2 * Copyright (C) 2005 - 2013 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>
25 MODULE_VERSION(DRV_VER);
26 MODULE_DEVICE_TABLE(pci, be_dev_ids);
27 MODULE_DESCRIPTION(DRV_DESC " " DRV_VER);
28 MODULE_AUTHOR("Emulex Corporation");
29 MODULE_LICENSE("GPL");
31 static unsigned int num_vfs;
32 module_param(num_vfs, uint, S_IRUGO);
33 MODULE_PARM_DESC(num_vfs, "Number of PCI VFs to initialize");
35 static ushort rx_frag_size = 2048;
36 module_param(rx_frag_size, ushort, S_IRUGO);
37 MODULE_PARM_DESC(rx_frag_size, "Size of a fragment that holds rcvd data.");
39 static DEFINE_PCI_DEVICE_TABLE(be_dev_ids) = {
40 { PCI_DEVICE(BE_VENDOR_ID, BE_DEVICE_ID1) },
41 { PCI_DEVICE(BE_VENDOR_ID, BE_DEVICE_ID2) },
42 { PCI_DEVICE(BE_VENDOR_ID, OC_DEVICE_ID1) },
43 { PCI_DEVICE(BE_VENDOR_ID, OC_DEVICE_ID2) },
44 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID3)},
45 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID4)},
46 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID5)},
47 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID6)},
50 MODULE_DEVICE_TABLE(pci, be_dev_ids);
51 /* UE Status Low CSR */
52 static const char * const ue_status_low_desc[] = {
86 /* UE Status High CSR */
87 static const char * const ue_status_hi_desc[] = {
122 /* Is BE in a multi-channel mode */
123 static inline bool be_is_mc(struct be_adapter *adapter) {
124 return (adapter->function_mode & FLEX10_MODE ||
125 adapter->function_mode & VNIC_MODE ||
126 adapter->function_mode & UMC_ENABLED);
129 static void be_queue_free(struct be_adapter *adapter, struct be_queue_info *q)
131 struct be_dma_mem *mem = &q->dma_mem;
133 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
139 static int be_queue_alloc(struct be_adapter *adapter, struct be_queue_info *q,
140 u16 len, u16 entry_size)
142 struct be_dma_mem *mem = &q->dma_mem;
144 memset(q, 0, sizeof(*q));
146 q->entry_size = entry_size;
147 mem->size = len * entry_size;
148 mem->va = dma_alloc_coherent(&adapter->pdev->dev, mem->size, &mem->dma,
149 GFP_KERNEL | __GFP_ZERO);
155 static void be_reg_intr_set(struct be_adapter *adapter, bool enable)
159 pci_read_config_dword(adapter->pdev, PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET,
161 enabled = reg & MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
163 if (!enabled && enable)
164 reg |= MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
165 else if (enabled && !enable)
166 reg &= ~MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
170 pci_write_config_dword(adapter->pdev,
171 PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET, reg);
174 static void be_intr_set(struct be_adapter *adapter, bool enable)
178 /* On lancer interrupts can't be controlled via this register */
179 if (lancer_chip(adapter))
182 if (adapter->eeh_error)
185 status = be_cmd_intr_set(adapter, enable);
187 be_reg_intr_set(adapter, enable);
190 static void be_rxq_notify(struct be_adapter *adapter, u16 qid, u16 posted)
193 val |= qid & DB_RQ_RING_ID_MASK;
194 val |= posted << DB_RQ_NUM_POSTED_SHIFT;
197 iowrite32(val, adapter->db + DB_RQ_OFFSET);
200 static void be_txq_notify(struct be_adapter *adapter, struct be_tx_obj *txo,
204 val |= txo->q.id & DB_TXULP_RING_ID_MASK;
205 val |= (posted & DB_TXULP_NUM_POSTED_MASK) << DB_TXULP_NUM_POSTED_SHIFT;
208 iowrite32(val, adapter->db + txo->db_offset);
211 static void be_eq_notify(struct be_adapter *adapter, u16 qid,
212 bool arm, bool clear_int, u16 num_popped)
215 val |= qid & DB_EQ_RING_ID_MASK;
216 val |= ((qid & DB_EQ_RING_ID_EXT_MASK) <<
217 DB_EQ_RING_ID_EXT_MASK_SHIFT);
219 if (adapter->eeh_error)
223 val |= 1 << DB_EQ_REARM_SHIFT;
225 val |= 1 << DB_EQ_CLR_SHIFT;
226 val |= 1 << DB_EQ_EVNT_SHIFT;
227 val |= num_popped << DB_EQ_NUM_POPPED_SHIFT;
228 iowrite32(val, adapter->db + DB_EQ_OFFSET);
231 void be_cq_notify(struct be_adapter *adapter, u16 qid, bool arm, u16 num_popped)
234 val |= qid & DB_CQ_RING_ID_MASK;
235 val |= ((qid & DB_CQ_RING_ID_EXT_MASK) <<
236 DB_CQ_RING_ID_EXT_MASK_SHIFT);
238 if (adapter->eeh_error)
242 val |= 1 << DB_CQ_REARM_SHIFT;
243 val |= num_popped << DB_CQ_NUM_POPPED_SHIFT;
244 iowrite32(val, adapter->db + DB_CQ_OFFSET);
247 static int be_mac_addr_set(struct net_device *netdev, void *p)
249 struct be_adapter *adapter = netdev_priv(netdev);
250 struct sockaddr *addr = p;
252 u8 current_mac[ETH_ALEN];
253 u32 pmac_id = adapter->pmac_id[0];
254 bool active_mac = true;
256 if (!is_valid_ether_addr(addr->sa_data))
257 return -EADDRNOTAVAIL;
259 /* For BE VF, MAC address is already activated by PF.
260 * Hence only operation left is updating netdev->devaddr.
261 * Update it if user is passing the same MAC which was used
262 * during configuring VF MAC from PF(Hypervisor).
264 if (!lancer_chip(adapter) && !be_physfn(adapter)) {
265 status = be_cmd_mac_addr_query(adapter, current_mac,
266 false, adapter->if_handle, 0);
267 if (!status && !memcmp(current_mac, addr->sa_data, ETH_ALEN))
273 if (!memcmp(addr->sa_data, netdev->dev_addr, ETH_ALEN))
276 /* For Lancer check if any MAC is active.
277 * If active, get its mac id.
279 if (lancer_chip(adapter) && !be_physfn(adapter))
280 be_cmd_get_mac_from_list(adapter, current_mac, &active_mac,
283 status = be_cmd_pmac_add(adapter, (u8 *)addr->sa_data,
285 &adapter->pmac_id[0], 0);
291 be_cmd_pmac_del(adapter, adapter->if_handle,
294 memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
297 dev_err(&adapter->pdev->dev, "MAC %pM set Failed\n", addr->sa_data);
301 /* BE2 supports only v0 cmd */
302 static void *hw_stats_from_cmd(struct be_adapter *adapter)
304 if (BE2_chip(adapter)) {
305 struct be_cmd_resp_get_stats_v0 *cmd = adapter->stats_cmd.va;
307 return &cmd->hw_stats;
309 struct be_cmd_resp_get_stats_v1 *cmd = adapter->stats_cmd.va;
311 return &cmd->hw_stats;
315 /* BE2 supports only v0 cmd */
316 static void *be_erx_stats_from_cmd(struct be_adapter *adapter)
318 if (BE2_chip(adapter)) {
319 struct be_hw_stats_v0 *hw_stats = hw_stats_from_cmd(adapter);
321 return &hw_stats->erx;
323 struct be_hw_stats_v1 *hw_stats = hw_stats_from_cmd(adapter);
325 return &hw_stats->erx;
329 static void populate_be_v0_stats(struct be_adapter *adapter)
331 struct be_hw_stats_v0 *hw_stats = hw_stats_from_cmd(adapter);
332 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
333 struct be_rxf_stats_v0 *rxf_stats = &hw_stats->rxf;
334 struct be_port_rxf_stats_v0 *port_stats =
335 &rxf_stats->port[adapter->port_num];
336 struct be_drv_stats *drvs = &adapter->drv_stats;
338 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
339 drvs->rx_pause_frames = port_stats->rx_pause_frames;
340 drvs->rx_crc_errors = port_stats->rx_crc_errors;
341 drvs->rx_control_frames = port_stats->rx_control_frames;
342 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
343 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
344 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
345 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
346 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
347 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
348 drvs->rxpp_fifo_overflow_drop = port_stats->rx_fifo_overflow;
349 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
350 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
351 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
352 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
353 drvs->rx_input_fifo_overflow_drop = port_stats->rx_input_fifo_overflow;
354 drvs->rx_dropped_header_too_small =
355 port_stats->rx_dropped_header_too_small;
356 drvs->rx_address_filtered =
357 port_stats->rx_address_filtered +
358 port_stats->rx_vlan_filtered;
359 drvs->rx_alignment_symbol_errors =
360 port_stats->rx_alignment_symbol_errors;
362 drvs->tx_pauseframes = port_stats->tx_pauseframes;
363 drvs->tx_controlframes = port_stats->tx_controlframes;
365 if (adapter->port_num)
366 drvs->jabber_events = rxf_stats->port1_jabber_events;
368 drvs->jabber_events = rxf_stats->port0_jabber_events;
369 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
370 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
371 drvs->forwarded_packets = rxf_stats->forwarded_packets;
372 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
373 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
374 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
375 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
378 static void populate_be_v1_stats(struct be_adapter *adapter)
380 struct be_hw_stats_v1 *hw_stats = hw_stats_from_cmd(adapter);
381 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
382 struct be_rxf_stats_v1 *rxf_stats = &hw_stats->rxf;
383 struct be_port_rxf_stats_v1 *port_stats =
384 &rxf_stats->port[adapter->port_num];
385 struct be_drv_stats *drvs = &adapter->drv_stats;
387 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
388 drvs->pmem_fifo_overflow_drop = port_stats->pmem_fifo_overflow_drop;
389 drvs->rx_priority_pause_frames = port_stats->rx_priority_pause_frames;
390 drvs->rx_pause_frames = port_stats->rx_pause_frames;
391 drvs->rx_crc_errors = port_stats->rx_crc_errors;
392 drvs->rx_control_frames = port_stats->rx_control_frames;
393 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
394 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
395 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
396 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
397 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
398 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
399 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
400 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
401 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
402 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
403 drvs->rx_dropped_header_too_small =
404 port_stats->rx_dropped_header_too_small;
405 drvs->rx_input_fifo_overflow_drop =
406 port_stats->rx_input_fifo_overflow_drop;
407 drvs->rx_address_filtered = port_stats->rx_address_filtered;
408 drvs->rx_alignment_symbol_errors =
409 port_stats->rx_alignment_symbol_errors;
410 drvs->rxpp_fifo_overflow_drop = port_stats->rxpp_fifo_overflow_drop;
411 drvs->tx_pauseframes = port_stats->tx_pauseframes;
412 drvs->tx_controlframes = port_stats->tx_controlframes;
413 drvs->tx_priority_pauseframes = port_stats->tx_priority_pauseframes;
414 drvs->jabber_events = port_stats->jabber_events;
415 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
416 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
417 drvs->forwarded_packets = rxf_stats->forwarded_packets;
418 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
419 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
420 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
421 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
424 static void populate_lancer_stats(struct be_adapter *adapter)
427 struct be_drv_stats *drvs = &adapter->drv_stats;
428 struct lancer_pport_stats *pport_stats =
429 pport_stats_from_cmd(adapter);
431 be_dws_le_to_cpu(pport_stats, sizeof(*pport_stats));
432 drvs->rx_pause_frames = pport_stats->rx_pause_frames_lo;
433 drvs->rx_crc_errors = pport_stats->rx_crc_errors_lo;
434 drvs->rx_control_frames = pport_stats->rx_control_frames_lo;
435 drvs->rx_in_range_errors = pport_stats->rx_in_range_errors;
436 drvs->rx_frame_too_long = pport_stats->rx_frames_too_long_lo;
437 drvs->rx_dropped_runt = pport_stats->rx_dropped_runt;
438 drvs->rx_ip_checksum_errs = pport_stats->rx_ip_checksum_errors;
439 drvs->rx_tcp_checksum_errs = pport_stats->rx_tcp_checksum_errors;
440 drvs->rx_udp_checksum_errs = pport_stats->rx_udp_checksum_errors;
441 drvs->rx_dropped_tcp_length =
442 pport_stats->rx_dropped_invalid_tcp_length;
443 drvs->rx_dropped_too_small = pport_stats->rx_dropped_too_small;
444 drvs->rx_dropped_too_short = pport_stats->rx_dropped_too_short;
445 drvs->rx_out_range_errors = pport_stats->rx_out_of_range_errors;
446 drvs->rx_dropped_header_too_small =
447 pport_stats->rx_dropped_header_too_small;
448 drvs->rx_input_fifo_overflow_drop = pport_stats->rx_fifo_overflow;
449 drvs->rx_address_filtered =
450 pport_stats->rx_address_filtered +
451 pport_stats->rx_vlan_filtered;
452 drvs->rx_alignment_symbol_errors = pport_stats->rx_symbol_errors_lo;
453 drvs->rxpp_fifo_overflow_drop = pport_stats->rx_fifo_overflow;
454 drvs->tx_pauseframes = pport_stats->tx_pause_frames_lo;
455 drvs->tx_controlframes = pport_stats->tx_control_frames_lo;
456 drvs->jabber_events = pport_stats->rx_jabbers;
457 drvs->forwarded_packets = pport_stats->num_forwards_lo;
458 drvs->rx_drops_mtu = pport_stats->rx_drops_mtu_lo;
459 drvs->rx_drops_too_many_frags =
460 pport_stats->rx_drops_too_many_frags_lo;
463 static void accumulate_16bit_val(u32 *acc, u16 val)
465 #define lo(x) (x & 0xFFFF)
466 #define hi(x) (x & 0xFFFF0000)
467 bool wrapped = val < lo(*acc);
468 u32 newacc = hi(*acc) + val;
472 ACCESS_ONCE(*acc) = newacc;
475 void populate_erx_stats(struct be_adapter *adapter,
476 struct be_rx_obj *rxo,
479 if (!BEx_chip(adapter))
480 rx_stats(rxo)->rx_drops_no_frags = erx_stat;
482 /* below erx HW counter can actually wrap around after
483 * 65535. Driver accumulates a 32-bit value
485 accumulate_16bit_val(&rx_stats(rxo)->rx_drops_no_frags,
489 void be_parse_stats(struct be_adapter *adapter)
491 struct be_erx_stats_v1 *erx = be_erx_stats_from_cmd(adapter);
492 struct be_rx_obj *rxo;
496 if (lancer_chip(adapter)) {
497 populate_lancer_stats(adapter);
499 if (BE2_chip(adapter))
500 populate_be_v0_stats(adapter);
502 /* for BE3 and Skyhawk */
503 populate_be_v1_stats(adapter);
505 /* as erx_v1 is longer than v0, ok to use v1 for v0 access */
506 for_all_rx_queues(adapter, rxo, i) {
507 erx_stat = erx->rx_drops_no_fragments[rxo->q.id];
508 populate_erx_stats(adapter, rxo, erx_stat);
513 static struct rtnl_link_stats64 *be_get_stats64(struct net_device *netdev,
514 struct rtnl_link_stats64 *stats)
516 struct be_adapter *adapter = netdev_priv(netdev);
517 struct be_drv_stats *drvs = &adapter->drv_stats;
518 struct be_rx_obj *rxo;
519 struct be_tx_obj *txo;
524 for_all_rx_queues(adapter, rxo, i) {
525 const struct be_rx_stats *rx_stats = rx_stats(rxo);
527 start = u64_stats_fetch_begin_bh(&rx_stats->sync);
528 pkts = rx_stats(rxo)->rx_pkts;
529 bytes = rx_stats(rxo)->rx_bytes;
530 } while (u64_stats_fetch_retry_bh(&rx_stats->sync, start));
531 stats->rx_packets += pkts;
532 stats->rx_bytes += bytes;
533 stats->multicast += rx_stats(rxo)->rx_mcast_pkts;
534 stats->rx_dropped += rx_stats(rxo)->rx_drops_no_skbs +
535 rx_stats(rxo)->rx_drops_no_frags;
538 for_all_tx_queues(adapter, txo, i) {
539 const struct be_tx_stats *tx_stats = tx_stats(txo);
541 start = u64_stats_fetch_begin_bh(&tx_stats->sync);
542 pkts = tx_stats(txo)->tx_pkts;
543 bytes = tx_stats(txo)->tx_bytes;
544 } while (u64_stats_fetch_retry_bh(&tx_stats->sync, start));
545 stats->tx_packets += pkts;
546 stats->tx_bytes += bytes;
549 /* bad pkts received */
550 stats->rx_errors = drvs->rx_crc_errors +
551 drvs->rx_alignment_symbol_errors +
552 drvs->rx_in_range_errors +
553 drvs->rx_out_range_errors +
554 drvs->rx_frame_too_long +
555 drvs->rx_dropped_too_small +
556 drvs->rx_dropped_too_short +
557 drvs->rx_dropped_header_too_small +
558 drvs->rx_dropped_tcp_length +
559 drvs->rx_dropped_runt;
561 /* detailed rx errors */
562 stats->rx_length_errors = drvs->rx_in_range_errors +
563 drvs->rx_out_range_errors +
564 drvs->rx_frame_too_long;
566 stats->rx_crc_errors = drvs->rx_crc_errors;
568 /* frame alignment errors */
569 stats->rx_frame_errors = drvs->rx_alignment_symbol_errors;
571 /* receiver fifo overrun */
572 /* drops_no_pbuf is no per i/f, it's per BE card */
573 stats->rx_fifo_errors = drvs->rxpp_fifo_overflow_drop +
574 drvs->rx_input_fifo_overflow_drop +
575 drvs->rx_drops_no_pbuf;
579 void be_link_status_update(struct be_adapter *adapter, u8 link_status)
581 struct net_device *netdev = adapter->netdev;
583 if (!(adapter->flags & BE_FLAGS_LINK_STATUS_INIT)) {
584 netif_carrier_off(netdev);
585 adapter->flags |= BE_FLAGS_LINK_STATUS_INIT;
588 if ((link_status & LINK_STATUS_MASK) == LINK_UP)
589 netif_carrier_on(netdev);
591 netif_carrier_off(netdev);
594 static void be_tx_stats_update(struct be_tx_obj *txo,
595 u32 wrb_cnt, u32 copied, u32 gso_segs, bool stopped)
597 struct be_tx_stats *stats = tx_stats(txo);
599 u64_stats_update_begin(&stats->sync);
601 stats->tx_wrbs += wrb_cnt;
602 stats->tx_bytes += copied;
603 stats->tx_pkts += (gso_segs ? gso_segs : 1);
606 u64_stats_update_end(&stats->sync);
609 /* Determine number of WRB entries needed to xmit data in an skb */
610 static u32 wrb_cnt_for_skb(struct be_adapter *adapter, struct sk_buff *skb,
613 int cnt = (skb->len > skb->data_len);
615 cnt += skb_shinfo(skb)->nr_frags;
617 /* to account for hdr wrb */
619 if (lancer_chip(adapter) || !(cnt & 1)) {
622 /* add a dummy to make it an even num */
626 BUG_ON(cnt > BE_MAX_TX_FRAG_COUNT);
630 static inline void wrb_fill(struct be_eth_wrb *wrb, u64 addr, int len)
632 wrb->frag_pa_hi = upper_32_bits(addr);
633 wrb->frag_pa_lo = addr & 0xFFFFFFFF;
634 wrb->frag_len = len & ETH_WRB_FRAG_LEN_MASK;
638 static inline u16 be_get_tx_vlan_tag(struct be_adapter *adapter,
644 vlan_tag = vlan_tx_tag_get(skb);
645 vlan_prio = (vlan_tag & VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
646 /* If vlan priority provided by OS is NOT in available bmap */
647 if (!(adapter->vlan_prio_bmap & (1 << vlan_prio)))
648 vlan_tag = (vlan_tag & ~VLAN_PRIO_MASK) |
649 adapter->recommended_prio;
654 static void wrb_fill_hdr(struct be_adapter *adapter, struct be_eth_hdr_wrb *hdr,
655 struct sk_buff *skb, u32 wrb_cnt, u32 len, bool skip_hw_vlan)
659 memset(hdr, 0, sizeof(*hdr));
661 AMAP_SET_BITS(struct amap_eth_hdr_wrb, crc, hdr, 1);
663 if (skb_is_gso(skb)) {
664 AMAP_SET_BITS(struct amap_eth_hdr_wrb, lso, hdr, 1);
665 AMAP_SET_BITS(struct amap_eth_hdr_wrb, lso_mss,
666 hdr, skb_shinfo(skb)->gso_size);
667 if (skb_is_gso_v6(skb) && !lancer_chip(adapter))
668 AMAP_SET_BITS(struct amap_eth_hdr_wrb, lso6, hdr, 1);
669 } else if (skb->ip_summed == CHECKSUM_PARTIAL) {
671 AMAP_SET_BITS(struct amap_eth_hdr_wrb, tcpcs, hdr, 1);
672 else if (is_udp_pkt(skb))
673 AMAP_SET_BITS(struct amap_eth_hdr_wrb, udpcs, hdr, 1);
676 if (vlan_tx_tag_present(skb)) {
677 AMAP_SET_BITS(struct amap_eth_hdr_wrb, vlan, hdr, 1);
678 vlan_tag = be_get_tx_vlan_tag(adapter, skb);
679 AMAP_SET_BITS(struct amap_eth_hdr_wrb, vlan_tag, hdr, vlan_tag);
682 /* To skip HW VLAN tagging: evt = 1, compl = 0 */
683 AMAP_SET_BITS(struct amap_eth_hdr_wrb, complete, hdr, !skip_hw_vlan);
684 AMAP_SET_BITS(struct amap_eth_hdr_wrb, event, hdr, 1);
685 AMAP_SET_BITS(struct amap_eth_hdr_wrb, num_wrb, hdr, wrb_cnt);
686 AMAP_SET_BITS(struct amap_eth_hdr_wrb, len, hdr, len);
689 static void unmap_tx_frag(struct device *dev, struct be_eth_wrb *wrb,
694 be_dws_le_to_cpu(wrb, sizeof(*wrb));
696 dma = (u64)wrb->frag_pa_hi << 32 | (u64)wrb->frag_pa_lo;
699 dma_unmap_single(dev, dma, wrb->frag_len,
702 dma_unmap_page(dev, dma, wrb->frag_len, DMA_TO_DEVICE);
706 static int make_tx_wrbs(struct be_adapter *adapter, struct be_queue_info *txq,
707 struct sk_buff *skb, u32 wrb_cnt, bool dummy_wrb,
712 struct device *dev = &adapter->pdev->dev;
713 struct sk_buff *first_skb = skb;
714 struct be_eth_wrb *wrb;
715 struct be_eth_hdr_wrb *hdr;
716 bool map_single = false;
719 hdr = queue_head_node(txq);
721 map_head = txq->head;
723 if (skb->len > skb->data_len) {
724 int len = skb_headlen(skb);
725 busaddr = dma_map_single(dev, skb->data, len, DMA_TO_DEVICE);
726 if (dma_mapping_error(dev, busaddr))
729 wrb = queue_head_node(txq);
730 wrb_fill(wrb, busaddr, len);
731 be_dws_cpu_to_le(wrb, sizeof(*wrb));
736 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
737 const struct skb_frag_struct *frag =
738 &skb_shinfo(skb)->frags[i];
739 busaddr = skb_frag_dma_map(dev, frag, 0,
740 skb_frag_size(frag), DMA_TO_DEVICE);
741 if (dma_mapping_error(dev, busaddr))
743 wrb = queue_head_node(txq);
744 wrb_fill(wrb, busaddr, skb_frag_size(frag));
745 be_dws_cpu_to_le(wrb, sizeof(*wrb));
747 copied += skb_frag_size(frag);
751 wrb = queue_head_node(txq);
753 be_dws_cpu_to_le(wrb, sizeof(*wrb));
757 wrb_fill_hdr(adapter, hdr, first_skb, wrb_cnt, copied, skip_hw_vlan);
758 be_dws_cpu_to_le(hdr, sizeof(*hdr));
762 txq->head = map_head;
764 wrb = queue_head_node(txq);
765 unmap_tx_frag(dev, wrb, map_single);
767 copied -= wrb->frag_len;
773 static struct sk_buff *be_insert_vlan_in_pkt(struct be_adapter *adapter,
779 skb = skb_share_check(skb, GFP_ATOMIC);
783 if (vlan_tx_tag_present(skb))
784 vlan_tag = be_get_tx_vlan_tag(adapter, skb);
786 if (qnq_async_evt_rcvd(adapter) && adapter->pvid) {
788 vlan_tag = adapter->pvid;
789 /* f/w workaround to set skip_hw_vlan = 1, informs the F/W to
790 * skip VLAN insertion
793 *skip_hw_vlan = true;
797 skb = __vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_tag);
803 /* Insert the outer VLAN, if any */
804 if (adapter->qnq_vid) {
805 vlan_tag = adapter->qnq_vid;
806 skb = __vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_tag);
810 *skip_hw_vlan = true;
816 static bool be_ipv6_exthdr_check(struct sk_buff *skb)
818 struct ethhdr *eh = (struct ethhdr *)skb->data;
819 u16 offset = ETH_HLEN;
821 if (eh->h_proto == htons(ETH_P_IPV6)) {
822 struct ipv6hdr *ip6h = (struct ipv6hdr *)(skb->data + offset);
824 offset += sizeof(struct ipv6hdr);
825 if (ip6h->nexthdr != NEXTHDR_TCP &&
826 ip6h->nexthdr != NEXTHDR_UDP) {
827 struct ipv6_opt_hdr *ehdr =
828 (struct ipv6_opt_hdr *) (skb->data + offset);
830 /* offending pkt: 2nd byte following IPv6 hdr is 0xff */
831 if (ehdr->hdrlen == 0xff)
838 static int be_vlan_tag_tx_chk(struct be_adapter *adapter, struct sk_buff *skb)
840 return vlan_tx_tag_present(skb) || adapter->pvid || adapter->qnq_vid;
843 static int be_ipv6_tx_stall_chk(struct be_adapter *adapter,
846 return BE3_chip(adapter) && be_ipv6_exthdr_check(skb);
849 static struct sk_buff *be_xmit_workarounds(struct be_adapter *adapter,
853 struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
854 unsigned int eth_hdr_len;
857 /* Lancer ASIC has a bug wherein packets that are 32 bytes or less
858 * may cause a transmit stall on that port. So the work-around is to
859 * pad such packets to a 36-byte length.
861 if (unlikely(lancer_chip(adapter) && skb->len <= 32)) {
862 if (skb_padto(skb, 36))
867 /* For padded packets, BE HW modifies tot_len field in IP header
868 * incorrecly when VLAN tag is inserted by HW.
869 * For padded packets, Lancer computes incorrect checksum.
871 eth_hdr_len = ntohs(skb->protocol) == ETH_P_8021Q ?
872 VLAN_ETH_HLEN : ETH_HLEN;
873 if (skb->len <= 60 &&
874 (lancer_chip(adapter) || vlan_tx_tag_present(skb)) &&
876 ip = (struct iphdr *)ip_hdr(skb);
877 pskb_trim(skb, eth_hdr_len + ntohs(ip->tot_len));
880 /* If vlan tag is already inlined in the packet, skip HW VLAN
881 * tagging in UMC mode
883 if ((adapter->function_mode & UMC_ENABLED) &&
884 veh->h_vlan_proto == htons(ETH_P_8021Q))
885 *skip_hw_vlan = true;
887 /* HW has a bug wherein it will calculate CSUM for VLAN
888 * pkts even though it is disabled.
889 * Manually insert VLAN in pkt.
891 if (skb->ip_summed != CHECKSUM_PARTIAL &&
892 vlan_tx_tag_present(skb)) {
893 skb = be_insert_vlan_in_pkt(adapter, skb, skip_hw_vlan);
898 /* HW may lockup when VLAN HW tagging is requested on
899 * certain ipv6 packets. Drop such pkts if the HW workaround to
900 * skip HW tagging is not enabled by FW.
902 if (unlikely(be_ipv6_tx_stall_chk(adapter, skb) &&
903 (adapter->pvid || adapter->qnq_vid) &&
904 !qnq_async_evt_rcvd(adapter)))
907 /* Manual VLAN tag insertion to prevent:
908 * ASIC lockup when the ASIC inserts VLAN tag into
909 * certain ipv6 packets. Insert VLAN tags in driver,
910 * and set event, completion, vlan bits accordingly
913 if (be_ipv6_tx_stall_chk(adapter, skb) &&
914 be_vlan_tag_tx_chk(adapter, skb)) {
915 skb = be_insert_vlan_in_pkt(adapter, skb, skip_hw_vlan);
922 dev_kfree_skb_any(skb);
926 static netdev_tx_t be_xmit(struct sk_buff *skb, struct net_device *netdev)
928 struct be_adapter *adapter = netdev_priv(netdev);
929 struct be_tx_obj *txo = &adapter->tx_obj[skb_get_queue_mapping(skb)];
930 struct be_queue_info *txq = &txo->q;
931 bool dummy_wrb, stopped = false;
932 u32 wrb_cnt = 0, copied = 0;
933 bool skip_hw_vlan = false;
934 u32 start = txq->head;
936 skb = be_xmit_workarounds(adapter, skb, &skip_hw_vlan);
940 wrb_cnt = wrb_cnt_for_skb(adapter, skb, &dummy_wrb);
942 copied = make_tx_wrbs(adapter, txq, skb, wrb_cnt, dummy_wrb,
945 int gso_segs = skb_shinfo(skb)->gso_segs;
947 /* record the sent skb in the sent_skb table */
948 BUG_ON(txo->sent_skb_list[start]);
949 txo->sent_skb_list[start] = skb;
951 /* Ensure txq has space for the next skb; Else stop the queue
952 * *BEFORE* ringing the tx doorbell, so that we serialze the
953 * tx compls of the current transmit which'll wake up the queue
955 atomic_add(wrb_cnt, &txq->used);
956 if ((BE_MAX_TX_FRAG_COUNT + atomic_read(&txq->used)) >=
958 netif_stop_subqueue(netdev, skb_get_queue_mapping(skb));
962 be_txq_notify(adapter, txo, wrb_cnt);
964 be_tx_stats_update(txo, wrb_cnt, copied, gso_segs, stopped);
967 dev_kfree_skb_any(skb);
972 static int be_change_mtu(struct net_device *netdev, int new_mtu)
974 struct be_adapter *adapter = netdev_priv(netdev);
975 if (new_mtu < BE_MIN_MTU ||
976 new_mtu > (BE_MAX_JUMBO_FRAME_SIZE -
977 (ETH_HLEN + ETH_FCS_LEN))) {
978 dev_info(&adapter->pdev->dev,
979 "MTU must be between %d and %d bytes\n",
981 (BE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN)));
984 dev_info(&adapter->pdev->dev, "MTU changed from %d to %d bytes\n",
985 netdev->mtu, new_mtu);
986 netdev->mtu = new_mtu;
991 * A max of 64 (BE_NUM_VLANS_SUPPORTED) vlans can be configured in BE.
992 * If the user configures more, place BE in vlan promiscuous mode.
994 static int be_vid_config(struct be_adapter *adapter)
996 u16 vids[BE_NUM_VLANS_SUPPORTED];
1000 /* No need to further configure vids if in promiscuous mode */
1001 if (adapter->promiscuous)
1004 if (adapter->vlans_added > adapter->max_vlans)
1005 goto set_vlan_promisc;
1007 /* Construct VLAN Table to give to HW */
1008 for (i = 0; i < VLAN_N_VID; i++)
1009 if (adapter->vlan_tag[i])
1010 vids[num++] = cpu_to_le16(i);
1012 status = be_cmd_vlan_config(adapter, adapter->if_handle,
1015 /* Set to VLAN promisc mode as setting VLAN filter failed */
1017 dev_info(&adapter->pdev->dev, "Exhausted VLAN HW filters.\n");
1018 dev_info(&adapter->pdev->dev, "Disabling HW VLAN filtering.\n");
1019 goto set_vlan_promisc;
1025 status = be_cmd_vlan_config(adapter, adapter->if_handle,
1030 static int be_vlan_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
1032 struct be_adapter *adapter = netdev_priv(netdev);
1035 if (!lancer_chip(adapter) && !be_physfn(adapter)) {
1040 /* Packets with VID 0 are always received by Lancer by default */
1041 if (lancer_chip(adapter) && vid == 0)
1044 adapter->vlan_tag[vid] = 1;
1045 if (adapter->vlans_added <= (adapter->max_vlans + 1))
1046 status = be_vid_config(adapter);
1049 adapter->vlans_added++;
1051 adapter->vlan_tag[vid] = 0;
1056 static int be_vlan_rem_vid(struct net_device *netdev, __be16 proto, u16 vid)
1058 struct be_adapter *adapter = netdev_priv(netdev);
1061 if (!lancer_chip(adapter) && !be_physfn(adapter)) {
1066 /* Packets with VID 0 are always received by Lancer by default */
1067 if (lancer_chip(adapter) && vid == 0)
1070 adapter->vlan_tag[vid] = 0;
1071 if (adapter->vlans_added <= adapter->max_vlans)
1072 status = be_vid_config(adapter);
1075 adapter->vlans_added--;
1077 adapter->vlan_tag[vid] = 1;
1082 static void be_set_rx_mode(struct net_device *netdev)
1084 struct be_adapter *adapter = netdev_priv(netdev);
1087 if (netdev->flags & IFF_PROMISC) {
1088 be_cmd_rx_filter(adapter, IFF_PROMISC, ON);
1089 adapter->promiscuous = true;
1093 /* BE was previously in promiscuous mode; disable it */
1094 if (adapter->promiscuous) {
1095 adapter->promiscuous = false;
1096 be_cmd_rx_filter(adapter, IFF_PROMISC, OFF);
1098 if (adapter->vlans_added)
1099 be_vid_config(adapter);
1102 /* Enable multicast promisc if num configured exceeds what we support */
1103 if (netdev->flags & IFF_ALLMULTI ||
1104 netdev_mc_count(netdev) > adapter->max_mcast_mac) {
1105 be_cmd_rx_filter(adapter, IFF_ALLMULTI, ON);
1109 if (netdev_uc_count(netdev) != adapter->uc_macs) {
1110 struct netdev_hw_addr *ha;
1111 int i = 1; /* First slot is claimed by the Primary MAC */
1113 for (; adapter->uc_macs > 0; adapter->uc_macs--, i++) {
1114 be_cmd_pmac_del(adapter, adapter->if_handle,
1115 adapter->pmac_id[i], 0);
1118 if (netdev_uc_count(netdev) > adapter->max_pmac_cnt) {
1119 be_cmd_rx_filter(adapter, IFF_PROMISC, ON);
1120 adapter->promiscuous = true;
1124 netdev_for_each_uc_addr(ha, adapter->netdev) {
1125 adapter->uc_macs++; /* First slot is for Primary MAC */
1126 be_cmd_pmac_add(adapter, (u8 *)ha->addr,
1128 &adapter->pmac_id[adapter->uc_macs], 0);
1132 status = be_cmd_rx_filter(adapter, IFF_MULTICAST, ON);
1134 /* Set to MCAST promisc mode if setting MULTICAST address fails */
1136 dev_info(&adapter->pdev->dev, "Exhausted multicast HW filters.\n");
1137 dev_info(&adapter->pdev->dev, "Disabling HW multicast filtering.\n");
1138 be_cmd_rx_filter(adapter, IFF_ALLMULTI, ON);
1144 static int be_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
1146 struct be_adapter *adapter = netdev_priv(netdev);
1147 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1149 bool active_mac = false;
1151 u8 old_mac[ETH_ALEN];
1153 if (!sriov_enabled(adapter))
1156 if (!is_valid_ether_addr(mac) || vf >= adapter->num_vfs)
1159 if (lancer_chip(adapter)) {
1160 status = be_cmd_get_mac_from_list(adapter, old_mac, &active_mac,
1162 if (!status && active_mac)
1163 be_cmd_pmac_del(adapter, vf_cfg->if_handle,
1166 status = be_cmd_set_mac_list(adapter, mac, 1, vf + 1);
1168 status = be_cmd_pmac_del(adapter, vf_cfg->if_handle,
1169 vf_cfg->pmac_id, vf + 1);
1171 status = be_cmd_pmac_add(adapter, mac, vf_cfg->if_handle,
1172 &vf_cfg->pmac_id, vf + 1);
1176 dev_err(&adapter->pdev->dev, "MAC %pM set on VF %d Failed\n",
1179 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
1184 static int be_get_vf_config(struct net_device *netdev, int vf,
1185 struct ifla_vf_info *vi)
1187 struct be_adapter *adapter = netdev_priv(netdev);
1188 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1190 if (!sriov_enabled(adapter))
1193 if (vf >= adapter->num_vfs)
1197 vi->tx_rate = vf_cfg->tx_rate;
1198 vi->vlan = vf_cfg->vlan_tag;
1200 memcpy(&vi->mac, vf_cfg->mac_addr, ETH_ALEN);
1205 static int be_set_vf_vlan(struct net_device *netdev,
1206 int vf, u16 vlan, u8 qos)
1208 struct be_adapter *adapter = netdev_priv(netdev);
1211 if (!sriov_enabled(adapter))
1214 if (vf >= adapter->num_vfs || vlan > 4095)
1218 if (adapter->vf_cfg[vf].vlan_tag != vlan) {
1219 /* If this is new value, program it. Else skip. */
1220 adapter->vf_cfg[vf].vlan_tag = vlan;
1222 status = be_cmd_set_hsw_config(adapter, vlan,
1223 vf + 1, adapter->vf_cfg[vf].if_handle);
1226 /* Reset Transparent Vlan Tagging. */
1227 adapter->vf_cfg[vf].vlan_tag = 0;
1228 vlan = adapter->vf_cfg[vf].def_vid;
1229 status = be_cmd_set_hsw_config(adapter, vlan, vf + 1,
1230 adapter->vf_cfg[vf].if_handle);
1235 dev_info(&adapter->pdev->dev,
1236 "VLAN %d config on VF %d failed\n", vlan, vf);
1240 static int be_set_vf_tx_rate(struct net_device *netdev,
1243 struct be_adapter *adapter = netdev_priv(netdev);
1246 if (!sriov_enabled(adapter))
1249 if (vf >= adapter->num_vfs)
1252 if (rate < 100 || rate > 10000) {
1253 dev_err(&adapter->pdev->dev,
1254 "tx rate must be between 100 and 10000 Mbps\n");
1258 if (lancer_chip(adapter))
1259 status = be_cmd_set_profile_config(adapter, rate / 10, vf + 1);
1261 status = be_cmd_set_qos(adapter, rate / 10, vf + 1);
1264 dev_err(&adapter->pdev->dev,
1265 "tx rate %d on VF %d failed\n", rate, vf);
1267 adapter->vf_cfg[vf].tx_rate = rate;
1271 static void be_eqd_update(struct be_adapter *adapter, struct be_eq_obj *eqo)
1273 struct be_rx_stats *stats = rx_stats(&adapter->rx_obj[eqo->idx]);
1274 ulong now = jiffies;
1275 ulong delta = now - stats->rx_jiffies;
1277 unsigned int start, eqd;
1279 if (!eqo->enable_aic) {
1284 if (eqo->idx >= adapter->num_rx_qs)
1287 stats = rx_stats(&adapter->rx_obj[eqo->idx]);
1289 /* Wrapped around */
1290 if (time_before(now, stats->rx_jiffies)) {
1291 stats->rx_jiffies = now;
1295 /* Update once a second */
1300 start = u64_stats_fetch_begin_bh(&stats->sync);
1301 pkts = stats->rx_pkts;
1302 } while (u64_stats_fetch_retry_bh(&stats->sync, start));
1304 stats->rx_pps = (unsigned long)(pkts - stats->rx_pkts_prev) / (delta / HZ);
1305 stats->rx_pkts_prev = pkts;
1306 stats->rx_jiffies = now;
1307 eqd = (stats->rx_pps / 110000) << 3;
1308 eqd = min(eqd, eqo->max_eqd);
1309 eqd = max(eqd, eqo->min_eqd);
1314 if (eqd != eqo->cur_eqd) {
1315 be_cmd_modify_eqd(adapter, eqo->q.id, eqd);
1320 static void be_rx_stats_update(struct be_rx_obj *rxo,
1321 struct be_rx_compl_info *rxcp)
1323 struct be_rx_stats *stats = rx_stats(rxo);
1325 u64_stats_update_begin(&stats->sync);
1327 stats->rx_bytes += rxcp->pkt_size;
1329 if (rxcp->pkt_type == BE_MULTICAST_PACKET)
1330 stats->rx_mcast_pkts++;
1332 stats->rx_compl_err++;
1333 u64_stats_update_end(&stats->sync);
1336 static inline bool csum_passed(struct be_rx_compl_info *rxcp)
1338 /* L4 checksum is not reliable for non TCP/UDP packets.
1339 * Also ignore ipcksm for ipv6 pkts */
1340 return (rxcp->tcpf || rxcp->udpf) && rxcp->l4_csum &&
1341 (rxcp->ip_csum || rxcp->ipv6);
1344 static struct be_rx_page_info *get_rx_page_info(struct be_rx_obj *rxo,
1347 struct be_adapter *adapter = rxo->adapter;
1348 struct be_rx_page_info *rx_page_info;
1349 struct be_queue_info *rxq = &rxo->q;
1351 rx_page_info = &rxo->page_info_tbl[frag_idx];
1352 BUG_ON(!rx_page_info->page);
1354 if (rx_page_info->last_page_user) {
1355 dma_unmap_page(&adapter->pdev->dev,
1356 dma_unmap_addr(rx_page_info, bus),
1357 adapter->big_page_size, DMA_FROM_DEVICE);
1358 rx_page_info->last_page_user = false;
1361 atomic_dec(&rxq->used);
1362 return rx_page_info;
1365 /* Throwaway the data in the Rx completion */
1366 static void be_rx_compl_discard(struct be_rx_obj *rxo,
1367 struct be_rx_compl_info *rxcp)
1369 struct be_queue_info *rxq = &rxo->q;
1370 struct be_rx_page_info *page_info;
1371 u16 i, num_rcvd = rxcp->num_rcvd;
1373 for (i = 0; i < num_rcvd; i++) {
1374 page_info = get_rx_page_info(rxo, rxcp->rxq_idx);
1375 put_page(page_info->page);
1376 memset(page_info, 0, sizeof(*page_info));
1377 index_inc(&rxcp->rxq_idx, rxq->len);
1382 * skb_fill_rx_data forms a complete skb for an ether frame
1383 * indicated by rxcp.
1385 static void skb_fill_rx_data(struct be_rx_obj *rxo, struct sk_buff *skb,
1386 struct be_rx_compl_info *rxcp)
1388 struct be_queue_info *rxq = &rxo->q;
1389 struct be_rx_page_info *page_info;
1391 u16 hdr_len, curr_frag_len, remaining;
1394 page_info = get_rx_page_info(rxo, rxcp->rxq_idx);
1395 start = page_address(page_info->page) + page_info->page_offset;
1398 /* Copy data in the first descriptor of this completion */
1399 curr_frag_len = min(rxcp->pkt_size, rx_frag_size);
1401 skb->len = curr_frag_len;
1402 if (curr_frag_len <= BE_HDR_LEN) { /* tiny packet */
1403 memcpy(skb->data, start, curr_frag_len);
1404 /* Complete packet has now been moved to data */
1405 put_page(page_info->page);
1407 skb->tail += curr_frag_len;
1410 memcpy(skb->data, start, hdr_len);
1411 skb_shinfo(skb)->nr_frags = 1;
1412 skb_frag_set_page(skb, 0, page_info->page);
1413 skb_shinfo(skb)->frags[0].page_offset =
1414 page_info->page_offset + hdr_len;
1415 skb_frag_size_set(&skb_shinfo(skb)->frags[0], curr_frag_len - hdr_len);
1416 skb->data_len = curr_frag_len - hdr_len;
1417 skb->truesize += rx_frag_size;
1418 skb->tail += hdr_len;
1420 page_info->page = NULL;
1422 if (rxcp->pkt_size <= rx_frag_size) {
1423 BUG_ON(rxcp->num_rcvd != 1);
1427 /* More frags present for this completion */
1428 index_inc(&rxcp->rxq_idx, rxq->len);
1429 remaining = rxcp->pkt_size - curr_frag_len;
1430 for (i = 1, j = 0; i < rxcp->num_rcvd; i++) {
1431 page_info = get_rx_page_info(rxo, rxcp->rxq_idx);
1432 curr_frag_len = min(remaining, rx_frag_size);
1434 /* Coalesce all frags from the same physical page in one slot */
1435 if (page_info->page_offset == 0) {
1438 skb_frag_set_page(skb, j, page_info->page);
1439 skb_shinfo(skb)->frags[j].page_offset =
1440 page_info->page_offset;
1441 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1442 skb_shinfo(skb)->nr_frags++;
1444 put_page(page_info->page);
1447 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1448 skb->len += curr_frag_len;
1449 skb->data_len += curr_frag_len;
1450 skb->truesize += rx_frag_size;
1451 remaining -= curr_frag_len;
1452 index_inc(&rxcp->rxq_idx, rxq->len);
1453 page_info->page = NULL;
1455 BUG_ON(j > MAX_SKB_FRAGS);
1458 /* Process the RX completion indicated by rxcp when GRO is disabled */
1459 static void be_rx_compl_process(struct be_rx_obj *rxo,
1460 struct be_rx_compl_info *rxcp)
1462 struct be_adapter *adapter = rxo->adapter;
1463 struct net_device *netdev = adapter->netdev;
1464 struct sk_buff *skb;
1466 skb = netdev_alloc_skb_ip_align(netdev, BE_RX_SKB_ALLOC_SIZE);
1467 if (unlikely(!skb)) {
1468 rx_stats(rxo)->rx_drops_no_skbs++;
1469 be_rx_compl_discard(rxo, rxcp);
1473 skb_fill_rx_data(rxo, skb, rxcp);
1475 if (likely((netdev->features & NETIF_F_RXCSUM) && csum_passed(rxcp)))
1476 skb->ip_summed = CHECKSUM_UNNECESSARY;
1478 skb_checksum_none_assert(skb);
1480 skb->protocol = eth_type_trans(skb, netdev);
1481 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1482 if (netdev->features & NETIF_F_RXHASH)
1483 skb->rxhash = rxcp->rss_hash;
1487 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1489 netif_receive_skb(skb);
1492 /* Process the RX completion indicated by rxcp when GRO is enabled */
1493 void be_rx_compl_process_gro(struct be_rx_obj *rxo, struct napi_struct *napi,
1494 struct be_rx_compl_info *rxcp)
1496 struct be_adapter *adapter = rxo->adapter;
1497 struct be_rx_page_info *page_info;
1498 struct sk_buff *skb = NULL;
1499 struct be_queue_info *rxq = &rxo->q;
1500 u16 remaining, curr_frag_len;
1503 skb = napi_get_frags(napi);
1505 be_rx_compl_discard(rxo, rxcp);
1509 remaining = rxcp->pkt_size;
1510 for (i = 0, j = -1; i < rxcp->num_rcvd; i++) {
1511 page_info = get_rx_page_info(rxo, rxcp->rxq_idx);
1513 curr_frag_len = min(remaining, rx_frag_size);
1515 /* Coalesce all frags from the same physical page in one slot */
1516 if (i == 0 || page_info->page_offset == 0) {
1517 /* First frag or Fresh page */
1519 skb_frag_set_page(skb, j, page_info->page);
1520 skb_shinfo(skb)->frags[j].page_offset =
1521 page_info->page_offset;
1522 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1524 put_page(page_info->page);
1526 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1527 skb->truesize += rx_frag_size;
1528 remaining -= curr_frag_len;
1529 index_inc(&rxcp->rxq_idx, rxq->len);
1530 memset(page_info, 0, sizeof(*page_info));
1532 BUG_ON(j > MAX_SKB_FRAGS);
1534 skb_shinfo(skb)->nr_frags = j + 1;
1535 skb->len = rxcp->pkt_size;
1536 skb->data_len = rxcp->pkt_size;
1537 skb->ip_summed = CHECKSUM_UNNECESSARY;
1538 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1539 if (adapter->netdev->features & NETIF_F_RXHASH)
1540 skb->rxhash = rxcp->rss_hash;
1543 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1545 napi_gro_frags(napi);
1548 static void be_parse_rx_compl_v1(struct be_eth_rx_compl *compl,
1549 struct be_rx_compl_info *rxcp)
1552 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, pktsize, compl);
1553 rxcp->vlanf = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, vtp, compl);
1554 rxcp->err = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, err, compl);
1555 rxcp->tcpf = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, tcpf, compl);
1556 rxcp->udpf = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, udpf, compl);
1558 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, ipcksm, compl);
1560 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, l4_cksm, compl);
1562 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, ip_version, compl);
1564 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, fragndx, compl);
1566 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, numfrags, compl);
1568 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, cast_enc, compl);
1570 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, rsshash, compl);
1572 rxcp->vtm = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, vtm,
1574 rxcp->vlan_tag = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, vlan_tag,
1577 rxcp->port = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, port, compl);
1580 static void be_parse_rx_compl_v0(struct be_eth_rx_compl *compl,
1581 struct be_rx_compl_info *rxcp)
1584 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, pktsize, compl);
1585 rxcp->vlanf = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, vtp, compl);
1586 rxcp->err = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, err, compl);
1587 rxcp->tcpf = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, tcpf, compl);
1588 rxcp->udpf = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, udpf, compl);
1590 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, ipcksm, compl);
1592 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, l4_cksm, compl);
1594 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, ip_version, compl);
1596 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, fragndx, compl);
1598 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, numfrags, compl);
1600 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, cast_enc, compl);
1602 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, rsshash, compl);
1604 rxcp->vtm = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, vtm,
1606 rxcp->vlan_tag = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, vlan_tag,
1609 rxcp->port = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, port, compl);
1610 rxcp->ip_frag = AMAP_GET_BITS(struct amap_eth_rx_compl_v0,
1614 static struct be_rx_compl_info *be_rx_compl_get(struct be_rx_obj *rxo)
1616 struct be_eth_rx_compl *compl = queue_tail_node(&rxo->cq);
1617 struct be_rx_compl_info *rxcp = &rxo->rxcp;
1618 struct be_adapter *adapter = rxo->adapter;
1620 /* For checking the valid bit it is Ok to use either definition as the
1621 * valid bit is at the same position in both v0 and v1 Rx compl */
1622 if (compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] == 0)
1626 be_dws_le_to_cpu(compl, sizeof(*compl));
1628 if (adapter->be3_native)
1629 be_parse_rx_compl_v1(compl, rxcp);
1631 be_parse_rx_compl_v0(compl, rxcp);
1637 /* vlanf could be wrongly set in some cards.
1638 * ignore if vtm is not set */
1639 if ((adapter->function_mode & FLEX10_MODE) && !rxcp->vtm)
1642 if (!lancer_chip(adapter))
1643 rxcp->vlan_tag = swab16(rxcp->vlan_tag);
1645 if (adapter->pvid == (rxcp->vlan_tag & VLAN_VID_MASK) &&
1646 !adapter->vlan_tag[rxcp->vlan_tag])
1650 /* As the compl has been parsed, reset it; we wont touch it again */
1651 compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] = 0;
1653 queue_tail_inc(&rxo->cq);
1657 static inline struct page *be_alloc_pages(u32 size, gfp_t gfp)
1659 u32 order = get_order(size);
1663 return alloc_pages(gfp, order);
1667 * Allocate a page, split it to fragments of size rx_frag_size and post as
1668 * receive buffers to BE
1670 static void be_post_rx_frags(struct be_rx_obj *rxo, gfp_t gfp)
1672 struct be_adapter *adapter = rxo->adapter;
1673 struct be_rx_page_info *page_info = NULL, *prev_page_info = NULL;
1674 struct be_queue_info *rxq = &rxo->q;
1675 struct page *pagep = NULL;
1676 struct be_eth_rx_d *rxd;
1677 u64 page_dmaaddr = 0, frag_dmaaddr;
1678 u32 posted, page_offset = 0;
1680 page_info = &rxo->page_info_tbl[rxq->head];
1681 for (posted = 0; posted < MAX_RX_POST && !page_info->page; posted++) {
1683 pagep = be_alloc_pages(adapter->big_page_size, gfp);
1684 if (unlikely(!pagep)) {
1685 rx_stats(rxo)->rx_post_fail++;
1688 page_dmaaddr = dma_map_page(&adapter->pdev->dev, pagep,
1689 0, adapter->big_page_size,
1691 page_info->page_offset = 0;
1694 page_info->page_offset = page_offset + rx_frag_size;
1696 page_offset = page_info->page_offset;
1697 page_info->page = pagep;
1698 dma_unmap_addr_set(page_info, bus, page_dmaaddr);
1699 frag_dmaaddr = page_dmaaddr + page_info->page_offset;
1701 rxd = queue_head_node(rxq);
1702 rxd->fragpa_lo = cpu_to_le32(frag_dmaaddr & 0xFFFFFFFF);
1703 rxd->fragpa_hi = cpu_to_le32(upper_32_bits(frag_dmaaddr));
1705 /* Any space left in the current big page for another frag? */
1706 if ((page_offset + rx_frag_size + rx_frag_size) >
1707 adapter->big_page_size) {
1709 page_info->last_page_user = true;
1712 prev_page_info = page_info;
1713 queue_head_inc(rxq);
1714 page_info = &rxo->page_info_tbl[rxq->head];
1717 prev_page_info->last_page_user = true;
1720 atomic_add(posted, &rxq->used);
1721 be_rxq_notify(adapter, rxq->id, posted);
1722 } else if (atomic_read(&rxq->used) == 0) {
1723 /* Let be_worker replenish when memory is available */
1724 rxo->rx_post_starved = true;
1728 static struct be_eth_tx_compl *be_tx_compl_get(struct be_queue_info *tx_cq)
1730 struct be_eth_tx_compl *txcp = queue_tail_node(tx_cq);
1732 if (txcp->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] == 0)
1736 be_dws_le_to_cpu(txcp, sizeof(*txcp));
1738 txcp->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] = 0;
1740 queue_tail_inc(tx_cq);
1744 static u16 be_tx_compl_process(struct be_adapter *adapter,
1745 struct be_tx_obj *txo, u16 last_index)
1747 struct be_queue_info *txq = &txo->q;
1748 struct be_eth_wrb *wrb;
1749 struct sk_buff **sent_skbs = txo->sent_skb_list;
1750 struct sk_buff *sent_skb;
1751 u16 cur_index, num_wrbs = 1; /* account for hdr wrb */
1752 bool unmap_skb_hdr = true;
1754 sent_skb = sent_skbs[txq->tail];
1756 sent_skbs[txq->tail] = NULL;
1758 /* skip header wrb */
1759 queue_tail_inc(txq);
1762 cur_index = txq->tail;
1763 wrb = queue_tail_node(txq);
1764 unmap_tx_frag(&adapter->pdev->dev, wrb,
1765 (unmap_skb_hdr && skb_headlen(sent_skb)));
1766 unmap_skb_hdr = false;
1769 queue_tail_inc(txq);
1770 } while (cur_index != last_index);
1772 kfree_skb(sent_skb);
1776 /* Return the number of events in the event queue */
1777 static inline int events_get(struct be_eq_obj *eqo)
1779 struct be_eq_entry *eqe;
1783 eqe = queue_tail_node(&eqo->q);
1790 queue_tail_inc(&eqo->q);
1796 /* Leaves the EQ is disarmed state */
1797 static void be_eq_clean(struct be_eq_obj *eqo)
1799 int num = events_get(eqo);
1801 be_eq_notify(eqo->adapter, eqo->q.id, false, true, num);
1804 static void be_rx_cq_clean(struct be_rx_obj *rxo)
1806 struct be_rx_page_info *page_info;
1807 struct be_queue_info *rxq = &rxo->q;
1808 struct be_queue_info *rx_cq = &rxo->cq;
1809 struct be_rx_compl_info *rxcp;
1810 struct be_adapter *adapter = rxo->adapter;
1814 /* Consume pending rx completions.
1815 * Wait for the flush completion (identified by zero num_rcvd)
1816 * to arrive. Notify CQ even when there are no more CQ entries
1817 * for HW to flush partially coalesced CQ entries.
1818 * In Lancer, there is no need to wait for flush compl.
1821 rxcp = be_rx_compl_get(rxo);
1823 if (lancer_chip(adapter))
1826 if (flush_wait++ > 10 || be_hw_error(adapter)) {
1827 dev_warn(&adapter->pdev->dev,
1828 "did not receive flush compl\n");
1831 be_cq_notify(adapter, rx_cq->id, true, 0);
1834 be_rx_compl_discard(rxo, rxcp);
1835 be_cq_notify(adapter, rx_cq->id, false, 1);
1836 if (rxcp->num_rcvd == 0)
1841 /* After cleanup, leave the CQ in unarmed state */
1842 be_cq_notify(adapter, rx_cq->id, false, 0);
1844 /* Then free posted rx buffers that were not used */
1845 tail = (rxq->head + rxq->len - atomic_read(&rxq->used)) % rxq->len;
1846 for (; atomic_read(&rxq->used) > 0; index_inc(&tail, rxq->len)) {
1847 page_info = get_rx_page_info(rxo, tail);
1848 put_page(page_info->page);
1849 memset(page_info, 0, sizeof(*page_info));
1851 BUG_ON(atomic_read(&rxq->used));
1852 rxq->tail = rxq->head = 0;
1855 static void be_tx_compl_clean(struct be_adapter *adapter)
1857 struct be_tx_obj *txo;
1858 struct be_queue_info *txq;
1859 struct be_eth_tx_compl *txcp;
1860 u16 end_idx, cmpl = 0, timeo = 0, num_wrbs = 0;
1861 struct sk_buff *sent_skb;
1863 int i, pending_txqs;
1865 /* Wait for a max of 200ms for all the tx-completions to arrive. */
1867 pending_txqs = adapter->num_tx_qs;
1869 for_all_tx_queues(adapter, txo, i) {
1871 while ((txcp = be_tx_compl_get(&txo->cq))) {
1873 AMAP_GET_BITS(struct amap_eth_tx_compl,
1875 num_wrbs += be_tx_compl_process(adapter, txo,
1880 be_cq_notify(adapter, txo->cq.id, false, cmpl);
1881 atomic_sub(num_wrbs, &txq->used);
1885 if (atomic_read(&txq->used) == 0)
1889 if (pending_txqs == 0 || ++timeo > 200)
1895 for_all_tx_queues(adapter, txo, i) {
1897 if (atomic_read(&txq->used))
1898 dev_err(&adapter->pdev->dev, "%d pending tx-compls\n",
1899 atomic_read(&txq->used));
1901 /* free posted tx for which compls will never arrive */
1902 while (atomic_read(&txq->used)) {
1903 sent_skb = txo->sent_skb_list[txq->tail];
1904 end_idx = txq->tail;
1905 num_wrbs = wrb_cnt_for_skb(adapter, sent_skb,
1907 index_adv(&end_idx, num_wrbs - 1, txq->len);
1908 num_wrbs = be_tx_compl_process(adapter, txo, end_idx);
1909 atomic_sub(num_wrbs, &txq->used);
1914 static void be_evt_queues_destroy(struct be_adapter *adapter)
1916 struct be_eq_obj *eqo;
1919 for_all_evt_queues(adapter, eqo, i) {
1920 if (eqo->q.created) {
1922 be_cmd_q_destroy(adapter, &eqo->q, QTYPE_EQ);
1924 be_queue_free(adapter, &eqo->q);
1928 static int be_evt_queues_create(struct be_adapter *adapter)
1930 struct be_queue_info *eq;
1931 struct be_eq_obj *eqo;
1934 adapter->num_evt_qs = num_irqs(adapter);
1936 for_all_evt_queues(adapter, eqo, i) {
1937 eqo->adapter = adapter;
1938 eqo->tx_budget = BE_TX_BUDGET;
1940 eqo->max_eqd = BE_MAX_EQD;
1941 eqo->enable_aic = true;
1944 rc = be_queue_alloc(adapter, eq, EVNT_Q_LEN,
1945 sizeof(struct be_eq_entry));
1949 rc = be_cmd_eq_create(adapter, eq, eqo->cur_eqd);
1956 static void be_mcc_queues_destroy(struct be_adapter *adapter)
1958 struct be_queue_info *q;
1960 q = &adapter->mcc_obj.q;
1962 be_cmd_q_destroy(adapter, q, QTYPE_MCCQ);
1963 be_queue_free(adapter, q);
1965 q = &adapter->mcc_obj.cq;
1967 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
1968 be_queue_free(adapter, q);
1971 /* Must be called only after TX qs are created as MCC shares TX EQ */
1972 static int be_mcc_queues_create(struct be_adapter *adapter)
1974 struct be_queue_info *q, *cq;
1976 cq = &adapter->mcc_obj.cq;
1977 if (be_queue_alloc(adapter, cq, MCC_CQ_LEN,
1978 sizeof(struct be_mcc_compl)))
1981 /* Use the default EQ for MCC completions */
1982 if (be_cmd_cq_create(adapter, cq, &mcc_eqo(adapter)->q, true, 0))
1985 q = &adapter->mcc_obj.q;
1986 if (be_queue_alloc(adapter, q, MCC_Q_LEN, sizeof(struct be_mcc_wrb)))
1987 goto mcc_cq_destroy;
1989 if (be_cmd_mccq_create(adapter, q, cq))
1995 be_queue_free(adapter, q);
1997 be_cmd_q_destroy(adapter, cq, QTYPE_CQ);
1999 be_queue_free(adapter, cq);
2004 static void be_tx_queues_destroy(struct be_adapter *adapter)
2006 struct be_queue_info *q;
2007 struct be_tx_obj *txo;
2010 for_all_tx_queues(adapter, txo, i) {
2013 be_cmd_q_destroy(adapter, q, QTYPE_TXQ);
2014 be_queue_free(adapter, q);
2018 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2019 be_queue_free(adapter, q);
2023 static int be_num_txqs_want(struct be_adapter *adapter)
2025 if ((!lancer_chip(adapter) && sriov_want(adapter)) ||
2026 be_is_mc(adapter) ||
2027 (!lancer_chip(adapter) && !be_physfn(adapter)) ||
2031 return adapter->max_tx_queues;
2034 static int be_tx_cqs_create(struct be_adapter *adapter)
2036 struct be_queue_info *cq, *eq;
2038 struct be_tx_obj *txo;
2041 adapter->num_tx_qs = be_num_txqs_want(adapter);
2042 if (adapter->num_tx_qs != MAX_TX_QS) {
2044 netif_set_real_num_tx_queues(adapter->netdev,
2045 adapter->num_tx_qs);
2049 for_all_tx_queues(adapter, txo, i) {
2051 status = be_queue_alloc(adapter, cq, TX_CQ_LEN,
2052 sizeof(struct be_eth_tx_compl));
2056 /* If num_evt_qs is less than num_tx_qs, then more than
2057 * one txq share an eq
2059 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2060 status = be_cmd_cq_create(adapter, cq, eq, false, 3);
2067 static int be_tx_qs_create(struct be_adapter *adapter)
2069 struct be_tx_obj *txo;
2072 for_all_tx_queues(adapter, txo, i) {
2073 status = be_queue_alloc(adapter, &txo->q, TX_Q_LEN,
2074 sizeof(struct be_eth_wrb));
2078 status = be_cmd_txq_create(adapter, txo);
2083 dev_info(&adapter->pdev->dev, "created %d TX queue(s)\n",
2084 adapter->num_tx_qs);
2088 static void be_rx_cqs_destroy(struct be_adapter *adapter)
2090 struct be_queue_info *q;
2091 struct be_rx_obj *rxo;
2094 for_all_rx_queues(adapter, rxo, i) {
2097 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2098 be_queue_free(adapter, q);
2102 static int be_rx_cqs_create(struct be_adapter *adapter)
2104 struct be_queue_info *eq, *cq;
2105 struct be_rx_obj *rxo;
2108 /* We'll create as many RSS rings as there are irqs.
2109 * But when there's only one irq there's no use creating RSS rings
2111 adapter->num_rx_qs = (num_irqs(adapter) > 1) ?
2112 num_irqs(adapter) + 1 : 1;
2113 if (adapter->num_rx_qs != MAX_RX_QS) {
2115 netif_set_real_num_rx_queues(adapter->netdev,
2116 adapter->num_rx_qs);
2120 adapter->big_page_size = (1 << get_order(rx_frag_size)) * PAGE_SIZE;
2121 for_all_rx_queues(adapter, rxo, i) {
2122 rxo->adapter = adapter;
2124 rc = be_queue_alloc(adapter, cq, RX_CQ_LEN,
2125 sizeof(struct be_eth_rx_compl));
2129 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2130 rc = be_cmd_cq_create(adapter, cq, eq, false, 3);
2135 dev_info(&adapter->pdev->dev,
2136 "created %d RSS queue(s) and 1 default RX queue\n",
2137 adapter->num_rx_qs - 1);
2141 static irqreturn_t be_intx(int irq, void *dev)
2143 struct be_eq_obj *eqo = dev;
2144 struct be_adapter *adapter = eqo->adapter;
2147 /* IRQ is not expected when NAPI is scheduled as the EQ
2148 * will not be armed.
2149 * But, this can happen on Lancer INTx where it takes
2150 * a while to de-assert INTx or in BE2 where occasionaly
2151 * an interrupt may be raised even when EQ is unarmed.
2152 * If NAPI is already scheduled, then counting & notifying
2153 * events will orphan them.
2155 if (napi_schedule_prep(&eqo->napi)) {
2156 num_evts = events_get(eqo);
2157 __napi_schedule(&eqo->napi);
2159 eqo->spurious_intr = 0;
2161 be_eq_notify(adapter, eqo->q.id, false, true, num_evts);
2163 /* Return IRQ_HANDLED only for the the first spurious intr
2164 * after a valid intr to stop the kernel from branding
2165 * this irq as a bad one!
2167 if (num_evts || eqo->spurious_intr++ == 0)
2173 static irqreturn_t be_msix(int irq, void *dev)
2175 struct be_eq_obj *eqo = dev;
2177 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
2178 napi_schedule(&eqo->napi);
2182 static inline bool do_gro(struct be_rx_compl_info *rxcp)
2184 return (rxcp->tcpf && !rxcp->err && rxcp->l4_csum) ? true : false;
2187 static int be_process_rx(struct be_rx_obj *rxo, struct napi_struct *napi,
2190 struct be_adapter *adapter = rxo->adapter;
2191 struct be_queue_info *rx_cq = &rxo->cq;
2192 struct be_rx_compl_info *rxcp;
2195 for (work_done = 0; work_done < budget; work_done++) {
2196 rxcp = be_rx_compl_get(rxo);
2200 /* Is it a flush compl that has no data */
2201 if (unlikely(rxcp->num_rcvd == 0))
2204 /* Discard compl with partial DMA Lancer B0 */
2205 if (unlikely(!rxcp->pkt_size)) {
2206 be_rx_compl_discard(rxo, rxcp);
2210 /* On BE drop pkts that arrive due to imperfect filtering in
2211 * promiscuous mode on some skews
2213 if (unlikely(rxcp->port != adapter->port_num &&
2214 !lancer_chip(adapter))) {
2215 be_rx_compl_discard(rxo, rxcp);
2220 be_rx_compl_process_gro(rxo, napi, rxcp);
2222 be_rx_compl_process(rxo, rxcp);
2224 be_rx_stats_update(rxo, rxcp);
2228 be_cq_notify(adapter, rx_cq->id, true, work_done);
2230 if (atomic_read(&rxo->q.used) < RX_FRAGS_REFILL_WM)
2231 be_post_rx_frags(rxo, GFP_ATOMIC);
2237 static bool be_process_tx(struct be_adapter *adapter, struct be_tx_obj *txo,
2238 int budget, int idx)
2240 struct be_eth_tx_compl *txcp;
2241 int num_wrbs = 0, work_done;
2243 for (work_done = 0; work_done < budget; work_done++) {
2244 txcp = be_tx_compl_get(&txo->cq);
2247 num_wrbs += be_tx_compl_process(adapter, txo,
2248 AMAP_GET_BITS(struct amap_eth_tx_compl,
2253 be_cq_notify(adapter, txo->cq.id, true, work_done);
2254 atomic_sub(num_wrbs, &txo->q.used);
2256 /* As Tx wrbs have been freed up, wake up netdev queue
2257 * if it was stopped due to lack of tx wrbs. */
2258 if (__netif_subqueue_stopped(adapter->netdev, idx) &&
2259 atomic_read(&txo->q.used) < txo->q.len / 2) {
2260 netif_wake_subqueue(adapter->netdev, idx);
2263 u64_stats_update_begin(&tx_stats(txo)->sync_compl);
2264 tx_stats(txo)->tx_compl += work_done;
2265 u64_stats_update_end(&tx_stats(txo)->sync_compl);
2267 return (work_done < budget); /* Done */
2270 int be_poll(struct napi_struct *napi, int budget)
2272 struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2273 struct be_adapter *adapter = eqo->adapter;
2274 int max_work = 0, work, i, num_evts;
2277 num_evts = events_get(eqo);
2279 /* Process all TXQs serviced by this EQ */
2280 for (i = eqo->idx; i < adapter->num_tx_qs; i += adapter->num_evt_qs) {
2281 tx_done = be_process_tx(adapter, &adapter->tx_obj[i],
2287 /* This loop will iterate twice for EQ0 in which
2288 * completions of the last RXQ (default one) are also processed
2289 * For other EQs the loop iterates only once
2291 for (i = eqo->idx; i < adapter->num_rx_qs; i += adapter->num_evt_qs) {
2292 work = be_process_rx(&adapter->rx_obj[i], napi, budget);
2293 max_work = max(work, max_work);
2296 if (is_mcc_eqo(eqo))
2297 be_process_mcc(adapter);
2299 if (max_work < budget) {
2300 napi_complete(napi);
2301 be_eq_notify(adapter, eqo->q.id, true, false, num_evts);
2303 /* As we'll continue in polling mode, count and clear events */
2304 be_eq_notify(adapter, eqo->q.id, false, false, num_evts);
2309 void be_detect_error(struct be_adapter *adapter)
2311 u32 ue_lo = 0, ue_hi = 0, ue_lo_mask = 0, ue_hi_mask = 0;
2312 u32 sliport_status = 0, sliport_err1 = 0, sliport_err2 = 0;
2315 if (be_hw_error(adapter))
2318 if (lancer_chip(adapter)) {
2319 sliport_status = ioread32(adapter->db + SLIPORT_STATUS_OFFSET);
2320 if (sliport_status & SLIPORT_STATUS_ERR_MASK) {
2321 sliport_err1 = ioread32(adapter->db +
2322 SLIPORT_ERROR1_OFFSET);
2323 sliport_err2 = ioread32(adapter->db +
2324 SLIPORT_ERROR2_OFFSET);
2327 pci_read_config_dword(adapter->pdev,
2328 PCICFG_UE_STATUS_LOW, &ue_lo);
2329 pci_read_config_dword(adapter->pdev,
2330 PCICFG_UE_STATUS_HIGH, &ue_hi);
2331 pci_read_config_dword(adapter->pdev,
2332 PCICFG_UE_STATUS_LOW_MASK, &ue_lo_mask);
2333 pci_read_config_dword(adapter->pdev,
2334 PCICFG_UE_STATUS_HI_MASK, &ue_hi_mask);
2336 ue_lo = (ue_lo & ~ue_lo_mask);
2337 ue_hi = (ue_hi & ~ue_hi_mask);
2340 /* On certain platforms BE hardware can indicate spurious UEs.
2341 * Allow the h/w to stop working completely in case of a real UE.
2342 * Hence not setting the hw_error for UE detection.
2344 if (sliport_status & SLIPORT_STATUS_ERR_MASK) {
2345 adapter->hw_error = true;
2346 dev_err(&adapter->pdev->dev,
2347 "Error detected in the card\n");
2350 if (sliport_status & SLIPORT_STATUS_ERR_MASK) {
2351 dev_err(&adapter->pdev->dev,
2352 "ERR: sliport status 0x%x\n", sliport_status);
2353 dev_err(&adapter->pdev->dev,
2354 "ERR: sliport error1 0x%x\n", sliport_err1);
2355 dev_err(&adapter->pdev->dev,
2356 "ERR: sliport error2 0x%x\n", sliport_err2);
2360 for (i = 0; ue_lo; ue_lo >>= 1, i++) {
2362 dev_err(&adapter->pdev->dev,
2363 "UE: %s bit set\n", ue_status_low_desc[i]);
2368 for (i = 0; ue_hi; ue_hi >>= 1, i++) {
2370 dev_err(&adapter->pdev->dev,
2371 "UE: %s bit set\n", ue_status_hi_desc[i]);
2377 static void be_msix_disable(struct be_adapter *adapter)
2379 if (msix_enabled(adapter)) {
2380 pci_disable_msix(adapter->pdev);
2381 adapter->num_msix_vec = 0;
2385 static uint be_num_rss_want(struct be_adapter *adapter)
2389 if ((adapter->function_caps & BE_FUNCTION_CAPS_RSS) &&
2390 (lancer_chip(adapter) ||
2391 (!sriov_want(adapter) && be_physfn(adapter)))) {
2392 num = adapter->max_rss_queues;
2393 num = min_t(u32, num, (u32)netif_get_num_default_rss_queues());
2398 static int be_msix_enable(struct be_adapter *adapter)
2400 #define BE_MIN_MSIX_VECTORS 1
2401 int i, status, num_vec, num_roce_vec = 0;
2402 struct device *dev = &adapter->pdev->dev;
2404 /* If RSS queues are not used, need a vec for default RX Q */
2405 num_vec = min(be_num_rss_want(adapter), num_online_cpus());
2406 if (be_roce_supported(adapter)) {
2407 num_roce_vec = min_t(u32, MAX_ROCE_MSIX_VECTORS,
2408 (num_online_cpus() + 1));
2409 num_roce_vec = min(num_roce_vec, MAX_ROCE_EQS);
2410 num_vec += num_roce_vec;
2411 num_vec = min(num_vec, MAX_MSIX_VECTORS);
2413 num_vec = max(num_vec, BE_MIN_MSIX_VECTORS);
2415 for (i = 0; i < num_vec; i++)
2416 adapter->msix_entries[i].entry = i;
2418 status = pci_enable_msix(adapter->pdev, adapter->msix_entries, num_vec);
2421 } else if (status >= BE_MIN_MSIX_VECTORS) {
2423 status = pci_enable_msix(adapter->pdev, adapter->msix_entries,
2429 dev_warn(dev, "MSIx enable failed\n");
2430 /* INTx is not supported in VFs, so fail probe if enable_msix fails */
2431 if (!be_physfn(adapter))
2435 if (be_roce_supported(adapter)) {
2436 if (num_vec > num_roce_vec) {
2437 adapter->num_msix_vec = num_vec - num_roce_vec;
2438 adapter->num_msix_roce_vec =
2439 num_vec - adapter->num_msix_vec;
2441 adapter->num_msix_vec = num_vec;
2442 adapter->num_msix_roce_vec = 0;
2445 adapter->num_msix_vec = num_vec;
2446 dev_info(dev, "enabled %d MSI-x vector(s)\n", adapter->num_msix_vec);
2450 static inline int be_msix_vec_get(struct be_adapter *adapter,
2451 struct be_eq_obj *eqo)
2453 return adapter->msix_entries[eqo->idx].vector;
2456 static int be_msix_register(struct be_adapter *adapter)
2458 struct net_device *netdev = adapter->netdev;
2459 struct be_eq_obj *eqo;
2462 for_all_evt_queues(adapter, eqo, i) {
2463 sprintf(eqo->desc, "%s-q%d", netdev->name, i);
2464 vec = be_msix_vec_get(adapter, eqo);
2465 status = request_irq(vec, be_msix, 0, eqo->desc, eqo);
2472 for (i--, eqo = &adapter->eq_obj[i]; i >= 0; i--, eqo--)
2473 free_irq(be_msix_vec_get(adapter, eqo), eqo);
2474 dev_warn(&adapter->pdev->dev, "MSIX Request IRQ failed - err %d\n",
2476 be_msix_disable(adapter);
2480 static int be_irq_register(struct be_adapter *adapter)
2482 struct net_device *netdev = adapter->netdev;
2485 if (msix_enabled(adapter)) {
2486 status = be_msix_register(adapter);
2489 /* INTx is not supported for VF */
2490 if (!be_physfn(adapter))
2494 /* INTx: only the first EQ is used */
2495 netdev->irq = adapter->pdev->irq;
2496 status = request_irq(netdev->irq, be_intx, IRQF_SHARED, netdev->name,
2497 &adapter->eq_obj[0]);
2499 dev_err(&adapter->pdev->dev,
2500 "INTx request IRQ failed - err %d\n", status);
2504 adapter->isr_registered = true;
2508 static void be_irq_unregister(struct be_adapter *adapter)
2510 struct net_device *netdev = adapter->netdev;
2511 struct be_eq_obj *eqo;
2514 if (!adapter->isr_registered)
2518 if (!msix_enabled(adapter)) {
2519 free_irq(netdev->irq, &adapter->eq_obj[0]);
2524 for_all_evt_queues(adapter, eqo, i)
2525 free_irq(be_msix_vec_get(adapter, eqo), eqo);
2528 adapter->isr_registered = false;
2531 static void be_rx_qs_destroy(struct be_adapter *adapter)
2533 struct be_queue_info *q;
2534 struct be_rx_obj *rxo;
2537 for_all_rx_queues(adapter, rxo, i) {
2540 be_cmd_rxq_destroy(adapter, q);
2541 be_rx_cq_clean(rxo);
2543 be_queue_free(adapter, q);
2547 static int be_close(struct net_device *netdev)
2549 struct be_adapter *adapter = netdev_priv(netdev);
2550 struct be_eq_obj *eqo;
2553 be_roce_dev_close(adapter);
2555 if (adapter->flags & BE_FLAGS_NAPI_ENABLED) {
2556 for_all_evt_queues(adapter, eqo, i)
2557 napi_disable(&eqo->napi);
2558 adapter->flags &= ~BE_FLAGS_NAPI_ENABLED;
2561 be_async_mcc_disable(adapter);
2563 /* Wait for all pending tx completions to arrive so that
2564 * all tx skbs are freed.
2566 netif_tx_disable(netdev);
2567 be_tx_compl_clean(adapter);
2569 be_rx_qs_destroy(adapter);
2571 for_all_evt_queues(adapter, eqo, i) {
2572 if (msix_enabled(adapter))
2573 synchronize_irq(be_msix_vec_get(adapter, eqo));
2575 synchronize_irq(netdev->irq);
2579 be_irq_unregister(adapter);
2584 static int be_rx_qs_create(struct be_adapter *adapter)
2586 struct be_rx_obj *rxo;
2590 for_all_rx_queues(adapter, rxo, i) {
2591 rc = be_queue_alloc(adapter, &rxo->q, RX_Q_LEN,
2592 sizeof(struct be_eth_rx_d));
2597 /* The FW would like the default RXQ to be created first */
2598 rxo = default_rxo(adapter);
2599 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id, rx_frag_size,
2600 adapter->if_handle, false, &rxo->rss_id);
2604 for_all_rss_queues(adapter, rxo, i) {
2605 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id,
2606 rx_frag_size, adapter->if_handle,
2607 true, &rxo->rss_id);
2612 if (be_multi_rxq(adapter)) {
2613 for (j = 0; j < 128; j += adapter->num_rx_qs - 1) {
2614 for_all_rss_queues(adapter, rxo, i) {
2617 rsstable[j + i] = rxo->rss_id;
2620 adapter->rss_flags = RSS_ENABLE_TCP_IPV4 | RSS_ENABLE_IPV4 |
2621 RSS_ENABLE_TCP_IPV6 | RSS_ENABLE_IPV6;
2623 if (!BEx_chip(adapter))
2624 adapter->rss_flags |= RSS_ENABLE_UDP_IPV4 |
2625 RSS_ENABLE_UDP_IPV6;
2627 rc = be_cmd_rss_config(adapter, rsstable, adapter->rss_flags,
2630 adapter->rss_flags = 0;
2635 /* First time posting */
2636 for_all_rx_queues(adapter, rxo, i)
2637 be_post_rx_frags(rxo, GFP_KERNEL);
2641 static int be_open(struct net_device *netdev)
2643 struct be_adapter *adapter = netdev_priv(netdev);
2644 struct be_eq_obj *eqo;
2645 struct be_rx_obj *rxo;
2646 struct be_tx_obj *txo;
2650 status = be_rx_qs_create(adapter);
2654 status = be_irq_register(adapter);
2658 for_all_rx_queues(adapter, rxo, i)
2659 be_cq_notify(adapter, rxo->cq.id, true, 0);
2661 for_all_tx_queues(adapter, txo, i)
2662 be_cq_notify(adapter, txo->cq.id, true, 0);
2664 be_async_mcc_enable(adapter);
2666 for_all_evt_queues(adapter, eqo, i) {
2667 napi_enable(&eqo->napi);
2668 be_eq_notify(adapter, eqo->q.id, true, false, 0);
2670 adapter->flags |= BE_FLAGS_NAPI_ENABLED;
2672 status = be_cmd_link_status_query(adapter, NULL, &link_status, 0);
2674 be_link_status_update(adapter, link_status);
2676 netif_tx_start_all_queues(netdev);
2677 be_roce_dev_open(adapter);
2680 be_close(adapter->netdev);
2684 static int be_setup_wol(struct be_adapter *adapter, bool enable)
2686 struct be_dma_mem cmd;
2690 memset(mac, 0, ETH_ALEN);
2692 cmd.size = sizeof(struct be_cmd_req_acpi_wol_magic_config);
2693 cmd.va = dma_alloc_coherent(&adapter->pdev->dev, cmd.size, &cmd.dma,
2694 GFP_KERNEL | __GFP_ZERO);
2699 status = pci_write_config_dword(adapter->pdev,
2700 PCICFG_PM_CONTROL_OFFSET, PCICFG_PM_CONTROL_MASK);
2702 dev_err(&adapter->pdev->dev,
2703 "Could not enable Wake-on-lan\n");
2704 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va,
2708 status = be_cmd_enable_magic_wol(adapter,
2709 adapter->netdev->dev_addr, &cmd);
2710 pci_enable_wake(adapter->pdev, PCI_D3hot, 1);
2711 pci_enable_wake(adapter->pdev, PCI_D3cold, 1);
2713 status = be_cmd_enable_magic_wol(adapter, mac, &cmd);
2714 pci_enable_wake(adapter->pdev, PCI_D3hot, 0);
2715 pci_enable_wake(adapter->pdev, PCI_D3cold, 0);
2718 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va, cmd.dma);
2723 * Generate a seed MAC address from the PF MAC Address using jhash.
2724 * MAC Address for VFs are assigned incrementally starting from the seed.
2725 * These addresses are programmed in the ASIC by the PF and the VF driver
2726 * queries for the MAC address during its probe.
2728 static int be_vf_eth_addr_config(struct be_adapter *adapter)
2733 struct be_vf_cfg *vf_cfg;
2735 be_vf_eth_addr_generate(adapter, mac);
2737 for_all_vfs(adapter, vf_cfg, vf) {
2738 if (lancer_chip(adapter)) {
2739 status = be_cmd_set_mac_list(adapter, mac, 1, vf + 1);
2741 status = be_cmd_pmac_add(adapter, mac,
2743 &vf_cfg->pmac_id, vf + 1);
2747 dev_err(&adapter->pdev->dev,
2748 "Mac address assignment failed for VF %d\n", vf);
2750 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
2757 static int be_vfs_mac_query(struct be_adapter *adapter)
2761 struct be_vf_cfg *vf_cfg;
2764 for_all_vfs(adapter, vf_cfg, vf) {
2765 be_cmd_get_mac_from_list(adapter, mac, &active,
2766 &vf_cfg->pmac_id, 0);
2768 status = be_cmd_mac_addr_query(adapter, mac, false,
2769 vf_cfg->if_handle, 0);
2772 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
2777 static void be_vf_clear(struct be_adapter *adapter)
2779 struct be_vf_cfg *vf_cfg;
2782 if (pci_vfs_assigned(adapter->pdev)) {
2783 dev_warn(&adapter->pdev->dev,
2784 "VFs are assigned to VMs: not disabling VFs\n");
2788 pci_disable_sriov(adapter->pdev);
2790 for_all_vfs(adapter, vf_cfg, vf) {
2791 if (lancer_chip(adapter))
2792 be_cmd_set_mac_list(adapter, NULL, 0, vf + 1);
2794 be_cmd_pmac_del(adapter, vf_cfg->if_handle,
2795 vf_cfg->pmac_id, vf + 1);
2797 be_cmd_if_destroy(adapter, vf_cfg->if_handle, vf + 1);
2800 kfree(adapter->vf_cfg);
2801 adapter->num_vfs = 0;
2804 static int be_clear(struct be_adapter *adapter)
2808 if (adapter->flags & BE_FLAGS_WORKER_SCHEDULED) {
2809 cancel_delayed_work_sync(&adapter->work);
2810 adapter->flags &= ~BE_FLAGS_WORKER_SCHEDULED;
2813 if (sriov_enabled(adapter))
2814 be_vf_clear(adapter);
2816 for (; adapter->uc_macs > 0; adapter->uc_macs--, i++)
2817 be_cmd_pmac_del(adapter, adapter->if_handle,
2818 adapter->pmac_id[i], 0);
2820 be_cmd_if_destroy(adapter, adapter->if_handle, 0);
2822 be_mcc_queues_destroy(adapter);
2823 be_rx_cqs_destroy(adapter);
2824 be_tx_queues_destroy(adapter);
2825 be_evt_queues_destroy(adapter);
2827 kfree(adapter->pmac_id);
2828 adapter->pmac_id = NULL;
2830 be_msix_disable(adapter);
2834 static int be_vfs_if_create(struct be_adapter *adapter)
2836 struct be_vf_cfg *vf_cfg;
2837 u32 cap_flags, en_flags, vf;
2840 cap_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
2841 BE_IF_FLAGS_MULTICAST;
2843 for_all_vfs(adapter, vf_cfg, vf) {
2844 if (!BE3_chip(adapter))
2845 be_cmd_get_profile_config(adapter, &cap_flags,
2848 /* If a FW profile exists, then cap_flags are updated */
2849 en_flags = cap_flags & (BE_IF_FLAGS_UNTAGGED |
2850 BE_IF_FLAGS_BROADCAST | BE_IF_FLAGS_MULTICAST);
2851 status = be_cmd_if_create(adapter, cap_flags, en_flags,
2852 &vf_cfg->if_handle, vf + 1);
2860 static int be_vf_setup_init(struct be_adapter *adapter)
2862 struct be_vf_cfg *vf_cfg;
2865 adapter->vf_cfg = kcalloc(adapter->num_vfs, sizeof(*vf_cfg),
2867 if (!adapter->vf_cfg)
2870 for_all_vfs(adapter, vf_cfg, vf) {
2871 vf_cfg->if_handle = -1;
2872 vf_cfg->pmac_id = -1;
2877 static int be_vf_setup(struct be_adapter *adapter)
2879 struct be_vf_cfg *vf_cfg;
2880 u16 def_vlan, lnk_speed;
2881 int status, old_vfs, vf;
2882 struct device *dev = &adapter->pdev->dev;
2884 old_vfs = pci_num_vf(adapter->pdev);
2886 dev_info(dev, "%d VFs are already enabled\n", old_vfs);
2887 if (old_vfs != num_vfs)
2888 dev_warn(dev, "Ignoring num_vfs=%d setting\n", num_vfs);
2889 adapter->num_vfs = old_vfs;
2891 if (num_vfs > adapter->dev_num_vfs)
2892 dev_info(dev, "Device supports %d VFs and not %d\n",
2893 adapter->dev_num_vfs, num_vfs);
2894 adapter->num_vfs = min_t(u16, num_vfs, adapter->dev_num_vfs);
2895 if (!adapter->num_vfs)
2899 status = be_vf_setup_init(adapter);
2904 for_all_vfs(adapter, vf_cfg, vf) {
2905 status = be_cmd_get_if_id(adapter, vf_cfg, vf);
2910 status = be_vfs_if_create(adapter);
2916 status = be_vfs_mac_query(adapter);
2920 status = be_vf_eth_addr_config(adapter);
2925 for_all_vfs(adapter, vf_cfg, vf) {
2926 /* BE3 FW, by default, caps VF TX-rate to 100mbps.
2927 * Allow full available bandwidth
2929 if (BE3_chip(adapter) && !old_vfs)
2930 be_cmd_set_qos(adapter, 1000, vf+1);
2932 status = be_cmd_link_status_query(adapter, &lnk_speed,
2935 vf_cfg->tx_rate = lnk_speed;
2937 status = be_cmd_get_hsw_config(adapter, &def_vlan,
2938 vf + 1, vf_cfg->if_handle);
2941 vf_cfg->def_vid = def_vlan;
2943 be_cmd_enable_vf(adapter, vf + 1);
2947 status = pci_enable_sriov(adapter->pdev, adapter->num_vfs);
2949 dev_err(dev, "SRIOV enable failed\n");
2950 adapter->num_vfs = 0;
2956 dev_err(dev, "VF setup failed\n");
2957 be_vf_clear(adapter);
2961 static void be_setup_init(struct be_adapter *adapter)
2963 adapter->vlan_prio_bmap = 0xff;
2964 adapter->phy.link_speed = -1;
2965 adapter->if_handle = -1;
2966 adapter->be3_native = false;
2967 adapter->promiscuous = false;
2968 if (be_physfn(adapter))
2969 adapter->cmd_privileges = MAX_PRIVILEGES;
2971 adapter->cmd_privileges = MIN_PRIVILEGES;
2974 static int be_get_mac_addr(struct be_adapter *adapter, u8 *mac, u32 if_handle,
2975 bool *active_mac, u32 *pmac_id)
2979 if (!is_zero_ether_addr(adapter->netdev->perm_addr)) {
2980 memcpy(mac, adapter->netdev->dev_addr, ETH_ALEN);
2981 if (!lancer_chip(adapter) && !be_physfn(adapter))
2984 *active_mac = false;
2989 if (lancer_chip(adapter)) {
2990 status = be_cmd_get_mac_from_list(adapter, mac,
2991 active_mac, pmac_id, 0);
2993 status = be_cmd_mac_addr_query(adapter, mac, false,
2994 if_handle, *pmac_id);
2996 } else if (be_physfn(adapter)) {
2997 /* For BE3, for PF get permanent MAC */
2998 status = be_cmd_mac_addr_query(adapter, mac, true, 0, 0);
2999 *active_mac = false;
3001 /* For BE3, for VF get soft MAC assigned by PF*/
3002 status = be_cmd_mac_addr_query(adapter, mac, false,
3009 static void be_get_resources(struct be_adapter *adapter)
3013 bool profile_present = false;
3016 if (!BEx_chip(adapter)) {
3017 status = be_cmd_get_func_config(adapter);
3019 profile_present = true;
3020 } else if (BE3_chip(adapter) && be_physfn(adapter)) {
3021 be_cmd_get_profile_config(adapter, NULL, &txq_count, 0);
3024 if (profile_present) {
3025 /* Sanity fixes for Lancer */
3026 adapter->max_pmac_cnt = min_t(u16, adapter->max_pmac_cnt,
3028 adapter->max_vlans = min_t(u16, adapter->max_vlans,
3029 BE_NUM_VLANS_SUPPORTED);
3030 adapter->max_mcast_mac = min_t(u16, adapter->max_mcast_mac,
3032 adapter->max_tx_queues = min_t(u16, adapter->max_tx_queues,
3034 adapter->max_rss_queues = min_t(u16, adapter->max_rss_queues,
3036 adapter->max_event_queues = min_t(u16,
3037 adapter->max_event_queues,
3040 if (adapter->max_rss_queues &&
3041 adapter->max_rss_queues == adapter->max_rx_queues)
3042 adapter->max_rss_queues -= 1;
3044 if (adapter->max_event_queues < adapter->max_rss_queues)
3045 adapter->max_rss_queues = adapter->max_event_queues;
3048 if (be_physfn(adapter))
3049 adapter->max_pmac_cnt = BE_UC_PMAC_COUNT;
3051 adapter->max_pmac_cnt = BE_VF_UC_PMAC_COUNT;
3053 if (adapter->function_mode & FLEX10_MODE)
3054 adapter->max_vlans = BE_NUM_VLANS_SUPPORTED/8;
3056 adapter->max_vlans = BE_NUM_VLANS_SUPPORTED;
3058 adapter->max_mcast_mac = BE_MAX_MC;
3059 adapter->max_tx_queues = txq_count ? txq_count : MAX_TX_QS;
3060 adapter->max_tx_queues = min_t(u16, adapter->max_tx_queues,
3062 adapter->max_rss_queues = (adapter->be3_native) ?
3063 BE3_MAX_RSS_QS : BE2_MAX_RSS_QS;
3064 adapter->max_event_queues = BE3_MAX_RSS_QS;
3066 adapter->if_cap_flags = BE_IF_FLAGS_UNTAGGED |
3067 BE_IF_FLAGS_BROADCAST |
3068 BE_IF_FLAGS_MULTICAST |
3069 BE_IF_FLAGS_PASS_L3L4_ERRORS |
3070 BE_IF_FLAGS_MCAST_PROMISCUOUS |
3071 BE_IF_FLAGS_VLAN_PROMISCUOUS |
3072 BE_IF_FLAGS_PROMISCUOUS;
3074 if (adapter->function_caps & BE_FUNCTION_CAPS_RSS)
3075 adapter->if_cap_flags |= BE_IF_FLAGS_RSS;
3078 pos = pci_find_ext_capability(adapter->pdev, PCI_EXT_CAP_ID_SRIOV);
3080 pci_read_config_word(adapter->pdev, pos + PCI_SRIOV_TOTAL_VF,
3082 if (BE3_chip(adapter))
3083 dev_num_vfs = min_t(u16, dev_num_vfs, MAX_VFS);
3084 adapter->dev_num_vfs = dev_num_vfs;
3088 /* Routine to query per function resource limits */
3089 static int be_get_config(struct be_adapter *adapter)
3093 status = be_cmd_query_fw_cfg(adapter, &adapter->port_num,
3094 &adapter->function_mode,
3095 &adapter->function_caps,
3096 &adapter->asic_rev);
3100 be_get_resources(adapter);
3102 /* primary mac needs 1 pmac entry */
3103 adapter->pmac_id = kcalloc(adapter->max_pmac_cnt + 1,
3104 sizeof(u32), GFP_KERNEL);
3105 if (!adapter->pmac_id) {
3114 static int be_setup(struct be_adapter *adapter)
3116 struct device *dev = &adapter->pdev->dev;
3123 be_setup_init(adapter);
3125 if (!lancer_chip(adapter))
3126 be_cmd_req_native_mode(adapter);
3128 status = be_get_config(adapter);
3132 status = be_msix_enable(adapter);
3136 status = be_evt_queues_create(adapter);
3140 status = be_tx_cqs_create(adapter);
3144 status = be_rx_cqs_create(adapter);
3148 status = be_mcc_queues_create(adapter);
3152 be_cmd_get_fn_privileges(adapter, &adapter->cmd_privileges, 0);
3153 /* In UMC mode FW does not return right privileges.
3154 * Override with correct privilege equivalent to PF.
3156 if (be_is_mc(adapter))
3157 adapter->cmd_privileges = MAX_PRIVILEGES;
3159 en_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3160 BE_IF_FLAGS_MULTICAST | BE_IF_FLAGS_PASS_L3L4_ERRORS;
3162 if (adapter->function_caps & BE_FUNCTION_CAPS_RSS)
3163 en_flags |= BE_IF_FLAGS_RSS;
3165 en_flags = en_flags & adapter->if_cap_flags;
3167 status = be_cmd_if_create(adapter, adapter->if_cap_flags, en_flags,
3168 &adapter->if_handle, 0);
3172 memset(mac, 0, ETH_ALEN);
3174 status = be_get_mac_addr(adapter, mac, adapter->if_handle,
3175 &active_mac, &adapter->pmac_id[0]);
3180 status = be_cmd_pmac_add(adapter, mac, adapter->if_handle,
3181 &adapter->pmac_id[0], 0);
3186 if (is_zero_ether_addr(adapter->netdev->dev_addr)) {
3187 memcpy(adapter->netdev->dev_addr, mac, ETH_ALEN);
3188 memcpy(adapter->netdev->perm_addr, mac, ETH_ALEN);
3191 status = be_tx_qs_create(adapter);
3195 be_cmd_get_fw_ver(adapter, adapter->fw_ver, adapter->fw_on_flash);
3197 if (adapter->vlans_added)
3198 be_vid_config(adapter);
3200 be_set_rx_mode(adapter->netdev);
3202 be_cmd_get_flow_control(adapter, &tx_fc, &rx_fc);
3204 if (rx_fc != adapter->rx_fc || tx_fc != adapter->tx_fc)
3205 be_cmd_set_flow_control(adapter, adapter->tx_fc,
3208 if (be_physfn(adapter)) {
3209 if (adapter->dev_num_vfs)
3210 be_vf_setup(adapter);
3212 dev_warn(dev, "device doesn't support SRIOV\n");
3215 status = be_cmd_get_phy_info(adapter);
3216 if (!status && be_pause_supported(adapter))
3217 adapter->phy.fc_autoneg = 1;
3219 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
3220 adapter->flags |= BE_FLAGS_WORKER_SCHEDULED;
3227 #ifdef CONFIG_NET_POLL_CONTROLLER
3228 static void be_netpoll(struct net_device *netdev)
3230 struct be_adapter *adapter = netdev_priv(netdev);
3231 struct be_eq_obj *eqo;
3234 for_all_evt_queues(adapter, eqo, i) {
3235 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
3236 napi_schedule(&eqo->napi);
3243 #define FW_FILE_HDR_SIGN "ServerEngines Corp. "
3244 char flash_cookie[2][16] = {"*** SE FLAS", "H DIRECTORY *** "};
3246 static bool be_flash_redboot(struct be_adapter *adapter,
3247 const u8 *p, u32 img_start, int image_size,
3254 crc_offset = hdr_size + img_start + image_size - 4;
3258 status = be_cmd_get_flash_crc(adapter, flashed_crc,
3261 dev_err(&adapter->pdev->dev,
3262 "could not get crc from flash, not flashing redboot\n");
3266 /*update redboot only if crc does not match*/
3267 if (!memcmp(flashed_crc, p, 4))
3273 static bool phy_flashing_required(struct be_adapter *adapter)
3275 return (adapter->phy.phy_type == TN_8022 &&
3276 adapter->phy.interface_type == PHY_TYPE_BASET_10GB);
3279 static bool is_comp_in_ufi(struct be_adapter *adapter,
3280 struct flash_section_info *fsec, int type)
3282 int i = 0, img_type = 0;
3283 struct flash_section_info_g2 *fsec_g2 = NULL;
3285 if (BE2_chip(adapter))
3286 fsec_g2 = (struct flash_section_info_g2 *)fsec;
3288 for (i = 0; i < MAX_FLASH_COMP; i++) {
3290 img_type = le32_to_cpu(fsec_g2->fsec_entry[i].type);
3292 img_type = le32_to_cpu(fsec->fsec_entry[i].type);
3294 if (img_type == type)
3301 struct flash_section_info *get_fsec_info(struct be_adapter *adapter,
3303 const struct firmware *fw)
3305 struct flash_section_info *fsec = NULL;
3306 const u8 *p = fw->data;
3309 while (p < (fw->data + fw->size)) {
3310 fsec = (struct flash_section_info *)p;
3311 if (!memcmp(flash_cookie, fsec->cookie, sizeof(flash_cookie)))
3318 static int be_flash(struct be_adapter *adapter, const u8 *img,
3319 struct be_dma_mem *flash_cmd, int optype, int img_size)
3321 u32 total_bytes = 0, flash_op, num_bytes = 0;
3323 struct be_cmd_write_flashrom *req = flash_cmd->va;
3325 total_bytes = img_size;
3326 while (total_bytes) {
3327 num_bytes = min_t(u32, 32*1024, total_bytes);
3329 total_bytes -= num_bytes;
3332 if (optype == OPTYPE_PHY_FW)
3333 flash_op = FLASHROM_OPER_PHY_FLASH;
3335 flash_op = FLASHROM_OPER_FLASH;
3337 if (optype == OPTYPE_PHY_FW)
3338 flash_op = FLASHROM_OPER_PHY_SAVE;
3340 flash_op = FLASHROM_OPER_SAVE;
3343 memcpy(req->data_buf, img, num_bytes);
3345 status = be_cmd_write_flashrom(adapter, flash_cmd, optype,
3346 flash_op, num_bytes);
3348 if (status == ILLEGAL_IOCTL_REQ &&
3349 optype == OPTYPE_PHY_FW)
3351 dev_err(&adapter->pdev->dev,
3352 "cmd to write to flash rom failed.\n");
3359 /* For BE2, BE3 and BE3-R */
3360 static int be_flash_BEx(struct be_adapter *adapter,
3361 const struct firmware *fw,
3362 struct be_dma_mem *flash_cmd,
3366 int status = 0, i, filehdr_size = 0;
3367 int img_hdrs_size = (num_of_images * sizeof(struct image_hdr));
3368 const u8 *p = fw->data;
3369 const struct flash_comp *pflashcomp;
3370 int num_comp, redboot;
3371 struct flash_section_info *fsec = NULL;
3373 struct flash_comp gen3_flash_types[] = {
3374 { FLASH_iSCSI_PRIMARY_IMAGE_START_g3, OPTYPE_ISCSI_ACTIVE,
3375 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_iSCSI},
3376 { FLASH_REDBOOT_START_g3, OPTYPE_REDBOOT,
3377 FLASH_REDBOOT_IMAGE_MAX_SIZE_g3, IMAGE_BOOT_CODE},
3378 { FLASH_iSCSI_BIOS_START_g3, OPTYPE_BIOS,
3379 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_ISCSI},
3380 { FLASH_PXE_BIOS_START_g3, OPTYPE_PXE_BIOS,
3381 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_PXE},
3382 { FLASH_FCoE_BIOS_START_g3, OPTYPE_FCOE_BIOS,
3383 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_FCoE},
3384 { FLASH_iSCSI_BACKUP_IMAGE_START_g3, OPTYPE_ISCSI_BACKUP,
3385 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_iSCSI},
3386 { FLASH_FCoE_PRIMARY_IMAGE_START_g3, OPTYPE_FCOE_FW_ACTIVE,
3387 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_FCoE},
3388 { FLASH_FCoE_BACKUP_IMAGE_START_g3, OPTYPE_FCOE_FW_BACKUP,
3389 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_FCoE},
3390 { FLASH_NCSI_START_g3, OPTYPE_NCSI_FW,
3391 FLASH_NCSI_IMAGE_MAX_SIZE_g3, IMAGE_NCSI},
3392 { FLASH_PHY_FW_START_g3, OPTYPE_PHY_FW,
3393 FLASH_PHY_FW_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_PHY}
3396 struct flash_comp gen2_flash_types[] = {
3397 { FLASH_iSCSI_PRIMARY_IMAGE_START_g2, OPTYPE_ISCSI_ACTIVE,
3398 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_iSCSI},
3399 { FLASH_REDBOOT_START_g2, OPTYPE_REDBOOT,
3400 FLASH_REDBOOT_IMAGE_MAX_SIZE_g2, IMAGE_BOOT_CODE},
3401 { FLASH_iSCSI_BIOS_START_g2, OPTYPE_BIOS,
3402 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_ISCSI},
3403 { FLASH_PXE_BIOS_START_g2, OPTYPE_PXE_BIOS,
3404 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_PXE},
3405 { FLASH_FCoE_BIOS_START_g2, OPTYPE_FCOE_BIOS,
3406 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_FCoE},
3407 { FLASH_iSCSI_BACKUP_IMAGE_START_g2, OPTYPE_ISCSI_BACKUP,
3408 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_iSCSI},
3409 { FLASH_FCoE_PRIMARY_IMAGE_START_g2, OPTYPE_FCOE_FW_ACTIVE,
3410 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_FCoE},
3411 { FLASH_FCoE_BACKUP_IMAGE_START_g2, OPTYPE_FCOE_FW_BACKUP,
3412 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_FCoE}
3415 if (BE3_chip(adapter)) {
3416 pflashcomp = gen3_flash_types;
3417 filehdr_size = sizeof(struct flash_file_hdr_g3);
3418 num_comp = ARRAY_SIZE(gen3_flash_types);
3420 pflashcomp = gen2_flash_types;
3421 filehdr_size = sizeof(struct flash_file_hdr_g2);
3422 num_comp = ARRAY_SIZE(gen2_flash_types);
3425 /* Get flash section info*/
3426 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
3428 dev_err(&adapter->pdev->dev,
3429 "Invalid Cookie. UFI corrupted ?\n");
3432 for (i = 0; i < num_comp; i++) {
3433 if (!is_comp_in_ufi(adapter, fsec, pflashcomp[i].img_type))
3436 if ((pflashcomp[i].optype == OPTYPE_NCSI_FW) &&
3437 memcmp(adapter->fw_ver, "3.102.148.0", 11) < 0)
3440 if (pflashcomp[i].optype == OPTYPE_PHY_FW &&
3441 !phy_flashing_required(adapter))
3444 if (pflashcomp[i].optype == OPTYPE_REDBOOT) {
3445 redboot = be_flash_redboot(adapter, fw->data,
3446 pflashcomp[i].offset, pflashcomp[i].size,
3447 filehdr_size + img_hdrs_size);
3453 p += filehdr_size + pflashcomp[i].offset + img_hdrs_size;
3454 if (p + pflashcomp[i].size > fw->data + fw->size)
3457 status = be_flash(adapter, p, flash_cmd, pflashcomp[i].optype,
3458 pflashcomp[i].size);
3460 dev_err(&adapter->pdev->dev,
3461 "Flashing section type %d failed.\n",
3462 pflashcomp[i].img_type);
3469 static int be_flash_skyhawk(struct be_adapter *adapter,
3470 const struct firmware *fw,
3471 struct be_dma_mem *flash_cmd, int num_of_images)
3473 int status = 0, i, filehdr_size = 0;
3474 int img_offset, img_size, img_optype, redboot;
3475 int img_hdrs_size = num_of_images * sizeof(struct image_hdr);
3476 const u8 *p = fw->data;
3477 struct flash_section_info *fsec = NULL;
3479 filehdr_size = sizeof(struct flash_file_hdr_g3);
3480 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
3482 dev_err(&adapter->pdev->dev,
3483 "Invalid Cookie. UFI corrupted ?\n");
3487 for (i = 0; i < le32_to_cpu(fsec->fsec_hdr.num_images); i++) {
3488 img_offset = le32_to_cpu(fsec->fsec_entry[i].offset);
3489 img_size = le32_to_cpu(fsec->fsec_entry[i].pad_size);
3491 switch (le32_to_cpu(fsec->fsec_entry[i].type)) {
3492 case IMAGE_FIRMWARE_iSCSI:
3493 img_optype = OPTYPE_ISCSI_ACTIVE;
3495 case IMAGE_BOOT_CODE:
3496 img_optype = OPTYPE_REDBOOT;
3498 case IMAGE_OPTION_ROM_ISCSI:
3499 img_optype = OPTYPE_BIOS;
3501 case IMAGE_OPTION_ROM_PXE:
3502 img_optype = OPTYPE_PXE_BIOS;
3504 case IMAGE_OPTION_ROM_FCoE:
3505 img_optype = OPTYPE_FCOE_BIOS;
3507 case IMAGE_FIRMWARE_BACKUP_iSCSI:
3508 img_optype = OPTYPE_ISCSI_BACKUP;
3511 img_optype = OPTYPE_NCSI_FW;
3517 if (img_optype == OPTYPE_REDBOOT) {
3518 redboot = be_flash_redboot(adapter, fw->data,
3519 img_offset, img_size,
3520 filehdr_size + img_hdrs_size);
3526 p += filehdr_size + img_offset + img_hdrs_size;
3527 if (p + img_size > fw->data + fw->size)
3530 status = be_flash(adapter, p, flash_cmd, img_optype, img_size);
3532 dev_err(&adapter->pdev->dev,
3533 "Flashing section type %d failed.\n",
3534 fsec->fsec_entry[i].type);
3541 static int lancer_fw_download(struct be_adapter *adapter,
3542 const struct firmware *fw)
3544 #define LANCER_FW_DOWNLOAD_CHUNK (32 * 1024)
3545 #define LANCER_FW_DOWNLOAD_LOCATION "/prg"
3546 struct be_dma_mem flash_cmd;
3547 const u8 *data_ptr = NULL;
3548 u8 *dest_image_ptr = NULL;
3549 size_t image_size = 0;
3551 u32 data_written = 0;
3557 if (!IS_ALIGNED(fw->size, sizeof(u32))) {
3558 dev_err(&adapter->pdev->dev,
3559 "FW Image not properly aligned. "
3560 "Length must be 4 byte aligned.\n");
3562 goto lancer_fw_exit;
3565 flash_cmd.size = sizeof(struct lancer_cmd_req_write_object)
3566 + LANCER_FW_DOWNLOAD_CHUNK;
3567 flash_cmd.va = dma_alloc_coherent(&adapter->pdev->dev, flash_cmd.size,
3568 &flash_cmd.dma, GFP_KERNEL);
3569 if (!flash_cmd.va) {
3571 goto lancer_fw_exit;
3574 dest_image_ptr = flash_cmd.va +
3575 sizeof(struct lancer_cmd_req_write_object);
3576 image_size = fw->size;
3577 data_ptr = fw->data;
3579 while (image_size) {
3580 chunk_size = min_t(u32, image_size, LANCER_FW_DOWNLOAD_CHUNK);
3582 /* Copy the image chunk content. */
3583 memcpy(dest_image_ptr, data_ptr, chunk_size);
3585 status = lancer_cmd_write_object(adapter, &flash_cmd,
3587 LANCER_FW_DOWNLOAD_LOCATION,
3588 &data_written, &change_status,
3593 offset += data_written;
3594 data_ptr += data_written;
3595 image_size -= data_written;
3599 /* Commit the FW written */
3600 status = lancer_cmd_write_object(adapter, &flash_cmd,
3602 LANCER_FW_DOWNLOAD_LOCATION,
3603 &data_written, &change_status,
3607 dma_free_coherent(&adapter->pdev->dev, flash_cmd.size, flash_cmd.va,
3610 dev_err(&adapter->pdev->dev,
3611 "Firmware load error. "
3612 "Status code: 0x%x Additional Status: 0x%x\n",
3613 status, add_status);
3614 goto lancer_fw_exit;
3617 if (change_status == LANCER_FW_RESET_NEEDED) {
3618 status = lancer_physdev_ctrl(adapter,
3619 PHYSDEV_CONTROL_FW_RESET_MASK);
3621 dev_err(&adapter->pdev->dev,
3622 "Adapter busy for FW reset.\n"
3623 "New FW will not be active.\n");
3624 goto lancer_fw_exit;
3626 } else if (change_status != LANCER_NO_RESET_NEEDED) {
3627 dev_err(&adapter->pdev->dev,
3628 "System reboot required for new FW"
3632 dev_info(&adapter->pdev->dev, "Firmware flashed successfully\n");
3639 #define UFI_TYPE3R 10
3641 static int be_get_ufi_type(struct be_adapter *adapter,
3642 struct flash_file_hdr_g3 *fhdr)
3645 goto be_get_ufi_exit;
3647 if (skyhawk_chip(adapter) && fhdr->build[0] == '4')
3649 else if (BE3_chip(adapter) && fhdr->build[0] == '3') {
3650 if (fhdr->asic_type_rev == 0x10)
3654 } else if (BE2_chip(adapter) && fhdr->build[0] == '2')
3658 dev_err(&adapter->pdev->dev,
3659 "UFI and Interface are not compatible for flashing\n");
3663 static int be_fw_download(struct be_adapter *adapter, const struct firmware* fw)
3665 struct flash_file_hdr_g3 *fhdr3;
3666 struct image_hdr *img_hdr_ptr = NULL;
3667 struct be_dma_mem flash_cmd;
3669 int status = 0, i = 0, num_imgs = 0, ufi_type = 0;
3671 flash_cmd.size = sizeof(struct be_cmd_write_flashrom);
3672 flash_cmd.va = dma_alloc_coherent(&adapter->pdev->dev, flash_cmd.size,
3673 &flash_cmd.dma, GFP_KERNEL);
3674 if (!flash_cmd.va) {
3680 fhdr3 = (struct flash_file_hdr_g3 *)p;
3682 ufi_type = be_get_ufi_type(adapter, fhdr3);
3684 num_imgs = le32_to_cpu(fhdr3->num_imgs);
3685 for (i = 0; i < num_imgs; i++) {
3686 img_hdr_ptr = (struct image_hdr *)(fw->data +
3687 (sizeof(struct flash_file_hdr_g3) +
3688 i * sizeof(struct image_hdr)));
3689 if (le32_to_cpu(img_hdr_ptr->imageid) == 1) {
3692 status = be_flash_skyhawk(adapter, fw,
3693 &flash_cmd, num_imgs);
3696 status = be_flash_BEx(adapter, fw, &flash_cmd,
3700 /* Do not flash this ufi on BE3-R cards */
3701 if (adapter->asic_rev < 0x10)
3702 status = be_flash_BEx(adapter, fw,
3707 dev_err(&adapter->pdev->dev,
3708 "Can't load BE3 UFI on BE3R\n");
3714 if (ufi_type == UFI_TYPE2)
3715 status = be_flash_BEx(adapter, fw, &flash_cmd, 0);
3716 else if (ufi_type == -1)
3719 dma_free_coherent(&adapter->pdev->dev, flash_cmd.size, flash_cmd.va,
3722 dev_err(&adapter->pdev->dev, "Firmware load error\n");
3726 dev_info(&adapter->pdev->dev, "Firmware flashed successfully\n");
3732 int be_load_fw(struct be_adapter *adapter, u8 *fw_file)
3734 const struct firmware *fw;
3737 if (!netif_running(adapter->netdev)) {
3738 dev_err(&adapter->pdev->dev,
3739 "Firmware load not allowed (interface is down)\n");
3743 status = request_firmware(&fw, fw_file, &adapter->pdev->dev);
3747 dev_info(&adapter->pdev->dev, "Flashing firmware file %s\n", fw_file);
3749 if (lancer_chip(adapter))
3750 status = lancer_fw_download(adapter, fw);
3752 status = be_fw_download(adapter, fw);
3755 be_cmd_get_fw_ver(adapter, adapter->fw_ver,
3756 adapter->fw_on_flash);
3759 release_firmware(fw);
3763 static const struct net_device_ops be_netdev_ops = {
3764 .ndo_open = be_open,
3765 .ndo_stop = be_close,
3766 .ndo_start_xmit = be_xmit,
3767 .ndo_set_rx_mode = be_set_rx_mode,
3768 .ndo_set_mac_address = be_mac_addr_set,
3769 .ndo_change_mtu = be_change_mtu,
3770 .ndo_get_stats64 = be_get_stats64,
3771 .ndo_validate_addr = eth_validate_addr,
3772 .ndo_vlan_rx_add_vid = be_vlan_add_vid,
3773 .ndo_vlan_rx_kill_vid = be_vlan_rem_vid,
3774 .ndo_set_vf_mac = be_set_vf_mac,
3775 .ndo_set_vf_vlan = be_set_vf_vlan,
3776 .ndo_set_vf_tx_rate = be_set_vf_tx_rate,
3777 .ndo_get_vf_config = be_get_vf_config,
3778 #ifdef CONFIG_NET_POLL_CONTROLLER
3779 .ndo_poll_controller = be_netpoll,
3783 static void be_netdev_init(struct net_device *netdev)
3785 struct be_adapter *adapter = netdev_priv(netdev);
3786 struct be_eq_obj *eqo;
3789 netdev->hw_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
3790 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM | NETIF_F_RXCSUM |
3791 NETIF_F_HW_VLAN_CTAG_TX;
3792 if (be_multi_rxq(adapter))
3793 netdev->hw_features |= NETIF_F_RXHASH;
3795 netdev->features |= netdev->hw_features |
3796 NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_FILTER;
3798 netdev->vlan_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
3799 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
3801 netdev->priv_flags |= IFF_UNICAST_FLT;
3803 netdev->flags |= IFF_MULTICAST;
3805 netif_set_gso_max_size(netdev, 65535 - ETH_HLEN);
3807 netdev->netdev_ops = &be_netdev_ops;
3809 SET_ETHTOOL_OPS(netdev, &be_ethtool_ops);
3811 for_all_evt_queues(adapter, eqo, i)
3812 netif_napi_add(netdev, &eqo->napi, be_poll, BE_NAPI_WEIGHT);
3815 static void be_unmap_pci_bars(struct be_adapter *adapter)
3818 pci_iounmap(adapter->pdev, adapter->csr);
3820 pci_iounmap(adapter->pdev, adapter->db);
3823 static int db_bar(struct be_adapter *adapter)
3825 if (lancer_chip(adapter) || !be_physfn(adapter))
3831 static int be_roce_map_pci_bars(struct be_adapter *adapter)
3833 if (skyhawk_chip(adapter)) {
3834 adapter->roce_db.size = 4096;
3835 adapter->roce_db.io_addr = pci_resource_start(adapter->pdev,
3837 adapter->roce_db.total_size = pci_resource_len(adapter->pdev,
3843 static int be_map_pci_bars(struct be_adapter *adapter)
3848 pci_read_config_dword(adapter->pdev, SLI_INTF_REG_OFFSET, &sli_intf);
3849 adapter->if_type = (sli_intf & SLI_INTF_IF_TYPE_MASK) >>
3850 SLI_INTF_IF_TYPE_SHIFT;
3852 if (BEx_chip(adapter) && be_physfn(adapter)) {
3853 adapter->csr = pci_iomap(adapter->pdev, 2, 0);
3854 if (adapter->csr == NULL)
3858 addr = pci_iomap(adapter->pdev, db_bar(adapter), 0);
3863 be_roce_map_pci_bars(adapter);
3867 be_unmap_pci_bars(adapter);
3871 static void be_ctrl_cleanup(struct be_adapter *adapter)
3873 struct be_dma_mem *mem = &adapter->mbox_mem_alloced;
3875 be_unmap_pci_bars(adapter);
3878 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
3881 mem = &adapter->rx_filter;
3883 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
3887 static int be_ctrl_init(struct be_adapter *adapter)
3889 struct be_dma_mem *mbox_mem_alloc = &adapter->mbox_mem_alloced;
3890 struct be_dma_mem *mbox_mem_align = &adapter->mbox_mem;
3891 struct be_dma_mem *rx_filter = &adapter->rx_filter;
3895 pci_read_config_dword(adapter->pdev, SLI_INTF_REG_OFFSET, &sli_intf);
3896 adapter->sli_family = (sli_intf & SLI_INTF_FAMILY_MASK) >>
3897 SLI_INTF_FAMILY_SHIFT;
3898 adapter->virtfn = (sli_intf & SLI_INTF_FT_MASK) ? 1 : 0;
3900 status = be_map_pci_bars(adapter);
3904 mbox_mem_alloc->size = sizeof(struct be_mcc_mailbox) + 16;
3905 mbox_mem_alloc->va = dma_alloc_coherent(&adapter->pdev->dev,
3906 mbox_mem_alloc->size,
3907 &mbox_mem_alloc->dma,
3909 if (!mbox_mem_alloc->va) {
3911 goto unmap_pci_bars;
3913 mbox_mem_align->size = sizeof(struct be_mcc_mailbox);
3914 mbox_mem_align->va = PTR_ALIGN(mbox_mem_alloc->va, 16);
3915 mbox_mem_align->dma = PTR_ALIGN(mbox_mem_alloc->dma, 16);
3916 memset(mbox_mem_align->va, 0, sizeof(struct be_mcc_mailbox));
3918 rx_filter->size = sizeof(struct be_cmd_req_rx_filter);
3919 rx_filter->va = dma_alloc_coherent(&adapter->pdev->dev, rx_filter->size,
3921 GFP_KERNEL | __GFP_ZERO);
3922 if (rx_filter->va == NULL) {
3927 mutex_init(&adapter->mbox_lock);
3928 spin_lock_init(&adapter->mcc_lock);
3929 spin_lock_init(&adapter->mcc_cq_lock);
3931 init_completion(&adapter->flash_compl);
3932 pci_save_state(adapter->pdev);
3936 dma_free_coherent(&adapter->pdev->dev, mbox_mem_alloc->size,
3937 mbox_mem_alloc->va, mbox_mem_alloc->dma);
3940 be_unmap_pci_bars(adapter);
3946 static void be_stats_cleanup(struct be_adapter *adapter)
3948 struct be_dma_mem *cmd = &adapter->stats_cmd;
3951 dma_free_coherent(&adapter->pdev->dev, cmd->size,
3955 static int be_stats_init(struct be_adapter *adapter)
3957 struct be_dma_mem *cmd = &adapter->stats_cmd;
3959 if (lancer_chip(adapter))
3960 cmd->size = sizeof(struct lancer_cmd_req_pport_stats);
3961 else if (BE2_chip(adapter))
3962 cmd->size = sizeof(struct be_cmd_req_get_stats_v0);
3964 /* BE3 and Skyhawk */
3965 cmd->size = sizeof(struct be_cmd_req_get_stats_v1);
3967 cmd->va = dma_alloc_coherent(&adapter->pdev->dev, cmd->size, &cmd->dma,
3968 GFP_KERNEL | __GFP_ZERO);
3969 if (cmd->va == NULL)
3974 static void be_remove(struct pci_dev *pdev)
3976 struct be_adapter *adapter = pci_get_drvdata(pdev);
3981 be_roce_dev_remove(adapter);
3982 be_intr_set(adapter, false);
3984 cancel_delayed_work_sync(&adapter->func_recovery_work);
3986 unregister_netdev(adapter->netdev);
3990 /* tell fw we're done with firing cmds */
3991 be_cmd_fw_clean(adapter);
3993 be_stats_cleanup(adapter);
3995 be_ctrl_cleanup(adapter);
3997 pci_disable_pcie_error_reporting(pdev);
3999 pci_set_drvdata(pdev, NULL);
4000 pci_release_regions(pdev);
4001 pci_disable_device(pdev);
4003 free_netdev(adapter->netdev);
4006 bool be_is_wol_supported(struct be_adapter *adapter)
4008 return ((adapter->wol_cap & BE_WOL_CAP) &&
4009 !be_is_wol_excluded(adapter)) ? true : false;
4012 u32 be_get_fw_log_level(struct be_adapter *adapter)
4014 struct be_dma_mem extfat_cmd;
4015 struct be_fat_conf_params *cfgs;
4020 if (lancer_chip(adapter))
4023 memset(&extfat_cmd, 0, sizeof(struct be_dma_mem));
4024 extfat_cmd.size = sizeof(struct be_cmd_resp_get_ext_fat_caps);
4025 extfat_cmd.va = pci_alloc_consistent(adapter->pdev, extfat_cmd.size,
4028 if (!extfat_cmd.va) {
4029 dev_err(&adapter->pdev->dev, "%s: Memory allocation failure\n",
4034 status = be_cmd_get_ext_fat_capabilites(adapter, &extfat_cmd);
4036 cfgs = (struct be_fat_conf_params *)(extfat_cmd.va +
4037 sizeof(struct be_cmd_resp_hdr));
4038 for (j = 0; j < le32_to_cpu(cfgs->module[0].num_modes); j++) {
4039 if (cfgs->module[0].trace_lvl[j].mode == MODE_UART)
4040 level = cfgs->module[0].trace_lvl[j].dbg_lvl;
4043 pci_free_consistent(adapter->pdev, extfat_cmd.size, extfat_cmd.va,
4049 static int be_get_initial_config(struct be_adapter *adapter)
4054 status = be_cmd_get_cntl_attributes(adapter);
4058 status = be_cmd_get_acpi_wol_cap(adapter);
4060 /* in case of a failure to get wol capabillities
4061 * check the exclusion list to determine WOL capability */
4062 if (!be_is_wol_excluded(adapter))
4063 adapter->wol_cap |= BE_WOL_CAP;
4066 if (be_is_wol_supported(adapter))
4067 adapter->wol = true;
4069 /* Must be a power of 2 or else MODULO will BUG_ON */
4070 adapter->be_get_temp_freq = 64;
4072 level = be_get_fw_log_level(adapter);
4073 adapter->msg_enable = level <= FW_LOG_LEVEL_DEFAULT ? NETIF_MSG_HW : 0;
4078 static int lancer_recover_func(struct be_adapter *adapter)
4080 struct device *dev = &adapter->pdev->dev;
4083 status = lancer_test_and_set_rdy_state(adapter);
4087 if (netif_running(adapter->netdev))
4088 be_close(adapter->netdev);
4092 be_clear_all_error(adapter);
4094 status = be_setup(adapter);
4098 if (netif_running(adapter->netdev)) {
4099 status = be_open(adapter->netdev);
4104 dev_err(dev, "Error recovery successful\n");
4107 if (status == -EAGAIN)
4108 dev_err(dev, "Waiting for resource provisioning\n");
4110 dev_err(dev, "Error recovery failed\n");
4115 static void be_func_recovery_task(struct work_struct *work)
4117 struct be_adapter *adapter =
4118 container_of(work, struct be_adapter, func_recovery_work.work);
4121 be_detect_error(adapter);
4123 if (adapter->hw_error && lancer_chip(adapter)) {
4126 netif_device_detach(adapter->netdev);
4129 status = lancer_recover_func(adapter);
4131 netif_device_attach(adapter->netdev);
4134 /* In Lancer, for all errors other than provisioning error (-EAGAIN),
4135 * no need to attempt further recovery.
4137 if (!status || status == -EAGAIN)
4138 schedule_delayed_work(&adapter->func_recovery_work,
4139 msecs_to_jiffies(1000));
4142 static void be_worker(struct work_struct *work)
4144 struct be_adapter *adapter =
4145 container_of(work, struct be_adapter, work.work);
4146 struct be_rx_obj *rxo;
4147 struct be_eq_obj *eqo;
4150 /* when interrupts are not yet enabled, just reap any pending
4151 * mcc completions */
4152 if (!netif_running(adapter->netdev)) {
4154 be_process_mcc(adapter);
4159 if (!adapter->stats_cmd_sent) {
4160 if (lancer_chip(adapter))
4161 lancer_cmd_get_pport_stats(adapter,
4162 &adapter->stats_cmd);
4164 be_cmd_get_stats(adapter, &adapter->stats_cmd);
4167 if (MODULO(adapter->work_counter, adapter->be_get_temp_freq) == 0)
4168 be_cmd_get_die_temperature(adapter);
4170 for_all_rx_queues(adapter, rxo, i) {
4171 if (rxo->rx_post_starved) {
4172 rxo->rx_post_starved = false;
4173 be_post_rx_frags(rxo, GFP_KERNEL);
4177 for_all_evt_queues(adapter, eqo, i)
4178 be_eqd_update(adapter, eqo);
4181 adapter->work_counter++;
4182 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
4185 /* If any VFs are already enabled don't FLR the PF */
4186 static bool be_reset_required(struct be_adapter *adapter)
4188 return pci_num_vf(adapter->pdev) ? false : true;
4191 static char *mc_name(struct be_adapter *adapter)
4193 if (adapter->function_mode & FLEX10_MODE)
4195 else if (adapter->function_mode & VNIC_MODE)
4197 else if (adapter->function_mode & UMC_ENABLED)
4203 static inline char *func_name(struct be_adapter *adapter)
4205 return be_physfn(adapter) ? "PF" : "VF";
4208 static int be_probe(struct pci_dev *pdev, const struct pci_device_id *pdev_id)
4211 struct be_adapter *adapter;
4212 struct net_device *netdev;
4215 status = pci_enable_device(pdev);
4219 status = pci_request_regions(pdev, DRV_NAME);
4222 pci_set_master(pdev);
4224 netdev = alloc_etherdev_mqs(sizeof(*adapter), MAX_TX_QS, MAX_RX_QS);
4225 if (netdev == NULL) {
4229 adapter = netdev_priv(netdev);
4230 adapter->pdev = pdev;
4231 pci_set_drvdata(pdev, adapter);
4232 adapter->netdev = netdev;
4233 SET_NETDEV_DEV(netdev, &pdev->dev);
4235 status = dma_set_mask(&pdev->dev, DMA_BIT_MASK(64));
4237 status = dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(64));
4239 dev_err(&pdev->dev, "dma_set_coherent_mask failed\n");
4242 netdev->features |= NETIF_F_HIGHDMA;
4244 status = dma_set_mask(&pdev->dev, DMA_BIT_MASK(32));
4246 status = dma_set_coherent_mask(&pdev->dev,
4249 dev_err(&pdev->dev, "Could not set PCI DMA Mask\n");
4254 status = pci_enable_pcie_error_reporting(pdev);
4256 dev_err(&pdev->dev, "Could not use PCIe error reporting\n");
4258 status = be_ctrl_init(adapter);
4262 /* sync up with fw's ready state */
4263 if (be_physfn(adapter)) {
4264 status = be_fw_wait_ready(adapter);
4269 if (be_reset_required(adapter)) {
4270 status = be_cmd_reset_function(adapter);
4274 /* Wait for interrupts to quiesce after an FLR */
4278 /* Allow interrupts for other ULPs running on NIC function */
4279 be_intr_set(adapter, true);
4281 /* tell fw we're ready to fire cmds */
4282 status = be_cmd_fw_init(adapter);
4286 status = be_stats_init(adapter);
4290 status = be_get_initial_config(adapter);
4294 INIT_DELAYED_WORK(&adapter->work, be_worker);
4295 INIT_DELAYED_WORK(&adapter->func_recovery_work, be_func_recovery_task);
4296 adapter->rx_fc = adapter->tx_fc = true;
4298 status = be_setup(adapter);
4302 be_netdev_init(netdev);
4303 status = register_netdev(netdev);
4307 be_roce_dev_add(adapter);
4309 schedule_delayed_work(&adapter->func_recovery_work,
4310 msecs_to_jiffies(1000));
4312 be_cmd_query_port_name(adapter, &port_name);
4314 dev_info(&pdev->dev, "%s: %s %s port %c\n", nic_name(pdev),
4315 func_name(adapter), mc_name(adapter), port_name);
4322 be_stats_cleanup(adapter);
4324 be_ctrl_cleanup(adapter);
4326 free_netdev(netdev);
4327 pci_set_drvdata(pdev, NULL);
4329 pci_release_regions(pdev);
4331 pci_disable_device(pdev);
4333 dev_err(&pdev->dev, "%s initialization failed\n", nic_name(pdev));
4337 static int be_suspend(struct pci_dev *pdev, pm_message_t state)
4339 struct be_adapter *adapter = pci_get_drvdata(pdev);
4340 struct net_device *netdev = adapter->netdev;
4343 be_setup_wol(adapter, true);
4345 cancel_delayed_work_sync(&adapter->func_recovery_work);
4347 netif_device_detach(netdev);
4348 if (netif_running(netdev)) {
4355 pci_save_state(pdev);
4356 pci_disable_device(pdev);
4357 pci_set_power_state(pdev, pci_choose_state(pdev, state));
4361 static int be_resume(struct pci_dev *pdev)
4364 struct be_adapter *adapter = pci_get_drvdata(pdev);
4365 struct net_device *netdev = adapter->netdev;
4367 netif_device_detach(netdev);
4369 status = pci_enable_device(pdev);
4373 pci_set_power_state(pdev, PCI_D0);
4374 pci_restore_state(pdev);
4376 status = be_fw_wait_ready(adapter);
4380 /* tell fw we're ready to fire cmds */
4381 status = be_cmd_fw_init(adapter);
4386 if (netif_running(netdev)) {
4392 schedule_delayed_work(&adapter->func_recovery_work,
4393 msecs_to_jiffies(1000));
4394 netif_device_attach(netdev);
4397 be_setup_wol(adapter, false);
4403 * An FLR will stop BE from DMAing any data.
4405 static void be_shutdown(struct pci_dev *pdev)
4407 struct be_adapter *adapter = pci_get_drvdata(pdev);
4412 cancel_delayed_work_sync(&adapter->work);
4413 cancel_delayed_work_sync(&adapter->func_recovery_work);
4415 netif_device_detach(adapter->netdev);
4417 be_cmd_reset_function(adapter);
4419 pci_disable_device(pdev);
4422 static pci_ers_result_t be_eeh_err_detected(struct pci_dev *pdev,
4423 pci_channel_state_t state)
4425 struct be_adapter *adapter = pci_get_drvdata(pdev);
4426 struct net_device *netdev = adapter->netdev;
4428 dev_err(&adapter->pdev->dev, "EEH error detected\n");
4430 if (!adapter->eeh_error) {
4431 adapter->eeh_error = true;
4433 cancel_delayed_work_sync(&adapter->func_recovery_work);
4436 netif_device_detach(netdev);
4437 if (netif_running(netdev))
4444 if (state == pci_channel_io_perm_failure)
4445 return PCI_ERS_RESULT_DISCONNECT;
4447 pci_disable_device(pdev);
4449 /* The error could cause the FW to trigger a flash debug dump.
4450 * Resetting the card while flash dump is in progress
4451 * can cause it not to recover; wait for it to finish.
4452 * Wait only for first function as it is needed only once per
4455 if (pdev->devfn == 0)
4458 return PCI_ERS_RESULT_NEED_RESET;
4461 static pci_ers_result_t be_eeh_reset(struct pci_dev *pdev)
4463 struct be_adapter *adapter = pci_get_drvdata(pdev);
4466 dev_info(&adapter->pdev->dev, "EEH reset\n");
4468 status = pci_enable_device(pdev);
4470 return PCI_ERS_RESULT_DISCONNECT;
4472 pci_set_master(pdev);
4473 pci_set_power_state(pdev, PCI_D0);
4474 pci_restore_state(pdev);
4476 /* Check if card is ok and fw is ready */
4477 dev_info(&adapter->pdev->dev,
4478 "Waiting for FW to be ready after EEH reset\n");
4479 status = be_fw_wait_ready(adapter);
4481 return PCI_ERS_RESULT_DISCONNECT;
4483 pci_cleanup_aer_uncorrect_error_status(pdev);
4484 be_clear_all_error(adapter);
4485 return PCI_ERS_RESULT_RECOVERED;
4488 static void be_eeh_resume(struct pci_dev *pdev)
4491 struct be_adapter *adapter = pci_get_drvdata(pdev);
4492 struct net_device *netdev = adapter->netdev;
4494 dev_info(&adapter->pdev->dev, "EEH resume\n");
4496 pci_save_state(pdev);
4498 status = be_cmd_reset_function(adapter);
4502 /* tell fw we're ready to fire cmds */
4503 status = be_cmd_fw_init(adapter);
4507 status = be_setup(adapter);
4511 if (netif_running(netdev)) {
4512 status = be_open(netdev);
4517 schedule_delayed_work(&adapter->func_recovery_work,
4518 msecs_to_jiffies(1000));
4519 netif_device_attach(netdev);
4522 dev_err(&adapter->pdev->dev, "EEH resume failed\n");
4525 static const struct pci_error_handlers be_eeh_handlers = {
4526 .error_detected = be_eeh_err_detected,
4527 .slot_reset = be_eeh_reset,
4528 .resume = be_eeh_resume,
4531 static struct pci_driver be_driver = {
4533 .id_table = be_dev_ids,
4535 .remove = be_remove,
4536 .suspend = be_suspend,
4537 .resume = be_resume,
4538 .shutdown = be_shutdown,
4539 .err_handler = &be_eeh_handlers
4542 static int __init be_init_module(void)
4544 if (rx_frag_size != 8192 && rx_frag_size != 4096 &&
4545 rx_frag_size != 2048) {
4546 printk(KERN_WARNING DRV_NAME
4547 " : Module param rx_frag_size must be 2048/4096/8192."
4549 rx_frag_size = 2048;
4552 return pci_register_driver(&be_driver);
4554 module_init(be_init_module);
4556 static void __exit be_exit_module(void)
4558 pci_unregister_driver(&be_driver);
4560 module_exit(be_exit_module);