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/aer.h>
27 #include <linux/prefetch.h>
28 
29 #include "e1000.h"
30 
31 #define DRV_EXTRAVERSION "-k"
32 
33 #define DRV_VERSION "3.2.6" DRV_EXTRAVERSION
34 char e1000e_driver_name[] = "e1000e";
35 const char e1000e_driver_version[] = DRV_VERSION;
36 
37 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
38 static int debug = -1;
39 module_param(debug, int, 0);
40 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
41 
42 static const struct e1000_info *e1000_info_tbl[] = {
43 	[board_82571]		= &e1000_82571_info,
44 	[board_82572]		= &e1000_82572_info,
45 	[board_82573]		= &e1000_82573_info,
46 	[board_82574]		= &e1000_82574_info,
47 	[board_82583]		= &e1000_82583_info,
48 	[board_80003es2lan]	= &e1000_es2_info,
49 	[board_ich8lan]		= &e1000_ich8_info,
50 	[board_ich9lan]		= &e1000_ich9_info,
51 	[board_ich10lan]	= &e1000_ich10_info,
52 	[board_pchlan]		= &e1000_pch_info,
53 	[board_pch2lan]		= &e1000_pch2_info,
54 	[board_pch_lpt]		= &e1000_pch_lpt_info,
55 	[board_pch_spt]		= &e1000_pch_spt_info,
56 	[board_pch_cnp]		= &e1000_pch_cnp_info,
57 };
58 
59 struct e1000_reg_info {
60 	u32 ofs;
61 	char *name;
62 };
63 
64 static const struct e1000_reg_info e1000_reg_info_tbl[] = {
65 	/* General Registers */
66 	{E1000_CTRL, "CTRL"},
67 	{E1000_STATUS, "STATUS"},
68 	{E1000_CTRL_EXT, "CTRL_EXT"},
69 
70 	/* Interrupt Registers */
71 	{E1000_ICR, "ICR"},
72 
73 	/* Rx Registers */
74 	{E1000_RCTL, "RCTL"},
75 	{E1000_RDLEN(0), "RDLEN"},
76 	{E1000_RDH(0), "RDH"},
77 	{E1000_RDT(0), "RDT"},
78 	{E1000_RDTR, "RDTR"},
79 	{E1000_RXDCTL(0), "RXDCTL"},
80 	{E1000_ERT, "ERT"},
81 	{E1000_RDBAL(0), "RDBAL"},
82 	{E1000_RDBAH(0), "RDBAH"},
83 	{E1000_RDFH, "RDFH"},
84 	{E1000_RDFT, "RDFT"},
85 	{E1000_RDFHS, "RDFHS"},
86 	{E1000_RDFTS, "RDFTS"},
87 	{E1000_RDFPC, "RDFPC"},
88 
89 	/* Tx Registers */
90 	{E1000_TCTL, "TCTL"},
91 	{E1000_TDBAL(0), "TDBAL"},
92 	{E1000_TDBAH(0), "TDBAH"},
93 	{E1000_TDLEN(0), "TDLEN"},
94 	{E1000_TDH(0), "TDH"},
95 	{E1000_TDT(0), "TDT"},
96 	{E1000_TIDV, "TIDV"},
97 	{E1000_TXDCTL(0), "TXDCTL"},
98 	{E1000_TADV, "TADV"},
99 	{E1000_TARC(0), "TARC"},
100 	{E1000_TDFH, "TDFH"},
101 	{E1000_TDFT, "TDFT"},
102 	{E1000_TDFHS, "TDFHS"},
103 	{E1000_TDFTS, "TDFTS"},
104 	{E1000_TDFPC, "TDFPC"},
105 
106 	/* List Terminator */
107 	{0, NULL}
108 };
109 
110 /**
111  * __ew32_prepare - prepare to write to MAC CSR register on certain parts
112  * @hw: pointer to the HW structure
113  *
114  * When updating the MAC CSR registers, the Manageability Engine (ME) could
115  * be accessing the registers at the same time.  Normally, this is handled in
116  * h/w by an arbiter but on some parts there is a bug that acknowledges Host
117  * accesses later than it should which could result in the register to have
118  * an incorrect value.  Workaround this by checking the FWSM register which
119  * has bit 24 set while ME is accessing MAC CSR registers, wait if it is set
120  * and try again a number of times.
121  **/
122 s32 __ew32_prepare(struct e1000_hw *hw)
123 {
124 	s32 i = E1000_ICH_FWSM_PCIM2PCI_COUNT;
125 
126 	while ((er32(FWSM) & E1000_ICH_FWSM_PCIM2PCI) && --i)
127 		udelay(50);
128 
129 	return i;
130 }
131 
132 void __ew32(struct e1000_hw *hw, unsigned long reg, u32 val)
133 {
134 	if (hw->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
135 		__ew32_prepare(hw);
136 
137 	writel(val, hw->hw_addr + reg);
138 }
139 
140 /**
141  * e1000_regdump - register printout routine
142  * @hw: pointer to the HW structure
143  * @reginfo: pointer to the register info table
144  **/
145 static void e1000_regdump(struct e1000_hw *hw, struct e1000_reg_info *reginfo)
146 {
147 	int n = 0;
148 	char rname[16];
149 	u32 regs[8];
150 
151 	switch (reginfo->ofs) {
152 	case E1000_RXDCTL(0):
153 		for (n = 0; n < 2; n++)
154 			regs[n] = __er32(hw, E1000_RXDCTL(n));
155 		break;
156 	case E1000_TXDCTL(0):
157 		for (n = 0; n < 2; n++)
158 			regs[n] = __er32(hw, E1000_TXDCTL(n));
159 		break;
160 	case E1000_TARC(0):
161 		for (n = 0; n < 2; n++)
162 			regs[n] = __er32(hw, E1000_TARC(n));
163 		break;
164 	default:
165 		pr_info("%-15s %08x\n",
166 			reginfo->name, __er32(hw, reginfo->ofs));
167 		return;
168 	}
169 
170 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-1]");
171 	pr_info("%-15s %08x %08x\n", rname, regs[0], regs[1]);
172 }
173 
174 static void e1000e_dump_ps_pages(struct e1000_adapter *adapter,
175 				 struct e1000_buffer *bi)
176 {
177 	int i;
178 	struct e1000_ps_page *ps_page;
179 
180 	for (i = 0; i < adapter->rx_ps_pages; i++) {
181 		ps_page = &bi->ps_pages[i];
182 
183 		if (ps_page->page) {
184 			pr_info("packet dump for ps_page %d:\n", i);
185 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
186 				       16, 1, page_address(ps_page->page),
187 				       PAGE_SIZE, true);
188 		}
189 	}
190 }
191 
192 /**
193  * e1000e_dump - Print registers, Tx-ring and Rx-ring
194  * @adapter: board private structure
195  **/
196 static void e1000e_dump(struct e1000_adapter *adapter)
197 {
198 	struct net_device *netdev = adapter->netdev;
199 	struct e1000_hw *hw = &adapter->hw;
200 	struct e1000_reg_info *reginfo;
201 	struct e1000_ring *tx_ring = adapter->tx_ring;
202 	struct e1000_tx_desc *tx_desc;
203 	struct my_u0 {
204 		__le64 a;
205 		__le64 b;
206 	} *u0;
207 	struct e1000_buffer *buffer_info;
208 	struct e1000_ring *rx_ring = adapter->rx_ring;
209 	union e1000_rx_desc_packet_split *rx_desc_ps;
210 	union e1000_rx_desc_extended *rx_desc;
211 	struct my_u1 {
212 		__le64 a;
213 		__le64 b;
214 		__le64 c;
215 		__le64 d;
216 	} *u1;
217 	u32 staterr;
218 	int i = 0;
219 
220 	if (!netif_msg_hw(adapter))
221 		return;
222 
223 	/* Print netdevice Info */
224 	if (netdev) {
225 		dev_info(&adapter->pdev->dev, "Net device Info\n");
226 		pr_info("Device Name     state            trans_start\n");
227 		pr_info("%-15s %016lX %016lX\n", netdev->name,
228 			netdev->state, dev_trans_start(netdev));
229 	}
230 
231 	/* Print Registers */
232 	dev_info(&adapter->pdev->dev, "Register Dump\n");
233 	pr_info(" Register Name   Value\n");
234 	for (reginfo = (struct e1000_reg_info *)e1000_reg_info_tbl;
235 	     reginfo->name; reginfo++) {
236 		e1000_regdump(hw, reginfo);
237 	}
238 
239 	/* Print Tx Ring Summary */
240 	if (!netdev || !netif_running(netdev))
241 		return;
242 
243 	dev_info(&adapter->pdev->dev, "Tx Ring Summary\n");
244 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
245 	buffer_info = &tx_ring->buffer_info[tx_ring->next_to_clean];
246 	pr_info(" %5d %5X %5X %016llX %04X %3X %016llX\n",
247 		0, tx_ring->next_to_use, tx_ring->next_to_clean,
248 		(unsigned long long)buffer_info->dma,
249 		buffer_info->length,
250 		buffer_info->next_to_watch,
251 		(unsigned long long)buffer_info->time_stamp);
252 
253 	/* Print Tx Ring */
254 	if (!netif_msg_tx_done(adapter))
255 		goto rx_ring_summary;
256 
257 	dev_info(&adapter->pdev->dev, "Tx Ring Dump\n");
258 
259 	/* Transmit Descriptor Formats - DEXT[29] is 0 (Legacy) or 1 (Extended)
260 	 *
261 	 * Legacy Transmit Descriptor
262 	 *   +--------------------------------------------------------------+
263 	 * 0 |         Buffer Address [63:0] (Reserved on Write Back)       |
264 	 *   +--------------------------------------------------------------+
265 	 * 8 | Special  |    CSS     | Status |  CMD    |  CSO   |  Length  |
266 	 *   +--------------------------------------------------------------+
267 	 *   63       48 47        36 35    32 31     24 23    16 15        0
268 	 *
269 	 * Extended Context Descriptor (DTYP=0x0) for TSO or checksum offload
270 	 *   63      48 47    40 39       32 31             16 15    8 7      0
271 	 *   +----------------------------------------------------------------+
272 	 * 0 |  TUCSE  | TUCS0  |   TUCSS   |     IPCSE       | IPCS0 | IPCSS |
273 	 *   +----------------------------------------------------------------+
274 	 * 8 |   MSS   | HDRLEN | RSV | STA | TUCMD | DTYP |      PAYLEN      |
275 	 *   +----------------------------------------------------------------+
276 	 *   63      48 47    40 39 36 35 32 31   24 23  20 19                0
277 	 *
278 	 * Extended Data Descriptor (DTYP=0x1)
279 	 *   +----------------------------------------------------------------+
280 	 * 0 |                     Buffer Address [63:0]                      |
281 	 *   +----------------------------------------------------------------+
282 	 * 8 | VLAN tag |  POPTS  | Rsvd | Status | Command | DTYP |  DTALEN  |
283 	 *   +----------------------------------------------------------------+
284 	 *   63       48 47     40 39  36 35    32 31     24 23  20 19        0
285 	 */
286 	pr_info("Tl[desc]     [address 63:0  ] [SpeCssSCmCsLen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Legacy format\n");
287 	pr_info("Tc[desc]     [Ce CoCsIpceCoS] [MssHlRSCm0Plen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Context format\n");
288 	pr_info("Td[desc]     [address 63:0  ] [VlaPoRSCm1Dlen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Data format\n");
289 	for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
290 		const char *next_desc;
291 		tx_desc = E1000_TX_DESC(*tx_ring, i);
292 		buffer_info = &tx_ring->buffer_info[i];
293 		u0 = (struct my_u0 *)tx_desc;
294 		if (i == tx_ring->next_to_use && i == tx_ring->next_to_clean)
295 			next_desc = " NTC/U";
296 		else if (i == tx_ring->next_to_use)
297 			next_desc = " NTU";
298 		else if (i == tx_ring->next_to_clean)
299 			next_desc = " NTC";
300 		else
301 			next_desc = "";
302 		pr_info("T%c[0x%03X]    %016llX %016llX %016llX %04X  %3X %016llX %p%s\n",
303 			(!(le64_to_cpu(u0->b) & BIT(29)) ? 'l' :
304 			 ((le64_to_cpu(u0->b) & BIT(20)) ? 'd' : 'c')),
305 			i,
306 			(unsigned long long)le64_to_cpu(u0->a),
307 			(unsigned long long)le64_to_cpu(u0->b),
308 			(unsigned long long)buffer_info->dma,
309 			buffer_info->length, buffer_info->next_to_watch,
310 			(unsigned long long)buffer_info->time_stamp,
311 			buffer_info->skb, next_desc);
312 
313 		if (netif_msg_pktdata(adapter) && buffer_info->skb)
314 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
315 				       16, 1, buffer_info->skb->data,
316 				       buffer_info->skb->len, true);
317 	}
318 
319 	/* Print Rx Ring Summary */
320 rx_ring_summary:
321 	dev_info(&adapter->pdev->dev, "Rx Ring Summary\n");
322 	pr_info("Queue [NTU] [NTC]\n");
323 	pr_info(" %5d %5X %5X\n",
324 		0, rx_ring->next_to_use, rx_ring->next_to_clean);
325 
326 	/* Print Rx Ring */
327 	if (!netif_msg_rx_status(adapter))
328 		return;
329 
330 	dev_info(&adapter->pdev->dev, "Rx Ring Dump\n");
331 	switch (adapter->rx_ps_pages) {
332 	case 1:
333 	case 2:
334 	case 3:
335 		/* [Extended] Packet Split Receive Descriptor Format
336 		 *
337 		 *    +-----------------------------------------------------+
338 		 *  0 |                Buffer Address 0 [63:0]              |
339 		 *    +-----------------------------------------------------+
340 		 *  8 |                Buffer Address 1 [63:0]              |
341 		 *    +-----------------------------------------------------+
342 		 * 16 |                Buffer Address 2 [63:0]              |
343 		 *    +-----------------------------------------------------+
344 		 * 24 |                Buffer Address 3 [63:0]              |
345 		 *    +-----------------------------------------------------+
346 		 */
347 		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");
348 		/* [Extended] Receive Descriptor (Write-Back) Format
349 		 *
350 		 *   63       48 47    32 31     13 12    8 7    4 3        0
351 		 *   +------------------------------------------------------+
352 		 * 0 | Packet   | IP     |  Rsvd   | MRQ   | Rsvd | MRQ RSS |
353 		 *   | Checksum | Ident  |         | Queue |      |  Type   |
354 		 *   +------------------------------------------------------+
355 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
356 		 *   +------------------------------------------------------+
357 		 *   63       48 47    32 31            20 19               0
358 		 */
359 		pr_info("RWB[desc]      [ck ipid mrqhsh] [vl   l0 ee  es] [ l3  l2  l1 hs] [reserved      ] ---------------- [bi->skb] <-- Ext Rx Write-Back format\n");
360 		for (i = 0; i < rx_ring->count; i++) {
361 			const char *next_desc;
362 			buffer_info = &rx_ring->buffer_info[i];
363 			rx_desc_ps = E1000_RX_DESC_PS(*rx_ring, i);
364 			u1 = (struct my_u1 *)rx_desc_ps;
365 			staterr =
366 			    le32_to_cpu(rx_desc_ps->wb.middle.status_error);
367 
368 			if (i == rx_ring->next_to_use)
369 				next_desc = " NTU";
370 			else if (i == rx_ring->next_to_clean)
371 				next_desc = " NTC";
372 			else
373 				next_desc = "";
374 
375 			if (staterr & E1000_RXD_STAT_DD) {
376 				/* Descriptor Done */
377 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX ---------------- %p%s\n",
378 					"RWB", i,
379 					(unsigned long long)le64_to_cpu(u1->a),
380 					(unsigned long long)le64_to_cpu(u1->b),
381 					(unsigned long long)le64_to_cpu(u1->c),
382 					(unsigned long long)le64_to_cpu(u1->d),
383 					buffer_info->skb, next_desc);
384 			} else {
385 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX %016llX %p%s\n",
386 					"R  ", i,
387 					(unsigned long long)le64_to_cpu(u1->a),
388 					(unsigned long long)le64_to_cpu(u1->b),
389 					(unsigned long long)le64_to_cpu(u1->c),
390 					(unsigned long long)le64_to_cpu(u1->d),
391 					(unsigned long long)buffer_info->dma,
392 					buffer_info->skb, next_desc);
393 
394 				if (netif_msg_pktdata(adapter))
395 					e1000e_dump_ps_pages(adapter,
396 							     buffer_info);
397 			}
398 		}
399 		break;
400 	default:
401 	case 0:
402 		/* Extended Receive Descriptor (Read) Format
403 		 *
404 		 *   +-----------------------------------------------------+
405 		 * 0 |                Buffer Address [63:0]                |
406 		 *   +-----------------------------------------------------+
407 		 * 8 |                      Reserved                       |
408 		 *   +-----------------------------------------------------+
409 		 */
410 		pr_info("R  [desc]      [buf addr 63:0 ] [reserved 63:0 ] [bi->dma       ] [bi->skb] <-- Ext (Read) format\n");
411 		/* Extended Receive Descriptor (Write-Back) Format
412 		 *
413 		 *   63       48 47    32 31    24 23            4 3        0
414 		 *   +------------------------------------------------------+
415 		 *   |     RSS Hash      |        |               |         |
416 		 * 0 +-------------------+  Rsvd  |   Reserved    | MRQ RSS |
417 		 *   | Packet   | IP     |        |               |  Type   |
418 		 *   | Checksum | Ident  |        |               |         |
419 		 *   +------------------------------------------------------+
420 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
421 		 *   +------------------------------------------------------+
422 		 *   63       48 47    32 31            20 19               0
423 		 */
424 		pr_info("RWB[desc]      [cs ipid    mrq] [vt   ln xe  xs] [bi->skb] <-- Ext (Write-Back) format\n");
425 
426 		for (i = 0; i < rx_ring->count; i++) {
427 			const char *next_desc;
428 
429 			buffer_info = &rx_ring->buffer_info[i];
430 			rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
431 			u1 = (struct my_u1 *)rx_desc;
432 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
433 
434 			if (i == rx_ring->next_to_use)
435 				next_desc = " NTU";
436 			else if (i == rx_ring->next_to_clean)
437 				next_desc = " NTC";
438 			else
439 				next_desc = "";
440 
441 			if (staterr & E1000_RXD_STAT_DD) {
442 				/* Descriptor Done */
443 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %p%s\n",
444 					"RWB", i,
445 					(unsigned long long)le64_to_cpu(u1->a),
446 					(unsigned long long)le64_to_cpu(u1->b),
447 					buffer_info->skb, next_desc);
448 			} else {
449 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %p%s\n",
450 					"R  ", i,
451 					(unsigned long long)le64_to_cpu(u1->a),
452 					(unsigned long long)le64_to_cpu(u1->b),
453 					(unsigned long long)buffer_info->dma,
454 					buffer_info->skb, next_desc);
455 
456 				if (netif_msg_pktdata(adapter) &&
457 				    buffer_info->skb)
458 					print_hex_dump(KERN_INFO, "",
459 						       DUMP_PREFIX_ADDRESS, 16,
460 						       1,
461 						       buffer_info->skb->data,
462 						       adapter->rx_buffer_len,
463 						       true);
464 			}
465 		}
466 	}
467 }
468 
469 /**
470  * e1000_desc_unused - calculate if we have unused descriptors
471  **/
472 static 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  **/
494 static 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  **/
519 static 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  * @staterr: descriptor extended error and status field as written by hardware
548  * @vlan: descriptor vlan field as written by hardware (no le/be conversion)
549  * @skb: pointer to sk_buff to be indicated to stack
550  **/
551 static void e1000_receive_skb(struct e1000_adapter *adapter,
552 			      struct net_device *netdev, struct sk_buff *skb,
553 			      u32 staterr, __le16 vlan)
554 {
555 	u16 tag = le16_to_cpu(vlan);
556 
557 	e1000e_rx_hwtstamp(adapter, staterr, skb);
558 
559 	skb->protocol = eth_type_trans(skb, netdev);
560 
561 	if (staterr & E1000_RXD_STAT_VP)
562 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), tag);
563 
564 	napi_gro_receive(&adapter->napi, skb);
565 }
566 
567 /**
568  * e1000_rx_checksum - Receive Checksum Offload
569  * @adapter: board private structure
570  * @status_err: receive descriptor status and error fields
571  * @csum: receive descriptor csum field
572  * @sk_buff: socket buffer with received data
573  **/
574 static 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 
606 static 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 	s32 ret_val = __ew32_prepare(hw);
611 
612 	writel(i, rx_ring->tail);
613 
614 	if (unlikely(!ret_val && (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 
623 static 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 	s32 ret_val = __ew32_prepare(hw);
628 
629 	writel(i, tx_ring->tail);
630 
631 	if (unlikely(!ret_val && (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  **/
644 static void e1000_alloc_rx_buffers(struct e1000_ring *rx_ring,
645 				   int cleaned_count, gfp_t gfp)
646 {
647 	struct e1000_adapter *adapter = rx_ring->adapter;
648 	struct net_device *netdev = adapter->netdev;
649 	struct pci_dev *pdev = adapter->pdev;
650 	union e1000_rx_desc_extended *rx_desc;
651 	struct e1000_buffer *buffer_info;
652 	struct sk_buff *skb;
653 	unsigned int i;
654 	unsigned int bufsz = adapter->rx_buffer_len;
655 
656 	i = rx_ring->next_to_use;
657 	buffer_info = &rx_ring->buffer_info[i];
658 
659 	while (cleaned_count--) {
660 		skb = buffer_info->skb;
661 		if (skb) {
662 			skb_trim(skb, 0);
663 			goto map_skb;
664 		}
665 
666 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
667 		if (!skb) {
668 			/* Better luck next round */
669 			adapter->alloc_rx_buff_failed++;
670 			break;
671 		}
672 
673 		buffer_info->skb = skb;
674 map_skb:
675 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
676 						  adapter->rx_buffer_len,
677 						  DMA_FROM_DEVICE);
678 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
679 			dev_err(&pdev->dev, "Rx DMA map failed\n");
680 			adapter->rx_dma_failed++;
681 			break;
682 		}
683 
684 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
685 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
686 
687 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
688 			/* Force memory writes to complete before letting h/w
689 			 * know there are new descriptors to fetch.  (Only
690 			 * applicable for weak-ordered memory model archs,
691 			 * such as IA-64).
692 			 */
693 			wmb();
694 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
695 				e1000e_update_rdt_wa(rx_ring, i);
696 			else
697 				writel(i, rx_ring->tail);
698 		}
699 		i++;
700 		if (i == rx_ring->count)
701 			i = 0;
702 		buffer_info = &rx_ring->buffer_info[i];
703 	}
704 
705 	rx_ring->next_to_use = i;
706 }
707 
708 /**
709  * e1000_alloc_rx_buffers_ps - Replace used receive buffers; packet split
710  * @rx_ring: Rx descriptor ring
711  **/
712 static void e1000_alloc_rx_buffers_ps(struct e1000_ring *rx_ring,
713 				      int cleaned_count, gfp_t gfp)
714 {
715 	struct e1000_adapter *adapter = rx_ring->adapter;
716 	struct net_device *netdev = adapter->netdev;
717 	struct pci_dev *pdev = adapter->pdev;
718 	union e1000_rx_desc_packet_split *rx_desc;
719 	struct e1000_buffer *buffer_info;
720 	struct e1000_ps_page *ps_page;
721 	struct sk_buff *skb;
722 	unsigned int i, j;
723 
724 	i = rx_ring->next_to_use;
725 	buffer_info = &rx_ring->buffer_info[i];
726 
727 	while (cleaned_count--) {
728 		rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
729 
730 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
731 			ps_page = &buffer_info->ps_pages[j];
732 			if (j >= adapter->rx_ps_pages) {
733 				/* all unused desc entries get hw null ptr */
734 				rx_desc->read.buffer_addr[j + 1] =
735 				    ~cpu_to_le64(0);
736 				continue;
737 			}
738 			if (!ps_page->page) {
739 				ps_page->page = alloc_page(gfp);
740 				if (!ps_page->page) {
741 					adapter->alloc_rx_buff_failed++;
742 					goto no_buffers;
743 				}
744 				ps_page->dma = dma_map_page(&pdev->dev,
745 							    ps_page->page,
746 							    0, PAGE_SIZE,
747 							    DMA_FROM_DEVICE);
748 				if (dma_mapping_error(&pdev->dev,
749 						      ps_page->dma)) {
750 					dev_err(&adapter->pdev->dev,
751 						"Rx DMA page map failed\n");
752 					adapter->rx_dma_failed++;
753 					goto no_buffers;
754 				}
755 			}
756 			/* Refresh the desc even if buffer_addrs
757 			 * didn't change because each write-back
758 			 * erases this info.
759 			 */
760 			rx_desc->read.buffer_addr[j + 1] =
761 			    cpu_to_le64(ps_page->dma);
762 		}
763 
764 		skb = __netdev_alloc_skb_ip_align(netdev, adapter->rx_ps_bsize0,
765 						  gfp);
766 
767 		if (!skb) {
768 			adapter->alloc_rx_buff_failed++;
769 			break;
770 		}
771 
772 		buffer_info->skb = skb;
773 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
774 						  adapter->rx_ps_bsize0,
775 						  DMA_FROM_DEVICE);
776 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
777 			dev_err(&pdev->dev, "Rx DMA map failed\n");
778 			adapter->rx_dma_failed++;
779 			/* cleanup skb */
780 			dev_kfree_skb_any(skb);
781 			buffer_info->skb = NULL;
782 			break;
783 		}
784 
785 		rx_desc->read.buffer_addr[0] = cpu_to_le64(buffer_info->dma);
786 
787 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
788 			/* Force memory writes to complete before letting h/w
789 			 * know there are new descriptors to fetch.  (Only
790 			 * applicable for weak-ordered memory model archs,
791 			 * such as IA-64).
792 			 */
793 			wmb();
794 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
795 				e1000e_update_rdt_wa(rx_ring, i << 1);
796 			else
797 				writel(i << 1, rx_ring->tail);
798 		}
799 
800 		i++;
801 		if (i == rx_ring->count)
802 			i = 0;
803 		buffer_info = &rx_ring->buffer_info[i];
804 	}
805 
806 no_buffers:
807 	rx_ring->next_to_use = i;
808 }
809 
810 /**
811  * e1000_alloc_jumbo_rx_buffers - Replace used jumbo receive buffers
812  * @rx_ring: Rx descriptor ring
813  * @cleaned_count: number of buffers to allocate this pass
814  **/
815 
816 static void e1000_alloc_jumbo_rx_buffers(struct e1000_ring *rx_ring,
817 					 int cleaned_count, gfp_t gfp)
818 {
819 	struct e1000_adapter *adapter = rx_ring->adapter;
820 	struct net_device *netdev = adapter->netdev;
821 	struct pci_dev *pdev = adapter->pdev;
822 	union e1000_rx_desc_extended *rx_desc;
823 	struct e1000_buffer *buffer_info;
824 	struct sk_buff *skb;
825 	unsigned int i;
826 	unsigned int bufsz = 256 - 16;	/* for skb_reserve */
827 
828 	i = rx_ring->next_to_use;
829 	buffer_info = &rx_ring->buffer_info[i];
830 
831 	while (cleaned_count--) {
832 		skb = buffer_info->skb;
833 		if (skb) {
834 			skb_trim(skb, 0);
835 			goto check_page;
836 		}
837 
838 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
839 		if (unlikely(!skb)) {
840 			/* Better luck next round */
841 			adapter->alloc_rx_buff_failed++;
842 			break;
843 		}
844 
845 		buffer_info->skb = skb;
846 check_page:
847 		/* allocate a new page if necessary */
848 		if (!buffer_info->page) {
849 			buffer_info->page = alloc_page(gfp);
850 			if (unlikely(!buffer_info->page)) {
851 				adapter->alloc_rx_buff_failed++;
852 				break;
853 			}
854 		}
855 
856 		if (!buffer_info->dma) {
857 			buffer_info->dma = dma_map_page(&pdev->dev,
858 							buffer_info->page, 0,
859 							PAGE_SIZE,
860 							DMA_FROM_DEVICE);
861 			if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
862 				adapter->alloc_rx_buff_failed++;
863 				break;
864 			}
865 		}
866 
867 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
868 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
869 
870 		if (unlikely(++i == rx_ring->count))
871 			i = 0;
872 		buffer_info = &rx_ring->buffer_info[i];
873 	}
874 
875 	if (likely(rx_ring->next_to_use != i)) {
876 		rx_ring->next_to_use = i;
877 		if (unlikely(i-- == 0))
878 			i = (rx_ring->count - 1);
879 
880 		/* Force memory writes to complete before letting h/w
881 		 * know there are new descriptors to fetch.  (Only
882 		 * applicable for weak-ordered memory model archs,
883 		 * such as IA-64).
884 		 */
885 		wmb();
886 		if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
887 			e1000e_update_rdt_wa(rx_ring, i);
888 		else
889 			writel(i, rx_ring->tail);
890 	}
891 }
892 
893 static inline void e1000_rx_hash(struct net_device *netdev, __le32 rss,
894 				 struct sk_buff *skb)
895 {
896 	if (netdev->features & NETIF_F_RXHASH)
897 		skb_set_hash(skb, le32_to_cpu(rss), PKT_HASH_TYPE_L3);
898 }
899 
900 /**
901  * e1000_clean_rx_irq - Send received data up the network stack
902  * @rx_ring: Rx descriptor ring
903  *
904  * the return value indicates whether actual cleaning was done, there
905  * is no guarantee that everything was cleaned
906  **/
907 static bool e1000_clean_rx_irq(struct e1000_ring *rx_ring, int *work_done,
908 			       int work_to_do)
909 {
910 	struct e1000_adapter *adapter = rx_ring->adapter;
911 	struct net_device *netdev = adapter->netdev;
912 	struct pci_dev *pdev = adapter->pdev;
913 	struct e1000_hw *hw = &adapter->hw;
914 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
915 	struct e1000_buffer *buffer_info, *next_buffer;
916 	u32 length, staterr;
917 	unsigned int i;
918 	int cleaned_count = 0;
919 	bool cleaned = false;
920 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
921 
922 	i = rx_ring->next_to_clean;
923 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
924 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
925 	buffer_info = &rx_ring->buffer_info[i];
926 
927 	while (staterr & E1000_RXD_STAT_DD) {
928 		struct sk_buff *skb;
929 
930 		if (*work_done >= work_to_do)
931 			break;
932 		(*work_done)++;
933 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
934 
935 		skb = buffer_info->skb;
936 		buffer_info->skb = NULL;
937 
938 		prefetch(skb->data - NET_IP_ALIGN);
939 
940 		i++;
941 		if (i == rx_ring->count)
942 			i = 0;
943 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
944 		prefetch(next_rxd);
945 
946 		next_buffer = &rx_ring->buffer_info[i];
947 
948 		cleaned = true;
949 		cleaned_count++;
950 		dma_unmap_single(&pdev->dev, buffer_info->dma,
951 				 adapter->rx_buffer_len, DMA_FROM_DEVICE);
952 		buffer_info->dma = 0;
953 
954 		length = le16_to_cpu(rx_desc->wb.upper.length);
955 
956 		/* !EOP means multiple descriptors were used to store a single
957 		 * packet, if that's the case we need to toss it.  In fact, we
958 		 * need to toss every packet with the EOP bit clear and the
959 		 * next frame that _does_ have the EOP bit set, as it is by
960 		 * definition only a frame fragment
961 		 */
962 		if (unlikely(!(staterr & E1000_RXD_STAT_EOP)))
963 			adapter->flags2 |= FLAG2_IS_DISCARDING;
964 
965 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
966 			/* All receives must fit into a single buffer */
967 			e_dbg("Receive packet consumed multiple buffers\n");
968 			/* recycle */
969 			buffer_info->skb = skb;
970 			if (staterr & E1000_RXD_STAT_EOP)
971 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
972 			goto next_desc;
973 		}
974 
975 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
976 			     !(netdev->features & NETIF_F_RXALL))) {
977 			/* recycle */
978 			buffer_info->skb = skb;
979 			goto next_desc;
980 		}
981 
982 		/* adjust length to remove Ethernet CRC */
983 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
984 			/* If configured to store CRC, don't subtract FCS,
985 			 * but keep the FCS bytes out of the total_rx_bytes
986 			 * counter
987 			 */
988 			if (netdev->features & NETIF_F_RXFCS)
989 				total_rx_bytes -= 4;
990 			else
991 				length -= 4;
992 		}
993 
994 		total_rx_bytes += length;
995 		total_rx_packets++;
996 
997 		/* code added for copybreak, this should improve
998 		 * performance for small packets with large amounts
999 		 * of reassembly being done in the stack
1000 		 */
1001 		if (length < copybreak) {
1002 			struct sk_buff *new_skb =
1003 				napi_alloc_skb(&adapter->napi, length);
1004 			if (new_skb) {
1005 				skb_copy_to_linear_data_offset(new_skb,
1006 							       -NET_IP_ALIGN,
1007 							       (skb->data -
1008 								NET_IP_ALIGN),
1009 							       (length +
1010 								NET_IP_ALIGN));
1011 				/* save the skb in buffer_info as good */
1012 				buffer_info->skb = skb;
1013 				skb = new_skb;
1014 			}
1015 			/* else just continue with the old one */
1016 		}
1017 		/* end copybreak code */
1018 		skb_put(skb, length);
1019 
1020 		/* Receive Checksum Offload */
1021 		e1000_rx_checksum(adapter, staterr, skb);
1022 
1023 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1024 
1025 		e1000_receive_skb(adapter, netdev, skb, staterr,
1026 				  rx_desc->wb.upper.vlan);
1027 
1028 next_desc:
1029 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1030 
1031 		/* return some buffers to hardware, one at a time is too slow */
1032 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1033 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1034 					      GFP_ATOMIC);
1035 			cleaned_count = 0;
1036 		}
1037 
1038 		/* use prefetched values */
1039 		rx_desc = next_rxd;
1040 		buffer_info = next_buffer;
1041 
1042 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1043 	}
1044 	rx_ring->next_to_clean = i;
1045 
1046 	cleaned_count = e1000_desc_unused(rx_ring);
1047 	if (cleaned_count)
1048 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1049 
1050 	adapter->total_rx_bytes += total_rx_bytes;
1051 	adapter->total_rx_packets += total_rx_packets;
1052 	return cleaned;
1053 }
1054 
1055 static void e1000_put_txbuf(struct e1000_ring *tx_ring,
1056 			    struct e1000_buffer *buffer_info,
1057 			    bool drop)
1058 {
1059 	struct e1000_adapter *adapter = tx_ring->adapter;
1060 
1061 	if (buffer_info->dma) {
1062 		if (buffer_info->mapped_as_page)
1063 			dma_unmap_page(&adapter->pdev->dev, buffer_info->dma,
1064 				       buffer_info->length, DMA_TO_DEVICE);
1065 		else
1066 			dma_unmap_single(&adapter->pdev->dev, buffer_info->dma,
1067 					 buffer_info->length, DMA_TO_DEVICE);
1068 		buffer_info->dma = 0;
1069 	}
1070 	if (buffer_info->skb) {
1071 		if (drop)
1072 			dev_kfree_skb_any(buffer_info->skb);
1073 		else
1074 			dev_consume_skb_any(buffer_info->skb);
1075 		buffer_info->skb = NULL;
1076 	}
1077 	buffer_info->time_stamp = 0;
1078 }
1079 
1080 static void e1000_print_hw_hang(struct work_struct *work)
1081 {
1082 	struct e1000_adapter *adapter = container_of(work,
1083 						     struct e1000_adapter,
1084 						     print_hang_task);
1085 	struct net_device *netdev = adapter->netdev;
1086 	struct e1000_ring *tx_ring = adapter->tx_ring;
1087 	unsigned int i = tx_ring->next_to_clean;
1088 	unsigned int eop = tx_ring->buffer_info[i].next_to_watch;
1089 	struct e1000_tx_desc *eop_desc = E1000_TX_DESC(*tx_ring, eop);
1090 	struct e1000_hw *hw = &adapter->hw;
1091 	u16 phy_status, phy_1000t_status, phy_ext_status;
1092 	u16 pci_status;
1093 
1094 	if (test_bit(__E1000_DOWN, &adapter->state))
1095 		return;
1096 
1097 	if (!adapter->tx_hang_recheck && (adapter->flags2 & FLAG2_DMA_BURST)) {
1098 		/* May be block on write-back, flush and detect again
1099 		 * flush pending descriptor writebacks to memory
1100 		 */
1101 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1102 		/* execute the writes immediately */
1103 		e1e_flush();
1104 		/* Due to rare timing issues, write to TIDV again to ensure
1105 		 * the write is successful
1106 		 */
1107 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1108 		/* execute the writes immediately */
1109 		e1e_flush();
1110 		adapter->tx_hang_recheck = true;
1111 		return;
1112 	}
1113 	adapter->tx_hang_recheck = false;
1114 
1115 	if (er32(TDH(0)) == er32(TDT(0))) {
1116 		e_dbg("false hang detected, ignoring\n");
1117 		return;
1118 	}
1119 
1120 	/* Real hang detected */
1121 	netif_stop_queue(netdev);
1122 
1123 	e1e_rphy(hw, MII_BMSR, &phy_status);
1124 	e1e_rphy(hw, MII_STAT1000, &phy_1000t_status);
1125 	e1e_rphy(hw, MII_ESTATUS, &phy_ext_status);
1126 
1127 	pci_read_config_word(adapter->pdev, PCI_STATUS, &pci_status);
1128 
1129 	/* detected Hardware unit hang */
1130 	e_err("Detected Hardware Unit Hang:\n"
1131 	      "  TDH                  <%x>\n"
1132 	      "  TDT                  <%x>\n"
1133 	      "  next_to_use          <%x>\n"
1134 	      "  next_to_clean        <%x>\n"
1135 	      "buffer_info[next_to_clean]:\n"
1136 	      "  time_stamp           <%lx>\n"
1137 	      "  next_to_watch        <%x>\n"
1138 	      "  jiffies              <%lx>\n"
1139 	      "  next_to_watch.status <%x>\n"
1140 	      "MAC Status             <%x>\n"
1141 	      "PHY Status             <%x>\n"
1142 	      "PHY 1000BASE-T Status  <%x>\n"
1143 	      "PHY Extended Status    <%x>\n"
1144 	      "PCI Status             <%x>\n",
1145 	      readl(tx_ring->head), readl(tx_ring->tail), tx_ring->next_to_use,
1146 	      tx_ring->next_to_clean, tx_ring->buffer_info[eop].time_stamp,
1147 	      eop, jiffies, eop_desc->upper.fields.status, er32(STATUS),
1148 	      phy_status, phy_1000t_status, phy_ext_status, pci_status);
1149 
1150 	e1000e_dump(adapter);
1151 
1152 	/* Suggest workaround for known h/w issue */
1153 	if ((hw->mac.type == e1000_pchlan) && (er32(CTRL) & E1000_CTRL_TFCE))
1154 		e_err("Try turning off Tx pause (flow control) via ethtool\n");
1155 }
1156 
1157 /**
1158  * e1000e_tx_hwtstamp_work - check for Tx time stamp
1159  * @work: pointer to work struct
1160  *
1161  * This work function polls the TSYNCTXCTL valid bit to determine when a
1162  * timestamp has been taken for the current stored skb.  The timestamp must
1163  * be for this skb because only one such packet is allowed in the queue.
1164  */
1165 static void e1000e_tx_hwtstamp_work(struct work_struct *work)
1166 {
1167 	struct e1000_adapter *adapter = container_of(work, struct e1000_adapter,
1168 						     tx_hwtstamp_work);
1169 	struct e1000_hw *hw = &adapter->hw;
1170 
1171 	if (er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID) {
1172 		struct sk_buff *skb = adapter->tx_hwtstamp_skb;
1173 		struct skb_shared_hwtstamps shhwtstamps;
1174 		u64 txstmp;
1175 
1176 		txstmp = er32(TXSTMPL);
1177 		txstmp |= (u64)er32(TXSTMPH) << 32;
1178 
1179 		e1000e_systim_to_hwtstamp(adapter, &shhwtstamps, txstmp);
1180 
1181 		/* Clear the global tx_hwtstamp_skb pointer and force writes
1182 		 * prior to notifying the stack of a Tx timestamp.
1183 		 */
1184 		adapter->tx_hwtstamp_skb = NULL;
1185 		wmb(); /* force write prior to skb_tstamp_tx */
1186 
1187 		skb_tstamp_tx(skb, &shhwtstamps);
1188 		dev_consume_skb_any(skb);
1189 	} else if (time_after(jiffies, adapter->tx_hwtstamp_start
1190 			      + adapter->tx_timeout_factor * HZ)) {
1191 		dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
1192 		adapter->tx_hwtstamp_skb = NULL;
1193 		adapter->tx_hwtstamp_timeouts++;
1194 		e_warn("clearing Tx timestamp hang\n");
1195 	} else {
1196 		/* reschedule to check later */
1197 		schedule_work(&adapter->tx_hwtstamp_work);
1198 	}
1199 }
1200 
1201 /**
1202  * e1000_clean_tx_irq - Reclaim resources after transmit completes
1203  * @tx_ring: Tx descriptor ring
1204  *
1205  * the return value indicates whether actual cleaning was done, there
1206  * is no guarantee that everything was cleaned
1207  **/
1208 static bool e1000_clean_tx_irq(struct e1000_ring *tx_ring)
1209 {
1210 	struct e1000_adapter *adapter = tx_ring->adapter;
1211 	struct net_device *netdev = adapter->netdev;
1212 	struct e1000_hw *hw = &adapter->hw;
1213 	struct e1000_tx_desc *tx_desc, *eop_desc;
1214 	struct e1000_buffer *buffer_info;
1215 	unsigned int i, eop;
1216 	unsigned int count = 0;
1217 	unsigned int total_tx_bytes = 0, total_tx_packets = 0;
1218 	unsigned int bytes_compl = 0, pkts_compl = 0;
1219 
1220 	i = tx_ring->next_to_clean;
1221 	eop = tx_ring->buffer_info[i].next_to_watch;
1222 	eop_desc = E1000_TX_DESC(*tx_ring, eop);
1223 
1224 	while ((eop_desc->upper.data & cpu_to_le32(E1000_TXD_STAT_DD)) &&
1225 	       (count < tx_ring->count)) {
1226 		bool cleaned = false;
1227 
1228 		dma_rmb();		/* read buffer_info after eop_desc */
1229 		for (; !cleaned; count++) {
1230 			tx_desc = E1000_TX_DESC(*tx_ring, i);
1231 			buffer_info = &tx_ring->buffer_info[i];
1232 			cleaned = (i == eop);
1233 
1234 			if (cleaned) {
1235 				total_tx_packets += buffer_info->segs;
1236 				total_tx_bytes += buffer_info->bytecount;
1237 				if (buffer_info->skb) {
1238 					bytes_compl += buffer_info->skb->len;
1239 					pkts_compl++;
1240 				}
1241 			}
1242 
1243 			e1000_put_txbuf(tx_ring, buffer_info, false);
1244 			tx_desc->upper.data = 0;
1245 
1246 			i++;
1247 			if (i == tx_ring->count)
1248 				i = 0;
1249 		}
1250 
1251 		if (i == tx_ring->next_to_use)
1252 			break;
1253 		eop = tx_ring->buffer_info[i].next_to_watch;
1254 		eop_desc = E1000_TX_DESC(*tx_ring, eop);
1255 	}
1256 
1257 	tx_ring->next_to_clean = i;
1258 
1259 	netdev_completed_queue(netdev, pkts_compl, bytes_compl);
1260 
1261 #define TX_WAKE_THRESHOLD 32
1262 	if (count && netif_carrier_ok(netdev) &&
1263 	    e1000_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD) {
1264 		/* Make sure that anybody stopping the queue after this
1265 		 * sees the new next_to_clean.
1266 		 */
1267 		smp_mb();
1268 
1269 		if (netif_queue_stopped(netdev) &&
1270 		    !(test_bit(__E1000_DOWN, &adapter->state))) {
1271 			netif_wake_queue(netdev);
1272 			++adapter->restart_queue;
1273 		}
1274 	}
1275 
1276 	if (adapter->detect_tx_hung) {
1277 		/* Detect a transmit hang in hardware, this serializes the
1278 		 * check with the clearing of time_stamp and movement of i
1279 		 */
1280 		adapter->detect_tx_hung = false;
1281 		if (tx_ring->buffer_info[i].time_stamp &&
1282 		    time_after(jiffies, tx_ring->buffer_info[i].time_stamp
1283 			       + (adapter->tx_timeout_factor * HZ)) &&
1284 		    !(er32(STATUS) & E1000_STATUS_TXOFF))
1285 			schedule_work(&adapter->print_hang_task);
1286 		else
1287 			adapter->tx_hang_recheck = false;
1288 	}
1289 	adapter->total_tx_bytes += total_tx_bytes;
1290 	adapter->total_tx_packets += total_tx_packets;
1291 	return count < tx_ring->count;
1292 }
1293 
1294 /**
1295  * e1000_clean_rx_irq_ps - Send received data up the network stack; packet split
1296  * @rx_ring: Rx descriptor ring
1297  *
1298  * the return value indicates whether actual cleaning was done, there
1299  * is no guarantee that everything was cleaned
1300  **/
1301 static bool e1000_clean_rx_irq_ps(struct e1000_ring *rx_ring, int *work_done,
1302 				  int work_to_do)
1303 {
1304 	struct e1000_adapter *adapter = rx_ring->adapter;
1305 	struct e1000_hw *hw = &adapter->hw;
1306 	union e1000_rx_desc_packet_split *rx_desc, *next_rxd;
1307 	struct net_device *netdev = adapter->netdev;
1308 	struct pci_dev *pdev = adapter->pdev;
1309 	struct e1000_buffer *buffer_info, *next_buffer;
1310 	struct e1000_ps_page *ps_page;
1311 	struct sk_buff *skb;
1312 	unsigned int i, j;
1313 	u32 length, staterr;
1314 	int cleaned_count = 0;
1315 	bool cleaned = false;
1316 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1317 
1318 	i = rx_ring->next_to_clean;
1319 	rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
1320 	staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1321 	buffer_info = &rx_ring->buffer_info[i];
1322 
1323 	while (staterr & E1000_RXD_STAT_DD) {
1324 		if (*work_done >= work_to_do)
1325 			break;
1326 		(*work_done)++;
1327 		skb = buffer_info->skb;
1328 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1329 
1330 		/* in the packet split case this is header only */
1331 		prefetch(skb->data - NET_IP_ALIGN);
1332 
1333 		i++;
1334 		if (i == rx_ring->count)
1335 			i = 0;
1336 		next_rxd = E1000_RX_DESC_PS(*rx_ring, i);
1337 		prefetch(next_rxd);
1338 
1339 		next_buffer = &rx_ring->buffer_info[i];
1340 
1341 		cleaned = true;
1342 		cleaned_count++;
1343 		dma_unmap_single(&pdev->dev, buffer_info->dma,
1344 				 adapter->rx_ps_bsize0, DMA_FROM_DEVICE);
1345 		buffer_info->dma = 0;
1346 
1347 		/* see !EOP comment in other Rx routine */
1348 		if (!(staterr & E1000_RXD_STAT_EOP))
1349 			adapter->flags2 |= FLAG2_IS_DISCARDING;
1350 
1351 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
1352 			e_dbg("Packet Split buffers didn't pick up the full packet\n");
1353 			dev_kfree_skb_irq(skb);
1354 			if (staterr & E1000_RXD_STAT_EOP)
1355 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1356 			goto next_desc;
1357 		}
1358 
1359 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1360 			     !(netdev->features & NETIF_F_RXALL))) {
1361 			dev_kfree_skb_irq(skb);
1362 			goto next_desc;
1363 		}
1364 
1365 		length = le16_to_cpu(rx_desc->wb.middle.length0);
1366 
1367 		if (!length) {
1368 			e_dbg("Last part of the packet spanning multiple descriptors\n");
1369 			dev_kfree_skb_irq(skb);
1370 			goto next_desc;
1371 		}
1372 
1373 		/* Good Receive */
1374 		skb_put(skb, length);
1375 
1376 		{
1377 			/* this looks ugly, but it seems compiler issues make
1378 			 * it more efficient than reusing j
1379 			 */
1380 			int l1 = le16_to_cpu(rx_desc->wb.upper.length[0]);
1381 
1382 			/* page alloc/put takes too long and effects small
1383 			 * packet throughput, so unsplit small packets and
1384 			 * save the alloc/put only valid in softirq (napi)
1385 			 * context to call kmap_*
1386 			 */
1387 			if (l1 && (l1 <= copybreak) &&
1388 			    ((length + l1) <= adapter->rx_ps_bsize0)) {
1389 				u8 *vaddr;
1390 
1391 				ps_page = &buffer_info->ps_pages[0];
1392 
1393 				/* there is no documentation about how to call
1394 				 * kmap_atomic, so we can't hold the mapping
1395 				 * very long
1396 				 */
1397 				dma_sync_single_for_cpu(&pdev->dev,
1398 							ps_page->dma,
1399 							PAGE_SIZE,
1400 							DMA_FROM_DEVICE);
1401 				vaddr = kmap_atomic(ps_page->page);
1402 				memcpy(skb_tail_pointer(skb), vaddr, l1);
1403 				kunmap_atomic(vaddr);
1404 				dma_sync_single_for_device(&pdev->dev,
1405 							   ps_page->dma,
1406 							   PAGE_SIZE,
1407 							   DMA_FROM_DEVICE);
1408 
1409 				/* remove the CRC */
1410 				if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1411 					if (!(netdev->features & NETIF_F_RXFCS))
1412 						l1 -= 4;
1413 				}
1414 
1415 				skb_put(skb, l1);
1416 				goto copydone;
1417 			}	/* if */
1418 		}
1419 
1420 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1421 			length = le16_to_cpu(rx_desc->wb.upper.length[j]);
1422 			if (!length)
1423 				break;
1424 
1425 			ps_page = &buffer_info->ps_pages[j];
1426 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1427 				       DMA_FROM_DEVICE);
1428 			ps_page->dma = 0;
1429 			skb_fill_page_desc(skb, j, ps_page->page, 0, length);
1430 			ps_page->page = NULL;
1431 			skb->len += length;
1432 			skb->data_len += length;
1433 			skb->truesize += PAGE_SIZE;
1434 		}
1435 
1436 		/* strip the ethernet crc, problem is we're using pages now so
1437 		 * this whole operation can get a little cpu intensive
1438 		 */
1439 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1440 			if (!(netdev->features & NETIF_F_RXFCS))
1441 				pskb_trim(skb, skb->len - 4);
1442 		}
1443 
1444 copydone:
1445 		total_rx_bytes += skb->len;
1446 		total_rx_packets++;
1447 
1448 		e1000_rx_checksum(adapter, staterr, skb);
1449 
1450 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1451 
1452 		if (rx_desc->wb.upper.header_status &
1453 		    cpu_to_le16(E1000_RXDPS_HDRSTAT_HDRSP))
1454 			adapter->rx_hdr_split++;
1455 
1456 		e1000_receive_skb(adapter, netdev, skb, staterr,
1457 				  rx_desc->wb.middle.vlan);
1458 
1459 next_desc:
1460 		rx_desc->wb.middle.status_error &= cpu_to_le32(~0xFF);
1461 		buffer_info->skb = NULL;
1462 
1463 		/* return some buffers to hardware, one at a time is too slow */
1464 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1465 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1466 					      GFP_ATOMIC);
1467 			cleaned_count = 0;
1468 		}
1469 
1470 		/* use prefetched values */
1471 		rx_desc = next_rxd;
1472 		buffer_info = next_buffer;
1473 
1474 		staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1475 	}
1476 	rx_ring->next_to_clean = i;
1477 
1478 	cleaned_count = e1000_desc_unused(rx_ring);
1479 	if (cleaned_count)
1480 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1481 
1482 	adapter->total_rx_bytes += total_rx_bytes;
1483 	adapter->total_rx_packets += total_rx_packets;
1484 	return cleaned;
1485 }
1486 
1487 /**
1488  * e1000_consume_page - helper function
1489  **/
1490 static void e1000_consume_page(struct e1000_buffer *bi, struct sk_buff *skb,
1491 			       u16 length)
1492 {
1493 	bi->page = NULL;
1494 	skb->len += length;
1495 	skb->data_len += length;
1496 	skb->truesize += PAGE_SIZE;
1497 }
1498 
1499 /**
1500  * e1000_clean_jumbo_rx_irq - Send received data up the network stack; legacy
1501  * @adapter: board private structure
1502  *
1503  * the return value indicates whether actual cleaning was done, there
1504  * is no guarantee that everything was cleaned
1505  **/
1506 static 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 					u8 *vaddr;
1605 					vaddr = kmap_atomic(buffer_info->page);
1606 					memcpy(skb_tail_pointer(skb), vaddr,
1607 					       length);
1608 					kunmap_atomic(vaddr);
1609 					/* re-use the page, so don't erase
1610 					 * buffer_info->page
1611 					 */
1612 					skb_put(skb, length);
1613 				} else {
1614 					skb_fill_page_desc(skb, 0,
1615 							   buffer_info->page, 0,
1616 							   length);
1617 					e1000_consume_page(buffer_info, skb,
1618 							   length);
1619 				}
1620 			}
1621 		}
1622 
1623 		/* Receive Checksum Offload */
1624 		e1000_rx_checksum(adapter, staterr, skb);
1625 
1626 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1627 
1628 		/* probably a little skewed due to removing CRC */
1629 		total_rx_bytes += skb->len;
1630 		total_rx_packets++;
1631 
1632 		/* eth type trans needs skb->data to point to something */
1633 		if (!pskb_may_pull(skb, ETH_HLEN)) {
1634 			e_err("pskb_may_pull failed.\n");
1635 			dev_kfree_skb_irq(skb);
1636 			goto next_desc;
1637 		}
1638 
1639 		e1000_receive_skb(adapter, netdev, skb, staterr,
1640 				  rx_desc->wb.upper.vlan);
1641 
1642 next_desc:
1643 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1644 
1645 		/* return some buffers to hardware, one at a time is too slow */
1646 		if (unlikely(cleaned_count >= E1000_RX_BUFFER_WRITE)) {
1647 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1648 					      GFP_ATOMIC);
1649 			cleaned_count = 0;
1650 		}
1651 
1652 		/* use prefetched values */
1653 		rx_desc = next_rxd;
1654 		buffer_info = next_buffer;
1655 
1656 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1657 	}
1658 	rx_ring->next_to_clean = i;
1659 
1660 	cleaned_count = e1000_desc_unused(rx_ring);
1661 	if (cleaned_count)
1662 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1663 
1664 	adapter->total_rx_bytes += total_rx_bytes;
1665 	adapter->total_rx_packets += total_rx_packets;
1666 	return cleaned;
1667 }
1668 
1669 /**
1670  * e1000_clean_rx_ring - Free Rx Buffers per Queue
1671  * @rx_ring: Rx descriptor ring
1672  **/
1673 static void e1000_clean_rx_ring(struct e1000_ring *rx_ring)
1674 {
1675 	struct e1000_adapter *adapter = rx_ring->adapter;
1676 	struct e1000_buffer *buffer_info;
1677 	struct e1000_ps_page *ps_page;
1678 	struct pci_dev *pdev = adapter->pdev;
1679 	unsigned int i, j;
1680 
1681 	/* Free all the Rx ring sk_buffs */
1682 	for (i = 0; i < rx_ring->count; i++) {
1683 		buffer_info = &rx_ring->buffer_info[i];
1684 		if (buffer_info->dma) {
1685 			if (adapter->clean_rx == e1000_clean_rx_irq)
1686 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1687 						 adapter->rx_buffer_len,
1688 						 DMA_FROM_DEVICE);
1689 			else if (adapter->clean_rx == e1000_clean_jumbo_rx_irq)
1690 				dma_unmap_page(&pdev->dev, buffer_info->dma,
1691 					       PAGE_SIZE, DMA_FROM_DEVICE);
1692 			else if (adapter->clean_rx == e1000_clean_rx_irq_ps)
1693 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1694 						 adapter->rx_ps_bsize0,
1695 						 DMA_FROM_DEVICE);
1696 			buffer_info->dma = 0;
1697 		}
1698 
1699 		if (buffer_info->page) {
1700 			put_page(buffer_info->page);
1701 			buffer_info->page = NULL;
1702 		}
1703 
1704 		if (buffer_info->skb) {
1705 			dev_kfree_skb(buffer_info->skb);
1706 			buffer_info->skb = NULL;
1707 		}
1708 
1709 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1710 			ps_page = &buffer_info->ps_pages[j];
1711 			if (!ps_page->page)
1712 				break;
1713 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1714 				       DMA_FROM_DEVICE);
1715 			ps_page->dma = 0;
1716 			put_page(ps_page->page);
1717 			ps_page->page = NULL;
1718 		}
1719 	}
1720 
1721 	/* there also may be some cached data from a chained receive */
1722 	if (rx_ring->rx_skb_top) {
1723 		dev_kfree_skb(rx_ring->rx_skb_top);
1724 		rx_ring->rx_skb_top = NULL;
1725 	}
1726 
1727 	/* Zero out the descriptor ring */
1728 	memset(rx_ring->desc, 0, rx_ring->size);
1729 
1730 	rx_ring->next_to_clean = 0;
1731 	rx_ring->next_to_use = 0;
1732 	adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1733 }
1734 
1735 static void e1000e_downshift_workaround(struct work_struct *work)
1736 {
1737 	struct e1000_adapter *adapter = container_of(work,
1738 						     struct e1000_adapter,
1739 						     downshift_task);
1740 
1741 	if (test_bit(__E1000_DOWN, &adapter->state))
1742 		return;
1743 
1744 	e1000e_gig_downshift_workaround_ich8lan(&adapter->hw);
1745 }
1746 
1747 /**
1748  * e1000_intr_msi - Interrupt Handler
1749  * @irq: interrupt number
1750  * @data: pointer to a network interface device structure
1751  **/
1752 static irqreturn_t e1000_intr_msi(int __always_unused irq, void *data)
1753 {
1754 	struct net_device *netdev = data;
1755 	struct e1000_adapter *adapter = netdev_priv(netdev);
1756 	struct e1000_hw *hw = &adapter->hw;
1757 	u32 icr = er32(ICR);
1758 
1759 	/* read ICR disables interrupts using IAM */
1760 	if (icr & E1000_ICR_LSC) {
1761 		hw->mac.get_link_status = true;
1762 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1763 		 * disconnect (LSC) before accessing any PHY registers
1764 		 */
1765 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1766 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1767 			schedule_work(&adapter->downshift_task);
1768 
1769 		/* 80003ES2LAN workaround-- For packet buffer work-around on
1770 		 * link down event; disable receives here in the ISR and reset
1771 		 * adapter in watchdog
1772 		 */
1773 		if (netif_carrier_ok(netdev) &&
1774 		    adapter->flags & FLAG_RX_NEEDS_RESTART) {
1775 			/* disable receives */
1776 			u32 rctl = er32(RCTL);
1777 
1778 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1779 			adapter->flags |= FLAG_RESTART_NOW;
1780 		}
1781 		/* guard against interrupt when we're going down */
1782 		if (!test_bit(__E1000_DOWN, &adapter->state))
1783 			mod_delayed_work(adapter->e1000_workqueue,
1784 					 &adapter->watchdog_task, HZ);
1785 	}
1786 
1787 	/* Reset on uncorrectable ECC error */
1788 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1789 		u32 pbeccsts = er32(PBECCSTS);
1790 
1791 		adapter->corr_errors +=
1792 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1793 		adapter->uncorr_errors +=
1794 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1795 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1796 
1797 		/* Do the reset outside of interrupt context */
1798 		schedule_work(&adapter->reset_task);
1799 
1800 		/* return immediately since reset is imminent */
1801 		return IRQ_HANDLED;
1802 	}
1803 
1804 	if (napi_schedule_prep(&adapter->napi)) {
1805 		adapter->total_tx_bytes = 0;
1806 		adapter->total_tx_packets = 0;
1807 		adapter->total_rx_bytes = 0;
1808 		adapter->total_rx_packets = 0;
1809 		__napi_schedule(&adapter->napi);
1810 	}
1811 
1812 	return IRQ_HANDLED;
1813 }
1814 
1815 /**
1816  * e1000_intr - Interrupt Handler
1817  * @irq: interrupt number
1818  * @data: pointer to a network interface device structure
1819  **/
1820 static irqreturn_t e1000_intr(int __always_unused irq, void *data)
1821 {
1822 	struct net_device *netdev = data;
1823 	struct e1000_adapter *adapter = netdev_priv(netdev);
1824 	struct e1000_hw *hw = &adapter->hw;
1825 	u32 rctl, icr = er32(ICR);
1826 
1827 	if (!icr || test_bit(__E1000_DOWN, &adapter->state))
1828 		return IRQ_NONE;	/* Not our interrupt */
1829 
1830 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
1831 	 * not set, then the adapter didn't send an interrupt
1832 	 */
1833 	if (!(icr & E1000_ICR_INT_ASSERTED))
1834 		return IRQ_NONE;
1835 
1836 	/* Interrupt Auto-Mask...upon reading ICR,
1837 	 * interrupts are masked.  No need for the
1838 	 * IMC write
1839 	 */
1840 
1841 	if (icr & E1000_ICR_LSC) {
1842 		hw->mac.get_link_status = true;
1843 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1844 		 * disconnect (LSC) before accessing any PHY registers
1845 		 */
1846 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1847 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1848 			schedule_work(&adapter->downshift_task);
1849 
1850 		/* 80003ES2LAN workaround--
1851 		 * For packet buffer work-around on link down event;
1852 		 * disable receives here in the ISR and
1853 		 * reset adapter in watchdog
1854 		 */
1855 		if (netif_carrier_ok(netdev) &&
1856 		    (adapter->flags & FLAG_RX_NEEDS_RESTART)) {
1857 			/* disable receives */
1858 			rctl = er32(RCTL);
1859 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1860 			adapter->flags |= FLAG_RESTART_NOW;
1861 		}
1862 		/* guard against interrupt when we're going down */
1863 		if (!test_bit(__E1000_DOWN, &adapter->state))
1864 			mod_delayed_work(adapter->e1000_workqueue,
1865 					 &adapter->watchdog_task, HZ);
1866 	}
1867 
1868 	/* Reset on uncorrectable ECC error */
1869 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1870 		u32 pbeccsts = er32(PBECCSTS);
1871 
1872 		adapter->corr_errors +=
1873 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1874 		adapter->uncorr_errors +=
1875 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1876 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1877 
1878 		/* Do the reset outside of interrupt context */
1879 		schedule_work(&adapter->reset_task);
1880 
1881 		/* return immediately since reset is imminent */
1882 		return IRQ_HANDLED;
1883 	}
1884 
1885 	if (napi_schedule_prep(&adapter->napi)) {
1886 		adapter->total_tx_bytes = 0;
1887 		adapter->total_tx_packets = 0;
1888 		adapter->total_rx_bytes = 0;
1889 		adapter->total_rx_packets = 0;
1890 		__napi_schedule(&adapter->napi);
1891 	}
1892 
1893 	return IRQ_HANDLED;
1894 }
1895 
1896 static irqreturn_t e1000_msix_other(int __always_unused irq, void *data)
1897 {
1898 	struct net_device *netdev = data;
1899 	struct e1000_adapter *adapter = netdev_priv(netdev);
1900 	struct e1000_hw *hw = &adapter->hw;
1901 	u32 icr = er32(ICR);
1902 
1903 	if (icr & adapter->eiac_mask)
1904 		ew32(ICS, (icr & adapter->eiac_mask));
1905 
1906 	if (icr & E1000_ICR_LSC) {
1907 		hw->mac.get_link_status = true;
1908 		/* guard against interrupt when we're going down */
1909 		if (!test_bit(__E1000_DOWN, &adapter->state))
1910 			mod_delayed_work(adapter->e1000_workqueue,
1911 					 &adapter->watchdog_task, HZ);
1912 	}
1913 
1914 	if (!test_bit(__E1000_DOWN, &adapter->state))
1915 		ew32(IMS, E1000_IMS_OTHER | IMS_OTHER_MASK);
1916 
1917 	return IRQ_HANDLED;
1918 }
1919 
1920 static irqreturn_t e1000_intr_msix_tx(int __always_unused irq, void *data)
1921 {
1922 	struct net_device *netdev = data;
1923 	struct e1000_adapter *adapter = netdev_priv(netdev);
1924 	struct e1000_hw *hw = &adapter->hw;
1925 	struct e1000_ring *tx_ring = adapter->tx_ring;
1926 
1927 	adapter->total_tx_bytes = 0;
1928 	adapter->total_tx_packets = 0;
1929 
1930 	if (!e1000_clean_tx_irq(tx_ring))
1931 		/* Ring was not completely cleaned, so fire another interrupt */
1932 		ew32(ICS, tx_ring->ims_val);
1933 
1934 	if (!test_bit(__E1000_DOWN, &adapter->state))
1935 		ew32(IMS, adapter->tx_ring->ims_val);
1936 
1937 	return IRQ_HANDLED;
1938 }
1939 
1940 static irqreturn_t e1000_intr_msix_rx(int __always_unused irq, void *data)
1941 {
1942 	struct net_device *netdev = data;
1943 	struct e1000_adapter *adapter = netdev_priv(netdev);
1944 	struct e1000_ring *rx_ring = adapter->rx_ring;
1945 
1946 	/* Write the ITR value calculated at the end of the
1947 	 * previous interrupt.
1948 	 */
1949 	if (rx_ring->set_itr) {
1950 		u32 itr = rx_ring->itr_val ?
1951 			  1000000000 / (rx_ring->itr_val * 256) : 0;
1952 
1953 		writel(itr, rx_ring->itr_register);
1954 		rx_ring->set_itr = 0;
1955 	}
1956 
1957 	if (napi_schedule_prep(&adapter->napi)) {
1958 		adapter->total_rx_bytes = 0;
1959 		adapter->total_rx_packets = 0;
1960 		__napi_schedule(&adapter->napi);
1961 	}
1962 	return IRQ_HANDLED;
1963 }
1964 
1965 /**
1966  * e1000_configure_msix - Configure MSI-X hardware
1967  *
1968  * e1000_configure_msix sets up the hardware to properly
1969  * generate MSI-X interrupts.
1970  **/
1971 static void e1000_configure_msix(struct e1000_adapter *adapter)
1972 {
1973 	struct e1000_hw *hw = &adapter->hw;
1974 	struct e1000_ring *rx_ring = adapter->rx_ring;
1975 	struct e1000_ring *tx_ring = adapter->tx_ring;
1976 	int vector = 0;
1977 	u32 ctrl_ext, ivar = 0;
1978 
1979 	adapter->eiac_mask = 0;
1980 
1981 	/* Workaround issue with spurious interrupts on 82574 in MSI-X mode */
1982 	if (hw->mac.type == e1000_82574) {
1983 		u32 rfctl = er32(RFCTL);
1984 
1985 		rfctl |= E1000_RFCTL_ACK_DIS;
1986 		ew32(RFCTL, rfctl);
1987 	}
1988 
1989 	/* Configure Rx vector */
1990 	rx_ring->ims_val = E1000_IMS_RXQ0;
1991 	adapter->eiac_mask |= rx_ring->ims_val;
1992 	if (rx_ring->itr_val)
1993 		writel(1000000000 / (rx_ring->itr_val * 256),
1994 		       rx_ring->itr_register);
1995 	else
1996 		writel(1, rx_ring->itr_register);
1997 	ivar = E1000_IVAR_INT_ALLOC_VALID | vector;
1998 
1999 	/* Configure Tx vector */
2000 	tx_ring->ims_val = E1000_IMS_TXQ0;
2001 	vector++;
2002 	if (tx_ring->itr_val)
2003 		writel(1000000000 / (tx_ring->itr_val * 256),
2004 		       tx_ring->itr_register);
2005 	else
2006 		writel(1, tx_ring->itr_register);
2007 	adapter->eiac_mask |= tx_ring->ims_val;
2008 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 8);
2009 
2010 	/* set vector for Other Causes, e.g. link changes */
2011 	vector++;
2012 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 16);
2013 	if (rx_ring->itr_val)
2014 		writel(1000000000 / (rx_ring->itr_val * 256),
2015 		       hw->hw_addr + E1000_EITR_82574(vector));
2016 	else
2017 		writel(1, hw->hw_addr + E1000_EITR_82574(vector));
2018 
2019 	/* Cause Tx interrupts on every write back */
2020 	ivar |= BIT(31);
2021 
2022 	ew32(IVAR, ivar);
2023 
2024 	/* enable MSI-X PBA support */
2025 	ctrl_ext = er32(CTRL_EXT) & ~E1000_CTRL_EXT_IAME;
2026 	ctrl_ext |= E1000_CTRL_EXT_PBA_CLR | E1000_CTRL_EXT_EIAME;
2027 	ew32(CTRL_EXT, ctrl_ext);
2028 	e1e_flush();
2029 }
2030 
2031 void e1000e_reset_interrupt_capability(struct e1000_adapter *adapter)
2032 {
2033 	if (adapter->msix_entries) {
2034 		pci_disable_msix(adapter->pdev);
2035 		kfree(adapter->msix_entries);
2036 		adapter->msix_entries = NULL;
2037 	} else if (adapter->flags & FLAG_MSI_ENABLED) {
2038 		pci_disable_msi(adapter->pdev);
2039 		adapter->flags &= ~FLAG_MSI_ENABLED;
2040 	}
2041 }
2042 
2043 /**
2044  * e1000e_set_interrupt_capability - set MSI or MSI-X if supported
2045  *
2046  * Attempt to configure interrupts using the best available
2047  * capabilities of the hardware and kernel.
2048  **/
2049 void e1000e_set_interrupt_capability(struct e1000_adapter *adapter)
2050 {
2051 	int err;
2052 	int i;
2053 
2054 	switch (adapter->int_mode) {
2055 	case E1000E_INT_MODE_MSIX:
2056 		if (adapter->flags & FLAG_HAS_MSIX) {
2057 			adapter->num_vectors = 3; /* RxQ0, TxQ0 and other */
2058 			adapter->msix_entries = kcalloc(adapter->num_vectors,
2059 							sizeof(struct
2060 							       msix_entry),
2061 							GFP_KERNEL);
2062 			if (adapter->msix_entries) {
2063 				struct e1000_adapter *a = adapter;
2064 
2065 				for (i = 0; i < adapter->num_vectors; i++)
2066 					adapter->msix_entries[i].entry = i;
2067 
2068 				err = pci_enable_msix_range(a->pdev,
2069 							    a->msix_entries,
2070 							    a->num_vectors,
2071 							    a->num_vectors);
2072 				if (err > 0)
2073 					return;
2074 			}
2075 			/* MSI-X failed, so fall through and try MSI */
2076 			e_err("Failed to initialize MSI-X interrupts.  Falling back to MSI interrupts.\n");
2077 			e1000e_reset_interrupt_capability(adapter);
2078 		}
2079 		adapter->int_mode = E1000E_INT_MODE_MSI;
2080 		/* Fall through */
2081 	case E1000E_INT_MODE_MSI:
2082 		if (!pci_enable_msi(adapter->pdev)) {
2083 			adapter->flags |= FLAG_MSI_ENABLED;
2084 		} else {
2085 			adapter->int_mode = E1000E_INT_MODE_LEGACY;
2086 			e_err("Failed to initialize MSI interrupts.  Falling back to legacy interrupts.\n");
2087 		}
2088 		/* Fall through */
2089 	case E1000E_INT_MODE_LEGACY:
2090 		/* Don't do anything; this is the system default */
2091 		break;
2092 	}
2093 
2094 	/* store the number of vectors being used */
2095 	adapter->num_vectors = 1;
2096 }
2097 
2098 /**
2099  * e1000_request_msix - Initialize MSI-X interrupts
2100  *
2101  * e1000_request_msix allocates MSI-X vectors and requests interrupts from the
2102  * kernel.
2103  **/
2104 static int e1000_request_msix(struct e1000_adapter *adapter)
2105 {
2106 	struct net_device *netdev = adapter->netdev;
2107 	int err = 0, vector = 0;
2108 
2109 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2110 		snprintf(adapter->rx_ring->name,
2111 			 sizeof(adapter->rx_ring->name) - 1,
2112 			 "%.14s-rx-0", netdev->name);
2113 	else
2114 		memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
2115 	err = request_irq(adapter->msix_entries[vector].vector,
2116 			  e1000_intr_msix_rx, 0, adapter->rx_ring->name,
2117 			  netdev);
2118 	if (err)
2119 		return err;
2120 	adapter->rx_ring->itr_register = adapter->hw.hw_addr +
2121 	    E1000_EITR_82574(vector);
2122 	adapter->rx_ring->itr_val = adapter->itr;
2123 	vector++;
2124 
2125 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2126 		snprintf(adapter->tx_ring->name,
2127 			 sizeof(adapter->tx_ring->name) - 1,
2128 			 "%.14s-tx-0", netdev->name);
2129 	else
2130 		memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
2131 	err = request_irq(adapter->msix_entries[vector].vector,
2132 			  e1000_intr_msix_tx, 0, adapter->tx_ring->name,
2133 			  netdev);
2134 	if (err)
2135 		return err;
2136 	adapter->tx_ring->itr_register = adapter->hw.hw_addr +
2137 	    E1000_EITR_82574(vector);
2138 	adapter->tx_ring->itr_val = adapter->itr;
2139 	vector++;
2140 
2141 	err = request_irq(adapter->msix_entries[vector].vector,
2142 			  e1000_msix_other, 0, netdev->name, netdev);
2143 	if (err)
2144 		return err;
2145 
2146 	e1000_configure_msix(adapter);
2147 
2148 	return 0;
2149 }
2150 
2151 /**
2152  * e1000_request_irq - initialize interrupts
2153  *
2154  * Attempts to configure interrupts using the best available
2155  * capabilities of the hardware and kernel.
2156  **/
2157 static int e1000_request_irq(struct e1000_adapter *adapter)
2158 {
2159 	struct net_device *netdev = adapter->netdev;
2160 	int err;
2161 
2162 	if (adapter->msix_entries) {
2163 		err = e1000_request_msix(adapter);
2164 		if (!err)
2165 			return err;
2166 		/* fall back to MSI */
2167 		e1000e_reset_interrupt_capability(adapter);
2168 		adapter->int_mode = E1000E_INT_MODE_MSI;
2169 		e1000e_set_interrupt_capability(adapter);
2170 	}
2171 	if (adapter->flags & FLAG_MSI_ENABLED) {
2172 		err = request_irq(adapter->pdev->irq, e1000_intr_msi, 0,
2173 				  netdev->name, netdev);
2174 		if (!err)
2175 			return err;
2176 
2177 		/* fall back to legacy interrupt */
2178 		e1000e_reset_interrupt_capability(adapter);
2179 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
2180 	}
2181 
2182 	err = request_irq(adapter->pdev->irq, e1000_intr, IRQF_SHARED,
2183 			  netdev->name, netdev);
2184 	if (err)
2185 		e_err("Unable to allocate interrupt, Error: %d\n", err);
2186 
2187 	return err;
2188 }
2189 
2190 static void e1000_free_irq(struct e1000_adapter *adapter)
2191 {
2192 	struct net_device *netdev = adapter->netdev;
2193 
2194 	if (adapter->msix_entries) {
2195 		int vector = 0;
2196 
2197 		free_irq(adapter->msix_entries[vector].vector, netdev);
2198 		vector++;
2199 
2200 		free_irq(adapter->msix_entries[vector].vector, netdev);
2201 		vector++;
2202 
2203 		/* Other Causes interrupt vector */
2204 		free_irq(adapter->msix_entries[vector].vector, netdev);
2205 		return;
2206 	}
2207 
2208 	free_irq(adapter->pdev->irq, netdev);
2209 }
2210 
2211 /**
2212  * e1000_irq_disable - Mask off interrupt generation on the NIC
2213  **/
2214 static void e1000_irq_disable(struct e1000_adapter *adapter)
2215 {
2216 	struct e1000_hw *hw = &adapter->hw;
2217 
2218 	ew32(IMC, ~0);
2219 	if (adapter->msix_entries)
2220 		ew32(EIAC_82574, 0);
2221 	e1e_flush();
2222 
2223 	if (adapter->msix_entries) {
2224 		int i;
2225 
2226 		for (i = 0; i < adapter->num_vectors; i++)
2227 			synchronize_irq(adapter->msix_entries[i].vector);
2228 	} else {
2229 		synchronize_irq(adapter->pdev->irq);
2230 	}
2231 }
2232 
2233 /**
2234  * e1000_irq_enable - Enable default interrupt generation settings
2235  **/
2236 static void e1000_irq_enable(struct e1000_adapter *adapter)
2237 {
2238 	struct e1000_hw *hw = &adapter->hw;
2239 
2240 	if (adapter->msix_entries) {
2241 		ew32(EIAC_82574, adapter->eiac_mask & E1000_EIAC_MASK_82574);
2242 		ew32(IMS, adapter->eiac_mask | E1000_IMS_OTHER |
2243 		     IMS_OTHER_MASK);
2244 	} else if (hw->mac.type >= e1000_pch_lpt) {
2245 		ew32(IMS, IMS_ENABLE_MASK | E1000_IMS_ECCER);
2246 	} else {
2247 		ew32(IMS, IMS_ENABLE_MASK);
2248 	}
2249 	e1e_flush();
2250 }
2251 
2252 /**
2253  * e1000e_get_hw_control - get control of the h/w from f/w
2254  * @adapter: address of board private structure
2255  *
2256  * e1000e_get_hw_control sets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2257  * For ASF and Pass Through versions of f/w this means that
2258  * the driver is loaded. For AMT version (only with 82573)
2259  * of the f/w this means that the network i/f is open.
2260  **/
2261 void e1000e_get_hw_control(struct e1000_adapter *adapter)
2262 {
2263 	struct e1000_hw *hw = &adapter->hw;
2264 	u32 ctrl_ext;
2265 	u32 swsm;
2266 
2267 	/* Let firmware know the driver has taken over */
2268 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2269 		swsm = er32(SWSM);
2270 		ew32(SWSM, swsm | E1000_SWSM_DRV_LOAD);
2271 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2272 		ctrl_ext = er32(CTRL_EXT);
2273 		ew32(CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2274 	}
2275 }
2276 
2277 /**
2278  * e1000e_release_hw_control - release control of the h/w to f/w
2279  * @adapter: address of board private structure
2280  *
2281  * e1000e_release_hw_control resets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2282  * For ASF and Pass Through versions of f/w this means that the
2283  * driver is no longer loaded. For AMT version (only with 82573) i
2284  * of the f/w this means that the network i/f is closed.
2285  *
2286  **/
2287 void e1000e_release_hw_control(struct e1000_adapter *adapter)
2288 {
2289 	struct e1000_hw *hw = &adapter->hw;
2290 	u32 ctrl_ext;
2291 	u32 swsm;
2292 
2293 	/* Let firmware taken over control of h/w */
2294 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2295 		swsm = er32(SWSM);
2296 		ew32(SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
2297 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2298 		ctrl_ext = er32(CTRL_EXT);
2299 		ew32(CTRL_EXT, ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2300 	}
2301 }
2302 
2303 /**
2304  * e1000_alloc_ring_dma - allocate memory for a ring structure
2305  **/
2306 static int e1000_alloc_ring_dma(struct e1000_adapter *adapter,
2307 				struct e1000_ring *ring)
2308 {
2309 	struct pci_dev *pdev = adapter->pdev;
2310 
2311 	ring->desc = dma_alloc_coherent(&pdev->dev, ring->size, &ring->dma,
2312 					GFP_KERNEL);
2313 	if (!ring->desc)
2314 		return -ENOMEM;
2315 
2316 	return 0;
2317 }
2318 
2319 /**
2320  * e1000e_setup_tx_resources - allocate Tx resources (Descriptors)
2321  * @tx_ring: Tx descriptor ring
2322  *
2323  * Return 0 on success, negative on failure
2324  **/
2325 int e1000e_setup_tx_resources(struct e1000_ring *tx_ring)
2326 {
2327 	struct e1000_adapter *adapter = tx_ring->adapter;
2328 	int err = -ENOMEM, size;
2329 
2330 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2331 	tx_ring->buffer_info = vzalloc(size);
2332 	if (!tx_ring->buffer_info)
2333 		goto err;
2334 
2335 	/* round up to nearest 4K */
2336 	tx_ring->size = tx_ring->count * sizeof(struct e1000_tx_desc);
2337 	tx_ring->size = ALIGN(tx_ring->size, 4096);
2338 
2339 	err = e1000_alloc_ring_dma(adapter, tx_ring);
2340 	if (err)
2341 		goto err;
2342 
2343 	tx_ring->next_to_use = 0;
2344 	tx_ring->next_to_clean = 0;
2345 
2346 	return 0;
2347 err:
2348 	vfree(tx_ring->buffer_info);
2349 	e_err("Unable to allocate memory for the transmit descriptor ring\n");
2350 	return err;
2351 }
2352 
2353 /**
2354  * e1000e_setup_rx_resources - allocate Rx resources (Descriptors)
2355  * @rx_ring: Rx descriptor ring
2356  *
2357  * Returns 0 on success, negative on failure
2358  **/
2359 int e1000e_setup_rx_resources(struct e1000_ring *rx_ring)
2360 {
2361 	struct e1000_adapter *adapter = rx_ring->adapter;
2362 	struct e1000_buffer *buffer_info;
2363 	int i, size, desc_len, err = -ENOMEM;
2364 
2365 	size = sizeof(struct e1000_buffer) * rx_ring->count;
2366 	rx_ring->buffer_info = vzalloc(size);
2367 	if (!rx_ring->buffer_info)
2368 		goto err;
2369 
2370 	for (i = 0; i < rx_ring->count; i++) {
2371 		buffer_info = &rx_ring->buffer_info[i];
2372 		buffer_info->ps_pages = kcalloc(PS_PAGE_BUFFERS,
2373 						sizeof(struct e1000_ps_page),
2374 						GFP_KERNEL);
2375 		if (!buffer_info->ps_pages)
2376 			goto err_pages;
2377 	}
2378 
2379 	desc_len = sizeof(union e1000_rx_desc_packet_split);
2380 
2381 	/* Round up to nearest 4K */
2382 	rx_ring->size = rx_ring->count * desc_len;
2383 	rx_ring->size = ALIGN(rx_ring->size, 4096);
2384 
2385 	err = e1000_alloc_ring_dma(adapter, rx_ring);
2386 	if (err)
2387 		goto err_pages;
2388 
2389 	rx_ring->next_to_clean = 0;
2390 	rx_ring->next_to_use = 0;
2391 	rx_ring->rx_skb_top = NULL;
2392 
2393 	return 0;
2394 
2395 err_pages:
2396 	for (i = 0; i < rx_ring->count; i++) {
2397 		buffer_info = &rx_ring->buffer_info[i];
2398 		kfree(buffer_info->ps_pages);
2399 	}
2400 err:
2401 	vfree(rx_ring->buffer_info);
2402 	e_err("Unable to allocate memory for the receive descriptor ring\n");
2403 	return err;
2404 }
2405 
2406 /**
2407  * e1000_clean_tx_ring - Free Tx Buffers
2408  * @tx_ring: Tx descriptor ring
2409  **/
2410 static void e1000_clean_tx_ring(struct e1000_ring *tx_ring)
2411 {
2412 	struct e1000_adapter *adapter = tx_ring->adapter;
2413 	struct e1000_buffer *buffer_info;
2414 	unsigned long size;
2415 	unsigned int i;
2416 
2417 	for (i = 0; i < tx_ring->count; i++) {
2418 		buffer_info = &tx_ring->buffer_info[i];
2419 		e1000_put_txbuf(tx_ring, buffer_info, false);
2420 	}
2421 
2422 	netdev_reset_queue(adapter->netdev);
2423 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2424 	memset(tx_ring->buffer_info, 0, size);
2425 
2426 	memset(tx_ring->desc, 0, tx_ring->size);
2427 
2428 	tx_ring->next_to_use = 0;
2429 	tx_ring->next_to_clean = 0;
2430 }
2431 
2432 /**
2433  * e1000e_free_tx_resources - Free Tx Resources per Queue
2434  * @tx_ring: Tx descriptor ring
2435  *
2436  * Free all transmit software resources
2437  **/
2438 void e1000e_free_tx_resources(struct e1000_ring *tx_ring)
2439 {
2440 	struct e1000_adapter *adapter = tx_ring->adapter;
2441 	struct pci_dev *pdev = adapter->pdev;
2442 
2443 	e1000_clean_tx_ring(tx_ring);
2444 
2445 	vfree(tx_ring->buffer_info);
2446 	tx_ring->buffer_info = NULL;
2447 
2448 	dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
2449 			  tx_ring->dma);
2450 	tx_ring->desc = NULL;
2451 }
2452 
2453 /**
2454  * e1000e_free_rx_resources - Free Rx Resources
2455  * @rx_ring: Rx descriptor ring
2456  *
2457  * Free all receive software resources
2458  **/
2459 void e1000e_free_rx_resources(struct e1000_ring *rx_ring)
2460 {
2461 	struct e1000_adapter *adapter = rx_ring->adapter;
2462 	struct pci_dev *pdev = adapter->pdev;
2463 	int i;
2464 
2465 	e1000_clean_rx_ring(rx_ring);
2466 
2467 	for (i = 0; i < rx_ring->count; i++)
2468 		kfree(rx_ring->buffer_info[i].ps_pages);
2469 
2470 	vfree(rx_ring->buffer_info);
2471 	rx_ring->buffer_info = NULL;
2472 
2473 	dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
2474 			  rx_ring->dma);
2475 	rx_ring->desc = NULL;
2476 }
2477 
2478 /**
2479  * e1000_update_itr - update the dynamic ITR value based on statistics
2480  * @adapter: pointer to adapter
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  **/
2494 static 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 
2537 static 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 		current_itr = 0;
2545 		new_itr = 4000;
2546 		goto set_itr_now;
2547 	}
2548 
2549 	if (adapter->flags2 & FLAG2_DISABLE_AIM) {
2550 		new_itr = 0;
2551 		goto set_itr_now;
2552 	}
2553 
2554 	adapter->tx_itr = e1000_update_itr(adapter->tx_itr,
2555 					   adapter->total_tx_packets,
2556 					   adapter->total_tx_bytes);
2557 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2558 	if (adapter->itr_setting == 3 && adapter->tx_itr == lowest_latency)
2559 		adapter->tx_itr = low_latency;
2560 
2561 	adapter->rx_itr = e1000_update_itr(adapter->rx_itr,
2562 					   adapter->total_rx_packets,
2563 					   adapter->total_rx_bytes);
2564 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2565 	if (adapter->itr_setting == 3 && adapter->rx_itr == lowest_latency)
2566 		adapter->rx_itr = low_latency;
2567 
2568 	current_itr = max(adapter->rx_itr, adapter->tx_itr);
2569 
2570 	/* counts and packets in update_itr are dependent on these numbers */
2571 	switch (current_itr) {
2572 	case lowest_latency:
2573 		new_itr = 70000;
2574 		break;
2575 	case low_latency:
2576 		new_itr = 20000;	/* aka hwitr = ~200 */
2577 		break;
2578 	case bulk_latency:
2579 		new_itr = 4000;
2580 		break;
2581 	default:
2582 		break;
2583 	}
2584 
2585 set_itr_now:
2586 	if (new_itr != adapter->itr) {
2587 		/* this attempts to bias the interrupt rate towards Bulk
2588 		 * by adding intermediate steps when interrupt rate is
2589 		 * increasing
2590 		 */
2591 		new_itr = new_itr > adapter->itr ?
2592 		    min(adapter->itr + (new_itr >> 2), new_itr) : new_itr;
2593 		adapter->itr = new_itr;
2594 		adapter->rx_ring->itr_val = new_itr;
2595 		if (adapter->msix_entries)
2596 			adapter->rx_ring->set_itr = 1;
2597 		else
2598 			e1000e_write_itr(adapter, new_itr);
2599 	}
2600 }
2601 
2602 /**
2603  * e1000e_write_itr - write the ITR value to the appropriate registers
2604  * @adapter: address of board private structure
2605  * @itr: new ITR value to program
2606  *
2607  * e1000e_write_itr determines if the adapter is in MSI-X mode
2608  * and, if so, writes the EITR registers with the ITR value.
2609  * Otherwise, it writes the ITR value into the ITR register.
2610  **/
2611 void e1000e_write_itr(struct e1000_adapter *adapter, u32 itr)
2612 {
2613 	struct e1000_hw *hw = &adapter->hw;
2614 	u32 new_itr = itr ? 1000000000 / (itr * 256) : 0;
2615 
2616 	if (adapter->msix_entries) {
2617 		int vector;
2618 
2619 		for (vector = 0; vector < adapter->num_vectors; vector++)
2620 			writel(new_itr, hw->hw_addr + E1000_EITR_82574(vector));
2621 	} else {
2622 		ew32(ITR, new_itr);
2623 	}
2624 }
2625 
2626 /**
2627  * e1000_alloc_queues - Allocate memory for all rings
2628  * @adapter: board private structure to initialize
2629  **/
2630 static int e1000_alloc_queues(struct e1000_adapter *adapter)
2631 {
2632 	int size = sizeof(struct e1000_ring);
2633 
2634 	adapter->tx_ring = kzalloc(size, GFP_KERNEL);
2635 	if (!adapter->tx_ring)
2636 		goto err;
2637 	adapter->tx_ring->count = adapter->tx_ring_count;
2638 	adapter->tx_ring->adapter = adapter;
2639 
2640 	adapter->rx_ring = kzalloc(size, GFP_KERNEL);
2641 	if (!adapter->rx_ring)
2642 		goto err;
2643 	adapter->rx_ring->count = adapter->rx_ring_count;
2644 	adapter->rx_ring->adapter = adapter;
2645 
2646 	return 0;
2647 err:
2648 	e_err("Unable to allocate memory for queues\n");
2649 	kfree(adapter->rx_ring);
2650 	kfree(adapter->tx_ring);
2651 	return -ENOMEM;
2652 }
2653 
2654 /**
2655  * e1000e_poll - NAPI Rx polling callback
2656  * @napi: struct associated with this polling callback
2657  * @budget: number of packets driver is allowed to process this poll
2658  **/
2659 static int e1000e_poll(struct napi_struct *napi, int budget)
2660 {
2661 	struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter,
2662 						     napi);
2663 	struct e1000_hw *hw = &adapter->hw;
2664 	struct net_device *poll_dev = adapter->netdev;
2665 	int tx_cleaned = 1, work_done = 0;
2666 
2667 	adapter = netdev_priv(poll_dev);
2668 
2669 	if (!adapter->msix_entries ||
2670 	    (adapter->rx_ring->ims_val & adapter->tx_ring->ims_val))
2671 		tx_cleaned = e1000_clean_tx_irq(adapter->tx_ring);
2672 
2673 	adapter->clean_rx(adapter->rx_ring, &work_done, budget);
2674 
2675 	if (!tx_cleaned || work_done == budget)
2676 		return budget;
2677 
2678 	/* Exit the polling mode, but don't re-enable interrupts if stack might
2679 	 * poll us due to busy-polling
2680 	 */
2681 	if (likely(napi_complete_done(napi, work_done))) {
2682 		if (adapter->itr_setting & 3)
2683 			e1000_set_itr(adapter);
2684 		if (!test_bit(__E1000_DOWN, &adapter->state)) {
2685 			if (adapter->msix_entries)
2686 				ew32(IMS, adapter->rx_ring->ims_val);
2687 			else
2688 				e1000_irq_enable(adapter);
2689 		}
2690 	}
2691 
2692 	return work_done;
2693 }
2694 
2695 static int e1000_vlan_rx_add_vid(struct net_device *netdev,
2696 				 __always_unused __be16 proto, u16 vid)
2697 {
2698 	struct e1000_adapter *adapter = netdev_priv(netdev);
2699 	struct e1000_hw *hw = &adapter->hw;
2700 	u32 vfta, index;
2701 
2702 	/* don't update vlan cookie if already programmed */
2703 	if ((adapter->hw.mng_cookie.status &
2704 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2705 	    (vid == adapter->mng_vlan_id))
2706 		return 0;
2707 
2708 	/* add VID to filter table */
2709 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2710 		index = (vid >> 5) & 0x7F;
2711 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2712 		vfta |= BIT((vid & 0x1F));
2713 		hw->mac.ops.write_vfta(hw, index, vfta);
2714 	}
2715 
2716 	set_bit(vid, adapter->active_vlans);
2717 
2718 	return 0;
2719 }
2720 
2721 static int e1000_vlan_rx_kill_vid(struct net_device *netdev,
2722 				  __always_unused __be16 proto, u16 vid)
2723 {
2724 	struct e1000_adapter *adapter = netdev_priv(netdev);
2725 	struct e1000_hw *hw = &adapter->hw;
2726 	u32 vfta, index;
2727 
2728 	if ((adapter->hw.mng_cookie.status &
2729 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2730 	    (vid == adapter->mng_vlan_id)) {
2731 		/* release control to f/w */
2732 		e1000e_release_hw_control(adapter);
2733 		return 0;
2734 	}
2735 
2736 	/* remove VID from filter table */
2737 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2738 		index = (vid >> 5) & 0x7F;
2739 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2740 		vfta &= ~BIT((vid & 0x1F));
2741 		hw->mac.ops.write_vfta(hw, index, vfta);
2742 	}
2743 
2744 	clear_bit(vid, adapter->active_vlans);
2745 
2746 	return 0;
2747 }
2748 
2749 /**
2750  * e1000e_vlan_filter_disable - helper to disable hw VLAN filtering
2751  * @adapter: board private structure to initialize
2752  **/
2753 static void e1000e_vlan_filter_disable(struct e1000_adapter *adapter)
2754 {
2755 	struct net_device *netdev = adapter->netdev;
2756 	struct e1000_hw *hw = &adapter->hw;
2757 	u32 rctl;
2758 
2759 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2760 		/* disable VLAN receive filtering */
2761 		rctl = er32(RCTL);
2762 		rctl &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
2763 		ew32(RCTL, rctl);
2764 
2765 		if (adapter->mng_vlan_id != (u16)E1000_MNG_VLAN_NONE) {
2766 			e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
2767 					       adapter->mng_vlan_id);
2768 			adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
2769 		}
2770 	}
2771 }
2772 
2773 /**
2774  * e1000e_vlan_filter_enable - helper to enable HW VLAN filtering
2775  * @adapter: board private structure to initialize
2776  **/
2777 static void e1000e_vlan_filter_enable(struct e1000_adapter *adapter)
2778 {
2779 	struct e1000_hw *hw = &adapter->hw;
2780 	u32 rctl;
2781 
2782 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2783 		/* enable VLAN receive filtering */
2784 		rctl = er32(RCTL);
2785 		rctl |= E1000_RCTL_VFE;
2786 		rctl &= ~E1000_RCTL_CFIEN;
2787 		ew32(RCTL, rctl);
2788 	}
2789 }
2790 
2791 /**
2792  * e1000e_vlan_strip_disable - helper to disable HW VLAN stripping
2793  * @adapter: board private structure to initialize
2794  **/
2795 static void e1000e_vlan_strip_disable(struct e1000_adapter *adapter)
2796 {
2797 	struct e1000_hw *hw = &adapter->hw;
2798 	u32 ctrl;
2799 
2800 	/* disable VLAN tag insert/strip */
2801 	ctrl = er32(CTRL);
2802 	ctrl &= ~E1000_CTRL_VME;
2803 	ew32(CTRL, ctrl);
2804 }
2805 
2806 /**
2807  * e1000e_vlan_strip_enable - helper to enable HW VLAN stripping
2808  * @adapter: board private structure to initialize
2809  **/
2810 static void e1000e_vlan_strip_enable(struct e1000_adapter *adapter)
2811 {
2812 	struct e1000_hw *hw = &adapter->hw;
2813 	u32 ctrl;
2814 
2815 	/* enable VLAN tag insert/strip */
2816 	ctrl = er32(CTRL);
2817 	ctrl |= E1000_CTRL_VME;
2818 	ew32(CTRL, ctrl);
2819 }
2820 
2821 static void e1000_update_mng_vlan(struct e1000_adapter *adapter)
2822 {
2823 	struct net_device *netdev = adapter->netdev;
2824 	u16 vid = adapter->hw.mng_cookie.vlan_id;
2825 	u16 old_vid = adapter->mng_vlan_id;
2826 
2827 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
2828 		e1000_vlan_rx_add_vid(netdev, htons(ETH_P_8021Q), vid);
2829 		adapter->mng_vlan_id = vid;
2830 	}
2831 
2832 	if ((old_vid != (u16)E1000_MNG_VLAN_NONE) && (vid != old_vid))
2833 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q), old_vid);
2834 }
2835 
2836 static void e1000_restore_vlan(struct e1000_adapter *adapter)
2837 {
2838 	u16 vid;
2839 
2840 	e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
2841 
2842 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
2843 	    e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
2844 }
2845 
2846 static void e1000_init_manageability_pt(struct e1000_adapter *adapter)
2847 {
2848 	struct e1000_hw *hw = &adapter->hw;
2849 	u32 manc, manc2h, mdef, i, j;
2850 
2851 	if (!(adapter->flags & FLAG_MNG_PT_ENABLED))
2852 		return;
2853 
2854 	manc = er32(MANC);
2855 
2856 	/* enable receiving management packets to the host. this will probably
2857 	 * generate destination unreachable messages from the host OS, but
2858 	 * the packets will be handled on SMBUS
2859 	 */
2860 	manc |= E1000_MANC_EN_MNG2HOST;
2861 	manc2h = er32(MANC2H);
2862 
2863 	switch (hw->mac.type) {
2864 	default:
2865 		manc2h |= (E1000_MANC2H_PORT_623 | E1000_MANC2H_PORT_664);
2866 		break;
2867 	case e1000_82574:
2868 	case e1000_82583:
2869 		/* Check if IPMI pass-through decision filter already exists;
2870 		 * if so, enable it.
2871 		 */
2872 		for (i = 0, j = 0; i < 8; i++) {
2873 			mdef = er32(MDEF(i));
2874 
2875 			/* Ignore filters with anything other than IPMI ports */
2876 			if (mdef & ~(E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2877 				continue;
2878 
2879 			/* Enable this decision filter in MANC2H */
2880 			if (mdef)
2881 				manc2h |= BIT(i);
2882 
2883 			j |= mdef;
2884 		}
2885 
2886 		if (j == (E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2887 			break;
2888 
2889 		/* Create new decision filter in an empty filter */
2890 		for (i = 0, j = 0; i < 8; i++)
2891 			if (er32(MDEF(i)) == 0) {
2892 				ew32(MDEF(i), (E1000_MDEF_PORT_623 |
2893 					       E1000_MDEF_PORT_664));
2894 				manc2h |= BIT(1);
2895 				j++;
2896 				break;
2897 			}
2898 
2899 		if (!j)
2900 			e_warn("Unable to create IPMI pass-through filter\n");
2901 		break;
2902 	}
2903 
2904 	ew32(MANC2H, manc2h);
2905 	ew32(MANC, manc);
2906 }
2907 
2908 /**
2909  * e1000_configure_tx - Configure Transmit Unit after Reset
2910  * @adapter: board private structure
2911  *
2912  * Configure the Tx unit of the MAC after a reset.
2913  **/
2914 static void e1000_configure_tx(struct e1000_adapter *adapter)
2915 {
2916 	struct e1000_hw *hw = &adapter->hw;
2917 	struct e1000_ring *tx_ring = adapter->tx_ring;
2918 	u64 tdba;
2919 	u32 tdlen, tctl, tarc;
2920 
2921 	/* Setup the HW Tx Head and Tail descriptor pointers */
2922 	tdba = tx_ring->dma;
2923 	tdlen = tx_ring->count * sizeof(struct e1000_tx_desc);
2924 	ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
2925 	ew32(TDBAH(0), (tdba >> 32));
2926 	ew32(TDLEN(0), tdlen);
2927 	ew32(TDH(0), 0);
2928 	ew32(TDT(0), 0);
2929 	tx_ring->head = adapter->hw.hw_addr + E1000_TDH(0);
2930 	tx_ring->tail = adapter->hw.hw_addr + E1000_TDT(0);
2931 
2932 	writel(0, tx_ring->head);
2933 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
2934 		e1000e_update_tdt_wa(tx_ring, 0);
2935 	else
2936 		writel(0, tx_ring->tail);
2937 
2938 	/* Set the Tx Interrupt Delay register */
2939 	ew32(TIDV, adapter->tx_int_delay);
2940 	/* Tx irq moderation */
2941 	ew32(TADV, adapter->tx_abs_int_delay);
2942 
2943 	if (adapter->flags2 & FLAG2_DMA_BURST) {
2944 		u32 txdctl = er32(TXDCTL(0));
2945 
2946 		txdctl &= ~(E1000_TXDCTL_PTHRESH | E1000_TXDCTL_HTHRESH |
2947 			    E1000_TXDCTL_WTHRESH);
2948 		/* set up some performance related parameters to encourage the
2949 		 * hardware to use the bus more efficiently in bursts, depends
2950 		 * on the tx_int_delay to be enabled,
2951 		 * wthresh = 1 ==> burst write is disabled to avoid Tx stalls
2952 		 * hthresh = 1 ==> prefetch when one or more available
2953 		 * pthresh = 0x1f ==> prefetch if internal cache 31 or less
2954 		 * BEWARE: this seems to work but should be considered first if
2955 		 * there are Tx hangs or other Tx related bugs
2956 		 */
2957 		txdctl |= E1000_TXDCTL_DMA_BURST_ENABLE;
2958 		ew32(TXDCTL(0), txdctl);
2959 	}
2960 	/* erratum work around: set txdctl the same for both queues */
2961 	ew32(TXDCTL(1), er32(TXDCTL(0)));
2962 
2963 	/* Program the Transmit Control Register */
2964 	tctl = er32(TCTL);
2965 	tctl &= ~E1000_TCTL_CT;
2966 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
2967 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
2968 
2969 	if (adapter->flags & FLAG_TARC_SPEED_MODE_BIT) {
2970 		tarc = er32(TARC(0));
2971 		/* set the speed mode bit, we'll clear it if we're not at
2972 		 * gigabit link later
2973 		 */
2974 #define SPEED_MODE_BIT BIT(21)
2975 		tarc |= SPEED_MODE_BIT;
2976 		ew32(TARC(0), tarc);
2977 	}
2978 
2979 	/* errata: program both queues to unweighted RR */
2980 	if (adapter->flags & FLAG_TARC_SET_BIT_ZERO) {
2981 		tarc = er32(TARC(0));
2982 		tarc |= 1;
2983 		ew32(TARC(0), tarc);
2984 		tarc = er32(TARC(1));
2985 		tarc |= 1;
2986 		ew32(TARC(1), tarc);
2987 	}
2988 
2989 	/* Setup Transmit Descriptor Settings for eop descriptor */
2990 	adapter->txd_cmd = E1000_TXD_CMD_EOP | E1000_TXD_CMD_IFCS;
2991 
2992 	/* only set IDE if we are delaying interrupts using the timers */
2993 	if (adapter->tx_int_delay)
2994 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
2995 
2996 	/* enable Report Status bit */
2997 	adapter->txd_cmd |= E1000_TXD_CMD_RS;
2998 
2999 	ew32(TCTL, tctl);
3000 
3001 	hw->mac.ops.config_collision_dist(hw);
3002 
3003 	/* SPT and KBL Si errata workaround to avoid data corruption */
3004 	if (hw->mac.type == e1000_pch_spt) {
3005 		u32 reg_val;
3006 
3007 		reg_val = er32(IOSFPC);
3008 		reg_val |= E1000_RCTL_RDMTS_HEX;
3009 		ew32(IOSFPC, reg_val);
3010 
3011 		reg_val = er32(TARC(0));
3012 		/* SPT and KBL Si errata workaround to avoid Tx hang.
3013 		 * Dropping the number of outstanding requests from
3014 		 * 3 to 2 in order to avoid a buffer overrun.
3015 		 */
3016 		reg_val &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3017 		reg_val |= E1000_TARC0_CB_MULTIQ_2_REQ;
3018 		ew32(TARC(0), reg_val);
3019 	}
3020 }
3021 
3022 /**
3023  * e1000_setup_rctl - configure the receive control registers
3024  * @adapter: Board private structure
3025  **/
3026 #define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
3027 			   (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
3028 static 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 			/* fall-through */
3147 		case 2:
3148 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE2_SHIFT;
3149 			/* fall-through */
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  **/
3186 static 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 		pm_qos_update_request(&adapter->pm_qos_req, lat);
3287 	} else {
3288 		pm_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  */
3305 static 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  **/
3343 static 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  **/
3400 static 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 
3451 static 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  **/
3491 s32 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 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3543 			/* Stable 24MHz frequency */
3544 			incperiod = INCPERIOD_24MHZ;
3545 			incvalue = INCVALUE_24MHZ;
3546 			shift = INCVALUE_SHIFT_24MHZ;
3547 			adapter->cc.shift = shift;
3548 		} else {
3549 			/* Stable 38400KHz frequency */
3550 			incperiod = INCPERIOD_38400KHZ;
3551 			incvalue = INCVALUE_38400KHZ;
3552 			shift = INCVALUE_SHIFT_38400KHZ;
3553 			adapter->cc.shift = shift;
3554 		}
3555 		break;
3556 	case e1000_82574:
3557 	case e1000_82583:
3558 		/* Stable 25MHz frequency */
3559 		incperiod = INCPERIOD_25MHZ;
3560 		incvalue = INCVALUE_25MHZ;
3561 		shift = INCVALUE_SHIFT_25MHZ;
3562 		adapter->cc.shift = shift;
3563 		break;
3564 	default:
3565 		return -EINVAL;
3566 	}
3567 
3568 	*timinca = ((incperiod << E1000_TIMINCA_INCPERIOD_SHIFT) |
3569 		    ((incvalue << shift) & E1000_TIMINCA_INCVALUE_MASK));
3570 
3571 	return 0;
3572 }
3573 
3574 /**
3575  * e1000e_config_hwtstamp - configure the hwtstamp registers and enable/disable
3576  * @adapter: board private structure
3577  *
3578  * Outgoing time stamping can be enabled and disabled. Play nice and
3579  * disable it when requested, although it shouldn't cause any overhead
3580  * when no packet needs it. At most one packet in the queue may be
3581  * marked for time stamping, otherwise it would be impossible to tell
3582  * for sure to which packet the hardware time stamp belongs.
3583  *
3584  * Incoming time stamping has to be configured via the hardware filters.
3585  * Not all combinations are supported, in particular event type has to be
3586  * specified. Matching the kind of event packet is not supported, with the
3587  * exception of "all V2 events regardless of level 2 or 4".
3588  **/
3589 static int e1000e_config_hwtstamp(struct e1000_adapter *adapter,
3590 				  struct hwtstamp_config *config)
3591 {
3592 	struct e1000_hw *hw = &adapter->hw;
3593 	u32 tsync_tx_ctl = E1000_TSYNCTXCTL_ENABLED;
3594 	u32 tsync_rx_ctl = E1000_TSYNCRXCTL_ENABLED;
3595 	u32 rxmtrl = 0;
3596 	u16 rxudp = 0;
3597 	bool is_l4 = false;
3598 	bool is_l2 = false;
3599 	u32 regval;
3600 
3601 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3602 		return -EINVAL;
3603 
3604 	/* flags reserved for future extensions - must be zero */
3605 	if (config->flags)
3606 		return -EINVAL;
3607 
3608 	switch (config->tx_type) {
3609 	case HWTSTAMP_TX_OFF:
3610 		tsync_tx_ctl = 0;
3611 		break;
3612 	case HWTSTAMP_TX_ON:
3613 		break;
3614 	default:
3615 		return -ERANGE;
3616 	}
3617 
3618 	switch (config->rx_filter) {
3619 	case HWTSTAMP_FILTER_NONE:
3620 		tsync_rx_ctl = 0;
3621 		break;
3622 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3623 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3624 		rxmtrl = E1000_RXMTRL_PTP_V1_SYNC_MESSAGE;
3625 		is_l4 = true;
3626 		break;
3627 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3628 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3629 		rxmtrl = E1000_RXMTRL_PTP_V1_DELAY_REQ_MESSAGE;
3630 		is_l4 = true;
3631 		break;
3632 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3633 		/* Also time stamps V2 L2 Path Delay Request/Response */
3634 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3635 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3636 		is_l2 = true;
3637 		break;
3638 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3639 		/* Also time stamps V2 L2 Path Delay Request/Response. */
3640 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3641 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3642 		is_l2 = true;
3643 		break;
3644 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3645 		/* Hardware cannot filter just V2 L4 Sync messages;
3646 		 * fall-through to V2 (both L2 and L4) Sync.
3647 		 */
3648 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
3649 		/* Also time stamps V2 Path Delay Request/Response. */
3650 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3651 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3652 		is_l2 = true;
3653 		is_l4 = true;
3654 		break;
3655 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3656 		/* Hardware cannot filter just V2 L4 Delay Request messages;
3657 		 * fall-through to V2 (both L2 and L4) Delay Request.
3658 		 */
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 		 * fall-through to all V2 (both L2 and L4) Events.
3670 		 */
3671 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
3672 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_EVENT_V2;
3673 		config->rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
3674 		is_l2 = true;
3675 		is_l4 = true;
3676 		break;
3677 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3678 		/* For V1, the hardware can only filter Sync messages or
3679 		 * Delay Request messages but not both so fall-through to
3680 		 * time stamp all packets.
3681 		 */
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  **/
3746 static 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  **/
3772 void 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  *
3783  * Power down the PHY so no link is implied when interface is down.
3784  * The PHY cannot be powered down if management or WoL is active.
3785  */
3786 static void e1000_power_down_phy(struct e1000_adapter *adapter)
3787 {
3788 	if (adapter->hw.phy.ops.power_down)
3789 		adapter->hw.phy.ops.power_down(&adapter->hw);
3790 }
3791 
3792 /**
3793  * e1000_flush_tx_ring - remove all descriptors from the tx_ring
3794  *
3795  * We want to clear all pending descriptors from the TX ring.
3796  * zeroing happens when the HW reads the regs. We  assign the ring itself as
3797  * the data of the next descriptor. We don't care about the data we are about
3798  * to reset the HW.
3799  */
3800 static void e1000_flush_tx_ring(struct e1000_adapter *adapter)
3801 {
3802 	struct e1000_hw *hw = &adapter->hw;
3803 	struct e1000_ring *tx_ring = adapter->tx_ring;
3804 	struct e1000_tx_desc *tx_desc = NULL;
3805 	u32 tdt, tctl, txd_lower = E1000_TXD_CMD_IFCS;
3806 	u16 size = 512;
3807 
3808 	tctl = er32(TCTL);
3809 	ew32(TCTL, tctl | E1000_TCTL_EN);
3810 	tdt = er32(TDT(0));
3811 	BUG_ON(tdt != tx_ring->next_to_use);
3812 	tx_desc =  E1000_TX_DESC(*tx_ring, tx_ring->next_to_use);
3813 	tx_desc->buffer_addr = tx_ring->dma;
3814 
3815 	tx_desc->lower.data = cpu_to_le32(txd_lower | size);
3816 	tx_desc->upper.data = 0;
3817 	/* flush descriptors to memory before notifying the HW */
3818 	wmb();
3819 	tx_ring->next_to_use++;
3820 	if (tx_ring->next_to_use == tx_ring->count)
3821 		tx_ring->next_to_use = 0;
3822 	ew32(TDT(0), tx_ring->next_to_use);
3823 	usleep_range(200, 250);
3824 }
3825 
3826 /**
3827  * e1000_flush_rx_ring - remove all descriptors from the rx_ring
3828  *
3829  * Mark all descriptors in the RX ring as consumed and disable the rx ring
3830  */
3831 static void e1000_flush_rx_ring(struct e1000_adapter *adapter)
3832 {
3833 	u32 rctl, rxdctl;
3834 	struct e1000_hw *hw = &adapter->hw;
3835 
3836 	rctl = er32(RCTL);
3837 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3838 	e1e_flush();
3839 	usleep_range(100, 150);
3840 
3841 	rxdctl = er32(RXDCTL(0));
3842 	/* zero the lower 14 bits (prefetch and host thresholds) */
3843 	rxdctl &= 0xffffc000;
3844 
3845 	/* update thresholds: prefetch threshold to 31, host threshold to 1
3846 	 * and make sure the granularity is "descriptors" and not "cache lines"
3847 	 */
3848 	rxdctl |= (0x1F | BIT(8) | E1000_RXDCTL_THRESH_UNIT_DESC);
3849 
3850 	ew32(RXDCTL(0), rxdctl);
3851 	/* momentarily enable the RX ring for the changes to take effect */
3852 	ew32(RCTL, rctl | E1000_RCTL_EN);
3853 	e1e_flush();
3854 	usleep_range(100, 150);
3855 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3856 }
3857 
3858 /**
3859  * e1000_flush_desc_rings - remove all descriptors from the descriptor rings
3860  *
3861  * In i219, the descriptor rings must be emptied before resetting the HW
3862  * or before changing the device state to D3 during runtime (runtime PM).
3863  *
3864  * Failure to do this will cause the HW to enter a unit hang state which can
3865  * only be released by PCI reset on the device
3866  *
3867  */
3868 
3869 static void e1000_flush_desc_rings(struct e1000_adapter *adapter)
3870 {
3871 	u16 hang_state;
3872 	u32 fext_nvm11, tdlen;
3873 	struct e1000_hw *hw = &adapter->hw;
3874 
3875 	/* First, disable MULR fix in FEXTNVM11 */
3876 	fext_nvm11 = er32(FEXTNVM11);
3877 	fext_nvm11 |= E1000_FEXTNVM11_DISABLE_MULR_FIX;
3878 	ew32(FEXTNVM11, fext_nvm11);
3879 	/* do nothing if we're not in faulty state, or if the queue is empty */
3880 	tdlen = er32(TDLEN(0));
3881 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3882 			     &hang_state);
3883 	if (!(hang_state & FLUSH_DESC_REQUIRED) || !tdlen)
3884 		return;
3885 	e1000_flush_tx_ring(adapter);
3886 	/* recheck, maybe the fault is caused by the rx ring */
3887 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3888 			     &hang_state);
3889 	if (hang_state & FLUSH_DESC_REQUIRED)
3890 		e1000_flush_rx_ring(adapter);
3891 }
3892 
3893 /**
3894  * e1000e_systim_reset - reset the timesync registers after a hardware reset
3895  * @adapter: board private structure
3896  *
3897  * When the MAC is reset, all hardware bits for timesync will be reset to the
3898  * default values. This function will restore the settings last in place.
3899  * Since the clock SYSTIME registers are reset, we will simply restore the
3900  * cyclecounter to the kernel real clock time.
3901  **/
3902 static void e1000e_systim_reset(struct e1000_adapter *adapter)
3903 {
3904 	struct ptp_clock_info *info = &adapter->ptp_clock_info;
3905 	struct e1000_hw *hw = &adapter->hw;
3906 	unsigned long flags;
3907 	u32 timinca;
3908 	s32 ret_val;
3909 
3910 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3911 		return;
3912 
3913 	if (info->adjfreq) {
3914 		/* restore the previous ptp frequency delta */
3915 		ret_val = info->adjfreq(info, adapter->ptp_delta);
3916 	} else {
3917 		/* set the default base frequency if no adjustment possible */
3918 		ret_val = e1000e_get_base_timinca(adapter, &timinca);
3919 		if (!ret_val)
3920 			ew32(TIMINCA, timinca);
3921 	}
3922 
3923 	if (ret_val) {
3924 		dev_warn(&adapter->pdev->dev,
3925 			 "Failed to restore TIMINCA clock rate delta: %d\n",
3926 			 ret_val);
3927 		return;
3928 	}
3929 
3930 	/* reset the systim ns time counter */
3931 	spin_lock_irqsave(&adapter->systim_lock, flags);
3932 	timecounter_init(&adapter->tc, &adapter->cc,
3933 			 ktime_to_ns(ktime_get_real()));
3934 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
3935 
3936 	/* restore the previous hwtstamp configuration settings */
3937 	e1000e_config_hwtstamp(adapter, &adapter->hwtstamp_config);
3938 }
3939 
3940 /**
3941  * e1000e_reset - bring the hardware into a known good state
3942  *
3943  * This function boots the hardware and enables some settings that
3944  * require a configuration cycle of the hardware - those cannot be
3945  * set/changed during runtime. After reset the device needs to be
3946  * properly configured for Rx, Tx etc.
3947  */
3948 void e1000e_reset(struct e1000_adapter *adapter)
3949 {
3950 	struct e1000_mac_info *mac = &adapter->hw.mac;
3951 	struct e1000_fc_info *fc = &adapter->hw.fc;
3952 	struct e1000_hw *hw = &adapter->hw;
3953 	u32 tx_space, min_tx_space, min_rx_space;
3954 	u32 pba = adapter->pba;
3955 	u16 hwm;
3956 
3957 	/* reset Packet Buffer Allocation to default */
3958 	ew32(PBA, pba);
3959 
3960 	if (adapter->max_frame_size > (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN)) {
3961 		/* To maintain wire speed transmits, the Tx FIFO should be
3962 		 * large enough to accommodate two full transmit packets,
3963 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
3964 		 * the Rx FIFO should be large enough to accommodate at least
3965 		 * one full receive packet and is similarly rounded up and
3966 		 * expressed in KB.
3967 		 */
3968 		pba = er32(PBA);
3969 		/* upper 16 bits has Tx packet buffer allocation size in KB */
3970 		tx_space = pba >> 16;
3971 		/* lower 16 bits has Rx packet buffer allocation size in KB */
3972 		pba &= 0xffff;
3973 		/* the Tx fifo also stores 16 bytes of information about the Tx
3974 		 * but don't include ethernet FCS because hardware appends it
3975 		 */
3976 		min_tx_space = (adapter->max_frame_size +
3977 				sizeof(struct e1000_tx_desc) - ETH_FCS_LEN) * 2;
3978 		min_tx_space = ALIGN(min_tx_space, 1024);
3979 		min_tx_space >>= 10;
3980 		/* software strips receive CRC, so leave room for it */
3981 		min_rx_space = adapter->max_frame_size;
3982 		min_rx_space = ALIGN(min_rx_space, 1024);
3983 		min_rx_space >>= 10;
3984 
3985 		/* If current Tx allocation is less than the min Tx FIFO size,
3986 		 * and the min Tx FIFO size is less than the current Rx FIFO
3987 		 * allocation, take space away from current Rx allocation
3988 		 */
3989 		if ((tx_space < min_tx_space) &&
3990 		    ((min_tx_space - tx_space) < pba)) {
3991 			pba -= min_tx_space - tx_space;
3992 
3993 			/* if short on Rx space, Rx wins and must trump Tx
3994 			 * adjustment
3995 			 */
3996 			if (pba < min_rx_space)
3997 				pba = min_rx_space;
3998 		}
3999 
4000 		ew32(PBA, pba);
4001 	}
4002 
4003 	/* flow control settings
4004 	 *
4005 	 * The high water mark must be low enough to fit one full frame
4006 	 * (or the size used for early receive) above it in the Rx FIFO.
4007 	 * Set it to the lower of:
4008 	 * - 90% of the Rx FIFO size, and
4009 	 * - the full Rx FIFO size minus one full frame
4010 	 */
4011 	if (adapter->flags & FLAG_DISABLE_FC_PAUSE_TIME)
4012 		fc->pause_time = 0xFFFF;
4013 	else
4014 		fc->pause_time = E1000_FC_PAUSE_TIME;
4015 	fc->send_xon = true;
4016 	fc->current_mode = fc->requested_mode;
4017 
4018 	switch (hw->mac.type) {
4019 	case e1000_ich9lan:
4020 	case e1000_ich10lan:
4021 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4022 			pba = 14;
4023 			ew32(PBA, pba);
4024 			fc->high_water = 0x2800;
4025 			fc->low_water = fc->high_water - 8;
4026 			break;
4027 		}
4028 		/* fall-through */
4029 	default:
4030 		hwm = min(((pba << 10) * 9 / 10),
4031 			  ((pba << 10) - adapter->max_frame_size));
4032 
4033 		fc->high_water = hwm & E1000_FCRTH_RTH;	/* 8-byte granularity */
4034 		fc->low_water = fc->high_water - 8;
4035 		break;
4036 	case e1000_pchlan:
4037 		/* Workaround PCH LOM adapter hangs with certain network
4038 		 * loads.  If hangs persist, try disabling Tx flow control.
4039 		 */
4040 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4041 			fc->high_water = 0x3500;
4042 			fc->low_water = 0x1500;
4043 		} else {
4044 			fc->high_water = 0x5000;
4045 			fc->low_water = 0x3000;
4046 		}
4047 		fc->refresh_time = 0x1000;
4048 		break;
4049 	case e1000_pch2lan:
4050 	case e1000_pch_lpt:
4051 	case e1000_pch_spt:
4052 	case e1000_pch_cnp:
4053 		/* fall-through */
4054 	case e1000_pch_tgp:
4055 		fc->refresh_time = 0xFFFF;
4056 		fc->pause_time = 0xFFFF;
4057 
4058 		if (adapter->netdev->mtu <= ETH_DATA_LEN) {
4059 			fc->high_water = 0x05C20;
4060 			fc->low_water = 0x05048;
4061 			break;
4062 		}
4063 
4064 		pba = 14;
4065 		ew32(PBA, pba);
4066 		fc->high_water = ((pba << 10) * 9 / 10) & E1000_FCRTH_RTH;
4067 		fc->low_water = ((pba << 10) * 8 / 10) & E1000_FCRTL_RTL;
4068 		break;
4069 	}
4070 
4071 	/* Alignment of Tx data is on an arbitrary byte boundary with the
4072 	 * maximum size per Tx descriptor limited only to the transmit
4073 	 * allocation of the packet buffer minus 96 bytes with an upper
4074 	 * limit of 24KB due to receive synchronization limitations.
4075 	 */
4076 	adapter->tx_fifo_limit = min_t(u32, ((er32(PBA) >> 16) << 10) - 96,
4077 				       24 << 10);
4078 
4079 	/* Disable Adaptive Interrupt Moderation if 2 full packets cannot
4080 	 * fit in receive buffer.
4081 	 */
4082 	if (adapter->itr_setting & 0x3) {
4083 		if ((adapter->max_frame_size * 2) > (pba << 10)) {
4084 			if (!(adapter->flags2 & FLAG2_DISABLE_AIM)) {
4085 				dev_info(&adapter->pdev->dev,
4086 					 "Interrupt Throttle Rate off\n");
4087 				adapter->flags2 |= FLAG2_DISABLE_AIM;
4088 				e1000e_write_itr(adapter, 0);
4089 			}
4090 		} else if (adapter->flags2 & FLAG2_DISABLE_AIM) {
4091 			dev_info(&adapter->pdev->dev,
4092 				 "Interrupt Throttle Rate on\n");
4093 			adapter->flags2 &= ~FLAG2_DISABLE_AIM;
4094 			adapter->itr = 20000;
4095 			e1000e_write_itr(adapter, adapter->itr);
4096 		}
4097 	}
4098 
4099 	if (hw->mac.type >= e1000_pch_spt)
4100 		e1000_flush_desc_rings(adapter);
4101 	/* Allow time for pending master requests to run */
4102 	mac->ops.reset_hw(hw);
4103 
4104 	/* For parts with AMT enabled, let the firmware know
4105 	 * that the network interface is in control
4106 	 */
4107 	if (adapter->flags & FLAG_HAS_AMT)
4108 		e1000e_get_hw_control(adapter);
4109 
4110 	ew32(WUC, 0);
4111 
4112 	if (mac->ops.init_hw(hw))
4113 		e_err("Hardware Error\n");
4114 
4115 	e1000_update_mng_vlan(adapter);
4116 
4117 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
4118 	ew32(VET, ETH_P_8021Q);
4119 
4120 	e1000e_reset_adaptive(hw);
4121 
4122 	/* restore systim and hwtstamp settings */
4123 	e1000e_systim_reset(adapter);
4124 
4125 	/* Set EEE advertisement as appropriate */
4126 	if (adapter->flags2 & FLAG2_HAS_EEE) {
4127 		s32 ret_val;
4128 		u16 adv_addr;
4129 
4130 		switch (hw->phy.type) {
4131 		case e1000_phy_82579:
4132 			adv_addr = I82579_EEE_ADVERTISEMENT;
4133 			break;
4134 		case e1000_phy_i217:
4135 			adv_addr = I217_EEE_ADVERTISEMENT;
4136 			break;
4137 		default:
4138 			dev_err(&adapter->pdev->dev,
4139 				"Invalid PHY type setting EEE advertisement\n");
4140 			return;
4141 		}
4142 
4143 		ret_val = hw->phy.ops.acquire(hw);
4144 		if (ret_val) {
4145 			dev_err(&adapter->pdev->dev,
4146 				"EEE advertisement - unable to acquire PHY\n");
4147 			return;
4148 		}
4149 
4150 		e1000_write_emi_reg_locked(hw, adv_addr,
4151 					   hw->dev_spec.ich8lan.eee_disable ?
4152 					   0 : adapter->eee_advert);
4153 
4154 		hw->phy.ops.release(hw);
4155 	}
4156 
4157 	if (!netif_running(adapter->netdev) &&
4158 	    !test_bit(__E1000_TESTING, &adapter->state))
4159 		e1000_power_down_phy(adapter);
4160 
4161 	e1000_get_phy_info(hw);
4162 
4163 	if ((adapter->flags & FLAG_HAS_SMART_POWER_DOWN) &&
4164 	    !(adapter->flags & FLAG_SMART_POWER_DOWN)) {
4165 		u16 phy_data = 0;
4166 		/* speed up time to link by disabling smart power down, ignore
4167 		 * the return value of this function because there is nothing
4168 		 * different we would do if it failed
4169 		 */
4170 		e1e_rphy(hw, IGP02E1000_PHY_POWER_MGMT, &phy_data);
4171 		phy_data &= ~IGP02E1000_PM_SPD;
4172 		e1e_wphy(hw, IGP02E1000_PHY_POWER_MGMT, phy_data);
4173 	}
4174 	if (hw->mac.type >= e1000_pch_spt && adapter->int_mode == 0) {
4175 		u32 reg;
4176 
4177 		/* Fextnvm7 @ 0xe4[2] = 1 */
4178 		reg = er32(FEXTNVM7);
4179 		reg |= E1000_FEXTNVM7_SIDE_CLK_UNGATE;
4180 		ew32(FEXTNVM7, reg);
4181 		/* Fextnvm9 @ 0x5bb4[13:12] = 11 */
4182 		reg = er32(FEXTNVM9);
4183 		reg |= E1000_FEXTNVM9_IOSFSB_CLKGATE_DIS |
4184 		       E1000_FEXTNVM9_IOSFSB_CLKREQ_DIS;
4185 		ew32(FEXTNVM9, reg);
4186 	}
4187 
4188 }
4189 
4190 /**
4191  * e1000e_trigger_lsc - trigger an LSC interrupt
4192  * @adapter:
4193  *
4194  * Fire a link status change interrupt to start the watchdog.
4195  **/
4196 static void e1000e_trigger_lsc(struct e1000_adapter *adapter)
4197 {
4198 	struct e1000_hw *hw = &adapter->hw;
4199 
4200 	if (adapter->msix_entries)
4201 		ew32(ICS, E1000_ICS_LSC | E1000_ICS_OTHER);
4202 	else
4203 		ew32(ICS, E1000_ICS_LSC);
4204 }
4205 
4206 void e1000e_up(struct e1000_adapter *adapter)
4207 {
4208 	/* hardware has been reset, we need to reload some things */
4209 	e1000_configure(adapter);
4210 
4211 	clear_bit(__E1000_DOWN, &adapter->state);
4212 
4213 	if (adapter->msix_entries)
4214 		e1000_configure_msix(adapter);
4215 	e1000_irq_enable(adapter);
4216 
4217 	/* Tx queue started by watchdog timer when link is up */
4218 
4219 	e1000e_trigger_lsc(adapter);
4220 }
4221 
4222 static void e1000e_flush_descriptors(struct e1000_adapter *adapter)
4223 {
4224 	struct e1000_hw *hw = &adapter->hw;
4225 
4226 	if (!(adapter->flags2 & FLAG2_DMA_BURST))
4227 		return;
4228 
4229 	/* flush pending descriptor writebacks to memory */
4230 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4231 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4232 
4233 	/* execute the writes immediately */
4234 	e1e_flush();
4235 
4236 	/* due to rare timing issues, write to TIDV/RDTR again to ensure the
4237 	 * write is successful
4238 	 */
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 
4246 static void e1000e_update_stats(struct e1000_adapter *adapter);
4247 
4248 /**
4249  * e1000e_down - quiesce the device and optionally reset the hardware
4250  * @adapter: board private structure
4251  * @reset: boolean flag to reset the hardware or not
4252  */
4253 void e1000e_down(struct e1000_adapter *adapter, bool reset)
4254 {
4255 	struct net_device *netdev = adapter->netdev;
4256 	struct e1000_hw *hw = &adapter->hw;
4257 	u32 tctl, rctl;
4258 
4259 	/* signal that we're down so the interrupt handler does not
4260 	 * reschedule our watchdog timer
4261 	 */
4262 	set_bit(__E1000_DOWN, &adapter->state);
4263 
4264 	netif_carrier_off(netdev);
4265 
4266 	/* disable receives in the hardware */
4267 	rctl = er32(RCTL);
4268 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
4269 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
4270 	/* flush and sleep below */
4271 
4272 	netif_stop_queue(netdev);
4273 
4274 	/* disable transmits in the hardware */
4275 	tctl = er32(TCTL);
4276 	tctl &= ~E1000_TCTL_EN;
4277 	ew32(TCTL, tctl);
4278 
4279 	/* flush both disables and wait for them to finish */
4280 	e1e_flush();
4281 	usleep_range(10000, 11000);
4282 
4283 	e1000_irq_disable(adapter);
4284 
4285 	napi_synchronize(&adapter->napi);
4286 
4287 	del_timer_sync(&adapter->phy_info_timer);
4288 
4289 	spin_lock(&adapter->stats64_lock);
4290 	e1000e_update_stats(adapter);
4291 	spin_unlock(&adapter->stats64_lock);
4292 
4293 	e1000e_flush_descriptors(adapter);
4294 
4295 	adapter->link_speed = 0;
4296 	adapter->link_duplex = 0;
4297 
4298 	/* Disable Si errata workaround on PCHx for jumbo frame flow */
4299 	if ((hw->mac.type >= e1000_pch2lan) &&
4300 	    (adapter->netdev->mtu > ETH_DATA_LEN) &&
4301 	    e1000_lv_jumbo_workaround_ich8lan(hw, false))
4302 		e_dbg("failed to disable jumbo frame workaround mode\n");
4303 
4304 	if (!pci_channel_offline(adapter->pdev)) {
4305 		if (reset)
4306 			e1000e_reset(adapter);
4307 		else if (hw->mac.type >= e1000_pch_spt)
4308 			e1000_flush_desc_rings(adapter);
4309 	}
4310 	e1000_clean_tx_ring(adapter->tx_ring);
4311 	e1000_clean_rx_ring(adapter->rx_ring);
4312 }
4313 
4314 void e1000e_reinit_locked(struct e1000_adapter *adapter)
4315 {
4316 	might_sleep();
4317 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
4318 		usleep_range(1000, 1100);
4319 	e1000e_down(adapter, true);
4320 	e1000e_up(adapter);
4321 	clear_bit(__E1000_RESETTING, &adapter->state);
4322 }
4323 
4324 /**
4325  * e1000e_sanitize_systim - sanitize raw cycle counter reads
4326  * @hw: pointer to the HW structure
4327  * @systim: PHC time value read, sanitized and returned
4328  * @sts: structure to hold system time before and after reading SYSTIML,
4329  * may be NULL
4330  *
4331  * Errata for 82574/82583 possible bad bits read from SYSTIMH/L:
4332  * check to see that the time is incrementing at a reasonable
4333  * rate and is a multiple of incvalue.
4334  **/
4335 static u64 e1000e_sanitize_systim(struct e1000_hw *hw, u64 systim,
4336 				  struct ptp_system_timestamp *sts)
4337 {
4338 	u64 time_delta, rem, temp;
4339 	u64 systim_next;
4340 	u32 incvalue;
4341 	int i;
4342 
4343 	incvalue = er32(TIMINCA) & E1000_TIMINCA_INCVALUE_MASK;
4344 	for (i = 0; i < E1000_MAX_82574_SYSTIM_REREADS; i++) {
4345 		/* latch SYSTIMH on read of SYSTIML */
4346 		ptp_read_system_prets(sts);
4347 		systim_next = (u64)er32(SYSTIML);
4348 		ptp_read_system_postts(sts);
4349 		systim_next |= (u64)er32(SYSTIMH) << 32;
4350 
4351 		time_delta = systim_next - systim;
4352 		temp = time_delta;
4353 		/* VMWare users have seen incvalue of zero, don't div / 0 */
4354 		rem = incvalue ? do_div(temp, incvalue) : (time_delta != 0);
4355 
4356 		systim = systim_next;
4357 
4358 		if ((time_delta < E1000_82574_SYSTIM_EPSILON) && (rem == 0))
4359 			break;
4360 	}
4361 
4362 	return systim;
4363 }
4364 
4365 /**
4366  * e1000e_read_systim - read SYSTIM register
4367  * @adapter: board private structure
4368  * @sts: structure which will contain system time before and after reading
4369  * SYSTIML, may be NULL
4370  **/
4371 u64 e1000e_read_systim(struct e1000_adapter *adapter,
4372 		       struct ptp_system_timestamp *sts)
4373 {
4374 	struct e1000_hw *hw = &adapter->hw;
4375 	u32 systimel, systimel_2, systimeh;
4376 	u64 systim;
4377 	/* SYSTIMH latching upon SYSTIML read does not work well.
4378 	 * This means that if SYSTIML overflows after we read it but before
4379 	 * we read SYSTIMH, the value of SYSTIMH has been incremented and we
4380 	 * will experience a huge non linear increment in the systime value
4381 	 * to fix that we test for overflow and if true, we re-read systime.
4382 	 */
4383 	ptp_read_system_prets(sts);
4384 	systimel = er32(SYSTIML);
4385 	ptp_read_system_postts(sts);
4386 	systimeh = er32(SYSTIMH);
4387 	/* Is systimel is so large that overflow is possible? */
4388 	if (systimel >= (u32)0xffffffff - E1000_TIMINCA_INCVALUE_MASK) {
4389 		ptp_read_system_prets(sts);
4390 		systimel_2 = er32(SYSTIML);
4391 		ptp_read_system_postts(sts);
4392 		if (systimel > systimel_2) {
4393 			/* There was an overflow, read again SYSTIMH, and use
4394 			 * systimel_2
4395 			 */
4396 			systimeh = er32(SYSTIMH);
4397 			systimel = systimel_2;
4398 		}
4399 	}
4400 	systim = (u64)systimel;
4401 	systim |= (u64)systimeh << 32;
4402 
4403 	if (adapter->flags2 & FLAG2_CHECK_SYSTIM_OVERFLOW)
4404 		systim = e1000e_sanitize_systim(hw, systim, sts);
4405 
4406 	return systim;
4407 }
4408 
4409 /**
4410  * e1000e_cyclecounter_read - read raw cycle counter (used by time counter)
4411  * @cc: cyclecounter structure
4412  **/
4413 static u64 e1000e_cyclecounter_read(const struct cyclecounter *cc)
4414 {
4415 	struct e1000_adapter *adapter = container_of(cc, struct e1000_adapter,
4416 						     cc);
4417 
4418 	return e1000e_read_systim(adapter, NULL);
4419 }
4420 
4421 /**
4422  * e1000_sw_init - Initialize general software structures (struct e1000_adapter)
4423  * @adapter: board private structure to initialize
4424  *
4425  * e1000_sw_init initializes the Adapter private data structure.
4426  * Fields are initialized based on PCI device information and
4427  * OS network device settings (MTU size).
4428  **/
4429 static int e1000_sw_init(struct e1000_adapter *adapter)
4430 {
4431 	struct net_device *netdev = adapter->netdev;
4432 
4433 	adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
4434 	adapter->rx_ps_bsize0 = 128;
4435 	adapter->max_frame_size = netdev->mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
4436 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4437 	adapter->tx_ring_count = E1000_DEFAULT_TXD;
4438 	adapter->rx_ring_count = E1000_DEFAULT_RXD;
4439 
4440 	spin_lock_init(&adapter->stats64_lock);
4441 
4442 	e1000e_set_interrupt_capability(adapter);
4443 
4444 	if (e1000_alloc_queues(adapter))
4445 		return -ENOMEM;
4446 
4447 	/* Setup hardware time stamping cyclecounter */
4448 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
4449 		adapter->cc.read = e1000e_cyclecounter_read;
4450 		adapter->cc.mask = CYCLECOUNTER_MASK(64);
4451 		adapter->cc.mult = 1;
4452 		/* cc.shift set in e1000e_get_base_tininca() */
4453 
4454 		spin_lock_init(&adapter->systim_lock);
4455 		INIT_WORK(&adapter->tx_hwtstamp_work, e1000e_tx_hwtstamp_work);
4456 	}
4457 
4458 	/* Explicitly disable IRQ since the NIC can be in any state. */
4459 	e1000_irq_disable(adapter);
4460 
4461 	set_bit(__E1000_DOWN, &adapter->state);
4462 	return 0;
4463 }
4464 
4465 /**
4466  * e1000_intr_msi_test - Interrupt Handler
4467  * @irq: interrupt number
4468  * @data: pointer to a network interface device structure
4469  **/
4470 static irqreturn_t e1000_intr_msi_test(int __always_unused irq, void *data)
4471 {
4472 	struct net_device *netdev = data;
4473 	struct e1000_adapter *adapter = netdev_priv(netdev);
4474 	struct e1000_hw *hw = &adapter->hw;
4475 	u32 icr = er32(ICR);
4476 
4477 	e_dbg("icr is %08X\n", icr);
4478 	if (icr & E1000_ICR_RXSEQ) {
4479 		adapter->flags &= ~FLAG_MSI_TEST_FAILED;
4480 		/* Force memory writes to complete before acknowledging the
4481 		 * interrupt is handled.
4482 		 */
4483 		wmb();
4484 	}
4485 
4486 	return IRQ_HANDLED;
4487 }
4488 
4489 /**
4490  * e1000_test_msi_interrupt - Returns 0 for successful test
4491  * @adapter: board private struct
4492  *
4493  * code flow taken from tg3.c
4494  **/
4495 static int e1000_test_msi_interrupt(struct e1000_adapter *adapter)
4496 {
4497 	struct net_device *netdev = adapter->netdev;
4498 	struct e1000_hw *hw = &adapter->hw;
4499 	int err;
4500 
4501 	/* poll_enable hasn't been called yet, so don't need disable */
4502 	/* clear any pending events */
4503 	er32(ICR);
4504 
4505 	/* free the real vector and request a test handler */
4506 	e1000_free_irq(adapter);
4507 	e1000e_reset_interrupt_capability(adapter);
4508 
4509 	/* Assume that the test fails, if it succeeds then the test
4510 	 * MSI irq handler will unset this flag
4511 	 */
4512 	adapter->flags |= FLAG_MSI_TEST_FAILED;
4513 
4514 	err = pci_enable_msi(adapter->pdev);
4515 	if (err)
4516 		goto msi_test_failed;
4517 
4518 	err = request_irq(adapter->pdev->irq, e1000_intr_msi_test, 0,
4519 			  netdev->name, netdev);
4520 	if (err) {
4521 		pci_disable_msi(adapter->pdev);
4522 		goto msi_test_failed;
4523 	}
4524 
4525 	/* Force memory writes to complete before enabling and firing an
4526 	 * interrupt.
4527 	 */
4528 	wmb();
4529 
4530 	e1000_irq_enable(adapter);
4531 
4532 	/* fire an unusual interrupt on the test handler */
4533 	ew32(ICS, E1000_ICS_RXSEQ);
4534 	e1e_flush();
4535 	msleep(100);
4536 
4537 	e1000_irq_disable(adapter);
4538 
4539 	rmb();			/* read flags after interrupt has been fired */
4540 
4541 	if (adapter->flags & FLAG_MSI_TEST_FAILED) {
4542 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
4543 		e_info("MSI interrupt test failed, using legacy interrupt.\n");
4544 	} else {
4545 		e_dbg("MSI interrupt test succeeded!\n");
4546 	}
4547 
4548 	free_irq(adapter->pdev->irq, netdev);
4549 	pci_disable_msi(adapter->pdev);
4550 
4551 msi_test_failed:
4552 	e1000e_set_interrupt_capability(adapter);
4553 	return e1000_request_irq(adapter);
4554 }
4555 
4556 /**
4557  * e1000_test_msi - Returns 0 if MSI test succeeds or INTx mode is restored
4558  * @adapter: board private struct
4559  *
4560  * code flow taken from tg3.c, called with e1000 interrupts disabled.
4561  **/
4562 static int e1000_test_msi(struct e1000_adapter *adapter)
4563 {
4564 	int err;
4565 	u16 pci_cmd;
4566 
4567 	if (!(adapter->flags & FLAG_MSI_ENABLED))
4568 		return 0;
4569 
4570 	/* disable SERR in case the MSI write causes a master abort */
4571 	pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4572 	if (pci_cmd & PCI_COMMAND_SERR)
4573 		pci_write_config_word(adapter->pdev, PCI_COMMAND,
4574 				      pci_cmd & ~PCI_COMMAND_SERR);
4575 
4576 	err = e1000_test_msi_interrupt(adapter);
4577 
4578 	/* re-enable SERR */
4579 	if (pci_cmd & PCI_COMMAND_SERR) {
4580 		pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4581 		pci_cmd |= PCI_COMMAND_SERR;
4582 		pci_write_config_word(adapter->pdev, PCI_COMMAND, pci_cmd);
4583 	}
4584 
4585 	return err;
4586 }
4587 
4588 /**
4589  * e1000e_open - Called when a network interface is made active
4590  * @netdev: network interface device structure
4591  *
4592  * Returns 0 on success, negative value on failure
4593  *
4594  * The open entry point is called when a network interface is made
4595  * active by the system (IFF_UP).  At this point all resources needed
4596  * for transmit and receive operations are allocated, the interrupt
4597  * handler is registered with the OS, the watchdog timer is started,
4598  * and the stack is notified that the interface is ready.
4599  **/
4600 int e1000e_open(struct net_device *netdev)
4601 {
4602 	struct e1000_adapter *adapter = netdev_priv(netdev);
4603 	struct e1000_hw *hw = &adapter->hw;
4604 	struct pci_dev *pdev = adapter->pdev;
4605 	int err;
4606 
4607 	/* disallow open during test */
4608 	if (test_bit(__E1000_TESTING, &adapter->state))
4609 		return -EBUSY;
4610 
4611 	pm_runtime_get_sync(&pdev->dev);
4612 
4613 	netif_carrier_off(netdev);
4614 	netif_stop_queue(netdev);
4615 
4616 	/* allocate transmit descriptors */
4617 	err = e1000e_setup_tx_resources(adapter->tx_ring);
4618 	if (err)
4619 		goto err_setup_tx;
4620 
4621 	/* allocate receive descriptors */
4622 	err = e1000e_setup_rx_resources(adapter->rx_ring);
4623 	if (err)
4624 		goto err_setup_rx;
4625 
4626 	/* If AMT is enabled, let the firmware know that the network
4627 	 * interface is now open and reset the part to a known state.
4628 	 */
4629 	if (adapter->flags & FLAG_HAS_AMT) {
4630 		e1000e_get_hw_control(adapter);
4631 		e1000e_reset(adapter);
4632 	}
4633 
4634 	e1000e_power_up_phy(adapter);
4635 
4636 	adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
4637 	if ((adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
4638 		e1000_update_mng_vlan(adapter);
4639 
4640 	/* DMA latency requirement to workaround jumbo issue */
4641 	pm_qos_add_request(&adapter->pm_qos_req, PM_QOS_CPU_DMA_LATENCY,
4642 			   PM_QOS_DEFAULT_VALUE);
4643 
4644 	/* before we allocate an interrupt, we must be ready to handle it.
4645 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4646 	 * as soon as we call pci_request_irq, so we have to setup our
4647 	 * clean_rx handler before we do so.
4648 	 */
4649 	e1000_configure(adapter);
4650 
4651 	err = e1000_request_irq(adapter);
4652 	if (err)
4653 		goto err_req_irq;
4654 
4655 	/* Work around PCIe errata with MSI interrupts causing some chipsets to
4656 	 * ignore e1000e MSI messages, which means we need to test our MSI
4657 	 * interrupt now
4658 	 */
4659 	if (adapter->int_mode != E1000E_INT_MODE_LEGACY) {
4660 		err = e1000_test_msi(adapter);
4661 		if (err) {
4662 			e_err("Interrupt allocation failed\n");
4663 			goto err_req_irq;
4664 		}
4665 	}
4666 
4667 	/* From here on the code is the same as e1000e_up() */
4668 	clear_bit(__E1000_DOWN, &adapter->state);
4669 
4670 	napi_enable(&adapter->napi);
4671 
4672 	e1000_irq_enable(adapter);
4673 
4674 	adapter->tx_hang_recheck = false;
4675 
4676 	hw->mac.get_link_status = true;
4677 	pm_runtime_put(&pdev->dev);
4678 
4679 	e1000e_trigger_lsc(adapter);
4680 
4681 	return 0;
4682 
4683 err_req_irq:
4684 	pm_qos_remove_request(&adapter->pm_qos_req);
4685 	e1000e_release_hw_control(adapter);
4686 	e1000_power_down_phy(adapter);
4687 	e1000e_free_rx_resources(adapter->rx_ring);
4688 err_setup_rx:
4689 	e1000e_free_tx_resources(adapter->tx_ring);
4690 err_setup_tx:
4691 	e1000e_reset(adapter);
4692 	pm_runtime_put_sync(&pdev->dev);
4693 
4694 	return err;
4695 }
4696 
4697 /**
4698  * e1000e_close - Disables a network interface
4699  * @netdev: network interface device structure
4700  *
4701  * Returns 0, this is not allowed to fail
4702  *
4703  * The close entry point is called when an interface is de-activated
4704  * by the OS.  The hardware is still under the drivers control, but
4705  * needs to be disabled.  A global MAC reset is issued to stop the
4706  * hardware, and all transmit and receive resources are freed.
4707  **/
4708 int e1000e_close(struct net_device *netdev)
4709 {
4710 	struct e1000_adapter *adapter = netdev_priv(netdev);
4711 	struct pci_dev *pdev = adapter->pdev;
4712 	int count = E1000_CHECK_RESET_COUNT;
4713 
4714 	while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
4715 		usleep_range(10000, 11000);
4716 
4717 	WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
4718 
4719 	pm_runtime_get_sync(&pdev->dev);
4720 
4721 	if (netif_device_present(netdev)) {
4722 		e1000e_down(adapter, true);
4723 		e1000_free_irq(adapter);
4724 
4725 		/* Link status message must follow this format */
4726 		pr_info("%s NIC Link is Down\n", netdev->name);
4727 	}
4728 
4729 	napi_disable(&adapter->napi);
4730 
4731 	e1000e_free_tx_resources(adapter->tx_ring);
4732 	e1000e_free_rx_resources(adapter->rx_ring);
4733 
4734 	/* kill manageability vlan ID if supported, but not if a vlan with
4735 	 * the same ID is registered on the host OS (let 8021q kill it)
4736 	 */
4737 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN)
4738 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
4739 				       adapter->mng_vlan_id);
4740 
4741 	/* If AMT is enabled, let the firmware know that the network
4742 	 * interface is now closed
4743 	 */
4744 	if ((adapter->flags & FLAG_HAS_AMT) &&
4745 	    !test_bit(__E1000_TESTING, &adapter->state))
4746 		e1000e_release_hw_control(adapter);
4747 
4748 	pm_qos_remove_request(&adapter->pm_qos_req);
4749 
4750 	pm_runtime_put_sync(&pdev->dev);
4751 
4752 	return 0;
4753 }
4754 
4755 /**
4756  * e1000_set_mac - Change the Ethernet Address of the NIC
4757  * @netdev: network interface device structure
4758  * @p: pointer to an address structure
4759  *
4760  * Returns 0 on success, negative on failure
4761  **/
4762 static int e1000_set_mac(struct net_device *netdev, void *p)
4763 {
4764 	struct e1000_adapter *adapter = netdev_priv(netdev);
4765 	struct e1000_hw *hw = &adapter->hw;
4766 	struct sockaddr *addr = p;
4767 
4768 	if (!is_valid_ether_addr(addr->sa_data))
4769 		return -EADDRNOTAVAIL;
4770 
4771 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
4772 	memcpy(adapter->hw.mac.addr, addr->sa_data, netdev->addr_len);
4773 
4774 	hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
4775 
4776 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA) {
4777 		/* activate the work around */
4778 		e1000e_set_laa_state_82571(&adapter->hw, 1);
4779 
4780 		/* Hold a copy of the LAA in RAR[14] This is done so that
4781 		 * between the time RAR[0] gets clobbered  and the time it
4782 		 * gets fixed (in e1000_watchdog), the actual LAA is in one
4783 		 * of the RARs and no incoming packets directed to this port
4784 		 * are dropped. Eventually the LAA will be in RAR[0] and
4785 		 * RAR[14]
4786 		 */
4787 		hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr,
4788 				    adapter->hw.mac.rar_entry_count - 1);
4789 	}
4790 
4791 	return 0;
4792 }
4793 
4794 /**
4795  * e1000e_update_phy_task - work thread to update phy
4796  * @work: pointer to our work struct
4797  *
4798  * this worker thread exists because we must acquire a
4799  * semaphore to read the phy, which we could msleep while
4800  * waiting for it, and we can't msleep in a timer.
4801  **/
4802 static void e1000e_update_phy_task(struct work_struct *work)
4803 {
4804 	struct e1000_adapter *adapter = container_of(work,
4805 						     struct e1000_adapter,
4806 						     update_phy_task);
4807 	struct e1000_hw *hw = &adapter->hw;
4808 
4809 	if (test_bit(__E1000_DOWN, &adapter->state))
4810 		return;
4811 
4812 	e1000_get_phy_info(hw);
4813 
4814 	/* Enable EEE on 82579 after link up */
4815 	if (hw->phy.type >= e1000_phy_82579)
4816 		e1000_set_eee_pchlan(hw);
4817 }
4818 
4819 /**
4820  * e1000_update_phy_info - timre call-back to update PHY info
4821  * @data: pointer to adapter cast into an unsigned long
4822  *
4823  * Need to wait a few seconds after link up to get diagnostic information from
4824  * the phy
4825  **/
4826 static void e1000_update_phy_info(struct timer_list *t)
4827 {
4828 	struct e1000_adapter *adapter = from_timer(adapter, t, phy_info_timer);
4829 
4830 	if (test_bit(__E1000_DOWN, &adapter->state))
4831 		return;
4832 
4833 	schedule_work(&adapter->update_phy_task);
4834 }
4835 
4836 /**
4837  * e1000e_update_phy_stats - Update the PHY statistics counters
4838  * @adapter: board private structure
4839  *
4840  * Read/clear the upper 16-bit PHY registers and read/accumulate lower
4841  **/
4842 static void e1000e_update_phy_stats(struct e1000_adapter *adapter)
4843 {
4844 	struct e1000_hw *hw = &adapter->hw;
4845 	s32 ret_val;
4846 	u16 phy_data;
4847 
4848 	ret_val = hw->phy.ops.acquire(hw);
4849 	if (ret_val)
4850 		return;
4851 
4852 	/* A page set is expensive so check if already on desired page.
4853 	 * If not, set to the page with the PHY status registers.
4854 	 */
4855 	hw->phy.addr = 1;
4856 	ret_val = e1000e_read_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4857 					   &phy_data);
4858 	if (ret_val)
4859 		goto release;
4860 	if (phy_data != (HV_STATS_PAGE << IGP_PAGE_SHIFT)) {
4861 		ret_val = hw->phy.ops.set_page(hw,
4862 					       HV_STATS_PAGE << IGP_PAGE_SHIFT);
4863 		if (ret_val)
4864 			goto release;
4865 	}
4866 
4867 	/* Single Collision Count */
4868 	hw->phy.ops.read_reg_page(hw, HV_SCC_UPPER, &phy_data);
4869 	ret_val = hw->phy.ops.read_reg_page(hw, HV_SCC_LOWER, &phy_data);
4870 	if (!ret_val)
4871 		adapter->stats.scc += phy_data;
4872 
4873 	/* Excessive Collision Count */
4874 	hw->phy.ops.read_reg_page(hw, HV_ECOL_UPPER, &phy_data);
4875 	ret_val = hw->phy.ops.read_reg_page(hw, HV_ECOL_LOWER, &phy_data);
4876 	if (!ret_val)
4877 		adapter->stats.ecol += phy_data;
4878 
4879 	/* Multiple Collision Count */
4880 	hw->phy.ops.read_reg_page(hw, HV_MCC_UPPER, &phy_data);
4881 	ret_val = hw->phy.ops.read_reg_page(hw, HV_MCC_LOWER, &phy_data);
4882 	if (!ret_val)
4883 		adapter->stats.mcc += phy_data;
4884 
4885 	/* Late Collision Count */
4886 	hw->phy.ops.read_reg_page(hw, HV_LATECOL_UPPER, &phy_data);
4887 	ret_val = hw->phy.ops.read_reg_page(hw, HV_LATECOL_LOWER, &phy_data);
4888 	if (!ret_val)
4889 		adapter->stats.latecol += phy_data;
4890 
4891 	/* Collision Count - also used for adaptive IFS */
4892 	hw->phy.ops.read_reg_page(hw, HV_COLC_UPPER, &phy_data);
4893 	ret_val = hw->phy.ops.read_reg_page(hw, HV_COLC_LOWER, &phy_data);
4894 	if (!ret_val)
4895 		hw->mac.collision_delta = phy_data;
4896 
4897 	/* Defer Count */
4898 	hw->phy.ops.read_reg_page(hw, HV_DC_UPPER, &phy_data);
4899 	ret_val = hw->phy.ops.read_reg_page(hw, HV_DC_LOWER, &phy_data);
4900 	if (!ret_val)
4901 		adapter->stats.dc += phy_data;
4902 
4903 	/* Transmit with no CRS */
4904 	hw->phy.ops.read_reg_page(hw, HV_TNCRS_UPPER, &phy_data);
4905 	ret_val = hw->phy.ops.read_reg_page(hw, HV_TNCRS_LOWER, &phy_data);
4906 	if (!ret_val)
4907 		adapter->stats.tncrs += phy_data;
4908 
4909 release:
4910 	hw->phy.ops.release(hw);
4911 }
4912 
4913 /**
4914  * e1000e_update_stats - Update the board statistics counters
4915  * @adapter: board private structure
4916  **/
4917 static void e1000e_update_stats(struct e1000_adapter *adapter)
4918 {
4919 	struct net_device *netdev = adapter->netdev;
4920 	struct e1000_hw *hw = &adapter->hw;
4921 	struct pci_dev *pdev = adapter->pdev;
4922 
4923 	/* Prevent stats update while adapter is being reset, or if the pci
4924 	 * connection is down.
4925 	 */
4926 	if (adapter->link_speed == 0)
4927 		return;
4928 	if (pci_channel_offline(pdev))
4929 		return;
4930 
4931 	adapter->stats.crcerrs += er32(CRCERRS);
4932 	adapter->stats.gprc += er32(GPRC);
4933 	adapter->stats.gorc += er32(GORCL);
4934 	er32(GORCH);		/* Clear gorc */
4935 	adapter->stats.bprc += er32(BPRC);
4936 	adapter->stats.mprc += er32(MPRC);
4937 	adapter->stats.roc += er32(ROC);
4938 
4939 	adapter->stats.mpc += er32(MPC);
4940 
4941 	/* Half-duplex statistics */
4942 	if (adapter->link_duplex == HALF_DUPLEX) {
4943 		if (adapter->flags2 & FLAG2_HAS_PHY_STATS) {
4944 			e1000e_update_phy_stats(adapter);
4945 		} else {
4946 			adapter->stats.scc += er32(SCC);
4947 			adapter->stats.ecol += er32(ECOL);
4948 			adapter->stats.mcc += er32(MCC);
4949 			adapter->stats.latecol += er32(LATECOL);
4950 			adapter->stats.dc += er32(DC);
4951 
4952 			hw->mac.collision_delta = er32(COLC);
4953 
4954 			if ((hw->mac.type != e1000_82574) &&
4955 			    (hw->mac.type != e1000_82583))
4956 				adapter->stats.tncrs += er32(TNCRS);
4957 		}
4958 		adapter->stats.colc += hw->mac.collision_delta;
4959 	}
4960 
4961 	adapter->stats.xonrxc += er32(XONRXC);
4962 	adapter->stats.xontxc += er32(XONTXC);
4963 	adapter->stats.xoffrxc += er32(XOFFRXC);
4964 	adapter->stats.xofftxc += er32(XOFFTXC);
4965 	adapter->stats.gptc += er32(GPTC);
4966 	adapter->stats.gotc += er32(GOTCL);
4967 	er32(GOTCH);		/* Clear gotc */
4968 	adapter->stats.rnbc += er32(RNBC);
4969 	adapter->stats.ruc += er32(RUC);
4970 
4971 	adapter->stats.mptc += er32(MPTC);
4972 	adapter->stats.bptc += er32(BPTC);
4973 
4974 	/* used for adaptive IFS */
4975 
4976 	hw->mac.tx_packet_delta = er32(TPT);
4977 	adapter->stats.tpt += hw->mac.tx_packet_delta;
4978 
4979 	adapter->stats.algnerrc += er32(ALGNERRC);
4980 	adapter->stats.rxerrc += er32(RXERRC);
4981 	adapter->stats.cexterr += er32(CEXTERR);
4982 	adapter->stats.tsctc += er32(TSCTC);
4983 	adapter->stats.tsctfc += er32(TSCTFC);
4984 
4985 	/* Fill out the OS statistics structure */
4986 	netdev->stats.multicast = adapter->stats.mprc;
4987 	netdev->stats.collisions = adapter->stats.colc;
4988 
4989 	/* Rx Errors */
4990 
4991 	/* RLEC on some newer hardware can be incorrect so build
4992 	 * our own version based on RUC and ROC
4993 	 */
4994 	netdev->stats.rx_errors = adapter->stats.rxerrc +
4995 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
4996 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
4997 	netdev->stats.rx_length_errors = adapter->stats.ruc +
4998 	    adapter->stats.roc;
4999 	netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
5000 	netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
5001 	netdev->stats.rx_missed_errors = adapter->stats.mpc;
5002 
5003 	/* Tx Errors */
5004 	netdev->stats.tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5005 	netdev->stats.tx_aborted_errors = adapter->stats.ecol;
5006 	netdev->stats.tx_window_errors = adapter->stats.latecol;
5007 	netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
5008 
5009 	/* Tx Dropped needs to be maintained elsewhere */
5010 
5011 	/* Management Stats */
5012 	adapter->stats.mgptc += er32(MGTPTC);
5013 	adapter->stats.mgprc += er32(MGTPRC);
5014 	adapter->stats.mgpdc += er32(MGTPDC);
5015 
5016 	/* Correctable ECC Errors */
5017 	if (hw->mac.type >= e1000_pch_lpt) {
5018 		u32 pbeccsts = er32(PBECCSTS);
5019 
5020 		adapter->corr_errors +=
5021 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
5022 		adapter->uncorr_errors +=
5023 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
5024 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
5025 	}
5026 }
5027 
5028 /**
5029  * e1000_phy_read_status - Update the PHY register status snapshot
5030  * @adapter: board private structure
5031  **/
5032 static void e1000_phy_read_status(struct e1000_adapter *adapter)
5033 {
5034 	struct e1000_hw *hw = &adapter->hw;
5035 	struct e1000_phy_regs *phy = &adapter->phy_regs;
5036 
5037 	if (!pm_runtime_suspended((&adapter->pdev->dev)->parent) &&
5038 	    (er32(STATUS) & E1000_STATUS_LU) &&
5039 	    (adapter->hw.phy.media_type == e1000_media_type_copper)) {
5040 		int ret_val;
5041 
5042 		ret_val = e1e_rphy(hw, MII_BMCR, &phy->bmcr);
5043 		ret_val |= e1e_rphy(hw, MII_BMSR, &phy->bmsr);
5044 		ret_val |= e1e_rphy(hw, MII_ADVERTISE, &phy->advertise);
5045 		ret_val |= e1e_rphy(hw, MII_LPA, &phy->lpa);
5046 		ret_val |= e1e_rphy(hw, MII_EXPANSION, &phy->expansion);
5047 		ret_val |= e1e_rphy(hw, MII_CTRL1000, &phy->ctrl1000);
5048 		ret_val |= e1e_rphy(hw, MII_STAT1000, &phy->stat1000);
5049 		ret_val |= e1e_rphy(hw, MII_ESTATUS, &phy->estatus);
5050 		if (ret_val)
5051 			e_warn("Error reading PHY register\n");
5052 	} else {
5053 		/* Do not read PHY registers if link is not up
5054 		 * Set values to typical power-on defaults
5055 		 */
5056 		phy->bmcr = (BMCR_SPEED1000 | BMCR_ANENABLE | BMCR_FULLDPLX);
5057 		phy->bmsr = (BMSR_100FULL | BMSR_100HALF | BMSR_10FULL |
5058 			     BMSR_10HALF | BMSR_ESTATEN | BMSR_ANEGCAPABLE |
5059 			     BMSR_ERCAP);
5060 		phy->advertise = (ADVERTISE_PAUSE_ASYM | ADVERTISE_PAUSE_CAP |
5061 				  ADVERTISE_ALL | ADVERTISE_CSMA);
5062 		phy->lpa = 0;
5063 		phy->expansion = EXPANSION_ENABLENPAGE;
5064 		phy->ctrl1000 = ADVERTISE_1000FULL;
5065 		phy->stat1000 = 0;
5066 		phy->estatus = (ESTATUS_1000_TFULL | ESTATUS_1000_THALF);
5067 	}
5068 }
5069 
5070 static void e1000_print_link_info(struct e1000_adapter *adapter)
5071 {
5072 	struct e1000_hw *hw = &adapter->hw;
5073 	u32 ctrl = er32(CTRL);
5074 
5075 	/* Link status message must follow this format for user tools */
5076 	pr_info("%s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5077 		adapter->netdev->name, adapter->link_speed,
5078 		adapter->link_duplex == FULL_DUPLEX ? "Full" : "Half",
5079 		(ctrl & E1000_CTRL_TFCE) && (ctrl & E1000_CTRL_RFCE) ? "Rx/Tx" :
5080 		(ctrl & E1000_CTRL_RFCE) ? "Rx" :
5081 		(ctrl & E1000_CTRL_TFCE) ? "Tx" : "None");
5082 }
5083 
5084 static bool e1000e_has_link(struct e1000_adapter *adapter)
5085 {
5086 	struct e1000_hw *hw = &adapter->hw;
5087 	bool link_active = false;
5088 	s32 ret_val = 0;
5089 
5090 	/* get_link_status is set on LSC (link status) interrupt or
5091 	 * Rx sequence error interrupt.  get_link_status will stay
5092 	 * true until the check_for_link establishes link
5093 	 * for copper adapters ONLY
5094 	 */
5095 	switch (hw->phy.media_type) {
5096 	case e1000_media_type_copper:
5097 		if (hw->mac.get_link_status) {
5098 			ret_val = hw->mac.ops.check_for_link(hw);
5099 			link_active = !hw->mac.get_link_status;
5100 		} else {
5101 			link_active = true;
5102 		}
5103 		break;
5104 	case e1000_media_type_fiber:
5105 		ret_val = hw->mac.ops.check_for_link(hw);
5106 		link_active = !!(er32(STATUS) & E1000_STATUS_LU);
5107 		break;
5108 	case e1000_media_type_internal_serdes:
5109 		ret_val = hw->mac.ops.check_for_link(hw);
5110 		link_active = hw->mac.serdes_has_link;
5111 		break;
5112 	default:
5113 	case e1000_media_type_unknown:
5114 		break;
5115 	}
5116 
5117 	if ((ret_val == -E1000_ERR_PHY) && (hw->phy.type == e1000_phy_igp_3) &&
5118 	    (er32(CTRL) & E1000_PHY_CTRL_GBE_DISABLE)) {
5119 		/* See e1000_kmrn_lock_loss_workaround_ich8lan() */
5120 		e_info("Gigabit has been disabled, downgrading speed\n");
5121 	}
5122 
5123 	return link_active;
5124 }
5125 
5126 static void e1000e_enable_receives(struct e1000_adapter *adapter)
5127 {
5128 	/* make sure the receive unit is started */
5129 	if ((adapter->flags & FLAG_RX_NEEDS_RESTART) &&
5130 	    (adapter->flags & FLAG_RESTART_NOW)) {
5131 		struct e1000_hw *hw = &adapter->hw;
5132 		u32 rctl = er32(RCTL);
5133 
5134 		ew32(RCTL, rctl | E1000_RCTL_EN);
5135 		adapter->flags &= ~FLAG_RESTART_NOW;
5136 	}
5137 }
5138 
5139 static void e1000e_check_82574_phy_workaround(struct e1000_adapter *adapter)
5140 {
5141 	struct e1000_hw *hw = &adapter->hw;
5142 
5143 	/* With 82574 controllers, PHY needs to be checked periodically
5144 	 * for hung state and reset, if two calls return true
5145 	 */
5146 	if (e1000_check_phy_82574(hw))
5147 		adapter->phy_hang_count++;
5148 	else
5149 		adapter->phy_hang_count = 0;
5150 
5151 	if (adapter->phy_hang_count > 1) {
5152 		adapter->phy_hang_count = 0;
5153 		e_dbg("PHY appears hung - resetting\n");
5154 		schedule_work(&adapter->reset_task);
5155 	}
5156 }
5157 
5158 static void e1000_watchdog_task(struct work_struct *work)
5159 {
5160 	struct e1000_adapter *adapter = container_of(work,
5161 						     struct e1000_adapter,
5162 						     watchdog_task.work);
5163 	struct net_device *netdev = adapter->netdev;
5164 	struct e1000_mac_info *mac = &adapter->hw.mac;
5165 	struct e1000_phy_info *phy = &adapter->hw.phy;
5166 	struct e1000_ring *tx_ring = adapter->tx_ring;
5167 	u32 dmoff_exit_timeout = 100, tries = 0;
5168 	struct e1000_hw *hw = &adapter->hw;
5169 	u32 link, tctl, pcim_state;
5170 
5171 	if (test_bit(__E1000_DOWN, &adapter->state))
5172 		return;
5173 
5174 	link = e1000e_has_link(adapter);
5175 	if ((netif_carrier_ok(netdev)) && link) {
5176 		/* Cancel scheduled suspend requests. */
5177 		pm_runtime_resume(netdev->dev.parent);
5178 
5179 		e1000e_enable_receives(adapter);
5180 		goto link_up;
5181 	}
5182 
5183 	if ((e1000e_enable_tx_pkt_filtering(hw)) &&
5184 	    (adapter->mng_vlan_id != adapter->hw.mng_cookie.vlan_id))
5185 		e1000_update_mng_vlan(adapter);
5186 
5187 	if (link) {
5188 		if (!netif_carrier_ok(netdev)) {
5189 			bool txb2b = true;
5190 
5191 			/* Cancel scheduled suspend requests. */
5192 			pm_runtime_resume(netdev->dev.parent);
5193 
5194 			/* Checking if MAC is in DMoff state*/
5195 			pcim_state = er32(STATUS);
5196 			while (pcim_state & E1000_STATUS_PCIM_STATE) {
5197 				if (tries++ == dmoff_exit_timeout) {
5198 					e_dbg("Error in exiting dmoff\n");
5199 					break;
5200 				}
5201 				usleep_range(10000, 20000);
5202 				pcim_state = er32(STATUS);
5203 
5204 				/* Checking if MAC exited DMoff state */
5205 				if (!(pcim_state & E1000_STATUS_PCIM_STATE))
5206 					e1000_phy_hw_reset(&adapter->hw);
5207 			}
5208 
5209 			/* update snapshot of PHY registers on LSC */
5210 			e1000_phy_read_status(adapter);
5211 			mac->ops.get_link_up_info(&adapter->hw,
5212 						  &adapter->link_speed,
5213 						  &adapter->link_duplex);
5214 			e1000_print_link_info(adapter);
5215 
5216 			/* check if SmartSpeed worked */
5217 			e1000e_check_downshift(hw);
5218 			if (phy->speed_downgraded)
5219 				netdev_warn(netdev,
5220 					    "Link Speed was downgraded by SmartSpeed\n");
5221 
5222 			/* On supported PHYs, check for duplex mismatch only
5223 			 * if link has autonegotiated at 10/100 half
5224 			 */
5225 			if ((hw->phy.type == e1000_phy_igp_3 ||
5226 			     hw->phy.type == e1000_phy_bm) &&
5227 			    hw->mac.autoneg &&
5228 			    (adapter->link_speed == SPEED_10 ||
5229 			     adapter->link_speed == SPEED_100) &&
5230 			    (adapter->link_duplex == HALF_DUPLEX)) {
5231 				u16 autoneg_exp;
5232 
5233 				e1e_rphy(hw, MII_EXPANSION, &autoneg_exp);
5234 
5235 				if (!(autoneg_exp & EXPANSION_NWAY))
5236 					e_info("Autonegotiated half duplex but link partner cannot autoneg.  Try forcing full duplex if link gets many collisions.\n");
5237 			}
5238 
5239 			/* adjust timeout factor according to speed/duplex */
5240 			adapter->tx_timeout_factor = 1;
5241 			switch (adapter->link_speed) {
5242 			case SPEED_10:
5243 				txb2b = false;
5244 				adapter->tx_timeout_factor = 16;
5245 				break;
5246 			case SPEED_100:
5247 				txb2b = false;
5248 				adapter->tx_timeout_factor = 10;
5249 				break;
5250 			}
5251 
5252 			/* workaround: re-program speed mode bit after
5253 			 * link-up event
5254 			 */
5255 			if ((adapter->flags & FLAG_TARC_SPEED_MODE_BIT) &&
5256 			    !txb2b) {
5257 				u32 tarc0;
5258 
5259 				tarc0 = er32(TARC(0));
5260 				tarc0 &= ~SPEED_MODE_BIT;
5261 				ew32(TARC(0), tarc0);
5262 			}
5263 
5264 			/* disable TSO for pcie and 10/100 speeds, to avoid
5265 			 * some hardware issues
5266 			 */
5267 			if (!(adapter->flags & FLAG_TSO_FORCE)) {
5268 				switch (adapter->link_speed) {
5269 				case SPEED_10:
5270 				case SPEED_100:
5271 					e_info("10/100 speed: disabling TSO\n");
5272 					netdev->features &= ~NETIF_F_TSO;
5273 					netdev->features &= ~NETIF_F_TSO6;
5274 					break;
5275 				case SPEED_1000:
5276 					netdev->features |= NETIF_F_TSO;
5277 					netdev->features |= NETIF_F_TSO6;
5278 					break;
5279 				default:
5280 					/* oops */
5281 					break;
5282 				}
5283 			}
5284 
5285 			/* enable transmits in the hardware, need to do this
5286 			 * after setting TARC(0)
5287 			 */
5288 			tctl = er32(TCTL);
5289 			tctl |= E1000_TCTL_EN;
5290 			ew32(TCTL, tctl);
5291 
5292 			/* Perform any post-link-up configuration before
5293 			 * reporting link up.
5294 			 */
5295 			if (phy->ops.cfg_on_link_up)
5296 				phy->ops.cfg_on_link_up(hw);
5297 
5298 			netif_wake_queue(netdev);
5299 			netif_carrier_on(netdev);
5300 
5301 			if (!test_bit(__E1000_DOWN, &adapter->state))
5302 				mod_timer(&adapter->phy_info_timer,
5303 					  round_jiffies(jiffies + 2 * HZ));
5304 		}
5305 	} else {
5306 		if (netif_carrier_ok(netdev)) {
5307 			adapter->link_speed = 0;
5308 			adapter->link_duplex = 0;
5309 			/* Link status message must follow this format */
5310 			pr_info("%s NIC Link is Down\n", adapter->netdev->name);
5311 			netif_carrier_off(netdev);
5312 			netif_stop_queue(netdev);
5313 			if (!test_bit(__E1000_DOWN, &adapter->state))
5314 				mod_timer(&adapter->phy_info_timer,
5315 					  round_jiffies(jiffies + 2 * HZ));
5316 
5317 			/* 8000ES2LAN requires a Rx packet buffer work-around
5318 			 * on link down event; reset the controller to flush
5319 			 * the Rx packet buffer.
5320 			 */
5321 			if (adapter->flags & FLAG_RX_NEEDS_RESTART)
5322 				adapter->flags |= FLAG_RESTART_NOW;
5323 			else
5324 				pm_schedule_suspend(netdev->dev.parent,
5325 						    LINK_TIMEOUT);
5326 		}
5327 	}
5328 
5329 link_up:
5330 	spin_lock(&adapter->stats64_lock);
5331 	e1000e_update_stats(adapter);
5332 
5333 	mac->tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
5334 	adapter->tpt_old = adapter->stats.tpt;
5335 	mac->collision_delta = adapter->stats.colc - adapter->colc_old;
5336 	adapter->colc_old = adapter->stats.colc;
5337 
5338 	adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
5339 	adapter->gorc_old = adapter->stats.gorc;
5340 	adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
5341 	adapter->gotc_old = adapter->stats.gotc;
5342 	spin_unlock(&adapter->stats64_lock);
5343 
5344 	/* If the link is lost the controller stops DMA, but
5345 	 * if there is queued Tx work it cannot be done.  So
5346 	 * reset the controller to flush the Tx packet buffers.
5347 	 */
5348 	if (!netif_carrier_ok(netdev) &&
5349 	    (e1000_desc_unused(tx_ring) + 1 < tx_ring->count))
5350 		adapter->flags |= FLAG_RESTART_NOW;
5351 
5352 	/* If reset is necessary, do it outside of interrupt context. */
5353 	if (adapter->flags & FLAG_RESTART_NOW) {
5354 		schedule_work(&adapter->reset_task);
5355 		/* return immediately since reset is imminent */
5356 		return;
5357 	}
5358 
5359 	e1000e_update_adaptive(&adapter->hw);
5360 
5361 	/* Simple mode for Interrupt Throttle Rate (ITR) */
5362 	if (adapter->itr_setting == 4) {
5363 		/* Symmetric Tx/Rx gets a reduced ITR=2000;
5364 		 * Total asymmetrical Tx or Rx gets ITR=8000;
5365 		 * everyone else is between 2000-8000.
5366 		 */
5367 		u32 goc = (adapter->gotc + adapter->gorc) / 10000;
5368 		u32 dif = (adapter->gotc > adapter->gorc ?
5369 			   adapter->gotc - adapter->gorc :
5370 			   adapter->gorc - adapter->gotc) / 10000;
5371 		u32 itr = goc > 0 ? (dif * 6000 / goc + 2000) : 8000;
5372 
5373 		e1000e_write_itr(adapter, itr);
5374 	}
5375 
5376 	/* Cause software interrupt to ensure Rx ring is cleaned */
5377 	if (adapter->msix_entries)
5378 		ew32(ICS, adapter->rx_ring->ims_val);
5379 	else
5380 		ew32(ICS, E1000_ICS_RXDMT0);
5381 
5382 	/* flush pending descriptors to memory before detecting Tx hang */
5383 	e1000e_flush_descriptors(adapter);
5384 
5385 	/* Force detection of hung controller every watchdog period */
5386 	adapter->detect_tx_hung = true;
5387 
5388 	/* With 82571 controllers, LAA may be overwritten due to controller
5389 	 * reset from the other port. Set the appropriate LAA in RAR[0]
5390 	 */
5391 	if (e1000e_get_laa_state_82571(hw))
5392 		hw->mac.ops.rar_set(hw, adapter->hw.mac.addr, 0);
5393 
5394 	if (adapter->flags2 & FLAG2_CHECK_PHY_HANG)
5395 		e1000e_check_82574_phy_workaround(adapter);
5396 
5397 	/* Clear valid timestamp stuck in RXSTMPL/H due to a Rx error */
5398 	if (adapter->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE) {
5399 		if ((adapter->flags2 & FLAG2_CHECK_RX_HWTSTAMP) &&
5400 		    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID)) {
5401 			er32(RXSTMPH);
5402 			adapter->rx_hwtstamp_cleared++;
5403 		} else {
5404 			adapter->flags2 |= FLAG2_CHECK_RX_HWTSTAMP;
5405 		}
5406 	}
5407 
5408 	/* Reset the timer */
5409 	if (!test_bit(__E1000_DOWN, &adapter->state))
5410 		queue_delayed_work(adapter->e1000_workqueue,
5411 				   &adapter->watchdog_task,
5412 				   round_jiffies(2 * HZ));
5413 }
5414 
5415 #define E1000_TX_FLAGS_CSUM		0x00000001
5416 #define E1000_TX_FLAGS_VLAN		0x00000002
5417 #define E1000_TX_FLAGS_TSO		0x00000004
5418 #define E1000_TX_FLAGS_IPV4		0x00000008
5419 #define E1000_TX_FLAGS_NO_FCS		0x00000010
5420 #define E1000_TX_FLAGS_HWTSTAMP		0x00000020
5421 #define E1000_TX_FLAGS_VLAN_MASK	0xffff0000
5422 #define E1000_TX_FLAGS_VLAN_SHIFT	16
5423 
5424 static int e1000_tso(struct e1000_ring *tx_ring, struct sk_buff *skb,
5425 		     __be16 protocol)
5426 {
5427 	struct e1000_context_desc *context_desc;
5428 	struct e1000_buffer *buffer_info;
5429 	unsigned int i;
5430 	u32 cmd_length = 0;
5431 	u16 ipcse = 0, mss;
5432 	u8 ipcss, ipcso, tucss, tucso, hdr_len;
5433 	int err;
5434 
5435 	if (!skb_is_gso(skb))
5436 		return 0;
5437 
5438 	err = skb_cow_head(skb, 0);
5439 	if (err < 0)
5440 		return err;
5441 
5442 	hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5443 	mss = skb_shinfo(skb)->gso_size;
5444 	if (protocol == htons(ETH_P_IP)) {
5445 		struct iphdr *iph = ip_hdr(skb);
5446 		iph->tot_len = 0;
5447 		iph->check = 0;
5448 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
5449 							 0, IPPROTO_TCP, 0);
5450 		cmd_length = E1000_TXD_CMD_IP;
5451 		ipcse = skb_transport_offset(skb) - 1;
5452 	} else if (skb_is_gso_v6(skb)) {
5453 		ipv6_hdr(skb)->payload_len = 0;
5454 		tcp_hdr(skb)->check = ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
5455 						       &ipv6_hdr(skb)->daddr,
5456 						       0, IPPROTO_TCP, 0);
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 
5492 static 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 
5546 static 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 
5625 dma_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 
5642 static 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
5709 static 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 
5746 static 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 
5769 static 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 
5778 static 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_transport_offset(skb) + tcp_hdrlen(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 *
5908 				     DIV_ROUND_UP(PAGE_SIZE,
5909 						  adapter->tx_fifo_limit) + 2));
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  **/
5932 static void e1000_tx_timeout(struct net_device *netdev)
5933 {
5934 	struct e1000_adapter *adapter = netdev_priv(netdev);
5935 
5936 	/* Do the reset outside of interrupt context */
5937 	adapter->tx_timeout_count++;
5938 	schedule_work(&adapter->reset_task);
5939 }
5940 
5941 static void e1000_reset_task(struct work_struct *work)
5942 {
5943 	struct e1000_adapter *adapter;
5944 	adapter = container_of(work, struct e1000_adapter, reset_task);
5945 
5946 	/* don't run the task if already down */
5947 	if (test_bit(__E1000_DOWN, &adapter->state))
5948 		return;
5949 
5950 	if (!(adapter->flags & FLAG_RESTART_NOW)) {
5951 		e1000e_dump(adapter);
5952 		e_err("Reset adapter unexpectedly\n");
5953 	}
5954 	e1000e_reinit_locked(adapter);
5955 }
5956 
5957 /**
5958  * e1000_get_stats64 - Get System Network Statistics
5959  * @netdev: network interface device structure
5960  * @stats: rtnl_link_stats64 pointer
5961  *
5962  * Returns the address of the device statistics structure.
5963  **/
5964 void e1000e_get_stats64(struct net_device *netdev,
5965 			struct rtnl_link_stats64 *stats)
5966 {
5967 	struct e1000_adapter *adapter = netdev_priv(netdev);
5968 
5969 	spin_lock(&adapter->stats64_lock);
5970 	e1000e_update_stats(adapter);
5971 	/* Fill out the OS statistics structure */
5972 	stats->rx_bytes = adapter->stats.gorc;
5973 	stats->rx_packets = adapter->stats.gprc;
5974 	stats->tx_bytes = adapter->stats.gotc;
5975 	stats->tx_packets = adapter->stats.gptc;
5976 	stats->multicast = adapter->stats.mprc;
5977 	stats->collisions = adapter->stats.colc;
5978 
5979 	/* Rx Errors */
5980 
5981 	/* RLEC on some newer hardware can be incorrect so build
5982 	 * our own version based on RUC and ROC
5983 	 */
5984 	stats->rx_errors = adapter->stats.rxerrc +
5985 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
5986 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5987 	stats->rx_length_errors = adapter->stats.ruc + adapter->stats.roc;
5988 	stats->rx_crc_errors = adapter->stats.crcerrs;
5989 	stats->rx_frame_errors = adapter->stats.algnerrc;
5990 	stats->rx_missed_errors = adapter->stats.mpc;
5991 
5992 	/* Tx Errors */
5993 	stats->tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5994 	stats->tx_aborted_errors = adapter->stats.ecol;
5995 	stats->tx_window_errors = adapter->stats.latecol;
5996 	stats->tx_carrier_errors = adapter->stats.tncrs;
5997 
5998 	/* Tx Dropped needs to be maintained elsewhere */
5999 
6000 	spin_unlock(&adapter->stats64_lock);
6001 }
6002 
6003 /**
6004  * e1000_change_mtu - Change the Maximum Transfer Unit
6005  * @netdev: network interface device structure
6006  * @new_mtu: new value for maximum frame size
6007  *
6008  * Returns 0 on success, negative on failure
6009  **/
6010 static int e1000_change_mtu(struct net_device *netdev, int new_mtu)
6011 {
6012 	struct e1000_adapter *adapter = netdev_priv(netdev);
6013 	int max_frame = new_mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
6014 
6015 	/* Jumbo frame support */
6016 	if ((new_mtu > ETH_DATA_LEN) &&
6017 	    !(adapter->flags & FLAG_HAS_JUMBO_FRAMES)) {
6018 		e_err("Jumbo Frames not supported.\n");
6019 		return -EINVAL;
6020 	}
6021 
6022 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
6023 	if ((adapter->hw.mac.type >= e1000_pch2lan) &&
6024 	    !(adapter->flags2 & FLAG2_CRC_STRIPPING) &&
6025 	    (new_mtu > ETH_DATA_LEN)) {
6026 		e_err("Jumbo Frames not supported on this device when CRC stripping is disabled.\n");
6027 		return -EINVAL;
6028 	}
6029 
6030 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
6031 		usleep_range(1000, 1100);
6032 	/* e1000e_down -> e1000e_reset dependent on max_frame_size & mtu */
6033 	adapter->max_frame_size = max_frame;
6034 	netdev_dbg(netdev, "changing MTU from %d to %d\n",
6035 		   netdev->mtu, new_mtu);
6036 	netdev->mtu = new_mtu;
6037 
6038 	pm_runtime_get_sync(netdev->dev.parent);
6039 
6040 	if (netif_running(netdev))
6041 		e1000e_down(adapter, true);
6042 
6043 	/* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
6044 	 * means we reserve 2 more, this pushes us to allocate from the next
6045 	 * larger slab size.
6046 	 * i.e. RXBUFFER_2048 --> size-4096 slab
6047 	 * However with the new *_jumbo_rx* routines, jumbo receives will use
6048 	 * fragmented skbs
6049 	 */
6050 
6051 	if (max_frame <= 2048)
6052 		adapter->rx_buffer_len = 2048;
6053 	else
6054 		adapter->rx_buffer_len = 4096;
6055 
6056 	/* adjust allocation if LPE protects us, and we aren't using SBP */
6057 	if (max_frame <= (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN))
6058 		adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
6059 
6060 	if (netif_running(netdev))
6061 		e1000e_up(adapter);
6062 	else
6063 		e1000e_reset(adapter);
6064 
6065 	pm_runtime_put_sync(netdev->dev.parent);
6066 
6067 	clear_bit(__E1000_RESETTING, &adapter->state);
6068 
6069 	return 0;
6070 }
6071 
6072 static int e1000_mii_ioctl(struct net_device *netdev, struct ifreq *ifr,
6073 			   int cmd)
6074 {
6075 	struct e1000_adapter *adapter = netdev_priv(netdev);
6076 	struct mii_ioctl_data *data = if_mii(ifr);
6077 
6078 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
6079 		return -EOPNOTSUPP;
6080 
6081 	switch (cmd) {
6082 	case SIOCGMIIPHY:
6083 		data->phy_id = adapter->hw.phy.addr;
6084 		break;
6085 	case SIOCGMIIREG:
6086 		e1000_phy_read_status(adapter);
6087 
6088 		switch (data->reg_num & 0x1F) {
6089 		case MII_BMCR:
6090 			data->val_out = adapter->phy_regs.bmcr;
6091 			break;
6092 		case MII_BMSR:
6093 			data->val_out = adapter->phy_regs.bmsr;
6094 			break;
6095 		case MII_PHYSID1:
6096 			data->val_out = (adapter->hw.phy.id >> 16);
6097 			break;
6098 		case MII_PHYSID2:
6099 			data->val_out = (adapter->hw.phy.id & 0xFFFF);
6100 			break;
6101 		case MII_ADVERTISE:
6102 			data->val_out = adapter->phy_regs.advertise;
6103 			break;
6104 		case MII_LPA:
6105 			data->val_out = adapter->phy_regs.lpa;
6106 			break;
6107 		case MII_EXPANSION:
6108 			data->val_out = adapter->phy_regs.expansion;
6109 			break;
6110 		case MII_CTRL1000:
6111 			data->val_out = adapter->phy_regs.ctrl1000;
6112 			break;
6113 		case MII_STAT1000:
6114 			data->val_out = adapter->phy_regs.stat1000;
6115 			break;
6116 		case MII_ESTATUS:
6117 			data->val_out = adapter->phy_regs.estatus;
6118 			break;
6119 		default:
6120 			return -EIO;
6121 		}
6122 		break;
6123 	case SIOCSMIIREG:
6124 	default:
6125 		return -EOPNOTSUPP;
6126 	}
6127 	return 0;
6128 }
6129 
6130 /**
6131  * e1000e_hwtstamp_ioctl - control hardware time stamping
6132  * @netdev: network interface device structure
6133  * @ifreq: interface request
6134  *
6135  * Outgoing time stamping can be enabled and disabled. Play nice and
6136  * disable it when requested, although it shouldn't cause any overhead
6137  * when no packet needs it. At most one packet in the queue may be
6138  * marked for time stamping, otherwise it would be impossible to tell
6139  * for sure to which packet the hardware time stamp belongs.
6140  *
6141  * Incoming time stamping has to be configured via the hardware filters.
6142  * Not all combinations are supported, in particular event type has to be
6143  * specified. Matching the kind of event packet is not supported, with the
6144  * exception of "all V2 events regardless of level 2 or 4".
6145  **/
6146 static int e1000e_hwtstamp_set(struct net_device *netdev, struct ifreq *ifr)
6147 {
6148 	struct e1000_adapter *adapter = netdev_priv(netdev);
6149 	struct hwtstamp_config config;
6150 	int ret_val;
6151 
6152 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
6153 		return -EFAULT;
6154 
6155 	ret_val = e1000e_config_hwtstamp(adapter, &config);
6156 	if (ret_val)
6157 		return ret_val;
6158 
6159 	switch (config.rx_filter) {
6160 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
6161 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
6162 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
6163 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
6164 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
6165 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
6166 		/* With V2 type filters which specify a Sync or Delay Request,
6167 		 * Path Delay Request/Response messages are also time stamped
6168 		 * by hardware so notify the caller the requested packets plus
6169 		 * some others are time stamped.
6170 		 */
6171 		config.rx_filter = HWTSTAMP_FILTER_SOME;
6172 		break;
6173 	default:
6174 		break;
6175 	}
6176 
6177 	return copy_to_user(ifr->ifr_data, &config,
6178 			    sizeof(config)) ? -EFAULT : 0;
6179 }
6180 
6181 static int e1000e_hwtstamp_get(struct net_device *netdev, struct ifreq *ifr)
6182 {
6183 	struct e1000_adapter *adapter = netdev_priv(netdev);
6184 
6185 	return copy_to_user(ifr->ifr_data, &adapter->hwtstamp_config,
6186 			    sizeof(adapter->hwtstamp_config)) ? -EFAULT : 0;
6187 }
6188 
6189 static int e1000_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
6190 {
6191 	switch (cmd) {
6192 	case SIOCGMIIPHY:
6193 	case SIOCGMIIREG:
6194 	case SIOCSMIIREG:
6195 		return e1000_mii_ioctl(netdev, ifr, cmd);
6196 	case SIOCSHWTSTAMP:
6197 		return e1000e_hwtstamp_set(netdev, ifr);
6198 	case SIOCGHWTSTAMP:
6199 		return e1000e_hwtstamp_get(netdev, ifr);
6200 	default:
6201 		return -EOPNOTSUPP;
6202 	}
6203 }
6204 
6205 static int e1000_init_phy_wakeup(struct e1000_adapter *adapter, u32 wufc)
6206 {
6207 	struct e1000_hw *hw = &adapter->hw;
6208 	u32 i, mac_reg, wuc;
6209 	u16 phy_reg, wuc_enable;
6210 	int retval;
6211 
6212 	/* copy MAC RARs to PHY RARs */
6213 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
6214 
6215 	retval = hw->phy.ops.acquire(hw);
6216 	if (retval) {
6217 		e_err("Could not acquire PHY\n");
6218 		return retval;
6219 	}
6220 
6221 	/* Enable access to wakeup registers on and set page to BM_WUC_PAGE */
6222 	retval = e1000_enable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6223 	if (retval)
6224 		goto release;
6225 
6226 	/* copy MAC MTA to PHY MTA - only needed for pchlan */
6227 	for (i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
6228 		mac_reg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
6229 		hw->phy.ops.write_reg_page(hw, BM_MTA(i),
6230 					   (u16)(mac_reg & 0xFFFF));
6231 		hw->phy.ops.write_reg_page(hw, BM_MTA(i) + 1,
6232 					   (u16)((mac_reg >> 16) & 0xFFFF));
6233 	}
6234 
6235 	/* configure PHY Rx Control register */
6236 	hw->phy.ops.read_reg_page(&adapter->hw, BM_RCTL, &phy_reg);
6237 	mac_reg = er32(RCTL);
6238 	if (mac_reg & E1000_RCTL_UPE)
6239 		phy_reg |= BM_RCTL_UPE;
6240 	if (mac_reg & E1000_RCTL_MPE)
6241 		phy_reg |= BM_RCTL_MPE;
6242 	phy_reg &= ~(BM_RCTL_MO_MASK);
6243 	if (mac_reg & E1000_RCTL_MO_3)
6244 		phy_reg |= (((mac_reg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
6245 			    << BM_RCTL_MO_SHIFT);
6246 	if (mac_reg & E1000_RCTL_BAM)
6247 		phy_reg |= BM_RCTL_BAM;
6248 	if (mac_reg & E1000_RCTL_PMCF)
6249 		phy_reg |= BM_RCTL_PMCF;
6250 	mac_reg = er32(CTRL);
6251 	if (mac_reg & E1000_CTRL_RFCE)
6252 		phy_reg |= BM_RCTL_RFCE;
6253 	hw->phy.ops.write_reg_page(&adapter->hw, BM_RCTL, phy_reg);
6254 
6255 	wuc = E1000_WUC_PME_EN;
6256 	if (wufc & (E1000_WUFC_MAG | E1000_WUFC_LNKC))
6257 		wuc |= E1000_WUC_APME;
6258 
6259 	/* enable PHY wakeup in MAC register */
6260 	ew32(WUFC, wufc);
6261 	ew32(WUC, (E1000_WUC_PHY_WAKE | E1000_WUC_APMPME |
6262 		   E1000_WUC_PME_STATUS | wuc));
6263 
6264 	/* configure and enable PHY wakeup in PHY registers */
6265 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUFC, wufc);
6266 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUC, wuc);
6267 
6268 	/* activate PHY wakeup */
6269 	wuc_enable |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
6270 	retval = e1000_disable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6271 	if (retval)
6272 		e_err("Could not set PHY Host Wakeup bit\n");
6273 release:
6274 	hw->phy.ops.release(hw);
6275 
6276 	return retval;
6277 }
6278 
6279 static void e1000e_flush_lpic(struct pci_dev *pdev)
6280 {
6281 	struct net_device *netdev = pci_get_drvdata(pdev);
6282 	struct e1000_adapter *adapter = netdev_priv(netdev);
6283 	struct e1000_hw *hw = &adapter->hw;
6284 	u32 ret_val;
6285 
6286 	pm_runtime_get_sync(netdev->dev.parent);
6287 
6288 	ret_val = hw->phy.ops.acquire(hw);
6289 	if (ret_val)
6290 		goto fl_out;
6291 
6292 	pr_info("EEE TX LPI TIMER: %08X\n",
6293 		er32(LPIC) >> E1000_LPIC_LPIET_SHIFT);
6294 
6295 	hw->phy.ops.release(hw);
6296 
6297 fl_out:
6298 	pm_runtime_put_sync(netdev->dev.parent);
6299 }
6300 
6301 #ifdef CONFIG_PM_SLEEP
6302 /* S0ix implementation */
6303 static void e1000e_s0ix_entry_flow(struct e1000_adapter *adapter)
6304 {
6305 	struct e1000_hw *hw = &adapter->hw;
6306 	u32 mac_data;
6307 	u16 phy_data;
6308 
6309 	/* Disable the periodic inband message,
6310 	 * don't request PCIe clock in K1 page770_17[10:9] = 10b
6311 	 */
6312 	e1e_rphy(hw, HV_PM_CTRL, &phy_data);
6313 	phy_data &= ~HV_PM_CTRL_K1_CLK_REQ;
6314 	phy_data |= BIT(10);
6315 	e1e_wphy(hw, HV_PM_CTRL, phy_data);
6316 
6317 	/* Make sure we don't exit K1 every time a new packet arrives
6318 	 * 772_29[5] = 1 CS_Mode_Stay_In_K1
6319 	 */
6320 	e1e_rphy(hw, I217_CGFREG, &phy_data);
6321 	phy_data |= BIT(5);
6322 	e1e_wphy(hw, I217_CGFREG, phy_data);
6323 
6324 	/* Change the MAC/PHY interface to SMBus
6325 	 * Force the SMBus in PHY page769_23[0] = 1
6326 	 * Force the SMBus in MAC CTRL_EXT[11] = 1
6327 	 */
6328 	e1e_rphy(hw, CV_SMB_CTRL, &phy_data);
6329 	phy_data |= CV_SMB_CTRL_FORCE_SMBUS;
6330 	e1e_wphy(hw, CV_SMB_CTRL, phy_data);
6331 	mac_data = er32(CTRL_EXT);
6332 	mac_data |= E1000_CTRL_EXT_FORCE_SMBUS;
6333 	ew32(CTRL_EXT, mac_data);
6334 
6335 	/* DFT control: PHY bit: page769_20[0] = 1
6336 	 * Gate PPW via EXTCNF_CTRL - set 0x0F00[7] = 1
6337 	 */
6338 	e1e_rphy(hw, I82579_DFT_CTRL, &phy_data);
6339 	phy_data |= BIT(0);
6340 	e1e_wphy(hw, I82579_DFT_CTRL, phy_data);
6341 
6342 	mac_data = er32(EXTCNF_CTRL);
6343 	mac_data |= E1000_EXTCNF_CTRL_GATE_PHY_CFG;
6344 	ew32(EXTCNF_CTRL, mac_data);
6345 
6346 	/* Check MAC Tx/Rx packet buffer pointers.
6347 	 * Reset MAC Tx/Rx packet buffer pointers to suppress any
6348 	 * pending traffic indication that would prevent power gating.
6349 	 */
6350 	mac_data = er32(TDFH);
6351 	if (mac_data)
6352 		ew32(TDFH, 0);
6353 	mac_data = er32(TDFT);
6354 	if (mac_data)
6355 		ew32(TDFT, 0);
6356 	mac_data = er32(TDFHS);
6357 	if (mac_data)
6358 		ew32(TDFHS, 0);
6359 	mac_data = er32(TDFTS);
6360 	if (mac_data)
6361 		ew32(TDFTS, 0);
6362 	mac_data = er32(TDFPC);
6363 	if (mac_data)
6364 		ew32(TDFPC, 0);
6365 	mac_data = er32(RDFH);
6366 	if (mac_data)
6367 		ew32(RDFH, 0);
6368 	mac_data = er32(RDFT);
6369 	if (mac_data)
6370 		ew32(RDFT, 0);
6371 	mac_data = er32(RDFHS);
6372 	if (mac_data)
6373 		ew32(RDFHS, 0);
6374 	mac_data = er32(RDFTS);
6375 	if (mac_data)
6376 		ew32(RDFTS, 0);
6377 	mac_data = er32(RDFPC);
6378 	if (mac_data)
6379 		ew32(RDFPC, 0);
6380 
6381 	/* Enable the Dynamic Power Gating in the MAC */
6382 	mac_data = er32(FEXTNVM7);
6383 	mac_data |= BIT(22);
6384 	ew32(FEXTNVM7, mac_data);
6385 
6386 	/* Disable the time synchronization clock */
6387 	mac_data = er32(FEXTNVM7);
6388 	mac_data |= BIT(31);
6389 	mac_data &= ~BIT(0);
6390 	ew32(FEXTNVM7, mac_data);
6391 
6392 	/* Dynamic Power Gating Enable */
6393 	mac_data = er32(CTRL_EXT);
6394 	mac_data |= BIT(3);
6395 	ew32(CTRL_EXT, mac_data);
6396 
6397 	/* Enable the Dynamic Clock Gating in the DMA and MAC */
6398 	mac_data = er32(CTRL_EXT);
6399 	mac_data |= E1000_CTRL_EXT_DMA_DYN_CLK_EN;
6400 	ew32(CTRL_EXT, mac_data);
6401 
6402 	/* No MAC DPG gating SLP_S0 in modern standby
6403 	 * Switch the logic of the lanphypc to use PMC counter
6404 	 */
6405 	mac_data = er32(FEXTNVM5);
6406 	mac_data |= BIT(7);
6407 	ew32(FEXTNVM5, mac_data);
6408 }
6409 
6410 static void e1000e_s0ix_exit_flow(struct e1000_adapter *adapter)
6411 {
6412 	struct e1000_hw *hw = &adapter->hw;
6413 	u32 mac_data;
6414 	u16 phy_data;
6415 
6416 	/* Disable the Dynamic Power Gating in the MAC */
6417 	mac_data = er32(FEXTNVM7);
6418 	mac_data &= 0xFFBFFFFF;
6419 	ew32(FEXTNVM7, mac_data);
6420 
6421 	/* Enable the time synchronization clock */
6422 	mac_data = er32(FEXTNVM7);
6423 	mac_data |= BIT(0);
6424 	ew32(FEXTNVM7, mac_data);
6425 
6426 	/* Disable Dynamic Power Gating */
6427 	mac_data = er32(CTRL_EXT);
6428 	mac_data &= 0xFFFFFFF7;
6429 	ew32(CTRL_EXT, mac_data);
6430 
6431 	/* Disable the Dynamic Clock Gating in the DMA and MAC */
6432 	mac_data = er32(CTRL_EXT);
6433 	mac_data &= 0xFFF7FFFF;
6434 	ew32(CTRL_EXT, mac_data);
6435 
6436 	/* Revert the lanphypc logic to use the internal Gbe counter
6437 	 * and not the PMC counter
6438 	 */
6439 	mac_data = er32(FEXTNVM5);
6440 	mac_data &= 0xFFFFFF7F;
6441 	ew32(FEXTNVM5, mac_data);
6442 
6443 	/* Enable the periodic inband message,
6444 	 * Request PCIe clock in K1 page770_17[10:9] =01b
6445 	 */
6446 	e1e_rphy(hw, HV_PM_CTRL, &phy_data);
6447 	phy_data &= 0xFBFF;
6448 	phy_data |= HV_PM_CTRL_K1_CLK_REQ;
6449 	e1e_wphy(hw, HV_PM_CTRL, phy_data);
6450 
6451 	/* Return back configuration
6452 	 * 772_29[5] = 0 CS_Mode_Stay_In_K1
6453 	 */
6454 	e1e_rphy(hw, I217_CGFREG, &phy_data);
6455 	phy_data &= 0xFFDF;
6456 	e1e_wphy(hw, I217_CGFREG, phy_data);
6457 
6458 	/* Change the MAC/PHY interface to Kumeran
6459 	 * Unforce the SMBus in PHY page769_23[0] = 0
6460 	 * Unforce the SMBus in MAC CTRL_EXT[11] = 0
6461 	 */
6462 	e1e_rphy(hw, CV_SMB_CTRL, &phy_data);
6463 	phy_data &= ~CV_SMB_CTRL_FORCE_SMBUS;
6464 	e1e_wphy(hw, CV_SMB_CTRL, phy_data);
6465 	mac_data = er32(CTRL_EXT);
6466 	mac_data &= ~E1000_CTRL_EXT_FORCE_SMBUS;
6467 	ew32(CTRL_EXT, mac_data);
6468 }
6469 #endif /* CONFIG_PM_SLEEP */
6470 
6471 static int e1000e_pm_freeze(struct device *dev)
6472 {
6473 	struct net_device *netdev = dev_get_drvdata(dev);
6474 	struct e1000_adapter *adapter = netdev_priv(netdev);
6475 	bool present;
6476 
6477 	rtnl_lock();
6478 
6479 	present = netif_device_present(netdev);
6480 	netif_device_detach(netdev);
6481 
6482 	if (present && netif_running(netdev)) {
6483 		int count = E1000_CHECK_RESET_COUNT;
6484 
6485 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6486 			usleep_range(10000, 11000);
6487 
6488 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6489 
6490 		/* Quiesce the device without resetting the hardware */
6491 		e1000e_down(adapter, false);
6492 		e1000_free_irq(adapter);
6493 	}
6494 	rtnl_unlock();
6495 
6496 	e1000e_reset_interrupt_capability(adapter);
6497 
6498 	/* Allow time for pending master requests to run */
6499 	e1000e_disable_pcie_master(&adapter->hw);
6500 
6501 	return 0;
6502 }
6503 
6504 static int __e1000_shutdown(struct pci_dev *pdev, bool runtime)
6505 {
6506 	struct net_device *netdev = pci_get_drvdata(pdev);
6507 	struct e1000_adapter *adapter = netdev_priv(netdev);
6508 	struct e1000_hw *hw = &adapter->hw;
6509 	u32 ctrl, ctrl_ext, rctl, status;
6510 	/* Runtime suspend should only enable wakeup for link changes */
6511 	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
6512 	int retval = 0;
6513 
6514 	status = er32(STATUS);
6515 	if (status & E1000_STATUS_LU)
6516 		wufc &= ~E1000_WUFC_LNKC;
6517 
6518 	if (wufc) {
6519 		e1000_setup_rctl(adapter);
6520 		e1000e_set_rx_mode(netdev);
6521 
6522 		/* turn on all-multi mode if wake on multicast is enabled */
6523 		if (wufc & E1000_WUFC_MC) {
6524 			rctl = er32(RCTL);
6525 			rctl |= E1000_RCTL_MPE;
6526 			ew32(RCTL, rctl);
6527 		}
6528 
6529 		ctrl = er32(CTRL);
6530 		ctrl |= E1000_CTRL_ADVD3WUC;
6531 		if (!(adapter->flags2 & FLAG2_HAS_PHY_WAKEUP))
6532 			ctrl |= E1000_CTRL_EN_PHY_PWR_MGMT;
6533 		ew32(CTRL, ctrl);
6534 
6535 		if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
6536 		    adapter->hw.phy.media_type ==
6537 		    e1000_media_type_internal_serdes) {
6538 			/* keep the laser running in D3 */
6539 			ctrl_ext = er32(CTRL_EXT);
6540 			ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
6541 			ew32(CTRL_EXT, ctrl_ext);
6542 		}
6543 
6544 		if (!runtime)
6545 			e1000e_power_up_phy(adapter);
6546 
6547 		if (adapter->flags & FLAG_IS_ICH)
6548 			e1000_suspend_workarounds_ich8lan(&adapter->hw);
6549 
6550 		if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6551 			/* enable wakeup by the PHY */
6552 			retval = e1000_init_phy_wakeup(adapter, wufc);
6553 			if (retval)
6554 				return retval;
6555 		} else {
6556 			/* enable wakeup by the MAC */
6557 			ew32(WUFC, wufc);
6558 			ew32(WUC, E1000_WUC_PME_EN);
6559 		}
6560 	} else {
6561 		ew32(WUC, 0);
6562 		ew32(WUFC, 0);
6563 
6564 		e1000_power_down_phy(adapter);
6565 	}
6566 
6567 	if (adapter->hw.phy.type == e1000_phy_igp_3) {
6568 		e1000e_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
6569 	} else if (hw->mac.type >= e1000_pch_lpt) {
6570 		if (!(wufc & (E1000_WUFC_EX | E1000_WUFC_MC | E1000_WUFC_BC)))
6571 			/* ULP does not support wake from unicast, multicast
6572 			 * or broadcast.
6573 			 */
6574 			retval = e1000_enable_ulp_lpt_lp(hw, !runtime);
6575 
6576 		if (retval)
6577 			return retval;
6578 	}
6579 
6580 	/* Ensure that the appropriate bits are set in LPI_CTRL
6581 	 * for EEE in Sx
6582 	 */
6583 	if ((hw->phy.type >= e1000_phy_i217) &&
6584 	    adapter->eee_advert && hw->dev_spec.ich8lan.eee_lp_ability) {
6585 		u16 lpi_ctrl = 0;
6586 
6587 		retval = hw->phy.ops.acquire(hw);
6588 		if (!retval) {
6589 			retval = e1e_rphy_locked(hw, I82579_LPI_CTRL,
6590 						 &lpi_ctrl);
6591 			if (!retval) {
6592 				if (adapter->eee_advert &
6593 				    hw->dev_spec.ich8lan.eee_lp_ability &
6594 				    I82579_EEE_100_SUPPORTED)
6595 					lpi_ctrl |= I82579_LPI_CTRL_100_ENABLE;
6596 				if (adapter->eee_advert &
6597 				    hw->dev_spec.ich8lan.eee_lp_ability &
6598 				    I82579_EEE_1000_SUPPORTED)
6599 					lpi_ctrl |= I82579_LPI_CTRL_1000_ENABLE;
6600 
6601 				retval = e1e_wphy_locked(hw, I82579_LPI_CTRL,
6602 							 lpi_ctrl);
6603 			}
6604 		}
6605 		hw->phy.ops.release(hw);
6606 	}
6607 
6608 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
6609 	 * would have already happened in close and is redundant.
6610 	 */
6611 	e1000e_release_hw_control(adapter);
6612 
6613 	pci_clear_master(pdev);
6614 
6615 	/* The pci-e switch on some quad port adapters will report a
6616 	 * correctable error when the MAC transitions from D0 to D3.  To
6617 	 * prevent this we need to mask off the correctable errors on the
6618 	 * downstream port of the pci-e switch.
6619 	 *
6620 	 * We don't have the associated upstream bridge while assigning
6621 	 * the PCI device into guest. For example, the KVM on power is
6622 	 * one of the cases.
6623 	 */
6624 	if (adapter->flags & FLAG_IS_QUAD_PORT) {
6625 		struct pci_dev *us_dev = pdev->bus->self;
6626 		u16 devctl;
6627 
6628 		if (!us_dev)
6629 			return 0;
6630 
6631 		pcie_capability_read_word(us_dev, PCI_EXP_DEVCTL, &devctl);
6632 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL,
6633 					   (devctl & ~PCI_EXP_DEVCTL_CERE));
6634 
6635 		pci_save_state(pdev);
6636 		pci_prepare_to_sleep(pdev);
6637 
6638 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL, devctl);
6639 	}
6640 
6641 	return 0;
6642 }
6643 
6644 /**
6645  * __e1000e_disable_aspm - Disable ASPM states
6646  * @pdev: pointer to PCI device struct
6647  * @state: bit-mask of ASPM states to disable
6648  * @locked: indication if this context holds pci_bus_sem locked.
6649  *
6650  * Some devices *must* have certain ASPM states disabled per hardware errata.
6651  **/
6652 static void __e1000e_disable_aspm(struct pci_dev *pdev, u16 state, int locked)
6653 {
6654 	struct pci_dev *parent = pdev->bus->self;
6655 	u16 aspm_dis_mask = 0;
6656 	u16 pdev_aspmc, parent_aspmc;
6657 
6658 	switch (state) {
6659 	case PCIE_LINK_STATE_L0S:
6660 	case PCIE_LINK_STATE_L0S | PCIE_LINK_STATE_L1:
6661 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L0S;
6662 		/* fall-through - can't have L1 without L0s */
6663 	case PCIE_LINK_STATE_L1:
6664 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L1;
6665 		break;
6666 	default:
6667 		return;
6668 	}
6669 
6670 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6671 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6672 
6673 	if (parent) {
6674 		pcie_capability_read_word(parent, PCI_EXP_LNKCTL,
6675 					  &parent_aspmc);
6676 		parent_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6677 	}
6678 
6679 	/* Nothing to do if the ASPM states to be disabled already are */
6680 	if (!(pdev_aspmc & aspm_dis_mask) &&
6681 	    (!parent || !(parent_aspmc & aspm_dis_mask)))
6682 		return;
6683 
6684 	dev_info(&pdev->dev, "Disabling ASPM %s %s\n",
6685 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L0S) ?
6686 		 "L0s" : "",
6687 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L1) ?
6688 		 "L1" : "");
6689 
6690 #ifdef CONFIG_PCIEASPM
6691 	if (locked)
6692 		pci_disable_link_state_locked(pdev, state);
6693 	else
6694 		pci_disable_link_state(pdev, state);
6695 
6696 	/* Double-check ASPM control.  If not disabled by the above, the
6697 	 * BIOS is preventing that from happening (or CONFIG_PCIEASPM is
6698 	 * not enabled); override by writing PCI config space directly.
6699 	 */
6700 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6701 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6702 
6703 	if (!(aspm_dis_mask & pdev_aspmc))
6704 		return;
6705 #endif
6706 
6707 	/* Both device and parent should have the same ASPM setting.
6708 	 * Disable ASPM in downstream component first and then upstream.
6709 	 */
6710 	pcie_capability_clear_word(pdev, PCI_EXP_LNKCTL, aspm_dis_mask);
6711 
6712 	if (parent)
6713 		pcie_capability_clear_word(parent, PCI_EXP_LNKCTL,
6714 					   aspm_dis_mask);
6715 }
6716 
6717 /**
6718  * e1000e_disable_aspm - Disable ASPM states.
6719  * @pdev: pointer to PCI device struct
6720  * @state: bit-mask of ASPM states to disable
6721  *
6722  * This function acquires the pci_bus_sem!
6723  * Some devices *must* have certain ASPM states disabled per hardware errata.
6724  **/
6725 static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6726 {
6727 	__e1000e_disable_aspm(pdev, state, 0);
6728 }
6729 
6730 /**
6731  * e1000e_disable_aspm_locked   Disable ASPM states.
6732  * @pdev: pointer to PCI device struct
6733  * @state: bit-mask of ASPM states to disable
6734  *
6735  * This function must be called with pci_bus_sem acquired!
6736  * Some devices *must* have certain ASPM states disabled per hardware errata.
6737  **/
6738 static void e1000e_disable_aspm_locked(struct pci_dev *pdev, u16 state)
6739 {
6740 	__e1000e_disable_aspm(pdev, state, 1);
6741 }
6742 
6743 static int e1000e_pm_thaw(struct device *dev)
6744 {
6745 	struct net_device *netdev = dev_get_drvdata(dev);
6746 	struct e1000_adapter *adapter = netdev_priv(netdev);
6747 	int rc = 0;
6748 
6749 	e1000e_set_interrupt_capability(adapter);
6750 
6751 	rtnl_lock();
6752 	if (netif_running(netdev)) {
6753 		rc = e1000_request_irq(adapter);
6754 		if (rc)
6755 			goto err_irq;
6756 
6757 		e1000e_up(adapter);
6758 	}
6759 
6760 	netif_device_attach(netdev);
6761 err_irq:
6762 	rtnl_unlock();
6763 
6764 	return rc;
6765 }
6766 
6767 #ifdef CONFIG_PM
6768 static int __e1000_resume(struct pci_dev *pdev)
6769 {
6770 	struct net_device *netdev = pci_get_drvdata(pdev);
6771 	struct e1000_adapter *adapter = netdev_priv(netdev);
6772 	struct e1000_hw *hw = &adapter->hw;
6773 	u16 aspm_disable_flag = 0;
6774 
6775 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6776 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6777 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6778 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6779 	if (aspm_disable_flag)
6780 		e1000e_disable_aspm(pdev, aspm_disable_flag);
6781 
6782 	pci_set_master(pdev);
6783 
6784 	if (hw->mac.type >= e1000_pch2lan)
6785 		e1000_resume_workarounds_pchlan(&adapter->hw);
6786 
6787 	e1000e_power_up_phy(adapter);
6788 
6789 	/* report the system wakeup cause from S3/S4 */
6790 	if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6791 		u16 phy_data;
6792 
6793 		e1e_rphy(&adapter->hw, BM_WUS, &phy_data);
6794 		if (phy_data) {
6795 			e_info("PHY Wakeup cause - %s\n",
6796 			       phy_data & E1000_WUS_EX ? "Unicast Packet" :
6797 			       phy_data & E1000_WUS_MC ? "Multicast Packet" :
6798 			       phy_data & E1000_WUS_BC ? "Broadcast Packet" :
6799 			       phy_data & E1000_WUS_MAG ? "Magic Packet" :
6800 			       phy_data & E1000_WUS_LNKC ?
6801 			       "Link Status Change" : "other");
6802 		}
6803 		e1e_wphy(&adapter->hw, BM_WUS, ~0);
6804 	} else {
6805 		u32 wus = er32(WUS);
6806 
6807 		if (wus) {
6808 			e_info("MAC Wakeup cause - %s\n",
6809 			       wus & E1000_WUS_EX ? "Unicast Packet" :
6810 			       wus & E1000_WUS_MC ? "Multicast Packet" :
6811 			       wus & E1000_WUS_BC ? "Broadcast Packet" :
6812 			       wus & E1000_WUS_MAG ? "Magic Packet" :
6813 			       wus & E1000_WUS_LNKC ? "Link Status Change" :
6814 			       "other");
6815 		}
6816 		ew32(WUS, ~0);
6817 	}
6818 
6819 	e1000e_reset(adapter);
6820 
6821 	e1000_init_manageability_pt(adapter);
6822 
6823 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6824 	 * is up.  For all other cases, let the f/w know that the h/w is now
6825 	 * under the control of the driver.
6826 	 */
6827 	if (!(adapter->flags & FLAG_HAS_AMT))
6828 		e1000e_get_hw_control(adapter);
6829 
6830 	return 0;
6831 }
6832 
6833 #ifdef CONFIG_PM_SLEEP
6834 static int e1000e_pm_suspend(struct device *dev)
6835 {
6836 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6837 	struct e1000_adapter *adapter = netdev_priv(netdev);
6838 	struct pci_dev *pdev = to_pci_dev(dev);
6839 	struct e1000_hw *hw = &adapter->hw;
6840 	int rc;
6841 
6842 	e1000e_flush_lpic(pdev);
6843 
6844 	e1000e_pm_freeze(dev);
6845 
6846 	rc = __e1000_shutdown(pdev, false);
6847 	if (rc)
6848 		e1000e_pm_thaw(dev);
6849 
6850 	/* Introduce S0ix implementation */
6851 	if (hw->mac.type >= e1000_pch_cnp)
6852 		e1000e_s0ix_entry_flow(adapter);
6853 
6854 	return rc;
6855 }
6856 
6857 static int e1000e_pm_resume(struct device *dev)
6858 {
6859 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6860 	struct e1000_adapter *adapter = netdev_priv(netdev);
6861 	struct pci_dev *pdev = to_pci_dev(dev);
6862 	struct e1000_hw *hw = &adapter->hw;
6863 	int rc;
6864 
6865 	/* Introduce S0ix implementation */
6866 	if (hw->mac.type >= e1000_pch_cnp)
6867 		e1000e_s0ix_exit_flow(adapter);
6868 
6869 	rc = __e1000_resume(pdev);
6870 	if (rc)
6871 		return rc;
6872 
6873 	return e1000e_pm_thaw(dev);
6874 }
6875 #endif /* CONFIG_PM_SLEEP */
6876 
6877 static int e1000e_pm_runtime_idle(struct device *dev)
6878 {
6879 	struct net_device *netdev = dev_get_drvdata(dev);
6880 	struct e1000_adapter *adapter = netdev_priv(netdev);
6881 	u16 eee_lp;
6882 
6883 	eee_lp = adapter->hw.dev_spec.ich8lan.eee_lp_ability;
6884 
6885 	if (!e1000e_has_link(adapter)) {
6886 		adapter->hw.dev_spec.ich8lan.eee_lp_ability = eee_lp;
6887 		pm_schedule_suspend(dev, 5 * MSEC_PER_SEC);
6888 	}
6889 
6890 	return -EBUSY;
6891 }
6892 
6893 static int e1000e_pm_runtime_resume(struct device *dev)
6894 {
6895 	struct pci_dev *pdev = to_pci_dev(dev);
6896 	struct net_device *netdev = pci_get_drvdata(pdev);
6897 	struct e1000_adapter *adapter = netdev_priv(netdev);
6898 	int rc;
6899 
6900 	rc = __e1000_resume(pdev);
6901 	if (rc)
6902 		return rc;
6903 
6904 	if (netdev->flags & IFF_UP)
6905 		e1000e_up(adapter);
6906 
6907 	return rc;
6908 }
6909 
6910 static int e1000e_pm_runtime_suspend(struct device *dev)
6911 {
6912 	struct pci_dev *pdev = to_pci_dev(dev);
6913 	struct net_device *netdev = pci_get_drvdata(pdev);
6914 	struct e1000_adapter *adapter = netdev_priv(netdev);
6915 
6916 	if (netdev->flags & IFF_UP) {
6917 		int count = E1000_CHECK_RESET_COUNT;
6918 
6919 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6920 			usleep_range(10000, 11000);
6921 
6922 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6923 
6924 		/* Down the device without resetting the hardware */
6925 		e1000e_down(adapter, false);
6926 	}
6927 
6928 	if (__e1000_shutdown(pdev, true)) {
6929 		e1000e_pm_runtime_resume(dev);
6930 		return -EBUSY;
6931 	}
6932 
6933 	return 0;
6934 }
6935 #endif /* CONFIG_PM */
6936 
6937 static void e1000_shutdown(struct pci_dev *pdev)
6938 {
6939 	e1000e_flush_lpic(pdev);
6940 
6941 	e1000e_pm_freeze(&pdev->dev);
6942 
6943 	__e1000_shutdown(pdev, false);
6944 }
6945 
6946 #ifdef CONFIG_NET_POLL_CONTROLLER
6947 
6948 static irqreturn_t e1000_intr_msix(int __always_unused irq, void *data)
6949 {
6950 	struct net_device *netdev = data;
6951 	struct e1000_adapter *adapter = netdev_priv(netdev);
6952 
6953 	if (adapter->msix_entries) {
6954 		int vector, msix_irq;
6955 
6956 		vector = 0;
6957 		msix_irq = adapter->msix_entries[vector].vector;
6958 		if (disable_hardirq(msix_irq))
6959 			e1000_intr_msix_rx(msix_irq, netdev);
6960 		enable_irq(msix_irq);
6961 
6962 		vector++;
6963 		msix_irq = adapter->msix_entries[vector].vector;
6964 		if (disable_hardirq(msix_irq))
6965 			e1000_intr_msix_tx(msix_irq, netdev);
6966 		enable_irq(msix_irq);
6967 
6968 		vector++;
6969 		msix_irq = adapter->msix_entries[vector].vector;
6970 		if (disable_hardirq(msix_irq))
6971 			e1000_msix_other(msix_irq, netdev);
6972 		enable_irq(msix_irq);
6973 	}
6974 
6975 	return IRQ_HANDLED;
6976 }
6977 
6978 /**
6979  * e1000_netpoll
6980  * @netdev: network interface device structure
6981  *
6982  * Polling 'interrupt' - used by things like netconsole to send skbs
6983  * without having to re-enable interrupts. It's not called while
6984  * the interrupt routine is executing.
6985  */
6986 static void e1000_netpoll(struct net_device *netdev)
6987 {
6988 	struct e1000_adapter *adapter = netdev_priv(netdev);
6989 
6990 	switch (adapter->int_mode) {
6991 	case E1000E_INT_MODE_MSIX:
6992 		e1000_intr_msix(adapter->pdev->irq, netdev);
6993 		break;
6994 	case E1000E_INT_MODE_MSI:
6995 		if (disable_hardirq(adapter->pdev->irq))
6996 			e1000_intr_msi(adapter->pdev->irq, netdev);
6997 		enable_irq(adapter->pdev->irq);
6998 		break;
6999 	default:		/* E1000E_INT_MODE_LEGACY */
7000 		if (disable_hardirq(adapter->pdev->irq))
7001 			e1000_intr(adapter->pdev->irq, netdev);
7002 		enable_irq(adapter->pdev->irq);
7003 		break;
7004 	}
7005 }
7006 #endif
7007 
7008 /**
7009  * e1000_io_error_detected - called when PCI error is detected
7010  * @pdev: Pointer to PCI device
7011  * @state: The current pci connection state
7012  *
7013  * This function is called after a PCI bus error affecting
7014  * this device has been detected.
7015  */
7016 static pci_ers_result_t e1000_io_error_detected(struct pci_dev *pdev,
7017 						pci_channel_state_t state)
7018 {
7019 	e1000e_pm_freeze(&pdev->dev);
7020 
7021 	if (state == pci_channel_io_perm_failure)
7022 		return PCI_ERS_RESULT_DISCONNECT;
7023 
7024 	pci_disable_device(pdev);
7025 
7026 	/* Request a slot slot reset. */
7027 	return PCI_ERS_RESULT_NEED_RESET;
7028 }
7029 
7030 /**
7031  * e1000_io_slot_reset - called after the pci bus has been reset.
7032  * @pdev: Pointer to PCI device
7033  *
7034  * Restart the card from scratch, as if from a cold-boot. Implementation
7035  * resembles the first-half of the e1000e_pm_resume routine.
7036  */
7037 static pci_ers_result_t e1000_io_slot_reset(struct pci_dev *pdev)
7038 {
7039 	struct net_device *netdev = pci_get_drvdata(pdev);
7040 	struct e1000_adapter *adapter = netdev_priv(netdev);
7041 	struct e1000_hw *hw = &adapter->hw;
7042 	u16 aspm_disable_flag = 0;
7043 	int err;
7044 	pci_ers_result_t result;
7045 
7046 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
7047 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7048 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
7049 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7050 	if (aspm_disable_flag)
7051 		e1000e_disable_aspm_locked(pdev, aspm_disable_flag);
7052 
7053 	err = pci_enable_device_mem(pdev);
7054 	if (err) {
7055 		dev_err(&pdev->dev,
7056 			"Cannot re-enable PCI device after reset.\n");
7057 		result = PCI_ERS_RESULT_DISCONNECT;
7058 	} else {
7059 		pdev->state_saved = true;
7060 		pci_restore_state(pdev);
7061 		pci_set_master(pdev);
7062 
7063 		pci_enable_wake(pdev, PCI_D3hot, 0);
7064 		pci_enable_wake(pdev, PCI_D3cold, 0);
7065 
7066 		e1000e_reset(adapter);
7067 		ew32(WUS, ~0);
7068 		result = PCI_ERS_RESULT_RECOVERED;
7069 	}
7070 
7071 	return result;
7072 }
7073 
7074 /**
7075  * e1000_io_resume - called when traffic can start flowing again.
7076  * @pdev: Pointer to PCI device
7077  *
7078  * This callback is called when the error recovery driver tells us that
7079  * its OK to resume normal operation. Implementation resembles the
7080  * second-half of the e1000e_pm_resume routine.
7081  */
7082 static void e1000_io_resume(struct pci_dev *pdev)
7083 {
7084 	struct net_device *netdev = pci_get_drvdata(pdev);
7085 	struct e1000_adapter *adapter = netdev_priv(netdev);
7086 
7087 	e1000_init_manageability_pt(adapter);
7088 
7089 	e1000e_pm_thaw(&pdev->dev);
7090 
7091 	/* If the controller has AMT, do not set DRV_LOAD until the interface
7092 	 * is up.  For all other cases, let the f/w know that the h/w is now
7093 	 * under the control of the driver.
7094 	 */
7095 	if (!(adapter->flags & FLAG_HAS_AMT))
7096 		e1000e_get_hw_control(adapter);
7097 }
7098 
7099 static void e1000_print_device_info(struct e1000_adapter *adapter)
7100 {
7101 	struct e1000_hw *hw = &adapter->hw;
7102 	struct net_device *netdev = adapter->netdev;
7103 	u32 ret_val;
7104 	u8 pba_str[E1000_PBANUM_LENGTH];
7105 
7106 	/* print bus type/speed/width info */
7107 	e_info("(PCI Express:2.5GT/s:%s) %pM\n",
7108 	       /* bus width */
7109 	       ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
7110 		"Width x1"),
7111 	       /* MAC address */
7112 	       netdev->dev_addr);
7113 	e_info("Intel(R) PRO/%s Network Connection\n",
7114 	       (hw->phy.type == e1000_phy_ife) ? "10/100" : "1000");
7115 	ret_val = e1000_read_pba_string_generic(hw, pba_str,
7116 						E1000_PBANUM_LENGTH);
7117 	if (ret_val)
7118 		strlcpy((char *)pba_str, "Unknown", sizeof(pba_str));
7119 	e_info("MAC: %d, PHY: %d, PBA No: %s\n",
7120 	       hw->mac.type, hw->phy.type, pba_str);
7121 }
7122 
7123 static void e1000_eeprom_checks(struct e1000_adapter *adapter)
7124 {
7125 	struct e1000_hw *hw = &adapter->hw;
7126 	int ret_val;
7127 	u16 buf = 0;
7128 
7129 	if (hw->mac.type != e1000_82573)
7130 		return;
7131 
7132 	ret_val = e1000_read_nvm(hw, NVM_INIT_CONTROL2_REG, 1, &buf);
7133 	le16_to_cpus(&buf);
7134 	if (!ret_val && (!(buf & BIT(0)))) {
7135 		/* Deep Smart Power Down (DSPD) */
7136 		dev_warn(&adapter->pdev->dev,
7137 			 "Warning: detected DSPD enabled in EEPROM\n");
7138 	}
7139 }
7140 
7141 static netdev_features_t e1000_fix_features(struct net_device *netdev,
7142 					    netdev_features_t features)
7143 {
7144 	struct e1000_adapter *adapter = netdev_priv(netdev);
7145 	struct e1000_hw *hw = &adapter->hw;
7146 
7147 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
7148 	if ((hw->mac.type >= e1000_pch2lan) && (netdev->mtu > ETH_DATA_LEN))
7149 		features &= ~NETIF_F_RXFCS;
7150 
7151 	/* Since there is no support for separate Rx/Tx vlan accel
7152 	 * enable/disable make sure Tx flag is always in same state as Rx.
7153 	 */
7154 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
7155 		features |= NETIF_F_HW_VLAN_CTAG_TX;
7156 	else
7157 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
7158 
7159 	return features;
7160 }
7161 
7162 static int e1000_set_features(struct net_device *netdev,
7163 			      netdev_features_t features)
7164 {
7165 	struct e1000_adapter *adapter = netdev_priv(netdev);
7166 	netdev_features_t changed = features ^ netdev->features;
7167 
7168 	if (changed & (NETIF_F_TSO | NETIF_F_TSO6))
7169 		adapter->flags |= FLAG_TSO_FORCE;
7170 
7171 	if (!(changed & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX |
7172 			 NETIF_F_RXCSUM | NETIF_F_RXHASH | NETIF_F_RXFCS |
7173 			 NETIF_F_RXALL)))
7174 		return 0;
7175 
7176 	if (changed & NETIF_F_RXFCS) {
7177 		if (features & NETIF_F_RXFCS) {
7178 			adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7179 		} else {
7180 			/* We need to take it back to defaults, which might mean
7181 			 * stripping is still disabled at the adapter level.
7182 			 */
7183 			if (adapter->flags2 & FLAG2_DFLT_CRC_STRIPPING)
7184 				adapter->flags2 |= FLAG2_CRC_STRIPPING;
7185 			else
7186 				adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7187 		}
7188 	}
7189 
7190 	netdev->features = features;
7191 
7192 	if (netif_running(netdev))
7193 		e1000e_reinit_locked(adapter);
7194 	else
7195 		e1000e_reset(adapter);
7196 
7197 	return 1;
7198 }
7199 
7200 static const struct net_device_ops e1000e_netdev_ops = {
7201 	.ndo_open		= e1000e_open,
7202 	.ndo_stop		= e1000e_close,
7203 	.ndo_start_xmit		= e1000_xmit_frame,
7204 	.ndo_get_stats64	= e1000e_get_stats64,
7205 	.ndo_set_rx_mode	= e1000e_set_rx_mode,
7206 	.ndo_set_mac_address	= e1000_set_mac,
7207 	.ndo_change_mtu		= e1000_change_mtu,
7208 	.ndo_do_ioctl		= e1000_ioctl,
7209 	.ndo_tx_timeout		= e1000_tx_timeout,
7210 	.ndo_validate_addr	= eth_validate_addr,
7211 
7212 	.ndo_vlan_rx_add_vid	= e1000_vlan_rx_add_vid,
7213 	.ndo_vlan_rx_kill_vid	= e1000_vlan_rx_kill_vid,
7214 #ifdef CONFIG_NET_POLL_CONTROLLER
7215 	.ndo_poll_controller	= e1000_netpoll,
7216 #endif
7217 	.ndo_set_features = e1000_set_features,
7218 	.ndo_fix_features = e1000_fix_features,
7219 	.ndo_features_check	= passthru_features_check,
7220 };
7221 
7222 /**
7223  * e1000_probe - Device Initialization Routine
7224  * @pdev: PCI device information struct
7225  * @ent: entry in e1000_pci_tbl
7226  *
7227  * Returns 0 on success, negative on failure
7228  *
7229  * e1000_probe initializes an adapter identified by a pci_dev structure.
7230  * The OS initialization, configuring of the adapter private structure,
7231  * and a hardware reset occur.
7232  **/
7233 static int e1000_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7234 {
7235 	struct net_device *netdev;
7236 	struct e1000_adapter *adapter;
7237 	struct e1000_hw *hw;
7238 	const struct e1000_info *ei = e1000_info_tbl[ent->driver_data];
7239 	resource_size_t mmio_start, mmio_len;
7240 	resource_size_t flash_start, flash_len;
7241 	static int cards_found;
7242 	u16 aspm_disable_flag = 0;
7243 	int bars, i, err, pci_using_dac;
7244 	u16 eeprom_data = 0;
7245 	u16 eeprom_apme_mask = E1000_EEPROM_APME;
7246 	s32 ret_val = 0;
7247 
7248 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L0S)
7249 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7250 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L1)
7251 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7252 	if (aspm_disable_flag)
7253 		e1000e_disable_aspm(pdev, aspm_disable_flag);
7254 
7255 	err = pci_enable_device_mem(pdev);
7256 	if (err)
7257 		return err;
7258 
7259 	pci_using_dac = 0;
7260 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
7261 	if (!err) {
7262 		pci_using_dac = 1;
7263 	} else {
7264 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
7265 		if (err) {
7266 			dev_err(&pdev->dev,
7267 				"No usable DMA configuration, aborting\n");
7268 			goto err_dma;
7269 		}
7270 	}
7271 
7272 	bars = pci_select_bars(pdev, IORESOURCE_MEM);
7273 	err = pci_request_selected_regions_exclusive(pdev, bars,
7274 						     e1000e_driver_name);
7275 	if (err)
7276 		goto err_pci_reg;
7277 
7278 	/* AER (Advanced Error Reporting) hooks */
7279 	pci_enable_pcie_error_reporting(pdev);
7280 
7281 	pci_set_master(pdev);
7282 	/* PCI config space info */
7283 	err = pci_save_state(pdev);
7284 	if (err)
7285 		goto err_alloc_etherdev;
7286 
7287 	err = -ENOMEM;
7288 	netdev = alloc_etherdev(sizeof(struct e1000_adapter));
7289 	if (!netdev)
7290 		goto err_alloc_etherdev;
7291 
7292 	SET_NETDEV_DEV(netdev, &pdev->dev);
7293 
7294 	netdev->irq = pdev->irq;
7295 
7296 	pci_set_drvdata(pdev, netdev);
7297 	adapter = netdev_priv(netdev);
7298 	hw = &adapter->hw;
7299 	adapter->netdev = netdev;
7300 	adapter->pdev = pdev;
7301 	adapter->ei = ei;
7302 	adapter->pba = ei->pba;
7303 	adapter->flags = ei->flags;
7304 	adapter->flags2 = ei->flags2;
7305 	adapter->hw.adapter = adapter;
7306 	adapter->hw.mac.type = ei->mac;
7307 	adapter->max_hw_frame_size = ei->max_hw_frame_size;
7308 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
7309 
7310 	mmio_start = pci_resource_start(pdev, 0);
7311 	mmio_len = pci_resource_len(pdev, 0);
7312 
7313 	err = -EIO;
7314 	adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
7315 	if (!adapter->hw.hw_addr)
7316 		goto err_ioremap;
7317 
7318 	if ((adapter->flags & FLAG_HAS_FLASH) &&
7319 	    (pci_resource_flags(pdev, 1) & IORESOURCE_MEM) &&
7320 	    (hw->mac.type < e1000_pch_spt)) {
7321 		flash_start = pci_resource_start(pdev, 1);
7322 		flash_len = pci_resource_len(pdev, 1);
7323 		adapter->hw.flash_address = ioremap(flash_start, flash_len);
7324 		if (!adapter->hw.flash_address)
7325 			goto err_flashmap;
7326 	}
7327 
7328 	/* Set default EEE advertisement */
7329 	if (adapter->flags2 & FLAG2_HAS_EEE)
7330 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
7331 
7332 	/* construct the net_device struct */
7333 	netdev->netdev_ops = &e1000e_netdev_ops;
7334 	e1000e_set_ethtool_ops(netdev);
7335 	netdev->watchdog_timeo = 5 * HZ;
7336 	netif_napi_add(netdev, &adapter->napi, e1000e_poll, 64);
7337 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
7338 
7339 	netdev->mem_start = mmio_start;
7340 	netdev->mem_end = mmio_start + mmio_len;
7341 
7342 	adapter->bd_number = cards_found++;
7343 
7344 	e1000e_check_options(adapter);
7345 
7346 	/* setup adapter struct */
7347 	err = e1000_sw_init(adapter);
7348 	if (err)
7349 		goto err_sw_init;
7350 
7351 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
7352 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
7353 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
7354 
7355 	err = ei->get_variants(adapter);
7356 	if (err)
7357 		goto err_hw_init;
7358 
7359 	if ((adapter->flags & FLAG_IS_ICH) &&
7360 	    (adapter->flags & FLAG_READ_ONLY_NVM) &&
7361 	    (hw->mac.type < e1000_pch_spt))
7362 		e1000e_write_protect_nvm_ich8lan(&adapter->hw);
7363 
7364 	hw->mac.ops.get_bus_info(&adapter->hw);
7365 
7366 	adapter->hw.phy.autoneg_wait_to_complete = 0;
7367 
7368 	/* Copper options */
7369 	if (adapter->hw.phy.media_type == e1000_media_type_copper) {
7370 		adapter->hw.phy.mdix = AUTO_ALL_MODES;
7371 		adapter->hw.phy.disable_polarity_correction = 0;
7372 		adapter->hw.phy.ms_type = e1000_ms_hw_default;
7373 	}
7374 
7375 	if (hw->phy.ops.check_reset_block && hw->phy.ops.check_reset_block(hw))
7376 		dev_info(&pdev->dev,
7377 			 "PHY reset is blocked due to SOL/IDER session.\n");
7378 
7379 	/* Set initial default active device features */
7380 	netdev->features = (NETIF_F_SG |
7381 			    NETIF_F_HW_VLAN_CTAG_RX |
7382 			    NETIF_F_HW_VLAN_CTAG_TX |
7383 			    NETIF_F_TSO |
7384 			    NETIF_F_TSO6 |
7385 			    NETIF_F_RXHASH |
7386 			    NETIF_F_RXCSUM |
7387 			    NETIF_F_HW_CSUM);
7388 
7389 	/* Set user-changeable features (subset of all device features) */
7390 	netdev->hw_features = netdev->features;
7391 	netdev->hw_features |= NETIF_F_RXFCS;
7392 	netdev->priv_flags |= IFF_SUPP_NOFCS;
7393 	netdev->hw_features |= NETIF_F_RXALL;
7394 
7395 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER)
7396 		netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
7397 
7398 	netdev->vlan_features |= (NETIF_F_SG |
7399 				  NETIF_F_TSO |
7400 				  NETIF_F_TSO6 |
7401 				  NETIF_F_HW_CSUM);
7402 
7403 	netdev->priv_flags |= IFF_UNICAST_FLT;
7404 
7405 	if (pci_using_dac) {
7406 		netdev->features |= NETIF_F_HIGHDMA;
7407 		netdev->vlan_features |= NETIF_F_HIGHDMA;
7408 	}
7409 
7410 	/* MTU range: 68 - max_hw_frame_size */
7411 	netdev->min_mtu = ETH_MIN_MTU;
7412 	netdev->max_mtu = adapter->max_hw_frame_size -
7413 			  (VLAN_ETH_HLEN + ETH_FCS_LEN);
7414 
7415 	if (e1000e_enable_mng_pass_thru(&adapter->hw))
7416 		adapter->flags |= FLAG_MNG_PT_ENABLED;
7417 
7418 	/* before reading the NVM, reset the controller to
7419 	 * put the device in a known good starting state
7420 	 */
7421 	adapter->hw.mac.ops.reset_hw(&adapter->hw);
7422 
7423 	/* systems with ASPM and others may see the checksum fail on the first
7424 	 * attempt. Let's give it a few tries
7425 	 */
7426 	for (i = 0;; i++) {
7427 		if (e1000_validate_nvm_checksum(&adapter->hw) >= 0)
7428 			break;
7429 		if (i == 2) {
7430 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
7431 			err = -EIO;
7432 			goto err_eeprom;
7433 		}
7434 	}
7435 
7436 	e1000_eeprom_checks(adapter);
7437 
7438 	/* copy the MAC address */
7439 	if (e1000e_read_mac_addr(&adapter->hw))
7440 		dev_err(&pdev->dev,
7441 			"NVM Read Error while reading MAC address\n");
7442 
7443 	memcpy(netdev->dev_addr, adapter->hw.mac.addr, netdev->addr_len);
7444 
7445 	if (!is_valid_ether_addr(netdev->dev_addr)) {
7446 		dev_err(&pdev->dev, "Invalid MAC Address: %pM\n",
7447 			netdev->dev_addr);
7448 		err = -EIO;
7449 		goto err_eeprom;
7450 	}
7451 
7452 	adapter->e1000_workqueue = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0,
7453 						   e1000e_driver_name);
7454 
7455 	if (!adapter->e1000_workqueue) {
7456 		err = -ENOMEM;
7457 		goto err_workqueue;
7458 	}
7459 
7460 	INIT_DELAYED_WORK(&adapter->watchdog_task, e1000_watchdog_task);
7461 	queue_delayed_work(adapter->e1000_workqueue, &adapter->watchdog_task,
7462 			   0);
7463 
7464 	timer_setup(&adapter->phy_info_timer, e1000_update_phy_info, 0);
7465 
7466 	INIT_WORK(&adapter->reset_task, e1000_reset_task);
7467 	INIT_WORK(&adapter->downshift_task, e1000e_downshift_workaround);
7468 	INIT_WORK(&adapter->update_phy_task, e1000e_update_phy_task);
7469 	INIT_WORK(&adapter->print_hang_task, e1000_print_hw_hang);
7470 
7471 	/* Initialize link parameters. User can change them with ethtool */
7472 	adapter->hw.mac.autoneg = 1;
7473 	adapter->fc_autoneg = true;
7474 	adapter->hw.fc.requested_mode = e1000_fc_default;
7475 	adapter->hw.fc.current_mode = e1000_fc_default;
7476 	adapter->hw.phy.autoneg_advertised = 0x2f;
7477 
7478 	/* Initial Wake on LAN setting - If APM wake is enabled in
7479 	 * the EEPROM, enable the ACPI Magic Packet filter
7480 	 */
7481 	if (adapter->flags & FLAG_APME_IN_WUC) {
7482 		/* APME bit in EEPROM is mapped to WUC.APME */
7483 		eeprom_data = er32(WUC);
7484 		eeprom_apme_mask = E1000_WUC_APME;
7485 		if ((hw->mac.type > e1000_ich10lan) &&
7486 		    (eeprom_data & E1000_WUC_PHY_WAKE))
7487 			adapter->flags2 |= FLAG2_HAS_PHY_WAKEUP;
7488 	} else if (adapter->flags & FLAG_APME_IN_CTRL3) {
7489 		if (adapter->flags & FLAG_APME_CHECK_PORT_B &&
7490 		    (adapter->hw.bus.func == 1))
7491 			ret_val = e1000_read_nvm(&adapter->hw,
7492 					      NVM_INIT_CONTROL3_PORT_B,
7493 					      1, &eeprom_data);
7494 		else
7495 			ret_val = e1000_read_nvm(&adapter->hw,
7496 					      NVM_INIT_CONTROL3_PORT_A,
7497 					      1, &eeprom_data);
7498 	}
7499 
7500 	/* fetch WoL from EEPROM */
7501 	if (ret_val)
7502 		e_dbg("NVM read error getting WoL initial values: %d\n", ret_val);
7503 	else if (eeprom_data & eeprom_apme_mask)
7504 		adapter->eeprom_wol |= E1000_WUFC_MAG;
7505 
7506 	/* now that we have the eeprom settings, apply the special cases
7507 	 * where the eeprom may be wrong or the board simply won't support
7508 	 * wake on lan on a particular port
7509 	 */
7510 	if (!(adapter->flags & FLAG_HAS_WOL))
7511 		adapter->eeprom_wol = 0;
7512 
7513 	/* initialize the wol settings based on the eeprom settings */
7514 	adapter->wol = adapter->eeprom_wol;
7515 
7516 	/* make sure adapter isn't asleep if manageability is enabled */
7517 	if (adapter->wol || (adapter->flags & FLAG_MNG_PT_ENABLED) ||
7518 	    (hw->mac.ops.check_mng_mode(hw)))
7519 		device_wakeup_enable(&pdev->dev);
7520 
7521 	/* save off EEPROM version number */
7522 	ret_val = e1000_read_nvm(&adapter->hw, 5, 1, &adapter->eeprom_vers);
7523 
7524 	if (ret_val) {
7525 		e_dbg("NVM read error getting EEPROM version: %d\n", ret_val);
7526 		adapter->eeprom_vers = 0;
7527 	}
7528 
7529 	/* init PTP hardware clock */
7530 	e1000e_ptp_init(adapter);
7531 
7532 	/* reset the hardware with the new settings */
7533 	e1000e_reset(adapter);
7534 
7535 	/* If the controller has AMT, do not set DRV_LOAD until the interface
7536 	 * is up.  For all other cases, let the f/w know that the h/w is now
7537 	 * under the control of the driver.
7538 	 */
7539 	if (!(adapter->flags & FLAG_HAS_AMT))
7540 		e1000e_get_hw_control(adapter);
7541 
7542 	strlcpy(netdev->name, "eth%d", sizeof(netdev->name));
7543 	err = register_netdev(netdev);
7544 	if (err)
7545 		goto err_register;
7546 
7547 	/* carrier off reporting is important to ethtool even BEFORE open */
7548 	netif_carrier_off(netdev);
7549 
7550 	e1000_print_device_info(adapter);
7551 
7552 	dev_pm_set_driver_flags(&pdev->dev, DPM_FLAG_NEVER_SKIP);
7553 
7554 	if (pci_dev_run_wake(pdev) && hw->mac.type < e1000_pch_cnp)
7555 		pm_runtime_put_noidle(&pdev->dev);
7556 
7557 	return 0;
7558 
7559 err_register:
7560 	flush_workqueue(adapter->e1000_workqueue);
7561 	destroy_workqueue(adapter->e1000_workqueue);
7562 err_workqueue:
7563 	if (!(adapter->flags & FLAG_HAS_AMT))
7564 		e1000e_release_hw_control(adapter);
7565 err_eeprom:
7566 	if (hw->phy.ops.check_reset_block && !hw->phy.ops.check_reset_block(hw))
7567 		e1000_phy_hw_reset(&adapter->hw);
7568 err_hw_init:
7569 	kfree(adapter->tx_ring);
7570 	kfree(adapter->rx_ring);
7571 err_sw_init:
7572 	if ((adapter->hw.flash_address) && (hw->mac.type < e1000_pch_spt))
7573 		iounmap(adapter->hw.flash_address);
7574 	e1000e_reset_interrupt_capability(adapter);
7575 err_flashmap:
7576 	iounmap(adapter->hw.hw_addr);
7577 err_ioremap:
7578 	free_netdev(netdev);
7579 err_alloc_etherdev:
7580 	pci_release_mem_regions(pdev);
7581 err_pci_reg:
7582 err_dma:
7583 	pci_disable_device(pdev);
7584 	return err;
7585 }
7586 
7587 /**
7588  * e1000_remove - Device Removal Routine
7589  * @pdev: PCI device information struct
7590  *
7591  * e1000_remove is called by the PCI subsystem to alert the driver
7592  * that it should release a PCI device.  The could be caused by a
7593  * Hot-Plug event, or because the driver is going to be removed from
7594  * memory.
7595  **/
7596 static void e1000_remove(struct pci_dev *pdev)
7597 {
7598 	struct net_device *netdev = pci_get_drvdata(pdev);
7599 	struct e1000_adapter *adapter = netdev_priv(netdev);
7600 
7601 	e1000e_ptp_remove(adapter);
7602 
7603 	/* The timers may be rescheduled, so explicitly disable them
7604 	 * from being rescheduled.
7605 	 */
7606 	set_bit(__E1000_DOWN, &adapter->state);
7607 	del_timer_sync(&adapter->phy_info_timer);
7608 
7609 	cancel_work_sync(&adapter->reset_task);
7610 	cancel_work_sync(&adapter->downshift_task);
7611 	cancel_work_sync(&adapter->update_phy_task);
7612 	cancel_work_sync(&adapter->print_hang_task);
7613 
7614 	cancel_delayed_work(&adapter->watchdog_task);
7615 	flush_workqueue(adapter->e1000_workqueue);
7616 	destroy_workqueue(adapter->e1000_workqueue);
7617 
7618 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
7619 		cancel_work_sync(&adapter->tx_hwtstamp_work);
7620 		if (adapter->tx_hwtstamp_skb) {
7621 			dev_consume_skb_any(adapter->tx_hwtstamp_skb);
7622 			adapter->tx_hwtstamp_skb = NULL;
7623 		}
7624 	}
7625 
7626 	unregister_netdev(netdev);
7627 
7628 	if (pci_dev_run_wake(pdev))
7629 		pm_runtime_get_noresume(&pdev->dev);
7630 
7631 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
7632 	 * would have already happened in close and is redundant.
7633 	 */
7634 	e1000e_release_hw_control(adapter);
7635 
7636 	e1000e_reset_interrupt_capability(adapter);
7637 	kfree(adapter->tx_ring);
7638 	kfree(adapter->rx_ring);
7639 
7640 	iounmap(adapter->hw.hw_addr);
7641 	if ((adapter->hw.flash_address) &&
7642 	    (adapter->hw.mac.type < e1000_pch_spt))
7643 		iounmap(adapter->hw.flash_address);
7644 	pci_release_mem_regions(pdev);
7645 
7646 	free_netdev(netdev);
7647 
7648 	/* AER disable */
7649 	pci_disable_pcie_error_reporting(pdev);
7650 
7651 	pci_disable_device(pdev);
7652 }
7653 
7654 /* PCI Error Recovery (ERS) */
7655 static const struct pci_error_handlers e1000_err_handler = {
7656 	.error_detected = e1000_io_error_detected,
7657 	.slot_reset = e1000_io_slot_reset,
7658 	.resume = e1000_io_resume,
7659 };
7660 
7661 static const struct pci_device_id e1000_pci_tbl[] = {
7662 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_COPPER), board_82571 },
7663 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_FIBER), board_82571 },
7664 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER), board_82571 },
7665 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER_LP),
7666 	  board_82571 },
7667 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_FIBER), board_82571 },
7668 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES), board_82571 },
7669 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_DUAL), board_82571 },
7670 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_QUAD), board_82571 },
7671 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571PT_QUAD_COPPER), board_82571 },
7672 
7673 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI), board_82572 },
7674 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_COPPER), board_82572 },
7675 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_FIBER), board_82572 },
7676 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_SERDES), board_82572 },
7677 
7678 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E), board_82573 },
7679 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E_IAMT), board_82573 },
7680 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573L), board_82573 },
7681 
7682 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574L), board_82574 },
7683 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574LA), board_82574 },
7684 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82583V), board_82583 },
7685 
7686 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_DPT),
7687 	  board_80003es2lan },
7688 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_SPT),
7689 	  board_80003es2lan },
7690 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_DPT),
7691 	  board_80003es2lan },
7692 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_SPT),
7693 	  board_80003es2lan },
7694 
7695 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE), board_ich8lan },
7696 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_G), board_ich8lan },
7697 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_GT), board_ich8lan },
7698 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_AMT), board_ich8lan },
7699 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_C), board_ich8lan },
7700 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M), board_ich8lan },
7701 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M_AMT), board_ich8lan },
7702 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_82567V_3), board_ich8lan },
7703 
7704 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE), board_ich9lan },
7705 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_G), board_ich9lan },
7706 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_GT), board_ich9lan },
7707 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_AMT), board_ich9lan },
7708 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_C), board_ich9lan },
7709 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_BM), board_ich9lan },
7710 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M), board_ich9lan },
7711 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_AMT), board_ich9lan },
7712 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_V), board_ich9lan },
7713 
7714 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LM), board_ich9lan },
7715 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LF), board_ich9lan },
7716 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_V), board_ich9lan },
7717 
7718 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LM), board_ich10lan },
7719 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LF), board_ich10lan },
7720 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_V), board_ich10lan },
7721 
7722 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LM), board_pchlan },
7723 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LC), board_pchlan },
7724 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DM), board_pchlan },
7725 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DC), board_pchlan },
7726 
7727 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_LM), board_pch2lan },
7728 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_V), board_pch2lan },
7729 
7730 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_LM), board_pch_lpt },
7731 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_V), board_pch_lpt },
7732 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_LM), board_pch_lpt },
7733 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_V), board_pch_lpt },
7734 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM2), board_pch_lpt },
7735 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V2), board_pch_lpt },
7736 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM3), board_pch_lpt },
7737 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V3), board_pch_lpt },
7738 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM), board_pch_spt },
7739 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V), board_pch_spt },
7740 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM2), board_pch_spt },
7741 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V2), board_pch_spt },
7742 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LBG_I219_LM3), board_pch_spt },
7743 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM4), board_pch_spt },
7744 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V4), board_pch_spt },
7745 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM5), board_pch_spt },
7746 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V5), board_pch_spt },
7747 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM6), board_pch_cnp },
7748 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V6), board_pch_cnp },
7749 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM7), board_pch_cnp },
7750 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V7), board_pch_cnp },
7751 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM8), board_pch_cnp },
7752 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V8), board_pch_cnp },
7753 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM9), board_pch_cnp },
7754 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V9), board_pch_cnp },
7755 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM10), board_pch_cnp },
7756 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V10), board_pch_cnp },
7757 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM11), board_pch_cnp },
7758 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V11), board_pch_cnp },
7759 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM12), board_pch_spt },
7760 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V12), board_pch_spt },
7761 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM13), board_pch_cnp },
7762 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V13), board_pch_cnp },
7763 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM14), board_pch_cnp },
7764 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V14), board_pch_cnp },
7765 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM15), board_pch_cnp },
7766 
7767 	{ 0, 0, 0, 0, 0, 0, 0 }	/* terminate list */
7768 };
7769 MODULE_DEVICE_TABLE(pci, e1000_pci_tbl);
7770 
7771 static const struct dev_pm_ops e1000_pm_ops = {
7772 #ifdef CONFIG_PM_SLEEP
7773 	.suspend	= e1000e_pm_suspend,
7774 	.resume		= e1000e_pm_resume,
7775 	.freeze		= e1000e_pm_freeze,
7776 	.thaw		= e1000e_pm_thaw,
7777 	.poweroff	= e1000e_pm_suspend,
7778 	.restore	= e1000e_pm_resume,
7779 #endif
7780 	SET_RUNTIME_PM_OPS(e1000e_pm_runtime_suspend, e1000e_pm_runtime_resume,
7781 			   e1000e_pm_runtime_idle)
7782 };
7783 
7784 /* PCI Device API Driver */
7785 static struct pci_driver e1000_driver = {
7786 	.name     = e1000e_driver_name,
7787 	.id_table = e1000_pci_tbl,
7788 	.probe    = e1000_probe,
7789 	.remove   = e1000_remove,
7790 	.driver   = {
7791 		.pm = &e1000_pm_ops,
7792 	},
7793 	.shutdown = e1000_shutdown,
7794 	.err_handler = &e1000_err_handler
7795 };
7796 
7797 /**
7798  * e1000_init_module - Driver Registration Routine
7799  *
7800  * e1000_init_module is the first routine called when the driver is
7801  * loaded. All it does is register with the PCI subsystem.
7802  **/
7803 static int __init e1000_init_module(void)
7804 {
7805 	pr_info("Intel(R) PRO/1000 Network Driver - %s\n",
7806 		e1000e_driver_version);
7807 	pr_info("Copyright(c) 1999 - 2015 Intel Corporation.\n");
7808 
7809 	return pci_register_driver(&e1000_driver);
7810 }
7811 module_init(e1000_init_module);
7812 
7813 /**
7814  * e1000_exit_module - Driver Exit Cleanup Routine
7815  *
7816  * e1000_exit_module is called just before the driver is removed
7817  * from memory.
7818  **/
7819 static void __exit e1000_exit_module(void)
7820 {
7821 	pci_unregister_driver(&e1000_driver);
7822 }
7823 module_exit(e1000_exit_module);
7824 
7825 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
7826 MODULE_DESCRIPTION("Intel(R) PRO/1000 Network Driver");
7827 MODULE_LICENSE("GPL v2");
7828 MODULE_VERSION(DRV_VERSION);
7829 
7830 /* netdev.c */
7831