_port;
1190 	unsigned int mem_bar;
1191 	u32 reg_base;
1192 
1193 	/* The following fields may be written more often */
1194 
1195 	struct delayed_work monitor_work ____cacheline_aligned_in_smp;
1196 	spinlock_t biu_lock;
1197 	int last_irq_cpu;
1198 	spinlock_t stats_lock;
1199 	atomic_t n_rx_noskb_drops;
1200 };
1201 
1202 /**
1203  * struct efx_probe_data - State after hardware probe
1204  * @pci_dev: The PCI device
1205  * @efx: Efx NIC details
1206  */
1207 struct efx_probe_data {
1208 	struct pci_dev *pci_dev;
1209 	struct efx_nic efx;
1210 };
1211 
1212 static inline struct efx_nic *efx_netdev_priv(struct net_device *dev)
1213 {
1214 	struct efx_probe_data **probe_ptr = netdev_priv(dev);
1215 	struct efx_probe_data *probe_data = *probe_ptr;
1216 
1217 	return &probe_data->efx;
1218 }
1219 
1220 static inline int efx_dev_registered(struct efx_nic *efx)
1221 {
1222 	return efx->net_dev->reg_state == NETREG_REGISTERED;
1223 }
1224 
1225 static inline unsigned int efx_port_num(struct efx_nic *efx)
1226 {
1227 	return efx->port_num;
1228 }
1229 
1230 struct efx_mtd_partition {
1231 	struct list_head node;
1232 	struct mtd_info mtd;
1233 	const char *dev_type_name;
1234 	const char *type_name;
1235 	char name[IFNAMSIZ + 20];
1236 };
1237 
1238 struct efx_udp_tunnel {
1239 #define TUNNEL_ENCAP_UDP_PORT_ENTRY_INVALID	0xffff
1240 	u16 type; /* TUNNEL_ENCAP_UDP_PORT_ENTRY_foo, see mcdi_pcol.h */
1241 	__be16 port;
1242 };
1243 
1244 /**
1245  * struct efx_nic_type - Efx device type definition
1246  * @mem_bar: Get the memory BAR
1247  * @mem_map_size: Get memory BAR mapped size
1248  * @probe: Probe the controller
1249  * @remove: Free resources allocated by probe()
1250  * @init: Initialise the controller
1251  * @dimension_resources: Dimension controller resources (buffer table,
1252  *	and VIs once the available interrupt resources are clear)
1253  * @fini: Shut down the controller
1254  * @monitor: Periodic function for polling link state and hardware monitor
1255  * @map_reset_reason: Map ethtool reset reason to a reset method
1256  * @map_reset_flags: Map ethtool reset flags to a reset method, if possible
1257  * @reset: Reset the controller hardware and possibly the PHY.  This will
1258  *	be called while the controller is uninitialised.
1259  * @probe_port: Probe the MAC and PHY
1260  * @remove_port: Free resources allocated by probe_port()
1261  * @handle_global_event: Handle a "global" event (may be %NULL)
1262  * @fini_dmaq: Flush and finalise DMA queues (RX and TX queues)
1263  * @prepare_flr: Prepare for an FLR
1264  * @finish_flr: Clean up after an FLR
1265  * @describe_stats: Describe statistics for ethtool
1266  * @update_stats: Update statistics not provided by event handling.
1267  *	Either argument may be %NULL.
1268  * @update_stats_atomic: Update statistics while in atomic context, if that
1269  *	is more limiting than @update_stats.  Otherwise, leave %NULL and
1270  *	driver core will call @update_stats.
1271  * @start_stats: Start the regular fetching of statistics
1272  * @pull_stats: Pull stats from the NIC and wait until they arrive.
1273  * @stop_stats: Stop the regular fetching of statistics
1274  * @push_irq_moderation: Apply interrupt moderation value
1275  * @reconfigure_port: Push loopback/power/txdis changes to the MAC and PHY
1276  * @prepare_enable_fc_tx: Prepare MAC to enable pause frame TX (may be %NULL)
1277  * @reconfigure_mac: Push MAC address, MTU, flow control and filter settings
1278  *	to the hardware.  Serialised by the mac_lock.
1279  * @check_mac_fault: Check MAC fault state. True if fault present.
1280  * @get_wol: Get WoL configuration from driver state
1281  * @set_wol: Push WoL configuration to the NIC
1282  * @resume_wol: Synchronise WoL state between driver and MC (e.g. after resume)
1283  * @get_fec_stats: Get standard FEC statistics.
1284  * @test_chip: Test registers. This is expected to reset the NIC.
1285  * @test_nvram: Test validity of NVRAM contents
1286  * @mcdi_request: Send an MCDI request with the given header and SDU.
1287  *	The SDU length may be any value from 0 up to the protocol-
1288  *	defined maximum, but its buffer will be padded to a multiple
1289  *	of 4 bytes.
1290  * @mcdi_poll_response: Test whether an MCDI response is available.
1291  * @mcdi_read_response: Read the MCDI response PDU.  The offset will
1292  *	be a multiple of 4.  The length may not be, but the buffer
1293  *	will be padded so it is safe to round up.
1294  * @mcdi_poll_reboot: Test whether the MCDI has rebooted.  If so,
1295  *	return an appropriate error code for aborting any current
1296  *	request; otherwise return 0.
1297  * @irq_enable_master: Enable IRQs on the NIC.  Each event queue must
1298  *	be separately enabled after this.
1299  * @irq_test_generate: Generate a test IRQ
1300  * @irq_disable_non_ev: Disable non-event IRQs on the NIC.  Each event
1301  *	queue must be separately disabled before this.
1302  * @irq_handle_msi: Handle MSI for a channel.  The @dev_id argument is
1303  *	a pointer to the &struct efx_msi_context for the channel.
1304  * @irq_handle_legacy: Handle legacy interrupt.  The @dev_id argument
1305  *	is a pointer to the &struct efx_nic.
1306  * @tx_probe: Allocate resources for TX queue (and select TXQ type)
1307  * @tx_init: Initialise TX queue on the NIC
1308  * @tx_remove: Free resources for TX queue
1309  * @tx_write: Write TX descriptors and doorbell
1310  * @tx_enqueue: Add an SKB to TX queue
1311  * @rx_push_rss_config: Write RSS hash key and indirection table to the NIC
1312  * @rx_pull_rss_config: Read RSS hash key and indirection table back from the NIC
1313  * @rx_push_rss_context_config: Write RSS hash key and indirection table for
1314  *	user RSS context to the NIC
1315  * @rx_pull_rss_context_config: Read RSS hash key and indirection table for user
1316  *	RSS context back from the NIC
1317  * @rx_probe: Allocate resources for RX queue
1318  * @rx_init: Initialise RX queue on the NIC
1319  * @rx_remove: Free resources for RX queue
1320  * @rx_write: Write RX descriptors and doorbell
1321  * @rx_defer_refill: Generate a refill reminder event
1322  * @rx_packet: Receive the queued RX buffer on a channel
1323  * @rx_buf_hash_valid: Determine whether the RX prefix contains a valid hash
1324  * @ev_probe: Allocate resources for event queue
1325  * @ev_init: Initialise event queue on the NIC
1326  * @ev_fini: Deinitialise event queue on the NIC
1327  * @ev_remove: Free resources for event queue
1328  * @ev_process: Process events for a queue, up to the given NAPI quota
1329  * @ev_read_ack: Acknowledge read events on a queue, rearming its IRQ
1330  * @ev_test_generate: Generate a test event
1331  * @filter_table_probe: Probe filter capabilities and set up filter software state
1332  * @filter_table_restore: Restore filters removed from hardware
1333  * @filter_table_remove: Remove filters from hardware and tear down software state
1334  * @filter_update_rx_scatter: Update filters after change to rx scatter setting
1335  * @filter_insert: add or replace a filter
1336  * @filter_remove_safe: remove a filter by ID, carefully
1337  * @filter_get_safe: retrieve a filter by ID, carefully
1338  * @filter_clear_rx: Remove all RX filters whose priority is less than or
1339  *	equal to the given priority and is not %EFX_FILTER_PRI_AUTO
1340  * @filter_count_rx_used: Get the number of filters in use at a given priority
1341  * @filter_get_rx_id_limit: Get maximum value of a filter id, plus 1
1342  * @filter_get_rx_ids: Get list of RX filters at a given priority
1343  * @filter_rfs_expire_one: Consider expiring a filter inserted for RFS.
1344  *	This must check whether the specified table entry is used by RFS
1345  *	and that rps_may_expire_flow() returns true for it.
1346  * @mtd_probe: Probe and add MTD partitions associated with this net device,
1347  *	 using efx_mtd_add()
1348  * @mtd_rename: Set an MTD partition name using the net device name
1349  * @mtd_read: Read from an MTD partition
1350  * @mtd_erase: Erase part of an MTD partition
1351  * @mtd_write: Write to an MTD partition
1352  * @mtd_sync: Wait for write-back to complete on MTD partition.  This
1353  *	also notifies the driver that a writer has finished using this
1354  *	partition.
1355  * @ptp_write_host_time: Send host time to MC as part of sync protocol
1356  * @ptp_set_ts_sync_events: Enable or disable sync events for inline RX
1357  *	timestamping, possibly only temporarily for the purposes of a reset.
1358  * @ptp_set_ts_config: Set hardware timestamp configuration.  The flags
1359  *	and tx_type will already have been validated but this operation
1360  *	must validate and update rx_filter.
1361  * @get_phys_port_id: Get the underlying physical port id.
1362  * @set_mac_address: Set the MAC address of the device
1363  * @tso_versions: Returns mask of firmware-assisted TSO versions supported.
1364  *	If %NULL, then device does not support any TSO version.
1365  * @udp_tnl_push_ports: Push the list of UDP tunnel ports to the NIC if required.
1366  * @udp_tnl_has_port: Check if a port has been added as UDP tunnel
1367  * @print_additional_fwver: Dump NIC-specific additional FW version info
1368  * @sensor_event: Handle a sensor event from MCDI
1369  * @rx_recycle_ring_size: Size of the RX recycle ring
1370  * @revision: Hardware architecture revision
1371  * @txd_ptr_tbl_base: TX descriptor ring base address
1372  * @rxd_ptr_tbl_base: RX descriptor ring base address
1373  * @buf_tbl_base: Buffer table base address
1374  * @evq_ptr_tbl_base: Event queue pointer table base address
1375  * @evq_rptr_tbl_base: Event queue read-pointer table base address
1376  * @max_dma_mask: Maximum possible DMA mask
1377  * @rx_prefix_size: Size of RX prefix before packet data
1378  * @rx_hash_offset: Offset of RX flow hash within prefix
1379  * @rx_ts_offset: Offset of timestamp within prefix
1380  * @rx_buffer_padding: Size of padding at end of RX packet
1381  * @can_rx_scatter: NIC is able to scatter packets to multiple buffers
1382  * @always_rx_scatter: NIC will always scatter packets to multiple buffers
1383  * @option_descriptors: NIC supports TX option descriptors
1384  * @min_interrupt_mode: Lowest capability interrupt mode supported
1385  *	from &enum efx_int_mode.
1386  * @timer_period_max: Maximum period of interrupt timer (in ticks)
1387  * @offload_features: net_device feature flags for protocol offload
1388  *	features implemented in hardware
1389  * @mcdi_max_ver: Maximum MCDI version supported
1390  * @hwtstamp_filters: Mask of hardware timestamp filter types supported
1391  */
1392 struct efx_nic_type {
1393 	bool is_vf;
1394 	unsigned int (*mem_bar)(struct efx_nic *efx);
1395 	unsigned int (*mem_map_size)(struct efx_nic *efx);
1396 	int (*probe)(struct efx_nic *efx);
1397 	void (*remove)(struct efx_nic *efx);
1398 	int (*init)(struct efx_nic *efx);
1399 	int (*dimension_resources)(struct efx_nic *efx);
1400 	void (*fini)(struct efx_nic *efx);
1401 	void (*monitor)(struct efx_nic *efx);
1402 	enum reset_type (*map_reset_reason)(enum reset_type reason);
1403 	int (*map_reset_flags)(u32 *flags);
1404 	int (*reset)(struct efx_nic *efx, enum reset_type method);
1405 	int (*probe_port)(struct efx_nic *efx);
1406 	void (*remove_port)(struct efx_nic *efx);
1407 	bool (*handle_global_event)(struct efx_channel *channel, efx_qword_t *);
1408 	int (*fini_dmaq)(struct efx_nic *efx);
1409 	void (*prepare_flr)(struct efx_nic *efx);
1410 	void (*finish_flr)(struct efx_nic *efx);
1411 	size_t (*describe_stats)(struct efx_nic *efx, u8 **names);
1412 	size_t (*update_stats)(struct efx_nic *efx, u64 *full_stats,
1413 			       struct rtnl_link_stats64 *core_stats);
1414 	size_t (*update_stats_atomic)(struct efx_nic *efx, u64 *full_stats,
1415 				      struct rtnl_link_stats64 *core_stats);
1416 	void (*start_stats)(struct efx_nic *efx);
1417 	void (*pull_stats)(struct efx_nic *efx);
1418 	void (*stop_stats)(struct efx_nic *efx);
1419 	void (*push_irq_moderation)(struct efx_channel *channel);
1420 	int (*reconfigure_port)(struct efx_nic *efx);
1421 	void (*prepare_enable_fc_tx)(struct efx_nic *efx);
1422 	int (*reconfigure_mac)(struct efx_nic *efx, bool mtu_only);
1423 	bool (*check_mac_fault)(struct efx_nic *efx);
1424 	void (*get_wol)(struct efx_nic *efx, struct ethtool_wolinfo *wol);
1425 	int (*set_wol)(struct efx_nic *efx, u32 type);
1426 	void (*resume_wol)(struct efx_nic *efx);
1427 	void (*get_fec_stats)(struct efx_nic *efx,
1428 			      struct ethtool_fec_stats *fec_stats);
1429 	unsigned int (*check_caps)(const struct efx_nic *efx,
1430 				   u8 flag,
1431 				   u32 offset);
1432 	int (*test_chip)(struct efx_nic *efx, struct efx_self_tests *tests);
1433 	int (*test_nvram)(struct efx_nic *efx);
1434 	void (*mcdi_request)(struct efx_nic *efx,
1435 			     const efx_dword_t *hdr, size_t hdr_len,
1436 			     const efx_dword_t *sdu, size_t sdu_len);
1437 	bool (*mcdi_poll_response)(struct efx_nic *efx);
1438 	void (*mcdi_read_response)(struct efx_nic *efx, efx_dword_t *pdu,
1439 				   size_t pdu_offset, size_t pdu_len);
1440 	int (*mcdi_poll_reboot)(struct efx_nic *efx);
1441 	void (*mcdi_reboot_detected)(struct efx_nic *efx);
1442 	void (*irq_enable_master)(struct efx_nic *efx);
1443 	int (*irq_test_generate)(struct efx_nic *efx);
1444 	void (*irq_disable_non_ev)(struct efx_nic *efx);
1445 	irqreturn_t (*irq_handle_msi)(int irq, void *dev_id);
1446 	irqreturn_t (*irq_handle_legacy)(int irq, void *dev_id);
1447 	int (*tx_probe)(struct efx_tx_queue *tx_queue);
1448 	void (*tx_init)(struct efx_tx_queue *tx_queue);
1449 	void (*tx_remove)(struct efx_tx_queue *tx_queue);
1450 	void (*tx_write)(struct efx_tx_queue *tx_queue);
1451 	netdev_tx_t (*tx_enqueue)(struct efx_tx_queue *tx_queue, struct sk_buff *skb);
1452 	unsigned int (*tx_limit_len)(struct efx_tx_queue *tx_queue,
1453 				     dma_addr_t dma_addr, unsigned int len);
1454 	int (*rx_push_rss_config)(struct efx_nic *efx, bool user,
1455 				  const u32 *rx_indir_table, const u8 *key);
1456 	int (*rx_pull_rss_config)(struct efx_nic *efx);
1457 	int (*rx_push_rss_context_config)(struct efx_nic *efx,
1458 					  struct efx_rss_context_priv *ctx,
1459 					  const u32 *rx_indir_table,
1460 					  const u8 *key, bool delete);
1461 	int (*rx_pull_rss_context_config)(struct efx_nic *efx,
1462 					  struct efx_rss_context *ctx);
1463 	void (*rx_restore_rss_contexts)(struct efx_nic *efx);
1464 	int (*rx_probe)(struct efx_rx_queue *rx_queue);
1465 	void (*rx_init)(struct efx_rx_queue *rx_queue);
1466 	void (*rx_remove)(struct efx_rx_queue *rx_queue);
1467 	void (*rx_write)(struct efx_rx_queue *rx_queue);
1468 	void (*rx_defer_refill)(struct efx_rx_queue *rx_queue);
1469 	void (*rx_packet)(struct efx_channel *channel);
1470 	bool (*rx_buf_hash_valid)(const u8 *prefix);
1471 	int (*ev_probe)(struct efx_channel *channel);
1472 	int (*ev_init)(struct efx_channel *channel);
1473 	void (*ev_fini)(struct efx_channel *channel);
1474 	void (*ev_remove)(struct efx_channel *channel);
1475 	int (*ev_process)(struct efx_channel *channel, int quota);
1476 	void (*ev_read_ack)(struct efx_channel *channel);
1477 	void (*ev_test_generate)(struct efx_channel *channel);
1478 	int (*filter_table_probe)(struct efx_nic *efx);
1479 	void (*filter_table_restore)(struct efx_nic *efx);
1480 	void (*filter_table_remove)(struct efx_nic *efx);
1481 	void (*filter_update_rx_scatter)(struct efx_nic *efx);
1482 	s32 (*filter_insert)(struct efx_nic *efx,
1483 			     struct efx_filter_spec *spec, bool replace);
1484 	int (*filter_remove_safe)(struct efx_nic *efx,
1485 				  enum efx_filter_priority priority,
1486 				  u32 filter_id);
1487 	int (*filter_get_safe)(struct efx_nic *efx,
1488 			       enum efx_filter_priority priority,
1489 			       u32 filter_id, struct efx_filter_spec *);
1490 	int (*filter_clear_rx)(struct efx_nic *efx,
1491 			       enum efx_filter_priority priority);
1492 	u32 (*filter_count_rx_used)(struct efx_nic *efx,
1493 				    enum efx_filter_priority priority);
1494 	u32 (*filter_get_rx_id_limit)(struct efx_nic *efx);
1495 	s32 (*filter_get_rx_ids)(struct efx_nic *efx,
1496 				 enum efx_filter_priority priority,
1497 				 u32 *buf, u32 size);
1498 #ifdef CONFIG_RFS_ACCEL
1499 	bool (*filter_rfs_expire_one)(struct efx_nic *efx, u32 flow_id,
1500 				      unsigned int index);
1501 #endif
1502 #ifdef CONFIG_SFC_MTD
1503 	int (*mtd_probe)(struct efx_nic *efx);
1504 	void (*mtd_rename)(struct efx_mtd_partition *part);
1505 	int (*mtd_read)(struct mtd_info *mtd, loff_t start, size_t len,
1506 			size_t *retlen, u8 *buffer);
1507 	int (*mtd_erase)(struct mtd_info *mtd, loff_t start, size_t len);
1508 	int (*mtd_write)(struct mtd_info *mtd, loff_t start, size_t len,
1509 			 size_t *retlen, const u8 *buffer);
1510 	int (*mtd_sync)(struct mtd_info *mtd);
1511 #endif
1512 	void (*ptp_write_host_time)(struct efx_nic *efx, u32 host_time);
1513 	int (*ptp_set_ts_sync_events)(struct efx_nic *efx, bool en, bool temp);
1514 	int (*ptp_set_ts_config)(struct efx_nic *efx,
1515 				 struct kernel_hwtstamp_config *init);
1516 	int (*sriov_configure)(struct efx_nic *efx, int num_vfs);
1517 	int (*vlan_rx_add_vid)(struct efx_nic *efx, __be16 proto, u16 vid);
1518 	int (*vlan_rx_kill_vid)(struct efx_nic *efx, __be16 proto, u16 vid);
1519 	int (*get_phys_port_id)(struct efx_nic *efx,
1520 				struct netdev_phys_item_id *ppid);
1521 	int (*sriov_init)(struct efx_nic *efx);
1522 	void (*sriov_fini)(struct efx_nic *efx);
1523 	bool (*sriov_wanted)(struct efx_nic *efx);
1524 	int (*sriov_set_vf_mac)(struct efx_nic *efx, int vf_i, const u8 *mac);
1525 	int (*sriov_set_vf_vlan)(struct efx_nic *efx, int vf_i, u16 vlan,
1526 				 u8 qos);
1527 	int (*sriov_set_vf_spoofchk)(struct efx_nic *efx, int vf_i,
1528 				     bool spoofchk);
1529 	int (*sriov_get_vf_config)(struct efx_nic *efx, int vf_i,
1530 				   struct ifla_vf_info *ivi);
1531 	int (*sriov_set_vf_link_state)(struct efx_nic *efx, int vf_i,
1532 				       int link_state);
1533 	int (*vswitching_probe)(struct efx_nic *efx);
1534 	int (*vswitching_restore)(struct efx_nic *efx);
1535 	void (*vswitching_remove)(struct efx_nic *efx);
1536 	int (*get_mac_address)(struct efx_nic *efx, unsigned char *perm_addr);
1537 	int (*set_mac_address)(struct efx_nic *efx);
1538 	u32 (*tso_versions)(struct efx_nic *efx);
1539 	int (*udp_tnl_push_ports)(struct efx_nic *efx);
1540 	bool (*udp_tnl_has_port)(struct efx_nic *efx, __be16 port);
1541 	size_t (*print_additional_fwver)(struct efx_nic *efx, char *buf,
1542 					 size_t len);
1543 	void (*sensor_event)(struct efx_nic *efx, efx_qword_t *ev);
1544 	unsigned int (*rx_recycle_ring_size)(const struct efx_nic *efx);
1545 
1546 	int revision;
1547 	unsigned int txd_ptr_tbl_base;
1548 	unsigned int rxd_ptr_tbl_base;
1549 	unsigned int buf_tbl_base;
1550 	unsigned int evq_ptr_tbl_base;
1551 	unsigned int evq_rptr_tbl_base;
1552 	u64 max_dma_mask;
1553 	unsigned int rx_prefix_size;
1554 	unsigned int rx_hash_offset;
1555 	unsigned int rx_ts_offset;
1556 	unsigned int rx_buffer_padding;
1557 	bool can_rx_scatter;
1558 	bool always_rx_scatter;
1559 	bool option_descriptors;
1560 	unsigned int min_interrupt_mode;
1561 	unsigned int timer_period_max;
1562 	netdev_features_t offload_features;
1563 	int mcdi_max_ver;
1564 	unsigned int max_rx_ip_filters;
1565 	u32 hwtstamp_filters;
1566 	unsigned int rx_hash_key_size;
1567 };
1568 
1569 /**************************************************************************
1570  *
1571  * Prototypes and inline functions
1572  *
1573  *************************************************************************/
1574 
1575 static inline struct efx_channel *
1576 efx_get_channel(struct efx_nic *efx, unsigned index)
1577 {
1578 	EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_channels);
1579 	return efx->channel[index];
1580 }
1581 
1582 /* Iterate over all used channels */
1583 #define efx_for_each_channel(_channel, _efx)				\
1584 	for (_channel = (_efx)->channel[0];				\
1585 	     _channel;							\
1586 	     _channel = (_channel->channel + 1 < (_efx)->n_channels) ?	\
1587 		     (_efx)->channel[_channel->channel + 1] : NULL)
1588 
1589 /* Iterate over all used channels in reverse */
1590 #define efx_for_each_channel_rev(_channel, _efx)			\
1591 	for (_channel = (_efx)->channel[(_efx)->n_channels - 1];	\
1592 	     _channel;							\
1593 	     _channel = _channel->channel ?				\
1594 		     (_efx)->channel[_channel->channel - 1] : NULL)
1595 
1596 static inline struct efx_channel *
1597 efx_get_tx_channel(struct efx_nic *efx, unsigned int index)
1598 {
1599 	EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_tx_channels);
1600 	return efx->channel[efx->tx_channel_offset + index];
1601 }
1602 
1603 static inline struct efx_channel *
1604 efx_get_xdp_channel(struct efx_nic *efx, unsigned int index)
1605 {
1606 	EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_xdp_channels);
1607 	return efx->channel[efx->xdp_channel_offset + index];
1608 }
1609 
1610 static inline bool efx_channel_is_xdp_tx(struct efx_channel *channel)
1611 {
1612 	return channel->channel - channel->efx->xdp_channel_offset <
1613 	       channel->efx->n_xdp_channels;
1614 }
1615 
1616 static inline bool efx_channel_has_tx_queues(struct efx_channel *channel)
1617 {
1618 	return channel && channel->channel >= channel->efx->tx_channel_offset;
1619 }
1620 
1621 static inline unsigned int efx_channel_num_tx_queues(struct efx_channel *channel)
1622 {
1623 	if (efx_channel_is_xdp_tx(channel))
1624 		return channel->efx->xdp_tx_per_channel;
1625 	return channel->efx->tx_queues_per_channel;
1626 }
1627 
1628 static inline struct efx_tx_queue *
1629 efx_channel_get_tx_queue(struct efx_channel *channel, unsigned int type)
1630 {
1631 	EFX_WARN_ON_ONCE_PARANOID(type >= EFX_TXQ_TYPES);
1632 	return channel->tx_queue_by_type[type];
1633 }
1634 
1635 static inline struct efx_tx_queue *
1636 efx_get_tx_queue(struct efx_nic *efx, unsigned int index, unsigned int type)
1637 {
1638 	struct efx_channel *channel = efx_get_tx_channel(efx, index);
1639 
1640 	return efx_channel_get_tx_queue(channel, type);
1641 }
1642 
1643 /* Iterate over all TX queues belonging to a channel */
1644 #define efx_for_each_channel_tx_queue(_tx_queue, _channel)		\
1645 	if (!efx_channel_has_tx_queues(_channel))			\
1646 		;							\
1647 	else								\
1648 		for (_tx_queue = (_channel)->tx_queue;			\
1649 		     _tx_queue < (_channel)->tx_queue +			\
1650 				 efx_channel_num_tx_queues(_channel);		\
1651 		     _tx_queue++)
1652 
1653 static inline bool efx_channel_has_rx_queue(struct efx_channel *channel)
1654 {
1655 	return channel->rx_queue.core_index >= 0;
1656 }
1657 
1658 static inline struct efx_rx_queue *
1659 efx_channel_get_rx_queue(struct efx_channel *channel)
1660 {
1661 	EFX_WARN_ON_ONCE_PARANOID(!efx_channel_has_rx_queue(channel));
1662 	return &channel->rx_queue;
1663 }
1664 
1665 /* Iterate over all RX queues belonging to a channel */
1666 #define efx_for_each_channel_rx_queue(_rx_queue, _channel)		\
1667 	if (!efx_channel_has_rx_queue(_channel))			\
1668 		;							\
1669 	else								\
1670 		for (_rx_queue = &(_channel)->rx_queue;			\
1671 		     _rx_queue;						\
1672 		     _rx_queue = NULL)
1673 
1674 static inline struct efx_channel *
1675 efx_rx_queue_channel(struct efx_rx_queue *rx_queue)
1676 {
1677 	return container_of(rx_queue, struct efx_channel, rx_queue);
1678 }
1679 
1680 static inline int efx_rx_queue_index(struct efx_rx_queue *rx_queue)
1681 {
1682 	return efx_rx_queue_channel(rx_queue)->channel;
1683 }
1684 
1685 /* Returns a pointer to the specified receive buffer in the RX
1686  * descriptor queue.
1687  */
1688 static inline struct efx_rx_buffer *efx_rx_buffer(struct efx_rx_queue *rx_queue,
1689 						  unsigned int index)
1690 {
1691 	return &rx_queue->buffer[index];
1692 }
1693 
1694 static inline struct efx_rx_buffer *
1695 efx_rx_buf_next(struct efx_rx_queue *rx_queue, struct efx_rx_buffer *rx_buf)
1696 {
1697 	if (unlikely(rx_buf == efx_rx_buffer(rx_queue, rx_queue->ptr_mask)))
1698 		return efx_rx_buffer(rx_queue, 0);
1699 	else
1700 		return rx_buf + 1;
1701 }
1702 
1703 /**
1704  * EFX_MAX_FRAME_LEN - calculate maximum frame length
1705  *
1706  * This calculates the maximum frame length that will be used for a
1707  * given MTU.  The frame length will be equal to the MTU plus a
1708  * constant amount of header space and padding.  This is the quantity
1709  * that the net driver will program into the MAC as the maximum frame
1710  * length.
1711  *
1712  * The 10G MAC requires 8-byte alignment on the frame
1713  * length, so we round up to the nearest 8.
1714  *
1715  * Re-clocking by the XGXS on RX can reduce an IPG to 32 bits (half an
1716  * XGMII cycle).  If the frame length reaches the maximum value in the
1717  * same cycle, the XMAC can miss the IPG altogether.  We work around
1718  * this by adding a further 16 bytes.
1719  */
1720 #define EFX_FRAME_PAD	16
1721 #define EFX_MAX_FRAME_LEN(mtu) \
1722 	(ALIGN(((mtu) + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN + EFX_FRAME_PAD), 8))
1723 
1724 static inline bool efx_xmit_with_hwtstamp(struct sk_buff *skb)
1725 {
1726 	return skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP;
1727 }
1728 static inline void efx_xmit_hwtstamp_pending(struct sk_buff *skb)
1729 {
1730 	skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
1731 }
1732 
1733 /* Get the max fill level of the TX queues on this channel */
1734 static inline unsigned int
1735 efx_channel_tx_fill_level(struct efx_channel *channel)
1736 {
1737 	struct efx_tx_queue *tx_queue;
1738 	unsigned int fill_level = 0;
1739 
1740 	efx_for_each_channel_tx_queue(tx_queue, channel)
1741 		fill_level = max(fill_level,
1742 				 tx_queue->insert_count - tx_queue->read_count);
1743 
1744 	return fill_level;
1745 }
1746 
1747 /* Conservative approximation of efx_channel_tx_fill_level using cached value */
1748 static inline unsigned int
1749 efx_channel_tx_old_fill_level(struct efx_channel *channel)
1750 {
1751 	struct efx_tx_queue *tx_queue;
1752 	unsigned int fill_level = 0;
1753 
1754 	efx_for_each_channel_tx_queue(tx_queue, channel)
1755 		fill_level = max(fill_level,
1756 				 tx_queue->insert_count - tx_queue->old_read_count);
1757 
1758 	return fill_level;
1759 }
1760 
1761 /* Get all supported features.
1762  * If a feature is not fixed, it is present in hw_features.
1763  * If a feature is fixed, it does not present in hw_features, but
1764  * always in features.
1765  */
1766 static inline netdev_features_t efx_supported_features(const struct efx_nic *efx)
1767 {
1768 	const struct net_device *net_dev = efx->net_dev;
1769 
1770 	return net_dev->features | net_dev->hw_features;
1771 }
1772 
1773 /* Get the current TX queue insert index. */
1774 static inline unsigned int
1775 efx_tx_queue_get_insert_index(const struct efx_tx_queue *tx_queue)
1776 {
1777 	return tx_queue->insert_count & tx_queue->ptr_mask;
1778 }
1779 
1780 /* Get a TX buffer. */
1781 static inline struct efx_tx_buffer *
1782 __efx_tx_queue_get_insert_buffer(const struct efx_tx_queue *tx_queue)
1783 {
1784 	return &tx_queue->buffer[efx_tx_queue_get_insert_index(tx_queue)];
1785 }
1786 
1787 /* Get a TX buffer, checking it's not currently in use. */
1788 static inline struct efx_tx_buffer *
1789 efx_tx_queue_get_insert_buffer(const struct efx_tx_queue *tx_queue)
1790 {
1791 	struct efx_tx_buffer *buffer =
1792 		__efx_tx_queue_get_insert_buffer(tx_queue);
1793 
1794 	EFX_WARN_ON_ONCE_PARANOID(buffer->len);
1795 	EFX_WARN_ON_ONCE_PARANOID(buffer->flags);
1796 	EFX_WARN_ON_ONCE_PARANOID(buffer->unmap_len);
1797 
1798 	return buffer;
1799 }
1800 
1801 #endif /* EFX_NET_DRIVER_H */
1802