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 struct device *dev = &adapter->pdev->dev;
1097 u16 vids[BE_NUM_VLANS_SUPPORTED];
1101 /* No need to further configure vids if in promiscuous mode */
1102 if (adapter->promiscuous)
1105 if (adapter->vlans_added > be_max_vlans(adapter))
1106 goto set_vlan_promisc;
1108 /* Construct VLAN Table to give to HW */
1109 for_each_set_bit(i, adapter->vids, VLAN_N_VID)
1110 vids[num++] = cpu_to_le16(i);
1112 status = be_cmd_vlan_config(adapter, adapter->if_handle, vids, num);
1114 /* Set to VLAN promisc mode as setting VLAN filter failed */
1115 if (addl_status(status) ==
1116 MCC_ADDL_STATUS_INSUFFICIENT_RESOURCES)
1117 goto set_vlan_promisc;
1118 dev_err(dev, "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);
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(dev, "Enable VLAN Promiscuous mode\n");
1141 adapter->flags |= BE_FLAGS_VLAN_PROMISC;
1143 dev_err(dev, "Failed to enable VLAN Promiscuous mode\n");
1147 static int be_vlan_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
1149 struct be_adapter *adapter = netdev_priv(netdev);
1152 /* Packets with VID 0 are always received by Lancer by default */
1153 if (lancer_chip(adapter) && vid == 0)
1156 if (test_bit(vid, adapter->vids))
1159 set_bit(vid, adapter->vids);
1160 adapter->vlans_added++;
1162 status = be_vid_config(adapter);
1164 adapter->vlans_added--;
1165 clear_bit(vid, adapter->vids);
1171 static int be_vlan_rem_vid(struct net_device *netdev, __be16 proto, u16 vid)
1173 struct be_adapter *adapter = netdev_priv(netdev);
1175 /* Packets with VID 0 are always received by Lancer by default */
1176 if (lancer_chip(adapter) && vid == 0)
1179 clear_bit(vid, adapter->vids);
1180 adapter->vlans_added--;
1182 return be_vid_config(adapter);
1185 static void be_clear_promisc(struct be_adapter *adapter)
1187 adapter->promiscuous = false;
1188 adapter->flags &= ~(BE_FLAGS_VLAN_PROMISC | BE_FLAGS_MCAST_PROMISC);
1190 be_cmd_rx_filter(adapter, IFF_PROMISC, OFF);
1193 static void be_set_rx_mode(struct net_device *netdev)
1195 struct be_adapter *adapter = netdev_priv(netdev);
1198 if (netdev->flags & IFF_PROMISC) {
1199 be_cmd_rx_filter(adapter, IFF_PROMISC, ON);
1200 adapter->promiscuous = true;
1204 /* BE was previously in promiscuous mode; disable it */
1205 if (adapter->promiscuous) {
1206 be_clear_promisc(adapter);
1207 if (adapter->vlans_added)
1208 be_vid_config(adapter);
1211 /* Enable multicast promisc if num configured exceeds what we support */
1212 if (netdev->flags & IFF_ALLMULTI ||
1213 netdev_mc_count(netdev) > be_max_mc(adapter))
1214 goto set_mcast_promisc;
1216 if (netdev_uc_count(netdev) != adapter->uc_macs) {
1217 struct netdev_hw_addr *ha;
1218 int i = 1; /* First slot is claimed by the Primary MAC */
1220 for (; adapter->uc_macs > 0; adapter->uc_macs--, i++) {
1221 be_cmd_pmac_del(adapter, adapter->if_handle,
1222 adapter->pmac_id[i], 0);
1225 if (netdev_uc_count(netdev) > be_max_uc(adapter)) {
1226 be_cmd_rx_filter(adapter, IFF_PROMISC, ON);
1227 adapter->promiscuous = true;
1231 netdev_for_each_uc_addr(ha, adapter->netdev) {
1232 adapter->uc_macs++; /* First slot is for Primary MAC */
1233 be_cmd_pmac_add(adapter, (u8 *)ha->addr,
1235 &adapter->pmac_id[adapter->uc_macs], 0);
1239 status = be_cmd_rx_filter(adapter, IFF_MULTICAST, ON);
1241 if (adapter->flags & BE_FLAGS_MCAST_PROMISC)
1242 adapter->flags &= ~BE_FLAGS_MCAST_PROMISC;
1247 if (adapter->flags & BE_FLAGS_MCAST_PROMISC)
1250 /* Set to MCAST promisc mode if setting MULTICAST address fails
1251 * or if num configured exceeds what we support
1253 status = be_cmd_rx_filter(adapter, IFF_ALLMULTI, ON);
1255 adapter->flags |= BE_FLAGS_MCAST_PROMISC;
1260 static int be_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
1262 struct be_adapter *adapter = netdev_priv(netdev);
1263 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1266 if (!sriov_enabled(adapter))
1269 if (!is_valid_ether_addr(mac) || vf >= adapter->num_vfs)
1272 /* Proceed further only if user provided MAC is different
1275 if (ether_addr_equal(mac, vf_cfg->mac_addr))
1278 if (BEx_chip(adapter)) {
1279 be_cmd_pmac_del(adapter, vf_cfg->if_handle, vf_cfg->pmac_id,
1282 status = be_cmd_pmac_add(adapter, mac, vf_cfg->if_handle,
1283 &vf_cfg->pmac_id, vf + 1);
1285 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
1290 dev_err(&adapter->pdev->dev, "MAC %pM set on VF %d Failed: %#x",
1292 return be_cmd_status(status);
1295 ether_addr_copy(vf_cfg->mac_addr, mac);
1300 static int be_get_vf_config(struct net_device *netdev, int vf,
1301 struct ifla_vf_info *vi)
1303 struct be_adapter *adapter = netdev_priv(netdev);
1304 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1306 if (!sriov_enabled(adapter))
1309 if (vf >= adapter->num_vfs)
1313 vi->max_tx_rate = vf_cfg->tx_rate;
1314 vi->min_tx_rate = 0;
1315 vi->vlan = vf_cfg->vlan_tag & VLAN_VID_MASK;
1316 vi->qos = vf_cfg->vlan_tag >> VLAN_PRIO_SHIFT;
1317 memcpy(&vi->mac, vf_cfg->mac_addr, ETH_ALEN);
1318 vi->linkstate = adapter->vf_cfg[vf].plink_tracking;
1323 static int be_set_vf_vlan(struct net_device *netdev, int vf, u16 vlan, u8 qos)
1325 struct be_adapter *adapter = netdev_priv(netdev);
1326 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1329 if (!sriov_enabled(adapter))
1332 if (vf >= adapter->num_vfs || vlan > 4095 || qos > 7)
1336 vlan |= qos << VLAN_PRIO_SHIFT;
1337 if (vf_cfg->vlan_tag != vlan)
1338 status = be_cmd_set_hsw_config(adapter, vlan, vf + 1,
1339 vf_cfg->if_handle, 0);
1341 /* Reset Transparent Vlan Tagging. */
1342 status = be_cmd_set_hsw_config(adapter, BE_RESET_VLAN_TAG_ID,
1343 vf + 1, vf_cfg->if_handle, 0);
1347 dev_err(&adapter->pdev->dev,
1348 "VLAN %d config on VF %d failed : %#x\n", vlan,
1350 return be_cmd_status(status);
1353 vf_cfg->vlan_tag = vlan;
1358 static int be_set_vf_tx_rate(struct net_device *netdev, int vf,
1359 int min_tx_rate, int max_tx_rate)
1361 struct be_adapter *adapter = netdev_priv(netdev);
1362 struct device *dev = &adapter->pdev->dev;
1363 int percent_rate, status = 0;
1367 if (!sriov_enabled(adapter))
1370 if (vf >= adapter->num_vfs)
1379 status = be_cmd_link_status_query(adapter, &link_speed,
1385 dev_err(dev, "TX-rate setting not allowed when link is down\n");
1390 if (max_tx_rate < 100 || max_tx_rate > link_speed) {
1391 dev_err(dev, "TX-rate must be between 100 and %d Mbps\n",
1397 /* On Skyhawk the QOS setting must be done only as a % value */
1398 percent_rate = link_speed / 100;
1399 if (skyhawk_chip(adapter) && (max_tx_rate % percent_rate)) {
1400 dev_err(dev, "TX-rate must be a multiple of %d Mbps\n",
1407 status = be_cmd_config_qos(adapter, max_tx_rate, link_speed, vf + 1);
1411 adapter->vf_cfg[vf].tx_rate = max_tx_rate;
1415 dev_err(dev, "TX-rate setting of %dMbps on VF%d failed\n",
1417 return be_cmd_status(status);
1419 static int be_set_vf_link_state(struct net_device *netdev, int vf,
1422 struct be_adapter *adapter = netdev_priv(netdev);
1425 if (!sriov_enabled(adapter))
1428 if (vf >= adapter->num_vfs)
1431 status = be_cmd_set_logical_link_config(adapter, link_state, vf+1);
1433 dev_err(&adapter->pdev->dev,
1434 "Link state change on VF %d failed: %#x\n", vf, status);
1435 return be_cmd_status(status);
1438 adapter->vf_cfg[vf].plink_tracking = link_state;
1443 static void be_aic_update(struct be_aic_obj *aic, u64 rx_pkts, u64 tx_pkts,
1446 aic->rx_pkts_prev = rx_pkts;
1447 aic->tx_reqs_prev = tx_pkts;
1451 static void be_eqd_update(struct be_adapter *adapter)
1453 struct be_set_eqd set_eqd[MAX_EVT_QS];
1454 int eqd, i, num = 0, start;
1455 struct be_aic_obj *aic;
1456 struct be_eq_obj *eqo;
1457 struct be_rx_obj *rxo;
1458 struct be_tx_obj *txo;
1459 u64 rx_pkts, tx_pkts;
1463 for_all_evt_queues(adapter, eqo, i) {
1464 aic = &adapter->aic_obj[eqo->idx];
1472 rxo = &adapter->rx_obj[eqo->idx];
1474 start = u64_stats_fetch_begin_irq(&rxo->stats.sync);
1475 rx_pkts = rxo->stats.rx_pkts;
1476 } while (u64_stats_fetch_retry_irq(&rxo->stats.sync, start));
1478 txo = &adapter->tx_obj[eqo->idx];
1480 start = u64_stats_fetch_begin_irq(&txo->stats.sync);
1481 tx_pkts = txo->stats.tx_reqs;
1482 } while (u64_stats_fetch_retry_irq(&txo->stats.sync, start));
1485 /* Skip, if wrapped around or first calculation */
1487 if (!aic->jiffies || time_before(now, aic->jiffies) ||
1488 rx_pkts < aic->rx_pkts_prev ||
1489 tx_pkts < aic->tx_reqs_prev) {
1490 be_aic_update(aic, rx_pkts, tx_pkts, now);
1494 delta = jiffies_to_msecs(now - aic->jiffies);
1495 pps = (((u32)(rx_pkts - aic->rx_pkts_prev) * 1000) / delta) +
1496 (((u32)(tx_pkts - aic->tx_reqs_prev) * 1000) / delta);
1497 eqd = (pps / 15000) << 2;
1501 eqd = min_t(u32, eqd, aic->max_eqd);
1502 eqd = max_t(u32, eqd, aic->min_eqd);
1504 be_aic_update(aic, rx_pkts, tx_pkts, now);
1506 if (eqd != aic->prev_eqd) {
1507 set_eqd[num].delay_multiplier = (eqd * 65)/100;
1508 set_eqd[num].eq_id = eqo->q.id;
1509 aic->prev_eqd = eqd;
1515 be_cmd_modify_eqd(adapter, set_eqd, num);
1518 static void be_rx_stats_update(struct be_rx_obj *rxo,
1519 struct be_rx_compl_info *rxcp)
1521 struct be_rx_stats *stats = rx_stats(rxo);
1523 u64_stats_update_begin(&stats->sync);
1525 stats->rx_bytes += rxcp->pkt_size;
1527 if (rxcp->pkt_type == BE_MULTICAST_PACKET)
1528 stats->rx_mcast_pkts++;
1530 stats->rx_compl_err++;
1531 u64_stats_update_end(&stats->sync);
1534 static inline bool csum_passed(struct be_rx_compl_info *rxcp)
1536 /* L4 checksum is not reliable for non TCP/UDP packets.
1537 * Also ignore ipcksm for ipv6 pkts
1539 return (rxcp->tcpf || rxcp->udpf) && rxcp->l4_csum &&
1540 (rxcp->ip_csum || rxcp->ipv6) && !rxcp->err;
1543 static struct be_rx_page_info *get_rx_page_info(struct be_rx_obj *rxo)
1545 struct be_adapter *adapter = rxo->adapter;
1546 struct be_rx_page_info *rx_page_info;
1547 struct be_queue_info *rxq = &rxo->q;
1548 u16 frag_idx = rxq->tail;
1550 rx_page_info = &rxo->page_info_tbl[frag_idx];
1551 BUG_ON(!rx_page_info->page);
1553 if (rx_page_info->last_frag) {
1554 dma_unmap_page(&adapter->pdev->dev,
1555 dma_unmap_addr(rx_page_info, bus),
1556 adapter->big_page_size, DMA_FROM_DEVICE);
1557 rx_page_info->last_frag = false;
1559 dma_sync_single_for_cpu(&adapter->pdev->dev,
1560 dma_unmap_addr(rx_page_info, bus),
1561 rx_frag_size, DMA_FROM_DEVICE);
1564 queue_tail_inc(rxq);
1565 atomic_dec(&rxq->used);
1566 return rx_page_info;
1569 /* Throwaway the data in the Rx completion */
1570 static void be_rx_compl_discard(struct be_rx_obj *rxo,
1571 struct be_rx_compl_info *rxcp)
1573 struct be_rx_page_info *page_info;
1574 u16 i, num_rcvd = rxcp->num_rcvd;
1576 for (i = 0; i < num_rcvd; i++) {
1577 page_info = get_rx_page_info(rxo);
1578 put_page(page_info->page);
1579 memset(page_info, 0, sizeof(*page_info));
1584 * skb_fill_rx_data forms a complete skb for an ether frame
1585 * indicated by rxcp.
1587 static void skb_fill_rx_data(struct be_rx_obj *rxo, struct sk_buff *skb,
1588 struct be_rx_compl_info *rxcp)
1590 struct be_rx_page_info *page_info;
1592 u16 hdr_len, curr_frag_len, remaining;
1595 page_info = get_rx_page_info(rxo);
1596 start = page_address(page_info->page) + page_info->page_offset;
1599 /* Copy data in the first descriptor of this completion */
1600 curr_frag_len = min(rxcp->pkt_size, rx_frag_size);
1602 skb->len = curr_frag_len;
1603 if (curr_frag_len <= BE_HDR_LEN) { /* tiny packet */
1604 memcpy(skb->data, start, curr_frag_len);
1605 /* Complete packet has now been moved to data */
1606 put_page(page_info->page);
1608 skb->tail += curr_frag_len;
1611 memcpy(skb->data, start, hdr_len);
1612 skb_shinfo(skb)->nr_frags = 1;
1613 skb_frag_set_page(skb, 0, page_info->page);
1614 skb_shinfo(skb)->frags[0].page_offset =
1615 page_info->page_offset + hdr_len;
1616 skb_frag_size_set(&skb_shinfo(skb)->frags[0],
1617 curr_frag_len - hdr_len);
1618 skb->data_len = curr_frag_len - hdr_len;
1619 skb->truesize += rx_frag_size;
1620 skb->tail += hdr_len;
1622 page_info->page = NULL;
1624 if (rxcp->pkt_size <= rx_frag_size) {
1625 BUG_ON(rxcp->num_rcvd != 1);
1629 /* More frags present for this completion */
1630 remaining = rxcp->pkt_size - curr_frag_len;
1631 for (i = 1, j = 0; i < rxcp->num_rcvd; i++) {
1632 page_info = get_rx_page_info(rxo);
1633 curr_frag_len = min(remaining, rx_frag_size);
1635 /* Coalesce all frags from the same physical page in one slot */
1636 if (page_info->page_offset == 0) {
1639 skb_frag_set_page(skb, j, page_info->page);
1640 skb_shinfo(skb)->frags[j].page_offset =
1641 page_info->page_offset;
1642 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1643 skb_shinfo(skb)->nr_frags++;
1645 put_page(page_info->page);
1648 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1649 skb->len += curr_frag_len;
1650 skb->data_len += curr_frag_len;
1651 skb->truesize += rx_frag_size;
1652 remaining -= curr_frag_len;
1653 page_info->page = NULL;
1655 BUG_ON(j > MAX_SKB_FRAGS);
1658 /* Process the RX completion indicated by rxcp when GRO is disabled */
1659 static void be_rx_compl_process(struct be_rx_obj *rxo, struct napi_struct *napi,
1660 struct be_rx_compl_info *rxcp)
1662 struct be_adapter *adapter = rxo->adapter;
1663 struct net_device *netdev = adapter->netdev;
1664 struct sk_buff *skb;
1666 skb = netdev_alloc_skb_ip_align(netdev, BE_RX_SKB_ALLOC_SIZE);
1667 if (unlikely(!skb)) {
1668 rx_stats(rxo)->rx_drops_no_skbs++;
1669 be_rx_compl_discard(rxo, rxcp);
1673 skb_fill_rx_data(rxo, skb, rxcp);
1675 if (likely((netdev->features & NETIF_F_RXCSUM) && csum_passed(rxcp)))
1676 skb->ip_summed = CHECKSUM_UNNECESSARY;
1678 skb_checksum_none_assert(skb);
1680 skb->protocol = eth_type_trans(skb, netdev);
1681 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1682 if (netdev->features & NETIF_F_RXHASH)
1683 skb_set_hash(skb, rxcp->rss_hash, PKT_HASH_TYPE_L3);
1685 skb->csum_level = rxcp->tunneled;
1686 skb_mark_napi_id(skb, napi);
1689 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1691 netif_receive_skb(skb);
1694 /* Process the RX completion indicated by rxcp when GRO is enabled */
1695 static void be_rx_compl_process_gro(struct be_rx_obj *rxo,
1696 struct napi_struct *napi,
1697 struct be_rx_compl_info *rxcp)
1699 struct be_adapter *adapter = rxo->adapter;
1700 struct be_rx_page_info *page_info;
1701 struct sk_buff *skb = NULL;
1702 u16 remaining, curr_frag_len;
1705 skb = napi_get_frags(napi);
1707 be_rx_compl_discard(rxo, rxcp);
1711 remaining = rxcp->pkt_size;
1712 for (i = 0, j = -1; i < rxcp->num_rcvd; i++) {
1713 page_info = get_rx_page_info(rxo);
1715 curr_frag_len = min(remaining, rx_frag_size);
1717 /* Coalesce all frags from the same physical page in one slot */
1718 if (i == 0 || page_info->page_offset == 0) {
1719 /* First frag or Fresh page */
1721 skb_frag_set_page(skb, j, page_info->page);
1722 skb_shinfo(skb)->frags[j].page_offset =
1723 page_info->page_offset;
1724 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1726 put_page(page_info->page);
1728 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1729 skb->truesize += rx_frag_size;
1730 remaining -= curr_frag_len;
1731 memset(page_info, 0, sizeof(*page_info));
1733 BUG_ON(j > MAX_SKB_FRAGS);
1735 skb_shinfo(skb)->nr_frags = j + 1;
1736 skb->len = rxcp->pkt_size;
1737 skb->data_len = rxcp->pkt_size;
1738 skb->ip_summed = CHECKSUM_UNNECESSARY;
1739 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1740 if (adapter->netdev->features & NETIF_F_RXHASH)
1741 skb_set_hash(skb, rxcp->rss_hash, PKT_HASH_TYPE_L3);
1743 skb->csum_level = rxcp->tunneled;
1744 skb_mark_napi_id(skb, napi);
1747 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1749 napi_gro_frags(napi);
1752 static void be_parse_rx_compl_v1(struct be_eth_rx_compl *compl,
1753 struct be_rx_compl_info *rxcp)
1755 rxcp->pkt_size = GET_RX_COMPL_V1_BITS(pktsize, compl);
1756 rxcp->vlanf = GET_RX_COMPL_V1_BITS(vtp, compl);
1757 rxcp->err = GET_RX_COMPL_V1_BITS(err, compl);
1758 rxcp->tcpf = GET_RX_COMPL_V1_BITS(tcpf, compl);
1759 rxcp->udpf = GET_RX_COMPL_V1_BITS(udpf, compl);
1760 rxcp->ip_csum = GET_RX_COMPL_V1_BITS(ipcksm, compl);
1761 rxcp->l4_csum = GET_RX_COMPL_V1_BITS(l4_cksm, compl);
1762 rxcp->ipv6 = GET_RX_COMPL_V1_BITS(ip_version, compl);
1763 rxcp->num_rcvd = GET_RX_COMPL_V1_BITS(numfrags, compl);
1764 rxcp->pkt_type = GET_RX_COMPL_V1_BITS(cast_enc, compl);
1765 rxcp->rss_hash = GET_RX_COMPL_V1_BITS(rsshash, compl);
1767 rxcp->qnq = GET_RX_COMPL_V1_BITS(qnq, compl);
1768 rxcp->vlan_tag = GET_RX_COMPL_V1_BITS(vlan_tag, compl);
1770 rxcp->port = GET_RX_COMPL_V1_BITS(port, compl);
1772 GET_RX_COMPL_V1_BITS(tunneled, compl);
1775 static void be_parse_rx_compl_v0(struct be_eth_rx_compl *compl,
1776 struct be_rx_compl_info *rxcp)
1778 rxcp->pkt_size = GET_RX_COMPL_V0_BITS(pktsize, compl);
1779 rxcp->vlanf = GET_RX_COMPL_V0_BITS(vtp, compl);
1780 rxcp->err = GET_RX_COMPL_V0_BITS(err, compl);
1781 rxcp->tcpf = GET_RX_COMPL_V0_BITS(tcpf, compl);
1782 rxcp->udpf = GET_RX_COMPL_V0_BITS(udpf, compl);
1783 rxcp->ip_csum = GET_RX_COMPL_V0_BITS(ipcksm, compl);
1784 rxcp->l4_csum = GET_RX_COMPL_V0_BITS(l4_cksm, compl);
1785 rxcp->ipv6 = GET_RX_COMPL_V0_BITS(ip_version, compl);
1786 rxcp->num_rcvd = GET_RX_COMPL_V0_BITS(numfrags, compl);
1787 rxcp->pkt_type = GET_RX_COMPL_V0_BITS(cast_enc, compl);
1788 rxcp->rss_hash = GET_RX_COMPL_V0_BITS(rsshash, compl);
1790 rxcp->qnq = GET_RX_COMPL_V0_BITS(qnq, compl);
1791 rxcp->vlan_tag = GET_RX_COMPL_V0_BITS(vlan_tag, compl);
1793 rxcp->port = GET_RX_COMPL_V0_BITS(port, compl);
1794 rxcp->ip_frag = GET_RX_COMPL_V0_BITS(ip_frag, compl);
1797 static struct be_rx_compl_info *be_rx_compl_get(struct be_rx_obj *rxo)
1799 struct be_eth_rx_compl *compl = queue_tail_node(&rxo->cq);
1800 struct be_rx_compl_info *rxcp = &rxo->rxcp;
1801 struct be_adapter *adapter = rxo->adapter;
1803 /* For checking the valid bit it is Ok to use either definition as the
1804 * valid bit is at the same position in both v0 and v1 Rx compl */
1805 if (compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] == 0)
1809 be_dws_le_to_cpu(compl, sizeof(*compl));
1811 if (adapter->be3_native)
1812 be_parse_rx_compl_v1(compl, rxcp);
1814 be_parse_rx_compl_v0(compl, rxcp);
1820 /* In QNQ modes, if qnq bit is not set, then the packet was
1821 * tagged only with the transparent outer vlan-tag and must
1822 * not be treated as a vlan packet by host
1824 if (be_is_qnq_mode(adapter) && !rxcp->qnq)
1827 if (!lancer_chip(adapter))
1828 rxcp->vlan_tag = swab16(rxcp->vlan_tag);
1830 if (adapter->pvid == (rxcp->vlan_tag & VLAN_VID_MASK) &&
1831 !test_bit(rxcp->vlan_tag, adapter->vids))
1835 /* As the compl has been parsed, reset it; we wont touch it again */
1836 compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] = 0;
1838 queue_tail_inc(&rxo->cq);
1842 static inline struct page *be_alloc_pages(u32 size, gfp_t gfp)
1844 u32 order = get_order(size);
1848 return alloc_pages(gfp, order);
1852 * Allocate a page, split it to fragments of size rx_frag_size and post as
1853 * receive buffers to BE
1855 static void be_post_rx_frags(struct be_rx_obj *rxo, gfp_t gfp, u32 frags_needed)
1857 struct be_adapter *adapter = rxo->adapter;
1858 struct be_rx_page_info *page_info = NULL, *prev_page_info = NULL;
1859 struct be_queue_info *rxq = &rxo->q;
1860 struct page *pagep = NULL;
1861 struct device *dev = &adapter->pdev->dev;
1862 struct be_eth_rx_d *rxd;
1863 u64 page_dmaaddr = 0, frag_dmaaddr;
1864 u32 posted, page_offset = 0, notify = 0;
1866 page_info = &rxo->page_info_tbl[rxq->head];
1867 for (posted = 0; posted < frags_needed && !page_info->page; posted++) {
1869 pagep = be_alloc_pages(adapter->big_page_size, gfp);
1870 if (unlikely(!pagep)) {
1871 rx_stats(rxo)->rx_post_fail++;
1874 page_dmaaddr = dma_map_page(dev, pagep, 0,
1875 adapter->big_page_size,
1877 if (dma_mapping_error(dev, page_dmaaddr)) {
1880 adapter->drv_stats.dma_map_errors++;
1886 page_offset += rx_frag_size;
1888 page_info->page_offset = page_offset;
1889 page_info->page = pagep;
1891 rxd = queue_head_node(rxq);
1892 frag_dmaaddr = page_dmaaddr + page_info->page_offset;
1893 rxd->fragpa_lo = cpu_to_le32(frag_dmaaddr & 0xFFFFFFFF);
1894 rxd->fragpa_hi = cpu_to_le32(upper_32_bits(frag_dmaaddr));
1896 /* Any space left in the current big page for another frag? */
1897 if ((page_offset + rx_frag_size + rx_frag_size) >
1898 adapter->big_page_size) {
1900 page_info->last_frag = true;
1901 dma_unmap_addr_set(page_info, bus, page_dmaaddr);
1903 dma_unmap_addr_set(page_info, bus, frag_dmaaddr);
1906 prev_page_info = page_info;
1907 queue_head_inc(rxq);
1908 page_info = &rxo->page_info_tbl[rxq->head];
1911 /* Mark the last frag of a page when we break out of the above loop
1912 * with no more slots available in the RXQ
1915 prev_page_info->last_frag = true;
1916 dma_unmap_addr_set(prev_page_info, bus, page_dmaaddr);
1920 atomic_add(posted, &rxq->used);
1921 if (rxo->rx_post_starved)
1922 rxo->rx_post_starved = false;
1924 notify = min(256u, posted);
1925 be_rxq_notify(adapter, rxq->id, notify);
1928 } else if (atomic_read(&rxq->used) == 0) {
1929 /* Let be_worker replenish when memory is available */
1930 rxo->rx_post_starved = true;
1934 static struct be_eth_tx_compl *be_tx_compl_get(struct be_queue_info *tx_cq)
1936 struct be_eth_tx_compl *txcp = queue_tail_node(tx_cq);
1938 if (txcp->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] == 0)
1942 be_dws_le_to_cpu(txcp, sizeof(*txcp));
1944 txcp->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] = 0;
1946 queue_tail_inc(tx_cq);
1950 static u16 be_tx_compl_process(struct be_adapter *adapter,
1951 struct be_tx_obj *txo, u16 last_index)
1953 struct be_queue_info *txq = &txo->q;
1954 struct be_eth_wrb *wrb;
1955 struct sk_buff **sent_skbs = txo->sent_skb_list;
1956 struct sk_buff *sent_skb;
1957 u16 cur_index, num_wrbs = 1; /* account for hdr wrb */
1958 bool unmap_skb_hdr = true;
1960 sent_skb = sent_skbs[txq->tail];
1962 sent_skbs[txq->tail] = NULL;
1964 /* skip header wrb */
1965 queue_tail_inc(txq);
1968 cur_index = txq->tail;
1969 wrb = queue_tail_node(txq);
1970 unmap_tx_frag(&adapter->pdev->dev, wrb,
1971 (unmap_skb_hdr && skb_headlen(sent_skb)));
1972 unmap_skb_hdr = false;
1975 queue_tail_inc(txq);
1976 } while (cur_index != last_index);
1978 dev_consume_skb_any(sent_skb);
1982 /* Return the number of events in the event queue */
1983 static inline int events_get(struct be_eq_obj *eqo)
1985 struct be_eq_entry *eqe;
1989 eqe = queue_tail_node(&eqo->q);
1996 queue_tail_inc(&eqo->q);
2002 /* Leaves the EQ is disarmed state */
2003 static void be_eq_clean(struct be_eq_obj *eqo)
2005 int num = events_get(eqo);
2007 be_eq_notify(eqo->adapter, eqo->q.id, false, true, num);
2010 static void be_rx_cq_clean(struct be_rx_obj *rxo)
2012 struct be_rx_page_info *page_info;
2013 struct be_queue_info *rxq = &rxo->q;
2014 struct be_queue_info *rx_cq = &rxo->cq;
2015 struct be_rx_compl_info *rxcp;
2016 struct be_adapter *adapter = rxo->adapter;
2019 /* Consume pending rx completions.
2020 * Wait for the flush completion (identified by zero num_rcvd)
2021 * to arrive. Notify CQ even when there are no more CQ entries
2022 * for HW to flush partially coalesced CQ entries.
2023 * In Lancer, there is no need to wait for flush compl.
2026 rxcp = be_rx_compl_get(rxo);
2028 if (lancer_chip(adapter))
2031 if (flush_wait++ > 10 || be_hw_error(adapter)) {
2032 dev_warn(&adapter->pdev->dev,
2033 "did not receive flush compl\n");
2036 be_cq_notify(adapter, rx_cq->id, true, 0);
2039 be_rx_compl_discard(rxo, rxcp);
2040 be_cq_notify(adapter, rx_cq->id, false, 1);
2041 if (rxcp->num_rcvd == 0)
2046 /* After cleanup, leave the CQ in unarmed state */
2047 be_cq_notify(adapter, rx_cq->id, false, 0);
2049 /* Then free posted rx buffers that were not used */
2050 while (atomic_read(&rxq->used) > 0) {
2051 page_info = get_rx_page_info(rxo);
2052 put_page(page_info->page);
2053 memset(page_info, 0, sizeof(*page_info));
2055 BUG_ON(atomic_read(&rxq->used));
2056 rxq->tail = rxq->head = 0;
2059 static void be_tx_compl_clean(struct be_adapter *adapter)
2061 struct be_tx_obj *txo;
2062 struct be_queue_info *txq;
2063 struct be_eth_tx_compl *txcp;
2064 u16 end_idx, cmpl = 0, timeo = 0, num_wrbs = 0;
2065 struct sk_buff *sent_skb;
2067 int i, pending_txqs;
2069 /* Stop polling for compls when HW has been silent for 10ms */
2071 pending_txqs = adapter->num_tx_qs;
2073 for_all_tx_queues(adapter, txo, i) {
2077 while ((txcp = be_tx_compl_get(&txo->cq))) {
2078 end_idx = GET_TX_COMPL_BITS(wrb_index, txcp);
2079 num_wrbs += be_tx_compl_process(adapter, txo,
2084 be_cq_notify(adapter, txo->cq.id, false, cmpl);
2085 atomic_sub(num_wrbs, &txq->used);
2088 if (atomic_read(&txq->used) == 0)
2092 if (pending_txqs == 0 || ++timeo > 10 || be_hw_error(adapter))
2098 for_all_tx_queues(adapter, txo, i) {
2100 if (atomic_read(&txq->used))
2101 dev_err(&adapter->pdev->dev, "%d pending tx-compls\n",
2102 atomic_read(&txq->used));
2104 /* free posted tx for which compls will never arrive */
2105 while (atomic_read(&txq->used)) {
2106 sent_skb = txo->sent_skb_list[txq->tail];
2107 end_idx = txq->tail;
2108 num_wrbs = wrb_cnt_for_skb(adapter, sent_skb,
2110 index_adv(&end_idx, num_wrbs - 1, txq->len);
2111 num_wrbs = be_tx_compl_process(adapter, txo, end_idx);
2112 atomic_sub(num_wrbs, &txq->used);
2117 static void be_evt_queues_destroy(struct be_adapter *adapter)
2119 struct be_eq_obj *eqo;
2122 for_all_evt_queues(adapter, eqo, i) {
2123 if (eqo->q.created) {
2125 be_cmd_q_destroy(adapter, &eqo->q, QTYPE_EQ);
2126 napi_hash_del(&eqo->napi);
2127 netif_napi_del(&eqo->napi);
2129 be_queue_free(adapter, &eqo->q);
2133 static int be_evt_queues_create(struct be_adapter *adapter)
2135 struct be_queue_info *eq;
2136 struct be_eq_obj *eqo;
2137 struct be_aic_obj *aic;
2140 adapter->num_evt_qs = min_t(u16, num_irqs(adapter),
2141 adapter->cfg_num_qs);
2143 for_all_evt_queues(adapter, eqo, i) {
2144 netif_napi_add(adapter->netdev, &eqo->napi, be_poll,
2146 napi_hash_add(&eqo->napi);
2147 aic = &adapter->aic_obj[i];
2148 eqo->adapter = adapter;
2150 aic->max_eqd = BE_MAX_EQD;
2154 rc = be_queue_alloc(adapter, eq, EVNT_Q_LEN,
2155 sizeof(struct be_eq_entry));
2159 rc = be_cmd_eq_create(adapter, eqo);
2166 static void be_mcc_queues_destroy(struct be_adapter *adapter)
2168 struct be_queue_info *q;
2170 q = &adapter->mcc_obj.q;
2172 be_cmd_q_destroy(adapter, q, QTYPE_MCCQ);
2173 be_queue_free(adapter, q);
2175 q = &adapter->mcc_obj.cq;
2177 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2178 be_queue_free(adapter, q);
2181 /* Must be called only after TX qs are created as MCC shares TX EQ */
2182 static int be_mcc_queues_create(struct be_adapter *adapter)
2184 struct be_queue_info *q, *cq;
2186 cq = &adapter->mcc_obj.cq;
2187 if (be_queue_alloc(adapter, cq, MCC_CQ_LEN,
2188 sizeof(struct be_mcc_compl)))
2191 /* Use the default EQ for MCC completions */
2192 if (be_cmd_cq_create(adapter, cq, &mcc_eqo(adapter)->q, true, 0))
2195 q = &adapter->mcc_obj.q;
2196 if (be_queue_alloc(adapter, q, MCC_Q_LEN, sizeof(struct be_mcc_wrb)))
2197 goto mcc_cq_destroy;
2199 if (be_cmd_mccq_create(adapter, q, cq))
2205 be_queue_free(adapter, q);
2207 be_cmd_q_destroy(adapter, cq, QTYPE_CQ);
2209 be_queue_free(adapter, cq);
2214 static void be_tx_queues_destroy(struct be_adapter *adapter)
2216 struct be_queue_info *q;
2217 struct be_tx_obj *txo;
2220 for_all_tx_queues(adapter, txo, i) {
2223 be_cmd_q_destroy(adapter, q, QTYPE_TXQ);
2224 be_queue_free(adapter, q);
2228 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2229 be_queue_free(adapter, q);
2233 static int be_tx_qs_create(struct be_adapter *adapter)
2235 struct be_queue_info *cq, *eq;
2236 struct be_tx_obj *txo;
2239 adapter->num_tx_qs = min(adapter->num_evt_qs, be_max_txqs(adapter));
2241 for_all_tx_queues(adapter, txo, i) {
2243 status = be_queue_alloc(adapter, cq, TX_CQ_LEN,
2244 sizeof(struct be_eth_tx_compl));
2248 u64_stats_init(&txo->stats.sync);
2249 u64_stats_init(&txo->stats.sync_compl);
2251 /* If num_evt_qs is less than num_tx_qs, then more than
2252 * one txq share an eq
2254 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2255 status = be_cmd_cq_create(adapter, cq, eq, false, 3);
2259 status = be_queue_alloc(adapter, &txo->q, TX_Q_LEN,
2260 sizeof(struct be_eth_wrb));
2264 status = be_cmd_txq_create(adapter, txo);
2269 dev_info(&adapter->pdev->dev, "created %d TX queue(s)\n",
2270 adapter->num_tx_qs);
2274 static void be_rx_cqs_destroy(struct be_adapter *adapter)
2276 struct be_queue_info *q;
2277 struct be_rx_obj *rxo;
2280 for_all_rx_queues(adapter, rxo, i) {
2283 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2284 be_queue_free(adapter, q);
2288 static int be_rx_cqs_create(struct be_adapter *adapter)
2290 struct be_queue_info *eq, *cq;
2291 struct be_rx_obj *rxo;
2294 /* We can create as many RSS rings as there are EQs. */
2295 adapter->num_rx_qs = adapter->num_evt_qs;
2297 /* We'll use RSS only if atleast 2 RSS rings are supported.
2298 * When RSS is used, we'll need a default RXQ for non-IP traffic.
2300 if (adapter->num_rx_qs > 1)
2301 adapter->num_rx_qs++;
2303 adapter->big_page_size = (1 << get_order(rx_frag_size)) * PAGE_SIZE;
2304 for_all_rx_queues(adapter, rxo, i) {
2305 rxo->adapter = adapter;
2307 rc = be_queue_alloc(adapter, cq, RX_CQ_LEN,
2308 sizeof(struct be_eth_rx_compl));
2312 u64_stats_init(&rxo->stats.sync);
2313 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2314 rc = be_cmd_cq_create(adapter, cq, eq, false, 3);
2319 dev_info(&adapter->pdev->dev,
2320 "created %d RSS queue(s) and 1 default RX queue\n",
2321 adapter->num_rx_qs - 1);
2325 static irqreturn_t be_intx(int irq, void *dev)
2327 struct be_eq_obj *eqo = dev;
2328 struct be_adapter *adapter = eqo->adapter;
2331 /* IRQ is not expected when NAPI is scheduled as the EQ
2332 * will not be armed.
2333 * But, this can happen on Lancer INTx where it takes
2334 * a while to de-assert INTx or in BE2 where occasionaly
2335 * an interrupt may be raised even when EQ is unarmed.
2336 * If NAPI is already scheduled, then counting & notifying
2337 * events will orphan them.
2339 if (napi_schedule_prep(&eqo->napi)) {
2340 num_evts = events_get(eqo);
2341 __napi_schedule(&eqo->napi);
2343 eqo->spurious_intr = 0;
2345 be_eq_notify(adapter, eqo->q.id, false, true, num_evts);
2347 /* Return IRQ_HANDLED only for the the first spurious intr
2348 * after a valid intr to stop the kernel from branding
2349 * this irq as a bad one!
2351 if (num_evts || eqo->spurious_intr++ == 0)
2357 static irqreturn_t be_msix(int irq, void *dev)
2359 struct be_eq_obj *eqo = dev;
2361 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
2362 napi_schedule(&eqo->napi);
2366 static inline bool do_gro(struct be_rx_compl_info *rxcp)
2368 return (rxcp->tcpf && !rxcp->err && rxcp->l4_csum) ? true : false;
2371 static int be_process_rx(struct be_rx_obj *rxo, struct napi_struct *napi,
2372 int budget, int polling)
2374 struct be_adapter *adapter = rxo->adapter;
2375 struct be_queue_info *rx_cq = &rxo->cq;
2376 struct be_rx_compl_info *rxcp;
2378 u32 frags_consumed = 0;
2380 for (work_done = 0; work_done < budget; work_done++) {
2381 rxcp = be_rx_compl_get(rxo);
2385 /* Is it a flush compl that has no data */
2386 if (unlikely(rxcp->num_rcvd == 0))
2389 /* Discard compl with partial DMA Lancer B0 */
2390 if (unlikely(!rxcp->pkt_size)) {
2391 be_rx_compl_discard(rxo, rxcp);
2395 /* On BE drop pkts that arrive due to imperfect filtering in
2396 * promiscuous mode on some skews
2398 if (unlikely(rxcp->port != adapter->port_num &&
2399 !lancer_chip(adapter))) {
2400 be_rx_compl_discard(rxo, rxcp);
2404 /* Don't do gro when we're busy_polling */
2405 if (do_gro(rxcp) && polling != BUSY_POLLING)
2406 be_rx_compl_process_gro(rxo, napi, rxcp);
2408 be_rx_compl_process(rxo, napi, rxcp);
2411 frags_consumed += rxcp->num_rcvd;
2412 be_rx_stats_update(rxo, rxcp);
2416 be_cq_notify(adapter, rx_cq->id, true, work_done);
2418 /* When an rx-obj gets into post_starved state, just
2419 * let be_worker do the posting.
2421 if (atomic_read(&rxo->q.used) < RX_FRAGS_REFILL_WM &&
2422 !rxo->rx_post_starved)
2423 be_post_rx_frags(rxo, GFP_ATOMIC,
2424 max_t(u32, MAX_RX_POST,
2431 static inline void be_update_tx_err(struct be_tx_obj *txo, u32 status)
2434 case BE_TX_COMP_HDR_PARSE_ERR:
2435 tx_stats(txo)->tx_hdr_parse_err++;
2437 case BE_TX_COMP_NDMA_ERR:
2438 tx_stats(txo)->tx_dma_err++;
2440 case BE_TX_COMP_ACL_ERR:
2441 tx_stats(txo)->tx_spoof_check_err++;
2446 static inline void lancer_update_tx_err(struct be_tx_obj *txo, u32 status)
2449 case LANCER_TX_COMP_LSO_ERR:
2450 tx_stats(txo)->tx_tso_err++;
2452 case LANCER_TX_COMP_HSW_DROP_MAC_ERR:
2453 case LANCER_TX_COMP_HSW_DROP_VLAN_ERR:
2454 tx_stats(txo)->tx_spoof_check_err++;
2456 case LANCER_TX_COMP_QINQ_ERR:
2457 tx_stats(txo)->tx_qinq_err++;
2459 case LANCER_TX_COMP_PARITY_ERR:
2460 tx_stats(txo)->tx_internal_parity_err++;
2462 case LANCER_TX_COMP_DMA_ERR:
2463 tx_stats(txo)->tx_dma_err++;
2468 static void be_process_tx(struct be_adapter *adapter, struct be_tx_obj *txo,
2471 struct be_eth_tx_compl *txcp;
2472 int num_wrbs = 0, work_done = 0;
2476 while ((txcp = be_tx_compl_get(&txo->cq))) {
2477 last_idx = GET_TX_COMPL_BITS(wrb_index, txcp);
2478 num_wrbs += be_tx_compl_process(adapter, txo, last_idx);
2481 compl_status = GET_TX_COMPL_BITS(status, txcp);
2483 if (lancer_chip(adapter))
2484 lancer_update_tx_err(txo, compl_status);
2486 be_update_tx_err(txo, compl_status);
2491 be_cq_notify(adapter, txo->cq.id, true, work_done);
2492 atomic_sub(num_wrbs, &txo->q.used);
2494 /* As Tx wrbs have been freed up, wake up netdev queue
2495 * if it was stopped due to lack of tx wrbs. */
2496 if (__netif_subqueue_stopped(adapter->netdev, idx) &&
2497 atomic_read(&txo->q.used) < txo->q.len / 2) {
2498 netif_wake_subqueue(adapter->netdev, idx);
2501 u64_stats_update_begin(&tx_stats(txo)->sync_compl);
2502 tx_stats(txo)->tx_compl += work_done;
2503 u64_stats_update_end(&tx_stats(txo)->sync_compl);
2507 int be_poll(struct napi_struct *napi, int budget)
2509 struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2510 struct be_adapter *adapter = eqo->adapter;
2511 int max_work = 0, work, i, num_evts;
2512 struct be_rx_obj *rxo;
2513 struct be_tx_obj *txo;
2515 num_evts = events_get(eqo);
2517 for_all_tx_queues_on_eq(adapter, eqo, txo, i)
2518 be_process_tx(adapter, txo, i);
2520 if (be_lock_napi(eqo)) {
2521 /* This loop will iterate twice for EQ0 in which
2522 * completions of the last RXQ (default one) are also processed
2523 * For other EQs the loop iterates only once
2525 for_all_rx_queues_on_eq(adapter, eqo, rxo, i) {
2526 work = be_process_rx(rxo, napi, budget, NAPI_POLLING);
2527 max_work = max(work, max_work);
2529 be_unlock_napi(eqo);
2534 if (is_mcc_eqo(eqo))
2535 be_process_mcc(adapter);
2537 if (max_work < budget) {
2538 napi_complete(napi);
2539 be_eq_notify(adapter, eqo->q.id, true, false, num_evts);
2541 /* As we'll continue in polling mode, count and clear events */
2542 be_eq_notify(adapter, eqo->q.id, false, false, num_evts);
2547 #ifdef CONFIG_NET_RX_BUSY_POLL
2548 static int be_busy_poll(struct napi_struct *napi)
2550 struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2551 struct be_adapter *adapter = eqo->adapter;
2552 struct be_rx_obj *rxo;
2555 if (!be_lock_busy_poll(eqo))
2556 return LL_FLUSH_BUSY;
2558 for_all_rx_queues_on_eq(adapter, eqo, rxo, i) {
2559 work = be_process_rx(rxo, napi, 4, BUSY_POLLING);
2564 be_unlock_busy_poll(eqo);
2569 void be_detect_error(struct be_adapter *adapter)
2571 u32 ue_lo = 0, ue_hi = 0, ue_lo_mask = 0, ue_hi_mask = 0;
2572 u32 sliport_status = 0, sliport_err1 = 0, sliport_err2 = 0;
2574 bool error_detected = false;
2575 struct device *dev = &adapter->pdev->dev;
2576 struct net_device *netdev = adapter->netdev;
2578 if (be_hw_error(adapter))
2581 if (lancer_chip(adapter)) {
2582 sliport_status = ioread32(adapter->db + SLIPORT_STATUS_OFFSET);
2583 if (sliport_status & SLIPORT_STATUS_ERR_MASK) {
2584 sliport_err1 = ioread32(adapter->db +
2585 SLIPORT_ERROR1_OFFSET);
2586 sliport_err2 = ioread32(adapter->db +
2587 SLIPORT_ERROR2_OFFSET);
2588 adapter->hw_error = true;
2589 /* Do not log error messages if its a FW reset */
2590 if (sliport_err1 == SLIPORT_ERROR_FW_RESET1 &&
2591 sliport_err2 == SLIPORT_ERROR_FW_RESET2) {
2592 dev_info(dev, "Firmware update in progress\n");
2594 error_detected = true;
2595 dev_err(dev, "Error detected in the card\n");
2596 dev_err(dev, "ERR: sliport status 0x%x\n",
2598 dev_err(dev, "ERR: sliport error1 0x%x\n",
2600 dev_err(dev, "ERR: sliport error2 0x%x\n",
2605 pci_read_config_dword(adapter->pdev,
2606 PCICFG_UE_STATUS_LOW, &ue_lo);
2607 pci_read_config_dword(adapter->pdev,
2608 PCICFG_UE_STATUS_HIGH, &ue_hi);
2609 pci_read_config_dword(adapter->pdev,
2610 PCICFG_UE_STATUS_LOW_MASK, &ue_lo_mask);
2611 pci_read_config_dword(adapter->pdev,
2612 PCICFG_UE_STATUS_HI_MASK, &ue_hi_mask);
2614 ue_lo = (ue_lo & ~ue_lo_mask);
2615 ue_hi = (ue_hi & ~ue_hi_mask);
2617 /* On certain platforms BE hardware can indicate spurious UEs.
2618 * Allow HW to stop working completely in case of a real UE.
2619 * Hence not setting the hw_error for UE detection.
2622 if (ue_lo || ue_hi) {
2623 error_detected = true;
2625 "Unrecoverable Error detected in the adapter");
2626 dev_err(dev, "Please reboot server to recover");
2627 if (skyhawk_chip(adapter))
2628 adapter->hw_error = true;
2629 for (i = 0; ue_lo; ue_lo >>= 1, i++) {
2631 dev_err(dev, "UE: %s bit set\n",
2632 ue_status_low_desc[i]);
2634 for (i = 0; ue_hi; ue_hi >>= 1, i++) {
2636 dev_err(dev, "UE: %s bit set\n",
2637 ue_status_hi_desc[i]);
2642 netif_carrier_off(netdev);
2645 static void be_msix_disable(struct be_adapter *adapter)
2647 if (msix_enabled(adapter)) {
2648 pci_disable_msix(adapter->pdev);
2649 adapter->num_msix_vec = 0;
2650 adapter->num_msix_roce_vec = 0;
2654 static int be_msix_enable(struct be_adapter *adapter)
2657 struct device *dev = &adapter->pdev->dev;
2659 /* If RoCE is supported, program the max number of NIC vectors that
2660 * may be configured via set-channels, along with vectors needed for
2661 * RoCe. Else, just program the number we'll use initially.
2663 if (be_roce_supported(adapter))
2664 num_vec = min_t(int, 2 * be_max_eqs(adapter),
2665 2 * num_online_cpus());
2667 num_vec = adapter->cfg_num_qs;
2669 for (i = 0; i < num_vec; i++)
2670 adapter->msix_entries[i].entry = i;
2672 num_vec = pci_enable_msix_range(adapter->pdev, adapter->msix_entries,
2673 MIN_MSIX_VECTORS, num_vec);
2677 if (be_roce_supported(adapter) && num_vec > MIN_MSIX_VECTORS) {
2678 adapter->num_msix_roce_vec = num_vec / 2;
2679 dev_info(dev, "enabled %d MSI-x vector(s) for RoCE\n",
2680 adapter->num_msix_roce_vec);
2683 adapter->num_msix_vec = num_vec - adapter->num_msix_roce_vec;
2685 dev_info(dev, "enabled %d MSI-x vector(s) for NIC\n",
2686 adapter->num_msix_vec);
2690 dev_warn(dev, "MSIx enable failed\n");
2692 /* INTx is not supported in VFs, so fail probe if enable_msix fails */
2693 if (!be_physfn(adapter))
2698 static inline int be_msix_vec_get(struct be_adapter *adapter,
2699 struct be_eq_obj *eqo)
2701 return adapter->msix_entries[eqo->msix_idx].vector;
2704 static int be_msix_register(struct be_adapter *adapter)
2706 struct net_device *netdev = adapter->netdev;
2707 struct be_eq_obj *eqo;
2710 for_all_evt_queues(adapter, eqo, i) {
2711 sprintf(eqo->desc, "%s-q%d", netdev->name, i);
2712 vec = be_msix_vec_get(adapter, eqo);
2713 status = request_irq(vec, be_msix, 0, eqo->desc, eqo);
2720 for (i--, eqo = &adapter->eq_obj[i]; i >= 0; i--, eqo--)
2721 free_irq(be_msix_vec_get(adapter, eqo), eqo);
2722 dev_warn(&adapter->pdev->dev, "MSIX Request IRQ failed - err %d\n",
2724 be_msix_disable(adapter);
2728 static int be_irq_register(struct be_adapter *adapter)
2730 struct net_device *netdev = adapter->netdev;
2733 if (msix_enabled(adapter)) {
2734 status = be_msix_register(adapter);
2737 /* INTx is not supported for VF */
2738 if (!be_physfn(adapter))
2742 /* INTx: only the first EQ is used */
2743 netdev->irq = adapter->pdev->irq;
2744 status = request_irq(netdev->irq, be_intx, IRQF_SHARED, netdev->name,
2745 &adapter->eq_obj[0]);
2747 dev_err(&adapter->pdev->dev,
2748 "INTx request IRQ failed - err %d\n", status);
2752 adapter->isr_registered = true;
2756 static void be_irq_unregister(struct be_adapter *adapter)
2758 struct net_device *netdev = adapter->netdev;
2759 struct be_eq_obj *eqo;
2762 if (!adapter->isr_registered)
2766 if (!msix_enabled(adapter)) {
2767 free_irq(netdev->irq, &adapter->eq_obj[0]);
2772 for_all_evt_queues(adapter, eqo, i)
2773 free_irq(be_msix_vec_get(adapter, eqo), eqo);
2776 adapter->isr_registered = false;
2779 static void be_rx_qs_destroy(struct be_adapter *adapter)
2781 struct be_queue_info *q;
2782 struct be_rx_obj *rxo;
2785 for_all_rx_queues(adapter, rxo, i) {
2788 be_cmd_rxq_destroy(adapter, q);
2789 be_rx_cq_clean(rxo);
2791 be_queue_free(adapter, q);
2795 static int be_close(struct net_device *netdev)
2797 struct be_adapter *adapter = netdev_priv(netdev);
2798 struct be_eq_obj *eqo;
2801 /* This protection is needed as be_close() may be called even when the
2802 * adapter is in cleared state (after eeh perm failure)
2804 if (!(adapter->flags & BE_FLAGS_SETUP_DONE))
2807 be_roce_dev_close(adapter);
2809 if (adapter->flags & BE_FLAGS_NAPI_ENABLED) {
2810 for_all_evt_queues(adapter, eqo, i) {
2811 napi_disable(&eqo->napi);
2812 be_disable_busy_poll(eqo);
2814 adapter->flags &= ~BE_FLAGS_NAPI_ENABLED;
2817 be_async_mcc_disable(adapter);
2819 /* Wait for all pending tx completions to arrive so that
2820 * all tx skbs are freed.
2822 netif_tx_disable(netdev);
2823 be_tx_compl_clean(adapter);
2825 be_rx_qs_destroy(adapter);
2827 for (i = 1; i < (adapter->uc_macs + 1); i++)
2828 be_cmd_pmac_del(adapter, adapter->if_handle,
2829 adapter->pmac_id[i], 0);
2830 adapter->uc_macs = 0;
2832 for_all_evt_queues(adapter, eqo, i) {
2833 if (msix_enabled(adapter))
2834 synchronize_irq(be_msix_vec_get(adapter, eqo));
2836 synchronize_irq(netdev->irq);
2840 be_irq_unregister(adapter);
2845 static int be_rx_qs_create(struct be_adapter *adapter)
2847 struct be_rx_obj *rxo;
2849 u8 rss_hkey[RSS_HASH_KEY_LEN];
2850 struct rss_info *rss = &adapter->rss_info;
2852 for_all_rx_queues(adapter, rxo, i) {
2853 rc = be_queue_alloc(adapter, &rxo->q, RX_Q_LEN,
2854 sizeof(struct be_eth_rx_d));
2859 /* The FW would like the default RXQ to be created first */
2860 rxo = default_rxo(adapter);
2861 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id, rx_frag_size,
2862 adapter->if_handle, false, &rxo->rss_id);
2866 for_all_rss_queues(adapter, rxo, i) {
2867 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id,
2868 rx_frag_size, adapter->if_handle,
2869 true, &rxo->rss_id);
2874 if (be_multi_rxq(adapter)) {
2875 for (j = 0; j < RSS_INDIR_TABLE_LEN;
2876 j += adapter->num_rx_qs - 1) {
2877 for_all_rss_queues(adapter, rxo, i) {
2878 if ((j + i) >= RSS_INDIR_TABLE_LEN)
2880 rss->rsstable[j + i] = rxo->rss_id;
2881 rss->rss_queue[j + i] = i;
2884 rss->rss_flags = RSS_ENABLE_TCP_IPV4 | RSS_ENABLE_IPV4 |
2885 RSS_ENABLE_TCP_IPV6 | RSS_ENABLE_IPV6;
2887 if (!BEx_chip(adapter))
2888 rss->rss_flags |= RSS_ENABLE_UDP_IPV4 |
2889 RSS_ENABLE_UDP_IPV6;
2891 /* Disable RSS, if only default RX Q is created */
2892 rss->rss_flags = RSS_ENABLE_NONE;
2895 get_random_bytes(rss_hkey, RSS_HASH_KEY_LEN);
2896 rc = be_cmd_rss_config(adapter, rss->rsstable, rss->rss_flags,
2899 rss->rss_flags = RSS_ENABLE_NONE;
2903 memcpy(rss->rss_hkey, rss_hkey, RSS_HASH_KEY_LEN);
2905 /* First time posting */
2906 for_all_rx_queues(adapter, rxo, i)
2907 be_post_rx_frags(rxo, GFP_KERNEL, MAX_RX_POST);
2911 static int be_open(struct net_device *netdev)
2913 struct be_adapter *adapter = netdev_priv(netdev);
2914 struct be_eq_obj *eqo;
2915 struct be_rx_obj *rxo;
2916 struct be_tx_obj *txo;
2920 status = be_rx_qs_create(adapter);
2924 status = be_irq_register(adapter);
2928 for_all_rx_queues(adapter, rxo, i)
2929 be_cq_notify(adapter, rxo->cq.id, true, 0);
2931 for_all_tx_queues(adapter, txo, i)
2932 be_cq_notify(adapter, txo->cq.id, true, 0);
2934 be_async_mcc_enable(adapter);
2936 for_all_evt_queues(adapter, eqo, i) {
2937 napi_enable(&eqo->napi);
2938 be_enable_busy_poll(eqo);
2939 be_eq_notify(adapter, eqo->q.id, true, true, 0);
2941 adapter->flags |= BE_FLAGS_NAPI_ENABLED;
2943 status = be_cmd_link_status_query(adapter, NULL, &link_status, 0);
2945 be_link_status_update(adapter, link_status);
2947 netif_tx_start_all_queues(netdev);
2948 be_roce_dev_open(adapter);
2950 #ifdef CONFIG_BE2NET_VXLAN
2951 if (skyhawk_chip(adapter))
2952 vxlan_get_rx_port(netdev);
2957 be_close(adapter->netdev);
2961 static int be_setup_wol(struct be_adapter *adapter, bool enable)
2963 struct be_dma_mem cmd;
2967 memset(mac, 0, ETH_ALEN);
2969 cmd.size = sizeof(struct be_cmd_req_acpi_wol_magic_config);
2970 cmd.va = dma_zalloc_coherent(&adapter->pdev->dev, cmd.size, &cmd.dma,
2976 status = pci_write_config_dword(adapter->pdev,
2977 PCICFG_PM_CONTROL_OFFSET,
2978 PCICFG_PM_CONTROL_MASK);
2980 dev_err(&adapter->pdev->dev,
2981 "Could not enable Wake-on-lan\n");
2982 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va,
2986 status = be_cmd_enable_magic_wol(adapter,
2987 adapter->netdev->dev_addr,
2989 pci_enable_wake(adapter->pdev, PCI_D3hot, 1);
2990 pci_enable_wake(adapter->pdev, PCI_D3cold, 1);
2992 status = be_cmd_enable_magic_wol(adapter, mac, &cmd);
2993 pci_enable_wake(adapter->pdev, PCI_D3hot, 0);
2994 pci_enable_wake(adapter->pdev, PCI_D3cold, 0);
2997 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va, cmd.dma);
3002 * Generate a seed MAC address from the PF MAC Address using jhash.
3003 * MAC Address for VFs are assigned incrementally starting from the seed.
3004 * These addresses are programmed in the ASIC by the PF and the VF driver
3005 * queries for the MAC address during its probe.
3007 static int be_vf_eth_addr_config(struct be_adapter *adapter)
3012 struct be_vf_cfg *vf_cfg;
3014 be_vf_eth_addr_generate(adapter, mac);
3016 for_all_vfs(adapter, vf_cfg, vf) {
3017 if (BEx_chip(adapter))
3018 status = be_cmd_pmac_add(adapter, mac,
3020 &vf_cfg->pmac_id, vf + 1);
3022 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
3026 dev_err(&adapter->pdev->dev,
3027 "Mac address assignment failed for VF %d\n",
3030 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
3037 static int be_vfs_mac_query(struct be_adapter *adapter)
3041 struct be_vf_cfg *vf_cfg;
3043 for_all_vfs(adapter, vf_cfg, vf) {
3044 status = be_cmd_get_active_mac(adapter, vf_cfg->pmac_id,
3045 mac, vf_cfg->if_handle,
3049 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
3054 static void be_vf_clear(struct be_adapter *adapter)
3056 struct be_vf_cfg *vf_cfg;
3059 if (pci_vfs_assigned(adapter->pdev)) {
3060 dev_warn(&adapter->pdev->dev,
3061 "VFs are assigned to VMs: not disabling VFs\n");
3065 pci_disable_sriov(adapter->pdev);
3067 for_all_vfs(adapter, vf_cfg, vf) {
3068 if (BEx_chip(adapter))
3069 be_cmd_pmac_del(adapter, vf_cfg->if_handle,
3070 vf_cfg->pmac_id, vf + 1);
3072 be_cmd_set_mac(adapter, NULL, vf_cfg->if_handle,
3075 be_cmd_if_destroy(adapter, vf_cfg->if_handle, vf + 1);
3078 kfree(adapter->vf_cfg);
3079 adapter->num_vfs = 0;
3080 adapter->flags &= ~BE_FLAGS_SRIOV_ENABLED;
3083 static void be_clear_queues(struct be_adapter *adapter)
3085 be_mcc_queues_destroy(adapter);
3086 be_rx_cqs_destroy(adapter);
3087 be_tx_queues_destroy(adapter);
3088 be_evt_queues_destroy(adapter);
3091 static void be_cancel_worker(struct be_adapter *adapter)
3093 if (adapter->flags & BE_FLAGS_WORKER_SCHEDULED) {
3094 cancel_delayed_work_sync(&adapter->work);
3095 adapter->flags &= ~BE_FLAGS_WORKER_SCHEDULED;
3099 static void be_mac_clear(struct be_adapter *adapter)
3103 if (adapter->pmac_id) {
3104 for (i = 0; i < (adapter->uc_macs + 1); i++)
3105 be_cmd_pmac_del(adapter, adapter->if_handle,
3106 adapter->pmac_id[i], 0);
3107 adapter->uc_macs = 0;
3109 kfree(adapter->pmac_id);
3110 adapter->pmac_id = NULL;
3114 #ifdef CONFIG_BE2NET_VXLAN
3115 static void be_disable_vxlan_offloads(struct be_adapter *adapter)
3117 if (adapter->flags & BE_FLAGS_VXLAN_OFFLOADS)
3118 be_cmd_manage_iface(adapter, adapter->if_handle,
3119 OP_CONVERT_TUNNEL_TO_NORMAL);
3121 if (adapter->vxlan_port)
3122 be_cmd_set_vxlan_port(adapter, 0);
3124 adapter->flags &= ~BE_FLAGS_VXLAN_OFFLOADS;
3125 adapter->vxlan_port = 0;
3129 static int be_clear(struct be_adapter *adapter)
3131 be_cancel_worker(adapter);
3133 if (sriov_enabled(adapter))
3134 be_vf_clear(adapter);
3136 /* Re-configure FW to distribute resources evenly across max-supported
3137 * number of VFs, only when VFs are not already enabled.
3139 if (be_physfn(adapter) && !pci_vfs_assigned(adapter->pdev))
3140 be_cmd_set_sriov_config(adapter, adapter->pool_res,
3141 pci_sriov_get_totalvfs(adapter->pdev));
3143 #ifdef CONFIG_BE2NET_VXLAN
3144 be_disable_vxlan_offloads(adapter);
3146 /* delete the primary mac along with the uc-mac list */
3147 be_mac_clear(adapter);
3149 be_cmd_if_destroy(adapter, adapter->if_handle, 0);
3151 be_clear_queues(adapter);
3153 be_msix_disable(adapter);
3154 adapter->flags &= ~BE_FLAGS_SETUP_DONE;
3158 static int be_vfs_if_create(struct be_adapter *adapter)
3160 struct be_resources res = {0};
3161 struct be_vf_cfg *vf_cfg;
3162 u32 cap_flags, en_flags, vf;
3165 cap_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3166 BE_IF_FLAGS_MULTICAST;
3168 for_all_vfs(adapter, vf_cfg, vf) {
3169 if (!BE3_chip(adapter)) {
3170 status = be_cmd_get_profile_config(adapter, &res,
3173 cap_flags = res.if_cap_flags;
3176 /* If a FW profile exists, then cap_flags are updated */
3177 en_flags = cap_flags & (BE_IF_FLAGS_UNTAGGED |
3178 BE_IF_FLAGS_BROADCAST |
3179 BE_IF_FLAGS_MULTICAST);
3181 be_cmd_if_create(adapter, cap_flags, en_flags,
3182 &vf_cfg->if_handle, vf + 1);
3190 static int be_vf_setup_init(struct be_adapter *adapter)
3192 struct be_vf_cfg *vf_cfg;
3195 adapter->vf_cfg = kcalloc(adapter->num_vfs, sizeof(*vf_cfg),
3197 if (!adapter->vf_cfg)
3200 for_all_vfs(adapter, vf_cfg, vf) {
3201 vf_cfg->if_handle = -1;
3202 vf_cfg->pmac_id = -1;
3207 static int be_vf_setup(struct be_adapter *adapter)
3209 struct device *dev = &adapter->pdev->dev;
3210 struct be_vf_cfg *vf_cfg;
3211 int status, old_vfs, vf;
3214 old_vfs = pci_num_vf(adapter->pdev);
3216 status = be_vf_setup_init(adapter);
3221 for_all_vfs(adapter, vf_cfg, vf) {
3222 status = be_cmd_get_if_id(adapter, vf_cfg, vf);
3227 status = be_vfs_mac_query(adapter);
3231 status = be_vfs_if_create(adapter);
3235 status = be_vf_eth_addr_config(adapter);
3240 for_all_vfs(adapter, vf_cfg, vf) {
3241 /* Allow VFs to programs MAC/VLAN filters */
3242 status = be_cmd_get_fn_privileges(adapter, &privileges, vf + 1);
3243 if (!status && !(privileges & BE_PRIV_FILTMGMT)) {
3244 status = be_cmd_set_fn_privileges(adapter,
3249 dev_info(dev, "VF%d has FILTMGMT privilege\n",
3253 /* Allow full available bandwidth */
3255 be_cmd_config_qos(adapter, 0, 0, vf + 1);
3258 be_cmd_enable_vf(adapter, vf + 1);
3259 be_cmd_set_logical_link_config(adapter,
3260 IFLA_VF_LINK_STATE_AUTO,
3266 status = pci_enable_sriov(adapter->pdev, adapter->num_vfs);
3268 dev_err(dev, "SRIOV enable failed\n");
3269 adapter->num_vfs = 0;
3274 adapter->flags |= BE_FLAGS_SRIOV_ENABLED;
3277 dev_err(dev, "VF setup failed\n");
3278 be_vf_clear(adapter);
3282 /* Converting function_mode bits on BE3 to SH mc_type enums */
3284 static u8 be_convert_mc_type(u32 function_mode)
3286 if (function_mode & VNIC_MODE && function_mode & QNQ_MODE)
3288 else if (function_mode & QNQ_MODE)
3290 else if (function_mode & VNIC_MODE)
3292 else if (function_mode & UMC_ENABLED)
3298 /* On BE2/BE3 FW does not suggest the supported limits */
3299 static void BEx_get_resources(struct be_adapter *adapter,
3300 struct be_resources *res)
3302 bool use_sriov = adapter->num_vfs ? 1 : 0;
3304 if (be_physfn(adapter))
3305 res->max_uc_mac = BE_UC_PMAC_COUNT;
3307 res->max_uc_mac = BE_VF_UC_PMAC_COUNT;
3309 adapter->mc_type = be_convert_mc_type(adapter->function_mode);
3311 if (be_is_mc(adapter)) {
3312 /* Assuming that there are 4 channels per port,
3313 * when multi-channel is enabled
3315 if (be_is_qnq_mode(adapter))
3316 res->max_vlans = BE_NUM_VLANS_SUPPORTED/8;
3318 /* In a non-qnq multichannel mode, the pvid
3319 * takes up one vlan entry
3321 res->max_vlans = (BE_NUM_VLANS_SUPPORTED / 4) - 1;
3323 res->max_vlans = BE_NUM_VLANS_SUPPORTED;
3326 res->max_mcast_mac = BE_MAX_MC;
3328 /* 1) For BE3 1Gb ports, FW does not support multiple TXQs
3329 * 2) Create multiple TX rings on a BE3-R multi-channel interface
3330 * *only* if it is RSS-capable.
3332 if (BE2_chip(adapter) || use_sriov || (adapter->port_num > 1) ||
3333 !be_physfn(adapter) || (be_is_mc(adapter) &&
3334 !(adapter->function_caps & BE_FUNCTION_CAPS_RSS))) {
3336 } else if (adapter->function_caps & BE_FUNCTION_CAPS_SUPER_NIC) {
3337 struct be_resources super_nic_res = {0};
3339 /* On a SuperNIC profile, the driver needs to use the
3340 * GET_PROFILE_CONFIG cmd to query the per-function TXQ limits
3342 be_cmd_get_profile_config(adapter, &super_nic_res, 0);
3343 /* Some old versions of BE3 FW don't report max_tx_qs value */
3344 res->max_tx_qs = super_nic_res.max_tx_qs ? : BE3_MAX_TX_QS;
3346 res->max_tx_qs = BE3_MAX_TX_QS;
3349 if ((adapter->function_caps & BE_FUNCTION_CAPS_RSS) &&
3350 !use_sriov && be_physfn(adapter))
3351 res->max_rss_qs = (adapter->be3_native) ?
3352 BE3_MAX_RSS_QS : BE2_MAX_RSS_QS;
3353 res->max_rx_qs = res->max_rss_qs + 1;
3355 if (be_physfn(adapter))
3356 res->max_evt_qs = (be_max_vfs(adapter) > 0) ?
3357 BE3_SRIOV_MAX_EVT_QS : BE3_MAX_EVT_QS;
3359 res->max_evt_qs = 1;
3361 res->if_cap_flags = BE_IF_CAP_FLAGS_WANT;
3362 if (!(adapter->function_caps & BE_FUNCTION_CAPS_RSS))
3363 res->if_cap_flags &= ~BE_IF_FLAGS_RSS;
3366 static void be_setup_init(struct be_adapter *adapter)
3368 adapter->vlan_prio_bmap = 0xff;
3369 adapter->phy.link_speed = -1;
3370 adapter->if_handle = -1;
3371 adapter->be3_native = false;
3372 adapter->promiscuous = false;
3373 if (be_physfn(adapter))
3374 adapter->cmd_privileges = MAX_PRIVILEGES;
3376 adapter->cmd_privileges = MIN_PRIVILEGES;
3379 static int be_get_sriov_config(struct be_adapter *adapter)
3381 struct device *dev = &adapter->pdev->dev;
3382 struct be_resources res = {0};
3383 int max_vfs, old_vfs;
3385 /* Some old versions of BE3 FW don't report max_vfs value */
3386 be_cmd_get_profile_config(adapter, &res, 0);
3388 if (BE3_chip(adapter) && !res.max_vfs) {
3389 max_vfs = pci_sriov_get_totalvfs(adapter->pdev);
3390 res.max_vfs = max_vfs > 0 ? min(MAX_VFS, max_vfs) : 0;
3393 adapter->pool_res = res;
3395 if (!be_max_vfs(adapter)) {
3397 dev_warn(dev, "SRIOV is disabled. Ignoring num_vfs\n");
3398 adapter->num_vfs = 0;
3402 pci_sriov_set_totalvfs(adapter->pdev, be_max_vfs(adapter));
3404 /* validate num_vfs module param */
3405 old_vfs = pci_num_vf(adapter->pdev);
3407 dev_info(dev, "%d VFs are already enabled\n", old_vfs);
3408 if (old_vfs != num_vfs)
3409 dev_warn(dev, "Ignoring num_vfs=%d setting\n", num_vfs);
3410 adapter->num_vfs = old_vfs;
3412 if (num_vfs > be_max_vfs(adapter)) {
3413 dev_info(dev, "Resources unavailable to init %d VFs\n",
3415 dev_info(dev, "Limiting to %d VFs\n",
3416 be_max_vfs(adapter));
3418 adapter->num_vfs = min_t(u16, num_vfs, be_max_vfs(adapter));
3424 static int be_get_resources(struct be_adapter *adapter)
3426 struct device *dev = &adapter->pdev->dev;
3427 struct be_resources res = {0};
3430 if (BEx_chip(adapter)) {
3431 BEx_get_resources(adapter, &res);
3435 /* For Lancer, SH etc read per-function resource limits from FW.
3436 * GET_FUNC_CONFIG returns per function guaranteed limits.
3437 * GET_PROFILE_CONFIG returns PCI-E related limits PF-pool limits
3439 if (!BEx_chip(adapter)) {
3440 status = be_cmd_get_func_config(adapter, &res);
3444 /* If RoCE may be enabled stash away half the EQs for RoCE */
3445 if (be_roce_supported(adapter))
3446 res.max_evt_qs /= 2;
3450 dev_info(dev, "Max: txqs %d, rxqs %d, rss %d, eqs %d, vfs %d\n",
3451 be_max_txqs(adapter), be_max_rxqs(adapter),
3452 be_max_rss(adapter), be_max_eqs(adapter),
3453 be_max_vfs(adapter));
3454 dev_info(dev, "Max: uc-macs %d, mc-macs %d, vlans %d\n",
3455 be_max_uc(adapter), be_max_mc(adapter),
3456 be_max_vlans(adapter));
3461 static void be_sriov_config(struct be_adapter *adapter)
3463 struct device *dev = &adapter->pdev->dev;
3466 status = be_get_sriov_config(adapter);
3468 dev_err(dev, "Failed to query SR-IOV configuration\n");
3469 dev_err(dev, "SR-IOV cannot be enabled\n");
3473 /* When the HW is in SRIOV capable configuration, the PF-pool
3474 * resources are equally distributed across the max-number of
3475 * VFs. The user may request only a subset of the max-vfs to be
3476 * enabled. Based on num_vfs, redistribute the resources across
3477 * num_vfs so that each VF will have access to more number of
3478 * resources. This facility is not available in BE3 FW.
3479 * Also, this is done by FW in Lancer chip.
3481 if (be_max_vfs(adapter) && !pci_num_vf(adapter->pdev)) {
3482 status = be_cmd_set_sriov_config(adapter,
3486 dev_err(dev, "Failed to optimize SR-IOV resources\n");
3490 static int be_get_config(struct be_adapter *adapter)
3495 status = be_cmd_query_fw_cfg(adapter);
3499 if (be_physfn(adapter)) {
3500 status = be_cmd_get_active_profile(adapter, &profile_id);
3502 dev_info(&adapter->pdev->dev,
3503 "Using profile 0x%x\n", profile_id);
3506 if (!BE2_chip(adapter) && be_physfn(adapter))
3507 be_sriov_config(adapter);
3509 status = be_get_resources(adapter);
3513 adapter->pmac_id = kcalloc(be_max_uc(adapter),
3514 sizeof(*adapter->pmac_id), GFP_KERNEL);
3515 if (!adapter->pmac_id)
3518 /* Sanitize cfg_num_qs based on HW and platform limits */
3519 adapter->cfg_num_qs = min(adapter->cfg_num_qs, be_max_qs(adapter));
3524 static int be_mac_setup(struct be_adapter *adapter)
3529 if (is_zero_ether_addr(adapter->netdev->dev_addr)) {
3530 status = be_cmd_get_perm_mac(adapter, mac);
3534 memcpy(adapter->netdev->dev_addr, mac, ETH_ALEN);
3535 memcpy(adapter->netdev->perm_addr, mac, ETH_ALEN);
3537 /* Maybe the HW was reset; dev_addr must be re-programmed */
3538 memcpy(mac, adapter->netdev->dev_addr, ETH_ALEN);
3541 /* For BE3-R VFs, the PF programs the initial MAC address */
3542 if (!(BEx_chip(adapter) && be_virtfn(adapter)))
3543 be_cmd_pmac_add(adapter, mac, adapter->if_handle,
3544 &adapter->pmac_id[0], 0);
3548 static void be_schedule_worker(struct be_adapter *adapter)
3550 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
3551 adapter->flags |= BE_FLAGS_WORKER_SCHEDULED;
3554 static int be_setup_queues(struct be_adapter *adapter)
3556 struct net_device *netdev = adapter->netdev;
3559 status = be_evt_queues_create(adapter);
3563 status = be_tx_qs_create(adapter);
3567 status = be_rx_cqs_create(adapter);
3571 status = be_mcc_queues_create(adapter);
3575 status = netif_set_real_num_rx_queues(netdev, adapter->num_rx_qs);
3579 status = netif_set_real_num_tx_queues(netdev, adapter->num_tx_qs);
3585 dev_err(&adapter->pdev->dev, "queue_setup failed\n");
3589 int be_update_queues(struct be_adapter *adapter)
3591 struct net_device *netdev = adapter->netdev;
3594 if (netif_running(netdev))
3597 be_cancel_worker(adapter);
3599 /* If any vectors have been shared with RoCE we cannot re-program
3602 if (!adapter->num_msix_roce_vec)
3603 be_msix_disable(adapter);
3605 be_clear_queues(adapter);
3607 if (!msix_enabled(adapter)) {
3608 status = be_msix_enable(adapter);
3613 status = be_setup_queues(adapter);
3617 be_schedule_worker(adapter);
3619 if (netif_running(netdev))
3620 status = be_open(netdev);
3625 static int be_setup(struct be_adapter *adapter)
3627 struct device *dev = &adapter->pdev->dev;
3628 u32 tx_fc, rx_fc, en_flags;
3631 be_setup_init(adapter);
3633 if (!lancer_chip(adapter))
3634 be_cmd_req_native_mode(adapter);
3636 status = be_get_config(adapter);
3640 status = be_msix_enable(adapter);
3644 en_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3645 BE_IF_FLAGS_MULTICAST | BE_IF_FLAGS_PASS_L3L4_ERRORS;
3646 if (adapter->function_caps & BE_FUNCTION_CAPS_RSS)
3647 en_flags |= BE_IF_FLAGS_RSS;
3648 en_flags = en_flags & be_if_cap_flags(adapter);
3649 status = be_cmd_if_create(adapter, be_if_cap_flags(adapter), en_flags,
3650 &adapter->if_handle, 0);
3654 /* Updating real_num_tx/rx_queues() requires rtnl_lock() */
3656 status = be_setup_queues(adapter);
3661 be_cmd_get_fn_privileges(adapter, &adapter->cmd_privileges, 0);
3663 status = be_mac_setup(adapter);
3667 be_cmd_get_fw_ver(adapter);
3668 dev_info(dev, "FW version is %s\n", adapter->fw_ver);
3670 if (BE2_chip(adapter) && fw_major_num(adapter->fw_ver) < 4) {
3671 dev_err(dev, "Firmware on card is old(%s), IRQs may not work",
3673 dev_err(dev, "Please upgrade firmware to version >= 4.0\n");
3676 if (adapter->vlans_added)
3677 be_vid_config(adapter);
3679 be_set_rx_mode(adapter->netdev);
3681 be_cmd_get_acpi_wol_cap(adapter);
3683 be_cmd_get_flow_control(adapter, &tx_fc, &rx_fc);
3685 if (rx_fc != adapter->rx_fc || tx_fc != adapter->tx_fc)
3686 be_cmd_set_flow_control(adapter, adapter->tx_fc,
3689 if (be_physfn(adapter))
3690 be_cmd_set_logical_link_config(adapter,
3691 IFLA_VF_LINK_STATE_AUTO, 0);
3693 if (adapter->num_vfs)
3694 be_vf_setup(adapter);
3696 status = be_cmd_get_phy_info(adapter);
3697 if (!status && be_pause_supported(adapter))
3698 adapter->phy.fc_autoneg = 1;
3700 be_schedule_worker(adapter);
3701 adapter->flags |= BE_FLAGS_SETUP_DONE;
3708 #ifdef CONFIG_NET_POLL_CONTROLLER
3709 static void be_netpoll(struct net_device *netdev)
3711 struct be_adapter *adapter = netdev_priv(netdev);
3712 struct be_eq_obj *eqo;
3715 for_all_evt_queues(adapter, eqo, i) {
3716 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
3717 napi_schedule(&eqo->napi);
3722 static char flash_cookie[2][16] = {"*** SE FLAS", "H DIRECTORY *** "};
3724 static bool phy_flashing_required(struct be_adapter *adapter)
3726 return (adapter->phy.phy_type == TN_8022 &&
3727 adapter->phy.interface_type == PHY_TYPE_BASET_10GB);
3730 static bool is_comp_in_ufi(struct be_adapter *adapter,
3731 struct flash_section_info *fsec, int type)
3733 int i = 0, img_type = 0;
3734 struct flash_section_info_g2 *fsec_g2 = NULL;
3736 if (BE2_chip(adapter))
3737 fsec_g2 = (struct flash_section_info_g2 *)fsec;
3739 for (i = 0; i < MAX_FLASH_COMP; i++) {
3741 img_type = le32_to_cpu(fsec_g2->fsec_entry[i].type);
3743 img_type = le32_to_cpu(fsec->fsec_entry[i].type);
3745 if (img_type == type)
3752 static struct flash_section_info *get_fsec_info(struct be_adapter *adapter,
3754 const struct firmware *fw)
3756 struct flash_section_info *fsec = NULL;
3757 const u8 *p = fw->data;
3760 while (p < (fw->data + fw->size)) {
3761 fsec = (struct flash_section_info *)p;
3762 if (!memcmp(flash_cookie, fsec->cookie, sizeof(flash_cookie)))
3769 static int be_check_flash_crc(struct be_adapter *adapter, const u8 *p,
3770 u32 img_offset, u32 img_size, int hdr_size,
3771 u16 img_optype, bool *crc_match)
3777 status = be_cmd_get_flash_crc(adapter, crc, img_optype, img_size - 4);
3781 crc_offset = hdr_size + img_offset + img_size - 4;
3783 /* Skip flashing, if crc of flashed region matches */
3784 if (!memcmp(crc, p + crc_offset, 4))
3792 static int be_flash(struct be_adapter *adapter, const u8 *img,
3793 struct be_dma_mem *flash_cmd, int optype, int img_size)
3795 struct be_cmd_write_flashrom *req = flash_cmd->va;
3796 u32 total_bytes, flash_op, num_bytes;
3799 total_bytes = img_size;
3800 while (total_bytes) {
3801 num_bytes = min_t(u32, 32*1024, total_bytes);
3803 total_bytes -= num_bytes;
3806 if (optype == OPTYPE_PHY_FW)
3807 flash_op = FLASHROM_OPER_PHY_FLASH;
3809 flash_op = FLASHROM_OPER_FLASH;
3811 if (optype == OPTYPE_PHY_FW)
3812 flash_op = FLASHROM_OPER_PHY_SAVE;
3814 flash_op = FLASHROM_OPER_SAVE;
3817 memcpy(req->data_buf, img, num_bytes);
3819 status = be_cmd_write_flashrom(adapter, flash_cmd, optype,
3820 flash_op, num_bytes);
3821 if (base_status(status) == MCC_STATUS_ILLEGAL_REQUEST &&
3822 optype == OPTYPE_PHY_FW)
3830 /* For BE2, BE3 and BE3-R */
3831 static int be_flash_BEx(struct be_adapter *adapter,
3832 const struct firmware *fw,
3833 struct be_dma_mem *flash_cmd, int num_of_images)
3835 int img_hdrs_size = (num_of_images * sizeof(struct image_hdr));
3836 struct device *dev = &adapter->pdev->dev;
3837 struct flash_section_info *fsec = NULL;
3838 int status, i, filehdr_size, num_comp;
3839 const struct flash_comp *pflashcomp;
3843 struct flash_comp gen3_flash_types[] = {
3844 { FLASH_iSCSI_PRIMARY_IMAGE_START_g3, OPTYPE_ISCSI_ACTIVE,
3845 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_iSCSI},
3846 { FLASH_REDBOOT_START_g3, OPTYPE_REDBOOT,
3847 FLASH_REDBOOT_IMAGE_MAX_SIZE_g3, IMAGE_BOOT_CODE},
3848 { FLASH_iSCSI_BIOS_START_g3, OPTYPE_BIOS,
3849 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_ISCSI},
3850 { FLASH_PXE_BIOS_START_g3, OPTYPE_PXE_BIOS,
3851 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_PXE},
3852 { FLASH_FCoE_BIOS_START_g3, OPTYPE_FCOE_BIOS,
3853 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_FCoE},
3854 { FLASH_iSCSI_BACKUP_IMAGE_START_g3, OPTYPE_ISCSI_BACKUP,
3855 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_iSCSI},
3856 { FLASH_FCoE_PRIMARY_IMAGE_START_g3, OPTYPE_FCOE_FW_ACTIVE,
3857 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_FCoE},
3858 { FLASH_FCoE_BACKUP_IMAGE_START_g3, OPTYPE_FCOE_FW_BACKUP,
3859 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_FCoE},
3860 { FLASH_NCSI_START_g3, OPTYPE_NCSI_FW,
3861 FLASH_NCSI_IMAGE_MAX_SIZE_g3, IMAGE_NCSI},
3862 { FLASH_PHY_FW_START_g3, OPTYPE_PHY_FW,
3863 FLASH_PHY_FW_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_PHY}
3866 struct flash_comp gen2_flash_types[] = {
3867 { FLASH_iSCSI_PRIMARY_IMAGE_START_g2, OPTYPE_ISCSI_ACTIVE,
3868 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_iSCSI},
3869 { FLASH_REDBOOT_START_g2, OPTYPE_REDBOOT,
3870 FLASH_REDBOOT_IMAGE_MAX_SIZE_g2, IMAGE_BOOT_CODE},
3871 { FLASH_iSCSI_BIOS_START_g2, OPTYPE_BIOS,
3872 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_ISCSI},
3873 { FLASH_PXE_BIOS_START_g2, OPTYPE_PXE_BIOS,
3874 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_PXE},
3875 { FLASH_FCoE_BIOS_START_g2, OPTYPE_FCOE_BIOS,
3876 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_FCoE},
3877 { FLASH_iSCSI_BACKUP_IMAGE_START_g2, OPTYPE_ISCSI_BACKUP,
3878 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_iSCSI},
3879 { FLASH_FCoE_PRIMARY_IMAGE_START_g2, OPTYPE_FCOE_FW_ACTIVE,
3880 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_FCoE},
3881 { FLASH_FCoE_BACKUP_IMAGE_START_g2, OPTYPE_FCOE_FW_BACKUP,
3882 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_FCoE}
3885 if (BE3_chip(adapter)) {
3886 pflashcomp = gen3_flash_types;
3887 filehdr_size = sizeof(struct flash_file_hdr_g3);
3888 num_comp = ARRAY_SIZE(gen3_flash_types);
3890 pflashcomp = gen2_flash_types;
3891 filehdr_size = sizeof(struct flash_file_hdr_g2);
3892 num_comp = ARRAY_SIZE(gen2_flash_types);
3895 /* Get flash section info*/
3896 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
3898 dev_err(dev, "Invalid Cookie. FW image may be corrupted\n");
3901 for (i = 0; i < num_comp; i++) {
3902 if (!is_comp_in_ufi(adapter, fsec, pflashcomp[i].img_type))
3905 if ((pflashcomp[i].optype == OPTYPE_NCSI_FW) &&
3906 memcmp(adapter->fw_ver, "3.102.148.0", 11) < 0)
3909 if (pflashcomp[i].optype == OPTYPE_PHY_FW &&
3910 !phy_flashing_required(adapter))
3913 if (pflashcomp[i].optype == OPTYPE_REDBOOT) {
3914 status = be_check_flash_crc(adapter, fw->data,
3915 pflashcomp[i].offset,
3919 OPTYPE_REDBOOT, &crc_match);
3922 "Could not get CRC for 0x%x region\n",
3923 pflashcomp[i].optype);
3931 p = fw->data + filehdr_size + pflashcomp[i].offset +
3933 if (p + pflashcomp[i].size > fw->data + fw->size)
3936 status = be_flash(adapter, p, flash_cmd, pflashcomp[i].optype,
3937 pflashcomp[i].size);
3939 dev_err(dev, "Flashing section type 0x%x failed\n",
3940 pflashcomp[i].img_type);
3947 static u16 be_get_img_optype(struct flash_section_entry fsec_entry)
3949 u32 img_type = le32_to_cpu(fsec_entry.type);
3950 u16 img_optype = le16_to_cpu(fsec_entry.optype);
3952 if (img_optype != 0xFFFF)
3956 case IMAGE_FIRMWARE_iSCSI:
3957 img_optype = OPTYPE_ISCSI_ACTIVE;
3959 case IMAGE_BOOT_CODE:
3960 img_optype = OPTYPE_REDBOOT;
3962 case IMAGE_OPTION_ROM_ISCSI:
3963 img_optype = OPTYPE_BIOS;
3965 case IMAGE_OPTION_ROM_PXE:
3966 img_optype = OPTYPE_PXE_BIOS;
3968 case IMAGE_OPTION_ROM_FCoE:
3969 img_optype = OPTYPE_FCOE_BIOS;
3971 case IMAGE_FIRMWARE_BACKUP_iSCSI:
3972 img_optype = OPTYPE_ISCSI_BACKUP;
3975 img_optype = OPTYPE_NCSI_FW;
3977 case IMAGE_FLASHISM_JUMPVECTOR:
3978 img_optype = OPTYPE_FLASHISM_JUMPVECTOR;
3980 case IMAGE_FIRMWARE_PHY:
3981 img_optype = OPTYPE_SH_PHY_FW;
3983 case IMAGE_REDBOOT_DIR:
3984 img_optype = OPTYPE_REDBOOT_DIR;
3986 case IMAGE_REDBOOT_CONFIG:
3987 img_optype = OPTYPE_REDBOOT_CONFIG;
3990 img_optype = OPTYPE_UFI_DIR;
3999 static int be_flash_skyhawk(struct be_adapter *adapter,
4000 const struct firmware *fw,
4001 struct be_dma_mem *flash_cmd, int num_of_images)
4003 int img_hdrs_size = num_of_images * sizeof(struct image_hdr);
4004 struct device *dev = &adapter->pdev->dev;
4005 struct flash_section_info *fsec = NULL;
4006 u32 img_offset, img_size, img_type;
4007 int status, i, filehdr_size;
4008 bool crc_match, old_fw_img;
4012 filehdr_size = sizeof(struct flash_file_hdr_g3);
4013 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
4015 dev_err(dev, "Invalid Cookie. FW image may be corrupted\n");
4019 for (i = 0; i < le32_to_cpu(fsec->fsec_hdr.num_images); i++) {
4020 img_offset = le32_to_cpu(fsec->fsec_entry[i].offset);
4021 img_size = le32_to_cpu(fsec->fsec_entry[i].pad_size);
4022 img_type = le32_to_cpu(fsec->fsec_entry[i].type);
4023 img_optype = be_get_img_optype(fsec->fsec_entry[i]);
4024 old_fw_img = fsec->fsec_entry[i].optype == 0xFFFF;
4026 if (img_optype == 0xFFFF)
4028 /* Don't bother verifying CRC if an old FW image is being
4034 status = be_check_flash_crc(adapter, fw->data, img_offset,
4035 img_size, filehdr_size +
4036 img_hdrs_size, img_optype,
4038 /* The current FW image on the card does not recognize the new
4039 * FLASH op_type. The FW download is partially complete.
4040 * Reboot the server now to enable FW image to recognize the
4041 * new FLASH op_type. To complete the remaining process,
4042 * download the same FW again after the reboot.
4044 if (base_status(status) == MCC_STATUS_ILLEGAL_REQUEST ||
4045 base_status(status) == MCC_STATUS_ILLEGAL_FIELD) {
4046 dev_err(dev, "Flash incomplete. Reset the server\n");
4047 dev_err(dev, "Download FW image again after reset\n");
4049 } else if (status) {
4050 dev_err(dev, "Could not get CRC for 0x%x region\n",
4059 p = fw->data + filehdr_size + img_offset + img_hdrs_size;
4060 if (p + img_size > fw->data + fw->size)
4063 status = be_flash(adapter, p, flash_cmd, img_optype, img_size);
4064 /* For old FW images ignore ILLEGAL_FIELD error or errors on
4068 (base_status(status) == MCC_STATUS_ILLEGAL_FIELD ||
4069 (img_optype == OPTYPE_UFI_DIR &&
4070 base_status(status) == MCC_STATUS_FAILED))) {
4072 } else if (status) {
4073 dev_err(dev, "Flashing section type 0x%x failed\n",
4081 static int lancer_fw_download(struct be_adapter *adapter,
4082 const struct firmware *fw)
4084 #define LANCER_FW_DOWNLOAD_CHUNK (32 * 1024)
4085 #define LANCER_FW_DOWNLOAD_LOCATION "/prg"
4086 struct device *dev = &adapter->pdev->dev;
4087 struct be_dma_mem flash_cmd;
4088 const u8 *data_ptr = NULL;
4089 u8 *dest_image_ptr = NULL;
4090 size_t image_size = 0;
4092 u32 data_written = 0;
4098 if (!IS_ALIGNED(fw->size, sizeof(u32))) {
4099 dev_err(dev, "FW image size should be multiple of 4\n");
4103 flash_cmd.size = sizeof(struct lancer_cmd_req_write_object)
4104 + LANCER_FW_DOWNLOAD_CHUNK;
4105 flash_cmd.va = dma_alloc_coherent(dev, flash_cmd.size,
4106 &flash_cmd.dma, GFP_KERNEL);
4110 dest_image_ptr = flash_cmd.va +
4111 sizeof(struct lancer_cmd_req_write_object);
4112 image_size = fw->size;
4113 data_ptr = fw->data;
4115 while (image_size) {
4116 chunk_size = min_t(u32, image_size, LANCER_FW_DOWNLOAD_CHUNK);
4118 /* Copy the image chunk content. */
4119 memcpy(dest_image_ptr, data_ptr, chunk_size);
4121 status = lancer_cmd_write_object(adapter, &flash_cmd,
4123 LANCER_FW_DOWNLOAD_LOCATION,
4124 &data_written, &change_status,
4129 offset += data_written;
4130 data_ptr += data_written;
4131 image_size -= data_written;
4135 /* Commit the FW written */
4136 status = lancer_cmd_write_object(adapter, &flash_cmd,
4138 LANCER_FW_DOWNLOAD_LOCATION,
4139 &data_written, &change_status,
4143 dma_free_coherent(dev, flash_cmd.size, flash_cmd.va, flash_cmd.dma);
4145 dev_err(dev, "Firmware load error\n");
4146 return be_cmd_status(status);
4149 dev_info(dev, "Firmware flashed successfully\n");
4151 if (change_status == LANCER_FW_RESET_NEEDED) {
4152 dev_info(dev, "Resetting adapter to activate new FW\n");
4153 status = lancer_physdev_ctrl(adapter,
4154 PHYSDEV_CONTROL_FW_RESET_MASK);
4156 dev_err(dev, "Adapter busy, could not reset FW\n");
4157 dev_err(dev, "Reboot server to activate new FW\n");
4159 } else if (change_status != LANCER_NO_RESET_NEEDED) {
4160 dev_info(dev, "Reboot server to activate new FW\n");
4168 #define UFI_TYPE3R 10
4170 static int be_get_ufi_type(struct be_adapter *adapter,
4171 struct flash_file_hdr_g3 *fhdr)
4174 goto be_get_ufi_exit;
4176 if (skyhawk_chip(adapter) && fhdr->build[0] == '4')
4178 else if (BE3_chip(adapter) && fhdr->build[0] == '3') {
4179 if (fhdr->asic_type_rev == 0x10)
4183 } else if (BE2_chip(adapter) && fhdr->build[0] == '2')
4187 dev_err(&adapter->pdev->dev,
4188 "UFI and Interface are not compatible for flashing\n");
4192 static int be_fw_download(struct be_adapter *adapter, const struct firmware* fw)
4194 struct flash_file_hdr_g3 *fhdr3;
4195 struct image_hdr *img_hdr_ptr = NULL;
4196 struct be_dma_mem flash_cmd;
4198 int status = 0, i = 0, num_imgs = 0, ufi_type = 0;
4200 flash_cmd.size = sizeof(struct be_cmd_write_flashrom);
4201 flash_cmd.va = dma_alloc_coherent(&adapter->pdev->dev, flash_cmd.size,
4202 &flash_cmd.dma, GFP_KERNEL);
4203 if (!flash_cmd.va) {
4209 fhdr3 = (struct flash_file_hdr_g3 *)p;
4211 ufi_type = be_get_ufi_type(adapter, fhdr3);
4213 num_imgs = le32_to_cpu(fhdr3->num_imgs);
4214 for (i = 0; i < num_imgs; i++) {
4215 img_hdr_ptr = (struct image_hdr *)(fw->data +
4216 (sizeof(struct flash_file_hdr_g3) +
4217 i * sizeof(struct image_hdr)));
4218 if (le32_to_cpu(img_hdr_ptr->imageid) == 1) {
4221 status = be_flash_skyhawk(adapter, fw,
4222 &flash_cmd, num_imgs);
4225 status = be_flash_BEx(adapter, fw, &flash_cmd,
4229 /* Do not flash this ufi on BE3-R cards */
4230 if (adapter->asic_rev < 0x10)
4231 status = be_flash_BEx(adapter, fw,
4236 dev_err(&adapter->pdev->dev,
4237 "Can't load BE3 UFI on BE3R\n");
4243 if (ufi_type == UFI_TYPE2)
4244 status = be_flash_BEx(adapter, fw, &flash_cmd, 0);
4245 else if (ufi_type == -1)
4248 dma_free_coherent(&adapter->pdev->dev, flash_cmd.size, flash_cmd.va,
4251 dev_err(&adapter->pdev->dev, "Firmware load error\n");
4255 dev_info(&adapter->pdev->dev, "Firmware flashed successfully\n");
4261 int be_load_fw(struct be_adapter *adapter, u8 *fw_file)
4263 const struct firmware *fw;
4266 if (!netif_running(adapter->netdev)) {
4267 dev_err(&adapter->pdev->dev,
4268 "Firmware load not allowed (interface is down)\n");
4272 status = request_firmware(&fw, fw_file, &adapter->pdev->dev);
4276 dev_info(&adapter->pdev->dev, "Flashing firmware file %s\n", fw_file);
4278 if (lancer_chip(adapter))
4279 status = lancer_fw_download(adapter, fw);
4281 status = be_fw_download(adapter, fw);
4284 be_cmd_get_fw_ver(adapter);
4287 release_firmware(fw);
4291 static int be_ndo_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh)
4293 struct be_adapter *adapter = netdev_priv(dev);
4294 struct nlattr *attr, *br_spec;
4299 if (!sriov_enabled(adapter))
4302 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4304 nla_for_each_nested(attr, br_spec, rem) {
4305 if (nla_type(attr) != IFLA_BRIDGE_MODE)
4308 mode = nla_get_u16(attr);
4309 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
4312 status = be_cmd_set_hsw_config(adapter, 0, 0,
4314 mode == BRIDGE_MODE_VEPA ?
4315 PORT_FWD_TYPE_VEPA :
4320 dev_info(&adapter->pdev->dev, "enabled switch mode: %s\n",
4321 mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
4326 dev_err(&adapter->pdev->dev, "Failed to set switch mode %s\n",
4327 mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
4332 static int be_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4333 struct net_device *dev, u32 filter_mask)
4335 struct be_adapter *adapter = netdev_priv(dev);
4339 if (!sriov_enabled(adapter))
4342 /* BE and Lancer chips support VEB mode only */
4343 if (BEx_chip(adapter) || lancer_chip(adapter)) {
4344 hsw_mode = PORT_FWD_TYPE_VEB;
4346 status = be_cmd_get_hsw_config(adapter, NULL, 0,
4347 adapter->if_handle, &hsw_mode);
4352 return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
4353 hsw_mode == PORT_FWD_TYPE_VEPA ?
4354 BRIDGE_MODE_VEPA : BRIDGE_MODE_VEB);
4357 #ifdef CONFIG_BE2NET_VXLAN
4358 static void be_add_vxlan_port(struct net_device *netdev, sa_family_t sa_family,
4361 struct be_adapter *adapter = netdev_priv(netdev);
4362 struct device *dev = &adapter->pdev->dev;
4365 if (lancer_chip(adapter) || BEx_chip(adapter))
4368 if (adapter->flags & BE_FLAGS_VXLAN_OFFLOADS) {
4369 dev_warn(dev, "Cannot add UDP port %d for VxLAN offloads\n",
4372 "Only one UDP port supported for VxLAN offloads\n");
4376 status = be_cmd_manage_iface(adapter, adapter->if_handle,
4377 OP_CONVERT_NORMAL_TO_TUNNEL);
4379 dev_warn(dev, "Failed to convert normal interface to tunnel\n");
4383 status = be_cmd_set_vxlan_port(adapter, port);
4385 dev_warn(dev, "Failed to add VxLAN port\n");
4388 adapter->flags |= BE_FLAGS_VXLAN_OFFLOADS;
4389 adapter->vxlan_port = port;
4391 dev_info(dev, "Enabled VxLAN offloads for UDP port %d\n",
4395 be_disable_vxlan_offloads(adapter);
4398 static void be_del_vxlan_port(struct net_device *netdev, sa_family_t sa_family,
4401 struct be_adapter *adapter = netdev_priv(netdev);
4403 if (lancer_chip(adapter) || BEx_chip(adapter))
4406 if (adapter->vxlan_port != port)
4409 be_disable_vxlan_offloads(adapter);
4411 dev_info(&adapter->pdev->dev,
4412 "Disabled VxLAN offloads for UDP port %d\n",
4417 static const struct net_device_ops be_netdev_ops = {
4418 .ndo_open = be_open,
4419 .ndo_stop = be_close,
4420 .ndo_start_xmit = be_xmit,
4421 .ndo_set_rx_mode = be_set_rx_mode,
4422 .ndo_set_mac_address = be_mac_addr_set,
4423 .ndo_change_mtu = be_change_mtu,
4424 .ndo_get_stats64 = be_get_stats64,
4425 .ndo_validate_addr = eth_validate_addr,
4426 .ndo_vlan_rx_add_vid = be_vlan_add_vid,
4427 .ndo_vlan_rx_kill_vid = be_vlan_rem_vid,
4428 .ndo_set_vf_mac = be_set_vf_mac,
4429 .ndo_set_vf_vlan = be_set_vf_vlan,
4430 .ndo_set_vf_rate = be_set_vf_tx_rate,
4431 .ndo_get_vf_config = be_get_vf_config,
4432 .ndo_set_vf_link_state = be_set_vf_link_state,
4433 #ifdef CONFIG_NET_POLL_CONTROLLER
4434 .ndo_poll_controller = be_netpoll,
4436 .ndo_bridge_setlink = be_ndo_bridge_setlink,
4437 .ndo_bridge_getlink = be_ndo_bridge_getlink,
4438 #ifdef CONFIG_NET_RX_BUSY_POLL
4439 .ndo_busy_poll = be_busy_poll,
4441 #ifdef CONFIG_BE2NET_VXLAN
4442 .ndo_add_vxlan_port = be_add_vxlan_port,
4443 .ndo_del_vxlan_port = be_del_vxlan_port,
4447 static void be_netdev_init(struct net_device *netdev)
4449 struct be_adapter *adapter = netdev_priv(netdev);
4451 if (skyhawk_chip(adapter)) {
4452 netdev->hw_enc_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
4453 NETIF_F_TSO | NETIF_F_TSO6 |
4454 NETIF_F_GSO_UDP_TUNNEL;
4455 netdev->hw_features |= NETIF_F_GSO_UDP_TUNNEL;
4457 netdev->hw_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
4458 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM | NETIF_F_RXCSUM |
4459 NETIF_F_HW_VLAN_CTAG_TX;
4460 if (be_multi_rxq(adapter))
4461 netdev->hw_features |= NETIF_F_RXHASH;
4463 netdev->features |= netdev->hw_features |
4464 NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_FILTER;
4466 netdev->vlan_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
4467 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
4469 netdev->priv_flags |= IFF_UNICAST_FLT;
4471 netdev->flags |= IFF_MULTICAST;
4473 netif_set_gso_max_size(netdev, 65535 - ETH_HLEN);
4475 netdev->netdev_ops = &be_netdev_ops;
4477 netdev->ethtool_ops = &be_ethtool_ops;
4480 static void be_unmap_pci_bars(struct be_adapter *adapter)
4483 pci_iounmap(adapter->pdev, adapter->csr);
4485 pci_iounmap(adapter->pdev, adapter->db);
4488 static int db_bar(struct be_adapter *adapter)
4490 if (lancer_chip(adapter) || !be_physfn(adapter))
4496 static int be_roce_map_pci_bars(struct be_adapter *adapter)
4498 if (skyhawk_chip(adapter)) {
4499 adapter->roce_db.size = 4096;
4500 adapter->roce_db.io_addr = pci_resource_start(adapter->pdev,
4502 adapter->roce_db.total_size = pci_resource_len(adapter->pdev,
4508 static int be_map_pci_bars(struct be_adapter *adapter)
4512 if (BEx_chip(adapter) && be_physfn(adapter)) {
4513 adapter->csr = pci_iomap(adapter->pdev, 2, 0);
4518 addr = pci_iomap(adapter->pdev, db_bar(adapter), 0);
4523 be_roce_map_pci_bars(adapter);
4527 dev_err(&adapter->pdev->dev, "Error in mapping PCI BARs\n");
4528 be_unmap_pci_bars(adapter);
4532 static void be_ctrl_cleanup(struct be_adapter *adapter)
4534 struct be_dma_mem *mem = &adapter->mbox_mem_alloced;
4536 be_unmap_pci_bars(adapter);
4539 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
4542 mem = &adapter->rx_filter;
4544 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
4548 static int be_ctrl_init(struct be_adapter *adapter)
4550 struct be_dma_mem *mbox_mem_alloc = &adapter->mbox_mem_alloced;
4551 struct be_dma_mem *mbox_mem_align = &adapter->mbox_mem;
4552 struct be_dma_mem *rx_filter = &adapter->rx_filter;
4556 pci_read_config_dword(adapter->pdev, SLI_INTF_REG_OFFSET, &sli_intf);
4557 adapter->sli_family = (sli_intf & SLI_INTF_FAMILY_MASK) >>
4558 SLI_INTF_FAMILY_SHIFT;
4559 adapter->virtfn = (sli_intf & SLI_INTF_FT_MASK) ? 1 : 0;
4561 status = be_map_pci_bars(adapter);
4565 mbox_mem_alloc->size = sizeof(struct be_mcc_mailbox) + 16;
4566 mbox_mem_alloc->va = dma_alloc_coherent(&adapter->pdev->dev,
4567 mbox_mem_alloc->size,
4568 &mbox_mem_alloc->dma,
4570 if (!mbox_mem_alloc->va) {
4572 goto unmap_pci_bars;
4574 mbox_mem_align->size = sizeof(struct be_mcc_mailbox);
4575 mbox_mem_align->va = PTR_ALIGN(mbox_mem_alloc->va, 16);
4576 mbox_mem_align->dma = PTR_ALIGN(mbox_mem_alloc->dma, 16);
4577 memset(mbox_mem_align->va, 0, sizeof(struct be_mcc_mailbox));
4579 rx_filter->size = sizeof(struct be_cmd_req_rx_filter);
4580 rx_filter->va = dma_zalloc_coherent(&adapter->pdev->dev,
4581 rx_filter->size, &rx_filter->dma,
4583 if (!rx_filter->va) {
4588 mutex_init(&adapter->mbox_lock);
4589 spin_lock_init(&adapter->mcc_lock);
4590 spin_lock_init(&adapter->mcc_cq_lock);
4592 init_completion(&adapter->et_cmd_compl);
4593 pci_save_state(adapter->pdev);
4597 dma_free_coherent(&adapter->pdev->dev, mbox_mem_alloc->size,
4598 mbox_mem_alloc->va, mbox_mem_alloc->dma);
4601 be_unmap_pci_bars(adapter);
4607 static void be_stats_cleanup(struct be_adapter *adapter)
4609 struct be_dma_mem *cmd = &adapter->stats_cmd;
4612 dma_free_coherent(&adapter->pdev->dev, cmd->size,
4616 static int be_stats_init(struct be_adapter *adapter)
4618 struct be_dma_mem *cmd = &adapter->stats_cmd;
4620 if (lancer_chip(adapter))
4621 cmd->size = sizeof(struct lancer_cmd_req_pport_stats);
4622 else if (BE2_chip(adapter))
4623 cmd->size = sizeof(struct be_cmd_req_get_stats_v0);
4624 else if (BE3_chip(adapter))
4625 cmd->size = sizeof(struct be_cmd_req_get_stats_v1);
4627 /* ALL non-BE ASICs */
4628 cmd->size = sizeof(struct be_cmd_req_get_stats_v2);
4630 cmd->va = dma_zalloc_coherent(&adapter->pdev->dev, cmd->size, &cmd->dma,
4637 static void be_remove(struct pci_dev *pdev)
4639 struct be_adapter *adapter = pci_get_drvdata(pdev);
4644 be_roce_dev_remove(adapter);
4645 be_intr_set(adapter, false);
4647 cancel_delayed_work_sync(&adapter->func_recovery_work);
4649 unregister_netdev(adapter->netdev);
4653 /* tell fw we're done with firing cmds */
4654 be_cmd_fw_clean(adapter);
4656 be_stats_cleanup(adapter);
4658 be_ctrl_cleanup(adapter);
4660 pci_disable_pcie_error_reporting(pdev);
4662 pci_release_regions(pdev);
4663 pci_disable_device(pdev);
4665 free_netdev(adapter->netdev);
4668 static int be_get_initial_config(struct be_adapter *adapter)
4672 status = be_cmd_get_cntl_attributes(adapter);
4676 /* Must be a power of 2 or else MODULO will BUG_ON */
4677 adapter->be_get_temp_freq = 64;
4679 if (BEx_chip(adapter)) {
4680 level = be_cmd_get_fw_log_level(adapter);
4681 adapter->msg_enable =
4682 level <= FW_LOG_LEVEL_DEFAULT ? NETIF_MSG_HW : 0;
4685 adapter->cfg_num_qs = netif_get_num_default_rss_queues();
4689 static int lancer_recover_func(struct be_adapter *adapter)
4691 struct device *dev = &adapter->pdev->dev;
4694 status = lancer_test_and_set_rdy_state(adapter);
4698 if (netif_running(adapter->netdev))
4699 be_close(adapter->netdev);
4703 be_clear_all_error(adapter);
4705 status = be_setup(adapter);
4709 if (netif_running(adapter->netdev)) {
4710 status = be_open(adapter->netdev);
4715 dev_err(dev, "Adapter recovery successful\n");
4718 if (status == -EAGAIN)
4719 dev_err(dev, "Waiting for resource provisioning\n");
4721 dev_err(dev, "Adapter recovery failed\n");
4726 static void be_func_recovery_task(struct work_struct *work)
4728 struct be_adapter *adapter =
4729 container_of(work, struct be_adapter, func_recovery_work.work);
4732 be_detect_error(adapter);
4734 if (adapter->hw_error && lancer_chip(adapter)) {
4737 netif_device_detach(adapter->netdev);
4740 status = lancer_recover_func(adapter);
4742 netif_device_attach(adapter->netdev);
4745 /* In Lancer, for all errors other than provisioning error (-EAGAIN),
4746 * no need to attempt further recovery.
4748 if (!status || status == -EAGAIN)
4749 schedule_delayed_work(&adapter->func_recovery_work,
4750 msecs_to_jiffies(1000));
4753 static void be_worker(struct work_struct *work)
4755 struct be_adapter *adapter =
4756 container_of(work, struct be_adapter, work.work);
4757 struct be_rx_obj *rxo;
4760 /* when interrupts are not yet enabled, just reap any pending
4761 * mcc completions */
4762 if (!netif_running(adapter->netdev)) {
4764 be_process_mcc(adapter);
4769 if (!adapter->stats_cmd_sent) {
4770 if (lancer_chip(adapter))
4771 lancer_cmd_get_pport_stats(adapter,
4772 &adapter->stats_cmd);
4774 be_cmd_get_stats(adapter, &adapter->stats_cmd);
4777 if (be_physfn(adapter) &&
4778 MODULO(adapter->work_counter, adapter->be_get_temp_freq) == 0)
4779 be_cmd_get_die_temperature(adapter);
4781 for_all_rx_queues(adapter, rxo, i) {
4782 /* Replenish RX-queues starved due to memory
4783 * allocation failures.
4785 if (rxo->rx_post_starved)
4786 be_post_rx_frags(rxo, GFP_KERNEL, MAX_RX_POST);
4789 be_eqd_update(adapter);
4792 adapter->work_counter++;
4793 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
4796 /* If any VFs are already enabled don't FLR the PF */
4797 static bool be_reset_required(struct be_adapter *adapter)
4799 return pci_num_vf(adapter->pdev) ? false : true;
4802 static char *mc_name(struct be_adapter *adapter)
4804 char *str = ""; /* default */
4806 switch (adapter->mc_type) {
4832 static inline char *func_name(struct be_adapter *adapter)
4834 return be_physfn(adapter) ? "PF" : "VF";
4837 static int be_probe(struct pci_dev *pdev, const struct pci_device_id *pdev_id)
4840 struct be_adapter *adapter;
4841 struct net_device *netdev;
4844 dev_info(&pdev->dev, "%s version is %s\n", DRV_NAME, DRV_VER);
4846 status = pci_enable_device(pdev);
4850 status = pci_request_regions(pdev, DRV_NAME);
4853 pci_set_master(pdev);
4855 netdev = alloc_etherdev_mqs(sizeof(*adapter), MAX_TX_QS, MAX_RX_QS);
4860 adapter = netdev_priv(netdev);
4861 adapter->pdev = pdev;
4862 pci_set_drvdata(pdev, adapter);
4863 adapter->netdev = netdev;
4864 SET_NETDEV_DEV(netdev, &pdev->dev);
4866 status = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
4868 netdev->features |= NETIF_F_HIGHDMA;
4870 status = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
4872 dev_err(&pdev->dev, "Could not set PCI DMA Mask\n");
4877 status = pci_enable_pcie_error_reporting(pdev);
4879 dev_info(&pdev->dev, "PCIe error reporting enabled\n");
4881 status = be_ctrl_init(adapter);
4885 /* sync up with fw's ready state */
4886 if (be_physfn(adapter)) {
4887 status = be_fw_wait_ready(adapter);
4892 if (be_reset_required(adapter)) {
4893 status = be_cmd_reset_function(adapter);
4897 /* Wait for interrupts to quiesce after an FLR */
4901 /* Allow interrupts for other ULPs running on NIC function */
4902 be_intr_set(adapter, true);
4904 /* tell fw we're ready to fire cmds */
4905 status = be_cmd_fw_init(adapter);
4909 status = be_stats_init(adapter);
4913 status = be_get_initial_config(adapter);
4917 INIT_DELAYED_WORK(&adapter->work, be_worker);
4918 INIT_DELAYED_WORK(&adapter->func_recovery_work, be_func_recovery_task);
4919 adapter->rx_fc = adapter->tx_fc = true;
4921 status = be_setup(adapter);
4925 be_netdev_init(netdev);
4926 status = register_netdev(netdev);
4930 be_roce_dev_add(adapter);
4932 schedule_delayed_work(&adapter->func_recovery_work,
4933 msecs_to_jiffies(1000));
4935 be_cmd_query_port_name(adapter, &port_name);
4937 dev_info(&pdev->dev, "%s: %s %s port %c\n", nic_name(pdev),
4938 func_name(adapter), mc_name(adapter), port_name);
4945 be_stats_cleanup(adapter);
4947 be_ctrl_cleanup(adapter);
4949 free_netdev(netdev);
4951 pci_release_regions(pdev);
4953 pci_disable_device(pdev);
4955 dev_err(&pdev->dev, "%s initialization failed\n", nic_name(pdev));
4959 static int be_suspend(struct pci_dev *pdev, pm_message_t state)
4961 struct be_adapter *adapter = pci_get_drvdata(pdev);
4962 struct net_device *netdev = adapter->netdev;
4964 if (adapter->wol_en)
4965 be_setup_wol(adapter, true);
4967 be_intr_set(adapter, false);
4968 cancel_delayed_work_sync(&adapter->func_recovery_work);
4970 netif_device_detach(netdev);
4971 if (netif_running(netdev)) {
4978 pci_save_state(pdev);
4979 pci_disable_device(pdev);
4980 pci_set_power_state(pdev, pci_choose_state(pdev, state));
4984 static int be_resume(struct pci_dev *pdev)
4987 struct be_adapter *adapter = pci_get_drvdata(pdev);
4988 struct net_device *netdev = adapter->netdev;
4990 netif_device_detach(netdev);
4992 status = pci_enable_device(pdev);
4996 pci_set_power_state(pdev, PCI_D0);
4997 pci_restore_state(pdev);
4999 status = be_fw_wait_ready(adapter);
5003 be_intr_set(adapter, true);
5004 /* tell fw we're ready to fire cmds */
5005 status = be_cmd_fw_init(adapter);
5010 if (netif_running(netdev)) {
5016 schedule_delayed_work(&adapter->func_recovery_work,
5017 msecs_to_jiffies(1000));
5018 netif_device_attach(netdev);
5020 if (adapter->wol_en)
5021 be_setup_wol(adapter, false);
5027 * An FLR will stop BE from DMAing any data.
5029 static void be_shutdown(struct pci_dev *pdev)
5031 struct be_adapter *adapter = pci_get_drvdata(pdev);
5036 be_roce_dev_shutdown(adapter);
5037 cancel_delayed_work_sync(&adapter->work);
5038 cancel_delayed_work_sync(&adapter->func_recovery_work);
5040 netif_device_detach(adapter->netdev);
5042 be_cmd_reset_function(adapter);
5044 pci_disable_device(pdev);
5047 static pci_ers_result_t be_eeh_err_detected(struct pci_dev *pdev,
5048 pci_channel_state_t state)
5050 struct be_adapter *adapter = pci_get_drvdata(pdev);
5051 struct net_device *netdev = adapter->netdev;
5053 dev_err(&adapter->pdev->dev, "EEH error detected\n");
5055 if (!adapter->eeh_error) {
5056 adapter->eeh_error = true;
5058 cancel_delayed_work_sync(&adapter->func_recovery_work);
5061 netif_device_detach(netdev);
5062 if (netif_running(netdev))
5069 if (state == pci_channel_io_perm_failure)
5070 return PCI_ERS_RESULT_DISCONNECT;
5072 pci_disable_device(pdev);
5074 /* The error could cause the FW to trigger a flash debug dump.
5075 * Resetting the card while flash dump is in progress
5076 * can cause it not to recover; wait for it to finish.
5077 * Wait only for first function as it is needed only once per
5080 if (pdev->devfn == 0)
5083 return PCI_ERS_RESULT_NEED_RESET;
5086 static pci_ers_result_t be_eeh_reset(struct pci_dev *pdev)
5088 struct be_adapter *adapter = pci_get_drvdata(pdev);
5091 dev_info(&adapter->pdev->dev, "EEH reset\n");
5093 status = pci_enable_device(pdev);
5095 return PCI_ERS_RESULT_DISCONNECT;
5097 pci_set_master(pdev);
5098 pci_set_power_state(pdev, PCI_D0);
5099 pci_restore_state(pdev);
5101 /* Check if card is ok and fw is ready */
5102 dev_info(&adapter->pdev->dev,
5103 "Waiting for FW to be ready after EEH reset\n");
5104 status = be_fw_wait_ready(adapter);
5106 return PCI_ERS_RESULT_DISCONNECT;
5108 pci_cleanup_aer_uncorrect_error_status(pdev);
5109 be_clear_all_error(adapter);
5110 return PCI_ERS_RESULT_RECOVERED;
5113 static void be_eeh_resume(struct pci_dev *pdev)
5116 struct be_adapter *adapter = pci_get_drvdata(pdev);
5117 struct net_device *netdev = adapter->netdev;
5119 dev_info(&adapter->pdev->dev, "EEH resume\n");
5121 pci_save_state(pdev);
5123 status = be_cmd_reset_function(adapter);
5127 /* On some BE3 FW versions, after a HW reset,
5128 * interrupts will remain disabled for each function.
5129 * So, explicitly enable interrupts
5131 be_intr_set(adapter, true);
5133 /* tell fw we're ready to fire cmds */
5134 status = be_cmd_fw_init(adapter);
5138 status = be_setup(adapter);
5142 if (netif_running(netdev)) {
5143 status = be_open(netdev);
5148 schedule_delayed_work(&adapter->func_recovery_work,
5149 msecs_to_jiffies(1000));
5150 netif_device_attach(netdev);
5153 dev_err(&adapter->pdev->dev, "EEH resume failed\n");
5156 static const struct pci_error_handlers be_eeh_handlers = {
5157 .error_detected = be_eeh_err_detected,
5158 .slot_reset = be_eeh_reset,
5159 .resume = be_eeh_resume,
5162 static struct pci_driver be_driver = {
5164 .id_table = be_dev_ids,
5166 .remove = be_remove,
5167 .suspend = be_suspend,
5168 .resume = be_resume,
5169 .shutdown = be_shutdown,
5170 .err_handler = &be_eeh_handlers
5173 static int __init be_init_module(void)
5175 if (rx_frag_size != 8192 && rx_frag_size != 4096 &&
5176 rx_frag_size != 2048) {
5177 printk(KERN_WARNING DRV_NAME
5178 " : Module param rx_frag_size must be 2048/4096/8192."
5180 rx_frag_size = 2048;
5183 return pci_register_driver(&be_driver);
5185 module_init(be_init_module);
5187 static void __exit be_exit_module(void)
5189 pci_unregister_driver(&be_driver);
5191 module_exit(be_exit_module);