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/rwsem.h>
29 #include <linux/vmalloc.h>
30 #include <linux/i2c.h>
31 #include <linux/mtd/mtd.h>
32 #include <net/busy_poll.h>
33 
34 #include "enum.h"
35 #include "bitfield.h"
36 #include "filter.h"
37 
38 /**************************************************************************
39  *
40  * Build definitions
41  *
42  **************************************************************************/
43 
44 #define EFX_DRIVER_VERSION	"4.1"
45 
46 #ifdef DEBUG
47 #define EFX_WARN_ON_ONCE_PARANOID(x) WARN_ON_ONCE(x)
48 #define EFX_WARN_ON_PARANOID(x) WARN_ON(x)
49 #else
50 #define EFX_WARN_ON_ONCE_PARANOID(x) do {} while (0)
51 #define EFX_WARN_ON_PARANOID(x) do {} while (0)
52 #endif
53 
54 /**************************************************************************
55  *
56  * Efx data structures
57  *
58  **************************************************************************/
59 
60 #define EFX_MAX_CHANNELS 32U
61 #define EFX_MAX_RX_QUEUES EFX_MAX_CHANNELS
62 #define EFX_EXTRA_CHANNEL_IOV	0
63 #define EFX_EXTRA_CHANNEL_PTP	1
64 #define EFX_MAX_EXTRA_CHANNELS	2U
65 
66 /* Checksum generation is a per-queue option in hardware, so each
67  * queue visible to the networking core is backed by two hardware TX
68  * queues. */
69 #define EFX_MAX_TX_TC		2
70 #define EFX_MAX_CORE_TX_QUEUES	(EFX_MAX_TX_TC * EFX_MAX_CHANNELS)
71 #define EFX_TXQ_TYPE_OFFLOAD	1	/* flag */
72 #define EFX_TXQ_TYPE_HIGHPRI	2	/* flag */
73 #define EFX_TXQ_TYPES		4
74 #define EFX_MAX_TX_QUEUES	(EFX_TXQ_TYPES * EFX_MAX_CHANNELS)
75 
76 /* Maximum possible MTU the driver supports */
77 #define EFX_MAX_MTU (9 * 1024)
78 
79 /* Minimum MTU, from RFC791 (IP) */
80 #define EFX_MIN_MTU 68
81 
82 /* Size of an RX scatter buffer.  Small enough to pack 2 into a 4K page,
83  * and should be a multiple of the cache line size.
84  */
85 #define EFX_RX_USR_BUF_SIZE	(2048 - 256)
86 
87 /* If possible, we should ensure cache line alignment at start and end
88  * of every buffer.  Otherwise, we just need to ensure 4-byte
89  * alignment of the network header.
90  */
91 #if NET_IP_ALIGN == 0
92 #define EFX_RX_BUF_ALIGNMENT	L1_CACHE_BYTES
93 #else
94 #define EFX_RX_BUF_ALIGNMENT	4
95 #endif
96 
97 /* Forward declare Precision Time Protocol (PTP) support structure. */
98 struct efx_ptp_data;
99 struct hwtstamp_config;
100 
101 struct efx_self_tests;
102 
103 /**
104  * struct efx_buffer - A general-purpose DMA buffer
105  * @addr: host base address of the buffer
106  * @dma_addr: DMA base address of the buffer
107  * @len: Buffer length, in bytes
108  *
109  * The NIC uses these buffers for its interrupt status registers and
110  * MAC stats dumps.
111  */
112 struct efx_buffer {
113 	void *addr;
114 	dma_addr_t dma_addr;
115 	unsigned int len;
116 };
117 
118 /**
119  * struct efx_special_buffer - DMA buffer entered into buffer table
120  * @buf: Standard &struct efx_buffer
121  * @index: Buffer index within controller;s buffer table
122  * @entries: Number of buffer table entries
123  *
124  * The NIC has a buffer table that maps buffers of size %EFX_BUF_SIZE.
125  * Event and descriptor rings are addressed via one or more buffer
126  * table entries (and so can be physically non-contiguous, although we
127  * currently do not take advantage of that).  On Falcon and Siena we
128  * have to take care of allocating and initialising the entries
129  * ourselves.  On later hardware this is managed by the firmware and
130  * @index and @entries are left as 0.
131  */
132 struct efx_special_buffer {
133 	struct efx_buffer buf;
134 	unsigned int index;
135 	unsigned int entries;
136 };
137 
138 /**
139  * struct efx_tx_buffer - buffer state for a TX descriptor
140  * @skb: When @flags & %EFX_TX_BUF_SKB, the associated socket buffer to be
141  *	freed when descriptor completes
142  * @option: When @flags & %EFX_TX_BUF_OPTION, a NIC-specific option descriptor.
143  * @dma_addr: DMA address of the fragment.
144  * @flags: Flags for allocation and DMA mapping type
145  * @len: Length of this fragment.
146  *	This field is zero when the queue slot is empty.
147  * @unmap_len: Length of this fragment to unmap
148  * @dma_offset: Offset of @dma_addr from the address of the backing DMA mapping.
149  * Only valid if @unmap_len != 0.
150  */
151 struct efx_tx_buffer {
152 	const struct sk_buff *skb;
153 	union {
154 		efx_qword_t option;
155 		dma_addr_t dma_addr;
156 	};
157 	unsigned short flags;
158 	unsigned short len;
159 	unsigned short unmap_len;
160 	unsigned short dma_offset;
161 };
162 #define EFX_TX_BUF_CONT		1	/* not last descriptor of packet */
163 #define EFX_TX_BUF_SKB		2	/* buffer is last part of skb */
164 #define EFX_TX_BUF_MAP_SINGLE	8	/* buffer was mapped with dma_map_single() */
165 #define EFX_TX_BUF_OPTION	0x10	/* empty buffer for option descriptor */
166 
167 /**
168  * struct efx_tx_queue - An Efx TX queue
169  *
170  * This is a ring buffer of TX fragments.
171  * Since the TX completion path always executes on the same
172  * CPU and the xmit path can operate on different CPUs,
173  * performance is increased by ensuring that the completion
174  * path and the xmit path operate on different cache lines.
175  * This is particularly important if the xmit path is always
176  * executing on one CPU which is different from the completion
177  * path.  There is also a cache line for members which are
178  * read but not written on the fast path.
179  *
180  * @efx: The associated Efx NIC
181  * @queue: DMA queue number
182  * @tso_version: Version of TSO in use for this queue.
183  * @channel: The associated channel
184  * @core_txq: The networking core TX queue structure
185  * @buffer: The software buffer ring
186  * @cb_page: Array of pages of copy buffers.  Carved up according to
187  *	%EFX_TX_CB_ORDER into %EFX_TX_CB_SIZE-sized chunks.
188  * @txd: The hardware descriptor ring
189  * @ptr_mask: The size of the ring minus 1.
190  * @piobuf: PIO buffer region for this TX queue (shared with its partner).
191  *	Size of the region is efx_piobuf_size.
192  * @piobuf_offset: Buffer offset to be specified in PIO descriptors
193  * @initialised: Has hardware queue been initialised?
194  * @handle_tso: TSO xmit preparation handler.  Sets up the TSO metadata and
195  *	may also map tx data, depending on the nature of the TSO implementation.
196  * @read_count: Current read pointer.
197  *	This is the number of buffers that have been removed from both rings.
198  * @old_write_count: The value of @write_count when last checked.
199  *	This is here for performance reasons.  The xmit path will
200  *	only get the up-to-date value of @write_count if this
201  *	variable indicates that the queue is empty.  This is to
202  *	avoid cache-line ping-pong between the xmit path and the
203  *	completion path.
204  * @merge_events: Number of TX merged completion events
205  * @insert_count: Current insert pointer
206  *	This is the number of buffers that have been added to the
207  *	software ring.
208  * @write_count: Current write pointer
209  *	This is the number of buffers that have been added to the
210  *	hardware ring.
211  * @packet_write_count: Completable write pointer
212  *	This is the write pointer of the last packet written.
213  *	Normally this will equal @write_count, but as option descriptors
214  *	don't produce completion events, they won't update this.
215  *	Filled in iff @efx->type->option_descriptors; only used for PIO.
216  *	Thus, this is written and used on EF10, and neither on farch.
217  * @old_read_count: The value of read_count when last checked.
218  *	This is here for performance reasons.  The xmit path will
219  *	only get the up-to-date value of read_count if this
220  *	variable indicates that the queue is full.  This is to
221  *	avoid cache-line ping-pong between the xmit path and the
222  *	completion path.
223  * @tso_bursts: Number of times TSO xmit invoked by kernel
224  * @tso_long_headers: Number of packets with headers too long for standard
225  *	blocks
226  * @tso_packets: Number of packets via the TSO xmit path
227  * @tso_fallbacks: Number of times TSO fallback used
228  * @pushes: Number of times the TX push feature has been used
229  * @pio_packets: Number of times the TX PIO feature has been used
230  * @xmit_more_available: Are any packets waiting to be pushed to the NIC
231  * @cb_packets: Number of times the TX copybreak feature has been used
232  * @empty_read_count: If the completion path has seen the queue as empty
233  *	and the transmission path has not yet checked this, the value of
234  *	@read_count bitwise-added to %EFX_EMPTY_COUNT_VALID; otherwise 0.
235  */
236 struct efx_tx_queue {
237 	/* Members which don't change on the fast path */
238 	struct efx_nic *efx ____cacheline_aligned_in_smp;
239 	unsigned queue;
240 	unsigned int tso_version;
241 	struct efx_channel *channel;
242 	struct netdev_queue *core_txq;
243 	struct efx_tx_buffer *buffer;
244 	struct efx_buffer *cb_page;
245 	struct efx_special_buffer txd;
246 	unsigned int ptr_mask;
247 	void __iomem *piobuf;
248 	unsigned int piobuf_offset;
249 	bool initialised;
250 
251 	/* Function pointers used in the fast path. */
252 	int (*handle_tso)(struct efx_tx_queue*, struct sk_buff*, bool *);
253 
254 	/* Members used mainly on the completion path */
255 	unsigned int read_count ____cacheline_aligned_in_smp;
256 	unsigned int old_write_count;
257 	unsigned int merge_events;
258 	unsigned int bytes_compl;
259 	unsigned int pkts_compl;
260 
261 	/* Members used only on the xmit path */
262 	unsigned int insert_count ____cacheline_aligned_in_smp;
263 	unsigned int write_count;
264 	unsigned int packet_write_count;
265 	unsigned int old_read_count;
266 	unsigned int tso_bursts;
267 	unsigned int tso_long_headers;
268 	unsigned int tso_packets;
269 	unsigned int tso_fallbacks;
270 	unsigned int pushes;
271 	unsigned int pio_packets;
272 	bool xmit_more_available;
273 	unsigned int cb_packets;
274 	/* Statistics to supplement MAC stats */
275 	unsigned long tx_packets;
276 
277 	/* Members shared between paths and sometimes updated */
278 	unsigned int empty_read_count ____cacheline_aligned_in_smp;
279 #define EFX_EMPTY_COUNT_VALID 0x80000000
280 	atomic_t flush_outstanding;
281 };
282 
283 #define EFX_TX_CB_ORDER	7
284 #define EFX_TX_CB_SIZE	(1 << EFX_TX_CB_ORDER) - NET_IP_ALIGN
285 
286 /**
287  * struct efx_rx_buffer - An Efx RX data buffer
288  * @dma_addr: DMA base address of the buffer
289  * @page: The associated page buffer.
290  *	Will be %NULL if the buffer slot is currently free.
291  * @page_offset: If pending: offset in @page of DMA base address.
292  *	If completed: offset in @page of Ethernet header.
293  * @len: If pending: length for DMA descriptor.
294  *	If completed: received length, excluding hash prefix.
295  * @flags: Flags for buffer and packet state.  These are only set on the
296  *	first buffer of a scattered packet.
297  */
298 struct efx_rx_buffer {
299 	dma_addr_t dma_addr;
300 	struct page *page;
301 	u16 page_offset;
302 	u16 len;
303 	u16 flags;
304 };
305 #define EFX_RX_BUF_LAST_IN_PAGE	0x0001
306 #define EFX_RX_PKT_CSUMMED	0x0002
307 #define EFX_RX_PKT_DISCARD	0x0004
308 #define EFX_RX_PKT_TCP		0x0040
309 #define EFX_RX_PKT_PREFIX_LEN	0x0080	/* length is in prefix only */
310 #define EFX_RX_PKT_CSUM_LEVEL	0x0200
311 
312 /**
313  * struct efx_rx_page_state - Page-based rx buffer state
314  *
315  * Inserted at the start of every page allocated for receive buffers.
316  * Used to facilitate sharing dma mappings between recycled rx buffers
317  * and those passed up to the kernel.
318  *
319  * @dma_addr: The dma address of this page.
320  */
321 struct efx_rx_page_state {
322 	dma_addr_t dma_addr;
323 
324 	unsigned int __pad[0] ____cacheline_aligned;
325 };
326 
327 /**
328  * struct efx_rx_queue - An Efx RX queue
329  * @efx: The associated Efx NIC
330  * @core_index:  Index of network core RX queue.  Will be >= 0 iff this
331  *	is associated with a real RX queue.
332  * @buffer: The software buffer ring
333  * @rxd: The hardware descriptor ring
334  * @ptr_mask: The size of the ring minus 1.
335  * @refill_enabled: Enable refill whenever fill level is low
336  * @flush_pending: Set when a RX flush is pending. Has the same lifetime as
337  *	@rxq_flush_pending.
338  * @added_count: Number of buffers added to the receive queue.
339  * @notified_count: Number of buffers given to NIC (<= @added_count).
340  * @removed_count: Number of buffers removed from the receive queue.
341  * @scatter_n: Used by NIC specific receive code.
342  * @scatter_len: Used by NIC specific receive code.
343  * @page_ring: The ring to store DMA mapped pages for reuse.
344  * @page_add: Counter to calculate the write pointer for the recycle ring.
345  * @page_remove: Counter to calculate the read pointer for the recycle ring.
346  * @page_recycle_count: The number of pages that have been recycled.
347  * @page_recycle_failed: The number of pages that couldn't be recycled because
348  *      the kernel still held a reference to them.
349  * @page_recycle_full: The number of pages that were released because the
350  *      recycle ring was full.
351  * @page_ptr_mask: The number of pages in the RX recycle ring minus 1.
352  * @max_fill: RX descriptor maximum fill level (<= ring size)
353  * @fast_fill_trigger: RX descriptor fill level that will trigger a fast fill
354  *	(<= @max_fill)
355  * @min_fill: RX descriptor minimum non-zero fill level.
356  *	This records the minimum fill level observed when a ring
357  *	refill was triggered.
358  * @recycle_count: RX buffer recycle counter.
359  * @slow_fill: Timer used to defer efx_nic_generate_fill_event().
360  */
361 struct efx_rx_queue {
362 	struct efx_nic *efx;
363 	int core_index;
364 	struct efx_rx_buffer *buffer;
365 	struct efx_special_buffer rxd;
366 	unsigned int ptr_mask;
367 	bool refill_enabled;
368 	bool flush_pending;
369 
370 	unsigned int added_count;
371 	unsigned int notified_count;
372 	unsigned int removed_count;
373 	unsigned int scatter_n;
374 	unsigned int scatter_len;
375 	struct page **page_ring;
376 	unsigned int page_add;
377 	unsigned int page_remove;
378 	unsigned int page_recycle_count;
379 	unsigned int page_recycle_failed;
380 	unsigned int page_recycle_full;
381 	unsigned int page_ptr_mask;
382 	unsigned int max_fill;
383 	unsigned int fast_fill_trigger;
384 	unsigned int min_fill;
385 	unsigned int min_overfill;
386 	unsigned int recycle_count;
387 	struct timer_list slow_fill;
388 	unsigned int slow_fill_count;
389 	/* Statistics to supplement MAC stats */
390 	unsigned long rx_packets;
391 };
392 
393 enum efx_sync_events_state {
394 	SYNC_EVENTS_DISABLED = 0,
395 	SYNC_EVENTS_QUIESCENT,
396 	SYNC_EVENTS_REQUESTED,
397 	SYNC_EVENTS_VALID,
398 };
399 
400 /**
401  * struct efx_channel - An Efx channel
402  *
403  * A channel comprises an event queue, at least one TX queue, at least
404  * one RX queue, and an associated tasklet for processing the event
405  * queue.
406  *
407  * @efx: Associated Efx NIC
408  * @channel: Channel instance number
409  * @type: Channel type definition
410  * @eventq_init: Event queue initialised flag
411  * @enabled: Channel enabled indicator
412  * @irq: IRQ number (MSI and MSI-X only)
413  * @irq_moderation_us: IRQ moderation value (in microseconds)
414  * @napi_dev: Net device used with NAPI
415  * @napi_str: NAPI control structure
416  * @state: state for NAPI vs busy polling
417  * @state_lock: lock protecting @state
418  * @eventq: Event queue buffer
419  * @eventq_mask: Event queue pointer mask
420  * @eventq_read_ptr: Event queue read pointer
421  * @event_test_cpu: Last CPU to handle interrupt or test event for this channel
422  * @irq_count: Number of IRQs since last adaptive moderation decision
423  * @irq_mod_score: IRQ moderation score
424  * @rps_flow_id: Flow IDs of filters allocated for accelerated RFS,
425  *      indexed by filter ID
426  * @n_rx_tobe_disc: Count of RX_TOBE_DISC errors
427  * @n_rx_ip_hdr_chksum_err: Count of RX IP header checksum errors
428  * @n_rx_tcp_udp_chksum_err: Count of RX TCP and UDP checksum errors
429  * @n_rx_mcast_mismatch: Count of unmatched multicast frames
430  * @n_rx_frm_trunc: Count of RX_FRM_TRUNC errors
431  * @n_rx_overlength: Count of RX_OVERLENGTH errors
432  * @n_skbuff_leaks: Count of skbuffs leaked due to RX overrun
433  * @n_rx_nodesc_trunc: Number of RX packets truncated and then dropped due to
434  *	lack of descriptors
435  * @n_rx_merge_events: Number of RX merged completion events
436  * @n_rx_merge_packets: Number of RX packets completed by merged events
437  * @rx_pkt_n_frags: Number of fragments in next packet to be delivered by
438  *	__efx_rx_packet(), or zero if there is none
439  * @rx_pkt_index: Ring index of first buffer for next packet to be delivered
440  *	by __efx_rx_packet(), if @rx_pkt_n_frags != 0
441  * @rx_queue: RX queue for this channel
442  * @tx_queue: TX queues for this channel
443  * @sync_events_state: Current state of sync events on this channel
444  * @sync_timestamp_major: Major part of the last ptp sync event
445  * @sync_timestamp_minor: Minor part of the last ptp sync event
446  */
447 struct efx_channel {
448 	struct efx_nic *efx;
449 	int channel;
450 	const struct efx_channel_type *type;
451 	bool eventq_init;
452 	bool enabled;
453 	int irq;
454 	unsigned int irq_moderation_us;
455 	struct net_device *napi_dev;
456 	struct napi_struct napi_str;
457 #ifdef CONFIG_NET_RX_BUSY_POLL
458 	unsigned long busy_poll_state;
459 #endif
460 	struct efx_special_buffer eventq;
461 	unsigned int eventq_mask;
462 	unsigned int eventq_read_ptr;
463 	int event_test_cpu;
464 
465 	unsigned int irq_count;
466 	unsigned int irq_mod_score;
467 #ifdef CONFIG_RFS_ACCEL
468 	unsigned int rfs_filters_added;
469 #define RPS_FLOW_ID_INVALID 0xFFFFFFFF
470 	u32 *rps_flow_id;
471 #endif
472 
473 	unsigned int n_rx_tobe_disc;
474 	unsigned int n_rx_ip_hdr_chksum_err;
475 	unsigned int n_rx_tcp_udp_chksum_err;
476 	unsigned int n_rx_outer_ip_hdr_chksum_err;
477 	unsigned int n_rx_outer_tcp_udp_chksum_err;
478 	unsigned int n_rx_inner_ip_hdr_chksum_err;
479 	unsigned int n_rx_inner_tcp_udp_chksum_err;
480 	unsigned int n_rx_eth_crc_err;
481 	unsigned int n_rx_mcast_mismatch;
482 	unsigned int n_rx_frm_trunc;
483 	unsigned int n_rx_overlength;
484 	unsigned int n_skbuff_leaks;
485 	unsigned int n_rx_nodesc_trunc;
486 	unsigned int n_rx_merge_events;
487 	unsigned int n_rx_merge_packets;
488 
489 	unsigned int rx_pkt_n_frags;
490 	unsigned int rx_pkt_index;
491 
492 	struct efx_rx_queue rx_queue;
493 	struct efx_tx_queue tx_queue[EFX_TXQ_TYPES];
494 
495 	enum efx_sync_events_state sync_events_state;
496 	u32 sync_timestamp_major;
497 	u32 sync_timestamp_minor;
498 };
499 
500 /**
501  * struct efx_msi_context - Context for each MSI
502  * @efx: The associated NIC
503  * @index: Index of the channel/IRQ
504  * @name: Name of the channel/IRQ
505  *
506  * Unlike &struct efx_channel, this is never reallocated and is always
507  * safe for the IRQ handler to access.
508  */
509 struct efx_msi_context {
510 	struct efx_nic *efx;
511 	unsigned int index;
512 	char name[IFNAMSIZ + 6];
513 };
514 
515 /**
516  * struct efx_channel_type - distinguishes traffic and extra channels
517  * @handle_no_channel: Handle failure to allocate an extra channel
518  * @pre_probe: Set up extra state prior to initialisation
519  * @post_remove: Tear down extra state after finalisation, if allocated.
520  *	May be called on channels that have not been probed.
521  * @get_name: Generate the channel's name (used for its IRQ handler)
522  * @copy: Copy the channel state prior to reallocation.  May be %NULL if
523  *	reallocation is not supported.
524  * @receive_skb: Handle an skb ready to be passed to netif_receive_skb()
525  * @keep_eventq: Flag for whether event queue should be kept initialised
526  *	while the device is stopped
527  */
528 struct efx_channel_type {
529 	void (*handle_no_channel)(struct efx_nic *);
530 	int (*pre_probe)(struct efx_channel *);
531 	void (*post_remove)(struct efx_channel *);
532 	void (*get_name)(struct efx_channel *, char *buf, size_t len);
533 	struct efx_channel *(*copy)(const struct efx_channel *);
534 	bool (*receive_skb)(struct efx_channel *, struct sk_buff *);
535 	bool keep_eventq;
536 };
537 
538 enum efx_led_mode {
539 	EFX_LED_OFF	= 0,
540 	EFX_LED_ON	= 1,
541 	EFX_LED_DEFAULT	= 2
542 };
543 
544 #define STRING_TABLE_LOOKUP(val, member) \
545 	((val) < member ## _max) ? member ## _names[val] : "(invalid)"
546 
547 extern const char *const efx_loopback_mode_names[];
548 extern const unsigned int efx_loopback_mode_max;
549 #define LOOPBACK_MODE(efx) \
550 	STRING_TABLE_LOOKUP((efx)->loopback_mode, efx_loopback_mode)
551 
552 extern const char *const efx_reset_type_names[];
553 extern const unsigned int efx_reset_type_max;
554 #define RESET_TYPE(type) \
555 	STRING_TABLE_LOOKUP(type, efx_reset_type)
556 
557 void efx_get_udp_tunnel_type_name(u16 type, char *buf, size_t buflen);
558 
559 enum efx_int_mode {
560 	/* Be careful if altering to correct macro below */
561 	EFX_INT_MODE_MSIX = 0,
562 	EFX_INT_MODE_MSI = 1,
563 	EFX_INT_MODE_LEGACY = 2,
564 	EFX_INT_MODE_MAX	/* Insert any new items before this */
565 };
566 #define EFX_INT_MODE_USE_MSI(x) (((x)->interrupt_mode) <= EFX_INT_MODE_MSI)
567 
568 enum nic_state {
569 	STATE_UNINIT = 0,	/* device being probed/removed or is frozen */
570 	STATE_READY = 1,	/* hardware ready and netdev registered */
571 	STATE_DISABLED = 2,	/* device disabled due to hardware errors */
572 	STATE_RECOVERY = 3,	/* device recovering from PCI error */
573 };
574 
575 /* Forward declaration */
576 struct efx_nic;
577 
578 /* Pseudo bit-mask flow control field */
579 #define EFX_FC_RX	FLOW_CTRL_RX
580 #define EFX_FC_TX	FLOW_CTRL_TX
581 #define EFX_FC_AUTO	4
582 
583 /**
584  * struct efx_link_state - Current state of the link
585  * @up: Link is up
586  * @fd: Link is full-duplex
587  * @fc: Actual flow control flags
588  * @speed: Link speed (Mbps)
589  */
590 struct efx_link_state {
591 	bool up;
592 	bool fd;
593 	u8 fc;
594 	unsigned int speed;
595 };
596 
597 static inline bool efx_link_state_equal(const struct efx_link_state *left,
598 					const struct efx_link_state *right)
599 {
600 	return left->up == right->up && left->fd == right->fd &&
601 		left->fc == right->fc && left->speed == right->speed;
602 }
603 
604 /**
605  * struct efx_phy_operations - Efx PHY operations table
606  * @probe: Probe PHY and initialise efx->mdio.mode_support, efx->mdio.mmds,
607  *	efx->loopback_modes.
608  * @init: Initialise PHY
609  * @fini: Shut down PHY
610  * @reconfigure: Reconfigure PHY (e.g. for new link parameters)
611  * @poll: Update @link_state and report whether it changed.
612  *	Serialised by the mac_lock.
613  * @get_link_ksettings: Get ethtool settings. Serialised by the mac_lock.
614  * @set_link_ksettings: Set ethtool settings. Serialised by the mac_lock.
615  * @set_npage_adv: Set abilities advertised in (Extended) Next Page
616  *	(only needed where AN bit is set in mmds)
617  * @test_alive: Test that PHY is 'alive' (online)
618  * @test_name: Get the name of a PHY-specific test/result
619  * @run_tests: Run tests and record results as appropriate (offline).
620  *	Flags are the ethtool tests flags.
621  */
622 struct efx_phy_operations {
623 	int (*probe) (struct efx_nic *efx);
624 	int (*init) (struct efx_nic *efx);
625 	void (*fini) (struct efx_nic *efx);
626 	void (*remove) (struct efx_nic *efx);
627 	int (*reconfigure) (struct efx_nic *efx);
628 	bool (*poll) (struct efx_nic *efx);
629 	void (*get_link_ksettings)(struct efx_nic *efx,
630 				   struct ethtool_link_ksettings *cmd);
631 	int (*set_link_ksettings)(struct efx_nic *efx,
632 				  const struct ethtool_link_ksettings *cmd);
633 	void (*set_npage_adv) (struct efx_nic *efx, u32);
634 	int (*test_alive) (struct efx_nic *efx);
635 	const char *(*test_name) (struct efx_nic *efx, unsigned int index);
636 	int (*run_tests) (struct efx_nic *efx, int *results, unsigned flags);
637 	int (*get_module_eeprom) (struct efx_nic *efx,
638 			       struct ethtool_eeprom *ee,
639 			       u8 *data);
640 	int (*get_module_info) (struct efx_nic *efx,
641 				struct ethtool_modinfo *modinfo);
642 };
643 
644 /**
645  * enum efx_phy_mode - PHY operating mode flags
646  * @PHY_MODE_NORMAL: on and should pass traffic
647  * @PHY_MODE_TX_DISABLED: on with TX disabled
648  * @PHY_MODE_LOW_POWER: set to low power through MDIO
649  * @PHY_MODE_OFF: switched off through external control
650  * @PHY_MODE_SPECIAL: on but will not pass traffic
651  */
652 enum efx_phy_mode {
653 	PHY_MODE_NORMAL		= 0,
654 	PHY_MODE_TX_DISABLED	= 1,
655 	PHY_MODE_LOW_POWER	= 2,
656 	PHY_MODE_OFF		= 4,
657 	PHY_MODE_SPECIAL	= 8,
658 };
659 
660 static inline bool efx_phy_mode_disabled(enum efx_phy_mode mode)
661 {
662 	return !!(mode & ~PHY_MODE_TX_DISABLED);
663 }
664 
665 /**
666  * struct efx_hw_stat_desc - Description of a hardware statistic
667  * @name: Name of the statistic as visible through ethtool, or %NULL if
668  *	it should not be exposed
669  * @dma_width: Width in bits (0 for non-DMA statistics)
670  * @offset: Offset within stats (ignored for non-DMA statistics)
671  */
672 struct efx_hw_stat_desc {
673 	const char *name;
674 	u16 dma_width;
675 	u16 offset;
676 };
677 
678 /* Number of bits used in a multicast filter hash address */
679 #define EFX_MCAST_HASH_BITS 8
680 
681 /* Number of (single-bit) entries in a multicast filter hash */
682 #define EFX_MCAST_HASH_ENTRIES (1 << EFX_MCAST_HASH_BITS)
683 
684 /* An Efx multicast filter hash */
685 union efx_multicast_hash {
686 	u8 byte[EFX_MCAST_HASH_ENTRIES / 8];
687 	efx_oword_t oword[EFX_MCAST_HASH_ENTRIES / sizeof(efx_oword_t) / 8];
688 };
689 
690 struct vfdi_status;
691 
692 /**
693  * struct efx_nic - an Efx NIC
694  * @name: Device name (net device name or bus id before net device registered)
695  * @pci_dev: The PCI device
696  * @node: List node for maintaning primary/secondary function lists
697  * @primary: &struct efx_nic instance for the primary function of this
698  *	controller.  May be the same structure, and may be %NULL if no
699  *	primary function is bound.  Serialised by rtnl_lock.
700  * @secondary_list: List of &struct efx_nic instances for the secondary PCI
701  *	functions of the controller, if this is for the primary function.
702  *	Serialised by rtnl_lock.
703  * @type: Controller type attributes
704  * @legacy_irq: IRQ number
705  * @workqueue: Workqueue for port reconfigures and the HW monitor.
706  *	Work items do not hold and must not acquire RTNL.
707  * @workqueue_name: Name of workqueue
708  * @reset_work: Scheduled reset workitem
709  * @membase_phys: Memory BAR value as physical address
710  * @membase: Memory BAR value
711  * @interrupt_mode: Interrupt mode
712  * @timer_quantum_ns: Interrupt timer quantum, in nanoseconds
713  * @timer_max_ns: Interrupt timer maximum value, in nanoseconds
714  * @irq_rx_adaptive: Adaptive IRQ moderation enabled for RX event queues
715  * @irq_rx_mod_step_us: Step size for IRQ moderation for RX event queues
716  * @irq_rx_moderation_us: IRQ moderation time for RX event queues
717  * @msg_enable: Log message enable flags
718  * @state: Device state number (%STATE_*). Serialised by the rtnl_lock.
719  * @reset_pending: Bitmask for pending resets
720  * @tx_queue: TX DMA queues
721  * @rx_queue: RX DMA queues
722  * @channel: Channels
723  * @msi_context: Context for each MSI
724  * @extra_channel_types: Types of extra (non-traffic) channels that
725  *	should be allocated for this NIC
726  * @rxq_entries: Size of receive queues requested by user.
727  * @txq_entries: Size of transmit queues requested by user.
728  * @txq_stop_thresh: TX queue fill level at or above which we stop it.
729  * @txq_wake_thresh: TX queue fill level at or below which we wake it.
730  * @tx_dc_base: Base qword address in SRAM of TX queue descriptor caches
731  * @rx_dc_base: Base qword address in SRAM of RX queue descriptor caches
732  * @sram_lim_qw: Qword address limit of SRAM
733  * @next_buffer_table: First available buffer table id
734  * @n_channels: Number of channels in use
735  * @n_rx_channels: Number of channels used for RX (= number of RX queues)
736  * @n_tx_channels: Number of channels used for TX
737  * @rx_ip_align: RX DMA address offset to have IP header aligned in
738  *	in accordance with NET_IP_ALIGN
739  * @rx_dma_len: Current maximum RX DMA length
740  * @rx_buffer_order: Order (log2) of number of pages for each RX buffer
741  * @rx_buffer_truesize: Amortised allocation size of an RX buffer,
742  *	for use in sk_buff::truesize
743  * @rx_prefix_size: Size of RX prefix before packet data
744  * @rx_packet_hash_offset: Offset of RX flow hash from start of packet data
745  *	(valid only if @rx_prefix_size != 0; always negative)
746  * @rx_packet_len_offset: Offset of RX packet length from start of packet data
747  *	(valid only for NICs that set %EFX_RX_PKT_PREFIX_LEN; always negative)
748  * @rx_packet_ts_offset: Offset of timestamp from start of packet data
749  *	(valid only if channel->sync_timestamps_enabled; always negative)
750  * @rx_hash_key: Toeplitz hash key for RSS
751  * @rx_indir_table: Indirection table for RSS
752  * @rx_scatter: Scatter mode enabled for receives
753  * @rss_active: RSS enabled on hardware
754  * @rx_hash_udp_4tuple: UDP 4-tuple hashing enabled
755  * @int_error_count: Number of internal errors seen recently
756  * @int_error_expire: Time at which error count will be expired
757  * @irq_soft_enabled: Are IRQs soft-enabled? If not, IRQ handler will
758  *	acknowledge but do nothing else.
759  * @irq_status: Interrupt status buffer
760  * @irq_zero_count: Number of legacy IRQs seen with queue flags == 0
761  * @irq_level: IRQ level/index for IRQs not triggered by an event queue
762  * @selftest_work: Work item for asynchronous self-test
763  * @mtd_list: List of MTDs attached to the NIC
764  * @nic_data: Hardware dependent state
765  * @mcdi: Management-Controller-to-Driver Interface state
766  * @mac_lock: MAC access lock. Protects @port_enabled, @phy_mode,
767  *	efx_monitor() and efx_reconfigure_port()
768  * @port_enabled: Port enabled indicator.
769  *	Serialises efx_stop_all(), efx_start_all(), efx_monitor() and
770  *	efx_mac_work() with kernel interfaces. Safe to read under any
771  *	one of the rtnl_lock, mac_lock, or netif_tx_lock, but all three must
772  *	be held to modify it.
773  * @port_initialized: Port initialized?
774  * @net_dev: Operating system network device. Consider holding the rtnl lock
775  * @fixed_features: Features which cannot be turned off
776  * @stats_buffer: DMA buffer for statistics
777  * @phy_type: PHY type
778  * @phy_op: PHY interface
779  * @phy_data: PHY private data (including PHY-specific stats)
780  * @mdio: PHY MDIO interface
781  * @mdio_bus: PHY MDIO bus ID (only used by Siena)
782  * @phy_mode: PHY operating mode. Serialised by @mac_lock.
783  * @link_advertising: Autonegotiation advertising flags
784  * @link_state: Current state of the link
785  * @n_link_state_changes: Number of times the link has changed state
786  * @unicast_filter: Flag for Falcon-arch simple unicast filter.
787  *	Protected by @mac_lock.
788  * @multicast_hash: Multicast hash table for Falcon-arch.
789  *	Protected by @mac_lock.
790  * @wanted_fc: Wanted flow control flags
791  * @fc_disable: When non-zero flow control is disabled. Typically used to
792  *	ensure that network back pressure doesn't delay dma queue flushes.
793  *	Serialised by the rtnl lock.
794  * @mac_work: Work item for changing MAC promiscuity and multicast hash
795  * @loopback_mode: Loopback status
796  * @loopback_modes: Supported loopback mode bitmask
797  * @loopback_selftest: Offline self-test private state
798  * @filter_sem: Filter table rw_semaphore, for freeing the table
799  * @filter_lock: Filter table lock, for mere content changes
800  * @filter_state: Architecture-dependent filter table state
801  * @rps_expire_channel: Next channel to check for expiry
802  * @rps_expire_index: Next index to check for expiry in
803  *	@rps_expire_channel's @rps_flow_id
804  * @active_queues: Count of RX and TX queues that haven't been flushed and drained.
805  * @rxq_flush_pending: Count of number of receive queues that need to be flushed.
806  *	Decremented when the efx_flush_rx_queue() is called.
807  * @rxq_flush_outstanding: Count of number of RX flushes started but not yet
808  *	completed (either success or failure). Not used when MCDI is used to
809  *	flush receive queues.
810  * @flush_wq: wait queue used by efx_nic_flush_queues() to wait for flush completions.
811  * @vf_count: Number of VFs intended to be enabled.
812  * @vf_init_count: Number of VFs that have been fully initialised.
813  * @vi_scale: log2 number of vnics per VF.
814  * @ptp_data: PTP state data
815  * @vpd_sn: Serial number read from VPD
816  * @monitor_work: Hardware monitor workitem
817  * @biu_lock: BIU (bus interface unit) lock
818  * @last_irq_cpu: Last CPU to handle a possible test interrupt.  This
819  *	field is used by efx_test_interrupts() to verify that an
820  *	interrupt has occurred.
821  * @stats_lock: Statistics update lock. Must be held when calling
822  *	efx_nic_type::{update,start,stop}_stats.
823  * @n_rx_noskb_drops: Count of RX packets dropped due to failure to allocate an skb
824  *
825  * This is stored in the private area of the &struct net_device.
826  */
827 struct efx_nic {
828 	/* The following fields should be written very rarely */
829 
830 	char name[IFNAMSIZ];
831 	struct list_head node;
832 	struct efx_nic *primary;
833 	struct list_head secondary_list;
834 	struct pci_dev *pci_dev;
835 	unsigned int port_num;
836 	const struct efx_nic_type *type;
837 	int legacy_irq;
838 	bool eeh_disabled_legacy_irq;
839 	struct workqueue_struct *workqueue;
840 	char workqueue_name[16];
841 	struct work_struct reset_work;
842 	resource_size_t membase_phys;
843 	void __iomem *membase;
844 
845 	enum efx_int_mode interrupt_mode;
846 	unsigned int timer_quantum_ns;
847 	unsigned int timer_max_ns;
848 	bool irq_rx_adaptive;
849 	unsigned int irq_mod_step_us;
850 	unsigned int irq_rx_moderation_us;
851 	u32 msg_enable;
852 
853 	enum nic_state state;
854 	unsigned long reset_pending;
855 
856 	struct efx_channel *channel[EFX_MAX_CHANNELS];
857 	struct efx_msi_context msi_context[EFX_MAX_CHANNELS];
858 	const struct efx_channel_type *
859 	extra_channel_type[EFX_MAX_EXTRA_CHANNELS];
860 
861 	unsigned rxq_entries;
862 	unsigned txq_entries;
863 	unsigned int txq_stop_thresh;
864 	unsigned int txq_wake_thresh;
865 
866 	unsigned tx_dc_base;
867 	unsigned rx_dc_base;
868 	unsigned sram_lim_qw;
869 	unsigned next_buffer_table;
870 
871 	unsigned int max_channels;
872 	unsigned int max_tx_channels;
873 	unsigned n_channels;
874 	unsigned n_rx_channels;
875 	unsigned rss_spread;
876 	unsigned tx_channel_offset;
877 	unsigned n_tx_channels;
878 	unsigned int rx_ip_align;
879 	unsigned int rx_dma_len;
880 	unsigned int rx_buffer_order;
881 	unsigned int rx_buffer_truesize;
882 	unsigned int rx_page_buf_step;
883 	unsigned int rx_bufs_per_page;
884 	unsigned int rx_pages_per_batch;
885 	unsigned int rx_prefix_size;
886 	int rx_packet_hash_offset;
887 	int rx_packet_len_offset;
888 	int rx_packet_ts_offset;
889 	u8 rx_hash_key[40];
890 	u32 rx_indir_table[128];
891 	bool rx_scatter;
892 	bool rss_active;
893 	bool rx_hash_udp_4tuple;
894 
895 	unsigned int_error_count;
896 	unsigned long int_error_expire;
897 
898 	bool irq_soft_enabled;
899 	struct efx_buffer irq_status;
900 	unsigned irq_zero_count;
901 	unsigned irq_level;
902 	struct delayed_work selftest_work;
903 
904 #ifdef CONFIG_SFC_MTD
905 	struct list_head mtd_list;
906 #endif
907 
908 	void *nic_data;
909 	struct efx_mcdi_data *mcdi;
910 
911 	struct mutex mac_lock;
912 	struct work_struct mac_work;
913 	bool port_enabled;
914 
915 	bool mc_bist_for_other_fn;
916 	bool port_initialized;
917 	struct net_device *net_dev;
918 
919 	netdev_features_t fixed_features;
920 
921 	struct efx_buffer stats_buffer;
922 	u64 rx_nodesc_drops_total;
923 	u64 rx_nodesc_drops_while_down;
924 	bool rx_nodesc_drops_prev_state;
925 
926 	unsigned int phy_type;
927 	const struct efx_phy_operations *phy_op;
928 	void *phy_data;
929 	struct mdio_if_info mdio;
930 	unsigned int mdio_bus;
931 	enum efx_phy_mode phy_mode;
932 
933 	u32 link_advertising;
934 	struct efx_link_state link_state;
935 	unsigned int n_link_state_changes;
936 
937 	bool unicast_filter;
938 	union efx_multicast_hash multicast_hash;
939 	u8 wanted_fc;
940 	unsigned fc_disable;
941 
942 	atomic_t rx_reset;
943 	enum efx_loopback_mode loopback_mode;
944 	u64 loopback_modes;
945 
946 	void *loopback_selftest;
947 
948 	struct rw_semaphore filter_sem;
949 	spinlock_t filter_lock;
950 	void *filter_state;
951 #ifdef CONFIG_RFS_ACCEL
952 	unsigned int rps_expire_channel;
953 	unsigned int rps_expire_index;
954 #endif
955 
956 	atomic_t active_queues;
957 	atomic_t rxq_flush_pending;
958 	atomic_t rxq_flush_outstanding;
959 	wait_queue_head_t flush_wq;
960 
961 #ifdef CONFIG_SFC_SRIOV
962 	unsigned vf_count;
963 	unsigned vf_init_count;
964 	unsigned vi_scale;
965 #endif
966 
967 	struct efx_ptp_data *ptp_data;
968 
969 	char *vpd_sn;
970 
971 	/* The following fields may be written more often */
972 
973 	struct delayed_work monitor_work ____cacheline_aligned_in_smp;
974 	spinlock_t biu_lock;
975 	int last_irq_cpu;
976 	spinlock_t stats_lock;
977 	atomic_t n_rx_noskb_drops;
978 };
979 
980 static inline int efx_dev_registered(struct efx_nic *efx)
981 {
982 	return efx->net_dev->reg_state == NETREG_REGISTERED;
983 }
984 
985 static inline unsigned int efx_port_num(struct efx_nic *efx)
986 {
987 	return efx->port_num;
988 }
989 
990 struct efx_mtd_partition {
991 	struct list_head node;
992 	struct mtd_info mtd;
993 	const char *dev_type_name;
994 	const char *type_name;
995 	char name[IFNAMSIZ + 20];
996 };
997 
998 struct efx_udp_tunnel {
999 	u16 type; /* TUNNEL_ENCAP_UDP_PORT_ENTRY_foo, see mcdi_pcol.h */
1000 	__be16 port;
1001 	/* Count of repeated adds of the same port.  Used only inside the list,
1002 	 * not in request arguments.
1003 	 */
1004 	u16 count;
1005 };
1006 
1007 /**
1008  * struct efx_nic_type - Efx device type definition
1009  * @mem_bar: Get the memory BAR
1010  * @mem_map_size: Get memory BAR mapped size
1011  * @probe: Probe the controller
1012  * @remove: Free resources allocated by probe()
1013  * @init: Initialise the controller
1014  * @dimension_resources: Dimension controller resources (buffer table,
1015  *	and VIs once the available interrupt resources are clear)
1016  * @fini: Shut down the controller
1017  * @monitor: Periodic function for polling link state and hardware monitor
1018  * @map_reset_reason: Map ethtool reset reason to a reset method
1019  * @map_reset_flags: Map ethtool reset flags to a reset method, if possible
1020  * @reset: Reset the controller hardware and possibly the PHY.  This will
1021  *	be called while the controller is uninitialised.
1022  * @probe_port: Probe the MAC and PHY
1023  * @remove_port: Free resources allocated by probe_port()
1024  * @handle_global_event: Handle a "global" event (may be %NULL)
1025  * @fini_dmaq: Flush and finalise DMA queues (RX and TX queues)
1026  * @prepare_flush: Prepare the hardware for flushing the DMA queues
1027  *	(for Falcon architecture)
1028  * @finish_flush: Clean up after flushing the DMA queues (for Falcon
1029  *	architecture)
1030  * @prepare_flr: Prepare for an FLR
1031  * @finish_flr: Clean up after an FLR
1032  * @describe_stats: Describe statistics for ethtool
1033  * @update_stats: Update statistics not provided by event handling.
1034  *	Either argument may be %NULL.
1035  * @start_stats: Start the regular fetching of statistics
1036  * @pull_stats: Pull stats from the NIC and wait until they arrive.
1037  * @stop_stats: Stop the regular fetching of statistics
1038  * @set_id_led: Set state of identifying LED or revert to automatic function
1039  * @push_irq_moderation: Apply interrupt moderation value
1040  * @reconfigure_port: Push loopback/power/txdis changes to the MAC and PHY
1041  * @prepare_enable_fc_tx: Prepare MAC to enable pause frame TX (may be %NULL)
1042  * @reconfigure_mac: Push MAC address, MTU, flow control and filter settings
1043  *	to the hardware.  Serialised by the mac_lock.
1044  * @check_mac_fault: Check MAC fault state. True if fault present.
1045  * @get_wol: Get WoL configuration from driver state
1046  * @set_wol: Push WoL configuration to the NIC
1047  * @resume_wol: Synchronise WoL state between driver and MC (e.g. after resume)
1048  * @test_chip: Test registers.  May use efx_farch_test_registers(), and is
1049  *	expected to reset the NIC.
1050  * @test_nvram: Test validity of NVRAM contents
1051  * @mcdi_request: Send an MCDI request with the given header and SDU.
1052  *	The SDU length may be any value from 0 up to the protocol-
1053  *	defined maximum, but its buffer will be padded to a multiple
1054  *	of 4 bytes.
1055  * @mcdi_poll_response: Test whether an MCDI response is available.
1056  * @mcdi_read_response: Read the MCDI response PDU.  The offset will
1057  *	be a multiple of 4.  The length may not be, but the buffer
1058  *	will be padded so it is safe to round up.
1059  * @mcdi_poll_reboot: Test whether the MCDI has rebooted.  If so,
1060  *	return an appropriate error code for aborting any current
1061  *	request; otherwise return 0.
1062  * @irq_enable_master: Enable IRQs on the NIC.  Each event queue must
1063  *	be separately enabled after this.
1064  * @irq_test_generate: Generate a test IRQ
1065  * @irq_disable_non_ev: Disable non-event IRQs on the NIC.  Each event
1066  *	queue must be separately disabled before this.
1067  * @irq_handle_msi: Handle MSI for a channel.  The @dev_id argument is
1068  *	a pointer to the &struct efx_msi_context for the channel.
1069  * @irq_handle_legacy: Handle legacy interrupt.  The @dev_id argument
1070  *	is a pointer to the &struct efx_nic.
1071  * @tx_probe: Allocate resources for TX queue
1072  * @tx_init: Initialise TX queue on the NIC
1073  * @tx_remove: Free resources for TX queue
1074  * @tx_write: Write TX descriptors and doorbell
1075  * @rx_push_rss_config: Write RSS hash key and indirection table to the NIC
1076  * @rx_pull_rss_config: Read RSS hash key and indirection table back from the NIC
1077  * @rx_probe: Allocate resources for RX queue
1078  * @rx_init: Initialise RX queue on the NIC
1079  * @rx_remove: Free resources for RX queue
1080  * @rx_write: Write RX descriptors and doorbell
1081  * @rx_defer_refill: Generate a refill reminder event
1082  * @ev_probe: Allocate resources for event queue
1083  * @ev_init: Initialise event queue on the NIC
1084  * @ev_fini: Deinitialise event queue on the NIC
1085  * @ev_remove: Free resources for event queue
1086  * @ev_process: Process events for a queue, up to the given NAPI quota
1087  * @ev_read_ack: Acknowledge read events on a queue, rearming its IRQ
1088  * @ev_test_generate: Generate a test event
1089  * @filter_table_probe: Probe filter capabilities and set up filter software state
1090  * @filter_table_restore: Restore filters removed from hardware
1091  * @filter_table_remove: Remove filters from hardware and tear down software state
1092  * @filter_update_rx_scatter: Update filters after change to rx scatter setting
1093  * @filter_insert: add or replace a filter
1094  * @filter_remove_safe: remove a filter by ID, carefully
1095  * @filter_get_safe: retrieve a filter by ID, carefully
1096  * @filter_clear_rx: Remove all RX filters whose priority is less than or
1097  *	equal to the given priority and is not %EFX_FILTER_PRI_AUTO
1098  * @filter_count_rx_used: Get the number of filters in use at a given priority
1099  * @filter_get_rx_id_limit: Get maximum value of a filter id, plus 1
1100  * @filter_get_rx_ids: Get list of RX filters at a given priority
1101  * @filter_rfs_insert: Add or replace a filter for RFS.  This must be
1102  *	atomic.  The hardware change may be asynchronous but should
1103  *	not be delayed for long.  It may fail if this can't be done
1104  *	atomically.
1105  * @filter_rfs_expire_one: Consider expiring a filter inserted for RFS.
1106  *	This must check whether the specified table entry is used by RFS
1107  *	and that rps_may_expire_flow() returns true for it.
1108  * @mtd_probe: Probe and add MTD partitions associated with this net device,
1109  *	 using efx_mtd_add()
1110  * @mtd_rename: Set an MTD partition name using the net device name
1111  * @mtd_read: Read from an MTD partition
1112  * @mtd_erase: Erase part of an MTD partition
1113  * @mtd_write: Write to an MTD partition
1114  * @mtd_sync: Wait for write-back to complete on MTD partition.  This
1115  *	also notifies the driver that a writer has finished using this
1116  *	partition.
1117  * @ptp_write_host_time: Send host time to MC as part of sync protocol
1118  * @ptp_set_ts_sync_events: Enable or disable sync events for inline RX
1119  *	timestamping, possibly only temporarily for the purposes of a reset.
1120  * @ptp_set_ts_config: Set hardware timestamp configuration.  The flags
1121  *	and tx_type will already have been validated but this operation
1122  *	must validate and update rx_filter.
1123  * @get_phys_port_id: Get the underlying physical port id.
1124  * @set_mac_address: Set the MAC address of the device
1125  * @tso_versions: Returns mask of firmware-assisted TSO versions supported.
1126  *	If %NULL, then device does not support any TSO version.
1127  * @udp_tnl_push_ports: Push the list of UDP tunnel ports to the NIC if required.
1128  * @udp_tnl_add_port: Add a UDP tunnel port
1129  * @udp_tnl_has_port: Check if a port has been added as UDP tunnel
1130  * @udp_tnl_del_port: Remove a UDP tunnel port
1131  * @revision: Hardware architecture revision
1132  * @txd_ptr_tbl_base: TX descriptor ring base address
1133  * @rxd_ptr_tbl_base: RX descriptor ring base address
1134  * @buf_tbl_base: Buffer table base address
1135  * @evq_ptr_tbl_base: Event queue pointer table base address
1136  * @evq_rptr_tbl_base: Event queue read-pointer table base address
1137  * @max_dma_mask: Maximum possible DMA mask
1138  * @rx_prefix_size: Size of RX prefix before packet data
1139  * @rx_hash_offset: Offset of RX flow hash within prefix
1140  * @rx_ts_offset: Offset of timestamp within prefix
1141  * @rx_buffer_padding: Size of padding at end of RX packet
1142  * @can_rx_scatter: NIC is able to scatter packets to multiple buffers
1143  * @always_rx_scatter: NIC will always scatter packets to multiple buffers
1144  * @option_descriptors: NIC supports TX option descriptors
1145  * @min_interrupt_mode: Lowest capability interrupt mode supported
1146  *	from &enum efx_int_mode.
1147  * @max_interrupt_mode: Highest capability interrupt mode supported
1148  *	from &enum efx_int_mode.
1149  * @timer_period_max: Maximum period of interrupt timer (in ticks)
1150  * @offload_features: net_device feature flags for protocol offload
1151  *	features implemented in hardware
1152  * @mcdi_max_ver: Maximum MCDI version supported
1153  * @hwtstamp_filters: Mask of hardware timestamp filter types supported
1154  */
1155 struct efx_nic_type {
1156 	bool is_vf;
1157 	unsigned int mem_bar;
1158 	unsigned int (*mem_map_size)(struct efx_nic *efx);
1159 	int (*probe)(struct efx_nic *efx);
1160 	void (*remove)(struct efx_nic *efx);
1161 	int (*init)(struct efx_nic *efx);
1162 	int (*dimension_resources)(struct efx_nic *efx);
1163 	void (*fini)(struct efx_nic *efx);
1164 	void (*monitor)(struct efx_nic *efx);
1165 	enum reset_type (*map_reset_reason)(enum reset_type reason);
1166 	int (*map_reset_flags)(u32 *flags);
1167 	int (*reset)(struct efx_nic *efx, enum reset_type method);
1168 	int (*probe_port)(struct efx_nic *efx);
1169 	void (*remove_port)(struct efx_nic *efx);
1170 	bool (*handle_global_event)(struct efx_channel *channel, efx_qword_t *);
1171 	int (*fini_dmaq)(struct efx_nic *efx);
1172 	void (*prepare_flush)(struct efx_nic *efx);
1173 	void (*finish_flush)(struct efx_nic *efx);
1174 	void (*prepare_flr)(struct efx_nic *efx);
1175 	void (*finish_flr)(struct efx_nic *efx);
1176 	size_t (*describe_stats)(struct efx_nic *efx, u8 *names);
1177 	size_t (*update_stats)(struct efx_nic *efx, u64 *full_stats,
1178 			       struct rtnl_link_stats64 *core_stats);
1179 	void (*start_stats)(struct efx_nic *efx);
1180 	void (*pull_stats)(struct efx_nic *efx);
1181 	void (*stop_stats)(struct efx_nic *efx);
1182 	void (*set_id_led)(struct efx_nic *efx, enum efx_led_mode mode);
1183 	void (*push_irq_moderation)(struct efx_channel *channel);
1184 	int (*reconfigure_port)(struct efx_nic *efx);
1185 	void (*prepare_enable_fc_tx)(struct efx_nic *efx);
1186 	int (*reconfigure_mac)(struct efx_nic *efx);
1187 	bool (*check_mac_fault)(struct efx_nic *efx);
1188 	void (*get_wol)(struct efx_nic *efx, struct ethtool_wolinfo *wol);
1189 	int (*set_wol)(struct efx_nic *efx, u32 type);
1190 	void (*resume_wol)(struct efx_nic *efx);
1191 	int (*test_chip)(struct efx_nic *efx, struct efx_self_tests *tests);
1192 	int (*test_nvram)(struct efx_nic *efx);
1193 	void (*mcdi_request)(struct efx_nic *efx,
1194 			     const efx_dword_t *hdr, size_t hdr_len,
1195 			     const efx_dword_t *sdu, size_t sdu_len);
1196 	bool (*mcdi_poll_response)(struct efx_nic *efx);
1197 	void (*mcdi_read_response)(struct efx_nic *efx, efx_dword_t *pdu,
1198 				   size_t pdu_offset, size_t pdu_len);
1199 	int (*mcdi_poll_reboot)(struct efx_nic *efx);
1200 	void (*mcdi_reboot_detected)(struct efx_nic *efx);
1201 	void (*irq_enable_master)(struct efx_nic *efx);
1202 	int (*irq_test_generate)(struct efx_nic *efx);
1203 	void (*irq_disable_non_ev)(struct efx_nic *efx);
1204 	irqreturn_t (*irq_handle_msi)(int irq, void *dev_id);
1205 	irqreturn_t (*irq_handle_legacy)(int irq, void *dev_id);
1206 	int (*tx_probe)(struct efx_tx_queue *tx_queue);
1207 	void (*tx_init)(struct efx_tx_queue *tx_queue);
1208 	void (*tx_remove)(struct efx_tx_queue *tx_queue);
1209 	void (*tx_write)(struct efx_tx_queue *tx_queue);
1210 	unsigned int (*tx_limit_len)(struct efx_tx_queue *tx_queue,
1211 				     dma_addr_t dma_addr, unsigned int len);
1212 	int (*rx_push_rss_config)(struct efx_nic *efx, bool user,
1213 				  const u32 *rx_indir_table, const u8 *key);
1214 	int (*rx_pull_rss_config)(struct efx_nic *efx);
1215 	int (*rx_probe)(struct efx_rx_queue *rx_queue);
1216 	void (*rx_init)(struct efx_rx_queue *rx_queue);
1217 	void (*rx_remove)(struct efx_rx_queue *rx_queue);
1218 	void (*rx_write)(struct efx_rx_queue *rx_queue);
1219 	void (*rx_defer_refill)(struct efx_rx_queue *rx_queue);
1220 	int (*ev_probe)(struct efx_channel *channel);
1221 	int (*ev_init)(struct efx_channel *channel);
1222 	void (*ev_fini)(struct efx_channel *channel);
1223 	void (*ev_remove)(struct efx_channel *channel);
1224 	int (*ev_process)(struct efx_channel *channel, int quota);
1225 	void (*ev_read_ack)(struct efx_channel *channel);
1226 	void (*ev_test_generate)(struct efx_channel *channel);
1227 	int (*filter_table_probe)(struct efx_nic *efx);
1228 	void (*filter_table_restore)(struct efx_nic *efx);
1229 	void (*filter_table_remove)(struct efx_nic *efx);
1230 	void (*filter_update_rx_scatter)(struct efx_nic *efx);
1231 	s32 (*filter_insert)(struct efx_nic *efx,
1232 			     struct efx_filter_spec *spec, bool replace);
1233 	int (*filter_remove_safe)(struct efx_nic *efx,
1234 				  enum efx_filter_priority priority,
1235 				  u32 filter_id);
1236 	int (*filter_get_safe)(struct efx_nic *efx,
1237 			       enum efx_filter_priority priority,
1238 			       u32 filter_id, struct efx_filter_spec *);
1239 	int (*filter_clear_rx)(struct efx_nic *efx,
1240 			       enum efx_filter_priority priority);
1241 	u32 (*filter_count_rx_used)(struct efx_nic *efx,
1242 				    enum efx_filter_priority priority);
1243 	u32 (*filter_get_rx_id_limit)(struct efx_nic *efx);
1244 	s32 (*filter_get_rx_ids)(struct efx_nic *efx,
1245 				 enum efx_filter_priority priority,
1246 				 u32 *buf, u32 size);
1247 #ifdef CONFIG_RFS_ACCEL
1248 	s32 (*filter_rfs_insert)(struct efx_nic *efx,
1249 				 struct efx_filter_spec *spec);
1250 	bool (*filter_rfs_expire_one)(struct efx_nic *efx, u32 flow_id,
1251 				      unsigned int index);
1252 #endif
1253 #ifdef CONFIG_SFC_MTD
1254 	int (*mtd_probe)(struct efx_nic *efx);
1255 	void (*mtd_rename)(struct efx_mtd_partition *part);
1256 	int (*mtd_read)(struct mtd_info *mtd, loff_t start, size_t len,
1257 			size_t *retlen, u8 *buffer);
1258 	int (*mtd_erase)(struct mtd_info *mtd, loff_t start, size_t len);
1259 	int (*mtd_write)(struct mtd_info *mtd, loff_t start, size_t len,
1260 			 size_t *retlen, const u8 *buffer);
1261 	int (*mtd_sync)(struct mtd_info *mtd);
1262 #endif
1263 	void (*ptp_write_host_time)(struct efx_nic *efx, u32 host_time);
1264 	int (*ptp_set_ts_sync_events)(struct efx_nic *efx, bool en, bool temp);
1265 	int (*ptp_set_ts_config)(struct efx_nic *efx,
1266 				 struct hwtstamp_config *init);
1267 	int (*sriov_configure)(struct efx_nic *efx, int num_vfs);
1268 	int (*vlan_rx_add_vid)(struct efx_nic *efx, __be16 proto, u16 vid);
1269 	int (*vlan_rx_kill_vid)(struct efx_nic *efx, __be16 proto, u16 vid);
1270 	int (*get_phys_port_id)(struct efx_nic *efx,
1271 				struct netdev_phys_item_id *ppid);
1272 	int (*sriov_init)(struct efx_nic *efx);
1273 	void (*sriov_fini)(struct efx_nic *efx);
1274 	bool (*sriov_wanted)(struct efx_nic *efx);
1275 	void (*sriov_reset)(struct efx_nic *efx);
1276 	void (*sriov_flr)(struct efx_nic *efx, unsigned vf_i);
1277 	int (*sriov_set_vf_mac)(struct efx_nic *efx, int vf_i, u8 *mac);
1278 	int (*sriov_set_vf_vlan)(struct efx_nic *efx, int vf_i, u16 vlan,
1279 				 u8 qos);
1280 	int (*sriov_set_vf_spoofchk)(struct efx_nic *efx, int vf_i,
1281 				     bool spoofchk);
1282 	int (*sriov_get_vf_config)(struct efx_nic *efx, int vf_i,
1283 				   struct ifla_vf_info *ivi);
1284 	int (*sriov_set_vf_link_state)(struct efx_nic *efx, int vf_i,
1285 				       int link_state);
1286 	int (*vswitching_probe)(struct efx_nic *efx);
1287 	int (*vswitching_restore)(struct efx_nic *efx);
1288 	void (*vswitching_remove)(struct efx_nic *efx);
1289 	int (*get_mac_address)(struct efx_nic *efx, unsigned char *perm_addr);
1290 	int (*set_mac_address)(struct efx_nic *efx);
1291 	u32 (*tso_versions)(struct efx_nic *efx);
1292 	int (*udp_tnl_push_ports)(struct efx_nic *efx);
1293 	int (*udp_tnl_add_port)(struct efx_nic *efx, struct efx_udp_tunnel tnl);
1294 	bool (*udp_tnl_has_port)(struct efx_nic *efx, __be16 port);
1295 	int (*udp_tnl_del_port)(struct efx_nic *efx, struct efx_udp_tunnel tnl);
1296 
1297 	int revision;
1298 	unsigned int txd_ptr_tbl_base;
1299 	unsigned int rxd_ptr_tbl_base;
1300 	unsigned int buf_tbl_base;
1301 	unsigned int evq_ptr_tbl_base;
1302 	unsigned int evq_rptr_tbl_base;
1303 	u64 max_dma_mask;
1304 	unsigned int rx_prefix_size;
1305 	unsigned int rx_hash_offset;
1306 	unsigned int rx_ts_offset;
1307 	unsigned int rx_buffer_padding;
1308 	bool can_rx_scatter;
1309 	bool always_rx_scatter;
1310 	bool option_descriptors;
1311 	unsigned int min_interrupt_mode;
1312 	unsigned int max_interrupt_mode;
1313 	unsigned int timer_period_max;
1314 	netdev_features_t offload_features;
1315 	int mcdi_max_ver;
1316 	unsigned int max_rx_ip_filters;
1317 	u32 hwtstamp_filters;
1318 	unsigned int rx_hash_key_size;
1319 };
1320 
1321 /**************************************************************************
1322  *
1323  * Prototypes and inline functions
1324  *
1325  *************************************************************************/
1326 
1327 static inline struct efx_channel *
1328 efx_get_channel(struct efx_nic *efx, unsigned index)
1329 {
1330 	EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_channels);
1331 	return efx->channel[index];
1332 }
1333 
1334 /* Iterate over all used channels */
1335 #define efx_for_each_channel(_channel, _efx)				\
1336 	for (_channel = (_efx)->channel[0];				\
1337 	     _channel;							\
1338 	     _channel = (_channel->channel + 1 < (_efx)->n_channels) ?	\
1339 		     (_efx)->channel[_channel->channel + 1] : NULL)
1340 
1341 /* Iterate over all used channels in reverse */
1342 #define efx_for_each_channel_rev(_channel, _efx)			\
1343 	for (_channel = (_efx)->channel[(_efx)->n_channels - 1];	\
1344 	     _channel;							\
1345 	     _channel = _channel->channel ?				\
1346 		     (_efx)->channel[_channel->channel - 1] : NULL)
1347 
1348 static inline struct efx_tx_queue *
1349 efx_get_tx_queue(struct efx_nic *efx, unsigned index, unsigned type)
1350 {
1351 	EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_tx_channels ||
1352 				  type >= EFX_TXQ_TYPES);
1353 	return &efx->channel[efx->tx_channel_offset + index]->tx_queue[type];
1354 }
1355 
1356 static inline bool efx_channel_has_tx_queues(struct efx_channel *channel)
1357 {
1358 	return channel->channel - channel->efx->tx_channel_offset <
1359 		channel->efx->n_tx_channels;
1360 }
1361 
1362 static inline struct efx_tx_queue *
1363 efx_channel_get_tx_queue(struct efx_channel *channel, unsigned type)
1364 {
1365 	EFX_WARN_ON_ONCE_PARANOID(!efx_channel_has_tx_queues(channel) ||
1366 				  type >= EFX_TXQ_TYPES);
1367 	return &channel->tx_queue[type];
1368 }
1369 
1370 static inline bool efx_tx_queue_used(struct efx_tx_queue *tx_queue)
1371 {
1372 	return !(tx_queue->efx->net_dev->num_tc < 2 &&
1373 		 tx_queue->queue & EFX_TXQ_TYPE_HIGHPRI);
1374 }
1375 
1376 /* Iterate over all TX queues belonging to a channel */
1377 #define efx_for_each_channel_tx_queue(_tx_queue, _channel)		\
1378 	if (!efx_channel_has_tx_queues(_channel))			\
1379 		;							\
1380 	else								\
1381 		for (_tx_queue = (_channel)->tx_queue;			\
1382 		     _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES && \
1383 			     efx_tx_queue_used(_tx_queue);		\
1384 		     _tx_queue++)
1385 
1386 /* Iterate over all possible TX queues belonging to a channel */
1387 #define efx_for_each_possible_channel_tx_queue(_tx_queue, _channel)	\
1388 	if (!efx_channel_has_tx_queues(_channel))			\
1389 		;							\
1390 	else								\
1391 		for (_tx_queue = (_channel)->tx_queue;			\
1392 		     _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES;	\
1393 		     _tx_queue++)
1394 
1395 static inline bool efx_channel_has_rx_queue(struct efx_channel *channel)
1396 {
1397 	return channel->rx_queue.core_index >= 0;
1398 }
1399 
1400 static inline struct efx_rx_queue *
1401 efx_channel_get_rx_queue(struct efx_channel *channel)
1402 {
1403 	EFX_WARN_ON_ONCE_PARANOID(!efx_channel_has_rx_queue(channel));
1404 	return &channel->rx_queue;
1405 }
1406 
1407 /* Iterate over all RX queues belonging to a channel */
1408 #define efx_for_each_channel_rx_queue(_rx_queue, _channel)		\
1409 	if (!efx_channel_has_rx_queue(_channel))			\
1410 		;							\
1411 	else								\
1412 		for (_rx_queue = &(_channel)->rx_queue;			\
1413 		     _rx_queue;						\
1414 		     _rx_queue = NULL)
1415 
1416 static inline struct efx_channel *
1417 efx_rx_queue_channel(struct efx_rx_queue *rx_queue)
1418 {
1419 	return container_of(rx_queue, struct efx_channel, rx_queue);
1420 }
1421 
1422 static inline int efx_rx_queue_index(struct efx_rx_queue *rx_queue)
1423 {
1424 	return efx_rx_queue_channel(rx_queue)->channel;
1425 }
1426 
1427 /* Returns a pointer to the specified receive buffer in the RX
1428  * descriptor queue.
1429  */
1430 static inline struct efx_rx_buffer *efx_rx_buffer(struct efx_rx_queue *rx_queue,
1431 						  unsigned int index)
1432 {
1433 	return &rx_queue->buffer[index];
1434 }
1435 
1436 /**
1437  * EFX_MAX_FRAME_LEN - calculate maximum frame length
1438  *
1439  * This calculates the maximum frame length that will be used for a
1440  * given MTU.  The frame length will be equal to the MTU plus a
1441  * constant amount of header space and padding.  This is the quantity
1442  * that the net driver will program into the MAC as the maximum frame
1443  * length.
1444  *
1445  * The 10G MAC requires 8-byte alignment on the frame
1446  * length, so we round up to the nearest 8.
1447  *
1448  * Re-clocking by the XGXS on RX can reduce an IPG to 32 bits (half an
1449  * XGMII cycle).  If the frame length reaches the maximum value in the
1450  * same cycle, the XMAC can miss the IPG altogether.  We work around
1451  * this by adding a further 16 bytes.
1452  */
1453 #define EFX_FRAME_PAD	16
1454 #define EFX_MAX_FRAME_LEN(mtu) \
1455 	(ALIGN(((mtu) + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN + EFX_FRAME_PAD), 8))
1456 
1457 static inline bool efx_xmit_with_hwtstamp(struct sk_buff *skb)
1458 {
1459 	return skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP;
1460 }
1461 static inline void efx_xmit_hwtstamp_pending(struct sk_buff *skb)
1462 {
1463 	skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
1464 }
1465 
1466 /* Get all supported features.
1467  * If a feature is not fixed, it is present in hw_features.
1468  * If a feature is fixed, it does not present in hw_features, but
1469  * always in features.
1470  */
1471 static inline netdev_features_t efx_supported_features(const struct efx_nic *efx)
1472 {
1473 	const struct net_device *net_dev = efx->net_dev;
1474 
1475 	return net_dev->features | net_dev->hw_features;
1476 }
1477 
1478 /* Get the current TX queue insert index. */
1479 static inline unsigned int
1480 efx_tx_queue_get_insert_index(const struct efx_tx_queue *tx_queue)
1481 {
1482 	return tx_queue->insert_count & tx_queue->ptr_mask;
1483 }
1484 
1485 /* Get a TX buffer. */
1486 static inline struct efx_tx_buffer *
1487 __efx_tx_queue_get_insert_buffer(const struct efx_tx_queue *tx_queue)
1488 {
1489 	return &tx_queue->buffer[efx_tx_queue_get_insert_index(tx_queue)];
1490 }
1491 
1492 /* Get a TX buffer, checking it's not currently in use. */
1493 static inline struct efx_tx_buffer *
1494 efx_tx_queue_get_insert_buffer(const struct efx_tx_queue *tx_queue)
1495 {
1496 	struct efx_tx_buffer *buffer =
1497 		__efx_tx_queue_get_insert_buffer(tx_queue);
1498 
1499 	EFX_WARN_ON_ONCE_PARANOID(buffer->len);
1500 	EFX_WARN_ON_ONCE_PARANOID(buffer->flags);
1501 	EFX_WARN_ON_ONCE_PARANOID(buffer->unmap_len);
1502 
1503 	return buffer;
1504 }
1505 
1506 #endif /* EFX_NET_DRIVER_H */
1507