xref: /linux/drivers/net/ethernet/intel/e1000e/netdev.c (revision 578294b8b60d2c630991f221838f9ec61ae89df0)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright(c) 1999 - 2018 Intel Corporation. */
3 
4 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
5 
6 #include <linux/module.h>
7 #include <linux/types.h>
8 #include <linux/init.h>
9 #include <linux/pci.h>
10 #include <linux/vmalloc.h>
11 #include <linux/pagemap.h>
12 #include <linux/delay.h>
13 #include <linux/netdevice.h>
14 #include <linux/interrupt.h>
15 #include <linux/tcp.h>
16 #include <linux/ipv6.h>
17 #include <linux/slab.h>
18 #include <net/checksum.h>
19 #include <net/ip6_checksum.h>
20 #include <linux/ethtool.h>
21 #include <linux/if_vlan.h>
22 #include <linux/cpu.h>
23 #include <linux/smp.h>
24 #include <linux/pm_qos.h>
25 #include <linux/pm_runtime.h>
26 #include <linux/prefetch.h>
27 #include <linux/suspend.h>
28 #include <linux/dmi.h>
29 
30 #include "e1000.h"
31 #define CREATE_TRACE_POINTS
32 #include "e1000e_trace.h"
33 
34 char e1000e_driver_name[] = "e1000e";
35 
36 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
37 static int debug = -1;
38 module_param(debug, int, 0);
39 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
40 
41 static const struct e1000_info *e1000_info_tbl[] = {
42 	[board_82571]		= &e1000_82571_info,
43 	[board_82572]		= &e1000_82572_info,
44 	[board_82573]		= &e1000_82573_info,
45 	[board_82574]		= &e1000_82574_info,
46 	[board_82583]		= &e1000_82583_info,
47 	[board_80003es2lan]	= &e1000_es2_info,
48 	[board_ich8lan]		= &e1000_ich8_info,
49 	[board_ich9lan]		= &e1000_ich9_info,
50 	[board_ich10lan]	= &e1000_ich10_info,
51 	[board_pchlan]		= &e1000_pch_info,
52 	[board_pch2lan]		= &e1000_pch2_info,
53 	[board_pch_lpt]		= &e1000_pch_lpt_info,
54 	[board_pch_spt]		= &e1000_pch_spt_info,
55 	[board_pch_cnp]		= &e1000_pch_cnp_info,
56 	[board_pch_tgp]		= &e1000_pch_tgp_info,
57 	[board_pch_adp]		= &e1000_pch_adp_info,
58 	[board_pch_mtp]		= &e1000_pch_mtp_info,
59 	[board_pch_ptp]		= &e1000_pch_ptp_info,
60 };
61 
62 static const struct dmi_system_id disable_k1_list[] = {
63 	{
64 		.ident = "Dell Pro 16 Plus PB16250",
65 		.matches = {
66 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
67 			DMI_MATCH(DMI_PRODUCT_NAME, "Dell Pro 16 Plus PB16250"),
68 		},
69 	},
70 	{}
71 };
72 
73 struct e1000_reg_info {
74 	u32 ofs;
75 	char *name;
76 };
77 
78 static const struct e1000_reg_info e1000_reg_info_tbl[] = {
79 	/* General Registers */
80 	{E1000_CTRL, "CTRL"},
81 	{E1000_STATUS, "STATUS"},
82 	{E1000_CTRL_EXT, "CTRL_EXT"},
83 
84 	/* Interrupt Registers */
85 	{E1000_ICR, "ICR"},
86 
87 	/* Rx Registers */
88 	{E1000_RCTL, "RCTL"},
89 	{E1000_RDLEN(0), "RDLEN"},
90 	{E1000_RDH(0), "RDH"},
91 	{E1000_RDT(0), "RDT"},
92 	{E1000_RDTR, "RDTR"},
93 	{E1000_RXDCTL(0), "RXDCTL"},
94 	{E1000_ERT, "ERT"},
95 	{E1000_RDBAL(0), "RDBAL"},
96 	{E1000_RDBAH(0), "RDBAH"},
97 	{E1000_RDFH, "RDFH"},
98 	{E1000_RDFT, "RDFT"},
99 	{E1000_RDFHS, "RDFHS"},
100 	{E1000_RDFTS, "RDFTS"},
101 	{E1000_RDFPC, "RDFPC"},
102 
103 	/* Tx Registers */
104 	{E1000_TCTL, "TCTL"},
105 	{E1000_TDBAL(0), "TDBAL"},
106 	{E1000_TDBAH(0), "TDBAH"},
107 	{E1000_TDLEN(0), "TDLEN"},
108 	{E1000_TDH(0), "TDH"},
109 	{E1000_TDT(0), "TDT"},
110 	{E1000_TIDV, "TIDV"},
111 	{E1000_TXDCTL(0), "TXDCTL"},
112 	{E1000_TADV, "TADV"},
113 	{E1000_TARC(0), "TARC"},
114 	{E1000_TDFH, "TDFH"},
115 	{E1000_TDFT, "TDFT"},
116 	{E1000_TDFHS, "TDFHS"},
117 	{E1000_TDFTS, "TDFTS"},
118 	{E1000_TDFPC, "TDFPC"},
119 
120 	/* List Terminator */
121 	{0, NULL}
122 };
123 
124 /**
125  * __ew32_prepare - prepare to write to MAC CSR register on certain parts
126  * @hw: pointer to the HW structure
127  *
128  * When updating the MAC CSR registers, the Manageability Engine (ME) could
129  * be accessing the registers at the same time.  Normally, this is handled in
130  * h/w by an arbiter but on some parts there is a bug that acknowledges Host
131  * accesses later than it should which could result in the register to have
132  * an incorrect value.  Workaround this by checking the FWSM register which
133  * has bit 24 set while ME is accessing MAC CSR registers, wait if it is set
134  * and try again a number of times.
135  **/
136 static void __ew32_prepare(struct e1000_hw *hw)
137 {
138 	s32 i = E1000_ICH_FWSM_PCIM2PCI_COUNT;
139 
140 	while ((er32(FWSM) & E1000_ICH_FWSM_PCIM2PCI) && --i)
141 		udelay(50);
142 }
143 
144 void __ew32(struct e1000_hw *hw, unsigned long reg, u32 val)
145 {
146 	if (hw->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
147 		__ew32_prepare(hw);
148 
149 	writel(val, hw->hw_addr + reg);
150 }
151 
152 /**
153  * e1000_regdump - register printout routine
154  * @hw: pointer to the HW structure
155  * @reginfo: pointer to the register info table
156  **/
157 static void e1000_regdump(struct e1000_hw *hw, struct e1000_reg_info *reginfo)
158 {
159 	int n = 0;
160 	char rname[16];
161 	u32 regs[8];
162 
163 	switch (reginfo->ofs) {
164 	case E1000_RXDCTL(0):
165 		for (n = 0; n < 2; n++)
166 			regs[n] = __er32(hw, E1000_RXDCTL(n));
167 		break;
168 	case E1000_TXDCTL(0):
169 		for (n = 0; n < 2; n++)
170 			regs[n] = __er32(hw, E1000_TXDCTL(n));
171 		break;
172 	case E1000_TARC(0):
173 		for (n = 0; n < 2; n++)
174 			regs[n] = __er32(hw, E1000_TARC(n));
175 		break;
176 	default:
177 		pr_info("%-15s %08x\n",
178 			reginfo->name, __er32(hw, reginfo->ofs));
179 		return;
180 	}
181 
182 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-1]");
183 	pr_info("%-15s %08x %08x\n", rname, regs[0], regs[1]);
184 }
185 
186 static void e1000e_dump_ps_pages(struct e1000_adapter *adapter,
187 				 struct e1000_buffer *bi)
188 {
189 	int i;
190 	struct e1000_ps_page *ps_page;
191 
192 	for (i = 0; i < adapter->rx_ps_pages; i++) {
193 		ps_page = &bi->ps_pages[i];
194 
195 		if (ps_page->page) {
196 			pr_info("packet dump for ps_page %d:\n", i);
197 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
198 				       16, 1, page_address(ps_page->page),
199 				       PAGE_SIZE, true);
200 		}
201 	}
202 }
203 
204 /**
205  * e1000e_dump - Print registers, Tx-ring and Rx-ring
206  * @adapter: board private structure
207  **/
208 static void e1000e_dump(struct e1000_adapter *adapter)
209 {
210 	struct net_device *netdev = adapter->netdev;
211 	struct e1000_hw *hw = &adapter->hw;
212 	struct e1000_reg_info *reginfo;
213 	struct e1000_ring *tx_ring = adapter->tx_ring;
214 	struct e1000_tx_desc *tx_desc;
215 	struct my_u0 {
216 		__le64 a;
217 		__le64 b;
218 	} *u0;
219 	struct e1000_buffer *buffer_info;
220 	struct e1000_ring *rx_ring = adapter->rx_ring;
221 	union e1000_rx_desc_packet_split *rx_desc_ps;
222 	union e1000_rx_desc_extended *rx_desc;
223 	struct my_u1 {
224 		__le64 a;
225 		__le64 b;
226 		__le64 c;
227 		__le64 d;
228 	} *u1;
229 	u32 staterr;
230 	int i = 0;
231 
232 	if (!netif_msg_hw(adapter))
233 		return;
234 
235 	/* Print netdevice Info */
236 	if (netdev) {
237 		dev_info(&adapter->pdev->dev, "Net device Info\n");
238 		pr_info("Device Name     state            trans_start\n");
239 		pr_info("%-15s %016lX %016lX\n", netdev->name,
240 			netdev->state, dev_trans_start(netdev));
241 	}
242 
243 	/* Print Registers */
244 	dev_info(&adapter->pdev->dev, "Register Dump\n");
245 	pr_info(" Register Name   Value\n");
246 	for (reginfo = (struct e1000_reg_info *)e1000_reg_info_tbl;
247 	     reginfo->name; reginfo++) {
248 		e1000_regdump(hw, reginfo);
249 	}
250 
251 	/* Print Tx Ring Summary */
252 	if (!netdev || !netif_running(netdev))
253 		return;
254 
255 	dev_info(&adapter->pdev->dev, "Tx Ring Summary\n");
256 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
257 	buffer_info = &tx_ring->buffer_info[tx_ring->next_to_clean];
258 	pr_info(" %5d %5X %5X %016llX %04X %3X %016llX\n",
259 		0, tx_ring->next_to_use, tx_ring->next_to_clean,
260 		(unsigned long long)buffer_info->dma,
261 		buffer_info->length,
262 		buffer_info->next_to_watch,
263 		(unsigned long long)buffer_info->time_stamp);
264 
265 	/* Print Tx Ring */
266 	if (!netif_msg_tx_done(adapter))
267 		goto rx_ring_summary;
268 
269 	dev_info(&adapter->pdev->dev, "Tx Ring Dump\n");
270 
271 	/* Transmit Descriptor Formats - DEXT[29] is 0 (Legacy) or 1 (Extended)
272 	 *
273 	 * Legacy Transmit Descriptor
274 	 *   +--------------------------------------------------------------+
275 	 * 0 |         Buffer Address [63:0] (Reserved on Write Back)       |
276 	 *   +--------------------------------------------------------------+
277 	 * 8 | Special  |    CSS     | Status |  CMD    |  CSO   |  Length  |
278 	 *   +--------------------------------------------------------------+
279 	 *   63       48 47        36 35    32 31     24 23    16 15        0
280 	 *
281 	 * Extended Context Descriptor (DTYP=0x0) for TSO or checksum offload
282 	 *   63      48 47    40 39       32 31             16 15    8 7      0
283 	 *   +----------------------------------------------------------------+
284 	 * 0 |  TUCSE  | TUCS0  |   TUCSS   |     IPCSE       | IPCS0 | IPCSS |
285 	 *   +----------------------------------------------------------------+
286 	 * 8 |   MSS   | HDRLEN | RSV | STA | TUCMD | DTYP |      PAYLEN      |
287 	 *   +----------------------------------------------------------------+
288 	 *   63      48 47    40 39 36 35 32 31   24 23  20 19                0
289 	 *
290 	 * Extended Data Descriptor (DTYP=0x1)
291 	 *   +----------------------------------------------------------------+
292 	 * 0 |                     Buffer Address [63:0]                      |
293 	 *   +----------------------------------------------------------------+
294 	 * 8 | VLAN tag |  POPTS  | Rsvd | Status | Command | DTYP |  DTALEN  |
295 	 *   +----------------------------------------------------------------+
296 	 *   63       48 47     40 39  36 35    32 31     24 23  20 19        0
297 	 */
298 	pr_info("Tl[desc]     [address 63:0  ] [SpeCssSCmCsLen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Legacy format\n");
299 	pr_info("Tc[desc]     [Ce CoCsIpceCoS] [MssHlRSCm0Plen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Context format\n");
300 	pr_info("Td[desc]     [address 63:0  ] [VlaPoRSCm1Dlen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Data format\n");
301 	for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
302 		const char *next_desc;
303 		tx_desc = E1000_TX_DESC(*tx_ring, i);
304 		buffer_info = &tx_ring->buffer_info[i];
305 		u0 = (struct my_u0 *)tx_desc;
306 		if (i == tx_ring->next_to_use && i == tx_ring->next_to_clean)
307 			next_desc = " NTC/U";
308 		else if (i == tx_ring->next_to_use)
309 			next_desc = " NTU";
310 		else if (i == tx_ring->next_to_clean)
311 			next_desc = " NTC";
312 		else
313 			next_desc = "";
314 		pr_info("T%c[0x%03X]    %016llX %016llX %016llX %04X  %3X %016llX %p%s\n",
315 			(!(le64_to_cpu(u0->b) & BIT(29)) ? 'l' :
316 			 ((le64_to_cpu(u0->b) & BIT(20)) ? 'd' : 'c')),
317 			i,
318 			(unsigned long long)le64_to_cpu(u0->a),
319 			(unsigned long long)le64_to_cpu(u0->b),
320 			(unsigned long long)buffer_info->dma,
321 			buffer_info->length, buffer_info->next_to_watch,
322 			(unsigned long long)buffer_info->time_stamp,
323 			buffer_info->skb, next_desc);
324 
325 		if (netif_msg_pktdata(adapter) && buffer_info->skb)
326 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
327 				       16, 1, buffer_info->skb->data,
328 				       buffer_info->skb->len, true);
329 	}
330 
331 	/* Print Rx Ring Summary */
332 rx_ring_summary:
333 	dev_info(&adapter->pdev->dev, "Rx Ring Summary\n");
334 	pr_info("Queue [NTU] [NTC]\n");
335 	pr_info(" %5d %5X %5X\n",
336 		0, rx_ring->next_to_use, rx_ring->next_to_clean);
337 
338 	/* Print Rx Ring */
339 	if (!netif_msg_rx_status(adapter))
340 		return;
341 
342 	dev_info(&adapter->pdev->dev, "Rx Ring Dump\n");
343 	switch (adapter->rx_ps_pages) {
344 	case 1:
345 	case 2:
346 	case 3:
347 		/* [Extended] Packet Split Receive Descriptor Format
348 		 *
349 		 *    +-----------------------------------------------------+
350 		 *  0 |                Buffer Address 0 [63:0]              |
351 		 *    +-----------------------------------------------------+
352 		 *  8 |                Buffer Address 1 [63:0]              |
353 		 *    +-----------------------------------------------------+
354 		 * 16 |                Buffer Address 2 [63:0]              |
355 		 *    +-----------------------------------------------------+
356 		 * 24 |                Buffer Address 3 [63:0]              |
357 		 *    +-----------------------------------------------------+
358 		 */
359 		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");
360 		/* [Extended] Receive Descriptor (Write-Back) Format
361 		 *
362 		 *   63       48 47    32 31     13 12    8 7    4 3        0
363 		 *   +------------------------------------------------------+
364 		 * 0 | Packet   | IP     |  Rsvd   | MRQ   | Rsvd | MRQ RSS |
365 		 *   | Checksum | Ident  |         | Queue |      |  Type   |
366 		 *   +------------------------------------------------------+
367 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
368 		 *   +------------------------------------------------------+
369 		 *   63       48 47    32 31            20 19               0
370 		 */
371 		pr_info("RWB[desc]      [ck ipid mrqhsh] [vl   l0 ee  es] [ l3  l2  l1 hs] [reserved      ] ---------------- [bi->skb] <-- Ext Rx Write-Back format\n");
372 		for (i = 0; i < rx_ring->count; i++) {
373 			const char *next_desc;
374 			buffer_info = &rx_ring->buffer_info[i];
375 			rx_desc_ps = E1000_RX_DESC_PS(*rx_ring, i);
376 			u1 = (struct my_u1 *)rx_desc_ps;
377 			staterr =
378 			    le32_to_cpu(rx_desc_ps->wb.middle.status_error);
379 
380 			if (i == rx_ring->next_to_use)
381 				next_desc = " NTU";
382 			else if (i == rx_ring->next_to_clean)
383 				next_desc = " NTC";
384 			else
385 				next_desc = "";
386 
387 			if (staterr & E1000_RXD_STAT_DD) {
388 				/* Descriptor Done */
389 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX ---------------- %p%s\n",
390 					"RWB", i,
391 					(unsigned long long)le64_to_cpu(u1->a),
392 					(unsigned long long)le64_to_cpu(u1->b),
393 					(unsigned long long)le64_to_cpu(u1->c),
394 					(unsigned long long)le64_to_cpu(u1->d),
395 					buffer_info->skb, next_desc);
396 			} else {
397 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX %016llX %p%s\n",
398 					"R  ", i,
399 					(unsigned long long)le64_to_cpu(u1->a),
400 					(unsigned long long)le64_to_cpu(u1->b),
401 					(unsigned long long)le64_to_cpu(u1->c),
402 					(unsigned long long)le64_to_cpu(u1->d),
403 					(unsigned long long)buffer_info->dma,
404 					buffer_info->skb, next_desc);
405 
406 				if (netif_msg_pktdata(adapter))
407 					e1000e_dump_ps_pages(adapter,
408 							     buffer_info);
409 			}
410 		}
411 		break;
412 	default:
413 	case 0:
414 		/* Extended Receive Descriptor (Read) Format
415 		 *
416 		 *   +-----------------------------------------------------+
417 		 * 0 |                Buffer Address [63:0]                |
418 		 *   +-----------------------------------------------------+
419 		 * 8 |                      Reserved                       |
420 		 *   +-----------------------------------------------------+
421 		 */
422 		pr_info("R  [desc]      [buf addr 63:0 ] [reserved 63:0 ] [bi->dma       ] [bi->skb] <-- Ext (Read) format\n");
423 		/* Extended Receive Descriptor (Write-Back) Format
424 		 *
425 		 *   63       48 47    32 31    24 23            4 3        0
426 		 *   +------------------------------------------------------+
427 		 *   |     RSS Hash      |        |               |         |
428 		 * 0 +-------------------+  Rsvd  |   Reserved    | MRQ RSS |
429 		 *   | Packet   | IP     |        |               |  Type   |
430 		 *   | Checksum | Ident  |        |               |         |
431 		 *   +------------------------------------------------------+
432 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
433 		 *   +------------------------------------------------------+
434 		 *   63       48 47    32 31            20 19               0
435 		 */
436 		pr_info("RWB[desc]      [cs ipid    mrq] [vt   ln xe  xs] [bi->skb] <-- Ext (Write-Back) format\n");
437 
438 		for (i = 0; i < rx_ring->count; i++) {
439 			const char *next_desc;
440 
441 			buffer_info = &rx_ring->buffer_info[i];
442 			rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
443 			u1 = (struct my_u1 *)rx_desc;
444 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
445 
446 			if (i == rx_ring->next_to_use)
447 				next_desc = " NTU";
448 			else if (i == rx_ring->next_to_clean)
449 				next_desc = " NTC";
450 			else
451 				next_desc = "";
452 
453 			if (staterr & E1000_RXD_STAT_DD) {
454 				/* Descriptor Done */
455 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %p%s\n",
456 					"RWB", i,
457 					(unsigned long long)le64_to_cpu(u1->a),
458 					(unsigned long long)le64_to_cpu(u1->b),
459 					buffer_info->skb, next_desc);
460 			} else {
461 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %p%s\n",
462 					"R  ", i,
463 					(unsigned long long)le64_to_cpu(u1->a),
464 					(unsigned long long)le64_to_cpu(u1->b),
465 					(unsigned long long)buffer_info->dma,
466 					buffer_info->skb, next_desc);
467 
468 				if (netif_msg_pktdata(adapter) &&
469 				    buffer_info->skb)
470 					print_hex_dump(KERN_INFO, "",
471 						       DUMP_PREFIX_ADDRESS, 16,
472 						       1,
473 						       buffer_info->skb->data,
474 						       adapter->rx_buffer_len,
475 						       true);
476 			}
477 		}
478 	}
479 }
480 
481 /**
482  * e1000_desc_unused - calculate if we have unused descriptors
483  * @ring: pointer to ring struct to perform calculation on
484  **/
485 static int e1000_desc_unused(struct e1000_ring *ring)
486 {
487 	if (ring->next_to_clean > ring->next_to_use)
488 		return ring->next_to_clean - ring->next_to_use - 1;
489 
490 	return ring->count + ring->next_to_clean - ring->next_to_use - 1;
491 }
492 
493 /**
494  * e1000e_systim_to_hwtstamp - convert system time value to hw time stamp
495  * @adapter: board private structure
496  * @hwtstamps: time stamp structure to update
497  * @systim: unsigned 64bit system time value.
498  *
499  * Convert the system time value stored in the RX/TXSTMP registers into a
500  * hwtstamp which can be used by the upper level time stamping functions.
501  *
502  * The 'systim_lock' spinlock is used to protect the consistency of the
503  * system time value. This is needed because reading the 64 bit time
504  * value involves reading two 32 bit registers. The first read latches the
505  * value.
506  **/
507 static void e1000e_systim_to_hwtstamp(struct e1000_adapter *adapter,
508 				      struct skb_shared_hwtstamps *hwtstamps,
509 				      u64 systim)
510 {
511 	u64 ns;
512 	unsigned long flags;
513 
514 	spin_lock_irqsave(&adapter->systim_lock, flags);
515 	ns = timecounter_cyc2time(&adapter->tc, systim);
516 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
517 
518 	memset(hwtstamps, 0, sizeof(*hwtstamps));
519 	hwtstamps->hwtstamp = ns_to_ktime(ns);
520 }
521 
522 /**
523  * e1000e_rx_hwtstamp - utility function which checks for Rx time stamp
524  * @adapter: board private structure
525  * @status: descriptor extended error and status field
526  * @skb: particular skb to include time stamp
527  *
528  * If the time stamp is valid, convert it into the timecounter ns value
529  * and store that result into the shhwtstamps structure which is passed
530  * up the network stack.
531  **/
532 static void e1000e_rx_hwtstamp(struct e1000_adapter *adapter, u32 status,
533 			       struct sk_buff *skb)
534 {
535 	struct e1000_hw *hw = &adapter->hw;
536 	u64 rxstmp;
537 
538 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP) ||
539 	    !(status & E1000_RXDEXT_STATERR_TST) ||
540 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID))
541 		return;
542 
543 	/* The Rx time stamp registers contain the time stamp.  No other
544 	 * received packet will be time stamped until the Rx time stamp
545 	 * registers are read.  Because only one packet can be time stamped
546 	 * at a time, the register values must belong to this packet and
547 	 * therefore none of the other additional attributes need to be
548 	 * compared.
549 	 */
550 	rxstmp = (u64)er32(RXSTMPL);
551 	rxstmp |= (u64)er32(RXSTMPH) << 32;
552 	e1000e_systim_to_hwtstamp(adapter, skb_hwtstamps(skb), rxstmp);
553 
554 	adapter->flags2 &= ~FLAG2_CHECK_RX_HWTSTAMP;
555 }
556 
557 /**
558  * e1000_receive_skb - helper function to handle Rx indications
559  * @adapter: board private structure
560  * @netdev: pointer to netdev struct
561  * @staterr: descriptor extended error and status field as written by hardware
562  * @vlan: descriptor vlan field as written by hardware (no le/be conversion)
563  * @skb: pointer to sk_buff to be indicated to stack
564  **/
565 static void e1000_receive_skb(struct e1000_adapter *adapter,
566 			      struct net_device *netdev, struct sk_buff *skb,
567 			      u32 staterr, __le16 vlan)
568 {
569 	u16 tag = le16_to_cpu(vlan);
570 
571 	e1000e_rx_hwtstamp(adapter, staterr, skb);
572 
573 	skb->protocol = eth_type_trans(skb, netdev);
574 
575 	if (staterr & E1000_RXD_STAT_VP)
576 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), tag);
577 
578 	napi_gro_receive(&adapter->napi, skb);
579 }
580 
581 /**
582  * e1000_rx_checksum - Receive Checksum Offload
583  * @adapter: board private structure
584  * @status_err: receive descriptor status and error fields
585  * @skb: socket buffer with received data
586  **/
587 static void e1000_rx_checksum(struct e1000_adapter *adapter, u32 status_err,
588 			      struct sk_buff *skb)
589 {
590 	u16 status = (u16)status_err;
591 	u8 errors = (u8)(status_err >> 24);
592 
593 	skb_checksum_none_assert(skb);
594 
595 	/* Rx checksum disabled */
596 	if (!(adapter->netdev->features & NETIF_F_RXCSUM))
597 		return;
598 
599 	/* Ignore Checksum bit is set */
600 	if (status & E1000_RXD_STAT_IXSM)
601 		return;
602 
603 	/* TCP/UDP checksum error bit or IP checksum error bit is set */
604 	if (errors & (E1000_RXD_ERR_TCPE | E1000_RXD_ERR_IPE)) {
605 		/* let the stack verify checksum errors */
606 		adapter->hw_csum_err++;
607 		return;
608 	}
609 
610 	/* TCP/UDP Checksum has not been calculated */
611 	if (!(status & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)))
612 		return;
613 
614 	/* It must be a TCP or UDP packet with a valid checksum */
615 	skb->ip_summed = CHECKSUM_UNNECESSARY;
616 	adapter->hw_csum_good++;
617 }
618 
619 static void e1000e_update_rdt_wa(struct e1000_ring *rx_ring, unsigned int i)
620 {
621 	struct e1000_adapter *adapter = rx_ring->adapter;
622 	struct e1000_hw *hw = &adapter->hw;
623 
624 	__ew32_prepare(hw);
625 	writel(i, rx_ring->tail);
626 
627 	if (unlikely(i != readl(rx_ring->tail))) {
628 		u32 rctl = er32(RCTL);
629 
630 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
631 		e_err("ME firmware caused invalid RDT - resetting\n");
632 		schedule_work(&adapter->reset_task);
633 	}
634 }
635 
636 static void e1000e_update_tdt_wa(struct e1000_ring *tx_ring, unsigned int i)
637 {
638 	struct e1000_adapter *adapter = tx_ring->adapter;
639 	struct e1000_hw *hw = &adapter->hw;
640 
641 	__ew32_prepare(hw);
642 	writel(i, tx_ring->tail);
643 
644 	if (unlikely(i != readl(tx_ring->tail))) {
645 		u32 tctl = er32(TCTL);
646 
647 		ew32(TCTL, tctl & ~E1000_TCTL_EN);
648 		e_err("ME firmware caused invalid TDT - resetting\n");
649 		schedule_work(&adapter->reset_task);
650 	}
651 }
652 
653 /**
654  * e1000_alloc_rx_buffers - Replace used receive buffers
655  * @rx_ring: Rx descriptor ring
656  * @cleaned_count: number to reallocate
657  * @gfp: flags for allocation
658  **/
659 static void e1000_alloc_rx_buffers(struct e1000_ring *rx_ring,
660 				   int cleaned_count, gfp_t gfp)
661 {
662 	struct e1000_adapter *adapter = rx_ring->adapter;
663 	struct net_device *netdev = adapter->netdev;
664 	struct pci_dev *pdev = adapter->pdev;
665 	union e1000_rx_desc_extended *rx_desc;
666 	struct e1000_buffer *buffer_info;
667 	struct sk_buff *skb;
668 	unsigned int i;
669 	unsigned int bufsz = adapter->rx_buffer_len;
670 
671 	i = rx_ring->next_to_use;
672 	buffer_info = &rx_ring->buffer_info[i];
673 
674 	while (cleaned_count--) {
675 		skb = buffer_info->skb;
676 		if (skb) {
677 			skb_trim(skb, 0);
678 			goto map_skb;
679 		}
680 
681 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
682 		if (!skb) {
683 			/* Better luck next round */
684 			adapter->alloc_rx_buff_failed++;
685 			break;
686 		}
687 
688 		buffer_info->skb = skb;
689 map_skb:
690 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
691 						  adapter->rx_buffer_len,
692 						  DMA_FROM_DEVICE);
693 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
694 			dev_err(&pdev->dev, "Rx DMA map failed\n");
695 			adapter->rx_dma_failed++;
696 			break;
697 		}
698 
699 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
700 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
701 
702 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
703 			/* Force memory writes to complete before letting h/w
704 			 * know there are new descriptors to fetch.  (Only
705 			 * applicable for weak-ordered memory model archs,
706 			 * such as IA-64).
707 			 */
708 			wmb();
709 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
710 				e1000e_update_rdt_wa(rx_ring, i);
711 			else
712 				writel(i, rx_ring->tail);
713 		}
714 		i++;
715 		if (i == rx_ring->count)
716 			i = 0;
717 		buffer_info = &rx_ring->buffer_info[i];
718 	}
719 
720 	rx_ring->next_to_use = i;
721 }
722 
723 /**
724  * e1000_alloc_rx_buffers_ps - Replace used receive buffers; packet split
725  * @rx_ring: Rx descriptor ring
726  * @cleaned_count: number to reallocate
727  * @gfp: flags for allocation
728  **/
729 static void e1000_alloc_rx_buffers_ps(struct e1000_ring *rx_ring,
730 				      int cleaned_count, gfp_t gfp)
731 {
732 	struct e1000_adapter *adapter = rx_ring->adapter;
733 	struct net_device *netdev = adapter->netdev;
734 	struct pci_dev *pdev = adapter->pdev;
735 	union e1000_rx_desc_packet_split *rx_desc;
736 	struct e1000_buffer *buffer_info;
737 	struct e1000_ps_page *ps_page;
738 	struct sk_buff *skb;
739 	unsigned int i, j;
740 
741 	i = rx_ring->next_to_use;
742 	buffer_info = &rx_ring->buffer_info[i];
743 
744 	while (cleaned_count--) {
745 		rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
746 
747 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
748 			ps_page = &buffer_info->ps_pages[j];
749 			if (j >= adapter->rx_ps_pages) {
750 				/* all unused desc entries get hw null ptr */
751 				rx_desc->read.buffer_addr[j + 1] =
752 				    ~cpu_to_le64(0);
753 				continue;
754 			}
755 			if (!ps_page->page) {
756 				ps_page->page = alloc_page(gfp);
757 				if (!ps_page->page) {
758 					adapter->alloc_rx_buff_failed++;
759 					goto no_buffers;
760 				}
761 				ps_page->dma = dma_map_page(&pdev->dev,
762 							    ps_page->page,
763 							    0, PAGE_SIZE,
764 							    DMA_FROM_DEVICE);
765 				if (dma_mapping_error(&pdev->dev,
766 						      ps_page->dma)) {
767 					dev_err(&adapter->pdev->dev,
768 						"Rx DMA page map failed\n");
769 					adapter->rx_dma_failed++;
770 					goto no_buffers;
771 				}
772 			}
773 			/* Refresh the desc even if buffer_addrs
774 			 * didn't change because each write-back
775 			 * erases this info.
776 			 */
777 			rx_desc->read.buffer_addr[j + 1] =
778 			    cpu_to_le64(ps_page->dma);
779 		}
780 
781 		skb = __netdev_alloc_skb_ip_align(netdev, adapter->rx_ps_bsize0,
782 						  gfp);
783 
784 		if (!skb) {
785 			adapter->alloc_rx_buff_failed++;
786 			break;
787 		}
788 
789 		buffer_info->skb = skb;
790 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
791 						  adapter->rx_ps_bsize0,
792 						  DMA_FROM_DEVICE);
793 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
794 			dev_err(&pdev->dev, "Rx DMA map failed\n");
795 			adapter->rx_dma_failed++;
796 			/* cleanup skb */
797 			dev_kfree_skb_any(skb);
798 			buffer_info->skb = NULL;
799 			break;
800 		}
801 
802 		rx_desc->read.buffer_addr[0] = cpu_to_le64(buffer_info->dma);
803 
804 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
805 			/* Force memory writes to complete before letting h/w
806 			 * know there are new descriptors to fetch.  (Only
807 			 * applicable for weak-ordered memory model archs,
808 			 * such as IA-64).
809 			 */
810 			wmb();
811 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
812 				e1000e_update_rdt_wa(rx_ring, i << 1);
813 			else
814 				writel(i << 1, rx_ring->tail);
815 		}
816 
817 		i++;
818 		if (i == rx_ring->count)
819 			i = 0;
820 		buffer_info = &rx_ring->buffer_info[i];
821 	}
822 
823 no_buffers:
824 	rx_ring->next_to_use = i;
825 }
826 
827 /**
828  * e1000_alloc_jumbo_rx_buffers - Replace used jumbo receive buffers
829  * @rx_ring: Rx descriptor ring
830  * @cleaned_count: number of buffers to allocate this pass
831  * @gfp: flags for allocation
832  **/
833 
834 static void e1000_alloc_jumbo_rx_buffers(struct e1000_ring *rx_ring,
835 					 int cleaned_count, gfp_t gfp)
836 {
837 	struct e1000_adapter *adapter = rx_ring->adapter;
838 	struct net_device *netdev = adapter->netdev;
839 	struct pci_dev *pdev = adapter->pdev;
840 	union e1000_rx_desc_extended *rx_desc;
841 	struct e1000_buffer *buffer_info;
842 	struct sk_buff *skb;
843 	unsigned int i;
844 	unsigned int bufsz = 256 - 16;	/* for skb_reserve */
845 
846 	i = rx_ring->next_to_use;
847 	buffer_info = &rx_ring->buffer_info[i];
848 
849 	while (cleaned_count--) {
850 		skb = buffer_info->skb;
851 		if (skb) {
852 			skb_trim(skb, 0);
853 			goto check_page;
854 		}
855 
856 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
857 		if (unlikely(!skb)) {
858 			/* Better luck next round */
859 			adapter->alloc_rx_buff_failed++;
860 			break;
861 		}
862 
863 		buffer_info->skb = skb;
864 check_page:
865 		/* allocate a new page if necessary */
866 		if (!buffer_info->page) {
867 			buffer_info->page = alloc_page(gfp);
868 			if (unlikely(!buffer_info->page)) {
869 				adapter->alloc_rx_buff_failed++;
870 				break;
871 			}
872 		}
873 
874 		if (!buffer_info->dma) {
875 			buffer_info->dma = dma_map_page(&pdev->dev,
876 							buffer_info->page, 0,
877 							PAGE_SIZE,
878 							DMA_FROM_DEVICE);
879 			if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
880 				adapter->alloc_rx_buff_failed++;
881 				break;
882 			}
883 		}
884 
885 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
886 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
887 
888 		if (unlikely(++i == rx_ring->count))
889 			i = 0;
890 		buffer_info = &rx_ring->buffer_info[i];
891 	}
892 
893 	if (likely(rx_ring->next_to_use != i)) {
894 		rx_ring->next_to_use = i;
895 		if (unlikely(i-- == 0))
896 			i = (rx_ring->count - 1);
897 
898 		/* Force memory writes to complete before letting h/w
899 		 * know there are new descriptors to fetch.  (Only
900 		 * applicable for weak-ordered memory model archs,
901 		 * such as IA-64).
902 		 */
903 		wmb();
904 		if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
905 			e1000e_update_rdt_wa(rx_ring, i);
906 		else
907 			writel(i, rx_ring->tail);
908 	}
909 }
910 
911 static inline void e1000_rx_hash(struct net_device *netdev, __le32 rss,
912 				 struct sk_buff *skb)
913 {
914 	if (netdev->features & NETIF_F_RXHASH)
915 		skb_set_hash(skb, le32_to_cpu(rss), PKT_HASH_TYPE_L3);
916 }
917 
918 /**
919  * e1000_clean_rx_irq - Send received data up the network stack
920  * @rx_ring: Rx descriptor ring
921  * @work_done: output parameter for indicating completed work
922  * @work_to_do: how many packets we can clean
923  *
924  * the return value indicates whether actual cleaning was done, there
925  * is no guarantee that everything was cleaned
926  **/
927 static bool e1000_clean_rx_irq(struct e1000_ring *rx_ring, int *work_done,
928 			       int work_to_do)
929 {
930 	struct e1000_adapter *adapter = rx_ring->adapter;
931 	struct net_device *netdev = adapter->netdev;
932 	struct pci_dev *pdev = adapter->pdev;
933 	struct e1000_hw *hw = &adapter->hw;
934 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
935 	struct e1000_buffer *buffer_info, *next_buffer;
936 	u32 length, staterr;
937 	unsigned int i;
938 	int cleaned_count = 0;
939 	bool cleaned = false;
940 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
941 
942 	i = rx_ring->next_to_clean;
943 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
944 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
945 	buffer_info = &rx_ring->buffer_info[i];
946 
947 	while (staterr & E1000_RXD_STAT_DD) {
948 		struct sk_buff *skb;
949 
950 		if (*work_done >= work_to_do)
951 			break;
952 		(*work_done)++;
953 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
954 
955 		skb = buffer_info->skb;
956 		buffer_info->skb = NULL;
957 
958 		prefetch(skb->data - NET_IP_ALIGN);
959 
960 		i++;
961 		if (i == rx_ring->count)
962 			i = 0;
963 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
964 		prefetch(next_rxd);
965 
966 		next_buffer = &rx_ring->buffer_info[i];
967 
968 		cleaned = true;
969 		cleaned_count++;
970 		dma_unmap_single(&pdev->dev, buffer_info->dma,
971 				 adapter->rx_buffer_len, DMA_FROM_DEVICE);
972 		buffer_info->dma = 0;
973 
974 		length = le16_to_cpu(rx_desc->wb.upper.length);
975 
976 		/* !EOP means multiple descriptors were used to store a single
977 		 * packet, if that's the case we need to toss it.  In fact, we
978 		 * need to toss every packet with the EOP bit clear and the
979 		 * next frame that _does_ have the EOP bit set, as it is by
980 		 * definition only a frame fragment
981 		 */
982 		if (unlikely(!(staterr & E1000_RXD_STAT_EOP)))
983 			adapter->flags2 |= FLAG2_IS_DISCARDING;
984 
985 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
986 			/* All receives must fit into a single buffer */
987 			e_dbg("Receive packet consumed multiple buffers\n");
988 			/* recycle */
989 			buffer_info->skb = skb;
990 			if (staterr & E1000_RXD_STAT_EOP)
991 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
992 			goto next_desc;
993 		}
994 
995 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
996 			     !(netdev->features & NETIF_F_RXALL))) {
997 			/* recycle */
998 			buffer_info->skb = skb;
999 			goto next_desc;
1000 		}
1001 
1002 		/* adjust length to remove Ethernet CRC */
1003 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1004 			/* If configured to store CRC, don't subtract FCS,
1005 			 * but keep the FCS bytes out of the total_rx_bytes
1006 			 * counter
1007 			 */
1008 			if (netdev->features & NETIF_F_RXFCS)
1009 				total_rx_bytes -= 4;
1010 			else
1011 				length -= 4;
1012 		}
1013 
1014 		total_rx_bytes += length;
1015 		total_rx_packets++;
1016 
1017 		/* code added for copybreak, this should improve
1018 		 * performance for small packets with large amounts
1019 		 * of reassembly being done in the stack
1020 		 */
1021 		if (length < copybreak) {
1022 			struct sk_buff *new_skb =
1023 				napi_alloc_skb(&adapter->napi, length);
1024 			if (new_skb) {
1025 				skb_copy_to_linear_data_offset(new_skb,
1026 							       -NET_IP_ALIGN,
1027 							       (skb->data -
1028 								NET_IP_ALIGN),
1029 							       (length +
1030 								NET_IP_ALIGN));
1031 				/* save the skb in buffer_info as good */
1032 				buffer_info->skb = skb;
1033 				skb = new_skb;
1034 			}
1035 			/* else just continue with the old one */
1036 		}
1037 		/* end copybreak code */
1038 		skb_put(skb, length);
1039 
1040 		/* Receive Checksum Offload */
1041 		e1000_rx_checksum(adapter, staterr, skb);
1042 
1043 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1044 
1045 		e1000_receive_skb(adapter, netdev, skb, staterr,
1046 				  rx_desc->wb.upper.vlan);
1047 
1048 next_desc:
1049 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1050 
1051 		/* return some buffers to hardware, one at a time is too slow */
1052 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1053 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1054 					      GFP_ATOMIC);
1055 			cleaned_count = 0;
1056 		}
1057 
1058 		/* use prefetched values */
1059 		rx_desc = next_rxd;
1060 		buffer_info = next_buffer;
1061 
1062 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1063 	}
1064 	rx_ring->next_to_clean = i;
1065 
1066 	cleaned_count = e1000_desc_unused(rx_ring);
1067 	if (cleaned_count)
1068 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1069 
1070 	adapter->total_rx_bytes += total_rx_bytes;
1071 	adapter->total_rx_packets += total_rx_packets;
1072 	return cleaned;
1073 }
1074 
1075 static void e1000_put_txbuf(struct e1000_ring *tx_ring,
1076 			    struct e1000_buffer *buffer_info,
1077 			    bool drop)
1078 {
1079 	struct e1000_adapter *adapter = tx_ring->adapter;
1080 
1081 	if (buffer_info->dma) {
1082 		if (buffer_info->mapped_as_page)
1083 			dma_unmap_page(&adapter->pdev->dev, buffer_info->dma,
1084 				       buffer_info->length, DMA_TO_DEVICE);
1085 		else
1086 			dma_unmap_single(&adapter->pdev->dev, buffer_info->dma,
1087 					 buffer_info->length, DMA_TO_DEVICE);
1088 		buffer_info->dma = 0;
1089 	}
1090 	if (buffer_info->skb) {
1091 		if (drop)
1092 			dev_kfree_skb_any(buffer_info->skb);
1093 		else
1094 			dev_consume_skb_any(buffer_info->skb);
1095 		buffer_info->skb = NULL;
1096 	}
1097 	buffer_info->time_stamp = 0;
1098 }
1099 
1100 static void e1000_print_hw_hang(struct work_struct *work)
1101 {
1102 	struct e1000_adapter *adapter = container_of(work,
1103 						     struct e1000_adapter,
1104 						     print_hang_task);
1105 	struct net_device *netdev = adapter->netdev;
1106 	struct e1000_ring *tx_ring = adapter->tx_ring;
1107 	unsigned int i = tx_ring->next_to_clean;
1108 	unsigned int eop = tx_ring->buffer_info[i].next_to_watch;
1109 	struct e1000_tx_desc *eop_desc = E1000_TX_DESC(*tx_ring, eop);
1110 	struct e1000_hw *hw = &adapter->hw;
1111 	u16 phy_status, phy_1000t_status, phy_ext_status;
1112 	u16 pci_status;
1113 
1114 	if (test_bit(__E1000_DOWN, &adapter->state))
1115 		return;
1116 
1117 	if (!adapter->tx_hang_recheck && (adapter->flags2 & FLAG2_DMA_BURST)) {
1118 		/* May be block on write-back, flush and detect again
1119 		 * flush pending descriptor writebacks to memory
1120 		 */
1121 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1122 		/* execute the writes immediately */
1123 		e1e_flush();
1124 		/* Due to rare timing issues, write to TIDV again to ensure
1125 		 * the write is successful
1126 		 */
1127 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1128 		/* execute the writes immediately */
1129 		e1e_flush();
1130 		adapter->tx_hang_recheck = true;
1131 		return;
1132 	}
1133 	adapter->tx_hang_recheck = false;
1134 
1135 	if (er32(TDH(0)) == er32(TDT(0))) {
1136 		e_dbg("false hang detected, ignoring\n");
1137 		return;
1138 	}
1139 
1140 	/* Real hang detected */
1141 	netif_stop_queue(netdev);
1142 
1143 	e1e_rphy(hw, MII_BMSR, &phy_status);
1144 	e1e_rphy(hw, MII_STAT1000, &phy_1000t_status);
1145 	e1e_rphy(hw, MII_ESTATUS, &phy_ext_status);
1146 
1147 	pci_read_config_word(adapter->pdev, PCI_STATUS, &pci_status);
1148 
1149 	/* detected Hardware unit hang */
1150 	e_err("Detected Hardware Unit Hang:\n"
1151 	      "  TDH                  <%x>\n"
1152 	      "  TDT                  <%x>\n"
1153 	      "  next_to_use          <%x>\n"
1154 	      "  next_to_clean        <%x>\n"
1155 	      "buffer_info[next_to_clean]:\n"
1156 	      "  time_stamp           <%lx>\n"
1157 	      "  next_to_watch        <%x>\n"
1158 	      "  jiffies              <%lx>\n"
1159 	      "  next_to_watch.status <%x>\n"
1160 	      "MAC Status             <%x>\n"
1161 	      "PHY Status             <%x>\n"
1162 	      "PHY 1000BASE-T Status  <%x>\n"
1163 	      "PHY Extended Status    <%x>\n"
1164 	      "PCI Status             <%x>\n",
1165 	      readl(tx_ring->head), readl(tx_ring->tail), tx_ring->next_to_use,
1166 	      tx_ring->next_to_clean, tx_ring->buffer_info[eop].time_stamp,
1167 	      eop, jiffies, eop_desc->upper.fields.status, er32(STATUS),
1168 	      phy_status, phy_1000t_status, phy_ext_status, pci_status);
1169 
1170 	e1000e_dump(adapter);
1171 
1172 	/* Suggest workaround for known h/w issue */
1173 	if ((hw->mac.type == e1000_pchlan) && (er32(CTRL) & E1000_CTRL_TFCE))
1174 		e_err("Try turning off Tx pause (flow control) via ethtool\n");
1175 }
1176 
1177 /**
1178  * e1000e_tx_hwtstamp_work - check for Tx time stamp
1179  * @work: pointer to work struct
1180  *
1181  * This work function polls the TSYNCTXCTL valid bit to determine when a
1182  * timestamp has been taken for the current stored skb.  The timestamp must
1183  * be for this skb because only one such packet is allowed in the queue.
1184  */
1185 static void e1000e_tx_hwtstamp_work(struct work_struct *work)
1186 {
1187 	struct e1000_adapter *adapter = container_of(work, struct e1000_adapter,
1188 						     tx_hwtstamp_work);
1189 	struct e1000_hw *hw = &adapter->hw;
1190 
1191 	if (er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID) {
1192 		struct sk_buff *skb = adapter->tx_hwtstamp_skb;
1193 		struct skb_shared_hwtstamps shhwtstamps;
1194 		u64 txstmp;
1195 
1196 		txstmp = er32(TXSTMPL);
1197 		txstmp |= (u64)er32(TXSTMPH) << 32;
1198 
1199 		e1000e_systim_to_hwtstamp(adapter, &shhwtstamps, txstmp);
1200 
1201 		/* Clear the global tx_hwtstamp_skb pointer and force writes
1202 		 * prior to notifying the stack of a Tx timestamp.
1203 		 */
1204 		adapter->tx_hwtstamp_skb = NULL;
1205 		wmb(); /* force write prior to skb_tstamp_tx */
1206 
1207 		skb_tstamp_tx(skb, &shhwtstamps);
1208 		dev_consume_skb_any(skb);
1209 	} else if (time_after(jiffies, adapter->tx_hwtstamp_start
1210 			      + adapter->tx_timeout_factor * HZ)) {
1211 		dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
1212 		adapter->tx_hwtstamp_skb = NULL;
1213 		adapter->tx_hwtstamp_timeouts++;
1214 		e_warn("clearing Tx timestamp hang\n");
1215 	} else {
1216 		/* reschedule to check later */
1217 		schedule_work(&adapter->tx_hwtstamp_work);
1218 	}
1219 }
1220 
1221 /**
1222  * e1000_clean_tx_irq - Reclaim resources after transmit completes
1223  * @tx_ring: Tx descriptor ring
1224  *
1225  * the return value indicates whether actual cleaning was done, there
1226  * is no guarantee that everything was cleaned
1227  **/
1228 static bool e1000_clean_tx_irq(struct e1000_ring *tx_ring)
1229 {
1230 	struct e1000_adapter *adapter = tx_ring->adapter;
1231 	struct net_device *netdev = adapter->netdev;
1232 	struct e1000_hw *hw = &adapter->hw;
1233 	struct e1000_tx_desc *tx_desc, *eop_desc;
1234 	struct e1000_buffer *buffer_info;
1235 	unsigned int i, eop;
1236 	unsigned int count = 0;
1237 	unsigned int total_tx_bytes = 0, total_tx_packets = 0;
1238 	unsigned int bytes_compl = 0, pkts_compl = 0;
1239 
1240 	i = tx_ring->next_to_clean;
1241 	eop = tx_ring->buffer_info[i].next_to_watch;
1242 	eop_desc = E1000_TX_DESC(*tx_ring, eop);
1243 
1244 	while ((eop_desc->upper.data & cpu_to_le32(E1000_TXD_STAT_DD)) &&
1245 	       (count < tx_ring->count)) {
1246 		bool cleaned = false;
1247 
1248 		dma_rmb();		/* read buffer_info after eop_desc */
1249 		for (; !cleaned; count++) {
1250 			tx_desc = E1000_TX_DESC(*tx_ring, i);
1251 			buffer_info = &tx_ring->buffer_info[i];
1252 			cleaned = (i == eop);
1253 
1254 			if (cleaned) {
1255 				total_tx_packets += buffer_info->segs;
1256 				total_tx_bytes += buffer_info->bytecount;
1257 				if (buffer_info->skb) {
1258 					bytes_compl += buffer_info->skb->len;
1259 					pkts_compl++;
1260 				}
1261 			}
1262 
1263 			e1000_put_txbuf(tx_ring, buffer_info, false);
1264 			tx_desc->upper.data = 0;
1265 
1266 			i++;
1267 			if (i == tx_ring->count)
1268 				i = 0;
1269 		}
1270 
1271 		if (i == tx_ring->next_to_use)
1272 			break;
1273 		eop = tx_ring->buffer_info[i].next_to_watch;
1274 		eop_desc = E1000_TX_DESC(*tx_ring, eop);
1275 	}
1276 
1277 	tx_ring->next_to_clean = i;
1278 
1279 	netdev_completed_queue(netdev, pkts_compl, bytes_compl);
1280 
1281 #define TX_WAKE_THRESHOLD 32
1282 	if (count && netif_carrier_ok(netdev) &&
1283 	    e1000_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD) {
1284 		/* Make sure that anybody stopping the queue after this
1285 		 * sees the new next_to_clean.
1286 		 */
1287 		smp_mb();
1288 
1289 		if (netif_queue_stopped(netdev) &&
1290 		    !(test_bit(__E1000_DOWN, &adapter->state))) {
1291 			netif_wake_queue(netdev);
1292 			++adapter->restart_queue;
1293 		}
1294 	}
1295 
1296 	if (adapter->detect_tx_hung) {
1297 		/* Detect a transmit hang in hardware, this serializes the
1298 		 * check with the clearing of time_stamp and movement of i
1299 		 */
1300 		adapter->detect_tx_hung = false;
1301 		if (tx_ring->buffer_info[i].time_stamp &&
1302 		    time_after(jiffies, tx_ring->buffer_info[i].time_stamp
1303 			       + (adapter->tx_timeout_factor * HZ)) &&
1304 		    !(er32(STATUS) & E1000_STATUS_TXOFF))
1305 			schedule_work(&adapter->print_hang_task);
1306 		else
1307 			adapter->tx_hang_recheck = false;
1308 	}
1309 	adapter->total_tx_bytes += total_tx_bytes;
1310 	adapter->total_tx_packets += total_tx_packets;
1311 	return count < tx_ring->count;
1312 }
1313 
1314 /**
1315  * e1000_clean_rx_irq_ps - Send received data up the network stack; packet split
1316  * @rx_ring: Rx descriptor ring
1317  * @work_done: output parameter for indicating completed work
1318  * @work_to_do: how many packets we can clean
1319  *
1320  * the return value indicates whether actual cleaning was done, there
1321  * is no guarantee that everything was cleaned
1322  **/
1323 static bool e1000_clean_rx_irq_ps(struct e1000_ring *rx_ring, int *work_done,
1324 				  int work_to_do)
1325 {
1326 	struct e1000_adapter *adapter = rx_ring->adapter;
1327 	struct e1000_hw *hw = &adapter->hw;
1328 	union e1000_rx_desc_packet_split *rx_desc, *next_rxd;
1329 	struct net_device *netdev = adapter->netdev;
1330 	struct pci_dev *pdev = adapter->pdev;
1331 	struct e1000_buffer *buffer_info, *next_buffer;
1332 	struct e1000_ps_page *ps_page;
1333 	struct sk_buff *skb;
1334 	unsigned int i, j;
1335 	u32 length, staterr;
1336 	int cleaned_count = 0;
1337 	bool cleaned = false;
1338 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1339 
1340 	i = rx_ring->next_to_clean;
1341 	rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
1342 	staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1343 	buffer_info = &rx_ring->buffer_info[i];
1344 
1345 	while (staterr & E1000_RXD_STAT_DD) {
1346 		if (*work_done >= work_to_do)
1347 			break;
1348 		(*work_done)++;
1349 		skb = buffer_info->skb;
1350 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1351 
1352 		/* in the packet split case this is header only */
1353 		prefetch(skb->data - NET_IP_ALIGN);
1354 
1355 		i++;
1356 		if (i == rx_ring->count)
1357 			i = 0;
1358 		next_rxd = E1000_RX_DESC_PS(*rx_ring, i);
1359 		prefetch(next_rxd);
1360 
1361 		next_buffer = &rx_ring->buffer_info[i];
1362 
1363 		cleaned = true;
1364 		cleaned_count++;
1365 		dma_unmap_single(&pdev->dev, buffer_info->dma,
1366 				 adapter->rx_ps_bsize0, DMA_FROM_DEVICE);
1367 		buffer_info->dma = 0;
1368 
1369 		/* see !EOP comment in other Rx routine */
1370 		if (!(staterr & E1000_RXD_STAT_EOP))
1371 			adapter->flags2 |= FLAG2_IS_DISCARDING;
1372 
1373 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
1374 			e_dbg("Packet Split buffers didn't pick up the full packet\n");
1375 			dev_kfree_skb_irq(skb);
1376 			if (staterr & E1000_RXD_STAT_EOP)
1377 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1378 			goto next_desc;
1379 		}
1380 
1381 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1382 			     !(netdev->features & NETIF_F_RXALL))) {
1383 			dev_kfree_skb_irq(skb);
1384 			goto next_desc;
1385 		}
1386 
1387 		length = le16_to_cpu(rx_desc->wb.middle.length0);
1388 
1389 		if (!length) {
1390 			e_dbg("Last part of the packet spanning multiple descriptors\n");
1391 			dev_kfree_skb_irq(skb);
1392 			goto next_desc;
1393 		}
1394 
1395 		/* Good Receive */
1396 		skb_put(skb, length);
1397 
1398 		{
1399 			/* this looks ugly, but it seems compiler issues make
1400 			 * it more efficient than reusing j
1401 			 */
1402 			int l1 = le16_to_cpu(rx_desc->wb.upper.length[0]);
1403 
1404 			/* page alloc/put takes too long and effects small
1405 			 * packet throughput, so unsplit small packets and
1406 			 * save the alloc/put
1407 			 */
1408 			if (l1 && (l1 <= copybreak) &&
1409 			    ((length + l1) <= adapter->rx_ps_bsize0)) {
1410 				ps_page = &buffer_info->ps_pages[0];
1411 
1412 				dma_sync_single_for_cpu(&pdev->dev,
1413 							ps_page->dma,
1414 							PAGE_SIZE,
1415 							DMA_FROM_DEVICE);
1416 				memcpy(skb_tail_pointer(skb),
1417 				       page_address(ps_page->page), l1);
1418 				dma_sync_single_for_device(&pdev->dev,
1419 							   ps_page->dma,
1420 							   PAGE_SIZE,
1421 							   DMA_FROM_DEVICE);
1422 
1423 				/* remove the CRC */
1424 				if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1425 					if (!(netdev->features & NETIF_F_RXFCS))
1426 						l1 -= 4;
1427 				}
1428 
1429 				skb_put(skb, l1);
1430 				goto copydone;
1431 			}	/* if */
1432 		}
1433 
1434 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1435 			length = le16_to_cpu(rx_desc->wb.upper.length[j]);
1436 			if (!length)
1437 				break;
1438 
1439 			ps_page = &buffer_info->ps_pages[j];
1440 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1441 				       DMA_FROM_DEVICE);
1442 			ps_page->dma = 0;
1443 			skb_fill_page_desc(skb, j, ps_page->page, 0, length);
1444 			ps_page->page = NULL;
1445 			skb->len += length;
1446 			skb->data_len += length;
1447 			skb->truesize += PAGE_SIZE;
1448 		}
1449 
1450 		/* strip the ethernet crc, problem is we're using pages now so
1451 		 * this whole operation can get a little cpu intensive
1452 		 */
1453 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1454 			if (!(netdev->features & NETIF_F_RXFCS))
1455 				pskb_trim(skb, skb->len - 4);
1456 		}
1457 
1458 copydone:
1459 		total_rx_bytes += skb->len;
1460 		total_rx_packets++;
1461 
1462 		e1000_rx_checksum(adapter, staterr, skb);
1463 
1464 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1465 
1466 		if (rx_desc->wb.upper.header_status &
1467 		    cpu_to_le16(E1000_RXDPS_HDRSTAT_HDRSP))
1468 			adapter->rx_hdr_split++;
1469 
1470 		e1000_receive_skb(adapter, netdev, skb, staterr,
1471 				  rx_desc->wb.middle.vlan);
1472 
1473 next_desc:
1474 		rx_desc->wb.middle.status_error &= cpu_to_le32(~0xFF);
1475 		buffer_info->skb = NULL;
1476 
1477 		/* return some buffers to hardware, one at a time is too slow */
1478 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1479 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1480 					      GFP_ATOMIC);
1481 			cleaned_count = 0;
1482 		}
1483 
1484 		/* use prefetched values */
1485 		rx_desc = next_rxd;
1486 		buffer_info = next_buffer;
1487 
1488 		staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1489 	}
1490 	rx_ring->next_to_clean = i;
1491 
1492 	cleaned_count = e1000_desc_unused(rx_ring);
1493 	if (cleaned_count)
1494 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1495 
1496 	adapter->total_rx_bytes += total_rx_bytes;
1497 	adapter->total_rx_packets += total_rx_packets;
1498 	return cleaned;
1499 }
1500 
1501 static void e1000_consume_page(struct e1000_buffer *bi, struct sk_buff *skb,
1502 			       u16 length)
1503 {
1504 	bi->page = NULL;
1505 	skb->len += length;
1506 	skb->data_len += length;
1507 	skb->truesize += PAGE_SIZE;
1508 }
1509 
1510 /**
1511  * e1000_clean_jumbo_rx_irq - Send received data up the network stack; legacy
1512  * @rx_ring: Rx descriptor ring
1513  * @work_done: output parameter for indicating completed work
1514  * @work_to_do: how many packets we can clean
1515  *
1516  * the return value indicates whether actual cleaning was done, there
1517  * is no guarantee that everything was cleaned
1518  **/
1519 static bool e1000_clean_jumbo_rx_irq(struct e1000_ring *rx_ring, int *work_done,
1520 				     int work_to_do)
1521 {
1522 	struct e1000_adapter *adapter = rx_ring->adapter;
1523 	struct net_device *netdev = adapter->netdev;
1524 	struct pci_dev *pdev = adapter->pdev;
1525 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
1526 	struct e1000_buffer *buffer_info, *next_buffer;
1527 	u32 length, staterr;
1528 	unsigned int i;
1529 	int cleaned_count = 0;
1530 	bool cleaned = false;
1531 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1532 	struct skb_shared_info *shinfo;
1533 
1534 	i = rx_ring->next_to_clean;
1535 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
1536 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1537 	buffer_info = &rx_ring->buffer_info[i];
1538 
1539 	while (staterr & E1000_RXD_STAT_DD) {
1540 		struct sk_buff *skb;
1541 
1542 		if (*work_done >= work_to_do)
1543 			break;
1544 		(*work_done)++;
1545 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1546 
1547 		skb = buffer_info->skb;
1548 		buffer_info->skb = NULL;
1549 
1550 		++i;
1551 		if (i == rx_ring->count)
1552 			i = 0;
1553 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
1554 		prefetch(next_rxd);
1555 
1556 		next_buffer = &rx_ring->buffer_info[i];
1557 
1558 		cleaned = true;
1559 		cleaned_count++;
1560 		dma_unmap_page(&pdev->dev, buffer_info->dma, PAGE_SIZE,
1561 			       DMA_FROM_DEVICE);
1562 		buffer_info->dma = 0;
1563 
1564 		length = le16_to_cpu(rx_desc->wb.upper.length);
1565 
1566 		/* errors is only valid for DD + EOP descriptors */
1567 		if (unlikely((staterr & E1000_RXD_STAT_EOP) &&
1568 			     ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1569 			      !(netdev->features & NETIF_F_RXALL)))) {
1570 			/* recycle both page and skb */
1571 			buffer_info->skb = skb;
1572 			/* an error means any chain goes out the window too */
1573 			if (rx_ring->rx_skb_top)
1574 				dev_kfree_skb_irq(rx_ring->rx_skb_top);
1575 			rx_ring->rx_skb_top = NULL;
1576 			goto next_desc;
1577 		}
1578 #define rxtop (rx_ring->rx_skb_top)
1579 		if (!(staterr & E1000_RXD_STAT_EOP)) {
1580 			/* this descriptor is only the beginning (or middle) */
1581 			if (!rxtop) {
1582 				/* this is the beginning of a chain */
1583 				rxtop = skb;
1584 				skb_fill_page_desc(rxtop, 0, buffer_info->page,
1585 						   0, length);
1586 			} else {
1587 				/* this is the middle of a chain */
1588 				shinfo = skb_shinfo(rxtop);
1589 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1590 						   buffer_info->page, 0,
1591 						   length);
1592 				/* re-use the skb, only consumed the page */
1593 				buffer_info->skb = skb;
1594 			}
1595 			e1000_consume_page(buffer_info, rxtop, length);
1596 			goto next_desc;
1597 		} else {
1598 			if (rxtop) {
1599 				/* end of the chain */
1600 				shinfo = skb_shinfo(rxtop);
1601 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1602 						   buffer_info->page, 0,
1603 						   length);
1604 				/* re-use the current skb, we only consumed the
1605 				 * page
1606 				 */
1607 				buffer_info->skb = skb;
1608 				skb = rxtop;
1609 				rxtop = NULL;
1610 				e1000_consume_page(buffer_info, skb, length);
1611 			} else {
1612 				/* no chain, got EOP, this buf is the packet
1613 				 * copybreak to save the put_page/alloc_page
1614 				 */
1615 				if (length <= copybreak &&
1616 				    skb_tailroom(skb) >= length) {
1617 					memcpy(skb_tail_pointer(skb),
1618 					       page_address(buffer_info->page),
1619 					       length);
1620 					/* re-use the page, so don't erase
1621 					 * buffer_info->page
1622 					 */
1623 					skb_put(skb, length);
1624 				} else {
1625 					skb_fill_page_desc(skb, 0,
1626 							   buffer_info->page, 0,
1627 							   length);
1628 					e1000_consume_page(buffer_info, skb,
1629 							   length);
1630 				}
1631 			}
1632 		}
1633 
1634 		/* Receive Checksum Offload */
1635 		e1000_rx_checksum(adapter, staterr, skb);
1636 
1637 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1638 
1639 		/* probably a little skewed due to removing CRC */
1640 		total_rx_bytes += skb->len;
1641 		total_rx_packets++;
1642 
1643 		/* eth type trans needs skb->data to point to something */
1644 		if (!pskb_may_pull(skb, ETH_HLEN)) {
1645 			e_err("pskb_may_pull failed.\n");
1646 			dev_kfree_skb_irq(skb);
1647 			goto next_desc;
1648 		}
1649 
1650 		e1000_receive_skb(adapter, netdev, skb, staterr,
1651 				  rx_desc->wb.upper.vlan);
1652 
1653 next_desc:
1654 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1655 
1656 		/* return some buffers to hardware, one at a time is too slow */
1657 		if (unlikely(cleaned_count >= E1000_RX_BUFFER_WRITE)) {
1658 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1659 					      GFP_ATOMIC);
1660 			cleaned_count = 0;
1661 		}
1662 
1663 		/* use prefetched values */
1664 		rx_desc = next_rxd;
1665 		buffer_info = next_buffer;
1666 
1667 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1668 	}
1669 	rx_ring->next_to_clean = i;
1670 
1671 	cleaned_count = e1000_desc_unused(rx_ring);
1672 	if (cleaned_count)
1673 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1674 
1675 	adapter->total_rx_bytes += total_rx_bytes;
1676 	adapter->total_rx_packets += total_rx_packets;
1677 	return cleaned;
1678 }
1679 
1680 /**
1681  * e1000_clean_rx_ring - Free Rx Buffers per Queue
1682  * @rx_ring: Rx descriptor ring
1683  **/
1684 static void e1000_clean_rx_ring(struct e1000_ring *rx_ring)
1685 {
1686 	struct e1000_adapter *adapter = rx_ring->adapter;
1687 	struct e1000_buffer *buffer_info;
1688 	struct e1000_ps_page *ps_page;
1689 	struct pci_dev *pdev = adapter->pdev;
1690 	unsigned int i, j;
1691 
1692 	/* Free all the Rx ring sk_buffs */
1693 	for (i = 0; i < rx_ring->count; i++) {
1694 		buffer_info = &rx_ring->buffer_info[i];
1695 		if (buffer_info->dma) {
1696 			if (adapter->clean_rx == e1000_clean_rx_irq)
1697 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1698 						 adapter->rx_buffer_len,
1699 						 DMA_FROM_DEVICE);
1700 			else if (adapter->clean_rx == e1000_clean_jumbo_rx_irq)
1701 				dma_unmap_page(&pdev->dev, buffer_info->dma,
1702 					       PAGE_SIZE, DMA_FROM_DEVICE);
1703 			else if (adapter->clean_rx == e1000_clean_rx_irq_ps)
1704 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1705 						 adapter->rx_ps_bsize0,
1706 						 DMA_FROM_DEVICE);
1707 			buffer_info->dma = 0;
1708 		}
1709 
1710 		if (buffer_info->page) {
1711 			put_page(buffer_info->page);
1712 			buffer_info->page = NULL;
1713 		}
1714 
1715 		if (buffer_info->skb) {
1716 			dev_kfree_skb(buffer_info->skb);
1717 			buffer_info->skb = NULL;
1718 		}
1719 
1720 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1721 			ps_page = &buffer_info->ps_pages[j];
1722 			if (!ps_page->page)
1723 				break;
1724 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1725 				       DMA_FROM_DEVICE);
1726 			ps_page->dma = 0;
1727 			put_page(ps_page->page);
1728 			ps_page->page = NULL;
1729 		}
1730 	}
1731 
1732 	/* there also may be some cached data from a chained receive */
1733 	if (rx_ring->rx_skb_top) {
1734 		dev_kfree_skb(rx_ring->rx_skb_top);
1735 		rx_ring->rx_skb_top = NULL;
1736 	}
1737 
1738 	/* Zero out the descriptor ring */
1739 	memset(rx_ring->desc, 0, rx_ring->size);
1740 
1741 	rx_ring->next_to_clean = 0;
1742 	rx_ring->next_to_use = 0;
1743 	adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1744 }
1745 
1746 static void e1000e_downshift_workaround(struct work_struct *work)
1747 {
1748 	struct e1000_adapter *adapter = container_of(work,
1749 						     struct e1000_adapter,
1750 						     downshift_task);
1751 
1752 	if (test_bit(__E1000_DOWN, &adapter->state))
1753 		return;
1754 
1755 	e1000e_gig_downshift_workaround_ich8lan(&adapter->hw);
1756 }
1757 
1758 /**
1759  * e1000_intr_msi - Interrupt Handler
1760  * @irq: interrupt number
1761  * @data: pointer to a network interface device structure
1762  **/
1763 static irqreturn_t e1000_intr_msi(int __always_unused irq, void *data)
1764 {
1765 	struct net_device *netdev = data;
1766 	struct e1000_adapter *adapter = netdev_priv(netdev);
1767 	struct e1000_hw *hw = &adapter->hw;
1768 	u32 icr = er32(ICR);
1769 
1770 	/* read ICR disables interrupts using IAM */
1771 	if (icr & E1000_ICR_LSC) {
1772 		hw->mac.get_link_status = true;
1773 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1774 		 * disconnect (LSC) before accessing any PHY registers
1775 		 */
1776 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1777 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1778 			schedule_work(&adapter->downshift_task);
1779 
1780 		/* 80003ES2LAN workaround-- For packet buffer work-around on
1781 		 * link down event; disable receives here in the ISR and reset
1782 		 * adapter in watchdog
1783 		 */
1784 		if (netif_carrier_ok(netdev) &&
1785 		    adapter->flags & FLAG_RX_NEEDS_RESTART) {
1786 			/* disable receives */
1787 			u32 rctl = er32(RCTL);
1788 
1789 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1790 			adapter->flags |= FLAG_RESTART_NOW;
1791 		}
1792 		/* guard against interrupt when we're going down */
1793 		if (!test_bit(__E1000_DOWN, &adapter->state))
1794 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1795 	}
1796 
1797 	/* Reset on uncorrectable ECC error */
1798 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1799 		u32 pbeccsts = er32(PBECCSTS);
1800 
1801 		adapter->corr_errors +=
1802 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1803 		adapter->uncorr_errors +=
1804 		    FIELD_GET(E1000_PBECCSTS_UNCORR_ERR_CNT_MASK, pbeccsts);
1805 
1806 		/* Do the reset outside of interrupt context */
1807 		schedule_work(&adapter->reset_task);
1808 
1809 		/* return immediately since reset is imminent */
1810 		return IRQ_HANDLED;
1811 	}
1812 
1813 	if (napi_schedule_prep(&adapter->napi)) {
1814 		adapter->total_tx_bytes = 0;
1815 		adapter->total_tx_packets = 0;
1816 		adapter->total_rx_bytes = 0;
1817 		adapter->total_rx_packets = 0;
1818 		__napi_schedule_irqoff(&adapter->napi);
1819 	}
1820 
1821 	return IRQ_HANDLED;
1822 }
1823 
1824 /**
1825  * e1000_intr - Interrupt Handler
1826  * @irq: interrupt number
1827  * @data: pointer to a network interface device structure
1828  **/
1829 static irqreturn_t e1000_intr(int __always_unused irq, void *data)
1830 {
1831 	struct net_device *netdev = data;
1832 	struct e1000_adapter *adapter = netdev_priv(netdev);
1833 	struct e1000_hw *hw = &adapter->hw;
1834 	u32 rctl, icr = er32(ICR);
1835 
1836 	if (!icr || test_bit(__E1000_DOWN, &adapter->state))
1837 		return IRQ_NONE;	/* Not our interrupt */
1838 
1839 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
1840 	 * not set, then the adapter didn't send an interrupt
1841 	 */
1842 	if (!(icr & E1000_ICR_INT_ASSERTED))
1843 		return IRQ_NONE;
1844 
1845 	/* Interrupt Auto-Mask...upon reading ICR,
1846 	 * interrupts are masked.  No need for the
1847 	 * IMC write
1848 	 */
1849 
1850 	if (icr & E1000_ICR_LSC) {
1851 		hw->mac.get_link_status = true;
1852 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1853 		 * disconnect (LSC) before accessing any PHY registers
1854 		 */
1855 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1856 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1857 			schedule_work(&adapter->downshift_task);
1858 
1859 		/* 80003ES2LAN workaround--
1860 		 * For packet buffer work-around on link down event;
1861 		 * disable receives here in the ISR and
1862 		 * reset adapter in watchdog
1863 		 */
1864 		if (netif_carrier_ok(netdev) &&
1865 		    (adapter->flags & FLAG_RX_NEEDS_RESTART)) {
1866 			/* disable receives */
1867 			rctl = er32(RCTL);
1868 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1869 			adapter->flags |= FLAG_RESTART_NOW;
1870 		}
1871 		/* guard against interrupt when we're going down */
1872 		if (!test_bit(__E1000_DOWN, &adapter->state))
1873 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1874 	}
1875 
1876 	/* Reset on uncorrectable ECC error */
1877 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1878 		u32 pbeccsts = er32(PBECCSTS);
1879 
1880 		adapter->corr_errors +=
1881 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1882 		adapter->uncorr_errors +=
1883 		    FIELD_GET(E1000_PBECCSTS_UNCORR_ERR_CNT_MASK, pbeccsts);
1884 
1885 		/* Do the reset outside of interrupt context */
1886 		schedule_work(&adapter->reset_task);
1887 
1888 		/* return immediately since reset is imminent */
1889 		return IRQ_HANDLED;
1890 	}
1891 
1892 	if (napi_schedule_prep(&adapter->napi)) {
1893 		adapter->total_tx_bytes = 0;
1894 		adapter->total_tx_packets = 0;
1895 		adapter->total_rx_bytes = 0;
1896 		adapter->total_rx_packets = 0;
1897 		__napi_schedule_irqoff(&adapter->napi);
1898 	}
1899 
1900 	return IRQ_HANDLED;
1901 }
1902 
1903 static irqreturn_t e1000_msix_other(int __always_unused irq, void *data)
1904 {
1905 	struct net_device *netdev = data;
1906 	struct e1000_adapter *adapter = netdev_priv(netdev);
1907 	struct e1000_hw *hw = &adapter->hw;
1908 	u32 icr = er32(ICR);
1909 
1910 	if (icr & adapter->eiac_mask)
1911 		ew32(ICS, (icr & adapter->eiac_mask));
1912 
1913 	if (icr & E1000_ICR_LSC) {
1914 		hw->mac.get_link_status = true;
1915 		/* guard against interrupt when we're going down */
1916 		if (!test_bit(__E1000_DOWN, &adapter->state))
1917 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1918 	}
1919 
1920 	if (!test_bit(__E1000_DOWN, &adapter->state))
1921 		ew32(IMS, E1000_IMS_OTHER | IMS_OTHER_MASK);
1922 
1923 	return IRQ_HANDLED;
1924 }
1925 
1926 static irqreturn_t e1000_intr_msix_tx(int __always_unused irq, void *data)
1927 {
1928 	struct net_device *netdev = data;
1929 	struct e1000_adapter *adapter = netdev_priv(netdev);
1930 	struct e1000_hw *hw = &adapter->hw;
1931 	struct e1000_ring *tx_ring = adapter->tx_ring;
1932 
1933 	adapter->total_tx_bytes = 0;
1934 	adapter->total_tx_packets = 0;
1935 
1936 	if (!e1000_clean_tx_irq(tx_ring))
1937 		/* Ring was not completely cleaned, so fire another interrupt */
1938 		ew32(ICS, tx_ring->ims_val);
1939 
1940 	if (!test_bit(__E1000_DOWN, &adapter->state))
1941 		ew32(IMS, adapter->tx_ring->ims_val);
1942 
1943 	return IRQ_HANDLED;
1944 }
1945 
1946 static irqreturn_t e1000_intr_msix_rx(int __always_unused irq, void *data)
1947 {
1948 	struct net_device *netdev = data;
1949 	struct e1000_adapter *adapter = netdev_priv(netdev);
1950 	struct e1000_ring *rx_ring = adapter->rx_ring;
1951 
1952 	/* Write the ITR value calculated at the end of the
1953 	 * previous interrupt.
1954 	 */
1955 	if (rx_ring->set_itr) {
1956 		u32 itr = rx_ring->itr_val ?
1957 			  1000000000 / (rx_ring->itr_val * 256) : 0;
1958 
1959 		writel(itr, rx_ring->itr_register);
1960 		rx_ring->set_itr = 0;
1961 	}
1962 
1963 	if (napi_schedule_prep(&adapter->napi)) {
1964 		adapter->total_rx_bytes = 0;
1965 		adapter->total_rx_packets = 0;
1966 		__napi_schedule_irqoff(&adapter->napi);
1967 	}
1968 	return IRQ_HANDLED;
1969 }
1970 
1971 /**
1972  * e1000_configure_msix - Configure MSI-X hardware
1973  * @adapter: board private structure
1974  *
1975  * e1000_configure_msix sets up the hardware to properly
1976  * generate MSI-X interrupts.
1977  **/
1978 static void e1000_configure_msix(struct e1000_adapter *adapter)
1979 {
1980 	struct e1000_hw *hw = &adapter->hw;
1981 	struct e1000_ring *rx_ring = adapter->rx_ring;
1982 	struct e1000_ring *tx_ring = adapter->tx_ring;
1983 	int vector = 0;
1984 	u32 ctrl_ext, ivar = 0;
1985 
1986 	adapter->eiac_mask = 0;
1987 
1988 	/* Workaround issue with spurious interrupts on 82574 in MSI-X mode */
1989 	if (hw->mac.type == e1000_82574) {
1990 		u32 rfctl = er32(RFCTL);
1991 
1992 		rfctl |= E1000_RFCTL_ACK_DIS;
1993 		ew32(RFCTL, rfctl);
1994 	}
1995 
1996 	/* Configure Rx vector */
1997 	rx_ring->ims_val = E1000_IMS_RXQ0;
1998 	adapter->eiac_mask |= rx_ring->ims_val;
1999 	if (rx_ring->itr_val)
2000 		writel(1000000000 / (rx_ring->itr_val * 256),
2001 		       rx_ring->itr_register);
2002 	else
2003 		writel(1, rx_ring->itr_register);
2004 	ivar = E1000_IVAR_INT_ALLOC_VALID | vector;
2005 
2006 	/* Configure Tx vector */
2007 	tx_ring->ims_val = E1000_IMS_TXQ0;
2008 	vector++;
2009 	if (tx_ring->itr_val)
2010 		writel(1000000000 / (tx_ring->itr_val * 256),
2011 		       tx_ring->itr_register);
2012 	else
2013 		writel(1, tx_ring->itr_register);
2014 	adapter->eiac_mask |= tx_ring->ims_val;
2015 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 8);
2016 
2017 	/* set vector for Other Causes, e.g. link changes */
2018 	vector++;
2019 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 16);
2020 	if (rx_ring->itr_val)
2021 		writel(1000000000 / (rx_ring->itr_val * 256),
2022 		       hw->hw_addr + E1000_EITR_82574(vector));
2023 	else
2024 		writel(1, hw->hw_addr + E1000_EITR_82574(vector));
2025 
2026 	/* Cause Tx interrupts on every write back */
2027 	ivar |= BIT(31);
2028 
2029 	ew32(IVAR, ivar);
2030 
2031 	/* enable MSI-X PBA support */
2032 	ctrl_ext = er32(CTRL_EXT) & ~E1000_CTRL_EXT_IAME;
2033 	ctrl_ext |= E1000_CTRL_EXT_PBA_CLR | E1000_CTRL_EXT_EIAME;
2034 	ew32(CTRL_EXT, ctrl_ext);
2035 	e1e_flush();
2036 }
2037 
2038 void e1000e_reset_interrupt_capability(struct e1000_adapter *adapter)
2039 {
2040 	if (adapter->msix_entries) {
2041 		pci_disable_msix(adapter->pdev);
2042 		kfree(adapter->msix_entries);
2043 		adapter->msix_entries = NULL;
2044 	} else if (adapter->flags & FLAG_MSI_ENABLED) {
2045 		pci_disable_msi(adapter->pdev);
2046 		adapter->flags &= ~FLAG_MSI_ENABLED;
2047 	}
2048 }
2049 
2050 /**
2051  * e1000e_set_interrupt_capability - set MSI or MSI-X if supported
2052  * @adapter: board private structure
2053  *
2054  * Attempt to configure interrupts using the best available
2055  * capabilities of the hardware and kernel.
2056  **/
2057 void e1000e_set_interrupt_capability(struct e1000_adapter *adapter)
2058 {
2059 	int err;
2060 	int i;
2061 
2062 	switch (adapter->int_mode) {
2063 	case E1000E_INT_MODE_MSIX:
2064 		if (adapter->flags & FLAG_HAS_MSIX) {
2065 			adapter->num_vectors = 3; /* RxQ0, TxQ0 and other */
2066 			adapter->msix_entries = kzalloc_objs(struct msix_entry,
2067 							     adapter->num_vectors);
2068 			if (adapter->msix_entries) {
2069 				struct e1000_adapter *a = adapter;
2070 
2071 				for (i = 0; i < adapter->num_vectors; i++)
2072 					adapter->msix_entries[i].entry = i;
2073 
2074 				err = pci_enable_msix_range(a->pdev,
2075 							    a->msix_entries,
2076 							    a->num_vectors,
2077 							    a->num_vectors);
2078 				if (err > 0)
2079 					return;
2080 			}
2081 			/* MSI-X failed, so fall through and try MSI */
2082 			e_err("Failed to initialize MSI-X interrupts.  Falling back to MSI interrupts.\n");
2083 			e1000e_reset_interrupt_capability(adapter);
2084 		}
2085 		adapter->int_mode = E1000E_INT_MODE_MSI;
2086 		fallthrough;
2087 	case E1000E_INT_MODE_MSI:
2088 		if (!pci_enable_msi(adapter->pdev)) {
2089 			adapter->flags |= FLAG_MSI_ENABLED;
2090 		} else {
2091 			adapter->int_mode = E1000E_INT_MODE_LEGACY;
2092 			e_err("Failed to initialize MSI interrupts.  Falling back to legacy interrupts.\n");
2093 		}
2094 		fallthrough;
2095 	case E1000E_INT_MODE_LEGACY:
2096 		/* Don't do anything; this is the system default */
2097 		break;
2098 	}
2099 
2100 	/* store the number of vectors being used */
2101 	adapter->num_vectors = 1;
2102 }
2103 
2104 /**
2105  * e1000_request_msix - Initialize MSI-X interrupts
2106  * @adapter: board private structure
2107  *
2108  * e1000_request_msix allocates MSI-X vectors and requests interrupts from the
2109  * kernel.
2110  **/
2111 static int e1000_request_msix(struct e1000_adapter *adapter)
2112 {
2113 	struct net_device *netdev = adapter->netdev;
2114 	int err = 0, vector = 0;
2115 
2116 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2117 		snprintf(adapter->rx_ring->name,
2118 			 sizeof(adapter->rx_ring->name) - 1,
2119 			 "%.14s-rx-0", netdev->name);
2120 	else
2121 		memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
2122 	err = request_irq(adapter->msix_entries[vector].vector,
2123 			  e1000_intr_msix_rx, 0, adapter->rx_ring->name,
2124 			  netdev);
2125 	if (err)
2126 		return err;
2127 	adapter->rx_ring->itr_register = adapter->hw.hw_addr +
2128 	    E1000_EITR_82574(vector);
2129 	adapter->rx_ring->itr_val = adapter->itr;
2130 	vector++;
2131 
2132 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2133 		snprintf(adapter->tx_ring->name,
2134 			 sizeof(adapter->tx_ring->name) - 1,
2135 			 "%.14s-tx-0", netdev->name);
2136 	else
2137 		memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
2138 	err = request_irq(adapter->msix_entries[vector].vector,
2139 			  e1000_intr_msix_tx, 0, adapter->tx_ring->name,
2140 			  netdev);
2141 	if (err)
2142 		return err;
2143 	adapter->tx_ring->itr_register = adapter->hw.hw_addr +
2144 	    E1000_EITR_82574(vector);
2145 	adapter->tx_ring->itr_val = adapter->itr;
2146 	vector++;
2147 
2148 	err = request_irq(adapter->msix_entries[vector].vector,
2149 			  e1000_msix_other, 0, netdev->name, netdev);
2150 	if (err)
2151 		return err;
2152 
2153 	e1000_configure_msix(adapter);
2154 
2155 	return 0;
2156 }
2157 
2158 /**
2159  * e1000_request_irq - initialize interrupts
2160  * @adapter: board private structure
2161  *
2162  * Attempts to configure interrupts using the best available
2163  * capabilities of the hardware and kernel.
2164  **/
2165 static int e1000_request_irq(struct e1000_adapter *adapter)
2166 {
2167 	struct net_device *netdev = adapter->netdev;
2168 	int err;
2169 
2170 	if (adapter->msix_entries) {
2171 		err = e1000_request_msix(adapter);
2172 		if (!err)
2173 			return err;
2174 		/* fall back to MSI */
2175 		e1000e_reset_interrupt_capability(adapter);
2176 		adapter->int_mode = E1000E_INT_MODE_MSI;
2177 		e1000e_set_interrupt_capability(adapter);
2178 	}
2179 	if (adapter->flags & FLAG_MSI_ENABLED) {
2180 		err = request_irq(adapter->pdev->irq, e1000_intr_msi, 0,
2181 				  netdev->name, netdev);
2182 		if (!err)
2183 			return err;
2184 
2185 		/* fall back to legacy interrupt */
2186 		e1000e_reset_interrupt_capability(adapter);
2187 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
2188 	}
2189 
2190 	err = request_irq(adapter->pdev->irq, e1000_intr, IRQF_SHARED,
2191 			  netdev->name, netdev);
2192 	if (err)
2193 		e_err("Unable to allocate interrupt, Error: %d\n", err);
2194 
2195 	return err;
2196 }
2197 
2198 static void e1000_free_irq(struct e1000_adapter *adapter)
2199 {
2200 	struct net_device *netdev = adapter->netdev;
2201 
2202 	if (adapter->msix_entries) {
2203 		int vector = 0;
2204 
2205 		free_irq(adapter->msix_entries[vector].vector, netdev);
2206 		vector++;
2207 
2208 		free_irq(adapter->msix_entries[vector].vector, netdev);
2209 		vector++;
2210 
2211 		/* Other Causes interrupt vector */
2212 		free_irq(adapter->msix_entries[vector].vector, netdev);
2213 		return;
2214 	}
2215 
2216 	free_irq(adapter->pdev->irq, netdev);
2217 }
2218 
2219 /**
2220  * e1000_irq_disable - Mask off interrupt generation on the NIC
2221  * @adapter: board private structure
2222  **/
2223 static void e1000_irq_disable(struct e1000_adapter *adapter)
2224 {
2225 	struct e1000_hw *hw = &adapter->hw;
2226 
2227 	ew32(IMC, ~0);
2228 	if (adapter->msix_entries)
2229 		ew32(EIAC_82574, 0);
2230 	e1e_flush();
2231 
2232 	if (adapter->msix_entries) {
2233 		int i;
2234 
2235 		for (i = 0; i < adapter->num_vectors; i++)
2236 			synchronize_irq(adapter->msix_entries[i].vector);
2237 	} else {
2238 		synchronize_irq(adapter->pdev->irq);
2239 	}
2240 }
2241 
2242 /**
2243  * e1000_irq_enable - Enable default interrupt generation settings
2244  * @adapter: board private structure
2245  **/
2246 static void e1000_irq_enable(struct e1000_adapter *adapter)
2247 {
2248 	struct e1000_hw *hw = &adapter->hw;
2249 
2250 	if (adapter->msix_entries) {
2251 		ew32(EIAC_82574, adapter->eiac_mask & E1000_EIAC_MASK_82574);
2252 		ew32(IMS, adapter->eiac_mask | E1000_IMS_OTHER |
2253 		     IMS_OTHER_MASK);
2254 	} else if (hw->mac.type >= e1000_pch_lpt) {
2255 		ew32(IMS, IMS_ENABLE_MASK | E1000_IMS_ECCER);
2256 	} else {
2257 		ew32(IMS, IMS_ENABLE_MASK);
2258 	}
2259 	e1e_flush();
2260 }
2261 
2262 /**
2263  * e1000e_get_hw_control - get control of the h/w from f/w
2264  * @adapter: address of board private structure
2265  *
2266  * e1000e_get_hw_control sets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2267  * For ASF and Pass Through versions of f/w this means that
2268  * the driver is loaded. For AMT version (only with 82573)
2269  * of the f/w this means that the network i/f is open.
2270  **/
2271 void e1000e_get_hw_control(struct e1000_adapter *adapter)
2272 {
2273 	struct e1000_hw *hw = &adapter->hw;
2274 	u32 ctrl_ext;
2275 	u32 swsm;
2276 
2277 	/* Let firmware know the driver has taken over */
2278 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2279 		swsm = er32(SWSM);
2280 		ew32(SWSM, swsm | E1000_SWSM_DRV_LOAD);
2281 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2282 		ctrl_ext = er32(CTRL_EXT);
2283 		ew32(CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2284 	}
2285 }
2286 
2287 /**
2288  * e1000e_release_hw_control - release control of the h/w to f/w
2289  * @adapter: address of board private structure
2290  *
2291  * e1000e_release_hw_control resets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2292  * For ASF and Pass Through versions of f/w this means that the
2293  * driver is no longer loaded. For AMT version (only with 82573) i
2294  * of the f/w this means that the network i/f is closed.
2295  *
2296  **/
2297 void e1000e_release_hw_control(struct e1000_adapter *adapter)
2298 {
2299 	struct e1000_hw *hw = &adapter->hw;
2300 	u32 ctrl_ext;
2301 	u32 swsm;
2302 
2303 	/* Let firmware taken over control of h/w */
2304 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2305 		swsm = er32(SWSM);
2306 		ew32(SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
2307 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2308 		ctrl_ext = er32(CTRL_EXT);
2309 		ew32(CTRL_EXT, ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2310 	}
2311 }
2312 
2313 /**
2314  * e1000_alloc_ring_dma - allocate memory for a ring structure
2315  * @adapter: board private structure
2316  * @ring: ring struct for which to allocate dma
2317  **/
2318 static int e1000_alloc_ring_dma(struct e1000_adapter *adapter,
2319 				struct e1000_ring *ring)
2320 {
2321 	struct pci_dev *pdev = adapter->pdev;
2322 
2323 	ring->desc = dma_alloc_coherent(&pdev->dev, ring->size, &ring->dma,
2324 					GFP_KERNEL);
2325 	if (!ring->desc)
2326 		return -ENOMEM;
2327 
2328 	return 0;
2329 }
2330 
2331 /**
2332  * e1000e_setup_tx_resources - allocate Tx resources (Descriptors)
2333  * @tx_ring: Tx descriptor ring
2334  *
2335  * Return 0 on success, negative on failure
2336  **/
2337 int e1000e_setup_tx_resources(struct e1000_ring *tx_ring)
2338 {
2339 	struct e1000_adapter *adapter = tx_ring->adapter;
2340 	int err = -ENOMEM, size;
2341 
2342 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2343 	tx_ring->buffer_info = vzalloc(size);
2344 	if (!tx_ring->buffer_info)
2345 		goto err;
2346 
2347 	/* round up to nearest 4K */
2348 	tx_ring->size = tx_ring->count * sizeof(struct e1000_tx_desc);
2349 	tx_ring->size = ALIGN(tx_ring->size, 4096);
2350 
2351 	err = e1000_alloc_ring_dma(adapter, tx_ring);
2352 	if (err)
2353 		goto err;
2354 
2355 	tx_ring->next_to_use = 0;
2356 	tx_ring->next_to_clean = 0;
2357 
2358 	return 0;
2359 err:
2360 	vfree(tx_ring->buffer_info);
2361 	e_err("Unable to allocate memory for the transmit descriptor ring\n");
2362 	return err;
2363 }
2364 
2365 /**
2366  * e1000e_setup_rx_resources - allocate Rx resources (Descriptors)
2367  * @rx_ring: Rx descriptor ring
2368  *
2369  * Returns 0 on success, negative on failure
2370  **/
2371 int e1000e_setup_rx_resources(struct e1000_ring *rx_ring)
2372 {
2373 	struct e1000_adapter *adapter = rx_ring->adapter;
2374 	struct e1000_buffer *buffer_info;
2375 	int i, size, desc_len, err = -ENOMEM;
2376 
2377 	size = sizeof(struct e1000_buffer) * rx_ring->count;
2378 	rx_ring->buffer_info = vzalloc(size);
2379 	if (!rx_ring->buffer_info)
2380 		goto err;
2381 
2382 	for (i = 0; i < rx_ring->count; i++) {
2383 		buffer_info = &rx_ring->buffer_info[i];
2384 		buffer_info->ps_pages = kzalloc_objs(struct e1000_ps_page,
2385 						     PS_PAGE_BUFFERS);
2386 		if (!buffer_info->ps_pages)
2387 			goto err_pages;
2388 	}
2389 
2390 	desc_len = sizeof(union e1000_rx_desc_packet_split);
2391 
2392 	/* Round up to nearest 4K */
2393 	rx_ring->size = rx_ring->count * desc_len;
2394 	rx_ring->size = ALIGN(rx_ring->size, 4096);
2395 
2396 	err = e1000_alloc_ring_dma(adapter, rx_ring);
2397 	if (err)
2398 		goto err_pages;
2399 
2400 	rx_ring->next_to_clean = 0;
2401 	rx_ring->next_to_use = 0;
2402 	rx_ring->rx_skb_top = NULL;
2403 
2404 	return 0;
2405 
2406 err_pages:
2407 	for (i = 0; i < rx_ring->count; i++) {
2408 		buffer_info = &rx_ring->buffer_info[i];
2409 		kfree(buffer_info->ps_pages);
2410 	}
2411 err:
2412 	vfree(rx_ring->buffer_info);
2413 	e_err("Unable to allocate memory for the receive descriptor ring\n");
2414 	return err;
2415 }
2416 
2417 /**
2418  * e1000_clean_tx_ring - Free Tx Buffers
2419  * @tx_ring: Tx descriptor ring
2420  **/
2421 static void e1000_clean_tx_ring(struct e1000_ring *tx_ring)
2422 {
2423 	struct e1000_adapter *adapter = tx_ring->adapter;
2424 	struct e1000_buffer *buffer_info;
2425 	unsigned long size;
2426 	unsigned int i;
2427 
2428 	for (i = 0; i < tx_ring->count; i++) {
2429 		buffer_info = &tx_ring->buffer_info[i];
2430 		e1000_put_txbuf(tx_ring, buffer_info, false);
2431 	}
2432 
2433 	netdev_reset_queue(adapter->netdev);
2434 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2435 	memset(tx_ring->buffer_info, 0, size);
2436 
2437 	memset(tx_ring->desc, 0, tx_ring->size);
2438 
2439 	tx_ring->next_to_use = 0;
2440 	tx_ring->next_to_clean = 0;
2441 }
2442 
2443 /**
2444  * e1000e_free_tx_resources - Free Tx Resources per Queue
2445  * @tx_ring: Tx descriptor ring
2446  *
2447  * Free all transmit software resources
2448  **/
2449 void e1000e_free_tx_resources(struct e1000_ring *tx_ring)
2450 {
2451 	struct e1000_adapter *adapter = tx_ring->adapter;
2452 	struct pci_dev *pdev = adapter->pdev;
2453 
2454 	e1000_clean_tx_ring(tx_ring);
2455 
2456 	vfree(tx_ring->buffer_info);
2457 	tx_ring->buffer_info = NULL;
2458 
2459 	dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
2460 			  tx_ring->dma);
2461 	tx_ring->desc = NULL;
2462 }
2463 
2464 /**
2465  * e1000e_free_rx_resources - Free Rx Resources
2466  * @rx_ring: Rx descriptor ring
2467  *
2468  * Free all receive software resources
2469  **/
2470 void e1000e_free_rx_resources(struct e1000_ring *rx_ring)
2471 {
2472 	struct e1000_adapter *adapter = rx_ring->adapter;
2473 	struct pci_dev *pdev = adapter->pdev;
2474 	int i;
2475 
2476 	e1000_clean_rx_ring(rx_ring);
2477 
2478 	for (i = 0; i < rx_ring->count; i++)
2479 		kfree(rx_ring->buffer_info[i].ps_pages);
2480 
2481 	vfree(rx_ring->buffer_info);
2482 	rx_ring->buffer_info = NULL;
2483 
2484 	dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
2485 			  rx_ring->dma);
2486 	rx_ring->desc = NULL;
2487 }
2488 
2489 /**
2490  * e1000_update_itr - update the dynamic ITR value based on statistics
2491  * @itr_setting: current adapter->itr
2492  * @packets: the number of packets during this measurement interval
2493  * @bytes: the number of bytes during this measurement interval
2494  *
2495  *      Stores a new ITR value based on packets and byte
2496  *      counts during the last interrupt.  The advantage of per interrupt
2497  *      computation is faster updates and more accurate ITR for the current
2498  *      traffic pattern.  Constants in this function were computed
2499  *      based on theoretical maximum wire speed and thresholds were set based
2500  *      on testing data as well as attempting to minimize response time
2501  *      while increasing bulk throughput.  This functionality is controlled
2502  *      by the InterruptThrottleRate module parameter.
2503  **/
2504 static unsigned int e1000_update_itr(u16 itr_setting, int packets, int bytes)
2505 {
2506 	unsigned int retval = itr_setting;
2507 
2508 	if (packets == 0)
2509 		return itr_setting;
2510 
2511 	switch (itr_setting) {
2512 	case lowest_latency:
2513 		/* handle TSO and jumbo frames */
2514 		if (bytes / packets > 8000)
2515 			retval = bulk_latency;
2516 		else if ((packets < 5) && (bytes > 512))
2517 			retval = low_latency;
2518 		break;
2519 	case low_latency:	/* 50 usec aka 20000 ints/s */
2520 		if (bytes > 10000) {
2521 			/* this if handles the TSO accounting */
2522 			if (bytes / packets > 8000)
2523 				retval = bulk_latency;
2524 			else if ((packets < 10) || ((bytes / packets) > 1200))
2525 				retval = bulk_latency;
2526 			else if ((packets > 35))
2527 				retval = lowest_latency;
2528 		} else if (bytes / packets > 2000) {
2529 			retval = bulk_latency;
2530 		} else if (packets <= 2 && bytes < 512) {
2531 			retval = lowest_latency;
2532 		}
2533 		break;
2534 	case bulk_latency:	/* 250 usec aka 4000 ints/s */
2535 		if (bytes > 25000) {
2536 			if (packets > 35)
2537 				retval = low_latency;
2538 		} else if (bytes < 6000) {
2539 			retval = low_latency;
2540 		}
2541 		break;
2542 	}
2543 
2544 	return retval;
2545 }
2546 
2547 static void e1000_set_itr(struct e1000_adapter *adapter)
2548 {
2549 	u16 current_itr;
2550 	u32 new_itr = adapter->itr;
2551 
2552 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
2553 	if (adapter->link_speed != SPEED_1000) {
2554 		new_itr = 4000;
2555 		goto set_itr_now;
2556 	}
2557 
2558 	if (adapter->flags2 & FLAG2_DISABLE_AIM) {
2559 		new_itr = 0;
2560 		goto set_itr_now;
2561 	}
2562 
2563 	adapter->tx_itr = e1000_update_itr(adapter->tx_itr,
2564 					   adapter->total_tx_packets,
2565 					   adapter->total_tx_bytes);
2566 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2567 	if (adapter->itr_setting == 3 && adapter->tx_itr == lowest_latency)
2568 		adapter->tx_itr = low_latency;
2569 
2570 	adapter->rx_itr = e1000_update_itr(adapter->rx_itr,
2571 					   adapter->total_rx_packets,
2572 					   adapter->total_rx_bytes);
2573 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2574 	if (adapter->itr_setting == 3 && adapter->rx_itr == lowest_latency)
2575 		adapter->rx_itr = low_latency;
2576 
2577 	current_itr = max(adapter->rx_itr, adapter->tx_itr);
2578 
2579 	/* counts and packets in update_itr are dependent on these numbers */
2580 	switch (current_itr) {
2581 	case lowest_latency:
2582 		new_itr = 70000;
2583 		break;
2584 	case low_latency:
2585 		new_itr = 20000;	/* aka hwitr = ~200 */
2586 		break;
2587 	case bulk_latency:
2588 		new_itr = 4000;
2589 		break;
2590 	default:
2591 		break;
2592 	}
2593 
2594 set_itr_now:
2595 	if (new_itr != adapter->itr) {
2596 		/* this attempts to bias the interrupt rate towards Bulk
2597 		 * by adding intermediate steps when interrupt rate is
2598 		 * increasing
2599 		 */
2600 		new_itr = new_itr > adapter->itr ?
2601 		    min(adapter->itr + (new_itr >> 2), new_itr) : new_itr;
2602 		adapter->itr = new_itr;
2603 		adapter->rx_ring->itr_val = new_itr;
2604 		if (adapter->msix_entries)
2605 			adapter->rx_ring->set_itr = 1;
2606 		else
2607 			e1000e_write_itr(adapter, new_itr);
2608 	}
2609 }
2610 
2611 /**
2612  * e1000e_write_itr - write the ITR value to the appropriate registers
2613  * @adapter: address of board private structure
2614  * @itr: new ITR value to program
2615  *
2616  * e1000e_write_itr determines if the adapter is in MSI-X mode
2617  * and, if so, writes the EITR registers with the ITR value.
2618  * Otherwise, it writes the ITR value into the ITR register.
2619  **/
2620 void e1000e_write_itr(struct e1000_adapter *adapter, u32 itr)
2621 {
2622 	struct e1000_hw *hw = &adapter->hw;
2623 	u32 new_itr = itr ? 1000000000 / (itr * 256) : 0;
2624 
2625 	if (adapter->msix_entries) {
2626 		int vector;
2627 
2628 		for (vector = 0; vector < adapter->num_vectors; vector++)
2629 			writel(new_itr, hw->hw_addr + E1000_EITR_82574(vector));
2630 	} else {
2631 		ew32(ITR, new_itr);
2632 	}
2633 }
2634 
2635 /**
2636  * e1000_alloc_queues - Allocate memory for all rings
2637  * @adapter: board private structure to initialize
2638  **/
2639 static int e1000_alloc_queues(struct e1000_adapter *adapter)
2640 {
2641 	int size = sizeof(struct e1000_ring);
2642 
2643 	adapter->tx_ring = kzalloc(size, GFP_KERNEL);
2644 	if (!adapter->tx_ring)
2645 		goto err;
2646 	adapter->tx_ring->count = adapter->tx_ring_count;
2647 	adapter->tx_ring->adapter = adapter;
2648 
2649 	adapter->rx_ring = kzalloc(size, GFP_KERNEL);
2650 	if (!adapter->rx_ring)
2651 		goto err;
2652 	adapter->rx_ring->count = adapter->rx_ring_count;
2653 	adapter->rx_ring->adapter = adapter;
2654 
2655 	return 0;
2656 err:
2657 	e_err("Unable to allocate memory for queues\n");
2658 	kfree(adapter->rx_ring);
2659 	kfree(adapter->tx_ring);
2660 	return -ENOMEM;
2661 }
2662 
2663 /**
2664  * e1000e_poll - NAPI Rx polling callback
2665  * @napi: struct associated with this polling callback
2666  * @budget: number of packets driver is allowed to process this poll
2667  **/
2668 static int e1000e_poll(struct napi_struct *napi, int budget)
2669 {
2670 	struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter,
2671 						     napi);
2672 	struct e1000_hw *hw = &adapter->hw;
2673 	struct net_device *poll_dev = adapter->netdev;
2674 	int tx_cleaned = 1, work_done = 0;
2675 
2676 	adapter = netdev_priv(poll_dev);
2677 
2678 	if (!adapter->msix_entries ||
2679 	    (adapter->rx_ring->ims_val & adapter->tx_ring->ims_val))
2680 		tx_cleaned = e1000_clean_tx_irq(adapter->tx_ring);
2681 
2682 	adapter->clean_rx(adapter->rx_ring, &work_done, budget);
2683 
2684 	if (!tx_cleaned || work_done == budget)
2685 		return budget;
2686 
2687 	/* Exit the polling mode, but don't re-enable interrupts if stack might
2688 	 * poll us due to busy-polling
2689 	 */
2690 	if (likely(napi_complete_done(napi, work_done))) {
2691 		if (adapter->itr_setting & 3)
2692 			e1000_set_itr(adapter);
2693 		if (!test_bit(__E1000_DOWN, &adapter->state)) {
2694 			if (adapter->msix_entries)
2695 				ew32(IMS, adapter->rx_ring->ims_val);
2696 			else
2697 				e1000_irq_enable(adapter);
2698 		}
2699 	}
2700 
2701 	return work_done;
2702 }
2703 
2704 static int e1000_vlan_rx_add_vid(struct net_device *netdev,
2705 				 __always_unused __be16 proto, u16 vid)
2706 {
2707 	struct e1000_adapter *adapter = netdev_priv(netdev);
2708 	struct e1000_hw *hw = &adapter->hw;
2709 	u32 vfta, index;
2710 
2711 	/* don't update vlan cookie if already programmed */
2712 	if ((adapter->hw.mng_cookie.status &
2713 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2714 	    (vid == adapter->mng_vlan_id))
2715 		return 0;
2716 
2717 	/* add VID to filter table */
2718 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2719 		index = (vid >> 5) & 0x7F;
2720 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2721 		vfta |= BIT((vid & 0x1F));
2722 		hw->mac.ops.write_vfta(hw, index, vfta);
2723 	}
2724 
2725 	set_bit(vid, adapter->active_vlans);
2726 
2727 	return 0;
2728 }
2729 
2730 static int e1000_vlan_rx_kill_vid(struct net_device *netdev,
2731 				  __always_unused __be16 proto, u16 vid)
2732 {
2733 	struct e1000_adapter *adapter = netdev_priv(netdev);
2734 	struct e1000_hw *hw = &adapter->hw;
2735 	u32 vfta, index;
2736 
2737 	if ((adapter->hw.mng_cookie.status &
2738 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2739 	    (vid == adapter->mng_vlan_id)) {
2740 		/* release control to f/w */
2741 		e1000e_release_hw_control(adapter);
2742 		return 0;
2743 	}
2744 
2745 	/* remove VID from filter table */
2746 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2747 		index = (vid >> 5) & 0x7F;
2748 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2749 		vfta &= ~BIT((vid & 0x1F));
2750 		hw->mac.ops.write_vfta(hw, index, vfta);
2751 	}
2752 
2753 	clear_bit(vid, adapter->active_vlans);
2754 
2755 	return 0;
2756 }
2757 
2758 /**
2759  * e1000e_vlan_filter_disable - helper to disable hw VLAN filtering
2760  * @adapter: board private structure to initialize
2761  **/
2762 static void e1000e_vlan_filter_disable(struct e1000_adapter *adapter)
2763 {
2764 	struct net_device *netdev = adapter->netdev;
2765 	struct e1000_hw *hw = &adapter->hw;
2766 	u32 rctl;
2767 
2768 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2769 		/* disable VLAN receive filtering */
2770 		rctl = er32(RCTL);
2771 		rctl &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
2772 		ew32(RCTL, rctl);
2773 
2774 		if (adapter->mng_vlan_id != E1000_MNG_VLAN_NONE) {
2775 			e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
2776 					       adapter->mng_vlan_id);
2777 			adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
2778 		}
2779 	}
2780 }
2781 
2782 /**
2783  * e1000e_vlan_filter_enable - helper to enable HW VLAN filtering
2784  * @adapter: board private structure to initialize
2785  **/
2786 static void e1000e_vlan_filter_enable(struct e1000_adapter *adapter)
2787 {
2788 	struct e1000_hw *hw = &adapter->hw;
2789 	u32 rctl;
2790 
2791 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2792 		/* enable VLAN receive filtering */
2793 		rctl = er32(RCTL);
2794 		rctl |= E1000_RCTL_VFE;
2795 		rctl &= ~E1000_RCTL_CFIEN;
2796 		ew32(RCTL, rctl);
2797 	}
2798 }
2799 
2800 /**
2801  * e1000e_vlan_strip_disable - helper to disable HW VLAN stripping
2802  * @adapter: board private structure to initialize
2803  **/
2804 static void e1000e_vlan_strip_disable(struct e1000_adapter *adapter)
2805 {
2806 	struct e1000_hw *hw = &adapter->hw;
2807 	u32 ctrl;
2808 
2809 	/* disable VLAN tag insert/strip */
2810 	ctrl = er32(CTRL);
2811 	ctrl &= ~E1000_CTRL_VME;
2812 	ew32(CTRL, ctrl);
2813 }
2814 
2815 /**
2816  * e1000e_vlan_strip_enable - helper to enable HW VLAN stripping
2817  * @adapter: board private structure to initialize
2818  **/
2819 static void e1000e_vlan_strip_enable(struct e1000_adapter *adapter)
2820 {
2821 	struct e1000_hw *hw = &adapter->hw;
2822 	u32 ctrl;
2823 
2824 	/* enable VLAN tag insert/strip */
2825 	ctrl = er32(CTRL);
2826 	ctrl |= E1000_CTRL_VME;
2827 	ew32(CTRL, ctrl);
2828 }
2829 
2830 static void e1000_update_mng_vlan(struct e1000_adapter *adapter)
2831 {
2832 	struct net_device *netdev = adapter->netdev;
2833 	u16 vid = adapter->hw.mng_cookie.vlan_id;
2834 	u16 old_vid = adapter->mng_vlan_id;
2835 
2836 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
2837 		e1000_vlan_rx_add_vid(netdev, htons(ETH_P_8021Q), vid);
2838 		adapter->mng_vlan_id = vid;
2839 	}
2840 
2841 	if (old_vid != E1000_MNG_VLAN_NONE && vid != old_vid)
2842 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q), old_vid);
2843 }
2844 
2845 static void e1000_restore_vlan(struct e1000_adapter *adapter)
2846 {
2847 	u16 vid;
2848 
2849 	e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
2850 
2851 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
2852 	    e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
2853 }
2854 
2855 static void e1000_init_manageability_pt(struct e1000_adapter *adapter)
2856 {
2857 	struct e1000_hw *hw = &adapter->hw;
2858 	u32 manc, manc2h, mdef, i, j;
2859 
2860 	if (!(adapter->flags & FLAG_MNG_PT_ENABLED))
2861 		return;
2862 
2863 	manc = er32(MANC);
2864 
2865 	/* enable receiving management packets to the host. this will probably
2866 	 * generate destination unreachable messages from the host OS, but
2867 	 * the packets will be handled on SMBUS
2868 	 */
2869 	manc |= E1000_MANC_EN_MNG2HOST;
2870 	manc2h = er32(MANC2H);
2871 
2872 	switch (hw->mac.type) {
2873 	default:
2874 		manc2h |= (E1000_MANC2H_PORT_623 | E1000_MANC2H_PORT_664);
2875 		break;
2876 	case e1000_82574:
2877 	case e1000_82583:
2878 		/* Check if IPMI pass-through decision filter already exists;
2879 		 * if so, enable it.
2880 		 */
2881 		for (i = 0, j = 0; i < 8; i++) {
2882 			mdef = er32(MDEF(i));
2883 
2884 			/* Ignore filters with anything other than IPMI ports */
2885 			if (mdef & ~(E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2886 				continue;
2887 
2888 			/* Enable this decision filter in MANC2H */
2889 			if (mdef)
2890 				manc2h |= BIT(i);
2891 
2892 			j |= mdef;
2893 		}
2894 
2895 		if (j == (E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2896 			break;
2897 
2898 		/* Create new decision filter in an empty filter */
2899 		for (i = 0, j = 0; i < 8; i++)
2900 			if (er32(MDEF(i)) == 0) {
2901 				ew32(MDEF(i), (E1000_MDEF_PORT_623 |
2902 					       E1000_MDEF_PORT_664));
2903 				manc2h |= BIT(1);
2904 				j++;
2905 				break;
2906 			}
2907 
2908 		if (!j)
2909 			e_warn("Unable to create IPMI pass-through filter\n");
2910 		break;
2911 	}
2912 
2913 	ew32(MANC2H, manc2h);
2914 	ew32(MANC, manc);
2915 }
2916 
2917 /**
2918  * e1000_configure_tx - Configure Transmit Unit after Reset
2919  * @adapter: board private structure
2920  *
2921  * Configure the Tx unit of the MAC after a reset.
2922  **/
2923 static void e1000_configure_tx(struct e1000_adapter *adapter)
2924 {
2925 	struct e1000_hw *hw = &adapter->hw;
2926 	struct e1000_ring *tx_ring = adapter->tx_ring;
2927 	u64 tdba;
2928 	u32 tdlen, tctl, tarc;
2929 
2930 	/* Setup the HW Tx Head and Tail descriptor pointers */
2931 	tdba = tx_ring->dma;
2932 	tdlen = tx_ring->count * sizeof(struct e1000_tx_desc);
2933 	ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
2934 	ew32(TDBAH(0), (tdba >> 32));
2935 	ew32(TDLEN(0), tdlen);
2936 	ew32(TDH(0), 0);
2937 	ew32(TDT(0), 0);
2938 	tx_ring->head = adapter->hw.hw_addr + E1000_TDH(0);
2939 	tx_ring->tail = adapter->hw.hw_addr + E1000_TDT(0);
2940 
2941 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
2942 		e1000e_update_tdt_wa(tx_ring, 0);
2943 
2944 	/* Set the Tx Interrupt Delay register */
2945 	ew32(TIDV, adapter->tx_int_delay);
2946 	/* Tx irq moderation */
2947 	ew32(TADV, adapter->tx_abs_int_delay);
2948 
2949 	if (adapter->flags2 & FLAG2_DMA_BURST) {
2950 		u32 txdctl = er32(TXDCTL(0));
2951 
2952 		txdctl &= ~(E1000_TXDCTL_PTHRESH | E1000_TXDCTL_HTHRESH |
2953 			    E1000_TXDCTL_WTHRESH);
2954 		/* set up some performance related parameters to encourage the
2955 		 * hardware to use the bus more efficiently in bursts, depends
2956 		 * on the tx_int_delay to be enabled,
2957 		 * wthresh = 1 ==> burst write is disabled to avoid Tx stalls
2958 		 * hthresh = 1 ==> prefetch when one or more available
2959 		 * pthresh = 0x1f ==> prefetch if internal cache 31 or less
2960 		 * BEWARE: this seems to work but should be considered first if
2961 		 * there are Tx hangs or other Tx related bugs
2962 		 */
2963 		txdctl |= E1000_TXDCTL_DMA_BURST_ENABLE;
2964 		ew32(TXDCTL(0), txdctl);
2965 	}
2966 	/* erratum work around: set txdctl the same for both queues */
2967 	ew32(TXDCTL(1), er32(TXDCTL(0)));
2968 
2969 	/* Program the Transmit Control Register */
2970 	tctl = er32(TCTL);
2971 	tctl &= ~E1000_TCTL_CT;
2972 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
2973 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
2974 
2975 	if (adapter->flags & FLAG_TARC_SPEED_MODE_BIT) {
2976 		tarc = er32(TARC(0));
2977 		/* set the speed mode bit, we'll clear it if we're not at
2978 		 * gigabit link later
2979 		 */
2980 #define SPEED_MODE_BIT BIT(21)
2981 		tarc |= SPEED_MODE_BIT;
2982 		ew32(TARC(0), tarc);
2983 	}
2984 
2985 	/* errata: program both queues to unweighted RR */
2986 	if (adapter->flags & FLAG_TARC_SET_BIT_ZERO) {
2987 		tarc = er32(TARC(0));
2988 		tarc |= 1;
2989 		ew32(TARC(0), tarc);
2990 		tarc = er32(TARC(1));
2991 		tarc |= 1;
2992 		ew32(TARC(1), tarc);
2993 	}
2994 
2995 	/* Setup Transmit Descriptor Settings for eop descriptor */
2996 	adapter->txd_cmd = E1000_TXD_CMD_EOP | E1000_TXD_CMD_IFCS;
2997 
2998 	/* only set IDE if we are delaying interrupts using the timers */
2999 	if (adapter->tx_int_delay)
3000 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
3001 
3002 	/* enable Report Status bit */
3003 	adapter->txd_cmd |= E1000_TXD_CMD_RS;
3004 
3005 	ew32(TCTL, tctl);
3006 
3007 	hw->mac.ops.config_collision_dist(hw);
3008 
3009 	/* SPT and KBL Si errata workaround to avoid data corruption */
3010 	if (hw->mac.type == e1000_pch_spt) {
3011 		u32 reg_val;
3012 
3013 		reg_val = er32(IOSFPC);
3014 		reg_val |= E1000_RCTL_RDMTS_HEX;
3015 		ew32(IOSFPC, reg_val);
3016 
3017 		reg_val = er32(TARC(0));
3018 		/* SPT and KBL Si errata workaround to avoid Tx hang.
3019 		 * Dropping the number of outstanding requests from
3020 		 * 3 to 2 in order to avoid a buffer overrun.
3021 		 */
3022 		reg_val &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3023 		reg_val |= E1000_TARC0_CB_MULTIQ_2_REQ;
3024 		ew32(TARC(0), reg_val);
3025 	}
3026 }
3027 
3028 #define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
3029 			   (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
3030 
3031 /**
3032  * e1000_setup_rctl - configure the receive control registers
3033  * @adapter: Board private structure
3034  **/
3035 static void e1000_setup_rctl(struct e1000_adapter *adapter)
3036 {
3037 	struct e1000_hw *hw = &adapter->hw;
3038 	u32 rctl, rfctl;
3039 	u32 pages = 0;
3040 
3041 	/* Workaround Si errata on PCHx - configure jumbo frame flow.
3042 	 * If jumbo frames not set, program related MAC/PHY registers
3043 	 * to h/w defaults
3044 	 */
3045 	if (hw->mac.type >= e1000_pch2lan) {
3046 		s32 ret_val;
3047 
3048 		if (adapter->netdev->mtu > ETH_DATA_LEN)
3049 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, true);
3050 		else
3051 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, false);
3052 
3053 		if (ret_val)
3054 			e_dbg("failed to enable|disable jumbo frame workaround mode\n");
3055 	}
3056 
3057 	/* Program MC offset vector base */
3058 	rctl = er32(RCTL);
3059 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3060 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3061 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3062 	    (adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3063 
3064 	/* Do not Store bad packets */
3065 	rctl &= ~E1000_RCTL_SBP;
3066 
3067 	/* Enable Long Packet receive */
3068 	if (adapter->netdev->mtu <= ETH_DATA_LEN)
3069 		rctl &= ~E1000_RCTL_LPE;
3070 	else
3071 		rctl |= E1000_RCTL_LPE;
3072 
3073 	/* Some systems expect that the CRC is included in SMBUS traffic. The
3074 	 * hardware strips the CRC before sending to both SMBUS (BMC) and to
3075 	 * host memory when this is enabled
3076 	 */
3077 	if (adapter->flags2 & FLAG2_CRC_STRIPPING)
3078 		rctl |= E1000_RCTL_SECRC;
3079 
3080 	/* Workaround Si errata on 82577 PHY - configure IPG for jumbos */
3081 	if ((hw->phy.type == e1000_phy_82577) && (rctl & E1000_RCTL_LPE)) {
3082 		u16 phy_data;
3083 
3084 		e1e_rphy(hw, PHY_REG(770, 26), &phy_data);
3085 		phy_data &= 0xfff8;
3086 		phy_data |= BIT(2);
3087 		e1e_wphy(hw, PHY_REG(770, 26), phy_data);
3088 
3089 		e1e_rphy(hw, 22, &phy_data);
3090 		phy_data &= 0x0fff;
3091 		phy_data |= BIT(14);
3092 		e1e_wphy(hw, 0x10, 0x2823);
3093 		e1e_wphy(hw, 0x11, 0x0003);
3094 		e1e_wphy(hw, 22, phy_data);
3095 	}
3096 
3097 	/* Setup buffer sizes */
3098 	rctl &= ~E1000_RCTL_SZ_4096;
3099 	rctl |= E1000_RCTL_BSEX;
3100 	switch (adapter->rx_buffer_len) {
3101 	case 2048:
3102 	default:
3103 		rctl |= E1000_RCTL_SZ_2048;
3104 		rctl &= ~E1000_RCTL_BSEX;
3105 		break;
3106 	case 4096:
3107 		rctl |= E1000_RCTL_SZ_4096;
3108 		break;
3109 	case 8192:
3110 		rctl |= E1000_RCTL_SZ_8192;
3111 		break;
3112 	case 16384:
3113 		rctl |= E1000_RCTL_SZ_16384;
3114 		break;
3115 	}
3116 
3117 	/* Enable Extended Status in all Receive Descriptors */
3118 	rfctl = er32(RFCTL);
3119 	rfctl |= E1000_RFCTL_EXTEN;
3120 	ew32(RFCTL, rfctl);
3121 
3122 	/* 82571 and greater support packet-split where the protocol
3123 	 * header is placed in skb->data and the packet data is
3124 	 * placed in pages hanging off of skb_shinfo(skb)->nr_frags.
3125 	 * In the case of a non-split, skb->data is linearly filled,
3126 	 * followed by the page buffers.  Therefore, skb->data is
3127 	 * sized to hold the largest protocol header.
3128 	 *
3129 	 * allocations using alloc_page take too long for regular MTU
3130 	 * so only enable packet split for jumbo frames
3131 	 *
3132 	 * Using pages when the page size is greater than 16k wastes
3133 	 * a lot of memory, since we allocate 3 pages at all times
3134 	 * per packet.
3135 	 */
3136 	pages = PAGE_USE_COUNT(adapter->netdev->mtu);
3137 	if ((pages <= 3) && (PAGE_SIZE <= 16384) && (rctl & E1000_RCTL_LPE))
3138 		adapter->rx_ps_pages = pages;
3139 	else
3140 		adapter->rx_ps_pages = 0;
3141 
3142 	if (adapter->rx_ps_pages) {
3143 		u32 psrctl = 0;
3144 
3145 		/* Enable Packet split descriptors */
3146 		rctl |= E1000_RCTL_DTYP_PS;
3147 
3148 		psrctl |= adapter->rx_ps_bsize0 >> E1000_PSRCTL_BSIZE0_SHIFT;
3149 
3150 		switch (adapter->rx_ps_pages) {
3151 		case 3:
3152 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE3_SHIFT;
3153 			fallthrough;
3154 		case 2:
3155 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE2_SHIFT;
3156 			fallthrough;
3157 		case 1:
3158 			psrctl |= PAGE_SIZE >> E1000_PSRCTL_BSIZE1_SHIFT;
3159 			break;
3160 		}
3161 
3162 		ew32(PSRCTL, psrctl);
3163 	}
3164 
3165 	/* This is useful for sniffing bad packets. */
3166 	if (adapter->netdev->features & NETIF_F_RXALL) {
3167 		/* UPE and MPE will be handled by normal PROMISC logic
3168 		 * in e1000e_set_rx_mode
3169 		 */
3170 		rctl |= (E1000_RCTL_SBP |	/* Receive bad packets */
3171 			 E1000_RCTL_BAM |	/* RX All Bcast Pkts */
3172 			 E1000_RCTL_PMCF);	/* RX All MAC Ctrl Pkts */
3173 
3174 		rctl &= ~(E1000_RCTL_VFE |	/* Disable VLAN filter */
3175 			  E1000_RCTL_DPF |	/* Allow filtered pause */
3176 			  E1000_RCTL_CFIEN);	/* Dis VLAN CFIEN Filter */
3177 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3178 		 * and that breaks VLANs.
3179 		 */
3180 	}
3181 
3182 	ew32(RCTL, rctl);
3183 	/* just started the receive unit, no need to restart */
3184 	adapter->flags &= ~FLAG_RESTART_NOW;
3185 }
3186 
3187 /**
3188  * e1000_configure_rx - Configure Receive Unit after Reset
3189  * @adapter: board private structure
3190  *
3191  * Configure the Rx unit of the MAC after a reset.
3192  **/
3193 static void e1000_configure_rx(struct e1000_adapter *adapter)
3194 {
3195 	struct e1000_hw *hw = &adapter->hw;
3196 	struct e1000_ring *rx_ring = adapter->rx_ring;
3197 	u64 rdba;
3198 	u32 rdlen, rctl, rxcsum, ctrl_ext;
3199 
3200 	if (adapter->rx_ps_pages) {
3201 		/* this is a 32 byte descriptor */
3202 		rdlen = rx_ring->count *
3203 		    sizeof(union e1000_rx_desc_packet_split);
3204 		adapter->clean_rx = e1000_clean_rx_irq_ps;
3205 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers_ps;
3206 	} else if (adapter->netdev->mtu > ETH_FRAME_LEN + ETH_FCS_LEN) {
3207 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3208 		adapter->clean_rx = e1000_clean_jumbo_rx_irq;
3209 		adapter->alloc_rx_buf = e1000_alloc_jumbo_rx_buffers;
3210 	} else {
3211 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3212 		adapter->clean_rx = e1000_clean_rx_irq;
3213 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers;
3214 	}
3215 
3216 	/* disable receives while setting up the descriptors */
3217 	rctl = er32(RCTL);
3218 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
3219 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
3220 	e1e_flush();
3221 	usleep_range(10000, 11000);
3222 
3223 	if (adapter->flags2 & FLAG2_DMA_BURST) {
3224 		/* set the writeback threshold (only takes effect if the RDTR
3225 		 * is set). set GRAN=1 and write back up to 0x4 worth, and
3226 		 * enable prefetching of 0x20 Rx descriptors
3227 		 * granularity = 01
3228 		 * wthresh = 04,
3229 		 * hthresh = 04,
3230 		 * pthresh = 0x20
3231 		 */
3232 		ew32(RXDCTL(0), E1000_RXDCTL_DMA_BURST_ENABLE);
3233 		ew32(RXDCTL(1), E1000_RXDCTL_DMA_BURST_ENABLE);
3234 	}
3235 
3236 	/* set the Receive Delay Timer Register */
3237 	ew32(RDTR, adapter->rx_int_delay);
3238 
3239 	/* irq moderation */
3240 	ew32(RADV, adapter->rx_abs_int_delay);
3241 	if ((adapter->itr_setting != 0) && (adapter->itr != 0))
3242 		e1000e_write_itr(adapter, adapter->itr);
3243 
3244 	ctrl_ext = er32(CTRL_EXT);
3245 	/* Auto-Mask interrupts upon ICR access */
3246 	ctrl_ext |= E1000_CTRL_EXT_IAME;
3247 	ew32(IAM, 0xffffffff);
3248 	ew32(CTRL_EXT, ctrl_ext);
3249 	e1e_flush();
3250 
3251 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
3252 	 * the Base and Length of the Rx Descriptor Ring
3253 	 */
3254 	rdba = rx_ring->dma;
3255 	ew32(RDBAL(0), (rdba & DMA_BIT_MASK(32)));
3256 	ew32(RDBAH(0), (rdba >> 32));
3257 	ew32(RDLEN(0), rdlen);
3258 	ew32(RDH(0), 0);
3259 	ew32(RDT(0), 0);
3260 	rx_ring->head = adapter->hw.hw_addr + E1000_RDH(0);
3261 	rx_ring->tail = adapter->hw.hw_addr + E1000_RDT(0);
3262 
3263 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
3264 		e1000e_update_rdt_wa(rx_ring, 0);
3265 
3266 	/* Enable Receive Checksum Offload for TCP and UDP */
3267 	rxcsum = er32(RXCSUM);
3268 	if (adapter->netdev->features & NETIF_F_RXCSUM)
3269 		rxcsum |= E1000_RXCSUM_TUOFL;
3270 	else
3271 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3272 	ew32(RXCSUM, rxcsum);
3273 
3274 	/* With jumbo frames, excessive C-state transition latencies result
3275 	 * in dropped transactions.
3276 	 */
3277 	if (adapter->netdev->mtu > ETH_DATA_LEN) {
3278 		u32 lat =
3279 		    ((er32(PBA) & E1000_PBA_RXA_MASK) * 1024 -
3280 		     adapter->max_frame_size) * 8 / 1000;
3281 
3282 		if (adapter->flags & FLAG_IS_ICH) {
3283 			u32 rxdctl = er32(RXDCTL(0));
3284 
3285 			ew32(RXDCTL(0), rxdctl | 0x3 | BIT(8));
3286 		}
3287 
3288 		dev_info(&adapter->pdev->dev,
3289 			 "Some CPU C-states have been disabled in order to enable jumbo frames\n");
3290 		cpu_latency_qos_update_request(&adapter->pm_qos_req, lat);
3291 	} else {
3292 		cpu_latency_qos_update_request(&adapter->pm_qos_req,
3293 					       PM_QOS_DEFAULT_VALUE);
3294 	}
3295 
3296 	/* Enable Receives */
3297 	ew32(RCTL, rctl);
3298 }
3299 
3300 /**
3301  * e1000e_write_mc_addr_list - write multicast addresses to MTA
3302  * @netdev: network interface device structure
3303  *
3304  * Writes multicast address list to the MTA hash table.
3305  * Returns: -ENOMEM on failure
3306  *                0 on no addresses written
3307  *                X on writing X addresses to MTA
3308  */
3309 static int e1000e_write_mc_addr_list(struct net_device *netdev)
3310 {
3311 	struct e1000_adapter *adapter = netdev_priv(netdev);
3312 	struct e1000_hw *hw = &adapter->hw;
3313 	struct netdev_hw_addr *ha;
3314 	u8 *mta_list;
3315 	int i;
3316 
3317 	if (netdev_mc_empty(netdev)) {
3318 		/* nothing to program, so clear mc list */
3319 		hw->mac.ops.update_mc_addr_list(hw, NULL, 0);
3320 		return 0;
3321 	}
3322 
3323 	mta_list = kcalloc(netdev_mc_count(netdev), ETH_ALEN, GFP_ATOMIC);
3324 	if (!mta_list)
3325 		return -ENOMEM;
3326 
3327 	/* update_mc_addr_list expects a packed array of only addresses. */
3328 	i = 0;
3329 	netdev_for_each_mc_addr(ha, netdev)
3330 	    memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3331 
3332 	hw->mac.ops.update_mc_addr_list(hw, mta_list, i);
3333 	kfree(mta_list);
3334 
3335 	return netdev_mc_count(netdev);
3336 }
3337 
3338 /**
3339  * e1000e_write_uc_addr_list - write unicast addresses to RAR table
3340  * @netdev: network interface device structure
3341  *
3342  * Writes unicast address list to the RAR table.
3343  * Returns: -ENOMEM on failure/insufficient address space
3344  *                0 on no addresses written
3345  *                X on writing X addresses to the RAR table
3346  **/
3347 static int e1000e_write_uc_addr_list(struct net_device *netdev)
3348 {
3349 	struct e1000_adapter *adapter = netdev_priv(netdev);
3350 	struct e1000_hw *hw = &adapter->hw;
3351 	unsigned int rar_entries;
3352 	int count = 0;
3353 
3354 	rar_entries = hw->mac.ops.rar_get_count(hw);
3355 
3356 	/* save a rar entry for our hardware address */
3357 	rar_entries--;
3358 
3359 	/* save a rar entry for the LAA workaround */
3360 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA)
3361 		rar_entries--;
3362 
3363 	/* return ENOMEM indicating insufficient memory for addresses */
3364 	if (netdev_uc_count(netdev) > rar_entries)
3365 		return -ENOMEM;
3366 
3367 	if (!netdev_uc_empty(netdev) && rar_entries) {
3368 		struct netdev_hw_addr *ha;
3369 
3370 		/* write the addresses in reverse order to avoid write
3371 		 * combining
3372 		 */
3373 		netdev_for_each_uc_addr(ha, netdev) {
3374 			int ret_val;
3375 
3376 			if (!rar_entries)
3377 				break;
3378 			ret_val = hw->mac.ops.rar_set(hw, ha->addr, rar_entries--);
3379 			if (ret_val < 0)
3380 				return -ENOMEM;
3381 			count++;
3382 		}
3383 	}
3384 
3385 	/* zero out the remaining RAR entries not used above */
3386 	for (; rar_entries > 0; rar_entries--) {
3387 		ew32(RAH(rar_entries), 0);
3388 		ew32(RAL(rar_entries), 0);
3389 	}
3390 	e1e_flush();
3391 
3392 	return count;
3393 }
3394 
3395 /**
3396  * e1000e_set_rx_mode - secondary unicast, Multicast and Promiscuous mode set
3397  * @netdev: network interface device structure
3398  *
3399  * The ndo_set_rx_mode entry point is called whenever the unicast or multicast
3400  * address list or the network interface flags are updated.  This routine is
3401  * responsible for configuring the hardware for proper unicast, multicast,
3402  * promiscuous mode, and all-multi behavior.
3403  **/
3404 static void e1000e_set_rx_mode(struct net_device *netdev)
3405 {
3406 	struct e1000_adapter *adapter = netdev_priv(netdev);
3407 	struct e1000_hw *hw = &adapter->hw;
3408 	u32 rctl;
3409 
3410 	if (pm_runtime_suspended(netdev->dev.parent))
3411 		return;
3412 
3413 	/* Check for Promiscuous and All Multicast modes */
3414 	rctl = er32(RCTL);
3415 
3416 	/* clear the affected bits */
3417 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
3418 
3419 	if (netdev->flags & IFF_PROMISC) {
3420 		rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
3421 		/* Do not hardware filter VLANs in promisc mode */
3422 		e1000e_vlan_filter_disable(adapter);
3423 	} else {
3424 		int count;
3425 
3426 		if (netdev->flags & IFF_ALLMULTI) {
3427 			rctl |= E1000_RCTL_MPE;
3428 		} else {
3429 			/* Write addresses to the MTA, if the attempt fails
3430 			 * then we should just turn on promiscuous mode so
3431 			 * that we can at least receive multicast traffic
3432 			 */
3433 			count = e1000e_write_mc_addr_list(netdev);
3434 			if (count < 0)
3435 				rctl |= E1000_RCTL_MPE;
3436 		}
3437 		e1000e_vlan_filter_enable(adapter);
3438 		/* Write addresses to available RAR registers, if there is not
3439 		 * sufficient space to store all the addresses then enable
3440 		 * unicast promiscuous mode
3441 		 */
3442 		count = e1000e_write_uc_addr_list(netdev);
3443 		if (count < 0)
3444 			rctl |= E1000_RCTL_UPE;
3445 	}
3446 
3447 	ew32(RCTL, rctl);
3448 
3449 	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
3450 		e1000e_vlan_strip_enable(adapter);
3451 	else
3452 		e1000e_vlan_strip_disable(adapter);
3453 }
3454 
3455 static void e1000e_setup_rss_hash(struct e1000_adapter *adapter)
3456 {
3457 	struct e1000_hw *hw = &adapter->hw;
3458 	u32 mrqc, rxcsum;
3459 	u32 rss_key[10];
3460 	int i;
3461 
3462 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3463 	for (i = 0; i < 10; i++)
3464 		ew32(RSSRK(i), rss_key[i]);
3465 
3466 	/* Direct all traffic to queue 0 */
3467 	for (i = 0; i < 32; i++)
3468 		ew32(RETA(i), 0);
3469 
3470 	/* Disable raw packet checksumming so that RSS hash is placed in
3471 	 * descriptor on writeback.
3472 	 */
3473 	rxcsum = er32(RXCSUM);
3474 	rxcsum |= E1000_RXCSUM_PCSD;
3475 
3476 	ew32(RXCSUM, rxcsum);
3477 
3478 	mrqc = (E1000_MRQC_RSS_FIELD_IPV4 |
3479 		E1000_MRQC_RSS_FIELD_IPV4_TCP |
3480 		E1000_MRQC_RSS_FIELD_IPV6 |
3481 		E1000_MRQC_RSS_FIELD_IPV6_TCP |
3482 		E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3483 
3484 	ew32(MRQC, mrqc);
3485 }
3486 
3487 /**
3488  * e1000e_get_base_timinca - get default SYSTIM time increment attributes
3489  * @adapter: board private structure
3490  * @timinca: pointer to returned time increment attributes
3491  *
3492  * Get attributes for incrementing the System Time Register SYSTIML/H at
3493  * the default base frequency, and set the cyclecounter shift value.
3494  **/
3495 s32 e1000e_get_base_timinca(struct e1000_adapter *adapter, u32 *timinca)
3496 {
3497 	struct e1000_hw *hw = &adapter->hw;
3498 	u32 incvalue, incperiod, shift;
3499 
3500 	/* Make sure clock is enabled on I217/I218/I219  before checking
3501 	 * the frequency
3502 	 */
3503 	if ((hw->mac.type >= e1000_pch_lpt) &&
3504 	    !(er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) &&
3505 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_ENABLED)) {
3506 		u32 fextnvm7 = er32(FEXTNVM7);
3507 
3508 		if (!(fextnvm7 & BIT(0))) {
3509 			ew32(FEXTNVM7, fextnvm7 | BIT(0));
3510 			e1e_flush();
3511 		}
3512 	}
3513 
3514 	switch (hw->mac.type) {
3515 	case e1000_pch2lan:
3516 		/* Stable 96MHz frequency */
3517 		incperiod = INCPERIOD_96MHZ;
3518 		incvalue = INCVALUE_96MHZ;
3519 		shift = INCVALUE_SHIFT_96MHZ;
3520 		adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3521 		break;
3522 	case e1000_pch_lpt:
3523 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3524 			/* Stable 96MHz frequency */
3525 			incperiod = INCPERIOD_96MHZ;
3526 			incvalue = INCVALUE_96MHZ;
3527 			shift = INCVALUE_SHIFT_96MHZ;
3528 			adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3529 		} else {
3530 			/* Stable 25MHz frequency */
3531 			incperiod = INCPERIOD_25MHZ;
3532 			incvalue = INCVALUE_25MHZ;
3533 			shift = INCVALUE_SHIFT_25MHZ;
3534 			adapter->cc.shift = shift;
3535 		}
3536 		break;
3537 	case e1000_pch_spt:
3538 		/* Stable 24MHz frequency */
3539 		incperiod = INCPERIOD_24MHZ;
3540 		incvalue = INCVALUE_24MHZ;
3541 		shift = INCVALUE_SHIFT_24MHZ;
3542 		adapter->cc.shift = shift;
3543 		break;
3544 	case e1000_pch_cnp:
3545 	case e1000_pch_tgp:
3546 	case e1000_pch_adp:
3547 	case e1000_pch_nvp:
3548 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3549 			/* Stable 24MHz frequency */
3550 			incperiod = INCPERIOD_24MHZ;
3551 			incvalue = INCVALUE_24MHZ;
3552 			shift = INCVALUE_SHIFT_24MHZ;
3553 			adapter->cc.shift = shift;
3554 		} else {
3555 			/* Stable 38400KHz frequency */
3556 			incperiod = INCPERIOD_38400KHZ;
3557 			incvalue = INCVALUE_38400KHZ;
3558 			shift = INCVALUE_SHIFT_38400KHZ;
3559 			adapter->cc.shift = shift;
3560 		}
3561 		break;
3562 	case e1000_pch_mtp:
3563 	case e1000_pch_lnp:
3564 	case e1000_pch_ptp:
3565 		/* System firmware can misreport this value, so set it to a
3566 		 * stable 38400KHz frequency.
3567 		 */
3568 		incperiod = INCPERIOD_38400KHZ;
3569 		incvalue = INCVALUE_38400KHZ;
3570 		shift = INCVALUE_SHIFT_38400KHZ;
3571 		adapter->cc.shift = shift;
3572 		break;
3573 	case e1000_82574:
3574 	case e1000_82583:
3575 		/* Stable 25MHz frequency */
3576 		incperiod = INCPERIOD_25MHZ;
3577 		incvalue = INCVALUE_25MHZ;
3578 		shift = INCVALUE_SHIFT_25MHZ;
3579 		adapter->cc.shift = shift;
3580 		break;
3581 	default:
3582 		return -EINVAL;
3583 	}
3584 
3585 	*timinca = ((incperiod << E1000_TIMINCA_INCPERIOD_SHIFT) |
3586 		    ((incvalue << shift) & E1000_TIMINCA_INCVALUE_MASK));
3587 
3588 	return 0;
3589 }
3590 
3591 /**
3592  * e1000e_config_hwtstamp - configure the hwtstamp registers and enable/disable
3593  * @adapter: board private structure
3594  * @config: timestamp configuration
3595  * @extack: netlink extended ACK for error report
3596  *
3597  * Outgoing time stamping can be enabled and disabled. Play nice and
3598  * disable it when requested, although it shouldn't cause any overhead
3599  * when no packet needs it. At most one packet in the queue may be
3600  * marked for time stamping, otherwise it would be impossible to tell
3601  * for sure to which packet the hardware time stamp belongs.
3602  *
3603  * Incoming time stamping has to be configured via the hardware filters.
3604  * Not all combinations are supported, in particular event type has to be
3605  * specified. Matching the kind of event packet is not supported, with the
3606  * exception of "all V2 events regardless of level 2 or 4".
3607  **/
3608 static int e1000e_config_hwtstamp(struct e1000_adapter *adapter,
3609 				  struct kernel_hwtstamp_config *config,
3610 				  struct netlink_ext_ack *extack)
3611 {
3612 	struct e1000_hw *hw = &adapter->hw;
3613 	u32 tsync_tx_ctl = E1000_TSYNCTXCTL_ENABLED;
3614 	u32 tsync_rx_ctl = E1000_TSYNCRXCTL_ENABLED;
3615 	u32 rxmtrl = 0;
3616 	u16 rxudp = 0;
3617 	bool is_l4 = false;
3618 	bool is_l2 = false;
3619 	u32 regval;
3620 
3621 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP)) {
3622 		NL_SET_ERR_MSG(extack, "No HW timestamp support");
3623 		return -EINVAL;
3624 	}
3625 
3626 	switch (config->tx_type) {
3627 	case HWTSTAMP_TX_OFF:
3628 		tsync_tx_ctl = 0;
3629 		break;
3630 	case HWTSTAMP_TX_ON:
3631 		break;
3632 	default:
3633 		NL_SET_ERR_MSG(extack, "Unsupported TX HW timestamp type");
3634 		return -ERANGE;
3635 	}
3636 
3637 	switch (config->rx_filter) {
3638 	case HWTSTAMP_FILTER_NONE:
3639 		tsync_rx_ctl = 0;
3640 		break;
3641 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3642 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3643 		rxmtrl = E1000_RXMTRL_PTP_V1_SYNC_MESSAGE;
3644 		is_l4 = true;
3645 		break;
3646 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3647 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3648 		rxmtrl = E1000_RXMTRL_PTP_V1_DELAY_REQ_MESSAGE;
3649 		is_l4 = true;
3650 		break;
3651 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3652 		/* Also time stamps V2 L2 Path Delay Request/Response */
3653 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3654 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3655 		is_l2 = true;
3656 		break;
3657 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3658 		/* Also time stamps V2 L2 Path Delay Request/Response. */
3659 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3660 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3661 		is_l2 = true;
3662 		break;
3663 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3664 		/* Hardware cannot filter just V2 L4 Sync messages */
3665 		fallthrough;
3666 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
3667 		/* Also time stamps V2 Path Delay Request/Response. */
3668 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3669 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3670 		is_l2 = true;
3671 		is_l4 = true;
3672 		break;
3673 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3674 		/* Hardware cannot filter just V2 L4 Delay Request messages */
3675 		fallthrough;
3676 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
3677 		/* Also time stamps V2 Path Delay Request/Response. */
3678 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3679 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3680 		is_l2 = true;
3681 		is_l4 = true;
3682 		break;
3683 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
3684 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
3685 		/* Hardware cannot filter just V2 L4 or L2 Event messages */
3686 		fallthrough;
3687 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
3688 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_EVENT_V2;
3689 		config->rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
3690 		is_l2 = true;
3691 		is_l4 = true;
3692 		break;
3693 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3694 		/* For V1, the hardware can only filter Sync messages or
3695 		 * Delay Request messages but not both so fall-through to
3696 		 * time stamp all packets.
3697 		 */
3698 		fallthrough;
3699 	case HWTSTAMP_FILTER_NTP_ALL:
3700 	case HWTSTAMP_FILTER_ALL:
3701 		is_l2 = true;
3702 		is_l4 = true;
3703 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_ALL;
3704 		config->rx_filter = HWTSTAMP_FILTER_ALL;
3705 		break;
3706 	default:
3707 		NL_SET_ERR_MSG(extack, "Unsupported RX HW timestamp filter");
3708 		return -ERANGE;
3709 	}
3710 
3711 	adapter->hwtstamp_config = *config;
3712 
3713 	/* enable/disable Tx h/w time stamping */
3714 	regval = er32(TSYNCTXCTL);
3715 	regval &= ~E1000_TSYNCTXCTL_ENABLED;
3716 	regval |= tsync_tx_ctl;
3717 	ew32(TSYNCTXCTL, regval);
3718 	if ((er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) !=
3719 	    (regval & E1000_TSYNCTXCTL_ENABLED)) {
3720 		NL_SET_ERR_MSG(extack,
3721 			       "Timesync Tx Control register not set as expected");
3722 		return -EAGAIN;
3723 	}
3724 
3725 	/* enable/disable Rx h/w time stamping */
3726 	regval = er32(TSYNCRXCTL);
3727 	regval &= ~(E1000_TSYNCRXCTL_ENABLED | E1000_TSYNCRXCTL_TYPE_MASK);
3728 	regval |= tsync_rx_ctl;
3729 	ew32(TSYNCRXCTL, regval);
3730 	if ((er32(TSYNCRXCTL) & (E1000_TSYNCRXCTL_ENABLED |
3731 				 E1000_TSYNCRXCTL_TYPE_MASK)) !=
3732 	    (regval & (E1000_TSYNCRXCTL_ENABLED |
3733 		       E1000_TSYNCRXCTL_TYPE_MASK))) {
3734 		NL_SET_ERR_MSG(extack,
3735 			       "Timesync Rx Control register not set as expected");
3736 		return -EAGAIN;
3737 	}
3738 
3739 	/* L2: define ethertype filter for time stamped packets */
3740 	if (is_l2)
3741 		rxmtrl |= ETH_P_1588;
3742 
3743 	/* define which PTP packets get time stamped */
3744 	ew32(RXMTRL, rxmtrl);
3745 
3746 	/* Filter by destination port */
3747 	if (is_l4) {
3748 		rxudp = PTP_EV_PORT;
3749 		cpu_to_be16s(&rxudp);
3750 	}
3751 	ew32(RXUDP, rxudp);
3752 
3753 	e1e_flush();
3754 
3755 	/* Clear TSYNCRXCTL_VALID & TSYNCTXCTL_VALID bit */
3756 	er32(RXSTMPH);
3757 	er32(TXSTMPH);
3758 
3759 	return 0;
3760 }
3761 
3762 /**
3763  * e1000_configure - configure the hardware for Rx and Tx
3764  * @adapter: private board structure
3765  **/
3766 static void e1000_configure(struct e1000_adapter *adapter)
3767 {
3768 	struct e1000_ring *rx_ring = adapter->rx_ring;
3769 
3770 	e1000e_set_rx_mode(adapter->netdev);
3771 
3772 	e1000_restore_vlan(adapter);
3773 	e1000_init_manageability_pt(adapter);
3774 
3775 	e1000_configure_tx(adapter);
3776 
3777 	if (adapter->netdev->features & NETIF_F_RXHASH)
3778 		e1000e_setup_rss_hash(adapter);
3779 	e1000_setup_rctl(adapter);
3780 	e1000_configure_rx(adapter);
3781 	adapter->alloc_rx_buf(rx_ring, e1000_desc_unused(rx_ring), GFP_KERNEL);
3782 }
3783 
3784 /**
3785  * e1000e_power_up_phy - restore link in case the phy was powered down
3786  * @adapter: address of board private structure
3787  *
3788  * The phy may be powered down to save power and turn off link when the
3789  * driver is unloaded and wake on lan is not enabled (among others)
3790  * *** this routine MUST be followed by a call to e1000e_reset ***
3791  **/
3792 void e1000e_power_up_phy(struct e1000_adapter *adapter)
3793 {
3794 	if (adapter->hw.phy.ops.power_up)
3795 		adapter->hw.phy.ops.power_up(&adapter->hw);
3796 
3797 	adapter->hw.mac.ops.setup_link(&adapter->hw);
3798 }
3799 
3800 /**
3801  * e1000_power_down_phy - Power down the PHY
3802  * @adapter: board private structure
3803  *
3804  * Power down the PHY so no link is implied when interface is down.
3805  * The PHY cannot be powered down if management or WoL is active.
3806  */
3807 static void e1000_power_down_phy(struct e1000_adapter *adapter)
3808 {
3809 	if (adapter->hw.phy.ops.power_down)
3810 		adapter->hw.phy.ops.power_down(&adapter->hw);
3811 }
3812 
3813 /**
3814  * e1000_flush_tx_ring - remove all descriptors from the tx_ring
3815  * @adapter: board private structure
3816  *
3817  * We want to clear all pending descriptors from the TX ring.
3818  * zeroing happens when the HW reads the regs. We  assign the ring itself as
3819  * the data of the next descriptor. We don't care about the data we are about
3820  * to reset the HW.
3821  */
3822 static void e1000_flush_tx_ring(struct e1000_adapter *adapter)
3823 {
3824 	struct e1000_hw *hw = &adapter->hw;
3825 	struct e1000_ring *tx_ring = adapter->tx_ring;
3826 	struct e1000_tx_desc *tx_desc = NULL;
3827 	u32 tdt, tctl, txd_lower = E1000_TXD_CMD_IFCS;
3828 	u16 size = 512;
3829 
3830 	tctl = er32(TCTL);
3831 	ew32(TCTL, tctl | E1000_TCTL_EN);
3832 	tdt = er32(TDT(0));
3833 	BUG_ON(tdt != tx_ring->next_to_use);
3834 	tx_desc =  E1000_TX_DESC(*tx_ring, tx_ring->next_to_use);
3835 	tx_desc->buffer_addr = cpu_to_le64(tx_ring->dma);
3836 
3837 	tx_desc->lower.data = cpu_to_le32(txd_lower | size);
3838 	tx_desc->upper.data = 0;
3839 	/* flush descriptors to memory before notifying the HW */
3840 	wmb();
3841 	tx_ring->next_to_use++;
3842 	if (tx_ring->next_to_use == tx_ring->count)
3843 		tx_ring->next_to_use = 0;
3844 	ew32(TDT(0), tx_ring->next_to_use);
3845 	usleep_range(200, 250);
3846 }
3847 
3848 /**
3849  * e1000_flush_rx_ring - remove all descriptors from the rx_ring
3850  * @adapter: board private structure
3851  *
3852  * Mark all descriptors in the RX ring as consumed and disable the rx ring
3853  */
3854 static void e1000_flush_rx_ring(struct e1000_adapter *adapter)
3855 {
3856 	u32 rctl, rxdctl;
3857 	struct e1000_hw *hw = &adapter->hw;
3858 
3859 	rctl = er32(RCTL);
3860 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3861 	e1e_flush();
3862 	usleep_range(100, 150);
3863 
3864 	rxdctl = er32(RXDCTL(0));
3865 	/* zero the lower 14 bits (prefetch and host thresholds) */
3866 	rxdctl &= 0xffffc000;
3867 
3868 	/* update thresholds: prefetch threshold to 31, host threshold to 1
3869 	 * and make sure the granularity is "descriptors" and not "cache lines"
3870 	 */
3871 	rxdctl |= (0x1F | BIT(8) | E1000_RXDCTL_THRESH_UNIT_DESC);
3872 
3873 	ew32(RXDCTL(0), rxdctl);
3874 	/* momentarily enable the RX ring for the changes to take effect */
3875 	ew32(RCTL, rctl | E1000_RCTL_EN);
3876 	e1e_flush();
3877 	usleep_range(100, 150);
3878 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3879 }
3880 
3881 /**
3882  * e1000_flush_desc_rings - remove all descriptors from the descriptor rings
3883  * @adapter: board private structure
3884  *
3885  * In i219, the descriptor rings must be emptied before resetting the HW
3886  * or before changing the device state to D3 during runtime (runtime PM).
3887  *
3888  * Failure to do this will cause the HW to enter a unit hang state which can
3889  * only be released by PCI reset on the device
3890  *
3891  */
3892 
3893 static void e1000_flush_desc_rings(struct e1000_adapter *adapter)
3894 {
3895 	u16 hang_state;
3896 	u32 fext_nvm11, tdlen;
3897 	struct e1000_hw *hw = &adapter->hw;
3898 
3899 	/* First, disable MULR fix in FEXTNVM11 */
3900 	fext_nvm11 = er32(FEXTNVM11);
3901 	fext_nvm11 |= E1000_FEXTNVM11_DISABLE_MULR_FIX;
3902 	ew32(FEXTNVM11, fext_nvm11);
3903 	/* do nothing if we're not in faulty state, or if the queue is empty */
3904 	tdlen = er32(TDLEN(0));
3905 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3906 			     &hang_state);
3907 	if (!(hang_state & FLUSH_DESC_REQUIRED) || !tdlen)
3908 		return;
3909 	e1000_flush_tx_ring(adapter);
3910 	/* recheck, maybe the fault is caused by the rx ring */
3911 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3912 			     &hang_state);
3913 	if (hang_state & FLUSH_DESC_REQUIRED)
3914 		e1000_flush_rx_ring(adapter);
3915 }
3916 
3917 /**
3918  * e1000e_systim_reset - reset the timesync registers after a hardware reset
3919  * @adapter: board private structure
3920  *
3921  * When the MAC is reset, all hardware bits for timesync will be reset to the
3922  * default values. This function will restore the settings last in place.
3923  * Since the clock SYSTIME registers are reset, we will simply restore the
3924  * cyclecounter to the kernel real clock time.
3925  **/
3926 static void e1000e_systim_reset(struct e1000_adapter *adapter)
3927 {
3928 	struct ptp_clock_info *info = &adapter->ptp_clock_info;
3929 	struct e1000_hw *hw = &adapter->hw;
3930 	struct netlink_ext_ack extack = {};
3931 	unsigned long flags;
3932 	u32 timinca;
3933 	s32 ret_val;
3934 
3935 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3936 		return;
3937 
3938 	if (info->adjfine) {
3939 		/* restore the previous ptp frequency delta */
3940 		ret_val = info->adjfine(info, adapter->ptp_delta);
3941 	} else {
3942 		/* set the default base frequency if no adjustment possible */
3943 		ret_val = e1000e_get_base_timinca(adapter, &timinca);
3944 		if (!ret_val)
3945 			ew32(TIMINCA, timinca);
3946 	}
3947 
3948 	if (ret_val) {
3949 		dev_warn(&adapter->pdev->dev,
3950 			 "Failed to restore TIMINCA clock rate delta: %d\n",
3951 			 ret_val);
3952 		return;
3953 	}
3954 
3955 	/* reset the systim ns time counter */
3956 	spin_lock_irqsave(&adapter->systim_lock, flags);
3957 	timecounter_init(&adapter->tc, &adapter->cc,
3958 			 ktime_get_real_ns());
3959 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
3960 
3961 	/* restore the previous hwtstamp configuration settings */
3962 	ret_val = e1000e_config_hwtstamp(adapter, &adapter->hwtstamp_config,
3963 					 &extack);
3964 	if (ret_val) {
3965 		if (extack._msg)
3966 			e_err("%s\n", extack._msg);
3967 	}
3968 }
3969 
3970 /**
3971  * e1000e_reset - bring the hardware into a known good state
3972  * @adapter: board private structure
3973  *
3974  * This function boots the hardware and enables some settings that
3975  * require a configuration cycle of the hardware - those cannot be
3976  * set/changed during runtime. After reset the device needs to be
3977  * properly configured for Rx, Tx etc.
3978  */
3979 void e1000e_reset(struct e1000_adapter *adapter)
3980 {
3981 	struct e1000_mac_info *mac = &adapter->hw.mac;
3982 	struct e1000_fc_info *fc = &adapter->hw.fc;
3983 	struct e1000_hw *hw = &adapter->hw;
3984 	u32 tx_space, min_tx_space, min_rx_space;
3985 	u32 pba = adapter->pba;
3986 	u16 hwm;
3987 
3988 	/* reset Packet Buffer Allocation to default */
3989 	ew32(PBA, pba);
3990 
3991 	if (adapter->max_frame_size > (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN)) {
3992 		/* To maintain wire speed transmits, the Tx FIFO should be
3993 		 * large enough to accommodate two full transmit packets,
3994 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
3995 		 * the Rx FIFO should be large enough to accommodate at least
3996 		 * one full receive packet and is similarly rounded up and
3997 		 * expressed in KB.
3998 		 */
3999 		pba = er32(PBA);
4000 		/* upper 16 bits has Tx packet buffer allocation size in KB */
4001 		tx_space = pba >> 16;
4002 		/* lower 16 bits has Rx packet buffer allocation size in KB */
4003 		pba &= 0xffff;
4004 		/* the Tx fifo also stores 16 bytes of information about the Tx
4005 		 * but don't include ethernet FCS because hardware appends it
4006 		 */
4007 		min_tx_space = (adapter->max_frame_size +
4008 				sizeof(struct e1000_tx_desc) - ETH_FCS_LEN) * 2;
4009 		min_tx_space = ALIGN(min_tx_space, 1024);
4010 		min_tx_space >>= 10;
4011 		/* software strips receive CRC, so leave room for it */
4012 		min_rx_space = adapter->max_frame_size;
4013 		min_rx_space = ALIGN(min_rx_space, 1024);
4014 		min_rx_space >>= 10;
4015 
4016 		/* If current Tx allocation is less than the min Tx FIFO size,
4017 		 * and the min Tx FIFO size is less than the current Rx FIFO
4018 		 * allocation, take space away from current Rx allocation
4019 		 */
4020 		if ((tx_space < min_tx_space) &&
4021 		    ((min_tx_space - tx_space) < pba)) {
4022 			pba -= min_tx_space - tx_space;
4023 
4024 			/* if short on Rx space, Rx wins and must trump Tx
4025 			 * adjustment
4026 			 */
4027 			if (pba < min_rx_space)
4028 				pba = min_rx_space;
4029 		}
4030 
4031 		ew32(PBA, pba);
4032 	}
4033 
4034 	/* flow control settings
4035 	 *
4036 	 * The high water mark must be low enough to fit one full frame
4037 	 * (or the size used for early receive) above it in the Rx FIFO.
4038 	 * Set it to the lower of:
4039 	 * - 90% of the Rx FIFO size, and
4040 	 * - the full Rx FIFO size minus one full frame
4041 	 */
4042 	if (adapter->flags & FLAG_DISABLE_FC_PAUSE_TIME)
4043 		fc->pause_time = 0xFFFF;
4044 	else
4045 		fc->pause_time = E1000_FC_PAUSE_TIME;
4046 	fc->send_xon = true;
4047 	fc->current_mode = fc->requested_mode;
4048 
4049 	switch (hw->mac.type) {
4050 	case e1000_ich9lan:
4051 	case e1000_ich10lan:
4052 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4053 			pba = 14;
4054 			ew32(PBA, pba);
4055 			fc->high_water = 0x2800;
4056 			fc->low_water = fc->high_water - 8;
4057 			break;
4058 		}
4059 		fallthrough;
4060 	default:
4061 		hwm = min(((pba << 10) * 9 / 10),
4062 			  ((pba << 10) - adapter->max_frame_size));
4063 
4064 		fc->high_water = hwm & E1000_FCRTH_RTH;	/* 8-byte granularity */
4065 		fc->low_water = fc->high_water - 8;
4066 		break;
4067 	case e1000_pchlan:
4068 		/* Workaround PCH LOM adapter hangs with certain network
4069 		 * loads.  If hangs persist, try disabling Tx flow control.
4070 		 */
4071 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4072 			fc->high_water = 0x3500;
4073 			fc->low_water = 0x1500;
4074 		} else {
4075 			fc->high_water = 0x5000;
4076 			fc->low_water = 0x3000;
4077 		}
4078 		fc->refresh_time = 0x1000;
4079 		break;
4080 	case e1000_pch2lan:
4081 	case e1000_pch_lpt:
4082 	case e1000_pch_spt:
4083 	case e1000_pch_cnp:
4084 	case e1000_pch_tgp:
4085 	case e1000_pch_adp:
4086 	case e1000_pch_mtp:
4087 	case e1000_pch_lnp:
4088 	case e1000_pch_ptp:
4089 	case e1000_pch_nvp:
4090 		fc->refresh_time = 0xFFFF;
4091 		fc->pause_time = 0xFFFF;
4092 
4093 		if (adapter->netdev->mtu <= ETH_DATA_LEN) {
4094 			fc->high_water = 0x05C20;
4095 			fc->low_water = 0x05048;
4096 			break;
4097 		}
4098 
4099 		pba = 14;
4100 		ew32(PBA, pba);
4101 		fc->high_water = ((pba << 10) * 9 / 10) & E1000_FCRTH_RTH;
4102 		fc->low_water = ((pba << 10) * 8 / 10) & E1000_FCRTL_RTL;
4103 		break;
4104 	}
4105 
4106 	/* Alignment of Tx data is on an arbitrary byte boundary with the
4107 	 * maximum size per Tx descriptor limited only to the transmit
4108 	 * allocation of the packet buffer minus 96 bytes with an upper
4109 	 * limit of 24KB due to receive synchronization limitations.
4110 	 */
4111 	adapter->tx_fifo_limit = min_t(u32, ((er32(PBA) >> 16) << 10) - 96,
4112 				       24 << 10);
4113 
4114 	/* Disable Adaptive Interrupt Moderation if 2 full packets cannot
4115 	 * fit in receive buffer.
4116 	 */
4117 	if (adapter->itr_setting & 0x3) {
4118 		if ((adapter->max_frame_size * 2) > (pba << 10)) {
4119 			if (!(adapter->flags2 & FLAG2_DISABLE_AIM)) {
4120 				dev_info(&adapter->pdev->dev,
4121 					 "Interrupt Throttle Rate off\n");
4122 				adapter->flags2 |= FLAG2_DISABLE_AIM;
4123 				e1000e_write_itr(adapter, 0);
4124 			}
4125 		} else if (adapter->flags2 & FLAG2_DISABLE_AIM) {
4126 			dev_info(&adapter->pdev->dev,
4127 				 "Interrupt Throttle Rate on\n");
4128 			adapter->flags2 &= ~FLAG2_DISABLE_AIM;
4129 			adapter->itr = 20000;
4130 			e1000e_write_itr(adapter, adapter->itr);
4131 		}
4132 	}
4133 
4134 	if (hw->mac.type >= e1000_pch_spt)
4135 		e1000_flush_desc_rings(adapter);
4136 	/* Allow time for pending master requests to run */
4137 	mac->ops.reset_hw(hw);
4138 
4139 	/* For parts with AMT enabled, let the firmware know
4140 	 * that the network interface is in control
4141 	 */
4142 	if (adapter->flags & FLAG_HAS_AMT)
4143 		e1000e_get_hw_control(adapter);
4144 
4145 	ew32(WUC, 0);
4146 
4147 	if (mac->ops.init_hw(hw))
4148 		e_err("Hardware Error\n");
4149 
4150 	e1000_update_mng_vlan(adapter);
4151 
4152 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
4153 	ew32(VET, ETH_P_8021Q);
4154 
4155 	e1000e_reset_adaptive(hw);
4156 
4157 	/* restore systim and hwtstamp settings */
4158 	e1000e_systim_reset(adapter);
4159 
4160 	/* Set EEE advertisement as appropriate */
4161 	if (adapter->flags2 & FLAG2_HAS_EEE) {
4162 		s32 ret_val;
4163 		u16 adv_addr;
4164 
4165 		switch (hw->phy.type) {
4166 		case e1000_phy_82579:
4167 			adv_addr = I82579_EEE_ADVERTISEMENT;
4168 			break;
4169 		case e1000_phy_i217:
4170 			adv_addr = I217_EEE_ADVERTISEMENT;
4171 			break;
4172 		default:
4173 			dev_err(&adapter->pdev->dev,
4174 				"Invalid PHY type setting EEE advertisement\n");
4175 			return;
4176 		}
4177 
4178 		ret_val = hw->phy.ops.acquire(hw);
4179 		if (ret_val) {
4180 			dev_err(&adapter->pdev->dev,
4181 				"EEE advertisement - unable to acquire PHY\n");
4182 			return;
4183 		}
4184 
4185 		e1000_write_emi_reg_locked(hw, adv_addr,
4186 					   hw->dev_spec.ich8lan.eee_disable ?
4187 					   0 : adapter->eee_advert);
4188 
4189 		hw->phy.ops.release(hw);
4190 	}
4191 
4192 	if (!netif_running(adapter->netdev) &&
4193 	    !test_bit(__E1000_TESTING, &adapter->state))
4194 		e1000_power_down_phy(adapter);
4195 
4196 	e1000_get_phy_info(hw);
4197 
4198 	if ((adapter->flags & FLAG_HAS_SMART_POWER_DOWN) &&
4199 	    !(adapter->flags & FLAG_SMART_POWER_DOWN)) {
4200 		u16 phy_data = 0;
4201 		/* speed up time to link by disabling smart power down, ignore
4202 		 * the return value of this function because there is nothing
4203 		 * different we would do if it failed
4204 		 */
4205 		e1e_rphy(hw, IGP02E1000_PHY_POWER_MGMT, &phy_data);
4206 		phy_data &= ~IGP02E1000_PM_SPD;
4207 		e1e_wphy(hw, IGP02E1000_PHY_POWER_MGMT, phy_data);
4208 	}
4209 	if (hw->mac.type >= e1000_pch_spt && adapter->int_mode == 0) {
4210 		u32 reg;
4211 
4212 		/* Fextnvm7 @ 0xe4[2] = 1 */
4213 		reg = er32(FEXTNVM7);
4214 		reg |= E1000_FEXTNVM7_SIDE_CLK_UNGATE;
4215 		ew32(FEXTNVM7, reg);
4216 		/* Fextnvm9 @ 0x5bb4[13:12] = 11 */
4217 		reg = er32(FEXTNVM9);
4218 		reg |= E1000_FEXTNVM9_IOSFSB_CLKGATE_DIS |
4219 		       E1000_FEXTNVM9_IOSFSB_CLKREQ_DIS;
4220 		ew32(FEXTNVM9, reg);
4221 	}
4222 
4223 }
4224 
4225 /**
4226  * e1000e_trigger_lsc - trigger an LSC interrupt
4227  * @adapter: board private structure
4228  *
4229  * Fire a link status change interrupt to start the watchdog.
4230  **/
4231 static void e1000e_trigger_lsc(struct e1000_adapter *adapter)
4232 {
4233 	struct e1000_hw *hw = &adapter->hw;
4234 
4235 	if (adapter->msix_entries)
4236 		ew32(ICS, E1000_ICS_LSC | E1000_ICS_OTHER);
4237 	else
4238 		ew32(ICS, E1000_ICS_LSC);
4239 }
4240 
4241 void e1000e_up(struct e1000_adapter *adapter)
4242 {
4243 	/* hardware has been reset, we need to reload some things */
4244 	e1000_configure(adapter);
4245 
4246 	clear_bit(__E1000_DOWN, &adapter->state);
4247 
4248 	if (adapter->msix_entries)
4249 		e1000_configure_msix(adapter);
4250 	e1000_irq_enable(adapter);
4251 
4252 	/* Tx queue started by watchdog timer when link is up */
4253 
4254 	e1000e_trigger_lsc(adapter);
4255 }
4256 
4257 static void e1000e_flush_descriptors(struct e1000_adapter *adapter)
4258 {
4259 	struct e1000_hw *hw = &adapter->hw;
4260 
4261 	if (!(adapter->flags2 & FLAG2_DMA_BURST))
4262 		return;
4263 
4264 	/* flush pending descriptor writebacks to memory */
4265 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4266 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4267 
4268 	/* execute the writes immediately */
4269 	e1e_flush();
4270 
4271 	/* due to rare timing issues, write to TIDV/RDTR again to ensure the
4272 	 * write is successful
4273 	 */
4274 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4275 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4276 
4277 	/* execute the writes immediately */
4278 	e1e_flush();
4279 }
4280 
4281 static void e1000e_update_stats(struct e1000_adapter *adapter);
4282 
4283 /**
4284  * e1000e_down - quiesce the device and optionally reset the hardware
4285  * @adapter: board private structure
4286  * @reset: boolean flag to reset the hardware or not
4287  */
4288 void e1000e_down(struct e1000_adapter *adapter, bool reset)
4289 {
4290 	struct net_device *netdev = adapter->netdev;
4291 	struct e1000_hw *hw = &adapter->hw;
4292 	u32 tctl, rctl;
4293 
4294 	/* signal that we're down so the interrupt handler does not
4295 	 * reschedule our watchdog timer
4296 	 */
4297 	set_bit(__E1000_DOWN, &adapter->state);
4298 
4299 	netif_carrier_off(netdev);
4300 
4301 	/* disable receives in the hardware */
4302 	rctl = er32(RCTL);
4303 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
4304 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
4305 	/* flush and sleep below */
4306 
4307 	netif_stop_queue(netdev);
4308 
4309 	/* disable transmits in the hardware */
4310 	tctl = er32(TCTL);
4311 	tctl &= ~E1000_TCTL_EN;
4312 	ew32(TCTL, tctl);
4313 
4314 	/* flush both disables and wait for them to finish */
4315 	e1e_flush();
4316 	usleep_range(10000, 11000);
4317 
4318 	e1000_irq_disable(adapter);
4319 
4320 	napi_synchronize(&adapter->napi);
4321 
4322 	timer_delete_sync(&adapter->watchdog_timer);
4323 	timer_delete_sync(&adapter->phy_info_timer);
4324 
4325 	spin_lock(&adapter->stats64_lock);
4326 	e1000e_update_stats(adapter);
4327 	spin_unlock(&adapter->stats64_lock);
4328 
4329 	e1000e_flush_descriptors(adapter);
4330 
4331 	adapter->link_speed = 0;
4332 	adapter->link_duplex = 0;
4333 
4334 	/* Disable Si errata workaround on PCHx for jumbo frame flow */
4335 	if ((hw->mac.type >= e1000_pch2lan) &&
4336 	    (adapter->netdev->mtu > ETH_DATA_LEN) &&
4337 	    e1000_lv_jumbo_workaround_ich8lan(hw, false))
4338 		e_dbg("failed to disable jumbo frame workaround mode\n");
4339 
4340 	if (!pci_channel_offline(adapter->pdev)) {
4341 		if (reset)
4342 			e1000e_reset(adapter);
4343 		else if (hw->mac.type >= e1000_pch_spt)
4344 			e1000_flush_desc_rings(adapter);
4345 	}
4346 	e1000_clean_tx_ring(adapter->tx_ring);
4347 	e1000_clean_rx_ring(adapter->rx_ring);
4348 }
4349 
4350 void e1000e_reinit_locked(struct e1000_adapter *adapter)
4351 {
4352 	might_sleep();
4353 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
4354 		usleep_range(1000, 1100);
4355 	e1000e_down(adapter, true);
4356 	e1000e_up(adapter);
4357 	clear_bit(__E1000_RESETTING, &adapter->state);
4358 }
4359 
4360 /**
4361  * e1000e_sanitize_systim - sanitize raw cycle counter reads
4362  * @hw: pointer to the HW structure
4363  * @systim: PHC time value read, sanitized and returned
4364  * @sts: structure to hold system time before and after reading SYSTIML,
4365  * may be NULL
4366  *
4367  * Errata for 82574/82583 possible bad bits read from SYSTIMH/L:
4368  * check to see that the time is incrementing at a reasonable
4369  * rate and is a multiple of incvalue.
4370  **/
4371 static u64 e1000e_sanitize_systim(struct e1000_hw *hw, u64 systim,
4372 				  struct ptp_system_timestamp *sts)
4373 {
4374 	u64 time_delta, rem, temp;
4375 	u64 systim_next;
4376 	u32 incvalue;
4377 	int i;
4378 
4379 	incvalue = er32(TIMINCA) & E1000_TIMINCA_INCVALUE_MASK;
4380 	for (i = 0; i < E1000_MAX_82574_SYSTIM_REREADS; i++) {
4381 		/* latch SYSTIMH on read of SYSTIML */
4382 		ptp_read_system_prets(sts);
4383 		systim_next = (u64)er32(SYSTIML);
4384 		ptp_read_system_postts(sts);
4385 		systim_next |= (u64)er32(SYSTIMH) << 32;
4386 
4387 		time_delta = systim_next - systim;
4388 		temp = time_delta;
4389 		/* VMWare users have seen incvalue of zero, don't div / 0 */
4390 		rem = incvalue ? do_div(temp, incvalue) : (time_delta != 0);
4391 
4392 		systim = systim_next;
4393 
4394 		if ((time_delta < E1000_82574_SYSTIM_EPSILON) && (rem == 0))
4395 			break;
4396 	}
4397 
4398 	return systim;
4399 }
4400 
4401 /**
4402  * e1000e_read_systim - read SYSTIM register
4403  * @adapter: board private structure
4404  * @sts: structure which will contain system time before and after reading
4405  * SYSTIML, may be NULL
4406  **/
4407 u64 e1000e_read_systim(struct e1000_adapter *adapter,
4408 		       struct ptp_system_timestamp *sts)
4409 {
4410 	struct e1000_hw *hw = &adapter->hw;
4411 	u32 systimel, systimel_2, systimeh;
4412 	u64 systim;
4413 	/* SYSTIMH latching upon SYSTIML read does not work well.
4414 	 * This means that if SYSTIML overflows after we read it but before
4415 	 * we read SYSTIMH, the value of SYSTIMH has been incremented and we
4416 	 * will experience a huge non linear increment in the systime value
4417 	 * to fix that we test for overflow and if true, we re-read systime.
4418 	 */
4419 	ptp_read_system_prets(sts);
4420 	systimel = er32(SYSTIML);
4421 	ptp_read_system_postts(sts);
4422 	systimeh = er32(SYSTIMH);
4423 	/* Is systimel is so large that overflow is possible? */
4424 	if (systimel >= (u32)0xffffffff - E1000_TIMINCA_INCVALUE_MASK) {
4425 		ptp_read_system_prets(sts);
4426 		systimel_2 = er32(SYSTIML);
4427 		ptp_read_system_postts(sts);
4428 		if (systimel > systimel_2) {
4429 			/* There was an overflow, read again SYSTIMH, and use
4430 			 * systimel_2
4431 			 */
4432 			systimeh = er32(SYSTIMH);
4433 			systimel = systimel_2;
4434 		}
4435 	}
4436 	systim = (u64)systimel;
4437 	systim |= (u64)systimeh << 32;
4438 
4439 	if (adapter->flags2 & FLAG2_CHECK_SYSTIM_OVERFLOW)
4440 		systim = e1000e_sanitize_systim(hw, systim, sts);
4441 
4442 	return systim;
4443 }
4444 
4445 /**
4446  * e1000e_cyclecounter_read - read raw cycle counter (used by time counter)
4447  * @cc: cyclecounter structure
4448  **/
4449 static u64 e1000e_cyclecounter_read(struct cyclecounter *cc)
4450 {
4451 	struct e1000_adapter *adapter = container_of(cc, struct e1000_adapter,
4452 						     cc);
4453 
4454 	return e1000e_read_systim(adapter, NULL);
4455 }
4456 
4457 /**
4458  * e1000_sw_init - Initialize general software structures (struct e1000_adapter)
4459  * @adapter: board private structure to initialize
4460  *
4461  * e1000_sw_init initializes the Adapter private data structure.
4462  * Fields are initialized based on PCI device information and
4463  * OS network device settings (MTU size).
4464  **/
4465 static int e1000_sw_init(struct e1000_adapter *adapter)
4466 {
4467 	struct net_device *netdev = adapter->netdev;
4468 
4469 	adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
4470 	adapter->rx_ps_bsize0 = 128;
4471 	adapter->max_frame_size = netdev->mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
4472 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4473 	adapter->tx_ring_count = E1000_DEFAULT_TXD;
4474 	adapter->rx_ring_count = E1000_DEFAULT_RXD;
4475 
4476 	spin_lock_init(&adapter->stats64_lock);
4477 
4478 	e1000e_set_interrupt_capability(adapter);
4479 
4480 	if (e1000_alloc_queues(adapter))
4481 		return -ENOMEM;
4482 
4483 	/* Setup hardware time stamping cyclecounter */
4484 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
4485 		adapter->cc.read = e1000e_cyclecounter_read;
4486 		adapter->cc.mask = CYCLECOUNTER_MASK(64);
4487 		adapter->cc.mult = 1;
4488 		/* cc.shift set in e1000e_get_base_tininca() */
4489 
4490 		spin_lock_init(&adapter->systim_lock);
4491 		INIT_WORK(&adapter->tx_hwtstamp_work, e1000e_tx_hwtstamp_work);
4492 	}
4493 
4494 	/* Explicitly disable IRQ since the NIC can be in any state. */
4495 	e1000_irq_disable(adapter);
4496 
4497 	set_bit(__E1000_DOWN, &adapter->state);
4498 	return 0;
4499 }
4500 
4501 /**
4502  * e1000_intr_msi_test - Interrupt Handler
4503  * @irq: interrupt number
4504  * @data: pointer to a network interface device structure
4505  **/
4506 static irqreturn_t e1000_intr_msi_test(int __always_unused irq, void *data)
4507 {
4508 	struct net_device *netdev = data;
4509 	struct e1000_adapter *adapter = netdev_priv(netdev);
4510 	struct e1000_hw *hw = &adapter->hw;
4511 	u32 icr = er32(ICR);
4512 
4513 	e_dbg("icr is %08X\n", icr);
4514 	if (icr & E1000_ICR_RXSEQ) {
4515 		adapter->flags &= ~FLAG_MSI_TEST_FAILED;
4516 		/* Force memory writes to complete before acknowledging the
4517 		 * interrupt is handled.
4518 		 */
4519 		wmb();
4520 	}
4521 
4522 	return IRQ_HANDLED;
4523 }
4524 
4525 /**
4526  * e1000_test_msi_interrupt - Returns 0 for successful test
4527  * @adapter: board private struct
4528  *
4529  * code flow taken from tg3.c
4530  **/
4531 static int e1000_test_msi_interrupt(struct e1000_adapter *adapter)
4532 {
4533 	struct net_device *netdev = adapter->netdev;
4534 	struct e1000_hw *hw = &adapter->hw;
4535 	int err;
4536 
4537 	/* poll_enable hasn't been called yet, so don't need disable */
4538 	/* clear any pending events */
4539 	er32(ICR);
4540 
4541 	/* free the real vector and request a test handler */
4542 	e1000_free_irq(adapter);
4543 	e1000e_reset_interrupt_capability(adapter);
4544 
4545 	/* Assume that the test fails, if it succeeds then the test
4546 	 * MSI irq handler will unset this flag
4547 	 */
4548 	adapter->flags |= FLAG_MSI_TEST_FAILED;
4549 
4550 	err = pci_enable_msi(adapter->pdev);
4551 	if (err)
4552 		goto msi_test_failed;
4553 
4554 	err = request_irq(adapter->pdev->irq, e1000_intr_msi_test, 0,
4555 			  netdev->name, netdev);
4556 	if (err) {
4557 		pci_disable_msi(adapter->pdev);
4558 		goto msi_test_failed;
4559 	}
4560 
4561 	/* Force memory writes to complete before enabling and firing an
4562 	 * interrupt.
4563 	 */
4564 	wmb();
4565 
4566 	e1000_irq_enable(adapter);
4567 
4568 	/* fire an unusual interrupt on the test handler */
4569 	ew32(ICS, E1000_ICS_RXSEQ);
4570 	e1e_flush();
4571 	msleep(100);
4572 
4573 	e1000_irq_disable(adapter);
4574 
4575 	rmb();			/* read flags after interrupt has been fired */
4576 
4577 	if (adapter->flags & FLAG_MSI_TEST_FAILED) {
4578 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
4579 		e_info("MSI interrupt test failed, using legacy interrupt.\n");
4580 	} else {
4581 		e_dbg("MSI interrupt test succeeded!\n");
4582 	}
4583 
4584 	free_irq(adapter->pdev->irq, netdev);
4585 	pci_disable_msi(adapter->pdev);
4586 
4587 msi_test_failed:
4588 	e1000e_set_interrupt_capability(adapter);
4589 	return e1000_request_irq(adapter);
4590 }
4591 
4592 /**
4593  * e1000_test_msi - Returns 0 if MSI test succeeds or INTx mode is restored
4594  * @adapter: board private struct
4595  *
4596  * code flow taken from tg3.c, called with e1000 interrupts disabled.
4597  **/
4598 static int e1000_test_msi(struct e1000_adapter *adapter)
4599 {
4600 	int err;
4601 	u16 pci_cmd;
4602 
4603 	if (!(adapter->flags & FLAG_MSI_ENABLED))
4604 		return 0;
4605 
4606 	/* disable SERR in case the MSI write causes a master abort */
4607 	pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4608 	if (pci_cmd & PCI_COMMAND_SERR)
4609 		pci_write_config_word(adapter->pdev, PCI_COMMAND,
4610 				      pci_cmd & ~PCI_COMMAND_SERR);
4611 
4612 	err = e1000_test_msi_interrupt(adapter);
4613 
4614 	/* re-enable SERR */
4615 	if (pci_cmd & PCI_COMMAND_SERR) {
4616 		pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4617 		pci_cmd |= PCI_COMMAND_SERR;
4618 		pci_write_config_word(adapter->pdev, PCI_COMMAND, pci_cmd);
4619 	}
4620 
4621 	return err;
4622 }
4623 
4624 /**
4625  * e1000e_open - Called when a network interface is made active
4626  * @netdev: network interface device structure
4627  *
4628  * Returns 0 on success, negative value on failure
4629  *
4630  * The open entry point is called when a network interface is made
4631  * active by the system (IFF_UP).  At this point all resources needed
4632  * for transmit and receive operations are allocated, the interrupt
4633  * handler is registered with the OS, the watchdog timer is started,
4634  * and the stack is notified that the interface is ready.
4635  **/
4636 int e1000e_open(struct net_device *netdev)
4637 {
4638 	struct e1000_adapter *adapter = netdev_priv(netdev);
4639 	struct e1000_hw *hw = &adapter->hw;
4640 	struct pci_dev *pdev = adapter->pdev;
4641 	int err;
4642 	int irq;
4643 
4644 	/* disallow open during test */
4645 	if (test_bit(__E1000_TESTING, &adapter->state))
4646 		return -EBUSY;
4647 
4648 	pm_runtime_get_sync(&pdev->dev);
4649 
4650 	netif_carrier_off(netdev);
4651 	netif_stop_queue(netdev);
4652 
4653 	/* allocate transmit descriptors */
4654 	err = e1000e_setup_tx_resources(adapter->tx_ring);
4655 	if (err)
4656 		goto err_setup_tx;
4657 
4658 	/* allocate receive descriptors */
4659 	err = e1000e_setup_rx_resources(adapter->rx_ring);
4660 	if (err)
4661 		goto err_setup_rx;
4662 
4663 	/* If AMT is enabled, let the firmware know that the network
4664 	 * interface is now open and reset the part to a known state.
4665 	 */
4666 	if (adapter->flags & FLAG_HAS_AMT) {
4667 		e1000e_get_hw_control(adapter);
4668 		e1000e_reset(adapter);
4669 	}
4670 
4671 	e1000e_power_up_phy(adapter);
4672 
4673 	adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
4674 	if ((adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
4675 		e1000_update_mng_vlan(adapter);
4676 
4677 	/* DMA latency requirement to workaround jumbo issue */
4678 	cpu_latency_qos_add_request(&adapter->pm_qos_req, PM_QOS_DEFAULT_VALUE);
4679 
4680 	/* before we allocate an interrupt, we must be ready to handle it.
4681 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4682 	 * as soon as we call pci_request_irq, so we have to setup our
4683 	 * clean_rx handler before we do so.
4684 	 */
4685 	e1000_configure(adapter);
4686 
4687 	err = e1000_request_irq(adapter);
4688 	if (err)
4689 		goto err_req_irq;
4690 
4691 	/* Work around PCIe errata with MSI interrupts causing some chipsets to
4692 	 * ignore e1000e MSI messages, which means we need to test our MSI
4693 	 * interrupt now
4694 	 */
4695 	if (adapter->int_mode != E1000E_INT_MODE_LEGACY) {
4696 		err = e1000_test_msi(adapter);
4697 		if (err) {
4698 			e_err("Interrupt allocation failed\n");
4699 			goto err_req_irq;
4700 		}
4701 	}
4702 
4703 	/* From here on the code is the same as e1000e_up() */
4704 	clear_bit(__E1000_DOWN, &adapter->state);
4705 
4706 	if (adapter->int_mode == E1000E_INT_MODE_MSIX)
4707 		irq = adapter->msix_entries[0].vector;
4708 	else
4709 		irq = adapter->pdev->irq;
4710 
4711 	netif_napi_set_irq(&adapter->napi, irq);
4712 	napi_enable(&adapter->napi);
4713 	netif_queue_set_napi(netdev, 0, NETDEV_QUEUE_TYPE_RX, &adapter->napi);
4714 	netif_queue_set_napi(netdev, 0, NETDEV_QUEUE_TYPE_TX, &adapter->napi);
4715 
4716 	e1000_irq_enable(adapter);
4717 
4718 	adapter->tx_hang_recheck = false;
4719 
4720 	hw->mac.get_link_status = true;
4721 	pm_runtime_put(&pdev->dev);
4722 
4723 	e1000e_trigger_lsc(adapter);
4724 
4725 	return 0;
4726 
4727 err_req_irq:
4728 	cpu_latency_qos_remove_request(&adapter->pm_qos_req);
4729 	e1000e_release_hw_control(adapter);
4730 	e1000_power_down_phy(adapter);
4731 	e1000e_free_rx_resources(adapter->rx_ring);
4732 err_setup_rx:
4733 	e1000e_free_tx_resources(adapter->tx_ring);
4734 err_setup_tx:
4735 	e1000e_reset(adapter);
4736 	pm_runtime_put_sync(&pdev->dev);
4737 
4738 	return err;
4739 }
4740 
4741 /**
4742  * e1000e_close - Disables a network interface
4743  * @netdev: network interface device structure
4744  *
4745  * Returns 0, this is not allowed to fail
4746  *
4747  * The close entry point is called when an interface is de-activated
4748  * by the OS.  The hardware is still under the drivers control, but
4749  * needs to be disabled.  A global MAC reset is issued to stop the
4750  * hardware, and all transmit and receive resources are freed.
4751  **/
4752 int e1000e_close(struct net_device *netdev)
4753 {
4754 	struct e1000_adapter *adapter = netdev_priv(netdev);
4755 	struct pci_dev *pdev = adapter->pdev;
4756 	int count = E1000_CHECK_RESET_COUNT;
4757 
4758 	while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
4759 		usleep_range(10000, 11000);
4760 
4761 	WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
4762 
4763 	pm_runtime_get_sync(&pdev->dev);
4764 
4765 	if (netif_device_present(netdev)) {
4766 		e1000e_down(adapter, true);
4767 		e1000_free_irq(adapter);
4768 
4769 		/* Link status message must follow this format */
4770 		netdev_info(netdev, "NIC Link is Down\n");
4771 	}
4772 
4773 	netif_queue_set_napi(netdev, 0, NETDEV_QUEUE_TYPE_RX, NULL);
4774 	netif_queue_set_napi(netdev, 0, NETDEV_QUEUE_TYPE_TX, NULL);
4775 	napi_disable(&adapter->napi);
4776 
4777 	e1000e_free_tx_resources(adapter->tx_ring);
4778 	e1000e_free_rx_resources(adapter->rx_ring);
4779 
4780 	/* kill manageability vlan ID if supported, but not if a vlan with
4781 	 * the same ID is registered on the host OS (let 8021q kill it)
4782 	 */
4783 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN)
4784 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
4785 				       adapter->mng_vlan_id);
4786 
4787 	/* If AMT is enabled, let the firmware know that the network
4788 	 * interface is now closed
4789 	 */
4790 	if ((adapter->flags & FLAG_HAS_AMT) &&
4791 	    !test_bit(__E1000_TESTING, &adapter->state))
4792 		e1000e_release_hw_control(adapter);
4793 
4794 	cpu_latency_qos_remove_request(&adapter->pm_qos_req);
4795 
4796 	pm_runtime_put_sync(&pdev->dev);
4797 
4798 	return 0;
4799 }
4800 
4801 /**
4802  * e1000_set_mac - Change the Ethernet Address of the NIC
4803  * @netdev: network interface device structure
4804  * @p: pointer to an address structure
4805  *
4806  * Returns 0 on success, negative on failure
4807  **/
4808 static int e1000_set_mac(struct net_device *netdev, void *p)
4809 {
4810 	struct e1000_adapter *adapter = netdev_priv(netdev);
4811 	struct e1000_hw *hw = &adapter->hw;
4812 	struct sockaddr *addr = p;
4813 
4814 	if (!is_valid_ether_addr(addr->sa_data))
4815 		return -EADDRNOTAVAIL;
4816 
4817 	eth_hw_addr_set(netdev, addr->sa_data);
4818 	memcpy(adapter->hw.mac.addr, addr->sa_data, netdev->addr_len);
4819 
4820 	hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
4821 
4822 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA) {
4823 		/* activate the work around */
4824 		e1000e_set_laa_state_82571(&adapter->hw, 1);
4825 
4826 		/* Hold a copy of the LAA in RAR[14] This is done so that
4827 		 * between the time RAR[0] gets clobbered  and the time it
4828 		 * gets fixed (in e1000_watchdog), the actual LAA is in one
4829 		 * of the RARs and no incoming packets directed to this port
4830 		 * are dropped. Eventually the LAA will be in RAR[0] and
4831 		 * RAR[14]
4832 		 */
4833 		hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr,
4834 				    adapter->hw.mac.rar_entry_count - 1);
4835 	}
4836 
4837 	return 0;
4838 }
4839 
4840 /**
4841  * e1000e_update_phy_task - work thread to update phy
4842  * @work: pointer to our work struct
4843  *
4844  * this worker thread exists because we must acquire a
4845  * semaphore to read the phy, which we could msleep while
4846  * waiting for it, and we can't msleep in a timer.
4847  **/
4848 static void e1000e_update_phy_task(struct work_struct *work)
4849 {
4850 	struct e1000_adapter *adapter = container_of(work,
4851 						     struct e1000_adapter,
4852 						     update_phy_task);
4853 	struct e1000_hw *hw = &adapter->hw;
4854 
4855 	if (test_bit(__E1000_DOWN, &adapter->state))
4856 		return;
4857 
4858 	e1000_get_phy_info(hw);
4859 
4860 	/* Enable EEE on 82579 after link up */
4861 	if (hw->phy.type >= e1000_phy_82579)
4862 		e1000_set_eee_pchlan(hw);
4863 }
4864 
4865 /**
4866  * e1000_update_phy_info - timre call-back to update PHY info
4867  * @t: pointer to timer_list containing private info adapter
4868  *
4869  * Need to wait a few seconds after link up to get diagnostic information from
4870  * the phy
4871  **/
4872 static void e1000_update_phy_info(struct timer_list *t)
4873 {
4874 	struct e1000_adapter *adapter = timer_container_of(adapter, t,
4875 							   phy_info_timer);
4876 
4877 	if (test_bit(__E1000_DOWN, &adapter->state))
4878 		return;
4879 
4880 	schedule_work(&adapter->update_phy_task);
4881 }
4882 
4883 /**
4884  * e1000e_update_phy_stats - Update the PHY statistics counters
4885  * @adapter: board private structure
4886  *
4887  * Read/clear the upper 16-bit PHY registers and read/accumulate lower
4888  **/
4889 static void e1000e_update_phy_stats(struct e1000_adapter *adapter)
4890 {
4891 	struct e1000_hw *hw = &adapter->hw;
4892 	s32 ret_val;
4893 	u16 phy_data;
4894 
4895 	ret_val = hw->phy.ops.acquire(hw);
4896 	if (ret_val)
4897 		return;
4898 
4899 	/* A page set is expensive so check if already on desired page.
4900 	 * If not, set to the page with the PHY status registers.
4901 	 */
4902 	hw->phy.addr = 1;
4903 	ret_val = e1000e_read_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4904 					   &phy_data);
4905 	if (ret_val)
4906 		goto release;
4907 	if (phy_data != (HV_STATS_PAGE << IGP_PAGE_SHIFT)) {
4908 		ret_val = hw->phy.ops.set_page(hw,
4909 					       HV_STATS_PAGE << IGP_PAGE_SHIFT);
4910 		if (ret_val)
4911 			goto release;
4912 	}
4913 
4914 	/* Single Collision Count */
4915 	hw->phy.ops.read_reg_page(hw, HV_SCC_UPPER, &phy_data);
4916 	ret_val = hw->phy.ops.read_reg_page(hw, HV_SCC_LOWER, &phy_data);
4917 	if (!ret_val)
4918 		adapter->stats.scc += phy_data;
4919 
4920 	/* Excessive Collision Count */
4921 	hw->phy.ops.read_reg_page(hw, HV_ECOL_UPPER, &phy_data);
4922 	ret_val = hw->phy.ops.read_reg_page(hw, HV_ECOL_LOWER, &phy_data);
4923 	if (!ret_val)
4924 		adapter->stats.ecol += phy_data;
4925 
4926 	/* Multiple Collision Count */
4927 	hw->phy.ops.read_reg_page(hw, HV_MCC_UPPER, &phy_data);
4928 	ret_val = hw->phy.ops.read_reg_page(hw, HV_MCC_LOWER, &phy_data);
4929 	if (!ret_val)
4930 		adapter->stats.mcc += phy_data;
4931 
4932 	/* Late Collision Count */
4933 	hw->phy.ops.read_reg_page(hw, HV_LATECOL_UPPER, &phy_data);
4934 	ret_val = hw->phy.ops.read_reg_page(hw, HV_LATECOL_LOWER, &phy_data);
4935 	if (!ret_val)
4936 		adapter->stats.latecol += phy_data;
4937 
4938 	/* Collision Count - also used for adaptive IFS */
4939 	hw->phy.ops.read_reg_page(hw, HV_COLC_UPPER, &phy_data);
4940 	ret_val = hw->phy.ops.read_reg_page(hw, HV_COLC_LOWER, &phy_data);
4941 	if (!ret_val)
4942 		hw->mac.collision_delta = phy_data;
4943 
4944 	/* Defer Count */
4945 	hw->phy.ops.read_reg_page(hw, HV_DC_UPPER, &phy_data);
4946 	ret_val = hw->phy.ops.read_reg_page(hw, HV_DC_LOWER, &phy_data);
4947 	if (!ret_val)
4948 		adapter->stats.dc += phy_data;
4949 
4950 	/* Transmit with no CRS */
4951 	hw->phy.ops.read_reg_page(hw, HV_TNCRS_UPPER, &phy_data);
4952 	ret_val = hw->phy.ops.read_reg_page(hw, HV_TNCRS_LOWER, &phy_data);
4953 	if (!ret_val)
4954 		adapter->stats.tncrs += phy_data;
4955 
4956 release:
4957 	hw->phy.ops.release(hw);
4958 }
4959 
4960 /**
4961  * e1000e_update_stats - Update the board statistics counters
4962  * @adapter: board private structure
4963  **/
4964 static void e1000e_update_stats(struct e1000_adapter *adapter)
4965 {
4966 	struct net_device *netdev = adapter->netdev;
4967 	struct e1000_hw *hw = &adapter->hw;
4968 	struct pci_dev *pdev = adapter->pdev;
4969 
4970 	/* Prevent stats update while adapter is being reset, or if the pci
4971 	 * connection is down.
4972 	 */
4973 	if (adapter->link_speed == 0)
4974 		return;
4975 	if (pci_channel_offline(pdev))
4976 		return;
4977 
4978 	adapter->stats.crcerrs += er32(CRCERRS);
4979 	adapter->stats.gprc += er32(GPRC);
4980 	adapter->stats.gorc += er32(GORCL);
4981 	er32(GORCH);		/* Clear gorc */
4982 	adapter->stats.bprc += er32(BPRC);
4983 	adapter->stats.mprc += er32(MPRC);
4984 	adapter->stats.roc += er32(ROC);
4985 
4986 	adapter->stats.mpc += er32(MPC);
4987 
4988 	/* Half-duplex statistics */
4989 	if (adapter->link_duplex == HALF_DUPLEX) {
4990 		if (adapter->flags2 & FLAG2_HAS_PHY_STATS) {
4991 			e1000e_update_phy_stats(adapter);
4992 		} else {
4993 			adapter->stats.scc += er32(SCC);
4994 			adapter->stats.ecol += er32(ECOL);
4995 			adapter->stats.mcc += er32(MCC);
4996 			adapter->stats.latecol += er32(LATECOL);
4997 			adapter->stats.dc += er32(DC);
4998 
4999 			hw->mac.collision_delta = er32(COLC);
5000 
5001 			if ((hw->mac.type != e1000_82574) &&
5002 			    (hw->mac.type != e1000_82583))
5003 				adapter->stats.tncrs += er32(TNCRS);
5004 		}
5005 		adapter->stats.colc += hw->mac.collision_delta;
5006 	}
5007 
5008 	adapter->stats.xonrxc += er32(XONRXC);
5009 	adapter->stats.xontxc += er32(XONTXC);
5010 	adapter->stats.xoffrxc += er32(XOFFRXC);
5011 	adapter->stats.xofftxc += er32(XOFFTXC);
5012 	adapter->stats.gptc += er32(GPTC);
5013 	adapter->stats.gotc += er32(GOTCL);
5014 	er32(GOTCH);		/* Clear gotc */
5015 	adapter->stats.rnbc += er32(RNBC);
5016 	adapter->stats.ruc += er32(RUC);
5017 
5018 	adapter->stats.mptc += er32(MPTC);
5019 	adapter->stats.bptc += er32(BPTC);
5020 
5021 	/* used for adaptive IFS */
5022 
5023 	hw->mac.tx_packet_delta = er32(TPT);
5024 	adapter->stats.tpt += hw->mac.tx_packet_delta;
5025 
5026 	adapter->stats.algnerrc += er32(ALGNERRC);
5027 	adapter->stats.rxerrc += er32(RXERRC);
5028 	adapter->stats.cexterr += er32(CEXTERR);
5029 	adapter->stats.tsctc += er32(TSCTC);
5030 	adapter->stats.tsctfc += er32(TSCTFC);
5031 
5032 	/* Fill out the OS statistics structure */
5033 	netdev->stats.multicast = adapter->stats.mprc;
5034 	netdev->stats.collisions = adapter->stats.colc;
5035 
5036 	/* Rx Errors */
5037 
5038 	/* RLEC on some newer hardware can be incorrect so build
5039 	 * our own version based on RUC and ROC
5040 	 */
5041 	netdev->stats.rx_errors = adapter->stats.rxerrc +
5042 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
5043 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5044 	netdev->stats.rx_length_errors = adapter->stats.ruc +
5045 	    adapter->stats.roc;
5046 	netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
5047 	netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
5048 	netdev->stats.rx_missed_errors = adapter->stats.mpc;
5049 
5050 	/* Tx Errors */
5051 	netdev->stats.tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5052 	netdev->stats.tx_aborted_errors = adapter->stats.ecol;
5053 	netdev->stats.tx_window_errors = adapter->stats.latecol;
5054 	netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
5055 
5056 	/* Tx Dropped needs to be maintained elsewhere */
5057 
5058 	/* Management Stats */
5059 	adapter->stats.mgptc += er32(MGTPTC);
5060 	adapter->stats.mgprc += er32(MGTPRC);
5061 	adapter->stats.mgpdc += er32(MGTPDC);
5062 
5063 	/* Correctable ECC Errors */
5064 	if (hw->mac.type >= e1000_pch_lpt) {
5065 		u32 pbeccsts = er32(PBECCSTS);
5066 
5067 		adapter->corr_errors +=
5068 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
5069 		adapter->uncorr_errors +=
5070 		    FIELD_GET(E1000_PBECCSTS_UNCORR_ERR_CNT_MASK, pbeccsts);
5071 	}
5072 }
5073 
5074 /**
5075  * e1000_phy_read_status - Update the PHY register status snapshot
5076  * @adapter: board private structure
5077  **/
5078 static void e1000_phy_read_status(struct e1000_adapter *adapter)
5079 {
5080 	struct e1000_hw *hw = &adapter->hw;
5081 	struct e1000_phy_regs *phy = &adapter->phy_regs;
5082 
5083 	if (!pm_runtime_suspended((&adapter->pdev->dev)->parent) &&
5084 	    (er32(STATUS) & E1000_STATUS_LU) &&
5085 	    (adapter->hw.phy.media_type == e1000_media_type_copper)) {
5086 		int ret_val;
5087 
5088 		ret_val = e1e_rphy(hw, MII_BMCR, &phy->bmcr);
5089 		ret_val |= e1e_rphy(hw, MII_BMSR, &phy->bmsr);
5090 		ret_val |= e1e_rphy(hw, MII_ADVERTISE, &phy->advertise);
5091 		ret_val |= e1e_rphy(hw, MII_LPA, &phy->lpa);
5092 		ret_val |= e1e_rphy(hw, MII_EXPANSION, &phy->expansion);
5093 		ret_val |= e1e_rphy(hw, MII_CTRL1000, &phy->ctrl1000);
5094 		ret_val |= e1e_rphy(hw, MII_STAT1000, &phy->stat1000);
5095 		ret_val |= e1e_rphy(hw, MII_ESTATUS, &phy->estatus);
5096 		if (ret_val)
5097 			e_warn("Error reading PHY register\n");
5098 	} else {
5099 		/* Do not read PHY registers if link is not up
5100 		 * Set values to typical power-on defaults
5101 		 */
5102 		phy->bmcr = (BMCR_SPEED1000 | BMCR_ANENABLE | BMCR_FULLDPLX);
5103 		phy->bmsr = (BMSR_100FULL | BMSR_100HALF | BMSR_10FULL |
5104 			     BMSR_10HALF | BMSR_ESTATEN | BMSR_ANEGCAPABLE |
5105 			     BMSR_ERCAP);
5106 		phy->advertise = (ADVERTISE_PAUSE_ASYM | ADVERTISE_PAUSE_CAP |
5107 				  ADVERTISE_ALL | ADVERTISE_CSMA);
5108 		phy->lpa = 0;
5109 		phy->expansion = EXPANSION_ENABLENPAGE;
5110 		phy->ctrl1000 = ADVERTISE_1000FULL;
5111 		phy->stat1000 = 0;
5112 		phy->estatus = (ESTATUS_1000_TFULL | ESTATUS_1000_THALF);
5113 	}
5114 }
5115 
5116 static void e1000_print_link_info(struct e1000_adapter *adapter)
5117 {
5118 	struct e1000_hw *hw = &adapter->hw;
5119 	u32 ctrl = er32(CTRL);
5120 
5121 	/* Link status message must follow this format for user tools */
5122 	netdev_info(adapter->netdev,
5123 		    "NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5124 		    adapter->link_speed,
5125 		    adapter->link_duplex == FULL_DUPLEX ? "Full" : "Half",
5126 		    (ctrl & E1000_CTRL_TFCE) && (ctrl & E1000_CTRL_RFCE) ? "Rx/Tx" :
5127 		    (ctrl & E1000_CTRL_RFCE) ? "Rx" :
5128 		    (ctrl & E1000_CTRL_TFCE) ? "Tx" : "None");
5129 }
5130 
5131 static bool e1000e_has_link(struct e1000_adapter *adapter)
5132 {
5133 	struct e1000_hw *hw = &adapter->hw;
5134 	bool link_active = false;
5135 	s32 ret_val = 0;
5136 
5137 	/* get_link_status is set on LSC (link status) interrupt or
5138 	 * Rx sequence error interrupt.  get_link_status will stay
5139 	 * true until the check_for_link establishes link
5140 	 * for copper adapters ONLY
5141 	 */
5142 	switch (hw->phy.media_type) {
5143 	case e1000_media_type_copper:
5144 		if (hw->mac.get_link_status) {
5145 			ret_val = hw->mac.ops.check_for_link(hw);
5146 			link_active = !hw->mac.get_link_status;
5147 		} else {
5148 			link_active = true;
5149 		}
5150 		break;
5151 	case e1000_media_type_fiber:
5152 		ret_val = hw->mac.ops.check_for_link(hw);
5153 		link_active = !!(er32(STATUS) & E1000_STATUS_LU);
5154 		break;
5155 	case e1000_media_type_internal_serdes:
5156 		ret_val = hw->mac.ops.check_for_link(hw);
5157 		link_active = hw->mac.serdes_has_link;
5158 		break;
5159 	default:
5160 	case e1000_media_type_unknown:
5161 		break;
5162 	}
5163 
5164 	if ((ret_val == -E1000_ERR_PHY) && (hw->phy.type == e1000_phy_igp_3) &&
5165 	    (er32(CTRL) & E1000_PHY_CTRL_GBE_DISABLE)) {
5166 		/* See e1000_kmrn_lock_loss_workaround_ich8lan() */
5167 		e_info("Gigabit has been disabled, downgrading speed\n");
5168 	}
5169 
5170 	return link_active;
5171 }
5172 
5173 static void e1000e_enable_receives(struct e1000_adapter *adapter)
5174 {
5175 	/* make sure the receive unit is started */
5176 	if ((adapter->flags & FLAG_RX_NEEDS_RESTART) &&
5177 	    (adapter->flags & FLAG_RESTART_NOW)) {
5178 		struct e1000_hw *hw = &adapter->hw;
5179 		u32 rctl = er32(RCTL);
5180 
5181 		ew32(RCTL, rctl | E1000_RCTL_EN);
5182 		adapter->flags &= ~FLAG_RESTART_NOW;
5183 	}
5184 }
5185 
5186 static void e1000e_check_82574_phy_workaround(struct e1000_adapter *adapter)
5187 {
5188 	struct e1000_hw *hw = &adapter->hw;
5189 
5190 	/* With 82574 controllers, PHY needs to be checked periodically
5191 	 * for hung state and reset, if two calls return true
5192 	 */
5193 	if (e1000_check_phy_82574(hw))
5194 		adapter->phy_hang_count++;
5195 	else
5196 		adapter->phy_hang_count = 0;
5197 
5198 	if (adapter->phy_hang_count > 1) {
5199 		adapter->phy_hang_count = 0;
5200 		e_dbg("PHY appears hung - resetting\n");
5201 		schedule_work(&adapter->reset_task);
5202 	}
5203 }
5204 
5205 /**
5206  * e1000_watchdog - Timer Call-back
5207  * @t: pointer to timer_list containing private info adapter
5208  **/
5209 static void e1000_watchdog(struct timer_list *t)
5210 {
5211 	struct e1000_adapter *adapter = timer_container_of(adapter, t,
5212 							   watchdog_timer);
5213 
5214 	/* Do the rest outside of interrupt context */
5215 	schedule_work(&adapter->watchdog_task);
5216 
5217 	/* TODO: make this use queue_delayed_work() */
5218 }
5219 
5220 static void e1000_watchdog_task(struct work_struct *work)
5221 {
5222 	struct e1000_adapter *adapter = container_of(work,
5223 						     struct e1000_adapter,
5224 						     watchdog_task);
5225 	struct net_device *netdev = adapter->netdev;
5226 	struct e1000_mac_info *mac = &adapter->hw.mac;
5227 	struct e1000_phy_info *phy = &adapter->hw.phy;
5228 	struct e1000_ring *tx_ring = adapter->tx_ring;
5229 	u32 dmoff_exit_timeout = 100, tries = 0;
5230 	struct e1000_hw *hw = &adapter->hw;
5231 	u32 link, tctl, pcim_state;
5232 
5233 	if (test_bit(__E1000_DOWN, &adapter->state))
5234 		return;
5235 
5236 	link = e1000e_has_link(adapter);
5237 	if ((netif_carrier_ok(netdev)) && link) {
5238 		/* Cancel scheduled suspend requests. */
5239 		pm_runtime_resume(netdev->dev.parent);
5240 
5241 		e1000e_enable_receives(adapter);
5242 		goto link_up;
5243 	}
5244 
5245 	if ((e1000e_enable_tx_pkt_filtering(hw)) &&
5246 	    (adapter->mng_vlan_id != adapter->hw.mng_cookie.vlan_id))
5247 		e1000_update_mng_vlan(adapter);
5248 
5249 	if (link) {
5250 		if (!netif_carrier_ok(netdev)) {
5251 			bool txb2b = true;
5252 
5253 			/* Cancel scheduled suspend requests. */
5254 			pm_runtime_resume(netdev->dev.parent);
5255 
5256 			/* Checking if MAC is in DMoff state*/
5257 			if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID) {
5258 				pcim_state = er32(STATUS);
5259 				while (pcim_state & E1000_STATUS_PCIM_STATE) {
5260 					if (tries++ == dmoff_exit_timeout) {
5261 						e_dbg("Error in exiting dmoff\n");
5262 						break;
5263 					}
5264 					usleep_range(10000, 20000);
5265 					pcim_state = er32(STATUS);
5266 
5267 					/* Checking if MAC exited DMoff state */
5268 					if (!(pcim_state & E1000_STATUS_PCIM_STATE))
5269 						e1000_phy_hw_reset(&adapter->hw);
5270 				}
5271 			}
5272 
5273 			/* update snapshot of PHY registers on LSC */
5274 			e1000_phy_read_status(adapter);
5275 			mac->ops.get_link_up_info(&adapter->hw,
5276 						  &adapter->link_speed,
5277 						  &adapter->link_duplex);
5278 			e1000_print_link_info(adapter);
5279 
5280 			/* check if SmartSpeed worked */
5281 			e1000e_check_downshift(hw);
5282 			if (phy->speed_downgraded)
5283 				netdev_warn(netdev,
5284 					    "Link Speed was downgraded by SmartSpeed\n");
5285 
5286 			/* On supported PHYs, check for duplex mismatch only
5287 			 * if link has autonegotiated at 10/100 half
5288 			 */
5289 			if ((hw->phy.type == e1000_phy_igp_3 ||
5290 			     hw->phy.type == e1000_phy_bm) &&
5291 			    hw->mac.autoneg &&
5292 			    (adapter->link_speed == SPEED_10 ||
5293 			     adapter->link_speed == SPEED_100) &&
5294 			    (adapter->link_duplex == HALF_DUPLEX)) {
5295 				u16 autoneg_exp;
5296 
5297 				e1e_rphy(hw, MII_EXPANSION, &autoneg_exp);
5298 
5299 				if (!(autoneg_exp & EXPANSION_NWAY))
5300 					e_info("Autonegotiated half duplex but link partner cannot autoneg.  Try forcing full duplex if link gets many collisions.\n");
5301 			}
5302 
5303 			/* adjust timeout factor according to speed/duplex */
5304 			adapter->tx_timeout_factor = 1;
5305 			switch (adapter->link_speed) {
5306 			case SPEED_10:
5307 				txb2b = false;
5308 				adapter->tx_timeout_factor = 16;
5309 				break;
5310 			case SPEED_100:
5311 				txb2b = false;
5312 				adapter->tx_timeout_factor = 10;
5313 				break;
5314 			}
5315 
5316 			/* workaround: re-program speed mode bit after
5317 			 * link-up event
5318 			 */
5319 			if ((adapter->flags & FLAG_TARC_SPEED_MODE_BIT) &&
5320 			    !txb2b) {
5321 				u32 tarc0;
5322 
5323 				tarc0 = er32(TARC(0));
5324 				tarc0 &= ~SPEED_MODE_BIT;
5325 				ew32(TARC(0), tarc0);
5326 			}
5327 
5328 			/* enable transmits in the hardware, need to do this
5329 			 * after setting TARC(0)
5330 			 */
5331 			tctl = er32(TCTL);
5332 			tctl |= E1000_TCTL_EN;
5333 			ew32(TCTL, tctl);
5334 
5335 			/* Perform any post-link-up configuration before
5336 			 * reporting link up.
5337 			 */
5338 			if (phy->ops.cfg_on_link_up)
5339 				phy->ops.cfg_on_link_up(hw);
5340 
5341 			netif_wake_queue(netdev);
5342 			netif_carrier_on(netdev);
5343 
5344 			if (!test_bit(__E1000_DOWN, &adapter->state))
5345 				mod_timer(&adapter->phy_info_timer,
5346 					  round_jiffies(jiffies + 2 * HZ));
5347 		}
5348 	} else {
5349 		if (netif_carrier_ok(netdev)) {
5350 			adapter->link_speed = 0;
5351 			adapter->link_duplex = 0;
5352 			/* Link status message must follow this format */
5353 			netdev_info(netdev, "NIC Link is Down\n");
5354 			netif_carrier_off(netdev);
5355 			netif_stop_queue(netdev);
5356 			if (!test_bit(__E1000_DOWN, &adapter->state))
5357 				mod_timer(&adapter->phy_info_timer,
5358 					  round_jiffies(jiffies + 2 * HZ));
5359 
5360 			/* 8000ES2LAN requires a Rx packet buffer work-around
5361 			 * on link down event; reset the controller to flush
5362 			 * the Rx packet buffer.
5363 			 */
5364 			if (adapter->flags & FLAG_RX_NEEDS_RESTART)
5365 				adapter->flags |= FLAG_RESTART_NOW;
5366 			else
5367 				pm_schedule_suspend(netdev->dev.parent,
5368 						    LINK_TIMEOUT);
5369 		}
5370 	}
5371 
5372 link_up:
5373 	spin_lock(&adapter->stats64_lock);
5374 	e1000e_update_stats(adapter);
5375 
5376 	mac->tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
5377 	adapter->tpt_old = adapter->stats.tpt;
5378 	mac->collision_delta = adapter->stats.colc - adapter->colc_old;
5379 	adapter->colc_old = adapter->stats.colc;
5380 
5381 	adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
5382 	adapter->gorc_old = adapter->stats.gorc;
5383 	adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
5384 	adapter->gotc_old = adapter->stats.gotc;
5385 	spin_unlock(&adapter->stats64_lock);
5386 
5387 	/* If the link is lost the controller stops DMA, but
5388 	 * if there is queued Tx work it cannot be done.  So
5389 	 * reset the controller to flush the Tx packet buffers.
5390 	 */
5391 	if (!netif_carrier_ok(netdev) &&
5392 	    (e1000_desc_unused(tx_ring) + 1 < tx_ring->count))
5393 		adapter->flags |= FLAG_RESTART_NOW;
5394 
5395 	/* If reset is necessary, do it outside of interrupt context. */
5396 	if (adapter->flags & FLAG_RESTART_NOW) {
5397 		schedule_work(&adapter->reset_task);
5398 		/* return immediately since reset is imminent */
5399 		return;
5400 	}
5401 
5402 	e1000e_update_adaptive(&adapter->hw);
5403 
5404 	/* Simple mode for Interrupt Throttle Rate (ITR) */
5405 	if (adapter->itr_setting == 4) {
5406 		/* Symmetric Tx/Rx gets a reduced ITR=2000;
5407 		 * Total asymmetrical Tx or Rx gets ITR=8000;
5408 		 * everyone else is between 2000-8000.
5409 		 */
5410 		u32 goc = (adapter->gotc + adapter->gorc) / 10000;
5411 		u32 dif = (adapter->gotc > adapter->gorc ?
5412 			   adapter->gotc - adapter->gorc :
5413 			   adapter->gorc - adapter->gotc) / 10000;
5414 		u32 itr = goc > 0 ? (dif * 6000 / goc + 2000) : 8000;
5415 
5416 		e1000e_write_itr(adapter, itr);
5417 	}
5418 
5419 	/* Cause software interrupt to ensure Rx ring is cleaned */
5420 	if (adapter->msix_entries)
5421 		ew32(ICS, adapter->rx_ring->ims_val);
5422 	else
5423 		ew32(ICS, E1000_ICS_RXDMT0);
5424 
5425 	/* flush pending descriptors to memory before detecting Tx hang */
5426 	e1000e_flush_descriptors(adapter);
5427 
5428 	/* Force detection of hung controller every watchdog period */
5429 	adapter->detect_tx_hung = true;
5430 
5431 	/* With 82571 controllers, LAA may be overwritten due to controller
5432 	 * reset from the other port. Set the appropriate LAA in RAR[0]
5433 	 */
5434 	if (e1000e_get_laa_state_82571(hw))
5435 		hw->mac.ops.rar_set(hw, adapter->hw.mac.addr, 0);
5436 
5437 	if (adapter->flags2 & FLAG2_CHECK_PHY_HANG)
5438 		e1000e_check_82574_phy_workaround(adapter);
5439 
5440 	/* Clear valid timestamp stuck in RXSTMPL/H due to a Rx error */
5441 	if (adapter->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE) {
5442 		if ((adapter->flags2 & FLAG2_CHECK_RX_HWTSTAMP) &&
5443 		    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID)) {
5444 			er32(RXSTMPH);
5445 			adapter->rx_hwtstamp_cleared++;
5446 		} else {
5447 			adapter->flags2 |= FLAG2_CHECK_RX_HWTSTAMP;
5448 		}
5449 	}
5450 
5451 	/* Reset the timer */
5452 	if (!test_bit(__E1000_DOWN, &adapter->state))
5453 		mod_timer(&adapter->watchdog_timer,
5454 			  round_jiffies(jiffies + 2 * HZ));
5455 }
5456 
5457 #define E1000_TX_FLAGS_CSUM		0x00000001
5458 #define E1000_TX_FLAGS_VLAN		0x00000002
5459 #define E1000_TX_FLAGS_TSO		0x00000004
5460 #define E1000_TX_FLAGS_IPV4		0x00000008
5461 #define E1000_TX_FLAGS_NO_FCS		0x00000010
5462 #define E1000_TX_FLAGS_HWTSTAMP		0x00000020
5463 #define E1000_TX_FLAGS_VLAN_MASK	0xffff0000
5464 #define E1000_TX_FLAGS_VLAN_SHIFT	16
5465 
5466 static int e1000_tso(struct e1000_ring *tx_ring, struct sk_buff *skb,
5467 		     __be16 protocol)
5468 {
5469 	struct e1000_context_desc *context_desc;
5470 	struct e1000_buffer *buffer_info;
5471 	unsigned int i;
5472 	u32 cmd_length = 0;
5473 	u16 ipcse = 0, mss;
5474 	u8 ipcss, ipcso, tucss, tucso, hdr_len;
5475 	int err;
5476 
5477 	if (!skb_is_gso(skb))
5478 		return 0;
5479 
5480 	err = skb_cow_head(skb, 0);
5481 	if (err < 0)
5482 		return err;
5483 
5484 	hdr_len = skb_tcp_all_headers(skb);
5485 	mss = skb_shinfo(skb)->gso_size;
5486 	if (protocol == htons(ETH_P_IP)) {
5487 		struct iphdr *iph = ip_hdr(skb);
5488 		iph->tot_len = 0;
5489 		iph->check = 0;
5490 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
5491 							 0, IPPROTO_TCP, 0);
5492 		cmd_length = E1000_TXD_CMD_IP;
5493 		ipcse = skb_transport_offset(skb) - 1;
5494 	} else if (skb_is_gso_v6(skb)) {
5495 		tcp_v6_gso_csum_prep(skb);
5496 		ipcse = 0;
5497 	}
5498 	ipcss = skb_network_offset(skb);
5499 	ipcso = (void *)&(ip_hdr(skb)->check) - (void *)skb->data;
5500 	tucss = skb_transport_offset(skb);
5501 	tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
5502 
5503 	cmd_length |= (E1000_TXD_CMD_DEXT | E1000_TXD_CMD_TSE |
5504 		       E1000_TXD_CMD_TCP | (skb->len - (hdr_len)));
5505 
5506 	i = tx_ring->next_to_use;
5507 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5508 	buffer_info = &tx_ring->buffer_info[i];
5509 
5510 	context_desc->lower_setup.ip_fields.ipcss = ipcss;
5511 	context_desc->lower_setup.ip_fields.ipcso = ipcso;
5512 	context_desc->lower_setup.ip_fields.ipcse = cpu_to_le16(ipcse);
5513 	context_desc->upper_setup.tcp_fields.tucss = tucss;
5514 	context_desc->upper_setup.tcp_fields.tucso = tucso;
5515 	context_desc->upper_setup.tcp_fields.tucse = 0;
5516 	context_desc->tcp_seg_setup.fields.mss = cpu_to_le16(mss);
5517 	context_desc->tcp_seg_setup.fields.hdr_len = hdr_len;
5518 	context_desc->cmd_and_length = cpu_to_le32(cmd_length);
5519 
5520 	buffer_info->time_stamp = jiffies;
5521 	buffer_info->next_to_watch = i;
5522 
5523 	i++;
5524 	if (i == tx_ring->count)
5525 		i = 0;
5526 	tx_ring->next_to_use = i;
5527 
5528 	return 1;
5529 }
5530 
5531 static bool e1000_tx_csum(struct e1000_ring *tx_ring, struct sk_buff *skb,
5532 			  __be16 protocol)
5533 {
5534 	struct e1000_adapter *adapter = tx_ring->adapter;
5535 	struct e1000_context_desc *context_desc;
5536 	struct e1000_buffer *buffer_info;
5537 	unsigned int i;
5538 	u8 css;
5539 	u32 cmd_len = E1000_TXD_CMD_DEXT;
5540 
5541 	if (skb->ip_summed != CHECKSUM_PARTIAL)
5542 		return false;
5543 
5544 	switch (protocol) {
5545 	case cpu_to_be16(ETH_P_IP):
5546 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
5547 			cmd_len |= E1000_TXD_CMD_TCP;
5548 		break;
5549 	case cpu_to_be16(ETH_P_IPV6):
5550 		/* XXX not handling all IPV6 headers */
5551 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
5552 			cmd_len |= E1000_TXD_CMD_TCP;
5553 		break;
5554 	default:
5555 		if (unlikely(net_ratelimit()))
5556 			e_warn("checksum_partial proto=%x!\n",
5557 			       be16_to_cpu(protocol));
5558 		break;
5559 	}
5560 
5561 	css = skb_checksum_start_offset(skb);
5562 
5563 	i = tx_ring->next_to_use;
5564 	buffer_info = &tx_ring->buffer_info[i];
5565 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5566 
5567 	context_desc->lower_setup.ip_config = 0;
5568 	context_desc->upper_setup.tcp_fields.tucss = css;
5569 	context_desc->upper_setup.tcp_fields.tucso = css + skb->csum_offset;
5570 	context_desc->upper_setup.tcp_fields.tucse = 0;
5571 	context_desc->tcp_seg_setup.data = 0;
5572 	context_desc->cmd_and_length = cpu_to_le32(cmd_len);
5573 
5574 	buffer_info->time_stamp = jiffies;
5575 	buffer_info->next_to_watch = i;
5576 
5577 	i++;
5578 	if (i == tx_ring->count)
5579 		i = 0;
5580 	tx_ring->next_to_use = i;
5581 
5582 	return true;
5583 }
5584 
5585 static int e1000_tx_map(struct e1000_ring *tx_ring, struct sk_buff *skb,
5586 			unsigned int first, unsigned int max_per_txd,
5587 			unsigned int nr_frags)
5588 {
5589 	struct e1000_adapter *adapter = tx_ring->adapter;
5590 	struct pci_dev *pdev = adapter->pdev;
5591 	struct e1000_buffer *buffer_info;
5592 	unsigned int len = skb_headlen(skb);
5593 	unsigned int offset = 0, size, count = 0, i;
5594 	unsigned int f, bytecount, segs;
5595 
5596 	i = tx_ring->next_to_use;
5597 
5598 	while (len) {
5599 		buffer_info = &tx_ring->buffer_info[i];
5600 		size = min(len, max_per_txd);
5601 
5602 		buffer_info->length = size;
5603 		buffer_info->time_stamp = jiffies;
5604 		buffer_info->next_to_watch = i;
5605 		buffer_info->dma = dma_map_single(&pdev->dev,
5606 						  skb->data + offset,
5607 						  size, DMA_TO_DEVICE);
5608 		buffer_info->mapped_as_page = false;
5609 		if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5610 			goto dma_error;
5611 
5612 		len -= size;
5613 		offset += size;
5614 		count++;
5615 
5616 		if (len) {
5617 			i++;
5618 			if (i == tx_ring->count)
5619 				i = 0;
5620 		}
5621 	}
5622 
5623 	for (f = 0; f < nr_frags; f++) {
5624 		const skb_frag_t *frag = &skb_shinfo(skb)->frags[f];
5625 
5626 		len = skb_frag_size(frag);
5627 		offset = 0;
5628 
5629 		while (len) {
5630 			i++;
5631 			if (i == tx_ring->count)
5632 				i = 0;
5633 
5634 			buffer_info = &tx_ring->buffer_info[i];
5635 			size = min(len, max_per_txd);
5636 
5637 			buffer_info->length = size;
5638 			buffer_info->time_stamp = jiffies;
5639 			buffer_info->next_to_watch = i;
5640 			buffer_info->dma = skb_frag_dma_map(&pdev->dev, frag,
5641 							    offset, size,
5642 							    DMA_TO_DEVICE);
5643 			buffer_info->mapped_as_page = true;
5644 			if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5645 				goto dma_error;
5646 
5647 			len -= size;
5648 			offset += size;
5649 			count++;
5650 		}
5651 	}
5652 
5653 	segs = skb_shinfo(skb)->gso_segs ? : 1;
5654 	/* multiply data chunks by size of headers */
5655 	bytecount = ((segs - 1) * skb_headlen(skb)) + skb->len;
5656 
5657 	tx_ring->buffer_info[i].skb = skb;
5658 	tx_ring->buffer_info[i].segs = segs;
5659 	tx_ring->buffer_info[i].bytecount = bytecount;
5660 	tx_ring->buffer_info[first].next_to_watch = i;
5661 
5662 	return count;
5663 
5664 dma_error:
5665 	dev_err(&pdev->dev, "Tx DMA map failed\n");
5666 	buffer_info->dma = 0;
5667 
5668 	while (count--) {
5669 		if (i == 0)
5670 			i += tx_ring->count;
5671 		i--;
5672 		buffer_info = &tx_ring->buffer_info[i];
5673 		e1000_put_txbuf(tx_ring, buffer_info, true);
5674 	}
5675 
5676 	return 0;
5677 }
5678 
5679 static void e1000_tx_queue(struct e1000_ring *tx_ring, int tx_flags, int count)
5680 {
5681 	struct e1000_adapter *adapter = tx_ring->adapter;
5682 	struct e1000_tx_desc *tx_desc = NULL;
5683 	struct e1000_buffer *buffer_info;
5684 	u32 txd_upper = 0, txd_lower = E1000_TXD_CMD_IFCS;
5685 	unsigned int i;
5686 
5687 	if (tx_flags & E1000_TX_FLAGS_TSO) {
5688 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D |
5689 		    E1000_TXD_CMD_TSE;
5690 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5691 
5692 		if (tx_flags & E1000_TX_FLAGS_IPV4)
5693 			txd_upper |= E1000_TXD_POPTS_IXSM << 8;
5694 	}
5695 
5696 	if (tx_flags & E1000_TX_FLAGS_CSUM) {
5697 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5698 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5699 	}
5700 
5701 	if (tx_flags & E1000_TX_FLAGS_VLAN) {
5702 		txd_lower |= E1000_TXD_CMD_VLE;
5703 		txd_upper |= (tx_flags & E1000_TX_FLAGS_VLAN_MASK);
5704 	}
5705 
5706 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5707 		txd_lower &= ~(E1000_TXD_CMD_IFCS);
5708 
5709 	if (unlikely(tx_flags & E1000_TX_FLAGS_HWTSTAMP)) {
5710 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5711 		txd_upper |= E1000_TXD_EXTCMD_TSTAMP;
5712 	}
5713 
5714 	i = tx_ring->next_to_use;
5715 
5716 	do {
5717 		buffer_info = &tx_ring->buffer_info[i];
5718 		tx_desc = E1000_TX_DESC(*tx_ring, i);
5719 		tx_desc->buffer_addr = cpu_to_le64(buffer_info->dma);
5720 		tx_desc->lower.data = cpu_to_le32(txd_lower |
5721 						  buffer_info->length);
5722 		tx_desc->upper.data = cpu_to_le32(txd_upper);
5723 
5724 		i++;
5725 		if (i == tx_ring->count)
5726 			i = 0;
5727 	} while (--count > 0);
5728 
5729 	tx_desc->lower.data |= cpu_to_le32(adapter->txd_cmd);
5730 
5731 	/* txd_cmd re-enables FCS, so we'll re-disable it here as desired. */
5732 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5733 		tx_desc->lower.data &= ~(cpu_to_le32(E1000_TXD_CMD_IFCS));
5734 
5735 	/* Force memory writes to complete before letting h/w
5736 	 * know there are new descriptors to fetch.  (Only
5737 	 * applicable for weak-ordered memory model archs,
5738 	 * such as IA-64).
5739 	 */
5740 	wmb();
5741 
5742 	tx_ring->next_to_use = i;
5743 }
5744 
5745 #define MINIMUM_DHCP_PACKET_SIZE 282
5746 static int e1000_transfer_dhcp_info(struct e1000_adapter *adapter,
5747 				    struct sk_buff *skb)
5748 {
5749 	struct e1000_hw *hw = &adapter->hw;
5750 	u16 length, offset;
5751 
5752 	if (skb_vlan_tag_present(skb) &&
5753 	    !((skb_vlan_tag_get(skb) == adapter->hw.mng_cookie.vlan_id) &&
5754 	      (adapter->hw.mng_cookie.status &
5755 	       E1000_MNG_DHCP_COOKIE_STATUS_VLAN)))
5756 		return 0;
5757 
5758 	if (skb->len <= MINIMUM_DHCP_PACKET_SIZE)
5759 		return 0;
5760 
5761 	if (((struct ethhdr *)skb->data)->h_proto != htons(ETH_P_IP))
5762 		return 0;
5763 
5764 	{
5765 		const struct iphdr *ip = (struct iphdr *)((u8 *)skb->data + 14);
5766 		struct udphdr *udp;
5767 
5768 		if (ip->protocol != IPPROTO_UDP)
5769 			return 0;
5770 
5771 		udp = (struct udphdr *)((u8 *)ip + (ip->ihl << 2));
5772 		if (ntohs(udp->dest) != 67)
5773 			return 0;
5774 
5775 		offset = (u8 *)udp + 8 - skb->data;
5776 		length = skb->len - offset;
5777 		return e1000e_mng_write_dhcp_info(hw, (u8 *)udp + 8, length);
5778 	}
5779 
5780 	return 0;
5781 }
5782 
5783 static int __e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5784 {
5785 	struct e1000_adapter *adapter = tx_ring->adapter;
5786 
5787 	netif_stop_queue(adapter->netdev);
5788 	/* Herbert's original patch had:
5789 	 *  smp_mb__after_netif_stop_queue();
5790 	 * but since that doesn't exist yet, just open code it.
5791 	 */
5792 	smp_mb();
5793 
5794 	/* We need to check again in a case another CPU has just
5795 	 * made room available.
5796 	 */
5797 	if (e1000_desc_unused(tx_ring) < size)
5798 		return -EBUSY;
5799 
5800 	/* A reprieve! */
5801 	netif_start_queue(adapter->netdev);
5802 	++adapter->restart_queue;
5803 	return 0;
5804 }
5805 
5806 static int e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5807 {
5808 	BUG_ON(size > tx_ring->count);
5809 
5810 	if (e1000_desc_unused(tx_ring) >= size)
5811 		return 0;
5812 	return __e1000_maybe_stop_tx(tx_ring, size);
5813 }
5814 
5815 static netdev_tx_t e1000_xmit_frame(struct sk_buff *skb,
5816 				    struct net_device *netdev)
5817 {
5818 	struct e1000_adapter *adapter = netdev_priv(netdev);
5819 	struct e1000_ring *tx_ring = adapter->tx_ring;
5820 	unsigned int first;
5821 	unsigned int tx_flags = 0;
5822 	unsigned int len = skb_headlen(skb);
5823 	unsigned int nr_frags;
5824 	unsigned int mss;
5825 	int count = 0;
5826 	int tso;
5827 	unsigned int f;
5828 	__be16 protocol = vlan_get_protocol(skb);
5829 
5830 	if (test_bit(__E1000_DOWN, &adapter->state)) {
5831 		dev_kfree_skb_any(skb);
5832 		return NETDEV_TX_OK;
5833 	}
5834 
5835 	if (skb->len <= 0) {
5836 		dev_kfree_skb_any(skb);
5837 		return NETDEV_TX_OK;
5838 	}
5839 
5840 	/* The minimum packet size with TCTL.PSP set is 17 bytes so
5841 	 * pad skb in order to meet this minimum size requirement
5842 	 */
5843 	if (skb_put_padto(skb, 17))
5844 		return NETDEV_TX_OK;
5845 
5846 	mss = skb_shinfo(skb)->gso_size;
5847 	if (mss) {
5848 		u8 hdr_len;
5849 
5850 		/* TSO Workaround for 82571/2/3 Controllers -- if skb->data
5851 		 * points to just header, pull a few bytes of payload from
5852 		 * frags into skb->data
5853 		 */
5854 		hdr_len = skb_tcp_all_headers(skb);
5855 		/* we do this workaround for ES2LAN, but it is un-necessary,
5856 		 * avoiding it could save a lot of cycles
5857 		 */
5858 		if (skb->data_len && (hdr_len == len)) {
5859 			unsigned int pull_size;
5860 
5861 			pull_size = min_t(unsigned int, 4, skb->data_len);
5862 			if (!__pskb_pull_tail(skb, pull_size)) {
5863 				e_err("__pskb_pull_tail failed.\n");
5864 				dev_kfree_skb_any(skb);
5865 				return NETDEV_TX_OK;
5866 			}
5867 			len = skb_headlen(skb);
5868 		}
5869 	}
5870 
5871 	/* reserve a descriptor for the offload context */
5872 	if ((mss) || (skb->ip_summed == CHECKSUM_PARTIAL))
5873 		count++;
5874 	count++;
5875 
5876 	count += DIV_ROUND_UP(len, adapter->tx_fifo_limit);
5877 
5878 	nr_frags = skb_shinfo(skb)->nr_frags;
5879 	for (f = 0; f < nr_frags; f++)
5880 		count += DIV_ROUND_UP(skb_frag_size(&skb_shinfo(skb)->frags[f]),
5881 				      adapter->tx_fifo_limit);
5882 
5883 	if (adapter->hw.mac.tx_pkt_filtering)
5884 		e1000_transfer_dhcp_info(adapter, skb);
5885 
5886 	/* need: count + 2 desc gap to keep tail from touching
5887 	 * head, otherwise try next time
5888 	 */
5889 	if (e1000_maybe_stop_tx(tx_ring, count + 2))
5890 		return NETDEV_TX_BUSY;
5891 
5892 	if (skb_vlan_tag_present(skb)) {
5893 		tx_flags |= E1000_TX_FLAGS_VLAN;
5894 		tx_flags |= (skb_vlan_tag_get(skb) <<
5895 			     E1000_TX_FLAGS_VLAN_SHIFT);
5896 	}
5897 
5898 	first = tx_ring->next_to_use;
5899 
5900 	tso = e1000_tso(tx_ring, skb, protocol);
5901 	if (tso < 0) {
5902 		dev_kfree_skb_any(skb);
5903 		return NETDEV_TX_OK;
5904 	}
5905 
5906 	if (tso)
5907 		tx_flags |= E1000_TX_FLAGS_TSO;
5908 	else if (e1000_tx_csum(tx_ring, skb, protocol))
5909 		tx_flags |= E1000_TX_FLAGS_CSUM;
5910 
5911 	/* Old method was to assume IPv4 packet by default if TSO was enabled.
5912 	 * 82571 hardware supports TSO capabilities for IPv6 as well...
5913 	 * no longer assume, we must.
5914 	 */
5915 	if (protocol == htons(ETH_P_IP))
5916 		tx_flags |= E1000_TX_FLAGS_IPV4;
5917 
5918 	if (unlikely(skb->no_fcs))
5919 		tx_flags |= E1000_TX_FLAGS_NO_FCS;
5920 
5921 	/* if count is 0 then mapping error has occurred */
5922 	count = e1000_tx_map(tx_ring, skb, first, adapter->tx_fifo_limit,
5923 			     nr_frags);
5924 	if (count) {
5925 		if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
5926 		    (adapter->flags & FLAG_HAS_HW_TIMESTAMP)) {
5927 			if (!adapter->tx_hwtstamp_skb) {
5928 				skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5929 				tx_flags |= E1000_TX_FLAGS_HWTSTAMP;
5930 				adapter->tx_hwtstamp_skb = skb_get(skb);
5931 				adapter->tx_hwtstamp_start = jiffies;
5932 				schedule_work(&adapter->tx_hwtstamp_work);
5933 			} else {
5934 				adapter->tx_hwtstamp_skipped++;
5935 			}
5936 		}
5937 
5938 		skb_tx_timestamp(skb);
5939 
5940 		netdev_sent_queue(netdev, skb->len);
5941 		e1000_tx_queue(tx_ring, tx_flags, count);
5942 		/* Make sure there is space in the ring for the next send. */
5943 		e1000_maybe_stop_tx(tx_ring,
5944 				    ((MAX_SKB_FRAGS + 1) *
5945 				     DIV_ROUND_UP(PAGE_SIZE,
5946 						  adapter->tx_fifo_limit) + 4));
5947 
5948 		if (!netdev_xmit_more() ||
5949 		    netif_xmit_stopped(netdev_get_tx_queue(netdev, 0))) {
5950 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
5951 				e1000e_update_tdt_wa(tx_ring,
5952 						     tx_ring->next_to_use);
5953 			else
5954 				writel(tx_ring->next_to_use, tx_ring->tail);
5955 		}
5956 	} else {
5957 		dev_kfree_skb_any(skb);
5958 		tx_ring->buffer_info[first].time_stamp = 0;
5959 		tx_ring->next_to_use = first;
5960 	}
5961 
5962 	return NETDEV_TX_OK;
5963 }
5964 
5965 /**
5966  * e1000_tx_timeout - Respond to a Tx Hang
5967  * @netdev: network interface device structure
5968  * @txqueue: index of the hung queue (unused)
5969  **/
5970 static void e1000_tx_timeout(struct net_device *netdev, unsigned int __always_unused txqueue)
5971 {
5972 	struct e1000_adapter *adapter = netdev_priv(netdev);
5973 
5974 	/* Do the reset outside of interrupt context */
5975 	adapter->tx_timeout_count++;
5976 	schedule_work(&adapter->reset_task);
5977 }
5978 
5979 static void e1000_reset_task(struct work_struct *work)
5980 {
5981 	struct e1000_adapter *adapter;
5982 	adapter = container_of(work, struct e1000_adapter, reset_task);
5983 
5984 	rtnl_lock();
5985 	/* don't run the task if already down */
5986 	if (test_bit(__E1000_DOWN, &adapter->state)) {
5987 		rtnl_unlock();
5988 		return;
5989 	}
5990 
5991 	if (!(adapter->flags & FLAG_RESTART_NOW)) {
5992 		e1000e_dump(adapter);
5993 		e_err("Reset adapter unexpectedly\n");
5994 	}
5995 	e1000e_reinit_locked(adapter);
5996 	rtnl_unlock();
5997 }
5998 
5999 /**
6000  * e1000e_get_stats64 - Get System Network Statistics
6001  * @netdev: network interface device structure
6002  * @stats: rtnl_link_stats64 pointer
6003  *
6004  * Returns the address of the device statistics structure.
6005  **/
6006 void e1000e_get_stats64(struct net_device *netdev,
6007 			struct rtnl_link_stats64 *stats)
6008 {
6009 	struct e1000_adapter *adapter = netdev_priv(netdev);
6010 
6011 	spin_lock(&adapter->stats64_lock);
6012 	e1000e_update_stats(adapter);
6013 	/* Fill out the OS statistics structure */
6014 	stats->rx_bytes = adapter->stats.gorc;
6015 	stats->rx_packets = adapter->stats.gprc;
6016 	stats->tx_bytes = adapter->stats.gotc;
6017 	stats->tx_packets = adapter->stats.gptc;
6018 	stats->multicast = adapter->stats.mprc;
6019 	stats->collisions = adapter->stats.colc;
6020 
6021 	/* Rx Errors */
6022 
6023 	/* RLEC on some newer hardware can be incorrect so build
6024 	 * our own version based on RUC and ROC
6025 	 */
6026 	stats->rx_errors = adapter->stats.rxerrc +
6027 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
6028 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
6029 	stats->rx_length_errors = adapter->stats.ruc + adapter->stats.roc;
6030 	stats->rx_crc_errors = adapter->stats.crcerrs;
6031 	stats->rx_frame_errors = adapter->stats.algnerrc;
6032 	stats->rx_missed_errors = adapter->stats.mpc;
6033 
6034 	/* Tx Errors */
6035 	stats->tx_errors = adapter->stats.ecol + adapter->stats.latecol;
6036 	stats->tx_aborted_errors = adapter->stats.ecol;
6037 	stats->tx_window_errors = adapter->stats.latecol;
6038 	stats->tx_carrier_errors = adapter->stats.tncrs;
6039 
6040 	/* Tx Dropped needs to be maintained elsewhere */
6041 
6042 	spin_unlock(&adapter->stats64_lock);
6043 }
6044 
6045 /**
6046  * e1000_change_mtu - Change the Maximum Transfer Unit
6047  * @netdev: network interface device structure
6048  * @new_mtu: new value for maximum frame size
6049  *
6050  * Returns 0 on success, negative on failure
6051  **/
6052 static int e1000_change_mtu(struct net_device *netdev, int new_mtu)
6053 {
6054 	struct e1000_adapter *adapter = netdev_priv(netdev);
6055 	int max_frame = new_mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
6056 
6057 	/* Jumbo frame support */
6058 	if ((new_mtu > ETH_DATA_LEN) &&
6059 	    !(adapter->flags & FLAG_HAS_JUMBO_FRAMES)) {
6060 		e_err("Jumbo Frames not supported.\n");
6061 		return -EINVAL;
6062 	}
6063 
6064 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
6065 	if ((adapter->hw.mac.type >= e1000_pch2lan) &&
6066 	    !(adapter->flags2 & FLAG2_CRC_STRIPPING) &&
6067 	    (new_mtu > ETH_DATA_LEN)) {
6068 		e_err("Jumbo Frames not supported on this device when CRC stripping is disabled.\n");
6069 		return -EINVAL;
6070 	}
6071 
6072 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
6073 		usleep_range(1000, 1100);
6074 	/* e1000e_down -> e1000e_reset dependent on max_frame_size & mtu */
6075 	adapter->max_frame_size = max_frame;
6076 	netdev_dbg(netdev, "changing MTU from %d to %d\n",
6077 		   netdev->mtu, new_mtu);
6078 	WRITE_ONCE(netdev->mtu, new_mtu);
6079 
6080 	pm_runtime_get_sync(netdev->dev.parent);
6081 
6082 	if (netif_running(netdev))
6083 		e1000e_down(adapter, true);
6084 
6085 	/* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
6086 	 * means we reserve 2 more, this pushes us to allocate from the next
6087 	 * larger slab size.
6088 	 * i.e. RXBUFFER_2048 --> size-4096 slab
6089 	 * However with the new *_jumbo_rx* routines, jumbo receives will use
6090 	 * fragmented skbs
6091 	 */
6092 
6093 	if (max_frame <= 2048)
6094 		adapter->rx_buffer_len = 2048;
6095 	else
6096 		adapter->rx_buffer_len = 4096;
6097 
6098 	/* adjust allocation if LPE protects us, and we aren't using SBP */
6099 	if (max_frame <= (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN))
6100 		adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
6101 
6102 	if (netif_running(netdev))
6103 		e1000e_up(adapter);
6104 	else
6105 		e1000e_reset(adapter);
6106 
6107 	pm_runtime_put_sync(netdev->dev.parent);
6108 
6109 	clear_bit(__E1000_RESETTING, &adapter->state);
6110 
6111 	return 0;
6112 }
6113 
6114 static int e1000_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
6115 {
6116 	struct e1000_adapter *adapter = netdev_priv(netdev);
6117 	struct mii_ioctl_data *data = if_mii(ifr);
6118 
6119 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
6120 		return -EOPNOTSUPP;
6121 
6122 	switch (cmd) {
6123 	case SIOCGMIIPHY:
6124 		data->phy_id = adapter->hw.phy.addr;
6125 		break;
6126 	case SIOCGMIIREG:
6127 		e1000_phy_read_status(adapter);
6128 
6129 		switch (data->reg_num & 0x1F) {
6130 		case MII_BMCR:
6131 			data->val_out = adapter->phy_regs.bmcr;
6132 			break;
6133 		case MII_BMSR:
6134 			data->val_out = adapter->phy_regs.bmsr;
6135 			break;
6136 		case MII_PHYSID1:
6137 			data->val_out = (adapter->hw.phy.id >> 16);
6138 			break;
6139 		case MII_PHYSID2:
6140 			data->val_out = (adapter->hw.phy.id & 0xFFFF);
6141 			break;
6142 		case MII_ADVERTISE:
6143 			data->val_out = adapter->phy_regs.advertise;
6144 			break;
6145 		case MII_LPA:
6146 			data->val_out = adapter->phy_regs.lpa;
6147 			break;
6148 		case MII_EXPANSION:
6149 			data->val_out = adapter->phy_regs.expansion;
6150 			break;
6151 		case MII_CTRL1000:
6152 			data->val_out = adapter->phy_regs.ctrl1000;
6153 			break;
6154 		case MII_STAT1000:
6155 			data->val_out = adapter->phy_regs.stat1000;
6156 			break;
6157 		case MII_ESTATUS:
6158 			data->val_out = adapter->phy_regs.estatus;
6159 			break;
6160 		default:
6161 			return -EIO;
6162 		}
6163 		break;
6164 	case SIOCSMIIREG:
6165 	default:
6166 		return -EOPNOTSUPP;
6167 	}
6168 	return 0;
6169 }
6170 
6171 /**
6172  * e1000e_hwtstamp_set - control hardware time stamping
6173  * @netdev: network interface device structure
6174  * @config: timestamp configuration
6175  * @extack: netlink extended ACK report
6176  *
6177  * Outgoing time stamping can be enabled and disabled. Play nice and
6178  * disable it when requested, although it shouldn't cause any overhead
6179  * when no packet needs it. At most one packet in the queue may be
6180  * marked for time stamping, otherwise it would be impossible to tell
6181  * for sure to which packet the hardware time stamp belongs.
6182  *
6183  * Incoming time stamping has to be configured via the hardware filters.
6184  * Not all combinations are supported, in particular event type has to be
6185  * specified. Matching the kind of event packet is not supported, with the
6186  * exception of "all V2 events regardless of level 2 or 4".
6187  **/
6188 static int e1000e_hwtstamp_set(struct net_device *netdev,
6189 			       struct kernel_hwtstamp_config *config,
6190 			       struct netlink_ext_ack *extack)
6191 {
6192 	struct e1000_adapter *adapter = netdev_priv(netdev);
6193 	int ret_val;
6194 
6195 	ret_val = e1000e_config_hwtstamp(adapter, config, extack);
6196 	if (ret_val)
6197 		return ret_val;
6198 
6199 	switch (config->rx_filter) {
6200 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
6201 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
6202 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
6203 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
6204 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
6205 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
6206 		/* With V2 type filters which specify a Sync or Delay Request,
6207 		 * Path Delay Request/Response messages are also time stamped
6208 		 * by hardware so notify the caller the requested packets plus
6209 		 * some others are time stamped.
6210 		 */
6211 		config->rx_filter = HWTSTAMP_FILTER_SOME;
6212 		break;
6213 	default:
6214 		break;
6215 	}
6216 
6217 	return 0;
6218 }
6219 
6220 static int e1000e_hwtstamp_get(struct net_device *netdev,
6221 			       struct kernel_hwtstamp_config *kernel_config)
6222 {
6223 	struct e1000_adapter *adapter = netdev_priv(netdev);
6224 
6225 	*kernel_config = adapter->hwtstamp_config;
6226 
6227 	return 0;
6228 }
6229 
6230 static int e1000_init_phy_wakeup(struct e1000_adapter *adapter, u32 wufc)
6231 {
6232 	struct e1000_hw *hw = &adapter->hw;
6233 	u32 i, mac_reg, wuc;
6234 	u16 phy_reg, wuc_enable;
6235 	int retval;
6236 
6237 	/* copy MAC RARs to PHY RARs */
6238 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
6239 
6240 	retval = hw->phy.ops.acquire(hw);
6241 	if (retval) {
6242 		e_err("Could not acquire PHY\n");
6243 		return retval;
6244 	}
6245 
6246 	/* Enable access to wakeup registers on and set page to BM_WUC_PAGE */
6247 	retval = e1000_enable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6248 	if (retval)
6249 		goto release;
6250 
6251 	/* copy MAC MTA to PHY MTA - only needed for pchlan */
6252 	for (i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
6253 		mac_reg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
6254 		hw->phy.ops.write_reg_page(hw, BM_MTA(i),
6255 					   (u16)(mac_reg & 0xFFFF));
6256 		hw->phy.ops.write_reg_page(hw, BM_MTA(i) + 1,
6257 					   (u16)((mac_reg >> 16) & 0xFFFF));
6258 	}
6259 
6260 	/* configure PHY Rx Control register */
6261 	hw->phy.ops.read_reg_page(&adapter->hw, BM_RCTL, &phy_reg);
6262 	mac_reg = er32(RCTL);
6263 	if (mac_reg & E1000_RCTL_UPE)
6264 		phy_reg |= BM_RCTL_UPE;
6265 	if (mac_reg & E1000_RCTL_MPE)
6266 		phy_reg |= BM_RCTL_MPE;
6267 	phy_reg &= ~(BM_RCTL_MO_MASK);
6268 	if (mac_reg & E1000_RCTL_MO_3)
6269 		phy_reg |= (FIELD_GET(E1000_RCTL_MO_3, mac_reg)
6270 			    << BM_RCTL_MO_SHIFT);
6271 	if (mac_reg & E1000_RCTL_BAM)
6272 		phy_reg |= BM_RCTL_BAM;
6273 	if (mac_reg & E1000_RCTL_PMCF)
6274 		phy_reg |= BM_RCTL_PMCF;
6275 	mac_reg = er32(CTRL);
6276 	if (mac_reg & E1000_CTRL_RFCE)
6277 		phy_reg |= BM_RCTL_RFCE;
6278 	hw->phy.ops.write_reg_page(&adapter->hw, BM_RCTL, phy_reg);
6279 
6280 	wuc = E1000_WUC_PME_EN;
6281 	if (wufc & (E1000_WUFC_MAG | E1000_WUFC_LNKC))
6282 		wuc |= E1000_WUC_APME;
6283 
6284 	/* enable PHY wakeup in MAC register */
6285 	ew32(WUFC, wufc);
6286 	ew32(WUC, (E1000_WUC_PHY_WAKE | E1000_WUC_APMPME |
6287 		   E1000_WUC_PME_STATUS | wuc));
6288 
6289 	/* configure and enable PHY wakeup in PHY registers */
6290 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUFC, wufc);
6291 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUC, wuc);
6292 
6293 	/* activate PHY wakeup */
6294 	wuc_enable |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
6295 	retval = e1000_disable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6296 	if (retval)
6297 		e_err("Could not set PHY Host Wakeup bit\n");
6298 release:
6299 	hw->phy.ops.release(hw);
6300 
6301 	return retval;
6302 }
6303 
6304 static void e1000e_flush_lpic(struct pci_dev *pdev)
6305 {
6306 	struct net_device *netdev = pci_get_drvdata(pdev);
6307 	struct e1000_adapter *adapter = netdev_priv(netdev);
6308 	struct e1000_hw *hw = &adapter->hw;
6309 	u32 ret_val;
6310 
6311 	pm_runtime_get_sync(netdev->dev.parent);
6312 
6313 	ret_val = hw->phy.ops.acquire(hw);
6314 	if (ret_val)
6315 		goto fl_out;
6316 
6317 	pr_info("EEE TX LPI TIMER: %08X\n",
6318 		er32(LPIC) >> E1000_LPIC_LPIET_SHIFT);
6319 
6320 	hw->phy.ops.release(hw);
6321 
6322 fl_out:
6323 	pm_runtime_put_sync(netdev->dev.parent);
6324 }
6325 
6326 /* S0ix implementation */
6327 static void e1000e_s0ix_entry_flow(struct e1000_adapter *adapter)
6328 {
6329 	struct e1000_hw *hw = &adapter->hw;
6330 	u32 mac_data;
6331 	u16 phy_data;
6332 
6333 	if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID &&
6334 	    hw->mac.type >= e1000_pch_adp) {
6335 		/* Request ME configure the device for S0ix */
6336 		mac_data = er32(H2ME);
6337 		mac_data |= E1000_H2ME_START_DPG;
6338 		mac_data &= ~E1000_H2ME_EXIT_DPG;
6339 		trace_e1000e_trace_mac_register(mac_data);
6340 		ew32(H2ME, mac_data);
6341 	} else {
6342 		/* Request driver configure the device to S0ix */
6343 		/* Disable the periodic inband message,
6344 		 * don't request PCIe clock in K1 page770_17[10:9] = 10b
6345 		 */
6346 		e1e_rphy(hw, HV_PM_CTRL, &phy_data);
6347 		phy_data &= ~HV_PM_CTRL_K1_CLK_REQ;
6348 		phy_data |= BIT(10);
6349 		e1e_wphy(hw, HV_PM_CTRL, phy_data);
6350 
6351 		/* Make sure we don't exit K1 every time a new packet arrives
6352 		 * 772_29[5] = 1 CS_Mode_Stay_In_K1
6353 		 */
6354 		e1e_rphy(hw, I217_CGFREG, &phy_data);
6355 		phy_data |= BIT(5);
6356 		e1e_wphy(hw, I217_CGFREG, phy_data);
6357 
6358 		/* Change the MAC/PHY interface to SMBus
6359 		 * Force the SMBus in PHY page769_23[0] = 1
6360 		 * Force the SMBus in MAC CTRL_EXT[11] = 1
6361 		 */
6362 		e1e_rphy(hw, CV_SMB_CTRL, &phy_data);
6363 		phy_data |= CV_SMB_CTRL_FORCE_SMBUS;
6364 		e1e_wphy(hw, CV_SMB_CTRL, phy_data);
6365 		mac_data = er32(CTRL_EXT);
6366 		mac_data |= E1000_CTRL_EXT_FORCE_SMBUS;
6367 		ew32(CTRL_EXT, mac_data);
6368 
6369 		/* DFT control: PHY bit: page769_20[0] = 1
6370 		 * page769_20[7] - PHY PLL stop
6371 		 * page769_20[8] - PHY go to the electrical idle
6372 		 * page769_20[9] - PHY serdes disable
6373 		 * Gate PPW via EXTCNF_CTRL - set 0x0F00[7] = 1
6374 		 */
6375 		e1e_rphy(hw, I82579_DFT_CTRL, &phy_data);
6376 		phy_data |= BIT(0);
6377 		phy_data |= BIT(7);
6378 		phy_data |= BIT(8);
6379 		phy_data |= BIT(9);
6380 		e1e_wphy(hw, I82579_DFT_CTRL, phy_data);
6381 
6382 		mac_data = er32(EXTCNF_CTRL);
6383 		mac_data |= E1000_EXTCNF_CTRL_GATE_PHY_CFG;
6384 		ew32(EXTCNF_CTRL, mac_data);
6385 
6386 		/* Disable disconnected cable conditioning for Power Gating */
6387 		mac_data = er32(DPGFR);
6388 		mac_data |= BIT(2);
6389 		ew32(DPGFR, mac_data);
6390 
6391 		/* Enable the Dynamic Clock Gating in the DMA and MAC */
6392 		mac_data = er32(CTRL_EXT);
6393 		mac_data |= E1000_CTRL_EXT_DMA_DYN_CLK_EN;
6394 		ew32(CTRL_EXT, mac_data);
6395 	}
6396 
6397 	/* Enable the Dynamic Power Gating in the MAC */
6398 	mac_data = er32(FEXTNVM7);
6399 	mac_data |= BIT(22);
6400 	ew32(FEXTNVM7, mac_data);
6401 
6402 	/* Don't wake from dynamic Power Gating with clock request */
6403 	mac_data = er32(FEXTNVM12);
6404 	mac_data |= BIT(12);
6405 	ew32(FEXTNVM12, mac_data);
6406 
6407 	/* Ungate PGCB clock */
6408 	mac_data = er32(FEXTNVM9);
6409 	mac_data &= ~BIT(28);
6410 	ew32(FEXTNVM9, mac_data);
6411 
6412 	/* Enable K1 off to enable mPHY Power Gating */
6413 	mac_data = er32(FEXTNVM6);
6414 	mac_data |= BIT(31);
6415 	ew32(FEXTNVM6, mac_data);
6416 
6417 	/* Enable mPHY power gating for any link and speed */
6418 	mac_data = er32(FEXTNVM8);
6419 	mac_data |= BIT(9);
6420 	ew32(FEXTNVM8, mac_data);
6421 
6422 	/* No MAC DPG gating SLP_S0 in modern standby
6423 	 * Switch the logic of the lanphypc to use PMC counter
6424 	 */
6425 	mac_data = er32(FEXTNVM5);
6426 	mac_data |= BIT(7);
6427 	ew32(FEXTNVM5, mac_data);
6428 
6429 	/* Disable the time synchronization clock */
6430 	mac_data = er32(FEXTNVM7);
6431 	mac_data |= BIT(31);
6432 	mac_data &= ~BIT(0);
6433 	ew32(FEXTNVM7, mac_data);
6434 
6435 	/* Dynamic Power Gating Enable */
6436 	mac_data = er32(CTRL_EXT);
6437 	mac_data |= BIT(3);
6438 	ew32(CTRL_EXT, mac_data);
6439 
6440 	/* Check MAC Tx/Rx packet buffer pointers.
6441 	 * Reset MAC Tx/Rx packet buffer pointers to suppress any
6442 	 * pending traffic indication that would prevent power gating.
6443 	 */
6444 	mac_data = er32(TDFH);
6445 	if (mac_data)
6446 		ew32(TDFH, 0);
6447 	mac_data = er32(TDFT);
6448 	if (mac_data)
6449 		ew32(TDFT, 0);
6450 	mac_data = er32(TDFHS);
6451 	if (mac_data)
6452 		ew32(TDFHS, 0);
6453 	mac_data = er32(TDFTS);
6454 	if (mac_data)
6455 		ew32(TDFTS, 0);
6456 	mac_data = er32(TDFPC);
6457 	if (mac_data)
6458 		ew32(TDFPC, 0);
6459 	mac_data = er32(RDFH);
6460 	if (mac_data)
6461 		ew32(RDFH, 0);
6462 	mac_data = er32(RDFT);
6463 	if (mac_data)
6464 		ew32(RDFT, 0);
6465 	mac_data = er32(RDFHS);
6466 	if (mac_data)
6467 		ew32(RDFHS, 0);
6468 	mac_data = er32(RDFTS);
6469 	if (mac_data)
6470 		ew32(RDFTS, 0);
6471 	mac_data = er32(RDFPC);
6472 	if (mac_data)
6473 		ew32(RDFPC, 0);
6474 }
6475 
6476 static void e1000e_s0ix_exit_flow(struct e1000_adapter *adapter)
6477 {
6478 	struct e1000_hw *hw = &adapter->hw;
6479 	bool firmware_bug = false;
6480 	u32 mac_data;
6481 	u16 phy_data;
6482 	u32 i = 0;
6483 
6484 	if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID &&
6485 	    hw->mac.type >= e1000_pch_adp) {
6486 		/* Keep the GPT clock enabled for CSME */
6487 		mac_data = er32(FEXTNVM);
6488 		mac_data |= BIT(3);
6489 		ew32(FEXTNVM, mac_data);
6490 		/* Request ME unconfigure the device from S0ix */
6491 		mac_data = er32(H2ME);
6492 		mac_data &= ~E1000_H2ME_START_DPG;
6493 		mac_data |= E1000_H2ME_EXIT_DPG;
6494 		trace_e1000e_trace_mac_register(mac_data);
6495 		ew32(H2ME, mac_data);
6496 
6497 		/* Poll up to 2.5 seconds for ME to unconfigure DPG.
6498 		 * If this takes more than 1 second, show a warning indicating a
6499 		 * firmware bug
6500 		 */
6501 		while (!(er32(EXFWSM) & E1000_EXFWSM_DPG_EXIT_DONE)) {
6502 			if (i > 100 && !firmware_bug)
6503 				firmware_bug = true;
6504 
6505 			if (i++ == 250) {
6506 				e_dbg("Timeout (firmware bug): %d msec\n",
6507 				      i * 10);
6508 				break;
6509 			}
6510 
6511 			usleep_range(10000, 11000);
6512 		}
6513 		if (firmware_bug)
6514 			e_warn("DPG_EXIT_DONE took %d msec. This is a firmware bug\n",
6515 			       i * 10);
6516 		else
6517 			e_dbg("DPG_EXIT_DONE cleared after %d msec\n", i * 10);
6518 	} else {
6519 		/* Request driver unconfigure the device from S0ix */
6520 
6521 		/* Cancel disable disconnected cable conditioning
6522 		 * for Power Gating
6523 		 */
6524 		mac_data = er32(DPGFR);
6525 		mac_data &= ~BIT(2);
6526 		ew32(DPGFR, mac_data);
6527 
6528 		/* Disable the Dynamic Clock Gating in the DMA and MAC */
6529 		mac_data = er32(CTRL_EXT);
6530 		mac_data &= 0xFFF7FFFF;
6531 		ew32(CTRL_EXT, mac_data);
6532 
6533 		/* Enable the periodic inband message,
6534 		 * Request PCIe clock in K1 page770_17[10:9] =01b
6535 		 */
6536 		e1e_rphy(hw, HV_PM_CTRL, &phy_data);
6537 		phy_data &= 0xFBFF;
6538 		phy_data |= HV_PM_CTRL_K1_CLK_REQ;
6539 		e1e_wphy(hw, HV_PM_CTRL, phy_data);
6540 
6541 		/* Return back configuration
6542 		 * 772_29[5] = 0 CS_Mode_Stay_In_K1
6543 		 */
6544 		e1e_rphy(hw, I217_CGFREG, &phy_data);
6545 		phy_data &= 0xFFDF;
6546 		e1e_wphy(hw, I217_CGFREG, phy_data);
6547 
6548 		/* Change the MAC/PHY interface to Kumeran
6549 		 * Unforce the SMBus in PHY page769_23[0] = 0
6550 		 * Unforce the SMBus in MAC CTRL_EXT[11] = 0
6551 		 */
6552 		e1e_rphy(hw, CV_SMB_CTRL, &phy_data);
6553 		phy_data &= ~CV_SMB_CTRL_FORCE_SMBUS;
6554 		e1e_wphy(hw, CV_SMB_CTRL, phy_data);
6555 		mac_data = er32(CTRL_EXT);
6556 		mac_data &= ~E1000_CTRL_EXT_FORCE_SMBUS;
6557 		ew32(CTRL_EXT, mac_data);
6558 	}
6559 
6560 	/* Disable Dynamic Power Gating */
6561 	mac_data = er32(CTRL_EXT);
6562 	mac_data &= 0xFFFFFFF7;
6563 	ew32(CTRL_EXT, mac_data);
6564 
6565 	/* Enable the time synchronization clock */
6566 	mac_data = er32(FEXTNVM7);
6567 	mac_data &= ~BIT(31);
6568 	mac_data |= BIT(0);
6569 	ew32(FEXTNVM7, mac_data);
6570 
6571 	/* Disable the Dynamic Power Gating in the MAC */
6572 	mac_data = er32(FEXTNVM7);
6573 	mac_data &= 0xFFBFFFFF;
6574 	ew32(FEXTNVM7, mac_data);
6575 
6576 	/* Disable mPHY power gating for any link and speed */
6577 	mac_data = er32(FEXTNVM8);
6578 	mac_data &= ~BIT(9);
6579 	ew32(FEXTNVM8, mac_data);
6580 
6581 	/* Disable K1 off */
6582 	mac_data = er32(FEXTNVM6);
6583 	mac_data &= ~BIT(31);
6584 	ew32(FEXTNVM6, mac_data);
6585 
6586 	/* Disable Ungate PGCB clock */
6587 	mac_data = er32(FEXTNVM9);
6588 	mac_data |= BIT(28);
6589 	ew32(FEXTNVM9, mac_data);
6590 
6591 	/* Cancel not waking from dynamic
6592 	 * Power Gating with clock request
6593 	 */
6594 	mac_data = er32(FEXTNVM12);
6595 	mac_data &= ~BIT(12);
6596 	ew32(FEXTNVM12, mac_data);
6597 
6598 	/* Revert the lanphypc logic to use the internal Gbe counter
6599 	 * and not the PMC counter
6600 	 */
6601 	mac_data = er32(FEXTNVM5);
6602 	mac_data &= 0xFFFFFF7F;
6603 	ew32(FEXTNVM5, mac_data);
6604 }
6605 
6606 static int e1000e_pm_freeze(struct device *dev)
6607 {
6608 	struct net_device *netdev = dev_get_drvdata(dev);
6609 	struct e1000_adapter *adapter = netdev_priv(netdev);
6610 	bool present;
6611 
6612 	rtnl_lock();
6613 
6614 	present = netif_device_present(netdev);
6615 	netif_device_detach(netdev);
6616 
6617 	if (present && netif_running(netdev)) {
6618 		int count = E1000_CHECK_RESET_COUNT;
6619 
6620 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6621 			usleep_range(10000, 11000);
6622 
6623 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6624 
6625 		/* Quiesce the device without resetting the hardware */
6626 		e1000e_down(adapter, false);
6627 		e1000_free_irq(adapter);
6628 	}
6629 	rtnl_unlock();
6630 
6631 	e1000e_reset_interrupt_capability(adapter);
6632 
6633 	/* Allow time for pending master requests to run */
6634 	e1000e_disable_pcie_master(&adapter->hw);
6635 
6636 	return 0;
6637 }
6638 
6639 static int __e1000_shutdown(struct pci_dev *pdev, bool runtime)
6640 {
6641 	struct net_device *netdev = pci_get_drvdata(pdev);
6642 	struct e1000_adapter *adapter = netdev_priv(netdev);
6643 	struct e1000_hw *hw = &adapter->hw;
6644 	u32 ctrl, ctrl_ext, rctl, status, wufc;
6645 	int retval = 0;
6646 
6647 	/* Runtime suspend should only enable wakeup for link changes */
6648 	if (runtime)
6649 		wufc = E1000_WUFC_LNKC;
6650 	else if (device_may_wakeup(&pdev->dev))
6651 		wufc = adapter->wol;
6652 	else
6653 		wufc = 0;
6654 
6655 	status = er32(STATUS);
6656 	if (status & E1000_STATUS_LU)
6657 		wufc &= ~E1000_WUFC_LNKC;
6658 
6659 	if (wufc) {
6660 		e1000_setup_rctl(adapter);
6661 		e1000e_set_rx_mode(netdev);
6662 
6663 		/* turn on all-multi mode if wake on multicast is enabled */
6664 		if (wufc & E1000_WUFC_MC) {
6665 			rctl = er32(RCTL);
6666 			rctl |= E1000_RCTL_MPE;
6667 			ew32(RCTL, rctl);
6668 		}
6669 
6670 		ctrl = er32(CTRL);
6671 		ctrl |= E1000_CTRL_ADVD3WUC;
6672 		if (!(adapter->flags2 & FLAG2_HAS_PHY_WAKEUP))
6673 			ctrl |= E1000_CTRL_EN_PHY_PWR_MGMT;
6674 		ew32(CTRL, ctrl);
6675 
6676 		if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
6677 		    adapter->hw.phy.media_type ==
6678 		    e1000_media_type_internal_serdes) {
6679 			/* keep the laser running in D3 */
6680 			ctrl_ext = er32(CTRL_EXT);
6681 			ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
6682 			ew32(CTRL_EXT, ctrl_ext);
6683 		}
6684 
6685 		if (!runtime)
6686 			e1000e_power_up_phy(adapter);
6687 
6688 		if (adapter->flags & FLAG_IS_ICH)
6689 			e1000_suspend_workarounds_ich8lan(&adapter->hw);
6690 
6691 		if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6692 			/* enable wakeup by the PHY */
6693 			retval = e1000_init_phy_wakeup(adapter, wufc);
6694 			if (retval) {
6695 				e_err("Failed to enable wakeup\n");
6696 				goto skip_phy_configurations;
6697 			}
6698 		} else {
6699 			/* enable wakeup by the MAC */
6700 			ew32(WUFC, wufc);
6701 			ew32(WUC, E1000_WUC_PME_EN);
6702 		}
6703 	} else {
6704 		ew32(WUC, 0);
6705 		ew32(WUFC, 0);
6706 
6707 		e1000_power_down_phy(adapter);
6708 	}
6709 
6710 	if (adapter->hw.phy.type == e1000_phy_igp_3) {
6711 		e1000e_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
6712 	} else if (hw->mac.type >= e1000_pch_lpt) {
6713 		if (wufc && !(wufc & (E1000_WUFC_EX | E1000_WUFC_MC | E1000_WUFC_BC))) {
6714 			/* ULP does not support wake from unicast, multicast
6715 			 * or broadcast.
6716 			 */
6717 			retval = e1000_enable_ulp_lpt_lp(hw, !runtime);
6718 			if (retval) {
6719 				e_err("Failed to enable ULP\n");
6720 				goto skip_phy_configurations;
6721 			}
6722 		}
6723 	}
6724 
6725 	/* Ensure that the appropriate bits are set in LPI_CTRL
6726 	 * for EEE in Sx
6727 	 */
6728 	if ((hw->phy.type >= e1000_phy_i217) &&
6729 	    adapter->eee_advert && hw->dev_spec.ich8lan.eee_lp_ability) {
6730 		u16 lpi_ctrl = 0;
6731 
6732 		retval = hw->phy.ops.acquire(hw);
6733 		if (!retval) {
6734 			retval = e1e_rphy_locked(hw, I82579_LPI_CTRL,
6735 						 &lpi_ctrl);
6736 			if (!retval) {
6737 				if (adapter->eee_advert &
6738 				    hw->dev_spec.ich8lan.eee_lp_ability &
6739 				    I82579_EEE_100_SUPPORTED)
6740 					lpi_ctrl |= I82579_LPI_CTRL_100_ENABLE;
6741 				if (adapter->eee_advert &
6742 				    hw->dev_spec.ich8lan.eee_lp_ability &
6743 				    I82579_EEE_1000_SUPPORTED)
6744 					lpi_ctrl |= I82579_LPI_CTRL_1000_ENABLE;
6745 
6746 				retval = e1e_wphy_locked(hw, I82579_LPI_CTRL,
6747 							 lpi_ctrl);
6748 			}
6749 		}
6750 		hw->phy.ops.release(hw);
6751 	}
6752 
6753 skip_phy_configurations:
6754 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
6755 	 * would have already happened in close and is redundant.
6756 	 */
6757 	e1000e_release_hw_control(adapter);
6758 
6759 	pci_clear_master(pdev);
6760 
6761 	/* The pci-e switch on some quad port adapters will report a
6762 	 * correctable error when the MAC transitions from D0 to D3.  To
6763 	 * prevent this we need to mask off the correctable errors on the
6764 	 * downstream port of the pci-e switch.
6765 	 *
6766 	 * We don't have the associated upstream bridge while assigning
6767 	 * the PCI device into guest. For example, the KVM on power is
6768 	 * one of the cases.
6769 	 */
6770 	if (adapter->flags & FLAG_IS_QUAD_PORT) {
6771 		struct pci_dev *us_dev = pdev->bus->self;
6772 		u16 devctl;
6773 
6774 		if (!us_dev)
6775 			return 0;
6776 
6777 		pcie_capability_read_word(us_dev, PCI_EXP_DEVCTL, &devctl);
6778 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL,
6779 					   (devctl & ~PCI_EXP_DEVCTL_CERE));
6780 
6781 		pci_save_state(pdev);
6782 		pci_prepare_to_sleep(pdev);
6783 
6784 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL, devctl);
6785 	}
6786 
6787 	return 0;
6788 }
6789 
6790 /**
6791  * __e1000e_disable_aspm - Disable ASPM states
6792  * @pdev: pointer to PCI device struct
6793  * @state: bit-mask of ASPM states to disable
6794  * @locked: indication if this context holds pci_bus_sem locked.
6795  *
6796  * Some devices *must* have certain ASPM states disabled per hardware errata.
6797  **/
6798 static void __e1000e_disable_aspm(struct pci_dev *pdev, u16 state, int locked)
6799 {
6800 	struct pci_dev *parent = pdev->bus->self;
6801 	u16 aspm_dis_mask = 0;
6802 	u16 pdev_aspmc, parent_aspmc;
6803 
6804 	switch (state) {
6805 	case PCIE_LINK_STATE_L0S:
6806 	case PCIE_LINK_STATE_L0S | PCIE_LINK_STATE_L1:
6807 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L0S;
6808 		fallthrough; /* can't have L1 without L0s */
6809 	case PCIE_LINK_STATE_L1:
6810 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L1;
6811 		break;
6812 	default:
6813 		return;
6814 	}
6815 
6816 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6817 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6818 
6819 	if (parent) {
6820 		pcie_capability_read_word(parent, PCI_EXP_LNKCTL,
6821 					  &parent_aspmc);
6822 		parent_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6823 	}
6824 
6825 	/* Nothing to do if the ASPM states to be disabled already are */
6826 	if (!(pdev_aspmc & aspm_dis_mask) &&
6827 	    (!parent || !(parent_aspmc & aspm_dis_mask)))
6828 		return;
6829 
6830 	dev_info(&pdev->dev, "Disabling ASPM %s %s\n",
6831 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L0S) ?
6832 		 "L0s" : "",
6833 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L1) ?
6834 		 "L1" : "");
6835 
6836 #ifdef CONFIG_PCIEASPM
6837 	if (locked)
6838 		pci_disable_link_state_locked(pdev, state);
6839 	else
6840 		pci_disable_link_state(pdev, state);
6841 
6842 	/* Double-check ASPM control.  If not disabled by the above, the
6843 	 * BIOS is preventing that from happening (or CONFIG_PCIEASPM is
6844 	 * not enabled); override by writing PCI config space directly.
6845 	 */
6846 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6847 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6848 
6849 	if (!(aspm_dis_mask & pdev_aspmc))
6850 		return;
6851 #endif
6852 
6853 	/* Both device and parent should have the same ASPM setting.
6854 	 * Disable ASPM in downstream component first and then upstream.
6855 	 */
6856 	pcie_capability_clear_word(pdev, PCI_EXP_LNKCTL, aspm_dis_mask);
6857 
6858 	if (parent)
6859 		pcie_capability_clear_word(parent, PCI_EXP_LNKCTL,
6860 					   aspm_dis_mask);
6861 }
6862 
6863 /**
6864  * e1000e_disable_aspm - Disable ASPM states.
6865  * @pdev: pointer to PCI device struct
6866  * @state: bit-mask of ASPM states to disable
6867  *
6868  * This function acquires the pci_bus_sem!
6869  * Some devices *must* have certain ASPM states disabled per hardware errata.
6870  **/
6871 static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6872 {
6873 	__e1000e_disable_aspm(pdev, state, 0);
6874 }
6875 
6876 /**
6877  * e1000e_disable_aspm_locked - Disable ASPM states.
6878  * @pdev: pointer to PCI device struct
6879  * @state: bit-mask of ASPM states to disable
6880  *
6881  * This function must be called with pci_bus_sem acquired!
6882  * Some devices *must* have certain ASPM states disabled per hardware errata.
6883  **/
6884 static void e1000e_disable_aspm_locked(struct pci_dev *pdev, u16 state)
6885 {
6886 	__e1000e_disable_aspm(pdev, state, 1);
6887 }
6888 
6889 static int e1000e_pm_thaw(struct device *dev)
6890 {
6891 	struct net_device *netdev = dev_get_drvdata(dev);
6892 	struct e1000_adapter *adapter = netdev_priv(netdev);
6893 	int rc = 0;
6894 
6895 	e1000e_set_interrupt_capability(adapter);
6896 
6897 	rtnl_lock();
6898 	if (netif_running(netdev)) {
6899 		rc = e1000_request_irq(adapter);
6900 		if (rc)
6901 			goto err_irq;
6902 
6903 		e1000e_up(adapter);
6904 	}
6905 
6906 	netif_device_attach(netdev);
6907 err_irq:
6908 	rtnl_unlock();
6909 
6910 	return rc;
6911 }
6912 
6913 static int __e1000_resume(struct pci_dev *pdev)
6914 {
6915 	struct net_device *netdev = pci_get_drvdata(pdev);
6916 	struct e1000_adapter *adapter = netdev_priv(netdev);
6917 	struct e1000_hw *hw = &adapter->hw;
6918 	u16 aspm_disable_flag = 0;
6919 
6920 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6921 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6922 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6923 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6924 	if (aspm_disable_flag)
6925 		e1000e_disable_aspm(pdev, aspm_disable_flag);
6926 
6927 	pci_set_master(pdev);
6928 
6929 	if (hw->mac.type >= e1000_pch2lan)
6930 		e1000_resume_workarounds_pchlan(&adapter->hw);
6931 
6932 	e1000e_power_up_phy(adapter);
6933 
6934 	/* report the system wakeup cause from S3/S4 */
6935 	if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6936 		u16 phy_data;
6937 
6938 		e1e_rphy(&adapter->hw, BM_WUS, &phy_data);
6939 		if (phy_data) {
6940 			e_info("PHY Wakeup cause - %s\n",
6941 			       phy_data & E1000_WUS_EX ? "Unicast Packet" :
6942 			       phy_data & E1000_WUS_MC ? "Multicast Packet" :
6943 			       phy_data & E1000_WUS_BC ? "Broadcast Packet" :
6944 			       phy_data & E1000_WUS_MAG ? "Magic Packet" :
6945 			       phy_data & E1000_WUS_LNKC ?
6946 			       "Link Status Change" : "other");
6947 		}
6948 		e1e_wphy(&adapter->hw, BM_WUS, ~0);
6949 	} else {
6950 		u32 wus = er32(WUS);
6951 
6952 		if (wus) {
6953 			e_info("MAC Wakeup cause - %s\n",
6954 			       wus & E1000_WUS_EX ? "Unicast Packet" :
6955 			       wus & E1000_WUS_MC ? "Multicast Packet" :
6956 			       wus & E1000_WUS_BC ? "Broadcast Packet" :
6957 			       wus & E1000_WUS_MAG ? "Magic Packet" :
6958 			       wus & E1000_WUS_LNKC ? "Link Status Change" :
6959 			       "other");
6960 		}
6961 		ew32(WUS, ~0);
6962 	}
6963 
6964 	e1000e_reset(adapter);
6965 
6966 	e1000_init_manageability_pt(adapter);
6967 
6968 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6969 	 * is up.  For all other cases, let the f/w know that the h/w is now
6970 	 * under the control of the driver.
6971 	 */
6972 	if (!(adapter->flags & FLAG_HAS_AMT))
6973 		e1000e_get_hw_control(adapter);
6974 
6975 	return 0;
6976 }
6977 
6978 static int e1000e_pm_prepare(struct device *dev)
6979 {
6980 	return pm_runtime_suspended(dev) &&
6981 		pm_suspend_via_firmware();
6982 }
6983 
6984 static int e1000e_pm_suspend(struct device *dev)
6985 {
6986 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6987 	struct e1000_adapter *adapter = netdev_priv(netdev);
6988 	struct pci_dev *pdev = to_pci_dev(dev);
6989 	int rc;
6990 
6991 	e1000e_flush_lpic(pdev);
6992 
6993 	e1000e_pm_freeze(dev);
6994 
6995 	rc = __e1000_shutdown(pdev, false);
6996 	if (!rc) {
6997 		/* Introduce S0ix implementation */
6998 		if (adapter->flags2 & FLAG2_ENABLE_S0IX_FLOWS)
6999 			e1000e_s0ix_entry_flow(adapter);
7000 	}
7001 
7002 	return 0;
7003 }
7004 
7005 static int e1000e_pm_resume(struct device *dev)
7006 {
7007 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
7008 	struct e1000_adapter *adapter = netdev_priv(netdev);
7009 	struct pci_dev *pdev = to_pci_dev(dev);
7010 	int rc;
7011 
7012 	/* Introduce S0ix implementation */
7013 	if (adapter->flags2 & FLAG2_ENABLE_S0IX_FLOWS)
7014 		e1000e_s0ix_exit_flow(adapter);
7015 
7016 	rc = __e1000_resume(pdev);
7017 	if (rc)
7018 		return rc;
7019 
7020 	return e1000e_pm_thaw(dev);
7021 }
7022 
7023 static __maybe_unused int e1000e_pm_runtime_idle(struct device *dev)
7024 {
7025 	struct net_device *netdev = dev_get_drvdata(dev);
7026 	struct e1000_adapter *adapter = netdev_priv(netdev);
7027 	u16 eee_lp;
7028 
7029 	eee_lp = adapter->hw.dev_spec.ich8lan.eee_lp_ability;
7030 
7031 	if (!e1000e_has_link(adapter)) {
7032 		adapter->hw.dev_spec.ich8lan.eee_lp_ability = eee_lp;
7033 		pm_schedule_suspend(dev, 5 * MSEC_PER_SEC);
7034 	}
7035 
7036 	return -EBUSY;
7037 }
7038 
7039 static int e1000e_pm_runtime_resume(struct device *dev)
7040 {
7041 	struct pci_dev *pdev = to_pci_dev(dev);
7042 	struct net_device *netdev = pci_get_drvdata(pdev);
7043 	struct e1000_adapter *adapter = netdev_priv(netdev);
7044 	int rc;
7045 
7046 	pdev->pme_poll = true;
7047 
7048 	rc = __e1000_resume(pdev);
7049 	if (rc)
7050 		return rc;
7051 
7052 	if (netdev->flags & IFF_UP)
7053 		e1000e_up(adapter);
7054 
7055 	return rc;
7056 }
7057 
7058 static int e1000e_pm_runtime_suspend(struct device *dev)
7059 {
7060 	struct pci_dev *pdev = to_pci_dev(dev);
7061 	struct net_device *netdev = pci_get_drvdata(pdev);
7062 	struct e1000_adapter *adapter = netdev_priv(netdev);
7063 
7064 	if (netdev->flags & IFF_UP) {
7065 		int count = E1000_CHECK_RESET_COUNT;
7066 
7067 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
7068 			usleep_range(10000, 11000);
7069 
7070 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
7071 
7072 		/* Down the device without resetting the hardware */
7073 		e1000e_down(adapter, false);
7074 	}
7075 
7076 	if (__e1000_shutdown(pdev, true)) {
7077 		e1000e_pm_runtime_resume(dev);
7078 		return -EBUSY;
7079 	}
7080 
7081 	return 0;
7082 }
7083 
7084 static void e1000_shutdown(struct pci_dev *pdev)
7085 {
7086 	e1000e_flush_lpic(pdev);
7087 
7088 	e1000e_pm_freeze(&pdev->dev);
7089 
7090 	__e1000_shutdown(pdev, false);
7091 }
7092 
7093 #ifdef CONFIG_NET_POLL_CONTROLLER
7094 
7095 static irqreturn_t e1000_intr_msix(int __always_unused irq, void *data)
7096 {
7097 	struct net_device *netdev = data;
7098 	struct e1000_adapter *adapter = netdev_priv(netdev);
7099 
7100 	if (adapter->msix_entries) {
7101 		int vector, msix_irq;
7102 
7103 		vector = 0;
7104 		msix_irq = adapter->msix_entries[vector].vector;
7105 		if (disable_hardirq(msix_irq))
7106 			e1000_intr_msix_rx(msix_irq, netdev);
7107 		enable_irq(msix_irq);
7108 
7109 		vector++;
7110 		msix_irq = adapter->msix_entries[vector].vector;
7111 		if (disable_hardirq(msix_irq))
7112 			e1000_intr_msix_tx(msix_irq, netdev);
7113 		enable_irq(msix_irq);
7114 
7115 		vector++;
7116 		msix_irq = adapter->msix_entries[vector].vector;
7117 		if (disable_hardirq(msix_irq))
7118 			e1000_msix_other(msix_irq, netdev);
7119 		enable_irq(msix_irq);
7120 	}
7121 
7122 	return IRQ_HANDLED;
7123 }
7124 
7125 /**
7126  * e1000_netpoll
7127  * @netdev: network interface device structure
7128  *
7129  * Polling 'interrupt' - used by things like netconsole to send skbs
7130  * without having to re-enable interrupts. It's not called while
7131  * the interrupt routine is executing.
7132  */
7133 static void e1000_netpoll(struct net_device *netdev)
7134 {
7135 	struct e1000_adapter *adapter = netdev_priv(netdev);
7136 
7137 	switch (adapter->int_mode) {
7138 	case E1000E_INT_MODE_MSIX:
7139 		e1000_intr_msix(adapter->pdev->irq, netdev);
7140 		break;
7141 	case E1000E_INT_MODE_MSI:
7142 		if (disable_hardirq(adapter->pdev->irq))
7143 			e1000_intr_msi(adapter->pdev->irq, netdev);
7144 		enable_irq(adapter->pdev->irq);
7145 		break;
7146 	default:		/* E1000E_INT_MODE_LEGACY */
7147 		if (disable_hardirq(adapter->pdev->irq))
7148 			e1000_intr(adapter->pdev->irq, netdev);
7149 		enable_irq(adapter->pdev->irq);
7150 		break;
7151 	}
7152 }
7153 #endif
7154 
7155 /**
7156  * e1000_io_error_detected - called when PCI error is detected
7157  * @pdev: Pointer to PCI device
7158  * @state: The current pci connection state
7159  *
7160  * This function is called after a PCI bus error affecting
7161  * this device has been detected.
7162  */
7163 static pci_ers_result_t e1000_io_error_detected(struct pci_dev *pdev,
7164 						pci_channel_state_t state)
7165 {
7166 	e1000e_pm_freeze(&pdev->dev);
7167 
7168 	if (state == pci_channel_io_perm_failure)
7169 		return PCI_ERS_RESULT_DISCONNECT;
7170 
7171 	pci_disable_device(pdev);
7172 
7173 	/* Request a slot reset. */
7174 	return PCI_ERS_RESULT_NEED_RESET;
7175 }
7176 
7177 /**
7178  * e1000_io_slot_reset - called after the pci bus has been reset.
7179  * @pdev: Pointer to PCI device
7180  *
7181  * Restart the card from scratch, as if from a cold-boot. Implementation
7182  * resembles the first-half of the e1000e_pm_resume routine.
7183  */
7184 static pci_ers_result_t e1000_io_slot_reset(struct pci_dev *pdev)
7185 {
7186 	struct net_device *netdev = pci_get_drvdata(pdev);
7187 	struct e1000_adapter *adapter = netdev_priv(netdev);
7188 	struct e1000_hw *hw = &adapter->hw;
7189 	u16 aspm_disable_flag = 0;
7190 	int err;
7191 	pci_ers_result_t result;
7192 
7193 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
7194 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7195 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
7196 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7197 	if (aspm_disable_flag)
7198 		e1000e_disable_aspm_locked(pdev, aspm_disable_flag);
7199 
7200 	err = pci_enable_device_mem(pdev);
7201 	if (err) {
7202 		dev_err(&pdev->dev,
7203 			"Cannot re-enable PCI device after reset.\n");
7204 		result = PCI_ERS_RESULT_DISCONNECT;
7205 	} else {
7206 		pci_restore_state(pdev);
7207 		pci_set_master(pdev);
7208 
7209 		pci_enable_wake(pdev, PCI_D3hot, 0);
7210 		pci_enable_wake(pdev, PCI_D3cold, 0);
7211 
7212 		e1000e_reset(adapter);
7213 		ew32(WUS, ~0);
7214 		result = PCI_ERS_RESULT_RECOVERED;
7215 	}
7216 
7217 	return result;
7218 }
7219 
7220 /**
7221  * e1000_io_resume - called when traffic can start flowing again.
7222  * @pdev: Pointer to PCI device
7223  *
7224  * This callback is called when the error recovery driver tells us that
7225  * its OK to resume normal operation. Implementation resembles the
7226  * second-half of the e1000e_pm_resume routine.
7227  */
7228 static void e1000_io_resume(struct pci_dev *pdev)
7229 {
7230 	struct net_device *netdev = pci_get_drvdata(pdev);
7231 	struct e1000_adapter *adapter = netdev_priv(netdev);
7232 
7233 	e1000_init_manageability_pt(adapter);
7234 
7235 	e1000e_pm_thaw(&pdev->dev);
7236 
7237 	/* If the controller has AMT, do not set DRV_LOAD until the interface
7238 	 * is up.  For all other cases, let the f/w know that the h/w is now
7239 	 * under the control of the driver.
7240 	 */
7241 	if (!(adapter->flags & FLAG_HAS_AMT))
7242 		e1000e_get_hw_control(adapter);
7243 }
7244 
7245 static void e1000_print_device_info(struct e1000_adapter *adapter)
7246 {
7247 	struct e1000_hw *hw = &adapter->hw;
7248 	struct net_device *netdev = adapter->netdev;
7249 	u32 ret_val;
7250 	u8 pba_str[E1000_PBANUM_LENGTH];
7251 
7252 	/* print bus type/speed/width info */
7253 	e_info("(PCI Express:2.5GT/s:%s) %pM\n",
7254 	       /* bus width */
7255 	       ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
7256 		"Width x1"),
7257 	       /* MAC address */
7258 	       netdev->dev_addr);
7259 	e_info("Intel(R) PRO/%s Network Connection\n",
7260 	       (hw->phy.type == e1000_phy_ife) ? "10/100" : "1000");
7261 	ret_val = e1000_read_pba_string_generic(hw, pba_str,
7262 						E1000_PBANUM_LENGTH);
7263 	if (ret_val)
7264 		strscpy((char *)pba_str, "Unknown", sizeof(pba_str));
7265 	e_info("MAC: %d, PHY: %d, PBA No: %s\n",
7266 	       hw->mac.type, hw->phy.type, pba_str);
7267 }
7268 
7269 static void e1000_eeprom_checks(struct e1000_adapter *adapter)
7270 {
7271 	struct e1000_hw *hw = &adapter->hw;
7272 	int ret_val;
7273 	u16 buf = 0;
7274 
7275 	if (hw->mac.type != e1000_82573)
7276 		return;
7277 
7278 	ret_val = e1000_read_nvm(hw, NVM_INIT_CONTROL2_REG, 1, &buf);
7279 	le16_to_cpus(&buf);
7280 	if (!ret_val && (!(buf & BIT(0)))) {
7281 		/* Deep Smart Power Down (DSPD) */
7282 		dev_warn(&adapter->pdev->dev,
7283 			 "Warning: detected DSPD enabled in EEPROM\n");
7284 	}
7285 }
7286 
7287 static netdev_features_t e1000_fix_features(struct net_device *netdev,
7288 					    netdev_features_t features)
7289 {
7290 	struct e1000_adapter *adapter = netdev_priv(netdev);
7291 	struct e1000_hw *hw = &adapter->hw;
7292 
7293 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
7294 	if ((hw->mac.type >= e1000_pch2lan) && (netdev->mtu > ETH_DATA_LEN))
7295 		features &= ~NETIF_F_RXFCS;
7296 
7297 	/* Since there is no support for separate Rx/Tx vlan accel
7298 	 * enable/disable make sure Tx flag is always in same state as Rx.
7299 	 */
7300 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
7301 		features |= NETIF_F_HW_VLAN_CTAG_TX;
7302 	else
7303 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
7304 
7305 	return features;
7306 }
7307 
7308 static int e1000_set_features(struct net_device *netdev,
7309 			      netdev_features_t features)
7310 {
7311 	struct e1000_adapter *adapter = netdev_priv(netdev);
7312 	netdev_features_t changed = features ^ netdev->features;
7313 
7314 	if (changed & (NETIF_F_TSO | NETIF_F_TSO6))
7315 		adapter->flags |= FLAG_TSO_FORCE;
7316 
7317 	if (!(changed & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX |
7318 			 NETIF_F_RXCSUM | NETIF_F_RXHASH | NETIF_F_RXFCS |
7319 			 NETIF_F_RXALL)))
7320 		return 0;
7321 
7322 	if (changed & NETIF_F_RXFCS) {
7323 		if (features & NETIF_F_RXFCS) {
7324 			adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7325 		} else {
7326 			/* We need to take it back to defaults, which might mean
7327 			 * stripping is still disabled at the adapter level.
7328 			 */
7329 			if (adapter->flags2 & FLAG2_DFLT_CRC_STRIPPING)
7330 				adapter->flags2 |= FLAG2_CRC_STRIPPING;
7331 			else
7332 				adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7333 		}
7334 	}
7335 
7336 	netdev->features = features;
7337 
7338 	if (netif_running(netdev))
7339 		e1000e_reinit_locked(adapter);
7340 	else
7341 		e1000e_reset(adapter);
7342 
7343 	return 1;
7344 }
7345 
7346 static const struct net_device_ops e1000e_netdev_ops = {
7347 	.ndo_open		= e1000e_open,
7348 	.ndo_stop		= e1000e_close,
7349 	.ndo_start_xmit		= e1000_xmit_frame,
7350 	.ndo_get_stats64	= e1000e_get_stats64,
7351 	.ndo_set_rx_mode	= e1000e_set_rx_mode,
7352 	.ndo_set_mac_address	= e1000_set_mac,
7353 	.ndo_change_mtu		= e1000_change_mtu,
7354 	.ndo_eth_ioctl		= e1000_ioctl,
7355 	.ndo_tx_timeout		= e1000_tx_timeout,
7356 	.ndo_validate_addr	= eth_validate_addr,
7357 
7358 	.ndo_vlan_rx_add_vid	= e1000_vlan_rx_add_vid,
7359 	.ndo_vlan_rx_kill_vid	= e1000_vlan_rx_kill_vid,
7360 #ifdef CONFIG_NET_POLL_CONTROLLER
7361 	.ndo_poll_controller	= e1000_netpoll,
7362 #endif
7363 	.ndo_set_features	= e1000_set_features,
7364 	.ndo_fix_features	= e1000_fix_features,
7365 	.ndo_features_check	= passthru_features_check,
7366 	.ndo_hwtstamp_get	= e1000e_hwtstamp_get,
7367 	.ndo_hwtstamp_set	= e1000e_hwtstamp_set,
7368 };
7369 
7370 /**
7371  * e1000_probe - Device Initialization Routine
7372  * @pdev: PCI device information struct
7373  * @ent: entry in e1000_pci_tbl
7374  *
7375  * Returns 0 on success, negative on failure
7376  *
7377  * e1000_probe initializes an adapter identified by a pci_dev structure.
7378  * The OS initialization, configuring of the adapter private structure,
7379  * and a hardware reset occur.
7380  **/
7381 static int e1000_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7382 {
7383 	struct net_device *netdev;
7384 	struct e1000_adapter *adapter;
7385 	struct e1000_hw *hw;
7386 	const struct e1000_info *ei = e1000_info_tbl[ent->driver_data];
7387 	resource_size_t mmio_start, mmio_len;
7388 	resource_size_t flash_start, flash_len;
7389 	static int cards_found;
7390 	u16 aspm_disable_flag = 0;
7391 	u16 eeprom_data = 0;
7392 	u16 eeprom_apme_mask = E1000_EEPROM_APME;
7393 	int bars, i, err;
7394 	s32 ret_val = 0;
7395 
7396 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L0S)
7397 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7398 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L1)
7399 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7400 	if (aspm_disable_flag)
7401 		e1000e_disable_aspm(pdev, aspm_disable_flag);
7402 
7403 	err = pci_enable_device_mem(pdev);
7404 	if (err)
7405 		return err;
7406 
7407 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
7408 	if (err) {
7409 		dev_err(&pdev->dev,
7410 			"No usable DMA configuration, aborting\n");
7411 		goto err_dma;
7412 	}
7413 
7414 	bars = pci_select_bars(pdev, IORESOURCE_MEM);
7415 	err = pci_request_selected_regions_exclusive(pdev, bars,
7416 						     e1000e_driver_name);
7417 	if (err)
7418 		goto err_pci_reg;
7419 
7420 	pci_set_master(pdev);
7421 	/* PCI config space info */
7422 	err = pci_save_state(pdev);
7423 	if (err)
7424 		goto err_alloc_etherdev;
7425 
7426 	err = -ENOMEM;
7427 	netdev = alloc_etherdev(sizeof(struct e1000_adapter));
7428 	if (!netdev)
7429 		goto err_alloc_etherdev;
7430 
7431 	SET_NETDEV_DEV(netdev, &pdev->dev);
7432 
7433 	netdev->irq = pdev->irq;
7434 
7435 	pci_set_drvdata(pdev, netdev);
7436 	adapter = netdev_priv(netdev);
7437 	hw = &adapter->hw;
7438 	adapter->netdev = netdev;
7439 	adapter->pdev = pdev;
7440 	adapter->ei = ei;
7441 	adapter->pba = ei->pba;
7442 	adapter->flags = ei->flags;
7443 	adapter->flags2 = ei->flags2;
7444 	adapter->hw.adapter = adapter;
7445 	adapter->hw.mac.type = ei->mac;
7446 	adapter->max_hw_frame_size = ei->max_hw_frame_size;
7447 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
7448 
7449 	mmio_start = pci_resource_start(pdev, 0);
7450 	mmio_len = pci_resource_len(pdev, 0);
7451 
7452 	err = -EIO;
7453 	adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
7454 	if (!adapter->hw.hw_addr)
7455 		goto err_ioremap;
7456 
7457 	if ((adapter->flags & FLAG_HAS_FLASH) &&
7458 	    (pci_resource_flags(pdev, 1) & IORESOURCE_MEM) &&
7459 	    (hw->mac.type < e1000_pch_spt)) {
7460 		flash_start = pci_resource_start(pdev, 1);
7461 		flash_len = pci_resource_len(pdev, 1);
7462 		adapter->hw.flash_address = ioremap(flash_start, flash_len);
7463 		if (!adapter->hw.flash_address)
7464 			goto err_flashmap;
7465 	}
7466 
7467 	/* Set default EEE advertisement */
7468 	if (adapter->flags2 & FLAG2_HAS_EEE)
7469 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
7470 
7471 	/* construct the net_device struct */
7472 	netdev->netdev_ops = &e1000e_netdev_ops;
7473 	e1000e_set_ethtool_ops(netdev);
7474 	netdev->watchdog_timeo = 5 * HZ;
7475 	netif_napi_add(netdev, &adapter->napi, e1000e_poll);
7476 	strscpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
7477 
7478 	netdev->mem_start = mmio_start;
7479 	netdev->mem_end = mmio_start + mmio_len;
7480 
7481 	adapter->bd_number = cards_found++;
7482 
7483 	e1000e_check_options(adapter);
7484 
7485 	/* setup adapter struct */
7486 	err = e1000_sw_init(adapter);
7487 	if (err)
7488 		goto err_sw_init;
7489 
7490 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
7491 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
7492 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
7493 
7494 	err = ei->get_variants(adapter);
7495 	if (err)
7496 		goto err_hw_init;
7497 
7498 	if ((adapter->flags & FLAG_IS_ICH) &&
7499 	    (adapter->flags & FLAG_READ_ONLY_NVM) &&
7500 	    (hw->mac.type < e1000_pch_spt))
7501 		e1000e_write_protect_nvm_ich8lan(&adapter->hw);
7502 
7503 	hw->mac.ops.get_bus_info(&adapter->hw);
7504 
7505 	adapter->hw.phy.autoneg_wait_to_complete = 0;
7506 
7507 	/* Copper options */
7508 	if (adapter->hw.phy.media_type == e1000_media_type_copper) {
7509 		adapter->hw.phy.mdix = AUTO_ALL_MODES;
7510 		adapter->hw.phy.disable_polarity_correction = 0;
7511 		adapter->hw.phy.ms_type = e1000_ms_hw_default;
7512 	}
7513 
7514 	if (hw->phy.ops.check_reset_block && hw->phy.ops.check_reset_block(hw))
7515 		dev_info(&pdev->dev,
7516 			 "PHY reset is blocked due to SOL/IDER session.\n");
7517 
7518 	/* Set initial default active device features */
7519 	netdev->features = (NETIF_F_SG |
7520 			    NETIF_F_HW_VLAN_CTAG_RX |
7521 			    NETIF_F_HW_VLAN_CTAG_TX |
7522 			    NETIF_F_TSO |
7523 			    NETIF_F_TSO6 |
7524 			    NETIF_F_RXHASH |
7525 			    NETIF_F_RXCSUM |
7526 			    NETIF_F_HW_CSUM);
7527 
7528 	/* disable TSO for pcie and 10/100 speeds to avoid
7529 	 * some hardware issues and for i219 to fix transfer
7530 	 * speed being capped at 60%
7531 	 */
7532 	if (!(adapter->flags & FLAG_TSO_FORCE)) {
7533 		switch (adapter->link_speed) {
7534 		case SPEED_10:
7535 		case SPEED_100:
7536 			e_info("10/100 speed: disabling TSO\n");
7537 			netdev->features &= ~NETIF_F_TSO;
7538 			netdev->features &= ~NETIF_F_TSO6;
7539 			break;
7540 		case SPEED_1000:
7541 			netdev->features |= NETIF_F_TSO;
7542 			netdev->features |= NETIF_F_TSO6;
7543 			break;
7544 		default:
7545 			/* oops */
7546 			break;
7547 		}
7548 		if (hw->mac.type == e1000_pch_spt) {
7549 			netdev->features &= ~NETIF_F_TSO;
7550 			netdev->features &= ~NETIF_F_TSO6;
7551 		}
7552 	}
7553 
7554 	/* Set user-changeable features (subset of all device features) */
7555 	netdev->hw_features = netdev->features;
7556 	netdev->hw_features |= NETIF_F_RXFCS;
7557 	netdev->priv_flags |= IFF_SUPP_NOFCS;
7558 	netdev->hw_features |= NETIF_F_RXALL;
7559 
7560 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER)
7561 		netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
7562 
7563 	netdev->vlan_features |= (NETIF_F_SG |
7564 				  NETIF_F_TSO |
7565 				  NETIF_F_TSO6 |
7566 				  NETIF_F_HW_CSUM);
7567 
7568 	netdev->priv_flags |= IFF_UNICAST_FLT;
7569 
7570 	netdev->features |= NETIF_F_HIGHDMA;
7571 	netdev->vlan_features |= NETIF_F_HIGHDMA;
7572 
7573 	/* MTU range: 68 - max_hw_frame_size */
7574 	netdev->min_mtu = ETH_MIN_MTU;
7575 	netdev->max_mtu = adapter->max_hw_frame_size -
7576 			  (VLAN_ETH_HLEN + ETH_FCS_LEN);
7577 
7578 	if (e1000e_enable_mng_pass_thru(&adapter->hw))
7579 		adapter->flags |= FLAG_MNG_PT_ENABLED;
7580 
7581 	/* before reading the NVM, reset the controller to
7582 	 * put the device in a known good starting state
7583 	 */
7584 	adapter->hw.mac.ops.reset_hw(&adapter->hw);
7585 
7586 	/* systems with ASPM and others may see the checksum fail on the first
7587 	 * attempt. Let's give it a few tries
7588 	 */
7589 	for (i = 0;; i++) {
7590 		if (e1000_validate_nvm_checksum(&adapter->hw) >= 0)
7591 			break;
7592 		if (i == 2) {
7593 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
7594 			err = -EIO;
7595 			goto err_eeprom;
7596 		}
7597 	}
7598 
7599 	e1000_eeprom_checks(adapter);
7600 
7601 	/* copy the MAC address */
7602 	if (e1000e_read_mac_addr(&adapter->hw))
7603 		dev_err(&pdev->dev,
7604 			"NVM Read Error while reading MAC address\n");
7605 
7606 	eth_hw_addr_set(netdev, adapter->hw.mac.addr);
7607 
7608 	if (!is_valid_ether_addr(netdev->dev_addr)) {
7609 		dev_err(&pdev->dev, "Invalid MAC Address: %pM\n",
7610 			netdev->dev_addr);
7611 		err = -EIO;
7612 		goto err_eeprom;
7613 	}
7614 
7615 	timer_setup(&adapter->watchdog_timer, e1000_watchdog, 0);
7616 	timer_setup(&adapter->phy_info_timer, e1000_update_phy_info, 0);
7617 
7618 	INIT_WORK(&adapter->reset_task, e1000_reset_task);
7619 	INIT_WORK(&adapter->watchdog_task, e1000_watchdog_task);
7620 	INIT_WORK(&adapter->downshift_task, e1000e_downshift_workaround);
7621 	INIT_WORK(&adapter->update_phy_task, e1000e_update_phy_task);
7622 	INIT_WORK(&adapter->print_hang_task, e1000_print_hw_hang);
7623 
7624 	/* Initialize link parameters. User can change them with ethtool */
7625 	adapter->hw.mac.autoneg = 1;
7626 	adapter->fc_autoneg = true;
7627 	adapter->hw.fc.requested_mode = e1000_fc_default;
7628 	adapter->hw.fc.current_mode = e1000_fc_default;
7629 	adapter->hw.phy.autoneg_advertised = 0x2f;
7630 
7631 	/* Initial Wake on LAN setting - If APM wake is enabled in
7632 	 * the EEPROM, enable the ACPI Magic Packet filter
7633 	 */
7634 	if (adapter->flags & FLAG_APME_IN_WUC) {
7635 		/* APME bit in EEPROM is mapped to WUC.APME */
7636 		eeprom_data = er32(WUC);
7637 		eeprom_apme_mask = E1000_WUC_APME;
7638 		if ((hw->mac.type > e1000_ich10lan) &&
7639 		    (eeprom_data & E1000_WUC_PHY_WAKE))
7640 			adapter->flags2 |= FLAG2_HAS_PHY_WAKEUP;
7641 	} else if (adapter->flags & FLAG_APME_IN_CTRL3) {
7642 		if (adapter->flags & FLAG_APME_CHECK_PORT_B &&
7643 		    (adapter->hw.bus.func == 1))
7644 			ret_val = e1000_read_nvm(&adapter->hw,
7645 					      NVM_INIT_CONTROL3_PORT_B,
7646 					      1, &eeprom_data);
7647 		else
7648 			ret_val = e1000_read_nvm(&adapter->hw,
7649 					      NVM_INIT_CONTROL3_PORT_A,
7650 					      1, &eeprom_data);
7651 	}
7652 
7653 	/* fetch WoL from EEPROM */
7654 	if (ret_val)
7655 		e_dbg("NVM read error getting WoL initial values: %d\n", ret_val);
7656 	else if (eeprom_data & eeprom_apme_mask)
7657 		adapter->eeprom_wol |= E1000_WUFC_MAG;
7658 
7659 	/* now that we have the eeprom settings, apply the special cases
7660 	 * where the eeprom may be wrong or the board simply won't support
7661 	 * wake on lan on a particular port
7662 	 */
7663 	if (!(adapter->flags & FLAG_HAS_WOL))
7664 		adapter->eeprom_wol = 0;
7665 
7666 	/* initialize the wol settings based on the eeprom settings */
7667 	adapter->wol = adapter->eeprom_wol;
7668 
7669 	/* make sure adapter isn't asleep if manageability is enabled */
7670 	if (adapter->wol || (adapter->flags & FLAG_MNG_PT_ENABLED) ||
7671 	    (hw->mac.ops.check_mng_mode(hw)))
7672 		device_wakeup_enable(&pdev->dev);
7673 
7674 	/* save off EEPROM version number */
7675 	ret_val = e1000_read_nvm(&adapter->hw, 5, 1, &adapter->eeprom_vers);
7676 
7677 	if (ret_val) {
7678 		e_dbg("NVM read error getting EEPROM version: %d\n", ret_val);
7679 		adapter->eeprom_vers = 0;
7680 	}
7681 
7682 	/* init PTP hardware clock */
7683 	e1000e_ptp_init(adapter);
7684 
7685 	/* disable K1 by default on known problematic systems */
7686 	if (hw->mac.type >= e1000_pch_mtp && dmi_check_system(disable_k1_list))
7687 		adapter->flags2 |= FLAG2_DISABLE_K1;
7688 
7689 	/* reset the hardware with the new settings */
7690 	e1000e_reset(adapter);
7691 
7692 	/* If the controller has AMT, do not set DRV_LOAD until the interface
7693 	 * is up.  For all other cases, let the f/w know that the h/w is now
7694 	 * under the control of the driver.
7695 	 */
7696 	if (!(adapter->flags & FLAG_HAS_AMT))
7697 		e1000e_get_hw_control(adapter);
7698 
7699 	if (hw->mac.type >= e1000_pch_cnp)
7700 		adapter->flags2 |= FLAG2_ENABLE_S0IX_FLOWS;
7701 
7702 	strscpy(netdev->name, "eth%d", sizeof(netdev->name));
7703 	err = register_netdev(netdev);
7704 	if (err)
7705 		goto err_register;
7706 
7707 	/* carrier off reporting is important to ethtool even BEFORE open */
7708 	netif_carrier_off(netdev);
7709 
7710 	e1000_print_device_info(adapter);
7711 
7712 	dev_pm_set_driver_flags(&pdev->dev, DPM_FLAG_SMART_PREPARE);
7713 
7714 	if (pci_dev_run_wake(pdev))
7715 		pm_runtime_put_noidle(&pdev->dev);
7716 
7717 	return 0;
7718 
7719 err_register:
7720 	if (!(adapter->flags & FLAG_HAS_AMT))
7721 		e1000e_release_hw_control(adapter);
7722 	e1000e_ptp_remove(adapter);
7723 err_eeprom:
7724 	if (hw->phy.ops.check_reset_block && !hw->phy.ops.check_reset_block(hw))
7725 		e1000_phy_hw_reset(&adapter->hw);
7726 err_hw_init:
7727 	kfree(adapter->tx_ring);
7728 	kfree(adapter->rx_ring);
7729 err_sw_init:
7730 	if ((adapter->hw.flash_address) && (hw->mac.type < e1000_pch_spt))
7731 		iounmap(adapter->hw.flash_address);
7732 	e1000e_reset_interrupt_capability(adapter);
7733 err_flashmap:
7734 	iounmap(adapter->hw.hw_addr);
7735 err_ioremap:
7736 	free_netdev(netdev);
7737 err_alloc_etherdev:
7738 	pci_release_mem_regions(pdev);
7739 err_pci_reg:
7740 err_dma:
7741 	pci_disable_device(pdev);
7742 	return err;
7743 }
7744 
7745 /**
7746  * e1000_remove - Device Removal Routine
7747  * @pdev: PCI device information struct
7748  *
7749  * e1000_remove is called by the PCI subsystem to alert the driver
7750  * that it should release a PCI device.  This could be caused by a
7751  * Hot-Plug event, or because the driver is going to be removed from
7752  * memory.
7753  **/
7754 static void e1000_remove(struct pci_dev *pdev)
7755 {
7756 	struct net_device *netdev = pci_get_drvdata(pdev);
7757 	struct e1000_adapter *adapter = netdev_priv(netdev);
7758 
7759 	e1000e_ptp_remove(adapter);
7760 
7761 	/* The timers may be rescheduled, so explicitly disable them
7762 	 * from being rescheduled.
7763 	 */
7764 	set_bit(__E1000_DOWN, &adapter->state);
7765 	timer_delete_sync(&adapter->watchdog_timer);
7766 	timer_delete_sync(&adapter->phy_info_timer);
7767 
7768 	cancel_work_sync(&adapter->reset_task);
7769 	cancel_work_sync(&adapter->watchdog_task);
7770 	cancel_work_sync(&adapter->downshift_task);
7771 	cancel_work_sync(&adapter->update_phy_task);
7772 	cancel_work_sync(&adapter->print_hang_task);
7773 
7774 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
7775 		cancel_work_sync(&adapter->tx_hwtstamp_work);
7776 		if (adapter->tx_hwtstamp_skb) {
7777 			dev_consume_skb_any(adapter->tx_hwtstamp_skb);
7778 			adapter->tx_hwtstamp_skb = NULL;
7779 		}
7780 	}
7781 
7782 	unregister_netdev(netdev);
7783 
7784 	if (pci_dev_run_wake(pdev))
7785 		pm_runtime_get_noresume(&pdev->dev);
7786 
7787 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
7788 	 * would have already happened in close and is redundant.
7789 	 */
7790 	e1000e_release_hw_control(adapter);
7791 
7792 	e1000e_reset_interrupt_capability(adapter);
7793 	kfree(adapter->tx_ring);
7794 	kfree(adapter->rx_ring);
7795 
7796 	iounmap(adapter->hw.hw_addr);
7797 	if ((adapter->hw.flash_address) &&
7798 	    (adapter->hw.mac.type < e1000_pch_spt))
7799 		iounmap(adapter->hw.flash_address);
7800 	pci_release_mem_regions(pdev);
7801 
7802 	free_netdev(netdev);
7803 
7804 	pci_disable_device(pdev);
7805 }
7806 
7807 /* PCI Error Recovery (ERS) */
7808 static const struct pci_error_handlers e1000_err_handler = {
7809 	.error_detected = e1000_io_error_detected,
7810 	.slot_reset = e1000_io_slot_reset,
7811 	.resume = e1000_io_resume,
7812 };
7813 
7814 static const struct pci_device_id e1000_pci_tbl[] = {
7815 	{
7816 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_COPPER),
7817 		.driver_data = board_82571,
7818 	}, {
7819 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_FIBER),
7820 		.driver_data = board_82571,
7821 	}, {
7822 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER),
7823 		.driver_data = board_82571,
7824 	}, {
7825 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER_LP),
7826 		.driver_data = board_82571,
7827 	}, {
7828 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_FIBER),
7829 		.driver_data = board_82571,
7830 	}, {
7831 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES),
7832 		.driver_data = board_82571,
7833 	}, {
7834 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_DUAL),
7835 		.driver_data = board_82571,
7836 	}, {
7837 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_QUAD),
7838 		.driver_data = board_82571,
7839 	}, {
7840 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82571PT_QUAD_COPPER),
7841 		.driver_data = board_82571,
7842 	},
7843 
7844 	{
7845 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI),
7846 		.driver_data = board_82572,
7847 	}, {
7848 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_COPPER),
7849 		.driver_data = board_82572,
7850 	}, {
7851 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_FIBER),
7852 		.driver_data = board_82572,
7853 	}, {
7854 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_SERDES),
7855 		.driver_data = board_82572,
7856 	},
7857 
7858 	{
7859 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E),
7860 		.driver_data = board_82573,
7861 	}, {
7862 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E_IAMT),
7863 		.driver_data = board_82573,
7864 	}, {
7865 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82573L),
7866 		.driver_data = board_82573,
7867 	},
7868 
7869 	{
7870 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82574L),
7871 		.driver_data = board_82574,
7872 	}, {
7873 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82574LA),
7874 		.driver_data = board_82574,
7875 	}, {
7876 		PCI_VDEVICE(INTEL, E1000_DEV_ID_82583V),
7877 		.driver_data = board_82583,
7878 	},
7879 
7880 	{
7881 		PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_DPT),
7882 		.driver_data = board_80003es2lan,
7883 	}, {
7884 		PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_SPT),
7885 		.driver_data = board_80003es2lan,
7886 	}, {
7887 		PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_DPT),
7888 		.driver_data = board_80003es2lan,
7889 	}, {
7890 		PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_SPT),
7891 		.driver_data = board_80003es2lan,
7892 	},
7893 
7894 	{
7895 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE),
7896 		.driver_data = board_ich8lan,
7897 	}, {
7898 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_G),
7899 		.driver_data = board_ich8lan,
7900 	}, {
7901 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_GT),
7902 		.driver_data = board_ich8lan,
7903 	}, {
7904 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_AMT),
7905 		.driver_data = board_ich8lan,
7906 	}, {
7907 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_C),
7908 		.driver_data = board_ich8lan,
7909 	}, {
7910 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M),
7911 		.driver_data = board_ich8lan,
7912 	}, {
7913 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M_AMT),
7914 		.driver_data = board_ich8lan,
7915 	}, {
7916 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_82567V_3),
7917 		.driver_data = board_ich8lan,
7918 	},
7919 
7920 	{
7921 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE),
7922 		.driver_data = board_ich9lan,
7923 	}, {
7924 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_G),
7925 		.driver_data = board_ich9lan,
7926 	}, {
7927 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_GT),
7928 		.driver_data = board_ich9lan,
7929 	}, {
7930 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_AMT),
7931 		.driver_data = board_ich9lan,
7932 	}, {
7933 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_C),
7934 		.driver_data = board_ich9lan,
7935 	}, {
7936 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_BM),
7937 		.driver_data = board_ich9lan,
7938 	}, {
7939 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M),
7940 		.driver_data = board_ich9lan,
7941 	}, {
7942 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_AMT),
7943 		.driver_data = board_ich9lan,
7944 	}, {
7945 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_V),
7946 		.driver_data = board_ich9lan
7947 	},
7948 
7949 	{
7950 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LM),
7951 		.driver_data = board_ich9lan,
7952 	}, {
7953 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LF),
7954 		.driver_data = board_ich9lan,
7955 	}, {
7956 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_V),
7957 		.driver_data = board_ich9lan,
7958 	},
7959 
7960 	{
7961 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LM),
7962 		.driver_data = board_ich10lan,
7963 	}, {
7964 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LF),
7965 		.driver_data = board_ich10lan,
7966 	}, {
7967 		PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_V),
7968 		.driver_data = board_ich10lan,
7969 	},
7970 
7971 	{
7972 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LM),
7973 		.driver_data = board_pchlan,
7974 	}, {
7975 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LC),
7976 		.driver_data = board_pchlan,
7977 	}, {
7978 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DM),
7979 		.driver_data = board_pchlan,
7980 	}, {
7981 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DC),
7982 		.driver_data = board_pchlan
7983 	},
7984 
7985 	{
7986 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_LM),
7987 		.driver_data = board_pch2lan,
7988 	}, {
7989 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_V),
7990 		.driver_data = board_pch2lan
7991 	},
7992 
7993 	{
7994 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_LM),
7995 		.driver_data = board_pch_lpt,
7996 	}, {
7997 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_V),
7998 		.driver_data = board_pch_lpt,
7999 	}, {
8000 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_LM),
8001 		.driver_data = board_pch_lpt,
8002 	}, {
8003 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_V),
8004 		.driver_data = board_pch_lpt,
8005 	}, {
8006 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM2),
8007 		.driver_data = board_pch_lpt,
8008 	}, {
8009 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V2),
8010 		.driver_data = board_pch_lpt,
8011 	}, {
8012 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM3),
8013 		.driver_data = board_pch_lpt,
8014 	}, {
8015 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V3),
8016 		.driver_data = board_pch_lpt,
8017 	}, {
8018 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM),
8019 		.driver_data = board_pch_spt,
8020 	}, {
8021 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V),
8022 		.driver_data = board_pch_spt,
8023 	}, {
8024 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM2),
8025 		.driver_data = board_pch_spt,
8026 	}, {
8027 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V2),
8028 		.driver_data = board_pch_spt,
8029 	}, {
8030 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LBG_I219_LM3),
8031 		.driver_data = board_pch_spt,
8032 	}, {
8033 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM4),
8034 		.driver_data = board_pch_spt,
8035 	}, {
8036 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V4),
8037 		.driver_data = board_pch_spt,
8038 	}, {
8039 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM5),
8040 		.driver_data = board_pch_spt,
8041 	}, {
8042 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V5),
8043 		.driver_data = board_pch_spt,
8044 	}, {
8045 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM6),
8046 		.driver_data = board_pch_cnp,
8047 	}, {
8048 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V6),
8049 		.driver_data = board_pch_cnp,
8050 	}, {
8051 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM7),
8052 		.driver_data = board_pch_cnp,
8053 	}, {
8054 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V7),
8055 		.driver_data = board_pch_cnp,
8056 	}, {
8057 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM8),
8058 		.driver_data = board_pch_cnp,
8059 	}, {
8060 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V8),
8061 		.driver_data = board_pch_cnp,
8062 	}, {
8063 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM9),
8064 		.driver_data = board_pch_cnp,
8065 	}, {
8066 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V9),
8067 		.driver_data = board_pch_cnp,
8068 	}, {
8069 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM10),
8070 		.driver_data = board_pch_cnp,
8071 	}, {
8072 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V10),
8073 		.driver_data = board_pch_cnp,
8074 	}, {
8075 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM11),
8076 		.driver_data = board_pch_cnp,
8077 	}, {
8078 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V11),
8079 		.driver_data = board_pch_cnp,
8080 	}, {
8081 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM12),
8082 		.driver_data = board_pch_spt,
8083 	}, {
8084 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V12),
8085 		.driver_data = board_pch_spt,
8086 	}, {
8087 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM13),
8088 		.driver_data = board_pch_tgp,
8089 	}, {
8090 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V13),
8091 		.driver_data = board_pch_tgp,
8092 	}, {
8093 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM14),
8094 		.driver_data = board_pch_tgp,
8095 	}, {
8096 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V14),
8097 		.driver_data = board_pch_tgp,
8098 	}, {
8099 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM15),
8100 		.driver_data = board_pch_tgp,
8101 	}, {
8102 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V15),
8103 		.driver_data = board_pch_tgp,
8104 	}, {
8105 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_LM23),
8106 		.driver_data = board_pch_adp,
8107 	}, {
8108 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_V23),
8109 		.driver_data = board_pch_adp,
8110 	}, {
8111 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM16),
8112 		.driver_data = board_pch_adp,
8113 	}, {
8114 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V16),
8115 		.driver_data = board_pch_adp,
8116 	}, {
8117 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM17),
8118 		.driver_data = board_pch_adp,
8119 	}, {
8120 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V17),
8121 		.driver_data = board_pch_adp,
8122 	}, {
8123 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_LM22),
8124 		.driver_data = board_pch_adp,
8125 	}, {
8126 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_V22),
8127 		.driver_data = board_pch_adp,
8128 	}, {
8129 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM19),
8130 		.driver_data = board_pch_adp,
8131 	}, {
8132 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V19),
8133 		.driver_data = board_pch_adp,
8134 	}, {
8135 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_LM18),
8136 		.driver_data = board_pch_mtp,
8137 	}, {
8138 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_V18),
8139 		.driver_data = board_pch_mtp,
8140 	}, {
8141 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_LM20),
8142 		.driver_data = board_pch_mtp,
8143 	}, {
8144 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_V20),
8145 		.driver_data = board_pch_mtp,
8146 	}, {
8147 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_LM21),
8148 		.driver_data = board_pch_mtp,
8149 	}, {
8150 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_V21),
8151 		.driver_data = board_pch_mtp,
8152 	}, {
8153 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ARL_I219_LM24),
8154 		.driver_data = board_pch_mtp,
8155 	}, {
8156 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ARL_I219_V24),
8157 		.driver_data = board_pch_mtp,
8158 	}, {
8159 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_LM25),
8160 		.driver_data = board_pch_ptp,
8161 	}, {
8162 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_V25),
8163 		.driver_data = board_pch_ptp,
8164 	}, {
8165 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_LM27),
8166 		.driver_data = board_pch_ptp,
8167 	}, {
8168 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_V27),
8169 		.driver_data = board_pch_ptp,
8170 	}, {
8171 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_NVL_I219_LM29),
8172 		.driver_data = board_pch_ptp,
8173 	}, {
8174 		PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_NVL_I219_V29),
8175 		.driver_data = board_pch_ptp
8176 	},
8177 
8178 	{ }	/* terminate list */
8179 };
8180 MODULE_DEVICE_TABLE(pci, e1000_pci_tbl);
8181 
8182 static const struct dev_pm_ops e1000e_pm_ops = {
8183 	.prepare	= e1000e_pm_prepare,
8184 	.suspend	= e1000e_pm_suspend,
8185 	.resume		= e1000e_pm_resume,
8186 	.freeze		= e1000e_pm_freeze,
8187 	.thaw		= e1000e_pm_thaw,
8188 	.poweroff	= e1000e_pm_suspend,
8189 	.restore	= e1000e_pm_resume,
8190 	RUNTIME_PM_OPS(e1000e_pm_runtime_suspend, e1000e_pm_runtime_resume,
8191 		       e1000e_pm_runtime_idle)
8192 };
8193 
8194 /* PCI Device API Driver */
8195 static struct pci_driver e1000_driver = {
8196 	.name     = e1000e_driver_name,
8197 	.id_table = e1000_pci_tbl,
8198 	.probe    = e1000_probe,
8199 	.remove   = e1000_remove,
8200 	.driver.pm = pm_ptr(&e1000e_pm_ops),
8201 	.shutdown = e1000_shutdown,
8202 	.err_handler = &e1000_err_handler
8203 };
8204 
8205 /**
8206  * e1000_init_module - Driver Registration Routine
8207  *
8208  * e1000_init_module is the first routine called when the driver is
8209  * loaded. All it does is register with the PCI subsystem.
8210  **/
8211 static int __init e1000_init_module(void)
8212 {
8213 	pr_info("Intel(R) PRO/1000 Network Driver\n");
8214 	pr_info("Copyright(c) 1999 - 2015 Intel Corporation.\n");
8215 
8216 	return pci_register_driver(&e1000_driver);
8217 }
8218 module_init(e1000_init_module);
8219 
8220 /**
8221  * e1000_exit_module - Driver Exit Cleanup Routine
8222  *
8223  * e1000_exit_module is called just before the driver is removed
8224  * from memory.
8225  **/
8226 static void __exit e1000_exit_module(void)
8227 {
8228 	pci_unregister_driver(&e1000_driver);
8229 }
8230 module_exit(e1000_exit_module);
8231 
8232 MODULE_DESCRIPTION("Intel(R) PRO/1000 Network Driver");
8233 MODULE_LICENSE("GPL v2");
8234 
8235 /* netdev.c */
8236