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.
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>
35 /**************************************************************************
39 **************************************************************************/
42 #define EFX_WARN_ON_ONCE_PARANOID(x) WARN_ON_ONCE(x)
43 #define EFX_WARN_ON_PARANOID(x) WARN_ON(x)
45 #define EFX_WARN_ON_ONCE_PARANOID(x) do {} while (0)
46 #define EFX_WARN_ON_PARANOID(x) do {} while (0)
49 /**************************************************************************
53 **************************************************************************/
55 #define EFX_MAX_CHANNELS 32U
56 #define EFX_MAX_RX_QUEUES EFX_MAX_CHANNELS
57 #define EFX_EXTRA_CHANNEL_IOV 0
58 #define EFX_EXTRA_CHANNEL_PTP 1
59 #define EFX_MAX_EXTRA_CHANNELS 2U
61 /* Checksum generation is a per-queue option in hardware, so each
62 * queue visible to the networking core is backed by two hardware TX
64 #define EFX_MAX_TX_TC 2
65 #define EFX_MAX_CORE_TX_QUEUES (EFX_MAX_TX_TC * EFX_MAX_CHANNELS)
66 #define EFX_TXQ_TYPE_OFFLOAD 1 /* flag */
67 #define EFX_TXQ_TYPE_HIGHPRI 2 /* flag */
68 #define EFX_TXQ_TYPES 4
69 #define EFX_MAX_TX_QUEUES (EFX_TXQ_TYPES * EFX_MAX_CHANNELS)
71 /* Maximum possible MTU the driver supports */
72 #define EFX_MAX_MTU (9 * 1024)
74 /* Minimum MTU, from RFC791 (IP) */
75 #define EFX_MIN_MTU 68
77 /* Size of an RX scatter buffer. Small enough to pack 2 into a 4K page,
78 * and should be a multiple of the cache line size.
80 #define EFX_RX_USR_BUF_SIZE (2048 - 256)
82 /* If possible, we should ensure cache line alignment at start and end
83 * of every buffer. Otherwise, we just need to ensure 4-byte
84 * alignment of the network header.
87 #define EFX_RX_BUF_ALIGNMENT L1_CACHE_BYTES
89 #define EFX_RX_BUF_ALIGNMENT 4
92 /* Non-standard XDP_PACKET_HEADROOM and tailroom to satisfy XDP_REDIRECT and
93 * still fit two standard MTU size packets into a single 4K page.
95 #define EFX_XDP_HEADROOM 128
96 #define EFX_XDP_TAILROOM SKB_DATA_ALIGN(sizeof(struct skb_shared_info))
98 /* Forward declare Precision Time Protocol (PTP) support structure. */
100 struct hwtstamp_config;
102 struct efx_self_tests;
105 * struct efx_buffer - A general-purpose DMA buffer
106 * @addr: host base address of the buffer
107 * @dma_addr: DMA base address of the buffer
108 * @len: Buffer length, in bytes
110 * The NIC uses these buffers for its interrupt status registers and
120 * struct efx_special_buffer - DMA buffer entered into buffer table
121 * @buf: Standard &struct efx_buffer
122 * @index: Buffer index within controller;s buffer table
123 * @entries: Number of buffer table entries
125 * The NIC has a buffer table that maps buffers of size %EFX_BUF_SIZE.
126 * Event and descriptor rings are addressed via one or more buffer
127 * table entries (and so can be physically non-contiguous, although we
128 * currently do not take advantage of that). On Falcon and Siena we
129 * have to take care of allocating and initialising the entries
130 * ourselves. On later hardware this is managed by the firmware and
131 * @index and @entries are left as 0.
133 struct efx_special_buffer {
134 struct efx_buffer buf;
136 unsigned int entries;
140 * struct efx_tx_buffer - buffer state for a TX descriptor
141 * @skb: When @flags & %EFX_TX_BUF_SKB, the associated socket buffer to be
142 * freed when descriptor completes
143 * @xdpf: When @flags & %EFX_TX_BUF_XDP, the XDP frame information; its @data
144 * member is the associated buffer to drop a page reference on.
145 * @option: When @flags & %EFX_TX_BUF_OPTION, an EF10-specific option
147 * @dma_addr: DMA address of the fragment.
148 * @flags: Flags for allocation and DMA mapping type
149 * @len: Length of this fragment.
150 * This field is zero when the queue slot is empty.
151 * @unmap_len: Length of this fragment to unmap
152 * @dma_offset: Offset of @dma_addr from the address of the backing DMA mapping.
153 * Only valid if @unmap_len != 0.
155 struct efx_tx_buffer {
157 const struct sk_buff *skb;
158 struct xdp_frame *xdpf;
161 efx_qword_t option; /* EF10 */
164 unsigned short flags;
166 unsigned short unmap_len;
167 unsigned short dma_offset;
169 #define EFX_TX_BUF_CONT 1 /* not last descriptor of packet */
170 #define EFX_TX_BUF_SKB 2 /* buffer is last part of skb */
171 #define EFX_TX_BUF_MAP_SINGLE 8 /* buffer was mapped with dma_map_single() */
172 #define EFX_TX_BUF_OPTION 0x10 /* empty buffer for option descriptor */
173 #define EFX_TX_BUF_XDP 0x20 /* buffer was sent with XDP */
174 #define EFX_TX_BUF_TSO_V3 0x40 /* empty buffer for a TSO_V3 descriptor */
177 * struct efx_tx_queue - An Efx TX queue
179 * This is a ring buffer of TX fragments.
180 * Since the TX completion path always executes on the same
181 * CPU and the xmit path can operate on different CPUs,
182 * performance is increased by ensuring that the completion
183 * path and the xmit path operate on different cache lines.
184 * This is particularly important if the xmit path is always
185 * executing on one CPU which is different from the completion
186 * path. There is also a cache line for members which are
187 * read but not written on the fast path.
189 * @efx: The associated Efx NIC
190 * @queue: DMA queue number
191 * @label: Label for TX completion events.
192 * Is our index within @channel->tx_queue array.
193 * @tso_version: Version of TSO in use for this queue.
194 * @channel: The associated channel
195 * @core_txq: The networking core TX queue structure
196 * @buffer: The software buffer ring
197 * @cb_page: Array of pages of copy buffers. Carved up according to
198 * %EFX_TX_CB_ORDER into %EFX_TX_CB_SIZE-sized chunks.
199 * @txd: The hardware descriptor ring
200 * @ptr_mask: The size of the ring minus 1.
201 * @piobuf: PIO buffer region for this TX queue (shared with its partner).
202 * Size of the region is efx_piobuf_size.
203 * @piobuf_offset: Buffer offset to be specified in PIO descriptors
204 * @initialised: Has hardware queue been initialised?
205 * @timestamping: Is timestamping enabled for this channel?
206 * @xdp_tx: Is this an XDP tx queue?
207 * @handle_tso: TSO xmit preparation handler. Sets up the TSO metadata and
208 * may also map tx data, depending on the nature of the TSO implementation.
209 * @read_count: Current read pointer.
210 * This is the number of buffers that have been removed from both rings.
211 * @old_write_count: The value of @write_count when last checked.
212 * This is here for performance reasons. The xmit path will
213 * only get the up-to-date value of @write_count if this
214 * variable indicates that the queue is empty. This is to
215 * avoid cache-line ping-pong between the xmit path and the
217 * @merge_events: Number of TX merged completion events
218 * @completed_timestamp_major: Top part of the most recent tx timestamp.
219 * @completed_timestamp_minor: Low part of the most recent tx timestamp.
220 * @insert_count: Current insert pointer
221 * This is the number of buffers that have been added to the
223 * @write_count: Current write pointer
224 * This is the number of buffers that have been added to the
226 * @packet_write_count: Completable write pointer
227 * This is the write pointer of the last packet written.
228 * Normally this will equal @write_count, but as option descriptors
229 * don't produce completion events, they won't update this.
230 * Filled in iff @efx->type->option_descriptors; only used for PIO.
231 * Thus, this is written and used on EF10, and neither on farch.
232 * @old_read_count: The value of read_count when last checked.
233 * This is here for performance reasons. The xmit path will
234 * only get the up-to-date value of read_count if this
235 * variable indicates that the queue is full. This is to
236 * avoid cache-line ping-pong between the xmit path and the
238 * @tso_bursts: Number of times TSO xmit invoked by kernel
239 * @tso_long_headers: Number of packets with headers too long for standard
241 * @tso_packets: Number of packets via the TSO xmit path
242 * @tso_fallbacks: Number of times TSO fallback used
243 * @pushes: Number of times the TX push feature has been used
244 * @pio_packets: Number of times the TX PIO feature has been used
245 * @xmit_pending: Are any packets waiting to be pushed to the NIC
246 * @cb_packets: Number of times the TX copybreak feature has been used
247 * @notify_count: Count of notified descriptors to the NIC
248 * @empty_read_count: If the completion path has seen the queue as empty
249 * and the transmission path has not yet checked this, the value of
250 * @read_count bitwise-added to %EFX_EMPTY_COUNT_VALID; otherwise 0.
252 struct efx_tx_queue {
253 /* Members which don't change on the fast path */
254 struct efx_nic *efx ____cacheline_aligned_in_smp;
257 unsigned int tso_version;
258 struct efx_channel *channel;
259 struct netdev_queue *core_txq;
260 struct efx_tx_buffer *buffer;
261 struct efx_buffer *cb_page;
262 struct efx_special_buffer txd;
263 unsigned int ptr_mask;
264 void __iomem *piobuf;
265 unsigned int piobuf_offset;
270 /* Function pointers used in the fast path. */
271 int (*handle_tso)(struct efx_tx_queue*, struct sk_buff*, bool *);
273 /* Members used mainly on the completion path */
274 unsigned int read_count ____cacheline_aligned_in_smp;
275 unsigned int old_write_count;
276 unsigned int merge_events;
277 unsigned int bytes_compl;
278 unsigned int pkts_compl;
279 u32 completed_timestamp_major;
280 u32 completed_timestamp_minor;
282 /* Members used only on the xmit path */
283 unsigned int insert_count ____cacheline_aligned_in_smp;
284 unsigned int write_count;
285 unsigned int packet_write_count;
286 unsigned int old_read_count;
287 unsigned int tso_bursts;
288 unsigned int tso_long_headers;
289 unsigned int tso_packets;
290 unsigned int tso_fallbacks;
292 unsigned int pio_packets;
294 unsigned int cb_packets;
295 unsigned int notify_count;
296 /* Statistics to supplement MAC stats */
297 unsigned long tx_packets;
299 /* Members shared between paths and sometimes updated */
300 unsigned int empty_read_count ____cacheline_aligned_in_smp;
301 #define EFX_EMPTY_COUNT_VALID 0x80000000
302 atomic_t flush_outstanding;
305 #define EFX_TX_CB_ORDER 7
306 #define EFX_TX_CB_SIZE (1 << EFX_TX_CB_ORDER) - NET_IP_ALIGN
309 * struct efx_rx_buffer - An Efx RX data buffer
310 * @dma_addr: DMA base address of the buffer
311 * @page: The associated page buffer.
312 * Will be %NULL if the buffer slot is currently free.
313 * @page_offset: If pending: offset in @page of DMA base address.
314 * If completed: offset in @page of Ethernet header.
315 * @len: If pending: length for DMA descriptor.
316 * If completed: received length, excluding hash prefix.
317 * @flags: Flags for buffer and packet state. These are only set on the
318 * first buffer of a scattered packet.
320 struct efx_rx_buffer {
327 #define EFX_RX_BUF_LAST_IN_PAGE 0x0001
328 #define EFX_RX_PKT_CSUMMED 0x0002
329 #define EFX_RX_PKT_DISCARD 0x0004
330 #define EFX_RX_PKT_TCP 0x0040
331 #define EFX_RX_PKT_PREFIX_LEN 0x0080 /* length is in prefix only */
332 #define EFX_RX_PKT_CSUM_LEVEL 0x0200
335 * struct efx_rx_page_state - Page-based rx buffer state
337 * Inserted at the start of every page allocated for receive buffers.
338 * Used to facilitate sharing dma mappings between recycled rx buffers
339 * and those passed up to the kernel.
341 * @dma_addr: The dma address of this page.
343 struct efx_rx_page_state {
346 unsigned int __pad[] ____cacheline_aligned;
350 * struct efx_rx_queue - An Efx RX queue
351 * @efx: The associated Efx NIC
352 * @core_index: Index of network core RX queue. Will be >= 0 iff this
353 * is associated with a real RX queue.
354 * @buffer: The software buffer ring
355 * @rxd: The hardware descriptor ring
356 * @ptr_mask: The size of the ring minus 1.
357 * @refill_enabled: Enable refill whenever fill level is low
358 * @flush_pending: Set when a RX flush is pending. Has the same lifetime as
359 * @rxq_flush_pending.
360 * @added_count: Number of buffers added to the receive queue.
361 * @notified_count: Number of buffers given to NIC (<= @added_count).
362 * @removed_count: Number of buffers removed from the receive queue.
363 * @scatter_n: Used by NIC specific receive code.
364 * @scatter_len: Used by NIC specific receive code.
365 * @page_ring: The ring to store DMA mapped pages for reuse.
366 * @page_add: Counter to calculate the write pointer for the recycle ring.
367 * @page_remove: Counter to calculate the read pointer for the recycle ring.
368 * @page_recycle_count: The number of pages that have been recycled.
369 * @page_recycle_failed: The number of pages that couldn't be recycled because
370 * the kernel still held a reference to them.
371 * @page_recycle_full: The number of pages that were released because the
372 * recycle ring was full.
373 * @page_ptr_mask: The number of pages in the RX recycle ring minus 1.
374 * @max_fill: RX descriptor maximum fill level (<= ring size)
375 * @fast_fill_trigger: RX descriptor fill level that will trigger a fast fill
377 * @min_fill: RX descriptor minimum non-zero fill level.
378 * This records the minimum fill level observed when a ring
379 * refill was triggered.
380 * @recycle_count: RX buffer recycle counter.
381 * @slow_fill: Timer used to defer efx_nic_generate_fill_event().
382 * @xdp_rxq_info: XDP specific RX queue information.
383 * @xdp_rxq_info_valid: Is xdp_rxq_info valid data?.
385 struct efx_rx_queue {
388 struct efx_rx_buffer *buffer;
389 struct efx_special_buffer rxd;
390 unsigned int ptr_mask;
394 unsigned int added_count;
395 unsigned int notified_count;
396 unsigned int removed_count;
397 unsigned int scatter_n;
398 unsigned int scatter_len;
399 struct page **page_ring;
400 unsigned int page_add;
401 unsigned int page_remove;
402 unsigned int page_recycle_count;
403 unsigned int page_recycle_failed;
404 unsigned int page_recycle_full;
405 unsigned int page_ptr_mask;
406 unsigned int max_fill;
407 unsigned int fast_fill_trigger;
408 unsigned int min_fill;
409 unsigned int min_overfill;
410 unsigned int recycle_count;
411 struct timer_list slow_fill;
412 unsigned int slow_fill_count;
413 /* Statistics to supplement MAC stats */
414 unsigned long rx_packets;
415 struct xdp_rxq_info xdp_rxq_info;
416 bool xdp_rxq_info_valid;
419 enum efx_sync_events_state {
420 SYNC_EVENTS_DISABLED = 0,
421 SYNC_EVENTS_QUIESCENT,
422 SYNC_EVENTS_REQUESTED,
427 * struct efx_channel - An Efx channel
429 * A channel comprises an event queue, at least one TX queue, at least
430 * one RX queue, and an associated tasklet for processing the event
433 * @efx: Associated Efx NIC
434 * @channel: Channel instance number
435 * @type: Channel type definition
436 * @eventq_init: Event queue initialised flag
437 * @enabled: Channel enabled indicator
438 * @irq: IRQ number (MSI and MSI-X only)
439 * @irq_moderation_us: IRQ moderation value (in microseconds)
440 * @napi_dev: Net device used with NAPI
441 * @napi_str: NAPI control structure
442 * @state: state for NAPI vs busy polling
443 * @state_lock: lock protecting @state
444 * @eventq: Event queue buffer
445 * @eventq_mask: Event queue pointer mask
446 * @eventq_read_ptr: Event queue read pointer
447 * @event_test_cpu: Last CPU to handle interrupt or test event for this channel
448 * @irq_count: Number of IRQs since last adaptive moderation decision
449 * @irq_mod_score: IRQ moderation score
450 * @rfs_filter_count: number of accelerated RFS filters currently in place;
451 * equals the count of @rps_flow_id slots filled
452 * @rfs_last_expiry: value of jiffies last time some accelerated RFS filters
453 * were checked for expiry
454 * @rfs_expire_index: next accelerated RFS filter ID to check for expiry
455 * @n_rfs_succeeded: number of successful accelerated RFS filter insertions
456 * @n_rfs_failed; number of failed accelerated RFS filter insertions
457 * @filter_work: Work item for efx_filter_rfs_expire()
458 * @rps_flow_id: Flow IDs of filters allocated for accelerated RFS,
459 * indexed by filter ID
460 * @n_rx_tobe_disc: Count of RX_TOBE_DISC errors
461 * @n_rx_ip_hdr_chksum_err: Count of RX IP header checksum errors
462 * @n_rx_tcp_udp_chksum_err: Count of RX TCP and UDP checksum errors
463 * @n_rx_mcast_mismatch: Count of unmatched multicast frames
464 * @n_rx_frm_trunc: Count of RX_FRM_TRUNC errors
465 * @n_rx_overlength: Count of RX_OVERLENGTH errors
466 * @n_skbuff_leaks: Count of skbuffs leaked due to RX overrun
467 * @n_rx_nodesc_trunc: Number of RX packets truncated and then dropped due to
468 * lack of descriptors
469 * @n_rx_merge_events: Number of RX merged completion events
470 * @n_rx_merge_packets: Number of RX packets completed by merged events
471 * @n_rx_xdp_drops: Count of RX packets intentionally dropped due to XDP
472 * @n_rx_xdp_bad_drops: Count of RX packets dropped due to XDP errors
473 * @n_rx_xdp_tx: Count of RX packets retransmitted due to XDP
474 * @n_rx_xdp_redirect: Count of RX packets redirected to a different NIC by XDP
475 * @rx_pkt_n_frags: Number of fragments in next packet to be delivered by
476 * __efx_rx_packet(), or zero if there is none
477 * @rx_pkt_index: Ring index of first buffer for next packet to be delivered
478 * by __efx_rx_packet(), if @rx_pkt_n_frags != 0
479 * @rx_list: list of SKBs from current RX, awaiting processing
480 * @rx_queue: RX queue for this channel
481 * @tx_queue: TX queues for this channel
482 * @sync_events_state: Current state of sync events on this channel
483 * @sync_timestamp_major: Major part of the last ptp sync event
484 * @sync_timestamp_minor: Minor part of the last ptp sync event
489 const struct efx_channel_type *type;
493 unsigned int irq_moderation_us;
494 struct net_device *napi_dev;
495 struct napi_struct napi_str;
496 #ifdef CONFIG_NET_RX_BUSY_POLL
497 unsigned long busy_poll_state;
499 struct efx_special_buffer eventq;
500 unsigned int eventq_mask;
501 unsigned int eventq_read_ptr;
504 unsigned int irq_count;
505 unsigned int irq_mod_score;
506 #ifdef CONFIG_RFS_ACCEL
507 unsigned int rfs_filter_count;
508 unsigned int rfs_last_expiry;
509 unsigned int rfs_expire_index;
510 unsigned int n_rfs_succeeded;
511 unsigned int n_rfs_failed;
512 struct delayed_work filter_work;
513 #define RPS_FLOW_ID_INVALID 0xFFFFFFFF
517 unsigned int n_rx_tobe_disc;
518 unsigned int n_rx_ip_hdr_chksum_err;
519 unsigned int n_rx_tcp_udp_chksum_err;
520 unsigned int n_rx_outer_ip_hdr_chksum_err;
521 unsigned int n_rx_outer_tcp_udp_chksum_err;
522 unsigned int n_rx_inner_ip_hdr_chksum_err;
523 unsigned int n_rx_inner_tcp_udp_chksum_err;
524 unsigned int n_rx_eth_crc_err;
525 unsigned int n_rx_mcast_mismatch;
526 unsigned int n_rx_frm_trunc;
527 unsigned int n_rx_overlength;
528 unsigned int n_skbuff_leaks;
529 unsigned int n_rx_nodesc_trunc;
530 unsigned int n_rx_merge_events;
531 unsigned int n_rx_merge_packets;
532 unsigned int n_rx_xdp_drops;
533 unsigned int n_rx_xdp_bad_drops;
534 unsigned int n_rx_xdp_tx;
535 unsigned int n_rx_xdp_redirect;
537 unsigned int rx_pkt_n_frags;
538 unsigned int rx_pkt_index;
540 struct list_head *rx_list;
542 struct efx_rx_queue rx_queue;
543 struct efx_tx_queue tx_queue[EFX_TXQ_TYPES];
545 enum efx_sync_events_state sync_events_state;
546 u32 sync_timestamp_major;
547 u32 sync_timestamp_minor;
551 * struct efx_msi_context - Context for each MSI
552 * @efx: The associated NIC
553 * @index: Index of the channel/IRQ
554 * @name: Name of the channel/IRQ
556 * Unlike &struct efx_channel, this is never reallocated and is always
557 * safe for the IRQ handler to access.
559 struct efx_msi_context {
562 char name[IFNAMSIZ + 6];
566 * struct efx_channel_type - distinguishes traffic and extra channels
567 * @handle_no_channel: Handle failure to allocate an extra channel
568 * @pre_probe: Set up extra state prior to initialisation
569 * @post_remove: Tear down extra state after finalisation, if allocated.
570 * May be called on channels that have not been probed.
571 * @get_name: Generate the channel's name (used for its IRQ handler)
572 * @copy: Copy the channel state prior to reallocation. May be %NULL if
573 * reallocation is not supported.
574 * @receive_skb: Handle an skb ready to be passed to netif_receive_skb()
575 * @want_txqs: Determine whether this channel should have TX queues
576 * created. If %NULL, TX queues are not created.
577 * @keep_eventq: Flag for whether event queue should be kept initialised
578 * while the device is stopped
579 * @want_pio: Flag for whether PIO buffers should be linked to this
580 * channel's TX queues.
582 struct efx_channel_type {
583 void (*handle_no_channel)(struct efx_nic *);
584 int (*pre_probe)(struct efx_channel *);
585 void (*post_remove)(struct efx_channel *);
586 void (*get_name)(struct efx_channel *, char *buf, size_t len);
587 struct efx_channel *(*copy)(const struct efx_channel *);
588 bool (*receive_skb)(struct efx_channel *, struct sk_buff *);
589 bool (*want_txqs)(struct efx_channel *);
600 #define STRING_TABLE_LOOKUP(val, member) \
601 ((val) < member ## _max) ? member ## _names[val] : "(invalid)"
603 extern const char *const efx_loopback_mode_names[];
604 extern const unsigned int efx_loopback_mode_max;
605 #define LOOPBACK_MODE(efx) \
606 STRING_TABLE_LOOKUP((efx)->loopback_mode, efx_loopback_mode)
608 extern const char *const efx_reset_type_names[];
609 extern const unsigned int efx_reset_type_max;
610 #define RESET_TYPE(type) \
611 STRING_TABLE_LOOKUP(type, efx_reset_type)
614 /* Be careful if altering to correct macro below */
615 EFX_INT_MODE_MSIX = 0,
616 EFX_INT_MODE_MSI = 1,
617 EFX_INT_MODE_LEGACY = 2,
618 EFX_INT_MODE_MAX /* Insert any new items before this */
620 #define EFX_INT_MODE_USE_MSI(x) (((x)->interrupt_mode) <= EFX_INT_MODE_MSI)
623 STATE_UNINIT = 0, /* device being probed/removed or is frozen */
624 STATE_READY = 1, /* hardware ready and netdev registered */
625 STATE_DISABLED = 2, /* device disabled due to hardware errors */
626 STATE_RECOVERY = 3, /* device recovering from PCI error */
629 /* Forward declaration */
632 /* Pseudo bit-mask flow control field */
633 #define EFX_FC_RX FLOW_CTRL_RX
634 #define EFX_FC_TX FLOW_CTRL_TX
635 #define EFX_FC_AUTO 4
638 * struct efx_link_state - Current state of the link
640 * @fd: Link is full-duplex
641 * @fc: Actual flow control flags
642 * @speed: Link speed (Mbps)
644 struct efx_link_state {
651 static inline bool efx_link_state_equal(const struct efx_link_state *left,
652 const struct efx_link_state *right)
654 return left->up == right->up && left->fd == right->fd &&
655 left->fc == right->fc && left->speed == right->speed;
659 * enum efx_phy_mode - PHY operating mode flags
660 * @PHY_MODE_NORMAL: on and should pass traffic
661 * @PHY_MODE_TX_DISABLED: on with TX disabled
662 * @PHY_MODE_LOW_POWER: set to low power through MDIO
663 * @PHY_MODE_OFF: switched off through external control
664 * @PHY_MODE_SPECIAL: on but will not pass traffic
668 PHY_MODE_TX_DISABLED = 1,
669 PHY_MODE_LOW_POWER = 2,
671 PHY_MODE_SPECIAL = 8,
674 static inline bool efx_phy_mode_disabled(enum efx_phy_mode mode)
676 return !!(mode & ~PHY_MODE_TX_DISABLED);
680 * struct efx_hw_stat_desc - Description of a hardware statistic
681 * @name: Name of the statistic as visible through ethtool, or %NULL if
682 * it should not be exposed
683 * @dma_width: Width in bits (0 for non-DMA statistics)
684 * @offset: Offset within stats (ignored for non-DMA statistics)
686 struct efx_hw_stat_desc {
692 /* Number of bits used in a multicast filter hash address */
693 #define EFX_MCAST_HASH_BITS 8
695 /* Number of (single-bit) entries in a multicast filter hash */
696 #define EFX_MCAST_HASH_ENTRIES (1 << EFX_MCAST_HASH_BITS)
698 /* An Efx multicast filter hash */
699 union efx_multicast_hash {
700 u8 byte[EFX_MCAST_HASH_ENTRIES / 8];
701 efx_oword_t oword[EFX_MCAST_HASH_ENTRIES / sizeof(efx_oword_t) / 8];
706 /* The reserved RSS context value */
707 #define EFX_MCDI_RSS_CONTEXT_INVALID 0xffffffff
709 * struct efx_rss_context - A user-defined RSS context for filtering
710 * @list: node of linked list on which this struct is stored
711 * @context_id: the RSS_CONTEXT_ID returned by MC firmware, or
712 * %EFX_MCDI_RSS_CONTEXT_INVALID if this context is not present on the NIC.
713 * For Siena, 0 if RSS is active, else %EFX_MCDI_RSS_CONTEXT_INVALID.
714 * @user_id: the rss_context ID exposed to userspace over ethtool.
715 * @rx_hash_udp_4tuple: UDP 4-tuple hashing enabled
716 * @rx_hash_key: Toeplitz hash key for this RSS context
717 * @indir_table: Indirection table for this RSS context
719 struct efx_rss_context {
720 struct list_head list;
723 bool rx_hash_udp_4tuple;
725 u32 rx_indir_table[128];
728 #ifdef CONFIG_RFS_ACCEL
729 /* Order of these is important, since filter_id >= %EFX_ARFS_FILTER_ID_PENDING
730 * is used to test if filter does or will exist.
732 #define EFX_ARFS_FILTER_ID_PENDING -1
733 #define EFX_ARFS_FILTER_ID_ERROR -2
734 #define EFX_ARFS_FILTER_ID_REMOVING -3
736 * struct efx_arfs_rule - record of an ARFS filter and its IDs
737 * @node: linkage into hash table
738 * @spec: details of the filter (used as key for hash table). Use efx->type to
739 * determine which member to use.
740 * @rxq_index: channel to which the filter will steer traffic.
741 * @arfs_id: filter ID which was returned to ARFS
742 * @filter_id: index in software filter table. May be
743 * %EFX_ARFS_FILTER_ID_PENDING if filter was not inserted yet,
744 * %EFX_ARFS_FILTER_ID_ERROR if filter insertion failed, or
745 * %EFX_ARFS_FILTER_ID_REMOVING if expiry is currently removing the filter.
747 struct efx_arfs_rule {
748 struct hlist_node node;
749 struct efx_filter_spec spec;
755 /* Size chosen so that the table is one page (4kB) */
756 #define EFX_ARFS_HASH_TABLE_SIZE 512
759 * struct efx_async_filter_insertion - Request to asynchronously insert a filter
760 * @net_dev: Reference to the netdevice
761 * @spec: The filter to insert
762 * @work: Workitem for this request
763 * @rxq_index: Identifies the channel for which this request was made
764 * @flow_id: Identifies the kernel-side flow for which this request was made
766 struct efx_async_filter_insertion {
767 struct net_device *net_dev;
768 struct efx_filter_spec spec;
769 struct work_struct work;
774 /* Maximum number of ARFS workitems that may be in flight on an efx_nic */
775 #define EFX_RPS_MAX_IN_FLIGHT 8
776 #endif /* CONFIG_RFS_ACCEL */
779 * struct efx_nic - an Efx NIC
780 * @name: Device name (net device name or bus id before net device registered)
781 * @pci_dev: The PCI device
782 * @node: List node for maintaning primary/secondary function lists
783 * @primary: &struct efx_nic instance for the primary function of this
784 * controller. May be the same structure, and may be %NULL if no
785 * primary function is bound. Serialised by rtnl_lock.
786 * @secondary_list: List of &struct efx_nic instances for the secondary PCI
787 * functions of the controller, if this is for the primary function.
788 * Serialised by rtnl_lock.
789 * @type: Controller type attributes
790 * @legacy_irq: IRQ number
791 * @workqueue: Workqueue for port reconfigures and the HW monitor.
792 * Work items do not hold and must not acquire RTNL.
793 * @workqueue_name: Name of workqueue
794 * @reset_work: Scheduled reset workitem
795 * @membase_phys: Memory BAR value as physical address
796 * @membase: Memory BAR value
797 * @vi_stride: step between per-VI registers / memory regions
798 * @interrupt_mode: Interrupt mode
799 * @timer_quantum_ns: Interrupt timer quantum, in nanoseconds
800 * @timer_max_ns: Interrupt timer maximum value, in nanoseconds
801 * @irq_rx_adaptive: Adaptive IRQ moderation enabled for RX event queues
802 * @irqs_hooked: Channel interrupts are hooked
803 * @irq_rx_mod_step_us: Step size for IRQ moderation for RX event queues
804 * @irq_rx_moderation_us: IRQ moderation time for RX event queues
805 * @msg_enable: Log message enable flags
806 * @state: Device state number (%STATE_*). Serialised by the rtnl_lock.
807 * @reset_pending: Bitmask for pending resets
808 * @tx_queue: TX DMA queues
809 * @rx_queue: RX DMA queues
811 * @msi_context: Context for each MSI
812 * @extra_channel_types: Types of extra (non-traffic) channels that
813 * should be allocated for this NIC
814 * @xdp_tx_queue_count: Number of entries in %xdp_tx_queues.
815 * @xdp_tx_queues: Array of pointers to tx queues used for XDP transmit.
816 * @rxq_entries: Size of receive queues requested by user.
817 * @txq_entries: Size of transmit queues requested by user.
818 * @txq_stop_thresh: TX queue fill level at or above which we stop it.
819 * @txq_wake_thresh: TX queue fill level at or below which we wake it.
820 * @tx_dc_base: Base qword address in SRAM of TX queue descriptor caches
821 * @rx_dc_base: Base qword address in SRAM of RX queue descriptor caches
822 * @sram_lim_qw: Qword address limit of SRAM
823 * @next_buffer_table: First available buffer table id
824 * @n_channels: Number of channels in use
825 * @n_rx_channels: Number of channels used for RX (= number of RX queues)
826 * @n_tx_channels: Number of channels used for TX
827 * @n_extra_tx_channels: Number of extra channels with TX queues
828 * @tx_queues_per_channel: number of TX queues probed on each channel
829 * @n_xdp_channels: Number of channels used for XDP TX
830 * @xdp_channel_offset: Offset of zeroth channel used for XPD TX.
831 * @xdp_tx_per_channel: Max number of TX queues on an XDP TX channel.
832 * @rx_ip_align: RX DMA address offset to have IP header aligned in
833 * in accordance with NET_IP_ALIGN
834 * @rx_dma_len: Current maximum RX DMA length
835 * @rx_buffer_order: Order (log2) of number of pages for each RX buffer
836 * @rx_buffer_truesize: Amortised allocation size of an RX buffer,
837 * for use in sk_buff::truesize
838 * @rx_prefix_size: Size of RX prefix before packet data
839 * @rx_packet_hash_offset: Offset of RX flow hash from start of packet data
840 * (valid only if @rx_prefix_size != 0; always negative)
841 * @rx_packet_len_offset: Offset of RX packet length from start of packet data
842 * (valid only for NICs that set %EFX_RX_PKT_PREFIX_LEN; always negative)
843 * @rx_packet_ts_offset: Offset of timestamp from start of packet data
844 * (valid only if channel->sync_timestamps_enabled; always negative)
845 * @rx_scatter: Scatter mode enabled for receives
846 * @rss_context: Main RSS context. Its @list member is the head of the list of
847 * RSS contexts created by user requests
848 * @rss_lock: Protects custom RSS context software state in @rss_context.list
849 * @vport_id: The function's vport ID, only relevant for PFs
850 * @int_error_count: Number of internal errors seen recently
851 * @int_error_expire: Time at which error count will be expired
852 * @must_realloc_vis: Flag: VIs have yet to be reallocated after MC reboot
853 * @irq_soft_enabled: Are IRQs soft-enabled? If not, IRQ handler will
854 * acknowledge but do nothing else.
855 * @irq_status: Interrupt status buffer
856 * @irq_zero_count: Number of legacy IRQs seen with queue flags == 0
857 * @irq_level: IRQ level/index for IRQs not triggered by an event queue
858 * @selftest_work: Work item for asynchronous self-test
859 * @mtd_list: List of MTDs attached to the NIC
860 * @nic_data: Hardware dependent state
861 * @mcdi: Management-Controller-to-Driver Interface state
862 * @mac_lock: MAC access lock. Protects @port_enabled, @phy_mode,
863 * efx_monitor() and efx_reconfigure_port()
864 * @port_enabled: Port enabled indicator.
865 * Serialises efx_stop_all(), efx_start_all(), efx_monitor() and
866 * efx_mac_work() with kernel interfaces. Safe to read under any
867 * one of the rtnl_lock, mac_lock, or netif_tx_lock, but all three must
868 * be held to modify it.
869 * @port_initialized: Port initialized?
870 * @net_dev: Operating system network device. Consider holding the rtnl lock
871 * @fixed_features: Features which cannot be turned off
872 * @num_mac_stats: Number of MAC stats reported by firmware (MAC_STATS_NUM_STATS
873 * field of %MC_CMD_GET_CAPABILITIES_V4 response, or %MC_CMD_MAC_NSTATS)
874 * @stats_buffer: DMA buffer for statistics
875 * @phy_type: PHY type
876 * @phy_data: PHY private data (including PHY-specific stats)
877 * @mdio: PHY MDIO interface
878 * @mdio_bus: PHY MDIO bus ID (only used by Siena)
879 * @phy_mode: PHY operating mode. Serialised by @mac_lock.
880 * @link_advertising: Autonegotiation advertising flags
881 * @fec_config: Forward Error Correction configuration flags. For bit positions
882 * see &enum ethtool_fec_config_bits.
883 * @link_state: Current state of the link
884 * @n_link_state_changes: Number of times the link has changed state
885 * @unicast_filter: Flag for Falcon-arch simple unicast filter.
886 * Protected by @mac_lock.
887 * @multicast_hash: Multicast hash table for Falcon-arch.
888 * Protected by @mac_lock.
889 * @wanted_fc: Wanted flow control flags
890 * @fc_disable: When non-zero flow control is disabled. Typically used to
891 * ensure that network back pressure doesn't delay dma queue flushes.
892 * Serialised by the rtnl lock.
893 * @mac_work: Work item for changing MAC promiscuity and multicast hash
894 * @loopback_mode: Loopback status
895 * @loopback_modes: Supported loopback mode bitmask
896 * @loopback_selftest: Offline self-test private state
897 * @xdp_prog: Current XDP programme for this interface
898 * @filter_sem: Filter table rw_semaphore, protects existence of @filter_state
899 * @filter_state: Architecture-dependent filter table state
900 * @rps_mutex: Protects RPS state of all channels
901 * @rps_slot_map: bitmap of in-flight entries in @rps_slot
902 * @rps_slot: array of ARFS insertion requests for efx_filter_rfs_work()
903 * @rps_hash_lock: Protects ARFS filter mapping state (@rps_hash_table and
905 * @rps_hash_table: Mapping between ARFS filters and their various IDs
906 * @rps_next_id: next arfs_id for an ARFS filter
907 * @active_queues: Count of RX and TX queues that haven't been flushed and drained.
908 * @rxq_flush_pending: Count of number of receive queues that need to be flushed.
909 * Decremented when the efx_flush_rx_queue() is called.
910 * @rxq_flush_outstanding: Count of number of RX flushes started but not yet
911 * completed (either success or failure). Not used when MCDI is used to
912 * flush receive queues.
913 * @flush_wq: wait queue used by efx_nic_flush_queues() to wait for flush completions.
914 * @vf_count: Number of VFs intended to be enabled.
915 * @vf_init_count: Number of VFs that have been fully initialised.
916 * @vi_scale: log2 number of vnics per VF.
917 * @ptp_data: PTP state data
918 * @ptp_warned: has this NIC seen and warned about unexpected PTP events?
919 * @vpd_sn: Serial number read from VPD
920 * @xdp_rxq_info_failed: Have any of the rx queues failed to initialise their
921 * xdp_rxq_info structures?
922 * @netdev_notifier: Netdevice notifier.
923 * @mem_bar: The BAR that is mapped into membase.
924 * @reg_base: Offset from the start of the bar to the function control window.
925 * @monitor_work: Hardware monitor workitem
926 * @biu_lock: BIU (bus interface unit) lock
927 * @last_irq_cpu: Last CPU to handle a possible test interrupt. This
928 * field is used by efx_test_interrupts() to verify that an
929 * interrupt has occurred.
930 * @stats_lock: Statistics update lock. Must be held when calling
931 * efx_nic_type::{update,start,stop}_stats.
932 * @n_rx_noskb_drops: Count of RX packets dropped due to failure to allocate an skb
934 * This is stored in the private area of the &struct net_device.
937 /* The following fields should be written very rarely */
940 struct list_head node;
941 struct efx_nic *primary;
942 struct list_head secondary_list;
943 struct pci_dev *pci_dev;
944 unsigned int port_num;
945 const struct efx_nic_type *type;
947 bool eeh_disabled_legacy_irq;
948 struct workqueue_struct *workqueue;
949 char workqueue_name[16];
950 struct work_struct reset_work;
951 resource_size_t membase_phys;
952 void __iomem *membase;
954 unsigned int vi_stride;
956 enum efx_int_mode interrupt_mode;
957 unsigned int timer_quantum_ns;
958 unsigned int timer_max_ns;
959 bool irq_rx_adaptive;
961 unsigned int irq_mod_step_us;
962 unsigned int irq_rx_moderation_us;
965 enum nic_state state;
966 unsigned long reset_pending;
968 struct efx_channel *channel[EFX_MAX_CHANNELS];
969 struct efx_msi_context msi_context[EFX_MAX_CHANNELS];
970 const struct efx_channel_type *
971 extra_channel_type[EFX_MAX_EXTRA_CHANNELS];
973 unsigned int xdp_tx_queue_count;
974 struct efx_tx_queue **xdp_tx_queues;
976 unsigned rxq_entries;
977 unsigned txq_entries;
978 unsigned int txq_stop_thresh;
979 unsigned int txq_wake_thresh;
983 unsigned sram_lim_qw;
984 unsigned next_buffer_table;
986 unsigned int max_channels;
987 unsigned int max_vis;
988 unsigned int max_tx_channels;
990 unsigned n_rx_channels;
992 unsigned tx_channel_offset;
993 unsigned n_tx_channels;
994 unsigned n_extra_tx_channels;
995 unsigned int tx_queues_per_channel;
996 unsigned int n_xdp_channels;
997 unsigned int xdp_channel_offset;
998 unsigned int xdp_tx_per_channel;
999 unsigned int rx_ip_align;
1000 unsigned int rx_dma_len;
1001 unsigned int rx_buffer_order;
1002 unsigned int rx_buffer_truesize;
1003 unsigned int rx_page_buf_step;
1004 unsigned int rx_bufs_per_page;
1005 unsigned int rx_pages_per_batch;
1006 unsigned int rx_prefix_size;
1007 int rx_packet_hash_offset;
1008 int rx_packet_len_offset;
1009 int rx_packet_ts_offset;
1011 struct efx_rss_context rss_context;
1012 struct mutex rss_lock;
1015 unsigned int_error_count;
1016 unsigned long int_error_expire;
1018 bool must_realloc_vis;
1019 bool irq_soft_enabled;
1020 struct efx_buffer irq_status;
1021 unsigned irq_zero_count;
1023 struct delayed_work selftest_work;
1025 #ifdef CONFIG_SFC_MTD
1026 struct list_head mtd_list;
1030 struct efx_mcdi_data *mcdi;
1032 struct mutex mac_lock;
1033 struct work_struct mac_work;
1036 bool mc_bist_for_other_fn;
1037 bool port_initialized;
1038 struct net_device *net_dev;
1040 netdev_features_t fixed_features;
1043 struct efx_buffer stats_buffer;
1044 u64 rx_nodesc_drops_total;
1045 u64 rx_nodesc_drops_while_down;
1046 bool rx_nodesc_drops_prev_state;
1048 unsigned int phy_type;
1050 struct mdio_if_info mdio;
1051 unsigned int mdio_bus;
1052 enum efx_phy_mode phy_mode;
1054 __ETHTOOL_DECLARE_LINK_MODE_MASK(link_advertising);
1056 struct efx_link_state link_state;
1057 unsigned int n_link_state_changes;
1059 bool unicast_filter;
1060 union efx_multicast_hash multicast_hash;
1062 unsigned fc_disable;
1065 enum efx_loopback_mode loopback_mode;
1068 void *loopback_selftest;
1069 /* We access loopback_selftest immediately before running XDP,
1070 * so we want them next to each other.
1072 struct bpf_prog __rcu *xdp_prog;
1074 struct rw_semaphore filter_sem;
1076 #ifdef CONFIG_RFS_ACCEL
1077 struct mutex rps_mutex;
1078 unsigned long rps_slot_map;
1079 struct efx_async_filter_insertion rps_slot[EFX_RPS_MAX_IN_FLIGHT];
1080 spinlock_t rps_hash_lock;
1081 struct hlist_head *rps_hash_table;
1085 atomic_t active_queues;
1086 atomic_t rxq_flush_pending;
1087 atomic_t rxq_flush_outstanding;
1088 wait_queue_head_t flush_wq;
1090 #ifdef CONFIG_SFC_SRIOV
1092 unsigned vf_init_count;
1096 struct efx_ptp_data *ptp_data;
1100 bool xdp_rxq_info_failed;
1102 struct notifier_block netdev_notifier;
1104 unsigned int mem_bar;
1107 /* The following fields may be written more often */
1109 struct delayed_work monitor_work ____cacheline_aligned_in_smp;
1110 spinlock_t biu_lock;
1112 spinlock_t stats_lock;
1113 atomic_t n_rx_noskb_drops;
1116 static inline int efx_dev_registered(struct efx_nic *efx)
1118 return efx->net_dev->reg_state == NETREG_REGISTERED;
1121 static inline unsigned int efx_port_num(struct efx_nic *efx)
1123 return efx->port_num;
1126 struct efx_mtd_partition {
1127 struct list_head node;
1128 struct mtd_info mtd;
1129 const char *dev_type_name;
1130 const char *type_name;
1131 char name[IFNAMSIZ + 20];
1134 struct efx_udp_tunnel {
1135 #define TUNNEL_ENCAP_UDP_PORT_ENTRY_INVALID 0xffff
1136 u16 type; /* TUNNEL_ENCAP_UDP_PORT_ENTRY_foo, see mcdi_pcol.h */
1141 * struct efx_nic_type - Efx device type definition
1142 * @mem_bar: Get the memory BAR
1143 * @mem_map_size: Get memory BAR mapped size
1144 * @probe: Probe the controller
1145 * @remove: Free resources allocated by probe()
1146 * @init: Initialise the controller
1147 * @dimension_resources: Dimension controller resources (buffer table,
1148 * and VIs once the available interrupt resources are clear)
1149 * @fini: Shut down the controller
1150 * @monitor: Periodic function for polling link state and hardware monitor
1151 * @map_reset_reason: Map ethtool reset reason to a reset method
1152 * @map_reset_flags: Map ethtool reset flags to a reset method, if possible
1153 * @reset: Reset the controller hardware and possibly the PHY. This will
1154 * be called while the controller is uninitialised.
1155 * @probe_port: Probe the MAC and PHY
1156 * @remove_port: Free resources allocated by probe_port()
1157 * @handle_global_event: Handle a "global" event (may be %NULL)
1158 * @fini_dmaq: Flush and finalise DMA queues (RX and TX queues)
1159 * @prepare_flush: Prepare the hardware for flushing the DMA queues
1160 * (for Falcon architecture)
1161 * @finish_flush: Clean up after flushing the DMA queues (for Falcon
1163 * @prepare_flr: Prepare for an FLR
1164 * @finish_flr: Clean up after an FLR
1165 * @describe_stats: Describe statistics for ethtool
1166 * @update_stats: Update statistics not provided by event handling.
1167 * Either argument may be %NULL.
1168 * @start_stats: Start the regular fetching of statistics
1169 * @pull_stats: Pull stats from the NIC and wait until they arrive.
1170 * @stop_stats: Stop the regular fetching of statistics
1171 * @push_irq_moderation: Apply interrupt moderation value
1172 * @reconfigure_port: Push loopback/power/txdis changes to the MAC and PHY
1173 * @prepare_enable_fc_tx: Prepare MAC to enable pause frame TX (may be %NULL)
1174 * @reconfigure_mac: Push MAC address, MTU, flow control and filter settings
1175 * to the hardware. Serialised by the mac_lock.
1176 * @check_mac_fault: Check MAC fault state. True if fault present.
1177 * @get_wol: Get WoL configuration from driver state
1178 * @set_wol: Push WoL configuration to the NIC
1179 * @resume_wol: Synchronise WoL state between driver and MC (e.g. after resume)
1180 * @test_chip: Test registers. May use efx_farch_test_registers(), and is
1181 * expected to reset the NIC.
1182 * @test_nvram: Test validity of NVRAM contents
1183 * @mcdi_request: Send an MCDI request with the given header and SDU.
1184 * The SDU length may be any value from 0 up to the protocol-
1185 * defined maximum, but its buffer will be padded to a multiple
1187 * @mcdi_poll_response: Test whether an MCDI response is available.
1188 * @mcdi_read_response: Read the MCDI response PDU. The offset will
1189 * be a multiple of 4. The length may not be, but the buffer
1190 * will be padded so it is safe to round up.
1191 * @mcdi_poll_reboot: Test whether the MCDI has rebooted. If so,
1192 * return an appropriate error code for aborting any current
1193 * request; otherwise return 0.
1194 * @irq_enable_master: Enable IRQs on the NIC. Each event queue must
1195 * be separately enabled after this.
1196 * @irq_test_generate: Generate a test IRQ
1197 * @irq_disable_non_ev: Disable non-event IRQs on the NIC. Each event
1198 * queue must be separately disabled before this.
1199 * @irq_handle_msi: Handle MSI for a channel. The @dev_id argument is
1200 * a pointer to the &struct efx_msi_context for the channel.
1201 * @irq_handle_legacy: Handle legacy interrupt. The @dev_id argument
1202 * is a pointer to the &struct efx_nic.
1203 * @tx_probe: Allocate resources for TX queue
1204 * @tx_init: Initialise TX queue on the NIC
1205 * @tx_remove: Free resources for TX queue
1206 * @tx_write: Write TX descriptors and doorbell
1207 * @tx_enqueue: Add an SKB to TX queue
1208 * @rx_push_rss_config: Write RSS hash key and indirection table to the NIC
1209 * @rx_pull_rss_config: Read RSS hash key and indirection table back from the NIC
1210 * @rx_push_rss_context_config: Write RSS hash key and indirection table for
1211 * user RSS context to the NIC
1212 * @rx_pull_rss_context_config: Read RSS hash key and indirection table for user
1213 * RSS context back from the NIC
1214 * @rx_probe: Allocate resources for RX queue
1215 * @rx_init: Initialise RX queue on the NIC
1216 * @rx_remove: Free resources for RX queue
1217 * @rx_write: Write RX descriptors and doorbell
1218 * @rx_defer_refill: Generate a refill reminder event
1219 * @rx_packet: Receive the queued RX buffer on a channel
1220 * @rx_buf_hash_valid: Determine whether the RX prefix contains a valid hash
1221 * @ev_probe: Allocate resources for event queue
1222 * @ev_init: Initialise event queue on the NIC
1223 * @ev_fini: Deinitialise event queue on the NIC
1224 * @ev_remove: Free resources for event queue
1225 * @ev_process: Process events for a queue, up to the given NAPI quota
1226 * @ev_read_ack: Acknowledge read events on a queue, rearming its IRQ
1227 * @ev_test_generate: Generate a test event
1228 * @filter_table_probe: Probe filter capabilities and set up filter software state
1229 * @filter_table_restore: Restore filters removed from hardware
1230 * @filter_table_remove: Remove filters from hardware and tear down software state
1231 * @filter_update_rx_scatter: Update filters after change to rx scatter setting
1232 * @filter_insert: add or replace a filter
1233 * @filter_remove_safe: remove a filter by ID, carefully
1234 * @filter_get_safe: retrieve a filter by ID, carefully
1235 * @filter_clear_rx: Remove all RX filters whose priority is less than or
1236 * equal to the given priority and is not %EFX_FILTER_PRI_AUTO
1237 * @filter_count_rx_used: Get the number of filters in use at a given priority
1238 * @filter_get_rx_id_limit: Get maximum value of a filter id, plus 1
1239 * @filter_get_rx_ids: Get list of RX filters at a given priority
1240 * @filter_rfs_expire_one: Consider expiring a filter inserted for RFS.
1241 * This must check whether the specified table entry is used by RFS
1242 * and that rps_may_expire_flow() returns true for it.
1243 * @mtd_probe: Probe and add MTD partitions associated with this net device,
1244 * using efx_mtd_add()
1245 * @mtd_rename: Set an MTD partition name using the net device name
1246 * @mtd_read: Read from an MTD partition
1247 * @mtd_erase: Erase part of an MTD partition
1248 * @mtd_write: Write to an MTD partition
1249 * @mtd_sync: Wait for write-back to complete on MTD partition. This
1250 * also notifies the driver that a writer has finished using this
1252 * @ptp_write_host_time: Send host time to MC as part of sync protocol
1253 * @ptp_set_ts_sync_events: Enable or disable sync events for inline RX
1254 * timestamping, possibly only temporarily for the purposes of a reset.
1255 * @ptp_set_ts_config: Set hardware timestamp configuration. The flags
1256 * and tx_type will already have been validated but this operation
1257 * must validate and update rx_filter.
1258 * @get_phys_port_id: Get the underlying physical port id.
1259 * @set_mac_address: Set the MAC address of the device
1260 * @tso_versions: Returns mask of firmware-assisted TSO versions supported.
1261 * If %NULL, then device does not support any TSO version.
1262 * @udp_tnl_push_ports: Push the list of UDP tunnel ports to the NIC if required.
1263 * @udp_tnl_has_port: Check if a port has been added as UDP tunnel
1264 * @print_additional_fwver: Dump NIC-specific additional FW version info
1265 * @sensor_event: Handle a sensor event from MCDI
1266 * @revision: Hardware architecture revision
1267 * @txd_ptr_tbl_base: TX descriptor ring base address
1268 * @rxd_ptr_tbl_base: RX descriptor ring base address
1269 * @buf_tbl_base: Buffer table base address
1270 * @evq_ptr_tbl_base: Event queue pointer table base address
1271 * @evq_rptr_tbl_base: Event queue read-pointer table base address
1272 * @max_dma_mask: Maximum possible DMA mask
1273 * @rx_prefix_size: Size of RX prefix before packet data
1274 * @rx_hash_offset: Offset of RX flow hash within prefix
1275 * @rx_ts_offset: Offset of timestamp within prefix
1276 * @rx_buffer_padding: Size of padding at end of RX packet
1277 * @can_rx_scatter: NIC is able to scatter packets to multiple buffers
1278 * @always_rx_scatter: NIC will always scatter packets to multiple buffers
1279 * @option_descriptors: NIC supports TX option descriptors
1280 * @min_interrupt_mode: Lowest capability interrupt mode supported
1281 * from &enum efx_int_mode.
1282 * @timer_period_max: Maximum period of interrupt timer (in ticks)
1283 * @offload_features: net_device feature flags for protocol offload
1284 * features implemented in hardware
1285 * @mcdi_max_ver: Maximum MCDI version supported
1286 * @hwtstamp_filters: Mask of hardware timestamp filter types supported
1288 struct efx_nic_type {
1290 unsigned int (*mem_bar)(struct efx_nic *efx);
1291 unsigned int (*mem_map_size)(struct efx_nic *efx);
1292 int (*probe)(struct efx_nic *efx);
1293 void (*remove)(struct efx_nic *efx);
1294 int (*init)(struct efx_nic *efx);
1295 int (*dimension_resources)(struct efx_nic *efx);
1296 void (*fini)(struct efx_nic *efx);
1297 void (*monitor)(struct efx_nic *efx);
1298 enum reset_type (*map_reset_reason)(enum reset_type reason);
1299 int (*map_reset_flags)(u32 *flags);
1300 int (*reset)(struct efx_nic *efx, enum reset_type method);
1301 int (*probe_port)(struct efx_nic *efx);
1302 void (*remove_port)(struct efx_nic *efx);
1303 bool (*handle_global_event)(struct efx_channel *channel, efx_qword_t *);
1304 int (*fini_dmaq)(struct efx_nic *efx);
1305 void (*prepare_flush)(struct efx_nic *efx);
1306 void (*finish_flush)(struct efx_nic *efx);
1307 void (*prepare_flr)(struct efx_nic *efx);
1308 void (*finish_flr)(struct efx_nic *efx);
1309 size_t (*describe_stats)(struct efx_nic *efx, u8 *names);
1310 size_t (*update_stats)(struct efx_nic *efx, u64 *full_stats,
1311 struct rtnl_link_stats64 *core_stats);
1312 void (*start_stats)(struct efx_nic *efx);
1313 void (*pull_stats)(struct efx_nic *efx);
1314 void (*stop_stats)(struct efx_nic *efx);
1315 void (*push_irq_moderation)(struct efx_channel *channel);
1316 int (*reconfigure_port)(struct efx_nic *efx);
1317 void (*prepare_enable_fc_tx)(struct efx_nic *efx);
1318 int (*reconfigure_mac)(struct efx_nic *efx, bool mtu_only);
1319 bool (*check_mac_fault)(struct efx_nic *efx);
1320 void (*get_wol)(struct efx_nic *efx, struct ethtool_wolinfo *wol);
1321 int (*set_wol)(struct efx_nic *efx, u32 type);
1322 void (*resume_wol)(struct efx_nic *efx);
1323 unsigned int (*check_caps)(const struct efx_nic *efx,
1326 int (*test_chip)(struct efx_nic *efx, struct efx_self_tests *tests);
1327 int (*test_nvram)(struct efx_nic *efx);
1328 void (*mcdi_request)(struct efx_nic *efx,
1329 const efx_dword_t *hdr, size_t hdr_len,
1330 const efx_dword_t *sdu, size_t sdu_len);
1331 bool (*mcdi_poll_response)(struct efx_nic *efx);
1332 void (*mcdi_read_response)(struct efx_nic *efx, efx_dword_t *pdu,
1333 size_t pdu_offset, size_t pdu_len);
1334 int (*mcdi_poll_reboot)(struct efx_nic *efx);
1335 void (*mcdi_reboot_detected)(struct efx_nic *efx);
1336 void (*irq_enable_master)(struct efx_nic *efx);
1337 int (*irq_test_generate)(struct efx_nic *efx);
1338 void (*irq_disable_non_ev)(struct efx_nic *efx);
1339 irqreturn_t (*irq_handle_msi)(int irq, void *dev_id);
1340 irqreturn_t (*irq_handle_legacy)(int irq, void *dev_id);
1341 int (*tx_probe)(struct efx_tx_queue *tx_queue);
1342 void (*tx_init)(struct efx_tx_queue *tx_queue);
1343 void (*tx_remove)(struct efx_tx_queue *tx_queue);
1344 void (*tx_write)(struct efx_tx_queue *tx_queue);
1345 netdev_tx_t (*tx_enqueue)(struct efx_tx_queue *tx_queue, struct sk_buff *skb);
1346 unsigned int (*tx_limit_len)(struct efx_tx_queue *tx_queue,
1347 dma_addr_t dma_addr, unsigned int len);
1348 int (*rx_push_rss_config)(struct efx_nic *efx, bool user,
1349 const u32 *rx_indir_table, const u8 *key);
1350 int (*rx_pull_rss_config)(struct efx_nic *efx);
1351 int (*rx_push_rss_context_config)(struct efx_nic *efx,
1352 struct efx_rss_context *ctx,
1353 const u32 *rx_indir_table,
1355 int (*rx_pull_rss_context_config)(struct efx_nic *efx,
1356 struct efx_rss_context *ctx);
1357 void (*rx_restore_rss_contexts)(struct efx_nic *efx);
1358 int (*rx_probe)(struct efx_rx_queue *rx_queue);
1359 void (*rx_init)(struct efx_rx_queue *rx_queue);
1360 void (*rx_remove)(struct efx_rx_queue *rx_queue);
1361 void (*rx_write)(struct efx_rx_queue *rx_queue);
1362 void (*rx_defer_refill)(struct efx_rx_queue *rx_queue);
1363 void (*rx_packet)(struct efx_channel *channel);
1364 bool (*rx_buf_hash_valid)(const u8 *prefix);
1365 int (*ev_probe)(struct efx_channel *channel);
1366 int (*ev_init)(struct efx_channel *channel);
1367 void (*ev_fini)(struct efx_channel *channel);
1368 void (*ev_remove)(struct efx_channel *channel);
1369 int (*ev_process)(struct efx_channel *channel, int quota);
1370 void (*ev_read_ack)(struct efx_channel *channel);
1371 void (*ev_test_generate)(struct efx_channel *channel);
1372 int (*filter_table_probe)(struct efx_nic *efx);
1373 void (*filter_table_restore)(struct efx_nic *efx);
1374 void (*filter_table_remove)(struct efx_nic *efx);
1375 void (*filter_update_rx_scatter)(struct efx_nic *efx);
1376 s32 (*filter_insert)(struct efx_nic *efx,
1377 struct efx_filter_spec *spec, bool replace);
1378 int (*filter_remove_safe)(struct efx_nic *efx,
1379 enum efx_filter_priority priority,
1381 int (*filter_get_safe)(struct efx_nic *efx,
1382 enum efx_filter_priority priority,
1383 u32 filter_id, struct efx_filter_spec *);
1384 int (*filter_clear_rx)(struct efx_nic *efx,
1385 enum efx_filter_priority priority);
1386 u32 (*filter_count_rx_used)(struct efx_nic *efx,
1387 enum efx_filter_priority priority);
1388 u32 (*filter_get_rx_id_limit)(struct efx_nic *efx);
1389 s32 (*filter_get_rx_ids)(struct efx_nic *efx,
1390 enum efx_filter_priority priority,
1391 u32 *buf, u32 size);
1392 #ifdef CONFIG_RFS_ACCEL
1393 bool (*filter_rfs_expire_one)(struct efx_nic *efx, u32 flow_id,
1394 unsigned int index);
1396 #ifdef CONFIG_SFC_MTD
1397 int (*mtd_probe)(struct efx_nic *efx);
1398 void (*mtd_rename)(struct efx_mtd_partition *part);
1399 int (*mtd_read)(struct mtd_info *mtd, loff_t start, size_t len,
1400 size_t *retlen, u8 *buffer);
1401 int (*mtd_erase)(struct mtd_info *mtd, loff_t start, size_t len);
1402 int (*mtd_write)(struct mtd_info *mtd, loff_t start, size_t len,
1403 size_t *retlen, const u8 *buffer);
1404 int (*mtd_sync)(struct mtd_info *mtd);
1406 void (*ptp_write_host_time)(struct efx_nic *efx, u32 host_time);
1407 int (*ptp_set_ts_sync_events)(struct efx_nic *efx, bool en, bool temp);
1408 int (*ptp_set_ts_config)(struct efx_nic *efx,
1409 struct hwtstamp_config *init);
1410 int (*sriov_configure)(struct efx_nic *efx, int num_vfs);
1411 int (*vlan_rx_add_vid)(struct efx_nic *efx, __be16 proto, u16 vid);
1412 int (*vlan_rx_kill_vid)(struct efx_nic *efx, __be16 proto, u16 vid);
1413 int (*get_phys_port_id)(struct efx_nic *efx,
1414 struct netdev_phys_item_id *ppid);
1415 int (*sriov_init)(struct efx_nic *efx);
1416 void (*sriov_fini)(struct efx_nic *efx);
1417 bool (*sriov_wanted)(struct efx_nic *efx);
1418 void (*sriov_reset)(struct efx_nic *efx);
1419 void (*sriov_flr)(struct efx_nic *efx, unsigned vf_i);
1420 int (*sriov_set_vf_mac)(struct efx_nic *efx, int vf_i, u8 *mac);
1421 int (*sriov_set_vf_vlan)(struct efx_nic *efx, int vf_i, u16 vlan,
1423 int (*sriov_set_vf_spoofchk)(struct efx_nic *efx, int vf_i,
1425 int (*sriov_get_vf_config)(struct efx_nic *efx, int vf_i,
1426 struct ifla_vf_info *ivi);
1427 int (*sriov_set_vf_link_state)(struct efx_nic *efx, int vf_i,
1429 int (*vswitching_probe)(struct efx_nic *efx);
1430 int (*vswitching_restore)(struct efx_nic *efx);
1431 void (*vswitching_remove)(struct efx_nic *efx);
1432 int (*get_mac_address)(struct efx_nic *efx, unsigned char *perm_addr);
1433 int (*set_mac_address)(struct efx_nic *efx);
1434 u32 (*tso_versions)(struct efx_nic *efx);
1435 int (*udp_tnl_push_ports)(struct efx_nic *efx);
1436 bool (*udp_tnl_has_port)(struct efx_nic *efx, __be16 port);
1437 size_t (*print_additional_fwver)(struct efx_nic *efx, char *buf,
1439 void (*sensor_event)(struct efx_nic *efx, efx_qword_t *ev);
1442 unsigned int txd_ptr_tbl_base;
1443 unsigned int rxd_ptr_tbl_base;
1444 unsigned int buf_tbl_base;
1445 unsigned int evq_ptr_tbl_base;
1446 unsigned int evq_rptr_tbl_base;
1448 unsigned int rx_prefix_size;
1449 unsigned int rx_hash_offset;
1450 unsigned int rx_ts_offset;
1451 unsigned int rx_buffer_padding;
1452 bool can_rx_scatter;
1453 bool always_rx_scatter;
1454 bool option_descriptors;
1455 unsigned int min_interrupt_mode;
1456 unsigned int timer_period_max;
1457 netdev_features_t offload_features;
1459 unsigned int max_rx_ip_filters;
1460 u32 hwtstamp_filters;
1461 unsigned int rx_hash_key_size;
1464 /**************************************************************************
1466 * Prototypes and inline functions
1468 *************************************************************************/
1470 static inline struct efx_channel *
1471 efx_get_channel(struct efx_nic *efx, unsigned index)
1473 EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_channels);
1474 return efx->channel[index];
1477 /* Iterate over all used channels */
1478 #define efx_for_each_channel(_channel, _efx) \
1479 for (_channel = (_efx)->channel[0]; \
1481 _channel = (_channel->channel + 1 < (_efx)->n_channels) ? \
1482 (_efx)->channel[_channel->channel + 1] : NULL)
1484 /* Iterate over all used channels in reverse */
1485 #define efx_for_each_channel_rev(_channel, _efx) \
1486 for (_channel = (_efx)->channel[(_efx)->n_channels - 1]; \
1488 _channel = _channel->channel ? \
1489 (_efx)->channel[_channel->channel - 1] : NULL)
1491 static inline struct efx_channel *
1492 efx_get_tx_channel(struct efx_nic *efx, unsigned int index)
1494 EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_tx_channels);
1495 return efx->channel[efx->tx_channel_offset + index];
1498 static inline struct efx_tx_queue *
1499 efx_get_tx_queue(struct efx_nic *efx, unsigned index, unsigned type)
1501 EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_tx_channels ||
1502 type >= efx->tx_queues_per_channel);
1503 return &efx->channel[efx->tx_channel_offset + index]->tx_queue[type];
1506 static inline struct efx_channel *
1507 efx_get_xdp_channel(struct efx_nic *efx, unsigned int index)
1509 EFX_WARN_ON_ONCE_PARANOID(index >= efx->n_xdp_channels);
1510 return efx->channel[efx->xdp_channel_offset + index];
1513 static inline bool efx_channel_is_xdp_tx(struct efx_channel *channel)
1515 return channel->channel - channel->efx->xdp_channel_offset <
1516 channel->efx->n_xdp_channels;
1519 static inline bool efx_channel_has_tx_queues(struct efx_channel *channel)
1524 static inline unsigned int efx_channel_num_tx_queues(struct efx_channel *channel)
1526 if (efx_channel_is_xdp_tx(channel))
1527 return channel->efx->xdp_tx_per_channel;
1528 return channel->efx->tx_queues_per_channel;
1531 static inline struct efx_tx_queue *
1532 efx_channel_get_tx_queue(struct efx_channel *channel, unsigned type)
1534 EFX_WARN_ON_ONCE_PARANOID(type >= efx_channel_num_tx_queues(channel));
1535 return &channel->tx_queue[type];
1538 /* Iterate over all TX queues belonging to a channel */
1539 #define efx_for_each_channel_tx_queue(_tx_queue, _channel) \
1540 if (!efx_channel_has_tx_queues(_channel)) \
1543 for (_tx_queue = (_channel)->tx_queue; \
1544 _tx_queue < (_channel)->tx_queue + \
1545 efx_channel_num_tx_queues(_channel); \
1548 static inline bool efx_channel_has_rx_queue(struct efx_channel *channel)
1550 return channel->rx_queue.core_index >= 0;
1553 static inline struct efx_rx_queue *
1554 efx_channel_get_rx_queue(struct efx_channel *channel)
1556 EFX_WARN_ON_ONCE_PARANOID(!efx_channel_has_rx_queue(channel));
1557 return &channel->rx_queue;
1560 /* Iterate over all RX queues belonging to a channel */
1561 #define efx_for_each_channel_rx_queue(_rx_queue, _channel) \
1562 if (!efx_channel_has_rx_queue(_channel)) \
1565 for (_rx_queue = &(_channel)->rx_queue; \
1569 static inline struct efx_channel *
1570 efx_rx_queue_channel(struct efx_rx_queue *rx_queue)
1572 return container_of(rx_queue, struct efx_channel, rx_queue);
1575 static inline int efx_rx_queue_index(struct efx_rx_queue *rx_queue)
1577 return efx_rx_queue_channel(rx_queue)->channel;
1580 /* Returns a pointer to the specified receive buffer in the RX
1583 static inline struct efx_rx_buffer *efx_rx_buffer(struct efx_rx_queue *rx_queue,
1586 return &rx_queue->buffer[index];
1589 static inline struct efx_rx_buffer *
1590 efx_rx_buf_next(struct efx_rx_queue *rx_queue, struct efx_rx_buffer *rx_buf)
1592 if (unlikely(rx_buf == efx_rx_buffer(rx_queue, rx_queue->ptr_mask)))
1593 return efx_rx_buffer(rx_queue, 0);
1599 * EFX_MAX_FRAME_LEN - calculate maximum frame length
1601 * This calculates the maximum frame length that will be used for a
1602 * given MTU. The frame length will be equal to the MTU plus a
1603 * constant amount of header space and padding. This is the quantity
1604 * that the net driver will program into the MAC as the maximum frame
1607 * The 10G MAC requires 8-byte alignment on the frame
1608 * length, so we round up to the nearest 8.
1610 * Re-clocking by the XGXS on RX can reduce an IPG to 32 bits (half an
1611 * XGMII cycle). If the frame length reaches the maximum value in the
1612 * same cycle, the XMAC can miss the IPG altogether. We work around
1613 * this by adding a further 16 bytes.
1615 #define EFX_FRAME_PAD 16
1616 #define EFX_MAX_FRAME_LEN(mtu) \
1617 (ALIGN(((mtu) + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN + EFX_FRAME_PAD), 8))
1619 static inline bool efx_xmit_with_hwtstamp(struct sk_buff *skb)
1621 return skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP;
1623 static inline void efx_xmit_hwtstamp_pending(struct sk_buff *skb)
1625 skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
1628 /* Get the max fill level of the TX queues on this channel */
1629 static inline unsigned int
1630 efx_channel_tx_fill_level(struct efx_channel *channel)
1632 struct efx_tx_queue *tx_queue;
1633 unsigned int fill_level = 0;
1635 efx_for_each_channel_tx_queue(tx_queue, channel)
1636 fill_level = max(fill_level,
1637 tx_queue->insert_count - tx_queue->read_count);
1642 /* Conservative approximation of efx_channel_tx_fill_level using cached value */
1643 static inline unsigned int
1644 efx_channel_tx_old_fill_level(struct efx_channel *channel)
1646 struct efx_tx_queue *tx_queue;
1647 unsigned int fill_level = 0;
1649 efx_for_each_channel_tx_queue(tx_queue, channel)
1650 fill_level = max(fill_level,
1651 tx_queue->insert_count - tx_queue->old_read_count);
1656 /* Get all supported features.
1657 * If a feature is not fixed, it is present in hw_features.
1658 * If a feature is fixed, it does not present in hw_features, but
1659 * always in features.
1661 static inline netdev_features_t efx_supported_features(const struct efx_nic *efx)
1663 const struct net_device *net_dev = efx->net_dev;
1665 return net_dev->features | net_dev->hw_features;
1668 /* Get the current TX queue insert index. */
1669 static inline unsigned int
1670 efx_tx_queue_get_insert_index(const struct efx_tx_queue *tx_queue)
1672 return tx_queue->insert_count & tx_queue->ptr_mask;
1675 /* Get a TX buffer. */
1676 static inline struct efx_tx_buffer *
1677 __efx_tx_queue_get_insert_buffer(const struct efx_tx_queue *tx_queue)
1679 return &tx_queue->buffer[efx_tx_queue_get_insert_index(tx_queue)];
1682 /* Get a TX buffer, checking it's not currently in use. */
1683 static inline struct efx_tx_buffer *
1684 efx_tx_queue_get_insert_buffer(const struct efx_tx_queue *tx_queue)
1686 struct efx_tx_buffer *buffer =
1687 __efx_tx_queue_get_insert_buffer(tx_queue);
1689 EFX_WARN_ON_ONCE_PARANOID(buffer->len);
1690 EFX_WARN_ON_ONCE_PARANOID(buffer->flags);
1691 EFX_WARN_ON_ONCE_PARANOID(buffer->unmap_len);
1696 #endif /* EFX_NET_DRIVER_H */