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_DESCRIPTION(DRV_DESC " " DRV_VER);
30 MODULE_AUTHOR("Emulex Corporation");
31 MODULE_LICENSE("GPL");
33 /* num_vfs module param is obsolete.
34 * Use sysfs method to enable/disable VFs.
36 static unsigned int num_vfs;
37 module_param(num_vfs, uint, S_IRUGO);
38 MODULE_PARM_DESC(num_vfs, "Number of PCI VFs to initialize");
40 static ushort rx_frag_size = 2048;
41 module_param(rx_frag_size, ushort, S_IRUGO);
42 MODULE_PARM_DESC(rx_frag_size, "Size of a fragment that holds rcvd data.");
44 static const struct pci_device_id be_dev_ids[] = {
45 { PCI_DEVICE(BE_VENDOR_ID, BE_DEVICE_ID1) },
46 { PCI_DEVICE(BE_VENDOR_ID, BE_DEVICE_ID2) },
47 { PCI_DEVICE(BE_VENDOR_ID, OC_DEVICE_ID1) },
48 { PCI_DEVICE(BE_VENDOR_ID, OC_DEVICE_ID2) },
49 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID3)},
50 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID4)},
51 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID5)},
52 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID6)},
55 MODULE_DEVICE_TABLE(pci, be_dev_ids);
56 /* UE Status Low CSR */
57 static const char * const ue_status_low_desc[] = {
92 /* UE Status High CSR */
93 static const char * const ue_status_hi_desc[] = {
128 static void be_queue_free(struct be_adapter *adapter, struct be_queue_info *q)
130 struct be_dma_mem *mem = &q->dma_mem;
133 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
139 static int be_queue_alloc(struct be_adapter *adapter, struct be_queue_info *q,
140 u16 len, u16 entry_size)
142 struct be_dma_mem *mem = &q->dma_mem;
144 memset(q, 0, sizeof(*q));
146 q->entry_size = entry_size;
147 mem->size = len * entry_size;
148 mem->va = dma_zalloc_coherent(&adapter->pdev->dev, mem->size, &mem->dma,
155 static void be_reg_intr_set(struct be_adapter *adapter, bool enable)
159 pci_read_config_dword(adapter->pdev, PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET,
161 enabled = reg & MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
163 if (!enabled && enable)
164 reg |= MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
165 else if (enabled && !enable)
166 reg &= ~MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
170 pci_write_config_dword(adapter->pdev,
171 PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET, reg);
174 static void be_intr_set(struct be_adapter *adapter, bool enable)
178 /* On lancer interrupts can't be controlled via this register */
179 if (lancer_chip(adapter))
182 if (adapter->eeh_error)
185 status = be_cmd_intr_set(adapter, enable);
187 be_reg_intr_set(adapter, enable);
190 static void be_rxq_notify(struct be_adapter *adapter, u16 qid, u16 posted)
194 val |= qid & DB_RQ_RING_ID_MASK;
195 val |= posted << DB_RQ_NUM_POSTED_SHIFT;
198 iowrite32(val, adapter->db + DB_RQ_OFFSET);
201 static void be_txq_notify(struct be_adapter *adapter, struct be_tx_obj *txo,
206 val |= txo->q.id & DB_TXULP_RING_ID_MASK;
207 val |= (posted & DB_TXULP_NUM_POSTED_MASK) << DB_TXULP_NUM_POSTED_SHIFT;
210 iowrite32(val, adapter->db + txo->db_offset);
213 static void be_eq_notify(struct be_adapter *adapter, u16 qid,
214 bool arm, bool clear_int, u16 num_popped)
218 val |= qid & DB_EQ_RING_ID_MASK;
219 val |= ((qid & DB_EQ_RING_ID_EXT_MASK) << DB_EQ_RING_ID_EXT_MASK_SHIFT);
221 if (adapter->eeh_error)
225 val |= 1 << DB_EQ_REARM_SHIFT;
227 val |= 1 << DB_EQ_CLR_SHIFT;
228 val |= 1 << DB_EQ_EVNT_SHIFT;
229 val |= num_popped << DB_EQ_NUM_POPPED_SHIFT;
230 iowrite32(val, adapter->db + DB_EQ_OFFSET);
233 void be_cq_notify(struct be_adapter *adapter, u16 qid, bool arm, u16 num_popped)
237 val |= qid & DB_CQ_RING_ID_MASK;
238 val |= ((qid & DB_CQ_RING_ID_EXT_MASK) <<
239 DB_CQ_RING_ID_EXT_MASK_SHIFT);
241 if (adapter->eeh_error)
245 val |= 1 << DB_CQ_REARM_SHIFT;
246 val |= num_popped << DB_CQ_NUM_POPPED_SHIFT;
247 iowrite32(val, adapter->db + DB_CQ_OFFSET);
250 static int be_mac_addr_set(struct net_device *netdev, void *p)
252 struct be_adapter *adapter = netdev_priv(netdev);
253 struct device *dev = &adapter->pdev->dev;
254 struct sockaddr *addr = p;
257 u32 old_pmac_id = adapter->pmac_id[0], curr_pmac_id = 0;
259 if (!is_valid_ether_addr(addr->sa_data))
260 return -EADDRNOTAVAIL;
262 /* Proceed further only if, User provided MAC is different
265 if (ether_addr_equal(addr->sa_data, netdev->dev_addr))
268 /* The PMAC_ADD cmd may fail if the VF doesn't have FILTMGMT
269 * privilege or if PF did not provision the new MAC address.
270 * On BE3, this cmd will always fail if the VF doesn't have the
271 * FILTMGMT privilege. This failure is OK, only if the PF programmed
272 * the MAC for the VF.
274 status = be_cmd_pmac_add(adapter, (u8 *)addr->sa_data,
275 adapter->if_handle, &adapter->pmac_id[0], 0);
277 curr_pmac_id = adapter->pmac_id[0];
279 /* Delete the old programmed MAC. This call may fail if the
280 * old MAC was already deleted by the PF driver.
282 if (adapter->pmac_id[0] != old_pmac_id)
283 be_cmd_pmac_del(adapter, adapter->if_handle,
287 /* Decide if the new MAC is successfully activated only after
290 status = be_cmd_get_active_mac(adapter, curr_pmac_id, mac,
291 adapter->if_handle, true, 0);
295 /* The MAC change did not happen, either due to lack of privilege
296 * or PF didn't pre-provision.
298 if (!ether_addr_equal(addr->sa_data, mac)) {
303 memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
304 dev_info(dev, "MAC address changed to %pM\n", mac);
307 dev_warn(dev, "MAC address change to %pM failed\n", addr->sa_data);
311 /* BE2 supports only v0 cmd */
312 static void *hw_stats_from_cmd(struct be_adapter *adapter)
314 if (BE2_chip(adapter)) {
315 struct be_cmd_resp_get_stats_v0 *cmd = adapter->stats_cmd.va;
317 return &cmd->hw_stats;
318 } else if (BE3_chip(adapter)) {
319 struct be_cmd_resp_get_stats_v1 *cmd = adapter->stats_cmd.va;
321 return &cmd->hw_stats;
323 struct be_cmd_resp_get_stats_v2 *cmd = adapter->stats_cmd.va;
325 return &cmd->hw_stats;
329 /* BE2 supports only v0 cmd */
330 static void *be_erx_stats_from_cmd(struct be_adapter *adapter)
332 if (BE2_chip(adapter)) {
333 struct be_hw_stats_v0 *hw_stats = hw_stats_from_cmd(adapter);
335 return &hw_stats->erx;
336 } else if (BE3_chip(adapter)) {
337 struct be_hw_stats_v1 *hw_stats = hw_stats_from_cmd(adapter);
339 return &hw_stats->erx;
341 struct be_hw_stats_v2 *hw_stats = hw_stats_from_cmd(adapter);
343 return &hw_stats->erx;
347 static void populate_be_v0_stats(struct be_adapter *adapter)
349 struct be_hw_stats_v0 *hw_stats = hw_stats_from_cmd(adapter);
350 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
351 struct be_rxf_stats_v0 *rxf_stats = &hw_stats->rxf;
352 struct be_port_rxf_stats_v0 *port_stats =
353 &rxf_stats->port[adapter->port_num];
354 struct be_drv_stats *drvs = &adapter->drv_stats;
356 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
357 drvs->rx_pause_frames = port_stats->rx_pause_frames;
358 drvs->rx_crc_errors = port_stats->rx_crc_errors;
359 drvs->rx_control_frames = port_stats->rx_control_frames;
360 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
361 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
362 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
363 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
364 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
365 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
366 drvs->rxpp_fifo_overflow_drop = port_stats->rx_fifo_overflow;
367 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
368 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
369 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
370 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
371 drvs->rx_input_fifo_overflow_drop = port_stats->rx_input_fifo_overflow;
372 drvs->rx_dropped_header_too_small =
373 port_stats->rx_dropped_header_too_small;
374 drvs->rx_address_filtered =
375 port_stats->rx_address_filtered +
376 port_stats->rx_vlan_filtered;
377 drvs->rx_alignment_symbol_errors =
378 port_stats->rx_alignment_symbol_errors;
380 drvs->tx_pauseframes = port_stats->tx_pauseframes;
381 drvs->tx_controlframes = port_stats->tx_controlframes;
383 if (adapter->port_num)
384 drvs->jabber_events = rxf_stats->port1_jabber_events;
386 drvs->jabber_events = rxf_stats->port0_jabber_events;
387 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
388 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
389 drvs->forwarded_packets = rxf_stats->forwarded_packets;
390 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
391 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
392 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
393 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
396 static void populate_be_v1_stats(struct be_adapter *adapter)
398 struct be_hw_stats_v1 *hw_stats = hw_stats_from_cmd(adapter);
399 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
400 struct be_rxf_stats_v1 *rxf_stats = &hw_stats->rxf;
401 struct be_port_rxf_stats_v1 *port_stats =
402 &rxf_stats->port[adapter->port_num];
403 struct be_drv_stats *drvs = &adapter->drv_stats;
405 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
406 drvs->pmem_fifo_overflow_drop = port_stats->pmem_fifo_overflow_drop;
407 drvs->rx_priority_pause_frames = port_stats->rx_priority_pause_frames;
408 drvs->rx_pause_frames = port_stats->rx_pause_frames;
409 drvs->rx_crc_errors = port_stats->rx_crc_errors;
410 drvs->rx_control_frames = port_stats->rx_control_frames;
411 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
412 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
413 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
414 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
415 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
416 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
417 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
418 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
419 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
420 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
421 drvs->rx_dropped_header_too_small =
422 port_stats->rx_dropped_header_too_small;
423 drvs->rx_input_fifo_overflow_drop =
424 port_stats->rx_input_fifo_overflow_drop;
425 drvs->rx_address_filtered = port_stats->rx_address_filtered;
426 drvs->rx_alignment_symbol_errors =
427 port_stats->rx_alignment_symbol_errors;
428 drvs->rxpp_fifo_overflow_drop = port_stats->rxpp_fifo_overflow_drop;
429 drvs->tx_pauseframes = port_stats->tx_pauseframes;
430 drvs->tx_controlframes = port_stats->tx_controlframes;
431 drvs->tx_priority_pauseframes = port_stats->tx_priority_pauseframes;
432 drvs->jabber_events = port_stats->jabber_events;
433 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
434 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
435 drvs->forwarded_packets = rxf_stats->forwarded_packets;
436 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
437 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
438 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
439 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
442 static void populate_be_v2_stats(struct be_adapter *adapter)
444 struct be_hw_stats_v2 *hw_stats = hw_stats_from_cmd(adapter);
445 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
446 struct be_rxf_stats_v2 *rxf_stats = &hw_stats->rxf;
447 struct be_port_rxf_stats_v2 *port_stats =
448 &rxf_stats->port[adapter->port_num];
449 struct be_drv_stats *drvs = &adapter->drv_stats;
451 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
452 drvs->pmem_fifo_overflow_drop = port_stats->pmem_fifo_overflow_drop;
453 drvs->rx_priority_pause_frames = port_stats->rx_priority_pause_frames;
454 drvs->rx_pause_frames = port_stats->rx_pause_frames;
455 drvs->rx_crc_errors = port_stats->rx_crc_errors;
456 drvs->rx_control_frames = port_stats->rx_control_frames;
457 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
458 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
459 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
460 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
461 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
462 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
463 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
464 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
465 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
466 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
467 drvs->rx_dropped_header_too_small =
468 port_stats->rx_dropped_header_too_small;
469 drvs->rx_input_fifo_overflow_drop =
470 port_stats->rx_input_fifo_overflow_drop;
471 drvs->rx_address_filtered = port_stats->rx_address_filtered;
472 drvs->rx_alignment_symbol_errors =
473 port_stats->rx_alignment_symbol_errors;
474 drvs->rxpp_fifo_overflow_drop = port_stats->rxpp_fifo_overflow_drop;
475 drvs->tx_pauseframes = port_stats->tx_pauseframes;
476 drvs->tx_controlframes = port_stats->tx_controlframes;
477 drvs->tx_priority_pauseframes = port_stats->tx_priority_pauseframes;
478 drvs->jabber_events = port_stats->jabber_events;
479 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
480 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
481 drvs->forwarded_packets = rxf_stats->forwarded_packets;
482 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
483 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
484 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
485 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
486 if (be_roce_supported(adapter)) {
487 drvs->rx_roce_bytes_lsd = port_stats->roce_bytes_received_lsd;
488 drvs->rx_roce_bytes_msd = port_stats->roce_bytes_received_msd;
489 drvs->rx_roce_frames = port_stats->roce_frames_received;
490 drvs->roce_drops_crc = port_stats->roce_drops_crc;
491 drvs->roce_drops_payload_len =
492 port_stats->roce_drops_payload_len;
496 static void populate_lancer_stats(struct be_adapter *adapter)
498 struct be_drv_stats *drvs = &adapter->drv_stats;
499 struct lancer_pport_stats *pport_stats = pport_stats_from_cmd(adapter);
501 be_dws_le_to_cpu(pport_stats, sizeof(*pport_stats));
502 drvs->rx_pause_frames = pport_stats->rx_pause_frames_lo;
503 drvs->rx_crc_errors = pport_stats->rx_crc_errors_lo;
504 drvs->rx_control_frames = pport_stats->rx_control_frames_lo;
505 drvs->rx_in_range_errors = pport_stats->rx_in_range_errors;
506 drvs->rx_frame_too_long = pport_stats->rx_frames_too_long_lo;
507 drvs->rx_dropped_runt = pport_stats->rx_dropped_runt;
508 drvs->rx_ip_checksum_errs = pport_stats->rx_ip_checksum_errors;
509 drvs->rx_tcp_checksum_errs = pport_stats->rx_tcp_checksum_errors;
510 drvs->rx_udp_checksum_errs = pport_stats->rx_udp_checksum_errors;
511 drvs->rx_dropped_tcp_length =
512 pport_stats->rx_dropped_invalid_tcp_length;
513 drvs->rx_dropped_too_small = pport_stats->rx_dropped_too_small;
514 drvs->rx_dropped_too_short = pport_stats->rx_dropped_too_short;
515 drvs->rx_out_range_errors = pport_stats->rx_out_of_range_errors;
516 drvs->rx_dropped_header_too_small =
517 pport_stats->rx_dropped_header_too_small;
518 drvs->rx_input_fifo_overflow_drop = pport_stats->rx_fifo_overflow;
519 drvs->rx_address_filtered =
520 pport_stats->rx_address_filtered +
521 pport_stats->rx_vlan_filtered;
522 drvs->rx_alignment_symbol_errors = pport_stats->rx_symbol_errors_lo;
523 drvs->rxpp_fifo_overflow_drop = pport_stats->rx_fifo_overflow;
524 drvs->tx_pauseframes = pport_stats->tx_pause_frames_lo;
525 drvs->tx_controlframes = pport_stats->tx_control_frames_lo;
526 drvs->jabber_events = pport_stats->rx_jabbers;
527 drvs->forwarded_packets = pport_stats->num_forwards_lo;
528 drvs->rx_drops_mtu = pport_stats->rx_drops_mtu_lo;
529 drvs->rx_drops_too_many_frags =
530 pport_stats->rx_drops_too_many_frags_lo;
533 static void accumulate_16bit_val(u32 *acc, u16 val)
535 #define lo(x) (x & 0xFFFF)
536 #define hi(x) (x & 0xFFFF0000)
537 bool wrapped = val < lo(*acc);
538 u32 newacc = hi(*acc) + val;
542 ACCESS_ONCE(*acc) = newacc;
545 static void populate_erx_stats(struct be_adapter *adapter,
546 struct be_rx_obj *rxo, u32 erx_stat)
548 if (!BEx_chip(adapter))
549 rx_stats(rxo)->rx_drops_no_frags = erx_stat;
551 /* below erx HW counter can actually wrap around after
552 * 65535. Driver accumulates a 32-bit value
554 accumulate_16bit_val(&rx_stats(rxo)->rx_drops_no_frags,
558 void be_parse_stats(struct be_adapter *adapter)
560 struct be_erx_stats_v2 *erx = be_erx_stats_from_cmd(adapter);
561 struct be_rx_obj *rxo;
565 if (lancer_chip(adapter)) {
566 populate_lancer_stats(adapter);
568 if (BE2_chip(adapter))
569 populate_be_v0_stats(adapter);
570 else if (BE3_chip(adapter))
572 populate_be_v1_stats(adapter);
574 populate_be_v2_stats(adapter);
576 /* erx_v2 is longer than v0, v1. use v2 for v0, v1 access */
577 for_all_rx_queues(adapter, rxo, i) {
578 erx_stat = erx->rx_drops_no_fragments[rxo->q.id];
579 populate_erx_stats(adapter, rxo, erx_stat);
584 static struct rtnl_link_stats64 *be_get_stats64(struct net_device *netdev,
585 struct rtnl_link_stats64 *stats)
587 struct be_adapter *adapter = netdev_priv(netdev);
588 struct be_drv_stats *drvs = &adapter->drv_stats;
589 struct be_rx_obj *rxo;
590 struct be_tx_obj *txo;
595 for_all_rx_queues(adapter, rxo, i) {
596 const struct be_rx_stats *rx_stats = rx_stats(rxo);
599 start = u64_stats_fetch_begin_irq(&rx_stats->sync);
600 pkts = rx_stats(rxo)->rx_pkts;
601 bytes = rx_stats(rxo)->rx_bytes;
602 } while (u64_stats_fetch_retry_irq(&rx_stats->sync, start));
603 stats->rx_packets += pkts;
604 stats->rx_bytes += bytes;
605 stats->multicast += rx_stats(rxo)->rx_mcast_pkts;
606 stats->rx_dropped += rx_stats(rxo)->rx_drops_no_skbs +
607 rx_stats(rxo)->rx_drops_no_frags;
610 for_all_tx_queues(adapter, txo, i) {
611 const struct be_tx_stats *tx_stats = tx_stats(txo);
614 start = u64_stats_fetch_begin_irq(&tx_stats->sync);
615 pkts = tx_stats(txo)->tx_pkts;
616 bytes = tx_stats(txo)->tx_bytes;
617 } while (u64_stats_fetch_retry_irq(&tx_stats->sync, start));
618 stats->tx_packets += pkts;
619 stats->tx_bytes += bytes;
622 /* bad pkts received */
623 stats->rx_errors = drvs->rx_crc_errors +
624 drvs->rx_alignment_symbol_errors +
625 drvs->rx_in_range_errors +
626 drvs->rx_out_range_errors +
627 drvs->rx_frame_too_long +
628 drvs->rx_dropped_too_small +
629 drvs->rx_dropped_too_short +
630 drvs->rx_dropped_header_too_small +
631 drvs->rx_dropped_tcp_length +
632 drvs->rx_dropped_runt;
634 /* detailed rx errors */
635 stats->rx_length_errors = drvs->rx_in_range_errors +
636 drvs->rx_out_range_errors +
637 drvs->rx_frame_too_long;
639 stats->rx_crc_errors = drvs->rx_crc_errors;
641 /* frame alignment errors */
642 stats->rx_frame_errors = drvs->rx_alignment_symbol_errors;
644 /* receiver fifo overrun */
645 /* drops_no_pbuf is no per i/f, it's per BE card */
646 stats->rx_fifo_errors = drvs->rxpp_fifo_overflow_drop +
647 drvs->rx_input_fifo_overflow_drop +
648 drvs->rx_drops_no_pbuf;
652 void be_link_status_update(struct be_adapter *adapter, u8 link_status)
654 struct net_device *netdev = adapter->netdev;
656 if (!(adapter->flags & BE_FLAGS_LINK_STATUS_INIT)) {
657 netif_carrier_off(netdev);
658 adapter->flags |= BE_FLAGS_LINK_STATUS_INIT;
662 netif_carrier_on(netdev);
664 netif_carrier_off(netdev);
667 static void be_tx_stats_update(struct be_tx_obj *txo, struct sk_buff *skb)
669 struct be_tx_stats *stats = tx_stats(txo);
671 u64_stats_update_begin(&stats->sync);
673 stats->tx_bytes += skb->len;
674 stats->tx_pkts += (skb_shinfo(skb)->gso_segs ? : 1);
675 u64_stats_update_end(&stats->sync);
678 /* Returns number of WRBs needed for the skb */
679 static u32 skb_wrb_cnt(struct sk_buff *skb)
681 /* +1 for the header wrb */
682 return 1 + (skb_headlen(skb) ? 1 : 0) + skb_shinfo(skb)->nr_frags;
685 static inline void wrb_fill(struct be_eth_wrb *wrb, u64 addr, int len)
687 wrb->frag_pa_hi = cpu_to_le32(upper_32_bits(addr));
688 wrb->frag_pa_lo = cpu_to_le32(lower_32_bits(addr));
689 wrb->frag_len = cpu_to_le32(len & ETH_WRB_FRAG_LEN_MASK);
693 /* A dummy wrb is just all zeros. Using a separate routine for dummy-wrb
694 * to avoid the swap and shift/mask operations in wrb_fill().
696 static inline void wrb_fill_dummy(struct be_eth_wrb *wrb)
704 static inline u16 be_get_tx_vlan_tag(struct be_adapter *adapter,
710 vlan_tag = skb_vlan_tag_get(skb);
711 vlan_prio = (vlan_tag & VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
712 /* If vlan priority provided by OS is NOT in available bmap */
713 if (!(adapter->vlan_prio_bmap & (1 << vlan_prio)))
714 vlan_tag = (vlan_tag & ~VLAN_PRIO_MASK) |
715 adapter->recommended_prio;
720 /* Used only for IP tunnel packets */
721 static u16 skb_inner_ip_proto(struct sk_buff *skb)
723 return (inner_ip_hdr(skb)->version == 4) ?
724 inner_ip_hdr(skb)->protocol : inner_ipv6_hdr(skb)->nexthdr;
727 static u16 skb_ip_proto(struct sk_buff *skb)
729 return (ip_hdr(skb)->version == 4) ?
730 ip_hdr(skb)->protocol : ipv6_hdr(skb)->nexthdr;
733 static inline bool be_is_txq_full(struct be_tx_obj *txo)
735 return atomic_read(&txo->q.used) + BE_MAX_TX_FRAG_COUNT >= txo->q.len;
738 static inline bool be_can_txq_wake(struct be_tx_obj *txo)
740 return atomic_read(&txo->q.used) < txo->q.len / 2;
743 static inline bool be_is_tx_compl_pending(struct be_tx_obj *txo)
745 return atomic_read(&txo->q.used) > txo->pend_wrb_cnt;
748 static void be_get_wrb_params_from_skb(struct be_adapter *adapter,
750 struct be_wrb_params *wrb_params)
754 if (skb_is_gso(skb)) {
755 BE_WRB_F_SET(wrb_params->features, LSO, 1);
756 wrb_params->lso_mss = skb_shinfo(skb)->gso_size;
757 if (skb_is_gso_v6(skb) && !lancer_chip(adapter))
758 BE_WRB_F_SET(wrb_params->features, LSO6, 1);
759 } else if (skb->ip_summed == CHECKSUM_PARTIAL) {
760 if (skb->encapsulation) {
761 BE_WRB_F_SET(wrb_params->features, IPCS, 1);
762 proto = skb_inner_ip_proto(skb);
764 proto = skb_ip_proto(skb);
766 if (proto == IPPROTO_TCP)
767 BE_WRB_F_SET(wrb_params->features, TCPCS, 1);
768 else if (proto == IPPROTO_UDP)
769 BE_WRB_F_SET(wrb_params->features, UDPCS, 1);
772 if (skb_vlan_tag_present(skb)) {
773 BE_WRB_F_SET(wrb_params->features, VLAN, 1);
774 wrb_params->vlan_tag = be_get_tx_vlan_tag(adapter, skb);
777 BE_WRB_F_SET(wrb_params->features, CRC, 1);
780 static void wrb_fill_hdr(struct be_adapter *adapter,
781 struct be_eth_hdr_wrb *hdr,
782 struct be_wrb_params *wrb_params,
785 memset(hdr, 0, sizeof(*hdr));
787 SET_TX_WRB_HDR_BITS(crc, hdr,
788 BE_WRB_F_GET(wrb_params->features, CRC));
789 SET_TX_WRB_HDR_BITS(ipcs, hdr,
790 BE_WRB_F_GET(wrb_params->features, IPCS));
791 SET_TX_WRB_HDR_BITS(tcpcs, hdr,
792 BE_WRB_F_GET(wrb_params->features, TCPCS));
793 SET_TX_WRB_HDR_BITS(udpcs, hdr,
794 BE_WRB_F_GET(wrb_params->features, UDPCS));
796 SET_TX_WRB_HDR_BITS(lso, hdr,
797 BE_WRB_F_GET(wrb_params->features, LSO));
798 SET_TX_WRB_HDR_BITS(lso6, hdr,
799 BE_WRB_F_GET(wrb_params->features, LSO6));
800 SET_TX_WRB_HDR_BITS(lso_mss, hdr, wrb_params->lso_mss);
802 /* Hack to skip HW VLAN tagging needs evt = 1, compl = 0. When this
803 * hack is not needed, the evt bit is set while ringing DB.
805 SET_TX_WRB_HDR_BITS(event, hdr,
806 BE_WRB_F_GET(wrb_params->features, VLAN_SKIP_HW));
807 SET_TX_WRB_HDR_BITS(vlan, hdr,
808 BE_WRB_F_GET(wrb_params->features, VLAN));
809 SET_TX_WRB_HDR_BITS(vlan_tag, hdr, wrb_params->vlan_tag);
811 SET_TX_WRB_HDR_BITS(num_wrb, hdr, skb_wrb_cnt(skb));
812 SET_TX_WRB_HDR_BITS(len, hdr, skb->len);
815 static void unmap_tx_frag(struct device *dev, struct be_eth_wrb *wrb,
819 u32 frag_len = le32_to_cpu(wrb->frag_len);
822 dma = (u64)le32_to_cpu(wrb->frag_pa_hi) << 32 |
823 (u64)le32_to_cpu(wrb->frag_pa_lo);
826 dma_unmap_single(dev, dma, frag_len, DMA_TO_DEVICE);
828 dma_unmap_page(dev, dma, frag_len, DMA_TO_DEVICE);
832 /* Grab a WRB header for xmit */
833 static u16 be_tx_get_wrb_hdr(struct be_tx_obj *txo)
835 u16 head = txo->q.head;
837 queue_head_inc(&txo->q);
841 /* Set up the WRB header for xmit */
842 static void be_tx_setup_wrb_hdr(struct be_adapter *adapter,
843 struct be_tx_obj *txo,
844 struct be_wrb_params *wrb_params,
845 struct sk_buff *skb, u16 head)
847 u32 num_frags = skb_wrb_cnt(skb);
848 struct be_queue_info *txq = &txo->q;
849 struct be_eth_hdr_wrb *hdr = queue_index_node(txq, head);
851 wrb_fill_hdr(adapter, hdr, wrb_params, skb);
852 be_dws_cpu_to_le(hdr, sizeof(*hdr));
854 BUG_ON(txo->sent_skb_list[head]);
855 txo->sent_skb_list[head] = skb;
856 txo->last_req_hdr = head;
857 atomic_add(num_frags, &txq->used);
858 txo->last_req_wrb_cnt = num_frags;
859 txo->pend_wrb_cnt += num_frags;
862 /* Setup a WRB fragment (buffer descriptor) for xmit */
863 static void be_tx_setup_wrb_frag(struct be_tx_obj *txo, dma_addr_t busaddr,
866 struct be_eth_wrb *wrb;
867 struct be_queue_info *txq = &txo->q;
869 wrb = queue_head_node(txq);
870 wrb_fill(wrb, busaddr, len);
874 /* Bring the queue back to the state it was in before be_xmit_enqueue() routine
875 * was invoked. The producer index is restored to the previous packet and the
876 * WRBs of the current packet are unmapped. Invoked to handle tx setup errors.
878 static void be_xmit_restore(struct be_adapter *adapter,
879 struct be_tx_obj *txo, u16 head, bool map_single,
883 struct be_eth_wrb *wrb;
884 struct be_queue_info *txq = &txo->q;
886 dev = &adapter->pdev->dev;
889 /* skip the first wrb (hdr); it's not mapped */
892 wrb = queue_head_node(txq);
893 unmap_tx_frag(dev, wrb, map_single);
895 copied -= le32_to_cpu(wrb->frag_len);
902 /* Enqueue the given packet for transmit. This routine allocates WRBs for the
903 * packet, dma maps the packet buffers and sets up the WRBs. Returns the number
904 * of WRBs used up by the packet.
906 static u32 be_xmit_enqueue(struct be_adapter *adapter, struct be_tx_obj *txo,
908 struct be_wrb_params *wrb_params)
910 u32 i, copied = 0, wrb_cnt = skb_wrb_cnt(skb);
911 struct device *dev = &adapter->pdev->dev;
912 struct be_queue_info *txq = &txo->q;
913 bool map_single = false;
914 u16 head = txq->head;
918 head = be_tx_get_wrb_hdr(txo);
920 if (skb->len > skb->data_len) {
921 len = skb_headlen(skb);
923 busaddr = dma_map_single(dev, skb->data, len, DMA_TO_DEVICE);
924 if (dma_mapping_error(dev, busaddr))
927 be_tx_setup_wrb_frag(txo, busaddr, len);
931 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
932 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i];
933 len = skb_frag_size(frag);
935 busaddr = skb_frag_dma_map(dev, frag, 0, len, DMA_TO_DEVICE);
936 if (dma_mapping_error(dev, busaddr))
938 be_tx_setup_wrb_frag(txo, busaddr, len);
942 be_tx_setup_wrb_hdr(adapter, txo, wrb_params, skb, head);
944 be_tx_stats_update(txo, skb);
948 adapter->drv_stats.dma_map_errors++;
949 be_xmit_restore(adapter, txo, head, map_single, copied);
953 static inline int qnq_async_evt_rcvd(struct be_adapter *adapter)
955 return adapter->flags & BE_FLAGS_QNQ_ASYNC_EVT_RCVD;
958 static struct sk_buff *be_insert_vlan_in_pkt(struct be_adapter *adapter,
965 skb = skb_share_check(skb, GFP_ATOMIC);
969 if (skb_vlan_tag_present(skb))
970 vlan_tag = be_get_tx_vlan_tag(adapter, skb);
972 if (qnq_async_evt_rcvd(adapter) && adapter->pvid) {
974 vlan_tag = adapter->pvid;
975 /* f/w workaround to set skip_hw_vlan = 1, informs the F/W to
976 * skip VLAN insertion
978 BE_WRB_F_SET(wrb_params->features, VLAN_SKIP_HW, 1);
982 skb = vlan_insert_tag_set_proto(skb, htons(ETH_P_8021Q),
989 /* Insert the outer VLAN, if any */
990 if (adapter->qnq_vid) {
991 vlan_tag = adapter->qnq_vid;
992 skb = vlan_insert_tag_set_proto(skb, htons(ETH_P_8021Q),
996 BE_WRB_F_SET(wrb_params->features, VLAN_SKIP_HW, 1);
1002 static bool be_ipv6_exthdr_check(struct sk_buff *skb)
1004 struct ethhdr *eh = (struct ethhdr *)skb->data;
1005 u16 offset = ETH_HLEN;
1007 if (eh->h_proto == htons(ETH_P_IPV6)) {
1008 struct ipv6hdr *ip6h = (struct ipv6hdr *)(skb->data + offset);
1010 offset += sizeof(struct ipv6hdr);
1011 if (ip6h->nexthdr != NEXTHDR_TCP &&
1012 ip6h->nexthdr != NEXTHDR_UDP) {
1013 struct ipv6_opt_hdr *ehdr =
1014 (struct ipv6_opt_hdr *)(skb->data + offset);
1016 /* offending pkt: 2nd byte following IPv6 hdr is 0xff */
1017 if (ehdr->hdrlen == 0xff)
1024 static int be_vlan_tag_tx_chk(struct be_adapter *adapter, struct sk_buff *skb)
1026 return skb_vlan_tag_present(skb) || adapter->pvid || adapter->qnq_vid;
1029 static int be_ipv6_tx_stall_chk(struct be_adapter *adapter, struct sk_buff *skb)
1031 return BE3_chip(adapter) && be_ipv6_exthdr_check(skb);
1034 static struct sk_buff *be_lancer_xmit_workarounds(struct be_adapter *adapter,
1035 struct sk_buff *skb,
1036 struct be_wrb_params
1039 struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
1040 unsigned int eth_hdr_len;
1043 /* For padded packets, BE HW modifies tot_len field in IP header
1044 * incorrecly when VLAN tag is inserted by HW.
1045 * For padded packets, Lancer computes incorrect checksum.
1047 eth_hdr_len = ntohs(skb->protocol) == ETH_P_8021Q ?
1048 VLAN_ETH_HLEN : ETH_HLEN;
1049 if (skb->len <= 60 &&
1050 (lancer_chip(adapter) || skb_vlan_tag_present(skb)) &&
1052 ip = (struct iphdr *)ip_hdr(skb);
1053 pskb_trim(skb, eth_hdr_len + ntohs(ip->tot_len));
1056 /* If vlan tag is already inlined in the packet, skip HW VLAN
1057 * tagging in pvid-tagging mode
1059 if (be_pvid_tagging_enabled(adapter) &&
1060 veh->h_vlan_proto == htons(ETH_P_8021Q))
1061 BE_WRB_F_SET(wrb_params->features, VLAN_SKIP_HW, 1);
1063 /* HW has a bug wherein it will calculate CSUM for VLAN
1064 * pkts even though it is disabled.
1065 * Manually insert VLAN in pkt.
1067 if (skb->ip_summed != CHECKSUM_PARTIAL &&
1068 skb_vlan_tag_present(skb)) {
1069 skb = be_insert_vlan_in_pkt(adapter, skb, wrb_params);
1074 /* HW may lockup when VLAN HW tagging is requested on
1075 * certain ipv6 packets. Drop such pkts if the HW workaround to
1076 * skip HW tagging is not enabled by FW.
1078 if (unlikely(be_ipv6_tx_stall_chk(adapter, skb) &&
1079 (adapter->pvid || adapter->qnq_vid) &&
1080 !qnq_async_evt_rcvd(adapter)))
1083 /* Manual VLAN tag insertion to prevent:
1084 * ASIC lockup when the ASIC inserts VLAN tag into
1085 * certain ipv6 packets. Insert VLAN tags in driver,
1086 * and set event, completion, vlan bits accordingly
1089 if (be_ipv6_tx_stall_chk(adapter, skb) &&
1090 be_vlan_tag_tx_chk(adapter, skb)) {
1091 skb = be_insert_vlan_in_pkt(adapter, skb, wrb_params);
1098 dev_kfree_skb_any(skb);
1103 static struct sk_buff *be_xmit_workarounds(struct be_adapter *adapter,
1104 struct sk_buff *skb,
1105 struct be_wrb_params *wrb_params)
1107 /* Lancer, SH-R ASICs have a bug wherein Packets that are 32 bytes or
1108 * less may cause a transmit stall on that port. So the work-around is
1109 * to pad short packets (<= 32 bytes) to a 36-byte length.
1111 if (unlikely(!BEx_chip(adapter) && skb->len <= 32)) {
1112 if (skb_put_padto(skb, 36))
1116 if (BEx_chip(adapter) || lancer_chip(adapter)) {
1117 skb = be_lancer_xmit_workarounds(adapter, skb, wrb_params);
1125 static void be_xmit_flush(struct be_adapter *adapter, struct be_tx_obj *txo)
1127 struct be_queue_info *txq = &txo->q;
1128 struct be_eth_hdr_wrb *hdr = queue_index_node(txq, txo->last_req_hdr);
1130 /* Mark the last request eventable if it hasn't been marked already */
1131 if (!(hdr->dw[2] & cpu_to_le32(TX_HDR_WRB_EVT)))
1132 hdr->dw[2] |= cpu_to_le32(TX_HDR_WRB_EVT | TX_HDR_WRB_COMPL);
1134 /* compose a dummy wrb if there are odd set of wrbs to notify */
1135 if (!lancer_chip(adapter) && (txo->pend_wrb_cnt & 1)) {
1136 wrb_fill_dummy(queue_head_node(txq));
1137 queue_head_inc(txq);
1138 atomic_inc(&txq->used);
1139 txo->pend_wrb_cnt++;
1140 hdr->dw[2] &= ~cpu_to_le32(TX_HDR_WRB_NUM_MASK <<
1141 TX_HDR_WRB_NUM_SHIFT);
1142 hdr->dw[2] |= cpu_to_le32((txo->last_req_wrb_cnt + 1) <<
1143 TX_HDR_WRB_NUM_SHIFT);
1145 be_txq_notify(adapter, txo, txo->pend_wrb_cnt);
1146 txo->pend_wrb_cnt = 0;
1149 static netdev_tx_t be_xmit(struct sk_buff *skb, struct net_device *netdev)
1151 struct be_adapter *adapter = netdev_priv(netdev);
1152 u16 q_idx = skb_get_queue_mapping(skb);
1153 struct be_tx_obj *txo = &adapter->tx_obj[q_idx];
1154 struct be_wrb_params wrb_params = { 0 };
1155 bool flush = !skb->xmit_more;
1158 skb = be_xmit_workarounds(adapter, skb, &wrb_params);
1162 be_get_wrb_params_from_skb(adapter, skb, &wrb_params);
1164 wrb_cnt = be_xmit_enqueue(adapter, txo, skb, &wrb_params);
1165 if (unlikely(!wrb_cnt)) {
1166 dev_kfree_skb_any(skb);
1170 if (be_is_txq_full(txo)) {
1171 netif_stop_subqueue(netdev, q_idx);
1172 tx_stats(txo)->tx_stops++;
1175 if (flush || __netif_subqueue_stopped(netdev, q_idx))
1176 be_xmit_flush(adapter, txo);
1178 return NETDEV_TX_OK;
1180 tx_stats(txo)->tx_drv_drops++;
1181 /* Flush the already enqueued tx requests */
1182 if (flush && txo->pend_wrb_cnt)
1183 be_xmit_flush(adapter, txo);
1185 return NETDEV_TX_OK;
1188 static int be_change_mtu(struct net_device *netdev, int new_mtu)
1190 struct be_adapter *adapter = netdev_priv(netdev);
1191 struct device *dev = &adapter->pdev->dev;
1193 if (new_mtu < BE_MIN_MTU || new_mtu > BE_MAX_MTU) {
1194 dev_info(dev, "MTU must be between %d and %d bytes\n",
1195 BE_MIN_MTU, BE_MAX_MTU);
1199 dev_info(dev, "MTU changed from %d to %d bytes\n",
1200 netdev->mtu, new_mtu);
1201 netdev->mtu = new_mtu;
1205 static inline bool be_in_all_promisc(struct be_adapter *adapter)
1207 return (adapter->if_flags & BE_IF_FLAGS_ALL_PROMISCUOUS) ==
1208 BE_IF_FLAGS_ALL_PROMISCUOUS;
1211 static int be_set_vlan_promisc(struct be_adapter *adapter)
1213 struct device *dev = &adapter->pdev->dev;
1216 if (adapter->if_flags & BE_IF_FLAGS_VLAN_PROMISCUOUS)
1219 status = be_cmd_rx_filter(adapter, BE_IF_FLAGS_VLAN_PROMISCUOUS, ON);
1221 dev_info(dev, "Enabled VLAN promiscuous mode\n");
1222 adapter->if_flags |= BE_IF_FLAGS_VLAN_PROMISCUOUS;
1224 dev_err(dev, "Failed to enable VLAN promiscuous mode\n");
1229 static int be_clear_vlan_promisc(struct be_adapter *adapter)
1231 struct device *dev = &adapter->pdev->dev;
1234 status = be_cmd_rx_filter(adapter, BE_IF_FLAGS_VLAN_PROMISCUOUS, OFF);
1236 dev_info(dev, "Disabling VLAN promiscuous mode\n");
1237 adapter->if_flags &= ~BE_IF_FLAGS_VLAN_PROMISCUOUS;
1243 * A max of 64 (BE_NUM_VLANS_SUPPORTED) vlans can be configured in BE.
1244 * If the user configures more, place BE in vlan promiscuous mode.
1246 static int be_vid_config(struct be_adapter *adapter)
1248 struct device *dev = &adapter->pdev->dev;
1249 u16 vids[BE_NUM_VLANS_SUPPORTED];
1253 /* No need to further configure vids if in promiscuous mode */
1254 if (be_in_all_promisc(adapter))
1257 if (adapter->vlans_added > be_max_vlans(adapter))
1258 return be_set_vlan_promisc(adapter);
1260 /* Construct VLAN Table to give to HW */
1261 for_each_set_bit(i, adapter->vids, VLAN_N_VID)
1262 vids[num++] = cpu_to_le16(i);
1264 status = be_cmd_vlan_config(adapter, adapter->if_handle, vids, num, 0);
1266 dev_err(dev, "Setting HW VLAN filtering failed\n");
1267 /* Set to VLAN promisc mode as setting VLAN filter failed */
1268 if (addl_status(status) ==
1269 MCC_ADDL_STATUS_INSUFFICIENT_RESOURCES)
1270 return be_set_vlan_promisc(adapter);
1271 } else if (adapter->if_flags & BE_IF_FLAGS_VLAN_PROMISCUOUS) {
1272 status = be_clear_vlan_promisc(adapter);
1277 static int be_vlan_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
1279 struct be_adapter *adapter = netdev_priv(netdev);
1282 /* Packets with VID 0 are always received by Lancer by default */
1283 if (lancer_chip(adapter) && vid == 0)
1286 if (test_bit(vid, adapter->vids))
1289 set_bit(vid, adapter->vids);
1290 adapter->vlans_added++;
1292 status = be_vid_config(adapter);
1294 adapter->vlans_added--;
1295 clear_bit(vid, adapter->vids);
1301 static int be_vlan_rem_vid(struct net_device *netdev, __be16 proto, u16 vid)
1303 struct be_adapter *adapter = netdev_priv(netdev);
1305 /* Packets with VID 0 are always received by Lancer by default */
1306 if (lancer_chip(adapter) && vid == 0)
1309 clear_bit(vid, adapter->vids);
1310 adapter->vlans_added--;
1312 return be_vid_config(adapter);
1315 static void be_clear_all_promisc(struct be_adapter *adapter)
1317 be_cmd_rx_filter(adapter, BE_IF_FLAGS_ALL_PROMISCUOUS, OFF);
1318 adapter->if_flags &= ~BE_IF_FLAGS_ALL_PROMISCUOUS;
1321 static void be_set_all_promisc(struct be_adapter *adapter)
1323 be_cmd_rx_filter(adapter, BE_IF_FLAGS_ALL_PROMISCUOUS, ON);
1324 adapter->if_flags |= BE_IF_FLAGS_ALL_PROMISCUOUS;
1327 static void be_set_mc_promisc(struct be_adapter *adapter)
1331 if (adapter->if_flags & BE_IF_FLAGS_MCAST_PROMISCUOUS)
1334 status = be_cmd_rx_filter(adapter, BE_IF_FLAGS_MCAST_PROMISCUOUS, ON);
1336 adapter->if_flags |= BE_IF_FLAGS_MCAST_PROMISCUOUS;
1339 static void be_set_mc_list(struct be_adapter *adapter)
1343 status = be_cmd_rx_filter(adapter, BE_IF_FLAGS_MULTICAST, ON);
1345 adapter->if_flags &= ~BE_IF_FLAGS_MCAST_PROMISCUOUS;
1347 be_set_mc_promisc(adapter);
1350 static void be_set_uc_list(struct be_adapter *adapter)
1352 struct netdev_hw_addr *ha;
1353 int i = 1; /* First slot is claimed by the Primary MAC */
1355 for (; adapter->uc_macs > 0; adapter->uc_macs--, i++)
1356 be_cmd_pmac_del(adapter, adapter->if_handle,
1357 adapter->pmac_id[i], 0);
1359 if (netdev_uc_count(adapter->netdev) > be_max_uc(adapter)) {
1360 be_set_all_promisc(adapter);
1364 netdev_for_each_uc_addr(ha, adapter->netdev) {
1365 adapter->uc_macs++; /* First slot is for Primary MAC */
1366 be_cmd_pmac_add(adapter, (u8 *)ha->addr, adapter->if_handle,
1367 &adapter->pmac_id[adapter->uc_macs], 0);
1371 static void be_clear_uc_list(struct be_adapter *adapter)
1375 for (i = 1; i < (adapter->uc_macs + 1); i++)
1376 be_cmd_pmac_del(adapter, adapter->if_handle,
1377 adapter->pmac_id[i], 0);
1378 adapter->uc_macs = 0;
1381 static void be_set_rx_mode(struct net_device *netdev)
1383 struct be_adapter *adapter = netdev_priv(netdev);
1385 if (netdev->flags & IFF_PROMISC) {
1386 be_set_all_promisc(adapter);
1390 /* Interface was previously in promiscuous mode; disable it */
1391 if (be_in_all_promisc(adapter)) {
1392 be_clear_all_promisc(adapter);
1393 if (adapter->vlans_added)
1394 be_vid_config(adapter);
1397 /* Enable multicast promisc if num configured exceeds what we support */
1398 if (netdev->flags & IFF_ALLMULTI ||
1399 netdev_mc_count(netdev) > be_max_mc(adapter)) {
1400 be_set_mc_promisc(adapter);
1404 if (netdev_uc_count(netdev) != adapter->uc_macs)
1405 be_set_uc_list(adapter);
1407 be_set_mc_list(adapter);
1410 static int be_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
1412 struct be_adapter *adapter = netdev_priv(netdev);
1413 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1416 if (!sriov_enabled(adapter))
1419 if (!is_valid_ether_addr(mac) || vf >= adapter->num_vfs)
1422 /* Proceed further only if user provided MAC is different
1425 if (ether_addr_equal(mac, vf_cfg->mac_addr))
1428 if (BEx_chip(adapter)) {
1429 be_cmd_pmac_del(adapter, vf_cfg->if_handle, vf_cfg->pmac_id,
1432 status = be_cmd_pmac_add(adapter, mac, vf_cfg->if_handle,
1433 &vf_cfg->pmac_id, vf + 1);
1435 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
1440 dev_err(&adapter->pdev->dev, "MAC %pM set on VF %d Failed: %#x",
1442 return be_cmd_status(status);
1445 ether_addr_copy(vf_cfg->mac_addr, mac);
1450 static int be_get_vf_config(struct net_device *netdev, int vf,
1451 struct ifla_vf_info *vi)
1453 struct be_adapter *adapter = netdev_priv(netdev);
1454 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1456 if (!sriov_enabled(adapter))
1459 if (vf >= adapter->num_vfs)
1463 vi->max_tx_rate = vf_cfg->tx_rate;
1464 vi->min_tx_rate = 0;
1465 vi->vlan = vf_cfg->vlan_tag & VLAN_VID_MASK;
1466 vi->qos = vf_cfg->vlan_tag >> VLAN_PRIO_SHIFT;
1467 memcpy(&vi->mac, vf_cfg->mac_addr, ETH_ALEN);
1468 vi->linkstate = adapter->vf_cfg[vf].plink_tracking;
1473 static int be_set_vf_tvt(struct be_adapter *adapter, int vf, u16 vlan)
1475 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1476 u16 vids[BE_NUM_VLANS_SUPPORTED];
1477 int vf_if_id = vf_cfg->if_handle;
1480 /* Enable Transparent VLAN Tagging */
1481 status = be_cmd_set_hsw_config(adapter, vlan, vf + 1, vf_if_id, 0);
1485 /* Clear pre-programmed VLAN filters on VF if any, if TVT is enabled */
1487 status = be_cmd_vlan_config(adapter, vf_if_id, vids, 1, vf + 1);
1489 dev_info(&adapter->pdev->dev,
1490 "Cleared guest VLANs on VF%d", vf);
1492 /* After TVT is enabled, disallow VFs to program VLAN filters */
1493 if (vf_cfg->privileges & BE_PRIV_FILTMGMT) {
1494 status = be_cmd_set_fn_privileges(adapter, vf_cfg->privileges &
1495 ~BE_PRIV_FILTMGMT, vf + 1);
1497 vf_cfg->privileges &= ~BE_PRIV_FILTMGMT;
1502 static int be_clear_vf_tvt(struct be_adapter *adapter, int vf)
1504 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1505 struct device *dev = &adapter->pdev->dev;
1508 /* Reset Transparent VLAN Tagging. */
1509 status = be_cmd_set_hsw_config(adapter, BE_RESET_VLAN_TAG_ID, vf + 1,
1510 vf_cfg->if_handle, 0);
1514 /* Allow VFs to program VLAN filtering */
1515 if (!(vf_cfg->privileges & BE_PRIV_FILTMGMT)) {
1516 status = be_cmd_set_fn_privileges(adapter, vf_cfg->privileges |
1517 BE_PRIV_FILTMGMT, vf + 1);
1519 vf_cfg->privileges |= BE_PRIV_FILTMGMT;
1520 dev_info(dev, "VF%d: FILTMGMT priv enabled", vf);
1525 "Disable/re-enable i/f in VM to clear Transparent VLAN tag");
1529 static int be_set_vf_vlan(struct net_device *netdev, int vf, u16 vlan, u8 qos)
1531 struct be_adapter *adapter = netdev_priv(netdev);
1532 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1535 if (!sriov_enabled(adapter))
1538 if (vf >= adapter->num_vfs || vlan > 4095 || qos > 7)
1542 vlan |= qos << VLAN_PRIO_SHIFT;
1543 status = be_set_vf_tvt(adapter, vf, vlan);
1545 status = be_clear_vf_tvt(adapter, vf);
1549 dev_err(&adapter->pdev->dev,
1550 "VLAN %d config on VF %d failed : %#x\n", vlan, vf,
1552 return be_cmd_status(status);
1555 vf_cfg->vlan_tag = vlan;
1559 static int be_set_vf_tx_rate(struct net_device *netdev, int vf,
1560 int min_tx_rate, int max_tx_rate)
1562 struct be_adapter *adapter = netdev_priv(netdev);
1563 struct device *dev = &adapter->pdev->dev;
1564 int percent_rate, status = 0;
1568 if (!sriov_enabled(adapter))
1571 if (vf >= adapter->num_vfs)
1580 status = be_cmd_link_status_query(adapter, &link_speed,
1586 dev_err(dev, "TX-rate setting not allowed when link is down\n");
1591 if (max_tx_rate < 100 || max_tx_rate > link_speed) {
1592 dev_err(dev, "TX-rate must be between 100 and %d Mbps\n",
1598 /* On Skyhawk the QOS setting must be done only as a % value */
1599 percent_rate = link_speed / 100;
1600 if (skyhawk_chip(adapter) && (max_tx_rate % percent_rate)) {
1601 dev_err(dev, "TX-rate must be a multiple of %d Mbps\n",
1608 status = be_cmd_config_qos(adapter, max_tx_rate, link_speed, vf + 1);
1612 adapter->vf_cfg[vf].tx_rate = max_tx_rate;
1616 dev_err(dev, "TX-rate setting of %dMbps on VF%d failed\n",
1618 return be_cmd_status(status);
1621 static int be_set_vf_link_state(struct net_device *netdev, int vf,
1624 struct be_adapter *adapter = netdev_priv(netdev);
1627 if (!sriov_enabled(adapter))
1630 if (vf >= adapter->num_vfs)
1633 status = be_cmd_set_logical_link_config(adapter, link_state, vf+1);
1635 dev_err(&adapter->pdev->dev,
1636 "Link state change on VF %d failed: %#x\n", vf, status);
1637 return be_cmd_status(status);
1640 adapter->vf_cfg[vf].plink_tracking = link_state;
1645 static void be_aic_update(struct be_aic_obj *aic, u64 rx_pkts, u64 tx_pkts,
1648 aic->rx_pkts_prev = rx_pkts;
1649 aic->tx_reqs_prev = tx_pkts;
1653 static void be_eqd_update(struct be_adapter *adapter)
1655 struct be_set_eqd set_eqd[MAX_EVT_QS];
1656 int eqd, i, num = 0, start;
1657 struct be_aic_obj *aic;
1658 struct be_eq_obj *eqo;
1659 struct be_rx_obj *rxo;
1660 struct be_tx_obj *txo;
1661 u64 rx_pkts, tx_pkts;
1665 for_all_evt_queues(adapter, eqo, i) {
1666 aic = &adapter->aic_obj[eqo->idx];
1674 rxo = &adapter->rx_obj[eqo->idx];
1676 start = u64_stats_fetch_begin_irq(&rxo->stats.sync);
1677 rx_pkts = rxo->stats.rx_pkts;
1678 } while (u64_stats_fetch_retry_irq(&rxo->stats.sync, start));
1680 txo = &adapter->tx_obj[eqo->idx];
1682 start = u64_stats_fetch_begin_irq(&txo->stats.sync);
1683 tx_pkts = txo->stats.tx_reqs;
1684 } while (u64_stats_fetch_retry_irq(&txo->stats.sync, start));
1686 /* Skip, if wrapped around or first calculation */
1688 if (!aic->jiffies || time_before(now, aic->jiffies) ||
1689 rx_pkts < aic->rx_pkts_prev ||
1690 tx_pkts < aic->tx_reqs_prev) {
1691 be_aic_update(aic, rx_pkts, tx_pkts, now);
1695 delta = jiffies_to_msecs(now - aic->jiffies);
1696 pps = (((u32)(rx_pkts - aic->rx_pkts_prev) * 1000) / delta) +
1697 (((u32)(tx_pkts - aic->tx_reqs_prev) * 1000) / delta);
1698 eqd = (pps / 15000) << 2;
1702 eqd = min_t(u32, eqd, aic->max_eqd);
1703 eqd = max_t(u32, eqd, aic->min_eqd);
1705 be_aic_update(aic, rx_pkts, tx_pkts, now);
1707 if (eqd != aic->prev_eqd) {
1708 set_eqd[num].delay_multiplier = (eqd * 65)/100;
1709 set_eqd[num].eq_id = eqo->q.id;
1710 aic->prev_eqd = eqd;
1716 be_cmd_modify_eqd(adapter, set_eqd, num);
1719 static void be_rx_stats_update(struct be_rx_obj *rxo,
1720 struct be_rx_compl_info *rxcp)
1722 struct be_rx_stats *stats = rx_stats(rxo);
1724 u64_stats_update_begin(&stats->sync);
1726 stats->rx_bytes += rxcp->pkt_size;
1728 if (rxcp->pkt_type == BE_MULTICAST_PACKET)
1729 stats->rx_mcast_pkts++;
1731 stats->rx_compl_err++;
1732 u64_stats_update_end(&stats->sync);
1735 static inline bool csum_passed(struct be_rx_compl_info *rxcp)
1737 /* L4 checksum is not reliable for non TCP/UDP packets.
1738 * Also ignore ipcksm for ipv6 pkts
1740 return (rxcp->tcpf || rxcp->udpf) && rxcp->l4_csum &&
1741 (rxcp->ip_csum || rxcp->ipv6) && !rxcp->err;
1744 static struct be_rx_page_info *get_rx_page_info(struct be_rx_obj *rxo)
1746 struct be_adapter *adapter = rxo->adapter;
1747 struct be_rx_page_info *rx_page_info;
1748 struct be_queue_info *rxq = &rxo->q;
1749 u16 frag_idx = rxq->tail;
1751 rx_page_info = &rxo->page_info_tbl[frag_idx];
1752 BUG_ON(!rx_page_info->page);
1754 if (rx_page_info->last_frag) {
1755 dma_unmap_page(&adapter->pdev->dev,
1756 dma_unmap_addr(rx_page_info, bus),
1757 adapter->big_page_size, DMA_FROM_DEVICE);
1758 rx_page_info->last_frag = false;
1760 dma_sync_single_for_cpu(&adapter->pdev->dev,
1761 dma_unmap_addr(rx_page_info, bus),
1762 rx_frag_size, DMA_FROM_DEVICE);
1765 queue_tail_inc(rxq);
1766 atomic_dec(&rxq->used);
1767 return rx_page_info;
1770 /* Throwaway the data in the Rx completion */
1771 static void be_rx_compl_discard(struct be_rx_obj *rxo,
1772 struct be_rx_compl_info *rxcp)
1774 struct be_rx_page_info *page_info;
1775 u16 i, num_rcvd = rxcp->num_rcvd;
1777 for (i = 0; i < num_rcvd; i++) {
1778 page_info = get_rx_page_info(rxo);
1779 put_page(page_info->page);
1780 memset(page_info, 0, sizeof(*page_info));
1785 * skb_fill_rx_data forms a complete skb for an ether frame
1786 * indicated by rxcp.
1788 static void skb_fill_rx_data(struct be_rx_obj *rxo, struct sk_buff *skb,
1789 struct be_rx_compl_info *rxcp)
1791 struct be_rx_page_info *page_info;
1793 u16 hdr_len, curr_frag_len, remaining;
1796 page_info = get_rx_page_info(rxo);
1797 start = page_address(page_info->page) + page_info->page_offset;
1800 /* Copy data in the first descriptor of this completion */
1801 curr_frag_len = min(rxcp->pkt_size, rx_frag_size);
1803 skb->len = curr_frag_len;
1804 if (curr_frag_len <= BE_HDR_LEN) { /* tiny packet */
1805 memcpy(skb->data, start, curr_frag_len);
1806 /* Complete packet has now been moved to data */
1807 put_page(page_info->page);
1809 skb->tail += curr_frag_len;
1812 memcpy(skb->data, start, hdr_len);
1813 skb_shinfo(skb)->nr_frags = 1;
1814 skb_frag_set_page(skb, 0, page_info->page);
1815 skb_shinfo(skb)->frags[0].page_offset =
1816 page_info->page_offset + hdr_len;
1817 skb_frag_size_set(&skb_shinfo(skb)->frags[0],
1818 curr_frag_len - hdr_len);
1819 skb->data_len = curr_frag_len - hdr_len;
1820 skb->truesize += rx_frag_size;
1821 skb->tail += hdr_len;
1823 page_info->page = NULL;
1825 if (rxcp->pkt_size <= rx_frag_size) {
1826 BUG_ON(rxcp->num_rcvd != 1);
1830 /* More frags present for this completion */
1831 remaining = rxcp->pkt_size - curr_frag_len;
1832 for (i = 1, j = 0; i < rxcp->num_rcvd; i++) {
1833 page_info = get_rx_page_info(rxo);
1834 curr_frag_len = min(remaining, rx_frag_size);
1836 /* Coalesce all frags from the same physical page in one slot */
1837 if (page_info->page_offset == 0) {
1840 skb_frag_set_page(skb, j, page_info->page);
1841 skb_shinfo(skb)->frags[j].page_offset =
1842 page_info->page_offset;
1843 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1844 skb_shinfo(skb)->nr_frags++;
1846 put_page(page_info->page);
1849 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1850 skb->len += curr_frag_len;
1851 skb->data_len += curr_frag_len;
1852 skb->truesize += rx_frag_size;
1853 remaining -= curr_frag_len;
1854 page_info->page = NULL;
1856 BUG_ON(j > MAX_SKB_FRAGS);
1859 /* Process the RX completion indicated by rxcp when GRO is disabled */
1860 static void be_rx_compl_process(struct be_rx_obj *rxo, struct napi_struct *napi,
1861 struct be_rx_compl_info *rxcp)
1863 struct be_adapter *adapter = rxo->adapter;
1864 struct net_device *netdev = adapter->netdev;
1865 struct sk_buff *skb;
1867 skb = netdev_alloc_skb_ip_align(netdev, BE_RX_SKB_ALLOC_SIZE);
1868 if (unlikely(!skb)) {
1869 rx_stats(rxo)->rx_drops_no_skbs++;
1870 be_rx_compl_discard(rxo, rxcp);
1874 skb_fill_rx_data(rxo, skb, rxcp);
1876 if (likely((netdev->features & NETIF_F_RXCSUM) && csum_passed(rxcp)))
1877 skb->ip_summed = CHECKSUM_UNNECESSARY;
1879 skb_checksum_none_assert(skb);
1881 skb->protocol = eth_type_trans(skb, netdev);
1882 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1883 if (netdev->features & NETIF_F_RXHASH)
1884 skb_set_hash(skb, rxcp->rss_hash, PKT_HASH_TYPE_L3);
1886 skb->csum_level = rxcp->tunneled;
1887 skb_mark_napi_id(skb, napi);
1890 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1892 netif_receive_skb(skb);
1895 /* Process the RX completion indicated by rxcp when GRO is enabled */
1896 static void be_rx_compl_process_gro(struct be_rx_obj *rxo,
1897 struct napi_struct *napi,
1898 struct be_rx_compl_info *rxcp)
1900 struct be_adapter *adapter = rxo->adapter;
1901 struct be_rx_page_info *page_info;
1902 struct sk_buff *skb = NULL;
1903 u16 remaining, curr_frag_len;
1906 skb = napi_get_frags(napi);
1908 be_rx_compl_discard(rxo, rxcp);
1912 remaining = rxcp->pkt_size;
1913 for (i = 0, j = -1; i < rxcp->num_rcvd; i++) {
1914 page_info = get_rx_page_info(rxo);
1916 curr_frag_len = min(remaining, rx_frag_size);
1918 /* Coalesce all frags from the same physical page in one slot */
1919 if (i == 0 || page_info->page_offset == 0) {
1920 /* First frag or Fresh page */
1922 skb_frag_set_page(skb, j, page_info->page);
1923 skb_shinfo(skb)->frags[j].page_offset =
1924 page_info->page_offset;
1925 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1927 put_page(page_info->page);
1929 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1930 skb->truesize += rx_frag_size;
1931 remaining -= curr_frag_len;
1932 memset(page_info, 0, sizeof(*page_info));
1934 BUG_ON(j > MAX_SKB_FRAGS);
1936 skb_shinfo(skb)->nr_frags = j + 1;
1937 skb->len = rxcp->pkt_size;
1938 skb->data_len = rxcp->pkt_size;
1939 skb->ip_summed = CHECKSUM_UNNECESSARY;
1940 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1941 if (adapter->netdev->features & NETIF_F_RXHASH)
1942 skb_set_hash(skb, rxcp->rss_hash, PKT_HASH_TYPE_L3);
1944 skb->csum_level = rxcp->tunneled;
1945 skb_mark_napi_id(skb, napi);
1948 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1950 napi_gro_frags(napi);
1953 static void be_parse_rx_compl_v1(struct be_eth_rx_compl *compl,
1954 struct be_rx_compl_info *rxcp)
1956 rxcp->pkt_size = GET_RX_COMPL_V1_BITS(pktsize, compl);
1957 rxcp->vlanf = GET_RX_COMPL_V1_BITS(vtp, compl);
1958 rxcp->err = GET_RX_COMPL_V1_BITS(err, compl);
1959 rxcp->tcpf = GET_RX_COMPL_V1_BITS(tcpf, compl);
1960 rxcp->udpf = GET_RX_COMPL_V1_BITS(udpf, compl);
1961 rxcp->ip_csum = GET_RX_COMPL_V1_BITS(ipcksm, compl);
1962 rxcp->l4_csum = GET_RX_COMPL_V1_BITS(l4_cksm, compl);
1963 rxcp->ipv6 = GET_RX_COMPL_V1_BITS(ip_version, compl);
1964 rxcp->num_rcvd = GET_RX_COMPL_V1_BITS(numfrags, compl);
1965 rxcp->pkt_type = GET_RX_COMPL_V1_BITS(cast_enc, compl);
1966 rxcp->rss_hash = GET_RX_COMPL_V1_BITS(rsshash, compl);
1968 rxcp->qnq = GET_RX_COMPL_V1_BITS(qnq, compl);
1969 rxcp->vlan_tag = GET_RX_COMPL_V1_BITS(vlan_tag, compl);
1971 rxcp->port = GET_RX_COMPL_V1_BITS(port, compl);
1973 GET_RX_COMPL_V1_BITS(tunneled, compl);
1976 static void be_parse_rx_compl_v0(struct be_eth_rx_compl *compl,
1977 struct be_rx_compl_info *rxcp)
1979 rxcp->pkt_size = GET_RX_COMPL_V0_BITS(pktsize, compl);
1980 rxcp->vlanf = GET_RX_COMPL_V0_BITS(vtp, compl);
1981 rxcp->err = GET_RX_COMPL_V0_BITS(err, compl);
1982 rxcp->tcpf = GET_RX_COMPL_V0_BITS(tcpf, compl);
1983 rxcp->udpf = GET_RX_COMPL_V0_BITS(udpf, compl);
1984 rxcp->ip_csum = GET_RX_COMPL_V0_BITS(ipcksm, compl);
1985 rxcp->l4_csum = GET_RX_COMPL_V0_BITS(l4_cksm, compl);
1986 rxcp->ipv6 = GET_RX_COMPL_V0_BITS(ip_version, compl);
1987 rxcp->num_rcvd = GET_RX_COMPL_V0_BITS(numfrags, compl);
1988 rxcp->pkt_type = GET_RX_COMPL_V0_BITS(cast_enc, compl);
1989 rxcp->rss_hash = GET_RX_COMPL_V0_BITS(rsshash, compl);
1991 rxcp->qnq = GET_RX_COMPL_V0_BITS(qnq, compl);
1992 rxcp->vlan_tag = GET_RX_COMPL_V0_BITS(vlan_tag, compl);
1994 rxcp->port = GET_RX_COMPL_V0_BITS(port, compl);
1995 rxcp->ip_frag = GET_RX_COMPL_V0_BITS(ip_frag, compl);
1998 static struct be_rx_compl_info *be_rx_compl_get(struct be_rx_obj *rxo)
2000 struct be_eth_rx_compl *compl = queue_tail_node(&rxo->cq);
2001 struct be_rx_compl_info *rxcp = &rxo->rxcp;
2002 struct be_adapter *adapter = rxo->adapter;
2004 /* For checking the valid bit it is Ok to use either definition as the
2005 * valid bit is at the same position in both v0 and v1 Rx compl */
2006 if (compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] == 0)
2010 be_dws_le_to_cpu(compl, sizeof(*compl));
2012 if (adapter->be3_native)
2013 be_parse_rx_compl_v1(compl, rxcp);
2015 be_parse_rx_compl_v0(compl, rxcp);
2021 /* In QNQ modes, if qnq bit is not set, then the packet was
2022 * tagged only with the transparent outer vlan-tag and must
2023 * not be treated as a vlan packet by host
2025 if (be_is_qnq_mode(adapter) && !rxcp->qnq)
2028 if (!lancer_chip(adapter))
2029 rxcp->vlan_tag = swab16(rxcp->vlan_tag);
2031 if (adapter->pvid == (rxcp->vlan_tag & VLAN_VID_MASK) &&
2032 !test_bit(rxcp->vlan_tag, adapter->vids))
2036 /* As the compl has been parsed, reset it; we wont touch it again */
2037 compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] = 0;
2039 queue_tail_inc(&rxo->cq);
2043 static inline struct page *be_alloc_pages(u32 size, gfp_t gfp)
2045 u32 order = get_order(size);
2049 return alloc_pages(gfp, order);
2053 * Allocate a page, split it to fragments of size rx_frag_size and post as
2054 * receive buffers to BE
2056 static void be_post_rx_frags(struct be_rx_obj *rxo, gfp_t gfp, u32 frags_needed)
2058 struct be_adapter *adapter = rxo->adapter;
2059 struct be_rx_page_info *page_info = NULL, *prev_page_info = NULL;
2060 struct be_queue_info *rxq = &rxo->q;
2061 struct page *pagep = NULL;
2062 struct device *dev = &adapter->pdev->dev;
2063 struct be_eth_rx_d *rxd;
2064 u64 page_dmaaddr = 0, frag_dmaaddr;
2065 u32 posted, page_offset = 0, notify = 0;
2067 page_info = &rxo->page_info_tbl[rxq->head];
2068 for (posted = 0; posted < frags_needed && !page_info->page; posted++) {
2070 pagep = be_alloc_pages(adapter->big_page_size, gfp);
2071 if (unlikely(!pagep)) {
2072 rx_stats(rxo)->rx_post_fail++;
2075 page_dmaaddr = dma_map_page(dev, pagep, 0,
2076 adapter->big_page_size,
2078 if (dma_mapping_error(dev, page_dmaaddr)) {
2081 adapter->drv_stats.dma_map_errors++;
2087 page_offset += rx_frag_size;
2089 page_info->page_offset = page_offset;
2090 page_info->page = pagep;
2092 rxd = queue_head_node(rxq);
2093 frag_dmaaddr = page_dmaaddr + page_info->page_offset;
2094 rxd->fragpa_lo = cpu_to_le32(frag_dmaaddr & 0xFFFFFFFF);
2095 rxd->fragpa_hi = cpu_to_le32(upper_32_bits(frag_dmaaddr));
2097 /* Any space left in the current big page for another frag? */
2098 if ((page_offset + rx_frag_size + rx_frag_size) >
2099 adapter->big_page_size) {
2101 page_info->last_frag = true;
2102 dma_unmap_addr_set(page_info, bus, page_dmaaddr);
2104 dma_unmap_addr_set(page_info, bus, frag_dmaaddr);
2107 prev_page_info = page_info;
2108 queue_head_inc(rxq);
2109 page_info = &rxo->page_info_tbl[rxq->head];
2112 /* Mark the last frag of a page when we break out of the above loop
2113 * with no more slots available in the RXQ
2116 prev_page_info->last_frag = true;
2117 dma_unmap_addr_set(prev_page_info, bus, page_dmaaddr);
2121 atomic_add(posted, &rxq->used);
2122 if (rxo->rx_post_starved)
2123 rxo->rx_post_starved = false;
2125 notify = min(256u, posted);
2126 be_rxq_notify(adapter, rxq->id, notify);
2129 } else if (atomic_read(&rxq->used) == 0) {
2130 /* Let be_worker replenish when memory is available */
2131 rxo->rx_post_starved = true;
2135 static struct be_tx_compl_info *be_tx_compl_get(struct be_tx_obj *txo)
2137 struct be_queue_info *tx_cq = &txo->cq;
2138 struct be_tx_compl_info *txcp = &txo->txcp;
2139 struct be_eth_tx_compl *compl = queue_tail_node(tx_cq);
2141 if (compl->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] == 0)
2144 /* Ensure load ordering of valid bit dword and other dwords below */
2146 be_dws_le_to_cpu(compl, sizeof(*compl));
2148 txcp->status = GET_TX_COMPL_BITS(status, compl);
2149 txcp->end_index = GET_TX_COMPL_BITS(wrb_index, compl);
2151 compl->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] = 0;
2152 queue_tail_inc(tx_cq);
2156 static u16 be_tx_compl_process(struct be_adapter *adapter,
2157 struct be_tx_obj *txo, u16 last_index)
2159 struct sk_buff **sent_skbs = txo->sent_skb_list;
2160 struct be_queue_info *txq = &txo->q;
2161 u16 frag_index, num_wrbs = 0;
2162 struct sk_buff *skb = NULL;
2163 bool unmap_skb_hdr = false;
2164 struct be_eth_wrb *wrb;
2167 if (sent_skbs[txq->tail]) {
2168 /* Free skb from prev req */
2170 dev_consume_skb_any(skb);
2171 skb = sent_skbs[txq->tail];
2172 sent_skbs[txq->tail] = NULL;
2173 queue_tail_inc(txq); /* skip hdr wrb */
2175 unmap_skb_hdr = true;
2177 wrb = queue_tail_node(txq);
2178 frag_index = txq->tail;
2179 unmap_tx_frag(&adapter->pdev->dev, wrb,
2180 (unmap_skb_hdr && skb_headlen(skb)));
2181 unmap_skb_hdr = false;
2182 queue_tail_inc(txq);
2184 } while (frag_index != last_index);
2185 dev_consume_skb_any(skb);
2190 /* Return the number of events in the event queue */
2191 static inline int events_get(struct be_eq_obj *eqo)
2193 struct be_eq_entry *eqe;
2197 eqe = queue_tail_node(&eqo->q);
2204 queue_tail_inc(&eqo->q);
2210 /* Leaves the EQ is disarmed state */
2211 static void be_eq_clean(struct be_eq_obj *eqo)
2213 int num = events_get(eqo);
2215 be_eq_notify(eqo->adapter, eqo->q.id, false, true, num);
2218 static void be_rx_cq_clean(struct be_rx_obj *rxo)
2220 struct be_rx_page_info *page_info;
2221 struct be_queue_info *rxq = &rxo->q;
2222 struct be_queue_info *rx_cq = &rxo->cq;
2223 struct be_rx_compl_info *rxcp;
2224 struct be_adapter *adapter = rxo->adapter;
2227 /* Consume pending rx completions.
2228 * Wait for the flush completion (identified by zero num_rcvd)
2229 * to arrive. Notify CQ even when there are no more CQ entries
2230 * for HW to flush partially coalesced CQ entries.
2231 * In Lancer, there is no need to wait for flush compl.
2234 rxcp = be_rx_compl_get(rxo);
2236 if (lancer_chip(adapter))
2239 if (flush_wait++ > 10 || be_hw_error(adapter)) {
2240 dev_warn(&adapter->pdev->dev,
2241 "did not receive flush compl\n");
2244 be_cq_notify(adapter, rx_cq->id, true, 0);
2247 be_rx_compl_discard(rxo, rxcp);
2248 be_cq_notify(adapter, rx_cq->id, false, 1);
2249 if (rxcp->num_rcvd == 0)
2254 /* After cleanup, leave the CQ in unarmed state */
2255 be_cq_notify(adapter, rx_cq->id, false, 0);
2257 /* Then free posted rx buffers that were not used */
2258 while (atomic_read(&rxq->used) > 0) {
2259 page_info = get_rx_page_info(rxo);
2260 put_page(page_info->page);
2261 memset(page_info, 0, sizeof(*page_info));
2263 BUG_ON(atomic_read(&rxq->used));
2268 static void be_tx_compl_clean(struct be_adapter *adapter)
2270 u16 end_idx, notified_idx, cmpl = 0, timeo = 0, num_wrbs = 0;
2271 struct device *dev = &adapter->pdev->dev;
2272 struct be_tx_compl_info *txcp;
2273 struct be_queue_info *txq;
2274 struct be_tx_obj *txo;
2275 int i, pending_txqs;
2277 /* Stop polling for compls when HW has been silent for 10ms */
2279 pending_txqs = adapter->num_tx_qs;
2281 for_all_tx_queues(adapter, txo, i) {
2285 while ((txcp = be_tx_compl_get(txo))) {
2287 be_tx_compl_process(adapter, txo,
2292 be_cq_notify(adapter, txo->cq.id, false, cmpl);
2293 atomic_sub(num_wrbs, &txq->used);
2296 if (!be_is_tx_compl_pending(txo))
2300 if (pending_txqs == 0 || ++timeo > 10 || be_hw_error(adapter))
2306 /* Free enqueued TX that was never notified to HW */
2307 for_all_tx_queues(adapter, txo, i) {
2310 if (atomic_read(&txq->used)) {
2311 dev_info(dev, "txq%d: cleaning %d pending tx-wrbs\n",
2312 i, atomic_read(&txq->used));
2313 notified_idx = txq->tail;
2314 end_idx = txq->tail;
2315 index_adv(&end_idx, atomic_read(&txq->used) - 1,
2317 /* Use the tx-compl process logic to handle requests
2318 * that were not sent to the HW.
2320 num_wrbs = be_tx_compl_process(adapter, txo, end_idx);
2321 atomic_sub(num_wrbs, &txq->used);
2322 BUG_ON(atomic_read(&txq->used));
2323 txo->pend_wrb_cnt = 0;
2324 /* Since hw was never notified of these requests,
2327 txq->head = notified_idx;
2328 txq->tail = notified_idx;
2333 static void be_evt_queues_destroy(struct be_adapter *adapter)
2335 struct be_eq_obj *eqo;
2338 for_all_evt_queues(adapter, eqo, i) {
2339 if (eqo->q.created) {
2341 be_cmd_q_destroy(adapter, &eqo->q, QTYPE_EQ);
2342 napi_hash_del(&eqo->napi);
2343 netif_napi_del(&eqo->napi);
2345 be_queue_free(adapter, &eqo->q);
2349 static int be_evt_queues_create(struct be_adapter *adapter)
2351 struct be_queue_info *eq;
2352 struct be_eq_obj *eqo;
2353 struct be_aic_obj *aic;
2356 adapter->num_evt_qs = min_t(u16, num_irqs(adapter),
2357 adapter->cfg_num_qs);
2359 for_all_evt_queues(adapter, eqo, i) {
2360 netif_napi_add(adapter->netdev, &eqo->napi, be_poll,
2362 napi_hash_add(&eqo->napi);
2363 aic = &adapter->aic_obj[i];
2364 eqo->adapter = adapter;
2366 aic->max_eqd = BE_MAX_EQD;
2370 rc = be_queue_alloc(adapter, eq, EVNT_Q_LEN,
2371 sizeof(struct be_eq_entry));
2375 rc = be_cmd_eq_create(adapter, eqo);
2382 static void be_mcc_queues_destroy(struct be_adapter *adapter)
2384 struct be_queue_info *q;
2386 q = &adapter->mcc_obj.q;
2388 be_cmd_q_destroy(adapter, q, QTYPE_MCCQ);
2389 be_queue_free(adapter, q);
2391 q = &adapter->mcc_obj.cq;
2393 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2394 be_queue_free(adapter, q);
2397 /* Must be called only after TX qs are created as MCC shares TX EQ */
2398 static int be_mcc_queues_create(struct be_adapter *adapter)
2400 struct be_queue_info *q, *cq;
2402 cq = &adapter->mcc_obj.cq;
2403 if (be_queue_alloc(adapter, cq, MCC_CQ_LEN,
2404 sizeof(struct be_mcc_compl)))
2407 /* Use the default EQ for MCC completions */
2408 if (be_cmd_cq_create(adapter, cq, &mcc_eqo(adapter)->q, true, 0))
2411 q = &adapter->mcc_obj.q;
2412 if (be_queue_alloc(adapter, q, MCC_Q_LEN, sizeof(struct be_mcc_wrb)))
2413 goto mcc_cq_destroy;
2415 if (be_cmd_mccq_create(adapter, q, cq))
2421 be_queue_free(adapter, q);
2423 be_cmd_q_destroy(adapter, cq, QTYPE_CQ);
2425 be_queue_free(adapter, cq);
2430 static void be_tx_queues_destroy(struct be_adapter *adapter)
2432 struct be_queue_info *q;
2433 struct be_tx_obj *txo;
2436 for_all_tx_queues(adapter, txo, i) {
2439 be_cmd_q_destroy(adapter, q, QTYPE_TXQ);
2440 be_queue_free(adapter, q);
2444 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2445 be_queue_free(adapter, q);
2449 static int be_tx_qs_create(struct be_adapter *adapter)
2451 struct be_queue_info *cq, *eq;
2452 struct be_tx_obj *txo;
2455 adapter->num_tx_qs = min(adapter->num_evt_qs, be_max_txqs(adapter));
2457 for_all_tx_queues(adapter, txo, i) {
2459 status = be_queue_alloc(adapter, cq, TX_CQ_LEN,
2460 sizeof(struct be_eth_tx_compl));
2464 u64_stats_init(&txo->stats.sync);
2465 u64_stats_init(&txo->stats.sync_compl);
2467 /* If num_evt_qs is less than num_tx_qs, then more than
2468 * one txq share an eq
2470 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2471 status = be_cmd_cq_create(adapter, cq, eq, false, 3);
2475 status = be_queue_alloc(adapter, &txo->q, TX_Q_LEN,
2476 sizeof(struct be_eth_wrb));
2480 status = be_cmd_txq_create(adapter, txo);
2485 dev_info(&adapter->pdev->dev, "created %d TX queue(s)\n",
2486 adapter->num_tx_qs);
2490 static void be_rx_cqs_destroy(struct be_adapter *adapter)
2492 struct be_queue_info *q;
2493 struct be_rx_obj *rxo;
2496 for_all_rx_queues(adapter, rxo, i) {
2499 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2500 be_queue_free(adapter, q);
2504 static int be_rx_cqs_create(struct be_adapter *adapter)
2506 struct be_queue_info *eq, *cq;
2507 struct be_rx_obj *rxo;
2510 /* We can create as many RSS rings as there are EQs. */
2511 adapter->num_rss_qs = adapter->num_evt_qs;
2513 /* We'll use RSS only if atleast 2 RSS rings are supported. */
2514 if (adapter->num_rss_qs <= 1)
2515 adapter->num_rss_qs = 0;
2517 adapter->num_rx_qs = adapter->num_rss_qs + adapter->need_def_rxq;
2519 /* When the interface is not capable of RSS rings (and there is no
2520 * need to create a default RXQ) we'll still need one RXQ
2522 if (adapter->num_rx_qs == 0)
2523 adapter->num_rx_qs = 1;
2525 adapter->big_page_size = (1 << get_order(rx_frag_size)) * PAGE_SIZE;
2526 for_all_rx_queues(adapter, rxo, i) {
2527 rxo->adapter = adapter;
2529 rc = be_queue_alloc(adapter, cq, RX_CQ_LEN,
2530 sizeof(struct be_eth_rx_compl));
2534 u64_stats_init(&rxo->stats.sync);
2535 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2536 rc = be_cmd_cq_create(adapter, cq, eq, false, 3);
2541 dev_info(&adapter->pdev->dev,
2542 "created %d RX queue(s)\n", adapter->num_rx_qs);
2546 static irqreturn_t be_intx(int irq, void *dev)
2548 struct be_eq_obj *eqo = dev;
2549 struct be_adapter *adapter = eqo->adapter;
2552 /* IRQ is not expected when NAPI is scheduled as the EQ
2553 * will not be armed.
2554 * But, this can happen on Lancer INTx where it takes
2555 * a while to de-assert INTx or in BE2 where occasionaly
2556 * an interrupt may be raised even when EQ is unarmed.
2557 * If NAPI is already scheduled, then counting & notifying
2558 * events will orphan them.
2560 if (napi_schedule_prep(&eqo->napi)) {
2561 num_evts = events_get(eqo);
2562 __napi_schedule(&eqo->napi);
2564 eqo->spurious_intr = 0;
2566 be_eq_notify(adapter, eqo->q.id, false, true, num_evts);
2568 /* Return IRQ_HANDLED only for the the first spurious intr
2569 * after a valid intr to stop the kernel from branding
2570 * this irq as a bad one!
2572 if (num_evts || eqo->spurious_intr++ == 0)
2578 static irqreturn_t be_msix(int irq, void *dev)
2580 struct be_eq_obj *eqo = dev;
2582 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
2583 napi_schedule(&eqo->napi);
2587 static inline bool do_gro(struct be_rx_compl_info *rxcp)
2589 return (rxcp->tcpf && !rxcp->err && rxcp->l4_csum) ? true : false;
2592 static int be_process_rx(struct be_rx_obj *rxo, struct napi_struct *napi,
2593 int budget, int polling)
2595 struct be_adapter *adapter = rxo->adapter;
2596 struct be_queue_info *rx_cq = &rxo->cq;
2597 struct be_rx_compl_info *rxcp;
2599 u32 frags_consumed = 0;
2601 for (work_done = 0; work_done < budget; work_done++) {
2602 rxcp = be_rx_compl_get(rxo);
2606 /* Is it a flush compl that has no data */
2607 if (unlikely(rxcp->num_rcvd == 0))
2610 /* Discard compl with partial DMA Lancer B0 */
2611 if (unlikely(!rxcp->pkt_size)) {
2612 be_rx_compl_discard(rxo, rxcp);
2616 /* On BE drop pkts that arrive due to imperfect filtering in
2617 * promiscuous mode on some skews
2619 if (unlikely(rxcp->port != adapter->port_num &&
2620 !lancer_chip(adapter))) {
2621 be_rx_compl_discard(rxo, rxcp);
2625 /* Don't do gro when we're busy_polling */
2626 if (do_gro(rxcp) && polling != BUSY_POLLING)
2627 be_rx_compl_process_gro(rxo, napi, rxcp);
2629 be_rx_compl_process(rxo, napi, rxcp);
2632 frags_consumed += rxcp->num_rcvd;
2633 be_rx_stats_update(rxo, rxcp);
2637 be_cq_notify(adapter, rx_cq->id, true, work_done);
2639 /* When an rx-obj gets into post_starved state, just
2640 * let be_worker do the posting.
2642 if (atomic_read(&rxo->q.used) < RX_FRAGS_REFILL_WM &&
2643 !rxo->rx_post_starved)
2644 be_post_rx_frags(rxo, GFP_ATOMIC,
2645 max_t(u32, MAX_RX_POST,
2652 static inline void be_update_tx_err(struct be_tx_obj *txo, u8 status)
2655 case BE_TX_COMP_HDR_PARSE_ERR:
2656 tx_stats(txo)->tx_hdr_parse_err++;
2658 case BE_TX_COMP_NDMA_ERR:
2659 tx_stats(txo)->tx_dma_err++;
2661 case BE_TX_COMP_ACL_ERR:
2662 tx_stats(txo)->tx_spoof_check_err++;
2667 static inline void lancer_update_tx_err(struct be_tx_obj *txo, u8 status)
2670 case LANCER_TX_COMP_LSO_ERR:
2671 tx_stats(txo)->tx_tso_err++;
2673 case LANCER_TX_COMP_HSW_DROP_MAC_ERR:
2674 case LANCER_TX_COMP_HSW_DROP_VLAN_ERR:
2675 tx_stats(txo)->tx_spoof_check_err++;
2677 case LANCER_TX_COMP_QINQ_ERR:
2678 tx_stats(txo)->tx_qinq_err++;
2680 case LANCER_TX_COMP_PARITY_ERR:
2681 tx_stats(txo)->tx_internal_parity_err++;
2683 case LANCER_TX_COMP_DMA_ERR:
2684 tx_stats(txo)->tx_dma_err++;
2689 static void be_process_tx(struct be_adapter *adapter, struct be_tx_obj *txo,
2692 int num_wrbs = 0, work_done = 0;
2693 struct be_tx_compl_info *txcp;
2695 while ((txcp = be_tx_compl_get(txo))) {
2696 num_wrbs += be_tx_compl_process(adapter, txo, txcp->end_index);
2700 if (lancer_chip(adapter))
2701 lancer_update_tx_err(txo, txcp->status);
2703 be_update_tx_err(txo, txcp->status);
2708 be_cq_notify(adapter, txo->cq.id, true, work_done);
2709 atomic_sub(num_wrbs, &txo->q.used);
2711 /* As Tx wrbs have been freed up, wake up netdev queue
2712 * if it was stopped due to lack of tx wrbs. */
2713 if (__netif_subqueue_stopped(adapter->netdev, idx) &&
2714 be_can_txq_wake(txo)) {
2715 netif_wake_subqueue(adapter->netdev, idx);
2718 u64_stats_update_begin(&tx_stats(txo)->sync_compl);
2719 tx_stats(txo)->tx_compl += work_done;
2720 u64_stats_update_end(&tx_stats(txo)->sync_compl);
2724 #ifdef CONFIG_NET_RX_BUSY_POLL
2725 static inline bool be_lock_napi(struct be_eq_obj *eqo)
2729 spin_lock(&eqo->lock); /* BH is already disabled */
2730 if (eqo->state & BE_EQ_LOCKED) {
2731 WARN_ON(eqo->state & BE_EQ_NAPI);
2732 eqo->state |= BE_EQ_NAPI_YIELD;
2735 eqo->state = BE_EQ_NAPI;
2737 spin_unlock(&eqo->lock);
2741 static inline void be_unlock_napi(struct be_eq_obj *eqo)
2743 spin_lock(&eqo->lock); /* BH is already disabled */
2745 WARN_ON(eqo->state & (BE_EQ_POLL | BE_EQ_NAPI_YIELD));
2746 eqo->state = BE_EQ_IDLE;
2748 spin_unlock(&eqo->lock);
2751 static inline bool be_lock_busy_poll(struct be_eq_obj *eqo)
2755 spin_lock_bh(&eqo->lock);
2756 if (eqo->state & BE_EQ_LOCKED) {
2757 eqo->state |= BE_EQ_POLL_YIELD;
2760 eqo->state |= BE_EQ_POLL;
2762 spin_unlock_bh(&eqo->lock);
2766 static inline void be_unlock_busy_poll(struct be_eq_obj *eqo)
2768 spin_lock_bh(&eqo->lock);
2770 WARN_ON(eqo->state & (BE_EQ_NAPI));
2771 eqo->state = BE_EQ_IDLE;
2773 spin_unlock_bh(&eqo->lock);
2776 static inline void be_enable_busy_poll(struct be_eq_obj *eqo)
2778 spin_lock_init(&eqo->lock);
2779 eqo->state = BE_EQ_IDLE;
2782 static inline void be_disable_busy_poll(struct be_eq_obj *eqo)
2786 /* It's enough to just acquire napi lock on the eqo to stop
2787 * be_busy_poll() from processing any queueus.
2789 while (!be_lock_napi(eqo))
2795 #else /* CONFIG_NET_RX_BUSY_POLL */
2797 static inline bool be_lock_napi(struct be_eq_obj *eqo)
2802 static inline void be_unlock_napi(struct be_eq_obj *eqo)
2806 static inline bool be_lock_busy_poll(struct be_eq_obj *eqo)
2811 static inline void be_unlock_busy_poll(struct be_eq_obj *eqo)
2815 static inline void be_enable_busy_poll(struct be_eq_obj *eqo)
2819 static inline void be_disable_busy_poll(struct be_eq_obj *eqo)
2822 #endif /* CONFIG_NET_RX_BUSY_POLL */
2824 int be_poll(struct napi_struct *napi, int budget)
2826 struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2827 struct be_adapter *adapter = eqo->adapter;
2828 int max_work = 0, work, i, num_evts;
2829 struct be_rx_obj *rxo;
2830 struct be_tx_obj *txo;
2832 num_evts = events_get(eqo);
2834 for_all_tx_queues_on_eq(adapter, eqo, txo, i)
2835 be_process_tx(adapter, txo, i);
2837 if (be_lock_napi(eqo)) {
2838 /* This loop will iterate twice for EQ0 in which
2839 * completions of the last RXQ (default one) are also processed
2840 * For other EQs the loop iterates only once
2842 for_all_rx_queues_on_eq(adapter, eqo, rxo, i) {
2843 work = be_process_rx(rxo, napi, budget, NAPI_POLLING);
2844 max_work = max(work, max_work);
2846 be_unlock_napi(eqo);
2851 if (is_mcc_eqo(eqo))
2852 be_process_mcc(adapter);
2854 if (max_work < budget) {
2855 napi_complete(napi);
2856 be_eq_notify(adapter, eqo->q.id, true, false, num_evts);
2858 /* As we'll continue in polling mode, count and clear events */
2859 be_eq_notify(adapter, eqo->q.id, false, false, num_evts);
2864 #ifdef CONFIG_NET_RX_BUSY_POLL
2865 static int be_busy_poll(struct napi_struct *napi)
2867 struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2868 struct be_adapter *adapter = eqo->adapter;
2869 struct be_rx_obj *rxo;
2872 if (!be_lock_busy_poll(eqo))
2873 return LL_FLUSH_BUSY;
2875 for_all_rx_queues_on_eq(adapter, eqo, rxo, i) {
2876 work = be_process_rx(rxo, napi, 4, BUSY_POLLING);
2881 be_unlock_busy_poll(eqo);
2886 void be_detect_error(struct be_adapter *adapter)
2888 u32 ue_lo = 0, ue_hi = 0, ue_lo_mask = 0, ue_hi_mask = 0;
2889 u32 sliport_status = 0, sliport_err1 = 0, sliport_err2 = 0;
2891 bool error_detected = false;
2892 struct device *dev = &adapter->pdev->dev;
2893 struct net_device *netdev = adapter->netdev;
2895 if (be_hw_error(adapter))
2898 if (lancer_chip(adapter)) {
2899 sliport_status = ioread32(adapter->db + SLIPORT_STATUS_OFFSET);
2900 if (sliport_status & SLIPORT_STATUS_ERR_MASK) {
2901 sliport_err1 = ioread32(adapter->db +
2902 SLIPORT_ERROR1_OFFSET);
2903 sliport_err2 = ioread32(adapter->db +
2904 SLIPORT_ERROR2_OFFSET);
2905 adapter->hw_error = true;
2906 error_detected = true;
2907 /* Do not log error messages if its a FW reset */
2908 if (sliport_err1 == SLIPORT_ERROR_FW_RESET1 &&
2909 sliport_err2 == SLIPORT_ERROR_FW_RESET2) {
2910 dev_info(dev, "Firmware update in progress\n");
2912 dev_err(dev, "Error detected in the card\n");
2913 dev_err(dev, "ERR: sliport status 0x%x\n",
2915 dev_err(dev, "ERR: sliport error1 0x%x\n",
2917 dev_err(dev, "ERR: sliport error2 0x%x\n",
2922 ue_lo = ioread32(adapter->pcicfg + PCICFG_UE_STATUS_LOW);
2923 ue_hi = ioread32(adapter->pcicfg + PCICFG_UE_STATUS_HIGH);
2924 ue_lo_mask = ioread32(adapter->pcicfg +
2925 PCICFG_UE_STATUS_LOW_MASK);
2926 ue_hi_mask = ioread32(adapter->pcicfg +
2927 PCICFG_UE_STATUS_HI_MASK);
2929 ue_lo = (ue_lo & ~ue_lo_mask);
2930 ue_hi = (ue_hi & ~ue_hi_mask);
2932 /* On certain platforms BE hardware can indicate spurious UEs.
2933 * Allow HW to stop working completely in case of a real UE.
2934 * Hence not setting the hw_error for UE detection.
2937 if (ue_lo || ue_hi) {
2938 error_detected = true;
2940 "Unrecoverable Error detected in the adapter");
2941 dev_err(dev, "Please reboot server to recover");
2942 if (skyhawk_chip(adapter))
2943 adapter->hw_error = true;
2944 for (i = 0; ue_lo; ue_lo >>= 1, i++) {
2946 dev_err(dev, "UE: %s bit set\n",
2947 ue_status_low_desc[i]);
2949 for (i = 0; ue_hi; ue_hi >>= 1, i++) {
2951 dev_err(dev, "UE: %s bit set\n",
2952 ue_status_hi_desc[i]);
2957 netif_carrier_off(netdev);
2960 static void be_msix_disable(struct be_adapter *adapter)
2962 if (msix_enabled(adapter)) {
2963 pci_disable_msix(adapter->pdev);
2964 adapter->num_msix_vec = 0;
2965 adapter->num_msix_roce_vec = 0;
2969 static int be_msix_enable(struct be_adapter *adapter)
2972 struct device *dev = &adapter->pdev->dev;
2974 /* If RoCE is supported, program the max number of NIC vectors that
2975 * may be configured via set-channels, along with vectors needed for
2976 * RoCe. Else, just program the number we'll use initially.
2978 if (be_roce_supported(adapter))
2979 num_vec = min_t(int, 2 * be_max_eqs(adapter),
2980 2 * num_online_cpus());
2982 num_vec = adapter->cfg_num_qs;
2984 for (i = 0; i < num_vec; i++)
2985 adapter->msix_entries[i].entry = i;
2987 num_vec = pci_enable_msix_range(adapter->pdev, adapter->msix_entries,
2988 MIN_MSIX_VECTORS, num_vec);
2992 if (be_roce_supported(adapter) && num_vec > MIN_MSIX_VECTORS) {
2993 adapter->num_msix_roce_vec = num_vec / 2;
2994 dev_info(dev, "enabled %d MSI-x vector(s) for RoCE\n",
2995 adapter->num_msix_roce_vec);
2998 adapter->num_msix_vec = num_vec - adapter->num_msix_roce_vec;
3000 dev_info(dev, "enabled %d MSI-x vector(s) for NIC\n",
3001 adapter->num_msix_vec);
3005 dev_warn(dev, "MSIx enable failed\n");
3007 /* INTx is not supported in VFs, so fail probe if enable_msix fails */
3008 if (!be_physfn(adapter))
3013 static inline int be_msix_vec_get(struct be_adapter *adapter,
3014 struct be_eq_obj *eqo)
3016 return adapter->msix_entries[eqo->msix_idx].vector;
3019 static int be_msix_register(struct be_adapter *adapter)
3021 struct net_device *netdev = adapter->netdev;
3022 struct be_eq_obj *eqo;
3025 for_all_evt_queues(adapter, eqo, i) {
3026 sprintf(eqo->desc, "%s-q%d", netdev->name, i);
3027 vec = be_msix_vec_get(adapter, eqo);
3028 status = request_irq(vec, be_msix, 0, eqo->desc, eqo);
3035 for (i--, eqo = &adapter->eq_obj[i]; i >= 0; i--, eqo--)
3036 free_irq(be_msix_vec_get(adapter, eqo), eqo);
3037 dev_warn(&adapter->pdev->dev, "MSIX Request IRQ failed - err %d\n",
3039 be_msix_disable(adapter);
3043 static int be_irq_register(struct be_adapter *adapter)
3045 struct net_device *netdev = adapter->netdev;
3048 if (msix_enabled(adapter)) {
3049 status = be_msix_register(adapter);
3052 /* INTx is not supported for VF */
3053 if (!be_physfn(adapter))
3057 /* INTx: only the first EQ is used */
3058 netdev->irq = adapter->pdev->irq;
3059 status = request_irq(netdev->irq, be_intx, IRQF_SHARED, netdev->name,
3060 &adapter->eq_obj[0]);
3062 dev_err(&adapter->pdev->dev,
3063 "INTx request IRQ failed - err %d\n", status);
3067 adapter->isr_registered = true;
3071 static void be_irq_unregister(struct be_adapter *adapter)
3073 struct net_device *netdev = adapter->netdev;
3074 struct be_eq_obj *eqo;
3077 if (!adapter->isr_registered)
3081 if (!msix_enabled(adapter)) {
3082 free_irq(netdev->irq, &adapter->eq_obj[0]);
3087 for_all_evt_queues(adapter, eqo, i)
3088 free_irq(be_msix_vec_get(adapter, eqo), eqo);
3091 adapter->isr_registered = false;
3094 static void be_rx_qs_destroy(struct be_adapter *adapter)
3096 struct be_queue_info *q;
3097 struct be_rx_obj *rxo;
3100 for_all_rx_queues(adapter, rxo, i) {
3103 be_cmd_rxq_destroy(adapter, q);
3104 be_rx_cq_clean(rxo);
3106 be_queue_free(adapter, q);
3110 static int be_close(struct net_device *netdev)
3112 struct be_adapter *adapter = netdev_priv(netdev);
3113 struct be_eq_obj *eqo;
3116 /* This protection is needed as be_close() may be called even when the
3117 * adapter is in cleared state (after eeh perm failure)
3119 if (!(adapter->flags & BE_FLAGS_SETUP_DONE))
3122 be_roce_dev_close(adapter);
3124 if (adapter->flags & BE_FLAGS_NAPI_ENABLED) {
3125 for_all_evt_queues(adapter, eqo, i) {
3126 napi_disable(&eqo->napi);
3127 be_disable_busy_poll(eqo);
3129 adapter->flags &= ~BE_FLAGS_NAPI_ENABLED;
3132 be_async_mcc_disable(adapter);
3134 /* Wait for all pending tx completions to arrive so that
3135 * all tx skbs are freed.
3137 netif_tx_disable(netdev);
3138 be_tx_compl_clean(adapter);
3140 be_rx_qs_destroy(adapter);
3141 be_clear_uc_list(adapter);
3143 for_all_evt_queues(adapter, eqo, i) {
3144 if (msix_enabled(adapter))
3145 synchronize_irq(be_msix_vec_get(adapter, eqo));
3147 synchronize_irq(netdev->irq);
3151 be_irq_unregister(adapter);
3156 static int be_rx_qs_create(struct be_adapter *adapter)
3158 struct rss_info *rss = &adapter->rss_info;
3159 u8 rss_key[RSS_HASH_KEY_LEN];
3160 struct be_rx_obj *rxo;
3163 for_all_rx_queues(adapter, rxo, i) {
3164 rc = be_queue_alloc(adapter, &rxo->q, RX_Q_LEN,
3165 sizeof(struct be_eth_rx_d));
3170 if (adapter->need_def_rxq || !adapter->num_rss_qs) {
3171 rxo = default_rxo(adapter);
3172 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id,
3173 rx_frag_size, adapter->if_handle,
3174 false, &rxo->rss_id);
3179 for_all_rss_queues(adapter, rxo, i) {
3180 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id,
3181 rx_frag_size, adapter->if_handle,
3182 true, &rxo->rss_id);
3187 if (be_multi_rxq(adapter)) {
3188 for (j = 0; j < RSS_INDIR_TABLE_LEN; j += adapter->num_rss_qs) {
3189 for_all_rss_queues(adapter, rxo, i) {
3190 if ((j + i) >= RSS_INDIR_TABLE_LEN)
3192 rss->rsstable[j + i] = rxo->rss_id;
3193 rss->rss_queue[j + i] = i;
3196 rss->rss_flags = RSS_ENABLE_TCP_IPV4 | RSS_ENABLE_IPV4 |
3197 RSS_ENABLE_TCP_IPV6 | RSS_ENABLE_IPV6;
3199 if (!BEx_chip(adapter))
3200 rss->rss_flags |= RSS_ENABLE_UDP_IPV4 |
3201 RSS_ENABLE_UDP_IPV6;
3203 /* Disable RSS, if only default RX Q is created */
3204 rss->rss_flags = RSS_ENABLE_NONE;
3207 netdev_rss_key_fill(rss_key, RSS_HASH_KEY_LEN);
3208 rc = be_cmd_rss_config(adapter, rss->rsstable, rss->rss_flags,
3211 rss->rss_flags = RSS_ENABLE_NONE;
3215 memcpy(rss->rss_hkey, rss_key, RSS_HASH_KEY_LEN);
3217 /* First time posting */
3218 for_all_rx_queues(adapter, rxo, i)
3219 be_post_rx_frags(rxo, GFP_KERNEL, MAX_RX_POST);
3223 static int be_open(struct net_device *netdev)
3225 struct be_adapter *adapter = netdev_priv(netdev);
3226 struct be_eq_obj *eqo;
3227 struct be_rx_obj *rxo;
3228 struct be_tx_obj *txo;
3232 status = be_rx_qs_create(adapter);
3236 status = be_irq_register(adapter);
3240 for_all_rx_queues(adapter, rxo, i)
3241 be_cq_notify(adapter, rxo->cq.id, true, 0);
3243 for_all_tx_queues(adapter, txo, i)
3244 be_cq_notify(adapter, txo->cq.id, true, 0);
3246 be_async_mcc_enable(adapter);
3248 for_all_evt_queues(adapter, eqo, i) {
3249 napi_enable(&eqo->napi);
3250 be_enable_busy_poll(eqo);
3251 be_eq_notify(adapter, eqo->q.id, true, true, 0);
3253 adapter->flags |= BE_FLAGS_NAPI_ENABLED;
3255 status = be_cmd_link_status_query(adapter, NULL, &link_status, 0);
3257 be_link_status_update(adapter, link_status);
3259 netif_tx_start_all_queues(netdev);
3260 be_roce_dev_open(adapter);
3262 #ifdef CONFIG_BE2NET_VXLAN
3263 if (skyhawk_chip(adapter))
3264 vxlan_get_rx_port(netdev);
3269 be_close(adapter->netdev);
3273 static int be_setup_wol(struct be_adapter *adapter, bool enable)
3275 struct be_dma_mem cmd;
3281 cmd.size = sizeof(struct be_cmd_req_acpi_wol_magic_config);
3282 cmd.va = dma_zalloc_coherent(&adapter->pdev->dev, cmd.size, &cmd.dma,
3288 status = pci_write_config_dword(adapter->pdev,
3289 PCICFG_PM_CONTROL_OFFSET,
3290 PCICFG_PM_CONTROL_MASK);
3292 dev_err(&adapter->pdev->dev,
3293 "Could not enable Wake-on-lan\n");
3294 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va,
3298 status = be_cmd_enable_magic_wol(adapter,
3299 adapter->netdev->dev_addr,
3301 pci_enable_wake(adapter->pdev, PCI_D3hot, 1);
3302 pci_enable_wake(adapter->pdev, PCI_D3cold, 1);
3304 status = be_cmd_enable_magic_wol(adapter, mac, &cmd);
3305 pci_enable_wake(adapter->pdev, PCI_D3hot, 0);
3306 pci_enable_wake(adapter->pdev, PCI_D3cold, 0);
3309 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va, cmd.dma);
3313 static void be_vf_eth_addr_generate(struct be_adapter *adapter, u8 *mac)
3317 addr = jhash(adapter->netdev->dev_addr, ETH_ALEN, 0);
3319 mac[5] = (u8)(addr & 0xFF);
3320 mac[4] = (u8)((addr >> 8) & 0xFF);
3321 mac[3] = (u8)((addr >> 16) & 0xFF);
3322 /* Use the OUI from the current MAC address */
3323 memcpy(mac, adapter->netdev->dev_addr, 3);
3327 * Generate a seed MAC address from the PF MAC Address using jhash.
3328 * MAC Address for VFs are assigned incrementally starting from the seed.
3329 * These addresses are programmed in the ASIC by the PF and the VF driver
3330 * queries for the MAC address during its probe.
3332 static int be_vf_eth_addr_config(struct be_adapter *adapter)
3337 struct be_vf_cfg *vf_cfg;
3339 be_vf_eth_addr_generate(adapter, mac);
3341 for_all_vfs(adapter, vf_cfg, vf) {
3342 if (BEx_chip(adapter))
3343 status = be_cmd_pmac_add(adapter, mac,
3345 &vf_cfg->pmac_id, vf + 1);
3347 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
3351 dev_err(&adapter->pdev->dev,
3352 "Mac address assignment failed for VF %d\n",
3355 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
3362 static int be_vfs_mac_query(struct be_adapter *adapter)
3366 struct be_vf_cfg *vf_cfg;
3368 for_all_vfs(adapter, vf_cfg, vf) {
3369 status = be_cmd_get_active_mac(adapter, vf_cfg->pmac_id,
3370 mac, vf_cfg->if_handle,
3374 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
3379 static void be_vf_clear(struct be_adapter *adapter)
3381 struct be_vf_cfg *vf_cfg;
3384 if (pci_vfs_assigned(adapter->pdev)) {
3385 dev_warn(&adapter->pdev->dev,
3386 "VFs are assigned to VMs: not disabling VFs\n");
3390 pci_disable_sriov(adapter->pdev);
3392 for_all_vfs(adapter, vf_cfg, vf) {
3393 if (BEx_chip(adapter))
3394 be_cmd_pmac_del(adapter, vf_cfg->if_handle,
3395 vf_cfg->pmac_id, vf + 1);
3397 be_cmd_set_mac(adapter, NULL, vf_cfg->if_handle,
3400 be_cmd_if_destroy(adapter, vf_cfg->if_handle, vf + 1);
3403 kfree(adapter->vf_cfg);
3404 adapter->num_vfs = 0;
3405 adapter->flags &= ~BE_FLAGS_SRIOV_ENABLED;
3408 static void be_clear_queues(struct be_adapter *adapter)
3410 be_mcc_queues_destroy(adapter);
3411 be_rx_cqs_destroy(adapter);
3412 be_tx_queues_destroy(adapter);
3413 be_evt_queues_destroy(adapter);
3416 static void be_cancel_worker(struct be_adapter *adapter)
3418 if (adapter->flags & BE_FLAGS_WORKER_SCHEDULED) {
3419 cancel_delayed_work_sync(&adapter->work);
3420 adapter->flags &= ~BE_FLAGS_WORKER_SCHEDULED;
3424 static void be_cancel_err_detection(struct be_adapter *adapter)
3426 if (adapter->flags & BE_FLAGS_ERR_DETECTION_SCHEDULED) {
3427 cancel_delayed_work_sync(&adapter->be_err_detection_work);
3428 adapter->flags &= ~BE_FLAGS_ERR_DETECTION_SCHEDULED;
3432 static void be_mac_clear(struct be_adapter *adapter)
3434 if (adapter->pmac_id) {
3435 be_cmd_pmac_del(adapter, adapter->if_handle,
3436 adapter->pmac_id[0], 0);
3437 kfree(adapter->pmac_id);
3438 adapter->pmac_id = NULL;
3442 #ifdef CONFIG_BE2NET_VXLAN
3443 static void be_disable_vxlan_offloads(struct be_adapter *adapter)
3445 struct net_device *netdev = adapter->netdev;
3447 if (adapter->flags & BE_FLAGS_VXLAN_OFFLOADS)
3448 be_cmd_manage_iface(adapter, adapter->if_handle,
3449 OP_CONVERT_TUNNEL_TO_NORMAL);
3451 if (adapter->vxlan_port)
3452 be_cmd_set_vxlan_port(adapter, 0);
3454 adapter->flags &= ~BE_FLAGS_VXLAN_OFFLOADS;
3455 adapter->vxlan_port = 0;
3457 netdev->hw_enc_features = 0;
3458 netdev->hw_features &= ~(NETIF_F_GSO_UDP_TUNNEL);
3459 netdev->features &= ~(NETIF_F_GSO_UDP_TUNNEL);
3463 static u16 be_calculate_vf_qs(struct be_adapter *adapter, u16 num_vfs)
3465 struct be_resources res = adapter->pool_res;
3468 /* Distribute the queue resources equally among the PF and it's VFs
3469 * Do not distribute queue resources in multi-channel configuration.
3471 if (num_vfs && !be_is_mc(adapter)) {
3472 /* If number of VFs requested is 8 less than max supported,
3473 * assign 8 queue pairs to the PF and divide the remaining
3474 * resources evenly among the VFs
3476 if (num_vfs < (be_max_vfs(adapter) - 8))
3477 num_vf_qs = (res.max_rss_qs - 8) / num_vfs;
3479 num_vf_qs = res.max_rss_qs / num_vfs;
3481 /* Skyhawk-R chip supports only MAX_RSS_IFACES RSS capable
3482 * interfaces per port. Provide RSS on VFs, only if number
3483 * of VFs requested is less than MAX_RSS_IFACES limit.
3485 if (num_vfs >= MAX_RSS_IFACES)
3491 static int be_clear(struct be_adapter *adapter)
3493 struct pci_dev *pdev = adapter->pdev;
3496 be_cancel_worker(adapter);
3498 if (sriov_enabled(adapter))
3499 be_vf_clear(adapter);
3501 /* Re-configure FW to distribute resources evenly across max-supported
3502 * number of VFs, only when VFs are not already enabled.
3504 if (skyhawk_chip(adapter) && be_physfn(adapter) &&
3505 !pci_vfs_assigned(pdev)) {
3506 num_vf_qs = be_calculate_vf_qs(adapter,
3507 pci_sriov_get_totalvfs(pdev));
3508 be_cmd_set_sriov_config(adapter, adapter->pool_res,
3509 pci_sriov_get_totalvfs(pdev),
3513 #ifdef CONFIG_BE2NET_VXLAN
3514 be_disable_vxlan_offloads(adapter);
3516 /* delete the primary mac along with the uc-mac list */
3517 be_mac_clear(adapter);
3519 be_cmd_if_destroy(adapter, adapter->if_handle, 0);
3521 be_clear_queues(adapter);
3523 be_msix_disable(adapter);
3524 adapter->flags &= ~BE_FLAGS_SETUP_DONE;
3528 static int be_if_create(struct be_adapter *adapter, u32 *if_handle,
3529 u32 cap_flags, u32 vf)
3533 en_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3534 BE_IF_FLAGS_MULTICAST | BE_IF_FLAGS_PASS_L3L4_ERRORS |
3535 BE_IF_FLAGS_RSS | BE_IF_FLAGS_DEFQ_RSS;
3537 en_flags &= cap_flags;
3539 return be_cmd_if_create(adapter, cap_flags, en_flags, if_handle, vf);
3542 static int be_vfs_if_create(struct be_adapter *adapter)
3544 struct be_resources res = {0};
3545 struct be_vf_cfg *vf_cfg;
3549 /* If a FW profile exists, then cap_flags are updated */
3550 cap_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3551 BE_IF_FLAGS_MULTICAST;
3553 for_all_vfs(adapter, vf_cfg, vf) {
3554 if (!BE3_chip(adapter)) {
3555 status = be_cmd_get_profile_config(adapter, &res,
3559 cap_flags = res.if_cap_flags;
3560 /* Prevent VFs from enabling VLAN promiscuous
3563 cap_flags &= ~BE_IF_FLAGS_VLAN_PROMISCUOUS;
3567 status = be_if_create(adapter, &vf_cfg->if_handle,
3576 static int be_vf_setup_init(struct be_adapter *adapter)
3578 struct be_vf_cfg *vf_cfg;
3581 adapter->vf_cfg = kcalloc(adapter->num_vfs, sizeof(*vf_cfg),
3583 if (!adapter->vf_cfg)
3586 for_all_vfs(adapter, vf_cfg, vf) {
3587 vf_cfg->if_handle = -1;
3588 vf_cfg->pmac_id = -1;
3593 static int be_vf_setup(struct be_adapter *adapter)
3595 struct device *dev = &adapter->pdev->dev;
3596 struct be_vf_cfg *vf_cfg;
3597 int status, old_vfs, vf;
3599 old_vfs = pci_num_vf(adapter->pdev);
3601 status = be_vf_setup_init(adapter);
3606 for_all_vfs(adapter, vf_cfg, vf) {
3607 status = be_cmd_get_if_id(adapter, vf_cfg, vf);
3612 status = be_vfs_mac_query(adapter);
3616 status = be_vfs_if_create(adapter);
3620 status = be_vf_eth_addr_config(adapter);
3625 for_all_vfs(adapter, vf_cfg, vf) {
3626 /* Allow VFs to programs MAC/VLAN filters */
3627 status = be_cmd_get_fn_privileges(adapter, &vf_cfg->privileges,
3629 if (!status && !(vf_cfg->privileges & BE_PRIV_FILTMGMT)) {
3630 status = be_cmd_set_fn_privileges(adapter,
3631 vf_cfg->privileges |
3635 vf_cfg->privileges |= BE_PRIV_FILTMGMT;
3636 dev_info(dev, "VF%d has FILTMGMT privilege\n",
3641 /* Allow full available bandwidth */
3643 be_cmd_config_qos(adapter, 0, 0, vf + 1);
3646 be_cmd_enable_vf(adapter, vf + 1);
3647 be_cmd_set_logical_link_config(adapter,
3648 IFLA_VF_LINK_STATE_AUTO,
3654 status = pci_enable_sriov(adapter->pdev, adapter->num_vfs);
3656 dev_err(dev, "SRIOV enable failed\n");
3657 adapter->num_vfs = 0;
3662 adapter->flags |= BE_FLAGS_SRIOV_ENABLED;
3665 dev_err(dev, "VF setup failed\n");
3666 be_vf_clear(adapter);
3670 /* Converting function_mode bits on BE3 to SH mc_type enums */
3672 static u8 be_convert_mc_type(u32 function_mode)
3674 if (function_mode & VNIC_MODE && function_mode & QNQ_MODE)
3676 else if (function_mode & QNQ_MODE)
3678 else if (function_mode & VNIC_MODE)
3680 else if (function_mode & UMC_ENABLED)
3686 /* On BE2/BE3 FW does not suggest the supported limits */
3687 static void BEx_get_resources(struct be_adapter *adapter,
3688 struct be_resources *res)
3690 bool use_sriov = adapter->num_vfs ? 1 : 0;
3692 if (be_physfn(adapter))
3693 res->max_uc_mac = BE_UC_PMAC_COUNT;
3695 res->max_uc_mac = BE_VF_UC_PMAC_COUNT;
3697 adapter->mc_type = be_convert_mc_type(adapter->function_mode);
3699 if (be_is_mc(adapter)) {
3700 /* Assuming that there are 4 channels per port,
3701 * when multi-channel is enabled
3703 if (be_is_qnq_mode(adapter))
3704 res->max_vlans = BE_NUM_VLANS_SUPPORTED/8;
3706 /* In a non-qnq multichannel mode, the pvid
3707 * takes up one vlan entry
3709 res->max_vlans = (BE_NUM_VLANS_SUPPORTED / 4) - 1;
3711 res->max_vlans = BE_NUM_VLANS_SUPPORTED;
3714 res->max_mcast_mac = BE_MAX_MC;
3716 /* 1) For BE3 1Gb ports, FW does not support multiple TXQs
3717 * 2) Create multiple TX rings on a BE3-R multi-channel interface
3718 * *only* if it is RSS-capable.
3720 if (BE2_chip(adapter) || use_sriov || (adapter->port_num > 1) ||
3721 !be_physfn(adapter) || (be_is_mc(adapter) &&
3722 !(adapter->function_caps & BE_FUNCTION_CAPS_RSS))) {
3724 } else if (adapter->function_caps & BE_FUNCTION_CAPS_SUPER_NIC) {
3725 struct be_resources super_nic_res = {0};
3727 /* On a SuperNIC profile, the driver needs to use the
3728 * GET_PROFILE_CONFIG cmd to query the per-function TXQ limits
3730 be_cmd_get_profile_config(adapter, &super_nic_res,
3731 RESOURCE_LIMITS, 0);
3732 /* Some old versions of BE3 FW don't report max_tx_qs value */
3733 res->max_tx_qs = super_nic_res.max_tx_qs ? : BE3_MAX_TX_QS;
3735 res->max_tx_qs = BE3_MAX_TX_QS;
3738 if ((adapter->function_caps & BE_FUNCTION_CAPS_RSS) &&
3739 !use_sriov && be_physfn(adapter))
3740 res->max_rss_qs = (adapter->be3_native) ?
3741 BE3_MAX_RSS_QS : BE2_MAX_RSS_QS;
3742 res->max_rx_qs = res->max_rss_qs + 1;
3744 if (be_physfn(adapter))
3745 res->max_evt_qs = (be_max_vfs(adapter) > 0) ?
3746 BE3_SRIOV_MAX_EVT_QS : BE3_MAX_EVT_QS;
3748 res->max_evt_qs = 1;
3750 res->if_cap_flags = BE_IF_CAP_FLAGS_WANT;
3751 res->if_cap_flags &= ~BE_IF_FLAGS_DEFQ_RSS;
3752 if (!(adapter->function_caps & BE_FUNCTION_CAPS_RSS))
3753 res->if_cap_flags &= ~BE_IF_FLAGS_RSS;
3756 static void be_setup_init(struct be_adapter *adapter)
3758 adapter->vlan_prio_bmap = 0xff;
3759 adapter->phy.link_speed = -1;
3760 adapter->if_handle = -1;
3761 adapter->be3_native = false;
3762 adapter->if_flags = 0;
3763 if (be_physfn(adapter))
3764 adapter->cmd_privileges = MAX_PRIVILEGES;
3766 adapter->cmd_privileges = MIN_PRIVILEGES;
3769 static int be_get_sriov_config(struct be_adapter *adapter)
3771 struct be_resources res = {0};
3772 int max_vfs, old_vfs;
3774 be_cmd_get_profile_config(adapter, &res, RESOURCE_LIMITS, 0);
3776 /* Some old versions of BE3 FW don't report max_vfs value */
3777 if (BE3_chip(adapter) && !res.max_vfs) {
3778 max_vfs = pci_sriov_get_totalvfs(adapter->pdev);
3779 res.max_vfs = max_vfs > 0 ? min(MAX_VFS, max_vfs) : 0;
3782 adapter->pool_res = res;
3784 /* If during previous unload of the driver, the VFs were not disabled,
3785 * then we cannot rely on the PF POOL limits for the TotalVFs value.
3786 * Instead use the TotalVFs value stored in the pci-dev struct.
3788 old_vfs = pci_num_vf(adapter->pdev);
3790 dev_info(&adapter->pdev->dev, "%d VFs are already enabled\n",
3793 adapter->pool_res.max_vfs =
3794 pci_sriov_get_totalvfs(adapter->pdev);
3795 adapter->num_vfs = old_vfs;
3801 static void be_alloc_sriov_res(struct be_adapter *adapter)
3803 int old_vfs = pci_num_vf(adapter->pdev);
3807 be_get_sriov_config(adapter);
3810 pci_sriov_set_totalvfs(adapter->pdev, be_max_vfs(adapter));
3812 /* When the HW is in SRIOV capable configuration, the PF-pool
3813 * resources are given to PF during driver load, if there are no
3814 * old VFs. This facility is not available in BE3 FW.
3815 * Also, this is done by FW in Lancer chip.
3817 if (skyhawk_chip(adapter) && be_max_vfs(adapter) && !old_vfs) {
3818 num_vf_qs = be_calculate_vf_qs(adapter, 0);
3819 status = be_cmd_set_sriov_config(adapter, adapter->pool_res, 0,
3822 dev_err(&adapter->pdev->dev,
3823 "Failed to optimize SRIOV resources\n");
3827 static int be_get_resources(struct be_adapter *adapter)
3829 struct device *dev = &adapter->pdev->dev;
3830 struct be_resources res = {0};
3833 if (BEx_chip(adapter)) {
3834 BEx_get_resources(adapter, &res);
3838 /* For Lancer, SH etc read per-function resource limits from FW.
3839 * GET_FUNC_CONFIG returns per function guaranteed limits.
3840 * GET_PROFILE_CONFIG returns PCI-E related limits PF-pool limits
3842 if (!BEx_chip(adapter)) {
3843 status = be_cmd_get_func_config(adapter, &res);
3847 /* If a deafault RXQ must be created, we'll use up one RSSQ*/
3848 if (res.max_rss_qs && res.max_rss_qs == res.max_rx_qs &&
3849 !(res.if_cap_flags & BE_IF_FLAGS_DEFQ_RSS))
3850 res.max_rss_qs -= 1;
3852 /* If RoCE may be enabled stash away half the EQs for RoCE */
3853 if (be_roce_supported(adapter))
3854 res.max_evt_qs /= 2;
3858 /* If FW supports RSS default queue, then skip creating non-RSS
3859 * queue for non-IP traffic.
3861 adapter->need_def_rxq = (be_if_cap_flags(adapter) &
3862 BE_IF_FLAGS_DEFQ_RSS) ? 0 : 1;
3864 dev_info(dev, "Max: txqs %d, rxqs %d, rss %d, eqs %d, vfs %d\n",
3865 be_max_txqs(adapter), be_max_rxqs(adapter),
3866 be_max_rss(adapter), be_max_eqs(adapter),
3867 be_max_vfs(adapter));
3868 dev_info(dev, "Max: uc-macs %d, mc-macs %d, vlans %d\n",
3869 be_max_uc(adapter), be_max_mc(adapter),
3870 be_max_vlans(adapter));
3872 /* Sanitize cfg_num_qs based on HW and platform limits */
3873 adapter->cfg_num_qs = min_t(u16, netif_get_num_default_rss_queues(),
3874 be_max_qs(adapter));
3878 static int be_get_config(struct be_adapter *adapter)
3883 status = be_cmd_get_cntl_attributes(adapter);
3887 status = be_cmd_query_fw_cfg(adapter);
3891 if (BEx_chip(adapter)) {
3892 level = be_cmd_get_fw_log_level(adapter);
3893 adapter->msg_enable =
3894 level <= FW_LOG_LEVEL_DEFAULT ? NETIF_MSG_HW : 0;
3897 be_cmd_get_acpi_wol_cap(adapter);
3899 be_cmd_query_port_name(adapter);
3901 if (be_physfn(adapter)) {
3902 status = be_cmd_get_active_profile(adapter, &profile_id);
3904 dev_info(&adapter->pdev->dev,
3905 "Using profile 0x%x\n", profile_id);
3908 status = be_get_resources(adapter);
3912 adapter->pmac_id = kcalloc(be_max_uc(adapter),
3913 sizeof(*adapter->pmac_id), GFP_KERNEL);
3914 if (!adapter->pmac_id)
3920 static int be_mac_setup(struct be_adapter *adapter)
3925 if (is_zero_ether_addr(adapter->netdev->dev_addr)) {
3926 status = be_cmd_get_perm_mac(adapter, mac);
3930 memcpy(adapter->netdev->dev_addr, mac, ETH_ALEN);
3931 memcpy(adapter->netdev->perm_addr, mac, ETH_ALEN);
3933 /* Maybe the HW was reset; dev_addr must be re-programmed */
3934 memcpy(mac, adapter->netdev->dev_addr, ETH_ALEN);
3937 /* For BE3-R VFs, the PF programs the initial MAC address */
3938 if (!(BEx_chip(adapter) && be_virtfn(adapter)))
3939 be_cmd_pmac_add(adapter, mac, adapter->if_handle,
3940 &adapter->pmac_id[0], 0);
3944 static void be_schedule_worker(struct be_adapter *adapter)
3946 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
3947 adapter->flags |= BE_FLAGS_WORKER_SCHEDULED;
3950 static void be_schedule_err_detection(struct be_adapter *adapter)
3952 schedule_delayed_work(&adapter->be_err_detection_work,
3953 msecs_to_jiffies(1000));
3954 adapter->flags |= BE_FLAGS_ERR_DETECTION_SCHEDULED;
3957 static int be_setup_queues(struct be_adapter *adapter)
3959 struct net_device *netdev = adapter->netdev;
3962 status = be_evt_queues_create(adapter);
3966 status = be_tx_qs_create(adapter);
3970 status = be_rx_cqs_create(adapter);
3974 status = be_mcc_queues_create(adapter);
3978 status = netif_set_real_num_rx_queues(netdev, adapter->num_rx_qs);
3982 status = netif_set_real_num_tx_queues(netdev, adapter->num_tx_qs);
3988 dev_err(&adapter->pdev->dev, "queue_setup failed\n");
3992 int be_update_queues(struct be_adapter *adapter)
3994 struct net_device *netdev = adapter->netdev;
3997 if (netif_running(netdev))
4000 be_cancel_worker(adapter);
4002 /* If any vectors have been shared with RoCE we cannot re-program
4005 if (!adapter->num_msix_roce_vec)
4006 be_msix_disable(adapter);
4008 be_clear_queues(adapter);
4010 if (!msix_enabled(adapter)) {
4011 status = be_msix_enable(adapter);
4016 status = be_setup_queues(adapter);
4020 be_schedule_worker(adapter);
4022 if (netif_running(netdev))
4023 status = be_open(netdev);
4028 static inline int fw_major_num(const char *fw_ver)
4030 int fw_major = 0, i;
4032 i = sscanf(fw_ver, "%d.", &fw_major);
4039 /* If any VFs are already enabled don't FLR the PF */
4040 static bool be_reset_required(struct be_adapter *adapter)
4042 return pci_num_vf(adapter->pdev) ? false : true;
4045 /* Wait for the FW to be ready and perform the required initialization */
4046 static int be_func_init(struct be_adapter *adapter)
4050 status = be_fw_wait_ready(adapter);
4054 if (be_reset_required(adapter)) {
4055 status = be_cmd_reset_function(adapter);
4059 /* Wait for interrupts to quiesce after an FLR */
4062 /* We can clear all errors when function reset succeeds */
4063 be_clear_all_error(adapter);
4066 /* Tell FW we're ready to fire cmds */
4067 status = be_cmd_fw_init(adapter);
4071 /* Allow interrupts for other ULPs running on NIC function */
4072 be_intr_set(adapter, true);
4077 static int be_setup(struct be_adapter *adapter)
4079 struct device *dev = &adapter->pdev->dev;
4082 status = be_func_init(adapter);
4086 be_setup_init(adapter);
4088 if (!lancer_chip(adapter))
4089 be_cmd_req_native_mode(adapter);
4091 if (!BE2_chip(adapter) && be_physfn(adapter))
4092 be_alloc_sriov_res(adapter);
4094 status = be_get_config(adapter);
4098 status = be_msix_enable(adapter);
4102 status = be_if_create(adapter, &adapter->if_handle,
4103 be_if_cap_flags(adapter), 0);
4107 /* Updating real_num_tx/rx_queues() requires rtnl_lock() */
4109 status = be_setup_queues(adapter);
4114 be_cmd_get_fn_privileges(adapter, &adapter->cmd_privileges, 0);
4116 status = be_mac_setup(adapter);
4120 be_cmd_get_fw_ver(adapter);
4121 dev_info(dev, "FW version is %s\n", adapter->fw_ver);
4123 if (BE2_chip(adapter) && fw_major_num(adapter->fw_ver) < 4) {
4124 dev_err(dev, "Firmware on card is old(%s), IRQs may not work",
4126 dev_err(dev, "Please upgrade firmware to version >= 4.0\n");
4129 if (adapter->vlans_added)
4130 be_vid_config(adapter);
4132 be_set_rx_mode(adapter->netdev);
4134 status = be_cmd_set_flow_control(adapter, adapter->tx_fc,
4137 be_cmd_get_flow_control(adapter, &adapter->tx_fc,
4140 dev_info(&adapter->pdev->dev, "HW Flow control - TX:%d RX:%d\n",
4141 adapter->tx_fc, adapter->rx_fc);
4143 if (be_physfn(adapter))
4144 be_cmd_set_logical_link_config(adapter,
4145 IFLA_VF_LINK_STATE_AUTO, 0);
4147 if (adapter->num_vfs)
4148 be_vf_setup(adapter);
4150 status = be_cmd_get_phy_info(adapter);
4151 if (!status && be_pause_supported(adapter))
4152 adapter->phy.fc_autoneg = 1;
4154 be_schedule_worker(adapter);
4155 adapter->flags |= BE_FLAGS_SETUP_DONE;
4162 #ifdef CONFIG_NET_POLL_CONTROLLER
4163 static void be_netpoll(struct net_device *netdev)
4165 struct be_adapter *adapter = netdev_priv(netdev);
4166 struct be_eq_obj *eqo;
4169 for_all_evt_queues(adapter, eqo, i) {
4170 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
4171 napi_schedule(&eqo->napi);
4176 static char flash_cookie[2][16] = {"*** SE FLAS", "H DIRECTORY *** "};
4178 static bool phy_flashing_required(struct be_adapter *adapter)
4180 return (adapter->phy.phy_type == PHY_TYPE_TN_8022 &&
4181 adapter->phy.interface_type == PHY_TYPE_BASET_10GB);
4184 static bool is_comp_in_ufi(struct be_adapter *adapter,
4185 struct flash_section_info *fsec, int type)
4187 int i = 0, img_type = 0;
4188 struct flash_section_info_g2 *fsec_g2 = NULL;
4190 if (BE2_chip(adapter))
4191 fsec_g2 = (struct flash_section_info_g2 *)fsec;
4193 for (i = 0; i < MAX_FLASH_COMP; i++) {
4195 img_type = le32_to_cpu(fsec_g2->fsec_entry[i].type);
4197 img_type = le32_to_cpu(fsec->fsec_entry[i].type);
4199 if (img_type == type)
4206 static struct flash_section_info *get_fsec_info(struct be_adapter *adapter,
4208 const struct firmware *fw)
4210 struct flash_section_info *fsec = NULL;
4211 const u8 *p = fw->data;
4214 while (p < (fw->data + fw->size)) {
4215 fsec = (struct flash_section_info *)p;
4216 if (!memcmp(flash_cookie, fsec->cookie, sizeof(flash_cookie)))
4223 static int be_check_flash_crc(struct be_adapter *adapter, const u8 *p,
4224 u32 img_offset, u32 img_size, int hdr_size,
4225 u16 img_optype, bool *crc_match)
4231 status = be_cmd_get_flash_crc(adapter, crc, img_optype, img_offset,
4236 crc_offset = hdr_size + img_offset + img_size - 4;
4238 /* Skip flashing, if crc of flashed region matches */
4239 if (!memcmp(crc, p + crc_offset, 4))
4247 static int be_flash(struct be_adapter *adapter, const u8 *img,
4248 struct be_dma_mem *flash_cmd, int optype, int img_size,
4251 u32 flash_op, num_bytes, total_bytes = img_size, bytes_sent = 0;
4252 struct be_cmd_write_flashrom *req = flash_cmd->va;
4255 while (total_bytes) {
4256 num_bytes = min_t(u32, 32*1024, total_bytes);
4258 total_bytes -= num_bytes;
4261 if (optype == OPTYPE_PHY_FW)
4262 flash_op = FLASHROM_OPER_PHY_FLASH;
4264 flash_op = FLASHROM_OPER_FLASH;
4266 if (optype == OPTYPE_PHY_FW)
4267 flash_op = FLASHROM_OPER_PHY_SAVE;
4269 flash_op = FLASHROM_OPER_SAVE;
4272 memcpy(req->data_buf, img, num_bytes);
4274 status = be_cmd_write_flashrom(adapter, flash_cmd, optype,
4275 flash_op, img_offset +
4276 bytes_sent, num_bytes);
4277 if (base_status(status) == MCC_STATUS_ILLEGAL_REQUEST &&
4278 optype == OPTYPE_PHY_FW)
4283 bytes_sent += num_bytes;
4288 /* For BE2, BE3 and BE3-R */
4289 static int be_flash_BEx(struct be_adapter *adapter,
4290 const struct firmware *fw,
4291 struct be_dma_mem *flash_cmd, int num_of_images)
4293 int img_hdrs_size = (num_of_images * sizeof(struct image_hdr));
4294 struct device *dev = &adapter->pdev->dev;
4295 struct flash_section_info *fsec = NULL;
4296 int status, i, filehdr_size, num_comp;
4297 const struct flash_comp *pflashcomp;
4301 struct flash_comp gen3_flash_types[] = {
4302 { FLASH_iSCSI_PRIMARY_IMAGE_START_g3, OPTYPE_ISCSI_ACTIVE,
4303 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_iSCSI},
4304 { FLASH_REDBOOT_START_g3, OPTYPE_REDBOOT,
4305 FLASH_REDBOOT_IMAGE_MAX_SIZE_g3, IMAGE_BOOT_CODE},
4306 { FLASH_iSCSI_BIOS_START_g3, OPTYPE_BIOS,
4307 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_ISCSI},
4308 { FLASH_PXE_BIOS_START_g3, OPTYPE_PXE_BIOS,
4309 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_PXE},
4310 { FLASH_FCoE_BIOS_START_g3, OPTYPE_FCOE_BIOS,
4311 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_FCoE},
4312 { FLASH_iSCSI_BACKUP_IMAGE_START_g3, OPTYPE_ISCSI_BACKUP,
4313 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_iSCSI},
4314 { FLASH_FCoE_PRIMARY_IMAGE_START_g3, OPTYPE_FCOE_FW_ACTIVE,
4315 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_FCoE},
4316 { FLASH_FCoE_BACKUP_IMAGE_START_g3, OPTYPE_FCOE_FW_BACKUP,
4317 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_FCoE},
4318 { FLASH_NCSI_START_g3, OPTYPE_NCSI_FW,
4319 FLASH_NCSI_IMAGE_MAX_SIZE_g3, IMAGE_NCSI},
4320 { FLASH_PHY_FW_START_g3, OPTYPE_PHY_FW,
4321 FLASH_PHY_FW_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_PHY}
4324 struct flash_comp gen2_flash_types[] = {
4325 { FLASH_iSCSI_PRIMARY_IMAGE_START_g2, OPTYPE_ISCSI_ACTIVE,
4326 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_iSCSI},
4327 { FLASH_REDBOOT_START_g2, OPTYPE_REDBOOT,
4328 FLASH_REDBOOT_IMAGE_MAX_SIZE_g2, IMAGE_BOOT_CODE},
4329 { FLASH_iSCSI_BIOS_START_g2, OPTYPE_BIOS,
4330 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_ISCSI},
4331 { FLASH_PXE_BIOS_START_g2, OPTYPE_PXE_BIOS,
4332 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_PXE},
4333 { FLASH_FCoE_BIOS_START_g2, OPTYPE_FCOE_BIOS,
4334 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_FCoE},
4335 { FLASH_iSCSI_BACKUP_IMAGE_START_g2, OPTYPE_ISCSI_BACKUP,
4336 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_iSCSI},
4337 { FLASH_FCoE_PRIMARY_IMAGE_START_g2, OPTYPE_FCOE_FW_ACTIVE,
4338 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_FCoE},
4339 { FLASH_FCoE_BACKUP_IMAGE_START_g2, OPTYPE_FCOE_FW_BACKUP,
4340 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_FCoE}
4343 if (BE3_chip(adapter)) {
4344 pflashcomp = gen3_flash_types;
4345 filehdr_size = sizeof(struct flash_file_hdr_g3);
4346 num_comp = ARRAY_SIZE(gen3_flash_types);
4348 pflashcomp = gen2_flash_types;
4349 filehdr_size = sizeof(struct flash_file_hdr_g2);
4350 num_comp = ARRAY_SIZE(gen2_flash_types);
4354 /* Get flash section info*/
4355 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
4357 dev_err(dev, "Invalid Cookie. FW image may be corrupted\n");
4360 for (i = 0; i < num_comp; i++) {
4361 if (!is_comp_in_ufi(adapter, fsec, pflashcomp[i].img_type))
4364 if ((pflashcomp[i].optype == OPTYPE_NCSI_FW) &&
4365 memcmp(adapter->fw_ver, "3.102.148.0", 11) < 0)
4368 if (pflashcomp[i].optype == OPTYPE_PHY_FW &&
4369 !phy_flashing_required(adapter))
4372 if (pflashcomp[i].optype == OPTYPE_REDBOOT) {
4373 status = be_check_flash_crc(adapter, fw->data,
4374 pflashcomp[i].offset,
4378 OPTYPE_REDBOOT, &crc_match);
4381 "Could not get CRC for 0x%x region\n",
4382 pflashcomp[i].optype);
4390 p = fw->data + filehdr_size + pflashcomp[i].offset +
4392 if (p + pflashcomp[i].size > fw->data + fw->size)
4395 status = be_flash(adapter, p, flash_cmd, pflashcomp[i].optype,
4396 pflashcomp[i].size, 0);
4398 dev_err(dev, "Flashing section type 0x%x failed\n",
4399 pflashcomp[i].img_type);
4406 static u16 be_get_img_optype(struct flash_section_entry fsec_entry)
4408 u32 img_type = le32_to_cpu(fsec_entry.type);
4409 u16 img_optype = le16_to_cpu(fsec_entry.optype);
4411 if (img_optype != 0xFFFF)
4415 case IMAGE_FIRMWARE_iSCSI:
4416 img_optype = OPTYPE_ISCSI_ACTIVE;
4418 case IMAGE_BOOT_CODE:
4419 img_optype = OPTYPE_REDBOOT;
4421 case IMAGE_OPTION_ROM_ISCSI:
4422 img_optype = OPTYPE_BIOS;
4424 case IMAGE_OPTION_ROM_PXE:
4425 img_optype = OPTYPE_PXE_BIOS;
4427 case IMAGE_OPTION_ROM_FCoE:
4428 img_optype = OPTYPE_FCOE_BIOS;
4430 case IMAGE_FIRMWARE_BACKUP_iSCSI:
4431 img_optype = OPTYPE_ISCSI_BACKUP;
4434 img_optype = OPTYPE_NCSI_FW;
4436 case IMAGE_FLASHISM_JUMPVECTOR:
4437 img_optype = OPTYPE_FLASHISM_JUMPVECTOR;
4439 case IMAGE_FIRMWARE_PHY:
4440 img_optype = OPTYPE_SH_PHY_FW;
4442 case IMAGE_REDBOOT_DIR:
4443 img_optype = OPTYPE_REDBOOT_DIR;
4445 case IMAGE_REDBOOT_CONFIG:
4446 img_optype = OPTYPE_REDBOOT_CONFIG;
4449 img_optype = OPTYPE_UFI_DIR;
4458 static int be_flash_skyhawk(struct be_adapter *adapter,
4459 const struct firmware *fw,
4460 struct be_dma_mem *flash_cmd, int num_of_images)
4462 int img_hdrs_size = num_of_images * sizeof(struct image_hdr);
4463 bool crc_match, old_fw_img, flash_offset_support = true;
4464 struct device *dev = &adapter->pdev->dev;
4465 struct flash_section_info *fsec = NULL;
4466 u32 img_offset, img_size, img_type;
4467 u16 img_optype, flash_optype;
4468 int status, i, filehdr_size;
4471 filehdr_size = sizeof(struct flash_file_hdr_g3);
4472 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
4474 dev_err(dev, "Invalid Cookie. FW image may be corrupted\n");
4479 for (i = 0; i < le32_to_cpu(fsec->fsec_hdr.num_images); i++) {
4480 img_offset = le32_to_cpu(fsec->fsec_entry[i].offset);
4481 img_size = le32_to_cpu(fsec->fsec_entry[i].pad_size);
4482 img_type = le32_to_cpu(fsec->fsec_entry[i].type);
4483 img_optype = be_get_img_optype(fsec->fsec_entry[i]);
4484 old_fw_img = fsec->fsec_entry[i].optype == 0xFFFF;
4486 if (img_optype == 0xFFFF)
4489 if (flash_offset_support)
4490 flash_optype = OPTYPE_OFFSET_SPECIFIED;
4492 flash_optype = img_optype;
4494 /* Don't bother verifying CRC if an old FW image is being
4500 status = be_check_flash_crc(adapter, fw->data, img_offset,
4501 img_size, filehdr_size +
4502 img_hdrs_size, flash_optype,
4504 if (base_status(status) == MCC_STATUS_ILLEGAL_REQUEST ||
4505 base_status(status) == MCC_STATUS_ILLEGAL_FIELD) {
4506 /* The current FW image on the card does not support
4507 * OFFSET based flashing. Retry using older mechanism
4508 * of OPTYPE based flashing
4510 if (flash_optype == OPTYPE_OFFSET_SPECIFIED) {
4511 flash_offset_support = false;
4515 /* The current FW image on the card does not recognize
4516 * the new FLASH op_type. The FW download is partially
4517 * complete. Reboot the server now to enable FW image
4518 * to recognize the new FLASH op_type. To complete the
4519 * remaining process, download the same FW again after
4522 dev_err(dev, "Flash incomplete. Reset the server\n");
4523 dev_err(dev, "Download FW image again after reset\n");
4525 } else if (status) {
4526 dev_err(dev, "Could not get CRC for 0x%x region\n",
4535 p = fw->data + filehdr_size + img_offset + img_hdrs_size;
4536 if (p + img_size > fw->data + fw->size)
4539 status = be_flash(adapter, p, flash_cmd, flash_optype, img_size,
4542 /* The current FW image on the card does not support OFFSET
4543 * based flashing. Retry using older mechanism of OPTYPE based
4546 if (base_status(status) == MCC_STATUS_ILLEGAL_FIELD &&
4547 flash_optype == OPTYPE_OFFSET_SPECIFIED) {
4548 flash_offset_support = false;
4552 /* For old FW images ignore ILLEGAL_FIELD error or errors on
4556 (base_status(status) == MCC_STATUS_ILLEGAL_FIELD ||
4557 (img_optype == OPTYPE_UFI_DIR &&
4558 base_status(status) == MCC_STATUS_FAILED))) {
4560 } else if (status) {
4561 dev_err(dev, "Flashing section type 0x%x failed\n",
4569 static int lancer_fw_download(struct be_adapter *adapter,
4570 const struct firmware *fw)
4572 #define LANCER_FW_DOWNLOAD_CHUNK (32 * 1024)
4573 #define LANCER_FW_DOWNLOAD_LOCATION "/prg"
4574 struct device *dev = &adapter->pdev->dev;
4575 struct be_dma_mem flash_cmd;
4576 const u8 *data_ptr = NULL;
4577 u8 *dest_image_ptr = NULL;
4578 size_t image_size = 0;
4580 u32 data_written = 0;
4586 if (!IS_ALIGNED(fw->size, sizeof(u32))) {
4587 dev_err(dev, "FW image size should be multiple of 4\n");
4591 flash_cmd.size = sizeof(struct lancer_cmd_req_write_object)
4592 + LANCER_FW_DOWNLOAD_CHUNK;
4593 flash_cmd.va = dma_alloc_coherent(dev, flash_cmd.size,
4594 &flash_cmd.dma, GFP_KERNEL);
4598 dest_image_ptr = flash_cmd.va +
4599 sizeof(struct lancer_cmd_req_write_object);
4600 image_size = fw->size;
4601 data_ptr = fw->data;
4603 while (image_size) {
4604 chunk_size = min_t(u32, image_size, LANCER_FW_DOWNLOAD_CHUNK);
4606 /* Copy the image chunk content. */
4607 memcpy(dest_image_ptr, data_ptr, chunk_size);
4609 status = lancer_cmd_write_object(adapter, &flash_cmd,
4611 LANCER_FW_DOWNLOAD_LOCATION,
4612 &data_written, &change_status,
4617 offset += data_written;
4618 data_ptr += data_written;
4619 image_size -= data_written;
4623 /* Commit the FW written */
4624 status = lancer_cmd_write_object(adapter, &flash_cmd,
4626 LANCER_FW_DOWNLOAD_LOCATION,
4627 &data_written, &change_status,
4631 dma_free_coherent(dev, flash_cmd.size, flash_cmd.va, flash_cmd.dma);
4633 dev_err(dev, "Firmware load error\n");
4634 return be_cmd_status(status);
4637 dev_info(dev, "Firmware flashed successfully\n");
4639 if (change_status == LANCER_FW_RESET_NEEDED) {
4640 dev_info(dev, "Resetting adapter to activate new FW\n");
4641 status = lancer_physdev_ctrl(adapter,
4642 PHYSDEV_CONTROL_FW_RESET_MASK);
4644 dev_err(dev, "Adapter busy, could not reset FW\n");
4645 dev_err(dev, "Reboot server to activate new FW\n");
4647 } else if (change_status != LANCER_NO_RESET_NEEDED) {
4648 dev_info(dev, "Reboot server to activate new FW\n");
4658 #define SH_P2_UFI 11
4660 static int be_get_ufi_type(struct be_adapter *adapter,
4661 struct flash_file_hdr_g3 *fhdr)
4664 dev_err(&adapter->pdev->dev, "Invalid FW UFI file");
4668 /* First letter of the build version is used to identify
4669 * which chip this image file is meant for.
4671 switch (fhdr->build[0]) {
4672 case BLD_STR_UFI_TYPE_SH:
4673 return (fhdr->asic_type_rev == ASIC_REV_P2) ? SH_P2_UFI :
4675 case BLD_STR_UFI_TYPE_BE3:
4676 return (fhdr->asic_type_rev == ASIC_REV_B0) ? BE3R_UFI :
4678 case BLD_STR_UFI_TYPE_BE2:
4685 /* Check if the flash image file is compatible with the adapter that
4687 * BE3 chips with asic-rev B0 must be flashed only with BE3R_UFI type.
4688 * Skyhawk chips with asic-rev P2 must be flashed only with SH_P2_UFI type.
4690 static bool be_check_ufi_compatibility(struct be_adapter *adapter,
4691 struct flash_file_hdr_g3 *fhdr)
4693 int ufi_type = be_get_ufi_type(adapter, fhdr);
4697 return skyhawk_chip(adapter);
4699 return (skyhawk_chip(adapter) &&
4700 adapter->asic_rev < ASIC_REV_P2);
4702 return BE3_chip(adapter);
4704 return (BE3_chip(adapter) && adapter->asic_rev < ASIC_REV_B0);
4706 return BE2_chip(adapter);
4712 static int be_fw_download(struct be_adapter *adapter, const struct firmware* fw)
4714 struct device *dev = &adapter->pdev->dev;
4715 struct flash_file_hdr_g3 *fhdr3;
4716 struct image_hdr *img_hdr_ptr;
4717 int status = 0, i, num_imgs;
4718 struct be_dma_mem flash_cmd;
4720 fhdr3 = (struct flash_file_hdr_g3 *)fw->data;
4721 if (!be_check_ufi_compatibility(adapter, fhdr3)) {
4722 dev_err(dev, "Flash image is not compatible with adapter\n");
4726 flash_cmd.size = sizeof(struct be_cmd_write_flashrom);
4727 flash_cmd.va = dma_alloc_coherent(dev, flash_cmd.size, &flash_cmd.dma,
4732 num_imgs = le32_to_cpu(fhdr3->num_imgs);
4733 for (i = 0; i < num_imgs; i++) {
4734 img_hdr_ptr = (struct image_hdr *)(fw->data +
4735 (sizeof(struct flash_file_hdr_g3) +
4736 i * sizeof(struct image_hdr)));
4737 if (!BE2_chip(adapter) &&
4738 le32_to_cpu(img_hdr_ptr->imageid) != 1)
4741 if (skyhawk_chip(adapter))
4742 status = be_flash_skyhawk(adapter, fw, &flash_cmd,
4745 status = be_flash_BEx(adapter, fw, &flash_cmd,
4749 dma_free_coherent(dev, flash_cmd.size, flash_cmd.va, flash_cmd.dma);
4751 dev_info(dev, "Firmware flashed successfully\n");
4756 int be_load_fw(struct be_adapter *adapter, u8 *fw_file)
4758 const struct firmware *fw;
4761 if (!netif_running(adapter->netdev)) {
4762 dev_err(&adapter->pdev->dev,
4763 "Firmware load not allowed (interface is down)\n");
4767 status = request_firmware(&fw, fw_file, &adapter->pdev->dev);
4771 dev_info(&adapter->pdev->dev, "Flashing firmware file %s\n", fw_file);
4773 if (lancer_chip(adapter))
4774 status = lancer_fw_download(adapter, fw);
4776 status = be_fw_download(adapter, fw);
4779 be_cmd_get_fw_ver(adapter);
4782 release_firmware(fw);
4786 static int be_ndo_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
4789 struct be_adapter *adapter = netdev_priv(dev);
4790 struct nlattr *attr, *br_spec;
4795 if (!sriov_enabled(adapter))
4798 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4802 nla_for_each_nested(attr, br_spec, rem) {
4803 if (nla_type(attr) != IFLA_BRIDGE_MODE)
4806 if (nla_len(attr) < sizeof(mode))
4809 mode = nla_get_u16(attr);
4810 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
4813 status = be_cmd_set_hsw_config(adapter, 0, 0,
4815 mode == BRIDGE_MODE_VEPA ?
4816 PORT_FWD_TYPE_VEPA :
4821 dev_info(&adapter->pdev->dev, "enabled switch mode: %s\n",
4822 mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
4827 dev_err(&adapter->pdev->dev, "Failed to set switch mode %s\n",
4828 mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
4833 static int be_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4834 struct net_device *dev, u32 filter_mask)
4836 struct be_adapter *adapter = netdev_priv(dev);
4840 if (!sriov_enabled(adapter))
4843 /* BE and Lancer chips support VEB mode only */
4844 if (BEx_chip(adapter) || lancer_chip(adapter)) {
4845 hsw_mode = PORT_FWD_TYPE_VEB;
4847 status = be_cmd_get_hsw_config(adapter, NULL, 0,
4848 adapter->if_handle, &hsw_mode);
4853 return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
4854 hsw_mode == PORT_FWD_TYPE_VEPA ?
4855 BRIDGE_MODE_VEPA : BRIDGE_MODE_VEB,
4859 #ifdef CONFIG_BE2NET_VXLAN
4860 /* VxLAN offload Notes:
4862 * The stack defines tunnel offload flags (hw_enc_features) for IP and doesn't
4863 * distinguish various types of transports (VxLAN, GRE, NVGRE ..). So, offload
4864 * is expected to work across all types of IP tunnels once exported. Skyhawk
4865 * supports offloads for either VxLAN or NVGRE, exclusively. So we export VxLAN
4866 * offloads in hw_enc_features only when a VxLAN port is added. If other (non
4867 * VxLAN) tunnels are configured while VxLAN offloads are enabled, offloads for
4868 * those other tunnels are unexported on the fly through ndo_features_check().
4870 * Skyhawk supports VxLAN offloads only for one UDP dport. So, if the stack
4871 * adds more than one port, disable offloads and don't re-enable them again
4872 * until after all the tunnels are removed.
4874 static void be_add_vxlan_port(struct net_device *netdev, sa_family_t sa_family,
4877 struct be_adapter *adapter = netdev_priv(netdev);
4878 struct device *dev = &adapter->pdev->dev;
4881 if (lancer_chip(adapter) || BEx_chip(adapter))
4884 if (adapter->flags & BE_FLAGS_VXLAN_OFFLOADS) {
4886 "Only one UDP port supported for VxLAN offloads\n");
4887 dev_info(dev, "Disabling VxLAN offloads\n");
4888 adapter->vxlan_port_count++;
4892 if (adapter->vxlan_port_count++ >= 1)
4895 status = be_cmd_manage_iface(adapter, adapter->if_handle,
4896 OP_CONVERT_NORMAL_TO_TUNNEL);
4898 dev_warn(dev, "Failed to convert normal interface to tunnel\n");
4902 status = be_cmd_set_vxlan_port(adapter, port);
4904 dev_warn(dev, "Failed to add VxLAN port\n");
4907 adapter->flags |= BE_FLAGS_VXLAN_OFFLOADS;
4908 adapter->vxlan_port = port;
4910 netdev->hw_enc_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
4911 NETIF_F_TSO | NETIF_F_TSO6 |
4912 NETIF_F_GSO_UDP_TUNNEL;
4913 netdev->hw_features |= NETIF_F_GSO_UDP_TUNNEL;
4914 netdev->features |= NETIF_F_GSO_UDP_TUNNEL;
4916 dev_info(dev, "Enabled VxLAN offloads for UDP port %d\n",
4920 be_disable_vxlan_offloads(adapter);
4923 static void be_del_vxlan_port(struct net_device *netdev, sa_family_t sa_family,
4926 struct be_adapter *adapter = netdev_priv(netdev);
4928 if (lancer_chip(adapter) || BEx_chip(adapter))
4931 if (adapter->vxlan_port != port)
4934 be_disable_vxlan_offloads(adapter);
4936 dev_info(&adapter->pdev->dev,
4937 "Disabled VxLAN offloads for UDP port %d\n",
4940 adapter->vxlan_port_count--;
4943 static netdev_features_t be_features_check(struct sk_buff *skb,
4944 struct net_device *dev,
4945 netdev_features_t features)
4947 struct be_adapter *adapter = netdev_priv(dev);
4950 /* The code below restricts offload features for some tunneled packets.
4951 * Offload features for normal (non tunnel) packets are unchanged.
4953 if (!skb->encapsulation ||
4954 !(adapter->flags & BE_FLAGS_VXLAN_OFFLOADS))
4957 /* It's an encapsulated packet and VxLAN offloads are enabled. We
4958 * should disable tunnel offload features if it's not a VxLAN packet,
4959 * as tunnel offloads have been enabled only for VxLAN. This is done to
4960 * allow other tunneled traffic like GRE work fine while VxLAN
4961 * offloads are configured in Skyhawk-R.
4963 switch (vlan_get_protocol(skb)) {
4964 case htons(ETH_P_IP):
4965 l4_hdr = ip_hdr(skb)->protocol;
4967 case htons(ETH_P_IPV6):
4968 l4_hdr = ipv6_hdr(skb)->nexthdr;
4974 if (l4_hdr != IPPROTO_UDP ||
4975 skb->inner_protocol_type != ENCAP_TYPE_ETHER ||
4976 skb->inner_protocol != htons(ETH_P_TEB) ||
4977 skb_inner_mac_header(skb) - skb_transport_header(skb) !=
4978 sizeof(struct udphdr) + sizeof(struct vxlanhdr))
4979 return features & ~(NETIF_F_ALL_CSUM | NETIF_F_GSO_MASK);
4985 static const struct net_device_ops be_netdev_ops = {
4986 .ndo_open = be_open,
4987 .ndo_stop = be_close,
4988 .ndo_start_xmit = be_xmit,
4989 .ndo_set_rx_mode = be_set_rx_mode,
4990 .ndo_set_mac_address = be_mac_addr_set,
4991 .ndo_change_mtu = be_change_mtu,
4992 .ndo_get_stats64 = be_get_stats64,
4993 .ndo_validate_addr = eth_validate_addr,
4994 .ndo_vlan_rx_add_vid = be_vlan_add_vid,
4995 .ndo_vlan_rx_kill_vid = be_vlan_rem_vid,
4996 .ndo_set_vf_mac = be_set_vf_mac,
4997 .ndo_set_vf_vlan = be_set_vf_vlan,
4998 .ndo_set_vf_rate = be_set_vf_tx_rate,
4999 .ndo_get_vf_config = be_get_vf_config,
5000 .ndo_set_vf_link_state = be_set_vf_link_state,
5001 #ifdef CONFIG_NET_POLL_CONTROLLER
5002 .ndo_poll_controller = be_netpoll,
5004 .ndo_bridge_setlink = be_ndo_bridge_setlink,
5005 .ndo_bridge_getlink = be_ndo_bridge_getlink,
5006 #ifdef CONFIG_NET_RX_BUSY_POLL
5007 .ndo_busy_poll = be_busy_poll,
5009 #ifdef CONFIG_BE2NET_VXLAN
5010 .ndo_add_vxlan_port = be_add_vxlan_port,
5011 .ndo_del_vxlan_port = be_del_vxlan_port,
5012 .ndo_features_check = be_features_check,
5016 static void be_netdev_init(struct net_device *netdev)
5018 struct be_adapter *adapter = netdev_priv(netdev);
5020 netdev->hw_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
5021 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM | NETIF_F_RXCSUM |
5022 NETIF_F_HW_VLAN_CTAG_TX;
5023 if (be_multi_rxq(adapter))
5024 netdev->hw_features |= NETIF_F_RXHASH;
5026 netdev->features |= netdev->hw_features |
5027 NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_FILTER;
5029 netdev->vlan_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
5030 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
5032 netdev->priv_flags |= IFF_UNICAST_FLT;
5034 netdev->flags |= IFF_MULTICAST;
5036 netif_set_gso_max_size(netdev, 65535 - ETH_HLEN);
5038 netdev->netdev_ops = &be_netdev_ops;
5040 netdev->ethtool_ops = &be_ethtool_ops;
5043 static void be_cleanup(struct be_adapter *adapter)
5045 struct net_device *netdev = adapter->netdev;
5048 netif_device_detach(netdev);
5049 if (netif_running(netdev))
5056 static int be_resume(struct be_adapter *adapter)
5058 struct net_device *netdev = adapter->netdev;
5061 status = be_setup(adapter);
5065 if (netif_running(netdev)) {
5066 status = be_open(netdev);
5071 netif_device_attach(netdev);
5076 static int be_err_recover(struct be_adapter *adapter)
5078 struct device *dev = &adapter->pdev->dev;
5081 status = be_resume(adapter);
5085 dev_info(dev, "Adapter recovery successful\n");
5088 if (be_physfn(adapter))
5089 dev_err(dev, "Adapter recovery failed\n");
5091 dev_err(dev, "Re-trying adapter recovery\n");
5096 static void be_err_detection_task(struct work_struct *work)
5098 struct be_adapter *adapter =
5099 container_of(work, struct be_adapter,
5100 be_err_detection_work.work);
5103 be_detect_error(adapter);
5105 if (adapter->hw_error) {
5106 be_cleanup(adapter);
5108 /* As of now error recovery support is in Lancer only */
5109 if (lancer_chip(adapter))
5110 status = be_err_recover(adapter);
5113 /* Always attempt recovery on VFs */
5114 if (!status || be_virtfn(adapter))
5115 be_schedule_err_detection(adapter);
5118 static void be_log_sfp_info(struct be_adapter *adapter)
5122 status = be_cmd_query_sfp_info(adapter);
5124 dev_err(&adapter->pdev->dev,
5125 "Unqualified SFP+ detected on %c from %s part no: %s",
5126 adapter->port_name, adapter->phy.vendor_name,
5127 adapter->phy.vendor_pn);
5129 adapter->flags &= ~BE_FLAGS_EVT_INCOMPATIBLE_SFP;
5132 static void be_worker(struct work_struct *work)
5134 struct be_adapter *adapter =
5135 container_of(work, struct be_adapter, work.work);
5136 struct be_rx_obj *rxo;
5139 /* when interrupts are not yet enabled, just reap any pending
5142 if (!netif_running(adapter->netdev)) {
5144 be_process_mcc(adapter);
5149 if (!adapter->stats_cmd_sent) {
5150 if (lancer_chip(adapter))
5151 lancer_cmd_get_pport_stats(adapter,
5152 &adapter->stats_cmd);
5154 be_cmd_get_stats(adapter, &adapter->stats_cmd);
5157 if (be_physfn(adapter) &&
5158 MODULO(adapter->work_counter, adapter->be_get_temp_freq) == 0)
5159 be_cmd_get_die_temperature(adapter);
5161 for_all_rx_queues(adapter, rxo, i) {
5162 /* Replenish RX-queues starved due to memory
5163 * allocation failures.
5165 if (rxo->rx_post_starved)
5166 be_post_rx_frags(rxo, GFP_KERNEL, MAX_RX_POST);
5169 be_eqd_update(adapter);
5171 if (adapter->flags & BE_FLAGS_EVT_INCOMPATIBLE_SFP)
5172 be_log_sfp_info(adapter);
5175 adapter->work_counter++;
5176 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
5179 static void be_unmap_pci_bars(struct be_adapter *adapter)
5182 pci_iounmap(adapter->pdev, adapter->csr);
5184 pci_iounmap(adapter->pdev, adapter->db);
5187 static int db_bar(struct be_adapter *adapter)
5189 if (lancer_chip(adapter) || !be_physfn(adapter))
5195 static int be_roce_map_pci_bars(struct be_adapter *adapter)
5197 if (skyhawk_chip(adapter)) {
5198 adapter->roce_db.size = 4096;
5199 adapter->roce_db.io_addr = pci_resource_start(adapter->pdev,
5201 adapter->roce_db.total_size = pci_resource_len(adapter->pdev,
5207 static int be_map_pci_bars(struct be_adapter *adapter)
5209 struct pci_dev *pdev = adapter->pdev;
5213 pci_read_config_dword(adapter->pdev, SLI_INTF_REG_OFFSET, &sli_intf);
5214 adapter->sli_family = (sli_intf & SLI_INTF_FAMILY_MASK) >>
5215 SLI_INTF_FAMILY_SHIFT;
5216 adapter->virtfn = (sli_intf & SLI_INTF_FT_MASK) ? 1 : 0;
5218 if (BEx_chip(adapter) && be_physfn(adapter)) {
5219 adapter->csr = pci_iomap(pdev, 2, 0);
5224 addr = pci_iomap(pdev, db_bar(adapter), 0);
5229 if (skyhawk_chip(adapter) || BEx_chip(adapter)) {
5230 if (be_physfn(adapter)) {
5231 /* PCICFG is the 2nd BAR in BE2 */
5232 addr = pci_iomap(pdev, BE2_chip(adapter) ? 1 : 0, 0);
5235 adapter->pcicfg = addr;
5237 adapter->pcicfg = adapter->db + SRIOV_VF_PCICFG_OFFSET;
5241 be_roce_map_pci_bars(adapter);
5245 dev_err(&pdev->dev, "Error in mapping PCI BARs\n");
5246 be_unmap_pci_bars(adapter);
5250 static void be_drv_cleanup(struct be_adapter *adapter)
5252 struct be_dma_mem *mem = &adapter->mbox_mem_alloced;
5253 struct device *dev = &adapter->pdev->dev;
5256 dma_free_coherent(dev, mem->size, mem->va, mem->dma);
5258 mem = &adapter->rx_filter;
5260 dma_free_coherent(dev, mem->size, mem->va, mem->dma);
5262 mem = &adapter->stats_cmd;
5264 dma_free_coherent(dev, mem->size, mem->va, mem->dma);
5267 /* Allocate and initialize various fields in be_adapter struct */
5268 static int be_drv_init(struct be_adapter *adapter)
5270 struct be_dma_mem *mbox_mem_alloc = &adapter->mbox_mem_alloced;
5271 struct be_dma_mem *mbox_mem_align = &adapter->mbox_mem;
5272 struct be_dma_mem *rx_filter = &adapter->rx_filter;
5273 struct be_dma_mem *stats_cmd = &adapter->stats_cmd;
5274 struct device *dev = &adapter->pdev->dev;
5277 mbox_mem_alloc->size = sizeof(struct be_mcc_mailbox) + 16;
5278 mbox_mem_alloc->va = dma_alloc_coherent(dev, mbox_mem_alloc->size,
5279 &mbox_mem_alloc->dma,
5281 if (!mbox_mem_alloc->va)
5284 mbox_mem_align->size = sizeof(struct be_mcc_mailbox);
5285 mbox_mem_align->va = PTR_ALIGN(mbox_mem_alloc->va, 16);
5286 mbox_mem_align->dma = PTR_ALIGN(mbox_mem_alloc->dma, 16);
5287 memset(mbox_mem_align->va, 0, sizeof(struct be_mcc_mailbox));
5289 rx_filter->size = sizeof(struct be_cmd_req_rx_filter);
5290 rx_filter->va = dma_zalloc_coherent(dev, rx_filter->size,
5291 &rx_filter->dma, GFP_KERNEL);
5292 if (!rx_filter->va) {
5297 if (lancer_chip(adapter))
5298 stats_cmd->size = sizeof(struct lancer_cmd_req_pport_stats);
5299 else if (BE2_chip(adapter))
5300 stats_cmd->size = sizeof(struct be_cmd_req_get_stats_v0);
5301 else if (BE3_chip(adapter))
5302 stats_cmd->size = sizeof(struct be_cmd_req_get_stats_v1);
5304 stats_cmd->size = sizeof(struct be_cmd_req_get_stats_v2);
5305 stats_cmd->va = dma_zalloc_coherent(dev, stats_cmd->size,
5306 &stats_cmd->dma, GFP_KERNEL);
5307 if (!stats_cmd->va) {
5309 goto free_rx_filter;
5312 mutex_init(&adapter->mbox_lock);
5313 spin_lock_init(&adapter->mcc_lock);
5314 spin_lock_init(&adapter->mcc_cq_lock);
5315 init_completion(&adapter->et_cmd_compl);
5317 pci_save_state(adapter->pdev);
5319 INIT_DELAYED_WORK(&adapter->work, be_worker);
5320 INIT_DELAYED_WORK(&adapter->be_err_detection_work,
5321 be_err_detection_task);
5323 adapter->rx_fc = true;
5324 adapter->tx_fc = true;
5326 /* Must be a power of 2 or else MODULO will BUG_ON */
5327 adapter->be_get_temp_freq = 64;
5332 dma_free_coherent(dev, rx_filter->size, rx_filter->va, rx_filter->dma);
5334 dma_free_coherent(dev, mbox_mem_alloc->size, mbox_mem_alloc->va,
5335 mbox_mem_alloc->dma);
5339 static void be_remove(struct pci_dev *pdev)
5341 struct be_adapter *adapter = pci_get_drvdata(pdev);
5346 be_roce_dev_remove(adapter);
5347 be_intr_set(adapter, false);
5349 be_cancel_err_detection(adapter);
5351 unregister_netdev(adapter->netdev);
5355 /* tell fw we're done with firing cmds */
5356 be_cmd_fw_clean(adapter);
5358 be_unmap_pci_bars(adapter);
5359 be_drv_cleanup(adapter);
5361 pci_disable_pcie_error_reporting(pdev);
5363 pci_release_regions(pdev);
5364 pci_disable_device(pdev);
5366 free_netdev(adapter->netdev);
5369 static char *mc_name(struct be_adapter *adapter)
5371 char *str = ""; /* default */
5373 switch (adapter->mc_type) {
5399 static inline char *func_name(struct be_adapter *adapter)
5401 return be_physfn(adapter) ? "PF" : "VF";
5404 static inline char *nic_name(struct pci_dev *pdev)
5406 switch (pdev->device) {
5413 return OC_NAME_LANCER;
5424 static int be_probe(struct pci_dev *pdev, const struct pci_device_id *pdev_id)
5426 struct be_adapter *adapter;
5427 struct net_device *netdev;
5430 dev_info(&pdev->dev, "%s version is %s\n", DRV_NAME, DRV_VER);
5432 status = pci_enable_device(pdev);
5436 status = pci_request_regions(pdev, DRV_NAME);
5439 pci_set_master(pdev);
5441 netdev = alloc_etherdev_mqs(sizeof(*adapter), MAX_TX_QS, MAX_RX_QS);
5446 adapter = netdev_priv(netdev);
5447 adapter->pdev = pdev;
5448 pci_set_drvdata(pdev, adapter);
5449 adapter->netdev = netdev;
5450 SET_NETDEV_DEV(netdev, &pdev->dev);
5452 status = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
5454 netdev->features |= NETIF_F_HIGHDMA;
5456 status = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
5458 dev_err(&pdev->dev, "Could not set PCI DMA Mask\n");
5463 status = pci_enable_pcie_error_reporting(pdev);
5465 dev_info(&pdev->dev, "PCIe error reporting enabled\n");
5467 status = be_map_pci_bars(adapter);
5471 status = be_drv_init(adapter);
5475 status = be_setup(adapter);
5479 be_netdev_init(netdev);
5480 status = register_netdev(netdev);
5484 be_roce_dev_add(adapter);
5486 be_schedule_err_detection(adapter);
5488 dev_info(&pdev->dev, "%s: %s %s port %c\n", nic_name(pdev),
5489 func_name(adapter), mc_name(adapter), adapter->port_name);
5496 be_drv_cleanup(adapter);
5498 be_unmap_pci_bars(adapter);
5500 free_netdev(netdev);
5502 pci_release_regions(pdev);
5504 pci_disable_device(pdev);
5506 dev_err(&pdev->dev, "%s initialization failed\n", nic_name(pdev));
5510 static int be_suspend(struct pci_dev *pdev, pm_message_t state)
5512 struct be_adapter *adapter = pci_get_drvdata(pdev);
5514 if (adapter->wol_en)
5515 be_setup_wol(adapter, true);
5517 be_intr_set(adapter, false);
5518 be_cancel_err_detection(adapter);
5520 be_cleanup(adapter);
5522 pci_save_state(pdev);
5523 pci_disable_device(pdev);
5524 pci_set_power_state(pdev, pci_choose_state(pdev, state));
5528 static int be_pci_resume(struct pci_dev *pdev)
5530 struct be_adapter *adapter = pci_get_drvdata(pdev);
5533 status = pci_enable_device(pdev);
5537 pci_set_power_state(pdev, PCI_D0);
5538 pci_restore_state(pdev);
5540 status = be_resume(adapter);
5544 be_schedule_err_detection(adapter);
5546 if (adapter->wol_en)
5547 be_setup_wol(adapter, false);
5553 * An FLR will stop BE from DMAing any data.
5555 static void be_shutdown(struct pci_dev *pdev)
5557 struct be_adapter *adapter = pci_get_drvdata(pdev);
5562 be_roce_dev_shutdown(adapter);
5563 cancel_delayed_work_sync(&adapter->work);
5564 be_cancel_err_detection(adapter);
5566 netif_device_detach(adapter->netdev);
5568 be_cmd_reset_function(adapter);
5570 pci_disable_device(pdev);
5573 static pci_ers_result_t be_eeh_err_detected(struct pci_dev *pdev,
5574 pci_channel_state_t state)
5576 struct be_adapter *adapter = pci_get_drvdata(pdev);
5578 dev_err(&adapter->pdev->dev, "EEH error detected\n");
5580 if (!adapter->eeh_error) {
5581 adapter->eeh_error = true;
5583 be_cancel_err_detection(adapter);
5585 be_cleanup(adapter);
5588 if (state == pci_channel_io_perm_failure)
5589 return PCI_ERS_RESULT_DISCONNECT;
5591 pci_disable_device(pdev);
5593 /* The error could cause the FW to trigger a flash debug dump.
5594 * Resetting the card while flash dump is in progress
5595 * can cause it not to recover; wait for it to finish.
5596 * Wait only for first function as it is needed only once per
5599 if (pdev->devfn == 0)
5602 return PCI_ERS_RESULT_NEED_RESET;
5605 static pci_ers_result_t be_eeh_reset(struct pci_dev *pdev)
5607 struct be_adapter *adapter = pci_get_drvdata(pdev);
5610 dev_info(&adapter->pdev->dev, "EEH reset\n");
5612 status = pci_enable_device(pdev);
5614 return PCI_ERS_RESULT_DISCONNECT;
5616 pci_set_master(pdev);
5617 pci_set_power_state(pdev, PCI_D0);
5618 pci_restore_state(pdev);
5620 /* Check if card is ok and fw is ready */
5621 dev_info(&adapter->pdev->dev,
5622 "Waiting for FW to be ready after EEH reset\n");
5623 status = be_fw_wait_ready(adapter);
5625 return PCI_ERS_RESULT_DISCONNECT;
5627 pci_cleanup_aer_uncorrect_error_status(pdev);
5628 be_clear_all_error(adapter);
5629 return PCI_ERS_RESULT_RECOVERED;
5632 static void be_eeh_resume(struct pci_dev *pdev)
5635 struct be_adapter *adapter = pci_get_drvdata(pdev);
5637 dev_info(&adapter->pdev->dev, "EEH resume\n");
5639 pci_save_state(pdev);
5641 status = be_resume(adapter);
5645 be_schedule_err_detection(adapter);
5648 dev_err(&adapter->pdev->dev, "EEH resume failed\n");
5651 static int be_pci_sriov_configure(struct pci_dev *pdev, int num_vfs)
5653 struct be_adapter *adapter = pci_get_drvdata(pdev);
5658 be_vf_clear(adapter);
5660 adapter->num_vfs = num_vfs;
5662 if (adapter->num_vfs == 0 && pci_vfs_assigned(pdev)) {
5663 dev_warn(&pdev->dev,
5664 "Cannot disable VFs while they are assigned\n");
5668 /* When the HW is in SRIOV capable configuration, the PF-pool resources
5669 * are equally distributed across the max-number of VFs. The user may
5670 * request only a subset of the max-vfs to be enabled.
5671 * Based on num_vfs, redistribute the resources across num_vfs so that
5672 * each VF will have access to more number of resources.
5673 * This facility is not available in BE3 FW.
5674 * Also, this is done by FW in Lancer chip.
5676 if (skyhawk_chip(adapter) && !pci_num_vf(pdev)) {
5677 num_vf_qs = be_calculate_vf_qs(adapter, adapter->num_vfs);
5678 status = be_cmd_set_sriov_config(adapter, adapter->pool_res,
5679 adapter->num_vfs, num_vf_qs);
5682 "Failed to optimize SR-IOV resources\n");
5685 status = be_get_resources(adapter);
5687 return be_cmd_status(status);
5689 /* Updating real_num_tx/rx_queues() requires rtnl_lock() */
5691 status = be_update_queues(adapter);
5694 return be_cmd_status(status);
5696 if (adapter->num_vfs)
5697 status = be_vf_setup(adapter);
5700 return adapter->num_vfs;
5705 static const struct pci_error_handlers be_eeh_handlers = {
5706 .error_detected = be_eeh_err_detected,
5707 .slot_reset = be_eeh_reset,
5708 .resume = be_eeh_resume,
5711 static struct pci_driver be_driver = {
5713 .id_table = be_dev_ids,
5715 .remove = be_remove,
5716 .suspend = be_suspend,
5717 .resume = be_pci_resume,
5718 .shutdown = be_shutdown,
5719 .sriov_configure = be_pci_sriov_configure,
5720 .err_handler = &be_eeh_handlers
5723 static int __init be_init_module(void)
5725 if (rx_frag_size != 8192 && rx_frag_size != 4096 &&
5726 rx_frag_size != 2048) {
5727 printk(KERN_WARNING DRV_NAME
5728 " : Module param rx_frag_size must be 2048/4096/8192."
5730 rx_frag_size = 2048;
5734 pr_info(DRV_NAME " : Module param num_vfs is obsolete.");
5735 pr_info(DRV_NAME " : Use sysfs method to enable VFs\n");
5738 return pci_register_driver(&be_driver);
5740 module_init(be_init_module);
5742 static void __exit be_exit_module(void)
5744 pci_unregister_driver(&be_driver);
5746 module_exit(be_exit_module);