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 #define EFX_DRIVER_VERSION	"4.1"
42 
43 #ifdef DEBUG
44 #define EFX_WARN_ON_ONCE_PARANOID(x) WARN_ON_ONCE(x)
45 #define EFX_WARN_ON_PARANOID(x) WARN_ON(x)
46 #else
47 #define EFX_WARN_ON_ONCE_PARANOID(x) do {} while (0)
48 #define EFX_WARN_ON_PARANOID(x) do {} while (0)
49 #endif
50 
51 /**************************************************************************
52  *
53  * Efx data structures
54  *
55  **************************************************************************/
56 
57 #define EFX_MAX_CHANNELS 32U
58 #define EFX_MAX_RX_QUEUES EFX_MAX_CHANNELS
59 #define EFX_EXTRA_CHANNEL_IOV	0
60 #define EFX_EXTRA_CHANNEL_PTP	1
61 #define EFX_MAX_EXTRA_CHANNELS	2U
62 
63 /* Checksum generation is a per-queue option in hardware, so each
64  * queue visible to the networking core is backed by two hardware TX
65  * queues. */
66 #define EFX_MAX_TX_TC		2
67 #define EFX_MAX_CORE_TX_QUEUES	(EFX_MAX_TX_TC * EFX_MAX_CHANNELS)
68 #define EFX_TXQ_TYPE_OFFLOAD	1	/* flag */
69 #define EFX_TXQ_TYPE_HIGHPRI	2	/* flag */
70 #define EFX_TXQ_TYPES		4
71 #define EFX_MAX_TX_QUEUES	(EFX_TXQ_TYPES * EFX_MAX_CHANNELS)
72 
73 /* Maximum possible MTU the driver supports */
74 #define EFX_MAX_MTU (9 * 1024)
75 
76 /* Minimum MTU, from RFC791 (IP) */
77 #define EFX_MIN_MTU 68
78 
79 /* Size of an RX scatter buffer.  Small enough to pack 2 into a 4K page,
80  * and should be a multiple of the cache line size.
81  */
82 #define EFX_RX_USR_BUF_SIZE	(2048 - 256)
83 
84 /* If possible, we should ensure cache line alignment at start and end
85  * of every buffer.  Otherwise, we just need to ensure 4-byte
86  * alignment of the network header.
87  */
88 #if NET_IP_ALIGN == 0
89 #define EFX_RX_BUF_ALIGNMENT	L1_CACHE_BYTES
90 #else
91 #define EFX_RX_BUF_ALIGNMENT	4
92 #endif
93 
94 /* Non-standard XDP_PACKET_HEADROOM and tailroom to satisfy XDP_REDIRECT and
95  * still fit two standard MTU size packets into a single 4K page.
96  */
97 #define EFX_XDP_HEADROOM	128
98 #define EFX_XDP_TAILROOM	SKB_DATA_ALIGN(sizeof(struct skb_shared_info))
99 
100 /* Forward declare Precision Time Protocol (PTP) support structure. */
101 struct efx_ptp_data;
102 struct hwtstamp_config;
103 
104 struct efx_self_tests;
105 
106 /**
107  * struct efx_buffer - A general-purpose DMA buffer
108  * @addr: host base address of the buffer
109  * @dma_addr: DMA base address of the buffer
110  * @len: Buffer length, in bytes
111  *
112  * The NIC uses these buffers for its interrupt status registers and
113  * MAC stats dumps.
114  */
115 struct efx_buffer {
116 	void *addr;
117 	dma_addr_t dma_addr;
118 	unsigned int len;
119 };
120 
121 /**
122  * struct efx_special_buffer - DMA buffer entered into buffer table
123  * @buf: Standard &struct efx_buffer
124  * @index: Buffer index within controller;s buffer table
125  * @entries: Number of buffer table entries
126  *
127  * The NIC has a buffer table that maps buffers of size %EFX_BUF_SIZE.
128  * Event and descriptor rings are addressed via one or more buffer
129  * table entries (and so can be physically non-contiguous, although we
130  * currently do not take advantage of that).  On Falcon and Siena we
131  * have to take care of allocating and initialising the entries
132  * ourselves.  On later hardware this is managed by the firmware and
133  * @index and @entries are left as 0.
134  */
135 struct efx_special_buffer {
136 	struct efx_buffer buf;
137 	unsigned int index;
138 	unsigned int entries;
139 };
140 
141 /**
142  * struct efx_tx_buffer - buffer state for a TX descriptor
143  * @skb: When @flags & %EFX_TX_BUF_SKB, the associated socket buffer to be
144  *	freed when descriptor completes
145  * @xdpf: When @flags & %EFX_TX_BUF_XDP, the XDP frame information; its @data
146  *	member is the associated buffer to drop a page reference on.
147  * @option: When @flags & %EFX_TX_BUF_OPTION, an EF10-specific option
148  *	descriptor.
149  * @dma_addr: DMA address of the fragment.
150  * @flags: Flags for allocation and DMA mapping type
151  * @len: Length of this fragment.
152  *	This field is zero when the queue slot is empty.
153  * @unmap_len: Length of this fragment to unmap
154  * @dma_offset: Offset of @dma_addr from the address of the backing DMA mapping.
155  * Only valid if @unmap_len != 0.
156  */
157 struct efx_tx_buffer {
158 	union {
159 		const struct sk_buff *skb;
160 		struct xdp_frame *xdpf;
161 	};
162 	union {
163 		efx_qword_t option;    /* EF10 */
164 		dma_addr_t dma_addr;
165 	};
166 	unsigned short flags;
167 	unsigned short len;
168 	unsigned short unmap_len;
169 	unsigned short dma_offset;
170 };
171 #define EFX_TX_BUF_CONT		1	/* not last descriptor of packet */
172 #define EFX_TX_BUF_SKB		2	/* buffer is last part of skb */
173 #define EFX_TX_BUF_MAP_SINGLE	8	/* buffer was mapped with dma_map_single() */
174 #define EFX_TX_BUF_OPTION	0x10	/* empty buffer for option descriptor */
175 #define EFX_TX_BUF_XDP		0x20	/* buffer was sent with XDP */
176 
177 /**
178  * struct efx_tx_queue - An Efx TX queue
179  *
180  * This is a ring buffer of TX fragments.
181  * Since the TX completion path always executes on the same
182  * CPU and the xmit path can operate on different CPUs,
183  * performance is increased by ensuring that the completion
184  * path and the xmit path operate on different cache lines.
185  * This is particularly important if the xmit path is always
186  * executing on one CPU which is different from the completion
187  * path.  There is also a cache line for members which are
188  * read but not written on the fast path.
189  *
190  * @efx: The associated Efx NIC
191  * @queue: DMA queue number
192  * @tso_version: Version of TSO in use for this queue.
193  * @channel: The associated channel
194  * @core_txq: The networking core TX queue structure
195  * @buffer: The software buffer ring
196  * @cb_page: Array of pages of copy buffers.  Carved up according to
197  *	%EFX_TX_CB_ORDER into %EFX_TX_CB_SIZE-sized chunks.
198  * @txd: The hardware descriptor ring
199  * @ptr_mask: The size of the ring minus 1.
200  * @piobuf: PIO buffer region for this TX queue (shared with its partner).
201  *	Size of the region is efx_piobuf_size.
202  * @piobuf_offset: Buffer offset to be specified in PIO descriptors
203  * @initialised: Has hardware queue been initialised?
204  * @timestamping: Is timestamping enabled for this channel?
205  * @xdp_tx: Is this an XDP tx queue?
206  * @handle_tso: TSO xmit preparation handler.  Sets up the TSO metadata and
207  *	may also map tx data, depending on the nature of the TSO implementation.
208  * @read_count: Current read pointer.
209  *	This is the number of buffers that have been removed from both rings.
210  * @old_write_count: The value of @write_count when last checked.
211  *	This is here for performance reasons.  The xmit path will
212  *	only get the up-to-date value of @write_count if this
213  *	variable indicates that the queue is empty.  This is to
214  *	avoid cache-line ping-pong between the xmit path and the
215  *	completion path.
216  * @merge_events: Number of TX merged completion events
217  * @completed_timestamp_major: Top part of the most recent tx timestamp.
218  * @completed_timestamp_minor: Low part of the most recent tx timestamp.
219  * @insert_count: Current insert pointer
220  *	This is the number of buffers that have been added to the
221  *	software ring.
222  * @write_count: Current write pointer
223  *	This is the number of buffers that have been added to the
224  *	hardware ring.
225  * @packet_write_count: Completable write pointer
226  *	This is the write pointer of the last packet written.
227  *	Normally this will equal @write_count, but as option descriptors
228  *	don't produce completion events, they won't update this.
229  *	Filled in iff @efx->type->option_descriptors; only used for PIO.
230  *	Thus, this is written and used on EF10, and neither on farch.
231  * @old_read_count: The value of read_count when last checked.
232  *	This is here for performance reasons.  The xmit path will
233  *	only get the up-to-date value of read_count if this
234  *	variable indicates that the queue is full.  This is to
235  *	avoid cache-line ping-pong between the xmit path and the
236  *	completion path.
237  * @tso_bursts: Number of times TSO xmit invoked by kernel
238  * @tso_long_headers: Number of packets with headers too long for standard
239  *	blocks
240  * @tso_packets: Number of packets via the TSO xmit path
241  * @tso_fallbacks: Number of times TSO fallback used
242  * @pushes: Number of times the TX push feature has been used
243  * @pio_packets: Number of times the TX PIO feature has been used
244  * @xmit_more_available: Are any packets waiting to be pushed to the NIC
245  * @cb_packets: Number of times the TX copybreak feature has been used
246  * @empty_read_count: If the completion path has seen the queue as empty
247  *	and the transmission path has not yet checked this, the value of
248  *	@read_count bitwise-added to %EFX_EMPTY_COUNT_VALID; otherwise 0.
249  */
250 struct efx_tx_queue {
251 	/* Members which don't change on the fast path */
252 	struct efx_nic *efx ____cacheline_aligned_in_smp;
253 	unsigned queue;
254 	unsigned int tso_version;
255 	struct efx_channel *channel;
256 	struct netdev_queue *core_txq;
257 	struct efx_tx_buffer *buffer;
258 	struct efx_buffer *cb_page;
259 	struct efx_special_buffer txd;
260 	unsigned int ptr_mask;
261 	void __iomem *piobuf;
262 	unsigned int piobuf_offset;
263 	bool initialised;
264 	bool timestamping;
265 	bool xdp_tx;
266 
267 	/* Function pointers used in the fast path. */
268 	int (*handle_tso)(struct efx_tx_queue*, struct sk_buff*, bool *);
269 
270 	/* Members used mainly on the completion path */
271 	unsigned int read_count ____cacheline_aligned_in_smp;
272 	unsigned int old_write_count;
273 	unsigned int merge_events;
274 	unsigned int bytes_compl;
275 	unsigned int pkts_compl;
276 	u32 completed_timestamp_major;
277 	u32 completed_timestamp_minor;
278 
279 	/* Members used only on the xmit path */
280 	unsigned int insert_count ____cacheline_aligned_in_smp;
281 	unsigned int write_count;
282 	unsigned int packet_write_count;
283 	unsigned int old_read_count;
284 	unsigned int tso_bursts;
285 	unsigned int tso_long_headers;
286 	unsigned int tso_packets;
287 	unsigned int tso_fallbacks;
288 	unsigned int pushes;
289 	unsigned int pio_packets;
290 	bool xmit_more_available;
291 	unsigned int cb_packets;
292 	/* Statistics to supplement MAC stats */
293 	unsigned long tx_packets;
294 
295 	/* Members shared between paths and sometimes updated */
296 	unsigned int empty_read_count ____cacheline_aligned_in_smp;
297 #define EFX_EMPTY_COUNT_VALID 0x80000000
298 	atomic_t flush_outstanding;
299 };
300 
301 #define EFX_TX_CB_ORDER	7
302 #define EFX_TX_CB_SIZE	(1 << EFX_TX_CB_ORDER) - NET_IP_ALIGN
303 
304 /**
305  * struct efx_rx_buffer - An Efx RX data buffer
306  * @dma_addr: DMA base address of the buffer
307  * @page: The associated page buffer.
308  *	Will be %NULL if the buffer slot is currently free.
309  * @page_offset: If pending: offset in @page of DMA base address.
310  *	If completed: offset in @page of Ethernet header.
311  * @len: If pending: length for DMA descriptor.
312  *	If completed: received length, excluding hash prefix.
313  * @flags: Flags for buffer and packet state.  These are only set on the
314  *	first buffer of a scattered packet.
315  */
316 struct efx_rx_buffer {
317 	dma_addr_t dma_addr;
318 	struct page *page;
319 	u16 page_offset;
320 	u16 len;
321 	u16 flags;
322 };
323 #define EFX_RX_BUF_LAST_IN_PAGE	0x0001
324 #define EFX_RX_PKT_CSUMMED	0x0002
325 #define EFX_RX_PKT_DISCARD	0x0004
326 #define EFX_RX_PKT_TCP		0x0040
327 #define EFX_RX_PKT_PREFIX_LEN	0x0080	/* length is in prefix only */
328 #define EFX_RX_PKT_CSUM_LEVEL	0x0200
329 
330 /**
331  * struct efx_rx_page_state - Page-based rx buffer state
332  *
333  * Inserted at the start of every page allocated for receive buffers.
334  * Used to facilitate sharing dma mappings between recycled rx buffers
335  * and those passed up to the kernel.
336  *
337  * @dma_addr: The dma address of this page.
338  */
339 struct efx_rx_page_state {
340 	dma_addr_t dma_addr;
341 
342 	unsigned int __pad[] ____cacheline_aligned;
343 };
344 
345 /**
346  * struct efx_rx_queue - An Efx RX queue
347  * @efx: The associated Efx NIC
348  * @core_index:  Index of network core RX queue.  Will be >= 0 iff this
349  *	is associated with a real RX queue.
350  * @buffer: The software buffer ring
351  * @rxd: The hardware descriptor ring
352  * @ptr_mask: The size of the ring minus 1.
353  * @refill_enabled: Enable refill whenever fill level is low
354  * @flush_pending: Set when a RX flush is pending. Has the same lifetime as
355  *	@rxq_flush_pending.
356  * @added_count: Number of buffers added to the receive queue.
357  * @notified_count: Number of buffers given to NIC (<= @added_count).
358  * @removed_count: Number of buffers removed from the receive queue.
359  * @scatter_n: Used by NIC specific receive code.
360  * @scatter_len: Used by NIC specific receive code.
361  * @page_ring: The ring to store DMA mapped pages for reuse.
362  * @page_add: Counter to calculate the write pointer for the recycle ring.
363  * @page_remove: Counter to calculate the read pointer for the recycle ring.
364  * @page_recycle_count: The number of pages that have been recycled.
365  * @page_recycle_failed: The number of pages that couldn't be recycled because
366  *      the kernel still held a reference to them.
367  * @page_recycle_full: The number of pages that were released because the
368  *      recycle ring was full.
369  * @page_ptr_mask: The number of pages in the RX recycle ring minus 1.
370  * @max_fill: RX descriptor maximum fill level (<= ring size)
371  * @fast_fill_trigger: RX descriptor fill level that will trigger a fast fill
372  *	(<= @max_fill)
373  * @min_fill: RX descriptor minimum non-zero fill level.
374  *	This records the minimum fill level observed when a ring
375  *	refill was triggered.
376  * @recycle_count: RX buffer recycle counter.
377  * @slow_fill: Timer used to defer efx_nic_generate_fill_event().
378  * @xdp_rxq_info: XDP specific RX queue information.
379  * @xdp_rxq_info_valid: Is xdp_rxq_info valid data?.
380  */
381 struct efx_rx_queue {
382 	struct efx_nic *efx;
383 	int core_index;
384 	struct efx_rx_buffer *buffer;
385 	struct efx_special_buffer rxd;
386 	unsigned int ptr_mask;
387 	bool refill_enabled;
388 	bool flush_pending;
389 
390 	unsigned int added_count;
391 	unsigned int notified_count;
392 	unsigned int removed_count;
393 	unsigned int scatter_n;
394 	unsigned int scatter_len;
395 	struct page **page_ring;
396 	unsigned int page_add;
397 	unsigned int page_remove;
398 	unsigned int page_recycle_count;
399 	unsigned int page_recycle_failed;
400 	unsigned int page_recycle_full;
401 	unsigned int page_ptr_mask;
402 	unsigned int max_fill;
403 	unsigned int fast_fill_trigger;
404 	unsigned int min_fill;
405 	unsigned int min_overfill;
406 	unsigned int recycle_count;
407 	struct timer_list slow_fill;
408 	unsigned int slow_fill_count;
409 	/* Statistics to supplement MAC stats */
410 	unsigned long rx_packets;
411 	struct xdp_rxq_info xdp_rxq_info;
412 	bool xdp_rxq_info_valid;
413 };
414 
415 enum efx_sync_events_state {
416 	SYNC_EVENTS_DISABLED = 0,
417 	SYNC_EVENTS_QUIESCENT,
418 	SYNC_EVENTS_REQUESTED,
419 	SYNC_EVENTS_VALID,
420 };
421 
422 /**
423  * struct efx_channel - An Efx channel
424  *
425  * A channel comprises an event queue, at least one TX queue, at least
426  * one RX queue, and an associated tasklet for processing the event
427  * queue.
428  *
429  * @efx: Associated Efx NIC
430  * @channel: Channel instance number
431  * @type: Channel type definition
432  * @eventq_init: Event queue initialised flag
433  * @enabled: Channel enabled indicator
434  * @irq: IRQ number (MSI and MSI-X only)
435  * @irq_moderation_us: IRQ moderation value (in microseconds)
436  * @napi_dev: Net device used with NAPI
437  * @napi_str: NAPI control structure
438  * @state: state for NAPI vs busy polling
439  * @state_lock: lock protecting @state
440  * @eventq: Event queue buffer
441  * @eventq_mask: Event queue pointer mask
442  * @eventq_read_ptr: Event queue read pointer
443  * @event_test_cpu: Last CPU to handle interrupt or test event for this channel
444  * @irq_count: Number of IRQs since last adaptive moderation decision
445  * @irq_mod_score: IRQ moderation score
446  * @rfs_filter_count: number of accelerated RFS filters currently in place;
447  *	equals the count of @rps_flow_id slots filled
448  * @rfs_last_expiry: value of jiffies last time some accelerated RFS filters
449  *	were checked for expiry
450  * @rfs_expire_index: next accelerated RFS filter ID to check for expiry
451  * @n_rfs_succeeded: number of successful accelerated RFS filter insertions
452  * @n_rfs_failed; number of failed accelerated RFS filter insertions
453  * @filter_work: Work item for efx_filter_rfs_expire()
454  * @rps_flow_id: Flow IDs of filters allocated for accelerated RFS,
455  *      indexed by filter ID
456  * @n_rx_tobe_disc: Count of RX_TOBE_DISC errors
457  * @n_rx_ip_hdr_chksum_err: Count of RX IP header checksum errors
458  * @n_rx_tcp_udp_chksum_err: Count of RX TCP and UDP checksum errors
459  * @n_rx_mcast_mismatch: Count of unmatched multicast frames
460  * @n_rx_frm_trunc: Count of RX_FRM_TRUNC errors
461  * @n_rx_overlength: Count of RX_OVERLENGTH errors
462  * @n_skbuff_leaks: Count of skbuffs leaked due to RX overrun
463  * @n_rx_nodesc_trunc: Number of RX packets truncated and then dropped due to
464  *	lack of descriptors
465  * @n_rx_merge_events: Number of RX merged completion events
466  * @n_rx_merge_packets: Number of RX packets completed by merged events
467  * @n_rx_xdp_drops: Count of RX packets intentionally dropped due to XDP
468  * @n_rx_xdp_bad_drops: Count of RX packets dropped due to XDP errors
469  * @n_rx_xdp_tx: Count of RX packets retransmitted due to XDP
470  * @n_rx_xdp_redirect: Count of RX packets redirected to a different NIC by XDP
471  * @rx_pkt_n_frags: Number of fragments in next packet to be delivered by
472  *	__efx_rx_packet(), or zero if there is none
473  * @rx_pkt_index: Ring index of first buffer for next packet to be delivered
474  *	by __efx_rx_packet(), if @rx_pkt_n_frags != 0
475  * @rx_list: list of SKBs from current RX, awaiting processing
476  * @rx_queue: RX queue for this channel
477  * @tx_queue: TX queues for this channel
478  * @sync_events_state: Current state of sync events on this channel
479  * @sync_timestamp_major: Major part of the last ptp sync event
480  * @sync_timestamp_minor: Minor part of the last ptp sync event
481  */
482 struct efx_channel {
483 	struct efx_nic *efx;
484 	int channel;
485 	const struct efx_channel_type *type;
486 	bool eventq_init;
487 	bool enabled;
488 	int irq;
489 	unsigned int irq_moderation_us;
490 	struct net_device *napi_dev;
491 	struct napi_struct napi_str;
492 #ifdef CONFIG_NET_RX_BUSY_POLL
493 	unsigned long busy_poll_state;
494 #endif
495 	struct efx_special_buffer eventq;
496 	unsigned int eventq_mask;
497 	unsigned int eventq_read_ptr;
498 	int event_test_cpu;
499 
500 	unsigned int irq_count;
501 	unsigned int irq_mod_score;
502 #ifdef CONFIG_RFS_ACCEL
503 	unsigned int rfs_filter_count;
504 	unsigned int rfs_last_expiry;
505 	unsigned int rfs_expire_index;
506 	unsigned int n_rfs_succeeded;
507 	unsigned int n_rfs_failed;
508 	struct delayed_work filter_work;
509 #define RPS_FLOW_ID_INVALID 0xFFFFFFFF
510 	u32 *rps_flow_id;
511 #endif
512 
513 	unsigned int n_rx_tobe_disc;
514 	unsigned int n_rx_ip_hdr_chksum_err;
515 	unsigned int n_rx_tcp_udp_chksum_err;
516 	unsigned int n_rx_outer_ip_hdr_chksum_err;
517 	unsigned int n_rx_outer_tcp_udp_chksum_err;
518 	unsigned int n_rx_inner_ip_hdr_chksum_err;
519 	unsigned int n_rx_inner_tcp_udp_chksum_err;
520 	unsigned int n_rx_eth_crc_err;
521 	unsigned int n_rx_mcast_mismatch;
522 	unsigned int n_rx_frm_trunc;
523 	unsigned int n_rx_overlength;
524 	unsigned int n_skbuff_leaks;
525 	unsigned int n_rx_nodesc_trunc;
526 	unsigned int n_rx_merge_events;
527 	unsigned int n_rx_merge_packets;
528 	unsigned int n_rx_xdp_drops;
529 	unsigned int n_rx_xdp_bad_drops;
530 	unsigned int n_rx_xdp_tx;
531 	unsigned int n_rx_xdp_redirect;
532 
533 	unsigned int rx_pkt_n_frags;
534 	unsigned int rx_pkt_index;
535 
536 	struct list_head *rx_list;
537 
538 	struct efx_rx_queue rx_queue;
539 	struct efx_tx_queue tx_queue[EFX_TXQ_TYPES];
540 
541 	enum efx_sync_events_state sync_events_state;
542 	u32 sync_timestamp_major;
543 	u32 sync_timestamp_minor;
544 };
545 
546 /**
547  * struct efx_msi_context - Context for each MSI
548  * @efx: The associated NIC
549  * @index: Index of the channel/IRQ
550  * @name: Name of the channel/IRQ
551  *
552  * Unlike &struct efx_channel, this is never reallocated and is always
553  * safe for the IRQ handler to access.
554  */
555 struct efx_msi_context {
556 	struct efx_nic *efx;
557 	unsigned int index;
558 	char name[IFNAMSIZ + 6];
559 };
560 
561 /**
562  * struct efx_channel_type - distinguishes traffic and extra channels
563  * @handle_no_channel: Handle failure to allocate an extra channel
564  * @pre_probe: Set up extra state prior to initialisation
565  * @post_remove: Tear down extra state after finalisation, if allocated.
566  *	May be called on channels that have not been probed.
567  * @get_name: Generate the channel's name (used for its IRQ handler)
568  * @copy: Copy the channel state prior to reallocation.  May be %NULL if
569  *	reallocation is not supported.
570  * @receive_skb: Handle an skb ready to be passed to netif_receive_skb()
571  * @want_txqs: Determine whether this channel should have TX queues
572  *	created.  If %NULL, TX queues are not created.
573  * @keep_eventq: Flag for whether event queue should be kept initialised
574  *	while the device is stopped
575  * @want_pio: Flag for whether PIO buffers should be linked to this
576  *	channel's TX queues.
577  */
578 struct efx_channel_type {
579 	void (*handle_no_channel)(struct efx_nic *);
580 	int (*pre_probe)(struct efx_channel *);
581 	void (*post_remove)(struct efx_channel *);
582 	void (*get_name)(struct efx_channel *, char *buf, size_t len);
583 	struct efx_channel *(*copy)(const struct efx_channel *);
584 	bool (*receive_skb)(struct efx_channel *, struct sk_buff *);
585 	bool (*want_txqs)(struct efx_channel *);
586 	bool keep_eventq;
587 	bool want_pio;
588 };
589 
590 enum efx_led_mode {
591 	EFX_LED_OFF	= 0,
592 	EFX_LED_ON	= 1,
593 	EFX_LED_DEFAULT	= 2
594 };
595 
596 #define STRING_TABLE_LOOKUP(val, member) \
597 	((val) < member ## _max) ? member ## _names[val] : "(invalid)"
598 
599 extern const char *const efx_loopback_mode_names[];
600 extern const unsigned int efx_loopback_mode_max;
601 #define LOOPBACK_MODE(efx) \
602 	STRING_TABLE_LOOKUP((efx)->loopback_mode, efx_loopback_mode)
603 
604 extern const char *const efx_reset_type_names[];
605 extern const unsigned int efx_reset_type_max;
606 #define RESET_TYPE(type) \
607 	STRING_TABLE_LOOKUP(type, efx_reset_type)
608 
609 void efx_get_udp_tunnel_type_name(u16 type, char *buf, size_t buflen);
610 
611 enum efx_int_mode {
612 	/* Be careful if altering to correct macro below */
613 	EFX_INT_MODE_MSIX = 0,
614 	EFX_INT_MODE_MSI = 1,
615 	EFX_INT_MODE_LEGACY = 2,
616 	EFX_INT_MODE_MAX	/* Insert any new items before this */
617 };
618 #define EFX_INT_MODE_USE_MSI(x) (((x)->interrupt_mode) <= EFX_INT_MODE_MSI)
619 
620 enum nic_state {
621 	STATE_UNINIT = 0,	/* device being probed/removed or is frozen */
622 	STATE_READY = 1,	/* hardware ready and netdev registered */
623 	STATE_DISABLED = 2,	/* device disabled due to hardware errors */
624 	STATE_RECOVERY = 3,	/* device recovering from PCI error */
625 };
626 
627 /* Forward declaration */
628 struct efx_nic;
629 
630 /* Pseudo bit-mask flow control field */
631 #define EFX_FC_RX	FLOW_CTRL_RX
632 #define EFX_FC_TX	FLOW_CTRL_TX
633 #define EFX_FC_AUTO	4
634 
635 /**
636  * struct efx_link_state - Current state of the link
637  * @up: Link is up
638  * @fd: Link is full-duplex
639  * @fc: Actual flow control flags
640  * @speed: Link speed (Mbps)
641  */
642 struct efx_link_state {
643 	bool up;
644 	bool fd;
645 	u8 fc;
646 	unsigned int speed;
647 };
648 
649 static inline bool efx_link_state_equal(const struct efx_link_state *left,
650 					const struct efx_link_state *right)
651 {
652 	return left->up == right->up && left->fd == right->fd &&
653 		left->fc == right->fc && left->speed == right->speed;
654 }
655 
656 /**
657  * struct efx_phy_operations - Efx PHY operations table
658  * @probe: Probe PHY and initialise efx->mdio.mode_support, efx->mdio.mmds,
659  *	efx->loopback_modes.
660  * @init: Initialise PHY
661  * @fini: Shut down PHY
662  * @reconfigure: Reconfigure PHY (e.g. for new link parameters)
663  * @poll: Update @link_state and report whether it changed.
664  *	Serialised by the mac_lock.
665  * @get_link_ksettings: Get ethtool settings. Serialised by the mac_lock.
666  * @set_link_ksettings: Set ethtool settings. Serialised by the mac_lock.
667  * @get_fecparam: Get Forward Error Correction settings. Serialised by mac_lock.
668  * @set_fecparam: Set Forward Error Correction settings. Serialised by mac_lock.
669  * @set_npage_adv: Set abilities advertised in (Extended) Next Page
670  *	(only needed where AN bit is set in mmds)
671  * @test_alive: Test that PHY is 'alive' (online)
672  * @test_name: Get the name of a PHY-specific test/result
673  * @run_tests: Run tests and record results as appropriate (offline).
674  *	Flags are the ethtool tests flags.
675  */
676 struct efx_phy_operations {
677 	int (*probe) (struct efx_nic *efx);
678 	int (*init) (struct efx_nic *efx);
679 	void (*fini) (struct efx_nic *efx);
680 	void (*remove) (struct efx_nic *efx);
681 	int (*reconfigure) (struct efx_nic *efx);
682 	bool (*poll) (struct efx_nic *efx);
683 	void (*get_link_ksettings)(struct efx_nic *efx,
684 				   struct ethtool_link_ksettings *cmd);
685 	int (*set_link_ksettings)(struct efx_nic *efx,
686 				  const struct ethtool_link_ksettings *cmd);
687 	int (*get_fecparam)(struct efx_nic *efx, struct ethtool_fecparam *fec);
688 	int (*set_fecparam)(struct efx_nic *efx,
689 			    const struct ethtool_fecparam *fec);
690 	void (*set_npage_adv) (struct efx_nic *efx, u32);
691 	int (*test_alive) (struct efx_nic *efx);
692 	const char *(*test_name) (struct efx_nic *efx, unsigned int index);
693 	int (*run_tests) (struct efx_nic *efx, int *results, unsigned flags);
694 	int (*get_module_eeprom) (struct efx_nic *efx,
695 			       struct ethtool_eeprom *ee,
696 			       u8 *data);
697 	int (*get_module_info) (struct efx_nic *efx,
698 				struct ethtool_modinfo *modinfo);
699 };
700 
701 /**
702  * enum efx_phy_mode - PHY operating mode flags
703  * @PHY_MODE_NORMAL: on and should pass traffic
704  * @PHY_MODE_TX_DISABLED: on with TX disabled
705  * @PHY_MODE_LOW_POWER: set to low power through MDIO
706  * @PHY_MODE_OFF: switched off through external control
707  * @PHY_MODE_SPECIAL: on but will not pass traffic
708  */
709 enum efx_phy_mode {
710 	PHY_MODE_NORMAL		= 0,
711 	PHY_MODE_TX_DISABLED	= 1,
712 	PHY_MODE_LOW_POWER	= 2,
713 	PHY_MODE_OFF		= 4,
714 	PHY_MODE_SPECIAL	= 8,
715 };
716 
717 static inline bool efx_phy_mode_disabled(enum efx_phy_mode mode)
718 {
719 	return !!(mode & ~PHY_MODE_TX_DISABLED);
720 }
721 
722 /**
723  * struct efx_hw_stat_desc - Description of a hardware statistic
724  * @name: Name of the statistic as visible through ethtool, or %NULL if
725  *	it should not be exposed
726  * @dma_width: Width in bits (0 for non-DMA statistics)
727  * @offset: Offset within stats (ignored for non-DMA statistics)
728  */
729 struct efx_hw_stat_desc {
730 	const char *name;
731 	u16 dma_width;
732 	u16 offset;
733 };
734 
735 /* Number of bits used in a multicast filter hash address */
736 #define EFX_MCAST_HASH_BITS 8
737 
738 /* Number of (single-bit) entries in a multicast filter hash */
739 #define EFX_MCAST_HASH_ENTRIES (1 << EFX_MCAST_HASH_BITS)
740 
741 /* An Efx multicast filter hash */
742 union efx_multicast_hash {
743 	u8 byte[EFX_MCAST_HASH_ENTRIES / 8];
744 	efx_oword_t oword[EFX_MCAST_HASH_ENTRIES / sizeof(efx_oword_t) / 8];
745 };
746 
747 struct vfdi_status;
748 
749 /* The reserved RSS context value */
750 #define EFX_MCDI_RSS_CONTEXT_INVALID	0xffffffff
751 /**
752  * struct efx_rss_context - A user-defined RSS context for filtering
753  * @list: node of linked list on which this struct is stored
754  * @context_id: the RSS_CONTEXT_ID returned by MC firmware, or
755  *	%EFX_MCDI_RSS_CONTEXT_INVALID if this context is not present on the NIC.
756  *	For Siena, 0 if RSS is active, else %EFX_MCDI_RSS_CONTEXT_INVALID.
757  * @user_id: the rss_context ID exposed to userspace over ethtool.
758  * @rx_hash_udp_4tuple: UDP 4-tuple hashing enabled
759  * @rx_hash_key: Toeplitz hash key for this RSS context
760  * @indir_table: Indirection table for this RSS context
761  */
762 struct efx_rss_context {
763 	struct list_head list;
764 	u32 context_id;
765 	u32 user_id;
766 	bool rx_hash_udp_4tuple;
767 	u8 rx_hash_key[40];
768 	u32 rx_indir_table[128];
769 };
770 
771 #ifdef CONFIG_RFS_ACCEL
772 /* Order of these is important, since filter_id >= %EFX_ARFS_FILTER_ID_PENDING
773  * is used to test if filter does or will exist.
774  */
775 #define EFX_ARFS_FILTER_ID_PENDING	-1
776 #define EFX_ARFS_FILTER_ID_ERROR	-2
777 #define EFX_ARFS_FILTER_ID_REMOVING	-3
778 /**
779  * struct efx_arfs_rule - record of an ARFS filter and its IDs
780  * @node: linkage into hash table
781  * @spec: details of the filter (used as key for hash table).  Use efx->type to
782  *	determine which member to use.
783  * @rxq_index: channel to which the filter will steer traffic.
784  * @arfs_id: filter ID which was returned to ARFS
785  * @filter_id: index in software filter table.  May be
786  *	%EFX_ARFS_FILTER_ID_PENDING if filter was not inserted yet,
787  *	%EFX_ARFS_FILTER_ID_ERROR if filter insertion failed, or
788  *	%EFX_ARFS_FILTER_ID_REMOVING if expiry is currently removing the filter.
789  */
790 struct efx_arfs_rule {
791 	struct hlist_node node;
792 	struct efx_filter_spec spec;
793 	u16 rxq_index;
794 	u16 arfs_id;
795 	s32 filter_id;
796 };
797 
798 /* Size chosen so that the table is one page (4kB) */
799 #define EFX_ARFS_HASH_TABLE_SIZE	512
800 
801 /**
802  * struct efx_async_filter_insertion - Request to asynchronously insert a filter
803  * @net_dev: Reference to the netdevice
804  * @spec: The filter to insert
805  * @work: Workitem for this request
806  * @rxq_index: Identifies the channel for which this request was made
807  * @flow_id: Identifies the kernel-side flow for which this request was made
808  */
809 struct efx_async_filter_insertion {
810 	struct net_device *net_dev;
811 	struct efx_filter_spec spec;
812 	struct work_struct work;
813 	u16 rxq_index;
814 	u32 flow_id;
815 };
816 
817 /* Maximum number of ARFS workitems that may be in flight on an efx_nic */
818 #define EFX_RPS_MAX_IN_FLIGHT	8
819 #endif /* CONFIG_RFS_ACCEL */
820 
821 /**
822  * struct efx_nic - an Efx NIC
823  * @name: Device name (net device name or bus id before net device registered)
824  * @pci_dev: The PCI device
825  * @node: List node for maintaning primary/secondary function lists
826  * @primary: &struct efx_nic instance for the primary function of this
827  *	controller.  May be the same structure, and may be %NULL if no
828  *	primary function is bound.  Serialised by rtnl_lock.
829  * @secondary_list: List of &struct efx_nic instances for the secondary PCI
830  *	functions of the controller, if this is for the primary function.
831  *	Serialised by rtnl_lock.
832  * @type: Controller type attributes
833  * @legacy_irq: IRQ number
834  * @workqueue: Workqueue for port reconfigures and the HW monitor.
835  *	Work items do not hold and must not acquire RTNL.
836  * @workqueue_name: Name of workqueue
837  * @reset_work: Scheduled reset workitem
838  * @membase_phys: Memory BAR value as physical address
839  * @membase: Memory BAR value
840  * @vi_stride: step between per-VI registers / memory regions
841  * @interrupt_mode: Interrupt mode
842  * @timer_quantum_ns: Interrupt timer quantum, in nanoseconds
843  * @timer_max_ns: Interrupt timer maximum value, in nanoseconds
844  * @irq_rx_adaptive: Adaptive IRQ moderation enabled for RX event queues
845  * @irq_rx_mod_step_us: Step size for IRQ moderation for RX event queues
846  * @irq_rx_moderation_us: IRQ moderation time for RX event queues
847  * @msg_enable: Log message enable flags
848  * @state: Device state number (%STATE_*). Serialised by the rtnl_lock.
849  * @reset_pending: Bitmask for pending resets
850  * @tx_queue: TX DMA queues
851  * @rx_queue: RX DMA queues
852  * @channel: Channels
853  * @msi_context: Context for each MSI
854  * @extra_channel_types: Types of extra (non-traffic) channels that
855  *	should be allocated for this NIC
856  * @xdp_tx_queue_count: Number of entries in %xdp_tx_queues.
857  * @xdp_tx_queues: Array of pointers to tx queues used for XDP transmit.
858  * @rxq_entries: Size of receive queues requested by user.
859  * @txq_entries: Size of transmit queues requested by user.
860  * @txq_stop_thresh: TX queue fill level at or above which we stop it.
861  * @txq_wake_thresh: TX queue fill level at or below which we wake it.
862  * @tx_dc_base: Base qword address in SRAM of TX queue descriptor caches
863  * @rx_dc_base: Base qword address in SRAM of RX queue descriptor caches
864  * @sram_lim_qw: Qword address limit of SRAM
865  * @next_buffer_table: First available buffer table id
866  * @n_channels: Number of channels in use
867  * @n_rx_channels: Number of channels used for RX (= number of RX queues)
868  * @n_tx_channels: Number of channels used for TX
869  * @n_extra_tx_channels: Number of extra channels with TX queues
870  * @n_xdp_channels: Number of channels used for XDP TX
871  * @xdp_channel_offset: Offset of zeroth channel used for XPD TX.
872  * @xdp_tx_per_channel: Max number of TX queues on an XDP TX channel.
873  * @rx_ip_align: RX DMA address offset to have IP header aligned in
874  *	in accordance with NET_IP_ALIGN
875  * @rx_dma_len: Current maximum RX DMA length
876  * @rx_buffer_order: Order (log2) of number of pages for each RX buffer
877  * @rx_buffer_truesize: Amortised allocation size of an RX buffer,
878  *	for use in sk_buff::truesize
879  * @rx_prefix_size: Size of RX prefix before packet data
880  * @rx_packet_hash_offset: Offset of RX flow hash from start of packet data
881  *	(valid only if @rx_prefix_size != 0; always negative)
882  * @rx_packet_len_offset: Offset of RX packet length from start of packet data
883  *	(valid only for NICs that set %EFX_RX_PKT_PREFIX_LEN; always negative)
884  * @rx_packet_ts_offset: Offset of timestamp from start of packet data
885  *	(valid only if channel->sync_timestamps_enabled; always negative)
886  * @rx_scatter: Scatter mode enabled for receives
887  * @rss_context: Main RSS context.  Its @list member is the head of the list of
888  *	RSS contexts created by user requests
889  * @rss_lock: Protects custom RSS context software state in @rss_context.list
890  * @int_error_count: Number of internal errors seen recently
891  * @int_error_expire: Time at which error count will be expired
892  * @irq_soft_enabled: Are IRQs soft-enabled? If not, IRQ handler will
893  *	acknowledge but do nothing else.
894  * @irq_status: Interrupt status buffer
895  * @irq_zero_count: Number of legacy IRQs seen with queue flags == 0
896  * @irq_level: IRQ level/index for IRQs not triggered by an event queue
897  * @selftest_work: Work item for asynchronous self-test
898  * @mtd_list: List of MTDs attached to the NIC
899  * @nic_data: Hardware dependent state
900  * @mcdi: Management-Controller-to-Driver Interface state
901  * @mac_lock: MAC access lock. Protects @port_enabled, @phy_mode,
902  *	efx_monitor() and efx_reconfigure_port()
903  * @port_enabled: Port enabled indicator.
904  *	Serialises efx_stop_all(), efx_start_all(), efx_monitor() and
905  *	efx_mac_work() with kernel interfaces. Safe to read under any
906  *	one of the rtnl_lock, mac_lock, or netif_tx_lock, but all three must
907  *	be held to modify it.
908  * @port_initialized: Port initialized?
909  * @net_dev: Operating system network device. Consider holding the rtnl lock
910  * @fixed_features: Features which cannot be turned off
911  * @num_mac_stats: Number of MAC stats reported by firmware (MAC_STATS_NUM_STATS
912  *	field of %MC_CMD_GET_CAPABILITIES_V4 response, or %MC_CMD_MAC_NSTATS)
913  * @stats_buffer: DMA buffer for statistics
914  * @phy_type: PHY type
915  * @phy_op: PHY interface
916  * @phy_data: PHY private data (including PHY-specific stats)
917  * @mdio: PHY MDIO interface
918  * @mdio_bus: PHY MDIO bus ID (only used by Siena)
919  * @phy_mode: PHY operating mode. Serialised by @mac_lock.
920  * @link_advertising: Autonegotiation advertising flags
921  * @fec_config: Forward Error Correction configuration flags.  For bit positions
922  *	see &enum ethtool_fec_config_bits.
923  * @link_state: Current state of the link
924  * @n_link_state_changes: Number of times the link has changed state
925  * @unicast_filter: Flag for Falcon-arch simple unicast filter.
926  *	Protected by @mac_lock.
927  * @multicast_hash: Multicast hash table for Falcon-arch.
928  *	Protected by @mac_lock.
929  * @wanted_fc: Wanted flow control flags
930  * @fc_disable: When non-zero flow control is disabled. Typically used to
931  *	ensure that network back pressure doesn't delay dma queue flushes.
932  *	Serialised by the rtnl lock.
933  * @mac_work: Work item for changing MAC promiscuity and multicast hash
934  * @loopback_mode: Loopback status
935  * @loopback_modes: Supported loopback mode bitmask
936  * @loopback_selftest: Offline self-test private state
937  * @xdp_prog: Current XDP programme for this interface
938  * @filter_sem: Filter table rw_semaphore, protects existence of @filter_state
939  * @filter_state: Architecture-dependent filter table state
940  * @rps_mutex: Protects RPS state of all channels
941  * @rps_slot_map: bitmap of in-flight entries in @rps_slot
942  * @rps_slot: array of ARFS insertion requests for efx_filter_rfs_work()
943  * @rps_hash_lock: Protects ARFS filter mapping state (@rps_hash_table and
944  *	@rps_next_id).
945  * @rps_hash_table: Mapping between ARFS filters and their various IDs
946  * @rps_next_id: next arfs_id for an ARFS filter
947  * @active_queues: Count of RX and TX queues that haven't been flushed and drained.
948  * @rxq_flush_pending: Count of number of receive queues that need to be flushed.
949  *	Decremented when the efx_flush_rx_queue() is called.
950  * @rxq_flush_outstanding: Count of number of RX flushes started but not yet
951  *	completed (either success or failure). Not used when MCDI is used to
952  *	flush receive queues.
953  * @flush_wq: wait queue used by efx_nic_flush_queues() to wait for flush completions.
954  * @vf_count: Number of VFs intended to be enabled.
955  * @vf_init_count: Number of VFs that have been fully initialised.
956  * @vi_scale: log2 number of vnics per VF.
957  * @ptp_data: PTP state data
958  * @ptp_warned: has this NIC seen and warned about unexpected PTP events?
959  * @vpd_sn: Serial number read from VPD
960  * @xdp_rxq_info_failed: Have any of the rx queues failed to initialise their
961  *      xdp_rxq_info structures?
962  * @monitor_work: Hardware monitor workitem
963  * @biu_lock: BIU (bus interface unit) lock
964  * @last_irq_cpu: Last CPU to handle a possible test interrupt.  This
965  *	field is used by efx_test_interrupts() to verify that an
966  *	interrupt has occurred.
967  * @stats_lock: Statistics update lock. Must be held when calling
968  *	efx_nic_type::{update,start,stop}_stats.
969  * @n_rx_noskb_drops: Count of RX packets dropped due to failure to allocate an skb
970  *
971  * This is stored in the private area of the &struct net_device.
972  */
973 struct efx_nic {
974 	/* The following fields should be written very rarely */
975 
976 	char name[IFNAMSIZ];
977 	struct list_head node;
978 	struct efx_nic *primary;
979 	struct list_head secondary_list;
980 	struct pci_dev *pci_dev;
981 	unsigned int port_num;
982 	const struct efx_nic_type *type;
983 	int legacy_irq;
984 	bool eeh_disabled_legacy_irq;
985 	struct workqueue_struct *workqueue;
986 	char workqueue_name[16];
987 	struct work_struct reset_work;
988 	resource_size_t membase_phys;
989 	void __iomem *membase;
990 
991 	unsigned int vi_stride;
992 
993 	enum efx_int_mode interrupt_mode;
994 	unsigned int timer_quantum_ns;
995 	unsigned int timer_max_ns;
996 	bool irq_rx_adaptive;
997 	unsigned int irq_mod_step_us;
998 	unsigned int irq_rx_moderation_us;
999 	u32 msg_enable;
1000 
1001 	enum nic_state state;
1002 	unsigned long reset_pending;
1003 
1004 	struct efx_channel *channel[EFX_MAX_CHANNELS];
1005 	struct efx_msi_context msi_context[EFX_MAX_CHANNELS];
1006 	const struct efx_channel_type *
1007 	extra_channel_type[EFX_MAX_EXTRA_CHANNELS];
1008 
1009 	unsigned int xdp_tx_queue_count;
1010 	struct efx_tx_queue **xdp_tx_queues;
1011 
1012 	unsigned rxq_entries;
1013 	unsigned txq_entries;
1014 	unsigned int txq_stop_thresh;
1015 	unsigned int txq_wake_thresh;
1016 
1017 	unsigned tx_dc_base;
1018 	unsigned rx_dc_base;
1019 	unsigned sram_lim_qw;
1020 	unsigned next_buffer_table;
1021 
1022 	unsigned int max_channels;
1023 	unsigned int max_tx_channels;
1024 	unsigned n_channels;
1025 	unsigned n_rx_channels;
1026 	unsigned rss_spread;
1027 	unsigned tx_channel_offset;
1028 	unsigned n_tx_channels;
1029 	unsigned n_extra_tx_channels;
1030 	unsigned int n_xdp_channels;
1031 	unsigned int xdp_channel_offset;
1032 	unsigned int xdp_tx_per_channel;
1033 	unsigned int rx_ip_align;
1034 	unsigned int rx_dma_len;
1035 	unsigned int rx_buffer_order;
1036 	unsigned int rx_buffer_truesize;
1037 	unsigned int rx_page_buf_step;
1038 	unsigned int rx_bufs_per_page;
1039 	unsigned int rx_pages_per_batch;
1040 	unsigned int rx_prefix_size;
1041 	int rx_packet_hash_offset;
1042 	int rx_packet_len_offset;
1043 	int rx_packet_ts_offset;
1044 	bool rx_scatter;
1045 	struct efx_rss_context rss_context;
1046 	struct mutex rss_lock;
1047 
1048 	unsigned int_error_count;
1049 	unsigned long int_error_expire;
1050 
1051 	bool irq_soft_enabled;
1052 	struct efx_buffer irq_status;
1053 	unsigned irq_zero_count;
1054 	unsigned irq_level;
1055 	struct delayed_work selftest_work;
1056 
1057 #ifdef CONFIG_SFC_MTD
1058 	struct list_head mtd_list;
1059 #endif
1060 
1061 	void *nic_data;
1062 	struct efx_mcdi_data *mcdi;
1063 
1064 	struct mutex mac_lock;
1065 	struct work_struct mac_work;
1066 	bool port_enabled;
1067 
1068 	bool mc_bist_for_other_fn;
1069 	bool port_initialized;
1070 	struct net_device *net_dev;
1071 
1072 	netdev_features_t fixed_features;
1073 
1074 	u16 num_mac_stats;
1075 	struct efx_buffer stats_buffer;
1076 	u64 rx_nodesc_drops_total;
1077 	u64 rx_nodesc_drops_while_down;
1078 	bool rx_nodesc_drops_prev_state;
1079 
1080 	unsigned int phy_type;
1081 	const struct efx_phy_operations *phy_op;
1082 	void *phy_data;
1083 	struct mdio_if_info mdio;
1084 	unsigned int mdio_bus;
1085 	enum efx_phy_mode phy_mode;
1086 
1087 	__ETHTOOL_DECLARE_LINK_MODE_MASK(link_advertising);
1088 	u32 fec_config;
1089 	struct efx_link_state link_state;
1090 	unsigned int n_link_state_changes;
1091 
1092 	bool unicast_filter;
1093 	union efx_multicast_hash multicast_hash;
1094 	u8 wanted_fc;
1095 	unsigned fc_disable;
1096 
1097 	atomic_t rx_reset;
1098 	enum efx_loopback_mode loopback_mode;
1099 	u64 loopback_modes;
1100 
1101 	void *loopback_selftest;
1102 	/* We access loopback_selftest immediately before running XDP,
1103 	 * so we want them next to each other.
1104 	 */
1105 	struct bpf_prog __rcu *xdp_prog;
1106 
1107 	struct rw_semaphore filter_sem;
1108 	void *filter_state;
1109 #ifdef CONFIG_RFS_ACCEL
1110 	struct mutex rps_mutex;
1111 	unsigned long rps_slot_map;
1112 	struct efx_async_filter_insertion rps_slot[EFX_RPS_MAX_IN_FLIGHT];
1113 	spinlock_t rps_hash_lock;
1114 	struct hlist_head *rps_hash_table;
1115 	u32 rps_next_id;
1116 #endif
1117 
1118 	atomic_t active_queues;
1119 	atomic_t rxq_flush_pending;
1120 	atomic_t rxq_flush_outstanding;
1121 	wait_queue_head_t flush_wq;
1122 
1123 #ifdef CONFIG_SFC_SRIOV
1124 	unsigned vf_count;
1125 	unsigned vf_init_count;
1126 	unsigned vi_scale;
1127 #endif
1128 
1129 	struct efx_ptp_data *ptp_data;
1130 	bool ptp_warned;
1131 
1132 	char *vpd_sn;
1133 	bool xdp_rxq_info_failed;
1134 
1135 	/* The following fields may be written more often */
1136 
1137 	struct delayed_work monitor_work ____cacheline_aligned_in_smp;
1138 	spinlock_t biu_lock;
1139 	int last_irq_cpu;
1140 	spinlock_t stats_lock;
1141 	atomic_t n_rx_noskb_drops;
1142 };
1143 
1144 static inline int efx_dev_registered(struct efx_nic *efx)
1145 {
1146 	return efx->net_dev->reg_state == NETREG_REGISTERED;
1147 }
1148 
1149 static inline unsigned int efx_port_num(struct efx_nic *efx)
1150 {
1151 	return efx->port_num;
1152 }
1153 
1154 struct efx_mtd_partition {
1155 	struct list_head node;
1156 	struct mtd_info mtd;
1157 	const char *dev_type_name;
1158 	const char *type_name;
1159 	char name[IFNAMSIZ + 20];
1160 };
1161 
1162 struct efx_udp_tunnel {
1163 	u16 type; /* TUNNEL_ENCAP_UDP_PORT_ENTRY_foo, see mcdi_pcol.h */
1164 	__be16 port;
1165 	/* Count of repeated adds of the same port.  Used only inside the list,
1166 	 * not in request arguments.
1167 	 */
1168 	u16 count;
1169 };
1170 
1171 /**
1172  * struct efx_nic_type - Efx device type definition
1173  * @mem_bar: Get the memory BAR
1174  * @mem_map_size: Get memory BAR mapped size
1175  * @probe: Probe the controller
1176  * @remove: Free resources allocated by probe()
1177  * @init: Initialise the controller
1178  * @dimension_resources: Dimension controller resources (buffer table,
1179  *	and VIs once the available interrupt resources are clear)
1180  * @fini: Shut down the controller
1181  * @monitor: Periodic function for polling link state and hardware monitor
1182  * @map_reset_reason: Map ethtool reset reason to a reset method
1183  * @map_reset_flags: Map ethtool reset flags to a reset method, if possible
1184  * @reset: Reset the controller hardware and possibly the PHY.  This will
1185  *	be called while the controller is uninitialised.
1186  * @probe_port: Probe the MAC and PHY
1187  * @remove_port: Free resources allocated by probe_port()
1188  * @handle_global_event: Handle a "global" event (may be %NULL)
1189  * @fini_dmaq: Flush and finalise DMA queues (RX and TX queues)
1190  * @prepare_flush: Prepare the hardware for flushing the DMA queues
1191  *	(for Falcon architecture)
1192  * @finish_flush: Clean up after flushing the DMA queues (for Falcon
1193  *	architecture)
1194  * @prepare_flr: Prepare for an FLR
1195  * @finish_flr: Clean up after an FLR
1196  * @describe_stats: Describe statistics for ethtool
1197  * @update_stats: Update statistics not provided by event handling.
1198  *	Either argument may be %NULL.
1199  * @start_stats: Start the regular fetching of statistics
1200  * @pull_stats: Pull stats from the NIC and wait until they arrive.
1201  * @stop_stats: Stop the regular fetching of statistics
1202  * @set_id_led: Set state of identifying LED or revert to automatic function
1203  * @push_irq_moderation: Apply interrupt moderation value
1204  * @reconfigure_port: Push loopback/power/txdis changes to the MAC and PHY
1205  * @prepare_enable_fc_tx: Prepare MAC to enable pause frame TX (may be %NULL)
1206  * @reconfigure_mac: Push MAC address, MTU, flow control and filter settings
1207  *	to the hardware.  Serialised by the mac_lock.
1208  * @check_mac_fault: Check MAC fault state. True if fault present.
1209  * @get_wol: Get WoL configuration from driver state
1210  * @set_wol: Push WoL configuration to the NIC
1211  * @resume_wol: Synchronise WoL state between driver and MC (e.g. after resume)
1212  * @test_chip: Test registers.  May use efx_farch_test_registers(), and is
1213  *	expected to reset the NIC.
1214  * @test_nvram: Test validity of NVRAM contents
1215  * @mcdi_request: Send an MCDI request with the given header and SDU.
1216  *	The SDU length may be any value from 0 up to the protocol-
1217  *	defined maximum, but its buffer will be padded to a multiple
1218  *	of 4 bytes.
1219  * @mcdi_poll_response: Test whether an MCDI response is available.
1220  * @mcdi_read_response: Read the MCDI response PDU.  The offset will
1221  *	be a multiple of 4.  The length may not be, but the buffer
1222  *	will be padded so it is safe to round up.
1223  * @mcdi_poll_reboot: Test whether the MCDI has rebooted.  If so,
1224  *	return an appropriate error code for aborting any current
1225  *	request; otherwise return 0.
1226  * @irq_enable_master: Enable IRQs on the NIC.  Each event queue must
1227  *	be separately enabled after this.
1228  * @irq_test_generate: Generate a test IRQ
1229  * @irq_disable_non_ev: Disable non-event IRQs on the NIC.  Each event
1230  *	queue must be separately disabled before this.
1231  * @irq_handle_msi: Handle MSI for a channel.  The @dev_id argument is
1232  *	a pointer to the &struct efx_msi_context for the channel.
1233  * @irq_handle_legacy: Handle legacy interrupt.  The @dev_id argument
1234  *	is a pointer to the &struct efx_nic.
1235  * @tx_probe: Allocate resources for TX queue
1236  * @tx_init: Initialise TX queue on the NIC
1237  * @tx_remove: Free resources for TX queue
1238  * @tx_write: Write TX descriptors and doorbell
1239  * @rx_push_rss_config: Write RSS hash key and indirection table to the NIC
1240  * @rx_pull_rss_config: Read RSS hash key and indirection table back from the NIC
1241  * @rx_push_rss_context_config: Write RSS hash key and indirection table for
1242  *	user RSS context to the NIC
1243  * @rx_pull_rss_context_config: Read RSS hash key and indirection table for user
1244  *	RSS context back from the NIC
1245  * @rx_probe: Allocate resources for RX queue
1246  * @rx_init: Initialise RX queue on the NIC
1247  * @rx_remove: Free resources for RX queue
1248  * @rx_write: Write RX descriptors and doorbell
1249  * @rx_defer_refill: Generate a refill reminder event
1250  * @ev_probe: Allocate resources for event queue
1251  * @ev_init: Initialise event queue on the NIC
1252  * @ev_fini: Deinitialise event queue on the NIC
1253  * @ev_remove: Free resources for event queue
1254  * @ev_process: Process events for a queue, up to the given NAPI quota
1255  * @ev_read_ack: Acknowledge read events on a queue, rearming its IRQ
1256  * @ev_test_generate: Generate a test event
1257  * @filter_table_probe: Probe filter capabilities and set up filter software state
1258  * @filter_table_restore: Restore filters removed from hardware
1259  * @filter_table_remove: Remove filters from hardware and tear down software state
1260  * @filter_update_rx_scatter: Update filters after change to rx scatter setting
1261  * @filter_insert: add or replace a filter
1262  * @filter_remove_safe: remove a filter by ID, carefully
1263  * @filter_get_safe: retrieve a filter by ID, carefully
1264  * @filter_clear_rx: Remove all RX filters whose priority is less than or
1265  *	equal to the given priority and is not %EFX_FILTER_PRI_AUTO
1266  * @filter_count_rx_used: Get the number of filters in use at a given priority
1267  * @filter_get_rx_id_limit: Get maximum value of a filter id, plus 1
1268  * @filter_get_rx_ids: Get list of RX filters at a given priority
1269  * @filter_rfs_expire_one: Consider expiring a filter inserted for RFS.
1270  *	This must check whether the specified table entry is used by RFS
1271  *	and that rps_may_expire_flow() returns true for it.
1272  * @mtd_probe: Probe and add MTD partitions associated with this net device,
1273  *	 using efx_mtd_add()
1274  * @mtd_rename: Set an MTD partition name using the net device name
1275  * @mtd_read: Read from an MTD partition
1276  * @mtd_erase: Erase part of an MTD partition
1277  * @mtd_write: Write to an MTD partition
1278  * @mtd_sync: Wait for write-back to complete on MTD partition.  This
1279  *	also notifies the driver that a writer has finished using this
1280  *	partition.
1281  * @ptp_write_host_time: Send host time to MC as part of sync protocol
1282  * @ptp_set_ts_sync_events: Enable or disable sync events for inline RX
1283  *	timestamping, possibly only temporarily for the purposes of a reset.
1284  * @ptp_set_ts_config: Set hardware timestamp configuration.  The flags
1285  *	and tx_type will already have been validated but this operation
1286  *	must validate and update rx_filter.
1287  * @get_phys_port_id: Get the underlying physical port id.
1288  * @set_mac_address: Set the MAC address of the device
1289  * @tso_versions: Returns mask of firmware-assisted TSO versions supported.
1290  *	If %NULL, then device does not support any TSO version.
1291  * @udp_tnl_push_ports: Push the list of UDP tunnel ports to the NIC if required.
1292  * @udp_tnl_add_port: Add a UDP tunnel port
1293  * @udp_tnl_has_port: Check if a port has been added as UDP tunnel
1294  * @udp_tnl_del_port: Remove a UDP tunnel port
1295  * @revision: Hardware architecture revision
1296  * @txd_ptr_tbl_base: TX descriptor ring base address
1297  * @rxd_ptr_tbl_base: RX descriptor ring base address
1298  * @buf_tbl_base: Buffer table base address
1299  * @evq_ptr_tbl_base: Event queue pointer table base address
1300  * @evq_rptr_tbl_base: Event queue read-pointer table base address
1301  * @max_dma_mask: Maximum possible DMA mask
1302  * @rx_prefix_size: Size of RX prefix before packet data
1303  * @rx_hash_offset: Offset of RX flow hash within prefix
1304  * @rx_ts_offset: Offset of timestamp within prefix
1305  * @rx_buffer_padding: Size of padding at end of RX packet
1306  * @can_rx_scatter: NIC is able to scatter packets to multiple buffers
1307  * @always_rx_scatter: NIC will always scatter packets to multiple buffers
1308  * @option_descriptors: NIC supports TX option descriptors
1309  * @min_interrupt_mode: Lowest capability interrupt mode supported
1310  *	from &enum efx_int_mode.
1311  * @max_interrupt_mode: Highest capability interrupt mode supported
1312  *	from &enum efx_int_mode.
1313  * @timer_period_max: Maximum period of interrupt timer (in ticks)
1314  * @offload_features: net_device feature flags for protocol offload
1315  *	features implemented in hardware
1316  * @mcdi_max_ver: Maximum MCDI version supported
1317  * @hwtstamp_filters: Mask of hardware timestamp filter types supported
1318  */
1319 struct efx_nic_type {
1320 	bool is_vf;
1321 	unsigned int (*mem_bar)(struct efx_nic *efx);
1322 	unsigned int (*mem_map_size)(struct efx_nic *efx);
1323 	int (*probe)(struct efx_nic *efx);
1324 	void (*remove)(struct efx_nic *efx);
1325 	int (*init)(struct efx_nic *efx);
1326 	int (*dimension_resources)(struct efx_nic *efx);
1327 	void (*fini)(struct efx_nic *efx);
1328 	void (*monitor)(struct efx_nic *efx);
1329 	enum reset_type (*map_reset_reason)(enum reset_type reason);
1330 	int (*map_reset_flags)(u32 *flags);
1331 	int (*reset)(struct efx_nic *efx, enum reset_type method);
1332 	int (*probe_port)(struct efx_nic *efx);
1333 	void (*remove_port)(struct efx_nic *efx);
1334 	bool (*handle_global_event)(struct efx_channel *channel, efx_qword_t *);
1335 	int (*fini_dmaq)(struct efx_nic *efx);
1336 	void (*prepare_flush)(struct efx_nic *efx);
1337 	void (*finish_flush)(struct efx_nic *efx);
1338 	void (*prepare_flr)(struct efx_nic *efx);
1339 	void (*finish_flr)(struct efx_nic *efx);
1340 	size_t (*describe_stats)(struct efx_nic *efx, u8 *names);
1341 	size_t (*update_stats)(struct efx_nic *efx, u64 *full_stats,
1342 			       struct rtnl_link_stats64 *core_stats);
1343 	void (*start_stats)(struct efx_nic *efx);
1344 	void (*pull_stats)(struct efx_nic *efx);
1345 	void (*stop_stats)(struct efx_nic *efx);
1346 	void (*set_id_led)(struct efx_nic *efx, enum efx_led_mode mode);
1347 	void (*push_irq_moderation)(struct efx_channel *channel);
1348 	int (*reconfigure_port)(struct efx_nic *efx);
1349 	void (*prepare_enable_fc_tx)(struct efx_nic *efx);
1350 	int (*reconfigure_mac)(struct efx_nic *efx);
1351 	bool (*check_mac_fault)(struct efx_nic *efx);
1352 	void (*get_wol)(struct efx_nic *efx, struct ethtool_wolinfo *wol);
1353 	int (*set_wol)(struct efx_nic *efx, u32 type);
1354 	void (*resume_wol)(struct efx_nic *efx);
1355 	int (*test_chip)(struct efx_nic *efx, struct efx_self_tests *tests);
1356 	int (*test_nvram)(struct efx_nic *efx);
1357 	void (*mcdi_request)(struct efx_nic *efx,
1358 			     const efx_dword_t *hdr, size_t hdr_len,
1359 			     const efx_dword_t *sdu, size_t sdu_len);
1360 	bool (*mcdi_poll_response)(struct efx_nic *efx);
1361 	void (*mcdi_read_response)(struct efx_nic *efx, efx_dword_t *pdu,
1362 				   size_t pdu_offset, size_t pdu_len);
1363 	int (*mcdi_poll_reboot)(struct efx_nic *efx);
1364 	void (*mcdi_reboot_detected)(struct efx_nic *efx);
1365 	void (*irq_enable_master)(struct efx_nic *efx);
1366 	int (*irq_test_generate)(struct efx_nic *efx);
1367 	void (*irq_disable_non_ev)(struct efx_nic *efx);
1368 	irqreturn_t (*irq_handle_msi)(int irq, void *dev_id);
1369 	irqreturn_t (*irq_handle_legacy)(int irq, void *dev_id);
1370 	int (*tx_probe)(struct efx_tx_queue *tx_queue);
1371 	void (*tx_init)(struct efx_tx_queue *tx_queue);
1372 	void (*tx_remove)(struct efx_tx_queue *tx_queue);
1373 	void (*tx_write)(struct efx_tx_queue *tx_queue);
1374 	unsigned int (*tx_limit_len)(struct efx_tx_queue *tx_queue,
1375 				     dma_addr_t dma_addr, unsigned int len);
1376 	int (*rx_push_rss_config)(struct efx_nic *efx, bool user,
1377 				  const u32 *rx_indir_table, const u8 *key);
1378 	int (*rx_pull_rss_config)(struct efx_nic *efx);
1379 	int (*rx_push_rss_context_config)(struct efx_nic *efx,
1380 					  struct efx_rss_context *ctx,
1381 					  const u32 *rx_indir_table,
1382 					  const u8 *key);
1383 	int (*rx_pull_rss_context_config)(struct efx_nic *efx,
1384 					  struct efx_rss_context *ctx);
1385 	void (*rx_restore_rss_contexts)(struct efx_nic *efx);
1386 	int (*rx_probe)(struct efx_rx_queue *rx_queue);
1387 	void (*rx_init)(struct efx_rx_queue *rx_queue);
1388 	void (*rx_remove)(struct efx_rx_queue *rx_queue);
1389 	void (*rx_write)(struct efx_rx_queue *rx_queue);
1390 	void (*rx_defer_refill)(struct efx_rx_queue *rx_queue);
1391 	int (*ev_probe)(struct efx_channel *channel);
1392 	int (*ev_init)(struct efx_channel *channel);
1393 	void (*ev_fini)(struct efx_channel *channel);
1394 	void (*ev_remove)(struct efx_channel *channel);
1395 	int (*ev_process)(struct efx_channel *channel, int quota);
1396 	void (*ev_read_ack)(struct efx_channel *channel);
1397 	void (*ev_test_generate)(struct efx_channel *channel);
1398 	int (*filter_table_probe)(struct efx_nic *efx);
1399 	void (*filter_table_restore)(struct efx_nic *efx);
1400 	void (*filter_table_remove)(struct efx_nic *efx);
1401 	void (*filter_update_rx_scatter)(struct efx_nic *efx);
1402 	s32 (*filter_insert)(struct efx_nic *efx,
1403 			     struct efx_filter_spec *spec, bool replace);
1404 	int (*filter_remove_safe)(struct efx_nic *efx,
1405 				  enum efx_filter_priority priority,
1406 				  u32 filter_id);
1407 	int (*filter_get_safe)(struct efx_nic *efx,
1408 			       enum efx_filter_priority priority,
1409 			       u32 filter_id, struct efx_filter_spec *);
1410 	int (*filter_clear_rx)(struct efx_nic *efx,
1411 			       enum efx_filter_priority priority);
1412 	u32 (*filter_count_rx_used)(struct efx_nic *efx,
1413 				    enum efx_filter_priority priority);
1414 	u32 (*filter_get_rx_id_limit)(struct efx_nic *efx);
1415 	s32 (*filter_get_rx_ids)(struct efx_nic *efx,
1416 				 enum efx_filter_priority priority,
1417 				 u32 *buf, u32 size);
1418 #ifdef CONFIG_RFS_ACCEL
1419 	bool (*filter_rfs_expire_one)(struct efx_nic *efx, u32 flow_id,
1420 				      unsigned int index);
1421 #endif
1422 #ifdef CONFIG_SFC_MTD
1423 	int (*mtd_probe)(struct efx_nic *efx);
1424 	void (*mtd_rename)(struct efx_mtd_partition *part);
1425 	int (*mtd_read)(struct mtd_info *mtd, loff_t start, size_t len,
1426 			size_t *retlen, u8 *buffer);
1427 	int (*mtd_erase)(struct mtd_info *mtd, loff_t start, size_t len);
1428 	int (*mtd_write)(struct mtd_info *mtd, loff_t start, size_t len,
1429 			 size_t *retlen, const u8 *buffer);
1430 	int (*mtd_sync)(struct mtd_info *mtd);
1431 #endif
1432 	void (*ptp_write_host_time)(struct efx_nic *efx, u32 host_time);
1433 	int (*ptp_set_ts_sync_events)(struct efx_nic *efx, bool en, bool temp);
1434 	int (*ptp_set_ts_config)(struct efx_nic *efx,
1435 				 struct hwtstamp_config *init);
1436 	int (*sriov_configure)(struct efx_nic *efx, int num_vfs);
1437 	int (*vlan_rx_add_vid)(struct efx_nic *efx, __be16 proto, u16 vid);
1438 	int (*vlan_rx_kill_vid)(struct efx_nic *efx, __be16 proto, u16 vid);
1439 	int (*get_phys_port_id)(struct efx_nic *efx,
1440 				struct netdev_phys_item_id *ppid);
1441 	int (*sriov_init)(struct efx_nic *efx);
1442 	void (*sriov_fini)(struct efx_nic *efx);
1443 	bool (*sriov_wanted)(struct efx_nic *efx);
1444 	void (*sriov_reset)(struct efx_nic *efx);
1445 	void (*sriov_flr)(struct efx_nic *efx, unsigned vf_i);
1446 	int (*sriov_set_vf_mac)(struct efx_nic *efx, int vf_i, u8 *mac);
1447 	int (*sriov_set_vf_vlan)(struct efx_nic *efx, int vf_i, u16 vlan,
1448 				 u8 qos);
1449 	int (*sriov_set_vf_spoofchk)(struct efx_nic *efx, int vf_i,
1450 				     bool spoofchk);
1451 	int (*sriov_get_vf_config)(struct efx_nic *efx, int vf_i,
1452 				   struct ifla_vf_info *ivi);
1453 	int (*sriov_set_vf_link_state)(struct efx_nic *efx, int vf_i,
1454 				       int link_state);
1455 	int (*vswitching_probe)(struct efx_nic *efx);
1456 	int (*vswitching_restore)(struct efx_nic *efx);
1457 	void (*vswitching_remove)(struct efx_nic *efx);
1458 	int (*get_mac_address)(struct efx_nic *efx, unsigned char *perm_addr);
1459 	int (*set_mac_address)(struct efx_nic *efx);
1460 	u32 (*tso_versions)(struct efx_nic *efx);
1461 	int (*udp_tnl_push_ports)(struct efx_nic *efx);
1462 	int (*udp_tnl_add_port)(struct efx_nic *efx, struct efx_udp_tunnel tnl);
1463 	bool (*udp_tnl_has_port)(struct efx_nic *efx, __be16 port);
1464 	int (*udp_tnl_del_port)(struct efx_nic *efx, struct efx_udp_tunnel tnl);
1465 
1466 	int revision;
1467 	unsigned int txd_ptr_tbl_base;
1468 	unsigned int rxd_ptr_tbl_base;
1469 	unsigned int buf_tbl_base;
1470 	unsigned int evq_ptr_tbl_base;
1471 	unsigned int evq_rptr_tbl_base;
1472 	u64 max_dma_mask;
1473 	unsigned int rx_prefix_size;
1474 	unsigned int rx_hash_offset;
1475 	unsigned int rx_ts_offset;
1476 	unsigned int rx_buffer_padding;
1477 	bool can_rx_scatter;
1478 	bool always_rx_scatter;
1479 	bool option_descriptors;
1480 	unsigned int min_interrupt_mode;
1481 	unsigned int max_interrupt_mode;
1482 	unsigned int timer_period_max;
1483 	netdev_features_t offload_features;
1484 	int mcdi_max_ver;
1485 	unsigned int max_rx_ip_filters;
1486 	u32 hwtstamp_filters;
1487 	unsigned int rx_hash_key_size;
1488 };
1489 
1490 /**************************************************************************
1491  *
1492  * Prototypes and inline functions
1493  *
1494  *************************************************************************/
1495 
1496 static inline struct efx_channel *
1497 efx_get_channel(struct efx_nic *efx, unsigned index)
1498 {
1499 	EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_channels);
1500 	return efx->channel[index];
1501 }
1502 
1503 /* Iterate over all used channels */
1504 #define efx_for_each_channel(_channel, _efx)				\
1505 	for (_channel = (_efx)->channel[0];				\
1506 	     _channel;							\
1507 	     _channel = (_channel->channel + 1 < (_efx)->n_channels) ?	\
1508 		     (_efx)->channel[_channel->channel + 1] : NULL)
1509 
1510 /* Iterate over all used channels in reverse */
1511 #define efx_for_each_channel_rev(_channel, _efx)			\
1512 	for (_channel = (_efx)->channel[(_efx)->n_channels - 1];	\
1513 	     _channel;							\
1514 	     _channel = _channel->channel ?				\
1515 		     (_efx)->channel[_channel->channel - 1] : NULL)
1516 
1517 static inline struct efx_tx_queue *
1518 efx_get_tx_queue(struct efx_nic *efx, unsigned index, unsigned type)
1519 {
1520 	EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_tx_channels ||
1521 				  type >= EFX_TXQ_TYPES);
1522 	return &efx->channel[efx->tx_channel_offset + index]->tx_queue[type];
1523 }
1524 
1525 static inline struct efx_channel *
1526 efx_get_xdp_channel(struct efx_nic *efx, unsigned int index)
1527 {
1528 	EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_xdp_channels);
1529 	return efx->channel[efx->xdp_channel_offset + index];
1530 }
1531 
1532 static inline bool efx_channel_is_xdp_tx(struct efx_channel *channel)
1533 {
1534 	return channel->channel - channel->efx->xdp_channel_offset <
1535 	       channel->efx->n_xdp_channels;
1536 }
1537 
1538 static inline bool efx_channel_has_tx_queues(struct efx_channel *channel)
1539 {
1540 	return true;
1541 }
1542 
1543 static inline struct efx_tx_queue *
1544 efx_channel_get_tx_queue(struct efx_channel *channel, unsigned type)
1545 {
1546 	EFX_WARN_ON_ONCE_PARANOID(!efx_channel_has_tx_queues(channel) ||
1547 				  type >= EFX_TXQ_TYPES);
1548 	return &channel->tx_queue[type];
1549 }
1550 
1551 static inline bool efx_tx_queue_used(struct efx_tx_queue *tx_queue)
1552 {
1553 	return !(tx_queue->efx->net_dev->num_tc < 2 &&
1554 		 tx_queue->queue & EFX_TXQ_TYPE_HIGHPRI);
1555 }
1556 
1557 /* Iterate over all TX queues belonging to a channel */
1558 #define efx_for_each_channel_tx_queue(_tx_queue, _channel)		\
1559 	if (!efx_channel_has_tx_queues(_channel))			\
1560 		;							\
1561 	else								\
1562 		for (_tx_queue = (_channel)->tx_queue;			\
1563 		     _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES && \
1564 			     (efx_tx_queue_used(_tx_queue) ||            \
1565 			      efx_channel_is_xdp_tx(_channel));		\
1566 		     _tx_queue++)
1567 
1568 /* Iterate over all possible TX queues belonging to a channel */
1569 #define efx_for_each_possible_channel_tx_queue(_tx_queue, _channel)	\
1570 	if (!efx_channel_has_tx_queues(_channel))			\
1571 		;							\
1572 	else								\
1573 		for (_tx_queue = (_channel)->tx_queue;			\
1574 		     _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES;	\
1575 		     _tx_queue++)
1576 
1577 static inline bool efx_channel_has_rx_queue(struct efx_channel *channel)
1578 {
1579 	return channel->rx_queue.core_index >= 0;
1580 }
1581 
1582 static inline struct efx_rx_queue *
1583 efx_channel_get_rx_queue(struct efx_channel *channel)
1584 {
1585 	EFX_WARN_ON_ONCE_PARANOID(!efx_channel_has_rx_queue(channel));
1586 	return &channel->rx_queue;
1587 }
1588 
1589 /* Iterate over all RX queues belonging to a channel */
1590 #define efx_for_each_channel_rx_queue(_rx_queue, _channel)		\
1591 	if (!efx_channel_has_rx_queue(_channel))			\
1592 		;							\
1593 	else								\
1594 		for (_rx_queue = &(_channel)->rx_queue;			\
1595 		     _rx_queue;						\
1596 		     _rx_queue = NULL)
1597 
1598 static inline struct efx_channel *
1599 efx_rx_queue_channel(struct efx_rx_queue *rx_queue)
1600 {
1601 	return container_of(rx_queue, struct efx_channel, rx_queue);
1602 }
1603 
1604 static inline int efx_rx_queue_index(struct efx_rx_queue *rx_queue)
1605 {
1606 	return efx_rx_queue_channel(rx_queue)->channel;
1607 }
1608 
1609 /* Returns a pointer to the specified receive buffer in the RX
1610  * descriptor queue.
1611  */
1612 static inline struct efx_rx_buffer *efx_rx_buffer(struct efx_rx_queue *rx_queue,
1613 						  unsigned int index)
1614 {
1615 	return &rx_queue->buffer[index];
1616 }
1617 
1618 static inline struct efx_rx_buffer *
1619 efx_rx_buf_next(struct efx_rx_queue *rx_queue, struct efx_rx_buffer *rx_buf)
1620 {
1621 	if (unlikely(rx_buf == efx_rx_buffer(rx_queue, rx_queue->ptr_mask)))
1622 		return efx_rx_buffer(rx_queue, 0);
1623 	else
1624 		return rx_buf + 1;
1625 }
1626 
1627 /**
1628  * EFX_MAX_FRAME_LEN - calculate maximum frame length
1629  *
1630  * This calculates the maximum frame length that will be used for a
1631  * given MTU.  The frame length will be equal to the MTU plus a
1632  * constant amount of header space and padding.  This is the quantity
1633  * that the net driver will program into the MAC as the maximum frame
1634  * length.
1635  *
1636  * The 10G MAC requires 8-byte alignment on the frame
1637  * length, so we round up to the nearest 8.
1638  *
1639  * Re-clocking by the XGXS on RX can reduce an IPG to 32 bits (half an
1640  * XGMII cycle).  If the frame length reaches the maximum value in the
1641  * same cycle, the XMAC can miss the IPG altogether.  We work around
1642  * this by adding a further 16 bytes.
1643  */
1644 #define EFX_FRAME_PAD	16
1645 #define EFX_MAX_FRAME_LEN(mtu) \
1646 	(ALIGN(((mtu) + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN + EFX_FRAME_PAD), 8))
1647 
1648 static inline bool efx_xmit_with_hwtstamp(struct sk_buff *skb)
1649 {
1650 	return skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP;
1651 }
1652 static inline void efx_xmit_hwtstamp_pending(struct sk_buff *skb)
1653 {
1654 	skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
1655 }
1656 
1657 /* Get all supported features.
1658  * If a feature is not fixed, it is present in hw_features.
1659  * If a feature is fixed, it does not present in hw_features, but
1660  * always in features.
1661  */
1662 static inline netdev_features_t efx_supported_features(const struct efx_nic *efx)
1663 {
1664 	const struct net_device *net_dev = efx->net_dev;
1665 
1666 	return net_dev->features | net_dev->hw_features;
1667 }
1668 
1669 /* Get the current TX queue insert index. */
1670 static inline unsigned int
1671 efx_tx_queue_get_insert_index(const struct efx_tx_queue *tx_queue)
1672 {
1673 	return tx_queue->insert_count & tx_queue->ptr_mask;
1674 }
1675 
1676 /* Get a TX buffer. */
1677 static inline struct efx_tx_buffer *
1678 __efx_tx_queue_get_insert_buffer(const struct efx_tx_queue *tx_queue)
1679 {
1680 	return &tx_queue->buffer[efx_tx_queue_get_insert_index(tx_queue)];
1681 }
1682 
1683 /* Get a TX buffer, checking it's not currently in use. */
1684 static inline struct efx_tx_buffer *
1685 efx_tx_queue_get_insert_buffer(const struct efx_tx_queue *tx_queue)
1686 {
1687 	struct efx_tx_buffer *buffer =
1688 		__efx_tx_queue_get_insert_buffer(tx_queue);
1689 
1690 	EFX_WARN_ON_ONCE_PARANOID(buffer->len);
1691 	EFX_WARN_ON_ONCE_PARANOID(buffer->flags);
1692 	EFX_WARN_ON_ONCE_PARANOID(buffer->unmap_len);
1693 
1694 	return buffer;
1695 }
1696 
1697 #endif /* EFX_NET_DRIVER_H */
1698