1 /****************************************************************************
2 * Driver for Solarflare Solarstorm network controllers and boards
3 * Copyright 2005-2006 Fen Systems Ltd.
4 * Copyright 2005-2011 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/vmalloc.h>
29 #include <linux/i2c.h>
34 /**************************************************************************
38 **************************************************************************/
40 #define EFX_DRIVER_VERSION "3.2"
43 #define EFX_BUG_ON_PARANOID(x) BUG_ON(x)
44 #define EFX_WARN_ON_PARANOID(x) WARN_ON(x)
46 #define EFX_BUG_ON_PARANOID(x) do {} while (0)
47 #define EFX_WARN_ON_PARANOID(x) do {} while (0)
50 /**************************************************************************
54 **************************************************************************/
56 #define EFX_MAX_CHANNELS 32U
57 #define EFX_MAX_RX_QUEUES EFX_MAX_CHANNELS
58 #define EFX_EXTRA_CHANNEL_IOV 0
59 #define EFX_EXTRA_CHANNEL_PTP 1
60 #define EFX_MAX_EXTRA_CHANNELS 2U
62 /* Checksum generation is a per-queue option in hardware, so each
63 * queue visible to the networking core is backed by two hardware TX
65 #define EFX_MAX_TX_TC 2
66 #define EFX_MAX_CORE_TX_QUEUES (EFX_MAX_TX_TC * EFX_MAX_CHANNELS)
67 #define EFX_TXQ_TYPE_OFFLOAD 1 /* flag */
68 #define EFX_TXQ_TYPE_HIGHPRI 2 /* flag */
69 #define EFX_TXQ_TYPES 4
70 #define EFX_MAX_TX_QUEUES (EFX_TXQ_TYPES * EFX_MAX_CHANNELS)
72 /* Forward declare Precision Time Protocol (PTP) support structure. */
75 struct efx_self_tests
;
78 * struct efx_special_buffer - An Efx special buffer
79 * @addr: CPU base address of the buffer
80 * @dma_addr: DMA base address of the buffer
81 * @len: Buffer length, in bytes
82 * @index: Buffer index within controller;s buffer table
83 * @entries: Number of buffer table entries
85 * Special buffers are used for the event queues and the TX and RX
86 * descriptor queues for each channel. They are *not* used for the
87 * actual transmit and receive buffers.
89 struct efx_special_buffer
{
98 * struct efx_tx_buffer - buffer state for a TX descriptor
99 * @skb: When @flags & %EFX_TX_BUF_SKB, the associated socket buffer to be
100 * freed when descriptor completes
101 * @heap_buf: When @flags & %EFX_TX_BUF_HEAP, the associated heap buffer to be
102 * freed when descriptor completes.
103 * @dma_addr: DMA address of the fragment.
104 * @flags: Flags for allocation and DMA mapping type
105 * @len: Length of this fragment.
106 * This field is zero when the queue slot is empty.
107 * @unmap_len: Length of this fragment to unmap
109 struct efx_tx_buffer
{
111 const struct sk_buff
*skb
;
115 unsigned short flags
;
117 unsigned short unmap_len
;
119 #define EFX_TX_BUF_CONT 1 /* not last descriptor of packet */
120 #define EFX_TX_BUF_SKB 2 /* buffer is last part of skb */
121 #define EFX_TX_BUF_HEAP 4 /* buffer was allocated with kmalloc() */
122 #define EFX_TX_BUF_MAP_SINGLE 8 /* buffer was mapped with dma_map_single() */
125 * struct efx_tx_queue - An Efx TX queue
127 * This is a ring buffer of TX fragments.
128 * Since the TX completion path always executes on the same
129 * CPU and the xmit path can operate on different CPUs,
130 * performance is increased by ensuring that the completion
131 * path and the xmit path operate on different cache lines.
132 * This is particularly important if the xmit path is always
133 * executing on one CPU which is different from the completion
134 * path. There is also a cache line for members which are
135 * read but not written on the fast path.
137 * @efx: The associated Efx NIC
138 * @queue: DMA queue number
139 * @channel: The associated channel
140 * @core_txq: The networking core TX queue structure
141 * @buffer: The software buffer ring
142 * @tsoh_page: Array of pages of TSO header buffers
143 * @txd: The hardware descriptor ring
144 * @ptr_mask: The size of the ring minus 1.
145 * @initialised: Has hardware queue been initialised?
146 * @read_count: Current read pointer.
147 * This is the number of buffers that have been removed from both rings.
148 * @old_write_count: The value of @write_count when last checked.
149 * This is here for performance reasons. The xmit path will
150 * only get the up-to-date value of @write_count if this
151 * variable indicates that the queue is empty. This is to
152 * avoid cache-line ping-pong between the xmit path and the
154 * @insert_count: Current insert pointer
155 * This is the number of buffers that have been added to the
157 * @write_count: Current write pointer
158 * This is the number of buffers that have been added to the
160 * @old_read_count: The value of read_count when last checked.
161 * This is here for performance reasons. The xmit path will
162 * only get the up-to-date value of read_count if this
163 * variable indicates that the queue is full. This is to
164 * avoid cache-line ping-pong between the xmit path and the
166 * @tso_bursts: Number of times TSO xmit invoked by kernel
167 * @tso_long_headers: Number of packets with headers too long for standard
169 * @tso_packets: Number of packets via the TSO xmit path
170 * @pushes: Number of times the TX push feature has been used
171 * @empty_read_count: If the completion path has seen the queue as empty
172 * and the transmission path has not yet checked this, the value of
173 * @read_count bitwise-added to %EFX_EMPTY_COUNT_VALID; otherwise 0.
175 struct efx_tx_queue
{
176 /* Members which don't change on the fast path */
177 struct efx_nic
*efx ____cacheline_aligned_in_smp
;
179 struct efx_channel
*channel
;
180 struct netdev_queue
*core_txq
;
181 struct efx_tx_buffer
*buffer
;
182 struct efx_buffer
*tsoh_page
;
183 struct efx_special_buffer txd
;
184 unsigned int ptr_mask
;
187 /* Members used mainly on the completion path */
188 unsigned int read_count ____cacheline_aligned_in_smp
;
189 unsigned int old_write_count
;
191 /* Members used only on the xmit path */
192 unsigned int insert_count ____cacheline_aligned_in_smp
;
193 unsigned int write_count
;
194 unsigned int old_read_count
;
195 unsigned int tso_bursts
;
196 unsigned int tso_long_headers
;
197 unsigned int tso_packets
;
200 /* Members shared between paths and sometimes updated */
201 unsigned int empty_read_count ____cacheline_aligned_in_smp
;
202 #define EFX_EMPTY_COUNT_VALID 0x80000000
203 atomic_t flush_outstanding
;
207 * struct efx_rx_buffer - An Efx RX data buffer
208 * @dma_addr: DMA base address of the buffer
209 * @page: The associated page buffer.
210 * Will be %NULL if the buffer slot is currently free.
211 * @page_offset: Offset within page
212 * @len: If pending: length for DMA descriptor.
213 * If completed: received length, excluding hash prefix.
214 * @flags: Flags for buffer and packet state.
216 struct efx_rx_buffer
{
223 #define EFX_RX_PKT_CSUMMED 0x0002
224 #define EFX_RX_PKT_DISCARD 0x0004
227 * struct efx_rx_page_state - Page-based rx buffer state
229 * Inserted at the start of every page allocated for receive buffers.
230 * Used to facilitate sharing dma mappings between recycled rx buffers
231 * and those passed up to the kernel.
233 * @refcnt: Number of struct efx_rx_buffer's referencing this page.
234 * When refcnt falls to zero, the page is unmapped for dma
235 * @dma_addr: The dma address of this page.
237 struct efx_rx_page_state
{
241 unsigned int __pad
[0] ____cacheline_aligned
;
245 * struct efx_rx_queue - An Efx RX queue
246 * @efx: The associated Efx NIC
247 * @core_index: Index of network core RX queue. Will be >= 0 iff this
248 * is associated with a real RX queue.
249 * @buffer: The software buffer ring
250 * @rxd: The hardware descriptor ring
251 * @ptr_mask: The size of the ring minus 1.
252 * @enabled: Receive queue enabled indicator.
253 * @flush_pending: Set when a RX flush is pending. Has the same lifetime as
254 * @rxq_flush_pending.
255 * @added_count: Number of buffers added to the receive queue.
256 * @notified_count: Number of buffers given to NIC (<= @added_count).
257 * @removed_count: Number of buffers removed from the receive queue.
258 * @max_fill: RX descriptor maximum fill level (<= ring size)
259 * @fast_fill_trigger: RX descriptor fill level that will trigger a fast fill
261 * @min_fill: RX descriptor minimum non-zero fill level.
262 * This records the minimum fill level observed when a ring
263 * refill was triggered.
264 * @slow_fill: Timer used to defer efx_nic_generate_fill_event().
266 struct efx_rx_queue
{
269 struct efx_rx_buffer
*buffer
;
270 struct efx_special_buffer rxd
;
271 unsigned int ptr_mask
;
275 unsigned int added_count
;
276 unsigned int notified_count
;
277 unsigned int removed_count
;
278 unsigned int max_fill
;
279 unsigned int fast_fill_trigger
;
280 unsigned int min_fill
;
281 unsigned int min_overfill
;
282 struct timer_list slow_fill
;
283 unsigned int slow_fill_count
;
287 * struct efx_buffer - An Efx general-purpose buffer
288 * @addr: host base address of the buffer
289 * @dma_addr: DMA base address of the buffer
290 * @len: Buffer length, in bytes
292 * The NIC uses these buffers for its interrupt status registers and
302 enum efx_rx_alloc_method
{
303 RX_ALLOC_METHOD_AUTO
= 0,
304 RX_ALLOC_METHOD_SKB
= 1,
305 RX_ALLOC_METHOD_PAGE
= 2,
309 * struct efx_channel - An Efx channel
311 * A channel comprises an event queue, at least one TX queue, at least
312 * one RX queue, and an associated tasklet for processing the event
315 * @efx: Associated Efx NIC
316 * @channel: Channel instance number
317 * @type: Channel type definition
318 * @enabled: Channel enabled indicator
319 * @irq: IRQ number (MSI and MSI-X only)
320 * @irq_moderation: IRQ moderation value (in hardware ticks)
321 * @napi_dev: Net device used with NAPI
322 * @napi_str: NAPI control structure
323 * @work_pending: Is work pending via NAPI?
324 * @eventq: Event queue buffer
325 * @eventq_mask: Event queue pointer mask
326 * @eventq_read_ptr: Event queue read pointer
327 * @event_test_cpu: Last CPU to handle interrupt or test event for this channel
328 * @irq_count: Number of IRQs since last adaptive moderation decision
329 * @irq_mod_score: IRQ moderation score
330 * @n_rx_tobe_disc: Count of RX_TOBE_DISC errors
331 * @n_rx_ip_hdr_chksum_err: Count of RX IP header checksum errors
332 * @n_rx_tcp_udp_chksum_err: Count of RX TCP and UDP checksum errors
333 * @n_rx_mcast_mismatch: Count of unmatched multicast frames
334 * @n_rx_frm_trunc: Count of RX_FRM_TRUNC errors
335 * @n_rx_overlength: Count of RX_OVERLENGTH errors
336 * @n_skbuff_leaks: Count of skbuffs leaked due to RX overrun
337 * @rx_queue: RX queue for this channel
338 * @tx_queue: TX queues for this channel
343 const struct efx_channel_type
*type
;
346 unsigned int irq_moderation
;
347 struct net_device
*napi_dev
;
348 struct napi_struct napi_str
;
350 struct efx_special_buffer eventq
;
351 unsigned int eventq_mask
;
352 unsigned int eventq_read_ptr
;
355 unsigned int irq_count
;
356 unsigned int irq_mod_score
;
357 #ifdef CONFIG_RFS_ACCEL
358 unsigned int rfs_filters_added
;
361 unsigned n_rx_tobe_disc
;
362 unsigned n_rx_ip_hdr_chksum_err
;
363 unsigned n_rx_tcp_udp_chksum_err
;
364 unsigned n_rx_mcast_mismatch
;
365 unsigned n_rx_frm_trunc
;
366 unsigned n_rx_overlength
;
367 unsigned n_skbuff_leaks
;
369 /* Used to pipeline received packets in order to optimise memory
370 * access with prefetches.
372 struct efx_rx_buffer
*rx_pkt
;
374 struct efx_rx_queue rx_queue
;
375 struct efx_tx_queue tx_queue
[EFX_TXQ_TYPES
];
379 * struct efx_channel_type - distinguishes traffic and extra channels
380 * @handle_no_channel: Handle failure to allocate an extra channel
381 * @pre_probe: Set up extra state prior to initialisation
382 * @post_remove: Tear down extra state after finalisation, if allocated.
383 * May be called on channels that have not been probed.
384 * @get_name: Generate the channel's name (used for its IRQ handler)
385 * @copy: Copy the channel state prior to reallocation. May be %NULL if
386 * reallocation is not supported.
387 * @receive_skb: Handle an skb ready to be passed to netif_receive_skb()
388 * @keep_eventq: Flag for whether event queue should be kept initialised
389 * while the device is stopped
391 struct efx_channel_type
{
392 void (*handle_no_channel
)(struct efx_nic
*);
393 int (*pre_probe
)(struct efx_channel
*);
394 void (*post_remove
)(struct efx_channel
*);
395 void (*get_name
)(struct efx_channel
*, char *buf
, size_t len
);
396 struct efx_channel
*(*copy
)(const struct efx_channel
*);
397 bool (*receive_skb
)(struct efx_channel
*, struct sk_buff
*);
407 #define STRING_TABLE_LOOKUP(val, member) \
408 ((val) < member ## _max) ? member ## _names[val] : "(invalid)"
410 extern const char *const efx_loopback_mode_names
[];
411 extern const unsigned int efx_loopback_mode_max
;
412 #define LOOPBACK_MODE(efx) \
413 STRING_TABLE_LOOKUP((efx)->loopback_mode, efx_loopback_mode)
415 extern const char *const efx_reset_type_names
[];
416 extern const unsigned int efx_reset_type_max
;
417 #define RESET_TYPE(type) \
418 STRING_TABLE_LOOKUP(type, efx_reset_type)
421 /* Be careful if altering to correct macro below */
422 EFX_INT_MODE_MSIX
= 0,
423 EFX_INT_MODE_MSI
= 1,
424 EFX_INT_MODE_LEGACY
= 2,
425 EFX_INT_MODE_MAX
/* Insert any new items before this */
427 #define EFX_INT_MODE_USE_MSI(x) (((x)->interrupt_mode) <= EFX_INT_MODE_MSI)
430 STATE_UNINIT
= 0, /* device being probed/removed or is frozen */
431 STATE_READY
= 1, /* hardware ready and netdev registered */
432 STATE_DISABLED
= 2, /* device disabled due to hardware errors */
433 STATE_RECOVERY
= 3, /* device recovering from PCI error */
437 * Alignment of page-allocated RX buffers
439 * Controls the number of bytes inserted at the start of an RX buffer.
440 * This is the equivalent of NET_IP_ALIGN [which controls the alignment
441 * of the skb->head for hardware DMA].
443 #ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
444 #define EFX_PAGE_IP_ALIGN 0
446 #define EFX_PAGE_IP_ALIGN NET_IP_ALIGN
450 * Alignment of the skb->head which wraps a page-allocated RX buffer
452 * The skb allocated to wrap an rx_buffer can have this alignment. Since
453 * the data is memcpy'd from the rx_buf, it does not need to be equal to
456 #define EFX_PAGE_SKB_ALIGN 2
458 /* Forward declaration */
461 /* Pseudo bit-mask flow control field */
462 #define EFX_FC_RX FLOW_CTRL_RX
463 #define EFX_FC_TX FLOW_CTRL_TX
464 #define EFX_FC_AUTO 4
467 * struct efx_link_state - Current state of the link
469 * @fd: Link is full-duplex
470 * @fc: Actual flow control flags
471 * @speed: Link speed (Mbps)
473 struct efx_link_state
{
480 static inline bool efx_link_state_equal(const struct efx_link_state
*left
,
481 const struct efx_link_state
*right
)
483 return left
->up
== right
->up
&& left
->fd
== right
->fd
&&
484 left
->fc
== right
->fc
&& left
->speed
== right
->speed
;
488 * struct efx_phy_operations - Efx PHY operations table
489 * @probe: Probe PHY and initialise efx->mdio.mode_support, efx->mdio.mmds,
490 * efx->loopback_modes.
491 * @init: Initialise PHY
492 * @fini: Shut down PHY
493 * @reconfigure: Reconfigure PHY (e.g. for new link parameters)
494 * @poll: Update @link_state and report whether it changed.
495 * Serialised by the mac_lock.
496 * @get_settings: Get ethtool settings. Serialised by the mac_lock.
497 * @set_settings: Set ethtool settings. Serialised by the mac_lock.
498 * @set_npage_adv: Set abilities advertised in (Extended) Next Page
499 * (only needed where AN bit is set in mmds)
500 * @test_alive: Test that PHY is 'alive' (online)
501 * @test_name: Get the name of a PHY-specific test/result
502 * @run_tests: Run tests and record results as appropriate (offline).
503 * Flags are the ethtool tests flags.
505 struct efx_phy_operations
{
506 int (*probe
) (struct efx_nic
*efx
);
507 int (*init
) (struct efx_nic
*efx
);
508 void (*fini
) (struct efx_nic
*efx
);
509 void (*remove
) (struct efx_nic
*efx
);
510 int (*reconfigure
) (struct efx_nic
*efx
);
511 bool (*poll
) (struct efx_nic
*efx
);
512 void (*get_settings
) (struct efx_nic
*efx
,
513 struct ethtool_cmd
*ecmd
);
514 int (*set_settings
) (struct efx_nic
*efx
,
515 struct ethtool_cmd
*ecmd
);
516 void (*set_npage_adv
) (struct efx_nic
*efx
, u32
);
517 int (*test_alive
) (struct efx_nic
*efx
);
518 const char *(*test_name
) (struct efx_nic
*efx
, unsigned int index
);
519 int (*run_tests
) (struct efx_nic
*efx
, int *results
, unsigned flags
);
520 int (*get_module_eeprom
) (struct efx_nic
*efx
,
521 struct ethtool_eeprom
*ee
,
523 int (*get_module_info
) (struct efx_nic
*efx
,
524 struct ethtool_modinfo
*modinfo
);
528 * enum efx_phy_mode - PHY operating mode flags
529 * @PHY_MODE_NORMAL: on and should pass traffic
530 * @PHY_MODE_TX_DISABLED: on with TX disabled
531 * @PHY_MODE_LOW_POWER: set to low power through MDIO
532 * @PHY_MODE_OFF: switched off through external control
533 * @PHY_MODE_SPECIAL: on but will not pass traffic
537 PHY_MODE_TX_DISABLED
= 1,
538 PHY_MODE_LOW_POWER
= 2,
540 PHY_MODE_SPECIAL
= 8,
543 static inline bool efx_phy_mode_disabled(enum efx_phy_mode mode
)
545 return !!(mode
& ~PHY_MODE_TX_DISABLED
);
549 * Efx extended statistics
551 * Not all statistics are provided by all supported MACs. The purpose
552 * is this structure is to contain the raw statistics provided by each
555 struct efx_mac_stats
{
573 u64 tx_15xx_to_jumbo
;
576 u64 tx_single_collision
;
577 u64 tx_multiple_collision
;
578 u64 tx_excessive_collision
;
580 u64 tx_late_collision
;
581 u64 tx_excessive_deferred
;
583 u64 tx_mac_src_error
;
603 u64 rx_15xx_to_jumbo
;
606 u64 rx_bad_64_to_15xx
;
607 u64 rx_bad_15xx_to_jumbo
;
611 u64 rx_false_carrier
;
615 u64 rx_internal_error
;
619 /* Number of bits used in a multicast filter hash address */
620 #define EFX_MCAST_HASH_BITS 8
622 /* Number of (single-bit) entries in a multicast filter hash */
623 #define EFX_MCAST_HASH_ENTRIES (1 << EFX_MCAST_HASH_BITS)
625 /* An Efx multicast filter hash */
626 union efx_multicast_hash
{
627 u8 byte
[EFX_MCAST_HASH_ENTRIES
/ 8];
628 efx_oword_t oword
[EFX_MCAST_HASH_ENTRIES
/ sizeof(efx_oword_t
) / 8];
631 struct efx_filter_state
;
636 * struct efx_nic - an Efx NIC
637 * @name: Device name (net device name or bus id before net device registered)
638 * @pci_dev: The PCI device
639 * @type: Controller type attributes
640 * @legacy_irq: IRQ number
641 * @legacy_irq_enabled: Are IRQs enabled on NIC (INT_EN_KER register)?
642 * @workqueue: Workqueue for port reconfigures and the HW monitor.
643 * Work items do not hold and must not acquire RTNL.
644 * @workqueue_name: Name of workqueue
645 * @reset_work: Scheduled reset workitem
646 * @membase_phys: Memory BAR value as physical address
647 * @membase: Memory BAR value
648 * @interrupt_mode: Interrupt mode
649 * @timer_quantum_ns: Interrupt timer quantum, in nanoseconds
650 * @irq_rx_adaptive: Adaptive IRQ moderation enabled for RX event queues
651 * @irq_rx_moderation: IRQ moderation time for RX event queues
652 * @msg_enable: Log message enable flags
653 * @state: Device state number (%STATE_*). Serialised by the rtnl_lock.
654 * @reset_pending: Bitmask for pending resets
655 * @tx_queue: TX DMA queues
656 * @rx_queue: RX DMA queues
658 * @channel_name: Names for channels and their IRQs
659 * @extra_channel_types: Types of extra (non-traffic) channels that
660 * should be allocated for this NIC
661 * @rxq_entries: Size of receive queues requested by user.
662 * @txq_entries: Size of transmit queues requested by user.
663 * @txq_stop_thresh: TX queue fill level at or above which we stop it.
664 * @txq_wake_thresh: TX queue fill level at or below which we wake it.
665 * @tx_dc_base: Base qword address in SRAM of TX queue descriptor caches
666 * @rx_dc_base: Base qword address in SRAM of RX queue descriptor caches
667 * @sram_lim_qw: Qword address limit of SRAM
668 * @next_buffer_table: First available buffer table id
669 * @n_channels: Number of channels in use
670 * @n_rx_channels: Number of channels used for RX (= number of RX queues)
671 * @n_tx_channels: Number of channels used for TX
672 * @rx_dma_len: Current maximum RX DMA length
673 * @rx_buffer_order: Order (log2) of number of pages for each RX buffer
674 * @rx_hash_key: Toeplitz hash key for RSS
675 * @rx_indir_table: Indirection table for RSS
676 * @int_error_count: Number of internal errors seen recently
677 * @int_error_expire: Time at which error count will be expired
678 * @irq_status: Interrupt status buffer
679 * @irq_zero_count: Number of legacy IRQs seen with queue flags == 0
680 * @irq_level: IRQ level/index for IRQs not triggered by an event queue
681 * @selftest_work: Work item for asynchronous self-test
682 * @mtd_list: List of MTDs attached to the NIC
683 * @nic_data: Hardware dependent state
684 * @mac_lock: MAC access lock. Protects @port_enabled, @phy_mode,
685 * efx_monitor() and efx_reconfigure_port()
686 * @port_enabled: Port enabled indicator.
687 * Serialises efx_stop_all(), efx_start_all(), efx_monitor() and
688 * efx_mac_work() with kernel interfaces. Safe to read under any
689 * one of the rtnl_lock, mac_lock, or netif_tx_lock, but all three must
690 * be held to modify it.
691 * @port_initialized: Port initialized?
692 * @net_dev: Operating system network device. Consider holding the rtnl lock
693 * @stats_buffer: DMA buffer for statistics
694 * @phy_type: PHY type
695 * @phy_op: PHY interface
696 * @phy_data: PHY private data (including PHY-specific stats)
697 * @mdio: PHY MDIO interface
698 * @mdio_bus: PHY MDIO bus ID (only used by Siena)
699 * @phy_mode: PHY operating mode. Serialised by @mac_lock.
700 * @link_advertising: Autonegotiation advertising flags
701 * @link_state: Current state of the link
702 * @n_link_state_changes: Number of times the link has changed state
703 * @promiscuous: Promiscuous flag. Protected by netif_tx_lock.
704 * @multicast_hash: Multicast hash table
705 * @wanted_fc: Wanted flow control flags
706 * @fc_disable: When non-zero flow control is disabled. Typically used to
707 * ensure that network back pressure doesn't delay dma queue flushes.
708 * Serialised by the rtnl lock.
709 * @mac_work: Work item for changing MAC promiscuity and multicast hash
710 * @loopback_mode: Loopback status
711 * @loopback_modes: Supported loopback mode bitmask
712 * @loopback_selftest: Offline self-test private state
713 * @drain_pending: Count of RX and TX queues that haven't been flushed and drained.
714 * @rxq_flush_pending: Count of number of receive queues that need to be flushed.
715 * Decremented when the efx_flush_rx_queue() is called.
716 * @rxq_flush_outstanding: Count of number of RX flushes started but not yet
717 * completed (either success or failure). Not used when MCDI is used to
718 * flush receive queues.
719 * @flush_wq: wait queue used by efx_nic_flush_queues() to wait for flush completions.
720 * @vf: Array of &struct efx_vf objects.
721 * @vf_count: Number of VFs intended to be enabled.
722 * @vf_init_count: Number of VFs that have been fully initialised.
723 * @vi_scale: log2 number of vnics per VF.
724 * @vf_buftbl_base: The zeroth buffer table index used to back VF queues.
725 * @vfdi_status: Common VFDI status page to be dmad to VF address space.
726 * @local_addr_list: List of local addresses. Protected by %local_lock.
727 * @local_page_list: List of DMA addressable pages used to broadcast
728 * %local_addr_list. Protected by %local_lock.
729 * @local_lock: Mutex protecting %local_addr_list and %local_page_list.
730 * @peer_work: Work item to broadcast peer addresses to VMs.
731 * @ptp_data: PTP state data
732 * @monitor_work: Hardware monitor workitem
733 * @biu_lock: BIU (bus interface unit) lock
734 * @last_irq_cpu: Last CPU to handle a possible test interrupt. This
735 * field is used by efx_test_interrupts() to verify that an
736 * interrupt has occurred.
737 * @n_rx_nodesc_drop_cnt: RX no descriptor drop count
738 * @mac_stats: MAC statistics. These include all statistics the MACs
739 * can provide. Generic code converts these into a standard
740 * &struct net_device_stats.
741 * @stats_lock: Statistics update lock. Serialises statistics fetches
742 * and access to @mac_stats.
744 * This is stored in the private area of the &struct net_device.
747 /* The following fields should be written very rarely */
750 struct pci_dev
*pci_dev
;
751 const struct efx_nic_type
*type
;
753 bool legacy_irq_enabled
;
754 struct workqueue_struct
*workqueue
;
755 char workqueue_name
[16];
756 struct work_struct reset_work
;
757 resource_size_t membase_phys
;
758 void __iomem
*membase
;
760 enum efx_int_mode interrupt_mode
;
761 unsigned int timer_quantum_ns
;
762 bool irq_rx_adaptive
;
763 unsigned int irq_rx_moderation
;
766 enum nic_state state
;
767 unsigned long reset_pending
;
769 struct efx_channel
*channel
[EFX_MAX_CHANNELS
];
770 char channel_name
[EFX_MAX_CHANNELS
][IFNAMSIZ
+ 6];
771 const struct efx_channel_type
*
772 extra_channel_type
[EFX_MAX_EXTRA_CHANNELS
];
774 unsigned rxq_entries
;
775 unsigned txq_entries
;
776 unsigned int txq_stop_thresh
;
777 unsigned int txq_wake_thresh
;
781 unsigned sram_lim_qw
;
782 unsigned next_buffer_table
;
784 unsigned n_rx_channels
;
786 unsigned tx_channel_offset
;
787 unsigned n_tx_channels
;
788 unsigned int rx_dma_len
;
789 unsigned int rx_buffer_order
;
791 u32 rx_indir_table
[128];
793 unsigned int_error_count
;
794 unsigned long int_error_expire
;
796 struct efx_buffer irq_status
;
797 unsigned irq_zero_count
;
799 struct delayed_work selftest_work
;
801 #ifdef CONFIG_SFC_MTD
802 struct list_head mtd_list
;
807 struct mutex mac_lock
;
808 struct work_struct mac_work
;
811 bool port_initialized
;
812 struct net_device
*net_dev
;
814 struct efx_buffer stats_buffer
;
816 unsigned int phy_type
;
817 const struct efx_phy_operations
*phy_op
;
819 struct mdio_if_info mdio
;
820 unsigned int mdio_bus
;
821 enum efx_phy_mode phy_mode
;
823 u32 link_advertising
;
824 struct efx_link_state link_state
;
825 unsigned int n_link_state_changes
;
828 union efx_multicast_hash multicast_hash
;
833 enum efx_loopback_mode loopback_mode
;
836 void *loopback_selftest
;
838 struct efx_filter_state
*filter_state
;
840 atomic_t drain_pending
;
841 atomic_t rxq_flush_pending
;
842 atomic_t rxq_flush_outstanding
;
843 wait_queue_head_t flush_wq
;
845 #ifdef CONFIG_SFC_SRIOV
846 struct efx_channel
*vfdi_channel
;
849 unsigned vf_init_count
;
851 unsigned vf_buftbl_base
;
852 struct efx_buffer vfdi_status
;
853 struct list_head local_addr_list
;
854 struct list_head local_page_list
;
855 struct mutex local_lock
;
856 struct work_struct peer_work
;
859 struct efx_ptp_data
*ptp_data
;
861 /* The following fields may be written more often */
863 struct delayed_work monitor_work ____cacheline_aligned_in_smp
;
866 unsigned n_rx_nodesc_drop_cnt
;
867 struct efx_mac_stats mac_stats
;
868 spinlock_t stats_lock
;
871 static inline int efx_dev_registered(struct efx_nic
*efx
)
873 return efx
->net_dev
->reg_state
== NETREG_REGISTERED
;
876 static inline unsigned int efx_port_num(struct efx_nic
*efx
)
878 return efx
->net_dev
->dev_id
;
882 * struct efx_nic_type - Efx device type definition
883 * @probe: Probe the controller
884 * @remove: Free resources allocated by probe()
885 * @init: Initialise the controller
886 * @dimension_resources: Dimension controller resources (buffer table,
887 * and VIs once the available interrupt resources are clear)
888 * @fini: Shut down the controller
889 * @monitor: Periodic function for polling link state and hardware monitor
890 * @map_reset_reason: Map ethtool reset reason to a reset method
891 * @map_reset_flags: Map ethtool reset flags to a reset method, if possible
892 * @reset: Reset the controller hardware and possibly the PHY. This will
893 * be called while the controller is uninitialised.
894 * @probe_port: Probe the MAC and PHY
895 * @remove_port: Free resources allocated by probe_port()
896 * @handle_global_event: Handle a "global" event (may be %NULL)
897 * @prepare_flush: Prepare the hardware for flushing the DMA queues
898 * @finish_flush: Clean up after flushing the DMA queues
899 * @update_stats: Update statistics not provided by event handling
900 * @start_stats: Start the regular fetching of statistics
901 * @stop_stats: Stop the regular fetching of statistics
902 * @set_id_led: Set state of identifying LED or revert to automatic function
903 * @push_irq_moderation: Apply interrupt moderation value
904 * @reconfigure_port: Push loopback/power/txdis changes to the MAC and PHY
905 * @reconfigure_mac: Push MAC address, MTU, flow control and filter settings
906 * to the hardware. Serialised by the mac_lock.
907 * @check_mac_fault: Check MAC fault state. True if fault present.
908 * @get_wol: Get WoL configuration from driver state
909 * @set_wol: Push WoL configuration to the NIC
910 * @resume_wol: Synchronise WoL state between driver and MC (e.g. after resume)
911 * @test_chip: Test registers. Should use efx_nic_test_registers(), and is
912 * expected to reset the NIC.
913 * @test_nvram: Test validity of NVRAM contents
914 * @revision: Hardware architecture revision
915 * @mem_map_size: Memory BAR mapped size
916 * @txd_ptr_tbl_base: TX descriptor ring base address
917 * @rxd_ptr_tbl_base: RX descriptor ring base address
918 * @buf_tbl_base: Buffer table base address
919 * @evq_ptr_tbl_base: Event queue pointer table base address
920 * @evq_rptr_tbl_base: Event queue read-pointer table base address
921 * @max_dma_mask: Maximum possible DMA mask
922 * @rx_buffer_hash_size: Size of hash at start of RX buffer
923 * @rx_buffer_padding: Size of padding at end of RX buffer
924 * @max_interrupt_mode: Highest capability interrupt mode supported
925 * from &enum efx_init_mode.
926 * @phys_addr_channels: Number of channels with physically addressed
928 * @timer_period_max: Maximum period of interrupt timer (in ticks)
929 * @offload_features: net_device feature flags for protocol offload
930 * features implemented in hardware
932 struct efx_nic_type
{
933 int (*probe
)(struct efx_nic
*efx
);
934 void (*remove
)(struct efx_nic
*efx
);
935 int (*init
)(struct efx_nic
*efx
);
936 void (*dimension_resources
)(struct efx_nic
*efx
);
937 void (*fini
)(struct efx_nic
*efx
);
938 void (*monitor
)(struct efx_nic
*efx
);
939 enum reset_type (*map_reset_reason
)(enum reset_type reason
);
940 int (*map_reset_flags
)(u32
*flags
);
941 int (*reset
)(struct efx_nic
*efx
, enum reset_type method
);
942 int (*probe_port
)(struct efx_nic
*efx
);
943 void (*remove_port
)(struct efx_nic
*efx
);
944 bool (*handle_global_event
)(struct efx_channel
*channel
, efx_qword_t
*);
945 void (*prepare_flush
)(struct efx_nic
*efx
);
946 void (*finish_flush
)(struct efx_nic
*efx
);
947 void (*update_stats
)(struct efx_nic
*efx
);
948 void (*start_stats
)(struct efx_nic
*efx
);
949 void (*stop_stats
)(struct efx_nic
*efx
);
950 void (*set_id_led
)(struct efx_nic
*efx
, enum efx_led_mode mode
);
951 void (*push_irq_moderation
)(struct efx_channel
*channel
);
952 int (*reconfigure_port
)(struct efx_nic
*efx
);
953 int (*reconfigure_mac
)(struct efx_nic
*efx
);
954 bool (*check_mac_fault
)(struct efx_nic
*efx
);
955 void (*get_wol
)(struct efx_nic
*efx
, struct ethtool_wolinfo
*wol
);
956 int (*set_wol
)(struct efx_nic
*efx
, u32 type
);
957 void (*resume_wol
)(struct efx_nic
*efx
);
958 int (*test_chip
)(struct efx_nic
*efx
, struct efx_self_tests
*tests
);
959 int (*test_nvram
)(struct efx_nic
*efx
);
962 unsigned int mem_map_size
;
963 unsigned int txd_ptr_tbl_base
;
964 unsigned int rxd_ptr_tbl_base
;
965 unsigned int buf_tbl_base
;
966 unsigned int evq_ptr_tbl_base
;
967 unsigned int evq_rptr_tbl_base
;
969 unsigned int rx_buffer_hash_size
;
970 unsigned int rx_buffer_padding
;
971 unsigned int max_interrupt_mode
;
972 unsigned int phys_addr_channels
;
973 unsigned int timer_period_max
;
974 netdev_features_t offload_features
;
977 /**************************************************************************
979 * Prototypes and inline functions
981 *************************************************************************/
983 static inline struct efx_channel
*
984 efx_get_channel(struct efx_nic
*efx
, unsigned index
)
986 EFX_BUG_ON_PARANOID(index
>= efx
->n_channels
);
987 return efx
->channel
[index
];
990 /* Iterate over all used channels */
991 #define efx_for_each_channel(_channel, _efx) \
992 for (_channel = (_efx)->channel[0]; \
994 _channel = (_channel->channel + 1 < (_efx)->n_channels) ? \
995 (_efx)->channel[_channel->channel + 1] : NULL)
997 /* Iterate over all used channels in reverse */
998 #define efx_for_each_channel_rev(_channel, _efx) \
999 for (_channel = (_efx)->channel[(_efx)->n_channels - 1]; \
1001 _channel = _channel->channel ? \
1002 (_efx)->channel[_channel->channel - 1] : NULL)
1004 static inline struct efx_tx_queue
*
1005 efx_get_tx_queue(struct efx_nic
*efx
, unsigned index
, unsigned type
)
1007 EFX_BUG_ON_PARANOID(index
>= efx
->n_tx_channels
||
1008 type
>= EFX_TXQ_TYPES
);
1009 return &efx
->channel
[efx
->tx_channel_offset
+ index
]->tx_queue
[type
];
1012 static inline bool efx_channel_has_tx_queues(struct efx_channel
*channel
)
1014 return channel
->channel
- channel
->efx
->tx_channel_offset
<
1015 channel
->efx
->n_tx_channels
;
1018 static inline struct efx_tx_queue
*
1019 efx_channel_get_tx_queue(struct efx_channel
*channel
, unsigned type
)
1021 EFX_BUG_ON_PARANOID(!efx_channel_has_tx_queues(channel
) ||
1022 type
>= EFX_TXQ_TYPES
);
1023 return &channel
->tx_queue
[type
];
1026 static inline bool efx_tx_queue_used(struct efx_tx_queue
*tx_queue
)
1028 return !(tx_queue
->efx
->net_dev
->num_tc
< 2 &&
1029 tx_queue
->queue
& EFX_TXQ_TYPE_HIGHPRI
);
1032 /* Iterate over all TX queues belonging to a channel */
1033 #define efx_for_each_channel_tx_queue(_tx_queue, _channel) \
1034 if (!efx_channel_has_tx_queues(_channel)) \
1037 for (_tx_queue = (_channel)->tx_queue; \
1038 _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES && \
1039 efx_tx_queue_used(_tx_queue); \
1042 /* Iterate over all possible TX queues belonging to a channel */
1043 #define efx_for_each_possible_channel_tx_queue(_tx_queue, _channel) \
1044 if (!efx_channel_has_tx_queues(_channel)) \
1047 for (_tx_queue = (_channel)->tx_queue; \
1048 _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES; \
1051 static inline bool efx_channel_has_rx_queue(struct efx_channel
*channel
)
1053 return channel
->rx_queue
.core_index
>= 0;
1056 static inline struct efx_rx_queue
*
1057 efx_channel_get_rx_queue(struct efx_channel
*channel
)
1059 EFX_BUG_ON_PARANOID(!efx_channel_has_rx_queue(channel
));
1060 return &channel
->rx_queue
;
1063 /* Iterate over all RX queues belonging to a channel */
1064 #define efx_for_each_channel_rx_queue(_rx_queue, _channel) \
1065 if (!efx_channel_has_rx_queue(_channel)) \
1068 for (_rx_queue = &(_channel)->rx_queue; \
1072 static inline struct efx_channel
*
1073 efx_rx_queue_channel(struct efx_rx_queue
*rx_queue
)
1075 return container_of(rx_queue
, struct efx_channel
, rx_queue
);
1078 static inline int efx_rx_queue_index(struct efx_rx_queue
*rx_queue
)
1080 return efx_rx_queue_channel(rx_queue
)->channel
;
1083 /* Returns a pointer to the specified receive buffer in the RX
1086 static inline struct efx_rx_buffer
*efx_rx_buffer(struct efx_rx_queue
*rx_queue
,
1089 return &rx_queue
->buffer
[index
];
1094 * EFX_MAX_FRAME_LEN - calculate maximum frame length
1096 * This calculates the maximum frame length that will be used for a
1097 * given MTU. The frame length will be equal to the MTU plus a
1098 * constant amount of header space and padding. This is the quantity
1099 * that the net driver will program into the MAC as the maximum frame
1102 * The 10G MAC requires 8-byte alignment on the frame
1103 * length, so we round up to the nearest 8.
1105 * Re-clocking by the XGXS on RX can reduce an IPG to 32 bits (half an
1106 * XGMII cycle). If the frame length reaches the maximum value in the
1107 * same cycle, the XMAC can miss the IPG altogether. We work around
1108 * this by adding a further 16 bytes.
1110 #define EFX_MAX_FRAME_LEN(mtu) \
1111 ((((mtu) + ETH_HLEN + VLAN_HLEN + 4/* FCS */ + 7) & ~7) + 16)
1113 static inline bool efx_xmit_with_hwtstamp(struct sk_buff
*skb
)
1115 return skb_shinfo(skb
)->tx_flags
& SKBTX_HW_TSTAMP
;
1117 static inline void efx_xmit_hwtstamp_pending(struct sk_buff
*skb
)
1119 skb_shinfo(skb
)->tx_flags
|= SKBTX_IN_PROGRESS
;
1122 #endif /* EFX_NET_DRIVER_H */