xref: /linux/drivers/net/ethernet/intel/e1000e/netdev.c (revision 69050f8d6d075dc01af7a5f2f550a8067510366f)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright(c) 1999 - 2018 Intel Corporation. */
3 
4 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
5 
6 #include <linux/module.h>
7 #include <linux/types.h>
8 #include <linux/init.h>
9 #include <linux/pci.h>
10 #include <linux/vmalloc.h>
11 #include <linux/pagemap.h>
12 #include <linux/delay.h>
13 #include <linux/netdevice.h>
14 #include <linux/interrupt.h>
15 #include <linux/tcp.h>
16 #include <linux/ipv6.h>
17 #include <linux/slab.h>
18 #include <net/checksum.h>
19 #include <net/ip6_checksum.h>
20 #include <linux/ethtool.h>
21 #include <linux/if_vlan.h>
22 #include <linux/cpu.h>
23 #include <linux/smp.h>
24 #include <linux/pm_qos.h>
25 #include <linux/pm_runtime.h>
26 #include <linux/prefetch.h>
27 #include <linux/suspend.h>
28 
29 #include "e1000.h"
30 #define CREATE_TRACE_POINTS
31 #include "e1000e_trace.h"
32 
33 char e1000e_driver_name[] = "e1000e";
34 
35 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
36 static int debug = -1;
37 module_param(debug, int, 0);
38 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
39 
40 static const struct e1000_info *e1000_info_tbl[] = {
41 	[board_82571]		= &e1000_82571_info,
42 	[board_82572]		= &e1000_82572_info,
43 	[board_82573]		= &e1000_82573_info,
44 	[board_82574]		= &e1000_82574_info,
45 	[board_82583]		= &e1000_82583_info,
46 	[board_80003es2lan]	= &e1000_es2_info,
47 	[board_ich8lan]		= &e1000_ich8_info,
48 	[board_ich9lan]		= &e1000_ich9_info,
49 	[board_ich10lan]	= &e1000_ich10_info,
50 	[board_pchlan]		= &e1000_pch_info,
51 	[board_pch2lan]		= &e1000_pch2_info,
52 	[board_pch_lpt]		= &e1000_pch_lpt_info,
53 	[board_pch_spt]		= &e1000_pch_spt_info,
54 	[board_pch_cnp]		= &e1000_pch_cnp_info,
55 	[board_pch_tgp]		= &e1000_pch_tgp_info,
56 	[board_pch_adp]		= &e1000_pch_adp_info,
57 	[board_pch_mtp]		= &e1000_pch_mtp_info,
58 };
59 
60 struct e1000_reg_info {
61 	u32 ofs;
62 	char *name;
63 };
64 
65 static const struct e1000_reg_info e1000_reg_info_tbl[] = {
66 	/* General Registers */
67 	{E1000_CTRL, "CTRL"},
68 	{E1000_STATUS, "STATUS"},
69 	{E1000_CTRL_EXT, "CTRL_EXT"},
70 
71 	/* Interrupt Registers */
72 	{E1000_ICR, "ICR"},
73 
74 	/* Rx Registers */
75 	{E1000_RCTL, "RCTL"},
76 	{E1000_RDLEN(0), "RDLEN"},
77 	{E1000_RDH(0), "RDH"},
78 	{E1000_RDT(0), "RDT"},
79 	{E1000_RDTR, "RDTR"},
80 	{E1000_RXDCTL(0), "RXDCTL"},
81 	{E1000_ERT, "ERT"},
82 	{E1000_RDBAL(0), "RDBAL"},
83 	{E1000_RDBAH(0), "RDBAH"},
84 	{E1000_RDFH, "RDFH"},
85 	{E1000_RDFT, "RDFT"},
86 	{E1000_RDFHS, "RDFHS"},
87 	{E1000_RDFTS, "RDFTS"},
88 	{E1000_RDFPC, "RDFPC"},
89 
90 	/* Tx Registers */
91 	{E1000_TCTL, "TCTL"},
92 	{E1000_TDBAL(0), "TDBAL"},
93 	{E1000_TDBAH(0), "TDBAH"},
94 	{E1000_TDLEN(0), "TDLEN"},
95 	{E1000_TDH(0), "TDH"},
96 	{E1000_TDT(0), "TDT"},
97 	{E1000_TIDV, "TIDV"},
98 	{E1000_TXDCTL(0), "TXDCTL"},
99 	{E1000_TADV, "TADV"},
100 	{E1000_TARC(0), "TARC"},
101 	{E1000_TDFH, "TDFH"},
102 	{E1000_TDFT, "TDFT"},
103 	{E1000_TDFHS, "TDFHS"},
104 	{E1000_TDFTS, "TDFTS"},
105 	{E1000_TDFPC, "TDFPC"},
106 
107 	/* List Terminator */
108 	{0, NULL}
109 };
110 
111 /**
112  * __ew32_prepare - prepare to write to MAC CSR register on certain parts
113  * @hw: pointer to the HW structure
114  *
115  * When updating the MAC CSR registers, the Manageability Engine (ME) could
116  * be accessing the registers at the same time.  Normally, this is handled in
117  * h/w by an arbiter but on some parts there is a bug that acknowledges Host
118  * accesses later than it should which could result in the register to have
119  * an incorrect value.  Workaround this by checking the FWSM register which
120  * has bit 24 set while ME is accessing MAC CSR registers, wait if it is set
121  * and try again a number of times.
122  **/
123 static void __ew32_prepare(struct e1000_hw *hw)
124 {
125 	s32 i = E1000_ICH_FWSM_PCIM2PCI_COUNT;
126 
127 	while ((er32(FWSM) & E1000_ICH_FWSM_PCIM2PCI) && --i)
128 		udelay(50);
129 }
130 
131 void __ew32(struct e1000_hw *hw, unsigned long reg, u32 val)
132 {
133 	if (hw->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
134 		__ew32_prepare(hw);
135 
136 	writel(val, hw->hw_addr + reg);
137 }
138 
139 /**
140  * e1000_regdump - register printout routine
141  * @hw: pointer to the HW structure
142  * @reginfo: pointer to the register info table
143  **/
144 static void e1000_regdump(struct e1000_hw *hw, struct e1000_reg_info *reginfo)
145 {
146 	int n = 0;
147 	char rname[16];
148 	u32 regs[8];
149 
150 	switch (reginfo->ofs) {
151 	case E1000_RXDCTL(0):
152 		for (n = 0; n < 2; n++)
153 			regs[n] = __er32(hw, E1000_RXDCTL(n));
154 		break;
155 	case E1000_TXDCTL(0):
156 		for (n = 0; n < 2; n++)
157 			regs[n] = __er32(hw, E1000_TXDCTL(n));
158 		break;
159 	case E1000_TARC(0):
160 		for (n = 0; n < 2; n++)
161 			regs[n] = __er32(hw, E1000_TARC(n));
162 		break;
163 	default:
164 		pr_info("%-15s %08x\n",
165 			reginfo->name, __er32(hw, reginfo->ofs));
166 		return;
167 	}
168 
169 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-1]");
170 	pr_info("%-15s %08x %08x\n", rname, regs[0], regs[1]);
171 }
172 
173 static void e1000e_dump_ps_pages(struct e1000_adapter *adapter,
174 				 struct e1000_buffer *bi)
175 {
176 	int i;
177 	struct e1000_ps_page *ps_page;
178 
179 	for (i = 0; i < adapter->rx_ps_pages; i++) {
180 		ps_page = &bi->ps_pages[i];
181 
182 		if (ps_page->page) {
183 			pr_info("packet dump for ps_page %d:\n", i);
184 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
185 				       16, 1, page_address(ps_page->page),
186 				       PAGE_SIZE, true);
187 		}
188 	}
189 }
190 
191 /**
192  * e1000e_dump - Print registers, Tx-ring and Rx-ring
193  * @adapter: board private structure
194  **/
195 static void e1000e_dump(struct e1000_adapter *adapter)
196 {
197 	struct net_device *netdev = adapter->netdev;
198 	struct e1000_hw *hw = &adapter->hw;
199 	struct e1000_reg_info *reginfo;
200 	struct e1000_ring *tx_ring = adapter->tx_ring;
201 	struct e1000_tx_desc *tx_desc;
202 	struct my_u0 {
203 		__le64 a;
204 		__le64 b;
205 	} *u0;
206 	struct e1000_buffer *buffer_info;
207 	struct e1000_ring *rx_ring = adapter->rx_ring;
208 	union e1000_rx_desc_packet_split *rx_desc_ps;
209 	union e1000_rx_desc_extended *rx_desc;
210 	struct my_u1 {
211 		__le64 a;
212 		__le64 b;
213 		__le64 c;
214 		__le64 d;
215 	} *u1;
216 	u32 staterr;
217 	int i = 0;
218 
219 	if (!netif_msg_hw(adapter))
220 		return;
221 
222 	/* Print netdevice Info */
223 	if (netdev) {
224 		dev_info(&adapter->pdev->dev, "Net device Info\n");
225 		pr_info("Device Name     state            trans_start\n");
226 		pr_info("%-15s %016lX %016lX\n", netdev->name,
227 			netdev->state, dev_trans_start(netdev));
228 	}
229 
230 	/* Print Registers */
231 	dev_info(&adapter->pdev->dev, "Register Dump\n");
232 	pr_info(" Register Name   Value\n");
233 	for (reginfo = (struct e1000_reg_info *)e1000_reg_info_tbl;
234 	     reginfo->name; reginfo++) {
235 		e1000_regdump(hw, reginfo);
236 	}
237 
238 	/* Print Tx Ring Summary */
239 	if (!netdev || !netif_running(netdev))
240 		return;
241 
242 	dev_info(&adapter->pdev->dev, "Tx Ring Summary\n");
243 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
244 	buffer_info = &tx_ring->buffer_info[tx_ring->next_to_clean];
245 	pr_info(" %5d %5X %5X %016llX %04X %3X %016llX\n",
246 		0, tx_ring->next_to_use, tx_ring->next_to_clean,
247 		(unsigned long long)buffer_info->dma,
248 		buffer_info->length,
249 		buffer_info->next_to_watch,
250 		(unsigned long long)buffer_info->time_stamp);
251 
252 	/* Print Tx Ring */
253 	if (!netif_msg_tx_done(adapter))
254 		goto rx_ring_summary;
255 
256 	dev_info(&adapter->pdev->dev, "Tx Ring Dump\n");
257 
258 	/* Transmit Descriptor Formats - DEXT[29] is 0 (Legacy) or 1 (Extended)
259 	 *
260 	 * Legacy Transmit Descriptor
261 	 *   +--------------------------------------------------------------+
262 	 * 0 |         Buffer Address [63:0] (Reserved on Write Back)       |
263 	 *   +--------------------------------------------------------------+
264 	 * 8 | Special  |    CSS     | Status |  CMD    |  CSO   |  Length  |
265 	 *   +--------------------------------------------------------------+
266 	 *   63       48 47        36 35    32 31     24 23    16 15        0
267 	 *
268 	 * Extended Context Descriptor (DTYP=0x0) for TSO or checksum offload
269 	 *   63      48 47    40 39       32 31             16 15    8 7      0
270 	 *   +----------------------------------------------------------------+
271 	 * 0 |  TUCSE  | TUCS0  |   TUCSS   |     IPCSE       | IPCS0 | IPCSS |
272 	 *   +----------------------------------------------------------------+
273 	 * 8 |   MSS   | HDRLEN | RSV | STA | TUCMD | DTYP |      PAYLEN      |
274 	 *   +----------------------------------------------------------------+
275 	 *   63      48 47    40 39 36 35 32 31   24 23  20 19                0
276 	 *
277 	 * Extended Data Descriptor (DTYP=0x1)
278 	 *   +----------------------------------------------------------------+
279 	 * 0 |                     Buffer Address [63:0]                      |
280 	 *   +----------------------------------------------------------------+
281 	 * 8 | VLAN tag |  POPTS  | Rsvd | Status | Command | DTYP |  DTALEN  |
282 	 *   +----------------------------------------------------------------+
283 	 *   63       48 47     40 39  36 35    32 31     24 23  20 19        0
284 	 */
285 	pr_info("Tl[desc]     [address 63:0  ] [SpeCssSCmCsLen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Legacy format\n");
286 	pr_info("Tc[desc]     [Ce CoCsIpceCoS] [MssHlRSCm0Plen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Context format\n");
287 	pr_info("Td[desc]     [address 63:0  ] [VlaPoRSCm1Dlen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Data format\n");
288 	for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
289 		const char *next_desc;
290 		tx_desc = E1000_TX_DESC(*tx_ring, i);
291 		buffer_info = &tx_ring->buffer_info[i];
292 		u0 = (struct my_u0 *)tx_desc;
293 		if (i == tx_ring->next_to_use && i == tx_ring->next_to_clean)
294 			next_desc = " NTC/U";
295 		else if (i == tx_ring->next_to_use)
296 			next_desc = " NTU";
297 		else if (i == tx_ring->next_to_clean)
298 			next_desc = " NTC";
299 		else
300 			next_desc = "";
301 		pr_info("T%c[0x%03X]    %016llX %016llX %016llX %04X  %3X %016llX %p%s\n",
302 			(!(le64_to_cpu(u0->b) & BIT(29)) ? 'l' :
303 			 ((le64_to_cpu(u0->b) & BIT(20)) ? 'd' : 'c')),
304 			i,
305 			(unsigned long long)le64_to_cpu(u0->a),
306 			(unsigned long long)le64_to_cpu(u0->b),
307 			(unsigned long long)buffer_info->dma,
308 			buffer_info->length, buffer_info->next_to_watch,
309 			(unsigned long long)buffer_info->time_stamp,
310 			buffer_info->skb, next_desc);
311 
312 		if (netif_msg_pktdata(adapter) && buffer_info->skb)
313 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
314 				       16, 1, buffer_info->skb->data,
315 				       buffer_info->skb->len, true);
316 	}
317 
318 	/* Print Rx Ring Summary */
319 rx_ring_summary:
320 	dev_info(&adapter->pdev->dev, "Rx Ring Summary\n");
321 	pr_info("Queue [NTU] [NTC]\n");
322 	pr_info(" %5d %5X %5X\n",
323 		0, rx_ring->next_to_use, rx_ring->next_to_clean);
324 
325 	/* Print Rx Ring */
326 	if (!netif_msg_rx_status(adapter))
327 		return;
328 
329 	dev_info(&adapter->pdev->dev, "Rx Ring Dump\n");
330 	switch (adapter->rx_ps_pages) {
331 	case 1:
332 	case 2:
333 	case 3:
334 		/* [Extended] Packet Split Receive Descriptor Format
335 		 *
336 		 *    +-----------------------------------------------------+
337 		 *  0 |                Buffer Address 0 [63:0]              |
338 		 *    +-----------------------------------------------------+
339 		 *  8 |                Buffer Address 1 [63:0]              |
340 		 *    +-----------------------------------------------------+
341 		 * 16 |                Buffer Address 2 [63:0]              |
342 		 *    +-----------------------------------------------------+
343 		 * 24 |                Buffer Address 3 [63:0]              |
344 		 *    +-----------------------------------------------------+
345 		 */
346 		pr_info("R  [desc]      [buffer 0 63:0 ] [buffer 1 63:0 ] [buffer 2 63:0 ] [buffer 3 63:0 ] [bi->dma       ] [bi->skb] <-- Ext Pkt Split format\n");
347 		/* [Extended] Receive Descriptor (Write-Back) Format
348 		 *
349 		 *   63       48 47    32 31     13 12    8 7    4 3        0
350 		 *   +------------------------------------------------------+
351 		 * 0 | Packet   | IP     |  Rsvd   | MRQ   | Rsvd | MRQ RSS |
352 		 *   | Checksum | Ident  |         | Queue |      |  Type   |
353 		 *   +------------------------------------------------------+
354 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
355 		 *   +------------------------------------------------------+
356 		 *   63       48 47    32 31            20 19               0
357 		 */
358 		pr_info("RWB[desc]      [ck ipid mrqhsh] [vl   l0 ee  es] [ l3  l2  l1 hs] [reserved      ] ---------------- [bi->skb] <-- Ext Rx Write-Back format\n");
359 		for (i = 0; i < rx_ring->count; i++) {
360 			const char *next_desc;
361 			buffer_info = &rx_ring->buffer_info[i];
362 			rx_desc_ps = E1000_RX_DESC_PS(*rx_ring, i);
363 			u1 = (struct my_u1 *)rx_desc_ps;
364 			staterr =
365 			    le32_to_cpu(rx_desc_ps->wb.middle.status_error);
366 
367 			if (i == rx_ring->next_to_use)
368 				next_desc = " NTU";
369 			else if (i == rx_ring->next_to_clean)
370 				next_desc = " NTC";
371 			else
372 				next_desc = "";
373 
374 			if (staterr & E1000_RXD_STAT_DD) {
375 				/* Descriptor Done */
376 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX ---------------- %p%s\n",
377 					"RWB", i,
378 					(unsigned long long)le64_to_cpu(u1->a),
379 					(unsigned long long)le64_to_cpu(u1->b),
380 					(unsigned long long)le64_to_cpu(u1->c),
381 					(unsigned long long)le64_to_cpu(u1->d),
382 					buffer_info->skb, next_desc);
383 			} else {
384 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX %016llX %p%s\n",
385 					"R  ", i,
386 					(unsigned long long)le64_to_cpu(u1->a),
387 					(unsigned long long)le64_to_cpu(u1->b),
388 					(unsigned long long)le64_to_cpu(u1->c),
389 					(unsigned long long)le64_to_cpu(u1->d),
390 					(unsigned long long)buffer_info->dma,
391 					buffer_info->skb, next_desc);
392 
393 				if (netif_msg_pktdata(adapter))
394 					e1000e_dump_ps_pages(adapter,
395 							     buffer_info);
396 			}
397 		}
398 		break;
399 	default:
400 	case 0:
401 		/* Extended Receive Descriptor (Read) Format
402 		 *
403 		 *   +-----------------------------------------------------+
404 		 * 0 |                Buffer Address [63:0]                |
405 		 *   +-----------------------------------------------------+
406 		 * 8 |                      Reserved                       |
407 		 *   +-----------------------------------------------------+
408 		 */
409 		pr_info("R  [desc]      [buf addr 63:0 ] [reserved 63:0 ] [bi->dma       ] [bi->skb] <-- Ext (Read) format\n");
410 		/* Extended Receive Descriptor (Write-Back) Format
411 		 *
412 		 *   63       48 47    32 31    24 23            4 3        0
413 		 *   +------------------------------------------------------+
414 		 *   |     RSS Hash      |        |               |         |
415 		 * 0 +-------------------+  Rsvd  |   Reserved    | MRQ RSS |
416 		 *   | Packet   | IP     |        |               |  Type   |
417 		 *   | Checksum | Ident  |        |               |         |
418 		 *   +------------------------------------------------------+
419 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
420 		 *   +------------------------------------------------------+
421 		 *   63       48 47    32 31            20 19               0
422 		 */
423 		pr_info("RWB[desc]      [cs ipid    mrq] [vt   ln xe  xs] [bi->skb] <-- Ext (Write-Back) format\n");
424 
425 		for (i = 0; i < rx_ring->count; i++) {
426 			const char *next_desc;
427 
428 			buffer_info = &rx_ring->buffer_info[i];
429 			rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
430 			u1 = (struct my_u1 *)rx_desc;
431 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
432 
433 			if (i == rx_ring->next_to_use)
434 				next_desc = " NTU";
435 			else if (i == rx_ring->next_to_clean)
436 				next_desc = " NTC";
437 			else
438 				next_desc = "";
439 
440 			if (staterr & E1000_RXD_STAT_DD) {
441 				/* Descriptor Done */
442 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %p%s\n",
443 					"RWB", i,
444 					(unsigned long long)le64_to_cpu(u1->a),
445 					(unsigned long long)le64_to_cpu(u1->b),
446 					buffer_info->skb, next_desc);
447 			} else {
448 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %p%s\n",
449 					"R  ", i,
450 					(unsigned long long)le64_to_cpu(u1->a),
451 					(unsigned long long)le64_to_cpu(u1->b),
452 					(unsigned long long)buffer_info->dma,
453 					buffer_info->skb, next_desc);
454 
455 				if (netif_msg_pktdata(adapter) &&
456 				    buffer_info->skb)
457 					print_hex_dump(KERN_INFO, "",
458 						       DUMP_PREFIX_ADDRESS, 16,
459 						       1,
460 						       buffer_info->skb->data,
461 						       adapter->rx_buffer_len,
462 						       true);
463 			}
464 		}
465 	}
466 }
467 
468 /**
469  * e1000_desc_unused - calculate if we have unused descriptors
470  * @ring: pointer to ring struct to perform calculation on
471  **/
472 static int e1000_desc_unused(struct e1000_ring *ring)
473 {
474 	if (ring->next_to_clean > ring->next_to_use)
475 		return ring->next_to_clean - ring->next_to_use - 1;
476 
477 	return ring->count + ring->next_to_clean - ring->next_to_use - 1;
478 }
479 
480 /**
481  * e1000e_systim_to_hwtstamp - convert system time value to hw time stamp
482  * @adapter: board private structure
483  * @hwtstamps: time stamp structure to update
484  * @systim: unsigned 64bit system time value.
485  *
486  * Convert the system time value stored in the RX/TXSTMP registers into a
487  * hwtstamp which can be used by the upper level time stamping functions.
488  *
489  * The 'systim_lock' spinlock is used to protect the consistency of the
490  * system time value. This is needed because reading the 64 bit time
491  * value involves reading two 32 bit registers. The first read latches the
492  * value.
493  **/
494 static void e1000e_systim_to_hwtstamp(struct e1000_adapter *adapter,
495 				      struct skb_shared_hwtstamps *hwtstamps,
496 				      u64 systim)
497 {
498 	u64 ns;
499 	unsigned long flags;
500 
501 	spin_lock_irqsave(&adapter->systim_lock, flags);
502 	ns = timecounter_cyc2time(&adapter->tc, systim);
503 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
504 
505 	memset(hwtstamps, 0, sizeof(*hwtstamps));
506 	hwtstamps->hwtstamp = ns_to_ktime(ns);
507 }
508 
509 /**
510  * e1000e_rx_hwtstamp - utility function which checks for Rx time stamp
511  * @adapter: board private structure
512  * @status: descriptor extended error and status field
513  * @skb: particular skb to include time stamp
514  *
515  * If the time stamp is valid, convert it into the timecounter ns value
516  * and store that result into the shhwtstamps structure which is passed
517  * up the network stack.
518  **/
519 static void e1000e_rx_hwtstamp(struct e1000_adapter *adapter, u32 status,
520 			       struct sk_buff *skb)
521 {
522 	struct e1000_hw *hw = &adapter->hw;
523 	u64 rxstmp;
524 
525 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP) ||
526 	    !(status & E1000_RXDEXT_STATERR_TST) ||
527 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID))
528 		return;
529 
530 	/* The Rx time stamp registers contain the time stamp.  No other
531 	 * received packet will be time stamped until the Rx time stamp
532 	 * registers are read.  Because only one packet can be time stamped
533 	 * at a time, the register values must belong to this packet and
534 	 * therefore none of the other additional attributes need to be
535 	 * compared.
536 	 */
537 	rxstmp = (u64)er32(RXSTMPL);
538 	rxstmp |= (u64)er32(RXSTMPH) << 32;
539 	e1000e_systim_to_hwtstamp(adapter, skb_hwtstamps(skb), rxstmp);
540 
541 	adapter->flags2 &= ~FLAG2_CHECK_RX_HWTSTAMP;
542 }
543 
544 /**
545  * e1000_receive_skb - helper function to handle Rx indications
546  * @adapter: board private structure
547  * @netdev: pointer to netdev struct
548  * @staterr: descriptor extended error and status field as written by hardware
549  * @vlan: descriptor vlan field as written by hardware (no le/be conversion)
550  * @skb: pointer to sk_buff to be indicated to stack
551  **/
552 static void e1000_receive_skb(struct e1000_adapter *adapter,
553 			      struct net_device *netdev, struct sk_buff *skb,
554 			      u32 staterr, __le16 vlan)
555 {
556 	u16 tag = le16_to_cpu(vlan);
557 
558 	e1000e_rx_hwtstamp(adapter, staterr, skb);
559 
560 	skb->protocol = eth_type_trans(skb, netdev);
561 
562 	if (staterr & E1000_RXD_STAT_VP)
563 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), tag);
564 
565 	napi_gro_receive(&adapter->napi, skb);
566 }
567 
568 /**
569  * e1000_rx_checksum - Receive Checksum Offload
570  * @adapter: board private structure
571  * @status_err: receive descriptor status and error fields
572  * @skb: socket buffer with received data
573  **/
574 static void e1000_rx_checksum(struct e1000_adapter *adapter, u32 status_err,
575 			      struct sk_buff *skb)
576 {
577 	u16 status = (u16)status_err;
578 	u8 errors = (u8)(status_err >> 24);
579 
580 	skb_checksum_none_assert(skb);
581 
582 	/* Rx checksum disabled */
583 	if (!(adapter->netdev->features & NETIF_F_RXCSUM))
584 		return;
585 
586 	/* Ignore Checksum bit is set */
587 	if (status & E1000_RXD_STAT_IXSM)
588 		return;
589 
590 	/* TCP/UDP checksum error bit or IP checksum error bit is set */
591 	if (errors & (E1000_RXD_ERR_TCPE | E1000_RXD_ERR_IPE)) {
592 		/* let the stack verify checksum errors */
593 		adapter->hw_csum_err++;
594 		return;
595 	}
596 
597 	/* TCP/UDP Checksum has not been calculated */
598 	if (!(status & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)))
599 		return;
600 
601 	/* It must be a TCP or UDP packet with a valid checksum */
602 	skb->ip_summed = CHECKSUM_UNNECESSARY;
603 	adapter->hw_csum_good++;
604 }
605 
606 static void e1000e_update_rdt_wa(struct e1000_ring *rx_ring, unsigned int i)
607 {
608 	struct e1000_adapter *adapter = rx_ring->adapter;
609 	struct e1000_hw *hw = &adapter->hw;
610 
611 	__ew32_prepare(hw);
612 	writel(i, rx_ring->tail);
613 
614 	if (unlikely(i != readl(rx_ring->tail))) {
615 		u32 rctl = er32(RCTL);
616 
617 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
618 		e_err("ME firmware caused invalid RDT - resetting\n");
619 		schedule_work(&adapter->reset_task);
620 	}
621 }
622 
623 static void e1000e_update_tdt_wa(struct e1000_ring *tx_ring, unsigned int i)
624 {
625 	struct e1000_adapter *adapter = tx_ring->adapter;
626 	struct e1000_hw *hw = &adapter->hw;
627 
628 	__ew32_prepare(hw);
629 	writel(i, tx_ring->tail);
630 
631 	if (unlikely(i != readl(tx_ring->tail))) {
632 		u32 tctl = er32(TCTL);
633 
634 		ew32(TCTL, tctl & ~E1000_TCTL_EN);
635 		e_err("ME firmware caused invalid TDT - resetting\n");
636 		schedule_work(&adapter->reset_task);
637 	}
638 }
639 
640 /**
641  * e1000_alloc_rx_buffers - Replace used receive buffers
642  * @rx_ring: Rx descriptor ring
643  * @cleaned_count: number to reallocate
644  * @gfp: flags for allocation
645  **/
646 static void e1000_alloc_rx_buffers(struct e1000_ring *rx_ring,
647 				   int cleaned_count, gfp_t gfp)
648 {
649 	struct e1000_adapter *adapter = rx_ring->adapter;
650 	struct net_device *netdev = adapter->netdev;
651 	struct pci_dev *pdev = adapter->pdev;
652 	union e1000_rx_desc_extended *rx_desc;
653 	struct e1000_buffer *buffer_info;
654 	struct sk_buff *skb;
655 	unsigned int i;
656 	unsigned int bufsz = adapter->rx_buffer_len;
657 
658 	i = rx_ring->next_to_use;
659 	buffer_info = &rx_ring->buffer_info[i];
660 
661 	while (cleaned_count--) {
662 		skb = buffer_info->skb;
663 		if (skb) {
664 			skb_trim(skb, 0);
665 			goto map_skb;
666 		}
667 
668 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
669 		if (!skb) {
670 			/* Better luck next round */
671 			adapter->alloc_rx_buff_failed++;
672 			break;
673 		}
674 
675 		buffer_info->skb = skb;
676 map_skb:
677 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
678 						  adapter->rx_buffer_len,
679 						  DMA_FROM_DEVICE);
680 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
681 			dev_err(&pdev->dev, "Rx DMA map failed\n");
682 			adapter->rx_dma_failed++;
683 			break;
684 		}
685 
686 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
687 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
688 
689 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
690 			/* Force memory writes to complete before letting h/w
691 			 * know there are new descriptors to fetch.  (Only
692 			 * applicable for weak-ordered memory model archs,
693 			 * such as IA-64).
694 			 */
695 			wmb();
696 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
697 				e1000e_update_rdt_wa(rx_ring, i);
698 			else
699 				writel(i, rx_ring->tail);
700 		}
701 		i++;
702 		if (i == rx_ring->count)
703 			i = 0;
704 		buffer_info = &rx_ring->buffer_info[i];
705 	}
706 
707 	rx_ring->next_to_use = i;
708 }
709 
710 /**
711  * e1000_alloc_rx_buffers_ps - Replace used receive buffers; packet split
712  * @rx_ring: Rx descriptor ring
713  * @cleaned_count: number to reallocate
714  * @gfp: flags for allocation
715  **/
716 static void e1000_alloc_rx_buffers_ps(struct e1000_ring *rx_ring,
717 				      int cleaned_count, gfp_t gfp)
718 {
719 	struct e1000_adapter *adapter = rx_ring->adapter;
720 	struct net_device *netdev = adapter->netdev;
721 	struct pci_dev *pdev = adapter->pdev;
722 	union e1000_rx_desc_packet_split *rx_desc;
723 	struct e1000_buffer *buffer_info;
724 	struct e1000_ps_page *ps_page;
725 	struct sk_buff *skb;
726 	unsigned int i, j;
727 
728 	i = rx_ring->next_to_use;
729 	buffer_info = &rx_ring->buffer_info[i];
730 
731 	while (cleaned_count--) {
732 		rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
733 
734 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
735 			ps_page = &buffer_info->ps_pages[j];
736 			if (j >= adapter->rx_ps_pages) {
737 				/* all unused desc entries get hw null ptr */
738 				rx_desc->read.buffer_addr[j + 1] =
739 				    ~cpu_to_le64(0);
740 				continue;
741 			}
742 			if (!ps_page->page) {
743 				ps_page->page = alloc_page(gfp);
744 				if (!ps_page->page) {
745 					adapter->alloc_rx_buff_failed++;
746 					goto no_buffers;
747 				}
748 				ps_page->dma = dma_map_page(&pdev->dev,
749 							    ps_page->page,
750 							    0, PAGE_SIZE,
751 							    DMA_FROM_DEVICE);
752 				if (dma_mapping_error(&pdev->dev,
753 						      ps_page->dma)) {
754 					dev_err(&adapter->pdev->dev,
755 						"Rx DMA page map failed\n");
756 					adapter->rx_dma_failed++;
757 					goto no_buffers;
758 				}
759 			}
760 			/* Refresh the desc even if buffer_addrs
761 			 * didn't change because each write-back
762 			 * erases this info.
763 			 */
764 			rx_desc->read.buffer_addr[j + 1] =
765 			    cpu_to_le64(ps_page->dma);
766 		}
767 
768 		skb = __netdev_alloc_skb_ip_align(netdev, adapter->rx_ps_bsize0,
769 						  gfp);
770 
771 		if (!skb) {
772 			adapter->alloc_rx_buff_failed++;
773 			break;
774 		}
775 
776 		buffer_info->skb = skb;
777 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
778 						  adapter->rx_ps_bsize0,
779 						  DMA_FROM_DEVICE);
780 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
781 			dev_err(&pdev->dev, "Rx DMA map failed\n");
782 			adapter->rx_dma_failed++;
783 			/* cleanup skb */
784 			dev_kfree_skb_any(skb);
785 			buffer_info->skb = NULL;
786 			break;
787 		}
788 
789 		rx_desc->read.buffer_addr[0] = cpu_to_le64(buffer_info->dma);
790 
791 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
792 			/* Force memory writes to complete before letting h/w
793 			 * know there are new descriptors to fetch.  (Only
794 			 * applicable for weak-ordered memory model archs,
795 			 * such as IA-64).
796 			 */
797 			wmb();
798 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
799 				e1000e_update_rdt_wa(rx_ring, i << 1);
800 			else
801 				writel(i << 1, rx_ring->tail);
802 		}
803 
804 		i++;
805 		if (i == rx_ring->count)
806 			i = 0;
807 		buffer_info = &rx_ring->buffer_info[i];
808 	}
809 
810 no_buffers:
811 	rx_ring->next_to_use = i;
812 }
813 
814 /**
815  * e1000_alloc_jumbo_rx_buffers - Replace used jumbo receive buffers
816  * @rx_ring: Rx descriptor ring
817  * @cleaned_count: number of buffers to allocate this pass
818  * @gfp: flags for allocation
819  **/
820 
821 static void e1000_alloc_jumbo_rx_buffers(struct e1000_ring *rx_ring,
822 					 int cleaned_count, gfp_t gfp)
823 {
824 	struct e1000_adapter *adapter = rx_ring->adapter;
825 	struct net_device *netdev = adapter->netdev;
826 	struct pci_dev *pdev = adapter->pdev;
827 	union e1000_rx_desc_extended *rx_desc;
828 	struct e1000_buffer *buffer_info;
829 	struct sk_buff *skb;
830 	unsigned int i;
831 	unsigned int bufsz = 256 - 16;	/* for skb_reserve */
832 
833 	i = rx_ring->next_to_use;
834 	buffer_info = &rx_ring->buffer_info[i];
835 
836 	while (cleaned_count--) {
837 		skb = buffer_info->skb;
838 		if (skb) {
839 			skb_trim(skb, 0);
840 			goto check_page;
841 		}
842 
843 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
844 		if (unlikely(!skb)) {
845 			/* Better luck next round */
846 			adapter->alloc_rx_buff_failed++;
847 			break;
848 		}
849 
850 		buffer_info->skb = skb;
851 check_page:
852 		/* allocate a new page if necessary */
853 		if (!buffer_info->page) {
854 			buffer_info->page = alloc_page(gfp);
855 			if (unlikely(!buffer_info->page)) {
856 				adapter->alloc_rx_buff_failed++;
857 				break;
858 			}
859 		}
860 
861 		if (!buffer_info->dma) {
862 			buffer_info->dma = dma_map_page(&pdev->dev,
863 							buffer_info->page, 0,
864 							PAGE_SIZE,
865 							DMA_FROM_DEVICE);
866 			if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
867 				adapter->alloc_rx_buff_failed++;
868 				break;
869 			}
870 		}
871 
872 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
873 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
874 
875 		if (unlikely(++i == rx_ring->count))
876 			i = 0;
877 		buffer_info = &rx_ring->buffer_info[i];
878 	}
879 
880 	if (likely(rx_ring->next_to_use != i)) {
881 		rx_ring->next_to_use = i;
882 		if (unlikely(i-- == 0))
883 			i = (rx_ring->count - 1);
884 
885 		/* Force memory writes to complete before letting h/w
886 		 * know there are new descriptors to fetch.  (Only
887 		 * applicable for weak-ordered memory model archs,
888 		 * such as IA-64).
889 		 */
890 		wmb();
891 		if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
892 			e1000e_update_rdt_wa(rx_ring, i);
893 		else
894 			writel(i, rx_ring->tail);
895 	}
896 }
897 
898 static inline void e1000_rx_hash(struct net_device *netdev, __le32 rss,
899 				 struct sk_buff *skb)
900 {
901 	if (netdev->features & NETIF_F_RXHASH)
902 		skb_set_hash(skb, le32_to_cpu(rss), PKT_HASH_TYPE_L3);
903 }
904 
905 /**
906  * e1000_clean_rx_irq - Send received data up the network stack
907  * @rx_ring: Rx descriptor ring
908  * @work_done: output parameter for indicating completed work
909  * @work_to_do: how many packets we can clean
910  *
911  * the return value indicates whether actual cleaning was done, there
912  * is no guarantee that everything was cleaned
913  **/
914 static bool e1000_clean_rx_irq(struct e1000_ring *rx_ring, int *work_done,
915 			       int work_to_do)
916 {
917 	struct e1000_adapter *adapter = rx_ring->adapter;
918 	struct net_device *netdev = adapter->netdev;
919 	struct pci_dev *pdev = adapter->pdev;
920 	struct e1000_hw *hw = &adapter->hw;
921 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
922 	struct e1000_buffer *buffer_info, *next_buffer;
923 	u32 length, staterr;
924 	unsigned int i;
925 	int cleaned_count = 0;
926 	bool cleaned = false;
927 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
928 
929 	i = rx_ring->next_to_clean;
930 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
931 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
932 	buffer_info = &rx_ring->buffer_info[i];
933 
934 	while (staterr & E1000_RXD_STAT_DD) {
935 		struct sk_buff *skb;
936 
937 		if (*work_done >= work_to_do)
938 			break;
939 		(*work_done)++;
940 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
941 
942 		skb = buffer_info->skb;
943 		buffer_info->skb = NULL;
944 
945 		prefetch(skb->data - NET_IP_ALIGN);
946 
947 		i++;
948 		if (i == rx_ring->count)
949 			i = 0;
950 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
951 		prefetch(next_rxd);
952 
953 		next_buffer = &rx_ring->buffer_info[i];
954 
955 		cleaned = true;
956 		cleaned_count++;
957 		dma_unmap_single(&pdev->dev, buffer_info->dma,
958 				 adapter->rx_buffer_len, DMA_FROM_DEVICE);
959 		buffer_info->dma = 0;
960 
961 		length = le16_to_cpu(rx_desc->wb.upper.length);
962 
963 		/* !EOP means multiple descriptors were used to store a single
964 		 * packet, if that's the case we need to toss it.  In fact, we
965 		 * need to toss every packet with the EOP bit clear and the
966 		 * next frame that _does_ have the EOP bit set, as it is by
967 		 * definition only a frame fragment
968 		 */
969 		if (unlikely(!(staterr & E1000_RXD_STAT_EOP)))
970 			adapter->flags2 |= FLAG2_IS_DISCARDING;
971 
972 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
973 			/* All receives must fit into a single buffer */
974 			e_dbg("Receive packet consumed multiple buffers\n");
975 			/* recycle */
976 			buffer_info->skb = skb;
977 			if (staterr & E1000_RXD_STAT_EOP)
978 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
979 			goto next_desc;
980 		}
981 
982 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
983 			     !(netdev->features & NETIF_F_RXALL))) {
984 			/* recycle */
985 			buffer_info->skb = skb;
986 			goto next_desc;
987 		}
988 
989 		/* adjust length to remove Ethernet CRC */
990 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
991 			/* If configured to store CRC, don't subtract FCS,
992 			 * but keep the FCS bytes out of the total_rx_bytes
993 			 * counter
994 			 */
995 			if (netdev->features & NETIF_F_RXFCS)
996 				total_rx_bytes -= 4;
997 			else
998 				length -= 4;
999 		}
1000 
1001 		total_rx_bytes += length;
1002 		total_rx_packets++;
1003 
1004 		/* code added for copybreak, this should improve
1005 		 * performance for small packets with large amounts
1006 		 * of reassembly being done in the stack
1007 		 */
1008 		if (length < copybreak) {
1009 			struct sk_buff *new_skb =
1010 				napi_alloc_skb(&adapter->napi, length);
1011 			if (new_skb) {
1012 				skb_copy_to_linear_data_offset(new_skb,
1013 							       -NET_IP_ALIGN,
1014 							       (skb->data -
1015 								NET_IP_ALIGN),
1016 							       (length +
1017 								NET_IP_ALIGN));
1018 				/* save the skb in buffer_info as good */
1019 				buffer_info->skb = skb;
1020 				skb = new_skb;
1021 			}
1022 			/* else just continue with the old one */
1023 		}
1024 		/* end copybreak code */
1025 		skb_put(skb, length);
1026 
1027 		/* Receive Checksum Offload */
1028 		e1000_rx_checksum(adapter, staterr, skb);
1029 
1030 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1031 
1032 		e1000_receive_skb(adapter, netdev, skb, staterr,
1033 				  rx_desc->wb.upper.vlan);
1034 
1035 next_desc:
1036 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1037 
1038 		/* return some buffers to hardware, one at a time is too slow */
1039 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1040 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1041 					      GFP_ATOMIC);
1042 			cleaned_count = 0;
1043 		}
1044 
1045 		/* use prefetched values */
1046 		rx_desc = next_rxd;
1047 		buffer_info = next_buffer;
1048 
1049 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1050 	}
1051 	rx_ring->next_to_clean = i;
1052 
1053 	cleaned_count = e1000_desc_unused(rx_ring);
1054 	if (cleaned_count)
1055 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1056 
1057 	adapter->total_rx_bytes += total_rx_bytes;
1058 	adapter->total_rx_packets += total_rx_packets;
1059 	return cleaned;
1060 }
1061 
1062 static void e1000_put_txbuf(struct e1000_ring *tx_ring,
1063 			    struct e1000_buffer *buffer_info,
1064 			    bool drop)
1065 {
1066 	struct e1000_adapter *adapter = tx_ring->adapter;
1067 
1068 	if (buffer_info->dma) {
1069 		if (buffer_info->mapped_as_page)
1070 			dma_unmap_page(&adapter->pdev->dev, buffer_info->dma,
1071 				       buffer_info->length, DMA_TO_DEVICE);
1072 		else
1073 			dma_unmap_single(&adapter->pdev->dev, buffer_info->dma,
1074 					 buffer_info->length, DMA_TO_DEVICE);
1075 		buffer_info->dma = 0;
1076 	}
1077 	if (buffer_info->skb) {
1078 		if (drop)
1079 			dev_kfree_skb_any(buffer_info->skb);
1080 		else
1081 			dev_consume_skb_any(buffer_info->skb);
1082 		buffer_info->skb = NULL;
1083 	}
1084 	buffer_info->time_stamp = 0;
1085 }
1086 
1087 static void e1000_print_hw_hang(struct work_struct *work)
1088 {
1089 	struct e1000_adapter *adapter = container_of(work,
1090 						     struct e1000_adapter,
1091 						     print_hang_task);
1092 	struct net_device *netdev = adapter->netdev;
1093 	struct e1000_ring *tx_ring = adapter->tx_ring;
1094 	unsigned int i = tx_ring->next_to_clean;
1095 	unsigned int eop = tx_ring->buffer_info[i].next_to_watch;
1096 	struct e1000_tx_desc *eop_desc = E1000_TX_DESC(*tx_ring, eop);
1097 	struct e1000_hw *hw = &adapter->hw;
1098 	u16 phy_status, phy_1000t_status, phy_ext_status;
1099 	u16 pci_status;
1100 
1101 	if (test_bit(__E1000_DOWN, &adapter->state))
1102 		return;
1103 
1104 	if (!adapter->tx_hang_recheck && (adapter->flags2 & FLAG2_DMA_BURST)) {
1105 		/* May be block on write-back, flush and detect again
1106 		 * flush pending descriptor writebacks to memory
1107 		 */
1108 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1109 		/* execute the writes immediately */
1110 		e1e_flush();
1111 		/* Due to rare timing issues, write to TIDV again to ensure
1112 		 * the write is successful
1113 		 */
1114 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1115 		/* execute the writes immediately */
1116 		e1e_flush();
1117 		adapter->tx_hang_recheck = true;
1118 		return;
1119 	}
1120 	adapter->tx_hang_recheck = false;
1121 
1122 	if (er32(TDH(0)) == er32(TDT(0))) {
1123 		e_dbg("false hang detected, ignoring\n");
1124 		return;
1125 	}
1126 
1127 	/* Real hang detected */
1128 	netif_stop_queue(netdev);
1129 
1130 	e1e_rphy(hw, MII_BMSR, &phy_status);
1131 	e1e_rphy(hw, MII_STAT1000, &phy_1000t_status);
1132 	e1e_rphy(hw, MII_ESTATUS, &phy_ext_status);
1133 
1134 	pci_read_config_word(adapter->pdev, PCI_STATUS, &pci_status);
1135 
1136 	/* detected Hardware unit hang */
1137 	e_err("Detected Hardware Unit Hang:\n"
1138 	      "  TDH                  <%x>\n"
1139 	      "  TDT                  <%x>\n"
1140 	      "  next_to_use          <%x>\n"
1141 	      "  next_to_clean        <%x>\n"
1142 	      "buffer_info[next_to_clean]:\n"
1143 	      "  time_stamp           <%lx>\n"
1144 	      "  next_to_watch        <%x>\n"
1145 	      "  jiffies              <%lx>\n"
1146 	      "  next_to_watch.status <%x>\n"
1147 	      "MAC Status             <%x>\n"
1148 	      "PHY Status             <%x>\n"
1149 	      "PHY 1000BASE-T Status  <%x>\n"
1150 	      "PHY Extended Status    <%x>\n"
1151 	      "PCI Status             <%x>\n",
1152 	      readl(tx_ring->head), readl(tx_ring->tail), tx_ring->next_to_use,
1153 	      tx_ring->next_to_clean, tx_ring->buffer_info[eop].time_stamp,
1154 	      eop, jiffies, eop_desc->upper.fields.status, er32(STATUS),
1155 	      phy_status, phy_1000t_status, phy_ext_status, pci_status);
1156 
1157 	e1000e_dump(adapter);
1158 
1159 	/* Suggest workaround for known h/w issue */
1160 	if ((hw->mac.type == e1000_pchlan) && (er32(CTRL) & E1000_CTRL_TFCE))
1161 		e_err("Try turning off Tx pause (flow control) via ethtool\n");
1162 }
1163 
1164 /**
1165  * e1000e_tx_hwtstamp_work - check for Tx time stamp
1166  * @work: pointer to work struct
1167  *
1168  * This work function polls the TSYNCTXCTL valid bit to determine when a
1169  * timestamp has been taken for the current stored skb.  The timestamp must
1170  * be for this skb because only one such packet is allowed in the queue.
1171  */
1172 static void e1000e_tx_hwtstamp_work(struct work_struct *work)
1173 {
1174 	struct e1000_adapter *adapter = container_of(work, struct e1000_adapter,
1175 						     tx_hwtstamp_work);
1176 	struct e1000_hw *hw = &adapter->hw;
1177 
1178 	if (er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID) {
1179 		struct sk_buff *skb = adapter->tx_hwtstamp_skb;
1180 		struct skb_shared_hwtstamps shhwtstamps;
1181 		u64 txstmp;
1182 
1183 		txstmp = er32(TXSTMPL);
1184 		txstmp |= (u64)er32(TXSTMPH) << 32;
1185 
1186 		e1000e_systim_to_hwtstamp(adapter, &shhwtstamps, txstmp);
1187 
1188 		/* Clear the global tx_hwtstamp_skb pointer and force writes
1189 		 * prior to notifying the stack of a Tx timestamp.
1190 		 */
1191 		adapter->tx_hwtstamp_skb = NULL;
1192 		wmb(); /* force write prior to skb_tstamp_tx */
1193 
1194 		skb_tstamp_tx(skb, &shhwtstamps);
1195 		dev_consume_skb_any(skb);
1196 	} else if (time_after(jiffies, adapter->tx_hwtstamp_start
1197 			      + adapter->tx_timeout_factor * HZ)) {
1198 		dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
1199 		adapter->tx_hwtstamp_skb = NULL;
1200 		adapter->tx_hwtstamp_timeouts++;
1201 		e_warn("clearing Tx timestamp hang\n");
1202 	} else {
1203 		/* reschedule to check later */
1204 		schedule_work(&adapter->tx_hwtstamp_work);
1205 	}
1206 }
1207 
1208 /**
1209  * e1000_clean_tx_irq - Reclaim resources after transmit completes
1210  * @tx_ring: Tx descriptor ring
1211  *
1212  * the return value indicates whether actual cleaning was done, there
1213  * is no guarantee that everything was cleaned
1214  **/
1215 static bool e1000_clean_tx_irq(struct e1000_ring *tx_ring)
1216 {
1217 	struct e1000_adapter *adapter = tx_ring->adapter;
1218 	struct net_device *netdev = adapter->netdev;
1219 	struct e1000_hw *hw = &adapter->hw;
1220 	struct e1000_tx_desc *tx_desc, *eop_desc;
1221 	struct e1000_buffer *buffer_info;
1222 	unsigned int i, eop;
1223 	unsigned int count = 0;
1224 	unsigned int total_tx_bytes = 0, total_tx_packets = 0;
1225 	unsigned int bytes_compl = 0, pkts_compl = 0;
1226 
1227 	i = tx_ring->next_to_clean;
1228 	eop = tx_ring->buffer_info[i].next_to_watch;
1229 	eop_desc = E1000_TX_DESC(*tx_ring, eop);
1230 
1231 	while ((eop_desc->upper.data & cpu_to_le32(E1000_TXD_STAT_DD)) &&
1232 	       (count < tx_ring->count)) {
1233 		bool cleaned = false;
1234 
1235 		dma_rmb();		/* read buffer_info after eop_desc */
1236 		for (; !cleaned; count++) {
1237 			tx_desc = E1000_TX_DESC(*tx_ring, i);
1238 			buffer_info = &tx_ring->buffer_info[i];
1239 			cleaned = (i == eop);
1240 
1241 			if (cleaned) {
1242 				total_tx_packets += buffer_info->segs;
1243 				total_tx_bytes += buffer_info->bytecount;
1244 				if (buffer_info->skb) {
1245 					bytes_compl += buffer_info->skb->len;
1246 					pkts_compl++;
1247 				}
1248 			}
1249 
1250 			e1000_put_txbuf(tx_ring, buffer_info, false);
1251 			tx_desc->upper.data = 0;
1252 
1253 			i++;
1254 			if (i == tx_ring->count)
1255 				i = 0;
1256 		}
1257 
1258 		if (i == tx_ring->next_to_use)
1259 			break;
1260 		eop = tx_ring->buffer_info[i].next_to_watch;
1261 		eop_desc = E1000_TX_DESC(*tx_ring, eop);
1262 	}
1263 
1264 	tx_ring->next_to_clean = i;
1265 
1266 	netdev_completed_queue(netdev, pkts_compl, bytes_compl);
1267 
1268 #define TX_WAKE_THRESHOLD 32
1269 	if (count && netif_carrier_ok(netdev) &&
1270 	    e1000_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD) {
1271 		/* Make sure that anybody stopping the queue after this
1272 		 * sees the new next_to_clean.
1273 		 */
1274 		smp_mb();
1275 
1276 		if (netif_queue_stopped(netdev) &&
1277 		    !(test_bit(__E1000_DOWN, &adapter->state))) {
1278 			netif_wake_queue(netdev);
1279 			++adapter->restart_queue;
1280 		}
1281 	}
1282 
1283 	if (adapter->detect_tx_hung) {
1284 		/* Detect a transmit hang in hardware, this serializes the
1285 		 * check with the clearing of time_stamp and movement of i
1286 		 */
1287 		adapter->detect_tx_hung = false;
1288 		if (tx_ring->buffer_info[i].time_stamp &&
1289 		    time_after(jiffies, tx_ring->buffer_info[i].time_stamp
1290 			       + (adapter->tx_timeout_factor * HZ)) &&
1291 		    !(er32(STATUS) & E1000_STATUS_TXOFF))
1292 			schedule_work(&adapter->print_hang_task);
1293 		else
1294 			adapter->tx_hang_recheck = false;
1295 	}
1296 	adapter->total_tx_bytes += total_tx_bytes;
1297 	adapter->total_tx_packets += total_tx_packets;
1298 	return count < tx_ring->count;
1299 }
1300 
1301 /**
1302  * e1000_clean_rx_irq_ps - Send received data up the network stack; packet split
1303  * @rx_ring: Rx descriptor ring
1304  * @work_done: output parameter for indicating completed work
1305  * @work_to_do: how many packets we can clean
1306  *
1307  * the return value indicates whether actual cleaning was done, there
1308  * is no guarantee that everything was cleaned
1309  **/
1310 static bool e1000_clean_rx_irq_ps(struct e1000_ring *rx_ring, int *work_done,
1311 				  int work_to_do)
1312 {
1313 	struct e1000_adapter *adapter = rx_ring->adapter;
1314 	struct e1000_hw *hw = &adapter->hw;
1315 	union e1000_rx_desc_packet_split *rx_desc, *next_rxd;
1316 	struct net_device *netdev = adapter->netdev;
1317 	struct pci_dev *pdev = adapter->pdev;
1318 	struct e1000_buffer *buffer_info, *next_buffer;
1319 	struct e1000_ps_page *ps_page;
1320 	struct sk_buff *skb;
1321 	unsigned int i, j;
1322 	u32 length, staterr;
1323 	int cleaned_count = 0;
1324 	bool cleaned = false;
1325 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1326 
1327 	i = rx_ring->next_to_clean;
1328 	rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
1329 	staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1330 	buffer_info = &rx_ring->buffer_info[i];
1331 
1332 	while (staterr & E1000_RXD_STAT_DD) {
1333 		if (*work_done >= work_to_do)
1334 			break;
1335 		(*work_done)++;
1336 		skb = buffer_info->skb;
1337 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1338 
1339 		/* in the packet split case this is header only */
1340 		prefetch(skb->data - NET_IP_ALIGN);
1341 
1342 		i++;
1343 		if (i == rx_ring->count)
1344 			i = 0;
1345 		next_rxd = E1000_RX_DESC_PS(*rx_ring, i);
1346 		prefetch(next_rxd);
1347 
1348 		next_buffer = &rx_ring->buffer_info[i];
1349 
1350 		cleaned = true;
1351 		cleaned_count++;
1352 		dma_unmap_single(&pdev->dev, buffer_info->dma,
1353 				 adapter->rx_ps_bsize0, DMA_FROM_DEVICE);
1354 		buffer_info->dma = 0;
1355 
1356 		/* see !EOP comment in other Rx routine */
1357 		if (!(staterr & E1000_RXD_STAT_EOP))
1358 			adapter->flags2 |= FLAG2_IS_DISCARDING;
1359 
1360 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
1361 			e_dbg("Packet Split buffers didn't pick up the full packet\n");
1362 			dev_kfree_skb_irq(skb);
1363 			if (staterr & E1000_RXD_STAT_EOP)
1364 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1365 			goto next_desc;
1366 		}
1367 
1368 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1369 			     !(netdev->features & NETIF_F_RXALL))) {
1370 			dev_kfree_skb_irq(skb);
1371 			goto next_desc;
1372 		}
1373 
1374 		length = le16_to_cpu(rx_desc->wb.middle.length0);
1375 
1376 		if (!length) {
1377 			e_dbg("Last part of the packet spanning multiple descriptors\n");
1378 			dev_kfree_skb_irq(skb);
1379 			goto next_desc;
1380 		}
1381 
1382 		/* Good Receive */
1383 		skb_put(skb, length);
1384 
1385 		{
1386 			/* this looks ugly, but it seems compiler issues make
1387 			 * it more efficient than reusing j
1388 			 */
1389 			int l1 = le16_to_cpu(rx_desc->wb.upper.length[0]);
1390 
1391 			/* page alloc/put takes too long and effects small
1392 			 * packet throughput, so unsplit small packets and
1393 			 * save the alloc/put
1394 			 */
1395 			if (l1 && (l1 <= copybreak) &&
1396 			    ((length + l1) <= adapter->rx_ps_bsize0)) {
1397 				ps_page = &buffer_info->ps_pages[0];
1398 
1399 				dma_sync_single_for_cpu(&pdev->dev,
1400 							ps_page->dma,
1401 							PAGE_SIZE,
1402 							DMA_FROM_DEVICE);
1403 				memcpy(skb_tail_pointer(skb),
1404 				       page_address(ps_page->page), l1);
1405 				dma_sync_single_for_device(&pdev->dev,
1406 							   ps_page->dma,
1407 							   PAGE_SIZE,
1408 							   DMA_FROM_DEVICE);
1409 
1410 				/* remove the CRC */
1411 				if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1412 					if (!(netdev->features & NETIF_F_RXFCS))
1413 						l1 -= 4;
1414 				}
1415 
1416 				skb_put(skb, l1);
1417 				goto copydone;
1418 			}	/* if */
1419 		}
1420 
1421 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1422 			length = le16_to_cpu(rx_desc->wb.upper.length[j]);
1423 			if (!length)
1424 				break;
1425 
1426 			ps_page = &buffer_info->ps_pages[j];
1427 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1428 				       DMA_FROM_DEVICE);
1429 			ps_page->dma = 0;
1430 			skb_fill_page_desc(skb, j, ps_page->page, 0, length);
1431 			ps_page->page = NULL;
1432 			skb->len += length;
1433 			skb->data_len += length;
1434 			skb->truesize += PAGE_SIZE;
1435 		}
1436 
1437 		/* strip the ethernet crc, problem is we're using pages now so
1438 		 * this whole operation can get a little cpu intensive
1439 		 */
1440 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1441 			if (!(netdev->features & NETIF_F_RXFCS))
1442 				pskb_trim(skb, skb->len - 4);
1443 		}
1444 
1445 copydone:
1446 		total_rx_bytes += skb->len;
1447 		total_rx_packets++;
1448 
1449 		e1000_rx_checksum(adapter, staterr, skb);
1450 
1451 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1452 
1453 		if (rx_desc->wb.upper.header_status &
1454 		    cpu_to_le16(E1000_RXDPS_HDRSTAT_HDRSP))
1455 			adapter->rx_hdr_split++;
1456 
1457 		e1000_receive_skb(adapter, netdev, skb, staterr,
1458 				  rx_desc->wb.middle.vlan);
1459 
1460 next_desc:
1461 		rx_desc->wb.middle.status_error &= cpu_to_le32(~0xFF);
1462 		buffer_info->skb = NULL;
1463 
1464 		/* return some buffers to hardware, one at a time is too slow */
1465 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1466 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1467 					      GFP_ATOMIC);
1468 			cleaned_count = 0;
1469 		}
1470 
1471 		/* use prefetched values */
1472 		rx_desc = next_rxd;
1473 		buffer_info = next_buffer;
1474 
1475 		staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1476 	}
1477 	rx_ring->next_to_clean = i;
1478 
1479 	cleaned_count = e1000_desc_unused(rx_ring);
1480 	if (cleaned_count)
1481 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1482 
1483 	adapter->total_rx_bytes += total_rx_bytes;
1484 	adapter->total_rx_packets += total_rx_packets;
1485 	return cleaned;
1486 }
1487 
1488 static void e1000_consume_page(struct e1000_buffer *bi, struct sk_buff *skb,
1489 			       u16 length)
1490 {
1491 	bi->page = NULL;
1492 	skb->len += length;
1493 	skb->data_len += length;
1494 	skb->truesize += PAGE_SIZE;
1495 }
1496 
1497 /**
1498  * e1000_clean_jumbo_rx_irq - Send received data up the network stack; legacy
1499  * @rx_ring: Rx descriptor ring
1500  * @work_done: output parameter for indicating completed work
1501  * @work_to_do: how many packets we can clean
1502  *
1503  * the return value indicates whether actual cleaning was done, there
1504  * is no guarantee that everything was cleaned
1505  **/
1506 static bool e1000_clean_jumbo_rx_irq(struct e1000_ring *rx_ring, int *work_done,
1507 				     int work_to_do)
1508 {
1509 	struct e1000_adapter *adapter = rx_ring->adapter;
1510 	struct net_device *netdev = adapter->netdev;
1511 	struct pci_dev *pdev = adapter->pdev;
1512 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
1513 	struct e1000_buffer *buffer_info, *next_buffer;
1514 	u32 length, staterr;
1515 	unsigned int i;
1516 	int cleaned_count = 0;
1517 	bool cleaned = false;
1518 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1519 	struct skb_shared_info *shinfo;
1520 
1521 	i = rx_ring->next_to_clean;
1522 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
1523 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1524 	buffer_info = &rx_ring->buffer_info[i];
1525 
1526 	while (staterr & E1000_RXD_STAT_DD) {
1527 		struct sk_buff *skb;
1528 
1529 		if (*work_done >= work_to_do)
1530 			break;
1531 		(*work_done)++;
1532 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1533 
1534 		skb = buffer_info->skb;
1535 		buffer_info->skb = NULL;
1536 
1537 		++i;
1538 		if (i == rx_ring->count)
1539 			i = 0;
1540 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
1541 		prefetch(next_rxd);
1542 
1543 		next_buffer = &rx_ring->buffer_info[i];
1544 
1545 		cleaned = true;
1546 		cleaned_count++;
1547 		dma_unmap_page(&pdev->dev, buffer_info->dma, PAGE_SIZE,
1548 			       DMA_FROM_DEVICE);
1549 		buffer_info->dma = 0;
1550 
1551 		length = le16_to_cpu(rx_desc->wb.upper.length);
1552 
1553 		/* errors is only valid for DD + EOP descriptors */
1554 		if (unlikely((staterr & E1000_RXD_STAT_EOP) &&
1555 			     ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1556 			      !(netdev->features & NETIF_F_RXALL)))) {
1557 			/* recycle both page and skb */
1558 			buffer_info->skb = skb;
1559 			/* an error means any chain goes out the window too */
1560 			if (rx_ring->rx_skb_top)
1561 				dev_kfree_skb_irq(rx_ring->rx_skb_top);
1562 			rx_ring->rx_skb_top = NULL;
1563 			goto next_desc;
1564 		}
1565 #define rxtop (rx_ring->rx_skb_top)
1566 		if (!(staterr & E1000_RXD_STAT_EOP)) {
1567 			/* this descriptor is only the beginning (or middle) */
1568 			if (!rxtop) {
1569 				/* this is the beginning of a chain */
1570 				rxtop = skb;
1571 				skb_fill_page_desc(rxtop, 0, buffer_info->page,
1572 						   0, length);
1573 			} else {
1574 				/* this is the middle of a chain */
1575 				shinfo = skb_shinfo(rxtop);
1576 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1577 						   buffer_info->page, 0,
1578 						   length);
1579 				/* re-use the skb, only consumed the page */
1580 				buffer_info->skb = skb;
1581 			}
1582 			e1000_consume_page(buffer_info, rxtop, length);
1583 			goto next_desc;
1584 		} else {
1585 			if (rxtop) {
1586 				/* end of the chain */
1587 				shinfo = skb_shinfo(rxtop);
1588 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1589 						   buffer_info->page, 0,
1590 						   length);
1591 				/* re-use the current skb, we only consumed the
1592 				 * page
1593 				 */
1594 				buffer_info->skb = skb;
1595 				skb = rxtop;
1596 				rxtop = NULL;
1597 				e1000_consume_page(buffer_info, skb, length);
1598 			} else {
1599 				/* no chain, got EOP, this buf is the packet
1600 				 * copybreak to save the put_page/alloc_page
1601 				 */
1602 				if (length <= copybreak &&
1603 				    skb_tailroom(skb) >= length) {
1604 					memcpy(skb_tail_pointer(skb),
1605 					       page_address(buffer_info->page),
1606 					       length);
1607 					/* re-use the page, so don't erase
1608 					 * buffer_info->page
1609 					 */
1610 					skb_put(skb, length);
1611 				} else {
1612 					skb_fill_page_desc(skb, 0,
1613 							   buffer_info->page, 0,
1614 							   length);
1615 					e1000_consume_page(buffer_info, skb,
1616 							   length);
1617 				}
1618 			}
1619 		}
1620 
1621 		/* Receive Checksum Offload */
1622 		e1000_rx_checksum(adapter, staterr, skb);
1623 
1624 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1625 
1626 		/* probably a little skewed due to removing CRC */
1627 		total_rx_bytes += skb->len;
1628 		total_rx_packets++;
1629 
1630 		/* eth type trans needs skb->data to point to something */
1631 		if (!pskb_may_pull(skb, ETH_HLEN)) {
1632 			e_err("pskb_may_pull failed.\n");
1633 			dev_kfree_skb_irq(skb);
1634 			goto next_desc;
1635 		}
1636 
1637 		e1000_receive_skb(adapter, netdev, skb, staterr,
1638 				  rx_desc->wb.upper.vlan);
1639 
1640 next_desc:
1641 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1642 
1643 		/* return some buffers to hardware, one at a time is too slow */
1644 		if (unlikely(cleaned_count >= E1000_RX_BUFFER_WRITE)) {
1645 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1646 					      GFP_ATOMIC);
1647 			cleaned_count = 0;
1648 		}
1649 
1650 		/* use prefetched values */
1651 		rx_desc = next_rxd;
1652 		buffer_info = next_buffer;
1653 
1654 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1655 	}
1656 	rx_ring->next_to_clean = i;
1657 
1658 	cleaned_count = e1000_desc_unused(rx_ring);
1659 	if (cleaned_count)
1660 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1661 
1662 	adapter->total_rx_bytes += total_rx_bytes;
1663 	adapter->total_rx_packets += total_rx_packets;
1664 	return cleaned;
1665 }
1666 
1667 /**
1668  * e1000_clean_rx_ring - Free Rx Buffers per Queue
1669  * @rx_ring: Rx descriptor ring
1670  **/
1671 static void e1000_clean_rx_ring(struct e1000_ring *rx_ring)
1672 {
1673 	struct e1000_adapter *adapter = rx_ring->adapter;
1674 	struct e1000_buffer *buffer_info;
1675 	struct e1000_ps_page *ps_page;
1676 	struct pci_dev *pdev = adapter->pdev;
1677 	unsigned int i, j;
1678 
1679 	/* Free all the Rx ring sk_buffs */
1680 	for (i = 0; i < rx_ring->count; i++) {
1681 		buffer_info = &rx_ring->buffer_info[i];
1682 		if (buffer_info->dma) {
1683 			if (adapter->clean_rx == e1000_clean_rx_irq)
1684 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1685 						 adapter->rx_buffer_len,
1686 						 DMA_FROM_DEVICE);
1687 			else if (adapter->clean_rx == e1000_clean_jumbo_rx_irq)
1688 				dma_unmap_page(&pdev->dev, buffer_info->dma,
1689 					       PAGE_SIZE, DMA_FROM_DEVICE);
1690 			else if (adapter->clean_rx == e1000_clean_rx_irq_ps)
1691 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1692 						 adapter->rx_ps_bsize0,
1693 						 DMA_FROM_DEVICE);
1694 			buffer_info->dma = 0;
1695 		}
1696 
1697 		if (buffer_info->page) {
1698 			put_page(buffer_info->page);
1699 			buffer_info->page = NULL;
1700 		}
1701 
1702 		if (buffer_info->skb) {
1703 			dev_kfree_skb(buffer_info->skb);
1704 			buffer_info->skb = NULL;
1705 		}
1706 
1707 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1708 			ps_page = &buffer_info->ps_pages[j];
1709 			if (!ps_page->page)
1710 				break;
1711 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1712 				       DMA_FROM_DEVICE);
1713 			ps_page->dma = 0;
1714 			put_page(ps_page->page);
1715 			ps_page->page = NULL;
1716 		}
1717 	}
1718 
1719 	/* there also may be some cached data from a chained receive */
1720 	if (rx_ring->rx_skb_top) {
1721 		dev_kfree_skb(rx_ring->rx_skb_top);
1722 		rx_ring->rx_skb_top = NULL;
1723 	}
1724 
1725 	/* Zero out the descriptor ring */
1726 	memset(rx_ring->desc, 0, rx_ring->size);
1727 
1728 	rx_ring->next_to_clean = 0;
1729 	rx_ring->next_to_use = 0;
1730 	adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1731 }
1732 
1733 static void e1000e_downshift_workaround(struct work_struct *work)
1734 {
1735 	struct e1000_adapter *adapter = container_of(work,
1736 						     struct e1000_adapter,
1737 						     downshift_task);
1738 
1739 	if (test_bit(__E1000_DOWN, &adapter->state))
1740 		return;
1741 
1742 	e1000e_gig_downshift_workaround_ich8lan(&adapter->hw);
1743 }
1744 
1745 /**
1746  * e1000_intr_msi - Interrupt Handler
1747  * @irq: interrupt number
1748  * @data: pointer to a network interface device structure
1749  **/
1750 static irqreturn_t e1000_intr_msi(int __always_unused irq, void *data)
1751 {
1752 	struct net_device *netdev = data;
1753 	struct e1000_adapter *adapter = netdev_priv(netdev);
1754 	struct e1000_hw *hw = &adapter->hw;
1755 	u32 icr = er32(ICR);
1756 
1757 	/* read ICR disables interrupts using IAM */
1758 	if (icr & E1000_ICR_LSC) {
1759 		hw->mac.get_link_status = true;
1760 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1761 		 * disconnect (LSC) before accessing any PHY registers
1762 		 */
1763 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1764 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1765 			schedule_work(&adapter->downshift_task);
1766 
1767 		/* 80003ES2LAN workaround-- For packet buffer work-around on
1768 		 * link down event; disable receives here in the ISR and reset
1769 		 * adapter in watchdog
1770 		 */
1771 		if (netif_carrier_ok(netdev) &&
1772 		    adapter->flags & FLAG_RX_NEEDS_RESTART) {
1773 			/* disable receives */
1774 			u32 rctl = er32(RCTL);
1775 
1776 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1777 			adapter->flags |= FLAG_RESTART_NOW;
1778 		}
1779 		/* guard against interrupt when we're going down */
1780 		if (!test_bit(__E1000_DOWN, &adapter->state))
1781 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1782 	}
1783 
1784 	/* Reset on uncorrectable ECC error */
1785 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1786 		u32 pbeccsts = er32(PBECCSTS);
1787 
1788 		adapter->corr_errors +=
1789 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1790 		adapter->uncorr_errors +=
1791 		    FIELD_GET(E1000_PBECCSTS_UNCORR_ERR_CNT_MASK, pbeccsts);
1792 
1793 		/* Do the reset outside of interrupt context */
1794 		schedule_work(&adapter->reset_task);
1795 
1796 		/* return immediately since reset is imminent */
1797 		return IRQ_HANDLED;
1798 	}
1799 
1800 	if (napi_schedule_prep(&adapter->napi)) {
1801 		adapter->total_tx_bytes = 0;
1802 		adapter->total_tx_packets = 0;
1803 		adapter->total_rx_bytes = 0;
1804 		adapter->total_rx_packets = 0;
1805 		__napi_schedule(&adapter->napi);
1806 	}
1807 
1808 	return IRQ_HANDLED;
1809 }
1810 
1811 /**
1812  * e1000_intr - Interrupt Handler
1813  * @irq: interrupt number
1814  * @data: pointer to a network interface device structure
1815  **/
1816 static irqreturn_t e1000_intr(int __always_unused irq, void *data)
1817 {
1818 	struct net_device *netdev = data;
1819 	struct e1000_adapter *adapter = netdev_priv(netdev);
1820 	struct e1000_hw *hw = &adapter->hw;
1821 	u32 rctl, icr = er32(ICR);
1822 
1823 	if (!icr || test_bit(__E1000_DOWN, &adapter->state))
1824 		return IRQ_NONE;	/* Not our interrupt */
1825 
1826 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
1827 	 * not set, then the adapter didn't send an interrupt
1828 	 */
1829 	if (!(icr & E1000_ICR_INT_ASSERTED))
1830 		return IRQ_NONE;
1831 
1832 	/* Interrupt Auto-Mask...upon reading ICR,
1833 	 * interrupts are masked.  No need for the
1834 	 * IMC write
1835 	 */
1836 
1837 	if (icr & E1000_ICR_LSC) {
1838 		hw->mac.get_link_status = true;
1839 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1840 		 * disconnect (LSC) before accessing any PHY registers
1841 		 */
1842 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1843 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1844 			schedule_work(&adapter->downshift_task);
1845 
1846 		/* 80003ES2LAN workaround--
1847 		 * For packet buffer work-around on link down event;
1848 		 * disable receives here in the ISR and
1849 		 * reset adapter in watchdog
1850 		 */
1851 		if (netif_carrier_ok(netdev) &&
1852 		    (adapter->flags & FLAG_RX_NEEDS_RESTART)) {
1853 			/* disable receives */
1854 			rctl = er32(RCTL);
1855 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1856 			adapter->flags |= FLAG_RESTART_NOW;
1857 		}
1858 		/* guard against interrupt when we're going down */
1859 		if (!test_bit(__E1000_DOWN, &adapter->state))
1860 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1861 	}
1862 
1863 	/* Reset on uncorrectable ECC error */
1864 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1865 		u32 pbeccsts = er32(PBECCSTS);
1866 
1867 		adapter->corr_errors +=
1868 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1869 		adapter->uncorr_errors +=
1870 		    FIELD_GET(E1000_PBECCSTS_UNCORR_ERR_CNT_MASK, pbeccsts);
1871 
1872 		/* Do the reset outside of interrupt context */
1873 		schedule_work(&adapter->reset_task);
1874 
1875 		/* return immediately since reset is imminent */
1876 		return IRQ_HANDLED;
1877 	}
1878 
1879 	if (napi_schedule_prep(&adapter->napi)) {
1880 		adapter->total_tx_bytes = 0;
1881 		adapter->total_tx_packets = 0;
1882 		adapter->total_rx_bytes = 0;
1883 		adapter->total_rx_packets = 0;
1884 		__napi_schedule(&adapter->napi);
1885 	}
1886 
1887 	return IRQ_HANDLED;
1888 }
1889 
1890 static irqreturn_t e1000_msix_other(int __always_unused irq, void *data)
1891 {
1892 	struct net_device *netdev = data;
1893 	struct e1000_adapter *adapter = netdev_priv(netdev);
1894 	struct e1000_hw *hw = &adapter->hw;
1895 	u32 icr = er32(ICR);
1896 
1897 	if (icr & adapter->eiac_mask)
1898 		ew32(ICS, (icr & adapter->eiac_mask));
1899 
1900 	if (icr & E1000_ICR_LSC) {
1901 		hw->mac.get_link_status = true;
1902 		/* guard against interrupt when we're going down */
1903 		if (!test_bit(__E1000_DOWN, &adapter->state))
1904 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1905 	}
1906 
1907 	if (!test_bit(__E1000_DOWN, &adapter->state))
1908 		ew32(IMS, E1000_IMS_OTHER | IMS_OTHER_MASK);
1909 
1910 	return IRQ_HANDLED;
1911 }
1912 
1913 static irqreturn_t e1000_intr_msix_tx(int __always_unused irq, void *data)
1914 {
1915 	struct net_device *netdev = data;
1916 	struct e1000_adapter *adapter = netdev_priv(netdev);
1917 	struct e1000_hw *hw = &adapter->hw;
1918 	struct e1000_ring *tx_ring = adapter->tx_ring;
1919 
1920 	adapter->total_tx_bytes = 0;
1921 	adapter->total_tx_packets = 0;
1922 
1923 	if (!e1000_clean_tx_irq(tx_ring))
1924 		/* Ring was not completely cleaned, so fire another interrupt */
1925 		ew32(ICS, tx_ring->ims_val);
1926 
1927 	if (!test_bit(__E1000_DOWN, &adapter->state))
1928 		ew32(IMS, adapter->tx_ring->ims_val);
1929 
1930 	return IRQ_HANDLED;
1931 }
1932 
1933 static irqreturn_t e1000_intr_msix_rx(int __always_unused irq, void *data)
1934 {
1935 	struct net_device *netdev = data;
1936 	struct e1000_adapter *adapter = netdev_priv(netdev);
1937 	struct e1000_ring *rx_ring = adapter->rx_ring;
1938 
1939 	/* Write the ITR value calculated at the end of the
1940 	 * previous interrupt.
1941 	 */
1942 	if (rx_ring->set_itr) {
1943 		u32 itr = rx_ring->itr_val ?
1944 			  1000000000 / (rx_ring->itr_val * 256) : 0;
1945 
1946 		writel(itr, rx_ring->itr_register);
1947 		rx_ring->set_itr = 0;
1948 	}
1949 
1950 	if (napi_schedule_prep(&adapter->napi)) {
1951 		adapter->total_rx_bytes = 0;
1952 		adapter->total_rx_packets = 0;
1953 		__napi_schedule(&adapter->napi);
1954 	}
1955 	return IRQ_HANDLED;
1956 }
1957 
1958 /**
1959  * e1000_configure_msix - Configure MSI-X hardware
1960  * @adapter: board private structure
1961  *
1962  * e1000_configure_msix sets up the hardware to properly
1963  * generate MSI-X interrupts.
1964  **/
1965 static void e1000_configure_msix(struct e1000_adapter *adapter)
1966 {
1967 	struct e1000_hw *hw = &adapter->hw;
1968 	struct e1000_ring *rx_ring = adapter->rx_ring;
1969 	struct e1000_ring *tx_ring = adapter->tx_ring;
1970 	int vector = 0;
1971 	u32 ctrl_ext, ivar = 0;
1972 
1973 	adapter->eiac_mask = 0;
1974 
1975 	/* Workaround issue with spurious interrupts on 82574 in MSI-X mode */
1976 	if (hw->mac.type == e1000_82574) {
1977 		u32 rfctl = er32(RFCTL);
1978 
1979 		rfctl |= E1000_RFCTL_ACK_DIS;
1980 		ew32(RFCTL, rfctl);
1981 	}
1982 
1983 	/* Configure Rx vector */
1984 	rx_ring->ims_val = E1000_IMS_RXQ0;
1985 	adapter->eiac_mask |= rx_ring->ims_val;
1986 	if (rx_ring->itr_val)
1987 		writel(1000000000 / (rx_ring->itr_val * 256),
1988 		       rx_ring->itr_register);
1989 	else
1990 		writel(1, rx_ring->itr_register);
1991 	ivar = E1000_IVAR_INT_ALLOC_VALID | vector;
1992 
1993 	/* Configure Tx vector */
1994 	tx_ring->ims_val = E1000_IMS_TXQ0;
1995 	vector++;
1996 	if (tx_ring->itr_val)
1997 		writel(1000000000 / (tx_ring->itr_val * 256),
1998 		       tx_ring->itr_register);
1999 	else
2000 		writel(1, tx_ring->itr_register);
2001 	adapter->eiac_mask |= tx_ring->ims_val;
2002 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 8);
2003 
2004 	/* set vector for Other Causes, e.g. link changes */
2005 	vector++;
2006 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 16);
2007 	if (rx_ring->itr_val)
2008 		writel(1000000000 / (rx_ring->itr_val * 256),
2009 		       hw->hw_addr + E1000_EITR_82574(vector));
2010 	else
2011 		writel(1, hw->hw_addr + E1000_EITR_82574(vector));
2012 
2013 	/* Cause Tx interrupts on every write back */
2014 	ivar |= BIT(31);
2015 
2016 	ew32(IVAR, ivar);
2017 
2018 	/* enable MSI-X PBA support */
2019 	ctrl_ext = er32(CTRL_EXT) & ~E1000_CTRL_EXT_IAME;
2020 	ctrl_ext |= E1000_CTRL_EXT_PBA_CLR | E1000_CTRL_EXT_EIAME;
2021 	ew32(CTRL_EXT, ctrl_ext);
2022 	e1e_flush();
2023 }
2024 
2025 void e1000e_reset_interrupt_capability(struct e1000_adapter *adapter)
2026 {
2027 	if (adapter->msix_entries) {
2028 		pci_disable_msix(adapter->pdev);
2029 		kfree(adapter->msix_entries);
2030 		adapter->msix_entries = NULL;
2031 	} else if (adapter->flags & FLAG_MSI_ENABLED) {
2032 		pci_disable_msi(adapter->pdev);
2033 		adapter->flags &= ~FLAG_MSI_ENABLED;
2034 	}
2035 }
2036 
2037 /**
2038  * e1000e_set_interrupt_capability - set MSI or MSI-X if supported
2039  * @adapter: board private structure
2040  *
2041  * Attempt to configure interrupts using the best available
2042  * capabilities of the hardware and kernel.
2043  **/
2044 void e1000e_set_interrupt_capability(struct e1000_adapter *adapter)
2045 {
2046 	int err;
2047 	int i;
2048 
2049 	switch (adapter->int_mode) {
2050 	case E1000E_INT_MODE_MSIX:
2051 		if (adapter->flags & FLAG_HAS_MSIX) {
2052 			adapter->num_vectors = 3; /* RxQ0, TxQ0 and other */
2053 			adapter->msix_entries = kzalloc_objs(struct msix_entry,
2054 							     adapter->num_vectors,
2055 							     GFP_KERNEL);
2056 			if (adapter->msix_entries) {
2057 				struct e1000_adapter *a = adapter;
2058 
2059 				for (i = 0; i < adapter->num_vectors; i++)
2060 					adapter->msix_entries[i].entry = i;
2061 
2062 				err = pci_enable_msix_range(a->pdev,
2063 							    a->msix_entries,
2064 							    a->num_vectors,
2065 							    a->num_vectors);
2066 				if (err > 0)
2067 					return;
2068 			}
2069 			/* MSI-X failed, so fall through and try MSI */
2070 			e_err("Failed to initialize MSI-X interrupts.  Falling back to MSI interrupts.\n");
2071 			e1000e_reset_interrupt_capability(adapter);
2072 		}
2073 		adapter->int_mode = E1000E_INT_MODE_MSI;
2074 		fallthrough;
2075 	case E1000E_INT_MODE_MSI:
2076 		if (!pci_enable_msi(adapter->pdev)) {
2077 			adapter->flags |= FLAG_MSI_ENABLED;
2078 		} else {
2079 			adapter->int_mode = E1000E_INT_MODE_LEGACY;
2080 			e_err("Failed to initialize MSI interrupts.  Falling back to legacy interrupts.\n");
2081 		}
2082 		fallthrough;
2083 	case E1000E_INT_MODE_LEGACY:
2084 		/* Don't do anything; this is the system default */
2085 		break;
2086 	}
2087 
2088 	/* store the number of vectors being used */
2089 	adapter->num_vectors = 1;
2090 }
2091 
2092 /**
2093  * e1000_request_msix - Initialize MSI-X interrupts
2094  * @adapter: board private structure
2095  *
2096  * e1000_request_msix allocates MSI-X vectors and requests interrupts from the
2097  * kernel.
2098  **/
2099 static int e1000_request_msix(struct e1000_adapter *adapter)
2100 {
2101 	struct net_device *netdev = adapter->netdev;
2102 	int err = 0, vector = 0;
2103 
2104 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2105 		snprintf(adapter->rx_ring->name,
2106 			 sizeof(adapter->rx_ring->name) - 1,
2107 			 "%.14s-rx-0", netdev->name);
2108 	else
2109 		memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
2110 	err = request_irq(adapter->msix_entries[vector].vector,
2111 			  e1000_intr_msix_rx, 0, adapter->rx_ring->name,
2112 			  netdev);
2113 	if (err)
2114 		return err;
2115 	adapter->rx_ring->itr_register = adapter->hw.hw_addr +
2116 	    E1000_EITR_82574(vector);
2117 	adapter->rx_ring->itr_val = adapter->itr;
2118 	vector++;
2119 
2120 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2121 		snprintf(adapter->tx_ring->name,
2122 			 sizeof(adapter->tx_ring->name) - 1,
2123 			 "%.14s-tx-0", netdev->name);
2124 	else
2125 		memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
2126 	err = request_irq(adapter->msix_entries[vector].vector,
2127 			  e1000_intr_msix_tx, 0, adapter->tx_ring->name,
2128 			  netdev);
2129 	if (err)
2130 		return err;
2131 	adapter->tx_ring->itr_register = adapter->hw.hw_addr +
2132 	    E1000_EITR_82574(vector);
2133 	adapter->tx_ring->itr_val = adapter->itr;
2134 	vector++;
2135 
2136 	err = request_irq(adapter->msix_entries[vector].vector,
2137 			  e1000_msix_other, 0, netdev->name, netdev);
2138 	if (err)
2139 		return err;
2140 
2141 	e1000_configure_msix(adapter);
2142 
2143 	return 0;
2144 }
2145 
2146 /**
2147  * e1000_request_irq - initialize interrupts
2148  * @adapter: board private structure
2149  *
2150  * Attempts to configure interrupts using the best available
2151  * capabilities of the hardware and kernel.
2152  **/
2153 static int e1000_request_irq(struct e1000_adapter *adapter)
2154 {
2155 	struct net_device *netdev = adapter->netdev;
2156 	int err;
2157 
2158 	if (adapter->msix_entries) {
2159 		err = e1000_request_msix(adapter);
2160 		if (!err)
2161 			return err;
2162 		/* fall back to MSI */
2163 		e1000e_reset_interrupt_capability(adapter);
2164 		adapter->int_mode = E1000E_INT_MODE_MSI;
2165 		e1000e_set_interrupt_capability(adapter);
2166 	}
2167 	if (adapter->flags & FLAG_MSI_ENABLED) {
2168 		err = request_irq(adapter->pdev->irq, e1000_intr_msi, 0,
2169 				  netdev->name, netdev);
2170 		if (!err)
2171 			return err;
2172 
2173 		/* fall back to legacy interrupt */
2174 		e1000e_reset_interrupt_capability(adapter);
2175 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
2176 	}
2177 
2178 	err = request_irq(adapter->pdev->irq, e1000_intr, IRQF_SHARED,
2179 			  netdev->name, netdev);
2180 	if (err)
2181 		e_err("Unable to allocate interrupt, Error: %d\n", err);
2182 
2183 	return err;
2184 }
2185 
2186 static void e1000_free_irq(struct e1000_adapter *adapter)
2187 {
2188 	struct net_device *netdev = adapter->netdev;
2189 
2190 	if (adapter->msix_entries) {
2191 		int vector = 0;
2192 
2193 		free_irq(adapter->msix_entries[vector].vector, netdev);
2194 		vector++;
2195 
2196 		free_irq(adapter->msix_entries[vector].vector, netdev);
2197 		vector++;
2198 
2199 		/* Other Causes interrupt vector */
2200 		free_irq(adapter->msix_entries[vector].vector, netdev);
2201 		return;
2202 	}
2203 
2204 	free_irq(adapter->pdev->irq, netdev);
2205 }
2206 
2207 /**
2208  * e1000_irq_disable - Mask off interrupt generation on the NIC
2209  * @adapter: board private structure
2210  **/
2211 static void e1000_irq_disable(struct e1000_adapter *adapter)
2212 {
2213 	struct e1000_hw *hw = &adapter->hw;
2214 
2215 	ew32(IMC, ~0);
2216 	if (adapter->msix_entries)
2217 		ew32(EIAC_82574, 0);
2218 	e1e_flush();
2219 
2220 	if (adapter->msix_entries) {
2221 		int i;
2222 
2223 		for (i = 0; i < adapter->num_vectors; i++)
2224 			synchronize_irq(adapter->msix_entries[i].vector);
2225 	} else {
2226 		synchronize_irq(adapter->pdev->irq);
2227 	}
2228 }
2229 
2230 /**
2231  * e1000_irq_enable - Enable default interrupt generation settings
2232  * @adapter: board private structure
2233  **/
2234 static void e1000_irq_enable(struct e1000_adapter *adapter)
2235 {
2236 	struct e1000_hw *hw = &adapter->hw;
2237 
2238 	if (adapter->msix_entries) {
2239 		ew32(EIAC_82574, adapter->eiac_mask & E1000_EIAC_MASK_82574);
2240 		ew32(IMS, adapter->eiac_mask | E1000_IMS_OTHER |
2241 		     IMS_OTHER_MASK);
2242 	} else if (hw->mac.type >= e1000_pch_lpt) {
2243 		ew32(IMS, IMS_ENABLE_MASK | E1000_IMS_ECCER);
2244 	} else {
2245 		ew32(IMS, IMS_ENABLE_MASK);
2246 	}
2247 	e1e_flush();
2248 }
2249 
2250 /**
2251  * e1000e_get_hw_control - get control of the h/w from f/w
2252  * @adapter: address of board private structure
2253  *
2254  * e1000e_get_hw_control sets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2255  * For ASF and Pass Through versions of f/w this means that
2256  * the driver is loaded. For AMT version (only with 82573)
2257  * of the f/w this means that the network i/f is open.
2258  **/
2259 void e1000e_get_hw_control(struct e1000_adapter *adapter)
2260 {
2261 	struct e1000_hw *hw = &adapter->hw;
2262 	u32 ctrl_ext;
2263 	u32 swsm;
2264 
2265 	/* Let firmware know the driver has taken over */
2266 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2267 		swsm = er32(SWSM);
2268 		ew32(SWSM, swsm | E1000_SWSM_DRV_LOAD);
2269 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2270 		ctrl_ext = er32(CTRL_EXT);
2271 		ew32(CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2272 	}
2273 }
2274 
2275 /**
2276  * e1000e_release_hw_control - release control of the h/w to f/w
2277  * @adapter: address of board private structure
2278  *
2279  * e1000e_release_hw_control resets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2280  * For ASF and Pass Through versions of f/w this means that the
2281  * driver is no longer loaded. For AMT version (only with 82573) i
2282  * of the f/w this means that the network i/f is closed.
2283  *
2284  **/
2285 void e1000e_release_hw_control(struct e1000_adapter *adapter)
2286 {
2287 	struct e1000_hw *hw = &adapter->hw;
2288 	u32 ctrl_ext;
2289 	u32 swsm;
2290 
2291 	/* Let firmware taken over control of h/w */
2292 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2293 		swsm = er32(SWSM);
2294 		ew32(SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
2295 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2296 		ctrl_ext = er32(CTRL_EXT);
2297 		ew32(CTRL_EXT, ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2298 	}
2299 }
2300 
2301 /**
2302  * e1000_alloc_ring_dma - allocate memory for a ring structure
2303  * @adapter: board private structure
2304  * @ring: ring struct for which to allocate dma
2305  **/
2306 static int e1000_alloc_ring_dma(struct e1000_adapter *adapter,
2307 				struct e1000_ring *ring)
2308 {
2309 	struct pci_dev *pdev = adapter->pdev;
2310 
2311 	ring->desc = dma_alloc_coherent(&pdev->dev, ring->size, &ring->dma,
2312 					GFP_KERNEL);
2313 	if (!ring->desc)
2314 		return -ENOMEM;
2315 
2316 	return 0;
2317 }
2318 
2319 /**
2320  * e1000e_setup_tx_resources - allocate Tx resources (Descriptors)
2321  * @tx_ring: Tx descriptor ring
2322  *
2323  * Return 0 on success, negative on failure
2324  **/
2325 int e1000e_setup_tx_resources(struct e1000_ring *tx_ring)
2326 {
2327 	struct e1000_adapter *adapter = tx_ring->adapter;
2328 	int err = -ENOMEM, size;
2329 
2330 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2331 	tx_ring->buffer_info = vzalloc(size);
2332 	if (!tx_ring->buffer_info)
2333 		goto err;
2334 
2335 	/* round up to nearest 4K */
2336 	tx_ring->size = tx_ring->count * sizeof(struct e1000_tx_desc);
2337 	tx_ring->size = ALIGN(tx_ring->size, 4096);
2338 
2339 	err = e1000_alloc_ring_dma(adapter, tx_ring);
2340 	if (err)
2341 		goto err;
2342 
2343 	tx_ring->next_to_use = 0;
2344 	tx_ring->next_to_clean = 0;
2345 
2346 	return 0;
2347 err:
2348 	vfree(tx_ring->buffer_info);
2349 	e_err("Unable to allocate memory for the transmit descriptor ring\n");
2350 	return err;
2351 }
2352 
2353 /**
2354  * e1000e_setup_rx_resources - allocate Rx resources (Descriptors)
2355  * @rx_ring: Rx descriptor ring
2356  *
2357  * Returns 0 on success, negative on failure
2358  **/
2359 int e1000e_setup_rx_resources(struct e1000_ring *rx_ring)
2360 {
2361 	struct e1000_adapter *adapter = rx_ring->adapter;
2362 	struct e1000_buffer *buffer_info;
2363 	int i, size, desc_len, err = -ENOMEM;
2364 
2365 	size = sizeof(struct e1000_buffer) * rx_ring->count;
2366 	rx_ring->buffer_info = vzalloc(size);
2367 	if (!rx_ring->buffer_info)
2368 		goto err;
2369 
2370 	for (i = 0; i < rx_ring->count; i++) {
2371 		buffer_info = &rx_ring->buffer_info[i];
2372 		buffer_info->ps_pages = kzalloc_objs(struct e1000_ps_page,
2373 						     PS_PAGE_BUFFERS,
2374 						     GFP_KERNEL);
2375 		if (!buffer_info->ps_pages)
2376 			goto err_pages;
2377 	}
2378 
2379 	desc_len = sizeof(union e1000_rx_desc_packet_split);
2380 
2381 	/* Round up to nearest 4K */
2382 	rx_ring->size = rx_ring->count * desc_len;
2383 	rx_ring->size = ALIGN(rx_ring->size, 4096);
2384 
2385 	err = e1000_alloc_ring_dma(adapter, rx_ring);
2386 	if (err)
2387 		goto err_pages;
2388 
2389 	rx_ring->next_to_clean = 0;
2390 	rx_ring->next_to_use = 0;
2391 	rx_ring->rx_skb_top = NULL;
2392 
2393 	return 0;
2394 
2395 err_pages:
2396 	for (i = 0; i < rx_ring->count; i++) {
2397 		buffer_info = &rx_ring->buffer_info[i];
2398 		kfree(buffer_info->ps_pages);
2399 	}
2400 err:
2401 	vfree(rx_ring->buffer_info);
2402 	e_err("Unable to allocate memory for the receive descriptor ring\n");
2403 	return err;
2404 }
2405 
2406 /**
2407  * e1000_clean_tx_ring - Free Tx Buffers
2408  * @tx_ring: Tx descriptor ring
2409  **/
2410 static void e1000_clean_tx_ring(struct e1000_ring *tx_ring)
2411 {
2412 	struct e1000_adapter *adapter = tx_ring->adapter;
2413 	struct e1000_buffer *buffer_info;
2414 	unsigned long size;
2415 	unsigned int i;
2416 
2417 	for (i = 0; i < tx_ring->count; i++) {
2418 		buffer_info = &tx_ring->buffer_info[i];
2419 		e1000_put_txbuf(tx_ring, buffer_info, false);
2420 	}
2421 
2422 	netdev_reset_queue(adapter->netdev);
2423 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2424 	memset(tx_ring->buffer_info, 0, size);
2425 
2426 	memset(tx_ring->desc, 0, tx_ring->size);
2427 
2428 	tx_ring->next_to_use = 0;
2429 	tx_ring->next_to_clean = 0;
2430 }
2431 
2432 /**
2433  * e1000e_free_tx_resources - Free Tx Resources per Queue
2434  * @tx_ring: Tx descriptor ring
2435  *
2436  * Free all transmit software resources
2437  **/
2438 void e1000e_free_tx_resources(struct e1000_ring *tx_ring)
2439 {
2440 	struct e1000_adapter *adapter = tx_ring->adapter;
2441 	struct pci_dev *pdev = adapter->pdev;
2442 
2443 	e1000_clean_tx_ring(tx_ring);
2444 
2445 	vfree(tx_ring->buffer_info);
2446 	tx_ring->buffer_info = NULL;
2447 
2448 	dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
2449 			  tx_ring->dma);
2450 	tx_ring->desc = NULL;
2451 }
2452 
2453 /**
2454  * e1000e_free_rx_resources - Free Rx Resources
2455  * @rx_ring: Rx descriptor ring
2456  *
2457  * Free all receive software resources
2458  **/
2459 void e1000e_free_rx_resources(struct e1000_ring *rx_ring)
2460 {
2461 	struct e1000_adapter *adapter = rx_ring->adapter;
2462 	struct pci_dev *pdev = adapter->pdev;
2463 	int i;
2464 
2465 	e1000_clean_rx_ring(rx_ring);
2466 
2467 	for (i = 0; i < rx_ring->count; i++)
2468 		kfree(rx_ring->buffer_info[i].ps_pages);
2469 
2470 	vfree(rx_ring->buffer_info);
2471 	rx_ring->buffer_info = NULL;
2472 
2473 	dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
2474 			  rx_ring->dma);
2475 	rx_ring->desc = NULL;
2476 }
2477 
2478 /**
2479  * e1000_update_itr - update the dynamic ITR value based on statistics
2480  * @itr_setting: current adapter->itr
2481  * @packets: the number of packets during this measurement interval
2482  * @bytes: the number of bytes during this measurement interval
2483  *
2484  *      Stores a new ITR value based on packets and byte
2485  *      counts during the last interrupt.  The advantage of per interrupt
2486  *      computation is faster updates and more accurate ITR for the current
2487  *      traffic pattern.  Constants in this function were computed
2488  *      based on theoretical maximum wire speed and thresholds were set based
2489  *      on testing data as well as attempting to minimize response time
2490  *      while increasing bulk throughput.  This functionality is controlled
2491  *      by the InterruptThrottleRate module parameter.
2492  **/
2493 static unsigned int e1000_update_itr(u16 itr_setting, int packets, int bytes)
2494 {
2495 	unsigned int retval = itr_setting;
2496 
2497 	if (packets == 0)
2498 		return itr_setting;
2499 
2500 	switch (itr_setting) {
2501 	case lowest_latency:
2502 		/* handle TSO and jumbo frames */
2503 		if (bytes / packets > 8000)
2504 			retval = bulk_latency;
2505 		else if ((packets < 5) && (bytes > 512))
2506 			retval = low_latency;
2507 		break;
2508 	case low_latency:	/* 50 usec aka 20000 ints/s */
2509 		if (bytes > 10000) {
2510 			/* this if handles the TSO accounting */
2511 			if (bytes / packets > 8000)
2512 				retval = bulk_latency;
2513 			else if ((packets < 10) || ((bytes / packets) > 1200))
2514 				retval = bulk_latency;
2515 			else if ((packets > 35))
2516 				retval = lowest_latency;
2517 		} else if (bytes / packets > 2000) {
2518 			retval = bulk_latency;
2519 		} else if (packets <= 2 && bytes < 512) {
2520 			retval = lowest_latency;
2521 		}
2522 		break;
2523 	case bulk_latency:	/* 250 usec aka 4000 ints/s */
2524 		if (bytes > 25000) {
2525 			if (packets > 35)
2526 				retval = low_latency;
2527 		} else if (bytes < 6000) {
2528 			retval = low_latency;
2529 		}
2530 		break;
2531 	}
2532 
2533 	return retval;
2534 }
2535 
2536 static void e1000_set_itr(struct e1000_adapter *adapter)
2537 {
2538 	u16 current_itr;
2539 	u32 new_itr = adapter->itr;
2540 
2541 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
2542 	if (adapter->link_speed != SPEED_1000) {
2543 		new_itr = 4000;
2544 		goto set_itr_now;
2545 	}
2546 
2547 	if (adapter->flags2 & FLAG2_DISABLE_AIM) {
2548 		new_itr = 0;
2549 		goto set_itr_now;
2550 	}
2551 
2552 	adapter->tx_itr = e1000_update_itr(adapter->tx_itr,
2553 					   adapter->total_tx_packets,
2554 					   adapter->total_tx_bytes);
2555 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2556 	if (adapter->itr_setting == 3 && adapter->tx_itr == lowest_latency)
2557 		adapter->tx_itr = low_latency;
2558 
2559 	adapter->rx_itr = e1000_update_itr(adapter->rx_itr,
2560 					   adapter->total_rx_packets,
2561 					   adapter->total_rx_bytes);
2562 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2563 	if (adapter->itr_setting == 3 && adapter->rx_itr == lowest_latency)
2564 		adapter->rx_itr = low_latency;
2565 
2566 	current_itr = max(adapter->rx_itr, adapter->tx_itr);
2567 
2568 	/* counts and packets in update_itr are dependent on these numbers */
2569 	switch (current_itr) {
2570 	case lowest_latency:
2571 		new_itr = 70000;
2572 		break;
2573 	case low_latency:
2574 		new_itr = 20000;	/* aka hwitr = ~200 */
2575 		break;
2576 	case bulk_latency:
2577 		new_itr = 4000;
2578 		break;
2579 	default:
2580 		break;
2581 	}
2582 
2583 set_itr_now:
2584 	if (new_itr != adapter->itr) {
2585 		/* this attempts to bias the interrupt rate towards Bulk
2586 		 * by adding intermediate steps when interrupt rate is
2587 		 * increasing
2588 		 */
2589 		new_itr = new_itr > adapter->itr ?
2590 		    min(adapter->itr + (new_itr >> 2), new_itr) : new_itr;
2591 		adapter->itr = new_itr;
2592 		adapter->rx_ring->itr_val = new_itr;
2593 		if (adapter->msix_entries)
2594 			adapter->rx_ring->set_itr = 1;
2595 		else
2596 			e1000e_write_itr(adapter, new_itr);
2597 	}
2598 }
2599 
2600 /**
2601  * e1000e_write_itr - write the ITR value to the appropriate registers
2602  * @adapter: address of board private structure
2603  * @itr: new ITR value to program
2604  *
2605  * e1000e_write_itr determines if the adapter is in MSI-X mode
2606  * and, if so, writes the EITR registers with the ITR value.
2607  * Otherwise, it writes the ITR value into the ITR register.
2608  **/
2609 void e1000e_write_itr(struct e1000_adapter *adapter, u32 itr)
2610 {
2611 	struct e1000_hw *hw = &adapter->hw;
2612 	u32 new_itr = itr ? 1000000000 / (itr * 256) : 0;
2613 
2614 	if (adapter->msix_entries) {
2615 		int vector;
2616 
2617 		for (vector = 0; vector < adapter->num_vectors; vector++)
2618 			writel(new_itr, hw->hw_addr + E1000_EITR_82574(vector));
2619 	} else {
2620 		ew32(ITR, new_itr);
2621 	}
2622 }
2623 
2624 /**
2625  * e1000_alloc_queues - Allocate memory for all rings
2626  * @adapter: board private structure to initialize
2627  **/
2628 static int e1000_alloc_queues(struct e1000_adapter *adapter)
2629 {
2630 	int size = sizeof(struct e1000_ring);
2631 
2632 	adapter->tx_ring = kzalloc(size, GFP_KERNEL);
2633 	if (!adapter->tx_ring)
2634 		goto err;
2635 	adapter->tx_ring->count = adapter->tx_ring_count;
2636 	adapter->tx_ring->adapter = adapter;
2637 
2638 	adapter->rx_ring = kzalloc(size, GFP_KERNEL);
2639 	if (!adapter->rx_ring)
2640 		goto err;
2641 	adapter->rx_ring->count = adapter->rx_ring_count;
2642 	adapter->rx_ring->adapter = adapter;
2643 
2644 	return 0;
2645 err:
2646 	e_err("Unable to allocate memory for queues\n");
2647 	kfree(adapter->rx_ring);
2648 	kfree(adapter->tx_ring);
2649 	return -ENOMEM;
2650 }
2651 
2652 /**
2653  * e1000e_poll - NAPI Rx polling callback
2654  * @napi: struct associated with this polling callback
2655  * @budget: number of packets driver is allowed to process this poll
2656  **/
2657 static int e1000e_poll(struct napi_struct *napi, int budget)
2658 {
2659 	struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter,
2660 						     napi);
2661 	struct e1000_hw *hw = &adapter->hw;
2662 	struct net_device *poll_dev = adapter->netdev;
2663 	int tx_cleaned = 1, work_done = 0;
2664 
2665 	adapter = netdev_priv(poll_dev);
2666 
2667 	if (!adapter->msix_entries ||
2668 	    (adapter->rx_ring->ims_val & adapter->tx_ring->ims_val))
2669 		tx_cleaned = e1000_clean_tx_irq(adapter->tx_ring);
2670 
2671 	adapter->clean_rx(adapter->rx_ring, &work_done, budget);
2672 
2673 	if (!tx_cleaned || work_done == budget)
2674 		return budget;
2675 
2676 	/* Exit the polling mode, but don't re-enable interrupts if stack might
2677 	 * poll us due to busy-polling
2678 	 */
2679 	if (likely(napi_complete_done(napi, work_done))) {
2680 		if (adapter->itr_setting & 3)
2681 			e1000_set_itr(adapter);
2682 		if (!test_bit(__E1000_DOWN, &adapter->state)) {
2683 			if (adapter->msix_entries)
2684 				ew32(IMS, adapter->rx_ring->ims_val);
2685 			else
2686 				e1000_irq_enable(adapter);
2687 		}
2688 	}
2689 
2690 	return work_done;
2691 }
2692 
2693 static int e1000_vlan_rx_add_vid(struct net_device *netdev,
2694 				 __always_unused __be16 proto, u16 vid)
2695 {
2696 	struct e1000_adapter *adapter = netdev_priv(netdev);
2697 	struct e1000_hw *hw = &adapter->hw;
2698 	u32 vfta, index;
2699 
2700 	/* don't update vlan cookie if already programmed */
2701 	if ((adapter->hw.mng_cookie.status &
2702 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2703 	    (vid == adapter->mng_vlan_id))
2704 		return 0;
2705 
2706 	/* add VID to filter table */
2707 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2708 		index = (vid >> 5) & 0x7F;
2709 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2710 		vfta |= BIT((vid & 0x1F));
2711 		hw->mac.ops.write_vfta(hw, index, vfta);
2712 	}
2713 
2714 	set_bit(vid, adapter->active_vlans);
2715 
2716 	return 0;
2717 }
2718 
2719 static int e1000_vlan_rx_kill_vid(struct net_device *netdev,
2720 				  __always_unused __be16 proto, u16 vid)
2721 {
2722 	struct e1000_adapter *adapter = netdev_priv(netdev);
2723 	struct e1000_hw *hw = &adapter->hw;
2724 	u32 vfta, index;
2725 
2726 	if ((adapter->hw.mng_cookie.status &
2727 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2728 	    (vid == adapter->mng_vlan_id)) {
2729 		/* release control to f/w */
2730 		e1000e_release_hw_control(adapter);
2731 		return 0;
2732 	}
2733 
2734 	/* remove VID from filter table */
2735 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2736 		index = (vid >> 5) & 0x7F;
2737 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2738 		vfta &= ~BIT((vid & 0x1F));
2739 		hw->mac.ops.write_vfta(hw, index, vfta);
2740 	}
2741 
2742 	clear_bit(vid, adapter->active_vlans);
2743 
2744 	return 0;
2745 }
2746 
2747 /**
2748  * e1000e_vlan_filter_disable - helper to disable hw VLAN filtering
2749  * @adapter: board private structure to initialize
2750  **/
2751 static void e1000e_vlan_filter_disable(struct e1000_adapter *adapter)
2752 {
2753 	struct net_device *netdev = adapter->netdev;
2754 	struct e1000_hw *hw = &adapter->hw;
2755 	u32 rctl;
2756 
2757 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2758 		/* disable VLAN receive filtering */
2759 		rctl = er32(RCTL);
2760 		rctl &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
2761 		ew32(RCTL, rctl);
2762 
2763 		if (adapter->mng_vlan_id != E1000_MNG_VLAN_NONE) {
2764 			e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
2765 					       adapter->mng_vlan_id);
2766 			adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
2767 		}
2768 	}
2769 }
2770 
2771 /**
2772  * e1000e_vlan_filter_enable - helper to enable HW VLAN filtering
2773  * @adapter: board private structure to initialize
2774  **/
2775 static void e1000e_vlan_filter_enable(struct e1000_adapter *adapter)
2776 {
2777 	struct e1000_hw *hw = &adapter->hw;
2778 	u32 rctl;
2779 
2780 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2781 		/* enable VLAN receive filtering */
2782 		rctl = er32(RCTL);
2783 		rctl |= E1000_RCTL_VFE;
2784 		rctl &= ~E1000_RCTL_CFIEN;
2785 		ew32(RCTL, rctl);
2786 	}
2787 }
2788 
2789 /**
2790  * e1000e_vlan_strip_disable - helper to disable HW VLAN stripping
2791  * @adapter: board private structure to initialize
2792  **/
2793 static void e1000e_vlan_strip_disable(struct e1000_adapter *adapter)
2794 {
2795 	struct e1000_hw *hw = &adapter->hw;
2796 	u32 ctrl;
2797 
2798 	/* disable VLAN tag insert/strip */
2799 	ctrl = er32(CTRL);
2800 	ctrl &= ~E1000_CTRL_VME;
2801 	ew32(CTRL, ctrl);
2802 }
2803 
2804 /**
2805  * e1000e_vlan_strip_enable - helper to enable HW VLAN stripping
2806  * @adapter: board private structure to initialize
2807  **/
2808 static void e1000e_vlan_strip_enable(struct e1000_adapter *adapter)
2809 {
2810 	struct e1000_hw *hw = &adapter->hw;
2811 	u32 ctrl;
2812 
2813 	/* enable VLAN tag insert/strip */
2814 	ctrl = er32(CTRL);
2815 	ctrl |= E1000_CTRL_VME;
2816 	ew32(CTRL, ctrl);
2817 }
2818 
2819 static void e1000_update_mng_vlan(struct e1000_adapter *adapter)
2820 {
2821 	struct net_device *netdev = adapter->netdev;
2822 	u16 vid = adapter->hw.mng_cookie.vlan_id;
2823 	u16 old_vid = adapter->mng_vlan_id;
2824 
2825 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
2826 		e1000_vlan_rx_add_vid(netdev, htons(ETH_P_8021Q), vid);
2827 		adapter->mng_vlan_id = vid;
2828 	}
2829 
2830 	if (old_vid != E1000_MNG_VLAN_NONE && vid != old_vid)
2831 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q), old_vid);
2832 }
2833 
2834 static void e1000_restore_vlan(struct e1000_adapter *adapter)
2835 {
2836 	u16 vid;
2837 
2838 	e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
2839 
2840 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
2841 	    e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
2842 }
2843 
2844 static void e1000_init_manageability_pt(struct e1000_adapter *adapter)
2845 {
2846 	struct e1000_hw *hw = &adapter->hw;
2847 	u32 manc, manc2h, mdef, i, j;
2848 
2849 	if (!(adapter->flags & FLAG_MNG_PT_ENABLED))
2850 		return;
2851 
2852 	manc = er32(MANC);
2853 
2854 	/* enable receiving management packets to the host. this will probably
2855 	 * generate destination unreachable messages from the host OS, but
2856 	 * the packets will be handled on SMBUS
2857 	 */
2858 	manc |= E1000_MANC_EN_MNG2HOST;
2859 	manc2h = er32(MANC2H);
2860 
2861 	switch (hw->mac.type) {
2862 	default:
2863 		manc2h |= (E1000_MANC2H_PORT_623 | E1000_MANC2H_PORT_664);
2864 		break;
2865 	case e1000_82574:
2866 	case e1000_82583:
2867 		/* Check if IPMI pass-through decision filter already exists;
2868 		 * if so, enable it.
2869 		 */
2870 		for (i = 0, j = 0; i < 8; i++) {
2871 			mdef = er32(MDEF(i));
2872 
2873 			/* Ignore filters with anything other than IPMI ports */
2874 			if (mdef & ~(E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2875 				continue;
2876 
2877 			/* Enable this decision filter in MANC2H */
2878 			if (mdef)
2879 				manc2h |= BIT(i);
2880 
2881 			j |= mdef;
2882 		}
2883 
2884 		if (j == (E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2885 			break;
2886 
2887 		/* Create new decision filter in an empty filter */
2888 		for (i = 0, j = 0; i < 8; i++)
2889 			if (er32(MDEF(i)) == 0) {
2890 				ew32(MDEF(i), (E1000_MDEF_PORT_623 |
2891 					       E1000_MDEF_PORT_664));
2892 				manc2h |= BIT(1);
2893 				j++;
2894 				break;
2895 			}
2896 
2897 		if (!j)
2898 			e_warn("Unable to create IPMI pass-through filter\n");
2899 		break;
2900 	}
2901 
2902 	ew32(MANC2H, manc2h);
2903 	ew32(MANC, manc);
2904 }
2905 
2906 /**
2907  * e1000_configure_tx - Configure Transmit Unit after Reset
2908  * @adapter: board private structure
2909  *
2910  * Configure the Tx unit of the MAC after a reset.
2911  **/
2912 static void e1000_configure_tx(struct e1000_adapter *adapter)
2913 {
2914 	struct e1000_hw *hw = &adapter->hw;
2915 	struct e1000_ring *tx_ring = adapter->tx_ring;
2916 	u64 tdba;
2917 	u32 tdlen, tctl, tarc;
2918 
2919 	/* Setup the HW Tx Head and Tail descriptor pointers */
2920 	tdba = tx_ring->dma;
2921 	tdlen = tx_ring->count * sizeof(struct e1000_tx_desc);
2922 	ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
2923 	ew32(TDBAH(0), (tdba >> 32));
2924 	ew32(TDLEN(0), tdlen);
2925 	ew32(TDH(0), 0);
2926 	ew32(TDT(0), 0);
2927 	tx_ring->head = adapter->hw.hw_addr + E1000_TDH(0);
2928 	tx_ring->tail = adapter->hw.hw_addr + E1000_TDT(0);
2929 
2930 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
2931 		e1000e_update_tdt_wa(tx_ring, 0);
2932 
2933 	/* Set the Tx Interrupt Delay register */
2934 	ew32(TIDV, adapter->tx_int_delay);
2935 	/* Tx irq moderation */
2936 	ew32(TADV, adapter->tx_abs_int_delay);
2937 
2938 	if (adapter->flags2 & FLAG2_DMA_BURST) {
2939 		u32 txdctl = er32(TXDCTL(0));
2940 
2941 		txdctl &= ~(E1000_TXDCTL_PTHRESH | E1000_TXDCTL_HTHRESH |
2942 			    E1000_TXDCTL_WTHRESH);
2943 		/* set up some performance related parameters to encourage the
2944 		 * hardware to use the bus more efficiently in bursts, depends
2945 		 * on the tx_int_delay to be enabled,
2946 		 * wthresh = 1 ==> burst write is disabled to avoid Tx stalls
2947 		 * hthresh = 1 ==> prefetch when one or more available
2948 		 * pthresh = 0x1f ==> prefetch if internal cache 31 or less
2949 		 * BEWARE: this seems to work but should be considered first if
2950 		 * there are Tx hangs or other Tx related bugs
2951 		 */
2952 		txdctl |= E1000_TXDCTL_DMA_BURST_ENABLE;
2953 		ew32(TXDCTL(0), txdctl);
2954 	}
2955 	/* erratum work around: set txdctl the same for both queues */
2956 	ew32(TXDCTL(1), er32(TXDCTL(0)));
2957 
2958 	/* Program the Transmit Control Register */
2959 	tctl = er32(TCTL);
2960 	tctl &= ~E1000_TCTL_CT;
2961 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
2962 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
2963 
2964 	if (adapter->flags & FLAG_TARC_SPEED_MODE_BIT) {
2965 		tarc = er32(TARC(0));
2966 		/* set the speed mode bit, we'll clear it if we're not at
2967 		 * gigabit link later
2968 		 */
2969 #define SPEED_MODE_BIT BIT(21)
2970 		tarc |= SPEED_MODE_BIT;
2971 		ew32(TARC(0), tarc);
2972 	}
2973 
2974 	/* errata: program both queues to unweighted RR */
2975 	if (adapter->flags & FLAG_TARC_SET_BIT_ZERO) {
2976 		tarc = er32(TARC(0));
2977 		tarc |= 1;
2978 		ew32(TARC(0), tarc);
2979 		tarc = er32(TARC(1));
2980 		tarc |= 1;
2981 		ew32(TARC(1), tarc);
2982 	}
2983 
2984 	/* Setup Transmit Descriptor Settings for eop descriptor */
2985 	adapter->txd_cmd = E1000_TXD_CMD_EOP | E1000_TXD_CMD_IFCS;
2986 
2987 	/* only set IDE if we are delaying interrupts using the timers */
2988 	if (adapter->tx_int_delay)
2989 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
2990 
2991 	/* enable Report Status bit */
2992 	adapter->txd_cmd |= E1000_TXD_CMD_RS;
2993 
2994 	ew32(TCTL, tctl);
2995 
2996 	hw->mac.ops.config_collision_dist(hw);
2997 
2998 	/* SPT and KBL Si errata workaround to avoid data corruption */
2999 	if (hw->mac.type == e1000_pch_spt) {
3000 		u32 reg_val;
3001 
3002 		reg_val = er32(IOSFPC);
3003 		reg_val |= E1000_RCTL_RDMTS_HEX;
3004 		ew32(IOSFPC, reg_val);
3005 
3006 		reg_val = er32(TARC(0));
3007 		/* SPT and KBL Si errata workaround to avoid Tx hang.
3008 		 * Dropping the number of outstanding requests from
3009 		 * 3 to 2 in order to avoid a buffer overrun.
3010 		 */
3011 		reg_val &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3012 		reg_val |= E1000_TARC0_CB_MULTIQ_2_REQ;
3013 		ew32(TARC(0), reg_val);
3014 	}
3015 }
3016 
3017 #define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
3018 			   (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
3019 
3020 /**
3021  * e1000_setup_rctl - configure the receive control registers
3022  * @adapter: Board private structure
3023  **/
3024 static void e1000_setup_rctl(struct e1000_adapter *adapter)
3025 {
3026 	struct e1000_hw *hw = &adapter->hw;
3027 	u32 rctl, rfctl;
3028 	u32 pages = 0;
3029 
3030 	/* Workaround Si errata on PCHx - configure jumbo frame flow.
3031 	 * If jumbo frames not set, program related MAC/PHY registers
3032 	 * to h/w defaults
3033 	 */
3034 	if (hw->mac.type >= e1000_pch2lan) {
3035 		s32 ret_val;
3036 
3037 		if (adapter->netdev->mtu > ETH_DATA_LEN)
3038 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, true);
3039 		else
3040 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, false);
3041 
3042 		if (ret_val)
3043 			e_dbg("failed to enable|disable jumbo frame workaround mode\n");
3044 	}
3045 
3046 	/* Program MC offset vector base */
3047 	rctl = er32(RCTL);
3048 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3049 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3050 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3051 	    (adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3052 
3053 	/* Do not Store bad packets */
3054 	rctl &= ~E1000_RCTL_SBP;
3055 
3056 	/* Enable Long Packet receive */
3057 	if (adapter->netdev->mtu <= ETH_DATA_LEN)
3058 		rctl &= ~E1000_RCTL_LPE;
3059 	else
3060 		rctl |= E1000_RCTL_LPE;
3061 
3062 	/* Some systems expect that the CRC is included in SMBUS traffic. The
3063 	 * hardware strips the CRC before sending to both SMBUS (BMC) and to
3064 	 * host memory when this is enabled
3065 	 */
3066 	if (adapter->flags2 & FLAG2_CRC_STRIPPING)
3067 		rctl |= E1000_RCTL_SECRC;
3068 
3069 	/* Workaround Si errata on 82577 PHY - configure IPG for jumbos */
3070 	if ((hw->phy.type == e1000_phy_82577) && (rctl & E1000_RCTL_LPE)) {
3071 		u16 phy_data;
3072 
3073 		e1e_rphy(hw, PHY_REG(770, 26), &phy_data);
3074 		phy_data &= 0xfff8;
3075 		phy_data |= BIT(2);
3076 		e1e_wphy(hw, PHY_REG(770, 26), phy_data);
3077 
3078 		e1e_rphy(hw, 22, &phy_data);
3079 		phy_data &= 0x0fff;
3080 		phy_data |= BIT(14);
3081 		e1e_wphy(hw, 0x10, 0x2823);
3082 		e1e_wphy(hw, 0x11, 0x0003);
3083 		e1e_wphy(hw, 22, phy_data);
3084 	}
3085 
3086 	/* Setup buffer sizes */
3087 	rctl &= ~E1000_RCTL_SZ_4096;
3088 	rctl |= E1000_RCTL_BSEX;
3089 	switch (adapter->rx_buffer_len) {
3090 	case 2048:
3091 	default:
3092 		rctl |= E1000_RCTL_SZ_2048;
3093 		rctl &= ~E1000_RCTL_BSEX;
3094 		break;
3095 	case 4096:
3096 		rctl |= E1000_RCTL_SZ_4096;
3097 		break;
3098 	case 8192:
3099 		rctl |= E1000_RCTL_SZ_8192;
3100 		break;
3101 	case 16384:
3102 		rctl |= E1000_RCTL_SZ_16384;
3103 		break;
3104 	}
3105 
3106 	/* Enable Extended Status in all Receive Descriptors */
3107 	rfctl = er32(RFCTL);
3108 	rfctl |= E1000_RFCTL_EXTEN;
3109 	ew32(RFCTL, rfctl);
3110 
3111 	/* 82571 and greater support packet-split where the protocol
3112 	 * header is placed in skb->data and the packet data is
3113 	 * placed in pages hanging off of skb_shinfo(skb)->nr_frags.
3114 	 * In the case of a non-split, skb->data is linearly filled,
3115 	 * followed by the page buffers.  Therefore, skb->data is
3116 	 * sized to hold the largest protocol header.
3117 	 *
3118 	 * allocations using alloc_page take too long for regular MTU
3119 	 * so only enable packet split for jumbo frames
3120 	 *
3121 	 * Using pages when the page size is greater than 16k wastes
3122 	 * a lot of memory, since we allocate 3 pages at all times
3123 	 * per packet.
3124 	 */
3125 	pages = PAGE_USE_COUNT(adapter->netdev->mtu);
3126 	if ((pages <= 3) && (PAGE_SIZE <= 16384) && (rctl & E1000_RCTL_LPE))
3127 		adapter->rx_ps_pages = pages;
3128 	else
3129 		adapter->rx_ps_pages = 0;
3130 
3131 	if (adapter->rx_ps_pages) {
3132 		u32 psrctl = 0;
3133 
3134 		/* Enable Packet split descriptors */
3135 		rctl |= E1000_RCTL_DTYP_PS;
3136 
3137 		psrctl |= adapter->rx_ps_bsize0 >> E1000_PSRCTL_BSIZE0_SHIFT;
3138 
3139 		switch (adapter->rx_ps_pages) {
3140 		case 3:
3141 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE3_SHIFT;
3142 			fallthrough;
3143 		case 2:
3144 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE2_SHIFT;
3145 			fallthrough;
3146 		case 1:
3147 			psrctl |= PAGE_SIZE >> E1000_PSRCTL_BSIZE1_SHIFT;
3148 			break;
3149 		}
3150 
3151 		ew32(PSRCTL, psrctl);
3152 	}
3153 
3154 	/* This is useful for sniffing bad packets. */
3155 	if (adapter->netdev->features & NETIF_F_RXALL) {
3156 		/* UPE and MPE will be handled by normal PROMISC logic
3157 		 * in e1000e_set_rx_mode
3158 		 */
3159 		rctl |= (E1000_RCTL_SBP |	/* Receive bad packets */
3160 			 E1000_RCTL_BAM |	/* RX All Bcast Pkts */
3161 			 E1000_RCTL_PMCF);	/* RX All MAC Ctrl Pkts */
3162 
3163 		rctl &= ~(E1000_RCTL_VFE |	/* Disable VLAN filter */
3164 			  E1000_RCTL_DPF |	/* Allow filtered pause */
3165 			  E1000_RCTL_CFIEN);	/* Dis VLAN CFIEN Filter */
3166 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3167 		 * and that breaks VLANs.
3168 		 */
3169 	}
3170 
3171 	ew32(RCTL, rctl);
3172 	/* just started the receive unit, no need to restart */
3173 	adapter->flags &= ~FLAG_RESTART_NOW;
3174 }
3175 
3176 /**
3177  * e1000_configure_rx - Configure Receive Unit after Reset
3178  * @adapter: board private structure
3179  *
3180  * Configure the Rx unit of the MAC after a reset.
3181  **/
3182 static void e1000_configure_rx(struct e1000_adapter *adapter)
3183 {
3184 	struct e1000_hw *hw = &adapter->hw;
3185 	struct e1000_ring *rx_ring = adapter->rx_ring;
3186 	u64 rdba;
3187 	u32 rdlen, rctl, rxcsum, ctrl_ext;
3188 
3189 	if (adapter->rx_ps_pages) {
3190 		/* this is a 32 byte descriptor */
3191 		rdlen = rx_ring->count *
3192 		    sizeof(union e1000_rx_desc_packet_split);
3193 		adapter->clean_rx = e1000_clean_rx_irq_ps;
3194 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers_ps;
3195 	} else if (adapter->netdev->mtu > ETH_FRAME_LEN + ETH_FCS_LEN) {
3196 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3197 		adapter->clean_rx = e1000_clean_jumbo_rx_irq;
3198 		adapter->alloc_rx_buf = e1000_alloc_jumbo_rx_buffers;
3199 	} else {
3200 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3201 		adapter->clean_rx = e1000_clean_rx_irq;
3202 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers;
3203 	}
3204 
3205 	/* disable receives while setting up the descriptors */
3206 	rctl = er32(RCTL);
3207 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
3208 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
3209 	e1e_flush();
3210 	usleep_range(10000, 11000);
3211 
3212 	if (adapter->flags2 & FLAG2_DMA_BURST) {
3213 		/* set the writeback threshold (only takes effect if the RDTR
3214 		 * is set). set GRAN=1 and write back up to 0x4 worth, and
3215 		 * enable prefetching of 0x20 Rx descriptors
3216 		 * granularity = 01
3217 		 * wthresh = 04,
3218 		 * hthresh = 04,
3219 		 * pthresh = 0x20
3220 		 */
3221 		ew32(RXDCTL(0), E1000_RXDCTL_DMA_BURST_ENABLE);
3222 		ew32(RXDCTL(1), E1000_RXDCTL_DMA_BURST_ENABLE);
3223 	}
3224 
3225 	/* set the Receive Delay Timer Register */
3226 	ew32(RDTR, adapter->rx_int_delay);
3227 
3228 	/* irq moderation */
3229 	ew32(RADV, adapter->rx_abs_int_delay);
3230 	if ((adapter->itr_setting != 0) && (adapter->itr != 0))
3231 		e1000e_write_itr(adapter, adapter->itr);
3232 
3233 	ctrl_ext = er32(CTRL_EXT);
3234 	/* Auto-Mask interrupts upon ICR access */
3235 	ctrl_ext |= E1000_CTRL_EXT_IAME;
3236 	ew32(IAM, 0xffffffff);
3237 	ew32(CTRL_EXT, ctrl_ext);
3238 	e1e_flush();
3239 
3240 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
3241 	 * the Base and Length of the Rx Descriptor Ring
3242 	 */
3243 	rdba = rx_ring->dma;
3244 	ew32(RDBAL(0), (rdba & DMA_BIT_MASK(32)));
3245 	ew32(RDBAH(0), (rdba >> 32));
3246 	ew32(RDLEN(0), rdlen);
3247 	ew32(RDH(0), 0);
3248 	ew32(RDT(0), 0);
3249 	rx_ring->head = adapter->hw.hw_addr + E1000_RDH(0);
3250 	rx_ring->tail = adapter->hw.hw_addr + E1000_RDT(0);
3251 
3252 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
3253 		e1000e_update_rdt_wa(rx_ring, 0);
3254 
3255 	/* Enable Receive Checksum Offload for TCP and UDP */
3256 	rxcsum = er32(RXCSUM);
3257 	if (adapter->netdev->features & NETIF_F_RXCSUM)
3258 		rxcsum |= E1000_RXCSUM_TUOFL;
3259 	else
3260 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3261 	ew32(RXCSUM, rxcsum);
3262 
3263 	/* With jumbo frames, excessive C-state transition latencies result
3264 	 * in dropped transactions.
3265 	 */
3266 	if (adapter->netdev->mtu > ETH_DATA_LEN) {
3267 		u32 lat =
3268 		    ((er32(PBA) & E1000_PBA_RXA_MASK) * 1024 -
3269 		     adapter->max_frame_size) * 8 / 1000;
3270 
3271 		if (adapter->flags & FLAG_IS_ICH) {
3272 			u32 rxdctl = er32(RXDCTL(0));
3273 
3274 			ew32(RXDCTL(0), rxdctl | 0x3 | BIT(8));
3275 		}
3276 
3277 		dev_info(&adapter->pdev->dev,
3278 			 "Some CPU C-states have been disabled in order to enable jumbo frames\n");
3279 		cpu_latency_qos_update_request(&adapter->pm_qos_req, lat);
3280 	} else {
3281 		cpu_latency_qos_update_request(&adapter->pm_qos_req,
3282 					       PM_QOS_DEFAULT_VALUE);
3283 	}
3284 
3285 	/* Enable Receives */
3286 	ew32(RCTL, rctl);
3287 }
3288 
3289 /**
3290  * e1000e_write_mc_addr_list - write multicast addresses to MTA
3291  * @netdev: network interface device structure
3292  *
3293  * Writes multicast address list to the MTA hash table.
3294  * Returns: -ENOMEM on failure
3295  *                0 on no addresses written
3296  *                X on writing X addresses to MTA
3297  */
3298 static int e1000e_write_mc_addr_list(struct net_device *netdev)
3299 {
3300 	struct e1000_adapter *adapter = netdev_priv(netdev);
3301 	struct e1000_hw *hw = &adapter->hw;
3302 	struct netdev_hw_addr *ha;
3303 	u8 *mta_list;
3304 	int i;
3305 
3306 	if (netdev_mc_empty(netdev)) {
3307 		/* nothing to program, so clear mc list */
3308 		hw->mac.ops.update_mc_addr_list(hw, NULL, 0);
3309 		return 0;
3310 	}
3311 
3312 	mta_list = kcalloc(netdev_mc_count(netdev), ETH_ALEN, GFP_ATOMIC);
3313 	if (!mta_list)
3314 		return -ENOMEM;
3315 
3316 	/* update_mc_addr_list expects a packed array of only addresses. */
3317 	i = 0;
3318 	netdev_for_each_mc_addr(ha, netdev)
3319 	    memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3320 
3321 	hw->mac.ops.update_mc_addr_list(hw, mta_list, i);
3322 	kfree(mta_list);
3323 
3324 	return netdev_mc_count(netdev);
3325 }
3326 
3327 /**
3328  * e1000e_write_uc_addr_list - write unicast addresses to RAR table
3329  * @netdev: network interface device structure
3330  *
3331  * Writes unicast address list to the RAR table.
3332  * Returns: -ENOMEM on failure/insufficient address space
3333  *                0 on no addresses written
3334  *                X on writing X addresses to the RAR table
3335  **/
3336 static int e1000e_write_uc_addr_list(struct net_device *netdev)
3337 {
3338 	struct e1000_adapter *adapter = netdev_priv(netdev);
3339 	struct e1000_hw *hw = &adapter->hw;
3340 	unsigned int rar_entries;
3341 	int count = 0;
3342 
3343 	rar_entries = hw->mac.ops.rar_get_count(hw);
3344 
3345 	/* save a rar entry for our hardware address */
3346 	rar_entries--;
3347 
3348 	/* save a rar entry for the LAA workaround */
3349 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA)
3350 		rar_entries--;
3351 
3352 	/* return ENOMEM indicating insufficient memory for addresses */
3353 	if (netdev_uc_count(netdev) > rar_entries)
3354 		return -ENOMEM;
3355 
3356 	if (!netdev_uc_empty(netdev) && rar_entries) {
3357 		struct netdev_hw_addr *ha;
3358 
3359 		/* write the addresses in reverse order to avoid write
3360 		 * combining
3361 		 */
3362 		netdev_for_each_uc_addr(ha, netdev) {
3363 			int ret_val;
3364 
3365 			if (!rar_entries)
3366 				break;
3367 			ret_val = hw->mac.ops.rar_set(hw, ha->addr, rar_entries--);
3368 			if (ret_val < 0)
3369 				return -ENOMEM;
3370 			count++;
3371 		}
3372 	}
3373 
3374 	/* zero out the remaining RAR entries not used above */
3375 	for (; rar_entries > 0; rar_entries--) {
3376 		ew32(RAH(rar_entries), 0);
3377 		ew32(RAL(rar_entries), 0);
3378 	}
3379 	e1e_flush();
3380 
3381 	return count;
3382 }
3383 
3384 /**
3385  * e1000e_set_rx_mode - secondary unicast, Multicast and Promiscuous mode set
3386  * @netdev: network interface device structure
3387  *
3388  * The ndo_set_rx_mode entry point is called whenever the unicast or multicast
3389  * address list or the network interface flags are updated.  This routine is
3390  * responsible for configuring the hardware for proper unicast, multicast,
3391  * promiscuous mode, and all-multi behavior.
3392  **/
3393 static void e1000e_set_rx_mode(struct net_device *netdev)
3394 {
3395 	struct e1000_adapter *adapter = netdev_priv(netdev);
3396 	struct e1000_hw *hw = &adapter->hw;
3397 	u32 rctl;
3398 
3399 	if (pm_runtime_suspended(netdev->dev.parent))
3400 		return;
3401 
3402 	/* Check for Promiscuous and All Multicast modes */
3403 	rctl = er32(RCTL);
3404 
3405 	/* clear the affected bits */
3406 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
3407 
3408 	if (netdev->flags & IFF_PROMISC) {
3409 		rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
3410 		/* Do not hardware filter VLANs in promisc mode */
3411 		e1000e_vlan_filter_disable(adapter);
3412 	} else {
3413 		int count;
3414 
3415 		if (netdev->flags & IFF_ALLMULTI) {
3416 			rctl |= E1000_RCTL_MPE;
3417 		} else {
3418 			/* Write addresses to the MTA, if the attempt fails
3419 			 * then we should just turn on promiscuous mode so
3420 			 * that we can at least receive multicast traffic
3421 			 */
3422 			count = e1000e_write_mc_addr_list(netdev);
3423 			if (count < 0)
3424 				rctl |= E1000_RCTL_MPE;
3425 		}
3426 		e1000e_vlan_filter_enable(adapter);
3427 		/* Write addresses to available RAR registers, if there is not
3428 		 * sufficient space to store all the addresses then enable
3429 		 * unicast promiscuous mode
3430 		 */
3431 		count = e1000e_write_uc_addr_list(netdev);
3432 		if (count < 0)
3433 			rctl |= E1000_RCTL_UPE;
3434 	}
3435 
3436 	ew32(RCTL, rctl);
3437 
3438 	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
3439 		e1000e_vlan_strip_enable(adapter);
3440 	else
3441 		e1000e_vlan_strip_disable(adapter);
3442 }
3443 
3444 static void e1000e_setup_rss_hash(struct e1000_adapter *adapter)
3445 {
3446 	struct e1000_hw *hw = &adapter->hw;
3447 	u32 mrqc, rxcsum;
3448 	u32 rss_key[10];
3449 	int i;
3450 
3451 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3452 	for (i = 0; i < 10; i++)
3453 		ew32(RSSRK(i), rss_key[i]);
3454 
3455 	/* Direct all traffic to queue 0 */
3456 	for (i = 0; i < 32; i++)
3457 		ew32(RETA(i), 0);
3458 
3459 	/* Disable raw packet checksumming so that RSS hash is placed in
3460 	 * descriptor on writeback.
3461 	 */
3462 	rxcsum = er32(RXCSUM);
3463 	rxcsum |= E1000_RXCSUM_PCSD;
3464 
3465 	ew32(RXCSUM, rxcsum);
3466 
3467 	mrqc = (E1000_MRQC_RSS_FIELD_IPV4 |
3468 		E1000_MRQC_RSS_FIELD_IPV4_TCP |
3469 		E1000_MRQC_RSS_FIELD_IPV6 |
3470 		E1000_MRQC_RSS_FIELD_IPV6_TCP |
3471 		E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3472 
3473 	ew32(MRQC, mrqc);
3474 }
3475 
3476 /**
3477  * e1000e_get_base_timinca - get default SYSTIM time increment attributes
3478  * @adapter: board private structure
3479  * @timinca: pointer to returned time increment attributes
3480  *
3481  * Get attributes for incrementing the System Time Register SYSTIML/H at
3482  * the default base frequency, and set the cyclecounter shift value.
3483  **/
3484 s32 e1000e_get_base_timinca(struct e1000_adapter *adapter, u32 *timinca)
3485 {
3486 	struct e1000_hw *hw = &adapter->hw;
3487 	u32 incvalue, incperiod, shift;
3488 
3489 	/* Make sure clock is enabled on I217/I218/I219  before checking
3490 	 * the frequency
3491 	 */
3492 	if ((hw->mac.type >= e1000_pch_lpt) &&
3493 	    !(er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) &&
3494 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_ENABLED)) {
3495 		u32 fextnvm7 = er32(FEXTNVM7);
3496 
3497 		if (!(fextnvm7 & BIT(0))) {
3498 			ew32(FEXTNVM7, fextnvm7 | BIT(0));
3499 			e1e_flush();
3500 		}
3501 	}
3502 
3503 	switch (hw->mac.type) {
3504 	case e1000_pch2lan:
3505 		/* Stable 96MHz frequency */
3506 		incperiod = INCPERIOD_96MHZ;
3507 		incvalue = INCVALUE_96MHZ;
3508 		shift = INCVALUE_SHIFT_96MHZ;
3509 		adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3510 		break;
3511 	case e1000_pch_lpt:
3512 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3513 			/* Stable 96MHz frequency */
3514 			incperiod = INCPERIOD_96MHZ;
3515 			incvalue = INCVALUE_96MHZ;
3516 			shift = INCVALUE_SHIFT_96MHZ;
3517 			adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3518 		} else {
3519 			/* Stable 25MHz frequency */
3520 			incperiod = INCPERIOD_25MHZ;
3521 			incvalue = INCVALUE_25MHZ;
3522 			shift = INCVALUE_SHIFT_25MHZ;
3523 			adapter->cc.shift = shift;
3524 		}
3525 		break;
3526 	case e1000_pch_spt:
3527 		/* Stable 24MHz frequency */
3528 		incperiod = INCPERIOD_24MHZ;
3529 		incvalue = INCVALUE_24MHZ;
3530 		shift = INCVALUE_SHIFT_24MHZ;
3531 		adapter->cc.shift = shift;
3532 		break;
3533 	case e1000_pch_cnp:
3534 	case e1000_pch_tgp:
3535 	case e1000_pch_adp:
3536 	case e1000_pch_nvp:
3537 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3538 			/* Stable 24MHz frequency */
3539 			incperiod = INCPERIOD_24MHZ;
3540 			incvalue = INCVALUE_24MHZ;
3541 			shift = INCVALUE_SHIFT_24MHZ;
3542 			adapter->cc.shift = shift;
3543 		} else {
3544 			/* Stable 38400KHz frequency */
3545 			incperiod = INCPERIOD_38400KHZ;
3546 			incvalue = INCVALUE_38400KHZ;
3547 			shift = INCVALUE_SHIFT_38400KHZ;
3548 			adapter->cc.shift = shift;
3549 		}
3550 		break;
3551 	case e1000_pch_mtp:
3552 	case e1000_pch_lnp:
3553 	case e1000_pch_ptp:
3554 		/* System firmware can misreport this value, so set it to a
3555 		 * stable 38400KHz frequency.
3556 		 */
3557 		incperiod = INCPERIOD_38400KHZ;
3558 		incvalue = INCVALUE_38400KHZ;
3559 		shift = INCVALUE_SHIFT_38400KHZ;
3560 		adapter->cc.shift = shift;
3561 		break;
3562 	case e1000_82574:
3563 	case e1000_82583:
3564 		/* Stable 25MHz frequency */
3565 		incperiod = INCPERIOD_25MHZ;
3566 		incvalue = INCVALUE_25MHZ;
3567 		shift = INCVALUE_SHIFT_25MHZ;
3568 		adapter->cc.shift = shift;
3569 		break;
3570 	default:
3571 		return -EINVAL;
3572 	}
3573 
3574 	*timinca = ((incperiod << E1000_TIMINCA_INCPERIOD_SHIFT) |
3575 		    ((incvalue << shift) & E1000_TIMINCA_INCVALUE_MASK));
3576 
3577 	return 0;
3578 }
3579 
3580 /**
3581  * e1000e_config_hwtstamp - configure the hwtstamp registers and enable/disable
3582  * @adapter: board private structure
3583  * @config: timestamp configuration
3584  * @extack: netlink extended ACK for error report
3585  *
3586  * Outgoing time stamping can be enabled and disabled. Play nice and
3587  * disable it when requested, although it shouldn't cause any overhead
3588  * when no packet needs it. At most one packet in the queue may be
3589  * marked for time stamping, otherwise it would be impossible to tell
3590  * for sure to which packet the hardware time stamp belongs.
3591  *
3592  * Incoming time stamping has to be configured via the hardware filters.
3593  * Not all combinations are supported, in particular event type has to be
3594  * specified. Matching the kind of event packet is not supported, with the
3595  * exception of "all V2 events regardless of level 2 or 4".
3596  **/
3597 static int e1000e_config_hwtstamp(struct e1000_adapter *adapter,
3598 				  struct kernel_hwtstamp_config *config,
3599 				  struct netlink_ext_ack *extack)
3600 {
3601 	struct e1000_hw *hw = &adapter->hw;
3602 	u32 tsync_tx_ctl = E1000_TSYNCTXCTL_ENABLED;
3603 	u32 tsync_rx_ctl = E1000_TSYNCRXCTL_ENABLED;
3604 	u32 rxmtrl = 0;
3605 	u16 rxudp = 0;
3606 	bool is_l4 = false;
3607 	bool is_l2 = false;
3608 	u32 regval;
3609 
3610 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP)) {
3611 		NL_SET_ERR_MSG(extack, "No HW timestamp support");
3612 		return -EINVAL;
3613 	}
3614 
3615 	switch (config->tx_type) {
3616 	case HWTSTAMP_TX_OFF:
3617 		tsync_tx_ctl = 0;
3618 		break;
3619 	case HWTSTAMP_TX_ON:
3620 		break;
3621 	default:
3622 		NL_SET_ERR_MSG(extack, "Unsupported TX HW timestamp type");
3623 		return -ERANGE;
3624 	}
3625 
3626 	switch (config->rx_filter) {
3627 	case HWTSTAMP_FILTER_NONE:
3628 		tsync_rx_ctl = 0;
3629 		break;
3630 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3631 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3632 		rxmtrl = E1000_RXMTRL_PTP_V1_SYNC_MESSAGE;
3633 		is_l4 = true;
3634 		break;
3635 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3636 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3637 		rxmtrl = E1000_RXMTRL_PTP_V1_DELAY_REQ_MESSAGE;
3638 		is_l4 = true;
3639 		break;
3640 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3641 		/* Also time stamps V2 L2 Path Delay Request/Response */
3642 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3643 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3644 		is_l2 = true;
3645 		break;
3646 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3647 		/* Also time stamps V2 L2 Path Delay Request/Response. */
3648 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3649 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3650 		is_l2 = true;
3651 		break;
3652 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3653 		/* Hardware cannot filter just V2 L4 Sync messages */
3654 		fallthrough;
3655 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
3656 		/* Also time stamps V2 Path Delay Request/Response. */
3657 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3658 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3659 		is_l2 = true;
3660 		is_l4 = true;
3661 		break;
3662 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3663 		/* Hardware cannot filter just V2 L4 Delay Request messages */
3664 		fallthrough;
3665 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
3666 		/* Also time stamps V2 Path Delay Request/Response. */
3667 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3668 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3669 		is_l2 = true;
3670 		is_l4 = true;
3671 		break;
3672 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
3673 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
3674 		/* Hardware cannot filter just V2 L4 or L2 Event messages */
3675 		fallthrough;
3676 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
3677 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_EVENT_V2;
3678 		config->rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
3679 		is_l2 = true;
3680 		is_l4 = true;
3681 		break;
3682 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3683 		/* For V1, the hardware can only filter Sync messages or
3684 		 * Delay Request messages but not both so fall-through to
3685 		 * time stamp all packets.
3686 		 */
3687 		fallthrough;
3688 	case HWTSTAMP_FILTER_NTP_ALL:
3689 	case HWTSTAMP_FILTER_ALL:
3690 		is_l2 = true;
3691 		is_l4 = true;
3692 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_ALL;
3693 		config->rx_filter = HWTSTAMP_FILTER_ALL;
3694 		break;
3695 	default:
3696 		NL_SET_ERR_MSG(extack, "Unsupported RX HW timestamp filter");
3697 		return -ERANGE;
3698 	}
3699 
3700 	adapter->hwtstamp_config = *config;
3701 
3702 	/* enable/disable Tx h/w time stamping */
3703 	regval = er32(TSYNCTXCTL);
3704 	regval &= ~E1000_TSYNCTXCTL_ENABLED;
3705 	regval |= tsync_tx_ctl;
3706 	ew32(TSYNCTXCTL, regval);
3707 	if ((er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) !=
3708 	    (regval & E1000_TSYNCTXCTL_ENABLED)) {
3709 		NL_SET_ERR_MSG(extack,
3710 			       "Timesync Tx Control register not set as expected");
3711 		return -EAGAIN;
3712 	}
3713 
3714 	/* enable/disable Rx h/w time stamping */
3715 	regval = er32(TSYNCRXCTL);
3716 	regval &= ~(E1000_TSYNCRXCTL_ENABLED | E1000_TSYNCRXCTL_TYPE_MASK);
3717 	regval |= tsync_rx_ctl;
3718 	ew32(TSYNCRXCTL, regval);
3719 	if ((er32(TSYNCRXCTL) & (E1000_TSYNCRXCTL_ENABLED |
3720 				 E1000_TSYNCRXCTL_TYPE_MASK)) !=
3721 	    (regval & (E1000_TSYNCRXCTL_ENABLED |
3722 		       E1000_TSYNCRXCTL_TYPE_MASK))) {
3723 		NL_SET_ERR_MSG(extack,
3724 			       "Timesync Rx Control register not set as expected");
3725 		return -EAGAIN;
3726 	}
3727 
3728 	/* L2: define ethertype filter for time stamped packets */
3729 	if (is_l2)
3730 		rxmtrl |= ETH_P_1588;
3731 
3732 	/* define which PTP packets get time stamped */
3733 	ew32(RXMTRL, rxmtrl);
3734 
3735 	/* Filter by destination port */
3736 	if (is_l4) {
3737 		rxudp = PTP_EV_PORT;
3738 		cpu_to_be16s(&rxudp);
3739 	}
3740 	ew32(RXUDP, rxudp);
3741 
3742 	e1e_flush();
3743 
3744 	/* Clear TSYNCRXCTL_VALID & TSYNCTXCTL_VALID bit */
3745 	er32(RXSTMPH);
3746 	er32(TXSTMPH);
3747 
3748 	return 0;
3749 }
3750 
3751 /**
3752  * e1000_configure - configure the hardware for Rx and Tx
3753  * @adapter: private board structure
3754  **/
3755 static void e1000_configure(struct e1000_adapter *adapter)
3756 {
3757 	struct e1000_ring *rx_ring = adapter->rx_ring;
3758 
3759 	e1000e_set_rx_mode(adapter->netdev);
3760 
3761 	e1000_restore_vlan(adapter);
3762 	e1000_init_manageability_pt(adapter);
3763 
3764 	e1000_configure_tx(adapter);
3765 
3766 	if (adapter->netdev->features & NETIF_F_RXHASH)
3767 		e1000e_setup_rss_hash(adapter);
3768 	e1000_setup_rctl(adapter);
3769 	e1000_configure_rx(adapter);
3770 	adapter->alloc_rx_buf(rx_ring, e1000_desc_unused(rx_ring), GFP_KERNEL);
3771 }
3772 
3773 /**
3774  * e1000e_power_up_phy - restore link in case the phy was powered down
3775  * @adapter: address of board private structure
3776  *
3777  * The phy may be powered down to save power and turn off link when the
3778  * driver is unloaded and wake on lan is not enabled (among others)
3779  * *** this routine MUST be followed by a call to e1000e_reset ***
3780  **/
3781 void e1000e_power_up_phy(struct e1000_adapter *adapter)
3782 {
3783 	if (adapter->hw.phy.ops.power_up)
3784 		adapter->hw.phy.ops.power_up(&adapter->hw);
3785 
3786 	adapter->hw.mac.ops.setup_link(&adapter->hw);
3787 }
3788 
3789 /**
3790  * e1000_power_down_phy - Power down the PHY
3791  * @adapter: board private structure
3792  *
3793  * Power down the PHY so no link is implied when interface is down.
3794  * The PHY cannot be powered down if management or WoL is active.
3795  */
3796 static void e1000_power_down_phy(struct e1000_adapter *adapter)
3797 {
3798 	if (adapter->hw.phy.ops.power_down)
3799 		adapter->hw.phy.ops.power_down(&adapter->hw);
3800 }
3801 
3802 /**
3803  * e1000_flush_tx_ring - remove all descriptors from the tx_ring
3804  * @adapter: board private structure
3805  *
3806  * We want to clear all pending descriptors from the TX ring.
3807  * zeroing happens when the HW reads the regs. We  assign the ring itself as
3808  * the data of the next descriptor. We don't care about the data we are about
3809  * to reset the HW.
3810  */
3811 static void e1000_flush_tx_ring(struct e1000_adapter *adapter)
3812 {
3813 	struct e1000_hw *hw = &adapter->hw;
3814 	struct e1000_ring *tx_ring = adapter->tx_ring;
3815 	struct e1000_tx_desc *tx_desc = NULL;
3816 	u32 tdt, tctl, txd_lower = E1000_TXD_CMD_IFCS;
3817 	u16 size = 512;
3818 
3819 	tctl = er32(TCTL);
3820 	ew32(TCTL, tctl | E1000_TCTL_EN);
3821 	tdt = er32(TDT(0));
3822 	BUG_ON(tdt != tx_ring->next_to_use);
3823 	tx_desc =  E1000_TX_DESC(*tx_ring, tx_ring->next_to_use);
3824 	tx_desc->buffer_addr = cpu_to_le64(tx_ring->dma);
3825 
3826 	tx_desc->lower.data = cpu_to_le32(txd_lower | size);
3827 	tx_desc->upper.data = 0;
3828 	/* flush descriptors to memory before notifying the HW */
3829 	wmb();
3830 	tx_ring->next_to_use++;
3831 	if (tx_ring->next_to_use == tx_ring->count)
3832 		tx_ring->next_to_use = 0;
3833 	ew32(TDT(0), tx_ring->next_to_use);
3834 	usleep_range(200, 250);
3835 }
3836 
3837 /**
3838  * e1000_flush_rx_ring - remove all descriptors from the rx_ring
3839  * @adapter: board private structure
3840  *
3841  * Mark all descriptors in the RX ring as consumed and disable the rx ring
3842  */
3843 static void e1000_flush_rx_ring(struct e1000_adapter *adapter)
3844 {
3845 	u32 rctl, rxdctl;
3846 	struct e1000_hw *hw = &adapter->hw;
3847 
3848 	rctl = er32(RCTL);
3849 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3850 	e1e_flush();
3851 	usleep_range(100, 150);
3852 
3853 	rxdctl = er32(RXDCTL(0));
3854 	/* zero the lower 14 bits (prefetch and host thresholds) */
3855 	rxdctl &= 0xffffc000;
3856 
3857 	/* update thresholds: prefetch threshold to 31, host threshold to 1
3858 	 * and make sure the granularity is "descriptors" and not "cache lines"
3859 	 */
3860 	rxdctl |= (0x1F | BIT(8) | E1000_RXDCTL_THRESH_UNIT_DESC);
3861 
3862 	ew32(RXDCTL(0), rxdctl);
3863 	/* momentarily enable the RX ring for the changes to take effect */
3864 	ew32(RCTL, rctl | E1000_RCTL_EN);
3865 	e1e_flush();
3866 	usleep_range(100, 150);
3867 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3868 }
3869 
3870 /**
3871  * e1000_flush_desc_rings - remove all descriptors from the descriptor rings
3872  * @adapter: board private structure
3873  *
3874  * In i219, the descriptor rings must be emptied before resetting the HW
3875  * or before changing the device state to D3 during runtime (runtime PM).
3876  *
3877  * Failure to do this will cause the HW to enter a unit hang state which can
3878  * only be released by PCI reset on the device
3879  *
3880  */
3881 
3882 static void e1000_flush_desc_rings(struct e1000_adapter *adapter)
3883 {
3884 	u16 hang_state;
3885 	u32 fext_nvm11, tdlen;
3886 	struct e1000_hw *hw = &adapter->hw;
3887 
3888 	/* First, disable MULR fix in FEXTNVM11 */
3889 	fext_nvm11 = er32(FEXTNVM11);
3890 	fext_nvm11 |= E1000_FEXTNVM11_DISABLE_MULR_FIX;
3891 	ew32(FEXTNVM11, fext_nvm11);
3892 	/* do nothing if we're not in faulty state, or if the queue is empty */
3893 	tdlen = er32(TDLEN(0));
3894 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3895 			     &hang_state);
3896 	if (!(hang_state & FLUSH_DESC_REQUIRED) || !tdlen)
3897 		return;
3898 	e1000_flush_tx_ring(adapter);
3899 	/* recheck, maybe the fault is caused by the rx ring */
3900 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3901 			     &hang_state);
3902 	if (hang_state & FLUSH_DESC_REQUIRED)
3903 		e1000_flush_rx_ring(adapter);
3904 }
3905 
3906 /**
3907  * e1000e_systim_reset - reset the timesync registers after a hardware reset
3908  * @adapter: board private structure
3909  *
3910  * When the MAC is reset, all hardware bits for timesync will be reset to the
3911  * default values. This function will restore the settings last in place.
3912  * Since the clock SYSTIME registers are reset, we will simply restore the
3913  * cyclecounter to the kernel real clock time.
3914  **/
3915 static void e1000e_systim_reset(struct e1000_adapter *adapter)
3916 {
3917 	struct ptp_clock_info *info = &adapter->ptp_clock_info;
3918 	struct e1000_hw *hw = &adapter->hw;
3919 	struct netlink_ext_ack extack = {};
3920 	unsigned long flags;
3921 	u32 timinca;
3922 	s32 ret_val;
3923 
3924 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3925 		return;
3926 
3927 	if (info->adjfine) {
3928 		/* restore the previous ptp frequency delta */
3929 		ret_val = info->adjfine(info, adapter->ptp_delta);
3930 	} else {
3931 		/* set the default base frequency if no adjustment possible */
3932 		ret_val = e1000e_get_base_timinca(adapter, &timinca);
3933 		if (!ret_val)
3934 			ew32(TIMINCA, timinca);
3935 	}
3936 
3937 	if (ret_val) {
3938 		dev_warn(&adapter->pdev->dev,
3939 			 "Failed to restore TIMINCA clock rate delta: %d\n",
3940 			 ret_val);
3941 		return;
3942 	}
3943 
3944 	/* reset the systim ns time counter */
3945 	spin_lock_irqsave(&adapter->systim_lock, flags);
3946 	timecounter_init(&adapter->tc, &adapter->cc,
3947 			 ktime_to_ns(ktime_get_real()));
3948 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
3949 
3950 	/* restore the previous hwtstamp configuration settings */
3951 	ret_val = e1000e_config_hwtstamp(adapter, &adapter->hwtstamp_config,
3952 					 &extack);
3953 	if (ret_val) {
3954 		if (extack._msg)
3955 			e_err("%s\n", extack._msg);
3956 	}
3957 }
3958 
3959 /**
3960  * e1000e_reset - bring the hardware into a known good state
3961  * @adapter: board private structure
3962  *
3963  * This function boots the hardware and enables some settings that
3964  * require a configuration cycle of the hardware - those cannot be
3965  * set/changed during runtime. After reset the device needs to be
3966  * properly configured for Rx, Tx etc.
3967  */
3968 void e1000e_reset(struct e1000_adapter *adapter)
3969 {
3970 	struct e1000_mac_info *mac = &adapter->hw.mac;
3971 	struct e1000_fc_info *fc = &adapter->hw.fc;
3972 	struct e1000_hw *hw = &adapter->hw;
3973 	u32 tx_space, min_tx_space, min_rx_space;
3974 	u32 pba = adapter->pba;
3975 	u16 hwm;
3976 
3977 	/* reset Packet Buffer Allocation to default */
3978 	ew32(PBA, pba);
3979 
3980 	if (adapter->max_frame_size > (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN)) {
3981 		/* To maintain wire speed transmits, the Tx FIFO should be
3982 		 * large enough to accommodate two full transmit packets,
3983 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
3984 		 * the Rx FIFO should be large enough to accommodate at least
3985 		 * one full receive packet and is similarly rounded up and
3986 		 * expressed in KB.
3987 		 */
3988 		pba = er32(PBA);
3989 		/* upper 16 bits has Tx packet buffer allocation size in KB */
3990 		tx_space = pba >> 16;
3991 		/* lower 16 bits has Rx packet buffer allocation size in KB */
3992 		pba &= 0xffff;
3993 		/* the Tx fifo also stores 16 bytes of information about the Tx
3994 		 * but don't include ethernet FCS because hardware appends it
3995 		 */
3996 		min_tx_space = (adapter->max_frame_size +
3997 				sizeof(struct e1000_tx_desc) - ETH_FCS_LEN) * 2;
3998 		min_tx_space = ALIGN(min_tx_space, 1024);
3999 		min_tx_space >>= 10;
4000 		/* software strips receive CRC, so leave room for it */
4001 		min_rx_space = adapter->max_frame_size;
4002 		min_rx_space = ALIGN(min_rx_space, 1024);
4003 		min_rx_space >>= 10;
4004 
4005 		/* If current Tx allocation is less than the min Tx FIFO size,
4006 		 * and the min Tx FIFO size is less than the current Rx FIFO
4007 		 * allocation, take space away from current Rx allocation
4008 		 */
4009 		if ((tx_space < min_tx_space) &&
4010 		    ((min_tx_space - tx_space) < pba)) {
4011 			pba -= min_tx_space - tx_space;
4012 
4013 			/* if short on Rx space, Rx wins and must trump Tx
4014 			 * adjustment
4015 			 */
4016 			if (pba < min_rx_space)
4017 				pba = min_rx_space;
4018 		}
4019 
4020 		ew32(PBA, pba);
4021 	}
4022 
4023 	/* flow control settings
4024 	 *
4025 	 * The high water mark must be low enough to fit one full frame
4026 	 * (or the size used for early receive) above it in the Rx FIFO.
4027 	 * Set it to the lower of:
4028 	 * - 90% of the Rx FIFO size, and
4029 	 * - the full Rx FIFO size minus one full frame
4030 	 */
4031 	if (adapter->flags & FLAG_DISABLE_FC_PAUSE_TIME)
4032 		fc->pause_time = 0xFFFF;
4033 	else
4034 		fc->pause_time = E1000_FC_PAUSE_TIME;
4035 	fc->send_xon = true;
4036 	fc->current_mode = fc->requested_mode;
4037 
4038 	switch (hw->mac.type) {
4039 	case e1000_ich9lan:
4040 	case e1000_ich10lan:
4041 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4042 			pba = 14;
4043 			ew32(PBA, pba);
4044 			fc->high_water = 0x2800;
4045 			fc->low_water = fc->high_water - 8;
4046 			break;
4047 		}
4048 		fallthrough;
4049 	default:
4050 		hwm = min(((pba << 10) * 9 / 10),
4051 			  ((pba << 10) - adapter->max_frame_size));
4052 
4053 		fc->high_water = hwm & E1000_FCRTH_RTH;	/* 8-byte granularity */
4054 		fc->low_water = fc->high_water - 8;
4055 		break;
4056 	case e1000_pchlan:
4057 		/* Workaround PCH LOM adapter hangs with certain network
4058 		 * loads.  If hangs persist, try disabling Tx flow control.
4059 		 */
4060 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4061 			fc->high_water = 0x3500;
4062 			fc->low_water = 0x1500;
4063 		} else {
4064 			fc->high_water = 0x5000;
4065 			fc->low_water = 0x3000;
4066 		}
4067 		fc->refresh_time = 0x1000;
4068 		break;
4069 	case e1000_pch2lan:
4070 	case e1000_pch_lpt:
4071 	case e1000_pch_spt:
4072 	case e1000_pch_cnp:
4073 	case e1000_pch_tgp:
4074 	case e1000_pch_adp:
4075 	case e1000_pch_mtp:
4076 	case e1000_pch_lnp:
4077 	case e1000_pch_ptp:
4078 	case e1000_pch_nvp:
4079 		fc->refresh_time = 0xFFFF;
4080 		fc->pause_time = 0xFFFF;
4081 
4082 		if (adapter->netdev->mtu <= ETH_DATA_LEN) {
4083 			fc->high_water = 0x05C20;
4084 			fc->low_water = 0x05048;
4085 			break;
4086 		}
4087 
4088 		pba = 14;
4089 		ew32(PBA, pba);
4090 		fc->high_water = ((pba << 10) * 9 / 10) & E1000_FCRTH_RTH;
4091 		fc->low_water = ((pba << 10) * 8 / 10) & E1000_FCRTL_RTL;
4092 		break;
4093 	}
4094 
4095 	/* Alignment of Tx data is on an arbitrary byte boundary with the
4096 	 * maximum size per Tx descriptor limited only to the transmit
4097 	 * allocation of the packet buffer minus 96 bytes with an upper
4098 	 * limit of 24KB due to receive synchronization limitations.
4099 	 */
4100 	adapter->tx_fifo_limit = min_t(u32, ((er32(PBA) >> 16) << 10) - 96,
4101 				       24 << 10);
4102 
4103 	/* Disable Adaptive Interrupt Moderation if 2 full packets cannot
4104 	 * fit in receive buffer.
4105 	 */
4106 	if (adapter->itr_setting & 0x3) {
4107 		if ((adapter->max_frame_size * 2) > (pba << 10)) {
4108 			if (!(adapter->flags2 & FLAG2_DISABLE_AIM)) {
4109 				dev_info(&adapter->pdev->dev,
4110 					 "Interrupt Throttle Rate off\n");
4111 				adapter->flags2 |= FLAG2_DISABLE_AIM;
4112 				e1000e_write_itr(adapter, 0);
4113 			}
4114 		} else if (adapter->flags2 & FLAG2_DISABLE_AIM) {
4115 			dev_info(&adapter->pdev->dev,
4116 				 "Interrupt Throttle Rate on\n");
4117 			adapter->flags2 &= ~FLAG2_DISABLE_AIM;
4118 			adapter->itr = 20000;
4119 			e1000e_write_itr(adapter, adapter->itr);
4120 		}
4121 	}
4122 
4123 	if (hw->mac.type >= e1000_pch_spt)
4124 		e1000_flush_desc_rings(adapter);
4125 	/* Allow time for pending master requests to run */
4126 	mac->ops.reset_hw(hw);
4127 
4128 	/* For parts with AMT enabled, let the firmware know
4129 	 * that the network interface is in control
4130 	 */
4131 	if (adapter->flags & FLAG_HAS_AMT)
4132 		e1000e_get_hw_control(adapter);
4133 
4134 	ew32(WUC, 0);
4135 
4136 	if (mac->ops.init_hw(hw))
4137 		e_err("Hardware Error\n");
4138 
4139 	e1000_update_mng_vlan(adapter);
4140 
4141 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
4142 	ew32(VET, ETH_P_8021Q);
4143 
4144 	e1000e_reset_adaptive(hw);
4145 
4146 	/* restore systim and hwtstamp settings */
4147 	e1000e_systim_reset(adapter);
4148 
4149 	/* Set EEE advertisement as appropriate */
4150 	if (adapter->flags2 & FLAG2_HAS_EEE) {
4151 		s32 ret_val;
4152 		u16 adv_addr;
4153 
4154 		switch (hw->phy.type) {
4155 		case e1000_phy_82579:
4156 			adv_addr = I82579_EEE_ADVERTISEMENT;
4157 			break;
4158 		case e1000_phy_i217:
4159 			adv_addr = I217_EEE_ADVERTISEMENT;
4160 			break;
4161 		default:
4162 			dev_err(&adapter->pdev->dev,
4163 				"Invalid PHY type setting EEE advertisement\n");
4164 			return;
4165 		}
4166 
4167 		ret_val = hw->phy.ops.acquire(hw);
4168 		if (ret_val) {
4169 			dev_err(&adapter->pdev->dev,
4170 				"EEE advertisement - unable to acquire PHY\n");
4171 			return;
4172 		}
4173 
4174 		e1000_write_emi_reg_locked(hw, adv_addr,
4175 					   hw->dev_spec.ich8lan.eee_disable ?
4176 					   0 : adapter->eee_advert);
4177 
4178 		hw->phy.ops.release(hw);
4179 	}
4180 
4181 	if (!netif_running(adapter->netdev) &&
4182 	    !test_bit(__E1000_TESTING, &adapter->state))
4183 		e1000_power_down_phy(adapter);
4184 
4185 	e1000_get_phy_info(hw);
4186 
4187 	if ((adapter->flags & FLAG_HAS_SMART_POWER_DOWN) &&
4188 	    !(adapter->flags & FLAG_SMART_POWER_DOWN)) {
4189 		u16 phy_data = 0;
4190 		/* speed up time to link by disabling smart power down, ignore
4191 		 * the return value of this function because there is nothing
4192 		 * different we would do if it failed
4193 		 */
4194 		e1e_rphy(hw, IGP02E1000_PHY_POWER_MGMT, &phy_data);
4195 		phy_data &= ~IGP02E1000_PM_SPD;
4196 		e1e_wphy(hw, IGP02E1000_PHY_POWER_MGMT, phy_data);
4197 	}
4198 	if (hw->mac.type >= e1000_pch_spt && adapter->int_mode == 0) {
4199 		u32 reg;
4200 
4201 		/* Fextnvm7 @ 0xe4[2] = 1 */
4202 		reg = er32(FEXTNVM7);
4203 		reg |= E1000_FEXTNVM7_SIDE_CLK_UNGATE;
4204 		ew32(FEXTNVM7, reg);
4205 		/* Fextnvm9 @ 0x5bb4[13:12] = 11 */
4206 		reg = er32(FEXTNVM9);
4207 		reg |= E1000_FEXTNVM9_IOSFSB_CLKGATE_DIS |
4208 		       E1000_FEXTNVM9_IOSFSB_CLKREQ_DIS;
4209 		ew32(FEXTNVM9, reg);
4210 	}
4211 
4212 }
4213 
4214 /**
4215  * e1000e_trigger_lsc - trigger an LSC interrupt
4216  * @adapter: board private structure
4217  *
4218  * Fire a link status change interrupt to start the watchdog.
4219  **/
4220 static void e1000e_trigger_lsc(struct e1000_adapter *adapter)
4221 {
4222 	struct e1000_hw *hw = &adapter->hw;
4223 
4224 	if (adapter->msix_entries)
4225 		ew32(ICS, E1000_ICS_LSC | E1000_ICS_OTHER);
4226 	else
4227 		ew32(ICS, E1000_ICS_LSC);
4228 }
4229 
4230 void e1000e_up(struct e1000_adapter *adapter)
4231 {
4232 	/* hardware has been reset, we need to reload some things */
4233 	e1000_configure(adapter);
4234 
4235 	clear_bit(__E1000_DOWN, &adapter->state);
4236 
4237 	if (adapter->msix_entries)
4238 		e1000_configure_msix(adapter);
4239 	e1000_irq_enable(adapter);
4240 
4241 	/* Tx queue started by watchdog timer when link is up */
4242 
4243 	e1000e_trigger_lsc(adapter);
4244 }
4245 
4246 static void e1000e_flush_descriptors(struct e1000_adapter *adapter)
4247 {
4248 	struct e1000_hw *hw = &adapter->hw;
4249 
4250 	if (!(adapter->flags2 & FLAG2_DMA_BURST))
4251 		return;
4252 
4253 	/* flush pending descriptor writebacks to memory */
4254 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4255 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4256 
4257 	/* execute the writes immediately */
4258 	e1e_flush();
4259 
4260 	/* due to rare timing issues, write to TIDV/RDTR again to ensure the
4261 	 * write is successful
4262 	 */
4263 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4264 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4265 
4266 	/* execute the writes immediately */
4267 	e1e_flush();
4268 }
4269 
4270 static void e1000e_update_stats(struct e1000_adapter *adapter);
4271 
4272 /**
4273  * e1000e_down - quiesce the device and optionally reset the hardware
4274  * @adapter: board private structure
4275  * @reset: boolean flag to reset the hardware or not
4276  */
4277 void e1000e_down(struct e1000_adapter *adapter, bool reset)
4278 {
4279 	struct net_device *netdev = adapter->netdev;
4280 	struct e1000_hw *hw = &adapter->hw;
4281 	u32 tctl, rctl;
4282 
4283 	/* signal that we're down so the interrupt handler does not
4284 	 * reschedule our watchdog timer
4285 	 */
4286 	set_bit(__E1000_DOWN, &adapter->state);
4287 
4288 	netif_carrier_off(netdev);
4289 
4290 	/* disable receives in the hardware */
4291 	rctl = er32(RCTL);
4292 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
4293 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
4294 	/* flush and sleep below */
4295 
4296 	netif_stop_queue(netdev);
4297 
4298 	/* disable transmits in the hardware */
4299 	tctl = er32(TCTL);
4300 	tctl &= ~E1000_TCTL_EN;
4301 	ew32(TCTL, tctl);
4302 
4303 	/* flush both disables and wait for them to finish */
4304 	e1e_flush();
4305 	usleep_range(10000, 11000);
4306 
4307 	e1000_irq_disable(adapter);
4308 
4309 	napi_synchronize(&adapter->napi);
4310 
4311 	timer_delete_sync(&adapter->watchdog_timer);
4312 	timer_delete_sync(&adapter->phy_info_timer);
4313 
4314 	spin_lock(&adapter->stats64_lock);
4315 	e1000e_update_stats(adapter);
4316 	spin_unlock(&adapter->stats64_lock);
4317 
4318 	e1000e_flush_descriptors(adapter);
4319 
4320 	adapter->link_speed = 0;
4321 	adapter->link_duplex = 0;
4322 
4323 	/* Disable Si errata workaround on PCHx for jumbo frame flow */
4324 	if ((hw->mac.type >= e1000_pch2lan) &&
4325 	    (adapter->netdev->mtu > ETH_DATA_LEN) &&
4326 	    e1000_lv_jumbo_workaround_ich8lan(hw, false))
4327 		e_dbg("failed to disable jumbo frame workaround mode\n");
4328 
4329 	if (!pci_channel_offline(adapter->pdev)) {
4330 		if (reset)
4331 			e1000e_reset(adapter);
4332 		else if (hw->mac.type >= e1000_pch_spt)
4333 			e1000_flush_desc_rings(adapter);
4334 	}
4335 	e1000_clean_tx_ring(adapter->tx_ring);
4336 	e1000_clean_rx_ring(adapter->rx_ring);
4337 }
4338 
4339 void e1000e_reinit_locked(struct e1000_adapter *adapter)
4340 {
4341 	might_sleep();
4342 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
4343 		usleep_range(1000, 1100);
4344 	e1000e_down(adapter, true);
4345 	e1000e_up(adapter);
4346 	clear_bit(__E1000_RESETTING, &adapter->state);
4347 }
4348 
4349 /**
4350  * e1000e_sanitize_systim - sanitize raw cycle counter reads
4351  * @hw: pointer to the HW structure
4352  * @systim: PHC time value read, sanitized and returned
4353  * @sts: structure to hold system time before and after reading SYSTIML,
4354  * may be NULL
4355  *
4356  * Errata for 82574/82583 possible bad bits read from SYSTIMH/L:
4357  * check to see that the time is incrementing at a reasonable
4358  * rate and is a multiple of incvalue.
4359  **/
4360 static u64 e1000e_sanitize_systim(struct e1000_hw *hw, u64 systim,
4361 				  struct ptp_system_timestamp *sts)
4362 {
4363 	u64 time_delta, rem, temp;
4364 	u64 systim_next;
4365 	u32 incvalue;
4366 	int i;
4367 
4368 	incvalue = er32(TIMINCA) & E1000_TIMINCA_INCVALUE_MASK;
4369 	for (i = 0; i < E1000_MAX_82574_SYSTIM_REREADS; i++) {
4370 		/* latch SYSTIMH on read of SYSTIML */
4371 		ptp_read_system_prets(sts);
4372 		systim_next = (u64)er32(SYSTIML);
4373 		ptp_read_system_postts(sts);
4374 		systim_next |= (u64)er32(SYSTIMH) << 32;
4375 
4376 		time_delta = systim_next - systim;
4377 		temp = time_delta;
4378 		/* VMWare users have seen incvalue of zero, don't div / 0 */
4379 		rem = incvalue ? do_div(temp, incvalue) : (time_delta != 0);
4380 
4381 		systim = systim_next;
4382 
4383 		if ((time_delta < E1000_82574_SYSTIM_EPSILON) && (rem == 0))
4384 			break;
4385 	}
4386 
4387 	return systim;
4388 }
4389 
4390 /**
4391  * e1000e_read_systim - read SYSTIM register
4392  * @adapter: board private structure
4393  * @sts: structure which will contain system time before and after reading
4394  * SYSTIML, may be NULL
4395  **/
4396 u64 e1000e_read_systim(struct e1000_adapter *adapter,
4397 		       struct ptp_system_timestamp *sts)
4398 {
4399 	struct e1000_hw *hw = &adapter->hw;
4400 	u32 systimel, systimel_2, systimeh;
4401 	u64 systim;
4402 	/* SYSTIMH latching upon SYSTIML read does not work well.
4403 	 * This means that if SYSTIML overflows after we read it but before
4404 	 * we read SYSTIMH, the value of SYSTIMH has been incremented and we
4405 	 * will experience a huge non linear increment in the systime value
4406 	 * to fix that we test for overflow and if true, we re-read systime.
4407 	 */
4408 	ptp_read_system_prets(sts);
4409 	systimel = er32(SYSTIML);
4410 	ptp_read_system_postts(sts);
4411 	systimeh = er32(SYSTIMH);
4412 	/* Is systimel is so large that overflow is possible? */
4413 	if (systimel >= (u32)0xffffffff - E1000_TIMINCA_INCVALUE_MASK) {
4414 		ptp_read_system_prets(sts);
4415 		systimel_2 = er32(SYSTIML);
4416 		ptp_read_system_postts(sts);
4417 		if (systimel > systimel_2) {
4418 			/* There was an overflow, read again SYSTIMH, and use
4419 			 * systimel_2
4420 			 */
4421 			systimeh = er32(SYSTIMH);
4422 			systimel = systimel_2;
4423 		}
4424 	}
4425 	systim = (u64)systimel;
4426 	systim |= (u64)systimeh << 32;
4427 
4428 	if (adapter->flags2 & FLAG2_CHECK_SYSTIM_OVERFLOW)
4429 		systim = e1000e_sanitize_systim(hw, systim, sts);
4430 
4431 	return systim;
4432 }
4433 
4434 /**
4435  * e1000e_cyclecounter_read - read raw cycle counter (used by time counter)
4436  * @cc: cyclecounter structure
4437  **/
4438 static u64 e1000e_cyclecounter_read(struct cyclecounter *cc)
4439 {
4440 	struct e1000_adapter *adapter = container_of(cc, struct e1000_adapter,
4441 						     cc);
4442 
4443 	return e1000e_read_systim(adapter, NULL);
4444 }
4445 
4446 /**
4447  * e1000_sw_init - Initialize general software structures (struct e1000_adapter)
4448  * @adapter: board private structure to initialize
4449  *
4450  * e1000_sw_init initializes the Adapter private data structure.
4451  * Fields are initialized based on PCI device information and
4452  * OS network device settings (MTU size).
4453  **/
4454 static int e1000_sw_init(struct e1000_adapter *adapter)
4455 {
4456 	struct net_device *netdev = adapter->netdev;
4457 
4458 	adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
4459 	adapter->rx_ps_bsize0 = 128;
4460 	adapter->max_frame_size = netdev->mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
4461 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4462 	adapter->tx_ring_count = E1000_DEFAULT_TXD;
4463 	adapter->rx_ring_count = E1000_DEFAULT_RXD;
4464 
4465 	spin_lock_init(&adapter->stats64_lock);
4466 
4467 	e1000e_set_interrupt_capability(adapter);
4468 
4469 	if (e1000_alloc_queues(adapter))
4470 		return -ENOMEM;
4471 
4472 	/* Setup hardware time stamping cyclecounter */
4473 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
4474 		adapter->cc.read = e1000e_cyclecounter_read;
4475 		adapter->cc.mask = CYCLECOUNTER_MASK(64);
4476 		adapter->cc.mult = 1;
4477 		/* cc.shift set in e1000e_get_base_tininca() */
4478 
4479 		spin_lock_init(&adapter->systim_lock);
4480 		INIT_WORK(&adapter->tx_hwtstamp_work, e1000e_tx_hwtstamp_work);
4481 	}
4482 
4483 	/* Explicitly disable IRQ since the NIC can be in any state. */
4484 	e1000_irq_disable(adapter);
4485 
4486 	set_bit(__E1000_DOWN, &adapter->state);
4487 	return 0;
4488 }
4489 
4490 /**
4491  * e1000_intr_msi_test - Interrupt Handler
4492  * @irq: interrupt number
4493  * @data: pointer to a network interface device structure
4494  **/
4495 static irqreturn_t e1000_intr_msi_test(int __always_unused irq, void *data)
4496 {
4497 	struct net_device *netdev = data;
4498 	struct e1000_adapter *adapter = netdev_priv(netdev);
4499 	struct e1000_hw *hw = &adapter->hw;
4500 	u32 icr = er32(ICR);
4501 
4502 	e_dbg("icr is %08X\n", icr);
4503 	if (icr & E1000_ICR_RXSEQ) {
4504 		adapter->flags &= ~FLAG_MSI_TEST_FAILED;
4505 		/* Force memory writes to complete before acknowledging the
4506 		 * interrupt is handled.
4507 		 */
4508 		wmb();
4509 	}
4510 
4511 	return IRQ_HANDLED;
4512 }
4513 
4514 /**
4515  * e1000_test_msi_interrupt - Returns 0 for successful test
4516  * @adapter: board private struct
4517  *
4518  * code flow taken from tg3.c
4519  **/
4520 static int e1000_test_msi_interrupt(struct e1000_adapter *adapter)
4521 {
4522 	struct net_device *netdev = adapter->netdev;
4523 	struct e1000_hw *hw = &adapter->hw;
4524 	int err;
4525 
4526 	/* poll_enable hasn't been called yet, so don't need disable */
4527 	/* clear any pending events */
4528 	er32(ICR);
4529 
4530 	/* free the real vector and request a test handler */
4531 	e1000_free_irq(adapter);
4532 	e1000e_reset_interrupt_capability(adapter);
4533 
4534 	/* Assume that the test fails, if it succeeds then the test
4535 	 * MSI irq handler will unset this flag
4536 	 */
4537 	adapter->flags |= FLAG_MSI_TEST_FAILED;
4538 
4539 	err = pci_enable_msi(adapter->pdev);
4540 	if (err)
4541 		goto msi_test_failed;
4542 
4543 	err = request_irq(adapter->pdev->irq, e1000_intr_msi_test, 0,
4544 			  netdev->name, netdev);
4545 	if (err) {
4546 		pci_disable_msi(adapter->pdev);
4547 		goto msi_test_failed;
4548 	}
4549 
4550 	/* Force memory writes to complete before enabling and firing an
4551 	 * interrupt.
4552 	 */
4553 	wmb();
4554 
4555 	e1000_irq_enable(adapter);
4556 
4557 	/* fire an unusual interrupt on the test handler */
4558 	ew32(ICS, E1000_ICS_RXSEQ);
4559 	e1e_flush();
4560 	msleep(100);
4561 
4562 	e1000_irq_disable(adapter);
4563 
4564 	rmb();			/* read flags after interrupt has been fired */
4565 
4566 	if (adapter->flags & FLAG_MSI_TEST_FAILED) {
4567 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
4568 		e_info("MSI interrupt test failed, using legacy interrupt.\n");
4569 	} else {
4570 		e_dbg("MSI interrupt test succeeded!\n");
4571 	}
4572 
4573 	free_irq(adapter->pdev->irq, netdev);
4574 	pci_disable_msi(adapter->pdev);
4575 
4576 msi_test_failed:
4577 	e1000e_set_interrupt_capability(adapter);
4578 	return e1000_request_irq(adapter);
4579 }
4580 
4581 /**
4582  * e1000_test_msi - Returns 0 if MSI test succeeds or INTx mode is restored
4583  * @adapter: board private struct
4584  *
4585  * code flow taken from tg3.c, called with e1000 interrupts disabled.
4586  **/
4587 static int e1000_test_msi(struct e1000_adapter *adapter)
4588 {
4589 	int err;
4590 	u16 pci_cmd;
4591 
4592 	if (!(adapter->flags & FLAG_MSI_ENABLED))
4593 		return 0;
4594 
4595 	/* disable SERR in case the MSI write causes a master abort */
4596 	pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4597 	if (pci_cmd & PCI_COMMAND_SERR)
4598 		pci_write_config_word(adapter->pdev, PCI_COMMAND,
4599 				      pci_cmd & ~PCI_COMMAND_SERR);
4600 
4601 	err = e1000_test_msi_interrupt(adapter);
4602 
4603 	/* re-enable SERR */
4604 	if (pci_cmd & PCI_COMMAND_SERR) {
4605 		pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4606 		pci_cmd |= PCI_COMMAND_SERR;
4607 		pci_write_config_word(adapter->pdev, PCI_COMMAND, pci_cmd);
4608 	}
4609 
4610 	return err;
4611 }
4612 
4613 /**
4614  * e1000e_open - Called when a network interface is made active
4615  * @netdev: network interface device structure
4616  *
4617  * Returns 0 on success, negative value on failure
4618  *
4619  * The open entry point is called when a network interface is made
4620  * active by the system (IFF_UP).  At this point all resources needed
4621  * for transmit and receive operations are allocated, the interrupt
4622  * handler is registered with the OS, the watchdog timer is started,
4623  * and the stack is notified that the interface is ready.
4624  **/
4625 int e1000e_open(struct net_device *netdev)
4626 {
4627 	struct e1000_adapter *adapter = netdev_priv(netdev);
4628 	struct e1000_hw *hw = &adapter->hw;
4629 	struct pci_dev *pdev = adapter->pdev;
4630 	int err;
4631 	int irq;
4632 
4633 	/* disallow open during test */
4634 	if (test_bit(__E1000_TESTING, &adapter->state))
4635 		return -EBUSY;
4636 
4637 	pm_runtime_get_sync(&pdev->dev);
4638 
4639 	netif_carrier_off(netdev);
4640 	netif_stop_queue(netdev);
4641 
4642 	/* allocate transmit descriptors */
4643 	err = e1000e_setup_tx_resources(adapter->tx_ring);
4644 	if (err)
4645 		goto err_setup_tx;
4646 
4647 	/* allocate receive descriptors */
4648 	err = e1000e_setup_rx_resources(adapter->rx_ring);
4649 	if (err)
4650 		goto err_setup_rx;
4651 
4652 	/* If AMT is enabled, let the firmware know that the network
4653 	 * interface is now open and reset the part to a known state.
4654 	 */
4655 	if (adapter->flags & FLAG_HAS_AMT) {
4656 		e1000e_get_hw_control(adapter);
4657 		e1000e_reset(adapter);
4658 	}
4659 
4660 	e1000e_power_up_phy(adapter);
4661 
4662 	adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
4663 	if ((adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
4664 		e1000_update_mng_vlan(adapter);
4665 
4666 	/* DMA latency requirement to workaround jumbo issue */
4667 	cpu_latency_qos_add_request(&adapter->pm_qos_req, PM_QOS_DEFAULT_VALUE);
4668 
4669 	/* before we allocate an interrupt, we must be ready to handle it.
4670 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4671 	 * as soon as we call pci_request_irq, so we have to setup our
4672 	 * clean_rx handler before we do so.
4673 	 */
4674 	e1000_configure(adapter);
4675 
4676 	err = e1000_request_irq(adapter);
4677 	if (err)
4678 		goto err_req_irq;
4679 
4680 	/* Work around PCIe errata with MSI interrupts causing some chipsets to
4681 	 * ignore e1000e MSI messages, which means we need to test our MSI
4682 	 * interrupt now
4683 	 */
4684 	if (adapter->int_mode != E1000E_INT_MODE_LEGACY) {
4685 		err = e1000_test_msi(adapter);
4686 		if (err) {
4687 			e_err("Interrupt allocation failed\n");
4688 			goto err_req_irq;
4689 		}
4690 	}
4691 
4692 	/* From here on the code is the same as e1000e_up() */
4693 	clear_bit(__E1000_DOWN, &adapter->state);
4694 
4695 	if (adapter->int_mode == E1000E_INT_MODE_MSIX)
4696 		irq = adapter->msix_entries[0].vector;
4697 	else
4698 		irq = adapter->pdev->irq;
4699 
4700 	netif_napi_set_irq(&adapter->napi, irq);
4701 	napi_enable(&adapter->napi);
4702 	netif_queue_set_napi(netdev, 0, NETDEV_QUEUE_TYPE_RX, &adapter->napi);
4703 	netif_queue_set_napi(netdev, 0, NETDEV_QUEUE_TYPE_TX, &adapter->napi);
4704 
4705 	e1000_irq_enable(adapter);
4706 
4707 	adapter->tx_hang_recheck = false;
4708 
4709 	hw->mac.get_link_status = true;
4710 	pm_runtime_put(&pdev->dev);
4711 
4712 	e1000e_trigger_lsc(adapter);
4713 
4714 	return 0;
4715 
4716 err_req_irq:
4717 	cpu_latency_qos_remove_request(&adapter->pm_qos_req);
4718 	e1000e_release_hw_control(adapter);
4719 	e1000_power_down_phy(adapter);
4720 	e1000e_free_rx_resources(adapter->rx_ring);
4721 err_setup_rx:
4722 	e1000e_free_tx_resources(adapter->tx_ring);
4723 err_setup_tx:
4724 	e1000e_reset(adapter);
4725 	pm_runtime_put_sync(&pdev->dev);
4726 
4727 	return err;
4728 }
4729 
4730 /**
4731  * e1000e_close - Disables a network interface
4732  * @netdev: network interface device structure
4733  *
4734  * Returns 0, this is not allowed to fail
4735  *
4736  * The close entry point is called when an interface is de-activated
4737  * by the OS.  The hardware is still under the drivers control, but
4738  * needs to be disabled.  A global MAC reset is issued to stop the
4739  * hardware, and all transmit and receive resources are freed.
4740  **/
4741 int e1000e_close(struct net_device *netdev)
4742 {
4743 	struct e1000_adapter *adapter = netdev_priv(netdev);
4744 	struct pci_dev *pdev = adapter->pdev;
4745 	int count = E1000_CHECK_RESET_COUNT;
4746 
4747 	while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
4748 		usleep_range(10000, 11000);
4749 
4750 	WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
4751 
4752 	pm_runtime_get_sync(&pdev->dev);
4753 
4754 	if (netif_device_present(netdev)) {
4755 		e1000e_down(adapter, true);
4756 		e1000_free_irq(adapter);
4757 
4758 		/* Link status message must follow this format */
4759 		netdev_info(netdev, "NIC Link is Down\n");
4760 	}
4761 
4762 	netif_queue_set_napi(netdev, 0, NETDEV_QUEUE_TYPE_RX, NULL);
4763 	netif_queue_set_napi(netdev, 0, NETDEV_QUEUE_TYPE_TX, NULL);
4764 	napi_disable(&adapter->napi);
4765 
4766 	e1000e_free_tx_resources(adapter->tx_ring);
4767 	e1000e_free_rx_resources(adapter->rx_ring);
4768 
4769 	/* kill manageability vlan ID if supported, but not if a vlan with
4770 	 * the same ID is registered on the host OS (let 8021q kill it)
4771 	 */
4772 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN)
4773 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
4774 				       adapter->mng_vlan_id);
4775 
4776 	/* If AMT is enabled, let the firmware know that the network
4777 	 * interface is now closed
4778 	 */
4779 	if ((adapter->flags & FLAG_HAS_AMT) &&
4780 	    !test_bit(__E1000_TESTING, &adapter->state))
4781 		e1000e_release_hw_control(adapter);
4782 
4783 	cpu_latency_qos_remove_request(&adapter->pm_qos_req);
4784 
4785 	pm_runtime_put_sync(&pdev->dev);
4786 
4787 	return 0;
4788 }
4789 
4790 /**
4791  * e1000_set_mac - Change the Ethernet Address of the NIC
4792  * @netdev: network interface device structure
4793  * @p: pointer to an address structure
4794  *
4795  * Returns 0 on success, negative on failure
4796  **/
4797 static int e1000_set_mac(struct net_device *netdev, void *p)
4798 {
4799 	struct e1000_adapter *adapter = netdev_priv(netdev);
4800 	struct e1000_hw *hw = &adapter->hw;
4801 	struct sockaddr *addr = p;
4802 
4803 	if (!is_valid_ether_addr(addr->sa_data))
4804 		return -EADDRNOTAVAIL;
4805 
4806 	eth_hw_addr_set(netdev, addr->sa_data);
4807 	memcpy(adapter->hw.mac.addr, addr->sa_data, netdev->addr_len);
4808 
4809 	hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
4810 
4811 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA) {
4812 		/* activate the work around */
4813 		e1000e_set_laa_state_82571(&adapter->hw, 1);
4814 
4815 		/* Hold a copy of the LAA in RAR[14] This is done so that
4816 		 * between the time RAR[0] gets clobbered  and the time it
4817 		 * gets fixed (in e1000_watchdog), the actual LAA is in one
4818 		 * of the RARs and no incoming packets directed to this port
4819 		 * are dropped. Eventually the LAA will be in RAR[0] and
4820 		 * RAR[14]
4821 		 */
4822 		hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr,
4823 				    adapter->hw.mac.rar_entry_count - 1);
4824 	}
4825 
4826 	return 0;
4827 }
4828 
4829 /**
4830  * e1000e_update_phy_task - work thread to update phy
4831  * @work: pointer to our work struct
4832  *
4833  * this worker thread exists because we must acquire a
4834  * semaphore to read the phy, which we could msleep while
4835  * waiting for it, and we can't msleep in a timer.
4836  **/
4837 static void e1000e_update_phy_task(struct work_struct *work)
4838 {
4839 	struct e1000_adapter *adapter = container_of(work,
4840 						     struct e1000_adapter,
4841 						     update_phy_task);
4842 	struct e1000_hw *hw = &adapter->hw;
4843 
4844 	if (test_bit(__E1000_DOWN, &adapter->state))
4845 		return;
4846 
4847 	e1000_get_phy_info(hw);
4848 
4849 	/* Enable EEE on 82579 after link up */
4850 	if (hw->phy.type >= e1000_phy_82579)
4851 		e1000_set_eee_pchlan(hw);
4852 }
4853 
4854 /**
4855  * e1000_update_phy_info - timre call-back to update PHY info
4856  * @t: pointer to timer_list containing private info adapter
4857  *
4858  * Need to wait a few seconds after link up to get diagnostic information from
4859  * the phy
4860  **/
4861 static void e1000_update_phy_info(struct timer_list *t)
4862 {
4863 	struct e1000_adapter *adapter = timer_container_of(adapter, t,
4864 							   phy_info_timer);
4865 
4866 	if (test_bit(__E1000_DOWN, &adapter->state))
4867 		return;
4868 
4869 	schedule_work(&adapter->update_phy_task);
4870 }
4871 
4872 /**
4873  * e1000e_update_phy_stats - Update the PHY statistics counters
4874  * @adapter: board private structure
4875  *
4876  * Read/clear the upper 16-bit PHY registers and read/accumulate lower
4877  **/
4878 static void e1000e_update_phy_stats(struct e1000_adapter *adapter)
4879 {
4880 	struct e1000_hw *hw = &adapter->hw;
4881 	s32 ret_val;
4882 	u16 phy_data;
4883 
4884 	ret_val = hw->phy.ops.acquire(hw);
4885 	if (ret_val)
4886 		return;
4887 
4888 	/* A page set is expensive so check if already on desired page.
4889 	 * If not, set to the page with the PHY status registers.
4890 	 */
4891 	hw->phy.addr = 1;
4892 	ret_val = e1000e_read_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4893 					   &phy_data);
4894 	if (ret_val)
4895 		goto release;
4896 	if (phy_data != (HV_STATS_PAGE << IGP_PAGE_SHIFT)) {
4897 		ret_val = hw->phy.ops.set_page(hw,
4898 					       HV_STATS_PAGE << IGP_PAGE_SHIFT);
4899 		if (ret_val)
4900 			goto release;
4901 	}
4902 
4903 	/* Single Collision Count */
4904 	hw->phy.ops.read_reg_page(hw, HV_SCC_UPPER, &phy_data);
4905 	ret_val = hw->phy.ops.read_reg_page(hw, HV_SCC_LOWER, &phy_data);
4906 	if (!ret_val)
4907 		adapter->stats.scc += phy_data;
4908 
4909 	/* Excessive Collision Count */
4910 	hw->phy.ops.read_reg_page(hw, HV_ECOL_UPPER, &phy_data);
4911 	ret_val = hw->phy.ops.read_reg_page(hw, HV_ECOL_LOWER, &phy_data);
4912 	if (!ret_val)
4913 		adapter->stats.ecol += phy_data;
4914 
4915 	/* Multiple Collision Count */
4916 	hw->phy.ops.read_reg_page(hw, HV_MCC_UPPER, &phy_data);
4917 	ret_val = hw->phy.ops.read_reg_page(hw, HV_MCC_LOWER, &phy_data);
4918 	if (!ret_val)
4919 		adapter->stats.mcc += phy_data;
4920 
4921 	/* Late Collision Count */
4922 	hw->phy.ops.read_reg_page(hw, HV_LATECOL_UPPER, &phy_data);
4923 	ret_val = hw->phy.ops.read_reg_page(hw, HV_LATECOL_LOWER, &phy_data);
4924 	if (!ret_val)
4925 		adapter->stats.latecol += phy_data;
4926 
4927 	/* Collision Count - also used for adaptive IFS */
4928 	hw->phy.ops.read_reg_page(hw, HV_COLC_UPPER, &phy_data);
4929 	ret_val = hw->phy.ops.read_reg_page(hw, HV_COLC_LOWER, &phy_data);
4930 	if (!ret_val)
4931 		hw->mac.collision_delta = phy_data;
4932 
4933 	/* Defer Count */
4934 	hw->phy.ops.read_reg_page(hw, HV_DC_UPPER, &phy_data);
4935 	ret_val = hw->phy.ops.read_reg_page(hw, HV_DC_LOWER, &phy_data);
4936 	if (!ret_val)
4937 		adapter->stats.dc += phy_data;
4938 
4939 	/* Transmit with no CRS */
4940 	hw->phy.ops.read_reg_page(hw, HV_TNCRS_UPPER, &phy_data);
4941 	ret_val = hw->phy.ops.read_reg_page(hw, HV_TNCRS_LOWER, &phy_data);
4942 	if (!ret_val)
4943 		adapter->stats.tncrs += phy_data;
4944 
4945 release:
4946 	hw->phy.ops.release(hw);
4947 }
4948 
4949 /**
4950  * e1000e_update_stats - Update the board statistics counters
4951  * @adapter: board private structure
4952  **/
4953 static void e1000e_update_stats(struct e1000_adapter *adapter)
4954 {
4955 	struct net_device *netdev = adapter->netdev;
4956 	struct e1000_hw *hw = &adapter->hw;
4957 	struct pci_dev *pdev = adapter->pdev;
4958 
4959 	/* Prevent stats update while adapter is being reset, or if the pci
4960 	 * connection is down.
4961 	 */
4962 	if (adapter->link_speed == 0)
4963 		return;
4964 	if (pci_channel_offline(pdev))
4965 		return;
4966 
4967 	adapter->stats.crcerrs += er32(CRCERRS);
4968 	adapter->stats.gprc += er32(GPRC);
4969 	adapter->stats.gorc += er32(GORCL);
4970 	er32(GORCH);		/* Clear gorc */
4971 	adapter->stats.bprc += er32(BPRC);
4972 	adapter->stats.mprc += er32(MPRC);
4973 	adapter->stats.roc += er32(ROC);
4974 
4975 	adapter->stats.mpc += er32(MPC);
4976 
4977 	/* Half-duplex statistics */
4978 	if (adapter->link_duplex == HALF_DUPLEX) {
4979 		if (adapter->flags2 & FLAG2_HAS_PHY_STATS) {
4980 			e1000e_update_phy_stats(adapter);
4981 		} else {
4982 			adapter->stats.scc += er32(SCC);
4983 			adapter->stats.ecol += er32(ECOL);
4984 			adapter->stats.mcc += er32(MCC);
4985 			adapter->stats.latecol += er32(LATECOL);
4986 			adapter->stats.dc += er32(DC);
4987 
4988 			hw->mac.collision_delta = er32(COLC);
4989 
4990 			if ((hw->mac.type != e1000_82574) &&
4991 			    (hw->mac.type != e1000_82583))
4992 				adapter->stats.tncrs += er32(TNCRS);
4993 		}
4994 		adapter->stats.colc += hw->mac.collision_delta;
4995 	}
4996 
4997 	adapter->stats.xonrxc += er32(XONRXC);
4998 	adapter->stats.xontxc += er32(XONTXC);
4999 	adapter->stats.xoffrxc += er32(XOFFRXC);
5000 	adapter->stats.xofftxc += er32(XOFFTXC);
5001 	adapter->stats.gptc += er32(GPTC);
5002 	adapter->stats.gotc += er32(GOTCL);
5003 	er32(GOTCH);		/* Clear gotc */
5004 	adapter->stats.rnbc += er32(RNBC);
5005 	adapter->stats.ruc += er32(RUC);
5006 
5007 	adapter->stats.mptc += er32(MPTC);
5008 	adapter->stats.bptc += er32(BPTC);
5009 
5010 	/* used for adaptive IFS */
5011 
5012 	hw->mac.tx_packet_delta = er32(TPT);
5013 	adapter->stats.tpt += hw->mac.tx_packet_delta;
5014 
5015 	adapter->stats.algnerrc += er32(ALGNERRC);
5016 	adapter->stats.rxerrc += er32(RXERRC);
5017 	adapter->stats.cexterr += er32(CEXTERR);
5018 	adapter->stats.tsctc += er32(TSCTC);
5019 	adapter->stats.tsctfc += er32(TSCTFC);
5020 
5021 	/* Fill out the OS statistics structure */
5022 	netdev->stats.multicast = adapter->stats.mprc;
5023 	netdev->stats.collisions = adapter->stats.colc;
5024 
5025 	/* Rx Errors */
5026 
5027 	/* RLEC on some newer hardware can be incorrect so build
5028 	 * our own version based on RUC and ROC
5029 	 */
5030 	netdev->stats.rx_errors = adapter->stats.rxerrc +
5031 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
5032 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5033 	netdev->stats.rx_length_errors = adapter->stats.ruc +
5034 	    adapter->stats.roc;
5035 	netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
5036 	netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
5037 	netdev->stats.rx_missed_errors = adapter->stats.mpc;
5038 
5039 	/* Tx Errors */
5040 	netdev->stats.tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5041 	netdev->stats.tx_aborted_errors = adapter->stats.ecol;
5042 	netdev->stats.tx_window_errors = adapter->stats.latecol;
5043 	netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
5044 
5045 	/* Tx Dropped needs to be maintained elsewhere */
5046 
5047 	/* Management Stats */
5048 	adapter->stats.mgptc += er32(MGTPTC);
5049 	adapter->stats.mgprc += er32(MGTPRC);
5050 	adapter->stats.mgpdc += er32(MGTPDC);
5051 
5052 	/* Correctable ECC Errors */
5053 	if (hw->mac.type >= e1000_pch_lpt) {
5054 		u32 pbeccsts = er32(PBECCSTS);
5055 
5056 		adapter->corr_errors +=
5057 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
5058 		adapter->uncorr_errors +=
5059 		    FIELD_GET(E1000_PBECCSTS_UNCORR_ERR_CNT_MASK, pbeccsts);
5060 	}
5061 }
5062 
5063 /**
5064  * e1000_phy_read_status - Update the PHY register status snapshot
5065  * @adapter: board private structure
5066  **/
5067 static void e1000_phy_read_status(struct e1000_adapter *adapter)
5068 {
5069 	struct e1000_hw *hw = &adapter->hw;
5070 	struct e1000_phy_regs *phy = &adapter->phy_regs;
5071 
5072 	if (!pm_runtime_suspended((&adapter->pdev->dev)->parent) &&
5073 	    (er32(STATUS) & E1000_STATUS_LU) &&
5074 	    (adapter->hw.phy.media_type == e1000_media_type_copper)) {
5075 		int ret_val;
5076 
5077 		ret_val = e1e_rphy(hw, MII_BMCR, &phy->bmcr);
5078 		ret_val |= e1e_rphy(hw, MII_BMSR, &phy->bmsr);
5079 		ret_val |= e1e_rphy(hw, MII_ADVERTISE, &phy->advertise);
5080 		ret_val |= e1e_rphy(hw, MII_LPA, &phy->lpa);
5081 		ret_val |= e1e_rphy(hw, MII_EXPANSION, &phy->expansion);
5082 		ret_val |= e1e_rphy(hw, MII_CTRL1000, &phy->ctrl1000);
5083 		ret_val |= e1e_rphy(hw, MII_STAT1000, &phy->stat1000);
5084 		ret_val |= e1e_rphy(hw, MII_ESTATUS, &phy->estatus);
5085 		if (ret_val)
5086 			e_warn("Error reading PHY register\n");
5087 	} else {
5088 		/* Do not read PHY registers if link is not up
5089 		 * Set values to typical power-on defaults
5090 		 */
5091 		phy->bmcr = (BMCR_SPEED1000 | BMCR_ANENABLE | BMCR_FULLDPLX);
5092 		phy->bmsr = (BMSR_100FULL | BMSR_100HALF | BMSR_10FULL |
5093 			     BMSR_10HALF | BMSR_ESTATEN | BMSR_ANEGCAPABLE |
5094 			     BMSR_ERCAP);
5095 		phy->advertise = (ADVERTISE_PAUSE_ASYM | ADVERTISE_PAUSE_CAP |
5096 				  ADVERTISE_ALL | ADVERTISE_CSMA);
5097 		phy->lpa = 0;
5098 		phy->expansion = EXPANSION_ENABLENPAGE;
5099 		phy->ctrl1000 = ADVERTISE_1000FULL;
5100 		phy->stat1000 = 0;
5101 		phy->estatus = (ESTATUS_1000_TFULL | ESTATUS_1000_THALF);
5102 	}
5103 }
5104 
5105 static void e1000_print_link_info(struct e1000_adapter *adapter)
5106 {
5107 	struct e1000_hw *hw = &adapter->hw;
5108 	u32 ctrl = er32(CTRL);
5109 
5110 	/* Link status message must follow this format for user tools */
5111 	netdev_info(adapter->netdev,
5112 		    "NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5113 		    adapter->link_speed,
5114 		    adapter->link_duplex == FULL_DUPLEX ? "Full" : "Half",
5115 		    (ctrl & E1000_CTRL_TFCE) && (ctrl & E1000_CTRL_RFCE) ? "Rx/Tx" :
5116 		    (ctrl & E1000_CTRL_RFCE) ? "Rx" :
5117 		    (ctrl & E1000_CTRL_TFCE) ? "Tx" : "None");
5118 }
5119 
5120 static bool e1000e_has_link(struct e1000_adapter *adapter)
5121 {
5122 	struct e1000_hw *hw = &adapter->hw;
5123 	bool link_active = false;
5124 	s32 ret_val = 0;
5125 
5126 	/* get_link_status is set on LSC (link status) interrupt or
5127 	 * Rx sequence error interrupt.  get_link_status will stay
5128 	 * true until the check_for_link establishes link
5129 	 * for copper adapters ONLY
5130 	 */
5131 	switch (hw->phy.media_type) {
5132 	case e1000_media_type_copper:
5133 		if (hw->mac.get_link_status) {
5134 			ret_val = hw->mac.ops.check_for_link(hw);
5135 			link_active = !hw->mac.get_link_status;
5136 		} else {
5137 			link_active = true;
5138 		}
5139 		break;
5140 	case e1000_media_type_fiber:
5141 		ret_val = hw->mac.ops.check_for_link(hw);
5142 		link_active = !!(er32(STATUS) & E1000_STATUS_LU);
5143 		break;
5144 	case e1000_media_type_internal_serdes:
5145 		ret_val = hw->mac.ops.check_for_link(hw);
5146 		link_active = hw->mac.serdes_has_link;
5147 		break;
5148 	default:
5149 	case e1000_media_type_unknown:
5150 		break;
5151 	}
5152 
5153 	if ((ret_val == -E1000_ERR_PHY) && (hw->phy.type == e1000_phy_igp_3) &&
5154 	    (er32(CTRL) & E1000_PHY_CTRL_GBE_DISABLE)) {
5155 		/* See e1000_kmrn_lock_loss_workaround_ich8lan() */
5156 		e_info("Gigabit has been disabled, downgrading speed\n");
5157 	}
5158 
5159 	return link_active;
5160 }
5161 
5162 static void e1000e_enable_receives(struct e1000_adapter *adapter)
5163 {
5164 	/* make sure the receive unit is started */
5165 	if ((adapter->flags & FLAG_RX_NEEDS_RESTART) &&
5166 	    (adapter->flags & FLAG_RESTART_NOW)) {
5167 		struct e1000_hw *hw = &adapter->hw;
5168 		u32 rctl = er32(RCTL);
5169 
5170 		ew32(RCTL, rctl | E1000_RCTL_EN);
5171 		adapter->flags &= ~FLAG_RESTART_NOW;
5172 	}
5173 }
5174 
5175 static void e1000e_check_82574_phy_workaround(struct e1000_adapter *adapter)
5176 {
5177 	struct e1000_hw *hw = &adapter->hw;
5178 
5179 	/* With 82574 controllers, PHY needs to be checked periodically
5180 	 * for hung state and reset, if two calls return true
5181 	 */
5182 	if (e1000_check_phy_82574(hw))
5183 		adapter->phy_hang_count++;
5184 	else
5185 		adapter->phy_hang_count = 0;
5186 
5187 	if (adapter->phy_hang_count > 1) {
5188 		adapter->phy_hang_count = 0;
5189 		e_dbg("PHY appears hung - resetting\n");
5190 		schedule_work(&adapter->reset_task);
5191 	}
5192 }
5193 
5194 /**
5195  * e1000_watchdog - Timer Call-back
5196  * @t: pointer to timer_list containing private info adapter
5197  **/
5198 static void e1000_watchdog(struct timer_list *t)
5199 {
5200 	struct e1000_adapter *adapter = timer_container_of(adapter, t,
5201 							   watchdog_timer);
5202 
5203 	/* Do the rest outside of interrupt context */
5204 	schedule_work(&adapter->watchdog_task);
5205 
5206 	/* TODO: make this use queue_delayed_work() */
5207 }
5208 
5209 static void e1000_watchdog_task(struct work_struct *work)
5210 {
5211 	struct e1000_adapter *adapter = container_of(work,
5212 						     struct e1000_adapter,
5213 						     watchdog_task);
5214 	struct net_device *netdev = adapter->netdev;
5215 	struct e1000_mac_info *mac = &adapter->hw.mac;
5216 	struct e1000_phy_info *phy = &adapter->hw.phy;
5217 	struct e1000_ring *tx_ring = adapter->tx_ring;
5218 	u32 dmoff_exit_timeout = 100, tries = 0;
5219 	struct e1000_hw *hw = &adapter->hw;
5220 	u32 link, tctl, pcim_state;
5221 
5222 	if (test_bit(__E1000_DOWN, &adapter->state))
5223 		return;
5224 
5225 	link = e1000e_has_link(adapter);
5226 	if ((netif_carrier_ok(netdev)) && link) {
5227 		/* Cancel scheduled suspend requests. */
5228 		pm_runtime_resume(netdev->dev.parent);
5229 
5230 		e1000e_enable_receives(adapter);
5231 		goto link_up;
5232 	}
5233 
5234 	if ((e1000e_enable_tx_pkt_filtering(hw)) &&
5235 	    (adapter->mng_vlan_id != adapter->hw.mng_cookie.vlan_id))
5236 		e1000_update_mng_vlan(adapter);
5237 
5238 	if (link) {
5239 		if (!netif_carrier_ok(netdev)) {
5240 			bool txb2b = true;
5241 
5242 			/* Cancel scheduled suspend requests. */
5243 			pm_runtime_resume(netdev->dev.parent);
5244 
5245 			/* Checking if MAC is in DMoff state*/
5246 			if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID) {
5247 				pcim_state = er32(STATUS);
5248 				while (pcim_state & E1000_STATUS_PCIM_STATE) {
5249 					if (tries++ == dmoff_exit_timeout) {
5250 						e_dbg("Error in exiting dmoff\n");
5251 						break;
5252 					}
5253 					usleep_range(10000, 20000);
5254 					pcim_state = er32(STATUS);
5255 
5256 					/* Checking if MAC exited DMoff state */
5257 					if (!(pcim_state & E1000_STATUS_PCIM_STATE))
5258 						e1000_phy_hw_reset(&adapter->hw);
5259 				}
5260 			}
5261 
5262 			/* update snapshot of PHY registers on LSC */
5263 			e1000_phy_read_status(adapter);
5264 			mac->ops.get_link_up_info(&adapter->hw,
5265 						  &adapter->link_speed,
5266 						  &adapter->link_duplex);
5267 			e1000_print_link_info(adapter);
5268 
5269 			/* check if SmartSpeed worked */
5270 			e1000e_check_downshift(hw);
5271 			if (phy->speed_downgraded)
5272 				netdev_warn(netdev,
5273 					    "Link Speed was downgraded by SmartSpeed\n");
5274 
5275 			/* On supported PHYs, check for duplex mismatch only
5276 			 * if link has autonegotiated at 10/100 half
5277 			 */
5278 			if ((hw->phy.type == e1000_phy_igp_3 ||
5279 			     hw->phy.type == e1000_phy_bm) &&
5280 			    hw->mac.autoneg &&
5281 			    (adapter->link_speed == SPEED_10 ||
5282 			     adapter->link_speed == SPEED_100) &&
5283 			    (adapter->link_duplex == HALF_DUPLEX)) {
5284 				u16 autoneg_exp;
5285 
5286 				e1e_rphy(hw, MII_EXPANSION, &autoneg_exp);
5287 
5288 				if (!(autoneg_exp & EXPANSION_NWAY))
5289 					e_info("Autonegotiated half duplex but link partner cannot autoneg.  Try forcing full duplex if link gets many collisions.\n");
5290 			}
5291 
5292 			/* adjust timeout factor according to speed/duplex */
5293 			adapter->tx_timeout_factor = 1;
5294 			switch (adapter->link_speed) {
5295 			case SPEED_10:
5296 				txb2b = false;
5297 				adapter->tx_timeout_factor = 16;
5298 				break;
5299 			case SPEED_100:
5300 				txb2b = false;
5301 				adapter->tx_timeout_factor = 10;
5302 				break;
5303 			}
5304 
5305 			/* workaround: re-program speed mode bit after
5306 			 * link-up event
5307 			 */
5308 			if ((adapter->flags & FLAG_TARC_SPEED_MODE_BIT) &&
5309 			    !txb2b) {
5310 				u32 tarc0;
5311 
5312 				tarc0 = er32(TARC(0));
5313 				tarc0 &= ~SPEED_MODE_BIT;
5314 				ew32(TARC(0), tarc0);
5315 			}
5316 
5317 			/* enable transmits in the hardware, need to do this
5318 			 * after setting TARC(0)
5319 			 */
5320 			tctl = er32(TCTL);
5321 			tctl |= E1000_TCTL_EN;
5322 			ew32(TCTL, tctl);
5323 
5324 			/* Perform any post-link-up configuration before
5325 			 * reporting link up.
5326 			 */
5327 			if (phy->ops.cfg_on_link_up)
5328 				phy->ops.cfg_on_link_up(hw);
5329 
5330 			netif_wake_queue(netdev);
5331 			netif_carrier_on(netdev);
5332 
5333 			if (!test_bit(__E1000_DOWN, &adapter->state))
5334 				mod_timer(&adapter->phy_info_timer,
5335 					  round_jiffies(jiffies + 2 * HZ));
5336 		}
5337 	} else {
5338 		if (netif_carrier_ok(netdev)) {
5339 			adapter->link_speed = 0;
5340 			adapter->link_duplex = 0;
5341 			/* Link status message must follow this format */
5342 			netdev_info(netdev, "NIC Link is Down\n");
5343 			netif_carrier_off(netdev);
5344 			netif_stop_queue(netdev);
5345 			if (!test_bit(__E1000_DOWN, &adapter->state))
5346 				mod_timer(&adapter->phy_info_timer,
5347 					  round_jiffies(jiffies + 2 * HZ));
5348 
5349 			/* 8000ES2LAN requires a Rx packet buffer work-around
5350 			 * on link down event; reset the controller to flush
5351 			 * the Rx packet buffer.
5352 			 */
5353 			if (adapter->flags & FLAG_RX_NEEDS_RESTART)
5354 				adapter->flags |= FLAG_RESTART_NOW;
5355 			else
5356 				pm_schedule_suspend(netdev->dev.parent,
5357 						    LINK_TIMEOUT);
5358 		}
5359 	}
5360 
5361 link_up:
5362 	spin_lock(&adapter->stats64_lock);
5363 	e1000e_update_stats(adapter);
5364 
5365 	mac->tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
5366 	adapter->tpt_old = adapter->stats.tpt;
5367 	mac->collision_delta = adapter->stats.colc - adapter->colc_old;
5368 	adapter->colc_old = adapter->stats.colc;
5369 
5370 	adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
5371 	adapter->gorc_old = adapter->stats.gorc;
5372 	adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
5373 	adapter->gotc_old = adapter->stats.gotc;
5374 	spin_unlock(&adapter->stats64_lock);
5375 
5376 	/* If the link is lost the controller stops DMA, but
5377 	 * if there is queued Tx work it cannot be done.  So
5378 	 * reset the controller to flush the Tx packet buffers.
5379 	 */
5380 	if (!netif_carrier_ok(netdev) &&
5381 	    (e1000_desc_unused(tx_ring) + 1 < tx_ring->count))
5382 		adapter->flags |= FLAG_RESTART_NOW;
5383 
5384 	/* If reset is necessary, do it outside of interrupt context. */
5385 	if (adapter->flags & FLAG_RESTART_NOW) {
5386 		schedule_work(&adapter->reset_task);
5387 		/* return immediately since reset is imminent */
5388 		return;
5389 	}
5390 
5391 	e1000e_update_adaptive(&adapter->hw);
5392 
5393 	/* Simple mode for Interrupt Throttle Rate (ITR) */
5394 	if (adapter->itr_setting == 4) {
5395 		/* Symmetric Tx/Rx gets a reduced ITR=2000;
5396 		 * Total asymmetrical Tx or Rx gets ITR=8000;
5397 		 * everyone else is between 2000-8000.
5398 		 */
5399 		u32 goc = (adapter->gotc + adapter->gorc) / 10000;
5400 		u32 dif = (adapter->gotc > adapter->gorc ?
5401 			   adapter->gotc - adapter->gorc :
5402 			   adapter->gorc - adapter->gotc) / 10000;
5403 		u32 itr = goc > 0 ? (dif * 6000 / goc + 2000) : 8000;
5404 
5405 		e1000e_write_itr(adapter, itr);
5406 	}
5407 
5408 	/* Cause software interrupt to ensure Rx ring is cleaned */
5409 	if (adapter->msix_entries)
5410 		ew32(ICS, adapter->rx_ring->ims_val);
5411 	else
5412 		ew32(ICS, E1000_ICS_RXDMT0);
5413 
5414 	/* flush pending descriptors to memory before detecting Tx hang */
5415 	e1000e_flush_descriptors(adapter);
5416 
5417 	/* Force detection of hung controller every watchdog period */
5418 	adapter->detect_tx_hung = true;
5419 
5420 	/* With 82571 controllers, LAA may be overwritten due to controller
5421 	 * reset from the other port. Set the appropriate LAA in RAR[0]
5422 	 */
5423 	if (e1000e_get_laa_state_82571(hw))
5424 		hw->mac.ops.rar_set(hw, adapter->hw.mac.addr, 0);
5425 
5426 	if (adapter->flags2 & FLAG2_CHECK_PHY_HANG)
5427 		e1000e_check_82574_phy_workaround(adapter);
5428 
5429 	/* Clear valid timestamp stuck in RXSTMPL/H due to a Rx error */
5430 	if (adapter->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE) {
5431 		if ((adapter->flags2 & FLAG2_CHECK_RX_HWTSTAMP) &&
5432 		    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID)) {
5433 			er32(RXSTMPH);
5434 			adapter->rx_hwtstamp_cleared++;
5435 		} else {
5436 			adapter->flags2 |= FLAG2_CHECK_RX_HWTSTAMP;
5437 		}
5438 	}
5439 
5440 	/* Reset the timer */
5441 	if (!test_bit(__E1000_DOWN, &adapter->state))
5442 		mod_timer(&adapter->watchdog_timer,
5443 			  round_jiffies(jiffies + 2 * HZ));
5444 }
5445 
5446 #define E1000_TX_FLAGS_CSUM		0x00000001
5447 #define E1000_TX_FLAGS_VLAN		0x00000002
5448 #define E1000_TX_FLAGS_TSO		0x00000004
5449 #define E1000_TX_FLAGS_IPV4		0x00000008
5450 #define E1000_TX_FLAGS_NO_FCS		0x00000010
5451 #define E1000_TX_FLAGS_HWTSTAMP		0x00000020
5452 #define E1000_TX_FLAGS_VLAN_MASK	0xffff0000
5453 #define E1000_TX_FLAGS_VLAN_SHIFT	16
5454 
5455 static int e1000_tso(struct e1000_ring *tx_ring, struct sk_buff *skb,
5456 		     __be16 protocol)
5457 {
5458 	struct e1000_context_desc *context_desc;
5459 	struct e1000_buffer *buffer_info;
5460 	unsigned int i;
5461 	u32 cmd_length = 0;
5462 	u16 ipcse = 0, mss;
5463 	u8 ipcss, ipcso, tucss, tucso, hdr_len;
5464 	int err;
5465 
5466 	if (!skb_is_gso(skb))
5467 		return 0;
5468 
5469 	err = skb_cow_head(skb, 0);
5470 	if (err < 0)
5471 		return err;
5472 
5473 	hdr_len = skb_tcp_all_headers(skb);
5474 	mss = skb_shinfo(skb)->gso_size;
5475 	if (protocol == htons(ETH_P_IP)) {
5476 		struct iphdr *iph = ip_hdr(skb);
5477 		iph->tot_len = 0;
5478 		iph->check = 0;
5479 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
5480 							 0, IPPROTO_TCP, 0);
5481 		cmd_length = E1000_TXD_CMD_IP;
5482 		ipcse = skb_transport_offset(skb) - 1;
5483 	} else if (skb_is_gso_v6(skb)) {
5484 		tcp_v6_gso_csum_prep(skb);
5485 		ipcse = 0;
5486 	}
5487 	ipcss = skb_network_offset(skb);
5488 	ipcso = (void *)&(ip_hdr(skb)->check) - (void *)skb->data;
5489 	tucss = skb_transport_offset(skb);
5490 	tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
5491 
5492 	cmd_length |= (E1000_TXD_CMD_DEXT | E1000_TXD_CMD_TSE |
5493 		       E1000_TXD_CMD_TCP | (skb->len - (hdr_len)));
5494 
5495 	i = tx_ring->next_to_use;
5496 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5497 	buffer_info = &tx_ring->buffer_info[i];
5498 
5499 	context_desc->lower_setup.ip_fields.ipcss = ipcss;
5500 	context_desc->lower_setup.ip_fields.ipcso = ipcso;
5501 	context_desc->lower_setup.ip_fields.ipcse = cpu_to_le16(ipcse);
5502 	context_desc->upper_setup.tcp_fields.tucss = tucss;
5503 	context_desc->upper_setup.tcp_fields.tucso = tucso;
5504 	context_desc->upper_setup.tcp_fields.tucse = 0;
5505 	context_desc->tcp_seg_setup.fields.mss = cpu_to_le16(mss);
5506 	context_desc->tcp_seg_setup.fields.hdr_len = hdr_len;
5507 	context_desc->cmd_and_length = cpu_to_le32(cmd_length);
5508 
5509 	buffer_info->time_stamp = jiffies;
5510 	buffer_info->next_to_watch = i;
5511 
5512 	i++;
5513 	if (i == tx_ring->count)
5514 		i = 0;
5515 	tx_ring->next_to_use = i;
5516 
5517 	return 1;
5518 }
5519 
5520 static bool e1000_tx_csum(struct e1000_ring *tx_ring, struct sk_buff *skb,
5521 			  __be16 protocol)
5522 {
5523 	struct e1000_adapter *adapter = tx_ring->adapter;
5524 	struct e1000_context_desc *context_desc;
5525 	struct e1000_buffer *buffer_info;
5526 	unsigned int i;
5527 	u8 css;
5528 	u32 cmd_len = E1000_TXD_CMD_DEXT;
5529 
5530 	if (skb->ip_summed != CHECKSUM_PARTIAL)
5531 		return false;
5532 
5533 	switch (protocol) {
5534 	case cpu_to_be16(ETH_P_IP):
5535 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
5536 			cmd_len |= E1000_TXD_CMD_TCP;
5537 		break;
5538 	case cpu_to_be16(ETH_P_IPV6):
5539 		/* XXX not handling all IPV6 headers */
5540 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
5541 			cmd_len |= E1000_TXD_CMD_TCP;
5542 		break;
5543 	default:
5544 		if (unlikely(net_ratelimit()))
5545 			e_warn("checksum_partial proto=%x!\n",
5546 			       be16_to_cpu(protocol));
5547 		break;
5548 	}
5549 
5550 	css = skb_checksum_start_offset(skb);
5551 
5552 	i = tx_ring->next_to_use;
5553 	buffer_info = &tx_ring->buffer_info[i];
5554 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5555 
5556 	context_desc->lower_setup.ip_config = 0;
5557 	context_desc->upper_setup.tcp_fields.tucss = css;
5558 	context_desc->upper_setup.tcp_fields.tucso = css + skb->csum_offset;
5559 	context_desc->upper_setup.tcp_fields.tucse = 0;
5560 	context_desc->tcp_seg_setup.data = 0;
5561 	context_desc->cmd_and_length = cpu_to_le32(cmd_len);
5562 
5563 	buffer_info->time_stamp = jiffies;
5564 	buffer_info->next_to_watch = i;
5565 
5566 	i++;
5567 	if (i == tx_ring->count)
5568 		i = 0;
5569 	tx_ring->next_to_use = i;
5570 
5571 	return true;
5572 }
5573 
5574 static int e1000_tx_map(struct e1000_ring *tx_ring, struct sk_buff *skb,
5575 			unsigned int first, unsigned int max_per_txd,
5576 			unsigned int nr_frags)
5577 {
5578 	struct e1000_adapter *adapter = tx_ring->adapter;
5579 	struct pci_dev *pdev = adapter->pdev;
5580 	struct e1000_buffer *buffer_info;
5581 	unsigned int len = skb_headlen(skb);
5582 	unsigned int offset = 0, size, count = 0, i;
5583 	unsigned int f, bytecount, segs;
5584 
5585 	i = tx_ring->next_to_use;
5586 
5587 	while (len) {
5588 		buffer_info = &tx_ring->buffer_info[i];
5589 		size = min(len, max_per_txd);
5590 
5591 		buffer_info->length = size;
5592 		buffer_info->time_stamp = jiffies;
5593 		buffer_info->next_to_watch = i;
5594 		buffer_info->dma = dma_map_single(&pdev->dev,
5595 						  skb->data + offset,
5596 						  size, DMA_TO_DEVICE);
5597 		buffer_info->mapped_as_page = false;
5598 		if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5599 			goto dma_error;
5600 
5601 		len -= size;
5602 		offset += size;
5603 		count++;
5604 
5605 		if (len) {
5606 			i++;
5607 			if (i == tx_ring->count)
5608 				i = 0;
5609 		}
5610 	}
5611 
5612 	for (f = 0; f < nr_frags; f++) {
5613 		const skb_frag_t *frag = &skb_shinfo(skb)->frags[f];
5614 
5615 		len = skb_frag_size(frag);
5616 		offset = 0;
5617 
5618 		while (len) {
5619 			i++;
5620 			if (i == tx_ring->count)
5621 				i = 0;
5622 
5623 			buffer_info = &tx_ring->buffer_info[i];
5624 			size = min(len, max_per_txd);
5625 
5626 			buffer_info->length = size;
5627 			buffer_info->time_stamp = jiffies;
5628 			buffer_info->next_to_watch = i;
5629 			buffer_info->dma = skb_frag_dma_map(&pdev->dev, frag,
5630 							    offset, size,
5631 							    DMA_TO_DEVICE);
5632 			buffer_info->mapped_as_page = true;
5633 			if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5634 				goto dma_error;
5635 
5636 			len -= size;
5637 			offset += size;
5638 			count++;
5639 		}
5640 	}
5641 
5642 	segs = skb_shinfo(skb)->gso_segs ? : 1;
5643 	/* multiply data chunks by size of headers */
5644 	bytecount = ((segs - 1) * skb_headlen(skb)) + skb->len;
5645 
5646 	tx_ring->buffer_info[i].skb = skb;
5647 	tx_ring->buffer_info[i].segs = segs;
5648 	tx_ring->buffer_info[i].bytecount = bytecount;
5649 	tx_ring->buffer_info[first].next_to_watch = i;
5650 
5651 	return count;
5652 
5653 dma_error:
5654 	dev_err(&pdev->dev, "Tx DMA map failed\n");
5655 	buffer_info->dma = 0;
5656 	if (count)
5657 		count--;
5658 
5659 	while (count--) {
5660 		if (i == 0)
5661 			i += tx_ring->count;
5662 		i--;
5663 		buffer_info = &tx_ring->buffer_info[i];
5664 		e1000_put_txbuf(tx_ring, buffer_info, true);
5665 	}
5666 
5667 	return 0;
5668 }
5669 
5670 static void e1000_tx_queue(struct e1000_ring *tx_ring, int tx_flags, int count)
5671 {
5672 	struct e1000_adapter *adapter = tx_ring->adapter;
5673 	struct e1000_tx_desc *tx_desc = NULL;
5674 	struct e1000_buffer *buffer_info;
5675 	u32 txd_upper = 0, txd_lower = E1000_TXD_CMD_IFCS;
5676 	unsigned int i;
5677 
5678 	if (tx_flags & E1000_TX_FLAGS_TSO) {
5679 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D |
5680 		    E1000_TXD_CMD_TSE;
5681 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5682 
5683 		if (tx_flags & E1000_TX_FLAGS_IPV4)
5684 			txd_upper |= E1000_TXD_POPTS_IXSM << 8;
5685 	}
5686 
5687 	if (tx_flags & E1000_TX_FLAGS_CSUM) {
5688 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5689 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5690 	}
5691 
5692 	if (tx_flags & E1000_TX_FLAGS_VLAN) {
5693 		txd_lower |= E1000_TXD_CMD_VLE;
5694 		txd_upper |= (tx_flags & E1000_TX_FLAGS_VLAN_MASK);
5695 	}
5696 
5697 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5698 		txd_lower &= ~(E1000_TXD_CMD_IFCS);
5699 
5700 	if (unlikely(tx_flags & E1000_TX_FLAGS_HWTSTAMP)) {
5701 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5702 		txd_upper |= E1000_TXD_EXTCMD_TSTAMP;
5703 	}
5704 
5705 	i = tx_ring->next_to_use;
5706 
5707 	do {
5708 		buffer_info = &tx_ring->buffer_info[i];
5709 		tx_desc = E1000_TX_DESC(*tx_ring, i);
5710 		tx_desc->buffer_addr = cpu_to_le64(buffer_info->dma);
5711 		tx_desc->lower.data = cpu_to_le32(txd_lower |
5712 						  buffer_info->length);
5713 		tx_desc->upper.data = cpu_to_le32(txd_upper);
5714 
5715 		i++;
5716 		if (i == tx_ring->count)
5717 			i = 0;
5718 	} while (--count > 0);
5719 
5720 	tx_desc->lower.data |= cpu_to_le32(adapter->txd_cmd);
5721 
5722 	/* txd_cmd re-enables FCS, so we'll re-disable it here as desired. */
5723 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5724 		tx_desc->lower.data &= ~(cpu_to_le32(E1000_TXD_CMD_IFCS));
5725 
5726 	/* Force memory writes to complete before letting h/w
5727 	 * know there are new descriptors to fetch.  (Only
5728 	 * applicable for weak-ordered memory model archs,
5729 	 * such as IA-64).
5730 	 */
5731 	wmb();
5732 
5733 	tx_ring->next_to_use = i;
5734 }
5735 
5736 #define MINIMUM_DHCP_PACKET_SIZE 282
5737 static int e1000_transfer_dhcp_info(struct e1000_adapter *adapter,
5738 				    struct sk_buff *skb)
5739 {
5740 	struct e1000_hw *hw = &adapter->hw;
5741 	u16 length, offset;
5742 
5743 	if (skb_vlan_tag_present(skb) &&
5744 	    !((skb_vlan_tag_get(skb) == adapter->hw.mng_cookie.vlan_id) &&
5745 	      (adapter->hw.mng_cookie.status &
5746 	       E1000_MNG_DHCP_COOKIE_STATUS_VLAN)))
5747 		return 0;
5748 
5749 	if (skb->len <= MINIMUM_DHCP_PACKET_SIZE)
5750 		return 0;
5751 
5752 	if (((struct ethhdr *)skb->data)->h_proto != htons(ETH_P_IP))
5753 		return 0;
5754 
5755 	{
5756 		const struct iphdr *ip = (struct iphdr *)((u8 *)skb->data + 14);
5757 		struct udphdr *udp;
5758 
5759 		if (ip->protocol != IPPROTO_UDP)
5760 			return 0;
5761 
5762 		udp = (struct udphdr *)((u8 *)ip + (ip->ihl << 2));
5763 		if (ntohs(udp->dest) != 67)
5764 			return 0;
5765 
5766 		offset = (u8 *)udp + 8 - skb->data;
5767 		length = skb->len - offset;
5768 		return e1000e_mng_write_dhcp_info(hw, (u8 *)udp + 8, length);
5769 	}
5770 
5771 	return 0;
5772 }
5773 
5774 static int __e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5775 {
5776 	struct e1000_adapter *adapter = tx_ring->adapter;
5777 
5778 	netif_stop_queue(adapter->netdev);
5779 	/* Herbert's original patch had:
5780 	 *  smp_mb__after_netif_stop_queue();
5781 	 * but since that doesn't exist yet, just open code it.
5782 	 */
5783 	smp_mb();
5784 
5785 	/* We need to check again in a case another CPU has just
5786 	 * made room available.
5787 	 */
5788 	if (e1000_desc_unused(tx_ring) < size)
5789 		return -EBUSY;
5790 
5791 	/* A reprieve! */
5792 	netif_start_queue(adapter->netdev);
5793 	++adapter->restart_queue;
5794 	return 0;
5795 }
5796 
5797 static int e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5798 {
5799 	BUG_ON(size > tx_ring->count);
5800 
5801 	if (e1000_desc_unused(tx_ring) >= size)
5802 		return 0;
5803 	return __e1000_maybe_stop_tx(tx_ring, size);
5804 }
5805 
5806 static netdev_tx_t e1000_xmit_frame(struct sk_buff *skb,
5807 				    struct net_device *netdev)
5808 {
5809 	struct e1000_adapter *adapter = netdev_priv(netdev);
5810 	struct e1000_ring *tx_ring = adapter->tx_ring;
5811 	unsigned int first;
5812 	unsigned int tx_flags = 0;
5813 	unsigned int len = skb_headlen(skb);
5814 	unsigned int nr_frags;
5815 	unsigned int mss;
5816 	int count = 0;
5817 	int tso;
5818 	unsigned int f;
5819 	__be16 protocol = vlan_get_protocol(skb);
5820 
5821 	if (test_bit(__E1000_DOWN, &adapter->state)) {
5822 		dev_kfree_skb_any(skb);
5823 		return NETDEV_TX_OK;
5824 	}
5825 
5826 	if (skb->len <= 0) {
5827 		dev_kfree_skb_any(skb);
5828 		return NETDEV_TX_OK;
5829 	}
5830 
5831 	/* The minimum packet size with TCTL.PSP set is 17 bytes so
5832 	 * pad skb in order to meet this minimum size requirement
5833 	 */
5834 	if (skb_put_padto(skb, 17))
5835 		return NETDEV_TX_OK;
5836 
5837 	mss = skb_shinfo(skb)->gso_size;
5838 	if (mss) {
5839 		u8 hdr_len;
5840 
5841 		/* TSO Workaround for 82571/2/3 Controllers -- if skb->data
5842 		 * points to just header, pull a few bytes of payload from
5843 		 * frags into skb->data
5844 		 */
5845 		hdr_len = skb_tcp_all_headers(skb);
5846 		/* we do this workaround for ES2LAN, but it is un-necessary,
5847 		 * avoiding it could save a lot of cycles
5848 		 */
5849 		if (skb->data_len && (hdr_len == len)) {
5850 			unsigned int pull_size;
5851 
5852 			pull_size = min_t(unsigned int, 4, skb->data_len);
5853 			if (!__pskb_pull_tail(skb, pull_size)) {
5854 				e_err("__pskb_pull_tail failed.\n");
5855 				dev_kfree_skb_any(skb);
5856 				return NETDEV_TX_OK;
5857 			}
5858 			len = skb_headlen(skb);
5859 		}
5860 	}
5861 
5862 	/* reserve a descriptor for the offload context */
5863 	if ((mss) || (skb->ip_summed == CHECKSUM_PARTIAL))
5864 		count++;
5865 	count++;
5866 
5867 	count += DIV_ROUND_UP(len, adapter->tx_fifo_limit);
5868 
5869 	nr_frags = skb_shinfo(skb)->nr_frags;
5870 	for (f = 0; f < nr_frags; f++)
5871 		count += DIV_ROUND_UP(skb_frag_size(&skb_shinfo(skb)->frags[f]),
5872 				      adapter->tx_fifo_limit);
5873 
5874 	if (adapter->hw.mac.tx_pkt_filtering)
5875 		e1000_transfer_dhcp_info(adapter, skb);
5876 
5877 	/* need: count + 2 desc gap to keep tail from touching
5878 	 * head, otherwise try next time
5879 	 */
5880 	if (e1000_maybe_stop_tx(tx_ring, count + 2))
5881 		return NETDEV_TX_BUSY;
5882 
5883 	if (skb_vlan_tag_present(skb)) {
5884 		tx_flags |= E1000_TX_FLAGS_VLAN;
5885 		tx_flags |= (skb_vlan_tag_get(skb) <<
5886 			     E1000_TX_FLAGS_VLAN_SHIFT);
5887 	}
5888 
5889 	first = tx_ring->next_to_use;
5890 
5891 	tso = e1000_tso(tx_ring, skb, protocol);
5892 	if (tso < 0) {
5893 		dev_kfree_skb_any(skb);
5894 		return NETDEV_TX_OK;
5895 	}
5896 
5897 	if (tso)
5898 		tx_flags |= E1000_TX_FLAGS_TSO;
5899 	else if (e1000_tx_csum(tx_ring, skb, protocol))
5900 		tx_flags |= E1000_TX_FLAGS_CSUM;
5901 
5902 	/* Old method was to assume IPv4 packet by default if TSO was enabled.
5903 	 * 82571 hardware supports TSO capabilities for IPv6 as well...
5904 	 * no longer assume, we must.
5905 	 */
5906 	if (protocol == htons(ETH_P_IP))
5907 		tx_flags |= E1000_TX_FLAGS_IPV4;
5908 
5909 	if (unlikely(skb->no_fcs))
5910 		tx_flags |= E1000_TX_FLAGS_NO_FCS;
5911 
5912 	/* if count is 0 then mapping error has occurred */
5913 	count = e1000_tx_map(tx_ring, skb, first, adapter->tx_fifo_limit,
5914 			     nr_frags);
5915 	if (count) {
5916 		if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
5917 		    (adapter->flags & FLAG_HAS_HW_TIMESTAMP)) {
5918 			if (!adapter->tx_hwtstamp_skb) {
5919 				skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5920 				tx_flags |= E1000_TX_FLAGS_HWTSTAMP;
5921 				adapter->tx_hwtstamp_skb = skb_get(skb);
5922 				adapter->tx_hwtstamp_start = jiffies;
5923 				schedule_work(&adapter->tx_hwtstamp_work);
5924 			} else {
5925 				adapter->tx_hwtstamp_skipped++;
5926 			}
5927 		}
5928 
5929 		skb_tx_timestamp(skb);
5930 
5931 		netdev_sent_queue(netdev, skb->len);
5932 		e1000_tx_queue(tx_ring, tx_flags, count);
5933 		/* Make sure there is space in the ring for the next send. */
5934 		e1000_maybe_stop_tx(tx_ring,
5935 				    ((MAX_SKB_FRAGS + 1) *
5936 				     DIV_ROUND_UP(PAGE_SIZE,
5937 						  adapter->tx_fifo_limit) + 4));
5938 
5939 		if (!netdev_xmit_more() ||
5940 		    netif_xmit_stopped(netdev_get_tx_queue(netdev, 0))) {
5941 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
5942 				e1000e_update_tdt_wa(tx_ring,
5943 						     tx_ring->next_to_use);
5944 			else
5945 				writel(tx_ring->next_to_use, tx_ring->tail);
5946 		}
5947 	} else {
5948 		dev_kfree_skb_any(skb);
5949 		tx_ring->buffer_info[first].time_stamp = 0;
5950 		tx_ring->next_to_use = first;
5951 	}
5952 
5953 	return NETDEV_TX_OK;
5954 }
5955 
5956 /**
5957  * e1000_tx_timeout - Respond to a Tx Hang
5958  * @netdev: network interface device structure
5959  * @txqueue: index of the hung queue (unused)
5960  **/
5961 static void e1000_tx_timeout(struct net_device *netdev, unsigned int __always_unused txqueue)
5962 {
5963 	struct e1000_adapter *adapter = netdev_priv(netdev);
5964 
5965 	/* Do the reset outside of interrupt context */
5966 	adapter->tx_timeout_count++;
5967 	schedule_work(&adapter->reset_task);
5968 }
5969 
5970 static void e1000_reset_task(struct work_struct *work)
5971 {
5972 	struct e1000_adapter *adapter;
5973 	adapter = container_of(work, struct e1000_adapter, reset_task);
5974 
5975 	rtnl_lock();
5976 	/* don't run the task if already down */
5977 	if (test_bit(__E1000_DOWN, &adapter->state)) {
5978 		rtnl_unlock();
5979 		return;
5980 	}
5981 
5982 	if (!(adapter->flags & FLAG_RESTART_NOW)) {
5983 		e1000e_dump(adapter);
5984 		e_err("Reset adapter unexpectedly\n");
5985 	}
5986 	e1000e_reinit_locked(adapter);
5987 	rtnl_unlock();
5988 }
5989 
5990 /**
5991  * e1000e_get_stats64 - Get System Network Statistics
5992  * @netdev: network interface device structure
5993  * @stats: rtnl_link_stats64 pointer
5994  *
5995  * Returns the address of the device statistics structure.
5996  **/
5997 void e1000e_get_stats64(struct net_device *netdev,
5998 			struct rtnl_link_stats64 *stats)
5999 {
6000 	struct e1000_adapter *adapter = netdev_priv(netdev);
6001 
6002 	spin_lock(&adapter->stats64_lock);
6003 	e1000e_update_stats(adapter);
6004 	/* Fill out the OS statistics structure */
6005 	stats->rx_bytes = adapter->stats.gorc;
6006 	stats->rx_packets = adapter->stats.gprc;
6007 	stats->tx_bytes = adapter->stats.gotc;
6008 	stats->tx_packets = adapter->stats.gptc;
6009 	stats->multicast = adapter->stats.mprc;
6010 	stats->collisions = adapter->stats.colc;
6011 
6012 	/* Rx Errors */
6013 
6014 	/* RLEC on some newer hardware can be incorrect so build
6015 	 * our own version based on RUC and ROC
6016 	 */
6017 	stats->rx_errors = adapter->stats.rxerrc +
6018 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
6019 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
6020 	stats->rx_length_errors = adapter->stats.ruc + adapter->stats.roc;
6021 	stats->rx_crc_errors = adapter->stats.crcerrs;
6022 	stats->rx_frame_errors = adapter->stats.algnerrc;
6023 	stats->rx_missed_errors = adapter->stats.mpc;
6024 
6025 	/* Tx Errors */
6026 	stats->tx_errors = adapter->stats.ecol + adapter->stats.latecol;
6027 	stats->tx_aborted_errors = adapter->stats.ecol;
6028 	stats->tx_window_errors = adapter->stats.latecol;
6029 	stats->tx_carrier_errors = adapter->stats.tncrs;
6030 
6031 	/* Tx Dropped needs to be maintained elsewhere */
6032 
6033 	spin_unlock(&adapter->stats64_lock);
6034 }
6035 
6036 /**
6037  * e1000_change_mtu - Change the Maximum Transfer Unit
6038  * @netdev: network interface device structure
6039  * @new_mtu: new value for maximum frame size
6040  *
6041  * Returns 0 on success, negative on failure
6042  **/
6043 static int e1000_change_mtu(struct net_device *netdev, int new_mtu)
6044 {
6045 	struct e1000_adapter *adapter = netdev_priv(netdev);
6046 	int max_frame = new_mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
6047 
6048 	/* Jumbo frame support */
6049 	if ((new_mtu > ETH_DATA_LEN) &&
6050 	    !(adapter->flags & FLAG_HAS_JUMBO_FRAMES)) {
6051 		e_err("Jumbo Frames not supported.\n");
6052 		return -EINVAL;
6053 	}
6054 
6055 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
6056 	if ((adapter->hw.mac.type >= e1000_pch2lan) &&
6057 	    !(adapter->flags2 & FLAG2_CRC_STRIPPING) &&
6058 	    (new_mtu > ETH_DATA_LEN)) {
6059 		e_err("Jumbo Frames not supported on this device when CRC stripping is disabled.\n");
6060 		return -EINVAL;
6061 	}
6062 
6063 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
6064 		usleep_range(1000, 1100);
6065 	/* e1000e_down -> e1000e_reset dependent on max_frame_size & mtu */
6066 	adapter->max_frame_size = max_frame;
6067 	netdev_dbg(netdev, "changing MTU from %d to %d\n",
6068 		   netdev->mtu, new_mtu);
6069 	WRITE_ONCE(netdev->mtu, new_mtu);
6070 
6071 	pm_runtime_get_sync(netdev->dev.parent);
6072 
6073 	if (netif_running(netdev))
6074 		e1000e_down(adapter, true);
6075 
6076 	/* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
6077 	 * means we reserve 2 more, this pushes us to allocate from the next
6078 	 * larger slab size.
6079 	 * i.e. RXBUFFER_2048 --> size-4096 slab
6080 	 * However with the new *_jumbo_rx* routines, jumbo receives will use
6081 	 * fragmented skbs
6082 	 */
6083 
6084 	if (max_frame <= 2048)
6085 		adapter->rx_buffer_len = 2048;
6086 	else
6087 		adapter->rx_buffer_len = 4096;
6088 
6089 	/* adjust allocation if LPE protects us, and we aren't using SBP */
6090 	if (max_frame <= (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN))
6091 		adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
6092 
6093 	if (netif_running(netdev))
6094 		e1000e_up(adapter);
6095 	else
6096 		e1000e_reset(adapter);
6097 
6098 	pm_runtime_put_sync(netdev->dev.parent);
6099 
6100 	clear_bit(__E1000_RESETTING, &adapter->state);
6101 
6102 	return 0;
6103 }
6104 
6105 static int e1000_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
6106 {
6107 	struct e1000_adapter *adapter = netdev_priv(netdev);
6108 	struct mii_ioctl_data *data = if_mii(ifr);
6109 
6110 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
6111 		return -EOPNOTSUPP;
6112 
6113 	switch (cmd) {
6114 	case SIOCGMIIPHY:
6115 		data->phy_id = adapter->hw.phy.addr;
6116 		break;
6117 	case SIOCGMIIREG:
6118 		e1000_phy_read_status(adapter);
6119 
6120 		switch (data->reg_num & 0x1F) {
6121 		case MII_BMCR:
6122 			data->val_out = adapter->phy_regs.bmcr;
6123 			break;
6124 		case MII_BMSR:
6125 			data->val_out = adapter->phy_regs.bmsr;
6126 			break;
6127 		case MII_PHYSID1:
6128 			data->val_out = (adapter->hw.phy.id >> 16);
6129 			break;
6130 		case MII_PHYSID2:
6131 			data->val_out = (adapter->hw.phy.id & 0xFFFF);
6132 			break;
6133 		case MII_ADVERTISE:
6134 			data->val_out = adapter->phy_regs.advertise;
6135 			break;
6136 		case MII_LPA:
6137 			data->val_out = adapter->phy_regs.lpa;
6138 			break;
6139 		case MII_EXPANSION:
6140 			data->val_out = adapter->phy_regs.expansion;
6141 			break;
6142 		case MII_CTRL1000:
6143 			data->val_out = adapter->phy_regs.ctrl1000;
6144 			break;
6145 		case MII_STAT1000:
6146 			data->val_out = adapter->phy_regs.stat1000;
6147 			break;
6148 		case MII_ESTATUS:
6149 			data->val_out = adapter->phy_regs.estatus;
6150 			break;
6151 		default:
6152 			return -EIO;
6153 		}
6154 		break;
6155 	case SIOCSMIIREG:
6156 	default:
6157 		return -EOPNOTSUPP;
6158 	}
6159 	return 0;
6160 }
6161 
6162 /**
6163  * e1000e_hwtstamp_set - control hardware time stamping
6164  * @netdev: network interface device structure
6165  * @config: timestamp configuration
6166  * @extack: netlink extended ACK report
6167  *
6168  * Outgoing time stamping can be enabled and disabled. Play nice and
6169  * disable it when requested, although it shouldn't cause any overhead
6170  * when no packet needs it. At most one packet in the queue may be
6171  * marked for time stamping, otherwise it would be impossible to tell
6172  * for sure to which packet the hardware time stamp belongs.
6173  *
6174  * Incoming time stamping has to be configured via the hardware filters.
6175  * Not all combinations are supported, in particular event type has to be
6176  * specified. Matching the kind of event packet is not supported, with the
6177  * exception of "all V2 events regardless of level 2 or 4".
6178  **/
6179 static int e1000e_hwtstamp_set(struct net_device *netdev,
6180 			       struct kernel_hwtstamp_config *config,
6181 			       struct netlink_ext_ack *extack)
6182 {
6183 	struct e1000_adapter *adapter = netdev_priv(netdev);
6184 	int ret_val;
6185 
6186 	ret_val = e1000e_config_hwtstamp(adapter, config, extack);
6187 	if (ret_val)
6188 		return ret_val;
6189 
6190 	switch (config->rx_filter) {
6191 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
6192 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
6193 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
6194 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
6195 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
6196 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
6197 		/* With V2 type filters which specify a Sync or Delay Request,
6198 		 * Path Delay Request/Response messages are also time stamped
6199 		 * by hardware so notify the caller the requested packets plus
6200 		 * some others are time stamped.
6201 		 */
6202 		config->rx_filter = HWTSTAMP_FILTER_SOME;
6203 		break;
6204 	default:
6205 		break;
6206 	}
6207 
6208 	return 0;
6209 }
6210 
6211 static int e1000e_hwtstamp_get(struct net_device *netdev,
6212 			       struct kernel_hwtstamp_config *kernel_config)
6213 {
6214 	struct e1000_adapter *adapter = netdev_priv(netdev);
6215 
6216 	*kernel_config = adapter->hwtstamp_config;
6217 
6218 	return 0;
6219 }
6220 
6221 static int e1000_init_phy_wakeup(struct e1000_adapter *adapter, u32 wufc)
6222 {
6223 	struct e1000_hw *hw = &adapter->hw;
6224 	u32 i, mac_reg, wuc;
6225 	u16 phy_reg, wuc_enable;
6226 	int retval;
6227 
6228 	/* copy MAC RARs to PHY RARs */
6229 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
6230 
6231 	retval = hw->phy.ops.acquire(hw);
6232 	if (retval) {
6233 		e_err("Could not acquire PHY\n");
6234 		return retval;
6235 	}
6236 
6237 	/* Enable access to wakeup registers on and set page to BM_WUC_PAGE */
6238 	retval = e1000_enable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6239 	if (retval)
6240 		goto release;
6241 
6242 	/* copy MAC MTA to PHY MTA - only needed for pchlan */
6243 	for (i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
6244 		mac_reg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
6245 		hw->phy.ops.write_reg_page(hw, BM_MTA(i),
6246 					   (u16)(mac_reg & 0xFFFF));
6247 		hw->phy.ops.write_reg_page(hw, BM_MTA(i) + 1,
6248 					   (u16)((mac_reg >> 16) & 0xFFFF));
6249 	}
6250 
6251 	/* configure PHY Rx Control register */
6252 	hw->phy.ops.read_reg_page(&adapter->hw, BM_RCTL, &phy_reg);
6253 	mac_reg = er32(RCTL);
6254 	if (mac_reg & E1000_RCTL_UPE)
6255 		phy_reg |= BM_RCTL_UPE;
6256 	if (mac_reg & E1000_RCTL_MPE)
6257 		phy_reg |= BM_RCTL_MPE;
6258 	phy_reg &= ~(BM_RCTL_MO_MASK);
6259 	if (mac_reg & E1000_RCTL_MO_3)
6260 		phy_reg |= (FIELD_GET(E1000_RCTL_MO_3, mac_reg)
6261 			    << BM_RCTL_MO_SHIFT);
6262 	if (mac_reg & E1000_RCTL_BAM)
6263 		phy_reg |= BM_RCTL_BAM;
6264 	if (mac_reg & E1000_RCTL_PMCF)
6265 		phy_reg |= BM_RCTL_PMCF;
6266 	mac_reg = er32(CTRL);
6267 	if (mac_reg & E1000_CTRL_RFCE)
6268 		phy_reg |= BM_RCTL_RFCE;
6269 	hw->phy.ops.write_reg_page(&adapter->hw, BM_RCTL, phy_reg);
6270 
6271 	wuc = E1000_WUC_PME_EN;
6272 	if (wufc & (E1000_WUFC_MAG | E1000_WUFC_LNKC))
6273 		wuc |= E1000_WUC_APME;
6274 
6275 	/* enable PHY wakeup in MAC register */
6276 	ew32(WUFC, wufc);
6277 	ew32(WUC, (E1000_WUC_PHY_WAKE | E1000_WUC_APMPME |
6278 		   E1000_WUC_PME_STATUS | wuc));
6279 
6280 	/* configure and enable PHY wakeup in PHY registers */
6281 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUFC, wufc);
6282 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUC, wuc);
6283 
6284 	/* activate PHY wakeup */
6285 	wuc_enable |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
6286 	retval = e1000_disable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6287 	if (retval)
6288 		e_err("Could not set PHY Host Wakeup bit\n");
6289 release:
6290 	hw->phy.ops.release(hw);
6291 
6292 	return retval;
6293 }
6294 
6295 static void e1000e_flush_lpic(struct pci_dev *pdev)
6296 {
6297 	struct net_device *netdev = pci_get_drvdata(pdev);
6298 	struct e1000_adapter *adapter = netdev_priv(netdev);
6299 	struct e1000_hw *hw = &adapter->hw;
6300 	u32 ret_val;
6301 
6302 	pm_runtime_get_sync(netdev->dev.parent);
6303 
6304 	ret_val = hw->phy.ops.acquire(hw);
6305 	if (ret_val)
6306 		goto fl_out;
6307 
6308 	pr_info("EEE TX LPI TIMER: %08X\n",
6309 		er32(LPIC) >> E1000_LPIC_LPIET_SHIFT);
6310 
6311 	hw->phy.ops.release(hw);
6312 
6313 fl_out:
6314 	pm_runtime_put_sync(netdev->dev.parent);
6315 }
6316 
6317 /* S0ix implementation */
6318 static void e1000e_s0ix_entry_flow(struct e1000_adapter *adapter)
6319 {
6320 	struct e1000_hw *hw = &adapter->hw;
6321 	u32 mac_data;
6322 	u16 phy_data;
6323 
6324 	if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID &&
6325 	    hw->mac.type >= e1000_pch_adp) {
6326 		/* Request ME configure the device for S0ix */
6327 		mac_data = er32(H2ME);
6328 		mac_data |= E1000_H2ME_START_DPG;
6329 		mac_data &= ~E1000_H2ME_EXIT_DPG;
6330 		trace_e1000e_trace_mac_register(mac_data);
6331 		ew32(H2ME, mac_data);
6332 	} else {
6333 		/* Request driver configure the device to S0ix */
6334 		/* Disable the periodic inband message,
6335 		 * don't request PCIe clock in K1 page770_17[10:9] = 10b
6336 		 */
6337 		e1e_rphy(hw, HV_PM_CTRL, &phy_data);
6338 		phy_data &= ~HV_PM_CTRL_K1_CLK_REQ;
6339 		phy_data |= BIT(10);
6340 		e1e_wphy(hw, HV_PM_CTRL, phy_data);
6341 
6342 		/* Make sure we don't exit K1 every time a new packet arrives
6343 		 * 772_29[5] = 1 CS_Mode_Stay_In_K1
6344 		 */
6345 		e1e_rphy(hw, I217_CGFREG, &phy_data);
6346 		phy_data |= BIT(5);
6347 		e1e_wphy(hw, I217_CGFREG, phy_data);
6348 
6349 		/* Change the MAC/PHY interface to SMBus
6350 		 * Force the SMBus in PHY page769_23[0] = 1
6351 		 * Force the SMBus in MAC CTRL_EXT[11] = 1
6352 		 */
6353 		e1e_rphy(hw, CV_SMB_CTRL, &phy_data);
6354 		phy_data |= CV_SMB_CTRL_FORCE_SMBUS;
6355 		e1e_wphy(hw, CV_SMB_CTRL, phy_data);
6356 		mac_data = er32(CTRL_EXT);
6357 		mac_data |= E1000_CTRL_EXT_FORCE_SMBUS;
6358 		ew32(CTRL_EXT, mac_data);
6359 
6360 		/* DFT control: PHY bit: page769_20[0] = 1
6361 		 * page769_20[7] - PHY PLL stop
6362 		 * page769_20[8] - PHY go to the electrical idle
6363 		 * page769_20[9] - PHY serdes disable
6364 		 * Gate PPW via EXTCNF_CTRL - set 0x0F00[7] = 1
6365 		 */
6366 		e1e_rphy(hw, I82579_DFT_CTRL, &phy_data);
6367 		phy_data |= BIT(0);
6368 		phy_data |= BIT(7);
6369 		phy_data |= BIT(8);
6370 		phy_data |= BIT(9);
6371 		e1e_wphy(hw, I82579_DFT_CTRL, phy_data);
6372 
6373 		mac_data = er32(EXTCNF_CTRL);
6374 		mac_data |= E1000_EXTCNF_CTRL_GATE_PHY_CFG;
6375 		ew32(EXTCNF_CTRL, mac_data);
6376 
6377 		/* Disable disconnected cable conditioning for Power Gating */
6378 		mac_data = er32(DPGFR);
6379 		mac_data |= BIT(2);
6380 		ew32(DPGFR, mac_data);
6381 
6382 		/* Enable the Dynamic Clock Gating in the DMA and MAC */
6383 		mac_data = er32(CTRL_EXT);
6384 		mac_data |= E1000_CTRL_EXT_DMA_DYN_CLK_EN;
6385 		ew32(CTRL_EXT, mac_data);
6386 	}
6387 
6388 	/* Enable the Dynamic Power Gating in the MAC */
6389 	mac_data = er32(FEXTNVM7);
6390 	mac_data |= BIT(22);
6391 	ew32(FEXTNVM7, mac_data);
6392 
6393 	/* Don't wake from dynamic Power Gating with clock request */
6394 	mac_data = er32(FEXTNVM12);
6395 	mac_data |= BIT(12);
6396 	ew32(FEXTNVM12, mac_data);
6397 
6398 	/* Ungate PGCB clock */
6399 	mac_data = er32(FEXTNVM9);
6400 	mac_data &= ~BIT(28);
6401 	ew32(FEXTNVM9, mac_data);
6402 
6403 	/* Enable K1 off to enable mPHY Power Gating */
6404 	mac_data = er32(FEXTNVM6);
6405 	mac_data |= BIT(31);
6406 	ew32(FEXTNVM6, mac_data);
6407 
6408 	/* Enable mPHY power gating for any link and speed */
6409 	mac_data = er32(FEXTNVM8);
6410 	mac_data |= BIT(9);
6411 	ew32(FEXTNVM8, mac_data);
6412 
6413 	/* No MAC DPG gating SLP_S0 in modern standby
6414 	 * Switch the logic of the lanphypc to use PMC counter
6415 	 */
6416 	mac_data = er32(FEXTNVM5);
6417 	mac_data |= BIT(7);
6418 	ew32(FEXTNVM5, mac_data);
6419 
6420 	/* Disable the time synchronization clock */
6421 	mac_data = er32(FEXTNVM7);
6422 	mac_data |= BIT(31);
6423 	mac_data &= ~BIT(0);
6424 	ew32(FEXTNVM7, mac_data);
6425 
6426 	/* Dynamic Power Gating Enable */
6427 	mac_data = er32(CTRL_EXT);
6428 	mac_data |= BIT(3);
6429 	ew32(CTRL_EXT, mac_data);
6430 
6431 	/* Check MAC Tx/Rx packet buffer pointers.
6432 	 * Reset MAC Tx/Rx packet buffer pointers to suppress any
6433 	 * pending traffic indication that would prevent power gating.
6434 	 */
6435 	mac_data = er32(TDFH);
6436 	if (mac_data)
6437 		ew32(TDFH, 0);
6438 	mac_data = er32(TDFT);
6439 	if (mac_data)
6440 		ew32(TDFT, 0);
6441 	mac_data = er32(TDFHS);
6442 	if (mac_data)
6443 		ew32(TDFHS, 0);
6444 	mac_data = er32(TDFTS);
6445 	if (mac_data)
6446 		ew32(TDFTS, 0);
6447 	mac_data = er32(TDFPC);
6448 	if (mac_data)
6449 		ew32(TDFPC, 0);
6450 	mac_data = er32(RDFH);
6451 	if (mac_data)
6452 		ew32(RDFH, 0);
6453 	mac_data = er32(RDFT);
6454 	if (mac_data)
6455 		ew32(RDFT, 0);
6456 	mac_data = er32(RDFHS);
6457 	if (mac_data)
6458 		ew32(RDFHS, 0);
6459 	mac_data = er32(RDFTS);
6460 	if (mac_data)
6461 		ew32(RDFTS, 0);
6462 	mac_data = er32(RDFPC);
6463 	if (mac_data)
6464 		ew32(RDFPC, 0);
6465 }
6466 
6467 static void e1000e_s0ix_exit_flow(struct e1000_adapter *adapter)
6468 {
6469 	struct e1000_hw *hw = &adapter->hw;
6470 	bool firmware_bug = false;
6471 	u32 mac_data;
6472 	u16 phy_data;
6473 	u32 i = 0;
6474 
6475 	if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID &&
6476 	    hw->mac.type >= e1000_pch_adp) {
6477 		/* Keep the GPT clock enabled for CSME */
6478 		mac_data = er32(FEXTNVM);
6479 		mac_data |= BIT(3);
6480 		ew32(FEXTNVM, mac_data);
6481 		/* Request ME unconfigure the device from S0ix */
6482 		mac_data = er32(H2ME);
6483 		mac_data &= ~E1000_H2ME_START_DPG;
6484 		mac_data |= E1000_H2ME_EXIT_DPG;
6485 		trace_e1000e_trace_mac_register(mac_data);
6486 		ew32(H2ME, mac_data);
6487 
6488 		/* Poll up to 2.5 seconds for ME to unconfigure DPG.
6489 		 * If this takes more than 1 second, show a warning indicating a
6490 		 * firmware bug
6491 		 */
6492 		while (!(er32(EXFWSM) & E1000_EXFWSM_DPG_EXIT_DONE)) {
6493 			if (i > 100 && !firmware_bug)
6494 				firmware_bug = true;
6495 
6496 			if (i++ == 250) {
6497 				e_dbg("Timeout (firmware bug): %d msec\n",
6498 				      i * 10);
6499 				break;
6500 			}
6501 
6502 			usleep_range(10000, 11000);
6503 		}
6504 		if (firmware_bug)
6505 			e_warn("DPG_EXIT_DONE took %d msec. This is a firmware bug\n",
6506 			       i * 10);
6507 		else
6508 			e_dbg("DPG_EXIT_DONE cleared after %d msec\n", i * 10);
6509 	} else {
6510 		/* Request driver unconfigure the device from S0ix */
6511 
6512 		/* Cancel disable disconnected cable conditioning
6513 		 * for Power Gating
6514 		 */
6515 		mac_data = er32(DPGFR);
6516 		mac_data &= ~BIT(2);
6517 		ew32(DPGFR, mac_data);
6518 
6519 		/* Disable the Dynamic Clock Gating in the DMA and MAC */
6520 		mac_data = er32(CTRL_EXT);
6521 		mac_data &= 0xFFF7FFFF;
6522 		ew32(CTRL_EXT, mac_data);
6523 
6524 		/* Enable the periodic inband message,
6525 		 * Request PCIe clock in K1 page770_17[10:9] =01b
6526 		 */
6527 		e1e_rphy(hw, HV_PM_CTRL, &phy_data);
6528 		phy_data &= 0xFBFF;
6529 		phy_data |= HV_PM_CTRL_K1_CLK_REQ;
6530 		e1e_wphy(hw, HV_PM_CTRL, phy_data);
6531 
6532 		/* Return back configuration
6533 		 * 772_29[5] = 0 CS_Mode_Stay_In_K1
6534 		 */
6535 		e1e_rphy(hw, I217_CGFREG, &phy_data);
6536 		phy_data &= 0xFFDF;
6537 		e1e_wphy(hw, I217_CGFREG, phy_data);
6538 
6539 		/* Change the MAC/PHY interface to Kumeran
6540 		 * Unforce the SMBus in PHY page769_23[0] = 0
6541 		 * Unforce the SMBus in MAC CTRL_EXT[11] = 0
6542 		 */
6543 		e1e_rphy(hw, CV_SMB_CTRL, &phy_data);
6544 		phy_data &= ~CV_SMB_CTRL_FORCE_SMBUS;
6545 		e1e_wphy(hw, CV_SMB_CTRL, phy_data);
6546 		mac_data = er32(CTRL_EXT);
6547 		mac_data &= ~E1000_CTRL_EXT_FORCE_SMBUS;
6548 		ew32(CTRL_EXT, mac_data);
6549 	}
6550 
6551 	/* Disable Dynamic Power Gating */
6552 	mac_data = er32(CTRL_EXT);
6553 	mac_data &= 0xFFFFFFF7;
6554 	ew32(CTRL_EXT, mac_data);
6555 
6556 	/* Enable the time synchronization clock */
6557 	mac_data = er32(FEXTNVM7);
6558 	mac_data &= ~BIT(31);
6559 	mac_data |= BIT(0);
6560 	ew32(FEXTNVM7, mac_data);
6561 
6562 	/* Disable the Dynamic Power Gating in the MAC */
6563 	mac_data = er32(FEXTNVM7);
6564 	mac_data &= 0xFFBFFFFF;
6565 	ew32(FEXTNVM7, mac_data);
6566 
6567 	/* Disable mPHY power gating for any link and speed */
6568 	mac_data = er32(FEXTNVM8);
6569 	mac_data &= ~BIT(9);
6570 	ew32(FEXTNVM8, mac_data);
6571 
6572 	/* Disable K1 off */
6573 	mac_data = er32(FEXTNVM6);
6574 	mac_data &= ~BIT(31);
6575 	ew32(FEXTNVM6, mac_data);
6576 
6577 	/* Disable Ungate PGCB clock */
6578 	mac_data = er32(FEXTNVM9);
6579 	mac_data |= BIT(28);
6580 	ew32(FEXTNVM9, mac_data);
6581 
6582 	/* Cancel not waking from dynamic
6583 	 * Power Gating with clock request
6584 	 */
6585 	mac_data = er32(FEXTNVM12);
6586 	mac_data &= ~BIT(12);
6587 	ew32(FEXTNVM12, mac_data);
6588 
6589 	/* Revert the lanphypc logic to use the internal Gbe counter
6590 	 * and not the PMC counter
6591 	 */
6592 	mac_data = er32(FEXTNVM5);
6593 	mac_data &= 0xFFFFFF7F;
6594 	ew32(FEXTNVM5, mac_data);
6595 }
6596 
6597 static int e1000e_pm_freeze(struct device *dev)
6598 {
6599 	struct net_device *netdev = dev_get_drvdata(dev);
6600 	struct e1000_adapter *adapter = netdev_priv(netdev);
6601 	bool present;
6602 
6603 	rtnl_lock();
6604 
6605 	present = netif_device_present(netdev);
6606 	netif_device_detach(netdev);
6607 
6608 	if (present && netif_running(netdev)) {
6609 		int count = E1000_CHECK_RESET_COUNT;
6610 
6611 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6612 			usleep_range(10000, 11000);
6613 
6614 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6615 
6616 		/* Quiesce the device without resetting the hardware */
6617 		e1000e_down(adapter, false);
6618 		e1000_free_irq(adapter);
6619 	}
6620 	rtnl_unlock();
6621 
6622 	e1000e_reset_interrupt_capability(adapter);
6623 
6624 	/* Allow time for pending master requests to run */
6625 	e1000e_disable_pcie_master(&adapter->hw);
6626 
6627 	return 0;
6628 }
6629 
6630 static int __e1000_shutdown(struct pci_dev *pdev, bool runtime)
6631 {
6632 	struct net_device *netdev = pci_get_drvdata(pdev);
6633 	struct e1000_adapter *adapter = netdev_priv(netdev);
6634 	struct e1000_hw *hw = &adapter->hw;
6635 	u32 ctrl, ctrl_ext, rctl, status, wufc;
6636 	int retval = 0;
6637 
6638 	/* Runtime suspend should only enable wakeup for link changes */
6639 	if (runtime)
6640 		wufc = E1000_WUFC_LNKC;
6641 	else if (device_may_wakeup(&pdev->dev))
6642 		wufc = adapter->wol;
6643 	else
6644 		wufc = 0;
6645 
6646 	status = er32(STATUS);
6647 	if (status & E1000_STATUS_LU)
6648 		wufc &= ~E1000_WUFC_LNKC;
6649 
6650 	if (wufc) {
6651 		e1000_setup_rctl(adapter);
6652 		e1000e_set_rx_mode(netdev);
6653 
6654 		/* turn on all-multi mode if wake on multicast is enabled */
6655 		if (wufc & E1000_WUFC_MC) {
6656 			rctl = er32(RCTL);
6657 			rctl |= E1000_RCTL_MPE;
6658 			ew32(RCTL, rctl);
6659 		}
6660 
6661 		ctrl = er32(CTRL);
6662 		ctrl |= E1000_CTRL_ADVD3WUC;
6663 		if (!(adapter->flags2 & FLAG2_HAS_PHY_WAKEUP))
6664 			ctrl |= E1000_CTRL_EN_PHY_PWR_MGMT;
6665 		ew32(CTRL, ctrl);
6666 
6667 		if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
6668 		    adapter->hw.phy.media_type ==
6669 		    e1000_media_type_internal_serdes) {
6670 			/* keep the laser running in D3 */
6671 			ctrl_ext = er32(CTRL_EXT);
6672 			ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
6673 			ew32(CTRL_EXT, ctrl_ext);
6674 		}
6675 
6676 		if (!runtime)
6677 			e1000e_power_up_phy(adapter);
6678 
6679 		if (adapter->flags & FLAG_IS_ICH)
6680 			e1000_suspend_workarounds_ich8lan(&adapter->hw);
6681 
6682 		if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6683 			/* enable wakeup by the PHY */
6684 			retval = e1000_init_phy_wakeup(adapter, wufc);
6685 			if (retval) {
6686 				e_err("Failed to enable wakeup\n");
6687 				goto skip_phy_configurations;
6688 			}
6689 		} else {
6690 			/* enable wakeup by the MAC */
6691 			ew32(WUFC, wufc);
6692 			ew32(WUC, E1000_WUC_PME_EN);
6693 		}
6694 	} else {
6695 		ew32(WUC, 0);
6696 		ew32(WUFC, 0);
6697 
6698 		e1000_power_down_phy(adapter);
6699 	}
6700 
6701 	if (adapter->hw.phy.type == e1000_phy_igp_3) {
6702 		e1000e_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
6703 	} else if (hw->mac.type >= e1000_pch_lpt) {
6704 		if (wufc && !(wufc & (E1000_WUFC_EX | E1000_WUFC_MC | E1000_WUFC_BC))) {
6705 			/* ULP does not support wake from unicast, multicast
6706 			 * or broadcast.
6707 			 */
6708 			retval = e1000_enable_ulp_lpt_lp(hw, !runtime);
6709 			if (retval) {
6710 				e_err("Failed to enable ULP\n");
6711 				goto skip_phy_configurations;
6712 			}
6713 		}
6714 	}
6715 
6716 	/* Ensure that the appropriate bits are set in LPI_CTRL
6717 	 * for EEE in Sx
6718 	 */
6719 	if ((hw->phy.type >= e1000_phy_i217) &&
6720 	    adapter->eee_advert && hw->dev_spec.ich8lan.eee_lp_ability) {
6721 		u16 lpi_ctrl = 0;
6722 
6723 		retval = hw->phy.ops.acquire(hw);
6724 		if (!retval) {
6725 			retval = e1e_rphy_locked(hw, I82579_LPI_CTRL,
6726 						 &lpi_ctrl);
6727 			if (!retval) {
6728 				if (adapter->eee_advert &
6729 				    hw->dev_spec.ich8lan.eee_lp_ability &
6730 				    I82579_EEE_100_SUPPORTED)
6731 					lpi_ctrl |= I82579_LPI_CTRL_100_ENABLE;
6732 				if (adapter->eee_advert &
6733 				    hw->dev_spec.ich8lan.eee_lp_ability &
6734 				    I82579_EEE_1000_SUPPORTED)
6735 					lpi_ctrl |= I82579_LPI_CTRL_1000_ENABLE;
6736 
6737 				retval = e1e_wphy_locked(hw, I82579_LPI_CTRL,
6738 							 lpi_ctrl);
6739 			}
6740 		}
6741 		hw->phy.ops.release(hw);
6742 	}
6743 
6744 skip_phy_configurations:
6745 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
6746 	 * would have already happened in close and is redundant.
6747 	 */
6748 	e1000e_release_hw_control(adapter);
6749 
6750 	pci_clear_master(pdev);
6751 
6752 	/* The pci-e switch on some quad port adapters will report a
6753 	 * correctable error when the MAC transitions from D0 to D3.  To
6754 	 * prevent this we need to mask off the correctable errors on the
6755 	 * downstream port of the pci-e switch.
6756 	 *
6757 	 * We don't have the associated upstream bridge while assigning
6758 	 * the PCI device into guest. For example, the KVM on power is
6759 	 * one of the cases.
6760 	 */
6761 	if (adapter->flags & FLAG_IS_QUAD_PORT) {
6762 		struct pci_dev *us_dev = pdev->bus->self;
6763 		u16 devctl;
6764 
6765 		if (!us_dev)
6766 			return 0;
6767 
6768 		pcie_capability_read_word(us_dev, PCI_EXP_DEVCTL, &devctl);
6769 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL,
6770 					   (devctl & ~PCI_EXP_DEVCTL_CERE));
6771 
6772 		pci_save_state(pdev);
6773 		pci_prepare_to_sleep(pdev);
6774 
6775 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL, devctl);
6776 	}
6777 
6778 	return 0;
6779 }
6780 
6781 /**
6782  * __e1000e_disable_aspm - Disable ASPM states
6783  * @pdev: pointer to PCI device struct
6784  * @state: bit-mask of ASPM states to disable
6785  * @locked: indication if this context holds pci_bus_sem locked.
6786  *
6787  * Some devices *must* have certain ASPM states disabled per hardware errata.
6788  **/
6789 static void __e1000e_disable_aspm(struct pci_dev *pdev, u16 state, int locked)
6790 {
6791 	struct pci_dev *parent = pdev->bus->self;
6792 	u16 aspm_dis_mask = 0;
6793 	u16 pdev_aspmc, parent_aspmc;
6794 
6795 	switch (state) {
6796 	case PCIE_LINK_STATE_L0S:
6797 	case PCIE_LINK_STATE_L0S | PCIE_LINK_STATE_L1:
6798 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L0S;
6799 		fallthrough; /* can't have L1 without L0s */
6800 	case PCIE_LINK_STATE_L1:
6801 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L1;
6802 		break;
6803 	default:
6804 		return;
6805 	}
6806 
6807 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6808 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6809 
6810 	if (parent) {
6811 		pcie_capability_read_word(parent, PCI_EXP_LNKCTL,
6812 					  &parent_aspmc);
6813 		parent_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6814 	}
6815 
6816 	/* Nothing to do if the ASPM states to be disabled already are */
6817 	if (!(pdev_aspmc & aspm_dis_mask) &&
6818 	    (!parent || !(parent_aspmc & aspm_dis_mask)))
6819 		return;
6820 
6821 	dev_info(&pdev->dev, "Disabling ASPM %s %s\n",
6822 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L0S) ?
6823 		 "L0s" : "",
6824 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L1) ?
6825 		 "L1" : "");
6826 
6827 #ifdef CONFIG_PCIEASPM
6828 	if (locked)
6829 		pci_disable_link_state_locked(pdev, state);
6830 	else
6831 		pci_disable_link_state(pdev, state);
6832 
6833 	/* Double-check ASPM control.  If not disabled by the above, the
6834 	 * BIOS is preventing that from happening (or CONFIG_PCIEASPM is
6835 	 * not enabled); override by writing PCI config space directly.
6836 	 */
6837 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6838 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6839 
6840 	if (!(aspm_dis_mask & pdev_aspmc))
6841 		return;
6842 #endif
6843 
6844 	/* Both device and parent should have the same ASPM setting.
6845 	 * Disable ASPM in downstream component first and then upstream.
6846 	 */
6847 	pcie_capability_clear_word(pdev, PCI_EXP_LNKCTL, aspm_dis_mask);
6848 
6849 	if (parent)
6850 		pcie_capability_clear_word(parent, PCI_EXP_LNKCTL,
6851 					   aspm_dis_mask);
6852 }
6853 
6854 /**
6855  * e1000e_disable_aspm - Disable ASPM states.
6856  * @pdev: pointer to PCI device struct
6857  * @state: bit-mask of ASPM states to disable
6858  *
6859  * This function acquires the pci_bus_sem!
6860  * Some devices *must* have certain ASPM states disabled per hardware errata.
6861  **/
6862 static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6863 {
6864 	__e1000e_disable_aspm(pdev, state, 0);
6865 }
6866 
6867 /**
6868  * e1000e_disable_aspm_locked - Disable ASPM states.
6869  * @pdev: pointer to PCI device struct
6870  * @state: bit-mask of ASPM states to disable
6871  *
6872  * This function must be called with pci_bus_sem acquired!
6873  * Some devices *must* have certain ASPM states disabled per hardware errata.
6874  **/
6875 static void e1000e_disable_aspm_locked(struct pci_dev *pdev, u16 state)
6876 {
6877 	__e1000e_disable_aspm(pdev, state, 1);
6878 }
6879 
6880 static int e1000e_pm_thaw(struct device *dev)
6881 {
6882 	struct net_device *netdev = dev_get_drvdata(dev);
6883 	struct e1000_adapter *adapter = netdev_priv(netdev);
6884 	int rc = 0;
6885 
6886 	e1000e_set_interrupt_capability(adapter);
6887 
6888 	rtnl_lock();
6889 	if (netif_running(netdev)) {
6890 		rc = e1000_request_irq(adapter);
6891 		if (rc)
6892 			goto err_irq;
6893 
6894 		e1000e_up(adapter);
6895 	}
6896 
6897 	netif_device_attach(netdev);
6898 err_irq:
6899 	rtnl_unlock();
6900 
6901 	return rc;
6902 }
6903 
6904 static int __e1000_resume(struct pci_dev *pdev)
6905 {
6906 	struct net_device *netdev = pci_get_drvdata(pdev);
6907 	struct e1000_adapter *adapter = netdev_priv(netdev);
6908 	struct e1000_hw *hw = &adapter->hw;
6909 	u16 aspm_disable_flag = 0;
6910 
6911 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6912 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6913 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6914 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6915 	if (aspm_disable_flag)
6916 		e1000e_disable_aspm(pdev, aspm_disable_flag);
6917 
6918 	pci_set_master(pdev);
6919 
6920 	if (hw->mac.type >= e1000_pch2lan)
6921 		e1000_resume_workarounds_pchlan(&adapter->hw);
6922 
6923 	e1000e_power_up_phy(adapter);
6924 
6925 	/* report the system wakeup cause from S3/S4 */
6926 	if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6927 		u16 phy_data;
6928 
6929 		e1e_rphy(&adapter->hw, BM_WUS, &phy_data);
6930 		if (phy_data) {
6931 			e_info("PHY Wakeup cause - %s\n",
6932 			       phy_data & E1000_WUS_EX ? "Unicast Packet" :
6933 			       phy_data & E1000_WUS_MC ? "Multicast Packet" :
6934 			       phy_data & E1000_WUS_BC ? "Broadcast Packet" :
6935 			       phy_data & E1000_WUS_MAG ? "Magic Packet" :
6936 			       phy_data & E1000_WUS_LNKC ?
6937 			       "Link Status Change" : "other");
6938 		}
6939 		e1e_wphy(&adapter->hw, BM_WUS, ~0);
6940 	} else {
6941 		u32 wus = er32(WUS);
6942 
6943 		if (wus) {
6944 			e_info("MAC Wakeup cause - %s\n",
6945 			       wus & E1000_WUS_EX ? "Unicast Packet" :
6946 			       wus & E1000_WUS_MC ? "Multicast Packet" :
6947 			       wus & E1000_WUS_BC ? "Broadcast Packet" :
6948 			       wus & E1000_WUS_MAG ? "Magic Packet" :
6949 			       wus & E1000_WUS_LNKC ? "Link Status Change" :
6950 			       "other");
6951 		}
6952 		ew32(WUS, ~0);
6953 	}
6954 
6955 	e1000e_reset(adapter);
6956 
6957 	e1000_init_manageability_pt(adapter);
6958 
6959 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6960 	 * is up.  For all other cases, let the f/w know that the h/w is now
6961 	 * under the control of the driver.
6962 	 */
6963 	if (!(adapter->flags & FLAG_HAS_AMT))
6964 		e1000e_get_hw_control(adapter);
6965 
6966 	return 0;
6967 }
6968 
6969 static int e1000e_pm_prepare(struct device *dev)
6970 {
6971 	return pm_runtime_suspended(dev) &&
6972 		pm_suspend_via_firmware();
6973 }
6974 
6975 static int e1000e_pm_suspend(struct device *dev)
6976 {
6977 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6978 	struct e1000_adapter *adapter = netdev_priv(netdev);
6979 	struct pci_dev *pdev = to_pci_dev(dev);
6980 	int rc;
6981 
6982 	e1000e_flush_lpic(pdev);
6983 
6984 	e1000e_pm_freeze(dev);
6985 
6986 	rc = __e1000_shutdown(pdev, false);
6987 	if (!rc) {
6988 		/* Introduce S0ix implementation */
6989 		if (adapter->flags2 & FLAG2_ENABLE_S0IX_FLOWS)
6990 			e1000e_s0ix_entry_flow(adapter);
6991 	}
6992 
6993 	return 0;
6994 }
6995 
6996 static int e1000e_pm_resume(struct device *dev)
6997 {
6998 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6999 	struct e1000_adapter *adapter = netdev_priv(netdev);
7000 	struct pci_dev *pdev = to_pci_dev(dev);
7001 	int rc;
7002 
7003 	/* Introduce S0ix implementation */
7004 	if (adapter->flags2 & FLAG2_ENABLE_S0IX_FLOWS)
7005 		e1000e_s0ix_exit_flow(adapter);
7006 
7007 	rc = __e1000_resume(pdev);
7008 	if (rc)
7009 		return rc;
7010 
7011 	return e1000e_pm_thaw(dev);
7012 }
7013 
7014 static __maybe_unused int e1000e_pm_runtime_idle(struct device *dev)
7015 {
7016 	struct net_device *netdev = dev_get_drvdata(dev);
7017 	struct e1000_adapter *adapter = netdev_priv(netdev);
7018 	u16 eee_lp;
7019 
7020 	eee_lp = adapter->hw.dev_spec.ich8lan.eee_lp_ability;
7021 
7022 	if (!e1000e_has_link(adapter)) {
7023 		adapter->hw.dev_spec.ich8lan.eee_lp_ability = eee_lp;
7024 		pm_schedule_suspend(dev, 5 * MSEC_PER_SEC);
7025 	}
7026 
7027 	return -EBUSY;
7028 }
7029 
7030 static int e1000e_pm_runtime_resume(struct device *dev)
7031 {
7032 	struct pci_dev *pdev = to_pci_dev(dev);
7033 	struct net_device *netdev = pci_get_drvdata(pdev);
7034 	struct e1000_adapter *adapter = netdev_priv(netdev);
7035 	int rc;
7036 
7037 	pdev->pme_poll = true;
7038 
7039 	rc = __e1000_resume(pdev);
7040 	if (rc)
7041 		return rc;
7042 
7043 	if (netdev->flags & IFF_UP)
7044 		e1000e_up(adapter);
7045 
7046 	return rc;
7047 }
7048 
7049 static int e1000e_pm_runtime_suspend(struct device *dev)
7050 {
7051 	struct pci_dev *pdev = to_pci_dev(dev);
7052 	struct net_device *netdev = pci_get_drvdata(pdev);
7053 	struct e1000_adapter *adapter = netdev_priv(netdev);
7054 
7055 	if (netdev->flags & IFF_UP) {
7056 		int count = E1000_CHECK_RESET_COUNT;
7057 
7058 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
7059 			usleep_range(10000, 11000);
7060 
7061 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
7062 
7063 		/* Down the device without resetting the hardware */
7064 		e1000e_down(adapter, false);
7065 	}
7066 
7067 	if (__e1000_shutdown(pdev, true)) {
7068 		e1000e_pm_runtime_resume(dev);
7069 		return -EBUSY;
7070 	}
7071 
7072 	return 0;
7073 }
7074 
7075 static void e1000_shutdown(struct pci_dev *pdev)
7076 {
7077 	e1000e_flush_lpic(pdev);
7078 
7079 	e1000e_pm_freeze(&pdev->dev);
7080 
7081 	__e1000_shutdown(pdev, false);
7082 }
7083 
7084 #ifdef CONFIG_NET_POLL_CONTROLLER
7085 
7086 static irqreturn_t e1000_intr_msix(int __always_unused irq, void *data)
7087 {
7088 	struct net_device *netdev = data;
7089 	struct e1000_adapter *adapter = netdev_priv(netdev);
7090 
7091 	if (adapter->msix_entries) {
7092 		int vector, msix_irq;
7093 
7094 		vector = 0;
7095 		msix_irq = adapter->msix_entries[vector].vector;
7096 		if (disable_hardirq(msix_irq))
7097 			e1000_intr_msix_rx(msix_irq, netdev);
7098 		enable_irq(msix_irq);
7099 
7100 		vector++;
7101 		msix_irq = adapter->msix_entries[vector].vector;
7102 		if (disable_hardirq(msix_irq))
7103 			e1000_intr_msix_tx(msix_irq, netdev);
7104 		enable_irq(msix_irq);
7105 
7106 		vector++;
7107 		msix_irq = adapter->msix_entries[vector].vector;
7108 		if (disable_hardirq(msix_irq))
7109 			e1000_msix_other(msix_irq, netdev);
7110 		enable_irq(msix_irq);
7111 	}
7112 
7113 	return IRQ_HANDLED;
7114 }
7115 
7116 /**
7117  * e1000_netpoll
7118  * @netdev: network interface device structure
7119  *
7120  * Polling 'interrupt' - used by things like netconsole to send skbs
7121  * without having to re-enable interrupts. It's not called while
7122  * the interrupt routine is executing.
7123  */
7124 static void e1000_netpoll(struct net_device *netdev)
7125 {
7126 	struct e1000_adapter *adapter = netdev_priv(netdev);
7127 
7128 	switch (adapter->int_mode) {
7129 	case E1000E_INT_MODE_MSIX:
7130 		e1000_intr_msix(adapter->pdev->irq, netdev);
7131 		break;
7132 	case E1000E_INT_MODE_MSI:
7133 		if (disable_hardirq(adapter->pdev->irq))
7134 			e1000_intr_msi(adapter->pdev->irq, netdev);
7135 		enable_irq(adapter->pdev->irq);
7136 		break;
7137 	default:		/* E1000E_INT_MODE_LEGACY */
7138 		if (disable_hardirq(adapter->pdev->irq))
7139 			e1000_intr(adapter->pdev->irq, netdev);
7140 		enable_irq(adapter->pdev->irq);
7141 		break;
7142 	}
7143 }
7144 #endif
7145 
7146 /**
7147  * e1000_io_error_detected - called when PCI error is detected
7148  * @pdev: Pointer to PCI device
7149  * @state: The current pci connection state
7150  *
7151  * This function is called after a PCI bus error affecting
7152  * this device has been detected.
7153  */
7154 static pci_ers_result_t e1000_io_error_detected(struct pci_dev *pdev,
7155 						pci_channel_state_t state)
7156 {
7157 	e1000e_pm_freeze(&pdev->dev);
7158 
7159 	if (state == pci_channel_io_perm_failure)
7160 		return PCI_ERS_RESULT_DISCONNECT;
7161 
7162 	pci_disable_device(pdev);
7163 
7164 	/* Request a slot reset. */
7165 	return PCI_ERS_RESULT_NEED_RESET;
7166 }
7167 
7168 /**
7169  * e1000_io_slot_reset - called after the pci bus has been reset.
7170  * @pdev: Pointer to PCI device
7171  *
7172  * Restart the card from scratch, as if from a cold-boot. Implementation
7173  * resembles the first-half of the e1000e_pm_resume routine.
7174  */
7175 static pci_ers_result_t e1000_io_slot_reset(struct pci_dev *pdev)
7176 {
7177 	struct net_device *netdev = pci_get_drvdata(pdev);
7178 	struct e1000_adapter *adapter = netdev_priv(netdev);
7179 	struct e1000_hw *hw = &adapter->hw;
7180 	u16 aspm_disable_flag = 0;
7181 	int err;
7182 	pci_ers_result_t result;
7183 
7184 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
7185 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7186 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
7187 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7188 	if (aspm_disable_flag)
7189 		e1000e_disable_aspm_locked(pdev, aspm_disable_flag);
7190 
7191 	err = pci_enable_device_mem(pdev);
7192 	if (err) {
7193 		dev_err(&pdev->dev,
7194 			"Cannot re-enable PCI device after reset.\n");
7195 		result = PCI_ERS_RESULT_DISCONNECT;
7196 	} else {
7197 		pci_restore_state(pdev);
7198 		pci_set_master(pdev);
7199 
7200 		pci_enable_wake(pdev, PCI_D3hot, 0);
7201 		pci_enable_wake(pdev, PCI_D3cold, 0);
7202 
7203 		e1000e_reset(adapter);
7204 		ew32(WUS, ~0);
7205 		result = PCI_ERS_RESULT_RECOVERED;
7206 	}
7207 
7208 	return result;
7209 }
7210 
7211 /**
7212  * e1000_io_resume - called when traffic can start flowing again.
7213  * @pdev: Pointer to PCI device
7214  *
7215  * This callback is called when the error recovery driver tells us that
7216  * its OK to resume normal operation. Implementation resembles the
7217  * second-half of the e1000e_pm_resume routine.
7218  */
7219 static void e1000_io_resume(struct pci_dev *pdev)
7220 {
7221 	struct net_device *netdev = pci_get_drvdata(pdev);
7222 	struct e1000_adapter *adapter = netdev_priv(netdev);
7223 
7224 	e1000_init_manageability_pt(adapter);
7225 
7226 	e1000e_pm_thaw(&pdev->dev);
7227 
7228 	/* If the controller has AMT, do not set DRV_LOAD until the interface
7229 	 * is up.  For all other cases, let the f/w know that the h/w is now
7230 	 * under the control of the driver.
7231 	 */
7232 	if (!(adapter->flags & FLAG_HAS_AMT))
7233 		e1000e_get_hw_control(adapter);
7234 }
7235 
7236 static void e1000_print_device_info(struct e1000_adapter *adapter)
7237 {
7238 	struct e1000_hw *hw = &adapter->hw;
7239 	struct net_device *netdev = adapter->netdev;
7240 	u32 ret_val;
7241 	u8 pba_str[E1000_PBANUM_LENGTH];
7242 
7243 	/* print bus type/speed/width info */
7244 	e_info("(PCI Express:2.5GT/s:%s) %pM\n",
7245 	       /* bus width */
7246 	       ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
7247 		"Width x1"),
7248 	       /* MAC address */
7249 	       netdev->dev_addr);
7250 	e_info("Intel(R) PRO/%s Network Connection\n",
7251 	       (hw->phy.type == e1000_phy_ife) ? "10/100" : "1000");
7252 	ret_val = e1000_read_pba_string_generic(hw, pba_str,
7253 						E1000_PBANUM_LENGTH);
7254 	if (ret_val)
7255 		strscpy((char *)pba_str, "Unknown", sizeof(pba_str));
7256 	e_info("MAC: %d, PHY: %d, PBA No: %s\n",
7257 	       hw->mac.type, hw->phy.type, pba_str);
7258 }
7259 
7260 static void e1000_eeprom_checks(struct e1000_adapter *adapter)
7261 {
7262 	struct e1000_hw *hw = &adapter->hw;
7263 	int ret_val;
7264 	u16 buf = 0;
7265 
7266 	if (hw->mac.type != e1000_82573)
7267 		return;
7268 
7269 	ret_val = e1000_read_nvm(hw, NVM_INIT_CONTROL2_REG, 1, &buf);
7270 	le16_to_cpus(&buf);
7271 	if (!ret_val && (!(buf & BIT(0)))) {
7272 		/* Deep Smart Power Down (DSPD) */
7273 		dev_warn(&adapter->pdev->dev,
7274 			 "Warning: detected DSPD enabled in EEPROM\n");
7275 	}
7276 }
7277 
7278 static netdev_features_t e1000_fix_features(struct net_device *netdev,
7279 					    netdev_features_t features)
7280 {
7281 	struct e1000_adapter *adapter = netdev_priv(netdev);
7282 	struct e1000_hw *hw = &adapter->hw;
7283 
7284 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
7285 	if ((hw->mac.type >= e1000_pch2lan) && (netdev->mtu > ETH_DATA_LEN))
7286 		features &= ~NETIF_F_RXFCS;
7287 
7288 	/* Since there is no support for separate Rx/Tx vlan accel
7289 	 * enable/disable make sure Tx flag is always in same state as Rx.
7290 	 */
7291 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
7292 		features |= NETIF_F_HW_VLAN_CTAG_TX;
7293 	else
7294 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
7295 
7296 	return features;
7297 }
7298 
7299 static int e1000_set_features(struct net_device *netdev,
7300 			      netdev_features_t features)
7301 {
7302 	struct e1000_adapter *adapter = netdev_priv(netdev);
7303 	netdev_features_t changed = features ^ netdev->features;
7304 
7305 	if (changed & (NETIF_F_TSO | NETIF_F_TSO6))
7306 		adapter->flags |= FLAG_TSO_FORCE;
7307 
7308 	if (!(changed & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX |
7309 			 NETIF_F_RXCSUM | NETIF_F_RXHASH | NETIF_F_RXFCS |
7310 			 NETIF_F_RXALL)))
7311 		return 0;
7312 
7313 	if (changed & NETIF_F_RXFCS) {
7314 		if (features & NETIF_F_RXFCS) {
7315 			adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7316 		} else {
7317 			/* We need to take it back to defaults, which might mean
7318 			 * stripping is still disabled at the adapter level.
7319 			 */
7320 			if (adapter->flags2 & FLAG2_DFLT_CRC_STRIPPING)
7321 				adapter->flags2 |= FLAG2_CRC_STRIPPING;
7322 			else
7323 				adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7324 		}
7325 	}
7326 
7327 	netdev->features = features;
7328 
7329 	if (netif_running(netdev))
7330 		e1000e_reinit_locked(adapter);
7331 	else
7332 		e1000e_reset(adapter);
7333 
7334 	return 1;
7335 }
7336 
7337 static const struct net_device_ops e1000e_netdev_ops = {
7338 	.ndo_open		= e1000e_open,
7339 	.ndo_stop		= e1000e_close,
7340 	.ndo_start_xmit		= e1000_xmit_frame,
7341 	.ndo_get_stats64	= e1000e_get_stats64,
7342 	.ndo_set_rx_mode	= e1000e_set_rx_mode,
7343 	.ndo_set_mac_address	= e1000_set_mac,
7344 	.ndo_change_mtu		= e1000_change_mtu,
7345 	.ndo_eth_ioctl		= e1000_ioctl,
7346 	.ndo_tx_timeout		= e1000_tx_timeout,
7347 	.ndo_validate_addr	= eth_validate_addr,
7348 
7349 	.ndo_vlan_rx_add_vid	= e1000_vlan_rx_add_vid,
7350 	.ndo_vlan_rx_kill_vid	= e1000_vlan_rx_kill_vid,
7351 #ifdef CONFIG_NET_POLL_CONTROLLER
7352 	.ndo_poll_controller	= e1000_netpoll,
7353 #endif
7354 	.ndo_set_features	= e1000_set_features,
7355 	.ndo_fix_features	= e1000_fix_features,
7356 	.ndo_features_check	= passthru_features_check,
7357 	.ndo_hwtstamp_get	= e1000e_hwtstamp_get,
7358 	.ndo_hwtstamp_set	= e1000e_hwtstamp_set,
7359 };
7360 
7361 /**
7362  * e1000_probe - Device Initialization Routine
7363  * @pdev: PCI device information struct
7364  * @ent: entry in e1000_pci_tbl
7365  *
7366  * Returns 0 on success, negative on failure
7367  *
7368  * e1000_probe initializes an adapter identified by a pci_dev structure.
7369  * The OS initialization, configuring of the adapter private structure,
7370  * and a hardware reset occur.
7371  **/
7372 static int e1000_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7373 {
7374 	struct net_device *netdev;
7375 	struct e1000_adapter *adapter;
7376 	struct e1000_hw *hw;
7377 	const struct e1000_info *ei = e1000_info_tbl[ent->driver_data];
7378 	resource_size_t mmio_start, mmio_len;
7379 	resource_size_t flash_start, flash_len;
7380 	static int cards_found;
7381 	u16 aspm_disable_flag = 0;
7382 	u16 eeprom_data = 0;
7383 	u16 eeprom_apme_mask = E1000_EEPROM_APME;
7384 	int bars, i, err;
7385 	s32 ret_val = 0;
7386 
7387 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L0S)
7388 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7389 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L1)
7390 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7391 	if (aspm_disable_flag)
7392 		e1000e_disable_aspm(pdev, aspm_disable_flag);
7393 
7394 	err = pci_enable_device_mem(pdev);
7395 	if (err)
7396 		return err;
7397 
7398 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
7399 	if (err) {
7400 		dev_err(&pdev->dev,
7401 			"No usable DMA configuration, aborting\n");
7402 		goto err_dma;
7403 	}
7404 
7405 	bars = pci_select_bars(pdev, IORESOURCE_MEM);
7406 	err = pci_request_selected_regions_exclusive(pdev, bars,
7407 						     e1000e_driver_name);
7408 	if (err)
7409 		goto err_pci_reg;
7410 
7411 	pci_set_master(pdev);
7412 	/* PCI config space info */
7413 	err = pci_save_state(pdev);
7414 	if (err)
7415 		goto err_alloc_etherdev;
7416 
7417 	err = -ENOMEM;
7418 	netdev = alloc_etherdev(sizeof(struct e1000_adapter));
7419 	if (!netdev)
7420 		goto err_alloc_etherdev;
7421 
7422 	SET_NETDEV_DEV(netdev, &pdev->dev);
7423 
7424 	netdev->irq = pdev->irq;
7425 
7426 	pci_set_drvdata(pdev, netdev);
7427 	adapter = netdev_priv(netdev);
7428 	hw = &adapter->hw;
7429 	adapter->netdev = netdev;
7430 	adapter->pdev = pdev;
7431 	adapter->ei = ei;
7432 	adapter->pba = ei->pba;
7433 	adapter->flags = ei->flags;
7434 	adapter->flags2 = ei->flags2;
7435 	adapter->hw.adapter = adapter;
7436 	adapter->hw.mac.type = ei->mac;
7437 	adapter->max_hw_frame_size = ei->max_hw_frame_size;
7438 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
7439 
7440 	mmio_start = pci_resource_start(pdev, 0);
7441 	mmio_len = pci_resource_len(pdev, 0);
7442 
7443 	err = -EIO;
7444 	adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
7445 	if (!adapter->hw.hw_addr)
7446 		goto err_ioremap;
7447 
7448 	if ((adapter->flags & FLAG_HAS_FLASH) &&
7449 	    (pci_resource_flags(pdev, 1) & IORESOURCE_MEM) &&
7450 	    (hw->mac.type < e1000_pch_spt)) {
7451 		flash_start = pci_resource_start(pdev, 1);
7452 		flash_len = pci_resource_len(pdev, 1);
7453 		adapter->hw.flash_address = ioremap(flash_start, flash_len);
7454 		if (!adapter->hw.flash_address)
7455 			goto err_flashmap;
7456 	}
7457 
7458 	/* Set default EEE advertisement */
7459 	if (adapter->flags2 & FLAG2_HAS_EEE)
7460 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
7461 
7462 	/* construct the net_device struct */
7463 	netdev->netdev_ops = &e1000e_netdev_ops;
7464 	e1000e_set_ethtool_ops(netdev);
7465 	netdev->watchdog_timeo = 5 * HZ;
7466 	netif_napi_add(netdev, &adapter->napi, e1000e_poll);
7467 	strscpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
7468 
7469 	netdev->mem_start = mmio_start;
7470 	netdev->mem_end = mmio_start + mmio_len;
7471 
7472 	adapter->bd_number = cards_found++;
7473 
7474 	e1000e_check_options(adapter);
7475 
7476 	/* setup adapter struct */
7477 	err = e1000_sw_init(adapter);
7478 	if (err)
7479 		goto err_sw_init;
7480 
7481 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
7482 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
7483 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
7484 
7485 	err = ei->get_variants(adapter);
7486 	if (err)
7487 		goto err_hw_init;
7488 
7489 	if ((adapter->flags & FLAG_IS_ICH) &&
7490 	    (adapter->flags & FLAG_READ_ONLY_NVM) &&
7491 	    (hw->mac.type < e1000_pch_spt))
7492 		e1000e_write_protect_nvm_ich8lan(&adapter->hw);
7493 
7494 	hw->mac.ops.get_bus_info(&adapter->hw);
7495 
7496 	adapter->hw.phy.autoneg_wait_to_complete = 0;
7497 
7498 	/* Copper options */
7499 	if (adapter->hw.phy.media_type == e1000_media_type_copper) {
7500 		adapter->hw.phy.mdix = AUTO_ALL_MODES;
7501 		adapter->hw.phy.disable_polarity_correction = 0;
7502 		adapter->hw.phy.ms_type = e1000_ms_hw_default;
7503 	}
7504 
7505 	if (hw->phy.ops.check_reset_block && hw->phy.ops.check_reset_block(hw))
7506 		dev_info(&pdev->dev,
7507 			 "PHY reset is blocked due to SOL/IDER session.\n");
7508 
7509 	/* Set initial default active device features */
7510 	netdev->features = (NETIF_F_SG |
7511 			    NETIF_F_HW_VLAN_CTAG_RX |
7512 			    NETIF_F_HW_VLAN_CTAG_TX |
7513 			    NETIF_F_TSO |
7514 			    NETIF_F_TSO6 |
7515 			    NETIF_F_RXHASH |
7516 			    NETIF_F_RXCSUM |
7517 			    NETIF_F_HW_CSUM);
7518 
7519 	/* disable TSO for pcie and 10/100 speeds to avoid
7520 	 * some hardware issues and for i219 to fix transfer
7521 	 * speed being capped at 60%
7522 	 */
7523 	if (!(adapter->flags & FLAG_TSO_FORCE)) {
7524 		switch (adapter->link_speed) {
7525 		case SPEED_10:
7526 		case SPEED_100:
7527 			e_info("10/100 speed: disabling TSO\n");
7528 			netdev->features &= ~NETIF_F_TSO;
7529 			netdev->features &= ~NETIF_F_TSO6;
7530 			break;
7531 		case SPEED_1000:
7532 			netdev->features |= NETIF_F_TSO;
7533 			netdev->features |= NETIF_F_TSO6;
7534 			break;
7535 		default:
7536 			/* oops */
7537 			break;
7538 		}
7539 		if (hw->mac.type == e1000_pch_spt) {
7540 			netdev->features &= ~NETIF_F_TSO;
7541 			netdev->features &= ~NETIF_F_TSO6;
7542 		}
7543 	}
7544 
7545 	/* Set user-changeable features (subset of all device features) */
7546 	netdev->hw_features = netdev->features;
7547 	netdev->hw_features |= NETIF_F_RXFCS;
7548 	netdev->priv_flags |= IFF_SUPP_NOFCS;
7549 	netdev->hw_features |= NETIF_F_RXALL;
7550 
7551 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER)
7552 		netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
7553 
7554 	netdev->vlan_features |= (NETIF_F_SG |
7555 				  NETIF_F_TSO |
7556 				  NETIF_F_TSO6 |
7557 				  NETIF_F_HW_CSUM);
7558 
7559 	netdev->priv_flags |= IFF_UNICAST_FLT;
7560 
7561 	netdev->features |= NETIF_F_HIGHDMA;
7562 	netdev->vlan_features |= NETIF_F_HIGHDMA;
7563 
7564 	/* MTU range: 68 - max_hw_frame_size */
7565 	netdev->min_mtu = ETH_MIN_MTU;
7566 	netdev->max_mtu = adapter->max_hw_frame_size -
7567 			  (VLAN_ETH_HLEN + ETH_FCS_LEN);
7568 
7569 	if (e1000e_enable_mng_pass_thru(&adapter->hw))
7570 		adapter->flags |= FLAG_MNG_PT_ENABLED;
7571 
7572 	/* before reading the NVM, reset the controller to
7573 	 * put the device in a known good starting state
7574 	 */
7575 	adapter->hw.mac.ops.reset_hw(&adapter->hw);
7576 
7577 	/* systems with ASPM and others may see the checksum fail on the first
7578 	 * attempt. Let's give it a few tries
7579 	 */
7580 	for (i = 0;; i++) {
7581 		if (e1000_validate_nvm_checksum(&adapter->hw) >= 0)
7582 			break;
7583 		if (i == 2) {
7584 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
7585 			err = -EIO;
7586 			goto err_eeprom;
7587 		}
7588 	}
7589 
7590 	e1000_eeprom_checks(adapter);
7591 
7592 	/* copy the MAC address */
7593 	if (e1000e_read_mac_addr(&adapter->hw))
7594 		dev_err(&pdev->dev,
7595 			"NVM Read Error while reading MAC address\n");
7596 
7597 	eth_hw_addr_set(netdev, adapter->hw.mac.addr);
7598 
7599 	if (!is_valid_ether_addr(netdev->dev_addr)) {
7600 		dev_err(&pdev->dev, "Invalid MAC Address: %pM\n",
7601 			netdev->dev_addr);
7602 		err = -EIO;
7603 		goto err_eeprom;
7604 	}
7605 
7606 	timer_setup(&adapter->watchdog_timer, e1000_watchdog, 0);
7607 	timer_setup(&adapter->phy_info_timer, e1000_update_phy_info, 0);
7608 
7609 	INIT_WORK(&adapter->reset_task, e1000_reset_task);
7610 	INIT_WORK(&adapter->watchdog_task, e1000_watchdog_task);
7611 	INIT_WORK(&adapter->downshift_task, e1000e_downshift_workaround);
7612 	INIT_WORK(&adapter->update_phy_task, e1000e_update_phy_task);
7613 	INIT_WORK(&adapter->print_hang_task, e1000_print_hw_hang);
7614 
7615 	/* Initialize link parameters. User can change them with ethtool */
7616 	adapter->hw.mac.autoneg = 1;
7617 	adapter->fc_autoneg = true;
7618 	adapter->hw.fc.requested_mode = e1000_fc_default;
7619 	adapter->hw.fc.current_mode = e1000_fc_default;
7620 	adapter->hw.phy.autoneg_advertised = 0x2f;
7621 
7622 	/* Initial Wake on LAN setting - If APM wake is enabled in
7623 	 * the EEPROM, enable the ACPI Magic Packet filter
7624 	 */
7625 	if (adapter->flags & FLAG_APME_IN_WUC) {
7626 		/* APME bit in EEPROM is mapped to WUC.APME */
7627 		eeprom_data = er32(WUC);
7628 		eeprom_apme_mask = E1000_WUC_APME;
7629 		if ((hw->mac.type > e1000_ich10lan) &&
7630 		    (eeprom_data & E1000_WUC_PHY_WAKE))
7631 			adapter->flags2 |= FLAG2_HAS_PHY_WAKEUP;
7632 	} else if (adapter->flags & FLAG_APME_IN_CTRL3) {
7633 		if (adapter->flags & FLAG_APME_CHECK_PORT_B &&
7634 		    (adapter->hw.bus.func == 1))
7635 			ret_val = e1000_read_nvm(&adapter->hw,
7636 					      NVM_INIT_CONTROL3_PORT_B,
7637 					      1, &eeprom_data);
7638 		else
7639 			ret_val = e1000_read_nvm(&adapter->hw,
7640 					      NVM_INIT_CONTROL3_PORT_A,
7641 					      1, &eeprom_data);
7642 	}
7643 
7644 	/* fetch WoL from EEPROM */
7645 	if (ret_val)
7646 		e_dbg("NVM read error getting WoL initial values: %d\n", ret_val);
7647 	else if (eeprom_data & eeprom_apme_mask)
7648 		adapter->eeprom_wol |= E1000_WUFC_MAG;
7649 
7650 	/* now that we have the eeprom settings, apply the special cases
7651 	 * where the eeprom may be wrong or the board simply won't support
7652 	 * wake on lan on a particular port
7653 	 */
7654 	if (!(adapter->flags & FLAG_HAS_WOL))
7655 		adapter->eeprom_wol = 0;
7656 
7657 	/* initialize the wol settings based on the eeprom settings */
7658 	adapter->wol = adapter->eeprom_wol;
7659 
7660 	/* make sure adapter isn't asleep if manageability is enabled */
7661 	if (adapter->wol || (adapter->flags & FLAG_MNG_PT_ENABLED) ||
7662 	    (hw->mac.ops.check_mng_mode(hw)))
7663 		device_wakeup_enable(&pdev->dev);
7664 
7665 	/* save off EEPROM version number */
7666 	ret_val = e1000_read_nvm(&adapter->hw, 5, 1, &adapter->eeprom_vers);
7667 
7668 	if (ret_val) {
7669 		e_dbg("NVM read error getting EEPROM version: %d\n", ret_val);
7670 		adapter->eeprom_vers = 0;
7671 	}
7672 
7673 	/* init PTP hardware clock */
7674 	e1000e_ptp_init(adapter);
7675 
7676 	if (hw->mac.type >= e1000_pch_mtp)
7677 		adapter->flags2 |= FLAG2_DISABLE_K1;
7678 
7679 	/* reset the hardware with the new settings */
7680 	e1000e_reset(adapter);
7681 
7682 	/* If the controller has AMT, do not set DRV_LOAD until the interface
7683 	 * is up.  For all other cases, let the f/w know that the h/w is now
7684 	 * under the control of the driver.
7685 	 */
7686 	if (!(adapter->flags & FLAG_HAS_AMT))
7687 		e1000e_get_hw_control(adapter);
7688 
7689 	if (hw->mac.type >= e1000_pch_cnp)
7690 		adapter->flags2 |= FLAG2_ENABLE_S0IX_FLOWS;
7691 
7692 	strscpy(netdev->name, "eth%d", sizeof(netdev->name));
7693 	err = register_netdev(netdev);
7694 	if (err)
7695 		goto err_register;
7696 
7697 	/* carrier off reporting is important to ethtool even BEFORE open */
7698 	netif_carrier_off(netdev);
7699 
7700 	e1000_print_device_info(adapter);
7701 
7702 	dev_pm_set_driver_flags(&pdev->dev, DPM_FLAG_SMART_PREPARE);
7703 
7704 	if (pci_dev_run_wake(pdev))
7705 		pm_runtime_put_noidle(&pdev->dev);
7706 
7707 	return 0;
7708 
7709 err_register:
7710 	if (!(adapter->flags & FLAG_HAS_AMT))
7711 		e1000e_release_hw_control(adapter);
7712 err_eeprom:
7713 	if (hw->phy.ops.check_reset_block && !hw->phy.ops.check_reset_block(hw))
7714 		e1000_phy_hw_reset(&adapter->hw);
7715 err_hw_init:
7716 	kfree(adapter->tx_ring);
7717 	kfree(adapter->rx_ring);
7718 err_sw_init:
7719 	if ((adapter->hw.flash_address) && (hw->mac.type < e1000_pch_spt))
7720 		iounmap(adapter->hw.flash_address);
7721 	e1000e_reset_interrupt_capability(adapter);
7722 err_flashmap:
7723 	iounmap(adapter->hw.hw_addr);
7724 err_ioremap:
7725 	free_netdev(netdev);
7726 err_alloc_etherdev:
7727 	pci_release_mem_regions(pdev);
7728 err_pci_reg:
7729 err_dma:
7730 	pci_disable_device(pdev);
7731 	return err;
7732 }
7733 
7734 /**
7735  * e1000_remove - Device Removal Routine
7736  * @pdev: PCI device information struct
7737  *
7738  * e1000_remove is called by the PCI subsystem to alert the driver
7739  * that it should release a PCI device.  This could be caused by a
7740  * Hot-Plug event, or because the driver is going to be removed from
7741  * memory.
7742  **/
7743 static void e1000_remove(struct pci_dev *pdev)
7744 {
7745 	struct net_device *netdev = pci_get_drvdata(pdev);
7746 	struct e1000_adapter *adapter = netdev_priv(netdev);
7747 
7748 	e1000e_ptp_remove(adapter);
7749 
7750 	/* The timers may be rescheduled, so explicitly disable them
7751 	 * from being rescheduled.
7752 	 */
7753 	set_bit(__E1000_DOWN, &adapter->state);
7754 	timer_delete_sync(&adapter->watchdog_timer);
7755 	timer_delete_sync(&adapter->phy_info_timer);
7756 
7757 	cancel_work_sync(&adapter->reset_task);
7758 	cancel_work_sync(&adapter->watchdog_task);
7759 	cancel_work_sync(&adapter->downshift_task);
7760 	cancel_work_sync(&adapter->update_phy_task);
7761 	cancel_work_sync(&adapter->print_hang_task);
7762 
7763 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
7764 		cancel_work_sync(&adapter->tx_hwtstamp_work);
7765 		if (adapter->tx_hwtstamp_skb) {
7766 			dev_consume_skb_any(adapter->tx_hwtstamp_skb);
7767 			adapter->tx_hwtstamp_skb = NULL;
7768 		}
7769 	}
7770 
7771 	unregister_netdev(netdev);
7772 
7773 	if (pci_dev_run_wake(pdev))
7774 		pm_runtime_get_noresume(&pdev->dev);
7775 
7776 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
7777 	 * would have already happened in close and is redundant.
7778 	 */
7779 	e1000e_release_hw_control(adapter);
7780 
7781 	e1000e_reset_interrupt_capability(adapter);
7782 	kfree(adapter->tx_ring);
7783 	kfree(adapter->rx_ring);
7784 
7785 	iounmap(adapter->hw.hw_addr);
7786 	if ((adapter->hw.flash_address) &&
7787 	    (adapter->hw.mac.type < e1000_pch_spt))
7788 		iounmap(adapter->hw.flash_address);
7789 	pci_release_mem_regions(pdev);
7790 
7791 	free_netdev(netdev);
7792 
7793 	pci_disable_device(pdev);
7794 }
7795 
7796 /* PCI Error Recovery (ERS) */
7797 static const struct pci_error_handlers e1000_err_handler = {
7798 	.error_detected = e1000_io_error_detected,
7799 	.slot_reset = e1000_io_slot_reset,
7800 	.resume = e1000_io_resume,
7801 };
7802 
7803 static const struct pci_device_id e1000_pci_tbl[] = {
7804 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_COPPER), board_82571 },
7805 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_FIBER), board_82571 },
7806 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER), board_82571 },
7807 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER_LP),
7808 	  board_82571 },
7809 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_FIBER), board_82571 },
7810 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES), board_82571 },
7811 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_DUAL), board_82571 },
7812 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_QUAD), board_82571 },
7813 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571PT_QUAD_COPPER), board_82571 },
7814 
7815 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI), board_82572 },
7816 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_COPPER), board_82572 },
7817 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_FIBER), board_82572 },
7818 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_SERDES), board_82572 },
7819 
7820 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E), board_82573 },
7821 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E_IAMT), board_82573 },
7822 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573L), board_82573 },
7823 
7824 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574L), board_82574 },
7825 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574LA), board_82574 },
7826 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82583V), board_82583 },
7827 
7828 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_DPT),
7829 	  board_80003es2lan },
7830 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_SPT),
7831 	  board_80003es2lan },
7832 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_DPT),
7833 	  board_80003es2lan },
7834 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_SPT),
7835 	  board_80003es2lan },
7836 
7837 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE), board_ich8lan },
7838 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_G), board_ich8lan },
7839 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_GT), board_ich8lan },
7840 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_AMT), board_ich8lan },
7841 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_C), board_ich8lan },
7842 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M), board_ich8lan },
7843 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M_AMT), board_ich8lan },
7844 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_82567V_3), board_ich8lan },
7845 
7846 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE), board_ich9lan },
7847 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_G), board_ich9lan },
7848 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_GT), board_ich9lan },
7849 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_AMT), board_ich9lan },
7850 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_C), board_ich9lan },
7851 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_BM), board_ich9lan },
7852 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M), board_ich9lan },
7853 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_AMT), board_ich9lan },
7854 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_V), board_ich9lan },
7855 
7856 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LM), board_ich9lan },
7857 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LF), board_ich9lan },
7858 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_V), board_ich9lan },
7859 
7860 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LM), board_ich10lan },
7861 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LF), board_ich10lan },
7862 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_V), board_ich10lan },
7863 
7864 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LM), board_pchlan },
7865 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LC), board_pchlan },
7866 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DM), board_pchlan },
7867 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DC), board_pchlan },
7868 
7869 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_LM), board_pch2lan },
7870 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_V), board_pch2lan },
7871 
7872 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_LM), board_pch_lpt },
7873 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_V), board_pch_lpt },
7874 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_LM), board_pch_lpt },
7875 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_V), board_pch_lpt },
7876 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM2), board_pch_lpt },
7877 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V2), board_pch_lpt },
7878 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM3), board_pch_lpt },
7879 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V3), board_pch_lpt },
7880 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM), board_pch_spt },
7881 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V), board_pch_spt },
7882 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM2), board_pch_spt },
7883 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V2), board_pch_spt },
7884 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LBG_I219_LM3), board_pch_spt },
7885 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM4), board_pch_spt },
7886 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V4), board_pch_spt },
7887 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM5), board_pch_spt },
7888 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V5), board_pch_spt },
7889 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM6), board_pch_cnp },
7890 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V6), board_pch_cnp },
7891 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM7), board_pch_cnp },
7892 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V7), board_pch_cnp },
7893 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM8), board_pch_cnp },
7894 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V8), board_pch_cnp },
7895 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM9), board_pch_cnp },
7896 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V9), board_pch_cnp },
7897 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM10), board_pch_cnp },
7898 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V10), board_pch_cnp },
7899 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM11), board_pch_cnp },
7900 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V11), board_pch_cnp },
7901 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM12), board_pch_spt },
7902 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V12), board_pch_spt },
7903 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM13), board_pch_tgp },
7904 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V13), board_pch_tgp },
7905 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM14), board_pch_tgp },
7906 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V14), board_pch_tgp },
7907 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM15), board_pch_tgp },
7908 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V15), board_pch_tgp },
7909 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_LM23), board_pch_adp },
7910 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_V23), board_pch_adp },
7911 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM16), board_pch_adp },
7912 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V16), board_pch_adp },
7913 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM17), board_pch_adp },
7914 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V17), board_pch_adp },
7915 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_LM22), board_pch_adp },
7916 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_V22), board_pch_adp },
7917 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM19), board_pch_adp },
7918 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V19), board_pch_adp },
7919 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_LM18), board_pch_mtp },
7920 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_V18), board_pch_mtp },
7921 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_LM20), board_pch_mtp },
7922 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_V20), board_pch_mtp },
7923 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_LM21), board_pch_mtp },
7924 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_V21), board_pch_mtp },
7925 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ARL_I219_LM24), board_pch_mtp },
7926 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ARL_I219_V24), board_pch_mtp },
7927 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_LM25), board_pch_mtp },
7928 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_V25), board_pch_mtp },
7929 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_LM26), board_pch_mtp },
7930 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_V26), board_pch_mtp },
7931 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_LM27), board_pch_mtp },
7932 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_V27), board_pch_mtp },
7933 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_NVL_I219_LM29), board_pch_mtp },
7934 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_NVL_I219_V29), board_pch_mtp },
7935 
7936 	{ 0, 0, 0, 0, 0, 0, 0 }	/* terminate list */
7937 };
7938 MODULE_DEVICE_TABLE(pci, e1000_pci_tbl);
7939 
7940 static const struct dev_pm_ops e1000e_pm_ops = {
7941 	.prepare	= e1000e_pm_prepare,
7942 	.suspend	= e1000e_pm_suspend,
7943 	.resume		= e1000e_pm_resume,
7944 	.freeze		= e1000e_pm_freeze,
7945 	.thaw		= e1000e_pm_thaw,
7946 	.poweroff	= e1000e_pm_suspend,
7947 	.restore	= e1000e_pm_resume,
7948 	RUNTIME_PM_OPS(e1000e_pm_runtime_suspend, e1000e_pm_runtime_resume,
7949 		       e1000e_pm_runtime_idle)
7950 };
7951 
7952 /* PCI Device API Driver */
7953 static struct pci_driver e1000_driver = {
7954 	.name     = e1000e_driver_name,
7955 	.id_table = e1000_pci_tbl,
7956 	.probe    = e1000_probe,
7957 	.remove   = e1000_remove,
7958 	.driver.pm = pm_ptr(&e1000e_pm_ops),
7959 	.shutdown = e1000_shutdown,
7960 	.err_handler = &e1000_err_handler
7961 };
7962 
7963 /**
7964  * e1000_init_module - Driver Registration Routine
7965  *
7966  * e1000_init_module is the first routine called when the driver is
7967  * loaded. All it does is register with the PCI subsystem.
7968  **/
7969 static int __init e1000_init_module(void)
7970 {
7971 	pr_info("Intel(R) PRO/1000 Network Driver\n");
7972 	pr_info("Copyright(c) 1999 - 2015 Intel Corporation.\n");
7973 
7974 	return pci_register_driver(&e1000_driver);
7975 }
7976 module_init(e1000_init_module);
7977 
7978 /**
7979  * e1000_exit_module - Driver Exit Cleanup Routine
7980  *
7981  * e1000_exit_module is called just before the driver is removed
7982  * from memory.
7983  **/
7984 static void __exit e1000_exit_module(void)
7985 {
7986 	pci_unregister_driver(&e1000_driver);
7987 }
7988 module_exit(e1000_exit_module);
7989 
7990 MODULE_DESCRIPTION("Intel(R) PRO/1000 Network Driver");
7991 MODULE_LICENSE("GPL v2");
7992 
7993 /* netdev.c */
7994