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) <<
214 DB_EQ_RING_ID_EXT_MASK_SHIFT);
216 if (adapter->eeh_error)
220 val |= 1 << DB_EQ_REARM_SHIFT;
222 val |= 1 << DB_EQ_CLR_SHIFT;
223 val |= 1 << DB_EQ_EVNT_SHIFT;
224 val |= num_popped << DB_EQ_NUM_POPPED_SHIFT;
225 iowrite32(val, adapter->db + DB_EQ_OFFSET);
228 void be_cq_notify(struct be_adapter *adapter, u16 qid, bool arm, u16 num_popped)
231 val |= qid & DB_CQ_RING_ID_MASK;
232 val |= ((qid & DB_CQ_RING_ID_EXT_MASK) <<
233 DB_CQ_RING_ID_EXT_MASK_SHIFT);
235 if (adapter->eeh_error)
239 val |= 1 << DB_CQ_REARM_SHIFT;
240 val |= num_popped << DB_CQ_NUM_POPPED_SHIFT;
241 iowrite32(val, adapter->db + DB_CQ_OFFSET);
244 static int be_mac_addr_set(struct net_device *netdev, void *p)
246 struct be_adapter *adapter = netdev_priv(netdev);
247 struct device *dev = &adapter->pdev->dev;
248 struct sockaddr *addr = p;
251 u32 old_pmac_id = adapter->pmac_id[0], curr_pmac_id = 0;
253 if (!is_valid_ether_addr(addr->sa_data))
254 return -EADDRNOTAVAIL;
256 /* Proceed further only if, User provided MAC is different
259 if (ether_addr_equal(addr->sa_data, netdev->dev_addr))
262 /* The PMAC_ADD cmd may fail if the VF doesn't have FILTMGMT
263 * privilege or if PF did not provision the new MAC address.
264 * On BE3, this cmd will always fail if the VF doesn't have the
265 * FILTMGMT privilege. This failure is OK, only if the PF programmed
266 * the MAC for the VF.
268 status = be_cmd_pmac_add(adapter, (u8 *)addr->sa_data,
269 adapter->if_handle, &adapter->pmac_id[0], 0);
271 curr_pmac_id = adapter->pmac_id[0];
273 /* Delete the old programmed MAC. This call may fail if the
274 * old MAC was already deleted by the PF driver.
276 if (adapter->pmac_id[0] != old_pmac_id)
277 be_cmd_pmac_del(adapter, adapter->if_handle,
281 /* Decide if the new MAC is successfully activated only after
284 status = be_cmd_get_active_mac(adapter, curr_pmac_id, mac,
285 adapter->if_handle, true, 0);
289 /* The MAC change did not happen, either due to lack of privilege
290 * or PF didn't pre-provision.
292 if (!ether_addr_equal(addr->sa_data, mac)) {
297 memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
298 dev_info(dev, "MAC address changed to %pM\n", mac);
301 dev_warn(dev, "MAC address change to %pM failed\n", addr->sa_data);
305 /* BE2 supports only v0 cmd */
306 static void *hw_stats_from_cmd(struct be_adapter *adapter)
308 if (BE2_chip(adapter)) {
309 struct be_cmd_resp_get_stats_v0 *cmd = adapter->stats_cmd.va;
311 return &cmd->hw_stats;
312 } else if (BE3_chip(adapter)) {
313 struct be_cmd_resp_get_stats_v1 *cmd = adapter->stats_cmd.va;
315 return &cmd->hw_stats;
317 struct be_cmd_resp_get_stats_v2 *cmd = adapter->stats_cmd.va;
319 return &cmd->hw_stats;
323 /* BE2 supports only v0 cmd */
324 static void *be_erx_stats_from_cmd(struct be_adapter *adapter)
326 if (BE2_chip(adapter)) {
327 struct be_hw_stats_v0 *hw_stats = hw_stats_from_cmd(adapter);
329 return &hw_stats->erx;
330 } else if (BE3_chip(adapter)) {
331 struct be_hw_stats_v1 *hw_stats = hw_stats_from_cmd(adapter);
333 return &hw_stats->erx;
335 struct be_hw_stats_v2 *hw_stats = hw_stats_from_cmd(adapter);
337 return &hw_stats->erx;
341 static void populate_be_v0_stats(struct be_adapter *adapter)
343 struct be_hw_stats_v0 *hw_stats = hw_stats_from_cmd(adapter);
344 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
345 struct be_rxf_stats_v0 *rxf_stats = &hw_stats->rxf;
346 struct be_port_rxf_stats_v0 *port_stats =
347 &rxf_stats->port[adapter->port_num];
348 struct be_drv_stats *drvs = &adapter->drv_stats;
350 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
351 drvs->rx_pause_frames = port_stats->rx_pause_frames;
352 drvs->rx_crc_errors = port_stats->rx_crc_errors;
353 drvs->rx_control_frames = port_stats->rx_control_frames;
354 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
355 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
356 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
357 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
358 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
359 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
360 drvs->rxpp_fifo_overflow_drop = port_stats->rx_fifo_overflow;
361 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
362 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
363 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
364 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
365 drvs->rx_input_fifo_overflow_drop = port_stats->rx_input_fifo_overflow;
366 drvs->rx_dropped_header_too_small =
367 port_stats->rx_dropped_header_too_small;
368 drvs->rx_address_filtered =
369 port_stats->rx_address_filtered +
370 port_stats->rx_vlan_filtered;
371 drvs->rx_alignment_symbol_errors =
372 port_stats->rx_alignment_symbol_errors;
374 drvs->tx_pauseframes = port_stats->tx_pauseframes;
375 drvs->tx_controlframes = port_stats->tx_controlframes;
377 if (adapter->port_num)
378 drvs->jabber_events = rxf_stats->port1_jabber_events;
380 drvs->jabber_events = rxf_stats->port0_jabber_events;
381 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
382 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
383 drvs->forwarded_packets = rxf_stats->forwarded_packets;
384 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
385 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
386 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
387 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
390 static void populate_be_v1_stats(struct be_adapter *adapter)
392 struct be_hw_stats_v1 *hw_stats = hw_stats_from_cmd(adapter);
393 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
394 struct be_rxf_stats_v1 *rxf_stats = &hw_stats->rxf;
395 struct be_port_rxf_stats_v1 *port_stats =
396 &rxf_stats->port[adapter->port_num];
397 struct be_drv_stats *drvs = &adapter->drv_stats;
399 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
400 drvs->pmem_fifo_overflow_drop = port_stats->pmem_fifo_overflow_drop;
401 drvs->rx_priority_pause_frames = port_stats->rx_priority_pause_frames;
402 drvs->rx_pause_frames = port_stats->rx_pause_frames;
403 drvs->rx_crc_errors = port_stats->rx_crc_errors;
404 drvs->rx_control_frames = port_stats->rx_control_frames;
405 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
406 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
407 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
408 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
409 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
410 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
411 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
412 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
413 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
414 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
415 drvs->rx_dropped_header_too_small =
416 port_stats->rx_dropped_header_too_small;
417 drvs->rx_input_fifo_overflow_drop =
418 port_stats->rx_input_fifo_overflow_drop;
419 drvs->rx_address_filtered = port_stats->rx_address_filtered;
420 drvs->rx_alignment_symbol_errors =
421 port_stats->rx_alignment_symbol_errors;
422 drvs->rxpp_fifo_overflow_drop = port_stats->rxpp_fifo_overflow_drop;
423 drvs->tx_pauseframes = port_stats->tx_pauseframes;
424 drvs->tx_controlframes = port_stats->tx_controlframes;
425 drvs->tx_priority_pauseframes = port_stats->tx_priority_pauseframes;
426 drvs->jabber_events = port_stats->jabber_events;
427 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
428 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
429 drvs->forwarded_packets = rxf_stats->forwarded_packets;
430 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
431 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
432 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
433 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
436 static void populate_be_v2_stats(struct be_adapter *adapter)
438 struct be_hw_stats_v2 *hw_stats = hw_stats_from_cmd(adapter);
439 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
440 struct be_rxf_stats_v2 *rxf_stats = &hw_stats->rxf;
441 struct be_port_rxf_stats_v2 *port_stats =
442 &rxf_stats->port[adapter->port_num];
443 struct be_drv_stats *drvs = &adapter->drv_stats;
445 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
446 drvs->pmem_fifo_overflow_drop = port_stats->pmem_fifo_overflow_drop;
447 drvs->rx_priority_pause_frames = port_stats->rx_priority_pause_frames;
448 drvs->rx_pause_frames = port_stats->rx_pause_frames;
449 drvs->rx_crc_errors = port_stats->rx_crc_errors;
450 drvs->rx_control_frames = port_stats->rx_control_frames;
451 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
452 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
453 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
454 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
455 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
456 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
457 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
458 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
459 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
460 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
461 drvs->rx_dropped_header_too_small =
462 port_stats->rx_dropped_header_too_small;
463 drvs->rx_input_fifo_overflow_drop =
464 port_stats->rx_input_fifo_overflow_drop;
465 drvs->rx_address_filtered = port_stats->rx_address_filtered;
466 drvs->rx_alignment_symbol_errors =
467 port_stats->rx_alignment_symbol_errors;
468 drvs->rxpp_fifo_overflow_drop = port_stats->rxpp_fifo_overflow_drop;
469 drvs->tx_pauseframes = port_stats->tx_pauseframes;
470 drvs->tx_controlframes = port_stats->tx_controlframes;
471 drvs->tx_priority_pauseframes = port_stats->tx_priority_pauseframes;
472 drvs->jabber_events = port_stats->jabber_events;
473 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
474 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
475 drvs->forwarded_packets = rxf_stats->forwarded_packets;
476 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
477 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
478 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
479 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
480 if (be_roce_supported(adapter)) {
481 drvs->rx_roce_bytes_lsd = port_stats->roce_bytes_received_lsd;
482 drvs->rx_roce_bytes_msd = port_stats->roce_bytes_received_msd;
483 drvs->rx_roce_frames = port_stats->roce_frames_received;
484 drvs->roce_drops_crc = port_stats->roce_drops_crc;
485 drvs->roce_drops_payload_len =
486 port_stats->roce_drops_payload_len;
490 static void populate_lancer_stats(struct be_adapter *adapter)
493 struct be_drv_stats *drvs = &adapter->drv_stats;
494 struct lancer_pport_stats *pport_stats =
495 pport_stats_from_cmd(adapter);
497 be_dws_le_to_cpu(pport_stats, sizeof(*pport_stats));
498 drvs->rx_pause_frames = pport_stats->rx_pause_frames_lo;
499 drvs->rx_crc_errors = pport_stats->rx_crc_errors_lo;
500 drvs->rx_control_frames = pport_stats->rx_control_frames_lo;
501 drvs->rx_in_range_errors = pport_stats->rx_in_range_errors;
502 drvs->rx_frame_too_long = pport_stats->rx_frames_too_long_lo;
503 drvs->rx_dropped_runt = pport_stats->rx_dropped_runt;
504 drvs->rx_ip_checksum_errs = pport_stats->rx_ip_checksum_errors;
505 drvs->rx_tcp_checksum_errs = pport_stats->rx_tcp_checksum_errors;
506 drvs->rx_udp_checksum_errs = pport_stats->rx_udp_checksum_errors;
507 drvs->rx_dropped_tcp_length =
508 pport_stats->rx_dropped_invalid_tcp_length;
509 drvs->rx_dropped_too_small = pport_stats->rx_dropped_too_small;
510 drvs->rx_dropped_too_short = pport_stats->rx_dropped_too_short;
511 drvs->rx_out_range_errors = pport_stats->rx_out_of_range_errors;
512 drvs->rx_dropped_header_too_small =
513 pport_stats->rx_dropped_header_too_small;
514 drvs->rx_input_fifo_overflow_drop = pport_stats->rx_fifo_overflow;
515 drvs->rx_address_filtered =
516 pport_stats->rx_address_filtered +
517 pport_stats->rx_vlan_filtered;
518 drvs->rx_alignment_symbol_errors = pport_stats->rx_symbol_errors_lo;
519 drvs->rxpp_fifo_overflow_drop = pport_stats->rx_fifo_overflow;
520 drvs->tx_pauseframes = pport_stats->tx_pause_frames_lo;
521 drvs->tx_controlframes = pport_stats->tx_control_frames_lo;
522 drvs->jabber_events = pport_stats->rx_jabbers;
523 drvs->forwarded_packets = pport_stats->num_forwards_lo;
524 drvs->rx_drops_mtu = pport_stats->rx_drops_mtu_lo;
525 drvs->rx_drops_too_many_frags =
526 pport_stats->rx_drops_too_many_frags_lo;
529 static void accumulate_16bit_val(u32 *acc, u16 val)
531 #define lo(x) (x & 0xFFFF)
532 #define hi(x) (x & 0xFFFF0000)
533 bool wrapped = val < lo(*acc);
534 u32 newacc = hi(*acc) + val;
538 ACCESS_ONCE(*acc) = newacc;
541 static void populate_erx_stats(struct be_adapter *adapter,
542 struct be_rx_obj *rxo,
545 if (!BEx_chip(adapter))
546 rx_stats(rxo)->rx_drops_no_frags = erx_stat;
548 /* below erx HW counter can actually wrap around after
549 * 65535. Driver accumulates a 32-bit value
551 accumulate_16bit_val(&rx_stats(rxo)->rx_drops_no_frags,
555 void be_parse_stats(struct be_adapter *adapter)
557 struct be_erx_stats_v2 *erx = be_erx_stats_from_cmd(adapter);
558 struct be_rx_obj *rxo;
562 if (lancer_chip(adapter)) {
563 populate_lancer_stats(adapter);
565 if (BE2_chip(adapter))
566 populate_be_v0_stats(adapter);
567 else if (BE3_chip(adapter))
569 populate_be_v1_stats(adapter);
571 populate_be_v2_stats(adapter);
573 /* erx_v2 is longer than v0, v1. use v2 for v0, v1 access */
574 for_all_rx_queues(adapter, rxo, i) {
575 erx_stat = erx->rx_drops_no_fragments[rxo->q.id];
576 populate_erx_stats(adapter, rxo, erx_stat);
581 static struct rtnl_link_stats64 *be_get_stats64(struct net_device *netdev,
582 struct rtnl_link_stats64 *stats)
584 struct be_adapter *adapter = netdev_priv(netdev);
585 struct be_drv_stats *drvs = &adapter->drv_stats;
586 struct be_rx_obj *rxo;
587 struct be_tx_obj *txo;
592 for_all_rx_queues(adapter, rxo, i) {
593 const struct be_rx_stats *rx_stats = rx_stats(rxo);
595 start = u64_stats_fetch_begin_irq(&rx_stats->sync);
596 pkts = rx_stats(rxo)->rx_pkts;
597 bytes = rx_stats(rxo)->rx_bytes;
598 } while (u64_stats_fetch_retry_irq(&rx_stats->sync, start));
599 stats->rx_packets += pkts;
600 stats->rx_bytes += bytes;
601 stats->multicast += rx_stats(rxo)->rx_mcast_pkts;
602 stats->rx_dropped += rx_stats(rxo)->rx_drops_no_skbs +
603 rx_stats(rxo)->rx_drops_no_frags;
606 for_all_tx_queues(adapter, txo, i) {
607 const struct be_tx_stats *tx_stats = tx_stats(txo);
609 start = u64_stats_fetch_begin_irq(&tx_stats->sync);
610 pkts = tx_stats(txo)->tx_pkts;
611 bytes = tx_stats(txo)->tx_bytes;
612 } while (u64_stats_fetch_retry_irq(&tx_stats->sync, start));
613 stats->tx_packets += pkts;
614 stats->tx_bytes += bytes;
617 /* bad pkts received */
618 stats->rx_errors = drvs->rx_crc_errors +
619 drvs->rx_alignment_symbol_errors +
620 drvs->rx_in_range_errors +
621 drvs->rx_out_range_errors +
622 drvs->rx_frame_too_long +
623 drvs->rx_dropped_too_small +
624 drvs->rx_dropped_too_short +
625 drvs->rx_dropped_header_too_small +
626 drvs->rx_dropped_tcp_length +
627 drvs->rx_dropped_runt;
629 /* detailed rx errors */
630 stats->rx_length_errors = drvs->rx_in_range_errors +
631 drvs->rx_out_range_errors +
632 drvs->rx_frame_too_long;
634 stats->rx_crc_errors = drvs->rx_crc_errors;
636 /* frame alignment errors */
637 stats->rx_frame_errors = drvs->rx_alignment_symbol_errors;
639 /* receiver fifo overrun */
640 /* drops_no_pbuf is no per i/f, it's per BE card */
641 stats->rx_fifo_errors = drvs->rxpp_fifo_overflow_drop +
642 drvs->rx_input_fifo_overflow_drop +
643 drvs->rx_drops_no_pbuf;
647 void be_link_status_update(struct be_adapter *adapter, u8 link_status)
649 struct net_device *netdev = adapter->netdev;
651 if (!(adapter->flags & BE_FLAGS_LINK_STATUS_INIT)) {
652 netif_carrier_off(netdev);
653 adapter->flags |= BE_FLAGS_LINK_STATUS_INIT;
657 netif_carrier_on(netdev);
659 netif_carrier_off(netdev);
662 static void be_tx_stats_update(struct be_tx_obj *txo,
663 u32 wrb_cnt, u32 copied, u32 gso_segs, bool stopped)
665 struct be_tx_stats *stats = tx_stats(txo);
667 u64_stats_update_begin(&stats->sync);
669 stats->tx_wrbs += wrb_cnt;
670 stats->tx_bytes += copied;
671 stats->tx_pkts += (gso_segs ? gso_segs : 1);
674 u64_stats_update_end(&stats->sync);
677 /* Determine number of WRB entries needed to xmit data in an skb */
678 static u32 wrb_cnt_for_skb(struct be_adapter *adapter, struct sk_buff *skb,
681 int cnt = (skb->len > skb->data_len);
683 cnt += skb_shinfo(skb)->nr_frags;
685 /* to account for hdr wrb */
687 if (lancer_chip(adapter) || !(cnt & 1)) {
690 /* add a dummy to make it an even num */
694 BUG_ON(cnt > BE_MAX_TX_FRAG_COUNT);
698 static inline void wrb_fill(struct be_eth_wrb *wrb, u64 addr, int len)
700 wrb->frag_pa_hi = upper_32_bits(addr);
701 wrb->frag_pa_lo = addr & 0xFFFFFFFF;
702 wrb->frag_len = len & ETH_WRB_FRAG_LEN_MASK;
706 static inline u16 be_get_tx_vlan_tag(struct be_adapter *adapter,
712 vlan_tag = vlan_tx_tag_get(skb);
713 vlan_prio = (vlan_tag & VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
714 /* If vlan priority provided by OS is NOT in available bmap */
715 if (!(adapter->vlan_prio_bmap & (1 << vlan_prio)))
716 vlan_tag = (vlan_tag & ~VLAN_PRIO_MASK) |
717 adapter->recommended_prio;
722 /* Used only for IP tunnel packets */
723 static u16 skb_inner_ip_proto(struct sk_buff *skb)
725 return (inner_ip_hdr(skb)->version == 4) ?
726 inner_ip_hdr(skb)->protocol : inner_ipv6_hdr(skb)->nexthdr;
729 static u16 skb_ip_proto(struct sk_buff *skb)
731 return (ip_hdr(skb)->version == 4) ?
732 ip_hdr(skb)->protocol : ipv6_hdr(skb)->nexthdr;
735 static void wrb_fill_hdr(struct be_adapter *adapter, struct be_eth_hdr_wrb *hdr,
736 struct sk_buff *skb, u32 wrb_cnt, u32 len, bool skip_hw_vlan)
740 memset(hdr, 0, sizeof(*hdr));
742 AMAP_SET_BITS(struct amap_eth_hdr_wrb, crc, hdr, 1);
744 if (skb_is_gso(skb)) {
745 AMAP_SET_BITS(struct amap_eth_hdr_wrb, lso, hdr, 1);
746 AMAP_SET_BITS(struct amap_eth_hdr_wrb, lso_mss,
747 hdr, skb_shinfo(skb)->gso_size);
748 if (skb_is_gso_v6(skb) && !lancer_chip(adapter))
749 AMAP_SET_BITS(struct amap_eth_hdr_wrb, lso6, hdr, 1);
750 } else if (skb->ip_summed == CHECKSUM_PARTIAL) {
751 if (skb->encapsulation) {
752 AMAP_SET_BITS(struct amap_eth_hdr_wrb, ipcs, hdr, 1);
753 proto = skb_inner_ip_proto(skb);
755 proto = skb_ip_proto(skb);
757 if (proto == IPPROTO_TCP)
758 AMAP_SET_BITS(struct amap_eth_hdr_wrb, tcpcs, hdr, 1);
759 else if (proto == IPPROTO_UDP)
760 AMAP_SET_BITS(struct amap_eth_hdr_wrb, udpcs, hdr, 1);
763 if (vlan_tx_tag_present(skb)) {
764 AMAP_SET_BITS(struct amap_eth_hdr_wrb, vlan, hdr, 1);
765 vlan_tag = be_get_tx_vlan_tag(adapter, skb);
766 AMAP_SET_BITS(struct amap_eth_hdr_wrb, vlan_tag, hdr, vlan_tag);
769 /* To skip HW VLAN tagging: evt = 1, compl = 0 */
770 AMAP_SET_BITS(struct amap_eth_hdr_wrb, complete, hdr, !skip_hw_vlan);
771 AMAP_SET_BITS(struct amap_eth_hdr_wrb, event, hdr, 1);
772 AMAP_SET_BITS(struct amap_eth_hdr_wrb, num_wrb, hdr, wrb_cnt);
773 AMAP_SET_BITS(struct amap_eth_hdr_wrb, len, hdr, len);
776 static void unmap_tx_frag(struct device *dev, struct be_eth_wrb *wrb,
781 be_dws_le_to_cpu(wrb, sizeof(*wrb));
783 dma = (u64)wrb->frag_pa_hi << 32 | (u64)wrb->frag_pa_lo;
786 dma_unmap_single(dev, dma, wrb->frag_len,
789 dma_unmap_page(dev, dma, wrb->frag_len, DMA_TO_DEVICE);
793 static int make_tx_wrbs(struct be_adapter *adapter, struct be_queue_info *txq,
794 struct sk_buff *skb, u32 wrb_cnt, bool dummy_wrb,
799 struct device *dev = &adapter->pdev->dev;
800 struct sk_buff *first_skb = skb;
801 struct be_eth_wrb *wrb;
802 struct be_eth_hdr_wrb *hdr;
803 bool map_single = false;
806 hdr = queue_head_node(txq);
808 map_head = txq->head;
810 if (skb->len > skb->data_len) {
811 int len = skb_headlen(skb);
812 busaddr = dma_map_single(dev, skb->data, len, DMA_TO_DEVICE);
813 if (dma_mapping_error(dev, busaddr))
816 wrb = queue_head_node(txq);
817 wrb_fill(wrb, busaddr, len);
818 be_dws_cpu_to_le(wrb, sizeof(*wrb));
823 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
824 const struct skb_frag_struct *frag =
825 &skb_shinfo(skb)->frags[i];
826 busaddr = skb_frag_dma_map(dev, frag, 0,
827 skb_frag_size(frag), DMA_TO_DEVICE);
828 if (dma_mapping_error(dev, busaddr))
830 wrb = queue_head_node(txq);
831 wrb_fill(wrb, busaddr, skb_frag_size(frag));
832 be_dws_cpu_to_le(wrb, sizeof(*wrb));
834 copied += skb_frag_size(frag);
838 wrb = queue_head_node(txq);
840 be_dws_cpu_to_le(wrb, sizeof(*wrb));
844 wrb_fill_hdr(adapter, hdr, first_skb, wrb_cnt, copied, skip_hw_vlan);
845 be_dws_cpu_to_le(hdr, sizeof(*hdr));
849 txq->head = map_head;
851 wrb = queue_head_node(txq);
852 unmap_tx_frag(dev, wrb, map_single);
854 copied -= wrb->frag_len;
860 static struct sk_buff *be_insert_vlan_in_pkt(struct be_adapter *adapter,
866 skb = skb_share_check(skb, GFP_ATOMIC);
870 if (vlan_tx_tag_present(skb))
871 vlan_tag = be_get_tx_vlan_tag(adapter, skb);
873 if (qnq_async_evt_rcvd(adapter) && adapter->pvid) {
875 vlan_tag = adapter->pvid;
876 /* f/w workaround to set skip_hw_vlan = 1, informs the F/W to
877 * skip VLAN insertion
880 *skip_hw_vlan = true;
884 skb = __vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_tag);
890 /* Insert the outer VLAN, if any */
891 if (adapter->qnq_vid) {
892 vlan_tag = adapter->qnq_vid;
893 skb = __vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_tag);
897 *skip_hw_vlan = true;
903 static bool be_ipv6_exthdr_check(struct sk_buff *skb)
905 struct ethhdr *eh = (struct ethhdr *)skb->data;
906 u16 offset = ETH_HLEN;
908 if (eh->h_proto == htons(ETH_P_IPV6)) {
909 struct ipv6hdr *ip6h = (struct ipv6hdr *)(skb->data + offset);
911 offset += sizeof(struct ipv6hdr);
912 if (ip6h->nexthdr != NEXTHDR_TCP &&
913 ip6h->nexthdr != NEXTHDR_UDP) {
914 struct ipv6_opt_hdr *ehdr =
915 (struct ipv6_opt_hdr *) (skb->data + offset);
917 /* offending pkt: 2nd byte following IPv6 hdr is 0xff */
918 if (ehdr->hdrlen == 0xff)
925 static int be_vlan_tag_tx_chk(struct be_adapter *adapter, struct sk_buff *skb)
927 return vlan_tx_tag_present(skb) || adapter->pvid || adapter->qnq_vid;
930 static int be_ipv6_tx_stall_chk(struct be_adapter *adapter,
933 return BE3_chip(adapter) && be_ipv6_exthdr_check(skb);
936 static struct sk_buff *be_lancer_xmit_workarounds(struct be_adapter *adapter,
940 struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
941 unsigned int eth_hdr_len;
944 /* For padded packets, BE HW modifies tot_len field in IP header
945 * incorrecly when VLAN tag is inserted by HW.
946 * For padded packets, Lancer computes incorrect checksum.
948 eth_hdr_len = ntohs(skb->protocol) == ETH_P_8021Q ?
949 VLAN_ETH_HLEN : ETH_HLEN;
950 if (skb->len <= 60 &&
951 (lancer_chip(adapter) || vlan_tx_tag_present(skb)) &&
953 ip = (struct iphdr *)ip_hdr(skb);
954 pskb_trim(skb, eth_hdr_len + ntohs(ip->tot_len));
957 /* If vlan tag is already inlined in the packet, skip HW VLAN
958 * tagging in pvid-tagging mode
960 if (be_pvid_tagging_enabled(adapter) &&
961 veh->h_vlan_proto == htons(ETH_P_8021Q))
962 *skip_hw_vlan = true;
964 /* HW has a bug wherein it will calculate CSUM for VLAN
965 * pkts even though it is disabled.
966 * Manually insert VLAN in pkt.
968 if (skb->ip_summed != CHECKSUM_PARTIAL &&
969 vlan_tx_tag_present(skb)) {
970 skb = be_insert_vlan_in_pkt(adapter, skb, skip_hw_vlan);
975 /* HW may lockup when VLAN HW tagging is requested on
976 * certain ipv6 packets. Drop such pkts if the HW workaround to
977 * skip HW tagging is not enabled by FW.
979 if (unlikely(be_ipv6_tx_stall_chk(adapter, skb) &&
980 (adapter->pvid || adapter->qnq_vid) &&
981 !qnq_async_evt_rcvd(adapter)))
984 /* Manual VLAN tag insertion to prevent:
985 * ASIC lockup when the ASIC inserts VLAN tag into
986 * certain ipv6 packets. Insert VLAN tags in driver,
987 * and set event, completion, vlan bits accordingly
990 if (be_ipv6_tx_stall_chk(adapter, skb) &&
991 be_vlan_tag_tx_chk(adapter, skb)) {
992 skb = be_insert_vlan_in_pkt(adapter, skb, skip_hw_vlan);
999 dev_kfree_skb_any(skb);
1004 static struct sk_buff *be_xmit_workarounds(struct be_adapter *adapter,
1005 struct sk_buff *skb,
1008 /* Lancer, SH-R ASICs have a bug wherein Packets that are 32 bytes or
1009 * less may cause a transmit stall on that port. So the work-around is
1010 * to pad short packets (<= 32 bytes) to a 36-byte length.
1012 if (unlikely(!BEx_chip(adapter) && skb->len <= 32)) {
1013 if (skb_padto(skb, 36))
1018 if (BEx_chip(adapter) || lancer_chip(adapter)) {
1019 skb = be_lancer_xmit_workarounds(adapter, skb, skip_hw_vlan);
1027 static netdev_tx_t be_xmit(struct sk_buff *skb, struct net_device *netdev)
1029 struct be_adapter *adapter = netdev_priv(netdev);
1030 struct be_tx_obj *txo = &adapter->tx_obj[skb_get_queue_mapping(skb)];
1031 struct be_queue_info *txq = &txo->q;
1032 bool dummy_wrb, stopped = false;
1033 u32 wrb_cnt = 0, copied = 0;
1034 bool skip_hw_vlan = false;
1035 u32 start = txq->head;
1037 skb = be_xmit_workarounds(adapter, skb, &skip_hw_vlan);
1039 tx_stats(txo)->tx_drv_drops++;
1040 return NETDEV_TX_OK;
1043 wrb_cnt = wrb_cnt_for_skb(adapter, skb, &dummy_wrb);
1045 copied = make_tx_wrbs(adapter, txq, skb, wrb_cnt, dummy_wrb,
1048 int gso_segs = skb_shinfo(skb)->gso_segs;
1050 /* record the sent skb in the sent_skb table */
1051 BUG_ON(txo->sent_skb_list[start]);
1052 txo->sent_skb_list[start] = skb;
1054 /* Ensure txq has space for the next skb; Else stop the queue
1055 * *BEFORE* ringing the tx doorbell, so that we serialze the
1056 * tx compls of the current transmit which'll wake up the queue
1058 atomic_add(wrb_cnt, &txq->used);
1059 if ((BE_MAX_TX_FRAG_COUNT + atomic_read(&txq->used)) >=
1061 netif_stop_subqueue(netdev, skb_get_queue_mapping(skb));
1065 be_txq_notify(adapter, txo, wrb_cnt);
1067 be_tx_stats_update(txo, wrb_cnt, copied, gso_segs, stopped);
1070 tx_stats(txo)->tx_drv_drops++;
1071 dev_kfree_skb_any(skb);
1073 return NETDEV_TX_OK;
1076 static int be_change_mtu(struct net_device *netdev, int new_mtu)
1078 struct be_adapter *adapter = netdev_priv(netdev);
1079 if (new_mtu < BE_MIN_MTU ||
1080 new_mtu > (BE_MAX_JUMBO_FRAME_SIZE -
1081 (ETH_HLEN + ETH_FCS_LEN))) {
1082 dev_info(&adapter->pdev->dev,
1083 "MTU must be between %d and %d bytes\n",
1085 (BE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN)));
1088 dev_info(&adapter->pdev->dev, "MTU changed from %d to %d bytes\n",
1089 netdev->mtu, new_mtu);
1090 netdev->mtu = new_mtu;
1095 * A max of 64 (BE_NUM_VLANS_SUPPORTED) vlans can be configured in BE.
1096 * If the user configures more, place BE in vlan promiscuous mode.
1098 static int be_vid_config(struct be_adapter *adapter)
1100 u16 vids[BE_NUM_VLANS_SUPPORTED];
1104 /* No need to further configure vids if in promiscuous mode */
1105 if (adapter->promiscuous)
1108 if (adapter->vlans_added > be_max_vlans(adapter))
1109 goto set_vlan_promisc;
1111 /* Construct VLAN Table to give to HW */
1112 for (i = 0; i < VLAN_N_VID; i++)
1113 if (adapter->vlan_tag[i])
1114 vids[num++] = cpu_to_le16(i);
1116 status = be_cmd_vlan_config(adapter, adapter->if_handle,
1120 /* Set to VLAN promisc mode as setting VLAN filter failed */
1121 if (status == MCC_ADDL_STS_INSUFFICIENT_RESOURCES)
1122 goto set_vlan_promisc;
1123 dev_err(&adapter->pdev->dev,
1124 "Setting HW VLAN filtering failed.\n");
1126 if (adapter->flags & BE_FLAGS_VLAN_PROMISC) {
1127 /* hw VLAN filtering re-enabled. */
1128 status = be_cmd_rx_filter(adapter,
1129 BE_FLAGS_VLAN_PROMISC, OFF);
1131 dev_info(&adapter->pdev->dev,
1132 "Disabling VLAN Promiscuous mode.\n");
1133 adapter->flags &= ~BE_FLAGS_VLAN_PROMISC;
1141 if (adapter->flags & BE_FLAGS_VLAN_PROMISC)
1144 status = be_cmd_rx_filter(adapter, BE_FLAGS_VLAN_PROMISC, ON);
1146 dev_info(&adapter->pdev->dev, "Enable VLAN Promiscuous mode\n");
1147 adapter->flags |= BE_FLAGS_VLAN_PROMISC;
1149 dev_err(&adapter->pdev->dev,
1150 "Failed to enable VLAN Promiscuous mode.\n");
1154 static int be_vlan_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
1156 struct be_adapter *adapter = netdev_priv(netdev);
1159 /* Packets with VID 0 are always received by Lancer by default */
1160 if (lancer_chip(adapter) && vid == 0)
1163 if (adapter->vlan_tag[vid])
1166 adapter->vlan_tag[vid] = 1;
1167 adapter->vlans_added++;
1169 status = be_vid_config(adapter);
1171 adapter->vlans_added--;
1172 adapter->vlan_tag[vid] = 0;
1178 static int be_vlan_rem_vid(struct net_device *netdev, __be16 proto, u16 vid)
1180 struct be_adapter *adapter = netdev_priv(netdev);
1183 /* Packets with VID 0 are always received by Lancer by default */
1184 if (lancer_chip(adapter) && vid == 0)
1187 adapter->vlan_tag[vid] = 0;
1188 status = be_vid_config(adapter);
1190 adapter->vlans_added--;
1192 adapter->vlan_tag[vid] = 1;
1197 static void be_clear_promisc(struct be_adapter *adapter)
1199 adapter->promiscuous = false;
1200 adapter->flags &= ~BE_FLAGS_VLAN_PROMISC;
1202 be_cmd_rx_filter(adapter, IFF_PROMISC, OFF);
1205 static void be_set_rx_mode(struct net_device *netdev)
1207 struct be_adapter *adapter = netdev_priv(netdev);
1210 if (netdev->flags & IFF_PROMISC) {
1211 be_cmd_rx_filter(adapter, IFF_PROMISC, ON);
1212 adapter->promiscuous = true;
1216 /* BE was previously in promiscuous mode; disable it */
1217 if (adapter->promiscuous) {
1218 be_clear_promisc(adapter);
1219 if (adapter->vlans_added)
1220 be_vid_config(adapter);
1223 /* Enable multicast promisc if num configured exceeds what we support */
1224 if (netdev->flags & IFF_ALLMULTI ||
1225 netdev_mc_count(netdev) > be_max_mc(adapter)) {
1226 be_cmd_rx_filter(adapter, IFF_ALLMULTI, ON);
1230 if (netdev_uc_count(netdev) != adapter->uc_macs) {
1231 struct netdev_hw_addr *ha;
1232 int i = 1; /* First slot is claimed by the Primary MAC */
1234 for (; adapter->uc_macs > 0; adapter->uc_macs--, i++) {
1235 be_cmd_pmac_del(adapter, adapter->if_handle,
1236 adapter->pmac_id[i], 0);
1239 if (netdev_uc_count(netdev) > be_max_uc(adapter)) {
1240 be_cmd_rx_filter(adapter, IFF_PROMISC, ON);
1241 adapter->promiscuous = true;
1245 netdev_for_each_uc_addr(ha, adapter->netdev) {
1246 adapter->uc_macs++; /* First slot is for Primary MAC */
1247 be_cmd_pmac_add(adapter, (u8 *)ha->addr,
1249 &adapter->pmac_id[adapter->uc_macs], 0);
1253 status = be_cmd_rx_filter(adapter, IFF_MULTICAST, ON);
1255 /* Set to MCAST promisc mode if setting MULTICAST address fails */
1257 dev_info(&adapter->pdev->dev, "Exhausted multicast HW filters.\n");
1258 dev_info(&adapter->pdev->dev, "Disabling HW multicast filtering.\n");
1259 be_cmd_rx_filter(adapter, IFF_ALLMULTI, ON);
1265 static int be_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
1267 struct be_adapter *adapter = netdev_priv(netdev);
1268 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1271 if (!sriov_enabled(adapter))
1274 if (!is_valid_ether_addr(mac) || vf >= adapter->num_vfs)
1277 if (BEx_chip(adapter)) {
1278 be_cmd_pmac_del(adapter, vf_cfg->if_handle, vf_cfg->pmac_id,
1281 status = be_cmd_pmac_add(adapter, mac, vf_cfg->if_handle,
1282 &vf_cfg->pmac_id, vf + 1);
1284 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
1289 dev_err(&adapter->pdev->dev, "MAC %pM set on VF %d Failed\n",
1292 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
1297 static int be_get_vf_config(struct net_device *netdev, int vf,
1298 struct ifla_vf_info *vi)
1300 struct be_adapter *adapter = netdev_priv(netdev);
1301 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1303 if (!sriov_enabled(adapter))
1306 if (vf >= adapter->num_vfs)
1310 vi->tx_rate = vf_cfg->tx_rate;
1311 vi->vlan = vf_cfg->vlan_tag & VLAN_VID_MASK;
1312 vi->qos = vf_cfg->vlan_tag >> VLAN_PRIO_SHIFT;
1313 memcpy(&vi->mac, vf_cfg->mac_addr, ETH_ALEN);
1314 vi->linkstate = adapter->vf_cfg[vf].plink_tracking;
1319 static int be_set_vf_vlan(struct net_device *netdev,
1320 int vf, u16 vlan, u8 qos)
1322 struct be_adapter *adapter = netdev_priv(netdev);
1323 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1326 if (!sriov_enabled(adapter))
1329 if (vf >= adapter->num_vfs || vlan > 4095 || qos > 7)
1333 vlan |= qos << VLAN_PRIO_SHIFT;
1334 if (vf_cfg->vlan_tag != vlan)
1335 status = be_cmd_set_hsw_config(adapter, vlan, vf + 1,
1336 vf_cfg->if_handle, 0);
1338 /* Reset Transparent Vlan Tagging. */
1339 status = be_cmd_set_hsw_config(adapter, BE_RESET_VLAN_TAG_ID,
1340 vf + 1, vf_cfg->if_handle, 0);
1344 vf_cfg->vlan_tag = vlan;
1346 dev_info(&adapter->pdev->dev,
1347 "VLAN %d config on VF %d failed\n", vlan, vf);
1351 static int be_set_vf_tx_rate(struct net_device *netdev,
1354 struct be_adapter *adapter = netdev_priv(netdev);
1357 if (!sriov_enabled(adapter))
1360 if (vf >= adapter->num_vfs)
1363 if (rate < 100 || rate > 10000) {
1364 dev_err(&adapter->pdev->dev,
1365 "tx rate must be between 100 and 10000 Mbps\n");
1369 status = be_cmd_config_qos(adapter, rate / 10, vf + 1);
1371 dev_err(&adapter->pdev->dev,
1372 "tx rate %d on VF %d failed\n", rate, vf);
1374 adapter->vf_cfg[vf].tx_rate = rate;
1377 static int be_set_vf_link_state(struct net_device *netdev, int vf,
1380 struct be_adapter *adapter = netdev_priv(netdev);
1383 if (!sriov_enabled(adapter))
1386 if (vf >= adapter->num_vfs)
1389 status = be_cmd_set_logical_link_config(adapter, link_state, vf+1);
1391 adapter->vf_cfg[vf].plink_tracking = link_state;
1396 static void be_aic_update(struct be_aic_obj *aic, u64 rx_pkts, u64 tx_pkts,
1399 aic->rx_pkts_prev = rx_pkts;
1400 aic->tx_reqs_prev = tx_pkts;
1404 static void be_eqd_update(struct be_adapter *adapter)
1406 struct be_set_eqd set_eqd[MAX_EVT_QS];
1407 int eqd, i, num = 0, start;
1408 struct be_aic_obj *aic;
1409 struct be_eq_obj *eqo;
1410 struct be_rx_obj *rxo;
1411 struct be_tx_obj *txo;
1412 u64 rx_pkts, tx_pkts;
1416 for_all_evt_queues(adapter, eqo, i) {
1417 aic = &adapter->aic_obj[eqo->idx];
1425 rxo = &adapter->rx_obj[eqo->idx];
1427 start = u64_stats_fetch_begin_irq(&rxo->stats.sync);
1428 rx_pkts = rxo->stats.rx_pkts;
1429 } while (u64_stats_fetch_retry_irq(&rxo->stats.sync, start));
1431 txo = &adapter->tx_obj[eqo->idx];
1433 start = u64_stats_fetch_begin_irq(&txo->stats.sync);
1434 tx_pkts = txo->stats.tx_reqs;
1435 } while (u64_stats_fetch_retry_irq(&txo->stats.sync, start));
1438 /* Skip, if wrapped around or first calculation */
1440 if (!aic->jiffies || time_before(now, aic->jiffies) ||
1441 rx_pkts < aic->rx_pkts_prev ||
1442 tx_pkts < aic->tx_reqs_prev) {
1443 be_aic_update(aic, rx_pkts, tx_pkts, now);
1447 delta = jiffies_to_msecs(now - aic->jiffies);
1448 pps = (((u32)(rx_pkts - aic->rx_pkts_prev) * 1000) / delta) +
1449 (((u32)(tx_pkts - aic->tx_reqs_prev) * 1000) / delta);
1450 eqd = (pps / 15000) << 2;
1454 eqd = min_t(u32, eqd, aic->max_eqd);
1455 eqd = max_t(u32, eqd, aic->min_eqd);
1457 be_aic_update(aic, rx_pkts, tx_pkts, now);
1459 if (eqd != aic->prev_eqd) {
1460 set_eqd[num].delay_multiplier = (eqd * 65)/100;
1461 set_eqd[num].eq_id = eqo->q.id;
1462 aic->prev_eqd = eqd;
1468 be_cmd_modify_eqd(adapter, set_eqd, num);
1471 static void be_rx_stats_update(struct be_rx_obj *rxo,
1472 struct be_rx_compl_info *rxcp)
1474 struct be_rx_stats *stats = rx_stats(rxo);
1476 u64_stats_update_begin(&stats->sync);
1478 stats->rx_bytes += rxcp->pkt_size;
1480 if (rxcp->pkt_type == BE_MULTICAST_PACKET)
1481 stats->rx_mcast_pkts++;
1483 stats->rx_compl_err++;
1484 u64_stats_update_end(&stats->sync);
1487 static inline bool csum_passed(struct be_rx_compl_info *rxcp)
1489 /* L4 checksum is not reliable for non TCP/UDP packets.
1490 * Also ignore ipcksm for ipv6 pkts
1492 return (rxcp->tcpf || rxcp->udpf) && rxcp->l4_csum &&
1493 (rxcp->ip_csum || rxcp->ipv6) && !rxcp->err;
1496 static struct be_rx_page_info *get_rx_page_info(struct be_rx_obj *rxo)
1498 struct be_adapter *adapter = rxo->adapter;
1499 struct be_rx_page_info *rx_page_info;
1500 struct be_queue_info *rxq = &rxo->q;
1501 u16 frag_idx = rxq->tail;
1503 rx_page_info = &rxo->page_info_tbl[frag_idx];
1504 BUG_ON(!rx_page_info->page);
1506 if (rx_page_info->last_frag) {
1507 dma_unmap_page(&adapter->pdev->dev,
1508 dma_unmap_addr(rx_page_info, bus),
1509 adapter->big_page_size, DMA_FROM_DEVICE);
1510 rx_page_info->last_frag = false;
1512 dma_sync_single_for_cpu(&adapter->pdev->dev,
1513 dma_unmap_addr(rx_page_info, bus),
1514 rx_frag_size, DMA_FROM_DEVICE);
1517 queue_tail_inc(rxq);
1518 atomic_dec(&rxq->used);
1519 return rx_page_info;
1522 /* Throwaway the data in the Rx completion */
1523 static void be_rx_compl_discard(struct be_rx_obj *rxo,
1524 struct be_rx_compl_info *rxcp)
1526 struct be_rx_page_info *page_info;
1527 u16 i, num_rcvd = rxcp->num_rcvd;
1529 for (i = 0; i < num_rcvd; i++) {
1530 page_info = get_rx_page_info(rxo);
1531 put_page(page_info->page);
1532 memset(page_info, 0, sizeof(*page_info));
1537 * skb_fill_rx_data forms a complete skb for an ether frame
1538 * indicated by rxcp.
1540 static void skb_fill_rx_data(struct be_rx_obj *rxo, struct sk_buff *skb,
1541 struct be_rx_compl_info *rxcp)
1543 struct be_rx_page_info *page_info;
1545 u16 hdr_len, curr_frag_len, remaining;
1548 page_info = get_rx_page_info(rxo);
1549 start = page_address(page_info->page) + page_info->page_offset;
1552 /* Copy data in the first descriptor of this completion */
1553 curr_frag_len = min(rxcp->pkt_size, rx_frag_size);
1555 skb->len = curr_frag_len;
1556 if (curr_frag_len <= BE_HDR_LEN) { /* tiny packet */
1557 memcpy(skb->data, start, curr_frag_len);
1558 /* Complete packet has now been moved to data */
1559 put_page(page_info->page);
1561 skb->tail += curr_frag_len;
1564 memcpy(skb->data, start, hdr_len);
1565 skb_shinfo(skb)->nr_frags = 1;
1566 skb_frag_set_page(skb, 0, page_info->page);
1567 skb_shinfo(skb)->frags[0].page_offset =
1568 page_info->page_offset + hdr_len;
1569 skb_frag_size_set(&skb_shinfo(skb)->frags[0], curr_frag_len - hdr_len);
1570 skb->data_len = curr_frag_len - hdr_len;
1571 skb->truesize += rx_frag_size;
1572 skb->tail += hdr_len;
1574 page_info->page = NULL;
1576 if (rxcp->pkt_size <= rx_frag_size) {
1577 BUG_ON(rxcp->num_rcvd != 1);
1581 /* More frags present for this completion */
1582 remaining = rxcp->pkt_size - curr_frag_len;
1583 for (i = 1, j = 0; i < rxcp->num_rcvd; i++) {
1584 page_info = get_rx_page_info(rxo);
1585 curr_frag_len = min(remaining, rx_frag_size);
1587 /* Coalesce all frags from the same physical page in one slot */
1588 if (page_info->page_offset == 0) {
1591 skb_frag_set_page(skb, j, page_info->page);
1592 skb_shinfo(skb)->frags[j].page_offset =
1593 page_info->page_offset;
1594 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1595 skb_shinfo(skb)->nr_frags++;
1597 put_page(page_info->page);
1600 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1601 skb->len += curr_frag_len;
1602 skb->data_len += curr_frag_len;
1603 skb->truesize += rx_frag_size;
1604 remaining -= curr_frag_len;
1605 page_info->page = NULL;
1607 BUG_ON(j > MAX_SKB_FRAGS);
1610 /* Process the RX completion indicated by rxcp when GRO is disabled */
1611 static void be_rx_compl_process(struct be_rx_obj *rxo, struct napi_struct *napi,
1612 struct be_rx_compl_info *rxcp)
1614 struct be_adapter *adapter = rxo->adapter;
1615 struct net_device *netdev = adapter->netdev;
1616 struct sk_buff *skb;
1618 skb = netdev_alloc_skb_ip_align(netdev, BE_RX_SKB_ALLOC_SIZE);
1619 if (unlikely(!skb)) {
1620 rx_stats(rxo)->rx_drops_no_skbs++;
1621 be_rx_compl_discard(rxo, rxcp);
1625 skb_fill_rx_data(rxo, skb, rxcp);
1627 if (likely((netdev->features & NETIF_F_RXCSUM) && csum_passed(rxcp)))
1628 skb->ip_summed = CHECKSUM_UNNECESSARY;
1630 skb_checksum_none_assert(skb);
1632 skb->protocol = eth_type_trans(skb, netdev);
1633 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1634 if (netdev->features & NETIF_F_RXHASH)
1635 skb_set_hash(skb, rxcp->rss_hash, PKT_HASH_TYPE_L3);
1637 skb->encapsulation = rxcp->tunneled;
1638 skb_mark_napi_id(skb, napi);
1641 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1643 netif_receive_skb(skb);
1646 /* Process the RX completion indicated by rxcp when GRO is enabled */
1647 static void be_rx_compl_process_gro(struct be_rx_obj *rxo,
1648 struct napi_struct *napi,
1649 struct be_rx_compl_info *rxcp)
1651 struct be_adapter *adapter = rxo->adapter;
1652 struct be_rx_page_info *page_info;
1653 struct sk_buff *skb = NULL;
1654 u16 remaining, curr_frag_len;
1657 skb = napi_get_frags(napi);
1659 be_rx_compl_discard(rxo, rxcp);
1663 remaining = rxcp->pkt_size;
1664 for (i = 0, j = -1; i < rxcp->num_rcvd; i++) {
1665 page_info = get_rx_page_info(rxo);
1667 curr_frag_len = min(remaining, rx_frag_size);
1669 /* Coalesce all frags from the same physical page in one slot */
1670 if (i == 0 || page_info->page_offset == 0) {
1671 /* First frag or Fresh page */
1673 skb_frag_set_page(skb, j, page_info->page);
1674 skb_shinfo(skb)->frags[j].page_offset =
1675 page_info->page_offset;
1676 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1678 put_page(page_info->page);
1680 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1681 skb->truesize += rx_frag_size;
1682 remaining -= curr_frag_len;
1683 memset(page_info, 0, sizeof(*page_info));
1685 BUG_ON(j > MAX_SKB_FRAGS);
1687 skb_shinfo(skb)->nr_frags = j + 1;
1688 skb->len = rxcp->pkt_size;
1689 skb->data_len = rxcp->pkt_size;
1690 skb->ip_summed = CHECKSUM_UNNECESSARY;
1691 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1692 if (adapter->netdev->features & NETIF_F_RXHASH)
1693 skb_set_hash(skb, rxcp->rss_hash, PKT_HASH_TYPE_L3);
1695 skb->encapsulation = rxcp->tunneled;
1696 skb_mark_napi_id(skb, napi);
1699 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1701 napi_gro_frags(napi);
1704 static void be_parse_rx_compl_v1(struct be_eth_rx_compl *compl,
1705 struct be_rx_compl_info *rxcp)
1708 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, pktsize, compl);
1709 rxcp->vlanf = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, vtp, compl);
1710 rxcp->err = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, err, compl);
1711 rxcp->tcpf = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, tcpf, compl);
1712 rxcp->udpf = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, udpf, compl);
1714 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, ipcksm, compl);
1716 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, l4_cksm, compl);
1718 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, ip_version, compl);
1720 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, numfrags, compl);
1722 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, cast_enc, compl);
1724 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, rsshash, compl);
1726 rxcp->qnq = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, qnq,
1728 rxcp->vlan_tag = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, vlan_tag,
1731 rxcp->port = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, port, compl);
1733 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, tunneled, compl);
1736 static void be_parse_rx_compl_v0(struct be_eth_rx_compl *compl,
1737 struct be_rx_compl_info *rxcp)
1740 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, pktsize, compl);
1741 rxcp->vlanf = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, vtp, compl);
1742 rxcp->err = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, err, compl);
1743 rxcp->tcpf = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, tcpf, compl);
1744 rxcp->udpf = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, udpf, compl);
1746 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, ipcksm, compl);
1748 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, l4_cksm, compl);
1750 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, ip_version, compl);
1752 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, numfrags, compl);
1754 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, cast_enc, compl);
1756 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, rsshash, compl);
1758 rxcp->qnq = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, qnq,
1760 rxcp->vlan_tag = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, vlan_tag,
1763 rxcp->port = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, port, compl);
1764 rxcp->ip_frag = AMAP_GET_BITS(struct amap_eth_rx_compl_v0,
1768 static struct be_rx_compl_info *be_rx_compl_get(struct be_rx_obj *rxo)
1770 struct be_eth_rx_compl *compl = queue_tail_node(&rxo->cq);
1771 struct be_rx_compl_info *rxcp = &rxo->rxcp;
1772 struct be_adapter *adapter = rxo->adapter;
1774 /* For checking the valid bit it is Ok to use either definition as the
1775 * valid bit is at the same position in both v0 and v1 Rx compl */
1776 if (compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] == 0)
1780 be_dws_le_to_cpu(compl, sizeof(*compl));
1782 if (adapter->be3_native)
1783 be_parse_rx_compl_v1(compl, rxcp);
1785 be_parse_rx_compl_v0(compl, rxcp);
1791 /* In QNQ modes, if qnq bit is not set, then the packet was
1792 * tagged only with the transparent outer vlan-tag and must
1793 * not be treated as a vlan packet by host
1795 if (be_is_qnq_mode(adapter) && !rxcp->qnq)
1798 if (!lancer_chip(adapter))
1799 rxcp->vlan_tag = swab16(rxcp->vlan_tag);
1801 if (adapter->pvid == (rxcp->vlan_tag & VLAN_VID_MASK) &&
1802 !adapter->vlan_tag[rxcp->vlan_tag])
1806 /* As the compl has been parsed, reset it; we wont touch it again */
1807 compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] = 0;
1809 queue_tail_inc(&rxo->cq);
1813 static inline struct page *be_alloc_pages(u32 size, gfp_t gfp)
1815 u32 order = get_order(size);
1819 return alloc_pages(gfp, order);
1823 * Allocate a page, split it to fragments of size rx_frag_size and post as
1824 * receive buffers to BE
1826 static void be_post_rx_frags(struct be_rx_obj *rxo, gfp_t gfp)
1828 struct be_adapter *adapter = rxo->adapter;
1829 struct be_rx_page_info *page_info = NULL, *prev_page_info = NULL;
1830 struct be_queue_info *rxq = &rxo->q;
1831 struct page *pagep = NULL;
1832 struct device *dev = &adapter->pdev->dev;
1833 struct be_eth_rx_d *rxd;
1834 u64 page_dmaaddr = 0, frag_dmaaddr;
1835 u32 posted, page_offset = 0;
1837 page_info = &rxo->page_info_tbl[rxq->head];
1838 for (posted = 0; posted < MAX_RX_POST && !page_info->page; posted++) {
1840 pagep = be_alloc_pages(adapter->big_page_size, gfp);
1841 if (unlikely(!pagep)) {
1842 rx_stats(rxo)->rx_post_fail++;
1845 page_dmaaddr = dma_map_page(dev, pagep, 0,
1846 adapter->big_page_size,
1848 if (dma_mapping_error(dev, page_dmaaddr)) {
1851 rx_stats(rxo)->rx_post_fail++;
1857 page_offset += rx_frag_size;
1859 page_info->page_offset = page_offset;
1860 page_info->page = pagep;
1862 rxd = queue_head_node(rxq);
1863 frag_dmaaddr = page_dmaaddr + page_info->page_offset;
1864 rxd->fragpa_lo = cpu_to_le32(frag_dmaaddr & 0xFFFFFFFF);
1865 rxd->fragpa_hi = cpu_to_le32(upper_32_bits(frag_dmaaddr));
1867 /* Any space left in the current big page for another frag? */
1868 if ((page_offset + rx_frag_size + rx_frag_size) >
1869 adapter->big_page_size) {
1871 page_info->last_frag = true;
1872 dma_unmap_addr_set(page_info, bus, page_dmaaddr);
1874 dma_unmap_addr_set(page_info, bus, frag_dmaaddr);
1877 prev_page_info = page_info;
1878 queue_head_inc(rxq);
1879 page_info = &rxo->page_info_tbl[rxq->head];
1882 /* Mark the last frag of a page when we break out of the above loop
1883 * with no more slots available in the RXQ
1886 prev_page_info->last_frag = true;
1887 dma_unmap_addr_set(prev_page_info, bus, page_dmaaddr);
1891 atomic_add(posted, &rxq->used);
1892 if (rxo->rx_post_starved)
1893 rxo->rx_post_starved = false;
1894 be_rxq_notify(adapter, rxq->id, posted);
1895 } else if (atomic_read(&rxq->used) == 0) {
1896 /* Let be_worker replenish when memory is available */
1897 rxo->rx_post_starved = true;
1901 static struct be_eth_tx_compl *be_tx_compl_get(struct be_queue_info *tx_cq)
1903 struct be_eth_tx_compl *txcp = queue_tail_node(tx_cq);
1905 if (txcp->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] == 0)
1909 be_dws_le_to_cpu(txcp, sizeof(*txcp));
1911 txcp->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] = 0;
1913 queue_tail_inc(tx_cq);
1917 static u16 be_tx_compl_process(struct be_adapter *adapter,
1918 struct be_tx_obj *txo, u16 last_index)
1920 struct be_queue_info *txq = &txo->q;
1921 struct be_eth_wrb *wrb;
1922 struct sk_buff **sent_skbs = txo->sent_skb_list;
1923 struct sk_buff *sent_skb;
1924 u16 cur_index, num_wrbs = 1; /* account for hdr wrb */
1925 bool unmap_skb_hdr = true;
1927 sent_skb = sent_skbs[txq->tail];
1929 sent_skbs[txq->tail] = NULL;
1931 /* skip header wrb */
1932 queue_tail_inc(txq);
1935 cur_index = txq->tail;
1936 wrb = queue_tail_node(txq);
1937 unmap_tx_frag(&adapter->pdev->dev, wrb,
1938 (unmap_skb_hdr && skb_headlen(sent_skb)));
1939 unmap_skb_hdr = false;
1942 queue_tail_inc(txq);
1943 } while (cur_index != last_index);
1945 dev_kfree_skb_any(sent_skb);
1949 /* Return the number of events in the event queue */
1950 static inline int events_get(struct be_eq_obj *eqo)
1952 struct be_eq_entry *eqe;
1956 eqe = queue_tail_node(&eqo->q);
1963 queue_tail_inc(&eqo->q);
1969 /* Leaves the EQ is disarmed state */
1970 static void be_eq_clean(struct be_eq_obj *eqo)
1972 int num = events_get(eqo);
1974 be_eq_notify(eqo->adapter, eqo->q.id, false, true, num);
1977 static void be_rx_cq_clean(struct be_rx_obj *rxo)
1979 struct be_rx_page_info *page_info;
1980 struct be_queue_info *rxq = &rxo->q;
1981 struct be_queue_info *rx_cq = &rxo->cq;
1982 struct be_rx_compl_info *rxcp;
1983 struct be_adapter *adapter = rxo->adapter;
1986 /* Consume pending rx completions.
1987 * Wait for the flush completion (identified by zero num_rcvd)
1988 * to arrive. Notify CQ even when there are no more CQ entries
1989 * for HW to flush partially coalesced CQ entries.
1990 * In Lancer, there is no need to wait for flush compl.
1993 rxcp = be_rx_compl_get(rxo);
1995 if (lancer_chip(adapter))
1998 if (flush_wait++ > 10 || be_hw_error(adapter)) {
1999 dev_warn(&adapter->pdev->dev,
2000 "did not receive flush compl\n");
2003 be_cq_notify(adapter, rx_cq->id, true, 0);
2006 be_rx_compl_discard(rxo, rxcp);
2007 be_cq_notify(adapter, rx_cq->id, false, 1);
2008 if (rxcp->num_rcvd == 0)
2013 /* After cleanup, leave the CQ in unarmed state */
2014 be_cq_notify(adapter, rx_cq->id, false, 0);
2016 /* Then free posted rx buffers that were not used */
2017 while (atomic_read(&rxq->used) > 0) {
2018 page_info = get_rx_page_info(rxo);
2019 put_page(page_info->page);
2020 memset(page_info, 0, sizeof(*page_info));
2022 BUG_ON(atomic_read(&rxq->used));
2023 rxq->tail = rxq->head = 0;
2026 static void be_tx_compl_clean(struct be_adapter *adapter)
2028 struct be_tx_obj *txo;
2029 struct be_queue_info *txq;
2030 struct be_eth_tx_compl *txcp;
2031 u16 end_idx, cmpl = 0, timeo = 0, num_wrbs = 0;
2032 struct sk_buff *sent_skb;
2034 int i, pending_txqs;
2036 /* Stop polling for compls when HW has been silent for 10ms */
2038 pending_txqs = adapter->num_tx_qs;
2040 for_all_tx_queues(adapter, txo, i) {
2044 while ((txcp = be_tx_compl_get(&txo->cq))) {
2046 AMAP_GET_BITS(struct amap_eth_tx_compl,
2048 num_wrbs += be_tx_compl_process(adapter, txo,
2053 be_cq_notify(adapter, txo->cq.id, false, cmpl);
2054 atomic_sub(num_wrbs, &txq->used);
2057 if (atomic_read(&txq->used) == 0)
2061 if (pending_txqs == 0 || ++timeo > 10 || be_hw_error(adapter))
2067 for_all_tx_queues(adapter, txo, i) {
2069 if (atomic_read(&txq->used))
2070 dev_err(&adapter->pdev->dev, "%d pending tx-compls\n",
2071 atomic_read(&txq->used));
2073 /* free posted tx for which compls will never arrive */
2074 while (atomic_read(&txq->used)) {
2075 sent_skb = txo->sent_skb_list[txq->tail];
2076 end_idx = txq->tail;
2077 num_wrbs = wrb_cnt_for_skb(adapter, sent_skb,
2079 index_adv(&end_idx, num_wrbs - 1, txq->len);
2080 num_wrbs = be_tx_compl_process(adapter, txo, end_idx);
2081 atomic_sub(num_wrbs, &txq->used);
2086 static void be_evt_queues_destroy(struct be_adapter *adapter)
2088 struct be_eq_obj *eqo;
2091 for_all_evt_queues(adapter, eqo, i) {
2092 if (eqo->q.created) {
2094 be_cmd_q_destroy(adapter, &eqo->q, QTYPE_EQ);
2095 napi_hash_del(&eqo->napi);
2096 netif_napi_del(&eqo->napi);
2098 be_queue_free(adapter, &eqo->q);
2102 static int be_evt_queues_create(struct be_adapter *adapter)
2104 struct be_queue_info *eq;
2105 struct be_eq_obj *eqo;
2106 struct be_aic_obj *aic;
2109 adapter->num_evt_qs = min_t(u16, num_irqs(adapter),
2110 adapter->cfg_num_qs);
2112 for_all_evt_queues(adapter, eqo, i) {
2113 netif_napi_add(adapter->netdev, &eqo->napi, be_poll,
2115 napi_hash_add(&eqo->napi);
2116 aic = &adapter->aic_obj[i];
2117 eqo->adapter = adapter;
2118 eqo->tx_budget = BE_TX_BUDGET;
2120 aic->max_eqd = BE_MAX_EQD;
2124 rc = be_queue_alloc(adapter, eq, EVNT_Q_LEN,
2125 sizeof(struct be_eq_entry));
2129 rc = be_cmd_eq_create(adapter, eqo);
2136 static void be_mcc_queues_destroy(struct be_adapter *adapter)
2138 struct be_queue_info *q;
2140 q = &adapter->mcc_obj.q;
2142 be_cmd_q_destroy(adapter, q, QTYPE_MCCQ);
2143 be_queue_free(adapter, q);
2145 q = &adapter->mcc_obj.cq;
2147 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2148 be_queue_free(adapter, q);
2151 /* Must be called only after TX qs are created as MCC shares TX EQ */
2152 static int be_mcc_queues_create(struct be_adapter *adapter)
2154 struct be_queue_info *q, *cq;
2156 cq = &adapter->mcc_obj.cq;
2157 if (be_queue_alloc(adapter, cq, MCC_CQ_LEN,
2158 sizeof(struct be_mcc_compl)))
2161 /* Use the default EQ for MCC completions */
2162 if (be_cmd_cq_create(adapter, cq, &mcc_eqo(adapter)->q, true, 0))
2165 q = &adapter->mcc_obj.q;
2166 if (be_queue_alloc(adapter, q, MCC_Q_LEN, sizeof(struct be_mcc_wrb)))
2167 goto mcc_cq_destroy;
2169 if (be_cmd_mccq_create(adapter, q, cq))
2175 be_queue_free(adapter, q);
2177 be_cmd_q_destroy(adapter, cq, QTYPE_CQ);
2179 be_queue_free(adapter, cq);
2184 static void be_tx_queues_destroy(struct be_adapter *adapter)
2186 struct be_queue_info *q;
2187 struct be_tx_obj *txo;
2190 for_all_tx_queues(adapter, txo, i) {
2193 be_cmd_q_destroy(adapter, q, QTYPE_TXQ);
2194 be_queue_free(adapter, q);
2198 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2199 be_queue_free(adapter, q);
2203 static int be_tx_qs_create(struct be_adapter *adapter)
2205 struct be_queue_info *cq, *eq;
2206 struct be_tx_obj *txo;
2209 adapter->num_tx_qs = min(adapter->num_evt_qs, be_max_txqs(adapter));
2211 for_all_tx_queues(adapter, txo, i) {
2213 status = be_queue_alloc(adapter, cq, TX_CQ_LEN,
2214 sizeof(struct be_eth_tx_compl));
2218 u64_stats_init(&txo->stats.sync);
2219 u64_stats_init(&txo->stats.sync_compl);
2221 /* If num_evt_qs is less than num_tx_qs, then more than
2222 * one txq share an eq
2224 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2225 status = be_cmd_cq_create(adapter, cq, eq, false, 3);
2229 status = be_queue_alloc(adapter, &txo->q, TX_Q_LEN,
2230 sizeof(struct be_eth_wrb));
2234 status = be_cmd_txq_create(adapter, txo);
2239 dev_info(&adapter->pdev->dev, "created %d TX queue(s)\n",
2240 adapter->num_tx_qs);
2244 static void be_rx_cqs_destroy(struct be_adapter *adapter)
2246 struct be_queue_info *q;
2247 struct be_rx_obj *rxo;
2250 for_all_rx_queues(adapter, rxo, i) {
2253 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2254 be_queue_free(adapter, q);
2258 static int be_rx_cqs_create(struct be_adapter *adapter)
2260 struct be_queue_info *eq, *cq;
2261 struct be_rx_obj *rxo;
2264 /* We can create as many RSS rings as there are EQs. */
2265 adapter->num_rx_qs = adapter->num_evt_qs;
2267 /* We'll use RSS only if atleast 2 RSS rings are supported.
2268 * When RSS is used, we'll need a default RXQ for non-IP traffic.
2270 if (adapter->num_rx_qs > 1)
2271 adapter->num_rx_qs++;
2273 adapter->big_page_size = (1 << get_order(rx_frag_size)) * PAGE_SIZE;
2274 for_all_rx_queues(adapter, rxo, i) {
2275 rxo->adapter = adapter;
2277 rc = be_queue_alloc(adapter, cq, RX_CQ_LEN,
2278 sizeof(struct be_eth_rx_compl));
2282 u64_stats_init(&rxo->stats.sync);
2283 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2284 rc = be_cmd_cq_create(adapter, cq, eq, false, 3);
2289 dev_info(&adapter->pdev->dev,
2290 "created %d RSS queue(s) and 1 default RX queue\n",
2291 adapter->num_rx_qs - 1);
2295 static irqreturn_t be_intx(int irq, void *dev)
2297 struct be_eq_obj *eqo = dev;
2298 struct be_adapter *adapter = eqo->adapter;
2301 /* IRQ is not expected when NAPI is scheduled as the EQ
2302 * will not be armed.
2303 * But, this can happen on Lancer INTx where it takes
2304 * a while to de-assert INTx or in BE2 where occasionaly
2305 * an interrupt may be raised even when EQ is unarmed.
2306 * If NAPI is already scheduled, then counting & notifying
2307 * events will orphan them.
2309 if (napi_schedule_prep(&eqo->napi)) {
2310 num_evts = events_get(eqo);
2311 __napi_schedule(&eqo->napi);
2313 eqo->spurious_intr = 0;
2315 be_eq_notify(adapter, eqo->q.id, false, true, num_evts);
2317 /* Return IRQ_HANDLED only for the the first spurious intr
2318 * after a valid intr to stop the kernel from branding
2319 * this irq as a bad one!
2321 if (num_evts || eqo->spurious_intr++ == 0)
2327 static irqreturn_t be_msix(int irq, void *dev)
2329 struct be_eq_obj *eqo = dev;
2331 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
2332 napi_schedule(&eqo->napi);
2336 static inline bool do_gro(struct be_rx_compl_info *rxcp)
2338 return (rxcp->tcpf && !rxcp->err && rxcp->l4_csum) ? true : false;
2341 static int be_process_rx(struct be_rx_obj *rxo, struct napi_struct *napi,
2342 int budget, int polling)
2344 struct be_adapter *adapter = rxo->adapter;
2345 struct be_queue_info *rx_cq = &rxo->cq;
2346 struct be_rx_compl_info *rxcp;
2349 for (work_done = 0; work_done < budget; work_done++) {
2350 rxcp = be_rx_compl_get(rxo);
2354 /* Is it a flush compl that has no data */
2355 if (unlikely(rxcp->num_rcvd == 0))
2358 /* Discard compl with partial DMA Lancer B0 */
2359 if (unlikely(!rxcp->pkt_size)) {
2360 be_rx_compl_discard(rxo, rxcp);
2364 /* On BE drop pkts that arrive due to imperfect filtering in
2365 * promiscuous mode on some skews
2367 if (unlikely(rxcp->port != adapter->port_num &&
2368 !lancer_chip(adapter))) {
2369 be_rx_compl_discard(rxo, rxcp);
2373 /* Don't do gro when we're busy_polling */
2374 if (do_gro(rxcp) && polling != BUSY_POLLING)
2375 be_rx_compl_process_gro(rxo, napi, rxcp);
2377 be_rx_compl_process(rxo, napi, rxcp);
2380 be_rx_stats_update(rxo, rxcp);
2384 be_cq_notify(adapter, rx_cq->id, true, work_done);
2386 /* When an rx-obj gets into post_starved state, just
2387 * let be_worker do the posting.
2389 if (atomic_read(&rxo->q.used) < RX_FRAGS_REFILL_WM &&
2390 !rxo->rx_post_starved)
2391 be_post_rx_frags(rxo, GFP_ATOMIC);
2397 static bool be_process_tx(struct be_adapter *adapter, struct be_tx_obj *txo,
2398 int budget, int idx)
2400 struct be_eth_tx_compl *txcp;
2401 int num_wrbs = 0, work_done;
2403 for (work_done = 0; work_done < budget; work_done++) {
2404 txcp = be_tx_compl_get(&txo->cq);
2407 num_wrbs += be_tx_compl_process(adapter, txo,
2408 AMAP_GET_BITS(struct amap_eth_tx_compl,
2413 be_cq_notify(adapter, txo->cq.id, true, work_done);
2414 atomic_sub(num_wrbs, &txo->q.used);
2416 /* As Tx wrbs have been freed up, wake up netdev queue
2417 * if it was stopped due to lack of tx wrbs. */
2418 if (__netif_subqueue_stopped(adapter->netdev, idx) &&
2419 atomic_read(&txo->q.used) < txo->q.len / 2) {
2420 netif_wake_subqueue(adapter->netdev, idx);
2423 u64_stats_update_begin(&tx_stats(txo)->sync_compl);
2424 tx_stats(txo)->tx_compl += work_done;
2425 u64_stats_update_end(&tx_stats(txo)->sync_compl);
2427 return (work_done < budget); /* Done */
2430 int be_poll(struct napi_struct *napi, int budget)
2432 struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2433 struct be_adapter *adapter = eqo->adapter;
2434 int max_work = 0, work, i, num_evts;
2435 struct be_rx_obj *rxo;
2438 num_evts = events_get(eqo);
2440 /* Process all TXQs serviced by this EQ */
2441 for (i = eqo->idx; i < adapter->num_tx_qs; i += adapter->num_evt_qs) {
2442 tx_done = be_process_tx(adapter, &adapter->tx_obj[i],
2448 if (be_lock_napi(eqo)) {
2449 /* This loop will iterate twice for EQ0 in which
2450 * completions of the last RXQ (default one) are also processed
2451 * For other EQs the loop iterates only once
2453 for_all_rx_queues_on_eq(adapter, eqo, rxo, i) {
2454 work = be_process_rx(rxo, napi, budget, NAPI_POLLING);
2455 max_work = max(work, max_work);
2457 be_unlock_napi(eqo);
2462 if (is_mcc_eqo(eqo))
2463 be_process_mcc(adapter);
2465 if (max_work < budget) {
2466 napi_complete(napi);
2467 be_eq_notify(adapter, eqo->q.id, true, false, num_evts);
2469 /* As we'll continue in polling mode, count and clear events */
2470 be_eq_notify(adapter, eqo->q.id, false, false, num_evts);
2475 #ifdef CONFIG_NET_RX_BUSY_POLL
2476 static int be_busy_poll(struct napi_struct *napi)
2478 struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2479 struct be_adapter *adapter = eqo->adapter;
2480 struct be_rx_obj *rxo;
2483 if (!be_lock_busy_poll(eqo))
2484 return LL_FLUSH_BUSY;
2486 for_all_rx_queues_on_eq(adapter, eqo, rxo, i) {
2487 work = be_process_rx(rxo, napi, 4, BUSY_POLLING);
2492 be_unlock_busy_poll(eqo);
2497 void be_detect_error(struct be_adapter *adapter)
2499 u32 ue_lo = 0, ue_hi = 0, ue_lo_mask = 0, ue_hi_mask = 0;
2500 u32 sliport_status = 0, sliport_err1 = 0, sliport_err2 = 0;
2502 bool error_detected = false;
2503 struct device *dev = &adapter->pdev->dev;
2504 struct net_device *netdev = adapter->netdev;
2506 if (be_hw_error(adapter))
2509 if (lancer_chip(adapter)) {
2510 sliport_status = ioread32(adapter->db + SLIPORT_STATUS_OFFSET);
2511 if (sliport_status & SLIPORT_STATUS_ERR_MASK) {
2512 sliport_err1 = ioread32(adapter->db +
2513 SLIPORT_ERROR1_OFFSET);
2514 sliport_err2 = ioread32(adapter->db +
2515 SLIPORT_ERROR2_OFFSET);
2516 adapter->hw_error = true;
2517 /* Do not log error messages if its a FW reset */
2518 if (sliport_err1 == SLIPORT_ERROR_FW_RESET1 &&
2519 sliport_err2 == SLIPORT_ERROR_FW_RESET2) {
2520 dev_info(dev, "Firmware update in progress\n");
2522 error_detected = true;
2523 dev_err(dev, "Error detected in the card\n");
2524 dev_err(dev, "ERR: sliport status 0x%x\n",
2526 dev_err(dev, "ERR: sliport error1 0x%x\n",
2528 dev_err(dev, "ERR: sliport error2 0x%x\n",
2533 pci_read_config_dword(adapter->pdev,
2534 PCICFG_UE_STATUS_LOW, &ue_lo);
2535 pci_read_config_dword(adapter->pdev,
2536 PCICFG_UE_STATUS_HIGH, &ue_hi);
2537 pci_read_config_dword(adapter->pdev,
2538 PCICFG_UE_STATUS_LOW_MASK, &ue_lo_mask);
2539 pci_read_config_dword(adapter->pdev,
2540 PCICFG_UE_STATUS_HI_MASK, &ue_hi_mask);
2542 ue_lo = (ue_lo & ~ue_lo_mask);
2543 ue_hi = (ue_hi & ~ue_hi_mask);
2545 /* On certain platforms BE hardware can indicate spurious UEs.
2546 * Allow HW to stop working completely in case of a real UE.
2547 * Hence not setting the hw_error for UE detection.
2550 if (ue_lo || ue_hi) {
2551 error_detected = true;
2553 "Unrecoverable Error detected in the adapter");
2554 dev_err(dev, "Please reboot server to recover");
2555 if (skyhawk_chip(adapter))
2556 adapter->hw_error = true;
2557 for (i = 0; ue_lo; ue_lo >>= 1, i++) {
2559 dev_err(dev, "UE: %s bit set\n",
2560 ue_status_low_desc[i]);
2562 for (i = 0; ue_hi; ue_hi >>= 1, i++) {
2564 dev_err(dev, "UE: %s bit set\n",
2565 ue_status_hi_desc[i]);
2570 netif_carrier_off(netdev);
2573 static void be_msix_disable(struct be_adapter *adapter)
2575 if (msix_enabled(adapter)) {
2576 pci_disable_msix(adapter->pdev);
2577 adapter->num_msix_vec = 0;
2578 adapter->num_msix_roce_vec = 0;
2582 static int be_msix_enable(struct be_adapter *adapter)
2585 struct device *dev = &adapter->pdev->dev;
2587 /* If RoCE is supported, program the max number of NIC vectors that
2588 * may be configured via set-channels, along with vectors needed for
2589 * RoCe. Else, just program the number we'll use initially.
2591 if (be_roce_supported(adapter))
2592 num_vec = min_t(int, 2 * be_max_eqs(adapter),
2593 2 * num_online_cpus());
2595 num_vec = adapter->cfg_num_qs;
2597 for (i = 0; i < num_vec; i++)
2598 adapter->msix_entries[i].entry = i;
2600 num_vec = pci_enable_msix_range(adapter->pdev, adapter->msix_entries,
2601 MIN_MSIX_VECTORS, num_vec);
2605 if (be_roce_supported(adapter) && num_vec > MIN_MSIX_VECTORS) {
2606 adapter->num_msix_roce_vec = num_vec / 2;
2607 dev_info(dev, "enabled %d MSI-x vector(s) for RoCE\n",
2608 adapter->num_msix_roce_vec);
2611 adapter->num_msix_vec = num_vec - adapter->num_msix_roce_vec;
2613 dev_info(dev, "enabled %d MSI-x vector(s) for NIC\n",
2614 adapter->num_msix_vec);
2618 dev_warn(dev, "MSIx enable failed\n");
2620 /* INTx is not supported in VFs, so fail probe if enable_msix fails */
2621 if (!be_physfn(adapter))
2626 static inline int be_msix_vec_get(struct be_adapter *adapter,
2627 struct be_eq_obj *eqo)
2629 return adapter->msix_entries[eqo->msix_idx].vector;
2632 static int be_msix_register(struct be_adapter *adapter)
2634 struct net_device *netdev = adapter->netdev;
2635 struct be_eq_obj *eqo;
2638 for_all_evt_queues(adapter, eqo, i) {
2639 sprintf(eqo->desc, "%s-q%d", netdev->name, i);
2640 vec = be_msix_vec_get(adapter, eqo);
2641 status = request_irq(vec, be_msix, 0, eqo->desc, eqo);
2648 for (i--, eqo = &adapter->eq_obj[i]; i >= 0; i--, eqo--)
2649 free_irq(be_msix_vec_get(adapter, eqo), eqo);
2650 dev_warn(&adapter->pdev->dev, "MSIX Request IRQ failed - err %d\n",
2652 be_msix_disable(adapter);
2656 static int be_irq_register(struct be_adapter *adapter)
2658 struct net_device *netdev = adapter->netdev;
2661 if (msix_enabled(adapter)) {
2662 status = be_msix_register(adapter);
2665 /* INTx is not supported for VF */
2666 if (!be_physfn(adapter))
2670 /* INTx: only the first EQ is used */
2671 netdev->irq = adapter->pdev->irq;
2672 status = request_irq(netdev->irq, be_intx, IRQF_SHARED, netdev->name,
2673 &adapter->eq_obj[0]);
2675 dev_err(&adapter->pdev->dev,
2676 "INTx request IRQ failed - err %d\n", status);
2680 adapter->isr_registered = true;
2684 static void be_irq_unregister(struct be_adapter *adapter)
2686 struct net_device *netdev = adapter->netdev;
2687 struct be_eq_obj *eqo;
2690 if (!adapter->isr_registered)
2694 if (!msix_enabled(adapter)) {
2695 free_irq(netdev->irq, &adapter->eq_obj[0]);
2700 for_all_evt_queues(adapter, eqo, i)
2701 free_irq(be_msix_vec_get(adapter, eqo), eqo);
2704 adapter->isr_registered = false;
2707 static void be_rx_qs_destroy(struct be_adapter *adapter)
2709 struct be_queue_info *q;
2710 struct be_rx_obj *rxo;
2713 for_all_rx_queues(adapter, rxo, i) {
2716 be_cmd_rxq_destroy(adapter, q);
2717 be_rx_cq_clean(rxo);
2719 be_queue_free(adapter, q);
2723 static int be_close(struct net_device *netdev)
2725 struct be_adapter *adapter = netdev_priv(netdev);
2726 struct be_eq_obj *eqo;
2729 /* This protection is needed as be_close() may be called even when the
2730 * adapter is in cleared state (after eeh perm failure)
2732 if (!(adapter->flags & BE_FLAGS_SETUP_DONE))
2735 be_roce_dev_close(adapter);
2737 if (adapter->flags & BE_FLAGS_NAPI_ENABLED) {
2738 for_all_evt_queues(adapter, eqo, i) {
2739 napi_disable(&eqo->napi);
2740 be_disable_busy_poll(eqo);
2742 adapter->flags &= ~BE_FLAGS_NAPI_ENABLED;
2745 be_async_mcc_disable(adapter);
2747 /* Wait for all pending tx completions to arrive so that
2748 * all tx skbs are freed.
2750 netif_tx_disable(netdev);
2751 be_tx_compl_clean(adapter);
2753 be_rx_qs_destroy(adapter);
2755 for (i = 1; i < (adapter->uc_macs + 1); i++)
2756 be_cmd_pmac_del(adapter, adapter->if_handle,
2757 adapter->pmac_id[i], 0);
2758 adapter->uc_macs = 0;
2760 for_all_evt_queues(adapter, eqo, i) {
2761 if (msix_enabled(adapter))
2762 synchronize_irq(be_msix_vec_get(adapter, eqo));
2764 synchronize_irq(netdev->irq);
2768 be_irq_unregister(adapter);
2773 static int be_rx_qs_create(struct be_adapter *adapter)
2775 struct be_rx_obj *rxo;
2779 for_all_rx_queues(adapter, rxo, i) {
2780 rc = be_queue_alloc(adapter, &rxo->q, RX_Q_LEN,
2781 sizeof(struct be_eth_rx_d));
2786 /* The FW would like the default RXQ to be created first */
2787 rxo = default_rxo(adapter);
2788 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id, rx_frag_size,
2789 adapter->if_handle, false, &rxo->rss_id);
2793 for_all_rss_queues(adapter, rxo, i) {
2794 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id,
2795 rx_frag_size, adapter->if_handle,
2796 true, &rxo->rss_id);
2801 if (be_multi_rxq(adapter)) {
2802 for (j = 0; j < 128; j += adapter->num_rx_qs - 1) {
2803 for_all_rss_queues(adapter, rxo, i) {
2806 rsstable[j + i] = rxo->rss_id;
2809 adapter->rss_flags = RSS_ENABLE_TCP_IPV4 | RSS_ENABLE_IPV4 |
2810 RSS_ENABLE_TCP_IPV6 | RSS_ENABLE_IPV6;
2812 if (!BEx_chip(adapter))
2813 adapter->rss_flags |= RSS_ENABLE_UDP_IPV4 |
2814 RSS_ENABLE_UDP_IPV6;
2816 /* Disable RSS, if only default RX Q is created */
2817 adapter->rss_flags = RSS_ENABLE_NONE;
2820 rc = be_cmd_rss_config(adapter, rsstable, adapter->rss_flags,
2823 adapter->rss_flags = RSS_ENABLE_NONE;
2827 /* First time posting */
2828 for_all_rx_queues(adapter, rxo, i)
2829 be_post_rx_frags(rxo, GFP_KERNEL);
2833 static int be_open(struct net_device *netdev)
2835 struct be_adapter *adapter = netdev_priv(netdev);
2836 struct be_eq_obj *eqo;
2837 struct be_rx_obj *rxo;
2838 struct be_tx_obj *txo;
2842 status = be_rx_qs_create(adapter);
2846 status = be_irq_register(adapter);
2850 for_all_rx_queues(adapter, rxo, i)
2851 be_cq_notify(adapter, rxo->cq.id, true, 0);
2853 for_all_tx_queues(adapter, txo, i)
2854 be_cq_notify(adapter, txo->cq.id, true, 0);
2856 be_async_mcc_enable(adapter);
2858 for_all_evt_queues(adapter, eqo, i) {
2859 napi_enable(&eqo->napi);
2860 be_enable_busy_poll(eqo);
2861 be_eq_notify(adapter, eqo->q.id, true, false, 0);
2863 adapter->flags |= BE_FLAGS_NAPI_ENABLED;
2865 status = be_cmd_link_status_query(adapter, NULL, &link_status, 0);
2867 be_link_status_update(adapter, link_status);
2869 netif_tx_start_all_queues(netdev);
2870 be_roce_dev_open(adapter);
2872 #ifdef CONFIG_BE2NET_VXLAN
2873 if (skyhawk_chip(adapter))
2874 vxlan_get_rx_port(netdev);
2879 be_close(adapter->netdev);
2883 static int be_setup_wol(struct be_adapter *adapter, bool enable)
2885 struct be_dma_mem cmd;
2889 memset(mac, 0, ETH_ALEN);
2891 cmd.size = sizeof(struct be_cmd_req_acpi_wol_magic_config);
2892 cmd.va = dma_zalloc_coherent(&adapter->pdev->dev, cmd.size, &cmd.dma,
2898 status = pci_write_config_dword(adapter->pdev,
2899 PCICFG_PM_CONTROL_OFFSET, PCICFG_PM_CONTROL_MASK);
2901 dev_err(&adapter->pdev->dev,
2902 "Could not enable Wake-on-lan\n");
2903 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va,
2907 status = be_cmd_enable_magic_wol(adapter,
2908 adapter->netdev->dev_addr, &cmd);
2909 pci_enable_wake(adapter->pdev, PCI_D3hot, 1);
2910 pci_enable_wake(adapter->pdev, PCI_D3cold, 1);
2912 status = be_cmd_enable_magic_wol(adapter, mac, &cmd);
2913 pci_enable_wake(adapter->pdev, PCI_D3hot, 0);
2914 pci_enable_wake(adapter->pdev, PCI_D3cold, 0);
2917 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va, cmd.dma);
2922 * Generate a seed MAC address from the PF MAC Address using jhash.
2923 * MAC Address for VFs are assigned incrementally starting from the seed.
2924 * These addresses are programmed in the ASIC by the PF and the VF driver
2925 * queries for the MAC address during its probe.
2927 static int be_vf_eth_addr_config(struct be_adapter *adapter)
2932 struct be_vf_cfg *vf_cfg;
2934 be_vf_eth_addr_generate(adapter, mac);
2936 for_all_vfs(adapter, vf_cfg, vf) {
2937 if (BEx_chip(adapter))
2938 status = be_cmd_pmac_add(adapter, mac,
2940 &vf_cfg->pmac_id, vf + 1);
2942 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
2946 dev_err(&adapter->pdev->dev,
2947 "Mac address assignment failed for VF %d\n", vf);
2949 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
2956 static int be_vfs_mac_query(struct be_adapter *adapter)
2960 struct be_vf_cfg *vf_cfg;
2962 for_all_vfs(adapter, vf_cfg, vf) {
2963 status = be_cmd_get_active_mac(adapter, vf_cfg->pmac_id,
2964 mac, vf_cfg->if_handle,
2968 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
2973 static void be_vf_clear(struct be_adapter *adapter)
2975 struct be_vf_cfg *vf_cfg;
2978 if (pci_vfs_assigned(adapter->pdev)) {
2979 dev_warn(&adapter->pdev->dev,
2980 "VFs are assigned to VMs: not disabling VFs\n");
2984 pci_disable_sriov(adapter->pdev);
2986 for_all_vfs(adapter, vf_cfg, vf) {
2987 if (BEx_chip(adapter))
2988 be_cmd_pmac_del(adapter, vf_cfg->if_handle,
2989 vf_cfg->pmac_id, vf + 1);
2991 be_cmd_set_mac(adapter, NULL, vf_cfg->if_handle,
2994 be_cmd_if_destroy(adapter, vf_cfg->if_handle, vf + 1);
2997 kfree(adapter->vf_cfg);
2998 adapter->num_vfs = 0;
3001 static void be_clear_queues(struct be_adapter *adapter)
3003 be_mcc_queues_destroy(adapter);
3004 be_rx_cqs_destroy(adapter);
3005 be_tx_queues_destroy(adapter);
3006 be_evt_queues_destroy(adapter);
3009 static void be_cancel_worker(struct be_adapter *adapter)
3011 if (adapter->flags & BE_FLAGS_WORKER_SCHEDULED) {
3012 cancel_delayed_work_sync(&adapter->work);
3013 adapter->flags &= ~BE_FLAGS_WORKER_SCHEDULED;
3017 static void be_mac_clear(struct be_adapter *adapter)
3021 if (adapter->pmac_id) {
3022 for (i = 0; i < (adapter->uc_macs + 1); i++)
3023 be_cmd_pmac_del(adapter, adapter->if_handle,
3024 adapter->pmac_id[i], 0);
3025 adapter->uc_macs = 0;
3027 kfree(adapter->pmac_id);
3028 adapter->pmac_id = NULL;
3032 #ifdef CONFIG_BE2NET_VXLAN
3033 static void be_disable_vxlan_offloads(struct be_adapter *adapter)
3035 if (adapter->flags & BE_FLAGS_VXLAN_OFFLOADS)
3036 be_cmd_manage_iface(adapter, adapter->if_handle,
3037 OP_CONVERT_TUNNEL_TO_NORMAL);
3039 if (adapter->vxlan_port)
3040 be_cmd_set_vxlan_port(adapter, 0);
3042 adapter->flags &= ~BE_FLAGS_VXLAN_OFFLOADS;
3043 adapter->vxlan_port = 0;
3047 static int be_clear(struct be_adapter *adapter)
3049 be_cancel_worker(adapter);
3051 if (sriov_enabled(adapter))
3052 be_vf_clear(adapter);
3054 #ifdef CONFIG_BE2NET_VXLAN
3055 be_disable_vxlan_offloads(adapter);
3057 /* delete the primary mac along with the uc-mac list */
3058 be_mac_clear(adapter);
3060 be_cmd_if_destroy(adapter, adapter->if_handle, 0);
3062 be_clear_queues(adapter);
3064 be_msix_disable(adapter);
3065 adapter->flags &= ~BE_FLAGS_SETUP_DONE;
3069 static int be_vfs_if_create(struct be_adapter *adapter)
3071 struct be_resources res = {0};
3072 struct be_vf_cfg *vf_cfg;
3073 u32 cap_flags, en_flags, vf;
3076 cap_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3077 BE_IF_FLAGS_MULTICAST;
3079 for_all_vfs(adapter, vf_cfg, vf) {
3080 if (!BE3_chip(adapter)) {
3081 status = be_cmd_get_profile_config(adapter, &res,
3084 cap_flags = res.if_cap_flags;
3087 /* If a FW profile exists, then cap_flags are updated */
3088 en_flags = cap_flags & (BE_IF_FLAGS_UNTAGGED |
3089 BE_IF_FLAGS_BROADCAST | BE_IF_FLAGS_MULTICAST);
3090 status = be_cmd_if_create(adapter, cap_flags, en_flags,
3091 &vf_cfg->if_handle, vf + 1);
3099 static int be_vf_setup_init(struct be_adapter *adapter)
3101 struct be_vf_cfg *vf_cfg;
3104 adapter->vf_cfg = kcalloc(adapter->num_vfs, sizeof(*vf_cfg),
3106 if (!adapter->vf_cfg)
3109 for_all_vfs(adapter, vf_cfg, vf) {
3110 vf_cfg->if_handle = -1;
3111 vf_cfg->pmac_id = -1;
3116 static int be_vf_setup(struct be_adapter *adapter)
3118 struct device *dev = &adapter->pdev->dev;
3119 struct be_vf_cfg *vf_cfg;
3120 int status, old_vfs, vf;
3124 old_vfs = pci_num_vf(adapter->pdev);
3126 dev_info(dev, "%d VFs are already enabled\n", old_vfs);
3127 if (old_vfs != num_vfs)
3128 dev_warn(dev, "Ignoring num_vfs=%d setting\n", num_vfs);
3129 adapter->num_vfs = old_vfs;
3131 if (num_vfs > be_max_vfs(adapter))
3132 dev_info(dev, "Device supports %d VFs and not %d\n",
3133 be_max_vfs(adapter), num_vfs);
3134 adapter->num_vfs = min_t(u16, num_vfs, be_max_vfs(adapter));
3135 if (!adapter->num_vfs)
3139 status = be_vf_setup_init(adapter);
3144 for_all_vfs(adapter, vf_cfg, vf) {
3145 status = be_cmd_get_if_id(adapter, vf_cfg, vf);
3150 status = be_vfs_if_create(adapter);
3156 status = be_vfs_mac_query(adapter);
3160 status = be_vf_eth_addr_config(adapter);
3165 for_all_vfs(adapter, vf_cfg, vf) {
3166 /* Allow VFs to programs MAC/VLAN filters */
3167 status = be_cmd_get_fn_privileges(adapter, &privileges, vf + 1);
3168 if (!status && !(privileges & BE_PRIV_FILTMGMT)) {
3169 status = be_cmd_set_fn_privileges(adapter,
3174 dev_info(dev, "VF%d has FILTMGMT privilege\n",
3178 /* BE3 FW, by default, caps VF TX-rate to 100mbps.
3179 * Allow full available bandwidth
3181 if (BE3_chip(adapter) && !old_vfs)
3182 be_cmd_config_qos(adapter, 1000, vf + 1);
3184 status = be_cmd_link_status_query(adapter, &lnk_speed,
3187 vf_cfg->tx_rate = lnk_speed;
3190 be_cmd_enable_vf(adapter, vf + 1);
3191 be_cmd_set_logical_link_config(adapter,
3192 IFLA_VF_LINK_STATE_AUTO,
3198 status = pci_enable_sriov(adapter->pdev, adapter->num_vfs);
3200 dev_err(dev, "SRIOV enable failed\n");
3201 adapter->num_vfs = 0;
3207 dev_err(dev, "VF setup failed\n");
3208 be_vf_clear(adapter);
3212 /* Converting function_mode bits on BE3 to SH mc_type enums */
3214 static u8 be_convert_mc_type(u32 function_mode)
3216 if (function_mode & VNIC_MODE && function_mode & FLEX10_MODE)
3218 else if (function_mode & FLEX10_MODE)
3220 else if (function_mode & VNIC_MODE)
3222 else if (function_mode & UMC_ENABLED)
3228 /* On BE2/BE3 FW does not suggest the supported limits */
3229 static void BEx_get_resources(struct be_adapter *adapter,
3230 struct be_resources *res)
3232 struct pci_dev *pdev = adapter->pdev;
3233 bool use_sriov = false;
3236 if (be_physfn(adapter) && BE3_chip(adapter)) {
3237 be_cmd_get_profile_config(adapter, res, 0);
3238 /* Some old versions of BE3 FW don't report max_vfs value */
3239 if (res->max_vfs == 0) {
3240 max_vfs = pci_sriov_get_totalvfs(pdev);
3241 res->max_vfs = max_vfs > 0 ? min(MAX_VFS, max_vfs) : 0;
3243 use_sriov = res->max_vfs && sriov_want(adapter);
3246 if (be_physfn(adapter))
3247 res->max_uc_mac = BE_UC_PMAC_COUNT;
3249 res->max_uc_mac = BE_VF_UC_PMAC_COUNT;
3251 adapter->mc_type = be_convert_mc_type(adapter->function_mode);
3253 if (be_is_mc(adapter)) {
3254 /* Assuming that there are 4 channels per port,
3255 * when multi-channel is enabled
3257 if (be_is_qnq_mode(adapter))
3258 res->max_vlans = BE_NUM_VLANS_SUPPORTED/8;
3260 /* In a non-qnq multichannel mode, the pvid
3261 * takes up one vlan entry
3263 res->max_vlans = (BE_NUM_VLANS_SUPPORTED / 4) - 1;
3265 res->max_vlans = BE_NUM_VLANS_SUPPORTED;
3268 res->max_mcast_mac = BE_MAX_MC;
3270 /* 1) For BE3 1Gb ports, FW does not support multiple TXQs
3271 * 2) Create multiple TX rings on a BE3-R multi-channel interface
3272 * *only* if it is RSS-capable.
3274 if (BE2_chip(adapter) || use_sriov || (adapter->port_num > 1) ||
3275 !be_physfn(adapter) || (be_is_mc(adapter) &&
3276 !(adapter->function_caps & BE_FUNCTION_CAPS_RSS)))
3279 res->max_tx_qs = BE3_MAX_TX_QS;
3281 if ((adapter->function_caps & BE_FUNCTION_CAPS_RSS) &&
3282 !use_sriov && be_physfn(adapter))
3283 res->max_rss_qs = (adapter->be3_native) ?
3284 BE3_MAX_RSS_QS : BE2_MAX_RSS_QS;
3285 res->max_rx_qs = res->max_rss_qs + 1;
3287 if (be_physfn(adapter))
3288 res->max_evt_qs = (res->max_vfs > 0) ?
3289 BE3_SRIOV_MAX_EVT_QS : BE3_MAX_EVT_QS;
3291 res->max_evt_qs = 1;
3293 res->if_cap_flags = BE_IF_CAP_FLAGS_WANT;
3294 if (!(adapter->function_caps & BE_FUNCTION_CAPS_RSS))
3295 res->if_cap_flags &= ~BE_IF_FLAGS_RSS;
3298 static void be_setup_init(struct be_adapter *adapter)
3300 adapter->vlan_prio_bmap = 0xff;
3301 adapter->phy.link_speed = -1;
3302 adapter->if_handle = -1;
3303 adapter->be3_native = false;
3304 adapter->promiscuous = false;
3305 if (be_physfn(adapter))
3306 adapter->cmd_privileges = MAX_PRIVILEGES;
3308 adapter->cmd_privileges = MIN_PRIVILEGES;
3311 static int be_get_resources(struct be_adapter *adapter)
3313 struct device *dev = &adapter->pdev->dev;
3314 struct be_resources res = {0};
3317 if (BEx_chip(adapter)) {
3318 BEx_get_resources(adapter, &res);
3322 /* For Lancer, SH etc read per-function resource limits from FW.
3323 * GET_FUNC_CONFIG returns per function guaranteed limits.
3324 * GET_PROFILE_CONFIG returns PCI-E related limits PF-pool limits
3326 if (!BEx_chip(adapter)) {
3327 status = be_cmd_get_func_config(adapter, &res);
3331 /* If RoCE may be enabled stash away half the EQs for RoCE */
3332 if (be_roce_supported(adapter))
3333 res.max_evt_qs /= 2;
3336 if (be_physfn(adapter)) {
3337 status = be_cmd_get_profile_config(adapter, &res, 0);
3340 adapter->res.max_vfs = res.max_vfs;
3343 dev_info(dev, "Max: txqs %d, rxqs %d, rss %d, eqs %d, vfs %d\n",
3344 be_max_txqs(adapter), be_max_rxqs(adapter),
3345 be_max_rss(adapter), be_max_eqs(adapter),
3346 be_max_vfs(adapter));
3347 dev_info(dev, "Max: uc-macs %d, mc-macs %d, vlans %d\n",
3348 be_max_uc(adapter), be_max_mc(adapter),
3349 be_max_vlans(adapter));
3355 /* Routine to query per function resource limits */
3356 static int be_get_config(struct be_adapter *adapter)
3361 status = be_cmd_query_fw_cfg(adapter, &adapter->port_num,
3362 &adapter->function_mode,
3363 &adapter->function_caps,
3364 &adapter->asic_rev);
3368 if (be_physfn(adapter)) {
3369 status = be_cmd_get_active_profile(adapter, &profile_id);
3371 dev_info(&adapter->pdev->dev,
3372 "Using profile 0x%x\n", profile_id);
3375 status = be_get_resources(adapter);
3379 adapter->pmac_id = kcalloc(be_max_uc(adapter),
3380 sizeof(*adapter->pmac_id), GFP_KERNEL);
3381 if (!adapter->pmac_id)
3384 /* Sanitize cfg_num_qs based on HW and platform limits */
3385 adapter->cfg_num_qs = min(adapter->cfg_num_qs, be_max_qs(adapter));
3390 static int be_mac_setup(struct be_adapter *adapter)
3395 if (is_zero_ether_addr(adapter->netdev->dev_addr)) {
3396 status = be_cmd_get_perm_mac(adapter, mac);
3400 memcpy(adapter->netdev->dev_addr, mac, ETH_ALEN);
3401 memcpy(adapter->netdev->perm_addr, mac, ETH_ALEN);
3403 /* Maybe the HW was reset; dev_addr must be re-programmed */
3404 memcpy(mac, adapter->netdev->dev_addr, ETH_ALEN);
3407 /* For BE3-R VFs, the PF programs the initial MAC address */
3408 if (!(BEx_chip(adapter) && be_virtfn(adapter)))
3409 be_cmd_pmac_add(adapter, mac, adapter->if_handle,
3410 &adapter->pmac_id[0], 0);
3414 static void be_schedule_worker(struct be_adapter *adapter)
3416 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
3417 adapter->flags |= BE_FLAGS_WORKER_SCHEDULED;
3420 static int be_setup_queues(struct be_adapter *adapter)
3422 struct net_device *netdev = adapter->netdev;
3425 status = be_evt_queues_create(adapter);
3429 status = be_tx_qs_create(adapter);
3433 status = be_rx_cqs_create(adapter);
3437 status = be_mcc_queues_create(adapter);
3441 status = netif_set_real_num_rx_queues(netdev, adapter->num_rx_qs);
3445 status = netif_set_real_num_tx_queues(netdev, adapter->num_tx_qs);
3451 dev_err(&adapter->pdev->dev, "queue_setup failed\n");
3455 int be_update_queues(struct be_adapter *adapter)
3457 struct net_device *netdev = adapter->netdev;
3460 if (netif_running(netdev))
3463 be_cancel_worker(adapter);
3465 /* If any vectors have been shared with RoCE we cannot re-program
3468 if (!adapter->num_msix_roce_vec)
3469 be_msix_disable(adapter);
3471 be_clear_queues(adapter);
3473 if (!msix_enabled(adapter)) {
3474 status = be_msix_enable(adapter);
3479 status = be_setup_queues(adapter);
3483 be_schedule_worker(adapter);
3485 if (netif_running(netdev))
3486 status = be_open(netdev);
3491 static int be_setup(struct be_adapter *adapter)
3493 struct device *dev = &adapter->pdev->dev;
3494 u32 tx_fc, rx_fc, en_flags;
3497 be_setup_init(adapter);
3499 if (!lancer_chip(adapter))
3500 be_cmd_req_native_mode(adapter);
3502 status = be_get_config(adapter);
3506 status = be_msix_enable(adapter);
3510 en_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3511 BE_IF_FLAGS_MULTICAST | BE_IF_FLAGS_PASS_L3L4_ERRORS;
3512 if (adapter->function_caps & BE_FUNCTION_CAPS_RSS)
3513 en_flags |= BE_IF_FLAGS_RSS;
3514 en_flags = en_flags & be_if_cap_flags(adapter);
3515 status = be_cmd_if_create(adapter, be_if_cap_flags(adapter), en_flags,
3516 &adapter->if_handle, 0);
3520 /* Updating real_num_tx/rx_queues() requires rtnl_lock() */
3522 status = be_setup_queues(adapter);
3527 be_cmd_get_fn_privileges(adapter, &adapter->cmd_privileges, 0);
3529 status = be_mac_setup(adapter);
3533 be_cmd_get_fw_ver(adapter, adapter->fw_ver, adapter->fw_on_flash);
3535 if (BE2_chip(adapter) && fw_major_num(adapter->fw_ver) < 4) {
3536 dev_err(dev, "Firmware on card is old(%s), IRQs may not work.",
3538 dev_err(dev, "Please upgrade firmware to version >= 4.0\n");
3541 if (adapter->vlans_added)
3542 be_vid_config(adapter);
3544 be_set_rx_mode(adapter->netdev);
3546 be_cmd_get_acpi_wol_cap(adapter);
3548 be_cmd_get_flow_control(adapter, &tx_fc, &rx_fc);
3550 if (rx_fc != adapter->rx_fc || tx_fc != adapter->tx_fc)
3551 be_cmd_set_flow_control(adapter, adapter->tx_fc,
3554 if (be_physfn(adapter))
3555 be_cmd_set_logical_link_config(adapter,
3556 IFLA_VF_LINK_STATE_AUTO, 0);
3558 if (sriov_want(adapter)) {
3559 if (be_max_vfs(adapter))
3560 be_vf_setup(adapter);
3562 dev_warn(dev, "device doesn't support SRIOV\n");
3565 status = be_cmd_get_phy_info(adapter);
3566 if (!status && be_pause_supported(adapter))
3567 adapter->phy.fc_autoneg = 1;
3569 be_schedule_worker(adapter);
3570 adapter->flags |= BE_FLAGS_SETUP_DONE;
3577 #ifdef CONFIG_NET_POLL_CONTROLLER
3578 static void be_netpoll(struct net_device *netdev)
3580 struct be_adapter *adapter = netdev_priv(netdev);
3581 struct be_eq_obj *eqo;
3584 for_all_evt_queues(adapter, eqo, i) {
3585 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
3586 napi_schedule(&eqo->napi);
3593 #define FW_FILE_HDR_SIGN "ServerEngines Corp. "
3594 static char flash_cookie[2][16] = {"*** SE FLAS", "H DIRECTORY *** "};
3596 static bool be_flash_redboot(struct be_adapter *adapter,
3597 const u8 *p, u32 img_start, int image_size,
3604 crc_offset = hdr_size + img_start + image_size - 4;
3608 status = be_cmd_get_flash_crc(adapter, flashed_crc,
3611 dev_err(&adapter->pdev->dev,
3612 "could not get crc from flash, not flashing redboot\n");
3616 /*update redboot only if crc does not match*/
3617 if (!memcmp(flashed_crc, p, 4))
3623 static bool phy_flashing_required(struct be_adapter *adapter)
3625 return (adapter->phy.phy_type == TN_8022 &&
3626 adapter->phy.interface_type == PHY_TYPE_BASET_10GB);
3629 static bool is_comp_in_ufi(struct be_adapter *adapter,
3630 struct flash_section_info *fsec, int type)
3632 int i = 0, img_type = 0;
3633 struct flash_section_info_g2 *fsec_g2 = NULL;
3635 if (BE2_chip(adapter))
3636 fsec_g2 = (struct flash_section_info_g2 *)fsec;
3638 for (i = 0; i < MAX_FLASH_COMP; i++) {
3640 img_type = le32_to_cpu(fsec_g2->fsec_entry[i].type);
3642 img_type = le32_to_cpu(fsec->fsec_entry[i].type);
3644 if (img_type == type)
3651 static struct flash_section_info *get_fsec_info(struct be_adapter *adapter,
3653 const struct firmware *fw)
3655 struct flash_section_info *fsec = NULL;
3656 const u8 *p = fw->data;
3659 while (p < (fw->data + fw->size)) {
3660 fsec = (struct flash_section_info *)p;
3661 if (!memcmp(flash_cookie, fsec->cookie, sizeof(flash_cookie)))
3668 static int be_flash(struct be_adapter *adapter, const u8 *img,
3669 struct be_dma_mem *flash_cmd, int optype, int img_size)
3671 u32 total_bytes = 0, flash_op, num_bytes = 0;
3673 struct be_cmd_write_flashrom *req = flash_cmd->va;
3675 total_bytes = img_size;
3676 while (total_bytes) {
3677 num_bytes = min_t(u32, 32*1024, total_bytes);
3679 total_bytes -= num_bytes;
3682 if (optype == OPTYPE_PHY_FW)
3683 flash_op = FLASHROM_OPER_PHY_FLASH;
3685 flash_op = FLASHROM_OPER_FLASH;
3687 if (optype == OPTYPE_PHY_FW)
3688 flash_op = FLASHROM_OPER_PHY_SAVE;
3690 flash_op = FLASHROM_OPER_SAVE;
3693 memcpy(req->data_buf, img, num_bytes);
3695 status = be_cmd_write_flashrom(adapter, flash_cmd, optype,
3696 flash_op, num_bytes);
3698 if (status == ILLEGAL_IOCTL_REQ &&
3699 optype == OPTYPE_PHY_FW)
3701 dev_err(&adapter->pdev->dev,
3702 "cmd to write to flash rom failed.\n");
3709 /* For BE2, BE3 and BE3-R */
3710 static int be_flash_BEx(struct be_adapter *adapter,
3711 const struct firmware *fw,
3712 struct be_dma_mem *flash_cmd,
3716 int status = 0, i, filehdr_size = 0;
3717 int img_hdrs_size = (num_of_images * sizeof(struct image_hdr));
3718 const u8 *p = fw->data;
3719 const struct flash_comp *pflashcomp;
3720 int num_comp, redboot;
3721 struct flash_section_info *fsec = NULL;
3723 struct flash_comp gen3_flash_types[] = {
3724 { FLASH_iSCSI_PRIMARY_IMAGE_START_g3, OPTYPE_ISCSI_ACTIVE,
3725 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_iSCSI},
3726 { FLASH_REDBOOT_START_g3, OPTYPE_REDBOOT,
3727 FLASH_REDBOOT_IMAGE_MAX_SIZE_g3, IMAGE_BOOT_CODE},
3728 { FLASH_iSCSI_BIOS_START_g3, OPTYPE_BIOS,
3729 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_ISCSI},
3730 { FLASH_PXE_BIOS_START_g3, OPTYPE_PXE_BIOS,
3731 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_PXE},
3732 { FLASH_FCoE_BIOS_START_g3, OPTYPE_FCOE_BIOS,
3733 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_FCoE},
3734 { FLASH_iSCSI_BACKUP_IMAGE_START_g3, OPTYPE_ISCSI_BACKUP,
3735 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_iSCSI},
3736 { FLASH_FCoE_PRIMARY_IMAGE_START_g3, OPTYPE_FCOE_FW_ACTIVE,
3737 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_FCoE},
3738 { FLASH_FCoE_BACKUP_IMAGE_START_g3, OPTYPE_FCOE_FW_BACKUP,
3739 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_FCoE},
3740 { FLASH_NCSI_START_g3, OPTYPE_NCSI_FW,
3741 FLASH_NCSI_IMAGE_MAX_SIZE_g3, IMAGE_NCSI},
3742 { FLASH_PHY_FW_START_g3, OPTYPE_PHY_FW,
3743 FLASH_PHY_FW_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_PHY}
3746 struct flash_comp gen2_flash_types[] = {
3747 { FLASH_iSCSI_PRIMARY_IMAGE_START_g2, OPTYPE_ISCSI_ACTIVE,
3748 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_iSCSI},
3749 { FLASH_REDBOOT_START_g2, OPTYPE_REDBOOT,
3750 FLASH_REDBOOT_IMAGE_MAX_SIZE_g2, IMAGE_BOOT_CODE},
3751 { FLASH_iSCSI_BIOS_START_g2, OPTYPE_BIOS,
3752 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_ISCSI},
3753 { FLASH_PXE_BIOS_START_g2, OPTYPE_PXE_BIOS,
3754 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_PXE},
3755 { FLASH_FCoE_BIOS_START_g2, OPTYPE_FCOE_BIOS,
3756 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_FCoE},
3757 { FLASH_iSCSI_BACKUP_IMAGE_START_g2, OPTYPE_ISCSI_BACKUP,
3758 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_iSCSI},
3759 { FLASH_FCoE_PRIMARY_IMAGE_START_g2, OPTYPE_FCOE_FW_ACTIVE,
3760 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_FCoE},
3761 { FLASH_FCoE_BACKUP_IMAGE_START_g2, OPTYPE_FCOE_FW_BACKUP,
3762 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_FCoE}
3765 if (BE3_chip(adapter)) {
3766 pflashcomp = gen3_flash_types;
3767 filehdr_size = sizeof(struct flash_file_hdr_g3);
3768 num_comp = ARRAY_SIZE(gen3_flash_types);
3770 pflashcomp = gen2_flash_types;
3771 filehdr_size = sizeof(struct flash_file_hdr_g2);
3772 num_comp = ARRAY_SIZE(gen2_flash_types);
3775 /* Get flash section info*/
3776 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
3778 dev_err(&adapter->pdev->dev,
3779 "Invalid Cookie. UFI corrupted ?\n");
3782 for (i = 0; i < num_comp; i++) {
3783 if (!is_comp_in_ufi(adapter, fsec, pflashcomp[i].img_type))
3786 if ((pflashcomp[i].optype == OPTYPE_NCSI_FW) &&
3787 memcmp(adapter->fw_ver, "3.102.148.0", 11) < 0)
3790 if (pflashcomp[i].optype == OPTYPE_PHY_FW &&
3791 !phy_flashing_required(adapter))
3794 if (pflashcomp[i].optype == OPTYPE_REDBOOT) {
3795 redboot = be_flash_redboot(adapter, fw->data,
3796 pflashcomp[i].offset, pflashcomp[i].size,
3797 filehdr_size + img_hdrs_size);
3803 p += filehdr_size + pflashcomp[i].offset + img_hdrs_size;
3804 if (p + pflashcomp[i].size > fw->data + fw->size)
3807 status = be_flash(adapter, p, flash_cmd, pflashcomp[i].optype,
3808 pflashcomp[i].size);
3810 dev_err(&adapter->pdev->dev,
3811 "Flashing section type %d failed.\n",
3812 pflashcomp[i].img_type);
3819 static int be_flash_skyhawk(struct be_adapter *adapter,
3820 const struct firmware *fw,
3821 struct be_dma_mem *flash_cmd, int num_of_images)
3823 int status = 0, i, filehdr_size = 0;
3824 int img_offset, img_size, img_optype, redboot;
3825 int img_hdrs_size = num_of_images * sizeof(struct image_hdr);
3826 const u8 *p = fw->data;
3827 struct flash_section_info *fsec = NULL;
3829 filehdr_size = sizeof(struct flash_file_hdr_g3);
3830 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
3832 dev_err(&adapter->pdev->dev,
3833 "Invalid Cookie. UFI corrupted ?\n");
3837 for (i = 0; i < le32_to_cpu(fsec->fsec_hdr.num_images); i++) {
3838 img_offset = le32_to_cpu(fsec->fsec_entry[i].offset);
3839 img_size = le32_to_cpu(fsec->fsec_entry[i].pad_size);
3841 switch (le32_to_cpu(fsec->fsec_entry[i].type)) {
3842 case IMAGE_FIRMWARE_iSCSI:
3843 img_optype = OPTYPE_ISCSI_ACTIVE;
3845 case IMAGE_BOOT_CODE:
3846 img_optype = OPTYPE_REDBOOT;
3848 case IMAGE_OPTION_ROM_ISCSI:
3849 img_optype = OPTYPE_BIOS;
3851 case IMAGE_OPTION_ROM_PXE:
3852 img_optype = OPTYPE_PXE_BIOS;
3854 case IMAGE_OPTION_ROM_FCoE:
3855 img_optype = OPTYPE_FCOE_BIOS;
3857 case IMAGE_FIRMWARE_BACKUP_iSCSI:
3858 img_optype = OPTYPE_ISCSI_BACKUP;
3861 img_optype = OPTYPE_NCSI_FW;
3867 if (img_optype == OPTYPE_REDBOOT) {
3868 redboot = be_flash_redboot(adapter, fw->data,
3869 img_offset, img_size,
3870 filehdr_size + img_hdrs_size);
3876 p += filehdr_size + img_offset + img_hdrs_size;
3877 if (p + img_size > fw->data + fw->size)
3880 status = be_flash(adapter, p, flash_cmd, img_optype, img_size);
3882 dev_err(&adapter->pdev->dev,
3883 "Flashing section type %d failed.\n",
3884 fsec->fsec_entry[i].type);
3891 static int lancer_fw_download(struct be_adapter *adapter,
3892 const struct firmware *fw)
3894 #define LANCER_FW_DOWNLOAD_CHUNK (32 * 1024)
3895 #define LANCER_FW_DOWNLOAD_LOCATION "/prg"
3896 struct be_dma_mem flash_cmd;
3897 const u8 *data_ptr = NULL;
3898 u8 *dest_image_ptr = NULL;
3899 size_t image_size = 0;
3901 u32 data_written = 0;
3907 if (!IS_ALIGNED(fw->size, sizeof(u32))) {
3908 dev_err(&adapter->pdev->dev,
3909 "FW Image not properly aligned. "
3910 "Length must be 4 byte aligned.\n");
3912 goto lancer_fw_exit;
3915 flash_cmd.size = sizeof(struct lancer_cmd_req_write_object)
3916 + LANCER_FW_DOWNLOAD_CHUNK;
3917 flash_cmd.va = dma_alloc_coherent(&adapter->pdev->dev, flash_cmd.size,
3918 &flash_cmd.dma, GFP_KERNEL);
3919 if (!flash_cmd.va) {
3921 goto lancer_fw_exit;
3924 dest_image_ptr = flash_cmd.va +
3925 sizeof(struct lancer_cmd_req_write_object);
3926 image_size = fw->size;
3927 data_ptr = fw->data;
3929 while (image_size) {
3930 chunk_size = min_t(u32, image_size, LANCER_FW_DOWNLOAD_CHUNK);
3932 /* Copy the image chunk content. */
3933 memcpy(dest_image_ptr, data_ptr, chunk_size);
3935 status = lancer_cmd_write_object(adapter, &flash_cmd,
3937 LANCER_FW_DOWNLOAD_LOCATION,
3938 &data_written, &change_status,
3943 offset += data_written;
3944 data_ptr += data_written;
3945 image_size -= data_written;
3949 /* Commit the FW written */
3950 status = lancer_cmd_write_object(adapter, &flash_cmd,
3952 LANCER_FW_DOWNLOAD_LOCATION,
3953 &data_written, &change_status,
3957 dma_free_coherent(&adapter->pdev->dev, flash_cmd.size, flash_cmd.va,
3960 dev_err(&adapter->pdev->dev,
3961 "Firmware load error. "
3962 "Status code: 0x%x Additional Status: 0x%x\n",
3963 status, add_status);
3964 goto lancer_fw_exit;
3967 if (change_status == LANCER_FW_RESET_NEEDED) {
3968 dev_info(&adapter->pdev->dev,
3969 "Resetting adapter to activate new FW\n");
3970 status = lancer_physdev_ctrl(adapter,
3971 PHYSDEV_CONTROL_FW_RESET_MASK);
3973 dev_err(&adapter->pdev->dev,
3974 "Adapter busy for FW reset.\n"
3975 "New FW will not be active.\n");
3976 goto lancer_fw_exit;
3978 } else if (change_status != LANCER_NO_RESET_NEEDED) {
3979 dev_err(&adapter->pdev->dev,
3980 "System reboot required for new FW"
3984 dev_info(&adapter->pdev->dev, "Firmware flashed successfully\n");
3991 #define UFI_TYPE3R 10
3993 static int be_get_ufi_type(struct be_adapter *adapter,
3994 struct flash_file_hdr_g3 *fhdr)
3997 goto be_get_ufi_exit;
3999 if (skyhawk_chip(adapter) && fhdr->build[0] == '4')
4001 else if (BE3_chip(adapter) && fhdr->build[0] == '3') {
4002 if (fhdr->asic_type_rev == 0x10)
4006 } else if (BE2_chip(adapter) && fhdr->build[0] == '2')
4010 dev_err(&adapter->pdev->dev,
4011 "UFI and Interface are not compatible for flashing\n");
4015 static int be_fw_download(struct be_adapter *adapter, const struct firmware* fw)
4017 struct flash_file_hdr_g3 *fhdr3;
4018 struct image_hdr *img_hdr_ptr = NULL;
4019 struct be_dma_mem flash_cmd;
4021 int status = 0, i = 0, num_imgs = 0, ufi_type = 0;
4023 flash_cmd.size = sizeof(struct be_cmd_write_flashrom);
4024 flash_cmd.va = dma_alloc_coherent(&adapter->pdev->dev, flash_cmd.size,
4025 &flash_cmd.dma, GFP_KERNEL);
4026 if (!flash_cmd.va) {
4032 fhdr3 = (struct flash_file_hdr_g3 *)p;
4034 ufi_type = be_get_ufi_type(adapter, fhdr3);
4036 num_imgs = le32_to_cpu(fhdr3->num_imgs);
4037 for (i = 0; i < num_imgs; i++) {
4038 img_hdr_ptr = (struct image_hdr *)(fw->data +
4039 (sizeof(struct flash_file_hdr_g3) +
4040 i * sizeof(struct image_hdr)));
4041 if (le32_to_cpu(img_hdr_ptr->imageid) == 1) {
4044 status = be_flash_skyhawk(adapter, fw,
4045 &flash_cmd, num_imgs);
4048 status = be_flash_BEx(adapter, fw, &flash_cmd,
4052 /* Do not flash this ufi on BE3-R cards */
4053 if (adapter->asic_rev < 0x10)
4054 status = be_flash_BEx(adapter, fw,
4059 dev_err(&adapter->pdev->dev,
4060 "Can't load BE3 UFI on BE3R\n");
4066 if (ufi_type == UFI_TYPE2)
4067 status = be_flash_BEx(adapter, fw, &flash_cmd, 0);
4068 else if (ufi_type == -1)
4071 dma_free_coherent(&adapter->pdev->dev, flash_cmd.size, flash_cmd.va,
4074 dev_err(&adapter->pdev->dev, "Firmware load error\n");
4078 dev_info(&adapter->pdev->dev, "Firmware flashed successfully\n");
4084 int be_load_fw(struct be_adapter *adapter, u8 *fw_file)
4086 const struct firmware *fw;
4089 if (!netif_running(adapter->netdev)) {
4090 dev_err(&adapter->pdev->dev,
4091 "Firmware load not allowed (interface is down)\n");
4095 status = request_firmware(&fw, fw_file, &adapter->pdev->dev);
4099 dev_info(&adapter->pdev->dev, "Flashing firmware file %s\n", fw_file);
4101 if (lancer_chip(adapter))
4102 status = lancer_fw_download(adapter, fw);
4104 status = be_fw_download(adapter, fw);
4107 be_cmd_get_fw_ver(adapter, adapter->fw_ver,
4108 adapter->fw_on_flash);
4111 release_firmware(fw);
4115 static int be_ndo_bridge_setlink(struct net_device *dev,
4116 struct nlmsghdr *nlh)
4118 struct be_adapter *adapter = netdev_priv(dev);
4119 struct nlattr *attr, *br_spec;
4124 if (!sriov_enabled(adapter))
4127 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4129 nla_for_each_nested(attr, br_spec, rem) {
4130 if (nla_type(attr) != IFLA_BRIDGE_MODE)
4133 mode = nla_get_u16(attr);
4134 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
4137 status = be_cmd_set_hsw_config(adapter, 0, 0,
4139 mode == BRIDGE_MODE_VEPA ?
4140 PORT_FWD_TYPE_VEPA :
4145 dev_info(&adapter->pdev->dev, "enabled switch mode: %s\n",
4146 mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
4151 dev_err(&adapter->pdev->dev, "Failed to set switch mode %s\n",
4152 mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
4157 static int be_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4158 struct net_device *dev,
4161 struct be_adapter *adapter = netdev_priv(dev);
4165 if (!sriov_enabled(adapter))
4168 /* BE and Lancer chips support VEB mode only */
4169 if (BEx_chip(adapter) || lancer_chip(adapter)) {
4170 hsw_mode = PORT_FWD_TYPE_VEB;
4172 status = be_cmd_get_hsw_config(adapter, NULL, 0,
4173 adapter->if_handle, &hsw_mode);
4178 return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
4179 hsw_mode == PORT_FWD_TYPE_VEPA ?
4180 BRIDGE_MODE_VEPA : BRIDGE_MODE_VEB);
4183 #ifdef CONFIG_BE2NET_VXLAN
4184 static void be_add_vxlan_port(struct net_device *netdev, sa_family_t sa_family,
4187 struct be_adapter *adapter = netdev_priv(netdev);
4188 struct device *dev = &adapter->pdev->dev;
4191 if (lancer_chip(adapter) || BEx_chip(adapter))
4194 if (adapter->flags & BE_FLAGS_VXLAN_OFFLOADS) {
4195 dev_warn(dev, "Cannot add UDP port %d for VxLAN offloads\n",
4198 "Only one UDP port supported for VxLAN offloads\n");
4202 status = be_cmd_manage_iface(adapter, adapter->if_handle,
4203 OP_CONVERT_NORMAL_TO_TUNNEL);
4205 dev_warn(dev, "Failed to convert normal interface to tunnel\n");
4209 status = be_cmd_set_vxlan_port(adapter, port);
4211 dev_warn(dev, "Failed to add VxLAN port\n");
4214 adapter->flags |= BE_FLAGS_VXLAN_OFFLOADS;
4215 adapter->vxlan_port = port;
4217 dev_info(dev, "Enabled VxLAN offloads for UDP port %d\n",
4221 be_disable_vxlan_offloads(adapter);
4225 static void be_del_vxlan_port(struct net_device *netdev, sa_family_t sa_family,
4228 struct be_adapter *adapter = netdev_priv(netdev);
4230 if (lancer_chip(adapter) || BEx_chip(adapter))
4233 if (adapter->vxlan_port != port)
4236 be_disable_vxlan_offloads(adapter);
4238 dev_info(&adapter->pdev->dev,
4239 "Disabled VxLAN offloads for UDP port %d\n",
4244 static const struct net_device_ops be_netdev_ops = {
4245 .ndo_open = be_open,
4246 .ndo_stop = be_close,
4247 .ndo_start_xmit = be_xmit,
4248 .ndo_set_rx_mode = be_set_rx_mode,
4249 .ndo_set_mac_address = be_mac_addr_set,
4250 .ndo_change_mtu = be_change_mtu,
4251 .ndo_get_stats64 = be_get_stats64,
4252 .ndo_validate_addr = eth_validate_addr,
4253 .ndo_vlan_rx_add_vid = be_vlan_add_vid,
4254 .ndo_vlan_rx_kill_vid = be_vlan_rem_vid,
4255 .ndo_set_vf_mac = be_set_vf_mac,
4256 .ndo_set_vf_vlan = be_set_vf_vlan,
4257 .ndo_set_vf_tx_rate = be_set_vf_tx_rate,
4258 .ndo_get_vf_config = be_get_vf_config,
4259 .ndo_set_vf_link_state = be_set_vf_link_state,
4260 #ifdef CONFIG_NET_POLL_CONTROLLER
4261 .ndo_poll_controller = be_netpoll,
4263 .ndo_bridge_setlink = be_ndo_bridge_setlink,
4264 .ndo_bridge_getlink = be_ndo_bridge_getlink,
4265 #ifdef CONFIG_NET_RX_BUSY_POLL
4266 .ndo_busy_poll = be_busy_poll,
4268 #ifdef CONFIG_BE2NET_VXLAN
4269 .ndo_add_vxlan_port = be_add_vxlan_port,
4270 .ndo_del_vxlan_port = be_del_vxlan_port,
4274 static void be_netdev_init(struct net_device *netdev)
4276 struct be_adapter *adapter = netdev_priv(netdev);
4278 if (skyhawk_chip(adapter)) {
4279 netdev->hw_enc_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
4280 NETIF_F_TSO | NETIF_F_TSO6 |
4281 NETIF_F_GSO_UDP_TUNNEL;
4282 netdev->hw_features |= NETIF_F_GSO_UDP_TUNNEL;
4284 netdev->hw_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
4285 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM | NETIF_F_RXCSUM |
4286 NETIF_F_HW_VLAN_CTAG_TX;
4287 if (be_multi_rxq(adapter))
4288 netdev->hw_features |= NETIF_F_RXHASH;
4290 netdev->features |= netdev->hw_features |
4291 NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_FILTER;
4293 netdev->vlan_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
4294 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
4296 netdev->priv_flags |= IFF_UNICAST_FLT;
4298 netdev->flags |= IFF_MULTICAST;
4300 netif_set_gso_max_size(netdev, 65535 - ETH_HLEN);
4302 netdev->netdev_ops = &be_netdev_ops;
4304 SET_ETHTOOL_OPS(netdev, &be_ethtool_ops);
4307 static void be_unmap_pci_bars(struct be_adapter *adapter)
4310 pci_iounmap(adapter->pdev, adapter->csr);
4312 pci_iounmap(adapter->pdev, adapter->db);
4315 static int db_bar(struct be_adapter *adapter)
4317 if (lancer_chip(adapter) || !be_physfn(adapter))
4323 static int be_roce_map_pci_bars(struct be_adapter *adapter)
4325 if (skyhawk_chip(adapter)) {
4326 adapter->roce_db.size = 4096;
4327 adapter->roce_db.io_addr = pci_resource_start(adapter->pdev,
4329 adapter->roce_db.total_size = pci_resource_len(adapter->pdev,
4335 static int be_map_pci_bars(struct be_adapter *adapter)
4339 if (BEx_chip(adapter) && be_physfn(adapter)) {
4340 adapter->csr = pci_iomap(adapter->pdev, 2, 0);
4341 if (adapter->csr == NULL)
4345 addr = pci_iomap(adapter->pdev, db_bar(adapter), 0);
4350 be_roce_map_pci_bars(adapter);
4354 be_unmap_pci_bars(adapter);
4358 static void be_ctrl_cleanup(struct be_adapter *adapter)
4360 struct be_dma_mem *mem = &adapter->mbox_mem_alloced;
4362 be_unmap_pci_bars(adapter);
4365 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
4368 mem = &adapter->rx_filter;
4370 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
4374 static int be_ctrl_init(struct be_adapter *adapter)
4376 struct be_dma_mem *mbox_mem_alloc = &adapter->mbox_mem_alloced;
4377 struct be_dma_mem *mbox_mem_align = &adapter->mbox_mem;
4378 struct be_dma_mem *rx_filter = &adapter->rx_filter;
4382 pci_read_config_dword(adapter->pdev, SLI_INTF_REG_OFFSET, &sli_intf);
4383 adapter->sli_family = (sli_intf & SLI_INTF_FAMILY_MASK) >>
4384 SLI_INTF_FAMILY_SHIFT;
4385 adapter->virtfn = (sli_intf & SLI_INTF_FT_MASK) ? 1 : 0;
4387 status = be_map_pci_bars(adapter);
4391 mbox_mem_alloc->size = sizeof(struct be_mcc_mailbox) + 16;
4392 mbox_mem_alloc->va = dma_alloc_coherent(&adapter->pdev->dev,
4393 mbox_mem_alloc->size,
4394 &mbox_mem_alloc->dma,
4396 if (!mbox_mem_alloc->va) {
4398 goto unmap_pci_bars;
4400 mbox_mem_align->size = sizeof(struct be_mcc_mailbox);
4401 mbox_mem_align->va = PTR_ALIGN(mbox_mem_alloc->va, 16);
4402 mbox_mem_align->dma = PTR_ALIGN(mbox_mem_alloc->dma, 16);
4403 memset(mbox_mem_align->va, 0, sizeof(struct be_mcc_mailbox));
4405 rx_filter->size = sizeof(struct be_cmd_req_rx_filter);
4406 rx_filter->va = dma_zalloc_coherent(&adapter->pdev->dev,
4407 rx_filter->size, &rx_filter->dma,
4409 if (rx_filter->va == NULL) {
4414 mutex_init(&adapter->mbox_lock);
4415 spin_lock_init(&adapter->mcc_lock);
4416 spin_lock_init(&adapter->mcc_cq_lock);
4418 init_completion(&adapter->et_cmd_compl);
4419 pci_save_state(adapter->pdev);
4423 dma_free_coherent(&adapter->pdev->dev, mbox_mem_alloc->size,
4424 mbox_mem_alloc->va, mbox_mem_alloc->dma);
4427 be_unmap_pci_bars(adapter);
4433 static void be_stats_cleanup(struct be_adapter *adapter)
4435 struct be_dma_mem *cmd = &adapter->stats_cmd;
4438 dma_free_coherent(&adapter->pdev->dev, cmd->size,
4442 static int be_stats_init(struct be_adapter *adapter)
4444 struct be_dma_mem *cmd = &adapter->stats_cmd;
4446 if (lancer_chip(adapter))
4447 cmd->size = sizeof(struct lancer_cmd_req_pport_stats);
4448 else if (BE2_chip(adapter))
4449 cmd->size = sizeof(struct be_cmd_req_get_stats_v0);
4450 else if (BE3_chip(adapter))
4451 cmd->size = sizeof(struct be_cmd_req_get_stats_v1);
4453 /* ALL non-BE ASICs */
4454 cmd->size = sizeof(struct be_cmd_req_get_stats_v2);
4456 cmd->va = dma_zalloc_coherent(&adapter->pdev->dev, cmd->size, &cmd->dma,
4458 if (cmd->va == NULL)
4463 static void be_remove(struct pci_dev *pdev)
4465 struct be_adapter *adapter = pci_get_drvdata(pdev);
4470 be_roce_dev_remove(adapter);
4471 be_intr_set(adapter, false);
4473 cancel_delayed_work_sync(&adapter->func_recovery_work);
4475 unregister_netdev(adapter->netdev);
4479 /* tell fw we're done with firing cmds */
4480 be_cmd_fw_clean(adapter);
4482 be_stats_cleanup(adapter);
4484 be_ctrl_cleanup(adapter);
4486 pci_disable_pcie_error_reporting(pdev);
4488 pci_release_regions(pdev);
4489 pci_disable_device(pdev);
4491 free_netdev(adapter->netdev);
4494 static int be_get_initial_config(struct be_adapter *adapter)
4498 status = be_cmd_get_cntl_attributes(adapter);
4502 /* Must be a power of 2 or else MODULO will BUG_ON */
4503 adapter->be_get_temp_freq = 64;
4505 if (BEx_chip(adapter)) {
4506 level = be_cmd_get_fw_log_level(adapter);
4507 adapter->msg_enable =
4508 level <= FW_LOG_LEVEL_DEFAULT ? NETIF_MSG_HW : 0;
4511 adapter->cfg_num_qs = netif_get_num_default_rss_queues();
4515 static int lancer_recover_func(struct be_adapter *adapter)
4517 struct device *dev = &adapter->pdev->dev;
4520 status = lancer_test_and_set_rdy_state(adapter);
4524 if (netif_running(adapter->netdev))
4525 be_close(adapter->netdev);
4529 be_clear_all_error(adapter);
4531 status = be_setup(adapter);
4535 if (netif_running(adapter->netdev)) {
4536 status = be_open(adapter->netdev);
4541 dev_err(dev, "Adapter recovery successful\n");
4544 if (status == -EAGAIN)
4545 dev_err(dev, "Waiting for resource provisioning\n");
4547 dev_err(dev, "Adapter recovery failed\n");
4552 static void be_func_recovery_task(struct work_struct *work)
4554 struct be_adapter *adapter =
4555 container_of(work, struct be_adapter, func_recovery_work.work);
4558 be_detect_error(adapter);
4560 if (adapter->hw_error && lancer_chip(adapter)) {
4563 netif_device_detach(adapter->netdev);
4566 status = lancer_recover_func(adapter);
4568 netif_device_attach(adapter->netdev);
4571 /* In Lancer, for all errors other than provisioning error (-EAGAIN),
4572 * no need to attempt further recovery.
4574 if (!status || status == -EAGAIN)
4575 schedule_delayed_work(&adapter->func_recovery_work,
4576 msecs_to_jiffies(1000));
4579 static void be_worker(struct work_struct *work)
4581 struct be_adapter *adapter =
4582 container_of(work, struct be_adapter, work.work);
4583 struct be_rx_obj *rxo;
4586 /* when interrupts are not yet enabled, just reap any pending
4587 * mcc completions */
4588 if (!netif_running(adapter->netdev)) {
4590 be_process_mcc(adapter);
4595 if (!adapter->stats_cmd_sent) {
4596 if (lancer_chip(adapter))
4597 lancer_cmd_get_pport_stats(adapter,
4598 &adapter->stats_cmd);
4600 be_cmd_get_stats(adapter, &adapter->stats_cmd);
4603 if (be_physfn(adapter) &&
4604 MODULO(adapter->work_counter, adapter->be_get_temp_freq) == 0)
4605 be_cmd_get_die_temperature(adapter);
4607 for_all_rx_queues(adapter, rxo, i) {
4608 /* Replenish RX-queues starved due to memory
4609 * allocation failures.
4611 if (rxo->rx_post_starved)
4612 be_post_rx_frags(rxo, GFP_KERNEL);
4615 be_eqd_update(adapter);
4618 adapter->work_counter++;
4619 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
4622 /* If any VFs are already enabled don't FLR the PF */
4623 static bool be_reset_required(struct be_adapter *adapter)
4625 return pci_num_vf(adapter->pdev) ? false : true;
4628 static char *mc_name(struct be_adapter *adapter)
4630 char *str = ""; /* default */
4632 switch (adapter->mc_type) {
4658 static inline char *func_name(struct be_adapter *adapter)
4660 return be_physfn(adapter) ? "PF" : "VF";
4663 static int be_probe(struct pci_dev *pdev, const struct pci_device_id *pdev_id)
4666 struct be_adapter *adapter;
4667 struct net_device *netdev;
4670 status = pci_enable_device(pdev);
4674 status = pci_request_regions(pdev, DRV_NAME);
4677 pci_set_master(pdev);
4679 netdev = alloc_etherdev_mqs(sizeof(*adapter), MAX_TX_QS, MAX_RX_QS);
4680 if (netdev == NULL) {
4684 adapter = netdev_priv(netdev);
4685 adapter->pdev = pdev;
4686 pci_set_drvdata(pdev, adapter);
4687 adapter->netdev = netdev;
4688 SET_NETDEV_DEV(netdev, &pdev->dev);
4690 status = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
4692 netdev->features |= NETIF_F_HIGHDMA;
4694 status = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
4696 dev_err(&pdev->dev, "Could not set PCI DMA Mask\n");
4701 if (be_physfn(adapter)) {
4702 status = pci_enable_pcie_error_reporting(pdev);
4704 dev_info(&pdev->dev, "PCIe error reporting enabled\n");
4707 status = be_ctrl_init(adapter);
4711 /* sync up with fw's ready state */
4712 if (be_physfn(adapter)) {
4713 status = be_fw_wait_ready(adapter);
4718 if (be_reset_required(adapter)) {
4719 status = be_cmd_reset_function(adapter);
4723 /* Wait for interrupts to quiesce after an FLR */
4727 /* Allow interrupts for other ULPs running on NIC function */
4728 be_intr_set(adapter, true);
4730 /* tell fw we're ready to fire cmds */
4731 status = be_cmd_fw_init(adapter);
4735 status = be_stats_init(adapter);
4739 status = be_get_initial_config(adapter);
4743 INIT_DELAYED_WORK(&adapter->work, be_worker);
4744 INIT_DELAYED_WORK(&adapter->func_recovery_work, be_func_recovery_task);
4745 adapter->rx_fc = adapter->tx_fc = true;
4747 status = be_setup(adapter);
4751 be_netdev_init(netdev);
4752 status = register_netdev(netdev);
4756 be_roce_dev_add(adapter);
4758 schedule_delayed_work(&adapter->func_recovery_work,
4759 msecs_to_jiffies(1000));
4761 be_cmd_query_port_name(adapter, &port_name);
4763 dev_info(&pdev->dev, "%s: %s %s port %c\n", nic_name(pdev),
4764 func_name(adapter), mc_name(adapter), port_name);
4771 be_stats_cleanup(adapter);
4773 be_ctrl_cleanup(adapter);
4775 free_netdev(netdev);
4777 pci_release_regions(pdev);
4779 pci_disable_device(pdev);
4781 dev_err(&pdev->dev, "%s initialization failed\n", nic_name(pdev));
4785 static int be_suspend(struct pci_dev *pdev, pm_message_t state)
4787 struct be_adapter *adapter = pci_get_drvdata(pdev);
4788 struct net_device *netdev = adapter->netdev;
4790 if (adapter->wol_en)
4791 be_setup_wol(adapter, true);
4793 be_intr_set(adapter, false);
4794 cancel_delayed_work_sync(&adapter->func_recovery_work);
4796 netif_device_detach(netdev);
4797 if (netif_running(netdev)) {
4804 pci_save_state(pdev);
4805 pci_disable_device(pdev);
4806 pci_set_power_state(pdev, pci_choose_state(pdev, state));
4810 static int be_resume(struct pci_dev *pdev)
4813 struct be_adapter *adapter = pci_get_drvdata(pdev);
4814 struct net_device *netdev = adapter->netdev;
4816 netif_device_detach(netdev);
4818 status = pci_enable_device(pdev);
4822 pci_set_power_state(pdev, PCI_D0);
4823 pci_restore_state(pdev);
4825 status = be_fw_wait_ready(adapter);
4829 be_intr_set(adapter, true);
4830 /* tell fw we're ready to fire cmds */
4831 status = be_cmd_fw_init(adapter);
4836 if (netif_running(netdev)) {
4842 schedule_delayed_work(&adapter->func_recovery_work,
4843 msecs_to_jiffies(1000));
4844 netif_device_attach(netdev);
4846 if (adapter->wol_en)
4847 be_setup_wol(adapter, false);
4853 * An FLR will stop BE from DMAing any data.
4855 static void be_shutdown(struct pci_dev *pdev)
4857 struct be_adapter *adapter = pci_get_drvdata(pdev);
4862 cancel_delayed_work_sync(&adapter->work);
4863 cancel_delayed_work_sync(&adapter->func_recovery_work);
4865 netif_device_detach(adapter->netdev);
4867 be_cmd_reset_function(adapter);
4869 pci_disable_device(pdev);
4872 static pci_ers_result_t be_eeh_err_detected(struct pci_dev *pdev,
4873 pci_channel_state_t state)
4875 struct be_adapter *adapter = pci_get_drvdata(pdev);
4876 struct net_device *netdev = adapter->netdev;
4878 dev_err(&adapter->pdev->dev, "EEH error detected\n");
4880 if (!adapter->eeh_error) {
4881 adapter->eeh_error = true;
4883 cancel_delayed_work_sync(&adapter->func_recovery_work);
4886 netif_device_detach(netdev);
4887 if (netif_running(netdev))
4894 if (state == pci_channel_io_perm_failure)
4895 return PCI_ERS_RESULT_DISCONNECT;
4897 pci_disable_device(pdev);
4899 /* The error could cause the FW to trigger a flash debug dump.
4900 * Resetting the card while flash dump is in progress
4901 * can cause it not to recover; wait for it to finish.
4902 * Wait only for first function as it is needed only once per
4905 if (pdev->devfn == 0)
4908 return PCI_ERS_RESULT_NEED_RESET;
4911 static pci_ers_result_t be_eeh_reset(struct pci_dev *pdev)
4913 struct be_adapter *adapter = pci_get_drvdata(pdev);
4916 dev_info(&adapter->pdev->dev, "EEH reset\n");
4918 status = pci_enable_device(pdev);
4920 return PCI_ERS_RESULT_DISCONNECT;
4922 pci_set_master(pdev);
4923 pci_set_power_state(pdev, PCI_D0);
4924 pci_restore_state(pdev);
4926 /* Check if card is ok and fw is ready */
4927 dev_info(&adapter->pdev->dev,
4928 "Waiting for FW to be ready after EEH reset\n");
4929 status = be_fw_wait_ready(adapter);
4931 return PCI_ERS_RESULT_DISCONNECT;
4933 pci_cleanup_aer_uncorrect_error_status(pdev);
4934 be_clear_all_error(adapter);
4935 return PCI_ERS_RESULT_RECOVERED;
4938 static void be_eeh_resume(struct pci_dev *pdev)
4941 struct be_adapter *adapter = pci_get_drvdata(pdev);
4942 struct net_device *netdev = adapter->netdev;
4944 dev_info(&adapter->pdev->dev, "EEH resume\n");
4946 pci_save_state(pdev);
4948 status = be_cmd_reset_function(adapter);
4952 /* On some BE3 FW versions, after a HW reset,
4953 * interrupts will remain disabled for each function.
4954 * So, explicitly enable interrupts
4956 be_intr_set(adapter, true);
4958 /* tell fw we're ready to fire cmds */
4959 status = be_cmd_fw_init(adapter);
4963 status = be_setup(adapter);
4967 if (netif_running(netdev)) {
4968 status = be_open(netdev);
4973 schedule_delayed_work(&adapter->func_recovery_work,
4974 msecs_to_jiffies(1000));
4975 netif_device_attach(netdev);
4978 dev_err(&adapter->pdev->dev, "EEH resume failed\n");
4981 static const struct pci_error_handlers be_eeh_handlers = {
4982 .error_detected = be_eeh_err_detected,
4983 .slot_reset = be_eeh_reset,
4984 .resume = be_eeh_resume,
4987 static struct pci_driver be_driver = {
4989 .id_table = be_dev_ids,
4991 .remove = be_remove,
4992 .suspend = be_suspend,
4993 .resume = be_resume,
4994 .shutdown = be_shutdown,
4995 .err_handler = &be_eeh_handlers
4998 static int __init be_init_module(void)
5000 if (rx_frag_size != 8192 && rx_frag_size != 4096 &&
5001 rx_frag_size != 2048) {
5002 printk(KERN_WARNING DRV_NAME
5003 " : Module param rx_frag_size must be 2048/4096/8192."
5005 rx_frag_size = 2048;
5008 return pci_register_driver(&be_driver);
5010 module_init(be_init_module);
5012 static void __exit be_exit_module(void)
5014 pci_unregister_driver(&be_driver);
5016 module_exit(be_exit_module);