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