1// SPDX-License-Identifier: GPL-2.0
2/* Copyright(c) 1999 - 2018 Intel Corporation. */
3
4#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
5
6#include <linux/module.h>
7#include <linux/types.h>
8#include <linux/init.h>
9#include <linux/pci.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/prefetch.h>
27#include <linux/suspend.h>
28
29#include "e1000.h"
30#define CREATE_TRACE_POINTS
31#include "e1000e_trace.h"
32
33char e1000e_driver_name[] = "e1000e";
34
35#define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
36static int debug = -1;
37module_param(debug, int, 0);
38MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
39
40static const struct e1000_info *e1000_info_tbl[] = {
41	[board_82571]		= &e1000_82571_info,
42	[board_82572]		= &e1000_82572_info,
43	[board_82573]		= &e1000_82573_info,
44	[board_82574]		= &e1000_82574_info,
45	[board_82583]		= &e1000_82583_info,
46	[board_80003es2lan]	= &e1000_es2_info,
47	[board_ich8lan]		= &e1000_ich8_info,
48	[board_ich9lan]		= &e1000_ich9_info,
49	[board_ich10lan]	= &e1000_ich10_info,
50	[board_pchlan]		= &e1000_pch_info,
51	[board_pch2lan]		= &e1000_pch2_info,
52	[board_pch_lpt]		= &e1000_pch_lpt_info,
53	[board_pch_spt]		= &e1000_pch_spt_info,
54	[board_pch_cnp]		= &e1000_pch_cnp_info,
55	[board_pch_tgp]		= &e1000_pch_tgp_info,
56	[board_pch_adp]		= &e1000_pch_adp_info,
57	[board_pch_mtp]		= &e1000_pch_mtp_info,
58};
59
60struct e1000_reg_info {
61	u32 ofs;
62	char *name;
63};
64
65static const struct e1000_reg_info e1000_reg_info_tbl[] = {
66	/* General Registers */
67	{E1000_CTRL, "CTRL"},
68	{E1000_STATUS, "STATUS"},
69	{E1000_CTRL_EXT, "CTRL_EXT"},
70
71	/* Interrupt Registers */
72	{E1000_ICR, "ICR"},
73
74	/* Rx Registers */
75	{E1000_RCTL, "RCTL"},
76	{E1000_RDLEN(0), "RDLEN"},
77	{E1000_RDH(0), "RDH"},
78	{E1000_RDT(0), "RDT"},
79	{E1000_RDTR, "RDTR"},
80	{E1000_RXDCTL(0), "RXDCTL"},
81	{E1000_ERT, "ERT"},
82	{E1000_RDBAL(0), "RDBAL"},
83	{E1000_RDBAH(0), "RDBAH"},
84	{E1000_RDFH, "RDFH"},
85	{E1000_RDFT, "RDFT"},
86	{E1000_RDFHS, "RDFHS"},
87	{E1000_RDFTS, "RDFTS"},
88	{E1000_RDFPC, "RDFPC"},
89
90	/* Tx Registers */
91	{E1000_TCTL, "TCTL"},
92	{E1000_TDBAL(0), "TDBAL"},
93	{E1000_TDBAH(0), "TDBAH"},
94	{E1000_TDLEN(0), "TDLEN"},
95	{E1000_TDH(0), "TDH"},
96	{E1000_TDT(0), "TDT"},
97	{E1000_TIDV, "TIDV"},
98	{E1000_TXDCTL(0), "TXDCTL"},
99	{E1000_TADV, "TADV"},
100	{E1000_TARC(0), "TARC"},
101	{E1000_TDFH, "TDFH"},
102	{E1000_TDFT, "TDFT"},
103	{E1000_TDFHS, "TDFHS"},
104	{E1000_TDFTS, "TDFTS"},
105	{E1000_TDFPC, "TDFPC"},
106
107	/* List Terminator */
108	{0, NULL}
109};
110
111/**
112 * __ew32_prepare - prepare to write to MAC CSR register on certain parts
113 * @hw: pointer to the HW structure
114 *
115 * When updating the MAC CSR registers, the Manageability Engine (ME) could
116 * be accessing the registers at the same time.  Normally, this is handled in
117 * h/w by an arbiter but on some parts there is a bug that acknowledges Host
118 * accesses later than it should which could result in the register to have
119 * an incorrect value.  Workaround this by checking the FWSM register which
120 * has bit 24 set while ME is accessing MAC CSR registers, wait if it is set
121 * and try again a number of times.
122 **/
123static void __ew32_prepare(struct e1000_hw *hw)
124{
125	s32 i = E1000_ICH_FWSM_PCIM2PCI_COUNT;
126
127	while ((er32(FWSM) & E1000_ICH_FWSM_PCIM2PCI) && --i)
128		udelay(50);
129}
130
131void __ew32(struct e1000_hw *hw, unsigned long reg, u32 val)
132{
133	if (hw->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
134		__ew32_prepare(hw);
135
136	writel(val, hw->hw_addr + reg);
137}
138
139/**
140 * e1000_regdump - register printout routine
141 * @hw: pointer to the HW structure
142 * @reginfo: pointer to the register info table
143 **/
144static void e1000_regdump(struct e1000_hw *hw, struct e1000_reg_info *reginfo)
145{
146	int n = 0;
147	char rname[16];
148	u32 regs[8];
149
150	switch (reginfo->ofs) {
151	case E1000_RXDCTL(0):
152		for (n = 0; n < 2; n++)
153			regs[n] = __er32(hw, E1000_RXDCTL(n));
154		break;
155	case E1000_TXDCTL(0):
156		for (n = 0; n < 2; n++)
157			regs[n] = __er32(hw, E1000_TXDCTL(n));
158		break;
159	case E1000_TARC(0):
160		for (n = 0; n < 2; n++)
161			regs[n] = __er32(hw, E1000_TARC(n));
162		break;
163	default:
164		pr_info("%-15s %08x\n",
165			reginfo->name, __er32(hw, reginfo->ofs));
166		return;
167	}
168
169	snprintf(rname, 16, "%s%s", reginfo->name, "[0-1]");
170	pr_info("%-15s %08x %08x\n", rname, regs[0], regs[1]);
171}
172
173static void e1000e_dump_ps_pages(struct e1000_adapter *adapter,
174				 struct e1000_buffer *bi)
175{
176	int i;
177	struct e1000_ps_page *ps_page;
178
179	for (i = 0; i < adapter->rx_ps_pages; i++) {
180		ps_page = &bi->ps_pages[i];
181
182		if (ps_page->page) {
183			pr_info("packet dump for ps_page %d:\n", i);
184			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
185				       16, 1, page_address(ps_page->page),
186				       PAGE_SIZE, true);
187		}
188	}
189}
190
191/**
192 * e1000e_dump - Print registers, Tx-ring and Rx-ring
193 * @adapter: board private structure
194 **/
195static void e1000e_dump(struct e1000_adapter *adapter)
196{
197	struct net_device *netdev = adapter->netdev;
198	struct e1000_hw *hw = &adapter->hw;
199	struct e1000_reg_info *reginfo;
200	struct e1000_ring *tx_ring = adapter->tx_ring;
201	struct e1000_tx_desc *tx_desc;
202	struct my_u0 {
203		__le64 a;
204		__le64 b;
205	} *u0;
206	struct e1000_buffer *buffer_info;
207	struct e1000_ring *rx_ring = adapter->rx_ring;
208	union e1000_rx_desc_packet_split *rx_desc_ps;
209	union e1000_rx_desc_extended *rx_desc;
210	struct my_u1 {
211		__le64 a;
212		__le64 b;
213		__le64 c;
214		__le64 d;
215	} *u1;
216	u32 staterr;
217	int i = 0;
218
219	if (!netif_msg_hw(adapter))
220		return;
221
222	/* Print netdevice Info */
223	if (netdev) {
224		dev_info(&adapter->pdev->dev, "Net device Info\n");
225		pr_info("Device Name     state            trans_start\n");
226		pr_info("%-15s %016lX %016lX\n", netdev->name,
227			netdev->state, dev_trans_start(netdev));
228	}
229
230	/* Print Registers */
231	dev_info(&adapter->pdev->dev, "Register Dump\n");
232	pr_info(" Register Name   Value\n");
233	for (reginfo = (struct e1000_reg_info *)e1000_reg_info_tbl;
234	     reginfo->name; reginfo++) {
235		e1000_regdump(hw, reginfo);
236	}
237
238	/* Print Tx Ring Summary */
239	if (!netdev || !netif_running(netdev))
240		return;
241
242	dev_info(&adapter->pdev->dev, "Tx Ring Summary\n");
243	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
244	buffer_info = &tx_ring->buffer_info[tx_ring->next_to_clean];
245	pr_info(" %5d %5X %5X %016llX %04X %3X %016llX\n",
246		0, tx_ring->next_to_use, tx_ring->next_to_clean,
247		(unsigned long long)buffer_info->dma,
248		buffer_info->length,
249		buffer_info->next_to_watch,
250		(unsigned long long)buffer_info->time_stamp);
251
252	/* Print Tx Ring */
253	if (!netif_msg_tx_done(adapter))
254		goto rx_ring_summary;
255
256	dev_info(&adapter->pdev->dev, "Tx Ring Dump\n");
257
258	/* Transmit Descriptor Formats - DEXT[29] is 0 (Legacy) or 1 (Extended)
259	 *
260	 * Legacy Transmit Descriptor
261	 *   +--------------------------------------------------------------+
262	 * 0 |         Buffer Address [63:0] (Reserved on Write Back)       |
263	 *   +--------------------------------------------------------------+
264	 * 8 | Special  |    CSS     | Status |  CMD    |  CSO   |  Length  |
265	 *   +--------------------------------------------------------------+
266	 *   63       48 47        36 35    32 31     24 23    16 15        0
267	 *
268	 * Extended Context Descriptor (DTYP=0x0) for TSO or checksum offload
269	 *   63      48 47    40 39       32 31             16 15    8 7      0
270	 *   +----------------------------------------------------------------+
271	 * 0 |  TUCSE  | TUCS0  |   TUCSS   |     IPCSE       | IPCS0 | IPCSS |
272	 *   +----------------------------------------------------------------+
273	 * 8 |   MSS   | HDRLEN | RSV | STA | TUCMD | DTYP |      PAYLEN      |
274	 *   +----------------------------------------------------------------+
275	 *   63      48 47    40 39 36 35 32 31   24 23  20 19                0
276	 *
277	 * Extended Data Descriptor (DTYP=0x1)
278	 *   +----------------------------------------------------------------+
279	 * 0 |                     Buffer Address [63:0]                      |
280	 *   +----------------------------------------------------------------+
281	 * 8 | VLAN tag |  POPTS  | Rsvd | Status | Command | DTYP |  DTALEN  |
282	 *   +----------------------------------------------------------------+
283	 *   63       48 47     40 39  36 35    32 31     24 23  20 19        0
284	 */
285	pr_info("Tl[desc]     [address 63:0  ] [SpeCssSCmCsLen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Legacy format\n");
286	pr_info("Tc[desc]     [Ce CoCsIpceCoS] [MssHlRSCm0Plen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Context format\n");
287	pr_info("Td[desc]     [address 63:0  ] [VlaPoRSCm1Dlen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Data format\n");
288	for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
289		const char *next_desc;
290		tx_desc = E1000_TX_DESC(*tx_ring, i);
291		buffer_info = &tx_ring->buffer_info[i];
292		u0 = (struct my_u0 *)tx_desc;
293		if (i == tx_ring->next_to_use && i == tx_ring->next_to_clean)
294			next_desc = " NTC/U";
295		else if (i == tx_ring->next_to_use)
296			next_desc = " NTU";
297		else if (i == tx_ring->next_to_clean)
298			next_desc = " NTC";
299		else
300			next_desc = "";
301		pr_info("T%c[0x%03X]    %016llX %016llX %016llX %04X  %3X %016llX %p%s\n",
302			(!(le64_to_cpu(u0->b) & BIT(29)) ? 'l' :
303			 ((le64_to_cpu(u0->b) & BIT(20)) ? 'd' : 'c')),
304			i,
305			(unsigned long long)le64_to_cpu(u0->a),
306			(unsigned long long)le64_to_cpu(u0->b),
307			(unsigned long long)buffer_info->dma,
308			buffer_info->length, buffer_info->next_to_watch,
309			(unsigned long long)buffer_info->time_stamp,
310			buffer_info->skb, next_desc);
311
312		if (netif_msg_pktdata(adapter) && buffer_info->skb)
313			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
314				       16, 1, buffer_info->skb->data,
315				       buffer_info->skb->len, true);
316	}
317
318	/* Print Rx Ring Summary */
319rx_ring_summary:
320	dev_info(&adapter->pdev->dev, "Rx Ring Summary\n");
321	pr_info("Queue [NTU] [NTC]\n");
322	pr_info(" %5d %5X %5X\n",
323		0, rx_ring->next_to_use, rx_ring->next_to_clean);
324
325	/* Print Rx Ring */
326	if (!netif_msg_rx_status(adapter))
327		return;
328
329	dev_info(&adapter->pdev->dev, "Rx Ring Dump\n");
330	switch (adapter->rx_ps_pages) {
331	case 1:
332	case 2:
333	case 3:
334		/* [Extended] Packet Split Receive Descriptor Format
335		 *
336		 *    +-----------------------------------------------------+
337		 *  0 |                Buffer Address 0 [63:0]              |
338		 *    +-----------------------------------------------------+
339		 *  8 |                Buffer Address 1 [63:0]              |
340		 *    +-----------------------------------------------------+
341		 * 16 |                Buffer Address 2 [63:0]              |
342		 *    +-----------------------------------------------------+
343		 * 24 |                Buffer Address 3 [63:0]              |
344		 *    +-----------------------------------------------------+
345		 */
346		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");
347		/* [Extended] Receive Descriptor (Write-Back) Format
348		 *
349		 *   63       48 47    32 31     13 12    8 7    4 3        0
350		 *   +------------------------------------------------------+
351		 * 0 | Packet   | IP     |  Rsvd   | MRQ   | Rsvd | MRQ RSS |
352		 *   | Checksum | Ident  |         | Queue |      |  Type   |
353		 *   +------------------------------------------------------+
354		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
355		 *   +------------------------------------------------------+
356		 *   63       48 47    32 31            20 19               0
357		 */
358		pr_info("RWB[desc]      [ck ipid mrqhsh] [vl   l0 ee  es] [ l3  l2  l1 hs] [reserved      ] ---------------- [bi->skb] <-- Ext Rx Write-Back format\n");
359		for (i = 0; i < rx_ring->count; i++) {
360			const char *next_desc;
361			buffer_info = &rx_ring->buffer_info[i];
362			rx_desc_ps = E1000_RX_DESC_PS(*rx_ring, i);
363			u1 = (struct my_u1 *)rx_desc_ps;
364			staterr =
365			    le32_to_cpu(rx_desc_ps->wb.middle.status_error);
366
367			if (i == rx_ring->next_to_use)
368				next_desc = " NTU";
369			else if (i == rx_ring->next_to_clean)
370				next_desc = " NTC";
371			else
372				next_desc = "";
373
374			if (staterr & E1000_RXD_STAT_DD) {
375				/* Descriptor Done */
376				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX ---------------- %p%s\n",
377					"RWB", i,
378					(unsigned long long)le64_to_cpu(u1->a),
379					(unsigned long long)le64_to_cpu(u1->b),
380					(unsigned long long)le64_to_cpu(u1->c),
381					(unsigned long long)le64_to_cpu(u1->d),
382					buffer_info->skb, next_desc);
383			} else {
384				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX %016llX %p%s\n",
385					"R  ", i,
386					(unsigned long long)le64_to_cpu(u1->a),
387					(unsigned long long)le64_to_cpu(u1->b),
388					(unsigned long long)le64_to_cpu(u1->c),
389					(unsigned long long)le64_to_cpu(u1->d),
390					(unsigned long long)buffer_info->dma,
391					buffer_info->skb, next_desc);
392
393				if (netif_msg_pktdata(adapter))
394					e1000e_dump_ps_pages(adapter,
395							     buffer_info);
396			}
397		}
398		break;
399	default:
400	case 0:
401		/* Extended Receive Descriptor (Read) Format
402		 *
403		 *   +-----------------------------------------------------+
404		 * 0 |                Buffer Address [63:0]                |
405		 *   +-----------------------------------------------------+
406		 * 8 |                      Reserved                       |
407		 *   +-----------------------------------------------------+
408		 */
409		pr_info("R  [desc]      [buf addr 63:0 ] [reserved 63:0 ] [bi->dma       ] [bi->skb] <-- Ext (Read) format\n");
410		/* Extended Receive Descriptor (Write-Back) Format
411		 *
412		 *   63       48 47    32 31    24 23            4 3        0
413		 *   +------------------------------------------------------+
414		 *   |     RSS Hash      |        |               |         |
415		 * 0 +-------------------+  Rsvd  |   Reserved    | MRQ RSS |
416		 *   | Packet   | IP     |        |               |  Type   |
417		 *   | Checksum | Ident  |        |               |         |
418		 *   +------------------------------------------------------+
419		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
420		 *   +------------------------------------------------------+
421		 *   63       48 47    32 31            20 19               0
422		 */
423		pr_info("RWB[desc]      [cs ipid    mrq] [vt   ln xe  xs] [bi->skb] <-- Ext (Write-Back) format\n");
424
425		for (i = 0; i < rx_ring->count; i++) {
426			const char *next_desc;
427
428			buffer_info = &rx_ring->buffer_info[i];
429			rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
430			u1 = (struct my_u1 *)rx_desc;
431			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
432
433			if (i == rx_ring->next_to_use)
434				next_desc = " NTU";
435			else if (i == rx_ring->next_to_clean)
436				next_desc = " NTC";
437			else
438				next_desc = "";
439
440			if (staterr & E1000_RXD_STAT_DD) {
441				/* Descriptor Done */
442				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %p%s\n",
443					"RWB", i,
444					(unsigned long long)le64_to_cpu(u1->a),
445					(unsigned long long)le64_to_cpu(u1->b),
446					buffer_info->skb, next_desc);
447			} else {
448				pr_info("%s[0x%03X]     %016llX %016llX %016llX %p%s\n",
449					"R  ", i,
450					(unsigned long long)le64_to_cpu(u1->a),
451					(unsigned long long)le64_to_cpu(u1->b),
452					(unsigned long long)buffer_info->dma,
453					buffer_info->skb, next_desc);
454
455				if (netif_msg_pktdata(adapter) &&
456				    buffer_info->skb)
457					print_hex_dump(KERN_INFO, "",
458						       DUMP_PREFIX_ADDRESS, 16,
459						       1,
460						       buffer_info->skb->data,
461						       adapter->rx_buffer_len,
462						       true);
463			}
464		}
465	}
466}
467
468/**
469 * e1000_desc_unused - calculate if we have unused descriptors
470 * @ring: pointer to ring struct to perform calculation on
471 **/
472static int e1000_desc_unused(struct e1000_ring *ring)
473{
474	if (ring->next_to_clean > ring->next_to_use)
475		return ring->next_to_clean - ring->next_to_use - 1;
476
477	return ring->count + ring->next_to_clean - ring->next_to_use - 1;
478}
479
480/**
481 * e1000e_systim_to_hwtstamp - convert system time value to hw time stamp
482 * @adapter: board private structure
483 * @hwtstamps: time stamp structure to update
484 * @systim: unsigned 64bit system time value.
485 *
486 * Convert the system time value stored in the RX/TXSTMP registers into a
487 * hwtstamp which can be used by the upper level time stamping functions.
488 *
489 * The 'systim_lock' spinlock is used to protect the consistency of the
490 * system time value. This is needed because reading the 64 bit time
491 * value involves reading two 32 bit registers. The first read latches the
492 * value.
493 **/
494static void e1000e_systim_to_hwtstamp(struct e1000_adapter *adapter,
495				      struct skb_shared_hwtstamps *hwtstamps,
496				      u64 systim)
497{
498	u64 ns;
499	unsigned long flags;
500
501	spin_lock_irqsave(&adapter->systim_lock, flags);
502	ns = timecounter_cyc2time(&adapter->tc, systim);
503	spin_unlock_irqrestore(&adapter->systim_lock, flags);
504
505	memset(hwtstamps, 0, sizeof(*hwtstamps));
506	hwtstamps->hwtstamp = ns_to_ktime(ns);
507}
508
509/**
510 * e1000e_rx_hwtstamp - utility function which checks for Rx time stamp
511 * @adapter: board private structure
512 * @status: descriptor extended error and status field
513 * @skb: particular skb to include time stamp
514 *
515 * If the time stamp is valid, convert it into the timecounter ns value
516 * and store that result into the shhwtstamps structure which is passed
517 * up the network stack.
518 **/
519static void e1000e_rx_hwtstamp(struct e1000_adapter *adapter, u32 status,
520			       struct sk_buff *skb)
521{
522	struct e1000_hw *hw = &adapter->hw;
523	u64 rxstmp;
524
525	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP) ||
526	    !(status & E1000_RXDEXT_STATERR_TST) ||
527	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID))
528		return;
529
530	/* The Rx time stamp registers contain the time stamp.  No other
531	 * received packet will be time stamped until the Rx time stamp
532	 * registers are read.  Because only one packet can be time stamped
533	 * at a time, the register values must belong to this packet and
534	 * therefore none of the other additional attributes need to be
535	 * compared.
536	 */
537	rxstmp = (u64)er32(RXSTMPL);
538	rxstmp |= (u64)er32(RXSTMPH) << 32;
539	e1000e_systim_to_hwtstamp(adapter, skb_hwtstamps(skb), rxstmp);
540
541	adapter->flags2 &= ~FLAG2_CHECK_RX_HWTSTAMP;
542}
543
544/**
545 * e1000_receive_skb - helper function to handle Rx indications
546 * @adapter: board private structure
547 * @netdev: pointer to netdev struct
548 * @staterr: descriptor extended error and status field as written by hardware
549 * @vlan: descriptor vlan field as written by hardware (no le/be conversion)
550 * @skb: pointer to sk_buff to be indicated to stack
551 **/
552static void e1000_receive_skb(struct e1000_adapter *adapter,
553			      struct net_device *netdev, struct sk_buff *skb,
554			      u32 staterr, __le16 vlan)
555{
556	u16 tag = le16_to_cpu(vlan);
557
558	e1000e_rx_hwtstamp(adapter, staterr, skb);
559
560	skb->protocol = eth_type_trans(skb, netdev);
561
562	if (staterr & E1000_RXD_STAT_VP)
563		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), tag);
564
565	napi_gro_receive(&adapter->napi, skb);
566}
567
568/**
569 * e1000_rx_checksum - Receive Checksum Offload
570 * @adapter: board private structure
571 * @status_err: receive descriptor status and error fields
572 * @skb: socket buffer with received data
573 **/
574static void e1000_rx_checksum(struct e1000_adapter *adapter, u32 status_err,
575			      struct sk_buff *skb)
576{
577	u16 status = (u16)status_err;
578	u8 errors = (u8)(status_err >> 24);
579
580	skb_checksum_none_assert(skb);
581
582	/* Rx checksum disabled */
583	if (!(adapter->netdev->features & NETIF_F_RXCSUM))
584		return;
585
586	/* Ignore Checksum bit is set */
587	if (status & E1000_RXD_STAT_IXSM)
588		return;
589
590	/* TCP/UDP checksum error bit or IP checksum error bit is set */
591	if (errors & (E1000_RXD_ERR_TCPE | E1000_RXD_ERR_IPE)) {
592		/* let the stack verify checksum errors */
593		adapter->hw_csum_err++;
594		return;
595	}
596
597	/* TCP/UDP Checksum has not been calculated */
598	if (!(status & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)))
599		return;
600
601	/* It must be a TCP or UDP packet with a valid checksum */
602	skb->ip_summed = CHECKSUM_UNNECESSARY;
603	adapter->hw_csum_good++;
604}
605
606static void e1000e_update_rdt_wa(struct e1000_ring *rx_ring, unsigned int i)
607{
608	struct e1000_adapter *adapter = rx_ring->adapter;
609	struct e1000_hw *hw = &adapter->hw;
610
611	__ew32_prepare(hw);
612	writel(i, rx_ring->tail);
613
614	if (unlikely(i != readl(rx_ring->tail))) {
615		u32 rctl = er32(RCTL);
616
617		ew32(RCTL, rctl & ~E1000_RCTL_EN);
618		e_err("ME firmware caused invalid RDT - resetting\n");
619		schedule_work(&adapter->reset_task);
620	}
621}
622
623static void e1000e_update_tdt_wa(struct e1000_ring *tx_ring, unsigned int i)
624{
625	struct e1000_adapter *adapter = tx_ring->adapter;
626	struct e1000_hw *hw = &adapter->hw;
627
628	__ew32_prepare(hw);
629	writel(i, tx_ring->tail);
630
631	if (unlikely(i != readl(tx_ring->tail))) {
632		u32 tctl = er32(TCTL);
633
634		ew32(TCTL, tctl & ~E1000_TCTL_EN);
635		e_err("ME firmware caused invalid TDT - resetting\n");
636		schedule_work(&adapter->reset_task);
637	}
638}
639
640/**
641 * e1000_alloc_rx_buffers - Replace used receive buffers
642 * @rx_ring: Rx descriptor ring
643 * @cleaned_count: number to reallocate
644 * @gfp: flags for allocation
645 **/
646static void e1000_alloc_rx_buffers(struct e1000_ring *rx_ring,
647				   int cleaned_count, gfp_t gfp)
648{
649	struct e1000_adapter *adapter = rx_ring->adapter;
650	struct net_device *netdev = adapter->netdev;
651	struct pci_dev *pdev = adapter->pdev;
652	union e1000_rx_desc_extended *rx_desc;
653	struct e1000_buffer *buffer_info;
654	struct sk_buff *skb;
655	unsigned int i;
656	unsigned int bufsz = adapter->rx_buffer_len;
657
658	i = rx_ring->next_to_use;
659	buffer_info = &rx_ring->buffer_info[i];
660
661	while (cleaned_count--) {
662		skb = buffer_info->skb;
663		if (skb) {
664			skb_trim(skb, 0);
665			goto map_skb;
666		}
667
668		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
669		if (!skb) {
670			/* Better luck next round */
671			adapter->alloc_rx_buff_failed++;
672			break;
673		}
674
675		buffer_info->skb = skb;
676map_skb:
677		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
678						  adapter->rx_buffer_len,
679						  DMA_FROM_DEVICE);
680		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
681			dev_err(&pdev->dev, "Rx DMA map failed\n");
682			adapter->rx_dma_failed++;
683			break;
684		}
685
686		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
687		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
688
689		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
690			/* Force memory writes to complete before letting h/w
691			 * know there are new descriptors to fetch.  (Only
692			 * applicable for weak-ordered memory model archs,
693			 * such as IA-64).
694			 */
695			wmb();
696			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
697				e1000e_update_rdt_wa(rx_ring, i);
698			else
699				writel(i, rx_ring->tail);
700		}
701		i++;
702		if (i == rx_ring->count)
703			i = 0;
704		buffer_info = &rx_ring->buffer_info[i];
705	}
706
707	rx_ring->next_to_use = i;
708}
709
710/**
711 * e1000_alloc_rx_buffers_ps - Replace used receive buffers; packet split
712 * @rx_ring: Rx descriptor ring
713 * @cleaned_count: number to reallocate
714 * @gfp: flags for allocation
715 **/
716static void e1000_alloc_rx_buffers_ps(struct e1000_ring *rx_ring,
717				      int cleaned_count, gfp_t gfp)
718{
719	struct e1000_adapter *adapter = rx_ring->adapter;
720	struct net_device *netdev = adapter->netdev;
721	struct pci_dev *pdev = adapter->pdev;
722	union e1000_rx_desc_packet_split *rx_desc;
723	struct e1000_buffer *buffer_info;
724	struct e1000_ps_page *ps_page;
725	struct sk_buff *skb;
726	unsigned int i, j;
727
728	i = rx_ring->next_to_use;
729	buffer_info = &rx_ring->buffer_info[i];
730
731	while (cleaned_count--) {
732		rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
733
734		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
735			ps_page = &buffer_info->ps_pages[j];
736			if (j >= adapter->rx_ps_pages) {
737				/* all unused desc entries get hw null ptr */
738				rx_desc->read.buffer_addr[j + 1] =
739				    ~cpu_to_le64(0);
740				continue;
741			}
742			if (!ps_page->page) {
743				ps_page->page = alloc_page(gfp);
744				if (!ps_page->page) {
745					adapter->alloc_rx_buff_failed++;
746					goto no_buffers;
747				}
748				ps_page->dma = dma_map_page(&pdev->dev,
749							    ps_page->page,
750							    0, PAGE_SIZE,
751							    DMA_FROM_DEVICE);
752				if (dma_mapping_error(&pdev->dev,
753						      ps_page->dma)) {
754					dev_err(&adapter->pdev->dev,
755						"Rx DMA page map failed\n");
756					adapter->rx_dma_failed++;
757					goto no_buffers;
758				}
759			}
760			/* Refresh the desc even if buffer_addrs
761			 * didn't change because each write-back
762			 * erases this info.
763			 */
764			rx_desc->read.buffer_addr[j + 1] =
765			    cpu_to_le64(ps_page->dma);
766		}
767
768		skb = __netdev_alloc_skb_ip_align(netdev, adapter->rx_ps_bsize0,
769						  gfp);
770
771		if (!skb) {
772			adapter->alloc_rx_buff_failed++;
773			break;
774		}
775
776		buffer_info->skb = skb;
777		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
778						  adapter->rx_ps_bsize0,
779						  DMA_FROM_DEVICE);
780		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
781			dev_err(&pdev->dev, "Rx DMA map failed\n");
782			adapter->rx_dma_failed++;
783			/* cleanup skb */
784			dev_kfree_skb_any(skb);
785			buffer_info->skb = NULL;
786			break;
787		}
788
789		rx_desc->read.buffer_addr[0] = cpu_to_le64(buffer_info->dma);
790
791		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
792			/* Force memory writes to complete before letting h/w
793			 * know there are new descriptors to fetch.  (Only
794			 * applicable for weak-ordered memory model archs,
795			 * such as IA-64).
796			 */
797			wmb();
798			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
799				e1000e_update_rdt_wa(rx_ring, i << 1);
800			else
801				writel(i << 1, rx_ring->tail);
802		}
803
804		i++;
805		if (i == rx_ring->count)
806			i = 0;
807		buffer_info = &rx_ring->buffer_info[i];
808	}
809
810no_buffers:
811	rx_ring->next_to_use = i;
812}
813
814/**
815 * e1000_alloc_jumbo_rx_buffers - Replace used jumbo receive buffers
816 * @rx_ring: Rx descriptor ring
817 * @cleaned_count: number of buffers to allocate this pass
818 * @gfp: flags for allocation
819 **/
820
821static void e1000_alloc_jumbo_rx_buffers(struct e1000_ring *rx_ring,
822					 int cleaned_count, gfp_t gfp)
823{
824	struct e1000_adapter *adapter = rx_ring->adapter;
825	struct net_device *netdev = adapter->netdev;
826	struct pci_dev *pdev = adapter->pdev;
827	union e1000_rx_desc_extended *rx_desc;
828	struct e1000_buffer *buffer_info;
829	struct sk_buff *skb;
830	unsigned int i;
831	unsigned int bufsz = 256 - 16;	/* for skb_reserve */
832
833	i = rx_ring->next_to_use;
834	buffer_info = &rx_ring->buffer_info[i];
835
836	while (cleaned_count--) {
837		skb = buffer_info->skb;
838		if (skb) {
839			skb_trim(skb, 0);
840			goto check_page;
841		}
842
843		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
844		if (unlikely(!skb)) {
845			/* Better luck next round */
846			adapter->alloc_rx_buff_failed++;
847			break;
848		}
849
850		buffer_info->skb = skb;
851check_page:
852		/* allocate a new page if necessary */
853		if (!buffer_info->page) {
854			buffer_info->page = alloc_page(gfp);
855			if (unlikely(!buffer_info->page)) {
856				adapter->alloc_rx_buff_failed++;
857				break;
858			}
859		}
860
861		if (!buffer_info->dma) {
862			buffer_info->dma = dma_map_page(&pdev->dev,
863							buffer_info->page, 0,
864							PAGE_SIZE,
865							DMA_FROM_DEVICE);
866			if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
867				adapter->alloc_rx_buff_failed++;
868				break;
869			}
870		}
871
872		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
873		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
874
875		if (unlikely(++i == rx_ring->count))
876			i = 0;
877		buffer_info = &rx_ring->buffer_info[i];
878	}
879
880	if (likely(rx_ring->next_to_use != i)) {
881		rx_ring->next_to_use = i;
882		if (unlikely(i-- == 0))
883			i = (rx_ring->count - 1);
884
885		/* Force memory writes to complete before letting h/w
886		 * know there are new descriptors to fetch.  (Only
887		 * applicable for weak-ordered memory model archs,
888		 * such as IA-64).
889		 */
890		wmb();
891		if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
892			e1000e_update_rdt_wa(rx_ring, i);
893		else
894			writel(i, rx_ring->tail);
895	}
896}
897
898static inline void e1000_rx_hash(struct net_device *netdev, __le32 rss,
899				 struct sk_buff *skb)
900{
901	if (netdev->features & NETIF_F_RXHASH)
902		skb_set_hash(skb, le32_to_cpu(rss), PKT_HASH_TYPE_L3);
903}
904
905/**
906 * e1000_clean_rx_irq - Send received data up the network stack
907 * @rx_ring: Rx descriptor ring
908 * @work_done: output parameter for indicating completed work
909 * @work_to_do: how many packets we can clean
910 *
911 * the return value indicates whether actual cleaning was done, there
912 * is no guarantee that everything was cleaned
913 **/
914static bool e1000_clean_rx_irq(struct e1000_ring *rx_ring, int *work_done,
915			       int work_to_do)
916{
917	struct e1000_adapter *adapter = rx_ring->adapter;
918	struct net_device *netdev = adapter->netdev;
919	struct pci_dev *pdev = adapter->pdev;
920	struct e1000_hw *hw = &adapter->hw;
921	union e1000_rx_desc_extended *rx_desc, *next_rxd;
922	struct e1000_buffer *buffer_info, *next_buffer;
923	u32 length, staterr;
924	unsigned int i;
925	int cleaned_count = 0;
926	bool cleaned = false;
927	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
928
929	i = rx_ring->next_to_clean;
930	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
931	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
932	buffer_info = &rx_ring->buffer_info[i];
933
934	while (staterr & E1000_RXD_STAT_DD) {
935		struct sk_buff *skb;
936
937		if (*work_done >= work_to_do)
938			break;
939		(*work_done)++;
940		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
941
942		skb = buffer_info->skb;
943		buffer_info->skb = NULL;
944
945		prefetch(skb->data - NET_IP_ALIGN);
946
947		i++;
948		if (i == rx_ring->count)
949			i = 0;
950		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
951		prefetch(next_rxd);
952
953		next_buffer = &rx_ring->buffer_info[i];
954
955		cleaned = true;
956		cleaned_count++;
957		dma_unmap_single(&pdev->dev, buffer_info->dma,
958				 adapter->rx_buffer_len, DMA_FROM_DEVICE);
959		buffer_info->dma = 0;
960
961		length = le16_to_cpu(rx_desc->wb.upper.length);
962
963		/* !EOP means multiple descriptors were used to store a single
964		 * packet, if that's the case we need to toss it.  In fact, we
965		 * need to toss every packet with the EOP bit clear and the
966		 * next frame that _does_ have the EOP bit set, as it is by
967		 * definition only a frame fragment
968		 */
969		if (unlikely(!(staterr & E1000_RXD_STAT_EOP)))
970			adapter->flags2 |= FLAG2_IS_DISCARDING;
971
972		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
973			/* All receives must fit into a single buffer */
974			e_dbg("Receive packet consumed multiple buffers\n");
975			/* recycle */
976			buffer_info->skb = skb;
977			if (staterr & E1000_RXD_STAT_EOP)
978				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
979			goto next_desc;
980		}
981
982		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
983			     !(netdev->features & NETIF_F_RXALL))) {
984			/* recycle */
985			buffer_info->skb = skb;
986			goto next_desc;
987		}
988
989		/* adjust length to remove Ethernet CRC */
990		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
991			/* If configured to store CRC, don't subtract FCS,
992			 * but keep the FCS bytes out of the total_rx_bytes
993			 * counter
994			 */
995			if (netdev->features & NETIF_F_RXFCS)
996				total_rx_bytes -= 4;
997			else
998				length -= 4;
999		}
1000
1001		total_rx_bytes += length;
1002		total_rx_packets++;
1003
1004		/* code added for copybreak, this should improve
1005		 * performance for small packets with large amounts
1006		 * of reassembly being done in the stack
1007		 */
1008		if (length < copybreak) {
1009			struct sk_buff *new_skb =
1010				napi_alloc_skb(&adapter->napi, length);
1011			if (new_skb) {
1012				skb_copy_to_linear_data_offset(new_skb,
1013							       -NET_IP_ALIGN,
1014							       (skb->data -
1015								NET_IP_ALIGN),
1016							       (length +
1017								NET_IP_ALIGN));
1018				/* save the skb in buffer_info as good */
1019				buffer_info->skb = skb;
1020				skb = new_skb;
1021			}
1022			/* else just continue with the old one */
1023		}
1024		/* end copybreak code */
1025		skb_put(skb, length);
1026
1027		/* Receive Checksum Offload */
1028		e1000_rx_checksum(adapter, staterr, skb);
1029
1030		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1031
1032		e1000_receive_skb(adapter, netdev, skb, staterr,
1033				  rx_desc->wb.upper.vlan);
1034
1035next_desc:
1036		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1037
1038		/* return some buffers to hardware, one at a time is too slow */
1039		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1040			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1041					      GFP_ATOMIC);
1042			cleaned_count = 0;
1043		}
1044
1045		/* use prefetched values */
1046		rx_desc = next_rxd;
1047		buffer_info = next_buffer;
1048
1049		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1050	}
1051	rx_ring->next_to_clean = i;
1052
1053	cleaned_count = e1000_desc_unused(rx_ring);
1054	if (cleaned_count)
1055		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1056
1057	adapter->total_rx_bytes += total_rx_bytes;
1058	adapter->total_rx_packets += total_rx_packets;
1059	return cleaned;
1060}
1061
1062static void e1000_put_txbuf(struct e1000_ring *tx_ring,
1063			    struct e1000_buffer *buffer_info,
1064			    bool drop)
1065{
1066	struct e1000_adapter *adapter = tx_ring->adapter;
1067
1068	if (buffer_info->dma) {
1069		if (buffer_info->mapped_as_page)
1070			dma_unmap_page(&adapter->pdev->dev, buffer_info->dma,
1071				       buffer_info->length, DMA_TO_DEVICE);
1072		else
1073			dma_unmap_single(&adapter->pdev->dev, buffer_info->dma,
1074					 buffer_info->length, DMA_TO_DEVICE);
1075		buffer_info->dma = 0;
1076	}
1077	if (buffer_info->skb) {
1078		if (drop)
1079			dev_kfree_skb_any(buffer_info->skb);
1080		else
1081			dev_consume_skb_any(buffer_info->skb);
1082		buffer_info->skb = NULL;
1083	}
1084	buffer_info->time_stamp = 0;
1085}
1086
1087static void e1000_print_hw_hang(struct work_struct *work)
1088{
1089	struct e1000_adapter *adapter = container_of(work,
1090						     struct e1000_adapter,
1091						     print_hang_task);
1092	struct net_device *netdev = adapter->netdev;
1093	struct e1000_ring *tx_ring = adapter->tx_ring;
1094	unsigned int i = tx_ring->next_to_clean;
1095	unsigned int eop = tx_ring->buffer_info[i].next_to_watch;
1096	struct e1000_tx_desc *eop_desc = E1000_TX_DESC(*tx_ring, eop);
1097	struct e1000_hw *hw = &adapter->hw;
1098	u16 phy_status, phy_1000t_status, phy_ext_status;
1099	u16 pci_status;
1100
1101	if (test_bit(__E1000_DOWN, &adapter->state))
1102		return;
1103
1104	if (!adapter->tx_hang_recheck && (adapter->flags2 & FLAG2_DMA_BURST)) {
1105		/* May be block on write-back, flush and detect again
1106		 * flush pending descriptor writebacks to memory
1107		 */
1108		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1109		/* execute the writes immediately */
1110		e1e_flush();
1111		/* Due to rare timing issues, write to TIDV again to ensure
1112		 * the write is successful
1113		 */
1114		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1115		/* execute the writes immediately */
1116		e1e_flush();
1117		adapter->tx_hang_recheck = true;
1118		return;
1119	}
1120	adapter->tx_hang_recheck = false;
1121
1122	if (er32(TDH(0)) == er32(TDT(0))) {
1123		e_dbg("false hang detected, ignoring\n");
1124		return;
1125	}
1126
1127	/* Real hang detected */
1128	netif_stop_queue(netdev);
1129
1130	e1e_rphy(hw, MII_BMSR, &phy_status);
1131	e1e_rphy(hw, MII_STAT1000, &phy_1000t_status);
1132	e1e_rphy(hw, MII_ESTATUS, &phy_ext_status);
1133
1134	pci_read_config_word(adapter->pdev, PCI_STATUS, &pci_status);
1135
1136	/* detected Hardware unit hang */
1137	e_err("Detected Hardware Unit Hang:\n"
1138	      "  TDH                  <%x>\n"
1139	      "  TDT                  <%x>\n"
1140	      "  next_to_use          <%x>\n"
1141	      "  next_to_clean        <%x>\n"
1142	      "buffer_info[next_to_clean]:\n"
1143	      "  time_stamp           <%lx>\n"
1144	      "  next_to_watch        <%x>\n"
1145	      "  jiffies              <%lx>\n"
1146	      "  next_to_watch.status <%x>\n"
1147	      "MAC Status             <%x>\n"
1148	      "PHY Status             <%x>\n"
1149	      "PHY 1000BASE-T Status  <%x>\n"
1150	      "PHY Extended Status    <%x>\n"
1151	      "PCI Status             <%x>\n",
1152	      readl(tx_ring->head), readl(tx_ring->tail), tx_ring->next_to_use,
1153	      tx_ring->next_to_clean, tx_ring->buffer_info[eop].time_stamp,
1154	      eop, jiffies, eop_desc->upper.fields.status, er32(STATUS),
1155	      phy_status, phy_1000t_status, phy_ext_status, pci_status);
1156
1157	e1000e_dump(adapter);
1158
1159	/* Suggest workaround for known h/w issue */
1160	if ((hw->mac.type == e1000_pchlan) && (er32(CTRL) & E1000_CTRL_TFCE))
1161		e_err("Try turning off Tx pause (flow control) via ethtool\n");
1162}
1163
1164/**
1165 * e1000e_tx_hwtstamp_work - check for Tx time stamp
1166 * @work: pointer to work struct
1167 *
1168 * This work function polls the TSYNCTXCTL valid bit to determine when a
1169 * timestamp has been taken for the current stored skb.  The timestamp must
1170 * be for this skb because only one such packet is allowed in the queue.
1171 */
1172static void e1000e_tx_hwtstamp_work(struct work_struct *work)
1173{
1174	struct e1000_adapter *adapter = container_of(work, struct e1000_adapter,
1175						     tx_hwtstamp_work);
1176	struct e1000_hw *hw = &adapter->hw;
1177
1178	if (er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID) {
1179		struct sk_buff *skb = adapter->tx_hwtstamp_skb;
1180		struct skb_shared_hwtstamps shhwtstamps;
1181		u64 txstmp;
1182
1183		txstmp = er32(TXSTMPL);
1184		txstmp |= (u64)er32(TXSTMPH) << 32;
1185
1186		e1000e_systim_to_hwtstamp(adapter, &shhwtstamps, txstmp);
1187
1188		/* Clear the global tx_hwtstamp_skb pointer and force writes
1189		 * prior to notifying the stack of a Tx timestamp.
1190		 */
1191		adapter->tx_hwtstamp_skb = NULL;
1192		wmb(); /* force write prior to skb_tstamp_tx */
1193
1194		skb_tstamp_tx(skb, &shhwtstamps);
1195		dev_consume_skb_any(skb);
1196	} else if (time_after(jiffies, adapter->tx_hwtstamp_start
1197			      + adapter->tx_timeout_factor * HZ)) {
1198		dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
1199		adapter->tx_hwtstamp_skb = NULL;
1200		adapter->tx_hwtstamp_timeouts++;
1201		e_warn("clearing Tx timestamp hang\n");
1202	} else {
1203		/* reschedule to check later */
1204		schedule_work(&adapter->tx_hwtstamp_work);
1205	}
1206}
1207
1208/**
1209 * e1000_clean_tx_irq - Reclaim resources after transmit completes
1210 * @tx_ring: Tx descriptor ring
1211 *
1212 * the return value indicates whether actual cleaning was done, there
1213 * is no guarantee that everything was cleaned
1214 **/
1215static bool e1000_clean_tx_irq(struct e1000_ring *tx_ring)
1216{
1217	struct e1000_adapter *adapter = tx_ring->adapter;
1218	struct net_device *netdev = adapter->netdev;
1219	struct e1000_hw *hw = &adapter->hw;
1220	struct e1000_tx_desc *tx_desc, *eop_desc;
1221	struct e1000_buffer *buffer_info;
1222	unsigned int i, eop;
1223	unsigned int count = 0;
1224	unsigned int total_tx_bytes = 0, total_tx_packets = 0;
1225	unsigned int bytes_compl = 0, pkts_compl = 0;
1226
1227	i = tx_ring->next_to_clean;
1228	eop = tx_ring->buffer_info[i].next_to_watch;
1229	eop_desc = E1000_TX_DESC(*tx_ring, eop);
1230
1231	while ((eop_desc->upper.data & cpu_to_le32(E1000_TXD_STAT_DD)) &&
1232	       (count < tx_ring->count)) {
1233		bool cleaned = false;
1234
1235		dma_rmb();		/* read buffer_info after eop_desc */
1236		for (; !cleaned; count++) {
1237			tx_desc = E1000_TX_DESC(*tx_ring, i);
1238			buffer_info = &tx_ring->buffer_info[i];
1239			cleaned = (i == eop);
1240
1241			if (cleaned) {
1242				total_tx_packets += buffer_info->segs;
1243				total_tx_bytes += buffer_info->bytecount;
1244				if (buffer_info->skb) {
1245					bytes_compl += buffer_info->skb->len;
1246					pkts_compl++;
1247				}
1248			}
1249
1250			e1000_put_txbuf(tx_ring, buffer_info, false);
1251			tx_desc->upper.data = 0;
1252
1253			i++;
1254			if (i == tx_ring->count)
1255				i = 0;
1256		}
1257
1258		if (i == tx_ring->next_to_use)
1259			break;
1260		eop = tx_ring->buffer_info[i].next_to_watch;
1261		eop_desc = E1000_TX_DESC(*tx_ring, eop);
1262	}
1263
1264	tx_ring->next_to_clean = i;
1265
1266	netdev_completed_queue(netdev, pkts_compl, bytes_compl);
1267
1268#define TX_WAKE_THRESHOLD 32
1269	if (count && netif_carrier_ok(netdev) &&
1270	    e1000_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD) {
1271		/* Make sure that anybody stopping the queue after this
1272		 * sees the new next_to_clean.
1273		 */
1274		smp_mb();
1275
1276		if (netif_queue_stopped(netdev) &&
1277		    !(test_bit(__E1000_DOWN, &adapter->state))) {
1278			netif_wake_queue(netdev);
1279			++adapter->restart_queue;
1280		}
1281	}
1282
1283	if (adapter->detect_tx_hung) {
1284		/* Detect a transmit hang in hardware, this serializes the
1285		 * check with the clearing of time_stamp and movement of i
1286		 */
1287		adapter->detect_tx_hung = false;
1288		if (tx_ring->buffer_info[i].time_stamp &&
1289		    time_after(jiffies, tx_ring->buffer_info[i].time_stamp
1290			       + (adapter->tx_timeout_factor * HZ)) &&
1291		    !(er32(STATUS) & E1000_STATUS_TXOFF))
1292			schedule_work(&adapter->print_hang_task);
1293		else
1294			adapter->tx_hang_recheck = false;
1295	}
1296	adapter->total_tx_bytes += total_tx_bytes;
1297	adapter->total_tx_packets += total_tx_packets;
1298	return count < tx_ring->count;
1299}
1300
1301/**
1302 * e1000_clean_rx_irq_ps - Send received data up the network stack; packet split
1303 * @rx_ring: Rx descriptor ring
1304 * @work_done: output parameter for indicating completed work
1305 * @work_to_do: how many packets we can clean
1306 *
1307 * the return value indicates whether actual cleaning was done, there
1308 * is no guarantee that everything was cleaned
1309 **/
1310static bool e1000_clean_rx_irq_ps(struct e1000_ring *rx_ring, int *work_done,
1311				  int work_to_do)
1312{
1313	struct e1000_adapter *adapter = rx_ring->adapter;
1314	struct e1000_hw *hw = &adapter->hw;
1315	union e1000_rx_desc_packet_split *rx_desc, *next_rxd;
1316	struct net_device *netdev = adapter->netdev;
1317	struct pci_dev *pdev = adapter->pdev;
1318	struct e1000_buffer *buffer_info, *next_buffer;
1319	struct e1000_ps_page *ps_page;
1320	struct sk_buff *skb;
1321	unsigned int i, j;
1322	u32 length, staterr;
1323	int cleaned_count = 0;
1324	bool cleaned = false;
1325	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1326
1327	i = rx_ring->next_to_clean;
1328	rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
1329	staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1330	buffer_info = &rx_ring->buffer_info[i];
1331
1332	while (staterr & E1000_RXD_STAT_DD) {
1333		if (*work_done >= work_to_do)
1334			break;
1335		(*work_done)++;
1336		skb = buffer_info->skb;
1337		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1338
1339		/* in the packet split case this is header only */
1340		prefetch(skb->data - NET_IP_ALIGN);
1341
1342		i++;
1343		if (i == rx_ring->count)
1344			i = 0;
1345		next_rxd = E1000_RX_DESC_PS(*rx_ring, i);
1346		prefetch(next_rxd);
1347
1348		next_buffer = &rx_ring->buffer_info[i];
1349
1350		cleaned = true;
1351		cleaned_count++;
1352		dma_unmap_single(&pdev->dev, buffer_info->dma,
1353				 adapter->rx_ps_bsize0, DMA_FROM_DEVICE);
1354		buffer_info->dma = 0;
1355
1356		/* see !EOP comment in other Rx routine */
1357		if (!(staterr & E1000_RXD_STAT_EOP))
1358			adapter->flags2 |= FLAG2_IS_DISCARDING;
1359
1360		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
1361			e_dbg("Packet Split buffers didn't pick up the full packet\n");
1362			dev_kfree_skb_irq(skb);
1363			if (staterr & E1000_RXD_STAT_EOP)
1364				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1365			goto next_desc;
1366		}
1367
1368		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1369			     !(netdev->features & NETIF_F_RXALL))) {
1370			dev_kfree_skb_irq(skb);
1371			goto next_desc;
1372		}
1373
1374		length = le16_to_cpu(rx_desc->wb.middle.length0);
1375
1376		if (!length) {
1377			e_dbg("Last part of the packet spanning multiple descriptors\n");
1378			dev_kfree_skb_irq(skb);
1379			goto next_desc;
1380		}
1381
1382		/* Good Receive */
1383		skb_put(skb, length);
1384
1385		{
1386			/* this looks ugly, but it seems compiler issues make
1387			 * it more efficient than reusing j
1388			 */
1389			int l1 = le16_to_cpu(rx_desc->wb.upper.length[0]);
1390
1391			/* page alloc/put takes too long and effects small
1392			 * packet throughput, so unsplit small packets and
1393			 * save the alloc/put
1394			 */
1395			if (l1 && (l1 <= copybreak) &&
1396			    ((length + l1) <= adapter->rx_ps_bsize0)) {
1397				ps_page = &buffer_info->ps_pages[0];
1398
1399				dma_sync_single_for_cpu(&pdev->dev,
1400							ps_page->dma,
1401							PAGE_SIZE,
1402							DMA_FROM_DEVICE);
1403				memcpy(skb_tail_pointer(skb),
1404				       page_address(ps_page->page), l1);
1405				dma_sync_single_for_device(&pdev->dev,
1406							   ps_page->dma,
1407							   PAGE_SIZE,
1408							   DMA_FROM_DEVICE);
1409
1410				/* remove the CRC */
1411				if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1412					if (!(netdev->features & NETIF_F_RXFCS))
1413						l1 -= 4;
1414				}
1415
1416				skb_put(skb, l1);
1417				goto copydone;
1418			}	/* if */
1419		}
1420
1421		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1422			length = le16_to_cpu(rx_desc->wb.upper.length[j]);
1423			if (!length)
1424				break;
1425
1426			ps_page = &buffer_info->ps_pages[j];
1427			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1428				       DMA_FROM_DEVICE);
1429			ps_page->dma = 0;
1430			skb_fill_page_desc(skb, j, ps_page->page, 0, length);
1431			ps_page->page = NULL;
1432			skb->len += length;
1433			skb->data_len += length;
1434			skb->truesize += PAGE_SIZE;
1435		}
1436
1437		/* strip the ethernet crc, problem is we're using pages now so
1438		 * this whole operation can get a little cpu intensive
1439		 */
1440		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1441			if (!(netdev->features & NETIF_F_RXFCS))
1442				pskb_trim(skb, skb->len - 4);
1443		}
1444
1445copydone:
1446		total_rx_bytes += skb->len;
1447		total_rx_packets++;
1448
1449		e1000_rx_checksum(adapter, staterr, skb);
1450
1451		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1452
1453		if (rx_desc->wb.upper.header_status &
1454		    cpu_to_le16(E1000_RXDPS_HDRSTAT_HDRSP))
1455			adapter->rx_hdr_split++;
1456
1457		e1000_receive_skb(adapter, netdev, skb, staterr,
1458				  rx_desc->wb.middle.vlan);
1459
1460next_desc:
1461		rx_desc->wb.middle.status_error &= cpu_to_le32(~0xFF);
1462		buffer_info->skb = NULL;
1463
1464		/* return some buffers to hardware, one at a time is too slow */
1465		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1466			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1467					      GFP_ATOMIC);
1468			cleaned_count = 0;
1469		}
1470
1471		/* use prefetched values */
1472		rx_desc = next_rxd;
1473		buffer_info = next_buffer;
1474
1475		staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1476	}
1477	rx_ring->next_to_clean = i;
1478
1479	cleaned_count = e1000_desc_unused(rx_ring);
1480	if (cleaned_count)
1481		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1482
1483	adapter->total_rx_bytes += total_rx_bytes;
1484	adapter->total_rx_packets += total_rx_packets;
1485	return cleaned;
1486}
1487
1488static void e1000_consume_page(struct e1000_buffer *bi, struct sk_buff *skb,
1489			       u16 length)
1490{
1491	bi->page = NULL;
1492	skb->len += length;
1493	skb->data_len += length;
1494	skb->truesize += PAGE_SIZE;
1495}
1496
1497/**
1498 * e1000_clean_jumbo_rx_irq - Send received data up the network stack; legacy
1499 * @rx_ring: Rx descriptor ring
1500 * @work_done: output parameter for indicating completed work
1501 * @work_to_do: how many packets we can clean
1502 *
1503 * the return value indicates whether actual cleaning was done, there
1504 * is no guarantee that everything was cleaned
1505 **/
1506static bool e1000_clean_jumbo_rx_irq(struct e1000_ring *rx_ring, int *work_done,
1507				     int work_to_do)
1508{
1509	struct e1000_adapter *adapter = rx_ring->adapter;
1510	struct net_device *netdev = adapter->netdev;
1511	struct pci_dev *pdev = adapter->pdev;
1512	union e1000_rx_desc_extended *rx_desc, *next_rxd;
1513	struct e1000_buffer *buffer_info, *next_buffer;
1514	u32 length, staterr;
1515	unsigned int i;
1516	int cleaned_count = 0;
1517	bool cleaned = false;
1518	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1519	struct skb_shared_info *shinfo;
1520
1521	i = rx_ring->next_to_clean;
1522	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
1523	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1524	buffer_info = &rx_ring->buffer_info[i];
1525
1526	while (staterr & E1000_RXD_STAT_DD) {
1527		struct sk_buff *skb;
1528
1529		if (*work_done >= work_to_do)
1530			break;
1531		(*work_done)++;
1532		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1533
1534		skb = buffer_info->skb;
1535		buffer_info->skb = NULL;
1536
1537		++i;
1538		if (i == rx_ring->count)
1539			i = 0;
1540		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
1541		prefetch(next_rxd);
1542
1543		next_buffer = &rx_ring->buffer_info[i];
1544
1545		cleaned = true;
1546		cleaned_count++;
1547		dma_unmap_page(&pdev->dev, buffer_info->dma, PAGE_SIZE,
1548			       DMA_FROM_DEVICE);
1549		buffer_info->dma = 0;
1550
1551		length = le16_to_cpu(rx_desc->wb.upper.length);
1552
1553		/* errors is only valid for DD + EOP descriptors */
1554		if (unlikely((staterr & E1000_RXD_STAT_EOP) &&
1555			     ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1556			      !(netdev->features & NETIF_F_RXALL)))) {
1557			/* recycle both page and skb */
1558			buffer_info->skb = skb;
1559			/* an error means any chain goes out the window too */
1560			if (rx_ring->rx_skb_top)
1561				dev_kfree_skb_irq(rx_ring->rx_skb_top);
1562			rx_ring->rx_skb_top = NULL;
1563			goto next_desc;
1564		}
1565#define rxtop (rx_ring->rx_skb_top)
1566		if (!(staterr & E1000_RXD_STAT_EOP)) {
1567			/* this descriptor is only the beginning (or middle) */
1568			if (!rxtop) {
1569				/* this is the beginning of a chain */
1570				rxtop = skb;
1571				skb_fill_page_desc(rxtop, 0, buffer_info->page,
1572						   0, length);
1573			} else {
1574				/* this is the middle of a chain */
1575				shinfo = skb_shinfo(rxtop);
1576				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1577						   buffer_info->page, 0,
1578						   length);
1579				/* re-use the skb, only consumed the page */
1580				buffer_info->skb = skb;
1581			}
1582			e1000_consume_page(buffer_info, rxtop, length);
1583			goto next_desc;
1584		} else {
1585			if (rxtop) {
1586				/* end of the chain */
1587				shinfo = skb_shinfo(rxtop);
1588				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1589						   buffer_info->page, 0,
1590						   length);
1591				/* re-use the current skb, we only consumed the
1592				 * page
1593				 */
1594				buffer_info->skb = skb;
1595				skb = rxtop;
1596				rxtop = NULL;
1597				e1000_consume_page(buffer_info, skb, length);
1598			} else {
1599				/* no chain, got EOP, this buf is the packet
1600				 * copybreak to save the put_page/alloc_page
1601				 */
1602				if (length <= copybreak &&
1603				    skb_tailroom(skb) >= length) {
1604					memcpy(skb_tail_pointer(skb),
1605					       page_address(buffer_info->page),
1606					       length);
1607					/* re-use the page, so don't erase
1608					 * buffer_info->page
1609					 */
1610					skb_put(skb, length);
1611				} else {
1612					skb_fill_page_desc(skb, 0,
1613							   buffer_info->page, 0,
1614							   length);
1615					e1000_consume_page(buffer_info, skb,
1616							   length);
1617				}
1618			}
1619		}
1620
1621		/* Receive Checksum Offload */
1622		e1000_rx_checksum(adapter, staterr, skb);
1623
1624		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1625
1626		/* probably a little skewed due to removing CRC */
1627		total_rx_bytes += skb->len;
1628		total_rx_packets++;
1629
1630		/* eth type trans needs skb->data to point to something */
1631		if (!pskb_may_pull(skb, ETH_HLEN)) {
1632			e_err("pskb_may_pull failed.\n");
1633			dev_kfree_skb_irq(skb);
1634			goto next_desc;
1635		}
1636
1637		e1000_receive_skb(adapter, netdev, skb, staterr,
1638				  rx_desc->wb.upper.vlan);
1639
1640next_desc:
1641		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1642
1643		/* return some buffers to hardware, one at a time is too slow */
1644		if (unlikely(cleaned_count >= E1000_RX_BUFFER_WRITE)) {
1645			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1646					      GFP_ATOMIC);
1647			cleaned_count = 0;
1648		}
1649
1650		/* use prefetched values */
1651		rx_desc = next_rxd;
1652		buffer_info = next_buffer;
1653
1654		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1655	}
1656	rx_ring->next_to_clean = i;
1657
1658	cleaned_count = e1000_desc_unused(rx_ring);
1659	if (cleaned_count)
1660		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1661
1662	adapter->total_rx_bytes += total_rx_bytes;
1663	adapter->total_rx_packets += total_rx_packets;
1664	return cleaned;
1665}
1666
1667/**
1668 * e1000_clean_rx_ring - Free Rx Buffers per Queue
1669 * @rx_ring: Rx descriptor ring
1670 **/
1671static void e1000_clean_rx_ring(struct e1000_ring *rx_ring)
1672{
1673	struct e1000_adapter *adapter = rx_ring->adapter;
1674	struct e1000_buffer *buffer_info;
1675	struct e1000_ps_page *ps_page;
1676	struct pci_dev *pdev = adapter->pdev;
1677	unsigned int i, j;
1678
1679	/* Free all the Rx ring sk_buffs */
1680	for (i = 0; i < rx_ring->count; i++) {
1681		buffer_info = &rx_ring->buffer_info[i];
1682		if (buffer_info->dma) {
1683			if (adapter->clean_rx == e1000_clean_rx_irq)
1684				dma_unmap_single(&pdev->dev, buffer_info->dma,
1685						 adapter->rx_buffer_len,
1686						 DMA_FROM_DEVICE);
1687			else if (adapter->clean_rx == e1000_clean_jumbo_rx_irq)
1688				dma_unmap_page(&pdev->dev, buffer_info->dma,
1689					       PAGE_SIZE, DMA_FROM_DEVICE);
1690			else if (adapter->clean_rx == e1000_clean_rx_irq_ps)
1691				dma_unmap_single(&pdev->dev, buffer_info->dma,
1692						 adapter->rx_ps_bsize0,
1693						 DMA_FROM_DEVICE);
1694			buffer_info->dma = 0;
1695		}
1696
1697		if (buffer_info->page) {
1698			put_page(buffer_info->page);
1699			buffer_info->page = NULL;
1700		}
1701
1702		if (buffer_info->skb) {
1703			dev_kfree_skb(buffer_info->skb);
1704			buffer_info->skb = NULL;
1705		}
1706
1707		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1708			ps_page = &buffer_info->ps_pages[j];
1709			if (!ps_page->page)
1710				break;
1711			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1712				       DMA_FROM_DEVICE);
1713			ps_page->dma = 0;
1714			put_page(ps_page->page);
1715			ps_page->page = NULL;
1716		}
1717	}
1718
1719	/* there also may be some cached data from a chained receive */
1720	if (rx_ring->rx_skb_top) {
1721		dev_kfree_skb(rx_ring->rx_skb_top);
1722		rx_ring->rx_skb_top = NULL;
1723	}
1724
1725	/* Zero out the descriptor ring */
1726	memset(rx_ring->desc, 0, rx_ring->size);
1727
1728	rx_ring->next_to_clean = 0;
1729	rx_ring->next_to_use = 0;
1730	adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1731}
1732
1733static void e1000e_downshift_workaround(struct work_struct *work)
1734{
1735	struct e1000_adapter *adapter = container_of(work,
1736						     struct e1000_adapter,
1737						     downshift_task);
1738
1739	if (test_bit(__E1000_DOWN, &adapter->state))
1740		return;
1741
1742	e1000e_gig_downshift_workaround_ich8lan(&adapter->hw);
1743}
1744
1745/**
1746 * e1000_intr_msi - Interrupt Handler
1747 * @irq: interrupt number
1748 * @data: pointer to a network interface device structure
1749 **/
1750static irqreturn_t e1000_intr_msi(int __always_unused irq, void *data)
1751{
1752	struct net_device *netdev = data;
1753	struct e1000_adapter *adapter = netdev_priv(netdev);
1754	struct e1000_hw *hw = &adapter->hw;
1755	u32 icr = er32(ICR);
1756
1757	/* read ICR disables interrupts using IAM */
1758	if (icr & E1000_ICR_LSC) {
1759		hw->mac.get_link_status = true;
1760		/* ICH8 workaround-- Call gig speed drop workaround on cable
1761		 * disconnect (LSC) before accessing any PHY registers
1762		 */
1763		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1764		    (!(er32(STATUS) & E1000_STATUS_LU)))
1765			schedule_work(&adapter->downshift_task);
1766
1767		/* 80003ES2LAN workaround-- For packet buffer work-around on
1768		 * link down event; disable receives here in the ISR and reset
1769		 * adapter in watchdog
1770		 */
1771		if (netif_carrier_ok(netdev) &&
1772		    adapter->flags & FLAG_RX_NEEDS_RESTART) {
1773			/* disable receives */
1774			u32 rctl = er32(RCTL);
1775
1776			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1777			adapter->flags |= FLAG_RESTART_NOW;
1778		}
1779		/* guard against interrupt when we're going down */
1780		if (!test_bit(__E1000_DOWN, &adapter->state))
1781			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1782	}
1783
1784	/* Reset on uncorrectable ECC error */
1785	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1786		u32 pbeccsts = er32(PBECCSTS);
1787
1788		adapter->corr_errors +=
1789		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1790		adapter->uncorr_errors +=
1791		    FIELD_GET(E1000_PBECCSTS_UNCORR_ERR_CNT_MASK, pbeccsts);
1792
1793		/* Do the reset outside of interrupt context */
1794		schedule_work(&adapter->reset_task);
1795
1796		/* return immediately since reset is imminent */
1797		return IRQ_HANDLED;
1798	}
1799
1800	if (napi_schedule_prep(&adapter->napi)) {
1801		adapter->total_tx_bytes = 0;
1802		adapter->total_tx_packets = 0;
1803		adapter->total_rx_bytes = 0;
1804		adapter->total_rx_packets = 0;
1805		__napi_schedule(&adapter->napi);
1806	}
1807
1808	return IRQ_HANDLED;
1809}
1810
1811/**
1812 * e1000_intr - Interrupt Handler
1813 * @irq: interrupt number
1814 * @data: pointer to a network interface device structure
1815 **/
1816static irqreturn_t e1000_intr(int __always_unused irq, void *data)
1817{
1818	struct net_device *netdev = data;
1819	struct e1000_adapter *adapter = netdev_priv(netdev);
1820	struct e1000_hw *hw = &adapter->hw;
1821	u32 rctl, icr = er32(ICR);
1822
1823	if (!icr || test_bit(__E1000_DOWN, &adapter->state))
1824		return IRQ_NONE;	/* Not our interrupt */
1825
1826	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
1827	 * not set, then the adapter didn't send an interrupt
1828	 */
1829	if (!(icr & E1000_ICR_INT_ASSERTED))
1830		return IRQ_NONE;
1831
1832	/* Interrupt Auto-Mask...upon reading ICR,
1833	 * interrupts are masked.  No need for the
1834	 * IMC write
1835	 */
1836
1837	if (icr & E1000_ICR_LSC) {
1838		hw->mac.get_link_status = true;
1839		/* ICH8 workaround-- Call gig speed drop workaround on cable
1840		 * disconnect (LSC) before accessing any PHY registers
1841		 */
1842		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1843		    (!(er32(STATUS) & E1000_STATUS_LU)))
1844			schedule_work(&adapter->downshift_task);
1845
1846		/* 80003ES2LAN workaround--
1847		 * For packet buffer work-around on link down event;
1848		 * disable receives here in the ISR and
1849		 * reset adapter in watchdog
1850		 */
1851		if (netif_carrier_ok(netdev) &&
1852		    (adapter->flags & FLAG_RX_NEEDS_RESTART)) {
1853			/* disable receives */
1854			rctl = er32(RCTL);
1855			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1856			adapter->flags |= FLAG_RESTART_NOW;
1857		}
1858		/* guard against interrupt when we're going down */
1859		if (!test_bit(__E1000_DOWN, &adapter->state))
1860			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1861	}
1862
1863	/* Reset on uncorrectable ECC error */
1864	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1865		u32 pbeccsts = er32(PBECCSTS);
1866
1867		adapter->corr_errors +=
1868		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1869		adapter->uncorr_errors +=
1870		    FIELD_GET(E1000_PBECCSTS_UNCORR_ERR_CNT_MASK, pbeccsts);
1871
1872		/* Do the reset outside of interrupt context */
1873		schedule_work(&adapter->reset_task);
1874
1875		/* return immediately since reset is imminent */
1876		return IRQ_HANDLED;
1877	}
1878
1879	if (napi_schedule_prep(&adapter->napi)) {
1880		adapter->total_tx_bytes = 0;
1881		adapter->total_tx_packets = 0;
1882		adapter->total_rx_bytes = 0;
1883		adapter->total_rx_packets = 0;
1884		__napi_schedule(&adapter->napi);
1885	}
1886
1887	return IRQ_HANDLED;
1888}
1889
1890static irqreturn_t e1000_msix_other(int __always_unused irq, void *data)
1891{
1892	struct net_device *netdev = data;
1893	struct e1000_adapter *adapter = netdev_priv(netdev);
1894	struct e1000_hw *hw = &adapter->hw;
1895	u32 icr = er32(ICR);
1896
1897	if (icr & adapter->eiac_mask)
1898		ew32(ICS, (icr & adapter->eiac_mask));
1899
1900	if (icr & E1000_ICR_LSC) {
1901		hw->mac.get_link_status = true;
1902		/* guard against interrupt when we're going down */
1903		if (!test_bit(__E1000_DOWN, &adapter->state))
1904			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1905	}
1906
1907	if (!test_bit(__E1000_DOWN, &adapter->state))
1908		ew32(IMS, E1000_IMS_OTHER | IMS_OTHER_MASK);
1909
1910	return IRQ_HANDLED;
1911}
1912
1913static irqreturn_t e1000_intr_msix_tx(int __always_unused irq, void *data)
1914{
1915	struct net_device *netdev = data;
1916	struct e1000_adapter *adapter = netdev_priv(netdev);
1917	struct e1000_hw *hw = &adapter->hw;
1918	struct e1000_ring *tx_ring = adapter->tx_ring;
1919
1920	adapter->total_tx_bytes = 0;
1921	adapter->total_tx_packets = 0;
1922
1923	if (!e1000_clean_tx_irq(tx_ring))
1924		/* Ring was not completely cleaned, so fire another interrupt */
1925		ew32(ICS, tx_ring->ims_val);
1926
1927	if (!test_bit(__E1000_DOWN, &adapter->state))
1928		ew32(IMS, adapter->tx_ring->ims_val);
1929
1930	return IRQ_HANDLED;
1931}
1932
1933static irqreturn_t e1000_intr_msix_rx(int __always_unused irq, void *data)
1934{
1935	struct net_device *netdev = data;
1936	struct e1000_adapter *adapter = netdev_priv(netdev);
1937	struct e1000_ring *rx_ring = adapter->rx_ring;
1938
1939	/* Write the ITR value calculated at the end of the
1940	 * previous interrupt.
1941	 */
1942	if (rx_ring->set_itr) {
1943		u32 itr = rx_ring->itr_val ?
1944			  1000000000 / (rx_ring->itr_val * 256) : 0;
1945
1946		writel(itr, rx_ring->itr_register);
1947		rx_ring->set_itr = 0;
1948	}
1949
1950	if (napi_schedule_prep(&adapter->napi)) {
1951		adapter->total_rx_bytes = 0;
1952		adapter->total_rx_packets = 0;
1953		__napi_schedule(&adapter->napi);
1954	}
1955	return IRQ_HANDLED;
1956}
1957
1958/**
1959 * e1000_configure_msix - Configure MSI-X hardware
1960 * @adapter: board private structure
1961 *
1962 * e1000_configure_msix sets up the hardware to properly
1963 * generate MSI-X interrupts.
1964 **/
1965static void e1000_configure_msix(struct e1000_adapter *adapter)
1966{
1967	struct e1000_hw *hw = &adapter->hw;
1968	struct e1000_ring *rx_ring = adapter->rx_ring;
1969	struct e1000_ring *tx_ring = adapter->tx_ring;
1970	int vector = 0;
1971	u32 ctrl_ext, ivar = 0;
1972
1973	adapter->eiac_mask = 0;
1974
1975	/* Workaround issue with spurious interrupts on 82574 in MSI-X mode */
1976	if (hw->mac.type == e1000_82574) {
1977		u32 rfctl = er32(RFCTL);
1978
1979		rfctl |= E1000_RFCTL_ACK_DIS;
1980		ew32(RFCTL, rfctl);
1981	}
1982
1983	/* Configure Rx vector */
1984	rx_ring->ims_val = E1000_IMS_RXQ0;
1985	adapter->eiac_mask |= rx_ring->ims_val;
1986	if (rx_ring->itr_val)
1987		writel(1000000000 / (rx_ring->itr_val * 256),
1988		       rx_ring->itr_register);
1989	else
1990		writel(1, rx_ring->itr_register);
1991	ivar = E1000_IVAR_INT_ALLOC_VALID | vector;
1992
1993	/* Configure Tx vector */
1994	tx_ring->ims_val = E1000_IMS_TXQ0;
1995	vector++;
1996	if (tx_ring->itr_val)
1997		writel(1000000000 / (tx_ring->itr_val * 256),
1998		       tx_ring->itr_register);
1999	else
2000		writel(1, tx_ring->itr_register);
2001	adapter->eiac_mask |= tx_ring->ims_val;
2002	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 8);
2003
2004	/* set vector for Other Causes, e.g. link changes */
2005	vector++;
2006	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 16);
2007	if (rx_ring->itr_val)
2008		writel(1000000000 / (rx_ring->itr_val * 256),
2009		       hw->hw_addr + E1000_EITR_82574(vector));
2010	else
2011		writel(1, hw->hw_addr + E1000_EITR_82574(vector));
2012
2013	/* Cause Tx interrupts on every write back */
2014	ivar |= BIT(31);
2015
2016	ew32(IVAR, ivar);
2017
2018	/* enable MSI-X PBA support */
2019	ctrl_ext = er32(CTRL_EXT) & ~E1000_CTRL_EXT_IAME;
2020	ctrl_ext |= E1000_CTRL_EXT_PBA_CLR | E1000_CTRL_EXT_EIAME;
2021	ew32(CTRL_EXT, ctrl_ext);
2022	e1e_flush();
2023}
2024
2025void e1000e_reset_interrupt_capability(struct e1000_adapter *adapter)
2026{
2027	if (adapter->msix_entries) {
2028		pci_disable_msix(adapter->pdev);
2029		kfree(adapter->msix_entries);
2030		adapter->msix_entries = NULL;
2031	} else if (adapter->flags & FLAG_MSI_ENABLED) {
2032		pci_disable_msi(adapter->pdev);
2033		adapter->flags &= ~FLAG_MSI_ENABLED;
2034	}
2035}
2036
2037/**
2038 * e1000e_set_interrupt_capability - set MSI or MSI-X if supported
2039 * @adapter: board private structure
2040 *
2041 * Attempt to configure interrupts using the best available
2042 * capabilities of the hardware and kernel.
2043 **/
2044void e1000e_set_interrupt_capability(struct e1000_adapter *adapter)
2045{
2046	int err;
2047	int i;
2048
2049	switch (adapter->int_mode) {
2050	case E1000E_INT_MODE_MSIX:
2051		if (adapter->flags & FLAG_HAS_MSIX) {
2052			adapter->num_vectors = 3; /* RxQ0, TxQ0 and other */
2053			adapter->msix_entries = kcalloc(adapter->num_vectors,
2054							sizeof(struct
2055							       msix_entry),
2056							GFP_KERNEL);
2057			if (adapter->msix_entries) {
2058				struct e1000_adapter *a = adapter;
2059
2060				for (i = 0; i < adapter->num_vectors; i++)
2061					adapter->msix_entries[i].entry = i;
2062
2063				err = pci_enable_msix_range(a->pdev,
2064							    a->msix_entries,
2065							    a->num_vectors,
2066							    a->num_vectors);
2067				if (err > 0)
2068					return;
2069			}
2070			/* MSI-X failed, so fall through and try MSI */
2071			e_err("Failed to initialize MSI-X interrupts.  Falling back to MSI interrupts.\n");
2072			e1000e_reset_interrupt_capability(adapter);
2073		}
2074		adapter->int_mode = E1000E_INT_MODE_MSI;
2075		fallthrough;
2076	case E1000E_INT_MODE_MSI:
2077		if (!pci_enable_msi(adapter->pdev)) {
2078			adapter->flags |= FLAG_MSI_ENABLED;
2079		} else {
2080			adapter->int_mode = E1000E_INT_MODE_LEGACY;
2081			e_err("Failed to initialize MSI interrupts.  Falling back to legacy interrupts.\n");
2082		}
2083		fallthrough;
2084	case E1000E_INT_MODE_LEGACY:
2085		/* Don't do anything; this is the system default */
2086		break;
2087	}
2088
2089	/* store the number of vectors being used */
2090	adapter->num_vectors = 1;
2091}
2092
2093/**
2094 * e1000_request_msix - Initialize MSI-X interrupts
2095 * @adapter: board private structure
2096 *
2097 * e1000_request_msix allocates MSI-X vectors and requests interrupts from the
2098 * kernel.
2099 **/
2100static int e1000_request_msix(struct e1000_adapter *adapter)
2101{
2102	struct net_device *netdev = adapter->netdev;
2103	int err = 0, vector = 0;
2104
2105	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2106		snprintf(adapter->rx_ring->name,
2107			 sizeof(adapter->rx_ring->name) - 1,
2108			 "%.14s-rx-0", netdev->name);
2109	else
2110		memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
2111	err = request_irq(adapter->msix_entries[vector].vector,
2112			  e1000_intr_msix_rx, 0, adapter->rx_ring->name,
2113			  netdev);
2114	if (err)
2115		return err;
2116	adapter->rx_ring->itr_register = adapter->hw.hw_addr +
2117	    E1000_EITR_82574(vector);
2118	adapter->rx_ring->itr_val = adapter->itr;
2119	vector++;
2120
2121	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2122		snprintf(adapter->tx_ring->name,
2123			 sizeof(adapter->tx_ring->name) - 1,
2124			 "%.14s-tx-0", netdev->name);
2125	else
2126		memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
2127	err = request_irq(adapter->msix_entries[vector].vector,
2128			  e1000_intr_msix_tx, 0, adapter->tx_ring->name,
2129			  netdev);
2130	if (err)
2131		return err;
2132	adapter->tx_ring->itr_register = adapter->hw.hw_addr +
2133	    E1000_EITR_82574(vector);
2134	adapter->tx_ring->itr_val = adapter->itr;
2135	vector++;
2136
2137	err = request_irq(adapter->msix_entries[vector].vector,
2138			  e1000_msix_other, 0, netdev->name, netdev);
2139	if (err)
2140		return err;
2141
2142	e1000_configure_msix(adapter);
2143
2144	return 0;
2145}
2146
2147/**
2148 * e1000_request_irq - initialize interrupts
2149 * @adapter: board private structure
2150 *
2151 * Attempts to configure interrupts using the best available
2152 * capabilities of the hardware and kernel.
2153 **/
2154static int e1000_request_irq(struct e1000_adapter *adapter)
2155{
2156	struct net_device *netdev = adapter->netdev;
2157	int err;
2158
2159	if (adapter->msix_entries) {
2160		err = e1000_request_msix(adapter);
2161		if (!err)
2162			return err;
2163		/* fall back to MSI */
2164		e1000e_reset_interrupt_capability(adapter);
2165		adapter->int_mode = E1000E_INT_MODE_MSI;
2166		e1000e_set_interrupt_capability(adapter);
2167	}
2168	if (adapter->flags & FLAG_MSI_ENABLED) {
2169		err = request_irq(adapter->pdev->irq, e1000_intr_msi, 0,
2170				  netdev->name, netdev);
2171		if (!err)
2172			return err;
2173
2174		/* fall back to legacy interrupt */
2175		e1000e_reset_interrupt_capability(adapter);
2176		adapter->int_mode = E1000E_INT_MODE_LEGACY;
2177	}
2178
2179	err = request_irq(adapter->pdev->irq, e1000_intr, IRQF_SHARED,
2180			  netdev->name, netdev);
2181	if (err)
2182		e_err("Unable to allocate interrupt, Error: %d\n", err);
2183
2184	return err;
2185}
2186
2187static void e1000_free_irq(struct e1000_adapter *adapter)
2188{
2189	struct net_device *netdev = adapter->netdev;
2190
2191	if (adapter->msix_entries) {
2192		int vector = 0;
2193
2194		free_irq(adapter->msix_entries[vector].vector, netdev);
2195		vector++;
2196
2197		free_irq(adapter->msix_entries[vector].vector, netdev);
2198		vector++;
2199
2200		/* Other Causes interrupt vector */
2201		free_irq(adapter->msix_entries[vector].vector, netdev);
2202		return;
2203	}
2204
2205	free_irq(adapter->pdev->irq, netdev);
2206}
2207
2208/**
2209 * e1000_irq_disable - Mask off interrupt generation on the NIC
2210 * @adapter: board private structure
2211 **/
2212static void e1000_irq_disable(struct e1000_adapter *adapter)
2213{
2214	struct e1000_hw *hw = &adapter->hw;
2215
2216	ew32(IMC, ~0);
2217	if (adapter->msix_entries)
2218		ew32(EIAC_82574, 0);
2219	e1e_flush();
2220
2221	if (adapter->msix_entries) {
2222		int i;
2223
2224		for (i = 0; i < adapter->num_vectors; i++)
2225			synchronize_irq(adapter->msix_entries[i].vector);
2226	} else {
2227		synchronize_irq(adapter->pdev->irq);
2228	}
2229}
2230
2231/**
2232 * e1000_irq_enable - Enable default interrupt generation settings
2233 * @adapter: board private structure
2234 **/
2235static void e1000_irq_enable(struct e1000_adapter *adapter)
2236{
2237	struct e1000_hw *hw = &adapter->hw;
2238
2239	if (adapter->msix_entries) {
2240		ew32(EIAC_82574, adapter->eiac_mask & E1000_EIAC_MASK_82574);
2241		ew32(IMS, adapter->eiac_mask | E1000_IMS_OTHER |
2242		     IMS_OTHER_MASK);
2243	} else if (hw->mac.type >= e1000_pch_lpt) {
2244		ew32(IMS, IMS_ENABLE_MASK | E1000_IMS_ECCER);
2245	} else {
2246		ew32(IMS, IMS_ENABLE_MASK);
2247	}
2248	e1e_flush();
2249}
2250
2251/**
2252 * e1000e_get_hw_control - get control of the h/w from f/w
2253 * @adapter: address of board private structure
2254 *
2255 * e1000e_get_hw_control sets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2256 * For ASF and Pass Through versions of f/w this means that
2257 * the driver is loaded. For AMT version (only with 82573)
2258 * of the f/w this means that the network i/f is open.
2259 **/
2260void e1000e_get_hw_control(struct e1000_adapter *adapter)
2261{
2262	struct e1000_hw *hw = &adapter->hw;
2263	u32 ctrl_ext;
2264	u32 swsm;
2265
2266	/* Let firmware know the driver has taken over */
2267	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2268		swsm = er32(SWSM);
2269		ew32(SWSM, swsm | E1000_SWSM_DRV_LOAD);
2270	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2271		ctrl_ext = er32(CTRL_EXT);
2272		ew32(CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2273	}
2274}
2275
2276/**
2277 * e1000e_release_hw_control - release control of the h/w to f/w
2278 * @adapter: address of board private structure
2279 *
2280 * e1000e_release_hw_control resets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2281 * For ASF and Pass Through versions of f/w this means that the
2282 * driver is no longer loaded. For AMT version (only with 82573) i
2283 * of the f/w this means that the network i/f is closed.
2284 *
2285 **/
2286void e1000e_release_hw_control(struct e1000_adapter *adapter)
2287{
2288	struct e1000_hw *hw = &adapter->hw;
2289	u32 ctrl_ext;
2290	u32 swsm;
2291
2292	/* Let firmware taken over control of h/w */
2293	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2294		swsm = er32(SWSM);
2295		ew32(SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
2296	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2297		ctrl_ext = er32(CTRL_EXT);
2298		ew32(CTRL_EXT, ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2299	}
2300}
2301
2302/**
2303 * e1000_alloc_ring_dma - allocate memory for a ring structure
2304 * @adapter: board private structure
2305 * @ring: ring struct for which to allocate dma
2306 **/
2307static int e1000_alloc_ring_dma(struct e1000_adapter *adapter,
2308				struct e1000_ring *ring)
2309{
2310	struct pci_dev *pdev = adapter->pdev;
2311
2312	ring->desc = dma_alloc_coherent(&pdev->dev, ring->size, &ring->dma,
2313					GFP_KERNEL);
2314	if (!ring->desc)
2315		return -ENOMEM;
2316
2317	return 0;
2318}
2319
2320/**
2321 * e1000e_setup_tx_resources - allocate Tx resources (Descriptors)
2322 * @tx_ring: Tx descriptor ring
2323 *
2324 * Return 0 on success, negative on failure
2325 **/
2326int e1000e_setup_tx_resources(struct e1000_ring *tx_ring)
2327{
2328	struct e1000_adapter *adapter = tx_ring->adapter;
2329	int err = -ENOMEM, size;
2330
2331	size = sizeof(struct e1000_buffer) * tx_ring->count;
2332	tx_ring->buffer_info = vzalloc(size);
2333	if (!tx_ring->buffer_info)
2334		goto err;
2335
2336	/* round up to nearest 4K */
2337	tx_ring->size = tx_ring->count * sizeof(struct e1000_tx_desc);
2338	tx_ring->size = ALIGN(tx_ring->size, 4096);
2339
2340	err = e1000_alloc_ring_dma(adapter, tx_ring);
2341	if (err)
2342		goto err;
2343
2344	tx_ring->next_to_use = 0;
2345	tx_ring->next_to_clean = 0;
2346
2347	return 0;
2348err:
2349	vfree(tx_ring->buffer_info);
2350	e_err("Unable to allocate memory for the transmit descriptor ring\n");
2351	return err;
2352}
2353
2354/**
2355 * e1000e_setup_rx_resources - allocate Rx resources (Descriptors)
2356 * @rx_ring: Rx descriptor ring
2357 *
2358 * Returns 0 on success, negative on failure
2359 **/
2360int e1000e_setup_rx_resources(struct e1000_ring *rx_ring)
2361{
2362	struct e1000_adapter *adapter = rx_ring->adapter;
2363	struct e1000_buffer *buffer_info;
2364	int i, size, desc_len, err = -ENOMEM;
2365
2366	size = sizeof(struct e1000_buffer) * rx_ring->count;
2367	rx_ring->buffer_info = vzalloc(size);
2368	if (!rx_ring->buffer_info)
2369		goto err;
2370
2371	for (i = 0; i < rx_ring->count; i++) {
2372		buffer_info = &rx_ring->buffer_info[i];
2373		buffer_info->ps_pages = kcalloc(PS_PAGE_BUFFERS,
2374						sizeof(struct e1000_ps_page),
2375						GFP_KERNEL);
2376		if (!buffer_info->ps_pages)
2377			goto err_pages;
2378	}
2379
2380	desc_len = sizeof(union e1000_rx_desc_packet_split);
2381
2382	/* Round up to nearest 4K */
2383	rx_ring->size = rx_ring->count * desc_len;
2384	rx_ring->size = ALIGN(rx_ring->size, 4096);
2385
2386	err = e1000_alloc_ring_dma(adapter, rx_ring);
2387	if (err)
2388		goto err_pages;
2389
2390	rx_ring->next_to_clean = 0;
2391	rx_ring->next_to_use = 0;
2392	rx_ring->rx_skb_top = NULL;
2393
2394	return 0;
2395
2396err_pages:
2397	for (i = 0; i < rx_ring->count; i++) {
2398		buffer_info = &rx_ring->buffer_info[i];
2399		kfree(buffer_info->ps_pages);
2400	}
2401err:
2402	vfree(rx_ring->buffer_info);
2403	e_err("Unable to allocate memory for the receive descriptor ring\n");
2404	return err;
2405}
2406
2407/**
2408 * e1000_clean_tx_ring - Free Tx Buffers
2409 * @tx_ring: Tx descriptor ring
2410 **/
2411static void e1000_clean_tx_ring(struct e1000_ring *tx_ring)
2412{
2413	struct e1000_adapter *adapter = tx_ring->adapter;
2414	struct e1000_buffer *buffer_info;
2415	unsigned long size;
2416	unsigned int i;
2417
2418	for (i = 0; i < tx_ring->count; i++) {
2419		buffer_info = &tx_ring->buffer_info[i];
2420		e1000_put_txbuf(tx_ring, buffer_info, false);
2421	}
2422
2423	netdev_reset_queue(adapter->netdev);
2424	size = sizeof(struct e1000_buffer) * tx_ring->count;
2425	memset(tx_ring->buffer_info, 0, size);
2426
2427	memset(tx_ring->desc, 0, tx_ring->size);
2428
2429	tx_ring->next_to_use = 0;
2430	tx_ring->next_to_clean = 0;
2431}
2432
2433/**
2434 * e1000e_free_tx_resources - Free Tx Resources per Queue
2435 * @tx_ring: Tx descriptor ring
2436 *
2437 * Free all transmit software resources
2438 **/
2439void e1000e_free_tx_resources(struct e1000_ring *tx_ring)
2440{
2441	struct e1000_adapter *adapter = tx_ring->adapter;
2442	struct pci_dev *pdev = adapter->pdev;
2443
2444	e1000_clean_tx_ring(tx_ring);
2445
2446	vfree(tx_ring->buffer_info);
2447	tx_ring->buffer_info = NULL;
2448
2449	dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
2450			  tx_ring->dma);
2451	tx_ring->desc = NULL;
2452}
2453
2454/**
2455 * e1000e_free_rx_resources - Free Rx Resources
2456 * @rx_ring: Rx descriptor ring
2457 *
2458 * Free all receive software resources
2459 **/
2460void e1000e_free_rx_resources(struct e1000_ring *rx_ring)
2461{
2462	struct e1000_adapter *adapter = rx_ring->adapter;
2463	struct pci_dev *pdev = adapter->pdev;
2464	int i;
2465
2466	e1000_clean_rx_ring(rx_ring);
2467
2468	for (i = 0; i < rx_ring->count; i++)
2469		kfree(rx_ring->buffer_info[i].ps_pages);
2470
2471	vfree(rx_ring->buffer_info);
2472	rx_ring->buffer_info = NULL;
2473
2474	dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
2475			  rx_ring->dma);
2476	rx_ring->desc = NULL;
2477}
2478
2479/**
2480 * e1000_update_itr - update the dynamic ITR value based on statistics
2481 * @itr_setting: current adapter->itr
2482 * @packets: the number of packets during this measurement interval
2483 * @bytes: the number of bytes during this measurement interval
2484 *
2485 *      Stores a new ITR value based on packets and byte
2486 *      counts during the last interrupt.  The advantage of per interrupt
2487 *      computation is faster updates and more accurate ITR for the current
2488 *      traffic pattern.  Constants in this function were computed
2489 *      based on theoretical maximum wire speed and thresholds were set based
2490 *      on testing data as well as attempting to minimize response time
2491 *      while increasing bulk throughput.  This functionality is controlled
2492 *      by the InterruptThrottleRate module parameter.
2493 **/
2494static unsigned int e1000_update_itr(u16 itr_setting, int packets, int bytes)
2495{
2496	unsigned int retval = itr_setting;
2497
2498	if (packets == 0)
2499		return itr_setting;
2500
2501	switch (itr_setting) {
2502	case lowest_latency:
2503		/* handle TSO and jumbo frames */
2504		if (bytes / packets > 8000)
2505			retval = bulk_latency;
2506		else if ((packets < 5) && (bytes > 512))
2507			retval = low_latency;
2508		break;
2509	case low_latency:	/* 50 usec aka 20000 ints/s */
2510		if (bytes > 10000) {
2511			/* this if handles the TSO accounting */
2512			if (bytes / packets > 8000)
2513				retval = bulk_latency;
2514			else if ((packets < 10) || ((bytes / packets) > 1200))
2515				retval = bulk_latency;
2516			else if ((packets > 35))
2517				retval = lowest_latency;
2518		} else if (bytes / packets > 2000) {
2519			retval = bulk_latency;
2520		} else if (packets <= 2 && bytes < 512) {
2521			retval = lowest_latency;
2522		}
2523		break;
2524	case bulk_latency:	/* 250 usec aka 4000 ints/s */
2525		if (bytes > 25000) {
2526			if (packets > 35)
2527				retval = low_latency;
2528		} else if (bytes < 6000) {
2529			retval = low_latency;
2530		}
2531		break;
2532	}
2533
2534	return retval;
2535}
2536
2537static void e1000_set_itr(struct e1000_adapter *adapter)
2538{
2539	u16 current_itr;
2540	u32 new_itr = adapter->itr;
2541
2542	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
2543	if (adapter->link_speed != SPEED_1000) {
2544		new_itr = 4000;
2545		goto set_itr_now;
2546	}
2547
2548	if (adapter->flags2 & FLAG2_DISABLE_AIM) {
2549		new_itr = 0;
2550		goto set_itr_now;
2551	}
2552
2553	adapter->tx_itr = e1000_update_itr(adapter->tx_itr,
2554					   adapter->total_tx_packets,
2555					   adapter->total_tx_bytes);
2556	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2557	if (adapter->itr_setting == 3 && adapter->tx_itr == lowest_latency)
2558		adapter->tx_itr = low_latency;
2559
2560	adapter->rx_itr = e1000_update_itr(adapter->rx_itr,
2561					   adapter->total_rx_packets,
2562					   adapter->total_rx_bytes);
2563	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2564	if (adapter->itr_setting == 3 && adapter->rx_itr == lowest_latency)
2565		adapter->rx_itr = low_latency;
2566
2567	current_itr = max(adapter->rx_itr, adapter->tx_itr);
2568
2569	/* counts and packets in update_itr are dependent on these numbers */
2570	switch (current_itr) {
2571	case lowest_latency:
2572		new_itr = 70000;
2573		break;
2574	case low_latency:
2575		new_itr = 20000;	/* aka hwitr = ~200 */
2576		break;
2577	case bulk_latency:
2578		new_itr = 4000;
2579		break;
2580	default:
2581		break;
2582	}
2583
2584set_itr_now:
2585	if (new_itr != adapter->itr) {
2586		/* this attempts to bias the interrupt rate towards Bulk
2587		 * by adding intermediate steps when interrupt rate is
2588		 * increasing
2589		 */
2590		new_itr = new_itr > adapter->itr ?
2591		    min(adapter->itr + (new_itr >> 2), new_itr) : new_itr;
2592		adapter->itr = new_itr;
2593		adapter->rx_ring->itr_val = new_itr;
2594		if (adapter->msix_entries)
2595			adapter->rx_ring->set_itr = 1;
2596		else
2597			e1000e_write_itr(adapter, new_itr);
2598	}
2599}
2600
2601/**
2602 * e1000e_write_itr - write the ITR value to the appropriate registers
2603 * @adapter: address of board private structure
2604 * @itr: new ITR value to program
2605 *
2606 * e1000e_write_itr determines if the adapter is in MSI-X mode
2607 * and, if so, writes the EITR registers with the ITR value.
2608 * Otherwise, it writes the ITR value into the ITR register.
2609 **/
2610void e1000e_write_itr(struct e1000_adapter *adapter, u32 itr)
2611{
2612	struct e1000_hw *hw = &adapter->hw;
2613	u32 new_itr = itr ? 1000000000 / (itr * 256) : 0;
2614
2615	if (adapter->msix_entries) {
2616		int vector;
2617
2618		for (vector = 0; vector < adapter->num_vectors; vector++)
2619			writel(new_itr, hw->hw_addr + E1000_EITR_82574(vector));
2620	} else {
2621		ew32(ITR, new_itr);
2622	}
2623}
2624
2625/**
2626 * e1000_alloc_queues - Allocate memory for all rings
2627 * @adapter: board private structure to initialize
2628 **/
2629static int e1000_alloc_queues(struct e1000_adapter *adapter)
2630{
2631	int size = sizeof(struct e1000_ring);
2632
2633	adapter->tx_ring = kzalloc(size, GFP_KERNEL);
2634	if (!adapter->tx_ring)
2635		goto err;
2636	adapter->tx_ring->count = adapter->tx_ring_count;
2637	adapter->tx_ring->adapter = adapter;
2638
2639	adapter->rx_ring = kzalloc(size, GFP_KERNEL);
2640	if (!adapter->rx_ring)
2641		goto err;
2642	adapter->rx_ring->count = adapter->rx_ring_count;
2643	adapter->rx_ring->adapter = adapter;
2644
2645	return 0;
2646err:
2647	e_err("Unable to allocate memory for queues\n");
2648	kfree(adapter->rx_ring);
2649	kfree(adapter->tx_ring);
2650	return -ENOMEM;
2651}
2652
2653/**
2654 * e1000e_poll - NAPI Rx polling callback
2655 * @napi: struct associated with this polling callback
2656 * @budget: number of packets driver is allowed to process this poll
2657 **/
2658static int e1000e_poll(struct napi_struct *napi, int budget)
2659{
2660	struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter,
2661						     napi);
2662	struct e1000_hw *hw = &adapter->hw;
2663	struct net_device *poll_dev = adapter->netdev;
2664	int tx_cleaned = 1, work_done = 0;
2665
2666	adapter = netdev_priv(poll_dev);
2667
2668	if (!adapter->msix_entries ||
2669	    (adapter->rx_ring->ims_val & adapter->tx_ring->ims_val))
2670		tx_cleaned = e1000_clean_tx_irq(adapter->tx_ring);
2671
2672	adapter->clean_rx(adapter->rx_ring, &work_done, budget);
2673
2674	if (!tx_cleaned || work_done == budget)
2675		return budget;
2676
2677	/* Exit the polling mode, but don't re-enable interrupts if stack might
2678	 * poll us due to busy-polling
2679	 */
2680	if (likely(napi_complete_done(napi, work_done))) {
2681		if (adapter->itr_setting & 3)
2682			e1000_set_itr(adapter);
2683		if (!test_bit(__E1000_DOWN, &adapter->state)) {
2684			if (adapter->msix_entries)
2685				ew32(IMS, adapter->rx_ring->ims_val);
2686			else
2687				e1000_irq_enable(adapter);
2688		}
2689	}
2690
2691	return work_done;
2692}
2693
2694static int e1000_vlan_rx_add_vid(struct net_device *netdev,
2695				 __always_unused __be16 proto, u16 vid)
2696{
2697	struct e1000_adapter *adapter = netdev_priv(netdev);
2698	struct e1000_hw *hw = &adapter->hw;
2699	u32 vfta, index;
2700
2701	/* don't update vlan cookie if already programmed */
2702	if ((adapter->hw.mng_cookie.status &
2703	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2704	    (vid == adapter->mng_vlan_id))
2705		return 0;
2706
2707	/* add VID to filter table */
2708	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2709		index = (vid >> 5) & 0x7F;
2710		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2711		vfta |= BIT((vid & 0x1F));
2712		hw->mac.ops.write_vfta(hw, index, vfta);
2713	}
2714
2715	set_bit(vid, adapter->active_vlans);
2716
2717	return 0;
2718}
2719
2720static int e1000_vlan_rx_kill_vid(struct net_device *netdev,
2721				  __always_unused __be16 proto, u16 vid)
2722{
2723	struct e1000_adapter *adapter = netdev_priv(netdev);
2724	struct e1000_hw *hw = &adapter->hw;
2725	u32 vfta, index;
2726
2727	if ((adapter->hw.mng_cookie.status &
2728	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2729	    (vid == adapter->mng_vlan_id)) {
2730		/* release control to f/w */
2731		e1000e_release_hw_control(adapter);
2732		return 0;
2733	}
2734
2735	/* remove VID from filter table */
2736	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2737		index = (vid >> 5) & 0x7F;
2738		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2739		vfta &= ~BIT((vid & 0x1F));
2740		hw->mac.ops.write_vfta(hw, index, vfta);
2741	}
2742
2743	clear_bit(vid, adapter->active_vlans);
2744
2745	return 0;
2746}
2747
2748/**
2749 * e1000e_vlan_filter_disable - helper to disable hw VLAN filtering
2750 * @adapter: board private structure to initialize
2751 **/
2752static void e1000e_vlan_filter_disable(struct e1000_adapter *adapter)
2753{
2754	struct net_device *netdev = adapter->netdev;
2755	struct e1000_hw *hw = &adapter->hw;
2756	u32 rctl;
2757
2758	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2759		/* disable VLAN receive filtering */
2760		rctl = er32(RCTL);
2761		rctl &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
2762		ew32(RCTL, rctl);
2763
2764		if (adapter->mng_vlan_id != (u16)E1000_MNG_VLAN_NONE) {
2765			e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
2766					       adapter->mng_vlan_id);
2767			adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
2768		}
2769	}
2770}
2771
2772/**
2773 * e1000e_vlan_filter_enable - helper to enable HW VLAN filtering
2774 * @adapter: board private structure to initialize
2775 **/
2776static void e1000e_vlan_filter_enable(struct e1000_adapter *adapter)
2777{
2778	struct e1000_hw *hw = &adapter->hw;
2779	u32 rctl;
2780
2781	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2782		/* enable VLAN receive filtering */
2783		rctl = er32(RCTL);
2784		rctl |= E1000_RCTL_VFE;
2785		rctl &= ~E1000_RCTL_CFIEN;
2786		ew32(RCTL, rctl);
2787	}
2788}
2789
2790/**
2791 * e1000e_vlan_strip_disable - helper to disable HW VLAN stripping
2792 * @adapter: board private structure to initialize
2793 **/
2794static void e1000e_vlan_strip_disable(struct e1000_adapter *adapter)
2795{
2796	struct e1000_hw *hw = &adapter->hw;
2797	u32 ctrl;
2798
2799	/* disable VLAN tag insert/strip */
2800	ctrl = er32(CTRL);
2801	ctrl &= ~E1000_CTRL_VME;
2802	ew32(CTRL, ctrl);
2803}
2804
2805/**
2806 * e1000e_vlan_strip_enable - helper to enable HW VLAN stripping
2807 * @adapter: board private structure to initialize
2808 **/
2809static void e1000e_vlan_strip_enable(struct e1000_adapter *adapter)
2810{
2811	struct e1000_hw *hw = &adapter->hw;
2812	u32 ctrl;
2813
2814	/* enable VLAN tag insert/strip */
2815	ctrl = er32(CTRL);
2816	ctrl |= E1000_CTRL_VME;
2817	ew32(CTRL, ctrl);
2818}
2819
2820static void e1000_update_mng_vlan(struct e1000_adapter *adapter)
2821{
2822	struct net_device *netdev = adapter->netdev;
2823	u16 vid = adapter->hw.mng_cookie.vlan_id;
2824	u16 old_vid = adapter->mng_vlan_id;
2825
2826	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
2827		e1000_vlan_rx_add_vid(netdev, htons(ETH_P_8021Q), vid);
2828		adapter->mng_vlan_id = vid;
2829	}
2830
2831	if ((old_vid != (u16)E1000_MNG_VLAN_NONE) && (vid != old_vid))
2832		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q), old_vid);
2833}
2834
2835static void e1000_restore_vlan(struct e1000_adapter *adapter)
2836{
2837	u16 vid;
2838
2839	e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
2840
2841	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
2842	    e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
2843}
2844
2845static void e1000_init_manageability_pt(struct e1000_adapter *adapter)
2846{
2847	struct e1000_hw *hw = &adapter->hw;
2848	u32 manc, manc2h, mdef, i, j;
2849
2850	if (!(adapter->flags & FLAG_MNG_PT_ENABLED))
2851		return;
2852
2853	manc = er32(MANC);
2854
2855	/* enable receiving management packets to the host. this will probably
2856	 * generate destination unreachable messages from the host OS, but
2857	 * the packets will be handled on SMBUS
2858	 */
2859	manc |= E1000_MANC_EN_MNG2HOST;
2860	manc2h = er32(MANC2H);
2861
2862	switch (hw->mac.type) {
2863	default:
2864		manc2h |= (E1000_MANC2H_PORT_623 | E1000_MANC2H_PORT_664);
2865		break;
2866	case e1000_82574:
2867	case e1000_82583:
2868		/* Check if IPMI pass-through decision filter already exists;
2869		 * if so, enable it.
2870		 */
2871		for (i = 0, j = 0; i < 8; i++) {
2872			mdef = er32(MDEF(i));
2873
2874			/* Ignore filters with anything other than IPMI ports */
2875			if (mdef & ~(E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2876				continue;
2877
2878			/* Enable this decision filter in MANC2H */
2879			if (mdef)
2880				manc2h |= BIT(i);
2881
2882			j |= mdef;
2883		}
2884
2885		if (j == (E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2886			break;
2887
2888		/* Create new decision filter in an empty filter */
2889		for (i = 0, j = 0; i < 8; i++)
2890			if (er32(MDEF(i)) == 0) {
2891				ew32(MDEF(i), (E1000_MDEF_PORT_623 |
2892					       E1000_MDEF_PORT_664));
2893				manc2h |= BIT(1);
2894				j++;
2895				break;
2896			}
2897
2898		if (!j)
2899			e_warn("Unable to create IPMI pass-through filter\n");
2900		break;
2901	}
2902
2903	ew32(MANC2H, manc2h);
2904	ew32(MANC, manc);
2905}
2906
2907/**
2908 * e1000_configure_tx - Configure Transmit Unit after Reset
2909 * @adapter: board private structure
2910 *
2911 * Configure the Tx unit of the MAC after a reset.
2912 **/
2913static void e1000_configure_tx(struct e1000_adapter *adapter)
2914{
2915	struct e1000_hw *hw = &adapter->hw;
2916	struct e1000_ring *tx_ring = adapter->tx_ring;
2917	u64 tdba;
2918	u32 tdlen, tctl, tarc;
2919
2920	/* Setup the HW Tx Head and Tail descriptor pointers */
2921	tdba = tx_ring->dma;
2922	tdlen = tx_ring->count * sizeof(struct e1000_tx_desc);
2923	ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
2924	ew32(TDBAH(0), (tdba >> 32));
2925	ew32(TDLEN(0), tdlen);
2926	ew32(TDH(0), 0);
2927	ew32(TDT(0), 0);
2928	tx_ring->head = adapter->hw.hw_addr + E1000_TDH(0);
2929	tx_ring->tail = adapter->hw.hw_addr + E1000_TDT(0);
2930
2931	writel(0, tx_ring->head);
2932	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
2933		e1000e_update_tdt_wa(tx_ring, 0);
2934	else
2935		writel(0, tx_ring->tail);
2936
2937	/* Set the Tx Interrupt Delay register */
2938	ew32(TIDV, adapter->tx_int_delay);
2939	/* Tx irq moderation */
2940	ew32(TADV, adapter->tx_abs_int_delay);
2941
2942	if (adapter->flags2 & FLAG2_DMA_BURST) {
2943		u32 txdctl = er32(TXDCTL(0));
2944
2945		txdctl &= ~(E1000_TXDCTL_PTHRESH | E1000_TXDCTL_HTHRESH |
2946			    E1000_TXDCTL_WTHRESH);
2947		/* set up some performance related parameters to encourage the
2948		 * hardware to use the bus more efficiently in bursts, depends
2949		 * on the tx_int_delay to be enabled,
2950		 * wthresh = 1 ==> burst write is disabled to avoid Tx stalls
2951		 * hthresh = 1 ==> prefetch when one or more available
2952		 * pthresh = 0x1f ==> prefetch if internal cache 31 or less
2953		 * BEWARE: this seems to work but should be considered first if
2954		 * there are Tx hangs or other Tx related bugs
2955		 */
2956		txdctl |= E1000_TXDCTL_DMA_BURST_ENABLE;
2957		ew32(TXDCTL(0), txdctl);
2958	}
2959	/* erratum work around: set txdctl the same for both queues */
2960	ew32(TXDCTL(1), er32(TXDCTL(0)));
2961
2962	/* Program the Transmit Control Register */
2963	tctl = er32(TCTL);
2964	tctl &= ~E1000_TCTL_CT;
2965	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
2966		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
2967
2968	if (adapter->flags & FLAG_TARC_SPEED_MODE_BIT) {
2969		tarc = er32(TARC(0));
2970		/* set the speed mode bit, we'll clear it if we're not at
2971		 * gigabit link later
2972		 */
2973#define SPEED_MODE_BIT BIT(21)
2974		tarc |= SPEED_MODE_BIT;
2975		ew32(TARC(0), tarc);
2976	}
2977
2978	/* errata: program both queues to unweighted RR */
2979	if (adapter->flags & FLAG_TARC_SET_BIT_ZERO) {
2980		tarc = er32(TARC(0));
2981		tarc |= 1;
2982		ew32(TARC(0), tarc);
2983		tarc = er32(TARC(1));
2984		tarc |= 1;
2985		ew32(TARC(1), tarc);
2986	}
2987
2988	/* Setup Transmit Descriptor Settings for eop descriptor */
2989	adapter->txd_cmd = E1000_TXD_CMD_EOP | E1000_TXD_CMD_IFCS;
2990
2991	/* only set IDE if we are delaying interrupts using the timers */
2992	if (adapter->tx_int_delay)
2993		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
2994
2995	/* enable Report Status bit */
2996	adapter->txd_cmd |= E1000_TXD_CMD_RS;
2997
2998	ew32(TCTL, tctl);
2999
3000	hw->mac.ops.config_collision_dist(hw);
3001
3002	/* SPT and KBL Si errata workaround to avoid data corruption */
3003	if (hw->mac.type == e1000_pch_spt) {
3004		u32 reg_val;
3005
3006		reg_val = er32(IOSFPC);
3007		reg_val |= E1000_RCTL_RDMTS_HEX;
3008		ew32(IOSFPC, reg_val);
3009
3010		reg_val = er32(TARC(0));
3011		/* SPT and KBL Si errata workaround to avoid Tx hang.
3012		 * Dropping the number of outstanding requests from
3013		 * 3 to 2 in order to avoid a buffer overrun.
3014		 */
3015		reg_val &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3016		reg_val |= E1000_TARC0_CB_MULTIQ_2_REQ;
3017		ew32(TARC(0), reg_val);
3018	}
3019}
3020
3021#define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
3022			   (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
3023
3024/**
3025 * e1000_setup_rctl - configure the receive control registers
3026 * @adapter: Board private structure
3027 **/
3028static void e1000_setup_rctl(struct e1000_adapter *adapter)
3029{
3030	struct e1000_hw *hw = &adapter->hw;
3031	u32 rctl, rfctl;
3032	u32 pages = 0;
3033
3034	/* Workaround Si errata on PCHx - configure jumbo frame flow.
3035	 * If jumbo frames not set, program related MAC/PHY registers
3036	 * to h/w defaults
3037	 */
3038	if (hw->mac.type >= e1000_pch2lan) {
3039		s32 ret_val;
3040
3041		if (adapter->netdev->mtu > ETH_DATA_LEN)
3042			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, true);
3043		else
3044			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, false);
3045
3046		if (ret_val)
3047			e_dbg("failed to enable|disable jumbo frame workaround mode\n");
3048	}
3049
3050	/* Program MC offset vector base */
3051	rctl = er32(RCTL);
3052	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3053	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3054	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3055	    (adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3056
3057	/* Do not Store bad packets */
3058	rctl &= ~E1000_RCTL_SBP;
3059
3060	/* Enable Long Packet receive */
3061	if (adapter->netdev->mtu <= ETH_DATA_LEN)
3062		rctl &= ~E1000_RCTL_LPE;
3063	else
3064		rctl |= E1000_RCTL_LPE;
3065
3066	/* Some systems expect that the CRC is included in SMBUS traffic. The
3067	 * hardware strips the CRC before sending to both SMBUS (BMC) and to
3068	 * host memory when this is enabled
3069	 */
3070	if (adapter->flags2 & FLAG2_CRC_STRIPPING)
3071		rctl |= E1000_RCTL_SECRC;
3072
3073	/* Workaround Si errata on 82577 PHY - configure IPG for jumbos */
3074	if ((hw->phy.type == e1000_phy_82577) && (rctl & E1000_RCTL_LPE)) {
3075		u16 phy_data;
3076
3077		e1e_rphy(hw, PHY_REG(770, 26), &phy_data);
3078		phy_data &= 0xfff8;
3079		phy_data |= BIT(2);
3080		e1e_wphy(hw, PHY_REG(770, 26), phy_data);
3081
3082		e1e_rphy(hw, 22, &phy_data);
3083		phy_data &= 0x0fff;
3084		phy_data |= BIT(14);
3085		e1e_wphy(hw, 0x10, 0x2823);
3086		e1e_wphy(hw, 0x11, 0x0003);
3087		e1e_wphy(hw, 22, phy_data);
3088	}
3089
3090	/* Setup buffer sizes */
3091	rctl &= ~E1000_RCTL_SZ_4096;
3092	rctl |= E1000_RCTL_BSEX;
3093	switch (adapter->rx_buffer_len) {
3094	case 2048:
3095	default:
3096		rctl |= E1000_RCTL_SZ_2048;
3097		rctl &= ~E1000_RCTL_BSEX;
3098		break;
3099	case 4096:
3100		rctl |= E1000_RCTL_SZ_4096;
3101		break;
3102	case 8192:
3103		rctl |= E1000_RCTL_SZ_8192;
3104		break;
3105	case 16384:
3106		rctl |= E1000_RCTL_SZ_16384;
3107		break;
3108	}
3109
3110	/* Enable Extended Status in all Receive Descriptors */
3111	rfctl = er32(RFCTL);
3112	rfctl |= E1000_RFCTL_EXTEN;
3113	ew32(RFCTL, rfctl);
3114
3115	/* 82571 and greater support packet-split where the protocol
3116	 * header is placed in skb->data and the packet data is
3117	 * placed in pages hanging off of skb_shinfo(skb)->nr_frags.
3118	 * In the case of a non-split, skb->data is linearly filled,
3119	 * followed by the page buffers.  Therefore, skb->data is
3120	 * sized to hold the largest protocol header.
3121	 *
3122	 * allocations using alloc_page take too long for regular MTU
3123	 * so only enable packet split for jumbo frames
3124	 *
3125	 * Using pages when the page size is greater than 16k wastes
3126	 * a lot of memory, since we allocate 3 pages at all times
3127	 * per packet.
3128	 */
3129	pages = PAGE_USE_COUNT(adapter->netdev->mtu);
3130	if ((pages <= 3) && (PAGE_SIZE <= 16384) && (rctl & E1000_RCTL_LPE))
3131		adapter->rx_ps_pages = pages;
3132	else
3133		adapter->rx_ps_pages = 0;
3134
3135	if (adapter->rx_ps_pages) {
3136		u32 psrctl = 0;
3137
3138		/* Enable Packet split descriptors */
3139		rctl |= E1000_RCTL_DTYP_PS;
3140
3141		psrctl |= adapter->rx_ps_bsize0 >> E1000_PSRCTL_BSIZE0_SHIFT;
3142
3143		switch (adapter->rx_ps_pages) {
3144		case 3:
3145			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE3_SHIFT;
3146			fallthrough;
3147		case 2:
3148			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE2_SHIFT;
3149			fallthrough;
3150		case 1:
3151			psrctl |= PAGE_SIZE >> E1000_PSRCTL_BSIZE1_SHIFT;
3152			break;
3153		}
3154
3155		ew32(PSRCTL, psrctl);
3156	}
3157
3158	/* This is useful for sniffing bad packets. */
3159	if (adapter->netdev->features & NETIF_F_RXALL) {
3160		/* UPE and MPE will be handled by normal PROMISC logic
3161		 * in e1000e_set_rx_mode
3162		 */
3163		rctl |= (E1000_RCTL_SBP |	/* Receive bad packets */
3164			 E1000_RCTL_BAM |	/* RX All Bcast Pkts */
3165			 E1000_RCTL_PMCF);	/* RX All MAC Ctrl Pkts */
3166
3167		rctl &= ~(E1000_RCTL_VFE |	/* Disable VLAN filter */
3168			  E1000_RCTL_DPF |	/* Allow filtered pause */
3169			  E1000_RCTL_CFIEN);	/* Dis VLAN CFIEN Filter */
3170		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3171		 * and that breaks VLANs.
3172		 */
3173	}
3174
3175	ew32(RCTL, rctl);
3176	/* just started the receive unit, no need to restart */
3177	adapter->flags &= ~FLAG_RESTART_NOW;
3178}
3179
3180/**
3181 * e1000_configure_rx - Configure Receive Unit after Reset
3182 * @adapter: board private structure
3183 *
3184 * Configure the Rx unit of the MAC after a reset.
3185 **/
3186static void e1000_configure_rx(struct e1000_adapter *adapter)
3187{
3188	struct e1000_hw *hw = &adapter->hw;
3189	struct e1000_ring *rx_ring = adapter->rx_ring;
3190	u64 rdba;
3191	u32 rdlen, rctl, rxcsum, ctrl_ext;
3192
3193	if (adapter->rx_ps_pages) {
3194		/* this is a 32 byte descriptor */
3195		rdlen = rx_ring->count *
3196		    sizeof(union e1000_rx_desc_packet_split);
3197		adapter->clean_rx = e1000_clean_rx_irq_ps;
3198		adapter->alloc_rx_buf = e1000_alloc_rx_buffers_ps;
3199	} else if (adapter->netdev->mtu > ETH_FRAME_LEN + ETH_FCS_LEN) {
3200		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3201		adapter->clean_rx = e1000_clean_jumbo_rx_irq;
3202		adapter->alloc_rx_buf = e1000_alloc_jumbo_rx_buffers;
3203	} else {
3204		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3205		adapter->clean_rx = e1000_clean_rx_irq;
3206		adapter->alloc_rx_buf = e1000_alloc_rx_buffers;
3207	}
3208
3209	/* disable receives while setting up the descriptors */
3210	rctl = er32(RCTL);
3211	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
3212		ew32(RCTL, rctl & ~E1000_RCTL_EN);
3213	e1e_flush();
3214	usleep_range(10000, 11000);
3215
3216	if (adapter->flags2 & FLAG2_DMA_BURST) {
3217		/* set the writeback threshold (only takes effect if the RDTR
3218		 * is set). set GRAN=1 and write back up to 0x4 worth, and
3219		 * enable prefetching of 0x20 Rx descriptors
3220		 * granularity = 01
3221		 * wthresh = 04,
3222		 * hthresh = 04,
3223		 * pthresh = 0x20
3224		 */
3225		ew32(RXDCTL(0), E1000_RXDCTL_DMA_BURST_ENABLE);
3226		ew32(RXDCTL(1), E1000_RXDCTL_DMA_BURST_ENABLE);
3227	}
3228
3229	/* set the Receive Delay Timer Register */
3230	ew32(RDTR, adapter->rx_int_delay);
3231
3232	/* irq moderation */
3233	ew32(RADV, adapter->rx_abs_int_delay);
3234	if ((adapter->itr_setting != 0) && (adapter->itr != 0))
3235		e1000e_write_itr(adapter, adapter->itr);
3236
3237	ctrl_ext = er32(CTRL_EXT);
3238	/* Auto-Mask interrupts upon ICR access */
3239	ctrl_ext |= E1000_CTRL_EXT_IAME;
3240	ew32(IAM, 0xffffffff);
3241	ew32(CTRL_EXT, ctrl_ext);
3242	e1e_flush();
3243
3244	/* Setup the HW Rx Head and Tail Descriptor Pointers and
3245	 * the Base and Length of the Rx Descriptor Ring
3246	 */
3247	rdba = rx_ring->dma;
3248	ew32(RDBAL(0), (rdba & DMA_BIT_MASK(32)));
3249	ew32(RDBAH(0), (rdba >> 32));
3250	ew32(RDLEN(0), rdlen);
3251	ew32(RDH(0), 0);
3252	ew32(RDT(0), 0);
3253	rx_ring->head = adapter->hw.hw_addr + E1000_RDH(0);
3254	rx_ring->tail = adapter->hw.hw_addr + E1000_RDT(0);
3255
3256	writel(0, rx_ring->head);
3257	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
3258		e1000e_update_rdt_wa(rx_ring, 0);
3259	else
3260		writel(0, rx_ring->tail);
3261
3262	/* Enable Receive Checksum Offload for TCP and UDP */
3263	rxcsum = er32(RXCSUM);
3264	if (adapter->netdev->features & NETIF_F_RXCSUM)
3265		rxcsum |= E1000_RXCSUM_TUOFL;
3266	else
3267		rxcsum &= ~E1000_RXCSUM_TUOFL;
3268	ew32(RXCSUM, rxcsum);
3269
3270	/* With jumbo frames, excessive C-state transition latencies result
3271	 * in dropped transactions.
3272	 */
3273	if (adapter->netdev->mtu > ETH_DATA_LEN) {
3274		u32 lat =
3275		    ((er32(PBA) & E1000_PBA_RXA_MASK) * 1024 -
3276		     adapter->max_frame_size) * 8 / 1000;
3277
3278		if (adapter->flags & FLAG_IS_ICH) {
3279			u32 rxdctl = er32(RXDCTL(0));
3280
3281			ew32(RXDCTL(0), rxdctl | 0x3 | BIT(8));
3282		}
3283
3284		dev_info(&adapter->pdev->dev,
3285			 "Some CPU C-states have been disabled in order to enable jumbo frames\n");
3286		cpu_latency_qos_update_request(&adapter->pm_qos_req, lat);
3287	} else {
3288		cpu_latency_qos_update_request(&adapter->pm_qos_req,
3289					       PM_QOS_DEFAULT_VALUE);
3290	}
3291
3292	/* Enable Receives */
3293	ew32(RCTL, rctl);
3294}
3295
3296/**
3297 * e1000e_write_mc_addr_list - write multicast addresses to MTA
3298 * @netdev: network interface device structure
3299 *
3300 * Writes multicast address list to the MTA hash table.
3301 * Returns: -ENOMEM on failure
3302 *                0 on no addresses written
3303 *                X on writing X addresses to MTA
3304 */
3305static int e1000e_write_mc_addr_list(struct net_device *netdev)
3306{
3307	struct e1000_adapter *adapter = netdev_priv(netdev);
3308	struct e1000_hw *hw = &adapter->hw;
3309	struct netdev_hw_addr *ha;
3310	u8 *mta_list;
3311	int i;
3312
3313	if (netdev_mc_empty(netdev)) {
3314		/* nothing to program, so clear mc list */
3315		hw->mac.ops.update_mc_addr_list(hw, NULL, 0);
3316		return 0;
3317	}
3318
3319	mta_list = kcalloc(netdev_mc_count(netdev), ETH_ALEN, GFP_ATOMIC);
3320	if (!mta_list)
3321		return -ENOMEM;
3322
3323	/* update_mc_addr_list expects a packed array of only addresses. */
3324	i = 0;
3325	netdev_for_each_mc_addr(ha, netdev)
3326	    memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3327
3328	hw->mac.ops.update_mc_addr_list(hw, mta_list, i);
3329	kfree(mta_list);
3330
3331	return netdev_mc_count(netdev);
3332}
3333
3334/**
3335 * e1000e_write_uc_addr_list - write unicast addresses to RAR table
3336 * @netdev: network interface device structure
3337 *
3338 * Writes unicast address list to the RAR table.
3339 * Returns: -ENOMEM on failure/insufficient address space
3340 *                0 on no addresses written
3341 *                X on writing X addresses to the RAR table
3342 **/
3343static int e1000e_write_uc_addr_list(struct net_device *netdev)
3344{
3345	struct e1000_adapter *adapter = netdev_priv(netdev);
3346	struct e1000_hw *hw = &adapter->hw;
3347	unsigned int rar_entries;
3348	int count = 0;
3349
3350	rar_entries = hw->mac.ops.rar_get_count(hw);
3351
3352	/* save a rar entry for our hardware address */
3353	rar_entries--;
3354
3355	/* save a rar entry for the LAA workaround */
3356	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA)
3357		rar_entries--;
3358
3359	/* return ENOMEM indicating insufficient memory for addresses */
3360	if (netdev_uc_count(netdev) > rar_entries)
3361		return -ENOMEM;
3362
3363	if (!netdev_uc_empty(netdev) && rar_entries) {
3364		struct netdev_hw_addr *ha;
3365
3366		/* write the addresses in reverse order to avoid write
3367		 * combining
3368		 */
3369		netdev_for_each_uc_addr(ha, netdev) {
3370			int ret_val;
3371
3372			if (!rar_entries)
3373				break;
3374			ret_val = hw->mac.ops.rar_set(hw, ha->addr, rar_entries--);
3375			if (ret_val < 0)
3376				return -ENOMEM;
3377			count++;
3378		}
3379	}
3380
3381	/* zero out the remaining RAR entries not used above */
3382	for (; rar_entries > 0; rar_entries--) {
3383		ew32(RAH(rar_entries), 0);
3384		ew32(RAL(rar_entries), 0);
3385	}
3386	e1e_flush();
3387
3388	return count;
3389}
3390
3391/**
3392 * e1000e_set_rx_mode - secondary unicast, Multicast and Promiscuous mode set
3393 * @netdev: network interface device structure
3394 *
3395 * The ndo_set_rx_mode entry point is called whenever the unicast or multicast
3396 * address list or the network interface flags are updated.  This routine is
3397 * responsible for configuring the hardware for proper unicast, multicast,
3398 * promiscuous mode, and all-multi behavior.
3399 **/
3400static void e1000e_set_rx_mode(struct net_device *netdev)
3401{
3402	struct e1000_adapter *adapter = netdev_priv(netdev);
3403	struct e1000_hw *hw = &adapter->hw;
3404	u32 rctl;
3405
3406	if (pm_runtime_suspended(netdev->dev.parent))
3407		return;
3408
3409	/* Check for Promiscuous and All Multicast modes */
3410	rctl = er32(RCTL);
3411
3412	/* clear the affected bits */
3413	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
3414
3415	if (netdev->flags & IFF_PROMISC) {
3416		rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
3417		/* Do not hardware filter VLANs in promisc mode */
3418		e1000e_vlan_filter_disable(adapter);
3419	} else {
3420		int count;
3421
3422		if (netdev->flags & IFF_ALLMULTI) {
3423			rctl |= E1000_RCTL_MPE;
3424		} else {
3425			/* Write addresses to the MTA, if the attempt fails
3426			 * then we should just turn on promiscuous mode so
3427			 * that we can at least receive multicast traffic
3428			 */
3429			count = e1000e_write_mc_addr_list(netdev);
3430			if (count < 0)
3431				rctl |= E1000_RCTL_MPE;
3432		}
3433		e1000e_vlan_filter_enable(adapter);
3434		/* Write addresses to available RAR registers, if there is not
3435		 * sufficient space to store all the addresses then enable
3436		 * unicast promiscuous mode
3437		 */
3438		count = e1000e_write_uc_addr_list(netdev);
3439		if (count < 0)
3440			rctl |= E1000_RCTL_UPE;
3441	}
3442
3443	ew32(RCTL, rctl);
3444
3445	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
3446		e1000e_vlan_strip_enable(adapter);
3447	else
3448		e1000e_vlan_strip_disable(adapter);
3449}
3450
3451static void e1000e_setup_rss_hash(struct e1000_adapter *adapter)
3452{
3453	struct e1000_hw *hw = &adapter->hw;
3454	u32 mrqc, rxcsum;
3455	u32 rss_key[10];
3456	int i;
3457
3458	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3459	for (i = 0; i < 10; i++)
3460		ew32(RSSRK(i), rss_key[i]);
3461
3462	/* Direct all traffic to queue 0 */
3463	for (i = 0; i < 32; i++)
3464		ew32(RETA(i), 0);
3465
3466	/* Disable raw packet checksumming so that RSS hash is placed in
3467	 * descriptor on writeback.
3468	 */
3469	rxcsum = er32(RXCSUM);
3470	rxcsum |= E1000_RXCSUM_PCSD;
3471
3472	ew32(RXCSUM, rxcsum);
3473
3474	mrqc = (E1000_MRQC_RSS_FIELD_IPV4 |
3475		E1000_MRQC_RSS_FIELD_IPV4_TCP |
3476		E1000_MRQC_RSS_FIELD_IPV6 |
3477		E1000_MRQC_RSS_FIELD_IPV6_TCP |
3478		E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3479
3480	ew32(MRQC, mrqc);
3481}
3482
3483/**
3484 * e1000e_get_base_timinca - get default SYSTIM time increment attributes
3485 * @adapter: board private structure
3486 * @timinca: pointer to returned time increment attributes
3487 *
3488 * Get attributes for incrementing the System Time Register SYSTIML/H at
3489 * the default base frequency, and set the cyclecounter shift value.
3490 **/
3491s32 e1000e_get_base_timinca(struct e1000_adapter *adapter, u32 *timinca)
3492{
3493	struct e1000_hw *hw = &adapter->hw;
3494	u32 incvalue, incperiod, shift;
3495
3496	/* Make sure clock is enabled on I217/I218/I219  before checking
3497	 * the frequency
3498	 */
3499	if ((hw->mac.type >= e1000_pch_lpt) &&
3500	    !(er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) &&
3501	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_ENABLED)) {
3502		u32 fextnvm7 = er32(FEXTNVM7);
3503
3504		if (!(fextnvm7 & BIT(0))) {
3505			ew32(FEXTNVM7, fextnvm7 | BIT(0));
3506			e1e_flush();
3507		}
3508	}
3509
3510	switch (hw->mac.type) {
3511	case e1000_pch2lan:
3512		/* Stable 96MHz frequency */
3513		incperiod = INCPERIOD_96MHZ;
3514		incvalue = INCVALUE_96MHZ;
3515		shift = INCVALUE_SHIFT_96MHZ;
3516		adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3517		break;
3518	case e1000_pch_lpt:
3519		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3520			/* Stable 96MHz frequency */
3521			incperiod = INCPERIOD_96MHZ;
3522			incvalue = INCVALUE_96MHZ;
3523			shift = INCVALUE_SHIFT_96MHZ;
3524			adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3525		} else {
3526			/* Stable 25MHz frequency */
3527			incperiod = INCPERIOD_25MHZ;
3528			incvalue = INCVALUE_25MHZ;
3529			shift = INCVALUE_SHIFT_25MHZ;
3530			adapter->cc.shift = shift;
3531		}
3532		break;
3533	case e1000_pch_spt:
3534		/* Stable 24MHz frequency */
3535		incperiod = INCPERIOD_24MHZ;
3536		incvalue = INCVALUE_24MHZ;
3537		shift = INCVALUE_SHIFT_24MHZ;
3538		adapter->cc.shift = shift;
3539		break;
3540	case e1000_pch_cnp:
3541	case e1000_pch_tgp:
3542	case e1000_pch_adp:
3543	case e1000_pch_mtp:
3544	case e1000_pch_lnp:
3545	case e1000_pch_ptp:
3546	case e1000_pch_nvp:
3547		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3548			/* Stable 24MHz frequency */
3549			incperiod = INCPERIOD_24MHZ;
3550			incvalue = INCVALUE_24MHZ;
3551			shift = INCVALUE_SHIFT_24MHZ;
3552			adapter->cc.shift = shift;
3553		} else {
3554			/* Stable 38400KHz frequency */
3555			incperiod = INCPERIOD_38400KHZ;
3556			incvalue = INCVALUE_38400KHZ;
3557			shift = INCVALUE_SHIFT_38400KHZ;
3558			adapter->cc.shift = shift;
3559		}
3560		break;
3561	case e1000_82574:
3562	case e1000_82583:
3563		/* Stable 25MHz frequency */
3564		incperiod = INCPERIOD_25MHZ;
3565		incvalue = INCVALUE_25MHZ;
3566		shift = INCVALUE_SHIFT_25MHZ;
3567		adapter->cc.shift = shift;
3568		break;
3569	default:
3570		return -EINVAL;
3571	}
3572
3573	*timinca = ((incperiod << E1000_TIMINCA_INCPERIOD_SHIFT) |
3574		    ((incvalue << shift) & E1000_TIMINCA_INCVALUE_MASK));
3575
3576	return 0;
3577}
3578
3579/**
3580 * e1000e_config_hwtstamp - configure the hwtstamp registers and enable/disable
3581 * @adapter: board private structure
3582 * @config: timestamp configuration
3583 *
3584 * Outgoing time stamping can be enabled and disabled. Play nice and
3585 * disable it when requested, although it shouldn't cause any overhead
3586 * when no packet needs it. At most one packet in the queue may be
3587 * marked for time stamping, otherwise it would be impossible to tell
3588 * for sure to which packet the hardware time stamp belongs.
3589 *
3590 * Incoming time stamping has to be configured via the hardware filters.
3591 * Not all combinations are supported, in particular event type has to be
3592 * specified. Matching the kind of event packet is not supported, with the
3593 * exception of "all V2 events regardless of level 2 or 4".
3594 **/
3595static int e1000e_config_hwtstamp(struct e1000_adapter *adapter,
3596				  struct hwtstamp_config *config)
3597{
3598	struct e1000_hw *hw = &adapter->hw;
3599	u32 tsync_tx_ctl = E1000_TSYNCTXCTL_ENABLED;
3600	u32 tsync_rx_ctl = E1000_TSYNCRXCTL_ENABLED;
3601	u32 rxmtrl = 0;
3602	u16 rxudp = 0;
3603	bool is_l4 = false;
3604	bool is_l2 = false;
3605	u32 regval;
3606
3607	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3608		return -EINVAL;
3609
3610	switch (config->tx_type) {
3611	case HWTSTAMP_TX_OFF:
3612		tsync_tx_ctl = 0;
3613		break;
3614	case HWTSTAMP_TX_ON:
3615		break;
3616	default:
3617		return -ERANGE;
3618	}
3619
3620	switch (config->rx_filter) {
3621	case HWTSTAMP_FILTER_NONE:
3622		tsync_rx_ctl = 0;
3623		break;
3624	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3625		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3626		rxmtrl = E1000_RXMTRL_PTP_V1_SYNC_MESSAGE;
3627		is_l4 = true;
3628		break;
3629	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3630		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3631		rxmtrl = E1000_RXMTRL_PTP_V1_DELAY_REQ_MESSAGE;
3632		is_l4 = true;
3633		break;
3634	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3635		/* Also time stamps V2 L2 Path Delay Request/Response */
3636		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3637		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3638		is_l2 = true;
3639		break;
3640	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3641		/* Also time stamps V2 L2 Path Delay Request/Response. */
3642		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3643		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3644		is_l2 = true;
3645		break;
3646	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3647		/* Hardware cannot filter just V2 L4 Sync messages */
3648		fallthrough;
3649	case HWTSTAMP_FILTER_PTP_V2_SYNC:
3650		/* Also time stamps V2 Path Delay Request/Response. */
3651		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3652		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3653		is_l2 = true;
3654		is_l4 = true;
3655		break;
3656	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3657		/* Hardware cannot filter just V2 L4 Delay Request messages */
3658		fallthrough;
3659	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
3660		/* Also time stamps V2 Path Delay Request/Response. */
3661		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3662		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3663		is_l2 = true;
3664		is_l4 = true;
3665		break;
3666	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
3667	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
3668		/* Hardware cannot filter just V2 L4 or L2 Event messages */
3669		fallthrough;
3670	case HWTSTAMP_FILTER_PTP_V2_EVENT:
3671		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_EVENT_V2;
3672		config->rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
3673		is_l2 = true;
3674		is_l4 = true;
3675		break;
3676	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3677		/* For V1, the hardware can only filter Sync messages or
3678		 * Delay Request messages but not both so fall-through to
3679		 * time stamp all packets.
3680		 */
3681		fallthrough;
3682	case HWTSTAMP_FILTER_NTP_ALL:
3683	case HWTSTAMP_FILTER_ALL:
3684		is_l2 = true;
3685		is_l4 = true;
3686		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_ALL;
3687		config->rx_filter = HWTSTAMP_FILTER_ALL;
3688		break;
3689	default:
3690		return -ERANGE;
3691	}
3692
3693	adapter->hwtstamp_config = *config;
3694
3695	/* enable/disable Tx h/w time stamping */
3696	regval = er32(TSYNCTXCTL);
3697	regval &= ~E1000_TSYNCTXCTL_ENABLED;
3698	regval |= tsync_tx_ctl;
3699	ew32(TSYNCTXCTL, regval);
3700	if ((er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) !=
3701	    (regval & E1000_TSYNCTXCTL_ENABLED)) {
3702		e_err("Timesync Tx Control register not set as expected\n");
3703		return -EAGAIN;
3704	}
3705
3706	/* enable/disable Rx h/w time stamping */
3707	regval = er32(TSYNCRXCTL);
3708	regval &= ~(E1000_TSYNCRXCTL_ENABLED | E1000_TSYNCRXCTL_TYPE_MASK);
3709	regval |= tsync_rx_ctl;
3710	ew32(TSYNCRXCTL, regval);
3711	if ((er32(TSYNCRXCTL) & (E1000_TSYNCRXCTL_ENABLED |
3712				 E1000_TSYNCRXCTL_TYPE_MASK)) !=
3713	    (regval & (E1000_TSYNCRXCTL_ENABLED |
3714		       E1000_TSYNCRXCTL_TYPE_MASK))) {
3715		e_err("Timesync Rx Control register not set as expected\n");
3716		return -EAGAIN;
3717	}
3718
3719	/* L2: define ethertype filter for time stamped packets */
3720	if (is_l2)
3721		rxmtrl |= ETH_P_1588;
3722
3723	/* define which PTP packets get time stamped */
3724	ew32(RXMTRL, rxmtrl);
3725
3726	/* Filter by destination port */
3727	if (is_l4) {
3728		rxudp = PTP_EV_PORT;
3729		cpu_to_be16s(&rxudp);
3730	}
3731	ew32(RXUDP, rxudp);
3732
3733	e1e_flush();
3734
3735	/* Clear TSYNCRXCTL_VALID & TSYNCTXCTL_VALID bit */
3736	er32(RXSTMPH);
3737	er32(TXSTMPH);
3738
3739	return 0;
3740}
3741
3742/**
3743 * e1000_configure - configure the hardware for Rx and Tx
3744 * @adapter: private board structure
3745 **/
3746static void e1000_configure(struct e1000_adapter *adapter)
3747{
3748	struct e1000_ring *rx_ring = adapter->rx_ring;
3749
3750	e1000e_set_rx_mode(adapter->netdev);
3751
3752	e1000_restore_vlan(adapter);
3753	e1000_init_manageability_pt(adapter);
3754
3755	e1000_configure_tx(adapter);
3756
3757	if (adapter->netdev->features & NETIF_F_RXHASH)
3758		e1000e_setup_rss_hash(adapter);
3759	e1000_setup_rctl(adapter);
3760	e1000_configure_rx(adapter);
3761	adapter->alloc_rx_buf(rx_ring, e1000_desc_unused(rx_ring), GFP_KERNEL);
3762}
3763
3764/**
3765 * e1000e_power_up_phy - restore link in case the phy was powered down
3766 * @adapter: address of board private structure
3767 *
3768 * The phy may be powered down to save power and turn off link when the
3769 * driver is unloaded and wake on lan is not enabled (among others)
3770 * *** this routine MUST be followed by a call to e1000e_reset ***
3771 **/
3772void e1000e_power_up_phy(struct e1000_adapter *adapter)
3773{
3774	if (adapter->hw.phy.ops.power_up)
3775		adapter->hw.phy.ops.power_up(&adapter->hw);
3776
3777	adapter->hw.mac.ops.setup_link(&adapter->hw);
3778}
3779
3780/**
3781 * e1000_power_down_phy - Power down the PHY
3782 * @adapter: board private structure
3783 *
3784 * Power down the PHY so no link is implied when interface is down.
3785 * The PHY cannot be powered down if management or WoL is active.
3786 */
3787static void e1000_power_down_phy(struct e1000_adapter *adapter)
3788{
3789	if (adapter->hw.phy.ops.power_down)
3790		adapter->hw.phy.ops.power_down(&adapter->hw);
3791}
3792
3793/**
3794 * e1000_flush_tx_ring - remove all descriptors from the tx_ring
3795 * @adapter: board private structure
3796 *
3797 * We want to clear all pending descriptors from the TX ring.
3798 * zeroing happens when the HW reads the regs. We  assign the ring itself as
3799 * the data of the next descriptor. We don't care about the data we are about
3800 * to reset the HW.
3801 */
3802static void e1000_flush_tx_ring(struct e1000_adapter *adapter)
3803{
3804	struct e1000_hw *hw = &adapter->hw;
3805	struct e1000_ring *tx_ring = adapter->tx_ring;
3806	struct e1000_tx_desc *tx_desc = NULL;
3807	u32 tdt, tctl, txd_lower = E1000_TXD_CMD_IFCS;
3808	u16 size = 512;
3809
3810	tctl = er32(TCTL);
3811	ew32(TCTL, tctl | E1000_TCTL_EN);
3812	tdt = er32(TDT(0));
3813	BUG_ON(tdt != tx_ring->next_to_use);
3814	tx_desc =  E1000_TX_DESC(*tx_ring, tx_ring->next_to_use);
3815	tx_desc->buffer_addr = cpu_to_le64(tx_ring->dma);
3816
3817	tx_desc->lower.data = cpu_to_le32(txd_lower | size);
3818	tx_desc->upper.data = 0;
3819	/* flush descriptors to memory before notifying the HW */
3820	wmb();
3821	tx_ring->next_to_use++;
3822	if (tx_ring->next_to_use == tx_ring->count)
3823		tx_ring->next_to_use = 0;
3824	ew32(TDT(0), tx_ring->next_to_use);
3825	usleep_range(200, 250);
3826}
3827
3828/**
3829 * e1000_flush_rx_ring - remove all descriptors from the rx_ring
3830 * @adapter: board private structure
3831 *
3832 * Mark all descriptors in the RX ring as consumed and disable the rx ring
3833 */
3834static void e1000_flush_rx_ring(struct e1000_adapter *adapter)
3835{
3836	u32 rctl, rxdctl;
3837	struct e1000_hw *hw = &adapter->hw;
3838
3839	rctl = er32(RCTL);
3840	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3841	e1e_flush();
3842	usleep_range(100, 150);
3843
3844	rxdctl = er32(RXDCTL(0));
3845	/* zero the lower 14 bits (prefetch and host thresholds) */
3846	rxdctl &= 0xffffc000;
3847
3848	/* update thresholds: prefetch threshold to 31, host threshold to 1
3849	 * and make sure the granularity is "descriptors" and not "cache lines"
3850	 */
3851	rxdctl |= (0x1F | BIT(8) | E1000_RXDCTL_THRESH_UNIT_DESC);
3852
3853	ew32(RXDCTL(0), rxdctl);
3854	/* momentarily enable the RX ring for the changes to take effect */
3855	ew32(RCTL, rctl | E1000_RCTL_EN);
3856	e1e_flush();
3857	usleep_range(100, 150);
3858	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3859}
3860
3861/**
3862 * e1000_flush_desc_rings - remove all descriptors from the descriptor rings
3863 * @adapter: board private structure
3864 *
3865 * In i219, the descriptor rings must be emptied before resetting the HW
3866 * or before changing the device state to D3 during runtime (runtime PM).
3867 *
3868 * Failure to do this will cause the HW to enter a unit hang state which can
3869 * only be released by PCI reset on the device
3870 *
3871 */
3872
3873static void e1000_flush_desc_rings(struct e1000_adapter *adapter)
3874{
3875	u16 hang_state;
3876	u32 fext_nvm11, tdlen;
3877	struct e1000_hw *hw = &adapter->hw;
3878
3879	/* First, disable MULR fix in FEXTNVM11 */
3880	fext_nvm11 = er32(FEXTNVM11);
3881	fext_nvm11 |= E1000_FEXTNVM11_DISABLE_MULR_FIX;
3882	ew32(FEXTNVM11, fext_nvm11);
3883	/* do nothing if we're not in faulty state, or if the queue is empty */
3884	tdlen = er32(TDLEN(0));
3885	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3886			     &hang_state);
3887	if (!(hang_state & FLUSH_DESC_REQUIRED) || !tdlen)
3888		return;
3889	e1000_flush_tx_ring(adapter);
3890	/* recheck, maybe the fault is caused by the rx ring */
3891	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3892			     &hang_state);
3893	if (hang_state & FLUSH_DESC_REQUIRED)
3894		e1000_flush_rx_ring(adapter);
3895}
3896
3897/**
3898 * e1000e_systim_reset - reset the timesync registers after a hardware reset
3899 * @adapter: board private structure
3900 *
3901 * When the MAC is reset, all hardware bits for timesync will be reset to the
3902 * default values. This function will restore the settings last in place.
3903 * Since the clock SYSTIME registers are reset, we will simply restore the
3904 * cyclecounter to the kernel real clock time.
3905 **/
3906static void e1000e_systim_reset(struct e1000_adapter *adapter)
3907{
3908	struct ptp_clock_info *info = &adapter->ptp_clock_info;
3909	struct e1000_hw *hw = &adapter->hw;
3910	unsigned long flags;
3911	u32 timinca;
3912	s32 ret_val;
3913
3914	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3915		return;
3916
3917	if (info->adjfine) {
3918		/* restore the previous ptp frequency delta */
3919		ret_val = info->adjfine(info, adapter->ptp_delta);
3920	} else {
3921		/* set the default base frequency if no adjustment possible */
3922		ret_val = e1000e_get_base_timinca(adapter, &timinca);
3923		if (!ret_val)
3924			ew32(TIMINCA, timinca);
3925	}
3926
3927	if (ret_val) {
3928		dev_warn(&adapter->pdev->dev,
3929			 "Failed to restore TIMINCA clock rate delta: %d\n",
3930			 ret_val);
3931		return;
3932	}
3933
3934	/* reset the systim ns time counter */
3935	spin_lock_irqsave(&adapter->systim_lock, flags);
3936	timecounter_init(&adapter->tc, &adapter->cc,
3937			 ktime_to_ns(ktime_get_real()));
3938	spin_unlock_irqrestore(&adapter->systim_lock, flags);
3939
3940	/* restore the previous hwtstamp configuration settings */
3941	e1000e_config_hwtstamp(adapter, &adapter->hwtstamp_config);
3942}
3943
3944/**
3945 * e1000e_reset - bring the hardware into a known good state
3946 * @adapter: board private structure
3947 *
3948 * This function boots the hardware and enables some settings that
3949 * require a configuration cycle of the hardware - those cannot be
3950 * set/changed during runtime. After reset the device needs to be
3951 * properly configured for Rx, Tx etc.
3952 */
3953void e1000e_reset(struct e1000_adapter *adapter)
3954{
3955	struct e1000_mac_info *mac = &adapter->hw.mac;
3956	struct e1000_fc_info *fc = &adapter->hw.fc;
3957	struct e1000_hw *hw = &adapter->hw;
3958	u32 tx_space, min_tx_space, min_rx_space;
3959	u32 pba = adapter->pba;
3960	u16 hwm;
3961
3962	/* reset Packet Buffer Allocation to default */
3963	ew32(PBA, pba);
3964
3965	if (adapter->max_frame_size > (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN)) {
3966		/* To maintain wire speed transmits, the Tx FIFO should be
3967		 * large enough to accommodate two full transmit packets,
3968		 * rounded up to the next 1KB and expressed in KB.  Likewise,
3969		 * the Rx FIFO should be large enough to accommodate at least
3970		 * one full receive packet and is similarly rounded up and
3971		 * expressed in KB.
3972		 */
3973		pba = er32(PBA);
3974		/* upper 16 bits has Tx packet buffer allocation size in KB */
3975		tx_space = pba >> 16;
3976		/* lower 16 bits has Rx packet buffer allocation size in KB */
3977		pba &= 0xffff;
3978		/* the Tx fifo also stores 16 bytes of information about the Tx
3979		 * but don't include ethernet FCS because hardware appends it
3980		 */
3981		min_tx_space = (adapter->max_frame_size +
3982				sizeof(struct e1000_tx_desc) - ETH_FCS_LEN) * 2;
3983		min_tx_space = ALIGN(min_tx_space, 1024);
3984		min_tx_space >>= 10;
3985		/* software strips receive CRC, so leave room for it */
3986		min_rx_space = adapter->max_frame_size;
3987		min_rx_space = ALIGN(min_rx_space, 1024);
3988		min_rx_space >>= 10;
3989
3990		/* If current Tx allocation is less than the min Tx FIFO size,
3991		 * and the min Tx FIFO size is less than the current Rx FIFO
3992		 * allocation, take space away from current Rx allocation
3993		 */
3994		if ((tx_space < min_tx_space) &&
3995		    ((min_tx_space - tx_space) < pba)) {
3996			pba -= min_tx_space - tx_space;
3997
3998			/* if short on Rx space, Rx wins and must trump Tx
3999			 * adjustment
4000			 */
4001			if (pba < min_rx_space)
4002				pba = min_rx_space;
4003		}
4004
4005		ew32(PBA, pba);
4006	}
4007
4008	/* flow control settings
4009	 *
4010	 * The high water mark must be low enough to fit one full frame
4011	 * (or the size used for early receive) above it in the Rx FIFO.
4012	 * Set it to the lower of:
4013	 * - 90% of the Rx FIFO size, and
4014	 * - the full Rx FIFO size minus one full frame
4015	 */
4016	if (adapter->flags & FLAG_DISABLE_FC_PAUSE_TIME)
4017		fc->pause_time = 0xFFFF;
4018	else
4019		fc->pause_time = E1000_FC_PAUSE_TIME;
4020	fc->send_xon = true;
4021	fc->current_mode = fc->requested_mode;
4022
4023	switch (hw->mac.type) {
4024	case e1000_ich9lan:
4025	case e1000_ich10lan:
4026		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4027			pba = 14;
4028			ew32(PBA, pba);
4029			fc->high_water = 0x2800;
4030			fc->low_water = fc->high_water - 8;
4031			break;
4032		}
4033		fallthrough;
4034	default:
4035		hwm = min(((pba << 10) * 9 / 10),
4036			  ((pba << 10) - adapter->max_frame_size));
4037
4038		fc->high_water = hwm & E1000_FCRTH_RTH;	/* 8-byte granularity */
4039		fc->low_water = fc->high_water - 8;
4040		break;
4041	case e1000_pchlan:
4042		/* Workaround PCH LOM adapter hangs with certain network
4043		 * loads.  If hangs persist, try disabling Tx flow control.
4044		 */
4045		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4046			fc->high_water = 0x3500;
4047			fc->low_water = 0x1500;
4048		} else {
4049			fc->high_water = 0x5000;
4050			fc->low_water = 0x3000;
4051		}
4052		fc->refresh_time = 0x1000;
4053		break;
4054	case e1000_pch2lan:
4055	case e1000_pch_lpt:
4056	case e1000_pch_spt:
4057	case e1000_pch_cnp:
4058	case e1000_pch_tgp:
4059	case e1000_pch_adp:
4060	case e1000_pch_mtp:
4061	case e1000_pch_lnp:
4062	case e1000_pch_ptp:
4063	case e1000_pch_nvp:
4064		fc->refresh_time = 0xFFFF;
4065		fc->pause_time = 0xFFFF;
4066
4067		if (adapter->netdev->mtu <= ETH_DATA_LEN) {
4068			fc->high_water = 0x05C20;
4069			fc->low_water = 0x05048;
4070			break;
4071		}
4072
4073		pba = 14;
4074		ew32(PBA, pba);
4075		fc->high_water = ((pba << 10) * 9 / 10) & E1000_FCRTH_RTH;
4076		fc->low_water = ((pba << 10) * 8 / 10) & E1000_FCRTL_RTL;
4077		break;
4078	}
4079
4080	/* Alignment of Tx data is on an arbitrary byte boundary with the
4081	 * maximum size per Tx descriptor limited only to the transmit
4082	 * allocation of the packet buffer minus 96 bytes with an upper
4083	 * limit of 24KB due to receive synchronization limitations.
4084	 */
4085	adapter->tx_fifo_limit = min_t(u32, ((er32(PBA) >> 16) << 10) - 96,
4086				       24 << 10);
4087
4088	/* Disable Adaptive Interrupt Moderation if 2 full packets cannot
4089	 * fit in receive buffer.
4090	 */
4091	if (adapter->itr_setting & 0x3) {
4092		if ((adapter->max_frame_size * 2) > (pba << 10)) {
4093			if (!(adapter->flags2 & FLAG2_DISABLE_AIM)) {
4094				dev_info(&adapter->pdev->dev,
4095					 "Interrupt Throttle Rate off\n");
4096				adapter->flags2 |= FLAG2_DISABLE_AIM;
4097				e1000e_write_itr(adapter, 0);
4098			}
4099		} else if (adapter->flags2 & FLAG2_DISABLE_AIM) {
4100			dev_info(&adapter->pdev->dev,
4101				 "Interrupt Throttle Rate on\n");
4102			adapter->flags2 &= ~FLAG2_DISABLE_AIM;
4103			adapter->itr = 20000;
4104			e1000e_write_itr(adapter, adapter->itr);
4105		}
4106	}
4107
4108	if (hw->mac.type >= e1000_pch_spt)
4109		e1000_flush_desc_rings(adapter);
4110	/* Allow time for pending master requests to run */
4111	mac->ops.reset_hw(hw);
4112
4113	/* For parts with AMT enabled, let the firmware know
4114	 * that the network interface is in control
4115	 */
4116	if (adapter->flags & FLAG_HAS_AMT)
4117		e1000e_get_hw_control(adapter);
4118
4119	ew32(WUC, 0);
4120
4121	if (mac->ops.init_hw(hw))
4122		e_err("Hardware Error\n");
4123
4124	e1000_update_mng_vlan(adapter);
4125
4126	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
4127	ew32(VET, ETH_P_8021Q);
4128
4129	e1000e_reset_adaptive(hw);
4130
4131	/* restore systim and hwtstamp settings */
4132	e1000e_systim_reset(adapter);
4133
4134	/* Set EEE advertisement as appropriate */
4135	if (adapter->flags2 & FLAG2_HAS_EEE) {
4136		s32 ret_val;
4137		u16 adv_addr;
4138
4139		switch (hw->phy.type) {
4140		case e1000_phy_82579:
4141			adv_addr = I82579_EEE_ADVERTISEMENT;
4142			break;
4143		case e1000_phy_i217:
4144			adv_addr = I217_EEE_ADVERTISEMENT;
4145			break;
4146		default:
4147			dev_err(&adapter->pdev->dev,
4148				"Invalid PHY type setting EEE advertisement\n");
4149			return;
4150		}
4151
4152		ret_val = hw->phy.ops.acquire(hw);
4153		if (ret_val) {
4154			dev_err(&adapter->pdev->dev,
4155				"EEE advertisement - unable to acquire PHY\n");
4156			return;
4157		}
4158
4159		e1000_write_emi_reg_locked(hw, adv_addr,
4160					   hw->dev_spec.ich8lan.eee_disable ?
4161					   0 : adapter->eee_advert);
4162
4163		hw->phy.ops.release(hw);
4164	}
4165
4166	if (!netif_running(adapter->netdev) &&
4167	    !test_bit(__E1000_TESTING, &adapter->state))
4168		e1000_power_down_phy(adapter);
4169
4170	e1000_get_phy_info(hw);
4171
4172	if ((adapter->flags & FLAG_HAS_SMART_POWER_DOWN) &&
4173	    !(adapter->flags & FLAG_SMART_POWER_DOWN)) {
4174		u16 phy_data = 0;
4175		/* speed up time to link by disabling smart power down, ignore
4176		 * the return value of this function because there is nothing
4177		 * different we would do if it failed
4178		 */
4179		e1e_rphy(hw, IGP02E1000_PHY_POWER_MGMT, &phy_data);
4180		phy_data &= ~IGP02E1000_PM_SPD;
4181		e1e_wphy(hw, IGP02E1000_PHY_POWER_MGMT, phy_data);
4182	}
4183	if (hw->mac.type >= e1000_pch_spt && adapter->int_mode == 0) {
4184		u32 reg;
4185
4186		/* Fextnvm7 @ 0xe4[2] = 1 */
4187		reg = er32(FEXTNVM7);
4188		reg |= E1000_FEXTNVM7_SIDE_CLK_UNGATE;
4189		ew32(FEXTNVM7, reg);
4190		/* Fextnvm9 @ 0x5bb4[13:12] = 11 */
4191		reg = er32(FEXTNVM9);
4192		reg |= E1000_FEXTNVM9_IOSFSB_CLKGATE_DIS |
4193		       E1000_FEXTNVM9_IOSFSB_CLKREQ_DIS;
4194		ew32(FEXTNVM9, reg);
4195	}
4196
4197}
4198
4199/**
4200 * e1000e_trigger_lsc - trigger an LSC interrupt
4201 * @adapter: board private structure
4202 *
4203 * Fire a link status change interrupt to start the watchdog.
4204 **/
4205static void e1000e_trigger_lsc(struct e1000_adapter *adapter)
4206{
4207	struct e1000_hw *hw = &adapter->hw;
4208
4209	if (adapter->msix_entries)
4210		ew32(ICS, E1000_ICS_LSC | E1000_ICS_OTHER);
4211	else
4212		ew32(ICS, E1000_ICS_LSC);
4213}
4214
4215void e1000e_up(struct e1000_adapter *adapter)
4216{
4217	/* hardware has been reset, we need to reload some things */
4218	e1000_configure(adapter);
4219
4220	clear_bit(__E1000_DOWN, &adapter->state);
4221
4222	if (adapter->msix_entries)
4223		e1000_configure_msix(adapter);
4224	e1000_irq_enable(adapter);
4225
4226	/* Tx queue started by watchdog timer when link is up */
4227
4228	e1000e_trigger_lsc(adapter);
4229}
4230
4231static void e1000e_flush_descriptors(struct e1000_adapter *adapter)
4232{
4233	struct e1000_hw *hw = &adapter->hw;
4234
4235	if (!(adapter->flags2 & FLAG2_DMA_BURST))
4236		return;
4237
4238	/* flush pending descriptor writebacks to memory */
4239	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4240	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4241
4242	/* execute the writes immediately */
4243	e1e_flush();
4244
4245	/* due to rare timing issues, write to TIDV/RDTR again to ensure the
4246	 * write is successful
4247	 */
4248	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4249	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4250
4251	/* execute the writes immediately */
4252	e1e_flush();
4253}
4254
4255static void e1000e_update_stats(struct e1000_adapter *adapter);
4256
4257/**
4258 * e1000e_down - quiesce the device and optionally reset the hardware
4259 * @adapter: board private structure
4260 * @reset: boolean flag to reset the hardware or not
4261 */
4262void e1000e_down(struct e1000_adapter *adapter, bool reset)
4263{
4264	struct net_device *netdev = adapter->netdev;
4265	struct e1000_hw *hw = &adapter->hw;
4266	u32 tctl, rctl;
4267
4268	/* signal that we're down so the interrupt handler does not
4269	 * reschedule our watchdog timer
4270	 */
4271	set_bit(__E1000_DOWN, &adapter->state);
4272
4273	netif_carrier_off(netdev);
4274
4275	/* disable receives in the hardware */
4276	rctl = er32(RCTL);
4277	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
4278		ew32(RCTL, rctl & ~E1000_RCTL_EN);
4279	/* flush and sleep below */
4280
4281	netif_stop_queue(netdev);
4282
4283	/* disable transmits in the hardware */
4284	tctl = er32(TCTL);
4285	tctl &= ~E1000_TCTL_EN;
4286	ew32(TCTL, tctl);
4287
4288	/* flush both disables and wait for them to finish */
4289	e1e_flush();
4290	usleep_range(10000, 11000);
4291
4292	e1000_irq_disable(adapter);
4293
4294	napi_synchronize(&adapter->napi);
4295
4296	del_timer_sync(&adapter->watchdog_timer);
4297	del_timer_sync(&adapter->phy_info_timer);
4298
4299	spin_lock(&adapter->stats64_lock);
4300	e1000e_update_stats(adapter);
4301	spin_unlock(&adapter->stats64_lock);
4302
4303	e1000e_flush_descriptors(adapter);
4304
4305	adapter->link_speed = 0;
4306	adapter->link_duplex = 0;
4307
4308	/* Disable Si errata workaround on PCHx for jumbo frame flow */
4309	if ((hw->mac.type >= e1000_pch2lan) &&
4310	    (adapter->netdev->mtu > ETH_DATA_LEN) &&
4311	    e1000_lv_jumbo_workaround_ich8lan(hw, false))
4312		e_dbg("failed to disable jumbo frame workaround mode\n");
4313
4314	if (!pci_channel_offline(adapter->pdev)) {
4315		if (reset)
4316			e1000e_reset(adapter);
4317		else if (hw->mac.type >= e1000_pch_spt)
4318			e1000_flush_desc_rings(adapter);
4319	}
4320	e1000_clean_tx_ring(adapter->tx_ring);
4321	e1000_clean_rx_ring(adapter->rx_ring);
4322}
4323
4324void e1000e_reinit_locked(struct e1000_adapter *adapter)
4325{
4326	might_sleep();
4327	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
4328		usleep_range(1000, 1100);
4329	e1000e_down(adapter, true);
4330	e1000e_up(adapter);
4331	clear_bit(__E1000_RESETTING, &adapter->state);
4332}
4333
4334/**
4335 * e1000e_sanitize_systim - sanitize raw cycle counter reads
4336 * @hw: pointer to the HW structure
4337 * @systim: PHC time value read, sanitized and returned
4338 * @sts: structure to hold system time before and after reading SYSTIML,
4339 * may be NULL
4340 *
4341 * Errata for 82574/82583 possible bad bits read from SYSTIMH/L:
4342 * check to see that the time is incrementing at a reasonable
4343 * rate and is a multiple of incvalue.
4344 **/
4345static u64 e1000e_sanitize_systim(struct e1000_hw *hw, u64 systim,
4346				  struct ptp_system_timestamp *sts)
4347{
4348	u64 time_delta, rem, temp;
4349	u64 systim_next;
4350	u32 incvalue;
4351	int i;
4352
4353	incvalue = er32(TIMINCA) & E1000_TIMINCA_INCVALUE_MASK;
4354	for (i = 0; i < E1000_MAX_82574_SYSTIM_REREADS; i++) {
4355		/* latch SYSTIMH on read of SYSTIML */
4356		ptp_read_system_prets(sts);
4357		systim_next = (u64)er32(SYSTIML);
4358		ptp_read_system_postts(sts);
4359		systim_next |= (u64)er32(SYSTIMH) << 32;
4360
4361		time_delta = systim_next - systim;
4362		temp = time_delta;
4363		/* VMWare users have seen incvalue of zero, don't div / 0 */
4364		rem = incvalue ? do_div(temp, incvalue) : (time_delta != 0);
4365
4366		systim = systim_next;
4367
4368		if ((time_delta < E1000_82574_SYSTIM_EPSILON) && (rem == 0))
4369			break;
4370	}
4371
4372	return systim;
4373}
4374
4375/**
4376 * e1000e_read_systim - read SYSTIM register
4377 * @adapter: board private structure
4378 * @sts: structure which will contain system time before and after reading
4379 * SYSTIML, may be NULL
4380 **/
4381u64 e1000e_read_systim(struct e1000_adapter *adapter,
4382		       struct ptp_system_timestamp *sts)
4383{
4384	struct e1000_hw *hw = &adapter->hw;
4385	u32 systimel, systimel_2, systimeh;
4386	u64 systim;
4387	/* SYSTIMH latching upon SYSTIML read does not work well.
4388	 * This means that if SYSTIML overflows after we read it but before
4389	 * we read SYSTIMH, the value of SYSTIMH has been incremented and we
4390	 * will experience a huge non linear increment in the systime value
4391	 * to fix that we test for overflow and if true, we re-read systime.
4392	 */
4393	ptp_read_system_prets(sts);
4394	systimel = er32(SYSTIML);
4395	ptp_read_system_postts(sts);
4396	systimeh = er32(SYSTIMH);
4397	/* Is systimel is so large that overflow is possible? */
4398	if (systimel >= (u32)0xffffffff - E1000_TIMINCA_INCVALUE_MASK) {
4399		ptp_read_system_prets(sts);
4400		systimel_2 = er32(SYSTIML);
4401		ptp_read_system_postts(sts);
4402		if (systimel > systimel_2) {
4403			/* There was an overflow, read again SYSTIMH, and use
4404			 * systimel_2
4405			 */
4406			systimeh = er32(SYSTIMH);
4407			systimel = systimel_2;
4408		}
4409	}
4410	systim = (u64)systimel;
4411	systim |= (u64)systimeh << 32;
4412
4413	if (adapter->flags2 & FLAG2_CHECK_SYSTIM_OVERFLOW)
4414		systim = e1000e_sanitize_systim(hw, systim, sts);
4415
4416	return systim;
4417}
4418
4419/**
4420 * e1000e_cyclecounter_read - read raw cycle counter (used by time counter)
4421 * @cc: cyclecounter structure
4422 **/
4423static u64 e1000e_cyclecounter_read(const struct cyclecounter *cc)
4424{
4425	struct e1000_adapter *adapter = container_of(cc, struct e1000_adapter,
4426						     cc);
4427
4428	return e1000e_read_systim(adapter, NULL);
4429}
4430
4431/**
4432 * e1000_sw_init - Initialize general software structures (struct e1000_adapter)
4433 * @adapter: board private structure to initialize
4434 *
4435 * e1000_sw_init initializes the Adapter private data structure.
4436 * Fields are initialized based on PCI device information and
4437 * OS network device settings (MTU size).
4438 **/
4439static int e1000_sw_init(struct e1000_adapter *adapter)
4440{
4441	struct net_device *netdev = adapter->netdev;
4442
4443	adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
4444	adapter->rx_ps_bsize0 = 128;
4445	adapter->max_frame_size = netdev->mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
4446	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4447	adapter->tx_ring_count = E1000_DEFAULT_TXD;
4448	adapter->rx_ring_count = E1000_DEFAULT_RXD;
4449
4450	spin_lock_init(&adapter->stats64_lock);
4451
4452	e1000e_set_interrupt_capability(adapter);
4453
4454	if (e1000_alloc_queues(adapter))
4455		return -ENOMEM;
4456
4457	/* Setup hardware time stamping cyclecounter */
4458	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
4459		adapter->cc.read = e1000e_cyclecounter_read;
4460		adapter->cc.mask = CYCLECOUNTER_MASK(64);
4461		adapter->cc.mult = 1;
4462		/* cc.shift set in e1000e_get_base_tininca() */
4463
4464		spin_lock_init(&adapter->systim_lock);
4465		INIT_WORK(&adapter->tx_hwtstamp_work, e1000e_tx_hwtstamp_work);
4466	}
4467
4468	/* Explicitly disable IRQ since the NIC can be in any state. */
4469	e1000_irq_disable(adapter);
4470
4471	set_bit(__E1000_DOWN, &adapter->state);
4472	return 0;
4473}
4474
4475/**
4476 * e1000_intr_msi_test - Interrupt Handler
4477 * @irq: interrupt number
4478 * @data: pointer to a network interface device structure
4479 **/
4480static irqreturn_t e1000_intr_msi_test(int __always_unused irq, void *data)
4481{
4482	struct net_device *netdev = data;
4483	struct e1000_adapter *adapter = netdev_priv(netdev);
4484	struct e1000_hw *hw = &adapter->hw;
4485	u32 icr = er32(ICR);
4486
4487	e_dbg("icr is %08X\n", icr);
4488	if (icr & E1000_ICR_RXSEQ) {
4489		adapter->flags &= ~FLAG_MSI_TEST_FAILED;
4490		/* Force memory writes to complete before acknowledging the
4491		 * interrupt is handled.
4492		 */
4493		wmb();
4494	}
4495
4496	return IRQ_HANDLED;
4497}
4498
4499/**
4500 * e1000_test_msi_interrupt - Returns 0 for successful test
4501 * @adapter: board private struct
4502 *
4503 * code flow taken from tg3.c
4504 **/
4505static int e1000_test_msi_interrupt(struct e1000_adapter *adapter)
4506{
4507	struct net_device *netdev = adapter->netdev;
4508	struct e1000_hw *hw = &adapter->hw;
4509	int err;
4510
4511	/* poll_enable hasn't been called yet, so don't need disable */
4512	/* clear any pending events */
4513	er32(ICR);
4514
4515	/* free the real vector and request a test handler */
4516	e1000_free_irq(adapter);
4517	e1000e_reset_interrupt_capability(adapter);
4518
4519	/* Assume that the test fails, if it succeeds then the test
4520	 * MSI irq handler will unset this flag
4521	 */
4522	adapter->flags |= FLAG_MSI_TEST_FAILED;
4523
4524	err = pci_enable_msi(adapter->pdev);
4525	if (err)
4526		goto msi_test_failed;
4527
4528	err = request_irq(adapter->pdev->irq, e1000_intr_msi_test, 0,
4529			  netdev->name, netdev);
4530	if (err) {
4531		pci_disable_msi(adapter->pdev);
4532		goto msi_test_failed;
4533	}
4534
4535	/* Force memory writes to complete before enabling and firing an
4536	 * interrupt.
4537	 */
4538	wmb();
4539
4540	e1000_irq_enable(adapter);
4541
4542	/* fire an unusual interrupt on the test handler */
4543	ew32(ICS, E1000_ICS_RXSEQ);
4544	e1e_flush();
4545	msleep(100);
4546
4547	e1000_irq_disable(adapter);
4548
4549	rmb();			/* read flags after interrupt has been fired */
4550
4551	if (adapter->flags & FLAG_MSI_TEST_FAILED) {
4552		adapter->int_mode = E1000E_INT_MODE_LEGACY;
4553		e_info("MSI interrupt test failed, using legacy interrupt.\n");
4554	} else {
4555		e_dbg("MSI interrupt test succeeded!\n");
4556	}
4557
4558	free_irq(adapter->pdev->irq, netdev);
4559	pci_disable_msi(adapter->pdev);
4560
4561msi_test_failed:
4562	e1000e_set_interrupt_capability(adapter);
4563	return e1000_request_irq(adapter);
4564}
4565
4566/**
4567 * e1000_test_msi - Returns 0 if MSI test succeeds or INTx mode is restored
4568 * @adapter: board private struct
4569 *
4570 * code flow taken from tg3.c, called with e1000 interrupts disabled.
4571 **/
4572static int e1000_test_msi(struct e1000_adapter *adapter)
4573{
4574	int err;
4575	u16 pci_cmd;
4576
4577	if (!(adapter->flags & FLAG_MSI_ENABLED))
4578		return 0;
4579
4580	/* disable SERR in case the MSI write causes a master abort */
4581	pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4582	if (pci_cmd & PCI_COMMAND_SERR)
4583		pci_write_config_word(adapter->pdev, PCI_COMMAND,
4584				      pci_cmd & ~PCI_COMMAND_SERR);
4585
4586	err = e1000_test_msi_interrupt(adapter);
4587
4588	/* re-enable SERR */
4589	if (pci_cmd & PCI_COMMAND_SERR) {
4590		pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4591		pci_cmd |= PCI_COMMAND_SERR;
4592		pci_write_config_word(adapter->pdev, PCI_COMMAND, pci_cmd);
4593	}
4594
4595	return err;
4596}
4597
4598/**
4599 * e1000e_open - Called when a network interface is made active
4600 * @netdev: network interface device structure
4601 *
4602 * Returns 0 on success, negative value on failure
4603 *
4604 * The open entry point is called when a network interface is made
4605 * active by the system (IFF_UP).  At this point all resources needed
4606 * for transmit and receive operations are allocated, the interrupt
4607 * handler is registered with the OS, the watchdog timer is started,
4608 * and the stack is notified that the interface is ready.
4609 **/
4610int e1000e_open(struct net_device *netdev)
4611{
4612	struct e1000_adapter *adapter = netdev_priv(netdev);
4613	struct e1000_hw *hw = &adapter->hw;
4614	struct pci_dev *pdev = adapter->pdev;
4615	int err;
4616
4617	/* disallow open during test */
4618	if (test_bit(__E1000_TESTING, &adapter->state))
4619		return -EBUSY;
4620
4621	pm_runtime_get_sync(&pdev->dev);
4622
4623	netif_carrier_off(netdev);
4624	netif_stop_queue(netdev);
4625
4626	/* allocate transmit descriptors */
4627	err = e1000e_setup_tx_resources(adapter->tx_ring);
4628	if (err)
4629		goto err_setup_tx;
4630
4631	/* allocate receive descriptors */
4632	err = e1000e_setup_rx_resources(adapter->rx_ring);
4633	if (err)
4634		goto err_setup_rx;
4635
4636	/* If AMT is enabled, let the firmware know that the network
4637	 * interface is now open and reset the part to a known state.
4638	 */
4639	if (adapter->flags & FLAG_HAS_AMT) {
4640		e1000e_get_hw_control(adapter);
4641		e1000e_reset(adapter);
4642	}
4643
4644	e1000e_power_up_phy(adapter);
4645
4646	adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
4647	if ((adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
4648		e1000_update_mng_vlan(adapter);
4649
4650	/* DMA latency requirement to workaround jumbo issue */
4651	cpu_latency_qos_add_request(&adapter->pm_qos_req, PM_QOS_DEFAULT_VALUE);
4652
4653	/* before we allocate an interrupt, we must be ready to handle it.
4654	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4655	 * as soon as we call pci_request_irq, so we have to setup our
4656	 * clean_rx handler before we do so.
4657	 */
4658	e1000_configure(adapter);
4659
4660	err = e1000_request_irq(adapter);
4661	if (err)
4662		goto err_req_irq;
4663
4664	/* Work around PCIe errata with MSI interrupts causing some chipsets to
4665	 * ignore e1000e MSI messages, which means we need to test our MSI
4666	 * interrupt now
4667	 */
4668	if (adapter->int_mode != E1000E_INT_MODE_LEGACY) {
4669		err = e1000_test_msi(adapter);
4670		if (err) {
4671			e_err("Interrupt allocation failed\n");
4672			goto err_req_irq;
4673		}
4674	}
4675
4676	/* From here on the code is the same as e1000e_up() */
4677	clear_bit(__E1000_DOWN, &adapter->state);
4678
4679	napi_enable(&adapter->napi);
4680
4681	e1000_irq_enable(adapter);
4682
4683	adapter->tx_hang_recheck = false;
4684
4685	hw->mac.get_link_status = true;
4686	pm_runtime_put(&pdev->dev);
4687
4688	e1000e_trigger_lsc(adapter);
4689
4690	return 0;
4691
4692err_req_irq:
4693	cpu_latency_qos_remove_request(&adapter->pm_qos_req);
4694	e1000e_release_hw_control(adapter);
4695	e1000_power_down_phy(adapter);
4696	e1000e_free_rx_resources(adapter->rx_ring);
4697err_setup_rx:
4698	e1000e_free_tx_resources(adapter->tx_ring);
4699err_setup_tx:
4700	e1000e_reset(adapter);
4701	pm_runtime_put_sync(&pdev->dev);
4702
4703	return err;
4704}
4705
4706/**
4707 * e1000e_close - Disables a network interface
4708 * @netdev: network interface device structure
4709 *
4710 * Returns 0, this is not allowed to fail
4711 *
4712 * The close entry point is called when an interface is de-activated
4713 * by the OS.  The hardware is still under the drivers control, but
4714 * needs to be disabled.  A global MAC reset is issued to stop the
4715 * hardware, and all transmit and receive resources are freed.
4716 **/
4717int e1000e_close(struct net_device *netdev)
4718{
4719	struct e1000_adapter *adapter = netdev_priv(netdev);
4720	struct pci_dev *pdev = adapter->pdev;
4721	int count = E1000_CHECK_RESET_COUNT;
4722
4723	while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
4724		usleep_range(10000, 11000);
4725
4726	WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
4727
4728	pm_runtime_get_sync(&pdev->dev);
4729
4730	if (netif_device_present(netdev)) {
4731		e1000e_down(adapter, true);
4732		e1000_free_irq(adapter);
4733
4734		/* Link status message must follow this format */
4735		netdev_info(netdev, "NIC Link is Down\n");
4736	}
4737
4738	napi_disable(&adapter->napi);
4739
4740	e1000e_free_tx_resources(adapter->tx_ring);
4741	e1000e_free_rx_resources(adapter->rx_ring);
4742
4743	/* kill manageability vlan ID if supported, but not if a vlan with
4744	 * the same ID is registered on the host OS (let 8021q kill it)
4745	 */
4746	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN)
4747		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
4748				       adapter->mng_vlan_id);
4749
4750	/* If AMT is enabled, let the firmware know that the network
4751	 * interface is now closed
4752	 */
4753	if ((adapter->flags & FLAG_HAS_AMT) &&
4754	    !test_bit(__E1000_TESTING, &adapter->state))
4755		e1000e_release_hw_control(adapter);
4756
4757	cpu_latency_qos_remove_request(&adapter->pm_qos_req);
4758
4759	pm_runtime_put_sync(&pdev->dev);
4760
4761	return 0;
4762}
4763
4764/**
4765 * e1000_set_mac - Change the Ethernet Address of the NIC
4766 * @netdev: network interface device structure
4767 * @p: pointer to an address structure
4768 *
4769 * Returns 0 on success, negative on failure
4770 **/
4771static int e1000_set_mac(struct net_device *netdev, void *p)
4772{
4773	struct e1000_adapter *adapter = netdev_priv(netdev);
4774	struct e1000_hw *hw = &adapter->hw;
4775	struct sockaddr *addr = p;
4776
4777	if (!is_valid_ether_addr(addr->sa_data))
4778		return -EADDRNOTAVAIL;
4779
4780	eth_hw_addr_set(netdev, addr->sa_data);
4781	memcpy(adapter->hw.mac.addr, addr->sa_data, netdev->addr_len);
4782
4783	hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
4784
4785	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA) {
4786		/* activate the work around */
4787		e1000e_set_laa_state_82571(&adapter->hw, 1);
4788
4789		/* Hold a copy of the LAA in RAR[14] This is done so that
4790		 * between the time RAR[0] gets clobbered  and the time it
4791		 * gets fixed (in e1000_watchdog), the actual LAA is in one
4792		 * of the RARs and no incoming packets directed to this port
4793		 * are dropped. Eventually the LAA will be in RAR[0] and
4794		 * RAR[14]
4795		 */
4796		hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr,
4797				    adapter->hw.mac.rar_entry_count - 1);
4798	}
4799
4800	return 0;
4801}
4802
4803/**
4804 * e1000e_update_phy_task - work thread to update phy
4805 * @work: pointer to our work struct
4806 *
4807 * this worker thread exists because we must acquire a
4808 * semaphore to read the phy, which we could msleep while
4809 * waiting for it, and we can't msleep in a timer.
4810 **/
4811static void e1000e_update_phy_task(struct work_struct *work)
4812{
4813	struct e1000_adapter *adapter = container_of(work,
4814						     struct e1000_adapter,
4815						     update_phy_task);
4816	struct e1000_hw *hw = &adapter->hw;
4817
4818	if (test_bit(__E1000_DOWN, &adapter->state))
4819		return;
4820
4821	e1000_get_phy_info(hw);
4822
4823	/* Enable EEE on 82579 after link up */
4824	if (hw->phy.type >= e1000_phy_82579)
4825		e1000_set_eee_pchlan(hw);
4826}
4827
4828/**
4829 * e1000_update_phy_info - timre call-back to update PHY info
4830 * @t: pointer to timer_list containing private info adapter
4831 *
4832 * Need to wait a few seconds after link up to get diagnostic information from
4833 * the phy
4834 **/
4835static void e1000_update_phy_info(struct timer_list *t)
4836{
4837	struct e1000_adapter *adapter = from_timer(adapter, t, phy_info_timer);
4838
4839	if (test_bit(__E1000_DOWN, &adapter->state))
4840		return;
4841
4842	schedule_work(&adapter->update_phy_task);
4843}
4844
4845/**
4846 * e1000e_update_phy_stats - Update the PHY statistics counters
4847 * @adapter: board private structure
4848 *
4849 * Read/clear the upper 16-bit PHY registers and read/accumulate lower
4850 **/
4851static void e1000e_update_phy_stats(struct e1000_adapter *adapter)
4852{
4853	struct e1000_hw *hw = &adapter->hw;
4854	s32 ret_val;
4855	u16 phy_data;
4856
4857	ret_val = hw->phy.ops.acquire(hw);
4858	if (ret_val)
4859		return;
4860
4861	/* A page set is expensive so check if already on desired page.
4862	 * If not, set to the page with the PHY status registers.
4863	 */
4864	hw->phy.addr = 1;
4865	ret_val = e1000e_read_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4866					   &phy_data);
4867	if (ret_val)
4868		goto release;
4869	if (phy_data != (HV_STATS_PAGE << IGP_PAGE_SHIFT)) {
4870		ret_val = hw->phy.ops.set_page(hw,
4871					       HV_STATS_PAGE << IGP_PAGE_SHIFT);
4872		if (ret_val)
4873			goto release;
4874	}
4875
4876	/* Single Collision Count */
4877	hw->phy.ops.read_reg_page(hw, HV_SCC_UPPER, &phy_data);
4878	ret_val = hw->phy.ops.read_reg_page(hw, HV_SCC_LOWER, &phy_data);
4879	if (!ret_val)
4880		adapter->stats.scc += phy_data;
4881
4882	/* Excessive Collision Count */
4883	hw->phy.ops.read_reg_page(hw, HV_ECOL_UPPER, &phy_data);
4884	ret_val = hw->phy.ops.read_reg_page(hw, HV_ECOL_LOWER, &phy_data);
4885	if (!ret_val)
4886		adapter->stats.ecol += phy_data;
4887
4888	/* Multiple Collision Count */
4889	hw->phy.ops.read_reg_page(hw, HV_MCC_UPPER, &phy_data);
4890	ret_val = hw->phy.ops.read_reg_page(hw, HV_MCC_LOWER, &phy_data);
4891	if (!ret_val)
4892		adapter->stats.mcc += phy_data;
4893
4894	/* Late Collision Count */
4895	hw->phy.ops.read_reg_page(hw, HV_LATECOL_UPPER, &phy_data);
4896	ret_val = hw->phy.ops.read_reg_page(hw, HV_LATECOL_LOWER, &phy_data);
4897	if (!ret_val)
4898		adapter->stats.latecol += phy_data;
4899
4900	/* Collision Count - also used for adaptive IFS */
4901	hw->phy.ops.read_reg_page(hw, HV_COLC_UPPER, &phy_data);
4902	ret_val = hw->phy.ops.read_reg_page(hw, HV_COLC_LOWER, &phy_data);
4903	if (!ret_val)
4904		hw->mac.collision_delta = phy_data;
4905
4906	/* Defer Count */
4907	hw->phy.ops.read_reg_page(hw, HV_DC_UPPER, &phy_data);
4908	ret_val = hw->phy.ops.read_reg_page(hw, HV_DC_LOWER, &phy_data);
4909	if (!ret_val)
4910		adapter->stats.dc += phy_data;
4911
4912	/* Transmit with no CRS */
4913	hw->phy.ops.read_reg_page(hw, HV_TNCRS_UPPER, &phy_data);
4914	ret_val = hw->phy.ops.read_reg_page(hw, HV_TNCRS_LOWER, &phy_data);
4915	if (!ret_val)
4916		adapter->stats.tncrs += phy_data;
4917
4918release:
4919	hw->phy.ops.release(hw);
4920}
4921
4922/**
4923 * e1000e_update_stats - Update the board statistics counters
4924 * @adapter: board private structure
4925 **/
4926static void e1000e_update_stats(struct e1000_adapter *adapter)
4927{
4928	struct net_device *netdev = adapter->netdev;
4929	struct e1000_hw *hw = &adapter->hw;
4930	struct pci_dev *pdev = adapter->pdev;
4931
4932	/* Prevent stats update while adapter is being reset, or if the pci
4933	 * connection is down.
4934	 */
4935	if (adapter->link_speed == 0)
4936		return;
4937	if (pci_channel_offline(pdev))
4938		return;
4939
4940	adapter->stats.crcerrs += er32(CRCERRS);
4941	adapter->stats.gprc += er32(GPRC);
4942	adapter->stats.gorc += er32(GORCL);
4943	er32(GORCH);		/* Clear gorc */
4944	adapter->stats.bprc += er32(BPRC);
4945	adapter->stats.mprc += er32(MPRC);
4946	adapter->stats.roc += er32(ROC);
4947
4948	adapter->stats.mpc += er32(MPC);
4949
4950	/* Half-duplex statistics */
4951	if (adapter->link_duplex == HALF_DUPLEX) {
4952		if (adapter->flags2 & FLAG2_HAS_PHY_STATS) {
4953			e1000e_update_phy_stats(adapter);
4954		} else {
4955			adapter->stats.scc += er32(SCC);
4956			adapter->stats.ecol += er32(ECOL);
4957			adapter->stats.mcc += er32(MCC);
4958			adapter->stats.latecol += er32(LATECOL);
4959			adapter->stats.dc += er32(DC);
4960
4961			hw->mac.collision_delta = er32(COLC);
4962
4963			if ((hw->mac.type != e1000_82574) &&
4964			    (hw->mac.type != e1000_82583))
4965				adapter->stats.tncrs += er32(TNCRS);
4966		}
4967		adapter->stats.colc += hw->mac.collision_delta;
4968	}
4969
4970	adapter->stats.xonrxc += er32(XONRXC);
4971	adapter->stats.xontxc += er32(XONTXC);
4972	adapter->stats.xoffrxc += er32(XOFFRXC);
4973	adapter->stats.xofftxc += er32(XOFFTXC);
4974	adapter->stats.gptc += er32(GPTC);
4975	adapter->stats.gotc += er32(GOTCL);
4976	er32(GOTCH);		/* Clear gotc */
4977	adapter->stats.rnbc += er32(RNBC);
4978	adapter->stats.ruc += er32(RUC);
4979
4980	adapter->stats.mptc += er32(MPTC);
4981	adapter->stats.bptc += er32(BPTC);
4982
4983	/* used for adaptive IFS */
4984
4985	hw->mac.tx_packet_delta = er32(TPT);
4986	adapter->stats.tpt += hw->mac.tx_packet_delta;
4987
4988	adapter->stats.algnerrc += er32(ALGNERRC);
4989	adapter->stats.rxerrc += er32(RXERRC);
4990	adapter->stats.cexterr += er32(CEXTERR);
4991	adapter->stats.tsctc += er32(TSCTC);
4992	adapter->stats.tsctfc += er32(TSCTFC);
4993
4994	/* Fill out the OS statistics structure */
4995	netdev->stats.multicast = adapter->stats.mprc;
4996	netdev->stats.collisions = adapter->stats.colc;
4997
4998	/* Rx Errors */
4999
5000	/* RLEC on some newer hardware can be incorrect so build
5001	 * our own version based on RUC and ROC
5002	 */
5003	netdev->stats.rx_errors = adapter->stats.rxerrc +
5004	    adapter->stats.crcerrs + adapter->stats.algnerrc +
5005	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5006	netdev->stats.rx_length_errors = adapter->stats.ruc +
5007	    adapter->stats.roc;
5008	netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
5009	netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
5010	netdev->stats.rx_missed_errors = adapter->stats.mpc;
5011
5012	/* Tx Errors */
5013	netdev->stats.tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5014	netdev->stats.tx_aborted_errors = adapter->stats.ecol;
5015	netdev->stats.tx_window_errors = adapter->stats.latecol;
5016	netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
5017
5018	/* Tx Dropped needs to be maintained elsewhere */
5019
5020	/* Management Stats */
5021	adapter->stats.mgptc += er32(MGTPTC);
5022	adapter->stats.mgprc += er32(MGTPRC);
5023	adapter->stats.mgpdc += er32(MGTPDC);
5024
5025	/* Correctable ECC Errors */
5026	if (hw->mac.type >= e1000_pch_lpt) {
5027		u32 pbeccsts = er32(PBECCSTS);
5028
5029		adapter->corr_errors +=
5030		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
5031		adapter->uncorr_errors +=
5032		    FIELD_GET(E1000_PBECCSTS_UNCORR_ERR_CNT_MASK, pbeccsts);
5033	}
5034}
5035
5036/**
5037 * e1000_phy_read_status - Update the PHY register status snapshot
5038 * @adapter: board private structure
5039 **/
5040static void e1000_phy_read_status(struct e1000_adapter *adapter)
5041{
5042	struct e1000_hw *hw = &adapter->hw;
5043	struct e1000_phy_regs *phy = &adapter->phy_regs;
5044
5045	if (!pm_runtime_suspended((&adapter->pdev->dev)->parent) &&
5046	    (er32(STATUS) & E1000_STATUS_LU) &&
5047	    (adapter->hw.phy.media_type == e1000_media_type_copper)) {
5048		int ret_val;
5049
5050		ret_val = e1e_rphy(hw, MII_BMCR, &phy->bmcr);
5051		ret_val |= e1e_rphy(hw, MII_BMSR, &phy->bmsr);
5052		ret_val |= e1e_rphy(hw, MII_ADVERTISE, &phy->advertise);
5053		ret_val |= e1e_rphy(hw, MII_LPA, &phy->lpa);
5054		ret_val |= e1e_rphy(hw, MII_EXPANSION, &phy->expansion);
5055		ret_val |= e1e_rphy(hw, MII_CTRL1000, &phy->ctrl1000);
5056		ret_val |= e1e_rphy(hw, MII_STAT1000, &phy->stat1000);
5057		ret_val |= e1e_rphy(hw, MII_ESTATUS, &phy->estatus);
5058		if (ret_val)
5059			e_warn("Error reading PHY register\n");
5060	} else {
5061		/* Do not read PHY registers if link is not up
5062		 * Set values to typical power-on defaults
5063		 */
5064		phy->bmcr = (BMCR_SPEED1000 | BMCR_ANENABLE | BMCR_FULLDPLX);
5065		phy->bmsr = (BMSR_100FULL | BMSR_100HALF | BMSR_10FULL |
5066			     BMSR_10HALF | BMSR_ESTATEN | BMSR_ANEGCAPABLE |
5067			     BMSR_ERCAP);
5068		phy->advertise = (ADVERTISE_PAUSE_ASYM | ADVERTISE_PAUSE_CAP |
5069				  ADVERTISE_ALL | ADVERTISE_CSMA);
5070		phy->lpa = 0;
5071		phy->expansion = EXPANSION_ENABLENPAGE;
5072		phy->ctrl1000 = ADVERTISE_1000FULL;
5073		phy->stat1000 = 0;
5074		phy->estatus = (ESTATUS_1000_TFULL | ESTATUS_1000_THALF);
5075	}
5076}
5077
5078static void e1000_print_link_info(struct e1000_adapter *adapter)
5079{
5080	struct e1000_hw *hw = &adapter->hw;
5081	u32 ctrl = er32(CTRL);
5082
5083	/* Link status message must follow this format for user tools */
5084	netdev_info(adapter->netdev,
5085		    "NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5086		    adapter->link_speed,
5087		    adapter->link_duplex == FULL_DUPLEX ? "Full" : "Half",
5088		    (ctrl & E1000_CTRL_TFCE) && (ctrl & E1000_CTRL_RFCE) ? "Rx/Tx" :
5089		    (ctrl & E1000_CTRL_RFCE) ? "Rx" :
5090		    (ctrl & E1000_CTRL_TFCE) ? "Tx" : "None");
5091}
5092
5093static bool e1000e_has_link(struct e1000_adapter *adapter)
5094{
5095	struct e1000_hw *hw = &adapter->hw;
5096	bool link_active = false;
5097	s32 ret_val = 0;
5098
5099	/* get_link_status is set on LSC (link status) interrupt or
5100	 * Rx sequence error interrupt.  get_link_status will stay
5101	 * true until the check_for_link establishes link
5102	 * for copper adapters ONLY
5103	 */
5104	switch (hw->phy.media_type) {
5105	case e1000_media_type_copper:
5106		if (hw->mac.get_link_status) {
5107			ret_val = hw->mac.ops.check_for_link(hw);
5108			link_active = !hw->mac.get_link_status;
5109		} else {
5110			link_active = true;
5111		}
5112		break;
5113	case e1000_media_type_fiber:
5114		ret_val = hw->mac.ops.check_for_link(hw);
5115		link_active = !!(er32(STATUS) & E1000_STATUS_LU);
5116		break;
5117	case e1000_media_type_internal_serdes:
5118		ret_val = hw->mac.ops.check_for_link(hw);
5119		link_active = hw->mac.serdes_has_link;
5120		break;
5121	default:
5122	case e1000_media_type_unknown:
5123		break;
5124	}
5125
5126	if ((ret_val == -E1000_ERR_PHY) && (hw->phy.type == e1000_phy_igp_3) &&
5127	    (er32(CTRL) & E1000_PHY_CTRL_GBE_DISABLE)) {
5128		/* See e1000_kmrn_lock_loss_workaround_ich8lan() */
5129		e_info("Gigabit has been disabled, downgrading speed\n");
5130	}
5131
5132	return link_active;
5133}
5134
5135static void e1000e_enable_receives(struct e1000_adapter *adapter)
5136{
5137	/* make sure the receive unit is started */
5138	if ((adapter->flags & FLAG_RX_NEEDS_RESTART) &&
5139	    (adapter->flags & FLAG_RESTART_NOW)) {
5140		struct e1000_hw *hw = &adapter->hw;
5141		u32 rctl = er32(RCTL);
5142
5143		ew32(RCTL, rctl | E1000_RCTL_EN);
5144		adapter->flags &= ~FLAG_RESTART_NOW;
5145	}
5146}
5147
5148static void e1000e_check_82574_phy_workaround(struct e1000_adapter *adapter)
5149{
5150	struct e1000_hw *hw = &adapter->hw;
5151
5152	/* With 82574 controllers, PHY needs to be checked periodically
5153	 * for hung state and reset, if two calls return true
5154	 */
5155	if (e1000_check_phy_82574(hw))
5156		adapter->phy_hang_count++;
5157	else
5158		adapter->phy_hang_count = 0;
5159
5160	if (adapter->phy_hang_count > 1) {
5161		adapter->phy_hang_count = 0;
5162		e_dbg("PHY appears hung - resetting\n");
5163		schedule_work(&adapter->reset_task);
5164	}
5165}
5166
5167/**
5168 * e1000_watchdog - Timer Call-back
5169 * @t: pointer to timer_list containing private info adapter
5170 **/
5171static void e1000_watchdog(struct timer_list *t)
5172{
5173	struct e1000_adapter *adapter = from_timer(adapter, t, watchdog_timer);
5174
5175	/* Do the rest outside of interrupt context */
5176	schedule_work(&adapter->watchdog_task);
5177
5178	/* TODO: make this use queue_delayed_work() */
5179}
5180
5181static void e1000_watchdog_task(struct work_struct *work)
5182{
5183	struct e1000_adapter *adapter = container_of(work,
5184						     struct e1000_adapter,
5185						     watchdog_task);
5186	struct net_device *netdev = adapter->netdev;
5187	struct e1000_mac_info *mac = &adapter->hw.mac;
5188	struct e1000_phy_info *phy = &adapter->hw.phy;
5189	struct e1000_ring *tx_ring = adapter->tx_ring;
5190	u32 dmoff_exit_timeout = 100, tries = 0;
5191	struct e1000_hw *hw = &adapter->hw;
5192	u32 link, tctl, pcim_state;
5193
5194	if (test_bit(__E1000_DOWN, &adapter->state))
5195		return;
5196
5197	link = e1000e_has_link(adapter);
5198	if ((netif_carrier_ok(netdev)) && link) {
5199		/* Cancel scheduled suspend requests. */
5200		pm_runtime_resume(netdev->dev.parent);
5201
5202		e1000e_enable_receives(adapter);
5203		goto link_up;
5204	}
5205
5206	if ((e1000e_enable_tx_pkt_filtering(hw)) &&
5207	    (adapter->mng_vlan_id != adapter->hw.mng_cookie.vlan_id))
5208		e1000_update_mng_vlan(adapter);
5209
5210	if (link) {
5211		if (!netif_carrier_ok(netdev)) {
5212			bool txb2b = true;
5213
5214			/* Cancel scheduled suspend requests. */
5215			pm_runtime_resume(netdev->dev.parent);
5216
5217			/* Checking if MAC is in DMoff state*/
5218			if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID) {
5219				pcim_state = er32(STATUS);
5220				while (pcim_state & E1000_STATUS_PCIM_STATE) {
5221					if (tries++ == dmoff_exit_timeout) {
5222						e_dbg("Error in exiting dmoff\n");
5223						break;
5224					}
5225					usleep_range(10000, 20000);
5226					pcim_state = er32(STATUS);
5227
5228					/* Checking if MAC exited DMoff state */
5229					if (!(pcim_state & E1000_STATUS_PCIM_STATE))
5230						e1000_phy_hw_reset(&adapter->hw);
5231				}
5232			}
5233
5234			/* update snapshot of PHY registers on LSC */
5235			e1000_phy_read_status(adapter);
5236			mac->ops.get_link_up_info(&adapter->hw,
5237						  &adapter->link_speed,
5238						  &adapter->link_duplex);
5239			e1000_print_link_info(adapter);
5240
5241			/* check if SmartSpeed worked */
5242			e1000e_check_downshift(hw);
5243			if (phy->speed_downgraded)
5244				netdev_warn(netdev,
5245					    "Link Speed was downgraded by SmartSpeed\n");
5246
5247			/* On supported PHYs, check for duplex mismatch only
5248			 * if link has autonegotiated at 10/100 half
5249			 */
5250			if ((hw->phy.type == e1000_phy_igp_3 ||
5251			     hw->phy.type == e1000_phy_bm) &&
5252			    hw->mac.autoneg &&
5253			    (adapter->link_speed == SPEED_10 ||
5254			     adapter->link_speed == SPEED_100) &&
5255			    (adapter->link_duplex == HALF_DUPLEX)) {
5256				u16 autoneg_exp;
5257
5258				e1e_rphy(hw, MII_EXPANSION, &autoneg_exp);
5259
5260				if (!(autoneg_exp & EXPANSION_NWAY))
5261					e_info("Autonegotiated half duplex but link partner cannot autoneg.  Try forcing full duplex if link gets many collisions.\n");
5262			}
5263
5264			/* adjust timeout factor according to speed/duplex */
5265			adapter->tx_timeout_factor = 1;
5266			switch (adapter->link_speed) {
5267			case SPEED_10:
5268				txb2b = false;
5269				adapter->tx_timeout_factor = 16;
5270				break;
5271			case SPEED_100:
5272				txb2b = false;
5273				adapter->tx_timeout_factor = 10;
5274				break;
5275			}
5276
5277			/* workaround: re-program speed mode bit after
5278			 * link-up event
5279			 */
5280			if ((adapter->flags & FLAG_TARC_SPEED_MODE_BIT) &&
5281			    !txb2b) {
5282				u32 tarc0;
5283
5284				tarc0 = er32(TARC(0));
5285				tarc0 &= ~SPEED_MODE_BIT;
5286				ew32(TARC(0), tarc0);
5287			}
5288
5289			/* enable transmits in the hardware, need to do this
5290			 * after setting TARC(0)
5291			 */
5292			tctl = er32(TCTL);
5293			tctl |= E1000_TCTL_EN;
5294			ew32(TCTL, tctl);
5295
5296			/* Perform any post-link-up configuration before
5297			 * reporting link up.
5298			 */
5299			if (phy->ops.cfg_on_link_up)
5300				phy->ops.cfg_on_link_up(hw);
5301
5302			netif_wake_queue(netdev);
5303			netif_carrier_on(netdev);
5304
5305			if (!test_bit(__E1000_DOWN, &adapter->state))
5306				mod_timer(&adapter->phy_info_timer,
5307					  round_jiffies(jiffies + 2 * HZ));
5308		}
5309	} else {
5310		if (netif_carrier_ok(netdev)) {
5311			adapter->link_speed = 0;
5312			adapter->link_duplex = 0;
5313			/* Link status message must follow this format */
5314			netdev_info(netdev, "NIC Link is Down\n");
5315			netif_carrier_off(netdev);
5316			netif_stop_queue(netdev);
5317			if (!test_bit(__E1000_DOWN, &adapter->state))
5318				mod_timer(&adapter->phy_info_timer,
5319					  round_jiffies(jiffies + 2 * HZ));
5320
5321			/* 8000ES2LAN requires a Rx packet buffer work-around
5322			 * on link down event; reset the controller to flush
5323			 * the Rx packet buffer.
5324			 */
5325			if (adapter->flags & FLAG_RX_NEEDS_RESTART)
5326				adapter->flags |= FLAG_RESTART_NOW;
5327			else
5328				pm_schedule_suspend(netdev->dev.parent,
5329						    LINK_TIMEOUT);
5330		}
5331	}
5332
5333link_up:
5334	spin_lock(&adapter->stats64_lock);
5335	e1000e_update_stats(adapter);
5336
5337	mac->tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
5338	adapter->tpt_old = adapter->stats.tpt;
5339	mac->collision_delta = adapter->stats.colc - adapter->colc_old;
5340	adapter->colc_old = adapter->stats.colc;
5341
5342	adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
5343	adapter->gorc_old = adapter->stats.gorc;
5344	adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
5345	adapter->gotc_old = adapter->stats.gotc;
5346	spin_unlock(&adapter->stats64_lock);
5347
5348	/* If the link is lost the controller stops DMA, but
5349	 * if there is queued Tx work it cannot be done.  So
5350	 * reset the controller to flush the Tx packet buffers.
5351	 */
5352	if (!netif_carrier_ok(netdev) &&
5353	    (e1000_desc_unused(tx_ring) + 1 < tx_ring->count))
5354		adapter->flags |= FLAG_RESTART_NOW;
5355
5356	/* If reset is necessary, do it outside of interrupt context. */
5357	if (adapter->flags & FLAG_RESTART_NOW) {
5358		schedule_work(&adapter->reset_task);
5359		/* return immediately since reset is imminent */
5360		return;
5361	}
5362
5363	e1000e_update_adaptive(&adapter->hw);
5364
5365	/* Simple mode for Interrupt Throttle Rate (ITR) */
5366	if (adapter->itr_setting == 4) {
5367		/* Symmetric Tx/Rx gets a reduced ITR=2000;
5368		 * Total asymmetrical Tx or Rx gets ITR=8000;
5369		 * everyone else is between 2000-8000.
5370		 */
5371		u32 goc = (adapter->gotc + adapter->gorc) / 10000;
5372		u32 dif = (adapter->gotc > adapter->gorc ?
5373			   adapter->gotc - adapter->gorc :
5374			   adapter->gorc - adapter->gotc) / 10000;
5375		u32 itr = goc > 0 ? (dif * 6000 / goc + 2000) : 8000;
5376
5377		e1000e_write_itr(adapter, itr);
5378	}
5379
5380	/* Cause software interrupt to ensure Rx ring is cleaned */
5381	if (adapter->msix_entries)
5382		ew32(ICS, adapter->rx_ring->ims_val);
5383	else
5384		ew32(ICS, E1000_ICS_RXDMT0);
5385
5386	/* flush pending descriptors to memory before detecting Tx hang */
5387	e1000e_flush_descriptors(adapter);
5388
5389	/* Force detection of hung controller every watchdog period */
5390	adapter->detect_tx_hung = true;
5391
5392	/* With 82571 controllers, LAA may be overwritten due to controller
5393	 * reset from the other port. Set the appropriate LAA in RAR[0]
5394	 */
5395	if (e1000e_get_laa_state_82571(hw))
5396		hw->mac.ops.rar_set(hw, adapter->hw.mac.addr, 0);
5397
5398	if (adapter->flags2 & FLAG2_CHECK_PHY_HANG)
5399		e1000e_check_82574_phy_workaround(adapter);
5400
5401	/* Clear valid timestamp stuck in RXSTMPL/H due to a Rx error */
5402	if (adapter->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE) {
5403		if ((adapter->flags2 & FLAG2_CHECK_RX_HWTSTAMP) &&
5404		    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID)) {
5405			er32(RXSTMPH);
5406			adapter->rx_hwtstamp_cleared++;
5407		} else {
5408			adapter->flags2 |= FLAG2_CHECK_RX_HWTSTAMP;
5409		}
5410	}
5411
5412	/* Reset the timer */
5413	if (!test_bit(__E1000_DOWN, &adapter->state))
5414		mod_timer(&adapter->watchdog_timer,
5415			  round_jiffies(jiffies + 2 * HZ));
5416}
5417
5418#define E1000_TX_FLAGS_CSUM		0x00000001
5419#define E1000_TX_FLAGS_VLAN		0x00000002
5420#define E1000_TX_FLAGS_TSO		0x00000004
5421#define E1000_TX_FLAGS_IPV4		0x00000008
5422#define E1000_TX_FLAGS_NO_FCS		0x00000010
5423#define E1000_TX_FLAGS_HWTSTAMP		0x00000020
5424#define E1000_TX_FLAGS_VLAN_MASK	0xffff0000
5425#define E1000_TX_FLAGS_VLAN_SHIFT	16
5426
5427static int e1000_tso(struct e1000_ring *tx_ring, struct sk_buff *skb,
5428		     __be16 protocol)
5429{
5430	struct e1000_context_desc *context_desc;
5431	struct e1000_buffer *buffer_info;
5432	unsigned int i;
5433	u32 cmd_length = 0;
5434	u16 ipcse = 0, mss;
5435	u8 ipcss, ipcso, tucss, tucso, hdr_len;
5436	int err;
5437
5438	if (!skb_is_gso(skb))
5439		return 0;
5440
5441	err = skb_cow_head(skb, 0);
5442	if (err < 0)
5443		return err;
5444
5445	hdr_len = skb_tcp_all_headers(skb);
5446	mss = skb_shinfo(skb)->gso_size;
5447	if (protocol == htons(ETH_P_IP)) {
5448		struct iphdr *iph = ip_hdr(skb);
5449		iph->tot_len = 0;
5450		iph->check = 0;
5451		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
5452							 0, IPPROTO_TCP, 0);
5453		cmd_length = E1000_TXD_CMD_IP;
5454		ipcse = skb_transport_offset(skb) - 1;
5455	} else if (skb_is_gso_v6(skb)) {
5456		tcp_v6_gso_csum_prep(skb);
5457		ipcse = 0;
5458	}
5459	ipcss = skb_network_offset(skb);
5460	ipcso = (void *)&(ip_hdr(skb)->check) - (void *)skb->data;
5461	tucss = skb_transport_offset(skb);
5462	tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
5463
5464	cmd_length |= (E1000_TXD_CMD_DEXT | E1000_TXD_CMD_TSE |
5465		       E1000_TXD_CMD_TCP | (skb->len - (hdr_len)));
5466
5467	i = tx_ring->next_to_use;
5468	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5469	buffer_info = &tx_ring->buffer_info[i];
5470
5471	context_desc->lower_setup.ip_fields.ipcss = ipcss;
5472	context_desc->lower_setup.ip_fields.ipcso = ipcso;
5473	context_desc->lower_setup.ip_fields.ipcse = cpu_to_le16(ipcse);
5474	context_desc->upper_setup.tcp_fields.tucss = tucss;
5475	context_desc->upper_setup.tcp_fields.tucso = tucso;
5476	context_desc->upper_setup.tcp_fields.tucse = 0;
5477	context_desc->tcp_seg_setup.fields.mss = cpu_to_le16(mss);
5478	context_desc->tcp_seg_setup.fields.hdr_len = hdr_len;
5479	context_desc->cmd_and_length = cpu_to_le32(cmd_length);
5480
5481	buffer_info->time_stamp = jiffies;
5482	buffer_info->next_to_watch = i;
5483
5484	i++;
5485	if (i == tx_ring->count)
5486		i = 0;
5487	tx_ring->next_to_use = i;
5488
5489	return 1;
5490}
5491
5492static bool e1000_tx_csum(struct e1000_ring *tx_ring, struct sk_buff *skb,
5493			  __be16 protocol)
5494{
5495	struct e1000_adapter *adapter = tx_ring->adapter;
5496	struct e1000_context_desc *context_desc;
5497	struct e1000_buffer *buffer_info;
5498	unsigned int i;
5499	u8 css;
5500	u32 cmd_len = E1000_TXD_CMD_DEXT;
5501
5502	if (skb->ip_summed != CHECKSUM_PARTIAL)
5503		return false;
5504
5505	switch (protocol) {
5506	case cpu_to_be16(ETH_P_IP):
5507		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
5508			cmd_len |= E1000_TXD_CMD_TCP;
5509		break;
5510	case cpu_to_be16(ETH_P_IPV6):
5511		/* XXX not handling all IPV6 headers */
5512		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
5513			cmd_len |= E1000_TXD_CMD_TCP;
5514		break;
5515	default:
5516		if (unlikely(net_ratelimit()))
5517			e_warn("checksum_partial proto=%x!\n",
5518			       be16_to_cpu(protocol));
5519		break;
5520	}
5521
5522	css = skb_checksum_start_offset(skb);
5523
5524	i = tx_ring->next_to_use;
5525	buffer_info = &tx_ring->buffer_info[i];
5526	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5527
5528	context_desc->lower_setup.ip_config = 0;
5529	context_desc->upper_setup.tcp_fields.tucss = css;
5530	context_desc->upper_setup.tcp_fields.tucso = css + skb->csum_offset;
5531	context_desc->upper_setup.tcp_fields.tucse = 0;
5532	context_desc->tcp_seg_setup.data = 0;
5533	context_desc->cmd_and_length = cpu_to_le32(cmd_len);
5534
5535	buffer_info->time_stamp = jiffies;
5536	buffer_info->next_to_watch = i;
5537
5538	i++;
5539	if (i == tx_ring->count)
5540		i = 0;
5541	tx_ring->next_to_use = i;
5542
5543	return true;
5544}
5545
5546static int e1000_tx_map(struct e1000_ring *tx_ring, struct sk_buff *skb,
5547			unsigned int first, unsigned int max_per_txd,
5548			unsigned int nr_frags)
5549{
5550	struct e1000_adapter *adapter = tx_ring->adapter;
5551	struct pci_dev *pdev = adapter->pdev;
5552	struct e1000_buffer *buffer_info;
5553	unsigned int len = skb_headlen(skb);
5554	unsigned int offset = 0, size, count = 0, i;
5555	unsigned int f, bytecount, segs;
5556
5557	i = tx_ring->next_to_use;
5558
5559	while (len) {
5560		buffer_info = &tx_ring->buffer_info[i];
5561		size = min(len, max_per_txd);
5562
5563		buffer_info->length = size;
5564		buffer_info->time_stamp = jiffies;
5565		buffer_info->next_to_watch = i;
5566		buffer_info->dma = dma_map_single(&pdev->dev,
5567						  skb->data + offset,
5568						  size, DMA_TO_DEVICE);
5569		buffer_info->mapped_as_page = false;
5570		if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5571			goto dma_error;
5572
5573		len -= size;
5574		offset += size;
5575		count++;
5576
5577		if (len) {
5578			i++;
5579			if (i == tx_ring->count)
5580				i = 0;
5581		}
5582	}
5583
5584	for (f = 0; f < nr_frags; f++) {
5585		const skb_frag_t *frag = &skb_shinfo(skb)->frags[f];
5586
5587		len = skb_frag_size(frag);
5588		offset = 0;
5589
5590		while (len) {
5591			i++;
5592			if (i == tx_ring->count)
5593				i = 0;
5594
5595			buffer_info = &tx_ring->buffer_info[i];
5596			size = min(len, max_per_txd);
5597
5598			buffer_info->length = size;
5599			buffer_info->time_stamp = jiffies;
5600			buffer_info->next_to_watch = i;
5601			buffer_info->dma = skb_frag_dma_map(&pdev->dev, frag,
5602							    offset, size,
5603							    DMA_TO_DEVICE);
5604			buffer_info->mapped_as_page = true;
5605			if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5606				goto dma_error;
5607
5608			len -= size;
5609			offset += size;
5610			count++;
5611		}
5612	}
5613
5614	segs = skb_shinfo(skb)->gso_segs ? : 1;
5615	/* multiply data chunks by size of headers */
5616	bytecount = ((segs - 1) * skb_headlen(skb)) + skb->len;
5617
5618	tx_ring->buffer_info[i].skb = skb;
5619	tx_ring->buffer_info[i].segs = segs;
5620	tx_ring->buffer_info[i].bytecount = bytecount;
5621	tx_ring->buffer_info[first].next_to_watch = i;
5622
5623	return count;
5624
5625dma_error:
5626	dev_err(&pdev->dev, "Tx DMA map failed\n");
5627	buffer_info->dma = 0;
5628	if (count)
5629		count--;
5630
5631	while (count--) {
5632		if (i == 0)
5633			i += tx_ring->count;
5634		i--;
5635		buffer_info = &tx_ring->buffer_info[i];
5636		e1000_put_txbuf(tx_ring, buffer_info, true);
5637	}
5638
5639	return 0;
5640}
5641
5642static void e1000_tx_queue(struct e1000_ring *tx_ring, int tx_flags, int count)
5643{
5644	struct e1000_adapter *adapter = tx_ring->adapter;
5645	struct e1000_tx_desc *tx_desc = NULL;
5646	struct e1000_buffer *buffer_info;
5647	u32 txd_upper = 0, txd_lower = E1000_TXD_CMD_IFCS;
5648	unsigned int i;
5649
5650	if (tx_flags & E1000_TX_FLAGS_TSO) {
5651		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D |
5652		    E1000_TXD_CMD_TSE;
5653		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5654
5655		if (tx_flags & E1000_TX_FLAGS_IPV4)
5656			txd_upper |= E1000_TXD_POPTS_IXSM << 8;
5657	}
5658
5659	if (tx_flags & E1000_TX_FLAGS_CSUM) {
5660		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5661		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5662	}
5663
5664	if (tx_flags & E1000_TX_FLAGS_VLAN) {
5665		txd_lower |= E1000_TXD_CMD_VLE;
5666		txd_upper |= (tx_flags & E1000_TX_FLAGS_VLAN_MASK);
5667	}
5668
5669	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5670		txd_lower &= ~(E1000_TXD_CMD_IFCS);
5671
5672	if (unlikely(tx_flags & E1000_TX_FLAGS_HWTSTAMP)) {
5673		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5674		txd_upper |= E1000_TXD_EXTCMD_TSTAMP;
5675	}
5676
5677	i = tx_ring->next_to_use;
5678
5679	do {
5680		buffer_info = &tx_ring->buffer_info[i];
5681		tx_desc = E1000_TX_DESC(*tx_ring, i);
5682		tx_desc->buffer_addr = cpu_to_le64(buffer_info->dma);
5683		tx_desc->lower.data = cpu_to_le32(txd_lower |
5684						  buffer_info->length);
5685		tx_desc->upper.data = cpu_to_le32(txd_upper);
5686
5687		i++;
5688		if (i == tx_ring->count)
5689			i = 0;
5690	} while (--count > 0);
5691
5692	tx_desc->lower.data |= cpu_to_le32(adapter->txd_cmd);
5693
5694	/* txd_cmd re-enables FCS, so we'll re-disable it here as desired. */
5695	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5696		tx_desc->lower.data &= ~(cpu_to_le32(E1000_TXD_CMD_IFCS));
5697
5698	/* Force memory writes to complete before letting h/w
5699	 * know there are new descriptors to fetch.  (Only
5700	 * applicable for weak-ordered memory model archs,
5701	 * such as IA-64).
5702	 */
5703	wmb();
5704
5705	tx_ring->next_to_use = i;
5706}
5707
5708#define MINIMUM_DHCP_PACKET_SIZE 282
5709static int e1000_transfer_dhcp_info(struct e1000_adapter *adapter,
5710				    struct sk_buff *skb)
5711{
5712	struct e1000_hw *hw = &adapter->hw;
5713	u16 length, offset;
5714
5715	if (skb_vlan_tag_present(skb) &&
5716	    !((skb_vlan_tag_get(skb) == adapter->hw.mng_cookie.vlan_id) &&
5717	      (adapter->hw.mng_cookie.status &
5718	       E1000_MNG_DHCP_COOKIE_STATUS_VLAN)))
5719		return 0;
5720
5721	if (skb->len <= MINIMUM_DHCP_PACKET_SIZE)
5722		return 0;
5723
5724	if (((struct ethhdr *)skb->data)->h_proto != htons(ETH_P_IP))
5725		return 0;
5726
5727	{
5728		const struct iphdr *ip = (struct iphdr *)((u8 *)skb->data + 14);
5729		struct udphdr *udp;
5730
5731		if (ip->protocol != IPPROTO_UDP)
5732			return 0;
5733
5734		udp = (struct udphdr *)((u8 *)ip + (ip->ihl << 2));
5735		if (ntohs(udp->dest) != 67)
5736			return 0;
5737
5738		offset = (u8 *)udp + 8 - skb->data;
5739		length = skb->len - offset;
5740		return e1000e_mng_write_dhcp_info(hw, (u8 *)udp + 8, length);
5741	}
5742
5743	return 0;
5744}
5745
5746static int __e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5747{
5748	struct e1000_adapter *adapter = tx_ring->adapter;
5749
5750	netif_stop_queue(adapter->netdev);
5751	/* Herbert's original patch had:
5752	 *  smp_mb__after_netif_stop_queue();
5753	 * but since that doesn't exist yet, just open code it.
5754	 */
5755	smp_mb();
5756
5757	/* We need to check again in a case another CPU has just
5758	 * made room available.
5759	 */
5760	if (e1000_desc_unused(tx_ring) < size)
5761		return -EBUSY;
5762
5763	/* A reprieve! */
5764	netif_start_queue(adapter->netdev);
5765	++adapter->restart_queue;
5766	return 0;
5767}
5768
5769static int e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5770{
5771	BUG_ON(size > tx_ring->count);
5772
5773	if (e1000_desc_unused(tx_ring) >= size)
5774		return 0;
5775	return __e1000_maybe_stop_tx(tx_ring, size);
5776}
5777
5778static netdev_tx_t e1000_xmit_frame(struct sk_buff *skb,
5779				    struct net_device *netdev)
5780{
5781	struct e1000_adapter *adapter = netdev_priv(netdev);
5782	struct e1000_ring *tx_ring = adapter->tx_ring;
5783	unsigned int first;
5784	unsigned int tx_flags = 0;
5785	unsigned int len = skb_headlen(skb);
5786	unsigned int nr_frags;
5787	unsigned int mss;
5788	int count = 0;
5789	int tso;
5790	unsigned int f;
5791	__be16 protocol = vlan_get_protocol(skb);
5792
5793	if (test_bit(__E1000_DOWN, &adapter->state)) {
5794		dev_kfree_skb_any(skb);
5795		return NETDEV_TX_OK;
5796	}
5797
5798	if (skb->len <= 0) {
5799		dev_kfree_skb_any(skb);
5800		return NETDEV_TX_OK;
5801	}
5802
5803	/* The minimum packet size with TCTL.PSP set is 17 bytes so
5804	 * pad skb in order to meet this minimum size requirement
5805	 */
5806	if (skb_put_padto(skb, 17))
5807		return NETDEV_TX_OK;
5808
5809	mss = skb_shinfo(skb)->gso_size;
5810	if (mss) {
5811		u8 hdr_len;
5812
5813		/* TSO Workaround for 82571/2/3 Controllers -- if skb->data
5814		 * points to just header, pull a few bytes of payload from
5815		 * frags into skb->data
5816		 */
5817		hdr_len = skb_tcp_all_headers(skb);
5818		/* we do this workaround for ES2LAN, but it is un-necessary,
5819		 * avoiding it could save a lot of cycles
5820		 */
5821		if (skb->data_len && (hdr_len == len)) {
5822			unsigned int pull_size;
5823
5824			pull_size = min_t(unsigned int, 4, skb->data_len);
5825			if (!__pskb_pull_tail(skb, pull_size)) {
5826				e_err("__pskb_pull_tail failed.\n");
5827				dev_kfree_skb_any(skb);
5828				return NETDEV_TX_OK;
5829			}
5830			len = skb_headlen(skb);
5831		}
5832	}
5833
5834	/* reserve a descriptor for the offload context */
5835	if ((mss) || (skb->ip_summed == CHECKSUM_PARTIAL))
5836		count++;
5837	count++;
5838
5839	count += DIV_ROUND_UP(len, adapter->tx_fifo_limit);
5840
5841	nr_frags = skb_shinfo(skb)->nr_frags;
5842	for (f = 0; f < nr_frags; f++)
5843		count += DIV_ROUND_UP(skb_frag_size(&skb_shinfo(skb)->frags[f]),
5844				      adapter->tx_fifo_limit);
5845
5846	if (adapter->hw.mac.tx_pkt_filtering)
5847		e1000_transfer_dhcp_info(adapter, skb);
5848
5849	/* need: count + 2 desc gap to keep tail from touching
5850	 * head, otherwise try next time
5851	 */
5852	if (e1000_maybe_stop_tx(tx_ring, count + 2))
5853		return NETDEV_TX_BUSY;
5854
5855	if (skb_vlan_tag_present(skb)) {
5856		tx_flags |= E1000_TX_FLAGS_VLAN;
5857		tx_flags |= (skb_vlan_tag_get(skb) <<
5858			     E1000_TX_FLAGS_VLAN_SHIFT);
5859	}
5860
5861	first = tx_ring->next_to_use;
5862
5863	tso = e1000_tso(tx_ring, skb, protocol);
5864	if (tso < 0) {
5865		dev_kfree_skb_any(skb);
5866		return NETDEV_TX_OK;
5867	}
5868
5869	if (tso)
5870		tx_flags |= E1000_TX_FLAGS_TSO;
5871	else if (e1000_tx_csum(tx_ring, skb, protocol))
5872		tx_flags |= E1000_TX_FLAGS_CSUM;
5873
5874	/* Old method was to assume IPv4 packet by default if TSO was enabled.
5875	 * 82571 hardware supports TSO capabilities for IPv6 as well...
5876	 * no longer assume, we must.
5877	 */
5878	if (protocol == htons(ETH_P_IP))
5879		tx_flags |= E1000_TX_FLAGS_IPV4;
5880
5881	if (unlikely(skb->no_fcs))
5882		tx_flags |= E1000_TX_FLAGS_NO_FCS;
5883
5884	/* if count is 0 then mapping error has occurred */
5885	count = e1000_tx_map(tx_ring, skb, first, adapter->tx_fifo_limit,
5886			     nr_frags);
5887	if (count) {
5888		if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
5889		    (adapter->flags & FLAG_HAS_HW_TIMESTAMP)) {
5890			if (!adapter->tx_hwtstamp_skb) {
5891				skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5892				tx_flags |= E1000_TX_FLAGS_HWTSTAMP;
5893				adapter->tx_hwtstamp_skb = skb_get(skb);
5894				adapter->tx_hwtstamp_start = jiffies;
5895				schedule_work(&adapter->tx_hwtstamp_work);
5896			} else {
5897				adapter->tx_hwtstamp_skipped++;
5898			}
5899		}
5900
5901		skb_tx_timestamp(skb);
5902
5903		netdev_sent_queue(netdev, skb->len);
5904		e1000_tx_queue(tx_ring, tx_flags, count);
5905		/* Make sure there is space in the ring for the next send. */
5906		e1000_maybe_stop_tx(tx_ring,
5907				    ((MAX_SKB_FRAGS + 1) *
5908				     DIV_ROUND_UP(PAGE_SIZE,
5909						  adapter->tx_fifo_limit) + 4));
5910
5911		if (!netdev_xmit_more() ||
5912		    netif_xmit_stopped(netdev_get_tx_queue(netdev, 0))) {
5913			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
5914				e1000e_update_tdt_wa(tx_ring,
5915						     tx_ring->next_to_use);
5916			else
5917				writel(tx_ring->next_to_use, tx_ring->tail);
5918		}
5919	} else {
5920		dev_kfree_skb_any(skb);
5921		tx_ring->buffer_info[first].time_stamp = 0;
5922		tx_ring->next_to_use = first;
5923	}
5924
5925	return NETDEV_TX_OK;
5926}
5927
5928/**
5929 * e1000_tx_timeout - Respond to a Tx Hang
5930 * @netdev: network interface device structure
5931 * @txqueue: index of the hung queue (unused)
5932 **/
5933static void e1000_tx_timeout(struct net_device *netdev, unsigned int __always_unused txqueue)
5934{
5935	struct e1000_adapter *adapter = netdev_priv(netdev);
5936
5937	/* Do the reset outside of interrupt context */
5938	adapter->tx_timeout_count++;
5939	schedule_work(&adapter->reset_task);
5940}
5941
5942static void e1000_reset_task(struct work_struct *work)
5943{
5944	struct e1000_adapter *adapter;
5945	adapter = container_of(work, struct e1000_adapter, reset_task);
5946
5947	rtnl_lock();
5948	/* don't run the task if already down */
5949	if (test_bit(__E1000_DOWN, &adapter->state)) {
5950		rtnl_unlock();
5951		return;
5952	}
5953
5954	if (!(adapter->flags & FLAG_RESTART_NOW)) {
5955		e1000e_dump(adapter);
5956		e_err("Reset adapter unexpectedly\n");
5957	}
5958	e1000e_reinit_locked(adapter);
5959	rtnl_unlock();
5960}
5961
5962/**
5963 * e1000e_get_stats64 - Get System Network Statistics
5964 * @netdev: network interface device structure
5965 * @stats: rtnl_link_stats64 pointer
5966 *
5967 * Returns the address of the device statistics structure.
5968 **/
5969void e1000e_get_stats64(struct net_device *netdev,
5970			struct rtnl_link_stats64 *stats)
5971{
5972	struct e1000_adapter *adapter = netdev_priv(netdev);
5973
5974	spin_lock(&adapter->stats64_lock);
5975	e1000e_update_stats(adapter);
5976	/* Fill out the OS statistics structure */
5977	stats->rx_bytes = adapter->stats.gorc;
5978	stats->rx_packets = adapter->stats.gprc;
5979	stats->tx_bytes = adapter->stats.gotc;
5980	stats->tx_packets = adapter->stats.gptc;
5981	stats->multicast = adapter->stats.mprc;
5982	stats->collisions = adapter->stats.colc;
5983
5984	/* Rx Errors */
5985
5986	/* RLEC on some newer hardware can be incorrect so build
5987	 * our own version based on RUC and ROC
5988	 */
5989	stats->rx_errors = adapter->stats.rxerrc +
5990	    adapter->stats.crcerrs + adapter->stats.algnerrc +
5991	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5992	stats->rx_length_errors = adapter->stats.ruc + adapter->stats.roc;
5993	stats->rx_crc_errors = adapter->stats.crcerrs;
5994	stats->rx_frame_errors = adapter->stats.algnerrc;
5995	stats->rx_missed_errors = adapter->stats.mpc;
5996
5997	/* Tx Errors */
5998	stats->tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5999	stats->tx_aborted_errors = adapter->stats.ecol;
6000	stats->tx_window_errors = adapter->stats.latecol;
6001	stats->tx_carrier_errors = adapter->stats.tncrs;
6002
6003	/* Tx Dropped needs to be maintained elsewhere */
6004
6005	spin_unlock(&adapter->stats64_lock);
6006}
6007
6008/**
6009 * e1000_change_mtu - Change the Maximum Transfer Unit
6010 * @netdev: network interface device structure
6011 * @new_mtu: new value for maximum frame size
6012 *
6013 * Returns 0 on success, negative on failure
6014 **/
6015static int e1000_change_mtu(struct net_device *netdev, int new_mtu)
6016{
6017	struct e1000_adapter *adapter = netdev_priv(netdev);
6018	int max_frame = new_mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
6019
6020	/* Jumbo frame support */
6021	if ((new_mtu > ETH_DATA_LEN) &&
6022	    !(adapter->flags & FLAG_HAS_JUMBO_FRAMES)) {
6023		e_err("Jumbo Frames not supported.\n");
6024		return -EINVAL;
6025	}
6026
6027	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
6028	if ((adapter->hw.mac.type >= e1000_pch2lan) &&
6029	    !(adapter->flags2 & FLAG2_CRC_STRIPPING) &&
6030	    (new_mtu > ETH_DATA_LEN)) {
6031		e_err("Jumbo Frames not supported on this device when CRC stripping is disabled.\n");
6032		return -EINVAL;
6033	}
6034
6035	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
6036		usleep_range(1000, 1100);
6037	/* e1000e_down -> e1000e_reset dependent on max_frame_size & mtu */
6038	adapter->max_frame_size = max_frame;
6039	netdev_dbg(netdev, "changing MTU from %d to %d\n",
6040		   netdev->mtu, new_mtu);
6041	netdev->mtu = new_mtu;
6042
6043	pm_runtime_get_sync(netdev->dev.parent);
6044
6045	if (netif_running(netdev))
6046		e1000e_down(adapter, true);
6047
6048	/* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
6049	 * means we reserve 2 more, this pushes us to allocate from the next
6050	 * larger slab size.
6051	 * i.e. RXBUFFER_2048 --> size-4096 slab
6052	 * However with the new *_jumbo_rx* routines, jumbo receives will use
6053	 * fragmented skbs
6054	 */
6055
6056	if (max_frame <= 2048)
6057		adapter->rx_buffer_len = 2048;
6058	else
6059		adapter->rx_buffer_len = 4096;
6060
6061	/* adjust allocation if LPE protects us, and we aren't using SBP */
6062	if (max_frame <= (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN))
6063		adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
6064
6065	if (netif_running(netdev))
6066		e1000e_up(adapter);
6067	else
6068		e1000e_reset(adapter);
6069
6070	pm_runtime_put_sync(netdev->dev.parent);
6071
6072	clear_bit(__E1000_RESETTING, &adapter->state);
6073
6074	return 0;
6075}
6076
6077static int e1000_mii_ioctl(struct net_device *netdev, struct ifreq *ifr,
6078			   int cmd)
6079{
6080	struct e1000_adapter *adapter = netdev_priv(netdev);
6081	struct mii_ioctl_data *data = if_mii(ifr);
6082
6083	if (adapter->hw.phy.media_type != e1000_media_type_copper)
6084		return -EOPNOTSUPP;
6085
6086	switch (cmd) {
6087	case SIOCGMIIPHY:
6088		data->phy_id = adapter->hw.phy.addr;
6089		break;
6090	case SIOCGMIIREG:
6091		e1000_phy_read_status(adapter);
6092
6093		switch (data->reg_num & 0x1F) {
6094		case MII_BMCR:
6095			data->val_out = adapter->phy_regs.bmcr;
6096			break;
6097		case MII_BMSR:
6098			data->val_out = adapter->phy_regs.bmsr;
6099			break;
6100		case MII_PHYSID1:
6101			data->val_out = (adapter->hw.phy.id >> 16);
6102			break;
6103		case MII_PHYSID2:
6104			data->val_out = (adapter->hw.phy.id & 0xFFFF);
6105			break;
6106		case MII_ADVERTISE:
6107			data->val_out = adapter->phy_regs.advertise;
6108			break;
6109		case MII_LPA:
6110			data->val_out = adapter->phy_regs.lpa;
6111			break;
6112		case MII_EXPANSION:
6113			data->val_out = adapter->phy_regs.expansion;
6114			break;
6115		case MII_CTRL1000:
6116			data->val_out = adapter->phy_regs.ctrl1000;
6117			break;
6118		case MII_STAT1000:
6119			data->val_out = adapter->phy_regs.stat1000;
6120			break;
6121		case MII_ESTATUS:
6122			data->val_out = adapter->phy_regs.estatus;
6123			break;
6124		default:
6125			return -EIO;
6126		}
6127		break;
6128	case SIOCSMIIREG:
6129	default:
6130		return -EOPNOTSUPP;
6131	}
6132	return 0;
6133}
6134
6135/**
6136 * e1000e_hwtstamp_set - control hardware time stamping
6137 * @netdev: network interface device structure
6138 * @ifr: interface request
6139 *
6140 * Outgoing time stamping can be enabled and disabled. Play nice and
6141 * disable it when requested, although it shouldn't cause any overhead
6142 * when no packet needs it. At most one packet in the queue may be
6143 * marked for time stamping, otherwise it would be impossible to tell
6144 * for sure to which packet the hardware time stamp belongs.
6145 *
6146 * Incoming time stamping has to be configured via the hardware filters.
6147 * Not all combinations are supported, in particular event type has to be
6148 * specified. Matching the kind of event packet is not supported, with the
6149 * exception of "all V2 events regardless of level 2 or 4".
6150 **/
6151static int e1000e_hwtstamp_set(struct net_device *netdev, struct ifreq *ifr)
6152{
6153	struct e1000_adapter *adapter = netdev_priv(netdev);
6154	struct hwtstamp_config config;
6155	int ret_val;
6156
6157	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
6158		return -EFAULT;
6159
6160	ret_val = e1000e_config_hwtstamp(adapter, &config);
6161	if (ret_val)
6162		return ret_val;
6163
6164	switch (config.rx_filter) {
6165	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
6166	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
6167	case HWTSTAMP_FILTER_PTP_V2_SYNC:
6168	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
6169	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
6170	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
6171		/* With V2 type filters which specify a Sync or Delay Request,
6172		 * Path Delay Request/Response messages are also time stamped
6173		 * by hardware so notify the caller the requested packets plus
6174		 * some others are time stamped.
6175		 */
6176		config.rx_filter = HWTSTAMP_FILTER_SOME;
6177		break;
6178	default:
6179		break;
6180	}
6181
6182	return copy_to_user(ifr->ifr_data, &config,
6183			    sizeof(config)) ? -EFAULT : 0;
6184}
6185
6186static int e1000e_hwtstamp_get(struct net_device *netdev, struct ifreq *ifr)
6187{
6188	struct e1000_adapter *adapter = netdev_priv(netdev);
6189
6190	return copy_to_user(ifr->ifr_data, &adapter->hwtstamp_config,
6191			    sizeof(adapter->hwtstamp_config)) ? -EFAULT : 0;
6192}
6193
6194static int e1000_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
6195{
6196	switch (cmd) {
6197	case SIOCGMIIPHY:
6198	case SIOCGMIIREG:
6199	case SIOCSMIIREG:
6200		return e1000_mii_ioctl(netdev, ifr, cmd);
6201	case SIOCSHWTSTAMP:
6202		return e1000e_hwtstamp_set(netdev, ifr);
6203	case SIOCGHWTSTAMP:
6204		return e1000e_hwtstamp_get(netdev, ifr);
6205	default:
6206		return -EOPNOTSUPP;
6207	}
6208}
6209
6210static int e1000_init_phy_wakeup(struct e1000_adapter *adapter, u32 wufc)
6211{
6212	struct e1000_hw *hw = &adapter->hw;
6213	u32 i, mac_reg, wuc;
6214	u16 phy_reg, wuc_enable;
6215	int retval;
6216
6217	/* copy MAC RARs to PHY RARs */
6218	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
6219
6220	retval = hw->phy.ops.acquire(hw);
6221	if (retval) {
6222		e_err("Could not acquire PHY\n");
6223		return retval;
6224	}
6225
6226	/* Enable access to wakeup registers on and set page to BM_WUC_PAGE */
6227	retval = e1000_enable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6228	if (retval)
6229		goto release;
6230
6231	/* copy MAC MTA to PHY MTA - only needed for pchlan */
6232	for (i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
6233		mac_reg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
6234		hw->phy.ops.write_reg_page(hw, BM_MTA(i),
6235					   (u16)(mac_reg & 0xFFFF));
6236		hw->phy.ops.write_reg_page(hw, BM_MTA(i) + 1,
6237					   (u16)((mac_reg >> 16) & 0xFFFF));
6238	}
6239
6240	/* configure PHY Rx Control register */
6241	hw->phy.ops.read_reg_page(&adapter->hw, BM_RCTL, &phy_reg);
6242	mac_reg = er32(RCTL);
6243	if (mac_reg & E1000_RCTL_UPE)
6244		phy_reg |= BM_RCTL_UPE;
6245	if (mac_reg & E1000_RCTL_MPE)
6246		phy_reg |= BM_RCTL_MPE;
6247	phy_reg &= ~(BM_RCTL_MO_MASK);
6248	if (mac_reg & E1000_RCTL_MO_3)
6249		phy_reg |= (FIELD_GET(E1000_RCTL_MO_3, mac_reg)
6250			    << BM_RCTL_MO_SHIFT);
6251	if (mac_reg & E1000_RCTL_BAM)
6252		phy_reg |= BM_RCTL_BAM;
6253	if (mac_reg & E1000_RCTL_PMCF)
6254		phy_reg |= BM_RCTL_PMCF;
6255	mac_reg = er32(CTRL);
6256	if (mac_reg & E1000_CTRL_RFCE)
6257		phy_reg |= BM_RCTL_RFCE;
6258	hw->phy.ops.write_reg_page(&adapter->hw, BM_RCTL, phy_reg);
6259
6260	wuc = E1000_WUC_PME_EN;
6261	if (wufc & (E1000_WUFC_MAG | E1000_WUFC_LNKC))
6262		wuc |= E1000_WUC_APME;
6263
6264	/* enable PHY wakeup in MAC register */
6265	ew32(WUFC, wufc);
6266	ew32(WUC, (E1000_WUC_PHY_WAKE | E1000_WUC_APMPME |
6267		   E1000_WUC_PME_STATUS | wuc));
6268
6269	/* configure and enable PHY wakeup in PHY registers */
6270	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUFC, wufc);
6271	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUC, wuc);
6272
6273	/* activate PHY wakeup */
6274	wuc_enable |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
6275	retval = e1000_disable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6276	if (retval)
6277		e_err("Could not set PHY Host Wakeup bit\n");
6278release:
6279	hw->phy.ops.release(hw);
6280
6281	return retval;
6282}
6283
6284static void e1000e_flush_lpic(struct pci_dev *pdev)
6285{
6286	struct net_device *netdev = pci_get_drvdata(pdev);
6287	struct e1000_adapter *adapter = netdev_priv(netdev);
6288	struct e1000_hw *hw = &adapter->hw;
6289	u32 ret_val;
6290
6291	pm_runtime_get_sync(netdev->dev.parent);
6292
6293	ret_val = hw->phy.ops.acquire(hw);
6294	if (ret_val)
6295		goto fl_out;
6296
6297	pr_info("EEE TX LPI TIMER: %08X\n",
6298		er32(LPIC) >> E1000_LPIC_LPIET_SHIFT);
6299
6300	hw->phy.ops.release(hw);
6301
6302fl_out:
6303	pm_runtime_put_sync(netdev->dev.parent);
6304}
6305
6306/* S0ix implementation */
6307static void e1000e_s0ix_entry_flow(struct e1000_adapter *adapter)
6308{
6309	struct e1000_hw *hw = &adapter->hw;
6310	u32 mac_data;
6311	u16 phy_data;
6312
6313	if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID &&
6314	    hw->mac.type >= e1000_pch_adp) {
6315		/* Request ME configure the device for S0ix */
6316		mac_data = er32(H2ME);
6317		mac_data |= E1000_H2ME_START_DPG;
6318		mac_data &= ~E1000_H2ME_EXIT_DPG;
6319		trace_e1000e_trace_mac_register(mac_data);
6320		ew32(H2ME, mac_data);
6321	} else {
6322		/* Request driver configure the device to S0ix */
6323		/* Disable the periodic inband message,
6324		 * don't request PCIe clock in K1 page770_17[10:9] = 10b
6325		 */
6326		e1e_rphy(hw, HV_PM_CTRL, &phy_data);
6327		phy_data &= ~HV_PM_CTRL_K1_CLK_REQ;
6328		phy_data |= BIT(10);
6329		e1e_wphy(hw, HV_PM_CTRL, phy_data);
6330
6331		/* Make sure we don't exit K1 every time a new packet arrives
6332		 * 772_29[5] = 1 CS_Mode_Stay_In_K1
6333		 */
6334		e1e_rphy(hw, I217_CGFREG, &phy_data);
6335		phy_data |= BIT(5);
6336		e1e_wphy(hw, I217_CGFREG, phy_data);
6337
6338		/* Change the MAC/PHY interface to SMBus
6339		 * Force the SMBus in PHY page769_23[0] = 1
6340		 * Force the SMBus in MAC CTRL_EXT[11] = 1
6341		 */
6342		e1e_rphy(hw, CV_SMB_CTRL, &phy_data);
6343		phy_data |= CV_SMB_CTRL_FORCE_SMBUS;
6344		e1e_wphy(hw, CV_SMB_CTRL, phy_data);
6345		mac_data = er32(CTRL_EXT);
6346		mac_data |= E1000_CTRL_EXT_FORCE_SMBUS;
6347		ew32(CTRL_EXT, mac_data);
6348
6349		/* DFT control: PHY bit: page769_20[0] = 1
6350		 * page769_20[7] - PHY PLL stop
6351		 * page769_20[8] - PHY go to the electrical idle
6352		 * page769_20[9] - PHY serdes disable
6353		 * Gate PPW via EXTCNF_CTRL - set 0x0F00[7] = 1
6354		 */
6355		e1e_rphy(hw, I82579_DFT_CTRL, &phy_data);
6356		phy_data |= BIT(0);
6357		phy_data |= BIT(7);
6358		phy_data |= BIT(8);
6359		phy_data |= BIT(9);
6360		e1e_wphy(hw, I82579_DFT_CTRL, phy_data);
6361
6362		mac_data = er32(EXTCNF_CTRL);
6363		mac_data |= E1000_EXTCNF_CTRL_GATE_PHY_CFG;
6364		ew32(EXTCNF_CTRL, mac_data);
6365
6366		/* Enable the Dynamic Power Gating in the MAC */
6367		mac_data = er32(FEXTNVM7);
6368		mac_data |= BIT(22);
6369		ew32(FEXTNVM7, mac_data);
6370
6371		/* Disable disconnected cable conditioning for Power Gating */
6372		mac_data = er32(DPGFR);
6373		mac_data |= BIT(2);
6374		ew32(DPGFR, mac_data);
6375
6376		/* Don't wake from dynamic Power Gating with clock request */
6377		mac_data = er32(FEXTNVM12);
6378		mac_data |= BIT(12);
6379		ew32(FEXTNVM12, mac_data);
6380
6381		/* Ungate PGCB clock */
6382		mac_data = er32(FEXTNVM9);
6383		mac_data &= ~BIT(28);
6384		ew32(FEXTNVM9, mac_data);
6385
6386		/* Enable K1 off to enable mPHY Power Gating */
6387		mac_data = er32(FEXTNVM6);
6388		mac_data |= BIT(31);
6389		ew32(FEXTNVM6, mac_data);
6390
6391		/* Enable mPHY power gating for any link and speed */
6392		mac_data = er32(FEXTNVM8);
6393		mac_data |= BIT(9);
6394		ew32(FEXTNVM8, mac_data);
6395
6396		/* Enable the Dynamic Clock Gating in the DMA and MAC */
6397		mac_data = er32(CTRL_EXT);
6398		mac_data |= E1000_CTRL_EXT_DMA_DYN_CLK_EN;
6399		ew32(CTRL_EXT, mac_data);
6400
6401		/* No MAC DPG gating SLP_S0 in modern standby
6402		 * Switch the logic of the lanphypc to use PMC counter
6403		 */
6404		mac_data = er32(FEXTNVM5);
6405		mac_data |= BIT(7);
6406		ew32(FEXTNVM5, mac_data);
6407	}
6408
6409	/* Disable the time synchronization clock */
6410	mac_data = er32(FEXTNVM7);
6411	mac_data |= BIT(31);
6412	mac_data &= ~BIT(0);
6413	ew32(FEXTNVM7, mac_data);
6414
6415	/* Dynamic Power Gating Enable */
6416	mac_data = er32(CTRL_EXT);
6417	mac_data |= BIT(3);
6418	ew32(CTRL_EXT, mac_data);
6419
6420	/* Check MAC Tx/Rx packet buffer pointers.
6421	 * Reset MAC Tx/Rx packet buffer pointers to suppress any
6422	 * pending traffic indication that would prevent power gating.
6423	 */
6424	mac_data = er32(TDFH);
6425	if (mac_data)
6426		ew32(TDFH, 0);
6427	mac_data = er32(TDFT);
6428	if (mac_data)
6429		ew32(TDFT, 0);
6430	mac_data = er32(TDFHS);
6431	if (mac_data)
6432		ew32(TDFHS, 0);
6433	mac_data = er32(TDFTS);
6434	if (mac_data)
6435		ew32(TDFTS, 0);
6436	mac_data = er32(TDFPC);
6437	if (mac_data)
6438		ew32(TDFPC, 0);
6439	mac_data = er32(RDFH);
6440	if (mac_data)
6441		ew32(RDFH, 0);
6442	mac_data = er32(RDFT);
6443	if (mac_data)
6444		ew32(RDFT, 0);
6445	mac_data = er32(RDFHS);
6446	if (mac_data)
6447		ew32(RDFHS, 0);
6448	mac_data = er32(RDFTS);
6449	if (mac_data)
6450		ew32(RDFTS, 0);
6451	mac_data = er32(RDFPC);
6452	if (mac_data)
6453		ew32(RDFPC, 0);
6454}
6455
6456static void e1000e_s0ix_exit_flow(struct e1000_adapter *adapter)
6457{
6458	struct e1000_hw *hw = &adapter->hw;
6459	bool firmware_bug = false;
6460	u32 mac_data;
6461	u16 phy_data;
6462	u32 i = 0;
6463
6464	if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID &&
6465	    hw->mac.type >= e1000_pch_adp) {
6466		/* Keep the GPT clock enabled for CSME */
6467		mac_data = er32(FEXTNVM);
6468		mac_data |= BIT(3);
6469		ew32(FEXTNVM, mac_data);
6470		/* Request ME unconfigure the device from S0ix */
6471		mac_data = er32(H2ME);
6472		mac_data &= ~E1000_H2ME_START_DPG;
6473		mac_data |= E1000_H2ME_EXIT_DPG;
6474		trace_e1000e_trace_mac_register(mac_data);
6475		ew32(H2ME, mac_data);
6476
6477		/* Poll up to 2.5 seconds for ME to unconfigure DPG.
6478		 * If this takes more than 1 second, show a warning indicating a
6479		 * firmware bug
6480		 */
6481		while (!(er32(EXFWSM) & E1000_EXFWSM_DPG_EXIT_DONE)) {
6482			if (i > 100 && !firmware_bug)
6483				firmware_bug = true;
6484
6485			if (i++ == 250) {
6486				e_dbg("Timeout (firmware bug): %d msec\n",
6487				      i * 10);
6488				break;
6489			}
6490
6491			usleep_range(10000, 11000);
6492		}
6493		if (firmware_bug)
6494			e_warn("DPG_EXIT_DONE took %d msec. This is a firmware bug\n",
6495			       i * 10);
6496		else
6497			e_dbg("DPG_EXIT_DONE cleared after %d msec\n", i * 10);
6498	} else {
6499		/* Request driver unconfigure the device from S0ix */
6500
6501		/* Disable the Dynamic Power Gating in the MAC */
6502		mac_data = er32(FEXTNVM7);
6503		mac_data &= 0xFFBFFFFF;
6504		ew32(FEXTNVM7, mac_data);
6505
6506		/* Disable mPHY power gating for any link and speed */
6507		mac_data = er32(FEXTNVM8);
6508		mac_data &= ~BIT(9);
6509		ew32(FEXTNVM8, mac_data);
6510
6511		/* Disable K1 off */
6512		mac_data = er32(FEXTNVM6);
6513		mac_data &= ~BIT(31);
6514		ew32(FEXTNVM6, mac_data);
6515
6516		/* Disable Ungate PGCB clock */
6517		mac_data = er32(FEXTNVM9);
6518		mac_data |= BIT(28);
6519		ew32(FEXTNVM9, mac_data);
6520
6521		/* Cancel not waking from dynamic
6522		 * Power Gating with clock request
6523		 */
6524		mac_data = er32(FEXTNVM12);
6525		mac_data &= ~BIT(12);
6526		ew32(FEXTNVM12, mac_data);
6527
6528		/* Cancel disable disconnected cable conditioning
6529		 * for Power Gating
6530		 */
6531		mac_data = er32(DPGFR);
6532		mac_data &= ~BIT(2);
6533		ew32(DPGFR, mac_data);
6534
6535		/* Disable the Dynamic Clock Gating in the DMA and MAC */
6536		mac_data = er32(CTRL_EXT);
6537		mac_data &= 0xFFF7FFFF;
6538		ew32(CTRL_EXT, mac_data);
6539
6540		/* Revert the lanphypc logic to use the internal Gbe counter
6541		 * and not the PMC counter
6542		 */
6543		mac_data = er32(FEXTNVM5);
6544		mac_data &= 0xFFFFFF7F;
6545		ew32(FEXTNVM5, mac_data);
6546
6547		/* Enable the periodic inband message,
6548		 * Request PCIe clock in K1 page770_17[10:9] =01b
6549		 */
6550		e1e_rphy(hw, HV_PM_CTRL, &phy_data);
6551		phy_data &= 0xFBFF;
6552		phy_data |= HV_PM_CTRL_K1_CLK_REQ;
6553		e1e_wphy(hw, HV_PM_CTRL, phy_data);
6554
6555		/* Return back configuration
6556		 * 772_29[5] = 0 CS_Mode_Stay_In_K1
6557		 */
6558		e1e_rphy(hw, I217_CGFREG, &phy_data);
6559		phy_data &= 0xFFDF;
6560		e1e_wphy(hw, I217_CGFREG, phy_data);
6561
6562		/* Change the MAC/PHY interface to Kumeran
6563		 * Unforce the SMBus in PHY page769_23[0] = 0
6564		 * Unforce the SMBus in MAC CTRL_EXT[11] = 0
6565		 */
6566		e1e_rphy(hw, CV_SMB_CTRL, &phy_data);
6567		phy_data &= ~CV_SMB_CTRL_FORCE_SMBUS;
6568		e1e_wphy(hw, CV_SMB_CTRL, phy_data);
6569		mac_data = er32(CTRL_EXT);
6570		mac_data &= ~E1000_CTRL_EXT_FORCE_SMBUS;
6571		ew32(CTRL_EXT, mac_data);
6572	}
6573
6574	/* Disable Dynamic Power Gating */
6575	mac_data = er32(CTRL_EXT);
6576	mac_data &= 0xFFFFFFF7;
6577	ew32(CTRL_EXT, mac_data);
6578
6579	/* Enable the time synchronization clock */
6580	mac_data = er32(FEXTNVM7);
6581	mac_data &= ~BIT(31);
6582	mac_data |= BIT(0);
6583	ew32(FEXTNVM7, mac_data);
6584}
6585
6586static int e1000e_pm_freeze(struct device *dev)
6587{
6588	struct net_device *netdev = dev_get_drvdata(dev);
6589	struct e1000_adapter *adapter = netdev_priv(netdev);
6590	bool present;
6591
6592	rtnl_lock();
6593
6594	present = netif_device_present(netdev);
6595	netif_device_detach(netdev);
6596
6597	if (present && netif_running(netdev)) {
6598		int count = E1000_CHECK_RESET_COUNT;
6599
6600		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6601			usleep_range(10000, 11000);
6602
6603		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6604
6605		/* Quiesce the device without resetting the hardware */
6606		e1000e_down(adapter, false);
6607		e1000_free_irq(adapter);
6608	}
6609	rtnl_unlock();
6610
6611	e1000e_reset_interrupt_capability(adapter);
6612
6613	/* Allow time for pending master requests to run */
6614	e1000e_disable_pcie_master(&adapter->hw);
6615
6616	return 0;
6617}
6618
6619static int __e1000_shutdown(struct pci_dev *pdev, bool runtime)
6620{
6621	struct net_device *netdev = pci_get_drvdata(pdev);
6622	struct e1000_adapter *adapter = netdev_priv(netdev);
6623	struct e1000_hw *hw = &adapter->hw;
6624	u32 ctrl, ctrl_ext, rctl, status, wufc;
6625	int retval = 0;
6626	u16 smb_ctrl;
6627
6628	/* Runtime suspend should only enable wakeup for link changes */
6629	if (runtime)
6630		wufc = E1000_WUFC_LNKC;
6631	else if (device_may_wakeup(&pdev->dev))
6632		wufc = adapter->wol;
6633	else
6634		wufc = 0;
6635
6636	status = er32(STATUS);
6637	if (status & E1000_STATUS_LU)
6638		wufc &= ~E1000_WUFC_LNKC;
6639
6640	if (wufc) {
6641		e1000_setup_rctl(adapter);
6642		e1000e_set_rx_mode(netdev);
6643
6644		/* turn on all-multi mode if wake on multicast is enabled */
6645		if (wufc & E1000_WUFC_MC) {
6646			rctl = er32(RCTL);
6647			rctl |= E1000_RCTL_MPE;
6648			ew32(RCTL, rctl);
6649		}
6650
6651		ctrl = er32(CTRL);
6652		ctrl |= E1000_CTRL_ADVD3WUC;
6653		if (!(adapter->flags2 & FLAG2_HAS_PHY_WAKEUP))
6654			ctrl |= E1000_CTRL_EN_PHY_PWR_MGMT;
6655		ew32(CTRL, ctrl);
6656
6657		if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
6658		    adapter->hw.phy.media_type ==
6659		    e1000_media_type_internal_serdes) {
6660			/* keep the laser running in D3 */
6661			ctrl_ext = er32(CTRL_EXT);
6662			ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
6663			ew32(CTRL_EXT, ctrl_ext);
6664		}
6665
6666		if (!runtime)
6667			e1000e_power_up_phy(adapter);
6668
6669		if (adapter->flags & FLAG_IS_ICH)
6670			e1000_suspend_workarounds_ich8lan(&adapter->hw);
6671
6672		if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6673			/* enable wakeup by the PHY */
6674			retval = e1000_init_phy_wakeup(adapter, wufc);
6675			if (retval)
6676				return retval;
6677		} else {
6678			/* enable wakeup by the MAC */
6679			ew32(WUFC, wufc);
6680			ew32(WUC, E1000_WUC_PME_EN);
6681		}
6682	} else {
6683		ew32(WUC, 0);
6684		ew32(WUFC, 0);
6685
6686		e1000_power_down_phy(adapter);
6687	}
6688
6689	if (adapter->hw.phy.type == e1000_phy_igp_3) {
6690		e1000e_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
6691	} else if (hw->mac.type >= e1000_pch_lpt) {
6692		if (wufc && !(wufc & (E1000_WUFC_EX | E1000_WUFC_MC | E1000_WUFC_BC))) {
6693			/* ULP does not support wake from unicast, multicast
6694			 * or broadcast.
6695			 */
6696			retval = e1000_enable_ulp_lpt_lp(hw, !runtime);
6697			if (retval)
6698				return retval;
6699		}
6700
6701		/* Force SMBUS to allow WOL */
6702		/* Switching PHY interface always returns MDI error
6703		 * so disable retry mechanism to avoid wasting time
6704		 */
6705		e1000e_disable_phy_retry(hw);
6706
6707		e1e_rphy(hw, CV_SMB_CTRL, &smb_ctrl);
6708		smb_ctrl |= CV_SMB_CTRL_FORCE_SMBUS;
6709		e1e_wphy(hw, CV_SMB_CTRL, smb_ctrl);
6710
6711		e1000e_enable_phy_retry(hw);
6712
6713		/* Force SMBus mode in MAC */
6714		ctrl_ext = er32(CTRL_EXT);
6715		ctrl_ext |= E1000_CTRL_EXT_FORCE_SMBUS;
6716		ew32(CTRL_EXT, ctrl_ext);
6717	}
6718
6719	/* Ensure that the appropriate bits are set in LPI_CTRL
6720	 * for EEE in Sx
6721	 */
6722	if ((hw->phy.type >= e1000_phy_i217) &&
6723	    adapter->eee_advert && hw->dev_spec.ich8lan.eee_lp_ability) {
6724		u16 lpi_ctrl = 0;
6725
6726		retval = hw->phy.ops.acquire(hw);
6727		if (!retval) {
6728			retval = e1e_rphy_locked(hw, I82579_LPI_CTRL,
6729						 &lpi_ctrl);
6730			if (!retval) {
6731				if (adapter->eee_advert &
6732				    hw->dev_spec.ich8lan.eee_lp_ability &
6733				    I82579_EEE_100_SUPPORTED)
6734					lpi_ctrl |= I82579_LPI_CTRL_100_ENABLE;
6735				if (adapter->eee_advert &
6736				    hw->dev_spec.ich8lan.eee_lp_ability &
6737				    I82579_EEE_1000_SUPPORTED)
6738					lpi_ctrl |= I82579_LPI_CTRL_1000_ENABLE;
6739
6740				retval = e1e_wphy_locked(hw, I82579_LPI_CTRL,
6741							 lpi_ctrl);
6742			}
6743		}
6744		hw->phy.ops.release(hw);
6745	}
6746
6747	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
6748	 * would have already happened in close and is redundant.
6749	 */
6750	e1000e_release_hw_control(adapter);
6751
6752	pci_clear_master(pdev);
6753
6754	/* The pci-e switch on some quad port adapters will report a
6755	 * correctable error when the MAC transitions from D0 to D3.  To
6756	 * prevent this we need to mask off the correctable errors on the
6757	 * downstream port of the pci-e switch.
6758	 *
6759	 * We don't have the associated upstream bridge while assigning
6760	 * the PCI device into guest. For example, the KVM on power is
6761	 * one of the cases.
6762	 */
6763	if (adapter->flags & FLAG_IS_QUAD_PORT) {
6764		struct pci_dev *us_dev = pdev->bus->self;
6765		u16 devctl;
6766
6767		if (!us_dev)
6768			return 0;
6769
6770		pcie_capability_read_word(us_dev, PCI_EXP_DEVCTL, &devctl);
6771		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL,
6772					   (devctl & ~PCI_EXP_DEVCTL_CERE));
6773
6774		pci_save_state(pdev);
6775		pci_prepare_to_sleep(pdev);
6776
6777		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL, devctl);
6778	}
6779
6780	return 0;
6781}
6782
6783/**
6784 * __e1000e_disable_aspm - Disable ASPM states
6785 * @pdev: pointer to PCI device struct
6786 * @state: bit-mask of ASPM states to disable
6787 * @locked: indication if this context holds pci_bus_sem locked.
6788 *
6789 * Some devices *must* have certain ASPM states disabled per hardware errata.
6790 **/
6791static void __e1000e_disable_aspm(struct pci_dev *pdev, u16 state, int locked)
6792{
6793	struct pci_dev *parent = pdev->bus->self;
6794	u16 aspm_dis_mask = 0;
6795	u16 pdev_aspmc, parent_aspmc;
6796
6797	switch (state) {
6798	case PCIE_LINK_STATE_L0S:
6799	case PCIE_LINK_STATE_L0S | PCIE_LINK_STATE_L1:
6800		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L0S;
6801		fallthrough; /* can't have L1 without L0s */
6802	case PCIE_LINK_STATE_L1:
6803		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L1;
6804		break;
6805	default:
6806		return;
6807	}
6808
6809	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6810	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6811
6812	if (parent) {
6813		pcie_capability_read_word(parent, PCI_EXP_LNKCTL,
6814					  &parent_aspmc);
6815		parent_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6816	}
6817
6818	/* Nothing to do if the ASPM states to be disabled already are */
6819	if (!(pdev_aspmc & aspm_dis_mask) &&
6820	    (!parent || !(parent_aspmc & aspm_dis_mask)))
6821		return;
6822
6823	dev_info(&pdev->dev, "Disabling ASPM %s %s\n",
6824		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L0S) ?
6825		 "L0s" : "",
6826		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L1) ?
6827		 "L1" : "");
6828
6829#ifdef CONFIG_PCIEASPM
6830	if (locked)
6831		pci_disable_link_state_locked(pdev, state);
6832	else
6833		pci_disable_link_state(pdev, state);
6834
6835	/* Double-check ASPM control.  If not disabled by the above, the
6836	 * BIOS is preventing that from happening (or CONFIG_PCIEASPM is
6837	 * not enabled); override by writing PCI config space directly.
6838	 */
6839	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6840	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6841
6842	if (!(aspm_dis_mask & pdev_aspmc))
6843		return;
6844#endif
6845
6846	/* Both device and parent should have the same ASPM setting.
6847	 * Disable ASPM in downstream component first and then upstream.
6848	 */
6849	pcie_capability_clear_word(pdev, PCI_EXP_LNKCTL, aspm_dis_mask);
6850
6851	if (parent)
6852		pcie_capability_clear_word(parent, PCI_EXP_LNKCTL,
6853					   aspm_dis_mask);
6854}
6855
6856/**
6857 * e1000e_disable_aspm - Disable ASPM states.
6858 * @pdev: pointer to PCI device struct
6859 * @state: bit-mask of ASPM states to disable
6860 *
6861 * This function acquires the pci_bus_sem!
6862 * Some devices *must* have certain ASPM states disabled per hardware errata.
6863 **/
6864static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6865{
6866	__e1000e_disable_aspm(pdev, state, 0);
6867}
6868
6869/**
6870 * e1000e_disable_aspm_locked - Disable ASPM states.
6871 * @pdev: pointer to PCI device struct
6872 * @state: bit-mask of ASPM states to disable
6873 *
6874 * This function must be called with pci_bus_sem acquired!
6875 * Some devices *must* have certain ASPM states disabled per hardware errata.
6876 **/
6877static void e1000e_disable_aspm_locked(struct pci_dev *pdev, u16 state)
6878{
6879	__e1000e_disable_aspm(pdev, state, 1);
6880}
6881
6882static int e1000e_pm_thaw(struct device *dev)
6883{
6884	struct net_device *netdev = dev_get_drvdata(dev);
6885	struct e1000_adapter *adapter = netdev_priv(netdev);
6886	int rc = 0;
6887
6888	e1000e_set_interrupt_capability(adapter);
6889
6890	rtnl_lock();
6891	if (netif_running(netdev)) {
6892		rc = e1000_request_irq(adapter);
6893		if (rc)
6894			goto err_irq;
6895
6896		e1000e_up(adapter);
6897	}
6898
6899	netif_device_attach(netdev);
6900err_irq:
6901	rtnl_unlock();
6902
6903	return rc;
6904}
6905
6906static int __e1000_resume(struct pci_dev *pdev)
6907{
6908	struct net_device *netdev = pci_get_drvdata(pdev);
6909	struct e1000_adapter *adapter = netdev_priv(netdev);
6910	struct e1000_hw *hw = &adapter->hw;
6911	u16 aspm_disable_flag = 0;
6912
6913	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6914		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6915	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6916		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6917	if (aspm_disable_flag)
6918		e1000e_disable_aspm(pdev, aspm_disable_flag);
6919
6920	pci_set_master(pdev);
6921
6922	if (hw->mac.type >= e1000_pch2lan)
6923		e1000_resume_workarounds_pchlan(&adapter->hw);
6924
6925	e1000e_power_up_phy(adapter);
6926
6927	/* report the system wakeup cause from S3/S4 */
6928	if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6929		u16 phy_data;
6930
6931		e1e_rphy(&adapter->hw, BM_WUS, &phy_data);
6932		if (phy_data) {
6933			e_info("PHY Wakeup cause - %s\n",
6934			       phy_data & E1000_WUS_EX ? "Unicast Packet" :
6935			       phy_data & E1000_WUS_MC ? "Multicast Packet" :
6936			       phy_data & E1000_WUS_BC ? "Broadcast Packet" :
6937			       phy_data & E1000_WUS_MAG ? "Magic Packet" :
6938			       phy_data & E1000_WUS_LNKC ?
6939			       "Link Status Change" : "other");
6940		}
6941		e1e_wphy(&adapter->hw, BM_WUS, ~0);
6942	} else {
6943		u32 wus = er32(WUS);
6944
6945		if (wus) {
6946			e_info("MAC Wakeup cause - %s\n",
6947			       wus & E1000_WUS_EX ? "Unicast Packet" :
6948			       wus & E1000_WUS_MC ? "Multicast Packet" :
6949			       wus & E1000_WUS_BC ? "Broadcast Packet" :
6950			       wus & E1000_WUS_MAG ? "Magic Packet" :
6951			       wus & E1000_WUS_LNKC ? "Link Status Change" :
6952			       "other");
6953		}
6954		ew32(WUS, ~0);
6955	}
6956
6957	e1000e_reset(adapter);
6958
6959	e1000_init_manageability_pt(adapter);
6960
6961	/* If the controller has AMT, do not set DRV_LOAD until the interface
6962	 * is up.  For all other cases, let the f/w know that the h/w is now
6963	 * under the control of the driver.
6964	 */
6965	if (!(adapter->flags & FLAG_HAS_AMT))
6966		e1000e_get_hw_control(adapter);
6967
6968	return 0;
6969}
6970
6971static __maybe_unused int e1000e_pm_prepare(struct device *dev)
6972{
6973	return pm_runtime_suspended(dev) &&
6974		pm_suspend_via_firmware();
6975}
6976
6977static __maybe_unused int e1000e_pm_suspend(struct device *dev)
6978{
6979	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6980	struct e1000_adapter *adapter = netdev_priv(netdev);
6981	struct pci_dev *pdev = to_pci_dev(dev);
6982	int rc;
6983
6984	e1000e_flush_lpic(pdev);
6985
6986	e1000e_pm_freeze(dev);
6987
6988	rc = __e1000_shutdown(pdev, false);
6989	if (rc) {
6990		e1000e_pm_thaw(dev);
6991	} else {
6992		/* Introduce S0ix implementation */
6993		if (adapter->flags2 & FLAG2_ENABLE_S0IX_FLOWS)
6994			e1000e_s0ix_entry_flow(adapter);
6995	}
6996
6997	return rc;
6998}
6999
7000static __maybe_unused int e1000e_pm_resume(struct device *dev)
7001{
7002	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
7003	struct e1000_adapter *adapter = netdev_priv(netdev);
7004	struct pci_dev *pdev = to_pci_dev(dev);
7005	int rc;
7006
7007	/* Introduce S0ix implementation */
7008	if (adapter->flags2 & FLAG2_ENABLE_S0IX_FLOWS)
7009		e1000e_s0ix_exit_flow(adapter);
7010
7011	rc = __e1000_resume(pdev);
7012	if (rc)
7013		return rc;
7014
7015	return e1000e_pm_thaw(dev);
7016}
7017
7018static __maybe_unused int e1000e_pm_runtime_idle(struct device *dev)
7019{
7020	struct net_device *netdev = dev_get_drvdata(dev);
7021	struct e1000_adapter *adapter = netdev_priv(netdev);
7022	u16 eee_lp;
7023
7024	eee_lp = adapter->hw.dev_spec.ich8lan.eee_lp_ability;
7025
7026	if (!e1000e_has_link(adapter)) {
7027		adapter->hw.dev_spec.ich8lan.eee_lp_ability = eee_lp;
7028		pm_schedule_suspend(dev, 5 * MSEC_PER_SEC);
7029	}
7030
7031	return -EBUSY;
7032}
7033
7034static __maybe_unused int e1000e_pm_runtime_resume(struct device *dev)
7035{
7036	struct pci_dev *pdev = to_pci_dev(dev);
7037	struct net_device *netdev = pci_get_drvdata(pdev);
7038	struct e1000_adapter *adapter = netdev_priv(netdev);
7039	int rc;
7040
7041	pdev->pme_poll = true;
7042
7043	rc = __e1000_resume(pdev);
7044	if (rc)
7045		return rc;
7046
7047	if (netdev->flags & IFF_UP)
7048		e1000e_up(adapter);
7049
7050	return rc;
7051}
7052
7053static __maybe_unused int e1000e_pm_runtime_suspend(struct device *dev)
7054{
7055	struct pci_dev *pdev = to_pci_dev(dev);
7056	struct net_device *netdev = pci_get_drvdata(pdev);
7057	struct e1000_adapter *adapter = netdev_priv(netdev);
7058
7059	if (netdev->flags & IFF_UP) {
7060		int count = E1000_CHECK_RESET_COUNT;
7061
7062		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
7063			usleep_range(10000, 11000);
7064
7065		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
7066
7067		/* Down the device without resetting the hardware */
7068		e1000e_down(adapter, false);
7069	}
7070
7071	if (__e1000_shutdown(pdev, true)) {
7072		e1000e_pm_runtime_resume(dev);
7073		return -EBUSY;
7074	}
7075
7076	return 0;
7077}
7078
7079static void e1000_shutdown(struct pci_dev *pdev)
7080{
7081	e1000e_flush_lpic(pdev);
7082
7083	e1000e_pm_freeze(&pdev->dev);
7084
7085	__e1000_shutdown(pdev, false);
7086}
7087
7088#ifdef CONFIG_NET_POLL_CONTROLLER
7089
7090static irqreturn_t e1000_intr_msix(int __always_unused irq, void *data)
7091{
7092	struct net_device *netdev = data;
7093	struct e1000_adapter *adapter = netdev_priv(netdev);
7094
7095	if (adapter->msix_entries) {
7096		int vector, msix_irq;
7097
7098		vector = 0;
7099		msix_irq = adapter->msix_entries[vector].vector;
7100		if (disable_hardirq(msix_irq))
7101			e1000_intr_msix_rx(msix_irq, netdev);
7102		enable_irq(msix_irq);
7103
7104		vector++;
7105		msix_irq = adapter->msix_entries[vector].vector;
7106		if (disable_hardirq(msix_irq))
7107			e1000_intr_msix_tx(msix_irq, netdev);
7108		enable_irq(msix_irq);
7109
7110		vector++;
7111		msix_irq = adapter->msix_entries[vector].vector;
7112		if (disable_hardirq(msix_irq))
7113			e1000_msix_other(msix_irq, netdev);
7114		enable_irq(msix_irq);
7115	}
7116
7117	return IRQ_HANDLED;
7118}
7119
7120/**
7121 * e1000_netpoll
7122 * @netdev: network interface device structure
7123 *
7124 * Polling 'interrupt' - used by things like netconsole to send skbs
7125 * without having to re-enable interrupts. It's not called while
7126 * the interrupt routine is executing.
7127 */
7128static void e1000_netpoll(struct net_device *netdev)
7129{
7130	struct e1000_adapter *adapter = netdev_priv(netdev);
7131
7132	switch (adapter->int_mode) {
7133	case E1000E_INT_MODE_MSIX:
7134		e1000_intr_msix(adapter->pdev->irq, netdev);
7135		break;
7136	case E1000E_INT_MODE_MSI:
7137		if (disable_hardirq(adapter->pdev->irq))
7138			e1000_intr_msi(adapter->pdev->irq, netdev);
7139		enable_irq(adapter->pdev->irq);
7140		break;
7141	default:		/* E1000E_INT_MODE_LEGACY */
7142		if (disable_hardirq(adapter->pdev->irq))
7143			e1000_intr(adapter->pdev->irq, netdev);
7144		enable_irq(adapter->pdev->irq);
7145		break;
7146	}
7147}
7148#endif
7149
7150/**
7151 * e1000_io_error_detected - called when PCI error is detected
7152 * @pdev: Pointer to PCI device
7153 * @state: The current pci connection state
7154 *
7155 * This function is called after a PCI bus error affecting
7156 * this device has been detected.
7157 */
7158static pci_ers_result_t e1000_io_error_detected(struct pci_dev *pdev,
7159						pci_channel_state_t state)
7160{
7161	e1000e_pm_freeze(&pdev->dev);
7162
7163	if (state == pci_channel_io_perm_failure)
7164		return PCI_ERS_RESULT_DISCONNECT;
7165
7166	pci_disable_device(pdev);
7167
7168	/* Request a slot reset. */
7169	return PCI_ERS_RESULT_NEED_RESET;
7170}
7171
7172/**
7173 * e1000_io_slot_reset - called after the pci bus has been reset.
7174 * @pdev: Pointer to PCI device
7175 *
7176 * Restart the card from scratch, as if from a cold-boot. Implementation
7177 * resembles the first-half of the e1000e_pm_resume routine.
7178 */
7179static pci_ers_result_t e1000_io_slot_reset(struct pci_dev *pdev)
7180{
7181	struct net_device *netdev = pci_get_drvdata(pdev);
7182	struct e1000_adapter *adapter = netdev_priv(netdev);
7183	struct e1000_hw *hw = &adapter->hw;
7184	u16 aspm_disable_flag = 0;
7185	int err;
7186	pci_ers_result_t result;
7187
7188	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
7189		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7190	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
7191		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7192	if (aspm_disable_flag)
7193		e1000e_disable_aspm_locked(pdev, aspm_disable_flag);
7194
7195	err = pci_enable_device_mem(pdev);
7196	if (err) {
7197		dev_err(&pdev->dev,
7198			"Cannot re-enable PCI device after reset.\n");
7199		result = PCI_ERS_RESULT_DISCONNECT;
7200	} else {
7201		pdev->state_saved = true;
7202		pci_restore_state(pdev);
7203		pci_set_master(pdev);
7204
7205		pci_enable_wake(pdev, PCI_D3hot, 0);
7206		pci_enable_wake(pdev, PCI_D3cold, 0);
7207
7208		e1000e_reset(adapter);
7209		ew32(WUS, ~0);
7210		result = PCI_ERS_RESULT_RECOVERED;
7211	}
7212
7213	return result;
7214}
7215
7216/**
7217 * e1000_io_resume - called when traffic can start flowing again.
7218 * @pdev: Pointer to PCI device
7219 *
7220 * This callback is called when the error recovery driver tells us that
7221 * its OK to resume normal operation. Implementation resembles the
7222 * second-half of the e1000e_pm_resume routine.
7223 */
7224static void e1000_io_resume(struct pci_dev *pdev)
7225{
7226	struct net_device *netdev = pci_get_drvdata(pdev);
7227	struct e1000_adapter *adapter = netdev_priv(netdev);
7228
7229	e1000_init_manageability_pt(adapter);
7230
7231	e1000e_pm_thaw(&pdev->dev);
7232
7233	/* If the controller has AMT, do not set DRV_LOAD until the interface
7234	 * is up.  For all other cases, let the f/w know that the h/w is now
7235	 * under the control of the driver.
7236	 */
7237	if (!(adapter->flags & FLAG_HAS_AMT))
7238		e1000e_get_hw_control(adapter);
7239}
7240
7241static void e1000_print_device_info(struct e1000_adapter *adapter)
7242{
7243	struct e1000_hw *hw = &adapter->hw;
7244	struct net_device *netdev = adapter->netdev;
7245	u32 ret_val;
7246	u8 pba_str[E1000_PBANUM_LENGTH];
7247
7248	/* print bus type/speed/width info */
7249	e_info("(PCI Express:2.5GT/s:%s) %pM\n",
7250	       /* bus width */
7251	       ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
7252		"Width x1"),
7253	       /* MAC address */
7254	       netdev->dev_addr);
7255	e_info("Intel(R) PRO/%s Network Connection\n",
7256	       (hw->phy.type == e1000_phy_ife) ? "10/100" : "1000");
7257	ret_val = e1000_read_pba_string_generic(hw, pba_str,
7258						E1000_PBANUM_LENGTH);
7259	if (ret_val)
7260		strscpy((char *)pba_str, "Unknown", sizeof(pba_str));
7261	e_info("MAC: %d, PHY: %d, PBA No: %s\n",
7262	       hw->mac.type, hw->phy.type, pba_str);
7263}
7264
7265static void e1000_eeprom_checks(struct e1000_adapter *adapter)
7266{
7267	struct e1000_hw *hw = &adapter->hw;
7268	int ret_val;
7269	u16 buf = 0;
7270
7271	if (hw->mac.type != e1000_82573)
7272		return;
7273
7274	ret_val = e1000_read_nvm(hw, NVM_INIT_CONTROL2_REG, 1, &buf);
7275	le16_to_cpus(&buf);
7276	if (!ret_val && (!(buf & BIT(0)))) {
7277		/* Deep Smart Power Down (DSPD) */
7278		dev_warn(&adapter->pdev->dev,
7279			 "Warning: detected DSPD enabled in EEPROM\n");
7280	}
7281}
7282
7283static netdev_features_t e1000_fix_features(struct net_device *netdev,
7284					    netdev_features_t features)
7285{
7286	struct e1000_adapter *adapter = netdev_priv(netdev);
7287	struct e1000_hw *hw = &adapter->hw;
7288
7289	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
7290	if ((hw->mac.type >= e1000_pch2lan) && (netdev->mtu > ETH_DATA_LEN))
7291		features &= ~NETIF_F_RXFCS;
7292
7293	/* Since there is no support for separate Rx/Tx vlan accel
7294	 * enable/disable make sure Tx flag is always in same state as Rx.
7295	 */
7296	if (features & NETIF_F_HW_VLAN_CTAG_RX)
7297		features |= NETIF_F_HW_VLAN_CTAG_TX;
7298	else
7299		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
7300
7301	return features;
7302}
7303
7304static int e1000_set_features(struct net_device *netdev,
7305			      netdev_features_t features)
7306{
7307	struct e1000_adapter *adapter = netdev_priv(netdev);
7308	netdev_features_t changed = features ^ netdev->features;
7309
7310	if (changed & (NETIF_F_TSO | NETIF_F_TSO6))
7311		adapter->flags |= FLAG_TSO_FORCE;
7312
7313	if (!(changed & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX |
7314			 NETIF_F_RXCSUM | NETIF_F_RXHASH | NETIF_F_RXFCS |
7315			 NETIF_F_RXALL)))
7316		return 0;
7317
7318	if (changed & NETIF_F_RXFCS) {
7319		if (features & NETIF_F_RXFCS) {
7320			adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7321		} else {
7322			/* We need to take it back to defaults, which might mean
7323			 * stripping is still disabled at the adapter level.
7324			 */
7325			if (adapter->flags2 & FLAG2_DFLT_CRC_STRIPPING)
7326				adapter->flags2 |= FLAG2_CRC_STRIPPING;
7327			else
7328				adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7329		}
7330	}
7331
7332	netdev->features = features;
7333
7334	if (netif_running(netdev))
7335		e1000e_reinit_locked(adapter);
7336	else
7337		e1000e_reset(adapter);
7338
7339	return 1;
7340}
7341
7342static const struct net_device_ops e1000e_netdev_ops = {
7343	.ndo_open		= e1000e_open,
7344	.ndo_stop		= e1000e_close,
7345	.ndo_start_xmit		= e1000_xmit_frame,
7346	.ndo_get_stats64	= e1000e_get_stats64,
7347	.ndo_set_rx_mode	= e1000e_set_rx_mode,
7348	.ndo_set_mac_address	= e1000_set_mac,
7349	.ndo_change_mtu		= e1000_change_mtu,
7350	.ndo_eth_ioctl		= e1000_ioctl,
7351	.ndo_tx_timeout		= e1000_tx_timeout,
7352	.ndo_validate_addr	= eth_validate_addr,
7353
7354	.ndo_vlan_rx_add_vid	= e1000_vlan_rx_add_vid,
7355	.ndo_vlan_rx_kill_vid	= e1000_vlan_rx_kill_vid,
7356#ifdef CONFIG_NET_POLL_CONTROLLER
7357	.ndo_poll_controller	= e1000_netpoll,
7358#endif
7359	.ndo_set_features = e1000_set_features,
7360	.ndo_fix_features = e1000_fix_features,
7361	.ndo_features_check	= passthru_features_check,
7362};
7363
7364/**
7365 * e1000_probe - Device Initialization Routine
7366 * @pdev: PCI device information struct
7367 * @ent: entry in e1000_pci_tbl
7368 *
7369 * Returns 0 on success, negative on failure
7370 *
7371 * e1000_probe initializes an adapter identified by a pci_dev structure.
7372 * The OS initialization, configuring of the adapter private structure,
7373 * and a hardware reset occur.
7374 **/
7375static int e1000_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7376{
7377	struct net_device *netdev;
7378	struct e1000_adapter *adapter;
7379	struct e1000_hw *hw;
7380	const struct e1000_info *ei = e1000_info_tbl[ent->driver_data];
7381	resource_size_t mmio_start, mmio_len;
7382	resource_size_t flash_start, flash_len;
7383	static int cards_found;
7384	u16 aspm_disable_flag = 0;
7385	u16 eeprom_data = 0;
7386	u16 eeprom_apme_mask = E1000_EEPROM_APME;
7387	int bars, i, err;
7388	s32 ret_val = 0;
7389
7390	if (ei->flags2 & FLAG2_DISABLE_ASPM_L0S)
7391		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7392	if (ei->flags2 & FLAG2_DISABLE_ASPM_L1)
7393		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7394	if (aspm_disable_flag)
7395		e1000e_disable_aspm(pdev, aspm_disable_flag);
7396
7397	err = pci_enable_device_mem(pdev);
7398	if (err)
7399		return err;
7400
7401	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
7402	if (err) {
7403		dev_err(&pdev->dev,
7404			"No usable DMA configuration, aborting\n");
7405		goto err_dma;
7406	}
7407
7408	bars = pci_select_bars(pdev, IORESOURCE_MEM);
7409	err = pci_request_selected_regions_exclusive(pdev, bars,
7410						     e1000e_driver_name);
7411	if (err)
7412		goto err_pci_reg;
7413
7414	pci_set_master(pdev);
7415	/* PCI config space info */
7416	err = pci_save_state(pdev);
7417	if (err)
7418		goto err_alloc_etherdev;
7419
7420	err = -ENOMEM;
7421	netdev = alloc_etherdev(sizeof(struct e1000_adapter));
7422	if (!netdev)
7423		goto err_alloc_etherdev;
7424
7425	SET_NETDEV_DEV(netdev, &pdev->dev);
7426
7427	netdev->irq = pdev->irq;
7428
7429	pci_set_drvdata(pdev, netdev);
7430	adapter = netdev_priv(netdev);
7431	hw = &adapter->hw;
7432	adapter->netdev = netdev;
7433	adapter->pdev = pdev;
7434	adapter->ei = ei;
7435	adapter->pba = ei->pba;
7436	adapter->flags = ei->flags;
7437	adapter->flags2 = ei->flags2;
7438	adapter->hw.adapter = adapter;
7439	adapter->hw.mac.type = ei->mac;
7440	adapter->max_hw_frame_size = ei->max_hw_frame_size;
7441	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
7442
7443	mmio_start = pci_resource_start(pdev, 0);
7444	mmio_len = pci_resource_len(pdev, 0);
7445
7446	err = -EIO;
7447	adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
7448	if (!adapter->hw.hw_addr)
7449		goto err_ioremap;
7450
7451	if ((adapter->flags & FLAG_HAS_FLASH) &&
7452	    (pci_resource_flags(pdev, 1) & IORESOURCE_MEM) &&
7453	    (hw->mac.type < e1000_pch_spt)) {
7454		flash_start = pci_resource_start(pdev, 1);
7455		flash_len = pci_resource_len(pdev, 1);
7456		adapter->hw.flash_address = ioremap(flash_start, flash_len);
7457		if (!adapter->hw.flash_address)
7458			goto err_flashmap;
7459	}
7460
7461	/* Set default EEE advertisement */
7462	if (adapter->flags2 & FLAG2_HAS_EEE)
7463		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
7464
7465	/* construct the net_device struct */
7466	netdev->netdev_ops = &e1000e_netdev_ops;
7467	e1000e_set_ethtool_ops(netdev);
7468	netdev->watchdog_timeo = 5 * HZ;
7469	netif_napi_add(netdev, &adapter->napi, e1000e_poll);
7470	strscpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
7471
7472	netdev->mem_start = mmio_start;
7473	netdev->mem_end = mmio_start + mmio_len;
7474
7475	adapter->bd_number = cards_found++;
7476
7477	e1000e_check_options(adapter);
7478
7479	/* setup adapter struct */
7480	err = e1000_sw_init(adapter);
7481	if (err)
7482		goto err_sw_init;
7483
7484	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
7485	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
7486	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
7487
7488	err = ei->get_variants(adapter);
7489	if (err)
7490		goto err_hw_init;
7491
7492	if ((adapter->flags & FLAG_IS_ICH) &&
7493	    (adapter->flags & FLAG_READ_ONLY_NVM) &&
7494	    (hw->mac.type < e1000_pch_spt))
7495		e1000e_write_protect_nvm_ich8lan(&adapter->hw);
7496
7497	hw->mac.ops.get_bus_info(&adapter->hw);
7498
7499	adapter->hw.phy.autoneg_wait_to_complete = 0;
7500
7501	/* Copper options */
7502	if (adapter->hw.phy.media_type == e1000_media_type_copper) {
7503		adapter->hw.phy.mdix = AUTO_ALL_MODES;
7504		adapter->hw.phy.disable_polarity_correction = 0;
7505		adapter->hw.phy.ms_type = e1000_ms_hw_default;
7506	}
7507
7508	if (hw->phy.ops.check_reset_block && hw->phy.ops.check_reset_block(hw))
7509		dev_info(&pdev->dev,
7510			 "PHY reset is blocked due to SOL/IDER session.\n");
7511
7512	/* Set initial default active device features */
7513	netdev->features = (NETIF_F_SG |
7514			    NETIF_F_HW_VLAN_CTAG_RX |
7515			    NETIF_F_HW_VLAN_CTAG_TX |
7516			    NETIF_F_TSO |
7517			    NETIF_F_TSO6 |
7518			    NETIF_F_RXHASH |
7519			    NETIF_F_RXCSUM |
7520			    NETIF_F_HW_CSUM);
7521
7522	/* disable TSO for pcie and 10/100 speeds to avoid
7523	 * some hardware issues and for i219 to fix transfer
7524	 * speed being capped at 60%
7525	 */
7526	if (!(adapter->flags & FLAG_TSO_FORCE)) {
7527		switch (adapter->link_speed) {
7528		case SPEED_10:
7529		case SPEED_100:
7530			e_info("10/100 speed: disabling TSO\n");
7531			netdev->features &= ~NETIF_F_TSO;
7532			netdev->features &= ~NETIF_F_TSO6;
7533			break;
7534		case SPEED_1000:
7535			netdev->features |= NETIF_F_TSO;
7536			netdev->features |= NETIF_F_TSO6;
7537			break;
7538		default:
7539			/* oops */
7540			break;
7541		}
7542		if (hw->mac.type == e1000_pch_spt) {
7543			netdev->features &= ~NETIF_F_TSO;
7544			netdev->features &= ~NETIF_F_TSO6;
7545		}
7546	}
7547
7548	/* Set user-changeable features (subset of all device features) */
7549	netdev->hw_features = netdev->features;
7550	netdev->hw_features |= NETIF_F_RXFCS;
7551	netdev->priv_flags |= IFF_SUPP_NOFCS;
7552	netdev->hw_features |= NETIF_F_RXALL;
7553
7554	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER)
7555		netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
7556
7557	netdev->vlan_features |= (NETIF_F_SG |
7558				  NETIF_F_TSO |
7559				  NETIF_F_TSO6 |
7560				  NETIF_F_HW_CSUM);
7561
7562	netdev->priv_flags |= IFF_UNICAST_FLT;
7563
7564	netdev->features |= NETIF_F_HIGHDMA;
7565	netdev->vlan_features |= NETIF_F_HIGHDMA;
7566
7567	/* MTU range: 68 - max_hw_frame_size */
7568	netdev->min_mtu = ETH_MIN_MTU;
7569	netdev->max_mtu = adapter->max_hw_frame_size -
7570			  (VLAN_ETH_HLEN + ETH_FCS_LEN);
7571
7572	if (e1000e_enable_mng_pass_thru(&adapter->hw))
7573		adapter->flags |= FLAG_MNG_PT_ENABLED;
7574
7575	/* before reading the NVM, reset the controller to
7576	 * put the device in a known good starting state
7577	 */
7578	adapter->hw.mac.ops.reset_hw(&adapter->hw);
7579
7580	/* systems with ASPM and others may see the checksum fail on the first
7581	 * attempt. Let's give it a few tries
7582	 */
7583	for (i = 0;; i++) {
7584		if (e1000_validate_nvm_checksum(&adapter->hw) >= 0)
7585			break;
7586		if (i == 2) {
7587			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
7588			err = -EIO;
7589			goto err_eeprom;
7590		}
7591	}
7592
7593	e1000_eeprom_checks(adapter);
7594
7595	/* copy the MAC address */
7596	if (e1000e_read_mac_addr(&adapter->hw))
7597		dev_err(&pdev->dev,
7598			"NVM Read Error while reading MAC address\n");
7599
7600	eth_hw_addr_set(netdev, adapter->hw.mac.addr);
7601
7602	if (!is_valid_ether_addr(netdev->dev_addr)) {
7603		dev_err(&pdev->dev, "Invalid MAC Address: %pM\n",
7604			netdev->dev_addr);
7605		err = -EIO;
7606		goto err_eeprom;
7607	}
7608
7609	timer_setup(&adapter->watchdog_timer, e1000_watchdog, 0);
7610	timer_setup(&adapter->phy_info_timer, e1000_update_phy_info, 0);
7611
7612	INIT_WORK(&adapter->reset_task, e1000_reset_task);
7613	INIT_WORK(&adapter->watchdog_task, e1000_watchdog_task);
7614	INIT_WORK(&adapter->downshift_task, e1000e_downshift_workaround);
7615	INIT_WORK(&adapter->update_phy_task, e1000e_update_phy_task);
7616	INIT_WORK(&adapter->print_hang_task, e1000_print_hw_hang);
7617
7618	/* Initialize link parameters. User can change them with ethtool */
7619	adapter->hw.mac.autoneg = 1;
7620	adapter->fc_autoneg = true;
7621	adapter->hw.fc.requested_mode = e1000_fc_default;
7622	adapter->hw.fc.current_mode = e1000_fc_default;
7623	adapter->hw.phy.autoneg_advertised = 0x2f;
7624
7625	/* Initial Wake on LAN setting - If APM wake is enabled in
7626	 * the EEPROM, enable the ACPI Magic Packet filter
7627	 */
7628	if (adapter->flags & FLAG_APME_IN_WUC) {
7629		/* APME bit in EEPROM is mapped to WUC.APME */
7630		eeprom_data = er32(WUC);
7631		eeprom_apme_mask = E1000_WUC_APME;
7632		if ((hw->mac.type > e1000_ich10lan) &&
7633		    (eeprom_data & E1000_WUC_PHY_WAKE))
7634			adapter->flags2 |= FLAG2_HAS_PHY_WAKEUP;
7635	} else if (adapter->flags & FLAG_APME_IN_CTRL3) {
7636		if (adapter->flags & FLAG_APME_CHECK_PORT_B &&
7637		    (adapter->hw.bus.func == 1))
7638			ret_val = e1000_read_nvm(&adapter->hw,
7639					      NVM_INIT_CONTROL3_PORT_B,
7640					      1, &eeprom_data);
7641		else
7642			ret_val = e1000_read_nvm(&adapter->hw,
7643					      NVM_INIT_CONTROL3_PORT_A,
7644					      1, &eeprom_data);
7645	}
7646
7647	/* fetch WoL from EEPROM */
7648	if (ret_val)
7649		e_dbg("NVM read error getting WoL initial values: %d\n", ret_val);
7650	else if (eeprom_data & eeprom_apme_mask)
7651		adapter->eeprom_wol |= E1000_WUFC_MAG;
7652
7653	/* now that we have the eeprom settings, apply the special cases
7654	 * where the eeprom may be wrong or the board simply won't support
7655	 * wake on lan on a particular port
7656	 */
7657	if (!(adapter->flags & FLAG_HAS_WOL))
7658		adapter->eeprom_wol = 0;
7659
7660	/* initialize the wol settings based on the eeprom settings */
7661	adapter->wol = adapter->eeprom_wol;
7662
7663	/* make sure adapter isn't asleep if manageability is enabled */
7664	if (adapter->wol || (adapter->flags & FLAG_MNG_PT_ENABLED) ||
7665	    (hw->mac.ops.check_mng_mode(hw)))
7666		device_wakeup_enable(&pdev->dev);
7667
7668	/* save off EEPROM version number */
7669	ret_val = e1000_read_nvm(&adapter->hw, 5, 1, &adapter->eeprom_vers);
7670
7671	if (ret_val) {
7672		e_dbg("NVM read error getting EEPROM version: %d\n", ret_val);
7673		adapter->eeprom_vers = 0;
7674	}
7675
7676	/* init PTP hardware clock */
7677	e1000e_ptp_init(adapter);
7678
7679	/* reset the hardware with the new settings */
7680	e1000e_reset(adapter);
7681
7682	/* If the controller has AMT, do not set DRV_LOAD until the interface
7683	 * is up.  For all other cases, let the f/w know that the h/w is now
7684	 * under the control of the driver.
7685	 */
7686	if (!(adapter->flags & FLAG_HAS_AMT))
7687		e1000e_get_hw_control(adapter);
7688
7689	if (hw->mac.type >= e1000_pch_cnp)
7690		adapter->flags2 |= FLAG2_ENABLE_S0IX_FLOWS;
7691
7692	strscpy(netdev->name, "eth%d", sizeof(netdev->name));
7693	err = register_netdev(netdev);
7694	if (err)
7695		goto err_register;
7696
7697	/* carrier off reporting is important to ethtool even BEFORE open */
7698	netif_carrier_off(netdev);
7699
7700	e1000_print_device_info(adapter);
7701
7702	dev_pm_set_driver_flags(&pdev->dev, DPM_FLAG_SMART_PREPARE);
7703
7704	if (pci_dev_run_wake(pdev))
7705		pm_runtime_put_noidle(&pdev->dev);
7706
7707	return 0;
7708
7709err_register:
7710	if (!(adapter->flags & FLAG_HAS_AMT))
7711		e1000e_release_hw_control(adapter);
7712err_eeprom:
7713	if (hw->phy.ops.check_reset_block && !hw->phy.ops.check_reset_block(hw))
7714		e1000_phy_hw_reset(&adapter->hw);
7715err_hw_init:
7716	kfree(adapter->tx_ring);
7717	kfree(adapter->rx_ring);
7718err_sw_init:
7719	if ((adapter->hw.flash_address) && (hw->mac.type < e1000_pch_spt))
7720		iounmap(adapter->hw.flash_address);
7721	e1000e_reset_interrupt_capability(adapter);
7722err_flashmap:
7723	iounmap(adapter->hw.hw_addr);
7724err_ioremap:
7725	free_netdev(netdev);
7726err_alloc_etherdev:
7727	pci_release_mem_regions(pdev);
7728err_pci_reg:
7729err_dma:
7730	pci_disable_device(pdev);
7731	return err;
7732}
7733
7734/**
7735 * e1000_remove - Device Removal Routine
7736 * @pdev: PCI device information struct
7737 *
7738 * e1000_remove is called by the PCI subsystem to alert the driver
7739 * that it should release a PCI device.  This could be caused by a
7740 * Hot-Plug event, or because the driver is going to be removed from
7741 * memory.
7742 **/
7743static void e1000_remove(struct pci_dev *pdev)
7744{
7745	struct net_device *netdev = pci_get_drvdata(pdev);
7746	struct e1000_adapter *adapter = netdev_priv(netdev);
7747
7748	e1000e_ptp_remove(adapter);
7749
7750	/* The timers may be rescheduled, so explicitly disable them
7751	 * from being rescheduled.
7752	 */
7753	set_bit(__E1000_DOWN, &adapter->state);
7754	del_timer_sync(&adapter->watchdog_timer);
7755	del_timer_sync(&adapter->phy_info_timer);
7756
7757	cancel_work_sync(&adapter->reset_task);
7758	cancel_work_sync(&adapter->watchdog_task);
7759	cancel_work_sync(&adapter->downshift_task);
7760	cancel_work_sync(&adapter->update_phy_task);
7761	cancel_work_sync(&adapter->print_hang_task);
7762
7763	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
7764		cancel_work_sync(&adapter->tx_hwtstamp_work);
7765		if (adapter->tx_hwtstamp_skb) {
7766			dev_consume_skb_any(adapter->tx_hwtstamp_skb);
7767			adapter->tx_hwtstamp_skb = NULL;
7768		}
7769	}
7770
7771	unregister_netdev(netdev);
7772
7773	if (pci_dev_run_wake(pdev))
7774		pm_runtime_get_noresume(&pdev->dev);
7775
7776	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
7777	 * would have already happened in close and is redundant.
7778	 */
7779	e1000e_release_hw_control(adapter);
7780
7781	e1000e_reset_interrupt_capability(adapter);
7782	kfree(adapter->tx_ring);
7783	kfree(adapter->rx_ring);
7784
7785	iounmap(adapter->hw.hw_addr);
7786	if ((adapter->hw.flash_address) &&
7787	    (adapter->hw.mac.type < e1000_pch_spt))
7788		iounmap(adapter->hw.flash_address);
7789	pci_release_mem_regions(pdev);
7790
7791	free_netdev(netdev);
7792
7793	pci_disable_device(pdev);
7794}
7795
7796/* PCI Error Recovery (ERS) */
7797static const struct pci_error_handlers e1000_err_handler = {
7798	.error_detected = e1000_io_error_detected,
7799	.slot_reset = e1000_io_slot_reset,
7800	.resume = e1000_io_resume,
7801};
7802
7803static const struct pci_device_id e1000_pci_tbl[] = {
7804	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_COPPER), board_82571 },
7805	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_FIBER), board_82571 },
7806	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER), board_82571 },
7807	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER_LP),
7808	  board_82571 },
7809	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_FIBER), board_82571 },
7810	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES), board_82571 },
7811	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_DUAL), board_82571 },
7812	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_QUAD), board_82571 },
7813	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571PT_QUAD_COPPER), board_82571 },
7814
7815	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI), board_82572 },
7816	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_COPPER), board_82572 },
7817	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_FIBER), board_82572 },
7818	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_SERDES), board_82572 },
7819
7820	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E), board_82573 },
7821	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E_IAMT), board_82573 },
7822	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573L), board_82573 },
7823
7824	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574L), board_82574 },
7825	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574LA), board_82574 },
7826	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82583V), board_82583 },
7827
7828	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_DPT),
7829	  board_80003es2lan },
7830	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_SPT),
7831	  board_80003es2lan },
7832	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_DPT),
7833	  board_80003es2lan },
7834	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_SPT),
7835	  board_80003es2lan },
7836
7837	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE), board_ich8lan },
7838	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_G), board_ich8lan },
7839	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_GT), board_ich8lan },
7840	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_AMT), board_ich8lan },
7841	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_C), board_ich8lan },
7842	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M), board_ich8lan },
7843	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M_AMT), board_ich8lan },
7844	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_82567V_3), board_ich8lan },
7845
7846	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE), board_ich9lan },
7847	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_G), board_ich9lan },
7848	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_GT), board_ich9lan },
7849	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_AMT), board_ich9lan },
7850	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_C), board_ich9lan },
7851	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_BM), board_ich9lan },
7852	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M), board_ich9lan },
7853	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_AMT), board_ich9lan },
7854	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_V), board_ich9lan },
7855
7856	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LM), board_ich9lan },
7857	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LF), board_ich9lan },
7858	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_V), board_ich9lan },
7859
7860	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LM), board_ich10lan },
7861	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LF), board_ich10lan },
7862	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_V), board_ich10lan },
7863
7864	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LM), board_pchlan },
7865	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LC), board_pchlan },
7866	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DM), board_pchlan },
7867	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DC), board_pchlan },
7868
7869	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_LM), board_pch2lan },
7870	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_V), board_pch2lan },
7871
7872	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_LM), board_pch_lpt },
7873	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_V), board_pch_lpt },
7874	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_LM), board_pch_lpt },
7875	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_V), board_pch_lpt },
7876	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM2), board_pch_lpt },
7877	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V2), board_pch_lpt },
7878	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM3), board_pch_lpt },
7879	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V3), board_pch_lpt },
7880	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM), board_pch_spt },
7881	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V), board_pch_spt },
7882	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM2), board_pch_spt },
7883	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V2), board_pch_spt },
7884	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LBG_I219_LM3), board_pch_spt },
7885	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM4), board_pch_spt },
7886	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V4), board_pch_spt },
7887	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM5), board_pch_spt },
7888	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V5), board_pch_spt },
7889	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM6), board_pch_cnp },
7890	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V6), board_pch_cnp },
7891	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM7), board_pch_cnp },
7892	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V7), board_pch_cnp },
7893	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM8), board_pch_cnp },
7894	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V8), board_pch_cnp },
7895	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM9), board_pch_cnp },
7896	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V9), board_pch_cnp },
7897	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM10), board_pch_cnp },
7898	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V10), board_pch_cnp },
7899	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM11), board_pch_cnp },
7900	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V11), board_pch_cnp },
7901	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM12), board_pch_spt },
7902	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V12), board_pch_spt },
7903	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM13), board_pch_tgp },
7904	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V13), board_pch_tgp },
7905	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM14), board_pch_tgp },
7906	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V14), board_pch_tgp },
7907	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM15), board_pch_tgp },
7908	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V15), board_pch_tgp },
7909	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_LM23), board_pch_adp },
7910	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_V23), board_pch_adp },
7911	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM16), board_pch_adp },
7912	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V16), board_pch_adp },
7913	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM17), board_pch_adp },
7914	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V17), board_pch_adp },
7915	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_LM22), board_pch_adp },
7916	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_V22), board_pch_adp },
7917	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_LM18), board_pch_mtp },
7918	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_V18), board_pch_mtp },
7919	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_LM19), board_pch_mtp },
7920	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_V19), board_pch_mtp },
7921	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_LM20), board_pch_mtp },
7922	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_V20), board_pch_mtp },
7923	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_LM21), board_pch_mtp },
7924	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_V21), board_pch_mtp },
7925	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ARL_I219_LM24), board_pch_mtp },
7926	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ARL_I219_V24), board_pch_mtp },
7927	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_LM25), board_pch_mtp },
7928	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_V25), board_pch_mtp },
7929	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_LM26), board_pch_mtp },
7930	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_V26), board_pch_mtp },
7931	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_LM27), board_pch_mtp },
7932	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_V27), board_pch_mtp },
7933	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_NVL_I219_LM29), board_pch_mtp },
7934	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_NVL_I219_V29), board_pch_mtp },
7935
7936	{ 0, 0, 0, 0, 0, 0, 0 }	/* terminate list */
7937};
7938MODULE_DEVICE_TABLE(pci, e1000_pci_tbl);
7939
7940static const struct dev_pm_ops e1000_pm_ops = {
7941#ifdef CONFIG_PM_SLEEP
7942	.prepare	= e1000e_pm_prepare,
7943	.suspend	= e1000e_pm_suspend,
7944	.resume		= e1000e_pm_resume,
7945	.freeze		= e1000e_pm_freeze,
7946	.thaw		= e1000e_pm_thaw,
7947	.poweroff	= e1000e_pm_suspend,
7948	.restore	= e1000e_pm_resume,
7949#endif
7950	SET_RUNTIME_PM_OPS(e1000e_pm_runtime_suspend, e1000e_pm_runtime_resume,
7951			   e1000e_pm_runtime_idle)
7952};
7953
7954/* PCI Device API Driver */
7955static struct pci_driver e1000_driver = {
7956	.name     = e1000e_driver_name,
7957	.id_table = e1000_pci_tbl,
7958	.probe    = e1000_probe,
7959	.remove   = e1000_remove,
7960	.driver   = {
7961		.pm = &e1000_pm_ops,
7962	},
7963	.shutdown = e1000_shutdown,
7964	.err_handler = &e1000_err_handler
7965};
7966
7967/**
7968 * e1000_init_module - Driver Registration Routine
7969 *
7970 * e1000_init_module is the first routine called when the driver is
7971 * loaded. All it does is register with the PCI subsystem.
7972 **/
7973static int __init e1000_init_module(void)
7974{
7975	pr_info("Intel(R) PRO/1000 Network Driver\n");
7976	pr_info("Copyright(c) 1999 - 2015 Intel Corporation.\n");
7977
7978	return pci_register_driver(&e1000_driver);
7979}
7980module_init(e1000_init_module);
7981
7982/**
7983 * e1000_exit_module - Driver Exit Cleanup Routine
7984 *
7985 * e1000_exit_module is called just before the driver is removed
7986 * from memory.
7987 **/
7988static void __exit e1000_exit_module(void)
7989{
7990	pci_unregister_driver(&e1000_driver);
7991}
7992module_exit(e1000_exit_module);
7993
7994MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
7995MODULE_DESCRIPTION("Intel(R) PRO/1000 Network Driver");
7996MODULE_LICENSE("GPL v2");
7997
7998/* netdev.c */
7999