xref: /linux/drivers/net/ethernet/sfc/net_driver.h (revision 23c48a124b469cee2eb0c75e6d22d366d1caa118)
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 
31 #include "enum.h"
32 #include "bitfield.h"
33 #include "filter.h"
34 
35 /**************************************************************************
36  *
37  * Build definitions
38  *
39  **************************************************************************/
40 
41 #ifdef DEBUG
42 #define EFX_WARN_ON_ONCE_PARANOID(x) WARN_ON_ONCE(x)
43 #define EFX_WARN_ON_PARANOID(x) WARN_ON(x)
44 #else
45 #define EFX_WARN_ON_ONCE_PARANOID(x) do {} while (0)
46 #define EFX_WARN_ON_PARANOID(x) do {} while (0)
47 #endif
48 
49 /**************************************************************************
50  *
51  * Efx data structures
52  *
53  **************************************************************************/
54 
55 #define EFX_MAX_CHANNELS 32U
56 #define EFX_MAX_RX_QUEUES EFX_MAX_CHANNELS
57 #define EFX_EXTRA_CHANNEL_IOV	0
58 #define EFX_EXTRA_CHANNEL_PTP	1
59 #define EFX_MAX_EXTRA_CHANNELS	2U
60 
61 /* Checksum generation is a per-queue option in hardware, so each
62  * queue visible to the networking core is backed by two hardware TX
63  * queues. */
64 #define EFX_MAX_TX_TC		2
65 #define EFX_MAX_CORE_TX_QUEUES	(EFX_MAX_TX_TC * EFX_MAX_CHANNELS)
66 #define EFX_TXQ_TYPE_OUTER_CSUM	1	/* Outer checksum offload */
67 #define EFX_TXQ_TYPE_INNER_CSUM	2	/* Inner checksum offload */
68 #define EFX_TXQ_TYPE_HIGHPRI	4	/* High-priority (for TC) */
69 #define EFX_TXQ_TYPES		8
70 /* HIGHPRI is Siena-only, and INNER_CSUM is EF10, so no need for both */
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_special_buffer - DMA buffer entered into buffer table
127  * @buf: Standard &struct efx_buffer
128  * @index: Buffer index within controller;s buffer table
129  * @entries: Number of buffer table entries
130  *
131  * The NIC has a buffer table that maps buffers of size %EFX_BUF_SIZE.
132  * Event and descriptor rings are addressed via one or more buffer
133  * table entries (and so can be physically non-contiguous, although we
134  * currently do not take advantage of that).  On Falcon and Siena we
135  * have to take care of allocating and initialising the entries
136  * ourselves.  On later hardware this is managed by the firmware and
137  * @index and @entries are left as 0.
138  */
139 struct efx_special_buffer {
140 	struct efx_buffer buf;
141 	unsigned int index;
142 	unsigned int entries;
143 };
144 
145 /**
146  * struct efx_tx_buffer - buffer state for a TX descriptor
147  * @skb: When @flags & %EFX_TX_BUF_SKB, the associated socket buffer to be
148  *	freed when descriptor completes
149  * @xdpf: When @flags & %EFX_TX_BUF_XDP, the XDP frame information; its @data
150  *	member is the associated buffer to drop a page reference on.
151  * @option: When @flags & %EFX_TX_BUF_OPTION, an EF10-specific option
152  *	descriptor.
153  * @dma_addr: DMA address of the fragment.
154  * @flags: Flags for allocation and DMA mapping type
155  * @len: Length of this fragment.
156  *	This field is zero when the queue slot is empty.
157  * @unmap_len: Length of this fragment to unmap
158  * @dma_offset: Offset of @dma_addr from the address of the backing DMA mapping.
159  * Only valid if @unmap_len != 0.
160  */
161 struct efx_tx_buffer {
162 	union {
163 		const struct sk_buff *skb;
164 		struct xdp_frame *xdpf;
165 	};
166 	union {
167 		efx_qword_t option;    /* EF10 */
168 		dma_addr_t dma_addr;
169 	};
170 	unsigned short flags;
171 	unsigned short len;
172 	unsigned short unmap_len;
173 	unsigned short dma_offset;
174 };
175 #define EFX_TX_BUF_CONT		1	/* not last descriptor of packet */
176 #define EFX_TX_BUF_SKB		2	/* buffer is last part of skb */
177 #define EFX_TX_BUF_MAP_SINGLE	8	/* buffer was mapped with dma_map_single() */
178 #define EFX_TX_BUF_OPTION	0x10	/* empty buffer for option descriptor */
179 #define EFX_TX_BUF_XDP		0x20	/* buffer was sent with XDP */
180 #define EFX_TX_BUF_TSO_V3	0x40	/* empty buffer for a TSO_V3 descriptor */
181 #define EFX_TX_BUF_EFV		0x100	/* buffer was sent from representor */
182 
183 /**
184  * struct efx_tx_queue - An Efx TX queue
185  *
186  * This is a ring buffer of TX fragments.
187  * Since the TX completion path always executes on the same
188  * CPU and the xmit path can operate on different CPUs,
189  * performance is increased by ensuring that the completion
190  * path and the xmit path operate on different cache lines.
191  * This is particularly important if the xmit path is always
192  * executing on one CPU which is different from the completion
193  * path.  There is also a cache line for members which are
194  * read but not written on the fast path.
195  *
196  * @efx: The associated Efx NIC
197  * @queue: DMA queue number
198  * @label: Label for TX completion events.
199  *	Is our index within @channel->tx_queue array.
200  * @type: configuration type of this TX queue.  A bitmask of %EFX_TXQ_TYPE_* flags.
201  * @tso_version: Version of TSO in use for this queue.
202  * @tso_encap: Is encapsulated TSO supported? Supported in TSOv2 on 8000 series.
203  * @channel: The associated channel
204  * @core_txq: The networking core TX queue structure
205  * @buffer: The software buffer ring
206  * @cb_page: Array of pages of copy buffers.  Carved up according to
207  *	%EFX_TX_CB_ORDER into %EFX_TX_CB_SIZE-sized chunks.
208  * @txd: The hardware descriptor ring
209  * @ptr_mask: The size of the ring minus 1.
210  * @piobuf: PIO buffer region for this TX queue (shared with its partner).
211  *	Size of the region is efx_piobuf_size.
212  * @piobuf_offset: Buffer offset to be specified in PIO descriptors
213  * @initialised: Has hardware queue been initialised?
214  * @timestamping: Is timestamping enabled for this channel?
215  * @xdp_tx: Is this an XDP tx queue?
216  * @read_count: Current read pointer.
217  *	This is the number of buffers that have been removed from both rings.
218  * @old_write_count: The value of @write_count when last checked.
219  *	This is here for performance reasons.  The xmit path will
220  *	only get the up-to-date value of @write_count if this
221  *	variable indicates that the queue is empty.  This is to
222  *	avoid cache-line ping-pong between the xmit path and the
223  *	completion path.
224  * @merge_events: Number of TX merged completion events
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 written and used on EF10, and neither on farch.
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  * @empty_read_count: If the completion path has seen the queue as empty
256  *	and the transmission path has not yet checked this, the value of
257  *	@read_count bitwise-added to %EFX_EMPTY_COUNT_VALID; otherwise 0.
258  */
259 struct efx_tx_queue {
260 	/* Members which don't change on the fast path */
261 	struct efx_nic *efx ____cacheline_aligned_in_smp;
262 	unsigned int queue;
263 	unsigned int label;
264 	unsigned int type;
265 	unsigned int tso_version;
266 	bool tso_encap;
267 	struct efx_channel *channel;
268 	struct netdev_queue *core_txq;
269 	struct efx_tx_buffer *buffer;
270 	struct efx_buffer *cb_page;
271 	struct efx_special_buffer txd;
272 	unsigned int ptr_mask;
273 	void __iomem *piobuf;
274 	unsigned int piobuf_offset;
275 	bool initialised;
276 	bool timestamping;
277 	bool xdp_tx;
278 
279 	/* Members used mainly on the completion path */
280 	unsigned int read_count ____cacheline_aligned_in_smp;
281 	unsigned int old_write_count;
282 	unsigned int merge_events;
283 	unsigned int bytes_compl;
284 	unsigned int pkts_compl;
285 	u32 completed_timestamp_major;
286 	u32 completed_timestamp_minor;
287 
288 	/* Members used only on the xmit path */
289 	unsigned int insert_count ____cacheline_aligned_in_smp;
290 	unsigned int write_count;
291 	unsigned int packet_write_count;
292 	unsigned int old_read_count;
293 	unsigned int tso_bursts;
294 	unsigned int tso_long_headers;
295 	unsigned int tso_packets;
296 	unsigned int tso_fallbacks;
297 	unsigned int pushes;
298 	unsigned int pio_packets;
299 	bool xmit_pending;
300 	unsigned int cb_packets;
301 	unsigned int notify_count;
302 	/* Statistics to supplement MAC stats */
303 	unsigned long tx_packets;
304 
305 	/* Members shared between paths and sometimes updated */
306 	unsigned int empty_read_count ____cacheline_aligned_in_smp;
307 #define EFX_EMPTY_COUNT_VALID 0x80000000
308 	atomic_t flush_outstanding;
309 };
310 
311 #define EFX_TX_CB_ORDER	7
312 #define EFX_TX_CB_SIZE	(1 << EFX_TX_CB_ORDER) - NET_IP_ALIGN
313 
314 /**
315  * struct efx_rx_buffer - An Efx RX data buffer
316  * @dma_addr: DMA base address of the buffer
317  * @page: The associated page buffer.
318  *	Will be %NULL if the buffer slot is currently free.
319  * @page_offset: If pending: offset in @page of DMA base address.
320  *	If completed: offset in @page of Ethernet header.
321  * @len: If pending: length for DMA descriptor.
322  *	If completed: received length, excluding hash prefix.
323  * @flags: Flags for buffer and packet state.  These are only set on the
324  *	first buffer of a scattered packet.
325  */
326 struct efx_rx_buffer {
327 	dma_addr_t dma_addr;
328 	struct page *page;
329 	u16 page_offset;
330 	u16 len;
331 	u16 flags;
332 };
333 #define EFX_RX_BUF_LAST_IN_PAGE	0x0001
334 #define EFX_RX_PKT_CSUMMED	0x0002
335 #define EFX_RX_PKT_DISCARD	0x0004
336 #define EFX_RX_PKT_TCP		0x0040
337 #define EFX_RX_PKT_PREFIX_LEN	0x0080	/* length is in prefix only */
338 #define EFX_RX_PKT_CSUM_LEVEL	0x0200
339 
340 /**
341  * struct efx_rx_page_state - Page-based rx buffer state
342  *
343  * Inserted at the start of every page allocated for receive buffers.
344  * Used to facilitate sharing dma mappings between recycled rx buffers
345  * and those passed up to the kernel.
346  *
347  * @dma_addr: The dma address of this page.
348  */
349 struct efx_rx_page_state {
350 	dma_addr_t dma_addr;
351 
352 	unsigned int __pad[] ____cacheline_aligned;
353 };
354 
355 /**
356  * struct efx_rx_queue - An Efx RX queue
357  * @efx: The associated Efx NIC
358  * @core_index:  Index of network core RX queue.  Will be >= 0 iff this
359  *	is associated with a real RX queue.
360  * @buffer: The software buffer ring
361  * @rxd: The hardware descriptor ring
362  * @ptr_mask: The size of the ring minus 1.
363  * @refill_enabled: Enable refill whenever fill level is low
364  * @flush_pending: Set when a RX flush is pending. Has the same lifetime as
365  *	@rxq_flush_pending.
366  * @added_count: Number of buffers added to the receive queue.
367  * @notified_count: Number of buffers given to NIC (<= @added_count).
368  * @removed_count: Number of buffers removed from the receive queue.
369  * @scatter_n: Used by NIC specific receive code.
370  * @scatter_len: Used by NIC specific receive code.
371  * @page_ring: The ring to store DMA mapped pages for reuse.
372  * @page_add: Counter to calculate the write pointer for the recycle ring.
373  * @page_remove: Counter to calculate the read pointer for the recycle ring.
374  * @page_recycle_count: The number of pages that have been recycled.
375  * @page_recycle_failed: The number of pages that couldn't be recycled because
376  *      the kernel still held a reference to them.
377  * @page_recycle_full: The number of pages that were released because the
378  *      recycle ring was full.
379  * @page_ptr_mask: The number of pages in the RX recycle ring minus 1.
380  * @max_fill: RX descriptor maximum fill level (<= ring size)
381  * @fast_fill_trigger: RX descriptor fill level that will trigger a fast fill
382  *	(<= @max_fill)
383  * @min_fill: RX descriptor minimum non-zero fill level.
384  *	This records the minimum fill level observed when a ring
385  *	refill was triggered.
386  * @recycle_count: RX buffer recycle counter.
387  * @slow_fill: Timer used to defer efx_nic_generate_fill_event().
388  * @xdp_rxq_info: XDP specific RX queue information.
389  * @xdp_rxq_info_valid: Is xdp_rxq_info valid data?.
390  */
391 struct efx_rx_queue {
392 	struct efx_nic *efx;
393 	int core_index;
394 	struct efx_rx_buffer *buffer;
395 	struct efx_special_buffer rxd;
396 	unsigned int ptr_mask;
397 	bool refill_enabled;
398 	bool flush_pending;
399 
400 	unsigned int added_count;
401 	unsigned int notified_count;
402 	unsigned int removed_count;
403 	unsigned int scatter_n;
404 	unsigned int scatter_len;
405 	struct page **page_ring;
406 	unsigned int page_add;
407 	unsigned int page_remove;
408 	unsigned int page_recycle_count;
409 	unsigned int page_recycle_failed;
410 	unsigned int page_recycle_full;
411 	unsigned int page_ptr_mask;
412 	unsigned int max_fill;
413 	unsigned int fast_fill_trigger;
414 	unsigned int min_fill;
415 	unsigned int min_overfill;
416 	unsigned int recycle_count;
417 	struct timer_list slow_fill;
418 	unsigned int slow_fill_count;
419 	/* Statistics to supplement MAC stats */
420 	unsigned long rx_packets;
421 	struct xdp_rxq_info xdp_rxq_info;
422 	bool xdp_rxq_info_valid;
423 };
424 
425 enum efx_sync_events_state {
426 	SYNC_EVENTS_DISABLED = 0,
427 	SYNC_EVENTS_QUIESCENT,
428 	SYNC_EVENTS_REQUESTED,
429 	SYNC_EVENTS_VALID,
430 };
431 
432 /**
433  * struct efx_channel - An Efx channel
434  *
435  * A channel comprises an event queue, at least one TX queue, at least
436  * one RX queue, and an associated tasklet for processing the event
437  * queue.
438  *
439  * @efx: Associated Efx NIC
440  * @channel: Channel instance number
441  * @type: Channel type definition
442  * @eventq_init: Event queue initialised flag
443  * @enabled: Channel enabled indicator
444  * @irq: IRQ number (MSI and MSI-X only)
445  * @irq_moderation_us: IRQ moderation value (in microseconds)
446  * @napi_dev: Net device used with NAPI
447  * @napi_str: NAPI control structure
448  * @state: state for NAPI vs busy polling
449  * @state_lock: lock protecting @state
450  * @eventq: Event queue buffer
451  * @eventq_mask: Event queue pointer mask
452  * @eventq_read_ptr: Event queue read pointer
453  * @event_test_cpu: Last CPU to handle interrupt or test event for this channel
454  * @irq_count: Number of IRQs since last adaptive moderation decision
455  * @irq_mod_score: IRQ moderation score
456  * @rfs_filter_count: number of accelerated RFS filters currently in place;
457  *	equals the count of @rps_flow_id slots filled
458  * @rfs_last_expiry: value of jiffies last time some accelerated RFS filters
459  *	were checked for expiry
460  * @rfs_expire_index: next accelerated RFS filter ID to check for expiry
461  * @n_rfs_succeeded: number of successful accelerated RFS filter insertions
462  * @n_rfs_failed: number of failed accelerated RFS filter insertions
463  * @filter_work: Work item for efx_filter_rfs_expire()
464  * @rps_flow_id: Flow IDs of filters allocated for accelerated RFS,
465  *      indexed by filter ID
466  * @n_rx_tobe_disc: Count of RX_TOBE_DISC errors
467  * @n_rx_ip_hdr_chksum_err: Count of RX IP header checksum errors
468  * @n_rx_tcp_udp_chksum_err: Count of RX TCP and UDP checksum errors
469  * @n_rx_mcast_mismatch: Count of unmatched multicast frames
470  * @n_rx_frm_trunc: Count of RX_FRM_TRUNC errors
471  * @n_rx_overlength: Count of RX_OVERLENGTH errors
472  * @n_skbuff_leaks: Count of skbuffs leaked due to RX overrun
473  * @n_rx_nodesc_trunc: Number of RX packets truncated and then dropped due to
474  *	lack of descriptors
475  * @n_rx_merge_events: Number of RX merged completion events
476  * @n_rx_merge_packets: Number of RX packets completed by merged events
477  * @n_rx_xdp_drops: Count of RX packets intentionally dropped due to XDP
478  * @n_rx_xdp_bad_drops: Count of RX packets dropped due to XDP errors
479  * @n_rx_xdp_tx: Count of RX packets retransmitted due to XDP
480  * @n_rx_xdp_redirect: Count of RX packets redirected to a different NIC by XDP
481  * @n_rx_mport_bad: Count of RX packets dropped because their ingress mport was
482  *	not recognised
483  * @rx_pkt_n_frags: Number of fragments in next packet to be delivered by
484  *	__efx_rx_packet(), or zero if there is none
485  * @rx_pkt_index: Ring index of first buffer for next packet to be delivered
486  *	by __efx_rx_packet(), if @rx_pkt_n_frags != 0
487  * @rx_list: list of SKBs from current RX, awaiting processing
488  * @rx_queue: RX queue for this channel
489  * @tx_queue: TX queues for this channel
490  * @tx_queue_by_type: pointers into @tx_queue, or %NULL, indexed by txq type
491  * @sync_events_state: Current state of sync events on this channel
492  * @sync_timestamp_major: Major part of the last ptp sync event
493  * @sync_timestamp_minor: Minor part of the last ptp sync event
494  */
495 struct efx_channel {
496 	struct efx_nic *efx;
497 	int channel;
498 	const struct efx_channel_type *type;
499 	bool eventq_init;
500 	bool enabled;
501 	int irq;
502 	unsigned int irq_moderation_us;
503 	struct net_device *napi_dev;
504 	struct napi_struct napi_str;
505 #ifdef CONFIG_NET_RX_BUSY_POLL
506 	unsigned long busy_poll_state;
507 #endif
508 	struct efx_special_buffer eventq;
509 	unsigned int eventq_mask;
510 	unsigned int eventq_read_ptr;
511 	int event_test_cpu;
512 
513 	unsigned int irq_count;
514 	unsigned int irq_mod_score;
515 #ifdef CONFIG_RFS_ACCEL
516 	unsigned int rfs_filter_count;
517 	unsigned int rfs_last_expiry;
518 	unsigned int rfs_expire_index;
519 	unsigned int n_rfs_succeeded;
520 	unsigned int n_rfs_failed;
521 	struct delayed_work filter_work;
522 #define RPS_FLOW_ID_INVALID 0xFFFFFFFF
523 	u32 *rps_flow_id;
524 #endif
525 
526 	unsigned int n_rx_tobe_disc;
527 	unsigned int n_rx_ip_hdr_chksum_err;
528 	unsigned int n_rx_tcp_udp_chksum_err;
529 	unsigned int n_rx_outer_ip_hdr_chksum_err;
530 	unsigned int n_rx_outer_tcp_udp_chksum_err;
531 	unsigned int n_rx_inner_ip_hdr_chksum_err;
532 	unsigned int n_rx_inner_tcp_udp_chksum_err;
533 	unsigned int n_rx_eth_crc_err;
534 	unsigned int n_rx_mcast_mismatch;
535 	unsigned int n_rx_frm_trunc;
536 	unsigned int n_rx_overlength;
537 	unsigned int n_skbuff_leaks;
538 	unsigned int n_rx_nodesc_trunc;
539 	unsigned int n_rx_merge_events;
540 	unsigned int n_rx_merge_packets;
541 	unsigned int n_rx_xdp_drops;
542 	unsigned int n_rx_xdp_bad_drops;
543 	unsigned int n_rx_xdp_tx;
544 	unsigned int n_rx_xdp_redirect;
545 	unsigned int n_rx_mport_bad;
546 
547 	unsigned int rx_pkt_n_frags;
548 	unsigned int rx_pkt_index;
549 
550 	struct list_head *rx_list;
551 
552 	struct efx_rx_queue rx_queue;
553 	struct efx_tx_queue tx_queue[EFX_MAX_TXQ_PER_CHANNEL];
554 	struct efx_tx_queue *tx_queue_by_type[EFX_TXQ_TYPES];
555 
556 	enum efx_sync_events_state sync_events_state;
557 	u32 sync_timestamp_major;
558 	u32 sync_timestamp_minor;
559 };
560 
561 /**
562  * struct efx_msi_context - Context for each MSI
563  * @efx: The associated NIC
564  * @index: Index of the channel/IRQ
565  * @name: Name of the channel/IRQ
566  *
567  * Unlike &struct efx_channel, this is never reallocated and is always
568  * safe for the IRQ handler to access.
569  */
570 struct efx_msi_context {
571 	struct efx_nic *efx;
572 	unsigned int index;
573 	char name[IFNAMSIZ + 6];
574 };
575 
576 /**
577  * struct efx_channel_type - distinguishes traffic and extra channels
578  * @handle_no_channel: Handle failure to allocate an extra channel
579  * @pre_probe: Set up extra state prior to initialisation
580  * @post_remove: Tear down extra state after finalisation, if allocated.
581  *	May be called on channels that have not been probed.
582  * @get_name: Generate the channel's name (used for its IRQ handler)
583  * @copy: Copy the channel state prior to reallocation.  May be %NULL if
584  *	reallocation is not supported.
585  * @receive_skb: Handle an skb ready to be passed to netif_receive_skb()
586  * @want_txqs: Determine whether this channel should have TX queues
587  *	created.  If %NULL, TX queues are not created.
588  * @keep_eventq: Flag for whether event queue should be kept initialised
589  *	while the device is stopped
590  * @want_pio: Flag for whether PIO buffers should be linked to this
591  *	channel's TX queues.
592  */
593 struct efx_channel_type {
594 	void (*handle_no_channel)(struct efx_nic *);
595 	int (*pre_probe)(struct efx_channel *);
596 	void (*post_remove)(struct efx_channel *);
597 	void (*get_name)(struct efx_channel *, char *buf, size_t len);
598 	struct efx_channel *(*copy)(const struct efx_channel *);
599 	bool (*receive_skb)(struct efx_channel *, struct sk_buff *);
600 	bool (*want_txqs)(struct efx_channel *);
601 	bool keep_eventq;
602 	bool want_pio;
603 };
604 
605 enum efx_led_mode {
606 	EFX_LED_OFF	= 0,
607 	EFX_LED_ON	= 1,
608 	EFX_LED_DEFAULT	= 2
609 };
610 
611 #define STRING_TABLE_LOOKUP(val, member) \
612 	((val) < member ## _max) ? member ## _names[val] : "(invalid)"
613 
614 extern const char *const efx_loopback_mode_names[];
615 extern const unsigned int efx_loopback_mode_max;
616 #define LOOPBACK_MODE(efx) \
617 	STRING_TABLE_LOOKUP((efx)->loopback_mode, efx_loopback_mode)
618 
619 enum efx_int_mode {
620 	/* Be careful if altering to correct macro below */
621 	EFX_INT_MODE_MSIX = 0,
622 	EFX_INT_MODE_MSI = 1,
623 	EFX_INT_MODE_LEGACY = 2,
624 	EFX_INT_MODE_MAX	/* Insert any new items before this */
625 };
626 #define EFX_INT_MODE_USE_MSI(x) (((x)->interrupt_mode) <= EFX_INT_MODE_MSI)
627 
628 enum nic_state {
629 	STATE_UNINIT = 0,	/* device being probed/removed */
630 	STATE_PROBED,		/* hardware probed */
631 	STATE_NET_DOWN,		/* netdev registered */
632 	STATE_NET_UP,		/* ready for traffic */
633 	STATE_DISABLED,		/* device disabled due to hardware errors */
634 
635 	STATE_RECOVERY = 0x100,/* recovering from PCI error */
636 	STATE_FROZEN = 0x200,	/* frozen by power management */
637 };
638 
639 static inline bool efx_net_active(enum nic_state state)
640 {
641 	return state == STATE_NET_DOWN || state == STATE_NET_UP;
642 }
643 
644 static inline bool efx_frozen(enum nic_state state)
645 {
646 	return state & STATE_FROZEN;
647 }
648 
649 static inline bool efx_recovering(enum nic_state state)
650 {
651 	return state & STATE_RECOVERY;
652 }
653 
654 static inline enum nic_state efx_freeze(enum nic_state state)
655 {
656 	WARN_ON(!efx_net_active(state));
657 	return state | STATE_FROZEN;
658 }
659 
660 static inline enum nic_state efx_thaw(enum nic_state state)
661 {
662 	WARN_ON(!efx_frozen(state));
663 	return state & ~STATE_FROZEN;
664 }
665 
666 static inline enum nic_state efx_recover(enum nic_state state)
667 {
668 	WARN_ON(!efx_net_active(state));
669 	return state | STATE_RECOVERY;
670 }
671 
672 static inline enum nic_state efx_recovered(enum nic_state state)
673 {
674 	WARN_ON(!efx_recovering(state));
675 	return state & ~STATE_RECOVERY;
676 }
677 
678 /* Forward declaration */
679 struct efx_nic;
680 
681 /* Pseudo bit-mask flow control field */
682 #define EFX_FC_RX	FLOW_CTRL_RX
683 #define EFX_FC_TX	FLOW_CTRL_TX
684 #define EFX_FC_AUTO	4
685 
686 /**
687  * struct efx_link_state - Current state of the link
688  * @up: Link is up
689  * @fd: Link is full-duplex
690  * @fc: Actual flow control flags
691  * @speed: Link speed (Mbps)
692  */
693 struct efx_link_state {
694 	bool up;
695 	bool fd;
696 	u8 fc;
697 	unsigned int speed;
698 };
699 
700 static inline bool efx_link_state_equal(const struct efx_link_state *left,
701 					const struct efx_link_state *right)
702 {
703 	return left->up == right->up && left->fd == right->fd &&
704 		left->fc == right->fc && left->speed == right->speed;
705 }
706 
707 /**
708  * enum efx_phy_mode - PHY operating mode flags
709  * @PHY_MODE_NORMAL: on and should pass traffic
710  * @PHY_MODE_TX_DISABLED: on with TX disabled
711  * @PHY_MODE_LOW_POWER: set to low power through MDIO
712  * @PHY_MODE_OFF: switched off through external control
713  * @PHY_MODE_SPECIAL: on but will not pass traffic
714  */
715 enum efx_phy_mode {
716 	PHY_MODE_NORMAL		= 0,
717 	PHY_MODE_TX_DISABLED	= 1,
718 	PHY_MODE_LOW_POWER	= 2,
719 	PHY_MODE_OFF		= 4,
720 	PHY_MODE_SPECIAL	= 8,
721 };
722 
723 static inline bool efx_phy_mode_disabled(enum efx_phy_mode mode)
724 {
725 	return !!(mode & ~PHY_MODE_TX_DISABLED);
726 }
727 
728 /**
729  * struct efx_hw_stat_desc - Description of a hardware statistic
730  * @name: Name of the statistic as visible through ethtool, or %NULL if
731  *	it should not be exposed
732  * @dma_width: Width in bits (0 for non-DMA statistics)
733  * @offset: Offset within stats (ignored for non-DMA statistics)
734  */
735 struct efx_hw_stat_desc {
736 	const char *name;
737 	u16 dma_width;
738 	u16 offset;
739 };
740 
741 /* Number of bits used in a multicast filter hash address */
742 #define EFX_MCAST_HASH_BITS 8
743 
744 /* Number of (single-bit) entries in a multicast filter hash */
745 #define EFX_MCAST_HASH_ENTRIES (1 << EFX_MCAST_HASH_BITS)
746 
747 /* An Efx multicast filter hash */
748 union efx_multicast_hash {
749 	u8 byte[EFX_MCAST_HASH_ENTRIES / 8];
750 	efx_oword_t oword[EFX_MCAST_HASH_ENTRIES / sizeof(efx_oword_t) / 8];
751 };
752 
753 struct vfdi_status;
754 
755 /* The reserved RSS context value */
756 #define EFX_MCDI_RSS_CONTEXT_INVALID	0xffffffff
757 /**
758  * struct efx_rss_context - A user-defined RSS context for filtering
759  * @list: node of linked list on which this struct is stored
760  * @context_id: the RSS_CONTEXT_ID returned by MC firmware, or
761  *	%EFX_MCDI_RSS_CONTEXT_INVALID if this context is not present on the NIC.
762  *	For Siena, 0 if RSS is active, else %EFX_MCDI_RSS_CONTEXT_INVALID.
763  * @user_id: the rss_context ID exposed to userspace over ethtool.
764  * @rx_hash_udp_4tuple: UDP 4-tuple hashing enabled
765  * @rx_hash_key: Toeplitz hash key for this RSS context
766  * @indir_table: Indirection table for this RSS context
767  */
768 struct efx_rss_context {
769 	struct list_head list;
770 	u32 context_id;
771 	u32 user_id;
772 	bool rx_hash_udp_4tuple;
773 	u8 rx_hash_key[40];
774 	u32 rx_indir_table[128];
775 };
776 
777 #ifdef CONFIG_RFS_ACCEL
778 /* Order of these is important, since filter_id >= %EFX_ARFS_FILTER_ID_PENDING
779  * is used to test if filter does or will exist.
780  */
781 #define EFX_ARFS_FILTER_ID_PENDING	-1
782 #define EFX_ARFS_FILTER_ID_ERROR	-2
783 #define EFX_ARFS_FILTER_ID_REMOVING	-3
784 /**
785  * struct efx_arfs_rule - record of an ARFS filter and its IDs
786  * @node: linkage into hash table
787  * @spec: details of the filter (used as key for hash table).  Use efx->type to
788  *	determine which member to use.
789  * @rxq_index: channel to which the filter will steer traffic.
790  * @arfs_id: filter ID which was returned to ARFS
791  * @filter_id: index in software filter table.  May be
792  *	%EFX_ARFS_FILTER_ID_PENDING if filter was not inserted yet,
793  *	%EFX_ARFS_FILTER_ID_ERROR if filter insertion failed, or
794  *	%EFX_ARFS_FILTER_ID_REMOVING if expiry is currently removing the filter.
795  */
796 struct efx_arfs_rule {
797 	struct hlist_node node;
798 	struct efx_filter_spec spec;
799 	u16 rxq_index;
800 	u16 arfs_id;
801 	s32 filter_id;
802 };
803 
804 /* Size chosen so that the table is one page (4kB) */
805 #define EFX_ARFS_HASH_TABLE_SIZE	512
806 
807 /**
808  * struct efx_async_filter_insertion - Request to asynchronously insert a filter
809  * @net_dev: Reference to the netdevice
810  * @spec: The filter to insert
811  * @work: Workitem for this request
812  * @rxq_index: Identifies the channel for which this request was made
813  * @flow_id: Identifies the kernel-side flow for which this request was made
814  */
815 struct efx_async_filter_insertion {
816 	struct net_device *net_dev;
817 	struct efx_filter_spec spec;
818 	struct work_struct work;
819 	u16 rxq_index;
820 	u32 flow_id;
821 };
822 
823 /* Maximum number of ARFS workitems that may be in flight on an efx_nic */
824 #define EFX_RPS_MAX_IN_FLIGHT	8
825 #endif /* CONFIG_RFS_ACCEL */
826 
827 enum efx_xdp_tx_queues_mode {
828 	EFX_XDP_TX_QUEUES_DEDICATED,	/* one queue per core, locking not needed */
829 	EFX_XDP_TX_QUEUES_SHARED,	/* each queue used by more than 1 core */
830 	EFX_XDP_TX_QUEUES_BORROWED	/* queues borrowed from net stack */
831 };
832 
833 /**
834  * struct efx_nic - an Efx NIC
835  * @name: Device name (net device name or bus id before net device registered)
836  * @pci_dev: The PCI device
837  * @node: List node for maintaning primary/secondary function lists
838  * @primary: &struct efx_nic instance for the primary function of this
839  *	controller.  May be the same structure, and may be %NULL if no
840  *	primary function is bound.  Serialised by rtnl_lock.
841  * @secondary_list: List of &struct efx_nic instances for the secondary PCI
842  *	functions of the controller, if this is for the primary function.
843  *	Serialised by rtnl_lock.
844  * @type: Controller type attributes
845  * @legacy_irq: IRQ number
846  * @workqueue: Workqueue for port reconfigures and the HW monitor.
847  *	Work items do not hold and must not acquire RTNL.
848  * @workqueue_name: Name of workqueue
849  * @reset_work: Scheduled reset workitem
850  * @membase_phys: Memory BAR value as physical address
851  * @membase: Memory BAR value
852  * @vi_stride: step between per-VI registers / memory regions
853  * @interrupt_mode: Interrupt mode
854  * @timer_quantum_ns: Interrupt timer quantum, in nanoseconds
855  * @timer_max_ns: Interrupt timer maximum value, in nanoseconds
856  * @irq_rx_adaptive: Adaptive IRQ moderation enabled for RX event queues
857  * @irqs_hooked: Channel interrupts are hooked
858  * @irq_rx_mod_step_us: Step size for IRQ moderation for RX event queues
859  * @irq_rx_moderation_us: IRQ moderation time for RX event queues
860  * @msg_enable: Log message enable flags
861  * @state: Device state number (%STATE_*). Serialised by the rtnl_lock.
862  * @reset_pending: Bitmask for pending resets
863  * @tx_queue: TX DMA queues
864  * @rx_queue: RX DMA queues
865  * @channel: Channels
866  * @msi_context: Context for each MSI
867  * @extra_channel_types: Types of extra (non-traffic) channels that
868  *	should be allocated for this NIC
869  * @xdp_tx_queue_count: Number of entries in %xdp_tx_queues.
870  * @xdp_tx_queues: Array of pointers to tx queues used for XDP transmit.
871  * @xdp_txq_queues_mode: XDP TX queues sharing strategy.
872  * @rxq_entries: Size of receive queues requested by user.
873  * @txq_entries: Size of transmit queues requested by user.
874  * @txq_stop_thresh: TX queue fill level at or above which we stop it.
875  * @txq_wake_thresh: TX queue fill level at or below which we wake it.
876  * @tx_dc_base: Base qword address in SRAM of TX queue descriptor caches
877  * @rx_dc_base: Base qword address in SRAM of RX queue descriptor caches
878  * @sram_lim_qw: Qword address limit of SRAM
879  * @next_buffer_table: First available buffer table id
880  * @n_channels: Number of channels in use
881  * @n_rx_channels: Number of channels used for RX (= number of RX queues)
882  * @n_tx_channels: Number of channels used for TX
883  * @n_extra_tx_channels: Number of extra channels with TX queues
884  * @tx_queues_per_channel: number of TX queues probed on each channel
885  * @n_xdp_channels: Number of channels used for XDP TX
886  * @xdp_channel_offset: Offset of zeroth channel used for XPD TX.
887  * @xdp_tx_per_channel: Max number of TX queues on an XDP TX channel.
888  * @rx_ip_align: RX DMA address offset to have IP header aligned in
889  *	in accordance with NET_IP_ALIGN
890  * @rx_dma_len: Current maximum RX DMA length
891  * @rx_buffer_order: Order (log2) of number of pages for each RX buffer
892  * @rx_buffer_truesize: Amortised allocation size of an RX buffer,
893  *	for use in sk_buff::truesize
894  * @rx_prefix_size: Size of RX prefix before packet data
895  * @rx_packet_hash_offset: Offset of RX flow hash from start of packet data
896  *	(valid only if @rx_prefix_size != 0; always negative)
897  * @rx_packet_len_offset: Offset of RX packet length from start of packet data
898  *	(valid only for NICs that set %EFX_RX_PKT_PREFIX_LEN; always negative)
899  * @rx_packet_ts_offset: Offset of timestamp from start of packet data
900  *	(valid only if channel->sync_timestamps_enabled; always negative)
901  * @rx_scatter: Scatter mode enabled for receives
902  * @rss_context: Main RSS context.  Its @list member is the head of the list of
903  *	RSS contexts created by user requests
904  * @rss_lock: Protects custom RSS context software state in @rss_context.list
905  * @vport_id: The function's vport ID, only relevant for PFs
906  * @int_error_count: Number of internal errors seen recently
907  * @int_error_expire: Time at which error count will be expired
908  * @must_realloc_vis: Flag: VIs have yet to be reallocated after MC reboot
909  * @irq_soft_enabled: Are IRQs soft-enabled? If not, IRQ handler will
910  *	acknowledge but do nothing else.
911  * @irq_status: Interrupt status buffer
912  * @irq_zero_count: Number of legacy IRQs seen with queue flags == 0
913  * @irq_level: IRQ level/index for IRQs not triggered by an event queue
914  * @selftest_work: Work item for asynchronous self-test
915  * @mtd_list: List of MTDs attached to the NIC
916  * @nic_data: Hardware dependent state
917  * @mcdi: Management-Controller-to-Driver Interface state
918  * @mac_lock: MAC access lock. Protects @port_enabled, @phy_mode,
919  *	efx_monitor() and efx_reconfigure_port()
920  * @port_enabled: Port enabled indicator.
921  *	Serialises efx_stop_all(), efx_start_all(), efx_monitor() and
922  *	efx_mac_work() with kernel interfaces. Safe to read under any
923  *	one of the rtnl_lock, mac_lock, or netif_tx_lock, but all three must
924  *	be held to modify it.
925  * @port_initialized: Port initialized?
926  * @net_dev: Operating system network device. Consider holding the rtnl lock
927  * @fixed_features: Features which cannot be turned off
928  * @num_mac_stats: Number of MAC stats reported by firmware (MAC_STATS_NUM_STATS
929  *	field of %MC_CMD_GET_CAPABILITIES_V4 response, or %MC_CMD_MAC_NSTATS)
930  * @stats_buffer: DMA buffer for statistics
931  * @phy_type: PHY type
932  * @phy_data: PHY private data (including PHY-specific stats)
933  * @mdio: PHY MDIO interface
934  * @mdio_bus: PHY MDIO bus ID (only used by Siena)
935  * @phy_mode: PHY operating mode. Serialised by @mac_lock.
936  * @link_advertising: Autonegotiation advertising flags
937  * @fec_config: Forward Error Correction configuration flags.  For bit positions
938  *	see &enum ethtool_fec_config_bits.
939  * @link_state: Current state of the link
940  * @n_link_state_changes: Number of times the link has changed state
941  * @unicast_filter: Flag for Falcon-arch simple unicast filter.
942  *	Protected by @mac_lock.
943  * @multicast_hash: Multicast hash table for Falcon-arch.
944  *	Protected by @mac_lock.
945  * @wanted_fc: Wanted flow control flags
946  * @fc_disable: When non-zero flow control is disabled. Typically used to
947  *	ensure that network back pressure doesn't delay dma queue flushes.
948  *	Serialised by the rtnl lock.
949  * @mac_work: Work item for changing MAC promiscuity and multicast hash
950  * @loopback_mode: Loopback status
951  * @loopback_modes: Supported loopback mode bitmask
952  * @loopback_selftest: Offline self-test private state
953  * @xdp_prog: Current XDP programme for this interface
954  * @filter_sem: Filter table rw_semaphore, protects existence of @filter_state
955  * @filter_state: Architecture-dependent filter table state
956  * @rps_mutex: Protects RPS state of all channels
957  * @rps_slot_map: bitmap of in-flight entries in @rps_slot
958  * @rps_slot: array of ARFS insertion requests for efx_filter_rfs_work()
959  * @rps_hash_lock: Protects ARFS filter mapping state (@rps_hash_table and
960  *	@rps_next_id).
961  * @rps_hash_table: Mapping between ARFS filters and their various IDs
962  * @rps_next_id: next arfs_id for an ARFS filter
963  * @active_queues: Count of RX and TX queues that haven't been flushed and drained.
964  * @rxq_flush_pending: Count of number of receive queues that need to be flushed.
965  *	Decremented when the efx_flush_rx_queue() is called.
966  * @rxq_flush_outstanding: Count of number of RX flushes started but not yet
967  *	completed (either success or failure). Not used when MCDI is used to
968  *	flush receive queues.
969  * @flush_wq: wait queue used by efx_nic_flush_queues() to wait for flush completions.
970  * @vf_count: Number of VFs intended to be enabled.
971  * @vf_init_count: Number of VFs that have been fully initialised.
972  * @vi_scale: log2 number of vnics per VF.
973  * @vf_reps_lock: Protects vf_reps list
974  * @vf_reps: local VF reps
975  * @ptp_data: PTP state data
976  * @ptp_warned: has this NIC seen and warned about unexpected PTP events?
977  * @vpd_sn: Serial number read from VPD
978  * @xdp_rxq_info_failed: Have any of the rx queues failed to initialise their
979  *      xdp_rxq_info structures?
980  * @netdev_notifier: Netdevice notifier.
981  * @tc: state for TC offload (EF100).
982  * @mem_bar: The BAR that is mapped into membase.
983  * @reg_base: Offset from the start of the bar to the function control window.
984  * @monitor_work: Hardware monitor workitem
985  * @biu_lock: BIU (bus interface unit) lock
986  * @last_irq_cpu: Last CPU to handle a possible test interrupt.  This
987  *	field is used by efx_test_interrupts() to verify that an
988  *	interrupt has occurred.
989  * @stats_lock: Statistics update lock. Must be held when calling
990  *	efx_nic_type::{update,start,stop}_stats.
991  * @n_rx_noskb_drops: Count of RX packets dropped due to failure to allocate an skb
992  *
993  * This is stored in the private area of the &struct net_device.
994  */
995 struct efx_nic {
996 	/* The following fields should be written very rarely */
997 
998 	char name[IFNAMSIZ];
999 	struct list_head node;
1000 	struct efx_nic *primary;
1001 	struct list_head secondary_list;
1002 	struct pci_dev *pci_dev;
1003 	unsigned int port_num;
1004 	const struct efx_nic_type *type;
1005 	int legacy_irq;
1006 	bool eeh_disabled_legacy_irq;
1007 	struct workqueue_struct *workqueue;
1008 	char workqueue_name[16];
1009 	struct work_struct reset_work;
1010 	resource_size_t membase_phys;
1011 	void __iomem *membase;
1012 
1013 	unsigned int vi_stride;
1014 
1015 	enum efx_int_mode interrupt_mode;
1016 	unsigned int timer_quantum_ns;
1017 	unsigned int timer_max_ns;
1018 	bool irq_rx_adaptive;
1019 	bool irqs_hooked;
1020 	unsigned int irq_mod_step_us;
1021 	unsigned int irq_rx_moderation_us;
1022 	u32 msg_enable;
1023 
1024 	enum nic_state state;
1025 	unsigned long reset_pending;
1026 
1027 	struct efx_channel *channel[EFX_MAX_CHANNELS];
1028 	struct efx_msi_context msi_context[EFX_MAX_CHANNELS];
1029 	const struct efx_channel_type *
1030 	extra_channel_type[EFX_MAX_EXTRA_CHANNELS];
1031 
1032 	unsigned int xdp_tx_queue_count;
1033 	struct efx_tx_queue **xdp_tx_queues;
1034 	enum efx_xdp_tx_queues_mode xdp_txq_queues_mode;
1035 
1036 	unsigned rxq_entries;
1037 	unsigned txq_entries;
1038 	unsigned int txq_stop_thresh;
1039 	unsigned int txq_wake_thresh;
1040 
1041 	unsigned tx_dc_base;
1042 	unsigned rx_dc_base;
1043 	unsigned sram_lim_qw;
1044 	unsigned next_buffer_table;
1045 
1046 	unsigned int max_channels;
1047 	unsigned int max_vis;
1048 	unsigned int max_tx_channels;
1049 	unsigned n_channels;
1050 	unsigned n_rx_channels;
1051 	unsigned rss_spread;
1052 	unsigned tx_channel_offset;
1053 	unsigned n_tx_channels;
1054 	unsigned n_extra_tx_channels;
1055 	unsigned int tx_queues_per_channel;
1056 	unsigned int n_xdp_channels;
1057 	unsigned int xdp_channel_offset;
1058 	unsigned int xdp_tx_per_channel;
1059 	unsigned int rx_ip_align;
1060 	unsigned int rx_dma_len;
1061 	unsigned int rx_buffer_order;
1062 	unsigned int rx_buffer_truesize;
1063 	unsigned int rx_page_buf_step;
1064 	unsigned int rx_bufs_per_page;
1065 	unsigned int rx_pages_per_batch;
1066 	unsigned int rx_prefix_size;
1067 	int rx_packet_hash_offset;
1068 	int rx_packet_len_offset;
1069 	int rx_packet_ts_offset;
1070 	bool rx_scatter;
1071 	struct efx_rss_context rss_context;
1072 	struct mutex rss_lock;
1073 	u32 vport_id;
1074 
1075 	unsigned int_error_count;
1076 	unsigned long int_error_expire;
1077 
1078 	bool must_realloc_vis;
1079 	bool irq_soft_enabled;
1080 	struct efx_buffer irq_status;
1081 	unsigned irq_zero_count;
1082 	unsigned irq_level;
1083 	struct delayed_work selftest_work;
1084 
1085 #ifdef CONFIG_SFC_MTD
1086 	struct list_head mtd_list;
1087 #endif
1088 
1089 	void *nic_data;
1090 	struct efx_mcdi_data *mcdi;
1091 
1092 	struct mutex mac_lock;
1093 	struct work_struct mac_work;
1094 	bool port_enabled;
1095 
1096 	bool mc_bist_for_other_fn;
1097 	bool port_initialized;
1098 	struct net_device *net_dev;
1099 
1100 	netdev_features_t fixed_features;
1101 
1102 	u16 num_mac_stats;
1103 	struct efx_buffer stats_buffer;
1104 	u64 rx_nodesc_drops_total;
1105 	u64 rx_nodesc_drops_while_down;
1106 	bool rx_nodesc_drops_prev_state;
1107 
1108 	unsigned int phy_type;
1109 	void *phy_data;
1110 	struct mdio_if_info mdio;
1111 	unsigned int mdio_bus;
1112 	enum efx_phy_mode phy_mode;
1113 
1114 	__ETHTOOL_DECLARE_LINK_MODE_MASK(link_advertising);
1115 	u32 fec_config;
1116 	struct efx_link_state link_state;
1117 	unsigned int n_link_state_changes;
1118 
1119 	bool unicast_filter;
1120 	union efx_multicast_hash multicast_hash;
1121 	u8 wanted_fc;
1122 	unsigned fc_disable;
1123 
1124 	atomic_t rx_reset;
1125 	enum efx_loopback_mode loopback_mode;
1126 	u64 loopback_modes;
1127 
1128 	void *loopback_selftest;
1129 	/* We access loopback_selftest immediately before running XDP,
1130 	 * so we want them next to each other.
1131 	 */
1132 	struct bpf_prog __rcu *xdp_prog;
1133 
1134 	struct rw_semaphore filter_sem;
1135 	void *filter_state;
1136 #ifdef CONFIG_RFS_ACCEL
1137 	struct mutex rps_mutex;
1138 	unsigned long rps_slot_map;
1139 	struct efx_async_filter_insertion rps_slot[EFX_RPS_MAX_IN_FLIGHT];
1140 	spinlock_t rps_hash_lock;
1141 	struct hlist_head *rps_hash_table;
1142 	u32 rps_next_id;
1143 #endif
1144 
1145 	atomic_t active_queues;
1146 	atomic_t rxq_flush_pending;
1147 	atomic_t rxq_flush_outstanding;
1148 	wait_queue_head_t flush_wq;
1149 
1150 #ifdef CONFIG_SFC_SRIOV
1151 	unsigned vf_count;
1152 	unsigned vf_init_count;
1153 	unsigned vi_scale;
1154 #endif
1155 	spinlock_t vf_reps_lock;
1156 	struct list_head vf_reps;
1157 
1158 	struct efx_ptp_data *ptp_data;
1159 	bool ptp_warned;
1160 
1161 	char *vpd_sn;
1162 	bool xdp_rxq_info_failed;
1163 
1164 	struct notifier_block netdev_notifier;
1165 	struct efx_tc_state *tc;
1166 
1167 	unsigned int mem_bar;
1168 	u32 reg_base;
1169 
1170 	/* The following fields may be written more often */
1171 
1172 	struct delayed_work monitor_work ____cacheline_aligned_in_smp;
1173 	spinlock_t biu_lock;
1174 	int last_irq_cpu;
1175 	spinlock_t stats_lock;
1176 	atomic_t n_rx_noskb_drops;
1177 };
1178 
1179 /**
1180  * struct efx_probe_data - State after hardware probe
1181  * @pci_dev: The PCI device
1182  * @efx: Efx NIC details
1183  */
1184 struct efx_probe_data {
1185 	struct pci_dev *pci_dev;
1186 	struct efx_nic efx;
1187 };
1188 
1189 static inline struct efx_nic *efx_netdev_priv(struct net_device *dev)
1190 {
1191 	struct efx_probe_data **probe_ptr = netdev_priv(dev);
1192 	struct efx_probe_data *probe_data = *probe_ptr;
1193 
1194 	return &probe_data->efx;
1195 }
1196 
1197 static inline int efx_dev_registered(struct efx_nic *efx)
1198 {
1199 	return efx->net_dev->reg_state == NETREG_REGISTERED;
1200 }
1201 
1202 static inline unsigned int efx_port_num(struct efx_nic *efx)
1203 {
1204 	return efx->port_num;
1205 }
1206 
1207 struct efx_mtd_partition {
1208 	struct list_head node;
1209 	struct mtd_info mtd;
1210 	const char *dev_type_name;
1211 	const char *type_name;
1212 	char name[IFNAMSIZ + 20];
1213 };
1214 
1215 struct efx_udp_tunnel {
1216 #define TUNNEL_ENCAP_UDP_PORT_ENTRY_INVALID	0xffff
1217 	u16 type; /* TUNNEL_ENCAP_UDP_PORT_ENTRY_foo, see mcdi_pcol.h */
1218 	__be16 port;
1219 };
1220 
1221 /**
1222  * struct efx_nic_type - Efx device type definition
1223  * @mem_bar: Get the memory BAR
1224  * @mem_map_size: Get memory BAR mapped size
1225  * @probe: Probe the controller
1226  * @remove: Free resources allocated by probe()
1227  * @init: Initialise the controller
1228  * @dimension_resources: Dimension controller resources (buffer table,
1229  *	and VIs once the available interrupt resources are clear)
1230  * @fini: Shut down the controller
1231  * @monitor: Periodic function for polling link state and hardware monitor
1232  * @map_reset_reason: Map ethtool reset reason to a reset method
1233  * @map_reset_flags: Map ethtool reset flags to a reset method, if possible
1234  * @reset: Reset the controller hardware and possibly the PHY.  This will
1235  *	be called while the controller is uninitialised.
1236  * @probe_port: Probe the MAC and PHY
1237  * @remove_port: Free resources allocated by probe_port()
1238  * @handle_global_event: Handle a "global" event (may be %NULL)
1239  * @fini_dmaq: Flush and finalise DMA queues (RX and TX queues)
1240  * @prepare_flush: Prepare the hardware for flushing the DMA queues
1241  *	(for Falcon architecture)
1242  * @finish_flush: Clean up after flushing the DMA queues (for Falcon
1243  *	architecture)
1244  * @prepare_flr: Prepare for an FLR
1245  * @finish_flr: Clean up after an FLR
1246  * @describe_stats: Describe statistics for ethtool
1247  * @update_stats: Update statistics not provided by event handling.
1248  *	Either argument may be %NULL.
1249  * @update_stats_atomic: Update statistics while in atomic context, if that
1250  *	is more limiting than @update_stats.  Otherwise, leave %NULL and
1251  *	driver core will call @update_stats.
1252  * @start_stats: Start the regular fetching of statistics
1253  * @pull_stats: Pull stats from the NIC and wait until they arrive.
1254  * @stop_stats: Stop the regular fetching of statistics
1255  * @push_irq_moderation: Apply interrupt moderation value
1256  * @reconfigure_port: Push loopback/power/txdis changes to the MAC and PHY
1257  * @prepare_enable_fc_tx: Prepare MAC to enable pause frame TX (may be %NULL)
1258  * @reconfigure_mac: Push MAC address, MTU, flow control and filter settings
1259  *	to the hardware.  Serialised by the mac_lock.
1260  * @check_mac_fault: Check MAC fault state. True if fault present.
1261  * @get_wol: Get WoL configuration from driver state
1262  * @set_wol: Push WoL configuration to the NIC
1263  * @resume_wol: Synchronise WoL state between driver and MC (e.g. after resume)
1264  * @get_fec_stats: Get standard FEC statistics.
1265  * @test_chip: Test registers.  May use efx_farch_test_registers(), and is
1266  *	expected to reset the NIC.
1267  * @test_nvram: Test validity of NVRAM contents
1268  * @mcdi_request: Send an MCDI request with the given header and SDU.
1269  *	The SDU length may be any value from 0 up to the protocol-
1270  *	defined maximum, but its buffer will be padded to a multiple
1271  *	of 4 bytes.
1272  * @mcdi_poll_response: Test whether an MCDI response is available.
1273  * @mcdi_read_response: Read the MCDI response PDU.  The offset will
1274  *	be a multiple of 4.  The length may not be, but the buffer
1275  *	will be padded so it is safe to round up.
1276  * @mcdi_poll_reboot: Test whether the MCDI has rebooted.  If so,
1277  *	return an appropriate error code for aborting any current
1278  *	request; otherwise return 0.
1279  * @irq_enable_master: Enable IRQs on the NIC.  Each event queue must
1280  *	be separately enabled after this.
1281  * @irq_test_generate: Generate a test IRQ
1282  * @irq_disable_non_ev: Disable non-event IRQs on the NIC.  Each event
1283  *	queue must be separately disabled before this.
1284  * @irq_handle_msi: Handle MSI for a channel.  The @dev_id argument is
1285  *	a pointer to the &struct efx_msi_context for the channel.
1286  * @irq_handle_legacy: Handle legacy interrupt.  The @dev_id argument
1287  *	is a pointer to the &struct efx_nic.
1288  * @tx_probe: Allocate resources for TX queue (and select TXQ type)
1289  * @tx_init: Initialise TX queue on the NIC
1290  * @tx_remove: Free resources for TX queue
1291  * @tx_write: Write TX descriptors and doorbell
1292  * @tx_enqueue: Add an SKB to TX queue
1293  * @rx_push_rss_config: Write RSS hash key and indirection table to the NIC
1294  * @rx_pull_rss_config: Read RSS hash key and indirection table back from the NIC
1295  * @rx_push_rss_context_config: Write RSS hash key and indirection table for
1296  *	user RSS context to the NIC
1297  * @rx_pull_rss_context_config: Read RSS hash key and indirection table for user
1298  *	RSS context back from the NIC
1299  * @rx_probe: Allocate resources for RX queue
1300  * @rx_init: Initialise RX queue on the NIC
1301  * @rx_remove: Free resources for RX queue
1302  * @rx_write: Write RX descriptors and doorbell
1303  * @rx_defer_refill: Generate a refill reminder event
1304  * @rx_packet: Receive the queued RX buffer on a channel
1305  * @rx_buf_hash_valid: Determine whether the RX prefix contains a valid hash
1306  * @ev_probe: Allocate resources for event queue
1307  * @ev_init: Initialise event queue on the NIC
1308  * @ev_fini: Deinitialise event queue on the NIC
1309  * @ev_remove: Free resources for event queue
1310  * @ev_process: Process events for a queue, up to the given NAPI quota
1311  * @ev_read_ack: Acknowledge read events on a queue, rearming its IRQ
1312  * @ev_test_generate: Generate a test event
1313  * @filter_table_probe: Probe filter capabilities and set up filter software state
1314  * @filter_table_restore: Restore filters removed from hardware
1315  * @filter_table_remove: Remove filters from hardware and tear down software state
1316  * @filter_update_rx_scatter: Update filters after change to rx scatter setting
1317  * @filter_insert: add or replace a filter
1318  * @filter_remove_safe: remove a filter by ID, carefully
1319  * @filter_get_safe: retrieve a filter by ID, carefully
1320  * @filter_clear_rx: Remove all RX filters whose priority is less than or
1321  *	equal to the given priority and is not %EFX_FILTER_PRI_AUTO
1322  * @filter_count_rx_used: Get the number of filters in use at a given priority
1323  * @filter_get_rx_id_limit: Get maximum value of a filter id, plus 1
1324  * @filter_get_rx_ids: Get list of RX filters at a given priority
1325  * @filter_rfs_expire_one: Consider expiring a filter inserted for RFS.
1326  *	This must check whether the specified table entry is used by RFS
1327  *	and that rps_may_expire_flow() returns true for it.
1328  * @mtd_probe: Probe and add MTD partitions associated with this net device,
1329  *	 using efx_mtd_add()
1330  * @mtd_rename: Set an MTD partition name using the net device name
1331  * @mtd_read: Read from an MTD partition
1332  * @mtd_erase: Erase part of an MTD partition
1333  * @mtd_write: Write to an MTD partition
1334  * @mtd_sync: Wait for write-back to complete on MTD partition.  This
1335  *	also notifies the driver that a writer has finished using this
1336  *	partition.
1337  * @ptp_write_host_time: Send host time to MC as part of sync protocol
1338  * @ptp_set_ts_sync_events: Enable or disable sync events for inline RX
1339  *	timestamping, possibly only temporarily for the purposes of a reset.
1340  * @ptp_set_ts_config: Set hardware timestamp configuration.  The flags
1341  *	and tx_type will already have been validated but this operation
1342  *	must validate and update rx_filter.
1343  * @get_phys_port_id: Get the underlying physical port id.
1344  * @set_mac_address: Set the MAC address of the device
1345  * @tso_versions: Returns mask of firmware-assisted TSO versions supported.
1346  *	If %NULL, then device does not support any TSO version.
1347  * @udp_tnl_push_ports: Push the list of UDP tunnel ports to the NIC if required.
1348  * @udp_tnl_has_port: Check if a port has been added as UDP tunnel
1349  * @print_additional_fwver: Dump NIC-specific additional FW version info
1350  * @sensor_event: Handle a sensor event from MCDI
1351  * @rx_recycle_ring_size: Size of the RX recycle ring
1352  * @revision: Hardware architecture revision
1353  * @txd_ptr_tbl_base: TX descriptor ring base address
1354  * @rxd_ptr_tbl_base: RX descriptor ring base address
1355  * @buf_tbl_base: Buffer table base address
1356  * @evq_ptr_tbl_base: Event queue pointer table base address
1357  * @evq_rptr_tbl_base: Event queue read-pointer table base address
1358  * @max_dma_mask: Maximum possible DMA mask
1359  * @rx_prefix_size: Size of RX prefix before packet data
1360  * @rx_hash_offset: Offset of RX flow hash within prefix
1361  * @rx_ts_offset: Offset of timestamp within prefix
1362  * @rx_buffer_padding: Size of padding at end of RX packet
1363  * @can_rx_scatter: NIC is able to scatter packets to multiple buffers
1364  * @always_rx_scatter: NIC will always scatter packets to multiple buffers
1365  * @option_descriptors: NIC supports TX option descriptors
1366  * @min_interrupt_mode: Lowest capability interrupt mode supported
1367  *	from &enum efx_int_mode.
1368  * @timer_period_max: Maximum period of interrupt timer (in ticks)
1369  * @offload_features: net_device feature flags for protocol offload
1370  *	features implemented in hardware
1371  * @mcdi_max_ver: Maximum MCDI version supported
1372  * @hwtstamp_filters: Mask of hardware timestamp filter types supported
1373  */
1374 struct efx_nic_type {
1375 	bool is_vf;
1376 	unsigned int (*mem_bar)(struct efx_nic *efx);
1377 	unsigned int (*mem_map_size)(struct efx_nic *efx);
1378 	int (*probe)(struct efx_nic *efx);
1379 	void (*remove)(struct efx_nic *efx);
1380 	int (*init)(struct efx_nic *efx);
1381 	int (*dimension_resources)(struct efx_nic *efx);
1382 	void (*fini)(struct efx_nic *efx);
1383 	void (*monitor)(struct efx_nic *efx);
1384 	enum reset_type (*map_reset_reason)(enum reset_type reason);
1385 	int (*map_reset_flags)(u32 *flags);
1386 	int (*reset)(struct efx_nic *efx, enum reset_type method);
1387 	int (*probe_port)(struct efx_nic *efx);
1388 	void (*remove_port)(struct efx_nic *efx);
1389 	bool (*handle_global_event)(struct efx_channel *channel, efx_qword_t *);
1390 	int (*fini_dmaq)(struct efx_nic *efx);
1391 	void (*prepare_flush)(struct efx_nic *efx);
1392 	void (*finish_flush)(struct efx_nic *efx);
1393 	void (*prepare_flr)(struct efx_nic *efx);
1394 	void (*finish_flr)(struct efx_nic *efx);
1395 	size_t (*describe_stats)(struct efx_nic *efx, u8 *names);
1396 	size_t (*update_stats)(struct efx_nic *efx, u64 *full_stats,
1397 			       struct rtnl_link_stats64 *core_stats);
1398 	size_t (*update_stats_atomic)(struct efx_nic *efx, u64 *full_stats,
1399 				      struct rtnl_link_stats64 *core_stats);
1400 	void (*start_stats)(struct efx_nic *efx);
1401 	void (*pull_stats)(struct efx_nic *efx);
1402 	void (*stop_stats)(struct efx_nic *efx);
1403 	void (*push_irq_moderation)(struct efx_channel *channel);
1404 	int (*reconfigure_port)(struct efx_nic *efx);
1405 	void (*prepare_enable_fc_tx)(struct efx_nic *efx);
1406 	int (*reconfigure_mac)(struct efx_nic *efx, bool mtu_only);
1407 	bool (*check_mac_fault)(struct efx_nic *efx);
1408 	void (*get_wol)(struct efx_nic *efx, struct ethtool_wolinfo *wol);
1409 	int (*set_wol)(struct efx_nic *efx, u32 type);
1410 	void (*resume_wol)(struct efx_nic *efx);
1411 	void (*get_fec_stats)(struct efx_nic *efx,
1412 			      struct ethtool_fec_stats *fec_stats);
1413 	unsigned int (*check_caps)(const struct efx_nic *efx,
1414 				   u8 flag,
1415 				   u32 offset);
1416 	int (*test_chip)(struct efx_nic *efx, struct efx_self_tests *tests);
1417 	int (*test_nvram)(struct efx_nic *efx);
1418 	void (*mcdi_request)(struct efx_nic *efx,
1419 			     const efx_dword_t *hdr, size_t hdr_len,
1420 			     const efx_dword_t *sdu, size_t sdu_len);
1421 	bool (*mcdi_poll_response)(struct efx_nic *efx);
1422 	void (*mcdi_read_response)(struct efx_nic *efx, efx_dword_t *pdu,
1423 				   size_t pdu_offset, size_t pdu_len);
1424 	int (*mcdi_poll_reboot)(struct efx_nic *efx);
1425 	void (*mcdi_reboot_detected)(struct efx_nic *efx);
1426 	void (*irq_enable_master)(struct efx_nic *efx);
1427 	int (*irq_test_generate)(struct efx_nic *efx);
1428 	void (*irq_disable_non_ev)(struct efx_nic *efx);
1429 	irqreturn_t (*irq_handle_msi)(int irq, void *dev_id);
1430 	irqreturn_t (*irq_handle_legacy)(int irq, void *dev_id);
1431 	int (*tx_probe)(struct efx_tx_queue *tx_queue);
1432 	void (*tx_init)(struct efx_tx_queue *tx_queue);
1433 	void (*tx_remove)(struct efx_tx_queue *tx_queue);
1434 	void (*tx_write)(struct efx_tx_queue *tx_queue);
1435 	netdev_tx_t (*tx_enqueue)(struct efx_tx_queue *tx_queue, struct sk_buff *skb);
1436 	unsigned int (*tx_limit_len)(struct efx_tx_queue *tx_queue,
1437 				     dma_addr_t dma_addr, unsigned int len);
1438 	int (*rx_push_rss_config)(struct efx_nic *efx, bool user,
1439 				  const u32 *rx_indir_table, const u8 *key);
1440 	int (*rx_pull_rss_config)(struct efx_nic *efx);
1441 	int (*rx_push_rss_context_config)(struct efx_nic *efx,
1442 					  struct efx_rss_context *ctx,
1443 					  const u32 *rx_indir_table,
1444 					  const u8 *key);
1445 	int (*rx_pull_rss_context_config)(struct efx_nic *efx,
1446 					  struct efx_rss_context *ctx);
1447 	void (*rx_restore_rss_contexts)(struct efx_nic *efx);
1448 	int (*rx_probe)(struct efx_rx_queue *rx_queue);
1449 	void (*rx_init)(struct efx_rx_queue *rx_queue);
1450 	void (*rx_remove)(struct efx_rx_queue *rx_queue);
1451 	void (*rx_write)(struct efx_rx_queue *rx_queue);
1452 	void (*rx_defer_refill)(struct efx_rx_queue *rx_queue);
1453 	void (*rx_packet)(struct efx_channel *channel);
1454 	bool (*rx_buf_hash_valid)(const u8 *prefix);
1455 	int (*ev_probe)(struct efx_channel *channel);
1456 	int (*ev_init)(struct efx_channel *channel);
1457 	void (*ev_fini)(struct efx_channel *channel);
1458 	void (*ev_remove)(struct efx_channel *channel);
1459 	int (*ev_process)(struct efx_channel *channel, int quota);
1460 	void (*ev_read_ack)(struct efx_channel *channel);
1461 	void (*ev_test_generate)(struct efx_channel *channel);
1462 	int (*filter_table_probe)(struct efx_nic *efx);
1463 	void (*filter_table_restore)(struct efx_nic *efx);
1464 	void (*filter_table_remove)(struct efx_nic *efx);
1465 	void (*filter_update_rx_scatter)(struct efx_nic *efx);
1466 	s32 (*filter_insert)(struct efx_nic *efx,
1467 			     struct efx_filter_spec *spec, bool replace);
1468 	int (*filter_remove_safe)(struct efx_nic *efx,
1469 				  enum efx_filter_priority priority,
1470 				  u32 filter_id);
1471 	int (*filter_get_safe)(struct efx_nic *efx,
1472 			       enum efx_filter_priority priority,
1473 			       u32 filter_id, struct efx_filter_spec *);
1474 	int (*filter_clear_rx)(struct efx_nic *efx,
1475 			       enum efx_filter_priority priority);
1476 	u32 (*filter_count_rx_used)(struct efx_nic *efx,
1477 				    enum efx_filter_priority priority);
1478 	u32 (*filter_get_rx_id_limit)(struct efx_nic *efx);
1479 	s32 (*filter_get_rx_ids)(struct efx_nic *efx,
1480 				 enum efx_filter_priority priority,
1481 				 u32 *buf, u32 size);
1482 #ifdef CONFIG_RFS_ACCEL
1483 	bool (*filter_rfs_expire_one)(struct efx_nic *efx, u32 flow_id,
1484 				      unsigned int index);
1485 #endif
1486 #ifdef CONFIG_SFC_MTD
1487 	int (*mtd_probe)(struct efx_nic *efx);
1488 	void (*mtd_rename)(struct efx_mtd_partition *part);
1489 	int (*mtd_read)(struct mtd_info *mtd, loff_t start, size_t len,
1490 			size_t *retlen, u8 *buffer);
1491 	int (*mtd_erase)(struct mtd_info *mtd, loff_t start, size_t len);
1492 	int (*mtd_write)(struct mtd_info *mtd, loff_t start, size_t len,
1493 			 size_t *retlen, const u8 *buffer);
1494 	int (*mtd_sync)(struct mtd_info *mtd);
1495 #endif
1496 	void (*ptp_write_host_time)(struct efx_nic *efx, u32 host_time);
1497 	int (*ptp_set_ts_sync_events)(struct efx_nic *efx, bool en, bool temp);
1498 	int (*ptp_set_ts_config)(struct efx_nic *efx,
1499 				 struct hwtstamp_config *init);
1500 	int (*sriov_configure)(struct efx_nic *efx, int num_vfs);
1501 	int (*vlan_rx_add_vid)(struct efx_nic *efx, __be16 proto, u16 vid);
1502 	int (*vlan_rx_kill_vid)(struct efx_nic *efx, __be16 proto, u16 vid);
1503 	int (*get_phys_port_id)(struct efx_nic *efx,
1504 				struct netdev_phys_item_id *ppid);
1505 	int (*sriov_init)(struct efx_nic *efx);
1506 	void (*sriov_fini)(struct efx_nic *efx);
1507 	bool (*sriov_wanted)(struct efx_nic *efx);
1508 	void (*sriov_reset)(struct efx_nic *efx);
1509 	void (*sriov_flr)(struct efx_nic *efx, unsigned vf_i);
1510 	int (*sriov_set_vf_mac)(struct efx_nic *efx, int vf_i, const u8 *mac);
1511 	int (*sriov_set_vf_vlan)(struct efx_nic *efx, int vf_i, u16 vlan,
1512 				 u8 qos);
1513 	int (*sriov_set_vf_spoofchk)(struct efx_nic *efx, int vf_i,
1514 				     bool spoofchk);
1515 	int (*sriov_get_vf_config)(struct efx_nic *efx, int vf_i,
1516 				   struct ifla_vf_info *ivi);
1517 	int (*sriov_set_vf_link_state)(struct efx_nic *efx, int vf_i,
1518 				       int link_state);
1519 	int (*vswitching_probe)(struct efx_nic *efx);
1520 	int (*vswitching_restore)(struct efx_nic *efx);
1521 	void (*vswitching_remove)(struct efx_nic *efx);
1522 	int (*get_mac_address)(struct efx_nic *efx, unsigned char *perm_addr);
1523 	int (*set_mac_address)(struct efx_nic *efx);
1524 	u32 (*tso_versions)(struct efx_nic *efx);
1525 	int (*udp_tnl_push_ports)(struct efx_nic *efx);
1526 	bool (*udp_tnl_has_port)(struct efx_nic *efx, __be16 port);
1527 	size_t (*print_additional_fwver)(struct efx_nic *efx, char *buf,
1528 					 size_t len);
1529 	void (*sensor_event)(struct efx_nic *efx, efx_qword_t *ev);
1530 	unsigned int (*rx_recycle_ring_size)(const struct efx_nic *efx);
1531 
1532 	int revision;
1533 	unsigned int txd_ptr_tbl_base;
1534 	unsigned int rxd_ptr_tbl_base;
1535 	unsigned int buf_tbl_base;
1536 	unsigned int evq_ptr_tbl_base;
1537 	unsigned int evq_rptr_tbl_base;
1538 	u64 max_dma_mask;
1539 	unsigned int rx_prefix_size;
1540 	unsigned int rx_hash_offset;
1541 	unsigned int rx_ts_offset;
1542 	unsigned int rx_buffer_padding;
1543 	bool can_rx_scatter;
1544 	bool always_rx_scatter;
1545 	bool option_descriptors;
1546 	unsigned int min_interrupt_mode;
1547 	unsigned int timer_period_max;
1548 	netdev_features_t offload_features;
1549 	int mcdi_max_ver;
1550 	unsigned int max_rx_ip_filters;
1551 	u32 hwtstamp_filters;
1552 	unsigned int rx_hash_key_size;
1553 };
1554 
1555 /**************************************************************************
1556  *
1557  * Prototypes and inline functions
1558  *
1559  *************************************************************************/
1560 
1561 static inline struct efx_channel *
1562 efx_get_channel(struct efx_nic *efx, unsigned index)
1563 {
1564 	EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_channels);
1565 	return efx->channel[index];
1566 }
1567 
1568 /* Iterate over all used channels */
1569 #define efx_for_each_channel(_channel, _efx)				\
1570 	for (_channel = (_efx)->channel[0];				\
1571 	     _channel;							\
1572 	     _channel = (_channel->channel + 1 < (_efx)->n_channels) ?	\
1573 		     (_efx)->channel[_channel->channel + 1] : NULL)
1574 
1575 /* Iterate over all used channels in reverse */
1576 #define efx_for_each_channel_rev(_channel, _efx)			\
1577 	for (_channel = (_efx)->channel[(_efx)->n_channels - 1];	\
1578 	     _channel;							\
1579 	     _channel = _channel->channel ?				\
1580 		     (_efx)->channel[_channel->channel - 1] : NULL)
1581 
1582 static inline struct efx_channel *
1583 efx_get_tx_channel(struct efx_nic *efx, unsigned int index)
1584 {
1585 	EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_tx_channels);
1586 	return efx->channel[efx->tx_channel_offset + index];
1587 }
1588 
1589 static inline struct efx_channel *
1590 efx_get_xdp_channel(struct efx_nic *efx, unsigned int index)
1591 {
1592 	EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_xdp_channels);
1593 	return efx->channel[efx->xdp_channel_offset + index];
1594 }
1595 
1596 static inline bool efx_channel_is_xdp_tx(struct efx_channel *channel)
1597 {
1598 	return channel->channel - channel->efx->xdp_channel_offset <
1599 	       channel->efx->n_xdp_channels;
1600 }
1601 
1602 static inline bool efx_channel_has_tx_queues(struct efx_channel *channel)
1603 {
1604 	return channel && channel->channel >= channel->efx->tx_channel_offset;
1605 }
1606 
1607 static inline unsigned int efx_channel_num_tx_queues(struct efx_channel *channel)
1608 {
1609 	if (efx_channel_is_xdp_tx(channel))
1610 		return channel->efx->xdp_tx_per_channel;
1611 	return channel->efx->tx_queues_per_channel;
1612 }
1613 
1614 static inline struct efx_tx_queue *
1615 efx_channel_get_tx_queue(struct efx_channel *channel, unsigned int type)
1616 {
1617 	EFX_WARN_ON_ONCE_PARANOID(type >= EFX_TXQ_TYPES);
1618 	return channel->tx_queue_by_type[type];
1619 }
1620 
1621 static inline struct efx_tx_queue *
1622 efx_get_tx_queue(struct efx_nic *efx, unsigned int index, unsigned int type)
1623 {
1624 	struct efx_channel *channel = efx_get_tx_channel(efx, index);
1625 
1626 	return efx_channel_get_tx_queue(channel, type);
1627 }
1628 
1629 /* Iterate over all TX queues belonging to a channel */
1630 #define efx_for_each_channel_tx_queue(_tx_queue, _channel)		\
1631 	if (!efx_channel_has_tx_queues(_channel))			\
1632 		;							\
1633 	else								\
1634 		for (_tx_queue = (_channel)->tx_queue;			\
1635 		     _tx_queue < (_channel)->tx_queue +			\
1636 				 efx_channel_num_tx_queues(_channel);		\
1637 		     _tx_queue++)
1638 
1639 static inline bool efx_channel_has_rx_queue(struct efx_channel *channel)
1640 {
1641 	return channel->rx_queue.core_index >= 0;
1642 }
1643 
1644 static inline struct efx_rx_queue *
1645 efx_channel_get_rx_queue(struct efx_channel *channel)
1646 {
1647 	EFX_WARN_ON_ONCE_PARANOID(!efx_channel_has_rx_queue(channel));
1648 	return &channel->rx_queue;
1649 }
1650 
1651 /* Iterate over all RX queues belonging to a channel */
1652 #define efx_for_each_channel_rx_queue(_rx_queue, _channel)		\
1653 	if (!efx_channel_has_rx_queue(_channel))			\
1654 		;							\
1655 	else								\
1656 		for (_rx_queue = &(_channel)->rx_queue;			\
1657 		     _rx_queue;						\
1658 		     _rx_queue = NULL)
1659 
1660 static inline struct efx_channel *
1661 efx_rx_queue_channel(struct efx_rx_queue *rx_queue)
1662 {
1663 	return container_of(rx_queue, struct efx_channel, rx_queue);
1664 }
1665 
1666 static inline int efx_rx_queue_index(struct efx_rx_queue *rx_queue)
1667 {
1668 	return efx_rx_queue_channel(rx_queue)->channel;
1669 }
1670 
1671 /* Returns a pointer to the specified receive buffer in the RX
1672  * descriptor queue.
1673  */
1674 static inline struct efx_rx_buffer *efx_rx_buffer(struct efx_rx_queue *rx_queue,
1675 						  unsigned int index)
1676 {
1677 	return &rx_queue->buffer[index];
1678 }
1679 
1680 static inline struct efx_rx_buffer *
1681 efx_rx_buf_next(struct efx_rx_queue *rx_queue, struct efx_rx_buffer *rx_buf)
1682 {
1683 	if (unlikely(rx_buf == efx_rx_buffer(rx_queue, rx_queue->ptr_mask)))
1684 		return efx_rx_buffer(rx_queue, 0);
1685 	else
1686 		return rx_buf + 1;
1687 }
1688 
1689 /**
1690  * EFX_MAX_FRAME_LEN - calculate maximum frame length
1691  *
1692  * This calculates the maximum frame length that will be used for a
1693  * given MTU.  The frame length will be equal to the MTU plus a
1694  * constant amount of header space and padding.  This is the quantity
1695  * that the net driver will program into the MAC as the maximum frame
1696  * length.
1697  *
1698  * The 10G MAC requires 8-byte alignment on the frame
1699  * length, so we round up to the nearest 8.
1700  *
1701  * Re-clocking by the XGXS on RX can reduce an IPG to 32 bits (half an
1702  * XGMII cycle).  If the frame length reaches the maximum value in the
1703  * same cycle, the XMAC can miss the IPG altogether.  We work around
1704  * this by adding a further 16 bytes.
1705  */
1706 #define EFX_FRAME_PAD	16
1707 #define EFX_MAX_FRAME_LEN(mtu) \
1708 	(ALIGN(((mtu) + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN + EFX_FRAME_PAD), 8))
1709 
1710 static inline bool efx_xmit_with_hwtstamp(struct sk_buff *skb)
1711 {
1712 	return skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP;
1713 }
1714 static inline void efx_xmit_hwtstamp_pending(struct sk_buff *skb)
1715 {
1716 	skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
1717 }
1718 
1719 /* Get the max fill level of the TX queues on this channel */
1720 static inline unsigned int
1721 efx_channel_tx_fill_level(struct efx_channel *channel)
1722 {
1723 	struct efx_tx_queue *tx_queue;
1724 	unsigned int fill_level = 0;
1725 
1726 	efx_for_each_channel_tx_queue(tx_queue, channel)
1727 		fill_level = max(fill_level,
1728 				 tx_queue->insert_count - tx_queue->read_count);
1729 
1730 	return fill_level;
1731 }
1732 
1733 /* Conservative approximation of efx_channel_tx_fill_level using cached value */
1734 static inline unsigned int
1735 efx_channel_tx_old_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->old_read_count);
1743 
1744 	return fill_level;
1745 }
1746 
1747 /* Get all supported features.
1748  * If a feature is not fixed, it is present in hw_features.
1749  * If a feature is fixed, it does not present in hw_features, but
1750  * always in features.
1751  */
1752 static inline netdev_features_t efx_supported_features(const struct efx_nic *efx)
1753 {
1754 	const struct net_device *net_dev = efx->net_dev;
1755 
1756 	return net_dev->features | net_dev->hw_features;
1757 }
1758 
1759 /* Get the current TX queue insert index. */
1760 static inline unsigned int
1761 efx_tx_queue_get_insert_index(const struct efx_tx_queue *tx_queue)
1762 {
1763 	return tx_queue->insert_count & tx_queue->ptr_mask;
1764 }
1765 
1766 /* Get a TX buffer. */
1767 static inline struct efx_tx_buffer *
1768 __efx_tx_queue_get_insert_buffer(const struct efx_tx_queue *tx_queue)
1769 {
1770 	return &tx_queue->buffer[efx_tx_queue_get_insert_index(tx_queue)];
1771 }
1772 
1773 /* Get a TX buffer, checking it's not currently in use. */
1774 static inline struct efx_tx_buffer *
1775 efx_tx_queue_get_insert_buffer(const struct efx_tx_queue *tx_queue)
1776 {
1777 	struct efx_tx_buffer *buffer =
1778 		__efx_tx_queue_get_insert_buffer(tx_queue);
1779 
1780 	EFX_WARN_ON_ONCE_PARANOID(buffer->len);
1781 	EFX_WARN_ON_ONCE_PARANOID(buffer->flags);
1782 	EFX_WARN_ON_ONCE_PARANOID(buffer->unmap_len);
1783 
1784 	return buffer;
1785 }
1786 
1787 #endif /* EFX_NET_DRIVER_H */
1788