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/i2c.h>
28 #include <linux/mtd/mtd.h>
29 #include <net/busy_poll.h>
30 #include <net/xdp.h>
31 
32 #include "enum.h"
33 #include "bitfield.h"
34 #include "filter.h"
35 
36 /**************************************************************************
37  *
38  * Build definitions
39  *
40  **************************************************************************/
41 
42 #define EFX_DRIVER_VERSION	"4.1"
43 
44 #ifdef DEBUG
45 #define EFX_WARN_ON_ONCE_PARANOID(x) WARN_ON_ONCE(x)
46 #define EFX_WARN_ON_PARANOID(x) WARN_ON(x)
47 #else
48 #define EFX_WARN_ON_ONCE_PARANOID(x) do {} while (0)
49 #define EFX_WARN_ON_PARANOID(x) do {} while (0)
50 #endif
51 
52 /**************************************************************************
53  *
54  * Efx data structures
55  *
56  **************************************************************************/
57 
58 #define EFX_MAX_CHANNELS 32U
59 #define EFX_MAX_RX_QUEUES EFX_MAX_CHANNELS
60 #define EFX_EXTRA_CHANNEL_IOV	0
61 #define EFX_EXTRA_CHANNEL_PTP	1
62 #define EFX_MAX_EXTRA_CHANNELS	2U
63 
64 /* Checksum generation is a per-queue option in hardware, so each
65  * queue visible to the networking core is backed by two hardware TX
66  * queues. */
67 #define EFX_MAX_TX_TC		2
68 #define EFX_MAX_CORE_TX_QUEUES	(EFX_MAX_TX_TC * EFX_MAX_CHANNELS)
69 #define EFX_TXQ_TYPE_OFFLOAD	1	/* flag */
70 #define EFX_TXQ_TYPE_HIGHPRI	2	/* flag */
71 #define EFX_TXQ_TYPES		4
72 #define EFX_MAX_TX_QUEUES	(EFX_TXQ_TYPES * 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 /* Size of an RX scatter buffer.  Small enough to pack 2 into a 4K page,
81  * and should be a multiple of the cache line size.
82  */
83 #define EFX_RX_USR_BUF_SIZE	(2048 - 256)
84 
85 /* If possible, we should ensure cache line alignment at start and end
86  * of every buffer.  Otherwise, we just need to ensure 4-byte
87  * alignment of the network header.
88  */
89 #if NET_IP_ALIGN == 0
90 #define EFX_RX_BUF_ALIGNMENT	L1_CACHE_BYTES
91 #else
92 #define EFX_RX_BUF_ALIGNMENT	4
93 #endif
94 
95 /* Forward declare Precision Time Protocol (PTP) support structure. */
96 struct efx_ptp_data;
97 struct hwtstamp_config;
98 
99 struct efx_self_tests;
100 
101 /**
102  * struct efx_buffer - A general-purpose DMA buffer
103  * @addr: host base address of the buffer
104  * @dma_addr: DMA base address of the buffer
105  * @len: Buffer length, in bytes
106  *
107  * The NIC uses these buffers for its interrupt status registers and
108  * MAC stats dumps.
109  */
110 struct efx_buffer {
111 	void *addr;
112 	dma_addr_t dma_addr;
113 	unsigned int len;
114 };
115 
116 /**
117  * struct efx_special_buffer - DMA buffer entered into buffer table
118  * @buf: Standard &struct efx_buffer
119  * @index: Buffer index within controller;s buffer table
120  * @entries: Number of buffer table entries
121  *
122  * The NIC has a buffer table that maps buffers of size %EFX_BUF_SIZE.
123  * Event and descriptor rings are addressed via one or more buffer
124  * table entries (and so can be physically non-contiguous, although we
125  * currently do not take advantage of that).  On Falcon and Siena we
126  * have to take care of allocating and initialising the entries
127  * ourselves.  On later hardware this is managed by the firmware and
128  * @index and @entries are left as 0.
129  */
130 struct efx_special_buffer {
131 	struct efx_buffer buf;
132 	unsigned int index;
133 	unsigned int entries;
134 };
135 
136 /**
137  * struct efx_tx_buffer - buffer state for a TX descriptor
138  * @skb: When @flags & %EFX_TX_BUF_SKB, the associated socket buffer to be
139  *	freed when descriptor completes
140  * @xdpf: When @flags & %EFX_TX_BUF_XDP, the XDP frame information; its @data
141  *	member is the associated buffer to drop a page reference on.
142  * @dma_addr: DMA address of the fragment.
143  * @flags: Flags for allocation and DMA mapping type
144  * @len: Length of this fragment.
145  *	This field is zero when the queue slot is empty.
146  * @unmap_len: Length of this fragment to unmap
147  * @dma_offset: Offset of @dma_addr from the address of the backing DMA mapping.
148  * Only valid if @unmap_len != 0.
149  */
150 struct efx_tx_buffer {
151 	union {
152 		const struct sk_buff *skb;
153 		struct xdp_frame *xdpf;
154 	};
155 	union {
156 		efx_qword_t option;
157 		dma_addr_t dma_addr;
158 	};
159 	unsigned short flags;
160 	unsigned short len;
161 	unsigned short unmap_len;
162 	unsigned short dma_offset;
163 };
164 #define EFX_TX_BUF_CONT		1	/* not last descriptor of packet */
165 #define EFX_TX_BUF_SKB		2	/* buffer is last part of skb */
166 #define EFX_TX_BUF_MAP_SINGLE	8	/* buffer was mapped with dma_map_single() */
167 #define EFX_TX_BUF_OPTION	0x10	/* empty buffer for option descriptor */
168 #define EFX_TX_BUF_XDP		0x20	/* buffer was sent with XDP */
169 
170 /**
171  * struct efx_tx_queue - An Efx TX queue
172  *
173  * This is a ring buffer of TX fragments.
174  * Since the TX completion path always executes on the same
175  * CPU and the xmit path can operate on different CPUs,
176  * performance is increased by ensuring that the completion
177  * path and the xmit path operate on different cache lines.
178  * This is particularly important if the xmit path is always
179  * executing on one CPU which is different from the completion
180  * path.  There is also a cache line for members which are
181  * read but not written on the fast path.
182  *
183  * @efx: The associated Efx NIC
184  * @queue: DMA queue number
185  * @tso_version: Version of TSO in use for this queue.
186  * @channel: The associated channel
187  * @core_txq: The networking core TX queue structure
188  * @buffer: The software buffer ring
189  * @cb_page: Array of pages of copy buffers.  Carved up according to
190  *	%EFX_TX_CB_ORDER into %EFX_TX_CB_SIZE-sized chunks.
191  * @txd: The hardware descriptor ring
192  * @ptr_mask: The size of the ring minus 1.
193  * @piobuf: PIO buffer region for this TX queue (shared with its partner).
194  *	Size of the region is efx_piobuf_size.
195  * @piobuf_offset: Buffer offset to be specified in PIO descriptors
196  * @initialised: Has hardware queue been initialised?
197  * @timestamping: Is timestamping enabled for this channel?
198  * @xdp_tx: Is this an XDP tx queue?
199  * @handle_tso: TSO xmit preparation handler.  Sets up the TSO metadata and
200  *	may also map tx data, depending on the nature of the TSO implementation.
201  * @read_count: Current read pointer.
202  *	This is the number of buffers that have been removed from both rings.
203  * @old_write_count: The value of @write_count when last checked.
204  *	This is here for performance reasons.  The xmit path will
205  *	only get the up-to-date value of @write_count if this
206  *	variable indicates that the queue is empty.  This is to
207  *	avoid cache-line ping-pong between the xmit path and the
208  *	completion path.
209  * @merge_events: Number of TX merged completion events
210  * @completed_desc_ptr: Most recent completed pointer - only used with
211  *      timestamping.
212  * @completed_timestamp_major: Top part of the most recent tx timestamp.
213  * @completed_timestamp_minor: Low part of the most recent tx timestamp.
214  * @insert_count: Current insert pointer
215  *	This is the number of buffers that have been added to the
216  *	software ring.
217  * @write_count: Current write pointer
218  *	This is the number of buffers that have been added to the
219  *	hardware ring.
220  * @packet_write_count: Completable write pointer
221  *	This is the write pointer of the last packet written.
222  *	Normally this will equal @write_count, but as option descriptors
223  *	don't produce completion events, they won't update this.
224  *	Filled in iff @efx->type->option_descriptors; only used for PIO.
225  *	Thus, this is written and used on EF10, and neither on farch.
226  * @old_read_count: The value of read_count when last checked.
227  *	This is here for performance reasons.  The xmit path will
228  *	only get the up-to-date value of read_count if this
229  *	variable indicates that the queue is full.  This is to
230  *	avoid cache-line ping-pong between the xmit path and the
231  *	completion path.
232  * @tso_bursts: Number of times TSO xmit invoked by kernel
233  * @tso_long_headers: Number of packets with headers too long for standard
234  *	blocks
235  * @tso_packets: Number of packets via the TSO xmit path
236  * @tso_fallbacks: Number of times TSO fallback used
237  * @pushes: Number of times the TX push feature has been used
238  * @pio_packets: Number of times the TX PIO feature has been used
239  * @xmit_more_available: Are any packets waiting to be pushed to the NIC
240  * @cb_packets: Number of times the TX copybreak feature has been used
241  * @empty_read_count: If the completion path has seen the queue as empty
242  *	and the transmission path has not yet checked this, the value of
243  *	@read_count bitwise-added to %EFX_EMPTY_COUNT_VALID; otherwise 0.
244  */
245 struct efx_tx_queue {
246 	/* Members which don't change on the fast path */
247 	struct efx_nic *efx ____cacheline_aligned_in_smp;
248 	unsigned queue;
249 	unsigned int tso_version;
250 	struct efx_channel *channel;
251 	struct netdev_queue *core_txq;
252 	struct efx_tx_buffer *buffer;
253 	struct efx_buffer *cb_page;
254 	struct efx_special_buffer txd;
255 	unsigned int ptr_mask;
256 	void __iomem *piobuf;
257 	unsigned int piobuf_offset;
258 	bool initialised;
259 	bool timestamping;
260 	bool xdp_tx;
261 
262 	/* Function pointers used in the fast path. */
263 	int (*handle_tso)(struct efx_tx_queue*, struct sk_buff*, bool *);
264 
265 	/* Members used mainly on the completion path */
266 	unsigned int read_count ____cacheline_aligned_in_smp;
267 	unsigned int old_write_count;
268 	unsigned int merge_events;
269 	unsigned int bytes_compl;
270 	unsigned int pkts_compl;
271 	unsigned int completed_desc_ptr;
272 	u32 completed_timestamp_major;
273 	u32 completed_timestamp_minor;
274 
275 	/* Members used only on the xmit path */
276 	unsigned int insert_count ____cacheline_aligned_in_smp;
277 	unsigned int write_count;
278 	unsigned int packet_write_count;
279 	unsigned int old_read_count;
280 	unsigned int tso_bursts;
281 	unsigned int tso_long_headers;
282 	unsigned int tso_packets;
283 	unsigned int tso_fallbacks;
284 	unsigned int pushes;
285 	unsigned int pio_packets;
286 	bool xmit_more_available;
287 	unsigned int cb_packets;
288 	/* Statistics to supplement MAC stats */
289 	unsigned long tx_packets;
290 
291 	/* Members shared between paths and sometimes updated */
292 	unsigned int empty_read_count ____cacheline_aligned_in_smp;
293 #define EFX_EMPTY_COUNT_VALID 0x80000000
294 	atomic_t flush_outstanding;
295 };
296 
297 #define EFX_TX_CB_ORDER	7
298 #define EFX_TX_CB_SIZE	(1 << EFX_TX_CB_ORDER) - NET_IP_ALIGN
299 
300 /**
301  * struct efx_rx_buffer - An Efx RX data buffer
302  * @dma_addr: DMA base address of the buffer
303  * @page: The associated page buffer.
304  *	Will be %NULL if the buffer slot is currently free.
305  * @page_offset: If pending: offset in @page of DMA base address.
306  *	If completed: offset in @page of Ethernet header.
307  * @len: If pending: length for DMA descriptor.
308  *	If completed: received length, excluding hash prefix.
309  * @flags: Flags for buffer and packet state.  These are only set on the
310  *	first buffer of a scattered packet.
311  */
312 struct efx_rx_buffer {
313 	dma_addr_t dma_addr;
314 	struct page *page;
315 	u16 page_offset;
316 	u16 len;
317 	u16 flags;
318 };
319 #define EFX_RX_BUF_LAST_IN_PAGE	0x0001
320 #define EFX_RX_PKT_CSUMMED	0x0002
321 #define EFX_RX_PKT_DISCARD	0x0004
322 #define EFX_RX_PKT_TCP		0x0040
323 #define EFX_RX_PKT_PREFIX_LEN	0x0080	/* length is in prefix only */
324 #define EFX_RX_PKT_CSUM_LEVEL	0x0200
325 
326 /**
327  * struct efx_rx_page_state - Page-based rx buffer state
328  *
329  * Inserted at the start of every page allocated for receive buffers.
330  * Used to facilitate sharing dma mappings between recycled rx buffers
331  * and those passed up to the kernel.
332  *
333  * @dma_addr: The dma address of this page.
334  */
335 struct efx_rx_page_state {
336 	dma_addr_t dma_addr;
337 
338 	unsigned int __pad[0] ____cacheline_aligned;
339 };
340 
341 /**
342  * struct efx_rx_queue - An Efx RX queue
343  * @efx: The associated Efx NIC
344  * @core_index:  Index of network core RX queue.  Will be >= 0 iff this
345  *	is associated with a real RX queue.
346  * @buffer: The software buffer ring
347  * @rxd: The hardware descriptor ring
348  * @ptr_mask: The size of the ring minus 1.
349  * @refill_enabled: Enable refill whenever fill level is low
350  * @flush_pending: Set when a RX flush is pending. Has the same lifetime as
351  *	@rxq_flush_pending.
352  * @added_count: Number of buffers added to the receive queue.
353  * @notified_count: Number of buffers given to NIC (<= @added_count).
354  * @removed_count: Number of buffers removed from the receive queue.
355  * @scatter_n: Used by NIC specific receive code.
356  * @scatter_len: Used by NIC specific receive code.
357  * @page_ring: The ring to store DMA mapped pages for reuse.
358  * @page_add: Counter to calculate the write pointer for the recycle ring.
359  * @page_remove: Counter to calculate the read pointer for the recycle ring.
360  * @page_recycle_count: The number of pages that have been recycled.
361  * @page_recycle_failed: The number of pages that couldn't be recycled because
362  *      the kernel still held a reference to them.
363  * @page_recycle_full: The number of pages that were released because the
364  *      recycle ring was full.
365  * @page_ptr_mask: The number of pages in the RX recycle ring minus 1.
366  * @max_fill: RX descriptor maximum fill level (<= ring size)
367  * @fast_fill_trigger: RX descriptor fill level that will trigger a fast fill
368  *	(<= @max_fill)
369  * @min_fill: RX descriptor minimum non-zero fill level.
370  *	This records the minimum fill level observed when a ring
371  *	refill was triggered.
372  * @recycle_count: RX buffer recycle counter.
373  * @slow_fill: Timer used to defer efx_nic_generate_fill_event().
374  * @xdp_rxq_info: XDP specific RX queue information.
375  * @xdp_rxq_info_valid: Is xdp_rxq_info valid data?.
376  */
377 struct efx_rx_queue {
378 	struct efx_nic *efx;
379 	int core_index;
380 	struct efx_rx_buffer *buffer;
381 	struct efx_special_buffer rxd;
382 	unsigned int ptr_mask;
383 	bool refill_enabled;
384 	bool flush_pending;
385 
386 	unsigned int added_count;
387 	unsigned int notified_count;
388 	unsigned int removed_count;
389 	unsigned int scatter_n;
390 	unsigned int scatter_len;
391 	struct page **page_ring;
392 	unsigned int page_add;
393 	unsigned int page_remove;
394 	unsigned int page_recycle_count;
395 	unsigned int page_recycle_failed;
396 	unsigned int page_recycle_full;
397 	unsigned int page_ptr_mask;
398 	unsigned int max_fill;
399 	unsigned int fast_fill_trigger;
400 	unsigned int min_fill;
401 	unsigned int min_overfill;
402 	unsigned int recycle_count;
403 	struct timer_list slow_fill;
404 	unsigned int slow_fill_count;
405 	/* Statistics to supplement MAC stats */
406 	unsigned long rx_packets;
407 	struct xdp_rxq_info xdp_rxq_info;
408 	bool xdp_rxq_info_valid;
409 };
410 
411 enum efx_sync_events_state {
412 	SYNC_EVENTS_DISABLED = 0,
413 	SYNC_EVENTS_QUIESCENT,
414 	SYNC_EVENTS_REQUESTED,
415 	SYNC_EVENTS_VALID,
416 };
417 
418 /**
419  * struct efx_channel - An Efx channel
420  *
421  * A channel comprises an event queue, at least one TX queue, at least
422  * one RX queue, and an associated tasklet for processing the event
423  * queue.
424  *
425  * @efx: Associated Efx NIC
426  * @channel: Channel instance number
427  * @type: Channel type definition
428  * @eventq_init: Event queue initialised flag
429  * @enabled: Channel enabled indicator
430  * @irq: IRQ number (MSI and MSI-X only)
431  * @irq_moderation_us: IRQ moderation value (in microseconds)
432  * @napi_dev: Net device used with NAPI
433  * @napi_str: NAPI control structure
434  * @state: state for NAPI vs busy polling
435  * @state_lock: lock protecting @state
436  * @eventq: Event queue buffer
437  * @eventq_mask: Event queue pointer mask
438  * @eventq_read_ptr: Event queue read pointer
439  * @event_test_cpu: Last CPU to handle interrupt or test event for this channel
440  * @irq_count: Number of IRQs since last adaptive moderation decision
441  * @irq_mod_score: IRQ moderation score
442  * @rfs_filter_count: number of accelerated RFS filters currently in place;
443  *	equals the count of @rps_flow_id slots filled
444  * @rfs_last_expiry: value of jiffies last time some accelerated RFS filters
445  *	were checked for expiry
446  * @rfs_expire_index: next accelerated RFS filter ID to check for expiry
447  * @n_rfs_succeeded: number of successful accelerated RFS filter insertions
448  * @n_rfs_failed; number of failed accelerated RFS filter insertions
449  * @filter_work: Work item for efx_filter_rfs_expire()
450  * @rps_flow_id: Flow IDs of filters allocated for accelerated RFS,
451  *      indexed by filter ID
452  * @n_rx_tobe_disc: Count of RX_TOBE_DISC errors
453  * @n_rx_ip_hdr_chksum_err: Count of RX IP header checksum errors
454  * @n_rx_tcp_udp_chksum_err: Count of RX TCP and UDP checksum errors
455  * @n_rx_mcast_mismatch: Count of unmatched multicast frames
456  * @n_rx_frm_trunc: Count of RX_FRM_TRUNC errors
457  * @n_rx_overlength: Count of RX_OVERLENGTH errors
458  * @n_skbuff_leaks: Count of skbuffs leaked due to RX overrun
459  * @n_rx_nodesc_trunc: Number of RX packets truncated and then dropped due to
460  *	lack of descriptors
461  * @n_rx_merge_events: Number of RX merged completion events
462  * @n_rx_merge_packets: Number of RX packets completed by merged events
463  * @n_rx_xdp_drops: Count of RX packets intentionally dropped due to XDP
464  * @n_rx_xdp_bad_drops: Count of RX packets dropped due to XDP errors
465  * @n_rx_xdp_tx: Count of RX packets retransmitted due to XDP
466  * @n_rx_xdp_redirect: Count of RX packets redirected to a different NIC by XDP
467  * @rx_pkt_n_frags: Number of fragments in next packet to be delivered by
468  *	__efx_rx_packet(), or zero if there is none
469  * @rx_pkt_index: Ring index of first buffer for next packet to be delivered
470  *	by __efx_rx_packet(), if @rx_pkt_n_frags != 0
471  * @rx_list: list of SKBs from current RX, awaiting processing
472  * @rx_queue: RX queue for this channel
473  * @tx_queue: TX queues for this channel
474  * @sync_events_state: Current state of sync events on this channel
475  * @sync_timestamp_major: Major part of the last ptp sync event
476  * @sync_timestamp_minor: Minor part of the last ptp sync event
477  */
478 struct efx_channel {
479 	struct efx_nic *efx;
480 	int channel;
481 	const struct efx_channel_type *type;
482 	bool eventq_init;
483 	bool enabled;
484 	int irq;
485 	unsigned int irq_moderation_us;
486 	struct net_device *napi_dev;
487 	struct napi_struct napi_str;
488 #ifdef CONFIG_NET_RX_BUSY_POLL
489 	unsigned long busy_poll_state;
490 #endif
491 	struct efx_special_buffer eventq;
492 	unsigned int eventq_mask;
493 	unsigned int eventq_read_ptr;
494 	int event_test_cpu;
495 
496 	unsigned int irq_count;
497 	unsigned int irq_mod_score;
498 #ifdef CONFIG_RFS_ACCEL
499 	unsigned int rfs_filter_count;
500 	unsigned int rfs_last_expiry;
501 	unsigned int rfs_expire_index;
502 	unsigned int n_rfs_succeeded;
503 	unsigned int n_rfs_failed;
504 	struct delayed_work filter_work;
505 #define RPS_FLOW_ID_INVALID 0xFFFFFFFF
506 	u32 *rps_flow_id;
507 #endif
508 
509 	unsigned int n_rx_tobe_disc;
510 	unsigned int n_rx_ip_hdr_chksum_err;
511 	unsigned int n_rx_tcp_udp_chksum_err;
512 	unsigned int n_rx_outer_ip_hdr_chksum_err;
513 	unsigned int n_rx_outer_tcp_udp_chksum_err;
514 	unsigned int n_rx_inner_ip_hdr_chksum_err;
515 	unsigned int n_rx_inner_tcp_udp_chksum_err;
516 	unsigned int n_rx_eth_crc_err;
517 	unsigned int n_rx_mcast_mismatch;
518 	unsigned int n_rx_frm_trunc;
519 	unsigned int n_rx_overlength;
520 	unsigned int n_skbuff_leaks;
521 	unsigned int n_rx_nodesc_trunc;
522 	unsigned int n_rx_merge_events;
523 	unsigned int n_rx_merge_packets;
524 	unsigned int n_rx_xdp_drops;
525 	unsigned int n_rx_xdp_bad_drops;
526 	unsigned int n_rx_xdp_tx;
527 	unsigned int n_rx_xdp_redirect;
528 
529 	unsigned int rx_pkt_n_frags;
530 	unsigned int rx_pkt_index;
531 
532 	struct list_head *rx_list;
533 
534 	struct efx_rx_queue rx_queue;
535 	struct efx_tx_queue tx_queue[EFX_TXQ_TYPES];
536 
537 	enum efx_sync_events_state sync_events_state;
538 	u32 sync_timestamp_major;
539 	u32 sync_timestamp_minor;
540 };
541 
542 /**
543  * struct efx_msi_context - Context for each MSI
544  * @efx: The associated NIC
545  * @index: Index of the channel/IRQ
546  * @name: Name of the channel/IRQ
547  *
548  * Unlike &struct efx_channel, this is never reallocated and is always
549  * safe for the IRQ handler to access.
550  */
551 struct efx_msi_context {
552 	struct efx_nic *efx;
553 	unsigned int index;
554 	char name[IFNAMSIZ + 6];
555 };
556 
557 /**
558  * struct efx_channel_type - distinguishes traffic and extra channels
559  * @handle_no_channel: Handle failure to allocate an extra channel
560  * @pre_probe: Set up extra state prior to initialisation
561  * @post_remove: Tear down extra state after finalisation, if allocated.
562  *	May be called on channels that have not been probed.
563  * @get_name: Generate the channel's name (used for its IRQ handler)
564  * @copy: Copy the channel state prior to reallocation.  May be %NULL if
565  *	reallocation is not supported.
566  * @receive_skb: Handle an skb ready to be passed to netif_receive_skb()
567  * @want_txqs: Determine whether this channel should have TX queues
568  *	created.  If %NULL, TX queues are not created.
569  * @keep_eventq: Flag for whether event queue should be kept initialised
570  *	while the device is stopped
571  * @want_pio: Flag for whether PIO buffers should be linked to this
572  *	channel's TX queues.
573  */
574 struct efx_channel_type {
575 	void (*handle_no_channel)(struct efx_nic *);
576 	int (*pre_probe)(struct efx_channel *);
577 	void (*post_remove)(struct efx_channel *);
578 	void (*get_name)(struct efx_channel *, char *buf, size_t len);
579 	struct efx_channel *(*copy)(const struct efx_channel *);
580 	bool (*receive_skb)(struct efx_channel *, struct sk_buff *);
581 	bool (*want_txqs)(struct efx_channel *);
582 	bool keep_eventq;
583 	bool want_pio;
584 };
585 
586 enum efx_led_mode {
587 	EFX_LED_OFF	= 0,
588 	EFX_LED_ON	= 1,
589 	EFX_LED_DEFAULT	= 2
590 };
591 
592 #define STRING_TABLE_LOOKUP(val, member) \
593 	((val) < member ## _max) ? member ## _names[val] : "(invalid)"
594 
595 extern const char *const efx_loopback_mode_names[];
596 extern const unsigned int efx_loopback_mode_max;
597 #define LOOPBACK_MODE(efx) \
598 	STRING_TABLE_LOOKUP((efx)->loopback_mode, efx_loopback_mode)
599 
600 extern const char *const efx_reset_type_names[];
601 extern const unsigned int efx_reset_type_max;
602 #define RESET_TYPE(type) \
603 	STRING_TABLE_LOOKUP(type, efx_reset_type)
604 
605 void efx_get_udp_tunnel_type_name(u16 type, char *buf, size_t buflen);
606 
607 enum efx_int_mode {
608 	/* Be careful if altering to correct macro below */
609 	EFX_INT_MODE_MSIX = 0,
610 	EFX_INT_MODE_MSI = 1,
611 	EFX_INT_MODE_LEGACY = 2,
612 	EFX_INT_MODE_MAX	/* Insert any new items before this */
613 };
614 #define EFX_INT_MODE_USE_MSI(x) (((x)->interrupt_mode) <= EFX_INT_MODE_MSI)
615 
616 enum nic_state {
617 	STATE_UNINIT = 0,	/* device being probed/removed or is frozen */
618 	STATE_READY = 1,	/* hardware ready and netdev registered */
619 	STATE_DISABLED = 2,	/* device disabled due to hardware errors */
620 	STATE_RECOVERY = 3,	/* device recovering from PCI error */
621 };
622 
623 /* Forward declaration */
624 struct efx_nic;
625 
626 /* Pseudo bit-mask flow control field */
627 #define EFX_FC_RX	FLOW_CTRL_RX
628 #define EFX_FC_TX	FLOW_CTRL_TX
629 #define EFX_FC_AUTO	4
630 
631 /**
632  * struct efx_link_state - Current state of the link
633  * @up: Link is up
634  * @fd: Link is full-duplex
635  * @fc: Actual flow control flags
636  * @speed: Link speed (Mbps)
637  */
638 struct efx_link_state {
639 	bool up;
640 	bool fd;
641 	u8 fc;
642 	unsigned int speed;
643 };
644 
645 static inline bool efx_link_state_equal(const struct efx_link_state *left,
646 					const struct efx_link_state *right)
647 {
648 	return left->up == right->up && left->fd == right->fd &&
649 		left->fc == right->fc && left->speed == right->speed;
650 }
651 
652 /**
653  * struct efx_phy_operations - Efx PHY operations table
654  * @probe: Probe PHY and initialise efx->mdio.mode_support, efx->mdio.mmds,
655  *	efx->loopback_modes.
656  * @init: Initialise PHY
657  * @fini: Shut down PHY
658  * @reconfigure: Reconfigure PHY (e.g. for new link parameters)
659  * @poll: Update @link_state and report whether it changed.
660  *	Serialised by the mac_lock.
661  * @get_link_ksettings: Get ethtool settings. Serialised by the mac_lock.
662  * @set_link_ksettings: Set ethtool settings. Serialised by the mac_lock.
663  * @get_fecparam: Get Forward Error Correction settings. Serialised by mac_lock.
664  * @set_fecparam: Set Forward Error Correction settings. Serialised by mac_lock.
665  * @set_npage_adv: Set abilities advertised in (Extended) Next Page
666  *	(only needed where AN bit is set in mmds)
667  * @test_alive: Test that PHY is 'alive' (online)
668  * @test_name: Get the name of a PHY-specific test/result
669  * @run_tests: Run tests and record results as appropriate (offline).
670  *	Flags are the ethtool tests flags.
671  */
672 struct efx_phy_operations {
673 	int (*probe) (struct efx_nic *efx);
674 	int (*init) (struct efx_nic *efx);
675 	void (*fini) (struct efx_nic *efx);
676 	void (*remove) (struct efx_nic *efx);
677 	int (*reconfigure) (struct efx_nic *efx);
678 	bool (*poll) (struct efx_nic *efx);
679 	void (*get_link_ksettings)(struct efx_nic *efx,
680 				   struct ethtool_link_ksettings *cmd);
681 	int (*set_link_ksettings)(struct efx_nic *efx,
682 				  const struct ethtool_link_ksettings *cmd);
683 	int (*get_fecparam)(struct efx_nic *efx, struct ethtool_fecparam *fec);
684 	int (*set_fecparam)(struct efx_nic *efx,
685 			    const struct ethtool_fecparam *fec);
686 	void (*set_npage_adv) (struct efx_nic *efx, u32);
687 	int (*test_alive) (struct efx_nic *efx);
688 	const char *(*test_name) (struct efx_nic *efx, unsigned int index);
689 	int (*run_tests) (struct efx_nic *efx, int *results, unsigned flags);
690 	int (*get_module_eeprom) (struct efx_nic *efx,
691 			       struct ethtool_eeprom *ee,
692 			       u8 *data);
693 	int (*get_module_info) (struct efx_nic *efx,
694 				struct ethtool_modinfo *modinfo);
695 };
696 
697 /**
698  * enum efx_phy_mode - PHY operating mode flags
699  * @PHY_MODE_NORMAL: on and should pass traffic
700  * @PHY_MODE_TX_DISABLED: on with TX disabled
701  * @PHY_MODE_LOW_POWER: set to low power through MDIO
702  * @PHY_MODE_OFF: switched off through external control
703  * @PHY_MODE_SPECIAL: on but will not pass traffic
704  */
705 enum efx_phy_mode {
706 	PHY_MODE_NORMAL		= 0,
707 	PHY_MODE_TX_DISABLED	= 1,
708 	PHY_MODE_LOW_POWER	= 2,
709 	PHY_MODE_OFF		= 4,
710 	PHY_MODE_SPECIAL	= 8,
711 };
712 
713 static inline bool efx_phy_mode_disabled(enum efx_phy_mode mode)
714 {
715 	return !!(mode & ~PHY_MODE_TX_DISABLED);
716 }
717 
718 /**
719  * struct efx_hw_stat_desc - Description of a hardware statistic
720  * @name: Name of the statistic as visible through ethtool, or %NULL if
721  *	it should not be exposed
722  * @dma_width: Width in bits (0 for non-DMA statistics)
723  * @offset: Offset within stats (ignored for non-DMA statistics)
724  */
725 struct efx_hw_stat_desc {
726 	const char *name;
727 	u16 dma_width;
728 	u16 offset;
729 };
730 
731 /* Number of bits used in a multicast filter hash address */
732 #define EFX_MCAST_HASH_BITS 8
733 
734 /* Number of (single-bit) entries in a multicast filter hash */
735 #define EFX_MCAST_HASH_ENTRIES (1 << EFX_MCAST_HASH_BITS)
736 
737 /* An Efx multicast filter hash */
738 union efx_multicast_hash {
739 	u8 byte[EFX_MCAST_HASH_ENTRIES / 8];
740 	efx_oword_t oword[EFX_MCAST_HASH_ENTRIES / sizeof(efx_oword_t) / 8];
741 };
742 
743 struct vfdi_status;
744 
745 /* The reserved RSS context value */
746 #define EFX_EF10_RSS_CONTEXT_INVALID	0xffffffff
747 /**
748  * struct efx_rss_context - A user-defined RSS context for filtering
749  * @list: node of linked list on which this struct is stored
750  * @context_id: the RSS_CONTEXT_ID returned by MC firmware, or
751  *	%EFX_EF10_RSS_CONTEXT_INVALID if this context is not present on the NIC.
752  *	For Siena, 0 if RSS is active, else %EFX_EF10_RSS_CONTEXT_INVALID.
753  * @user_id: the rss_context ID exposed to userspace over ethtool.
754  * @rx_hash_udp_4tuple: UDP 4-tuple hashing enabled
755  * @rx_hash_key: Toeplitz hash key for this RSS context
756  * @indir_table: Indirection table for this RSS context
757  */
758 struct efx_rss_context {
759 	struct list_head list;
760 	u32 context_id;
761 	u32 user_id;
762 	bool rx_hash_udp_4tuple;
763 	u8 rx_hash_key[40];
764 	u32 rx_indir_table[128];
765 };
766 
767 #ifdef CONFIG_RFS_ACCEL
768 /* Order of these is important, since filter_id >= %EFX_ARFS_FILTER_ID_PENDING
769  * is used to test if filter does or will exist.
770  */
771 #define EFX_ARFS_FILTER_ID_PENDING	-1
772 #define EFX_ARFS_FILTER_ID_ERROR	-2
773 #define EFX_ARFS_FILTER_ID_REMOVING	-3
774 /**
775  * struct efx_arfs_rule - record of an ARFS filter and its IDs
776  * @node: linkage into hash table
777  * @spec: details of the filter (used as key for hash table).  Use efx->type to
778  *	determine which member to use.
779  * @rxq_index: channel to which the filter will steer traffic.
780  * @arfs_id: filter ID which was returned to ARFS
781  * @filter_id: index in software filter table.  May be
782  *	%EFX_ARFS_FILTER_ID_PENDING if filter was not inserted yet,
783  *	%EFX_ARFS_FILTER_ID_ERROR if filter insertion failed, or
784  *	%EFX_ARFS_FILTER_ID_REMOVING if expiry is currently removing the filter.
785  */
786 struct efx_arfs_rule {
787 	struct hlist_node node;
788 	struct efx_filter_spec spec;
789 	u16 rxq_index;
790 	u16 arfs_id;
791 	s32 filter_id;
792 };
793 
794 /* Size chosen so that the table is one page (4kB) */
795 #define EFX_ARFS_HASH_TABLE_SIZE	512
796 
797 /**
798  * struct efx_async_filter_insertion - Request to asynchronously insert a filter
799  * @net_dev: Reference to the netdevice
800  * @spec: The filter to insert
801  * @work: Workitem for this request
802  * @rxq_index: Identifies the channel for which this request was made
803  * @flow_id: Identifies the kernel-side flow for which this request was made
804  */
805 struct efx_async_filter_insertion {
806 	struct net_device *net_dev;
807 	struct efx_filter_spec spec;
808 	struct work_struct work;
809 	u16 rxq_index;
810 	u32 flow_id;
811 };
812 
813 /* Maximum number of ARFS workitems that may be in flight on an efx_nic */
814 #define EFX_RPS_MAX_IN_FLIGHT	8
815 #endif /* CONFIG_RFS_ACCEL */
816 
817 /**
818  * struct efx_nic - an Efx NIC
819  * @name: Device name (net device name or bus id before net device registered)
820  * @pci_dev: The PCI device
821  * @node: List node for maintaning primary/secondary function lists
822  * @primary: &struct efx_nic instance for the primary function of this
823  *	controller.  May be the same structure, and may be %NULL if no
824  *	primary function is bound.  Serialised by rtnl_lock.
825  * @secondary_list: List of &struct efx_nic instances for the secondary PCI
826  *	functions of the controller, if this is for the primary function.
827  *	Serialised by rtnl_lock.
828  * @type: Controller type attributes
829  * @legacy_irq: IRQ number
830  * @workqueue: Workqueue for port reconfigures and the HW monitor.
831  *	Work items do not hold and must not acquire RTNL.
832  * @workqueue_name: Name of workqueue
833  * @reset_work: Scheduled reset workitem
834  * @membase_phys: Memory BAR value as physical address
835  * @membase: Memory BAR value
836  * @vi_stride: step between per-VI registers / memory regions
837  * @interrupt_mode: Interrupt mode
838  * @timer_quantum_ns: Interrupt timer quantum, in nanoseconds
839  * @timer_max_ns: Interrupt timer maximum value, in nanoseconds
840  * @irq_rx_adaptive: Adaptive IRQ moderation enabled for RX event queues
841  * @irq_rx_mod_step_us: Step size for IRQ moderation for RX event queues
842  * @irq_rx_moderation_us: IRQ moderation time for RX event queues
843  * @msg_enable: Log message enable flags
844  * @state: Device state number (%STATE_*). Serialised by the rtnl_lock.
845  * @reset_pending: Bitmask for pending resets
846  * @tx_queue: TX DMA queues
847  * @rx_queue: RX DMA queues
848  * @channel: Channels
849  * @msi_context: Context for each MSI
850  * @extra_channel_types: Types of extra (non-traffic) channels that
851  *	should be allocated for this NIC
852  * @xdp_tx_queue_count: Number of entries in %xdp_tx_queues.
853  * @xdp_tx_queues: Array of pointers to tx queues used for XDP transmit.
854  * @rxq_entries: Size of receive queues requested by user.
855  * @txq_entries: Size of transmit queues requested by user.
856  * @txq_stop_thresh: TX queue fill level at or above which we stop it.
857  * @txq_wake_thresh: TX queue fill level at or below which we wake it.
858  * @tx_dc_base: Base qword address in SRAM of TX queue descriptor caches
859  * @rx_dc_base: Base qword address in SRAM of RX queue descriptor caches
860  * @sram_lim_qw: Qword address limit of SRAM
861  * @next_buffer_table: First available buffer table id
862  * @n_channels: Number of channels in use
863  * @n_rx_channels: Number of channels used for RX (= number of RX queues)
864  * @n_tx_channels: Number of channels used for TX
865  * @n_extra_tx_channels: Number of extra channels with TX queues
866  * @n_xdp_channels: Number of channels used for XDP TX
867  * @xdp_channel_offset: Offset of zeroth channel used for XPD TX.
868  * @xdp_tx_per_channel: Max number of TX queues on an XDP TX channel.
869  * @rx_ip_align: RX DMA address offset to have IP header aligned in
870  *	in accordance with NET_IP_ALIGN
871  * @rx_dma_len: Current maximum RX DMA length
872  * @rx_buffer_order: Order (log2) of number of pages for each RX buffer
873  * @rx_buffer_truesize: Amortised allocation size of an RX buffer,
874  *	for use in sk_buff::truesize
875  * @rx_prefix_size: Size of RX prefix before packet data
876  * @rx_packet_hash_offset: Offset of RX flow hash from start of packet data
877  *	(valid only if @rx_prefix_size != 0; always negative)
878  * @rx_packet_len_offset: Offset of RX packet length from start of packet data
879  *	(valid only for NICs that set %EFX_RX_PKT_PREFIX_LEN; always negative)
880  * @rx_packet_ts_offset: Offset of timestamp from start of packet data
881  *	(valid only if channel->sync_timestamps_enabled; always negative)
882  * @rx_scatter: Scatter mode enabled for receives
883  * @rss_context: Main RSS context.  Its @list member is the head of the list of
884  *	RSS contexts created by user requests
885  * @rss_lock: Protects custom RSS context software state in @rss_context.list
886  * @int_error_count: Number of internal errors seen recently
887  * @int_error_expire: Time at which error count will be expired
888  * @irq_soft_enabled: Are IRQs soft-enabled? If not, IRQ handler will
889  *	acknowledge but do nothing else.
890  * @irq_status: Interrupt status buffer
891  * @irq_zero_count: Number of legacy IRQs seen with queue flags == 0
892  * @irq_level: IRQ level/index for IRQs not triggered by an event queue
893  * @selftest_work: Work item for asynchronous self-test
894  * @mtd_list: List of MTDs attached to the NIC
895  * @nic_data: Hardware dependent state
896  * @mcdi: Management-Controller-to-Driver Interface state
897  * @mac_lock: MAC access lock. Protects @port_enabled, @phy_mode,
898  *	efx_monitor() and efx_reconfigure_port()
899  * @port_enabled: Port enabled indicator.
900  *	Serialises efx_stop_all(), efx_start_all(), efx_monitor() and
901  *	efx_mac_work() with kernel interfaces. Safe to read under any
902  *	one of the rtnl_lock, mac_lock, or netif_tx_lock, but all three must
903  *	be held to modify it.
904  * @port_initialized: Port initialized?
905  * @net_dev: Operating system network device. Consider holding the rtnl lock
906  * @fixed_features: Features which cannot be turned off
907  * @num_mac_stats: Number of MAC stats reported by firmware (MAC_STATS_NUM_STATS
908  *	field of %MC_CMD_GET_CAPABILITIES_V4 response, or %MC_CMD_MAC_NSTATS)
909  * @stats_buffer: DMA buffer for statistics
910  * @phy_type: PHY type
911  * @phy_op: PHY interface
912  * @phy_data: PHY private data (including PHY-specific stats)
913  * @mdio: PHY MDIO interface
914  * @mdio_bus: PHY MDIO bus ID (only used by Siena)
915  * @phy_mode: PHY operating mode. Serialised by @mac_lock.
916  * @link_advertising: Autonegotiation advertising flags
917  * @fec_config: Forward Error Correction configuration flags.  For bit positions
918  *	see &enum ethtool_fec_config_bits.
919  * @link_state: Current state of the link
920  * @n_link_state_changes: Number of times the link has changed state
921  * @unicast_filter: Flag for Falcon-arch simple unicast filter.
922  *	Protected by @mac_lock.
923  * @multicast_hash: Multicast hash table for Falcon-arch.
924  *	Protected by @mac_lock.
925  * @wanted_fc: Wanted flow control flags
926  * @fc_disable: When non-zero flow control is disabled. Typically used to
927  *	ensure that network back pressure doesn't delay dma queue flushes.
928  *	Serialised by the rtnl lock.
929  * @mac_work: Work item for changing MAC promiscuity and multicast hash
930  * @loopback_mode: Loopback status
931  * @loopback_modes: Supported loopback mode bitmask
932  * @loopback_selftest: Offline self-test private state
933  * @xdp_prog: Current XDP programme for this interface
934  * @filter_sem: Filter table rw_semaphore, protects existence of @filter_state
935  * @filter_state: Architecture-dependent filter table state
936  * @rps_mutex: Protects RPS state of all channels
937  * @rps_slot_map: bitmap of in-flight entries in @rps_slot
938  * @rps_slot: array of ARFS insertion requests for efx_filter_rfs_work()
939  * @rps_hash_lock: Protects ARFS filter mapping state (@rps_hash_table and
940  *	@rps_next_id).
941  * @rps_hash_table: Mapping between ARFS filters and their various IDs
942  * @rps_next_id: next arfs_id for an ARFS filter
943  * @active_queues: Count of RX and TX queues that haven't been flushed and drained.
944  * @rxq_flush_pending: Count of number of receive queues that need to be flushed.
945  *	Decremented when the efx_flush_rx_queue() is called.
946  * @rxq_flush_outstanding: Count of number of RX flushes started but not yet
947  *	completed (either success or failure). Not used when MCDI is used to
948  *	flush receive queues.
949  * @flush_wq: wait queue used by efx_nic_flush_queues() to wait for flush completions.
950  * @vf_count: Number of VFs intended to be enabled.
951  * @vf_init_count: Number of VFs that have been fully initialised.
952  * @vi_scale: log2 number of vnics per VF.
953  * @ptp_data: PTP state data
954  * @ptp_warned: has this NIC seen and warned about unexpected PTP events?
955  * @vpd_sn: Serial number read from VPD
956  * @xdp_rxq_info_failed: Have any of the rx queues failed to initialise their
957  *      xdp_rxq_info structures?
958  * @monitor_work: Hardware monitor workitem
959  * @biu_lock: BIU (bus interface unit) lock
960  * @last_irq_cpu: Last CPU to handle a possible test interrupt.  This
961  *	field is used by efx_test_interrupts() to verify that an
962  *	interrupt has occurred.
963  * @stats_lock: Statistics update lock. Must be held when calling
964  *	efx_nic_type::{update,start,stop}_stats.
965  * @n_rx_noskb_drops: Count of RX packets dropped due to failure to allocate an skb
966  *
967  * This is stored in the private area of the &struct net_device.
968  */
969 struct efx_nic {
970 	/* The following fields should be written very rarely */
971 
972 	char name[IFNAMSIZ];
973 	struct list_head node;
974 	struct efx_nic *primary;
975 	struct list_head secondary_list;
976 	struct pci_dev *pci_dev;
977 	unsigned int port_num;
978 	const struct efx_nic_type *type;
979 	int legacy_irq;
980 	bool eeh_disabled_legacy_irq;
981 	struct workqueue_struct *workqueue;
982 	char workqueue_name[16];
983 	struct work_struct reset_work;
984 	resource_size_t membase_phys;
985 	void __iomem *membase;
986 
987 	unsigned int vi_stride;
988 
989 	enum efx_int_mode interrupt_mode;
990 	unsigned int timer_quantum_ns;
991 	unsigned int timer_max_ns;
992 	bool irq_rx_adaptive;
993 	unsigned int irq_mod_step_us;
994 	unsigned int irq_rx_moderation_us;
995 	u32 msg_enable;
996 
997 	enum nic_state state;
998 	unsigned long reset_pending;
999 
1000 	struct efx_channel *channel[EFX_MAX_CHANNELS];
1001 	struct efx_msi_context msi_context[EFX_MAX_CHANNELS];
1002 	const struct efx_channel_type *
1003 	extra_channel_type[EFX_MAX_EXTRA_CHANNELS];
1004 
1005 	unsigned int xdp_tx_queue_count;
1006 	struct efx_tx_queue **xdp_tx_queues;
1007 
1008 	unsigned rxq_entries;
1009 	unsigned txq_entries;
1010 	unsigned int txq_stop_thresh;
1011 	unsigned int txq_wake_thresh;
1012 
1013 	unsigned tx_dc_base;
1014 	unsigned rx_dc_base;
1015 	unsigned sram_lim_qw;
1016 	unsigned next_buffer_table;
1017 
1018 	unsigned int max_channels;
1019 	unsigned int max_tx_channels;
1020 	unsigned n_channels;
1021 	unsigned n_rx_channels;
1022 	unsigned rss_spread;
1023 	unsigned tx_channel_offset;
1024 	unsigned n_tx_channels;
1025 	unsigned n_extra_tx_channels;
1026 	unsigned int n_xdp_channels;
1027 	unsigned int xdp_channel_offset;
1028 	unsigned int xdp_tx_per_channel;
1029 	unsigned int rx_ip_align;
1030 	unsigned int rx_dma_len;
1031 	unsigned int rx_buffer_order;
1032 	unsigned int rx_buffer_truesize;
1033 	unsigned int rx_page_buf_step;
1034 	unsigned int rx_bufs_per_page;
1035 	unsigned int rx_pages_per_batch;
1036 	unsigned int rx_prefix_size;
1037 	int rx_packet_hash_offset;
1038 	int rx_packet_len_offset;
1039 	int rx_packet_ts_offset;
1040 	bool rx_scatter;
1041 	struct efx_rss_context rss_context;
1042 	struct mutex rss_lock;
1043 
1044 	unsigned int_error_count;
1045 	unsigned long int_error_expire;
1046 
1047 	bool irq_soft_enabled;
1048 	struct efx_buffer irq_status;
1049 	unsigned irq_zero_count;
1050 	unsigned irq_level;
1051 	struct delayed_work selftest_work;
1052 
1053 #ifdef CONFIG_SFC_MTD
1054 	struct list_head mtd_list;
1055 #endif
1056 
1057 	void *nic_data;
1058 	struct efx_mcdi_data *mcdi;
1059 
1060 	struct mutex mac_lock;
1061 	struct work_struct mac_work;
1062 	bool port_enabled;
1063 
1064 	bool mc_bist_for_other_fn;
1065 	bool port_initialized;
1066 	struct net_device *net_dev;
1067 
1068 	netdev_features_t fixed_features;
1069 
1070 	u16 num_mac_stats;
1071 	struct efx_buffer stats_buffer;
1072 	u64 rx_nodesc_drops_total;
1073 	u64 rx_nodesc_drops_while_down;
1074 	bool rx_nodesc_drops_prev_state;
1075 
1076 	unsigned int phy_type;
1077 	const struct efx_phy_operations *phy_op;
1078 	void *phy_data;
1079 	struct mdio_if_info mdio;
1080 	unsigned int mdio_bus;
1081 	enum efx_phy_mode phy_mode;
1082 
1083 	__ETHTOOL_DECLARE_LINK_MODE_MASK(link_advertising);
1084 	u32 fec_config;
1085 	struct efx_link_state link_state;
1086 	unsigned int n_link_state_changes;
1087 
1088 	bool unicast_filter;
1089 	union efx_multicast_hash multicast_hash;
1090 	u8 wanted_fc;
1091 	unsigned fc_disable;
1092 
1093 	atomic_t rx_reset;
1094 	enum efx_loopback_mode loopback_mode;
1095 	u64 loopback_modes;
1096 
1097 	void *loopback_selftest;
1098 	/* We access loopback_selftest immediately before running XDP,
1099 	 * so we want them next to each other.
1100 	 */
1101 	struct bpf_prog __rcu *xdp_prog;
1102 
1103 	struct rw_semaphore filter_sem;
1104 	void *filter_state;
1105 #ifdef CONFIG_RFS_ACCEL
1106 	struct mutex rps_mutex;
1107 	unsigned long rps_slot_map;
1108 	struct efx_async_filter_insertion rps_slot[EFX_RPS_MAX_IN_FLIGHT];
1109 	spinlock_t rps_hash_lock;
1110 	struct hlist_head *rps_hash_table;
1111 	u32 rps_next_id;
1112 #endif
1113 
1114 	atomic_t active_queues;
1115 	atomic_t rxq_flush_pending;
1116 	atomic_t rxq_flush_outstanding;
1117 	wait_queue_head_t flush_wq;
1118 
1119 #ifdef CONFIG_SFC_SRIOV
1120 	unsigned vf_count;
1121 	unsigned vf_init_count;
1122 	unsigned vi_scale;
1123 #endif
1124 
1125 	struct efx_ptp_data *ptp_data;
1126 	bool ptp_warned;
1127 
1128 	char *vpd_sn;
1129 	bool xdp_rxq_info_failed;
1130 
1131 	/* The following fields may be written more often */
1132 
1133 	struct delayed_work monitor_work ____cacheline_aligned_in_smp;
1134 	spinlock_t biu_lock;
1135 	int last_irq_cpu;
1136 	spinlock_t stats_lock;
1137 	atomic_t n_rx_noskb_drops;
1138 };
1139 
1140 static inline int efx_dev_registered(struct efx_nic *efx)
1141 {
1142 	return efx->net_dev->reg_state == NETREG_REGISTERED;
1143 }
1144 
1145 static inline unsigned int efx_port_num(struct efx_nic *efx)
1146 {
1147 	return efx->port_num;
1148 }
1149 
1150 struct efx_mtd_partition {
1151 	struct list_head node;
1152 	struct mtd_info mtd;
1153 	const char *dev_type_name;
1154 	const char *type_name;
1155 	char name[IFNAMSIZ + 20];
1156 };
1157 
1158 struct efx_udp_tunnel {
1159 	u16 type; /* TUNNEL_ENCAP_UDP_PORT_ENTRY_foo, see mcdi_pcol.h */
1160 	__be16 port;
1161 	/* Count of repeated adds of the same port.  Used only inside the list,
1162 	 * not in request arguments.
1163 	 */
1164 	u16 count;
1165 };
1166 
1167 /**
1168  * struct efx_nic_type - Efx device type definition
1169  * @mem_bar: Get the memory BAR
1170  * @mem_map_size: Get memory BAR mapped size
1171  * @probe: Probe the controller
1172  * @remove: Free resources allocated by probe()
1173  * @init: Initialise the controller
1174  * @dimension_resources: Dimension controller resources (buffer table,
1175  *	and VIs once the available interrupt resources are clear)
1176  * @fini: Shut down the controller
1177  * @monitor: Periodic function for polling link state and hardware monitor
1178  * @map_reset_reason: Map ethtool reset reason to a reset method
1179  * @map_reset_flags: Map ethtool reset flags to a reset method, if possible
1180  * @reset: Reset the controller hardware and possibly the PHY.  This will
1181  *	be called while the controller is uninitialised.
1182  * @probe_port: Probe the MAC and PHY
1183  * @remove_port: Free resources allocated by probe_port()
1184  * @handle_global_event: Handle a "global" event (may be %NULL)
1185  * @fini_dmaq: Flush and finalise DMA queues (RX and TX queues)
1186  * @prepare_flush: Prepare the hardware for flushing the DMA queues
1187  *	(for Falcon architecture)
1188  * @finish_flush: Clean up after flushing the DMA queues (for Falcon
1189  *	architecture)
1190  * @prepare_flr: Prepare for an FLR
1191  * @finish_flr: Clean up after an FLR
1192  * @describe_stats: Describe statistics for ethtool
1193  * @update_stats: Update statistics not provided by event handling.
1194  *	Either argument may be %NULL.
1195  * @start_stats: Start the regular fetching of statistics
1196  * @pull_stats: Pull stats from the NIC and wait until they arrive.
1197  * @stop_stats: Stop the regular fetching of statistics
1198  * @set_id_led: Set state of identifying LED or revert to automatic function
1199  * @push_irq_moderation: Apply interrupt moderation value
1200  * @reconfigure_port: Push loopback/power/txdis changes to the MAC and PHY
1201  * @prepare_enable_fc_tx: Prepare MAC to enable pause frame TX (may be %NULL)
1202  * @reconfigure_mac: Push MAC address, MTU, flow control and filter settings
1203  *	to the hardware.  Serialised by the mac_lock.
1204  * @check_mac_fault: Check MAC fault state. True if fault present.
1205  * @get_wol: Get WoL configuration from driver state
1206  * @set_wol: Push WoL configuration to the NIC
1207  * @resume_wol: Synchronise WoL state between driver and MC (e.g. after resume)
1208  * @test_chip: Test registers.  May use efx_farch_test_registers(), and is
1209  *	expected to reset the NIC.
1210  * @test_nvram: Test validity of NVRAM contents
1211  * @mcdi_request: Send an MCDI request with the given header and SDU.
1212  *	The SDU length may be any value from 0 up to the protocol-
1213  *	defined maximum, but its buffer will be padded to a multiple
1214  *	of 4 bytes.
1215  * @mcdi_poll_response: Test whether an MCDI response is available.
1216  * @mcdi_read_response: Read the MCDI response PDU.  The offset will
1217  *	be a multiple of 4.  The length may not be, but the buffer
1218  *	will be padded so it is safe to round up.
1219  * @mcdi_poll_reboot: Test whether the MCDI has rebooted.  If so,
1220  *	return an appropriate error code for aborting any current
1221  *	request; otherwise return 0.
1222  * @irq_enable_master: Enable IRQs on the NIC.  Each event queue must
1223  *	be separately enabled after this.
1224  * @irq_test_generate: Generate a test IRQ
1225  * @irq_disable_non_ev: Disable non-event IRQs on the NIC.  Each event
1226  *	queue must be separately disabled before this.
1227  * @irq_handle_msi: Handle MSI for a channel.  The @dev_id argument is
1228  *	a pointer to the &struct efx_msi_context for the channel.
1229  * @irq_handle_legacy: Handle legacy interrupt.  The @dev_id argument
1230  *	is a pointer to the &struct efx_nic.
1231  * @tx_probe: Allocate resources for TX queue
1232  * @tx_init: Initialise TX queue on the NIC
1233  * @tx_remove: Free resources for TX queue
1234  * @tx_write: Write TX descriptors and doorbell
1235  * @rx_push_rss_config: Write RSS hash key and indirection table to the NIC
1236  * @rx_pull_rss_config: Read RSS hash key and indirection table back from the NIC
1237  * @rx_push_rss_context_config: Write RSS hash key and indirection table for
1238  *	user RSS context to the NIC
1239  * @rx_pull_rss_context_config: Read RSS hash key and indirection table for user
1240  *	RSS context back from the NIC
1241  * @rx_probe: Allocate resources for RX queue
1242  * @rx_init: Initialise RX queue on the NIC
1243  * @rx_remove: Free resources for RX queue
1244  * @rx_write: Write RX descriptors and doorbell
1245  * @rx_defer_refill: Generate a refill reminder event
1246  * @ev_probe: Allocate resources for event queue
1247  * @ev_init: Initialise event queue on the NIC
1248  * @ev_fini: Deinitialise event queue on the NIC
1249  * @ev_remove: Free resources for event queue
1250  * @ev_process: Process events for a queue, up to the given NAPI quota
1251  * @ev_read_ack: Acknowledge read events on a queue, rearming its IRQ
1252  * @ev_test_generate: Generate a test event
1253  * @filter_table_probe: Probe filter capabilities and set up filter software state
1254  * @filter_table_restore: Restore filters removed from hardware
1255  * @filter_table_remove: Remove filters from hardware and tear down software state
1256  * @filter_update_rx_scatter: Update filters after change to rx scatter setting
1257  * @filter_insert: add or replace a filter
1258  * @filter_remove_safe: remove a filter by ID, carefully
1259  * @filter_get_safe: retrieve a filter by ID, carefully
1260  * @filter_clear_rx: Remove all RX filters whose priority is less than or
1261  *	equal to the given priority and is not %EFX_FILTER_PRI_AUTO
1262  * @filter_count_rx_used: Get the number of filters in use at a given priority
1263  * @filter_get_rx_id_limit: Get maximum value of a filter id, plus 1
1264  * @filter_get_rx_ids: Get list of RX filters at a given priority
1265  * @filter_rfs_expire_one: Consider expiring a filter inserted for RFS.
1266  *	This must check whether the specified table entry is used by RFS
1267  *	and that rps_may_expire_flow() returns true for it.
1268  * @mtd_probe: Probe and add MTD partitions associated with this net device,
1269  *	 using efx_mtd_add()
1270  * @mtd_rename: Set an MTD partition name using the net device name
1271  * @mtd_read: Read from an MTD partition
1272  * @mtd_erase: Erase part of an MTD partition
1273  * @mtd_write: Write to an MTD partition
1274  * @mtd_sync: Wait for write-back to complete on MTD partition.  This
1275  *	also notifies the driver that a writer has finished using this
1276  *	partition.
1277  * @ptp_write_host_time: Send host time to MC as part of sync protocol
1278  * @ptp_set_ts_sync_events: Enable or disable sync events for inline RX
1279  *	timestamping, possibly only temporarily for the purposes of a reset.
1280  * @ptp_set_ts_config: Set hardware timestamp configuration.  The flags
1281  *	and tx_type will already have been validated but this operation
1282  *	must validate and update rx_filter.
1283  * @get_phys_port_id: Get the underlying physical port id.
1284  * @set_mac_address: Set the MAC address of the device
1285  * @tso_versions: Returns mask of firmware-assisted TSO versions supported.
1286  *	If %NULL, then device does not support any TSO version.
1287  * @udp_tnl_push_ports: Push the list of UDP tunnel ports to the NIC if required.
1288  * @udp_tnl_add_port: Add a UDP tunnel port
1289  * @udp_tnl_has_port: Check if a port has been added as UDP tunnel
1290  * @udp_tnl_del_port: Remove a UDP tunnel port
1291  * @revision: Hardware architecture revision
1292  * @txd_ptr_tbl_base: TX descriptor ring base address
1293  * @rxd_ptr_tbl_base: RX descriptor ring base address
1294  * @buf_tbl_base: Buffer table base address
1295  * @evq_ptr_tbl_base: Event queue pointer table base address
1296  * @evq_rptr_tbl_base: Event queue read-pointer table base address
1297  * @max_dma_mask: Maximum possible DMA mask
1298  * @rx_prefix_size: Size of RX prefix before packet data
1299  * @rx_hash_offset: Offset of RX flow hash within prefix
1300  * @rx_ts_offset: Offset of timestamp within prefix
1301  * @rx_buffer_padding: Size of padding at end of RX packet
1302  * @can_rx_scatter: NIC is able to scatter packets to multiple buffers
1303  * @always_rx_scatter: NIC will always scatter packets to multiple buffers
1304  * @option_descriptors: NIC supports TX option descriptors
1305  * @min_interrupt_mode: Lowest capability interrupt mode supported
1306  *	from &enum efx_int_mode.
1307  * @max_interrupt_mode: Highest capability interrupt mode supported
1308  *	from &enum efx_int_mode.
1309  * @timer_period_max: Maximum period of interrupt timer (in ticks)
1310  * @offload_features: net_device feature flags for protocol offload
1311  *	features implemented in hardware
1312  * @mcdi_max_ver: Maximum MCDI version supported
1313  * @hwtstamp_filters: Mask of hardware timestamp filter types supported
1314  */
1315 struct efx_nic_type {
1316 	bool is_vf;
1317 	unsigned int (*mem_bar)(struct efx_nic *efx);
1318 	unsigned int (*mem_map_size)(struct efx_nic *efx);
1319 	int (*probe)(struct efx_nic *efx);
1320 	void (*remove)(struct efx_nic *efx);
1321 	int (*init)(struct efx_nic *efx);
1322 	int (*dimension_resources)(struct efx_nic *efx);
1323 	void (*fini)(struct efx_nic *efx);
1324 	void (*monitor)(struct efx_nic *efx);
1325 	enum reset_type (*map_reset_reason)(enum reset_type reason);
1326 	int (*map_reset_flags)(u32 *flags);
1327 	int (*reset)(struct efx_nic *efx, enum reset_type method);
1328 	int (*probe_port)(struct efx_nic *efx);
1329 	void (*remove_port)(struct efx_nic *efx);
1330 	bool (*handle_global_event)(struct efx_channel *channel, efx_qword_t *);
1331 	int (*fini_dmaq)(struct efx_nic *efx);
1332 	void (*prepare_flush)(struct efx_nic *efx);
1333 	void (*finish_flush)(struct efx_nic *efx);
1334 	void (*prepare_flr)(struct efx_nic *efx);
1335 	void (*finish_flr)(struct efx_nic *efx);
1336 	size_t (*describe_stats)(struct efx_nic *efx, u8 *names);
1337 	size_t (*update_stats)(struct efx_nic *efx, u64 *full_stats,
1338 			       struct rtnl_link_stats64 *core_stats);
1339 	void (*start_stats)(struct efx_nic *efx);
1340 	void (*pull_stats)(struct efx_nic *efx);
1341 	void (*stop_stats)(struct efx_nic *efx);
1342 	void (*set_id_led)(struct efx_nic *efx, enum efx_led_mode mode);
1343 	void (*push_irq_moderation)(struct efx_channel *channel);
1344 	int (*reconfigure_port)(struct efx_nic *efx);
1345 	void (*prepare_enable_fc_tx)(struct efx_nic *efx);
1346 	int (*reconfigure_mac)(struct efx_nic *efx);
1347 	bool (*check_mac_fault)(struct efx_nic *efx);
1348 	void (*get_wol)(struct efx_nic *efx, struct ethtool_wolinfo *wol);
1349 	int (*set_wol)(struct efx_nic *efx, u32 type);
1350 	void (*resume_wol)(struct efx_nic *efx);
1351 	int (*test_chip)(struct efx_nic *efx, struct efx_self_tests *tests);
1352 	int (*test_nvram)(struct efx_nic *efx);
1353 	void (*mcdi_request)(struct efx_nic *efx,
1354 			     const efx_dword_t *hdr, size_t hdr_len,
1355 			     const efx_dword_t *sdu, size_t sdu_len);
1356 	bool (*mcdi_poll_response)(struct efx_nic *efx);
1357 	void (*mcdi_read_response)(struct efx_nic *efx, efx_dword_t *pdu,
1358 				   size_t pdu_offset, size_t pdu_len);
1359 	int (*mcdi_poll_reboot)(struct efx_nic *efx);
1360 	void (*mcdi_reboot_detected)(struct efx_nic *efx);
1361 	void (*irq_enable_master)(struct efx_nic *efx);
1362 	int (*irq_test_generate)(struct efx_nic *efx);
1363 	void (*irq_disable_non_ev)(struct efx_nic *efx);
1364 	irqreturn_t (*irq_handle_msi)(int irq, void *dev_id);
1365 	irqreturn_t (*irq_handle_legacy)(int irq, void *dev_id);
1366 	int (*tx_probe)(struct efx_tx_queue *tx_queue);
1367 	void (*tx_init)(struct efx_tx_queue *tx_queue);
1368 	void (*tx_remove)(struct efx_tx_queue *tx_queue);
1369 	void (*tx_write)(struct efx_tx_queue *tx_queue);
1370 	unsigned int (*tx_limit_len)(struct efx_tx_queue *tx_queue,
1371 				     dma_addr_t dma_addr, unsigned int len);
1372 	int (*rx_push_rss_config)(struct efx_nic *efx, bool user,
1373 				  const u32 *rx_indir_table, const u8 *key);
1374 	int (*rx_pull_rss_config)(struct efx_nic *efx);
1375 	int (*rx_push_rss_context_config)(struct efx_nic *efx,
1376 					  struct efx_rss_context *ctx,
1377 					  const u32 *rx_indir_table,
1378 					  const u8 *key);
1379 	int (*rx_pull_rss_context_config)(struct efx_nic *efx,
1380 					  struct efx_rss_context *ctx);
1381 	void (*rx_restore_rss_contexts)(struct efx_nic *efx);
1382 	int (*rx_probe)(struct efx_rx_queue *rx_queue);
1383 	void (*rx_init)(struct efx_rx_queue *rx_queue);
1384 	void (*rx_remove)(struct efx_rx_queue *rx_queue);
1385 	void (*rx_write)(struct efx_rx_queue *rx_queue);
1386 	void (*rx_defer_refill)(struct efx_rx_queue *rx_queue);
1387 	int (*ev_probe)(struct efx_channel *channel);
1388 	int (*ev_init)(struct efx_channel *channel);
1389 	void (*ev_fini)(struct efx_channel *channel);
1390 	void (*ev_remove)(struct efx_channel *channel);
1391 	int (*ev_process)(struct efx_channel *channel, int quota);
1392 	void (*ev_read_ack)(struct efx_channel *channel);
1393 	void (*ev_test_generate)(struct efx_channel *channel);
1394 	int (*filter_table_probe)(struct efx_nic *efx);
1395 	void (*filter_table_restore)(struct efx_nic *efx);
1396 	void (*filter_table_remove)(struct efx_nic *efx);
1397 	void (*filter_update_rx_scatter)(struct efx_nic *efx);
1398 	s32 (*filter_insert)(struct efx_nic *efx,
1399 			     struct efx_filter_spec *spec, bool replace);
1400 	int (*filter_remove_safe)(struct efx_nic *efx,
1401 				  enum efx_filter_priority priority,
1402 				  u32 filter_id);
1403 	int (*filter_get_safe)(struct efx_nic *efx,
1404 			       enum efx_filter_priority priority,
1405 			       u32 filter_id, struct efx_filter_spec *);
1406 	int (*filter_clear_rx)(struct efx_nic *efx,
1407 			       enum efx_filter_priority priority);
1408 	u32 (*filter_count_rx_used)(struct efx_nic *efx,
1409 				    enum efx_filter_priority priority);
1410 	u32 (*filter_get_rx_id_limit)(struct efx_nic *efx);
1411 	s32 (*filter_get_rx_ids)(struct efx_nic *efx,
1412 				 enum efx_filter_priority priority,
1413 				 u32 *buf, u32 size);
1414 #ifdef CONFIG_RFS_ACCEL
1415 	bool (*filter_rfs_expire_one)(struct efx_nic *efx, u32 flow_id,
1416 				      unsigned int index);
1417 #endif
1418 #ifdef CONFIG_SFC_MTD
1419 	int (*mtd_probe)(struct efx_nic *efx);
1420 	void (*mtd_rename)(struct efx_mtd_partition *part);
1421 	int (*mtd_read)(struct mtd_info *mtd, loff_t start, size_t len,
1422 			size_t *retlen, u8 *buffer);
1423 	int (*mtd_erase)(struct mtd_info *mtd, loff_t start, size_t len);
1424 	int (*mtd_write)(struct mtd_info *mtd, loff_t start, size_t len,
1425 			 size_t *retlen, const u8 *buffer);
1426 	int (*mtd_sync)(struct mtd_info *mtd);
1427 #endif
1428 	void (*ptp_write_host_time)(struct efx_nic *efx, u32 host_time);
1429 	int (*ptp_set_ts_sync_events)(struct efx_nic *efx, bool en, bool temp);
1430 	int (*ptp_set_ts_config)(struct efx_nic *efx,
1431 				 struct hwtstamp_config *init);
1432 	int (*sriov_configure)(struct efx_nic *efx, int num_vfs);
1433 	int (*vlan_rx_add_vid)(struct efx_nic *efx, __be16 proto, u16 vid);
1434 	int (*vlan_rx_kill_vid)(struct efx_nic *efx, __be16 proto, u16 vid);
1435 	int (*get_phys_port_id)(struct efx_nic *efx,
1436 				struct netdev_phys_item_id *ppid);
1437 	int (*sriov_init)(struct efx_nic *efx);
1438 	void (*sriov_fini)(struct efx_nic *efx);
1439 	bool (*sriov_wanted)(struct efx_nic *efx);
1440 	void (*sriov_reset)(struct efx_nic *efx);
1441 	void (*sriov_flr)(struct efx_nic *efx, unsigned vf_i);
1442 	int (*sriov_set_vf_mac)(struct efx_nic *efx, int vf_i, u8 *mac);
1443 	int (*sriov_set_vf_vlan)(struct efx_nic *efx, int vf_i, u16 vlan,
1444 				 u8 qos);
1445 	int (*sriov_set_vf_spoofchk)(struct efx_nic *efx, int vf_i,
1446 				     bool spoofchk);
1447 	int (*sriov_get_vf_config)(struct efx_nic *efx, int vf_i,
1448 				   struct ifla_vf_info *ivi);
1449 	int (*sriov_set_vf_link_state)(struct efx_nic *efx, int vf_i,
1450 				       int link_state);
1451 	int (*vswitching_probe)(struct efx_nic *efx);
1452 	int (*vswitching_restore)(struct efx_nic *efx);
1453 	void (*vswitching_remove)(struct efx_nic *efx);
1454 	int (*get_mac_address)(struct efx_nic *efx, unsigned char *perm_addr);
1455 	int (*set_mac_address)(struct efx_nic *efx);
1456 	u32 (*tso_versions)(struct efx_nic *efx);
1457 	int (*udp_tnl_push_ports)(struct efx_nic *efx);
1458 	int (*udp_tnl_add_port)(struct efx_nic *efx, struct efx_udp_tunnel tnl);
1459 	bool (*udp_tnl_has_port)(struct efx_nic *efx, __be16 port);
1460 	int (*udp_tnl_del_port)(struct efx_nic *efx, struct efx_udp_tunnel tnl);
1461 
1462 	int revision;
1463 	unsigned int txd_ptr_tbl_base;
1464 	unsigned int rxd_ptr_tbl_base;
1465 	unsigned int buf_tbl_base;
1466 	unsigned int evq_ptr_tbl_base;
1467 	unsigned int evq_rptr_tbl_base;
1468 	u64 max_dma_mask;
1469 	unsigned int rx_prefix_size;
1470 	unsigned int rx_hash_offset;
1471 	unsigned int rx_ts_offset;
1472 	unsigned int rx_buffer_padding;
1473 	bool can_rx_scatter;
1474 	bool always_rx_scatter;
1475 	bool option_descriptors;
1476 	unsigned int min_interrupt_mode;
1477 	unsigned int max_interrupt_mode;
1478 	unsigned int timer_period_max;
1479 	netdev_features_t offload_features;
1480 	int mcdi_max_ver;
1481 	unsigned int max_rx_ip_filters;
1482 	u32 hwtstamp_filters;
1483 	unsigned int rx_hash_key_size;
1484 };
1485 
1486 /**************************************************************************
1487  *
1488  * Prototypes and inline functions
1489  *
1490  *************************************************************************/
1491 
1492 static inline struct efx_channel *
1493 efx_get_channel(struct efx_nic *efx, unsigned index)
1494 {
1495 	EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_channels);
1496 	return efx->channel[index];
1497 }
1498 
1499 /* Iterate over all used channels */
1500 #define efx_for_each_channel(_channel, _efx)				\
1501 	for (_channel = (_efx)->channel[0];				\
1502 	     _channel;							\
1503 	     _channel = (_channel->channel + 1 < (_efx)->n_channels) ?	\
1504 		     (_efx)->channel[_channel->channel + 1] : NULL)
1505 
1506 /* Iterate over all used channels in reverse */
1507 #define efx_for_each_channel_rev(_channel, _efx)			\
1508 	for (_channel = (_efx)->channel[(_efx)->n_channels - 1];	\
1509 	     _channel;							\
1510 	     _channel = _channel->channel ?				\
1511 		     (_efx)->channel[_channel->channel - 1] : NULL)
1512 
1513 static inline struct efx_tx_queue *
1514 efx_get_tx_queue(struct efx_nic *efx, unsigned index, unsigned type)
1515 {
1516 	EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_tx_channels ||
1517 				  type >= EFX_TXQ_TYPES);
1518 	return &efx->channel[efx->tx_channel_offset + index]->tx_queue[type];
1519 }
1520 
1521 static inline struct efx_channel *
1522 efx_get_xdp_channel(struct efx_nic *efx, unsigned int index)
1523 {
1524 	EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_xdp_channels);
1525 	return efx->channel[efx->xdp_channel_offset + index];
1526 }
1527 
1528 static inline bool efx_channel_is_xdp_tx(struct efx_channel *channel)
1529 {
1530 	return channel->channel - channel->efx->xdp_channel_offset <
1531 	       channel->efx->n_xdp_channels;
1532 }
1533 
1534 static inline bool efx_channel_has_tx_queues(struct efx_channel *channel)
1535 {
1536 	return efx_channel_is_xdp_tx(channel) ||
1537 	       (channel->type && channel->type->want_txqs &&
1538 		channel->type->want_txqs(channel));
1539 }
1540 
1541 static inline struct efx_tx_queue *
1542 efx_channel_get_tx_queue(struct efx_channel *channel, unsigned type)
1543 {
1544 	EFX_WARN_ON_ONCE_PARANOID(!efx_channel_has_tx_queues(channel) ||
1545 				  type >= EFX_TXQ_TYPES);
1546 	return &channel->tx_queue[type];
1547 }
1548 
1549 static inline bool efx_tx_queue_used(struct efx_tx_queue *tx_queue)
1550 {
1551 	return !(tx_queue->efx->net_dev->num_tc < 2 &&
1552 		 tx_queue->queue & EFX_TXQ_TYPE_HIGHPRI);
1553 }
1554 
1555 /* Iterate over all TX queues belonging to a channel */
1556 #define efx_for_each_channel_tx_queue(_tx_queue, _channel)		\
1557 	if (!efx_channel_has_tx_queues(_channel))			\
1558 		;							\
1559 	else								\
1560 		for (_tx_queue = (_channel)->tx_queue;			\
1561 		     _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES && \
1562 			     (efx_tx_queue_used(_tx_queue) ||            \
1563 			      efx_channel_is_xdp_tx(_channel));		\
1564 		     _tx_queue++)
1565 
1566 /* Iterate over all possible TX queues belonging to a channel */
1567 #define efx_for_each_possible_channel_tx_queue(_tx_queue, _channel)	\
1568 	if (!efx_channel_has_tx_queues(_channel))			\
1569 		;							\
1570 	else								\
1571 		for (_tx_queue = (_channel)->tx_queue;			\
1572 		     _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES;	\
1573 		     _tx_queue++)
1574 
1575 static inline bool efx_channel_has_rx_queue(struct efx_channel *channel)
1576 {
1577 	return channel->rx_queue.core_index >= 0;
1578 }
1579 
1580 static inline struct efx_rx_queue *
1581 efx_channel_get_rx_queue(struct efx_channel *channel)
1582 {
1583 	EFX_WARN_ON_ONCE_PARANOID(!efx_channel_has_rx_queue(channel));
1584 	return &channel->rx_queue;
1585 }
1586 
1587 /* Iterate over all RX queues belonging to a channel */
1588 #define efx_for_each_channel_rx_queue(_rx_queue, _channel)		\
1589 	if (!efx_channel_has_rx_queue(_channel))			\
1590 		;							\
1591 	else								\
1592 		for (_rx_queue = &(_channel)->rx_queue;			\
1593 		     _rx_queue;						\
1594 		     _rx_queue = NULL)
1595 
1596 static inline struct efx_channel *
1597 efx_rx_queue_channel(struct efx_rx_queue *rx_queue)
1598 {
1599 	return container_of(rx_queue, struct efx_channel, rx_queue);
1600 }
1601 
1602 static inline int efx_rx_queue_index(struct efx_rx_queue *rx_queue)
1603 {
1604 	return efx_rx_queue_channel(rx_queue)->channel;
1605 }
1606 
1607 /* Returns a pointer to the specified receive buffer in the RX
1608  * descriptor queue.
1609  */
1610 static inline struct efx_rx_buffer *efx_rx_buffer(struct efx_rx_queue *rx_queue,
1611 						  unsigned int index)
1612 {
1613 	return &rx_queue->buffer[index];
1614 }
1615 
1616 /**
1617  * EFX_MAX_FRAME_LEN - calculate maximum frame length
1618  *
1619  * This calculates the maximum frame length that will be used for a
1620  * given MTU.  The frame length will be equal to the MTU plus a
1621  * constant amount of header space and padding.  This is the quantity
1622  * that the net driver will program into the MAC as the maximum frame
1623  * length.
1624  *
1625  * The 10G MAC requires 8-byte alignment on the frame
1626  * length, so we round up to the nearest 8.
1627  *
1628  * Re-clocking by the XGXS on RX can reduce an IPG to 32 bits (half an
1629  * XGMII cycle).  If the frame length reaches the maximum value in the
1630  * same cycle, the XMAC can miss the IPG altogether.  We work around
1631  * this by adding a further 16 bytes.
1632  */
1633 #define EFX_FRAME_PAD	16
1634 #define EFX_MAX_FRAME_LEN(mtu) \
1635 	(ALIGN(((mtu) + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN + EFX_FRAME_PAD), 8))
1636 
1637 static inline bool efx_xmit_with_hwtstamp(struct sk_buff *skb)
1638 {
1639 	return skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP;
1640 }
1641 static inline void efx_xmit_hwtstamp_pending(struct sk_buff *skb)
1642 {
1643 	skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
1644 }
1645 
1646 /* Get all supported features.
1647  * If a feature is not fixed, it is present in hw_features.
1648  * If a feature is fixed, it does not present in hw_features, but
1649  * always in features.
1650  */
1651 static inline netdev_features_t efx_supported_features(const struct efx_nic *efx)
1652 {
1653 	const struct net_device *net_dev = efx->net_dev;
1654 
1655 	return net_dev->features | net_dev->hw_features;
1656 }
1657 
1658 /* Get the current TX queue insert index. */
1659 static inline unsigned int
1660 efx_tx_queue_get_insert_index(const struct efx_tx_queue *tx_queue)
1661 {
1662 	return tx_queue->insert_count & tx_queue->ptr_mask;
1663 }
1664 
1665 /* Get a TX buffer. */
1666 static inline struct efx_tx_buffer *
1667 __efx_tx_queue_get_insert_buffer(const struct efx_tx_queue *tx_queue)
1668 {
1669 	return &tx_queue->buffer[efx_tx_queue_get_insert_index(tx_queue)];
1670 }
1671 
1672 /* Get a TX buffer, checking it's not currently in use. */
1673 static inline struct efx_tx_buffer *
1674 efx_tx_queue_get_insert_buffer(const struct efx_tx_queue *tx_queue)
1675 {
1676 	struct efx_tx_buffer *buffer =
1677 		__efx_tx_queue_get_insert_buffer(tx_queue);
1678 
1679 	EFX_WARN_ON_ONCE_PARANOID(buffer->len);
1680 	EFX_WARN_ON_ONCE_PARANOID(buffer->flags);
1681 	EFX_WARN_ON_ONCE_PARANOID(buffer->unmap_len);
1682 
1683 	return buffer;
1684 }
1685 
1686 #endif /* EFX_NET_DRIVER_H */
1687