1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright(c) 1999 - 2018 Intel Corporation. */
4 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
6 #include <linux/module.h>
7 #include <linux/types.h>
8 #include <linux/init.h>
10 #include <linux/vmalloc.h>
11 #include <linux/pagemap.h>
12 #include <linux/delay.h>
13 #include <linux/netdevice.h>
14 #include <linux/interrupt.h>
15 #include <linux/tcp.h>
16 #include <linux/ipv6.h>
17 #include <linux/slab.h>
18 #include <net/checksum.h>
19 #include <net/ip6_checksum.h>
20 #include <linux/ethtool.h>
21 #include <linux/if_vlan.h>
22 #include <linux/cpu.h>
23 #include <linux/smp.h>
24 #include <linux/pm_qos.h>
25 #include <linux/pm_runtime.h>
26 #include <linux/aer.h>
27 #include <linux/prefetch.h>
28 #include <linux/suspend.h>
32 char e1000e_driver_name[] = "e1000e";
34 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
35 static int debug = -1;
36 module_param(debug, int, 0);
37 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
39 static const struct e1000_info *e1000_info_tbl[] = {
40 [board_82571] = &e1000_82571_info,
41 [board_82572] = &e1000_82572_info,
42 [board_82573] = &e1000_82573_info,
43 [board_82574] = &e1000_82574_info,
44 [board_82583] = &e1000_82583_info,
45 [board_80003es2lan] = &e1000_es2_info,
46 [board_ich8lan] = &e1000_ich8_info,
47 [board_ich9lan] = &e1000_ich9_info,
48 [board_ich10lan] = &e1000_ich10_info,
49 [board_pchlan] = &e1000_pch_info,
50 [board_pch2lan] = &e1000_pch2_info,
51 [board_pch_lpt] = &e1000_pch_lpt_info,
52 [board_pch_spt] = &e1000_pch_spt_info,
53 [board_pch_cnp] = &e1000_pch_cnp_info,
54 [board_pch_tgp] = &e1000_pch_tgp_info,
57 struct e1000_reg_info {
62 static const struct e1000_reg_info e1000_reg_info_tbl[] = {
63 /* General Registers */
65 {E1000_STATUS, "STATUS"},
66 {E1000_CTRL_EXT, "CTRL_EXT"},
68 /* Interrupt Registers */
73 {E1000_RDLEN(0), "RDLEN"},
74 {E1000_RDH(0), "RDH"},
75 {E1000_RDT(0), "RDT"},
77 {E1000_RXDCTL(0), "RXDCTL"},
79 {E1000_RDBAL(0), "RDBAL"},
80 {E1000_RDBAH(0), "RDBAH"},
83 {E1000_RDFHS, "RDFHS"},
84 {E1000_RDFTS, "RDFTS"},
85 {E1000_RDFPC, "RDFPC"},
89 {E1000_TDBAL(0), "TDBAL"},
90 {E1000_TDBAH(0), "TDBAH"},
91 {E1000_TDLEN(0), "TDLEN"},
92 {E1000_TDH(0), "TDH"},
93 {E1000_TDT(0), "TDT"},
95 {E1000_TXDCTL(0), "TXDCTL"},
97 {E1000_TARC(0), "TARC"},
100 {E1000_TDFHS, "TDFHS"},
101 {E1000_TDFTS, "TDFTS"},
102 {E1000_TDFPC, "TDFPC"},
104 /* List Terminator */
109 * __ew32_prepare - prepare to write to MAC CSR register on certain parts
110 * @hw: pointer to the HW structure
112 * When updating the MAC CSR registers, the Manageability Engine (ME) could
113 * be accessing the registers at the same time. Normally, this is handled in
114 * h/w by an arbiter but on some parts there is a bug that acknowledges Host
115 * accesses later than it should which could result in the register to have
116 * an incorrect value. Workaround this by checking the FWSM register which
117 * has bit 24 set while ME is accessing MAC CSR registers, wait if it is set
118 * and try again a number of times.
120 static void __ew32_prepare(struct e1000_hw *hw)
122 s32 i = E1000_ICH_FWSM_PCIM2PCI_COUNT;
124 while ((er32(FWSM) & E1000_ICH_FWSM_PCIM2PCI) && --i)
128 void __ew32(struct e1000_hw *hw, unsigned long reg, u32 val)
130 if (hw->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
133 writel(val, hw->hw_addr + reg);
137 * e1000_regdump - register printout routine
138 * @hw: pointer to the HW structure
139 * @reginfo: pointer to the register info table
141 static void e1000_regdump(struct e1000_hw *hw, struct e1000_reg_info *reginfo)
147 switch (reginfo->ofs) {
148 case E1000_RXDCTL(0):
149 for (n = 0; n < 2; n++)
150 regs[n] = __er32(hw, E1000_RXDCTL(n));
152 case E1000_TXDCTL(0):
153 for (n = 0; n < 2; n++)
154 regs[n] = __er32(hw, E1000_TXDCTL(n));
157 for (n = 0; n < 2; n++)
158 regs[n] = __er32(hw, E1000_TARC(n));
161 pr_info("%-15s %08x\n",
162 reginfo->name, __er32(hw, reginfo->ofs));
166 snprintf(rname, 16, "%s%s", reginfo->name, "[0-1]");
167 pr_info("%-15s %08x %08x\n", rname, regs[0], regs[1]);
170 static void e1000e_dump_ps_pages(struct e1000_adapter *adapter,
171 struct e1000_buffer *bi)
174 struct e1000_ps_page *ps_page;
176 for (i = 0; i < adapter->rx_ps_pages; i++) {
177 ps_page = &bi->ps_pages[i];
180 pr_info("packet dump for ps_page %d:\n", i);
181 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
182 16, 1, page_address(ps_page->page),
189 * e1000e_dump - Print registers, Tx-ring and Rx-ring
190 * @adapter: board private structure
192 static void e1000e_dump(struct e1000_adapter *adapter)
194 struct net_device *netdev = adapter->netdev;
195 struct e1000_hw *hw = &adapter->hw;
196 struct e1000_reg_info *reginfo;
197 struct e1000_ring *tx_ring = adapter->tx_ring;
198 struct e1000_tx_desc *tx_desc;
203 struct e1000_buffer *buffer_info;
204 struct e1000_ring *rx_ring = adapter->rx_ring;
205 union e1000_rx_desc_packet_split *rx_desc_ps;
206 union e1000_rx_desc_extended *rx_desc;
216 if (!netif_msg_hw(adapter))
219 /* Print netdevice Info */
221 dev_info(&adapter->pdev->dev, "Net device Info\n");
222 pr_info("Device Name state trans_start\n");
223 pr_info("%-15s %016lX %016lX\n", netdev->name,
224 netdev->state, dev_trans_start(netdev));
227 /* Print Registers */
228 dev_info(&adapter->pdev->dev, "Register Dump\n");
229 pr_info(" Register Name Value\n");
230 for (reginfo = (struct e1000_reg_info *)e1000_reg_info_tbl;
231 reginfo->name; reginfo++) {
232 e1000_regdump(hw, reginfo);
235 /* Print Tx Ring Summary */
236 if (!netdev || !netif_running(netdev))
239 dev_info(&adapter->pdev->dev, "Tx Ring Summary\n");
240 pr_info("Queue [NTU] [NTC] [bi(ntc)->dma ] leng ntw timestamp\n");
241 buffer_info = &tx_ring->buffer_info[tx_ring->next_to_clean];
242 pr_info(" %5d %5X %5X %016llX %04X %3X %016llX\n",
243 0, tx_ring->next_to_use, tx_ring->next_to_clean,
244 (unsigned long long)buffer_info->dma,
246 buffer_info->next_to_watch,
247 (unsigned long long)buffer_info->time_stamp);
250 if (!netif_msg_tx_done(adapter))
251 goto rx_ring_summary;
253 dev_info(&adapter->pdev->dev, "Tx Ring Dump\n");
255 /* Transmit Descriptor Formats - DEXT[29] is 0 (Legacy) or 1 (Extended)
257 * Legacy Transmit Descriptor
258 * +--------------------------------------------------------------+
259 * 0 | Buffer Address [63:0] (Reserved on Write Back) |
260 * +--------------------------------------------------------------+
261 * 8 | Special | CSS | Status | CMD | CSO | Length |
262 * +--------------------------------------------------------------+
263 * 63 48 47 36 35 32 31 24 23 16 15 0
265 * Extended Context Descriptor (DTYP=0x0) for TSO or checksum offload
266 * 63 48 47 40 39 32 31 16 15 8 7 0
267 * +----------------------------------------------------------------+
268 * 0 | TUCSE | TUCS0 | TUCSS | IPCSE | IPCS0 | IPCSS |
269 * +----------------------------------------------------------------+
270 * 8 | MSS | HDRLEN | RSV | STA | TUCMD | DTYP | PAYLEN |
271 * +----------------------------------------------------------------+
272 * 63 48 47 40 39 36 35 32 31 24 23 20 19 0
274 * Extended Data Descriptor (DTYP=0x1)
275 * +----------------------------------------------------------------+
276 * 0 | Buffer Address [63:0] |
277 * +----------------------------------------------------------------+
278 * 8 | VLAN tag | POPTS | Rsvd | Status | Command | DTYP | DTALEN |
279 * +----------------------------------------------------------------+
280 * 63 48 47 40 39 36 35 32 31 24 23 20 19 0
282 pr_info("Tl[desc] [address 63:0 ] [SpeCssSCmCsLen] [bi->dma ] leng ntw timestamp bi->skb <-- Legacy format\n");
283 pr_info("Tc[desc] [Ce CoCsIpceCoS] [MssHlRSCm0Plen] [bi->dma ] leng ntw timestamp bi->skb <-- Ext Context format\n");
284 pr_info("Td[desc] [address 63:0 ] [VlaPoRSCm1Dlen] [bi->dma ] leng ntw timestamp bi->skb <-- Ext Data format\n");
285 for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
286 const char *next_desc;
287 tx_desc = E1000_TX_DESC(*tx_ring, i);
288 buffer_info = &tx_ring->buffer_info[i];
289 u0 = (struct my_u0 *)tx_desc;
290 if (i == tx_ring->next_to_use && i == tx_ring->next_to_clean)
291 next_desc = " NTC/U";
292 else if (i == tx_ring->next_to_use)
294 else if (i == tx_ring->next_to_clean)
298 pr_info("T%c[0x%03X] %016llX %016llX %016llX %04X %3X %016llX %p%s\n",
299 (!(le64_to_cpu(u0->b) & BIT(29)) ? 'l' :
300 ((le64_to_cpu(u0->b) & BIT(20)) ? 'd' : 'c')),
302 (unsigned long long)le64_to_cpu(u0->a),
303 (unsigned long long)le64_to_cpu(u0->b),
304 (unsigned long long)buffer_info->dma,
305 buffer_info->length, buffer_info->next_to_watch,
306 (unsigned long long)buffer_info->time_stamp,
307 buffer_info->skb, next_desc);
309 if (netif_msg_pktdata(adapter) && buffer_info->skb)
310 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
311 16, 1, buffer_info->skb->data,
312 buffer_info->skb->len, true);
315 /* Print Rx Ring Summary */
317 dev_info(&adapter->pdev->dev, "Rx Ring Summary\n");
318 pr_info("Queue [NTU] [NTC]\n");
319 pr_info(" %5d %5X %5X\n",
320 0, rx_ring->next_to_use, rx_ring->next_to_clean);
323 if (!netif_msg_rx_status(adapter))
326 dev_info(&adapter->pdev->dev, "Rx Ring Dump\n");
327 switch (adapter->rx_ps_pages) {
331 /* [Extended] Packet Split Receive Descriptor Format
333 * +-----------------------------------------------------+
334 * 0 | Buffer Address 0 [63:0] |
335 * +-----------------------------------------------------+
336 * 8 | Buffer Address 1 [63:0] |
337 * +-----------------------------------------------------+
338 * 16 | Buffer Address 2 [63:0] |
339 * +-----------------------------------------------------+
340 * 24 | Buffer Address 3 [63:0] |
341 * +-----------------------------------------------------+
343 pr_info("R [desc] [buffer 0 63:0 ] [buffer 1 63:0 ] [buffer 2 63:0 ] [buffer 3 63:0 ] [bi->dma ] [bi->skb] <-- Ext Pkt Split format\n");
344 /* [Extended] Receive Descriptor (Write-Back) Format
346 * 63 48 47 32 31 13 12 8 7 4 3 0
347 * +------------------------------------------------------+
348 * 0 | Packet | IP | Rsvd | MRQ | Rsvd | MRQ RSS |
349 * | Checksum | Ident | | Queue | | Type |
350 * +------------------------------------------------------+
351 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
352 * +------------------------------------------------------+
353 * 63 48 47 32 31 20 19 0
355 pr_info("RWB[desc] [ck ipid mrqhsh] [vl l0 ee es] [ l3 l2 l1 hs] [reserved ] ---------------- [bi->skb] <-- Ext Rx Write-Back format\n");
356 for (i = 0; i < rx_ring->count; i++) {
357 const char *next_desc;
358 buffer_info = &rx_ring->buffer_info[i];
359 rx_desc_ps = E1000_RX_DESC_PS(*rx_ring, i);
360 u1 = (struct my_u1 *)rx_desc_ps;
362 le32_to_cpu(rx_desc_ps->wb.middle.status_error);
364 if (i == rx_ring->next_to_use)
366 else if (i == rx_ring->next_to_clean)
371 if (staterr & E1000_RXD_STAT_DD) {
372 /* Descriptor Done */
373 pr_info("%s[0x%03X] %016llX %016llX %016llX %016llX ---------------- %p%s\n",
375 (unsigned long long)le64_to_cpu(u1->a),
376 (unsigned long long)le64_to_cpu(u1->b),
377 (unsigned long long)le64_to_cpu(u1->c),
378 (unsigned long long)le64_to_cpu(u1->d),
379 buffer_info->skb, next_desc);
381 pr_info("%s[0x%03X] %016llX %016llX %016llX %016llX %016llX %p%s\n",
383 (unsigned long long)le64_to_cpu(u1->a),
384 (unsigned long long)le64_to_cpu(u1->b),
385 (unsigned long long)le64_to_cpu(u1->c),
386 (unsigned long long)le64_to_cpu(u1->d),
387 (unsigned long long)buffer_info->dma,
388 buffer_info->skb, next_desc);
390 if (netif_msg_pktdata(adapter))
391 e1000e_dump_ps_pages(adapter,
398 /* Extended Receive Descriptor (Read) Format
400 * +-----------------------------------------------------+
401 * 0 | Buffer Address [63:0] |
402 * +-----------------------------------------------------+
404 * +-----------------------------------------------------+
406 pr_info("R [desc] [buf addr 63:0 ] [reserved 63:0 ] [bi->dma ] [bi->skb] <-- Ext (Read) format\n");
407 /* Extended Receive Descriptor (Write-Back) Format
409 * 63 48 47 32 31 24 23 4 3 0
410 * +------------------------------------------------------+
412 * 0 +-------------------+ Rsvd | Reserved | MRQ RSS |
413 * | Packet | IP | | | Type |
414 * | Checksum | Ident | | | |
415 * +------------------------------------------------------+
416 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
417 * +------------------------------------------------------+
418 * 63 48 47 32 31 20 19 0
420 pr_info("RWB[desc] [cs ipid mrq] [vt ln xe xs] [bi->skb] <-- Ext (Write-Back) format\n");
422 for (i = 0; i < rx_ring->count; i++) {
423 const char *next_desc;
425 buffer_info = &rx_ring->buffer_info[i];
426 rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
427 u1 = (struct my_u1 *)rx_desc;
428 staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
430 if (i == rx_ring->next_to_use)
432 else if (i == rx_ring->next_to_clean)
437 if (staterr & E1000_RXD_STAT_DD) {
438 /* Descriptor Done */
439 pr_info("%s[0x%03X] %016llX %016llX ---------------- %p%s\n",
441 (unsigned long long)le64_to_cpu(u1->a),
442 (unsigned long long)le64_to_cpu(u1->b),
443 buffer_info->skb, next_desc);
445 pr_info("%s[0x%03X] %016llX %016llX %016llX %p%s\n",
447 (unsigned long long)le64_to_cpu(u1->a),
448 (unsigned long long)le64_to_cpu(u1->b),
449 (unsigned long long)buffer_info->dma,
450 buffer_info->skb, next_desc);
452 if (netif_msg_pktdata(adapter) &&
454 print_hex_dump(KERN_INFO, "",
455 DUMP_PREFIX_ADDRESS, 16,
457 buffer_info->skb->data,
458 adapter->rx_buffer_len,
466 * e1000_desc_unused - calculate if we have unused descriptors
467 * @ring: pointer to ring struct to perform calculation on
469 static int e1000_desc_unused(struct e1000_ring *ring)
471 if (ring->next_to_clean > ring->next_to_use)
472 return ring->next_to_clean - ring->next_to_use - 1;
474 return ring->count + ring->next_to_clean - ring->next_to_use - 1;
478 * e1000e_systim_to_hwtstamp - convert system time value to hw time stamp
479 * @adapter: board private structure
480 * @hwtstamps: time stamp structure to update
481 * @systim: unsigned 64bit system time value.
483 * Convert the system time value stored in the RX/TXSTMP registers into a
484 * hwtstamp which can be used by the upper level time stamping functions.
486 * The 'systim_lock' spinlock is used to protect the consistency of the
487 * system time value. This is needed because reading the 64 bit time
488 * value involves reading two 32 bit registers. The first read latches the
491 static void e1000e_systim_to_hwtstamp(struct e1000_adapter *adapter,
492 struct skb_shared_hwtstamps *hwtstamps,
498 spin_lock_irqsave(&adapter->systim_lock, flags);
499 ns = timecounter_cyc2time(&adapter->tc, systim);
500 spin_unlock_irqrestore(&adapter->systim_lock, flags);
502 memset(hwtstamps, 0, sizeof(*hwtstamps));
503 hwtstamps->hwtstamp = ns_to_ktime(ns);
507 * e1000e_rx_hwtstamp - utility function which checks for Rx time stamp
508 * @adapter: board private structure
509 * @status: descriptor extended error and status field
510 * @skb: particular skb to include time stamp
512 * If the time stamp is valid, convert it into the timecounter ns value
513 * and store that result into the shhwtstamps structure which is passed
514 * up the network stack.
516 static void e1000e_rx_hwtstamp(struct e1000_adapter *adapter, u32 status,
519 struct e1000_hw *hw = &adapter->hw;
522 if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP) ||
523 !(status & E1000_RXDEXT_STATERR_TST) ||
524 !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID))
527 /* The Rx time stamp registers contain the time stamp. No other
528 * received packet will be time stamped until the Rx time stamp
529 * registers are read. Because only one packet can be time stamped
530 * at a time, the register values must belong to this packet and
531 * therefore none of the other additional attributes need to be
534 rxstmp = (u64)er32(RXSTMPL);
535 rxstmp |= (u64)er32(RXSTMPH) << 32;
536 e1000e_systim_to_hwtstamp(adapter, skb_hwtstamps(skb), rxstmp);
538 adapter->flags2 &= ~FLAG2_CHECK_RX_HWTSTAMP;
542 * e1000_receive_skb - helper function to handle Rx indications
543 * @adapter: board private structure
544 * @netdev: pointer to netdev struct
545 * @staterr: descriptor extended error and status field as written by hardware
546 * @vlan: descriptor vlan field as written by hardware (no le/be conversion)
547 * @skb: pointer to sk_buff to be indicated to stack
549 static void e1000_receive_skb(struct e1000_adapter *adapter,
550 struct net_device *netdev, struct sk_buff *skb,
551 u32 staterr, __le16 vlan)
553 u16 tag = le16_to_cpu(vlan);
555 e1000e_rx_hwtstamp(adapter, staterr, skb);
557 skb->protocol = eth_type_trans(skb, netdev);
559 if (staterr & E1000_RXD_STAT_VP)
560 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), tag);
562 napi_gro_receive(&adapter->napi, skb);
566 * e1000_rx_checksum - Receive Checksum Offload
567 * @adapter: board private structure
568 * @status_err: receive descriptor status and error fields
569 * @skb: socket buffer with received data
571 static void e1000_rx_checksum(struct e1000_adapter *adapter, u32 status_err,
574 u16 status = (u16)status_err;
575 u8 errors = (u8)(status_err >> 24);
577 skb_checksum_none_assert(skb);
579 /* Rx checksum disabled */
580 if (!(adapter->netdev->features & NETIF_F_RXCSUM))
583 /* Ignore Checksum bit is set */
584 if (status & E1000_RXD_STAT_IXSM)
587 /* TCP/UDP checksum error bit or IP checksum error bit is set */
588 if (errors & (E1000_RXD_ERR_TCPE | E1000_RXD_ERR_IPE)) {
589 /* let the stack verify checksum errors */
590 adapter->hw_csum_err++;
594 /* TCP/UDP Checksum has not been calculated */
595 if (!(status & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)))
598 /* It must be a TCP or UDP packet with a valid checksum */
599 skb->ip_summed = CHECKSUM_UNNECESSARY;
600 adapter->hw_csum_good++;
603 static void e1000e_update_rdt_wa(struct e1000_ring *rx_ring, unsigned int i)
605 struct e1000_adapter *adapter = rx_ring->adapter;
606 struct e1000_hw *hw = &adapter->hw;
609 writel(i, rx_ring->tail);
611 if (unlikely(i != readl(rx_ring->tail))) {
612 u32 rctl = er32(RCTL);
614 ew32(RCTL, rctl & ~E1000_RCTL_EN);
615 e_err("ME firmware caused invalid RDT - resetting\n");
616 schedule_work(&adapter->reset_task);
620 static void e1000e_update_tdt_wa(struct e1000_ring *tx_ring, unsigned int i)
622 struct e1000_adapter *adapter = tx_ring->adapter;
623 struct e1000_hw *hw = &adapter->hw;
626 writel(i, tx_ring->tail);
628 if (unlikely(i != readl(tx_ring->tail))) {
629 u32 tctl = er32(TCTL);
631 ew32(TCTL, tctl & ~E1000_TCTL_EN);
632 e_err("ME firmware caused invalid TDT - resetting\n");
633 schedule_work(&adapter->reset_task);
638 * e1000_alloc_rx_buffers - Replace used receive buffers
639 * @rx_ring: Rx descriptor ring
640 * @cleaned_count: number to reallocate
641 * @gfp: flags for allocation
643 static void e1000_alloc_rx_buffers(struct e1000_ring *rx_ring,
644 int cleaned_count, gfp_t gfp)
646 struct e1000_adapter *adapter = rx_ring->adapter;
647 struct net_device *netdev = adapter->netdev;
648 struct pci_dev *pdev = adapter->pdev;
649 union e1000_rx_desc_extended *rx_desc;
650 struct e1000_buffer *buffer_info;
653 unsigned int bufsz = adapter->rx_buffer_len;
655 i = rx_ring->next_to_use;
656 buffer_info = &rx_ring->buffer_info[i];
658 while (cleaned_count--) {
659 skb = buffer_info->skb;
665 skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
667 /* Better luck next round */
668 adapter->alloc_rx_buff_failed++;
672 buffer_info->skb = skb;
674 buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
675 adapter->rx_buffer_len,
677 if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
678 dev_err(&pdev->dev, "Rx DMA map failed\n");
679 adapter->rx_dma_failed++;
683 rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
684 rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
686 if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
687 /* Force memory writes to complete before letting h/w
688 * know there are new descriptors to fetch. (Only
689 * applicable for weak-ordered memory model archs,
693 if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
694 e1000e_update_rdt_wa(rx_ring, i);
696 writel(i, rx_ring->tail);
699 if (i == rx_ring->count)
701 buffer_info = &rx_ring->buffer_info[i];
704 rx_ring->next_to_use = i;
708 * e1000_alloc_rx_buffers_ps - Replace used receive buffers; packet split
709 * @rx_ring: Rx descriptor ring
710 * @cleaned_count: number to reallocate
711 * @gfp: flags for allocation
713 static void e1000_alloc_rx_buffers_ps(struct e1000_ring *rx_ring,
714 int cleaned_count, gfp_t gfp)
716 struct e1000_adapter *adapter = rx_ring->adapter;
717 struct net_device *netdev = adapter->netdev;
718 struct pci_dev *pdev = adapter->pdev;
719 union e1000_rx_desc_packet_split *rx_desc;
720 struct e1000_buffer *buffer_info;
721 struct e1000_ps_page *ps_page;
725 i = rx_ring->next_to_use;
726 buffer_info = &rx_ring->buffer_info[i];
728 while (cleaned_count--) {
729 rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
731 for (j = 0; j < PS_PAGE_BUFFERS; j++) {
732 ps_page = &buffer_info->ps_pages[j];
733 if (j >= adapter->rx_ps_pages) {
734 /* all unused desc entries get hw null ptr */
735 rx_desc->read.buffer_addr[j + 1] =
739 if (!ps_page->page) {
740 ps_page->page = alloc_page(gfp);
741 if (!ps_page->page) {
742 adapter->alloc_rx_buff_failed++;
745 ps_page->dma = dma_map_page(&pdev->dev,
749 if (dma_mapping_error(&pdev->dev,
751 dev_err(&adapter->pdev->dev,
752 "Rx DMA page map failed\n");
753 adapter->rx_dma_failed++;
757 /* Refresh the desc even if buffer_addrs
758 * didn't change because each write-back
761 rx_desc->read.buffer_addr[j + 1] =
762 cpu_to_le64(ps_page->dma);
765 skb = __netdev_alloc_skb_ip_align(netdev, adapter->rx_ps_bsize0,
769 adapter->alloc_rx_buff_failed++;
773 buffer_info->skb = skb;
774 buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
775 adapter->rx_ps_bsize0,
777 if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
778 dev_err(&pdev->dev, "Rx DMA map failed\n");
779 adapter->rx_dma_failed++;
781 dev_kfree_skb_any(skb);
782 buffer_info->skb = NULL;
786 rx_desc->read.buffer_addr[0] = cpu_to_le64(buffer_info->dma);
788 if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
789 /* Force memory writes to complete before letting h/w
790 * know there are new descriptors to fetch. (Only
791 * applicable for weak-ordered memory model archs,
795 if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
796 e1000e_update_rdt_wa(rx_ring, i << 1);
798 writel(i << 1, rx_ring->tail);
802 if (i == rx_ring->count)
804 buffer_info = &rx_ring->buffer_info[i];
808 rx_ring->next_to_use = i;
812 * e1000_alloc_jumbo_rx_buffers - Replace used jumbo receive buffers
813 * @rx_ring: Rx descriptor ring
814 * @cleaned_count: number of buffers to allocate this pass
815 * @gfp: flags for allocation
818 static void e1000_alloc_jumbo_rx_buffers(struct e1000_ring *rx_ring,
819 int cleaned_count, gfp_t gfp)
821 struct e1000_adapter *adapter = rx_ring->adapter;
822 struct net_device *netdev = adapter->netdev;
823 struct pci_dev *pdev = adapter->pdev;
824 union e1000_rx_desc_extended *rx_desc;
825 struct e1000_buffer *buffer_info;
828 unsigned int bufsz = 256 - 16; /* for skb_reserve */
830 i = rx_ring->next_to_use;
831 buffer_info = &rx_ring->buffer_info[i];
833 while (cleaned_count--) {
834 skb = buffer_info->skb;
840 skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
841 if (unlikely(!skb)) {
842 /* Better luck next round */
843 adapter->alloc_rx_buff_failed++;
847 buffer_info->skb = skb;
849 /* allocate a new page if necessary */
850 if (!buffer_info->page) {
851 buffer_info->page = alloc_page(gfp);
852 if (unlikely(!buffer_info->page)) {
853 adapter->alloc_rx_buff_failed++;
858 if (!buffer_info->dma) {
859 buffer_info->dma = dma_map_page(&pdev->dev,
860 buffer_info->page, 0,
863 if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
864 adapter->alloc_rx_buff_failed++;
869 rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
870 rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
872 if (unlikely(++i == rx_ring->count))
874 buffer_info = &rx_ring->buffer_info[i];
877 if (likely(rx_ring->next_to_use != i)) {
878 rx_ring->next_to_use = i;
879 if (unlikely(i-- == 0))
880 i = (rx_ring->count - 1);
882 /* Force memory writes to complete before letting h/w
883 * know there are new descriptors to fetch. (Only
884 * applicable for weak-ordered memory model archs,
888 if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
889 e1000e_update_rdt_wa(rx_ring, i);
891 writel(i, rx_ring->tail);
895 static inline void e1000_rx_hash(struct net_device *netdev, __le32 rss,
898 if (netdev->features & NETIF_F_RXHASH)
899 skb_set_hash(skb, le32_to_cpu(rss), PKT_HASH_TYPE_L3);
903 * e1000_clean_rx_irq - Send received data up the network stack
904 * @rx_ring: Rx descriptor ring
905 * @work_done: output parameter for indicating completed work
906 * @work_to_do: how many packets we can clean
908 * the return value indicates whether actual cleaning was done, there
909 * is no guarantee that everything was cleaned
911 static bool e1000_clean_rx_irq(struct e1000_ring *rx_ring, int *work_done,
914 struct e1000_adapter *adapter = rx_ring->adapter;
915 struct net_device *netdev = adapter->netdev;
916 struct pci_dev *pdev = adapter->pdev;
917 struct e1000_hw *hw = &adapter->hw;
918 union e1000_rx_desc_extended *rx_desc, *next_rxd;
919 struct e1000_buffer *buffer_info, *next_buffer;
922 int cleaned_count = 0;
923 bool cleaned = false;
924 unsigned int total_rx_bytes = 0, total_rx_packets = 0;
926 i = rx_ring->next_to_clean;
927 rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
928 staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
929 buffer_info = &rx_ring->buffer_info[i];
931 while (staterr & E1000_RXD_STAT_DD) {
934 if (*work_done >= work_to_do)
937 dma_rmb(); /* read descriptor and rx_buffer_info after status DD */
939 skb = buffer_info->skb;
940 buffer_info->skb = NULL;
942 prefetch(skb->data - NET_IP_ALIGN);
945 if (i == rx_ring->count)
947 next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
950 next_buffer = &rx_ring->buffer_info[i];
954 dma_unmap_single(&pdev->dev, buffer_info->dma,
955 adapter->rx_buffer_len, DMA_FROM_DEVICE);
956 buffer_info->dma = 0;
958 length = le16_to_cpu(rx_desc->wb.upper.length);
960 /* !EOP means multiple descriptors were used to store a single
961 * packet, if that's the case we need to toss it. In fact, we
962 * need to toss every packet with the EOP bit clear and the
963 * next frame that _does_ have the EOP bit set, as it is by
964 * definition only a frame fragment
966 if (unlikely(!(staterr & E1000_RXD_STAT_EOP)))
967 adapter->flags2 |= FLAG2_IS_DISCARDING;
969 if (adapter->flags2 & FLAG2_IS_DISCARDING) {
970 /* All receives must fit into a single buffer */
971 e_dbg("Receive packet consumed multiple buffers\n");
973 buffer_info->skb = skb;
974 if (staterr & E1000_RXD_STAT_EOP)
975 adapter->flags2 &= ~FLAG2_IS_DISCARDING;
979 if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
980 !(netdev->features & NETIF_F_RXALL))) {
982 buffer_info->skb = skb;
986 /* adjust length to remove Ethernet CRC */
987 if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
988 /* If configured to store CRC, don't subtract FCS,
989 * but keep the FCS bytes out of the total_rx_bytes
992 if (netdev->features & NETIF_F_RXFCS)
998 total_rx_bytes += length;
1001 /* code added for copybreak, this should improve
1002 * performance for small packets with large amounts
1003 * of reassembly being done in the stack
1005 if (length < copybreak) {
1006 struct sk_buff *new_skb =
1007 napi_alloc_skb(&adapter->napi, length);
1009 skb_copy_to_linear_data_offset(new_skb,
1015 /* save the skb in buffer_info as good */
1016 buffer_info->skb = skb;
1019 /* else just continue with the old one */
1021 /* end copybreak code */
1022 skb_put(skb, length);
1024 /* Receive Checksum Offload */
1025 e1000_rx_checksum(adapter, staterr, skb);
1027 e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1029 e1000_receive_skb(adapter, netdev, skb, staterr,
1030 rx_desc->wb.upper.vlan);
1033 rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1035 /* return some buffers to hardware, one at a time is too slow */
1036 if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1037 adapter->alloc_rx_buf(rx_ring, cleaned_count,
1042 /* use prefetched values */
1044 buffer_info = next_buffer;
1046 staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1048 rx_ring->next_to_clean = i;
1050 cleaned_count = e1000_desc_unused(rx_ring);
1052 adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1054 adapter->total_rx_bytes += total_rx_bytes;
1055 adapter->total_rx_packets += total_rx_packets;
1059 static void e1000_put_txbuf(struct e1000_ring *tx_ring,
1060 struct e1000_buffer *buffer_info,
1063 struct e1000_adapter *adapter = tx_ring->adapter;
1065 if (buffer_info->dma) {
1066 if (buffer_info->mapped_as_page)
1067 dma_unmap_page(&adapter->pdev->dev, buffer_info->dma,
1068 buffer_info->length, DMA_TO_DEVICE);
1070 dma_unmap_single(&adapter->pdev->dev, buffer_info->dma,
1071 buffer_info->length, DMA_TO_DEVICE);
1072 buffer_info->dma = 0;
1074 if (buffer_info->skb) {
1076 dev_kfree_skb_any(buffer_info->skb);
1078 dev_consume_skb_any(buffer_info->skb);
1079 buffer_info->skb = NULL;
1081 buffer_info->time_stamp = 0;
1084 static void e1000_print_hw_hang(struct work_struct *work)
1086 struct e1000_adapter *adapter = container_of(work,
1087 struct e1000_adapter,
1089 struct net_device *netdev = adapter->netdev;
1090 struct e1000_ring *tx_ring = adapter->tx_ring;
1091 unsigned int i = tx_ring->next_to_clean;
1092 unsigned int eop = tx_ring->buffer_info[i].next_to_watch;
1093 struct e1000_tx_desc *eop_desc = E1000_TX_DESC(*tx_ring, eop);
1094 struct e1000_hw *hw = &adapter->hw;
1095 u16 phy_status, phy_1000t_status, phy_ext_status;
1098 if (test_bit(__E1000_DOWN, &adapter->state))
1101 if (!adapter->tx_hang_recheck && (adapter->flags2 & FLAG2_DMA_BURST)) {
1102 /* May be block on write-back, flush and detect again
1103 * flush pending descriptor writebacks to memory
1105 ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1106 /* execute the writes immediately */
1108 /* Due to rare timing issues, write to TIDV again to ensure
1109 * the write is successful
1111 ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1112 /* execute the writes immediately */
1114 adapter->tx_hang_recheck = true;
1117 adapter->tx_hang_recheck = false;
1119 if (er32(TDH(0)) == er32(TDT(0))) {
1120 e_dbg("false hang detected, ignoring\n");
1124 /* Real hang detected */
1125 netif_stop_queue(netdev);
1127 e1e_rphy(hw, MII_BMSR, &phy_status);
1128 e1e_rphy(hw, MII_STAT1000, &phy_1000t_status);
1129 e1e_rphy(hw, MII_ESTATUS, &phy_ext_status);
1131 pci_read_config_word(adapter->pdev, PCI_STATUS, &pci_status);
1133 /* detected Hardware unit hang */
1134 e_err("Detected Hardware Unit Hang:\n"
1137 " next_to_use <%x>\n"
1138 " next_to_clean <%x>\n"
1139 "buffer_info[next_to_clean]:\n"
1140 " time_stamp <%lx>\n"
1141 " next_to_watch <%x>\n"
1143 " next_to_watch.status <%x>\n"
1146 "PHY 1000BASE-T Status <%x>\n"
1147 "PHY Extended Status <%x>\n"
1148 "PCI Status <%x>\n",
1149 readl(tx_ring->head), readl(tx_ring->tail), tx_ring->next_to_use,
1150 tx_ring->next_to_clean, tx_ring->buffer_info[eop].time_stamp,
1151 eop, jiffies, eop_desc->upper.fields.status, er32(STATUS),
1152 phy_status, phy_1000t_status, phy_ext_status, pci_status);
1154 e1000e_dump(adapter);
1156 /* Suggest workaround for known h/w issue */
1157 if ((hw->mac.type == e1000_pchlan) && (er32(CTRL) & E1000_CTRL_TFCE))
1158 e_err("Try turning off Tx pause (flow control) via ethtool\n");
1162 * e1000e_tx_hwtstamp_work - check for Tx time stamp
1163 * @work: pointer to work struct
1165 * This work function polls the TSYNCTXCTL valid bit to determine when a
1166 * timestamp has been taken for the current stored skb. The timestamp must
1167 * be for this skb because only one such packet is allowed in the queue.
1169 static void e1000e_tx_hwtstamp_work(struct work_struct *work)
1171 struct e1000_adapter *adapter = container_of(work, struct e1000_adapter,
1173 struct e1000_hw *hw = &adapter->hw;
1175 if (er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID) {
1176 struct sk_buff *skb = adapter->tx_hwtstamp_skb;
1177 struct skb_shared_hwtstamps shhwtstamps;
1180 txstmp = er32(TXSTMPL);
1181 txstmp |= (u64)er32(TXSTMPH) << 32;
1183 e1000e_systim_to_hwtstamp(adapter, &shhwtstamps, txstmp);
1185 /* Clear the global tx_hwtstamp_skb pointer and force writes
1186 * prior to notifying the stack of a Tx timestamp.
1188 adapter->tx_hwtstamp_skb = NULL;
1189 wmb(); /* force write prior to skb_tstamp_tx */
1191 skb_tstamp_tx(skb, &shhwtstamps);
1192 dev_consume_skb_any(skb);
1193 } else if (time_after(jiffies, adapter->tx_hwtstamp_start
1194 + adapter->tx_timeout_factor * HZ)) {
1195 dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
1196 adapter->tx_hwtstamp_skb = NULL;
1197 adapter->tx_hwtstamp_timeouts++;
1198 e_warn("clearing Tx timestamp hang\n");
1200 /* reschedule to check later */
1201 schedule_work(&adapter->tx_hwtstamp_work);
1206 * e1000_clean_tx_irq - Reclaim resources after transmit completes
1207 * @tx_ring: Tx descriptor ring
1209 * the return value indicates whether actual cleaning was done, there
1210 * is no guarantee that everything was cleaned
1212 static bool e1000_clean_tx_irq(struct e1000_ring *tx_ring)
1214 struct e1000_adapter *adapter = tx_ring->adapter;
1215 struct net_device *netdev = adapter->netdev;
1216 struct e1000_hw *hw = &adapter->hw;
1217 struct e1000_tx_desc *tx_desc, *eop_desc;
1218 struct e1000_buffer *buffer_info;
1219 unsigned int i, eop;
1220 unsigned int count = 0;
1221 unsigned int total_tx_bytes = 0, total_tx_packets = 0;
1222 unsigned int bytes_compl = 0, pkts_compl = 0;
1224 i = tx_ring->next_to_clean;
1225 eop = tx_ring->buffer_info[i].next_to_watch;
1226 eop_desc = E1000_TX_DESC(*tx_ring, eop);
1228 while ((eop_desc->upper.data & cpu_to_le32(E1000_TXD_STAT_DD)) &&
1229 (count < tx_ring->count)) {
1230 bool cleaned = false;
1232 dma_rmb(); /* read buffer_info after eop_desc */
1233 for (; !cleaned; count++) {
1234 tx_desc = E1000_TX_DESC(*tx_ring, i);
1235 buffer_info = &tx_ring->buffer_info[i];
1236 cleaned = (i == eop);
1239 total_tx_packets += buffer_info->segs;
1240 total_tx_bytes += buffer_info->bytecount;
1241 if (buffer_info->skb) {
1242 bytes_compl += buffer_info->skb->len;
1247 e1000_put_txbuf(tx_ring, buffer_info, false);
1248 tx_desc->upper.data = 0;
1251 if (i == tx_ring->count)
1255 if (i == tx_ring->next_to_use)
1257 eop = tx_ring->buffer_info[i].next_to_watch;
1258 eop_desc = E1000_TX_DESC(*tx_ring, eop);
1261 tx_ring->next_to_clean = i;
1263 netdev_completed_queue(netdev, pkts_compl, bytes_compl);
1265 #define TX_WAKE_THRESHOLD 32
1266 if (count && netif_carrier_ok(netdev) &&
1267 e1000_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD) {
1268 /* Make sure that anybody stopping the queue after this
1269 * sees the new next_to_clean.
1273 if (netif_queue_stopped(netdev) &&
1274 !(test_bit(__E1000_DOWN, &adapter->state))) {
1275 netif_wake_queue(netdev);
1276 ++adapter->restart_queue;
1280 if (adapter->detect_tx_hung) {
1281 /* Detect a transmit hang in hardware, this serializes the
1282 * check with the clearing of time_stamp and movement of i
1284 adapter->detect_tx_hung = false;
1285 if (tx_ring->buffer_info[i].time_stamp &&
1286 time_after(jiffies, tx_ring->buffer_info[i].time_stamp
1287 + (adapter->tx_timeout_factor * HZ)) &&
1288 !(er32(STATUS) & E1000_STATUS_TXOFF))
1289 schedule_work(&adapter->print_hang_task);
1291 adapter->tx_hang_recheck = false;
1293 adapter->total_tx_bytes += total_tx_bytes;
1294 adapter->total_tx_packets += total_tx_packets;
1295 return count < tx_ring->count;
1299 * e1000_clean_rx_irq_ps - Send received data up the network stack; packet split
1300 * @rx_ring: Rx descriptor ring
1301 * @work_done: output parameter for indicating completed work
1302 * @work_to_do: how many packets we can clean
1304 * the return value indicates whether actual cleaning was done, there
1305 * is no guarantee that everything was cleaned
1307 static bool e1000_clean_rx_irq_ps(struct e1000_ring *rx_ring, int *work_done,
1310 struct e1000_adapter *adapter = rx_ring->adapter;
1311 struct e1000_hw *hw = &adapter->hw;
1312 union e1000_rx_desc_packet_split *rx_desc, *next_rxd;
1313 struct net_device *netdev = adapter->netdev;
1314 struct pci_dev *pdev = adapter->pdev;
1315 struct e1000_buffer *buffer_info, *next_buffer;
1316 struct e1000_ps_page *ps_page;
1317 struct sk_buff *skb;
1319 u32 length, staterr;
1320 int cleaned_count = 0;
1321 bool cleaned = false;
1322 unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1324 i = rx_ring->next_to_clean;
1325 rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
1326 staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1327 buffer_info = &rx_ring->buffer_info[i];
1329 while (staterr & E1000_RXD_STAT_DD) {
1330 if (*work_done >= work_to_do)
1333 skb = buffer_info->skb;
1334 dma_rmb(); /* read descriptor and rx_buffer_info after status DD */
1336 /* in the packet split case this is header only */
1337 prefetch(skb->data - NET_IP_ALIGN);
1340 if (i == rx_ring->count)
1342 next_rxd = E1000_RX_DESC_PS(*rx_ring, i);
1345 next_buffer = &rx_ring->buffer_info[i];
1349 dma_unmap_single(&pdev->dev, buffer_info->dma,
1350 adapter->rx_ps_bsize0, DMA_FROM_DEVICE);
1351 buffer_info->dma = 0;
1353 /* see !EOP comment in other Rx routine */
1354 if (!(staterr & E1000_RXD_STAT_EOP))
1355 adapter->flags2 |= FLAG2_IS_DISCARDING;
1357 if (adapter->flags2 & FLAG2_IS_DISCARDING) {
1358 e_dbg("Packet Split buffers didn't pick up the full packet\n");
1359 dev_kfree_skb_irq(skb);
1360 if (staterr & E1000_RXD_STAT_EOP)
1361 adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1365 if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1366 !(netdev->features & NETIF_F_RXALL))) {
1367 dev_kfree_skb_irq(skb);
1371 length = le16_to_cpu(rx_desc->wb.middle.length0);
1374 e_dbg("Last part of the packet spanning multiple descriptors\n");
1375 dev_kfree_skb_irq(skb);
1380 skb_put(skb, length);
1383 /* this looks ugly, but it seems compiler issues make
1384 * it more efficient than reusing j
1386 int l1 = le16_to_cpu(rx_desc->wb.upper.length[0]);
1388 /* page alloc/put takes too long and effects small
1389 * packet throughput, so unsplit small packets and
1390 * save the alloc/put only valid in softirq (napi)
1391 * context to call kmap_*
1393 if (l1 && (l1 <= copybreak) &&
1394 ((length + l1) <= adapter->rx_ps_bsize0)) {
1397 ps_page = &buffer_info->ps_pages[0];
1399 /* there is no documentation about how to call
1400 * kmap_atomic, so we can't hold the mapping
1403 dma_sync_single_for_cpu(&pdev->dev,
1407 vaddr = kmap_atomic(ps_page->page);
1408 memcpy(skb_tail_pointer(skb), vaddr, l1);
1409 kunmap_atomic(vaddr);
1410 dma_sync_single_for_device(&pdev->dev,
1415 /* remove the CRC */
1416 if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1417 if (!(netdev->features & NETIF_F_RXFCS))
1426 for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1427 length = le16_to_cpu(rx_desc->wb.upper.length[j]);
1431 ps_page = &buffer_info->ps_pages[j];
1432 dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1435 skb_fill_page_desc(skb, j, ps_page->page, 0, length);
1436 ps_page->page = NULL;
1438 skb->data_len += length;
1439 skb->truesize += PAGE_SIZE;
1442 /* strip the ethernet crc, problem is we're using pages now so
1443 * this whole operation can get a little cpu intensive
1445 if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1446 if (!(netdev->features & NETIF_F_RXFCS))
1447 pskb_trim(skb, skb->len - 4);
1451 total_rx_bytes += skb->len;
1454 e1000_rx_checksum(adapter, staterr, skb);
1456 e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1458 if (rx_desc->wb.upper.header_status &
1459 cpu_to_le16(E1000_RXDPS_HDRSTAT_HDRSP))
1460 adapter->rx_hdr_split++;
1462 e1000_receive_skb(adapter, netdev, skb, staterr,
1463 rx_desc->wb.middle.vlan);
1466 rx_desc->wb.middle.status_error &= cpu_to_le32(~0xFF);
1467 buffer_info->skb = NULL;
1469 /* return some buffers to hardware, one at a time is too slow */
1470 if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1471 adapter->alloc_rx_buf(rx_ring, cleaned_count,
1476 /* use prefetched values */
1478 buffer_info = next_buffer;
1480 staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1482 rx_ring->next_to_clean = i;
1484 cleaned_count = e1000_desc_unused(rx_ring);
1486 adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1488 adapter->total_rx_bytes += total_rx_bytes;
1489 adapter->total_rx_packets += total_rx_packets;
1493 static void e1000_consume_page(struct e1000_buffer *bi, struct sk_buff *skb,
1498 skb->data_len += length;
1499 skb->truesize += PAGE_SIZE;
1503 * e1000_clean_jumbo_rx_irq - Send received data up the network stack; legacy
1504 * @rx_ring: Rx descriptor ring
1505 * @work_done: output parameter for indicating completed work
1506 * @work_to_do: how many packets we can clean
1508 * the return value indicates whether actual cleaning was done, there
1509 * is no guarantee that everything was cleaned
1511 static bool e1000_clean_jumbo_rx_irq(struct e1000_ring *rx_ring, int *work_done,
1514 struct e1000_adapter *adapter = rx_ring->adapter;
1515 struct net_device *netdev = adapter->netdev;
1516 struct pci_dev *pdev = adapter->pdev;
1517 union e1000_rx_desc_extended *rx_desc, *next_rxd;
1518 struct e1000_buffer *buffer_info, *next_buffer;
1519 u32 length, staterr;
1521 int cleaned_count = 0;
1522 bool cleaned = false;
1523 unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1524 struct skb_shared_info *shinfo;
1526 i = rx_ring->next_to_clean;
1527 rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
1528 staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1529 buffer_info = &rx_ring->buffer_info[i];
1531 while (staterr & E1000_RXD_STAT_DD) {
1532 struct sk_buff *skb;
1534 if (*work_done >= work_to_do)
1537 dma_rmb(); /* read descriptor and rx_buffer_info after status DD */
1539 skb = buffer_info->skb;
1540 buffer_info->skb = NULL;
1543 if (i == rx_ring->count)
1545 next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
1548 next_buffer = &rx_ring->buffer_info[i];
1552 dma_unmap_page(&pdev->dev, buffer_info->dma, PAGE_SIZE,
1554 buffer_info->dma = 0;
1556 length = le16_to_cpu(rx_desc->wb.upper.length);
1558 /* errors is only valid for DD + EOP descriptors */
1559 if (unlikely((staterr & E1000_RXD_STAT_EOP) &&
1560 ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1561 !(netdev->features & NETIF_F_RXALL)))) {
1562 /* recycle both page and skb */
1563 buffer_info->skb = skb;
1564 /* an error means any chain goes out the window too */
1565 if (rx_ring->rx_skb_top)
1566 dev_kfree_skb_irq(rx_ring->rx_skb_top);
1567 rx_ring->rx_skb_top = NULL;
1570 #define rxtop (rx_ring->rx_skb_top)
1571 if (!(staterr & E1000_RXD_STAT_EOP)) {
1572 /* this descriptor is only the beginning (or middle) */
1574 /* this is the beginning of a chain */
1576 skb_fill_page_desc(rxtop, 0, buffer_info->page,
1579 /* this is the middle of a chain */
1580 shinfo = skb_shinfo(rxtop);
1581 skb_fill_page_desc(rxtop, shinfo->nr_frags,
1582 buffer_info->page, 0,
1584 /* re-use the skb, only consumed the page */
1585 buffer_info->skb = skb;
1587 e1000_consume_page(buffer_info, rxtop, length);
1591 /* end of the chain */
1592 shinfo = skb_shinfo(rxtop);
1593 skb_fill_page_desc(rxtop, shinfo->nr_frags,
1594 buffer_info->page, 0,
1596 /* re-use the current skb, we only consumed the
1599 buffer_info->skb = skb;
1602 e1000_consume_page(buffer_info, skb, length);
1604 /* no chain, got EOP, this buf is the packet
1605 * copybreak to save the put_page/alloc_page
1607 if (length <= copybreak &&
1608 skb_tailroom(skb) >= length) {
1610 vaddr = kmap_atomic(buffer_info->page);
1611 memcpy(skb_tail_pointer(skb), vaddr,
1613 kunmap_atomic(vaddr);
1614 /* re-use the page, so don't erase
1617 skb_put(skb, length);
1619 skb_fill_page_desc(skb, 0,
1620 buffer_info->page, 0,
1622 e1000_consume_page(buffer_info, skb,
1628 /* Receive Checksum Offload */
1629 e1000_rx_checksum(adapter, staterr, skb);
1631 e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1633 /* probably a little skewed due to removing CRC */
1634 total_rx_bytes += skb->len;
1637 /* eth type trans needs skb->data to point to something */
1638 if (!pskb_may_pull(skb, ETH_HLEN)) {
1639 e_err("pskb_may_pull failed.\n");
1640 dev_kfree_skb_irq(skb);
1644 e1000_receive_skb(adapter, netdev, skb, staterr,
1645 rx_desc->wb.upper.vlan);
1648 rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1650 /* return some buffers to hardware, one at a time is too slow */
1651 if (unlikely(cleaned_count >= E1000_RX_BUFFER_WRITE)) {
1652 adapter->alloc_rx_buf(rx_ring, cleaned_count,
1657 /* use prefetched values */
1659 buffer_info = next_buffer;
1661 staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1663 rx_ring->next_to_clean = i;
1665 cleaned_count = e1000_desc_unused(rx_ring);
1667 adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1669 adapter->total_rx_bytes += total_rx_bytes;
1670 adapter->total_rx_packets += total_rx_packets;
1675 * e1000_clean_rx_ring - Free Rx Buffers per Queue
1676 * @rx_ring: Rx descriptor ring
1678 static void e1000_clean_rx_ring(struct e1000_ring *rx_ring)
1680 struct e1000_adapter *adapter = rx_ring->adapter;
1681 struct e1000_buffer *buffer_info;
1682 struct e1000_ps_page *ps_page;
1683 struct pci_dev *pdev = adapter->pdev;
1686 /* Free all the Rx ring sk_buffs */
1687 for (i = 0; i < rx_ring->count; i++) {
1688 buffer_info = &rx_ring->buffer_info[i];
1689 if (buffer_info->dma) {
1690 if (adapter->clean_rx == e1000_clean_rx_irq)
1691 dma_unmap_single(&pdev->dev, buffer_info->dma,
1692 adapter->rx_buffer_len,
1694 else if (adapter->clean_rx == e1000_clean_jumbo_rx_irq)
1695 dma_unmap_page(&pdev->dev, buffer_info->dma,
1696 PAGE_SIZE, DMA_FROM_DEVICE);
1697 else if (adapter->clean_rx == e1000_clean_rx_irq_ps)
1698 dma_unmap_single(&pdev->dev, buffer_info->dma,
1699 adapter->rx_ps_bsize0,
1701 buffer_info->dma = 0;
1704 if (buffer_info->page) {
1705 put_page(buffer_info->page);
1706 buffer_info->page = NULL;
1709 if (buffer_info->skb) {
1710 dev_kfree_skb(buffer_info->skb);
1711 buffer_info->skb = NULL;
1714 for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1715 ps_page = &buffer_info->ps_pages[j];
1718 dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1721 put_page(ps_page->page);
1722 ps_page->page = NULL;
1726 /* there also may be some cached data from a chained receive */
1727 if (rx_ring->rx_skb_top) {
1728 dev_kfree_skb(rx_ring->rx_skb_top);
1729 rx_ring->rx_skb_top = NULL;
1732 /* Zero out the descriptor ring */
1733 memset(rx_ring->desc, 0, rx_ring->size);
1735 rx_ring->next_to_clean = 0;
1736 rx_ring->next_to_use = 0;
1737 adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1740 static void e1000e_downshift_workaround(struct work_struct *work)
1742 struct e1000_adapter *adapter = container_of(work,
1743 struct e1000_adapter,
1746 if (test_bit(__E1000_DOWN, &adapter->state))
1749 e1000e_gig_downshift_workaround_ich8lan(&adapter->hw);
1753 * e1000_intr_msi - Interrupt Handler
1754 * @irq: interrupt number
1755 * @data: pointer to a network interface device structure
1757 static irqreturn_t e1000_intr_msi(int __always_unused irq, void *data)
1759 struct net_device *netdev = data;
1760 struct e1000_adapter *adapter = netdev_priv(netdev);
1761 struct e1000_hw *hw = &adapter->hw;
1762 u32 icr = er32(ICR);
1764 /* read ICR disables interrupts using IAM */
1765 if (icr & E1000_ICR_LSC) {
1766 hw->mac.get_link_status = true;
1767 /* ICH8 workaround-- Call gig speed drop workaround on cable
1768 * disconnect (LSC) before accessing any PHY registers
1770 if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1771 (!(er32(STATUS) & E1000_STATUS_LU)))
1772 schedule_work(&adapter->downshift_task);
1774 /* 80003ES2LAN workaround-- For packet buffer work-around on
1775 * link down event; disable receives here in the ISR and reset
1776 * adapter in watchdog
1778 if (netif_carrier_ok(netdev) &&
1779 adapter->flags & FLAG_RX_NEEDS_RESTART) {
1780 /* disable receives */
1781 u32 rctl = er32(RCTL);
1783 ew32(RCTL, rctl & ~E1000_RCTL_EN);
1784 adapter->flags |= FLAG_RESTART_NOW;
1786 /* guard against interrupt when we're going down */
1787 if (!test_bit(__E1000_DOWN, &adapter->state))
1788 mod_timer(&adapter->watchdog_timer, jiffies + 1);
1791 /* Reset on uncorrectable ECC error */
1792 if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1793 u32 pbeccsts = er32(PBECCSTS);
1795 adapter->corr_errors +=
1796 pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1797 adapter->uncorr_errors +=
1798 (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1799 E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1801 /* Do the reset outside of interrupt context */
1802 schedule_work(&adapter->reset_task);
1804 /* return immediately since reset is imminent */
1808 if (napi_schedule_prep(&adapter->napi)) {
1809 adapter->total_tx_bytes = 0;
1810 adapter->total_tx_packets = 0;
1811 adapter->total_rx_bytes = 0;
1812 adapter->total_rx_packets = 0;
1813 __napi_schedule(&adapter->napi);
1820 * e1000_intr - Interrupt Handler
1821 * @irq: interrupt number
1822 * @data: pointer to a network interface device structure
1824 static irqreturn_t e1000_intr(int __always_unused irq, void *data)
1826 struct net_device *netdev = data;
1827 struct e1000_adapter *adapter = netdev_priv(netdev);
1828 struct e1000_hw *hw = &adapter->hw;
1829 u32 rctl, icr = er32(ICR);
1831 if (!icr || test_bit(__E1000_DOWN, &adapter->state))
1832 return IRQ_NONE; /* Not our interrupt */
1834 /* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
1835 * not set, then the adapter didn't send an interrupt
1837 if (!(icr & E1000_ICR_INT_ASSERTED))
1840 /* Interrupt Auto-Mask...upon reading ICR,
1841 * interrupts are masked. No need for the
1845 if (icr & E1000_ICR_LSC) {
1846 hw->mac.get_link_status = true;
1847 /* ICH8 workaround-- Call gig speed drop workaround on cable
1848 * disconnect (LSC) before accessing any PHY registers
1850 if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1851 (!(er32(STATUS) & E1000_STATUS_LU)))
1852 schedule_work(&adapter->downshift_task);
1854 /* 80003ES2LAN workaround--
1855 * For packet buffer work-around on link down event;
1856 * disable receives here in the ISR and
1857 * reset adapter in watchdog
1859 if (netif_carrier_ok(netdev) &&
1860 (adapter->flags & FLAG_RX_NEEDS_RESTART)) {
1861 /* disable receives */
1863 ew32(RCTL, rctl & ~E1000_RCTL_EN);
1864 adapter->flags |= FLAG_RESTART_NOW;
1866 /* guard against interrupt when we're going down */
1867 if (!test_bit(__E1000_DOWN, &adapter->state))
1868 mod_timer(&adapter->watchdog_timer, jiffies + 1);
1871 /* Reset on uncorrectable ECC error */
1872 if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1873 u32 pbeccsts = er32(PBECCSTS);
1875 adapter->corr_errors +=
1876 pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1877 adapter->uncorr_errors +=
1878 (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1879 E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1881 /* Do the reset outside of interrupt context */
1882 schedule_work(&adapter->reset_task);
1884 /* return immediately since reset is imminent */
1888 if (napi_schedule_prep(&adapter->napi)) {
1889 adapter->total_tx_bytes = 0;
1890 adapter->total_tx_packets = 0;
1891 adapter->total_rx_bytes = 0;
1892 adapter->total_rx_packets = 0;
1893 __napi_schedule(&adapter->napi);
1899 static irqreturn_t e1000_msix_other(int __always_unused irq, void *data)
1901 struct net_device *netdev = data;
1902 struct e1000_adapter *adapter = netdev_priv(netdev);
1903 struct e1000_hw *hw = &adapter->hw;
1904 u32 icr = er32(ICR);
1906 if (icr & adapter->eiac_mask)
1907 ew32(ICS, (icr & adapter->eiac_mask));
1909 if (icr & E1000_ICR_LSC) {
1910 hw->mac.get_link_status = true;
1911 /* guard against interrupt when we're going down */
1912 if (!test_bit(__E1000_DOWN, &adapter->state))
1913 mod_timer(&adapter->watchdog_timer, jiffies + 1);
1916 if (!test_bit(__E1000_DOWN, &adapter->state))
1917 ew32(IMS, E1000_IMS_OTHER | IMS_OTHER_MASK);
1922 static irqreturn_t e1000_intr_msix_tx(int __always_unused irq, void *data)
1924 struct net_device *netdev = data;
1925 struct e1000_adapter *adapter = netdev_priv(netdev);
1926 struct e1000_hw *hw = &adapter->hw;
1927 struct e1000_ring *tx_ring = adapter->tx_ring;
1929 adapter->total_tx_bytes = 0;
1930 adapter->total_tx_packets = 0;
1932 if (!e1000_clean_tx_irq(tx_ring))
1933 /* Ring was not completely cleaned, so fire another interrupt */
1934 ew32(ICS, tx_ring->ims_val);
1936 if (!test_bit(__E1000_DOWN, &adapter->state))
1937 ew32(IMS, adapter->tx_ring->ims_val);
1942 static irqreturn_t e1000_intr_msix_rx(int __always_unused irq, void *data)
1944 struct net_device *netdev = data;
1945 struct e1000_adapter *adapter = netdev_priv(netdev);
1946 struct e1000_ring *rx_ring = adapter->rx_ring;
1948 /* Write the ITR value calculated at the end of the
1949 * previous interrupt.
1951 if (rx_ring->set_itr) {
1952 u32 itr = rx_ring->itr_val ?
1953 1000000000 / (rx_ring->itr_val * 256) : 0;
1955 writel(itr, rx_ring->itr_register);
1956 rx_ring->set_itr = 0;
1959 if (napi_schedule_prep(&adapter->napi)) {
1960 adapter->total_rx_bytes = 0;
1961 adapter->total_rx_packets = 0;
1962 __napi_schedule(&adapter->napi);
1968 * e1000_configure_msix - Configure MSI-X hardware
1969 * @adapter: board private structure
1971 * e1000_configure_msix sets up the hardware to properly
1972 * generate MSI-X interrupts.
1974 static void e1000_configure_msix(struct e1000_adapter *adapter)
1976 struct e1000_hw *hw = &adapter->hw;
1977 struct e1000_ring *rx_ring = adapter->rx_ring;
1978 struct e1000_ring *tx_ring = adapter->tx_ring;
1980 u32 ctrl_ext, ivar = 0;
1982 adapter->eiac_mask = 0;
1984 /* Workaround issue with spurious interrupts on 82574 in MSI-X mode */
1985 if (hw->mac.type == e1000_82574) {
1986 u32 rfctl = er32(RFCTL);
1988 rfctl |= E1000_RFCTL_ACK_DIS;
1992 /* Configure Rx vector */
1993 rx_ring->ims_val = E1000_IMS_RXQ0;
1994 adapter->eiac_mask |= rx_ring->ims_val;
1995 if (rx_ring->itr_val)
1996 writel(1000000000 / (rx_ring->itr_val * 256),
1997 rx_ring->itr_register);
1999 writel(1, rx_ring->itr_register);
2000 ivar = E1000_IVAR_INT_ALLOC_VALID | vector;
2002 /* Configure Tx vector */
2003 tx_ring->ims_val = E1000_IMS_TXQ0;
2005 if (tx_ring->itr_val)
2006 writel(1000000000 / (tx_ring->itr_val * 256),
2007 tx_ring->itr_register);
2009 writel(1, tx_ring->itr_register);
2010 adapter->eiac_mask |= tx_ring->ims_val;
2011 ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 8);
2013 /* set vector for Other Causes, e.g. link changes */
2015 ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 16);
2016 if (rx_ring->itr_val)
2017 writel(1000000000 / (rx_ring->itr_val * 256),
2018 hw->hw_addr + E1000_EITR_82574(vector));
2020 writel(1, hw->hw_addr + E1000_EITR_82574(vector));
2022 /* Cause Tx interrupts on every write back */
2027 /* enable MSI-X PBA support */
2028 ctrl_ext = er32(CTRL_EXT) & ~E1000_CTRL_EXT_IAME;
2029 ctrl_ext |= E1000_CTRL_EXT_PBA_CLR | E1000_CTRL_EXT_EIAME;
2030 ew32(CTRL_EXT, ctrl_ext);
2034 void e1000e_reset_interrupt_capability(struct e1000_adapter *adapter)
2036 if (adapter->msix_entries) {
2037 pci_disable_msix(adapter->pdev);
2038 kfree(adapter->msix_entries);
2039 adapter->msix_entries = NULL;
2040 } else if (adapter->flags & FLAG_MSI_ENABLED) {
2041 pci_disable_msi(adapter->pdev);
2042 adapter->flags &= ~FLAG_MSI_ENABLED;
2047 * e1000e_set_interrupt_capability - set MSI or MSI-X if supported
2048 * @adapter: board private structure
2050 * Attempt to configure interrupts using the best available
2051 * capabilities of the hardware and kernel.
2053 void e1000e_set_interrupt_capability(struct e1000_adapter *adapter)
2058 switch (adapter->int_mode) {
2059 case E1000E_INT_MODE_MSIX:
2060 if (adapter->flags & FLAG_HAS_MSIX) {
2061 adapter->num_vectors = 3; /* RxQ0, TxQ0 and other */
2062 adapter->msix_entries = kcalloc(adapter->num_vectors,
2066 if (adapter->msix_entries) {
2067 struct e1000_adapter *a = adapter;
2069 for (i = 0; i < adapter->num_vectors; i++)
2070 adapter->msix_entries[i].entry = i;
2072 err = pci_enable_msix_range(a->pdev,
2079 /* MSI-X failed, so fall through and try MSI */
2080 e_err("Failed to initialize MSI-X interrupts. Falling back to MSI interrupts.\n");
2081 e1000e_reset_interrupt_capability(adapter);
2083 adapter->int_mode = E1000E_INT_MODE_MSI;
2085 case E1000E_INT_MODE_MSI:
2086 if (!pci_enable_msi(adapter->pdev)) {
2087 adapter->flags |= FLAG_MSI_ENABLED;
2089 adapter->int_mode = E1000E_INT_MODE_LEGACY;
2090 e_err("Failed to initialize MSI interrupts. Falling back to legacy interrupts.\n");
2093 case E1000E_INT_MODE_LEGACY:
2094 /* Don't do anything; this is the system default */
2098 /* store the number of vectors being used */
2099 adapter->num_vectors = 1;
2103 * e1000_request_msix - Initialize MSI-X interrupts
2104 * @adapter: board private structure
2106 * e1000_request_msix allocates MSI-X vectors and requests interrupts from the
2109 static int e1000_request_msix(struct e1000_adapter *adapter)
2111 struct net_device *netdev = adapter->netdev;
2112 int err = 0, vector = 0;
2114 if (strlen(netdev->name) < (IFNAMSIZ - 5))
2115 snprintf(adapter->rx_ring->name,
2116 sizeof(adapter->rx_ring->name) - 1,
2117 "%.14s-rx-0", netdev->name);
2119 memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
2120 err = request_irq(adapter->msix_entries[vector].vector,
2121 e1000_intr_msix_rx, 0, adapter->rx_ring->name,
2125 adapter->rx_ring->itr_register = adapter->hw.hw_addr +
2126 E1000_EITR_82574(vector);
2127 adapter->rx_ring->itr_val = adapter->itr;
2130 if (strlen(netdev->name) < (IFNAMSIZ - 5))
2131 snprintf(adapter->tx_ring->name,
2132 sizeof(adapter->tx_ring->name) - 1,
2133 "%.14s-tx-0", netdev->name);
2135 memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
2136 err = request_irq(adapter->msix_entries[vector].vector,
2137 e1000_intr_msix_tx, 0, adapter->tx_ring->name,
2141 adapter->tx_ring->itr_register = adapter->hw.hw_addr +
2142 E1000_EITR_82574(vector);
2143 adapter->tx_ring->itr_val = adapter->itr;
2146 err = request_irq(adapter->msix_entries[vector].vector,
2147 e1000_msix_other, 0, netdev->name, netdev);
2151 e1000_configure_msix(adapter);
2157 * e1000_request_irq - initialize interrupts
2158 * @adapter: board private structure
2160 * Attempts to configure interrupts using the best available
2161 * capabilities of the hardware and kernel.
2163 static int e1000_request_irq(struct e1000_adapter *adapter)
2165 struct net_device *netdev = adapter->netdev;
2168 if (adapter->msix_entries) {
2169 err = e1000_request_msix(adapter);
2172 /* fall back to MSI */
2173 e1000e_reset_interrupt_capability(adapter);
2174 adapter->int_mode = E1000E_INT_MODE_MSI;
2175 e1000e_set_interrupt_capability(adapter);
2177 if (adapter->flags & FLAG_MSI_ENABLED) {
2178 err = request_irq(adapter->pdev->irq, e1000_intr_msi, 0,
2179 netdev->name, netdev);
2183 /* fall back to legacy interrupt */
2184 e1000e_reset_interrupt_capability(adapter);
2185 adapter->int_mode = E1000E_INT_MODE_LEGACY;
2188 err = request_irq(adapter->pdev->irq, e1000_intr, IRQF_SHARED,
2189 netdev->name, netdev);
2191 e_err("Unable to allocate interrupt, Error: %d\n", err);
2196 static void e1000_free_irq(struct e1000_adapter *adapter)
2198 struct net_device *netdev = adapter->netdev;
2200 if (adapter->msix_entries) {
2203 free_irq(adapter->msix_entries[vector].vector, netdev);
2206 free_irq(adapter->msix_entries[vector].vector, netdev);
2209 /* Other Causes interrupt vector */
2210 free_irq(adapter->msix_entries[vector].vector, netdev);
2214 free_irq(adapter->pdev->irq, netdev);
2218 * e1000_irq_disable - Mask off interrupt generation on the NIC
2219 * @adapter: board private structure
2221 static void e1000_irq_disable(struct e1000_adapter *adapter)
2223 struct e1000_hw *hw = &adapter->hw;
2226 if (adapter->msix_entries)
2227 ew32(EIAC_82574, 0);
2230 if (adapter->msix_entries) {
2233 for (i = 0; i < adapter->num_vectors; i++)
2234 synchronize_irq(adapter->msix_entries[i].vector);
2236 synchronize_irq(adapter->pdev->irq);
2241 * e1000_irq_enable - Enable default interrupt generation settings
2242 * @adapter: board private structure
2244 static void e1000_irq_enable(struct e1000_adapter *adapter)
2246 struct e1000_hw *hw = &adapter->hw;
2248 if (adapter->msix_entries) {
2249 ew32(EIAC_82574, adapter->eiac_mask & E1000_EIAC_MASK_82574);
2250 ew32(IMS, adapter->eiac_mask | E1000_IMS_OTHER |
2252 } else if (hw->mac.type >= e1000_pch_lpt) {
2253 ew32(IMS, IMS_ENABLE_MASK | E1000_IMS_ECCER);
2255 ew32(IMS, IMS_ENABLE_MASK);
2261 * e1000e_get_hw_control - get control of the h/w from f/w
2262 * @adapter: address of board private structure
2264 * e1000e_get_hw_control sets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2265 * For ASF and Pass Through versions of f/w this means that
2266 * the driver is loaded. For AMT version (only with 82573)
2267 * of the f/w this means that the network i/f is open.
2269 void e1000e_get_hw_control(struct e1000_adapter *adapter)
2271 struct e1000_hw *hw = &adapter->hw;
2275 /* Let firmware know the driver has taken over */
2276 if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2278 ew32(SWSM, swsm | E1000_SWSM_DRV_LOAD);
2279 } else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2280 ctrl_ext = er32(CTRL_EXT);
2281 ew32(CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2286 * e1000e_release_hw_control - release control of the h/w to f/w
2287 * @adapter: address of board private structure
2289 * e1000e_release_hw_control resets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2290 * For ASF and Pass Through versions of f/w this means that the
2291 * driver is no longer loaded. For AMT version (only with 82573) i
2292 * of the f/w this means that the network i/f is closed.
2295 void e1000e_release_hw_control(struct e1000_adapter *adapter)
2297 struct e1000_hw *hw = &adapter->hw;
2301 /* Let firmware taken over control of h/w */
2302 if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2304 ew32(SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
2305 } else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2306 ctrl_ext = er32(CTRL_EXT);
2307 ew32(CTRL_EXT, ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2312 * e1000_alloc_ring_dma - allocate memory for a ring structure
2313 * @adapter: board private structure
2314 * @ring: ring struct for which to allocate dma
2316 static int e1000_alloc_ring_dma(struct e1000_adapter *adapter,
2317 struct e1000_ring *ring)
2319 struct pci_dev *pdev = adapter->pdev;
2321 ring->desc = dma_alloc_coherent(&pdev->dev, ring->size, &ring->dma,
2330 * e1000e_setup_tx_resources - allocate Tx resources (Descriptors)
2331 * @tx_ring: Tx descriptor ring
2333 * Return 0 on success, negative on failure
2335 int e1000e_setup_tx_resources(struct e1000_ring *tx_ring)
2337 struct e1000_adapter *adapter = tx_ring->adapter;
2338 int err = -ENOMEM, size;
2340 size = sizeof(struct e1000_buffer) * tx_ring->count;
2341 tx_ring->buffer_info = vzalloc(size);
2342 if (!tx_ring->buffer_info)
2345 /* round up to nearest 4K */
2346 tx_ring->size = tx_ring->count * sizeof(struct e1000_tx_desc);
2347 tx_ring->size = ALIGN(tx_ring->size, 4096);
2349 err = e1000_alloc_ring_dma(adapter, tx_ring);
2353 tx_ring->next_to_use = 0;
2354 tx_ring->next_to_clean = 0;
2358 vfree(tx_ring->buffer_info);
2359 e_err("Unable to allocate memory for the transmit descriptor ring\n");
2364 * e1000e_setup_rx_resources - allocate Rx resources (Descriptors)
2365 * @rx_ring: Rx descriptor ring
2367 * Returns 0 on success, negative on failure
2369 int e1000e_setup_rx_resources(struct e1000_ring *rx_ring)
2371 struct e1000_adapter *adapter = rx_ring->adapter;
2372 struct e1000_buffer *buffer_info;
2373 int i, size, desc_len, err = -ENOMEM;
2375 size = sizeof(struct e1000_buffer) * rx_ring->count;
2376 rx_ring->buffer_info = vzalloc(size);
2377 if (!rx_ring->buffer_info)
2380 for (i = 0; i < rx_ring->count; i++) {
2381 buffer_info = &rx_ring->buffer_info[i];
2382 buffer_info->ps_pages = kcalloc(PS_PAGE_BUFFERS,
2383 sizeof(struct e1000_ps_page),
2385 if (!buffer_info->ps_pages)
2389 desc_len = sizeof(union e1000_rx_desc_packet_split);
2391 /* Round up to nearest 4K */
2392 rx_ring->size = rx_ring->count * desc_len;
2393 rx_ring->size = ALIGN(rx_ring->size, 4096);
2395 err = e1000_alloc_ring_dma(adapter, rx_ring);
2399 rx_ring->next_to_clean = 0;
2400 rx_ring->next_to_use = 0;
2401 rx_ring->rx_skb_top = NULL;
2406 for (i = 0; i < rx_ring->count; i++) {
2407 buffer_info = &rx_ring->buffer_info[i];
2408 kfree(buffer_info->ps_pages);
2411 vfree(rx_ring->buffer_info);
2412 e_err("Unable to allocate memory for the receive descriptor ring\n");
2417 * e1000_clean_tx_ring - Free Tx Buffers
2418 * @tx_ring: Tx descriptor ring
2420 static void e1000_clean_tx_ring(struct e1000_ring *tx_ring)
2422 struct e1000_adapter *adapter = tx_ring->adapter;
2423 struct e1000_buffer *buffer_info;
2427 for (i = 0; i < tx_ring->count; i++) {
2428 buffer_info = &tx_ring->buffer_info[i];
2429 e1000_put_txbuf(tx_ring, buffer_info, false);
2432 netdev_reset_queue(adapter->netdev);
2433 size = sizeof(struct e1000_buffer) * tx_ring->count;
2434 memset(tx_ring->buffer_info, 0, size);
2436 memset(tx_ring->desc, 0, tx_ring->size);
2438 tx_ring->next_to_use = 0;
2439 tx_ring->next_to_clean = 0;
2443 * e1000e_free_tx_resources - Free Tx Resources per Queue
2444 * @tx_ring: Tx descriptor ring
2446 * Free all transmit software resources
2448 void e1000e_free_tx_resources(struct e1000_ring *tx_ring)
2450 struct e1000_adapter *adapter = tx_ring->adapter;
2451 struct pci_dev *pdev = adapter->pdev;
2453 e1000_clean_tx_ring(tx_ring);
2455 vfree(tx_ring->buffer_info);
2456 tx_ring->buffer_info = NULL;
2458 dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
2460 tx_ring->desc = NULL;
2464 * e1000e_free_rx_resources - Free Rx Resources
2465 * @rx_ring: Rx descriptor ring
2467 * Free all receive software resources
2469 void e1000e_free_rx_resources(struct e1000_ring *rx_ring)
2471 struct e1000_adapter *adapter = rx_ring->adapter;
2472 struct pci_dev *pdev = adapter->pdev;
2475 e1000_clean_rx_ring(rx_ring);
2477 for (i = 0; i < rx_ring->count; i++)
2478 kfree(rx_ring->buffer_info[i].ps_pages);
2480 vfree(rx_ring->buffer_info);
2481 rx_ring->buffer_info = NULL;
2483 dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
2485 rx_ring->desc = NULL;
2489 * e1000_update_itr - update the dynamic ITR value based on statistics
2490 * @itr_setting: current adapter->itr
2491 * @packets: the number of packets during this measurement interval
2492 * @bytes: the number of bytes during this measurement interval
2494 * Stores a new ITR value based on packets and byte
2495 * counts during the last interrupt. The advantage of per interrupt
2496 * computation is faster updates and more accurate ITR for the current
2497 * traffic pattern. Constants in this function were computed
2498 * based on theoretical maximum wire speed and thresholds were set based
2499 * on testing data as well as attempting to minimize response time
2500 * while increasing bulk throughput. This functionality is controlled
2501 * by the InterruptThrottleRate module parameter.
2503 static unsigned int e1000_update_itr(u16 itr_setting, int packets, int bytes)
2505 unsigned int retval = itr_setting;
2510 switch (itr_setting) {
2511 case lowest_latency:
2512 /* handle TSO and jumbo frames */
2513 if (bytes / packets > 8000)
2514 retval = bulk_latency;
2515 else if ((packets < 5) && (bytes > 512))
2516 retval = low_latency;
2518 case low_latency: /* 50 usec aka 20000 ints/s */
2519 if (bytes > 10000) {
2520 /* this if handles the TSO accounting */
2521 if (bytes / packets > 8000)
2522 retval = bulk_latency;
2523 else if ((packets < 10) || ((bytes / packets) > 1200))
2524 retval = bulk_latency;
2525 else if ((packets > 35))
2526 retval = lowest_latency;
2527 } else if (bytes / packets > 2000) {
2528 retval = bulk_latency;
2529 } else if (packets <= 2 && bytes < 512) {
2530 retval = lowest_latency;
2533 case bulk_latency: /* 250 usec aka 4000 ints/s */
2534 if (bytes > 25000) {
2536 retval = low_latency;
2537 } else if (bytes < 6000) {
2538 retval = low_latency;
2546 static void e1000_set_itr(struct e1000_adapter *adapter)
2549 u32 new_itr = adapter->itr;
2551 /* for non-gigabit speeds, just fix the interrupt rate at 4000 */
2552 if (adapter->link_speed != SPEED_1000) {
2557 if (adapter->flags2 & FLAG2_DISABLE_AIM) {
2562 adapter->tx_itr = e1000_update_itr(adapter->tx_itr,
2563 adapter->total_tx_packets,
2564 adapter->total_tx_bytes);
2565 /* conservative mode (itr 3) eliminates the lowest_latency setting */
2566 if (adapter->itr_setting == 3 && adapter->tx_itr == lowest_latency)
2567 adapter->tx_itr = low_latency;
2569 adapter->rx_itr = e1000_update_itr(adapter->rx_itr,
2570 adapter->total_rx_packets,
2571 adapter->total_rx_bytes);
2572 /* conservative mode (itr 3) eliminates the lowest_latency setting */
2573 if (adapter->itr_setting == 3 && adapter->rx_itr == lowest_latency)
2574 adapter->rx_itr = low_latency;
2576 current_itr = max(adapter->rx_itr, adapter->tx_itr);
2578 /* counts and packets in update_itr are dependent on these numbers */
2579 switch (current_itr) {
2580 case lowest_latency:
2584 new_itr = 20000; /* aka hwitr = ~200 */
2594 if (new_itr != adapter->itr) {
2595 /* this attempts to bias the interrupt rate towards Bulk
2596 * by adding intermediate steps when interrupt rate is
2599 new_itr = new_itr > adapter->itr ?
2600 min(adapter->itr + (new_itr >> 2), new_itr) : new_itr;
2601 adapter->itr = new_itr;
2602 adapter->rx_ring->itr_val = new_itr;
2603 if (adapter->msix_entries)
2604 adapter->rx_ring->set_itr = 1;
2606 e1000e_write_itr(adapter, new_itr);
2611 * e1000e_write_itr - write the ITR value to the appropriate registers
2612 * @adapter: address of board private structure
2613 * @itr: new ITR value to program
2615 * e1000e_write_itr determines if the adapter is in MSI-X mode
2616 * and, if so, writes the EITR registers with the ITR value.
2617 * Otherwise, it writes the ITR value into the ITR register.
2619 void e1000e_write_itr(struct e1000_adapter *adapter, u32 itr)
2621 struct e1000_hw *hw = &adapter->hw;
2622 u32 new_itr = itr ? 1000000000 / (itr * 256) : 0;
2624 if (adapter->msix_entries) {
2627 for (vector = 0; vector < adapter->num_vectors; vector++)
2628 writel(new_itr, hw->hw_addr + E1000_EITR_82574(vector));
2635 * e1000_alloc_queues - Allocate memory for all rings
2636 * @adapter: board private structure to initialize
2638 static int e1000_alloc_queues(struct e1000_adapter *adapter)
2640 int size = sizeof(struct e1000_ring);
2642 adapter->tx_ring = kzalloc(size, GFP_KERNEL);
2643 if (!adapter->tx_ring)
2645 adapter->tx_ring->count = adapter->tx_ring_count;
2646 adapter->tx_ring->adapter = adapter;
2648 adapter->rx_ring = kzalloc(size, GFP_KERNEL);
2649 if (!adapter->rx_ring)
2651 adapter->rx_ring->count = adapter->rx_ring_count;
2652 adapter->rx_ring->adapter = adapter;
2656 e_err("Unable to allocate memory for queues\n");
2657 kfree(adapter->rx_ring);
2658 kfree(adapter->tx_ring);
2663 * e1000e_poll - NAPI Rx polling callback
2664 * @napi: struct associated with this polling callback
2665 * @budget: number of packets driver is allowed to process this poll
2667 static int e1000e_poll(struct napi_struct *napi, int budget)
2669 struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter,
2671 struct e1000_hw *hw = &adapter->hw;
2672 struct net_device *poll_dev = adapter->netdev;
2673 int tx_cleaned = 1, work_done = 0;
2675 adapter = netdev_priv(poll_dev);
2677 if (!adapter->msix_entries ||
2678 (adapter->rx_ring->ims_val & adapter->tx_ring->ims_val))
2679 tx_cleaned = e1000_clean_tx_irq(adapter->tx_ring);
2681 adapter->clean_rx(adapter->rx_ring, &work_done, budget);
2683 if (!tx_cleaned || work_done == budget)
2686 /* Exit the polling mode, but don't re-enable interrupts if stack might
2687 * poll us due to busy-polling
2689 if (likely(napi_complete_done(napi, work_done))) {
2690 if (adapter->itr_setting & 3)
2691 e1000_set_itr(adapter);
2692 if (!test_bit(__E1000_DOWN, &adapter->state)) {
2693 if (adapter->msix_entries)
2694 ew32(IMS, adapter->rx_ring->ims_val);
2696 e1000_irq_enable(adapter);
2703 static int e1000_vlan_rx_add_vid(struct net_device *netdev,
2704 __always_unused __be16 proto, u16 vid)
2706 struct e1000_adapter *adapter = netdev_priv(netdev);
2707 struct e1000_hw *hw = &adapter->hw;
2710 /* don't update vlan cookie if already programmed */
2711 if ((adapter->hw.mng_cookie.status &
2712 E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2713 (vid == adapter->mng_vlan_id))
2716 /* add VID to filter table */
2717 if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2718 index = (vid >> 5) & 0x7F;
2719 vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2720 vfta |= BIT((vid & 0x1F));
2721 hw->mac.ops.write_vfta(hw, index, vfta);
2724 set_bit(vid, adapter->active_vlans);
2729 static int e1000_vlan_rx_kill_vid(struct net_device *netdev,
2730 __always_unused __be16 proto, u16 vid)
2732 struct e1000_adapter *adapter = netdev_priv(netdev);
2733 struct e1000_hw *hw = &adapter->hw;
2736 if ((adapter->hw.mng_cookie.status &
2737 E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2738 (vid == adapter->mng_vlan_id)) {
2739 /* release control to f/w */
2740 e1000e_release_hw_control(adapter);
2744 /* remove VID from filter table */
2745 if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2746 index = (vid >> 5) & 0x7F;
2747 vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2748 vfta &= ~BIT((vid & 0x1F));
2749 hw->mac.ops.write_vfta(hw, index, vfta);
2752 clear_bit(vid, adapter->active_vlans);
2758 * e1000e_vlan_filter_disable - helper to disable hw VLAN filtering
2759 * @adapter: board private structure to initialize
2761 static void e1000e_vlan_filter_disable(struct e1000_adapter *adapter)
2763 struct net_device *netdev = adapter->netdev;
2764 struct e1000_hw *hw = &adapter->hw;
2767 if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2768 /* disable VLAN receive filtering */
2770 rctl &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
2773 if (adapter->mng_vlan_id != (u16)E1000_MNG_VLAN_NONE) {
2774 e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
2775 adapter->mng_vlan_id);
2776 adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
2782 * e1000e_vlan_filter_enable - helper to enable HW VLAN filtering
2783 * @adapter: board private structure to initialize
2785 static void e1000e_vlan_filter_enable(struct e1000_adapter *adapter)
2787 struct e1000_hw *hw = &adapter->hw;
2790 if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2791 /* enable VLAN receive filtering */
2793 rctl |= E1000_RCTL_VFE;
2794 rctl &= ~E1000_RCTL_CFIEN;
2800 * e1000e_vlan_strip_disable - helper to disable HW VLAN stripping
2801 * @adapter: board private structure to initialize
2803 static void e1000e_vlan_strip_disable(struct e1000_adapter *adapter)
2805 struct e1000_hw *hw = &adapter->hw;
2808 /* disable VLAN tag insert/strip */
2810 ctrl &= ~E1000_CTRL_VME;
2815 * e1000e_vlan_strip_enable - helper to enable HW VLAN stripping
2816 * @adapter: board private structure to initialize
2818 static void e1000e_vlan_strip_enable(struct e1000_adapter *adapter)
2820 struct e1000_hw *hw = &adapter->hw;
2823 /* enable VLAN tag insert/strip */
2825 ctrl |= E1000_CTRL_VME;
2829 static void e1000_update_mng_vlan(struct e1000_adapter *adapter)
2831 struct net_device *netdev = adapter->netdev;
2832 u16 vid = adapter->hw.mng_cookie.vlan_id;
2833 u16 old_vid = adapter->mng_vlan_id;
2835 if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
2836 e1000_vlan_rx_add_vid(netdev, htons(ETH_P_8021Q), vid);
2837 adapter->mng_vlan_id = vid;
2840 if ((old_vid != (u16)E1000_MNG_VLAN_NONE) && (vid != old_vid))
2841 e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q), old_vid);
2844 static void e1000_restore_vlan(struct e1000_adapter *adapter)
2848 e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
2850 for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
2851 e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
2854 static void e1000_init_manageability_pt(struct e1000_adapter *adapter)
2856 struct e1000_hw *hw = &adapter->hw;
2857 u32 manc, manc2h, mdef, i, j;
2859 if (!(adapter->flags & FLAG_MNG_PT_ENABLED))
2864 /* enable receiving management packets to the host. this will probably
2865 * generate destination unreachable messages from the host OS, but
2866 * the packets will be handled on SMBUS
2868 manc |= E1000_MANC_EN_MNG2HOST;
2869 manc2h = er32(MANC2H);
2871 switch (hw->mac.type) {
2873 manc2h |= (E1000_MANC2H_PORT_623 | E1000_MANC2H_PORT_664);
2877 /* Check if IPMI pass-through decision filter already exists;
2880 for (i = 0, j = 0; i < 8; i++) {
2881 mdef = er32(MDEF(i));
2883 /* Ignore filters with anything other than IPMI ports */
2884 if (mdef & ~(E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2887 /* Enable this decision filter in MANC2H */
2894 if (j == (E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2897 /* Create new decision filter in an empty filter */
2898 for (i = 0, j = 0; i < 8; i++)
2899 if (er32(MDEF(i)) == 0) {
2900 ew32(MDEF(i), (E1000_MDEF_PORT_623 |
2901 E1000_MDEF_PORT_664));
2908 e_warn("Unable to create IPMI pass-through filter\n");
2912 ew32(MANC2H, manc2h);
2917 * e1000_configure_tx - Configure Transmit Unit after Reset
2918 * @adapter: board private structure
2920 * Configure the Tx unit of the MAC after a reset.
2922 static void e1000_configure_tx(struct e1000_adapter *adapter)
2924 struct e1000_hw *hw = &adapter->hw;
2925 struct e1000_ring *tx_ring = adapter->tx_ring;
2927 u32 tdlen, tctl, tarc;
2929 /* Setup the HW Tx Head and Tail descriptor pointers */
2930 tdba = tx_ring->dma;
2931 tdlen = tx_ring->count * sizeof(struct e1000_tx_desc);
2932 ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
2933 ew32(TDBAH(0), (tdba >> 32));
2934 ew32(TDLEN(0), tdlen);
2937 tx_ring->head = adapter->hw.hw_addr + E1000_TDH(0);
2938 tx_ring->tail = adapter->hw.hw_addr + E1000_TDT(0);
2940 writel(0, tx_ring->head);
2941 if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
2942 e1000e_update_tdt_wa(tx_ring, 0);
2944 writel(0, tx_ring->tail);
2946 /* Set the Tx Interrupt Delay register */
2947 ew32(TIDV, adapter->tx_int_delay);
2948 /* Tx irq moderation */
2949 ew32(TADV, adapter->tx_abs_int_delay);
2951 if (adapter->flags2 & FLAG2_DMA_BURST) {
2952 u32 txdctl = er32(TXDCTL(0));
2954 txdctl &= ~(E1000_TXDCTL_PTHRESH | E1000_TXDCTL_HTHRESH |
2955 E1000_TXDCTL_WTHRESH);
2956 /* set up some performance related parameters to encourage the
2957 * hardware to use the bus more efficiently in bursts, depends
2958 * on the tx_int_delay to be enabled,
2959 * wthresh = 1 ==> burst write is disabled to avoid Tx stalls
2960 * hthresh = 1 ==> prefetch when one or more available
2961 * pthresh = 0x1f ==> prefetch if internal cache 31 or less
2962 * BEWARE: this seems to work but should be considered first if
2963 * there are Tx hangs or other Tx related bugs
2965 txdctl |= E1000_TXDCTL_DMA_BURST_ENABLE;
2966 ew32(TXDCTL(0), txdctl);
2968 /* erratum work around: set txdctl the same for both queues */
2969 ew32(TXDCTL(1), er32(TXDCTL(0)));
2971 /* Program the Transmit Control Register */
2973 tctl &= ~E1000_TCTL_CT;
2974 tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
2975 (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
2977 if (adapter->flags & FLAG_TARC_SPEED_MODE_BIT) {
2978 tarc = er32(TARC(0));
2979 /* set the speed mode bit, we'll clear it if we're not at
2980 * gigabit link later
2982 #define SPEED_MODE_BIT BIT(21)
2983 tarc |= SPEED_MODE_BIT;
2984 ew32(TARC(0), tarc);
2987 /* errata: program both queues to unweighted RR */
2988 if (adapter->flags & FLAG_TARC_SET_BIT_ZERO) {
2989 tarc = er32(TARC(0));
2991 ew32(TARC(0), tarc);
2992 tarc = er32(TARC(1));
2994 ew32(TARC(1), tarc);
2997 /* Setup Transmit Descriptor Settings for eop descriptor */
2998 adapter->txd_cmd = E1000_TXD_CMD_EOP | E1000_TXD_CMD_IFCS;
3000 /* only set IDE if we are delaying interrupts using the timers */
3001 if (adapter->tx_int_delay)
3002 adapter->txd_cmd |= E1000_TXD_CMD_IDE;
3004 /* enable Report Status bit */
3005 adapter->txd_cmd |= E1000_TXD_CMD_RS;
3009 hw->mac.ops.config_collision_dist(hw);
3011 /* SPT and KBL Si errata workaround to avoid data corruption */
3012 if (hw->mac.type == e1000_pch_spt) {
3015 reg_val = er32(IOSFPC);
3016 reg_val |= E1000_RCTL_RDMTS_HEX;
3017 ew32(IOSFPC, reg_val);
3019 reg_val = er32(TARC(0));
3020 /* SPT and KBL Si errata workaround to avoid Tx hang.
3021 * Dropping the number of outstanding requests from
3022 * 3 to 2 in order to avoid a buffer overrun.
3024 reg_val &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3025 reg_val |= E1000_TARC0_CB_MULTIQ_2_REQ;
3026 ew32(TARC(0), reg_val);
3030 #define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
3031 (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
3034 * e1000_setup_rctl - configure the receive control registers
3035 * @adapter: Board private structure
3037 static void e1000_setup_rctl(struct e1000_adapter *adapter)
3039 struct e1000_hw *hw = &adapter->hw;
3043 /* Workaround Si errata on PCHx - configure jumbo frame flow.
3044 * If jumbo frames not set, program related MAC/PHY registers
3047 if (hw->mac.type >= e1000_pch2lan) {
3050 if (adapter->netdev->mtu > ETH_DATA_LEN)
3051 ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, true);
3053 ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, false);
3056 e_dbg("failed to enable|disable jumbo frame workaround mode\n");
3059 /* Program MC offset vector base */
3061 rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3062 rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3063 E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3064 (adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3066 /* Do not Store bad packets */
3067 rctl &= ~E1000_RCTL_SBP;
3069 /* Enable Long Packet receive */
3070 if (adapter->netdev->mtu <= ETH_DATA_LEN)
3071 rctl &= ~E1000_RCTL_LPE;
3073 rctl |= E1000_RCTL_LPE;
3075 /* Some systems expect that the CRC is included in SMBUS traffic. The
3076 * hardware strips the CRC before sending to both SMBUS (BMC) and to
3077 * host memory when this is enabled
3079 if (adapter->flags2 & FLAG2_CRC_STRIPPING)
3080 rctl |= E1000_RCTL_SECRC;
3082 /* Workaround Si errata on 82577 PHY - configure IPG for jumbos */
3083 if ((hw->phy.type == e1000_phy_82577) && (rctl & E1000_RCTL_LPE)) {
3086 e1e_rphy(hw, PHY_REG(770, 26), &phy_data);
3089 e1e_wphy(hw, PHY_REG(770, 26), phy_data);
3091 e1e_rphy(hw, 22, &phy_data);
3093 phy_data |= BIT(14);
3094 e1e_wphy(hw, 0x10, 0x2823);
3095 e1e_wphy(hw, 0x11, 0x0003);
3096 e1e_wphy(hw, 22, phy_data);
3099 /* Setup buffer sizes */
3100 rctl &= ~E1000_RCTL_SZ_4096;
3101 rctl |= E1000_RCTL_BSEX;
3102 switch (adapter->rx_buffer_len) {
3105 rctl |= E1000_RCTL_SZ_2048;
3106 rctl &= ~E1000_RCTL_BSEX;
3109 rctl |= E1000_RCTL_SZ_4096;
3112 rctl |= E1000_RCTL_SZ_8192;
3115 rctl |= E1000_RCTL_SZ_16384;
3119 /* Enable Extended Status in all Receive Descriptors */
3120 rfctl = er32(RFCTL);
3121 rfctl |= E1000_RFCTL_EXTEN;
3124 /* 82571 and greater support packet-split where the protocol
3125 * header is placed in skb->data and the packet data is
3126 * placed in pages hanging off of skb_shinfo(skb)->nr_frags.
3127 * In the case of a non-split, skb->data is linearly filled,
3128 * followed by the page buffers. Therefore, skb->data is
3129 * sized to hold the largest protocol header.
3131 * allocations using alloc_page take too long for regular MTU
3132 * so only enable packet split for jumbo frames
3134 * Using pages when the page size is greater than 16k wastes
3135 * a lot of memory, since we allocate 3 pages at all times
3138 pages = PAGE_USE_COUNT(adapter->netdev->mtu);
3139 if ((pages <= 3) && (PAGE_SIZE <= 16384) && (rctl & E1000_RCTL_LPE))
3140 adapter->rx_ps_pages = pages;
3142 adapter->rx_ps_pages = 0;
3144 if (adapter->rx_ps_pages) {
3147 /* Enable Packet split descriptors */
3148 rctl |= E1000_RCTL_DTYP_PS;
3150 psrctl |= adapter->rx_ps_bsize0 >> E1000_PSRCTL_BSIZE0_SHIFT;
3152 switch (adapter->rx_ps_pages) {
3154 psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE3_SHIFT;
3157 psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE2_SHIFT;
3160 psrctl |= PAGE_SIZE >> E1000_PSRCTL_BSIZE1_SHIFT;
3164 ew32(PSRCTL, psrctl);
3167 /* This is useful for sniffing bad packets. */
3168 if (adapter->netdev->features & NETIF_F_RXALL) {
3169 /* UPE and MPE will be handled by normal PROMISC logic
3170 * in e1000e_set_rx_mode
3172 rctl |= (E1000_RCTL_SBP | /* Receive bad packets */
3173 E1000_RCTL_BAM | /* RX All Bcast Pkts */
3174 E1000_RCTL_PMCF); /* RX All MAC Ctrl Pkts */
3176 rctl &= ~(E1000_RCTL_VFE | /* Disable VLAN filter */
3177 E1000_RCTL_DPF | /* Allow filtered pause */
3178 E1000_RCTL_CFIEN); /* Dis VLAN CFIEN Filter */
3179 /* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3180 * and that breaks VLANs.
3185 /* just started the receive unit, no need to restart */
3186 adapter->flags &= ~FLAG_RESTART_NOW;
3190 * e1000_configure_rx - Configure Receive Unit after Reset
3191 * @adapter: board private structure
3193 * Configure the Rx unit of the MAC after a reset.
3195 static void e1000_configure_rx(struct e1000_adapter *adapter)
3197 struct e1000_hw *hw = &adapter->hw;
3198 struct e1000_ring *rx_ring = adapter->rx_ring;
3200 u32 rdlen, rctl, rxcsum, ctrl_ext;
3202 if (adapter->rx_ps_pages) {
3203 /* this is a 32 byte descriptor */
3204 rdlen = rx_ring->count *
3205 sizeof(union e1000_rx_desc_packet_split);
3206 adapter->clean_rx = e1000_clean_rx_irq_ps;
3207 adapter->alloc_rx_buf = e1000_alloc_rx_buffers_ps;
3208 } else if (adapter->netdev->mtu > ETH_FRAME_LEN + ETH_FCS_LEN) {
3209 rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3210 adapter->clean_rx = e1000_clean_jumbo_rx_irq;
3211 adapter->alloc_rx_buf = e1000_alloc_jumbo_rx_buffers;
3213 rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3214 adapter->clean_rx = e1000_clean_rx_irq;
3215 adapter->alloc_rx_buf = e1000_alloc_rx_buffers;
3218 /* disable receives while setting up the descriptors */
3220 if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
3221 ew32(RCTL, rctl & ~E1000_RCTL_EN);
3223 usleep_range(10000, 11000);
3225 if (adapter->flags2 & FLAG2_DMA_BURST) {
3226 /* set the writeback threshold (only takes effect if the RDTR
3227 * is set). set GRAN=1 and write back up to 0x4 worth, and
3228 * enable prefetching of 0x20 Rx descriptors
3234 ew32(RXDCTL(0), E1000_RXDCTL_DMA_BURST_ENABLE);
3235 ew32(RXDCTL(1), E1000_RXDCTL_DMA_BURST_ENABLE);
3238 /* set the Receive Delay Timer Register */
3239 ew32(RDTR, adapter->rx_int_delay);
3241 /* irq moderation */
3242 ew32(RADV, adapter->rx_abs_int_delay);
3243 if ((adapter->itr_setting != 0) && (adapter->itr != 0))
3244 e1000e_write_itr(adapter, adapter->itr);
3246 ctrl_ext = er32(CTRL_EXT);
3247 /* Auto-Mask interrupts upon ICR access */
3248 ctrl_ext |= E1000_CTRL_EXT_IAME;
3249 ew32(IAM, 0xffffffff);
3250 ew32(CTRL_EXT, ctrl_ext);
3253 /* Setup the HW Rx Head and Tail Descriptor Pointers and
3254 * the Base and Length of the Rx Descriptor Ring
3256 rdba = rx_ring->dma;
3257 ew32(RDBAL(0), (rdba & DMA_BIT_MASK(32)));
3258 ew32(RDBAH(0), (rdba >> 32));
3259 ew32(RDLEN(0), rdlen);
3262 rx_ring->head = adapter->hw.hw_addr + E1000_RDH(0);
3263 rx_ring->tail = adapter->hw.hw_addr + E1000_RDT(0);
3265 writel(0, rx_ring->head);
3266 if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
3267 e1000e_update_rdt_wa(rx_ring, 0);
3269 writel(0, rx_ring->tail);
3271 /* Enable Receive Checksum Offload for TCP and UDP */
3272 rxcsum = er32(RXCSUM);
3273 if (adapter->netdev->features & NETIF_F_RXCSUM)
3274 rxcsum |= E1000_RXCSUM_TUOFL;
3276 rxcsum &= ~E1000_RXCSUM_TUOFL;
3277 ew32(RXCSUM, rxcsum);
3279 /* With jumbo frames, excessive C-state transition latencies result
3280 * in dropped transactions.
3282 if (adapter->netdev->mtu > ETH_DATA_LEN) {
3284 ((er32(PBA) & E1000_PBA_RXA_MASK) * 1024 -
3285 adapter->max_frame_size) * 8 / 1000;
3287 if (adapter->flags & FLAG_IS_ICH) {
3288 u32 rxdctl = er32(RXDCTL(0));
3290 ew32(RXDCTL(0), rxdctl | 0x3 | BIT(8));
3293 dev_info(&adapter->pdev->dev,
3294 "Some CPU C-states have been disabled in order to enable jumbo frames\n");
3295 cpu_latency_qos_update_request(&adapter->pm_qos_req, lat);
3297 cpu_latency_qos_update_request(&adapter->pm_qos_req,
3298 PM_QOS_DEFAULT_VALUE);
3301 /* Enable Receives */
3306 * e1000e_write_mc_addr_list - write multicast addresses to MTA
3307 * @netdev: network interface device structure
3309 * Writes multicast address list to the MTA hash table.
3310 * Returns: -ENOMEM on failure
3311 * 0 on no addresses written
3312 * X on writing X addresses to MTA
3314 static int e1000e_write_mc_addr_list(struct net_device *netdev)
3316 struct e1000_adapter *adapter = netdev_priv(netdev);
3317 struct e1000_hw *hw = &adapter->hw;
3318 struct netdev_hw_addr *ha;
3322 if (netdev_mc_empty(netdev)) {
3323 /* nothing to program, so clear mc list */
3324 hw->mac.ops.update_mc_addr_list(hw, NULL, 0);
3328 mta_list = kcalloc(netdev_mc_count(netdev), ETH_ALEN, GFP_ATOMIC);
3332 /* update_mc_addr_list expects a packed array of only addresses. */
3334 netdev_for_each_mc_addr(ha, netdev)
3335 memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3337 hw->mac.ops.update_mc_addr_list(hw, mta_list, i);
3340 return netdev_mc_count(netdev);
3344 * e1000e_write_uc_addr_list - write unicast addresses to RAR table
3345 * @netdev: network interface device structure
3347 * Writes unicast address list to the RAR table.
3348 * Returns: -ENOMEM on failure/insufficient address space
3349 * 0 on no addresses written
3350 * X on writing X addresses to the RAR table
3352 static int e1000e_write_uc_addr_list(struct net_device *netdev)
3354 struct e1000_adapter *adapter = netdev_priv(netdev);
3355 struct e1000_hw *hw = &adapter->hw;
3356 unsigned int rar_entries;
3359 rar_entries = hw->mac.ops.rar_get_count(hw);
3361 /* save a rar entry for our hardware address */
3364 /* save a rar entry for the LAA workaround */
3365 if (adapter->flags & FLAG_RESET_OVERWRITES_LAA)
3368 /* return ENOMEM indicating insufficient memory for addresses */
3369 if (netdev_uc_count(netdev) > rar_entries)
3372 if (!netdev_uc_empty(netdev) && rar_entries) {
3373 struct netdev_hw_addr *ha;
3375 /* write the addresses in reverse order to avoid write
3378 netdev_for_each_uc_addr(ha, netdev) {
3383 ret_val = hw->mac.ops.rar_set(hw, ha->addr, rar_entries--);
3390 /* zero out the remaining RAR entries not used above */
3391 for (; rar_entries > 0; rar_entries--) {
3392 ew32(RAH(rar_entries), 0);
3393 ew32(RAL(rar_entries), 0);
3401 * e1000e_set_rx_mode - secondary unicast, Multicast and Promiscuous mode set
3402 * @netdev: network interface device structure
3404 * The ndo_set_rx_mode entry point is called whenever the unicast or multicast
3405 * address list or the network interface flags are updated. This routine is
3406 * responsible for configuring the hardware for proper unicast, multicast,
3407 * promiscuous mode, and all-multi behavior.
3409 static void e1000e_set_rx_mode(struct net_device *netdev)
3411 struct e1000_adapter *adapter = netdev_priv(netdev);
3412 struct e1000_hw *hw = &adapter->hw;
3415 if (pm_runtime_suspended(netdev->dev.parent))
3418 /* Check for Promiscuous and All Multicast modes */
3421 /* clear the affected bits */
3422 rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
3424 if (netdev->flags & IFF_PROMISC) {
3425 rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
3426 /* Do not hardware filter VLANs in promisc mode */
3427 e1000e_vlan_filter_disable(adapter);
3431 if (netdev->flags & IFF_ALLMULTI) {
3432 rctl |= E1000_RCTL_MPE;
3434 /* Write addresses to the MTA, if the attempt fails
3435 * then we should just turn on promiscuous mode so
3436 * that we can at least receive multicast traffic
3438 count = e1000e_write_mc_addr_list(netdev);
3440 rctl |= E1000_RCTL_MPE;
3442 e1000e_vlan_filter_enable(adapter);
3443 /* Write addresses to available RAR registers, if there is not
3444 * sufficient space to store all the addresses then enable
3445 * unicast promiscuous mode
3447 count = e1000e_write_uc_addr_list(netdev);
3449 rctl |= E1000_RCTL_UPE;
3454 if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
3455 e1000e_vlan_strip_enable(adapter);
3457 e1000e_vlan_strip_disable(adapter);
3460 static void e1000e_setup_rss_hash(struct e1000_adapter *adapter)
3462 struct e1000_hw *hw = &adapter->hw;
3467 netdev_rss_key_fill(rss_key, sizeof(rss_key));
3468 for (i = 0; i < 10; i++)
3469 ew32(RSSRK(i), rss_key[i]);
3471 /* Direct all traffic to queue 0 */
3472 for (i = 0; i < 32; i++)
3475 /* Disable raw packet checksumming so that RSS hash is placed in
3476 * descriptor on writeback.
3478 rxcsum = er32(RXCSUM);
3479 rxcsum |= E1000_RXCSUM_PCSD;
3481 ew32(RXCSUM, rxcsum);
3483 mrqc = (E1000_MRQC_RSS_FIELD_IPV4 |
3484 E1000_MRQC_RSS_FIELD_IPV4_TCP |
3485 E1000_MRQC_RSS_FIELD_IPV6 |
3486 E1000_MRQC_RSS_FIELD_IPV6_TCP |
3487 E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3493 * e1000e_get_base_timinca - get default SYSTIM time increment attributes
3494 * @adapter: board private structure
3495 * @timinca: pointer to returned time increment attributes
3497 * Get attributes for incrementing the System Time Register SYSTIML/H at
3498 * the default base frequency, and set the cyclecounter shift value.
3500 s32 e1000e_get_base_timinca(struct e1000_adapter *adapter, u32 *timinca)
3502 struct e1000_hw *hw = &adapter->hw;
3503 u32 incvalue, incperiod, shift;
3505 /* Make sure clock is enabled on I217/I218/I219 before checking
3508 if ((hw->mac.type >= e1000_pch_lpt) &&
3509 !(er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) &&
3510 !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_ENABLED)) {
3511 u32 fextnvm7 = er32(FEXTNVM7);
3513 if (!(fextnvm7 & BIT(0))) {
3514 ew32(FEXTNVM7, fextnvm7 | BIT(0));
3519 switch (hw->mac.type) {
3521 /* Stable 96MHz frequency */
3522 incperiod = INCPERIOD_96MHZ;
3523 incvalue = INCVALUE_96MHZ;
3524 shift = INCVALUE_SHIFT_96MHZ;
3525 adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3528 if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3529 /* Stable 96MHz frequency */
3530 incperiod = INCPERIOD_96MHZ;
3531 incvalue = INCVALUE_96MHZ;
3532 shift = INCVALUE_SHIFT_96MHZ;
3533 adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3535 /* Stable 25MHz frequency */
3536 incperiod = INCPERIOD_25MHZ;
3537 incvalue = INCVALUE_25MHZ;
3538 shift = INCVALUE_SHIFT_25MHZ;
3539 adapter->cc.shift = shift;
3543 /* Stable 24MHz frequency */
3544 incperiod = INCPERIOD_24MHZ;
3545 incvalue = INCVALUE_24MHZ;
3546 shift = INCVALUE_SHIFT_24MHZ;
3547 adapter->cc.shift = shift;
3554 if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3555 /* Stable 24MHz frequency */
3556 incperiod = INCPERIOD_24MHZ;
3557 incvalue = INCVALUE_24MHZ;
3558 shift = INCVALUE_SHIFT_24MHZ;
3559 adapter->cc.shift = shift;
3561 /* Stable 38400KHz frequency */
3562 incperiod = INCPERIOD_38400KHZ;
3563 incvalue = INCVALUE_38400KHZ;
3564 shift = INCVALUE_SHIFT_38400KHZ;
3565 adapter->cc.shift = shift;
3570 /* Stable 25MHz frequency */
3571 incperiod = INCPERIOD_25MHZ;
3572 incvalue = INCVALUE_25MHZ;
3573 shift = INCVALUE_SHIFT_25MHZ;
3574 adapter->cc.shift = shift;
3580 *timinca = ((incperiod << E1000_TIMINCA_INCPERIOD_SHIFT) |
3581 ((incvalue << shift) & E1000_TIMINCA_INCVALUE_MASK));
3587 * e1000e_config_hwtstamp - configure the hwtstamp registers and enable/disable
3588 * @adapter: board private structure
3589 * @config: timestamp configuration
3591 * Outgoing time stamping can be enabled and disabled. Play nice and
3592 * disable it when requested, although it shouldn't cause any overhead
3593 * when no packet needs it. At most one packet in the queue may be
3594 * marked for time stamping, otherwise it would be impossible to tell
3595 * for sure to which packet the hardware time stamp belongs.
3597 * Incoming time stamping has to be configured via the hardware filters.
3598 * Not all combinations are supported, in particular event type has to be
3599 * specified. Matching the kind of event packet is not supported, with the
3600 * exception of "all V2 events regardless of level 2 or 4".
3602 static int e1000e_config_hwtstamp(struct e1000_adapter *adapter,
3603 struct hwtstamp_config *config)
3605 struct e1000_hw *hw = &adapter->hw;
3606 u32 tsync_tx_ctl = E1000_TSYNCTXCTL_ENABLED;
3607 u32 tsync_rx_ctl = E1000_TSYNCRXCTL_ENABLED;
3614 if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3617 switch (config->tx_type) {
3618 case HWTSTAMP_TX_OFF:
3621 case HWTSTAMP_TX_ON:
3627 switch (config->rx_filter) {
3628 case HWTSTAMP_FILTER_NONE:
3631 case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3632 tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3633 rxmtrl = E1000_RXMTRL_PTP_V1_SYNC_MESSAGE;
3636 case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3637 tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3638 rxmtrl = E1000_RXMTRL_PTP_V1_DELAY_REQ_MESSAGE;
3641 case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3642 /* Also time stamps V2 L2 Path Delay Request/Response */
3643 tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3644 rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3647 case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3648 /* Also time stamps V2 L2 Path Delay Request/Response. */
3649 tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3650 rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3653 case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3654 /* Hardware cannot filter just V2 L4 Sync messages */
3656 case HWTSTAMP_FILTER_PTP_V2_SYNC:
3657 /* Also time stamps V2 Path Delay Request/Response. */
3658 tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3659 rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3663 case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3664 /* Hardware cannot filter just V2 L4 Delay Request messages */
3666 case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
3667 /* Also time stamps V2 Path Delay Request/Response. */
3668 tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3669 rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3673 case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
3674 case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
3675 /* Hardware cannot filter just V2 L4 or L2 Event messages */
3677 case HWTSTAMP_FILTER_PTP_V2_EVENT:
3678 tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_EVENT_V2;
3679 config->rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
3683 case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3684 /* For V1, the hardware can only filter Sync messages or
3685 * Delay Request messages but not both so fall-through to
3686 * time stamp all packets.
3689 case HWTSTAMP_FILTER_NTP_ALL:
3690 case HWTSTAMP_FILTER_ALL:
3693 tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_ALL;
3694 config->rx_filter = HWTSTAMP_FILTER_ALL;
3700 adapter->hwtstamp_config = *config;
3702 /* enable/disable Tx h/w time stamping */
3703 regval = er32(TSYNCTXCTL);
3704 regval &= ~E1000_TSYNCTXCTL_ENABLED;
3705 regval |= tsync_tx_ctl;
3706 ew32(TSYNCTXCTL, regval);
3707 if ((er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) !=
3708 (regval & E1000_TSYNCTXCTL_ENABLED)) {
3709 e_err("Timesync Tx Control register not set as expected\n");
3713 /* enable/disable Rx h/w time stamping */
3714 regval = er32(TSYNCRXCTL);
3715 regval &= ~(E1000_TSYNCRXCTL_ENABLED | E1000_TSYNCRXCTL_TYPE_MASK);
3716 regval |= tsync_rx_ctl;
3717 ew32(TSYNCRXCTL, regval);
3718 if ((er32(TSYNCRXCTL) & (E1000_TSYNCRXCTL_ENABLED |
3719 E1000_TSYNCRXCTL_TYPE_MASK)) !=
3720 (regval & (E1000_TSYNCRXCTL_ENABLED |
3721 E1000_TSYNCRXCTL_TYPE_MASK))) {
3722 e_err("Timesync Rx Control register not set as expected\n");
3726 /* L2: define ethertype filter for time stamped packets */
3728 rxmtrl |= ETH_P_1588;
3730 /* define which PTP packets get time stamped */
3731 ew32(RXMTRL, rxmtrl);
3733 /* Filter by destination port */
3735 rxudp = PTP_EV_PORT;
3736 cpu_to_be16s(&rxudp);
3742 /* Clear TSYNCRXCTL_VALID & TSYNCTXCTL_VALID bit */
3750 * e1000_configure - configure the hardware for Rx and Tx
3751 * @adapter: private board structure
3753 static void e1000_configure(struct e1000_adapter *adapter)
3755 struct e1000_ring *rx_ring = adapter->rx_ring;
3757 e1000e_set_rx_mode(adapter->netdev);
3759 e1000_restore_vlan(adapter);
3760 e1000_init_manageability_pt(adapter);
3762 e1000_configure_tx(adapter);
3764 if (adapter->netdev->features & NETIF_F_RXHASH)
3765 e1000e_setup_rss_hash(adapter);
3766 e1000_setup_rctl(adapter);
3767 e1000_configure_rx(adapter);
3768 adapter->alloc_rx_buf(rx_ring, e1000_desc_unused(rx_ring), GFP_KERNEL);
3772 * e1000e_power_up_phy - restore link in case the phy was powered down
3773 * @adapter: address of board private structure
3775 * The phy may be powered down to save power and turn off link when the
3776 * driver is unloaded and wake on lan is not enabled (among others)
3777 * *** this routine MUST be followed by a call to e1000e_reset ***
3779 void e1000e_power_up_phy(struct e1000_adapter *adapter)
3781 if (adapter->hw.phy.ops.power_up)
3782 adapter->hw.phy.ops.power_up(&adapter->hw);
3784 adapter->hw.mac.ops.setup_link(&adapter->hw);
3788 * e1000_power_down_phy - Power down the PHY
3789 * @adapter: board private structure
3791 * Power down the PHY so no link is implied when interface is down.
3792 * The PHY cannot be powered down if management or WoL is active.
3794 static void e1000_power_down_phy(struct e1000_adapter *adapter)
3796 if (adapter->hw.phy.ops.power_down)
3797 adapter->hw.phy.ops.power_down(&adapter->hw);
3801 * e1000_flush_tx_ring - remove all descriptors from the tx_ring
3802 * @adapter: board private structure
3804 * We want to clear all pending descriptors from the TX ring.
3805 * zeroing happens when the HW reads the regs. We assign the ring itself as
3806 * the data of the next descriptor. We don't care about the data we are about
3809 static void e1000_flush_tx_ring(struct e1000_adapter *adapter)
3811 struct e1000_hw *hw = &adapter->hw;
3812 struct e1000_ring *tx_ring = adapter->tx_ring;
3813 struct e1000_tx_desc *tx_desc = NULL;
3814 u32 tdt, tctl, txd_lower = E1000_TXD_CMD_IFCS;
3818 ew32(TCTL, tctl | E1000_TCTL_EN);
3820 BUG_ON(tdt != tx_ring->next_to_use);
3821 tx_desc = E1000_TX_DESC(*tx_ring, tx_ring->next_to_use);
3822 tx_desc->buffer_addr = cpu_to_le64(tx_ring->dma);
3824 tx_desc->lower.data = cpu_to_le32(txd_lower | size);
3825 tx_desc->upper.data = 0;
3826 /* flush descriptors to memory before notifying the HW */
3828 tx_ring->next_to_use++;
3829 if (tx_ring->next_to_use == tx_ring->count)
3830 tx_ring->next_to_use = 0;
3831 ew32(TDT(0), tx_ring->next_to_use);
3832 usleep_range(200, 250);
3836 * e1000_flush_rx_ring - remove all descriptors from the rx_ring
3837 * @adapter: board private structure
3839 * Mark all descriptors in the RX ring as consumed and disable the rx ring
3841 static void e1000_flush_rx_ring(struct e1000_adapter *adapter)
3844 struct e1000_hw *hw = &adapter->hw;
3847 ew32(RCTL, rctl & ~E1000_RCTL_EN);
3849 usleep_range(100, 150);
3851 rxdctl = er32(RXDCTL(0));
3852 /* zero the lower 14 bits (prefetch and host thresholds) */
3853 rxdctl &= 0xffffc000;
3855 /* update thresholds: prefetch threshold to 31, host threshold to 1
3856 * and make sure the granularity is "descriptors" and not "cache lines"
3858 rxdctl |= (0x1F | BIT(8) | E1000_RXDCTL_THRESH_UNIT_DESC);
3860 ew32(RXDCTL(0), rxdctl);
3861 /* momentarily enable the RX ring for the changes to take effect */
3862 ew32(RCTL, rctl | E1000_RCTL_EN);
3864 usleep_range(100, 150);
3865 ew32(RCTL, rctl & ~E1000_RCTL_EN);
3869 * e1000_flush_desc_rings - remove all descriptors from the descriptor rings
3870 * @adapter: board private structure
3872 * In i219, the descriptor rings must be emptied before resetting the HW
3873 * or before changing the device state to D3 during runtime (runtime PM).
3875 * Failure to do this will cause the HW to enter a unit hang state which can
3876 * only be released by PCI reset on the device
3880 static void e1000_flush_desc_rings(struct e1000_adapter *adapter)
3883 u32 fext_nvm11, tdlen;
3884 struct e1000_hw *hw = &adapter->hw;
3886 /* First, disable MULR fix in FEXTNVM11 */
3887 fext_nvm11 = er32(FEXTNVM11);
3888 fext_nvm11 |= E1000_FEXTNVM11_DISABLE_MULR_FIX;
3889 ew32(FEXTNVM11, fext_nvm11);
3890 /* do nothing if we're not in faulty state, or if the queue is empty */
3891 tdlen = er32(TDLEN(0));
3892 pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3894 if (!(hang_state & FLUSH_DESC_REQUIRED) || !tdlen)
3896 e1000_flush_tx_ring(adapter);
3897 /* recheck, maybe the fault is caused by the rx ring */
3898 pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3900 if (hang_state & FLUSH_DESC_REQUIRED)
3901 e1000_flush_rx_ring(adapter);
3905 * e1000e_systim_reset - reset the timesync registers after a hardware reset
3906 * @adapter: board private structure
3908 * When the MAC is reset, all hardware bits for timesync will be reset to the
3909 * default values. This function will restore the settings last in place.
3910 * Since the clock SYSTIME registers are reset, we will simply restore the
3911 * cyclecounter to the kernel real clock time.
3913 static void e1000e_systim_reset(struct e1000_adapter *adapter)
3915 struct ptp_clock_info *info = &adapter->ptp_clock_info;
3916 struct e1000_hw *hw = &adapter->hw;
3917 unsigned long flags;
3921 if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3924 if (info->adjfreq) {
3925 /* restore the previous ptp frequency delta */
3926 ret_val = info->adjfreq(info, adapter->ptp_delta);
3928 /* set the default base frequency if no adjustment possible */
3929 ret_val = e1000e_get_base_timinca(adapter, &timinca);
3931 ew32(TIMINCA, timinca);
3935 dev_warn(&adapter->pdev->dev,
3936 "Failed to restore TIMINCA clock rate delta: %d\n",
3941 /* reset the systim ns time counter */
3942 spin_lock_irqsave(&adapter->systim_lock, flags);
3943 timecounter_init(&adapter->tc, &adapter->cc,
3944 ktime_to_ns(ktime_get_real()));
3945 spin_unlock_irqrestore(&adapter->systim_lock, flags);
3947 /* restore the previous hwtstamp configuration settings */
3948 e1000e_config_hwtstamp(adapter, &adapter->hwtstamp_config);
3952 * e1000e_reset - bring the hardware into a known good state
3953 * @adapter: board private structure
3955 * This function boots the hardware and enables some settings that
3956 * require a configuration cycle of the hardware - those cannot be
3957 * set/changed during runtime. After reset the device needs to be
3958 * properly configured for Rx, Tx etc.
3960 void e1000e_reset(struct e1000_adapter *adapter)
3962 struct e1000_mac_info *mac = &adapter->hw.mac;
3963 struct e1000_fc_info *fc = &adapter->hw.fc;
3964 struct e1000_hw *hw = &adapter->hw;
3965 u32 tx_space, min_tx_space, min_rx_space;
3966 u32 pba = adapter->pba;
3969 /* reset Packet Buffer Allocation to default */
3972 if (adapter->max_frame_size > (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN)) {
3973 /* To maintain wire speed transmits, the Tx FIFO should be
3974 * large enough to accommodate two full transmit packets,
3975 * rounded up to the next 1KB and expressed in KB. Likewise,
3976 * the Rx FIFO should be large enough to accommodate at least
3977 * one full receive packet and is similarly rounded up and
3981 /* upper 16 bits has Tx packet buffer allocation size in KB */
3982 tx_space = pba >> 16;
3983 /* lower 16 bits has Rx packet buffer allocation size in KB */
3985 /* the Tx fifo also stores 16 bytes of information about the Tx
3986 * but don't include ethernet FCS because hardware appends it
3988 min_tx_space = (adapter->max_frame_size +
3989 sizeof(struct e1000_tx_desc) - ETH_FCS_LEN) * 2;
3990 min_tx_space = ALIGN(min_tx_space, 1024);
3991 min_tx_space >>= 10;
3992 /* software strips receive CRC, so leave room for it */
3993 min_rx_space = adapter->max_frame_size;
3994 min_rx_space = ALIGN(min_rx_space, 1024);
3995 min_rx_space >>= 10;
3997 /* If current Tx allocation is less than the min Tx FIFO size,
3998 * and the min Tx FIFO size is less than the current Rx FIFO
3999 * allocation, take space away from current Rx allocation
4001 if ((tx_space < min_tx_space) &&
4002 ((min_tx_space - tx_space) < pba)) {
4003 pba -= min_tx_space - tx_space;
4005 /* if short on Rx space, Rx wins and must trump Tx
4008 if (pba < min_rx_space)
4015 /* flow control settings
4017 * The high water mark must be low enough to fit one full frame
4018 * (or the size used for early receive) above it in the Rx FIFO.
4019 * Set it to the lower of:
4020 * - 90% of the Rx FIFO size, and
4021 * - the full Rx FIFO size minus one full frame
4023 if (adapter->flags & FLAG_DISABLE_FC_PAUSE_TIME)
4024 fc->pause_time = 0xFFFF;
4026 fc->pause_time = E1000_FC_PAUSE_TIME;
4027 fc->send_xon = true;
4028 fc->current_mode = fc->requested_mode;
4030 switch (hw->mac.type) {
4032 case e1000_ich10lan:
4033 if (adapter->netdev->mtu > ETH_DATA_LEN) {
4036 fc->high_water = 0x2800;
4037 fc->low_water = fc->high_water - 8;
4042 hwm = min(((pba << 10) * 9 / 10),
4043 ((pba << 10) - adapter->max_frame_size));
4045 fc->high_water = hwm & E1000_FCRTH_RTH; /* 8-byte granularity */
4046 fc->low_water = fc->high_water - 8;
4049 /* Workaround PCH LOM adapter hangs with certain network
4050 * loads. If hangs persist, try disabling Tx flow control.
4052 if (adapter->netdev->mtu > ETH_DATA_LEN) {
4053 fc->high_water = 0x3500;
4054 fc->low_water = 0x1500;
4056 fc->high_water = 0x5000;
4057 fc->low_water = 0x3000;
4059 fc->refresh_time = 0x1000;
4069 fc->refresh_time = 0xFFFF;
4070 fc->pause_time = 0xFFFF;
4072 if (adapter->netdev->mtu <= ETH_DATA_LEN) {
4073 fc->high_water = 0x05C20;
4074 fc->low_water = 0x05048;
4080 fc->high_water = ((pba << 10) * 9 / 10) & E1000_FCRTH_RTH;
4081 fc->low_water = ((pba << 10) * 8 / 10) & E1000_FCRTL_RTL;
4085 /* Alignment of Tx data is on an arbitrary byte boundary with the
4086 * maximum size per Tx descriptor limited only to the transmit
4087 * allocation of the packet buffer minus 96 bytes with an upper
4088 * limit of 24KB due to receive synchronization limitations.
4090 adapter->tx_fifo_limit = min_t(u32, ((er32(PBA) >> 16) << 10) - 96,
4093 /* Disable Adaptive Interrupt Moderation if 2 full packets cannot
4094 * fit in receive buffer.
4096 if (adapter->itr_setting & 0x3) {
4097 if ((adapter->max_frame_size * 2) > (pba << 10)) {
4098 if (!(adapter->flags2 & FLAG2_DISABLE_AIM)) {
4099 dev_info(&adapter->pdev->dev,
4100 "Interrupt Throttle Rate off\n");
4101 adapter->flags2 |= FLAG2_DISABLE_AIM;
4102 e1000e_write_itr(adapter, 0);
4104 } else if (adapter->flags2 & FLAG2_DISABLE_AIM) {
4105 dev_info(&adapter->pdev->dev,
4106 "Interrupt Throttle Rate on\n");
4107 adapter->flags2 &= ~FLAG2_DISABLE_AIM;
4108 adapter->itr = 20000;
4109 e1000e_write_itr(adapter, adapter->itr);
4113 if (hw->mac.type >= e1000_pch_spt)
4114 e1000_flush_desc_rings(adapter);
4115 /* Allow time for pending master requests to run */
4116 mac->ops.reset_hw(hw);
4118 /* For parts with AMT enabled, let the firmware know
4119 * that the network interface is in control
4121 if (adapter->flags & FLAG_HAS_AMT)
4122 e1000e_get_hw_control(adapter);
4126 if (mac->ops.init_hw(hw))
4127 e_err("Hardware Error\n");
4129 e1000_update_mng_vlan(adapter);
4131 /* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
4132 ew32(VET, ETH_P_8021Q);
4134 e1000e_reset_adaptive(hw);
4136 /* restore systim and hwtstamp settings */
4137 e1000e_systim_reset(adapter);
4139 /* Set EEE advertisement as appropriate */
4140 if (adapter->flags2 & FLAG2_HAS_EEE) {
4144 switch (hw->phy.type) {
4145 case e1000_phy_82579:
4146 adv_addr = I82579_EEE_ADVERTISEMENT;
4148 case e1000_phy_i217:
4149 adv_addr = I217_EEE_ADVERTISEMENT;
4152 dev_err(&adapter->pdev->dev,
4153 "Invalid PHY type setting EEE advertisement\n");
4157 ret_val = hw->phy.ops.acquire(hw);
4159 dev_err(&adapter->pdev->dev,
4160 "EEE advertisement - unable to acquire PHY\n");
4164 e1000_write_emi_reg_locked(hw, adv_addr,
4165 hw->dev_spec.ich8lan.eee_disable ?
4166 0 : adapter->eee_advert);
4168 hw->phy.ops.release(hw);
4171 if (!netif_running(adapter->netdev) &&
4172 !test_bit(__E1000_TESTING, &adapter->state))
4173 e1000_power_down_phy(adapter);
4175 e1000_get_phy_info(hw);
4177 if ((adapter->flags & FLAG_HAS_SMART_POWER_DOWN) &&
4178 !(adapter->flags & FLAG_SMART_POWER_DOWN)) {
4180 /* speed up time to link by disabling smart power down, ignore
4181 * the return value of this function because there is nothing
4182 * different we would do if it failed
4184 e1e_rphy(hw, IGP02E1000_PHY_POWER_MGMT, &phy_data);
4185 phy_data &= ~IGP02E1000_PM_SPD;
4186 e1e_wphy(hw, IGP02E1000_PHY_POWER_MGMT, phy_data);
4188 if (hw->mac.type >= e1000_pch_spt && adapter->int_mode == 0) {
4191 /* Fextnvm7 @ 0xe4[2] = 1 */
4192 reg = er32(FEXTNVM7);
4193 reg |= E1000_FEXTNVM7_SIDE_CLK_UNGATE;
4194 ew32(FEXTNVM7, reg);
4195 /* Fextnvm9 @ 0x5bb4[13:12] = 11 */
4196 reg = er32(FEXTNVM9);
4197 reg |= E1000_FEXTNVM9_IOSFSB_CLKGATE_DIS |
4198 E1000_FEXTNVM9_IOSFSB_CLKREQ_DIS;
4199 ew32(FEXTNVM9, reg);
4205 * e1000e_trigger_lsc - trigger an LSC interrupt
4208 * Fire a link status change interrupt to start the watchdog.
4210 static void e1000e_trigger_lsc(struct e1000_adapter *adapter)
4212 struct e1000_hw *hw = &adapter->hw;
4214 if (adapter->msix_entries)
4215 ew32(ICS, E1000_ICS_LSC | E1000_ICS_OTHER);
4217 ew32(ICS, E1000_ICS_LSC);
4220 void e1000e_up(struct e1000_adapter *adapter)
4222 /* hardware has been reset, we need to reload some things */
4223 e1000_configure(adapter);
4225 clear_bit(__E1000_DOWN, &adapter->state);
4227 if (adapter->msix_entries)
4228 e1000_configure_msix(adapter);
4229 e1000_irq_enable(adapter);
4231 /* Tx queue started by watchdog timer when link is up */
4233 e1000e_trigger_lsc(adapter);
4236 static void e1000e_flush_descriptors(struct e1000_adapter *adapter)
4238 struct e1000_hw *hw = &adapter->hw;
4240 if (!(adapter->flags2 & FLAG2_DMA_BURST))
4243 /* flush pending descriptor writebacks to memory */
4244 ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4245 ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4247 /* execute the writes immediately */
4250 /* due to rare timing issues, write to TIDV/RDTR again to ensure the
4251 * write is successful
4253 ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4254 ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4256 /* execute the writes immediately */
4260 static void e1000e_update_stats(struct e1000_adapter *adapter);
4263 * e1000e_down - quiesce the device and optionally reset the hardware
4264 * @adapter: board private structure
4265 * @reset: boolean flag to reset the hardware or not
4267 void e1000e_down(struct e1000_adapter *adapter, bool reset)
4269 struct net_device *netdev = adapter->netdev;
4270 struct e1000_hw *hw = &adapter->hw;
4273 /* signal that we're down so the interrupt handler does not
4274 * reschedule our watchdog timer
4276 set_bit(__E1000_DOWN, &adapter->state);
4278 netif_carrier_off(netdev);
4280 /* disable receives in the hardware */
4282 if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
4283 ew32(RCTL, rctl & ~E1000_RCTL_EN);
4284 /* flush and sleep below */
4286 netif_stop_queue(netdev);
4288 /* disable transmits in the hardware */
4290 tctl &= ~E1000_TCTL_EN;
4293 /* flush both disables and wait for them to finish */
4295 usleep_range(10000, 11000);
4297 e1000_irq_disable(adapter);
4299 napi_synchronize(&adapter->napi);
4301 del_timer_sync(&adapter->watchdog_timer);
4302 del_timer_sync(&adapter->phy_info_timer);
4304 spin_lock(&adapter->stats64_lock);
4305 e1000e_update_stats(adapter);
4306 spin_unlock(&adapter->stats64_lock);
4308 e1000e_flush_descriptors(adapter);
4310 adapter->link_speed = 0;
4311 adapter->link_duplex = 0;
4313 /* Disable Si errata workaround on PCHx for jumbo frame flow */
4314 if ((hw->mac.type >= e1000_pch2lan) &&
4315 (adapter->netdev->mtu > ETH_DATA_LEN) &&
4316 e1000_lv_jumbo_workaround_ich8lan(hw, false))
4317 e_dbg("failed to disable jumbo frame workaround mode\n");
4319 if (!pci_channel_offline(adapter->pdev)) {
4321 e1000e_reset(adapter);
4322 else if (hw->mac.type >= e1000_pch_spt)
4323 e1000_flush_desc_rings(adapter);
4325 e1000_clean_tx_ring(adapter->tx_ring);
4326 e1000_clean_rx_ring(adapter->rx_ring);
4329 void e1000e_reinit_locked(struct e1000_adapter *adapter)
4332 while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
4333 usleep_range(1000, 1100);
4334 e1000e_down(adapter, true);
4336 clear_bit(__E1000_RESETTING, &adapter->state);
4340 * e1000e_sanitize_systim - sanitize raw cycle counter reads
4341 * @hw: pointer to the HW structure
4342 * @systim: PHC time value read, sanitized and returned
4343 * @sts: structure to hold system time before and after reading SYSTIML,
4346 * Errata for 82574/82583 possible bad bits read from SYSTIMH/L:
4347 * check to see that the time is incrementing at a reasonable
4348 * rate and is a multiple of incvalue.
4350 static u64 e1000e_sanitize_systim(struct e1000_hw *hw, u64 systim,
4351 struct ptp_system_timestamp *sts)
4353 u64 time_delta, rem, temp;
4358 incvalue = er32(TIMINCA) & E1000_TIMINCA_INCVALUE_MASK;
4359 for (i = 0; i < E1000_MAX_82574_SYSTIM_REREADS; i++) {
4360 /* latch SYSTIMH on read of SYSTIML */
4361 ptp_read_system_prets(sts);
4362 systim_next = (u64)er32(SYSTIML);
4363 ptp_read_system_postts(sts);
4364 systim_next |= (u64)er32(SYSTIMH) << 32;
4366 time_delta = systim_next - systim;
4368 /* VMWare users have seen incvalue of zero, don't div / 0 */
4369 rem = incvalue ? do_div(temp, incvalue) : (time_delta != 0);
4371 systim = systim_next;
4373 if ((time_delta < E1000_82574_SYSTIM_EPSILON) && (rem == 0))
4381 * e1000e_read_systim - read SYSTIM register
4382 * @adapter: board private structure
4383 * @sts: structure which will contain system time before and after reading
4384 * SYSTIML, may be NULL
4386 u64 e1000e_read_systim(struct e1000_adapter *adapter,
4387 struct ptp_system_timestamp *sts)
4389 struct e1000_hw *hw = &adapter->hw;
4390 u32 systimel, systimel_2, systimeh;
4392 /* SYSTIMH latching upon SYSTIML read does not work well.
4393 * This means that if SYSTIML overflows after we read it but before
4394 * we read SYSTIMH, the value of SYSTIMH has been incremented and we
4395 * will experience a huge non linear increment in the systime value
4396 * to fix that we test for overflow and if true, we re-read systime.
4398 ptp_read_system_prets(sts);
4399 systimel = er32(SYSTIML);
4400 ptp_read_system_postts(sts);
4401 systimeh = er32(SYSTIMH);
4402 /* Is systimel is so large that overflow is possible? */
4403 if (systimel >= (u32)0xffffffff - E1000_TIMINCA_INCVALUE_MASK) {
4404 ptp_read_system_prets(sts);
4405 systimel_2 = er32(SYSTIML);
4406 ptp_read_system_postts(sts);
4407 if (systimel > systimel_2) {
4408 /* There was an overflow, read again SYSTIMH, and use
4411 systimeh = er32(SYSTIMH);
4412 systimel = systimel_2;
4415 systim = (u64)systimel;
4416 systim |= (u64)systimeh << 32;
4418 if (adapter->flags2 & FLAG2_CHECK_SYSTIM_OVERFLOW)
4419 systim = e1000e_sanitize_systim(hw, systim, sts);
4425 * e1000e_cyclecounter_read - read raw cycle counter (used by time counter)
4426 * @cc: cyclecounter structure
4428 static u64 e1000e_cyclecounter_read(const struct cyclecounter *cc)
4430 struct e1000_adapter *adapter = container_of(cc, struct e1000_adapter,
4433 return e1000e_read_systim(adapter, NULL);
4437 * e1000_sw_init - Initialize general software structures (struct e1000_adapter)
4438 * @adapter: board private structure to initialize
4440 * e1000_sw_init initializes the Adapter private data structure.
4441 * Fields are initialized based on PCI device information and
4442 * OS network device settings (MTU size).
4444 static int e1000_sw_init(struct e1000_adapter *adapter)
4446 struct net_device *netdev = adapter->netdev;
4448 adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
4449 adapter->rx_ps_bsize0 = 128;
4450 adapter->max_frame_size = netdev->mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
4451 adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4452 adapter->tx_ring_count = E1000_DEFAULT_TXD;
4453 adapter->rx_ring_count = E1000_DEFAULT_RXD;
4455 spin_lock_init(&adapter->stats64_lock);
4457 e1000e_set_interrupt_capability(adapter);
4459 if (e1000_alloc_queues(adapter))
4462 /* Setup hardware time stamping cyclecounter */
4463 if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
4464 adapter->cc.read = e1000e_cyclecounter_read;
4465 adapter->cc.mask = CYCLECOUNTER_MASK(64);
4466 adapter->cc.mult = 1;
4467 /* cc.shift set in e1000e_get_base_tininca() */
4469 spin_lock_init(&adapter->systim_lock);
4470 INIT_WORK(&adapter->tx_hwtstamp_work, e1000e_tx_hwtstamp_work);
4473 /* Explicitly disable IRQ since the NIC can be in any state. */
4474 e1000_irq_disable(adapter);
4476 set_bit(__E1000_DOWN, &adapter->state);
4481 * e1000_intr_msi_test - Interrupt Handler
4482 * @irq: interrupt number
4483 * @data: pointer to a network interface device structure
4485 static irqreturn_t e1000_intr_msi_test(int __always_unused irq, void *data)
4487 struct net_device *netdev = data;
4488 struct e1000_adapter *adapter = netdev_priv(netdev);
4489 struct e1000_hw *hw = &adapter->hw;
4490 u32 icr = er32(ICR);
4492 e_dbg("icr is %08X\n", icr);
4493 if (icr & E1000_ICR_RXSEQ) {
4494 adapter->flags &= ~FLAG_MSI_TEST_FAILED;
4495 /* Force memory writes to complete before acknowledging the
4496 * interrupt is handled.
4505 * e1000_test_msi_interrupt - Returns 0 for successful test
4506 * @adapter: board private struct
4508 * code flow taken from tg3.c
4510 static int e1000_test_msi_interrupt(struct e1000_adapter *adapter)
4512 struct net_device *netdev = adapter->netdev;
4513 struct e1000_hw *hw = &adapter->hw;
4516 /* poll_enable hasn't been called yet, so don't need disable */
4517 /* clear any pending events */
4520 /* free the real vector and request a test handler */
4521 e1000_free_irq(adapter);
4522 e1000e_reset_interrupt_capability(adapter);
4524 /* Assume that the test fails, if it succeeds then the test
4525 * MSI irq handler will unset this flag
4527 adapter->flags |= FLAG_MSI_TEST_FAILED;
4529 err = pci_enable_msi(adapter->pdev);
4531 goto msi_test_failed;
4533 err = request_irq(adapter->pdev->irq, e1000_intr_msi_test, 0,
4534 netdev->name, netdev);
4536 pci_disable_msi(adapter->pdev);
4537 goto msi_test_failed;
4540 /* Force memory writes to complete before enabling and firing an
4545 e1000_irq_enable(adapter);
4547 /* fire an unusual interrupt on the test handler */
4548 ew32(ICS, E1000_ICS_RXSEQ);
4552 e1000_irq_disable(adapter);
4554 rmb(); /* read flags after interrupt has been fired */
4556 if (adapter->flags & FLAG_MSI_TEST_FAILED) {
4557 adapter->int_mode = E1000E_INT_MODE_LEGACY;
4558 e_info("MSI interrupt test failed, using legacy interrupt.\n");
4560 e_dbg("MSI interrupt test succeeded!\n");
4563 free_irq(adapter->pdev->irq, netdev);
4564 pci_disable_msi(adapter->pdev);
4567 e1000e_set_interrupt_capability(adapter);
4568 return e1000_request_irq(adapter);
4572 * e1000_test_msi - Returns 0 if MSI test succeeds or INTx mode is restored
4573 * @adapter: board private struct
4575 * code flow taken from tg3.c, called with e1000 interrupts disabled.
4577 static int e1000_test_msi(struct e1000_adapter *adapter)
4582 if (!(adapter->flags & FLAG_MSI_ENABLED))
4585 /* disable SERR in case the MSI write causes a master abort */
4586 pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4587 if (pci_cmd & PCI_COMMAND_SERR)
4588 pci_write_config_word(adapter->pdev, PCI_COMMAND,
4589 pci_cmd & ~PCI_COMMAND_SERR);
4591 err = e1000_test_msi_interrupt(adapter);
4593 /* re-enable SERR */
4594 if (pci_cmd & PCI_COMMAND_SERR) {
4595 pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4596 pci_cmd |= PCI_COMMAND_SERR;
4597 pci_write_config_word(adapter->pdev, PCI_COMMAND, pci_cmd);
4604 * e1000e_open - Called when a network interface is made active
4605 * @netdev: network interface device structure
4607 * Returns 0 on success, negative value on failure
4609 * The open entry point is called when a network interface is made
4610 * active by the system (IFF_UP). At this point all resources needed
4611 * for transmit and receive operations are allocated, the interrupt
4612 * handler is registered with the OS, the watchdog timer is started,
4613 * and the stack is notified that the interface is ready.
4615 int e1000e_open(struct net_device *netdev)
4617 struct e1000_adapter *adapter = netdev_priv(netdev);
4618 struct e1000_hw *hw = &adapter->hw;
4619 struct pci_dev *pdev = adapter->pdev;
4622 /* disallow open during test */
4623 if (test_bit(__E1000_TESTING, &adapter->state))
4626 pm_runtime_get_sync(&pdev->dev);
4628 netif_carrier_off(netdev);
4629 netif_stop_queue(netdev);
4631 /* allocate transmit descriptors */
4632 err = e1000e_setup_tx_resources(adapter->tx_ring);
4636 /* allocate receive descriptors */
4637 err = e1000e_setup_rx_resources(adapter->rx_ring);
4641 /* If AMT is enabled, let the firmware know that the network
4642 * interface is now open and reset the part to a known state.
4644 if (adapter->flags & FLAG_HAS_AMT) {
4645 e1000e_get_hw_control(adapter);
4646 e1000e_reset(adapter);
4649 e1000e_power_up_phy(adapter);
4651 adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
4652 if ((adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
4653 e1000_update_mng_vlan(adapter);
4655 /* DMA latency requirement to workaround jumbo issue */
4656 cpu_latency_qos_add_request(&adapter->pm_qos_req, PM_QOS_DEFAULT_VALUE);
4658 /* before we allocate an interrupt, we must be ready to handle it.
4659 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4660 * as soon as we call pci_request_irq, so we have to setup our
4661 * clean_rx handler before we do so.
4663 e1000_configure(adapter);
4665 err = e1000_request_irq(adapter);
4669 /* Work around PCIe errata with MSI interrupts causing some chipsets to
4670 * ignore e1000e MSI messages, which means we need to test our MSI
4673 if (adapter->int_mode != E1000E_INT_MODE_LEGACY) {
4674 err = e1000_test_msi(adapter);
4676 e_err("Interrupt allocation failed\n");
4681 /* From here on the code is the same as e1000e_up() */
4682 clear_bit(__E1000_DOWN, &adapter->state);
4684 napi_enable(&adapter->napi);
4686 e1000_irq_enable(adapter);
4688 adapter->tx_hang_recheck = false;
4690 hw->mac.get_link_status = true;
4691 pm_runtime_put(&pdev->dev);
4693 e1000e_trigger_lsc(adapter);
4698 cpu_latency_qos_remove_request(&adapter->pm_qos_req);
4699 e1000e_release_hw_control(adapter);
4700 e1000_power_down_phy(adapter);
4701 e1000e_free_rx_resources(adapter->rx_ring);
4703 e1000e_free_tx_resources(adapter->tx_ring);
4705 e1000e_reset(adapter);
4706 pm_runtime_put_sync(&pdev->dev);
4712 * e1000e_close - Disables a network interface
4713 * @netdev: network interface device structure
4715 * Returns 0, this is not allowed to fail
4717 * The close entry point is called when an interface is de-activated
4718 * by the OS. The hardware is still under the drivers control, but
4719 * needs to be disabled. A global MAC reset is issued to stop the
4720 * hardware, and all transmit and receive resources are freed.
4722 int e1000e_close(struct net_device *netdev)
4724 struct e1000_adapter *adapter = netdev_priv(netdev);
4725 struct pci_dev *pdev = adapter->pdev;
4726 int count = E1000_CHECK_RESET_COUNT;
4728 while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
4729 usleep_range(10000, 11000);
4731 WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
4733 pm_runtime_get_sync(&pdev->dev);
4735 if (netif_device_present(netdev)) {
4736 e1000e_down(adapter, true);
4737 e1000_free_irq(adapter);
4739 /* Link status message must follow this format */
4740 netdev_info(netdev, "NIC Link is Down\n");
4743 napi_disable(&adapter->napi);
4745 e1000e_free_tx_resources(adapter->tx_ring);
4746 e1000e_free_rx_resources(adapter->rx_ring);
4748 /* kill manageability vlan ID if supported, but not if a vlan with
4749 * the same ID is registered on the host OS (let 8021q kill it)
4751 if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN)
4752 e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
4753 adapter->mng_vlan_id);
4755 /* If AMT is enabled, let the firmware know that the network
4756 * interface is now closed
4758 if ((adapter->flags & FLAG_HAS_AMT) &&
4759 !test_bit(__E1000_TESTING, &adapter->state))
4760 e1000e_release_hw_control(adapter);
4762 cpu_latency_qos_remove_request(&adapter->pm_qos_req);
4764 pm_runtime_put_sync(&pdev->dev);
4770 * e1000_set_mac - Change the Ethernet Address of the NIC
4771 * @netdev: network interface device structure
4772 * @p: pointer to an address structure
4774 * Returns 0 on success, negative on failure
4776 static int e1000_set_mac(struct net_device *netdev, void *p)
4778 struct e1000_adapter *adapter = netdev_priv(netdev);
4779 struct e1000_hw *hw = &adapter->hw;
4780 struct sockaddr *addr = p;
4782 if (!is_valid_ether_addr(addr->sa_data))
4783 return -EADDRNOTAVAIL;
4785 eth_hw_addr_set(netdev, addr->sa_data);
4786 memcpy(adapter->hw.mac.addr, addr->sa_data, netdev->addr_len);
4788 hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
4790 if (adapter->flags & FLAG_RESET_OVERWRITES_LAA) {
4791 /* activate the work around */
4792 e1000e_set_laa_state_82571(&adapter->hw, 1);
4794 /* Hold a copy of the LAA in RAR[14] This is done so that
4795 * between the time RAR[0] gets clobbered and the time it
4796 * gets fixed (in e1000_watchdog), the actual LAA is in one
4797 * of the RARs and no incoming packets directed to this port
4798 * are dropped. Eventually the LAA will be in RAR[0] and
4801 hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr,
4802 adapter->hw.mac.rar_entry_count - 1);
4809 * e1000e_update_phy_task - work thread to update phy
4810 * @work: pointer to our work struct
4812 * this worker thread exists because we must acquire a
4813 * semaphore to read the phy, which we could msleep while
4814 * waiting for it, and we can't msleep in a timer.
4816 static void e1000e_update_phy_task(struct work_struct *work)
4818 struct e1000_adapter *adapter = container_of(work,
4819 struct e1000_adapter,
4821 struct e1000_hw *hw = &adapter->hw;
4823 if (test_bit(__E1000_DOWN, &adapter->state))
4826 e1000_get_phy_info(hw);
4828 /* Enable EEE on 82579 after link up */
4829 if (hw->phy.type >= e1000_phy_82579)
4830 e1000_set_eee_pchlan(hw);
4834 * e1000_update_phy_info - timre call-back to update PHY info
4835 * @t: pointer to timer_list containing private info adapter
4837 * Need to wait a few seconds after link up to get diagnostic information from
4840 static void e1000_update_phy_info(struct timer_list *t)
4842 struct e1000_adapter *adapter = from_timer(adapter, t, phy_info_timer);
4844 if (test_bit(__E1000_DOWN, &adapter->state))
4847 schedule_work(&adapter->update_phy_task);
4851 * e1000e_update_phy_stats - Update the PHY statistics counters
4852 * @adapter: board private structure
4854 * Read/clear the upper 16-bit PHY registers and read/accumulate lower
4856 static void e1000e_update_phy_stats(struct e1000_adapter *adapter)
4858 struct e1000_hw *hw = &adapter->hw;
4862 ret_val = hw->phy.ops.acquire(hw);
4866 /* A page set is expensive so check if already on desired page.
4867 * If not, set to the page with the PHY status registers.
4870 ret_val = e1000e_read_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4874 if (phy_data != (HV_STATS_PAGE << IGP_PAGE_SHIFT)) {
4875 ret_val = hw->phy.ops.set_page(hw,
4876 HV_STATS_PAGE << IGP_PAGE_SHIFT);
4881 /* Single Collision Count */
4882 hw->phy.ops.read_reg_page(hw, HV_SCC_UPPER, &phy_data);
4883 ret_val = hw->phy.ops.read_reg_page(hw, HV_SCC_LOWER, &phy_data);
4885 adapter->stats.scc += phy_data;
4887 /* Excessive Collision Count */
4888 hw->phy.ops.read_reg_page(hw, HV_ECOL_UPPER, &phy_data);
4889 ret_val = hw->phy.ops.read_reg_page(hw, HV_ECOL_LOWER, &phy_data);
4891 adapter->stats.ecol += phy_data;
4893 /* Multiple Collision Count */
4894 hw->phy.ops.read_reg_page(hw, HV_MCC_UPPER, &phy_data);
4895 ret_val = hw->phy.ops.read_reg_page(hw, HV_MCC_LOWER, &phy_data);
4897 adapter->stats.mcc += phy_data;
4899 /* Late Collision Count */
4900 hw->phy.ops.read_reg_page(hw, HV_LATECOL_UPPER, &phy_data);
4901 ret_val = hw->phy.ops.read_reg_page(hw, HV_LATECOL_LOWER, &phy_data);
4903 adapter->stats.latecol += phy_data;
4905 /* Collision Count - also used for adaptive IFS */
4906 hw->phy.ops.read_reg_page(hw, HV_COLC_UPPER, &phy_data);
4907 ret_val = hw->phy.ops.read_reg_page(hw, HV_COLC_LOWER, &phy_data);
4909 hw->mac.collision_delta = phy_data;
4912 hw->phy.ops.read_reg_page(hw, HV_DC_UPPER, &phy_data);
4913 ret_val = hw->phy.ops.read_reg_page(hw, HV_DC_LOWER, &phy_data);
4915 adapter->stats.dc += phy_data;
4917 /* Transmit with no CRS */
4918 hw->phy.ops.read_reg_page(hw, HV_TNCRS_UPPER, &phy_data);
4919 ret_val = hw->phy.ops.read_reg_page(hw, HV_TNCRS_LOWER, &phy_data);
4921 adapter->stats.tncrs += phy_data;
4924 hw->phy.ops.release(hw);
4928 * e1000e_update_stats - Update the board statistics counters
4929 * @adapter: board private structure
4931 static void e1000e_update_stats(struct e1000_adapter *adapter)
4933 struct net_device *netdev = adapter->netdev;
4934 struct e1000_hw *hw = &adapter->hw;
4935 struct pci_dev *pdev = adapter->pdev;
4937 /* Prevent stats update while adapter is being reset, or if the pci
4938 * connection is down.
4940 if (adapter->link_speed == 0)
4942 if (pci_channel_offline(pdev))
4945 adapter->stats.crcerrs += er32(CRCERRS);
4946 adapter->stats.gprc += er32(GPRC);
4947 adapter->stats.gorc += er32(GORCL);
4948 er32(GORCH); /* Clear gorc */
4949 adapter->stats.bprc += er32(BPRC);
4950 adapter->stats.mprc += er32(MPRC);
4951 adapter->stats.roc += er32(ROC);
4953 adapter->stats.mpc += er32(MPC);
4955 /* Half-duplex statistics */
4956 if (adapter->link_duplex == HALF_DUPLEX) {
4957 if (adapter->flags2 & FLAG2_HAS_PHY_STATS) {
4958 e1000e_update_phy_stats(adapter);
4960 adapter->stats.scc += er32(SCC);
4961 adapter->stats.ecol += er32(ECOL);
4962 adapter->stats.mcc += er32(MCC);
4963 adapter->stats.latecol += er32(LATECOL);
4964 adapter->stats.dc += er32(DC);
4966 hw->mac.collision_delta = er32(COLC);
4968 if ((hw->mac.type != e1000_82574) &&
4969 (hw->mac.type != e1000_82583))
4970 adapter->stats.tncrs += er32(TNCRS);
4972 adapter->stats.colc += hw->mac.collision_delta;
4975 adapter->stats.xonrxc += er32(XONRXC);
4976 adapter->stats.xontxc += er32(XONTXC);
4977 adapter->stats.xoffrxc += er32(XOFFRXC);
4978 adapter->stats.xofftxc += er32(XOFFTXC);
4979 adapter->stats.gptc += er32(GPTC);
4980 adapter->stats.gotc += er32(GOTCL);
4981 er32(GOTCH); /* Clear gotc */
4982 adapter->stats.rnbc += er32(RNBC);
4983 adapter->stats.ruc += er32(RUC);
4985 adapter->stats.mptc += er32(MPTC);
4986 adapter->stats.bptc += er32(BPTC);
4988 /* used for adaptive IFS */
4990 hw->mac.tx_packet_delta = er32(TPT);
4991 adapter->stats.tpt += hw->mac.tx_packet_delta;
4993 adapter->stats.algnerrc += er32(ALGNERRC);
4994 adapter->stats.rxerrc += er32(RXERRC);
4995 adapter->stats.cexterr += er32(CEXTERR);
4996 adapter->stats.tsctc += er32(TSCTC);
4997 adapter->stats.tsctfc += er32(TSCTFC);
4999 /* Fill out the OS statistics structure */
5000 netdev->stats.multicast = adapter->stats.mprc;
5001 netdev->stats.collisions = adapter->stats.colc;
5005 /* RLEC on some newer hardware can be incorrect so build
5006 * our own version based on RUC and ROC
5008 netdev->stats.rx_errors = adapter->stats.rxerrc +
5009 adapter->stats.crcerrs + adapter->stats.algnerrc +
5010 adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5011 netdev->stats.rx_length_errors = adapter->stats.ruc +
5013 netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
5014 netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
5015 netdev->stats.rx_missed_errors = adapter->stats.mpc;
5018 netdev->stats.tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5019 netdev->stats.tx_aborted_errors = adapter->stats.ecol;
5020 netdev->stats.tx_window_errors = adapter->stats.latecol;
5021 netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
5023 /* Tx Dropped needs to be maintained elsewhere */
5025 /* Management Stats */
5026 adapter->stats.mgptc += er32(MGTPTC);
5027 adapter->stats.mgprc += er32(MGTPRC);
5028 adapter->stats.mgpdc += er32(MGTPDC);
5030 /* Correctable ECC Errors */
5031 if (hw->mac.type >= e1000_pch_lpt) {
5032 u32 pbeccsts = er32(PBECCSTS);
5034 adapter->corr_errors +=
5035 pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
5036 adapter->uncorr_errors +=
5037 (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
5038 E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
5043 * e1000_phy_read_status - Update the PHY register status snapshot
5044 * @adapter: board private structure
5046 static void e1000_phy_read_status(struct e1000_adapter *adapter)
5048 struct e1000_hw *hw = &adapter->hw;
5049 struct e1000_phy_regs *phy = &adapter->phy_regs;
5051 if (!pm_runtime_suspended((&adapter->pdev->dev)->parent) &&
5052 (er32(STATUS) & E1000_STATUS_LU) &&
5053 (adapter->hw.phy.media_type == e1000_media_type_copper)) {
5056 ret_val = e1e_rphy(hw, MII_BMCR, &phy->bmcr);
5057 ret_val |= e1e_rphy(hw, MII_BMSR, &phy->bmsr);
5058 ret_val |= e1e_rphy(hw, MII_ADVERTISE, &phy->advertise);
5059 ret_val |= e1e_rphy(hw, MII_LPA, &phy->lpa);
5060 ret_val |= e1e_rphy(hw, MII_EXPANSION, &phy->expansion);
5061 ret_val |= e1e_rphy(hw, MII_CTRL1000, &phy->ctrl1000);
5062 ret_val |= e1e_rphy(hw, MII_STAT1000, &phy->stat1000);
5063 ret_val |= e1e_rphy(hw, MII_ESTATUS, &phy->estatus);
5065 e_warn("Error reading PHY register\n");
5067 /* Do not read PHY registers if link is not up
5068 * Set values to typical power-on defaults
5070 phy->bmcr = (BMCR_SPEED1000 | BMCR_ANENABLE | BMCR_FULLDPLX);
5071 phy->bmsr = (BMSR_100FULL | BMSR_100HALF | BMSR_10FULL |
5072 BMSR_10HALF | BMSR_ESTATEN | BMSR_ANEGCAPABLE |
5074 phy->advertise = (ADVERTISE_PAUSE_ASYM | ADVERTISE_PAUSE_CAP |
5075 ADVERTISE_ALL | ADVERTISE_CSMA);
5077 phy->expansion = EXPANSION_ENABLENPAGE;
5078 phy->ctrl1000 = ADVERTISE_1000FULL;
5080 phy->estatus = (ESTATUS_1000_TFULL | ESTATUS_1000_THALF);
5084 static void e1000_print_link_info(struct e1000_adapter *adapter)
5086 struct e1000_hw *hw = &adapter->hw;
5087 u32 ctrl = er32(CTRL);
5089 /* Link status message must follow this format for user tools */
5090 netdev_info(adapter->netdev,
5091 "NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5092 adapter->link_speed,
5093 adapter->link_duplex == FULL_DUPLEX ? "Full" : "Half",
5094 (ctrl & E1000_CTRL_TFCE) && (ctrl & E1000_CTRL_RFCE) ? "Rx/Tx" :
5095 (ctrl & E1000_CTRL_RFCE) ? "Rx" :
5096 (ctrl & E1000_CTRL_TFCE) ? "Tx" : "None");
5099 static bool e1000e_has_link(struct e1000_adapter *adapter)
5101 struct e1000_hw *hw = &adapter->hw;
5102 bool link_active = false;
5105 /* get_link_status is set on LSC (link status) interrupt or
5106 * Rx sequence error interrupt. get_link_status will stay
5107 * true until the check_for_link establishes link
5108 * for copper adapters ONLY
5110 switch (hw->phy.media_type) {
5111 case e1000_media_type_copper:
5112 if (hw->mac.get_link_status) {
5113 ret_val = hw->mac.ops.check_for_link(hw);
5114 link_active = !hw->mac.get_link_status;
5119 case e1000_media_type_fiber:
5120 ret_val = hw->mac.ops.check_for_link(hw);
5121 link_active = !!(er32(STATUS) & E1000_STATUS_LU);
5123 case e1000_media_type_internal_serdes:
5124 ret_val = hw->mac.ops.check_for_link(hw);
5125 link_active = hw->mac.serdes_has_link;
5128 case e1000_media_type_unknown:
5132 if ((ret_val == -E1000_ERR_PHY) && (hw->phy.type == e1000_phy_igp_3) &&
5133 (er32(CTRL) & E1000_PHY_CTRL_GBE_DISABLE)) {
5134 /* See e1000_kmrn_lock_loss_workaround_ich8lan() */
5135 e_info("Gigabit has been disabled, downgrading speed\n");
5141 static void e1000e_enable_receives(struct e1000_adapter *adapter)
5143 /* make sure the receive unit is started */
5144 if ((adapter->flags & FLAG_RX_NEEDS_RESTART) &&
5145 (adapter->flags & FLAG_RESTART_NOW)) {
5146 struct e1000_hw *hw = &adapter->hw;
5147 u32 rctl = er32(RCTL);
5149 ew32(RCTL, rctl | E1000_RCTL_EN);
5150 adapter->flags &= ~FLAG_RESTART_NOW;
5154 static void e1000e_check_82574_phy_workaround(struct e1000_adapter *adapter)
5156 struct e1000_hw *hw = &adapter->hw;
5158 /* With 82574 controllers, PHY needs to be checked periodically
5159 * for hung state and reset, if two calls return true
5161 if (e1000_check_phy_82574(hw))
5162 adapter->phy_hang_count++;
5164 adapter->phy_hang_count = 0;
5166 if (adapter->phy_hang_count > 1) {
5167 adapter->phy_hang_count = 0;
5168 e_dbg("PHY appears hung - resetting\n");
5169 schedule_work(&adapter->reset_task);
5174 * e1000_watchdog - Timer Call-back
5175 * @t: pointer to timer_list containing private info adapter
5177 static void e1000_watchdog(struct timer_list *t)
5179 struct e1000_adapter *adapter = from_timer(adapter, t, watchdog_timer);
5181 /* Do the rest outside of interrupt context */
5182 schedule_work(&adapter->watchdog_task);
5184 /* TODO: make this use queue_delayed_work() */
5187 static void e1000_watchdog_task(struct work_struct *work)
5189 struct e1000_adapter *adapter = container_of(work,
5190 struct e1000_adapter,
5192 struct net_device *netdev = adapter->netdev;
5193 struct e1000_mac_info *mac = &adapter->hw.mac;
5194 struct e1000_phy_info *phy = &adapter->hw.phy;
5195 struct e1000_ring *tx_ring = adapter->tx_ring;
5196 u32 dmoff_exit_timeout = 100, tries = 0;
5197 struct e1000_hw *hw = &adapter->hw;
5198 u32 link, tctl, pcim_state;
5200 if (test_bit(__E1000_DOWN, &adapter->state))
5203 link = e1000e_has_link(adapter);
5204 if ((netif_carrier_ok(netdev)) && link) {
5205 /* Cancel scheduled suspend requests. */
5206 pm_runtime_resume(netdev->dev.parent);
5208 e1000e_enable_receives(adapter);
5212 if ((e1000e_enable_tx_pkt_filtering(hw)) &&
5213 (adapter->mng_vlan_id != adapter->hw.mng_cookie.vlan_id))
5214 e1000_update_mng_vlan(adapter);
5217 if (!netif_carrier_ok(netdev)) {
5220 /* Cancel scheduled suspend requests. */
5221 pm_runtime_resume(netdev->dev.parent);
5223 /* Checking if MAC is in DMoff state*/
5224 if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID) {
5225 pcim_state = er32(STATUS);
5226 while (pcim_state & E1000_STATUS_PCIM_STATE) {
5227 if (tries++ == dmoff_exit_timeout) {
5228 e_dbg("Error in exiting dmoff\n");
5231 usleep_range(10000, 20000);
5232 pcim_state = er32(STATUS);
5234 /* Checking if MAC exited DMoff state */
5235 if (!(pcim_state & E1000_STATUS_PCIM_STATE))
5236 e1000_phy_hw_reset(&adapter->hw);
5240 /* update snapshot of PHY registers on LSC */
5241 e1000_phy_read_status(adapter);
5242 mac->ops.get_link_up_info(&adapter->hw,
5243 &adapter->link_speed,
5244 &adapter->link_duplex);
5245 e1000_print_link_info(adapter);
5247 /* check if SmartSpeed worked */
5248 e1000e_check_downshift(hw);
5249 if (phy->speed_downgraded)
5251 "Link Speed was downgraded by SmartSpeed\n");
5253 /* On supported PHYs, check for duplex mismatch only
5254 * if link has autonegotiated at 10/100 half
5256 if ((hw->phy.type == e1000_phy_igp_3 ||
5257 hw->phy.type == e1000_phy_bm) &&
5259 (adapter->link_speed == SPEED_10 ||
5260 adapter->link_speed == SPEED_100) &&
5261 (adapter->link_duplex == HALF_DUPLEX)) {
5264 e1e_rphy(hw, MII_EXPANSION, &autoneg_exp);
5266 if (!(autoneg_exp & EXPANSION_NWAY))
5267 e_info("Autonegotiated half duplex but link partner cannot autoneg. Try forcing full duplex if link gets many collisions.\n");
5270 /* adjust timeout factor according to speed/duplex */
5271 adapter->tx_timeout_factor = 1;
5272 switch (adapter->link_speed) {
5275 adapter->tx_timeout_factor = 16;
5279 adapter->tx_timeout_factor = 10;
5283 /* workaround: re-program speed mode bit after
5286 if ((adapter->flags & FLAG_TARC_SPEED_MODE_BIT) &&
5290 tarc0 = er32(TARC(0));
5291 tarc0 &= ~SPEED_MODE_BIT;
5292 ew32(TARC(0), tarc0);
5295 /* disable TSO for pcie and 10/100 speeds, to avoid
5296 * some hardware issues
5298 if (!(adapter->flags & FLAG_TSO_FORCE)) {
5299 switch (adapter->link_speed) {
5302 e_info("10/100 speed: disabling TSO\n");
5303 netdev->features &= ~NETIF_F_TSO;
5304 netdev->features &= ~NETIF_F_TSO6;
5307 netdev->features |= NETIF_F_TSO;
5308 netdev->features |= NETIF_F_TSO6;
5314 if (hw->mac.type == e1000_pch_spt) {
5315 netdev->features &= ~NETIF_F_TSO;
5316 netdev->features &= ~NETIF_F_TSO6;
5320 /* enable transmits in the hardware, need to do this
5321 * after setting TARC(0)
5324 tctl |= E1000_TCTL_EN;
5327 /* Perform any post-link-up configuration before
5328 * reporting link up.
5330 if (phy->ops.cfg_on_link_up)
5331 phy->ops.cfg_on_link_up(hw);
5333 netif_wake_queue(netdev);
5334 netif_carrier_on(netdev);
5336 if (!test_bit(__E1000_DOWN, &adapter->state))
5337 mod_timer(&adapter->phy_info_timer,
5338 round_jiffies(jiffies + 2 * HZ));
5341 if (netif_carrier_ok(netdev)) {
5342 adapter->link_speed = 0;
5343 adapter->link_duplex = 0;
5344 /* Link status message must follow this format */
5345 netdev_info(netdev, "NIC Link is Down\n");
5346 netif_carrier_off(netdev);
5347 netif_stop_queue(netdev);
5348 if (!test_bit(__E1000_DOWN, &adapter->state))
5349 mod_timer(&adapter->phy_info_timer,
5350 round_jiffies(jiffies + 2 * HZ));
5352 /* 8000ES2LAN requires a Rx packet buffer work-around
5353 * on link down event; reset the controller to flush
5354 * the Rx packet buffer.
5356 if (adapter->flags & FLAG_RX_NEEDS_RESTART)
5357 adapter->flags |= FLAG_RESTART_NOW;
5359 pm_schedule_suspend(netdev->dev.parent,
5365 spin_lock(&adapter->stats64_lock);
5366 e1000e_update_stats(adapter);
5368 mac->tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
5369 adapter->tpt_old = adapter->stats.tpt;
5370 mac->collision_delta = adapter->stats.colc - adapter->colc_old;
5371 adapter->colc_old = adapter->stats.colc;
5373 adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
5374 adapter->gorc_old = adapter->stats.gorc;
5375 adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
5376 adapter->gotc_old = adapter->stats.gotc;
5377 spin_unlock(&adapter->stats64_lock);
5379 /* If the link is lost the controller stops DMA, but
5380 * if there is queued Tx work it cannot be done. So
5381 * reset the controller to flush the Tx packet buffers.
5383 if (!netif_carrier_ok(netdev) &&
5384 (e1000_desc_unused(tx_ring) + 1 < tx_ring->count))
5385 adapter->flags |= FLAG_RESTART_NOW;
5387 /* If reset is necessary, do it outside of interrupt context. */
5388 if (adapter->flags & FLAG_RESTART_NOW) {
5389 schedule_work(&adapter->reset_task);
5390 /* return immediately since reset is imminent */
5394 e1000e_update_adaptive(&adapter->hw);
5396 /* Simple mode for Interrupt Throttle Rate (ITR) */
5397 if (adapter->itr_setting == 4) {
5398 /* Symmetric Tx/Rx gets a reduced ITR=2000;
5399 * Total asymmetrical Tx or Rx gets ITR=8000;
5400 * everyone else is between 2000-8000.
5402 u32 goc = (adapter->gotc + adapter->gorc) / 10000;
5403 u32 dif = (adapter->gotc > adapter->gorc ?
5404 adapter->gotc - adapter->gorc :
5405 adapter->gorc - adapter->gotc) / 10000;
5406 u32 itr = goc > 0 ? (dif * 6000 / goc + 2000) : 8000;
5408 e1000e_write_itr(adapter, itr);
5411 /* Cause software interrupt to ensure Rx ring is cleaned */
5412 if (adapter->msix_entries)
5413 ew32(ICS, adapter->rx_ring->ims_val);
5415 ew32(ICS, E1000_ICS_RXDMT0);
5417 /* flush pending descriptors to memory before detecting Tx hang */
5418 e1000e_flush_descriptors(adapter);
5420 /* Force detection of hung controller every watchdog period */
5421 adapter->detect_tx_hung = true;
5423 /* With 82571 controllers, LAA may be overwritten due to controller
5424 * reset from the other port. Set the appropriate LAA in RAR[0]
5426 if (e1000e_get_laa_state_82571(hw))
5427 hw->mac.ops.rar_set(hw, adapter->hw.mac.addr, 0);
5429 if (adapter->flags2 & FLAG2_CHECK_PHY_HANG)
5430 e1000e_check_82574_phy_workaround(adapter);
5432 /* Clear valid timestamp stuck in RXSTMPL/H due to a Rx error */
5433 if (adapter->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE) {
5434 if ((adapter->flags2 & FLAG2_CHECK_RX_HWTSTAMP) &&
5435 (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID)) {
5437 adapter->rx_hwtstamp_cleared++;
5439 adapter->flags2 |= FLAG2_CHECK_RX_HWTSTAMP;
5443 /* Reset the timer */
5444 if (!test_bit(__E1000_DOWN, &adapter->state))
5445 mod_timer(&adapter->watchdog_timer,
5446 round_jiffies(jiffies + 2 * HZ));
5449 #define E1000_TX_FLAGS_CSUM 0x00000001
5450 #define E1000_TX_FLAGS_VLAN 0x00000002
5451 #define E1000_TX_FLAGS_TSO 0x00000004
5452 #define E1000_TX_FLAGS_IPV4 0x00000008
5453 #define E1000_TX_FLAGS_NO_FCS 0x00000010
5454 #define E1000_TX_FLAGS_HWTSTAMP 0x00000020
5455 #define E1000_TX_FLAGS_VLAN_MASK 0xffff0000
5456 #define E1000_TX_FLAGS_VLAN_SHIFT 16
5458 static int e1000_tso(struct e1000_ring *tx_ring, struct sk_buff *skb,
5461 struct e1000_context_desc *context_desc;
5462 struct e1000_buffer *buffer_info;
5466 u8 ipcss, ipcso, tucss, tucso, hdr_len;
5469 if (!skb_is_gso(skb))
5472 err = skb_cow_head(skb, 0);
5476 hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5477 mss = skb_shinfo(skb)->gso_size;
5478 if (protocol == htons(ETH_P_IP)) {
5479 struct iphdr *iph = ip_hdr(skb);
5482 tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
5484 cmd_length = E1000_TXD_CMD_IP;
5485 ipcse = skb_transport_offset(skb) - 1;
5486 } else if (skb_is_gso_v6(skb)) {
5487 tcp_v6_gso_csum_prep(skb);
5490 ipcss = skb_network_offset(skb);
5491 ipcso = (void *)&(ip_hdr(skb)->check) - (void *)skb->data;
5492 tucss = skb_transport_offset(skb);
5493 tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
5495 cmd_length |= (E1000_TXD_CMD_DEXT | E1000_TXD_CMD_TSE |
5496 E1000_TXD_CMD_TCP | (skb->len - (hdr_len)));
5498 i = tx_ring->next_to_use;
5499 context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5500 buffer_info = &tx_ring->buffer_info[i];
5502 context_desc->lower_setup.ip_fields.ipcss = ipcss;
5503 context_desc->lower_setup.ip_fields.ipcso = ipcso;
5504 context_desc->lower_setup.ip_fields.ipcse = cpu_to_le16(ipcse);
5505 context_desc->upper_setup.tcp_fields.tucss = tucss;
5506 context_desc->upper_setup.tcp_fields.tucso = tucso;
5507 context_desc->upper_setup.tcp_fields.tucse = 0;
5508 context_desc->tcp_seg_setup.fields.mss = cpu_to_le16(mss);
5509 context_desc->tcp_seg_setup.fields.hdr_len = hdr_len;
5510 context_desc->cmd_and_length = cpu_to_le32(cmd_length);
5512 buffer_info->time_stamp = jiffies;
5513 buffer_info->next_to_watch = i;
5516 if (i == tx_ring->count)
5518 tx_ring->next_to_use = i;
5523 static bool e1000_tx_csum(struct e1000_ring *tx_ring, struct sk_buff *skb,
5526 struct e1000_adapter *adapter = tx_ring->adapter;
5527 struct e1000_context_desc *context_desc;
5528 struct e1000_buffer *buffer_info;
5531 u32 cmd_len = E1000_TXD_CMD_DEXT;
5533 if (skb->ip_summed != CHECKSUM_PARTIAL)
5537 case cpu_to_be16(ETH_P_IP):
5538 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
5539 cmd_len |= E1000_TXD_CMD_TCP;
5541 case cpu_to_be16(ETH_P_IPV6):
5542 /* XXX not handling all IPV6 headers */
5543 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
5544 cmd_len |= E1000_TXD_CMD_TCP;
5547 if (unlikely(net_ratelimit()))
5548 e_warn("checksum_partial proto=%x!\n",
5549 be16_to_cpu(protocol));
5553 css = skb_checksum_start_offset(skb);
5555 i = tx_ring->next_to_use;
5556 buffer_info = &tx_ring->buffer_info[i];
5557 context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5559 context_desc->lower_setup.ip_config = 0;
5560 context_desc->upper_setup.tcp_fields.tucss = css;
5561 context_desc->upper_setup.tcp_fields.tucso = css + skb->csum_offset;
5562 context_desc->upper_setup.tcp_fields.tucse = 0;
5563 context_desc->tcp_seg_setup.data = 0;
5564 context_desc->cmd_and_length = cpu_to_le32(cmd_len);
5566 buffer_info->time_stamp = jiffies;
5567 buffer_info->next_to_watch = i;
5570 if (i == tx_ring->count)
5572 tx_ring->next_to_use = i;
5577 static int e1000_tx_map(struct e1000_ring *tx_ring, struct sk_buff *skb,
5578 unsigned int first, unsigned int max_per_txd,
5579 unsigned int nr_frags)
5581 struct e1000_adapter *adapter = tx_ring->adapter;
5582 struct pci_dev *pdev = adapter->pdev;
5583 struct e1000_buffer *buffer_info;
5584 unsigned int len = skb_headlen(skb);
5585 unsigned int offset = 0, size, count = 0, i;
5586 unsigned int f, bytecount, segs;
5588 i = tx_ring->next_to_use;
5591 buffer_info = &tx_ring->buffer_info[i];
5592 size = min(len, max_per_txd);
5594 buffer_info->length = size;
5595 buffer_info->time_stamp = jiffies;
5596 buffer_info->next_to_watch = i;
5597 buffer_info->dma = dma_map_single(&pdev->dev,
5599 size, DMA_TO_DEVICE);
5600 buffer_info->mapped_as_page = false;
5601 if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5610 if (i == tx_ring->count)
5615 for (f = 0; f < nr_frags; f++) {
5616 const skb_frag_t *frag = &skb_shinfo(skb)->frags[f];
5618 len = skb_frag_size(frag);
5623 if (i == tx_ring->count)
5626 buffer_info = &tx_ring->buffer_info[i];
5627 size = min(len, max_per_txd);
5629 buffer_info->length = size;
5630 buffer_info->time_stamp = jiffies;
5631 buffer_info->next_to_watch = i;
5632 buffer_info->dma = skb_frag_dma_map(&pdev->dev, frag,
5635 buffer_info->mapped_as_page = true;
5636 if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5645 segs = skb_shinfo(skb)->gso_segs ? : 1;
5646 /* multiply data chunks by size of headers */
5647 bytecount = ((segs - 1) * skb_headlen(skb)) + skb->len;
5649 tx_ring->buffer_info[i].skb = skb;
5650 tx_ring->buffer_info[i].segs = segs;
5651 tx_ring->buffer_info[i].bytecount = bytecount;
5652 tx_ring->buffer_info[first].next_to_watch = i;
5657 dev_err(&pdev->dev, "Tx DMA map failed\n");
5658 buffer_info->dma = 0;
5664 i += tx_ring->count;
5666 buffer_info = &tx_ring->buffer_info[i];
5667 e1000_put_txbuf(tx_ring, buffer_info, true);
5673 static void e1000_tx_queue(struct e1000_ring *tx_ring, int tx_flags, int count)
5675 struct e1000_adapter *adapter = tx_ring->adapter;
5676 struct e1000_tx_desc *tx_desc = NULL;
5677 struct e1000_buffer *buffer_info;
5678 u32 txd_upper = 0, txd_lower = E1000_TXD_CMD_IFCS;
5681 if (tx_flags & E1000_TX_FLAGS_TSO) {
5682 txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D |
5684 txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5686 if (tx_flags & E1000_TX_FLAGS_IPV4)
5687 txd_upper |= E1000_TXD_POPTS_IXSM << 8;
5690 if (tx_flags & E1000_TX_FLAGS_CSUM) {
5691 txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5692 txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5695 if (tx_flags & E1000_TX_FLAGS_VLAN) {
5696 txd_lower |= E1000_TXD_CMD_VLE;
5697 txd_upper |= (tx_flags & E1000_TX_FLAGS_VLAN_MASK);
5700 if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5701 txd_lower &= ~(E1000_TXD_CMD_IFCS);
5703 if (unlikely(tx_flags & E1000_TX_FLAGS_HWTSTAMP)) {
5704 txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5705 txd_upper |= E1000_TXD_EXTCMD_TSTAMP;
5708 i = tx_ring->next_to_use;
5711 buffer_info = &tx_ring->buffer_info[i];
5712 tx_desc = E1000_TX_DESC(*tx_ring, i);
5713 tx_desc->buffer_addr = cpu_to_le64(buffer_info->dma);
5714 tx_desc->lower.data = cpu_to_le32(txd_lower |
5715 buffer_info->length);
5716 tx_desc->upper.data = cpu_to_le32(txd_upper);
5719 if (i == tx_ring->count)
5721 } while (--count > 0);
5723 tx_desc->lower.data |= cpu_to_le32(adapter->txd_cmd);
5725 /* txd_cmd re-enables FCS, so we'll re-disable it here as desired. */
5726 if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5727 tx_desc->lower.data &= ~(cpu_to_le32(E1000_TXD_CMD_IFCS));
5729 /* Force memory writes to complete before letting h/w
5730 * know there are new descriptors to fetch. (Only
5731 * applicable for weak-ordered memory model archs,
5736 tx_ring->next_to_use = i;
5739 #define MINIMUM_DHCP_PACKET_SIZE 282
5740 static int e1000_transfer_dhcp_info(struct e1000_adapter *adapter,
5741 struct sk_buff *skb)
5743 struct e1000_hw *hw = &adapter->hw;
5746 if (skb_vlan_tag_present(skb) &&
5747 !((skb_vlan_tag_get(skb) == adapter->hw.mng_cookie.vlan_id) &&
5748 (adapter->hw.mng_cookie.status &
5749 E1000_MNG_DHCP_COOKIE_STATUS_VLAN)))
5752 if (skb->len <= MINIMUM_DHCP_PACKET_SIZE)
5755 if (((struct ethhdr *)skb->data)->h_proto != htons(ETH_P_IP))
5759 const struct iphdr *ip = (struct iphdr *)((u8 *)skb->data + 14);
5762 if (ip->protocol != IPPROTO_UDP)
5765 udp = (struct udphdr *)((u8 *)ip + (ip->ihl << 2));
5766 if (ntohs(udp->dest) != 67)
5769 offset = (u8 *)udp + 8 - skb->data;
5770 length = skb->len - offset;
5771 return e1000e_mng_write_dhcp_info(hw, (u8 *)udp + 8, length);
5777 static int __e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5779 struct e1000_adapter *adapter = tx_ring->adapter;
5781 netif_stop_queue(adapter->netdev);
5782 /* Herbert's original patch had:
5783 * smp_mb__after_netif_stop_queue();
5784 * but since that doesn't exist yet, just open code it.
5788 /* We need to check again in a case another CPU has just
5789 * made room available.
5791 if (e1000_desc_unused(tx_ring) < size)
5795 netif_start_queue(adapter->netdev);
5796 ++adapter->restart_queue;
5800 static int e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5802 BUG_ON(size > tx_ring->count);
5804 if (e1000_desc_unused(tx_ring) >= size)
5806 return __e1000_maybe_stop_tx(tx_ring, size);
5809 static netdev_tx_t e1000_xmit_frame(struct sk_buff *skb,
5810 struct net_device *netdev)
5812 struct e1000_adapter *adapter = netdev_priv(netdev);
5813 struct e1000_ring *tx_ring = adapter->tx_ring;
5815 unsigned int tx_flags = 0;
5816 unsigned int len = skb_headlen(skb);
5817 unsigned int nr_frags;
5822 __be16 protocol = vlan_get_protocol(skb);
5824 if (test_bit(__E1000_DOWN, &adapter->state)) {
5825 dev_kfree_skb_any(skb);
5826 return NETDEV_TX_OK;
5829 if (skb->len <= 0) {
5830 dev_kfree_skb_any(skb);
5831 return NETDEV_TX_OK;
5834 /* The minimum packet size with TCTL.PSP set is 17 bytes so
5835 * pad skb in order to meet this minimum size requirement
5837 if (skb_put_padto(skb, 17))
5838 return NETDEV_TX_OK;
5840 mss = skb_shinfo(skb)->gso_size;
5844 /* TSO Workaround for 82571/2/3 Controllers -- if skb->data
5845 * points to just header, pull a few bytes of payload from
5846 * frags into skb->data
5848 hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5849 /* we do this workaround for ES2LAN, but it is un-necessary,
5850 * avoiding it could save a lot of cycles
5852 if (skb->data_len && (hdr_len == len)) {
5853 unsigned int pull_size;
5855 pull_size = min_t(unsigned int, 4, skb->data_len);
5856 if (!__pskb_pull_tail(skb, pull_size)) {
5857 e_err("__pskb_pull_tail failed.\n");
5858 dev_kfree_skb_any(skb);
5859 return NETDEV_TX_OK;
5861 len = skb_headlen(skb);
5865 /* reserve a descriptor for the offload context */
5866 if ((mss) || (skb->ip_summed == CHECKSUM_PARTIAL))
5870 count += DIV_ROUND_UP(len, adapter->tx_fifo_limit);
5872 nr_frags = skb_shinfo(skb)->nr_frags;
5873 for (f = 0; f < nr_frags; f++)
5874 count += DIV_ROUND_UP(skb_frag_size(&skb_shinfo(skb)->frags[f]),
5875 adapter->tx_fifo_limit);
5877 if (adapter->hw.mac.tx_pkt_filtering)
5878 e1000_transfer_dhcp_info(adapter, skb);
5880 /* need: count + 2 desc gap to keep tail from touching
5881 * head, otherwise try next time
5883 if (e1000_maybe_stop_tx(tx_ring, count + 2))
5884 return NETDEV_TX_BUSY;
5886 if (skb_vlan_tag_present(skb)) {
5887 tx_flags |= E1000_TX_FLAGS_VLAN;
5888 tx_flags |= (skb_vlan_tag_get(skb) <<
5889 E1000_TX_FLAGS_VLAN_SHIFT);
5892 first = tx_ring->next_to_use;
5894 tso = e1000_tso(tx_ring, skb, protocol);
5896 dev_kfree_skb_any(skb);
5897 return NETDEV_TX_OK;
5901 tx_flags |= E1000_TX_FLAGS_TSO;
5902 else if (e1000_tx_csum(tx_ring, skb, protocol))
5903 tx_flags |= E1000_TX_FLAGS_CSUM;
5905 /* Old method was to assume IPv4 packet by default if TSO was enabled.
5906 * 82571 hardware supports TSO capabilities for IPv6 as well...
5907 * no longer assume, we must.
5909 if (protocol == htons(ETH_P_IP))
5910 tx_flags |= E1000_TX_FLAGS_IPV4;
5912 if (unlikely(skb->no_fcs))
5913 tx_flags |= E1000_TX_FLAGS_NO_FCS;
5915 /* if count is 0 then mapping error has occurred */
5916 count = e1000_tx_map(tx_ring, skb, first, adapter->tx_fifo_limit,
5919 if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
5920 (adapter->flags & FLAG_HAS_HW_TIMESTAMP)) {
5921 if (!adapter->tx_hwtstamp_skb) {
5922 skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5923 tx_flags |= E1000_TX_FLAGS_HWTSTAMP;
5924 adapter->tx_hwtstamp_skb = skb_get(skb);
5925 adapter->tx_hwtstamp_start = jiffies;
5926 schedule_work(&adapter->tx_hwtstamp_work);
5928 adapter->tx_hwtstamp_skipped++;
5932 skb_tx_timestamp(skb);
5934 netdev_sent_queue(netdev, skb->len);
5935 e1000_tx_queue(tx_ring, tx_flags, count);
5936 /* Make sure there is space in the ring for the next send. */
5937 e1000_maybe_stop_tx(tx_ring,
5939 DIV_ROUND_UP(PAGE_SIZE,
5940 adapter->tx_fifo_limit) + 2));
5942 if (!netdev_xmit_more() ||
5943 netif_xmit_stopped(netdev_get_tx_queue(netdev, 0))) {
5944 if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
5945 e1000e_update_tdt_wa(tx_ring,
5946 tx_ring->next_to_use);
5948 writel(tx_ring->next_to_use, tx_ring->tail);
5951 dev_kfree_skb_any(skb);
5952 tx_ring->buffer_info[first].time_stamp = 0;
5953 tx_ring->next_to_use = first;
5956 return NETDEV_TX_OK;
5960 * e1000_tx_timeout - Respond to a Tx Hang
5961 * @netdev: network interface device structure
5962 * @txqueue: index of the hung queue (unused)
5964 static void e1000_tx_timeout(struct net_device *netdev, unsigned int __always_unused txqueue)
5966 struct e1000_adapter *adapter = netdev_priv(netdev);
5968 /* Do the reset outside of interrupt context */
5969 adapter->tx_timeout_count++;
5970 schedule_work(&adapter->reset_task);
5973 static void e1000_reset_task(struct work_struct *work)
5975 struct e1000_adapter *adapter;
5976 adapter = container_of(work, struct e1000_adapter, reset_task);
5979 /* don't run the task if already down */
5980 if (test_bit(__E1000_DOWN, &adapter->state)) {
5985 if (!(adapter->flags & FLAG_RESTART_NOW)) {
5986 e1000e_dump(adapter);
5987 e_err("Reset adapter unexpectedly\n");
5989 e1000e_reinit_locked(adapter);
5994 * e1000e_get_stats64 - Get System Network Statistics
5995 * @netdev: network interface device structure
5996 * @stats: rtnl_link_stats64 pointer
5998 * Returns the address of the device statistics structure.
6000 void e1000e_get_stats64(struct net_device *netdev,
6001 struct rtnl_link_stats64 *stats)
6003 struct e1000_adapter *adapter = netdev_priv(netdev);
6005 spin_lock(&adapter->stats64_lock);
6006 e1000e_update_stats(adapter);
6007 /* Fill out the OS statistics structure */
6008 stats->rx_bytes = adapter->stats.gorc;
6009 stats->rx_packets = adapter->stats.gprc;
6010 stats->tx_bytes = adapter->stats.gotc;
6011 stats->tx_packets = adapter->stats.gptc;
6012 stats->multicast = adapter->stats.mprc;
6013 stats->collisions = adapter->stats.colc;
6017 /* RLEC on some newer hardware can be incorrect so build
6018 * our own version based on RUC and ROC
6020 stats->rx_errors = adapter->stats.rxerrc +
6021 adapter->stats.crcerrs + adapter->stats.algnerrc +
6022 adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
6023 stats->rx_length_errors = adapter->stats.ruc + adapter->stats.roc;
6024 stats->rx_crc_errors = adapter->stats.crcerrs;
6025 stats->rx_frame_errors = adapter->stats.algnerrc;
6026 stats->rx_missed_errors = adapter->stats.mpc;
6029 stats->tx_errors = adapter->stats.ecol + adapter->stats.latecol;
6030 stats->tx_aborted_errors = adapter->stats.ecol;
6031 stats->tx_window_errors = adapter->stats.latecol;
6032 stats->tx_carrier_errors = adapter->stats.tncrs;
6034 /* Tx Dropped needs to be maintained elsewhere */
6036 spin_unlock(&adapter->stats64_lock);
6040 * e1000_change_mtu - Change the Maximum Transfer Unit
6041 * @netdev: network interface device structure
6042 * @new_mtu: new value for maximum frame size
6044 * Returns 0 on success, negative on failure
6046 static int e1000_change_mtu(struct net_device *netdev, int new_mtu)
6048 struct e1000_adapter *adapter = netdev_priv(netdev);
6049 int max_frame = new_mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
6051 /* Jumbo frame support */
6052 if ((new_mtu > ETH_DATA_LEN) &&
6053 !(adapter->flags & FLAG_HAS_JUMBO_FRAMES)) {
6054 e_err("Jumbo Frames not supported.\n");
6058 /* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
6059 if ((adapter->hw.mac.type >= e1000_pch2lan) &&
6060 !(adapter->flags2 & FLAG2_CRC_STRIPPING) &&
6061 (new_mtu > ETH_DATA_LEN)) {
6062 e_err("Jumbo Frames not supported on this device when CRC stripping is disabled.\n");
6066 while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
6067 usleep_range(1000, 1100);
6068 /* e1000e_down -> e1000e_reset dependent on max_frame_size & mtu */
6069 adapter->max_frame_size = max_frame;
6070 netdev_dbg(netdev, "changing MTU from %d to %d\n",
6071 netdev->mtu, new_mtu);
6072 netdev->mtu = new_mtu;
6074 pm_runtime_get_sync(netdev->dev.parent);
6076 if (netif_running(netdev))
6077 e1000e_down(adapter, true);
6079 /* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
6080 * means we reserve 2 more, this pushes us to allocate from the next
6082 * i.e. RXBUFFER_2048 --> size-4096 slab
6083 * However with the new *_jumbo_rx* routines, jumbo receives will use
6087 if (max_frame <= 2048)
6088 adapter->rx_buffer_len = 2048;
6090 adapter->rx_buffer_len = 4096;
6092 /* adjust allocation if LPE protects us, and we aren't using SBP */
6093 if (max_frame <= (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN))
6094 adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
6096 if (netif_running(netdev))
6099 e1000e_reset(adapter);
6101 pm_runtime_put_sync(netdev->dev.parent);
6103 clear_bit(__E1000_RESETTING, &adapter->state);
6108 static int e1000_mii_ioctl(struct net_device *netdev, struct ifreq *ifr,
6111 struct e1000_adapter *adapter = netdev_priv(netdev);
6112 struct mii_ioctl_data *data = if_mii(ifr);
6114 if (adapter->hw.phy.media_type != e1000_media_type_copper)
6119 data->phy_id = adapter->hw.phy.addr;
6122 e1000_phy_read_status(adapter);
6124 switch (data->reg_num & 0x1F) {
6126 data->val_out = adapter->phy_regs.bmcr;
6129 data->val_out = adapter->phy_regs.bmsr;
6132 data->val_out = (adapter->hw.phy.id >> 16);
6135 data->val_out = (adapter->hw.phy.id & 0xFFFF);
6138 data->val_out = adapter->phy_regs.advertise;
6141 data->val_out = adapter->phy_regs.lpa;
6144 data->val_out = adapter->phy_regs.expansion;
6147 data->val_out = adapter->phy_regs.ctrl1000;
6150 data->val_out = adapter->phy_regs.stat1000;
6153 data->val_out = adapter->phy_regs.estatus;
6167 * e1000e_hwtstamp_set - control hardware time stamping
6168 * @netdev: network interface device structure
6169 * @ifr: interface request
6171 * Outgoing time stamping can be enabled and disabled. Play nice and
6172 * disable it when requested, although it shouldn't cause any overhead
6173 * when no packet needs it. At most one packet in the queue may be
6174 * marked for time stamping, otherwise it would be impossible to tell
6175 * for sure to which packet the hardware time stamp belongs.
6177 * Incoming time stamping has to be configured via the hardware filters.
6178 * Not all combinations are supported, in particular event type has to be
6179 * specified. Matching the kind of event packet is not supported, with the
6180 * exception of "all V2 events regardless of level 2 or 4".
6182 static int e1000e_hwtstamp_set(struct net_device *netdev, struct ifreq *ifr)
6184 struct e1000_adapter *adapter = netdev_priv(netdev);
6185 struct hwtstamp_config config;
6188 if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
6191 ret_val = e1000e_config_hwtstamp(adapter, &config);
6195 switch (config.rx_filter) {
6196 case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
6197 case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
6198 case HWTSTAMP_FILTER_PTP_V2_SYNC:
6199 case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
6200 case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
6201 case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
6202 /* With V2 type filters which specify a Sync or Delay Request,
6203 * Path Delay Request/Response messages are also time stamped
6204 * by hardware so notify the caller the requested packets plus
6205 * some others are time stamped.
6207 config.rx_filter = HWTSTAMP_FILTER_SOME;
6213 return copy_to_user(ifr->ifr_data, &config,
6214 sizeof(config)) ? -EFAULT : 0;
6217 static int e1000e_hwtstamp_get(struct net_device *netdev, struct ifreq *ifr)
6219 struct e1000_adapter *adapter = netdev_priv(netdev);
6221 return copy_to_user(ifr->ifr_data, &adapter->hwtstamp_config,
6222 sizeof(adapter->hwtstamp_config)) ? -EFAULT : 0;
6225 static int e1000_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
6231 return e1000_mii_ioctl(netdev, ifr, cmd);
6233 return e1000e_hwtstamp_set(netdev, ifr);
6235 return e1000e_hwtstamp_get(netdev, ifr);
6241 static int e1000_init_phy_wakeup(struct e1000_adapter *adapter, u32 wufc)
6243 struct e1000_hw *hw = &adapter->hw;
6244 u32 i, mac_reg, wuc;
6245 u16 phy_reg, wuc_enable;
6248 /* copy MAC RARs to PHY RARs */
6249 e1000_copy_rx_addrs_to_phy_ich8lan(hw);
6251 retval = hw->phy.ops.acquire(hw);
6253 e_err("Could not acquire PHY\n");
6257 /* Enable access to wakeup registers on and set page to BM_WUC_PAGE */
6258 retval = e1000_enable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6262 /* copy MAC MTA to PHY MTA - only needed for pchlan */
6263 for (i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
6264 mac_reg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
6265 hw->phy.ops.write_reg_page(hw, BM_MTA(i),
6266 (u16)(mac_reg & 0xFFFF));
6267 hw->phy.ops.write_reg_page(hw, BM_MTA(i) + 1,
6268 (u16)((mac_reg >> 16) & 0xFFFF));
6271 /* configure PHY Rx Control register */
6272 hw->phy.ops.read_reg_page(&adapter->hw, BM_RCTL, &phy_reg);
6273 mac_reg = er32(RCTL);
6274 if (mac_reg & E1000_RCTL_UPE)
6275 phy_reg |= BM_RCTL_UPE;
6276 if (mac_reg & E1000_RCTL_MPE)
6277 phy_reg |= BM_RCTL_MPE;
6278 phy_reg &= ~(BM_RCTL_MO_MASK);
6279 if (mac_reg & E1000_RCTL_MO_3)
6280 phy_reg |= (((mac_reg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
6281 << BM_RCTL_MO_SHIFT);
6282 if (mac_reg & E1000_RCTL_BAM)
6283 phy_reg |= BM_RCTL_BAM;
6284 if (mac_reg & E1000_RCTL_PMCF)
6285 phy_reg |= BM_RCTL_PMCF;
6286 mac_reg = er32(CTRL);
6287 if (mac_reg & E1000_CTRL_RFCE)
6288 phy_reg |= BM_RCTL_RFCE;
6289 hw->phy.ops.write_reg_page(&adapter->hw, BM_RCTL, phy_reg);
6291 wuc = E1000_WUC_PME_EN;
6292 if (wufc & (E1000_WUFC_MAG | E1000_WUFC_LNKC))
6293 wuc |= E1000_WUC_APME;
6295 /* enable PHY wakeup in MAC register */
6297 ew32(WUC, (E1000_WUC_PHY_WAKE | E1000_WUC_APMPME |
6298 E1000_WUC_PME_STATUS | wuc));
6300 /* configure and enable PHY wakeup in PHY registers */
6301 hw->phy.ops.write_reg_page(&adapter->hw, BM_WUFC, wufc);
6302 hw->phy.ops.write_reg_page(&adapter->hw, BM_WUC, wuc);
6304 /* activate PHY wakeup */
6305 wuc_enable |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
6306 retval = e1000_disable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6308 e_err("Could not set PHY Host Wakeup bit\n");
6310 hw->phy.ops.release(hw);
6315 static void e1000e_flush_lpic(struct pci_dev *pdev)
6317 struct net_device *netdev = pci_get_drvdata(pdev);
6318 struct e1000_adapter *adapter = netdev_priv(netdev);
6319 struct e1000_hw *hw = &adapter->hw;
6322 pm_runtime_get_sync(netdev->dev.parent);
6324 ret_val = hw->phy.ops.acquire(hw);
6328 pr_info("EEE TX LPI TIMER: %08X\n",
6329 er32(LPIC) >> E1000_LPIC_LPIET_SHIFT);
6331 hw->phy.ops.release(hw);
6334 pm_runtime_put_sync(netdev->dev.parent);
6337 /* S0ix implementation */
6338 static void e1000e_s0ix_entry_flow(struct e1000_adapter *adapter)
6340 struct e1000_hw *hw = &adapter->hw;
6344 if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID) {
6345 /* Request ME configure the device for S0ix */
6346 mac_data = er32(H2ME);
6347 mac_data |= E1000_H2ME_START_DPG;
6348 mac_data &= ~E1000_H2ME_EXIT_DPG;
6349 ew32(H2ME, mac_data);
6351 /* Request driver configure the device to S0ix */
6352 /* Disable the periodic inband message,
6353 * don't request PCIe clock in K1 page770_17[10:9] = 10b
6355 e1e_rphy(hw, HV_PM_CTRL, &phy_data);
6356 phy_data &= ~HV_PM_CTRL_K1_CLK_REQ;
6357 phy_data |= BIT(10);
6358 e1e_wphy(hw, HV_PM_CTRL, phy_data);
6360 /* Make sure we don't exit K1 every time a new packet arrives
6361 * 772_29[5] = 1 CS_Mode_Stay_In_K1
6363 e1e_rphy(hw, I217_CGFREG, &phy_data);
6365 e1e_wphy(hw, I217_CGFREG, phy_data);
6367 /* Change the MAC/PHY interface to SMBus
6368 * Force the SMBus in PHY page769_23[0] = 1
6369 * Force the SMBus in MAC CTRL_EXT[11] = 1
6371 e1e_rphy(hw, CV_SMB_CTRL, &phy_data);
6372 phy_data |= CV_SMB_CTRL_FORCE_SMBUS;
6373 e1e_wphy(hw, CV_SMB_CTRL, phy_data);
6374 mac_data = er32(CTRL_EXT);
6375 mac_data |= E1000_CTRL_EXT_FORCE_SMBUS;
6376 ew32(CTRL_EXT, mac_data);
6378 /* DFT control: PHY bit: page769_20[0] = 1
6379 * page769_20[7] - PHY PLL stop
6380 * page769_20[8] - PHY go to the electrical idle
6381 * page769_20[9] - PHY serdes disable
6382 * Gate PPW via EXTCNF_CTRL - set 0x0F00[7] = 1
6384 e1e_rphy(hw, I82579_DFT_CTRL, &phy_data);
6389 e1e_wphy(hw, I82579_DFT_CTRL, phy_data);
6391 mac_data = er32(EXTCNF_CTRL);
6392 mac_data |= E1000_EXTCNF_CTRL_GATE_PHY_CFG;
6393 ew32(EXTCNF_CTRL, mac_data);
6395 /* Enable the Dynamic Power Gating in the MAC */
6396 mac_data = er32(FEXTNVM7);
6397 mac_data |= BIT(22);
6398 ew32(FEXTNVM7, mac_data);
6400 /* Disable disconnected cable conditioning for Power Gating */
6401 mac_data = er32(DPGFR);
6403 ew32(DPGFR, mac_data);
6405 /* Don't wake from dynamic Power Gating with clock request */
6406 mac_data = er32(FEXTNVM12);
6407 mac_data |= BIT(12);
6408 ew32(FEXTNVM12, mac_data);
6410 /* Ungate PGCB clock */
6411 mac_data = er32(FEXTNVM9);
6412 mac_data &= ~BIT(28);
6413 ew32(FEXTNVM9, mac_data);
6415 /* Enable K1 off to enable mPHY Power Gating */
6416 mac_data = er32(FEXTNVM6);
6417 mac_data |= BIT(31);
6418 ew32(FEXTNVM6, mac_data);
6420 /* Enable mPHY power gating for any link and speed */
6421 mac_data = er32(FEXTNVM8);
6423 ew32(FEXTNVM8, mac_data);
6425 /* Enable the Dynamic Clock Gating in the DMA and MAC */
6426 mac_data = er32(CTRL_EXT);
6427 mac_data |= E1000_CTRL_EXT_DMA_DYN_CLK_EN;
6428 ew32(CTRL_EXT, mac_data);
6430 /* No MAC DPG gating SLP_S0 in modern standby
6431 * Switch the logic of the lanphypc to use PMC counter
6433 mac_data = er32(FEXTNVM5);
6435 ew32(FEXTNVM5, mac_data);
6438 /* Disable the time synchronization clock */
6439 mac_data = er32(FEXTNVM7);
6440 mac_data |= BIT(31);
6441 mac_data &= ~BIT(0);
6442 ew32(FEXTNVM7, mac_data);
6444 /* Dynamic Power Gating Enable */
6445 mac_data = er32(CTRL_EXT);
6447 ew32(CTRL_EXT, mac_data);
6449 /* Check MAC Tx/Rx packet buffer pointers.
6450 * Reset MAC Tx/Rx packet buffer pointers to suppress any
6451 * pending traffic indication that would prevent power gating.
6453 mac_data = er32(TDFH);
6456 mac_data = er32(TDFT);
6459 mac_data = er32(TDFHS);
6462 mac_data = er32(TDFTS);
6465 mac_data = er32(TDFPC);
6468 mac_data = er32(RDFH);
6471 mac_data = er32(RDFT);
6474 mac_data = er32(RDFHS);
6477 mac_data = er32(RDFTS);
6480 mac_data = er32(RDFPC);
6485 static void e1000e_s0ix_exit_flow(struct e1000_adapter *adapter)
6487 struct e1000_hw *hw = &adapter->hw;
6488 bool firmware_bug = false;
6493 if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID) {
6494 /* Request ME unconfigure the device from S0ix */
6495 mac_data = er32(H2ME);
6496 mac_data &= ~E1000_H2ME_START_DPG;
6497 mac_data |= E1000_H2ME_EXIT_DPG;
6498 ew32(H2ME, mac_data);
6500 /* Poll up to 2.5 seconds for ME to unconfigure DPG.
6501 * If this takes more than 1 second, show a warning indicating a
6504 while (!(er32(EXFWSM) & E1000_EXFWSM_DPG_EXIT_DONE)) {
6505 if (i > 100 && !firmware_bug)
6506 firmware_bug = true;
6509 e_dbg("Timeout (firmware bug): %d msec\n",
6514 usleep_range(10000, 11000);
6517 e_warn("DPG_EXIT_DONE took %d msec. This is a firmware bug\n",
6520 e_dbg("DPG_EXIT_DONE cleared after %d msec\n", i * 10);
6522 /* Request driver unconfigure the device from S0ix */
6524 /* Disable the Dynamic Power Gating in the MAC */
6525 mac_data = er32(FEXTNVM7);
6526 mac_data &= 0xFFBFFFFF;
6527 ew32(FEXTNVM7, mac_data);
6529 /* Disable mPHY power gating for any link and speed */
6530 mac_data = er32(FEXTNVM8);
6531 mac_data &= ~BIT(9);
6532 ew32(FEXTNVM8, mac_data);
6534 /* Disable K1 off */
6535 mac_data = er32(FEXTNVM6);
6536 mac_data &= ~BIT(31);
6537 ew32(FEXTNVM6, mac_data);
6539 /* Disable Ungate PGCB clock */
6540 mac_data = er32(FEXTNVM9);
6541 mac_data |= BIT(28);
6542 ew32(FEXTNVM9, mac_data);
6544 /* Cancel not waking from dynamic
6545 * Power Gating with clock request
6547 mac_data = er32(FEXTNVM12);
6548 mac_data &= ~BIT(12);
6549 ew32(FEXTNVM12, mac_data);
6551 /* Cancel disable disconnected cable conditioning
6554 mac_data = er32(DPGFR);
6555 mac_data &= ~BIT(2);
6556 ew32(DPGFR, mac_data);
6558 /* Disable the Dynamic Clock Gating in the DMA and MAC */
6559 mac_data = er32(CTRL_EXT);
6560 mac_data &= 0xFFF7FFFF;
6561 ew32(CTRL_EXT, mac_data);
6563 /* Revert the lanphypc logic to use the internal Gbe counter
6564 * and not the PMC counter
6566 mac_data = er32(FEXTNVM5);
6567 mac_data &= 0xFFFFFF7F;
6568 ew32(FEXTNVM5, mac_data);
6570 /* Enable the periodic inband message,
6571 * Request PCIe clock in K1 page770_17[10:9] =01b
6573 e1e_rphy(hw, HV_PM_CTRL, &phy_data);
6575 phy_data |= HV_PM_CTRL_K1_CLK_REQ;
6576 e1e_wphy(hw, HV_PM_CTRL, phy_data);
6578 /* Return back configuration
6579 * 772_29[5] = 0 CS_Mode_Stay_In_K1
6581 e1e_rphy(hw, I217_CGFREG, &phy_data);
6583 e1e_wphy(hw, I217_CGFREG, phy_data);
6585 /* Change the MAC/PHY interface to Kumeran
6586 * Unforce the SMBus in PHY page769_23[0] = 0
6587 * Unforce the SMBus in MAC CTRL_EXT[11] = 0
6589 e1e_rphy(hw, CV_SMB_CTRL, &phy_data);
6590 phy_data &= ~CV_SMB_CTRL_FORCE_SMBUS;
6591 e1e_wphy(hw, CV_SMB_CTRL, phy_data);
6592 mac_data = er32(CTRL_EXT);
6593 mac_data &= ~E1000_CTRL_EXT_FORCE_SMBUS;
6594 ew32(CTRL_EXT, mac_data);
6597 /* Disable Dynamic Power Gating */
6598 mac_data = er32(CTRL_EXT);
6599 mac_data &= 0xFFFFFFF7;
6600 ew32(CTRL_EXT, mac_data);
6602 /* Enable the time synchronization clock */
6603 mac_data = er32(FEXTNVM7);
6604 mac_data &= ~BIT(31);
6606 ew32(FEXTNVM7, mac_data);
6609 static int e1000e_pm_freeze(struct device *dev)
6611 struct net_device *netdev = dev_get_drvdata(dev);
6612 struct e1000_adapter *adapter = netdev_priv(netdev);
6617 present = netif_device_present(netdev);
6618 netif_device_detach(netdev);
6620 if (present && netif_running(netdev)) {
6621 int count = E1000_CHECK_RESET_COUNT;
6623 while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6624 usleep_range(10000, 11000);
6626 WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6628 /* Quiesce the device without resetting the hardware */
6629 e1000e_down(adapter, false);
6630 e1000_free_irq(adapter);
6634 e1000e_reset_interrupt_capability(adapter);
6636 /* Allow time for pending master requests to run */
6637 e1000e_disable_pcie_master(&adapter->hw);
6642 static int __e1000_shutdown(struct pci_dev *pdev, bool runtime)
6644 struct net_device *netdev = pci_get_drvdata(pdev);
6645 struct e1000_adapter *adapter = netdev_priv(netdev);
6646 struct e1000_hw *hw = &adapter->hw;
6647 u32 ctrl, ctrl_ext, rctl, status, wufc;
6650 /* Runtime suspend should only enable wakeup for link changes */
6652 wufc = E1000_WUFC_LNKC;
6653 else if (device_may_wakeup(&pdev->dev))
6654 wufc = adapter->wol;
6658 status = er32(STATUS);
6659 if (status & E1000_STATUS_LU)
6660 wufc &= ~E1000_WUFC_LNKC;
6663 e1000_setup_rctl(adapter);
6664 e1000e_set_rx_mode(netdev);
6666 /* turn on all-multi mode if wake on multicast is enabled */
6667 if (wufc & E1000_WUFC_MC) {
6669 rctl |= E1000_RCTL_MPE;
6674 ctrl |= E1000_CTRL_ADVD3WUC;
6675 if (!(adapter->flags2 & FLAG2_HAS_PHY_WAKEUP))
6676 ctrl |= E1000_CTRL_EN_PHY_PWR_MGMT;
6679 if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
6680 adapter->hw.phy.media_type ==
6681 e1000_media_type_internal_serdes) {
6682 /* keep the laser running in D3 */
6683 ctrl_ext = er32(CTRL_EXT);
6684 ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
6685 ew32(CTRL_EXT, ctrl_ext);
6689 e1000e_power_up_phy(adapter);
6691 if (adapter->flags & FLAG_IS_ICH)
6692 e1000_suspend_workarounds_ich8lan(&adapter->hw);
6694 if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6695 /* enable wakeup by the PHY */
6696 retval = e1000_init_phy_wakeup(adapter, wufc);
6700 /* enable wakeup by the MAC */
6702 ew32(WUC, E1000_WUC_PME_EN);
6708 e1000_power_down_phy(adapter);
6711 if (adapter->hw.phy.type == e1000_phy_igp_3) {
6712 e1000e_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
6713 } else if (hw->mac.type >= e1000_pch_lpt) {
6714 if (wufc && !(wufc & (E1000_WUFC_EX | E1000_WUFC_MC | E1000_WUFC_BC)))
6715 /* ULP does not support wake from unicast, multicast
6718 retval = e1000_enable_ulp_lpt_lp(hw, !runtime);
6724 /* Ensure that the appropriate bits are set in LPI_CTRL
6727 if ((hw->phy.type >= e1000_phy_i217) &&
6728 adapter->eee_advert && hw->dev_spec.ich8lan.eee_lp_ability) {
6731 retval = hw->phy.ops.acquire(hw);
6733 retval = e1e_rphy_locked(hw, I82579_LPI_CTRL,
6736 if (adapter->eee_advert &
6737 hw->dev_spec.ich8lan.eee_lp_ability &
6738 I82579_EEE_100_SUPPORTED)
6739 lpi_ctrl |= I82579_LPI_CTRL_100_ENABLE;
6740 if (adapter->eee_advert &
6741 hw->dev_spec.ich8lan.eee_lp_ability &
6742 I82579_EEE_1000_SUPPORTED)
6743 lpi_ctrl |= I82579_LPI_CTRL_1000_ENABLE;
6745 retval = e1e_wphy_locked(hw, I82579_LPI_CTRL,
6749 hw->phy.ops.release(hw);
6752 /* Release control of h/w to f/w. If f/w is AMT enabled, this
6753 * would have already happened in close and is redundant.
6755 e1000e_release_hw_control(adapter);
6757 pci_clear_master(pdev);
6759 /* The pci-e switch on some quad port adapters will report a
6760 * correctable error when the MAC transitions from D0 to D3. To
6761 * prevent this we need to mask off the correctable errors on the
6762 * downstream port of the pci-e switch.
6764 * We don't have the associated upstream bridge while assigning
6765 * the PCI device into guest. For example, the KVM on power is
6768 if (adapter->flags & FLAG_IS_QUAD_PORT) {
6769 struct pci_dev *us_dev = pdev->bus->self;
6775 pcie_capability_read_word(us_dev, PCI_EXP_DEVCTL, &devctl);
6776 pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL,
6777 (devctl & ~PCI_EXP_DEVCTL_CERE));
6779 pci_save_state(pdev);
6780 pci_prepare_to_sleep(pdev);
6782 pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL, devctl);
6789 * __e1000e_disable_aspm - Disable ASPM states
6790 * @pdev: pointer to PCI device struct
6791 * @state: bit-mask of ASPM states to disable
6792 * @locked: indication if this context holds pci_bus_sem locked.
6794 * Some devices *must* have certain ASPM states disabled per hardware errata.
6796 static void __e1000e_disable_aspm(struct pci_dev *pdev, u16 state, int locked)
6798 struct pci_dev *parent = pdev->bus->self;
6799 u16 aspm_dis_mask = 0;
6800 u16 pdev_aspmc, parent_aspmc;
6803 case PCIE_LINK_STATE_L0S:
6804 case PCIE_LINK_STATE_L0S | PCIE_LINK_STATE_L1:
6805 aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L0S;
6806 fallthrough; /* can't have L1 without L0s */
6807 case PCIE_LINK_STATE_L1:
6808 aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L1;
6814 pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6815 pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6818 pcie_capability_read_word(parent, PCI_EXP_LNKCTL,
6820 parent_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6823 /* Nothing to do if the ASPM states to be disabled already are */
6824 if (!(pdev_aspmc & aspm_dis_mask) &&
6825 (!parent || !(parent_aspmc & aspm_dis_mask)))
6828 dev_info(&pdev->dev, "Disabling ASPM %s %s\n",
6829 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L0S) ?
6831 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L1) ?
6834 #ifdef CONFIG_PCIEASPM
6836 pci_disable_link_state_locked(pdev, state);
6838 pci_disable_link_state(pdev, state);
6840 /* Double-check ASPM control. If not disabled by the above, the
6841 * BIOS is preventing that from happening (or CONFIG_PCIEASPM is
6842 * not enabled); override by writing PCI config space directly.
6844 pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6845 pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6847 if (!(aspm_dis_mask & pdev_aspmc))
6851 /* Both device and parent should have the same ASPM setting.
6852 * Disable ASPM in downstream component first and then upstream.
6854 pcie_capability_clear_word(pdev, PCI_EXP_LNKCTL, aspm_dis_mask);
6857 pcie_capability_clear_word(parent, PCI_EXP_LNKCTL,
6862 * e1000e_disable_aspm - Disable ASPM states.
6863 * @pdev: pointer to PCI device struct
6864 * @state: bit-mask of ASPM states to disable
6866 * This function acquires the pci_bus_sem!
6867 * Some devices *must* have certain ASPM states disabled per hardware errata.
6869 static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6871 __e1000e_disable_aspm(pdev, state, 0);
6875 * e1000e_disable_aspm_locked - Disable ASPM states.
6876 * @pdev: pointer to PCI device struct
6877 * @state: bit-mask of ASPM states to disable
6879 * This function must be called with pci_bus_sem acquired!
6880 * Some devices *must* have certain ASPM states disabled per hardware errata.
6882 static void e1000e_disable_aspm_locked(struct pci_dev *pdev, u16 state)
6884 __e1000e_disable_aspm(pdev, state, 1);
6887 static int e1000e_pm_thaw(struct device *dev)
6889 struct net_device *netdev = dev_get_drvdata(dev);
6890 struct e1000_adapter *adapter = netdev_priv(netdev);
6893 e1000e_set_interrupt_capability(adapter);
6896 if (netif_running(netdev)) {
6897 rc = e1000_request_irq(adapter);
6904 netif_device_attach(netdev);
6911 static int __e1000_resume(struct pci_dev *pdev)
6913 struct net_device *netdev = pci_get_drvdata(pdev);
6914 struct e1000_adapter *adapter = netdev_priv(netdev);
6915 struct e1000_hw *hw = &adapter->hw;
6916 u16 aspm_disable_flag = 0;
6918 if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6919 aspm_disable_flag = PCIE_LINK_STATE_L0S;
6920 if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6921 aspm_disable_flag |= PCIE_LINK_STATE_L1;
6922 if (aspm_disable_flag)
6923 e1000e_disable_aspm(pdev, aspm_disable_flag);
6925 pci_set_master(pdev);
6927 if (hw->mac.type >= e1000_pch2lan)
6928 e1000_resume_workarounds_pchlan(&adapter->hw);
6930 e1000e_power_up_phy(adapter);
6932 /* report the system wakeup cause from S3/S4 */
6933 if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6936 e1e_rphy(&adapter->hw, BM_WUS, &phy_data);
6938 e_info("PHY Wakeup cause - %s\n",
6939 phy_data & E1000_WUS_EX ? "Unicast Packet" :
6940 phy_data & E1000_WUS_MC ? "Multicast Packet" :
6941 phy_data & E1000_WUS_BC ? "Broadcast Packet" :
6942 phy_data & E1000_WUS_MAG ? "Magic Packet" :
6943 phy_data & E1000_WUS_LNKC ?
6944 "Link Status Change" : "other");
6946 e1e_wphy(&adapter->hw, BM_WUS, ~0);
6948 u32 wus = er32(WUS);
6951 e_info("MAC Wakeup cause - %s\n",
6952 wus & E1000_WUS_EX ? "Unicast Packet" :
6953 wus & E1000_WUS_MC ? "Multicast Packet" :
6954 wus & E1000_WUS_BC ? "Broadcast Packet" :
6955 wus & E1000_WUS_MAG ? "Magic Packet" :
6956 wus & E1000_WUS_LNKC ? "Link Status Change" :
6962 e1000e_reset(adapter);
6964 e1000_init_manageability_pt(adapter);
6966 /* If the controller has AMT, do not set DRV_LOAD until the interface
6967 * is up. For all other cases, let the f/w know that the h/w is now
6968 * under the control of the driver.
6970 if (!(adapter->flags & FLAG_HAS_AMT))
6971 e1000e_get_hw_control(adapter);
6976 static __maybe_unused int e1000e_pm_prepare(struct device *dev)
6978 return pm_runtime_suspended(dev) &&
6979 pm_suspend_via_firmware();
6982 static __maybe_unused int e1000e_pm_suspend(struct device *dev)
6984 struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6985 struct e1000_adapter *adapter = netdev_priv(netdev);
6986 struct pci_dev *pdev = to_pci_dev(dev);
6989 e1000e_flush_lpic(pdev);
6991 e1000e_pm_freeze(dev);
6993 rc = __e1000_shutdown(pdev, false);
6995 e1000e_pm_thaw(dev);
6997 /* Introduce S0ix implementation */
6998 if (adapter->flags2 & FLAG2_ENABLE_S0IX_FLOWS)
6999 e1000e_s0ix_entry_flow(adapter);
7005 static __maybe_unused int e1000e_pm_resume(struct device *dev)
7007 struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
7008 struct e1000_adapter *adapter = netdev_priv(netdev);
7009 struct pci_dev *pdev = to_pci_dev(dev);
7012 /* Introduce S0ix implementation */
7013 if (adapter->flags2 & FLAG2_ENABLE_S0IX_FLOWS)
7014 e1000e_s0ix_exit_flow(adapter);
7016 rc = __e1000_resume(pdev);
7020 return e1000e_pm_thaw(dev);
7023 static __maybe_unused int e1000e_pm_runtime_idle(struct device *dev)
7025 struct net_device *netdev = dev_get_drvdata(dev);
7026 struct e1000_adapter *adapter = netdev_priv(netdev);
7029 eee_lp = adapter->hw.dev_spec.ich8lan.eee_lp_ability;
7031 if (!e1000e_has_link(adapter)) {
7032 adapter->hw.dev_spec.ich8lan.eee_lp_ability = eee_lp;
7033 pm_schedule_suspend(dev, 5 * MSEC_PER_SEC);
7039 static __maybe_unused int e1000e_pm_runtime_resume(struct device *dev)
7041 struct pci_dev *pdev = to_pci_dev(dev);
7042 struct net_device *netdev = pci_get_drvdata(pdev);
7043 struct e1000_adapter *adapter = netdev_priv(netdev);
7046 rc = __e1000_resume(pdev);
7050 if (netdev->flags & IFF_UP)
7056 static __maybe_unused int e1000e_pm_runtime_suspend(struct device *dev)
7058 struct pci_dev *pdev = to_pci_dev(dev);
7059 struct net_device *netdev = pci_get_drvdata(pdev);
7060 struct e1000_adapter *adapter = netdev_priv(netdev);
7062 if (netdev->flags & IFF_UP) {
7063 int count = E1000_CHECK_RESET_COUNT;
7065 while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
7066 usleep_range(10000, 11000);
7068 WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
7070 /* Down the device without resetting the hardware */
7071 e1000e_down(adapter, false);
7074 if (__e1000_shutdown(pdev, true)) {
7075 e1000e_pm_runtime_resume(dev);
7082 static void e1000_shutdown(struct pci_dev *pdev)
7084 e1000e_flush_lpic(pdev);
7086 e1000e_pm_freeze(&pdev->dev);
7088 __e1000_shutdown(pdev, false);
7091 #ifdef CONFIG_NET_POLL_CONTROLLER
7093 static irqreturn_t e1000_intr_msix(int __always_unused irq, void *data)
7095 struct net_device *netdev = data;
7096 struct e1000_adapter *adapter = netdev_priv(netdev);
7098 if (adapter->msix_entries) {
7099 int vector, msix_irq;
7102 msix_irq = adapter->msix_entries[vector].vector;
7103 if (disable_hardirq(msix_irq))
7104 e1000_intr_msix_rx(msix_irq, netdev);
7105 enable_irq(msix_irq);
7108 msix_irq = adapter->msix_entries[vector].vector;
7109 if (disable_hardirq(msix_irq))
7110 e1000_intr_msix_tx(msix_irq, netdev);
7111 enable_irq(msix_irq);
7114 msix_irq = adapter->msix_entries[vector].vector;
7115 if (disable_hardirq(msix_irq))
7116 e1000_msix_other(msix_irq, netdev);
7117 enable_irq(msix_irq);
7125 * @netdev: network interface device structure
7127 * Polling 'interrupt' - used by things like netconsole to send skbs
7128 * without having to re-enable interrupts. It's not called while
7129 * the interrupt routine is executing.
7131 static void e1000_netpoll(struct net_device *netdev)
7133 struct e1000_adapter *adapter = netdev_priv(netdev);
7135 switch (adapter->int_mode) {
7136 case E1000E_INT_MODE_MSIX:
7137 e1000_intr_msix(adapter->pdev->irq, netdev);
7139 case E1000E_INT_MODE_MSI:
7140 if (disable_hardirq(adapter->pdev->irq))
7141 e1000_intr_msi(adapter->pdev->irq, netdev);
7142 enable_irq(adapter->pdev->irq);
7144 default: /* E1000E_INT_MODE_LEGACY */
7145 if (disable_hardirq(adapter->pdev->irq))
7146 e1000_intr(adapter->pdev->irq, netdev);
7147 enable_irq(adapter->pdev->irq);
7154 * e1000_io_error_detected - called when PCI error is detected
7155 * @pdev: Pointer to PCI device
7156 * @state: The current pci connection state
7158 * This function is called after a PCI bus error affecting
7159 * this device has been detected.
7161 static pci_ers_result_t e1000_io_error_detected(struct pci_dev *pdev,
7162 pci_channel_state_t state)
7164 e1000e_pm_freeze(&pdev->dev);
7166 if (state == pci_channel_io_perm_failure)
7167 return PCI_ERS_RESULT_DISCONNECT;
7169 pci_disable_device(pdev);
7171 /* Request a slot reset. */
7172 return PCI_ERS_RESULT_NEED_RESET;
7176 * e1000_io_slot_reset - called after the pci bus has been reset.
7177 * @pdev: Pointer to PCI device
7179 * Restart the card from scratch, as if from a cold-boot. Implementation
7180 * resembles the first-half of the e1000e_pm_resume routine.
7182 static pci_ers_result_t e1000_io_slot_reset(struct pci_dev *pdev)
7184 struct net_device *netdev = pci_get_drvdata(pdev);
7185 struct e1000_adapter *adapter = netdev_priv(netdev);
7186 struct e1000_hw *hw = &adapter->hw;
7187 u16 aspm_disable_flag = 0;
7189 pci_ers_result_t result;
7191 if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
7192 aspm_disable_flag = PCIE_LINK_STATE_L0S;
7193 if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
7194 aspm_disable_flag |= PCIE_LINK_STATE_L1;
7195 if (aspm_disable_flag)
7196 e1000e_disable_aspm_locked(pdev, aspm_disable_flag);
7198 err = pci_enable_device_mem(pdev);
7201 "Cannot re-enable PCI device after reset.\n");
7202 result = PCI_ERS_RESULT_DISCONNECT;
7204 pdev->state_saved = true;
7205 pci_restore_state(pdev);
7206 pci_set_master(pdev);
7208 pci_enable_wake(pdev, PCI_D3hot, 0);
7209 pci_enable_wake(pdev, PCI_D3cold, 0);
7211 e1000e_reset(adapter);
7213 result = PCI_ERS_RESULT_RECOVERED;
7220 * e1000_io_resume - called when traffic can start flowing again.
7221 * @pdev: Pointer to PCI device
7223 * This callback is called when the error recovery driver tells us that
7224 * its OK to resume normal operation. Implementation resembles the
7225 * second-half of the e1000e_pm_resume routine.
7227 static void e1000_io_resume(struct pci_dev *pdev)
7229 struct net_device *netdev = pci_get_drvdata(pdev);
7230 struct e1000_adapter *adapter = netdev_priv(netdev);
7232 e1000_init_manageability_pt(adapter);
7234 e1000e_pm_thaw(&pdev->dev);
7236 /* If the controller has AMT, do not set DRV_LOAD until the interface
7237 * is up. For all other cases, let the f/w know that the h/w is now
7238 * under the control of the driver.
7240 if (!(adapter->flags & FLAG_HAS_AMT))
7241 e1000e_get_hw_control(adapter);
7244 static void e1000_print_device_info(struct e1000_adapter *adapter)
7246 struct e1000_hw *hw = &adapter->hw;
7247 struct net_device *netdev = adapter->netdev;
7249 u8 pba_str[E1000_PBANUM_LENGTH];
7251 /* print bus type/speed/width info */
7252 e_info("(PCI Express:2.5GT/s:%s) %pM\n",
7254 ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
7258 e_info("Intel(R) PRO/%s Network Connection\n",
7259 (hw->phy.type == e1000_phy_ife) ? "10/100" : "1000");
7260 ret_val = e1000_read_pba_string_generic(hw, pba_str,
7261 E1000_PBANUM_LENGTH);
7263 strlcpy((char *)pba_str, "Unknown", sizeof(pba_str));
7264 e_info("MAC: %d, PHY: %d, PBA No: %s\n",
7265 hw->mac.type, hw->phy.type, pba_str);
7268 static void e1000_eeprom_checks(struct e1000_adapter *adapter)
7270 struct e1000_hw *hw = &adapter->hw;
7274 if (hw->mac.type != e1000_82573)
7277 ret_val = e1000_read_nvm(hw, NVM_INIT_CONTROL2_REG, 1, &buf);
7279 if (!ret_val && (!(buf & BIT(0)))) {
7280 /* Deep Smart Power Down (DSPD) */
7281 dev_warn(&adapter->pdev->dev,
7282 "Warning: detected DSPD enabled in EEPROM\n");
7286 static netdev_features_t e1000_fix_features(struct net_device *netdev,
7287 netdev_features_t features)
7289 struct e1000_adapter *adapter = netdev_priv(netdev);
7290 struct e1000_hw *hw = &adapter->hw;
7292 /* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
7293 if ((hw->mac.type >= e1000_pch2lan) && (netdev->mtu > ETH_DATA_LEN))
7294 features &= ~NETIF_F_RXFCS;
7296 /* Since there is no support for separate Rx/Tx vlan accel
7297 * enable/disable make sure Tx flag is always in same state as Rx.
7299 if (features & NETIF_F_HW_VLAN_CTAG_RX)
7300 features |= NETIF_F_HW_VLAN_CTAG_TX;
7302 features &= ~NETIF_F_HW_VLAN_CTAG_TX;
7307 static int e1000_set_features(struct net_device *netdev,
7308 netdev_features_t features)
7310 struct e1000_adapter *adapter = netdev_priv(netdev);
7311 netdev_features_t changed = features ^ netdev->features;
7313 if (changed & (NETIF_F_TSO | NETIF_F_TSO6))
7314 adapter->flags |= FLAG_TSO_FORCE;
7316 if (!(changed & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX |
7317 NETIF_F_RXCSUM | NETIF_F_RXHASH | NETIF_F_RXFCS |
7321 if (changed & NETIF_F_RXFCS) {
7322 if (features & NETIF_F_RXFCS) {
7323 adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7325 /* We need to take it back to defaults, which might mean
7326 * stripping is still disabled at the adapter level.
7328 if (adapter->flags2 & FLAG2_DFLT_CRC_STRIPPING)
7329 adapter->flags2 |= FLAG2_CRC_STRIPPING;
7331 adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7335 netdev->features = features;
7337 if (netif_running(netdev))
7338 e1000e_reinit_locked(adapter);
7340 e1000e_reset(adapter);
7345 static const struct net_device_ops e1000e_netdev_ops = {
7346 .ndo_open = e1000e_open,
7347 .ndo_stop = e1000e_close,
7348 .ndo_start_xmit = e1000_xmit_frame,
7349 .ndo_get_stats64 = e1000e_get_stats64,
7350 .ndo_set_rx_mode = e1000e_set_rx_mode,
7351 .ndo_set_mac_address = e1000_set_mac,
7352 .ndo_change_mtu = e1000_change_mtu,
7353 .ndo_eth_ioctl = e1000_ioctl,
7354 .ndo_tx_timeout = e1000_tx_timeout,
7355 .ndo_validate_addr = eth_validate_addr,
7357 .ndo_vlan_rx_add_vid = e1000_vlan_rx_add_vid,
7358 .ndo_vlan_rx_kill_vid = e1000_vlan_rx_kill_vid,
7359 #ifdef CONFIG_NET_POLL_CONTROLLER
7360 .ndo_poll_controller = e1000_netpoll,
7362 .ndo_set_features = e1000_set_features,
7363 .ndo_fix_features = e1000_fix_features,
7364 .ndo_features_check = passthru_features_check,
7368 * e1000_probe - Device Initialization Routine
7369 * @pdev: PCI device information struct
7370 * @ent: entry in e1000_pci_tbl
7372 * Returns 0 on success, negative on failure
7374 * e1000_probe initializes an adapter identified by a pci_dev structure.
7375 * The OS initialization, configuring of the adapter private structure,
7376 * and a hardware reset occur.
7378 static int e1000_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7380 struct net_device *netdev;
7381 struct e1000_adapter *adapter;
7382 struct e1000_hw *hw;
7383 const struct e1000_info *ei = e1000_info_tbl[ent->driver_data];
7384 resource_size_t mmio_start, mmio_len;
7385 resource_size_t flash_start, flash_len;
7386 static int cards_found;
7387 u16 aspm_disable_flag = 0;
7388 int bars, i, err, pci_using_dac;
7389 u16 eeprom_data = 0;
7390 u16 eeprom_apme_mask = E1000_EEPROM_APME;
7393 if (ei->flags2 & FLAG2_DISABLE_ASPM_L0S)
7394 aspm_disable_flag = PCIE_LINK_STATE_L0S;
7395 if (ei->flags2 & FLAG2_DISABLE_ASPM_L1)
7396 aspm_disable_flag |= PCIE_LINK_STATE_L1;
7397 if (aspm_disable_flag)
7398 e1000e_disable_aspm(pdev, aspm_disable_flag);
7400 err = pci_enable_device_mem(pdev);
7405 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
7409 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
7412 "No usable DMA configuration, aborting\n");
7417 bars = pci_select_bars(pdev, IORESOURCE_MEM);
7418 err = pci_request_selected_regions_exclusive(pdev, bars,
7419 e1000e_driver_name);
7423 /* AER (Advanced Error Reporting) hooks */
7424 pci_enable_pcie_error_reporting(pdev);
7426 pci_set_master(pdev);
7427 /* PCI config space info */
7428 err = pci_save_state(pdev);
7430 goto err_alloc_etherdev;
7433 netdev = alloc_etherdev(sizeof(struct e1000_adapter));
7435 goto err_alloc_etherdev;
7437 SET_NETDEV_DEV(netdev, &pdev->dev);
7439 netdev->irq = pdev->irq;
7441 pci_set_drvdata(pdev, netdev);
7442 adapter = netdev_priv(netdev);
7444 adapter->netdev = netdev;
7445 adapter->pdev = pdev;
7447 adapter->pba = ei->pba;
7448 adapter->flags = ei->flags;
7449 adapter->flags2 = ei->flags2;
7450 adapter->hw.adapter = adapter;
7451 adapter->hw.mac.type = ei->mac;
7452 adapter->max_hw_frame_size = ei->max_hw_frame_size;
7453 adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
7455 mmio_start = pci_resource_start(pdev, 0);
7456 mmio_len = pci_resource_len(pdev, 0);
7459 adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
7460 if (!adapter->hw.hw_addr)
7463 if ((adapter->flags & FLAG_HAS_FLASH) &&
7464 (pci_resource_flags(pdev, 1) & IORESOURCE_MEM) &&
7465 (hw->mac.type < e1000_pch_spt)) {
7466 flash_start = pci_resource_start(pdev, 1);
7467 flash_len = pci_resource_len(pdev, 1);
7468 adapter->hw.flash_address = ioremap(flash_start, flash_len);
7469 if (!adapter->hw.flash_address)
7473 /* Set default EEE advertisement */
7474 if (adapter->flags2 & FLAG2_HAS_EEE)
7475 adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
7477 /* construct the net_device struct */
7478 netdev->netdev_ops = &e1000e_netdev_ops;
7479 e1000e_set_ethtool_ops(netdev);
7480 netdev->watchdog_timeo = 5 * HZ;
7481 netif_napi_add(netdev, &adapter->napi, e1000e_poll, 64);
7482 strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
7484 netdev->mem_start = mmio_start;
7485 netdev->mem_end = mmio_start + mmio_len;
7487 adapter->bd_number = cards_found++;
7489 e1000e_check_options(adapter);
7491 /* setup adapter struct */
7492 err = e1000_sw_init(adapter);
7496 memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
7497 memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
7498 memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
7500 err = ei->get_variants(adapter);
7504 if ((adapter->flags & FLAG_IS_ICH) &&
7505 (adapter->flags & FLAG_READ_ONLY_NVM) &&
7506 (hw->mac.type < e1000_pch_spt))
7507 e1000e_write_protect_nvm_ich8lan(&adapter->hw);
7509 hw->mac.ops.get_bus_info(&adapter->hw);
7511 adapter->hw.phy.autoneg_wait_to_complete = 0;
7513 /* Copper options */
7514 if (adapter->hw.phy.media_type == e1000_media_type_copper) {
7515 adapter->hw.phy.mdix = AUTO_ALL_MODES;
7516 adapter->hw.phy.disable_polarity_correction = 0;
7517 adapter->hw.phy.ms_type = e1000_ms_hw_default;
7520 if (hw->phy.ops.check_reset_block && hw->phy.ops.check_reset_block(hw))
7521 dev_info(&pdev->dev,
7522 "PHY reset is blocked due to SOL/IDER session.\n");
7524 /* Set initial default active device features */
7525 netdev->features = (NETIF_F_SG |
7526 NETIF_F_HW_VLAN_CTAG_RX |
7527 NETIF_F_HW_VLAN_CTAG_TX |
7534 /* Set user-changeable features (subset of all device features) */
7535 netdev->hw_features = netdev->features;
7536 netdev->hw_features |= NETIF_F_RXFCS;
7537 netdev->priv_flags |= IFF_SUPP_NOFCS;
7538 netdev->hw_features |= NETIF_F_RXALL;
7540 if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER)
7541 netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
7543 netdev->vlan_features |= (NETIF_F_SG |
7548 netdev->priv_flags |= IFF_UNICAST_FLT;
7550 if (pci_using_dac) {
7551 netdev->features |= NETIF_F_HIGHDMA;
7552 netdev->vlan_features |= NETIF_F_HIGHDMA;
7555 /* MTU range: 68 - max_hw_frame_size */
7556 netdev->min_mtu = ETH_MIN_MTU;
7557 netdev->max_mtu = adapter->max_hw_frame_size -
7558 (VLAN_ETH_HLEN + ETH_FCS_LEN);
7560 if (e1000e_enable_mng_pass_thru(&adapter->hw))
7561 adapter->flags |= FLAG_MNG_PT_ENABLED;
7563 /* before reading the NVM, reset the controller to
7564 * put the device in a known good starting state
7566 adapter->hw.mac.ops.reset_hw(&adapter->hw);
7568 /* systems with ASPM and others may see the checksum fail on the first
7569 * attempt. Let's give it a few tries
7572 if (e1000_validate_nvm_checksum(&adapter->hw) >= 0)
7575 dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
7581 e1000_eeprom_checks(adapter);
7583 /* copy the MAC address */
7584 if (e1000e_read_mac_addr(&adapter->hw))
7586 "NVM Read Error while reading MAC address\n");
7588 eth_hw_addr_set(netdev, adapter->hw.mac.addr);
7590 if (!is_valid_ether_addr(netdev->dev_addr)) {
7591 dev_err(&pdev->dev, "Invalid MAC Address: %pM\n",
7597 timer_setup(&adapter->watchdog_timer, e1000_watchdog, 0);
7598 timer_setup(&adapter->phy_info_timer, e1000_update_phy_info, 0);
7600 INIT_WORK(&adapter->reset_task, e1000_reset_task);
7601 INIT_WORK(&adapter->watchdog_task, e1000_watchdog_task);
7602 INIT_WORK(&adapter->downshift_task, e1000e_downshift_workaround);
7603 INIT_WORK(&adapter->update_phy_task, e1000e_update_phy_task);
7604 INIT_WORK(&adapter->print_hang_task, e1000_print_hw_hang);
7606 /* Initialize link parameters. User can change them with ethtool */
7607 adapter->hw.mac.autoneg = 1;
7608 adapter->fc_autoneg = true;
7609 adapter->hw.fc.requested_mode = e1000_fc_default;
7610 adapter->hw.fc.current_mode = e1000_fc_default;
7611 adapter->hw.phy.autoneg_advertised = 0x2f;
7613 /* Initial Wake on LAN setting - If APM wake is enabled in
7614 * the EEPROM, enable the ACPI Magic Packet filter
7616 if (adapter->flags & FLAG_APME_IN_WUC) {
7617 /* APME bit in EEPROM is mapped to WUC.APME */
7618 eeprom_data = er32(WUC);
7619 eeprom_apme_mask = E1000_WUC_APME;
7620 if ((hw->mac.type > e1000_ich10lan) &&
7621 (eeprom_data & E1000_WUC_PHY_WAKE))
7622 adapter->flags2 |= FLAG2_HAS_PHY_WAKEUP;
7623 } else if (adapter->flags & FLAG_APME_IN_CTRL3) {
7624 if (adapter->flags & FLAG_APME_CHECK_PORT_B &&
7625 (adapter->hw.bus.func == 1))
7626 ret_val = e1000_read_nvm(&adapter->hw,
7627 NVM_INIT_CONTROL3_PORT_B,
7630 ret_val = e1000_read_nvm(&adapter->hw,
7631 NVM_INIT_CONTROL3_PORT_A,
7635 /* fetch WoL from EEPROM */
7637 e_dbg("NVM read error getting WoL initial values: %d\n", ret_val);
7638 else if (eeprom_data & eeprom_apme_mask)
7639 adapter->eeprom_wol |= E1000_WUFC_MAG;
7641 /* now that we have the eeprom settings, apply the special cases
7642 * where the eeprom may be wrong or the board simply won't support
7643 * wake on lan on a particular port
7645 if (!(adapter->flags & FLAG_HAS_WOL))
7646 adapter->eeprom_wol = 0;
7648 /* initialize the wol settings based on the eeprom settings */
7649 adapter->wol = adapter->eeprom_wol;
7651 /* make sure adapter isn't asleep if manageability is enabled */
7652 if (adapter->wol || (adapter->flags & FLAG_MNG_PT_ENABLED) ||
7653 (hw->mac.ops.check_mng_mode(hw)))
7654 device_wakeup_enable(&pdev->dev);
7656 /* save off EEPROM version number */
7657 ret_val = e1000_read_nvm(&adapter->hw, 5, 1, &adapter->eeprom_vers);
7660 e_dbg("NVM read error getting EEPROM version: %d\n", ret_val);
7661 adapter->eeprom_vers = 0;
7664 /* init PTP hardware clock */
7665 e1000e_ptp_init(adapter);
7667 /* reset the hardware with the new settings */
7668 e1000e_reset(adapter);
7670 /* If the controller has AMT, do not set DRV_LOAD until the interface
7671 * is up. For all other cases, let the f/w know that the h/w is now
7672 * under the control of the driver.
7674 if (!(adapter->flags & FLAG_HAS_AMT))
7675 e1000e_get_hw_control(adapter);
7677 if (hw->mac.type >= e1000_pch_cnp)
7678 adapter->flags2 |= FLAG2_ENABLE_S0IX_FLOWS;
7680 strlcpy(netdev->name, "eth%d", sizeof(netdev->name));
7681 err = register_netdev(netdev);
7685 /* carrier off reporting is important to ethtool even BEFORE open */
7686 netif_carrier_off(netdev);
7688 e1000_print_device_info(adapter);
7690 dev_pm_set_driver_flags(&pdev->dev, DPM_FLAG_SMART_PREPARE);
7692 if (pci_dev_run_wake(pdev) && hw->mac.type != e1000_pch_cnp)
7693 pm_runtime_put_noidle(&pdev->dev);
7698 if (!(adapter->flags & FLAG_HAS_AMT))
7699 e1000e_release_hw_control(adapter);
7701 if (hw->phy.ops.check_reset_block && !hw->phy.ops.check_reset_block(hw))
7702 e1000_phy_hw_reset(&adapter->hw);
7704 kfree(adapter->tx_ring);
7705 kfree(adapter->rx_ring);
7707 if ((adapter->hw.flash_address) && (hw->mac.type < e1000_pch_spt))
7708 iounmap(adapter->hw.flash_address);
7709 e1000e_reset_interrupt_capability(adapter);
7711 iounmap(adapter->hw.hw_addr);
7713 free_netdev(netdev);
7715 pci_disable_pcie_error_reporting(pdev);
7716 pci_release_mem_regions(pdev);
7719 pci_disable_device(pdev);
7724 * e1000_remove - Device Removal Routine
7725 * @pdev: PCI device information struct
7727 * e1000_remove is called by the PCI subsystem to alert the driver
7728 * that it should release a PCI device. This could be caused by a
7729 * Hot-Plug event, or because the driver is going to be removed from
7732 static void e1000_remove(struct pci_dev *pdev)
7734 struct net_device *netdev = pci_get_drvdata(pdev);
7735 struct e1000_adapter *adapter = netdev_priv(netdev);
7737 e1000e_ptp_remove(adapter);
7739 /* The timers may be rescheduled, so explicitly disable them
7740 * from being rescheduled.
7742 set_bit(__E1000_DOWN, &adapter->state);
7743 del_timer_sync(&adapter->watchdog_timer);
7744 del_timer_sync(&adapter->phy_info_timer);
7746 cancel_work_sync(&adapter->reset_task);
7747 cancel_work_sync(&adapter->watchdog_task);
7748 cancel_work_sync(&adapter->downshift_task);
7749 cancel_work_sync(&adapter->update_phy_task);
7750 cancel_work_sync(&adapter->print_hang_task);
7752 if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
7753 cancel_work_sync(&adapter->tx_hwtstamp_work);
7754 if (adapter->tx_hwtstamp_skb) {
7755 dev_consume_skb_any(adapter->tx_hwtstamp_skb);
7756 adapter->tx_hwtstamp_skb = NULL;
7760 unregister_netdev(netdev);
7762 if (pci_dev_run_wake(pdev))
7763 pm_runtime_get_noresume(&pdev->dev);
7765 /* Release control of h/w to f/w. If f/w is AMT enabled, this
7766 * would have already happened in close and is redundant.
7768 e1000e_release_hw_control(adapter);
7770 e1000e_reset_interrupt_capability(adapter);
7771 kfree(adapter->tx_ring);
7772 kfree(adapter->rx_ring);
7774 iounmap(adapter->hw.hw_addr);
7775 if ((adapter->hw.flash_address) &&
7776 (adapter->hw.mac.type < e1000_pch_spt))
7777 iounmap(adapter->hw.flash_address);
7778 pci_release_mem_regions(pdev);
7780 free_netdev(netdev);
7783 pci_disable_pcie_error_reporting(pdev);
7785 pci_disable_device(pdev);
7788 /* PCI Error Recovery (ERS) */
7789 static const struct pci_error_handlers e1000_err_handler = {
7790 .error_detected = e1000_io_error_detected,
7791 .slot_reset = e1000_io_slot_reset,
7792 .resume = e1000_io_resume,
7795 static const struct pci_device_id e1000_pci_tbl[] = {
7796 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_COPPER), board_82571 },
7797 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_FIBER), board_82571 },
7798 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER), board_82571 },
7799 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER_LP),
7801 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_FIBER), board_82571 },
7802 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES), board_82571 },
7803 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_DUAL), board_82571 },
7804 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_QUAD), board_82571 },
7805 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571PT_QUAD_COPPER), board_82571 },
7807 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI), board_82572 },
7808 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_COPPER), board_82572 },
7809 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_FIBER), board_82572 },
7810 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_SERDES), board_82572 },
7812 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E), board_82573 },
7813 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E_IAMT), board_82573 },
7814 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82573L), board_82573 },
7816 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82574L), board_82574 },
7817 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82574LA), board_82574 },
7818 { PCI_VDEVICE(INTEL, E1000_DEV_ID_82583V), board_82583 },
7820 { PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_DPT),
7821 board_80003es2lan },
7822 { PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_SPT),
7823 board_80003es2lan },
7824 { PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_DPT),
7825 board_80003es2lan },
7826 { PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_SPT),
7827 board_80003es2lan },
7829 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE), board_ich8lan },
7830 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_G), board_ich8lan },
7831 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_GT), board_ich8lan },
7832 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_AMT), board_ich8lan },
7833 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_C), board_ich8lan },
7834 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M), board_ich8lan },
7835 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M_AMT), board_ich8lan },
7836 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_82567V_3), board_ich8lan },
7838 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE), board_ich9lan },
7839 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_G), board_ich9lan },
7840 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_GT), board_ich9lan },
7841 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_AMT), board_ich9lan },
7842 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_C), board_ich9lan },
7843 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_BM), board_ich9lan },
7844 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M), board_ich9lan },
7845 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_AMT), board_ich9lan },
7846 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_V), board_ich9lan },
7848 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LM), board_ich9lan },
7849 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LF), board_ich9lan },
7850 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_V), board_ich9lan },
7852 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LM), board_ich10lan },
7853 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LF), board_ich10lan },
7854 { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_V), board_ich10lan },
7856 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LM), board_pchlan },
7857 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LC), board_pchlan },
7858 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DM), board_pchlan },
7859 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DC), board_pchlan },
7861 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_LM), board_pch2lan },
7862 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_V), board_pch2lan },
7864 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_LM), board_pch_lpt },
7865 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_V), board_pch_lpt },
7866 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_LM), board_pch_lpt },
7867 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_V), board_pch_lpt },
7868 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM2), board_pch_lpt },
7869 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V2), board_pch_lpt },
7870 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM3), board_pch_lpt },
7871 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V3), board_pch_lpt },
7872 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM), board_pch_spt },
7873 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V), board_pch_spt },
7874 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM2), board_pch_spt },
7875 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V2), board_pch_spt },
7876 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LBG_I219_LM3), board_pch_spt },
7877 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM4), board_pch_spt },
7878 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V4), board_pch_spt },
7879 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM5), board_pch_spt },
7880 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V5), board_pch_spt },
7881 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM6), board_pch_cnp },
7882 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V6), board_pch_cnp },
7883 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM7), board_pch_cnp },
7884 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V7), board_pch_cnp },
7885 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM8), board_pch_cnp },
7886 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V8), board_pch_cnp },
7887 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM9), board_pch_cnp },
7888 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V9), board_pch_cnp },
7889 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM10), board_pch_cnp },
7890 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V10), board_pch_cnp },
7891 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM11), board_pch_cnp },
7892 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V11), board_pch_cnp },
7893 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM12), board_pch_spt },
7894 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V12), board_pch_spt },
7895 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM13), board_pch_tgp },
7896 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V13), board_pch_tgp },
7897 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM14), board_pch_tgp },
7898 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V14), board_pch_tgp },
7899 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM15), board_pch_tgp },
7900 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V15), board_pch_tgp },
7901 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_LM23), board_pch_tgp },
7902 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_V23), board_pch_tgp },
7903 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM16), board_pch_tgp },
7904 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V16), board_pch_tgp },
7905 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM17), board_pch_tgp },
7906 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V17), board_pch_tgp },
7907 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_LM22), board_pch_tgp },
7908 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_V22), board_pch_tgp },
7909 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_LM18), board_pch_tgp },
7910 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_V18), board_pch_tgp },
7911 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_LM19), board_pch_tgp },
7912 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_V19), board_pch_tgp },
7913 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_LM20), board_pch_tgp },
7914 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_V20), board_pch_tgp },
7915 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_LM21), board_pch_tgp },
7916 { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_V21), board_pch_tgp },
7918 { 0, 0, 0, 0, 0, 0, 0 } /* terminate list */
7920 MODULE_DEVICE_TABLE(pci, e1000_pci_tbl);
7922 static const struct dev_pm_ops e1000_pm_ops = {
7923 #ifdef CONFIG_PM_SLEEP
7924 .prepare = e1000e_pm_prepare,
7925 .suspend = e1000e_pm_suspend,
7926 .resume = e1000e_pm_resume,
7927 .freeze = e1000e_pm_freeze,
7928 .thaw = e1000e_pm_thaw,
7929 .poweroff = e1000e_pm_suspend,
7930 .restore = e1000e_pm_resume,
7932 SET_RUNTIME_PM_OPS(e1000e_pm_runtime_suspend, e1000e_pm_runtime_resume,
7933 e1000e_pm_runtime_idle)
7936 /* PCI Device API Driver */
7937 static struct pci_driver e1000_driver = {
7938 .name = e1000e_driver_name,
7939 .id_table = e1000_pci_tbl,
7940 .probe = e1000_probe,
7941 .remove = e1000_remove,
7943 .pm = &e1000_pm_ops,
7945 .shutdown = e1000_shutdown,
7946 .err_handler = &e1000_err_handler
7950 * e1000_init_module - Driver Registration Routine
7952 * e1000_init_module is the first routine called when the driver is
7953 * loaded. All it does is register with the PCI subsystem.
7955 static int __init e1000_init_module(void)
7957 pr_info("Intel(R) PRO/1000 Network Driver\n");
7958 pr_info("Copyright(c) 1999 - 2015 Intel Corporation.\n");
7960 return pci_register_driver(&e1000_driver);
7962 module_init(e1000_init_module);
7965 * e1000_exit_module - Driver Exit Cleanup Routine
7967 * e1000_exit_module is called just before the driver is removed
7970 static void __exit e1000_exit_module(void)
7972 pci_unregister_driver(&e1000_driver);
7974 module_exit(e1000_exit_module);
7976 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
7977 MODULE_DESCRIPTION("Intel(R) PRO/1000 Network Driver");
7978 MODULE_LICENSE("GPL v2");