1 /****************************************************************************
2  * Driver for Solarflare network controllers and boards
3  * Copyright 2005-2006 Fen Systems Ltd.
4  * Copyright 2005-2013 Solarflare Communications Inc.
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms of the GNU General Public License version 2 as published
8  * by the Free Software Foundation, incorporated herein by reference.
9  */
10 
11 /* Common definitions for all Efx net driver code */
12 
13 #ifndef EFX_NET_DRIVER_H
14 #define EFX_NET_DRIVER_H
15 
16 #include <linux/netdevice.h>
17 #include <linux/etherdevice.h>
18 #include <linux/ethtool.h>
19 #include <linux/if_vlan.h>
20 #include <linux/timer.h>
21 #include <linux/mdio.h>
22 #include <linux/list.h>
23 #include <linux/pci.h>
24 #include <linux/device.h>
25 #include <linux/highmem.h>
26 #include <linux/workqueue.h>
27 #include <linux/mutex.h>
28 #include <linux/vmalloc.h>
29 #include <linux/i2c.h>
30 #include <linux/mtd/mtd.h>
31 
32 #include "enum.h"
33 #include "bitfield.h"
34 #include "filter.h"
35 
36 /**************************************************************************
37  *
38  * Build definitions
39  *
40  **************************************************************************/
41 
42 #define EFX_DRIVER_VERSION	"4.0"
43 
44 #ifdef DEBUG
45 #define EFX_BUG_ON_PARANOID(x) BUG_ON(x)
46 #define EFX_WARN_ON_PARANOID(x) WARN_ON(x)
47 #else
48 #define EFX_BUG_ON_PARANOID(x) do {} while (0)
49 #define EFX_WARN_ON_PARANOID(x) do {} while (0)
50 #endif
51 
52 /**************************************************************************
53  *
54  * Efx data structures
55  *
56  **************************************************************************/
57 
58 #define EFX_MAX_CHANNELS 32U
59 #define EFX_MAX_RX_QUEUES EFX_MAX_CHANNELS
60 #define EFX_EXTRA_CHANNEL_IOV	0
61 #define EFX_EXTRA_CHANNEL_PTP	1
62 #define EFX_MAX_EXTRA_CHANNELS	2U
63 
64 /* Checksum generation is a per-queue option in hardware, so each
65  * queue visible to the networking core is backed by two hardware TX
66  * queues. */
67 #define EFX_MAX_TX_TC		2
68 #define EFX_MAX_CORE_TX_QUEUES	(EFX_MAX_TX_TC * EFX_MAX_CHANNELS)
69 #define EFX_TXQ_TYPE_OFFLOAD	1	/* flag */
70 #define EFX_TXQ_TYPE_HIGHPRI	2	/* flag */
71 #define EFX_TXQ_TYPES		4
72 #define EFX_MAX_TX_QUEUES	(EFX_TXQ_TYPES * EFX_MAX_CHANNELS)
73 
74 /* Maximum possible MTU the driver supports */
75 #define EFX_MAX_MTU (9 * 1024)
76 
77 /* Size of an RX scatter buffer.  Small enough to pack 2 into a 4K page,
78  * and should be a multiple of the cache line size.
79  */
80 #define EFX_RX_USR_BUF_SIZE	(2048 - 256)
81 
82 /* If possible, we should ensure cache line alignment at start and end
83  * of every buffer.  Otherwise, we just need to ensure 4-byte
84  * alignment of the network header.
85  */
86 #if NET_IP_ALIGN == 0
87 #define EFX_RX_BUF_ALIGNMENT	L1_CACHE_BYTES
88 #else
89 #define EFX_RX_BUF_ALIGNMENT	4
90 #endif
91 
92 /* Forward declare Precision Time Protocol (PTP) support structure. */
93 struct efx_ptp_data;
94 
95 struct efx_self_tests;
96 
97 /**
98  * struct efx_buffer - A general-purpose DMA buffer
99  * @addr: host base address of the buffer
100  * @dma_addr: DMA base address of the buffer
101  * @len: Buffer length, in bytes
102  *
103  * The NIC uses these buffers for its interrupt status registers and
104  * MAC stats dumps.
105  */
106 struct efx_buffer {
107 	void *addr;
108 	dma_addr_t dma_addr;
109 	unsigned int len;
110 };
111 
112 /**
113  * struct efx_special_buffer - DMA buffer entered into buffer table
114  * @buf: Standard &struct efx_buffer
115  * @index: Buffer index within controller;s buffer table
116  * @entries: Number of buffer table entries
117  *
118  * The NIC has a buffer table that maps buffers of size %EFX_BUF_SIZE.
119  * Event and descriptor rings are addressed via one or more buffer
120  * table entries (and so can be physically non-contiguous, although we
121  * currently do not take advantage of that).  On Falcon and Siena we
122  * have to take care of allocating and initialising the entries
123  * ourselves.  On later hardware this is managed by the firmware and
124  * @index and @entries are left as 0.
125  */
126 struct efx_special_buffer {
127 	struct efx_buffer buf;
128 	unsigned int index;
129 	unsigned int entries;
130 };
131 
132 /**
133  * struct efx_tx_buffer - buffer state for a TX descriptor
134  * @skb: When @flags & %EFX_TX_BUF_SKB, the associated socket buffer to be
135  *	freed when descriptor completes
136  * @heap_buf: When @flags & %EFX_TX_BUF_HEAP, the associated heap buffer to be
137  *	freed when descriptor completes.
138  * @option: When @flags & %EFX_TX_BUF_OPTION, a NIC-specific option descriptor.
139  * @dma_addr: DMA address of the fragment.
140  * @flags: Flags for allocation and DMA mapping type
141  * @len: Length of this fragment.
142  *	This field is zero when the queue slot is empty.
143  * @unmap_len: Length of this fragment to unmap
144  */
145 struct efx_tx_buffer {
146 	union {
147 		const struct sk_buff *skb;
148 		void *heap_buf;
149 	};
150 	union {
151 		efx_qword_t option;
152 		dma_addr_t dma_addr;
153 	};
154 	unsigned short flags;
155 	unsigned short len;
156 	unsigned short unmap_len;
157 };
158 #define EFX_TX_BUF_CONT		1	/* not last descriptor of packet */
159 #define EFX_TX_BUF_SKB		2	/* buffer is last part of skb */
160 #define EFX_TX_BUF_HEAP		4	/* buffer was allocated with kmalloc() */
161 #define EFX_TX_BUF_MAP_SINGLE	8	/* buffer was mapped with dma_map_single() */
162 #define EFX_TX_BUF_OPTION	0x10	/* empty buffer for option descriptor */
163 
164 /**
165  * struct efx_tx_queue - An Efx TX queue
166  *
167  * This is a ring buffer of TX fragments.
168  * Since the TX completion path always executes on the same
169  * CPU and the xmit path can operate on different CPUs,
170  * performance is increased by ensuring that the completion
171  * path and the xmit path operate on different cache lines.
172  * This is particularly important if the xmit path is always
173  * executing on one CPU which is different from the completion
174  * path.  There is also a cache line for members which are
175  * read but not written on the fast path.
176  *
177  * @efx: The associated Efx NIC
178  * @queue: DMA queue number
179  * @channel: The associated channel
180  * @core_txq: The networking core TX queue structure
181  * @buffer: The software buffer ring
182  * @tsoh_page: Array of pages of TSO header buffers
183  * @txd: The hardware descriptor ring
184  * @ptr_mask: The size of the ring minus 1.
185  * @initialised: Has hardware queue been initialised?
186  * @read_count: Current read pointer.
187  *	This is the number of buffers that have been removed from both rings.
188  * @old_write_count: The value of @write_count when last checked.
189  *	This is here for performance reasons.  The xmit path will
190  *	only get the up-to-date value of @write_count if this
191  *	variable indicates that the queue is empty.  This is to
192  *	avoid cache-line ping-pong between the xmit path and the
193  *	completion path.
194  * @merge_events: Number of TX merged completion events
195  * @insert_count: Current insert pointer
196  *	This is the number of buffers that have been added to the
197  *	software ring.
198  * @write_count: Current write pointer
199  *	This is the number of buffers that have been added to the
200  *	hardware ring.
201  * @old_read_count: The value of read_count when last checked.
202  *	This is here for performance reasons.  The xmit path will
203  *	only get the up-to-date value of read_count if this
204  *	variable indicates that the queue is full.  This is to
205  *	avoid cache-line ping-pong between the xmit path and the
206  *	completion path.
207  * @tso_bursts: Number of times TSO xmit invoked by kernel
208  * @tso_long_headers: Number of packets with headers too long for standard
209  *	blocks
210  * @tso_packets: Number of packets via the TSO xmit path
211  * @pushes: Number of times the TX push feature has been used
212  * @empty_read_count: If the completion path has seen the queue as empty
213  *	and the transmission path has not yet checked this, the value of
214  *	@read_count bitwise-added to %EFX_EMPTY_COUNT_VALID; otherwise 0.
215  */
216 struct efx_tx_queue {
217 	/* Members which don't change on the fast path */
218 	struct efx_nic *efx ____cacheline_aligned_in_smp;
219 	unsigned queue;
220 	struct efx_channel *channel;
221 	struct netdev_queue *core_txq;
222 	struct efx_tx_buffer *buffer;
223 	struct efx_buffer *tsoh_page;
224 	struct efx_special_buffer txd;
225 	unsigned int ptr_mask;
226 	bool initialised;
227 
228 	/* Members used mainly on the completion path */
229 	unsigned int read_count ____cacheline_aligned_in_smp;
230 	unsigned int old_write_count;
231 	unsigned int merge_events;
232 
233 	/* Members used only on the xmit path */
234 	unsigned int insert_count ____cacheline_aligned_in_smp;
235 	unsigned int write_count;
236 	unsigned int old_read_count;
237 	unsigned int tso_bursts;
238 	unsigned int tso_long_headers;
239 	unsigned int tso_packets;
240 	unsigned int pushes;
241 
242 	/* Members shared between paths and sometimes updated */
243 	unsigned int empty_read_count ____cacheline_aligned_in_smp;
244 #define EFX_EMPTY_COUNT_VALID 0x80000000
245 	atomic_t flush_outstanding;
246 };
247 
248 /**
249  * struct efx_rx_buffer - An Efx RX data buffer
250  * @dma_addr: DMA base address of the buffer
251  * @page: The associated page buffer.
252  *	Will be %NULL if the buffer slot is currently free.
253  * @page_offset: If pending: offset in @page of DMA base address.
254  *	If completed: offset in @page of Ethernet header.
255  * @len: If pending: length for DMA descriptor.
256  *	If completed: received length, excluding hash prefix.
257  * @flags: Flags for buffer and packet state.  These are only set on the
258  *	first buffer of a scattered packet.
259  */
260 struct efx_rx_buffer {
261 	dma_addr_t dma_addr;
262 	struct page *page;
263 	u16 page_offset;
264 	u16 len;
265 	u16 flags;
266 };
267 #define EFX_RX_BUF_LAST_IN_PAGE	0x0001
268 #define EFX_RX_PKT_CSUMMED	0x0002
269 #define EFX_RX_PKT_DISCARD	0x0004
270 #define EFX_RX_PKT_TCP		0x0040
271 #define EFX_RX_PKT_PREFIX_LEN	0x0080	/* length is in prefix only */
272 
273 /**
274  * struct efx_rx_page_state - Page-based rx buffer state
275  *
276  * Inserted at the start of every page allocated for receive buffers.
277  * Used to facilitate sharing dma mappings between recycled rx buffers
278  * and those passed up to the kernel.
279  *
280  * @refcnt: Number of struct efx_rx_buffer's referencing this page.
281  *	When refcnt falls to zero, the page is unmapped for dma
282  * @dma_addr: The dma address of this page.
283  */
284 struct efx_rx_page_state {
285 	unsigned refcnt;
286 	dma_addr_t dma_addr;
287 
288 	unsigned int __pad[0] ____cacheline_aligned;
289 };
290 
291 /**
292  * struct efx_rx_queue - An Efx RX queue
293  * @efx: The associated Efx NIC
294  * @core_index:  Index of network core RX queue.  Will be >= 0 iff this
295  *	is associated with a real RX queue.
296  * @buffer: The software buffer ring
297  * @rxd: The hardware descriptor ring
298  * @ptr_mask: The size of the ring minus 1.
299  * @refill_enabled: Enable refill whenever fill level is low
300  * @flush_pending: Set when a RX flush is pending. Has the same lifetime as
301  *	@rxq_flush_pending.
302  * @added_count: Number of buffers added to the receive queue.
303  * @notified_count: Number of buffers given to NIC (<= @added_count).
304  * @removed_count: Number of buffers removed from the receive queue.
305  * @scatter_n: Used by NIC specific receive code.
306  * @scatter_len: Used by NIC specific receive code.
307  * @page_ring: The ring to store DMA mapped pages for reuse.
308  * @page_add: Counter to calculate the write pointer for the recycle ring.
309  * @page_remove: Counter to calculate the read pointer for the recycle ring.
310  * @page_recycle_count: The number of pages that have been recycled.
311  * @page_recycle_failed: The number of pages that couldn't be recycled because
312  *      the kernel still held a reference to them.
313  * @page_recycle_full: The number of pages that were released because the
314  *      recycle ring was full.
315  * @page_ptr_mask: The number of pages in the RX recycle ring minus 1.
316  * @max_fill: RX descriptor maximum fill level (<= ring size)
317  * @fast_fill_trigger: RX descriptor fill level that will trigger a fast fill
318  *	(<= @max_fill)
319  * @min_fill: RX descriptor minimum non-zero fill level.
320  *	This records the minimum fill level observed when a ring
321  *	refill was triggered.
322  * @recycle_count: RX buffer recycle counter.
323  * @slow_fill: Timer used to defer efx_nic_generate_fill_event().
324  */
325 struct efx_rx_queue {
326 	struct efx_nic *efx;
327 	int core_index;
328 	struct efx_rx_buffer *buffer;
329 	struct efx_special_buffer rxd;
330 	unsigned int ptr_mask;
331 	bool refill_enabled;
332 	bool flush_pending;
333 
334 	unsigned int added_count;
335 	unsigned int notified_count;
336 	unsigned int removed_count;
337 	unsigned int scatter_n;
338 	unsigned int scatter_len;
339 	struct page **page_ring;
340 	unsigned int page_add;
341 	unsigned int page_remove;
342 	unsigned int page_recycle_count;
343 	unsigned int page_recycle_failed;
344 	unsigned int page_recycle_full;
345 	unsigned int page_ptr_mask;
346 	unsigned int max_fill;
347 	unsigned int fast_fill_trigger;
348 	unsigned int min_fill;
349 	unsigned int min_overfill;
350 	unsigned int recycle_count;
351 	struct timer_list slow_fill;
352 	unsigned int slow_fill_count;
353 };
354 
355 enum efx_rx_alloc_method {
356 	RX_ALLOC_METHOD_AUTO = 0,
357 	RX_ALLOC_METHOD_SKB = 1,
358 	RX_ALLOC_METHOD_PAGE = 2,
359 };
360 
361 /**
362  * struct efx_channel - An Efx channel
363  *
364  * A channel comprises an event queue, at least one TX queue, at least
365  * one RX queue, and an associated tasklet for processing the event
366  * queue.
367  *
368  * @efx: Associated Efx NIC
369  * @channel: Channel instance number
370  * @type: Channel type definition
371  * @eventq_init: Event queue initialised flag
372  * @enabled: Channel enabled indicator
373  * @irq: IRQ number (MSI and MSI-X only)
374  * @irq_moderation: IRQ moderation value (in hardware ticks)
375  * @napi_dev: Net device used with NAPI
376  * @napi_str: NAPI control structure
377  * @eventq: Event queue buffer
378  * @eventq_mask: Event queue pointer mask
379  * @eventq_read_ptr: Event queue read pointer
380  * @event_test_cpu: Last CPU to handle interrupt or test event for this channel
381  * @irq_count: Number of IRQs since last adaptive moderation decision
382  * @irq_mod_score: IRQ moderation score
383  * @n_rx_tobe_disc: Count of RX_TOBE_DISC errors
384  * @n_rx_ip_hdr_chksum_err: Count of RX IP header checksum errors
385  * @n_rx_tcp_udp_chksum_err: Count of RX TCP and UDP checksum errors
386  * @n_rx_mcast_mismatch: Count of unmatched multicast frames
387  * @n_rx_frm_trunc: Count of RX_FRM_TRUNC errors
388  * @n_rx_overlength: Count of RX_OVERLENGTH errors
389  * @n_skbuff_leaks: Count of skbuffs leaked due to RX overrun
390  * @n_rx_nodesc_trunc: Number of RX packets truncated and then dropped due to
391  *	lack of descriptors
392  * @n_rx_merge_events: Number of RX merged completion events
393  * @n_rx_merge_packets: Number of RX packets completed by merged events
394  * @rx_pkt_n_frags: Number of fragments in next packet to be delivered by
395  *	__efx_rx_packet(), or zero if there is none
396  * @rx_pkt_index: Ring index of first buffer for next packet to be delivered
397  *	by __efx_rx_packet(), if @rx_pkt_n_frags != 0
398  * @rx_queue: RX queue for this channel
399  * @tx_queue: TX queues for this channel
400  */
401 struct efx_channel {
402 	struct efx_nic *efx;
403 	int channel;
404 	const struct efx_channel_type *type;
405 	bool eventq_init;
406 	bool enabled;
407 	int irq;
408 	unsigned int irq_moderation;
409 	struct net_device *napi_dev;
410 	struct napi_struct napi_str;
411 	struct efx_special_buffer eventq;
412 	unsigned int eventq_mask;
413 	unsigned int eventq_read_ptr;
414 	int event_test_cpu;
415 
416 	unsigned int irq_count;
417 	unsigned int irq_mod_score;
418 #ifdef CONFIG_RFS_ACCEL
419 	unsigned int rfs_filters_added;
420 #endif
421 
422 	unsigned n_rx_tobe_disc;
423 	unsigned n_rx_ip_hdr_chksum_err;
424 	unsigned n_rx_tcp_udp_chksum_err;
425 	unsigned n_rx_mcast_mismatch;
426 	unsigned n_rx_frm_trunc;
427 	unsigned n_rx_overlength;
428 	unsigned n_skbuff_leaks;
429 	unsigned int n_rx_nodesc_trunc;
430 	unsigned int n_rx_merge_events;
431 	unsigned int n_rx_merge_packets;
432 
433 	unsigned int rx_pkt_n_frags;
434 	unsigned int rx_pkt_index;
435 
436 	struct efx_rx_queue rx_queue;
437 	struct efx_tx_queue tx_queue[EFX_TXQ_TYPES];
438 };
439 
440 /**
441  * struct efx_msi_context - Context for each MSI
442  * @efx: The associated NIC
443  * @index: Index of the channel/IRQ
444  * @name: Name of the channel/IRQ
445  *
446  * Unlike &struct efx_channel, this is never reallocated and is always
447  * safe for the IRQ handler to access.
448  */
449 struct efx_msi_context {
450 	struct efx_nic *efx;
451 	unsigned int index;
452 	char name[IFNAMSIZ + 6];
453 };
454 
455 /**
456  * struct efx_channel_type - distinguishes traffic and extra channels
457  * @handle_no_channel: Handle failure to allocate an extra channel
458  * @pre_probe: Set up extra state prior to initialisation
459  * @post_remove: Tear down extra state after finalisation, if allocated.
460  *	May be called on channels that have not been probed.
461  * @get_name: Generate the channel's name (used for its IRQ handler)
462  * @copy: Copy the channel state prior to reallocation.  May be %NULL if
463  *	reallocation is not supported.
464  * @receive_skb: Handle an skb ready to be passed to netif_receive_skb()
465  * @keep_eventq: Flag for whether event queue should be kept initialised
466  *	while the device is stopped
467  */
468 struct efx_channel_type {
469 	void (*handle_no_channel)(struct efx_nic *);
470 	int (*pre_probe)(struct efx_channel *);
471 	void (*post_remove)(struct efx_channel *);
472 	void (*get_name)(struct efx_channel *, char *buf, size_t len);
473 	struct efx_channel *(*copy)(const struct efx_channel *);
474 	bool (*receive_skb)(struct efx_channel *, struct sk_buff *);
475 	bool keep_eventq;
476 };
477 
478 enum efx_led_mode {
479 	EFX_LED_OFF	= 0,
480 	EFX_LED_ON	= 1,
481 	EFX_LED_DEFAULT	= 2
482 };
483 
484 #define STRING_TABLE_LOOKUP(val, member) \
485 	((val) < member ## _max) ? member ## _names[val] : "(invalid)"
486 
487 extern const char *const efx_loopback_mode_names[];
488 extern const unsigned int efx_loopback_mode_max;
489 #define LOOPBACK_MODE(efx) \
490 	STRING_TABLE_LOOKUP((efx)->loopback_mode, efx_loopback_mode)
491 
492 extern const char *const efx_reset_type_names[];
493 extern const unsigned int efx_reset_type_max;
494 #define RESET_TYPE(type) \
495 	STRING_TABLE_LOOKUP(type, efx_reset_type)
496 
497 enum efx_int_mode {
498 	/* Be careful if altering to correct macro below */
499 	EFX_INT_MODE_MSIX = 0,
500 	EFX_INT_MODE_MSI = 1,
501 	EFX_INT_MODE_LEGACY = 2,
502 	EFX_INT_MODE_MAX	/* Insert any new items before this */
503 };
504 #define EFX_INT_MODE_USE_MSI(x) (((x)->interrupt_mode) <= EFX_INT_MODE_MSI)
505 
506 enum nic_state {
507 	STATE_UNINIT = 0,	/* device being probed/removed or is frozen */
508 	STATE_READY = 1,	/* hardware ready and netdev registered */
509 	STATE_DISABLED = 2,	/* device disabled due to hardware errors */
510 	STATE_RECOVERY = 3,	/* device recovering from PCI error */
511 };
512 
513 /*
514  * Alignment of the skb->head which wraps a page-allocated RX buffer
515  *
516  * The skb allocated to wrap an rx_buffer can have this alignment. Since
517  * the data is memcpy'd from the rx_buf, it does not need to be equal to
518  * NET_IP_ALIGN.
519  */
520 #define EFX_PAGE_SKB_ALIGN 2
521 
522 /* Forward declaration */
523 struct efx_nic;
524 
525 /* Pseudo bit-mask flow control field */
526 #define EFX_FC_RX	FLOW_CTRL_RX
527 #define EFX_FC_TX	FLOW_CTRL_TX
528 #define EFX_FC_AUTO	4
529 
530 /**
531  * struct efx_link_state - Current state of the link
532  * @up: Link is up
533  * @fd: Link is full-duplex
534  * @fc: Actual flow control flags
535  * @speed: Link speed (Mbps)
536  */
537 struct efx_link_state {
538 	bool up;
539 	bool fd;
540 	u8 fc;
541 	unsigned int speed;
542 };
543 
544 static inline bool efx_link_state_equal(const struct efx_link_state *left,
545 					const struct efx_link_state *right)
546 {
547 	return left->up == right->up && left->fd == right->fd &&
548 		left->fc == right->fc && left->speed == right->speed;
549 }
550 
551 /**
552  * struct efx_phy_operations - Efx PHY operations table
553  * @probe: Probe PHY and initialise efx->mdio.mode_support, efx->mdio.mmds,
554  *	efx->loopback_modes.
555  * @init: Initialise PHY
556  * @fini: Shut down PHY
557  * @reconfigure: Reconfigure PHY (e.g. for new link parameters)
558  * @poll: Update @link_state and report whether it changed.
559  *	Serialised by the mac_lock.
560  * @get_settings: Get ethtool settings. Serialised by the mac_lock.
561  * @set_settings: Set ethtool settings. Serialised by the mac_lock.
562  * @set_npage_adv: Set abilities advertised in (Extended) Next Page
563  *	(only needed where AN bit is set in mmds)
564  * @test_alive: Test that PHY is 'alive' (online)
565  * @test_name: Get the name of a PHY-specific test/result
566  * @run_tests: Run tests and record results as appropriate (offline).
567  *	Flags are the ethtool tests flags.
568  */
569 struct efx_phy_operations {
570 	int (*probe) (struct efx_nic *efx);
571 	int (*init) (struct efx_nic *efx);
572 	void (*fini) (struct efx_nic *efx);
573 	void (*remove) (struct efx_nic *efx);
574 	int (*reconfigure) (struct efx_nic *efx);
575 	bool (*poll) (struct efx_nic *efx);
576 	void (*get_settings) (struct efx_nic *efx,
577 			      struct ethtool_cmd *ecmd);
578 	int (*set_settings) (struct efx_nic *efx,
579 			     struct ethtool_cmd *ecmd);
580 	void (*set_npage_adv) (struct efx_nic *efx, u32);
581 	int (*test_alive) (struct efx_nic *efx);
582 	const char *(*test_name) (struct efx_nic *efx, unsigned int index);
583 	int (*run_tests) (struct efx_nic *efx, int *results, unsigned flags);
584 	int (*get_module_eeprom) (struct efx_nic *efx,
585 			       struct ethtool_eeprom *ee,
586 			       u8 *data);
587 	int (*get_module_info) (struct efx_nic *efx,
588 				struct ethtool_modinfo *modinfo);
589 };
590 
591 /**
592  * enum efx_phy_mode - PHY operating mode flags
593  * @PHY_MODE_NORMAL: on and should pass traffic
594  * @PHY_MODE_TX_DISABLED: on with TX disabled
595  * @PHY_MODE_LOW_POWER: set to low power through MDIO
596  * @PHY_MODE_OFF: switched off through external control
597  * @PHY_MODE_SPECIAL: on but will not pass traffic
598  */
599 enum efx_phy_mode {
600 	PHY_MODE_NORMAL		= 0,
601 	PHY_MODE_TX_DISABLED	= 1,
602 	PHY_MODE_LOW_POWER	= 2,
603 	PHY_MODE_OFF		= 4,
604 	PHY_MODE_SPECIAL	= 8,
605 };
606 
607 static inline bool efx_phy_mode_disabled(enum efx_phy_mode mode)
608 {
609 	return !!(mode & ~PHY_MODE_TX_DISABLED);
610 }
611 
612 /**
613  * struct efx_hw_stat_desc - Description of a hardware statistic
614  * @name: Name of the statistic as visible through ethtool, or %NULL if
615  *	it should not be exposed
616  * @dma_width: Width in bits (0 for non-DMA statistics)
617  * @offset: Offset within stats (ignored for non-DMA statistics)
618  */
619 struct efx_hw_stat_desc {
620 	const char *name;
621 	u16 dma_width;
622 	u16 offset;
623 };
624 
625 /* Number of bits used in a multicast filter hash address */
626 #define EFX_MCAST_HASH_BITS 8
627 
628 /* Number of (single-bit) entries in a multicast filter hash */
629 #define EFX_MCAST_HASH_ENTRIES (1 << EFX_MCAST_HASH_BITS)
630 
631 /* An Efx multicast filter hash */
632 union efx_multicast_hash {
633 	u8 byte[EFX_MCAST_HASH_ENTRIES / 8];
634 	efx_oword_t oword[EFX_MCAST_HASH_ENTRIES / sizeof(efx_oword_t) / 8];
635 };
636 
637 struct efx_vf;
638 struct vfdi_status;
639 
640 /**
641  * struct efx_nic - an Efx NIC
642  * @name: Device name (net device name or bus id before net device registered)
643  * @pci_dev: The PCI device
644  * @type: Controller type attributes
645  * @legacy_irq: IRQ number
646  * @workqueue: Workqueue for port reconfigures and the HW monitor.
647  *	Work items do not hold and must not acquire RTNL.
648  * @workqueue_name: Name of workqueue
649  * @reset_work: Scheduled reset workitem
650  * @membase_phys: Memory BAR value as physical address
651  * @membase: Memory BAR value
652  * @interrupt_mode: Interrupt mode
653  * @timer_quantum_ns: Interrupt timer quantum, in nanoseconds
654  * @irq_rx_adaptive: Adaptive IRQ moderation enabled for RX event queues
655  * @irq_rx_moderation: IRQ moderation time for RX event queues
656  * @msg_enable: Log message enable flags
657  * @state: Device state number (%STATE_*). Serialised by the rtnl_lock.
658  * @reset_pending: Bitmask for pending resets
659  * @tx_queue: TX DMA queues
660  * @rx_queue: RX DMA queues
661  * @channel: Channels
662  * @msi_context: Context for each MSI
663  * @extra_channel_types: Types of extra (non-traffic) channels that
664  *	should be allocated for this NIC
665  * @rxq_entries: Size of receive queues requested by user.
666  * @txq_entries: Size of transmit queues requested by user.
667  * @txq_stop_thresh: TX queue fill level at or above which we stop it.
668  * @txq_wake_thresh: TX queue fill level at or below which we wake it.
669  * @tx_dc_base: Base qword address in SRAM of TX queue descriptor caches
670  * @rx_dc_base: Base qword address in SRAM of RX queue descriptor caches
671  * @sram_lim_qw: Qword address limit of SRAM
672  * @next_buffer_table: First available buffer table id
673  * @n_channels: Number of channels in use
674  * @n_rx_channels: Number of channels used for RX (= number of RX queues)
675  * @n_tx_channels: Number of channels used for TX
676  * @rx_dma_len: Current maximum RX DMA length
677  * @rx_buffer_order: Order (log2) of number of pages for each RX buffer
678  * @rx_buffer_truesize: Amortised allocation size of an RX buffer,
679  *	for use in sk_buff::truesize
680  * @rx_prefix_size: Size of RX prefix before packet data
681  * @rx_packet_hash_offset: Offset of RX flow hash from start of packet data
682  *	(valid only if @rx_prefix_size != 0; always negative)
683  * @rx_packet_len_offset: Offset of RX packet length from start of packet data
684  *	(valid only for NICs that set %EFX_RX_PKT_PREFIX_LEN; always negative)
685  * @rx_hash_key: Toeplitz hash key for RSS
686  * @rx_indir_table: Indirection table for RSS
687  * @rx_scatter: Scatter mode enabled for receives
688  * @int_error_count: Number of internal errors seen recently
689  * @int_error_expire: Time at which error count will be expired
690  * @irq_soft_enabled: Are IRQs soft-enabled? If not, IRQ handler will
691  *	acknowledge but do nothing else.
692  * @irq_status: Interrupt status buffer
693  * @irq_zero_count: Number of legacy IRQs seen with queue flags == 0
694  * @irq_level: IRQ level/index for IRQs not triggered by an event queue
695  * @selftest_work: Work item for asynchronous self-test
696  * @mtd_list: List of MTDs attached to the NIC
697  * @nic_data: Hardware dependent state
698  * @mcdi: Management-Controller-to-Driver Interface state
699  * @mac_lock: MAC access lock. Protects @port_enabled, @phy_mode,
700  *	efx_monitor() and efx_reconfigure_port()
701  * @port_enabled: Port enabled indicator.
702  *	Serialises efx_stop_all(), efx_start_all(), efx_monitor() and
703  *	efx_mac_work() with kernel interfaces. Safe to read under any
704  *	one of the rtnl_lock, mac_lock, or netif_tx_lock, but all three must
705  *	be held to modify it.
706  * @port_initialized: Port initialized?
707  * @net_dev: Operating system network device. Consider holding the rtnl lock
708  * @stats_buffer: DMA buffer for statistics
709  * @phy_type: PHY type
710  * @phy_op: PHY interface
711  * @phy_data: PHY private data (including PHY-specific stats)
712  * @mdio: PHY MDIO interface
713  * @mdio_bus: PHY MDIO bus ID (only used by Siena)
714  * @phy_mode: PHY operating mode. Serialised by @mac_lock.
715  * @link_advertising: Autonegotiation advertising flags
716  * @link_state: Current state of the link
717  * @n_link_state_changes: Number of times the link has changed state
718  * @unicast_filter: Flag for Falcon-arch simple unicast filter.
719  *	Protected by @mac_lock.
720  * @multicast_hash: Multicast hash table for Falcon-arch.
721  *	Protected by @mac_lock.
722  * @wanted_fc: Wanted flow control flags
723  * @fc_disable: When non-zero flow control is disabled. Typically used to
724  *	ensure that network back pressure doesn't delay dma queue flushes.
725  *	Serialised by the rtnl lock.
726  * @mac_work: Work item for changing MAC promiscuity and multicast hash
727  * @loopback_mode: Loopback status
728  * @loopback_modes: Supported loopback mode bitmask
729  * @loopback_selftest: Offline self-test private state
730  * @filter_lock: Filter table lock
731  * @filter_state: Architecture-dependent filter table state
732  * @rps_flow_id: Flow IDs of filters allocated for accelerated RFS,
733  *	indexed by filter ID
734  * @rps_expire_index: Next index to check for expiry in @rps_flow_id
735  * @active_queues: Count of RX and TX queues that haven't been flushed and drained.
736  * @rxq_flush_pending: Count of number of receive queues that need to be flushed.
737  *	Decremented when the efx_flush_rx_queue() is called.
738  * @rxq_flush_outstanding: Count of number of RX flushes started but not yet
739  *	completed (either success or failure). Not used when MCDI is used to
740  *	flush receive queues.
741  * @flush_wq: wait queue used by efx_nic_flush_queues() to wait for flush completions.
742  * @vf: Array of &struct efx_vf objects.
743  * @vf_count: Number of VFs intended to be enabled.
744  * @vf_init_count: Number of VFs that have been fully initialised.
745  * @vi_scale: log2 number of vnics per VF.
746  * @vf_buftbl_base: The zeroth buffer table index used to back VF queues.
747  * @vfdi_status: Common VFDI status page to be dmad to VF address space.
748  * @local_addr_list: List of local addresses. Protected by %local_lock.
749  * @local_page_list: List of DMA addressable pages used to broadcast
750  *	%local_addr_list. Protected by %local_lock.
751  * @local_lock: Mutex protecting %local_addr_list and %local_page_list.
752  * @peer_work: Work item to broadcast peer addresses to VMs.
753  * @ptp_data: PTP state data
754  * @monitor_work: Hardware monitor workitem
755  * @biu_lock: BIU (bus interface unit) lock
756  * @last_irq_cpu: Last CPU to handle a possible test interrupt.  This
757  *	field is used by efx_test_interrupts() to verify that an
758  *	interrupt has occurred.
759  * @stats_lock: Statistics update lock. Must be held when calling
760  *	efx_nic_type::{update,start,stop}_stats.
761  *
762  * This is stored in the private area of the &struct net_device.
763  */
764 struct efx_nic {
765 	/* The following fields should be written very rarely */
766 
767 	char name[IFNAMSIZ];
768 	struct pci_dev *pci_dev;
769 	unsigned int port_num;
770 	const struct efx_nic_type *type;
771 	int legacy_irq;
772 	bool eeh_disabled_legacy_irq;
773 	struct workqueue_struct *workqueue;
774 	char workqueue_name[16];
775 	struct work_struct reset_work;
776 	resource_size_t membase_phys;
777 	void __iomem *membase;
778 
779 	enum efx_int_mode interrupt_mode;
780 	unsigned int timer_quantum_ns;
781 	bool irq_rx_adaptive;
782 	unsigned int irq_rx_moderation;
783 	u32 msg_enable;
784 
785 	enum nic_state state;
786 	unsigned long reset_pending;
787 
788 	struct efx_channel *channel[EFX_MAX_CHANNELS];
789 	struct efx_msi_context msi_context[EFX_MAX_CHANNELS];
790 	const struct efx_channel_type *
791 	extra_channel_type[EFX_MAX_EXTRA_CHANNELS];
792 
793 	unsigned rxq_entries;
794 	unsigned txq_entries;
795 	unsigned int txq_stop_thresh;
796 	unsigned int txq_wake_thresh;
797 
798 	unsigned tx_dc_base;
799 	unsigned rx_dc_base;
800 	unsigned sram_lim_qw;
801 	unsigned next_buffer_table;
802 
803 	unsigned int max_channels;
804 	unsigned n_channels;
805 	unsigned n_rx_channels;
806 	unsigned rss_spread;
807 	unsigned tx_channel_offset;
808 	unsigned n_tx_channels;
809 	unsigned int rx_dma_len;
810 	unsigned int rx_buffer_order;
811 	unsigned int rx_buffer_truesize;
812 	unsigned int rx_page_buf_step;
813 	unsigned int rx_bufs_per_page;
814 	unsigned int rx_pages_per_batch;
815 	unsigned int rx_prefix_size;
816 	int rx_packet_hash_offset;
817 	int rx_packet_len_offset;
818 	u8 rx_hash_key[40];
819 	u32 rx_indir_table[128];
820 	bool rx_scatter;
821 
822 	unsigned int_error_count;
823 	unsigned long int_error_expire;
824 
825 	bool irq_soft_enabled;
826 	struct efx_buffer irq_status;
827 	unsigned irq_zero_count;
828 	unsigned irq_level;
829 	struct delayed_work selftest_work;
830 
831 #ifdef CONFIG_SFC_MTD
832 	struct list_head mtd_list;
833 #endif
834 
835 	void *nic_data;
836 	struct efx_mcdi_data *mcdi;
837 
838 	struct mutex mac_lock;
839 	struct work_struct mac_work;
840 	bool port_enabled;
841 
842 	bool port_initialized;
843 	struct net_device *net_dev;
844 
845 	struct efx_buffer stats_buffer;
846 
847 	unsigned int phy_type;
848 	const struct efx_phy_operations *phy_op;
849 	void *phy_data;
850 	struct mdio_if_info mdio;
851 	unsigned int mdio_bus;
852 	enum efx_phy_mode phy_mode;
853 
854 	u32 link_advertising;
855 	struct efx_link_state link_state;
856 	unsigned int n_link_state_changes;
857 
858 	bool unicast_filter;
859 	union efx_multicast_hash multicast_hash;
860 	u8 wanted_fc;
861 	unsigned fc_disable;
862 
863 	atomic_t rx_reset;
864 	enum efx_loopback_mode loopback_mode;
865 	u64 loopback_modes;
866 
867 	void *loopback_selftest;
868 
869 	spinlock_t filter_lock;
870 	void *filter_state;
871 #ifdef CONFIG_RFS_ACCEL
872 	u32 *rps_flow_id;
873 	unsigned int rps_expire_index;
874 #endif
875 
876 	atomic_t active_queues;
877 	atomic_t rxq_flush_pending;
878 	atomic_t rxq_flush_outstanding;
879 	wait_queue_head_t flush_wq;
880 
881 #ifdef CONFIG_SFC_SRIOV
882 	struct efx_channel *vfdi_channel;
883 	struct efx_vf *vf;
884 	unsigned vf_count;
885 	unsigned vf_init_count;
886 	unsigned vi_scale;
887 	unsigned vf_buftbl_base;
888 	struct efx_buffer vfdi_status;
889 	struct list_head local_addr_list;
890 	struct list_head local_page_list;
891 	struct mutex local_lock;
892 	struct work_struct peer_work;
893 #endif
894 
895 	struct efx_ptp_data *ptp_data;
896 
897 	/* The following fields may be written more often */
898 
899 	struct delayed_work monitor_work ____cacheline_aligned_in_smp;
900 	spinlock_t biu_lock;
901 	int last_irq_cpu;
902 	spinlock_t stats_lock;
903 };
904 
905 static inline int efx_dev_registered(struct efx_nic *efx)
906 {
907 	return efx->net_dev->reg_state == NETREG_REGISTERED;
908 }
909 
910 static inline unsigned int efx_port_num(struct efx_nic *efx)
911 {
912 	return efx->port_num;
913 }
914 
915 struct efx_mtd_partition {
916 	struct list_head node;
917 	struct mtd_info mtd;
918 	const char *dev_type_name;
919 	const char *type_name;
920 	char name[IFNAMSIZ + 20];
921 };
922 
923 /**
924  * struct efx_nic_type - Efx device type definition
925  * @mem_map_size: Get memory BAR mapped size
926  * @probe: Probe the controller
927  * @remove: Free resources allocated by probe()
928  * @init: Initialise the controller
929  * @dimension_resources: Dimension controller resources (buffer table,
930  *	and VIs once the available interrupt resources are clear)
931  * @fini: Shut down the controller
932  * @monitor: Periodic function for polling link state and hardware monitor
933  * @map_reset_reason: Map ethtool reset reason to a reset method
934  * @map_reset_flags: Map ethtool reset flags to a reset method, if possible
935  * @reset: Reset the controller hardware and possibly the PHY.  This will
936  *	be called while the controller is uninitialised.
937  * @probe_port: Probe the MAC and PHY
938  * @remove_port: Free resources allocated by probe_port()
939  * @handle_global_event: Handle a "global" event (may be %NULL)
940  * @fini_dmaq: Flush and finalise DMA queues (RX and TX queues)
941  * @prepare_flush: Prepare the hardware for flushing the DMA queues
942  *	(for Falcon architecture)
943  * @finish_flush: Clean up after flushing the DMA queues (for Falcon
944  *	architecture)
945  * @describe_stats: Describe statistics for ethtool
946  * @update_stats: Update statistics not provided by event handling.
947  *	Either argument may be %NULL.
948  * @start_stats: Start the regular fetching of statistics
949  * @stop_stats: Stop the regular fetching of statistics
950  * @set_id_led: Set state of identifying LED or revert to automatic function
951  * @push_irq_moderation: Apply interrupt moderation value
952  * @reconfigure_port: Push loopback/power/txdis changes to the MAC and PHY
953  * @prepare_enable_fc_tx: Prepare MAC to enable pause frame TX (may be %NULL)
954  * @reconfigure_mac: Push MAC address, MTU, flow control and filter settings
955  *	to the hardware.  Serialised by the mac_lock.
956  * @check_mac_fault: Check MAC fault state. True if fault present.
957  * @get_wol: Get WoL configuration from driver state
958  * @set_wol: Push WoL configuration to the NIC
959  * @resume_wol: Synchronise WoL state between driver and MC (e.g. after resume)
960  * @test_chip: Test registers.  May use efx_farch_test_registers(), and is
961  *	expected to reset the NIC.
962  * @test_nvram: Test validity of NVRAM contents
963  * @mcdi_request: Send an MCDI request with the given header and SDU.
964  *	The SDU length may be any value from 0 up to the protocol-
965  *	defined maximum, but its buffer will be padded to a multiple
966  *	of 4 bytes.
967  * @mcdi_poll_response: Test whether an MCDI response is available.
968  * @mcdi_read_response: Read the MCDI response PDU.  The offset will
969  *	be a multiple of 4.  The length may not be, but the buffer
970  *	will be padded so it is safe to round up.
971  * @mcdi_poll_reboot: Test whether the MCDI has rebooted.  If so,
972  *	return an appropriate error code for aborting any current
973  *	request; otherwise return 0.
974  * @irq_enable_master: Enable IRQs on the NIC.  Each event queue must
975  *	be separately enabled after this.
976  * @irq_test_generate: Generate a test IRQ
977  * @irq_disable_non_ev: Disable non-event IRQs on the NIC.  Each event
978  *	queue must be separately disabled before this.
979  * @irq_handle_msi: Handle MSI for a channel.  The @dev_id argument is
980  *	a pointer to the &struct efx_msi_context for the channel.
981  * @irq_handle_legacy: Handle legacy interrupt.  The @dev_id argument
982  *	is a pointer to the &struct efx_nic.
983  * @tx_probe: Allocate resources for TX queue
984  * @tx_init: Initialise TX queue on the NIC
985  * @tx_remove: Free resources for TX queue
986  * @tx_write: Write TX descriptors and doorbell
987  * @rx_push_indir_table: Write RSS indirection table to the NIC
988  * @rx_probe: Allocate resources for RX queue
989  * @rx_init: Initialise RX queue on the NIC
990  * @rx_remove: Free resources for RX queue
991  * @rx_write: Write RX descriptors and doorbell
992  * @rx_defer_refill: Generate a refill reminder event
993  * @ev_probe: Allocate resources for event queue
994  * @ev_init: Initialise event queue on the NIC
995  * @ev_fini: Deinitialise event queue on the NIC
996  * @ev_remove: Free resources for event queue
997  * @ev_process: Process events for a queue, up to the given NAPI quota
998  * @ev_read_ack: Acknowledge read events on a queue, rearming its IRQ
999  * @ev_test_generate: Generate a test event
1000  * @filter_table_probe: Probe filter capabilities and set up filter software state
1001  * @filter_table_restore: Restore filters removed from hardware
1002  * @filter_table_remove: Remove filters from hardware and tear down software state
1003  * @filter_update_rx_scatter: Update filters after change to rx scatter setting
1004  * @filter_insert: add or replace a filter
1005  * @filter_remove_safe: remove a filter by ID, carefully
1006  * @filter_get_safe: retrieve a filter by ID, carefully
1007  * @filter_clear_rx: remove RX filters by priority
1008  * @filter_count_rx_used: Get the number of filters in use at a given priority
1009  * @filter_get_rx_id_limit: Get maximum value of a filter id, plus 1
1010  * @filter_get_rx_ids: Get list of RX filters at a given priority
1011  * @filter_rfs_insert: Add or replace a filter for RFS.  This must be
1012  *	atomic.  The hardware change may be asynchronous but should
1013  *	not be delayed for long.  It may fail if this can't be done
1014  *	atomically.
1015  * @filter_rfs_expire_one: Consider expiring a filter inserted for RFS.
1016  *	This must check whether the specified table entry is used by RFS
1017  *	and that rps_may_expire_flow() returns true for it.
1018  * @mtd_probe: Probe and add MTD partitions associated with this net device,
1019  *	 using efx_mtd_add()
1020  * @mtd_rename: Set an MTD partition name using the net device name
1021  * @mtd_read: Read from an MTD partition
1022  * @mtd_erase: Erase part of an MTD partition
1023  * @mtd_write: Write to an MTD partition
1024  * @mtd_sync: Wait for write-back to complete on MTD partition.  This
1025  *	also notifies the driver that a writer has finished using this
1026  *	partition.
1027  * @revision: Hardware architecture revision
1028  * @txd_ptr_tbl_base: TX descriptor ring base address
1029  * @rxd_ptr_tbl_base: RX descriptor ring base address
1030  * @buf_tbl_base: Buffer table base address
1031  * @evq_ptr_tbl_base: Event queue pointer table base address
1032  * @evq_rptr_tbl_base: Event queue read-pointer table base address
1033  * @max_dma_mask: Maximum possible DMA mask
1034  * @rx_prefix_size: Size of RX prefix before packet data
1035  * @rx_hash_offset: Offset of RX flow hash within prefix
1036  * @rx_buffer_padding: Size of padding at end of RX packet
1037  * @can_rx_scatter: NIC is able to scatter packets to multiple buffers
1038  * @always_rx_scatter: NIC will always scatter packets to multiple buffers
1039  * @max_interrupt_mode: Highest capability interrupt mode supported
1040  *	from &enum efx_init_mode.
1041  * @timer_period_max: Maximum period of interrupt timer (in ticks)
1042  * @offload_features: net_device feature flags for protocol offload
1043  *	features implemented in hardware
1044  * @mcdi_max_ver: Maximum MCDI version supported
1045  */
1046 struct efx_nic_type {
1047 	unsigned int (*mem_map_size)(struct efx_nic *efx);
1048 	int (*probe)(struct efx_nic *efx);
1049 	void (*remove)(struct efx_nic *efx);
1050 	int (*init)(struct efx_nic *efx);
1051 	int (*dimension_resources)(struct efx_nic *efx);
1052 	void (*fini)(struct efx_nic *efx);
1053 	void (*monitor)(struct efx_nic *efx);
1054 	enum reset_type (*map_reset_reason)(enum reset_type reason);
1055 	int (*map_reset_flags)(u32 *flags);
1056 	int (*reset)(struct efx_nic *efx, enum reset_type method);
1057 	int (*probe_port)(struct efx_nic *efx);
1058 	void (*remove_port)(struct efx_nic *efx);
1059 	bool (*handle_global_event)(struct efx_channel *channel, efx_qword_t *);
1060 	int (*fini_dmaq)(struct efx_nic *efx);
1061 	void (*prepare_flush)(struct efx_nic *efx);
1062 	void (*finish_flush)(struct efx_nic *efx);
1063 	size_t (*describe_stats)(struct efx_nic *efx, u8 *names);
1064 	size_t (*update_stats)(struct efx_nic *efx, u64 *full_stats,
1065 			       struct rtnl_link_stats64 *core_stats);
1066 	void (*start_stats)(struct efx_nic *efx);
1067 	void (*stop_stats)(struct efx_nic *efx);
1068 	void (*set_id_led)(struct efx_nic *efx, enum efx_led_mode mode);
1069 	void (*push_irq_moderation)(struct efx_channel *channel);
1070 	int (*reconfigure_port)(struct efx_nic *efx);
1071 	void (*prepare_enable_fc_tx)(struct efx_nic *efx);
1072 	int (*reconfigure_mac)(struct efx_nic *efx);
1073 	bool (*check_mac_fault)(struct efx_nic *efx);
1074 	void (*get_wol)(struct efx_nic *efx, struct ethtool_wolinfo *wol);
1075 	int (*set_wol)(struct efx_nic *efx, u32 type);
1076 	void (*resume_wol)(struct efx_nic *efx);
1077 	int (*test_chip)(struct efx_nic *efx, struct efx_self_tests *tests);
1078 	int (*test_nvram)(struct efx_nic *efx);
1079 	void (*mcdi_request)(struct efx_nic *efx,
1080 			     const efx_dword_t *hdr, size_t hdr_len,
1081 			     const efx_dword_t *sdu, size_t sdu_len);
1082 	bool (*mcdi_poll_response)(struct efx_nic *efx);
1083 	void (*mcdi_read_response)(struct efx_nic *efx, efx_dword_t *pdu,
1084 				   size_t pdu_offset, size_t pdu_len);
1085 	int (*mcdi_poll_reboot)(struct efx_nic *efx);
1086 	void (*irq_enable_master)(struct efx_nic *efx);
1087 	void (*irq_test_generate)(struct efx_nic *efx);
1088 	void (*irq_disable_non_ev)(struct efx_nic *efx);
1089 	irqreturn_t (*irq_handle_msi)(int irq, void *dev_id);
1090 	irqreturn_t (*irq_handle_legacy)(int irq, void *dev_id);
1091 	int (*tx_probe)(struct efx_tx_queue *tx_queue);
1092 	void (*tx_init)(struct efx_tx_queue *tx_queue);
1093 	void (*tx_remove)(struct efx_tx_queue *tx_queue);
1094 	void (*tx_write)(struct efx_tx_queue *tx_queue);
1095 	void (*rx_push_indir_table)(struct efx_nic *efx);
1096 	int (*rx_probe)(struct efx_rx_queue *rx_queue);
1097 	void (*rx_init)(struct efx_rx_queue *rx_queue);
1098 	void (*rx_remove)(struct efx_rx_queue *rx_queue);
1099 	void (*rx_write)(struct efx_rx_queue *rx_queue);
1100 	void (*rx_defer_refill)(struct efx_rx_queue *rx_queue);
1101 	int (*ev_probe)(struct efx_channel *channel);
1102 	int (*ev_init)(struct efx_channel *channel);
1103 	void (*ev_fini)(struct efx_channel *channel);
1104 	void (*ev_remove)(struct efx_channel *channel);
1105 	int (*ev_process)(struct efx_channel *channel, int quota);
1106 	void (*ev_read_ack)(struct efx_channel *channel);
1107 	void (*ev_test_generate)(struct efx_channel *channel);
1108 	int (*filter_table_probe)(struct efx_nic *efx);
1109 	void (*filter_table_restore)(struct efx_nic *efx);
1110 	void (*filter_table_remove)(struct efx_nic *efx);
1111 	void (*filter_update_rx_scatter)(struct efx_nic *efx);
1112 	s32 (*filter_insert)(struct efx_nic *efx,
1113 			     struct efx_filter_spec *spec, bool replace);
1114 	int (*filter_remove_safe)(struct efx_nic *efx,
1115 				  enum efx_filter_priority priority,
1116 				  u32 filter_id);
1117 	int (*filter_get_safe)(struct efx_nic *efx,
1118 			       enum efx_filter_priority priority,
1119 			       u32 filter_id, struct efx_filter_spec *);
1120 	void (*filter_clear_rx)(struct efx_nic *efx,
1121 				enum efx_filter_priority priority);
1122 	u32 (*filter_count_rx_used)(struct efx_nic *efx,
1123 				    enum efx_filter_priority priority);
1124 	u32 (*filter_get_rx_id_limit)(struct efx_nic *efx);
1125 	s32 (*filter_get_rx_ids)(struct efx_nic *efx,
1126 				 enum efx_filter_priority priority,
1127 				 u32 *buf, u32 size);
1128 #ifdef CONFIG_RFS_ACCEL
1129 	s32 (*filter_rfs_insert)(struct efx_nic *efx,
1130 				 struct efx_filter_spec *spec);
1131 	bool (*filter_rfs_expire_one)(struct efx_nic *efx, u32 flow_id,
1132 				      unsigned int index);
1133 #endif
1134 #ifdef CONFIG_SFC_MTD
1135 	int (*mtd_probe)(struct efx_nic *efx);
1136 	void (*mtd_rename)(struct efx_mtd_partition *part);
1137 	int (*mtd_read)(struct mtd_info *mtd, loff_t start, size_t len,
1138 			size_t *retlen, u8 *buffer);
1139 	int (*mtd_erase)(struct mtd_info *mtd, loff_t start, size_t len);
1140 	int (*mtd_write)(struct mtd_info *mtd, loff_t start, size_t len,
1141 			 size_t *retlen, const u8 *buffer);
1142 	int (*mtd_sync)(struct mtd_info *mtd);
1143 #endif
1144 	void (*ptp_write_host_time)(struct efx_nic *efx, u32 host_time);
1145 
1146 	int revision;
1147 	unsigned int txd_ptr_tbl_base;
1148 	unsigned int rxd_ptr_tbl_base;
1149 	unsigned int buf_tbl_base;
1150 	unsigned int evq_ptr_tbl_base;
1151 	unsigned int evq_rptr_tbl_base;
1152 	u64 max_dma_mask;
1153 	unsigned int rx_prefix_size;
1154 	unsigned int rx_hash_offset;
1155 	unsigned int rx_buffer_padding;
1156 	bool can_rx_scatter;
1157 	bool always_rx_scatter;
1158 	unsigned int max_interrupt_mode;
1159 	unsigned int timer_period_max;
1160 	netdev_features_t offload_features;
1161 	int mcdi_max_ver;
1162 	unsigned int max_rx_ip_filters;
1163 };
1164 
1165 /**************************************************************************
1166  *
1167  * Prototypes and inline functions
1168  *
1169  *************************************************************************/
1170 
1171 static inline struct efx_channel *
1172 efx_get_channel(struct efx_nic *efx, unsigned index)
1173 {
1174 	EFX_BUG_ON_PARANOID(index >= efx->n_channels);
1175 	return efx->channel[index];
1176 }
1177 
1178 /* Iterate over all used channels */
1179 #define efx_for_each_channel(_channel, _efx)				\
1180 	for (_channel = (_efx)->channel[0];				\
1181 	     _channel;							\
1182 	     _channel = (_channel->channel + 1 < (_efx)->n_channels) ?	\
1183 		     (_efx)->channel[_channel->channel + 1] : NULL)
1184 
1185 /* Iterate over all used channels in reverse */
1186 #define efx_for_each_channel_rev(_channel, _efx)			\
1187 	for (_channel = (_efx)->channel[(_efx)->n_channels - 1];	\
1188 	     _channel;							\
1189 	     _channel = _channel->channel ?				\
1190 		     (_efx)->channel[_channel->channel - 1] : NULL)
1191 
1192 static inline struct efx_tx_queue *
1193 efx_get_tx_queue(struct efx_nic *efx, unsigned index, unsigned type)
1194 {
1195 	EFX_BUG_ON_PARANOID(index >= efx->n_tx_channels ||
1196 			    type >= EFX_TXQ_TYPES);
1197 	return &efx->channel[efx->tx_channel_offset + index]->tx_queue[type];
1198 }
1199 
1200 static inline bool efx_channel_has_tx_queues(struct efx_channel *channel)
1201 {
1202 	return channel->channel - channel->efx->tx_channel_offset <
1203 		channel->efx->n_tx_channels;
1204 }
1205 
1206 static inline struct efx_tx_queue *
1207 efx_channel_get_tx_queue(struct efx_channel *channel, unsigned type)
1208 {
1209 	EFX_BUG_ON_PARANOID(!efx_channel_has_tx_queues(channel) ||
1210 			    type >= EFX_TXQ_TYPES);
1211 	return &channel->tx_queue[type];
1212 }
1213 
1214 static inline bool efx_tx_queue_used(struct efx_tx_queue *tx_queue)
1215 {
1216 	return !(tx_queue->efx->net_dev->num_tc < 2 &&
1217 		 tx_queue->queue & EFX_TXQ_TYPE_HIGHPRI);
1218 }
1219 
1220 /* Iterate over all TX queues belonging to a channel */
1221 #define efx_for_each_channel_tx_queue(_tx_queue, _channel)		\
1222 	if (!efx_channel_has_tx_queues(_channel))			\
1223 		;							\
1224 	else								\
1225 		for (_tx_queue = (_channel)->tx_queue;			\
1226 		     _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES && \
1227 			     efx_tx_queue_used(_tx_queue);		\
1228 		     _tx_queue++)
1229 
1230 /* Iterate over all possible TX queues belonging to a channel */
1231 #define efx_for_each_possible_channel_tx_queue(_tx_queue, _channel)	\
1232 	if (!efx_channel_has_tx_queues(_channel))			\
1233 		;							\
1234 	else								\
1235 		for (_tx_queue = (_channel)->tx_queue;			\
1236 		     _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES;	\
1237 		     _tx_queue++)
1238 
1239 static inline bool efx_channel_has_rx_queue(struct efx_channel *channel)
1240 {
1241 	return channel->rx_queue.core_index >= 0;
1242 }
1243 
1244 static inline struct efx_rx_queue *
1245 efx_channel_get_rx_queue(struct efx_channel *channel)
1246 {
1247 	EFX_BUG_ON_PARANOID(!efx_channel_has_rx_queue(channel));
1248 	return &channel->rx_queue;
1249 }
1250 
1251 /* Iterate over all RX queues belonging to a channel */
1252 #define efx_for_each_channel_rx_queue(_rx_queue, _channel)		\
1253 	if (!efx_channel_has_rx_queue(_channel))			\
1254 		;							\
1255 	else								\
1256 		for (_rx_queue = &(_channel)->rx_queue;			\
1257 		     _rx_queue;						\
1258 		     _rx_queue = NULL)
1259 
1260 static inline struct efx_channel *
1261 efx_rx_queue_channel(struct efx_rx_queue *rx_queue)
1262 {
1263 	return container_of(rx_queue, struct efx_channel, rx_queue);
1264 }
1265 
1266 static inline int efx_rx_queue_index(struct efx_rx_queue *rx_queue)
1267 {
1268 	return efx_rx_queue_channel(rx_queue)->channel;
1269 }
1270 
1271 /* Returns a pointer to the specified receive buffer in the RX
1272  * descriptor queue.
1273  */
1274 static inline struct efx_rx_buffer *efx_rx_buffer(struct efx_rx_queue *rx_queue,
1275 						  unsigned int index)
1276 {
1277 	return &rx_queue->buffer[index];
1278 }
1279 
1280 
1281 /**
1282  * EFX_MAX_FRAME_LEN - calculate maximum frame length
1283  *
1284  * This calculates the maximum frame length that will be used for a
1285  * given MTU.  The frame length will be equal to the MTU plus a
1286  * constant amount of header space and padding.  This is the quantity
1287  * that the net driver will program into the MAC as the maximum frame
1288  * length.
1289  *
1290  * The 10G MAC requires 8-byte alignment on the frame
1291  * length, so we round up to the nearest 8.
1292  *
1293  * Re-clocking by the XGXS on RX can reduce an IPG to 32 bits (half an
1294  * XGMII cycle).  If the frame length reaches the maximum value in the
1295  * same cycle, the XMAC can miss the IPG altogether.  We work around
1296  * this by adding a further 16 bytes.
1297  */
1298 #define EFX_MAX_FRAME_LEN(mtu) \
1299 	((((mtu) + ETH_HLEN + VLAN_HLEN + 4/* FCS */ + 7) & ~7) + 16)
1300 
1301 static inline bool efx_xmit_with_hwtstamp(struct sk_buff *skb)
1302 {
1303 	return skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP;
1304 }
1305 static inline void efx_xmit_hwtstamp_pending(struct sk_buff *skb)
1306 {
1307 	skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
1308 }
1309 
1310 #endif /* EFX_NET_DRIVER_H */
1311