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[] = {
90 /* UE Status High CSR */
91 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;
131 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
137 static int be_queue_alloc(struct be_adapter *adapter, struct be_queue_info *q,
138 u16 len, u16 entry_size)
140 struct be_dma_mem *mem = &q->dma_mem;
142 memset(q, 0, sizeof(*q));
144 q->entry_size = entry_size;
145 mem->size = len * entry_size;
146 mem->va = dma_zalloc_coherent(&adapter->pdev->dev, mem->size, &mem->dma,
153 static void be_reg_intr_set(struct be_adapter *adapter, bool enable)
157 pci_read_config_dword(adapter->pdev, PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET,
159 enabled = reg & MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
161 if (!enabled && enable)
162 reg |= MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
163 else if (enabled && !enable)
164 reg &= ~MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
168 pci_write_config_dword(adapter->pdev,
169 PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET, reg);
172 static void be_intr_set(struct be_adapter *adapter, bool enable)
176 /* On lancer interrupts can't be controlled via this register */
177 if (lancer_chip(adapter))
180 if (adapter->eeh_error)
183 status = be_cmd_intr_set(adapter, enable);
185 be_reg_intr_set(adapter, enable);
188 static void be_rxq_notify(struct be_adapter *adapter, u16 qid, u16 posted)
192 val |= qid & DB_RQ_RING_ID_MASK;
193 val |= posted << DB_RQ_NUM_POSTED_SHIFT;
196 iowrite32(val, adapter->db + DB_RQ_OFFSET);
199 static void be_txq_notify(struct be_adapter *adapter, struct be_tx_obj *txo,
204 val |= txo->q.id & DB_TXULP_RING_ID_MASK;
205 val |= (posted & DB_TXULP_NUM_POSTED_MASK) << DB_TXULP_NUM_POSTED_SHIFT;
208 iowrite32(val, adapter->db + txo->db_offset);
211 static void be_eq_notify(struct be_adapter *adapter, u16 qid,
212 bool arm, bool clear_int, u16 num_popped)
216 val |= qid & DB_EQ_RING_ID_MASK;
217 val |= ((qid & DB_EQ_RING_ID_EXT_MASK) << DB_EQ_RING_ID_EXT_MASK_SHIFT);
219 if (adapter->eeh_error)
223 val |= 1 << DB_EQ_REARM_SHIFT;
225 val |= 1 << DB_EQ_CLR_SHIFT;
226 val |= 1 << DB_EQ_EVNT_SHIFT;
227 val |= num_popped << DB_EQ_NUM_POPPED_SHIFT;
228 iowrite32(val, adapter->db + DB_EQ_OFFSET);
231 void be_cq_notify(struct be_adapter *adapter, u16 qid, bool arm, u16 num_popped)
235 val |= qid & DB_CQ_RING_ID_MASK;
236 val |= ((qid & DB_CQ_RING_ID_EXT_MASK) <<
237 DB_CQ_RING_ID_EXT_MASK_SHIFT);
239 if (adapter->eeh_error)
243 val |= 1 << DB_CQ_REARM_SHIFT;
244 val |= num_popped << DB_CQ_NUM_POPPED_SHIFT;
245 iowrite32(val, adapter->db + DB_CQ_OFFSET);
248 static int be_mac_addr_set(struct net_device *netdev, void *p)
250 struct be_adapter *adapter = netdev_priv(netdev);
251 struct device *dev = &adapter->pdev->dev;
252 struct sockaddr *addr = p;
255 u32 old_pmac_id = adapter->pmac_id[0], curr_pmac_id = 0;
257 if (!is_valid_ether_addr(addr->sa_data))
258 return -EADDRNOTAVAIL;
260 /* Proceed further only if, User provided MAC is different
263 if (ether_addr_equal(addr->sa_data, netdev->dev_addr))
266 /* The PMAC_ADD cmd may fail if the VF doesn't have FILTMGMT
267 * privilege or if PF did not provision the new MAC address.
268 * On BE3, this cmd will always fail if the VF doesn't have the
269 * FILTMGMT privilege. This failure is OK, only if the PF programmed
270 * the MAC for the VF.
272 status = be_cmd_pmac_add(adapter, (u8 *)addr->sa_data,
273 adapter->if_handle, &adapter->pmac_id[0], 0);
275 curr_pmac_id = adapter->pmac_id[0];
277 /* Delete the old programmed MAC. This call may fail if the
278 * old MAC was already deleted by the PF driver.
280 if (adapter->pmac_id[0] != old_pmac_id)
281 be_cmd_pmac_del(adapter, adapter->if_handle,
285 /* Decide if the new MAC is successfully activated only after
288 status = be_cmd_get_active_mac(adapter, curr_pmac_id, mac,
289 adapter->if_handle, true, 0);
293 /* The MAC change did not happen, either due to lack of privilege
294 * or PF didn't pre-provision.
296 if (!ether_addr_equal(addr->sa_data, mac)) {
301 memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
302 dev_info(dev, "MAC address changed to %pM\n", mac);
305 dev_warn(dev, "MAC address change to %pM failed\n", addr->sa_data);
309 /* BE2 supports only v0 cmd */
310 static void *hw_stats_from_cmd(struct be_adapter *adapter)
312 if (BE2_chip(adapter)) {
313 struct be_cmd_resp_get_stats_v0 *cmd = adapter->stats_cmd.va;
315 return &cmd->hw_stats;
316 } else if (BE3_chip(adapter)) {
317 struct be_cmd_resp_get_stats_v1 *cmd = adapter->stats_cmd.va;
319 return &cmd->hw_stats;
321 struct be_cmd_resp_get_stats_v2 *cmd = adapter->stats_cmd.va;
323 return &cmd->hw_stats;
327 /* BE2 supports only v0 cmd */
328 static void *be_erx_stats_from_cmd(struct be_adapter *adapter)
330 if (BE2_chip(adapter)) {
331 struct be_hw_stats_v0 *hw_stats = hw_stats_from_cmd(adapter);
333 return &hw_stats->erx;
334 } else if (BE3_chip(adapter)) {
335 struct be_hw_stats_v1 *hw_stats = hw_stats_from_cmd(adapter);
337 return &hw_stats->erx;
339 struct be_hw_stats_v2 *hw_stats = hw_stats_from_cmd(adapter);
341 return &hw_stats->erx;
345 static void populate_be_v0_stats(struct be_adapter *adapter)
347 struct be_hw_stats_v0 *hw_stats = hw_stats_from_cmd(adapter);
348 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
349 struct be_rxf_stats_v0 *rxf_stats = &hw_stats->rxf;
350 struct be_port_rxf_stats_v0 *port_stats =
351 &rxf_stats->port[adapter->port_num];
352 struct be_drv_stats *drvs = &adapter->drv_stats;
354 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
355 drvs->rx_pause_frames = port_stats->rx_pause_frames;
356 drvs->rx_crc_errors = port_stats->rx_crc_errors;
357 drvs->rx_control_frames = port_stats->rx_control_frames;
358 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
359 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
360 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
361 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
362 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
363 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
364 drvs->rxpp_fifo_overflow_drop = port_stats->rx_fifo_overflow;
365 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
366 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
367 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
368 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
369 drvs->rx_input_fifo_overflow_drop = port_stats->rx_input_fifo_overflow;
370 drvs->rx_dropped_header_too_small =
371 port_stats->rx_dropped_header_too_small;
372 drvs->rx_address_filtered =
373 port_stats->rx_address_filtered +
374 port_stats->rx_vlan_filtered;
375 drvs->rx_alignment_symbol_errors =
376 port_stats->rx_alignment_symbol_errors;
378 drvs->tx_pauseframes = port_stats->tx_pauseframes;
379 drvs->tx_controlframes = port_stats->tx_controlframes;
381 if (adapter->port_num)
382 drvs->jabber_events = rxf_stats->port1_jabber_events;
384 drvs->jabber_events = rxf_stats->port0_jabber_events;
385 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
386 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
387 drvs->forwarded_packets = rxf_stats->forwarded_packets;
388 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
389 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
390 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
391 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
394 static void populate_be_v1_stats(struct be_adapter *adapter)
396 struct be_hw_stats_v1 *hw_stats = hw_stats_from_cmd(adapter);
397 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
398 struct be_rxf_stats_v1 *rxf_stats = &hw_stats->rxf;
399 struct be_port_rxf_stats_v1 *port_stats =
400 &rxf_stats->port[adapter->port_num];
401 struct be_drv_stats *drvs = &adapter->drv_stats;
403 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
404 drvs->pmem_fifo_overflow_drop = port_stats->pmem_fifo_overflow_drop;
405 drvs->rx_priority_pause_frames = port_stats->rx_priority_pause_frames;
406 drvs->rx_pause_frames = port_stats->rx_pause_frames;
407 drvs->rx_crc_errors = port_stats->rx_crc_errors;
408 drvs->rx_control_frames = port_stats->rx_control_frames;
409 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
410 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
411 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
412 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
413 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
414 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
415 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
416 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
417 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
418 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
419 drvs->rx_dropped_header_too_small =
420 port_stats->rx_dropped_header_too_small;
421 drvs->rx_input_fifo_overflow_drop =
422 port_stats->rx_input_fifo_overflow_drop;
423 drvs->rx_address_filtered = port_stats->rx_address_filtered;
424 drvs->rx_alignment_symbol_errors =
425 port_stats->rx_alignment_symbol_errors;
426 drvs->rxpp_fifo_overflow_drop = port_stats->rxpp_fifo_overflow_drop;
427 drvs->tx_pauseframes = port_stats->tx_pauseframes;
428 drvs->tx_controlframes = port_stats->tx_controlframes;
429 drvs->tx_priority_pauseframes = port_stats->tx_priority_pauseframes;
430 drvs->jabber_events = port_stats->jabber_events;
431 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
432 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
433 drvs->forwarded_packets = rxf_stats->forwarded_packets;
434 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
435 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
436 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
437 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
440 static void populate_be_v2_stats(struct be_adapter *adapter)
442 struct be_hw_stats_v2 *hw_stats = hw_stats_from_cmd(adapter);
443 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
444 struct be_rxf_stats_v2 *rxf_stats = &hw_stats->rxf;
445 struct be_port_rxf_stats_v2 *port_stats =
446 &rxf_stats->port[adapter->port_num];
447 struct be_drv_stats *drvs = &adapter->drv_stats;
449 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
450 drvs->pmem_fifo_overflow_drop = port_stats->pmem_fifo_overflow_drop;
451 drvs->rx_priority_pause_frames = port_stats->rx_priority_pause_frames;
452 drvs->rx_pause_frames = port_stats->rx_pause_frames;
453 drvs->rx_crc_errors = port_stats->rx_crc_errors;
454 drvs->rx_control_frames = port_stats->rx_control_frames;
455 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
456 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
457 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
458 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
459 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
460 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
461 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
462 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
463 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
464 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
465 drvs->rx_dropped_header_too_small =
466 port_stats->rx_dropped_header_too_small;
467 drvs->rx_input_fifo_overflow_drop =
468 port_stats->rx_input_fifo_overflow_drop;
469 drvs->rx_address_filtered = port_stats->rx_address_filtered;
470 drvs->rx_alignment_symbol_errors =
471 port_stats->rx_alignment_symbol_errors;
472 drvs->rxpp_fifo_overflow_drop = port_stats->rxpp_fifo_overflow_drop;
473 drvs->tx_pauseframes = port_stats->tx_pauseframes;
474 drvs->tx_controlframes = port_stats->tx_controlframes;
475 drvs->tx_priority_pauseframes = port_stats->tx_priority_pauseframes;
476 drvs->jabber_events = port_stats->jabber_events;
477 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
478 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
479 drvs->forwarded_packets = rxf_stats->forwarded_packets;
480 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
481 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
482 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
483 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
484 if (be_roce_supported(adapter)) {
485 drvs->rx_roce_bytes_lsd = port_stats->roce_bytes_received_lsd;
486 drvs->rx_roce_bytes_msd = port_stats->roce_bytes_received_msd;
487 drvs->rx_roce_frames = port_stats->roce_frames_received;
488 drvs->roce_drops_crc = port_stats->roce_drops_crc;
489 drvs->roce_drops_payload_len =
490 port_stats->roce_drops_payload_len;
494 static void populate_lancer_stats(struct be_adapter *adapter)
497 struct be_drv_stats *drvs = &adapter->drv_stats;
498 struct lancer_pport_stats *pport_stats = pport_stats_from_cmd(adapter);
500 be_dws_le_to_cpu(pport_stats, sizeof(*pport_stats));
501 drvs->rx_pause_frames = pport_stats->rx_pause_frames_lo;
502 drvs->rx_crc_errors = pport_stats->rx_crc_errors_lo;
503 drvs->rx_control_frames = pport_stats->rx_control_frames_lo;
504 drvs->rx_in_range_errors = pport_stats->rx_in_range_errors;
505 drvs->rx_frame_too_long = pport_stats->rx_frames_too_long_lo;
506 drvs->rx_dropped_runt = pport_stats->rx_dropped_runt;
507 drvs->rx_ip_checksum_errs = pport_stats->rx_ip_checksum_errors;
508 drvs->rx_tcp_checksum_errs = pport_stats->rx_tcp_checksum_errors;
509 drvs->rx_udp_checksum_errs = pport_stats->rx_udp_checksum_errors;
510 drvs->rx_dropped_tcp_length =
511 pport_stats->rx_dropped_invalid_tcp_length;
512 drvs->rx_dropped_too_small = pport_stats->rx_dropped_too_small;
513 drvs->rx_dropped_too_short = pport_stats->rx_dropped_too_short;
514 drvs->rx_out_range_errors = pport_stats->rx_out_of_range_errors;
515 drvs->rx_dropped_header_too_small =
516 pport_stats->rx_dropped_header_too_small;
517 drvs->rx_input_fifo_overflow_drop = pport_stats->rx_fifo_overflow;
518 drvs->rx_address_filtered =
519 pport_stats->rx_address_filtered +
520 pport_stats->rx_vlan_filtered;
521 drvs->rx_alignment_symbol_errors = pport_stats->rx_symbol_errors_lo;
522 drvs->rxpp_fifo_overflow_drop = pport_stats->rx_fifo_overflow;
523 drvs->tx_pauseframes = pport_stats->tx_pause_frames_lo;
524 drvs->tx_controlframes = pport_stats->tx_control_frames_lo;
525 drvs->jabber_events = pport_stats->rx_jabbers;
526 drvs->forwarded_packets = pport_stats->num_forwards_lo;
527 drvs->rx_drops_mtu = pport_stats->rx_drops_mtu_lo;
528 drvs->rx_drops_too_many_frags =
529 pport_stats->rx_drops_too_many_frags_lo;
532 static void accumulate_16bit_val(u32 *acc, u16 val)
534 #define lo(x) (x & 0xFFFF)
535 #define hi(x) (x & 0xFFFF0000)
536 bool wrapped = val < lo(*acc);
537 u32 newacc = hi(*acc) + val;
541 ACCESS_ONCE(*acc) = newacc;
544 static void populate_erx_stats(struct be_adapter *adapter,
545 struct be_rx_obj *rxo, u32 erx_stat)
547 if (!BEx_chip(adapter))
548 rx_stats(rxo)->rx_drops_no_frags = erx_stat;
550 /* below erx HW counter can actually wrap around after
551 * 65535. Driver accumulates a 32-bit value
553 accumulate_16bit_val(&rx_stats(rxo)->rx_drops_no_frags,
557 void be_parse_stats(struct be_adapter *adapter)
559 struct be_erx_stats_v2 *erx = be_erx_stats_from_cmd(adapter);
560 struct be_rx_obj *rxo;
564 if (lancer_chip(adapter)) {
565 populate_lancer_stats(adapter);
567 if (BE2_chip(adapter))
568 populate_be_v0_stats(adapter);
569 else if (BE3_chip(adapter))
571 populate_be_v1_stats(adapter);
573 populate_be_v2_stats(adapter);
575 /* erx_v2 is longer than v0, v1. use v2 for v0, v1 access */
576 for_all_rx_queues(adapter, rxo, i) {
577 erx_stat = erx->rx_drops_no_fragments[rxo->q.id];
578 populate_erx_stats(adapter, rxo, erx_stat);
583 static struct rtnl_link_stats64 *be_get_stats64(struct net_device *netdev,
584 struct rtnl_link_stats64 *stats)
586 struct be_adapter *adapter = netdev_priv(netdev);
587 struct be_drv_stats *drvs = &adapter->drv_stats;
588 struct be_rx_obj *rxo;
589 struct be_tx_obj *txo;
594 for_all_rx_queues(adapter, rxo, i) {
595 const struct be_rx_stats *rx_stats = rx_stats(rxo);
598 start = u64_stats_fetch_begin_irq(&rx_stats->sync);
599 pkts = rx_stats(rxo)->rx_pkts;
600 bytes = rx_stats(rxo)->rx_bytes;
601 } while (u64_stats_fetch_retry_irq(&rx_stats->sync, start));
602 stats->rx_packets += pkts;
603 stats->rx_bytes += bytes;
604 stats->multicast += rx_stats(rxo)->rx_mcast_pkts;
605 stats->rx_dropped += rx_stats(rxo)->rx_drops_no_skbs +
606 rx_stats(rxo)->rx_drops_no_frags;
609 for_all_tx_queues(adapter, txo, i) {
610 const struct be_tx_stats *tx_stats = tx_stats(txo);
613 start = u64_stats_fetch_begin_irq(&tx_stats->sync);
614 pkts = tx_stats(txo)->tx_pkts;
615 bytes = tx_stats(txo)->tx_bytes;
616 } while (u64_stats_fetch_retry_irq(&tx_stats->sync, start));
617 stats->tx_packets += pkts;
618 stats->tx_bytes += bytes;
621 /* bad pkts received */
622 stats->rx_errors = drvs->rx_crc_errors +
623 drvs->rx_alignment_symbol_errors +
624 drvs->rx_in_range_errors +
625 drvs->rx_out_range_errors +
626 drvs->rx_frame_too_long +
627 drvs->rx_dropped_too_small +
628 drvs->rx_dropped_too_short +
629 drvs->rx_dropped_header_too_small +
630 drvs->rx_dropped_tcp_length +
631 drvs->rx_dropped_runt;
633 /* detailed rx errors */
634 stats->rx_length_errors = drvs->rx_in_range_errors +
635 drvs->rx_out_range_errors +
636 drvs->rx_frame_too_long;
638 stats->rx_crc_errors = drvs->rx_crc_errors;
640 /* frame alignment errors */
641 stats->rx_frame_errors = drvs->rx_alignment_symbol_errors;
643 /* receiver fifo overrun */
644 /* drops_no_pbuf is no per i/f, it's per BE card */
645 stats->rx_fifo_errors = drvs->rxpp_fifo_overflow_drop +
646 drvs->rx_input_fifo_overflow_drop +
647 drvs->rx_drops_no_pbuf;
651 void be_link_status_update(struct be_adapter *adapter, u8 link_status)
653 struct net_device *netdev = adapter->netdev;
655 if (!(adapter->flags & BE_FLAGS_LINK_STATUS_INIT)) {
656 netif_carrier_off(netdev);
657 adapter->flags |= BE_FLAGS_LINK_STATUS_INIT;
661 netif_carrier_on(netdev);
663 netif_carrier_off(netdev);
666 static void be_tx_stats_update(struct be_tx_obj *txo,
667 u32 wrb_cnt, u32 copied, u32 gso_segs,
670 struct be_tx_stats *stats = tx_stats(txo);
672 u64_stats_update_begin(&stats->sync);
674 stats->tx_wrbs += wrb_cnt;
675 stats->tx_bytes += copied;
676 stats->tx_pkts += (gso_segs ? gso_segs : 1);
679 u64_stats_update_end(&stats->sync);
682 /* Determine number of WRB entries needed to xmit data in an skb */
683 static u32 wrb_cnt_for_skb(struct be_adapter *adapter, struct sk_buff *skb,
686 int cnt = (skb->len > skb->data_len);
688 cnt += skb_shinfo(skb)->nr_frags;
690 /* to account for hdr wrb */
692 if (lancer_chip(adapter) || !(cnt & 1)) {
695 /* add a dummy to make it an even num */
699 BUG_ON(cnt > BE_MAX_TX_FRAG_COUNT);
703 static inline void wrb_fill(struct be_eth_wrb *wrb, u64 addr, int len)
705 wrb->frag_pa_hi = upper_32_bits(addr);
706 wrb->frag_pa_lo = addr & 0xFFFFFFFF;
707 wrb->frag_len = len & ETH_WRB_FRAG_LEN_MASK;
711 static inline u16 be_get_tx_vlan_tag(struct be_adapter *adapter,
717 vlan_tag = vlan_tx_tag_get(skb);
718 vlan_prio = (vlan_tag & VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
719 /* If vlan priority provided by OS is NOT in available bmap */
720 if (!(adapter->vlan_prio_bmap & (1 << vlan_prio)))
721 vlan_tag = (vlan_tag & ~VLAN_PRIO_MASK) |
722 adapter->recommended_prio;
727 /* Used only for IP tunnel packets */
728 static u16 skb_inner_ip_proto(struct sk_buff *skb)
730 return (inner_ip_hdr(skb)->version == 4) ?
731 inner_ip_hdr(skb)->protocol : inner_ipv6_hdr(skb)->nexthdr;
734 static u16 skb_ip_proto(struct sk_buff *skb)
736 return (ip_hdr(skb)->version == 4) ?
737 ip_hdr(skb)->protocol : ipv6_hdr(skb)->nexthdr;
740 static void wrb_fill_hdr(struct be_adapter *adapter, struct be_eth_hdr_wrb *hdr,
741 struct sk_buff *skb, u32 wrb_cnt, u32 len,
746 memset(hdr, 0, sizeof(*hdr));
748 SET_TX_WRB_HDR_BITS(crc, hdr, 1);
750 if (skb_is_gso(skb)) {
751 SET_TX_WRB_HDR_BITS(lso, hdr, 1);
752 SET_TX_WRB_HDR_BITS(lso_mss, hdr, skb_shinfo(skb)->gso_size);
753 if (skb_is_gso_v6(skb) && !lancer_chip(adapter))
754 SET_TX_WRB_HDR_BITS(lso6, hdr, 1);
755 } else if (skb->ip_summed == CHECKSUM_PARTIAL) {
756 if (skb->encapsulation) {
757 SET_TX_WRB_HDR_BITS(ipcs, hdr, 1);
758 proto = skb_inner_ip_proto(skb);
760 proto = skb_ip_proto(skb);
762 if (proto == IPPROTO_TCP)
763 SET_TX_WRB_HDR_BITS(tcpcs, hdr, 1);
764 else if (proto == IPPROTO_UDP)
765 SET_TX_WRB_HDR_BITS(udpcs, hdr, 1);
768 if (vlan_tx_tag_present(skb)) {
769 SET_TX_WRB_HDR_BITS(vlan, hdr, 1);
770 vlan_tag = be_get_tx_vlan_tag(adapter, skb);
771 SET_TX_WRB_HDR_BITS(vlan_tag, hdr, vlan_tag);
774 /* To skip HW VLAN tagging: evt = 1, compl = 0 */
775 SET_TX_WRB_HDR_BITS(complete, hdr, !skip_hw_vlan);
776 SET_TX_WRB_HDR_BITS(event, hdr, 1);
777 SET_TX_WRB_HDR_BITS(num_wrb, hdr, wrb_cnt);
778 SET_TX_WRB_HDR_BITS(len, hdr, len);
781 static void unmap_tx_frag(struct device *dev, struct be_eth_wrb *wrb,
786 be_dws_le_to_cpu(wrb, sizeof(*wrb));
788 dma = (u64)wrb->frag_pa_hi << 32 | (u64)wrb->frag_pa_lo;
791 dma_unmap_single(dev, dma, wrb->frag_len,
794 dma_unmap_page(dev, dma, wrb->frag_len, DMA_TO_DEVICE);
798 static int make_tx_wrbs(struct be_adapter *adapter, struct be_queue_info *txq,
799 struct sk_buff *skb, u32 wrb_cnt, bool dummy_wrb,
804 struct device *dev = &adapter->pdev->dev;
805 struct sk_buff *first_skb = skb;
806 struct be_eth_wrb *wrb;
807 struct be_eth_hdr_wrb *hdr;
808 bool map_single = false;
811 hdr = queue_head_node(txq);
813 map_head = txq->head;
815 if (skb->len > skb->data_len) {
816 int len = skb_headlen(skb);
818 busaddr = dma_map_single(dev, skb->data, len, DMA_TO_DEVICE);
819 if (dma_mapping_error(dev, busaddr))
822 wrb = queue_head_node(txq);
823 wrb_fill(wrb, busaddr, len);
824 be_dws_cpu_to_le(wrb, sizeof(*wrb));
829 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
830 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i];
832 busaddr = skb_frag_dma_map(dev, frag, 0,
833 skb_frag_size(frag), DMA_TO_DEVICE);
834 if (dma_mapping_error(dev, busaddr))
836 wrb = queue_head_node(txq);
837 wrb_fill(wrb, busaddr, skb_frag_size(frag));
838 be_dws_cpu_to_le(wrb, sizeof(*wrb));
840 copied += skb_frag_size(frag);
844 wrb = queue_head_node(txq);
846 be_dws_cpu_to_le(wrb, sizeof(*wrb));
850 wrb_fill_hdr(adapter, hdr, first_skb, wrb_cnt, copied, skip_hw_vlan);
851 be_dws_cpu_to_le(hdr, sizeof(*hdr));
855 txq->head = map_head;
857 wrb = queue_head_node(txq);
858 unmap_tx_frag(dev, wrb, map_single);
860 copied -= wrb->frag_len;
861 adapter->drv_stats.dma_map_errors++;
867 static struct sk_buff *be_insert_vlan_in_pkt(struct be_adapter *adapter,
873 skb = skb_share_check(skb, GFP_ATOMIC);
877 if (vlan_tx_tag_present(skb))
878 vlan_tag = be_get_tx_vlan_tag(adapter, skb);
880 if (qnq_async_evt_rcvd(adapter) && adapter->pvid) {
882 vlan_tag = adapter->pvid;
883 /* f/w workaround to set skip_hw_vlan = 1, informs the F/W to
884 * skip VLAN insertion
887 *skip_hw_vlan = true;
891 skb = __vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_tag);
897 /* Insert the outer VLAN, if any */
898 if (adapter->qnq_vid) {
899 vlan_tag = adapter->qnq_vid;
900 skb = __vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_tag);
904 *skip_hw_vlan = true;
910 static bool be_ipv6_exthdr_check(struct sk_buff *skb)
912 struct ethhdr *eh = (struct ethhdr *)skb->data;
913 u16 offset = ETH_HLEN;
915 if (eh->h_proto == htons(ETH_P_IPV6)) {
916 struct ipv6hdr *ip6h = (struct ipv6hdr *)(skb->data + offset);
918 offset += sizeof(struct ipv6hdr);
919 if (ip6h->nexthdr != NEXTHDR_TCP &&
920 ip6h->nexthdr != NEXTHDR_UDP) {
921 struct ipv6_opt_hdr *ehdr =
922 (struct ipv6_opt_hdr *) (skb->data + offset);
924 /* offending pkt: 2nd byte following IPv6 hdr is 0xff */
925 if (ehdr->hdrlen == 0xff)
932 static int be_vlan_tag_tx_chk(struct be_adapter *adapter, struct sk_buff *skb)
934 return vlan_tx_tag_present(skb) || adapter->pvid || adapter->qnq_vid;
937 static int be_ipv6_tx_stall_chk(struct be_adapter *adapter, struct sk_buff *skb)
939 return BE3_chip(adapter) && be_ipv6_exthdr_check(skb);
942 static struct sk_buff *be_lancer_xmit_workarounds(struct be_adapter *adapter,
946 struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
947 unsigned int eth_hdr_len;
950 /* For padded packets, BE HW modifies tot_len field in IP header
951 * incorrecly when VLAN tag is inserted by HW.
952 * For padded packets, Lancer computes incorrect checksum.
954 eth_hdr_len = ntohs(skb->protocol) == ETH_P_8021Q ?
955 VLAN_ETH_HLEN : ETH_HLEN;
956 if (skb->len <= 60 &&
957 (lancer_chip(adapter) || vlan_tx_tag_present(skb)) &&
959 ip = (struct iphdr *)ip_hdr(skb);
960 pskb_trim(skb, eth_hdr_len + ntohs(ip->tot_len));
963 /* If vlan tag is already inlined in the packet, skip HW VLAN
964 * tagging in pvid-tagging mode
966 if (be_pvid_tagging_enabled(adapter) &&
967 veh->h_vlan_proto == htons(ETH_P_8021Q))
968 *skip_hw_vlan = true;
970 /* HW has a bug wherein it will calculate CSUM for VLAN
971 * pkts even though it is disabled.
972 * Manually insert VLAN in pkt.
974 if (skb->ip_summed != CHECKSUM_PARTIAL &&
975 vlan_tx_tag_present(skb)) {
976 skb = be_insert_vlan_in_pkt(adapter, skb, skip_hw_vlan);
981 /* HW may lockup when VLAN HW tagging is requested on
982 * certain ipv6 packets. Drop such pkts if the HW workaround to
983 * skip HW tagging is not enabled by FW.
985 if (unlikely(be_ipv6_tx_stall_chk(adapter, skb) &&
986 (adapter->pvid || adapter->qnq_vid) &&
987 !qnq_async_evt_rcvd(adapter)))
990 /* Manual VLAN tag insertion to prevent:
991 * ASIC lockup when the ASIC inserts VLAN tag into
992 * certain ipv6 packets. Insert VLAN tags in driver,
993 * and set event, completion, vlan bits accordingly
996 if (be_ipv6_tx_stall_chk(adapter, skb) &&
997 be_vlan_tag_tx_chk(adapter, skb)) {
998 skb = be_insert_vlan_in_pkt(adapter, skb, skip_hw_vlan);
1005 dev_kfree_skb_any(skb);
1010 static struct sk_buff *be_xmit_workarounds(struct be_adapter *adapter,
1011 struct sk_buff *skb,
1014 /* Lancer, SH-R ASICs have a bug wherein Packets that are 32 bytes or
1015 * less may cause a transmit stall on that port. So the work-around is
1016 * to pad short packets (<= 32 bytes) to a 36-byte length.
1018 if (unlikely(!BEx_chip(adapter) && skb->len <= 32)) {
1019 if (skb_padto(skb, 36))
1024 if (BEx_chip(adapter) || lancer_chip(adapter)) {
1025 skb = be_lancer_xmit_workarounds(adapter, skb, skip_hw_vlan);
1033 static netdev_tx_t be_xmit(struct sk_buff *skb, struct net_device *netdev)
1035 struct be_adapter *adapter = netdev_priv(netdev);
1036 struct be_tx_obj *txo = &adapter->tx_obj[skb_get_queue_mapping(skb)];
1037 struct be_queue_info *txq = &txo->q;
1038 bool dummy_wrb, stopped = false;
1039 u32 wrb_cnt = 0, copied = 0;
1040 bool skip_hw_vlan = false;
1041 u32 start = txq->head;
1043 skb = be_xmit_workarounds(adapter, skb, &skip_hw_vlan);
1045 tx_stats(txo)->tx_drv_drops++;
1046 return NETDEV_TX_OK;
1049 wrb_cnt = wrb_cnt_for_skb(adapter, skb, &dummy_wrb);
1051 copied = make_tx_wrbs(adapter, txq, skb, wrb_cnt, dummy_wrb,
1054 int gso_segs = skb_shinfo(skb)->gso_segs;
1056 /* record the sent skb in the sent_skb table */
1057 BUG_ON(txo->sent_skb_list[start]);
1058 txo->sent_skb_list[start] = skb;
1060 /* Ensure txq has space for the next skb; Else stop the queue
1061 * *BEFORE* ringing the tx doorbell, so that we serialze the
1062 * tx compls of the current transmit which'll wake up the queue
1064 atomic_add(wrb_cnt, &txq->used);
1065 if ((BE_MAX_TX_FRAG_COUNT + atomic_read(&txq->used)) >=
1067 netif_stop_subqueue(netdev, skb_get_queue_mapping(skb));
1071 be_txq_notify(adapter, txo, wrb_cnt);
1073 be_tx_stats_update(txo, wrb_cnt, copied, gso_segs, stopped);
1076 tx_stats(txo)->tx_drv_drops++;
1077 dev_kfree_skb_any(skb);
1079 return NETDEV_TX_OK;
1082 static int be_change_mtu(struct net_device *netdev, int new_mtu)
1084 struct be_adapter *adapter = netdev_priv(netdev);
1085 struct device *dev = &adapter->pdev->dev;
1087 if (new_mtu < BE_MIN_MTU || new_mtu > BE_MAX_MTU) {
1088 dev_info(dev, "MTU must be between %d and %d bytes\n",
1089 BE_MIN_MTU, BE_MAX_MTU);
1093 dev_info(dev, "MTU changed from %d to %d bytes\n",
1094 netdev->mtu, new_mtu);
1095 netdev->mtu = new_mtu;
1100 * A max of 64 (BE_NUM_VLANS_SUPPORTED) vlans can be configured in BE.
1101 * If the user configures more, place BE in vlan promiscuous mode.
1103 static int be_vid_config(struct be_adapter *adapter)
1105 struct device *dev = &adapter->pdev->dev;
1106 u16 vids[BE_NUM_VLANS_SUPPORTED];
1110 /* No need to further configure vids if in promiscuous mode */
1111 if (adapter->promiscuous)
1114 if (adapter->vlans_added > be_max_vlans(adapter))
1115 goto set_vlan_promisc;
1117 /* Construct VLAN Table to give to HW */
1118 for_each_set_bit(i, adapter->vids, VLAN_N_VID)
1119 vids[num++] = cpu_to_le16(i);
1121 status = be_cmd_vlan_config(adapter, adapter->if_handle, vids, num);
1123 /* Set to VLAN promisc mode as setting VLAN filter failed */
1124 if (addl_status(status) ==
1125 MCC_ADDL_STATUS_INSUFFICIENT_RESOURCES)
1126 goto set_vlan_promisc;
1127 dev_err(dev, "Setting HW VLAN filtering failed\n");
1129 if (adapter->flags & BE_FLAGS_VLAN_PROMISC) {
1130 /* hw VLAN filtering re-enabled. */
1131 status = be_cmd_rx_filter(adapter,
1132 BE_FLAGS_VLAN_PROMISC, OFF);
1135 "Disabling VLAN Promiscuous mode\n");
1136 adapter->flags &= ~BE_FLAGS_VLAN_PROMISC;
1144 if (adapter->flags & BE_FLAGS_VLAN_PROMISC)
1147 status = be_cmd_rx_filter(adapter, BE_FLAGS_VLAN_PROMISC, ON);
1149 dev_info(dev, "Enable VLAN Promiscuous mode\n");
1150 adapter->flags |= BE_FLAGS_VLAN_PROMISC;
1152 dev_err(dev, "Failed to enable VLAN Promiscuous mode\n");
1156 static int be_vlan_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
1158 struct be_adapter *adapter = netdev_priv(netdev);
1161 /* Packets with VID 0 are always received by Lancer by default */
1162 if (lancer_chip(adapter) && vid == 0)
1165 if (test_bit(vid, adapter->vids))
1168 set_bit(vid, adapter->vids);
1169 adapter->vlans_added++;
1171 status = be_vid_config(adapter);
1173 adapter->vlans_added--;
1174 clear_bit(vid, adapter->vids);
1180 static int be_vlan_rem_vid(struct net_device *netdev, __be16 proto, u16 vid)
1182 struct be_adapter *adapter = netdev_priv(netdev);
1184 /* Packets with VID 0 are always received by Lancer by default */
1185 if (lancer_chip(adapter) && vid == 0)
1188 clear_bit(vid, adapter->vids);
1189 adapter->vlans_added--;
1191 return be_vid_config(adapter);
1194 static void be_clear_promisc(struct be_adapter *adapter)
1196 adapter->promiscuous = false;
1197 adapter->flags &= ~(BE_FLAGS_VLAN_PROMISC | BE_FLAGS_MCAST_PROMISC);
1199 be_cmd_rx_filter(adapter, IFF_PROMISC, OFF);
1202 static void be_set_rx_mode(struct net_device *netdev)
1204 struct be_adapter *adapter = netdev_priv(netdev);
1207 if (netdev->flags & IFF_PROMISC) {
1208 be_cmd_rx_filter(adapter, IFF_PROMISC, ON);
1209 adapter->promiscuous = true;
1213 /* BE was previously in promiscuous mode; disable it */
1214 if (adapter->promiscuous) {
1215 be_clear_promisc(adapter);
1216 if (adapter->vlans_added)
1217 be_vid_config(adapter);
1220 /* Enable multicast promisc if num configured exceeds what we support */
1221 if (netdev->flags & IFF_ALLMULTI ||
1222 netdev_mc_count(netdev) > be_max_mc(adapter))
1223 goto set_mcast_promisc;
1225 if (netdev_uc_count(netdev) != adapter->uc_macs) {
1226 struct netdev_hw_addr *ha;
1227 int i = 1; /* First slot is claimed by the Primary MAC */
1229 for (; adapter->uc_macs > 0; adapter->uc_macs--, i++) {
1230 be_cmd_pmac_del(adapter, adapter->if_handle,
1231 adapter->pmac_id[i], 0);
1234 if (netdev_uc_count(netdev) > be_max_uc(adapter)) {
1235 be_cmd_rx_filter(adapter, IFF_PROMISC, ON);
1236 adapter->promiscuous = true;
1240 netdev_for_each_uc_addr(ha, adapter->netdev) {
1241 adapter->uc_macs++; /* First slot is for Primary MAC */
1242 be_cmd_pmac_add(adapter, (u8 *)ha->addr,
1244 &adapter->pmac_id[adapter->uc_macs], 0);
1248 status = be_cmd_rx_filter(adapter, IFF_MULTICAST, ON);
1250 if (adapter->flags & BE_FLAGS_MCAST_PROMISC)
1251 adapter->flags &= ~BE_FLAGS_MCAST_PROMISC;
1256 if (adapter->flags & BE_FLAGS_MCAST_PROMISC)
1259 /* Set to MCAST promisc mode if setting MULTICAST address fails
1260 * or if num configured exceeds what we support
1262 status = be_cmd_rx_filter(adapter, IFF_ALLMULTI, ON);
1264 adapter->flags |= BE_FLAGS_MCAST_PROMISC;
1269 static int be_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
1271 struct be_adapter *adapter = netdev_priv(netdev);
1272 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1275 if (!sriov_enabled(adapter))
1278 if (!is_valid_ether_addr(mac) || vf >= adapter->num_vfs)
1281 /* Proceed further only if user provided MAC is different
1284 if (ether_addr_equal(mac, vf_cfg->mac_addr))
1287 if (BEx_chip(adapter)) {
1288 be_cmd_pmac_del(adapter, vf_cfg->if_handle, vf_cfg->pmac_id,
1291 status = be_cmd_pmac_add(adapter, mac, vf_cfg->if_handle,
1292 &vf_cfg->pmac_id, vf + 1);
1294 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
1299 dev_err(&adapter->pdev->dev, "MAC %pM set on VF %d Failed: %#x",
1301 return be_cmd_status(status);
1304 ether_addr_copy(vf_cfg->mac_addr, mac);
1309 static int be_get_vf_config(struct net_device *netdev, int vf,
1310 struct ifla_vf_info *vi)
1312 struct be_adapter *adapter = netdev_priv(netdev);
1313 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1315 if (!sriov_enabled(adapter))
1318 if (vf >= adapter->num_vfs)
1322 vi->max_tx_rate = vf_cfg->tx_rate;
1323 vi->min_tx_rate = 0;
1324 vi->vlan = vf_cfg->vlan_tag & VLAN_VID_MASK;
1325 vi->qos = vf_cfg->vlan_tag >> VLAN_PRIO_SHIFT;
1326 memcpy(&vi->mac, vf_cfg->mac_addr, ETH_ALEN);
1327 vi->linkstate = adapter->vf_cfg[vf].plink_tracking;
1332 static int be_set_vf_vlan(struct net_device *netdev, int vf, u16 vlan, u8 qos)
1334 struct be_adapter *adapter = netdev_priv(netdev);
1335 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1338 if (!sriov_enabled(adapter))
1341 if (vf >= adapter->num_vfs || vlan > 4095 || qos > 7)
1345 vlan |= qos << VLAN_PRIO_SHIFT;
1346 if (vf_cfg->vlan_tag != vlan)
1347 status = be_cmd_set_hsw_config(adapter, vlan, vf + 1,
1348 vf_cfg->if_handle, 0);
1350 /* Reset Transparent Vlan Tagging. */
1351 status = be_cmd_set_hsw_config(adapter, BE_RESET_VLAN_TAG_ID,
1352 vf + 1, vf_cfg->if_handle, 0);
1356 dev_err(&adapter->pdev->dev,
1357 "VLAN %d config on VF %d failed : %#x\n", vlan,
1359 return be_cmd_status(status);
1362 vf_cfg->vlan_tag = vlan;
1367 static int be_set_vf_tx_rate(struct net_device *netdev, int vf,
1368 int min_tx_rate, int max_tx_rate)
1370 struct be_adapter *adapter = netdev_priv(netdev);
1371 struct device *dev = &adapter->pdev->dev;
1372 int percent_rate, status = 0;
1376 if (!sriov_enabled(adapter))
1379 if (vf >= adapter->num_vfs)
1388 status = be_cmd_link_status_query(adapter, &link_speed,
1394 dev_err(dev, "TX-rate setting not allowed when link is down\n");
1399 if (max_tx_rate < 100 || max_tx_rate > link_speed) {
1400 dev_err(dev, "TX-rate must be between 100 and %d Mbps\n",
1406 /* On Skyhawk the QOS setting must be done only as a % value */
1407 percent_rate = link_speed / 100;
1408 if (skyhawk_chip(adapter) && (max_tx_rate % percent_rate)) {
1409 dev_err(dev, "TX-rate must be a multiple of %d Mbps\n",
1416 status = be_cmd_config_qos(adapter, max_tx_rate, link_speed, vf + 1);
1420 adapter->vf_cfg[vf].tx_rate = max_tx_rate;
1424 dev_err(dev, "TX-rate setting of %dMbps on VF%d failed\n",
1426 return be_cmd_status(status);
1429 static int be_set_vf_link_state(struct net_device *netdev, int vf,
1432 struct be_adapter *adapter = netdev_priv(netdev);
1435 if (!sriov_enabled(adapter))
1438 if (vf >= adapter->num_vfs)
1441 status = be_cmd_set_logical_link_config(adapter, link_state, vf+1);
1443 dev_err(&adapter->pdev->dev,
1444 "Link state change on VF %d failed: %#x\n", vf, status);
1445 return be_cmd_status(status);
1448 adapter->vf_cfg[vf].plink_tracking = link_state;
1453 static void be_aic_update(struct be_aic_obj *aic, u64 rx_pkts, u64 tx_pkts,
1456 aic->rx_pkts_prev = rx_pkts;
1457 aic->tx_reqs_prev = tx_pkts;
1461 static void be_eqd_update(struct be_adapter *adapter)
1463 struct be_set_eqd set_eqd[MAX_EVT_QS];
1464 int eqd, i, num = 0, start;
1465 struct be_aic_obj *aic;
1466 struct be_eq_obj *eqo;
1467 struct be_rx_obj *rxo;
1468 struct be_tx_obj *txo;
1469 u64 rx_pkts, tx_pkts;
1473 for_all_evt_queues(adapter, eqo, i) {
1474 aic = &adapter->aic_obj[eqo->idx];
1482 rxo = &adapter->rx_obj[eqo->idx];
1484 start = u64_stats_fetch_begin_irq(&rxo->stats.sync);
1485 rx_pkts = rxo->stats.rx_pkts;
1486 } while (u64_stats_fetch_retry_irq(&rxo->stats.sync, start));
1488 txo = &adapter->tx_obj[eqo->idx];
1490 start = u64_stats_fetch_begin_irq(&txo->stats.sync);
1491 tx_pkts = txo->stats.tx_reqs;
1492 } while (u64_stats_fetch_retry_irq(&txo->stats.sync, start));
1494 /* Skip, if wrapped around or first calculation */
1496 if (!aic->jiffies || time_before(now, aic->jiffies) ||
1497 rx_pkts < aic->rx_pkts_prev ||
1498 tx_pkts < aic->tx_reqs_prev) {
1499 be_aic_update(aic, rx_pkts, tx_pkts, now);
1503 delta = jiffies_to_msecs(now - aic->jiffies);
1504 pps = (((u32)(rx_pkts - aic->rx_pkts_prev) * 1000) / delta) +
1505 (((u32)(tx_pkts - aic->tx_reqs_prev) * 1000) / delta);
1506 eqd = (pps / 15000) << 2;
1510 eqd = min_t(u32, eqd, aic->max_eqd);
1511 eqd = max_t(u32, eqd, aic->min_eqd);
1513 be_aic_update(aic, rx_pkts, tx_pkts, now);
1515 if (eqd != aic->prev_eqd) {
1516 set_eqd[num].delay_multiplier = (eqd * 65)/100;
1517 set_eqd[num].eq_id = eqo->q.id;
1518 aic->prev_eqd = eqd;
1524 be_cmd_modify_eqd(adapter, set_eqd, num);
1527 static void be_rx_stats_update(struct be_rx_obj *rxo,
1528 struct be_rx_compl_info *rxcp)
1530 struct be_rx_stats *stats = rx_stats(rxo);
1532 u64_stats_update_begin(&stats->sync);
1534 stats->rx_bytes += rxcp->pkt_size;
1536 if (rxcp->pkt_type == BE_MULTICAST_PACKET)
1537 stats->rx_mcast_pkts++;
1539 stats->rx_compl_err++;
1540 u64_stats_update_end(&stats->sync);
1543 static inline bool csum_passed(struct be_rx_compl_info *rxcp)
1545 /* L4 checksum is not reliable for non TCP/UDP packets.
1546 * Also ignore ipcksm for ipv6 pkts
1548 return (rxcp->tcpf || rxcp->udpf) && rxcp->l4_csum &&
1549 (rxcp->ip_csum || rxcp->ipv6) && !rxcp->err;
1552 static struct be_rx_page_info *get_rx_page_info(struct be_rx_obj *rxo)
1554 struct be_adapter *adapter = rxo->adapter;
1555 struct be_rx_page_info *rx_page_info;
1556 struct be_queue_info *rxq = &rxo->q;
1557 u16 frag_idx = rxq->tail;
1559 rx_page_info = &rxo->page_info_tbl[frag_idx];
1560 BUG_ON(!rx_page_info->page);
1562 if (rx_page_info->last_frag) {
1563 dma_unmap_page(&adapter->pdev->dev,
1564 dma_unmap_addr(rx_page_info, bus),
1565 adapter->big_page_size, DMA_FROM_DEVICE);
1566 rx_page_info->last_frag = false;
1568 dma_sync_single_for_cpu(&adapter->pdev->dev,
1569 dma_unmap_addr(rx_page_info, bus),
1570 rx_frag_size, DMA_FROM_DEVICE);
1573 queue_tail_inc(rxq);
1574 atomic_dec(&rxq->used);
1575 return rx_page_info;
1578 /* Throwaway the data in the Rx completion */
1579 static void be_rx_compl_discard(struct be_rx_obj *rxo,
1580 struct be_rx_compl_info *rxcp)
1582 struct be_rx_page_info *page_info;
1583 u16 i, num_rcvd = rxcp->num_rcvd;
1585 for (i = 0; i < num_rcvd; i++) {
1586 page_info = get_rx_page_info(rxo);
1587 put_page(page_info->page);
1588 memset(page_info, 0, sizeof(*page_info));
1593 * skb_fill_rx_data forms a complete skb for an ether frame
1594 * indicated by rxcp.
1596 static void skb_fill_rx_data(struct be_rx_obj *rxo, struct sk_buff *skb,
1597 struct be_rx_compl_info *rxcp)
1599 struct be_rx_page_info *page_info;
1601 u16 hdr_len, curr_frag_len, remaining;
1604 page_info = get_rx_page_info(rxo);
1605 start = page_address(page_info->page) + page_info->page_offset;
1608 /* Copy data in the first descriptor of this completion */
1609 curr_frag_len = min(rxcp->pkt_size, rx_frag_size);
1611 skb->len = curr_frag_len;
1612 if (curr_frag_len <= BE_HDR_LEN) { /* tiny packet */
1613 memcpy(skb->data, start, curr_frag_len);
1614 /* Complete packet has now been moved to data */
1615 put_page(page_info->page);
1617 skb->tail += curr_frag_len;
1620 memcpy(skb->data, start, hdr_len);
1621 skb_shinfo(skb)->nr_frags = 1;
1622 skb_frag_set_page(skb, 0, page_info->page);
1623 skb_shinfo(skb)->frags[0].page_offset =
1624 page_info->page_offset + hdr_len;
1625 skb_frag_size_set(&skb_shinfo(skb)->frags[0],
1626 curr_frag_len - hdr_len);
1627 skb->data_len = curr_frag_len - hdr_len;
1628 skb->truesize += rx_frag_size;
1629 skb->tail += hdr_len;
1631 page_info->page = NULL;
1633 if (rxcp->pkt_size <= rx_frag_size) {
1634 BUG_ON(rxcp->num_rcvd != 1);
1638 /* More frags present for this completion */
1639 remaining = rxcp->pkt_size - curr_frag_len;
1640 for (i = 1, j = 0; i < rxcp->num_rcvd; i++) {
1641 page_info = get_rx_page_info(rxo);
1642 curr_frag_len = min(remaining, rx_frag_size);
1644 /* Coalesce all frags from the same physical page in one slot */
1645 if (page_info->page_offset == 0) {
1648 skb_frag_set_page(skb, j, page_info->page);
1649 skb_shinfo(skb)->frags[j].page_offset =
1650 page_info->page_offset;
1651 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1652 skb_shinfo(skb)->nr_frags++;
1654 put_page(page_info->page);
1657 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1658 skb->len += curr_frag_len;
1659 skb->data_len += curr_frag_len;
1660 skb->truesize += rx_frag_size;
1661 remaining -= curr_frag_len;
1662 page_info->page = NULL;
1664 BUG_ON(j > MAX_SKB_FRAGS);
1667 /* Process the RX completion indicated by rxcp when GRO is disabled */
1668 static void be_rx_compl_process(struct be_rx_obj *rxo, struct napi_struct *napi,
1669 struct be_rx_compl_info *rxcp)
1671 struct be_adapter *adapter = rxo->adapter;
1672 struct net_device *netdev = adapter->netdev;
1673 struct sk_buff *skb;
1675 skb = netdev_alloc_skb_ip_align(netdev, BE_RX_SKB_ALLOC_SIZE);
1676 if (unlikely(!skb)) {
1677 rx_stats(rxo)->rx_drops_no_skbs++;
1678 be_rx_compl_discard(rxo, rxcp);
1682 skb_fill_rx_data(rxo, skb, rxcp);
1684 if (likely((netdev->features & NETIF_F_RXCSUM) && csum_passed(rxcp)))
1685 skb->ip_summed = CHECKSUM_UNNECESSARY;
1687 skb_checksum_none_assert(skb);
1689 skb->protocol = eth_type_trans(skb, netdev);
1690 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1691 if (netdev->features & NETIF_F_RXHASH)
1692 skb_set_hash(skb, rxcp->rss_hash, PKT_HASH_TYPE_L3);
1694 skb->csum_level = rxcp->tunneled;
1695 skb_mark_napi_id(skb, napi);
1698 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1700 netif_receive_skb(skb);
1703 /* Process the RX completion indicated by rxcp when GRO is enabled */
1704 static void be_rx_compl_process_gro(struct be_rx_obj *rxo,
1705 struct napi_struct *napi,
1706 struct be_rx_compl_info *rxcp)
1708 struct be_adapter *adapter = rxo->adapter;
1709 struct be_rx_page_info *page_info;
1710 struct sk_buff *skb = NULL;
1711 u16 remaining, curr_frag_len;
1714 skb = napi_get_frags(napi);
1716 be_rx_compl_discard(rxo, rxcp);
1720 remaining = rxcp->pkt_size;
1721 for (i = 0, j = -1; i < rxcp->num_rcvd; i++) {
1722 page_info = get_rx_page_info(rxo);
1724 curr_frag_len = min(remaining, rx_frag_size);
1726 /* Coalesce all frags from the same physical page in one slot */
1727 if (i == 0 || page_info->page_offset == 0) {
1728 /* First frag or Fresh page */
1730 skb_frag_set_page(skb, j, page_info->page);
1731 skb_shinfo(skb)->frags[j].page_offset =
1732 page_info->page_offset;
1733 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1735 put_page(page_info->page);
1737 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1738 skb->truesize += rx_frag_size;
1739 remaining -= curr_frag_len;
1740 memset(page_info, 0, sizeof(*page_info));
1742 BUG_ON(j > MAX_SKB_FRAGS);
1744 skb_shinfo(skb)->nr_frags = j + 1;
1745 skb->len = rxcp->pkt_size;
1746 skb->data_len = rxcp->pkt_size;
1747 skb->ip_summed = CHECKSUM_UNNECESSARY;
1748 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1749 if (adapter->netdev->features & NETIF_F_RXHASH)
1750 skb_set_hash(skb, rxcp->rss_hash, PKT_HASH_TYPE_L3);
1752 skb->csum_level = rxcp->tunneled;
1753 skb_mark_napi_id(skb, napi);
1756 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1758 napi_gro_frags(napi);
1761 static void be_parse_rx_compl_v1(struct be_eth_rx_compl *compl,
1762 struct be_rx_compl_info *rxcp)
1764 rxcp->pkt_size = GET_RX_COMPL_V1_BITS(pktsize, compl);
1765 rxcp->vlanf = GET_RX_COMPL_V1_BITS(vtp, compl);
1766 rxcp->err = GET_RX_COMPL_V1_BITS(err, compl);
1767 rxcp->tcpf = GET_RX_COMPL_V1_BITS(tcpf, compl);
1768 rxcp->udpf = GET_RX_COMPL_V1_BITS(udpf, compl);
1769 rxcp->ip_csum = GET_RX_COMPL_V1_BITS(ipcksm, compl);
1770 rxcp->l4_csum = GET_RX_COMPL_V1_BITS(l4_cksm, compl);
1771 rxcp->ipv6 = GET_RX_COMPL_V1_BITS(ip_version, compl);
1772 rxcp->num_rcvd = GET_RX_COMPL_V1_BITS(numfrags, compl);
1773 rxcp->pkt_type = GET_RX_COMPL_V1_BITS(cast_enc, compl);
1774 rxcp->rss_hash = GET_RX_COMPL_V1_BITS(rsshash, compl);
1776 rxcp->qnq = GET_RX_COMPL_V1_BITS(qnq, compl);
1777 rxcp->vlan_tag = GET_RX_COMPL_V1_BITS(vlan_tag, compl);
1779 rxcp->port = GET_RX_COMPL_V1_BITS(port, compl);
1781 GET_RX_COMPL_V1_BITS(tunneled, compl);
1784 static void be_parse_rx_compl_v0(struct be_eth_rx_compl *compl,
1785 struct be_rx_compl_info *rxcp)
1787 rxcp->pkt_size = GET_RX_COMPL_V0_BITS(pktsize, compl);
1788 rxcp->vlanf = GET_RX_COMPL_V0_BITS(vtp, compl);
1789 rxcp->err = GET_RX_COMPL_V0_BITS(err, compl);
1790 rxcp->tcpf = GET_RX_COMPL_V0_BITS(tcpf, compl);
1791 rxcp->udpf = GET_RX_COMPL_V0_BITS(udpf, compl);
1792 rxcp->ip_csum = GET_RX_COMPL_V0_BITS(ipcksm, compl);
1793 rxcp->l4_csum = GET_RX_COMPL_V0_BITS(l4_cksm, compl);
1794 rxcp->ipv6 = GET_RX_COMPL_V0_BITS(ip_version, compl);
1795 rxcp->num_rcvd = GET_RX_COMPL_V0_BITS(numfrags, compl);
1796 rxcp->pkt_type = GET_RX_COMPL_V0_BITS(cast_enc, compl);
1797 rxcp->rss_hash = GET_RX_COMPL_V0_BITS(rsshash, compl);
1799 rxcp->qnq = GET_RX_COMPL_V0_BITS(qnq, compl);
1800 rxcp->vlan_tag = GET_RX_COMPL_V0_BITS(vlan_tag, compl);
1802 rxcp->port = GET_RX_COMPL_V0_BITS(port, compl);
1803 rxcp->ip_frag = GET_RX_COMPL_V0_BITS(ip_frag, compl);
1806 static struct be_rx_compl_info *be_rx_compl_get(struct be_rx_obj *rxo)
1808 struct be_eth_rx_compl *compl = queue_tail_node(&rxo->cq);
1809 struct be_rx_compl_info *rxcp = &rxo->rxcp;
1810 struct be_adapter *adapter = rxo->adapter;
1812 /* For checking the valid bit it is Ok to use either definition as the
1813 * valid bit is at the same position in both v0 and v1 Rx compl */
1814 if (compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] == 0)
1818 be_dws_le_to_cpu(compl, sizeof(*compl));
1820 if (adapter->be3_native)
1821 be_parse_rx_compl_v1(compl, rxcp);
1823 be_parse_rx_compl_v0(compl, rxcp);
1829 /* In QNQ modes, if qnq bit is not set, then the packet was
1830 * tagged only with the transparent outer vlan-tag and must
1831 * not be treated as a vlan packet by host
1833 if (be_is_qnq_mode(adapter) && !rxcp->qnq)
1836 if (!lancer_chip(adapter))
1837 rxcp->vlan_tag = swab16(rxcp->vlan_tag);
1839 if (adapter->pvid == (rxcp->vlan_tag & VLAN_VID_MASK) &&
1840 !test_bit(rxcp->vlan_tag, adapter->vids))
1844 /* As the compl has been parsed, reset it; we wont touch it again */
1845 compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] = 0;
1847 queue_tail_inc(&rxo->cq);
1851 static inline struct page *be_alloc_pages(u32 size, gfp_t gfp)
1853 u32 order = get_order(size);
1857 return alloc_pages(gfp, order);
1861 * Allocate a page, split it to fragments of size rx_frag_size and post as
1862 * receive buffers to BE
1864 static void be_post_rx_frags(struct be_rx_obj *rxo, gfp_t gfp, u32 frags_needed)
1866 struct be_adapter *adapter = rxo->adapter;
1867 struct be_rx_page_info *page_info = NULL, *prev_page_info = NULL;
1868 struct be_queue_info *rxq = &rxo->q;
1869 struct page *pagep = NULL;
1870 struct device *dev = &adapter->pdev->dev;
1871 struct be_eth_rx_d *rxd;
1872 u64 page_dmaaddr = 0, frag_dmaaddr;
1873 u32 posted, page_offset = 0, notify = 0;
1875 page_info = &rxo->page_info_tbl[rxq->head];
1876 for (posted = 0; posted < frags_needed && !page_info->page; posted++) {
1878 pagep = be_alloc_pages(adapter->big_page_size, gfp);
1879 if (unlikely(!pagep)) {
1880 rx_stats(rxo)->rx_post_fail++;
1883 page_dmaaddr = dma_map_page(dev, pagep, 0,
1884 adapter->big_page_size,
1886 if (dma_mapping_error(dev, page_dmaaddr)) {
1889 adapter->drv_stats.dma_map_errors++;
1895 page_offset += rx_frag_size;
1897 page_info->page_offset = page_offset;
1898 page_info->page = pagep;
1900 rxd = queue_head_node(rxq);
1901 frag_dmaaddr = page_dmaaddr + page_info->page_offset;
1902 rxd->fragpa_lo = cpu_to_le32(frag_dmaaddr & 0xFFFFFFFF);
1903 rxd->fragpa_hi = cpu_to_le32(upper_32_bits(frag_dmaaddr));
1905 /* Any space left in the current big page for another frag? */
1906 if ((page_offset + rx_frag_size + rx_frag_size) >
1907 adapter->big_page_size) {
1909 page_info->last_frag = true;
1910 dma_unmap_addr_set(page_info, bus, page_dmaaddr);
1912 dma_unmap_addr_set(page_info, bus, frag_dmaaddr);
1915 prev_page_info = page_info;
1916 queue_head_inc(rxq);
1917 page_info = &rxo->page_info_tbl[rxq->head];
1920 /* Mark the last frag of a page when we break out of the above loop
1921 * with no more slots available in the RXQ
1924 prev_page_info->last_frag = true;
1925 dma_unmap_addr_set(prev_page_info, bus, page_dmaaddr);
1929 atomic_add(posted, &rxq->used);
1930 if (rxo->rx_post_starved)
1931 rxo->rx_post_starved = false;
1933 notify = min(256u, posted);
1934 be_rxq_notify(adapter, rxq->id, notify);
1937 } else if (atomic_read(&rxq->used) == 0) {
1938 /* Let be_worker replenish when memory is available */
1939 rxo->rx_post_starved = true;
1943 static struct be_eth_tx_compl *be_tx_compl_get(struct be_queue_info *tx_cq)
1945 struct be_eth_tx_compl *txcp = queue_tail_node(tx_cq);
1947 if (txcp->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] == 0)
1951 be_dws_le_to_cpu(txcp, sizeof(*txcp));
1953 txcp->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] = 0;
1955 queue_tail_inc(tx_cq);
1959 static u16 be_tx_compl_process(struct be_adapter *adapter,
1960 struct be_tx_obj *txo, u16 last_index)
1962 struct be_queue_info *txq = &txo->q;
1963 struct be_eth_wrb *wrb;
1964 struct sk_buff **sent_skbs = txo->sent_skb_list;
1965 struct sk_buff *sent_skb;
1966 u16 cur_index, num_wrbs = 1; /* account for hdr wrb */
1967 bool unmap_skb_hdr = true;
1969 sent_skb = sent_skbs[txq->tail];
1971 sent_skbs[txq->tail] = NULL;
1973 /* skip header wrb */
1974 queue_tail_inc(txq);
1977 cur_index = txq->tail;
1978 wrb = queue_tail_node(txq);
1979 unmap_tx_frag(&adapter->pdev->dev, wrb,
1980 (unmap_skb_hdr && skb_headlen(sent_skb)));
1981 unmap_skb_hdr = false;
1984 queue_tail_inc(txq);
1985 } while (cur_index != last_index);
1987 dev_consume_skb_any(sent_skb);
1991 /* Return the number of events in the event queue */
1992 static inline int events_get(struct be_eq_obj *eqo)
1994 struct be_eq_entry *eqe;
1998 eqe = queue_tail_node(&eqo->q);
2005 queue_tail_inc(&eqo->q);
2011 /* Leaves the EQ is disarmed state */
2012 static void be_eq_clean(struct be_eq_obj *eqo)
2014 int num = events_get(eqo);
2016 be_eq_notify(eqo->adapter, eqo->q.id, false, true, num);
2019 static void be_rx_cq_clean(struct be_rx_obj *rxo)
2021 struct be_rx_page_info *page_info;
2022 struct be_queue_info *rxq = &rxo->q;
2023 struct be_queue_info *rx_cq = &rxo->cq;
2024 struct be_rx_compl_info *rxcp;
2025 struct be_adapter *adapter = rxo->adapter;
2028 /* Consume pending rx completions.
2029 * Wait for the flush completion (identified by zero num_rcvd)
2030 * to arrive. Notify CQ even when there are no more CQ entries
2031 * for HW to flush partially coalesced CQ entries.
2032 * In Lancer, there is no need to wait for flush compl.
2035 rxcp = be_rx_compl_get(rxo);
2037 if (lancer_chip(adapter))
2040 if (flush_wait++ > 10 || be_hw_error(adapter)) {
2041 dev_warn(&adapter->pdev->dev,
2042 "did not receive flush compl\n");
2045 be_cq_notify(adapter, rx_cq->id, true, 0);
2048 be_rx_compl_discard(rxo, rxcp);
2049 be_cq_notify(adapter, rx_cq->id, false, 1);
2050 if (rxcp->num_rcvd == 0)
2055 /* After cleanup, leave the CQ in unarmed state */
2056 be_cq_notify(adapter, rx_cq->id, false, 0);
2058 /* Then free posted rx buffers that were not used */
2059 while (atomic_read(&rxq->used) > 0) {
2060 page_info = get_rx_page_info(rxo);
2061 put_page(page_info->page);
2062 memset(page_info, 0, sizeof(*page_info));
2064 BUG_ON(atomic_read(&rxq->used));
2065 rxq->tail = rxq->head = 0;
2068 static void be_tx_compl_clean(struct be_adapter *adapter)
2070 struct be_tx_obj *txo;
2071 struct be_queue_info *txq;
2072 struct be_eth_tx_compl *txcp;
2073 u16 end_idx, cmpl = 0, timeo = 0, num_wrbs = 0;
2074 struct sk_buff *sent_skb;
2076 int i, pending_txqs;
2078 /* Stop polling for compls when HW has been silent for 10ms */
2080 pending_txqs = adapter->num_tx_qs;
2082 for_all_tx_queues(adapter, txo, i) {
2086 while ((txcp = be_tx_compl_get(&txo->cq))) {
2087 end_idx = GET_TX_COMPL_BITS(wrb_index, txcp);
2088 num_wrbs += be_tx_compl_process(adapter, txo,
2093 be_cq_notify(adapter, txo->cq.id, false, cmpl);
2094 atomic_sub(num_wrbs, &txq->used);
2097 if (atomic_read(&txq->used) == 0)
2101 if (pending_txqs == 0 || ++timeo > 10 || be_hw_error(adapter))
2107 for_all_tx_queues(adapter, txo, i) {
2109 if (atomic_read(&txq->used))
2110 dev_err(&adapter->pdev->dev, "%d pending tx-compls\n",
2111 atomic_read(&txq->used));
2113 /* free posted tx for which compls will never arrive */
2114 while (atomic_read(&txq->used)) {
2115 sent_skb = txo->sent_skb_list[txq->tail];
2116 end_idx = txq->tail;
2117 num_wrbs = wrb_cnt_for_skb(adapter, sent_skb,
2119 index_adv(&end_idx, num_wrbs - 1, txq->len);
2120 num_wrbs = be_tx_compl_process(adapter, txo, end_idx);
2121 atomic_sub(num_wrbs, &txq->used);
2126 static void be_evt_queues_destroy(struct be_adapter *adapter)
2128 struct be_eq_obj *eqo;
2131 for_all_evt_queues(adapter, eqo, i) {
2132 if (eqo->q.created) {
2134 be_cmd_q_destroy(adapter, &eqo->q, QTYPE_EQ);
2135 napi_hash_del(&eqo->napi);
2136 netif_napi_del(&eqo->napi);
2138 be_queue_free(adapter, &eqo->q);
2142 static int be_evt_queues_create(struct be_adapter *adapter)
2144 struct be_queue_info *eq;
2145 struct be_eq_obj *eqo;
2146 struct be_aic_obj *aic;
2149 adapter->num_evt_qs = min_t(u16, num_irqs(adapter),
2150 adapter->cfg_num_qs);
2152 for_all_evt_queues(adapter, eqo, i) {
2153 netif_napi_add(adapter->netdev, &eqo->napi, be_poll,
2155 napi_hash_add(&eqo->napi);
2156 aic = &adapter->aic_obj[i];
2157 eqo->adapter = adapter;
2159 aic->max_eqd = BE_MAX_EQD;
2163 rc = be_queue_alloc(adapter, eq, EVNT_Q_LEN,
2164 sizeof(struct be_eq_entry));
2168 rc = be_cmd_eq_create(adapter, eqo);
2175 static void be_mcc_queues_destroy(struct be_adapter *adapter)
2177 struct be_queue_info *q;
2179 q = &adapter->mcc_obj.q;
2181 be_cmd_q_destroy(adapter, q, QTYPE_MCCQ);
2182 be_queue_free(adapter, q);
2184 q = &adapter->mcc_obj.cq;
2186 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2187 be_queue_free(adapter, q);
2190 /* Must be called only after TX qs are created as MCC shares TX EQ */
2191 static int be_mcc_queues_create(struct be_adapter *adapter)
2193 struct be_queue_info *q, *cq;
2195 cq = &adapter->mcc_obj.cq;
2196 if (be_queue_alloc(adapter, cq, MCC_CQ_LEN,
2197 sizeof(struct be_mcc_compl)))
2200 /* Use the default EQ for MCC completions */
2201 if (be_cmd_cq_create(adapter, cq, &mcc_eqo(adapter)->q, true, 0))
2204 q = &adapter->mcc_obj.q;
2205 if (be_queue_alloc(adapter, q, MCC_Q_LEN, sizeof(struct be_mcc_wrb)))
2206 goto mcc_cq_destroy;
2208 if (be_cmd_mccq_create(adapter, q, cq))
2214 be_queue_free(adapter, q);
2216 be_cmd_q_destroy(adapter, cq, QTYPE_CQ);
2218 be_queue_free(adapter, cq);
2223 static void be_tx_queues_destroy(struct be_adapter *adapter)
2225 struct be_queue_info *q;
2226 struct be_tx_obj *txo;
2229 for_all_tx_queues(adapter, txo, i) {
2232 be_cmd_q_destroy(adapter, q, QTYPE_TXQ);
2233 be_queue_free(adapter, q);
2237 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2238 be_queue_free(adapter, q);
2242 static int be_tx_qs_create(struct be_adapter *adapter)
2244 struct be_queue_info *cq, *eq;
2245 struct be_tx_obj *txo;
2248 adapter->num_tx_qs = min(adapter->num_evt_qs, be_max_txqs(adapter));
2250 for_all_tx_queues(adapter, txo, i) {
2252 status = be_queue_alloc(adapter, cq, TX_CQ_LEN,
2253 sizeof(struct be_eth_tx_compl));
2257 u64_stats_init(&txo->stats.sync);
2258 u64_stats_init(&txo->stats.sync_compl);
2260 /* If num_evt_qs is less than num_tx_qs, then more than
2261 * one txq share an eq
2263 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2264 status = be_cmd_cq_create(adapter, cq, eq, false, 3);
2268 status = be_queue_alloc(adapter, &txo->q, TX_Q_LEN,
2269 sizeof(struct be_eth_wrb));
2273 status = be_cmd_txq_create(adapter, txo);
2278 dev_info(&adapter->pdev->dev, "created %d TX queue(s)\n",
2279 adapter->num_tx_qs);
2283 static void be_rx_cqs_destroy(struct be_adapter *adapter)
2285 struct be_queue_info *q;
2286 struct be_rx_obj *rxo;
2289 for_all_rx_queues(adapter, rxo, i) {
2292 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2293 be_queue_free(adapter, q);
2297 static int be_rx_cqs_create(struct be_adapter *adapter)
2299 struct be_queue_info *eq, *cq;
2300 struct be_rx_obj *rxo;
2303 /* We can create as many RSS rings as there are EQs. */
2304 adapter->num_rx_qs = adapter->num_evt_qs;
2306 /* We'll use RSS only if atleast 2 RSS rings are supported.
2307 * When RSS is used, we'll need a default RXQ for non-IP traffic.
2309 if (adapter->num_rx_qs > 1)
2310 adapter->num_rx_qs++;
2312 adapter->big_page_size = (1 << get_order(rx_frag_size)) * PAGE_SIZE;
2313 for_all_rx_queues(adapter, rxo, i) {
2314 rxo->adapter = adapter;
2316 rc = be_queue_alloc(adapter, cq, RX_CQ_LEN,
2317 sizeof(struct be_eth_rx_compl));
2321 u64_stats_init(&rxo->stats.sync);
2322 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2323 rc = be_cmd_cq_create(adapter, cq, eq, false, 3);
2328 dev_info(&adapter->pdev->dev,
2329 "created %d RSS queue(s) and 1 default RX queue\n",
2330 adapter->num_rx_qs - 1);
2334 static irqreturn_t be_intx(int irq, void *dev)
2336 struct be_eq_obj *eqo = dev;
2337 struct be_adapter *adapter = eqo->adapter;
2340 /* IRQ is not expected when NAPI is scheduled as the EQ
2341 * will not be armed.
2342 * But, this can happen on Lancer INTx where it takes
2343 * a while to de-assert INTx or in BE2 where occasionaly
2344 * an interrupt may be raised even when EQ is unarmed.
2345 * If NAPI is already scheduled, then counting & notifying
2346 * events will orphan them.
2348 if (napi_schedule_prep(&eqo->napi)) {
2349 num_evts = events_get(eqo);
2350 __napi_schedule(&eqo->napi);
2352 eqo->spurious_intr = 0;
2354 be_eq_notify(adapter, eqo->q.id, false, true, num_evts);
2356 /* Return IRQ_HANDLED only for the the first spurious intr
2357 * after a valid intr to stop the kernel from branding
2358 * this irq as a bad one!
2360 if (num_evts || eqo->spurious_intr++ == 0)
2366 static irqreturn_t be_msix(int irq, void *dev)
2368 struct be_eq_obj *eqo = dev;
2370 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
2371 napi_schedule(&eqo->napi);
2375 static inline bool do_gro(struct be_rx_compl_info *rxcp)
2377 return (rxcp->tcpf && !rxcp->err && rxcp->l4_csum) ? true : false;
2380 static int be_process_rx(struct be_rx_obj *rxo, struct napi_struct *napi,
2381 int budget, int polling)
2383 struct be_adapter *adapter = rxo->adapter;
2384 struct be_queue_info *rx_cq = &rxo->cq;
2385 struct be_rx_compl_info *rxcp;
2387 u32 frags_consumed = 0;
2389 for (work_done = 0; work_done < budget; work_done++) {
2390 rxcp = be_rx_compl_get(rxo);
2394 /* Is it a flush compl that has no data */
2395 if (unlikely(rxcp->num_rcvd == 0))
2398 /* Discard compl with partial DMA Lancer B0 */
2399 if (unlikely(!rxcp->pkt_size)) {
2400 be_rx_compl_discard(rxo, rxcp);
2404 /* On BE drop pkts that arrive due to imperfect filtering in
2405 * promiscuous mode on some skews
2407 if (unlikely(rxcp->port != adapter->port_num &&
2408 !lancer_chip(adapter))) {
2409 be_rx_compl_discard(rxo, rxcp);
2413 /* Don't do gro when we're busy_polling */
2414 if (do_gro(rxcp) && polling != BUSY_POLLING)
2415 be_rx_compl_process_gro(rxo, napi, rxcp);
2417 be_rx_compl_process(rxo, napi, rxcp);
2420 frags_consumed += rxcp->num_rcvd;
2421 be_rx_stats_update(rxo, rxcp);
2425 be_cq_notify(adapter, rx_cq->id, true, work_done);
2427 /* When an rx-obj gets into post_starved state, just
2428 * let be_worker do the posting.
2430 if (atomic_read(&rxo->q.used) < RX_FRAGS_REFILL_WM &&
2431 !rxo->rx_post_starved)
2432 be_post_rx_frags(rxo, GFP_ATOMIC,
2433 max_t(u32, MAX_RX_POST,
2440 static inline void be_update_tx_err(struct be_tx_obj *txo, u32 status)
2443 case BE_TX_COMP_HDR_PARSE_ERR:
2444 tx_stats(txo)->tx_hdr_parse_err++;
2446 case BE_TX_COMP_NDMA_ERR:
2447 tx_stats(txo)->tx_dma_err++;
2449 case BE_TX_COMP_ACL_ERR:
2450 tx_stats(txo)->tx_spoof_check_err++;
2455 static inline void lancer_update_tx_err(struct be_tx_obj *txo, u32 status)
2458 case LANCER_TX_COMP_LSO_ERR:
2459 tx_stats(txo)->tx_tso_err++;
2461 case LANCER_TX_COMP_HSW_DROP_MAC_ERR:
2462 case LANCER_TX_COMP_HSW_DROP_VLAN_ERR:
2463 tx_stats(txo)->tx_spoof_check_err++;
2465 case LANCER_TX_COMP_QINQ_ERR:
2466 tx_stats(txo)->tx_qinq_err++;
2468 case LANCER_TX_COMP_PARITY_ERR:
2469 tx_stats(txo)->tx_internal_parity_err++;
2471 case LANCER_TX_COMP_DMA_ERR:
2472 tx_stats(txo)->tx_dma_err++;
2477 static void be_process_tx(struct be_adapter *adapter, struct be_tx_obj *txo,
2480 struct be_eth_tx_compl *txcp;
2481 int num_wrbs = 0, work_done = 0;
2485 while ((txcp = be_tx_compl_get(&txo->cq))) {
2486 last_idx = GET_TX_COMPL_BITS(wrb_index, txcp);
2487 num_wrbs += be_tx_compl_process(adapter, txo, last_idx);
2490 compl_status = GET_TX_COMPL_BITS(status, txcp);
2492 if (lancer_chip(adapter))
2493 lancer_update_tx_err(txo, compl_status);
2495 be_update_tx_err(txo, compl_status);
2500 be_cq_notify(adapter, txo->cq.id, true, work_done);
2501 atomic_sub(num_wrbs, &txo->q.used);
2503 /* As Tx wrbs have been freed up, wake up netdev queue
2504 * if it was stopped due to lack of tx wrbs. */
2505 if (__netif_subqueue_stopped(adapter->netdev, idx) &&
2506 atomic_read(&txo->q.used) < txo->q.len / 2) {
2507 netif_wake_subqueue(adapter->netdev, idx);
2510 u64_stats_update_begin(&tx_stats(txo)->sync_compl);
2511 tx_stats(txo)->tx_compl += work_done;
2512 u64_stats_update_end(&tx_stats(txo)->sync_compl);
2516 int be_poll(struct napi_struct *napi, int budget)
2518 struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2519 struct be_adapter *adapter = eqo->adapter;
2520 int max_work = 0, work, i, num_evts;
2521 struct be_rx_obj *rxo;
2522 struct be_tx_obj *txo;
2524 num_evts = events_get(eqo);
2526 for_all_tx_queues_on_eq(adapter, eqo, txo, i)
2527 be_process_tx(adapter, txo, i);
2529 if (be_lock_napi(eqo)) {
2530 /* This loop will iterate twice for EQ0 in which
2531 * completions of the last RXQ (default one) are also processed
2532 * For other EQs the loop iterates only once
2534 for_all_rx_queues_on_eq(adapter, eqo, rxo, i) {
2535 work = be_process_rx(rxo, napi, budget, NAPI_POLLING);
2536 max_work = max(work, max_work);
2538 be_unlock_napi(eqo);
2543 if (is_mcc_eqo(eqo))
2544 be_process_mcc(adapter);
2546 if (max_work < budget) {
2547 napi_complete(napi);
2548 be_eq_notify(adapter, eqo->q.id, true, false, num_evts);
2550 /* As we'll continue in polling mode, count and clear events */
2551 be_eq_notify(adapter, eqo->q.id, false, false, num_evts);
2556 #ifdef CONFIG_NET_RX_BUSY_POLL
2557 static int be_busy_poll(struct napi_struct *napi)
2559 struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2560 struct be_adapter *adapter = eqo->adapter;
2561 struct be_rx_obj *rxo;
2564 if (!be_lock_busy_poll(eqo))
2565 return LL_FLUSH_BUSY;
2567 for_all_rx_queues_on_eq(adapter, eqo, rxo, i) {
2568 work = be_process_rx(rxo, napi, 4, BUSY_POLLING);
2573 be_unlock_busy_poll(eqo);
2578 void be_detect_error(struct be_adapter *adapter)
2580 u32 ue_lo = 0, ue_hi = 0, ue_lo_mask = 0, ue_hi_mask = 0;
2581 u32 sliport_status = 0, sliport_err1 = 0, sliport_err2 = 0;
2583 bool error_detected = false;
2584 struct device *dev = &adapter->pdev->dev;
2585 struct net_device *netdev = adapter->netdev;
2587 if (be_hw_error(adapter))
2590 if (lancer_chip(adapter)) {
2591 sliport_status = ioread32(adapter->db + SLIPORT_STATUS_OFFSET);
2592 if (sliport_status & SLIPORT_STATUS_ERR_MASK) {
2593 sliport_err1 = ioread32(adapter->db +
2594 SLIPORT_ERROR1_OFFSET);
2595 sliport_err2 = ioread32(adapter->db +
2596 SLIPORT_ERROR2_OFFSET);
2597 adapter->hw_error = true;
2598 /* Do not log error messages if its a FW reset */
2599 if (sliport_err1 == SLIPORT_ERROR_FW_RESET1 &&
2600 sliport_err2 == SLIPORT_ERROR_FW_RESET2) {
2601 dev_info(dev, "Firmware update in progress\n");
2603 error_detected = true;
2604 dev_err(dev, "Error detected in the card\n");
2605 dev_err(dev, "ERR: sliport status 0x%x\n",
2607 dev_err(dev, "ERR: sliport error1 0x%x\n",
2609 dev_err(dev, "ERR: sliport error2 0x%x\n",
2614 pci_read_config_dword(adapter->pdev,
2615 PCICFG_UE_STATUS_LOW, &ue_lo);
2616 pci_read_config_dword(adapter->pdev,
2617 PCICFG_UE_STATUS_HIGH, &ue_hi);
2618 pci_read_config_dword(adapter->pdev,
2619 PCICFG_UE_STATUS_LOW_MASK, &ue_lo_mask);
2620 pci_read_config_dword(adapter->pdev,
2621 PCICFG_UE_STATUS_HI_MASK, &ue_hi_mask);
2623 ue_lo = (ue_lo & ~ue_lo_mask);
2624 ue_hi = (ue_hi & ~ue_hi_mask);
2626 /* On certain platforms BE hardware can indicate spurious UEs.
2627 * Allow HW to stop working completely in case of a real UE.
2628 * Hence not setting the hw_error for UE detection.
2631 if (ue_lo || ue_hi) {
2632 error_detected = true;
2634 "Unrecoverable Error detected in the adapter");
2635 dev_err(dev, "Please reboot server to recover");
2636 if (skyhawk_chip(adapter))
2637 adapter->hw_error = true;
2638 for (i = 0; ue_lo; ue_lo >>= 1, i++) {
2640 dev_err(dev, "UE: %s bit set\n",
2641 ue_status_low_desc[i]);
2643 for (i = 0; ue_hi; ue_hi >>= 1, i++) {
2645 dev_err(dev, "UE: %s bit set\n",
2646 ue_status_hi_desc[i]);
2651 netif_carrier_off(netdev);
2654 static void be_msix_disable(struct be_adapter *adapter)
2656 if (msix_enabled(adapter)) {
2657 pci_disable_msix(adapter->pdev);
2658 adapter->num_msix_vec = 0;
2659 adapter->num_msix_roce_vec = 0;
2663 static int be_msix_enable(struct be_adapter *adapter)
2666 struct device *dev = &adapter->pdev->dev;
2668 /* If RoCE is supported, program the max number of NIC vectors that
2669 * may be configured via set-channels, along with vectors needed for
2670 * RoCe. Else, just program the number we'll use initially.
2672 if (be_roce_supported(adapter))
2673 num_vec = min_t(int, 2 * be_max_eqs(adapter),
2674 2 * num_online_cpus());
2676 num_vec = adapter->cfg_num_qs;
2678 for (i = 0; i < num_vec; i++)
2679 adapter->msix_entries[i].entry = i;
2681 num_vec = pci_enable_msix_range(adapter->pdev, adapter->msix_entries,
2682 MIN_MSIX_VECTORS, num_vec);
2686 if (be_roce_supported(adapter) && num_vec > MIN_MSIX_VECTORS) {
2687 adapter->num_msix_roce_vec = num_vec / 2;
2688 dev_info(dev, "enabled %d MSI-x vector(s) for RoCE\n",
2689 adapter->num_msix_roce_vec);
2692 adapter->num_msix_vec = num_vec - adapter->num_msix_roce_vec;
2694 dev_info(dev, "enabled %d MSI-x vector(s) for NIC\n",
2695 adapter->num_msix_vec);
2699 dev_warn(dev, "MSIx enable failed\n");
2701 /* INTx is not supported in VFs, so fail probe if enable_msix fails */
2702 if (!be_physfn(adapter))
2707 static inline int be_msix_vec_get(struct be_adapter *adapter,
2708 struct be_eq_obj *eqo)
2710 return adapter->msix_entries[eqo->msix_idx].vector;
2713 static int be_msix_register(struct be_adapter *adapter)
2715 struct net_device *netdev = adapter->netdev;
2716 struct be_eq_obj *eqo;
2719 for_all_evt_queues(adapter, eqo, i) {
2720 sprintf(eqo->desc, "%s-q%d", netdev->name, i);
2721 vec = be_msix_vec_get(adapter, eqo);
2722 status = request_irq(vec, be_msix, 0, eqo->desc, eqo);
2729 for (i--, eqo = &adapter->eq_obj[i]; i >= 0; i--, eqo--)
2730 free_irq(be_msix_vec_get(adapter, eqo), eqo);
2731 dev_warn(&adapter->pdev->dev, "MSIX Request IRQ failed - err %d\n",
2733 be_msix_disable(adapter);
2737 static int be_irq_register(struct be_adapter *adapter)
2739 struct net_device *netdev = adapter->netdev;
2742 if (msix_enabled(adapter)) {
2743 status = be_msix_register(adapter);
2746 /* INTx is not supported for VF */
2747 if (!be_physfn(adapter))
2751 /* INTx: only the first EQ is used */
2752 netdev->irq = adapter->pdev->irq;
2753 status = request_irq(netdev->irq, be_intx, IRQF_SHARED, netdev->name,
2754 &adapter->eq_obj[0]);
2756 dev_err(&adapter->pdev->dev,
2757 "INTx request IRQ failed - err %d\n", status);
2761 adapter->isr_registered = true;
2765 static void be_irq_unregister(struct be_adapter *adapter)
2767 struct net_device *netdev = adapter->netdev;
2768 struct be_eq_obj *eqo;
2771 if (!adapter->isr_registered)
2775 if (!msix_enabled(adapter)) {
2776 free_irq(netdev->irq, &adapter->eq_obj[0]);
2781 for_all_evt_queues(adapter, eqo, i)
2782 free_irq(be_msix_vec_get(adapter, eqo), eqo);
2785 adapter->isr_registered = false;
2788 static void be_rx_qs_destroy(struct be_adapter *adapter)
2790 struct be_queue_info *q;
2791 struct be_rx_obj *rxo;
2794 for_all_rx_queues(adapter, rxo, i) {
2797 be_cmd_rxq_destroy(adapter, q);
2798 be_rx_cq_clean(rxo);
2800 be_queue_free(adapter, q);
2804 static int be_close(struct net_device *netdev)
2806 struct be_adapter *adapter = netdev_priv(netdev);
2807 struct be_eq_obj *eqo;
2810 /* This protection is needed as be_close() may be called even when the
2811 * adapter is in cleared state (after eeh perm failure)
2813 if (!(adapter->flags & BE_FLAGS_SETUP_DONE))
2816 be_roce_dev_close(adapter);
2818 if (adapter->flags & BE_FLAGS_NAPI_ENABLED) {
2819 for_all_evt_queues(adapter, eqo, i) {
2820 napi_disable(&eqo->napi);
2821 be_disable_busy_poll(eqo);
2823 adapter->flags &= ~BE_FLAGS_NAPI_ENABLED;
2826 be_async_mcc_disable(adapter);
2828 /* Wait for all pending tx completions to arrive so that
2829 * all tx skbs are freed.
2831 netif_tx_disable(netdev);
2832 be_tx_compl_clean(adapter);
2834 be_rx_qs_destroy(adapter);
2836 for (i = 1; i < (adapter->uc_macs + 1); i++)
2837 be_cmd_pmac_del(adapter, adapter->if_handle,
2838 adapter->pmac_id[i], 0);
2839 adapter->uc_macs = 0;
2841 for_all_evt_queues(adapter, eqo, i) {
2842 if (msix_enabled(adapter))
2843 synchronize_irq(be_msix_vec_get(adapter, eqo));
2845 synchronize_irq(netdev->irq);
2849 be_irq_unregister(adapter);
2854 static int be_rx_qs_create(struct be_adapter *adapter)
2856 struct be_rx_obj *rxo;
2858 u8 rss_hkey[RSS_HASH_KEY_LEN];
2859 struct rss_info *rss = &adapter->rss_info;
2861 for_all_rx_queues(adapter, rxo, i) {
2862 rc = be_queue_alloc(adapter, &rxo->q, RX_Q_LEN,
2863 sizeof(struct be_eth_rx_d));
2868 /* The FW would like the default RXQ to be created first */
2869 rxo = default_rxo(adapter);
2870 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id, rx_frag_size,
2871 adapter->if_handle, false, &rxo->rss_id);
2875 for_all_rss_queues(adapter, rxo, i) {
2876 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id,
2877 rx_frag_size, adapter->if_handle,
2878 true, &rxo->rss_id);
2883 if (be_multi_rxq(adapter)) {
2884 for (j = 0; j < RSS_INDIR_TABLE_LEN;
2885 j += adapter->num_rx_qs - 1) {
2886 for_all_rss_queues(adapter, rxo, i) {
2887 if ((j + i) >= RSS_INDIR_TABLE_LEN)
2889 rss->rsstable[j + i] = rxo->rss_id;
2890 rss->rss_queue[j + i] = i;
2893 rss->rss_flags = RSS_ENABLE_TCP_IPV4 | RSS_ENABLE_IPV4 |
2894 RSS_ENABLE_TCP_IPV6 | RSS_ENABLE_IPV6;
2896 if (!BEx_chip(adapter))
2897 rss->rss_flags |= RSS_ENABLE_UDP_IPV4 |
2898 RSS_ENABLE_UDP_IPV6;
2900 /* Disable RSS, if only default RX Q is created */
2901 rss->rss_flags = RSS_ENABLE_NONE;
2904 get_random_bytes(rss_hkey, RSS_HASH_KEY_LEN);
2905 rc = be_cmd_rss_config(adapter, rss->rsstable, rss->rss_flags,
2908 rss->rss_flags = RSS_ENABLE_NONE;
2912 memcpy(rss->rss_hkey, rss_hkey, RSS_HASH_KEY_LEN);
2914 /* First time posting */
2915 for_all_rx_queues(adapter, rxo, i)
2916 be_post_rx_frags(rxo, GFP_KERNEL, MAX_RX_POST);
2920 static int be_open(struct net_device *netdev)
2922 struct be_adapter *adapter = netdev_priv(netdev);
2923 struct be_eq_obj *eqo;
2924 struct be_rx_obj *rxo;
2925 struct be_tx_obj *txo;
2929 status = be_rx_qs_create(adapter);
2933 status = be_irq_register(adapter);
2937 for_all_rx_queues(adapter, rxo, i)
2938 be_cq_notify(adapter, rxo->cq.id, true, 0);
2940 for_all_tx_queues(adapter, txo, i)
2941 be_cq_notify(adapter, txo->cq.id, true, 0);
2943 be_async_mcc_enable(adapter);
2945 for_all_evt_queues(adapter, eqo, i) {
2946 napi_enable(&eqo->napi);
2947 be_enable_busy_poll(eqo);
2948 be_eq_notify(adapter, eqo->q.id, true, true, 0);
2950 adapter->flags |= BE_FLAGS_NAPI_ENABLED;
2952 status = be_cmd_link_status_query(adapter, NULL, &link_status, 0);
2954 be_link_status_update(adapter, link_status);
2956 netif_tx_start_all_queues(netdev);
2957 be_roce_dev_open(adapter);
2959 #ifdef CONFIG_BE2NET_VXLAN
2960 if (skyhawk_chip(adapter))
2961 vxlan_get_rx_port(netdev);
2966 be_close(adapter->netdev);
2970 static int be_setup_wol(struct be_adapter *adapter, bool enable)
2972 struct be_dma_mem cmd;
2976 memset(mac, 0, ETH_ALEN);
2978 cmd.size = sizeof(struct be_cmd_req_acpi_wol_magic_config);
2979 cmd.va = dma_zalloc_coherent(&adapter->pdev->dev, cmd.size, &cmd.dma,
2985 status = pci_write_config_dword(adapter->pdev,
2986 PCICFG_PM_CONTROL_OFFSET,
2987 PCICFG_PM_CONTROL_MASK);
2989 dev_err(&adapter->pdev->dev,
2990 "Could not enable Wake-on-lan\n");
2991 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va,
2995 status = be_cmd_enable_magic_wol(adapter,
2996 adapter->netdev->dev_addr,
2998 pci_enable_wake(adapter->pdev, PCI_D3hot, 1);
2999 pci_enable_wake(adapter->pdev, PCI_D3cold, 1);
3001 status = be_cmd_enable_magic_wol(adapter, mac, &cmd);
3002 pci_enable_wake(adapter->pdev, PCI_D3hot, 0);
3003 pci_enable_wake(adapter->pdev, PCI_D3cold, 0);
3006 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va, cmd.dma);
3011 * Generate a seed MAC address from the PF MAC Address using jhash.
3012 * MAC Address for VFs are assigned incrementally starting from the seed.
3013 * These addresses are programmed in the ASIC by the PF and the VF driver
3014 * queries for the MAC address during its probe.
3016 static int be_vf_eth_addr_config(struct be_adapter *adapter)
3021 struct be_vf_cfg *vf_cfg;
3023 be_vf_eth_addr_generate(adapter, mac);
3025 for_all_vfs(adapter, vf_cfg, vf) {
3026 if (BEx_chip(adapter))
3027 status = be_cmd_pmac_add(adapter, mac,
3029 &vf_cfg->pmac_id, vf + 1);
3031 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
3035 dev_err(&adapter->pdev->dev,
3036 "Mac address assignment failed for VF %d\n",
3039 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
3046 static int be_vfs_mac_query(struct be_adapter *adapter)
3050 struct be_vf_cfg *vf_cfg;
3052 for_all_vfs(adapter, vf_cfg, vf) {
3053 status = be_cmd_get_active_mac(adapter, vf_cfg->pmac_id,
3054 mac, vf_cfg->if_handle,
3058 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
3063 static void be_vf_clear(struct be_adapter *adapter)
3065 struct be_vf_cfg *vf_cfg;
3068 if (pci_vfs_assigned(adapter->pdev)) {
3069 dev_warn(&adapter->pdev->dev,
3070 "VFs are assigned to VMs: not disabling VFs\n");
3074 pci_disable_sriov(adapter->pdev);
3076 for_all_vfs(adapter, vf_cfg, vf) {
3077 if (BEx_chip(adapter))
3078 be_cmd_pmac_del(adapter, vf_cfg->if_handle,
3079 vf_cfg->pmac_id, vf + 1);
3081 be_cmd_set_mac(adapter, NULL, vf_cfg->if_handle,
3084 be_cmd_if_destroy(adapter, vf_cfg->if_handle, vf + 1);
3087 kfree(adapter->vf_cfg);
3088 adapter->num_vfs = 0;
3089 adapter->flags &= ~BE_FLAGS_SRIOV_ENABLED;
3092 static void be_clear_queues(struct be_adapter *adapter)
3094 be_mcc_queues_destroy(adapter);
3095 be_rx_cqs_destroy(adapter);
3096 be_tx_queues_destroy(adapter);
3097 be_evt_queues_destroy(adapter);
3100 static void be_cancel_worker(struct be_adapter *adapter)
3102 if (adapter->flags & BE_FLAGS_WORKER_SCHEDULED) {
3103 cancel_delayed_work_sync(&adapter->work);
3104 adapter->flags &= ~BE_FLAGS_WORKER_SCHEDULED;
3108 static void be_mac_clear(struct be_adapter *adapter)
3112 if (adapter->pmac_id) {
3113 for (i = 0; i < (adapter->uc_macs + 1); i++)
3114 be_cmd_pmac_del(adapter, adapter->if_handle,
3115 adapter->pmac_id[i], 0);
3116 adapter->uc_macs = 0;
3118 kfree(adapter->pmac_id);
3119 adapter->pmac_id = NULL;
3123 #ifdef CONFIG_BE2NET_VXLAN
3124 static void be_disable_vxlan_offloads(struct be_adapter *adapter)
3126 if (adapter->flags & BE_FLAGS_VXLAN_OFFLOADS)
3127 be_cmd_manage_iface(adapter, adapter->if_handle,
3128 OP_CONVERT_TUNNEL_TO_NORMAL);
3130 if (adapter->vxlan_port)
3131 be_cmd_set_vxlan_port(adapter, 0);
3133 adapter->flags &= ~BE_FLAGS_VXLAN_OFFLOADS;
3134 adapter->vxlan_port = 0;
3138 static int be_clear(struct be_adapter *adapter)
3140 be_cancel_worker(adapter);
3142 if (sriov_enabled(adapter))
3143 be_vf_clear(adapter);
3145 /* Re-configure FW to distribute resources evenly across max-supported
3146 * number of VFs, only when VFs are not already enabled.
3148 if (be_physfn(adapter) && !pci_vfs_assigned(adapter->pdev))
3149 be_cmd_set_sriov_config(adapter, adapter->pool_res,
3150 pci_sriov_get_totalvfs(adapter->pdev));
3152 #ifdef CONFIG_BE2NET_VXLAN
3153 be_disable_vxlan_offloads(adapter);
3155 /* delete the primary mac along with the uc-mac list */
3156 be_mac_clear(adapter);
3158 be_cmd_if_destroy(adapter, adapter->if_handle, 0);
3160 be_clear_queues(adapter);
3162 be_msix_disable(adapter);
3163 adapter->flags &= ~BE_FLAGS_SETUP_DONE;
3167 static int be_vfs_if_create(struct be_adapter *adapter)
3169 struct be_resources res = {0};
3170 struct be_vf_cfg *vf_cfg;
3171 u32 cap_flags, en_flags, vf;
3174 cap_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3175 BE_IF_FLAGS_MULTICAST;
3177 for_all_vfs(adapter, vf_cfg, vf) {
3178 if (!BE3_chip(adapter)) {
3179 status = be_cmd_get_profile_config(adapter, &res,
3182 cap_flags = res.if_cap_flags;
3185 /* If a FW profile exists, then cap_flags are updated */
3186 en_flags = cap_flags & (BE_IF_FLAGS_UNTAGGED |
3187 BE_IF_FLAGS_BROADCAST |
3188 BE_IF_FLAGS_MULTICAST);
3190 be_cmd_if_create(adapter, cap_flags, en_flags,
3191 &vf_cfg->if_handle, vf + 1);
3199 static int be_vf_setup_init(struct be_adapter *adapter)
3201 struct be_vf_cfg *vf_cfg;
3204 adapter->vf_cfg = kcalloc(adapter->num_vfs, sizeof(*vf_cfg),
3206 if (!adapter->vf_cfg)
3209 for_all_vfs(adapter, vf_cfg, vf) {
3210 vf_cfg->if_handle = -1;
3211 vf_cfg->pmac_id = -1;
3216 static int be_vf_setup(struct be_adapter *adapter)
3218 struct device *dev = &adapter->pdev->dev;
3219 struct be_vf_cfg *vf_cfg;
3220 int status, old_vfs, vf;
3223 old_vfs = pci_num_vf(adapter->pdev);
3225 status = be_vf_setup_init(adapter);
3230 for_all_vfs(adapter, vf_cfg, vf) {
3231 status = be_cmd_get_if_id(adapter, vf_cfg, vf);
3236 status = be_vfs_mac_query(adapter);
3240 status = be_vfs_if_create(adapter);
3244 status = be_vf_eth_addr_config(adapter);
3249 for_all_vfs(adapter, vf_cfg, vf) {
3250 /* Allow VFs to programs MAC/VLAN filters */
3251 status = be_cmd_get_fn_privileges(adapter, &privileges, vf + 1);
3252 if (!status && !(privileges & BE_PRIV_FILTMGMT)) {
3253 status = be_cmd_set_fn_privileges(adapter,
3258 dev_info(dev, "VF%d has FILTMGMT privilege\n",
3262 /* Allow full available bandwidth */
3264 be_cmd_config_qos(adapter, 0, 0, vf + 1);
3267 be_cmd_enable_vf(adapter, vf + 1);
3268 be_cmd_set_logical_link_config(adapter,
3269 IFLA_VF_LINK_STATE_AUTO,
3275 status = pci_enable_sriov(adapter->pdev, adapter->num_vfs);
3277 dev_err(dev, "SRIOV enable failed\n");
3278 adapter->num_vfs = 0;
3283 adapter->flags |= BE_FLAGS_SRIOV_ENABLED;
3286 dev_err(dev, "VF setup failed\n");
3287 be_vf_clear(adapter);
3291 /* Converting function_mode bits on BE3 to SH mc_type enums */
3293 static u8 be_convert_mc_type(u32 function_mode)
3295 if (function_mode & VNIC_MODE && function_mode & QNQ_MODE)
3297 else if (function_mode & QNQ_MODE)
3299 else if (function_mode & VNIC_MODE)
3301 else if (function_mode & UMC_ENABLED)
3307 /* On BE2/BE3 FW does not suggest the supported limits */
3308 static void BEx_get_resources(struct be_adapter *adapter,
3309 struct be_resources *res)
3311 bool use_sriov = adapter->num_vfs ? 1 : 0;
3313 if (be_physfn(adapter))
3314 res->max_uc_mac = BE_UC_PMAC_COUNT;
3316 res->max_uc_mac = BE_VF_UC_PMAC_COUNT;
3318 adapter->mc_type = be_convert_mc_type(adapter->function_mode);
3320 if (be_is_mc(adapter)) {
3321 /* Assuming that there are 4 channels per port,
3322 * when multi-channel is enabled
3324 if (be_is_qnq_mode(adapter))
3325 res->max_vlans = BE_NUM_VLANS_SUPPORTED/8;
3327 /* In a non-qnq multichannel mode, the pvid
3328 * takes up one vlan entry
3330 res->max_vlans = (BE_NUM_VLANS_SUPPORTED / 4) - 1;
3332 res->max_vlans = BE_NUM_VLANS_SUPPORTED;
3335 res->max_mcast_mac = BE_MAX_MC;
3337 /* 1) For BE3 1Gb ports, FW does not support multiple TXQs
3338 * 2) Create multiple TX rings on a BE3-R multi-channel interface
3339 * *only* if it is RSS-capable.
3341 if (BE2_chip(adapter) || use_sriov || (adapter->port_num > 1) ||
3342 !be_physfn(adapter) || (be_is_mc(adapter) &&
3343 !(adapter->function_caps & BE_FUNCTION_CAPS_RSS))) {
3345 } else if (adapter->function_caps & BE_FUNCTION_CAPS_SUPER_NIC) {
3346 struct be_resources super_nic_res = {0};
3348 /* On a SuperNIC profile, the driver needs to use the
3349 * GET_PROFILE_CONFIG cmd to query the per-function TXQ limits
3351 be_cmd_get_profile_config(adapter, &super_nic_res, 0);
3352 /* Some old versions of BE3 FW don't report max_tx_qs value */
3353 res->max_tx_qs = super_nic_res.max_tx_qs ? : BE3_MAX_TX_QS;
3355 res->max_tx_qs = BE3_MAX_TX_QS;
3358 if ((adapter->function_caps & BE_FUNCTION_CAPS_RSS) &&
3359 !use_sriov && be_physfn(adapter))
3360 res->max_rss_qs = (adapter->be3_native) ?
3361 BE3_MAX_RSS_QS : BE2_MAX_RSS_QS;
3362 res->max_rx_qs = res->max_rss_qs + 1;
3364 if (be_physfn(adapter))
3365 res->max_evt_qs = (be_max_vfs(adapter) > 0) ?
3366 BE3_SRIOV_MAX_EVT_QS : BE3_MAX_EVT_QS;
3368 res->max_evt_qs = 1;
3370 res->if_cap_flags = BE_IF_CAP_FLAGS_WANT;
3371 if (!(adapter->function_caps & BE_FUNCTION_CAPS_RSS))
3372 res->if_cap_flags &= ~BE_IF_FLAGS_RSS;
3375 static void be_setup_init(struct be_adapter *adapter)
3377 adapter->vlan_prio_bmap = 0xff;
3378 adapter->phy.link_speed = -1;
3379 adapter->if_handle = -1;
3380 adapter->be3_native = false;
3381 adapter->promiscuous = false;
3382 if (be_physfn(adapter))
3383 adapter->cmd_privileges = MAX_PRIVILEGES;
3385 adapter->cmd_privileges = MIN_PRIVILEGES;
3388 static int be_get_sriov_config(struct be_adapter *adapter)
3390 struct device *dev = &adapter->pdev->dev;
3391 struct be_resources res = {0};
3392 int max_vfs, old_vfs;
3394 /* Some old versions of BE3 FW don't report max_vfs value */
3395 be_cmd_get_profile_config(adapter, &res, 0);
3397 if (BE3_chip(adapter) && !res.max_vfs) {
3398 max_vfs = pci_sriov_get_totalvfs(adapter->pdev);
3399 res.max_vfs = max_vfs > 0 ? min(MAX_VFS, max_vfs) : 0;
3402 adapter->pool_res = res;
3404 if (!be_max_vfs(adapter)) {
3406 dev_warn(dev, "SRIOV is disabled. Ignoring num_vfs\n");
3407 adapter->num_vfs = 0;
3411 pci_sriov_set_totalvfs(adapter->pdev, be_max_vfs(adapter));
3413 /* validate num_vfs module param */
3414 old_vfs = pci_num_vf(adapter->pdev);
3416 dev_info(dev, "%d VFs are already enabled\n", old_vfs);
3417 if (old_vfs != num_vfs)
3418 dev_warn(dev, "Ignoring num_vfs=%d setting\n", num_vfs);
3419 adapter->num_vfs = old_vfs;
3421 if (num_vfs > be_max_vfs(adapter)) {
3422 dev_info(dev, "Resources unavailable to init %d VFs\n",
3424 dev_info(dev, "Limiting to %d VFs\n",
3425 be_max_vfs(adapter));
3427 adapter->num_vfs = min_t(u16, num_vfs, be_max_vfs(adapter));
3433 static int be_get_resources(struct be_adapter *adapter)
3435 struct device *dev = &adapter->pdev->dev;
3436 struct be_resources res = {0};
3439 if (BEx_chip(adapter)) {
3440 BEx_get_resources(adapter, &res);
3444 /* For Lancer, SH etc read per-function resource limits from FW.
3445 * GET_FUNC_CONFIG returns per function guaranteed limits.
3446 * GET_PROFILE_CONFIG returns PCI-E related limits PF-pool limits
3448 if (!BEx_chip(adapter)) {
3449 status = be_cmd_get_func_config(adapter, &res);
3453 /* If RoCE may be enabled stash away half the EQs for RoCE */
3454 if (be_roce_supported(adapter))
3455 res.max_evt_qs /= 2;
3459 dev_info(dev, "Max: txqs %d, rxqs %d, rss %d, eqs %d, vfs %d\n",
3460 be_max_txqs(adapter), be_max_rxqs(adapter),
3461 be_max_rss(adapter), be_max_eqs(adapter),
3462 be_max_vfs(adapter));
3463 dev_info(dev, "Max: uc-macs %d, mc-macs %d, vlans %d\n",
3464 be_max_uc(adapter), be_max_mc(adapter),
3465 be_max_vlans(adapter));
3470 static void be_sriov_config(struct be_adapter *adapter)
3472 struct device *dev = &adapter->pdev->dev;
3475 status = be_get_sriov_config(adapter);
3477 dev_err(dev, "Failed to query SR-IOV configuration\n");
3478 dev_err(dev, "SR-IOV cannot be enabled\n");
3482 /* When the HW is in SRIOV capable configuration, the PF-pool
3483 * resources are equally distributed across the max-number of
3484 * VFs. The user may request only a subset of the max-vfs to be
3485 * enabled. Based on num_vfs, redistribute the resources across
3486 * num_vfs so that each VF will have access to more number of
3487 * resources. This facility is not available in BE3 FW.
3488 * Also, this is done by FW in Lancer chip.
3490 if (be_max_vfs(adapter) && !pci_num_vf(adapter->pdev)) {
3491 status = be_cmd_set_sriov_config(adapter,
3495 dev_err(dev, "Failed to optimize SR-IOV resources\n");
3499 static int be_get_config(struct be_adapter *adapter)
3504 status = be_cmd_query_fw_cfg(adapter);
3508 if (be_physfn(adapter)) {
3509 status = be_cmd_get_active_profile(adapter, &profile_id);
3511 dev_info(&adapter->pdev->dev,
3512 "Using profile 0x%x\n", profile_id);
3515 if (!BE2_chip(adapter) && be_physfn(adapter))
3516 be_sriov_config(adapter);
3518 status = be_get_resources(adapter);
3522 adapter->pmac_id = kcalloc(be_max_uc(adapter),
3523 sizeof(*adapter->pmac_id), GFP_KERNEL);
3524 if (!adapter->pmac_id)
3527 /* Sanitize cfg_num_qs based on HW and platform limits */
3528 adapter->cfg_num_qs = min(adapter->cfg_num_qs, be_max_qs(adapter));
3533 static int be_mac_setup(struct be_adapter *adapter)
3538 if (is_zero_ether_addr(adapter->netdev->dev_addr)) {
3539 status = be_cmd_get_perm_mac(adapter, mac);
3543 memcpy(adapter->netdev->dev_addr, mac, ETH_ALEN);
3544 memcpy(adapter->netdev->perm_addr, mac, ETH_ALEN);
3546 /* Maybe the HW was reset; dev_addr must be re-programmed */
3547 memcpy(mac, adapter->netdev->dev_addr, ETH_ALEN);
3550 /* For BE3-R VFs, the PF programs the initial MAC address */
3551 if (!(BEx_chip(adapter) && be_virtfn(adapter)))
3552 be_cmd_pmac_add(adapter, mac, adapter->if_handle,
3553 &adapter->pmac_id[0], 0);
3557 static void be_schedule_worker(struct be_adapter *adapter)
3559 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
3560 adapter->flags |= BE_FLAGS_WORKER_SCHEDULED;
3563 static int be_setup_queues(struct be_adapter *adapter)
3565 struct net_device *netdev = adapter->netdev;
3568 status = be_evt_queues_create(adapter);
3572 status = be_tx_qs_create(adapter);
3576 status = be_rx_cqs_create(adapter);
3580 status = be_mcc_queues_create(adapter);
3584 status = netif_set_real_num_rx_queues(netdev, adapter->num_rx_qs);
3588 status = netif_set_real_num_tx_queues(netdev, adapter->num_tx_qs);
3594 dev_err(&adapter->pdev->dev, "queue_setup failed\n");
3598 int be_update_queues(struct be_adapter *adapter)
3600 struct net_device *netdev = adapter->netdev;
3603 if (netif_running(netdev))
3606 be_cancel_worker(adapter);
3608 /* If any vectors have been shared with RoCE we cannot re-program
3611 if (!adapter->num_msix_roce_vec)
3612 be_msix_disable(adapter);
3614 be_clear_queues(adapter);
3616 if (!msix_enabled(adapter)) {
3617 status = be_msix_enable(adapter);
3622 status = be_setup_queues(adapter);
3626 be_schedule_worker(adapter);
3628 if (netif_running(netdev))
3629 status = be_open(netdev);
3634 static int be_setup(struct be_adapter *adapter)
3636 struct device *dev = &adapter->pdev->dev;
3637 u32 tx_fc, rx_fc, en_flags;
3640 be_setup_init(adapter);
3642 if (!lancer_chip(adapter))
3643 be_cmd_req_native_mode(adapter);
3645 status = be_get_config(adapter);
3649 status = be_msix_enable(adapter);
3653 en_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3654 BE_IF_FLAGS_MULTICAST | BE_IF_FLAGS_PASS_L3L4_ERRORS;
3655 if (adapter->function_caps & BE_FUNCTION_CAPS_RSS)
3656 en_flags |= BE_IF_FLAGS_RSS;
3657 en_flags = en_flags & be_if_cap_flags(adapter);
3658 status = be_cmd_if_create(adapter, be_if_cap_flags(adapter), en_flags,
3659 &adapter->if_handle, 0);
3663 /* Updating real_num_tx/rx_queues() requires rtnl_lock() */
3665 status = be_setup_queues(adapter);
3670 be_cmd_get_fn_privileges(adapter, &adapter->cmd_privileges, 0);
3672 status = be_mac_setup(adapter);
3676 be_cmd_get_fw_ver(adapter);
3677 dev_info(dev, "FW version is %s\n", adapter->fw_ver);
3679 if (BE2_chip(adapter) && fw_major_num(adapter->fw_ver) < 4) {
3680 dev_err(dev, "Firmware on card is old(%s), IRQs may not work",
3682 dev_err(dev, "Please upgrade firmware to version >= 4.0\n");
3685 if (adapter->vlans_added)
3686 be_vid_config(adapter);
3688 be_set_rx_mode(adapter->netdev);
3690 be_cmd_get_acpi_wol_cap(adapter);
3692 be_cmd_get_flow_control(adapter, &tx_fc, &rx_fc);
3694 if (rx_fc != adapter->rx_fc || tx_fc != adapter->tx_fc)
3695 be_cmd_set_flow_control(adapter, adapter->tx_fc,
3698 if (be_physfn(adapter))
3699 be_cmd_set_logical_link_config(adapter,
3700 IFLA_VF_LINK_STATE_AUTO, 0);
3702 if (adapter->num_vfs)
3703 be_vf_setup(adapter);
3705 status = be_cmd_get_phy_info(adapter);
3706 if (!status && be_pause_supported(adapter))
3707 adapter->phy.fc_autoneg = 1;
3709 be_schedule_worker(adapter);
3710 adapter->flags |= BE_FLAGS_SETUP_DONE;
3717 #ifdef CONFIG_NET_POLL_CONTROLLER
3718 static void be_netpoll(struct net_device *netdev)
3720 struct be_adapter *adapter = netdev_priv(netdev);
3721 struct be_eq_obj *eqo;
3724 for_all_evt_queues(adapter, eqo, i) {
3725 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
3726 napi_schedule(&eqo->napi);
3731 static char flash_cookie[2][16] = {"*** SE FLAS", "H DIRECTORY *** "};
3733 static bool phy_flashing_required(struct be_adapter *adapter)
3735 return (adapter->phy.phy_type == TN_8022 &&
3736 adapter->phy.interface_type == PHY_TYPE_BASET_10GB);
3739 static bool is_comp_in_ufi(struct be_adapter *adapter,
3740 struct flash_section_info *fsec, int type)
3742 int i = 0, img_type = 0;
3743 struct flash_section_info_g2 *fsec_g2 = NULL;
3745 if (BE2_chip(adapter))
3746 fsec_g2 = (struct flash_section_info_g2 *)fsec;
3748 for (i = 0; i < MAX_FLASH_COMP; i++) {
3750 img_type = le32_to_cpu(fsec_g2->fsec_entry[i].type);
3752 img_type = le32_to_cpu(fsec->fsec_entry[i].type);
3754 if (img_type == type)
3761 static struct flash_section_info *get_fsec_info(struct be_adapter *adapter,
3763 const struct firmware *fw)
3765 struct flash_section_info *fsec = NULL;
3766 const u8 *p = fw->data;
3769 while (p < (fw->data + fw->size)) {
3770 fsec = (struct flash_section_info *)p;
3771 if (!memcmp(flash_cookie, fsec->cookie, sizeof(flash_cookie)))
3778 static int be_check_flash_crc(struct be_adapter *adapter, const u8 *p,
3779 u32 img_offset, u32 img_size, int hdr_size,
3780 u16 img_optype, bool *crc_match)
3786 status = be_cmd_get_flash_crc(adapter, crc, img_optype, img_size - 4);
3790 crc_offset = hdr_size + img_offset + img_size - 4;
3792 /* Skip flashing, if crc of flashed region matches */
3793 if (!memcmp(crc, p + crc_offset, 4))
3801 static int be_flash(struct be_adapter *adapter, const u8 *img,
3802 struct be_dma_mem *flash_cmd, int optype, int img_size)
3804 struct be_cmd_write_flashrom *req = flash_cmd->va;
3805 u32 total_bytes, flash_op, num_bytes;
3808 total_bytes = img_size;
3809 while (total_bytes) {
3810 num_bytes = min_t(u32, 32*1024, total_bytes);
3812 total_bytes -= num_bytes;
3815 if (optype == OPTYPE_PHY_FW)
3816 flash_op = FLASHROM_OPER_PHY_FLASH;
3818 flash_op = FLASHROM_OPER_FLASH;
3820 if (optype == OPTYPE_PHY_FW)
3821 flash_op = FLASHROM_OPER_PHY_SAVE;
3823 flash_op = FLASHROM_OPER_SAVE;
3826 memcpy(req->data_buf, img, num_bytes);
3828 status = be_cmd_write_flashrom(adapter, flash_cmd, optype,
3829 flash_op, num_bytes);
3830 if (base_status(status) == MCC_STATUS_ILLEGAL_REQUEST &&
3831 optype == OPTYPE_PHY_FW)
3839 /* For BE2, BE3 and BE3-R */
3840 static int be_flash_BEx(struct be_adapter *adapter,
3841 const struct firmware *fw,
3842 struct be_dma_mem *flash_cmd, int num_of_images)
3844 int img_hdrs_size = (num_of_images * sizeof(struct image_hdr));
3845 struct device *dev = &adapter->pdev->dev;
3846 struct flash_section_info *fsec = NULL;
3847 int status, i, filehdr_size, num_comp;
3848 const struct flash_comp *pflashcomp;
3852 struct flash_comp gen3_flash_types[] = {
3853 { FLASH_iSCSI_PRIMARY_IMAGE_START_g3, OPTYPE_ISCSI_ACTIVE,
3854 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_iSCSI},
3855 { FLASH_REDBOOT_START_g3, OPTYPE_REDBOOT,
3856 FLASH_REDBOOT_IMAGE_MAX_SIZE_g3, IMAGE_BOOT_CODE},
3857 { FLASH_iSCSI_BIOS_START_g3, OPTYPE_BIOS,
3858 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_ISCSI},
3859 { FLASH_PXE_BIOS_START_g3, OPTYPE_PXE_BIOS,
3860 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_PXE},
3861 { FLASH_FCoE_BIOS_START_g3, OPTYPE_FCOE_BIOS,
3862 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_FCoE},
3863 { FLASH_iSCSI_BACKUP_IMAGE_START_g3, OPTYPE_ISCSI_BACKUP,
3864 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_iSCSI},
3865 { FLASH_FCoE_PRIMARY_IMAGE_START_g3, OPTYPE_FCOE_FW_ACTIVE,
3866 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_FCoE},
3867 { FLASH_FCoE_BACKUP_IMAGE_START_g3, OPTYPE_FCOE_FW_BACKUP,
3868 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_FCoE},
3869 { FLASH_NCSI_START_g3, OPTYPE_NCSI_FW,
3870 FLASH_NCSI_IMAGE_MAX_SIZE_g3, IMAGE_NCSI},
3871 { FLASH_PHY_FW_START_g3, OPTYPE_PHY_FW,
3872 FLASH_PHY_FW_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_PHY}
3875 struct flash_comp gen2_flash_types[] = {
3876 { FLASH_iSCSI_PRIMARY_IMAGE_START_g2, OPTYPE_ISCSI_ACTIVE,
3877 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_iSCSI},
3878 { FLASH_REDBOOT_START_g2, OPTYPE_REDBOOT,
3879 FLASH_REDBOOT_IMAGE_MAX_SIZE_g2, IMAGE_BOOT_CODE},
3880 { FLASH_iSCSI_BIOS_START_g2, OPTYPE_BIOS,
3881 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_ISCSI},
3882 { FLASH_PXE_BIOS_START_g2, OPTYPE_PXE_BIOS,
3883 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_PXE},
3884 { FLASH_FCoE_BIOS_START_g2, OPTYPE_FCOE_BIOS,
3885 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_FCoE},
3886 { FLASH_iSCSI_BACKUP_IMAGE_START_g2, OPTYPE_ISCSI_BACKUP,
3887 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_iSCSI},
3888 { FLASH_FCoE_PRIMARY_IMAGE_START_g2, OPTYPE_FCOE_FW_ACTIVE,
3889 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_FCoE},
3890 { FLASH_FCoE_BACKUP_IMAGE_START_g2, OPTYPE_FCOE_FW_BACKUP,
3891 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_FCoE}
3894 if (BE3_chip(adapter)) {
3895 pflashcomp = gen3_flash_types;
3896 filehdr_size = sizeof(struct flash_file_hdr_g3);
3897 num_comp = ARRAY_SIZE(gen3_flash_types);
3899 pflashcomp = gen2_flash_types;
3900 filehdr_size = sizeof(struct flash_file_hdr_g2);
3901 num_comp = ARRAY_SIZE(gen2_flash_types);
3904 /* Get flash section info*/
3905 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
3907 dev_err(dev, "Invalid Cookie. FW image may be corrupted\n");
3910 for (i = 0; i < num_comp; i++) {
3911 if (!is_comp_in_ufi(adapter, fsec, pflashcomp[i].img_type))
3914 if ((pflashcomp[i].optype == OPTYPE_NCSI_FW) &&
3915 memcmp(adapter->fw_ver, "3.102.148.0", 11) < 0)
3918 if (pflashcomp[i].optype == OPTYPE_PHY_FW &&
3919 !phy_flashing_required(adapter))
3922 if (pflashcomp[i].optype == OPTYPE_REDBOOT) {
3923 status = be_check_flash_crc(adapter, fw->data,
3924 pflashcomp[i].offset,
3928 OPTYPE_REDBOOT, &crc_match);
3931 "Could not get CRC for 0x%x region\n",
3932 pflashcomp[i].optype);
3940 p = fw->data + filehdr_size + pflashcomp[i].offset +
3942 if (p + pflashcomp[i].size > fw->data + fw->size)
3945 status = be_flash(adapter, p, flash_cmd, pflashcomp[i].optype,
3946 pflashcomp[i].size);
3948 dev_err(dev, "Flashing section type 0x%x failed\n",
3949 pflashcomp[i].img_type);
3956 static u16 be_get_img_optype(struct flash_section_entry fsec_entry)
3958 u32 img_type = le32_to_cpu(fsec_entry.type);
3959 u16 img_optype = le16_to_cpu(fsec_entry.optype);
3961 if (img_optype != 0xFFFF)
3965 case IMAGE_FIRMWARE_iSCSI:
3966 img_optype = OPTYPE_ISCSI_ACTIVE;
3968 case IMAGE_BOOT_CODE:
3969 img_optype = OPTYPE_REDBOOT;
3971 case IMAGE_OPTION_ROM_ISCSI:
3972 img_optype = OPTYPE_BIOS;
3974 case IMAGE_OPTION_ROM_PXE:
3975 img_optype = OPTYPE_PXE_BIOS;
3977 case IMAGE_OPTION_ROM_FCoE:
3978 img_optype = OPTYPE_FCOE_BIOS;
3980 case IMAGE_FIRMWARE_BACKUP_iSCSI:
3981 img_optype = OPTYPE_ISCSI_BACKUP;
3984 img_optype = OPTYPE_NCSI_FW;
3986 case IMAGE_FLASHISM_JUMPVECTOR:
3987 img_optype = OPTYPE_FLASHISM_JUMPVECTOR;
3989 case IMAGE_FIRMWARE_PHY:
3990 img_optype = OPTYPE_SH_PHY_FW;
3992 case IMAGE_REDBOOT_DIR:
3993 img_optype = OPTYPE_REDBOOT_DIR;
3995 case IMAGE_REDBOOT_CONFIG:
3996 img_optype = OPTYPE_REDBOOT_CONFIG;
3999 img_optype = OPTYPE_UFI_DIR;
4008 static int be_flash_skyhawk(struct be_adapter *adapter,
4009 const struct firmware *fw,
4010 struct be_dma_mem *flash_cmd, int num_of_images)
4012 int img_hdrs_size = num_of_images * sizeof(struct image_hdr);
4013 struct device *dev = &adapter->pdev->dev;
4014 struct flash_section_info *fsec = NULL;
4015 u32 img_offset, img_size, img_type;
4016 int status, i, filehdr_size;
4017 bool crc_match, old_fw_img;
4021 filehdr_size = sizeof(struct flash_file_hdr_g3);
4022 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
4024 dev_err(dev, "Invalid Cookie. FW image may be corrupted\n");
4028 for (i = 0; i < le32_to_cpu(fsec->fsec_hdr.num_images); i++) {
4029 img_offset = le32_to_cpu(fsec->fsec_entry[i].offset);
4030 img_size = le32_to_cpu(fsec->fsec_entry[i].pad_size);
4031 img_type = le32_to_cpu(fsec->fsec_entry[i].type);
4032 img_optype = be_get_img_optype(fsec->fsec_entry[i]);
4033 old_fw_img = fsec->fsec_entry[i].optype == 0xFFFF;
4035 if (img_optype == 0xFFFF)
4037 /* Don't bother verifying CRC if an old FW image is being
4043 status = be_check_flash_crc(adapter, fw->data, img_offset,
4044 img_size, filehdr_size +
4045 img_hdrs_size, img_optype,
4047 /* The current FW image on the card does not recognize the new
4048 * FLASH op_type. The FW download is partially complete.
4049 * Reboot the server now to enable FW image to recognize the
4050 * new FLASH op_type. To complete the remaining process,
4051 * download the same FW again after the reboot.
4053 if (base_status(status) == MCC_STATUS_ILLEGAL_REQUEST ||
4054 base_status(status) == MCC_STATUS_ILLEGAL_FIELD) {
4055 dev_err(dev, "Flash incomplete. Reset the server\n");
4056 dev_err(dev, "Download FW image again after reset\n");
4058 } else if (status) {
4059 dev_err(dev, "Could not get CRC for 0x%x region\n",
4068 p = fw->data + filehdr_size + img_offset + img_hdrs_size;
4069 if (p + img_size > fw->data + fw->size)
4072 status = be_flash(adapter, p, flash_cmd, img_optype, img_size);
4073 /* For old FW images ignore ILLEGAL_FIELD error or errors on
4077 (base_status(status) == MCC_STATUS_ILLEGAL_FIELD ||
4078 (img_optype == OPTYPE_UFI_DIR &&
4079 base_status(status) == MCC_STATUS_FAILED))) {
4081 } else if (status) {
4082 dev_err(dev, "Flashing section type 0x%x failed\n",
4090 static int lancer_fw_download(struct be_adapter *adapter,
4091 const struct firmware *fw)
4093 #define LANCER_FW_DOWNLOAD_CHUNK (32 * 1024)
4094 #define LANCER_FW_DOWNLOAD_LOCATION "/prg"
4095 struct device *dev = &adapter->pdev->dev;
4096 struct be_dma_mem flash_cmd;
4097 const u8 *data_ptr = NULL;
4098 u8 *dest_image_ptr = NULL;
4099 size_t image_size = 0;
4101 u32 data_written = 0;
4107 if (!IS_ALIGNED(fw->size, sizeof(u32))) {
4108 dev_err(dev, "FW image size should be multiple of 4\n");
4112 flash_cmd.size = sizeof(struct lancer_cmd_req_write_object)
4113 + LANCER_FW_DOWNLOAD_CHUNK;
4114 flash_cmd.va = dma_alloc_coherent(dev, flash_cmd.size,
4115 &flash_cmd.dma, GFP_KERNEL);
4119 dest_image_ptr = flash_cmd.va +
4120 sizeof(struct lancer_cmd_req_write_object);
4121 image_size = fw->size;
4122 data_ptr = fw->data;
4124 while (image_size) {
4125 chunk_size = min_t(u32, image_size, LANCER_FW_DOWNLOAD_CHUNK);
4127 /* Copy the image chunk content. */
4128 memcpy(dest_image_ptr, data_ptr, chunk_size);
4130 status = lancer_cmd_write_object(adapter, &flash_cmd,
4132 LANCER_FW_DOWNLOAD_LOCATION,
4133 &data_written, &change_status,
4138 offset += data_written;
4139 data_ptr += data_written;
4140 image_size -= data_written;
4144 /* Commit the FW written */
4145 status = lancer_cmd_write_object(adapter, &flash_cmd,
4147 LANCER_FW_DOWNLOAD_LOCATION,
4148 &data_written, &change_status,
4152 dma_free_coherent(dev, flash_cmd.size, flash_cmd.va, flash_cmd.dma);
4154 dev_err(dev, "Firmware load error\n");
4155 return be_cmd_status(status);
4158 dev_info(dev, "Firmware flashed successfully\n");
4160 if (change_status == LANCER_FW_RESET_NEEDED) {
4161 dev_info(dev, "Resetting adapter to activate new FW\n");
4162 status = lancer_physdev_ctrl(adapter,
4163 PHYSDEV_CONTROL_FW_RESET_MASK);
4165 dev_err(dev, "Adapter busy, could not reset FW\n");
4166 dev_err(dev, "Reboot server to activate new FW\n");
4168 } else if (change_status != LANCER_NO_RESET_NEEDED) {
4169 dev_info(dev, "Reboot server to activate new FW\n");
4177 #define UFI_TYPE3R 10
4179 static int be_get_ufi_type(struct be_adapter *adapter,
4180 struct flash_file_hdr_g3 *fhdr)
4183 goto be_get_ufi_exit;
4185 if (skyhawk_chip(adapter) && fhdr->build[0] == '4')
4187 else if (BE3_chip(adapter) && fhdr->build[0] == '3') {
4188 if (fhdr->asic_type_rev == 0x10)
4192 } else if (BE2_chip(adapter) && fhdr->build[0] == '2')
4196 dev_err(&adapter->pdev->dev,
4197 "UFI and Interface are not compatible for flashing\n");
4201 static int be_fw_download(struct be_adapter *adapter, const struct firmware* fw)
4203 struct flash_file_hdr_g3 *fhdr3;
4204 struct image_hdr *img_hdr_ptr = NULL;
4205 struct be_dma_mem flash_cmd;
4207 int status = 0, i = 0, num_imgs = 0, ufi_type = 0;
4209 flash_cmd.size = sizeof(struct be_cmd_write_flashrom);
4210 flash_cmd.va = dma_alloc_coherent(&adapter->pdev->dev, flash_cmd.size,
4211 &flash_cmd.dma, GFP_KERNEL);
4212 if (!flash_cmd.va) {
4218 fhdr3 = (struct flash_file_hdr_g3 *)p;
4220 ufi_type = be_get_ufi_type(adapter, fhdr3);
4222 num_imgs = le32_to_cpu(fhdr3->num_imgs);
4223 for (i = 0; i < num_imgs; i++) {
4224 img_hdr_ptr = (struct image_hdr *)(fw->data +
4225 (sizeof(struct flash_file_hdr_g3) +
4226 i * sizeof(struct image_hdr)));
4227 if (le32_to_cpu(img_hdr_ptr->imageid) == 1) {
4230 status = be_flash_skyhawk(adapter, fw,
4231 &flash_cmd, num_imgs);
4234 status = be_flash_BEx(adapter, fw, &flash_cmd,
4238 /* Do not flash this ufi on BE3-R cards */
4239 if (adapter->asic_rev < 0x10)
4240 status = be_flash_BEx(adapter, fw,
4245 dev_err(&adapter->pdev->dev,
4246 "Can't load BE3 UFI on BE3R\n");
4252 if (ufi_type == UFI_TYPE2)
4253 status = be_flash_BEx(adapter, fw, &flash_cmd, 0);
4254 else if (ufi_type == -1)
4257 dma_free_coherent(&adapter->pdev->dev, flash_cmd.size, flash_cmd.va,
4260 dev_err(&adapter->pdev->dev, "Firmware load error\n");
4264 dev_info(&adapter->pdev->dev, "Firmware flashed successfully\n");
4270 int be_load_fw(struct be_adapter *adapter, u8 *fw_file)
4272 const struct firmware *fw;
4275 if (!netif_running(adapter->netdev)) {
4276 dev_err(&adapter->pdev->dev,
4277 "Firmware load not allowed (interface is down)\n");
4281 status = request_firmware(&fw, fw_file, &adapter->pdev->dev);
4285 dev_info(&adapter->pdev->dev, "Flashing firmware file %s\n", fw_file);
4287 if (lancer_chip(adapter))
4288 status = lancer_fw_download(adapter, fw);
4290 status = be_fw_download(adapter, fw);
4293 be_cmd_get_fw_ver(adapter);
4296 release_firmware(fw);
4300 static int be_ndo_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh)
4302 struct be_adapter *adapter = netdev_priv(dev);
4303 struct nlattr *attr, *br_spec;
4308 if (!sriov_enabled(adapter))
4311 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4313 nla_for_each_nested(attr, br_spec, rem) {
4314 if (nla_type(attr) != IFLA_BRIDGE_MODE)
4317 mode = nla_get_u16(attr);
4318 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
4321 status = be_cmd_set_hsw_config(adapter, 0, 0,
4323 mode == BRIDGE_MODE_VEPA ?
4324 PORT_FWD_TYPE_VEPA :
4329 dev_info(&adapter->pdev->dev, "enabled switch mode: %s\n",
4330 mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
4335 dev_err(&adapter->pdev->dev, "Failed to set switch mode %s\n",
4336 mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
4341 static int be_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4342 struct net_device *dev, u32 filter_mask)
4344 struct be_adapter *adapter = netdev_priv(dev);
4348 if (!sriov_enabled(adapter))
4351 /* BE and Lancer chips support VEB mode only */
4352 if (BEx_chip(adapter) || lancer_chip(adapter)) {
4353 hsw_mode = PORT_FWD_TYPE_VEB;
4355 status = be_cmd_get_hsw_config(adapter, NULL, 0,
4356 adapter->if_handle, &hsw_mode);
4361 return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
4362 hsw_mode == PORT_FWD_TYPE_VEPA ?
4363 BRIDGE_MODE_VEPA : BRIDGE_MODE_VEB);
4366 #ifdef CONFIG_BE2NET_VXLAN
4367 static void be_add_vxlan_port(struct net_device *netdev, sa_family_t sa_family,
4370 struct be_adapter *adapter = netdev_priv(netdev);
4371 struct device *dev = &adapter->pdev->dev;
4374 if (lancer_chip(adapter) || BEx_chip(adapter))
4377 if (adapter->flags & BE_FLAGS_VXLAN_OFFLOADS) {
4378 dev_warn(dev, "Cannot add UDP port %d for VxLAN offloads\n",
4381 "Only one UDP port supported for VxLAN offloads\n");
4385 status = be_cmd_manage_iface(adapter, adapter->if_handle,
4386 OP_CONVERT_NORMAL_TO_TUNNEL);
4388 dev_warn(dev, "Failed to convert normal interface to tunnel\n");
4392 status = be_cmd_set_vxlan_port(adapter, port);
4394 dev_warn(dev, "Failed to add VxLAN port\n");
4397 adapter->flags |= BE_FLAGS_VXLAN_OFFLOADS;
4398 adapter->vxlan_port = port;
4400 dev_info(dev, "Enabled VxLAN offloads for UDP port %d\n",
4404 be_disable_vxlan_offloads(adapter);
4407 static void be_del_vxlan_port(struct net_device *netdev, sa_family_t sa_family,
4410 struct be_adapter *adapter = netdev_priv(netdev);
4412 if (lancer_chip(adapter) || BEx_chip(adapter))
4415 if (adapter->vxlan_port != port)
4418 be_disable_vxlan_offloads(adapter);
4420 dev_info(&adapter->pdev->dev,
4421 "Disabled VxLAN offloads for UDP port %d\n",
4426 static const struct net_device_ops be_netdev_ops = {
4427 .ndo_open = be_open,
4428 .ndo_stop = be_close,
4429 .ndo_start_xmit = be_xmit,
4430 .ndo_set_rx_mode = be_set_rx_mode,
4431 .ndo_set_mac_address = be_mac_addr_set,
4432 .ndo_change_mtu = be_change_mtu,
4433 .ndo_get_stats64 = be_get_stats64,
4434 .ndo_validate_addr = eth_validate_addr,
4435 .ndo_vlan_rx_add_vid = be_vlan_add_vid,
4436 .ndo_vlan_rx_kill_vid = be_vlan_rem_vid,
4437 .ndo_set_vf_mac = be_set_vf_mac,
4438 .ndo_set_vf_vlan = be_set_vf_vlan,
4439 .ndo_set_vf_rate = be_set_vf_tx_rate,
4440 .ndo_get_vf_config = be_get_vf_config,
4441 .ndo_set_vf_link_state = be_set_vf_link_state,
4442 #ifdef CONFIG_NET_POLL_CONTROLLER
4443 .ndo_poll_controller = be_netpoll,
4445 .ndo_bridge_setlink = be_ndo_bridge_setlink,
4446 .ndo_bridge_getlink = be_ndo_bridge_getlink,
4447 #ifdef CONFIG_NET_RX_BUSY_POLL
4448 .ndo_busy_poll = be_busy_poll,
4450 #ifdef CONFIG_BE2NET_VXLAN
4451 .ndo_add_vxlan_port = be_add_vxlan_port,
4452 .ndo_del_vxlan_port = be_del_vxlan_port,
4456 static void be_netdev_init(struct net_device *netdev)
4458 struct be_adapter *adapter = netdev_priv(netdev);
4460 if (skyhawk_chip(adapter)) {
4461 netdev->hw_enc_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
4462 NETIF_F_TSO | NETIF_F_TSO6 |
4463 NETIF_F_GSO_UDP_TUNNEL;
4464 netdev->hw_features |= NETIF_F_GSO_UDP_TUNNEL;
4466 netdev->hw_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
4467 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM | NETIF_F_RXCSUM |
4468 NETIF_F_HW_VLAN_CTAG_TX;
4469 if (be_multi_rxq(adapter))
4470 netdev->hw_features |= NETIF_F_RXHASH;
4472 netdev->features |= netdev->hw_features |
4473 NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_FILTER;
4475 netdev->vlan_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
4476 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
4478 netdev->priv_flags |= IFF_UNICAST_FLT;
4480 netdev->flags |= IFF_MULTICAST;
4482 netif_set_gso_max_size(netdev, 65535 - ETH_HLEN);
4484 netdev->netdev_ops = &be_netdev_ops;
4486 netdev->ethtool_ops = &be_ethtool_ops;
4489 static void be_unmap_pci_bars(struct be_adapter *adapter)
4492 pci_iounmap(adapter->pdev, adapter->csr);
4494 pci_iounmap(adapter->pdev, adapter->db);
4497 static int db_bar(struct be_adapter *adapter)
4499 if (lancer_chip(adapter) || !be_physfn(adapter))
4505 static int be_roce_map_pci_bars(struct be_adapter *adapter)
4507 if (skyhawk_chip(adapter)) {
4508 adapter->roce_db.size = 4096;
4509 adapter->roce_db.io_addr = pci_resource_start(adapter->pdev,
4511 adapter->roce_db.total_size = pci_resource_len(adapter->pdev,
4517 static int be_map_pci_bars(struct be_adapter *adapter)
4521 if (BEx_chip(adapter) && be_physfn(adapter)) {
4522 adapter->csr = pci_iomap(adapter->pdev, 2, 0);
4527 addr = pci_iomap(adapter->pdev, db_bar(adapter), 0);
4532 be_roce_map_pci_bars(adapter);
4536 dev_err(&adapter->pdev->dev, "Error in mapping PCI BARs\n");
4537 be_unmap_pci_bars(adapter);
4541 static void be_ctrl_cleanup(struct be_adapter *adapter)
4543 struct be_dma_mem *mem = &adapter->mbox_mem_alloced;
4545 be_unmap_pci_bars(adapter);
4548 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
4551 mem = &adapter->rx_filter;
4553 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
4557 static int be_ctrl_init(struct be_adapter *adapter)
4559 struct be_dma_mem *mbox_mem_alloc = &adapter->mbox_mem_alloced;
4560 struct be_dma_mem *mbox_mem_align = &adapter->mbox_mem;
4561 struct be_dma_mem *rx_filter = &adapter->rx_filter;
4565 pci_read_config_dword(adapter->pdev, SLI_INTF_REG_OFFSET, &sli_intf);
4566 adapter->sli_family = (sli_intf & SLI_INTF_FAMILY_MASK) >>
4567 SLI_INTF_FAMILY_SHIFT;
4568 adapter->virtfn = (sli_intf & SLI_INTF_FT_MASK) ? 1 : 0;
4570 status = be_map_pci_bars(adapter);
4574 mbox_mem_alloc->size = sizeof(struct be_mcc_mailbox) + 16;
4575 mbox_mem_alloc->va = dma_alloc_coherent(&adapter->pdev->dev,
4576 mbox_mem_alloc->size,
4577 &mbox_mem_alloc->dma,
4579 if (!mbox_mem_alloc->va) {
4581 goto unmap_pci_bars;
4583 mbox_mem_align->size = sizeof(struct be_mcc_mailbox);
4584 mbox_mem_align->va = PTR_ALIGN(mbox_mem_alloc->va, 16);
4585 mbox_mem_align->dma = PTR_ALIGN(mbox_mem_alloc->dma, 16);
4586 memset(mbox_mem_align->va, 0, sizeof(struct be_mcc_mailbox));
4588 rx_filter->size = sizeof(struct be_cmd_req_rx_filter);
4589 rx_filter->va = dma_zalloc_coherent(&adapter->pdev->dev,
4590 rx_filter->size, &rx_filter->dma,
4592 if (!rx_filter->va) {
4597 mutex_init(&adapter->mbox_lock);
4598 spin_lock_init(&adapter->mcc_lock);
4599 spin_lock_init(&adapter->mcc_cq_lock);
4601 init_completion(&adapter->et_cmd_compl);
4602 pci_save_state(adapter->pdev);
4606 dma_free_coherent(&adapter->pdev->dev, mbox_mem_alloc->size,
4607 mbox_mem_alloc->va, mbox_mem_alloc->dma);
4610 be_unmap_pci_bars(adapter);
4616 static void be_stats_cleanup(struct be_adapter *adapter)
4618 struct be_dma_mem *cmd = &adapter->stats_cmd;
4621 dma_free_coherent(&adapter->pdev->dev, cmd->size,
4625 static int be_stats_init(struct be_adapter *adapter)
4627 struct be_dma_mem *cmd = &adapter->stats_cmd;
4629 if (lancer_chip(adapter))
4630 cmd->size = sizeof(struct lancer_cmd_req_pport_stats);
4631 else if (BE2_chip(adapter))
4632 cmd->size = sizeof(struct be_cmd_req_get_stats_v0);
4633 else if (BE3_chip(adapter))
4634 cmd->size = sizeof(struct be_cmd_req_get_stats_v1);
4636 /* ALL non-BE ASICs */
4637 cmd->size = sizeof(struct be_cmd_req_get_stats_v2);
4639 cmd->va = dma_zalloc_coherent(&adapter->pdev->dev, cmd->size, &cmd->dma,
4646 static void be_remove(struct pci_dev *pdev)
4648 struct be_adapter *adapter = pci_get_drvdata(pdev);
4653 be_roce_dev_remove(adapter);
4654 be_intr_set(adapter, false);
4656 cancel_delayed_work_sync(&adapter->func_recovery_work);
4658 unregister_netdev(adapter->netdev);
4662 /* tell fw we're done with firing cmds */
4663 be_cmd_fw_clean(adapter);
4665 be_stats_cleanup(adapter);
4667 be_ctrl_cleanup(adapter);
4669 pci_disable_pcie_error_reporting(pdev);
4671 pci_release_regions(pdev);
4672 pci_disable_device(pdev);
4674 free_netdev(adapter->netdev);
4677 static int be_get_initial_config(struct be_adapter *adapter)
4681 status = be_cmd_get_cntl_attributes(adapter);
4685 /* Must be a power of 2 or else MODULO will BUG_ON */
4686 adapter->be_get_temp_freq = 64;
4688 if (BEx_chip(adapter)) {
4689 level = be_cmd_get_fw_log_level(adapter);
4690 adapter->msg_enable =
4691 level <= FW_LOG_LEVEL_DEFAULT ? NETIF_MSG_HW : 0;
4694 adapter->cfg_num_qs = netif_get_num_default_rss_queues();
4698 static int lancer_recover_func(struct be_adapter *adapter)
4700 struct device *dev = &adapter->pdev->dev;
4703 status = lancer_test_and_set_rdy_state(adapter);
4707 if (netif_running(adapter->netdev))
4708 be_close(adapter->netdev);
4712 be_clear_all_error(adapter);
4714 status = be_setup(adapter);
4718 if (netif_running(adapter->netdev)) {
4719 status = be_open(adapter->netdev);
4724 dev_err(dev, "Adapter recovery successful\n");
4727 if (status == -EAGAIN)
4728 dev_err(dev, "Waiting for resource provisioning\n");
4730 dev_err(dev, "Adapter recovery failed\n");
4735 static void be_func_recovery_task(struct work_struct *work)
4737 struct be_adapter *adapter =
4738 container_of(work, struct be_adapter, func_recovery_work.work);
4741 be_detect_error(adapter);
4743 if (adapter->hw_error && lancer_chip(adapter)) {
4746 netif_device_detach(adapter->netdev);
4749 status = lancer_recover_func(adapter);
4751 netif_device_attach(adapter->netdev);
4754 /* In Lancer, for all errors other than provisioning error (-EAGAIN),
4755 * no need to attempt further recovery.
4757 if (!status || status == -EAGAIN)
4758 schedule_delayed_work(&adapter->func_recovery_work,
4759 msecs_to_jiffies(1000));
4762 static void be_worker(struct work_struct *work)
4764 struct be_adapter *adapter =
4765 container_of(work, struct be_adapter, work.work);
4766 struct be_rx_obj *rxo;
4769 /* when interrupts are not yet enabled, just reap any pending
4770 * mcc completions */
4771 if (!netif_running(adapter->netdev)) {
4773 be_process_mcc(adapter);
4778 if (!adapter->stats_cmd_sent) {
4779 if (lancer_chip(adapter))
4780 lancer_cmd_get_pport_stats(adapter,
4781 &adapter->stats_cmd);
4783 be_cmd_get_stats(adapter, &adapter->stats_cmd);
4786 if (be_physfn(adapter) &&
4787 MODULO(adapter->work_counter, adapter->be_get_temp_freq) == 0)
4788 be_cmd_get_die_temperature(adapter);
4790 for_all_rx_queues(adapter, rxo, i) {
4791 /* Replenish RX-queues starved due to memory
4792 * allocation failures.
4794 if (rxo->rx_post_starved)
4795 be_post_rx_frags(rxo, GFP_KERNEL, MAX_RX_POST);
4798 be_eqd_update(adapter);
4801 adapter->work_counter++;
4802 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
4805 /* If any VFs are already enabled don't FLR the PF */
4806 static bool be_reset_required(struct be_adapter *adapter)
4808 return pci_num_vf(adapter->pdev) ? false : true;
4811 static char *mc_name(struct be_adapter *adapter)
4813 char *str = ""; /* default */
4815 switch (adapter->mc_type) {
4841 static inline char *func_name(struct be_adapter *adapter)
4843 return be_physfn(adapter) ? "PF" : "VF";
4846 static int be_probe(struct pci_dev *pdev, const struct pci_device_id *pdev_id)
4849 struct be_adapter *adapter;
4850 struct net_device *netdev;
4853 dev_info(&pdev->dev, "%s version is %s\n", DRV_NAME, DRV_VER);
4855 status = pci_enable_device(pdev);
4859 status = pci_request_regions(pdev, DRV_NAME);
4862 pci_set_master(pdev);
4864 netdev = alloc_etherdev_mqs(sizeof(*adapter), MAX_TX_QS, MAX_RX_QS);
4869 adapter = netdev_priv(netdev);
4870 adapter->pdev = pdev;
4871 pci_set_drvdata(pdev, adapter);
4872 adapter->netdev = netdev;
4873 SET_NETDEV_DEV(netdev, &pdev->dev);
4875 status = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
4877 netdev->features |= NETIF_F_HIGHDMA;
4879 status = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
4881 dev_err(&pdev->dev, "Could not set PCI DMA Mask\n");
4886 status = pci_enable_pcie_error_reporting(pdev);
4888 dev_info(&pdev->dev, "PCIe error reporting enabled\n");
4890 status = be_ctrl_init(adapter);
4894 /* sync up with fw's ready state */
4895 if (be_physfn(adapter)) {
4896 status = be_fw_wait_ready(adapter);
4901 if (be_reset_required(adapter)) {
4902 status = be_cmd_reset_function(adapter);
4906 /* Wait for interrupts to quiesce after an FLR */
4910 /* Allow interrupts for other ULPs running on NIC function */
4911 be_intr_set(adapter, true);
4913 /* tell fw we're ready to fire cmds */
4914 status = be_cmd_fw_init(adapter);
4918 status = be_stats_init(adapter);
4922 status = be_get_initial_config(adapter);
4926 INIT_DELAYED_WORK(&adapter->work, be_worker);
4927 INIT_DELAYED_WORK(&adapter->func_recovery_work, be_func_recovery_task);
4928 adapter->rx_fc = adapter->tx_fc = true;
4930 status = be_setup(adapter);
4934 be_netdev_init(netdev);
4935 status = register_netdev(netdev);
4939 be_roce_dev_add(adapter);
4941 schedule_delayed_work(&adapter->func_recovery_work,
4942 msecs_to_jiffies(1000));
4944 be_cmd_query_port_name(adapter, &port_name);
4946 dev_info(&pdev->dev, "%s: %s %s port %c\n", nic_name(pdev),
4947 func_name(adapter), mc_name(adapter), port_name);
4954 be_stats_cleanup(adapter);
4956 be_ctrl_cleanup(adapter);
4958 free_netdev(netdev);
4960 pci_release_regions(pdev);
4962 pci_disable_device(pdev);
4964 dev_err(&pdev->dev, "%s initialization failed\n", nic_name(pdev));
4968 static int be_suspend(struct pci_dev *pdev, pm_message_t state)
4970 struct be_adapter *adapter = pci_get_drvdata(pdev);
4971 struct net_device *netdev = adapter->netdev;
4973 if (adapter->wol_en)
4974 be_setup_wol(adapter, true);
4976 be_intr_set(adapter, false);
4977 cancel_delayed_work_sync(&adapter->func_recovery_work);
4979 netif_device_detach(netdev);
4980 if (netif_running(netdev)) {
4987 pci_save_state(pdev);
4988 pci_disable_device(pdev);
4989 pci_set_power_state(pdev, pci_choose_state(pdev, state));
4993 static int be_resume(struct pci_dev *pdev)
4996 struct be_adapter *adapter = pci_get_drvdata(pdev);
4997 struct net_device *netdev = adapter->netdev;
4999 netif_device_detach(netdev);
5001 status = pci_enable_device(pdev);
5005 pci_set_power_state(pdev, PCI_D0);
5006 pci_restore_state(pdev);
5008 status = be_fw_wait_ready(adapter);
5012 be_intr_set(adapter, true);
5013 /* tell fw we're ready to fire cmds */
5014 status = be_cmd_fw_init(adapter);
5019 if (netif_running(netdev)) {
5025 schedule_delayed_work(&adapter->func_recovery_work,
5026 msecs_to_jiffies(1000));
5027 netif_device_attach(netdev);
5029 if (adapter->wol_en)
5030 be_setup_wol(adapter, false);
5036 * An FLR will stop BE from DMAing any data.
5038 static void be_shutdown(struct pci_dev *pdev)
5040 struct be_adapter *adapter = pci_get_drvdata(pdev);
5045 be_roce_dev_shutdown(adapter);
5046 cancel_delayed_work_sync(&adapter->work);
5047 cancel_delayed_work_sync(&adapter->func_recovery_work);
5049 netif_device_detach(adapter->netdev);
5051 be_cmd_reset_function(adapter);
5053 pci_disable_device(pdev);
5056 static pci_ers_result_t be_eeh_err_detected(struct pci_dev *pdev,
5057 pci_channel_state_t state)
5059 struct be_adapter *adapter = pci_get_drvdata(pdev);
5060 struct net_device *netdev = adapter->netdev;
5062 dev_err(&adapter->pdev->dev, "EEH error detected\n");
5064 if (!adapter->eeh_error) {
5065 adapter->eeh_error = true;
5067 cancel_delayed_work_sync(&adapter->func_recovery_work);
5070 netif_device_detach(netdev);
5071 if (netif_running(netdev))
5078 if (state == pci_channel_io_perm_failure)
5079 return PCI_ERS_RESULT_DISCONNECT;
5081 pci_disable_device(pdev);
5083 /* The error could cause the FW to trigger a flash debug dump.
5084 * Resetting the card while flash dump is in progress
5085 * can cause it not to recover; wait for it to finish.
5086 * Wait only for first function as it is needed only once per
5089 if (pdev->devfn == 0)
5092 return PCI_ERS_RESULT_NEED_RESET;
5095 static pci_ers_result_t be_eeh_reset(struct pci_dev *pdev)
5097 struct be_adapter *adapter = pci_get_drvdata(pdev);
5100 dev_info(&adapter->pdev->dev, "EEH reset\n");
5102 status = pci_enable_device(pdev);
5104 return PCI_ERS_RESULT_DISCONNECT;
5106 pci_set_master(pdev);
5107 pci_set_power_state(pdev, PCI_D0);
5108 pci_restore_state(pdev);
5110 /* Check if card is ok and fw is ready */
5111 dev_info(&adapter->pdev->dev,
5112 "Waiting for FW to be ready after EEH reset\n");
5113 status = be_fw_wait_ready(adapter);
5115 return PCI_ERS_RESULT_DISCONNECT;
5117 pci_cleanup_aer_uncorrect_error_status(pdev);
5118 be_clear_all_error(adapter);
5119 return PCI_ERS_RESULT_RECOVERED;
5122 static void be_eeh_resume(struct pci_dev *pdev)
5125 struct be_adapter *adapter = pci_get_drvdata(pdev);
5126 struct net_device *netdev = adapter->netdev;
5128 dev_info(&adapter->pdev->dev, "EEH resume\n");
5130 pci_save_state(pdev);
5132 status = be_cmd_reset_function(adapter);
5136 /* On some BE3 FW versions, after a HW reset,
5137 * interrupts will remain disabled for each function.
5138 * So, explicitly enable interrupts
5140 be_intr_set(adapter, true);
5142 /* tell fw we're ready to fire cmds */
5143 status = be_cmd_fw_init(adapter);
5147 status = be_setup(adapter);
5151 if (netif_running(netdev)) {
5152 status = be_open(netdev);
5157 schedule_delayed_work(&adapter->func_recovery_work,
5158 msecs_to_jiffies(1000));
5159 netif_device_attach(netdev);
5162 dev_err(&adapter->pdev->dev, "EEH resume failed\n");
5165 static const struct pci_error_handlers be_eeh_handlers = {
5166 .error_detected = be_eeh_err_detected,
5167 .slot_reset = be_eeh_reset,
5168 .resume = be_eeh_resume,
5171 static struct pci_driver be_driver = {
5173 .id_table = be_dev_ids,
5175 .remove = be_remove,
5176 .suspend = be_suspend,
5177 .resume = be_resume,
5178 .shutdown = be_shutdown,
5179 .err_handler = &be_eeh_handlers
5182 static int __init be_init_module(void)
5184 if (rx_frag_size != 8192 && rx_frag_size != 4096 &&
5185 rx_frag_size != 2048) {
5186 printk(KERN_WARNING DRV_NAME
5187 " : Module param rx_frag_size must be 2048/4096/8192."
5189 rx_frag_size = 2048;
5192 return pci_register_driver(&be_driver);
5194 module_init(be_init_module);
5196 static void __exit be_exit_module(void)
5198 pci_unregister_driver(&be_driver);
5200 module_exit(be_exit_module);