xref: /linux/drivers/net/ethernet/intel/e1000e/netdev.c (revision 42874e4eb35bdfc54f8514685e50434098ba4f6c)
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 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1792 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1793 
1794 		/* Do the reset outside of interrupt context */
1795 		schedule_work(&adapter->reset_task);
1796 
1797 		/* return immediately since reset is imminent */
1798 		return IRQ_HANDLED;
1799 	}
1800 
1801 	if (napi_schedule_prep(&adapter->napi)) {
1802 		adapter->total_tx_bytes = 0;
1803 		adapter->total_tx_packets = 0;
1804 		adapter->total_rx_bytes = 0;
1805 		adapter->total_rx_packets = 0;
1806 		__napi_schedule(&adapter->napi);
1807 	}
1808 
1809 	return IRQ_HANDLED;
1810 }
1811 
1812 /**
1813  * e1000_intr - Interrupt Handler
1814  * @irq: interrupt number
1815  * @data: pointer to a network interface device structure
1816  **/
1817 static irqreturn_t e1000_intr(int __always_unused irq, void *data)
1818 {
1819 	struct net_device *netdev = data;
1820 	struct e1000_adapter *adapter = netdev_priv(netdev);
1821 	struct e1000_hw *hw = &adapter->hw;
1822 	u32 rctl, icr = er32(ICR);
1823 
1824 	if (!icr || test_bit(__E1000_DOWN, &adapter->state))
1825 		return IRQ_NONE;	/* Not our interrupt */
1826 
1827 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
1828 	 * not set, then the adapter didn't send an interrupt
1829 	 */
1830 	if (!(icr & E1000_ICR_INT_ASSERTED))
1831 		return IRQ_NONE;
1832 
1833 	/* Interrupt Auto-Mask...upon reading ICR,
1834 	 * interrupts are masked.  No need for the
1835 	 * IMC write
1836 	 */
1837 
1838 	if (icr & E1000_ICR_LSC) {
1839 		hw->mac.get_link_status = true;
1840 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1841 		 * disconnect (LSC) before accessing any PHY registers
1842 		 */
1843 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1844 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1845 			schedule_work(&adapter->downshift_task);
1846 
1847 		/* 80003ES2LAN workaround--
1848 		 * For packet buffer work-around on link down event;
1849 		 * disable receives here in the ISR and
1850 		 * reset adapter in watchdog
1851 		 */
1852 		if (netif_carrier_ok(netdev) &&
1853 		    (adapter->flags & FLAG_RX_NEEDS_RESTART)) {
1854 			/* disable receives */
1855 			rctl = er32(RCTL);
1856 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1857 			adapter->flags |= FLAG_RESTART_NOW;
1858 		}
1859 		/* guard against interrupt when we're going down */
1860 		if (!test_bit(__E1000_DOWN, &adapter->state))
1861 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1862 	}
1863 
1864 	/* Reset on uncorrectable ECC error */
1865 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1866 		u32 pbeccsts = er32(PBECCSTS);
1867 
1868 		adapter->corr_errors +=
1869 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1870 		adapter->uncorr_errors +=
1871 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1872 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1873 
1874 		/* Do the reset outside of interrupt context */
1875 		schedule_work(&adapter->reset_task);
1876 
1877 		/* return immediately since reset is imminent */
1878 		return IRQ_HANDLED;
1879 	}
1880 
1881 	if (napi_schedule_prep(&adapter->napi)) {
1882 		adapter->total_tx_bytes = 0;
1883 		adapter->total_tx_packets = 0;
1884 		adapter->total_rx_bytes = 0;
1885 		adapter->total_rx_packets = 0;
1886 		__napi_schedule(&adapter->napi);
1887 	}
1888 
1889 	return IRQ_HANDLED;
1890 }
1891 
1892 static irqreturn_t e1000_msix_other(int __always_unused irq, void *data)
1893 {
1894 	struct net_device *netdev = data;
1895 	struct e1000_adapter *adapter = netdev_priv(netdev);
1896 	struct e1000_hw *hw = &adapter->hw;
1897 	u32 icr = er32(ICR);
1898 
1899 	if (icr & adapter->eiac_mask)
1900 		ew32(ICS, (icr & adapter->eiac_mask));
1901 
1902 	if (icr & E1000_ICR_LSC) {
1903 		hw->mac.get_link_status = true;
1904 		/* guard against interrupt when we're going down */
1905 		if (!test_bit(__E1000_DOWN, &adapter->state))
1906 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1907 	}
1908 
1909 	if (!test_bit(__E1000_DOWN, &adapter->state))
1910 		ew32(IMS, E1000_IMS_OTHER | IMS_OTHER_MASK);
1911 
1912 	return IRQ_HANDLED;
1913 }
1914 
1915 static irqreturn_t e1000_intr_msix_tx(int __always_unused irq, void *data)
1916 {
1917 	struct net_device *netdev = data;
1918 	struct e1000_adapter *adapter = netdev_priv(netdev);
1919 	struct e1000_hw *hw = &adapter->hw;
1920 	struct e1000_ring *tx_ring = adapter->tx_ring;
1921 
1922 	adapter->total_tx_bytes = 0;
1923 	adapter->total_tx_packets = 0;
1924 
1925 	if (!e1000_clean_tx_irq(tx_ring))
1926 		/* Ring was not completely cleaned, so fire another interrupt */
1927 		ew32(ICS, tx_ring->ims_val);
1928 
1929 	if (!test_bit(__E1000_DOWN, &adapter->state))
1930 		ew32(IMS, adapter->tx_ring->ims_val);
1931 
1932 	return IRQ_HANDLED;
1933 }
1934 
1935 static irqreturn_t e1000_intr_msix_rx(int __always_unused irq, void *data)
1936 {
1937 	struct net_device *netdev = data;
1938 	struct e1000_adapter *adapter = netdev_priv(netdev);
1939 	struct e1000_ring *rx_ring = adapter->rx_ring;
1940 
1941 	/* Write the ITR value calculated at the end of the
1942 	 * previous interrupt.
1943 	 */
1944 	if (rx_ring->set_itr) {
1945 		u32 itr = rx_ring->itr_val ?
1946 			  1000000000 / (rx_ring->itr_val * 256) : 0;
1947 
1948 		writel(itr, rx_ring->itr_register);
1949 		rx_ring->set_itr = 0;
1950 	}
1951 
1952 	if (napi_schedule_prep(&adapter->napi)) {
1953 		adapter->total_rx_bytes = 0;
1954 		adapter->total_rx_packets = 0;
1955 		__napi_schedule(&adapter->napi);
1956 	}
1957 	return IRQ_HANDLED;
1958 }
1959 
1960 /**
1961  * e1000_configure_msix - Configure MSI-X hardware
1962  * @adapter: board private structure
1963  *
1964  * e1000_configure_msix sets up the hardware to properly
1965  * generate MSI-X interrupts.
1966  **/
1967 static void e1000_configure_msix(struct e1000_adapter *adapter)
1968 {
1969 	struct e1000_hw *hw = &adapter->hw;
1970 	struct e1000_ring *rx_ring = adapter->rx_ring;
1971 	struct e1000_ring *tx_ring = adapter->tx_ring;
1972 	int vector = 0;
1973 	u32 ctrl_ext, ivar = 0;
1974 
1975 	adapter->eiac_mask = 0;
1976 
1977 	/* Workaround issue with spurious interrupts on 82574 in MSI-X mode */
1978 	if (hw->mac.type == e1000_82574) {
1979 		u32 rfctl = er32(RFCTL);
1980 
1981 		rfctl |= E1000_RFCTL_ACK_DIS;
1982 		ew32(RFCTL, rfctl);
1983 	}
1984 
1985 	/* Configure Rx vector */
1986 	rx_ring->ims_val = E1000_IMS_RXQ0;
1987 	adapter->eiac_mask |= rx_ring->ims_val;
1988 	if (rx_ring->itr_val)
1989 		writel(1000000000 / (rx_ring->itr_val * 256),
1990 		       rx_ring->itr_register);
1991 	else
1992 		writel(1, rx_ring->itr_register);
1993 	ivar = E1000_IVAR_INT_ALLOC_VALID | vector;
1994 
1995 	/* Configure Tx vector */
1996 	tx_ring->ims_val = E1000_IMS_TXQ0;
1997 	vector++;
1998 	if (tx_ring->itr_val)
1999 		writel(1000000000 / (tx_ring->itr_val * 256),
2000 		       tx_ring->itr_register);
2001 	else
2002 		writel(1, tx_ring->itr_register);
2003 	adapter->eiac_mask |= tx_ring->ims_val;
2004 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 8);
2005 
2006 	/* set vector for Other Causes, e.g. link changes */
2007 	vector++;
2008 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 16);
2009 	if (rx_ring->itr_val)
2010 		writel(1000000000 / (rx_ring->itr_val * 256),
2011 		       hw->hw_addr + E1000_EITR_82574(vector));
2012 	else
2013 		writel(1, hw->hw_addr + E1000_EITR_82574(vector));
2014 
2015 	/* Cause Tx interrupts on every write back */
2016 	ivar |= BIT(31);
2017 
2018 	ew32(IVAR, ivar);
2019 
2020 	/* enable MSI-X PBA support */
2021 	ctrl_ext = er32(CTRL_EXT) & ~E1000_CTRL_EXT_IAME;
2022 	ctrl_ext |= E1000_CTRL_EXT_PBA_CLR | E1000_CTRL_EXT_EIAME;
2023 	ew32(CTRL_EXT, ctrl_ext);
2024 	e1e_flush();
2025 }
2026 
2027 void e1000e_reset_interrupt_capability(struct e1000_adapter *adapter)
2028 {
2029 	if (adapter->msix_entries) {
2030 		pci_disable_msix(adapter->pdev);
2031 		kfree(adapter->msix_entries);
2032 		adapter->msix_entries = NULL;
2033 	} else if (adapter->flags & FLAG_MSI_ENABLED) {
2034 		pci_disable_msi(adapter->pdev);
2035 		adapter->flags &= ~FLAG_MSI_ENABLED;
2036 	}
2037 }
2038 
2039 /**
2040  * e1000e_set_interrupt_capability - set MSI or MSI-X if supported
2041  * @adapter: board private structure
2042  *
2043  * Attempt to configure interrupts using the best available
2044  * capabilities of the hardware and kernel.
2045  **/
2046 void e1000e_set_interrupt_capability(struct e1000_adapter *adapter)
2047 {
2048 	int err;
2049 	int i;
2050 
2051 	switch (adapter->int_mode) {
2052 	case E1000E_INT_MODE_MSIX:
2053 		if (adapter->flags & FLAG_HAS_MSIX) {
2054 			adapter->num_vectors = 3; /* RxQ0, TxQ0 and other */
2055 			adapter->msix_entries = kcalloc(adapter->num_vectors,
2056 							sizeof(struct
2057 							       msix_entry),
2058 							GFP_KERNEL);
2059 			if (adapter->msix_entries) {
2060 				struct e1000_adapter *a = adapter;
2061 
2062 				for (i = 0; i < adapter->num_vectors; i++)
2063 					adapter->msix_entries[i].entry = i;
2064 
2065 				err = pci_enable_msix_range(a->pdev,
2066 							    a->msix_entries,
2067 							    a->num_vectors,
2068 							    a->num_vectors);
2069 				if (err > 0)
2070 					return;
2071 			}
2072 			/* MSI-X failed, so fall through and try MSI */
2073 			e_err("Failed to initialize MSI-X interrupts.  Falling back to MSI interrupts.\n");
2074 			e1000e_reset_interrupt_capability(adapter);
2075 		}
2076 		adapter->int_mode = E1000E_INT_MODE_MSI;
2077 		fallthrough;
2078 	case E1000E_INT_MODE_MSI:
2079 		if (!pci_enable_msi(adapter->pdev)) {
2080 			adapter->flags |= FLAG_MSI_ENABLED;
2081 		} else {
2082 			adapter->int_mode = E1000E_INT_MODE_LEGACY;
2083 			e_err("Failed to initialize MSI interrupts.  Falling back to legacy interrupts.\n");
2084 		}
2085 		fallthrough;
2086 	case E1000E_INT_MODE_LEGACY:
2087 		/* Don't do anything; this is the system default */
2088 		break;
2089 	}
2090 
2091 	/* store the number of vectors being used */
2092 	adapter->num_vectors = 1;
2093 }
2094 
2095 /**
2096  * e1000_request_msix - Initialize MSI-X interrupts
2097  * @adapter: board private structure
2098  *
2099  * e1000_request_msix allocates MSI-X vectors and requests interrupts from the
2100  * kernel.
2101  **/
2102 static int e1000_request_msix(struct e1000_adapter *adapter)
2103 {
2104 	struct net_device *netdev = adapter->netdev;
2105 	int err = 0, vector = 0;
2106 
2107 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2108 		snprintf(adapter->rx_ring->name,
2109 			 sizeof(adapter->rx_ring->name) - 1,
2110 			 "%.14s-rx-0", netdev->name);
2111 	else
2112 		memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
2113 	err = request_irq(adapter->msix_entries[vector].vector,
2114 			  e1000_intr_msix_rx, 0, adapter->rx_ring->name,
2115 			  netdev);
2116 	if (err)
2117 		return err;
2118 	adapter->rx_ring->itr_register = adapter->hw.hw_addr +
2119 	    E1000_EITR_82574(vector);
2120 	adapter->rx_ring->itr_val = adapter->itr;
2121 	vector++;
2122 
2123 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2124 		snprintf(adapter->tx_ring->name,
2125 			 sizeof(adapter->tx_ring->name) - 1,
2126 			 "%.14s-tx-0", netdev->name);
2127 	else
2128 		memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
2129 	err = request_irq(adapter->msix_entries[vector].vector,
2130 			  e1000_intr_msix_tx, 0, adapter->tx_ring->name,
2131 			  netdev);
2132 	if (err)
2133 		return err;
2134 	adapter->tx_ring->itr_register = adapter->hw.hw_addr +
2135 	    E1000_EITR_82574(vector);
2136 	adapter->tx_ring->itr_val = adapter->itr;
2137 	vector++;
2138 
2139 	err = request_irq(adapter->msix_entries[vector].vector,
2140 			  e1000_msix_other, 0, netdev->name, netdev);
2141 	if (err)
2142 		return err;
2143 
2144 	e1000_configure_msix(adapter);
2145 
2146 	return 0;
2147 }
2148 
2149 /**
2150  * e1000_request_irq - initialize interrupts
2151  * @adapter: board private structure
2152  *
2153  * Attempts to configure interrupts using the best available
2154  * capabilities of the hardware and kernel.
2155  **/
2156 static int e1000_request_irq(struct e1000_adapter *adapter)
2157 {
2158 	struct net_device *netdev = adapter->netdev;
2159 	int err;
2160 
2161 	if (adapter->msix_entries) {
2162 		err = e1000_request_msix(adapter);
2163 		if (!err)
2164 			return err;
2165 		/* fall back to MSI */
2166 		e1000e_reset_interrupt_capability(adapter);
2167 		adapter->int_mode = E1000E_INT_MODE_MSI;
2168 		e1000e_set_interrupt_capability(adapter);
2169 	}
2170 	if (adapter->flags & FLAG_MSI_ENABLED) {
2171 		err = request_irq(adapter->pdev->irq, e1000_intr_msi, 0,
2172 				  netdev->name, netdev);
2173 		if (!err)
2174 			return err;
2175 
2176 		/* fall back to legacy interrupt */
2177 		e1000e_reset_interrupt_capability(adapter);
2178 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
2179 	}
2180 
2181 	err = request_irq(adapter->pdev->irq, e1000_intr, IRQF_SHARED,
2182 			  netdev->name, netdev);
2183 	if (err)
2184 		e_err("Unable to allocate interrupt, Error: %d\n", err);
2185 
2186 	return err;
2187 }
2188 
2189 static void e1000_free_irq(struct e1000_adapter *adapter)
2190 {
2191 	struct net_device *netdev = adapter->netdev;
2192 
2193 	if (adapter->msix_entries) {
2194 		int vector = 0;
2195 
2196 		free_irq(adapter->msix_entries[vector].vector, netdev);
2197 		vector++;
2198 
2199 		free_irq(adapter->msix_entries[vector].vector, netdev);
2200 		vector++;
2201 
2202 		/* Other Causes interrupt vector */
2203 		free_irq(adapter->msix_entries[vector].vector, netdev);
2204 		return;
2205 	}
2206 
2207 	free_irq(adapter->pdev->irq, netdev);
2208 }
2209 
2210 /**
2211  * e1000_irq_disable - Mask off interrupt generation on the NIC
2212  * @adapter: board private structure
2213  **/
2214 static void e1000_irq_disable(struct e1000_adapter *adapter)
2215 {
2216 	struct e1000_hw *hw = &adapter->hw;
2217 
2218 	ew32(IMC, ~0);
2219 	if (adapter->msix_entries)
2220 		ew32(EIAC_82574, 0);
2221 	e1e_flush();
2222 
2223 	if (adapter->msix_entries) {
2224 		int i;
2225 
2226 		for (i = 0; i < adapter->num_vectors; i++)
2227 			synchronize_irq(adapter->msix_entries[i].vector);
2228 	} else {
2229 		synchronize_irq(adapter->pdev->irq);
2230 	}
2231 }
2232 
2233 /**
2234  * e1000_irq_enable - Enable default interrupt generation settings
2235  * @adapter: board private structure
2236  **/
2237 static void e1000_irq_enable(struct e1000_adapter *adapter)
2238 {
2239 	struct e1000_hw *hw = &adapter->hw;
2240 
2241 	if (adapter->msix_entries) {
2242 		ew32(EIAC_82574, adapter->eiac_mask & E1000_EIAC_MASK_82574);
2243 		ew32(IMS, adapter->eiac_mask | E1000_IMS_OTHER |
2244 		     IMS_OTHER_MASK);
2245 	} else if (hw->mac.type >= e1000_pch_lpt) {
2246 		ew32(IMS, IMS_ENABLE_MASK | E1000_IMS_ECCER);
2247 	} else {
2248 		ew32(IMS, IMS_ENABLE_MASK);
2249 	}
2250 	e1e_flush();
2251 }
2252 
2253 /**
2254  * e1000e_get_hw_control - get control of the h/w from f/w
2255  * @adapter: address of board private structure
2256  *
2257  * e1000e_get_hw_control sets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2258  * For ASF and Pass Through versions of f/w this means that
2259  * the driver is loaded. For AMT version (only with 82573)
2260  * of the f/w this means that the network i/f is open.
2261  **/
2262 void e1000e_get_hw_control(struct e1000_adapter *adapter)
2263 {
2264 	struct e1000_hw *hw = &adapter->hw;
2265 	u32 ctrl_ext;
2266 	u32 swsm;
2267 
2268 	/* Let firmware know the driver has taken over */
2269 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2270 		swsm = er32(SWSM);
2271 		ew32(SWSM, swsm | E1000_SWSM_DRV_LOAD);
2272 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2273 		ctrl_ext = er32(CTRL_EXT);
2274 		ew32(CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2275 	}
2276 }
2277 
2278 /**
2279  * e1000e_release_hw_control - release control of the h/w to f/w
2280  * @adapter: address of board private structure
2281  *
2282  * e1000e_release_hw_control resets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2283  * For ASF and Pass Through versions of f/w this means that the
2284  * driver is no longer loaded. For AMT version (only with 82573) i
2285  * of the f/w this means that the network i/f is closed.
2286  *
2287  **/
2288 void e1000e_release_hw_control(struct e1000_adapter *adapter)
2289 {
2290 	struct e1000_hw *hw = &adapter->hw;
2291 	u32 ctrl_ext;
2292 	u32 swsm;
2293 
2294 	/* Let firmware taken over control of h/w */
2295 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2296 		swsm = er32(SWSM);
2297 		ew32(SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
2298 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2299 		ctrl_ext = er32(CTRL_EXT);
2300 		ew32(CTRL_EXT, ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2301 	}
2302 }
2303 
2304 /**
2305  * e1000_alloc_ring_dma - allocate memory for a ring structure
2306  * @adapter: board private structure
2307  * @ring: ring struct for which to allocate dma
2308  **/
2309 static int e1000_alloc_ring_dma(struct e1000_adapter *adapter,
2310 				struct e1000_ring *ring)
2311 {
2312 	struct pci_dev *pdev = adapter->pdev;
2313 
2314 	ring->desc = dma_alloc_coherent(&pdev->dev, ring->size, &ring->dma,
2315 					GFP_KERNEL);
2316 	if (!ring->desc)
2317 		return -ENOMEM;
2318 
2319 	return 0;
2320 }
2321 
2322 /**
2323  * e1000e_setup_tx_resources - allocate Tx resources (Descriptors)
2324  * @tx_ring: Tx descriptor ring
2325  *
2326  * Return 0 on success, negative on failure
2327  **/
2328 int e1000e_setup_tx_resources(struct e1000_ring *tx_ring)
2329 {
2330 	struct e1000_adapter *adapter = tx_ring->adapter;
2331 	int err = -ENOMEM, size;
2332 
2333 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2334 	tx_ring->buffer_info = vzalloc(size);
2335 	if (!tx_ring->buffer_info)
2336 		goto err;
2337 
2338 	/* round up to nearest 4K */
2339 	tx_ring->size = tx_ring->count * sizeof(struct e1000_tx_desc);
2340 	tx_ring->size = ALIGN(tx_ring->size, 4096);
2341 
2342 	err = e1000_alloc_ring_dma(adapter, tx_ring);
2343 	if (err)
2344 		goto err;
2345 
2346 	tx_ring->next_to_use = 0;
2347 	tx_ring->next_to_clean = 0;
2348 
2349 	return 0;
2350 err:
2351 	vfree(tx_ring->buffer_info);
2352 	e_err("Unable to allocate memory for the transmit descriptor ring\n");
2353 	return err;
2354 }
2355 
2356 /**
2357  * e1000e_setup_rx_resources - allocate Rx resources (Descriptors)
2358  * @rx_ring: Rx descriptor ring
2359  *
2360  * Returns 0 on success, negative on failure
2361  **/
2362 int e1000e_setup_rx_resources(struct e1000_ring *rx_ring)
2363 {
2364 	struct e1000_adapter *adapter = rx_ring->adapter;
2365 	struct e1000_buffer *buffer_info;
2366 	int i, size, desc_len, err = -ENOMEM;
2367 
2368 	size = sizeof(struct e1000_buffer) * rx_ring->count;
2369 	rx_ring->buffer_info = vzalloc(size);
2370 	if (!rx_ring->buffer_info)
2371 		goto err;
2372 
2373 	for (i = 0; i < rx_ring->count; i++) {
2374 		buffer_info = &rx_ring->buffer_info[i];
2375 		buffer_info->ps_pages = kcalloc(PS_PAGE_BUFFERS,
2376 						sizeof(struct e1000_ps_page),
2377 						GFP_KERNEL);
2378 		if (!buffer_info->ps_pages)
2379 			goto err_pages;
2380 	}
2381 
2382 	desc_len = sizeof(union e1000_rx_desc_packet_split);
2383 
2384 	/* Round up to nearest 4K */
2385 	rx_ring->size = rx_ring->count * desc_len;
2386 	rx_ring->size = ALIGN(rx_ring->size, 4096);
2387 
2388 	err = e1000_alloc_ring_dma(adapter, rx_ring);
2389 	if (err)
2390 		goto err_pages;
2391 
2392 	rx_ring->next_to_clean = 0;
2393 	rx_ring->next_to_use = 0;
2394 	rx_ring->rx_skb_top = NULL;
2395 
2396 	return 0;
2397 
2398 err_pages:
2399 	for (i = 0; i < rx_ring->count; i++) {
2400 		buffer_info = &rx_ring->buffer_info[i];
2401 		kfree(buffer_info->ps_pages);
2402 	}
2403 err:
2404 	vfree(rx_ring->buffer_info);
2405 	e_err("Unable to allocate memory for the receive descriptor ring\n");
2406 	return err;
2407 }
2408 
2409 /**
2410  * e1000_clean_tx_ring - Free Tx Buffers
2411  * @tx_ring: Tx descriptor ring
2412  **/
2413 static void e1000_clean_tx_ring(struct e1000_ring *tx_ring)
2414 {
2415 	struct e1000_adapter *adapter = tx_ring->adapter;
2416 	struct e1000_buffer *buffer_info;
2417 	unsigned long size;
2418 	unsigned int i;
2419 
2420 	for (i = 0; i < tx_ring->count; i++) {
2421 		buffer_info = &tx_ring->buffer_info[i];
2422 		e1000_put_txbuf(tx_ring, buffer_info, false);
2423 	}
2424 
2425 	netdev_reset_queue(adapter->netdev);
2426 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2427 	memset(tx_ring->buffer_info, 0, size);
2428 
2429 	memset(tx_ring->desc, 0, tx_ring->size);
2430 
2431 	tx_ring->next_to_use = 0;
2432 	tx_ring->next_to_clean = 0;
2433 }
2434 
2435 /**
2436  * e1000e_free_tx_resources - Free Tx Resources per Queue
2437  * @tx_ring: Tx descriptor ring
2438  *
2439  * Free all transmit software resources
2440  **/
2441 void e1000e_free_tx_resources(struct e1000_ring *tx_ring)
2442 {
2443 	struct e1000_adapter *adapter = tx_ring->adapter;
2444 	struct pci_dev *pdev = adapter->pdev;
2445 
2446 	e1000_clean_tx_ring(tx_ring);
2447 
2448 	vfree(tx_ring->buffer_info);
2449 	tx_ring->buffer_info = NULL;
2450 
2451 	dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
2452 			  tx_ring->dma);
2453 	tx_ring->desc = NULL;
2454 }
2455 
2456 /**
2457  * e1000e_free_rx_resources - Free Rx Resources
2458  * @rx_ring: Rx descriptor ring
2459  *
2460  * Free all receive software resources
2461  **/
2462 void e1000e_free_rx_resources(struct e1000_ring *rx_ring)
2463 {
2464 	struct e1000_adapter *adapter = rx_ring->adapter;
2465 	struct pci_dev *pdev = adapter->pdev;
2466 	int i;
2467 
2468 	e1000_clean_rx_ring(rx_ring);
2469 
2470 	for (i = 0; i < rx_ring->count; i++)
2471 		kfree(rx_ring->buffer_info[i].ps_pages);
2472 
2473 	vfree(rx_ring->buffer_info);
2474 	rx_ring->buffer_info = NULL;
2475 
2476 	dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
2477 			  rx_ring->dma);
2478 	rx_ring->desc = NULL;
2479 }
2480 
2481 /**
2482  * e1000_update_itr - update the dynamic ITR value based on statistics
2483  * @itr_setting: current adapter->itr
2484  * @packets: the number of packets during this measurement interval
2485  * @bytes: the number of bytes during this measurement interval
2486  *
2487  *      Stores a new ITR value based on packets and byte
2488  *      counts during the last interrupt.  The advantage of per interrupt
2489  *      computation is faster updates and more accurate ITR for the current
2490  *      traffic pattern.  Constants in this function were computed
2491  *      based on theoretical maximum wire speed and thresholds were set based
2492  *      on testing data as well as attempting to minimize response time
2493  *      while increasing bulk throughput.  This functionality is controlled
2494  *      by the InterruptThrottleRate module parameter.
2495  **/
2496 static unsigned int e1000_update_itr(u16 itr_setting, int packets, int bytes)
2497 {
2498 	unsigned int retval = itr_setting;
2499 
2500 	if (packets == 0)
2501 		return itr_setting;
2502 
2503 	switch (itr_setting) {
2504 	case lowest_latency:
2505 		/* handle TSO and jumbo frames */
2506 		if (bytes / packets > 8000)
2507 			retval = bulk_latency;
2508 		else if ((packets < 5) && (bytes > 512))
2509 			retval = low_latency;
2510 		break;
2511 	case low_latency:	/* 50 usec aka 20000 ints/s */
2512 		if (bytes > 10000) {
2513 			/* this if handles the TSO accounting */
2514 			if (bytes / packets > 8000)
2515 				retval = bulk_latency;
2516 			else if ((packets < 10) || ((bytes / packets) > 1200))
2517 				retval = bulk_latency;
2518 			else if ((packets > 35))
2519 				retval = lowest_latency;
2520 		} else if (bytes / packets > 2000) {
2521 			retval = bulk_latency;
2522 		} else if (packets <= 2 && bytes < 512) {
2523 			retval = lowest_latency;
2524 		}
2525 		break;
2526 	case bulk_latency:	/* 250 usec aka 4000 ints/s */
2527 		if (bytes > 25000) {
2528 			if (packets > 35)
2529 				retval = low_latency;
2530 		} else if (bytes < 6000) {
2531 			retval = low_latency;
2532 		}
2533 		break;
2534 	}
2535 
2536 	return retval;
2537 }
2538 
2539 static void e1000_set_itr(struct e1000_adapter *adapter)
2540 {
2541 	u16 current_itr;
2542 	u32 new_itr = adapter->itr;
2543 
2544 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
2545 	if (adapter->link_speed != SPEED_1000) {
2546 		new_itr = 4000;
2547 		goto set_itr_now;
2548 	}
2549 
2550 	if (adapter->flags2 & FLAG2_DISABLE_AIM) {
2551 		new_itr = 0;
2552 		goto set_itr_now;
2553 	}
2554 
2555 	adapter->tx_itr = e1000_update_itr(adapter->tx_itr,
2556 					   adapter->total_tx_packets,
2557 					   adapter->total_tx_bytes);
2558 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2559 	if (adapter->itr_setting == 3 && adapter->tx_itr == lowest_latency)
2560 		adapter->tx_itr = low_latency;
2561 
2562 	adapter->rx_itr = e1000_update_itr(adapter->rx_itr,
2563 					   adapter->total_rx_packets,
2564 					   adapter->total_rx_bytes);
2565 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2566 	if (adapter->itr_setting == 3 && adapter->rx_itr == lowest_latency)
2567 		adapter->rx_itr = low_latency;
2568 
2569 	current_itr = max(adapter->rx_itr, adapter->tx_itr);
2570 
2571 	/* counts and packets in update_itr are dependent on these numbers */
2572 	switch (current_itr) {
2573 	case lowest_latency:
2574 		new_itr = 70000;
2575 		break;
2576 	case low_latency:
2577 		new_itr = 20000;	/* aka hwitr = ~200 */
2578 		break;
2579 	case bulk_latency:
2580 		new_itr = 4000;
2581 		break;
2582 	default:
2583 		break;
2584 	}
2585 
2586 set_itr_now:
2587 	if (new_itr != adapter->itr) {
2588 		/* this attempts to bias the interrupt rate towards Bulk
2589 		 * by adding intermediate steps when interrupt rate is
2590 		 * increasing
2591 		 */
2592 		new_itr = new_itr > adapter->itr ?
2593 		    min(adapter->itr + (new_itr >> 2), new_itr) : new_itr;
2594 		adapter->itr = new_itr;
2595 		adapter->rx_ring->itr_val = new_itr;
2596 		if (adapter->msix_entries)
2597 			adapter->rx_ring->set_itr = 1;
2598 		else
2599 			e1000e_write_itr(adapter, new_itr);
2600 	}
2601 }
2602 
2603 /**
2604  * e1000e_write_itr - write the ITR value to the appropriate registers
2605  * @adapter: address of board private structure
2606  * @itr: new ITR value to program
2607  *
2608  * e1000e_write_itr determines if the adapter is in MSI-X mode
2609  * and, if so, writes the EITR registers with the ITR value.
2610  * Otherwise, it writes the ITR value into the ITR register.
2611  **/
2612 void e1000e_write_itr(struct e1000_adapter *adapter, u32 itr)
2613 {
2614 	struct e1000_hw *hw = &adapter->hw;
2615 	u32 new_itr = itr ? 1000000000 / (itr * 256) : 0;
2616 
2617 	if (adapter->msix_entries) {
2618 		int vector;
2619 
2620 		for (vector = 0; vector < adapter->num_vectors; vector++)
2621 			writel(new_itr, hw->hw_addr + E1000_EITR_82574(vector));
2622 	} else {
2623 		ew32(ITR, new_itr);
2624 	}
2625 }
2626 
2627 /**
2628  * e1000_alloc_queues - Allocate memory for all rings
2629  * @adapter: board private structure to initialize
2630  **/
2631 static int e1000_alloc_queues(struct e1000_adapter *adapter)
2632 {
2633 	int size = sizeof(struct e1000_ring);
2634 
2635 	adapter->tx_ring = kzalloc(size, GFP_KERNEL);
2636 	if (!adapter->tx_ring)
2637 		goto err;
2638 	adapter->tx_ring->count = adapter->tx_ring_count;
2639 	adapter->tx_ring->adapter = adapter;
2640 
2641 	adapter->rx_ring = kzalloc(size, GFP_KERNEL);
2642 	if (!adapter->rx_ring)
2643 		goto err;
2644 	adapter->rx_ring->count = adapter->rx_ring_count;
2645 	adapter->rx_ring->adapter = adapter;
2646 
2647 	return 0;
2648 err:
2649 	e_err("Unable to allocate memory for queues\n");
2650 	kfree(adapter->rx_ring);
2651 	kfree(adapter->tx_ring);
2652 	return -ENOMEM;
2653 }
2654 
2655 /**
2656  * e1000e_poll - NAPI Rx polling callback
2657  * @napi: struct associated with this polling callback
2658  * @budget: number of packets driver is allowed to process this poll
2659  **/
2660 static int e1000e_poll(struct napi_struct *napi, int budget)
2661 {
2662 	struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter,
2663 						     napi);
2664 	struct e1000_hw *hw = &adapter->hw;
2665 	struct net_device *poll_dev = adapter->netdev;
2666 	int tx_cleaned = 1, work_done = 0;
2667 
2668 	adapter = netdev_priv(poll_dev);
2669 
2670 	if (!adapter->msix_entries ||
2671 	    (adapter->rx_ring->ims_val & adapter->tx_ring->ims_val))
2672 		tx_cleaned = e1000_clean_tx_irq(adapter->tx_ring);
2673 
2674 	adapter->clean_rx(adapter->rx_ring, &work_done, budget);
2675 
2676 	if (!tx_cleaned || work_done == budget)
2677 		return budget;
2678 
2679 	/* Exit the polling mode, but don't re-enable interrupts if stack might
2680 	 * poll us due to busy-polling
2681 	 */
2682 	if (likely(napi_complete_done(napi, work_done))) {
2683 		if (adapter->itr_setting & 3)
2684 			e1000_set_itr(adapter);
2685 		if (!test_bit(__E1000_DOWN, &adapter->state)) {
2686 			if (adapter->msix_entries)
2687 				ew32(IMS, adapter->rx_ring->ims_val);
2688 			else
2689 				e1000_irq_enable(adapter);
2690 		}
2691 	}
2692 
2693 	return work_done;
2694 }
2695 
2696 static int e1000_vlan_rx_add_vid(struct net_device *netdev,
2697 				 __always_unused __be16 proto, u16 vid)
2698 {
2699 	struct e1000_adapter *adapter = netdev_priv(netdev);
2700 	struct e1000_hw *hw = &adapter->hw;
2701 	u32 vfta, index;
2702 
2703 	/* don't update vlan cookie if already programmed */
2704 	if ((adapter->hw.mng_cookie.status &
2705 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2706 	    (vid == adapter->mng_vlan_id))
2707 		return 0;
2708 
2709 	/* add VID to filter table */
2710 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2711 		index = (vid >> 5) & 0x7F;
2712 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2713 		vfta |= BIT((vid & 0x1F));
2714 		hw->mac.ops.write_vfta(hw, index, vfta);
2715 	}
2716 
2717 	set_bit(vid, adapter->active_vlans);
2718 
2719 	return 0;
2720 }
2721 
2722 static int e1000_vlan_rx_kill_vid(struct net_device *netdev,
2723 				  __always_unused __be16 proto, u16 vid)
2724 {
2725 	struct e1000_adapter *adapter = netdev_priv(netdev);
2726 	struct e1000_hw *hw = &adapter->hw;
2727 	u32 vfta, index;
2728 
2729 	if ((adapter->hw.mng_cookie.status &
2730 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2731 	    (vid == adapter->mng_vlan_id)) {
2732 		/* release control to f/w */
2733 		e1000e_release_hw_control(adapter);
2734 		return 0;
2735 	}
2736 
2737 	/* remove VID from filter table */
2738 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2739 		index = (vid >> 5) & 0x7F;
2740 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2741 		vfta &= ~BIT((vid & 0x1F));
2742 		hw->mac.ops.write_vfta(hw, index, vfta);
2743 	}
2744 
2745 	clear_bit(vid, adapter->active_vlans);
2746 
2747 	return 0;
2748 }
2749 
2750 /**
2751  * e1000e_vlan_filter_disable - helper to disable hw VLAN filtering
2752  * @adapter: board private structure to initialize
2753  **/
2754 static void e1000e_vlan_filter_disable(struct e1000_adapter *adapter)
2755 {
2756 	struct net_device *netdev = adapter->netdev;
2757 	struct e1000_hw *hw = &adapter->hw;
2758 	u32 rctl;
2759 
2760 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2761 		/* disable VLAN receive filtering */
2762 		rctl = er32(RCTL);
2763 		rctl &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
2764 		ew32(RCTL, rctl);
2765 
2766 		if (adapter->mng_vlan_id != (u16)E1000_MNG_VLAN_NONE) {
2767 			e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
2768 					       adapter->mng_vlan_id);
2769 			adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
2770 		}
2771 	}
2772 }
2773 
2774 /**
2775  * e1000e_vlan_filter_enable - helper to enable HW VLAN filtering
2776  * @adapter: board private structure to initialize
2777  **/
2778 static void e1000e_vlan_filter_enable(struct e1000_adapter *adapter)
2779 {
2780 	struct e1000_hw *hw = &adapter->hw;
2781 	u32 rctl;
2782 
2783 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2784 		/* enable VLAN receive filtering */
2785 		rctl = er32(RCTL);
2786 		rctl |= E1000_RCTL_VFE;
2787 		rctl &= ~E1000_RCTL_CFIEN;
2788 		ew32(RCTL, rctl);
2789 	}
2790 }
2791 
2792 /**
2793  * e1000e_vlan_strip_disable - helper to disable HW VLAN stripping
2794  * @adapter: board private structure to initialize
2795  **/
2796 static void e1000e_vlan_strip_disable(struct e1000_adapter *adapter)
2797 {
2798 	struct e1000_hw *hw = &adapter->hw;
2799 	u32 ctrl;
2800 
2801 	/* disable VLAN tag insert/strip */
2802 	ctrl = er32(CTRL);
2803 	ctrl &= ~E1000_CTRL_VME;
2804 	ew32(CTRL, ctrl);
2805 }
2806 
2807 /**
2808  * e1000e_vlan_strip_enable - helper to enable HW VLAN stripping
2809  * @adapter: board private structure to initialize
2810  **/
2811 static void e1000e_vlan_strip_enable(struct e1000_adapter *adapter)
2812 {
2813 	struct e1000_hw *hw = &adapter->hw;
2814 	u32 ctrl;
2815 
2816 	/* enable VLAN tag insert/strip */
2817 	ctrl = er32(CTRL);
2818 	ctrl |= E1000_CTRL_VME;
2819 	ew32(CTRL, ctrl);
2820 }
2821 
2822 static void e1000_update_mng_vlan(struct e1000_adapter *adapter)
2823 {
2824 	struct net_device *netdev = adapter->netdev;
2825 	u16 vid = adapter->hw.mng_cookie.vlan_id;
2826 	u16 old_vid = adapter->mng_vlan_id;
2827 
2828 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
2829 		e1000_vlan_rx_add_vid(netdev, htons(ETH_P_8021Q), vid);
2830 		adapter->mng_vlan_id = vid;
2831 	}
2832 
2833 	if ((old_vid != (u16)E1000_MNG_VLAN_NONE) && (vid != old_vid))
2834 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q), old_vid);
2835 }
2836 
2837 static void e1000_restore_vlan(struct e1000_adapter *adapter)
2838 {
2839 	u16 vid;
2840 
2841 	e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
2842 
2843 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
2844 	    e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
2845 }
2846 
2847 static void e1000_init_manageability_pt(struct e1000_adapter *adapter)
2848 {
2849 	struct e1000_hw *hw = &adapter->hw;
2850 	u32 manc, manc2h, mdef, i, j;
2851 
2852 	if (!(adapter->flags & FLAG_MNG_PT_ENABLED))
2853 		return;
2854 
2855 	manc = er32(MANC);
2856 
2857 	/* enable receiving management packets to the host. this will probably
2858 	 * generate destination unreachable messages from the host OS, but
2859 	 * the packets will be handled on SMBUS
2860 	 */
2861 	manc |= E1000_MANC_EN_MNG2HOST;
2862 	manc2h = er32(MANC2H);
2863 
2864 	switch (hw->mac.type) {
2865 	default:
2866 		manc2h |= (E1000_MANC2H_PORT_623 | E1000_MANC2H_PORT_664);
2867 		break;
2868 	case e1000_82574:
2869 	case e1000_82583:
2870 		/* Check if IPMI pass-through decision filter already exists;
2871 		 * if so, enable it.
2872 		 */
2873 		for (i = 0, j = 0; i < 8; i++) {
2874 			mdef = er32(MDEF(i));
2875 
2876 			/* Ignore filters with anything other than IPMI ports */
2877 			if (mdef & ~(E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2878 				continue;
2879 
2880 			/* Enable this decision filter in MANC2H */
2881 			if (mdef)
2882 				manc2h |= BIT(i);
2883 
2884 			j |= mdef;
2885 		}
2886 
2887 		if (j == (E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2888 			break;
2889 
2890 		/* Create new decision filter in an empty filter */
2891 		for (i = 0, j = 0; i < 8; i++)
2892 			if (er32(MDEF(i)) == 0) {
2893 				ew32(MDEF(i), (E1000_MDEF_PORT_623 |
2894 					       E1000_MDEF_PORT_664));
2895 				manc2h |= BIT(1);
2896 				j++;
2897 				break;
2898 			}
2899 
2900 		if (!j)
2901 			e_warn("Unable to create IPMI pass-through filter\n");
2902 		break;
2903 	}
2904 
2905 	ew32(MANC2H, manc2h);
2906 	ew32(MANC, manc);
2907 }
2908 
2909 /**
2910  * e1000_configure_tx - Configure Transmit Unit after Reset
2911  * @adapter: board private structure
2912  *
2913  * Configure the Tx unit of the MAC after a reset.
2914  **/
2915 static void e1000_configure_tx(struct e1000_adapter *adapter)
2916 {
2917 	struct e1000_hw *hw = &adapter->hw;
2918 	struct e1000_ring *tx_ring = adapter->tx_ring;
2919 	u64 tdba;
2920 	u32 tdlen, tctl, tarc;
2921 
2922 	/* Setup the HW Tx Head and Tail descriptor pointers */
2923 	tdba = tx_ring->dma;
2924 	tdlen = tx_ring->count * sizeof(struct e1000_tx_desc);
2925 	ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
2926 	ew32(TDBAH(0), (tdba >> 32));
2927 	ew32(TDLEN(0), tdlen);
2928 	ew32(TDH(0), 0);
2929 	ew32(TDT(0), 0);
2930 	tx_ring->head = adapter->hw.hw_addr + E1000_TDH(0);
2931 	tx_ring->tail = adapter->hw.hw_addr + E1000_TDT(0);
2932 
2933 	writel(0, tx_ring->head);
2934 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
2935 		e1000e_update_tdt_wa(tx_ring, 0);
2936 	else
2937 		writel(0, tx_ring->tail);
2938 
2939 	/* Set the Tx Interrupt Delay register */
2940 	ew32(TIDV, adapter->tx_int_delay);
2941 	/* Tx irq moderation */
2942 	ew32(TADV, adapter->tx_abs_int_delay);
2943 
2944 	if (adapter->flags2 & FLAG2_DMA_BURST) {
2945 		u32 txdctl = er32(TXDCTL(0));
2946 
2947 		txdctl &= ~(E1000_TXDCTL_PTHRESH | E1000_TXDCTL_HTHRESH |
2948 			    E1000_TXDCTL_WTHRESH);
2949 		/* set up some performance related parameters to encourage the
2950 		 * hardware to use the bus more efficiently in bursts, depends
2951 		 * on the tx_int_delay to be enabled,
2952 		 * wthresh = 1 ==> burst write is disabled to avoid Tx stalls
2953 		 * hthresh = 1 ==> prefetch when one or more available
2954 		 * pthresh = 0x1f ==> prefetch if internal cache 31 or less
2955 		 * BEWARE: this seems to work but should be considered first if
2956 		 * there are Tx hangs or other Tx related bugs
2957 		 */
2958 		txdctl |= E1000_TXDCTL_DMA_BURST_ENABLE;
2959 		ew32(TXDCTL(0), txdctl);
2960 	}
2961 	/* erratum work around: set txdctl the same for both queues */
2962 	ew32(TXDCTL(1), er32(TXDCTL(0)));
2963 
2964 	/* Program the Transmit Control Register */
2965 	tctl = er32(TCTL);
2966 	tctl &= ~E1000_TCTL_CT;
2967 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
2968 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
2969 
2970 	if (adapter->flags & FLAG_TARC_SPEED_MODE_BIT) {
2971 		tarc = er32(TARC(0));
2972 		/* set the speed mode bit, we'll clear it if we're not at
2973 		 * gigabit link later
2974 		 */
2975 #define SPEED_MODE_BIT BIT(21)
2976 		tarc |= SPEED_MODE_BIT;
2977 		ew32(TARC(0), tarc);
2978 	}
2979 
2980 	/* errata: program both queues to unweighted RR */
2981 	if (adapter->flags & FLAG_TARC_SET_BIT_ZERO) {
2982 		tarc = er32(TARC(0));
2983 		tarc |= 1;
2984 		ew32(TARC(0), tarc);
2985 		tarc = er32(TARC(1));
2986 		tarc |= 1;
2987 		ew32(TARC(1), tarc);
2988 	}
2989 
2990 	/* Setup Transmit Descriptor Settings for eop descriptor */
2991 	adapter->txd_cmd = E1000_TXD_CMD_EOP | E1000_TXD_CMD_IFCS;
2992 
2993 	/* only set IDE if we are delaying interrupts using the timers */
2994 	if (adapter->tx_int_delay)
2995 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
2996 
2997 	/* enable Report Status bit */
2998 	adapter->txd_cmd |= E1000_TXD_CMD_RS;
2999 
3000 	ew32(TCTL, tctl);
3001 
3002 	hw->mac.ops.config_collision_dist(hw);
3003 
3004 	/* SPT and KBL Si errata workaround to avoid data corruption */
3005 	if (hw->mac.type == e1000_pch_spt) {
3006 		u32 reg_val;
3007 
3008 		reg_val = er32(IOSFPC);
3009 		reg_val |= E1000_RCTL_RDMTS_HEX;
3010 		ew32(IOSFPC, reg_val);
3011 
3012 		reg_val = er32(TARC(0));
3013 		/* SPT and KBL Si errata workaround to avoid Tx hang.
3014 		 * Dropping the number of outstanding requests from
3015 		 * 3 to 2 in order to avoid a buffer overrun.
3016 		 */
3017 		reg_val &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3018 		reg_val |= E1000_TARC0_CB_MULTIQ_2_REQ;
3019 		ew32(TARC(0), reg_val);
3020 	}
3021 }
3022 
3023 #define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
3024 			   (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
3025 
3026 /**
3027  * e1000_setup_rctl - configure the receive control registers
3028  * @adapter: Board private structure
3029  **/
3030 static void e1000_setup_rctl(struct e1000_adapter *adapter)
3031 {
3032 	struct e1000_hw *hw = &adapter->hw;
3033 	u32 rctl, rfctl;
3034 	u32 pages = 0;
3035 
3036 	/* Workaround Si errata on PCHx - configure jumbo frame flow.
3037 	 * If jumbo frames not set, program related MAC/PHY registers
3038 	 * to h/w defaults
3039 	 */
3040 	if (hw->mac.type >= e1000_pch2lan) {
3041 		s32 ret_val;
3042 
3043 		if (adapter->netdev->mtu > ETH_DATA_LEN)
3044 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, true);
3045 		else
3046 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, false);
3047 
3048 		if (ret_val)
3049 			e_dbg("failed to enable|disable jumbo frame workaround mode\n");
3050 	}
3051 
3052 	/* Program MC offset vector base */
3053 	rctl = er32(RCTL);
3054 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3055 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3056 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3057 	    (adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3058 
3059 	/* Do not Store bad packets */
3060 	rctl &= ~E1000_RCTL_SBP;
3061 
3062 	/* Enable Long Packet receive */
3063 	if (adapter->netdev->mtu <= ETH_DATA_LEN)
3064 		rctl &= ~E1000_RCTL_LPE;
3065 	else
3066 		rctl |= E1000_RCTL_LPE;
3067 
3068 	/* Some systems expect that the CRC is included in SMBUS traffic. The
3069 	 * hardware strips the CRC before sending to both SMBUS (BMC) and to
3070 	 * host memory when this is enabled
3071 	 */
3072 	if (adapter->flags2 & FLAG2_CRC_STRIPPING)
3073 		rctl |= E1000_RCTL_SECRC;
3074 
3075 	/* Workaround Si errata on 82577 PHY - configure IPG for jumbos */
3076 	if ((hw->phy.type == e1000_phy_82577) && (rctl & E1000_RCTL_LPE)) {
3077 		u16 phy_data;
3078 
3079 		e1e_rphy(hw, PHY_REG(770, 26), &phy_data);
3080 		phy_data &= 0xfff8;
3081 		phy_data |= BIT(2);
3082 		e1e_wphy(hw, PHY_REG(770, 26), phy_data);
3083 
3084 		e1e_rphy(hw, 22, &phy_data);
3085 		phy_data &= 0x0fff;
3086 		phy_data |= BIT(14);
3087 		e1e_wphy(hw, 0x10, 0x2823);
3088 		e1e_wphy(hw, 0x11, 0x0003);
3089 		e1e_wphy(hw, 22, phy_data);
3090 	}
3091 
3092 	/* Setup buffer sizes */
3093 	rctl &= ~E1000_RCTL_SZ_4096;
3094 	rctl |= E1000_RCTL_BSEX;
3095 	switch (adapter->rx_buffer_len) {
3096 	case 2048:
3097 	default:
3098 		rctl |= E1000_RCTL_SZ_2048;
3099 		rctl &= ~E1000_RCTL_BSEX;
3100 		break;
3101 	case 4096:
3102 		rctl |= E1000_RCTL_SZ_4096;
3103 		break;
3104 	case 8192:
3105 		rctl |= E1000_RCTL_SZ_8192;
3106 		break;
3107 	case 16384:
3108 		rctl |= E1000_RCTL_SZ_16384;
3109 		break;
3110 	}
3111 
3112 	/* Enable Extended Status in all Receive Descriptors */
3113 	rfctl = er32(RFCTL);
3114 	rfctl |= E1000_RFCTL_EXTEN;
3115 	ew32(RFCTL, rfctl);
3116 
3117 	/* 82571 and greater support packet-split where the protocol
3118 	 * header is placed in skb->data and the packet data is
3119 	 * placed in pages hanging off of skb_shinfo(skb)->nr_frags.
3120 	 * In the case of a non-split, skb->data is linearly filled,
3121 	 * followed by the page buffers.  Therefore, skb->data is
3122 	 * sized to hold the largest protocol header.
3123 	 *
3124 	 * allocations using alloc_page take too long for regular MTU
3125 	 * so only enable packet split for jumbo frames
3126 	 *
3127 	 * Using pages when the page size is greater than 16k wastes
3128 	 * a lot of memory, since we allocate 3 pages at all times
3129 	 * per packet.
3130 	 */
3131 	pages = PAGE_USE_COUNT(adapter->netdev->mtu);
3132 	if ((pages <= 3) && (PAGE_SIZE <= 16384) && (rctl & E1000_RCTL_LPE))
3133 		adapter->rx_ps_pages = pages;
3134 	else
3135 		adapter->rx_ps_pages = 0;
3136 
3137 	if (adapter->rx_ps_pages) {
3138 		u32 psrctl = 0;
3139 
3140 		/* Enable Packet split descriptors */
3141 		rctl |= E1000_RCTL_DTYP_PS;
3142 
3143 		psrctl |= adapter->rx_ps_bsize0 >> E1000_PSRCTL_BSIZE0_SHIFT;
3144 
3145 		switch (adapter->rx_ps_pages) {
3146 		case 3:
3147 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE3_SHIFT;
3148 			fallthrough;
3149 		case 2:
3150 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE2_SHIFT;
3151 			fallthrough;
3152 		case 1:
3153 			psrctl |= PAGE_SIZE >> E1000_PSRCTL_BSIZE1_SHIFT;
3154 			break;
3155 		}
3156 
3157 		ew32(PSRCTL, psrctl);
3158 	}
3159 
3160 	/* This is useful for sniffing bad packets. */
3161 	if (adapter->netdev->features & NETIF_F_RXALL) {
3162 		/* UPE and MPE will be handled by normal PROMISC logic
3163 		 * in e1000e_set_rx_mode
3164 		 */
3165 		rctl |= (E1000_RCTL_SBP |	/* Receive bad packets */
3166 			 E1000_RCTL_BAM |	/* RX All Bcast Pkts */
3167 			 E1000_RCTL_PMCF);	/* RX All MAC Ctrl Pkts */
3168 
3169 		rctl &= ~(E1000_RCTL_VFE |	/* Disable VLAN filter */
3170 			  E1000_RCTL_DPF |	/* Allow filtered pause */
3171 			  E1000_RCTL_CFIEN);	/* Dis VLAN CFIEN Filter */
3172 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3173 		 * and that breaks VLANs.
3174 		 */
3175 	}
3176 
3177 	ew32(RCTL, rctl);
3178 	/* just started the receive unit, no need to restart */
3179 	adapter->flags &= ~FLAG_RESTART_NOW;
3180 }
3181 
3182 /**
3183  * e1000_configure_rx - Configure Receive Unit after Reset
3184  * @adapter: board private structure
3185  *
3186  * Configure the Rx unit of the MAC after a reset.
3187  **/
3188 static void e1000_configure_rx(struct e1000_adapter *adapter)
3189 {
3190 	struct e1000_hw *hw = &adapter->hw;
3191 	struct e1000_ring *rx_ring = adapter->rx_ring;
3192 	u64 rdba;
3193 	u32 rdlen, rctl, rxcsum, ctrl_ext;
3194 
3195 	if (adapter->rx_ps_pages) {
3196 		/* this is a 32 byte descriptor */
3197 		rdlen = rx_ring->count *
3198 		    sizeof(union e1000_rx_desc_packet_split);
3199 		adapter->clean_rx = e1000_clean_rx_irq_ps;
3200 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers_ps;
3201 	} else if (adapter->netdev->mtu > ETH_FRAME_LEN + ETH_FCS_LEN) {
3202 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3203 		adapter->clean_rx = e1000_clean_jumbo_rx_irq;
3204 		adapter->alloc_rx_buf = e1000_alloc_jumbo_rx_buffers;
3205 	} else {
3206 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3207 		adapter->clean_rx = e1000_clean_rx_irq;
3208 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers;
3209 	}
3210 
3211 	/* disable receives while setting up the descriptors */
3212 	rctl = er32(RCTL);
3213 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
3214 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
3215 	e1e_flush();
3216 	usleep_range(10000, 11000);
3217 
3218 	if (adapter->flags2 & FLAG2_DMA_BURST) {
3219 		/* set the writeback threshold (only takes effect if the RDTR
3220 		 * is set). set GRAN=1 and write back up to 0x4 worth, and
3221 		 * enable prefetching of 0x20 Rx descriptors
3222 		 * granularity = 01
3223 		 * wthresh = 04,
3224 		 * hthresh = 04,
3225 		 * pthresh = 0x20
3226 		 */
3227 		ew32(RXDCTL(0), E1000_RXDCTL_DMA_BURST_ENABLE);
3228 		ew32(RXDCTL(1), E1000_RXDCTL_DMA_BURST_ENABLE);
3229 	}
3230 
3231 	/* set the Receive Delay Timer Register */
3232 	ew32(RDTR, adapter->rx_int_delay);
3233 
3234 	/* irq moderation */
3235 	ew32(RADV, adapter->rx_abs_int_delay);
3236 	if ((adapter->itr_setting != 0) && (adapter->itr != 0))
3237 		e1000e_write_itr(adapter, adapter->itr);
3238 
3239 	ctrl_ext = er32(CTRL_EXT);
3240 	/* Auto-Mask interrupts upon ICR access */
3241 	ctrl_ext |= E1000_CTRL_EXT_IAME;
3242 	ew32(IAM, 0xffffffff);
3243 	ew32(CTRL_EXT, ctrl_ext);
3244 	e1e_flush();
3245 
3246 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
3247 	 * the Base and Length of the Rx Descriptor Ring
3248 	 */
3249 	rdba = rx_ring->dma;
3250 	ew32(RDBAL(0), (rdba & DMA_BIT_MASK(32)));
3251 	ew32(RDBAH(0), (rdba >> 32));
3252 	ew32(RDLEN(0), rdlen);
3253 	ew32(RDH(0), 0);
3254 	ew32(RDT(0), 0);
3255 	rx_ring->head = adapter->hw.hw_addr + E1000_RDH(0);
3256 	rx_ring->tail = adapter->hw.hw_addr + E1000_RDT(0);
3257 
3258 	writel(0, rx_ring->head);
3259 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
3260 		e1000e_update_rdt_wa(rx_ring, 0);
3261 	else
3262 		writel(0, rx_ring->tail);
3263 
3264 	/* Enable Receive Checksum Offload for TCP and UDP */
3265 	rxcsum = er32(RXCSUM);
3266 	if (adapter->netdev->features & NETIF_F_RXCSUM)
3267 		rxcsum |= E1000_RXCSUM_TUOFL;
3268 	else
3269 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3270 	ew32(RXCSUM, rxcsum);
3271 
3272 	/* With jumbo frames, excessive C-state transition latencies result
3273 	 * in dropped transactions.
3274 	 */
3275 	if (adapter->netdev->mtu > ETH_DATA_LEN) {
3276 		u32 lat =
3277 		    ((er32(PBA) & E1000_PBA_RXA_MASK) * 1024 -
3278 		     adapter->max_frame_size) * 8 / 1000;
3279 
3280 		if (adapter->flags & FLAG_IS_ICH) {
3281 			u32 rxdctl = er32(RXDCTL(0));
3282 
3283 			ew32(RXDCTL(0), rxdctl | 0x3 | BIT(8));
3284 		}
3285 
3286 		dev_info(&adapter->pdev->dev,
3287 			 "Some CPU C-states have been disabled in order to enable jumbo frames\n");
3288 		cpu_latency_qos_update_request(&adapter->pm_qos_req, lat);
3289 	} else {
3290 		cpu_latency_qos_update_request(&adapter->pm_qos_req,
3291 					       PM_QOS_DEFAULT_VALUE);
3292 	}
3293 
3294 	/* Enable Receives */
3295 	ew32(RCTL, rctl);
3296 }
3297 
3298 /**
3299  * e1000e_write_mc_addr_list - write multicast addresses to MTA
3300  * @netdev: network interface device structure
3301  *
3302  * Writes multicast address list to the MTA hash table.
3303  * Returns: -ENOMEM on failure
3304  *                0 on no addresses written
3305  *                X on writing X addresses to MTA
3306  */
3307 static int e1000e_write_mc_addr_list(struct net_device *netdev)
3308 {
3309 	struct e1000_adapter *adapter = netdev_priv(netdev);
3310 	struct e1000_hw *hw = &adapter->hw;
3311 	struct netdev_hw_addr *ha;
3312 	u8 *mta_list;
3313 	int i;
3314 
3315 	if (netdev_mc_empty(netdev)) {
3316 		/* nothing to program, so clear mc list */
3317 		hw->mac.ops.update_mc_addr_list(hw, NULL, 0);
3318 		return 0;
3319 	}
3320 
3321 	mta_list = kcalloc(netdev_mc_count(netdev), ETH_ALEN, GFP_ATOMIC);
3322 	if (!mta_list)
3323 		return -ENOMEM;
3324 
3325 	/* update_mc_addr_list expects a packed array of only addresses. */
3326 	i = 0;
3327 	netdev_for_each_mc_addr(ha, netdev)
3328 	    memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3329 
3330 	hw->mac.ops.update_mc_addr_list(hw, mta_list, i);
3331 	kfree(mta_list);
3332 
3333 	return netdev_mc_count(netdev);
3334 }
3335 
3336 /**
3337  * e1000e_write_uc_addr_list - write unicast addresses to RAR table
3338  * @netdev: network interface device structure
3339  *
3340  * Writes unicast address list to the RAR table.
3341  * Returns: -ENOMEM on failure/insufficient address space
3342  *                0 on no addresses written
3343  *                X on writing X addresses to the RAR table
3344  **/
3345 static int e1000e_write_uc_addr_list(struct net_device *netdev)
3346 {
3347 	struct e1000_adapter *adapter = netdev_priv(netdev);
3348 	struct e1000_hw *hw = &adapter->hw;
3349 	unsigned int rar_entries;
3350 	int count = 0;
3351 
3352 	rar_entries = hw->mac.ops.rar_get_count(hw);
3353 
3354 	/* save a rar entry for our hardware address */
3355 	rar_entries--;
3356 
3357 	/* save a rar entry for the LAA workaround */
3358 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA)
3359 		rar_entries--;
3360 
3361 	/* return ENOMEM indicating insufficient memory for addresses */
3362 	if (netdev_uc_count(netdev) > rar_entries)
3363 		return -ENOMEM;
3364 
3365 	if (!netdev_uc_empty(netdev) && rar_entries) {
3366 		struct netdev_hw_addr *ha;
3367 
3368 		/* write the addresses in reverse order to avoid write
3369 		 * combining
3370 		 */
3371 		netdev_for_each_uc_addr(ha, netdev) {
3372 			int ret_val;
3373 
3374 			if (!rar_entries)
3375 				break;
3376 			ret_val = hw->mac.ops.rar_set(hw, ha->addr, rar_entries--);
3377 			if (ret_val < 0)
3378 				return -ENOMEM;
3379 			count++;
3380 		}
3381 	}
3382 
3383 	/* zero out the remaining RAR entries not used above */
3384 	for (; rar_entries > 0; rar_entries--) {
3385 		ew32(RAH(rar_entries), 0);
3386 		ew32(RAL(rar_entries), 0);
3387 	}
3388 	e1e_flush();
3389 
3390 	return count;
3391 }
3392 
3393 /**
3394  * e1000e_set_rx_mode - secondary unicast, Multicast and Promiscuous mode set
3395  * @netdev: network interface device structure
3396  *
3397  * The ndo_set_rx_mode entry point is called whenever the unicast or multicast
3398  * address list or the network interface flags are updated.  This routine is
3399  * responsible for configuring the hardware for proper unicast, multicast,
3400  * promiscuous mode, and all-multi behavior.
3401  **/
3402 static void e1000e_set_rx_mode(struct net_device *netdev)
3403 {
3404 	struct e1000_adapter *adapter = netdev_priv(netdev);
3405 	struct e1000_hw *hw = &adapter->hw;
3406 	u32 rctl;
3407 
3408 	if (pm_runtime_suspended(netdev->dev.parent))
3409 		return;
3410 
3411 	/* Check for Promiscuous and All Multicast modes */
3412 	rctl = er32(RCTL);
3413 
3414 	/* clear the affected bits */
3415 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
3416 
3417 	if (netdev->flags & IFF_PROMISC) {
3418 		rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
3419 		/* Do not hardware filter VLANs in promisc mode */
3420 		e1000e_vlan_filter_disable(adapter);
3421 	} else {
3422 		int count;
3423 
3424 		if (netdev->flags & IFF_ALLMULTI) {
3425 			rctl |= E1000_RCTL_MPE;
3426 		} else {
3427 			/* Write addresses to the MTA, if the attempt fails
3428 			 * then we should just turn on promiscuous mode so
3429 			 * that we can at least receive multicast traffic
3430 			 */
3431 			count = e1000e_write_mc_addr_list(netdev);
3432 			if (count < 0)
3433 				rctl |= E1000_RCTL_MPE;
3434 		}
3435 		e1000e_vlan_filter_enable(adapter);
3436 		/* Write addresses to available RAR registers, if there is not
3437 		 * sufficient space to store all the addresses then enable
3438 		 * unicast promiscuous mode
3439 		 */
3440 		count = e1000e_write_uc_addr_list(netdev);
3441 		if (count < 0)
3442 			rctl |= E1000_RCTL_UPE;
3443 	}
3444 
3445 	ew32(RCTL, rctl);
3446 
3447 	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
3448 		e1000e_vlan_strip_enable(adapter);
3449 	else
3450 		e1000e_vlan_strip_disable(adapter);
3451 }
3452 
3453 static void e1000e_setup_rss_hash(struct e1000_adapter *adapter)
3454 {
3455 	struct e1000_hw *hw = &adapter->hw;
3456 	u32 mrqc, rxcsum;
3457 	u32 rss_key[10];
3458 	int i;
3459 
3460 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3461 	for (i = 0; i < 10; i++)
3462 		ew32(RSSRK(i), rss_key[i]);
3463 
3464 	/* Direct all traffic to queue 0 */
3465 	for (i = 0; i < 32; i++)
3466 		ew32(RETA(i), 0);
3467 
3468 	/* Disable raw packet checksumming so that RSS hash is placed in
3469 	 * descriptor on writeback.
3470 	 */
3471 	rxcsum = er32(RXCSUM);
3472 	rxcsum |= E1000_RXCSUM_PCSD;
3473 
3474 	ew32(RXCSUM, rxcsum);
3475 
3476 	mrqc = (E1000_MRQC_RSS_FIELD_IPV4 |
3477 		E1000_MRQC_RSS_FIELD_IPV4_TCP |
3478 		E1000_MRQC_RSS_FIELD_IPV6 |
3479 		E1000_MRQC_RSS_FIELD_IPV6_TCP |
3480 		E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3481 
3482 	ew32(MRQC, mrqc);
3483 }
3484 
3485 /**
3486  * e1000e_get_base_timinca - get default SYSTIM time increment attributes
3487  * @adapter: board private structure
3488  * @timinca: pointer to returned time increment attributes
3489  *
3490  * Get attributes for incrementing the System Time Register SYSTIML/H at
3491  * the default base frequency, and set the cyclecounter shift value.
3492  **/
3493 s32 e1000e_get_base_timinca(struct e1000_adapter *adapter, u32 *timinca)
3494 {
3495 	struct e1000_hw *hw = &adapter->hw;
3496 	u32 incvalue, incperiod, shift;
3497 
3498 	/* Make sure clock is enabled on I217/I218/I219  before checking
3499 	 * the frequency
3500 	 */
3501 	if ((hw->mac.type >= e1000_pch_lpt) &&
3502 	    !(er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) &&
3503 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_ENABLED)) {
3504 		u32 fextnvm7 = er32(FEXTNVM7);
3505 
3506 		if (!(fextnvm7 & BIT(0))) {
3507 			ew32(FEXTNVM7, fextnvm7 | BIT(0));
3508 			e1e_flush();
3509 		}
3510 	}
3511 
3512 	switch (hw->mac.type) {
3513 	case e1000_pch2lan:
3514 		/* Stable 96MHz frequency */
3515 		incperiod = INCPERIOD_96MHZ;
3516 		incvalue = INCVALUE_96MHZ;
3517 		shift = INCVALUE_SHIFT_96MHZ;
3518 		adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3519 		break;
3520 	case e1000_pch_lpt:
3521 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3522 			/* Stable 96MHz frequency */
3523 			incperiod = INCPERIOD_96MHZ;
3524 			incvalue = INCVALUE_96MHZ;
3525 			shift = INCVALUE_SHIFT_96MHZ;
3526 			adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3527 		} else {
3528 			/* Stable 25MHz frequency */
3529 			incperiod = INCPERIOD_25MHZ;
3530 			incvalue = INCVALUE_25MHZ;
3531 			shift = INCVALUE_SHIFT_25MHZ;
3532 			adapter->cc.shift = shift;
3533 		}
3534 		break;
3535 	case e1000_pch_spt:
3536 		/* Stable 24MHz frequency */
3537 		incperiod = INCPERIOD_24MHZ;
3538 		incvalue = INCVALUE_24MHZ;
3539 		shift = INCVALUE_SHIFT_24MHZ;
3540 		adapter->cc.shift = shift;
3541 		break;
3542 	case e1000_pch_cnp:
3543 	case e1000_pch_tgp:
3544 	case e1000_pch_adp:
3545 	case e1000_pch_mtp:
3546 	case e1000_pch_lnp:
3547 	case e1000_pch_ptp:
3548 	case e1000_pch_nvp:
3549 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3550 			/* Stable 24MHz frequency */
3551 			incperiod = INCPERIOD_24MHZ;
3552 			incvalue = INCVALUE_24MHZ;
3553 			shift = INCVALUE_SHIFT_24MHZ;
3554 			adapter->cc.shift = shift;
3555 		} else {
3556 			/* Stable 38400KHz frequency */
3557 			incperiod = INCPERIOD_38400KHZ;
3558 			incvalue = INCVALUE_38400KHZ;
3559 			shift = INCVALUE_SHIFT_38400KHZ;
3560 			adapter->cc.shift = shift;
3561 		}
3562 		break;
3563 	case e1000_82574:
3564 	case e1000_82583:
3565 		/* Stable 25MHz frequency */
3566 		incperiod = INCPERIOD_25MHZ;
3567 		incvalue = INCVALUE_25MHZ;
3568 		shift = INCVALUE_SHIFT_25MHZ;
3569 		adapter->cc.shift = shift;
3570 		break;
3571 	default:
3572 		return -EINVAL;
3573 	}
3574 
3575 	*timinca = ((incperiod << E1000_TIMINCA_INCPERIOD_SHIFT) |
3576 		    ((incvalue << shift) & E1000_TIMINCA_INCVALUE_MASK));
3577 
3578 	return 0;
3579 }
3580 
3581 /**
3582  * e1000e_config_hwtstamp - configure the hwtstamp registers and enable/disable
3583  * @adapter: board private structure
3584  * @config: timestamp configuration
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 hwtstamp_config *config)
3599 {
3600 	struct e1000_hw *hw = &adapter->hw;
3601 	u32 tsync_tx_ctl = E1000_TSYNCTXCTL_ENABLED;
3602 	u32 tsync_rx_ctl = E1000_TSYNCRXCTL_ENABLED;
3603 	u32 rxmtrl = 0;
3604 	u16 rxudp = 0;
3605 	bool is_l4 = false;
3606 	bool is_l2 = false;
3607 	u32 regval;
3608 
3609 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3610 		return -EINVAL;
3611 
3612 	switch (config->tx_type) {
3613 	case HWTSTAMP_TX_OFF:
3614 		tsync_tx_ctl = 0;
3615 		break;
3616 	case HWTSTAMP_TX_ON:
3617 		break;
3618 	default:
3619 		return -ERANGE;
3620 	}
3621 
3622 	switch (config->rx_filter) {
3623 	case HWTSTAMP_FILTER_NONE:
3624 		tsync_rx_ctl = 0;
3625 		break;
3626 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3627 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3628 		rxmtrl = E1000_RXMTRL_PTP_V1_SYNC_MESSAGE;
3629 		is_l4 = true;
3630 		break;
3631 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3632 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3633 		rxmtrl = E1000_RXMTRL_PTP_V1_DELAY_REQ_MESSAGE;
3634 		is_l4 = true;
3635 		break;
3636 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3637 		/* Also time stamps V2 L2 Path Delay Request/Response */
3638 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3639 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3640 		is_l2 = true;
3641 		break;
3642 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3643 		/* Also time stamps V2 L2 Path Delay Request/Response. */
3644 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3645 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3646 		is_l2 = true;
3647 		break;
3648 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3649 		/* Hardware cannot filter just V2 L4 Sync messages */
3650 		fallthrough;
3651 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
3652 		/* Also time stamps V2 Path Delay Request/Response. */
3653 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3654 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3655 		is_l2 = true;
3656 		is_l4 = true;
3657 		break;
3658 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3659 		/* Hardware cannot filter just V2 L4 Delay Request messages */
3660 		fallthrough;
3661 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
3662 		/* Also time stamps V2 Path Delay Request/Response. */
3663 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3664 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3665 		is_l2 = true;
3666 		is_l4 = true;
3667 		break;
3668 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
3669 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
3670 		/* Hardware cannot filter just V2 L4 or L2 Event messages */
3671 		fallthrough;
3672 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
3673 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_EVENT_V2;
3674 		config->rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
3675 		is_l2 = true;
3676 		is_l4 = true;
3677 		break;
3678 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3679 		/* For V1, the hardware can only filter Sync messages or
3680 		 * Delay Request messages but not both so fall-through to
3681 		 * time stamp all packets.
3682 		 */
3683 		fallthrough;
3684 	case HWTSTAMP_FILTER_NTP_ALL:
3685 	case HWTSTAMP_FILTER_ALL:
3686 		is_l2 = true;
3687 		is_l4 = true;
3688 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_ALL;
3689 		config->rx_filter = HWTSTAMP_FILTER_ALL;
3690 		break;
3691 	default:
3692 		return -ERANGE;
3693 	}
3694 
3695 	adapter->hwtstamp_config = *config;
3696 
3697 	/* enable/disable Tx h/w time stamping */
3698 	regval = er32(TSYNCTXCTL);
3699 	regval &= ~E1000_TSYNCTXCTL_ENABLED;
3700 	regval |= tsync_tx_ctl;
3701 	ew32(TSYNCTXCTL, regval);
3702 	if ((er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) !=
3703 	    (regval & E1000_TSYNCTXCTL_ENABLED)) {
3704 		e_err("Timesync Tx Control register not set as expected\n");
3705 		return -EAGAIN;
3706 	}
3707 
3708 	/* enable/disable Rx h/w time stamping */
3709 	regval = er32(TSYNCRXCTL);
3710 	regval &= ~(E1000_TSYNCRXCTL_ENABLED | E1000_TSYNCRXCTL_TYPE_MASK);
3711 	regval |= tsync_rx_ctl;
3712 	ew32(TSYNCRXCTL, regval);
3713 	if ((er32(TSYNCRXCTL) & (E1000_TSYNCRXCTL_ENABLED |
3714 				 E1000_TSYNCRXCTL_TYPE_MASK)) !=
3715 	    (regval & (E1000_TSYNCRXCTL_ENABLED |
3716 		       E1000_TSYNCRXCTL_TYPE_MASK))) {
3717 		e_err("Timesync Rx Control register not set as expected\n");
3718 		return -EAGAIN;
3719 	}
3720 
3721 	/* L2: define ethertype filter for time stamped packets */
3722 	if (is_l2)
3723 		rxmtrl |= ETH_P_1588;
3724 
3725 	/* define which PTP packets get time stamped */
3726 	ew32(RXMTRL, rxmtrl);
3727 
3728 	/* Filter by destination port */
3729 	if (is_l4) {
3730 		rxudp = PTP_EV_PORT;
3731 		cpu_to_be16s(&rxudp);
3732 	}
3733 	ew32(RXUDP, rxudp);
3734 
3735 	e1e_flush();
3736 
3737 	/* Clear TSYNCRXCTL_VALID & TSYNCTXCTL_VALID bit */
3738 	er32(RXSTMPH);
3739 	er32(TXSTMPH);
3740 
3741 	return 0;
3742 }
3743 
3744 /**
3745  * e1000_configure - configure the hardware for Rx and Tx
3746  * @adapter: private board structure
3747  **/
3748 static void e1000_configure(struct e1000_adapter *adapter)
3749 {
3750 	struct e1000_ring *rx_ring = adapter->rx_ring;
3751 
3752 	e1000e_set_rx_mode(adapter->netdev);
3753 
3754 	e1000_restore_vlan(adapter);
3755 	e1000_init_manageability_pt(adapter);
3756 
3757 	e1000_configure_tx(adapter);
3758 
3759 	if (adapter->netdev->features & NETIF_F_RXHASH)
3760 		e1000e_setup_rss_hash(adapter);
3761 	e1000_setup_rctl(adapter);
3762 	e1000_configure_rx(adapter);
3763 	adapter->alloc_rx_buf(rx_ring, e1000_desc_unused(rx_ring), GFP_KERNEL);
3764 }
3765 
3766 /**
3767  * e1000e_power_up_phy - restore link in case the phy was powered down
3768  * @adapter: address of board private structure
3769  *
3770  * The phy may be powered down to save power and turn off link when the
3771  * driver is unloaded and wake on lan is not enabled (among others)
3772  * *** this routine MUST be followed by a call to e1000e_reset ***
3773  **/
3774 void e1000e_power_up_phy(struct e1000_adapter *adapter)
3775 {
3776 	if (adapter->hw.phy.ops.power_up)
3777 		adapter->hw.phy.ops.power_up(&adapter->hw);
3778 
3779 	adapter->hw.mac.ops.setup_link(&adapter->hw);
3780 }
3781 
3782 /**
3783  * e1000_power_down_phy - Power down the PHY
3784  * @adapter: board private structure
3785  *
3786  * Power down the PHY so no link is implied when interface is down.
3787  * The PHY cannot be powered down if management or WoL is active.
3788  */
3789 static void e1000_power_down_phy(struct e1000_adapter *adapter)
3790 {
3791 	if (adapter->hw.phy.ops.power_down)
3792 		adapter->hw.phy.ops.power_down(&adapter->hw);
3793 }
3794 
3795 /**
3796  * e1000_flush_tx_ring - remove all descriptors from the tx_ring
3797  * @adapter: board private structure
3798  *
3799  * We want to clear all pending descriptors from the TX ring.
3800  * zeroing happens when the HW reads the regs. We  assign the ring itself as
3801  * the data of the next descriptor. We don't care about the data we are about
3802  * to reset the HW.
3803  */
3804 static void e1000_flush_tx_ring(struct e1000_adapter *adapter)
3805 {
3806 	struct e1000_hw *hw = &adapter->hw;
3807 	struct e1000_ring *tx_ring = adapter->tx_ring;
3808 	struct e1000_tx_desc *tx_desc = NULL;
3809 	u32 tdt, tctl, txd_lower = E1000_TXD_CMD_IFCS;
3810 	u16 size = 512;
3811 
3812 	tctl = er32(TCTL);
3813 	ew32(TCTL, tctl | E1000_TCTL_EN);
3814 	tdt = er32(TDT(0));
3815 	BUG_ON(tdt != tx_ring->next_to_use);
3816 	tx_desc =  E1000_TX_DESC(*tx_ring, tx_ring->next_to_use);
3817 	tx_desc->buffer_addr = cpu_to_le64(tx_ring->dma);
3818 
3819 	tx_desc->lower.data = cpu_to_le32(txd_lower | size);
3820 	tx_desc->upper.data = 0;
3821 	/* flush descriptors to memory before notifying the HW */
3822 	wmb();
3823 	tx_ring->next_to_use++;
3824 	if (tx_ring->next_to_use == tx_ring->count)
3825 		tx_ring->next_to_use = 0;
3826 	ew32(TDT(0), tx_ring->next_to_use);
3827 	usleep_range(200, 250);
3828 }
3829 
3830 /**
3831  * e1000_flush_rx_ring - remove all descriptors from the rx_ring
3832  * @adapter: board private structure
3833  *
3834  * Mark all descriptors in the RX ring as consumed and disable the rx ring
3835  */
3836 static void e1000_flush_rx_ring(struct e1000_adapter *adapter)
3837 {
3838 	u32 rctl, rxdctl;
3839 	struct e1000_hw *hw = &adapter->hw;
3840 
3841 	rctl = er32(RCTL);
3842 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3843 	e1e_flush();
3844 	usleep_range(100, 150);
3845 
3846 	rxdctl = er32(RXDCTL(0));
3847 	/* zero the lower 14 bits (prefetch and host thresholds) */
3848 	rxdctl &= 0xffffc000;
3849 
3850 	/* update thresholds: prefetch threshold to 31, host threshold to 1
3851 	 * and make sure the granularity is "descriptors" and not "cache lines"
3852 	 */
3853 	rxdctl |= (0x1F | BIT(8) | E1000_RXDCTL_THRESH_UNIT_DESC);
3854 
3855 	ew32(RXDCTL(0), rxdctl);
3856 	/* momentarily enable the RX ring for the changes to take effect */
3857 	ew32(RCTL, rctl | E1000_RCTL_EN);
3858 	e1e_flush();
3859 	usleep_range(100, 150);
3860 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3861 }
3862 
3863 /**
3864  * e1000_flush_desc_rings - remove all descriptors from the descriptor rings
3865  * @adapter: board private structure
3866  *
3867  * In i219, the descriptor rings must be emptied before resetting the HW
3868  * or before changing the device state to D3 during runtime (runtime PM).
3869  *
3870  * Failure to do this will cause the HW to enter a unit hang state which can
3871  * only be released by PCI reset on the device
3872  *
3873  */
3874 
3875 static void e1000_flush_desc_rings(struct e1000_adapter *adapter)
3876 {
3877 	u16 hang_state;
3878 	u32 fext_nvm11, tdlen;
3879 	struct e1000_hw *hw = &adapter->hw;
3880 
3881 	/* First, disable MULR fix in FEXTNVM11 */
3882 	fext_nvm11 = er32(FEXTNVM11);
3883 	fext_nvm11 |= E1000_FEXTNVM11_DISABLE_MULR_FIX;
3884 	ew32(FEXTNVM11, fext_nvm11);
3885 	/* do nothing if we're not in faulty state, or if the queue is empty */
3886 	tdlen = er32(TDLEN(0));
3887 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3888 			     &hang_state);
3889 	if (!(hang_state & FLUSH_DESC_REQUIRED) || !tdlen)
3890 		return;
3891 	e1000_flush_tx_ring(adapter);
3892 	/* recheck, maybe the fault is caused by the rx ring */
3893 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3894 			     &hang_state);
3895 	if (hang_state & FLUSH_DESC_REQUIRED)
3896 		e1000_flush_rx_ring(adapter);
3897 }
3898 
3899 /**
3900  * e1000e_systim_reset - reset the timesync registers after a hardware reset
3901  * @adapter: board private structure
3902  *
3903  * When the MAC is reset, all hardware bits for timesync will be reset to the
3904  * default values. This function will restore the settings last in place.
3905  * Since the clock SYSTIME registers are reset, we will simply restore the
3906  * cyclecounter to the kernel real clock time.
3907  **/
3908 static void e1000e_systim_reset(struct e1000_adapter *adapter)
3909 {
3910 	struct ptp_clock_info *info = &adapter->ptp_clock_info;
3911 	struct e1000_hw *hw = &adapter->hw;
3912 	unsigned long flags;
3913 	u32 timinca;
3914 	s32 ret_val;
3915 
3916 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3917 		return;
3918 
3919 	if (info->adjfine) {
3920 		/* restore the previous ptp frequency delta */
3921 		ret_val = info->adjfine(info, adapter->ptp_delta);
3922 	} else {
3923 		/* set the default base frequency if no adjustment possible */
3924 		ret_val = e1000e_get_base_timinca(adapter, &timinca);
3925 		if (!ret_val)
3926 			ew32(TIMINCA, timinca);
3927 	}
3928 
3929 	if (ret_val) {
3930 		dev_warn(&adapter->pdev->dev,
3931 			 "Failed to restore TIMINCA clock rate delta: %d\n",
3932 			 ret_val);
3933 		return;
3934 	}
3935 
3936 	/* reset the systim ns time counter */
3937 	spin_lock_irqsave(&adapter->systim_lock, flags);
3938 	timecounter_init(&adapter->tc, &adapter->cc,
3939 			 ktime_to_ns(ktime_get_real()));
3940 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
3941 
3942 	/* restore the previous hwtstamp configuration settings */
3943 	e1000e_config_hwtstamp(adapter, &adapter->hwtstamp_config);
3944 }
3945 
3946 /**
3947  * e1000e_reset - bring the hardware into a known good state
3948  * @adapter: board private structure
3949  *
3950  * This function boots the hardware and enables some settings that
3951  * require a configuration cycle of the hardware - those cannot be
3952  * set/changed during runtime. After reset the device needs to be
3953  * properly configured for Rx, Tx etc.
3954  */
3955 void e1000e_reset(struct e1000_adapter *adapter)
3956 {
3957 	struct e1000_mac_info *mac = &adapter->hw.mac;
3958 	struct e1000_fc_info *fc = &adapter->hw.fc;
3959 	struct e1000_hw *hw = &adapter->hw;
3960 	u32 tx_space, min_tx_space, min_rx_space;
3961 	u32 pba = adapter->pba;
3962 	u16 hwm;
3963 
3964 	/* reset Packet Buffer Allocation to default */
3965 	ew32(PBA, pba);
3966 
3967 	if (adapter->max_frame_size > (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN)) {
3968 		/* To maintain wire speed transmits, the Tx FIFO should be
3969 		 * large enough to accommodate two full transmit packets,
3970 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
3971 		 * the Rx FIFO should be large enough to accommodate at least
3972 		 * one full receive packet and is similarly rounded up and
3973 		 * expressed in KB.
3974 		 */
3975 		pba = er32(PBA);
3976 		/* upper 16 bits has Tx packet buffer allocation size in KB */
3977 		tx_space = pba >> 16;
3978 		/* lower 16 bits has Rx packet buffer allocation size in KB */
3979 		pba &= 0xffff;
3980 		/* the Tx fifo also stores 16 bytes of information about the Tx
3981 		 * but don't include ethernet FCS because hardware appends it
3982 		 */
3983 		min_tx_space = (adapter->max_frame_size +
3984 				sizeof(struct e1000_tx_desc) - ETH_FCS_LEN) * 2;
3985 		min_tx_space = ALIGN(min_tx_space, 1024);
3986 		min_tx_space >>= 10;
3987 		/* software strips receive CRC, so leave room for it */
3988 		min_rx_space = adapter->max_frame_size;
3989 		min_rx_space = ALIGN(min_rx_space, 1024);
3990 		min_rx_space >>= 10;
3991 
3992 		/* If current Tx allocation is less than the min Tx FIFO size,
3993 		 * and the min Tx FIFO size is less than the current Rx FIFO
3994 		 * allocation, take space away from current Rx allocation
3995 		 */
3996 		if ((tx_space < min_tx_space) &&
3997 		    ((min_tx_space - tx_space) < pba)) {
3998 			pba -= min_tx_space - tx_space;
3999 
4000 			/* if short on Rx space, Rx wins and must trump Tx
4001 			 * adjustment
4002 			 */
4003 			if (pba < min_rx_space)
4004 				pba = min_rx_space;
4005 		}
4006 
4007 		ew32(PBA, pba);
4008 	}
4009 
4010 	/* flow control settings
4011 	 *
4012 	 * The high water mark must be low enough to fit one full frame
4013 	 * (or the size used for early receive) above it in the Rx FIFO.
4014 	 * Set it to the lower of:
4015 	 * - 90% of the Rx FIFO size, and
4016 	 * - the full Rx FIFO size minus one full frame
4017 	 */
4018 	if (adapter->flags & FLAG_DISABLE_FC_PAUSE_TIME)
4019 		fc->pause_time = 0xFFFF;
4020 	else
4021 		fc->pause_time = E1000_FC_PAUSE_TIME;
4022 	fc->send_xon = true;
4023 	fc->current_mode = fc->requested_mode;
4024 
4025 	switch (hw->mac.type) {
4026 	case e1000_ich9lan:
4027 	case e1000_ich10lan:
4028 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4029 			pba = 14;
4030 			ew32(PBA, pba);
4031 			fc->high_water = 0x2800;
4032 			fc->low_water = fc->high_water - 8;
4033 			break;
4034 		}
4035 		fallthrough;
4036 	default:
4037 		hwm = min(((pba << 10) * 9 / 10),
4038 			  ((pba << 10) - adapter->max_frame_size));
4039 
4040 		fc->high_water = hwm & E1000_FCRTH_RTH;	/* 8-byte granularity */
4041 		fc->low_water = fc->high_water - 8;
4042 		break;
4043 	case e1000_pchlan:
4044 		/* Workaround PCH LOM adapter hangs with certain network
4045 		 * loads.  If hangs persist, try disabling Tx flow control.
4046 		 */
4047 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4048 			fc->high_water = 0x3500;
4049 			fc->low_water = 0x1500;
4050 		} else {
4051 			fc->high_water = 0x5000;
4052 			fc->low_water = 0x3000;
4053 		}
4054 		fc->refresh_time = 0x1000;
4055 		break;
4056 	case e1000_pch2lan:
4057 	case e1000_pch_lpt:
4058 	case e1000_pch_spt:
4059 	case e1000_pch_cnp:
4060 	case e1000_pch_tgp:
4061 	case e1000_pch_adp:
4062 	case e1000_pch_mtp:
4063 	case e1000_pch_lnp:
4064 	case e1000_pch_ptp:
4065 	case e1000_pch_nvp:
4066 		fc->refresh_time = 0xFFFF;
4067 		fc->pause_time = 0xFFFF;
4068 
4069 		if (adapter->netdev->mtu <= ETH_DATA_LEN) {
4070 			fc->high_water = 0x05C20;
4071 			fc->low_water = 0x05048;
4072 			break;
4073 		}
4074 
4075 		pba = 14;
4076 		ew32(PBA, pba);
4077 		fc->high_water = ((pba << 10) * 9 / 10) & E1000_FCRTH_RTH;
4078 		fc->low_water = ((pba << 10) * 8 / 10) & E1000_FCRTL_RTL;
4079 		break;
4080 	}
4081 
4082 	/* Alignment of Tx data is on an arbitrary byte boundary with the
4083 	 * maximum size per Tx descriptor limited only to the transmit
4084 	 * allocation of the packet buffer minus 96 bytes with an upper
4085 	 * limit of 24KB due to receive synchronization limitations.
4086 	 */
4087 	adapter->tx_fifo_limit = min_t(u32, ((er32(PBA) >> 16) << 10) - 96,
4088 				       24 << 10);
4089 
4090 	/* Disable Adaptive Interrupt Moderation if 2 full packets cannot
4091 	 * fit in receive buffer.
4092 	 */
4093 	if (adapter->itr_setting & 0x3) {
4094 		if ((adapter->max_frame_size * 2) > (pba << 10)) {
4095 			if (!(adapter->flags2 & FLAG2_DISABLE_AIM)) {
4096 				dev_info(&adapter->pdev->dev,
4097 					 "Interrupt Throttle Rate off\n");
4098 				adapter->flags2 |= FLAG2_DISABLE_AIM;
4099 				e1000e_write_itr(adapter, 0);
4100 			}
4101 		} else if (adapter->flags2 & FLAG2_DISABLE_AIM) {
4102 			dev_info(&adapter->pdev->dev,
4103 				 "Interrupt Throttle Rate on\n");
4104 			adapter->flags2 &= ~FLAG2_DISABLE_AIM;
4105 			adapter->itr = 20000;
4106 			e1000e_write_itr(adapter, adapter->itr);
4107 		}
4108 	}
4109 
4110 	if (hw->mac.type >= e1000_pch_spt)
4111 		e1000_flush_desc_rings(adapter);
4112 	/* Allow time for pending master requests to run */
4113 	mac->ops.reset_hw(hw);
4114 
4115 	/* For parts with AMT enabled, let the firmware know
4116 	 * that the network interface is in control
4117 	 */
4118 	if (adapter->flags & FLAG_HAS_AMT)
4119 		e1000e_get_hw_control(adapter);
4120 
4121 	ew32(WUC, 0);
4122 
4123 	if (mac->ops.init_hw(hw))
4124 		e_err("Hardware Error\n");
4125 
4126 	e1000_update_mng_vlan(adapter);
4127 
4128 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
4129 	ew32(VET, ETH_P_8021Q);
4130 
4131 	e1000e_reset_adaptive(hw);
4132 
4133 	/* restore systim and hwtstamp settings */
4134 	e1000e_systim_reset(adapter);
4135 
4136 	/* Set EEE advertisement as appropriate */
4137 	if (adapter->flags2 & FLAG2_HAS_EEE) {
4138 		s32 ret_val;
4139 		u16 adv_addr;
4140 
4141 		switch (hw->phy.type) {
4142 		case e1000_phy_82579:
4143 			adv_addr = I82579_EEE_ADVERTISEMENT;
4144 			break;
4145 		case e1000_phy_i217:
4146 			adv_addr = I217_EEE_ADVERTISEMENT;
4147 			break;
4148 		default:
4149 			dev_err(&adapter->pdev->dev,
4150 				"Invalid PHY type setting EEE advertisement\n");
4151 			return;
4152 		}
4153 
4154 		ret_val = hw->phy.ops.acquire(hw);
4155 		if (ret_val) {
4156 			dev_err(&adapter->pdev->dev,
4157 				"EEE advertisement - unable to acquire PHY\n");
4158 			return;
4159 		}
4160 
4161 		e1000_write_emi_reg_locked(hw, adv_addr,
4162 					   hw->dev_spec.ich8lan.eee_disable ?
4163 					   0 : adapter->eee_advert);
4164 
4165 		hw->phy.ops.release(hw);
4166 	}
4167 
4168 	if (!netif_running(adapter->netdev) &&
4169 	    !test_bit(__E1000_TESTING, &adapter->state))
4170 		e1000_power_down_phy(adapter);
4171 
4172 	e1000_get_phy_info(hw);
4173 
4174 	if ((adapter->flags & FLAG_HAS_SMART_POWER_DOWN) &&
4175 	    !(adapter->flags & FLAG_SMART_POWER_DOWN)) {
4176 		u16 phy_data = 0;
4177 		/* speed up time to link by disabling smart power down, ignore
4178 		 * the return value of this function because there is nothing
4179 		 * different we would do if it failed
4180 		 */
4181 		e1e_rphy(hw, IGP02E1000_PHY_POWER_MGMT, &phy_data);
4182 		phy_data &= ~IGP02E1000_PM_SPD;
4183 		e1e_wphy(hw, IGP02E1000_PHY_POWER_MGMT, phy_data);
4184 	}
4185 	if (hw->mac.type >= e1000_pch_spt && adapter->int_mode == 0) {
4186 		u32 reg;
4187 
4188 		/* Fextnvm7 @ 0xe4[2] = 1 */
4189 		reg = er32(FEXTNVM7);
4190 		reg |= E1000_FEXTNVM7_SIDE_CLK_UNGATE;
4191 		ew32(FEXTNVM7, reg);
4192 		/* Fextnvm9 @ 0x5bb4[13:12] = 11 */
4193 		reg = er32(FEXTNVM9);
4194 		reg |= E1000_FEXTNVM9_IOSFSB_CLKGATE_DIS |
4195 		       E1000_FEXTNVM9_IOSFSB_CLKREQ_DIS;
4196 		ew32(FEXTNVM9, reg);
4197 	}
4198 
4199 }
4200 
4201 /**
4202  * e1000e_trigger_lsc - trigger an LSC interrupt
4203  * @adapter: board private structure
4204  *
4205  * Fire a link status change interrupt to start the watchdog.
4206  **/
4207 static void e1000e_trigger_lsc(struct e1000_adapter *adapter)
4208 {
4209 	struct e1000_hw *hw = &adapter->hw;
4210 
4211 	if (adapter->msix_entries)
4212 		ew32(ICS, E1000_ICS_LSC | E1000_ICS_OTHER);
4213 	else
4214 		ew32(ICS, E1000_ICS_LSC);
4215 }
4216 
4217 void e1000e_up(struct e1000_adapter *adapter)
4218 {
4219 	/* hardware has been reset, we need to reload some things */
4220 	e1000_configure(adapter);
4221 
4222 	clear_bit(__E1000_DOWN, &adapter->state);
4223 
4224 	if (adapter->msix_entries)
4225 		e1000_configure_msix(adapter);
4226 	e1000_irq_enable(adapter);
4227 
4228 	/* Tx queue started by watchdog timer when link is up */
4229 
4230 	e1000e_trigger_lsc(adapter);
4231 }
4232 
4233 static void e1000e_flush_descriptors(struct e1000_adapter *adapter)
4234 {
4235 	struct e1000_hw *hw = &adapter->hw;
4236 
4237 	if (!(adapter->flags2 & FLAG2_DMA_BURST))
4238 		return;
4239 
4240 	/* flush pending descriptor writebacks to memory */
4241 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4242 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4243 
4244 	/* execute the writes immediately */
4245 	e1e_flush();
4246 
4247 	/* due to rare timing issues, write to TIDV/RDTR again to ensure the
4248 	 * write is successful
4249 	 */
4250 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4251 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4252 
4253 	/* execute the writes immediately */
4254 	e1e_flush();
4255 }
4256 
4257 static void e1000e_update_stats(struct e1000_adapter *adapter);
4258 
4259 /**
4260  * e1000e_down - quiesce the device and optionally reset the hardware
4261  * @adapter: board private structure
4262  * @reset: boolean flag to reset the hardware or not
4263  */
4264 void e1000e_down(struct e1000_adapter *adapter, bool reset)
4265 {
4266 	struct net_device *netdev = adapter->netdev;
4267 	struct e1000_hw *hw = &adapter->hw;
4268 	u32 tctl, rctl;
4269 
4270 	/* signal that we're down so the interrupt handler does not
4271 	 * reschedule our watchdog timer
4272 	 */
4273 	set_bit(__E1000_DOWN, &adapter->state);
4274 
4275 	netif_carrier_off(netdev);
4276 
4277 	/* disable receives in the hardware */
4278 	rctl = er32(RCTL);
4279 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
4280 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
4281 	/* flush and sleep below */
4282 
4283 	netif_stop_queue(netdev);
4284 
4285 	/* disable transmits in the hardware */
4286 	tctl = er32(TCTL);
4287 	tctl &= ~E1000_TCTL_EN;
4288 	ew32(TCTL, tctl);
4289 
4290 	/* flush both disables and wait for them to finish */
4291 	e1e_flush();
4292 	usleep_range(10000, 11000);
4293 
4294 	e1000_irq_disable(adapter);
4295 
4296 	napi_synchronize(&adapter->napi);
4297 
4298 	del_timer_sync(&adapter->watchdog_timer);
4299 	del_timer_sync(&adapter->phy_info_timer);
4300 
4301 	spin_lock(&adapter->stats64_lock);
4302 	e1000e_update_stats(adapter);
4303 	spin_unlock(&adapter->stats64_lock);
4304 
4305 	e1000e_flush_descriptors(adapter);
4306 
4307 	adapter->link_speed = 0;
4308 	adapter->link_duplex = 0;
4309 
4310 	/* Disable Si errata workaround on PCHx for jumbo frame flow */
4311 	if ((hw->mac.type >= e1000_pch2lan) &&
4312 	    (adapter->netdev->mtu > ETH_DATA_LEN) &&
4313 	    e1000_lv_jumbo_workaround_ich8lan(hw, false))
4314 		e_dbg("failed to disable jumbo frame workaround mode\n");
4315 
4316 	if (!pci_channel_offline(adapter->pdev)) {
4317 		if (reset)
4318 			e1000e_reset(adapter);
4319 		else if (hw->mac.type >= e1000_pch_spt)
4320 			e1000_flush_desc_rings(adapter);
4321 	}
4322 	e1000_clean_tx_ring(adapter->tx_ring);
4323 	e1000_clean_rx_ring(adapter->rx_ring);
4324 }
4325 
4326 void e1000e_reinit_locked(struct e1000_adapter *adapter)
4327 {
4328 	might_sleep();
4329 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
4330 		usleep_range(1000, 1100);
4331 	e1000e_down(adapter, true);
4332 	e1000e_up(adapter);
4333 	clear_bit(__E1000_RESETTING, &adapter->state);
4334 }
4335 
4336 /**
4337  * e1000e_sanitize_systim - sanitize raw cycle counter reads
4338  * @hw: pointer to the HW structure
4339  * @systim: PHC time value read, sanitized and returned
4340  * @sts: structure to hold system time before and after reading SYSTIML,
4341  * may be NULL
4342  *
4343  * Errata for 82574/82583 possible bad bits read from SYSTIMH/L:
4344  * check to see that the time is incrementing at a reasonable
4345  * rate and is a multiple of incvalue.
4346  **/
4347 static u64 e1000e_sanitize_systim(struct e1000_hw *hw, u64 systim,
4348 				  struct ptp_system_timestamp *sts)
4349 {
4350 	u64 time_delta, rem, temp;
4351 	u64 systim_next;
4352 	u32 incvalue;
4353 	int i;
4354 
4355 	incvalue = er32(TIMINCA) & E1000_TIMINCA_INCVALUE_MASK;
4356 	for (i = 0; i < E1000_MAX_82574_SYSTIM_REREADS; i++) {
4357 		/* latch SYSTIMH on read of SYSTIML */
4358 		ptp_read_system_prets(sts);
4359 		systim_next = (u64)er32(SYSTIML);
4360 		ptp_read_system_postts(sts);
4361 		systim_next |= (u64)er32(SYSTIMH) << 32;
4362 
4363 		time_delta = systim_next - systim;
4364 		temp = time_delta;
4365 		/* VMWare users have seen incvalue of zero, don't div / 0 */
4366 		rem = incvalue ? do_div(temp, incvalue) : (time_delta != 0);
4367 
4368 		systim = systim_next;
4369 
4370 		if ((time_delta < E1000_82574_SYSTIM_EPSILON) && (rem == 0))
4371 			break;
4372 	}
4373 
4374 	return systim;
4375 }
4376 
4377 /**
4378  * e1000e_read_systim - read SYSTIM register
4379  * @adapter: board private structure
4380  * @sts: structure which will contain system time before and after reading
4381  * SYSTIML, may be NULL
4382  **/
4383 u64 e1000e_read_systim(struct e1000_adapter *adapter,
4384 		       struct ptp_system_timestamp *sts)
4385 {
4386 	struct e1000_hw *hw = &adapter->hw;
4387 	u32 systimel, systimel_2, systimeh;
4388 	u64 systim;
4389 	/* SYSTIMH latching upon SYSTIML read does not work well.
4390 	 * This means that if SYSTIML overflows after we read it but before
4391 	 * we read SYSTIMH, the value of SYSTIMH has been incremented and we
4392 	 * will experience a huge non linear increment in the systime value
4393 	 * to fix that we test for overflow and if true, we re-read systime.
4394 	 */
4395 	ptp_read_system_prets(sts);
4396 	systimel = er32(SYSTIML);
4397 	ptp_read_system_postts(sts);
4398 	systimeh = er32(SYSTIMH);
4399 	/* Is systimel is so large that overflow is possible? */
4400 	if (systimel >= (u32)0xffffffff - E1000_TIMINCA_INCVALUE_MASK) {
4401 		ptp_read_system_prets(sts);
4402 		systimel_2 = er32(SYSTIML);
4403 		ptp_read_system_postts(sts);
4404 		if (systimel > systimel_2) {
4405 			/* There was an overflow, read again SYSTIMH, and use
4406 			 * systimel_2
4407 			 */
4408 			systimeh = er32(SYSTIMH);
4409 			systimel = systimel_2;
4410 		}
4411 	}
4412 	systim = (u64)systimel;
4413 	systim |= (u64)systimeh << 32;
4414 
4415 	if (adapter->flags2 & FLAG2_CHECK_SYSTIM_OVERFLOW)
4416 		systim = e1000e_sanitize_systim(hw, systim, sts);
4417 
4418 	return systim;
4419 }
4420 
4421 /**
4422  * e1000e_cyclecounter_read - read raw cycle counter (used by time counter)
4423  * @cc: cyclecounter structure
4424  **/
4425 static u64 e1000e_cyclecounter_read(const struct cyclecounter *cc)
4426 {
4427 	struct e1000_adapter *adapter = container_of(cc, struct e1000_adapter,
4428 						     cc);
4429 
4430 	return e1000e_read_systim(adapter, NULL);
4431 }
4432 
4433 /**
4434  * e1000_sw_init - Initialize general software structures (struct e1000_adapter)
4435  * @adapter: board private structure to initialize
4436  *
4437  * e1000_sw_init initializes the Adapter private data structure.
4438  * Fields are initialized based on PCI device information and
4439  * OS network device settings (MTU size).
4440  **/
4441 static int e1000_sw_init(struct e1000_adapter *adapter)
4442 {
4443 	struct net_device *netdev = adapter->netdev;
4444 
4445 	adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
4446 	adapter->rx_ps_bsize0 = 128;
4447 	adapter->max_frame_size = netdev->mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
4448 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4449 	adapter->tx_ring_count = E1000_DEFAULT_TXD;
4450 	adapter->rx_ring_count = E1000_DEFAULT_RXD;
4451 
4452 	spin_lock_init(&adapter->stats64_lock);
4453 
4454 	e1000e_set_interrupt_capability(adapter);
4455 
4456 	if (e1000_alloc_queues(adapter))
4457 		return -ENOMEM;
4458 
4459 	/* Setup hardware time stamping cyclecounter */
4460 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
4461 		adapter->cc.read = e1000e_cyclecounter_read;
4462 		adapter->cc.mask = CYCLECOUNTER_MASK(64);
4463 		adapter->cc.mult = 1;
4464 		/* cc.shift set in e1000e_get_base_tininca() */
4465 
4466 		spin_lock_init(&adapter->systim_lock);
4467 		INIT_WORK(&adapter->tx_hwtstamp_work, e1000e_tx_hwtstamp_work);
4468 	}
4469 
4470 	/* Explicitly disable IRQ since the NIC can be in any state. */
4471 	e1000_irq_disable(adapter);
4472 
4473 	set_bit(__E1000_DOWN, &adapter->state);
4474 	return 0;
4475 }
4476 
4477 /**
4478  * e1000_intr_msi_test - Interrupt Handler
4479  * @irq: interrupt number
4480  * @data: pointer to a network interface device structure
4481  **/
4482 static irqreturn_t e1000_intr_msi_test(int __always_unused irq, void *data)
4483 {
4484 	struct net_device *netdev = data;
4485 	struct e1000_adapter *adapter = netdev_priv(netdev);
4486 	struct e1000_hw *hw = &adapter->hw;
4487 	u32 icr = er32(ICR);
4488 
4489 	e_dbg("icr is %08X\n", icr);
4490 	if (icr & E1000_ICR_RXSEQ) {
4491 		adapter->flags &= ~FLAG_MSI_TEST_FAILED;
4492 		/* Force memory writes to complete before acknowledging the
4493 		 * interrupt is handled.
4494 		 */
4495 		wmb();
4496 	}
4497 
4498 	return IRQ_HANDLED;
4499 }
4500 
4501 /**
4502  * e1000_test_msi_interrupt - Returns 0 for successful test
4503  * @adapter: board private struct
4504  *
4505  * code flow taken from tg3.c
4506  **/
4507 static int e1000_test_msi_interrupt(struct e1000_adapter *adapter)
4508 {
4509 	struct net_device *netdev = adapter->netdev;
4510 	struct e1000_hw *hw = &adapter->hw;
4511 	int err;
4512 
4513 	/* poll_enable hasn't been called yet, so don't need disable */
4514 	/* clear any pending events */
4515 	er32(ICR);
4516 
4517 	/* free the real vector and request a test handler */
4518 	e1000_free_irq(adapter);
4519 	e1000e_reset_interrupt_capability(adapter);
4520 
4521 	/* Assume that the test fails, if it succeeds then the test
4522 	 * MSI irq handler will unset this flag
4523 	 */
4524 	adapter->flags |= FLAG_MSI_TEST_FAILED;
4525 
4526 	err = pci_enable_msi(adapter->pdev);
4527 	if (err)
4528 		goto msi_test_failed;
4529 
4530 	err = request_irq(adapter->pdev->irq, e1000_intr_msi_test, 0,
4531 			  netdev->name, netdev);
4532 	if (err) {
4533 		pci_disable_msi(adapter->pdev);
4534 		goto msi_test_failed;
4535 	}
4536 
4537 	/* Force memory writes to complete before enabling and firing an
4538 	 * interrupt.
4539 	 */
4540 	wmb();
4541 
4542 	e1000_irq_enable(adapter);
4543 
4544 	/* fire an unusual interrupt on the test handler */
4545 	ew32(ICS, E1000_ICS_RXSEQ);
4546 	e1e_flush();
4547 	msleep(100);
4548 
4549 	e1000_irq_disable(adapter);
4550 
4551 	rmb();			/* read flags after interrupt has been fired */
4552 
4553 	if (adapter->flags & FLAG_MSI_TEST_FAILED) {
4554 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
4555 		e_info("MSI interrupt test failed, using legacy interrupt.\n");
4556 	} else {
4557 		e_dbg("MSI interrupt test succeeded!\n");
4558 	}
4559 
4560 	free_irq(adapter->pdev->irq, netdev);
4561 	pci_disable_msi(adapter->pdev);
4562 
4563 msi_test_failed:
4564 	e1000e_set_interrupt_capability(adapter);
4565 	return e1000_request_irq(adapter);
4566 }
4567 
4568 /**
4569  * e1000_test_msi - Returns 0 if MSI test succeeds or INTx mode is restored
4570  * @adapter: board private struct
4571  *
4572  * code flow taken from tg3.c, called with e1000 interrupts disabled.
4573  **/
4574 static int e1000_test_msi(struct e1000_adapter *adapter)
4575 {
4576 	int err;
4577 	u16 pci_cmd;
4578 
4579 	if (!(adapter->flags & FLAG_MSI_ENABLED))
4580 		return 0;
4581 
4582 	/* disable SERR in case the MSI write causes a master abort */
4583 	pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4584 	if (pci_cmd & PCI_COMMAND_SERR)
4585 		pci_write_config_word(adapter->pdev, PCI_COMMAND,
4586 				      pci_cmd & ~PCI_COMMAND_SERR);
4587 
4588 	err = e1000_test_msi_interrupt(adapter);
4589 
4590 	/* re-enable SERR */
4591 	if (pci_cmd & PCI_COMMAND_SERR) {
4592 		pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4593 		pci_cmd |= PCI_COMMAND_SERR;
4594 		pci_write_config_word(adapter->pdev, PCI_COMMAND, pci_cmd);
4595 	}
4596 
4597 	return err;
4598 }
4599 
4600 /**
4601  * e1000e_open - Called when a network interface is made active
4602  * @netdev: network interface device structure
4603  *
4604  * Returns 0 on success, negative value on failure
4605  *
4606  * The open entry point is called when a network interface is made
4607  * active by the system (IFF_UP).  At this point all resources needed
4608  * for transmit and receive operations are allocated, the interrupt
4609  * handler is registered with the OS, the watchdog timer is started,
4610  * and the stack is notified that the interface is ready.
4611  **/
4612 int e1000e_open(struct net_device *netdev)
4613 {
4614 	struct e1000_adapter *adapter = netdev_priv(netdev);
4615 	struct e1000_hw *hw = &adapter->hw;
4616 	struct pci_dev *pdev = adapter->pdev;
4617 	int err;
4618 
4619 	/* disallow open during test */
4620 	if (test_bit(__E1000_TESTING, &adapter->state))
4621 		return -EBUSY;
4622 
4623 	pm_runtime_get_sync(&pdev->dev);
4624 
4625 	netif_carrier_off(netdev);
4626 	netif_stop_queue(netdev);
4627 
4628 	/* allocate transmit descriptors */
4629 	err = e1000e_setup_tx_resources(adapter->tx_ring);
4630 	if (err)
4631 		goto err_setup_tx;
4632 
4633 	/* allocate receive descriptors */
4634 	err = e1000e_setup_rx_resources(adapter->rx_ring);
4635 	if (err)
4636 		goto err_setup_rx;
4637 
4638 	/* If AMT is enabled, let the firmware know that the network
4639 	 * interface is now open and reset the part to a known state.
4640 	 */
4641 	if (adapter->flags & FLAG_HAS_AMT) {
4642 		e1000e_get_hw_control(adapter);
4643 		e1000e_reset(adapter);
4644 	}
4645 
4646 	e1000e_power_up_phy(adapter);
4647 
4648 	adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
4649 	if ((adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
4650 		e1000_update_mng_vlan(adapter);
4651 
4652 	/* DMA latency requirement to workaround jumbo issue */
4653 	cpu_latency_qos_add_request(&adapter->pm_qos_req, PM_QOS_DEFAULT_VALUE);
4654 
4655 	/* before we allocate an interrupt, we must be ready to handle it.
4656 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4657 	 * as soon as we call pci_request_irq, so we have to setup our
4658 	 * clean_rx handler before we do so.
4659 	 */
4660 	e1000_configure(adapter);
4661 
4662 	err = e1000_request_irq(adapter);
4663 	if (err)
4664 		goto err_req_irq;
4665 
4666 	/* Work around PCIe errata with MSI interrupts causing some chipsets to
4667 	 * ignore e1000e MSI messages, which means we need to test our MSI
4668 	 * interrupt now
4669 	 */
4670 	if (adapter->int_mode != E1000E_INT_MODE_LEGACY) {
4671 		err = e1000_test_msi(adapter);
4672 		if (err) {
4673 			e_err("Interrupt allocation failed\n");
4674 			goto err_req_irq;
4675 		}
4676 	}
4677 
4678 	/* From here on the code is the same as e1000e_up() */
4679 	clear_bit(__E1000_DOWN, &adapter->state);
4680 
4681 	napi_enable(&adapter->napi);
4682 
4683 	e1000_irq_enable(adapter);
4684 
4685 	adapter->tx_hang_recheck = false;
4686 
4687 	hw->mac.get_link_status = true;
4688 	pm_runtime_put(&pdev->dev);
4689 
4690 	e1000e_trigger_lsc(adapter);
4691 
4692 	return 0;
4693 
4694 err_req_irq:
4695 	cpu_latency_qos_remove_request(&adapter->pm_qos_req);
4696 	e1000e_release_hw_control(adapter);
4697 	e1000_power_down_phy(adapter);
4698 	e1000e_free_rx_resources(adapter->rx_ring);
4699 err_setup_rx:
4700 	e1000e_free_tx_resources(adapter->tx_ring);
4701 err_setup_tx:
4702 	e1000e_reset(adapter);
4703 	pm_runtime_put_sync(&pdev->dev);
4704 
4705 	return err;
4706 }
4707 
4708 /**
4709  * e1000e_close - Disables a network interface
4710  * @netdev: network interface device structure
4711  *
4712  * Returns 0, this is not allowed to fail
4713  *
4714  * The close entry point is called when an interface is de-activated
4715  * by the OS.  The hardware is still under the drivers control, but
4716  * needs to be disabled.  A global MAC reset is issued to stop the
4717  * hardware, and all transmit and receive resources are freed.
4718  **/
4719 int e1000e_close(struct net_device *netdev)
4720 {
4721 	struct e1000_adapter *adapter = netdev_priv(netdev);
4722 	struct pci_dev *pdev = adapter->pdev;
4723 	int count = E1000_CHECK_RESET_COUNT;
4724 
4725 	while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
4726 		usleep_range(10000, 11000);
4727 
4728 	WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
4729 
4730 	pm_runtime_get_sync(&pdev->dev);
4731 
4732 	if (netif_device_present(netdev)) {
4733 		e1000e_down(adapter, true);
4734 		e1000_free_irq(adapter);
4735 
4736 		/* Link status message must follow this format */
4737 		netdev_info(netdev, "NIC Link is Down\n");
4738 	}
4739 
4740 	napi_disable(&adapter->napi);
4741 
4742 	e1000e_free_tx_resources(adapter->tx_ring);
4743 	e1000e_free_rx_resources(adapter->rx_ring);
4744 
4745 	/* kill manageability vlan ID if supported, but not if a vlan with
4746 	 * the same ID is registered on the host OS (let 8021q kill it)
4747 	 */
4748 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN)
4749 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
4750 				       adapter->mng_vlan_id);
4751 
4752 	/* If AMT is enabled, let the firmware know that the network
4753 	 * interface is now closed
4754 	 */
4755 	if ((adapter->flags & FLAG_HAS_AMT) &&
4756 	    !test_bit(__E1000_TESTING, &adapter->state))
4757 		e1000e_release_hw_control(adapter);
4758 
4759 	cpu_latency_qos_remove_request(&adapter->pm_qos_req);
4760 
4761 	pm_runtime_put_sync(&pdev->dev);
4762 
4763 	return 0;
4764 }
4765 
4766 /**
4767  * e1000_set_mac - Change the Ethernet Address of the NIC
4768  * @netdev: network interface device structure
4769  * @p: pointer to an address structure
4770  *
4771  * Returns 0 on success, negative on failure
4772  **/
4773 static int e1000_set_mac(struct net_device *netdev, void *p)
4774 {
4775 	struct e1000_adapter *adapter = netdev_priv(netdev);
4776 	struct e1000_hw *hw = &adapter->hw;
4777 	struct sockaddr *addr = p;
4778 
4779 	if (!is_valid_ether_addr(addr->sa_data))
4780 		return -EADDRNOTAVAIL;
4781 
4782 	eth_hw_addr_set(netdev, addr->sa_data);
4783 	memcpy(adapter->hw.mac.addr, addr->sa_data, netdev->addr_len);
4784 
4785 	hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
4786 
4787 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA) {
4788 		/* activate the work around */
4789 		e1000e_set_laa_state_82571(&adapter->hw, 1);
4790 
4791 		/* Hold a copy of the LAA in RAR[14] This is done so that
4792 		 * between the time RAR[0] gets clobbered  and the time it
4793 		 * gets fixed (in e1000_watchdog), the actual LAA is in one
4794 		 * of the RARs and no incoming packets directed to this port
4795 		 * are dropped. Eventually the LAA will be in RAR[0] and
4796 		 * RAR[14]
4797 		 */
4798 		hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr,
4799 				    adapter->hw.mac.rar_entry_count - 1);
4800 	}
4801 
4802 	return 0;
4803 }
4804 
4805 /**
4806  * e1000e_update_phy_task - work thread to update phy
4807  * @work: pointer to our work struct
4808  *
4809  * this worker thread exists because we must acquire a
4810  * semaphore to read the phy, which we could msleep while
4811  * waiting for it, and we can't msleep in a timer.
4812  **/
4813 static void e1000e_update_phy_task(struct work_struct *work)
4814 {
4815 	struct e1000_adapter *adapter = container_of(work,
4816 						     struct e1000_adapter,
4817 						     update_phy_task);
4818 	struct e1000_hw *hw = &adapter->hw;
4819 
4820 	if (test_bit(__E1000_DOWN, &adapter->state))
4821 		return;
4822 
4823 	e1000_get_phy_info(hw);
4824 
4825 	/* Enable EEE on 82579 after link up */
4826 	if (hw->phy.type >= e1000_phy_82579)
4827 		e1000_set_eee_pchlan(hw);
4828 }
4829 
4830 /**
4831  * e1000_update_phy_info - timre call-back to update PHY info
4832  * @t: pointer to timer_list containing private info adapter
4833  *
4834  * Need to wait a few seconds after link up to get diagnostic information from
4835  * the phy
4836  **/
4837 static void e1000_update_phy_info(struct timer_list *t)
4838 {
4839 	struct e1000_adapter *adapter = from_timer(adapter, t, phy_info_timer);
4840 
4841 	if (test_bit(__E1000_DOWN, &adapter->state))
4842 		return;
4843 
4844 	schedule_work(&adapter->update_phy_task);
4845 }
4846 
4847 /**
4848  * e1000e_update_phy_stats - Update the PHY statistics counters
4849  * @adapter: board private structure
4850  *
4851  * Read/clear the upper 16-bit PHY registers and read/accumulate lower
4852  **/
4853 static void e1000e_update_phy_stats(struct e1000_adapter *adapter)
4854 {
4855 	struct e1000_hw *hw = &adapter->hw;
4856 	s32 ret_val;
4857 	u16 phy_data;
4858 
4859 	ret_val = hw->phy.ops.acquire(hw);
4860 	if (ret_val)
4861 		return;
4862 
4863 	/* A page set is expensive so check if already on desired page.
4864 	 * If not, set to the page with the PHY status registers.
4865 	 */
4866 	hw->phy.addr = 1;
4867 	ret_val = e1000e_read_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4868 					   &phy_data);
4869 	if (ret_val)
4870 		goto release;
4871 	if (phy_data != (HV_STATS_PAGE << IGP_PAGE_SHIFT)) {
4872 		ret_val = hw->phy.ops.set_page(hw,
4873 					       HV_STATS_PAGE << IGP_PAGE_SHIFT);
4874 		if (ret_val)
4875 			goto release;
4876 	}
4877 
4878 	/* Single Collision Count */
4879 	hw->phy.ops.read_reg_page(hw, HV_SCC_UPPER, &phy_data);
4880 	ret_val = hw->phy.ops.read_reg_page(hw, HV_SCC_LOWER, &phy_data);
4881 	if (!ret_val)
4882 		adapter->stats.scc += phy_data;
4883 
4884 	/* Excessive Collision Count */
4885 	hw->phy.ops.read_reg_page(hw, HV_ECOL_UPPER, &phy_data);
4886 	ret_val = hw->phy.ops.read_reg_page(hw, HV_ECOL_LOWER, &phy_data);
4887 	if (!ret_val)
4888 		adapter->stats.ecol += phy_data;
4889 
4890 	/* Multiple Collision Count */
4891 	hw->phy.ops.read_reg_page(hw, HV_MCC_UPPER, &phy_data);
4892 	ret_val = hw->phy.ops.read_reg_page(hw, HV_MCC_LOWER, &phy_data);
4893 	if (!ret_val)
4894 		adapter->stats.mcc += phy_data;
4895 
4896 	/* Late Collision Count */
4897 	hw->phy.ops.read_reg_page(hw, HV_LATECOL_UPPER, &phy_data);
4898 	ret_val = hw->phy.ops.read_reg_page(hw, HV_LATECOL_LOWER, &phy_data);
4899 	if (!ret_val)
4900 		adapter->stats.latecol += phy_data;
4901 
4902 	/* Collision Count - also used for adaptive IFS */
4903 	hw->phy.ops.read_reg_page(hw, HV_COLC_UPPER, &phy_data);
4904 	ret_val = hw->phy.ops.read_reg_page(hw, HV_COLC_LOWER, &phy_data);
4905 	if (!ret_val)
4906 		hw->mac.collision_delta = phy_data;
4907 
4908 	/* Defer Count */
4909 	hw->phy.ops.read_reg_page(hw, HV_DC_UPPER, &phy_data);
4910 	ret_val = hw->phy.ops.read_reg_page(hw, HV_DC_LOWER, &phy_data);
4911 	if (!ret_val)
4912 		adapter->stats.dc += phy_data;
4913 
4914 	/* Transmit with no CRS */
4915 	hw->phy.ops.read_reg_page(hw, HV_TNCRS_UPPER, &phy_data);
4916 	ret_val = hw->phy.ops.read_reg_page(hw, HV_TNCRS_LOWER, &phy_data);
4917 	if (!ret_val)
4918 		adapter->stats.tncrs += phy_data;
4919 
4920 release:
4921 	hw->phy.ops.release(hw);
4922 }
4923 
4924 /**
4925  * e1000e_update_stats - Update the board statistics counters
4926  * @adapter: board private structure
4927  **/
4928 static void e1000e_update_stats(struct e1000_adapter *adapter)
4929 {
4930 	struct net_device *netdev = adapter->netdev;
4931 	struct e1000_hw *hw = &adapter->hw;
4932 	struct pci_dev *pdev = adapter->pdev;
4933 
4934 	/* Prevent stats update while adapter is being reset, or if the pci
4935 	 * connection is down.
4936 	 */
4937 	if (adapter->link_speed == 0)
4938 		return;
4939 	if (pci_channel_offline(pdev))
4940 		return;
4941 
4942 	adapter->stats.crcerrs += er32(CRCERRS);
4943 	adapter->stats.gprc += er32(GPRC);
4944 	adapter->stats.gorc += er32(GORCL);
4945 	er32(GORCH);		/* Clear gorc */
4946 	adapter->stats.bprc += er32(BPRC);
4947 	adapter->stats.mprc += er32(MPRC);
4948 	adapter->stats.roc += er32(ROC);
4949 
4950 	adapter->stats.mpc += er32(MPC);
4951 
4952 	/* Half-duplex statistics */
4953 	if (adapter->link_duplex == HALF_DUPLEX) {
4954 		if (adapter->flags2 & FLAG2_HAS_PHY_STATS) {
4955 			e1000e_update_phy_stats(adapter);
4956 		} else {
4957 			adapter->stats.scc += er32(SCC);
4958 			adapter->stats.ecol += er32(ECOL);
4959 			adapter->stats.mcc += er32(MCC);
4960 			adapter->stats.latecol += er32(LATECOL);
4961 			adapter->stats.dc += er32(DC);
4962 
4963 			hw->mac.collision_delta = er32(COLC);
4964 
4965 			if ((hw->mac.type != e1000_82574) &&
4966 			    (hw->mac.type != e1000_82583))
4967 				adapter->stats.tncrs += er32(TNCRS);
4968 		}
4969 		adapter->stats.colc += hw->mac.collision_delta;
4970 	}
4971 
4972 	adapter->stats.xonrxc += er32(XONRXC);
4973 	adapter->stats.xontxc += er32(XONTXC);
4974 	adapter->stats.xoffrxc += er32(XOFFRXC);
4975 	adapter->stats.xofftxc += er32(XOFFTXC);
4976 	adapter->stats.gptc += er32(GPTC);
4977 	adapter->stats.gotc += er32(GOTCL);
4978 	er32(GOTCH);		/* Clear gotc */
4979 	adapter->stats.rnbc += er32(RNBC);
4980 	adapter->stats.ruc += er32(RUC);
4981 
4982 	adapter->stats.mptc += er32(MPTC);
4983 	adapter->stats.bptc += er32(BPTC);
4984 
4985 	/* used for adaptive IFS */
4986 
4987 	hw->mac.tx_packet_delta = er32(TPT);
4988 	adapter->stats.tpt += hw->mac.tx_packet_delta;
4989 
4990 	adapter->stats.algnerrc += er32(ALGNERRC);
4991 	adapter->stats.rxerrc += er32(RXERRC);
4992 	adapter->stats.cexterr += er32(CEXTERR);
4993 	adapter->stats.tsctc += er32(TSCTC);
4994 	adapter->stats.tsctfc += er32(TSCTFC);
4995 
4996 	/* Fill out the OS statistics structure */
4997 	netdev->stats.multicast = adapter->stats.mprc;
4998 	netdev->stats.collisions = adapter->stats.colc;
4999 
5000 	/* Rx Errors */
5001 
5002 	/* RLEC on some newer hardware can be incorrect so build
5003 	 * our own version based on RUC and ROC
5004 	 */
5005 	netdev->stats.rx_errors = adapter->stats.rxerrc +
5006 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
5007 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5008 	netdev->stats.rx_length_errors = adapter->stats.ruc +
5009 	    adapter->stats.roc;
5010 	netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
5011 	netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
5012 	netdev->stats.rx_missed_errors = adapter->stats.mpc;
5013 
5014 	/* Tx Errors */
5015 	netdev->stats.tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5016 	netdev->stats.tx_aborted_errors = adapter->stats.ecol;
5017 	netdev->stats.tx_window_errors = adapter->stats.latecol;
5018 	netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
5019 
5020 	/* Tx Dropped needs to be maintained elsewhere */
5021 
5022 	/* Management Stats */
5023 	adapter->stats.mgptc += er32(MGTPTC);
5024 	adapter->stats.mgprc += er32(MGTPRC);
5025 	adapter->stats.mgpdc += er32(MGTPDC);
5026 
5027 	/* Correctable ECC Errors */
5028 	if (hw->mac.type >= e1000_pch_lpt) {
5029 		u32 pbeccsts = er32(PBECCSTS);
5030 
5031 		adapter->corr_errors +=
5032 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
5033 		adapter->uncorr_errors +=
5034 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
5035 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
5036 	}
5037 }
5038 
5039 /**
5040  * e1000_phy_read_status - Update the PHY register status snapshot
5041  * @adapter: board private structure
5042  **/
5043 static void e1000_phy_read_status(struct e1000_adapter *adapter)
5044 {
5045 	struct e1000_hw *hw = &adapter->hw;
5046 	struct e1000_phy_regs *phy = &adapter->phy_regs;
5047 
5048 	if (!pm_runtime_suspended((&adapter->pdev->dev)->parent) &&
5049 	    (er32(STATUS) & E1000_STATUS_LU) &&
5050 	    (adapter->hw.phy.media_type == e1000_media_type_copper)) {
5051 		int ret_val;
5052 
5053 		ret_val = e1e_rphy(hw, MII_BMCR, &phy->bmcr);
5054 		ret_val |= e1e_rphy(hw, MII_BMSR, &phy->bmsr);
5055 		ret_val |= e1e_rphy(hw, MII_ADVERTISE, &phy->advertise);
5056 		ret_val |= e1e_rphy(hw, MII_LPA, &phy->lpa);
5057 		ret_val |= e1e_rphy(hw, MII_EXPANSION, &phy->expansion);
5058 		ret_val |= e1e_rphy(hw, MII_CTRL1000, &phy->ctrl1000);
5059 		ret_val |= e1e_rphy(hw, MII_STAT1000, &phy->stat1000);
5060 		ret_val |= e1e_rphy(hw, MII_ESTATUS, &phy->estatus);
5061 		if (ret_val)
5062 			e_warn("Error reading PHY register\n");
5063 	} else {
5064 		/* Do not read PHY registers if link is not up
5065 		 * Set values to typical power-on defaults
5066 		 */
5067 		phy->bmcr = (BMCR_SPEED1000 | BMCR_ANENABLE | BMCR_FULLDPLX);
5068 		phy->bmsr = (BMSR_100FULL | BMSR_100HALF | BMSR_10FULL |
5069 			     BMSR_10HALF | BMSR_ESTATEN | BMSR_ANEGCAPABLE |
5070 			     BMSR_ERCAP);
5071 		phy->advertise = (ADVERTISE_PAUSE_ASYM | ADVERTISE_PAUSE_CAP |
5072 				  ADVERTISE_ALL | ADVERTISE_CSMA);
5073 		phy->lpa = 0;
5074 		phy->expansion = EXPANSION_ENABLENPAGE;
5075 		phy->ctrl1000 = ADVERTISE_1000FULL;
5076 		phy->stat1000 = 0;
5077 		phy->estatus = (ESTATUS_1000_TFULL | ESTATUS_1000_THALF);
5078 	}
5079 }
5080 
5081 static void e1000_print_link_info(struct e1000_adapter *adapter)
5082 {
5083 	struct e1000_hw *hw = &adapter->hw;
5084 	u32 ctrl = er32(CTRL);
5085 
5086 	/* Link status message must follow this format for user tools */
5087 	netdev_info(adapter->netdev,
5088 		    "NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5089 		    adapter->link_speed,
5090 		    adapter->link_duplex == FULL_DUPLEX ? "Full" : "Half",
5091 		    (ctrl & E1000_CTRL_TFCE) && (ctrl & E1000_CTRL_RFCE) ? "Rx/Tx" :
5092 		    (ctrl & E1000_CTRL_RFCE) ? "Rx" :
5093 		    (ctrl & E1000_CTRL_TFCE) ? "Tx" : "None");
5094 }
5095 
5096 static bool e1000e_has_link(struct e1000_adapter *adapter)
5097 {
5098 	struct e1000_hw *hw = &adapter->hw;
5099 	bool link_active = false;
5100 	s32 ret_val = 0;
5101 
5102 	/* get_link_status is set on LSC (link status) interrupt or
5103 	 * Rx sequence error interrupt.  get_link_status will stay
5104 	 * true until the check_for_link establishes link
5105 	 * for copper adapters ONLY
5106 	 */
5107 	switch (hw->phy.media_type) {
5108 	case e1000_media_type_copper:
5109 		if (hw->mac.get_link_status) {
5110 			ret_val = hw->mac.ops.check_for_link(hw);
5111 			link_active = !hw->mac.get_link_status;
5112 		} else {
5113 			link_active = true;
5114 		}
5115 		break;
5116 	case e1000_media_type_fiber:
5117 		ret_val = hw->mac.ops.check_for_link(hw);
5118 		link_active = !!(er32(STATUS) & E1000_STATUS_LU);
5119 		break;
5120 	case e1000_media_type_internal_serdes:
5121 		ret_val = hw->mac.ops.check_for_link(hw);
5122 		link_active = hw->mac.serdes_has_link;
5123 		break;
5124 	default:
5125 	case e1000_media_type_unknown:
5126 		break;
5127 	}
5128 
5129 	if ((ret_val == -E1000_ERR_PHY) && (hw->phy.type == e1000_phy_igp_3) &&
5130 	    (er32(CTRL) & E1000_PHY_CTRL_GBE_DISABLE)) {
5131 		/* See e1000_kmrn_lock_loss_workaround_ich8lan() */
5132 		e_info("Gigabit has been disabled, downgrading speed\n");
5133 	}
5134 
5135 	return link_active;
5136 }
5137 
5138 static void e1000e_enable_receives(struct e1000_adapter *adapter)
5139 {
5140 	/* make sure the receive unit is started */
5141 	if ((adapter->flags & FLAG_RX_NEEDS_RESTART) &&
5142 	    (adapter->flags & FLAG_RESTART_NOW)) {
5143 		struct e1000_hw *hw = &adapter->hw;
5144 		u32 rctl = er32(RCTL);
5145 
5146 		ew32(RCTL, rctl | E1000_RCTL_EN);
5147 		adapter->flags &= ~FLAG_RESTART_NOW;
5148 	}
5149 }
5150 
5151 static void e1000e_check_82574_phy_workaround(struct e1000_adapter *adapter)
5152 {
5153 	struct e1000_hw *hw = &adapter->hw;
5154 
5155 	/* With 82574 controllers, PHY needs to be checked periodically
5156 	 * for hung state and reset, if two calls return true
5157 	 */
5158 	if (e1000_check_phy_82574(hw))
5159 		adapter->phy_hang_count++;
5160 	else
5161 		adapter->phy_hang_count = 0;
5162 
5163 	if (adapter->phy_hang_count > 1) {
5164 		adapter->phy_hang_count = 0;
5165 		e_dbg("PHY appears hung - resetting\n");
5166 		schedule_work(&adapter->reset_task);
5167 	}
5168 }
5169 
5170 /**
5171  * e1000_watchdog - Timer Call-back
5172  * @t: pointer to timer_list containing private info adapter
5173  **/
5174 static void e1000_watchdog(struct timer_list *t)
5175 {
5176 	struct e1000_adapter *adapter = from_timer(adapter, t, watchdog_timer);
5177 
5178 	/* Do the rest outside of interrupt context */
5179 	schedule_work(&adapter->watchdog_task);
5180 
5181 	/* TODO: make this use queue_delayed_work() */
5182 }
5183 
5184 static void e1000_watchdog_task(struct work_struct *work)
5185 {
5186 	struct e1000_adapter *adapter = container_of(work,
5187 						     struct e1000_adapter,
5188 						     watchdog_task);
5189 	struct net_device *netdev = adapter->netdev;
5190 	struct e1000_mac_info *mac = &adapter->hw.mac;
5191 	struct e1000_phy_info *phy = &adapter->hw.phy;
5192 	struct e1000_ring *tx_ring = adapter->tx_ring;
5193 	u32 dmoff_exit_timeout = 100, tries = 0;
5194 	struct e1000_hw *hw = &adapter->hw;
5195 	u32 link, tctl, pcim_state;
5196 
5197 	if (test_bit(__E1000_DOWN, &adapter->state))
5198 		return;
5199 
5200 	link = e1000e_has_link(adapter);
5201 	if ((netif_carrier_ok(netdev)) && link) {
5202 		/* Cancel scheduled suspend requests. */
5203 		pm_runtime_resume(netdev->dev.parent);
5204 
5205 		e1000e_enable_receives(adapter);
5206 		goto link_up;
5207 	}
5208 
5209 	if ((e1000e_enable_tx_pkt_filtering(hw)) &&
5210 	    (adapter->mng_vlan_id != adapter->hw.mng_cookie.vlan_id))
5211 		e1000_update_mng_vlan(adapter);
5212 
5213 	if (link) {
5214 		if (!netif_carrier_ok(netdev)) {
5215 			bool txb2b = true;
5216 
5217 			/* Cancel scheduled suspend requests. */
5218 			pm_runtime_resume(netdev->dev.parent);
5219 
5220 			/* Checking if MAC is in DMoff state*/
5221 			if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID) {
5222 				pcim_state = er32(STATUS);
5223 				while (pcim_state & E1000_STATUS_PCIM_STATE) {
5224 					if (tries++ == dmoff_exit_timeout) {
5225 						e_dbg("Error in exiting dmoff\n");
5226 						break;
5227 					}
5228 					usleep_range(10000, 20000);
5229 					pcim_state = er32(STATUS);
5230 
5231 					/* Checking if MAC exited DMoff state */
5232 					if (!(pcim_state & E1000_STATUS_PCIM_STATE))
5233 						e1000_phy_hw_reset(&adapter->hw);
5234 				}
5235 			}
5236 
5237 			/* update snapshot of PHY registers on LSC */
5238 			e1000_phy_read_status(adapter);
5239 			mac->ops.get_link_up_info(&adapter->hw,
5240 						  &adapter->link_speed,
5241 						  &adapter->link_duplex);
5242 			e1000_print_link_info(adapter);
5243 
5244 			/* check if SmartSpeed worked */
5245 			e1000e_check_downshift(hw);
5246 			if (phy->speed_downgraded)
5247 				netdev_warn(netdev,
5248 					    "Link Speed was downgraded by SmartSpeed\n");
5249 
5250 			/* On supported PHYs, check for duplex mismatch only
5251 			 * if link has autonegotiated at 10/100 half
5252 			 */
5253 			if ((hw->phy.type == e1000_phy_igp_3 ||
5254 			     hw->phy.type == e1000_phy_bm) &&
5255 			    hw->mac.autoneg &&
5256 			    (adapter->link_speed == SPEED_10 ||
5257 			     adapter->link_speed == SPEED_100) &&
5258 			    (adapter->link_duplex == HALF_DUPLEX)) {
5259 				u16 autoneg_exp;
5260 
5261 				e1e_rphy(hw, MII_EXPANSION, &autoneg_exp);
5262 
5263 				if (!(autoneg_exp & EXPANSION_NWAY))
5264 					e_info("Autonegotiated half duplex but link partner cannot autoneg.  Try forcing full duplex if link gets many collisions.\n");
5265 			}
5266 
5267 			/* adjust timeout factor according to speed/duplex */
5268 			adapter->tx_timeout_factor = 1;
5269 			switch (adapter->link_speed) {
5270 			case SPEED_10:
5271 				txb2b = false;
5272 				adapter->tx_timeout_factor = 16;
5273 				break;
5274 			case SPEED_100:
5275 				txb2b = false;
5276 				adapter->tx_timeout_factor = 10;
5277 				break;
5278 			}
5279 
5280 			/* workaround: re-program speed mode bit after
5281 			 * link-up event
5282 			 */
5283 			if ((adapter->flags & FLAG_TARC_SPEED_MODE_BIT) &&
5284 			    !txb2b) {
5285 				u32 tarc0;
5286 
5287 				tarc0 = er32(TARC(0));
5288 				tarc0 &= ~SPEED_MODE_BIT;
5289 				ew32(TARC(0), tarc0);
5290 			}
5291 
5292 			/* enable transmits in the hardware, need to do this
5293 			 * after setting TARC(0)
5294 			 */
5295 			tctl = er32(TCTL);
5296 			tctl |= E1000_TCTL_EN;
5297 			ew32(TCTL, tctl);
5298 
5299 			/* Perform any post-link-up configuration before
5300 			 * reporting link up.
5301 			 */
5302 			if (phy->ops.cfg_on_link_up)
5303 				phy->ops.cfg_on_link_up(hw);
5304 
5305 			netif_wake_queue(netdev);
5306 			netif_carrier_on(netdev);
5307 
5308 			if (!test_bit(__E1000_DOWN, &adapter->state))
5309 				mod_timer(&adapter->phy_info_timer,
5310 					  round_jiffies(jiffies + 2 * HZ));
5311 		}
5312 	} else {
5313 		if (netif_carrier_ok(netdev)) {
5314 			adapter->link_speed = 0;
5315 			adapter->link_duplex = 0;
5316 			/* Link status message must follow this format */
5317 			netdev_info(netdev, "NIC Link is Down\n");
5318 			netif_carrier_off(netdev);
5319 			netif_stop_queue(netdev);
5320 			if (!test_bit(__E1000_DOWN, &adapter->state))
5321 				mod_timer(&adapter->phy_info_timer,
5322 					  round_jiffies(jiffies + 2 * HZ));
5323 
5324 			/* 8000ES2LAN requires a Rx packet buffer work-around
5325 			 * on link down event; reset the controller to flush
5326 			 * the Rx packet buffer.
5327 			 */
5328 			if (adapter->flags & FLAG_RX_NEEDS_RESTART)
5329 				adapter->flags |= FLAG_RESTART_NOW;
5330 			else
5331 				pm_schedule_suspend(netdev->dev.parent,
5332 						    LINK_TIMEOUT);
5333 		}
5334 	}
5335 
5336 link_up:
5337 	spin_lock(&adapter->stats64_lock);
5338 	e1000e_update_stats(adapter);
5339 
5340 	mac->tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
5341 	adapter->tpt_old = adapter->stats.tpt;
5342 	mac->collision_delta = adapter->stats.colc - adapter->colc_old;
5343 	adapter->colc_old = adapter->stats.colc;
5344 
5345 	adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
5346 	adapter->gorc_old = adapter->stats.gorc;
5347 	adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
5348 	adapter->gotc_old = adapter->stats.gotc;
5349 	spin_unlock(&adapter->stats64_lock);
5350 
5351 	/* If the link is lost the controller stops DMA, but
5352 	 * if there is queued Tx work it cannot be done.  So
5353 	 * reset the controller to flush the Tx packet buffers.
5354 	 */
5355 	if (!netif_carrier_ok(netdev) &&
5356 	    (e1000_desc_unused(tx_ring) + 1 < tx_ring->count))
5357 		adapter->flags |= FLAG_RESTART_NOW;
5358 
5359 	/* If reset is necessary, do it outside of interrupt context. */
5360 	if (adapter->flags & FLAG_RESTART_NOW) {
5361 		schedule_work(&adapter->reset_task);
5362 		/* return immediately since reset is imminent */
5363 		return;
5364 	}
5365 
5366 	e1000e_update_adaptive(&adapter->hw);
5367 
5368 	/* Simple mode for Interrupt Throttle Rate (ITR) */
5369 	if (adapter->itr_setting == 4) {
5370 		/* Symmetric Tx/Rx gets a reduced ITR=2000;
5371 		 * Total asymmetrical Tx or Rx gets ITR=8000;
5372 		 * everyone else is between 2000-8000.
5373 		 */
5374 		u32 goc = (adapter->gotc + adapter->gorc) / 10000;
5375 		u32 dif = (adapter->gotc > adapter->gorc ?
5376 			   adapter->gotc - adapter->gorc :
5377 			   adapter->gorc - adapter->gotc) / 10000;
5378 		u32 itr = goc > 0 ? (dif * 6000 / goc + 2000) : 8000;
5379 
5380 		e1000e_write_itr(adapter, itr);
5381 	}
5382 
5383 	/* Cause software interrupt to ensure Rx ring is cleaned */
5384 	if (adapter->msix_entries)
5385 		ew32(ICS, adapter->rx_ring->ims_val);
5386 	else
5387 		ew32(ICS, E1000_ICS_RXDMT0);
5388 
5389 	/* flush pending descriptors to memory before detecting Tx hang */
5390 	e1000e_flush_descriptors(adapter);
5391 
5392 	/* Force detection of hung controller every watchdog period */
5393 	adapter->detect_tx_hung = true;
5394 
5395 	/* With 82571 controllers, LAA may be overwritten due to controller
5396 	 * reset from the other port. Set the appropriate LAA in RAR[0]
5397 	 */
5398 	if (e1000e_get_laa_state_82571(hw))
5399 		hw->mac.ops.rar_set(hw, adapter->hw.mac.addr, 0);
5400 
5401 	if (adapter->flags2 & FLAG2_CHECK_PHY_HANG)
5402 		e1000e_check_82574_phy_workaround(adapter);
5403 
5404 	/* Clear valid timestamp stuck in RXSTMPL/H due to a Rx error */
5405 	if (adapter->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE) {
5406 		if ((adapter->flags2 & FLAG2_CHECK_RX_HWTSTAMP) &&
5407 		    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID)) {
5408 			er32(RXSTMPH);
5409 			adapter->rx_hwtstamp_cleared++;
5410 		} else {
5411 			adapter->flags2 |= FLAG2_CHECK_RX_HWTSTAMP;
5412 		}
5413 	}
5414 
5415 	/* Reset the timer */
5416 	if (!test_bit(__E1000_DOWN, &adapter->state))
5417 		mod_timer(&adapter->watchdog_timer,
5418 			  round_jiffies(jiffies + 2 * HZ));
5419 }
5420 
5421 #define E1000_TX_FLAGS_CSUM		0x00000001
5422 #define E1000_TX_FLAGS_VLAN		0x00000002
5423 #define E1000_TX_FLAGS_TSO		0x00000004
5424 #define E1000_TX_FLAGS_IPV4		0x00000008
5425 #define E1000_TX_FLAGS_NO_FCS		0x00000010
5426 #define E1000_TX_FLAGS_HWTSTAMP		0x00000020
5427 #define E1000_TX_FLAGS_VLAN_MASK	0xffff0000
5428 #define E1000_TX_FLAGS_VLAN_SHIFT	16
5429 
5430 static int e1000_tso(struct e1000_ring *tx_ring, struct sk_buff *skb,
5431 		     __be16 protocol)
5432 {
5433 	struct e1000_context_desc *context_desc;
5434 	struct e1000_buffer *buffer_info;
5435 	unsigned int i;
5436 	u32 cmd_length = 0;
5437 	u16 ipcse = 0, mss;
5438 	u8 ipcss, ipcso, tucss, tucso, hdr_len;
5439 	int err;
5440 
5441 	if (!skb_is_gso(skb))
5442 		return 0;
5443 
5444 	err = skb_cow_head(skb, 0);
5445 	if (err < 0)
5446 		return err;
5447 
5448 	hdr_len = skb_tcp_all_headers(skb);
5449 	mss = skb_shinfo(skb)->gso_size;
5450 	if (protocol == htons(ETH_P_IP)) {
5451 		struct iphdr *iph = ip_hdr(skb);
5452 		iph->tot_len = 0;
5453 		iph->check = 0;
5454 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
5455 							 0, IPPROTO_TCP, 0);
5456 		cmd_length = E1000_TXD_CMD_IP;
5457 		ipcse = skb_transport_offset(skb) - 1;
5458 	} else if (skb_is_gso_v6(skb)) {
5459 		tcp_v6_gso_csum_prep(skb);
5460 		ipcse = 0;
5461 	}
5462 	ipcss = skb_network_offset(skb);
5463 	ipcso = (void *)&(ip_hdr(skb)->check) - (void *)skb->data;
5464 	tucss = skb_transport_offset(skb);
5465 	tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
5466 
5467 	cmd_length |= (E1000_TXD_CMD_DEXT | E1000_TXD_CMD_TSE |
5468 		       E1000_TXD_CMD_TCP | (skb->len - (hdr_len)));
5469 
5470 	i = tx_ring->next_to_use;
5471 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5472 	buffer_info = &tx_ring->buffer_info[i];
5473 
5474 	context_desc->lower_setup.ip_fields.ipcss = ipcss;
5475 	context_desc->lower_setup.ip_fields.ipcso = ipcso;
5476 	context_desc->lower_setup.ip_fields.ipcse = cpu_to_le16(ipcse);
5477 	context_desc->upper_setup.tcp_fields.tucss = tucss;
5478 	context_desc->upper_setup.tcp_fields.tucso = tucso;
5479 	context_desc->upper_setup.tcp_fields.tucse = 0;
5480 	context_desc->tcp_seg_setup.fields.mss = cpu_to_le16(mss);
5481 	context_desc->tcp_seg_setup.fields.hdr_len = hdr_len;
5482 	context_desc->cmd_and_length = cpu_to_le32(cmd_length);
5483 
5484 	buffer_info->time_stamp = jiffies;
5485 	buffer_info->next_to_watch = i;
5486 
5487 	i++;
5488 	if (i == tx_ring->count)
5489 		i = 0;
5490 	tx_ring->next_to_use = i;
5491 
5492 	return 1;
5493 }
5494 
5495 static bool e1000_tx_csum(struct e1000_ring *tx_ring, struct sk_buff *skb,
5496 			  __be16 protocol)
5497 {
5498 	struct e1000_adapter *adapter = tx_ring->adapter;
5499 	struct e1000_context_desc *context_desc;
5500 	struct e1000_buffer *buffer_info;
5501 	unsigned int i;
5502 	u8 css;
5503 	u32 cmd_len = E1000_TXD_CMD_DEXT;
5504 
5505 	if (skb->ip_summed != CHECKSUM_PARTIAL)
5506 		return false;
5507 
5508 	switch (protocol) {
5509 	case cpu_to_be16(ETH_P_IP):
5510 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
5511 			cmd_len |= E1000_TXD_CMD_TCP;
5512 		break;
5513 	case cpu_to_be16(ETH_P_IPV6):
5514 		/* XXX not handling all IPV6 headers */
5515 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
5516 			cmd_len |= E1000_TXD_CMD_TCP;
5517 		break;
5518 	default:
5519 		if (unlikely(net_ratelimit()))
5520 			e_warn("checksum_partial proto=%x!\n",
5521 			       be16_to_cpu(protocol));
5522 		break;
5523 	}
5524 
5525 	css = skb_checksum_start_offset(skb);
5526 
5527 	i = tx_ring->next_to_use;
5528 	buffer_info = &tx_ring->buffer_info[i];
5529 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5530 
5531 	context_desc->lower_setup.ip_config = 0;
5532 	context_desc->upper_setup.tcp_fields.tucss = css;
5533 	context_desc->upper_setup.tcp_fields.tucso = css + skb->csum_offset;
5534 	context_desc->upper_setup.tcp_fields.tucse = 0;
5535 	context_desc->tcp_seg_setup.data = 0;
5536 	context_desc->cmd_and_length = cpu_to_le32(cmd_len);
5537 
5538 	buffer_info->time_stamp = jiffies;
5539 	buffer_info->next_to_watch = i;
5540 
5541 	i++;
5542 	if (i == tx_ring->count)
5543 		i = 0;
5544 	tx_ring->next_to_use = i;
5545 
5546 	return true;
5547 }
5548 
5549 static int e1000_tx_map(struct e1000_ring *tx_ring, struct sk_buff *skb,
5550 			unsigned int first, unsigned int max_per_txd,
5551 			unsigned int nr_frags)
5552 {
5553 	struct e1000_adapter *adapter = tx_ring->adapter;
5554 	struct pci_dev *pdev = adapter->pdev;
5555 	struct e1000_buffer *buffer_info;
5556 	unsigned int len = skb_headlen(skb);
5557 	unsigned int offset = 0, size, count = 0, i;
5558 	unsigned int f, bytecount, segs;
5559 
5560 	i = tx_ring->next_to_use;
5561 
5562 	while (len) {
5563 		buffer_info = &tx_ring->buffer_info[i];
5564 		size = min(len, max_per_txd);
5565 
5566 		buffer_info->length = size;
5567 		buffer_info->time_stamp = jiffies;
5568 		buffer_info->next_to_watch = i;
5569 		buffer_info->dma = dma_map_single(&pdev->dev,
5570 						  skb->data + offset,
5571 						  size, DMA_TO_DEVICE);
5572 		buffer_info->mapped_as_page = false;
5573 		if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5574 			goto dma_error;
5575 
5576 		len -= size;
5577 		offset += size;
5578 		count++;
5579 
5580 		if (len) {
5581 			i++;
5582 			if (i == tx_ring->count)
5583 				i = 0;
5584 		}
5585 	}
5586 
5587 	for (f = 0; f < nr_frags; f++) {
5588 		const skb_frag_t *frag = &skb_shinfo(skb)->frags[f];
5589 
5590 		len = skb_frag_size(frag);
5591 		offset = 0;
5592 
5593 		while (len) {
5594 			i++;
5595 			if (i == tx_ring->count)
5596 				i = 0;
5597 
5598 			buffer_info = &tx_ring->buffer_info[i];
5599 			size = min(len, max_per_txd);
5600 
5601 			buffer_info->length = size;
5602 			buffer_info->time_stamp = jiffies;
5603 			buffer_info->next_to_watch = i;
5604 			buffer_info->dma = skb_frag_dma_map(&pdev->dev, frag,
5605 							    offset, size,
5606 							    DMA_TO_DEVICE);
5607 			buffer_info->mapped_as_page = true;
5608 			if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5609 				goto dma_error;
5610 
5611 			len -= size;
5612 			offset += size;
5613 			count++;
5614 		}
5615 	}
5616 
5617 	segs = skb_shinfo(skb)->gso_segs ? : 1;
5618 	/* multiply data chunks by size of headers */
5619 	bytecount = ((segs - 1) * skb_headlen(skb)) + skb->len;
5620 
5621 	tx_ring->buffer_info[i].skb = skb;
5622 	tx_ring->buffer_info[i].segs = segs;
5623 	tx_ring->buffer_info[i].bytecount = bytecount;
5624 	tx_ring->buffer_info[first].next_to_watch = i;
5625 
5626 	return count;
5627 
5628 dma_error:
5629 	dev_err(&pdev->dev, "Tx DMA map failed\n");
5630 	buffer_info->dma = 0;
5631 	if (count)
5632 		count--;
5633 
5634 	while (count--) {
5635 		if (i == 0)
5636 			i += tx_ring->count;
5637 		i--;
5638 		buffer_info = &tx_ring->buffer_info[i];
5639 		e1000_put_txbuf(tx_ring, buffer_info, true);
5640 	}
5641 
5642 	return 0;
5643 }
5644 
5645 static void e1000_tx_queue(struct e1000_ring *tx_ring, int tx_flags, int count)
5646 {
5647 	struct e1000_adapter *adapter = tx_ring->adapter;
5648 	struct e1000_tx_desc *tx_desc = NULL;
5649 	struct e1000_buffer *buffer_info;
5650 	u32 txd_upper = 0, txd_lower = E1000_TXD_CMD_IFCS;
5651 	unsigned int i;
5652 
5653 	if (tx_flags & E1000_TX_FLAGS_TSO) {
5654 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D |
5655 		    E1000_TXD_CMD_TSE;
5656 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5657 
5658 		if (tx_flags & E1000_TX_FLAGS_IPV4)
5659 			txd_upper |= E1000_TXD_POPTS_IXSM << 8;
5660 	}
5661 
5662 	if (tx_flags & E1000_TX_FLAGS_CSUM) {
5663 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5664 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5665 	}
5666 
5667 	if (tx_flags & E1000_TX_FLAGS_VLAN) {
5668 		txd_lower |= E1000_TXD_CMD_VLE;
5669 		txd_upper |= (tx_flags & E1000_TX_FLAGS_VLAN_MASK);
5670 	}
5671 
5672 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5673 		txd_lower &= ~(E1000_TXD_CMD_IFCS);
5674 
5675 	if (unlikely(tx_flags & E1000_TX_FLAGS_HWTSTAMP)) {
5676 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5677 		txd_upper |= E1000_TXD_EXTCMD_TSTAMP;
5678 	}
5679 
5680 	i = tx_ring->next_to_use;
5681 
5682 	do {
5683 		buffer_info = &tx_ring->buffer_info[i];
5684 		tx_desc = E1000_TX_DESC(*tx_ring, i);
5685 		tx_desc->buffer_addr = cpu_to_le64(buffer_info->dma);
5686 		tx_desc->lower.data = cpu_to_le32(txd_lower |
5687 						  buffer_info->length);
5688 		tx_desc->upper.data = cpu_to_le32(txd_upper);
5689 
5690 		i++;
5691 		if (i == tx_ring->count)
5692 			i = 0;
5693 	} while (--count > 0);
5694 
5695 	tx_desc->lower.data |= cpu_to_le32(adapter->txd_cmd);
5696 
5697 	/* txd_cmd re-enables FCS, so we'll re-disable it here as desired. */
5698 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5699 		tx_desc->lower.data &= ~(cpu_to_le32(E1000_TXD_CMD_IFCS));
5700 
5701 	/* Force memory writes to complete before letting h/w
5702 	 * know there are new descriptors to fetch.  (Only
5703 	 * applicable for weak-ordered memory model archs,
5704 	 * such as IA-64).
5705 	 */
5706 	wmb();
5707 
5708 	tx_ring->next_to_use = i;
5709 }
5710 
5711 #define MINIMUM_DHCP_PACKET_SIZE 282
5712 static int e1000_transfer_dhcp_info(struct e1000_adapter *adapter,
5713 				    struct sk_buff *skb)
5714 {
5715 	struct e1000_hw *hw = &adapter->hw;
5716 	u16 length, offset;
5717 
5718 	if (skb_vlan_tag_present(skb) &&
5719 	    !((skb_vlan_tag_get(skb) == adapter->hw.mng_cookie.vlan_id) &&
5720 	      (adapter->hw.mng_cookie.status &
5721 	       E1000_MNG_DHCP_COOKIE_STATUS_VLAN)))
5722 		return 0;
5723 
5724 	if (skb->len <= MINIMUM_DHCP_PACKET_SIZE)
5725 		return 0;
5726 
5727 	if (((struct ethhdr *)skb->data)->h_proto != htons(ETH_P_IP))
5728 		return 0;
5729 
5730 	{
5731 		const struct iphdr *ip = (struct iphdr *)((u8 *)skb->data + 14);
5732 		struct udphdr *udp;
5733 
5734 		if (ip->protocol != IPPROTO_UDP)
5735 			return 0;
5736 
5737 		udp = (struct udphdr *)((u8 *)ip + (ip->ihl << 2));
5738 		if (ntohs(udp->dest) != 67)
5739 			return 0;
5740 
5741 		offset = (u8 *)udp + 8 - skb->data;
5742 		length = skb->len - offset;
5743 		return e1000e_mng_write_dhcp_info(hw, (u8 *)udp + 8, length);
5744 	}
5745 
5746 	return 0;
5747 }
5748 
5749 static int __e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5750 {
5751 	struct e1000_adapter *adapter = tx_ring->adapter;
5752 
5753 	netif_stop_queue(adapter->netdev);
5754 	/* Herbert's original patch had:
5755 	 *  smp_mb__after_netif_stop_queue();
5756 	 * but since that doesn't exist yet, just open code it.
5757 	 */
5758 	smp_mb();
5759 
5760 	/* We need to check again in a case another CPU has just
5761 	 * made room available.
5762 	 */
5763 	if (e1000_desc_unused(tx_ring) < size)
5764 		return -EBUSY;
5765 
5766 	/* A reprieve! */
5767 	netif_start_queue(adapter->netdev);
5768 	++adapter->restart_queue;
5769 	return 0;
5770 }
5771 
5772 static int e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5773 {
5774 	BUG_ON(size > tx_ring->count);
5775 
5776 	if (e1000_desc_unused(tx_ring) >= size)
5777 		return 0;
5778 	return __e1000_maybe_stop_tx(tx_ring, size);
5779 }
5780 
5781 static netdev_tx_t e1000_xmit_frame(struct sk_buff *skb,
5782 				    struct net_device *netdev)
5783 {
5784 	struct e1000_adapter *adapter = netdev_priv(netdev);
5785 	struct e1000_ring *tx_ring = adapter->tx_ring;
5786 	unsigned int first;
5787 	unsigned int tx_flags = 0;
5788 	unsigned int len = skb_headlen(skb);
5789 	unsigned int nr_frags;
5790 	unsigned int mss;
5791 	int count = 0;
5792 	int tso;
5793 	unsigned int f;
5794 	__be16 protocol = vlan_get_protocol(skb);
5795 
5796 	if (test_bit(__E1000_DOWN, &adapter->state)) {
5797 		dev_kfree_skb_any(skb);
5798 		return NETDEV_TX_OK;
5799 	}
5800 
5801 	if (skb->len <= 0) {
5802 		dev_kfree_skb_any(skb);
5803 		return NETDEV_TX_OK;
5804 	}
5805 
5806 	/* The minimum packet size with TCTL.PSP set is 17 bytes so
5807 	 * pad skb in order to meet this minimum size requirement
5808 	 */
5809 	if (skb_put_padto(skb, 17))
5810 		return NETDEV_TX_OK;
5811 
5812 	mss = skb_shinfo(skb)->gso_size;
5813 	if (mss) {
5814 		u8 hdr_len;
5815 
5816 		/* TSO Workaround for 82571/2/3 Controllers -- if skb->data
5817 		 * points to just header, pull a few bytes of payload from
5818 		 * frags into skb->data
5819 		 */
5820 		hdr_len = skb_tcp_all_headers(skb);
5821 		/* we do this workaround for ES2LAN, but it is un-necessary,
5822 		 * avoiding it could save a lot of cycles
5823 		 */
5824 		if (skb->data_len && (hdr_len == len)) {
5825 			unsigned int pull_size;
5826 
5827 			pull_size = min_t(unsigned int, 4, skb->data_len);
5828 			if (!__pskb_pull_tail(skb, pull_size)) {
5829 				e_err("__pskb_pull_tail failed.\n");
5830 				dev_kfree_skb_any(skb);
5831 				return NETDEV_TX_OK;
5832 			}
5833 			len = skb_headlen(skb);
5834 		}
5835 	}
5836 
5837 	/* reserve a descriptor for the offload context */
5838 	if ((mss) || (skb->ip_summed == CHECKSUM_PARTIAL))
5839 		count++;
5840 	count++;
5841 
5842 	count += DIV_ROUND_UP(len, adapter->tx_fifo_limit);
5843 
5844 	nr_frags = skb_shinfo(skb)->nr_frags;
5845 	for (f = 0; f < nr_frags; f++)
5846 		count += DIV_ROUND_UP(skb_frag_size(&skb_shinfo(skb)->frags[f]),
5847 				      adapter->tx_fifo_limit);
5848 
5849 	if (adapter->hw.mac.tx_pkt_filtering)
5850 		e1000_transfer_dhcp_info(adapter, skb);
5851 
5852 	/* need: count + 2 desc gap to keep tail from touching
5853 	 * head, otherwise try next time
5854 	 */
5855 	if (e1000_maybe_stop_tx(tx_ring, count + 2))
5856 		return NETDEV_TX_BUSY;
5857 
5858 	if (skb_vlan_tag_present(skb)) {
5859 		tx_flags |= E1000_TX_FLAGS_VLAN;
5860 		tx_flags |= (skb_vlan_tag_get(skb) <<
5861 			     E1000_TX_FLAGS_VLAN_SHIFT);
5862 	}
5863 
5864 	first = tx_ring->next_to_use;
5865 
5866 	tso = e1000_tso(tx_ring, skb, protocol);
5867 	if (tso < 0) {
5868 		dev_kfree_skb_any(skb);
5869 		return NETDEV_TX_OK;
5870 	}
5871 
5872 	if (tso)
5873 		tx_flags |= E1000_TX_FLAGS_TSO;
5874 	else if (e1000_tx_csum(tx_ring, skb, protocol))
5875 		tx_flags |= E1000_TX_FLAGS_CSUM;
5876 
5877 	/* Old method was to assume IPv4 packet by default if TSO was enabled.
5878 	 * 82571 hardware supports TSO capabilities for IPv6 as well...
5879 	 * no longer assume, we must.
5880 	 */
5881 	if (protocol == htons(ETH_P_IP))
5882 		tx_flags |= E1000_TX_FLAGS_IPV4;
5883 
5884 	if (unlikely(skb->no_fcs))
5885 		tx_flags |= E1000_TX_FLAGS_NO_FCS;
5886 
5887 	/* if count is 0 then mapping error has occurred */
5888 	count = e1000_tx_map(tx_ring, skb, first, adapter->tx_fifo_limit,
5889 			     nr_frags);
5890 	if (count) {
5891 		if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
5892 		    (adapter->flags & FLAG_HAS_HW_TIMESTAMP)) {
5893 			if (!adapter->tx_hwtstamp_skb) {
5894 				skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5895 				tx_flags |= E1000_TX_FLAGS_HWTSTAMP;
5896 				adapter->tx_hwtstamp_skb = skb_get(skb);
5897 				adapter->tx_hwtstamp_start = jiffies;
5898 				schedule_work(&adapter->tx_hwtstamp_work);
5899 			} else {
5900 				adapter->tx_hwtstamp_skipped++;
5901 			}
5902 		}
5903 
5904 		skb_tx_timestamp(skb);
5905 
5906 		netdev_sent_queue(netdev, skb->len);
5907 		e1000_tx_queue(tx_ring, tx_flags, count);
5908 		/* Make sure there is space in the ring for the next send. */
5909 		e1000_maybe_stop_tx(tx_ring,
5910 				    ((MAX_SKB_FRAGS + 1) *
5911 				     DIV_ROUND_UP(PAGE_SIZE,
5912 						  adapter->tx_fifo_limit) + 4));
5913 
5914 		if (!netdev_xmit_more() ||
5915 		    netif_xmit_stopped(netdev_get_tx_queue(netdev, 0))) {
5916 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
5917 				e1000e_update_tdt_wa(tx_ring,
5918 						     tx_ring->next_to_use);
5919 			else
5920 				writel(tx_ring->next_to_use, tx_ring->tail);
5921 		}
5922 	} else {
5923 		dev_kfree_skb_any(skb);
5924 		tx_ring->buffer_info[first].time_stamp = 0;
5925 		tx_ring->next_to_use = first;
5926 	}
5927 
5928 	return NETDEV_TX_OK;
5929 }
5930 
5931 /**
5932  * e1000_tx_timeout - Respond to a Tx Hang
5933  * @netdev: network interface device structure
5934  * @txqueue: index of the hung queue (unused)
5935  **/
5936 static void e1000_tx_timeout(struct net_device *netdev, unsigned int __always_unused txqueue)
5937 {
5938 	struct e1000_adapter *adapter = netdev_priv(netdev);
5939 
5940 	/* Do the reset outside of interrupt context */
5941 	adapter->tx_timeout_count++;
5942 	schedule_work(&adapter->reset_task);
5943 }
5944 
5945 static void e1000_reset_task(struct work_struct *work)
5946 {
5947 	struct e1000_adapter *adapter;
5948 	adapter = container_of(work, struct e1000_adapter, reset_task);
5949 
5950 	rtnl_lock();
5951 	/* don't run the task if already down */
5952 	if (test_bit(__E1000_DOWN, &adapter->state)) {
5953 		rtnl_unlock();
5954 		return;
5955 	}
5956 
5957 	if (!(adapter->flags & FLAG_RESTART_NOW)) {
5958 		e1000e_dump(adapter);
5959 		e_err("Reset adapter unexpectedly\n");
5960 	}
5961 	e1000e_reinit_locked(adapter);
5962 	rtnl_unlock();
5963 }
5964 
5965 /**
5966  * e1000e_get_stats64 - Get System Network Statistics
5967  * @netdev: network interface device structure
5968  * @stats: rtnl_link_stats64 pointer
5969  *
5970  * Returns the address of the device statistics structure.
5971  **/
5972 void e1000e_get_stats64(struct net_device *netdev,
5973 			struct rtnl_link_stats64 *stats)
5974 {
5975 	struct e1000_adapter *adapter = netdev_priv(netdev);
5976 
5977 	spin_lock(&adapter->stats64_lock);
5978 	e1000e_update_stats(adapter);
5979 	/* Fill out the OS statistics structure */
5980 	stats->rx_bytes = adapter->stats.gorc;
5981 	stats->rx_packets = adapter->stats.gprc;
5982 	stats->tx_bytes = adapter->stats.gotc;
5983 	stats->tx_packets = adapter->stats.gptc;
5984 	stats->multicast = adapter->stats.mprc;
5985 	stats->collisions = adapter->stats.colc;
5986 
5987 	/* Rx Errors */
5988 
5989 	/* RLEC on some newer hardware can be incorrect so build
5990 	 * our own version based on RUC and ROC
5991 	 */
5992 	stats->rx_errors = adapter->stats.rxerrc +
5993 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
5994 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5995 	stats->rx_length_errors = adapter->stats.ruc + adapter->stats.roc;
5996 	stats->rx_crc_errors = adapter->stats.crcerrs;
5997 	stats->rx_frame_errors = adapter->stats.algnerrc;
5998 	stats->rx_missed_errors = adapter->stats.mpc;
5999 
6000 	/* Tx Errors */
6001 	stats->tx_errors = adapter->stats.ecol + adapter->stats.latecol;
6002 	stats->tx_aborted_errors = adapter->stats.ecol;
6003 	stats->tx_window_errors = adapter->stats.latecol;
6004 	stats->tx_carrier_errors = adapter->stats.tncrs;
6005 
6006 	/* Tx Dropped needs to be maintained elsewhere */
6007 
6008 	spin_unlock(&adapter->stats64_lock);
6009 }
6010 
6011 /**
6012  * e1000_change_mtu - Change the Maximum Transfer Unit
6013  * @netdev: network interface device structure
6014  * @new_mtu: new value for maximum frame size
6015  *
6016  * Returns 0 on success, negative on failure
6017  **/
6018 static int e1000_change_mtu(struct net_device *netdev, int new_mtu)
6019 {
6020 	struct e1000_adapter *adapter = netdev_priv(netdev);
6021 	int max_frame = new_mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
6022 
6023 	/* Jumbo frame support */
6024 	if ((new_mtu > ETH_DATA_LEN) &&
6025 	    !(adapter->flags & FLAG_HAS_JUMBO_FRAMES)) {
6026 		e_err("Jumbo Frames not supported.\n");
6027 		return -EINVAL;
6028 	}
6029 
6030 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
6031 	if ((adapter->hw.mac.type >= e1000_pch2lan) &&
6032 	    !(adapter->flags2 & FLAG2_CRC_STRIPPING) &&
6033 	    (new_mtu > ETH_DATA_LEN)) {
6034 		e_err("Jumbo Frames not supported on this device when CRC stripping is disabled.\n");
6035 		return -EINVAL;
6036 	}
6037 
6038 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
6039 		usleep_range(1000, 1100);
6040 	/* e1000e_down -> e1000e_reset dependent on max_frame_size & mtu */
6041 	adapter->max_frame_size = max_frame;
6042 	netdev_dbg(netdev, "changing MTU from %d to %d\n",
6043 		   netdev->mtu, new_mtu);
6044 	netdev->mtu = new_mtu;
6045 
6046 	pm_runtime_get_sync(netdev->dev.parent);
6047 
6048 	if (netif_running(netdev))
6049 		e1000e_down(adapter, true);
6050 
6051 	/* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
6052 	 * means we reserve 2 more, this pushes us to allocate from the next
6053 	 * larger slab size.
6054 	 * i.e. RXBUFFER_2048 --> size-4096 slab
6055 	 * However with the new *_jumbo_rx* routines, jumbo receives will use
6056 	 * fragmented skbs
6057 	 */
6058 
6059 	if (max_frame <= 2048)
6060 		adapter->rx_buffer_len = 2048;
6061 	else
6062 		adapter->rx_buffer_len = 4096;
6063 
6064 	/* adjust allocation if LPE protects us, and we aren't using SBP */
6065 	if (max_frame <= (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN))
6066 		adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
6067 
6068 	if (netif_running(netdev))
6069 		e1000e_up(adapter);
6070 	else
6071 		e1000e_reset(adapter);
6072 
6073 	pm_runtime_put_sync(netdev->dev.parent);
6074 
6075 	clear_bit(__E1000_RESETTING, &adapter->state);
6076 
6077 	return 0;
6078 }
6079 
6080 static int e1000_mii_ioctl(struct net_device *netdev, struct ifreq *ifr,
6081 			   int cmd)
6082 {
6083 	struct e1000_adapter *adapter = netdev_priv(netdev);
6084 	struct mii_ioctl_data *data = if_mii(ifr);
6085 
6086 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
6087 		return -EOPNOTSUPP;
6088 
6089 	switch (cmd) {
6090 	case SIOCGMIIPHY:
6091 		data->phy_id = adapter->hw.phy.addr;
6092 		break;
6093 	case SIOCGMIIREG:
6094 		e1000_phy_read_status(adapter);
6095 
6096 		switch (data->reg_num & 0x1F) {
6097 		case MII_BMCR:
6098 			data->val_out = adapter->phy_regs.bmcr;
6099 			break;
6100 		case MII_BMSR:
6101 			data->val_out = adapter->phy_regs.bmsr;
6102 			break;
6103 		case MII_PHYSID1:
6104 			data->val_out = (adapter->hw.phy.id >> 16);
6105 			break;
6106 		case MII_PHYSID2:
6107 			data->val_out = (adapter->hw.phy.id & 0xFFFF);
6108 			break;
6109 		case MII_ADVERTISE:
6110 			data->val_out = adapter->phy_regs.advertise;
6111 			break;
6112 		case MII_LPA:
6113 			data->val_out = adapter->phy_regs.lpa;
6114 			break;
6115 		case MII_EXPANSION:
6116 			data->val_out = adapter->phy_regs.expansion;
6117 			break;
6118 		case MII_CTRL1000:
6119 			data->val_out = adapter->phy_regs.ctrl1000;
6120 			break;
6121 		case MII_STAT1000:
6122 			data->val_out = adapter->phy_regs.stat1000;
6123 			break;
6124 		case MII_ESTATUS:
6125 			data->val_out = adapter->phy_regs.estatus;
6126 			break;
6127 		default:
6128 			return -EIO;
6129 		}
6130 		break;
6131 	case SIOCSMIIREG:
6132 	default:
6133 		return -EOPNOTSUPP;
6134 	}
6135 	return 0;
6136 }
6137 
6138 /**
6139  * e1000e_hwtstamp_set - control hardware time stamping
6140  * @netdev: network interface device structure
6141  * @ifr: interface request
6142  *
6143  * Outgoing time stamping can be enabled and disabled. Play nice and
6144  * disable it when requested, although it shouldn't cause any overhead
6145  * when no packet needs it. At most one packet in the queue may be
6146  * marked for time stamping, otherwise it would be impossible to tell
6147  * for sure to which packet the hardware time stamp belongs.
6148  *
6149  * Incoming time stamping has to be configured via the hardware filters.
6150  * Not all combinations are supported, in particular event type has to be
6151  * specified. Matching the kind of event packet is not supported, with the
6152  * exception of "all V2 events regardless of level 2 or 4".
6153  **/
6154 static int e1000e_hwtstamp_set(struct net_device *netdev, struct ifreq *ifr)
6155 {
6156 	struct e1000_adapter *adapter = netdev_priv(netdev);
6157 	struct hwtstamp_config config;
6158 	int ret_val;
6159 
6160 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
6161 		return -EFAULT;
6162 
6163 	ret_val = e1000e_config_hwtstamp(adapter, &config);
6164 	if (ret_val)
6165 		return ret_val;
6166 
6167 	switch (config.rx_filter) {
6168 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
6169 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
6170 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
6171 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
6172 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
6173 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
6174 		/* With V2 type filters which specify a Sync or Delay Request,
6175 		 * Path Delay Request/Response messages are also time stamped
6176 		 * by hardware so notify the caller the requested packets plus
6177 		 * some others are time stamped.
6178 		 */
6179 		config.rx_filter = HWTSTAMP_FILTER_SOME;
6180 		break;
6181 	default:
6182 		break;
6183 	}
6184 
6185 	return copy_to_user(ifr->ifr_data, &config,
6186 			    sizeof(config)) ? -EFAULT : 0;
6187 }
6188 
6189 static int e1000e_hwtstamp_get(struct net_device *netdev, struct ifreq *ifr)
6190 {
6191 	struct e1000_adapter *adapter = netdev_priv(netdev);
6192 
6193 	return copy_to_user(ifr->ifr_data, &adapter->hwtstamp_config,
6194 			    sizeof(adapter->hwtstamp_config)) ? -EFAULT : 0;
6195 }
6196 
6197 static int e1000_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
6198 {
6199 	switch (cmd) {
6200 	case SIOCGMIIPHY:
6201 	case SIOCGMIIREG:
6202 	case SIOCSMIIREG:
6203 		return e1000_mii_ioctl(netdev, ifr, cmd);
6204 	case SIOCSHWTSTAMP:
6205 		return e1000e_hwtstamp_set(netdev, ifr);
6206 	case SIOCGHWTSTAMP:
6207 		return e1000e_hwtstamp_get(netdev, ifr);
6208 	default:
6209 		return -EOPNOTSUPP;
6210 	}
6211 }
6212 
6213 static int e1000_init_phy_wakeup(struct e1000_adapter *adapter, u32 wufc)
6214 {
6215 	struct e1000_hw *hw = &adapter->hw;
6216 	u32 i, mac_reg, wuc;
6217 	u16 phy_reg, wuc_enable;
6218 	int retval;
6219 
6220 	/* copy MAC RARs to PHY RARs */
6221 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
6222 
6223 	retval = hw->phy.ops.acquire(hw);
6224 	if (retval) {
6225 		e_err("Could not acquire PHY\n");
6226 		return retval;
6227 	}
6228 
6229 	/* Enable access to wakeup registers on and set page to BM_WUC_PAGE */
6230 	retval = e1000_enable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6231 	if (retval)
6232 		goto release;
6233 
6234 	/* copy MAC MTA to PHY MTA - only needed for pchlan */
6235 	for (i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
6236 		mac_reg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
6237 		hw->phy.ops.write_reg_page(hw, BM_MTA(i),
6238 					   (u16)(mac_reg & 0xFFFF));
6239 		hw->phy.ops.write_reg_page(hw, BM_MTA(i) + 1,
6240 					   (u16)((mac_reg >> 16) & 0xFFFF));
6241 	}
6242 
6243 	/* configure PHY Rx Control register */
6244 	hw->phy.ops.read_reg_page(&adapter->hw, BM_RCTL, &phy_reg);
6245 	mac_reg = er32(RCTL);
6246 	if (mac_reg & E1000_RCTL_UPE)
6247 		phy_reg |= BM_RCTL_UPE;
6248 	if (mac_reg & E1000_RCTL_MPE)
6249 		phy_reg |= BM_RCTL_MPE;
6250 	phy_reg &= ~(BM_RCTL_MO_MASK);
6251 	if (mac_reg & E1000_RCTL_MO_3)
6252 		phy_reg |= (((mac_reg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
6253 			    << BM_RCTL_MO_SHIFT);
6254 	if (mac_reg & E1000_RCTL_BAM)
6255 		phy_reg |= BM_RCTL_BAM;
6256 	if (mac_reg & E1000_RCTL_PMCF)
6257 		phy_reg |= BM_RCTL_PMCF;
6258 	mac_reg = er32(CTRL);
6259 	if (mac_reg & E1000_CTRL_RFCE)
6260 		phy_reg |= BM_RCTL_RFCE;
6261 	hw->phy.ops.write_reg_page(&adapter->hw, BM_RCTL, phy_reg);
6262 
6263 	wuc = E1000_WUC_PME_EN;
6264 	if (wufc & (E1000_WUFC_MAG | E1000_WUFC_LNKC))
6265 		wuc |= E1000_WUC_APME;
6266 
6267 	/* enable PHY wakeup in MAC register */
6268 	ew32(WUFC, wufc);
6269 	ew32(WUC, (E1000_WUC_PHY_WAKE | E1000_WUC_APMPME |
6270 		   E1000_WUC_PME_STATUS | wuc));
6271 
6272 	/* configure and enable PHY wakeup in PHY registers */
6273 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUFC, wufc);
6274 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUC, wuc);
6275 
6276 	/* activate PHY wakeup */
6277 	wuc_enable |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
6278 	retval = e1000_disable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6279 	if (retval)
6280 		e_err("Could not set PHY Host Wakeup bit\n");
6281 release:
6282 	hw->phy.ops.release(hw);
6283 
6284 	return retval;
6285 }
6286 
6287 static void e1000e_flush_lpic(struct pci_dev *pdev)
6288 {
6289 	struct net_device *netdev = pci_get_drvdata(pdev);
6290 	struct e1000_adapter *adapter = netdev_priv(netdev);
6291 	struct e1000_hw *hw = &adapter->hw;
6292 	u32 ret_val;
6293 
6294 	pm_runtime_get_sync(netdev->dev.parent);
6295 
6296 	ret_val = hw->phy.ops.acquire(hw);
6297 	if (ret_val)
6298 		goto fl_out;
6299 
6300 	pr_info("EEE TX LPI TIMER: %08X\n",
6301 		er32(LPIC) >> E1000_LPIC_LPIET_SHIFT);
6302 
6303 	hw->phy.ops.release(hw);
6304 
6305 fl_out:
6306 	pm_runtime_put_sync(netdev->dev.parent);
6307 }
6308 
6309 /* S0ix implementation */
6310 static void e1000e_s0ix_entry_flow(struct e1000_adapter *adapter)
6311 {
6312 	struct e1000_hw *hw = &adapter->hw;
6313 	u32 mac_data;
6314 	u16 phy_data;
6315 
6316 	if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID &&
6317 	    hw->mac.type >= e1000_pch_adp) {
6318 		/* Request ME configure the device for S0ix */
6319 		mac_data = er32(H2ME);
6320 		mac_data |= E1000_H2ME_START_DPG;
6321 		mac_data &= ~E1000_H2ME_EXIT_DPG;
6322 		trace_e1000e_trace_mac_register(mac_data);
6323 		ew32(H2ME, mac_data);
6324 	} else {
6325 		/* Request driver configure the device to S0ix */
6326 		/* Disable the periodic inband message,
6327 		 * don't request PCIe clock in K1 page770_17[10:9] = 10b
6328 		 */
6329 		e1e_rphy(hw, HV_PM_CTRL, &phy_data);
6330 		phy_data &= ~HV_PM_CTRL_K1_CLK_REQ;
6331 		phy_data |= BIT(10);
6332 		e1e_wphy(hw, HV_PM_CTRL, phy_data);
6333 
6334 		/* Make sure we don't exit K1 every time a new packet arrives
6335 		 * 772_29[5] = 1 CS_Mode_Stay_In_K1
6336 		 */
6337 		e1e_rphy(hw, I217_CGFREG, &phy_data);
6338 		phy_data |= BIT(5);
6339 		e1e_wphy(hw, I217_CGFREG, phy_data);
6340 
6341 		/* Change the MAC/PHY interface to SMBus
6342 		 * Force the SMBus in PHY page769_23[0] = 1
6343 		 * Force the SMBus in MAC CTRL_EXT[11] = 1
6344 		 */
6345 		e1e_rphy(hw, CV_SMB_CTRL, &phy_data);
6346 		phy_data |= CV_SMB_CTRL_FORCE_SMBUS;
6347 		e1e_wphy(hw, CV_SMB_CTRL, phy_data);
6348 		mac_data = er32(CTRL_EXT);
6349 		mac_data |= E1000_CTRL_EXT_FORCE_SMBUS;
6350 		ew32(CTRL_EXT, mac_data);
6351 
6352 		/* DFT control: PHY bit: page769_20[0] = 1
6353 		 * page769_20[7] - PHY PLL stop
6354 		 * page769_20[8] - PHY go to the electrical idle
6355 		 * page769_20[9] - PHY serdes disable
6356 		 * Gate PPW via EXTCNF_CTRL - set 0x0F00[7] = 1
6357 		 */
6358 		e1e_rphy(hw, I82579_DFT_CTRL, &phy_data);
6359 		phy_data |= BIT(0);
6360 		phy_data |= BIT(7);
6361 		phy_data |= BIT(8);
6362 		phy_data |= BIT(9);
6363 		e1e_wphy(hw, I82579_DFT_CTRL, phy_data);
6364 
6365 		mac_data = er32(EXTCNF_CTRL);
6366 		mac_data |= E1000_EXTCNF_CTRL_GATE_PHY_CFG;
6367 		ew32(EXTCNF_CTRL, mac_data);
6368 
6369 		/* Enable the Dynamic Power Gating in the MAC */
6370 		mac_data = er32(FEXTNVM7);
6371 		mac_data |= BIT(22);
6372 		ew32(FEXTNVM7, mac_data);
6373 
6374 		/* Disable disconnected cable conditioning for Power Gating */
6375 		mac_data = er32(DPGFR);
6376 		mac_data |= BIT(2);
6377 		ew32(DPGFR, mac_data);
6378 
6379 		/* Don't wake from dynamic Power Gating with clock request */
6380 		mac_data = er32(FEXTNVM12);
6381 		mac_data |= BIT(12);
6382 		ew32(FEXTNVM12, mac_data);
6383 
6384 		/* Ungate PGCB clock */
6385 		mac_data = er32(FEXTNVM9);
6386 		mac_data &= ~BIT(28);
6387 		ew32(FEXTNVM9, mac_data);
6388 
6389 		/* Enable K1 off to enable mPHY Power Gating */
6390 		mac_data = er32(FEXTNVM6);
6391 		mac_data |= BIT(31);
6392 		ew32(FEXTNVM6, mac_data);
6393 
6394 		/* Enable mPHY power gating for any link and speed */
6395 		mac_data = er32(FEXTNVM8);
6396 		mac_data |= BIT(9);
6397 		ew32(FEXTNVM8, mac_data);
6398 
6399 		/* Enable the Dynamic Clock Gating in the DMA and MAC */
6400 		mac_data = er32(CTRL_EXT);
6401 		mac_data |= E1000_CTRL_EXT_DMA_DYN_CLK_EN;
6402 		ew32(CTRL_EXT, mac_data);
6403 
6404 		/* No MAC DPG gating SLP_S0 in modern standby
6405 		 * Switch the logic of the lanphypc to use PMC counter
6406 		 */
6407 		mac_data = er32(FEXTNVM5);
6408 		mac_data |= BIT(7);
6409 		ew32(FEXTNVM5, mac_data);
6410 	}
6411 
6412 	/* Disable the time synchronization clock */
6413 	mac_data = er32(FEXTNVM7);
6414 	mac_data |= BIT(31);
6415 	mac_data &= ~BIT(0);
6416 	ew32(FEXTNVM7, mac_data);
6417 
6418 	/* Dynamic Power Gating Enable */
6419 	mac_data = er32(CTRL_EXT);
6420 	mac_data |= BIT(3);
6421 	ew32(CTRL_EXT, mac_data);
6422 
6423 	/* Check MAC Tx/Rx packet buffer pointers.
6424 	 * Reset MAC Tx/Rx packet buffer pointers to suppress any
6425 	 * pending traffic indication that would prevent power gating.
6426 	 */
6427 	mac_data = er32(TDFH);
6428 	if (mac_data)
6429 		ew32(TDFH, 0);
6430 	mac_data = er32(TDFT);
6431 	if (mac_data)
6432 		ew32(TDFT, 0);
6433 	mac_data = er32(TDFHS);
6434 	if (mac_data)
6435 		ew32(TDFHS, 0);
6436 	mac_data = er32(TDFTS);
6437 	if (mac_data)
6438 		ew32(TDFTS, 0);
6439 	mac_data = er32(TDFPC);
6440 	if (mac_data)
6441 		ew32(TDFPC, 0);
6442 	mac_data = er32(RDFH);
6443 	if (mac_data)
6444 		ew32(RDFH, 0);
6445 	mac_data = er32(RDFT);
6446 	if (mac_data)
6447 		ew32(RDFT, 0);
6448 	mac_data = er32(RDFHS);
6449 	if (mac_data)
6450 		ew32(RDFHS, 0);
6451 	mac_data = er32(RDFTS);
6452 	if (mac_data)
6453 		ew32(RDFTS, 0);
6454 	mac_data = er32(RDFPC);
6455 	if (mac_data)
6456 		ew32(RDFPC, 0);
6457 }
6458 
6459 static void e1000e_s0ix_exit_flow(struct e1000_adapter *adapter)
6460 {
6461 	struct e1000_hw *hw = &adapter->hw;
6462 	bool firmware_bug = false;
6463 	u32 mac_data;
6464 	u16 phy_data;
6465 	u32 i = 0;
6466 
6467 	if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID &&
6468 	    hw->mac.type >= e1000_pch_adp) {
6469 		/* Keep the GPT clock enabled for CSME */
6470 		mac_data = er32(FEXTNVM);
6471 		mac_data |= BIT(3);
6472 		ew32(FEXTNVM, mac_data);
6473 		/* Request ME unconfigure the device from S0ix */
6474 		mac_data = er32(H2ME);
6475 		mac_data &= ~E1000_H2ME_START_DPG;
6476 		mac_data |= E1000_H2ME_EXIT_DPG;
6477 		trace_e1000e_trace_mac_register(mac_data);
6478 		ew32(H2ME, mac_data);
6479 
6480 		/* Poll up to 2.5 seconds for ME to unconfigure DPG.
6481 		 * If this takes more than 1 second, show a warning indicating a
6482 		 * firmware bug
6483 		 */
6484 		while (!(er32(EXFWSM) & E1000_EXFWSM_DPG_EXIT_DONE)) {
6485 			if (i > 100 && !firmware_bug)
6486 				firmware_bug = true;
6487 
6488 			if (i++ == 250) {
6489 				e_dbg("Timeout (firmware bug): %d msec\n",
6490 				      i * 10);
6491 				break;
6492 			}
6493 
6494 			usleep_range(10000, 11000);
6495 		}
6496 		if (firmware_bug)
6497 			e_warn("DPG_EXIT_DONE took %d msec. This is a firmware bug\n",
6498 			       i * 10);
6499 		else
6500 			e_dbg("DPG_EXIT_DONE cleared after %d msec\n", i * 10);
6501 	} else {
6502 		/* Request driver unconfigure the device from S0ix */
6503 
6504 		/* Disable the Dynamic Power Gating in the MAC */
6505 		mac_data = er32(FEXTNVM7);
6506 		mac_data &= 0xFFBFFFFF;
6507 		ew32(FEXTNVM7, mac_data);
6508 
6509 		/* Disable mPHY power gating for any link and speed */
6510 		mac_data = er32(FEXTNVM8);
6511 		mac_data &= ~BIT(9);
6512 		ew32(FEXTNVM8, mac_data);
6513 
6514 		/* Disable K1 off */
6515 		mac_data = er32(FEXTNVM6);
6516 		mac_data &= ~BIT(31);
6517 		ew32(FEXTNVM6, mac_data);
6518 
6519 		/* Disable Ungate PGCB clock */
6520 		mac_data = er32(FEXTNVM9);
6521 		mac_data |= BIT(28);
6522 		ew32(FEXTNVM9, mac_data);
6523 
6524 		/* Cancel not waking from dynamic
6525 		 * Power Gating with clock request
6526 		 */
6527 		mac_data = er32(FEXTNVM12);
6528 		mac_data &= ~BIT(12);
6529 		ew32(FEXTNVM12, mac_data);
6530 
6531 		/* Cancel disable disconnected cable conditioning
6532 		 * for Power Gating
6533 		 */
6534 		mac_data = er32(DPGFR);
6535 		mac_data &= ~BIT(2);
6536 		ew32(DPGFR, mac_data);
6537 
6538 		/* Disable the Dynamic Clock Gating in the DMA and MAC */
6539 		mac_data = er32(CTRL_EXT);
6540 		mac_data &= 0xFFF7FFFF;
6541 		ew32(CTRL_EXT, mac_data);
6542 
6543 		/* Revert the lanphypc logic to use the internal Gbe counter
6544 		 * and not the PMC counter
6545 		 */
6546 		mac_data = er32(FEXTNVM5);
6547 		mac_data &= 0xFFFFFF7F;
6548 		ew32(FEXTNVM5, mac_data);
6549 
6550 		/* Enable the periodic inband message,
6551 		 * Request PCIe clock in K1 page770_17[10:9] =01b
6552 		 */
6553 		e1e_rphy(hw, HV_PM_CTRL, &phy_data);
6554 		phy_data &= 0xFBFF;
6555 		phy_data |= HV_PM_CTRL_K1_CLK_REQ;
6556 		e1e_wphy(hw, HV_PM_CTRL, phy_data);
6557 
6558 		/* Return back configuration
6559 		 * 772_29[5] = 0 CS_Mode_Stay_In_K1
6560 		 */
6561 		e1e_rphy(hw, I217_CGFREG, &phy_data);
6562 		phy_data &= 0xFFDF;
6563 		e1e_wphy(hw, I217_CGFREG, phy_data);
6564 
6565 		/* Change the MAC/PHY interface to Kumeran
6566 		 * Unforce the SMBus in PHY page769_23[0] = 0
6567 		 * Unforce the SMBus in MAC CTRL_EXT[11] = 0
6568 		 */
6569 		e1e_rphy(hw, CV_SMB_CTRL, &phy_data);
6570 		phy_data &= ~CV_SMB_CTRL_FORCE_SMBUS;
6571 		e1e_wphy(hw, CV_SMB_CTRL, phy_data);
6572 		mac_data = er32(CTRL_EXT);
6573 		mac_data &= ~E1000_CTRL_EXT_FORCE_SMBUS;
6574 		ew32(CTRL_EXT, mac_data);
6575 	}
6576 
6577 	/* Disable Dynamic Power Gating */
6578 	mac_data = er32(CTRL_EXT);
6579 	mac_data &= 0xFFFFFFF7;
6580 	ew32(CTRL_EXT, mac_data);
6581 
6582 	/* Enable the time synchronization clock */
6583 	mac_data = er32(FEXTNVM7);
6584 	mac_data &= ~BIT(31);
6585 	mac_data |= BIT(0);
6586 	ew32(FEXTNVM7, mac_data);
6587 }
6588 
6589 static int e1000e_pm_freeze(struct device *dev)
6590 {
6591 	struct net_device *netdev = dev_get_drvdata(dev);
6592 	struct e1000_adapter *adapter = netdev_priv(netdev);
6593 	bool present;
6594 
6595 	rtnl_lock();
6596 
6597 	present = netif_device_present(netdev);
6598 	netif_device_detach(netdev);
6599 
6600 	if (present && netif_running(netdev)) {
6601 		int count = E1000_CHECK_RESET_COUNT;
6602 
6603 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6604 			usleep_range(10000, 11000);
6605 
6606 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6607 
6608 		/* Quiesce the device without resetting the hardware */
6609 		e1000e_down(adapter, false);
6610 		e1000_free_irq(adapter);
6611 	}
6612 	rtnl_unlock();
6613 
6614 	e1000e_reset_interrupt_capability(adapter);
6615 
6616 	/* Allow time for pending master requests to run */
6617 	e1000e_disable_pcie_master(&adapter->hw);
6618 
6619 	return 0;
6620 }
6621 
6622 static int __e1000_shutdown(struct pci_dev *pdev, bool runtime)
6623 {
6624 	struct net_device *netdev = pci_get_drvdata(pdev);
6625 	struct e1000_adapter *adapter = netdev_priv(netdev);
6626 	struct e1000_hw *hw = &adapter->hw;
6627 	u32 ctrl, ctrl_ext, rctl, status, wufc;
6628 	int retval = 0;
6629 
6630 	/* Runtime suspend should only enable wakeup for link changes */
6631 	if (runtime)
6632 		wufc = E1000_WUFC_LNKC;
6633 	else if (device_may_wakeup(&pdev->dev))
6634 		wufc = adapter->wol;
6635 	else
6636 		wufc = 0;
6637 
6638 	status = er32(STATUS);
6639 	if (status & E1000_STATUS_LU)
6640 		wufc &= ~E1000_WUFC_LNKC;
6641 
6642 	if (wufc) {
6643 		e1000_setup_rctl(adapter);
6644 		e1000e_set_rx_mode(netdev);
6645 
6646 		/* turn on all-multi mode if wake on multicast is enabled */
6647 		if (wufc & E1000_WUFC_MC) {
6648 			rctl = er32(RCTL);
6649 			rctl |= E1000_RCTL_MPE;
6650 			ew32(RCTL, rctl);
6651 		}
6652 
6653 		ctrl = er32(CTRL);
6654 		ctrl |= E1000_CTRL_ADVD3WUC;
6655 		if (!(adapter->flags2 & FLAG2_HAS_PHY_WAKEUP))
6656 			ctrl |= E1000_CTRL_EN_PHY_PWR_MGMT;
6657 		ew32(CTRL, ctrl);
6658 
6659 		if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
6660 		    adapter->hw.phy.media_type ==
6661 		    e1000_media_type_internal_serdes) {
6662 			/* keep the laser running in D3 */
6663 			ctrl_ext = er32(CTRL_EXT);
6664 			ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
6665 			ew32(CTRL_EXT, ctrl_ext);
6666 		}
6667 
6668 		if (!runtime)
6669 			e1000e_power_up_phy(adapter);
6670 
6671 		if (adapter->flags & FLAG_IS_ICH)
6672 			e1000_suspend_workarounds_ich8lan(&adapter->hw);
6673 
6674 		if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6675 			/* enable wakeup by the PHY */
6676 			retval = e1000_init_phy_wakeup(adapter, wufc);
6677 			if (retval)
6678 				return retval;
6679 		} else {
6680 			/* enable wakeup by the MAC */
6681 			ew32(WUFC, wufc);
6682 			ew32(WUC, E1000_WUC_PME_EN);
6683 		}
6684 	} else {
6685 		ew32(WUC, 0);
6686 		ew32(WUFC, 0);
6687 
6688 		e1000_power_down_phy(adapter);
6689 	}
6690 
6691 	if (adapter->hw.phy.type == e1000_phy_igp_3) {
6692 		e1000e_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
6693 	} else if (hw->mac.type >= e1000_pch_lpt) {
6694 		if (wufc && !(wufc & (E1000_WUFC_EX | E1000_WUFC_MC | E1000_WUFC_BC)))
6695 			/* ULP does not support wake from unicast, multicast
6696 			 * or broadcast.
6697 			 */
6698 			retval = e1000_enable_ulp_lpt_lp(hw, !runtime);
6699 
6700 		if (retval)
6701 			return retval;
6702 	}
6703 
6704 	/* Ensure that the appropriate bits are set in LPI_CTRL
6705 	 * for EEE in Sx
6706 	 */
6707 	if ((hw->phy.type >= e1000_phy_i217) &&
6708 	    adapter->eee_advert && hw->dev_spec.ich8lan.eee_lp_ability) {
6709 		u16 lpi_ctrl = 0;
6710 
6711 		retval = hw->phy.ops.acquire(hw);
6712 		if (!retval) {
6713 			retval = e1e_rphy_locked(hw, I82579_LPI_CTRL,
6714 						 &lpi_ctrl);
6715 			if (!retval) {
6716 				if (adapter->eee_advert &
6717 				    hw->dev_spec.ich8lan.eee_lp_ability &
6718 				    I82579_EEE_100_SUPPORTED)
6719 					lpi_ctrl |= I82579_LPI_CTRL_100_ENABLE;
6720 				if (adapter->eee_advert &
6721 				    hw->dev_spec.ich8lan.eee_lp_ability &
6722 				    I82579_EEE_1000_SUPPORTED)
6723 					lpi_ctrl |= I82579_LPI_CTRL_1000_ENABLE;
6724 
6725 				retval = e1e_wphy_locked(hw, I82579_LPI_CTRL,
6726 							 lpi_ctrl);
6727 			}
6728 		}
6729 		hw->phy.ops.release(hw);
6730 	}
6731 
6732 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
6733 	 * would have already happened in close and is redundant.
6734 	 */
6735 	e1000e_release_hw_control(adapter);
6736 
6737 	pci_clear_master(pdev);
6738 
6739 	/* The pci-e switch on some quad port adapters will report a
6740 	 * correctable error when the MAC transitions from D0 to D3.  To
6741 	 * prevent this we need to mask off the correctable errors on the
6742 	 * downstream port of the pci-e switch.
6743 	 *
6744 	 * We don't have the associated upstream bridge while assigning
6745 	 * the PCI device into guest. For example, the KVM on power is
6746 	 * one of the cases.
6747 	 */
6748 	if (adapter->flags & FLAG_IS_QUAD_PORT) {
6749 		struct pci_dev *us_dev = pdev->bus->self;
6750 		u16 devctl;
6751 
6752 		if (!us_dev)
6753 			return 0;
6754 
6755 		pcie_capability_read_word(us_dev, PCI_EXP_DEVCTL, &devctl);
6756 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL,
6757 					   (devctl & ~PCI_EXP_DEVCTL_CERE));
6758 
6759 		pci_save_state(pdev);
6760 		pci_prepare_to_sleep(pdev);
6761 
6762 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL, devctl);
6763 	}
6764 
6765 	return 0;
6766 }
6767 
6768 /**
6769  * __e1000e_disable_aspm - Disable ASPM states
6770  * @pdev: pointer to PCI device struct
6771  * @state: bit-mask of ASPM states to disable
6772  * @locked: indication if this context holds pci_bus_sem locked.
6773  *
6774  * Some devices *must* have certain ASPM states disabled per hardware errata.
6775  **/
6776 static void __e1000e_disable_aspm(struct pci_dev *pdev, u16 state, int locked)
6777 {
6778 	struct pci_dev *parent = pdev->bus->self;
6779 	u16 aspm_dis_mask = 0;
6780 	u16 pdev_aspmc, parent_aspmc;
6781 
6782 	switch (state) {
6783 	case PCIE_LINK_STATE_L0S:
6784 	case PCIE_LINK_STATE_L0S | PCIE_LINK_STATE_L1:
6785 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L0S;
6786 		fallthrough; /* can't have L1 without L0s */
6787 	case PCIE_LINK_STATE_L1:
6788 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L1;
6789 		break;
6790 	default:
6791 		return;
6792 	}
6793 
6794 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6795 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6796 
6797 	if (parent) {
6798 		pcie_capability_read_word(parent, PCI_EXP_LNKCTL,
6799 					  &parent_aspmc);
6800 		parent_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6801 	}
6802 
6803 	/* Nothing to do if the ASPM states to be disabled already are */
6804 	if (!(pdev_aspmc & aspm_dis_mask) &&
6805 	    (!parent || !(parent_aspmc & aspm_dis_mask)))
6806 		return;
6807 
6808 	dev_info(&pdev->dev, "Disabling ASPM %s %s\n",
6809 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L0S) ?
6810 		 "L0s" : "",
6811 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L1) ?
6812 		 "L1" : "");
6813 
6814 #ifdef CONFIG_PCIEASPM
6815 	if (locked)
6816 		pci_disable_link_state_locked(pdev, state);
6817 	else
6818 		pci_disable_link_state(pdev, state);
6819 
6820 	/* Double-check ASPM control.  If not disabled by the above, the
6821 	 * BIOS is preventing that from happening (or CONFIG_PCIEASPM is
6822 	 * not enabled); override by writing PCI config space directly.
6823 	 */
6824 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6825 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6826 
6827 	if (!(aspm_dis_mask & pdev_aspmc))
6828 		return;
6829 #endif
6830 
6831 	/* Both device and parent should have the same ASPM setting.
6832 	 * Disable ASPM in downstream component first and then upstream.
6833 	 */
6834 	pcie_capability_clear_word(pdev, PCI_EXP_LNKCTL, aspm_dis_mask);
6835 
6836 	if (parent)
6837 		pcie_capability_clear_word(parent, PCI_EXP_LNKCTL,
6838 					   aspm_dis_mask);
6839 }
6840 
6841 /**
6842  * e1000e_disable_aspm - Disable ASPM states.
6843  * @pdev: pointer to PCI device struct
6844  * @state: bit-mask of ASPM states to disable
6845  *
6846  * This function acquires the pci_bus_sem!
6847  * Some devices *must* have certain ASPM states disabled per hardware errata.
6848  **/
6849 static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6850 {
6851 	__e1000e_disable_aspm(pdev, state, 0);
6852 }
6853 
6854 /**
6855  * e1000e_disable_aspm_locked - Disable ASPM states.
6856  * @pdev: pointer to PCI device struct
6857  * @state: bit-mask of ASPM states to disable
6858  *
6859  * This function must be called with pci_bus_sem acquired!
6860  * Some devices *must* have certain ASPM states disabled per hardware errata.
6861  **/
6862 static void e1000e_disable_aspm_locked(struct pci_dev *pdev, u16 state)
6863 {
6864 	__e1000e_disable_aspm(pdev, state, 1);
6865 }
6866 
6867 static int e1000e_pm_thaw(struct device *dev)
6868 {
6869 	struct net_device *netdev = dev_get_drvdata(dev);
6870 	struct e1000_adapter *adapter = netdev_priv(netdev);
6871 	int rc = 0;
6872 
6873 	e1000e_set_interrupt_capability(adapter);
6874 
6875 	rtnl_lock();
6876 	if (netif_running(netdev)) {
6877 		rc = e1000_request_irq(adapter);
6878 		if (rc)
6879 			goto err_irq;
6880 
6881 		e1000e_up(adapter);
6882 	}
6883 
6884 	netif_device_attach(netdev);
6885 err_irq:
6886 	rtnl_unlock();
6887 
6888 	return rc;
6889 }
6890 
6891 static int __e1000_resume(struct pci_dev *pdev)
6892 {
6893 	struct net_device *netdev = pci_get_drvdata(pdev);
6894 	struct e1000_adapter *adapter = netdev_priv(netdev);
6895 	struct e1000_hw *hw = &adapter->hw;
6896 	u16 aspm_disable_flag = 0;
6897 
6898 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6899 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6900 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6901 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6902 	if (aspm_disable_flag)
6903 		e1000e_disable_aspm(pdev, aspm_disable_flag);
6904 
6905 	pci_set_master(pdev);
6906 
6907 	if (hw->mac.type >= e1000_pch2lan)
6908 		e1000_resume_workarounds_pchlan(&adapter->hw);
6909 
6910 	e1000e_power_up_phy(adapter);
6911 
6912 	/* report the system wakeup cause from S3/S4 */
6913 	if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6914 		u16 phy_data;
6915 
6916 		e1e_rphy(&adapter->hw, BM_WUS, &phy_data);
6917 		if (phy_data) {
6918 			e_info("PHY Wakeup cause - %s\n",
6919 			       phy_data & E1000_WUS_EX ? "Unicast Packet" :
6920 			       phy_data & E1000_WUS_MC ? "Multicast Packet" :
6921 			       phy_data & E1000_WUS_BC ? "Broadcast Packet" :
6922 			       phy_data & E1000_WUS_MAG ? "Magic Packet" :
6923 			       phy_data & E1000_WUS_LNKC ?
6924 			       "Link Status Change" : "other");
6925 		}
6926 		e1e_wphy(&adapter->hw, BM_WUS, ~0);
6927 	} else {
6928 		u32 wus = er32(WUS);
6929 
6930 		if (wus) {
6931 			e_info("MAC Wakeup cause - %s\n",
6932 			       wus & E1000_WUS_EX ? "Unicast Packet" :
6933 			       wus & E1000_WUS_MC ? "Multicast Packet" :
6934 			       wus & E1000_WUS_BC ? "Broadcast Packet" :
6935 			       wus & E1000_WUS_MAG ? "Magic Packet" :
6936 			       wus & E1000_WUS_LNKC ? "Link Status Change" :
6937 			       "other");
6938 		}
6939 		ew32(WUS, ~0);
6940 	}
6941 
6942 	e1000e_reset(adapter);
6943 
6944 	e1000_init_manageability_pt(adapter);
6945 
6946 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6947 	 * is up.  For all other cases, let the f/w know that the h/w is now
6948 	 * under the control of the driver.
6949 	 */
6950 	if (!(adapter->flags & FLAG_HAS_AMT))
6951 		e1000e_get_hw_control(adapter);
6952 
6953 	return 0;
6954 }
6955 
6956 static __maybe_unused int e1000e_pm_prepare(struct device *dev)
6957 {
6958 	return pm_runtime_suspended(dev) &&
6959 		pm_suspend_via_firmware();
6960 }
6961 
6962 static __maybe_unused int e1000e_pm_suspend(struct device *dev)
6963 {
6964 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6965 	struct e1000_adapter *adapter = netdev_priv(netdev);
6966 	struct pci_dev *pdev = to_pci_dev(dev);
6967 	int rc;
6968 
6969 	e1000e_flush_lpic(pdev);
6970 
6971 	e1000e_pm_freeze(dev);
6972 
6973 	rc = __e1000_shutdown(pdev, false);
6974 	if (rc) {
6975 		e1000e_pm_thaw(dev);
6976 	} else {
6977 		/* Introduce S0ix implementation */
6978 		if (adapter->flags2 & FLAG2_ENABLE_S0IX_FLOWS)
6979 			e1000e_s0ix_entry_flow(adapter);
6980 	}
6981 
6982 	return rc;
6983 }
6984 
6985 static __maybe_unused int e1000e_pm_resume(struct device *dev)
6986 {
6987 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6988 	struct e1000_adapter *adapter = netdev_priv(netdev);
6989 	struct pci_dev *pdev = to_pci_dev(dev);
6990 	int rc;
6991 
6992 	/* Introduce S0ix implementation */
6993 	if (adapter->flags2 & FLAG2_ENABLE_S0IX_FLOWS)
6994 		e1000e_s0ix_exit_flow(adapter);
6995 
6996 	rc = __e1000_resume(pdev);
6997 	if (rc)
6998 		return rc;
6999 
7000 	return e1000e_pm_thaw(dev);
7001 }
7002 
7003 static __maybe_unused int e1000e_pm_runtime_idle(struct device *dev)
7004 {
7005 	struct net_device *netdev = dev_get_drvdata(dev);
7006 	struct e1000_adapter *adapter = netdev_priv(netdev);
7007 	u16 eee_lp;
7008 
7009 	eee_lp = adapter->hw.dev_spec.ich8lan.eee_lp_ability;
7010 
7011 	if (!e1000e_has_link(adapter)) {
7012 		adapter->hw.dev_spec.ich8lan.eee_lp_ability = eee_lp;
7013 		pm_schedule_suspend(dev, 5 * MSEC_PER_SEC);
7014 	}
7015 
7016 	return -EBUSY;
7017 }
7018 
7019 static __maybe_unused int e1000e_pm_runtime_resume(struct device *dev)
7020 {
7021 	struct pci_dev *pdev = to_pci_dev(dev);
7022 	struct net_device *netdev = pci_get_drvdata(pdev);
7023 	struct e1000_adapter *adapter = netdev_priv(netdev);
7024 	int rc;
7025 
7026 	pdev->pme_poll = true;
7027 
7028 	rc = __e1000_resume(pdev);
7029 	if (rc)
7030 		return rc;
7031 
7032 	if (netdev->flags & IFF_UP)
7033 		e1000e_up(adapter);
7034 
7035 	return rc;
7036 }
7037 
7038 static __maybe_unused int e1000e_pm_runtime_suspend(struct device *dev)
7039 {
7040 	struct pci_dev *pdev = to_pci_dev(dev);
7041 	struct net_device *netdev = pci_get_drvdata(pdev);
7042 	struct e1000_adapter *adapter = netdev_priv(netdev);
7043 
7044 	if (netdev->flags & IFF_UP) {
7045 		int count = E1000_CHECK_RESET_COUNT;
7046 
7047 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
7048 			usleep_range(10000, 11000);
7049 
7050 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
7051 
7052 		/* Down the device without resetting the hardware */
7053 		e1000e_down(adapter, false);
7054 	}
7055 
7056 	if (__e1000_shutdown(pdev, true)) {
7057 		e1000e_pm_runtime_resume(dev);
7058 		return -EBUSY;
7059 	}
7060 
7061 	return 0;
7062 }
7063 
7064 static void e1000_shutdown(struct pci_dev *pdev)
7065 {
7066 	e1000e_flush_lpic(pdev);
7067 
7068 	e1000e_pm_freeze(&pdev->dev);
7069 
7070 	__e1000_shutdown(pdev, false);
7071 }
7072 
7073 #ifdef CONFIG_NET_POLL_CONTROLLER
7074 
7075 static irqreturn_t e1000_intr_msix(int __always_unused irq, void *data)
7076 {
7077 	struct net_device *netdev = data;
7078 	struct e1000_adapter *adapter = netdev_priv(netdev);
7079 
7080 	if (adapter->msix_entries) {
7081 		int vector, msix_irq;
7082 
7083 		vector = 0;
7084 		msix_irq = adapter->msix_entries[vector].vector;
7085 		if (disable_hardirq(msix_irq))
7086 			e1000_intr_msix_rx(msix_irq, netdev);
7087 		enable_irq(msix_irq);
7088 
7089 		vector++;
7090 		msix_irq = adapter->msix_entries[vector].vector;
7091 		if (disable_hardirq(msix_irq))
7092 			e1000_intr_msix_tx(msix_irq, netdev);
7093 		enable_irq(msix_irq);
7094 
7095 		vector++;
7096 		msix_irq = adapter->msix_entries[vector].vector;
7097 		if (disable_hardirq(msix_irq))
7098 			e1000_msix_other(msix_irq, netdev);
7099 		enable_irq(msix_irq);
7100 	}
7101 
7102 	return IRQ_HANDLED;
7103 }
7104 
7105 /**
7106  * e1000_netpoll
7107  * @netdev: network interface device structure
7108  *
7109  * Polling 'interrupt' - used by things like netconsole to send skbs
7110  * without having to re-enable interrupts. It's not called while
7111  * the interrupt routine is executing.
7112  */
7113 static void e1000_netpoll(struct net_device *netdev)
7114 {
7115 	struct e1000_adapter *adapter = netdev_priv(netdev);
7116 
7117 	switch (adapter->int_mode) {
7118 	case E1000E_INT_MODE_MSIX:
7119 		e1000_intr_msix(adapter->pdev->irq, netdev);
7120 		break;
7121 	case E1000E_INT_MODE_MSI:
7122 		if (disable_hardirq(adapter->pdev->irq))
7123 			e1000_intr_msi(adapter->pdev->irq, netdev);
7124 		enable_irq(adapter->pdev->irq);
7125 		break;
7126 	default:		/* E1000E_INT_MODE_LEGACY */
7127 		if (disable_hardirq(adapter->pdev->irq))
7128 			e1000_intr(adapter->pdev->irq, netdev);
7129 		enable_irq(adapter->pdev->irq);
7130 		break;
7131 	}
7132 }
7133 #endif
7134 
7135 /**
7136  * e1000_io_error_detected - called when PCI error is detected
7137  * @pdev: Pointer to PCI device
7138  * @state: The current pci connection state
7139  *
7140  * This function is called after a PCI bus error affecting
7141  * this device has been detected.
7142  */
7143 static pci_ers_result_t e1000_io_error_detected(struct pci_dev *pdev,
7144 						pci_channel_state_t state)
7145 {
7146 	e1000e_pm_freeze(&pdev->dev);
7147 
7148 	if (state == pci_channel_io_perm_failure)
7149 		return PCI_ERS_RESULT_DISCONNECT;
7150 
7151 	pci_disable_device(pdev);
7152 
7153 	/* Request a slot reset. */
7154 	return PCI_ERS_RESULT_NEED_RESET;
7155 }
7156 
7157 /**
7158  * e1000_io_slot_reset - called after the pci bus has been reset.
7159  * @pdev: Pointer to PCI device
7160  *
7161  * Restart the card from scratch, as if from a cold-boot. Implementation
7162  * resembles the first-half of the e1000e_pm_resume routine.
7163  */
7164 static pci_ers_result_t e1000_io_slot_reset(struct pci_dev *pdev)
7165 {
7166 	struct net_device *netdev = pci_get_drvdata(pdev);
7167 	struct e1000_adapter *adapter = netdev_priv(netdev);
7168 	struct e1000_hw *hw = &adapter->hw;
7169 	u16 aspm_disable_flag = 0;
7170 	int err;
7171 	pci_ers_result_t result;
7172 
7173 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
7174 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7175 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
7176 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7177 	if (aspm_disable_flag)
7178 		e1000e_disable_aspm_locked(pdev, aspm_disable_flag);
7179 
7180 	err = pci_enable_device_mem(pdev);
7181 	if (err) {
7182 		dev_err(&pdev->dev,
7183 			"Cannot re-enable PCI device after reset.\n");
7184 		result = PCI_ERS_RESULT_DISCONNECT;
7185 	} else {
7186 		pdev->state_saved = true;
7187 		pci_restore_state(pdev);
7188 		pci_set_master(pdev);
7189 
7190 		pci_enable_wake(pdev, PCI_D3hot, 0);
7191 		pci_enable_wake(pdev, PCI_D3cold, 0);
7192 
7193 		e1000e_reset(adapter);
7194 		ew32(WUS, ~0);
7195 		result = PCI_ERS_RESULT_RECOVERED;
7196 	}
7197 
7198 	return result;
7199 }
7200 
7201 /**
7202  * e1000_io_resume - called when traffic can start flowing again.
7203  * @pdev: Pointer to PCI device
7204  *
7205  * This callback is called when the error recovery driver tells us that
7206  * its OK to resume normal operation. Implementation resembles the
7207  * second-half of the e1000e_pm_resume routine.
7208  */
7209 static void e1000_io_resume(struct pci_dev *pdev)
7210 {
7211 	struct net_device *netdev = pci_get_drvdata(pdev);
7212 	struct e1000_adapter *adapter = netdev_priv(netdev);
7213 
7214 	e1000_init_manageability_pt(adapter);
7215 
7216 	e1000e_pm_thaw(&pdev->dev);
7217 
7218 	/* If the controller has AMT, do not set DRV_LOAD until the interface
7219 	 * is up.  For all other cases, let the f/w know that the h/w is now
7220 	 * under the control of the driver.
7221 	 */
7222 	if (!(adapter->flags & FLAG_HAS_AMT))
7223 		e1000e_get_hw_control(adapter);
7224 }
7225 
7226 static void e1000_print_device_info(struct e1000_adapter *adapter)
7227 {
7228 	struct e1000_hw *hw = &adapter->hw;
7229 	struct net_device *netdev = adapter->netdev;
7230 	u32 ret_val;
7231 	u8 pba_str[E1000_PBANUM_LENGTH];
7232 
7233 	/* print bus type/speed/width info */
7234 	e_info("(PCI Express:2.5GT/s:%s) %pM\n",
7235 	       /* bus width */
7236 	       ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
7237 		"Width x1"),
7238 	       /* MAC address */
7239 	       netdev->dev_addr);
7240 	e_info("Intel(R) PRO/%s Network Connection\n",
7241 	       (hw->phy.type == e1000_phy_ife) ? "10/100" : "1000");
7242 	ret_val = e1000_read_pba_string_generic(hw, pba_str,
7243 						E1000_PBANUM_LENGTH);
7244 	if (ret_val)
7245 		strscpy((char *)pba_str, "Unknown", sizeof(pba_str));
7246 	e_info("MAC: %d, PHY: %d, PBA No: %s\n",
7247 	       hw->mac.type, hw->phy.type, pba_str);
7248 }
7249 
7250 static void e1000_eeprom_checks(struct e1000_adapter *adapter)
7251 {
7252 	struct e1000_hw *hw = &adapter->hw;
7253 	int ret_val;
7254 	u16 buf = 0;
7255 
7256 	if (hw->mac.type != e1000_82573)
7257 		return;
7258 
7259 	ret_val = e1000_read_nvm(hw, NVM_INIT_CONTROL2_REG, 1, &buf);
7260 	le16_to_cpus(&buf);
7261 	if (!ret_val && (!(buf & BIT(0)))) {
7262 		/* Deep Smart Power Down (DSPD) */
7263 		dev_warn(&adapter->pdev->dev,
7264 			 "Warning: detected DSPD enabled in EEPROM\n");
7265 	}
7266 }
7267 
7268 static netdev_features_t e1000_fix_features(struct net_device *netdev,
7269 					    netdev_features_t features)
7270 {
7271 	struct e1000_adapter *adapter = netdev_priv(netdev);
7272 	struct e1000_hw *hw = &adapter->hw;
7273 
7274 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
7275 	if ((hw->mac.type >= e1000_pch2lan) && (netdev->mtu > ETH_DATA_LEN))
7276 		features &= ~NETIF_F_RXFCS;
7277 
7278 	/* Since there is no support for separate Rx/Tx vlan accel
7279 	 * enable/disable make sure Tx flag is always in same state as Rx.
7280 	 */
7281 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
7282 		features |= NETIF_F_HW_VLAN_CTAG_TX;
7283 	else
7284 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
7285 
7286 	return features;
7287 }
7288 
7289 static int e1000_set_features(struct net_device *netdev,
7290 			      netdev_features_t features)
7291 {
7292 	struct e1000_adapter *adapter = netdev_priv(netdev);
7293 	netdev_features_t changed = features ^ netdev->features;
7294 
7295 	if (changed & (NETIF_F_TSO | NETIF_F_TSO6))
7296 		adapter->flags |= FLAG_TSO_FORCE;
7297 
7298 	if (!(changed & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX |
7299 			 NETIF_F_RXCSUM | NETIF_F_RXHASH | NETIF_F_RXFCS |
7300 			 NETIF_F_RXALL)))
7301 		return 0;
7302 
7303 	if (changed & NETIF_F_RXFCS) {
7304 		if (features & NETIF_F_RXFCS) {
7305 			adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7306 		} else {
7307 			/* We need to take it back to defaults, which might mean
7308 			 * stripping is still disabled at the adapter level.
7309 			 */
7310 			if (adapter->flags2 & FLAG2_DFLT_CRC_STRIPPING)
7311 				adapter->flags2 |= FLAG2_CRC_STRIPPING;
7312 			else
7313 				adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7314 		}
7315 	}
7316 
7317 	netdev->features = features;
7318 
7319 	if (netif_running(netdev))
7320 		e1000e_reinit_locked(adapter);
7321 	else
7322 		e1000e_reset(adapter);
7323 
7324 	return 1;
7325 }
7326 
7327 static const struct net_device_ops e1000e_netdev_ops = {
7328 	.ndo_open		= e1000e_open,
7329 	.ndo_stop		= e1000e_close,
7330 	.ndo_start_xmit		= e1000_xmit_frame,
7331 	.ndo_get_stats64	= e1000e_get_stats64,
7332 	.ndo_set_rx_mode	= e1000e_set_rx_mode,
7333 	.ndo_set_mac_address	= e1000_set_mac,
7334 	.ndo_change_mtu		= e1000_change_mtu,
7335 	.ndo_eth_ioctl		= e1000_ioctl,
7336 	.ndo_tx_timeout		= e1000_tx_timeout,
7337 	.ndo_validate_addr	= eth_validate_addr,
7338 
7339 	.ndo_vlan_rx_add_vid	= e1000_vlan_rx_add_vid,
7340 	.ndo_vlan_rx_kill_vid	= e1000_vlan_rx_kill_vid,
7341 #ifdef CONFIG_NET_POLL_CONTROLLER
7342 	.ndo_poll_controller	= e1000_netpoll,
7343 #endif
7344 	.ndo_set_features = e1000_set_features,
7345 	.ndo_fix_features = e1000_fix_features,
7346 	.ndo_features_check	= passthru_features_check,
7347 };
7348 
7349 /**
7350  * e1000_probe - Device Initialization Routine
7351  * @pdev: PCI device information struct
7352  * @ent: entry in e1000_pci_tbl
7353  *
7354  * Returns 0 on success, negative on failure
7355  *
7356  * e1000_probe initializes an adapter identified by a pci_dev structure.
7357  * The OS initialization, configuring of the adapter private structure,
7358  * and a hardware reset occur.
7359  **/
7360 static int e1000_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7361 {
7362 	struct net_device *netdev;
7363 	struct e1000_adapter *adapter;
7364 	struct e1000_hw *hw;
7365 	const struct e1000_info *ei = e1000_info_tbl[ent->driver_data];
7366 	resource_size_t mmio_start, mmio_len;
7367 	resource_size_t flash_start, flash_len;
7368 	static int cards_found;
7369 	u16 aspm_disable_flag = 0;
7370 	u16 eeprom_data = 0;
7371 	u16 eeprom_apme_mask = E1000_EEPROM_APME;
7372 	int bars, i, err;
7373 	s32 ret_val = 0;
7374 
7375 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L0S)
7376 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7377 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L1)
7378 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7379 	if (aspm_disable_flag)
7380 		e1000e_disable_aspm(pdev, aspm_disable_flag);
7381 
7382 	err = pci_enable_device_mem(pdev);
7383 	if (err)
7384 		return err;
7385 
7386 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
7387 	if (err) {
7388 		dev_err(&pdev->dev,
7389 			"No usable DMA configuration, aborting\n");
7390 		goto err_dma;
7391 	}
7392 
7393 	bars = pci_select_bars(pdev, IORESOURCE_MEM);
7394 	err = pci_request_selected_regions_exclusive(pdev, bars,
7395 						     e1000e_driver_name);
7396 	if (err)
7397 		goto err_pci_reg;
7398 
7399 	pci_set_master(pdev);
7400 	/* PCI config space info */
7401 	err = pci_save_state(pdev);
7402 	if (err)
7403 		goto err_alloc_etherdev;
7404 
7405 	err = -ENOMEM;
7406 	netdev = alloc_etherdev(sizeof(struct e1000_adapter));
7407 	if (!netdev)
7408 		goto err_alloc_etherdev;
7409 
7410 	SET_NETDEV_DEV(netdev, &pdev->dev);
7411 
7412 	netdev->irq = pdev->irq;
7413 
7414 	pci_set_drvdata(pdev, netdev);
7415 	adapter = netdev_priv(netdev);
7416 	hw = &adapter->hw;
7417 	adapter->netdev = netdev;
7418 	adapter->pdev = pdev;
7419 	adapter->ei = ei;
7420 	adapter->pba = ei->pba;
7421 	adapter->flags = ei->flags;
7422 	adapter->flags2 = ei->flags2;
7423 	adapter->hw.adapter = adapter;
7424 	adapter->hw.mac.type = ei->mac;
7425 	adapter->max_hw_frame_size = ei->max_hw_frame_size;
7426 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
7427 
7428 	mmio_start = pci_resource_start(pdev, 0);
7429 	mmio_len = pci_resource_len(pdev, 0);
7430 
7431 	err = -EIO;
7432 	adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
7433 	if (!adapter->hw.hw_addr)
7434 		goto err_ioremap;
7435 
7436 	if ((adapter->flags & FLAG_HAS_FLASH) &&
7437 	    (pci_resource_flags(pdev, 1) & IORESOURCE_MEM) &&
7438 	    (hw->mac.type < e1000_pch_spt)) {
7439 		flash_start = pci_resource_start(pdev, 1);
7440 		flash_len = pci_resource_len(pdev, 1);
7441 		adapter->hw.flash_address = ioremap(flash_start, flash_len);
7442 		if (!adapter->hw.flash_address)
7443 			goto err_flashmap;
7444 	}
7445 
7446 	/* Set default EEE advertisement */
7447 	if (adapter->flags2 & FLAG2_HAS_EEE)
7448 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
7449 
7450 	/* construct the net_device struct */
7451 	netdev->netdev_ops = &e1000e_netdev_ops;
7452 	e1000e_set_ethtool_ops(netdev);
7453 	netdev->watchdog_timeo = 5 * HZ;
7454 	netif_napi_add(netdev, &adapter->napi, e1000e_poll);
7455 	strscpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
7456 
7457 	netdev->mem_start = mmio_start;
7458 	netdev->mem_end = mmio_start + mmio_len;
7459 
7460 	adapter->bd_number = cards_found++;
7461 
7462 	e1000e_check_options(adapter);
7463 
7464 	/* setup adapter struct */
7465 	err = e1000_sw_init(adapter);
7466 	if (err)
7467 		goto err_sw_init;
7468 
7469 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
7470 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
7471 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
7472 
7473 	err = ei->get_variants(adapter);
7474 	if (err)
7475 		goto err_hw_init;
7476 
7477 	if ((adapter->flags & FLAG_IS_ICH) &&
7478 	    (adapter->flags & FLAG_READ_ONLY_NVM) &&
7479 	    (hw->mac.type < e1000_pch_spt))
7480 		e1000e_write_protect_nvm_ich8lan(&adapter->hw);
7481 
7482 	hw->mac.ops.get_bus_info(&adapter->hw);
7483 
7484 	adapter->hw.phy.autoneg_wait_to_complete = 0;
7485 
7486 	/* Copper options */
7487 	if (adapter->hw.phy.media_type == e1000_media_type_copper) {
7488 		adapter->hw.phy.mdix = AUTO_ALL_MODES;
7489 		adapter->hw.phy.disable_polarity_correction = 0;
7490 		adapter->hw.phy.ms_type = e1000_ms_hw_default;
7491 	}
7492 
7493 	if (hw->phy.ops.check_reset_block && hw->phy.ops.check_reset_block(hw))
7494 		dev_info(&pdev->dev,
7495 			 "PHY reset is blocked due to SOL/IDER session.\n");
7496 
7497 	/* Set initial default active device features */
7498 	netdev->features = (NETIF_F_SG |
7499 			    NETIF_F_HW_VLAN_CTAG_RX |
7500 			    NETIF_F_HW_VLAN_CTAG_TX |
7501 			    NETIF_F_TSO |
7502 			    NETIF_F_TSO6 |
7503 			    NETIF_F_RXHASH |
7504 			    NETIF_F_RXCSUM |
7505 			    NETIF_F_HW_CSUM);
7506 
7507 	/* disable TSO for pcie and 10/100 speeds to avoid
7508 	 * some hardware issues and for i219 to fix transfer
7509 	 * speed being capped at 60%
7510 	 */
7511 	if (!(adapter->flags & FLAG_TSO_FORCE)) {
7512 		switch (adapter->link_speed) {
7513 		case SPEED_10:
7514 		case SPEED_100:
7515 			e_info("10/100 speed: disabling TSO\n");
7516 			netdev->features &= ~NETIF_F_TSO;
7517 			netdev->features &= ~NETIF_F_TSO6;
7518 			break;
7519 		case SPEED_1000:
7520 			netdev->features |= NETIF_F_TSO;
7521 			netdev->features |= NETIF_F_TSO6;
7522 			break;
7523 		default:
7524 			/* oops */
7525 			break;
7526 		}
7527 		if (hw->mac.type == e1000_pch_spt) {
7528 			netdev->features &= ~NETIF_F_TSO;
7529 			netdev->features &= ~NETIF_F_TSO6;
7530 		}
7531 	}
7532 
7533 	/* Set user-changeable features (subset of all device features) */
7534 	netdev->hw_features = netdev->features;
7535 	netdev->hw_features |= NETIF_F_RXFCS;
7536 	netdev->priv_flags |= IFF_SUPP_NOFCS;
7537 	netdev->hw_features |= NETIF_F_RXALL;
7538 
7539 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER)
7540 		netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
7541 
7542 	netdev->vlan_features |= (NETIF_F_SG |
7543 				  NETIF_F_TSO |
7544 				  NETIF_F_TSO6 |
7545 				  NETIF_F_HW_CSUM);
7546 
7547 	netdev->priv_flags |= IFF_UNICAST_FLT;
7548 
7549 	netdev->features |= NETIF_F_HIGHDMA;
7550 	netdev->vlan_features |= NETIF_F_HIGHDMA;
7551 
7552 	/* MTU range: 68 - max_hw_frame_size */
7553 	netdev->min_mtu = ETH_MIN_MTU;
7554 	netdev->max_mtu = adapter->max_hw_frame_size -
7555 			  (VLAN_ETH_HLEN + ETH_FCS_LEN);
7556 
7557 	if (e1000e_enable_mng_pass_thru(&adapter->hw))
7558 		adapter->flags |= FLAG_MNG_PT_ENABLED;
7559 
7560 	/* before reading the NVM, reset the controller to
7561 	 * put the device in a known good starting state
7562 	 */
7563 	adapter->hw.mac.ops.reset_hw(&adapter->hw);
7564 
7565 	/* systems with ASPM and others may see the checksum fail on the first
7566 	 * attempt. Let's give it a few tries
7567 	 */
7568 	for (i = 0;; i++) {
7569 		if (e1000_validate_nvm_checksum(&adapter->hw) >= 0)
7570 			break;
7571 		if (i == 2) {
7572 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
7573 			err = -EIO;
7574 			goto err_eeprom;
7575 		}
7576 	}
7577 
7578 	e1000_eeprom_checks(adapter);
7579 
7580 	/* copy the MAC address */
7581 	if (e1000e_read_mac_addr(&adapter->hw))
7582 		dev_err(&pdev->dev,
7583 			"NVM Read Error while reading MAC address\n");
7584 
7585 	eth_hw_addr_set(netdev, adapter->hw.mac.addr);
7586 
7587 	if (!is_valid_ether_addr(netdev->dev_addr)) {
7588 		dev_err(&pdev->dev, "Invalid MAC Address: %pM\n",
7589 			netdev->dev_addr);
7590 		err = -EIO;
7591 		goto err_eeprom;
7592 	}
7593 
7594 	timer_setup(&adapter->watchdog_timer, e1000_watchdog, 0);
7595 	timer_setup(&adapter->phy_info_timer, e1000_update_phy_info, 0);
7596 
7597 	INIT_WORK(&adapter->reset_task, e1000_reset_task);
7598 	INIT_WORK(&adapter->watchdog_task, e1000_watchdog_task);
7599 	INIT_WORK(&adapter->downshift_task, e1000e_downshift_workaround);
7600 	INIT_WORK(&adapter->update_phy_task, e1000e_update_phy_task);
7601 	INIT_WORK(&adapter->print_hang_task, e1000_print_hw_hang);
7602 
7603 	/* Initialize link parameters. User can change them with ethtool */
7604 	adapter->hw.mac.autoneg = 1;
7605 	adapter->fc_autoneg = true;
7606 	adapter->hw.fc.requested_mode = e1000_fc_default;
7607 	adapter->hw.fc.current_mode = e1000_fc_default;
7608 	adapter->hw.phy.autoneg_advertised = 0x2f;
7609 
7610 	/* Initial Wake on LAN setting - If APM wake is enabled in
7611 	 * the EEPROM, enable the ACPI Magic Packet filter
7612 	 */
7613 	if (adapter->flags & FLAG_APME_IN_WUC) {
7614 		/* APME bit in EEPROM is mapped to WUC.APME */
7615 		eeprom_data = er32(WUC);
7616 		eeprom_apme_mask = E1000_WUC_APME;
7617 		if ((hw->mac.type > e1000_ich10lan) &&
7618 		    (eeprom_data & E1000_WUC_PHY_WAKE))
7619 			adapter->flags2 |= FLAG2_HAS_PHY_WAKEUP;
7620 	} else if (adapter->flags & FLAG_APME_IN_CTRL3) {
7621 		if (adapter->flags & FLAG_APME_CHECK_PORT_B &&
7622 		    (adapter->hw.bus.func == 1))
7623 			ret_val = e1000_read_nvm(&adapter->hw,
7624 					      NVM_INIT_CONTROL3_PORT_B,
7625 					      1, &eeprom_data);
7626 		else
7627 			ret_val = e1000_read_nvm(&adapter->hw,
7628 					      NVM_INIT_CONTROL3_PORT_A,
7629 					      1, &eeprom_data);
7630 	}
7631 
7632 	/* fetch WoL from EEPROM */
7633 	if (ret_val)
7634 		e_dbg("NVM read error getting WoL initial values: %d\n", ret_val);
7635 	else if (eeprom_data & eeprom_apme_mask)
7636 		adapter->eeprom_wol |= E1000_WUFC_MAG;
7637 
7638 	/* now that we have the eeprom settings, apply the special cases
7639 	 * where the eeprom may be wrong or the board simply won't support
7640 	 * wake on lan on a particular port
7641 	 */
7642 	if (!(adapter->flags & FLAG_HAS_WOL))
7643 		adapter->eeprom_wol = 0;
7644 
7645 	/* initialize the wol settings based on the eeprom settings */
7646 	adapter->wol = adapter->eeprom_wol;
7647 
7648 	/* make sure adapter isn't asleep if manageability is enabled */
7649 	if (adapter->wol || (adapter->flags & FLAG_MNG_PT_ENABLED) ||
7650 	    (hw->mac.ops.check_mng_mode(hw)))
7651 		device_wakeup_enable(&pdev->dev);
7652 
7653 	/* save off EEPROM version number */
7654 	ret_val = e1000_read_nvm(&adapter->hw, 5, 1, &adapter->eeprom_vers);
7655 
7656 	if (ret_val) {
7657 		e_dbg("NVM read error getting EEPROM version: %d\n", ret_val);
7658 		adapter->eeprom_vers = 0;
7659 	}
7660 
7661 	/* init PTP hardware clock */
7662 	e1000e_ptp_init(adapter);
7663 
7664 	/* reset the hardware with the new settings */
7665 	e1000e_reset(adapter);
7666 
7667 	/* If the controller has AMT, do not set DRV_LOAD until the interface
7668 	 * is up.  For all other cases, let the f/w know that the h/w is now
7669 	 * under the control of the driver.
7670 	 */
7671 	if (!(adapter->flags & FLAG_HAS_AMT))
7672 		e1000e_get_hw_control(adapter);
7673 
7674 	if (hw->mac.type >= e1000_pch_cnp)
7675 		adapter->flags2 |= FLAG2_ENABLE_S0IX_FLOWS;
7676 
7677 	strscpy(netdev->name, "eth%d", sizeof(netdev->name));
7678 	err = register_netdev(netdev);
7679 	if (err)
7680 		goto err_register;
7681 
7682 	/* carrier off reporting is important to ethtool even BEFORE open */
7683 	netif_carrier_off(netdev);
7684 
7685 	e1000_print_device_info(adapter);
7686 
7687 	dev_pm_set_driver_flags(&pdev->dev, DPM_FLAG_SMART_PREPARE);
7688 
7689 	if (pci_dev_run_wake(pdev))
7690 		pm_runtime_put_noidle(&pdev->dev);
7691 
7692 	return 0;
7693 
7694 err_register:
7695 	if (!(adapter->flags & FLAG_HAS_AMT))
7696 		e1000e_release_hw_control(adapter);
7697 err_eeprom:
7698 	if (hw->phy.ops.check_reset_block && !hw->phy.ops.check_reset_block(hw))
7699 		e1000_phy_hw_reset(&adapter->hw);
7700 err_hw_init:
7701 	kfree(adapter->tx_ring);
7702 	kfree(adapter->rx_ring);
7703 err_sw_init:
7704 	if ((adapter->hw.flash_address) && (hw->mac.type < e1000_pch_spt))
7705 		iounmap(adapter->hw.flash_address);
7706 	e1000e_reset_interrupt_capability(adapter);
7707 err_flashmap:
7708 	iounmap(adapter->hw.hw_addr);
7709 err_ioremap:
7710 	free_netdev(netdev);
7711 err_alloc_etherdev:
7712 	pci_release_mem_regions(pdev);
7713 err_pci_reg:
7714 err_dma:
7715 	pci_disable_device(pdev);
7716 	return err;
7717 }
7718 
7719 /**
7720  * e1000_remove - Device Removal Routine
7721  * @pdev: PCI device information struct
7722  *
7723  * e1000_remove is called by the PCI subsystem to alert the driver
7724  * that it should release a PCI device.  This could be caused by a
7725  * Hot-Plug event, or because the driver is going to be removed from
7726  * memory.
7727  **/
7728 static void e1000_remove(struct pci_dev *pdev)
7729 {
7730 	struct net_device *netdev = pci_get_drvdata(pdev);
7731 	struct e1000_adapter *adapter = netdev_priv(netdev);
7732 
7733 	e1000e_ptp_remove(adapter);
7734 
7735 	/* The timers may be rescheduled, so explicitly disable them
7736 	 * from being rescheduled.
7737 	 */
7738 	set_bit(__E1000_DOWN, &adapter->state);
7739 	del_timer_sync(&adapter->watchdog_timer);
7740 	del_timer_sync(&adapter->phy_info_timer);
7741 
7742 	cancel_work_sync(&adapter->reset_task);
7743 	cancel_work_sync(&adapter->watchdog_task);
7744 	cancel_work_sync(&adapter->downshift_task);
7745 	cancel_work_sync(&adapter->update_phy_task);
7746 	cancel_work_sync(&adapter->print_hang_task);
7747 
7748 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
7749 		cancel_work_sync(&adapter->tx_hwtstamp_work);
7750 		if (adapter->tx_hwtstamp_skb) {
7751 			dev_consume_skb_any(adapter->tx_hwtstamp_skb);
7752 			adapter->tx_hwtstamp_skb = NULL;
7753 		}
7754 	}
7755 
7756 	unregister_netdev(netdev);
7757 
7758 	if (pci_dev_run_wake(pdev))
7759 		pm_runtime_get_noresume(&pdev->dev);
7760 
7761 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
7762 	 * would have already happened in close and is redundant.
7763 	 */
7764 	e1000e_release_hw_control(adapter);
7765 
7766 	e1000e_reset_interrupt_capability(adapter);
7767 	kfree(adapter->tx_ring);
7768 	kfree(adapter->rx_ring);
7769 
7770 	iounmap(adapter->hw.hw_addr);
7771 	if ((adapter->hw.flash_address) &&
7772 	    (adapter->hw.mac.type < e1000_pch_spt))
7773 		iounmap(adapter->hw.flash_address);
7774 	pci_release_mem_regions(pdev);
7775 
7776 	free_netdev(netdev);
7777 
7778 	pci_disable_device(pdev);
7779 }
7780 
7781 /* PCI Error Recovery (ERS) */
7782 static const struct pci_error_handlers e1000_err_handler = {
7783 	.error_detected = e1000_io_error_detected,
7784 	.slot_reset = e1000_io_slot_reset,
7785 	.resume = e1000_io_resume,
7786 };
7787 
7788 static const struct pci_device_id e1000_pci_tbl[] = {
7789 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_COPPER), board_82571 },
7790 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_FIBER), board_82571 },
7791 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER), board_82571 },
7792 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER_LP),
7793 	  board_82571 },
7794 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_FIBER), board_82571 },
7795 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES), board_82571 },
7796 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_DUAL), board_82571 },
7797 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_QUAD), board_82571 },
7798 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571PT_QUAD_COPPER), board_82571 },
7799 
7800 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI), board_82572 },
7801 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_COPPER), board_82572 },
7802 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_FIBER), board_82572 },
7803 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_SERDES), board_82572 },
7804 
7805 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E), board_82573 },
7806 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E_IAMT), board_82573 },
7807 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573L), board_82573 },
7808 
7809 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574L), board_82574 },
7810 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574LA), board_82574 },
7811 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82583V), board_82583 },
7812 
7813 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_DPT),
7814 	  board_80003es2lan },
7815 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_SPT),
7816 	  board_80003es2lan },
7817 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_DPT),
7818 	  board_80003es2lan },
7819 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_SPT),
7820 	  board_80003es2lan },
7821 
7822 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE), board_ich8lan },
7823 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_G), board_ich8lan },
7824 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_GT), board_ich8lan },
7825 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_AMT), board_ich8lan },
7826 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_C), board_ich8lan },
7827 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M), board_ich8lan },
7828 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M_AMT), board_ich8lan },
7829 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_82567V_3), board_ich8lan },
7830 
7831 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE), board_ich9lan },
7832 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_G), board_ich9lan },
7833 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_GT), board_ich9lan },
7834 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_AMT), board_ich9lan },
7835 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_C), board_ich9lan },
7836 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_BM), board_ich9lan },
7837 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M), board_ich9lan },
7838 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_AMT), board_ich9lan },
7839 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_V), board_ich9lan },
7840 
7841 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LM), board_ich9lan },
7842 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LF), board_ich9lan },
7843 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_V), board_ich9lan },
7844 
7845 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LM), board_ich10lan },
7846 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LF), board_ich10lan },
7847 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_V), board_ich10lan },
7848 
7849 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LM), board_pchlan },
7850 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LC), board_pchlan },
7851 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DM), board_pchlan },
7852 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DC), board_pchlan },
7853 
7854 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_LM), board_pch2lan },
7855 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_V), board_pch2lan },
7856 
7857 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_LM), board_pch_lpt },
7858 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_V), board_pch_lpt },
7859 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_LM), board_pch_lpt },
7860 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_V), board_pch_lpt },
7861 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM2), board_pch_lpt },
7862 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V2), board_pch_lpt },
7863 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM3), board_pch_lpt },
7864 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V3), board_pch_lpt },
7865 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM), board_pch_spt },
7866 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V), board_pch_spt },
7867 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM2), board_pch_spt },
7868 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V2), board_pch_spt },
7869 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LBG_I219_LM3), board_pch_spt },
7870 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM4), board_pch_spt },
7871 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V4), board_pch_spt },
7872 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM5), board_pch_spt },
7873 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V5), board_pch_spt },
7874 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM6), board_pch_cnp },
7875 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V6), board_pch_cnp },
7876 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM7), board_pch_cnp },
7877 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V7), board_pch_cnp },
7878 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM8), board_pch_cnp },
7879 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V8), board_pch_cnp },
7880 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM9), board_pch_cnp },
7881 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V9), board_pch_cnp },
7882 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM10), board_pch_cnp },
7883 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V10), board_pch_cnp },
7884 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM11), board_pch_cnp },
7885 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V11), board_pch_cnp },
7886 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM12), board_pch_spt },
7887 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V12), board_pch_spt },
7888 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM13), board_pch_tgp },
7889 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V13), board_pch_tgp },
7890 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM14), board_pch_tgp },
7891 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V14), board_pch_tgp },
7892 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM15), board_pch_tgp },
7893 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V15), board_pch_tgp },
7894 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_LM23), board_pch_adp },
7895 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_V23), board_pch_adp },
7896 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM16), board_pch_adp },
7897 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V16), board_pch_adp },
7898 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM17), board_pch_adp },
7899 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V17), board_pch_adp },
7900 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_LM22), board_pch_adp },
7901 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_V22), board_pch_adp },
7902 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_LM18), board_pch_mtp },
7903 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_V18), board_pch_mtp },
7904 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_LM19), board_pch_mtp },
7905 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_V19), board_pch_mtp },
7906 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_LM20), board_pch_mtp },
7907 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_V20), board_pch_mtp },
7908 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_LM21), board_pch_mtp },
7909 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_V21), board_pch_mtp },
7910 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ARL_I219_LM24), board_pch_mtp },
7911 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ARL_I219_V24), board_pch_mtp },
7912 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_LM25), board_pch_mtp },
7913 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_V25), board_pch_mtp },
7914 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_LM26), board_pch_mtp },
7915 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_V26), board_pch_mtp },
7916 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_LM27), board_pch_mtp },
7917 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_V27), board_pch_mtp },
7918 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_NVL_I219_LM29), board_pch_mtp },
7919 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_NVL_I219_V29), board_pch_mtp },
7920 
7921 	{ 0, 0, 0, 0, 0, 0, 0 }	/* terminate list */
7922 };
7923 MODULE_DEVICE_TABLE(pci, e1000_pci_tbl);
7924 
7925 static const struct dev_pm_ops e1000_pm_ops = {
7926 #ifdef CONFIG_PM_SLEEP
7927 	.prepare	= e1000e_pm_prepare,
7928 	.suspend	= e1000e_pm_suspend,
7929 	.resume		= e1000e_pm_resume,
7930 	.freeze		= e1000e_pm_freeze,
7931 	.thaw		= e1000e_pm_thaw,
7932 	.poweroff	= e1000e_pm_suspend,
7933 	.restore	= e1000e_pm_resume,
7934 #endif
7935 	SET_RUNTIME_PM_OPS(e1000e_pm_runtime_suspend, e1000e_pm_runtime_resume,
7936 			   e1000e_pm_runtime_idle)
7937 };
7938 
7939 /* PCI Device API Driver */
7940 static struct pci_driver e1000_driver = {
7941 	.name     = e1000e_driver_name,
7942 	.id_table = e1000_pci_tbl,
7943 	.probe    = e1000_probe,
7944 	.remove   = e1000_remove,
7945 	.driver   = {
7946 		.pm = &e1000_pm_ops,
7947 	},
7948 	.shutdown = e1000_shutdown,
7949 	.err_handler = &e1000_err_handler
7950 };
7951 
7952 /**
7953  * e1000_init_module - Driver Registration Routine
7954  *
7955  * e1000_init_module is the first routine called when the driver is
7956  * loaded. All it does is register with the PCI subsystem.
7957  **/
7958 static int __init e1000_init_module(void)
7959 {
7960 	pr_info("Intel(R) PRO/1000 Network Driver\n");
7961 	pr_info("Copyright(c) 1999 - 2015 Intel Corporation.\n");
7962 
7963 	return pci_register_driver(&e1000_driver);
7964 }
7965 module_init(e1000_init_module);
7966 
7967 /**
7968  * e1000_exit_module - Driver Exit Cleanup Routine
7969  *
7970  * e1000_exit_module is called just before the driver is removed
7971  * from memory.
7972  **/
7973 static void __exit e1000_exit_module(void)
7974 {
7975 	pci_unregister_driver(&e1000_driver);
7976 }
7977 module_exit(e1000_exit_module);
7978 
7979 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
7980 MODULE_DESCRIPTION("Intel(R) PRO/1000 Network Driver");
7981 MODULE_LICENSE("GPL v2");
7982 
7983 /* netdev.c */
7984