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 device *dev = &adapter->pdev->dev;
251 struct sockaddr *addr = p;
254 u32 old_pmac_id = adapter->pmac_id[0], curr_pmac_id = 0;
256 if (!is_valid_ether_addr(addr->sa_data))
257 return -EADDRNOTAVAIL;
259 /* The PMAC_ADD cmd may fail if the VF doesn't have FILTMGMT
260 * privilege or if PF did not provision the new MAC address.
261 * On BE3, this cmd will always fail if the VF doesn't have the
262 * FILTMGMT privilege. This failure is OK, only if the PF programmed
263 * the MAC for the VF.
265 status = be_cmd_pmac_add(adapter, (u8 *)addr->sa_data,
266 adapter->if_handle, &adapter->pmac_id[0], 0);
268 curr_pmac_id = adapter->pmac_id[0];
270 /* Delete the old programmed MAC. This call may fail if the
271 * old MAC was already deleted by the PF driver.
273 if (adapter->pmac_id[0] != old_pmac_id)
274 be_cmd_pmac_del(adapter, adapter->if_handle,
278 /* Decide if the new MAC is successfully activated only after
281 status = be_cmd_get_active_mac(adapter, curr_pmac_id, mac);
285 /* The MAC change did not happen, either due to lack of privilege
286 * or PF didn't pre-provision.
288 if (memcmp(addr->sa_data, mac, ETH_ALEN)) {
293 memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
294 dev_info(dev, "MAC address changed to %pM\n", mac);
297 dev_warn(dev, "MAC address change to %pM 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 static 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];
1150 if (!sriov_enabled(adapter))
1153 if (!is_valid_ether_addr(mac) || vf >= adapter->num_vfs)
1156 if (BEx_chip(adapter)) {
1157 be_cmd_pmac_del(adapter, vf_cfg->if_handle, vf_cfg->pmac_id,
1160 status = be_cmd_pmac_add(adapter, mac, vf_cfg->if_handle,
1161 &vf_cfg->pmac_id, vf + 1);
1163 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
1168 dev_err(&adapter->pdev->dev, "MAC %pM set on VF %d Failed\n",
1171 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
1176 static int be_get_vf_config(struct net_device *netdev, int vf,
1177 struct ifla_vf_info *vi)
1179 struct be_adapter *adapter = netdev_priv(netdev);
1180 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1182 if (!sriov_enabled(adapter))
1185 if (vf >= adapter->num_vfs)
1189 vi->tx_rate = vf_cfg->tx_rate;
1190 vi->vlan = vf_cfg->vlan_tag;
1192 memcpy(&vi->mac, vf_cfg->mac_addr, ETH_ALEN);
1197 static int be_set_vf_vlan(struct net_device *netdev,
1198 int vf, u16 vlan, u8 qos)
1200 struct be_adapter *adapter = netdev_priv(netdev);
1203 if (!sriov_enabled(adapter))
1206 if (vf >= adapter->num_vfs || vlan > 4095)
1210 if (adapter->vf_cfg[vf].vlan_tag != vlan) {
1211 /* If this is new value, program it. Else skip. */
1212 adapter->vf_cfg[vf].vlan_tag = vlan;
1214 status = be_cmd_set_hsw_config(adapter, vlan,
1215 vf + 1, adapter->vf_cfg[vf].if_handle);
1218 /* Reset Transparent Vlan Tagging. */
1219 adapter->vf_cfg[vf].vlan_tag = 0;
1220 vlan = adapter->vf_cfg[vf].def_vid;
1221 status = be_cmd_set_hsw_config(adapter, vlan, vf + 1,
1222 adapter->vf_cfg[vf].if_handle);
1227 dev_info(&adapter->pdev->dev,
1228 "VLAN %d config on VF %d failed\n", vlan, vf);
1232 static int be_set_vf_tx_rate(struct net_device *netdev,
1235 struct be_adapter *adapter = netdev_priv(netdev);
1238 if (!sriov_enabled(adapter))
1241 if (vf >= adapter->num_vfs)
1244 if (rate < 100 || rate > 10000) {
1245 dev_err(&adapter->pdev->dev,
1246 "tx rate must be between 100 and 10000 Mbps\n");
1250 if (lancer_chip(adapter))
1251 status = be_cmd_set_profile_config(adapter, rate / 10, vf + 1);
1253 status = be_cmd_set_qos(adapter, rate / 10, vf + 1);
1256 dev_err(&adapter->pdev->dev,
1257 "tx rate %d on VF %d failed\n", rate, vf);
1259 adapter->vf_cfg[vf].tx_rate = rate;
1263 static void be_eqd_update(struct be_adapter *adapter, struct be_eq_obj *eqo)
1265 struct be_rx_stats *stats = rx_stats(&adapter->rx_obj[eqo->idx]);
1266 ulong now = jiffies;
1267 ulong delta = now - stats->rx_jiffies;
1269 unsigned int start, eqd;
1271 if (!eqo->enable_aic) {
1276 if (eqo->idx >= adapter->num_rx_qs)
1279 stats = rx_stats(&adapter->rx_obj[eqo->idx]);
1281 /* Wrapped around */
1282 if (time_before(now, stats->rx_jiffies)) {
1283 stats->rx_jiffies = now;
1287 /* Update once a second */
1292 start = u64_stats_fetch_begin_bh(&stats->sync);
1293 pkts = stats->rx_pkts;
1294 } while (u64_stats_fetch_retry_bh(&stats->sync, start));
1296 stats->rx_pps = (unsigned long)(pkts - stats->rx_pkts_prev) / (delta / HZ);
1297 stats->rx_pkts_prev = pkts;
1298 stats->rx_jiffies = now;
1299 eqd = (stats->rx_pps / 110000) << 3;
1300 eqd = min(eqd, eqo->max_eqd);
1301 eqd = max(eqd, eqo->min_eqd);
1306 if (eqd != eqo->cur_eqd) {
1307 be_cmd_modify_eqd(adapter, eqo->q.id, eqd);
1312 static void be_rx_stats_update(struct be_rx_obj *rxo,
1313 struct be_rx_compl_info *rxcp)
1315 struct be_rx_stats *stats = rx_stats(rxo);
1317 u64_stats_update_begin(&stats->sync);
1319 stats->rx_bytes += rxcp->pkt_size;
1321 if (rxcp->pkt_type == BE_MULTICAST_PACKET)
1322 stats->rx_mcast_pkts++;
1324 stats->rx_compl_err++;
1325 u64_stats_update_end(&stats->sync);
1328 static inline bool csum_passed(struct be_rx_compl_info *rxcp)
1330 /* L4 checksum is not reliable for non TCP/UDP packets.
1331 * Also ignore ipcksm for ipv6 pkts */
1332 return (rxcp->tcpf || rxcp->udpf) && rxcp->l4_csum &&
1333 (rxcp->ip_csum || rxcp->ipv6);
1336 static struct be_rx_page_info *get_rx_page_info(struct be_rx_obj *rxo,
1339 struct be_adapter *adapter = rxo->adapter;
1340 struct be_rx_page_info *rx_page_info;
1341 struct be_queue_info *rxq = &rxo->q;
1343 rx_page_info = &rxo->page_info_tbl[frag_idx];
1344 BUG_ON(!rx_page_info->page);
1346 if (rx_page_info->last_page_user) {
1347 dma_unmap_page(&adapter->pdev->dev,
1348 dma_unmap_addr(rx_page_info, bus),
1349 adapter->big_page_size, DMA_FROM_DEVICE);
1350 rx_page_info->last_page_user = false;
1353 atomic_dec(&rxq->used);
1354 return rx_page_info;
1357 /* Throwaway the data in the Rx completion */
1358 static void be_rx_compl_discard(struct be_rx_obj *rxo,
1359 struct be_rx_compl_info *rxcp)
1361 struct be_queue_info *rxq = &rxo->q;
1362 struct be_rx_page_info *page_info;
1363 u16 i, num_rcvd = rxcp->num_rcvd;
1365 for (i = 0; i < num_rcvd; i++) {
1366 page_info = get_rx_page_info(rxo, rxcp->rxq_idx);
1367 put_page(page_info->page);
1368 memset(page_info, 0, sizeof(*page_info));
1369 index_inc(&rxcp->rxq_idx, rxq->len);
1374 * skb_fill_rx_data forms a complete skb for an ether frame
1375 * indicated by rxcp.
1377 static void skb_fill_rx_data(struct be_rx_obj *rxo, struct sk_buff *skb,
1378 struct be_rx_compl_info *rxcp)
1380 struct be_queue_info *rxq = &rxo->q;
1381 struct be_rx_page_info *page_info;
1383 u16 hdr_len, curr_frag_len, remaining;
1386 page_info = get_rx_page_info(rxo, rxcp->rxq_idx);
1387 start = page_address(page_info->page) + page_info->page_offset;
1390 /* Copy data in the first descriptor of this completion */
1391 curr_frag_len = min(rxcp->pkt_size, rx_frag_size);
1393 skb->len = curr_frag_len;
1394 if (curr_frag_len <= BE_HDR_LEN) { /* tiny packet */
1395 memcpy(skb->data, start, curr_frag_len);
1396 /* Complete packet has now been moved to data */
1397 put_page(page_info->page);
1399 skb->tail += curr_frag_len;
1402 memcpy(skb->data, start, hdr_len);
1403 skb_shinfo(skb)->nr_frags = 1;
1404 skb_frag_set_page(skb, 0, page_info->page);
1405 skb_shinfo(skb)->frags[0].page_offset =
1406 page_info->page_offset + hdr_len;
1407 skb_frag_size_set(&skb_shinfo(skb)->frags[0], curr_frag_len - hdr_len);
1408 skb->data_len = curr_frag_len - hdr_len;
1409 skb->truesize += rx_frag_size;
1410 skb->tail += hdr_len;
1412 page_info->page = NULL;
1414 if (rxcp->pkt_size <= rx_frag_size) {
1415 BUG_ON(rxcp->num_rcvd != 1);
1419 /* More frags present for this completion */
1420 index_inc(&rxcp->rxq_idx, rxq->len);
1421 remaining = rxcp->pkt_size - curr_frag_len;
1422 for (i = 1, j = 0; i < rxcp->num_rcvd; i++) {
1423 page_info = get_rx_page_info(rxo, rxcp->rxq_idx);
1424 curr_frag_len = min(remaining, rx_frag_size);
1426 /* Coalesce all frags from the same physical page in one slot */
1427 if (page_info->page_offset == 0) {
1430 skb_frag_set_page(skb, j, page_info->page);
1431 skb_shinfo(skb)->frags[j].page_offset =
1432 page_info->page_offset;
1433 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1434 skb_shinfo(skb)->nr_frags++;
1436 put_page(page_info->page);
1439 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1440 skb->len += curr_frag_len;
1441 skb->data_len += curr_frag_len;
1442 skb->truesize += rx_frag_size;
1443 remaining -= curr_frag_len;
1444 index_inc(&rxcp->rxq_idx, rxq->len);
1445 page_info->page = NULL;
1447 BUG_ON(j > MAX_SKB_FRAGS);
1450 /* Process the RX completion indicated by rxcp when GRO is disabled */
1451 static void be_rx_compl_process(struct be_rx_obj *rxo,
1452 struct be_rx_compl_info *rxcp)
1454 struct be_adapter *adapter = rxo->adapter;
1455 struct net_device *netdev = adapter->netdev;
1456 struct sk_buff *skb;
1458 skb = netdev_alloc_skb_ip_align(netdev, BE_RX_SKB_ALLOC_SIZE);
1459 if (unlikely(!skb)) {
1460 rx_stats(rxo)->rx_drops_no_skbs++;
1461 be_rx_compl_discard(rxo, rxcp);
1465 skb_fill_rx_data(rxo, skb, rxcp);
1467 if (likely((netdev->features & NETIF_F_RXCSUM) && csum_passed(rxcp)))
1468 skb->ip_summed = CHECKSUM_UNNECESSARY;
1470 skb_checksum_none_assert(skb);
1472 skb->protocol = eth_type_trans(skb, netdev);
1473 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1474 if (netdev->features & NETIF_F_RXHASH)
1475 skb->rxhash = rxcp->rss_hash;
1479 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1481 netif_receive_skb(skb);
1484 /* Process the RX completion indicated by rxcp when GRO is enabled */
1485 static void be_rx_compl_process_gro(struct be_rx_obj *rxo,
1486 struct napi_struct *napi,
1487 struct be_rx_compl_info *rxcp)
1489 struct be_adapter *adapter = rxo->adapter;
1490 struct be_rx_page_info *page_info;
1491 struct sk_buff *skb = NULL;
1492 struct be_queue_info *rxq = &rxo->q;
1493 u16 remaining, curr_frag_len;
1496 skb = napi_get_frags(napi);
1498 be_rx_compl_discard(rxo, rxcp);
1502 remaining = rxcp->pkt_size;
1503 for (i = 0, j = -1; i < rxcp->num_rcvd; i++) {
1504 page_info = get_rx_page_info(rxo, rxcp->rxq_idx);
1506 curr_frag_len = min(remaining, rx_frag_size);
1508 /* Coalesce all frags from the same physical page in one slot */
1509 if (i == 0 || page_info->page_offset == 0) {
1510 /* First frag or Fresh page */
1512 skb_frag_set_page(skb, j, page_info->page);
1513 skb_shinfo(skb)->frags[j].page_offset =
1514 page_info->page_offset;
1515 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1517 put_page(page_info->page);
1519 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1520 skb->truesize += rx_frag_size;
1521 remaining -= curr_frag_len;
1522 index_inc(&rxcp->rxq_idx, rxq->len);
1523 memset(page_info, 0, sizeof(*page_info));
1525 BUG_ON(j > MAX_SKB_FRAGS);
1527 skb_shinfo(skb)->nr_frags = j + 1;
1528 skb->len = rxcp->pkt_size;
1529 skb->data_len = rxcp->pkt_size;
1530 skb->ip_summed = CHECKSUM_UNNECESSARY;
1531 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1532 if (adapter->netdev->features & NETIF_F_RXHASH)
1533 skb->rxhash = rxcp->rss_hash;
1536 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1538 napi_gro_frags(napi);
1541 static void be_parse_rx_compl_v1(struct be_eth_rx_compl *compl,
1542 struct be_rx_compl_info *rxcp)
1545 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, pktsize, compl);
1546 rxcp->vlanf = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, vtp, compl);
1547 rxcp->err = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, err, compl);
1548 rxcp->tcpf = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, tcpf, compl);
1549 rxcp->udpf = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, udpf, compl);
1551 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, ipcksm, compl);
1553 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, l4_cksm, compl);
1555 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, ip_version, compl);
1557 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, fragndx, compl);
1559 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, numfrags, compl);
1561 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, cast_enc, compl);
1563 AMAP_GET_BITS(struct amap_eth_rx_compl_v1, rsshash, compl);
1565 rxcp->vtm = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, vtm,
1567 rxcp->vlan_tag = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, vlan_tag,
1570 rxcp->port = AMAP_GET_BITS(struct amap_eth_rx_compl_v1, port, compl);
1573 static void be_parse_rx_compl_v0(struct be_eth_rx_compl *compl,
1574 struct be_rx_compl_info *rxcp)
1577 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, pktsize, compl);
1578 rxcp->vlanf = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, vtp, compl);
1579 rxcp->err = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, err, compl);
1580 rxcp->tcpf = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, tcpf, compl);
1581 rxcp->udpf = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, udpf, compl);
1583 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, ipcksm, compl);
1585 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, l4_cksm, compl);
1587 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, ip_version, compl);
1589 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, fragndx, compl);
1591 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, numfrags, compl);
1593 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, cast_enc, compl);
1595 AMAP_GET_BITS(struct amap_eth_rx_compl_v0, rsshash, compl);
1597 rxcp->vtm = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, vtm,
1599 rxcp->vlan_tag = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, vlan_tag,
1602 rxcp->port = AMAP_GET_BITS(struct amap_eth_rx_compl_v0, port, compl);
1603 rxcp->ip_frag = AMAP_GET_BITS(struct amap_eth_rx_compl_v0,
1607 static struct be_rx_compl_info *be_rx_compl_get(struct be_rx_obj *rxo)
1609 struct be_eth_rx_compl *compl = queue_tail_node(&rxo->cq);
1610 struct be_rx_compl_info *rxcp = &rxo->rxcp;
1611 struct be_adapter *adapter = rxo->adapter;
1613 /* For checking the valid bit it is Ok to use either definition as the
1614 * valid bit is at the same position in both v0 and v1 Rx compl */
1615 if (compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] == 0)
1619 be_dws_le_to_cpu(compl, sizeof(*compl));
1621 if (adapter->be3_native)
1622 be_parse_rx_compl_v1(compl, rxcp);
1624 be_parse_rx_compl_v0(compl, rxcp);
1630 /* vlanf could be wrongly set in some cards.
1631 * ignore if vtm is not set */
1632 if ((adapter->function_mode & FLEX10_MODE) && !rxcp->vtm)
1635 if (!lancer_chip(adapter))
1636 rxcp->vlan_tag = swab16(rxcp->vlan_tag);
1638 if (adapter->pvid == (rxcp->vlan_tag & VLAN_VID_MASK) &&
1639 !adapter->vlan_tag[rxcp->vlan_tag])
1643 /* As the compl has been parsed, reset it; we wont touch it again */
1644 compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] = 0;
1646 queue_tail_inc(&rxo->cq);
1650 static inline struct page *be_alloc_pages(u32 size, gfp_t gfp)
1652 u32 order = get_order(size);
1656 return alloc_pages(gfp, order);
1660 * Allocate a page, split it to fragments of size rx_frag_size and post as
1661 * receive buffers to BE
1663 static void be_post_rx_frags(struct be_rx_obj *rxo, gfp_t gfp)
1665 struct be_adapter *adapter = rxo->adapter;
1666 struct be_rx_page_info *page_info = NULL, *prev_page_info = NULL;
1667 struct be_queue_info *rxq = &rxo->q;
1668 struct page *pagep = NULL;
1669 struct be_eth_rx_d *rxd;
1670 u64 page_dmaaddr = 0, frag_dmaaddr;
1671 u32 posted, page_offset = 0;
1673 page_info = &rxo->page_info_tbl[rxq->head];
1674 for (posted = 0; posted < MAX_RX_POST && !page_info->page; posted++) {
1676 pagep = be_alloc_pages(adapter->big_page_size, gfp);
1677 if (unlikely(!pagep)) {
1678 rx_stats(rxo)->rx_post_fail++;
1681 page_dmaaddr = dma_map_page(&adapter->pdev->dev, pagep,
1682 0, adapter->big_page_size,
1684 page_info->page_offset = 0;
1687 page_info->page_offset = page_offset + rx_frag_size;
1689 page_offset = page_info->page_offset;
1690 page_info->page = pagep;
1691 dma_unmap_addr_set(page_info, bus, page_dmaaddr);
1692 frag_dmaaddr = page_dmaaddr + page_info->page_offset;
1694 rxd = queue_head_node(rxq);
1695 rxd->fragpa_lo = cpu_to_le32(frag_dmaaddr & 0xFFFFFFFF);
1696 rxd->fragpa_hi = cpu_to_le32(upper_32_bits(frag_dmaaddr));
1698 /* Any space left in the current big page for another frag? */
1699 if ((page_offset + rx_frag_size + rx_frag_size) >
1700 adapter->big_page_size) {
1702 page_info->last_page_user = true;
1705 prev_page_info = page_info;
1706 queue_head_inc(rxq);
1707 page_info = &rxo->page_info_tbl[rxq->head];
1710 prev_page_info->last_page_user = true;
1713 atomic_add(posted, &rxq->used);
1714 be_rxq_notify(adapter, rxq->id, posted);
1715 } else if (atomic_read(&rxq->used) == 0) {
1716 /* Let be_worker replenish when memory is available */
1717 rxo->rx_post_starved = true;
1721 static struct be_eth_tx_compl *be_tx_compl_get(struct be_queue_info *tx_cq)
1723 struct be_eth_tx_compl *txcp = queue_tail_node(tx_cq);
1725 if (txcp->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] == 0)
1729 be_dws_le_to_cpu(txcp, sizeof(*txcp));
1731 txcp->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] = 0;
1733 queue_tail_inc(tx_cq);
1737 static u16 be_tx_compl_process(struct be_adapter *adapter,
1738 struct be_tx_obj *txo, u16 last_index)
1740 struct be_queue_info *txq = &txo->q;
1741 struct be_eth_wrb *wrb;
1742 struct sk_buff **sent_skbs = txo->sent_skb_list;
1743 struct sk_buff *sent_skb;
1744 u16 cur_index, num_wrbs = 1; /* account for hdr wrb */
1745 bool unmap_skb_hdr = true;
1747 sent_skb = sent_skbs[txq->tail];
1749 sent_skbs[txq->tail] = NULL;
1751 /* skip header wrb */
1752 queue_tail_inc(txq);
1755 cur_index = txq->tail;
1756 wrb = queue_tail_node(txq);
1757 unmap_tx_frag(&adapter->pdev->dev, wrb,
1758 (unmap_skb_hdr && skb_headlen(sent_skb)));
1759 unmap_skb_hdr = false;
1762 queue_tail_inc(txq);
1763 } while (cur_index != last_index);
1765 kfree_skb(sent_skb);
1769 /* Return the number of events in the event queue */
1770 static inline int events_get(struct be_eq_obj *eqo)
1772 struct be_eq_entry *eqe;
1776 eqe = queue_tail_node(&eqo->q);
1783 queue_tail_inc(&eqo->q);
1789 /* Leaves the EQ is disarmed state */
1790 static void be_eq_clean(struct be_eq_obj *eqo)
1792 int num = events_get(eqo);
1794 be_eq_notify(eqo->adapter, eqo->q.id, false, true, num);
1797 static void be_rx_cq_clean(struct be_rx_obj *rxo)
1799 struct be_rx_page_info *page_info;
1800 struct be_queue_info *rxq = &rxo->q;
1801 struct be_queue_info *rx_cq = &rxo->cq;
1802 struct be_rx_compl_info *rxcp;
1803 struct be_adapter *adapter = rxo->adapter;
1807 /* Consume pending rx completions.
1808 * Wait for the flush completion (identified by zero num_rcvd)
1809 * to arrive. Notify CQ even when there are no more CQ entries
1810 * for HW to flush partially coalesced CQ entries.
1811 * In Lancer, there is no need to wait for flush compl.
1814 rxcp = be_rx_compl_get(rxo);
1816 if (lancer_chip(adapter))
1819 if (flush_wait++ > 10 || be_hw_error(adapter)) {
1820 dev_warn(&adapter->pdev->dev,
1821 "did not receive flush compl\n");
1824 be_cq_notify(adapter, rx_cq->id, true, 0);
1827 be_rx_compl_discard(rxo, rxcp);
1828 be_cq_notify(adapter, rx_cq->id, false, 1);
1829 if (rxcp->num_rcvd == 0)
1834 /* After cleanup, leave the CQ in unarmed state */
1835 be_cq_notify(adapter, rx_cq->id, false, 0);
1837 /* Then free posted rx buffers that were not used */
1838 tail = (rxq->head + rxq->len - atomic_read(&rxq->used)) % rxq->len;
1839 for (; atomic_read(&rxq->used) > 0; index_inc(&tail, rxq->len)) {
1840 page_info = get_rx_page_info(rxo, tail);
1841 put_page(page_info->page);
1842 memset(page_info, 0, sizeof(*page_info));
1844 BUG_ON(atomic_read(&rxq->used));
1845 rxq->tail = rxq->head = 0;
1848 static void be_tx_compl_clean(struct be_adapter *adapter)
1850 struct be_tx_obj *txo;
1851 struct be_queue_info *txq;
1852 struct be_eth_tx_compl *txcp;
1853 u16 end_idx, cmpl = 0, timeo = 0, num_wrbs = 0;
1854 struct sk_buff *sent_skb;
1856 int i, pending_txqs;
1858 /* Wait for a max of 200ms for all the tx-completions to arrive. */
1860 pending_txqs = adapter->num_tx_qs;
1862 for_all_tx_queues(adapter, txo, i) {
1864 while ((txcp = be_tx_compl_get(&txo->cq))) {
1866 AMAP_GET_BITS(struct amap_eth_tx_compl,
1868 num_wrbs += be_tx_compl_process(adapter, txo,
1873 be_cq_notify(adapter, txo->cq.id, false, cmpl);
1874 atomic_sub(num_wrbs, &txq->used);
1878 if (atomic_read(&txq->used) == 0)
1882 if (pending_txqs == 0 || ++timeo > 200)
1888 for_all_tx_queues(adapter, txo, i) {
1890 if (atomic_read(&txq->used))
1891 dev_err(&adapter->pdev->dev, "%d pending tx-compls\n",
1892 atomic_read(&txq->used));
1894 /* free posted tx for which compls will never arrive */
1895 while (atomic_read(&txq->used)) {
1896 sent_skb = txo->sent_skb_list[txq->tail];
1897 end_idx = txq->tail;
1898 num_wrbs = wrb_cnt_for_skb(adapter, sent_skb,
1900 index_adv(&end_idx, num_wrbs - 1, txq->len);
1901 num_wrbs = be_tx_compl_process(adapter, txo, end_idx);
1902 atomic_sub(num_wrbs, &txq->used);
1907 static void be_evt_queues_destroy(struct be_adapter *adapter)
1909 struct be_eq_obj *eqo;
1912 for_all_evt_queues(adapter, eqo, i) {
1913 if (eqo->q.created) {
1915 be_cmd_q_destroy(adapter, &eqo->q, QTYPE_EQ);
1917 be_queue_free(adapter, &eqo->q);
1921 static int be_evt_queues_create(struct be_adapter *adapter)
1923 struct be_queue_info *eq;
1924 struct be_eq_obj *eqo;
1927 adapter->num_evt_qs = num_irqs(adapter);
1929 for_all_evt_queues(adapter, eqo, i) {
1930 eqo->adapter = adapter;
1931 eqo->tx_budget = BE_TX_BUDGET;
1933 eqo->max_eqd = BE_MAX_EQD;
1934 eqo->enable_aic = true;
1937 rc = be_queue_alloc(adapter, eq, EVNT_Q_LEN,
1938 sizeof(struct be_eq_entry));
1942 rc = be_cmd_eq_create(adapter, eq, eqo->cur_eqd);
1949 static void be_mcc_queues_destroy(struct be_adapter *adapter)
1951 struct be_queue_info *q;
1953 q = &adapter->mcc_obj.q;
1955 be_cmd_q_destroy(adapter, q, QTYPE_MCCQ);
1956 be_queue_free(adapter, q);
1958 q = &adapter->mcc_obj.cq;
1960 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
1961 be_queue_free(adapter, q);
1964 /* Must be called only after TX qs are created as MCC shares TX EQ */
1965 static int be_mcc_queues_create(struct be_adapter *adapter)
1967 struct be_queue_info *q, *cq;
1969 cq = &adapter->mcc_obj.cq;
1970 if (be_queue_alloc(adapter, cq, MCC_CQ_LEN,
1971 sizeof(struct be_mcc_compl)))
1974 /* Use the default EQ for MCC completions */
1975 if (be_cmd_cq_create(adapter, cq, &mcc_eqo(adapter)->q, true, 0))
1978 q = &adapter->mcc_obj.q;
1979 if (be_queue_alloc(adapter, q, MCC_Q_LEN, sizeof(struct be_mcc_wrb)))
1980 goto mcc_cq_destroy;
1982 if (be_cmd_mccq_create(adapter, q, cq))
1988 be_queue_free(adapter, q);
1990 be_cmd_q_destroy(adapter, cq, QTYPE_CQ);
1992 be_queue_free(adapter, cq);
1997 static void be_tx_queues_destroy(struct be_adapter *adapter)
1999 struct be_queue_info *q;
2000 struct be_tx_obj *txo;
2003 for_all_tx_queues(adapter, txo, i) {
2006 be_cmd_q_destroy(adapter, q, QTYPE_TXQ);
2007 be_queue_free(adapter, q);
2011 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2012 be_queue_free(adapter, q);
2016 static int be_num_txqs_want(struct be_adapter *adapter)
2018 if ((!lancer_chip(adapter) && sriov_want(adapter)) ||
2019 be_is_mc(adapter) ||
2020 (!lancer_chip(adapter) && !be_physfn(adapter)) ||
2024 return adapter->max_tx_queues;
2027 static int be_tx_cqs_create(struct be_adapter *adapter)
2029 struct be_queue_info *cq, *eq;
2031 struct be_tx_obj *txo;
2034 adapter->num_tx_qs = be_num_txqs_want(adapter);
2035 if (adapter->num_tx_qs != MAX_TX_QS) {
2037 netif_set_real_num_tx_queues(adapter->netdev,
2038 adapter->num_tx_qs);
2042 for_all_tx_queues(adapter, txo, i) {
2044 status = be_queue_alloc(adapter, cq, TX_CQ_LEN,
2045 sizeof(struct be_eth_tx_compl));
2049 /* If num_evt_qs is less than num_tx_qs, then more than
2050 * one txq share an eq
2052 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2053 status = be_cmd_cq_create(adapter, cq, eq, false, 3);
2060 static int be_tx_qs_create(struct be_adapter *adapter)
2062 struct be_tx_obj *txo;
2065 for_all_tx_queues(adapter, txo, i) {
2066 status = be_queue_alloc(adapter, &txo->q, TX_Q_LEN,
2067 sizeof(struct be_eth_wrb));
2071 status = be_cmd_txq_create(adapter, txo);
2076 dev_info(&adapter->pdev->dev, "created %d TX queue(s)\n",
2077 adapter->num_tx_qs);
2081 static void be_rx_cqs_destroy(struct be_adapter *adapter)
2083 struct be_queue_info *q;
2084 struct be_rx_obj *rxo;
2087 for_all_rx_queues(adapter, rxo, i) {
2090 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2091 be_queue_free(adapter, q);
2095 static int be_rx_cqs_create(struct be_adapter *adapter)
2097 struct be_queue_info *eq, *cq;
2098 struct be_rx_obj *rxo;
2101 /* We'll create as many RSS rings as there are irqs.
2102 * But when there's only one irq there's no use creating RSS rings
2104 adapter->num_rx_qs = (num_irqs(adapter) > 1) ?
2105 num_irqs(adapter) + 1 : 1;
2106 if (adapter->num_rx_qs != MAX_RX_QS) {
2108 netif_set_real_num_rx_queues(adapter->netdev,
2109 adapter->num_rx_qs);
2113 adapter->big_page_size = (1 << get_order(rx_frag_size)) * PAGE_SIZE;
2114 for_all_rx_queues(adapter, rxo, i) {
2115 rxo->adapter = adapter;
2117 rc = be_queue_alloc(adapter, cq, RX_CQ_LEN,
2118 sizeof(struct be_eth_rx_compl));
2122 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2123 rc = be_cmd_cq_create(adapter, cq, eq, false, 3);
2128 dev_info(&adapter->pdev->dev,
2129 "created %d RSS queue(s) and 1 default RX queue\n",
2130 adapter->num_rx_qs - 1);
2134 static irqreturn_t be_intx(int irq, void *dev)
2136 struct be_eq_obj *eqo = dev;
2137 struct be_adapter *adapter = eqo->adapter;
2140 /* IRQ is not expected when NAPI is scheduled as the EQ
2141 * will not be armed.
2142 * But, this can happen on Lancer INTx where it takes
2143 * a while to de-assert INTx or in BE2 where occasionaly
2144 * an interrupt may be raised even when EQ is unarmed.
2145 * If NAPI is already scheduled, then counting & notifying
2146 * events will orphan them.
2148 if (napi_schedule_prep(&eqo->napi)) {
2149 num_evts = events_get(eqo);
2150 __napi_schedule(&eqo->napi);
2152 eqo->spurious_intr = 0;
2154 be_eq_notify(adapter, eqo->q.id, false, true, num_evts);
2156 /* Return IRQ_HANDLED only for the the first spurious intr
2157 * after a valid intr to stop the kernel from branding
2158 * this irq as a bad one!
2160 if (num_evts || eqo->spurious_intr++ == 0)
2166 static irqreturn_t be_msix(int irq, void *dev)
2168 struct be_eq_obj *eqo = dev;
2170 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
2171 napi_schedule(&eqo->napi);
2175 static inline bool do_gro(struct be_rx_compl_info *rxcp)
2177 return (rxcp->tcpf && !rxcp->err && rxcp->l4_csum) ? true : false;
2180 static int be_process_rx(struct be_rx_obj *rxo, struct napi_struct *napi,
2183 struct be_adapter *adapter = rxo->adapter;
2184 struct be_queue_info *rx_cq = &rxo->cq;
2185 struct be_rx_compl_info *rxcp;
2188 for (work_done = 0; work_done < budget; work_done++) {
2189 rxcp = be_rx_compl_get(rxo);
2193 /* Is it a flush compl that has no data */
2194 if (unlikely(rxcp->num_rcvd == 0))
2197 /* Discard compl with partial DMA Lancer B0 */
2198 if (unlikely(!rxcp->pkt_size)) {
2199 be_rx_compl_discard(rxo, rxcp);
2203 /* On BE drop pkts that arrive due to imperfect filtering in
2204 * promiscuous mode on some skews
2206 if (unlikely(rxcp->port != adapter->port_num &&
2207 !lancer_chip(adapter))) {
2208 be_rx_compl_discard(rxo, rxcp);
2213 be_rx_compl_process_gro(rxo, napi, rxcp);
2215 be_rx_compl_process(rxo, rxcp);
2217 be_rx_stats_update(rxo, rxcp);
2221 be_cq_notify(adapter, rx_cq->id, true, work_done);
2223 if (atomic_read(&rxo->q.used) < RX_FRAGS_REFILL_WM)
2224 be_post_rx_frags(rxo, GFP_ATOMIC);
2230 static bool be_process_tx(struct be_adapter *adapter, struct be_tx_obj *txo,
2231 int budget, int idx)
2233 struct be_eth_tx_compl *txcp;
2234 int num_wrbs = 0, work_done;
2236 for (work_done = 0; work_done < budget; work_done++) {
2237 txcp = be_tx_compl_get(&txo->cq);
2240 num_wrbs += be_tx_compl_process(adapter, txo,
2241 AMAP_GET_BITS(struct amap_eth_tx_compl,
2246 be_cq_notify(adapter, txo->cq.id, true, work_done);
2247 atomic_sub(num_wrbs, &txo->q.used);
2249 /* As Tx wrbs have been freed up, wake up netdev queue
2250 * if it was stopped due to lack of tx wrbs. */
2251 if (__netif_subqueue_stopped(adapter->netdev, idx) &&
2252 atomic_read(&txo->q.used) < txo->q.len / 2) {
2253 netif_wake_subqueue(adapter->netdev, idx);
2256 u64_stats_update_begin(&tx_stats(txo)->sync_compl);
2257 tx_stats(txo)->tx_compl += work_done;
2258 u64_stats_update_end(&tx_stats(txo)->sync_compl);
2260 return (work_done < budget); /* Done */
2263 static int be_poll(struct napi_struct *napi, int budget)
2265 struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2266 struct be_adapter *adapter = eqo->adapter;
2267 int max_work = 0, work, i, num_evts;
2270 num_evts = events_get(eqo);
2272 /* Process all TXQs serviced by this EQ */
2273 for (i = eqo->idx; i < adapter->num_tx_qs; i += adapter->num_evt_qs) {
2274 tx_done = be_process_tx(adapter, &adapter->tx_obj[i],
2280 /* This loop will iterate twice for EQ0 in which
2281 * completions of the last RXQ (default one) are also processed
2282 * For other EQs the loop iterates only once
2284 for (i = eqo->idx; i < adapter->num_rx_qs; i += adapter->num_evt_qs) {
2285 work = be_process_rx(&adapter->rx_obj[i], napi, budget);
2286 max_work = max(work, max_work);
2289 if (is_mcc_eqo(eqo))
2290 be_process_mcc(adapter);
2292 if (max_work < budget) {
2293 napi_complete(napi);
2294 be_eq_notify(adapter, eqo->q.id, true, false, num_evts);
2296 /* As we'll continue in polling mode, count and clear events */
2297 be_eq_notify(adapter, eqo->q.id, false, false, num_evts);
2302 void be_detect_error(struct be_adapter *adapter)
2304 u32 ue_lo = 0, ue_hi = 0, ue_lo_mask = 0, ue_hi_mask = 0;
2305 u32 sliport_status = 0, sliport_err1 = 0, sliport_err2 = 0;
2308 if (be_hw_error(adapter))
2311 if (lancer_chip(adapter)) {
2312 sliport_status = ioread32(adapter->db + SLIPORT_STATUS_OFFSET);
2313 if (sliport_status & SLIPORT_STATUS_ERR_MASK) {
2314 sliport_err1 = ioread32(adapter->db +
2315 SLIPORT_ERROR1_OFFSET);
2316 sliport_err2 = ioread32(adapter->db +
2317 SLIPORT_ERROR2_OFFSET);
2320 pci_read_config_dword(adapter->pdev,
2321 PCICFG_UE_STATUS_LOW, &ue_lo);
2322 pci_read_config_dword(adapter->pdev,
2323 PCICFG_UE_STATUS_HIGH, &ue_hi);
2324 pci_read_config_dword(adapter->pdev,
2325 PCICFG_UE_STATUS_LOW_MASK, &ue_lo_mask);
2326 pci_read_config_dword(adapter->pdev,
2327 PCICFG_UE_STATUS_HI_MASK, &ue_hi_mask);
2329 ue_lo = (ue_lo & ~ue_lo_mask);
2330 ue_hi = (ue_hi & ~ue_hi_mask);
2333 /* On certain platforms BE hardware can indicate spurious UEs.
2334 * Allow the h/w to stop working completely in case of a real UE.
2335 * Hence not setting the hw_error for UE detection.
2337 if (sliport_status & SLIPORT_STATUS_ERR_MASK) {
2338 adapter->hw_error = true;
2339 dev_err(&adapter->pdev->dev,
2340 "Error detected in the card\n");
2343 if (sliport_status & SLIPORT_STATUS_ERR_MASK) {
2344 dev_err(&adapter->pdev->dev,
2345 "ERR: sliport status 0x%x\n", sliport_status);
2346 dev_err(&adapter->pdev->dev,
2347 "ERR: sliport error1 0x%x\n", sliport_err1);
2348 dev_err(&adapter->pdev->dev,
2349 "ERR: sliport error2 0x%x\n", sliport_err2);
2353 for (i = 0; ue_lo; ue_lo >>= 1, i++) {
2355 dev_err(&adapter->pdev->dev,
2356 "UE: %s bit set\n", ue_status_low_desc[i]);
2361 for (i = 0; ue_hi; ue_hi >>= 1, i++) {
2363 dev_err(&adapter->pdev->dev,
2364 "UE: %s bit set\n", ue_status_hi_desc[i]);
2370 static void be_msix_disable(struct be_adapter *adapter)
2372 if (msix_enabled(adapter)) {
2373 pci_disable_msix(adapter->pdev);
2374 adapter->num_msix_vec = 0;
2378 static uint be_num_rss_want(struct be_adapter *adapter)
2382 if ((adapter->function_caps & BE_FUNCTION_CAPS_RSS) &&
2383 (lancer_chip(adapter) ||
2384 (!sriov_want(adapter) && be_physfn(adapter)))) {
2385 num = adapter->max_rss_queues;
2386 num = min_t(u32, num, (u32)netif_get_num_default_rss_queues());
2391 static int be_msix_enable(struct be_adapter *adapter)
2393 #define BE_MIN_MSIX_VECTORS 1
2394 int i, status, num_vec, num_roce_vec = 0;
2395 struct device *dev = &adapter->pdev->dev;
2397 /* If RSS queues are not used, need a vec for default RX Q */
2398 num_vec = min(be_num_rss_want(adapter), num_online_cpus());
2399 if (be_roce_supported(adapter)) {
2400 num_roce_vec = min_t(u32, MAX_ROCE_MSIX_VECTORS,
2401 (num_online_cpus() + 1));
2402 num_roce_vec = min(num_roce_vec, MAX_ROCE_EQS);
2403 num_vec += num_roce_vec;
2404 num_vec = min(num_vec, MAX_MSIX_VECTORS);
2406 num_vec = max(num_vec, BE_MIN_MSIX_VECTORS);
2408 for (i = 0; i < num_vec; i++)
2409 adapter->msix_entries[i].entry = i;
2411 status = pci_enable_msix(adapter->pdev, adapter->msix_entries, num_vec);
2414 } else if (status >= BE_MIN_MSIX_VECTORS) {
2416 status = pci_enable_msix(adapter->pdev, adapter->msix_entries,
2422 dev_warn(dev, "MSIx enable failed\n");
2423 /* INTx is not supported in VFs, so fail probe if enable_msix fails */
2424 if (!be_physfn(adapter))
2428 if (be_roce_supported(adapter)) {
2429 if (num_vec > num_roce_vec) {
2430 adapter->num_msix_vec = num_vec - num_roce_vec;
2431 adapter->num_msix_roce_vec =
2432 num_vec - adapter->num_msix_vec;
2434 adapter->num_msix_vec = num_vec;
2435 adapter->num_msix_roce_vec = 0;
2438 adapter->num_msix_vec = num_vec;
2439 dev_info(dev, "enabled %d MSI-x vector(s)\n", adapter->num_msix_vec);
2443 static inline int be_msix_vec_get(struct be_adapter *adapter,
2444 struct be_eq_obj *eqo)
2446 return adapter->msix_entries[eqo->idx].vector;
2449 static int be_msix_register(struct be_adapter *adapter)
2451 struct net_device *netdev = adapter->netdev;
2452 struct be_eq_obj *eqo;
2455 for_all_evt_queues(adapter, eqo, i) {
2456 sprintf(eqo->desc, "%s-q%d", netdev->name, i);
2457 vec = be_msix_vec_get(adapter, eqo);
2458 status = request_irq(vec, be_msix, 0, eqo->desc, eqo);
2465 for (i--, eqo = &adapter->eq_obj[i]; i >= 0; i--, eqo--)
2466 free_irq(be_msix_vec_get(adapter, eqo), eqo);
2467 dev_warn(&adapter->pdev->dev, "MSIX Request IRQ failed - err %d\n",
2469 be_msix_disable(adapter);
2473 static int be_irq_register(struct be_adapter *adapter)
2475 struct net_device *netdev = adapter->netdev;
2478 if (msix_enabled(adapter)) {
2479 status = be_msix_register(adapter);
2482 /* INTx is not supported for VF */
2483 if (!be_physfn(adapter))
2487 /* INTx: only the first EQ is used */
2488 netdev->irq = adapter->pdev->irq;
2489 status = request_irq(netdev->irq, be_intx, IRQF_SHARED, netdev->name,
2490 &adapter->eq_obj[0]);
2492 dev_err(&adapter->pdev->dev,
2493 "INTx request IRQ failed - err %d\n", status);
2497 adapter->isr_registered = true;
2501 static void be_irq_unregister(struct be_adapter *adapter)
2503 struct net_device *netdev = adapter->netdev;
2504 struct be_eq_obj *eqo;
2507 if (!adapter->isr_registered)
2511 if (!msix_enabled(adapter)) {
2512 free_irq(netdev->irq, &adapter->eq_obj[0]);
2517 for_all_evt_queues(adapter, eqo, i)
2518 free_irq(be_msix_vec_get(adapter, eqo), eqo);
2521 adapter->isr_registered = false;
2524 static void be_rx_qs_destroy(struct be_adapter *adapter)
2526 struct be_queue_info *q;
2527 struct be_rx_obj *rxo;
2530 for_all_rx_queues(adapter, rxo, i) {
2533 be_cmd_rxq_destroy(adapter, q);
2534 be_rx_cq_clean(rxo);
2536 be_queue_free(adapter, q);
2540 static int be_close(struct net_device *netdev)
2542 struct be_adapter *adapter = netdev_priv(netdev);
2543 struct be_eq_obj *eqo;
2546 be_roce_dev_close(adapter);
2548 if (adapter->flags & BE_FLAGS_NAPI_ENABLED) {
2549 for_all_evt_queues(adapter, eqo, i)
2550 napi_disable(&eqo->napi);
2551 adapter->flags &= ~BE_FLAGS_NAPI_ENABLED;
2554 be_async_mcc_disable(adapter);
2556 /* Wait for all pending tx completions to arrive so that
2557 * all tx skbs are freed.
2559 netif_tx_disable(netdev);
2560 be_tx_compl_clean(adapter);
2562 be_rx_qs_destroy(adapter);
2564 for_all_evt_queues(adapter, eqo, i) {
2565 if (msix_enabled(adapter))
2566 synchronize_irq(be_msix_vec_get(adapter, eqo));
2568 synchronize_irq(netdev->irq);
2572 be_irq_unregister(adapter);
2577 static int be_rx_qs_create(struct be_adapter *adapter)
2579 struct be_rx_obj *rxo;
2583 for_all_rx_queues(adapter, rxo, i) {
2584 rc = be_queue_alloc(adapter, &rxo->q, RX_Q_LEN,
2585 sizeof(struct be_eth_rx_d));
2590 /* The FW would like the default RXQ to be created first */
2591 rxo = default_rxo(adapter);
2592 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id, rx_frag_size,
2593 adapter->if_handle, false, &rxo->rss_id);
2597 for_all_rss_queues(adapter, rxo, i) {
2598 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id,
2599 rx_frag_size, adapter->if_handle,
2600 true, &rxo->rss_id);
2605 if (be_multi_rxq(adapter)) {
2606 for (j = 0; j < 128; j += adapter->num_rx_qs - 1) {
2607 for_all_rss_queues(adapter, rxo, i) {
2610 rsstable[j + i] = rxo->rss_id;
2613 adapter->rss_flags = RSS_ENABLE_TCP_IPV4 | RSS_ENABLE_IPV4 |
2614 RSS_ENABLE_TCP_IPV6 | RSS_ENABLE_IPV6;
2616 if (!BEx_chip(adapter))
2617 adapter->rss_flags |= RSS_ENABLE_UDP_IPV4 |
2618 RSS_ENABLE_UDP_IPV6;
2620 rc = be_cmd_rss_config(adapter, rsstable, adapter->rss_flags,
2623 adapter->rss_flags = 0;
2628 /* First time posting */
2629 for_all_rx_queues(adapter, rxo, i)
2630 be_post_rx_frags(rxo, GFP_KERNEL);
2634 static int be_open(struct net_device *netdev)
2636 struct be_adapter *adapter = netdev_priv(netdev);
2637 struct be_eq_obj *eqo;
2638 struct be_rx_obj *rxo;
2639 struct be_tx_obj *txo;
2643 status = be_rx_qs_create(adapter);
2647 status = be_irq_register(adapter);
2651 for_all_rx_queues(adapter, rxo, i)
2652 be_cq_notify(adapter, rxo->cq.id, true, 0);
2654 for_all_tx_queues(adapter, txo, i)
2655 be_cq_notify(adapter, txo->cq.id, true, 0);
2657 be_async_mcc_enable(adapter);
2659 for_all_evt_queues(adapter, eqo, i) {
2660 napi_enable(&eqo->napi);
2661 be_eq_notify(adapter, eqo->q.id, true, false, 0);
2663 adapter->flags |= BE_FLAGS_NAPI_ENABLED;
2665 status = be_cmd_link_status_query(adapter, NULL, &link_status, 0);
2667 be_link_status_update(adapter, link_status);
2669 netif_tx_start_all_queues(netdev);
2670 be_roce_dev_open(adapter);
2673 be_close(adapter->netdev);
2677 static int be_setup_wol(struct be_adapter *adapter, bool enable)
2679 struct be_dma_mem cmd;
2683 memset(mac, 0, ETH_ALEN);
2685 cmd.size = sizeof(struct be_cmd_req_acpi_wol_magic_config);
2686 cmd.va = dma_alloc_coherent(&adapter->pdev->dev, cmd.size, &cmd.dma,
2687 GFP_KERNEL | __GFP_ZERO);
2692 status = pci_write_config_dword(adapter->pdev,
2693 PCICFG_PM_CONTROL_OFFSET, PCICFG_PM_CONTROL_MASK);
2695 dev_err(&adapter->pdev->dev,
2696 "Could not enable Wake-on-lan\n");
2697 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va,
2701 status = be_cmd_enable_magic_wol(adapter,
2702 adapter->netdev->dev_addr, &cmd);
2703 pci_enable_wake(adapter->pdev, PCI_D3hot, 1);
2704 pci_enable_wake(adapter->pdev, PCI_D3cold, 1);
2706 status = be_cmd_enable_magic_wol(adapter, mac, &cmd);
2707 pci_enable_wake(adapter->pdev, PCI_D3hot, 0);
2708 pci_enable_wake(adapter->pdev, PCI_D3cold, 0);
2711 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va, cmd.dma);
2716 * Generate a seed MAC address from the PF MAC Address using jhash.
2717 * MAC Address for VFs are assigned incrementally starting from the seed.
2718 * These addresses are programmed in the ASIC by the PF and the VF driver
2719 * queries for the MAC address during its probe.
2721 static int be_vf_eth_addr_config(struct be_adapter *adapter)
2726 struct be_vf_cfg *vf_cfg;
2728 be_vf_eth_addr_generate(adapter, mac);
2730 for_all_vfs(adapter, vf_cfg, vf) {
2731 if (BEx_chip(adapter))
2732 status = be_cmd_pmac_add(adapter, mac,
2734 &vf_cfg->pmac_id, vf + 1);
2736 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
2740 dev_err(&adapter->pdev->dev,
2741 "Mac address assignment failed for VF %d\n", vf);
2743 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
2750 static int be_vfs_mac_query(struct be_adapter *adapter)
2754 struct be_vf_cfg *vf_cfg;
2755 bool active = false;
2757 for_all_vfs(adapter, vf_cfg, vf) {
2758 be_cmd_get_mac_from_list(adapter, mac, &active,
2759 &vf_cfg->pmac_id, 0);
2761 status = be_cmd_mac_addr_query(adapter, mac, false,
2762 vf_cfg->if_handle, 0);
2765 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
2770 static void be_vf_clear(struct be_adapter *adapter)
2772 struct be_vf_cfg *vf_cfg;
2775 if (pci_vfs_assigned(adapter->pdev)) {
2776 dev_warn(&adapter->pdev->dev,
2777 "VFs are assigned to VMs: not disabling VFs\n");
2781 pci_disable_sriov(adapter->pdev);
2783 for_all_vfs(adapter, vf_cfg, vf) {
2784 if (BEx_chip(adapter))
2785 be_cmd_pmac_del(adapter, vf_cfg->if_handle,
2786 vf_cfg->pmac_id, vf + 1);
2788 be_cmd_set_mac(adapter, NULL, vf_cfg->if_handle,
2791 be_cmd_if_destroy(adapter, vf_cfg->if_handle, vf + 1);
2794 kfree(adapter->vf_cfg);
2795 adapter->num_vfs = 0;
2798 static int be_clear(struct be_adapter *adapter)
2802 if (adapter->flags & BE_FLAGS_WORKER_SCHEDULED) {
2803 cancel_delayed_work_sync(&adapter->work);
2804 adapter->flags &= ~BE_FLAGS_WORKER_SCHEDULED;
2807 if (sriov_enabled(adapter))
2808 be_vf_clear(adapter);
2810 /* delete the primary mac along with the uc-mac list */
2811 for (i = 0; i < (adapter->uc_macs + 1); i++)
2812 be_cmd_pmac_del(adapter, adapter->if_handle,
2813 adapter->pmac_id[i], 0);
2814 adapter->uc_macs = 0;
2816 be_cmd_if_destroy(adapter, adapter->if_handle, 0);
2818 be_mcc_queues_destroy(adapter);
2819 be_rx_cqs_destroy(adapter);
2820 be_tx_queues_destroy(adapter);
2821 be_evt_queues_destroy(adapter);
2823 kfree(adapter->pmac_id);
2824 adapter->pmac_id = NULL;
2826 be_msix_disable(adapter);
2830 static int be_vfs_if_create(struct be_adapter *adapter)
2832 struct be_vf_cfg *vf_cfg;
2833 u32 cap_flags, en_flags, vf;
2836 cap_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
2837 BE_IF_FLAGS_MULTICAST;
2839 for_all_vfs(adapter, vf_cfg, vf) {
2840 if (!BE3_chip(adapter))
2841 be_cmd_get_profile_config(adapter, &cap_flags,
2844 /* If a FW profile exists, then cap_flags are updated */
2845 en_flags = cap_flags & (BE_IF_FLAGS_UNTAGGED |
2846 BE_IF_FLAGS_BROADCAST | BE_IF_FLAGS_MULTICAST);
2847 status = be_cmd_if_create(adapter, cap_flags, en_flags,
2848 &vf_cfg->if_handle, vf + 1);
2856 static int be_vf_setup_init(struct be_adapter *adapter)
2858 struct be_vf_cfg *vf_cfg;
2861 adapter->vf_cfg = kcalloc(adapter->num_vfs, sizeof(*vf_cfg),
2863 if (!adapter->vf_cfg)
2866 for_all_vfs(adapter, vf_cfg, vf) {
2867 vf_cfg->if_handle = -1;
2868 vf_cfg->pmac_id = -1;
2873 static int be_vf_setup(struct be_adapter *adapter)
2875 struct be_vf_cfg *vf_cfg;
2876 u16 def_vlan, lnk_speed;
2877 int status, old_vfs, vf;
2878 struct device *dev = &adapter->pdev->dev;
2881 old_vfs = pci_num_vf(adapter->pdev);
2883 dev_info(dev, "%d VFs are already enabled\n", old_vfs);
2884 if (old_vfs != num_vfs)
2885 dev_warn(dev, "Ignoring num_vfs=%d setting\n", num_vfs);
2886 adapter->num_vfs = old_vfs;
2888 if (num_vfs > adapter->dev_num_vfs)
2889 dev_info(dev, "Device supports %d VFs and not %d\n",
2890 adapter->dev_num_vfs, num_vfs);
2891 adapter->num_vfs = min_t(u16, num_vfs, adapter->dev_num_vfs);
2892 if (!adapter->num_vfs)
2896 status = be_vf_setup_init(adapter);
2901 for_all_vfs(adapter, vf_cfg, vf) {
2902 status = be_cmd_get_if_id(adapter, vf_cfg, vf);
2907 status = be_vfs_if_create(adapter);
2913 status = be_vfs_mac_query(adapter);
2917 status = be_vf_eth_addr_config(adapter);
2922 for_all_vfs(adapter, vf_cfg, vf) {
2923 /* Allow VFs to programs MAC/VLAN filters */
2924 status = be_cmd_get_fn_privileges(adapter, &privileges, vf + 1);
2925 if (!status && !(privileges & BE_PRIV_FILTMGMT)) {
2926 status = be_cmd_set_fn_privileges(adapter,
2931 dev_info(dev, "VF%d has FILTMGMT privilege\n",
2935 /* BE3 FW, by default, caps VF TX-rate to 100mbps.
2936 * Allow full available bandwidth
2938 if (BE3_chip(adapter) && !old_vfs)
2939 be_cmd_set_qos(adapter, 1000, vf+1);
2941 status = be_cmd_link_status_query(adapter, &lnk_speed,
2944 vf_cfg->tx_rate = lnk_speed;
2946 status = be_cmd_get_hsw_config(adapter, &def_vlan,
2947 vf + 1, vf_cfg->if_handle);
2950 vf_cfg->def_vid = def_vlan;
2952 be_cmd_enable_vf(adapter, vf + 1);
2956 status = pci_enable_sriov(adapter->pdev, adapter->num_vfs);
2958 dev_err(dev, "SRIOV enable failed\n");
2959 adapter->num_vfs = 0;
2965 dev_err(dev, "VF setup failed\n");
2966 be_vf_clear(adapter);
2970 static void be_setup_init(struct be_adapter *adapter)
2972 adapter->vlan_prio_bmap = 0xff;
2973 adapter->phy.link_speed = -1;
2974 adapter->if_handle = -1;
2975 adapter->be3_native = false;
2976 adapter->promiscuous = false;
2977 if (be_physfn(adapter))
2978 adapter->cmd_privileges = MAX_PRIVILEGES;
2980 adapter->cmd_privileges = MIN_PRIVILEGES;
2983 static void be_get_resources(struct be_adapter *adapter)
2987 bool profile_present = false;
2990 if (!BEx_chip(adapter)) {
2991 status = be_cmd_get_func_config(adapter);
2993 profile_present = true;
2994 } else if (BE3_chip(adapter) && be_physfn(adapter)) {
2995 be_cmd_get_profile_config(adapter, NULL, &txq_count, 0);
2998 if (profile_present) {
2999 adapter->max_tx_queues = min_t(u16, adapter->max_tx_queues,
3001 adapter->max_rss_queues = min_t(u16, adapter->max_rss_queues,
3003 adapter->max_event_queues = min_t(u16,
3004 adapter->max_event_queues,
3007 if (adapter->max_rss_queues &&
3008 adapter->max_rss_queues == adapter->max_rx_queues)
3009 adapter->max_rss_queues -= 1;
3011 if (adapter->max_event_queues < adapter->max_rss_queues)
3012 adapter->max_rss_queues = adapter->max_event_queues;
3015 if (be_physfn(adapter))
3016 adapter->max_pmac_cnt = BE_UC_PMAC_COUNT;
3018 adapter->max_pmac_cnt = BE_VF_UC_PMAC_COUNT;
3020 if (adapter->function_mode & FLEX10_MODE)
3021 adapter->max_vlans = BE_NUM_VLANS_SUPPORTED/8;
3023 adapter->max_vlans = BE_NUM_VLANS_SUPPORTED;
3025 adapter->max_mcast_mac = BE_MAX_MC;
3026 adapter->max_tx_queues = txq_count ? txq_count : MAX_TX_QS;
3027 adapter->max_tx_queues = min_t(u16, adapter->max_tx_queues,
3029 adapter->max_rss_queues = (adapter->be3_native) ?
3030 BE3_MAX_RSS_QS : BE2_MAX_RSS_QS;
3031 adapter->max_event_queues = BE3_MAX_RSS_QS;
3033 adapter->if_cap_flags = BE_IF_FLAGS_UNTAGGED |
3034 BE_IF_FLAGS_BROADCAST |
3035 BE_IF_FLAGS_MULTICAST |
3036 BE_IF_FLAGS_PASS_L3L4_ERRORS |
3037 BE_IF_FLAGS_MCAST_PROMISCUOUS |
3038 BE_IF_FLAGS_VLAN_PROMISCUOUS |
3039 BE_IF_FLAGS_PROMISCUOUS;
3041 if (adapter->function_caps & BE_FUNCTION_CAPS_RSS)
3042 adapter->if_cap_flags |= BE_IF_FLAGS_RSS;
3045 pos = pci_find_ext_capability(adapter->pdev, PCI_EXT_CAP_ID_SRIOV);
3047 pci_read_config_word(adapter->pdev, pos + PCI_SRIOV_TOTAL_VF,
3049 if (BE3_chip(adapter))
3050 dev_num_vfs = min_t(u16, dev_num_vfs, MAX_VFS);
3051 adapter->dev_num_vfs = dev_num_vfs;
3055 /* Routine to query per function resource limits */
3056 static int be_get_config(struct be_adapter *adapter)
3060 status = be_cmd_query_fw_cfg(adapter, &adapter->port_num,
3061 &adapter->function_mode,
3062 &adapter->function_caps,
3063 &adapter->asic_rev);
3067 be_get_resources(adapter);
3069 /* primary mac needs 1 pmac entry */
3070 adapter->pmac_id = kcalloc(adapter->max_pmac_cnt + 1,
3071 sizeof(u32), GFP_KERNEL);
3072 if (!adapter->pmac_id) {
3081 static int be_mac_setup(struct be_adapter *adapter)
3086 if (is_zero_ether_addr(adapter->netdev->dev_addr)) {
3087 status = be_cmd_get_perm_mac(adapter, mac);
3091 memcpy(adapter->netdev->dev_addr, mac, ETH_ALEN);
3092 memcpy(adapter->netdev->perm_addr, mac, ETH_ALEN);
3094 /* Maybe the HW was reset; dev_addr must be re-programmed */
3095 memcpy(mac, adapter->netdev->dev_addr, ETH_ALEN);
3098 /* On BE3 VFs this cmd may fail due to lack of privilege.
3099 * Ignore the failure as in this case pmac_id is fetched
3100 * in the IFACE_CREATE cmd.
3102 be_cmd_pmac_add(adapter, mac, adapter->if_handle,
3103 &adapter->pmac_id[0], 0);
3107 static int be_setup(struct be_adapter *adapter)
3109 struct device *dev = &adapter->pdev->dev;
3114 be_setup_init(adapter);
3116 if (!lancer_chip(adapter))
3117 be_cmd_req_native_mode(adapter);
3119 status = be_get_config(adapter);
3123 status = be_msix_enable(adapter);
3127 status = be_evt_queues_create(adapter);
3131 status = be_tx_cqs_create(adapter);
3135 status = be_rx_cqs_create(adapter);
3139 status = be_mcc_queues_create(adapter);
3143 be_cmd_get_fn_privileges(adapter, &adapter->cmd_privileges, 0);
3144 /* In UMC mode FW does not return right privileges.
3145 * Override with correct privilege equivalent to PF.
3147 if (be_is_mc(adapter))
3148 adapter->cmd_privileges = MAX_PRIVILEGES;
3150 en_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3151 BE_IF_FLAGS_MULTICAST | BE_IF_FLAGS_PASS_L3L4_ERRORS;
3152 if (adapter->function_caps & BE_FUNCTION_CAPS_RSS)
3153 en_flags |= BE_IF_FLAGS_RSS;
3154 en_flags = en_flags & adapter->if_cap_flags;
3155 status = be_cmd_if_create(adapter, adapter->if_cap_flags, en_flags,
3156 &adapter->if_handle, 0);
3160 status = be_mac_setup(adapter);
3164 status = be_tx_qs_create(adapter);
3168 be_cmd_get_fw_ver(adapter, adapter->fw_ver, adapter->fw_on_flash);
3170 if (adapter->vlans_added)
3171 be_vid_config(adapter);
3173 be_set_rx_mode(adapter->netdev);
3175 be_cmd_get_flow_control(adapter, &tx_fc, &rx_fc);
3177 if (rx_fc != adapter->rx_fc || tx_fc != adapter->tx_fc)
3178 be_cmd_set_flow_control(adapter, adapter->tx_fc,
3181 if (be_physfn(adapter)) {
3182 if (adapter->dev_num_vfs)
3183 be_vf_setup(adapter);
3185 dev_warn(dev, "device doesn't support SRIOV\n");
3188 status = be_cmd_get_phy_info(adapter);
3189 if (!status && be_pause_supported(adapter))
3190 adapter->phy.fc_autoneg = 1;
3192 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
3193 adapter->flags |= BE_FLAGS_WORKER_SCHEDULED;
3200 #ifdef CONFIG_NET_POLL_CONTROLLER
3201 static void be_netpoll(struct net_device *netdev)
3203 struct be_adapter *adapter = netdev_priv(netdev);
3204 struct be_eq_obj *eqo;
3207 for_all_evt_queues(adapter, eqo, i) {
3208 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
3209 napi_schedule(&eqo->napi);
3216 #define FW_FILE_HDR_SIGN "ServerEngines Corp. "
3217 static char flash_cookie[2][16] = {"*** SE FLAS", "H DIRECTORY *** "};
3219 static bool be_flash_redboot(struct be_adapter *adapter,
3220 const u8 *p, u32 img_start, int image_size,
3227 crc_offset = hdr_size + img_start + image_size - 4;
3231 status = be_cmd_get_flash_crc(adapter, flashed_crc,
3234 dev_err(&adapter->pdev->dev,
3235 "could not get crc from flash, not flashing redboot\n");
3239 /*update redboot only if crc does not match*/
3240 if (!memcmp(flashed_crc, p, 4))
3246 static bool phy_flashing_required(struct be_adapter *adapter)
3248 return (adapter->phy.phy_type == TN_8022 &&
3249 adapter->phy.interface_type == PHY_TYPE_BASET_10GB);
3252 static bool is_comp_in_ufi(struct be_adapter *adapter,
3253 struct flash_section_info *fsec, int type)
3255 int i = 0, img_type = 0;
3256 struct flash_section_info_g2 *fsec_g2 = NULL;
3258 if (BE2_chip(adapter))
3259 fsec_g2 = (struct flash_section_info_g2 *)fsec;
3261 for (i = 0; i < MAX_FLASH_COMP; i++) {
3263 img_type = le32_to_cpu(fsec_g2->fsec_entry[i].type);
3265 img_type = le32_to_cpu(fsec->fsec_entry[i].type);
3267 if (img_type == type)
3274 static struct flash_section_info *get_fsec_info(struct be_adapter *adapter,
3276 const struct firmware *fw)
3278 struct flash_section_info *fsec = NULL;
3279 const u8 *p = fw->data;
3282 while (p < (fw->data + fw->size)) {
3283 fsec = (struct flash_section_info *)p;
3284 if (!memcmp(flash_cookie, fsec->cookie, sizeof(flash_cookie)))
3291 static int be_flash(struct be_adapter *adapter, const u8 *img,
3292 struct be_dma_mem *flash_cmd, int optype, int img_size)
3294 u32 total_bytes = 0, flash_op, num_bytes = 0;
3296 struct be_cmd_write_flashrom *req = flash_cmd->va;
3298 total_bytes = img_size;
3299 while (total_bytes) {
3300 num_bytes = min_t(u32, 32*1024, total_bytes);
3302 total_bytes -= num_bytes;
3305 if (optype == OPTYPE_PHY_FW)
3306 flash_op = FLASHROM_OPER_PHY_FLASH;
3308 flash_op = FLASHROM_OPER_FLASH;
3310 if (optype == OPTYPE_PHY_FW)
3311 flash_op = FLASHROM_OPER_PHY_SAVE;
3313 flash_op = FLASHROM_OPER_SAVE;
3316 memcpy(req->data_buf, img, num_bytes);
3318 status = be_cmd_write_flashrom(adapter, flash_cmd, optype,
3319 flash_op, num_bytes);
3321 if (status == ILLEGAL_IOCTL_REQ &&
3322 optype == OPTYPE_PHY_FW)
3324 dev_err(&adapter->pdev->dev,
3325 "cmd to write to flash rom failed.\n");
3332 /* For BE2, BE3 and BE3-R */
3333 static int be_flash_BEx(struct be_adapter *adapter,
3334 const struct firmware *fw,
3335 struct be_dma_mem *flash_cmd,
3339 int status = 0, i, filehdr_size = 0;
3340 int img_hdrs_size = (num_of_images * sizeof(struct image_hdr));
3341 const u8 *p = fw->data;
3342 const struct flash_comp *pflashcomp;
3343 int num_comp, redboot;
3344 struct flash_section_info *fsec = NULL;
3346 struct flash_comp gen3_flash_types[] = {
3347 { FLASH_iSCSI_PRIMARY_IMAGE_START_g3, OPTYPE_ISCSI_ACTIVE,
3348 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_iSCSI},
3349 { FLASH_REDBOOT_START_g3, OPTYPE_REDBOOT,
3350 FLASH_REDBOOT_IMAGE_MAX_SIZE_g3, IMAGE_BOOT_CODE},
3351 { FLASH_iSCSI_BIOS_START_g3, OPTYPE_BIOS,
3352 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_ISCSI},
3353 { FLASH_PXE_BIOS_START_g3, OPTYPE_PXE_BIOS,
3354 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_PXE},
3355 { FLASH_FCoE_BIOS_START_g3, OPTYPE_FCOE_BIOS,
3356 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_FCoE},
3357 { FLASH_iSCSI_BACKUP_IMAGE_START_g3, OPTYPE_ISCSI_BACKUP,
3358 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_iSCSI},
3359 { FLASH_FCoE_PRIMARY_IMAGE_START_g3, OPTYPE_FCOE_FW_ACTIVE,
3360 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_FCoE},
3361 { FLASH_FCoE_BACKUP_IMAGE_START_g3, OPTYPE_FCOE_FW_BACKUP,
3362 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_FCoE},
3363 { FLASH_NCSI_START_g3, OPTYPE_NCSI_FW,
3364 FLASH_NCSI_IMAGE_MAX_SIZE_g3, IMAGE_NCSI},
3365 { FLASH_PHY_FW_START_g3, OPTYPE_PHY_FW,
3366 FLASH_PHY_FW_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_PHY}
3369 struct flash_comp gen2_flash_types[] = {
3370 { FLASH_iSCSI_PRIMARY_IMAGE_START_g2, OPTYPE_ISCSI_ACTIVE,
3371 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_iSCSI},
3372 { FLASH_REDBOOT_START_g2, OPTYPE_REDBOOT,
3373 FLASH_REDBOOT_IMAGE_MAX_SIZE_g2, IMAGE_BOOT_CODE},
3374 { FLASH_iSCSI_BIOS_START_g2, OPTYPE_BIOS,
3375 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_ISCSI},
3376 { FLASH_PXE_BIOS_START_g2, OPTYPE_PXE_BIOS,
3377 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_PXE},
3378 { FLASH_FCoE_BIOS_START_g2, OPTYPE_FCOE_BIOS,
3379 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_FCoE},
3380 { FLASH_iSCSI_BACKUP_IMAGE_START_g2, OPTYPE_ISCSI_BACKUP,
3381 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_iSCSI},
3382 { FLASH_FCoE_PRIMARY_IMAGE_START_g2, OPTYPE_FCOE_FW_ACTIVE,
3383 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_FCoE},
3384 { FLASH_FCoE_BACKUP_IMAGE_START_g2, OPTYPE_FCOE_FW_BACKUP,
3385 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_FCoE}
3388 if (BE3_chip(adapter)) {
3389 pflashcomp = gen3_flash_types;
3390 filehdr_size = sizeof(struct flash_file_hdr_g3);
3391 num_comp = ARRAY_SIZE(gen3_flash_types);
3393 pflashcomp = gen2_flash_types;
3394 filehdr_size = sizeof(struct flash_file_hdr_g2);
3395 num_comp = ARRAY_SIZE(gen2_flash_types);
3398 /* Get flash section info*/
3399 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
3401 dev_err(&adapter->pdev->dev,
3402 "Invalid Cookie. UFI corrupted ?\n");
3405 for (i = 0; i < num_comp; i++) {
3406 if (!is_comp_in_ufi(adapter, fsec, pflashcomp[i].img_type))
3409 if ((pflashcomp[i].optype == OPTYPE_NCSI_FW) &&
3410 memcmp(adapter->fw_ver, "3.102.148.0", 11) < 0)
3413 if (pflashcomp[i].optype == OPTYPE_PHY_FW &&
3414 !phy_flashing_required(adapter))
3417 if (pflashcomp[i].optype == OPTYPE_REDBOOT) {
3418 redboot = be_flash_redboot(adapter, fw->data,
3419 pflashcomp[i].offset, pflashcomp[i].size,
3420 filehdr_size + img_hdrs_size);
3426 p += filehdr_size + pflashcomp[i].offset + img_hdrs_size;
3427 if (p + pflashcomp[i].size > fw->data + fw->size)
3430 status = be_flash(adapter, p, flash_cmd, pflashcomp[i].optype,
3431 pflashcomp[i].size);
3433 dev_err(&adapter->pdev->dev,
3434 "Flashing section type %d failed.\n",
3435 pflashcomp[i].img_type);
3442 static int be_flash_skyhawk(struct be_adapter *adapter,
3443 const struct firmware *fw,
3444 struct be_dma_mem *flash_cmd, int num_of_images)
3446 int status = 0, i, filehdr_size = 0;
3447 int img_offset, img_size, img_optype, redboot;
3448 int img_hdrs_size = num_of_images * sizeof(struct image_hdr);
3449 const u8 *p = fw->data;
3450 struct flash_section_info *fsec = NULL;
3452 filehdr_size = sizeof(struct flash_file_hdr_g3);
3453 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
3455 dev_err(&adapter->pdev->dev,
3456 "Invalid Cookie. UFI corrupted ?\n");
3460 for (i = 0; i < le32_to_cpu(fsec->fsec_hdr.num_images); i++) {
3461 img_offset = le32_to_cpu(fsec->fsec_entry[i].offset);
3462 img_size = le32_to_cpu(fsec->fsec_entry[i].pad_size);
3464 switch (le32_to_cpu(fsec->fsec_entry[i].type)) {
3465 case IMAGE_FIRMWARE_iSCSI:
3466 img_optype = OPTYPE_ISCSI_ACTIVE;
3468 case IMAGE_BOOT_CODE:
3469 img_optype = OPTYPE_REDBOOT;
3471 case IMAGE_OPTION_ROM_ISCSI:
3472 img_optype = OPTYPE_BIOS;
3474 case IMAGE_OPTION_ROM_PXE:
3475 img_optype = OPTYPE_PXE_BIOS;
3477 case IMAGE_OPTION_ROM_FCoE:
3478 img_optype = OPTYPE_FCOE_BIOS;
3480 case IMAGE_FIRMWARE_BACKUP_iSCSI:
3481 img_optype = OPTYPE_ISCSI_BACKUP;
3484 img_optype = OPTYPE_NCSI_FW;
3490 if (img_optype == OPTYPE_REDBOOT) {
3491 redboot = be_flash_redboot(adapter, fw->data,
3492 img_offset, img_size,
3493 filehdr_size + img_hdrs_size);
3499 p += filehdr_size + img_offset + img_hdrs_size;
3500 if (p + img_size > fw->data + fw->size)
3503 status = be_flash(adapter, p, flash_cmd, img_optype, img_size);
3505 dev_err(&adapter->pdev->dev,
3506 "Flashing section type %d failed.\n",
3507 fsec->fsec_entry[i].type);
3514 static int lancer_fw_download(struct be_adapter *adapter,
3515 const struct firmware *fw)
3517 #define LANCER_FW_DOWNLOAD_CHUNK (32 * 1024)
3518 #define LANCER_FW_DOWNLOAD_LOCATION "/prg"
3519 struct be_dma_mem flash_cmd;
3520 const u8 *data_ptr = NULL;
3521 u8 *dest_image_ptr = NULL;
3522 size_t image_size = 0;
3524 u32 data_written = 0;
3530 if (!IS_ALIGNED(fw->size, sizeof(u32))) {
3531 dev_err(&adapter->pdev->dev,
3532 "FW Image not properly aligned. "
3533 "Length must be 4 byte aligned.\n");
3535 goto lancer_fw_exit;
3538 flash_cmd.size = sizeof(struct lancer_cmd_req_write_object)
3539 + LANCER_FW_DOWNLOAD_CHUNK;
3540 flash_cmd.va = dma_alloc_coherent(&adapter->pdev->dev, flash_cmd.size,
3541 &flash_cmd.dma, GFP_KERNEL);
3542 if (!flash_cmd.va) {
3544 goto lancer_fw_exit;
3547 dest_image_ptr = flash_cmd.va +
3548 sizeof(struct lancer_cmd_req_write_object);
3549 image_size = fw->size;
3550 data_ptr = fw->data;
3552 while (image_size) {
3553 chunk_size = min_t(u32, image_size, LANCER_FW_DOWNLOAD_CHUNK);
3555 /* Copy the image chunk content. */
3556 memcpy(dest_image_ptr, data_ptr, chunk_size);
3558 status = lancer_cmd_write_object(adapter, &flash_cmd,
3560 LANCER_FW_DOWNLOAD_LOCATION,
3561 &data_written, &change_status,
3566 offset += data_written;
3567 data_ptr += data_written;
3568 image_size -= data_written;
3572 /* Commit the FW written */
3573 status = lancer_cmd_write_object(adapter, &flash_cmd,
3575 LANCER_FW_DOWNLOAD_LOCATION,
3576 &data_written, &change_status,
3580 dma_free_coherent(&adapter->pdev->dev, flash_cmd.size, flash_cmd.va,
3583 dev_err(&adapter->pdev->dev,
3584 "Firmware load error. "
3585 "Status code: 0x%x Additional Status: 0x%x\n",
3586 status, add_status);
3587 goto lancer_fw_exit;
3590 if (change_status == LANCER_FW_RESET_NEEDED) {
3591 status = lancer_physdev_ctrl(adapter,
3592 PHYSDEV_CONTROL_FW_RESET_MASK);
3594 dev_err(&adapter->pdev->dev,
3595 "Adapter busy for FW reset.\n"
3596 "New FW will not be active.\n");
3597 goto lancer_fw_exit;
3599 } else if (change_status != LANCER_NO_RESET_NEEDED) {
3600 dev_err(&adapter->pdev->dev,
3601 "System reboot required for new FW"
3605 dev_info(&adapter->pdev->dev, "Firmware flashed successfully\n");
3612 #define UFI_TYPE3R 10
3614 static int be_get_ufi_type(struct be_adapter *adapter,
3615 struct flash_file_hdr_g3 *fhdr)
3618 goto be_get_ufi_exit;
3620 if (skyhawk_chip(adapter) && fhdr->build[0] == '4')
3622 else if (BE3_chip(adapter) && fhdr->build[0] == '3') {
3623 if (fhdr->asic_type_rev == 0x10)
3627 } else if (BE2_chip(adapter) && fhdr->build[0] == '2')
3631 dev_err(&adapter->pdev->dev,
3632 "UFI and Interface are not compatible for flashing\n");
3636 static int be_fw_download(struct be_adapter *adapter, const struct firmware* fw)
3638 struct flash_file_hdr_g3 *fhdr3;
3639 struct image_hdr *img_hdr_ptr = NULL;
3640 struct be_dma_mem flash_cmd;
3642 int status = 0, i = 0, num_imgs = 0, ufi_type = 0;
3644 flash_cmd.size = sizeof(struct be_cmd_write_flashrom);
3645 flash_cmd.va = dma_alloc_coherent(&adapter->pdev->dev, flash_cmd.size,
3646 &flash_cmd.dma, GFP_KERNEL);
3647 if (!flash_cmd.va) {
3653 fhdr3 = (struct flash_file_hdr_g3 *)p;
3655 ufi_type = be_get_ufi_type(adapter, fhdr3);
3657 num_imgs = le32_to_cpu(fhdr3->num_imgs);
3658 for (i = 0; i < num_imgs; i++) {
3659 img_hdr_ptr = (struct image_hdr *)(fw->data +
3660 (sizeof(struct flash_file_hdr_g3) +
3661 i * sizeof(struct image_hdr)));
3662 if (le32_to_cpu(img_hdr_ptr->imageid) == 1) {
3665 status = be_flash_skyhawk(adapter, fw,
3666 &flash_cmd, num_imgs);
3669 status = be_flash_BEx(adapter, fw, &flash_cmd,
3673 /* Do not flash this ufi on BE3-R cards */
3674 if (adapter->asic_rev < 0x10)
3675 status = be_flash_BEx(adapter, fw,
3680 dev_err(&adapter->pdev->dev,
3681 "Can't load BE3 UFI on BE3R\n");
3687 if (ufi_type == UFI_TYPE2)
3688 status = be_flash_BEx(adapter, fw, &flash_cmd, 0);
3689 else if (ufi_type == -1)
3692 dma_free_coherent(&adapter->pdev->dev, flash_cmd.size, flash_cmd.va,
3695 dev_err(&adapter->pdev->dev, "Firmware load error\n");
3699 dev_info(&adapter->pdev->dev, "Firmware flashed successfully\n");
3705 int be_load_fw(struct be_adapter *adapter, u8 *fw_file)
3707 const struct firmware *fw;
3710 if (!netif_running(adapter->netdev)) {
3711 dev_err(&adapter->pdev->dev,
3712 "Firmware load not allowed (interface is down)\n");
3716 status = request_firmware(&fw, fw_file, &adapter->pdev->dev);
3720 dev_info(&adapter->pdev->dev, "Flashing firmware file %s\n", fw_file);
3722 if (lancer_chip(adapter))
3723 status = lancer_fw_download(adapter, fw);
3725 status = be_fw_download(adapter, fw);
3728 be_cmd_get_fw_ver(adapter, adapter->fw_ver,
3729 adapter->fw_on_flash);
3732 release_firmware(fw);
3736 static const struct net_device_ops be_netdev_ops = {
3737 .ndo_open = be_open,
3738 .ndo_stop = be_close,
3739 .ndo_start_xmit = be_xmit,
3740 .ndo_set_rx_mode = be_set_rx_mode,
3741 .ndo_set_mac_address = be_mac_addr_set,
3742 .ndo_change_mtu = be_change_mtu,
3743 .ndo_get_stats64 = be_get_stats64,
3744 .ndo_validate_addr = eth_validate_addr,
3745 .ndo_vlan_rx_add_vid = be_vlan_add_vid,
3746 .ndo_vlan_rx_kill_vid = be_vlan_rem_vid,
3747 .ndo_set_vf_mac = be_set_vf_mac,
3748 .ndo_set_vf_vlan = be_set_vf_vlan,
3749 .ndo_set_vf_tx_rate = be_set_vf_tx_rate,
3750 .ndo_get_vf_config = be_get_vf_config,
3751 #ifdef CONFIG_NET_POLL_CONTROLLER
3752 .ndo_poll_controller = be_netpoll,
3756 static void be_netdev_init(struct net_device *netdev)
3758 struct be_adapter *adapter = netdev_priv(netdev);
3759 struct be_eq_obj *eqo;
3762 netdev->hw_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
3763 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM | NETIF_F_RXCSUM |
3764 NETIF_F_HW_VLAN_CTAG_TX;
3765 if (be_multi_rxq(adapter))
3766 netdev->hw_features |= NETIF_F_RXHASH;
3768 netdev->features |= netdev->hw_features |
3769 NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_FILTER;
3771 netdev->vlan_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
3772 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
3774 netdev->priv_flags |= IFF_UNICAST_FLT;
3776 netdev->flags |= IFF_MULTICAST;
3778 netif_set_gso_max_size(netdev, 65535 - ETH_HLEN);
3780 netdev->netdev_ops = &be_netdev_ops;
3782 SET_ETHTOOL_OPS(netdev, &be_ethtool_ops);
3784 for_all_evt_queues(adapter, eqo, i)
3785 netif_napi_add(netdev, &eqo->napi, be_poll, BE_NAPI_WEIGHT);
3788 static void be_unmap_pci_bars(struct be_adapter *adapter)
3791 pci_iounmap(adapter->pdev, adapter->csr);
3793 pci_iounmap(adapter->pdev, adapter->db);
3796 static int db_bar(struct be_adapter *adapter)
3798 if (lancer_chip(adapter) || !be_physfn(adapter))
3804 static int be_roce_map_pci_bars(struct be_adapter *adapter)
3806 if (skyhawk_chip(adapter)) {
3807 adapter->roce_db.size = 4096;
3808 adapter->roce_db.io_addr = pci_resource_start(adapter->pdev,
3810 adapter->roce_db.total_size = pci_resource_len(adapter->pdev,
3816 static int be_map_pci_bars(struct be_adapter *adapter)
3821 pci_read_config_dword(adapter->pdev, SLI_INTF_REG_OFFSET, &sli_intf);
3822 adapter->if_type = (sli_intf & SLI_INTF_IF_TYPE_MASK) >>
3823 SLI_INTF_IF_TYPE_SHIFT;
3825 if (BEx_chip(adapter) && be_physfn(adapter)) {
3826 adapter->csr = pci_iomap(adapter->pdev, 2, 0);
3827 if (adapter->csr == NULL)
3831 addr = pci_iomap(adapter->pdev, db_bar(adapter), 0);
3836 be_roce_map_pci_bars(adapter);
3840 be_unmap_pci_bars(adapter);
3844 static void be_ctrl_cleanup(struct be_adapter *adapter)
3846 struct be_dma_mem *mem = &adapter->mbox_mem_alloced;
3848 be_unmap_pci_bars(adapter);
3851 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
3854 mem = &adapter->rx_filter;
3856 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
3860 static int be_ctrl_init(struct be_adapter *adapter)
3862 struct be_dma_mem *mbox_mem_alloc = &adapter->mbox_mem_alloced;
3863 struct be_dma_mem *mbox_mem_align = &adapter->mbox_mem;
3864 struct be_dma_mem *rx_filter = &adapter->rx_filter;
3868 pci_read_config_dword(adapter->pdev, SLI_INTF_REG_OFFSET, &sli_intf);
3869 adapter->sli_family = (sli_intf & SLI_INTF_FAMILY_MASK) >>
3870 SLI_INTF_FAMILY_SHIFT;
3871 adapter->virtfn = (sli_intf & SLI_INTF_FT_MASK) ? 1 : 0;
3873 status = be_map_pci_bars(adapter);
3877 mbox_mem_alloc->size = sizeof(struct be_mcc_mailbox) + 16;
3878 mbox_mem_alloc->va = dma_alloc_coherent(&adapter->pdev->dev,
3879 mbox_mem_alloc->size,
3880 &mbox_mem_alloc->dma,
3882 if (!mbox_mem_alloc->va) {
3884 goto unmap_pci_bars;
3886 mbox_mem_align->size = sizeof(struct be_mcc_mailbox);
3887 mbox_mem_align->va = PTR_ALIGN(mbox_mem_alloc->va, 16);
3888 mbox_mem_align->dma = PTR_ALIGN(mbox_mem_alloc->dma, 16);
3889 memset(mbox_mem_align->va, 0, sizeof(struct be_mcc_mailbox));
3891 rx_filter->size = sizeof(struct be_cmd_req_rx_filter);
3892 rx_filter->va = dma_alloc_coherent(&adapter->pdev->dev, rx_filter->size,
3894 GFP_KERNEL | __GFP_ZERO);
3895 if (rx_filter->va == NULL) {
3900 mutex_init(&adapter->mbox_lock);
3901 spin_lock_init(&adapter->mcc_lock);
3902 spin_lock_init(&adapter->mcc_cq_lock);
3904 init_completion(&adapter->flash_compl);
3905 pci_save_state(adapter->pdev);
3909 dma_free_coherent(&adapter->pdev->dev, mbox_mem_alloc->size,
3910 mbox_mem_alloc->va, mbox_mem_alloc->dma);
3913 be_unmap_pci_bars(adapter);
3919 static void be_stats_cleanup(struct be_adapter *adapter)
3921 struct be_dma_mem *cmd = &adapter->stats_cmd;
3924 dma_free_coherent(&adapter->pdev->dev, cmd->size,
3928 static int be_stats_init(struct be_adapter *adapter)
3930 struct be_dma_mem *cmd = &adapter->stats_cmd;
3932 if (lancer_chip(adapter))
3933 cmd->size = sizeof(struct lancer_cmd_req_pport_stats);
3934 else if (BE2_chip(adapter))
3935 cmd->size = sizeof(struct be_cmd_req_get_stats_v0);
3937 /* BE3 and Skyhawk */
3938 cmd->size = sizeof(struct be_cmd_req_get_stats_v1);
3940 cmd->va = dma_alloc_coherent(&adapter->pdev->dev, cmd->size, &cmd->dma,
3941 GFP_KERNEL | __GFP_ZERO);
3942 if (cmd->va == NULL)
3947 static void be_remove(struct pci_dev *pdev)
3949 struct be_adapter *adapter = pci_get_drvdata(pdev);
3954 be_roce_dev_remove(adapter);
3955 be_intr_set(adapter, false);
3957 cancel_delayed_work_sync(&adapter->func_recovery_work);
3959 unregister_netdev(adapter->netdev);
3963 /* tell fw we're done with firing cmds */
3964 be_cmd_fw_clean(adapter);
3966 be_stats_cleanup(adapter);
3968 be_ctrl_cleanup(adapter);
3970 pci_disable_pcie_error_reporting(pdev);
3972 pci_set_drvdata(pdev, NULL);
3973 pci_release_regions(pdev);
3974 pci_disable_device(pdev);
3976 free_netdev(adapter->netdev);
3979 bool be_is_wol_supported(struct be_adapter *adapter)
3981 return ((adapter->wol_cap & BE_WOL_CAP) &&
3982 !be_is_wol_excluded(adapter)) ? true : false;
3985 u32 be_get_fw_log_level(struct be_adapter *adapter)
3987 struct be_dma_mem extfat_cmd;
3988 struct be_fat_conf_params *cfgs;
3993 if (lancer_chip(adapter))
3996 memset(&extfat_cmd, 0, sizeof(struct be_dma_mem));
3997 extfat_cmd.size = sizeof(struct be_cmd_resp_get_ext_fat_caps);
3998 extfat_cmd.va = pci_alloc_consistent(adapter->pdev, extfat_cmd.size,
4001 if (!extfat_cmd.va) {
4002 dev_err(&adapter->pdev->dev, "%s: Memory allocation failure\n",
4007 status = be_cmd_get_ext_fat_capabilites(adapter, &extfat_cmd);
4009 cfgs = (struct be_fat_conf_params *)(extfat_cmd.va +
4010 sizeof(struct be_cmd_resp_hdr));
4011 for (j = 0; j < le32_to_cpu(cfgs->module[0].num_modes); j++) {
4012 if (cfgs->module[0].trace_lvl[j].mode == MODE_UART)
4013 level = cfgs->module[0].trace_lvl[j].dbg_lvl;
4016 pci_free_consistent(adapter->pdev, extfat_cmd.size, extfat_cmd.va,
4022 static int be_get_initial_config(struct be_adapter *adapter)
4027 status = be_cmd_get_cntl_attributes(adapter);
4031 status = be_cmd_get_acpi_wol_cap(adapter);
4033 /* in case of a failure to get wol capabillities
4034 * check the exclusion list to determine WOL capability */
4035 if (!be_is_wol_excluded(adapter))
4036 adapter->wol_cap |= BE_WOL_CAP;
4039 if (be_is_wol_supported(adapter))
4040 adapter->wol = true;
4042 /* Must be a power of 2 or else MODULO will BUG_ON */
4043 adapter->be_get_temp_freq = 64;
4045 level = be_get_fw_log_level(adapter);
4046 adapter->msg_enable = level <= FW_LOG_LEVEL_DEFAULT ? NETIF_MSG_HW : 0;
4051 static int lancer_recover_func(struct be_adapter *adapter)
4053 struct device *dev = &adapter->pdev->dev;
4056 status = lancer_test_and_set_rdy_state(adapter);
4060 if (netif_running(adapter->netdev))
4061 be_close(adapter->netdev);
4065 be_clear_all_error(adapter);
4067 status = be_setup(adapter);
4071 if (netif_running(adapter->netdev)) {
4072 status = be_open(adapter->netdev);
4077 dev_err(dev, "Error recovery successful\n");
4080 if (status == -EAGAIN)
4081 dev_err(dev, "Waiting for resource provisioning\n");
4083 dev_err(dev, "Error recovery failed\n");
4088 static void be_func_recovery_task(struct work_struct *work)
4090 struct be_adapter *adapter =
4091 container_of(work, struct be_adapter, func_recovery_work.work);
4094 be_detect_error(adapter);
4096 if (adapter->hw_error && lancer_chip(adapter)) {
4099 netif_device_detach(adapter->netdev);
4102 status = lancer_recover_func(adapter);
4104 netif_device_attach(adapter->netdev);
4107 /* In Lancer, for all errors other than provisioning error (-EAGAIN),
4108 * no need to attempt further recovery.
4110 if (!status || status == -EAGAIN)
4111 schedule_delayed_work(&adapter->func_recovery_work,
4112 msecs_to_jiffies(1000));
4115 static void be_worker(struct work_struct *work)
4117 struct be_adapter *adapter =
4118 container_of(work, struct be_adapter, work.work);
4119 struct be_rx_obj *rxo;
4120 struct be_eq_obj *eqo;
4123 /* when interrupts are not yet enabled, just reap any pending
4124 * mcc completions */
4125 if (!netif_running(adapter->netdev)) {
4127 be_process_mcc(adapter);
4132 if (!adapter->stats_cmd_sent) {
4133 if (lancer_chip(adapter))
4134 lancer_cmd_get_pport_stats(adapter,
4135 &adapter->stats_cmd);
4137 be_cmd_get_stats(adapter, &adapter->stats_cmd);
4140 if (be_physfn(adapter) &&
4141 MODULO(adapter->work_counter, adapter->be_get_temp_freq) == 0)
4142 be_cmd_get_die_temperature(adapter);
4144 for_all_rx_queues(adapter, rxo, i) {
4145 if (rxo->rx_post_starved) {
4146 rxo->rx_post_starved = false;
4147 be_post_rx_frags(rxo, GFP_KERNEL);
4151 for_all_evt_queues(adapter, eqo, i)
4152 be_eqd_update(adapter, eqo);
4155 adapter->work_counter++;
4156 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
4159 /* If any VFs are already enabled don't FLR the PF */
4160 static bool be_reset_required(struct be_adapter *adapter)
4162 return pci_num_vf(adapter->pdev) ? false : true;
4165 static char *mc_name(struct be_adapter *adapter)
4167 if (adapter->function_mode & FLEX10_MODE)
4169 else if (adapter->function_mode & VNIC_MODE)
4171 else if (adapter->function_mode & UMC_ENABLED)
4177 static inline char *func_name(struct be_adapter *adapter)
4179 return be_physfn(adapter) ? "PF" : "VF";
4182 static int be_probe(struct pci_dev *pdev, const struct pci_device_id *pdev_id)
4185 struct be_adapter *adapter;
4186 struct net_device *netdev;
4189 status = pci_enable_device(pdev);
4193 status = pci_request_regions(pdev, DRV_NAME);
4196 pci_set_master(pdev);
4198 netdev = alloc_etherdev_mqs(sizeof(*adapter), MAX_TX_QS, MAX_RX_QS);
4199 if (netdev == NULL) {
4203 adapter = netdev_priv(netdev);
4204 adapter->pdev = pdev;
4205 pci_set_drvdata(pdev, adapter);
4206 adapter->netdev = netdev;
4207 SET_NETDEV_DEV(netdev, &pdev->dev);
4209 status = dma_set_mask(&pdev->dev, DMA_BIT_MASK(64));
4211 status = dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(64));
4213 dev_err(&pdev->dev, "dma_set_coherent_mask failed\n");
4216 netdev->features |= NETIF_F_HIGHDMA;
4218 status = dma_set_mask(&pdev->dev, DMA_BIT_MASK(32));
4220 status = dma_set_coherent_mask(&pdev->dev,
4223 dev_err(&pdev->dev, "Could not set PCI DMA Mask\n");
4228 status = pci_enable_pcie_error_reporting(pdev);
4230 dev_info(&pdev->dev, "Could not use PCIe error reporting\n");
4232 status = be_ctrl_init(adapter);
4236 /* sync up with fw's ready state */
4237 if (be_physfn(adapter)) {
4238 status = be_fw_wait_ready(adapter);
4243 if (be_reset_required(adapter)) {
4244 status = be_cmd_reset_function(adapter);
4248 /* Wait for interrupts to quiesce after an FLR */
4252 /* Allow interrupts for other ULPs running on NIC function */
4253 be_intr_set(adapter, true);
4255 /* tell fw we're ready to fire cmds */
4256 status = be_cmd_fw_init(adapter);
4260 status = be_stats_init(adapter);
4264 status = be_get_initial_config(adapter);
4268 INIT_DELAYED_WORK(&adapter->work, be_worker);
4269 INIT_DELAYED_WORK(&adapter->func_recovery_work, be_func_recovery_task);
4270 adapter->rx_fc = adapter->tx_fc = true;
4272 status = be_setup(adapter);
4276 be_netdev_init(netdev);
4277 status = register_netdev(netdev);
4281 be_roce_dev_add(adapter);
4283 schedule_delayed_work(&adapter->func_recovery_work,
4284 msecs_to_jiffies(1000));
4286 be_cmd_query_port_name(adapter, &port_name);
4288 dev_info(&pdev->dev, "%s: %s %s port %c\n", nic_name(pdev),
4289 func_name(adapter), mc_name(adapter), port_name);
4296 be_stats_cleanup(adapter);
4298 be_ctrl_cleanup(adapter);
4300 free_netdev(netdev);
4301 pci_set_drvdata(pdev, NULL);
4303 pci_release_regions(pdev);
4305 pci_disable_device(pdev);
4307 dev_err(&pdev->dev, "%s initialization failed\n", nic_name(pdev));
4311 static int be_suspend(struct pci_dev *pdev, pm_message_t state)
4313 struct be_adapter *adapter = pci_get_drvdata(pdev);
4314 struct net_device *netdev = adapter->netdev;
4317 be_setup_wol(adapter, true);
4319 cancel_delayed_work_sync(&adapter->func_recovery_work);
4321 netif_device_detach(netdev);
4322 if (netif_running(netdev)) {
4329 pci_save_state(pdev);
4330 pci_disable_device(pdev);
4331 pci_set_power_state(pdev, pci_choose_state(pdev, state));
4335 static int be_resume(struct pci_dev *pdev)
4338 struct be_adapter *adapter = pci_get_drvdata(pdev);
4339 struct net_device *netdev = adapter->netdev;
4341 netif_device_detach(netdev);
4343 status = pci_enable_device(pdev);
4347 pci_set_power_state(pdev, PCI_D0);
4348 pci_restore_state(pdev);
4350 /* tell fw we're ready to fire cmds */
4351 status = be_cmd_fw_init(adapter);
4356 if (netif_running(netdev)) {
4362 schedule_delayed_work(&adapter->func_recovery_work,
4363 msecs_to_jiffies(1000));
4364 netif_device_attach(netdev);
4367 be_setup_wol(adapter, false);
4373 * An FLR will stop BE from DMAing any data.
4375 static void be_shutdown(struct pci_dev *pdev)
4377 struct be_adapter *adapter = pci_get_drvdata(pdev);
4382 cancel_delayed_work_sync(&adapter->work);
4383 cancel_delayed_work_sync(&adapter->func_recovery_work);
4385 netif_device_detach(adapter->netdev);
4387 be_cmd_reset_function(adapter);
4389 pci_disable_device(pdev);
4392 static pci_ers_result_t be_eeh_err_detected(struct pci_dev *pdev,
4393 pci_channel_state_t state)
4395 struct be_adapter *adapter = pci_get_drvdata(pdev);
4396 struct net_device *netdev = adapter->netdev;
4398 dev_err(&adapter->pdev->dev, "EEH error detected\n");
4400 if (!adapter->eeh_error) {
4401 adapter->eeh_error = true;
4403 cancel_delayed_work_sync(&adapter->func_recovery_work);
4406 netif_device_detach(netdev);
4407 if (netif_running(netdev))
4414 if (state == pci_channel_io_perm_failure)
4415 return PCI_ERS_RESULT_DISCONNECT;
4417 pci_disable_device(pdev);
4419 /* The error could cause the FW to trigger a flash debug dump.
4420 * Resetting the card while flash dump is in progress
4421 * can cause it not to recover; wait for it to finish.
4422 * Wait only for first function as it is needed only once per
4425 if (pdev->devfn == 0)
4428 return PCI_ERS_RESULT_NEED_RESET;
4431 static pci_ers_result_t be_eeh_reset(struct pci_dev *pdev)
4433 struct be_adapter *adapter = pci_get_drvdata(pdev);
4436 dev_info(&adapter->pdev->dev, "EEH reset\n");
4438 status = pci_enable_device(pdev);
4440 return PCI_ERS_RESULT_DISCONNECT;
4442 pci_set_master(pdev);
4443 pci_set_power_state(pdev, PCI_D0);
4444 pci_restore_state(pdev);
4446 /* Check if card is ok and fw is ready */
4447 dev_info(&adapter->pdev->dev,
4448 "Waiting for FW to be ready after EEH reset\n");
4449 status = be_fw_wait_ready(adapter);
4451 return PCI_ERS_RESULT_DISCONNECT;
4453 pci_cleanup_aer_uncorrect_error_status(pdev);
4454 be_clear_all_error(adapter);
4455 return PCI_ERS_RESULT_RECOVERED;
4458 static void be_eeh_resume(struct pci_dev *pdev)
4461 struct be_adapter *adapter = pci_get_drvdata(pdev);
4462 struct net_device *netdev = adapter->netdev;
4464 dev_info(&adapter->pdev->dev, "EEH resume\n");
4466 pci_save_state(pdev);
4468 status = be_cmd_reset_function(adapter);
4472 /* tell fw we're ready to fire cmds */
4473 status = be_cmd_fw_init(adapter);
4477 status = be_setup(adapter);
4481 if (netif_running(netdev)) {
4482 status = be_open(netdev);
4487 schedule_delayed_work(&adapter->func_recovery_work,
4488 msecs_to_jiffies(1000));
4489 netif_device_attach(netdev);
4492 dev_err(&adapter->pdev->dev, "EEH resume failed\n");
4495 static const struct pci_error_handlers be_eeh_handlers = {
4496 .error_detected = be_eeh_err_detected,
4497 .slot_reset = be_eeh_reset,
4498 .resume = be_eeh_resume,
4501 static struct pci_driver be_driver = {
4503 .id_table = be_dev_ids,
4505 .remove = be_remove,
4506 .suspend = be_suspend,
4507 .resume = be_resume,
4508 .shutdown = be_shutdown,
4509 .err_handler = &be_eeh_handlers
4512 static int __init be_init_module(void)
4514 if (rx_frag_size != 8192 && rx_frag_size != 4096 &&
4515 rx_frag_size != 2048) {
4516 printk(KERN_WARNING DRV_NAME
4517 " : Module param rx_frag_size must be 2048/4096/8192."
4519 rx_frag_size = 2048;
4522 return pci_register_driver(&be_driver);
4524 module_init(be_init_module);
4526 static void __exit be_exit_module(void)
4528 pci_unregister_driver(&be_driver);
4530 module_exit(be_exit_module);