treewide: remove redundant IS_ERR() before error code check
[linux/fpc-iii.git] / drivers / net / ethernet / sfc / net_driver.h
blob9f9886f222c864d9dfb06bf6c39ce2900c064648
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /****************************************************************************
3 * Driver for Solarflare network controllers and boards
4 * Copyright 2005-2006 Fen Systems Ltd.
5 * Copyright 2005-2013 Solarflare Communications Inc.
6 */
8 /* Common definitions for all Efx net driver code */
10 #ifndef EFX_NET_DRIVER_H
11 #define EFX_NET_DRIVER_H
13 #include <linux/netdevice.h>
14 #include <linux/etherdevice.h>
15 #include <linux/ethtool.h>
16 #include <linux/if_vlan.h>
17 #include <linux/timer.h>
18 #include <linux/mdio.h>
19 #include <linux/list.h>
20 #include <linux/pci.h>
21 #include <linux/device.h>
22 #include <linux/highmem.h>
23 #include <linux/workqueue.h>
24 #include <linux/mutex.h>
25 #include <linux/rwsem.h>
26 #include <linux/vmalloc.h>
27 #include <linux/mtd/mtd.h>
28 #include <net/busy_poll.h>
29 #include <net/xdp.h>
31 #include "enum.h"
32 #include "bitfield.h"
33 #include "filter.h"
35 /**************************************************************************
37 * Build definitions
39 **************************************************************************/
41 #define EFX_DRIVER_VERSION "4.1"
43 #ifdef DEBUG
44 #define EFX_WARN_ON_ONCE_PARANOID(x) WARN_ON_ONCE(x)
45 #define EFX_WARN_ON_PARANOID(x) WARN_ON(x)
46 #else
47 #define EFX_WARN_ON_ONCE_PARANOID(x) do {} while (0)
48 #define EFX_WARN_ON_PARANOID(x) do {} while (0)
49 #endif
51 /**************************************************************************
53 * Efx data structures
55 **************************************************************************/
57 #define EFX_MAX_CHANNELS 32U
58 #define EFX_MAX_RX_QUEUES EFX_MAX_CHANNELS
59 #define EFX_EXTRA_CHANNEL_IOV 0
60 #define EFX_EXTRA_CHANNEL_PTP 1
61 #define EFX_MAX_EXTRA_CHANNELS 2U
63 /* Checksum generation is a per-queue option in hardware, so each
64 * queue visible to the networking core is backed by two hardware TX
65 * queues. */
66 #define EFX_MAX_TX_TC 2
67 #define EFX_MAX_CORE_TX_QUEUES (EFX_MAX_TX_TC * EFX_MAX_CHANNELS)
68 #define EFX_TXQ_TYPE_OFFLOAD 1 /* flag */
69 #define EFX_TXQ_TYPE_HIGHPRI 2 /* flag */
70 #define EFX_TXQ_TYPES 4
71 #define EFX_MAX_TX_QUEUES (EFX_TXQ_TYPES * EFX_MAX_CHANNELS)
73 /* Maximum possible MTU the driver supports */
74 #define EFX_MAX_MTU (9 * 1024)
76 /* Minimum MTU, from RFC791 (IP) */
77 #define EFX_MIN_MTU 68
79 /* Size of an RX scatter buffer. Small enough to pack 2 into a 4K page,
80 * and should be a multiple of the cache line size.
82 #define EFX_RX_USR_BUF_SIZE (2048 - 256)
84 /* If possible, we should ensure cache line alignment at start and end
85 * of every buffer. Otherwise, we just need to ensure 4-byte
86 * alignment of the network header.
88 #if NET_IP_ALIGN == 0
89 #define EFX_RX_BUF_ALIGNMENT L1_CACHE_BYTES
90 #else
91 #define EFX_RX_BUF_ALIGNMENT 4
92 #endif
94 /* Forward declare Precision Time Protocol (PTP) support structure. */
95 struct efx_ptp_data;
96 struct hwtstamp_config;
98 struct efx_self_tests;
101 * struct efx_buffer - A general-purpose DMA buffer
102 * @addr: host base address of the buffer
103 * @dma_addr: DMA base address of the buffer
104 * @len: Buffer length, in bytes
106 * The NIC uses these buffers for its interrupt status registers and
107 * MAC stats dumps.
109 struct efx_buffer {
110 void *addr;
111 dma_addr_t dma_addr;
112 unsigned int len;
116 * struct efx_special_buffer - DMA buffer entered into buffer table
117 * @buf: Standard &struct efx_buffer
118 * @index: Buffer index within controller;s buffer table
119 * @entries: Number of buffer table entries
121 * The NIC has a buffer table that maps buffers of size %EFX_BUF_SIZE.
122 * Event and descriptor rings are addressed via one or more buffer
123 * table entries (and so can be physically non-contiguous, although we
124 * currently do not take advantage of that). On Falcon and Siena we
125 * have to take care of allocating and initialising the entries
126 * ourselves. On later hardware this is managed by the firmware and
127 * @index and @entries are left as 0.
129 struct efx_special_buffer {
130 struct efx_buffer buf;
131 unsigned int index;
132 unsigned int entries;
136 * struct efx_tx_buffer - buffer state for a TX descriptor
137 * @skb: When @flags & %EFX_TX_BUF_SKB, the associated socket buffer to be
138 * freed when descriptor completes
139 * @xdpf: When @flags & %EFX_TX_BUF_XDP, the XDP frame information; its @data
140 * member is the associated buffer to drop a page reference on.
141 * @option: When @flags & %EFX_TX_BUF_OPTION, an EF10-specific option
142 * 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.
151 struct efx_tx_buffer {
152 union {
153 const struct sk_buff *skb;
154 struct xdp_frame *xdpf;
156 union {
157 efx_qword_t option; /* EF10 */
158 dma_addr_t dma_addr;
160 unsigned short flags;
161 unsigned short len;
162 unsigned short unmap_len;
163 unsigned short dma_offset;
165 #define EFX_TX_BUF_CONT 1 /* not last descriptor of packet */
166 #define EFX_TX_BUF_SKB 2 /* buffer is last part of skb */
167 #define EFX_TX_BUF_MAP_SINGLE 8 /* buffer was mapped with dma_map_single() */
168 #define EFX_TX_BUF_OPTION 0x10 /* empty buffer for option descriptor */
169 #define EFX_TX_BUF_XDP 0x20 /* buffer was sent with XDP */
172 * struct efx_tx_queue - An Efx TX queue
174 * This is a ring buffer of TX fragments.
175 * Since the TX completion path always executes on the same
176 * CPU and the xmit path can operate on different CPUs,
177 * performance is increased by ensuring that the completion
178 * path and the xmit path operate on different cache lines.
179 * This is particularly important if the xmit path is always
180 * executing on one CPU which is different from the completion
181 * path. There is also a cache line for members which are
182 * read but not written on the fast path.
184 * @efx: The associated Efx NIC
185 * @queue: DMA queue number
186 * @tso_version: Version of TSO in use for this queue.
187 * @channel: The associated channel
188 * @core_txq: The networking core TX queue structure
189 * @buffer: The software buffer ring
190 * @cb_page: Array of pages of copy buffers. Carved up according to
191 * %EFX_TX_CB_ORDER into %EFX_TX_CB_SIZE-sized chunks.
192 * @txd: The hardware descriptor ring
193 * @ptr_mask: The size of the ring minus 1.
194 * @piobuf: PIO buffer region for this TX queue (shared with its partner).
195 * Size of the region is efx_piobuf_size.
196 * @piobuf_offset: Buffer offset to be specified in PIO descriptors
197 * @initialised: Has hardware queue been initialised?
198 * @timestamping: Is timestamping enabled for this channel?
199 * @xdp_tx: Is this an XDP tx queue?
200 * @handle_tso: TSO xmit preparation handler. Sets up the TSO metadata and
201 * may also map tx data, depending on the nature of the TSO implementation.
202 * @read_count: Current read pointer.
203 * This is the number of buffers that have been removed from both rings.
204 * @old_write_count: The value of @write_count when last checked.
205 * This is here for performance reasons. The xmit path will
206 * only get the up-to-date value of @write_count if this
207 * variable indicates that the queue is empty. This is to
208 * avoid cache-line ping-pong between the xmit path and the
209 * completion path.
210 * @merge_events: Number of TX merged completion events
211 * @completed_desc_ptr: Most recent completed pointer - only used with
212 * timestamping.
213 * @completed_timestamp_major: Top part of the most recent tx timestamp.
214 * @completed_timestamp_minor: Low part of the most recent tx timestamp.
215 * @insert_count: Current insert pointer
216 * This is the number of buffers that have been added to the
217 * software ring.
218 * @write_count: Current write pointer
219 * This is the number of buffers that have been added to the
220 * hardware ring.
221 * @packet_write_count: Completable write pointer
222 * This is the write pointer of the last packet written.
223 * Normally this will equal @write_count, but as option descriptors
224 * don't produce completion events, they won't update this.
225 * Filled in iff @efx->type->option_descriptors; only used for PIO.
226 * Thus, this is written and used on EF10, and neither on farch.
227 * @old_read_count: The value of read_count when last checked.
228 * This is here for performance reasons. The xmit path will
229 * only get the up-to-date value of read_count if this
230 * variable indicates that the queue is full. This is to
231 * avoid cache-line ping-pong between the xmit path and the
232 * completion path.
233 * @tso_bursts: Number of times TSO xmit invoked by kernel
234 * @tso_long_headers: Number of packets with headers too long for standard
235 * blocks
236 * @tso_packets: Number of packets via the TSO xmit path
237 * @tso_fallbacks: Number of times TSO fallback used
238 * @pushes: Number of times the TX push feature has been used
239 * @pio_packets: Number of times the TX PIO feature has been used
240 * @xmit_more_available: Are any packets waiting to be pushed to the NIC
241 * @cb_packets: Number of times the TX copybreak feature has been used
242 * @empty_read_count: If the completion path has seen the queue as empty
243 * and the transmission path has not yet checked this, the value of
244 * @read_count bitwise-added to %EFX_EMPTY_COUNT_VALID; otherwise 0.
246 struct efx_tx_queue {
247 /* Members which don't change on the fast path */
248 struct efx_nic *efx ____cacheline_aligned_in_smp;
249 unsigned queue;
250 unsigned int tso_version;
251 struct efx_channel *channel;
252 struct netdev_queue *core_txq;
253 struct efx_tx_buffer *buffer;
254 struct efx_buffer *cb_page;
255 struct efx_special_buffer txd;
256 unsigned int ptr_mask;
257 void __iomem *piobuf;
258 unsigned int piobuf_offset;
259 bool initialised;
260 bool timestamping;
261 bool xdp_tx;
263 /* Function pointers used in the fast path. */
264 int (*handle_tso)(struct efx_tx_queue*, struct sk_buff*, bool *);
266 /* Members used mainly on the completion path */
267 unsigned int read_count ____cacheline_aligned_in_smp;
268 unsigned int old_write_count;
269 unsigned int merge_events;
270 unsigned int bytes_compl;
271 unsigned int pkts_compl;
272 unsigned int completed_desc_ptr;
273 u32 completed_timestamp_major;
274 u32 completed_timestamp_minor;
276 /* Members used only on the xmit path */
277 unsigned int insert_count ____cacheline_aligned_in_smp;
278 unsigned int write_count;
279 unsigned int packet_write_count;
280 unsigned int old_read_count;
281 unsigned int tso_bursts;
282 unsigned int tso_long_headers;
283 unsigned int tso_packets;
284 unsigned int tso_fallbacks;
285 unsigned int pushes;
286 unsigned int pio_packets;
287 bool xmit_more_available;
288 unsigned int cb_packets;
289 /* Statistics to supplement MAC stats */
290 unsigned long tx_packets;
292 /* Members shared between paths and sometimes updated */
293 unsigned int empty_read_count ____cacheline_aligned_in_smp;
294 #define EFX_EMPTY_COUNT_VALID 0x80000000
295 atomic_t flush_outstanding;
298 #define EFX_TX_CB_ORDER 7
299 #define EFX_TX_CB_SIZE (1 << EFX_TX_CB_ORDER) - NET_IP_ALIGN
302 * struct efx_rx_buffer - An Efx RX data buffer
303 * @dma_addr: DMA base address of the buffer
304 * @page: The associated page buffer.
305 * Will be %NULL if the buffer slot is currently free.
306 * @page_offset: If pending: offset in @page of DMA base address.
307 * If completed: offset in @page of Ethernet header.
308 * @len: If pending: length for DMA descriptor.
309 * If completed: received length, excluding hash prefix.
310 * @flags: Flags for buffer and packet state. These are only set on the
311 * first buffer of a scattered packet.
313 struct efx_rx_buffer {
314 dma_addr_t dma_addr;
315 struct page *page;
316 u16 page_offset;
317 u16 len;
318 u16 flags;
320 #define EFX_RX_BUF_LAST_IN_PAGE 0x0001
321 #define EFX_RX_PKT_CSUMMED 0x0002
322 #define EFX_RX_PKT_DISCARD 0x0004
323 #define EFX_RX_PKT_TCP 0x0040
324 #define EFX_RX_PKT_PREFIX_LEN 0x0080 /* length is in prefix only */
325 #define EFX_RX_PKT_CSUM_LEVEL 0x0200
328 * struct efx_rx_page_state - Page-based rx buffer state
330 * Inserted at the start of every page allocated for receive buffers.
331 * Used to facilitate sharing dma mappings between recycled rx buffers
332 * and those passed up to the kernel.
334 * @dma_addr: The dma address of this page.
336 struct efx_rx_page_state {
337 dma_addr_t dma_addr;
339 unsigned int __pad[0] ____cacheline_aligned;
343 * struct efx_rx_queue - An Efx RX queue
344 * @efx: The associated Efx NIC
345 * @core_index: Index of network core RX queue. Will be >= 0 iff this
346 * is associated with a real RX queue.
347 * @buffer: The software buffer ring
348 * @rxd: The hardware descriptor ring
349 * @ptr_mask: The size of the ring minus 1.
350 * @refill_enabled: Enable refill whenever fill level is low
351 * @flush_pending: Set when a RX flush is pending. Has the same lifetime as
352 * @rxq_flush_pending.
353 * @added_count: Number of buffers added to the receive queue.
354 * @notified_count: Number of buffers given to NIC (<= @added_count).
355 * @removed_count: Number of buffers removed from the receive queue.
356 * @scatter_n: Used by NIC specific receive code.
357 * @scatter_len: Used by NIC specific receive code.
358 * @page_ring: The ring to store DMA mapped pages for reuse.
359 * @page_add: Counter to calculate the write pointer for the recycle ring.
360 * @page_remove: Counter to calculate the read pointer for the recycle ring.
361 * @page_recycle_count: The number of pages that have been recycled.
362 * @page_recycle_failed: The number of pages that couldn't be recycled because
363 * the kernel still held a reference to them.
364 * @page_recycle_full: The number of pages that were released because the
365 * recycle ring was full.
366 * @page_ptr_mask: The number of pages in the RX recycle ring minus 1.
367 * @max_fill: RX descriptor maximum fill level (<= ring size)
368 * @fast_fill_trigger: RX descriptor fill level that will trigger a fast fill
369 * (<= @max_fill)
370 * @min_fill: RX descriptor minimum non-zero fill level.
371 * This records the minimum fill level observed when a ring
372 * refill was triggered.
373 * @recycle_count: RX buffer recycle counter.
374 * @slow_fill: Timer used to defer efx_nic_generate_fill_event().
375 * @xdp_rxq_info: XDP specific RX queue information.
376 * @xdp_rxq_info_valid: Is xdp_rxq_info valid data?.
378 struct efx_rx_queue {
379 struct efx_nic *efx;
380 int core_index;
381 struct efx_rx_buffer *buffer;
382 struct efx_special_buffer rxd;
383 unsigned int ptr_mask;
384 bool refill_enabled;
385 bool flush_pending;
387 unsigned int added_count;
388 unsigned int notified_count;
389 unsigned int removed_count;
390 unsigned int scatter_n;
391 unsigned int scatter_len;
392 struct page **page_ring;
393 unsigned int page_add;
394 unsigned int page_remove;
395 unsigned int page_recycle_count;
396 unsigned int page_recycle_failed;
397 unsigned int page_recycle_full;
398 unsigned int page_ptr_mask;
399 unsigned int max_fill;
400 unsigned int fast_fill_trigger;
401 unsigned int min_fill;
402 unsigned int min_overfill;
403 unsigned int recycle_count;
404 struct timer_list slow_fill;
405 unsigned int slow_fill_count;
406 /* Statistics to supplement MAC stats */
407 unsigned long rx_packets;
408 struct xdp_rxq_info xdp_rxq_info;
409 bool xdp_rxq_info_valid;
412 enum efx_sync_events_state {
413 SYNC_EVENTS_DISABLED = 0,
414 SYNC_EVENTS_QUIESCENT,
415 SYNC_EVENTS_REQUESTED,
416 SYNC_EVENTS_VALID,
420 * struct efx_channel - An Efx channel
422 * A channel comprises an event queue, at least one TX queue, at least
423 * one RX queue, and an associated tasklet for processing the event
424 * queue.
426 * @efx: Associated Efx NIC
427 * @channel: Channel instance number
428 * @type: Channel type definition
429 * @eventq_init: Event queue initialised flag
430 * @enabled: Channel enabled indicator
431 * @irq: IRQ number (MSI and MSI-X only)
432 * @irq_moderation_us: IRQ moderation value (in microseconds)
433 * @napi_dev: Net device used with NAPI
434 * @napi_str: NAPI control structure
435 * @state: state for NAPI vs busy polling
436 * @state_lock: lock protecting @state
437 * @eventq: Event queue buffer
438 * @eventq_mask: Event queue pointer mask
439 * @eventq_read_ptr: Event queue read pointer
440 * @event_test_cpu: Last CPU to handle interrupt or test event for this channel
441 * @irq_count: Number of IRQs since last adaptive moderation decision
442 * @irq_mod_score: IRQ moderation score
443 * @rfs_filter_count: number of accelerated RFS filters currently in place;
444 * equals the count of @rps_flow_id slots filled
445 * @rfs_last_expiry: value of jiffies last time some accelerated RFS filters
446 * were checked for expiry
447 * @rfs_expire_index: next accelerated RFS filter ID to check for expiry
448 * @n_rfs_succeeded: number of successful accelerated RFS filter insertions
449 * @n_rfs_failed; number of failed accelerated RFS filter insertions
450 * @filter_work: Work item for efx_filter_rfs_expire()
451 * @rps_flow_id: Flow IDs of filters allocated for accelerated RFS,
452 * indexed by filter ID
453 * @n_rx_tobe_disc: Count of RX_TOBE_DISC errors
454 * @n_rx_ip_hdr_chksum_err: Count of RX IP header checksum errors
455 * @n_rx_tcp_udp_chksum_err: Count of RX TCP and UDP checksum errors
456 * @n_rx_mcast_mismatch: Count of unmatched multicast frames
457 * @n_rx_frm_trunc: Count of RX_FRM_TRUNC errors
458 * @n_rx_overlength: Count of RX_OVERLENGTH errors
459 * @n_skbuff_leaks: Count of skbuffs leaked due to RX overrun
460 * @n_rx_nodesc_trunc: Number of RX packets truncated and then dropped due to
461 * lack of descriptors
462 * @n_rx_merge_events: Number of RX merged completion events
463 * @n_rx_merge_packets: Number of RX packets completed by merged events
464 * @n_rx_xdp_drops: Count of RX packets intentionally dropped due to XDP
465 * @n_rx_xdp_bad_drops: Count of RX packets dropped due to XDP errors
466 * @n_rx_xdp_tx: Count of RX packets retransmitted due to XDP
467 * @n_rx_xdp_redirect: Count of RX packets redirected to a different NIC by XDP
468 * @rx_pkt_n_frags: Number of fragments in next packet to be delivered by
469 * __efx_rx_packet(), or zero if there is none
470 * @rx_pkt_index: Ring index of first buffer for next packet to be delivered
471 * by __efx_rx_packet(), if @rx_pkt_n_frags != 0
472 * @rx_list: list of SKBs from current RX, awaiting processing
473 * @rx_queue: RX queue for this channel
474 * @tx_queue: TX queues for this channel
475 * @sync_events_state: Current state of sync events on this channel
476 * @sync_timestamp_major: Major part of the last ptp sync event
477 * @sync_timestamp_minor: Minor part of the last ptp sync event
479 struct efx_channel {
480 struct efx_nic *efx;
481 int channel;
482 const struct efx_channel_type *type;
483 bool eventq_init;
484 bool enabled;
485 int irq;
486 unsigned int irq_moderation_us;
487 struct net_device *napi_dev;
488 struct napi_struct napi_str;
489 #ifdef CONFIG_NET_RX_BUSY_POLL
490 unsigned long busy_poll_state;
491 #endif
492 struct efx_special_buffer eventq;
493 unsigned int eventq_mask;
494 unsigned int eventq_read_ptr;
495 int event_test_cpu;
497 unsigned int irq_count;
498 unsigned int irq_mod_score;
499 #ifdef CONFIG_RFS_ACCEL
500 unsigned int rfs_filter_count;
501 unsigned int rfs_last_expiry;
502 unsigned int rfs_expire_index;
503 unsigned int n_rfs_succeeded;
504 unsigned int n_rfs_failed;
505 struct delayed_work filter_work;
506 #define RPS_FLOW_ID_INVALID 0xFFFFFFFF
507 u32 *rps_flow_id;
508 #endif
510 unsigned int n_rx_tobe_disc;
511 unsigned int n_rx_ip_hdr_chksum_err;
512 unsigned int n_rx_tcp_udp_chksum_err;
513 unsigned int n_rx_outer_ip_hdr_chksum_err;
514 unsigned int n_rx_outer_tcp_udp_chksum_err;
515 unsigned int n_rx_inner_ip_hdr_chksum_err;
516 unsigned int n_rx_inner_tcp_udp_chksum_err;
517 unsigned int n_rx_eth_crc_err;
518 unsigned int n_rx_mcast_mismatch;
519 unsigned int n_rx_frm_trunc;
520 unsigned int n_rx_overlength;
521 unsigned int n_skbuff_leaks;
522 unsigned int n_rx_nodesc_trunc;
523 unsigned int n_rx_merge_events;
524 unsigned int n_rx_merge_packets;
525 unsigned int n_rx_xdp_drops;
526 unsigned int n_rx_xdp_bad_drops;
527 unsigned int n_rx_xdp_tx;
528 unsigned int n_rx_xdp_redirect;
530 unsigned int rx_pkt_n_frags;
531 unsigned int rx_pkt_index;
533 struct list_head *rx_list;
535 struct efx_rx_queue rx_queue;
536 struct efx_tx_queue tx_queue[EFX_TXQ_TYPES];
538 enum efx_sync_events_state sync_events_state;
539 u32 sync_timestamp_major;
540 u32 sync_timestamp_minor;
544 * struct efx_msi_context - Context for each MSI
545 * @efx: The associated NIC
546 * @index: Index of the channel/IRQ
547 * @name: Name of the channel/IRQ
549 * Unlike &struct efx_channel, this is never reallocated and is always
550 * safe for the IRQ handler to access.
552 struct efx_msi_context {
553 struct efx_nic *efx;
554 unsigned int index;
555 char name[IFNAMSIZ + 6];
559 * struct efx_channel_type - distinguishes traffic and extra channels
560 * @handle_no_channel: Handle failure to allocate an extra channel
561 * @pre_probe: Set up extra state prior to initialisation
562 * @post_remove: Tear down extra state after finalisation, if allocated.
563 * May be called on channels that have not been probed.
564 * @get_name: Generate the channel's name (used for its IRQ handler)
565 * @copy: Copy the channel state prior to reallocation. May be %NULL if
566 * reallocation is not supported.
567 * @receive_skb: Handle an skb ready to be passed to netif_receive_skb()
568 * @want_txqs: Determine whether this channel should have TX queues
569 * created. If %NULL, TX queues are not created.
570 * @keep_eventq: Flag for whether event queue should be kept initialised
571 * while the device is stopped
572 * @want_pio: Flag for whether PIO buffers should be linked to this
573 * channel's TX queues.
575 struct efx_channel_type {
576 void (*handle_no_channel)(struct efx_nic *);
577 int (*pre_probe)(struct efx_channel *);
578 void (*post_remove)(struct efx_channel *);
579 void (*get_name)(struct efx_channel *, char *buf, size_t len);
580 struct efx_channel *(*copy)(const struct efx_channel *);
581 bool (*receive_skb)(struct efx_channel *, struct sk_buff *);
582 bool (*want_txqs)(struct efx_channel *);
583 bool keep_eventq;
584 bool want_pio;
587 enum efx_led_mode {
588 EFX_LED_OFF = 0,
589 EFX_LED_ON = 1,
590 EFX_LED_DEFAULT = 2
593 #define STRING_TABLE_LOOKUP(val, member) \
594 ((val) < member ## _max) ? member ## _names[val] : "(invalid)"
596 extern const char *const efx_loopback_mode_names[];
597 extern const unsigned int efx_loopback_mode_max;
598 #define LOOPBACK_MODE(efx) \
599 STRING_TABLE_LOOKUP((efx)->loopback_mode, efx_loopback_mode)
601 extern const char *const efx_reset_type_names[];
602 extern const unsigned int efx_reset_type_max;
603 #define RESET_TYPE(type) \
604 STRING_TABLE_LOOKUP(type, efx_reset_type)
606 void efx_get_udp_tunnel_type_name(u16 type, char *buf, size_t buflen);
608 enum efx_int_mode {
609 /* Be careful if altering to correct macro below */
610 EFX_INT_MODE_MSIX = 0,
611 EFX_INT_MODE_MSI = 1,
612 EFX_INT_MODE_LEGACY = 2,
613 EFX_INT_MODE_MAX /* Insert any new items before this */
615 #define EFX_INT_MODE_USE_MSI(x) (((x)->interrupt_mode) <= EFX_INT_MODE_MSI)
617 enum nic_state {
618 STATE_UNINIT = 0, /* device being probed/removed or is frozen */
619 STATE_READY = 1, /* hardware ready and netdev registered */
620 STATE_DISABLED = 2, /* device disabled due to hardware errors */
621 STATE_RECOVERY = 3, /* device recovering from PCI error */
624 /* Forward declaration */
625 struct efx_nic;
627 /* Pseudo bit-mask flow control field */
628 #define EFX_FC_RX FLOW_CTRL_RX
629 #define EFX_FC_TX FLOW_CTRL_TX
630 #define EFX_FC_AUTO 4
633 * struct efx_link_state - Current state of the link
634 * @up: Link is up
635 * @fd: Link is full-duplex
636 * @fc: Actual flow control flags
637 * @speed: Link speed (Mbps)
639 struct efx_link_state {
640 bool up;
641 bool fd;
642 u8 fc;
643 unsigned int speed;
646 static inline bool efx_link_state_equal(const struct efx_link_state *left,
647 const struct efx_link_state *right)
649 return left->up == right->up && left->fd == right->fd &&
650 left->fc == right->fc && left->speed == right->speed;
654 * struct efx_phy_operations - Efx PHY operations table
655 * @probe: Probe PHY and initialise efx->mdio.mode_support, efx->mdio.mmds,
656 * efx->loopback_modes.
657 * @init: Initialise PHY
658 * @fini: Shut down PHY
659 * @reconfigure: Reconfigure PHY (e.g. for new link parameters)
660 * @poll: Update @link_state and report whether it changed.
661 * Serialised by the mac_lock.
662 * @get_link_ksettings: Get ethtool settings. Serialised by the mac_lock.
663 * @set_link_ksettings: Set ethtool settings. Serialised by the mac_lock.
664 * @get_fecparam: Get Forward Error Correction settings. Serialised by mac_lock.
665 * @set_fecparam: Set Forward Error Correction settings. Serialised by mac_lock.
666 * @set_npage_adv: Set abilities advertised in (Extended) Next Page
667 * (only needed where AN bit is set in mmds)
668 * @test_alive: Test that PHY is 'alive' (online)
669 * @test_name: Get the name of a PHY-specific test/result
670 * @run_tests: Run tests and record results as appropriate (offline).
671 * Flags are the ethtool tests flags.
673 struct efx_phy_operations {
674 int (*probe) (struct efx_nic *efx);
675 int (*init) (struct efx_nic *efx);
676 void (*fini) (struct efx_nic *efx);
677 void (*remove) (struct efx_nic *efx);
678 int (*reconfigure) (struct efx_nic *efx);
679 bool (*poll) (struct efx_nic *efx);
680 void (*get_link_ksettings)(struct efx_nic *efx,
681 struct ethtool_link_ksettings *cmd);
682 int (*set_link_ksettings)(struct efx_nic *efx,
683 const struct ethtool_link_ksettings *cmd);
684 int (*get_fecparam)(struct efx_nic *efx, struct ethtool_fecparam *fec);
685 int (*set_fecparam)(struct efx_nic *efx,
686 const struct ethtool_fecparam *fec);
687 void (*set_npage_adv) (struct efx_nic *efx, u32);
688 int (*test_alive) (struct efx_nic *efx);
689 const char *(*test_name) (struct efx_nic *efx, unsigned int index);
690 int (*run_tests) (struct efx_nic *efx, int *results, unsigned flags);
691 int (*get_module_eeprom) (struct efx_nic *efx,
692 struct ethtool_eeprom *ee,
693 u8 *data);
694 int (*get_module_info) (struct efx_nic *efx,
695 struct ethtool_modinfo *modinfo);
699 * enum efx_phy_mode - PHY operating mode flags
700 * @PHY_MODE_NORMAL: on and should pass traffic
701 * @PHY_MODE_TX_DISABLED: on with TX disabled
702 * @PHY_MODE_LOW_POWER: set to low power through MDIO
703 * @PHY_MODE_OFF: switched off through external control
704 * @PHY_MODE_SPECIAL: on but will not pass traffic
706 enum efx_phy_mode {
707 PHY_MODE_NORMAL = 0,
708 PHY_MODE_TX_DISABLED = 1,
709 PHY_MODE_LOW_POWER = 2,
710 PHY_MODE_OFF = 4,
711 PHY_MODE_SPECIAL = 8,
714 static inline bool efx_phy_mode_disabled(enum efx_phy_mode mode)
716 return !!(mode & ~PHY_MODE_TX_DISABLED);
720 * struct efx_hw_stat_desc - Description of a hardware statistic
721 * @name: Name of the statistic as visible through ethtool, or %NULL if
722 * it should not be exposed
723 * @dma_width: Width in bits (0 for non-DMA statistics)
724 * @offset: Offset within stats (ignored for non-DMA statistics)
726 struct efx_hw_stat_desc {
727 const char *name;
728 u16 dma_width;
729 u16 offset;
732 /* Number of bits used in a multicast filter hash address */
733 #define EFX_MCAST_HASH_BITS 8
735 /* Number of (single-bit) entries in a multicast filter hash */
736 #define EFX_MCAST_HASH_ENTRIES (1 << EFX_MCAST_HASH_BITS)
738 /* An Efx multicast filter hash */
739 union efx_multicast_hash {
740 u8 byte[EFX_MCAST_HASH_ENTRIES / 8];
741 efx_oword_t oword[EFX_MCAST_HASH_ENTRIES / sizeof(efx_oword_t) / 8];
744 struct vfdi_status;
746 /* The reserved RSS context value */
747 #define EFX_MCDI_RSS_CONTEXT_INVALID 0xffffffff
749 * struct efx_rss_context - A user-defined RSS context for filtering
750 * @list: node of linked list on which this struct is stored
751 * @context_id: the RSS_CONTEXT_ID returned by MC firmware, or
752 * %EFX_MCDI_RSS_CONTEXT_INVALID if this context is not present on the NIC.
753 * For Siena, 0 if RSS is active, else %EFX_MCDI_RSS_CONTEXT_INVALID.
754 * @user_id: the rss_context ID exposed to userspace over ethtool.
755 * @rx_hash_udp_4tuple: UDP 4-tuple hashing enabled
756 * @rx_hash_key: Toeplitz hash key for this RSS context
757 * @indir_table: Indirection table for this RSS context
759 struct efx_rss_context {
760 struct list_head list;
761 u32 context_id;
762 u32 user_id;
763 bool rx_hash_udp_4tuple;
764 u8 rx_hash_key[40];
765 u32 rx_indir_table[128];
768 #ifdef CONFIG_RFS_ACCEL
769 /* Order of these is important, since filter_id >= %EFX_ARFS_FILTER_ID_PENDING
770 * is used to test if filter does or will exist.
772 #define EFX_ARFS_FILTER_ID_PENDING -1
773 #define EFX_ARFS_FILTER_ID_ERROR -2
774 #define EFX_ARFS_FILTER_ID_REMOVING -3
776 * struct efx_arfs_rule - record of an ARFS filter and its IDs
777 * @node: linkage into hash table
778 * @spec: details of the filter (used as key for hash table). Use efx->type to
779 * determine which member to use.
780 * @rxq_index: channel to which the filter will steer traffic.
781 * @arfs_id: filter ID which was returned to ARFS
782 * @filter_id: index in software filter table. May be
783 * %EFX_ARFS_FILTER_ID_PENDING if filter was not inserted yet,
784 * %EFX_ARFS_FILTER_ID_ERROR if filter insertion failed, or
785 * %EFX_ARFS_FILTER_ID_REMOVING if expiry is currently removing the filter.
787 struct efx_arfs_rule {
788 struct hlist_node node;
789 struct efx_filter_spec spec;
790 u16 rxq_index;
791 u16 arfs_id;
792 s32 filter_id;
795 /* Size chosen so that the table is one page (4kB) */
796 #define EFX_ARFS_HASH_TABLE_SIZE 512
799 * struct efx_async_filter_insertion - Request to asynchronously insert a filter
800 * @net_dev: Reference to the netdevice
801 * @spec: The filter to insert
802 * @work: Workitem for this request
803 * @rxq_index: Identifies the channel for which this request was made
804 * @flow_id: Identifies the kernel-side flow for which this request was made
806 struct efx_async_filter_insertion {
807 struct net_device *net_dev;
808 struct efx_filter_spec spec;
809 struct work_struct work;
810 u16 rxq_index;
811 u32 flow_id;
814 /* Maximum number of ARFS workitems that may be in flight on an efx_nic */
815 #define EFX_RPS_MAX_IN_FLIGHT 8
816 #endif /* CONFIG_RFS_ACCEL */
819 * struct efx_nic - an Efx NIC
820 * @name: Device name (net device name or bus id before net device registered)
821 * @pci_dev: The PCI device
822 * @node: List node for maintaning primary/secondary function lists
823 * @primary: &struct efx_nic instance for the primary function of this
824 * controller. May be the same structure, and may be %NULL if no
825 * primary function is bound. Serialised by rtnl_lock.
826 * @secondary_list: List of &struct efx_nic instances for the secondary PCI
827 * functions of the controller, if this is for the primary function.
828 * Serialised by rtnl_lock.
829 * @type: Controller type attributes
830 * @legacy_irq: IRQ number
831 * @workqueue: Workqueue for port reconfigures and the HW monitor.
832 * Work items do not hold and must not acquire RTNL.
833 * @workqueue_name: Name of workqueue
834 * @reset_work: Scheduled reset workitem
835 * @membase_phys: Memory BAR value as physical address
836 * @membase: Memory BAR value
837 * @vi_stride: step between per-VI registers / memory regions
838 * @interrupt_mode: Interrupt mode
839 * @timer_quantum_ns: Interrupt timer quantum, in nanoseconds
840 * @timer_max_ns: Interrupt timer maximum value, in nanoseconds
841 * @irq_rx_adaptive: Adaptive IRQ moderation enabled for RX event queues
842 * @irq_rx_mod_step_us: Step size for IRQ moderation for RX event queues
843 * @irq_rx_moderation_us: IRQ moderation time for RX event queues
844 * @msg_enable: Log message enable flags
845 * @state: Device state number (%STATE_*). Serialised by the rtnl_lock.
846 * @reset_pending: Bitmask for pending resets
847 * @tx_queue: TX DMA queues
848 * @rx_queue: RX DMA queues
849 * @channel: Channels
850 * @msi_context: Context for each MSI
851 * @extra_channel_types: Types of extra (non-traffic) channels that
852 * should be allocated for this NIC
853 * @xdp_tx_queue_count: Number of entries in %xdp_tx_queues.
854 * @xdp_tx_queues: Array of pointers to tx queues used for XDP transmit.
855 * @rxq_entries: Size of receive queues requested by user.
856 * @txq_entries: Size of transmit queues requested by user.
857 * @txq_stop_thresh: TX queue fill level at or above which we stop it.
858 * @txq_wake_thresh: TX queue fill level at or below which we wake it.
859 * @tx_dc_base: Base qword address in SRAM of TX queue descriptor caches
860 * @rx_dc_base: Base qword address in SRAM of RX queue descriptor caches
861 * @sram_lim_qw: Qword address limit of SRAM
862 * @next_buffer_table: First available buffer table id
863 * @n_channels: Number of channels in use
864 * @n_rx_channels: Number of channels used for RX (= number of RX queues)
865 * @n_tx_channels: Number of channels used for TX
866 * @n_extra_tx_channels: Number of extra channels with TX queues
867 * @n_xdp_channels: Number of channels used for XDP TX
868 * @xdp_channel_offset: Offset of zeroth channel used for XPD TX.
869 * @xdp_tx_per_channel: Max number of TX queues on an XDP TX channel.
870 * @rx_ip_align: RX DMA address offset to have IP header aligned in
871 * in accordance with NET_IP_ALIGN
872 * @rx_dma_len: Current maximum RX DMA length
873 * @rx_buffer_order: Order (log2) of number of pages for each RX buffer
874 * @rx_buffer_truesize: Amortised allocation size of an RX buffer,
875 * for use in sk_buff::truesize
876 * @rx_prefix_size: Size of RX prefix before packet data
877 * @rx_packet_hash_offset: Offset of RX flow hash from start of packet data
878 * (valid only if @rx_prefix_size != 0; always negative)
879 * @rx_packet_len_offset: Offset of RX packet length from start of packet data
880 * (valid only for NICs that set %EFX_RX_PKT_PREFIX_LEN; always negative)
881 * @rx_packet_ts_offset: Offset of timestamp from start of packet data
882 * (valid only if channel->sync_timestamps_enabled; always negative)
883 * @rx_scatter: Scatter mode enabled for receives
884 * @rss_context: Main RSS context. Its @list member is the head of the list of
885 * RSS contexts created by user requests
886 * @rss_lock: Protects custom RSS context software state in @rss_context.list
887 * @int_error_count: Number of internal errors seen recently
888 * @int_error_expire: Time at which error count will be expired
889 * @irq_soft_enabled: Are IRQs soft-enabled? If not, IRQ handler will
890 * acknowledge but do nothing else.
891 * @irq_status: Interrupt status buffer
892 * @irq_zero_count: Number of legacy IRQs seen with queue flags == 0
893 * @irq_level: IRQ level/index for IRQs not triggered by an event queue
894 * @selftest_work: Work item for asynchronous self-test
895 * @mtd_list: List of MTDs attached to the NIC
896 * @nic_data: Hardware dependent state
897 * @mcdi: Management-Controller-to-Driver Interface state
898 * @mac_lock: MAC access lock. Protects @port_enabled, @phy_mode,
899 * efx_monitor() and efx_reconfigure_port()
900 * @port_enabled: Port enabled indicator.
901 * Serialises efx_stop_all(), efx_start_all(), efx_monitor() and
902 * efx_mac_work() with kernel interfaces. Safe to read under any
903 * one of the rtnl_lock, mac_lock, or netif_tx_lock, but all three must
904 * be held to modify it.
905 * @port_initialized: Port initialized?
906 * @net_dev: Operating system network device. Consider holding the rtnl lock
907 * @fixed_features: Features which cannot be turned off
908 * @num_mac_stats: Number of MAC stats reported by firmware (MAC_STATS_NUM_STATS
909 * field of %MC_CMD_GET_CAPABILITIES_V4 response, or %MC_CMD_MAC_NSTATS)
910 * @stats_buffer: DMA buffer for statistics
911 * @phy_type: PHY type
912 * @phy_op: PHY interface
913 * @phy_data: PHY private data (including PHY-specific stats)
914 * @mdio: PHY MDIO interface
915 * @mdio_bus: PHY MDIO bus ID (only used by Siena)
916 * @phy_mode: PHY operating mode. Serialised by @mac_lock.
917 * @link_advertising: Autonegotiation advertising flags
918 * @fec_config: Forward Error Correction configuration flags. For bit positions
919 * see &enum ethtool_fec_config_bits.
920 * @link_state: Current state of the link
921 * @n_link_state_changes: Number of times the link has changed state
922 * @unicast_filter: Flag for Falcon-arch simple unicast filter.
923 * Protected by @mac_lock.
924 * @multicast_hash: Multicast hash table for Falcon-arch.
925 * Protected by @mac_lock.
926 * @wanted_fc: Wanted flow control flags
927 * @fc_disable: When non-zero flow control is disabled. Typically used to
928 * ensure that network back pressure doesn't delay dma queue flushes.
929 * Serialised by the rtnl lock.
930 * @mac_work: Work item for changing MAC promiscuity and multicast hash
931 * @loopback_mode: Loopback status
932 * @loopback_modes: Supported loopback mode bitmask
933 * @loopback_selftest: Offline self-test private state
934 * @xdp_prog: Current XDP programme for this interface
935 * @filter_sem: Filter table rw_semaphore, protects existence of @filter_state
936 * @filter_state: Architecture-dependent filter table state
937 * @rps_mutex: Protects RPS state of all channels
938 * @rps_slot_map: bitmap of in-flight entries in @rps_slot
939 * @rps_slot: array of ARFS insertion requests for efx_filter_rfs_work()
940 * @rps_hash_lock: Protects ARFS filter mapping state (@rps_hash_table and
941 * @rps_next_id).
942 * @rps_hash_table: Mapping between ARFS filters and their various IDs
943 * @rps_next_id: next arfs_id for an ARFS filter
944 * @active_queues: Count of RX and TX queues that haven't been flushed and drained.
945 * @rxq_flush_pending: Count of number of receive queues that need to be flushed.
946 * Decremented when the efx_flush_rx_queue() is called.
947 * @rxq_flush_outstanding: Count of number of RX flushes started but not yet
948 * completed (either success or failure). Not used when MCDI is used to
949 * flush receive queues.
950 * @flush_wq: wait queue used by efx_nic_flush_queues() to wait for flush completions.
951 * @vf_count: Number of VFs intended to be enabled.
952 * @vf_init_count: Number of VFs that have been fully initialised.
953 * @vi_scale: log2 number of vnics per VF.
954 * @ptp_data: PTP state data
955 * @ptp_warned: has this NIC seen and warned about unexpected PTP events?
956 * @vpd_sn: Serial number read from VPD
957 * @xdp_rxq_info_failed: Have any of the rx queues failed to initialise their
958 * xdp_rxq_info structures?
959 * @monitor_work: Hardware monitor workitem
960 * @biu_lock: BIU (bus interface unit) lock
961 * @last_irq_cpu: Last CPU to handle a possible test interrupt. This
962 * field is used by efx_test_interrupts() to verify that an
963 * interrupt has occurred.
964 * @stats_lock: Statistics update lock. Must be held when calling
965 * efx_nic_type::{update,start,stop}_stats.
966 * @n_rx_noskb_drops: Count of RX packets dropped due to failure to allocate an skb
968 * This is stored in the private area of the &struct net_device.
970 struct efx_nic {
971 /* The following fields should be written very rarely */
973 char name[IFNAMSIZ];
974 struct list_head node;
975 struct efx_nic *primary;
976 struct list_head secondary_list;
977 struct pci_dev *pci_dev;
978 unsigned int port_num;
979 const struct efx_nic_type *type;
980 int legacy_irq;
981 bool eeh_disabled_legacy_irq;
982 struct workqueue_struct *workqueue;
983 char workqueue_name[16];
984 struct work_struct reset_work;
985 resource_size_t membase_phys;
986 void __iomem *membase;
988 unsigned int vi_stride;
990 enum efx_int_mode interrupt_mode;
991 unsigned int timer_quantum_ns;
992 unsigned int timer_max_ns;
993 bool irq_rx_adaptive;
994 unsigned int irq_mod_step_us;
995 unsigned int irq_rx_moderation_us;
996 u32 msg_enable;
998 enum nic_state state;
999 unsigned long reset_pending;
1001 struct efx_channel *channel[EFX_MAX_CHANNELS];
1002 struct efx_msi_context msi_context[EFX_MAX_CHANNELS];
1003 const struct efx_channel_type *
1004 extra_channel_type[EFX_MAX_EXTRA_CHANNELS];
1006 unsigned int xdp_tx_queue_count;
1007 struct efx_tx_queue **xdp_tx_queues;
1009 unsigned rxq_entries;
1010 unsigned txq_entries;
1011 unsigned int txq_stop_thresh;
1012 unsigned int txq_wake_thresh;
1014 unsigned tx_dc_base;
1015 unsigned rx_dc_base;
1016 unsigned sram_lim_qw;
1017 unsigned next_buffer_table;
1019 unsigned int max_channels;
1020 unsigned int max_tx_channels;
1021 unsigned n_channels;
1022 unsigned n_rx_channels;
1023 unsigned rss_spread;
1024 unsigned tx_channel_offset;
1025 unsigned n_tx_channels;
1026 unsigned n_extra_tx_channels;
1027 unsigned int n_xdp_channels;
1028 unsigned int xdp_channel_offset;
1029 unsigned int xdp_tx_per_channel;
1030 unsigned int rx_ip_align;
1031 unsigned int rx_dma_len;
1032 unsigned int rx_buffer_order;
1033 unsigned int rx_buffer_truesize;
1034 unsigned int rx_page_buf_step;
1035 unsigned int rx_bufs_per_page;
1036 unsigned int rx_pages_per_batch;
1037 unsigned int rx_prefix_size;
1038 int rx_packet_hash_offset;
1039 int rx_packet_len_offset;
1040 int rx_packet_ts_offset;
1041 bool rx_scatter;
1042 struct efx_rss_context rss_context;
1043 struct mutex rss_lock;
1045 unsigned int_error_count;
1046 unsigned long int_error_expire;
1048 bool irq_soft_enabled;
1049 struct efx_buffer irq_status;
1050 unsigned irq_zero_count;
1051 unsigned irq_level;
1052 struct delayed_work selftest_work;
1054 #ifdef CONFIG_SFC_MTD
1055 struct list_head mtd_list;
1056 #endif
1058 void *nic_data;
1059 struct efx_mcdi_data *mcdi;
1061 struct mutex mac_lock;
1062 struct work_struct mac_work;
1063 bool port_enabled;
1065 bool mc_bist_for_other_fn;
1066 bool port_initialized;
1067 struct net_device *net_dev;
1069 netdev_features_t fixed_features;
1071 u16 num_mac_stats;
1072 struct efx_buffer stats_buffer;
1073 u64 rx_nodesc_drops_total;
1074 u64 rx_nodesc_drops_while_down;
1075 bool rx_nodesc_drops_prev_state;
1077 unsigned int phy_type;
1078 const struct efx_phy_operations *phy_op;
1079 void *phy_data;
1080 struct mdio_if_info mdio;
1081 unsigned int mdio_bus;
1082 enum efx_phy_mode phy_mode;
1084 __ETHTOOL_DECLARE_LINK_MODE_MASK(link_advertising);
1085 u32 fec_config;
1086 struct efx_link_state link_state;
1087 unsigned int n_link_state_changes;
1089 bool unicast_filter;
1090 union efx_multicast_hash multicast_hash;
1091 u8 wanted_fc;
1092 unsigned fc_disable;
1094 atomic_t rx_reset;
1095 enum efx_loopback_mode loopback_mode;
1096 u64 loopback_modes;
1098 void *loopback_selftest;
1099 /* We access loopback_selftest immediately before running XDP,
1100 * so we want them next to each other.
1102 struct bpf_prog __rcu *xdp_prog;
1104 struct rw_semaphore filter_sem;
1105 void *filter_state;
1106 #ifdef CONFIG_RFS_ACCEL
1107 struct mutex rps_mutex;
1108 unsigned long rps_slot_map;
1109 struct efx_async_filter_insertion rps_slot[EFX_RPS_MAX_IN_FLIGHT];
1110 spinlock_t rps_hash_lock;
1111 struct hlist_head *rps_hash_table;
1112 u32 rps_next_id;
1113 #endif
1115 atomic_t active_queues;
1116 atomic_t rxq_flush_pending;
1117 atomic_t rxq_flush_outstanding;
1118 wait_queue_head_t flush_wq;
1120 #ifdef CONFIG_SFC_SRIOV
1121 unsigned vf_count;
1122 unsigned vf_init_count;
1123 unsigned vi_scale;
1124 #endif
1126 struct efx_ptp_data *ptp_data;
1127 bool ptp_warned;
1129 char *vpd_sn;
1130 bool xdp_rxq_info_failed;
1132 /* The following fields may be written more often */
1134 struct delayed_work monitor_work ____cacheline_aligned_in_smp;
1135 spinlock_t biu_lock;
1136 int last_irq_cpu;
1137 spinlock_t stats_lock;
1138 atomic_t n_rx_noskb_drops;
1141 static inline int efx_dev_registered(struct efx_nic *efx)
1143 return efx->net_dev->reg_state == NETREG_REGISTERED;
1146 static inline unsigned int efx_port_num(struct efx_nic *efx)
1148 return efx->port_num;
1151 struct efx_mtd_partition {
1152 struct list_head node;
1153 struct mtd_info mtd;
1154 const char *dev_type_name;
1155 const char *type_name;
1156 char name[IFNAMSIZ + 20];
1159 struct efx_udp_tunnel {
1160 u16 type; /* TUNNEL_ENCAP_UDP_PORT_ENTRY_foo, see mcdi_pcol.h */
1161 __be16 port;
1162 /* Count of repeated adds of the same port. Used only inside the list,
1163 * not in request arguments.
1165 u16 count;
1169 * struct efx_nic_type - Efx device type definition
1170 * @mem_bar: Get the memory BAR
1171 * @mem_map_size: Get memory BAR mapped size
1172 * @probe: Probe the controller
1173 * @remove: Free resources allocated by probe()
1174 * @init: Initialise the controller
1175 * @dimension_resources: Dimension controller resources (buffer table,
1176 * and VIs once the available interrupt resources are clear)
1177 * @fini: Shut down the controller
1178 * @monitor: Periodic function for polling link state and hardware monitor
1179 * @map_reset_reason: Map ethtool reset reason to a reset method
1180 * @map_reset_flags: Map ethtool reset flags to a reset method, if possible
1181 * @reset: Reset the controller hardware and possibly the PHY. This will
1182 * be called while the controller is uninitialised.
1183 * @probe_port: Probe the MAC and PHY
1184 * @remove_port: Free resources allocated by probe_port()
1185 * @handle_global_event: Handle a "global" event (may be %NULL)
1186 * @fini_dmaq: Flush and finalise DMA queues (RX and TX queues)
1187 * @prepare_flush: Prepare the hardware for flushing the DMA queues
1188 * (for Falcon architecture)
1189 * @finish_flush: Clean up after flushing the DMA queues (for Falcon
1190 * architecture)
1191 * @prepare_flr: Prepare for an FLR
1192 * @finish_flr: Clean up after an FLR
1193 * @describe_stats: Describe statistics for ethtool
1194 * @update_stats: Update statistics not provided by event handling.
1195 * Either argument may be %NULL.
1196 * @start_stats: Start the regular fetching of statistics
1197 * @pull_stats: Pull stats from the NIC and wait until they arrive.
1198 * @stop_stats: Stop the regular fetching of statistics
1199 * @set_id_led: Set state of identifying LED or revert to automatic function
1200 * @push_irq_moderation: Apply interrupt moderation value
1201 * @reconfigure_port: Push loopback/power/txdis changes to the MAC and PHY
1202 * @prepare_enable_fc_tx: Prepare MAC to enable pause frame TX (may be %NULL)
1203 * @reconfigure_mac: Push MAC address, MTU, flow control and filter settings
1204 * to the hardware. Serialised by the mac_lock.
1205 * @check_mac_fault: Check MAC fault state. True if fault present.
1206 * @get_wol: Get WoL configuration from driver state
1207 * @set_wol: Push WoL configuration to the NIC
1208 * @resume_wol: Synchronise WoL state between driver and MC (e.g. after resume)
1209 * @test_chip: Test registers. May use efx_farch_test_registers(), and is
1210 * expected to reset the NIC.
1211 * @test_nvram: Test validity of NVRAM contents
1212 * @mcdi_request: Send an MCDI request with the given header and SDU.
1213 * The SDU length may be any value from 0 up to the protocol-
1214 * defined maximum, but its buffer will be padded to a multiple
1215 * of 4 bytes.
1216 * @mcdi_poll_response: Test whether an MCDI response is available.
1217 * @mcdi_read_response: Read the MCDI response PDU. The offset will
1218 * be a multiple of 4. The length may not be, but the buffer
1219 * will be padded so it is safe to round up.
1220 * @mcdi_poll_reboot: Test whether the MCDI has rebooted. If so,
1221 * return an appropriate error code for aborting any current
1222 * request; otherwise return 0.
1223 * @irq_enable_master: Enable IRQs on the NIC. Each event queue must
1224 * be separately enabled after this.
1225 * @irq_test_generate: Generate a test IRQ
1226 * @irq_disable_non_ev: Disable non-event IRQs on the NIC. Each event
1227 * queue must be separately disabled before this.
1228 * @irq_handle_msi: Handle MSI for a channel. The @dev_id argument is
1229 * a pointer to the &struct efx_msi_context for the channel.
1230 * @irq_handle_legacy: Handle legacy interrupt. The @dev_id argument
1231 * is a pointer to the &struct efx_nic.
1232 * @tx_probe: Allocate resources for TX queue
1233 * @tx_init: Initialise TX queue on the NIC
1234 * @tx_remove: Free resources for TX queue
1235 * @tx_write: Write TX descriptors and doorbell
1236 * @rx_push_rss_config: Write RSS hash key and indirection table to the NIC
1237 * @rx_pull_rss_config: Read RSS hash key and indirection table back from the NIC
1238 * @rx_push_rss_context_config: Write RSS hash key and indirection table for
1239 * user RSS context to the NIC
1240 * @rx_pull_rss_context_config: Read RSS hash key and indirection table for user
1241 * RSS context back from the NIC
1242 * @rx_probe: Allocate resources for RX queue
1243 * @rx_init: Initialise RX queue on the NIC
1244 * @rx_remove: Free resources for RX queue
1245 * @rx_write: Write RX descriptors and doorbell
1246 * @rx_defer_refill: Generate a refill reminder event
1247 * @ev_probe: Allocate resources for event queue
1248 * @ev_init: Initialise event queue on the NIC
1249 * @ev_fini: Deinitialise event queue on the NIC
1250 * @ev_remove: Free resources for event queue
1251 * @ev_process: Process events for a queue, up to the given NAPI quota
1252 * @ev_read_ack: Acknowledge read events on a queue, rearming its IRQ
1253 * @ev_test_generate: Generate a test event
1254 * @filter_table_probe: Probe filter capabilities and set up filter software state
1255 * @filter_table_restore: Restore filters removed from hardware
1256 * @filter_table_remove: Remove filters from hardware and tear down software state
1257 * @filter_update_rx_scatter: Update filters after change to rx scatter setting
1258 * @filter_insert: add or replace a filter
1259 * @filter_remove_safe: remove a filter by ID, carefully
1260 * @filter_get_safe: retrieve a filter by ID, carefully
1261 * @filter_clear_rx: Remove all RX filters whose priority is less than or
1262 * equal to the given priority and is not %EFX_FILTER_PRI_AUTO
1263 * @filter_count_rx_used: Get the number of filters in use at a given priority
1264 * @filter_get_rx_id_limit: Get maximum value of a filter id, plus 1
1265 * @filter_get_rx_ids: Get list of RX filters at a given priority
1266 * @filter_rfs_expire_one: Consider expiring a filter inserted for RFS.
1267 * This must check whether the specified table entry is used by RFS
1268 * and that rps_may_expire_flow() returns true for it.
1269 * @mtd_probe: Probe and add MTD partitions associated with this net device,
1270 * using efx_mtd_add()
1271 * @mtd_rename: Set an MTD partition name using the net device name
1272 * @mtd_read: Read from an MTD partition
1273 * @mtd_erase: Erase part of an MTD partition
1274 * @mtd_write: Write to an MTD partition
1275 * @mtd_sync: Wait for write-back to complete on MTD partition. This
1276 * also notifies the driver that a writer has finished using this
1277 * partition.
1278 * @ptp_write_host_time: Send host time to MC as part of sync protocol
1279 * @ptp_set_ts_sync_events: Enable or disable sync events for inline RX
1280 * timestamping, possibly only temporarily for the purposes of a reset.
1281 * @ptp_set_ts_config: Set hardware timestamp configuration. The flags
1282 * and tx_type will already have been validated but this operation
1283 * must validate and update rx_filter.
1284 * @get_phys_port_id: Get the underlying physical port id.
1285 * @set_mac_address: Set the MAC address of the device
1286 * @tso_versions: Returns mask of firmware-assisted TSO versions supported.
1287 * If %NULL, then device does not support any TSO version.
1288 * @udp_tnl_push_ports: Push the list of UDP tunnel ports to the NIC if required.
1289 * @udp_tnl_add_port: Add a UDP tunnel port
1290 * @udp_tnl_has_port: Check if a port has been added as UDP tunnel
1291 * @udp_tnl_del_port: Remove a UDP tunnel port
1292 * @revision: Hardware architecture revision
1293 * @txd_ptr_tbl_base: TX descriptor ring base address
1294 * @rxd_ptr_tbl_base: RX descriptor ring base address
1295 * @buf_tbl_base: Buffer table base address
1296 * @evq_ptr_tbl_base: Event queue pointer table base address
1297 * @evq_rptr_tbl_base: Event queue read-pointer table base address
1298 * @max_dma_mask: Maximum possible DMA mask
1299 * @rx_prefix_size: Size of RX prefix before packet data
1300 * @rx_hash_offset: Offset of RX flow hash within prefix
1301 * @rx_ts_offset: Offset of timestamp within prefix
1302 * @rx_buffer_padding: Size of padding at end of RX packet
1303 * @can_rx_scatter: NIC is able to scatter packets to multiple buffers
1304 * @always_rx_scatter: NIC will always scatter packets to multiple buffers
1305 * @option_descriptors: NIC supports TX option descriptors
1306 * @min_interrupt_mode: Lowest capability interrupt mode supported
1307 * from &enum efx_int_mode.
1308 * @max_interrupt_mode: Highest capability interrupt mode supported
1309 * from &enum efx_int_mode.
1310 * @timer_period_max: Maximum period of interrupt timer (in ticks)
1311 * @offload_features: net_device feature flags for protocol offload
1312 * features implemented in hardware
1313 * @mcdi_max_ver: Maximum MCDI version supported
1314 * @hwtstamp_filters: Mask of hardware timestamp filter types supported
1316 struct efx_nic_type {
1317 bool is_vf;
1318 unsigned int (*mem_bar)(struct efx_nic *efx);
1319 unsigned int (*mem_map_size)(struct efx_nic *efx);
1320 int (*probe)(struct efx_nic *efx);
1321 void (*remove)(struct efx_nic *efx);
1322 int (*init)(struct efx_nic *efx);
1323 int (*dimension_resources)(struct efx_nic *efx);
1324 void (*fini)(struct efx_nic *efx);
1325 void (*monitor)(struct efx_nic *efx);
1326 enum reset_type (*map_reset_reason)(enum reset_type reason);
1327 int (*map_reset_flags)(u32 *flags);
1328 int (*reset)(struct efx_nic *efx, enum reset_type method);
1329 int (*probe_port)(struct efx_nic *efx);
1330 void (*remove_port)(struct efx_nic *efx);
1331 bool (*handle_global_event)(struct efx_channel *channel, efx_qword_t *);
1332 int (*fini_dmaq)(struct efx_nic *efx);
1333 void (*prepare_flush)(struct efx_nic *efx);
1334 void (*finish_flush)(struct efx_nic *efx);
1335 void (*prepare_flr)(struct efx_nic *efx);
1336 void (*finish_flr)(struct efx_nic *efx);
1337 size_t (*describe_stats)(struct efx_nic *efx, u8 *names);
1338 size_t (*update_stats)(struct efx_nic *efx, u64 *full_stats,
1339 struct rtnl_link_stats64 *core_stats);
1340 void (*start_stats)(struct efx_nic *efx);
1341 void (*pull_stats)(struct efx_nic *efx);
1342 void (*stop_stats)(struct efx_nic *efx);
1343 void (*set_id_led)(struct efx_nic *efx, enum efx_led_mode mode);
1344 void (*push_irq_moderation)(struct efx_channel *channel);
1345 int (*reconfigure_port)(struct efx_nic *efx);
1346 void (*prepare_enable_fc_tx)(struct efx_nic *efx);
1347 int (*reconfigure_mac)(struct efx_nic *efx);
1348 bool (*check_mac_fault)(struct efx_nic *efx);
1349 void (*get_wol)(struct efx_nic *efx, struct ethtool_wolinfo *wol);
1350 int (*set_wol)(struct efx_nic *efx, u32 type);
1351 void (*resume_wol)(struct efx_nic *efx);
1352 int (*test_chip)(struct efx_nic *efx, struct efx_self_tests *tests);
1353 int (*test_nvram)(struct efx_nic *efx);
1354 void (*mcdi_request)(struct efx_nic *efx,
1355 const efx_dword_t *hdr, size_t hdr_len,
1356 const efx_dword_t *sdu, size_t sdu_len);
1357 bool (*mcdi_poll_response)(struct efx_nic *efx);
1358 void (*mcdi_read_response)(struct efx_nic *efx, efx_dword_t *pdu,
1359 size_t pdu_offset, size_t pdu_len);
1360 int (*mcdi_poll_reboot)(struct efx_nic *efx);
1361 void (*mcdi_reboot_detected)(struct efx_nic *efx);
1362 void (*irq_enable_master)(struct efx_nic *efx);
1363 int (*irq_test_generate)(struct efx_nic *efx);
1364 void (*irq_disable_non_ev)(struct efx_nic *efx);
1365 irqreturn_t (*irq_handle_msi)(int irq, void *dev_id);
1366 irqreturn_t (*irq_handle_legacy)(int irq, void *dev_id);
1367 int (*tx_probe)(struct efx_tx_queue *tx_queue);
1368 void (*tx_init)(struct efx_tx_queue *tx_queue);
1369 void (*tx_remove)(struct efx_tx_queue *tx_queue);
1370 void (*tx_write)(struct efx_tx_queue *tx_queue);
1371 unsigned int (*tx_limit_len)(struct efx_tx_queue *tx_queue,
1372 dma_addr_t dma_addr, unsigned int len);
1373 int (*rx_push_rss_config)(struct efx_nic *efx, bool user,
1374 const u32 *rx_indir_table, const u8 *key);
1375 int (*rx_pull_rss_config)(struct efx_nic *efx);
1376 int (*rx_push_rss_context_config)(struct efx_nic *efx,
1377 struct efx_rss_context *ctx,
1378 const u32 *rx_indir_table,
1379 const u8 *key);
1380 int (*rx_pull_rss_context_config)(struct efx_nic *efx,
1381 struct efx_rss_context *ctx);
1382 void (*rx_restore_rss_contexts)(struct efx_nic *efx);
1383 int (*rx_probe)(struct efx_rx_queue *rx_queue);
1384 void (*rx_init)(struct efx_rx_queue *rx_queue);
1385 void (*rx_remove)(struct efx_rx_queue *rx_queue);
1386 void (*rx_write)(struct efx_rx_queue *rx_queue);
1387 void (*rx_defer_refill)(struct efx_rx_queue *rx_queue);
1388 int (*ev_probe)(struct efx_channel *channel);
1389 int (*ev_init)(struct efx_channel *channel);
1390 void (*ev_fini)(struct efx_channel *channel);
1391 void (*ev_remove)(struct efx_channel *channel);
1392 int (*ev_process)(struct efx_channel *channel, int quota);
1393 void (*ev_read_ack)(struct efx_channel *channel);
1394 void (*ev_test_generate)(struct efx_channel *channel);
1395 int (*filter_table_probe)(struct efx_nic *efx);
1396 void (*filter_table_restore)(struct efx_nic *efx);
1397 void (*filter_table_remove)(struct efx_nic *efx);
1398 void (*filter_update_rx_scatter)(struct efx_nic *efx);
1399 s32 (*filter_insert)(struct efx_nic *efx,
1400 struct efx_filter_spec *spec, bool replace);
1401 int (*filter_remove_safe)(struct efx_nic *efx,
1402 enum efx_filter_priority priority,
1403 u32 filter_id);
1404 int (*filter_get_safe)(struct efx_nic *efx,
1405 enum efx_filter_priority priority,
1406 u32 filter_id, struct efx_filter_spec *);
1407 int (*filter_clear_rx)(struct efx_nic *efx,
1408 enum efx_filter_priority priority);
1409 u32 (*filter_count_rx_used)(struct efx_nic *efx,
1410 enum efx_filter_priority priority);
1411 u32 (*filter_get_rx_id_limit)(struct efx_nic *efx);
1412 s32 (*filter_get_rx_ids)(struct efx_nic *efx,
1413 enum efx_filter_priority priority,
1414 u32 *buf, u32 size);
1415 #ifdef CONFIG_RFS_ACCEL
1416 bool (*filter_rfs_expire_one)(struct efx_nic *efx, u32 flow_id,
1417 unsigned int index);
1418 #endif
1419 #ifdef CONFIG_SFC_MTD
1420 int (*mtd_probe)(struct efx_nic *efx);
1421 void (*mtd_rename)(struct efx_mtd_partition *part);
1422 int (*mtd_read)(struct mtd_info *mtd, loff_t start, size_t len,
1423 size_t *retlen, u8 *buffer);
1424 int (*mtd_erase)(struct mtd_info *mtd, loff_t start, size_t len);
1425 int (*mtd_write)(struct mtd_info *mtd, loff_t start, size_t len,
1426 size_t *retlen, const u8 *buffer);
1427 int (*mtd_sync)(struct mtd_info *mtd);
1428 #endif
1429 void (*ptp_write_host_time)(struct efx_nic *efx, u32 host_time);
1430 int (*ptp_set_ts_sync_events)(struct efx_nic *efx, bool en, bool temp);
1431 int (*ptp_set_ts_config)(struct efx_nic *efx,
1432 struct hwtstamp_config *init);
1433 int (*sriov_configure)(struct efx_nic *efx, int num_vfs);
1434 int (*vlan_rx_add_vid)(struct efx_nic *efx, __be16 proto, u16 vid);
1435 int (*vlan_rx_kill_vid)(struct efx_nic *efx, __be16 proto, u16 vid);
1436 int (*get_phys_port_id)(struct efx_nic *efx,
1437 struct netdev_phys_item_id *ppid);
1438 int (*sriov_init)(struct efx_nic *efx);
1439 void (*sriov_fini)(struct efx_nic *efx);
1440 bool (*sriov_wanted)(struct efx_nic *efx);
1441 void (*sriov_reset)(struct efx_nic *efx);
1442 void (*sriov_flr)(struct efx_nic *efx, unsigned vf_i);
1443 int (*sriov_set_vf_mac)(struct efx_nic *efx, int vf_i, u8 *mac);
1444 int (*sriov_set_vf_vlan)(struct efx_nic *efx, int vf_i, u16 vlan,
1445 u8 qos);
1446 int (*sriov_set_vf_spoofchk)(struct efx_nic *efx, int vf_i,
1447 bool spoofchk);
1448 int (*sriov_get_vf_config)(struct efx_nic *efx, int vf_i,
1449 struct ifla_vf_info *ivi);
1450 int (*sriov_set_vf_link_state)(struct efx_nic *efx, int vf_i,
1451 int link_state);
1452 int (*vswitching_probe)(struct efx_nic *efx);
1453 int (*vswitching_restore)(struct efx_nic *efx);
1454 void (*vswitching_remove)(struct efx_nic *efx);
1455 int (*get_mac_address)(struct efx_nic *efx, unsigned char *perm_addr);
1456 int (*set_mac_address)(struct efx_nic *efx);
1457 u32 (*tso_versions)(struct efx_nic *efx);
1458 int (*udp_tnl_push_ports)(struct efx_nic *efx);
1459 int (*udp_tnl_add_port)(struct efx_nic *efx, struct efx_udp_tunnel tnl);
1460 bool (*udp_tnl_has_port)(struct efx_nic *efx, __be16 port);
1461 int (*udp_tnl_del_port)(struct efx_nic *efx, struct efx_udp_tunnel tnl);
1463 int revision;
1464 unsigned int txd_ptr_tbl_base;
1465 unsigned int rxd_ptr_tbl_base;
1466 unsigned int buf_tbl_base;
1467 unsigned int evq_ptr_tbl_base;
1468 unsigned int evq_rptr_tbl_base;
1469 u64 max_dma_mask;
1470 unsigned int rx_prefix_size;
1471 unsigned int rx_hash_offset;
1472 unsigned int rx_ts_offset;
1473 unsigned int rx_buffer_padding;
1474 bool can_rx_scatter;
1475 bool always_rx_scatter;
1476 bool option_descriptors;
1477 unsigned int min_interrupt_mode;
1478 unsigned int max_interrupt_mode;
1479 unsigned int timer_period_max;
1480 netdev_features_t offload_features;
1481 int mcdi_max_ver;
1482 unsigned int max_rx_ip_filters;
1483 u32 hwtstamp_filters;
1484 unsigned int rx_hash_key_size;
1487 /**************************************************************************
1489 * Prototypes and inline functions
1491 *************************************************************************/
1493 static inline struct efx_channel *
1494 efx_get_channel(struct efx_nic *efx, unsigned index)
1496 EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_channels);
1497 return efx->channel[index];
1500 /* Iterate over all used channels */
1501 #define efx_for_each_channel(_channel, _efx) \
1502 for (_channel = (_efx)->channel[0]; \
1503 _channel; \
1504 _channel = (_channel->channel + 1 < (_efx)->n_channels) ? \
1505 (_efx)->channel[_channel->channel + 1] : NULL)
1507 /* Iterate over all used channels in reverse */
1508 #define efx_for_each_channel_rev(_channel, _efx) \
1509 for (_channel = (_efx)->channel[(_efx)->n_channels - 1]; \
1510 _channel; \
1511 _channel = _channel->channel ? \
1512 (_efx)->channel[_channel->channel - 1] : NULL)
1514 static inline struct efx_tx_queue *
1515 efx_get_tx_queue(struct efx_nic *efx, unsigned index, unsigned type)
1517 EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_tx_channels ||
1518 type >= EFX_TXQ_TYPES);
1519 return &efx->channel[efx->tx_channel_offset + index]->tx_queue[type];
1522 static inline struct efx_channel *
1523 efx_get_xdp_channel(struct efx_nic *efx, unsigned int index)
1525 EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_xdp_channels);
1526 return efx->channel[efx->xdp_channel_offset + index];
1529 static inline bool efx_channel_is_xdp_tx(struct efx_channel *channel)
1531 return channel->channel - channel->efx->xdp_channel_offset <
1532 channel->efx->n_xdp_channels;
1535 static inline bool efx_channel_has_tx_queues(struct efx_channel *channel)
1537 return true;
1540 static inline struct efx_tx_queue *
1541 efx_channel_get_tx_queue(struct efx_channel *channel, unsigned type)
1543 EFX_WARN_ON_ONCE_PARANOID(!efx_channel_has_tx_queues(channel) ||
1544 type >= EFX_TXQ_TYPES);
1545 return &channel->tx_queue[type];
1548 static inline bool efx_tx_queue_used(struct efx_tx_queue *tx_queue)
1550 return !(tx_queue->efx->net_dev->num_tc < 2 &&
1551 tx_queue->queue & EFX_TXQ_TYPE_HIGHPRI);
1554 /* Iterate over all TX queues belonging to a channel */
1555 #define efx_for_each_channel_tx_queue(_tx_queue, _channel) \
1556 if (!efx_channel_has_tx_queues(_channel)) \
1558 else \
1559 for (_tx_queue = (_channel)->tx_queue; \
1560 _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES && \
1561 (efx_tx_queue_used(_tx_queue) || \
1562 efx_channel_is_xdp_tx(_channel)); \
1563 _tx_queue++)
1565 /* Iterate over all possible TX queues belonging to a channel */
1566 #define efx_for_each_possible_channel_tx_queue(_tx_queue, _channel) \
1567 if (!efx_channel_has_tx_queues(_channel)) \
1569 else \
1570 for (_tx_queue = (_channel)->tx_queue; \
1571 _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES; \
1572 _tx_queue++)
1574 static inline bool efx_channel_has_rx_queue(struct efx_channel *channel)
1576 return channel->rx_queue.core_index >= 0;
1579 static inline struct efx_rx_queue *
1580 efx_channel_get_rx_queue(struct efx_channel *channel)
1582 EFX_WARN_ON_ONCE_PARANOID(!efx_channel_has_rx_queue(channel));
1583 return &channel->rx_queue;
1586 /* Iterate over all RX queues belonging to a channel */
1587 #define efx_for_each_channel_rx_queue(_rx_queue, _channel) \
1588 if (!efx_channel_has_rx_queue(_channel)) \
1590 else \
1591 for (_rx_queue = &(_channel)->rx_queue; \
1592 _rx_queue; \
1593 _rx_queue = NULL)
1595 static inline struct efx_channel *
1596 efx_rx_queue_channel(struct efx_rx_queue *rx_queue)
1598 return container_of(rx_queue, struct efx_channel, rx_queue);
1601 static inline int efx_rx_queue_index(struct efx_rx_queue *rx_queue)
1603 return efx_rx_queue_channel(rx_queue)->channel;
1606 /* Returns a pointer to the specified receive buffer in the RX
1607 * descriptor queue.
1609 static inline struct efx_rx_buffer *efx_rx_buffer(struct efx_rx_queue *rx_queue,
1610 unsigned int index)
1612 return &rx_queue->buffer[index];
1615 static inline struct efx_rx_buffer *
1616 efx_rx_buf_next(struct efx_rx_queue *rx_queue, struct efx_rx_buffer *rx_buf)
1618 if (unlikely(rx_buf == efx_rx_buffer(rx_queue, rx_queue->ptr_mask)))
1619 return efx_rx_buffer(rx_queue, 0);
1620 else
1621 return rx_buf + 1;
1625 * EFX_MAX_FRAME_LEN - calculate maximum frame length
1627 * This calculates the maximum frame length that will be used for a
1628 * given MTU. The frame length will be equal to the MTU plus a
1629 * constant amount of header space and padding. This is the quantity
1630 * that the net driver will program into the MAC as the maximum frame
1631 * length.
1633 * The 10G MAC requires 8-byte alignment on the frame
1634 * length, so we round up to the nearest 8.
1636 * Re-clocking by the XGXS on RX can reduce an IPG to 32 bits (half an
1637 * XGMII cycle). If the frame length reaches the maximum value in the
1638 * same cycle, the XMAC can miss the IPG altogether. We work around
1639 * this by adding a further 16 bytes.
1641 #define EFX_FRAME_PAD 16
1642 #define EFX_MAX_FRAME_LEN(mtu) \
1643 (ALIGN(((mtu) + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN + EFX_FRAME_PAD), 8))
1645 static inline bool efx_xmit_with_hwtstamp(struct sk_buff *skb)
1647 return skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP;
1649 static inline void efx_xmit_hwtstamp_pending(struct sk_buff *skb)
1651 skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
1654 /* Get all supported features.
1655 * If a feature is not fixed, it is present in hw_features.
1656 * If a feature is fixed, it does not present in hw_features, but
1657 * always in features.
1659 static inline netdev_features_t efx_supported_features(const struct efx_nic *efx)
1661 const struct net_device *net_dev = efx->net_dev;
1663 return net_dev->features | net_dev->hw_features;
1666 /* Get the current TX queue insert index. */
1667 static inline unsigned int
1668 efx_tx_queue_get_insert_index(const struct efx_tx_queue *tx_queue)
1670 return tx_queue->insert_count & tx_queue->ptr_mask;
1673 /* Get a TX buffer. */
1674 static inline struct efx_tx_buffer *
1675 __efx_tx_queue_get_insert_buffer(const struct efx_tx_queue *tx_queue)
1677 return &tx_queue->buffer[efx_tx_queue_get_insert_index(tx_queue)];
1680 /* Get a TX buffer, checking it's not currently in use. */
1681 static inline struct efx_tx_buffer *
1682 efx_tx_queue_get_insert_buffer(const struct efx_tx_queue *tx_queue)
1684 struct efx_tx_buffer *buffer =
1685 __efx_tx_queue_get_insert_buffer(tx_queue);
1687 EFX_WARN_ON_ONCE_PARANOID(buffer->len);
1688 EFX_WARN_ON_ONCE_PARANOID(buffer->flags);
1689 EFX_WARN_ON_ONCE_PARANOID(buffer->unmap_len);
1691 return buffer;
1694 #endif /* EFX_NET_DRIVER_H */