2 * Copyright (C) 2005 - 2014 Emulex
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License version 2
7 * as published by the Free Software Foundation. The full GNU General
8 * Public License is included in this distribution in the file called COPYING.
10 * Contact Information:
11 * linux-drivers@emulex.com
15 * Costa Mesa, CA 92626
18 #include <linux/prefetch.h>
19 #include <linux/module.h>
22 #include <asm/div64.h>
23 #include <linux/aer.h>
24 #include <linux/if_bridge.h>
25 #include <net/busy_poll.h>
26 #include <net/vxlan.h>
28 MODULE_VERSION(DRV_VER);
29 MODULE_DEVICE_TABLE(pci, be_dev_ids);
30 MODULE_DESCRIPTION(DRV_DESC " " DRV_VER);
31 MODULE_AUTHOR("Emulex Corporation");
32 MODULE_LICENSE("GPL");
34 static unsigned int num_vfs;
35 module_param(num_vfs, uint, S_IRUGO);
36 MODULE_PARM_DESC(num_vfs, "Number of PCI VFs to initialize");
38 static ushort rx_frag_size = 2048;
39 module_param(rx_frag_size, ushort, S_IRUGO);
40 MODULE_PARM_DESC(rx_frag_size, "Size of a fragment that holds rcvd data.");
42 static const struct pci_device_id be_dev_ids[] = {
43 { PCI_DEVICE(BE_VENDOR_ID, BE_DEVICE_ID1) },
44 { PCI_DEVICE(BE_VENDOR_ID, BE_DEVICE_ID2) },
45 { PCI_DEVICE(BE_VENDOR_ID, OC_DEVICE_ID1) },
46 { PCI_DEVICE(BE_VENDOR_ID, OC_DEVICE_ID2) },
47 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID3)},
48 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID4)},
49 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID5)},
50 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID6)},
53 MODULE_DEVICE_TABLE(pci, be_dev_ids);
54 /* UE Status Low CSR */
55 static const char * const ue_status_low_desc[] = {
89 /* UE Status High CSR */
90 static const char * const ue_status_hi_desc[] = {
126 static void be_queue_free(struct be_adapter *adapter, struct be_queue_info *q)
128 struct be_dma_mem *mem = &q->dma_mem;
130 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
136 static int be_queue_alloc(struct be_adapter *adapter, struct be_queue_info *q,
137 u16 len, u16 entry_size)
139 struct be_dma_mem *mem = &q->dma_mem;
141 memset(q, 0, sizeof(*q));
143 q->entry_size = entry_size;
144 mem->size = len * entry_size;
145 mem->va = dma_zalloc_coherent(&adapter->pdev->dev, mem->size, &mem->dma,
152 static void be_reg_intr_set(struct be_adapter *adapter, bool enable)
156 pci_read_config_dword(adapter->pdev, PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET,
158 enabled = reg & MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
160 if (!enabled && enable)
161 reg |= MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
162 else if (enabled && !enable)
163 reg &= ~MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
167 pci_write_config_dword(adapter->pdev,
168 PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET, reg);
171 static void be_intr_set(struct be_adapter *adapter, bool enable)
175 /* On lancer interrupts can't be controlled via this register */
176 if (lancer_chip(adapter))
179 if (adapter->eeh_error)
182 status = be_cmd_intr_set(adapter, enable);
184 be_reg_intr_set(adapter, enable);
187 static void be_rxq_notify(struct be_adapter *adapter, u16 qid, u16 posted)
190 val |= qid & DB_RQ_RING_ID_MASK;
191 val |= posted << DB_RQ_NUM_POSTED_SHIFT;
194 iowrite32(val, adapter->db + DB_RQ_OFFSET);
197 static void be_txq_notify(struct be_adapter *adapter, struct be_tx_obj *txo,
201 val |= txo->q.id & DB_TXULP_RING_ID_MASK;
202 val |= (posted & DB_TXULP_NUM_POSTED_MASK) << DB_TXULP_NUM_POSTED_SHIFT;
205 iowrite32(val, adapter->db + txo->db_offset);
208 static void be_eq_notify(struct be_adapter *adapter, u16 qid,
209 bool arm, bool clear_int, u16 num_popped)
212 val |= qid & DB_EQ_RING_ID_MASK;
213 val |= ((qid & DB_EQ_RING_ID_EXT_MASK) << DB_EQ_RING_ID_EXT_MASK_SHIFT);
215 if (adapter->eeh_error)
219 val |= 1 << DB_EQ_REARM_SHIFT;
221 val |= 1 << DB_EQ_CLR_SHIFT;
222 val |= 1 << DB_EQ_EVNT_SHIFT;
223 val |= num_popped << DB_EQ_NUM_POPPED_SHIFT;
224 iowrite32(val, adapter->db + DB_EQ_OFFSET);
227 void be_cq_notify(struct be_adapter *adapter, u16 qid, bool arm, u16 num_popped)
230 val |= qid & DB_CQ_RING_ID_MASK;
231 val |= ((qid & DB_CQ_RING_ID_EXT_MASK) <<
232 DB_CQ_RING_ID_EXT_MASK_SHIFT);
234 if (adapter->eeh_error)
238 val |= 1 << DB_CQ_REARM_SHIFT;
239 val |= num_popped << DB_CQ_NUM_POPPED_SHIFT;
240 iowrite32(val, adapter->db + DB_CQ_OFFSET);
243 static int be_mac_addr_set(struct net_device *netdev, void *p)
245 struct be_adapter *adapter = netdev_priv(netdev);
246 struct device *dev = &adapter->pdev->dev;
247 struct sockaddr *addr = p;
250 u32 old_pmac_id = adapter->pmac_id[0], curr_pmac_id = 0;
252 if (!is_valid_ether_addr(addr->sa_data))
253 return -EADDRNOTAVAIL;
255 /* Proceed further only if, User provided MAC is different
258 if (ether_addr_equal(addr->sa_data, netdev->dev_addr))
261 /* The PMAC_ADD cmd may fail if the VF doesn't have FILTMGMT
262 * privilege or if PF did not provision the new MAC address.
263 * On BE3, this cmd will always fail if the VF doesn't have the
264 * FILTMGMT privilege. This failure is OK, only if the PF programmed
265 * the MAC for the VF.
267 status = be_cmd_pmac_add(adapter, (u8 *)addr->sa_data,
268 adapter->if_handle, &adapter->pmac_id[0], 0);
270 curr_pmac_id = adapter->pmac_id[0];
272 /* Delete the old programmed MAC. This call may fail if the
273 * old MAC was already deleted by the PF driver.
275 if (adapter->pmac_id[0] != old_pmac_id)
276 be_cmd_pmac_del(adapter, adapter->if_handle,
280 /* Decide if the new MAC is successfully activated only after
283 status = be_cmd_get_active_mac(adapter, curr_pmac_id, mac,
284 adapter->if_handle, true, 0);
288 /* The MAC change did not happen, either due to lack of privilege
289 * or PF didn't pre-provision.
291 if (!ether_addr_equal(addr->sa_data, mac)) {
296 memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
297 dev_info(dev, "MAC address changed to %pM\n", mac);
300 dev_warn(dev, "MAC address change to %pM failed\n", addr->sa_data);
304 /* BE2 supports only v0 cmd */
305 static void *hw_stats_from_cmd(struct be_adapter *adapter)
307 if (BE2_chip(adapter)) {
308 struct be_cmd_resp_get_stats_v0 *cmd = adapter->stats_cmd.va;
310 return &cmd->hw_stats;
311 } else if (BE3_chip(adapter)) {
312 struct be_cmd_resp_get_stats_v1 *cmd = adapter->stats_cmd.va;
314 return &cmd->hw_stats;
316 struct be_cmd_resp_get_stats_v2 *cmd = adapter->stats_cmd.va;
318 return &cmd->hw_stats;
322 /* BE2 supports only v0 cmd */
323 static void *be_erx_stats_from_cmd(struct be_adapter *adapter)
325 if (BE2_chip(adapter)) {
326 struct be_hw_stats_v0 *hw_stats = hw_stats_from_cmd(adapter);
328 return &hw_stats->erx;
329 } else if (BE3_chip(adapter)) {
330 struct be_hw_stats_v1 *hw_stats = hw_stats_from_cmd(adapter);
332 return &hw_stats->erx;
334 struct be_hw_stats_v2 *hw_stats = hw_stats_from_cmd(adapter);
336 return &hw_stats->erx;
340 static void populate_be_v0_stats(struct be_adapter *adapter)
342 struct be_hw_stats_v0 *hw_stats = hw_stats_from_cmd(adapter);
343 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
344 struct be_rxf_stats_v0 *rxf_stats = &hw_stats->rxf;
345 struct be_port_rxf_stats_v0 *port_stats =
346 &rxf_stats->port[adapter->port_num];
347 struct be_drv_stats *drvs = &adapter->drv_stats;
349 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
350 drvs->rx_pause_frames = port_stats->rx_pause_frames;
351 drvs->rx_crc_errors = port_stats->rx_crc_errors;
352 drvs->rx_control_frames = port_stats->rx_control_frames;
353 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
354 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
355 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
356 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
357 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
358 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
359 drvs->rxpp_fifo_overflow_drop = port_stats->rx_fifo_overflow;
360 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
361 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
362 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
363 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
364 drvs->rx_input_fifo_overflow_drop = port_stats->rx_input_fifo_overflow;
365 drvs->rx_dropped_header_too_small =
366 port_stats->rx_dropped_header_too_small;
367 drvs->rx_address_filtered =
368 port_stats->rx_address_filtered +
369 port_stats->rx_vlan_filtered;
370 drvs->rx_alignment_symbol_errors =
371 port_stats->rx_alignment_symbol_errors;
373 drvs->tx_pauseframes = port_stats->tx_pauseframes;
374 drvs->tx_controlframes = port_stats->tx_controlframes;
376 if (adapter->port_num)
377 drvs->jabber_events = rxf_stats->port1_jabber_events;
379 drvs->jabber_events = rxf_stats->port0_jabber_events;
380 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
381 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
382 drvs->forwarded_packets = rxf_stats->forwarded_packets;
383 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
384 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
385 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
386 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
389 static void populate_be_v1_stats(struct be_adapter *adapter)
391 struct be_hw_stats_v1 *hw_stats = hw_stats_from_cmd(adapter);
392 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
393 struct be_rxf_stats_v1 *rxf_stats = &hw_stats->rxf;
394 struct be_port_rxf_stats_v1 *port_stats =
395 &rxf_stats->port[adapter->port_num];
396 struct be_drv_stats *drvs = &adapter->drv_stats;
398 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
399 drvs->pmem_fifo_overflow_drop = port_stats->pmem_fifo_overflow_drop;
400 drvs->rx_priority_pause_frames = port_stats->rx_priority_pause_frames;
401 drvs->rx_pause_frames = port_stats->rx_pause_frames;
402 drvs->rx_crc_errors = port_stats->rx_crc_errors;
403 drvs->rx_control_frames = port_stats->rx_control_frames;
404 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
405 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
406 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
407 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
408 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
409 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
410 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
411 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
412 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
413 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
414 drvs->rx_dropped_header_too_small =
415 port_stats->rx_dropped_header_too_small;
416 drvs->rx_input_fifo_overflow_drop =
417 port_stats->rx_input_fifo_overflow_drop;
418 drvs->rx_address_filtered = port_stats->rx_address_filtered;
419 drvs->rx_alignment_symbol_errors =
420 port_stats->rx_alignment_symbol_errors;
421 drvs->rxpp_fifo_overflow_drop = port_stats->rxpp_fifo_overflow_drop;
422 drvs->tx_pauseframes = port_stats->tx_pauseframes;
423 drvs->tx_controlframes = port_stats->tx_controlframes;
424 drvs->tx_priority_pauseframes = port_stats->tx_priority_pauseframes;
425 drvs->jabber_events = port_stats->jabber_events;
426 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
427 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
428 drvs->forwarded_packets = rxf_stats->forwarded_packets;
429 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
430 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
431 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
432 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
435 static void populate_be_v2_stats(struct be_adapter *adapter)
437 struct be_hw_stats_v2 *hw_stats = hw_stats_from_cmd(adapter);
438 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
439 struct be_rxf_stats_v2 *rxf_stats = &hw_stats->rxf;
440 struct be_port_rxf_stats_v2 *port_stats =
441 &rxf_stats->port[adapter->port_num];
442 struct be_drv_stats *drvs = &adapter->drv_stats;
444 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
445 drvs->pmem_fifo_overflow_drop = port_stats->pmem_fifo_overflow_drop;
446 drvs->rx_priority_pause_frames = port_stats->rx_priority_pause_frames;
447 drvs->rx_pause_frames = port_stats->rx_pause_frames;
448 drvs->rx_crc_errors = port_stats->rx_crc_errors;
449 drvs->rx_control_frames = port_stats->rx_control_frames;
450 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
451 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
452 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
453 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
454 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
455 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
456 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
457 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
458 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
459 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
460 drvs->rx_dropped_header_too_small =
461 port_stats->rx_dropped_header_too_small;
462 drvs->rx_input_fifo_overflow_drop =
463 port_stats->rx_input_fifo_overflow_drop;
464 drvs->rx_address_filtered = port_stats->rx_address_filtered;
465 drvs->rx_alignment_symbol_errors =
466 port_stats->rx_alignment_symbol_errors;
467 drvs->rxpp_fifo_overflow_drop = port_stats->rxpp_fifo_overflow_drop;
468 drvs->tx_pauseframes = port_stats->tx_pauseframes;
469 drvs->tx_controlframes = port_stats->tx_controlframes;
470 drvs->tx_priority_pauseframes = port_stats->tx_priority_pauseframes;
471 drvs->jabber_events = port_stats->jabber_events;
472 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
473 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
474 drvs->forwarded_packets = rxf_stats->forwarded_packets;
475 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
476 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
477 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
478 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
479 if (be_roce_supported(adapter)) {
480 drvs->rx_roce_bytes_lsd = port_stats->roce_bytes_received_lsd;
481 drvs->rx_roce_bytes_msd = port_stats->roce_bytes_received_msd;
482 drvs->rx_roce_frames = port_stats->roce_frames_received;
483 drvs->roce_drops_crc = port_stats->roce_drops_crc;
484 drvs->roce_drops_payload_len =
485 port_stats->roce_drops_payload_len;
489 static void populate_lancer_stats(struct be_adapter *adapter)
492 struct be_drv_stats *drvs = &adapter->drv_stats;
493 struct lancer_pport_stats *pport_stats = pport_stats_from_cmd(adapter);
495 be_dws_le_to_cpu(pport_stats, sizeof(*pport_stats));
496 drvs->rx_pause_frames = pport_stats->rx_pause_frames_lo;
497 drvs->rx_crc_errors = pport_stats->rx_crc_errors_lo;
498 drvs->rx_control_frames = pport_stats->rx_control_frames_lo;
499 drvs->rx_in_range_errors = pport_stats->rx_in_range_errors;
500 drvs->rx_frame_too_long = pport_stats->rx_frames_too_long_lo;
501 drvs->rx_dropped_runt = pport_stats->rx_dropped_runt;
502 drvs->rx_ip_checksum_errs = pport_stats->rx_ip_checksum_errors;
503 drvs->rx_tcp_checksum_errs = pport_stats->rx_tcp_checksum_errors;
504 drvs->rx_udp_checksum_errs = pport_stats->rx_udp_checksum_errors;
505 drvs->rx_dropped_tcp_length =
506 pport_stats->rx_dropped_invalid_tcp_length;
507 drvs->rx_dropped_too_small = pport_stats->rx_dropped_too_small;
508 drvs->rx_dropped_too_short = pport_stats->rx_dropped_too_short;
509 drvs->rx_out_range_errors = pport_stats->rx_out_of_range_errors;
510 drvs->rx_dropped_header_too_small =
511 pport_stats->rx_dropped_header_too_small;
512 drvs->rx_input_fifo_overflow_drop = pport_stats->rx_fifo_overflow;
513 drvs->rx_address_filtered =
514 pport_stats->rx_address_filtered +
515 pport_stats->rx_vlan_filtered;
516 drvs->rx_alignment_symbol_errors = pport_stats->rx_symbol_errors_lo;
517 drvs->rxpp_fifo_overflow_drop = pport_stats->rx_fifo_overflow;
518 drvs->tx_pauseframes = pport_stats->tx_pause_frames_lo;
519 drvs->tx_controlframes = pport_stats->tx_control_frames_lo;
520 drvs->jabber_events = pport_stats->rx_jabbers;
521 drvs->forwarded_packets = pport_stats->num_forwards_lo;
522 drvs->rx_drops_mtu = pport_stats->rx_drops_mtu_lo;
523 drvs->rx_drops_too_many_frags =
524 pport_stats->rx_drops_too_many_frags_lo;
527 static void accumulate_16bit_val(u32 *acc, u16 val)
529 #define lo(x) (x & 0xFFFF)
530 #define hi(x) (x & 0xFFFF0000)
531 bool wrapped = val < lo(*acc);
532 u32 newacc = hi(*acc) + val;
536 ACCESS_ONCE(*acc) = newacc;
539 static void populate_erx_stats(struct be_adapter *adapter,
540 struct be_rx_obj *rxo, u32 erx_stat)
542 if (!BEx_chip(adapter))
543 rx_stats(rxo)->rx_drops_no_frags = erx_stat;
545 /* below erx HW counter can actually wrap around after
546 * 65535. Driver accumulates a 32-bit value
548 accumulate_16bit_val(&rx_stats(rxo)->rx_drops_no_frags,
552 void be_parse_stats(struct be_adapter *adapter)
554 struct be_erx_stats_v2 *erx = be_erx_stats_from_cmd(adapter);
555 struct be_rx_obj *rxo;
559 if (lancer_chip(adapter)) {
560 populate_lancer_stats(adapter);
562 if (BE2_chip(adapter))
563 populate_be_v0_stats(adapter);
564 else if (BE3_chip(adapter))
566 populate_be_v1_stats(adapter);
568 populate_be_v2_stats(adapter);
570 /* erx_v2 is longer than v0, v1. use v2 for v0, v1 access */
571 for_all_rx_queues(adapter, rxo, i) {
572 erx_stat = erx->rx_drops_no_fragments[rxo->q.id];
573 populate_erx_stats(adapter, rxo, erx_stat);
578 static struct rtnl_link_stats64 *be_get_stats64(struct net_device *netdev,
579 struct rtnl_link_stats64 *stats)
581 struct be_adapter *adapter = netdev_priv(netdev);
582 struct be_drv_stats *drvs = &adapter->drv_stats;
583 struct be_rx_obj *rxo;
584 struct be_tx_obj *txo;
589 for_all_rx_queues(adapter, rxo, i) {
590 const struct be_rx_stats *rx_stats = rx_stats(rxo);
592 start = u64_stats_fetch_begin_irq(&rx_stats->sync);
593 pkts = rx_stats(rxo)->rx_pkts;
594 bytes = rx_stats(rxo)->rx_bytes;
595 } while (u64_stats_fetch_retry_irq(&rx_stats->sync, start));
596 stats->rx_packets += pkts;
597 stats->rx_bytes += bytes;
598 stats->multicast += rx_stats(rxo)->rx_mcast_pkts;
599 stats->rx_dropped += rx_stats(rxo)->rx_drops_no_skbs +
600 rx_stats(rxo)->rx_drops_no_frags;
603 for_all_tx_queues(adapter, txo, i) {
604 const struct be_tx_stats *tx_stats = tx_stats(txo);
606 start = u64_stats_fetch_begin_irq(&tx_stats->sync);
607 pkts = tx_stats(txo)->tx_pkts;
608 bytes = tx_stats(txo)->tx_bytes;
609 } while (u64_stats_fetch_retry_irq(&tx_stats->sync, start));
610 stats->tx_packets += pkts;
611 stats->tx_bytes += bytes;
614 /* bad pkts received */
615 stats->rx_errors = drvs->rx_crc_errors +
616 drvs->rx_alignment_symbol_errors +
617 drvs->rx_in_range_errors +
618 drvs->rx_out_range_errors +
619 drvs->rx_frame_too_long +
620 drvs->rx_dropped_too_small +
621 drvs->rx_dropped_too_short +
622 drvs->rx_dropped_header_too_small +
623 drvs->rx_dropped_tcp_length +
624 drvs->rx_dropped_runt;
626 /* detailed rx errors */
627 stats->rx_length_errors = drvs->rx_in_range_errors +
628 drvs->rx_out_range_errors +
629 drvs->rx_frame_too_long;
631 stats->rx_crc_errors = drvs->rx_crc_errors;
633 /* frame alignment errors */
634 stats->rx_frame_errors = drvs->rx_alignment_symbol_errors;
636 /* receiver fifo overrun */
637 /* drops_no_pbuf is no per i/f, it's per BE card */
638 stats->rx_fifo_errors = drvs->rxpp_fifo_overflow_drop +
639 drvs->rx_input_fifo_overflow_drop +
640 drvs->rx_drops_no_pbuf;
644 void be_link_status_update(struct be_adapter *adapter, u8 link_status)
646 struct net_device *netdev = adapter->netdev;
648 if (!(adapter->flags & BE_FLAGS_LINK_STATUS_INIT)) {
649 netif_carrier_off(netdev);
650 adapter->flags |= BE_FLAGS_LINK_STATUS_INIT;
654 netif_carrier_on(netdev);
656 netif_carrier_off(netdev);
659 static void be_tx_stats_update(struct be_tx_obj *txo,
660 u32 wrb_cnt, u32 copied, u32 gso_segs,
663 struct be_tx_stats *stats = tx_stats(txo);
665 u64_stats_update_begin(&stats->sync);
667 stats->tx_wrbs += wrb_cnt;
668 stats->tx_bytes += copied;
669 stats->tx_pkts += (gso_segs ? gso_segs : 1);
672 u64_stats_update_end(&stats->sync);
675 /* Determine number of WRB entries needed to xmit data in an skb */
676 static u32 wrb_cnt_for_skb(struct be_adapter *adapter, struct sk_buff *skb,
679 int cnt = (skb->len > skb->data_len);
681 cnt += skb_shinfo(skb)->nr_frags;
683 /* to account for hdr wrb */
685 if (lancer_chip(adapter) || !(cnt & 1)) {
688 /* add a dummy to make it an even num */
692 BUG_ON(cnt > BE_MAX_TX_FRAG_COUNT);
696 static inline void wrb_fill(struct be_eth_wrb *wrb, u64 addr, int len)
698 wrb->frag_pa_hi = upper_32_bits(addr);
699 wrb->frag_pa_lo = addr & 0xFFFFFFFF;
700 wrb->frag_len = len & ETH_WRB_FRAG_LEN_MASK;
704 static inline u16 be_get_tx_vlan_tag(struct be_adapter *adapter,
710 vlan_tag = vlan_tx_tag_get(skb);
711 vlan_prio = (vlan_tag & VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
712 /* If vlan priority provided by OS is NOT in available bmap */
713 if (!(adapter->vlan_prio_bmap & (1 << vlan_prio)))
714 vlan_tag = (vlan_tag & ~VLAN_PRIO_MASK) |
715 adapter->recommended_prio;
720 /* Used only for IP tunnel packets */
721 static u16 skb_inner_ip_proto(struct sk_buff *skb)
723 return (inner_ip_hdr(skb)->version == 4) ?
724 inner_ip_hdr(skb)->protocol : inner_ipv6_hdr(skb)->nexthdr;
727 static u16 skb_ip_proto(struct sk_buff *skb)
729 return (ip_hdr(skb)->version == 4) ?
730 ip_hdr(skb)->protocol : ipv6_hdr(skb)->nexthdr;
733 static void wrb_fill_hdr(struct be_adapter *adapter, struct be_eth_hdr_wrb *hdr,
734 struct sk_buff *skb, u32 wrb_cnt, u32 len,
739 memset(hdr, 0, sizeof(*hdr));
741 SET_TX_WRB_HDR_BITS(crc, hdr, 1);
743 if (skb_is_gso(skb)) {
744 SET_TX_WRB_HDR_BITS(lso, hdr, 1);
745 SET_TX_WRB_HDR_BITS(lso_mss, hdr, skb_shinfo(skb)->gso_size);
746 if (skb_is_gso_v6(skb) && !lancer_chip(adapter))
747 SET_TX_WRB_HDR_BITS(lso6, hdr, 1);
748 } else if (skb->ip_summed == CHECKSUM_PARTIAL) {
749 if (skb->encapsulation) {
750 SET_TX_WRB_HDR_BITS(ipcs, hdr, 1);
751 proto = skb_inner_ip_proto(skb);
753 proto = skb_ip_proto(skb);
755 if (proto == IPPROTO_TCP)
756 SET_TX_WRB_HDR_BITS(tcpcs, hdr, 1);
757 else if (proto == IPPROTO_UDP)
758 SET_TX_WRB_HDR_BITS(udpcs, hdr, 1);
761 if (vlan_tx_tag_present(skb)) {
762 SET_TX_WRB_HDR_BITS(vlan, hdr, 1);
763 vlan_tag = be_get_tx_vlan_tag(adapter, skb);
764 SET_TX_WRB_HDR_BITS(vlan_tag, hdr, vlan_tag);
767 /* To skip HW VLAN tagging: evt = 1, compl = 0 */
768 SET_TX_WRB_HDR_BITS(complete, hdr, !skip_hw_vlan);
769 SET_TX_WRB_HDR_BITS(event, hdr, 1);
770 SET_TX_WRB_HDR_BITS(num_wrb, hdr, wrb_cnt);
771 SET_TX_WRB_HDR_BITS(len, hdr, len);
774 static void unmap_tx_frag(struct device *dev, struct be_eth_wrb *wrb,
779 be_dws_le_to_cpu(wrb, sizeof(*wrb));
781 dma = (u64)wrb->frag_pa_hi << 32 | (u64)wrb->frag_pa_lo;
784 dma_unmap_single(dev, dma, wrb->frag_len,
787 dma_unmap_page(dev, dma, wrb->frag_len, DMA_TO_DEVICE);
791 static int make_tx_wrbs(struct be_adapter *adapter, struct be_queue_info *txq,
792 struct sk_buff *skb, u32 wrb_cnt, bool dummy_wrb,
797 struct device *dev = &adapter->pdev->dev;
798 struct sk_buff *first_skb = skb;
799 struct be_eth_wrb *wrb;
800 struct be_eth_hdr_wrb *hdr;
801 bool map_single = false;
804 hdr = queue_head_node(txq);
806 map_head = txq->head;
808 if (skb->len > skb->data_len) {
809 int len = skb_headlen(skb);
810 busaddr = dma_map_single(dev, skb->data, len, DMA_TO_DEVICE);
811 if (dma_mapping_error(dev, busaddr))
814 wrb = queue_head_node(txq);
815 wrb_fill(wrb, busaddr, len);
816 be_dws_cpu_to_le(wrb, sizeof(*wrb));
821 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
822 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i];
823 busaddr = skb_frag_dma_map(dev, frag, 0,
824 skb_frag_size(frag), DMA_TO_DEVICE);
825 if (dma_mapping_error(dev, busaddr))
827 wrb = queue_head_node(txq);
828 wrb_fill(wrb, busaddr, skb_frag_size(frag));
829 be_dws_cpu_to_le(wrb, sizeof(*wrb));
831 copied += skb_frag_size(frag);
835 wrb = queue_head_node(txq);
837 be_dws_cpu_to_le(wrb, sizeof(*wrb));
841 wrb_fill_hdr(adapter, hdr, first_skb, wrb_cnt, copied, skip_hw_vlan);
842 be_dws_cpu_to_le(hdr, sizeof(*hdr));
846 txq->head = map_head;
848 wrb = queue_head_node(txq);
849 unmap_tx_frag(dev, wrb, map_single);
851 copied -= wrb->frag_len;
852 adapter->drv_stats.dma_map_errors++;
858 static struct sk_buff *be_insert_vlan_in_pkt(struct be_adapter *adapter,
864 skb = skb_share_check(skb, GFP_ATOMIC);
868 if (vlan_tx_tag_present(skb))
869 vlan_tag = be_get_tx_vlan_tag(adapter, skb);
871 if (qnq_async_evt_rcvd(adapter) && adapter->pvid) {
873 vlan_tag = adapter->pvid;
874 /* f/w workaround to set skip_hw_vlan = 1, informs the F/W to
875 * skip VLAN insertion
878 *skip_hw_vlan = true;
882 skb = __vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_tag);
888 /* Insert the outer VLAN, if any */
889 if (adapter->qnq_vid) {
890 vlan_tag = adapter->qnq_vid;
891 skb = __vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_tag);
895 *skip_hw_vlan = true;
901 static bool be_ipv6_exthdr_check(struct sk_buff *skb)
903 struct ethhdr *eh = (struct ethhdr *)skb->data;
904 u16 offset = ETH_HLEN;
906 if (eh->h_proto == htons(ETH_P_IPV6)) {
907 struct ipv6hdr *ip6h = (struct ipv6hdr *)(skb->data + offset);
909 offset += sizeof(struct ipv6hdr);
910 if (ip6h->nexthdr != NEXTHDR_TCP &&
911 ip6h->nexthdr != NEXTHDR_UDP) {
912 struct ipv6_opt_hdr *ehdr =
913 (struct ipv6_opt_hdr *) (skb->data + offset);
915 /* offending pkt: 2nd byte following IPv6 hdr is 0xff */
916 if (ehdr->hdrlen == 0xff)
923 static int be_vlan_tag_tx_chk(struct be_adapter *adapter, struct sk_buff *skb)
925 return vlan_tx_tag_present(skb) || adapter->pvid || adapter->qnq_vid;
928 static int be_ipv6_tx_stall_chk(struct be_adapter *adapter, struct sk_buff *skb)
930 return BE3_chip(adapter) && be_ipv6_exthdr_check(skb);
933 static struct sk_buff *be_lancer_xmit_workarounds(struct be_adapter *adapter,
937 struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
938 unsigned int eth_hdr_len;
941 /* For padded packets, BE HW modifies tot_len field in IP header
942 * incorrecly when VLAN tag is inserted by HW.
943 * For padded packets, Lancer computes incorrect checksum.
945 eth_hdr_len = ntohs(skb->protocol) == ETH_P_8021Q ?
946 VLAN_ETH_HLEN : ETH_HLEN;
947 if (skb->len <= 60 &&
948 (lancer_chip(adapter) || vlan_tx_tag_present(skb)) &&
950 ip = (struct iphdr *)ip_hdr(skb);
951 pskb_trim(skb, eth_hdr_len + ntohs(ip->tot_len));
954 /* If vlan tag is already inlined in the packet, skip HW VLAN
955 * tagging in pvid-tagging mode
957 if (be_pvid_tagging_enabled(adapter) &&
958 veh->h_vlan_proto == htons(ETH_P_8021Q))
959 *skip_hw_vlan = true;
961 /* HW has a bug wherein it will calculate CSUM for VLAN
962 * pkts even though it is disabled.
963 * Manually insert VLAN in pkt.
965 if (skb->ip_summed != CHECKSUM_PARTIAL &&
966 vlan_tx_tag_present(skb)) {
967 skb = be_insert_vlan_in_pkt(adapter, skb, skip_hw_vlan);
972 /* HW may lockup when VLAN HW tagging is requested on
973 * certain ipv6 packets. Drop such pkts if the HW workaround to
974 * skip HW tagging is not enabled by FW.
976 if (unlikely(be_ipv6_tx_stall_chk(adapter, skb) &&
977 (adapter->pvid || adapter->qnq_vid) &&
978 !qnq_async_evt_rcvd(adapter)))
981 /* Manual VLAN tag insertion to prevent:
982 * ASIC lockup when the ASIC inserts VLAN tag into
983 * certain ipv6 packets. Insert VLAN tags in driver,
984 * and set event, completion, vlan bits accordingly
987 if (be_ipv6_tx_stall_chk(adapter, skb) &&
988 be_vlan_tag_tx_chk(adapter, skb)) {
989 skb = be_insert_vlan_in_pkt(adapter, skb, skip_hw_vlan);
996 dev_kfree_skb_any(skb);
1001 static struct sk_buff *be_xmit_workarounds(struct be_adapter *adapter,
1002 struct sk_buff *skb,
1005 /* Lancer, SH-R ASICs have a bug wherein Packets that are 32 bytes or
1006 * less may cause a transmit stall on that port. So the work-around is
1007 * to pad short packets (<= 32 bytes) to a 36-byte length.
1009 if (unlikely(!BEx_chip(adapter) && skb->len <= 32)) {
1010 if (skb_padto(skb, 36))
1015 if (BEx_chip(adapter) || lancer_chip(adapter)) {
1016 skb = be_lancer_xmit_workarounds(adapter, skb, skip_hw_vlan);
1024 static netdev_tx_t be_xmit(struct sk_buff *skb, struct net_device *netdev)
1026 struct be_adapter *adapter = netdev_priv(netdev);
1027 struct be_tx_obj *txo = &adapter->tx_obj[skb_get_queue_mapping(skb)];
1028 struct be_queue_info *txq = &txo->q;
1029 bool dummy_wrb, stopped = false;
1030 u32 wrb_cnt = 0, copied = 0;
1031 bool skip_hw_vlan = false;
1032 u32 start = txq->head;
1034 skb = be_xmit_workarounds(adapter, skb, &skip_hw_vlan);
1036 tx_stats(txo)->tx_drv_drops++;
1037 return NETDEV_TX_OK;
1040 wrb_cnt = wrb_cnt_for_skb(adapter, skb, &dummy_wrb);
1042 copied = make_tx_wrbs(adapter, txq, skb, wrb_cnt, dummy_wrb,
1045 int gso_segs = skb_shinfo(skb)->gso_segs;
1047 /* record the sent skb in the sent_skb table */
1048 BUG_ON(txo->sent_skb_list[start]);
1049 txo->sent_skb_list[start] = skb;
1051 /* Ensure txq has space for the next skb; Else stop the queue
1052 * *BEFORE* ringing the tx doorbell, so that we serialze the
1053 * tx compls of the current transmit which'll wake up the queue
1055 atomic_add(wrb_cnt, &txq->used);
1056 if ((BE_MAX_TX_FRAG_COUNT + atomic_read(&txq->used)) >=
1058 netif_stop_subqueue(netdev, skb_get_queue_mapping(skb));
1062 be_txq_notify(adapter, txo, wrb_cnt);
1064 be_tx_stats_update(txo, wrb_cnt, copied, gso_segs, stopped);
1067 tx_stats(txo)->tx_drv_drops++;
1068 dev_kfree_skb_any(skb);
1070 return NETDEV_TX_OK;
1073 static int be_change_mtu(struct net_device *netdev, int new_mtu)
1075 struct be_adapter *adapter = netdev_priv(netdev);
1076 struct device *dev = &adapter->pdev->dev;
1078 if (new_mtu < BE_MIN_MTU || new_mtu > BE_MAX_MTU) {
1079 dev_info(dev, "MTU must be between %d and %d bytes\n",
1080 BE_MIN_MTU, BE_MAX_MTU);
1084 dev_info(dev, "MTU changed from %d to %d bytes\n",
1085 netdev->mtu, new_mtu);
1086 netdev->mtu = new_mtu;
1091 * A max of 64 (BE_NUM_VLANS_SUPPORTED) vlans can be configured in BE.
1092 * If the user configures more, place BE in vlan promiscuous mode.
1094 static int be_vid_config(struct be_adapter *adapter)
1096 u16 vids[BE_NUM_VLANS_SUPPORTED];
1100 /* No need to further configure vids if in promiscuous mode */
1101 if (adapter->promiscuous)
1104 if (adapter->vlans_added > be_max_vlans(adapter))
1105 goto set_vlan_promisc;
1107 /* Construct VLAN Table to give to HW */
1108 for_each_set_bit(i, adapter->vids, VLAN_N_VID)
1109 vids[num++] = cpu_to_le16(i);
1111 status = be_cmd_vlan_config(adapter, adapter->if_handle, vids, num);
1113 /* Set to VLAN promisc mode as setting VLAN filter failed */
1114 if (addl_status(status) ==
1115 MCC_ADDL_STATUS_INSUFFICIENT_RESOURCES)
1116 goto set_vlan_promisc;
1117 dev_err(&adapter->pdev->dev,
1118 "Setting HW VLAN filtering failed.\n");
1120 if (adapter->flags & BE_FLAGS_VLAN_PROMISC) {
1121 /* hw VLAN filtering re-enabled. */
1122 status = be_cmd_rx_filter(adapter,
1123 BE_FLAGS_VLAN_PROMISC, OFF);
1125 dev_info(&adapter->pdev->dev,
1126 "Disabling VLAN Promiscuous mode.\n");
1127 adapter->flags &= ~BE_FLAGS_VLAN_PROMISC;
1135 if (adapter->flags & BE_FLAGS_VLAN_PROMISC)
1138 status = be_cmd_rx_filter(adapter, BE_FLAGS_VLAN_PROMISC, ON);
1140 dev_info(&adapter->pdev->dev, "Enable VLAN Promiscuous mode\n");
1141 adapter->flags |= BE_FLAGS_VLAN_PROMISC;
1143 dev_err(&adapter->pdev->dev,
1144 "Failed to enable VLAN Promiscuous mode.\n");
1148 static int be_vlan_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
1150 struct be_adapter *adapter = netdev_priv(netdev);
1153 /* Packets with VID 0 are always received by Lancer by default */
1154 if (lancer_chip(adapter) && vid == 0)
1157 if (test_bit(vid, adapter->vids))
1160 set_bit(vid, adapter->vids);
1161 adapter->vlans_added++;
1163 status = be_vid_config(adapter);
1165 adapter->vlans_added--;
1166 clear_bit(vid, adapter->vids);
1172 static int be_vlan_rem_vid(struct net_device *netdev, __be16 proto, u16 vid)
1174 struct be_adapter *adapter = netdev_priv(netdev);
1176 /* Packets with VID 0 are always received by Lancer by default */
1177 if (lancer_chip(adapter) && vid == 0)
1180 clear_bit(vid, adapter->vids);
1181 adapter->vlans_added--;
1183 return be_vid_config(adapter);
1186 static void be_clear_promisc(struct be_adapter *adapter)
1188 adapter->promiscuous = false;
1189 adapter->flags &= ~(BE_FLAGS_VLAN_PROMISC | BE_FLAGS_MCAST_PROMISC);
1191 be_cmd_rx_filter(adapter, IFF_PROMISC, OFF);
1194 static void be_set_rx_mode(struct net_device *netdev)
1196 struct be_adapter *adapter = netdev_priv(netdev);
1199 if (netdev->flags & IFF_PROMISC) {
1200 be_cmd_rx_filter(adapter, IFF_PROMISC, ON);
1201 adapter->promiscuous = true;
1205 /* BE was previously in promiscuous mode; disable it */
1206 if (adapter->promiscuous) {
1207 be_clear_promisc(adapter);
1208 if (adapter->vlans_added)
1209 be_vid_config(adapter);
1212 /* Enable multicast promisc if num configured exceeds what we support */
1213 if (netdev->flags & IFF_ALLMULTI ||
1214 netdev_mc_count(netdev) > be_max_mc(adapter))
1215 goto set_mcast_promisc;
1217 if (netdev_uc_count(netdev) != adapter->uc_macs) {
1218 struct netdev_hw_addr *ha;
1219 int i = 1; /* First slot is claimed by the Primary MAC */
1221 for (; adapter->uc_macs > 0; adapter->uc_macs--, i++) {
1222 be_cmd_pmac_del(adapter, adapter->if_handle,
1223 adapter->pmac_id[i], 0);
1226 if (netdev_uc_count(netdev) > be_max_uc(adapter)) {
1227 be_cmd_rx_filter(adapter, IFF_PROMISC, ON);
1228 adapter->promiscuous = true;
1232 netdev_for_each_uc_addr(ha, adapter->netdev) {
1233 adapter->uc_macs++; /* First slot is for Primary MAC */
1234 be_cmd_pmac_add(adapter, (u8 *)ha->addr,
1236 &adapter->pmac_id[adapter->uc_macs], 0);
1240 status = be_cmd_rx_filter(adapter, IFF_MULTICAST, ON);
1242 if (adapter->flags & BE_FLAGS_MCAST_PROMISC)
1243 adapter->flags &= ~BE_FLAGS_MCAST_PROMISC;
1248 if (adapter->flags & BE_FLAGS_MCAST_PROMISC)
1251 /* Set to MCAST promisc mode if setting MULTICAST address fails
1252 * or if num configured exceeds what we support
1254 status = be_cmd_rx_filter(adapter, IFF_ALLMULTI, ON);
1256 adapter->flags |= BE_FLAGS_MCAST_PROMISC;
1261 static int be_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
1263 struct be_adapter *adapter = netdev_priv(netdev);
1264 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1267 if (!sriov_enabled(adapter))
1270 if (!is_valid_ether_addr(mac) || vf >= adapter->num_vfs)
1273 /* Proceed further only if user provided MAC is different
1276 if (ether_addr_equal(mac, vf_cfg->mac_addr))
1279 if (BEx_chip(adapter)) {
1280 be_cmd_pmac_del(adapter, vf_cfg->if_handle, vf_cfg->pmac_id,
1283 status = be_cmd_pmac_add(adapter, mac, vf_cfg->if_handle,
1284 &vf_cfg->pmac_id, vf + 1);
1286 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
1291 dev_err(&adapter->pdev->dev, "MAC %pM set on VF %d Failed: %#x",
1293 return be_cmd_status(status);
1296 ether_addr_copy(vf_cfg->mac_addr, mac);
1301 static int be_get_vf_config(struct net_device *netdev, int vf,
1302 struct ifla_vf_info *vi)
1304 struct be_adapter *adapter = netdev_priv(netdev);
1305 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1307 if (!sriov_enabled(adapter))
1310 if (vf >= adapter->num_vfs)
1314 vi->max_tx_rate = vf_cfg->tx_rate;
1315 vi->min_tx_rate = 0;
1316 vi->vlan = vf_cfg->vlan_tag & VLAN_VID_MASK;
1317 vi->qos = vf_cfg->vlan_tag >> VLAN_PRIO_SHIFT;
1318 memcpy(&vi->mac, vf_cfg->mac_addr, ETH_ALEN);
1319 vi->linkstate = adapter->vf_cfg[vf].plink_tracking;
1324 static int be_set_vf_vlan(struct net_device *netdev, int vf, u16 vlan, u8 qos)
1326 struct be_adapter *adapter = netdev_priv(netdev);
1327 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1330 if (!sriov_enabled(adapter))
1333 if (vf >= adapter->num_vfs || vlan > 4095 || qos > 7)
1337 vlan |= qos << VLAN_PRIO_SHIFT;
1338 if (vf_cfg->vlan_tag != vlan)
1339 status = be_cmd_set_hsw_config(adapter, vlan, vf + 1,
1340 vf_cfg->if_handle, 0);
1342 /* Reset Transparent Vlan Tagging. */
1343 status = be_cmd_set_hsw_config(adapter, BE_RESET_VLAN_TAG_ID,
1344 vf + 1, vf_cfg->if_handle, 0);
1348 dev_err(&adapter->pdev->dev,
1349 "VLAN %d config on VF %d failed : %#x\n", vlan,
1351 return be_cmd_status(status);
1354 vf_cfg->vlan_tag = vlan;
1359 static int be_set_vf_tx_rate(struct net_device *netdev, int vf,
1360 int min_tx_rate, int max_tx_rate)
1362 struct be_adapter *adapter = netdev_priv(netdev);
1363 struct device *dev = &adapter->pdev->dev;
1364 int percent_rate, status = 0;
1368 if (!sriov_enabled(adapter))
1371 if (vf >= adapter->num_vfs)
1380 status = be_cmd_link_status_query(adapter, &link_speed,
1386 dev_err(dev, "TX-rate setting not allowed when link is down\n");
1391 if (max_tx_rate < 100 || max_tx_rate > link_speed) {
1392 dev_err(dev, "TX-rate must be between 100 and %d Mbps\n",
1398 /* On Skyhawk the QOS setting must be done only as a % value */
1399 percent_rate = link_speed / 100;
1400 if (skyhawk_chip(adapter) && (max_tx_rate % percent_rate)) {
1401 dev_err(dev, "TX-rate must be a multiple of %d Mbps\n",
1408 status = be_cmd_config_qos(adapter, max_tx_rate, link_speed, vf + 1);
1412 adapter->vf_cfg[vf].tx_rate = max_tx_rate;
1416 dev_err(dev, "TX-rate setting of %dMbps on VF%d failed\n",
1418 return be_cmd_status(status);
1420 static int be_set_vf_link_state(struct net_device *netdev, int vf,
1423 struct be_adapter *adapter = netdev_priv(netdev);
1426 if (!sriov_enabled(adapter))
1429 if (vf >= adapter->num_vfs)
1432 status = be_cmd_set_logical_link_config(adapter, link_state, vf+1);
1434 dev_err(&adapter->pdev->dev,
1435 "Link state change on VF %d failed: %#x\n", vf, status);
1436 return be_cmd_status(status);
1439 adapter->vf_cfg[vf].plink_tracking = link_state;
1444 static void be_aic_update(struct be_aic_obj *aic, u64 rx_pkts, u64 tx_pkts,
1447 aic->rx_pkts_prev = rx_pkts;
1448 aic->tx_reqs_prev = tx_pkts;
1452 static void be_eqd_update(struct be_adapter *adapter)
1454 struct be_set_eqd set_eqd[MAX_EVT_QS];
1455 int eqd, i, num = 0, start;
1456 struct be_aic_obj *aic;
1457 struct be_eq_obj *eqo;
1458 struct be_rx_obj *rxo;
1459 struct be_tx_obj *txo;
1460 u64 rx_pkts, tx_pkts;
1464 for_all_evt_queues(adapter, eqo, i) {
1465 aic = &adapter->aic_obj[eqo->idx];
1473 rxo = &adapter->rx_obj[eqo->idx];
1475 start = u64_stats_fetch_begin_irq(&rxo->stats.sync);
1476 rx_pkts = rxo->stats.rx_pkts;
1477 } while (u64_stats_fetch_retry_irq(&rxo->stats.sync, start));
1479 txo = &adapter->tx_obj[eqo->idx];
1481 start = u64_stats_fetch_begin_irq(&txo->stats.sync);
1482 tx_pkts = txo->stats.tx_reqs;
1483 } while (u64_stats_fetch_retry_irq(&txo->stats.sync, start));
1486 /* Skip, if wrapped around or first calculation */
1488 if (!aic->jiffies || time_before(now, aic->jiffies) ||
1489 rx_pkts < aic->rx_pkts_prev ||
1490 tx_pkts < aic->tx_reqs_prev) {
1491 be_aic_update(aic, rx_pkts, tx_pkts, now);
1495 delta = jiffies_to_msecs(now - aic->jiffies);
1496 pps = (((u32)(rx_pkts - aic->rx_pkts_prev) * 1000) / delta) +
1497 (((u32)(tx_pkts - aic->tx_reqs_prev) * 1000) / delta);
1498 eqd = (pps / 15000) << 2;
1502 eqd = min_t(u32, eqd, aic->max_eqd);
1503 eqd = max_t(u32, eqd, aic->min_eqd);
1505 be_aic_update(aic, rx_pkts, tx_pkts, now);
1507 if (eqd != aic->prev_eqd) {
1508 set_eqd[num].delay_multiplier = (eqd * 65)/100;
1509 set_eqd[num].eq_id = eqo->q.id;
1510 aic->prev_eqd = eqd;
1516 be_cmd_modify_eqd(adapter, set_eqd, num);
1519 static void be_rx_stats_update(struct be_rx_obj *rxo,
1520 struct be_rx_compl_info *rxcp)
1522 struct be_rx_stats *stats = rx_stats(rxo);
1524 u64_stats_update_begin(&stats->sync);
1526 stats->rx_bytes += rxcp->pkt_size;
1528 if (rxcp->pkt_type == BE_MULTICAST_PACKET)
1529 stats->rx_mcast_pkts++;
1531 stats->rx_compl_err++;
1532 u64_stats_update_end(&stats->sync);
1535 static inline bool csum_passed(struct be_rx_compl_info *rxcp)
1537 /* L4 checksum is not reliable for non TCP/UDP packets.
1538 * Also ignore ipcksm for ipv6 pkts
1540 return (rxcp->tcpf || rxcp->udpf) && rxcp->l4_csum &&
1541 (rxcp->ip_csum || rxcp->ipv6) && !rxcp->err;
1544 static struct be_rx_page_info *get_rx_page_info(struct be_rx_obj *rxo)
1546 struct be_adapter *adapter = rxo->adapter;
1547 struct be_rx_page_info *rx_page_info;
1548 struct be_queue_info *rxq = &rxo->q;
1549 u16 frag_idx = rxq->tail;
1551 rx_page_info = &rxo->page_info_tbl[frag_idx];
1552 BUG_ON(!rx_page_info->page);
1554 if (rx_page_info->last_frag) {
1555 dma_unmap_page(&adapter->pdev->dev,
1556 dma_unmap_addr(rx_page_info, bus),
1557 adapter->big_page_size, DMA_FROM_DEVICE);
1558 rx_page_info->last_frag = false;
1560 dma_sync_single_for_cpu(&adapter->pdev->dev,
1561 dma_unmap_addr(rx_page_info, bus),
1562 rx_frag_size, DMA_FROM_DEVICE);
1565 queue_tail_inc(rxq);
1566 atomic_dec(&rxq->used);
1567 return rx_page_info;
1570 /* Throwaway the data in the Rx completion */
1571 static void be_rx_compl_discard(struct be_rx_obj *rxo,
1572 struct be_rx_compl_info *rxcp)
1574 struct be_rx_page_info *page_info;
1575 u16 i, num_rcvd = rxcp->num_rcvd;
1577 for (i = 0; i < num_rcvd; i++) {
1578 page_info = get_rx_page_info(rxo);
1579 put_page(page_info->page);
1580 memset(page_info, 0, sizeof(*page_info));
1585 * skb_fill_rx_data forms a complete skb for an ether frame
1586 * indicated by rxcp.
1588 static void skb_fill_rx_data(struct be_rx_obj *rxo, struct sk_buff *skb,
1589 struct be_rx_compl_info *rxcp)
1591 struct be_rx_page_info *page_info;
1593 u16 hdr_len, curr_frag_len, remaining;
1596 page_info = get_rx_page_info(rxo);
1597 start = page_address(page_info->page) + page_info->page_offset;
1600 /* Copy data in the first descriptor of this completion */
1601 curr_frag_len = min(rxcp->pkt_size, rx_frag_size);
1603 skb->len = curr_frag_len;
1604 if (curr_frag_len <= BE_HDR_LEN) { /* tiny packet */
1605 memcpy(skb->data, start, curr_frag_len);
1606 /* Complete packet has now been moved to data */
1607 put_page(page_info->page);
1609 skb->tail += curr_frag_len;
1612 memcpy(skb->data, start, hdr_len);
1613 skb_shinfo(skb)->nr_frags = 1;
1614 skb_frag_set_page(skb, 0, page_info->page);
1615 skb_shinfo(skb)->frags[0].page_offset =
1616 page_info->page_offset + hdr_len;
1617 skb_frag_size_set(&skb_shinfo(skb)->frags[0],
1618 curr_frag_len - hdr_len);
1619 skb->data_len = curr_frag_len - hdr_len;
1620 skb->truesize += rx_frag_size;
1621 skb->tail += hdr_len;
1623 page_info->page = NULL;
1625 if (rxcp->pkt_size <= rx_frag_size) {
1626 BUG_ON(rxcp->num_rcvd != 1);
1630 /* More frags present for this completion */
1631 remaining = rxcp->pkt_size - curr_frag_len;
1632 for (i = 1, j = 0; i < rxcp->num_rcvd; i++) {
1633 page_info = get_rx_page_info(rxo);
1634 curr_frag_len = min(remaining, rx_frag_size);
1636 /* Coalesce all frags from the same physical page in one slot */
1637 if (page_info->page_offset == 0) {
1640 skb_frag_set_page(skb, j, page_info->page);
1641 skb_shinfo(skb)->frags[j].page_offset =
1642 page_info->page_offset;
1643 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1644 skb_shinfo(skb)->nr_frags++;
1646 put_page(page_info->page);
1649 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1650 skb->len += curr_frag_len;
1651 skb->data_len += curr_frag_len;
1652 skb->truesize += rx_frag_size;
1653 remaining -= curr_frag_len;
1654 page_info->page = NULL;
1656 BUG_ON(j > MAX_SKB_FRAGS);
1659 /* Process the RX completion indicated by rxcp when GRO is disabled */
1660 static void be_rx_compl_process(struct be_rx_obj *rxo, struct napi_struct *napi,
1661 struct be_rx_compl_info *rxcp)
1663 struct be_adapter *adapter = rxo->adapter;
1664 struct net_device *netdev = adapter->netdev;
1665 struct sk_buff *skb;
1667 skb = netdev_alloc_skb_ip_align(netdev, BE_RX_SKB_ALLOC_SIZE);
1668 if (unlikely(!skb)) {
1669 rx_stats(rxo)->rx_drops_no_skbs++;
1670 be_rx_compl_discard(rxo, rxcp);
1674 skb_fill_rx_data(rxo, skb, rxcp);
1676 if (likely((netdev->features & NETIF_F_RXCSUM) && csum_passed(rxcp)))
1677 skb->ip_summed = CHECKSUM_UNNECESSARY;
1679 skb_checksum_none_assert(skb);
1681 skb->protocol = eth_type_trans(skb, netdev);
1682 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1683 if (netdev->features & NETIF_F_RXHASH)
1684 skb_set_hash(skb, rxcp->rss_hash, PKT_HASH_TYPE_L3);
1686 skb->csum_level = rxcp->tunneled;
1687 skb_mark_napi_id(skb, napi);
1690 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1692 netif_receive_skb(skb);
1695 /* Process the RX completion indicated by rxcp when GRO is enabled */
1696 static void be_rx_compl_process_gro(struct be_rx_obj *rxo,
1697 struct napi_struct *napi,
1698 struct be_rx_compl_info *rxcp)
1700 struct be_adapter *adapter = rxo->adapter;
1701 struct be_rx_page_info *page_info;
1702 struct sk_buff *skb = NULL;
1703 u16 remaining, curr_frag_len;
1706 skb = napi_get_frags(napi);
1708 be_rx_compl_discard(rxo, rxcp);
1712 remaining = rxcp->pkt_size;
1713 for (i = 0, j = -1; i < rxcp->num_rcvd; i++) {
1714 page_info = get_rx_page_info(rxo);
1716 curr_frag_len = min(remaining, rx_frag_size);
1718 /* Coalesce all frags from the same physical page in one slot */
1719 if (i == 0 || page_info->page_offset == 0) {
1720 /* First frag or Fresh page */
1722 skb_frag_set_page(skb, j, page_info->page);
1723 skb_shinfo(skb)->frags[j].page_offset =
1724 page_info->page_offset;
1725 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1727 put_page(page_info->page);
1729 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1730 skb->truesize += rx_frag_size;
1731 remaining -= curr_frag_len;
1732 memset(page_info, 0, sizeof(*page_info));
1734 BUG_ON(j > MAX_SKB_FRAGS);
1736 skb_shinfo(skb)->nr_frags = j + 1;
1737 skb->len = rxcp->pkt_size;
1738 skb->data_len = rxcp->pkt_size;
1739 skb->ip_summed = CHECKSUM_UNNECESSARY;
1740 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1741 if (adapter->netdev->features & NETIF_F_RXHASH)
1742 skb_set_hash(skb, rxcp->rss_hash, PKT_HASH_TYPE_L3);
1744 skb->csum_level = rxcp->tunneled;
1745 skb_mark_napi_id(skb, napi);
1748 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1750 napi_gro_frags(napi);
1753 static void be_parse_rx_compl_v1(struct be_eth_rx_compl *compl,
1754 struct be_rx_compl_info *rxcp)
1756 rxcp->pkt_size = GET_RX_COMPL_V1_BITS(pktsize, compl);
1757 rxcp->vlanf = GET_RX_COMPL_V1_BITS(vtp, compl);
1758 rxcp->err = GET_RX_COMPL_V1_BITS(err, compl);
1759 rxcp->tcpf = GET_RX_COMPL_V1_BITS(tcpf, compl);
1760 rxcp->udpf = GET_RX_COMPL_V1_BITS(udpf, compl);
1761 rxcp->ip_csum = GET_RX_COMPL_V1_BITS(ipcksm, compl);
1762 rxcp->l4_csum = GET_RX_COMPL_V1_BITS(l4_cksm, compl);
1763 rxcp->ipv6 = GET_RX_COMPL_V1_BITS(ip_version, compl);
1764 rxcp->num_rcvd = GET_RX_COMPL_V1_BITS(numfrags, compl);
1765 rxcp->pkt_type = GET_RX_COMPL_V1_BITS(cast_enc, compl);
1766 rxcp->rss_hash = GET_RX_COMPL_V1_BITS(rsshash, compl);
1768 rxcp->qnq = GET_RX_COMPL_V1_BITS(qnq, compl);
1769 rxcp->vlan_tag = GET_RX_COMPL_V1_BITS(vlan_tag, compl);
1771 rxcp->port = GET_RX_COMPL_V1_BITS(port, compl);
1773 GET_RX_COMPL_V1_BITS(tunneled, compl);
1776 static void be_parse_rx_compl_v0(struct be_eth_rx_compl *compl,
1777 struct be_rx_compl_info *rxcp)
1779 rxcp->pkt_size = GET_RX_COMPL_V0_BITS(pktsize, compl);
1780 rxcp->vlanf = GET_RX_COMPL_V0_BITS(vtp, compl);
1781 rxcp->err = GET_RX_COMPL_V0_BITS(err, compl);
1782 rxcp->tcpf = GET_RX_COMPL_V0_BITS(tcpf, compl);
1783 rxcp->udpf = GET_RX_COMPL_V0_BITS(udpf, compl);
1784 rxcp->ip_csum = GET_RX_COMPL_V0_BITS(ipcksm, compl);
1785 rxcp->l4_csum = GET_RX_COMPL_V0_BITS(l4_cksm, compl);
1786 rxcp->ipv6 = GET_RX_COMPL_V0_BITS(ip_version, compl);
1787 rxcp->num_rcvd = GET_RX_COMPL_V0_BITS(numfrags, compl);
1788 rxcp->pkt_type = GET_RX_COMPL_V0_BITS(cast_enc, compl);
1789 rxcp->rss_hash = GET_RX_COMPL_V0_BITS(rsshash, compl);
1791 rxcp->qnq = GET_RX_COMPL_V0_BITS(qnq, compl);
1792 rxcp->vlan_tag = GET_RX_COMPL_V0_BITS(vlan_tag, compl);
1794 rxcp->port = GET_RX_COMPL_V0_BITS(port, compl);
1795 rxcp->ip_frag = GET_RX_COMPL_V0_BITS(ip_frag, compl);
1798 static struct be_rx_compl_info *be_rx_compl_get(struct be_rx_obj *rxo)
1800 struct be_eth_rx_compl *compl = queue_tail_node(&rxo->cq);
1801 struct be_rx_compl_info *rxcp = &rxo->rxcp;
1802 struct be_adapter *adapter = rxo->adapter;
1804 /* For checking the valid bit it is Ok to use either definition as the
1805 * valid bit is at the same position in both v0 and v1 Rx compl */
1806 if (compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] == 0)
1810 be_dws_le_to_cpu(compl, sizeof(*compl));
1812 if (adapter->be3_native)
1813 be_parse_rx_compl_v1(compl, rxcp);
1815 be_parse_rx_compl_v0(compl, rxcp);
1821 /* In QNQ modes, if qnq bit is not set, then the packet was
1822 * tagged only with the transparent outer vlan-tag and must
1823 * not be treated as a vlan packet by host
1825 if (be_is_qnq_mode(adapter) && !rxcp->qnq)
1828 if (!lancer_chip(adapter))
1829 rxcp->vlan_tag = swab16(rxcp->vlan_tag);
1831 if (adapter->pvid == (rxcp->vlan_tag & VLAN_VID_MASK) &&
1832 !test_bit(rxcp->vlan_tag, adapter->vids))
1836 /* As the compl has been parsed, reset it; we wont touch it again */
1837 compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] = 0;
1839 queue_tail_inc(&rxo->cq);
1843 static inline struct page *be_alloc_pages(u32 size, gfp_t gfp)
1845 u32 order = get_order(size);
1849 return alloc_pages(gfp, order);
1853 * Allocate a page, split it to fragments of size rx_frag_size and post as
1854 * receive buffers to BE
1856 static void be_post_rx_frags(struct be_rx_obj *rxo, gfp_t gfp)
1858 struct be_adapter *adapter = rxo->adapter;
1859 struct be_rx_page_info *page_info = NULL, *prev_page_info = NULL;
1860 struct be_queue_info *rxq = &rxo->q;
1861 struct page *pagep = NULL;
1862 struct device *dev = &adapter->pdev->dev;
1863 struct be_eth_rx_d *rxd;
1864 u64 page_dmaaddr = 0, frag_dmaaddr;
1865 u32 posted, page_offset = 0;
1867 page_info = &rxo->page_info_tbl[rxq->head];
1868 for (posted = 0; posted < MAX_RX_POST && !page_info->page; posted++) {
1870 pagep = be_alloc_pages(adapter->big_page_size, gfp);
1871 if (unlikely(!pagep)) {
1872 rx_stats(rxo)->rx_post_fail++;
1875 page_dmaaddr = dma_map_page(dev, pagep, 0,
1876 adapter->big_page_size,
1878 if (dma_mapping_error(dev, page_dmaaddr)) {
1881 adapter->drv_stats.dma_map_errors++;
1887 page_offset += rx_frag_size;
1889 page_info->page_offset = page_offset;
1890 page_info->page = pagep;
1892 rxd = queue_head_node(rxq);
1893 frag_dmaaddr = page_dmaaddr + page_info->page_offset;
1894 rxd->fragpa_lo = cpu_to_le32(frag_dmaaddr & 0xFFFFFFFF);
1895 rxd->fragpa_hi = cpu_to_le32(upper_32_bits(frag_dmaaddr));
1897 /* Any space left in the current big page for another frag? */
1898 if ((page_offset + rx_frag_size + rx_frag_size) >
1899 adapter->big_page_size) {
1901 page_info->last_frag = true;
1902 dma_unmap_addr_set(page_info, bus, page_dmaaddr);
1904 dma_unmap_addr_set(page_info, bus, frag_dmaaddr);
1907 prev_page_info = page_info;
1908 queue_head_inc(rxq);
1909 page_info = &rxo->page_info_tbl[rxq->head];
1912 /* Mark the last frag of a page when we break out of the above loop
1913 * with no more slots available in the RXQ
1916 prev_page_info->last_frag = true;
1917 dma_unmap_addr_set(prev_page_info, bus, page_dmaaddr);
1921 atomic_add(posted, &rxq->used);
1922 if (rxo->rx_post_starved)
1923 rxo->rx_post_starved = false;
1924 be_rxq_notify(adapter, rxq->id, posted);
1925 } else if (atomic_read(&rxq->used) == 0) {
1926 /* Let be_worker replenish when memory is available */
1927 rxo->rx_post_starved = true;
1931 static struct be_eth_tx_compl *be_tx_compl_get(struct be_queue_info *tx_cq)
1933 struct be_eth_tx_compl *txcp = queue_tail_node(tx_cq);
1935 if (txcp->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] == 0)
1939 be_dws_le_to_cpu(txcp, sizeof(*txcp));
1941 txcp->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] = 0;
1943 queue_tail_inc(tx_cq);
1947 static u16 be_tx_compl_process(struct be_adapter *adapter,
1948 struct be_tx_obj *txo, u16 last_index)
1950 struct be_queue_info *txq = &txo->q;
1951 struct be_eth_wrb *wrb;
1952 struct sk_buff **sent_skbs = txo->sent_skb_list;
1953 struct sk_buff *sent_skb;
1954 u16 cur_index, num_wrbs = 1; /* account for hdr wrb */
1955 bool unmap_skb_hdr = true;
1957 sent_skb = sent_skbs[txq->tail];
1959 sent_skbs[txq->tail] = NULL;
1961 /* skip header wrb */
1962 queue_tail_inc(txq);
1965 cur_index = txq->tail;
1966 wrb = queue_tail_node(txq);
1967 unmap_tx_frag(&adapter->pdev->dev, wrb,
1968 (unmap_skb_hdr && skb_headlen(sent_skb)));
1969 unmap_skb_hdr = false;
1972 queue_tail_inc(txq);
1973 } while (cur_index != last_index);
1975 dev_consume_skb_any(sent_skb);
1979 /* Return the number of events in the event queue */
1980 static inline int events_get(struct be_eq_obj *eqo)
1982 struct be_eq_entry *eqe;
1986 eqe = queue_tail_node(&eqo->q);
1993 queue_tail_inc(&eqo->q);
1999 /* Leaves the EQ is disarmed state */
2000 static void be_eq_clean(struct be_eq_obj *eqo)
2002 int num = events_get(eqo);
2004 be_eq_notify(eqo->adapter, eqo->q.id, false, true, num);
2007 static void be_rx_cq_clean(struct be_rx_obj *rxo)
2009 struct be_rx_page_info *page_info;
2010 struct be_queue_info *rxq = &rxo->q;
2011 struct be_queue_info *rx_cq = &rxo->cq;
2012 struct be_rx_compl_info *rxcp;
2013 struct be_adapter *adapter = rxo->adapter;
2016 /* Consume pending rx completions.
2017 * Wait for the flush completion (identified by zero num_rcvd)
2018 * to arrive. Notify CQ even when there are no more CQ entries
2019 * for HW to flush partially coalesced CQ entries.
2020 * In Lancer, there is no need to wait for flush compl.
2023 rxcp = be_rx_compl_get(rxo);
2025 if (lancer_chip(adapter))
2028 if (flush_wait++ > 10 || be_hw_error(adapter)) {
2029 dev_warn(&adapter->pdev->dev,
2030 "did not receive flush compl\n");
2033 be_cq_notify(adapter, rx_cq->id, true, 0);
2036 be_rx_compl_discard(rxo, rxcp);
2037 be_cq_notify(adapter, rx_cq->id, false, 1);
2038 if (rxcp->num_rcvd == 0)
2043 /* After cleanup, leave the CQ in unarmed state */
2044 be_cq_notify(adapter, rx_cq->id, false, 0);
2046 /* Then free posted rx buffers that were not used */
2047 while (atomic_read(&rxq->used) > 0) {
2048 page_info = get_rx_page_info(rxo);
2049 put_page(page_info->page);
2050 memset(page_info, 0, sizeof(*page_info));
2052 BUG_ON(atomic_read(&rxq->used));
2053 rxq->tail = rxq->head = 0;
2056 static void be_tx_compl_clean(struct be_adapter *adapter)
2058 struct be_tx_obj *txo;
2059 struct be_queue_info *txq;
2060 struct be_eth_tx_compl *txcp;
2061 u16 end_idx, cmpl = 0, timeo = 0, num_wrbs = 0;
2062 struct sk_buff *sent_skb;
2064 int i, pending_txqs;
2066 /* Stop polling for compls when HW has been silent for 10ms */
2068 pending_txqs = adapter->num_tx_qs;
2070 for_all_tx_queues(adapter, txo, i) {
2074 while ((txcp = be_tx_compl_get(&txo->cq))) {
2075 end_idx = GET_TX_COMPL_BITS(wrb_index, txcp);
2076 num_wrbs += be_tx_compl_process(adapter, txo,
2081 be_cq_notify(adapter, txo->cq.id, false, cmpl);
2082 atomic_sub(num_wrbs, &txq->used);
2085 if (atomic_read(&txq->used) == 0)
2089 if (pending_txqs == 0 || ++timeo > 10 || be_hw_error(adapter))
2095 for_all_tx_queues(adapter, txo, i) {
2097 if (atomic_read(&txq->used))
2098 dev_err(&adapter->pdev->dev, "%d pending tx-compls\n",
2099 atomic_read(&txq->used));
2101 /* free posted tx for which compls will never arrive */
2102 while (atomic_read(&txq->used)) {
2103 sent_skb = txo->sent_skb_list[txq->tail];
2104 end_idx = txq->tail;
2105 num_wrbs = wrb_cnt_for_skb(adapter, sent_skb,
2107 index_adv(&end_idx, num_wrbs - 1, txq->len);
2108 num_wrbs = be_tx_compl_process(adapter, txo, end_idx);
2109 atomic_sub(num_wrbs, &txq->used);
2114 static void be_evt_queues_destroy(struct be_adapter *adapter)
2116 struct be_eq_obj *eqo;
2119 for_all_evt_queues(adapter, eqo, i) {
2120 if (eqo->q.created) {
2122 be_cmd_q_destroy(adapter, &eqo->q, QTYPE_EQ);
2123 napi_hash_del(&eqo->napi);
2124 netif_napi_del(&eqo->napi);
2126 be_queue_free(adapter, &eqo->q);
2130 static int be_evt_queues_create(struct be_adapter *adapter)
2132 struct be_queue_info *eq;
2133 struct be_eq_obj *eqo;
2134 struct be_aic_obj *aic;
2137 adapter->num_evt_qs = min_t(u16, num_irqs(adapter),
2138 adapter->cfg_num_qs);
2140 for_all_evt_queues(adapter, eqo, i) {
2141 netif_napi_add(adapter->netdev, &eqo->napi, be_poll,
2143 napi_hash_add(&eqo->napi);
2144 aic = &adapter->aic_obj[i];
2145 eqo->adapter = adapter;
2147 aic->max_eqd = BE_MAX_EQD;
2151 rc = be_queue_alloc(adapter, eq, EVNT_Q_LEN,
2152 sizeof(struct be_eq_entry));
2156 rc = be_cmd_eq_create(adapter, eqo);
2163 static void be_mcc_queues_destroy(struct be_adapter *adapter)
2165 struct be_queue_info *q;
2167 q = &adapter->mcc_obj.q;
2169 be_cmd_q_destroy(adapter, q, QTYPE_MCCQ);
2170 be_queue_free(adapter, q);
2172 q = &adapter->mcc_obj.cq;
2174 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2175 be_queue_free(adapter, q);
2178 /* Must be called only after TX qs are created as MCC shares TX EQ */
2179 static int be_mcc_queues_create(struct be_adapter *adapter)
2181 struct be_queue_info *q, *cq;
2183 cq = &adapter->mcc_obj.cq;
2184 if (be_queue_alloc(adapter, cq, MCC_CQ_LEN,
2185 sizeof(struct be_mcc_compl)))
2188 /* Use the default EQ for MCC completions */
2189 if (be_cmd_cq_create(adapter, cq, &mcc_eqo(adapter)->q, true, 0))
2192 q = &adapter->mcc_obj.q;
2193 if (be_queue_alloc(adapter, q, MCC_Q_LEN, sizeof(struct be_mcc_wrb)))
2194 goto mcc_cq_destroy;
2196 if (be_cmd_mccq_create(adapter, q, cq))
2202 be_queue_free(adapter, q);
2204 be_cmd_q_destroy(adapter, cq, QTYPE_CQ);
2206 be_queue_free(adapter, cq);
2211 static void be_tx_queues_destroy(struct be_adapter *adapter)
2213 struct be_queue_info *q;
2214 struct be_tx_obj *txo;
2217 for_all_tx_queues(adapter, txo, i) {
2220 be_cmd_q_destroy(adapter, q, QTYPE_TXQ);
2221 be_queue_free(adapter, q);
2225 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2226 be_queue_free(adapter, q);
2230 static int be_tx_qs_create(struct be_adapter *adapter)
2232 struct be_queue_info *cq, *eq;
2233 struct be_tx_obj *txo;
2236 adapter->num_tx_qs = min(adapter->num_evt_qs, be_max_txqs(adapter));
2238 for_all_tx_queues(adapter, txo, i) {
2240 status = be_queue_alloc(adapter, cq, TX_CQ_LEN,
2241 sizeof(struct be_eth_tx_compl));
2245 u64_stats_init(&txo->stats.sync);
2246 u64_stats_init(&txo->stats.sync_compl);
2248 /* If num_evt_qs is less than num_tx_qs, then more than
2249 * one txq share an eq
2251 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2252 status = be_cmd_cq_create(adapter, cq, eq, false, 3);
2256 status = be_queue_alloc(adapter, &txo->q, TX_Q_LEN,
2257 sizeof(struct be_eth_wrb));
2261 status = be_cmd_txq_create(adapter, txo);
2266 dev_info(&adapter->pdev->dev, "created %d TX queue(s)\n",
2267 adapter->num_tx_qs);
2271 static void be_rx_cqs_destroy(struct be_adapter *adapter)
2273 struct be_queue_info *q;
2274 struct be_rx_obj *rxo;
2277 for_all_rx_queues(adapter, rxo, i) {
2280 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2281 be_queue_free(adapter, q);
2285 static int be_rx_cqs_create(struct be_adapter *adapter)
2287 struct be_queue_info *eq, *cq;
2288 struct be_rx_obj *rxo;
2291 /* We can create as many RSS rings as there are EQs. */
2292 adapter->num_rx_qs = adapter->num_evt_qs;
2294 /* We'll use RSS only if atleast 2 RSS rings are supported.
2295 * When RSS is used, we'll need a default RXQ for non-IP traffic.
2297 if (adapter->num_rx_qs > 1)
2298 adapter->num_rx_qs++;
2300 adapter->big_page_size = (1 << get_order(rx_frag_size)) * PAGE_SIZE;
2301 for_all_rx_queues(adapter, rxo, i) {
2302 rxo->adapter = adapter;
2304 rc = be_queue_alloc(adapter, cq, RX_CQ_LEN,
2305 sizeof(struct be_eth_rx_compl));
2309 u64_stats_init(&rxo->stats.sync);
2310 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2311 rc = be_cmd_cq_create(adapter, cq, eq, false, 3);
2316 dev_info(&adapter->pdev->dev,
2317 "created %d RSS queue(s) and 1 default RX queue\n",
2318 adapter->num_rx_qs - 1);
2322 static irqreturn_t be_intx(int irq, void *dev)
2324 struct be_eq_obj *eqo = dev;
2325 struct be_adapter *adapter = eqo->adapter;
2328 /* IRQ is not expected when NAPI is scheduled as the EQ
2329 * will not be armed.
2330 * But, this can happen on Lancer INTx where it takes
2331 * a while to de-assert INTx or in BE2 where occasionaly
2332 * an interrupt may be raised even when EQ is unarmed.
2333 * If NAPI is already scheduled, then counting & notifying
2334 * events will orphan them.
2336 if (napi_schedule_prep(&eqo->napi)) {
2337 num_evts = events_get(eqo);
2338 __napi_schedule(&eqo->napi);
2340 eqo->spurious_intr = 0;
2342 be_eq_notify(adapter, eqo->q.id, false, true, num_evts);
2344 /* Return IRQ_HANDLED only for the the first spurious intr
2345 * after a valid intr to stop the kernel from branding
2346 * this irq as a bad one!
2348 if (num_evts || eqo->spurious_intr++ == 0)
2354 static irqreturn_t be_msix(int irq, void *dev)
2356 struct be_eq_obj *eqo = dev;
2358 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
2359 napi_schedule(&eqo->napi);
2363 static inline bool do_gro(struct be_rx_compl_info *rxcp)
2365 return (rxcp->tcpf && !rxcp->err && rxcp->l4_csum) ? true : false;
2368 static int be_process_rx(struct be_rx_obj *rxo, struct napi_struct *napi,
2369 int budget, int polling)
2371 struct be_adapter *adapter = rxo->adapter;
2372 struct be_queue_info *rx_cq = &rxo->cq;
2373 struct be_rx_compl_info *rxcp;
2376 for (work_done = 0; work_done < budget; work_done++) {
2377 rxcp = be_rx_compl_get(rxo);
2381 /* Is it a flush compl that has no data */
2382 if (unlikely(rxcp->num_rcvd == 0))
2385 /* Discard compl with partial DMA Lancer B0 */
2386 if (unlikely(!rxcp->pkt_size)) {
2387 be_rx_compl_discard(rxo, rxcp);
2391 /* On BE drop pkts that arrive due to imperfect filtering in
2392 * promiscuous mode on some skews
2394 if (unlikely(rxcp->port != adapter->port_num &&
2395 !lancer_chip(adapter))) {
2396 be_rx_compl_discard(rxo, rxcp);
2400 /* Don't do gro when we're busy_polling */
2401 if (do_gro(rxcp) && polling != BUSY_POLLING)
2402 be_rx_compl_process_gro(rxo, napi, rxcp);
2404 be_rx_compl_process(rxo, napi, rxcp);
2407 be_rx_stats_update(rxo, rxcp);
2411 be_cq_notify(adapter, rx_cq->id, true, work_done);
2413 /* When an rx-obj gets into post_starved state, just
2414 * let be_worker do the posting.
2416 if (atomic_read(&rxo->q.used) < RX_FRAGS_REFILL_WM &&
2417 !rxo->rx_post_starved)
2418 be_post_rx_frags(rxo, GFP_ATOMIC);
2424 static inline void be_update_tx_err(struct be_tx_obj *txo, u32 status)
2427 case BE_TX_COMP_HDR_PARSE_ERR:
2428 tx_stats(txo)->tx_hdr_parse_err++;
2430 case BE_TX_COMP_NDMA_ERR:
2431 tx_stats(txo)->tx_dma_err++;
2433 case BE_TX_COMP_ACL_ERR:
2434 tx_stats(txo)->tx_spoof_check_err++;
2439 static inline void lancer_update_tx_err(struct be_tx_obj *txo, u32 status)
2442 case LANCER_TX_COMP_LSO_ERR:
2443 tx_stats(txo)->tx_tso_err++;
2445 case LANCER_TX_COMP_HSW_DROP_MAC_ERR:
2446 case LANCER_TX_COMP_HSW_DROP_VLAN_ERR:
2447 tx_stats(txo)->tx_spoof_check_err++;
2449 case LANCER_TX_COMP_QINQ_ERR:
2450 tx_stats(txo)->tx_qinq_err++;
2452 case LANCER_TX_COMP_PARITY_ERR:
2453 tx_stats(txo)->tx_internal_parity_err++;
2455 case LANCER_TX_COMP_DMA_ERR:
2456 tx_stats(txo)->tx_dma_err++;
2461 static void be_process_tx(struct be_adapter *adapter, struct be_tx_obj *txo,
2464 struct be_eth_tx_compl *txcp;
2465 int num_wrbs = 0, work_done = 0;
2469 while ((txcp = be_tx_compl_get(&txo->cq))) {
2470 last_idx = GET_TX_COMPL_BITS(wrb_index, txcp);
2471 num_wrbs += be_tx_compl_process(adapter, txo, last_idx);
2474 compl_status = GET_TX_COMPL_BITS(status, txcp);
2476 if (lancer_chip(adapter))
2477 lancer_update_tx_err(txo, compl_status);
2479 be_update_tx_err(txo, compl_status);
2484 be_cq_notify(adapter, txo->cq.id, true, work_done);
2485 atomic_sub(num_wrbs, &txo->q.used);
2487 /* As Tx wrbs have been freed up, wake up netdev queue
2488 * if it was stopped due to lack of tx wrbs. */
2489 if (__netif_subqueue_stopped(adapter->netdev, idx) &&
2490 atomic_read(&txo->q.used) < txo->q.len / 2) {
2491 netif_wake_subqueue(adapter->netdev, idx);
2494 u64_stats_update_begin(&tx_stats(txo)->sync_compl);
2495 tx_stats(txo)->tx_compl += work_done;
2496 u64_stats_update_end(&tx_stats(txo)->sync_compl);
2500 int be_poll(struct napi_struct *napi, int budget)
2502 struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2503 struct be_adapter *adapter = eqo->adapter;
2504 int max_work = 0, work, i, num_evts;
2505 struct be_rx_obj *rxo;
2506 struct be_tx_obj *txo;
2508 num_evts = events_get(eqo);
2510 for_all_tx_queues_on_eq(adapter, eqo, txo, i)
2511 be_process_tx(adapter, txo, i);
2513 if (be_lock_napi(eqo)) {
2514 /* This loop will iterate twice for EQ0 in which
2515 * completions of the last RXQ (default one) are also processed
2516 * For other EQs the loop iterates only once
2518 for_all_rx_queues_on_eq(adapter, eqo, rxo, i) {
2519 work = be_process_rx(rxo, napi, budget, NAPI_POLLING);
2520 max_work = max(work, max_work);
2522 be_unlock_napi(eqo);
2527 if (is_mcc_eqo(eqo))
2528 be_process_mcc(adapter);
2530 if (max_work < budget) {
2531 napi_complete(napi);
2532 be_eq_notify(adapter, eqo->q.id, true, false, num_evts);
2534 /* As we'll continue in polling mode, count and clear events */
2535 be_eq_notify(adapter, eqo->q.id, false, false, num_evts);
2540 #ifdef CONFIG_NET_RX_BUSY_POLL
2541 static int be_busy_poll(struct napi_struct *napi)
2543 struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2544 struct be_adapter *adapter = eqo->adapter;
2545 struct be_rx_obj *rxo;
2548 if (!be_lock_busy_poll(eqo))
2549 return LL_FLUSH_BUSY;
2551 for_all_rx_queues_on_eq(adapter, eqo, rxo, i) {
2552 work = be_process_rx(rxo, napi, 4, BUSY_POLLING);
2557 be_unlock_busy_poll(eqo);
2562 void be_detect_error(struct be_adapter *adapter)
2564 u32 ue_lo = 0, ue_hi = 0, ue_lo_mask = 0, ue_hi_mask = 0;
2565 u32 sliport_status = 0, sliport_err1 = 0, sliport_err2 = 0;
2567 bool error_detected = false;
2568 struct device *dev = &adapter->pdev->dev;
2569 struct net_device *netdev = adapter->netdev;
2571 if (be_hw_error(adapter))
2574 if (lancer_chip(adapter)) {
2575 sliport_status = ioread32(adapter->db + SLIPORT_STATUS_OFFSET);
2576 if (sliport_status & SLIPORT_STATUS_ERR_MASK) {
2577 sliport_err1 = ioread32(adapter->db +
2578 SLIPORT_ERROR1_OFFSET);
2579 sliport_err2 = ioread32(adapter->db +
2580 SLIPORT_ERROR2_OFFSET);
2581 adapter->hw_error = true;
2582 /* Do not log error messages if its a FW reset */
2583 if (sliport_err1 == SLIPORT_ERROR_FW_RESET1 &&
2584 sliport_err2 == SLIPORT_ERROR_FW_RESET2) {
2585 dev_info(dev, "Firmware update in progress\n");
2587 error_detected = true;
2588 dev_err(dev, "Error detected in the card\n");
2589 dev_err(dev, "ERR: sliport status 0x%x\n",
2591 dev_err(dev, "ERR: sliport error1 0x%x\n",
2593 dev_err(dev, "ERR: sliport error2 0x%x\n",
2598 pci_read_config_dword(adapter->pdev,
2599 PCICFG_UE_STATUS_LOW, &ue_lo);
2600 pci_read_config_dword(adapter->pdev,
2601 PCICFG_UE_STATUS_HIGH, &ue_hi);
2602 pci_read_config_dword(adapter->pdev,
2603 PCICFG_UE_STATUS_LOW_MASK, &ue_lo_mask);
2604 pci_read_config_dword(adapter->pdev,
2605 PCICFG_UE_STATUS_HI_MASK, &ue_hi_mask);
2607 ue_lo = (ue_lo & ~ue_lo_mask);
2608 ue_hi = (ue_hi & ~ue_hi_mask);
2610 /* On certain platforms BE hardware can indicate spurious UEs.
2611 * Allow HW to stop working completely in case of a real UE.
2612 * Hence not setting the hw_error for UE detection.
2615 if (ue_lo || ue_hi) {
2616 error_detected = true;
2618 "Unrecoverable Error detected in the adapter");
2619 dev_err(dev, "Please reboot server to recover");
2620 if (skyhawk_chip(adapter))
2621 adapter->hw_error = true;
2622 for (i = 0; ue_lo; ue_lo >>= 1, i++) {
2624 dev_err(dev, "UE: %s bit set\n",
2625 ue_status_low_desc[i]);
2627 for (i = 0; ue_hi; ue_hi >>= 1, i++) {
2629 dev_err(dev, "UE: %s bit set\n",
2630 ue_status_hi_desc[i]);
2635 netif_carrier_off(netdev);
2638 static void be_msix_disable(struct be_adapter *adapter)
2640 if (msix_enabled(adapter)) {
2641 pci_disable_msix(adapter->pdev);
2642 adapter->num_msix_vec = 0;
2643 adapter->num_msix_roce_vec = 0;
2647 static int be_msix_enable(struct be_adapter *adapter)
2650 struct device *dev = &adapter->pdev->dev;
2652 /* If RoCE is supported, program the max number of NIC vectors that
2653 * may be configured via set-channels, along with vectors needed for
2654 * RoCe. Else, just program the number we'll use initially.
2656 if (be_roce_supported(adapter))
2657 num_vec = min_t(int, 2 * be_max_eqs(adapter),
2658 2 * num_online_cpus());
2660 num_vec = adapter->cfg_num_qs;
2662 for (i = 0; i < num_vec; i++)
2663 adapter->msix_entries[i].entry = i;
2665 num_vec = pci_enable_msix_range(adapter->pdev, adapter->msix_entries,
2666 MIN_MSIX_VECTORS, num_vec);
2670 if (be_roce_supported(adapter) && num_vec > MIN_MSIX_VECTORS) {
2671 adapter->num_msix_roce_vec = num_vec / 2;
2672 dev_info(dev, "enabled %d MSI-x vector(s) for RoCE\n",
2673 adapter->num_msix_roce_vec);
2676 adapter->num_msix_vec = num_vec - adapter->num_msix_roce_vec;
2678 dev_info(dev, "enabled %d MSI-x vector(s) for NIC\n",
2679 adapter->num_msix_vec);
2683 dev_warn(dev, "MSIx enable failed\n");
2685 /* INTx is not supported in VFs, so fail probe if enable_msix fails */
2686 if (!be_physfn(adapter))
2691 static inline int be_msix_vec_get(struct be_adapter *adapter,
2692 struct be_eq_obj *eqo)
2694 return adapter->msix_entries[eqo->msix_idx].vector;
2697 static int be_msix_register(struct be_adapter *adapter)
2699 struct net_device *netdev = adapter->netdev;
2700 struct be_eq_obj *eqo;
2703 for_all_evt_queues(adapter, eqo, i) {
2704 sprintf(eqo->desc, "%s-q%d", netdev->name, i);
2705 vec = be_msix_vec_get(adapter, eqo);
2706 status = request_irq(vec, be_msix, 0, eqo->desc, eqo);
2713 for (i--, eqo = &adapter->eq_obj[i]; i >= 0; i--, eqo--)
2714 free_irq(be_msix_vec_get(adapter, eqo), eqo);
2715 dev_warn(&adapter->pdev->dev, "MSIX Request IRQ failed - err %d\n",
2717 be_msix_disable(adapter);
2721 static int be_irq_register(struct be_adapter *adapter)
2723 struct net_device *netdev = adapter->netdev;
2726 if (msix_enabled(adapter)) {
2727 status = be_msix_register(adapter);
2730 /* INTx is not supported for VF */
2731 if (!be_physfn(adapter))
2735 /* INTx: only the first EQ is used */
2736 netdev->irq = adapter->pdev->irq;
2737 status = request_irq(netdev->irq, be_intx, IRQF_SHARED, netdev->name,
2738 &adapter->eq_obj[0]);
2740 dev_err(&adapter->pdev->dev,
2741 "INTx request IRQ failed - err %d\n", status);
2745 adapter->isr_registered = true;
2749 static void be_irq_unregister(struct be_adapter *adapter)
2751 struct net_device *netdev = adapter->netdev;
2752 struct be_eq_obj *eqo;
2755 if (!adapter->isr_registered)
2759 if (!msix_enabled(adapter)) {
2760 free_irq(netdev->irq, &adapter->eq_obj[0]);
2765 for_all_evt_queues(adapter, eqo, i)
2766 free_irq(be_msix_vec_get(adapter, eqo), eqo);
2769 adapter->isr_registered = false;
2772 static void be_rx_qs_destroy(struct be_adapter *adapter)
2774 struct be_queue_info *q;
2775 struct be_rx_obj *rxo;
2778 for_all_rx_queues(adapter, rxo, i) {
2781 be_cmd_rxq_destroy(adapter, q);
2782 be_rx_cq_clean(rxo);
2784 be_queue_free(adapter, q);
2788 static int be_close(struct net_device *netdev)
2790 struct be_adapter *adapter = netdev_priv(netdev);
2791 struct be_eq_obj *eqo;
2794 /* This protection is needed as be_close() may be called even when the
2795 * adapter is in cleared state (after eeh perm failure)
2797 if (!(adapter->flags & BE_FLAGS_SETUP_DONE))
2800 be_roce_dev_close(adapter);
2802 if (adapter->flags & BE_FLAGS_NAPI_ENABLED) {
2803 for_all_evt_queues(adapter, eqo, i) {
2804 napi_disable(&eqo->napi);
2805 be_disable_busy_poll(eqo);
2807 adapter->flags &= ~BE_FLAGS_NAPI_ENABLED;
2810 be_async_mcc_disable(adapter);
2812 /* Wait for all pending tx completions to arrive so that
2813 * all tx skbs are freed.
2815 netif_tx_disable(netdev);
2816 be_tx_compl_clean(adapter);
2818 be_rx_qs_destroy(adapter);
2820 for (i = 1; i < (adapter->uc_macs + 1); i++)
2821 be_cmd_pmac_del(adapter, adapter->if_handle,
2822 adapter->pmac_id[i], 0);
2823 adapter->uc_macs = 0;
2825 for_all_evt_queues(adapter, eqo, i) {
2826 if (msix_enabled(adapter))
2827 synchronize_irq(be_msix_vec_get(adapter, eqo));
2829 synchronize_irq(netdev->irq);
2833 be_irq_unregister(adapter);
2838 static int be_rx_qs_create(struct be_adapter *adapter)
2840 struct be_rx_obj *rxo;
2842 u8 rss_hkey[RSS_HASH_KEY_LEN];
2843 struct rss_info *rss = &adapter->rss_info;
2845 for_all_rx_queues(adapter, rxo, i) {
2846 rc = be_queue_alloc(adapter, &rxo->q, RX_Q_LEN,
2847 sizeof(struct be_eth_rx_d));
2852 /* The FW would like the default RXQ to be created first */
2853 rxo = default_rxo(adapter);
2854 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id, rx_frag_size,
2855 adapter->if_handle, false, &rxo->rss_id);
2859 for_all_rss_queues(adapter, rxo, i) {
2860 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id,
2861 rx_frag_size, adapter->if_handle,
2862 true, &rxo->rss_id);
2867 if (be_multi_rxq(adapter)) {
2868 for (j = 0; j < RSS_INDIR_TABLE_LEN;
2869 j += adapter->num_rx_qs - 1) {
2870 for_all_rss_queues(adapter, rxo, i) {
2871 if ((j + i) >= RSS_INDIR_TABLE_LEN)
2873 rss->rsstable[j + i] = rxo->rss_id;
2874 rss->rss_queue[j + i] = i;
2877 rss->rss_flags = RSS_ENABLE_TCP_IPV4 | RSS_ENABLE_IPV4 |
2878 RSS_ENABLE_TCP_IPV6 | RSS_ENABLE_IPV6;
2880 if (!BEx_chip(adapter))
2881 rss->rss_flags |= RSS_ENABLE_UDP_IPV4 |
2882 RSS_ENABLE_UDP_IPV6;
2884 /* Disable RSS, if only default RX Q is created */
2885 rss->rss_flags = RSS_ENABLE_NONE;
2888 get_random_bytes(rss_hkey, RSS_HASH_KEY_LEN);
2889 rc = be_cmd_rss_config(adapter, rss->rsstable, rss->rss_flags,
2892 rss->rss_flags = RSS_ENABLE_NONE;
2896 memcpy(rss->rss_hkey, rss_hkey, RSS_HASH_KEY_LEN);
2898 /* First time posting */
2899 for_all_rx_queues(adapter, rxo, i)
2900 be_post_rx_frags(rxo, GFP_KERNEL);
2904 static int be_open(struct net_device *netdev)
2906 struct be_adapter *adapter = netdev_priv(netdev);
2907 struct be_eq_obj *eqo;
2908 struct be_rx_obj *rxo;
2909 struct be_tx_obj *txo;
2913 status = be_rx_qs_create(adapter);
2917 status = be_irq_register(adapter);
2921 for_all_rx_queues(adapter, rxo, i)
2922 be_cq_notify(adapter, rxo->cq.id, true, 0);
2924 for_all_tx_queues(adapter, txo, i)
2925 be_cq_notify(adapter, txo->cq.id, true, 0);
2927 be_async_mcc_enable(adapter);
2929 for_all_evt_queues(adapter, eqo, i) {
2930 napi_enable(&eqo->napi);
2931 be_enable_busy_poll(eqo);
2932 be_eq_notify(adapter, eqo->q.id, true, true, 0);
2934 adapter->flags |= BE_FLAGS_NAPI_ENABLED;
2936 status = be_cmd_link_status_query(adapter, NULL, &link_status, 0);
2938 be_link_status_update(adapter, link_status);
2940 netif_tx_start_all_queues(netdev);
2941 be_roce_dev_open(adapter);
2943 #ifdef CONFIG_BE2NET_VXLAN
2944 if (skyhawk_chip(adapter))
2945 vxlan_get_rx_port(netdev);
2950 be_close(adapter->netdev);
2954 static int be_setup_wol(struct be_adapter *adapter, bool enable)
2956 struct be_dma_mem cmd;
2960 memset(mac, 0, ETH_ALEN);
2962 cmd.size = sizeof(struct be_cmd_req_acpi_wol_magic_config);
2963 cmd.va = dma_zalloc_coherent(&adapter->pdev->dev, cmd.size, &cmd.dma,
2969 status = pci_write_config_dword(adapter->pdev,
2970 PCICFG_PM_CONTROL_OFFSET,
2971 PCICFG_PM_CONTROL_MASK);
2973 dev_err(&adapter->pdev->dev,
2974 "Could not enable Wake-on-lan\n");
2975 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va,
2979 status = be_cmd_enable_magic_wol(adapter,
2980 adapter->netdev->dev_addr,
2982 pci_enable_wake(adapter->pdev, PCI_D3hot, 1);
2983 pci_enable_wake(adapter->pdev, PCI_D3cold, 1);
2985 status = be_cmd_enable_magic_wol(adapter, mac, &cmd);
2986 pci_enable_wake(adapter->pdev, PCI_D3hot, 0);
2987 pci_enable_wake(adapter->pdev, PCI_D3cold, 0);
2990 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va, cmd.dma);
2995 * Generate a seed MAC address from the PF MAC Address using jhash.
2996 * MAC Address for VFs are assigned incrementally starting from the seed.
2997 * These addresses are programmed in the ASIC by the PF and the VF driver
2998 * queries for the MAC address during its probe.
3000 static int be_vf_eth_addr_config(struct be_adapter *adapter)
3005 struct be_vf_cfg *vf_cfg;
3007 be_vf_eth_addr_generate(adapter, mac);
3009 for_all_vfs(adapter, vf_cfg, vf) {
3010 if (BEx_chip(adapter))
3011 status = be_cmd_pmac_add(adapter, mac,
3013 &vf_cfg->pmac_id, vf + 1);
3015 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
3019 dev_err(&adapter->pdev->dev,
3020 "Mac address assignment failed for VF %d\n",
3023 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
3030 static int be_vfs_mac_query(struct be_adapter *adapter)
3034 struct be_vf_cfg *vf_cfg;
3036 for_all_vfs(adapter, vf_cfg, vf) {
3037 status = be_cmd_get_active_mac(adapter, vf_cfg->pmac_id,
3038 mac, vf_cfg->if_handle,
3042 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
3047 static void be_vf_clear(struct be_adapter *adapter)
3049 struct be_vf_cfg *vf_cfg;
3052 if (pci_vfs_assigned(adapter->pdev)) {
3053 dev_warn(&adapter->pdev->dev,
3054 "VFs are assigned to VMs: not disabling VFs\n");
3058 pci_disable_sriov(adapter->pdev);
3060 for_all_vfs(adapter, vf_cfg, vf) {
3061 if (BEx_chip(adapter))
3062 be_cmd_pmac_del(adapter, vf_cfg->if_handle,
3063 vf_cfg->pmac_id, vf + 1);
3065 be_cmd_set_mac(adapter, NULL, vf_cfg->if_handle,
3068 be_cmd_if_destroy(adapter, vf_cfg->if_handle, vf + 1);
3071 kfree(adapter->vf_cfg);
3072 adapter->num_vfs = 0;
3073 adapter->flags &= ~BE_FLAGS_SRIOV_ENABLED;
3076 static void be_clear_queues(struct be_adapter *adapter)
3078 be_mcc_queues_destroy(adapter);
3079 be_rx_cqs_destroy(adapter);
3080 be_tx_queues_destroy(adapter);
3081 be_evt_queues_destroy(adapter);
3084 static void be_cancel_worker(struct be_adapter *adapter)
3086 if (adapter->flags & BE_FLAGS_WORKER_SCHEDULED) {
3087 cancel_delayed_work_sync(&adapter->work);
3088 adapter->flags &= ~BE_FLAGS_WORKER_SCHEDULED;
3092 static void be_mac_clear(struct be_adapter *adapter)
3096 if (adapter->pmac_id) {
3097 for (i = 0; i < (adapter->uc_macs + 1); i++)
3098 be_cmd_pmac_del(adapter, adapter->if_handle,
3099 adapter->pmac_id[i], 0);
3100 adapter->uc_macs = 0;
3102 kfree(adapter->pmac_id);
3103 adapter->pmac_id = NULL;
3107 #ifdef CONFIG_BE2NET_VXLAN
3108 static void be_disable_vxlan_offloads(struct be_adapter *adapter)
3110 if (adapter->flags & BE_FLAGS_VXLAN_OFFLOADS)
3111 be_cmd_manage_iface(adapter, adapter->if_handle,
3112 OP_CONVERT_TUNNEL_TO_NORMAL);
3114 if (adapter->vxlan_port)
3115 be_cmd_set_vxlan_port(adapter, 0);
3117 adapter->flags &= ~BE_FLAGS_VXLAN_OFFLOADS;
3118 adapter->vxlan_port = 0;
3122 static int be_clear(struct be_adapter *adapter)
3124 be_cancel_worker(adapter);
3126 if (sriov_enabled(adapter))
3127 be_vf_clear(adapter);
3129 /* Re-configure FW to distribute resources evenly across max-supported
3130 * number of VFs, only when VFs are not already enabled.
3132 if (be_physfn(adapter) && !pci_vfs_assigned(adapter->pdev))
3133 be_cmd_set_sriov_config(adapter, adapter->pool_res,
3134 pci_sriov_get_totalvfs(adapter->pdev));
3136 #ifdef CONFIG_BE2NET_VXLAN
3137 be_disable_vxlan_offloads(adapter);
3139 /* delete the primary mac along with the uc-mac list */
3140 be_mac_clear(adapter);
3142 be_cmd_if_destroy(adapter, adapter->if_handle, 0);
3144 be_clear_queues(adapter);
3146 be_msix_disable(adapter);
3147 adapter->flags &= ~BE_FLAGS_SETUP_DONE;
3151 static int be_vfs_if_create(struct be_adapter *adapter)
3153 struct be_resources res = {0};
3154 struct be_vf_cfg *vf_cfg;
3155 u32 cap_flags, en_flags, vf;
3158 cap_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3159 BE_IF_FLAGS_MULTICAST;
3161 for_all_vfs(adapter, vf_cfg, vf) {
3162 if (!BE3_chip(adapter)) {
3163 status = be_cmd_get_profile_config(adapter, &res,
3166 cap_flags = res.if_cap_flags;
3169 /* If a FW profile exists, then cap_flags are updated */
3170 en_flags = cap_flags & (BE_IF_FLAGS_UNTAGGED |
3171 BE_IF_FLAGS_BROADCAST |
3172 BE_IF_FLAGS_MULTICAST);
3174 be_cmd_if_create(adapter, cap_flags, en_flags,
3175 &vf_cfg->if_handle, vf + 1);
3183 static int be_vf_setup_init(struct be_adapter *adapter)
3185 struct be_vf_cfg *vf_cfg;
3188 adapter->vf_cfg = kcalloc(adapter->num_vfs, sizeof(*vf_cfg),
3190 if (!adapter->vf_cfg)
3193 for_all_vfs(adapter, vf_cfg, vf) {
3194 vf_cfg->if_handle = -1;
3195 vf_cfg->pmac_id = -1;
3200 static int be_vf_setup(struct be_adapter *adapter)
3202 struct device *dev = &adapter->pdev->dev;
3203 struct be_vf_cfg *vf_cfg;
3204 int status, old_vfs, vf;
3207 old_vfs = pci_num_vf(adapter->pdev);
3209 status = be_vf_setup_init(adapter);
3214 for_all_vfs(adapter, vf_cfg, vf) {
3215 status = be_cmd_get_if_id(adapter, vf_cfg, vf);
3220 status = be_vfs_mac_query(adapter);
3224 status = be_vfs_if_create(adapter);
3228 status = be_vf_eth_addr_config(adapter);
3233 for_all_vfs(adapter, vf_cfg, vf) {
3234 /* Allow VFs to programs MAC/VLAN filters */
3235 status = be_cmd_get_fn_privileges(adapter, &privileges, vf + 1);
3236 if (!status && !(privileges & BE_PRIV_FILTMGMT)) {
3237 status = be_cmd_set_fn_privileges(adapter,
3242 dev_info(dev, "VF%d has FILTMGMT privilege\n",
3246 /* Allow full available bandwidth */
3248 be_cmd_config_qos(adapter, 0, 0, vf + 1);
3251 be_cmd_enable_vf(adapter, vf + 1);
3252 be_cmd_set_logical_link_config(adapter,
3253 IFLA_VF_LINK_STATE_AUTO,
3259 status = pci_enable_sriov(adapter->pdev, adapter->num_vfs);
3261 dev_err(dev, "SRIOV enable failed\n");
3262 adapter->num_vfs = 0;
3267 adapter->flags |= BE_FLAGS_SRIOV_ENABLED;
3270 dev_err(dev, "VF setup failed\n");
3271 be_vf_clear(adapter);
3275 /* Converting function_mode bits on BE3 to SH mc_type enums */
3277 static u8 be_convert_mc_type(u32 function_mode)
3279 if (function_mode & VNIC_MODE && function_mode & QNQ_MODE)
3281 else if (function_mode & QNQ_MODE)
3283 else if (function_mode & VNIC_MODE)
3285 else if (function_mode & UMC_ENABLED)
3291 /* On BE2/BE3 FW does not suggest the supported limits */
3292 static void BEx_get_resources(struct be_adapter *adapter,
3293 struct be_resources *res)
3295 bool use_sriov = adapter->num_vfs ? 1 : 0;
3297 if (be_physfn(adapter))
3298 res->max_uc_mac = BE_UC_PMAC_COUNT;
3300 res->max_uc_mac = BE_VF_UC_PMAC_COUNT;
3302 adapter->mc_type = be_convert_mc_type(adapter->function_mode);
3304 if (be_is_mc(adapter)) {
3305 /* Assuming that there are 4 channels per port,
3306 * when multi-channel is enabled
3308 if (be_is_qnq_mode(adapter))
3309 res->max_vlans = BE_NUM_VLANS_SUPPORTED/8;
3311 /* In a non-qnq multichannel mode, the pvid
3312 * takes up one vlan entry
3314 res->max_vlans = (BE_NUM_VLANS_SUPPORTED / 4) - 1;
3316 res->max_vlans = BE_NUM_VLANS_SUPPORTED;
3319 res->max_mcast_mac = BE_MAX_MC;
3321 /* 1) For BE3 1Gb ports, FW does not support multiple TXQs
3322 * 2) Create multiple TX rings on a BE3-R multi-channel interface
3323 * *only* if it is RSS-capable.
3325 if (BE2_chip(adapter) || use_sriov || (adapter->port_num > 1) ||
3326 !be_physfn(adapter) || (be_is_mc(adapter) &&
3327 !(adapter->function_caps & BE_FUNCTION_CAPS_RSS))) {
3329 } else if (adapter->function_caps & BE_FUNCTION_CAPS_SUPER_NIC) {
3330 struct be_resources super_nic_res = {0};
3332 /* On a SuperNIC profile, the driver needs to use the
3333 * GET_PROFILE_CONFIG cmd to query the per-function TXQ limits
3335 be_cmd_get_profile_config(adapter, &super_nic_res, 0);
3336 /* Some old versions of BE3 FW don't report max_tx_qs value */
3337 res->max_tx_qs = super_nic_res.max_tx_qs ? : BE3_MAX_TX_QS;
3339 res->max_tx_qs = BE3_MAX_TX_QS;
3342 if ((adapter->function_caps & BE_FUNCTION_CAPS_RSS) &&
3343 !use_sriov && be_physfn(adapter))
3344 res->max_rss_qs = (adapter->be3_native) ?
3345 BE3_MAX_RSS_QS : BE2_MAX_RSS_QS;
3346 res->max_rx_qs = res->max_rss_qs + 1;
3348 if (be_physfn(adapter))
3349 res->max_evt_qs = (be_max_vfs(adapter) > 0) ?
3350 BE3_SRIOV_MAX_EVT_QS : BE3_MAX_EVT_QS;
3352 res->max_evt_qs = 1;
3354 res->if_cap_flags = BE_IF_CAP_FLAGS_WANT;
3355 if (!(adapter->function_caps & BE_FUNCTION_CAPS_RSS))
3356 res->if_cap_flags &= ~BE_IF_FLAGS_RSS;
3359 static void be_setup_init(struct be_adapter *adapter)
3361 adapter->vlan_prio_bmap = 0xff;
3362 adapter->phy.link_speed = -1;
3363 adapter->if_handle = -1;
3364 adapter->be3_native = false;
3365 adapter->promiscuous = false;
3366 if (be_physfn(adapter))
3367 adapter->cmd_privileges = MAX_PRIVILEGES;
3369 adapter->cmd_privileges = MIN_PRIVILEGES;
3372 static int be_get_sriov_config(struct be_adapter *adapter)
3374 struct device *dev = &adapter->pdev->dev;
3375 struct be_resources res = {0};
3376 int max_vfs, old_vfs;
3378 /* Some old versions of BE3 FW don't report max_vfs value */
3379 be_cmd_get_profile_config(adapter, &res, 0);
3381 if (BE3_chip(adapter) && !res.max_vfs) {
3382 max_vfs = pci_sriov_get_totalvfs(adapter->pdev);
3383 res.max_vfs = max_vfs > 0 ? min(MAX_VFS, max_vfs) : 0;
3386 adapter->pool_res = res;
3388 if (!be_max_vfs(adapter)) {
3390 dev_warn(dev, "device doesn't support SRIOV\n");
3391 adapter->num_vfs = 0;
3395 pci_sriov_set_totalvfs(adapter->pdev, be_max_vfs(adapter));
3397 /* validate num_vfs module param */
3398 old_vfs = pci_num_vf(adapter->pdev);
3400 dev_info(dev, "%d VFs are already enabled\n", old_vfs);
3401 if (old_vfs != num_vfs)
3402 dev_warn(dev, "Ignoring num_vfs=%d setting\n", num_vfs);
3403 adapter->num_vfs = old_vfs;
3405 if (num_vfs > be_max_vfs(adapter)) {
3406 dev_info(dev, "Resources unavailable to init %d VFs\n",
3408 dev_info(dev, "Limiting to %d VFs\n",
3409 be_max_vfs(adapter));
3411 adapter->num_vfs = min_t(u16, num_vfs, be_max_vfs(adapter));
3417 static int be_get_resources(struct be_adapter *adapter)
3419 struct device *dev = &adapter->pdev->dev;
3420 struct be_resources res = {0};
3423 if (BEx_chip(adapter)) {
3424 BEx_get_resources(adapter, &res);
3428 /* For Lancer, SH etc read per-function resource limits from FW.
3429 * GET_FUNC_CONFIG returns per function guaranteed limits.
3430 * GET_PROFILE_CONFIG returns PCI-E related limits PF-pool limits
3432 if (!BEx_chip(adapter)) {
3433 status = be_cmd_get_func_config(adapter, &res);
3437 /* If RoCE may be enabled stash away half the EQs for RoCE */
3438 if (be_roce_supported(adapter))
3439 res.max_evt_qs /= 2;
3443 dev_info(dev, "Max: txqs %d, rxqs %d, rss %d, eqs %d, vfs %d\n",
3444 be_max_txqs(adapter), be_max_rxqs(adapter),
3445 be_max_rss(adapter), be_max_eqs(adapter),
3446 be_max_vfs(adapter));
3447 dev_info(dev, "Max: uc-macs %d, mc-macs %d, vlans %d\n",
3448 be_max_uc(adapter), be_max_mc(adapter),
3449 be_max_vlans(adapter));
3454 static void be_sriov_config(struct be_adapter *adapter)
3456 struct device *dev = &adapter->pdev->dev;
3459 status = be_get_sriov_config(adapter);
3461 dev_err(dev, "Failed to query SR-IOV configuration\n");
3462 dev_err(dev, "SR-IOV cannot be enabled\n");
3466 /* When the HW is in SRIOV capable configuration, the PF-pool
3467 * resources are equally distributed across the max-number of
3468 * VFs. The user may request only a subset of the max-vfs to be
3469 * enabled. Based on num_vfs, redistribute the resources across
3470 * num_vfs so that each VF will have access to more number of
3471 * resources. This facility is not available in BE3 FW.
3472 * Also, this is done by FW in Lancer chip.
3474 if (be_max_vfs(adapter) && !pci_num_vf(adapter->pdev)) {
3475 status = be_cmd_set_sriov_config(adapter,
3479 dev_err(dev, "Failed to optimize SR-IOV resources\n");
3483 static int be_get_config(struct be_adapter *adapter)
3488 status = be_cmd_query_fw_cfg(adapter);
3492 if (be_physfn(adapter)) {
3493 status = be_cmd_get_active_profile(adapter, &profile_id);
3495 dev_info(&adapter->pdev->dev,
3496 "Using profile 0x%x\n", profile_id);
3499 if (!BE2_chip(adapter) && be_physfn(adapter))
3500 be_sriov_config(adapter);
3502 status = be_get_resources(adapter);
3506 adapter->pmac_id = kcalloc(be_max_uc(adapter),
3507 sizeof(*adapter->pmac_id), GFP_KERNEL);
3508 if (!adapter->pmac_id)
3511 /* Sanitize cfg_num_qs based on HW and platform limits */
3512 adapter->cfg_num_qs = min(adapter->cfg_num_qs, be_max_qs(adapter));
3517 static int be_mac_setup(struct be_adapter *adapter)
3522 if (is_zero_ether_addr(adapter->netdev->dev_addr)) {
3523 status = be_cmd_get_perm_mac(adapter, mac);
3527 memcpy(adapter->netdev->dev_addr, mac, ETH_ALEN);
3528 memcpy(adapter->netdev->perm_addr, mac, ETH_ALEN);
3530 /* Maybe the HW was reset; dev_addr must be re-programmed */
3531 memcpy(mac, adapter->netdev->dev_addr, ETH_ALEN);
3534 /* For BE3-R VFs, the PF programs the initial MAC address */
3535 if (!(BEx_chip(adapter) && be_virtfn(adapter)))
3536 be_cmd_pmac_add(adapter, mac, adapter->if_handle,
3537 &adapter->pmac_id[0], 0);
3541 static void be_schedule_worker(struct be_adapter *adapter)
3543 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
3544 adapter->flags |= BE_FLAGS_WORKER_SCHEDULED;
3547 static int be_setup_queues(struct be_adapter *adapter)
3549 struct net_device *netdev = adapter->netdev;
3552 status = be_evt_queues_create(adapter);
3556 status = be_tx_qs_create(adapter);
3560 status = be_rx_cqs_create(adapter);
3564 status = be_mcc_queues_create(adapter);
3568 status = netif_set_real_num_rx_queues(netdev, adapter->num_rx_qs);
3572 status = netif_set_real_num_tx_queues(netdev, adapter->num_tx_qs);
3578 dev_err(&adapter->pdev->dev, "queue_setup failed\n");
3582 int be_update_queues(struct be_adapter *adapter)
3584 struct net_device *netdev = adapter->netdev;
3587 if (netif_running(netdev))
3590 be_cancel_worker(adapter);
3592 /* If any vectors have been shared with RoCE we cannot re-program
3595 if (!adapter->num_msix_roce_vec)
3596 be_msix_disable(adapter);
3598 be_clear_queues(adapter);
3600 if (!msix_enabled(adapter)) {
3601 status = be_msix_enable(adapter);
3606 status = be_setup_queues(adapter);
3610 be_schedule_worker(adapter);
3612 if (netif_running(netdev))
3613 status = be_open(netdev);
3618 static int be_setup(struct be_adapter *adapter)
3620 struct device *dev = &adapter->pdev->dev;
3621 u32 tx_fc, rx_fc, en_flags;
3624 be_setup_init(adapter);
3626 if (!lancer_chip(adapter))
3627 be_cmd_req_native_mode(adapter);
3629 status = be_get_config(adapter);
3633 status = be_msix_enable(adapter);
3637 en_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3638 BE_IF_FLAGS_MULTICAST | BE_IF_FLAGS_PASS_L3L4_ERRORS;
3639 if (adapter->function_caps & BE_FUNCTION_CAPS_RSS)
3640 en_flags |= BE_IF_FLAGS_RSS;
3641 en_flags = en_flags & be_if_cap_flags(adapter);
3642 status = be_cmd_if_create(adapter, be_if_cap_flags(adapter), en_flags,
3643 &adapter->if_handle, 0);
3647 /* Updating real_num_tx/rx_queues() requires rtnl_lock() */
3649 status = be_setup_queues(adapter);
3654 be_cmd_get_fn_privileges(adapter, &adapter->cmd_privileges, 0);
3656 status = be_mac_setup(adapter);
3660 be_cmd_get_fw_ver(adapter);
3661 dev_info(dev, "FW version is %s\n", adapter->fw_ver);
3663 if (BE2_chip(adapter) && fw_major_num(adapter->fw_ver) < 4) {
3664 dev_err(dev, "Firmware on card is old(%s), IRQs may not work.",
3666 dev_err(dev, "Please upgrade firmware to version >= 4.0\n");
3669 if (adapter->vlans_added)
3670 be_vid_config(adapter);
3672 be_set_rx_mode(adapter->netdev);
3674 be_cmd_get_acpi_wol_cap(adapter);
3676 be_cmd_get_flow_control(adapter, &tx_fc, &rx_fc);
3678 if (rx_fc != adapter->rx_fc || tx_fc != adapter->tx_fc)
3679 be_cmd_set_flow_control(adapter, adapter->tx_fc,
3682 if (be_physfn(adapter))
3683 be_cmd_set_logical_link_config(adapter,
3684 IFLA_VF_LINK_STATE_AUTO, 0);
3686 if (adapter->num_vfs)
3687 be_vf_setup(adapter);
3689 status = be_cmd_get_phy_info(adapter);
3690 if (!status && be_pause_supported(adapter))
3691 adapter->phy.fc_autoneg = 1;
3693 be_schedule_worker(adapter);
3694 adapter->flags |= BE_FLAGS_SETUP_DONE;
3701 #ifdef CONFIG_NET_POLL_CONTROLLER
3702 static void be_netpoll(struct net_device *netdev)
3704 struct be_adapter *adapter = netdev_priv(netdev);
3705 struct be_eq_obj *eqo;
3708 for_all_evt_queues(adapter, eqo, i) {
3709 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
3710 napi_schedule(&eqo->napi);
3717 static char flash_cookie[2][16] = {"*** SE FLAS", "H DIRECTORY *** "};
3719 static bool phy_flashing_required(struct be_adapter *adapter)
3721 return (adapter->phy.phy_type == TN_8022 &&
3722 adapter->phy.interface_type == PHY_TYPE_BASET_10GB);
3725 static bool is_comp_in_ufi(struct be_adapter *adapter,
3726 struct flash_section_info *fsec, int type)
3728 int i = 0, img_type = 0;
3729 struct flash_section_info_g2 *fsec_g2 = NULL;
3731 if (BE2_chip(adapter))
3732 fsec_g2 = (struct flash_section_info_g2 *)fsec;
3734 for (i = 0; i < MAX_FLASH_COMP; i++) {
3736 img_type = le32_to_cpu(fsec_g2->fsec_entry[i].type);
3738 img_type = le32_to_cpu(fsec->fsec_entry[i].type);
3740 if (img_type == type)
3747 static struct flash_section_info *get_fsec_info(struct be_adapter *adapter,
3749 const struct firmware *fw)
3751 struct flash_section_info *fsec = NULL;
3752 const u8 *p = fw->data;
3755 while (p < (fw->data + fw->size)) {
3756 fsec = (struct flash_section_info *)p;
3757 if (!memcmp(flash_cookie, fsec->cookie, sizeof(flash_cookie)))
3764 static int be_check_flash_crc(struct be_adapter *adapter, const u8 *p,
3765 u32 img_offset, u32 img_size, int hdr_size,
3766 u16 img_optype, bool *crc_match)
3772 status = be_cmd_get_flash_crc(adapter, crc, img_optype, img_size - 4);
3776 crc_offset = hdr_size + img_offset + img_size - 4;
3778 /* Skip flashing, if crc of flashed region matches */
3779 if (!memcmp(crc, p + crc_offset, 4))
3787 static int be_flash(struct be_adapter *adapter, const u8 *img,
3788 struct be_dma_mem *flash_cmd, int optype, int img_size)
3790 struct be_cmd_write_flashrom *req = flash_cmd->va;
3791 u32 total_bytes, flash_op, num_bytes;
3794 total_bytes = img_size;
3795 while (total_bytes) {
3796 num_bytes = min_t(u32, 32*1024, total_bytes);
3798 total_bytes -= num_bytes;
3801 if (optype == OPTYPE_PHY_FW)
3802 flash_op = FLASHROM_OPER_PHY_FLASH;
3804 flash_op = FLASHROM_OPER_FLASH;
3806 if (optype == OPTYPE_PHY_FW)
3807 flash_op = FLASHROM_OPER_PHY_SAVE;
3809 flash_op = FLASHROM_OPER_SAVE;
3812 memcpy(req->data_buf, img, num_bytes);
3814 status = be_cmd_write_flashrom(adapter, flash_cmd, optype,
3815 flash_op, num_bytes);
3816 if (base_status(status) == MCC_STATUS_ILLEGAL_REQUEST &&
3817 optype == OPTYPE_PHY_FW)
3825 /* For BE2, BE3 and BE3-R */
3826 static int be_flash_BEx(struct be_adapter *adapter,
3827 const struct firmware *fw,
3828 struct be_dma_mem *flash_cmd, int num_of_images)
3830 int img_hdrs_size = (num_of_images * sizeof(struct image_hdr));
3831 struct device *dev = &adapter->pdev->dev;
3832 struct flash_section_info *fsec = NULL;
3833 int status, i, filehdr_size, num_comp;
3834 const struct flash_comp *pflashcomp;
3838 struct flash_comp gen3_flash_types[] = {
3839 { FLASH_iSCSI_PRIMARY_IMAGE_START_g3, OPTYPE_ISCSI_ACTIVE,
3840 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_iSCSI},
3841 { FLASH_REDBOOT_START_g3, OPTYPE_REDBOOT,
3842 FLASH_REDBOOT_IMAGE_MAX_SIZE_g3, IMAGE_BOOT_CODE},
3843 { FLASH_iSCSI_BIOS_START_g3, OPTYPE_BIOS,
3844 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_ISCSI},
3845 { FLASH_PXE_BIOS_START_g3, OPTYPE_PXE_BIOS,
3846 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_PXE},
3847 { FLASH_FCoE_BIOS_START_g3, OPTYPE_FCOE_BIOS,
3848 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_FCoE},
3849 { FLASH_iSCSI_BACKUP_IMAGE_START_g3, OPTYPE_ISCSI_BACKUP,
3850 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_iSCSI},
3851 { FLASH_FCoE_PRIMARY_IMAGE_START_g3, OPTYPE_FCOE_FW_ACTIVE,
3852 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_FCoE},
3853 { FLASH_FCoE_BACKUP_IMAGE_START_g3, OPTYPE_FCOE_FW_BACKUP,
3854 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_FCoE},
3855 { FLASH_NCSI_START_g3, OPTYPE_NCSI_FW,
3856 FLASH_NCSI_IMAGE_MAX_SIZE_g3, IMAGE_NCSI},
3857 { FLASH_PHY_FW_START_g3, OPTYPE_PHY_FW,
3858 FLASH_PHY_FW_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_PHY}
3861 struct flash_comp gen2_flash_types[] = {
3862 { FLASH_iSCSI_PRIMARY_IMAGE_START_g2, OPTYPE_ISCSI_ACTIVE,
3863 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_iSCSI},
3864 { FLASH_REDBOOT_START_g2, OPTYPE_REDBOOT,
3865 FLASH_REDBOOT_IMAGE_MAX_SIZE_g2, IMAGE_BOOT_CODE},
3866 { FLASH_iSCSI_BIOS_START_g2, OPTYPE_BIOS,
3867 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_ISCSI},
3868 { FLASH_PXE_BIOS_START_g2, OPTYPE_PXE_BIOS,
3869 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_PXE},
3870 { FLASH_FCoE_BIOS_START_g2, OPTYPE_FCOE_BIOS,
3871 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_FCoE},
3872 { FLASH_iSCSI_BACKUP_IMAGE_START_g2, OPTYPE_ISCSI_BACKUP,
3873 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_iSCSI},
3874 { FLASH_FCoE_PRIMARY_IMAGE_START_g2, OPTYPE_FCOE_FW_ACTIVE,
3875 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_FCoE},
3876 { FLASH_FCoE_BACKUP_IMAGE_START_g2, OPTYPE_FCOE_FW_BACKUP,
3877 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_FCoE}
3880 if (BE3_chip(adapter)) {
3881 pflashcomp = gen3_flash_types;
3882 filehdr_size = sizeof(struct flash_file_hdr_g3);
3883 num_comp = ARRAY_SIZE(gen3_flash_types);
3885 pflashcomp = gen2_flash_types;
3886 filehdr_size = sizeof(struct flash_file_hdr_g2);
3887 num_comp = ARRAY_SIZE(gen2_flash_types);
3890 /* Get flash section info*/
3891 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
3893 dev_err(dev, "Invalid Cookie. FW image may be corrupted\n");
3896 for (i = 0; i < num_comp; i++) {
3897 if (!is_comp_in_ufi(adapter, fsec, pflashcomp[i].img_type))
3900 if ((pflashcomp[i].optype == OPTYPE_NCSI_FW) &&
3901 memcmp(adapter->fw_ver, "3.102.148.0", 11) < 0)
3904 if (pflashcomp[i].optype == OPTYPE_PHY_FW &&
3905 !phy_flashing_required(adapter))
3908 if (pflashcomp[i].optype == OPTYPE_REDBOOT) {
3909 status = be_check_flash_crc(adapter, fw->data,
3910 pflashcomp[i].offset,
3914 OPTYPE_REDBOOT, &crc_match);
3917 "Could not get CRC for 0x%x region\n",
3918 pflashcomp[i].optype);
3926 p = fw->data + filehdr_size + pflashcomp[i].offset +
3928 if (p + pflashcomp[i].size > fw->data + fw->size)
3931 status = be_flash(adapter, p, flash_cmd, pflashcomp[i].optype,
3932 pflashcomp[i].size);
3934 dev_err(dev, "Flashing section type 0x%x failed\n",
3935 pflashcomp[i].img_type);
3942 static u16 be_get_img_optype(struct flash_section_entry fsec_entry)
3944 u32 img_type = le32_to_cpu(fsec_entry.type);
3945 u16 img_optype = le16_to_cpu(fsec_entry.optype);
3947 if (img_optype != 0xFFFF)
3951 case IMAGE_FIRMWARE_iSCSI:
3952 img_optype = OPTYPE_ISCSI_ACTIVE;
3954 case IMAGE_BOOT_CODE:
3955 img_optype = OPTYPE_REDBOOT;
3957 case IMAGE_OPTION_ROM_ISCSI:
3958 img_optype = OPTYPE_BIOS;
3960 case IMAGE_OPTION_ROM_PXE:
3961 img_optype = OPTYPE_PXE_BIOS;
3963 case IMAGE_OPTION_ROM_FCoE:
3964 img_optype = OPTYPE_FCOE_BIOS;
3966 case IMAGE_FIRMWARE_BACKUP_iSCSI:
3967 img_optype = OPTYPE_ISCSI_BACKUP;
3970 img_optype = OPTYPE_NCSI_FW;
3972 case IMAGE_FLASHISM_JUMPVECTOR:
3973 img_optype = OPTYPE_FLASHISM_JUMPVECTOR;
3975 case IMAGE_FIRMWARE_PHY:
3976 img_optype = OPTYPE_SH_PHY_FW;
3978 case IMAGE_REDBOOT_DIR:
3979 img_optype = OPTYPE_REDBOOT_DIR;
3981 case IMAGE_REDBOOT_CONFIG:
3982 img_optype = OPTYPE_REDBOOT_CONFIG;
3985 img_optype = OPTYPE_UFI_DIR;
3994 static int be_flash_skyhawk(struct be_adapter *adapter,
3995 const struct firmware *fw,
3996 struct be_dma_mem *flash_cmd, int num_of_images)
3998 int img_hdrs_size = num_of_images * sizeof(struct image_hdr);
3999 struct device *dev = &adapter->pdev->dev;
4000 struct flash_section_info *fsec = NULL;
4001 u32 img_offset, img_size, img_type;
4002 int status, i, filehdr_size;
4003 bool crc_match, old_fw_img;
4007 filehdr_size = sizeof(struct flash_file_hdr_g3);
4008 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
4010 dev_err(dev, "Invalid Cookie. FW image may be corrupted\n");
4014 for (i = 0; i < le32_to_cpu(fsec->fsec_hdr.num_images); i++) {
4015 img_offset = le32_to_cpu(fsec->fsec_entry[i].offset);
4016 img_size = le32_to_cpu(fsec->fsec_entry[i].pad_size);
4017 img_type = le32_to_cpu(fsec->fsec_entry[i].type);
4018 img_optype = be_get_img_optype(fsec->fsec_entry[i]);
4019 old_fw_img = fsec->fsec_entry[i].optype == 0xFFFF;
4021 if (img_optype == 0xFFFF)
4023 /* Don't bother verifying CRC if an old FW image is being
4029 status = be_check_flash_crc(adapter, fw->data, img_offset,
4030 img_size, filehdr_size +
4031 img_hdrs_size, img_optype,
4033 /* The current FW image on the card does not recognize the new
4034 * FLASH op_type. The FW download is partially complete.
4035 * Reboot the server now to enable FW image to recognize the
4036 * new FLASH op_type. To complete the remaining process,
4037 * download the same FW again after the reboot.
4039 if (base_status(status) == MCC_STATUS_ILLEGAL_REQUEST ||
4040 base_status(status) == MCC_STATUS_ILLEGAL_FIELD) {
4041 dev_err(dev, "Flash incomplete. Reset the server\n");
4042 dev_err(dev, "Download FW image again after reset\n");
4044 } else if (status) {
4045 dev_err(dev, "Could not get CRC for 0x%x region\n",
4054 p = fw->data + filehdr_size + img_offset + img_hdrs_size;
4055 if (p + img_size > fw->data + fw->size)
4058 status = be_flash(adapter, p, flash_cmd, img_optype, img_size);
4059 /* For old FW images ignore ILLEGAL_FIELD error or errors on
4063 (base_status(status) == MCC_STATUS_ILLEGAL_FIELD ||
4064 (img_optype == OPTYPE_UFI_DIR &&
4065 base_status(status) == MCC_STATUS_FAILED))) {
4067 } else if (status) {
4068 dev_err(dev, "Flashing section type 0x%x failed\n",
4076 static int lancer_fw_download(struct be_adapter *adapter,
4077 const struct firmware *fw)
4079 #define LANCER_FW_DOWNLOAD_CHUNK (32 * 1024)
4080 #define LANCER_FW_DOWNLOAD_LOCATION "/prg"
4081 struct device *dev = &adapter->pdev->dev;
4082 struct be_dma_mem flash_cmd;
4083 const u8 *data_ptr = NULL;
4084 u8 *dest_image_ptr = NULL;
4085 size_t image_size = 0;
4087 u32 data_written = 0;
4093 if (!IS_ALIGNED(fw->size, sizeof(u32))) {
4094 dev_err(dev, "FW image size should be multiple of 4\n");
4098 flash_cmd.size = sizeof(struct lancer_cmd_req_write_object)
4099 + LANCER_FW_DOWNLOAD_CHUNK;
4100 flash_cmd.va = dma_alloc_coherent(dev, flash_cmd.size,
4101 &flash_cmd.dma, GFP_KERNEL);
4105 dest_image_ptr = flash_cmd.va +
4106 sizeof(struct lancer_cmd_req_write_object);
4107 image_size = fw->size;
4108 data_ptr = fw->data;
4110 while (image_size) {
4111 chunk_size = min_t(u32, image_size, LANCER_FW_DOWNLOAD_CHUNK);
4113 /* Copy the image chunk content. */
4114 memcpy(dest_image_ptr, data_ptr, chunk_size);
4116 status = lancer_cmd_write_object(adapter, &flash_cmd,
4118 LANCER_FW_DOWNLOAD_LOCATION,
4119 &data_written, &change_status,
4124 offset += data_written;
4125 data_ptr += data_written;
4126 image_size -= data_written;
4130 /* Commit the FW written */
4131 status = lancer_cmd_write_object(adapter, &flash_cmd,
4133 LANCER_FW_DOWNLOAD_LOCATION,
4134 &data_written, &change_status,
4138 dma_free_coherent(dev, flash_cmd.size, flash_cmd.va, flash_cmd.dma);
4140 dev_err(dev, "Firmware load error\n");
4141 return be_cmd_status(status);
4144 dev_info(dev, "Firmware flashed successfully\n");
4146 if (change_status == LANCER_FW_RESET_NEEDED) {
4147 dev_info(dev, "Resetting adapter to activate new FW\n");
4148 status = lancer_physdev_ctrl(adapter,
4149 PHYSDEV_CONTROL_FW_RESET_MASK);
4151 dev_err(dev, "Adapter busy, could not reset FW\n");
4152 dev_err(dev, "Reboot server to activate new FW\n");
4154 } else if (change_status != LANCER_NO_RESET_NEEDED) {
4155 dev_info(dev, "Reboot server to activate new FW\n");
4163 #define UFI_TYPE3R 10
4165 static int be_get_ufi_type(struct be_adapter *adapter,
4166 struct flash_file_hdr_g3 *fhdr)
4169 goto be_get_ufi_exit;
4171 if (skyhawk_chip(adapter) && fhdr->build[0] == '4')
4173 else if (BE3_chip(adapter) && fhdr->build[0] == '3') {
4174 if (fhdr->asic_type_rev == 0x10)
4178 } else if (BE2_chip(adapter) && fhdr->build[0] == '2')
4182 dev_err(&adapter->pdev->dev,
4183 "UFI and Interface are not compatible for flashing\n");
4187 static int be_fw_download(struct be_adapter *adapter, const struct firmware* fw)
4189 struct flash_file_hdr_g3 *fhdr3;
4190 struct image_hdr *img_hdr_ptr = NULL;
4191 struct be_dma_mem flash_cmd;
4193 int status = 0, i = 0, num_imgs = 0, ufi_type = 0;
4195 flash_cmd.size = sizeof(struct be_cmd_write_flashrom);
4196 flash_cmd.va = dma_alloc_coherent(&adapter->pdev->dev, flash_cmd.size,
4197 &flash_cmd.dma, GFP_KERNEL);
4198 if (!flash_cmd.va) {
4204 fhdr3 = (struct flash_file_hdr_g3 *)p;
4206 ufi_type = be_get_ufi_type(adapter, fhdr3);
4208 num_imgs = le32_to_cpu(fhdr3->num_imgs);
4209 for (i = 0; i < num_imgs; i++) {
4210 img_hdr_ptr = (struct image_hdr *)(fw->data +
4211 (sizeof(struct flash_file_hdr_g3) +
4212 i * sizeof(struct image_hdr)));
4213 if (le32_to_cpu(img_hdr_ptr->imageid) == 1) {
4216 status = be_flash_skyhawk(adapter, fw,
4217 &flash_cmd, num_imgs);
4220 status = be_flash_BEx(adapter, fw, &flash_cmd,
4224 /* Do not flash this ufi on BE3-R cards */
4225 if (adapter->asic_rev < 0x10)
4226 status = be_flash_BEx(adapter, fw,
4231 dev_err(&adapter->pdev->dev,
4232 "Can't load BE3 UFI on BE3R\n");
4238 if (ufi_type == UFI_TYPE2)
4239 status = be_flash_BEx(adapter, fw, &flash_cmd, 0);
4240 else if (ufi_type == -1)
4243 dma_free_coherent(&adapter->pdev->dev, flash_cmd.size, flash_cmd.va,
4246 dev_err(&adapter->pdev->dev, "Firmware load error\n");
4250 dev_info(&adapter->pdev->dev, "Firmware flashed successfully\n");
4256 int be_load_fw(struct be_adapter *adapter, u8 *fw_file)
4258 const struct firmware *fw;
4261 if (!netif_running(adapter->netdev)) {
4262 dev_err(&adapter->pdev->dev,
4263 "Firmware load not allowed (interface is down)\n");
4267 status = request_firmware(&fw, fw_file, &adapter->pdev->dev);
4271 dev_info(&adapter->pdev->dev, "Flashing firmware file %s\n", fw_file);
4273 if (lancer_chip(adapter))
4274 status = lancer_fw_download(adapter, fw);
4276 status = be_fw_download(adapter, fw);
4279 be_cmd_get_fw_ver(adapter);
4282 release_firmware(fw);
4286 static int be_ndo_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh)
4288 struct be_adapter *adapter = netdev_priv(dev);
4289 struct nlattr *attr, *br_spec;
4294 if (!sriov_enabled(adapter))
4297 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4299 nla_for_each_nested(attr, br_spec, rem) {
4300 if (nla_type(attr) != IFLA_BRIDGE_MODE)
4303 mode = nla_get_u16(attr);
4304 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
4307 status = be_cmd_set_hsw_config(adapter, 0, 0,
4309 mode == BRIDGE_MODE_VEPA ?
4310 PORT_FWD_TYPE_VEPA :
4315 dev_info(&adapter->pdev->dev, "enabled switch mode: %s\n",
4316 mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
4321 dev_err(&adapter->pdev->dev, "Failed to set switch mode %s\n",
4322 mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
4327 static int be_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4328 struct net_device *dev, u32 filter_mask)
4330 struct be_adapter *adapter = netdev_priv(dev);
4334 if (!sriov_enabled(adapter))
4337 /* BE and Lancer chips support VEB mode only */
4338 if (BEx_chip(adapter) || lancer_chip(adapter)) {
4339 hsw_mode = PORT_FWD_TYPE_VEB;
4341 status = be_cmd_get_hsw_config(adapter, NULL, 0,
4342 adapter->if_handle, &hsw_mode);
4347 return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
4348 hsw_mode == PORT_FWD_TYPE_VEPA ?
4349 BRIDGE_MODE_VEPA : BRIDGE_MODE_VEB);
4352 #ifdef CONFIG_BE2NET_VXLAN
4353 static void be_add_vxlan_port(struct net_device *netdev, sa_family_t sa_family,
4356 struct be_adapter *adapter = netdev_priv(netdev);
4357 struct device *dev = &adapter->pdev->dev;
4360 if (lancer_chip(adapter) || BEx_chip(adapter))
4363 if (adapter->flags & BE_FLAGS_VXLAN_OFFLOADS) {
4364 dev_warn(dev, "Cannot add UDP port %d for VxLAN offloads\n",
4367 "Only one UDP port supported for VxLAN offloads\n");
4371 status = be_cmd_manage_iface(adapter, adapter->if_handle,
4372 OP_CONVERT_NORMAL_TO_TUNNEL);
4374 dev_warn(dev, "Failed to convert normal interface to tunnel\n");
4378 status = be_cmd_set_vxlan_port(adapter, port);
4380 dev_warn(dev, "Failed to add VxLAN port\n");
4383 adapter->flags |= BE_FLAGS_VXLAN_OFFLOADS;
4384 adapter->vxlan_port = port;
4386 dev_info(dev, "Enabled VxLAN offloads for UDP port %d\n",
4390 be_disable_vxlan_offloads(adapter);
4394 static void be_del_vxlan_port(struct net_device *netdev, sa_family_t sa_family,
4397 struct be_adapter *adapter = netdev_priv(netdev);
4399 if (lancer_chip(adapter) || BEx_chip(adapter))
4402 if (adapter->vxlan_port != port)
4405 be_disable_vxlan_offloads(adapter);
4407 dev_info(&adapter->pdev->dev,
4408 "Disabled VxLAN offloads for UDP port %d\n",
4413 static const struct net_device_ops be_netdev_ops = {
4414 .ndo_open = be_open,
4415 .ndo_stop = be_close,
4416 .ndo_start_xmit = be_xmit,
4417 .ndo_set_rx_mode = be_set_rx_mode,
4418 .ndo_set_mac_address = be_mac_addr_set,
4419 .ndo_change_mtu = be_change_mtu,
4420 .ndo_get_stats64 = be_get_stats64,
4421 .ndo_validate_addr = eth_validate_addr,
4422 .ndo_vlan_rx_add_vid = be_vlan_add_vid,
4423 .ndo_vlan_rx_kill_vid = be_vlan_rem_vid,
4424 .ndo_set_vf_mac = be_set_vf_mac,
4425 .ndo_set_vf_vlan = be_set_vf_vlan,
4426 .ndo_set_vf_rate = be_set_vf_tx_rate,
4427 .ndo_get_vf_config = be_get_vf_config,
4428 .ndo_set_vf_link_state = be_set_vf_link_state,
4429 #ifdef CONFIG_NET_POLL_CONTROLLER
4430 .ndo_poll_controller = be_netpoll,
4432 .ndo_bridge_setlink = be_ndo_bridge_setlink,
4433 .ndo_bridge_getlink = be_ndo_bridge_getlink,
4434 #ifdef CONFIG_NET_RX_BUSY_POLL
4435 .ndo_busy_poll = be_busy_poll,
4437 #ifdef CONFIG_BE2NET_VXLAN
4438 .ndo_add_vxlan_port = be_add_vxlan_port,
4439 .ndo_del_vxlan_port = be_del_vxlan_port,
4443 static void be_netdev_init(struct net_device *netdev)
4445 struct be_adapter *adapter = netdev_priv(netdev);
4447 if (skyhawk_chip(adapter)) {
4448 netdev->hw_enc_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
4449 NETIF_F_TSO | NETIF_F_TSO6 |
4450 NETIF_F_GSO_UDP_TUNNEL;
4451 netdev->hw_features |= NETIF_F_GSO_UDP_TUNNEL;
4453 netdev->hw_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
4454 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM | NETIF_F_RXCSUM |
4455 NETIF_F_HW_VLAN_CTAG_TX;
4456 if (be_multi_rxq(adapter))
4457 netdev->hw_features |= NETIF_F_RXHASH;
4459 netdev->features |= netdev->hw_features |
4460 NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_FILTER;
4462 netdev->vlan_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
4463 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
4465 netdev->priv_flags |= IFF_UNICAST_FLT;
4467 netdev->flags |= IFF_MULTICAST;
4469 netif_set_gso_max_size(netdev, 65535 - ETH_HLEN);
4471 netdev->netdev_ops = &be_netdev_ops;
4473 netdev->ethtool_ops = &be_ethtool_ops;
4476 static void be_unmap_pci_bars(struct be_adapter *adapter)
4479 pci_iounmap(adapter->pdev, adapter->csr);
4481 pci_iounmap(adapter->pdev, adapter->db);
4484 static int db_bar(struct be_adapter *adapter)
4486 if (lancer_chip(adapter) || !be_physfn(adapter))
4492 static int be_roce_map_pci_bars(struct be_adapter *adapter)
4494 if (skyhawk_chip(adapter)) {
4495 adapter->roce_db.size = 4096;
4496 adapter->roce_db.io_addr = pci_resource_start(adapter->pdev,
4498 adapter->roce_db.total_size = pci_resource_len(adapter->pdev,
4504 static int be_map_pci_bars(struct be_adapter *adapter)
4508 if (BEx_chip(adapter) && be_physfn(adapter)) {
4509 adapter->csr = pci_iomap(adapter->pdev, 2, 0);
4514 addr = pci_iomap(adapter->pdev, db_bar(adapter), 0);
4519 be_roce_map_pci_bars(adapter);
4523 dev_err(&adapter->pdev->dev, "Error in mapping PCI BARs\n");
4524 be_unmap_pci_bars(adapter);
4528 static void be_ctrl_cleanup(struct be_adapter *adapter)
4530 struct be_dma_mem *mem = &adapter->mbox_mem_alloced;
4532 be_unmap_pci_bars(adapter);
4535 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
4538 mem = &adapter->rx_filter;
4540 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
4544 static int be_ctrl_init(struct be_adapter *adapter)
4546 struct be_dma_mem *mbox_mem_alloc = &adapter->mbox_mem_alloced;
4547 struct be_dma_mem *mbox_mem_align = &adapter->mbox_mem;
4548 struct be_dma_mem *rx_filter = &adapter->rx_filter;
4552 pci_read_config_dword(adapter->pdev, SLI_INTF_REG_OFFSET, &sli_intf);
4553 adapter->sli_family = (sli_intf & SLI_INTF_FAMILY_MASK) >>
4554 SLI_INTF_FAMILY_SHIFT;
4555 adapter->virtfn = (sli_intf & SLI_INTF_FT_MASK) ? 1 : 0;
4557 status = be_map_pci_bars(adapter);
4561 mbox_mem_alloc->size = sizeof(struct be_mcc_mailbox) + 16;
4562 mbox_mem_alloc->va = dma_alloc_coherent(&adapter->pdev->dev,
4563 mbox_mem_alloc->size,
4564 &mbox_mem_alloc->dma,
4566 if (!mbox_mem_alloc->va) {
4568 goto unmap_pci_bars;
4570 mbox_mem_align->size = sizeof(struct be_mcc_mailbox);
4571 mbox_mem_align->va = PTR_ALIGN(mbox_mem_alloc->va, 16);
4572 mbox_mem_align->dma = PTR_ALIGN(mbox_mem_alloc->dma, 16);
4573 memset(mbox_mem_align->va, 0, sizeof(struct be_mcc_mailbox));
4575 rx_filter->size = sizeof(struct be_cmd_req_rx_filter);
4576 rx_filter->va = dma_zalloc_coherent(&adapter->pdev->dev,
4577 rx_filter->size, &rx_filter->dma,
4579 if (!rx_filter->va) {
4584 mutex_init(&adapter->mbox_lock);
4585 spin_lock_init(&adapter->mcc_lock);
4586 spin_lock_init(&adapter->mcc_cq_lock);
4588 init_completion(&adapter->et_cmd_compl);
4589 pci_save_state(adapter->pdev);
4593 dma_free_coherent(&adapter->pdev->dev, mbox_mem_alloc->size,
4594 mbox_mem_alloc->va, mbox_mem_alloc->dma);
4597 be_unmap_pci_bars(adapter);
4603 static void be_stats_cleanup(struct be_adapter *adapter)
4605 struct be_dma_mem *cmd = &adapter->stats_cmd;
4608 dma_free_coherent(&adapter->pdev->dev, cmd->size,
4612 static int be_stats_init(struct be_adapter *adapter)
4614 struct be_dma_mem *cmd = &adapter->stats_cmd;
4616 if (lancer_chip(adapter))
4617 cmd->size = sizeof(struct lancer_cmd_req_pport_stats);
4618 else if (BE2_chip(adapter))
4619 cmd->size = sizeof(struct be_cmd_req_get_stats_v0);
4620 else if (BE3_chip(adapter))
4621 cmd->size = sizeof(struct be_cmd_req_get_stats_v1);
4623 /* ALL non-BE ASICs */
4624 cmd->size = sizeof(struct be_cmd_req_get_stats_v2);
4626 cmd->va = dma_zalloc_coherent(&adapter->pdev->dev, cmd->size, &cmd->dma,
4633 static void be_remove(struct pci_dev *pdev)
4635 struct be_adapter *adapter = pci_get_drvdata(pdev);
4640 be_roce_dev_remove(adapter);
4641 be_intr_set(adapter, false);
4643 cancel_delayed_work_sync(&adapter->func_recovery_work);
4645 unregister_netdev(adapter->netdev);
4649 /* tell fw we're done with firing cmds */
4650 be_cmd_fw_clean(adapter);
4652 be_stats_cleanup(adapter);
4654 be_ctrl_cleanup(adapter);
4656 pci_disable_pcie_error_reporting(pdev);
4658 pci_release_regions(pdev);
4659 pci_disable_device(pdev);
4661 free_netdev(adapter->netdev);
4664 static int be_get_initial_config(struct be_adapter *adapter)
4668 status = be_cmd_get_cntl_attributes(adapter);
4672 /* Must be a power of 2 or else MODULO will BUG_ON */
4673 adapter->be_get_temp_freq = 64;
4675 if (BEx_chip(adapter)) {
4676 level = be_cmd_get_fw_log_level(adapter);
4677 adapter->msg_enable =
4678 level <= FW_LOG_LEVEL_DEFAULT ? NETIF_MSG_HW : 0;
4681 adapter->cfg_num_qs = netif_get_num_default_rss_queues();
4685 static int lancer_recover_func(struct be_adapter *adapter)
4687 struct device *dev = &adapter->pdev->dev;
4690 status = lancer_test_and_set_rdy_state(adapter);
4694 if (netif_running(adapter->netdev))
4695 be_close(adapter->netdev);
4699 be_clear_all_error(adapter);
4701 status = be_setup(adapter);
4705 if (netif_running(adapter->netdev)) {
4706 status = be_open(adapter->netdev);
4711 dev_err(dev, "Adapter recovery successful\n");
4714 if (status == -EAGAIN)
4715 dev_err(dev, "Waiting for resource provisioning\n");
4717 dev_err(dev, "Adapter recovery failed\n");
4722 static void be_func_recovery_task(struct work_struct *work)
4724 struct be_adapter *adapter =
4725 container_of(work, struct be_adapter, func_recovery_work.work);
4728 be_detect_error(adapter);
4730 if (adapter->hw_error && lancer_chip(adapter)) {
4733 netif_device_detach(adapter->netdev);
4736 status = lancer_recover_func(adapter);
4738 netif_device_attach(adapter->netdev);
4741 /* In Lancer, for all errors other than provisioning error (-EAGAIN),
4742 * no need to attempt further recovery.
4744 if (!status || status == -EAGAIN)
4745 schedule_delayed_work(&adapter->func_recovery_work,
4746 msecs_to_jiffies(1000));
4749 static void be_worker(struct work_struct *work)
4751 struct be_adapter *adapter =
4752 container_of(work, struct be_adapter, work.work);
4753 struct be_rx_obj *rxo;
4756 /* when interrupts are not yet enabled, just reap any pending
4757 * mcc completions */
4758 if (!netif_running(adapter->netdev)) {
4760 be_process_mcc(adapter);
4765 if (!adapter->stats_cmd_sent) {
4766 if (lancer_chip(adapter))
4767 lancer_cmd_get_pport_stats(adapter,
4768 &adapter->stats_cmd);
4770 be_cmd_get_stats(adapter, &adapter->stats_cmd);
4773 if (be_physfn(adapter) &&
4774 MODULO(adapter->work_counter, adapter->be_get_temp_freq) == 0)
4775 be_cmd_get_die_temperature(adapter);
4777 for_all_rx_queues(adapter, rxo, i) {
4778 /* Replenish RX-queues starved due to memory
4779 * allocation failures.
4781 if (rxo->rx_post_starved)
4782 be_post_rx_frags(rxo, GFP_KERNEL);
4785 be_eqd_update(adapter);
4788 adapter->work_counter++;
4789 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
4792 /* If any VFs are already enabled don't FLR the PF */
4793 static bool be_reset_required(struct be_adapter *adapter)
4795 return pci_num_vf(adapter->pdev) ? false : true;
4798 static char *mc_name(struct be_adapter *adapter)
4800 char *str = ""; /* default */
4802 switch (adapter->mc_type) {
4828 static inline char *func_name(struct be_adapter *adapter)
4830 return be_physfn(adapter) ? "PF" : "VF";
4833 static int be_probe(struct pci_dev *pdev, const struct pci_device_id *pdev_id)
4836 struct be_adapter *adapter;
4837 struct net_device *netdev;
4840 dev_info(&pdev->dev, "%s version is %s\n", DRV_NAME, DRV_VER);
4842 status = pci_enable_device(pdev);
4846 status = pci_request_regions(pdev, DRV_NAME);
4849 pci_set_master(pdev);
4851 netdev = alloc_etherdev_mqs(sizeof(*adapter), MAX_TX_QS, MAX_RX_QS);
4856 adapter = netdev_priv(netdev);
4857 adapter->pdev = pdev;
4858 pci_set_drvdata(pdev, adapter);
4859 adapter->netdev = netdev;
4860 SET_NETDEV_DEV(netdev, &pdev->dev);
4862 status = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
4864 netdev->features |= NETIF_F_HIGHDMA;
4866 status = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
4868 dev_err(&pdev->dev, "Could not set PCI DMA Mask\n");
4873 if (be_physfn(adapter)) {
4874 status = pci_enable_pcie_error_reporting(pdev);
4876 dev_info(&pdev->dev, "PCIe error reporting enabled\n");
4879 status = be_ctrl_init(adapter);
4883 /* sync up with fw's ready state */
4884 if (be_physfn(adapter)) {
4885 status = be_fw_wait_ready(adapter);
4890 if (be_reset_required(adapter)) {
4891 status = be_cmd_reset_function(adapter);
4895 /* Wait for interrupts to quiesce after an FLR */
4899 /* Allow interrupts for other ULPs running on NIC function */
4900 be_intr_set(adapter, true);
4902 /* tell fw we're ready to fire cmds */
4903 status = be_cmd_fw_init(adapter);
4907 status = be_stats_init(adapter);
4911 status = be_get_initial_config(adapter);
4915 INIT_DELAYED_WORK(&adapter->work, be_worker);
4916 INIT_DELAYED_WORK(&adapter->func_recovery_work, be_func_recovery_task);
4917 adapter->rx_fc = adapter->tx_fc = true;
4919 status = be_setup(adapter);
4923 be_netdev_init(netdev);
4924 status = register_netdev(netdev);
4928 be_roce_dev_add(adapter);
4930 schedule_delayed_work(&adapter->func_recovery_work,
4931 msecs_to_jiffies(1000));
4933 be_cmd_query_port_name(adapter, &port_name);
4935 dev_info(&pdev->dev, "%s: %s %s port %c\n", nic_name(pdev),
4936 func_name(adapter), mc_name(adapter), port_name);
4943 be_stats_cleanup(adapter);
4945 be_ctrl_cleanup(adapter);
4947 free_netdev(netdev);
4949 pci_release_regions(pdev);
4951 pci_disable_device(pdev);
4953 dev_err(&pdev->dev, "%s initialization failed\n", nic_name(pdev));
4957 static int be_suspend(struct pci_dev *pdev, pm_message_t state)
4959 struct be_adapter *adapter = pci_get_drvdata(pdev);
4960 struct net_device *netdev = adapter->netdev;
4962 if (adapter->wol_en)
4963 be_setup_wol(adapter, true);
4965 be_intr_set(adapter, false);
4966 cancel_delayed_work_sync(&adapter->func_recovery_work);
4968 netif_device_detach(netdev);
4969 if (netif_running(netdev)) {
4976 pci_save_state(pdev);
4977 pci_disable_device(pdev);
4978 pci_set_power_state(pdev, pci_choose_state(pdev, state));
4982 static int be_resume(struct pci_dev *pdev)
4985 struct be_adapter *adapter = pci_get_drvdata(pdev);
4986 struct net_device *netdev = adapter->netdev;
4988 netif_device_detach(netdev);
4990 status = pci_enable_device(pdev);
4994 pci_set_power_state(pdev, PCI_D0);
4995 pci_restore_state(pdev);
4997 status = be_fw_wait_ready(adapter);
5001 be_intr_set(adapter, true);
5002 /* tell fw we're ready to fire cmds */
5003 status = be_cmd_fw_init(adapter);
5008 if (netif_running(netdev)) {
5014 schedule_delayed_work(&adapter->func_recovery_work,
5015 msecs_to_jiffies(1000));
5016 netif_device_attach(netdev);
5018 if (adapter->wol_en)
5019 be_setup_wol(adapter, false);
5025 * An FLR will stop BE from DMAing any data.
5027 static void be_shutdown(struct pci_dev *pdev)
5029 struct be_adapter *adapter = pci_get_drvdata(pdev);
5034 be_roce_dev_shutdown(adapter);
5035 cancel_delayed_work_sync(&adapter->work);
5036 cancel_delayed_work_sync(&adapter->func_recovery_work);
5038 netif_device_detach(adapter->netdev);
5040 be_cmd_reset_function(adapter);
5042 pci_disable_device(pdev);
5045 static pci_ers_result_t be_eeh_err_detected(struct pci_dev *pdev,
5046 pci_channel_state_t state)
5048 struct be_adapter *adapter = pci_get_drvdata(pdev);
5049 struct net_device *netdev = adapter->netdev;
5051 dev_err(&adapter->pdev->dev, "EEH error detected\n");
5053 if (!adapter->eeh_error) {
5054 adapter->eeh_error = true;
5056 cancel_delayed_work_sync(&adapter->func_recovery_work);
5059 netif_device_detach(netdev);
5060 if (netif_running(netdev))
5067 if (state == pci_channel_io_perm_failure)
5068 return PCI_ERS_RESULT_DISCONNECT;
5070 pci_disable_device(pdev);
5072 /* The error could cause the FW to trigger a flash debug dump.
5073 * Resetting the card while flash dump is in progress
5074 * can cause it not to recover; wait for it to finish.
5075 * Wait only for first function as it is needed only once per
5078 if (pdev->devfn == 0)
5081 return PCI_ERS_RESULT_NEED_RESET;
5084 static pci_ers_result_t be_eeh_reset(struct pci_dev *pdev)
5086 struct be_adapter *adapter = pci_get_drvdata(pdev);
5089 dev_info(&adapter->pdev->dev, "EEH reset\n");
5091 status = pci_enable_device(pdev);
5093 return PCI_ERS_RESULT_DISCONNECT;
5095 pci_set_master(pdev);
5096 pci_set_power_state(pdev, PCI_D0);
5097 pci_restore_state(pdev);
5099 /* Check if card is ok and fw is ready */
5100 dev_info(&adapter->pdev->dev,
5101 "Waiting for FW to be ready after EEH reset\n");
5102 status = be_fw_wait_ready(adapter);
5104 return PCI_ERS_RESULT_DISCONNECT;
5106 pci_cleanup_aer_uncorrect_error_status(pdev);
5107 be_clear_all_error(adapter);
5108 return PCI_ERS_RESULT_RECOVERED;
5111 static void be_eeh_resume(struct pci_dev *pdev)
5114 struct be_adapter *adapter = pci_get_drvdata(pdev);
5115 struct net_device *netdev = adapter->netdev;
5117 dev_info(&adapter->pdev->dev, "EEH resume\n");
5119 pci_save_state(pdev);
5121 status = be_cmd_reset_function(adapter);
5125 /* On some BE3 FW versions, after a HW reset,
5126 * interrupts will remain disabled for each function.
5127 * So, explicitly enable interrupts
5129 be_intr_set(adapter, true);
5131 /* tell fw we're ready to fire cmds */
5132 status = be_cmd_fw_init(adapter);
5136 status = be_setup(adapter);
5140 if (netif_running(netdev)) {
5141 status = be_open(netdev);
5146 schedule_delayed_work(&adapter->func_recovery_work,
5147 msecs_to_jiffies(1000));
5148 netif_device_attach(netdev);
5151 dev_err(&adapter->pdev->dev, "EEH resume failed\n");
5154 static const struct pci_error_handlers be_eeh_handlers = {
5155 .error_detected = be_eeh_err_detected,
5156 .slot_reset = be_eeh_reset,
5157 .resume = be_eeh_resume,
5160 static struct pci_driver be_driver = {
5162 .id_table = be_dev_ids,
5164 .remove = be_remove,
5165 .suspend = be_suspend,
5166 .resume = be_resume,
5167 .shutdown = be_shutdown,
5168 .err_handler = &be_eeh_handlers
5171 static int __init be_init_module(void)
5173 if (rx_frag_size != 8192 && rx_frag_size != 4096 &&
5174 rx_frag_size != 2048) {
5175 printk(KERN_WARNING DRV_NAME
5176 " : Module param rx_frag_size must be 2048/4096/8192."
5178 rx_frag_size = 2048;
5181 return pci_register_driver(&be_driver);
5183 module_init(be_init_module);
5185 static void __exit be_exit_module(void)
5187 pci_unregister_driver(&be_driver);
5189 module_exit(be_exit_module);