1 /****************************************************************************
2 * Driver for Solarflare network controllers and boards
3 * Copyright 2005-2006 Fen Systems Ltd.
4 * Copyright 2005-2013 Solarflare Communications Inc.
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License version 2 as published
8 * by the Free Software Foundation, incorporated herein by reference.
11 /* Common definitions for all Efx net driver code */
13 #ifndef EFX_NET_DRIVER_H
14 #define EFX_NET_DRIVER_H
16 #include <linux/netdevice.h>
17 #include <linux/etherdevice.h>
18 #include <linux/ethtool.h>
19 #include <linux/if_vlan.h>
20 #include <linux/timer.h>
21 #include <linux/mdio.h>
22 #include <linux/list.h>
23 #include <linux/pci.h>
24 #include <linux/device.h>
25 #include <linux/highmem.h>
26 #include <linux/workqueue.h>
27 #include <linux/mutex.h>
28 #include <linux/rwsem.h>
29 #include <linux/vmalloc.h>
30 #include <linux/i2c.h>
31 #include <linux/mtd/mtd.h>
32 #include <net/busy_poll.h>
38 /**************************************************************************
42 **************************************************************************/
44 #define EFX_DRIVER_VERSION "4.1"
47 #define EFX_WARN_ON_ONCE_PARANOID(x) WARN_ON_ONCE(x)
48 #define EFX_WARN_ON_PARANOID(x) WARN_ON(x)
50 #define EFX_WARN_ON_ONCE_PARANOID(x) do {} while (0)
51 #define EFX_WARN_ON_PARANOID(x) do {} while (0)
54 /**************************************************************************
58 **************************************************************************/
60 #define EFX_MAX_CHANNELS 32U
61 #define EFX_MAX_RX_QUEUES EFX_MAX_CHANNELS
62 #define EFX_EXTRA_CHANNEL_IOV 0
63 #define EFX_EXTRA_CHANNEL_PTP 1
64 #define EFX_MAX_EXTRA_CHANNELS 2U
66 /* Checksum generation is a per-queue option in hardware, so each
67 * queue visible to the networking core is backed by two hardware TX
69 #define EFX_MAX_TX_TC 2
70 #define EFX_MAX_CORE_TX_QUEUES (EFX_MAX_TX_TC * EFX_MAX_CHANNELS)
71 #define EFX_TXQ_TYPE_OFFLOAD 1 /* flag */
72 #define EFX_TXQ_TYPE_HIGHPRI 2 /* flag */
73 #define EFX_TXQ_TYPES 4
74 #define EFX_MAX_TX_QUEUES (EFX_TXQ_TYPES * EFX_MAX_CHANNELS)
76 /* Maximum possible MTU the driver supports */
77 #define EFX_MAX_MTU (9 * 1024)
79 /* Minimum MTU, from RFC791 (IP) */
80 #define EFX_MIN_MTU 68
82 /* Size of an RX scatter buffer. Small enough to pack 2 into a 4K page,
83 * and should be a multiple of the cache line size.
85 #define EFX_RX_USR_BUF_SIZE (2048 - 256)
87 /* If possible, we should ensure cache line alignment at start and end
88 * of every buffer. Otherwise, we just need to ensure 4-byte
89 * alignment of the network header.
92 #define EFX_RX_BUF_ALIGNMENT L1_CACHE_BYTES
94 #define EFX_RX_BUF_ALIGNMENT 4
97 /* Forward declare Precision Time Protocol (PTP) support structure. */
99 struct hwtstamp_config
;
101 struct efx_self_tests
;
104 * struct efx_buffer - A general-purpose DMA buffer
105 * @addr: host base address of the buffer
106 * @dma_addr: DMA base address of the buffer
107 * @len: Buffer length, in bytes
109 * The NIC uses these buffers for its interrupt status registers and
119 * struct efx_special_buffer - DMA buffer entered into buffer table
120 * @buf: Standard &struct efx_buffer
121 * @index: Buffer index within controller;s buffer table
122 * @entries: Number of buffer table entries
124 * The NIC has a buffer table that maps buffers of size %EFX_BUF_SIZE.
125 * Event and descriptor rings are addressed via one or more buffer
126 * table entries (and so can be physically non-contiguous, although we
127 * currently do not take advantage of that). On Falcon and Siena we
128 * have to take care of allocating and initialising the entries
129 * ourselves. On later hardware this is managed by the firmware and
130 * @index and @entries are left as 0.
132 struct efx_special_buffer
{
133 struct efx_buffer buf
;
135 unsigned int entries
;
139 * struct efx_tx_buffer - buffer state for a TX descriptor
140 * @skb: When @flags & %EFX_TX_BUF_SKB, the associated socket buffer to be
141 * freed when descriptor completes
142 * @option: When @flags & %EFX_TX_BUF_OPTION, a NIC-specific option descriptor.
143 * @dma_addr: DMA address of the fragment.
144 * @flags: Flags for allocation and DMA mapping type
145 * @len: Length of this fragment.
146 * This field is zero when the queue slot is empty.
147 * @unmap_len: Length of this fragment to unmap
148 * @dma_offset: Offset of @dma_addr from the address of the backing DMA mapping.
149 * Only valid if @unmap_len != 0.
151 struct efx_tx_buffer
{
152 const struct sk_buff
*skb
;
157 unsigned short flags
;
159 unsigned short unmap_len
;
160 unsigned short dma_offset
;
162 #define EFX_TX_BUF_CONT 1 /* not last descriptor of packet */
163 #define EFX_TX_BUF_SKB 2 /* buffer is last part of skb */
164 #define EFX_TX_BUF_MAP_SINGLE 8 /* buffer was mapped with dma_map_single() */
165 #define EFX_TX_BUF_OPTION 0x10 /* empty buffer for option descriptor */
168 * struct efx_tx_queue - An Efx TX queue
170 * This is a ring buffer of TX fragments.
171 * Since the TX completion path always executes on the same
172 * CPU and the xmit path can operate on different CPUs,
173 * performance is increased by ensuring that the completion
174 * path and the xmit path operate on different cache lines.
175 * This is particularly important if the xmit path is always
176 * executing on one CPU which is different from the completion
177 * path. There is also a cache line for members which are
178 * read but not written on the fast path.
180 * @efx: The associated Efx NIC
181 * @queue: DMA queue number
182 * @tso_version: Version of TSO in use for this queue.
183 * @channel: The associated channel
184 * @core_txq: The networking core TX queue structure
185 * @buffer: The software buffer ring
186 * @cb_page: Array of pages of copy buffers. Carved up according to
187 * %EFX_TX_CB_ORDER into %EFX_TX_CB_SIZE-sized chunks.
188 * @txd: The hardware descriptor ring
189 * @ptr_mask: The size of the ring minus 1.
190 * @piobuf: PIO buffer region for this TX queue (shared with its partner).
191 * Size of the region is efx_piobuf_size.
192 * @piobuf_offset: Buffer offset to be specified in PIO descriptors
193 * @initialised: Has hardware queue been initialised?
194 * @timestamping: Is timestamping enabled for this channel?
195 * @handle_tso: TSO xmit preparation handler. Sets up the TSO metadata and
196 * may also map tx data, depending on the nature of the TSO implementation.
197 * @read_count: Current read pointer.
198 * This is the number of buffers that have been removed from both rings.
199 * @old_write_count: The value of @write_count when last checked.
200 * This is here for performance reasons. The xmit path will
201 * only get the up-to-date value of @write_count if this
202 * variable indicates that the queue is empty. This is to
203 * avoid cache-line ping-pong between the xmit path and the
205 * @merge_events: Number of TX merged completion events
206 * @completed_desc_ptr: Most recent completed pointer - only used with
208 * @completed_timestamp_major: Top part of the most recent tx timestamp.
209 * @completed_timestamp_minor: Low part of the most recent tx timestamp.
210 * @insert_count: Current insert pointer
211 * This is the number of buffers that have been added to the
213 * @write_count: Current write pointer
214 * This is the number of buffers that have been added to the
216 * @packet_write_count: Completable write pointer
217 * This is the write pointer of the last packet written.
218 * Normally this will equal @write_count, but as option descriptors
219 * don't produce completion events, they won't update this.
220 * Filled in iff @efx->type->option_descriptors; only used for PIO.
221 * Thus, this is written and used on EF10, and neither on farch.
222 * @old_read_count: The value of read_count when last checked.
223 * This is here for performance reasons. The xmit path will
224 * only get the up-to-date value of read_count if this
225 * variable indicates that the queue is full. This is to
226 * avoid cache-line ping-pong between the xmit path and the
228 * @tso_bursts: Number of times TSO xmit invoked by kernel
229 * @tso_long_headers: Number of packets with headers too long for standard
231 * @tso_packets: Number of packets via the TSO xmit path
232 * @tso_fallbacks: Number of times TSO fallback used
233 * @pushes: Number of times the TX push feature has been used
234 * @pio_packets: Number of times the TX PIO feature has been used
235 * @xmit_more_available: Are any packets waiting to be pushed to the NIC
236 * @cb_packets: Number of times the TX copybreak feature has been used
237 * @empty_read_count: If the completion path has seen the queue as empty
238 * and the transmission path has not yet checked this, the value of
239 * @read_count bitwise-added to %EFX_EMPTY_COUNT_VALID; otherwise 0.
241 struct efx_tx_queue
{
242 /* Members which don't change on the fast path */
243 struct efx_nic
*efx ____cacheline_aligned_in_smp
;
245 unsigned int tso_version
;
246 struct efx_channel
*channel
;
247 struct netdev_queue
*core_txq
;
248 struct efx_tx_buffer
*buffer
;
249 struct efx_buffer
*cb_page
;
250 struct efx_special_buffer txd
;
251 unsigned int ptr_mask
;
252 void __iomem
*piobuf
;
253 unsigned int piobuf_offset
;
257 /* Function pointers used in the fast path. */
258 int (*handle_tso
)(struct efx_tx_queue
*, struct sk_buff
*, bool *);
260 /* Members used mainly on the completion path */
261 unsigned int read_count ____cacheline_aligned_in_smp
;
262 unsigned int old_write_count
;
263 unsigned int merge_events
;
264 unsigned int bytes_compl
;
265 unsigned int pkts_compl
;
266 unsigned int completed_desc_ptr
;
267 u32 completed_timestamp_major
;
268 u32 completed_timestamp_minor
;
270 /* Members used only on the xmit path */
271 unsigned int insert_count ____cacheline_aligned_in_smp
;
272 unsigned int write_count
;
273 unsigned int packet_write_count
;
274 unsigned int old_read_count
;
275 unsigned int tso_bursts
;
276 unsigned int tso_long_headers
;
277 unsigned int tso_packets
;
278 unsigned int tso_fallbacks
;
280 unsigned int pio_packets
;
281 bool xmit_more_available
;
282 unsigned int cb_packets
;
283 /* Statistics to supplement MAC stats */
284 unsigned long tx_packets
;
286 /* Members shared between paths and sometimes updated */
287 unsigned int empty_read_count ____cacheline_aligned_in_smp
;
288 #define EFX_EMPTY_COUNT_VALID 0x80000000
289 atomic_t flush_outstanding
;
292 #define EFX_TX_CB_ORDER 7
293 #define EFX_TX_CB_SIZE (1 << EFX_TX_CB_ORDER) - NET_IP_ALIGN
296 * struct efx_rx_buffer - An Efx RX data buffer
297 * @dma_addr: DMA base address of the buffer
298 * @page: The associated page buffer.
299 * Will be %NULL if the buffer slot is currently free.
300 * @page_offset: If pending: offset in @page of DMA base address.
301 * If completed: offset in @page of Ethernet header.
302 * @len: If pending: length for DMA descriptor.
303 * If completed: received length, excluding hash prefix.
304 * @flags: Flags for buffer and packet state. These are only set on the
305 * first buffer of a scattered packet.
307 struct efx_rx_buffer
{
314 #define EFX_RX_BUF_LAST_IN_PAGE 0x0001
315 #define EFX_RX_PKT_CSUMMED 0x0002
316 #define EFX_RX_PKT_DISCARD 0x0004
317 #define EFX_RX_PKT_TCP 0x0040
318 #define EFX_RX_PKT_PREFIX_LEN 0x0080 /* length is in prefix only */
319 #define EFX_RX_PKT_CSUM_LEVEL 0x0200
322 * struct efx_rx_page_state - Page-based rx buffer state
324 * Inserted at the start of every page allocated for receive buffers.
325 * Used to facilitate sharing dma mappings between recycled rx buffers
326 * and those passed up to the kernel.
328 * @dma_addr: The dma address of this page.
330 struct efx_rx_page_state
{
333 unsigned int __pad
[0] ____cacheline_aligned
;
337 * struct efx_rx_queue - An Efx RX queue
338 * @efx: The associated Efx NIC
339 * @core_index: Index of network core RX queue. Will be >= 0 iff this
340 * is associated with a real RX queue.
341 * @buffer: The software buffer ring
342 * @rxd: The hardware descriptor ring
343 * @ptr_mask: The size of the ring minus 1.
344 * @refill_enabled: Enable refill whenever fill level is low
345 * @flush_pending: Set when a RX flush is pending. Has the same lifetime as
346 * @rxq_flush_pending.
347 * @added_count: Number of buffers added to the receive queue.
348 * @notified_count: Number of buffers given to NIC (<= @added_count).
349 * @removed_count: Number of buffers removed from the receive queue.
350 * @scatter_n: Used by NIC specific receive code.
351 * @scatter_len: Used by NIC specific receive code.
352 * @page_ring: The ring to store DMA mapped pages for reuse.
353 * @page_add: Counter to calculate the write pointer for the recycle ring.
354 * @page_remove: Counter to calculate the read pointer for the recycle ring.
355 * @page_recycle_count: The number of pages that have been recycled.
356 * @page_recycle_failed: The number of pages that couldn't be recycled because
357 * the kernel still held a reference to them.
358 * @page_recycle_full: The number of pages that were released because the
359 * recycle ring was full.
360 * @page_ptr_mask: The number of pages in the RX recycle ring minus 1.
361 * @max_fill: RX descriptor maximum fill level (<= ring size)
362 * @fast_fill_trigger: RX descriptor fill level that will trigger a fast fill
364 * @min_fill: RX descriptor minimum non-zero fill level.
365 * This records the minimum fill level observed when a ring
366 * refill was triggered.
367 * @recycle_count: RX buffer recycle counter.
368 * @slow_fill: Timer used to defer efx_nic_generate_fill_event().
370 struct efx_rx_queue
{
373 struct efx_rx_buffer
*buffer
;
374 struct efx_special_buffer rxd
;
375 unsigned int ptr_mask
;
379 unsigned int added_count
;
380 unsigned int notified_count
;
381 unsigned int removed_count
;
382 unsigned int scatter_n
;
383 unsigned int scatter_len
;
384 struct page
**page_ring
;
385 unsigned int page_add
;
386 unsigned int page_remove
;
387 unsigned int page_recycle_count
;
388 unsigned int page_recycle_failed
;
389 unsigned int page_recycle_full
;
390 unsigned int page_ptr_mask
;
391 unsigned int max_fill
;
392 unsigned int fast_fill_trigger
;
393 unsigned int min_fill
;
394 unsigned int min_overfill
;
395 unsigned int recycle_count
;
396 struct timer_list slow_fill
;
397 unsigned int slow_fill_count
;
398 /* Statistics to supplement MAC stats */
399 unsigned long rx_packets
;
402 enum efx_sync_events_state
{
403 SYNC_EVENTS_DISABLED
= 0,
404 SYNC_EVENTS_QUIESCENT
,
405 SYNC_EVENTS_REQUESTED
,
410 * struct efx_channel - An Efx channel
412 * A channel comprises an event queue, at least one TX queue, at least
413 * one RX queue, and an associated tasklet for processing the event
416 * @efx: Associated Efx NIC
417 * @channel: Channel instance number
418 * @type: Channel type definition
419 * @eventq_init: Event queue initialised flag
420 * @enabled: Channel enabled indicator
421 * @irq: IRQ number (MSI and MSI-X only)
422 * @irq_moderation_us: IRQ moderation value (in microseconds)
423 * @napi_dev: Net device used with NAPI
424 * @napi_str: NAPI control structure
425 * @state: state for NAPI vs busy polling
426 * @state_lock: lock protecting @state
427 * @eventq: Event queue buffer
428 * @eventq_mask: Event queue pointer mask
429 * @eventq_read_ptr: Event queue read pointer
430 * @event_test_cpu: Last CPU to handle interrupt or test event for this channel
431 * @irq_count: Number of IRQs since last adaptive moderation decision
432 * @irq_mod_score: IRQ moderation score
433 * @filter_work: Work item for efx_filter_rfs_expire()
434 * @rps_flow_id: Flow IDs of filters allocated for accelerated RFS,
435 * indexed by filter ID
436 * @n_rx_tobe_disc: Count of RX_TOBE_DISC errors
437 * @n_rx_ip_hdr_chksum_err: Count of RX IP header checksum errors
438 * @n_rx_tcp_udp_chksum_err: Count of RX TCP and UDP checksum errors
439 * @n_rx_mcast_mismatch: Count of unmatched multicast frames
440 * @n_rx_frm_trunc: Count of RX_FRM_TRUNC errors
441 * @n_rx_overlength: Count of RX_OVERLENGTH errors
442 * @n_skbuff_leaks: Count of skbuffs leaked due to RX overrun
443 * @n_rx_nodesc_trunc: Number of RX packets truncated and then dropped due to
444 * lack of descriptors
445 * @n_rx_merge_events: Number of RX merged completion events
446 * @n_rx_merge_packets: Number of RX packets completed by merged events
447 * @rx_pkt_n_frags: Number of fragments in next packet to be delivered by
448 * __efx_rx_packet(), or zero if there is none
449 * @rx_pkt_index: Ring index of first buffer for next packet to be delivered
450 * by __efx_rx_packet(), if @rx_pkt_n_frags != 0
451 * @rx_queue: RX queue for this channel
452 * @tx_queue: TX queues for this channel
453 * @sync_events_state: Current state of sync events on this channel
454 * @sync_timestamp_major: Major part of the last ptp sync event
455 * @sync_timestamp_minor: Minor part of the last ptp sync event
460 const struct efx_channel_type
*type
;
464 unsigned int irq_moderation_us
;
465 struct net_device
*napi_dev
;
466 struct napi_struct napi_str
;
467 #ifdef CONFIG_NET_RX_BUSY_POLL
468 unsigned long busy_poll_state
;
470 struct efx_special_buffer eventq
;
471 unsigned int eventq_mask
;
472 unsigned int eventq_read_ptr
;
475 unsigned int irq_count
;
476 unsigned int irq_mod_score
;
477 #ifdef CONFIG_RFS_ACCEL
478 unsigned int rfs_filters_added
;
479 struct work_struct filter_work
;
480 #define RPS_FLOW_ID_INVALID 0xFFFFFFFF
484 unsigned int n_rx_tobe_disc
;
485 unsigned int n_rx_ip_hdr_chksum_err
;
486 unsigned int n_rx_tcp_udp_chksum_err
;
487 unsigned int n_rx_outer_ip_hdr_chksum_err
;
488 unsigned int n_rx_outer_tcp_udp_chksum_err
;
489 unsigned int n_rx_inner_ip_hdr_chksum_err
;
490 unsigned int n_rx_inner_tcp_udp_chksum_err
;
491 unsigned int n_rx_eth_crc_err
;
492 unsigned int n_rx_mcast_mismatch
;
493 unsigned int n_rx_frm_trunc
;
494 unsigned int n_rx_overlength
;
495 unsigned int n_skbuff_leaks
;
496 unsigned int n_rx_nodesc_trunc
;
497 unsigned int n_rx_merge_events
;
498 unsigned int n_rx_merge_packets
;
500 unsigned int rx_pkt_n_frags
;
501 unsigned int rx_pkt_index
;
503 struct efx_rx_queue rx_queue
;
504 struct efx_tx_queue tx_queue
[EFX_TXQ_TYPES
];
506 enum efx_sync_events_state sync_events_state
;
507 u32 sync_timestamp_major
;
508 u32 sync_timestamp_minor
;
512 * struct efx_msi_context - Context for each MSI
513 * @efx: The associated NIC
514 * @index: Index of the channel/IRQ
515 * @name: Name of the channel/IRQ
517 * Unlike &struct efx_channel, this is never reallocated and is always
518 * safe for the IRQ handler to access.
520 struct efx_msi_context
{
523 char name
[IFNAMSIZ
+ 6];
527 * struct efx_channel_type - distinguishes traffic and extra channels
528 * @handle_no_channel: Handle failure to allocate an extra channel
529 * @pre_probe: Set up extra state prior to initialisation
530 * @post_remove: Tear down extra state after finalisation, if allocated.
531 * May be called on channels that have not been probed.
532 * @get_name: Generate the channel's name (used for its IRQ handler)
533 * @copy: Copy the channel state prior to reallocation. May be %NULL if
534 * reallocation is not supported.
535 * @receive_skb: Handle an skb ready to be passed to netif_receive_skb()
536 * @want_txqs: Determine whether this channel should have TX queues
537 * created. If %NULL, TX queues are not created.
538 * @keep_eventq: Flag for whether event queue should be kept initialised
539 * while the device is stopped
540 * @want_pio: Flag for whether PIO buffers should be linked to this
541 * channel's TX queues.
543 struct efx_channel_type
{
544 void (*handle_no_channel
)(struct efx_nic
*);
545 int (*pre_probe
)(struct efx_channel
*);
546 void (*post_remove
)(struct efx_channel
*);
547 void (*get_name
)(struct efx_channel
*, char *buf
, size_t len
);
548 struct efx_channel
*(*copy
)(const struct efx_channel
*);
549 bool (*receive_skb
)(struct efx_channel
*, struct sk_buff
*);
550 bool (*want_txqs
)(struct efx_channel
*);
561 #define STRING_TABLE_LOOKUP(val, member) \
562 ((val) < member ## _max) ? member ## _names[val] : "(invalid)"
564 extern const char *const efx_loopback_mode_names
[];
565 extern const unsigned int efx_loopback_mode_max
;
566 #define LOOPBACK_MODE(efx) \
567 STRING_TABLE_LOOKUP((efx)->loopback_mode, efx_loopback_mode)
569 extern const char *const efx_reset_type_names
[];
570 extern const unsigned int efx_reset_type_max
;
571 #define RESET_TYPE(type) \
572 STRING_TABLE_LOOKUP(type, efx_reset_type)
574 void efx_get_udp_tunnel_type_name(u16 type
, char *buf
, size_t buflen
);
577 /* Be careful if altering to correct macro below */
578 EFX_INT_MODE_MSIX
= 0,
579 EFX_INT_MODE_MSI
= 1,
580 EFX_INT_MODE_LEGACY
= 2,
581 EFX_INT_MODE_MAX
/* Insert any new items before this */
583 #define EFX_INT_MODE_USE_MSI(x) (((x)->interrupt_mode) <= EFX_INT_MODE_MSI)
586 STATE_UNINIT
= 0, /* device being probed/removed or is frozen */
587 STATE_READY
= 1, /* hardware ready and netdev registered */
588 STATE_DISABLED
= 2, /* device disabled due to hardware errors */
589 STATE_RECOVERY
= 3, /* device recovering from PCI error */
592 /* Forward declaration */
595 /* Pseudo bit-mask flow control field */
596 #define EFX_FC_RX FLOW_CTRL_RX
597 #define EFX_FC_TX FLOW_CTRL_TX
598 #define EFX_FC_AUTO 4
601 * struct efx_link_state - Current state of the link
603 * @fd: Link is full-duplex
604 * @fc: Actual flow control flags
605 * @speed: Link speed (Mbps)
607 struct efx_link_state
{
614 static inline bool efx_link_state_equal(const struct efx_link_state
*left
,
615 const struct efx_link_state
*right
)
617 return left
->up
== right
->up
&& left
->fd
== right
->fd
&&
618 left
->fc
== right
->fc
&& left
->speed
== right
->speed
;
622 * struct efx_phy_operations - Efx PHY operations table
623 * @probe: Probe PHY and initialise efx->mdio.mode_support, efx->mdio.mmds,
624 * efx->loopback_modes.
625 * @init: Initialise PHY
626 * @fini: Shut down PHY
627 * @reconfigure: Reconfigure PHY (e.g. for new link parameters)
628 * @poll: Update @link_state and report whether it changed.
629 * Serialised by the mac_lock.
630 * @get_link_ksettings: Get ethtool settings. Serialised by the mac_lock.
631 * @set_link_ksettings: Set ethtool settings. Serialised by the mac_lock.
632 * @get_fecparam: Get Forward Error Correction settings. Serialised by mac_lock.
633 * @set_fecparam: Set Forward Error Correction settings. Serialised by mac_lock.
634 * @set_npage_adv: Set abilities advertised in (Extended) Next Page
635 * (only needed where AN bit is set in mmds)
636 * @test_alive: Test that PHY is 'alive' (online)
637 * @test_name: Get the name of a PHY-specific test/result
638 * @run_tests: Run tests and record results as appropriate (offline).
639 * Flags are the ethtool tests flags.
641 struct efx_phy_operations
{
642 int (*probe
) (struct efx_nic
*efx
);
643 int (*init
) (struct efx_nic
*efx
);
644 void (*fini
) (struct efx_nic
*efx
);
645 void (*remove
) (struct efx_nic
*efx
);
646 int (*reconfigure
) (struct efx_nic
*efx
);
647 bool (*poll
) (struct efx_nic
*efx
);
648 void (*get_link_ksettings
)(struct efx_nic
*efx
,
649 struct ethtool_link_ksettings
*cmd
);
650 int (*set_link_ksettings
)(struct efx_nic
*efx
,
651 const struct ethtool_link_ksettings
*cmd
);
652 int (*get_fecparam
)(struct efx_nic
*efx
, struct ethtool_fecparam
*fec
);
653 int (*set_fecparam
)(struct efx_nic
*efx
,
654 const struct ethtool_fecparam
*fec
);
655 void (*set_npage_adv
) (struct efx_nic
*efx
, u32
);
656 int (*test_alive
) (struct efx_nic
*efx
);
657 const char *(*test_name
) (struct efx_nic
*efx
, unsigned int index
);
658 int (*run_tests
) (struct efx_nic
*efx
, int *results
, unsigned flags
);
659 int (*get_module_eeprom
) (struct efx_nic
*efx
,
660 struct ethtool_eeprom
*ee
,
662 int (*get_module_info
) (struct efx_nic
*efx
,
663 struct ethtool_modinfo
*modinfo
);
667 * enum efx_phy_mode - PHY operating mode flags
668 * @PHY_MODE_NORMAL: on and should pass traffic
669 * @PHY_MODE_TX_DISABLED: on with TX disabled
670 * @PHY_MODE_LOW_POWER: set to low power through MDIO
671 * @PHY_MODE_OFF: switched off through external control
672 * @PHY_MODE_SPECIAL: on but will not pass traffic
676 PHY_MODE_TX_DISABLED
= 1,
677 PHY_MODE_LOW_POWER
= 2,
679 PHY_MODE_SPECIAL
= 8,
682 static inline bool efx_phy_mode_disabled(enum efx_phy_mode mode
)
684 return !!(mode
& ~PHY_MODE_TX_DISABLED
);
688 * struct efx_hw_stat_desc - Description of a hardware statistic
689 * @name: Name of the statistic as visible through ethtool, or %NULL if
690 * it should not be exposed
691 * @dma_width: Width in bits (0 for non-DMA statistics)
692 * @offset: Offset within stats (ignored for non-DMA statistics)
694 struct efx_hw_stat_desc
{
700 /* Number of bits used in a multicast filter hash address */
701 #define EFX_MCAST_HASH_BITS 8
703 /* Number of (single-bit) entries in a multicast filter hash */
704 #define EFX_MCAST_HASH_ENTRIES (1 << EFX_MCAST_HASH_BITS)
706 /* An Efx multicast filter hash */
707 union efx_multicast_hash
{
708 u8 byte
[EFX_MCAST_HASH_ENTRIES
/ 8];
709 efx_oword_t oword
[EFX_MCAST_HASH_ENTRIES
/ sizeof(efx_oword_t
) / 8];
714 /* The reserved RSS context value */
715 #define EFX_EF10_RSS_CONTEXT_INVALID 0xffffffff
717 * struct efx_rss_context - A user-defined RSS context for filtering
718 * @list: node of linked list on which this struct is stored
719 * @context_id: the RSS_CONTEXT_ID returned by MC firmware, or
720 * %EFX_EF10_RSS_CONTEXT_INVALID if this context is not present on the NIC.
721 * For Siena, 0 if RSS is active, else %EFX_EF10_RSS_CONTEXT_INVALID.
722 * @user_id: the rss_context ID exposed to userspace over ethtool.
723 * @rx_hash_udp_4tuple: UDP 4-tuple hashing enabled
724 * @rx_hash_key: Toeplitz hash key for this RSS context
725 * @indir_table: Indirection table for this RSS context
727 struct efx_rss_context
{
728 struct list_head list
;
731 bool rx_hash_udp_4tuple
;
733 u32 rx_indir_table
[128];
737 * struct efx_nic - an Efx NIC
738 * @name: Device name (net device name or bus id before net device registered)
739 * @pci_dev: The PCI device
740 * @node: List node for maintaning primary/secondary function lists
741 * @primary: &struct efx_nic instance for the primary function of this
742 * controller. May be the same structure, and may be %NULL if no
743 * primary function is bound. Serialised by rtnl_lock.
744 * @secondary_list: List of &struct efx_nic instances for the secondary PCI
745 * functions of the controller, if this is for the primary function.
746 * Serialised by rtnl_lock.
747 * @type: Controller type attributes
748 * @legacy_irq: IRQ number
749 * @workqueue: Workqueue for port reconfigures and the HW monitor.
750 * Work items do not hold and must not acquire RTNL.
751 * @workqueue_name: Name of workqueue
752 * @reset_work: Scheduled reset workitem
753 * @membase_phys: Memory BAR value as physical address
754 * @membase: Memory BAR value
755 * @vi_stride: step between per-VI registers / memory regions
756 * @interrupt_mode: Interrupt mode
757 * @timer_quantum_ns: Interrupt timer quantum, in nanoseconds
758 * @timer_max_ns: Interrupt timer maximum value, in nanoseconds
759 * @irq_rx_adaptive: Adaptive IRQ moderation enabled for RX event queues
760 * @irq_rx_mod_step_us: Step size for IRQ moderation for RX event queues
761 * @irq_rx_moderation_us: IRQ moderation time for RX event queues
762 * @msg_enable: Log message enable flags
763 * @state: Device state number (%STATE_*). Serialised by the rtnl_lock.
764 * @reset_pending: Bitmask for pending resets
765 * @tx_queue: TX DMA queues
766 * @rx_queue: RX DMA queues
768 * @msi_context: Context for each MSI
769 * @extra_channel_types: Types of extra (non-traffic) channels that
770 * should be allocated for this NIC
771 * @rxq_entries: Size of receive queues requested by user.
772 * @txq_entries: Size of transmit queues requested by user.
773 * @txq_stop_thresh: TX queue fill level at or above which we stop it.
774 * @txq_wake_thresh: TX queue fill level at or below which we wake it.
775 * @tx_dc_base: Base qword address in SRAM of TX queue descriptor caches
776 * @rx_dc_base: Base qword address in SRAM of RX queue descriptor caches
777 * @sram_lim_qw: Qword address limit of SRAM
778 * @next_buffer_table: First available buffer table id
779 * @n_channels: Number of channels in use
780 * @n_rx_channels: Number of channels used for RX (= number of RX queues)
781 * @n_tx_channels: Number of channels used for TX
782 * @n_extra_tx_channels: Number of extra channels with TX queues
783 * @rx_ip_align: RX DMA address offset to have IP header aligned in
784 * in accordance with NET_IP_ALIGN
785 * @rx_dma_len: Current maximum RX DMA length
786 * @rx_buffer_order: Order (log2) of number of pages for each RX buffer
787 * @rx_buffer_truesize: Amortised allocation size of an RX buffer,
788 * for use in sk_buff::truesize
789 * @rx_prefix_size: Size of RX prefix before packet data
790 * @rx_packet_hash_offset: Offset of RX flow hash from start of packet data
791 * (valid only if @rx_prefix_size != 0; always negative)
792 * @rx_packet_len_offset: Offset of RX packet length from start of packet data
793 * (valid only for NICs that set %EFX_RX_PKT_PREFIX_LEN; always negative)
794 * @rx_packet_ts_offset: Offset of timestamp from start of packet data
795 * (valid only if channel->sync_timestamps_enabled; always negative)
796 * @rx_scatter: Scatter mode enabled for receives
797 * @rss_context: Main RSS context. Its @list member is the head of the list of
798 * RSS contexts created by user requests
799 * @rss_lock: Protects custom RSS context software state in @rss_context.list
800 * @int_error_count: Number of internal errors seen recently
801 * @int_error_expire: Time at which error count will be expired
802 * @irq_soft_enabled: Are IRQs soft-enabled? If not, IRQ handler will
803 * acknowledge but do nothing else.
804 * @irq_status: Interrupt status buffer
805 * @irq_zero_count: Number of legacy IRQs seen with queue flags == 0
806 * @irq_level: IRQ level/index for IRQs not triggered by an event queue
807 * @selftest_work: Work item for asynchronous self-test
808 * @mtd_list: List of MTDs attached to the NIC
809 * @nic_data: Hardware dependent state
810 * @mcdi: Management-Controller-to-Driver Interface state
811 * @mac_lock: MAC access lock. Protects @port_enabled, @phy_mode,
812 * efx_monitor() and efx_reconfigure_port()
813 * @port_enabled: Port enabled indicator.
814 * Serialises efx_stop_all(), efx_start_all(), efx_monitor() and
815 * efx_mac_work() with kernel interfaces. Safe to read under any
816 * one of the rtnl_lock, mac_lock, or netif_tx_lock, but all three must
817 * be held to modify it.
818 * @port_initialized: Port initialized?
819 * @net_dev: Operating system network device. Consider holding the rtnl lock
820 * @fixed_features: Features which cannot be turned off
821 * @num_mac_stats: Number of MAC stats reported by firmware (MAC_STATS_NUM_STATS
822 * field of %MC_CMD_GET_CAPABILITIES_V4 response, or %MC_CMD_MAC_NSTATS)
823 * @stats_buffer: DMA buffer for statistics
824 * @phy_type: PHY type
825 * @phy_op: PHY interface
826 * @phy_data: PHY private data (including PHY-specific stats)
827 * @mdio: PHY MDIO interface
828 * @mdio_bus: PHY MDIO bus ID (only used by Siena)
829 * @phy_mode: PHY operating mode. Serialised by @mac_lock.
830 * @link_advertising: Autonegotiation advertising flags
831 * @fec_config: Forward Error Correction configuration flags. For bit positions
832 * see &enum ethtool_fec_config_bits.
833 * @link_state: Current state of the link
834 * @n_link_state_changes: Number of times the link has changed state
835 * @unicast_filter: Flag for Falcon-arch simple unicast filter.
836 * Protected by @mac_lock.
837 * @multicast_hash: Multicast hash table for Falcon-arch.
838 * Protected by @mac_lock.
839 * @wanted_fc: Wanted flow control flags
840 * @fc_disable: When non-zero flow control is disabled. Typically used to
841 * ensure that network back pressure doesn't delay dma queue flushes.
842 * Serialised by the rtnl lock.
843 * @mac_work: Work item for changing MAC promiscuity and multicast hash
844 * @loopback_mode: Loopback status
845 * @loopback_modes: Supported loopback mode bitmask
846 * @loopback_selftest: Offline self-test private state
847 * @filter_sem: Filter table rw_semaphore, protects existence of @filter_state
848 * @filter_state: Architecture-dependent filter table state
849 * @rps_mutex: Protects RPS state of all channels
850 * @rps_expire_channel: Next channel to check for expiry
851 * @rps_expire_index: Next index to check for expiry in
852 * @rps_expire_channel's @rps_flow_id
853 * @active_queues: Count of RX and TX queues that haven't been flushed and drained.
854 * @rxq_flush_pending: Count of number of receive queues that need to be flushed.
855 * Decremented when the efx_flush_rx_queue() is called.
856 * @rxq_flush_outstanding: Count of number of RX flushes started but not yet
857 * completed (either success or failure). Not used when MCDI is used to
858 * flush receive queues.
859 * @flush_wq: wait queue used by efx_nic_flush_queues() to wait for flush completions.
860 * @vf_count: Number of VFs intended to be enabled.
861 * @vf_init_count: Number of VFs that have been fully initialised.
862 * @vi_scale: log2 number of vnics per VF.
863 * @ptp_data: PTP state data
864 * @ptp_warned: has this NIC seen and warned about unexpected PTP events?
865 * @vpd_sn: Serial number read from VPD
866 * @monitor_work: Hardware monitor workitem
867 * @biu_lock: BIU (bus interface unit) lock
868 * @last_irq_cpu: Last CPU to handle a possible test interrupt. This
869 * field is used by efx_test_interrupts() to verify that an
870 * interrupt has occurred.
871 * @stats_lock: Statistics update lock. Must be held when calling
872 * efx_nic_type::{update,start,stop}_stats.
873 * @n_rx_noskb_drops: Count of RX packets dropped due to failure to allocate an skb
875 * This is stored in the private area of the &struct net_device.
878 /* The following fields should be written very rarely */
881 struct list_head node
;
882 struct efx_nic
*primary
;
883 struct list_head secondary_list
;
884 struct pci_dev
*pci_dev
;
885 unsigned int port_num
;
886 const struct efx_nic_type
*type
;
888 bool eeh_disabled_legacy_irq
;
889 struct workqueue_struct
*workqueue
;
890 char workqueue_name
[16];
891 struct work_struct reset_work
;
892 resource_size_t membase_phys
;
893 void __iomem
*membase
;
895 unsigned int vi_stride
;
897 enum efx_int_mode interrupt_mode
;
898 unsigned int timer_quantum_ns
;
899 unsigned int timer_max_ns
;
900 bool irq_rx_adaptive
;
901 unsigned int irq_mod_step_us
;
902 unsigned int irq_rx_moderation_us
;
905 enum nic_state state
;
906 unsigned long reset_pending
;
908 struct efx_channel
*channel
[EFX_MAX_CHANNELS
];
909 struct efx_msi_context msi_context
[EFX_MAX_CHANNELS
];
910 const struct efx_channel_type
*
911 extra_channel_type
[EFX_MAX_EXTRA_CHANNELS
];
913 unsigned rxq_entries
;
914 unsigned txq_entries
;
915 unsigned int txq_stop_thresh
;
916 unsigned int txq_wake_thresh
;
920 unsigned sram_lim_qw
;
921 unsigned next_buffer_table
;
923 unsigned int max_channels
;
924 unsigned int max_tx_channels
;
926 unsigned n_rx_channels
;
928 unsigned tx_channel_offset
;
929 unsigned n_tx_channels
;
930 unsigned n_extra_tx_channels
;
931 unsigned int rx_ip_align
;
932 unsigned int rx_dma_len
;
933 unsigned int rx_buffer_order
;
934 unsigned int rx_buffer_truesize
;
935 unsigned int rx_page_buf_step
;
936 unsigned int rx_bufs_per_page
;
937 unsigned int rx_pages_per_batch
;
938 unsigned int rx_prefix_size
;
939 int rx_packet_hash_offset
;
940 int rx_packet_len_offset
;
941 int rx_packet_ts_offset
;
943 struct efx_rss_context rss_context
;
944 struct mutex rss_lock
;
946 unsigned int_error_count
;
947 unsigned long int_error_expire
;
949 bool irq_soft_enabled
;
950 struct efx_buffer irq_status
;
951 unsigned irq_zero_count
;
953 struct delayed_work selftest_work
;
955 #ifdef CONFIG_SFC_MTD
956 struct list_head mtd_list
;
960 struct efx_mcdi_data
*mcdi
;
962 struct mutex mac_lock
;
963 struct work_struct mac_work
;
966 bool mc_bist_for_other_fn
;
967 bool port_initialized
;
968 struct net_device
*net_dev
;
970 netdev_features_t fixed_features
;
973 struct efx_buffer stats_buffer
;
974 u64 rx_nodesc_drops_total
;
975 u64 rx_nodesc_drops_while_down
;
976 bool rx_nodesc_drops_prev_state
;
978 unsigned int phy_type
;
979 const struct efx_phy_operations
*phy_op
;
981 struct mdio_if_info mdio
;
982 unsigned int mdio_bus
;
983 enum efx_phy_mode phy_mode
;
985 __ETHTOOL_DECLARE_LINK_MODE_MASK(link_advertising
);
987 struct efx_link_state link_state
;
988 unsigned int n_link_state_changes
;
991 union efx_multicast_hash multicast_hash
;
996 enum efx_loopback_mode loopback_mode
;
999 void *loopback_selftest
;
1001 struct rw_semaphore filter_sem
;
1003 #ifdef CONFIG_RFS_ACCEL
1004 struct mutex rps_mutex
;
1005 unsigned int rps_expire_channel
;
1006 unsigned int rps_expire_index
;
1009 atomic_t active_queues
;
1010 atomic_t rxq_flush_pending
;
1011 atomic_t rxq_flush_outstanding
;
1012 wait_queue_head_t flush_wq
;
1014 #ifdef CONFIG_SFC_SRIOV
1016 unsigned vf_init_count
;
1020 struct efx_ptp_data
*ptp_data
;
1025 /* The following fields may be written more often */
1027 struct delayed_work monitor_work ____cacheline_aligned_in_smp
;
1028 spinlock_t biu_lock
;
1030 spinlock_t stats_lock
;
1031 atomic_t n_rx_noskb_drops
;
1034 static inline int efx_dev_registered(struct efx_nic
*efx
)
1036 return efx
->net_dev
->reg_state
== NETREG_REGISTERED
;
1039 static inline unsigned int efx_port_num(struct efx_nic
*efx
)
1041 return efx
->port_num
;
1044 struct efx_mtd_partition
{
1045 struct list_head node
;
1046 struct mtd_info mtd
;
1047 const char *dev_type_name
;
1048 const char *type_name
;
1049 char name
[IFNAMSIZ
+ 20];
1052 struct efx_udp_tunnel
{
1053 u16 type
; /* TUNNEL_ENCAP_UDP_PORT_ENTRY_foo, see mcdi_pcol.h */
1055 /* Count of repeated adds of the same port. Used only inside the list,
1056 * not in request arguments.
1062 * struct efx_nic_type - Efx device type definition
1063 * @mem_bar: Get the memory BAR
1064 * @mem_map_size: Get memory BAR mapped size
1065 * @probe: Probe the controller
1066 * @remove: Free resources allocated by probe()
1067 * @init: Initialise the controller
1068 * @dimension_resources: Dimension controller resources (buffer table,
1069 * and VIs once the available interrupt resources are clear)
1070 * @fini: Shut down the controller
1071 * @monitor: Periodic function for polling link state and hardware monitor
1072 * @map_reset_reason: Map ethtool reset reason to a reset method
1073 * @map_reset_flags: Map ethtool reset flags to a reset method, if possible
1074 * @reset: Reset the controller hardware and possibly the PHY. This will
1075 * be called while the controller is uninitialised.
1076 * @probe_port: Probe the MAC and PHY
1077 * @remove_port: Free resources allocated by probe_port()
1078 * @handle_global_event: Handle a "global" event (may be %NULL)
1079 * @fini_dmaq: Flush and finalise DMA queues (RX and TX queues)
1080 * @prepare_flush: Prepare the hardware for flushing the DMA queues
1081 * (for Falcon architecture)
1082 * @finish_flush: Clean up after flushing the DMA queues (for Falcon
1084 * @prepare_flr: Prepare for an FLR
1085 * @finish_flr: Clean up after an FLR
1086 * @describe_stats: Describe statistics for ethtool
1087 * @update_stats: Update statistics not provided by event handling.
1088 * Either argument may be %NULL.
1089 * @start_stats: Start the regular fetching of statistics
1090 * @pull_stats: Pull stats from the NIC and wait until they arrive.
1091 * @stop_stats: Stop the regular fetching of statistics
1092 * @set_id_led: Set state of identifying LED or revert to automatic function
1093 * @push_irq_moderation: Apply interrupt moderation value
1094 * @reconfigure_port: Push loopback/power/txdis changes to the MAC and PHY
1095 * @prepare_enable_fc_tx: Prepare MAC to enable pause frame TX (may be %NULL)
1096 * @reconfigure_mac: Push MAC address, MTU, flow control and filter settings
1097 * to the hardware. Serialised by the mac_lock.
1098 * @check_mac_fault: Check MAC fault state. True if fault present.
1099 * @get_wol: Get WoL configuration from driver state
1100 * @set_wol: Push WoL configuration to the NIC
1101 * @resume_wol: Synchronise WoL state between driver and MC (e.g. after resume)
1102 * @test_chip: Test registers. May use efx_farch_test_registers(), and is
1103 * expected to reset the NIC.
1104 * @test_nvram: Test validity of NVRAM contents
1105 * @mcdi_request: Send an MCDI request with the given header and SDU.
1106 * The SDU length may be any value from 0 up to the protocol-
1107 * defined maximum, but its buffer will be padded to a multiple
1109 * @mcdi_poll_response: Test whether an MCDI response is available.
1110 * @mcdi_read_response: Read the MCDI response PDU. The offset will
1111 * be a multiple of 4. The length may not be, but the buffer
1112 * will be padded so it is safe to round up.
1113 * @mcdi_poll_reboot: Test whether the MCDI has rebooted. If so,
1114 * return an appropriate error code for aborting any current
1115 * request; otherwise return 0.
1116 * @irq_enable_master: Enable IRQs on the NIC. Each event queue must
1117 * be separately enabled after this.
1118 * @irq_test_generate: Generate a test IRQ
1119 * @irq_disable_non_ev: Disable non-event IRQs on the NIC. Each event
1120 * queue must be separately disabled before this.
1121 * @irq_handle_msi: Handle MSI for a channel. The @dev_id argument is
1122 * a pointer to the &struct efx_msi_context for the channel.
1123 * @irq_handle_legacy: Handle legacy interrupt. The @dev_id argument
1124 * is a pointer to the &struct efx_nic.
1125 * @tx_probe: Allocate resources for TX queue
1126 * @tx_init: Initialise TX queue on the NIC
1127 * @tx_remove: Free resources for TX queue
1128 * @tx_write: Write TX descriptors and doorbell
1129 * @rx_push_rss_config: Write RSS hash key and indirection table to the NIC
1130 * @rx_pull_rss_config: Read RSS hash key and indirection table back from the NIC
1131 * @rx_push_rss_context_config: Write RSS hash key and indirection table for
1132 * user RSS context to the NIC
1133 * @rx_pull_rss_context_config: Read RSS hash key and indirection table for user
1134 * RSS context back from the NIC
1135 * @rx_probe: Allocate resources for RX queue
1136 * @rx_init: Initialise RX queue on the NIC
1137 * @rx_remove: Free resources for RX queue
1138 * @rx_write: Write RX descriptors and doorbell
1139 * @rx_defer_refill: Generate a refill reminder event
1140 * @ev_probe: Allocate resources for event queue
1141 * @ev_init: Initialise event queue on the NIC
1142 * @ev_fini: Deinitialise event queue on the NIC
1143 * @ev_remove: Free resources for event queue
1144 * @ev_process: Process events for a queue, up to the given NAPI quota
1145 * @ev_read_ack: Acknowledge read events on a queue, rearming its IRQ
1146 * @ev_test_generate: Generate a test event
1147 * @filter_table_probe: Probe filter capabilities and set up filter software state
1148 * @filter_table_restore: Restore filters removed from hardware
1149 * @filter_table_remove: Remove filters from hardware and tear down software state
1150 * @filter_update_rx_scatter: Update filters after change to rx scatter setting
1151 * @filter_insert: add or replace a filter
1152 * @filter_remove_safe: remove a filter by ID, carefully
1153 * @filter_get_safe: retrieve a filter by ID, carefully
1154 * @filter_clear_rx: Remove all RX filters whose priority is less than or
1155 * equal to the given priority and is not %EFX_FILTER_PRI_AUTO
1156 * @filter_count_rx_used: Get the number of filters in use at a given priority
1157 * @filter_get_rx_id_limit: Get maximum value of a filter id, plus 1
1158 * @filter_get_rx_ids: Get list of RX filters at a given priority
1159 * @filter_rfs_expire_one: Consider expiring a filter inserted for RFS.
1160 * This must check whether the specified table entry is used by RFS
1161 * and that rps_may_expire_flow() returns true for it.
1162 * @mtd_probe: Probe and add MTD partitions associated with this net device,
1163 * using efx_mtd_add()
1164 * @mtd_rename: Set an MTD partition name using the net device name
1165 * @mtd_read: Read from an MTD partition
1166 * @mtd_erase: Erase part of an MTD partition
1167 * @mtd_write: Write to an MTD partition
1168 * @mtd_sync: Wait for write-back to complete on MTD partition. This
1169 * also notifies the driver that a writer has finished using this
1171 * @ptp_write_host_time: Send host time to MC as part of sync protocol
1172 * @ptp_set_ts_sync_events: Enable or disable sync events for inline RX
1173 * timestamping, possibly only temporarily for the purposes of a reset.
1174 * @ptp_set_ts_config: Set hardware timestamp configuration. The flags
1175 * and tx_type will already have been validated but this operation
1176 * must validate and update rx_filter.
1177 * @get_phys_port_id: Get the underlying physical port id.
1178 * @set_mac_address: Set the MAC address of the device
1179 * @tso_versions: Returns mask of firmware-assisted TSO versions supported.
1180 * If %NULL, then device does not support any TSO version.
1181 * @udp_tnl_push_ports: Push the list of UDP tunnel ports to the NIC if required.
1182 * @udp_tnl_add_port: Add a UDP tunnel port
1183 * @udp_tnl_has_port: Check if a port has been added as UDP tunnel
1184 * @udp_tnl_del_port: Remove a UDP tunnel port
1185 * @revision: Hardware architecture revision
1186 * @txd_ptr_tbl_base: TX descriptor ring base address
1187 * @rxd_ptr_tbl_base: RX descriptor ring base address
1188 * @buf_tbl_base: Buffer table base address
1189 * @evq_ptr_tbl_base: Event queue pointer table base address
1190 * @evq_rptr_tbl_base: Event queue read-pointer table base address
1191 * @max_dma_mask: Maximum possible DMA mask
1192 * @rx_prefix_size: Size of RX prefix before packet data
1193 * @rx_hash_offset: Offset of RX flow hash within prefix
1194 * @rx_ts_offset: Offset of timestamp within prefix
1195 * @rx_buffer_padding: Size of padding at end of RX packet
1196 * @can_rx_scatter: NIC is able to scatter packets to multiple buffers
1197 * @always_rx_scatter: NIC will always scatter packets to multiple buffers
1198 * @option_descriptors: NIC supports TX option descriptors
1199 * @min_interrupt_mode: Lowest capability interrupt mode supported
1200 * from &enum efx_int_mode.
1201 * @max_interrupt_mode: Highest capability interrupt mode supported
1202 * from &enum efx_int_mode.
1203 * @timer_period_max: Maximum period of interrupt timer (in ticks)
1204 * @offload_features: net_device feature flags for protocol offload
1205 * features implemented in hardware
1206 * @mcdi_max_ver: Maximum MCDI version supported
1207 * @hwtstamp_filters: Mask of hardware timestamp filter types supported
1209 struct efx_nic_type
{
1211 unsigned int (*mem_bar
)(struct efx_nic
*efx
);
1212 unsigned int (*mem_map_size
)(struct efx_nic
*efx
);
1213 int (*probe
)(struct efx_nic
*efx
);
1214 void (*remove
)(struct efx_nic
*efx
);
1215 int (*init
)(struct efx_nic
*efx
);
1216 int (*dimension_resources
)(struct efx_nic
*efx
);
1217 void (*fini
)(struct efx_nic
*efx
);
1218 void (*monitor
)(struct efx_nic
*efx
);
1219 enum reset_type (*map_reset_reason
)(enum reset_type reason
);
1220 int (*map_reset_flags
)(u32
*flags
);
1221 int (*reset
)(struct efx_nic
*efx
, enum reset_type method
);
1222 int (*probe_port
)(struct efx_nic
*efx
);
1223 void (*remove_port
)(struct efx_nic
*efx
);
1224 bool (*handle_global_event
)(struct efx_channel
*channel
, efx_qword_t
*);
1225 int (*fini_dmaq
)(struct efx_nic
*efx
);
1226 void (*prepare_flush
)(struct efx_nic
*efx
);
1227 void (*finish_flush
)(struct efx_nic
*efx
);
1228 void (*prepare_flr
)(struct efx_nic
*efx
);
1229 void (*finish_flr
)(struct efx_nic
*efx
);
1230 size_t (*describe_stats
)(struct efx_nic
*efx
, u8
*names
);
1231 size_t (*update_stats
)(struct efx_nic
*efx
, u64
*full_stats
,
1232 struct rtnl_link_stats64
*core_stats
);
1233 void (*start_stats
)(struct efx_nic
*efx
);
1234 void (*pull_stats
)(struct efx_nic
*efx
);
1235 void (*stop_stats
)(struct efx_nic
*efx
);
1236 void (*set_id_led
)(struct efx_nic
*efx
, enum efx_led_mode mode
);
1237 void (*push_irq_moderation
)(struct efx_channel
*channel
);
1238 int (*reconfigure_port
)(struct efx_nic
*efx
);
1239 void (*prepare_enable_fc_tx
)(struct efx_nic
*efx
);
1240 int (*reconfigure_mac
)(struct efx_nic
*efx
);
1241 bool (*check_mac_fault
)(struct efx_nic
*efx
);
1242 void (*get_wol
)(struct efx_nic
*efx
, struct ethtool_wolinfo
*wol
);
1243 int (*set_wol
)(struct efx_nic
*efx
, u32 type
);
1244 void (*resume_wol
)(struct efx_nic
*efx
);
1245 int (*test_chip
)(struct efx_nic
*efx
, struct efx_self_tests
*tests
);
1246 int (*test_nvram
)(struct efx_nic
*efx
);
1247 void (*mcdi_request
)(struct efx_nic
*efx
,
1248 const efx_dword_t
*hdr
, size_t hdr_len
,
1249 const efx_dword_t
*sdu
, size_t sdu_len
);
1250 bool (*mcdi_poll_response
)(struct efx_nic
*efx
);
1251 void (*mcdi_read_response
)(struct efx_nic
*efx
, efx_dword_t
*pdu
,
1252 size_t pdu_offset
, size_t pdu_len
);
1253 int (*mcdi_poll_reboot
)(struct efx_nic
*efx
);
1254 void (*mcdi_reboot_detected
)(struct efx_nic
*efx
);
1255 void (*irq_enable_master
)(struct efx_nic
*efx
);
1256 int (*irq_test_generate
)(struct efx_nic
*efx
);
1257 void (*irq_disable_non_ev
)(struct efx_nic
*efx
);
1258 irqreturn_t (*irq_handle_msi
)(int irq
, void *dev_id
);
1259 irqreturn_t (*irq_handle_legacy
)(int irq
, void *dev_id
);
1260 int (*tx_probe
)(struct efx_tx_queue
*tx_queue
);
1261 void (*tx_init
)(struct efx_tx_queue
*tx_queue
);
1262 void (*tx_remove
)(struct efx_tx_queue
*tx_queue
);
1263 void (*tx_write
)(struct efx_tx_queue
*tx_queue
);
1264 unsigned int (*tx_limit_len
)(struct efx_tx_queue
*tx_queue
,
1265 dma_addr_t dma_addr
, unsigned int len
);
1266 int (*rx_push_rss_config
)(struct efx_nic
*efx
, bool user
,
1267 const u32
*rx_indir_table
, const u8
*key
);
1268 int (*rx_pull_rss_config
)(struct efx_nic
*efx
);
1269 int (*rx_push_rss_context_config
)(struct efx_nic
*efx
,
1270 struct efx_rss_context
*ctx
,
1271 const u32
*rx_indir_table
,
1273 int (*rx_pull_rss_context_config
)(struct efx_nic
*efx
,
1274 struct efx_rss_context
*ctx
);
1275 void (*rx_restore_rss_contexts
)(struct efx_nic
*efx
);
1276 int (*rx_probe
)(struct efx_rx_queue
*rx_queue
);
1277 void (*rx_init
)(struct efx_rx_queue
*rx_queue
);
1278 void (*rx_remove
)(struct efx_rx_queue
*rx_queue
);
1279 void (*rx_write
)(struct efx_rx_queue
*rx_queue
);
1280 void (*rx_defer_refill
)(struct efx_rx_queue
*rx_queue
);
1281 int (*ev_probe
)(struct efx_channel
*channel
);
1282 int (*ev_init
)(struct efx_channel
*channel
);
1283 void (*ev_fini
)(struct efx_channel
*channel
);
1284 void (*ev_remove
)(struct efx_channel
*channel
);
1285 int (*ev_process
)(struct efx_channel
*channel
, int quota
);
1286 void (*ev_read_ack
)(struct efx_channel
*channel
);
1287 void (*ev_test_generate
)(struct efx_channel
*channel
);
1288 int (*filter_table_probe
)(struct efx_nic
*efx
);
1289 void (*filter_table_restore
)(struct efx_nic
*efx
);
1290 void (*filter_table_remove
)(struct efx_nic
*efx
);
1291 void (*filter_update_rx_scatter
)(struct efx_nic
*efx
);
1292 s32 (*filter_insert
)(struct efx_nic
*efx
,
1293 struct efx_filter_spec
*spec
, bool replace
);
1294 int (*filter_remove_safe
)(struct efx_nic
*efx
,
1295 enum efx_filter_priority priority
,
1297 int (*filter_get_safe
)(struct efx_nic
*efx
,
1298 enum efx_filter_priority priority
,
1299 u32 filter_id
, struct efx_filter_spec
*);
1300 int (*filter_clear_rx
)(struct efx_nic
*efx
,
1301 enum efx_filter_priority priority
);
1302 u32 (*filter_count_rx_used
)(struct efx_nic
*efx
,
1303 enum efx_filter_priority priority
);
1304 u32 (*filter_get_rx_id_limit
)(struct efx_nic
*efx
);
1305 s32 (*filter_get_rx_ids
)(struct efx_nic
*efx
,
1306 enum efx_filter_priority priority
,
1307 u32
*buf
, u32 size
);
1308 #ifdef CONFIG_RFS_ACCEL
1309 bool (*filter_rfs_expire_one
)(struct efx_nic
*efx
, u32 flow_id
,
1310 unsigned int index
);
1312 #ifdef CONFIG_SFC_MTD
1313 int (*mtd_probe
)(struct efx_nic
*efx
);
1314 void (*mtd_rename
)(struct efx_mtd_partition
*part
);
1315 int (*mtd_read
)(struct mtd_info
*mtd
, loff_t start
, size_t len
,
1316 size_t *retlen
, u8
*buffer
);
1317 int (*mtd_erase
)(struct mtd_info
*mtd
, loff_t start
, size_t len
);
1318 int (*mtd_write
)(struct mtd_info
*mtd
, loff_t start
, size_t len
,
1319 size_t *retlen
, const u8
*buffer
);
1320 int (*mtd_sync
)(struct mtd_info
*mtd
);
1322 void (*ptp_write_host_time
)(struct efx_nic
*efx
, u32 host_time
);
1323 int (*ptp_set_ts_sync_events
)(struct efx_nic
*efx
, bool en
, bool temp
);
1324 int (*ptp_set_ts_config
)(struct efx_nic
*efx
,
1325 struct hwtstamp_config
*init
);
1326 int (*sriov_configure
)(struct efx_nic
*efx
, int num_vfs
);
1327 int (*vlan_rx_add_vid
)(struct efx_nic
*efx
, __be16 proto
, u16 vid
);
1328 int (*vlan_rx_kill_vid
)(struct efx_nic
*efx
, __be16 proto
, u16 vid
);
1329 int (*get_phys_port_id
)(struct efx_nic
*efx
,
1330 struct netdev_phys_item_id
*ppid
);
1331 int (*sriov_init
)(struct efx_nic
*efx
);
1332 void (*sriov_fini
)(struct efx_nic
*efx
);
1333 bool (*sriov_wanted
)(struct efx_nic
*efx
);
1334 void (*sriov_reset
)(struct efx_nic
*efx
);
1335 void (*sriov_flr
)(struct efx_nic
*efx
, unsigned vf_i
);
1336 int (*sriov_set_vf_mac
)(struct efx_nic
*efx
, int vf_i
, u8
*mac
);
1337 int (*sriov_set_vf_vlan
)(struct efx_nic
*efx
, int vf_i
, u16 vlan
,
1339 int (*sriov_set_vf_spoofchk
)(struct efx_nic
*efx
, int vf_i
,
1341 int (*sriov_get_vf_config
)(struct efx_nic
*efx
, int vf_i
,
1342 struct ifla_vf_info
*ivi
);
1343 int (*sriov_set_vf_link_state
)(struct efx_nic
*efx
, int vf_i
,
1345 int (*vswitching_probe
)(struct efx_nic
*efx
);
1346 int (*vswitching_restore
)(struct efx_nic
*efx
);
1347 void (*vswitching_remove
)(struct efx_nic
*efx
);
1348 int (*get_mac_address
)(struct efx_nic
*efx
, unsigned char *perm_addr
);
1349 int (*set_mac_address
)(struct efx_nic
*efx
);
1350 u32 (*tso_versions
)(struct efx_nic
*efx
);
1351 int (*udp_tnl_push_ports
)(struct efx_nic
*efx
);
1352 int (*udp_tnl_add_port
)(struct efx_nic
*efx
, struct efx_udp_tunnel tnl
);
1353 bool (*udp_tnl_has_port
)(struct efx_nic
*efx
, __be16 port
);
1354 int (*udp_tnl_del_port
)(struct efx_nic
*efx
, struct efx_udp_tunnel tnl
);
1357 unsigned int txd_ptr_tbl_base
;
1358 unsigned int rxd_ptr_tbl_base
;
1359 unsigned int buf_tbl_base
;
1360 unsigned int evq_ptr_tbl_base
;
1361 unsigned int evq_rptr_tbl_base
;
1363 unsigned int rx_prefix_size
;
1364 unsigned int rx_hash_offset
;
1365 unsigned int rx_ts_offset
;
1366 unsigned int rx_buffer_padding
;
1367 bool can_rx_scatter
;
1368 bool always_rx_scatter
;
1369 bool option_descriptors
;
1370 unsigned int min_interrupt_mode
;
1371 unsigned int max_interrupt_mode
;
1372 unsigned int timer_period_max
;
1373 netdev_features_t offload_features
;
1375 unsigned int max_rx_ip_filters
;
1376 u32 hwtstamp_filters
;
1377 unsigned int rx_hash_key_size
;
1380 /**************************************************************************
1382 * Prototypes and inline functions
1384 *************************************************************************/
1386 static inline struct efx_channel
*
1387 efx_get_channel(struct efx_nic
*efx
, unsigned index
)
1389 EFX_WARN_ON_ONCE_PARANOID(index
>= efx
->n_channels
);
1390 return efx
->channel
[index
];
1393 /* Iterate over all used channels */
1394 #define efx_for_each_channel(_channel, _efx) \
1395 for (_channel = (_efx)->channel[0]; \
1397 _channel = (_channel->channel + 1 < (_efx)->n_channels) ? \
1398 (_efx)->channel[_channel->channel + 1] : NULL)
1400 /* Iterate over all used channels in reverse */
1401 #define efx_for_each_channel_rev(_channel, _efx) \
1402 for (_channel = (_efx)->channel[(_efx)->n_channels - 1]; \
1404 _channel = _channel->channel ? \
1405 (_efx)->channel[_channel->channel - 1] : NULL)
1407 static inline struct efx_tx_queue
*
1408 efx_get_tx_queue(struct efx_nic
*efx
, unsigned index
, unsigned type
)
1410 EFX_WARN_ON_ONCE_PARANOID(index
>= efx
->n_tx_channels
||
1411 type
>= EFX_TXQ_TYPES
);
1412 return &efx
->channel
[efx
->tx_channel_offset
+ index
]->tx_queue
[type
];
1415 static inline bool efx_channel_has_tx_queues(struct efx_channel
*channel
)
1417 return channel
->type
&& channel
->type
->want_txqs
&&
1418 channel
->type
->want_txqs(channel
);
1421 static inline struct efx_tx_queue
*
1422 efx_channel_get_tx_queue(struct efx_channel
*channel
, unsigned type
)
1424 EFX_WARN_ON_ONCE_PARANOID(!efx_channel_has_tx_queues(channel
) ||
1425 type
>= EFX_TXQ_TYPES
);
1426 return &channel
->tx_queue
[type
];
1429 static inline bool efx_tx_queue_used(struct efx_tx_queue
*tx_queue
)
1431 return !(tx_queue
->efx
->net_dev
->num_tc
< 2 &&
1432 tx_queue
->queue
& EFX_TXQ_TYPE_HIGHPRI
);
1435 /* Iterate over all TX queues belonging to a channel */
1436 #define efx_for_each_channel_tx_queue(_tx_queue, _channel) \
1437 if (!efx_channel_has_tx_queues(_channel)) \
1440 for (_tx_queue = (_channel)->tx_queue; \
1441 _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES && \
1442 efx_tx_queue_used(_tx_queue); \
1445 /* Iterate over all possible TX queues belonging to a channel */
1446 #define efx_for_each_possible_channel_tx_queue(_tx_queue, _channel) \
1447 if (!efx_channel_has_tx_queues(_channel)) \
1450 for (_tx_queue = (_channel)->tx_queue; \
1451 _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES; \
1454 static inline bool efx_channel_has_rx_queue(struct efx_channel
*channel
)
1456 return channel
->rx_queue
.core_index
>= 0;
1459 static inline struct efx_rx_queue
*
1460 efx_channel_get_rx_queue(struct efx_channel
*channel
)
1462 EFX_WARN_ON_ONCE_PARANOID(!efx_channel_has_rx_queue(channel
));
1463 return &channel
->rx_queue
;
1466 /* Iterate over all RX queues belonging to a channel */
1467 #define efx_for_each_channel_rx_queue(_rx_queue, _channel) \
1468 if (!efx_channel_has_rx_queue(_channel)) \
1471 for (_rx_queue = &(_channel)->rx_queue; \
1475 static inline struct efx_channel
*
1476 efx_rx_queue_channel(struct efx_rx_queue
*rx_queue
)
1478 return container_of(rx_queue
, struct efx_channel
, rx_queue
);
1481 static inline int efx_rx_queue_index(struct efx_rx_queue
*rx_queue
)
1483 return efx_rx_queue_channel(rx_queue
)->channel
;
1486 /* Returns a pointer to the specified receive buffer in the RX
1489 static inline struct efx_rx_buffer
*efx_rx_buffer(struct efx_rx_queue
*rx_queue
,
1492 return &rx_queue
->buffer
[index
];
1496 * EFX_MAX_FRAME_LEN - calculate maximum frame length
1498 * This calculates the maximum frame length that will be used for a
1499 * given MTU. The frame length will be equal to the MTU plus a
1500 * constant amount of header space and padding. This is the quantity
1501 * that the net driver will program into the MAC as the maximum frame
1504 * The 10G MAC requires 8-byte alignment on the frame
1505 * length, so we round up to the nearest 8.
1507 * Re-clocking by the XGXS on RX can reduce an IPG to 32 bits (half an
1508 * XGMII cycle). If the frame length reaches the maximum value in the
1509 * same cycle, the XMAC can miss the IPG altogether. We work around
1510 * this by adding a further 16 bytes.
1512 #define EFX_FRAME_PAD 16
1513 #define EFX_MAX_FRAME_LEN(mtu) \
1514 (ALIGN(((mtu) + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN + EFX_FRAME_PAD), 8))
1516 static inline bool efx_xmit_with_hwtstamp(struct sk_buff
*skb
)
1518 return skb_shinfo(skb
)->tx_flags
& SKBTX_HW_TSTAMP
;
1520 static inline void efx_xmit_hwtstamp_pending(struct sk_buff
*skb
)
1522 skb_shinfo(skb
)->tx_flags
|= SKBTX_IN_PROGRESS
;
1525 /* Get all supported features.
1526 * If a feature is not fixed, it is present in hw_features.
1527 * If a feature is fixed, it does not present in hw_features, but
1528 * always in features.
1530 static inline netdev_features_t
efx_supported_features(const struct efx_nic
*efx
)
1532 const struct net_device
*net_dev
= efx
->net_dev
;
1534 return net_dev
->features
| net_dev
->hw_features
;
1537 /* Get the current TX queue insert index. */
1538 static inline unsigned int
1539 efx_tx_queue_get_insert_index(const struct efx_tx_queue
*tx_queue
)
1541 return tx_queue
->insert_count
& tx_queue
->ptr_mask
;
1544 /* Get a TX buffer. */
1545 static inline struct efx_tx_buffer
*
1546 __efx_tx_queue_get_insert_buffer(const struct efx_tx_queue
*tx_queue
)
1548 return &tx_queue
->buffer
[efx_tx_queue_get_insert_index(tx_queue
)];
1551 /* Get a TX buffer, checking it's not currently in use. */
1552 static inline struct efx_tx_buffer
*
1553 efx_tx_queue_get_insert_buffer(const struct efx_tx_queue
*tx_queue
)
1555 struct efx_tx_buffer
*buffer
=
1556 __efx_tx_queue_get_insert_buffer(tx_queue
);
1558 EFX_WARN_ON_ONCE_PARANOID(buffer
->len
);
1559 EFX_WARN_ON_ONCE_PARANOID(buffer
->flags
);
1560 EFX_WARN_ON_ONCE_PARANOID(buffer
->unmap_len
);
1565 #endif /* EFX_NET_DRIVER_H */