2 * Copyright (C) 2005 - 2014 Emulex
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License version 2
7 * as published by the Free Software Foundation. The full GNU General
8 * Public License is included in this distribution in the file called COPYING.
10 * Contact Information:
11 * linux-drivers@emulex.com
15 * Costa Mesa, CA 92626
18 #include <linux/prefetch.h>
19 #include <linux/module.h>
22 #include <asm/div64.h>
23 #include <linux/aer.h>
24 #include <linux/if_bridge.h>
25 #include <net/busy_poll.h>
26 #include <net/vxlan.h>
28 MODULE_VERSION(DRV_VER);
29 MODULE_DEVICE_TABLE(pci, be_dev_ids);
30 MODULE_DESCRIPTION(DRV_DESC " " DRV_VER);
31 MODULE_AUTHOR("Emulex Corporation");
32 MODULE_LICENSE("GPL");
34 static unsigned int num_vfs;
35 module_param(num_vfs, uint, S_IRUGO);
36 MODULE_PARM_DESC(num_vfs, "Number of PCI VFs to initialize");
38 static ushort rx_frag_size = 2048;
39 module_param(rx_frag_size, ushort, S_IRUGO);
40 MODULE_PARM_DESC(rx_frag_size, "Size of a fragment that holds rcvd data.");
42 static DEFINE_PCI_DEVICE_TABLE(be_dev_ids) = {
43 { PCI_DEVICE(BE_VENDOR_ID, BE_DEVICE_ID1) },
44 { PCI_DEVICE(BE_VENDOR_ID, BE_DEVICE_ID2) },
45 { PCI_DEVICE(BE_VENDOR_ID, OC_DEVICE_ID1) },
46 { PCI_DEVICE(BE_VENDOR_ID, OC_DEVICE_ID2) },
47 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID3)},
48 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID4)},
49 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID5)},
50 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID6)},
53 MODULE_DEVICE_TABLE(pci, be_dev_ids);
54 /* UE Status Low CSR */
55 static const char * const ue_status_low_desc[] = {
89 /* UE Status High CSR */
90 static const char * const ue_status_hi_desc[] = {
126 static void be_queue_free(struct be_adapter *adapter, struct be_queue_info *q)
128 struct be_dma_mem *mem = &q->dma_mem;
130 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
136 static int be_queue_alloc(struct be_adapter *adapter, struct be_queue_info *q,
137 u16 len, u16 entry_size)
139 struct be_dma_mem *mem = &q->dma_mem;
141 memset(q, 0, sizeof(*q));
143 q->entry_size = entry_size;
144 mem->size = len * entry_size;
145 mem->va = dma_zalloc_coherent(&adapter->pdev->dev, mem->size, &mem->dma,
152 static void be_reg_intr_set(struct be_adapter *adapter, bool enable)
156 pci_read_config_dword(adapter->pdev, PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET,
158 enabled = reg & MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
160 if (!enabled && enable)
161 reg |= MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
162 else if (enabled && !enable)
163 reg &= ~MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
167 pci_write_config_dword(adapter->pdev,
168 PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET, reg);
171 static void be_intr_set(struct be_adapter *adapter, bool enable)
175 /* On lancer interrupts can't be controlled via this register */
176 if (lancer_chip(adapter))
179 if (adapter->eeh_error)
182 status = be_cmd_intr_set(adapter, enable);
184 be_reg_intr_set(adapter, enable);
187 static void be_rxq_notify(struct be_adapter *adapter, u16 qid, u16 posted)
190 val |= qid & DB_RQ_RING_ID_MASK;
191 val |= posted << DB_RQ_NUM_POSTED_SHIFT;
194 iowrite32(val, adapter->db + DB_RQ_OFFSET);
197 static void be_txq_notify(struct be_adapter *adapter, struct be_tx_obj *txo,
201 val |= txo->q.id & DB_TXULP_RING_ID_MASK;
202 val |= (posted & DB_TXULP_NUM_POSTED_MASK) << DB_TXULP_NUM_POSTED_SHIFT;
205 iowrite32(val, adapter->db + txo->db_offset);
208 static void be_eq_notify(struct be_adapter *adapter, u16 qid,
209 bool arm, bool clear_int, u16 num_popped)
212 val |= qid & DB_EQ_RING_ID_MASK;
213 val |= ((qid & DB_EQ_RING_ID_EXT_MASK) << DB_EQ_RING_ID_EXT_MASK_SHIFT);
215 if (adapter->eeh_error)
219 val |= 1 << DB_EQ_REARM_SHIFT;
221 val |= 1 << DB_EQ_CLR_SHIFT;
222 val |= 1 << DB_EQ_EVNT_SHIFT;
223 val |= num_popped << DB_EQ_NUM_POPPED_SHIFT;
224 iowrite32(val, adapter->db + DB_EQ_OFFSET);
227 void be_cq_notify(struct be_adapter *adapter, u16 qid, bool arm, u16 num_popped)
230 val |= qid & DB_CQ_RING_ID_MASK;
231 val |= ((qid & DB_CQ_RING_ID_EXT_MASK) <<
232 DB_CQ_RING_ID_EXT_MASK_SHIFT);
234 if (adapter->eeh_error)
238 val |= 1 << DB_CQ_REARM_SHIFT;
239 val |= num_popped << DB_CQ_NUM_POPPED_SHIFT;
240 iowrite32(val, adapter->db + DB_CQ_OFFSET);
243 static int be_mac_addr_set(struct net_device *netdev, void *p)
245 struct be_adapter *adapter = netdev_priv(netdev);
246 struct device *dev = &adapter->pdev->dev;
247 struct sockaddr *addr = p;
250 u32 old_pmac_id = adapter->pmac_id[0], curr_pmac_id = 0;
252 if (!is_valid_ether_addr(addr->sa_data))
253 return -EADDRNOTAVAIL;
255 /* Proceed further only if, User provided MAC is different
258 if (ether_addr_equal(addr->sa_data, netdev->dev_addr))
261 /* The PMAC_ADD cmd may fail if the VF doesn't have FILTMGMT
262 * privilege or if PF did not provision the new MAC address.
263 * On BE3, this cmd will always fail if the VF doesn't have the
264 * FILTMGMT privilege. This failure is OK, only if the PF programmed
265 * the MAC for the VF.
267 status = be_cmd_pmac_add(adapter, (u8 *)addr->sa_data,
268 adapter->if_handle, &adapter->pmac_id[0], 0);
270 curr_pmac_id = adapter->pmac_id[0];
272 /* Delete the old programmed MAC. This call may fail if the
273 * old MAC was already deleted by the PF driver.
275 if (adapter->pmac_id[0] != old_pmac_id)
276 be_cmd_pmac_del(adapter, adapter->if_handle,
280 /* Decide if the new MAC is successfully activated only after
283 status = be_cmd_get_active_mac(adapter, curr_pmac_id, mac,
284 adapter->if_handle, true, 0);
288 /* The MAC change did not happen, either due to lack of privilege
289 * or PF didn't pre-provision.
291 if (!ether_addr_equal(addr->sa_data, mac)) {
296 memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
297 dev_info(dev, "MAC address changed to %pM\n", mac);
300 dev_warn(dev, "MAC address change to %pM failed\n", addr->sa_data);
304 /* BE2 supports only v0 cmd */
305 static void *hw_stats_from_cmd(struct be_adapter *adapter)
307 if (BE2_chip(adapter)) {
308 struct be_cmd_resp_get_stats_v0 *cmd = adapter->stats_cmd.va;
310 return &cmd->hw_stats;
311 } else if (BE3_chip(adapter)) {
312 struct be_cmd_resp_get_stats_v1 *cmd = adapter->stats_cmd.va;
314 return &cmd->hw_stats;
316 struct be_cmd_resp_get_stats_v2 *cmd = adapter->stats_cmd.va;
318 return &cmd->hw_stats;
322 /* BE2 supports only v0 cmd */
323 static void *be_erx_stats_from_cmd(struct be_adapter *adapter)
325 if (BE2_chip(adapter)) {
326 struct be_hw_stats_v0 *hw_stats = hw_stats_from_cmd(adapter);
328 return &hw_stats->erx;
329 } else if (BE3_chip(adapter)) {
330 struct be_hw_stats_v1 *hw_stats = hw_stats_from_cmd(adapter);
332 return &hw_stats->erx;
334 struct be_hw_stats_v2 *hw_stats = hw_stats_from_cmd(adapter);
336 return &hw_stats->erx;
340 static void populate_be_v0_stats(struct be_adapter *adapter)
342 struct be_hw_stats_v0 *hw_stats = hw_stats_from_cmd(adapter);
343 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
344 struct be_rxf_stats_v0 *rxf_stats = &hw_stats->rxf;
345 struct be_port_rxf_stats_v0 *port_stats =
346 &rxf_stats->port[adapter->port_num];
347 struct be_drv_stats *drvs = &adapter->drv_stats;
349 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
350 drvs->rx_pause_frames = port_stats->rx_pause_frames;
351 drvs->rx_crc_errors = port_stats->rx_crc_errors;
352 drvs->rx_control_frames = port_stats->rx_control_frames;
353 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
354 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
355 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
356 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
357 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
358 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
359 drvs->rxpp_fifo_overflow_drop = port_stats->rx_fifo_overflow;
360 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
361 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
362 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
363 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
364 drvs->rx_input_fifo_overflow_drop = port_stats->rx_input_fifo_overflow;
365 drvs->rx_dropped_header_too_small =
366 port_stats->rx_dropped_header_too_small;
367 drvs->rx_address_filtered =
368 port_stats->rx_address_filtered +
369 port_stats->rx_vlan_filtered;
370 drvs->rx_alignment_symbol_errors =
371 port_stats->rx_alignment_symbol_errors;
373 drvs->tx_pauseframes = port_stats->tx_pauseframes;
374 drvs->tx_controlframes = port_stats->tx_controlframes;
376 if (adapter->port_num)
377 drvs->jabber_events = rxf_stats->port1_jabber_events;
379 drvs->jabber_events = rxf_stats->port0_jabber_events;
380 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
381 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
382 drvs->forwarded_packets = rxf_stats->forwarded_packets;
383 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
384 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
385 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
386 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
389 static void populate_be_v1_stats(struct be_adapter *adapter)
391 struct be_hw_stats_v1 *hw_stats = hw_stats_from_cmd(adapter);
392 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
393 struct be_rxf_stats_v1 *rxf_stats = &hw_stats->rxf;
394 struct be_port_rxf_stats_v1 *port_stats =
395 &rxf_stats->port[adapter->port_num];
396 struct be_drv_stats *drvs = &adapter->drv_stats;
398 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
399 drvs->pmem_fifo_overflow_drop = port_stats->pmem_fifo_overflow_drop;
400 drvs->rx_priority_pause_frames = port_stats->rx_priority_pause_frames;
401 drvs->rx_pause_frames = port_stats->rx_pause_frames;
402 drvs->rx_crc_errors = port_stats->rx_crc_errors;
403 drvs->rx_control_frames = port_stats->rx_control_frames;
404 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
405 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
406 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
407 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
408 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
409 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
410 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
411 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
412 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
413 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
414 drvs->rx_dropped_header_too_small =
415 port_stats->rx_dropped_header_too_small;
416 drvs->rx_input_fifo_overflow_drop =
417 port_stats->rx_input_fifo_overflow_drop;
418 drvs->rx_address_filtered = port_stats->rx_address_filtered;
419 drvs->rx_alignment_symbol_errors =
420 port_stats->rx_alignment_symbol_errors;
421 drvs->rxpp_fifo_overflow_drop = port_stats->rxpp_fifo_overflow_drop;
422 drvs->tx_pauseframes = port_stats->tx_pauseframes;
423 drvs->tx_controlframes = port_stats->tx_controlframes;
424 drvs->tx_priority_pauseframes = port_stats->tx_priority_pauseframes;
425 drvs->jabber_events = port_stats->jabber_events;
426 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
427 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
428 drvs->forwarded_packets = rxf_stats->forwarded_packets;
429 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
430 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
431 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
432 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
435 static void populate_be_v2_stats(struct be_adapter *adapter)
437 struct be_hw_stats_v2 *hw_stats = hw_stats_from_cmd(adapter);
438 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
439 struct be_rxf_stats_v2 *rxf_stats = &hw_stats->rxf;
440 struct be_port_rxf_stats_v2 *port_stats =
441 &rxf_stats->port[adapter->port_num];
442 struct be_drv_stats *drvs = &adapter->drv_stats;
444 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
445 drvs->pmem_fifo_overflow_drop = port_stats->pmem_fifo_overflow_drop;
446 drvs->rx_priority_pause_frames = port_stats->rx_priority_pause_frames;
447 drvs->rx_pause_frames = port_stats->rx_pause_frames;
448 drvs->rx_crc_errors = port_stats->rx_crc_errors;
449 drvs->rx_control_frames = port_stats->rx_control_frames;
450 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
451 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
452 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
453 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
454 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
455 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
456 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
457 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
458 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
459 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
460 drvs->rx_dropped_header_too_small =
461 port_stats->rx_dropped_header_too_small;
462 drvs->rx_input_fifo_overflow_drop =
463 port_stats->rx_input_fifo_overflow_drop;
464 drvs->rx_address_filtered = port_stats->rx_address_filtered;
465 drvs->rx_alignment_symbol_errors =
466 port_stats->rx_alignment_symbol_errors;
467 drvs->rxpp_fifo_overflow_drop = port_stats->rxpp_fifo_overflow_drop;
468 drvs->tx_pauseframes = port_stats->tx_pauseframes;
469 drvs->tx_controlframes = port_stats->tx_controlframes;
470 drvs->tx_priority_pauseframes = port_stats->tx_priority_pauseframes;
471 drvs->jabber_events = port_stats->jabber_events;
472 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
473 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
474 drvs->forwarded_packets = rxf_stats->forwarded_packets;
475 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
476 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
477 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
478 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
479 if (be_roce_supported(adapter)) {
480 drvs->rx_roce_bytes_lsd = port_stats->roce_bytes_received_lsd;
481 drvs->rx_roce_bytes_msd = port_stats->roce_bytes_received_msd;
482 drvs->rx_roce_frames = port_stats->roce_frames_received;
483 drvs->roce_drops_crc = port_stats->roce_drops_crc;
484 drvs->roce_drops_payload_len =
485 port_stats->roce_drops_payload_len;
489 static void populate_lancer_stats(struct be_adapter *adapter)
492 struct be_drv_stats *drvs = &adapter->drv_stats;
493 struct lancer_pport_stats *pport_stats = pport_stats_from_cmd(adapter);
495 be_dws_le_to_cpu(pport_stats, sizeof(*pport_stats));
496 drvs->rx_pause_frames = pport_stats->rx_pause_frames_lo;
497 drvs->rx_crc_errors = pport_stats->rx_crc_errors_lo;
498 drvs->rx_control_frames = pport_stats->rx_control_frames_lo;
499 drvs->rx_in_range_errors = pport_stats->rx_in_range_errors;
500 drvs->rx_frame_too_long = pport_stats->rx_frames_too_long_lo;
501 drvs->rx_dropped_runt = pport_stats->rx_dropped_runt;
502 drvs->rx_ip_checksum_errs = pport_stats->rx_ip_checksum_errors;
503 drvs->rx_tcp_checksum_errs = pport_stats->rx_tcp_checksum_errors;
504 drvs->rx_udp_checksum_errs = pport_stats->rx_udp_checksum_errors;
505 drvs->rx_dropped_tcp_length =
506 pport_stats->rx_dropped_invalid_tcp_length;
507 drvs->rx_dropped_too_small = pport_stats->rx_dropped_too_small;
508 drvs->rx_dropped_too_short = pport_stats->rx_dropped_too_short;
509 drvs->rx_out_range_errors = pport_stats->rx_out_of_range_errors;
510 drvs->rx_dropped_header_too_small =
511 pport_stats->rx_dropped_header_too_small;
512 drvs->rx_input_fifo_overflow_drop = pport_stats->rx_fifo_overflow;
513 drvs->rx_address_filtered =
514 pport_stats->rx_address_filtered +
515 pport_stats->rx_vlan_filtered;
516 drvs->rx_alignment_symbol_errors = pport_stats->rx_symbol_errors_lo;
517 drvs->rxpp_fifo_overflow_drop = pport_stats->rx_fifo_overflow;
518 drvs->tx_pauseframes = pport_stats->tx_pause_frames_lo;
519 drvs->tx_controlframes = pport_stats->tx_control_frames_lo;
520 drvs->jabber_events = pport_stats->rx_jabbers;
521 drvs->forwarded_packets = pport_stats->num_forwards_lo;
522 drvs->rx_drops_mtu = pport_stats->rx_drops_mtu_lo;
523 drvs->rx_drops_too_many_frags =
524 pport_stats->rx_drops_too_many_frags_lo;
527 static void accumulate_16bit_val(u32 *acc, u16 val)
529 #define lo(x) (x & 0xFFFF)
530 #define hi(x) (x & 0xFFFF0000)
531 bool wrapped = val < lo(*acc);
532 u32 newacc = hi(*acc) + val;
536 ACCESS_ONCE(*acc) = newacc;
539 static void populate_erx_stats(struct be_adapter *adapter,
540 struct be_rx_obj *rxo, u32 erx_stat)
542 if (!BEx_chip(adapter))
543 rx_stats(rxo)->rx_drops_no_frags = erx_stat;
545 /* below erx HW counter can actually wrap around after
546 * 65535. Driver accumulates a 32-bit value
548 accumulate_16bit_val(&rx_stats(rxo)->rx_drops_no_frags,
552 void be_parse_stats(struct be_adapter *adapter)
554 struct be_erx_stats_v2 *erx = be_erx_stats_from_cmd(adapter);
555 struct be_rx_obj *rxo;
559 if (lancer_chip(adapter)) {
560 populate_lancer_stats(adapter);
562 if (BE2_chip(adapter))
563 populate_be_v0_stats(adapter);
564 else if (BE3_chip(adapter))
566 populate_be_v1_stats(adapter);
568 populate_be_v2_stats(adapter);
570 /* erx_v2 is longer than v0, v1. use v2 for v0, v1 access */
571 for_all_rx_queues(adapter, rxo, i) {
572 erx_stat = erx->rx_drops_no_fragments[rxo->q.id];
573 populate_erx_stats(adapter, rxo, erx_stat);
578 static struct rtnl_link_stats64 *be_get_stats64(struct net_device *netdev,
579 struct rtnl_link_stats64 *stats)
581 struct be_adapter *adapter = netdev_priv(netdev);
582 struct be_drv_stats *drvs = &adapter->drv_stats;
583 struct be_rx_obj *rxo;
584 struct be_tx_obj *txo;
589 for_all_rx_queues(adapter, rxo, i) {
590 const struct be_rx_stats *rx_stats = rx_stats(rxo);
592 start = u64_stats_fetch_begin_irq(&rx_stats->sync);
593 pkts = rx_stats(rxo)->rx_pkts;
594 bytes = rx_stats(rxo)->rx_bytes;
595 } while (u64_stats_fetch_retry_irq(&rx_stats->sync, start));
596 stats->rx_packets += pkts;
597 stats->rx_bytes += bytes;
598 stats->multicast += rx_stats(rxo)->rx_mcast_pkts;
599 stats->rx_dropped += rx_stats(rxo)->rx_drops_no_skbs +
600 rx_stats(rxo)->rx_drops_no_frags;
603 for_all_tx_queues(adapter, txo, i) {
604 const struct be_tx_stats *tx_stats = tx_stats(txo);
606 start = u64_stats_fetch_begin_irq(&tx_stats->sync);
607 pkts = tx_stats(txo)->tx_pkts;
608 bytes = tx_stats(txo)->tx_bytes;
609 } while (u64_stats_fetch_retry_irq(&tx_stats->sync, start));
610 stats->tx_packets += pkts;
611 stats->tx_bytes += bytes;
614 /* bad pkts received */
615 stats->rx_errors = drvs->rx_crc_errors +
616 drvs->rx_alignment_symbol_errors +
617 drvs->rx_in_range_errors +
618 drvs->rx_out_range_errors +
619 drvs->rx_frame_too_long +
620 drvs->rx_dropped_too_small +
621 drvs->rx_dropped_too_short +
622 drvs->rx_dropped_header_too_small +
623 drvs->rx_dropped_tcp_length +
624 drvs->rx_dropped_runt;
626 /* detailed rx errors */
627 stats->rx_length_errors = drvs->rx_in_range_errors +
628 drvs->rx_out_range_errors +
629 drvs->rx_frame_too_long;
631 stats->rx_crc_errors = drvs->rx_crc_errors;
633 /* frame alignment errors */
634 stats->rx_frame_errors = drvs->rx_alignment_symbol_errors;
636 /* receiver fifo overrun */
637 /* drops_no_pbuf is no per i/f, it's per BE card */
638 stats->rx_fifo_errors = drvs->rxpp_fifo_overflow_drop +
639 drvs->rx_input_fifo_overflow_drop +
640 drvs->rx_drops_no_pbuf;
644 void be_link_status_update(struct be_adapter *adapter, u8 link_status)
646 struct net_device *netdev = adapter->netdev;
648 if (!(adapter->flags & BE_FLAGS_LINK_STATUS_INIT)) {
649 netif_carrier_off(netdev);
650 adapter->flags |= BE_FLAGS_LINK_STATUS_INIT;
654 netif_carrier_on(netdev);
656 netif_carrier_off(netdev);
659 static void be_tx_stats_update(struct be_tx_obj *txo,
660 u32 wrb_cnt, u32 copied, u32 gso_segs,
663 struct be_tx_stats *stats = tx_stats(txo);
665 u64_stats_update_begin(&stats->sync);
667 stats->tx_wrbs += wrb_cnt;
668 stats->tx_bytes += copied;
669 stats->tx_pkts += (gso_segs ? gso_segs : 1);
672 u64_stats_update_end(&stats->sync);
675 /* Determine number of WRB entries needed to xmit data in an skb */
676 static u32 wrb_cnt_for_skb(struct be_adapter *adapter, struct sk_buff *skb,
679 int cnt = (skb->len > skb->data_len);
681 cnt += skb_shinfo(skb)->nr_frags;
683 /* to account for hdr wrb */
685 if (lancer_chip(adapter) || !(cnt & 1)) {
688 /* add a dummy to make it an even num */
692 BUG_ON(cnt > BE_MAX_TX_FRAG_COUNT);
696 static inline void wrb_fill(struct be_eth_wrb *wrb, u64 addr, int len)
698 wrb->frag_pa_hi = upper_32_bits(addr);
699 wrb->frag_pa_lo = addr & 0xFFFFFFFF;
700 wrb->frag_len = len & ETH_WRB_FRAG_LEN_MASK;
704 static inline u16 be_get_tx_vlan_tag(struct be_adapter *adapter,
710 vlan_tag = vlan_tx_tag_get(skb);
711 vlan_prio = (vlan_tag & VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
712 /* If vlan priority provided by OS is NOT in available bmap */
713 if (!(adapter->vlan_prio_bmap & (1 << vlan_prio)))
714 vlan_tag = (vlan_tag & ~VLAN_PRIO_MASK) |
715 adapter->recommended_prio;
720 /* Used only for IP tunnel packets */
721 static u16 skb_inner_ip_proto(struct sk_buff *skb)
723 return (inner_ip_hdr(skb)->version == 4) ?
724 inner_ip_hdr(skb)->protocol : inner_ipv6_hdr(skb)->nexthdr;
727 static u16 skb_ip_proto(struct sk_buff *skb)
729 return (ip_hdr(skb)->version == 4) ?
730 ip_hdr(skb)->protocol : ipv6_hdr(skb)->nexthdr;
733 static void wrb_fill_hdr(struct be_adapter *adapter, struct be_eth_hdr_wrb *hdr,
734 struct sk_buff *skb, u32 wrb_cnt, u32 len,
739 memset(hdr, 0, sizeof(*hdr));
741 AMAP_SET_BITS(struct amap_eth_hdr_wrb, crc, hdr, 1);
743 if (skb_is_gso(skb)) {
744 AMAP_SET_BITS(struct amap_eth_hdr_wrb, lso, hdr, 1);
745 AMAP_SET_BITS(struct amap_eth_hdr_wrb, lso_mss,
746 hdr, skb_shinfo(skb)->gso_size);
747 if (skb_is_gso_v6(skb) && !lancer_chip(adapter))
748 AMAP_SET_BITS(struct amap_eth_hdr_wrb, lso6, hdr, 1);
749 } else if (skb->ip_summed == CHECKSUM_PARTIAL) {
750 if (skb->encapsulation) {
751 AMAP_SET_BITS(struct amap_eth_hdr_wrb, ipcs, hdr, 1);
752 proto = skb_inner_ip_proto(skb);
754 proto = skb_ip_proto(skb);
756 if (proto == IPPROTO_TCP)
757 AMAP_SET_BITS(struct amap_eth_hdr_wrb, tcpcs, hdr, 1);
758 else if (proto == IPPROTO_UDP)
759 AMAP_SET_BITS(struct amap_eth_hdr_wrb, udpcs, hdr, 1);
762 if (vlan_tx_tag_present(skb)) {
763 AMAP_SET_BITS(struct amap_eth_hdr_wrb, vlan, hdr, 1);
764 vlan_tag = be_get_tx_vlan_tag(adapter, skb);
765 AMAP_SET_BITS(struct amap_eth_hdr_wrb, vlan_tag, hdr, vlan_tag);
768 /* To skip HW VLAN tagging: evt = 1, compl = 0 */
769 AMAP_SET_BITS(struct amap_eth_hdr_wrb, complete, hdr, !skip_hw_vlan);
770 AMAP_SET_BITS(struct amap_eth_hdr_wrb, event, hdr, 1);
771 AMAP_SET_BITS(struct amap_eth_hdr_wrb, num_wrb, hdr, wrb_cnt);
772 AMAP_SET_BITS(struct amap_eth_hdr_wrb, len, hdr, len);
775 static void unmap_tx_frag(struct device *dev, struct be_eth_wrb *wrb,
780 be_dws_le_to_cpu(wrb, sizeof(*wrb));
782 dma = (u64)wrb->frag_pa_hi << 32 | (u64)wrb->frag_pa_lo;
785 dma_unmap_single(dev, dma, wrb->frag_len,
788 dma_unmap_page(dev, dma, wrb->frag_len, DMA_TO_DEVICE);
792 static int make_tx_wrbs(struct be_adapter *adapter, struct be_queue_info *txq,
793 struct sk_buff *skb, u32 wrb_cnt, bool dummy_wrb,
798 struct device *dev = &adapter->pdev->dev;
799 struct sk_buff *first_skb = skb;
800 struct be_eth_wrb *wrb;
801 struct be_eth_hdr_wrb *hdr;
802 bool map_single = false;
805 hdr = queue_head_node(txq);
807 map_head = txq->head;
809 if (skb->len > skb->data_len) {
810 int len = skb_headlen(skb);
811 busaddr = dma_map_single(dev, skb->data, len, DMA_TO_DEVICE);
812 if (dma_mapping_error(dev, busaddr))
815 wrb = queue_head_node(txq);
816 wrb_fill(wrb, busaddr, len);
817 be_dws_cpu_to_le(wrb, sizeof(*wrb));
822 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
823 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i];
824 busaddr = skb_frag_dma_map(dev, frag, 0,
825 skb_frag_size(frag), DMA_TO_DEVICE);
826 if (dma_mapping_error(dev, busaddr))
828 wrb = queue_head_node(txq);
829 wrb_fill(wrb, busaddr, skb_frag_size(frag));
830 be_dws_cpu_to_le(wrb, sizeof(*wrb));
832 copied += skb_frag_size(frag);
836 wrb = queue_head_node(txq);
838 be_dws_cpu_to_le(wrb, sizeof(*wrb));
842 wrb_fill_hdr(adapter, hdr, first_skb, wrb_cnt, copied, skip_hw_vlan);
843 be_dws_cpu_to_le(hdr, sizeof(*hdr));
847 txq->head = map_head;
849 wrb = queue_head_node(txq);
850 unmap_tx_frag(dev, wrb, map_single);
852 copied -= wrb->frag_len;
858 static struct sk_buff *be_insert_vlan_in_pkt(struct be_adapter *adapter,
864 skb = skb_share_check(skb, GFP_ATOMIC);
868 if (vlan_tx_tag_present(skb))
869 vlan_tag = be_get_tx_vlan_tag(adapter, skb);
871 if (qnq_async_evt_rcvd(adapter) && adapter->pvid) {
873 vlan_tag = adapter->pvid;
874 /* f/w workaround to set skip_hw_vlan = 1, informs the F/W to
875 * skip VLAN insertion
878 *skip_hw_vlan = true;
882 skb = __vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_tag);
888 /* Insert the outer VLAN, if any */
889 if (adapter->qnq_vid) {
890 vlan_tag = adapter->qnq_vid;
891 skb = __vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_tag);
895 *skip_hw_vlan = true;
901 static bool be_ipv6_exthdr_check(struct sk_buff *skb)
903 struct ethhdr *eh = (struct ethhdr *)skb->data;
904 u16 offset = ETH_HLEN;
906 if (eh->h_proto == htons(ETH_P_IPV6)) {
907 struct ipv6hdr *ip6h = (struct ipv6hdr *)(skb->data + offset);
909 offset += sizeof(struct ipv6hdr);
910 if (ip6h->nexthdr != NEXTHDR_TCP &&
911 ip6h->nexthdr != NEXTHDR_UDP) {
912 struct ipv6_opt_hdr *ehdr =
913 (struct ipv6_opt_hdr *) (skb->data + offset);
915 /* offending pkt: 2nd byte following IPv6 hdr is 0xff */
916 if (ehdr->hdrlen == 0xff)
923 static int be_vlan_tag_tx_chk(struct be_adapter *adapter, struct sk_buff *skb)
925 return vlan_tx_tag_present(skb) || adapter->pvid || adapter->qnq_vid;
928 static int be_ipv6_tx_stall_chk(struct be_adapter *adapter, struct sk_buff *skb)
930 return BE3_chip(adapter) && be_ipv6_exthdr_check(skb);
933 static struct sk_buff *be_lancer_xmit_workarounds(struct be_adapter *adapter,
937 struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
938 unsigned int eth_hdr_len;
941 /* For padded packets, BE HW modifies tot_len field in IP header
942 * incorrecly when VLAN tag is inserted by HW.
943 * For padded packets, Lancer computes incorrect checksum.
945 eth_hdr_len = ntohs(skb->protocol) == ETH_P_8021Q ?
946 VLAN_ETH_HLEN : ETH_HLEN;
947 if (skb->len <= 60 &&
948 (lancer_chip(adapter) || vlan_tx_tag_present(skb)) &&
950 ip = (struct iphdr *)ip_hdr(skb);
951 pskb_trim(skb, eth_hdr_len + ntohs(ip->tot_len));
954 /* If vlan tag is already inlined in the packet, skip HW VLAN
955 * tagging in pvid-tagging mode
957 if (be_pvid_tagging_enabled(adapter) &&
958 veh->h_vlan_proto == htons(ETH_P_8021Q))
959 *skip_hw_vlan = true;
961 /* HW has a bug wherein it will calculate CSUM for VLAN
962 * pkts even though it is disabled.
963 * Manually insert VLAN in pkt.
965 if (skb->ip_summed != CHECKSUM_PARTIAL &&
966 vlan_tx_tag_present(skb)) {
967 skb = be_insert_vlan_in_pkt(adapter, skb, skip_hw_vlan);
972 /* HW may lockup when VLAN HW tagging is requested on
973 * certain ipv6 packets. Drop such pkts if the HW workaround to
974 * skip HW tagging is not enabled by FW.
976 if (unlikely(be_ipv6_tx_stall_chk(adapter, skb) &&
977 (adapter->pvid || adapter->qnq_vid) &&
978 !qnq_async_evt_rcvd(adapter)))
981 /* Manual VLAN tag insertion to prevent:
982 * ASIC lockup when the ASIC inserts VLAN tag into
983 * certain ipv6 packets. Insert VLAN tags in driver,
984 * and set event, completion, vlan bits accordingly
987 if (be_ipv6_tx_stall_chk(adapter, skb) &&
988 be_vlan_tag_tx_chk(adapter, skb)) {
989 skb = be_insert_vlan_in_pkt(adapter, skb, skip_hw_vlan);
996 dev_kfree_skb_any(skb);
1001 static struct sk_buff *be_xmit_workarounds(struct be_adapter *adapter,
1002 struct sk_buff *skb,
1005 /* Lancer, SH-R ASICs have a bug wherein Packets that are 32 bytes or
1006 * less may cause a transmit stall on that port. So the work-around is
1007 * to pad short packets (<= 32 bytes) to a 36-byte length.
1009 if (unlikely(!BEx_chip(adapter) && skb->len <= 32)) {
1010 if (skb_padto(skb, 36))
1015 if (BEx_chip(adapter) || lancer_chip(adapter)) {
1016 skb = be_lancer_xmit_workarounds(adapter, skb, skip_hw_vlan);
1024 static netdev_tx_t be_xmit(struct sk_buff *skb, struct net_device *netdev)
1026 struct be_adapter *adapter = netdev_priv(netdev);
1027 struct be_tx_obj *txo = &adapter->tx_obj[skb_get_queue_mapping(skb)];
1028 struct be_queue_info *txq = &txo->q;
1029 bool dummy_wrb, stopped = false;
1030 u32 wrb_cnt = 0, copied = 0;
1031 bool skip_hw_vlan = false;
1032 u32 start = txq->head;
1034 skb = be_xmit_workarounds(adapter, skb, &skip_hw_vlan);
1036 tx_stats(txo)->tx_drv_drops++;
1037 return NETDEV_TX_OK;
1040 wrb_cnt = wrb_cnt_for_skb(adapter, skb, &dummy_wrb);
1042 copied = make_tx_wrbs(adapter, txq, skb, wrb_cnt, dummy_wrb,
1045 int gso_segs = skb_shinfo(skb)->gso_segs;
1047 /* record the sent skb in the sent_skb table */
1048 BUG_ON(txo->sent_skb_list[start]);
1049 txo->sent_skb_list[start] = skb;
1051 /* Ensure txq has space for the next skb; Else stop the queue
1052 * *BEFORE* ringing the tx doorbell, so that we serialze the
1053 * tx compls of the current transmit which'll wake up the queue
1055 atomic_add(wrb_cnt, &txq->used);
1056 if ((BE_MAX_TX_FRAG_COUNT + atomic_read(&txq->used)) >=
1058 netif_stop_subqueue(netdev, skb_get_queue_mapping(skb));
1062 be_txq_notify(adapter, txo, wrb_cnt);
1064 be_tx_stats_update(txo, wrb_cnt, copied, gso_segs, stopped);
1067 tx_stats(txo)->tx_drv_drops++;
1068 dev_kfree_skb_any(skb);
1070 return NETDEV_TX_OK;
1073 static int be_change_mtu(struct net_device *netdev, int new_mtu)
1075 struct be_adapter *adapter = netdev_priv(netdev);
1076 if (new_mtu < BE_MIN_MTU ||
1077 new_mtu > (BE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN))) {
1078 dev_info(&adapter->pdev->dev,
1079 "MTU must be between %d and %d bytes\n",
1081 (BE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN)));
1084 dev_info(&adapter->pdev->dev, "MTU changed from %d to %d bytes\n",
1085 netdev->mtu, new_mtu);
1086 netdev->mtu = new_mtu;
1091 * A max of 64 (BE_NUM_VLANS_SUPPORTED) vlans can be configured in BE.
1092 * If the user configures more, place BE in vlan promiscuous mode.
1094 static int be_vid_config(struct be_adapter *adapter)
1096 u16 vids[BE_NUM_VLANS_SUPPORTED];
1100 /* No need to further configure vids if in promiscuous mode */
1101 if (adapter->promiscuous)
1104 if (adapter->vlans_added > be_max_vlans(adapter))
1105 goto set_vlan_promisc;
1107 /* Construct VLAN Table to give to HW */
1108 for_each_set_bit(i, adapter->vids, VLAN_N_VID)
1109 vids[num++] = cpu_to_le16(i);
1111 status = be_cmd_vlan_config(adapter, adapter->if_handle, vids, num);
1113 /* Set to VLAN promisc mode as setting VLAN filter failed */
1114 if (addl_status(status) ==
1115 MCC_ADDL_STATUS_INSUFFICIENT_RESOURCES)
1116 goto set_vlan_promisc;
1117 dev_err(&adapter->pdev->dev,
1118 "Setting HW VLAN filtering failed.\n");
1120 if (adapter->flags & BE_FLAGS_VLAN_PROMISC) {
1121 /* hw VLAN filtering re-enabled. */
1122 status = be_cmd_rx_filter(adapter,
1123 BE_FLAGS_VLAN_PROMISC, OFF);
1125 dev_info(&adapter->pdev->dev,
1126 "Disabling VLAN Promiscuous mode.\n");
1127 adapter->flags &= ~BE_FLAGS_VLAN_PROMISC;
1135 if (adapter->flags & BE_FLAGS_VLAN_PROMISC)
1138 status = be_cmd_rx_filter(adapter, BE_FLAGS_VLAN_PROMISC, ON);
1140 dev_info(&adapter->pdev->dev, "Enable VLAN Promiscuous mode\n");
1141 adapter->flags |= BE_FLAGS_VLAN_PROMISC;
1143 dev_err(&adapter->pdev->dev,
1144 "Failed to enable VLAN Promiscuous mode.\n");
1148 static int be_vlan_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
1150 struct be_adapter *adapter = netdev_priv(netdev);
1153 /* Packets with VID 0 are always received by Lancer by default */
1154 if (lancer_chip(adapter) && vid == 0)
1157 if (test_bit(vid, adapter->vids))
1160 set_bit(vid, adapter->vids);
1161 adapter->vlans_added++;
1163 status = be_vid_config(adapter);
1165 adapter->vlans_added--;
1166 clear_bit(vid, adapter->vids);
1172 static int be_vlan_rem_vid(struct net_device *netdev, __be16 proto, u16 vid)
1174 struct be_adapter *adapter = netdev_priv(netdev);
1176 /* Packets with VID 0 are always received by Lancer by default */
1177 if (lancer_chip(adapter) && vid == 0)
1180 clear_bit(vid, adapter->vids);
1181 adapter->vlans_added--;
1183 return be_vid_config(adapter);
1186 static void be_clear_promisc(struct be_adapter *adapter)
1188 adapter->promiscuous = false;
1189 adapter->flags &= ~(BE_FLAGS_VLAN_PROMISC | BE_FLAGS_MCAST_PROMISC);
1191 be_cmd_rx_filter(adapter, IFF_PROMISC, OFF);
1194 static void be_set_rx_mode(struct net_device *netdev)
1196 struct be_adapter *adapter = netdev_priv(netdev);
1199 if (netdev->flags & IFF_PROMISC) {
1200 be_cmd_rx_filter(adapter, IFF_PROMISC, ON);
1201 adapter->promiscuous = true;
1205 /* BE was previously in promiscuous mode; disable it */
1206 if (adapter->promiscuous) {
1207 be_clear_promisc(adapter);
1208 if (adapter->vlans_added)
1209 be_vid_config(adapter);
1212 /* Enable multicast promisc if num configured exceeds what we support */
1213 if (netdev->flags & IFF_ALLMULTI ||
1214 netdev_mc_count(netdev) > be_max_mc(adapter))
1215 goto set_mcast_promisc;
1217 if (netdev_uc_count(netdev) != adapter->uc_macs) {
1218 struct netdev_hw_addr *ha;
1219 int i = 1; /* First slot is claimed by the Primary MAC */
1221 for (; adapter->uc_macs > 0; adapter->uc_macs--, i++) {
1222 be_cmd_pmac_del(adapter, adapter->if_handle,
1223 adapter->pmac_id[i], 0);
1226 if (netdev_uc_count(netdev) > be_max_uc(adapter)) {
1227 be_cmd_rx_filter(adapter, IFF_PROMISC, ON);
1228 adapter->promiscuous = true;
1232 netdev_for_each_uc_addr(ha, adapter->netdev) {
1233 adapter->uc_macs++; /* First slot is for Primary MAC */
1234 be_cmd_pmac_add(adapter, (u8 *)ha->addr,
1236 &adapter->pmac_id[adapter->uc_macs], 0);
1240 status = be_cmd_rx_filter(adapter, IFF_MULTICAST, ON);
1242 if (adapter->flags & BE_FLAGS_MCAST_PROMISC)
1243 adapter->flags &= ~BE_FLAGS_MCAST_PROMISC;
1248 if (adapter->flags & BE_FLAGS_MCAST_PROMISC)
1251 /* Set to MCAST promisc mode if setting MULTICAST address fails
1252 * or if num configured exceeds what we support
1254 status = be_cmd_rx_filter(adapter, IFF_ALLMULTI, ON);
1256 adapter->flags |= BE_FLAGS_MCAST_PROMISC;
1261 static int be_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
1263 struct be_adapter *adapter = netdev_priv(netdev);
1264 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1267 if (!sriov_enabled(adapter))
1270 if (!is_valid_ether_addr(mac) || vf >= adapter->num_vfs)
1273 if (BEx_chip(adapter)) {
1274 be_cmd_pmac_del(adapter, vf_cfg->if_handle, vf_cfg->pmac_id,
1277 status = be_cmd_pmac_add(adapter, mac, vf_cfg->if_handle,
1278 &vf_cfg->pmac_id, vf + 1);
1280 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
1285 dev_err(&adapter->pdev->dev, "MAC %pM set on VF %d Failed: %#x",
1287 return be_cmd_status(status);
1290 ether_addr_copy(vf_cfg->mac_addr, mac);
1295 static int be_get_vf_config(struct net_device *netdev, int vf,
1296 struct ifla_vf_info *vi)
1298 struct be_adapter *adapter = netdev_priv(netdev);
1299 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1301 if (!sriov_enabled(adapter))
1304 if (vf >= adapter->num_vfs)
1308 vi->max_tx_rate = vf_cfg->tx_rate;
1309 vi->min_tx_rate = 0;
1310 vi->vlan = vf_cfg->vlan_tag & VLAN_VID_MASK;
1311 vi->qos = vf_cfg->vlan_tag >> VLAN_PRIO_SHIFT;
1312 memcpy(&vi->mac, vf_cfg->mac_addr, ETH_ALEN);
1313 vi->linkstate = adapter->vf_cfg[vf].plink_tracking;
1318 static int be_set_vf_vlan(struct net_device *netdev, int vf, u16 vlan, u8 qos)
1320 struct be_adapter *adapter = netdev_priv(netdev);
1321 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1324 if (!sriov_enabled(adapter))
1327 if (vf >= adapter->num_vfs || vlan > 4095 || qos > 7)
1331 vlan |= qos << VLAN_PRIO_SHIFT;
1332 if (vf_cfg->vlan_tag != vlan)
1333 status = be_cmd_set_hsw_config(adapter, vlan, vf + 1,
1334 vf_cfg->if_handle, 0);
1336 /* Reset Transparent Vlan Tagging. */
1337 status = be_cmd_set_hsw_config(adapter, BE_RESET_VLAN_TAG_ID,
1338 vf + 1, vf_cfg->if_handle, 0);
1342 dev_err(&adapter->pdev->dev,
1343 "VLAN %d config on VF %d failed : %#x\n", vlan,
1345 return be_cmd_status(status);
1348 vf_cfg->vlan_tag = vlan;
1353 static int be_set_vf_tx_rate(struct net_device *netdev, int vf,
1354 int min_tx_rate, int max_tx_rate)
1356 struct be_adapter *adapter = netdev_priv(netdev);
1357 struct device *dev = &adapter->pdev->dev;
1358 int percent_rate, status = 0;
1362 if (!sriov_enabled(adapter))
1365 if (vf >= adapter->num_vfs)
1374 status = be_cmd_link_status_query(adapter, &link_speed,
1380 dev_err(dev, "TX-rate setting not allowed when link is down\n");
1385 if (max_tx_rate < 100 || max_tx_rate > link_speed) {
1386 dev_err(dev, "TX-rate must be between 100 and %d Mbps\n",
1392 /* On Skyhawk the QOS setting must be done only as a % value */
1393 percent_rate = link_speed / 100;
1394 if (skyhawk_chip(adapter) && (max_tx_rate % percent_rate)) {
1395 dev_err(dev, "TX-rate must be a multiple of %d Mbps\n",
1402 status = be_cmd_config_qos(adapter, max_tx_rate, link_speed, vf + 1);
1406 adapter->vf_cfg[vf].tx_rate = max_tx_rate;
1410 dev_err(dev, "TX-rate setting of %dMbps on VF%d failed\n",
1412 return be_cmd_status(status);
1414 static int be_set_vf_link_state(struct net_device *netdev, int vf,
1417 struct be_adapter *adapter = netdev_priv(netdev);
1420 if (!sriov_enabled(adapter))
1423 if (vf >= adapter->num_vfs)
1426 status = be_cmd_set_logical_link_config(adapter, link_state, vf+1);
1428 dev_err(&adapter->pdev->dev,
1429 "Link state change on VF %d failed: %#x\n", vf, status);
1430 return be_cmd_status(status);
1433 adapter->vf_cfg[vf].plink_tracking = link_state;
1438 static void be_aic_update(struct be_aic_obj *aic, u64 rx_pkts, u64 tx_pkts,
1441 aic->rx_pkts_prev = rx_pkts;
1442 aic->tx_reqs_prev = tx_pkts;
1446 static void be_eqd_update(struct be_adapter *adapter)
1448 struct be_set_eqd set_eqd[MAX_EVT_QS];
1449 int eqd, i, num = 0, start;
1450 struct be_aic_obj *aic;
1451 struct be_eq_obj *eqo;
1452 struct be_rx_obj *rxo;
1453 struct be_tx_obj *txo;
1454 u64 rx_pkts, tx_pkts;
1458 for_all_evt_queues(adapter, eqo, i) {
1459 aic = &adapter->aic_obj[eqo->idx];
1467 rxo = &adapter->rx_obj[eqo->idx];
1469 start = u64_stats_fetch_begin_irq(&rxo->stats.sync);
1470 rx_pkts = rxo->stats.rx_pkts;
1471 } while (u64_stats_fetch_retry_irq(&rxo->stats.sync, start));
1473 txo = &adapter->tx_obj[eqo->idx];
1475 start = u64_stats_fetch_begin_irq(&txo->stats.sync);
1476 tx_pkts = txo->stats.tx_reqs;
1477 } while (u64_stats_fetch_retry_irq(&txo->stats.sync, start));
1480 /* Skip, if wrapped around or first calculation */
1482 if (!aic->jiffies || time_before(now, aic->jiffies) ||
1483 rx_pkts < aic->rx_pkts_prev ||
1484 tx_pkts < aic->tx_reqs_prev) {
1485 be_aic_update(aic, rx_pkts, tx_pkts, now);
1489 delta = jiffies_to_msecs(now - aic->jiffies);
1490 pps = (((u32)(rx_pkts - aic->rx_pkts_prev) * 1000) / delta) +
1491 (((u32)(tx_pkts - aic->tx_reqs_prev) * 1000) / delta);
1492 eqd = (pps / 15000) << 2;
1496 eqd = min_t(u32, eqd, aic->max_eqd);
1497 eqd = max_t(u32, eqd, aic->min_eqd);
1499 be_aic_update(aic, rx_pkts, tx_pkts, now);
1501 if (eqd != aic->prev_eqd) {
1502 set_eqd[num].delay_multiplier = (eqd * 65)/100;
1503 set_eqd[num].eq_id = eqo->q.id;
1504 aic->prev_eqd = eqd;
1510 be_cmd_modify_eqd(adapter, set_eqd, num);
1513 static void be_rx_stats_update(struct be_rx_obj *rxo,
1514 struct be_rx_compl_info *rxcp)
1516 struct be_rx_stats *stats = rx_stats(rxo);
1518 u64_stats_update_begin(&stats->sync);
1520 stats->rx_bytes += rxcp->pkt_size;
1522 if (rxcp->pkt_type == BE_MULTICAST_PACKET)
1523 stats->rx_mcast_pkts++;
1525 stats->rx_compl_err++;
1526 u64_stats_update_end(&stats->sync);
1529 static inline bool csum_passed(struct be_rx_compl_info *rxcp)
1531 /* L4 checksum is not reliable for non TCP/UDP packets.
1532 * Also ignore ipcksm for ipv6 pkts
1534 return (rxcp->tcpf || rxcp->udpf) && rxcp->l4_csum &&
1535 (rxcp->ip_csum || rxcp->ipv6) && !rxcp->err;
1538 static struct be_rx_page_info *get_rx_page_info(struct be_rx_obj *rxo)
1540 struct be_adapter *adapter = rxo->adapter;
1541 struct be_rx_page_info *rx_page_info;
1542 struct be_queue_info *rxq = &rxo->q;
1543 u16 frag_idx = rxq->tail;
1545 rx_page_info = &rxo->page_info_tbl[frag_idx];
1546 BUG_ON(!rx_page_info->page);
1548 if (rx_page_info->last_frag) {
1549 dma_unmap_page(&adapter->pdev->dev,
1550 dma_unmap_addr(rx_page_info, bus),
1551 adapter->big_page_size, DMA_FROM_DEVICE);
1552 rx_page_info->last_frag = false;
1554 dma_sync_single_for_cpu(&adapter->pdev->dev,
1555 dma_unmap_addr(rx_page_info, bus),
1556 rx_frag_size, DMA_FROM_DEVICE);
1559 queue_tail_inc(rxq);
1560 atomic_dec(&rxq->used);
1561 return rx_page_info;
1564 /* Throwaway the data in the Rx completion */
1565 static void be_rx_compl_discard(struct be_rx_obj *rxo,
1566 struct be_rx_compl_info *rxcp)
1568 struct be_rx_page_info *page_info;
1569 u16 i, num_rcvd = rxcp->num_rcvd;
1571 for (i = 0; i < num_rcvd; i++) {
1572 page_info = get_rx_page_info(rxo);
1573 put_page(page_info->page);
1574 memset(page_info, 0, sizeof(*page_info));
1579 * skb_fill_rx_data forms a complete skb for an ether frame
1580 * indicated by rxcp.
1582 static void skb_fill_rx_data(struct be_rx_obj *rxo, struct sk_buff *skb,
1583 struct be_rx_compl_info *rxcp)
1585 struct be_rx_page_info *page_info;
1587 u16 hdr_len, curr_frag_len, remaining;
1590 page_info = get_rx_page_info(rxo);
1591 start = page_address(page_info->page) + page_info->page_offset;
1594 /* Copy data in the first descriptor of this completion */
1595 curr_frag_len = min(rxcp->pkt_size, rx_frag_size);
1597 skb->len = curr_frag_len;
1598 if (curr_frag_len <= BE_HDR_LEN) { /* tiny packet */
1599 memcpy(skb->data, start, curr_frag_len);
1600 /* Complete packet has now been moved to data */
1601 put_page(page_info->page);
1603 skb->tail += curr_frag_len;
1606 memcpy(skb->data, start, hdr_len);
1607 skb_shinfo(skb)->nr_frags = 1;
1608 skb_frag_set_page(skb, 0, page_info->page);
1609 skb_shinfo(skb)->frags[0].page_offset =
1610 page_info->page_offset + hdr_len;
1611 skb_frag_size_set(&skb_shinfo(skb)->frags[0],
1612 curr_frag_len - hdr_len);
1613 skb->data_len = curr_frag_len - hdr_len;
1614 skb->truesize += rx_frag_size;
1615 skb->tail += hdr_len;
1617 page_info->page = NULL;
1619 if (rxcp->pkt_size <= rx_frag_size) {
1620 BUG_ON(rxcp->num_rcvd != 1);
1624 /* More frags present for this completion */
1625 remaining = rxcp->pkt_size - curr_frag_len;
1626 for (i = 1, j = 0; i < rxcp->num_rcvd; i++) {
1627 page_info = get_rx_page_info(rxo);
1628 curr_frag_len = min(remaining, rx_frag_size);
1630 /* Coalesce all frags from the same physical page in one slot */
1631 if (page_info->page_offset == 0) {
1634 skb_frag_set_page(skb, j, page_info->page);
1635 skb_shinfo(skb)->frags[j].page_offset =
1636 page_info->page_offset;
1637 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1638 skb_shinfo(skb)->nr_frags++;
1640 put_page(page_info->page);
1643 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1644 skb->len += curr_frag_len;
1645 skb->data_len += curr_frag_len;
1646 skb->truesize += rx_frag_size;
1647 remaining -= curr_frag_len;
1648 page_info->page = NULL;
1650 BUG_ON(j > MAX_SKB_FRAGS);
1653 /* Process the RX completion indicated by rxcp when GRO is disabled */
1654 static void be_rx_compl_process(struct be_rx_obj *rxo, struct napi_struct *napi,
1655 struct be_rx_compl_info *rxcp)
1657 struct be_adapter *adapter = rxo->adapter;
1658 struct net_device *netdev = adapter->netdev;
1659 struct sk_buff *skb;
1661 skb = netdev_alloc_skb_ip_align(netdev, BE_RX_SKB_ALLOC_SIZE);
1662 if (unlikely(!skb)) {
1663 rx_stats(rxo)->rx_drops_no_skbs++;
1664 be_rx_compl_discard(rxo, rxcp);
1668 skb_fill_rx_data(rxo, skb, rxcp);
1670 if (likely((netdev->features & NETIF_F_RXCSUM) && csum_passed(rxcp)))
1671 skb->ip_summed = CHECKSUM_UNNECESSARY;
1673 skb_checksum_none_assert(skb);
1675 skb->protocol = eth_type_trans(skb, netdev);
1676 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1677 if (netdev->features & NETIF_F_RXHASH)
1678 skb_set_hash(skb, rxcp->rss_hash, PKT_HASH_TYPE_L3);
1680 skb->encapsulation = rxcp->tunneled;
1681 skb_mark_napi_id(skb, napi);
1684 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1686 netif_receive_skb(skb);
1689 /* Process the RX completion indicated by rxcp when GRO is enabled */
1690 static void be_rx_compl_process_gro(struct be_rx_obj *rxo,
1691 struct napi_struct *napi,
1692 struct be_rx_compl_info *rxcp)
1694 struct be_adapter *adapter = rxo->adapter;
1695 struct be_rx_page_info *page_info;
1696 struct sk_buff *skb = NULL;
1697 u16 remaining, curr_frag_len;
1700 skb = napi_get_frags(napi);
1702 be_rx_compl_discard(rxo, rxcp);
1706 remaining = rxcp->pkt_size;
1707 for (i = 0, j = -1; i < rxcp->num_rcvd; i++) {
1708 page_info = get_rx_page_info(rxo);
1710 curr_frag_len = min(remaining, rx_frag_size);
1712 /* Coalesce all frags from the same physical page in one slot */
1713 if (i == 0 || page_info->page_offset == 0) {
1714 /* First frag or Fresh page */
1716 skb_frag_set_page(skb, j, page_info->page);
1717 skb_shinfo(skb)->frags[j].page_offset =
1718 page_info->page_offset;
1719 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1721 put_page(page_info->page);
1723 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1724 skb->truesize += rx_frag_size;
1725 remaining -= curr_frag_len;
1726 memset(page_info, 0, sizeof(*page_info));
1728 BUG_ON(j > MAX_SKB_FRAGS);
1730 skb_shinfo(skb)->nr_frags = j + 1;
1731 skb->len = rxcp->pkt_size;
1732 skb->data_len = rxcp->pkt_size;
1733 skb->ip_summed = CHECKSUM_UNNECESSARY;
1734 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1735 if (adapter->netdev->features & NETIF_F_RXHASH)
1736 skb_set_hash(skb, rxcp->rss_hash, PKT_HASH_TYPE_L3);
1738 skb->encapsulation = rxcp->tunneled;
1739 skb_mark_napi_id(skb, napi);
1742 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1744 napi_gro_frags(napi);
1747 static void be_parse_rx_compl_v1(struct be_eth_rx_compl *compl,
1748 struct be_rx_compl_info *rxcp)
1751 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, pktsize, compl);
1752 rxcp->vlanf = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, vtp, compl);
1753 rxcp->err = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, err, compl);
1754 rxcp->tcpf = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, tcpf, compl);
1755 rxcp->udpf = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, udpf, compl);
1757 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, ipcksm, compl);
1759 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, l4_cksm, compl);
1761 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, ip_version, compl);
1763 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, numfrags, compl);
1765 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, cast_enc, compl);
1767 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, rsshash, compl);
1769 rxcp->qnq = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, qnq,
1771 rxcp->vlan_tag = AMAP_GET_BITS(struct amap_eth_rx_compl_v1,
1774 rxcp->port = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, port, compl);
1776 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, tunneled, compl);
1779 static void be_parse_rx_compl_v0(struct be_eth_rx_compl *compl,
1780 struct be_rx_compl_info *rxcp)
1783 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, pktsize, compl);
1784 rxcp->vlanf = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, vtp, compl);
1785 rxcp->err = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, err, compl);
1786 rxcp->tcpf = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, tcpf, compl);
1787 rxcp->udpf = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, udpf, compl);
1789 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, ipcksm, compl);
1791 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, l4_cksm, compl);
1793 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, ip_version, compl);
1795 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, numfrags, compl);
1797 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, cast_enc, compl);
1799 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, rsshash, compl);
1801 rxcp->qnq = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, qnq,
1803 rxcp->vlan_tag = AMAP_GET_BITS(struct amap_eth_rx_compl_v0,
1806 rxcp->port = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, port, compl);
1807 rxcp->ip_frag = AMAP_GET_BITS(struct amap_eth_rx_compl_v0,
1811 static struct be_rx_compl_info *be_rx_compl_get(struct be_rx_obj *rxo)
1813 struct be_eth_rx_compl *compl = queue_tail_node(&rxo->cq);
1814 struct be_rx_compl_info *rxcp = &rxo->rxcp;
1815 struct be_adapter *adapter = rxo->adapter;
1817 /* For checking the valid bit it is Ok to use either definition as the
1818 * valid bit is at the same position in both v0 and v1 Rx compl */
1819 if (compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] == 0)
1823 be_dws_le_to_cpu(compl, sizeof(*compl));
1825 if (adapter->be3_native)
1826 be_parse_rx_compl_v1(compl, rxcp);
1828 be_parse_rx_compl_v0(compl, rxcp);
1834 /* In QNQ modes, if qnq bit is not set, then the packet was
1835 * tagged only with the transparent outer vlan-tag and must
1836 * not be treated as a vlan packet by host
1838 if (be_is_qnq_mode(adapter) && !rxcp->qnq)
1841 if (!lancer_chip(adapter))
1842 rxcp->vlan_tag = swab16(rxcp->vlan_tag);
1844 if (adapter->pvid == (rxcp->vlan_tag & VLAN_VID_MASK) &&
1845 !test_bit(rxcp->vlan_tag, adapter->vids))
1849 /* As the compl has been parsed, reset it; we wont touch it again */
1850 compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] = 0;
1852 queue_tail_inc(&rxo->cq);
1856 static inline struct page *be_alloc_pages(u32 size, gfp_t gfp)
1858 u32 order = get_order(size);
1862 return alloc_pages(gfp, order);
1866 * Allocate a page, split it to fragments of size rx_frag_size and post as
1867 * receive buffers to BE
1869 static void be_post_rx_frags(struct be_rx_obj *rxo, gfp_t gfp)
1871 struct be_adapter *adapter = rxo->adapter;
1872 struct be_rx_page_info *page_info = NULL, *prev_page_info = NULL;
1873 struct be_queue_info *rxq = &rxo->q;
1874 struct page *pagep = NULL;
1875 struct device *dev = &adapter->pdev->dev;
1876 struct be_eth_rx_d *rxd;
1877 u64 page_dmaaddr = 0, frag_dmaaddr;
1878 u32 posted, page_offset = 0;
1880 page_info = &rxo->page_info_tbl[rxq->head];
1881 for (posted = 0; posted < MAX_RX_POST && !page_info->page; posted++) {
1883 pagep = be_alloc_pages(adapter->big_page_size, gfp);
1884 if (unlikely(!pagep)) {
1885 rx_stats(rxo)->rx_post_fail++;
1888 page_dmaaddr = dma_map_page(dev, pagep, 0,
1889 adapter->big_page_size,
1891 if (dma_mapping_error(dev, page_dmaaddr)) {
1894 rx_stats(rxo)->rx_post_fail++;
1900 page_offset += rx_frag_size;
1902 page_info->page_offset = page_offset;
1903 page_info->page = pagep;
1905 rxd = queue_head_node(rxq);
1906 frag_dmaaddr = page_dmaaddr + page_info->page_offset;
1907 rxd->fragpa_lo = cpu_to_le32(frag_dmaaddr & 0xFFFFFFFF);
1908 rxd->fragpa_hi = cpu_to_le32(upper_32_bits(frag_dmaaddr));
1910 /* Any space left in the current big page for another frag? */
1911 if ((page_offset + rx_frag_size + rx_frag_size) >
1912 adapter->big_page_size) {
1914 page_info->last_frag = true;
1915 dma_unmap_addr_set(page_info, bus, page_dmaaddr);
1917 dma_unmap_addr_set(page_info, bus, frag_dmaaddr);
1920 prev_page_info = page_info;
1921 queue_head_inc(rxq);
1922 page_info = &rxo->page_info_tbl[rxq->head];
1925 /* Mark the last frag of a page when we break out of the above loop
1926 * with no more slots available in the RXQ
1929 prev_page_info->last_frag = true;
1930 dma_unmap_addr_set(prev_page_info, bus, page_dmaaddr);
1934 atomic_add(posted, &rxq->used);
1935 if (rxo->rx_post_starved)
1936 rxo->rx_post_starved = false;
1937 be_rxq_notify(adapter, rxq->id, posted);
1938 } else if (atomic_read(&rxq->used) == 0) {
1939 /* Let be_worker replenish when memory is available */
1940 rxo->rx_post_starved = true;
1944 static struct be_eth_tx_compl *be_tx_compl_get(struct be_queue_info *tx_cq)
1946 struct be_eth_tx_compl *txcp = queue_tail_node(tx_cq);
1948 if (txcp->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] == 0)
1952 be_dws_le_to_cpu(txcp, sizeof(*txcp));
1954 txcp->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] = 0;
1956 queue_tail_inc(tx_cq);
1960 static u16 be_tx_compl_process(struct be_adapter *adapter,
1961 struct be_tx_obj *txo, u16 last_index)
1963 struct be_queue_info *txq = &txo->q;
1964 struct be_eth_wrb *wrb;
1965 struct sk_buff **sent_skbs = txo->sent_skb_list;
1966 struct sk_buff *sent_skb;
1967 u16 cur_index, num_wrbs = 1; /* account for hdr wrb */
1968 bool unmap_skb_hdr = true;
1970 sent_skb = sent_skbs[txq->tail];
1972 sent_skbs[txq->tail] = NULL;
1974 /* skip header wrb */
1975 queue_tail_inc(txq);
1978 cur_index = txq->tail;
1979 wrb = queue_tail_node(txq);
1980 unmap_tx_frag(&adapter->pdev->dev, wrb,
1981 (unmap_skb_hdr && skb_headlen(sent_skb)));
1982 unmap_skb_hdr = false;
1985 queue_tail_inc(txq);
1986 } while (cur_index != last_index);
1988 dev_kfree_skb_any(sent_skb);
1992 /* Return the number of events in the event queue */
1993 static inline int events_get(struct be_eq_obj *eqo)
1995 struct be_eq_entry *eqe;
1999 eqe = queue_tail_node(&eqo->q);
2006 queue_tail_inc(&eqo->q);
2012 /* Leaves the EQ is disarmed state */
2013 static void be_eq_clean(struct be_eq_obj *eqo)
2015 int num = events_get(eqo);
2017 be_eq_notify(eqo->adapter, eqo->q.id, false, true, num);
2020 static void be_rx_cq_clean(struct be_rx_obj *rxo)
2022 struct be_rx_page_info *page_info;
2023 struct be_queue_info *rxq = &rxo->q;
2024 struct be_queue_info *rx_cq = &rxo->cq;
2025 struct be_rx_compl_info *rxcp;
2026 struct be_adapter *adapter = rxo->adapter;
2029 /* Consume pending rx completions.
2030 * Wait for the flush completion (identified by zero num_rcvd)
2031 * to arrive. Notify CQ even when there are no more CQ entries
2032 * for HW to flush partially coalesced CQ entries.
2033 * In Lancer, there is no need to wait for flush compl.
2036 rxcp = be_rx_compl_get(rxo);
2038 if (lancer_chip(adapter))
2041 if (flush_wait++ > 10 || be_hw_error(adapter)) {
2042 dev_warn(&adapter->pdev->dev,
2043 "did not receive flush compl\n");
2046 be_cq_notify(adapter, rx_cq->id, true, 0);
2049 be_rx_compl_discard(rxo, rxcp);
2050 be_cq_notify(adapter, rx_cq->id, false, 1);
2051 if (rxcp->num_rcvd == 0)
2056 /* After cleanup, leave the CQ in unarmed state */
2057 be_cq_notify(adapter, rx_cq->id, false, 0);
2059 /* Then free posted rx buffers that were not used */
2060 while (atomic_read(&rxq->used) > 0) {
2061 page_info = get_rx_page_info(rxo);
2062 put_page(page_info->page);
2063 memset(page_info, 0, sizeof(*page_info));
2065 BUG_ON(atomic_read(&rxq->used));
2066 rxq->tail = rxq->head = 0;
2069 static void be_tx_compl_clean(struct be_adapter *adapter)
2071 struct be_tx_obj *txo;
2072 struct be_queue_info *txq;
2073 struct be_eth_tx_compl *txcp;
2074 u16 end_idx, cmpl = 0, timeo = 0, num_wrbs = 0;
2075 struct sk_buff *sent_skb;
2077 int i, pending_txqs;
2079 /* Stop polling for compls when HW has been silent for 10ms */
2081 pending_txqs = adapter->num_tx_qs;
2083 for_all_tx_queues(adapter, txo, i) {
2087 while ((txcp = be_tx_compl_get(&txo->cq))) {
2089 AMAP_GET_BITS(struct amap_eth_tx_compl,
2091 num_wrbs += be_tx_compl_process(adapter, txo,
2096 be_cq_notify(adapter, txo->cq.id, false, cmpl);
2097 atomic_sub(num_wrbs, &txq->used);
2100 if (atomic_read(&txq->used) == 0)
2104 if (pending_txqs == 0 || ++timeo > 10 || be_hw_error(adapter))
2110 for_all_tx_queues(adapter, txo, i) {
2112 if (atomic_read(&txq->used))
2113 dev_err(&adapter->pdev->dev, "%d pending tx-compls\n",
2114 atomic_read(&txq->used));
2116 /* free posted tx for which compls will never arrive */
2117 while (atomic_read(&txq->used)) {
2118 sent_skb = txo->sent_skb_list[txq->tail];
2119 end_idx = txq->tail;
2120 num_wrbs = wrb_cnt_for_skb(adapter, sent_skb,
2122 index_adv(&end_idx, num_wrbs - 1, txq->len);
2123 num_wrbs = be_tx_compl_process(adapter, txo, end_idx);
2124 atomic_sub(num_wrbs, &txq->used);
2129 static void be_evt_queues_destroy(struct be_adapter *adapter)
2131 struct be_eq_obj *eqo;
2134 for_all_evt_queues(adapter, eqo, i) {
2135 if (eqo->q.created) {
2137 be_cmd_q_destroy(adapter, &eqo->q, QTYPE_EQ);
2138 napi_hash_del(&eqo->napi);
2139 netif_napi_del(&eqo->napi);
2141 be_queue_free(adapter, &eqo->q);
2145 static int be_evt_queues_create(struct be_adapter *adapter)
2147 struct be_queue_info *eq;
2148 struct be_eq_obj *eqo;
2149 struct be_aic_obj *aic;
2152 adapter->num_evt_qs = min_t(u16, num_irqs(adapter),
2153 adapter->cfg_num_qs);
2155 for_all_evt_queues(adapter, eqo, i) {
2156 netif_napi_add(adapter->netdev, &eqo->napi, be_poll,
2158 napi_hash_add(&eqo->napi);
2159 aic = &adapter->aic_obj[i];
2160 eqo->adapter = adapter;
2161 eqo->tx_budget = BE_TX_BUDGET;
2163 aic->max_eqd = BE_MAX_EQD;
2167 rc = be_queue_alloc(adapter, eq, EVNT_Q_LEN,
2168 sizeof(struct be_eq_entry));
2172 rc = be_cmd_eq_create(adapter, eqo);
2179 static void be_mcc_queues_destroy(struct be_adapter *adapter)
2181 struct be_queue_info *q;
2183 q = &adapter->mcc_obj.q;
2185 be_cmd_q_destroy(adapter, q, QTYPE_MCCQ);
2186 be_queue_free(adapter, q);
2188 q = &adapter->mcc_obj.cq;
2190 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2191 be_queue_free(adapter, q);
2194 /* Must be called only after TX qs are created as MCC shares TX EQ */
2195 static int be_mcc_queues_create(struct be_adapter *adapter)
2197 struct be_queue_info *q, *cq;
2199 cq = &adapter->mcc_obj.cq;
2200 if (be_queue_alloc(adapter, cq, MCC_CQ_LEN,
2201 sizeof(struct be_mcc_compl)))
2204 /* Use the default EQ for MCC completions */
2205 if (be_cmd_cq_create(adapter, cq, &mcc_eqo(adapter)->q, true, 0))
2208 q = &adapter->mcc_obj.q;
2209 if (be_queue_alloc(adapter, q, MCC_Q_LEN, sizeof(struct be_mcc_wrb)))
2210 goto mcc_cq_destroy;
2212 if (be_cmd_mccq_create(adapter, q, cq))
2218 be_queue_free(adapter, q);
2220 be_cmd_q_destroy(adapter, cq, QTYPE_CQ);
2222 be_queue_free(adapter, cq);
2227 static void be_tx_queues_destroy(struct be_adapter *adapter)
2229 struct be_queue_info *q;
2230 struct be_tx_obj *txo;
2233 for_all_tx_queues(adapter, txo, i) {
2236 be_cmd_q_destroy(adapter, q, QTYPE_TXQ);
2237 be_queue_free(adapter, q);
2241 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2242 be_queue_free(adapter, q);
2246 static int be_tx_qs_create(struct be_adapter *adapter)
2248 struct be_queue_info *cq, *eq;
2249 struct be_tx_obj *txo;
2252 adapter->num_tx_qs = min(adapter->num_evt_qs, be_max_txqs(adapter));
2254 for_all_tx_queues(adapter, txo, i) {
2256 status = be_queue_alloc(adapter, cq, TX_CQ_LEN,
2257 sizeof(struct be_eth_tx_compl));
2261 u64_stats_init(&txo->stats.sync);
2262 u64_stats_init(&txo->stats.sync_compl);
2264 /* If num_evt_qs is less than num_tx_qs, then more than
2265 * one txq share an eq
2267 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2268 status = be_cmd_cq_create(adapter, cq, eq, false, 3);
2272 status = be_queue_alloc(adapter, &txo->q, TX_Q_LEN,
2273 sizeof(struct be_eth_wrb));
2277 status = be_cmd_txq_create(adapter, txo);
2282 dev_info(&adapter->pdev->dev, "created %d TX queue(s)\n",
2283 adapter->num_tx_qs);
2287 static void be_rx_cqs_destroy(struct be_adapter *adapter)
2289 struct be_queue_info *q;
2290 struct be_rx_obj *rxo;
2293 for_all_rx_queues(adapter, rxo, i) {
2296 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2297 be_queue_free(adapter, q);
2301 static int be_rx_cqs_create(struct be_adapter *adapter)
2303 struct be_queue_info *eq, *cq;
2304 struct be_rx_obj *rxo;
2307 /* We can create as many RSS rings as there are EQs. */
2308 adapter->num_rx_qs = adapter->num_evt_qs;
2310 /* We'll use RSS only if atleast 2 RSS rings are supported.
2311 * When RSS is used, we'll need a default RXQ for non-IP traffic.
2313 if (adapter->num_rx_qs > 1)
2314 adapter->num_rx_qs++;
2316 adapter->big_page_size = (1 << get_order(rx_frag_size)) * PAGE_SIZE;
2317 for_all_rx_queues(adapter, rxo, i) {
2318 rxo->adapter = adapter;
2320 rc = be_queue_alloc(adapter, cq, RX_CQ_LEN,
2321 sizeof(struct be_eth_rx_compl));
2325 u64_stats_init(&rxo->stats.sync);
2326 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2327 rc = be_cmd_cq_create(adapter, cq, eq, false, 3);
2332 dev_info(&adapter->pdev->dev,
2333 "created %d RSS queue(s) and 1 default RX queue\n",
2334 adapter->num_rx_qs - 1);
2338 static irqreturn_t be_intx(int irq, void *dev)
2340 struct be_eq_obj *eqo = dev;
2341 struct be_adapter *adapter = eqo->adapter;
2344 /* IRQ is not expected when NAPI is scheduled as the EQ
2345 * will not be armed.
2346 * But, this can happen on Lancer INTx where it takes
2347 * a while to de-assert INTx or in BE2 where occasionaly
2348 * an interrupt may be raised even when EQ is unarmed.
2349 * If NAPI is already scheduled, then counting & notifying
2350 * events will orphan them.
2352 if (napi_schedule_prep(&eqo->napi)) {
2353 num_evts = events_get(eqo);
2354 __napi_schedule(&eqo->napi);
2356 eqo->spurious_intr = 0;
2358 be_eq_notify(adapter, eqo->q.id, false, true, num_evts);
2360 /* Return IRQ_HANDLED only for the the first spurious intr
2361 * after a valid intr to stop the kernel from branding
2362 * this irq as a bad one!
2364 if (num_evts || eqo->spurious_intr++ == 0)
2370 static irqreturn_t be_msix(int irq, void *dev)
2372 struct be_eq_obj *eqo = dev;
2374 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
2375 napi_schedule(&eqo->napi);
2379 static inline bool do_gro(struct be_rx_compl_info *rxcp)
2381 return (rxcp->tcpf && !rxcp->err && rxcp->l4_csum) ? true : false;
2384 static int be_process_rx(struct be_rx_obj *rxo, struct napi_struct *napi,
2385 int budget, int polling)
2387 struct be_adapter *adapter = rxo->adapter;
2388 struct be_queue_info *rx_cq = &rxo->cq;
2389 struct be_rx_compl_info *rxcp;
2392 for (work_done = 0; work_done < budget; work_done++) {
2393 rxcp = be_rx_compl_get(rxo);
2397 /* Is it a flush compl that has no data */
2398 if (unlikely(rxcp->num_rcvd == 0))
2401 /* Discard compl with partial DMA Lancer B0 */
2402 if (unlikely(!rxcp->pkt_size)) {
2403 be_rx_compl_discard(rxo, rxcp);
2407 /* On BE drop pkts that arrive due to imperfect filtering in
2408 * promiscuous mode on some skews
2410 if (unlikely(rxcp->port != adapter->port_num &&
2411 !lancer_chip(adapter))) {
2412 be_rx_compl_discard(rxo, rxcp);
2416 /* Don't do gro when we're busy_polling */
2417 if (do_gro(rxcp) && polling != BUSY_POLLING)
2418 be_rx_compl_process_gro(rxo, napi, rxcp);
2420 be_rx_compl_process(rxo, napi, rxcp);
2423 be_rx_stats_update(rxo, rxcp);
2427 be_cq_notify(adapter, rx_cq->id, true, work_done);
2429 /* When an rx-obj gets into post_starved state, just
2430 * let be_worker do the posting.
2432 if (atomic_read(&rxo->q.used) < RX_FRAGS_REFILL_WM &&
2433 !rxo->rx_post_starved)
2434 be_post_rx_frags(rxo, GFP_ATOMIC);
2440 static bool be_process_tx(struct be_adapter *adapter, struct be_tx_obj *txo,
2441 int budget, int idx)
2443 struct be_eth_tx_compl *txcp;
2444 int num_wrbs = 0, work_done;
2446 for (work_done = 0; work_done < budget; work_done++) {
2447 txcp = be_tx_compl_get(&txo->cq);
2450 num_wrbs += be_tx_compl_process(adapter, txo,
2451 AMAP_GET_BITS(struct
2457 be_cq_notify(adapter, txo->cq.id, true, work_done);
2458 atomic_sub(num_wrbs, &txo->q.used);
2460 /* As Tx wrbs have been freed up, wake up netdev queue
2461 * if it was stopped due to lack of tx wrbs. */
2462 if (__netif_subqueue_stopped(adapter->netdev, idx) &&
2463 atomic_read(&txo->q.used) < txo->q.len / 2) {
2464 netif_wake_subqueue(adapter->netdev, idx);
2467 u64_stats_update_begin(&tx_stats(txo)->sync_compl);
2468 tx_stats(txo)->tx_compl += work_done;
2469 u64_stats_update_end(&tx_stats(txo)->sync_compl);
2471 return (work_done < budget); /* Done */
2474 int be_poll(struct napi_struct *napi, int budget)
2476 struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2477 struct be_adapter *adapter = eqo->adapter;
2478 int max_work = 0, work, i, num_evts;
2479 struct be_rx_obj *rxo;
2482 num_evts = events_get(eqo);
2484 /* Process all TXQs serviced by this EQ */
2485 for (i = eqo->idx; i < adapter->num_tx_qs; i += adapter->num_evt_qs) {
2486 tx_done = be_process_tx(adapter, &adapter->tx_obj[i],
2492 if (be_lock_napi(eqo)) {
2493 /* This loop will iterate twice for EQ0 in which
2494 * completions of the last RXQ (default one) are also processed
2495 * For other EQs the loop iterates only once
2497 for_all_rx_queues_on_eq(adapter, eqo, rxo, i) {
2498 work = be_process_rx(rxo, napi, budget, NAPI_POLLING);
2499 max_work = max(work, max_work);
2501 be_unlock_napi(eqo);
2506 if (is_mcc_eqo(eqo))
2507 be_process_mcc(adapter);
2509 if (max_work < budget) {
2510 napi_complete(napi);
2511 be_eq_notify(adapter, eqo->q.id, true, false, num_evts);
2513 /* As we'll continue in polling mode, count and clear events */
2514 be_eq_notify(adapter, eqo->q.id, false, false, num_evts);
2519 #ifdef CONFIG_NET_RX_BUSY_POLL
2520 static int be_busy_poll(struct napi_struct *napi)
2522 struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2523 struct be_adapter *adapter = eqo->adapter;
2524 struct be_rx_obj *rxo;
2527 if (!be_lock_busy_poll(eqo))
2528 return LL_FLUSH_BUSY;
2530 for_all_rx_queues_on_eq(adapter, eqo, rxo, i) {
2531 work = be_process_rx(rxo, napi, 4, BUSY_POLLING);
2536 be_unlock_busy_poll(eqo);
2541 void be_detect_error(struct be_adapter *adapter)
2543 u32 ue_lo = 0, ue_hi = 0, ue_lo_mask = 0, ue_hi_mask = 0;
2544 u32 sliport_status = 0, sliport_err1 = 0, sliport_err2 = 0;
2546 bool error_detected = false;
2547 struct device *dev = &adapter->pdev->dev;
2548 struct net_device *netdev = adapter->netdev;
2550 if (be_hw_error(adapter))
2553 if (lancer_chip(adapter)) {
2554 sliport_status = ioread32(adapter->db + SLIPORT_STATUS_OFFSET);
2555 if (sliport_status & SLIPORT_STATUS_ERR_MASK) {
2556 sliport_err1 = ioread32(adapter->db +
2557 SLIPORT_ERROR1_OFFSET);
2558 sliport_err2 = ioread32(adapter->db +
2559 SLIPORT_ERROR2_OFFSET);
2560 adapter->hw_error = true;
2561 /* Do not log error messages if its a FW reset */
2562 if (sliport_err1 == SLIPORT_ERROR_FW_RESET1 &&
2563 sliport_err2 == SLIPORT_ERROR_FW_RESET2) {
2564 dev_info(dev, "Firmware update in progress\n");
2566 error_detected = true;
2567 dev_err(dev, "Error detected in the card\n");
2568 dev_err(dev, "ERR: sliport status 0x%x\n",
2570 dev_err(dev, "ERR: sliport error1 0x%x\n",
2572 dev_err(dev, "ERR: sliport error2 0x%x\n",
2577 pci_read_config_dword(adapter->pdev,
2578 PCICFG_UE_STATUS_LOW, &ue_lo);
2579 pci_read_config_dword(adapter->pdev,
2580 PCICFG_UE_STATUS_HIGH, &ue_hi);
2581 pci_read_config_dword(adapter->pdev,
2582 PCICFG_UE_STATUS_LOW_MASK, &ue_lo_mask);
2583 pci_read_config_dword(adapter->pdev,
2584 PCICFG_UE_STATUS_HI_MASK, &ue_hi_mask);
2586 ue_lo = (ue_lo & ~ue_lo_mask);
2587 ue_hi = (ue_hi & ~ue_hi_mask);
2589 /* On certain platforms BE hardware can indicate spurious UEs.
2590 * Allow HW to stop working completely in case of a real UE.
2591 * Hence not setting the hw_error for UE detection.
2594 if (ue_lo || ue_hi) {
2595 error_detected = true;
2597 "Unrecoverable Error detected in the adapter");
2598 dev_err(dev, "Please reboot server to recover");
2599 if (skyhawk_chip(adapter))
2600 adapter->hw_error = true;
2601 for (i = 0; ue_lo; ue_lo >>= 1, i++) {
2603 dev_err(dev, "UE: %s bit set\n",
2604 ue_status_low_desc[i]);
2606 for (i = 0; ue_hi; ue_hi >>= 1, i++) {
2608 dev_err(dev, "UE: %s bit set\n",
2609 ue_status_hi_desc[i]);
2614 netif_carrier_off(netdev);
2617 static void be_msix_disable(struct be_adapter *adapter)
2619 if (msix_enabled(adapter)) {
2620 pci_disable_msix(adapter->pdev);
2621 adapter->num_msix_vec = 0;
2622 adapter->num_msix_roce_vec = 0;
2626 static int be_msix_enable(struct be_adapter *adapter)
2629 struct device *dev = &adapter->pdev->dev;
2631 /* If RoCE is supported, program the max number of NIC vectors that
2632 * may be configured via set-channels, along with vectors needed for
2633 * RoCe. Else, just program the number we'll use initially.
2635 if (be_roce_supported(adapter))
2636 num_vec = min_t(int, 2 * be_max_eqs(adapter),
2637 2 * num_online_cpus());
2639 num_vec = adapter->cfg_num_qs;
2641 for (i = 0; i < num_vec; i++)
2642 adapter->msix_entries[i].entry = i;
2644 num_vec = pci_enable_msix_range(adapter->pdev, adapter->msix_entries,
2645 MIN_MSIX_VECTORS, num_vec);
2649 if (be_roce_supported(adapter) && num_vec > MIN_MSIX_VECTORS) {
2650 adapter->num_msix_roce_vec = num_vec / 2;
2651 dev_info(dev, "enabled %d MSI-x vector(s) for RoCE\n",
2652 adapter->num_msix_roce_vec);
2655 adapter->num_msix_vec = num_vec - adapter->num_msix_roce_vec;
2657 dev_info(dev, "enabled %d MSI-x vector(s) for NIC\n",
2658 adapter->num_msix_vec);
2662 dev_warn(dev, "MSIx enable failed\n");
2664 /* INTx is not supported in VFs, so fail probe if enable_msix fails */
2665 if (!be_physfn(adapter))
2670 static inline int be_msix_vec_get(struct be_adapter *adapter,
2671 struct be_eq_obj *eqo)
2673 return adapter->msix_entries[eqo->msix_idx].vector;
2676 static int be_msix_register(struct be_adapter *adapter)
2678 struct net_device *netdev = adapter->netdev;
2679 struct be_eq_obj *eqo;
2682 for_all_evt_queues(adapter, eqo, i) {
2683 sprintf(eqo->desc, "%s-q%d", netdev->name, i);
2684 vec = be_msix_vec_get(adapter, eqo);
2685 status = request_irq(vec, be_msix, 0, eqo->desc, eqo);
2692 for (i--, eqo = &adapter->eq_obj[i]; i >= 0; i--, eqo--)
2693 free_irq(be_msix_vec_get(adapter, eqo), eqo);
2694 dev_warn(&adapter->pdev->dev, "MSIX Request IRQ failed - err %d\n",
2696 be_msix_disable(adapter);
2700 static int be_irq_register(struct be_adapter *adapter)
2702 struct net_device *netdev = adapter->netdev;
2705 if (msix_enabled(adapter)) {
2706 status = be_msix_register(adapter);
2709 /* INTx is not supported for VF */
2710 if (!be_physfn(adapter))
2714 /* INTx: only the first EQ is used */
2715 netdev->irq = adapter->pdev->irq;
2716 status = request_irq(netdev->irq, be_intx, IRQF_SHARED, netdev->name,
2717 &adapter->eq_obj[0]);
2719 dev_err(&adapter->pdev->dev,
2720 "INTx request IRQ failed - err %d\n", status);
2724 adapter->isr_registered = true;
2728 static void be_irq_unregister(struct be_adapter *adapter)
2730 struct net_device *netdev = adapter->netdev;
2731 struct be_eq_obj *eqo;
2734 if (!adapter->isr_registered)
2738 if (!msix_enabled(adapter)) {
2739 free_irq(netdev->irq, &adapter->eq_obj[0]);
2744 for_all_evt_queues(adapter, eqo, i)
2745 free_irq(be_msix_vec_get(adapter, eqo), eqo);
2748 adapter->isr_registered = false;
2751 static void be_rx_qs_destroy(struct be_adapter *adapter)
2753 struct be_queue_info *q;
2754 struct be_rx_obj *rxo;
2757 for_all_rx_queues(adapter, rxo, i) {
2760 be_cmd_rxq_destroy(adapter, q);
2761 be_rx_cq_clean(rxo);
2763 be_queue_free(adapter, q);
2767 static int be_close(struct net_device *netdev)
2769 struct be_adapter *adapter = netdev_priv(netdev);
2770 struct be_eq_obj *eqo;
2773 /* This protection is needed as be_close() may be called even when the
2774 * adapter is in cleared state (after eeh perm failure)
2776 if (!(adapter->flags & BE_FLAGS_SETUP_DONE))
2779 be_roce_dev_close(adapter);
2781 if (adapter->flags & BE_FLAGS_NAPI_ENABLED) {
2782 for_all_evt_queues(adapter, eqo, i) {
2783 napi_disable(&eqo->napi);
2784 be_disable_busy_poll(eqo);
2786 adapter->flags &= ~BE_FLAGS_NAPI_ENABLED;
2789 be_async_mcc_disable(adapter);
2791 /* Wait for all pending tx completions to arrive so that
2792 * all tx skbs are freed.
2794 netif_tx_disable(netdev);
2795 be_tx_compl_clean(adapter);
2797 be_rx_qs_destroy(adapter);
2799 for (i = 1; i < (adapter->uc_macs + 1); i++)
2800 be_cmd_pmac_del(adapter, adapter->if_handle,
2801 adapter->pmac_id[i], 0);
2802 adapter->uc_macs = 0;
2804 for_all_evt_queues(adapter, eqo, i) {
2805 if (msix_enabled(adapter))
2806 synchronize_irq(be_msix_vec_get(adapter, eqo));
2808 synchronize_irq(netdev->irq);
2812 be_irq_unregister(adapter);
2817 static int be_rx_qs_create(struct be_adapter *adapter)
2819 struct be_rx_obj *rxo;
2821 u8 rss_hkey[RSS_HASH_KEY_LEN];
2822 struct rss_info *rss = &adapter->rss_info;
2824 for_all_rx_queues(adapter, rxo, i) {
2825 rc = be_queue_alloc(adapter, &rxo->q, RX_Q_LEN,
2826 sizeof(struct be_eth_rx_d));
2831 /* The FW would like the default RXQ to be created first */
2832 rxo = default_rxo(adapter);
2833 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id, rx_frag_size,
2834 adapter->if_handle, false, &rxo->rss_id);
2838 for_all_rss_queues(adapter, rxo, i) {
2839 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id,
2840 rx_frag_size, adapter->if_handle,
2841 true, &rxo->rss_id);
2846 if (be_multi_rxq(adapter)) {
2847 for (j = 0; j < RSS_INDIR_TABLE_LEN;
2848 j += adapter->num_rx_qs - 1) {
2849 for_all_rss_queues(adapter, rxo, i) {
2850 if ((j + i) >= RSS_INDIR_TABLE_LEN)
2852 rss->rsstable[j + i] = rxo->rss_id;
2853 rss->rss_queue[j + i] = i;
2856 rss->rss_flags = RSS_ENABLE_TCP_IPV4 | RSS_ENABLE_IPV4 |
2857 RSS_ENABLE_TCP_IPV6 | RSS_ENABLE_IPV6;
2859 if (!BEx_chip(adapter))
2860 rss->rss_flags |= RSS_ENABLE_UDP_IPV4 |
2861 RSS_ENABLE_UDP_IPV6;
2863 /* Disable RSS, if only default RX Q is created */
2864 rss->rss_flags = RSS_ENABLE_NONE;
2867 get_random_bytes(rss_hkey, RSS_HASH_KEY_LEN);
2868 rc = be_cmd_rss_config(adapter, rss->rsstable, rss->rss_flags,
2871 rss->rss_flags = RSS_ENABLE_NONE;
2875 memcpy(rss->rss_hkey, rss_hkey, RSS_HASH_KEY_LEN);
2877 /* First time posting */
2878 for_all_rx_queues(adapter, rxo, i)
2879 be_post_rx_frags(rxo, GFP_KERNEL);
2883 static int be_open(struct net_device *netdev)
2885 struct be_adapter *adapter = netdev_priv(netdev);
2886 struct be_eq_obj *eqo;
2887 struct be_rx_obj *rxo;
2888 struct be_tx_obj *txo;
2892 status = be_rx_qs_create(adapter);
2896 status = be_irq_register(adapter);
2900 for_all_rx_queues(adapter, rxo, i)
2901 be_cq_notify(adapter, rxo->cq.id, true, 0);
2903 for_all_tx_queues(adapter, txo, i)
2904 be_cq_notify(adapter, txo->cq.id, true, 0);
2906 be_async_mcc_enable(adapter);
2908 for_all_evt_queues(adapter, eqo, i) {
2909 napi_enable(&eqo->napi);
2910 be_enable_busy_poll(eqo);
2911 be_eq_notify(adapter, eqo->q.id, true, true, 0);
2913 adapter->flags |= BE_FLAGS_NAPI_ENABLED;
2915 status = be_cmd_link_status_query(adapter, NULL, &link_status, 0);
2917 be_link_status_update(adapter, link_status);
2919 netif_tx_start_all_queues(netdev);
2920 be_roce_dev_open(adapter);
2922 #ifdef CONFIG_BE2NET_VXLAN
2923 if (skyhawk_chip(adapter))
2924 vxlan_get_rx_port(netdev);
2929 be_close(adapter->netdev);
2933 static int be_setup_wol(struct be_adapter *adapter, bool enable)
2935 struct be_dma_mem cmd;
2939 memset(mac, 0, ETH_ALEN);
2941 cmd.size = sizeof(struct be_cmd_req_acpi_wol_magic_config);
2942 cmd.va = dma_zalloc_coherent(&adapter->pdev->dev, cmd.size, &cmd.dma,
2948 status = pci_write_config_dword(adapter->pdev,
2949 PCICFG_PM_CONTROL_OFFSET,
2950 PCICFG_PM_CONTROL_MASK);
2952 dev_err(&adapter->pdev->dev,
2953 "Could not enable Wake-on-lan\n");
2954 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va,
2958 status = be_cmd_enable_magic_wol(adapter,
2959 adapter->netdev->dev_addr,
2961 pci_enable_wake(adapter->pdev, PCI_D3hot, 1);
2962 pci_enable_wake(adapter->pdev, PCI_D3cold, 1);
2964 status = be_cmd_enable_magic_wol(adapter, mac, &cmd);
2965 pci_enable_wake(adapter->pdev, PCI_D3hot, 0);
2966 pci_enable_wake(adapter->pdev, PCI_D3cold, 0);
2969 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va, cmd.dma);
2974 * Generate a seed MAC address from the PF MAC Address using jhash.
2975 * MAC Address for VFs are assigned incrementally starting from the seed.
2976 * These addresses are programmed in the ASIC by the PF and the VF driver
2977 * queries for the MAC address during its probe.
2979 static int be_vf_eth_addr_config(struct be_adapter *adapter)
2984 struct be_vf_cfg *vf_cfg;
2986 be_vf_eth_addr_generate(adapter, mac);
2988 for_all_vfs(adapter, vf_cfg, vf) {
2989 if (BEx_chip(adapter))
2990 status = be_cmd_pmac_add(adapter, mac,
2992 &vf_cfg->pmac_id, vf + 1);
2994 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
2998 dev_err(&adapter->pdev->dev,
2999 "Mac address assignment failed for VF %d\n",
3002 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
3009 static int be_vfs_mac_query(struct be_adapter *adapter)
3013 struct be_vf_cfg *vf_cfg;
3015 for_all_vfs(adapter, vf_cfg, vf) {
3016 status = be_cmd_get_active_mac(adapter, vf_cfg->pmac_id,
3017 mac, vf_cfg->if_handle,
3021 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
3026 static void be_vf_clear(struct be_adapter *adapter)
3028 struct be_vf_cfg *vf_cfg;
3031 if (pci_vfs_assigned(adapter->pdev)) {
3032 dev_warn(&adapter->pdev->dev,
3033 "VFs are assigned to VMs: not disabling VFs\n");
3037 pci_disable_sriov(adapter->pdev);
3039 for_all_vfs(adapter, vf_cfg, vf) {
3040 if (BEx_chip(adapter))
3041 be_cmd_pmac_del(adapter, vf_cfg->if_handle,
3042 vf_cfg->pmac_id, vf + 1);
3044 be_cmd_set_mac(adapter, NULL, vf_cfg->if_handle,
3047 be_cmd_if_destroy(adapter, vf_cfg->if_handle, vf + 1);
3050 kfree(adapter->vf_cfg);
3051 adapter->num_vfs = 0;
3054 static void be_clear_queues(struct be_adapter *adapter)
3056 be_mcc_queues_destroy(adapter);
3057 be_rx_cqs_destroy(adapter);
3058 be_tx_queues_destroy(adapter);
3059 be_evt_queues_destroy(adapter);
3062 static void be_cancel_worker(struct be_adapter *adapter)
3064 if (adapter->flags & BE_FLAGS_WORKER_SCHEDULED) {
3065 cancel_delayed_work_sync(&adapter->work);
3066 adapter->flags &= ~BE_FLAGS_WORKER_SCHEDULED;
3070 static void be_mac_clear(struct be_adapter *adapter)
3074 if (adapter->pmac_id) {
3075 for (i = 0; i < (adapter->uc_macs + 1); i++)
3076 be_cmd_pmac_del(adapter, adapter->if_handle,
3077 adapter->pmac_id[i], 0);
3078 adapter->uc_macs = 0;
3080 kfree(adapter->pmac_id);
3081 adapter->pmac_id = NULL;
3085 #ifdef CONFIG_BE2NET_VXLAN
3086 static void be_disable_vxlan_offloads(struct be_adapter *adapter)
3088 if (adapter->flags & BE_FLAGS_VXLAN_OFFLOADS)
3089 be_cmd_manage_iface(adapter, adapter->if_handle,
3090 OP_CONVERT_TUNNEL_TO_NORMAL);
3092 if (adapter->vxlan_port)
3093 be_cmd_set_vxlan_port(adapter, 0);
3095 adapter->flags &= ~BE_FLAGS_VXLAN_OFFLOADS;
3096 adapter->vxlan_port = 0;
3100 static int be_clear(struct be_adapter *adapter)
3102 be_cancel_worker(adapter);
3104 if (sriov_enabled(adapter))
3105 be_vf_clear(adapter);
3107 /* Re-configure FW to distribute resources evenly across max-supported
3108 * number of VFs, only when VFs are not already enabled.
3110 if (be_physfn(adapter) && !pci_vfs_assigned(adapter->pdev))
3111 be_cmd_set_sriov_config(adapter, adapter->pool_res,
3112 pci_sriov_get_totalvfs(adapter->pdev));
3114 #ifdef CONFIG_BE2NET_VXLAN
3115 be_disable_vxlan_offloads(adapter);
3117 /* delete the primary mac along with the uc-mac list */
3118 be_mac_clear(adapter);
3120 be_cmd_if_destroy(adapter, adapter->if_handle, 0);
3122 be_clear_queues(adapter);
3124 be_msix_disable(adapter);
3125 adapter->flags &= ~BE_FLAGS_SETUP_DONE;
3129 static int be_vfs_if_create(struct be_adapter *adapter)
3131 struct be_resources res = {0};
3132 struct be_vf_cfg *vf_cfg;
3133 u32 cap_flags, en_flags, vf;
3136 cap_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3137 BE_IF_FLAGS_MULTICAST;
3139 for_all_vfs(adapter, vf_cfg, vf) {
3140 if (!BE3_chip(adapter)) {
3141 status = be_cmd_get_profile_config(adapter, &res,
3144 cap_flags = res.if_cap_flags;
3147 /* If a FW profile exists, then cap_flags are updated */
3148 en_flags = cap_flags & (BE_IF_FLAGS_UNTAGGED |
3149 BE_IF_FLAGS_BROADCAST |
3150 BE_IF_FLAGS_MULTICAST);
3152 be_cmd_if_create(adapter, cap_flags, en_flags,
3153 &vf_cfg->if_handle, vf + 1);
3161 static int be_vf_setup_init(struct be_adapter *adapter)
3163 struct be_vf_cfg *vf_cfg;
3166 adapter->vf_cfg = kcalloc(adapter->num_vfs, sizeof(*vf_cfg),
3168 if (!adapter->vf_cfg)
3171 for_all_vfs(adapter, vf_cfg, vf) {
3172 vf_cfg->if_handle = -1;
3173 vf_cfg->pmac_id = -1;
3178 static int be_vf_setup(struct be_adapter *adapter)
3180 struct device *dev = &adapter->pdev->dev;
3181 struct be_vf_cfg *vf_cfg;
3182 int status, old_vfs, vf;
3185 old_vfs = pci_num_vf(adapter->pdev);
3187 status = be_vf_setup_init(adapter);
3192 for_all_vfs(adapter, vf_cfg, vf) {
3193 status = be_cmd_get_if_id(adapter, vf_cfg, vf);
3198 status = be_vfs_mac_query(adapter);
3202 status = be_vfs_if_create(adapter);
3206 status = be_vf_eth_addr_config(adapter);
3211 for_all_vfs(adapter, vf_cfg, vf) {
3212 /* Allow VFs to programs MAC/VLAN filters */
3213 status = be_cmd_get_fn_privileges(adapter, &privileges, vf + 1);
3214 if (!status && !(privileges & BE_PRIV_FILTMGMT)) {
3215 status = be_cmd_set_fn_privileges(adapter,
3220 dev_info(dev, "VF%d has FILTMGMT privilege\n",
3224 /* Allow full available bandwidth */
3226 be_cmd_config_qos(adapter, 0, 0, vf + 1);
3229 be_cmd_enable_vf(adapter, vf + 1);
3230 be_cmd_set_logical_link_config(adapter,
3231 IFLA_VF_LINK_STATE_AUTO,
3237 status = pci_enable_sriov(adapter->pdev, adapter->num_vfs);
3239 dev_err(dev, "SRIOV enable failed\n");
3240 adapter->num_vfs = 0;
3246 dev_err(dev, "VF setup failed\n");
3247 be_vf_clear(adapter);
3251 /* Converting function_mode bits on BE3 to SH mc_type enums */
3253 static u8 be_convert_mc_type(u32 function_mode)
3255 if (function_mode & VNIC_MODE && function_mode & QNQ_MODE)
3257 else if (function_mode & QNQ_MODE)
3259 else if (function_mode & VNIC_MODE)
3261 else if (function_mode & UMC_ENABLED)
3267 /* On BE2/BE3 FW does not suggest the supported limits */
3268 static void BEx_get_resources(struct be_adapter *adapter,
3269 struct be_resources *res)
3271 bool use_sriov = adapter->num_vfs ? 1 : 0;
3273 if (be_physfn(adapter))
3274 res->max_uc_mac = BE_UC_PMAC_COUNT;
3276 res->max_uc_mac = BE_VF_UC_PMAC_COUNT;
3278 adapter->mc_type = be_convert_mc_type(adapter->function_mode);
3280 if (be_is_mc(adapter)) {
3281 /* Assuming that there are 4 channels per port,
3282 * when multi-channel is enabled
3284 if (be_is_qnq_mode(adapter))
3285 res->max_vlans = BE_NUM_VLANS_SUPPORTED/8;
3287 /* In a non-qnq multichannel mode, the pvid
3288 * takes up one vlan entry
3290 res->max_vlans = (BE_NUM_VLANS_SUPPORTED / 4) - 1;
3292 res->max_vlans = BE_NUM_VLANS_SUPPORTED;
3295 res->max_mcast_mac = BE_MAX_MC;
3297 /* 1) For BE3 1Gb ports, FW does not support multiple TXQs
3298 * 2) Create multiple TX rings on a BE3-R multi-channel interface
3299 * *only* if it is RSS-capable.
3301 if (BE2_chip(adapter) || use_sriov || (adapter->port_num > 1) ||
3302 !be_physfn(adapter) || (be_is_mc(adapter) &&
3303 !(adapter->function_caps & BE_FUNCTION_CAPS_RSS)))
3306 res->max_tx_qs = BE3_MAX_TX_QS;
3308 if ((adapter->function_caps & BE_FUNCTION_CAPS_RSS) &&
3309 !use_sriov && be_physfn(adapter))
3310 res->max_rss_qs = (adapter->be3_native) ?
3311 BE3_MAX_RSS_QS : BE2_MAX_RSS_QS;
3312 res->max_rx_qs = res->max_rss_qs + 1;
3314 if (be_physfn(adapter))
3315 res->max_evt_qs = (res->max_vfs > 0) ?
3316 BE3_SRIOV_MAX_EVT_QS : BE3_MAX_EVT_QS;
3318 res->max_evt_qs = 1;
3320 res->if_cap_flags = BE_IF_CAP_FLAGS_WANT;
3321 if (!(adapter->function_caps & BE_FUNCTION_CAPS_RSS))
3322 res->if_cap_flags &= ~BE_IF_FLAGS_RSS;
3325 static void be_setup_init(struct be_adapter *adapter)
3327 adapter->vlan_prio_bmap = 0xff;
3328 adapter->phy.link_speed = -1;
3329 adapter->if_handle = -1;
3330 adapter->be3_native = false;
3331 adapter->promiscuous = false;
3332 if (be_physfn(adapter))
3333 adapter->cmd_privileges = MAX_PRIVILEGES;
3335 adapter->cmd_privileges = MIN_PRIVILEGES;
3338 static int be_get_sriov_config(struct be_adapter *adapter)
3340 struct device *dev = &adapter->pdev->dev;
3341 struct be_resources res = {0};
3342 int status, max_vfs, old_vfs;
3344 status = be_cmd_get_profile_config(adapter, &res, 0);
3348 adapter->pool_res = res;
3350 /* Some old versions of BE3 FW don't report max_vfs value */
3351 if (BE3_chip(adapter) && !res.max_vfs) {
3352 max_vfs = pci_sriov_get_totalvfs(adapter->pdev);
3353 res.max_vfs = max_vfs > 0 ? min(MAX_VFS, max_vfs) : 0;
3356 adapter->pool_res.max_vfs = res.max_vfs;
3357 pci_sriov_set_totalvfs(adapter->pdev, be_max_vfs(adapter));
3359 if (!be_max_vfs(adapter)) {
3361 dev_warn(dev, "device doesn't support SRIOV\n");
3362 adapter->num_vfs = 0;
3366 /* validate num_vfs module param */
3367 old_vfs = pci_num_vf(adapter->pdev);
3369 dev_info(dev, "%d VFs are already enabled\n", old_vfs);
3370 if (old_vfs != num_vfs)
3371 dev_warn(dev, "Ignoring num_vfs=%d setting\n", num_vfs);
3372 adapter->num_vfs = old_vfs;
3374 if (num_vfs > be_max_vfs(adapter)) {
3375 dev_info(dev, "Resources unavailable to init %d VFs\n",
3377 dev_info(dev, "Limiting to %d VFs\n",
3378 be_max_vfs(adapter));
3380 adapter->num_vfs = min_t(u16, num_vfs, be_max_vfs(adapter));
3386 static int be_get_resources(struct be_adapter *adapter)
3388 struct device *dev = &adapter->pdev->dev;
3389 struct be_resources res = {0};
3392 if (BEx_chip(adapter)) {
3393 BEx_get_resources(adapter, &res);
3397 /* For Lancer, SH etc read per-function resource limits from FW.
3398 * GET_FUNC_CONFIG returns per function guaranteed limits.
3399 * GET_PROFILE_CONFIG returns PCI-E related limits PF-pool limits
3401 if (!BEx_chip(adapter)) {
3402 status = be_cmd_get_func_config(adapter, &res);
3406 /* If RoCE may be enabled stash away half the EQs for RoCE */
3407 if (be_roce_supported(adapter))
3408 res.max_evt_qs /= 2;
3411 dev_info(dev, "Max: txqs %d, rxqs %d, rss %d, eqs %d, vfs %d\n",
3412 be_max_txqs(adapter), be_max_rxqs(adapter),
3413 be_max_rss(adapter), be_max_eqs(adapter),
3414 be_max_vfs(adapter));
3415 dev_info(dev, "Max: uc-macs %d, mc-macs %d, vlans %d\n",
3416 be_max_uc(adapter), be_max_mc(adapter),
3417 be_max_vlans(adapter));
3423 static int be_get_config(struct be_adapter *adapter)
3428 status = be_cmd_query_fw_cfg(adapter, &adapter->port_num,
3429 &adapter->function_mode,
3430 &adapter->function_caps,
3431 &adapter->asic_rev);
3435 if (be_physfn(adapter)) {
3436 status = be_cmd_get_active_profile(adapter, &profile_id);
3438 dev_info(&adapter->pdev->dev,
3439 "Using profile 0x%x\n", profile_id);
3441 status = be_get_sriov_config(adapter);
3445 /* When the HW is in SRIOV capable configuration, the PF-pool
3446 * resources are equally distributed across the max-number of
3447 * VFs. The user may request only a subset of the max-vfs to be
3448 * enabled. Based on num_vfs, redistribute the resources across
3449 * num_vfs so that each VF will have access to more number of
3450 * resources. This facility is not available in BE3 FW.
3451 * Also, this is done by FW in Lancer chip.
3453 if (!pci_num_vf(adapter->pdev)) {
3454 status = be_cmd_set_sriov_config(adapter,
3462 status = be_get_resources(adapter);
3466 adapter->pmac_id = kcalloc(be_max_uc(adapter),
3467 sizeof(*adapter->pmac_id), GFP_KERNEL);
3468 if (!adapter->pmac_id)
3471 /* Sanitize cfg_num_qs based on HW and platform limits */
3472 adapter->cfg_num_qs = min(adapter->cfg_num_qs, be_max_qs(adapter));
3477 static int be_mac_setup(struct be_adapter *adapter)
3482 if (is_zero_ether_addr(adapter->netdev->dev_addr)) {
3483 status = be_cmd_get_perm_mac(adapter, mac);
3487 memcpy(adapter->netdev->dev_addr, mac, ETH_ALEN);
3488 memcpy(adapter->netdev->perm_addr, mac, ETH_ALEN);
3490 /* Maybe the HW was reset; dev_addr must be re-programmed */
3491 memcpy(mac, adapter->netdev->dev_addr, ETH_ALEN);
3494 /* For BE3-R VFs, the PF programs the initial MAC address */
3495 if (!(BEx_chip(adapter) && be_virtfn(adapter)))
3496 be_cmd_pmac_add(adapter, mac, adapter->if_handle,
3497 &adapter->pmac_id[0], 0);
3501 static void be_schedule_worker(struct be_adapter *adapter)
3503 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
3504 adapter->flags |= BE_FLAGS_WORKER_SCHEDULED;
3507 static int be_setup_queues(struct be_adapter *adapter)
3509 struct net_device *netdev = adapter->netdev;
3512 status = be_evt_queues_create(adapter);
3516 status = be_tx_qs_create(adapter);
3520 status = be_rx_cqs_create(adapter);
3524 status = be_mcc_queues_create(adapter);
3528 status = netif_set_real_num_rx_queues(netdev, adapter->num_rx_qs);
3532 status = netif_set_real_num_tx_queues(netdev, adapter->num_tx_qs);
3538 dev_err(&adapter->pdev->dev, "queue_setup failed\n");
3542 int be_update_queues(struct be_adapter *adapter)
3544 struct net_device *netdev = adapter->netdev;
3547 if (netif_running(netdev))
3550 be_cancel_worker(adapter);
3552 /* If any vectors have been shared with RoCE we cannot re-program
3555 if (!adapter->num_msix_roce_vec)
3556 be_msix_disable(adapter);
3558 be_clear_queues(adapter);
3560 if (!msix_enabled(adapter)) {
3561 status = be_msix_enable(adapter);
3566 status = be_setup_queues(adapter);
3570 be_schedule_worker(adapter);
3572 if (netif_running(netdev))
3573 status = be_open(netdev);
3578 static int be_setup(struct be_adapter *adapter)
3580 struct device *dev = &adapter->pdev->dev;
3581 u32 tx_fc, rx_fc, en_flags;
3584 be_setup_init(adapter);
3586 if (!lancer_chip(adapter))
3587 be_cmd_req_native_mode(adapter);
3589 status = be_get_config(adapter);
3593 status = be_msix_enable(adapter);
3597 en_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3598 BE_IF_FLAGS_MULTICAST | BE_IF_FLAGS_PASS_L3L4_ERRORS;
3599 if (adapter->function_caps & BE_FUNCTION_CAPS_RSS)
3600 en_flags |= BE_IF_FLAGS_RSS;
3601 en_flags = en_flags & be_if_cap_flags(adapter);
3602 status = be_cmd_if_create(adapter, be_if_cap_flags(adapter), en_flags,
3603 &adapter->if_handle, 0);
3607 /* Updating real_num_tx/rx_queues() requires rtnl_lock() */
3609 status = be_setup_queues(adapter);
3614 be_cmd_get_fn_privileges(adapter, &adapter->cmd_privileges, 0);
3616 status = be_mac_setup(adapter);
3620 be_cmd_get_fw_ver(adapter, adapter->fw_ver, adapter->fw_on_flash);
3622 if (BE2_chip(adapter) && fw_major_num(adapter->fw_ver) < 4) {
3623 dev_err(dev, "Firmware on card is old(%s), IRQs may not work.",
3625 dev_err(dev, "Please upgrade firmware to version >= 4.0\n");
3628 if (adapter->vlans_added)
3629 be_vid_config(adapter);
3631 be_set_rx_mode(adapter->netdev);
3633 be_cmd_get_acpi_wol_cap(adapter);
3635 be_cmd_get_flow_control(adapter, &tx_fc, &rx_fc);
3637 if (rx_fc != adapter->rx_fc || tx_fc != adapter->tx_fc)
3638 be_cmd_set_flow_control(adapter, adapter->tx_fc,
3641 if (be_physfn(adapter))
3642 be_cmd_set_logical_link_config(adapter,
3643 IFLA_VF_LINK_STATE_AUTO, 0);
3645 if (adapter->num_vfs)
3646 be_vf_setup(adapter);
3648 status = be_cmd_get_phy_info(adapter);
3649 if (!status && be_pause_supported(adapter))
3650 adapter->phy.fc_autoneg = 1;
3652 be_schedule_worker(adapter);
3653 adapter->flags |= BE_FLAGS_SETUP_DONE;
3660 #ifdef CONFIG_NET_POLL_CONTROLLER
3661 static void be_netpoll(struct net_device *netdev)
3663 struct be_adapter *adapter = netdev_priv(netdev);
3664 struct be_eq_obj *eqo;
3667 for_all_evt_queues(adapter, eqo, i) {
3668 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
3669 napi_schedule(&eqo->napi);
3676 static char flash_cookie[2][16] = {"*** SE FLAS", "H DIRECTORY *** "};
3678 static bool phy_flashing_required(struct be_adapter *adapter)
3680 return (adapter->phy.phy_type == TN_8022 &&
3681 adapter->phy.interface_type == PHY_TYPE_BASET_10GB);
3684 static bool is_comp_in_ufi(struct be_adapter *adapter,
3685 struct flash_section_info *fsec, int type)
3687 int i = 0, img_type = 0;
3688 struct flash_section_info_g2 *fsec_g2 = NULL;
3690 if (BE2_chip(adapter))
3691 fsec_g2 = (struct flash_section_info_g2 *)fsec;
3693 for (i = 0; i < MAX_FLASH_COMP; i++) {
3695 img_type = le32_to_cpu(fsec_g2->fsec_entry[i].type);
3697 img_type = le32_to_cpu(fsec->fsec_entry[i].type);
3699 if (img_type == type)
3706 static struct flash_section_info *get_fsec_info(struct be_adapter *adapter,
3708 const struct firmware *fw)
3710 struct flash_section_info *fsec = NULL;
3711 const u8 *p = fw->data;
3714 while (p < (fw->data + fw->size)) {
3715 fsec = (struct flash_section_info *)p;
3716 if (!memcmp(flash_cookie, fsec->cookie, sizeof(flash_cookie)))
3723 static int be_check_flash_crc(struct be_adapter *adapter, const u8 *p,
3724 u32 img_offset, u32 img_size, int hdr_size,
3725 u16 img_optype, bool *crc_match)
3731 status = be_cmd_get_flash_crc(adapter, crc, img_optype, img_size - 4);
3735 crc_offset = hdr_size + img_offset + img_size - 4;
3737 /* Skip flashing, if crc of flashed region matches */
3738 if (!memcmp(crc, p + crc_offset, 4))
3746 static int be_flash(struct be_adapter *adapter, const u8 *img,
3747 struct be_dma_mem *flash_cmd, int optype, int img_size)
3749 struct be_cmd_write_flashrom *req = flash_cmd->va;
3750 u32 total_bytes, flash_op, num_bytes;
3753 total_bytes = img_size;
3754 while (total_bytes) {
3755 num_bytes = min_t(u32, 32*1024, total_bytes);
3757 total_bytes -= num_bytes;
3760 if (optype == OPTYPE_PHY_FW)
3761 flash_op = FLASHROM_OPER_PHY_FLASH;
3763 flash_op = FLASHROM_OPER_FLASH;
3765 if (optype == OPTYPE_PHY_FW)
3766 flash_op = FLASHROM_OPER_PHY_SAVE;
3768 flash_op = FLASHROM_OPER_SAVE;
3771 memcpy(req->data_buf, img, num_bytes);
3773 status = be_cmd_write_flashrom(adapter, flash_cmd, optype,
3774 flash_op, num_bytes);
3775 if (base_status(status) == MCC_STATUS_ILLEGAL_REQUEST &&
3776 optype == OPTYPE_PHY_FW)
3784 /* For BE2, BE3 and BE3-R */
3785 static int be_flash_BEx(struct be_adapter *adapter,
3786 const struct firmware *fw,
3787 struct be_dma_mem *flash_cmd, int num_of_images)
3789 int img_hdrs_size = (num_of_images * sizeof(struct image_hdr));
3790 struct device *dev = &adapter->pdev->dev;
3791 struct flash_section_info *fsec = NULL;
3792 int status, i, filehdr_size, num_comp;
3793 const struct flash_comp *pflashcomp;
3797 struct flash_comp gen3_flash_types[] = {
3798 { FLASH_iSCSI_PRIMARY_IMAGE_START_g3, OPTYPE_ISCSI_ACTIVE,
3799 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_iSCSI},
3800 { FLASH_REDBOOT_START_g3, OPTYPE_REDBOOT,
3801 FLASH_REDBOOT_IMAGE_MAX_SIZE_g3, IMAGE_BOOT_CODE},
3802 { FLASH_iSCSI_BIOS_START_g3, OPTYPE_BIOS,
3803 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_ISCSI},
3804 { FLASH_PXE_BIOS_START_g3, OPTYPE_PXE_BIOS,
3805 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_PXE},
3806 { FLASH_FCoE_BIOS_START_g3, OPTYPE_FCOE_BIOS,
3807 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_FCoE},
3808 { FLASH_iSCSI_BACKUP_IMAGE_START_g3, OPTYPE_ISCSI_BACKUP,
3809 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_iSCSI},
3810 { FLASH_FCoE_PRIMARY_IMAGE_START_g3, OPTYPE_FCOE_FW_ACTIVE,
3811 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_FCoE},
3812 { FLASH_FCoE_BACKUP_IMAGE_START_g3, OPTYPE_FCOE_FW_BACKUP,
3813 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_FCoE},
3814 { FLASH_NCSI_START_g3, OPTYPE_NCSI_FW,
3815 FLASH_NCSI_IMAGE_MAX_SIZE_g3, IMAGE_NCSI},
3816 { FLASH_PHY_FW_START_g3, OPTYPE_PHY_FW,
3817 FLASH_PHY_FW_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_PHY}
3820 struct flash_comp gen2_flash_types[] = {
3821 { FLASH_iSCSI_PRIMARY_IMAGE_START_g2, OPTYPE_ISCSI_ACTIVE,
3822 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_iSCSI},
3823 { FLASH_REDBOOT_START_g2, OPTYPE_REDBOOT,
3824 FLASH_REDBOOT_IMAGE_MAX_SIZE_g2, IMAGE_BOOT_CODE},
3825 { FLASH_iSCSI_BIOS_START_g2, OPTYPE_BIOS,
3826 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_ISCSI},
3827 { FLASH_PXE_BIOS_START_g2, OPTYPE_PXE_BIOS,
3828 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_PXE},
3829 { FLASH_FCoE_BIOS_START_g2, OPTYPE_FCOE_BIOS,
3830 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_FCoE},
3831 { FLASH_iSCSI_BACKUP_IMAGE_START_g2, OPTYPE_ISCSI_BACKUP,
3832 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_iSCSI},
3833 { FLASH_FCoE_PRIMARY_IMAGE_START_g2, OPTYPE_FCOE_FW_ACTIVE,
3834 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_FCoE},
3835 { FLASH_FCoE_BACKUP_IMAGE_START_g2, OPTYPE_FCOE_FW_BACKUP,
3836 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_FCoE}
3839 if (BE3_chip(adapter)) {
3840 pflashcomp = gen3_flash_types;
3841 filehdr_size = sizeof(struct flash_file_hdr_g3);
3842 num_comp = ARRAY_SIZE(gen3_flash_types);
3844 pflashcomp = gen2_flash_types;
3845 filehdr_size = sizeof(struct flash_file_hdr_g2);
3846 num_comp = ARRAY_SIZE(gen2_flash_types);
3849 /* Get flash section info*/
3850 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
3852 dev_err(dev, "Invalid Cookie. FW image may be corrupted\n");
3855 for (i = 0; i < num_comp; i++) {
3856 if (!is_comp_in_ufi(adapter, fsec, pflashcomp[i].img_type))
3859 if ((pflashcomp[i].optype == OPTYPE_NCSI_FW) &&
3860 memcmp(adapter->fw_ver, "3.102.148.0", 11) < 0)
3863 if (pflashcomp[i].optype == OPTYPE_PHY_FW &&
3864 !phy_flashing_required(adapter))
3867 if (pflashcomp[i].optype == OPTYPE_REDBOOT) {
3868 status = be_check_flash_crc(adapter, fw->data,
3869 pflashcomp[i].offset,
3873 OPTYPE_REDBOOT, &crc_match);
3876 "Could not get CRC for 0x%x region\n",
3877 pflashcomp[i].optype);
3885 p = fw->data + filehdr_size + pflashcomp[i].offset +
3887 if (p + pflashcomp[i].size > fw->data + fw->size)
3890 status = be_flash(adapter, p, flash_cmd, pflashcomp[i].optype,
3891 pflashcomp[i].size);
3893 dev_err(dev, "Flashing section type 0x%x failed\n",
3894 pflashcomp[i].img_type);
3901 static u16 be_get_img_optype(struct flash_section_entry fsec_entry)
3903 u32 img_type = le32_to_cpu(fsec_entry.type);
3904 u16 img_optype = le16_to_cpu(fsec_entry.optype);
3906 if (img_optype != 0xFFFF)
3910 case IMAGE_FIRMWARE_iSCSI:
3911 img_optype = OPTYPE_ISCSI_ACTIVE;
3913 case IMAGE_BOOT_CODE:
3914 img_optype = OPTYPE_REDBOOT;
3916 case IMAGE_OPTION_ROM_ISCSI:
3917 img_optype = OPTYPE_BIOS;
3919 case IMAGE_OPTION_ROM_PXE:
3920 img_optype = OPTYPE_PXE_BIOS;
3922 case IMAGE_OPTION_ROM_FCoE:
3923 img_optype = OPTYPE_FCOE_BIOS;
3925 case IMAGE_FIRMWARE_BACKUP_iSCSI:
3926 img_optype = OPTYPE_ISCSI_BACKUP;
3929 img_optype = OPTYPE_NCSI_FW;
3931 case IMAGE_FLASHISM_JUMPVECTOR:
3932 img_optype = OPTYPE_FLASHISM_JUMPVECTOR;
3934 case IMAGE_FIRMWARE_PHY:
3935 img_optype = OPTYPE_SH_PHY_FW;
3937 case IMAGE_REDBOOT_DIR:
3938 img_optype = OPTYPE_REDBOOT_DIR;
3940 case IMAGE_REDBOOT_CONFIG:
3941 img_optype = OPTYPE_REDBOOT_CONFIG;
3944 img_optype = OPTYPE_UFI_DIR;
3953 static int be_flash_skyhawk(struct be_adapter *adapter,
3954 const struct firmware *fw,
3955 struct be_dma_mem *flash_cmd, int num_of_images)
3957 int img_hdrs_size = num_of_images * sizeof(struct image_hdr);
3958 struct device *dev = &adapter->pdev->dev;
3959 struct flash_section_info *fsec = NULL;
3960 u32 img_offset, img_size, img_type;
3961 int status, i, filehdr_size;
3962 bool crc_match, old_fw_img;
3966 filehdr_size = sizeof(struct flash_file_hdr_g3);
3967 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
3969 dev_err(dev, "Invalid Cookie. FW image may be corrupted\n");
3973 for (i = 0; i < le32_to_cpu(fsec->fsec_hdr.num_images); i++) {
3974 img_offset = le32_to_cpu(fsec->fsec_entry[i].offset);
3975 img_size = le32_to_cpu(fsec->fsec_entry[i].pad_size);
3976 img_type = le32_to_cpu(fsec->fsec_entry[i].type);
3977 img_optype = be_get_img_optype(fsec->fsec_entry[i]);
3978 old_fw_img = fsec->fsec_entry[i].optype == 0xFFFF;
3980 if (img_optype == 0xFFFF)
3982 /* Don't bother verifying CRC if an old FW image is being
3988 status = be_check_flash_crc(adapter, fw->data, img_offset,
3989 img_size, filehdr_size +
3990 img_hdrs_size, img_optype,
3992 /* The current FW image on the card does not recognize the new
3993 * FLASH op_type. The FW download is partially complete.
3994 * Reboot the server now to enable FW image to recognize the
3995 * new FLASH op_type. To complete the remaining process,
3996 * download the same FW again after the reboot.
3998 if (base_status(status) == MCC_STATUS_ILLEGAL_REQUEST ||
3999 base_status(status) == MCC_STATUS_ILLEGAL_FIELD) {
4000 dev_err(dev, "Flash incomplete. Reset the server\n");
4001 dev_err(dev, "Download FW image again after reset\n");
4003 } else if (status) {
4004 dev_err(dev, "Could not get CRC for 0x%x region\n",
4013 p = fw->data + filehdr_size + img_offset + img_hdrs_size;
4014 if (p + img_size > fw->data + fw->size)
4017 status = be_flash(adapter, p, flash_cmd, img_optype, img_size);
4018 /* For old FW images ignore ILLEGAL_FIELD error or errors on
4022 (base_status(status) == MCC_STATUS_ILLEGAL_FIELD ||
4023 (img_optype == OPTYPE_UFI_DIR &&
4024 base_status(status) == MCC_STATUS_FAILED))) {
4026 } else if (status) {
4027 dev_err(dev, "Flashing section type 0x%x failed\n",
4035 static int lancer_fw_download(struct be_adapter *adapter,
4036 const struct firmware *fw)
4038 #define LANCER_FW_DOWNLOAD_CHUNK (32 * 1024)
4039 #define LANCER_FW_DOWNLOAD_LOCATION "/prg"
4040 struct be_dma_mem flash_cmd;
4041 const u8 *data_ptr = NULL;
4042 u8 *dest_image_ptr = NULL;
4043 size_t image_size = 0;
4045 u32 data_written = 0;
4051 if (!IS_ALIGNED(fw->size, sizeof(u32))) {
4052 dev_err(&adapter->pdev->dev,
4053 "FW Image not properly aligned. "
4054 "Length must be 4 byte aligned.\n");
4056 goto lancer_fw_exit;
4059 flash_cmd.size = sizeof(struct lancer_cmd_req_write_object)
4060 + LANCER_FW_DOWNLOAD_CHUNK;
4061 flash_cmd.va = dma_alloc_coherent(&adapter->pdev->dev, flash_cmd.size,
4062 &flash_cmd.dma, GFP_KERNEL);
4063 if (!flash_cmd.va) {
4065 goto lancer_fw_exit;
4068 dest_image_ptr = flash_cmd.va +
4069 sizeof(struct lancer_cmd_req_write_object);
4070 image_size = fw->size;
4071 data_ptr = fw->data;
4073 while (image_size) {
4074 chunk_size = min_t(u32, image_size, LANCER_FW_DOWNLOAD_CHUNK);
4076 /* Copy the image chunk content. */
4077 memcpy(dest_image_ptr, data_ptr, chunk_size);
4079 status = lancer_cmd_write_object(adapter, &flash_cmd,
4081 LANCER_FW_DOWNLOAD_LOCATION,
4082 &data_written, &change_status,
4087 offset += data_written;
4088 data_ptr += data_written;
4089 image_size -= data_written;
4093 /* Commit the FW written */
4094 status = lancer_cmd_write_object(adapter, &flash_cmd,
4096 LANCER_FW_DOWNLOAD_LOCATION,
4097 &data_written, &change_status,
4101 dma_free_coherent(&adapter->pdev->dev, flash_cmd.size, flash_cmd.va,
4104 dev_err(&adapter->pdev->dev,
4105 "Firmware load error. "
4106 "Status code: 0x%x Additional Status: 0x%x\n",
4107 status, add_status);
4108 goto lancer_fw_exit;
4111 if (change_status == LANCER_FW_RESET_NEEDED) {
4112 dev_info(&adapter->pdev->dev,
4113 "Resetting adapter to activate new FW\n");
4114 status = lancer_physdev_ctrl(adapter,
4115 PHYSDEV_CONTROL_FW_RESET_MASK);
4117 dev_err(&adapter->pdev->dev,
4118 "Adapter busy for FW reset.\n"
4119 "New FW will not be active.\n");
4120 goto lancer_fw_exit;
4122 } else if (change_status != LANCER_NO_RESET_NEEDED) {
4123 dev_err(&adapter->pdev->dev,
4124 "System reboot required for new FW to be active\n");
4127 dev_info(&adapter->pdev->dev, "Firmware flashed successfully\n");
4134 #define UFI_TYPE3R 10
4136 static int be_get_ufi_type(struct be_adapter *adapter,
4137 struct flash_file_hdr_g3 *fhdr)
4140 goto be_get_ufi_exit;
4142 if (skyhawk_chip(adapter) && fhdr->build[0] == '4')
4144 else if (BE3_chip(adapter) && fhdr->build[0] == '3') {
4145 if (fhdr->asic_type_rev == 0x10)
4149 } else if (BE2_chip(adapter) && fhdr->build[0] == '2')
4153 dev_err(&adapter->pdev->dev,
4154 "UFI and Interface are not compatible for flashing\n");
4158 static int be_fw_download(struct be_adapter *adapter, const struct firmware* fw)
4160 struct flash_file_hdr_g3 *fhdr3;
4161 struct image_hdr *img_hdr_ptr = NULL;
4162 struct be_dma_mem flash_cmd;
4164 int status = 0, i = 0, num_imgs = 0, ufi_type = 0;
4166 flash_cmd.size = sizeof(struct be_cmd_write_flashrom);
4167 flash_cmd.va = dma_alloc_coherent(&adapter->pdev->dev, flash_cmd.size,
4168 &flash_cmd.dma, GFP_KERNEL);
4169 if (!flash_cmd.va) {
4175 fhdr3 = (struct flash_file_hdr_g3 *)p;
4177 ufi_type = be_get_ufi_type(adapter, fhdr3);
4179 num_imgs = le32_to_cpu(fhdr3->num_imgs);
4180 for (i = 0; i < num_imgs; i++) {
4181 img_hdr_ptr = (struct image_hdr *)(fw->data +
4182 (sizeof(struct flash_file_hdr_g3) +
4183 i * sizeof(struct image_hdr)));
4184 if (le32_to_cpu(img_hdr_ptr->imageid) == 1) {
4187 status = be_flash_skyhawk(adapter, fw,
4188 &flash_cmd, num_imgs);
4191 status = be_flash_BEx(adapter, fw, &flash_cmd,
4195 /* Do not flash this ufi on BE3-R cards */
4196 if (adapter->asic_rev < 0x10)
4197 status = be_flash_BEx(adapter, fw,
4202 dev_err(&adapter->pdev->dev,
4203 "Can't load BE3 UFI on BE3R\n");
4209 if (ufi_type == UFI_TYPE2)
4210 status = be_flash_BEx(adapter, fw, &flash_cmd, 0);
4211 else if (ufi_type == -1)
4214 dma_free_coherent(&adapter->pdev->dev, flash_cmd.size, flash_cmd.va,
4217 dev_err(&adapter->pdev->dev, "Firmware load error\n");
4221 dev_info(&adapter->pdev->dev, "Firmware flashed successfully\n");
4227 int be_load_fw(struct be_adapter *adapter, u8 *fw_file)
4229 const struct firmware *fw;
4232 if (!netif_running(adapter->netdev)) {
4233 dev_err(&adapter->pdev->dev,
4234 "Firmware load not allowed (interface is down)\n");
4238 status = request_firmware(&fw, fw_file, &adapter->pdev->dev);
4242 dev_info(&adapter->pdev->dev, "Flashing firmware file %s\n", fw_file);
4244 if (lancer_chip(adapter))
4245 status = lancer_fw_download(adapter, fw);
4247 status = be_fw_download(adapter, fw);
4250 be_cmd_get_fw_ver(adapter, adapter->fw_ver,
4251 adapter->fw_on_flash);
4254 release_firmware(fw);
4258 static int be_ndo_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh)
4260 struct be_adapter *adapter = netdev_priv(dev);
4261 struct nlattr *attr, *br_spec;
4266 if (!sriov_enabled(adapter))
4269 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4271 nla_for_each_nested(attr, br_spec, rem) {
4272 if (nla_type(attr) != IFLA_BRIDGE_MODE)
4275 mode = nla_get_u16(attr);
4276 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
4279 status = be_cmd_set_hsw_config(adapter, 0, 0,
4281 mode == BRIDGE_MODE_VEPA ?
4282 PORT_FWD_TYPE_VEPA :
4287 dev_info(&adapter->pdev->dev, "enabled switch mode: %s\n",
4288 mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
4293 dev_err(&adapter->pdev->dev, "Failed to set switch mode %s\n",
4294 mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
4299 static int be_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4300 struct net_device *dev, u32 filter_mask)
4302 struct be_adapter *adapter = netdev_priv(dev);
4306 if (!sriov_enabled(adapter))
4309 /* BE and Lancer chips support VEB mode only */
4310 if (BEx_chip(adapter) || lancer_chip(adapter)) {
4311 hsw_mode = PORT_FWD_TYPE_VEB;
4313 status = be_cmd_get_hsw_config(adapter, NULL, 0,
4314 adapter->if_handle, &hsw_mode);
4319 return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
4320 hsw_mode == PORT_FWD_TYPE_VEPA ?
4321 BRIDGE_MODE_VEPA : BRIDGE_MODE_VEB);
4324 #ifdef CONFIG_BE2NET_VXLAN
4325 static void be_add_vxlan_port(struct net_device *netdev, sa_family_t sa_family,
4328 struct be_adapter *adapter = netdev_priv(netdev);
4329 struct device *dev = &adapter->pdev->dev;
4332 if (lancer_chip(adapter) || BEx_chip(adapter))
4335 if (adapter->flags & BE_FLAGS_VXLAN_OFFLOADS) {
4336 dev_warn(dev, "Cannot add UDP port %d for VxLAN offloads\n",
4339 "Only one UDP port supported for VxLAN offloads\n");
4343 status = be_cmd_manage_iface(adapter, adapter->if_handle,
4344 OP_CONVERT_NORMAL_TO_TUNNEL);
4346 dev_warn(dev, "Failed to convert normal interface to tunnel\n");
4350 status = be_cmd_set_vxlan_port(adapter, port);
4352 dev_warn(dev, "Failed to add VxLAN port\n");
4355 adapter->flags |= BE_FLAGS_VXLAN_OFFLOADS;
4356 adapter->vxlan_port = port;
4358 dev_info(dev, "Enabled VxLAN offloads for UDP port %d\n",
4362 be_disable_vxlan_offloads(adapter);
4366 static void be_del_vxlan_port(struct net_device *netdev, sa_family_t sa_family,
4369 struct be_adapter *adapter = netdev_priv(netdev);
4371 if (lancer_chip(adapter) || BEx_chip(adapter))
4374 if (adapter->vxlan_port != port)
4377 be_disable_vxlan_offloads(adapter);
4379 dev_info(&adapter->pdev->dev,
4380 "Disabled VxLAN offloads for UDP port %d\n",
4385 static const struct net_device_ops be_netdev_ops = {
4386 .ndo_open = be_open,
4387 .ndo_stop = be_close,
4388 .ndo_start_xmit = be_xmit,
4389 .ndo_set_rx_mode = be_set_rx_mode,
4390 .ndo_set_mac_address = be_mac_addr_set,
4391 .ndo_change_mtu = be_change_mtu,
4392 .ndo_get_stats64 = be_get_stats64,
4393 .ndo_validate_addr = eth_validate_addr,
4394 .ndo_vlan_rx_add_vid = be_vlan_add_vid,
4395 .ndo_vlan_rx_kill_vid = be_vlan_rem_vid,
4396 .ndo_set_vf_mac = be_set_vf_mac,
4397 .ndo_set_vf_vlan = be_set_vf_vlan,
4398 .ndo_set_vf_rate = be_set_vf_tx_rate,
4399 .ndo_get_vf_config = be_get_vf_config,
4400 .ndo_set_vf_link_state = be_set_vf_link_state,
4401 #ifdef CONFIG_NET_POLL_CONTROLLER
4402 .ndo_poll_controller = be_netpoll,
4404 .ndo_bridge_setlink = be_ndo_bridge_setlink,
4405 .ndo_bridge_getlink = be_ndo_bridge_getlink,
4406 #ifdef CONFIG_NET_RX_BUSY_POLL
4407 .ndo_busy_poll = be_busy_poll,
4409 #ifdef CONFIG_BE2NET_VXLAN
4410 .ndo_add_vxlan_port = be_add_vxlan_port,
4411 .ndo_del_vxlan_port = be_del_vxlan_port,
4415 static void be_netdev_init(struct net_device *netdev)
4417 struct be_adapter *adapter = netdev_priv(netdev);
4419 if (skyhawk_chip(adapter)) {
4420 netdev->hw_enc_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
4421 NETIF_F_TSO | NETIF_F_TSO6 |
4422 NETIF_F_GSO_UDP_TUNNEL;
4423 netdev->hw_features |= NETIF_F_GSO_UDP_TUNNEL;
4425 netdev->hw_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
4426 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM | NETIF_F_RXCSUM |
4427 NETIF_F_HW_VLAN_CTAG_TX;
4428 if (be_multi_rxq(adapter))
4429 netdev->hw_features |= NETIF_F_RXHASH;
4431 netdev->features |= netdev->hw_features |
4432 NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_FILTER;
4434 netdev->vlan_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
4435 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
4437 netdev->priv_flags |= IFF_UNICAST_FLT;
4439 netdev->flags |= IFF_MULTICAST;
4441 netif_set_gso_max_size(netdev, 65535 - ETH_HLEN);
4443 netdev->netdev_ops = &be_netdev_ops;
4445 netdev->ethtool_ops = &be_ethtool_ops;
4448 static void be_unmap_pci_bars(struct be_adapter *adapter)
4451 pci_iounmap(adapter->pdev, adapter->csr);
4453 pci_iounmap(adapter->pdev, adapter->db);
4456 static int db_bar(struct be_adapter *adapter)
4458 if (lancer_chip(adapter) || !be_physfn(adapter))
4464 static int be_roce_map_pci_bars(struct be_adapter *adapter)
4466 if (skyhawk_chip(adapter)) {
4467 adapter->roce_db.size = 4096;
4468 adapter->roce_db.io_addr = pci_resource_start(adapter->pdev,
4470 adapter->roce_db.total_size = pci_resource_len(adapter->pdev,
4476 static int be_map_pci_bars(struct be_adapter *adapter)
4480 if (BEx_chip(adapter) && be_physfn(adapter)) {
4481 adapter->csr = pci_iomap(adapter->pdev, 2, 0);
4482 if (adapter->csr == NULL)
4486 addr = pci_iomap(adapter->pdev, db_bar(adapter), 0);
4491 be_roce_map_pci_bars(adapter);
4495 be_unmap_pci_bars(adapter);
4499 static void be_ctrl_cleanup(struct be_adapter *adapter)
4501 struct be_dma_mem *mem = &adapter->mbox_mem_alloced;
4503 be_unmap_pci_bars(adapter);
4506 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
4509 mem = &adapter->rx_filter;
4511 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
4515 static int be_ctrl_init(struct be_adapter *adapter)
4517 struct be_dma_mem *mbox_mem_alloc = &adapter->mbox_mem_alloced;
4518 struct be_dma_mem *mbox_mem_align = &adapter->mbox_mem;
4519 struct be_dma_mem *rx_filter = &adapter->rx_filter;
4523 pci_read_config_dword(adapter->pdev, SLI_INTF_REG_OFFSET, &sli_intf);
4524 adapter->sli_family = (sli_intf & SLI_INTF_FAMILY_MASK) >>
4525 SLI_INTF_FAMILY_SHIFT;
4526 adapter->virtfn = (sli_intf & SLI_INTF_FT_MASK) ? 1 : 0;
4528 status = be_map_pci_bars(adapter);
4532 mbox_mem_alloc->size = sizeof(struct be_mcc_mailbox) + 16;
4533 mbox_mem_alloc->va = dma_alloc_coherent(&adapter->pdev->dev,
4534 mbox_mem_alloc->size,
4535 &mbox_mem_alloc->dma,
4537 if (!mbox_mem_alloc->va) {
4539 goto unmap_pci_bars;
4541 mbox_mem_align->size = sizeof(struct be_mcc_mailbox);
4542 mbox_mem_align->va = PTR_ALIGN(mbox_mem_alloc->va, 16);
4543 mbox_mem_align->dma = PTR_ALIGN(mbox_mem_alloc->dma, 16);
4544 memset(mbox_mem_align->va, 0, sizeof(struct be_mcc_mailbox));
4546 rx_filter->size = sizeof(struct be_cmd_req_rx_filter);
4547 rx_filter->va = dma_zalloc_coherent(&adapter->pdev->dev,
4548 rx_filter->size, &rx_filter->dma,
4550 if (rx_filter->va == NULL) {
4555 mutex_init(&adapter->mbox_lock);
4556 spin_lock_init(&adapter->mcc_lock);
4557 spin_lock_init(&adapter->mcc_cq_lock);
4559 init_completion(&adapter->et_cmd_compl);
4560 pci_save_state(adapter->pdev);
4564 dma_free_coherent(&adapter->pdev->dev, mbox_mem_alloc->size,
4565 mbox_mem_alloc->va, mbox_mem_alloc->dma);
4568 be_unmap_pci_bars(adapter);
4574 static void be_stats_cleanup(struct be_adapter *adapter)
4576 struct be_dma_mem *cmd = &adapter->stats_cmd;
4579 dma_free_coherent(&adapter->pdev->dev, cmd->size,
4583 static int be_stats_init(struct be_adapter *adapter)
4585 struct be_dma_mem *cmd = &adapter->stats_cmd;
4587 if (lancer_chip(adapter))
4588 cmd->size = sizeof(struct lancer_cmd_req_pport_stats);
4589 else if (BE2_chip(adapter))
4590 cmd->size = sizeof(struct be_cmd_req_get_stats_v0);
4591 else if (BE3_chip(adapter))
4592 cmd->size = sizeof(struct be_cmd_req_get_stats_v1);
4594 /* ALL non-BE ASICs */
4595 cmd->size = sizeof(struct be_cmd_req_get_stats_v2);
4597 cmd->va = dma_zalloc_coherent(&adapter->pdev->dev, cmd->size, &cmd->dma,
4599 if (cmd->va == NULL)
4604 static void be_remove(struct pci_dev *pdev)
4606 struct be_adapter *adapter = pci_get_drvdata(pdev);
4611 be_roce_dev_remove(adapter);
4612 be_intr_set(adapter, false);
4614 cancel_delayed_work_sync(&adapter->func_recovery_work);
4616 unregister_netdev(adapter->netdev);
4620 /* tell fw we're done with firing cmds */
4621 be_cmd_fw_clean(adapter);
4623 be_stats_cleanup(adapter);
4625 be_ctrl_cleanup(adapter);
4627 pci_disable_pcie_error_reporting(pdev);
4629 pci_release_regions(pdev);
4630 pci_disable_device(pdev);
4632 free_netdev(adapter->netdev);
4635 static int be_get_initial_config(struct be_adapter *adapter)
4639 status = be_cmd_get_cntl_attributes(adapter);
4643 /* Must be a power of 2 or else MODULO will BUG_ON */
4644 adapter->be_get_temp_freq = 64;
4646 if (BEx_chip(adapter)) {
4647 level = be_cmd_get_fw_log_level(adapter);
4648 adapter->msg_enable =
4649 level <= FW_LOG_LEVEL_DEFAULT ? NETIF_MSG_HW : 0;
4652 adapter->cfg_num_qs = netif_get_num_default_rss_queues();
4656 static int lancer_recover_func(struct be_adapter *adapter)
4658 struct device *dev = &adapter->pdev->dev;
4661 status = lancer_test_and_set_rdy_state(adapter);
4665 if (netif_running(adapter->netdev))
4666 be_close(adapter->netdev);
4670 be_clear_all_error(adapter);
4672 status = be_setup(adapter);
4676 if (netif_running(adapter->netdev)) {
4677 status = be_open(adapter->netdev);
4682 dev_err(dev, "Adapter recovery successful\n");
4685 if (status == -EAGAIN)
4686 dev_err(dev, "Waiting for resource provisioning\n");
4688 dev_err(dev, "Adapter recovery failed\n");
4693 static void be_func_recovery_task(struct work_struct *work)
4695 struct be_adapter *adapter =
4696 container_of(work, struct be_adapter, func_recovery_work.work);
4699 be_detect_error(adapter);
4701 if (adapter->hw_error && lancer_chip(adapter)) {
4704 netif_device_detach(adapter->netdev);
4707 status = lancer_recover_func(adapter);
4709 netif_device_attach(adapter->netdev);
4712 /* In Lancer, for all errors other than provisioning error (-EAGAIN),
4713 * no need to attempt further recovery.
4715 if (!status || status == -EAGAIN)
4716 schedule_delayed_work(&adapter->func_recovery_work,
4717 msecs_to_jiffies(1000));
4720 static void be_worker(struct work_struct *work)
4722 struct be_adapter *adapter =
4723 container_of(work, struct be_adapter, work.work);
4724 struct be_rx_obj *rxo;
4727 /* when interrupts are not yet enabled, just reap any pending
4728 * mcc completions */
4729 if (!netif_running(adapter->netdev)) {
4731 be_process_mcc(adapter);
4736 if (!adapter->stats_cmd_sent) {
4737 if (lancer_chip(adapter))
4738 lancer_cmd_get_pport_stats(adapter,
4739 &adapter->stats_cmd);
4741 be_cmd_get_stats(adapter, &adapter->stats_cmd);
4744 if (be_physfn(adapter) &&
4745 MODULO(adapter->work_counter, adapter->be_get_temp_freq) == 0)
4746 be_cmd_get_die_temperature(adapter);
4748 for_all_rx_queues(adapter, rxo, i) {
4749 /* Replenish RX-queues starved due to memory
4750 * allocation failures.
4752 if (rxo->rx_post_starved)
4753 be_post_rx_frags(rxo, GFP_KERNEL);
4756 be_eqd_update(adapter);
4759 adapter->work_counter++;
4760 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
4763 /* If any VFs are already enabled don't FLR the PF */
4764 static bool be_reset_required(struct be_adapter *adapter)
4766 return pci_num_vf(adapter->pdev) ? false : true;
4769 static char *mc_name(struct be_adapter *adapter)
4771 char *str = ""; /* default */
4773 switch (adapter->mc_type) {
4799 static inline char *func_name(struct be_adapter *adapter)
4801 return be_physfn(adapter) ? "PF" : "VF";
4804 static int be_probe(struct pci_dev *pdev, const struct pci_device_id *pdev_id)
4807 struct be_adapter *adapter;
4808 struct net_device *netdev;
4811 status = pci_enable_device(pdev);
4815 status = pci_request_regions(pdev, DRV_NAME);
4818 pci_set_master(pdev);
4820 netdev = alloc_etherdev_mqs(sizeof(*adapter), MAX_TX_QS, MAX_RX_QS);
4821 if (netdev == NULL) {
4825 adapter = netdev_priv(netdev);
4826 adapter->pdev = pdev;
4827 pci_set_drvdata(pdev, adapter);
4828 adapter->netdev = netdev;
4829 SET_NETDEV_DEV(netdev, &pdev->dev);
4831 status = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
4833 netdev->features |= NETIF_F_HIGHDMA;
4835 status = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
4837 dev_err(&pdev->dev, "Could not set PCI DMA Mask\n");
4842 if (be_physfn(adapter)) {
4843 status = pci_enable_pcie_error_reporting(pdev);
4845 dev_info(&pdev->dev, "PCIe error reporting enabled\n");
4848 status = be_ctrl_init(adapter);
4852 /* sync up with fw's ready state */
4853 if (be_physfn(adapter)) {
4854 status = be_fw_wait_ready(adapter);
4859 if (be_reset_required(adapter)) {
4860 status = be_cmd_reset_function(adapter);
4864 /* Wait for interrupts to quiesce after an FLR */
4868 /* Allow interrupts for other ULPs running on NIC function */
4869 be_intr_set(adapter, true);
4871 /* tell fw we're ready to fire cmds */
4872 status = be_cmd_fw_init(adapter);
4876 status = be_stats_init(adapter);
4880 status = be_get_initial_config(adapter);
4884 INIT_DELAYED_WORK(&adapter->work, be_worker);
4885 INIT_DELAYED_WORK(&adapter->func_recovery_work, be_func_recovery_task);
4886 adapter->rx_fc = adapter->tx_fc = true;
4888 status = be_setup(adapter);
4892 be_netdev_init(netdev);
4893 status = register_netdev(netdev);
4897 be_roce_dev_add(adapter);
4899 schedule_delayed_work(&adapter->func_recovery_work,
4900 msecs_to_jiffies(1000));
4902 be_cmd_query_port_name(adapter, &port_name);
4904 dev_info(&pdev->dev, "%s: %s %s port %c\n", nic_name(pdev),
4905 func_name(adapter), mc_name(adapter), port_name);
4912 be_stats_cleanup(adapter);
4914 be_ctrl_cleanup(adapter);
4916 free_netdev(netdev);
4918 pci_release_regions(pdev);
4920 pci_disable_device(pdev);
4922 dev_err(&pdev->dev, "%s initialization failed\n", nic_name(pdev));
4926 static int be_suspend(struct pci_dev *pdev, pm_message_t state)
4928 struct be_adapter *adapter = pci_get_drvdata(pdev);
4929 struct net_device *netdev = adapter->netdev;
4931 if (adapter->wol_en)
4932 be_setup_wol(adapter, true);
4934 be_intr_set(adapter, false);
4935 cancel_delayed_work_sync(&adapter->func_recovery_work);
4937 netif_device_detach(netdev);
4938 if (netif_running(netdev)) {
4945 pci_save_state(pdev);
4946 pci_disable_device(pdev);
4947 pci_set_power_state(pdev, pci_choose_state(pdev, state));
4951 static int be_resume(struct pci_dev *pdev)
4954 struct be_adapter *adapter = pci_get_drvdata(pdev);
4955 struct net_device *netdev = adapter->netdev;
4957 netif_device_detach(netdev);
4959 status = pci_enable_device(pdev);
4963 pci_set_power_state(pdev, PCI_D0);
4964 pci_restore_state(pdev);
4966 status = be_fw_wait_ready(adapter);
4970 be_intr_set(adapter, true);
4971 /* tell fw we're ready to fire cmds */
4972 status = be_cmd_fw_init(adapter);
4977 if (netif_running(netdev)) {
4983 schedule_delayed_work(&adapter->func_recovery_work,
4984 msecs_to_jiffies(1000));
4985 netif_device_attach(netdev);
4987 if (adapter->wol_en)
4988 be_setup_wol(adapter, false);
4994 * An FLR will stop BE from DMAing any data.
4996 static void be_shutdown(struct pci_dev *pdev)
4998 struct be_adapter *adapter = pci_get_drvdata(pdev);
5003 cancel_delayed_work_sync(&adapter->work);
5004 cancel_delayed_work_sync(&adapter->func_recovery_work);
5006 netif_device_detach(adapter->netdev);
5008 be_cmd_reset_function(adapter);
5010 pci_disable_device(pdev);
5013 static pci_ers_result_t be_eeh_err_detected(struct pci_dev *pdev,
5014 pci_channel_state_t state)
5016 struct be_adapter *adapter = pci_get_drvdata(pdev);
5017 struct net_device *netdev = adapter->netdev;
5019 dev_err(&adapter->pdev->dev, "EEH error detected\n");
5021 if (!adapter->eeh_error) {
5022 adapter->eeh_error = true;
5024 cancel_delayed_work_sync(&adapter->func_recovery_work);
5027 netif_device_detach(netdev);
5028 if (netif_running(netdev))
5035 if (state == pci_channel_io_perm_failure)
5036 return PCI_ERS_RESULT_DISCONNECT;
5038 pci_disable_device(pdev);
5040 /* The error could cause the FW to trigger a flash debug dump.
5041 * Resetting the card while flash dump is in progress
5042 * can cause it not to recover; wait for it to finish.
5043 * Wait only for first function as it is needed only once per
5046 if (pdev->devfn == 0)
5049 return PCI_ERS_RESULT_NEED_RESET;
5052 static pci_ers_result_t be_eeh_reset(struct pci_dev *pdev)
5054 struct be_adapter *adapter = pci_get_drvdata(pdev);
5057 dev_info(&adapter->pdev->dev, "EEH reset\n");
5059 status = pci_enable_device(pdev);
5061 return PCI_ERS_RESULT_DISCONNECT;
5063 pci_set_master(pdev);
5064 pci_set_power_state(pdev, PCI_D0);
5065 pci_restore_state(pdev);
5067 /* Check if card is ok and fw is ready */
5068 dev_info(&adapter->pdev->dev,
5069 "Waiting for FW to be ready after EEH reset\n");
5070 status = be_fw_wait_ready(adapter);
5072 return PCI_ERS_RESULT_DISCONNECT;
5074 pci_cleanup_aer_uncorrect_error_status(pdev);
5075 be_clear_all_error(adapter);
5076 return PCI_ERS_RESULT_RECOVERED;
5079 static void be_eeh_resume(struct pci_dev *pdev)
5082 struct be_adapter *adapter = pci_get_drvdata(pdev);
5083 struct net_device *netdev = adapter->netdev;
5085 dev_info(&adapter->pdev->dev, "EEH resume\n");
5087 pci_save_state(pdev);
5089 status = be_cmd_reset_function(adapter);
5093 /* On some BE3 FW versions, after a HW reset,
5094 * interrupts will remain disabled for each function.
5095 * So, explicitly enable interrupts
5097 be_intr_set(adapter, true);
5099 /* tell fw we're ready to fire cmds */
5100 status = be_cmd_fw_init(adapter);
5104 status = be_setup(adapter);
5108 if (netif_running(netdev)) {
5109 status = be_open(netdev);
5114 schedule_delayed_work(&adapter->func_recovery_work,
5115 msecs_to_jiffies(1000));
5116 netif_device_attach(netdev);
5119 dev_err(&adapter->pdev->dev, "EEH resume failed\n");
5122 static const struct pci_error_handlers be_eeh_handlers = {
5123 .error_detected = be_eeh_err_detected,
5124 .slot_reset = be_eeh_reset,
5125 .resume = be_eeh_resume,
5128 static struct pci_driver be_driver = {
5130 .id_table = be_dev_ids,
5132 .remove = be_remove,
5133 .suspend = be_suspend,
5134 .resume = be_resume,
5135 .shutdown = be_shutdown,
5136 .err_handler = &be_eeh_handlers
5139 static int __init be_init_module(void)
5141 if (rx_frag_size != 8192 && rx_frag_size != 4096 &&
5142 rx_frag_size != 2048) {
5143 printk(KERN_WARNING DRV_NAME
5144 " : Module param rx_frag_size must be 2048/4096/8192."
5146 rx_frag_size = 2048;
5149 return pci_register_driver(&be_driver);
5151 module_init(be_init_module);
5153 static void __exit be_exit_module(void)
5155 pci_unregister_driver(&be_driver);
5157 module_exit(be_exit_module);