1 /* Intel PRO/1000 Linux driver
2  * Copyright(c) 1999 - 2015 Intel Corporation.
3  *
4  * This program is free software; you can redistribute it and/or modify it
5  * under the terms and conditions of the GNU General Public License,
6  * version 2, as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope it will be useful, but WITHOUT
9  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
11  * more details.
12  *
13  * The full GNU General Public License is included in this distribution in
14  * the file called "COPYING".
15  *
16  * Contact Information:
17  * Linux NICS <linux.nics@intel.com>
18  * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
19  * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
20  */
21 
22 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
23 
24 #include <linux/module.h>
25 #include <linux/types.h>
26 #include <linux/init.h>
27 #include <linux/pci.h>
28 #include <linux/vmalloc.h>
29 #include <linux/pagemap.h>
30 #include <linux/delay.h>
31 #include <linux/netdevice.h>
32 #include <linux/interrupt.h>
33 #include <linux/tcp.h>
34 #include <linux/ipv6.h>
35 #include <linux/slab.h>
36 #include <net/checksum.h>
37 #include <net/ip6_checksum.h>
38 #include <linux/ethtool.h>
39 #include <linux/if_vlan.h>
40 #include <linux/cpu.h>
41 #include <linux/smp.h>
42 #include <linux/pm_qos.h>
43 #include <linux/pm_runtime.h>
44 #include <linux/aer.h>
45 #include <linux/prefetch.h>
46 
47 #include "e1000.h"
48 
49 #define DRV_EXTRAVERSION "-k"
50 
51 #define DRV_VERSION "3.2.6" DRV_EXTRAVERSION
52 char e1000e_driver_name[] = "e1000e";
53 const char e1000e_driver_version[] = DRV_VERSION;
54 
55 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
56 static int debug = -1;
57 module_param(debug, int, 0);
58 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
59 
60 static const struct e1000_info *e1000_info_tbl[] = {
61 	[board_82571]		= &e1000_82571_info,
62 	[board_82572]		= &e1000_82572_info,
63 	[board_82573]		= &e1000_82573_info,
64 	[board_82574]		= &e1000_82574_info,
65 	[board_82583]		= &e1000_82583_info,
66 	[board_80003es2lan]	= &e1000_es2_info,
67 	[board_ich8lan]		= &e1000_ich8_info,
68 	[board_ich9lan]		= &e1000_ich9_info,
69 	[board_ich10lan]	= &e1000_ich10_info,
70 	[board_pchlan]		= &e1000_pch_info,
71 	[board_pch2lan]		= &e1000_pch2_info,
72 	[board_pch_lpt]		= &e1000_pch_lpt_info,
73 	[board_pch_spt]		= &e1000_pch_spt_info,
74 	[board_pch_cnp]		= &e1000_pch_cnp_info,
75 };
76 
77 struct e1000_reg_info {
78 	u32 ofs;
79 	char *name;
80 };
81 
82 static const struct e1000_reg_info e1000_reg_info_tbl[] = {
83 	/* General Registers */
84 	{E1000_CTRL, "CTRL"},
85 	{E1000_STATUS, "STATUS"},
86 	{E1000_CTRL_EXT, "CTRL_EXT"},
87 
88 	/* Interrupt Registers */
89 	{E1000_ICR, "ICR"},
90 
91 	/* Rx Registers */
92 	{E1000_RCTL, "RCTL"},
93 	{E1000_RDLEN(0), "RDLEN"},
94 	{E1000_RDH(0), "RDH"},
95 	{E1000_RDT(0), "RDT"},
96 	{E1000_RDTR, "RDTR"},
97 	{E1000_RXDCTL(0), "RXDCTL"},
98 	{E1000_ERT, "ERT"},
99 	{E1000_RDBAL(0), "RDBAL"},
100 	{E1000_RDBAH(0), "RDBAH"},
101 	{E1000_RDFH, "RDFH"},
102 	{E1000_RDFT, "RDFT"},
103 	{E1000_RDFHS, "RDFHS"},
104 	{E1000_RDFTS, "RDFTS"},
105 	{E1000_RDFPC, "RDFPC"},
106 
107 	/* Tx Registers */
108 	{E1000_TCTL, "TCTL"},
109 	{E1000_TDBAL(0), "TDBAL"},
110 	{E1000_TDBAH(0), "TDBAH"},
111 	{E1000_TDLEN(0), "TDLEN"},
112 	{E1000_TDH(0), "TDH"},
113 	{E1000_TDT(0), "TDT"},
114 	{E1000_TIDV, "TIDV"},
115 	{E1000_TXDCTL(0), "TXDCTL"},
116 	{E1000_TADV, "TADV"},
117 	{E1000_TARC(0), "TARC"},
118 	{E1000_TDFH, "TDFH"},
119 	{E1000_TDFT, "TDFT"},
120 	{E1000_TDFHS, "TDFHS"},
121 	{E1000_TDFTS, "TDFTS"},
122 	{E1000_TDFPC, "TDFPC"},
123 
124 	/* List Terminator */
125 	{0, NULL}
126 };
127 
128 /**
129  * __ew32_prepare - prepare to write to MAC CSR register on certain parts
130  * @hw: pointer to the HW structure
131  *
132  * When updating the MAC CSR registers, the Manageability Engine (ME) could
133  * be accessing the registers at the same time.  Normally, this is handled in
134  * h/w by an arbiter but on some parts there is a bug that acknowledges Host
135  * accesses later than it should which could result in the register to have
136  * an incorrect value.  Workaround this by checking the FWSM register which
137  * has bit 24 set while ME is accessing MAC CSR registers, wait if it is set
138  * and try again a number of times.
139  **/
140 s32 __ew32_prepare(struct e1000_hw *hw)
141 {
142 	s32 i = E1000_ICH_FWSM_PCIM2PCI_COUNT;
143 
144 	while ((er32(FWSM) & E1000_ICH_FWSM_PCIM2PCI) && --i)
145 		udelay(50);
146 
147 	return i;
148 }
149 
150 void __ew32(struct e1000_hw *hw, unsigned long reg, u32 val)
151 {
152 	if (hw->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
153 		__ew32_prepare(hw);
154 
155 	writel(val, hw->hw_addr + reg);
156 }
157 
158 /**
159  * e1000_regdump - register printout routine
160  * @hw: pointer to the HW structure
161  * @reginfo: pointer to the register info table
162  **/
163 static void e1000_regdump(struct e1000_hw *hw, struct e1000_reg_info *reginfo)
164 {
165 	int n = 0;
166 	char rname[16];
167 	u32 regs[8];
168 
169 	switch (reginfo->ofs) {
170 	case E1000_RXDCTL(0):
171 		for (n = 0; n < 2; n++)
172 			regs[n] = __er32(hw, E1000_RXDCTL(n));
173 		break;
174 	case E1000_TXDCTL(0):
175 		for (n = 0; n < 2; n++)
176 			regs[n] = __er32(hw, E1000_TXDCTL(n));
177 		break;
178 	case E1000_TARC(0):
179 		for (n = 0; n < 2; n++)
180 			regs[n] = __er32(hw, E1000_TARC(n));
181 		break;
182 	default:
183 		pr_info("%-15s %08x\n",
184 			reginfo->name, __er32(hw, reginfo->ofs));
185 		return;
186 	}
187 
188 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-1]");
189 	pr_info("%-15s %08x %08x\n", rname, regs[0], regs[1]);
190 }
191 
192 static void e1000e_dump_ps_pages(struct e1000_adapter *adapter,
193 				 struct e1000_buffer *bi)
194 {
195 	int i;
196 	struct e1000_ps_page *ps_page;
197 
198 	for (i = 0; i < adapter->rx_ps_pages; i++) {
199 		ps_page = &bi->ps_pages[i];
200 
201 		if (ps_page->page) {
202 			pr_info("packet dump for ps_page %d:\n", i);
203 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
204 				       16, 1, page_address(ps_page->page),
205 				       PAGE_SIZE, true);
206 		}
207 	}
208 }
209 
210 /**
211  * e1000e_dump - Print registers, Tx-ring and Rx-ring
212  * @adapter: board private structure
213  **/
214 static void e1000e_dump(struct e1000_adapter *adapter)
215 {
216 	struct net_device *netdev = adapter->netdev;
217 	struct e1000_hw *hw = &adapter->hw;
218 	struct e1000_reg_info *reginfo;
219 	struct e1000_ring *tx_ring = adapter->tx_ring;
220 	struct e1000_tx_desc *tx_desc;
221 	struct my_u0 {
222 		__le64 a;
223 		__le64 b;
224 	} *u0;
225 	struct e1000_buffer *buffer_info;
226 	struct e1000_ring *rx_ring = adapter->rx_ring;
227 	union e1000_rx_desc_packet_split *rx_desc_ps;
228 	union e1000_rx_desc_extended *rx_desc;
229 	struct my_u1 {
230 		__le64 a;
231 		__le64 b;
232 		__le64 c;
233 		__le64 d;
234 	} *u1;
235 	u32 staterr;
236 	int i = 0;
237 
238 	if (!netif_msg_hw(adapter))
239 		return;
240 
241 	/* Print netdevice Info */
242 	if (netdev) {
243 		dev_info(&adapter->pdev->dev, "Net device Info\n");
244 		pr_info("Device Name     state            trans_start\n");
245 		pr_info("%-15s %016lX %016lX\n", netdev->name,
246 			netdev->state, dev_trans_start(netdev));
247 	}
248 
249 	/* Print Registers */
250 	dev_info(&adapter->pdev->dev, "Register Dump\n");
251 	pr_info(" Register Name   Value\n");
252 	for (reginfo = (struct e1000_reg_info *)e1000_reg_info_tbl;
253 	     reginfo->name; reginfo++) {
254 		e1000_regdump(hw, reginfo);
255 	}
256 
257 	/* Print Tx Ring Summary */
258 	if (!netdev || !netif_running(netdev))
259 		return;
260 
261 	dev_info(&adapter->pdev->dev, "Tx Ring Summary\n");
262 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
263 	buffer_info = &tx_ring->buffer_info[tx_ring->next_to_clean];
264 	pr_info(" %5d %5X %5X %016llX %04X %3X %016llX\n",
265 		0, tx_ring->next_to_use, tx_ring->next_to_clean,
266 		(unsigned long long)buffer_info->dma,
267 		buffer_info->length,
268 		buffer_info->next_to_watch,
269 		(unsigned long long)buffer_info->time_stamp);
270 
271 	/* Print Tx Ring */
272 	if (!netif_msg_tx_done(adapter))
273 		goto rx_ring_summary;
274 
275 	dev_info(&adapter->pdev->dev, "Tx Ring Dump\n");
276 
277 	/* Transmit Descriptor Formats - DEXT[29] is 0 (Legacy) or 1 (Extended)
278 	 *
279 	 * Legacy Transmit Descriptor
280 	 *   +--------------------------------------------------------------+
281 	 * 0 |         Buffer Address [63:0] (Reserved on Write Back)       |
282 	 *   +--------------------------------------------------------------+
283 	 * 8 | Special  |    CSS     | Status |  CMD    |  CSO   |  Length  |
284 	 *   +--------------------------------------------------------------+
285 	 *   63       48 47        36 35    32 31     24 23    16 15        0
286 	 *
287 	 * Extended Context Descriptor (DTYP=0x0) for TSO or checksum offload
288 	 *   63      48 47    40 39       32 31             16 15    8 7      0
289 	 *   +----------------------------------------------------------------+
290 	 * 0 |  TUCSE  | TUCS0  |   TUCSS   |     IPCSE       | IPCS0 | IPCSS |
291 	 *   +----------------------------------------------------------------+
292 	 * 8 |   MSS   | HDRLEN | RSV | STA | TUCMD | DTYP |      PAYLEN      |
293 	 *   +----------------------------------------------------------------+
294 	 *   63      48 47    40 39 36 35 32 31   24 23  20 19                0
295 	 *
296 	 * Extended Data Descriptor (DTYP=0x1)
297 	 *   +----------------------------------------------------------------+
298 	 * 0 |                     Buffer Address [63:0]                      |
299 	 *   +----------------------------------------------------------------+
300 	 * 8 | VLAN tag |  POPTS  | Rsvd | Status | Command | DTYP |  DTALEN  |
301 	 *   +----------------------------------------------------------------+
302 	 *   63       48 47     40 39  36 35    32 31     24 23  20 19        0
303 	 */
304 	pr_info("Tl[desc]     [address 63:0  ] [SpeCssSCmCsLen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Legacy format\n");
305 	pr_info("Tc[desc]     [Ce CoCsIpceCoS] [MssHlRSCm0Plen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Context format\n");
306 	pr_info("Td[desc]     [address 63:0  ] [VlaPoRSCm1Dlen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Data format\n");
307 	for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
308 		const char *next_desc;
309 		tx_desc = E1000_TX_DESC(*tx_ring, i);
310 		buffer_info = &tx_ring->buffer_info[i];
311 		u0 = (struct my_u0 *)tx_desc;
312 		if (i == tx_ring->next_to_use && i == tx_ring->next_to_clean)
313 			next_desc = " NTC/U";
314 		else if (i == tx_ring->next_to_use)
315 			next_desc = " NTU";
316 		else if (i == tx_ring->next_to_clean)
317 			next_desc = " NTC";
318 		else
319 			next_desc = "";
320 		pr_info("T%c[0x%03X]    %016llX %016llX %016llX %04X  %3X %016llX %p%s\n",
321 			(!(le64_to_cpu(u0->b) & BIT(29)) ? 'l' :
322 			 ((le64_to_cpu(u0->b) & BIT(20)) ? 'd' : 'c')),
323 			i,
324 			(unsigned long long)le64_to_cpu(u0->a),
325 			(unsigned long long)le64_to_cpu(u0->b),
326 			(unsigned long long)buffer_info->dma,
327 			buffer_info->length, buffer_info->next_to_watch,
328 			(unsigned long long)buffer_info->time_stamp,
329 			buffer_info->skb, next_desc);
330 
331 		if (netif_msg_pktdata(adapter) && buffer_info->skb)
332 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
333 				       16, 1, buffer_info->skb->data,
334 				       buffer_info->skb->len, true);
335 	}
336 
337 	/* Print Rx Ring Summary */
338 rx_ring_summary:
339 	dev_info(&adapter->pdev->dev, "Rx Ring Summary\n");
340 	pr_info("Queue [NTU] [NTC]\n");
341 	pr_info(" %5d %5X %5X\n",
342 		0, rx_ring->next_to_use, rx_ring->next_to_clean);
343 
344 	/* Print Rx Ring */
345 	if (!netif_msg_rx_status(adapter))
346 		return;
347 
348 	dev_info(&adapter->pdev->dev, "Rx Ring Dump\n");
349 	switch (adapter->rx_ps_pages) {
350 	case 1:
351 	case 2:
352 	case 3:
353 		/* [Extended] Packet Split Receive Descriptor Format
354 		 *
355 		 *    +-----------------------------------------------------+
356 		 *  0 |                Buffer Address 0 [63:0]              |
357 		 *    +-----------------------------------------------------+
358 		 *  8 |                Buffer Address 1 [63:0]              |
359 		 *    +-----------------------------------------------------+
360 		 * 16 |                Buffer Address 2 [63:0]              |
361 		 *    +-----------------------------------------------------+
362 		 * 24 |                Buffer Address 3 [63:0]              |
363 		 *    +-----------------------------------------------------+
364 		 */
365 		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");
366 		/* [Extended] Receive Descriptor (Write-Back) Format
367 		 *
368 		 *   63       48 47    32 31     13 12    8 7    4 3        0
369 		 *   +------------------------------------------------------+
370 		 * 0 | Packet   | IP     |  Rsvd   | MRQ   | Rsvd | MRQ RSS |
371 		 *   | Checksum | Ident  |         | Queue |      |  Type   |
372 		 *   +------------------------------------------------------+
373 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
374 		 *   +------------------------------------------------------+
375 		 *   63       48 47    32 31            20 19               0
376 		 */
377 		pr_info("RWB[desc]      [ck ipid mrqhsh] [vl   l0 ee  es] [ l3  l2  l1 hs] [reserved      ] ---------------- [bi->skb] <-- Ext Rx Write-Back format\n");
378 		for (i = 0; i < rx_ring->count; i++) {
379 			const char *next_desc;
380 			buffer_info = &rx_ring->buffer_info[i];
381 			rx_desc_ps = E1000_RX_DESC_PS(*rx_ring, i);
382 			u1 = (struct my_u1 *)rx_desc_ps;
383 			staterr =
384 			    le32_to_cpu(rx_desc_ps->wb.middle.status_error);
385 
386 			if (i == rx_ring->next_to_use)
387 				next_desc = " NTU";
388 			else if (i == rx_ring->next_to_clean)
389 				next_desc = " NTC";
390 			else
391 				next_desc = "";
392 
393 			if (staterr & E1000_RXD_STAT_DD) {
394 				/* Descriptor Done */
395 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX ---------------- %p%s\n",
396 					"RWB", i,
397 					(unsigned long long)le64_to_cpu(u1->a),
398 					(unsigned long long)le64_to_cpu(u1->b),
399 					(unsigned long long)le64_to_cpu(u1->c),
400 					(unsigned long long)le64_to_cpu(u1->d),
401 					buffer_info->skb, next_desc);
402 			} else {
403 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX %016llX %p%s\n",
404 					"R  ", i,
405 					(unsigned long long)le64_to_cpu(u1->a),
406 					(unsigned long long)le64_to_cpu(u1->b),
407 					(unsigned long long)le64_to_cpu(u1->c),
408 					(unsigned long long)le64_to_cpu(u1->d),
409 					(unsigned long long)buffer_info->dma,
410 					buffer_info->skb, next_desc);
411 
412 				if (netif_msg_pktdata(adapter))
413 					e1000e_dump_ps_pages(adapter,
414 							     buffer_info);
415 			}
416 		}
417 		break;
418 	default:
419 	case 0:
420 		/* Extended Receive Descriptor (Read) Format
421 		 *
422 		 *   +-----------------------------------------------------+
423 		 * 0 |                Buffer Address [63:0]                |
424 		 *   +-----------------------------------------------------+
425 		 * 8 |                      Reserved                       |
426 		 *   +-----------------------------------------------------+
427 		 */
428 		pr_info("R  [desc]      [buf addr 63:0 ] [reserved 63:0 ] [bi->dma       ] [bi->skb] <-- Ext (Read) format\n");
429 		/* Extended Receive Descriptor (Write-Back) Format
430 		 *
431 		 *   63       48 47    32 31    24 23            4 3        0
432 		 *   +------------------------------------------------------+
433 		 *   |     RSS Hash      |        |               |         |
434 		 * 0 +-------------------+  Rsvd  |   Reserved    | MRQ RSS |
435 		 *   | Packet   | IP     |        |               |  Type   |
436 		 *   | Checksum | Ident  |        |               |         |
437 		 *   +------------------------------------------------------+
438 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
439 		 *   +------------------------------------------------------+
440 		 *   63       48 47    32 31            20 19               0
441 		 */
442 		pr_info("RWB[desc]      [cs ipid    mrq] [vt   ln xe  xs] [bi->skb] <-- Ext (Write-Back) format\n");
443 
444 		for (i = 0; i < rx_ring->count; i++) {
445 			const char *next_desc;
446 
447 			buffer_info = &rx_ring->buffer_info[i];
448 			rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
449 			u1 = (struct my_u1 *)rx_desc;
450 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
451 
452 			if (i == rx_ring->next_to_use)
453 				next_desc = " NTU";
454 			else if (i == rx_ring->next_to_clean)
455 				next_desc = " NTC";
456 			else
457 				next_desc = "";
458 
459 			if (staterr & E1000_RXD_STAT_DD) {
460 				/* Descriptor Done */
461 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %p%s\n",
462 					"RWB", i,
463 					(unsigned long long)le64_to_cpu(u1->a),
464 					(unsigned long long)le64_to_cpu(u1->b),
465 					buffer_info->skb, next_desc);
466 			} else {
467 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %p%s\n",
468 					"R  ", i,
469 					(unsigned long long)le64_to_cpu(u1->a),
470 					(unsigned long long)le64_to_cpu(u1->b),
471 					(unsigned long long)buffer_info->dma,
472 					buffer_info->skb, next_desc);
473 
474 				if (netif_msg_pktdata(adapter) &&
475 				    buffer_info->skb)
476 					print_hex_dump(KERN_INFO, "",
477 						       DUMP_PREFIX_ADDRESS, 16,
478 						       1,
479 						       buffer_info->skb->data,
480 						       adapter->rx_buffer_len,
481 						       true);
482 			}
483 		}
484 	}
485 }
486 
487 /**
488  * e1000_desc_unused - calculate if we have unused descriptors
489  **/
490 static int e1000_desc_unused(struct e1000_ring *ring)
491 {
492 	if (ring->next_to_clean > ring->next_to_use)
493 		return ring->next_to_clean - ring->next_to_use - 1;
494 
495 	return ring->count + ring->next_to_clean - ring->next_to_use - 1;
496 }
497 
498 /**
499  * e1000e_systim_to_hwtstamp - convert system time value to hw time stamp
500  * @adapter: board private structure
501  * @hwtstamps: time stamp structure to update
502  * @systim: unsigned 64bit system time value.
503  *
504  * Convert the system time value stored in the RX/TXSTMP registers into a
505  * hwtstamp which can be used by the upper level time stamping functions.
506  *
507  * The 'systim_lock' spinlock is used to protect the consistency of the
508  * system time value. This is needed because reading the 64 bit time
509  * value involves reading two 32 bit registers. The first read latches the
510  * value.
511  **/
512 static void e1000e_systim_to_hwtstamp(struct e1000_adapter *adapter,
513 				      struct skb_shared_hwtstamps *hwtstamps,
514 				      u64 systim)
515 {
516 	u64 ns;
517 	unsigned long flags;
518 
519 	spin_lock_irqsave(&adapter->systim_lock, flags);
520 	ns = timecounter_cyc2time(&adapter->tc, systim);
521 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
522 
523 	memset(hwtstamps, 0, sizeof(*hwtstamps));
524 	hwtstamps->hwtstamp = ns_to_ktime(ns);
525 }
526 
527 /**
528  * e1000e_rx_hwtstamp - utility function which checks for Rx time stamp
529  * @adapter: board private structure
530  * @status: descriptor extended error and status field
531  * @skb: particular skb to include time stamp
532  *
533  * If the time stamp is valid, convert it into the timecounter ns value
534  * and store that result into the shhwtstamps structure which is passed
535  * up the network stack.
536  **/
537 static void e1000e_rx_hwtstamp(struct e1000_adapter *adapter, u32 status,
538 			       struct sk_buff *skb)
539 {
540 	struct e1000_hw *hw = &adapter->hw;
541 	u64 rxstmp;
542 
543 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP) ||
544 	    !(status & E1000_RXDEXT_STATERR_TST) ||
545 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID))
546 		return;
547 
548 	/* The Rx time stamp registers contain the time stamp.  No other
549 	 * received packet will be time stamped until the Rx time stamp
550 	 * registers are read.  Because only one packet can be time stamped
551 	 * at a time, the register values must belong to this packet and
552 	 * therefore none of the other additional attributes need to be
553 	 * compared.
554 	 */
555 	rxstmp = (u64)er32(RXSTMPL);
556 	rxstmp |= (u64)er32(RXSTMPH) << 32;
557 	e1000e_systim_to_hwtstamp(adapter, skb_hwtstamps(skb), rxstmp);
558 
559 	adapter->flags2 &= ~FLAG2_CHECK_RX_HWTSTAMP;
560 }
561 
562 /**
563  * e1000_receive_skb - helper function to handle Rx indications
564  * @adapter: board private structure
565  * @staterr: descriptor extended error and status field as written by hardware
566  * @vlan: descriptor vlan field as written by hardware (no le/be conversion)
567  * @skb: pointer to sk_buff to be indicated to stack
568  **/
569 static void e1000_receive_skb(struct e1000_adapter *adapter,
570 			      struct net_device *netdev, struct sk_buff *skb,
571 			      u32 staterr, __le16 vlan)
572 {
573 	u16 tag = le16_to_cpu(vlan);
574 
575 	e1000e_rx_hwtstamp(adapter, staterr, skb);
576 
577 	skb->protocol = eth_type_trans(skb, netdev);
578 
579 	if (staterr & E1000_RXD_STAT_VP)
580 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), tag);
581 
582 	napi_gro_receive(&adapter->napi, skb);
583 }
584 
585 /**
586  * e1000_rx_checksum - Receive Checksum Offload
587  * @adapter: board private structure
588  * @status_err: receive descriptor status and error fields
589  * @csum: receive descriptor csum field
590  * @sk_buff: socket buffer with received data
591  **/
592 static void e1000_rx_checksum(struct e1000_adapter *adapter, u32 status_err,
593 			      struct sk_buff *skb)
594 {
595 	u16 status = (u16)status_err;
596 	u8 errors = (u8)(status_err >> 24);
597 
598 	skb_checksum_none_assert(skb);
599 
600 	/* Rx checksum disabled */
601 	if (!(adapter->netdev->features & NETIF_F_RXCSUM))
602 		return;
603 
604 	/* Ignore Checksum bit is set */
605 	if (status & E1000_RXD_STAT_IXSM)
606 		return;
607 
608 	/* TCP/UDP checksum error bit or IP checksum error bit is set */
609 	if (errors & (E1000_RXD_ERR_TCPE | E1000_RXD_ERR_IPE)) {
610 		/* let the stack verify checksum errors */
611 		adapter->hw_csum_err++;
612 		return;
613 	}
614 
615 	/* TCP/UDP Checksum has not been calculated */
616 	if (!(status & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)))
617 		return;
618 
619 	/* It must be a TCP or UDP packet with a valid checksum */
620 	skb->ip_summed = CHECKSUM_UNNECESSARY;
621 	adapter->hw_csum_good++;
622 }
623 
624 static void e1000e_update_rdt_wa(struct e1000_ring *rx_ring, unsigned int i)
625 {
626 	struct e1000_adapter *adapter = rx_ring->adapter;
627 	struct e1000_hw *hw = &adapter->hw;
628 	s32 ret_val = __ew32_prepare(hw);
629 
630 	writel(i, rx_ring->tail);
631 
632 	if (unlikely(!ret_val && (i != readl(rx_ring->tail)))) {
633 		u32 rctl = er32(RCTL);
634 
635 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
636 		e_err("ME firmware caused invalid RDT - resetting\n");
637 		schedule_work(&adapter->reset_task);
638 	}
639 }
640 
641 static void e1000e_update_tdt_wa(struct e1000_ring *tx_ring, unsigned int i)
642 {
643 	struct e1000_adapter *adapter = tx_ring->adapter;
644 	struct e1000_hw *hw = &adapter->hw;
645 	s32 ret_val = __ew32_prepare(hw);
646 
647 	writel(i, tx_ring->tail);
648 
649 	if (unlikely(!ret_val && (i != readl(tx_ring->tail)))) {
650 		u32 tctl = er32(TCTL);
651 
652 		ew32(TCTL, tctl & ~E1000_TCTL_EN);
653 		e_err("ME firmware caused invalid TDT - resetting\n");
654 		schedule_work(&adapter->reset_task);
655 	}
656 }
657 
658 /**
659  * e1000_alloc_rx_buffers - Replace used receive buffers
660  * @rx_ring: Rx descriptor ring
661  **/
662 static void e1000_alloc_rx_buffers(struct e1000_ring *rx_ring,
663 				   int cleaned_count, gfp_t gfp)
664 {
665 	struct e1000_adapter *adapter = rx_ring->adapter;
666 	struct net_device *netdev = adapter->netdev;
667 	struct pci_dev *pdev = adapter->pdev;
668 	union e1000_rx_desc_extended *rx_desc;
669 	struct e1000_buffer *buffer_info;
670 	struct sk_buff *skb;
671 	unsigned int i;
672 	unsigned int bufsz = adapter->rx_buffer_len;
673 
674 	i = rx_ring->next_to_use;
675 	buffer_info = &rx_ring->buffer_info[i];
676 
677 	while (cleaned_count--) {
678 		skb = buffer_info->skb;
679 		if (skb) {
680 			skb_trim(skb, 0);
681 			goto map_skb;
682 		}
683 
684 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
685 		if (!skb) {
686 			/* Better luck next round */
687 			adapter->alloc_rx_buff_failed++;
688 			break;
689 		}
690 
691 		buffer_info->skb = skb;
692 map_skb:
693 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
694 						  adapter->rx_buffer_len,
695 						  DMA_FROM_DEVICE);
696 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
697 			dev_err(&pdev->dev, "Rx DMA map failed\n");
698 			adapter->rx_dma_failed++;
699 			break;
700 		}
701 
702 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
703 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
704 
705 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
706 			/* Force memory writes to complete before letting h/w
707 			 * know there are new descriptors to fetch.  (Only
708 			 * applicable for weak-ordered memory model archs,
709 			 * such as IA-64).
710 			 */
711 			wmb();
712 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
713 				e1000e_update_rdt_wa(rx_ring, i);
714 			else
715 				writel(i, rx_ring->tail);
716 		}
717 		i++;
718 		if (i == rx_ring->count)
719 			i = 0;
720 		buffer_info = &rx_ring->buffer_info[i];
721 	}
722 
723 	rx_ring->next_to_use = i;
724 }
725 
726 /**
727  * e1000_alloc_rx_buffers_ps - Replace used receive buffers; packet split
728  * @rx_ring: Rx descriptor ring
729  **/
730 static void e1000_alloc_rx_buffers_ps(struct e1000_ring *rx_ring,
731 				      int cleaned_count, gfp_t gfp)
732 {
733 	struct e1000_adapter *adapter = rx_ring->adapter;
734 	struct net_device *netdev = adapter->netdev;
735 	struct pci_dev *pdev = adapter->pdev;
736 	union e1000_rx_desc_packet_split *rx_desc;
737 	struct e1000_buffer *buffer_info;
738 	struct e1000_ps_page *ps_page;
739 	struct sk_buff *skb;
740 	unsigned int i, j;
741 
742 	i = rx_ring->next_to_use;
743 	buffer_info = &rx_ring->buffer_info[i];
744 
745 	while (cleaned_count--) {
746 		rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
747 
748 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
749 			ps_page = &buffer_info->ps_pages[j];
750 			if (j >= adapter->rx_ps_pages) {
751 				/* all unused desc entries get hw null ptr */
752 				rx_desc->read.buffer_addr[j + 1] =
753 				    ~cpu_to_le64(0);
754 				continue;
755 			}
756 			if (!ps_page->page) {
757 				ps_page->page = alloc_page(gfp);
758 				if (!ps_page->page) {
759 					adapter->alloc_rx_buff_failed++;
760 					goto no_buffers;
761 				}
762 				ps_page->dma = dma_map_page(&pdev->dev,
763 							    ps_page->page,
764 							    0, PAGE_SIZE,
765 							    DMA_FROM_DEVICE);
766 				if (dma_mapping_error(&pdev->dev,
767 						      ps_page->dma)) {
768 					dev_err(&adapter->pdev->dev,
769 						"Rx DMA page map failed\n");
770 					adapter->rx_dma_failed++;
771 					goto no_buffers;
772 				}
773 			}
774 			/* Refresh the desc even if buffer_addrs
775 			 * didn't change because each write-back
776 			 * erases this info.
777 			 */
778 			rx_desc->read.buffer_addr[j + 1] =
779 			    cpu_to_le64(ps_page->dma);
780 		}
781 
782 		skb = __netdev_alloc_skb_ip_align(netdev, adapter->rx_ps_bsize0,
783 						  gfp);
784 
785 		if (!skb) {
786 			adapter->alloc_rx_buff_failed++;
787 			break;
788 		}
789 
790 		buffer_info->skb = skb;
791 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
792 						  adapter->rx_ps_bsize0,
793 						  DMA_FROM_DEVICE);
794 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
795 			dev_err(&pdev->dev, "Rx DMA map failed\n");
796 			adapter->rx_dma_failed++;
797 			/* cleanup skb */
798 			dev_kfree_skb_any(skb);
799 			buffer_info->skb = NULL;
800 			break;
801 		}
802 
803 		rx_desc->read.buffer_addr[0] = cpu_to_le64(buffer_info->dma);
804 
805 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
806 			/* Force memory writes to complete before letting h/w
807 			 * know there are new descriptors to fetch.  (Only
808 			 * applicable for weak-ordered memory model archs,
809 			 * such as IA-64).
810 			 */
811 			wmb();
812 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
813 				e1000e_update_rdt_wa(rx_ring, i << 1);
814 			else
815 				writel(i << 1, rx_ring->tail);
816 		}
817 
818 		i++;
819 		if (i == rx_ring->count)
820 			i = 0;
821 		buffer_info = &rx_ring->buffer_info[i];
822 	}
823 
824 no_buffers:
825 	rx_ring->next_to_use = i;
826 }
827 
828 /**
829  * e1000_alloc_jumbo_rx_buffers - Replace used jumbo receive buffers
830  * @rx_ring: Rx descriptor ring
831  * @cleaned_count: number of buffers to allocate this pass
832  **/
833 
834 static void e1000_alloc_jumbo_rx_buffers(struct e1000_ring *rx_ring,
835 					 int cleaned_count, gfp_t gfp)
836 {
837 	struct e1000_adapter *adapter = rx_ring->adapter;
838 	struct net_device *netdev = adapter->netdev;
839 	struct pci_dev *pdev = adapter->pdev;
840 	union e1000_rx_desc_extended *rx_desc;
841 	struct e1000_buffer *buffer_info;
842 	struct sk_buff *skb;
843 	unsigned int i;
844 	unsigned int bufsz = 256 - 16;	/* for skb_reserve */
845 
846 	i = rx_ring->next_to_use;
847 	buffer_info = &rx_ring->buffer_info[i];
848 
849 	while (cleaned_count--) {
850 		skb = buffer_info->skb;
851 		if (skb) {
852 			skb_trim(skb, 0);
853 			goto check_page;
854 		}
855 
856 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
857 		if (unlikely(!skb)) {
858 			/* Better luck next round */
859 			adapter->alloc_rx_buff_failed++;
860 			break;
861 		}
862 
863 		buffer_info->skb = skb;
864 check_page:
865 		/* allocate a new page if necessary */
866 		if (!buffer_info->page) {
867 			buffer_info->page = alloc_page(gfp);
868 			if (unlikely(!buffer_info->page)) {
869 				adapter->alloc_rx_buff_failed++;
870 				break;
871 			}
872 		}
873 
874 		if (!buffer_info->dma) {
875 			buffer_info->dma = dma_map_page(&pdev->dev,
876 							buffer_info->page, 0,
877 							PAGE_SIZE,
878 							DMA_FROM_DEVICE);
879 			if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
880 				adapter->alloc_rx_buff_failed++;
881 				break;
882 			}
883 		}
884 
885 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
886 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
887 
888 		if (unlikely(++i == rx_ring->count))
889 			i = 0;
890 		buffer_info = &rx_ring->buffer_info[i];
891 	}
892 
893 	if (likely(rx_ring->next_to_use != i)) {
894 		rx_ring->next_to_use = i;
895 		if (unlikely(i-- == 0))
896 			i = (rx_ring->count - 1);
897 
898 		/* Force memory writes to complete before letting h/w
899 		 * know there are new descriptors to fetch.  (Only
900 		 * applicable for weak-ordered memory model archs,
901 		 * such as IA-64).
902 		 */
903 		wmb();
904 		if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
905 			e1000e_update_rdt_wa(rx_ring, i);
906 		else
907 			writel(i, rx_ring->tail);
908 	}
909 }
910 
911 static inline void e1000_rx_hash(struct net_device *netdev, __le32 rss,
912 				 struct sk_buff *skb)
913 {
914 	if (netdev->features & NETIF_F_RXHASH)
915 		skb_set_hash(skb, le32_to_cpu(rss), PKT_HASH_TYPE_L3);
916 }
917 
918 /**
919  * e1000_clean_rx_irq - Send received data up the network stack
920  * @rx_ring: Rx descriptor ring
921  *
922  * the return value indicates whether actual cleaning was done, there
923  * is no guarantee that everything was cleaned
924  **/
925 static bool e1000_clean_rx_irq(struct e1000_ring *rx_ring, int *work_done,
926 			       int work_to_do)
927 {
928 	struct e1000_adapter *adapter = rx_ring->adapter;
929 	struct net_device *netdev = adapter->netdev;
930 	struct pci_dev *pdev = adapter->pdev;
931 	struct e1000_hw *hw = &adapter->hw;
932 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
933 	struct e1000_buffer *buffer_info, *next_buffer;
934 	u32 length, staterr;
935 	unsigned int i;
936 	int cleaned_count = 0;
937 	bool cleaned = false;
938 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
939 
940 	i = rx_ring->next_to_clean;
941 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
942 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
943 	buffer_info = &rx_ring->buffer_info[i];
944 
945 	while (staterr & E1000_RXD_STAT_DD) {
946 		struct sk_buff *skb;
947 
948 		if (*work_done >= work_to_do)
949 			break;
950 		(*work_done)++;
951 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
952 
953 		skb = buffer_info->skb;
954 		buffer_info->skb = NULL;
955 
956 		prefetch(skb->data - NET_IP_ALIGN);
957 
958 		i++;
959 		if (i == rx_ring->count)
960 			i = 0;
961 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
962 		prefetch(next_rxd);
963 
964 		next_buffer = &rx_ring->buffer_info[i];
965 
966 		cleaned = true;
967 		cleaned_count++;
968 		dma_unmap_single(&pdev->dev, buffer_info->dma,
969 				 adapter->rx_buffer_len, DMA_FROM_DEVICE);
970 		buffer_info->dma = 0;
971 
972 		length = le16_to_cpu(rx_desc->wb.upper.length);
973 
974 		/* !EOP means multiple descriptors were used to store a single
975 		 * packet, if that's the case we need to toss it.  In fact, we
976 		 * need to toss every packet with the EOP bit clear and the
977 		 * next frame that _does_ have the EOP bit set, as it is by
978 		 * definition only a frame fragment
979 		 */
980 		if (unlikely(!(staterr & E1000_RXD_STAT_EOP)))
981 			adapter->flags2 |= FLAG2_IS_DISCARDING;
982 
983 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
984 			/* All receives must fit into a single buffer */
985 			e_dbg("Receive packet consumed multiple buffers\n");
986 			/* recycle */
987 			buffer_info->skb = skb;
988 			if (staterr & E1000_RXD_STAT_EOP)
989 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
990 			goto next_desc;
991 		}
992 
993 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
994 			     !(netdev->features & NETIF_F_RXALL))) {
995 			/* recycle */
996 			buffer_info->skb = skb;
997 			goto next_desc;
998 		}
999 
1000 		/* adjust length to remove Ethernet CRC */
1001 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1002 			/* If configured to store CRC, don't subtract FCS,
1003 			 * but keep the FCS bytes out of the total_rx_bytes
1004 			 * counter
1005 			 */
1006 			if (netdev->features & NETIF_F_RXFCS)
1007 				total_rx_bytes -= 4;
1008 			else
1009 				length -= 4;
1010 		}
1011 
1012 		total_rx_bytes += length;
1013 		total_rx_packets++;
1014 
1015 		/* code added for copybreak, this should improve
1016 		 * performance for small packets with large amounts
1017 		 * of reassembly being done in the stack
1018 		 */
1019 		if (length < copybreak) {
1020 			struct sk_buff *new_skb =
1021 				napi_alloc_skb(&adapter->napi, length);
1022 			if (new_skb) {
1023 				skb_copy_to_linear_data_offset(new_skb,
1024 							       -NET_IP_ALIGN,
1025 							       (skb->data -
1026 								NET_IP_ALIGN),
1027 							       (length +
1028 								NET_IP_ALIGN));
1029 				/* save the skb in buffer_info as good */
1030 				buffer_info->skb = skb;
1031 				skb = new_skb;
1032 			}
1033 			/* else just continue with the old one */
1034 		}
1035 		/* end copybreak code */
1036 		skb_put(skb, length);
1037 
1038 		/* Receive Checksum Offload */
1039 		e1000_rx_checksum(adapter, staterr, skb);
1040 
1041 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1042 
1043 		e1000_receive_skb(adapter, netdev, skb, staterr,
1044 				  rx_desc->wb.upper.vlan);
1045 
1046 next_desc:
1047 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1048 
1049 		/* return some buffers to hardware, one at a time is too slow */
1050 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1051 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1052 					      GFP_ATOMIC);
1053 			cleaned_count = 0;
1054 		}
1055 
1056 		/* use prefetched values */
1057 		rx_desc = next_rxd;
1058 		buffer_info = next_buffer;
1059 
1060 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1061 	}
1062 	rx_ring->next_to_clean = i;
1063 
1064 	cleaned_count = e1000_desc_unused(rx_ring);
1065 	if (cleaned_count)
1066 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1067 
1068 	adapter->total_rx_bytes += total_rx_bytes;
1069 	adapter->total_rx_packets += total_rx_packets;
1070 	return cleaned;
1071 }
1072 
1073 static void e1000_put_txbuf(struct e1000_ring *tx_ring,
1074 			    struct e1000_buffer *buffer_info,
1075 			    bool drop)
1076 {
1077 	struct e1000_adapter *adapter = tx_ring->adapter;
1078 
1079 	if (buffer_info->dma) {
1080 		if (buffer_info->mapped_as_page)
1081 			dma_unmap_page(&adapter->pdev->dev, buffer_info->dma,
1082 				       buffer_info->length, DMA_TO_DEVICE);
1083 		else
1084 			dma_unmap_single(&adapter->pdev->dev, buffer_info->dma,
1085 					 buffer_info->length, DMA_TO_DEVICE);
1086 		buffer_info->dma = 0;
1087 	}
1088 	if (buffer_info->skb) {
1089 		if (drop)
1090 			dev_kfree_skb_any(buffer_info->skb);
1091 		else
1092 			dev_consume_skb_any(buffer_info->skb);
1093 		buffer_info->skb = NULL;
1094 	}
1095 	buffer_info->time_stamp = 0;
1096 }
1097 
1098 static void e1000_print_hw_hang(struct work_struct *work)
1099 {
1100 	struct e1000_adapter *adapter = container_of(work,
1101 						     struct e1000_adapter,
1102 						     print_hang_task);
1103 	struct net_device *netdev = adapter->netdev;
1104 	struct e1000_ring *tx_ring = adapter->tx_ring;
1105 	unsigned int i = tx_ring->next_to_clean;
1106 	unsigned int eop = tx_ring->buffer_info[i].next_to_watch;
1107 	struct e1000_tx_desc *eop_desc = E1000_TX_DESC(*tx_ring, eop);
1108 	struct e1000_hw *hw = &adapter->hw;
1109 	u16 phy_status, phy_1000t_status, phy_ext_status;
1110 	u16 pci_status;
1111 
1112 	if (test_bit(__E1000_DOWN, &adapter->state))
1113 		return;
1114 
1115 	if (!adapter->tx_hang_recheck && (adapter->flags2 & FLAG2_DMA_BURST)) {
1116 		/* May be block on write-back, flush and detect again
1117 		 * flush pending descriptor writebacks to memory
1118 		 */
1119 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1120 		/* execute the writes immediately */
1121 		e1e_flush();
1122 		/* Due to rare timing issues, write to TIDV again to ensure
1123 		 * the write is successful
1124 		 */
1125 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1126 		/* execute the writes immediately */
1127 		e1e_flush();
1128 		adapter->tx_hang_recheck = true;
1129 		return;
1130 	}
1131 	adapter->tx_hang_recheck = false;
1132 
1133 	if (er32(TDH(0)) == er32(TDT(0))) {
1134 		e_dbg("false hang detected, ignoring\n");
1135 		return;
1136 	}
1137 
1138 	/* Real hang detected */
1139 	netif_stop_queue(netdev);
1140 
1141 	e1e_rphy(hw, MII_BMSR, &phy_status);
1142 	e1e_rphy(hw, MII_STAT1000, &phy_1000t_status);
1143 	e1e_rphy(hw, MII_ESTATUS, &phy_ext_status);
1144 
1145 	pci_read_config_word(adapter->pdev, PCI_STATUS, &pci_status);
1146 
1147 	/* detected Hardware unit hang */
1148 	e_err("Detected Hardware Unit Hang:\n"
1149 	      "  TDH                  <%x>\n"
1150 	      "  TDT                  <%x>\n"
1151 	      "  next_to_use          <%x>\n"
1152 	      "  next_to_clean        <%x>\n"
1153 	      "buffer_info[next_to_clean]:\n"
1154 	      "  time_stamp           <%lx>\n"
1155 	      "  next_to_watch        <%x>\n"
1156 	      "  jiffies              <%lx>\n"
1157 	      "  next_to_watch.status <%x>\n"
1158 	      "MAC Status             <%x>\n"
1159 	      "PHY Status             <%x>\n"
1160 	      "PHY 1000BASE-T Status  <%x>\n"
1161 	      "PHY Extended Status    <%x>\n"
1162 	      "PCI Status             <%x>\n",
1163 	      readl(tx_ring->head), readl(tx_ring->tail), tx_ring->next_to_use,
1164 	      tx_ring->next_to_clean, tx_ring->buffer_info[eop].time_stamp,
1165 	      eop, jiffies, eop_desc->upper.fields.status, er32(STATUS),
1166 	      phy_status, phy_1000t_status, phy_ext_status, pci_status);
1167 
1168 	e1000e_dump(adapter);
1169 
1170 	/* Suggest workaround for known h/w issue */
1171 	if ((hw->mac.type == e1000_pchlan) && (er32(CTRL) & E1000_CTRL_TFCE))
1172 		e_err("Try turning off Tx pause (flow control) via ethtool\n");
1173 }
1174 
1175 /**
1176  * e1000e_tx_hwtstamp_work - check for Tx time stamp
1177  * @work: pointer to work struct
1178  *
1179  * This work function polls the TSYNCTXCTL valid bit to determine when a
1180  * timestamp has been taken for the current stored skb.  The timestamp must
1181  * be for this skb because only one such packet is allowed in the queue.
1182  */
1183 static void e1000e_tx_hwtstamp_work(struct work_struct *work)
1184 {
1185 	struct e1000_adapter *adapter = container_of(work, struct e1000_adapter,
1186 						     tx_hwtstamp_work);
1187 	struct e1000_hw *hw = &adapter->hw;
1188 
1189 	if (er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID) {
1190 		struct sk_buff *skb = adapter->tx_hwtstamp_skb;
1191 		struct skb_shared_hwtstamps shhwtstamps;
1192 		u64 txstmp;
1193 
1194 		txstmp = er32(TXSTMPL);
1195 		txstmp |= (u64)er32(TXSTMPH) << 32;
1196 
1197 		e1000e_systim_to_hwtstamp(adapter, &shhwtstamps, txstmp);
1198 
1199 		/* Clear the global tx_hwtstamp_skb pointer and force writes
1200 		 * prior to notifying the stack of a Tx timestamp.
1201 		 */
1202 		adapter->tx_hwtstamp_skb = NULL;
1203 		wmb(); /* force write prior to skb_tstamp_tx */
1204 
1205 		skb_tstamp_tx(skb, &shhwtstamps);
1206 		dev_consume_skb_any(skb);
1207 	} else if (time_after(jiffies, adapter->tx_hwtstamp_start
1208 			      + adapter->tx_timeout_factor * HZ)) {
1209 		dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
1210 		adapter->tx_hwtstamp_skb = NULL;
1211 		adapter->tx_hwtstamp_timeouts++;
1212 		e_warn("clearing Tx timestamp hang\n");
1213 	} else {
1214 		/* reschedule to check later */
1215 		schedule_work(&adapter->tx_hwtstamp_work);
1216 	}
1217 }
1218 
1219 /**
1220  * e1000_clean_tx_irq - Reclaim resources after transmit completes
1221  * @tx_ring: Tx descriptor ring
1222  *
1223  * the return value indicates whether actual cleaning was done, there
1224  * is no guarantee that everything was cleaned
1225  **/
1226 static bool e1000_clean_tx_irq(struct e1000_ring *tx_ring)
1227 {
1228 	struct e1000_adapter *adapter = tx_ring->adapter;
1229 	struct net_device *netdev = adapter->netdev;
1230 	struct e1000_hw *hw = &adapter->hw;
1231 	struct e1000_tx_desc *tx_desc, *eop_desc;
1232 	struct e1000_buffer *buffer_info;
1233 	unsigned int i, eop;
1234 	unsigned int count = 0;
1235 	unsigned int total_tx_bytes = 0, total_tx_packets = 0;
1236 	unsigned int bytes_compl = 0, pkts_compl = 0;
1237 
1238 	i = tx_ring->next_to_clean;
1239 	eop = tx_ring->buffer_info[i].next_to_watch;
1240 	eop_desc = E1000_TX_DESC(*tx_ring, eop);
1241 
1242 	while ((eop_desc->upper.data & cpu_to_le32(E1000_TXD_STAT_DD)) &&
1243 	       (count < tx_ring->count)) {
1244 		bool cleaned = false;
1245 
1246 		dma_rmb();		/* read buffer_info after eop_desc */
1247 		for (; !cleaned; count++) {
1248 			tx_desc = E1000_TX_DESC(*tx_ring, i);
1249 			buffer_info = &tx_ring->buffer_info[i];
1250 			cleaned = (i == eop);
1251 
1252 			if (cleaned) {
1253 				total_tx_packets += buffer_info->segs;
1254 				total_tx_bytes += buffer_info->bytecount;
1255 				if (buffer_info->skb) {
1256 					bytes_compl += buffer_info->skb->len;
1257 					pkts_compl++;
1258 				}
1259 			}
1260 
1261 			e1000_put_txbuf(tx_ring, buffer_info, false);
1262 			tx_desc->upper.data = 0;
1263 
1264 			i++;
1265 			if (i == tx_ring->count)
1266 				i = 0;
1267 		}
1268 
1269 		if (i == tx_ring->next_to_use)
1270 			break;
1271 		eop = tx_ring->buffer_info[i].next_to_watch;
1272 		eop_desc = E1000_TX_DESC(*tx_ring, eop);
1273 	}
1274 
1275 	tx_ring->next_to_clean = i;
1276 
1277 	netdev_completed_queue(netdev, pkts_compl, bytes_compl);
1278 
1279 #define TX_WAKE_THRESHOLD 32
1280 	if (count && netif_carrier_ok(netdev) &&
1281 	    e1000_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD) {
1282 		/* Make sure that anybody stopping the queue after this
1283 		 * sees the new next_to_clean.
1284 		 */
1285 		smp_mb();
1286 
1287 		if (netif_queue_stopped(netdev) &&
1288 		    !(test_bit(__E1000_DOWN, &adapter->state))) {
1289 			netif_wake_queue(netdev);
1290 			++adapter->restart_queue;
1291 		}
1292 	}
1293 
1294 	if (adapter->detect_tx_hung) {
1295 		/* Detect a transmit hang in hardware, this serializes the
1296 		 * check with the clearing of time_stamp and movement of i
1297 		 */
1298 		adapter->detect_tx_hung = false;
1299 		if (tx_ring->buffer_info[i].time_stamp &&
1300 		    time_after(jiffies, tx_ring->buffer_info[i].time_stamp
1301 			       + (adapter->tx_timeout_factor * HZ)) &&
1302 		    !(er32(STATUS) & E1000_STATUS_TXOFF))
1303 			schedule_work(&adapter->print_hang_task);
1304 		else
1305 			adapter->tx_hang_recheck = false;
1306 	}
1307 	adapter->total_tx_bytes += total_tx_bytes;
1308 	adapter->total_tx_packets += total_tx_packets;
1309 	return count < tx_ring->count;
1310 }
1311 
1312 /**
1313  * e1000_clean_rx_irq_ps - Send received data up the network stack; packet split
1314  * @rx_ring: Rx descriptor ring
1315  *
1316  * the return value indicates whether actual cleaning was done, there
1317  * is no guarantee that everything was cleaned
1318  **/
1319 static bool e1000_clean_rx_irq_ps(struct e1000_ring *rx_ring, int *work_done,
1320 				  int work_to_do)
1321 {
1322 	struct e1000_adapter *adapter = rx_ring->adapter;
1323 	struct e1000_hw *hw = &adapter->hw;
1324 	union e1000_rx_desc_packet_split *rx_desc, *next_rxd;
1325 	struct net_device *netdev = adapter->netdev;
1326 	struct pci_dev *pdev = adapter->pdev;
1327 	struct e1000_buffer *buffer_info, *next_buffer;
1328 	struct e1000_ps_page *ps_page;
1329 	struct sk_buff *skb;
1330 	unsigned int i, j;
1331 	u32 length, staterr;
1332 	int cleaned_count = 0;
1333 	bool cleaned = false;
1334 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1335 
1336 	i = rx_ring->next_to_clean;
1337 	rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
1338 	staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1339 	buffer_info = &rx_ring->buffer_info[i];
1340 
1341 	while (staterr & E1000_RXD_STAT_DD) {
1342 		if (*work_done >= work_to_do)
1343 			break;
1344 		(*work_done)++;
1345 		skb = buffer_info->skb;
1346 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1347 
1348 		/* in the packet split case this is header only */
1349 		prefetch(skb->data - NET_IP_ALIGN);
1350 
1351 		i++;
1352 		if (i == rx_ring->count)
1353 			i = 0;
1354 		next_rxd = E1000_RX_DESC_PS(*rx_ring, i);
1355 		prefetch(next_rxd);
1356 
1357 		next_buffer = &rx_ring->buffer_info[i];
1358 
1359 		cleaned = true;
1360 		cleaned_count++;
1361 		dma_unmap_single(&pdev->dev, buffer_info->dma,
1362 				 adapter->rx_ps_bsize0, DMA_FROM_DEVICE);
1363 		buffer_info->dma = 0;
1364 
1365 		/* see !EOP comment in other Rx routine */
1366 		if (!(staterr & E1000_RXD_STAT_EOP))
1367 			adapter->flags2 |= FLAG2_IS_DISCARDING;
1368 
1369 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
1370 			e_dbg("Packet Split buffers didn't pick up the full packet\n");
1371 			dev_kfree_skb_irq(skb);
1372 			if (staterr & E1000_RXD_STAT_EOP)
1373 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1374 			goto next_desc;
1375 		}
1376 
1377 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1378 			     !(netdev->features & NETIF_F_RXALL))) {
1379 			dev_kfree_skb_irq(skb);
1380 			goto next_desc;
1381 		}
1382 
1383 		length = le16_to_cpu(rx_desc->wb.middle.length0);
1384 
1385 		if (!length) {
1386 			e_dbg("Last part of the packet spanning multiple descriptors\n");
1387 			dev_kfree_skb_irq(skb);
1388 			goto next_desc;
1389 		}
1390 
1391 		/* Good Receive */
1392 		skb_put(skb, length);
1393 
1394 		{
1395 			/* this looks ugly, but it seems compiler issues make
1396 			 * it more efficient than reusing j
1397 			 */
1398 			int l1 = le16_to_cpu(rx_desc->wb.upper.length[0]);
1399 
1400 			/* page alloc/put takes too long and effects small
1401 			 * packet throughput, so unsplit small packets and
1402 			 * save the alloc/put only valid in softirq (napi)
1403 			 * context to call kmap_*
1404 			 */
1405 			if (l1 && (l1 <= copybreak) &&
1406 			    ((length + l1) <= adapter->rx_ps_bsize0)) {
1407 				u8 *vaddr;
1408 
1409 				ps_page = &buffer_info->ps_pages[0];
1410 
1411 				/* there is no documentation about how to call
1412 				 * kmap_atomic, so we can't hold the mapping
1413 				 * very long
1414 				 */
1415 				dma_sync_single_for_cpu(&pdev->dev,
1416 							ps_page->dma,
1417 							PAGE_SIZE,
1418 							DMA_FROM_DEVICE);
1419 				vaddr = kmap_atomic(ps_page->page);
1420 				memcpy(skb_tail_pointer(skb), vaddr, l1);
1421 				kunmap_atomic(vaddr);
1422 				dma_sync_single_for_device(&pdev->dev,
1423 							   ps_page->dma,
1424 							   PAGE_SIZE,
1425 							   DMA_FROM_DEVICE);
1426 
1427 				/* remove the CRC */
1428 				if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1429 					if (!(netdev->features & NETIF_F_RXFCS))
1430 						l1 -= 4;
1431 				}
1432 
1433 				skb_put(skb, l1);
1434 				goto copydone;
1435 			}	/* if */
1436 		}
1437 
1438 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1439 			length = le16_to_cpu(rx_desc->wb.upper.length[j]);
1440 			if (!length)
1441 				break;
1442 
1443 			ps_page = &buffer_info->ps_pages[j];
1444 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1445 				       DMA_FROM_DEVICE);
1446 			ps_page->dma = 0;
1447 			skb_fill_page_desc(skb, j, ps_page->page, 0, length);
1448 			ps_page->page = NULL;
1449 			skb->len += length;
1450 			skb->data_len += length;
1451 			skb->truesize += PAGE_SIZE;
1452 		}
1453 
1454 		/* strip the ethernet crc, problem is we're using pages now so
1455 		 * this whole operation can get a little cpu intensive
1456 		 */
1457 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1458 			if (!(netdev->features & NETIF_F_RXFCS))
1459 				pskb_trim(skb, skb->len - 4);
1460 		}
1461 
1462 copydone:
1463 		total_rx_bytes += skb->len;
1464 		total_rx_packets++;
1465 
1466 		e1000_rx_checksum(adapter, staterr, skb);
1467 
1468 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1469 
1470 		if (rx_desc->wb.upper.header_status &
1471 		    cpu_to_le16(E1000_RXDPS_HDRSTAT_HDRSP))
1472 			adapter->rx_hdr_split++;
1473 
1474 		e1000_receive_skb(adapter, netdev, skb, staterr,
1475 				  rx_desc->wb.middle.vlan);
1476 
1477 next_desc:
1478 		rx_desc->wb.middle.status_error &= cpu_to_le32(~0xFF);
1479 		buffer_info->skb = NULL;
1480 
1481 		/* return some buffers to hardware, one at a time is too slow */
1482 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1483 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1484 					      GFP_ATOMIC);
1485 			cleaned_count = 0;
1486 		}
1487 
1488 		/* use prefetched values */
1489 		rx_desc = next_rxd;
1490 		buffer_info = next_buffer;
1491 
1492 		staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1493 	}
1494 	rx_ring->next_to_clean = i;
1495 
1496 	cleaned_count = e1000_desc_unused(rx_ring);
1497 	if (cleaned_count)
1498 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1499 
1500 	adapter->total_rx_bytes += total_rx_bytes;
1501 	adapter->total_rx_packets += total_rx_packets;
1502 	return cleaned;
1503 }
1504 
1505 /**
1506  * e1000_consume_page - helper function
1507  **/
1508 static void e1000_consume_page(struct e1000_buffer *bi, struct sk_buff *skb,
1509 			       u16 length)
1510 {
1511 	bi->page = NULL;
1512 	skb->len += length;
1513 	skb->data_len += length;
1514 	skb->truesize += PAGE_SIZE;
1515 }
1516 
1517 /**
1518  * e1000_clean_jumbo_rx_irq - Send received data up the network stack; legacy
1519  * @adapter: board private structure
1520  *
1521  * the return value indicates whether actual cleaning was done, there
1522  * is no guarantee that everything was cleaned
1523  **/
1524 static bool e1000_clean_jumbo_rx_irq(struct e1000_ring *rx_ring, int *work_done,
1525 				     int work_to_do)
1526 {
1527 	struct e1000_adapter *adapter = rx_ring->adapter;
1528 	struct net_device *netdev = adapter->netdev;
1529 	struct pci_dev *pdev = adapter->pdev;
1530 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
1531 	struct e1000_buffer *buffer_info, *next_buffer;
1532 	u32 length, staterr;
1533 	unsigned int i;
1534 	int cleaned_count = 0;
1535 	bool cleaned = false;
1536 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1537 	struct skb_shared_info *shinfo;
1538 
1539 	i = rx_ring->next_to_clean;
1540 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
1541 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1542 	buffer_info = &rx_ring->buffer_info[i];
1543 
1544 	while (staterr & E1000_RXD_STAT_DD) {
1545 		struct sk_buff *skb;
1546 
1547 		if (*work_done >= work_to_do)
1548 			break;
1549 		(*work_done)++;
1550 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1551 
1552 		skb = buffer_info->skb;
1553 		buffer_info->skb = NULL;
1554 
1555 		++i;
1556 		if (i == rx_ring->count)
1557 			i = 0;
1558 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
1559 		prefetch(next_rxd);
1560 
1561 		next_buffer = &rx_ring->buffer_info[i];
1562 
1563 		cleaned = true;
1564 		cleaned_count++;
1565 		dma_unmap_page(&pdev->dev, buffer_info->dma, PAGE_SIZE,
1566 			       DMA_FROM_DEVICE);
1567 		buffer_info->dma = 0;
1568 
1569 		length = le16_to_cpu(rx_desc->wb.upper.length);
1570 
1571 		/* errors is only valid for DD + EOP descriptors */
1572 		if (unlikely((staterr & E1000_RXD_STAT_EOP) &&
1573 			     ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1574 			      !(netdev->features & NETIF_F_RXALL)))) {
1575 			/* recycle both page and skb */
1576 			buffer_info->skb = skb;
1577 			/* an error means any chain goes out the window too */
1578 			if (rx_ring->rx_skb_top)
1579 				dev_kfree_skb_irq(rx_ring->rx_skb_top);
1580 			rx_ring->rx_skb_top = NULL;
1581 			goto next_desc;
1582 		}
1583 #define rxtop (rx_ring->rx_skb_top)
1584 		if (!(staterr & E1000_RXD_STAT_EOP)) {
1585 			/* this descriptor is only the beginning (or middle) */
1586 			if (!rxtop) {
1587 				/* this is the beginning of a chain */
1588 				rxtop = skb;
1589 				skb_fill_page_desc(rxtop, 0, buffer_info->page,
1590 						   0, length);
1591 			} else {
1592 				/* this is the middle of a chain */
1593 				shinfo = skb_shinfo(rxtop);
1594 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1595 						   buffer_info->page, 0,
1596 						   length);
1597 				/* re-use the skb, only consumed the page */
1598 				buffer_info->skb = skb;
1599 			}
1600 			e1000_consume_page(buffer_info, rxtop, length);
1601 			goto next_desc;
1602 		} else {
1603 			if (rxtop) {
1604 				/* end of the chain */
1605 				shinfo = skb_shinfo(rxtop);
1606 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1607 						   buffer_info->page, 0,
1608 						   length);
1609 				/* re-use the current skb, we only consumed the
1610 				 * page
1611 				 */
1612 				buffer_info->skb = skb;
1613 				skb = rxtop;
1614 				rxtop = NULL;
1615 				e1000_consume_page(buffer_info, skb, length);
1616 			} else {
1617 				/* no chain, got EOP, this buf is the packet
1618 				 * copybreak to save the put_page/alloc_page
1619 				 */
1620 				if (length <= copybreak &&
1621 				    skb_tailroom(skb) >= length) {
1622 					u8 *vaddr;
1623 					vaddr = kmap_atomic(buffer_info->page);
1624 					memcpy(skb_tail_pointer(skb), vaddr,
1625 					       length);
1626 					kunmap_atomic(vaddr);
1627 					/* re-use the page, so don't erase
1628 					 * buffer_info->page
1629 					 */
1630 					skb_put(skb, length);
1631 				} else {
1632 					skb_fill_page_desc(skb, 0,
1633 							   buffer_info->page, 0,
1634 							   length);
1635 					e1000_consume_page(buffer_info, skb,
1636 							   length);
1637 				}
1638 			}
1639 		}
1640 
1641 		/* Receive Checksum Offload */
1642 		e1000_rx_checksum(adapter, staterr, skb);
1643 
1644 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1645 
1646 		/* probably a little skewed due to removing CRC */
1647 		total_rx_bytes += skb->len;
1648 		total_rx_packets++;
1649 
1650 		/* eth type trans needs skb->data to point to something */
1651 		if (!pskb_may_pull(skb, ETH_HLEN)) {
1652 			e_err("pskb_may_pull failed.\n");
1653 			dev_kfree_skb_irq(skb);
1654 			goto next_desc;
1655 		}
1656 
1657 		e1000_receive_skb(adapter, netdev, skb, staterr,
1658 				  rx_desc->wb.upper.vlan);
1659 
1660 next_desc:
1661 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1662 
1663 		/* return some buffers to hardware, one at a time is too slow */
1664 		if (unlikely(cleaned_count >= E1000_RX_BUFFER_WRITE)) {
1665 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1666 					      GFP_ATOMIC);
1667 			cleaned_count = 0;
1668 		}
1669 
1670 		/* use prefetched values */
1671 		rx_desc = next_rxd;
1672 		buffer_info = next_buffer;
1673 
1674 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1675 	}
1676 	rx_ring->next_to_clean = i;
1677 
1678 	cleaned_count = e1000_desc_unused(rx_ring);
1679 	if (cleaned_count)
1680 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1681 
1682 	adapter->total_rx_bytes += total_rx_bytes;
1683 	adapter->total_rx_packets += total_rx_packets;
1684 	return cleaned;
1685 }
1686 
1687 /**
1688  * e1000_clean_rx_ring - Free Rx Buffers per Queue
1689  * @rx_ring: Rx descriptor ring
1690  **/
1691 static void e1000_clean_rx_ring(struct e1000_ring *rx_ring)
1692 {
1693 	struct e1000_adapter *adapter = rx_ring->adapter;
1694 	struct e1000_buffer *buffer_info;
1695 	struct e1000_ps_page *ps_page;
1696 	struct pci_dev *pdev = adapter->pdev;
1697 	unsigned int i, j;
1698 
1699 	/* Free all the Rx ring sk_buffs */
1700 	for (i = 0; i < rx_ring->count; i++) {
1701 		buffer_info = &rx_ring->buffer_info[i];
1702 		if (buffer_info->dma) {
1703 			if (adapter->clean_rx == e1000_clean_rx_irq)
1704 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1705 						 adapter->rx_buffer_len,
1706 						 DMA_FROM_DEVICE);
1707 			else if (adapter->clean_rx == e1000_clean_jumbo_rx_irq)
1708 				dma_unmap_page(&pdev->dev, buffer_info->dma,
1709 					       PAGE_SIZE, DMA_FROM_DEVICE);
1710 			else if (adapter->clean_rx == e1000_clean_rx_irq_ps)
1711 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1712 						 adapter->rx_ps_bsize0,
1713 						 DMA_FROM_DEVICE);
1714 			buffer_info->dma = 0;
1715 		}
1716 
1717 		if (buffer_info->page) {
1718 			put_page(buffer_info->page);
1719 			buffer_info->page = NULL;
1720 		}
1721 
1722 		if (buffer_info->skb) {
1723 			dev_kfree_skb(buffer_info->skb);
1724 			buffer_info->skb = NULL;
1725 		}
1726 
1727 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1728 			ps_page = &buffer_info->ps_pages[j];
1729 			if (!ps_page->page)
1730 				break;
1731 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1732 				       DMA_FROM_DEVICE);
1733 			ps_page->dma = 0;
1734 			put_page(ps_page->page);
1735 			ps_page->page = NULL;
1736 		}
1737 	}
1738 
1739 	/* there also may be some cached data from a chained receive */
1740 	if (rx_ring->rx_skb_top) {
1741 		dev_kfree_skb(rx_ring->rx_skb_top);
1742 		rx_ring->rx_skb_top = NULL;
1743 	}
1744 
1745 	/* Zero out the descriptor ring */
1746 	memset(rx_ring->desc, 0, rx_ring->size);
1747 
1748 	rx_ring->next_to_clean = 0;
1749 	rx_ring->next_to_use = 0;
1750 	adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1751 }
1752 
1753 static void e1000e_downshift_workaround(struct work_struct *work)
1754 {
1755 	struct e1000_adapter *adapter = container_of(work,
1756 						     struct e1000_adapter,
1757 						     downshift_task);
1758 
1759 	if (test_bit(__E1000_DOWN, &adapter->state))
1760 		return;
1761 
1762 	e1000e_gig_downshift_workaround_ich8lan(&adapter->hw);
1763 }
1764 
1765 /**
1766  * e1000_intr_msi - Interrupt Handler
1767  * @irq: interrupt number
1768  * @data: pointer to a network interface device structure
1769  **/
1770 static irqreturn_t e1000_intr_msi(int __always_unused irq, void *data)
1771 {
1772 	struct net_device *netdev = data;
1773 	struct e1000_adapter *adapter = netdev_priv(netdev);
1774 	struct e1000_hw *hw = &adapter->hw;
1775 	u32 icr = er32(ICR);
1776 
1777 	/* read ICR disables interrupts using IAM */
1778 	if (icr & E1000_ICR_LSC) {
1779 		hw->mac.get_link_status = true;
1780 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1781 		 * disconnect (LSC) before accessing any PHY registers
1782 		 */
1783 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1784 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1785 			schedule_work(&adapter->downshift_task);
1786 
1787 		/* 80003ES2LAN workaround-- For packet buffer work-around on
1788 		 * link down event; disable receives here in the ISR and reset
1789 		 * adapter in watchdog
1790 		 */
1791 		if (netif_carrier_ok(netdev) &&
1792 		    adapter->flags & FLAG_RX_NEEDS_RESTART) {
1793 			/* disable receives */
1794 			u32 rctl = er32(RCTL);
1795 
1796 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1797 			adapter->flags |= FLAG_RESTART_NOW;
1798 		}
1799 		/* guard against interrupt when we're going down */
1800 		if (!test_bit(__E1000_DOWN, &adapter->state))
1801 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1802 	}
1803 
1804 	/* Reset on uncorrectable ECC error */
1805 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1806 		u32 pbeccsts = er32(PBECCSTS);
1807 
1808 		adapter->corr_errors +=
1809 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1810 		adapter->uncorr_errors +=
1811 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1812 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1813 
1814 		/* Do the reset outside of interrupt context */
1815 		schedule_work(&adapter->reset_task);
1816 
1817 		/* return immediately since reset is imminent */
1818 		return IRQ_HANDLED;
1819 	}
1820 
1821 	if (napi_schedule_prep(&adapter->napi)) {
1822 		adapter->total_tx_bytes = 0;
1823 		adapter->total_tx_packets = 0;
1824 		adapter->total_rx_bytes = 0;
1825 		adapter->total_rx_packets = 0;
1826 		__napi_schedule(&adapter->napi);
1827 	}
1828 
1829 	return IRQ_HANDLED;
1830 }
1831 
1832 /**
1833  * e1000_intr - Interrupt Handler
1834  * @irq: interrupt number
1835  * @data: pointer to a network interface device structure
1836  **/
1837 static irqreturn_t e1000_intr(int __always_unused irq, void *data)
1838 {
1839 	struct net_device *netdev = data;
1840 	struct e1000_adapter *adapter = netdev_priv(netdev);
1841 	struct e1000_hw *hw = &adapter->hw;
1842 	u32 rctl, icr = er32(ICR);
1843 
1844 	if (!icr || test_bit(__E1000_DOWN, &adapter->state))
1845 		return IRQ_NONE;	/* Not our interrupt */
1846 
1847 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
1848 	 * not set, then the adapter didn't send an interrupt
1849 	 */
1850 	if (!(icr & E1000_ICR_INT_ASSERTED))
1851 		return IRQ_NONE;
1852 
1853 	/* Interrupt Auto-Mask...upon reading ICR,
1854 	 * interrupts are masked.  No need for the
1855 	 * IMC write
1856 	 */
1857 
1858 	if (icr & E1000_ICR_LSC) {
1859 		hw->mac.get_link_status = true;
1860 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1861 		 * disconnect (LSC) before accessing any PHY registers
1862 		 */
1863 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1864 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1865 			schedule_work(&adapter->downshift_task);
1866 
1867 		/* 80003ES2LAN workaround--
1868 		 * For packet buffer work-around on link down event;
1869 		 * disable receives here in the ISR and
1870 		 * reset adapter in watchdog
1871 		 */
1872 		if (netif_carrier_ok(netdev) &&
1873 		    (adapter->flags & FLAG_RX_NEEDS_RESTART)) {
1874 			/* disable receives */
1875 			rctl = er32(RCTL);
1876 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1877 			adapter->flags |= FLAG_RESTART_NOW;
1878 		}
1879 		/* guard against interrupt when we're going down */
1880 		if (!test_bit(__E1000_DOWN, &adapter->state))
1881 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1882 	}
1883 
1884 	/* Reset on uncorrectable ECC error */
1885 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1886 		u32 pbeccsts = er32(PBECCSTS);
1887 
1888 		adapter->corr_errors +=
1889 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1890 		adapter->uncorr_errors +=
1891 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1892 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1893 
1894 		/* Do the reset outside of interrupt context */
1895 		schedule_work(&adapter->reset_task);
1896 
1897 		/* return immediately since reset is imminent */
1898 		return IRQ_HANDLED;
1899 	}
1900 
1901 	if (napi_schedule_prep(&adapter->napi)) {
1902 		adapter->total_tx_bytes = 0;
1903 		adapter->total_tx_packets = 0;
1904 		adapter->total_rx_bytes = 0;
1905 		adapter->total_rx_packets = 0;
1906 		__napi_schedule(&adapter->napi);
1907 	}
1908 
1909 	return IRQ_HANDLED;
1910 }
1911 
1912 static irqreturn_t e1000_msix_other(int __always_unused irq, void *data)
1913 {
1914 	struct net_device *netdev = data;
1915 	struct e1000_adapter *adapter = netdev_priv(netdev);
1916 	struct e1000_hw *hw = &adapter->hw;
1917 	u32 icr;
1918 	bool enable = true;
1919 
1920 	icr = er32(ICR);
1921 	if (icr & E1000_ICR_RXO) {
1922 		ew32(ICR, E1000_ICR_RXO);
1923 		enable = false;
1924 		/* napi poll will re-enable Other, make sure it runs */
1925 		if (napi_schedule_prep(&adapter->napi)) {
1926 			adapter->total_rx_bytes = 0;
1927 			adapter->total_rx_packets = 0;
1928 			__napi_schedule(&adapter->napi);
1929 		}
1930 	}
1931 	if (icr & E1000_ICR_LSC) {
1932 		ew32(ICR, E1000_ICR_LSC);
1933 		hw->mac.get_link_status = true;
1934 		/* guard against interrupt when we're going down */
1935 		if (!test_bit(__E1000_DOWN, &adapter->state))
1936 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1937 	}
1938 
1939 	if (enable && !test_bit(__E1000_DOWN, &adapter->state))
1940 		ew32(IMS, E1000_IMS_OTHER);
1941 
1942 	return IRQ_HANDLED;
1943 }
1944 
1945 static irqreturn_t e1000_intr_msix_tx(int __always_unused irq, void *data)
1946 {
1947 	struct net_device *netdev = data;
1948 	struct e1000_adapter *adapter = netdev_priv(netdev);
1949 	struct e1000_hw *hw = &adapter->hw;
1950 	struct e1000_ring *tx_ring = adapter->tx_ring;
1951 
1952 	adapter->total_tx_bytes = 0;
1953 	adapter->total_tx_packets = 0;
1954 
1955 	if (!e1000_clean_tx_irq(tx_ring))
1956 		/* Ring was not completely cleaned, so fire another interrupt */
1957 		ew32(ICS, tx_ring->ims_val);
1958 
1959 	if (!test_bit(__E1000_DOWN, &adapter->state))
1960 		ew32(IMS, adapter->tx_ring->ims_val);
1961 
1962 	return IRQ_HANDLED;
1963 }
1964 
1965 static irqreturn_t e1000_intr_msix_rx(int __always_unused irq, void *data)
1966 {
1967 	struct net_device *netdev = data;
1968 	struct e1000_adapter *adapter = netdev_priv(netdev);
1969 	struct e1000_ring *rx_ring = adapter->rx_ring;
1970 
1971 	/* Write the ITR value calculated at the end of the
1972 	 * previous interrupt.
1973 	 */
1974 	if (rx_ring->set_itr) {
1975 		u32 itr = rx_ring->itr_val ?
1976 			  1000000000 / (rx_ring->itr_val * 256) : 0;
1977 
1978 		writel(itr, rx_ring->itr_register);
1979 		rx_ring->set_itr = 0;
1980 	}
1981 
1982 	if (napi_schedule_prep(&adapter->napi)) {
1983 		adapter->total_rx_bytes = 0;
1984 		adapter->total_rx_packets = 0;
1985 		__napi_schedule(&adapter->napi);
1986 	}
1987 	return IRQ_HANDLED;
1988 }
1989 
1990 /**
1991  * e1000_configure_msix - Configure MSI-X hardware
1992  *
1993  * e1000_configure_msix sets up the hardware to properly
1994  * generate MSI-X interrupts.
1995  **/
1996 static void e1000_configure_msix(struct e1000_adapter *adapter)
1997 {
1998 	struct e1000_hw *hw = &adapter->hw;
1999 	struct e1000_ring *rx_ring = adapter->rx_ring;
2000 	struct e1000_ring *tx_ring = adapter->tx_ring;
2001 	int vector = 0;
2002 	u32 ctrl_ext, ivar = 0;
2003 
2004 	adapter->eiac_mask = 0;
2005 
2006 	/* Workaround issue with spurious interrupts on 82574 in MSI-X mode */
2007 	if (hw->mac.type == e1000_82574) {
2008 		u32 rfctl = er32(RFCTL);
2009 
2010 		rfctl |= E1000_RFCTL_ACK_DIS;
2011 		ew32(RFCTL, rfctl);
2012 	}
2013 
2014 	/* Configure Rx vector */
2015 	rx_ring->ims_val = E1000_IMS_RXQ0;
2016 	adapter->eiac_mask |= rx_ring->ims_val;
2017 	if (rx_ring->itr_val)
2018 		writel(1000000000 / (rx_ring->itr_val * 256),
2019 		       rx_ring->itr_register);
2020 	else
2021 		writel(1, rx_ring->itr_register);
2022 	ivar = E1000_IVAR_INT_ALLOC_VALID | vector;
2023 
2024 	/* Configure Tx vector */
2025 	tx_ring->ims_val = E1000_IMS_TXQ0;
2026 	vector++;
2027 	if (tx_ring->itr_val)
2028 		writel(1000000000 / (tx_ring->itr_val * 256),
2029 		       tx_ring->itr_register);
2030 	else
2031 		writel(1, tx_ring->itr_register);
2032 	adapter->eiac_mask |= tx_ring->ims_val;
2033 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 8);
2034 
2035 	/* set vector for Other Causes, e.g. link changes */
2036 	vector++;
2037 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 16);
2038 	if (rx_ring->itr_val)
2039 		writel(1000000000 / (rx_ring->itr_val * 256),
2040 		       hw->hw_addr + E1000_EITR_82574(vector));
2041 	else
2042 		writel(1, hw->hw_addr + E1000_EITR_82574(vector));
2043 	adapter->eiac_mask |= E1000_IMS_OTHER;
2044 
2045 	/* Cause Tx interrupts on every write back */
2046 	ivar |= BIT(31);
2047 
2048 	ew32(IVAR, ivar);
2049 
2050 	/* enable MSI-X PBA support */
2051 	ctrl_ext = er32(CTRL_EXT) & ~E1000_CTRL_EXT_IAME;
2052 	ctrl_ext |= E1000_CTRL_EXT_PBA_CLR | E1000_CTRL_EXT_EIAME;
2053 	ew32(CTRL_EXT, ctrl_ext);
2054 	e1e_flush();
2055 }
2056 
2057 void e1000e_reset_interrupt_capability(struct e1000_adapter *adapter)
2058 {
2059 	if (adapter->msix_entries) {
2060 		pci_disable_msix(adapter->pdev);
2061 		kfree(adapter->msix_entries);
2062 		adapter->msix_entries = NULL;
2063 	} else if (adapter->flags & FLAG_MSI_ENABLED) {
2064 		pci_disable_msi(adapter->pdev);
2065 		adapter->flags &= ~FLAG_MSI_ENABLED;
2066 	}
2067 }
2068 
2069 /**
2070  * e1000e_set_interrupt_capability - set MSI or MSI-X if supported
2071  *
2072  * Attempt to configure interrupts using the best available
2073  * capabilities of the hardware and kernel.
2074  **/
2075 void e1000e_set_interrupt_capability(struct e1000_adapter *adapter)
2076 {
2077 	int err;
2078 	int i;
2079 
2080 	switch (adapter->int_mode) {
2081 	case E1000E_INT_MODE_MSIX:
2082 		if (adapter->flags & FLAG_HAS_MSIX) {
2083 			adapter->num_vectors = 3; /* RxQ0, TxQ0 and other */
2084 			adapter->msix_entries = kcalloc(adapter->num_vectors,
2085 							sizeof(struct
2086 							       msix_entry),
2087 							GFP_KERNEL);
2088 			if (adapter->msix_entries) {
2089 				struct e1000_adapter *a = adapter;
2090 
2091 				for (i = 0; i < adapter->num_vectors; i++)
2092 					adapter->msix_entries[i].entry = i;
2093 
2094 				err = pci_enable_msix_range(a->pdev,
2095 							    a->msix_entries,
2096 							    a->num_vectors,
2097 							    a->num_vectors);
2098 				if (err > 0)
2099 					return;
2100 			}
2101 			/* MSI-X failed, so fall through and try MSI */
2102 			e_err("Failed to initialize MSI-X interrupts.  Falling back to MSI interrupts.\n");
2103 			e1000e_reset_interrupt_capability(adapter);
2104 		}
2105 		adapter->int_mode = E1000E_INT_MODE_MSI;
2106 		/* Fall through */
2107 	case E1000E_INT_MODE_MSI:
2108 		if (!pci_enable_msi(adapter->pdev)) {
2109 			adapter->flags |= FLAG_MSI_ENABLED;
2110 		} else {
2111 			adapter->int_mode = E1000E_INT_MODE_LEGACY;
2112 			e_err("Failed to initialize MSI interrupts.  Falling back to legacy interrupts.\n");
2113 		}
2114 		/* Fall through */
2115 	case E1000E_INT_MODE_LEGACY:
2116 		/* Don't do anything; this is the system default */
2117 		break;
2118 	}
2119 
2120 	/* store the number of vectors being used */
2121 	adapter->num_vectors = 1;
2122 }
2123 
2124 /**
2125  * e1000_request_msix - Initialize MSI-X interrupts
2126  *
2127  * e1000_request_msix allocates MSI-X vectors and requests interrupts from the
2128  * kernel.
2129  **/
2130 static int e1000_request_msix(struct e1000_adapter *adapter)
2131 {
2132 	struct net_device *netdev = adapter->netdev;
2133 	int err = 0, vector = 0;
2134 
2135 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2136 		snprintf(adapter->rx_ring->name,
2137 			 sizeof(adapter->rx_ring->name) - 1,
2138 			 "%s-rx-0", netdev->name);
2139 	else
2140 		memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
2141 	err = request_irq(adapter->msix_entries[vector].vector,
2142 			  e1000_intr_msix_rx, 0, adapter->rx_ring->name,
2143 			  netdev);
2144 	if (err)
2145 		return err;
2146 	adapter->rx_ring->itr_register = adapter->hw.hw_addr +
2147 	    E1000_EITR_82574(vector);
2148 	adapter->rx_ring->itr_val = adapter->itr;
2149 	vector++;
2150 
2151 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2152 		snprintf(adapter->tx_ring->name,
2153 			 sizeof(adapter->tx_ring->name) - 1,
2154 			 "%s-tx-0", netdev->name);
2155 	else
2156 		memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
2157 	err = request_irq(adapter->msix_entries[vector].vector,
2158 			  e1000_intr_msix_tx, 0, adapter->tx_ring->name,
2159 			  netdev);
2160 	if (err)
2161 		return err;
2162 	adapter->tx_ring->itr_register = adapter->hw.hw_addr +
2163 	    E1000_EITR_82574(vector);
2164 	adapter->tx_ring->itr_val = adapter->itr;
2165 	vector++;
2166 
2167 	err = request_irq(adapter->msix_entries[vector].vector,
2168 			  e1000_msix_other, 0, netdev->name, netdev);
2169 	if (err)
2170 		return err;
2171 
2172 	e1000_configure_msix(adapter);
2173 
2174 	return 0;
2175 }
2176 
2177 /**
2178  * e1000_request_irq - initialize interrupts
2179  *
2180  * Attempts to configure interrupts using the best available
2181  * capabilities of the hardware and kernel.
2182  **/
2183 static int e1000_request_irq(struct e1000_adapter *adapter)
2184 {
2185 	struct net_device *netdev = adapter->netdev;
2186 	int err;
2187 
2188 	if (adapter->msix_entries) {
2189 		err = e1000_request_msix(adapter);
2190 		if (!err)
2191 			return err;
2192 		/* fall back to MSI */
2193 		e1000e_reset_interrupt_capability(adapter);
2194 		adapter->int_mode = E1000E_INT_MODE_MSI;
2195 		e1000e_set_interrupt_capability(adapter);
2196 	}
2197 	if (adapter->flags & FLAG_MSI_ENABLED) {
2198 		err = request_irq(adapter->pdev->irq, e1000_intr_msi, 0,
2199 				  netdev->name, netdev);
2200 		if (!err)
2201 			return err;
2202 
2203 		/* fall back to legacy interrupt */
2204 		e1000e_reset_interrupt_capability(adapter);
2205 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
2206 	}
2207 
2208 	err = request_irq(adapter->pdev->irq, e1000_intr, IRQF_SHARED,
2209 			  netdev->name, netdev);
2210 	if (err)
2211 		e_err("Unable to allocate interrupt, Error: %d\n", err);
2212 
2213 	return err;
2214 }
2215 
2216 static void e1000_free_irq(struct e1000_adapter *adapter)
2217 {
2218 	struct net_device *netdev = adapter->netdev;
2219 
2220 	if (adapter->msix_entries) {
2221 		int vector = 0;
2222 
2223 		free_irq(adapter->msix_entries[vector].vector, netdev);
2224 		vector++;
2225 
2226 		free_irq(adapter->msix_entries[vector].vector, netdev);
2227 		vector++;
2228 
2229 		/* Other Causes interrupt vector */
2230 		free_irq(adapter->msix_entries[vector].vector, netdev);
2231 		return;
2232 	}
2233 
2234 	free_irq(adapter->pdev->irq, netdev);
2235 }
2236 
2237 /**
2238  * e1000_irq_disable - Mask off interrupt generation on the NIC
2239  **/
2240 static void e1000_irq_disable(struct e1000_adapter *adapter)
2241 {
2242 	struct e1000_hw *hw = &adapter->hw;
2243 
2244 	ew32(IMC, ~0);
2245 	if (adapter->msix_entries)
2246 		ew32(EIAC_82574, 0);
2247 	e1e_flush();
2248 
2249 	if (adapter->msix_entries) {
2250 		int i;
2251 
2252 		for (i = 0; i < adapter->num_vectors; i++)
2253 			synchronize_irq(adapter->msix_entries[i].vector);
2254 	} else {
2255 		synchronize_irq(adapter->pdev->irq);
2256 	}
2257 }
2258 
2259 /**
2260  * e1000_irq_enable - Enable default interrupt generation settings
2261  **/
2262 static void e1000_irq_enable(struct e1000_adapter *adapter)
2263 {
2264 	struct e1000_hw *hw = &adapter->hw;
2265 
2266 	if (adapter->msix_entries) {
2267 		ew32(EIAC_82574, adapter->eiac_mask & E1000_EIAC_MASK_82574);
2268 		ew32(IMS, adapter->eiac_mask | E1000_IMS_LSC);
2269 	} else if (hw->mac.type >= e1000_pch_lpt) {
2270 		ew32(IMS, IMS_ENABLE_MASK | E1000_IMS_ECCER);
2271 	} else {
2272 		ew32(IMS, IMS_ENABLE_MASK);
2273 	}
2274 	e1e_flush();
2275 }
2276 
2277 /**
2278  * e1000e_get_hw_control - get control of the h/w from f/w
2279  * @adapter: address of board private structure
2280  *
2281  * e1000e_get_hw_control sets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2282  * For ASF and Pass Through versions of f/w this means that
2283  * the driver is loaded. For AMT version (only with 82573)
2284  * of the f/w this means that the network i/f is open.
2285  **/
2286 void e1000e_get_hw_control(struct e1000_adapter *adapter)
2287 {
2288 	struct e1000_hw *hw = &adapter->hw;
2289 	u32 ctrl_ext;
2290 	u32 swsm;
2291 
2292 	/* Let firmware know the driver has taken over */
2293 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2294 		swsm = er32(SWSM);
2295 		ew32(SWSM, swsm | E1000_SWSM_DRV_LOAD);
2296 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2297 		ctrl_ext = er32(CTRL_EXT);
2298 		ew32(CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2299 	}
2300 }
2301 
2302 /**
2303  * e1000e_release_hw_control - release control of the h/w to f/w
2304  * @adapter: address of board private structure
2305  *
2306  * e1000e_release_hw_control resets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2307  * For ASF and Pass Through versions of f/w this means that the
2308  * driver is no longer loaded. For AMT version (only with 82573) i
2309  * of the f/w this means that the network i/f is closed.
2310  *
2311  **/
2312 void e1000e_release_hw_control(struct e1000_adapter *adapter)
2313 {
2314 	struct e1000_hw *hw = &adapter->hw;
2315 	u32 ctrl_ext;
2316 	u32 swsm;
2317 
2318 	/* Let firmware taken over control of h/w */
2319 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2320 		swsm = er32(SWSM);
2321 		ew32(SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
2322 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2323 		ctrl_ext = er32(CTRL_EXT);
2324 		ew32(CTRL_EXT, ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2325 	}
2326 }
2327 
2328 /**
2329  * e1000_alloc_ring_dma - allocate memory for a ring structure
2330  **/
2331 static int e1000_alloc_ring_dma(struct e1000_adapter *adapter,
2332 				struct e1000_ring *ring)
2333 {
2334 	struct pci_dev *pdev = adapter->pdev;
2335 
2336 	ring->desc = dma_alloc_coherent(&pdev->dev, ring->size, &ring->dma,
2337 					GFP_KERNEL);
2338 	if (!ring->desc)
2339 		return -ENOMEM;
2340 
2341 	return 0;
2342 }
2343 
2344 /**
2345  * e1000e_setup_tx_resources - allocate Tx resources (Descriptors)
2346  * @tx_ring: Tx descriptor ring
2347  *
2348  * Return 0 on success, negative on failure
2349  **/
2350 int e1000e_setup_tx_resources(struct e1000_ring *tx_ring)
2351 {
2352 	struct e1000_adapter *adapter = tx_ring->adapter;
2353 	int err = -ENOMEM, size;
2354 
2355 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2356 	tx_ring->buffer_info = vzalloc(size);
2357 	if (!tx_ring->buffer_info)
2358 		goto err;
2359 
2360 	/* round up to nearest 4K */
2361 	tx_ring->size = tx_ring->count * sizeof(struct e1000_tx_desc);
2362 	tx_ring->size = ALIGN(tx_ring->size, 4096);
2363 
2364 	err = e1000_alloc_ring_dma(adapter, tx_ring);
2365 	if (err)
2366 		goto err;
2367 
2368 	tx_ring->next_to_use = 0;
2369 	tx_ring->next_to_clean = 0;
2370 
2371 	return 0;
2372 err:
2373 	vfree(tx_ring->buffer_info);
2374 	e_err("Unable to allocate memory for the transmit descriptor ring\n");
2375 	return err;
2376 }
2377 
2378 /**
2379  * e1000e_setup_rx_resources - allocate Rx resources (Descriptors)
2380  * @rx_ring: Rx descriptor ring
2381  *
2382  * Returns 0 on success, negative on failure
2383  **/
2384 int e1000e_setup_rx_resources(struct e1000_ring *rx_ring)
2385 {
2386 	struct e1000_adapter *adapter = rx_ring->adapter;
2387 	struct e1000_buffer *buffer_info;
2388 	int i, size, desc_len, err = -ENOMEM;
2389 
2390 	size = sizeof(struct e1000_buffer) * rx_ring->count;
2391 	rx_ring->buffer_info = vzalloc(size);
2392 	if (!rx_ring->buffer_info)
2393 		goto err;
2394 
2395 	for (i = 0; i < rx_ring->count; i++) {
2396 		buffer_info = &rx_ring->buffer_info[i];
2397 		buffer_info->ps_pages = kcalloc(PS_PAGE_BUFFERS,
2398 						sizeof(struct e1000_ps_page),
2399 						GFP_KERNEL);
2400 		if (!buffer_info->ps_pages)
2401 			goto err_pages;
2402 	}
2403 
2404 	desc_len = sizeof(union e1000_rx_desc_packet_split);
2405 
2406 	/* Round up to nearest 4K */
2407 	rx_ring->size = rx_ring->count * desc_len;
2408 	rx_ring->size = ALIGN(rx_ring->size, 4096);
2409 
2410 	err = e1000_alloc_ring_dma(adapter, rx_ring);
2411 	if (err)
2412 		goto err_pages;
2413 
2414 	rx_ring->next_to_clean = 0;
2415 	rx_ring->next_to_use = 0;
2416 	rx_ring->rx_skb_top = NULL;
2417 
2418 	return 0;
2419 
2420 err_pages:
2421 	for (i = 0; i < rx_ring->count; i++) {
2422 		buffer_info = &rx_ring->buffer_info[i];
2423 		kfree(buffer_info->ps_pages);
2424 	}
2425 err:
2426 	vfree(rx_ring->buffer_info);
2427 	e_err("Unable to allocate memory for the receive descriptor ring\n");
2428 	return err;
2429 }
2430 
2431 /**
2432  * e1000_clean_tx_ring - Free Tx Buffers
2433  * @tx_ring: Tx descriptor ring
2434  **/
2435 static void e1000_clean_tx_ring(struct e1000_ring *tx_ring)
2436 {
2437 	struct e1000_adapter *adapter = tx_ring->adapter;
2438 	struct e1000_buffer *buffer_info;
2439 	unsigned long size;
2440 	unsigned int i;
2441 
2442 	for (i = 0; i < tx_ring->count; i++) {
2443 		buffer_info = &tx_ring->buffer_info[i];
2444 		e1000_put_txbuf(tx_ring, buffer_info, false);
2445 	}
2446 
2447 	netdev_reset_queue(adapter->netdev);
2448 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2449 	memset(tx_ring->buffer_info, 0, size);
2450 
2451 	memset(tx_ring->desc, 0, tx_ring->size);
2452 
2453 	tx_ring->next_to_use = 0;
2454 	tx_ring->next_to_clean = 0;
2455 }
2456 
2457 /**
2458  * e1000e_free_tx_resources - Free Tx Resources per Queue
2459  * @tx_ring: Tx descriptor ring
2460  *
2461  * Free all transmit software resources
2462  **/
2463 void e1000e_free_tx_resources(struct e1000_ring *tx_ring)
2464 {
2465 	struct e1000_adapter *adapter = tx_ring->adapter;
2466 	struct pci_dev *pdev = adapter->pdev;
2467 
2468 	e1000_clean_tx_ring(tx_ring);
2469 
2470 	vfree(tx_ring->buffer_info);
2471 	tx_ring->buffer_info = NULL;
2472 
2473 	dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
2474 			  tx_ring->dma);
2475 	tx_ring->desc = NULL;
2476 }
2477 
2478 /**
2479  * e1000e_free_rx_resources - Free Rx Resources
2480  * @rx_ring: Rx descriptor ring
2481  *
2482  * Free all receive software resources
2483  **/
2484 void e1000e_free_rx_resources(struct e1000_ring *rx_ring)
2485 {
2486 	struct e1000_adapter *adapter = rx_ring->adapter;
2487 	struct pci_dev *pdev = adapter->pdev;
2488 	int i;
2489 
2490 	e1000_clean_rx_ring(rx_ring);
2491 
2492 	for (i = 0; i < rx_ring->count; i++)
2493 		kfree(rx_ring->buffer_info[i].ps_pages);
2494 
2495 	vfree(rx_ring->buffer_info);
2496 	rx_ring->buffer_info = NULL;
2497 
2498 	dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
2499 			  rx_ring->dma);
2500 	rx_ring->desc = NULL;
2501 }
2502 
2503 /**
2504  * e1000_update_itr - update the dynamic ITR value based on statistics
2505  * @adapter: pointer to adapter
2506  * @itr_setting: current adapter->itr
2507  * @packets: the number of packets during this measurement interval
2508  * @bytes: the number of bytes during this measurement interval
2509  *
2510  *      Stores a new ITR value based on packets and byte
2511  *      counts during the last interrupt.  The advantage of per interrupt
2512  *      computation is faster updates and more accurate ITR for the current
2513  *      traffic pattern.  Constants in this function were computed
2514  *      based on theoretical maximum wire speed and thresholds were set based
2515  *      on testing data as well as attempting to minimize response time
2516  *      while increasing bulk throughput.  This functionality is controlled
2517  *      by the InterruptThrottleRate module parameter.
2518  **/
2519 static unsigned int e1000_update_itr(u16 itr_setting, int packets, int bytes)
2520 {
2521 	unsigned int retval = itr_setting;
2522 
2523 	if (packets == 0)
2524 		return itr_setting;
2525 
2526 	switch (itr_setting) {
2527 	case lowest_latency:
2528 		/* handle TSO and jumbo frames */
2529 		if (bytes / packets > 8000)
2530 			retval = bulk_latency;
2531 		else if ((packets < 5) && (bytes > 512))
2532 			retval = low_latency;
2533 		break;
2534 	case low_latency:	/* 50 usec aka 20000 ints/s */
2535 		if (bytes > 10000) {
2536 			/* this if handles the TSO accounting */
2537 			if (bytes / packets > 8000)
2538 				retval = bulk_latency;
2539 			else if ((packets < 10) || ((bytes / packets) > 1200))
2540 				retval = bulk_latency;
2541 			else if ((packets > 35))
2542 				retval = lowest_latency;
2543 		} else if (bytes / packets > 2000) {
2544 			retval = bulk_latency;
2545 		} else if (packets <= 2 && bytes < 512) {
2546 			retval = lowest_latency;
2547 		}
2548 		break;
2549 	case bulk_latency:	/* 250 usec aka 4000 ints/s */
2550 		if (bytes > 25000) {
2551 			if (packets > 35)
2552 				retval = low_latency;
2553 		} else if (bytes < 6000) {
2554 			retval = low_latency;
2555 		}
2556 		break;
2557 	}
2558 
2559 	return retval;
2560 }
2561 
2562 static void e1000_set_itr(struct e1000_adapter *adapter)
2563 {
2564 	u16 current_itr;
2565 	u32 new_itr = adapter->itr;
2566 
2567 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
2568 	if (adapter->link_speed != SPEED_1000) {
2569 		current_itr = 0;
2570 		new_itr = 4000;
2571 		goto set_itr_now;
2572 	}
2573 
2574 	if (adapter->flags2 & FLAG2_DISABLE_AIM) {
2575 		new_itr = 0;
2576 		goto set_itr_now;
2577 	}
2578 
2579 	adapter->tx_itr = e1000_update_itr(adapter->tx_itr,
2580 					   adapter->total_tx_packets,
2581 					   adapter->total_tx_bytes);
2582 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2583 	if (adapter->itr_setting == 3 && adapter->tx_itr == lowest_latency)
2584 		adapter->tx_itr = low_latency;
2585 
2586 	adapter->rx_itr = e1000_update_itr(adapter->rx_itr,
2587 					   adapter->total_rx_packets,
2588 					   adapter->total_rx_bytes);
2589 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2590 	if (adapter->itr_setting == 3 && adapter->rx_itr == lowest_latency)
2591 		adapter->rx_itr = low_latency;
2592 
2593 	current_itr = max(adapter->rx_itr, adapter->tx_itr);
2594 
2595 	/* counts and packets in update_itr are dependent on these numbers */
2596 	switch (current_itr) {
2597 	case lowest_latency:
2598 		new_itr = 70000;
2599 		break;
2600 	case low_latency:
2601 		new_itr = 20000;	/* aka hwitr = ~200 */
2602 		break;
2603 	case bulk_latency:
2604 		new_itr = 4000;
2605 		break;
2606 	default:
2607 		break;
2608 	}
2609 
2610 set_itr_now:
2611 	if (new_itr != adapter->itr) {
2612 		/* this attempts to bias the interrupt rate towards Bulk
2613 		 * by adding intermediate steps when interrupt rate is
2614 		 * increasing
2615 		 */
2616 		new_itr = new_itr > adapter->itr ?
2617 		    min(adapter->itr + (new_itr >> 2), new_itr) : new_itr;
2618 		adapter->itr = new_itr;
2619 		adapter->rx_ring->itr_val = new_itr;
2620 		if (adapter->msix_entries)
2621 			adapter->rx_ring->set_itr = 1;
2622 		else
2623 			e1000e_write_itr(adapter, new_itr);
2624 	}
2625 }
2626 
2627 /**
2628  * e1000e_write_itr - write the ITR value to the appropriate registers
2629  * @adapter: address of board private structure
2630  * @itr: new ITR value to program
2631  *
2632  * e1000e_write_itr determines if the adapter is in MSI-X mode
2633  * and, if so, writes the EITR registers with the ITR value.
2634  * Otherwise, it writes the ITR value into the ITR register.
2635  **/
2636 void e1000e_write_itr(struct e1000_adapter *adapter, u32 itr)
2637 {
2638 	struct e1000_hw *hw = &adapter->hw;
2639 	u32 new_itr = itr ? 1000000000 / (itr * 256) : 0;
2640 
2641 	if (adapter->msix_entries) {
2642 		int vector;
2643 
2644 		for (vector = 0; vector < adapter->num_vectors; vector++)
2645 			writel(new_itr, hw->hw_addr + E1000_EITR_82574(vector));
2646 	} else {
2647 		ew32(ITR, new_itr);
2648 	}
2649 }
2650 
2651 /**
2652  * e1000_alloc_queues - Allocate memory for all rings
2653  * @adapter: board private structure to initialize
2654  **/
2655 static int e1000_alloc_queues(struct e1000_adapter *adapter)
2656 {
2657 	int size = sizeof(struct e1000_ring);
2658 
2659 	adapter->tx_ring = kzalloc(size, GFP_KERNEL);
2660 	if (!adapter->tx_ring)
2661 		goto err;
2662 	adapter->tx_ring->count = adapter->tx_ring_count;
2663 	adapter->tx_ring->adapter = adapter;
2664 
2665 	adapter->rx_ring = kzalloc(size, GFP_KERNEL);
2666 	if (!adapter->rx_ring)
2667 		goto err;
2668 	adapter->rx_ring->count = adapter->rx_ring_count;
2669 	adapter->rx_ring->adapter = adapter;
2670 
2671 	return 0;
2672 err:
2673 	e_err("Unable to allocate memory for queues\n");
2674 	kfree(adapter->rx_ring);
2675 	kfree(adapter->tx_ring);
2676 	return -ENOMEM;
2677 }
2678 
2679 /**
2680  * e1000e_poll - NAPI Rx polling callback
2681  * @napi: struct associated with this polling callback
2682  * @weight: number of packets driver is allowed to process this poll
2683  **/
2684 static int e1000e_poll(struct napi_struct *napi, int weight)
2685 {
2686 	struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter,
2687 						     napi);
2688 	struct e1000_hw *hw = &adapter->hw;
2689 	struct net_device *poll_dev = adapter->netdev;
2690 	int tx_cleaned = 1, work_done = 0;
2691 
2692 	adapter = netdev_priv(poll_dev);
2693 
2694 	if (!adapter->msix_entries ||
2695 	    (adapter->rx_ring->ims_val & adapter->tx_ring->ims_val))
2696 		tx_cleaned = e1000_clean_tx_irq(adapter->tx_ring);
2697 
2698 	adapter->clean_rx(adapter->rx_ring, &work_done, weight);
2699 
2700 	if (!tx_cleaned)
2701 		work_done = weight;
2702 
2703 	/* If weight not fully consumed, exit the polling mode */
2704 	if (work_done < weight) {
2705 		if (adapter->itr_setting & 3)
2706 			e1000_set_itr(adapter);
2707 		napi_complete_done(napi, work_done);
2708 		if (!test_bit(__E1000_DOWN, &adapter->state)) {
2709 			if (adapter->msix_entries)
2710 				ew32(IMS, adapter->rx_ring->ims_val |
2711 				     E1000_IMS_OTHER);
2712 			else
2713 				e1000_irq_enable(adapter);
2714 		}
2715 	}
2716 
2717 	return work_done;
2718 }
2719 
2720 static int e1000_vlan_rx_add_vid(struct net_device *netdev,
2721 				 __always_unused __be16 proto, u16 vid)
2722 {
2723 	struct e1000_adapter *adapter = netdev_priv(netdev);
2724 	struct e1000_hw *hw = &adapter->hw;
2725 	u32 vfta, index;
2726 
2727 	/* don't update vlan cookie if already programmed */
2728 	if ((adapter->hw.mng_cookie.status &
2729 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2730 	    (vid == adapter->mng_vlan_id))
2731 		return 0;
2732 
2733 	/* add VID to filter table */
2734 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2735 		index = (vid >> 5) & 0x7F;
2736 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2737 		vfta |= BIT((vid & 0x1F));
2738 		hw->mac.ops.write_vfta(hw, index, vfta);
2739 	}
2740 
2741 	set_bit(vid, adapter->active_vlans);
2742 
2743 	return 0;
2744 }
2745 
2746 static int e1000_vlan_rx_kill_vid(struct net_device *netdev,
2747 				  __always_unused __be16 proto, u16 vid)
2748 {
2749 	struct e1000_adapter *adapter = netdev_priv(netdev);
2750 	struct e1000_hw *hw = &adapter->hw;
2751 	u32 vfta, index;
2752 
2753 	if ((adapter->hw.mng_cookie.status &
2754 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2755 	    (vid == adapter->mng_vlan_id)) {
2756 		/* release control to f/w */
2757 		e1000e_release_hw_control(adapter);
2758 		return 0;
2759 	}
2760 
2761 	/* remove VID from filter table */
2762 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2763 		index = (vid >> 5) & 0x7F;
2764 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2765 		vfta &= ~BIT((vid & 0x1F));
2766 		hw->mac.ops.write_vfta(hw, index, vfta);
2767 	}
2768 
2769 	clear_bit(vid, adapter->active_vlans);
2770 
2771 	return 0;
2772 }
2773 
2774 /**
2775  * e1000e_vlan_filter_disable - helper to disable hw VLAN filtering
2776  * @adapter: board private structure to initialize
2777  **/
2778 static void e1000e_vlan_filter_disable(struct e1000_adapter *adapter)
2779 {
2780 	struct net_device *netdev = adapter->netdev;
2781 	struct e1000_hw *hw = &adapter->hw;
2782 	u32 rctl;
2783 
2784 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2785 		/* disable VLAN receive filtering */
2786 		rctl = er32(RCTL);
2787 		rctl &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
2788 		ew32(RCTL, rctl);
2789 
2790 		if (adapter->mng_vlan_id != (u16)E1000_MNG_VLAN_NONE) {
2791 			e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
2792 					       adapter->mng_vlan_id);
2793 			adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
2794 		}
2795 	}
2796 }
2797 
2798 /**
2799  * e1000e_vlan_filter_enable - helper to enable HW VLAN filtering
2800  * @adapter: board private structure to initialize
2801  **/
2802 static void e1000e_vlan_filter_enable(struct e1000_adapter *adapter)
2803 {
2804 	struct e1000_hw *hw = &adapter->hw;
2805 	u32 rctl;
2806 
2807 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2808 		/* enable VLAN receive filtering */
2809 		rctl = er32(RCTL);
2810 		rctl |= E1000_RCTL_VFE;
2811 		rctl &= ~E1000_RCTL_CFIEN;
2812 		ew32(RCTL, rctl);
2813 	}
2814 }
2815 
2816 /**
2817  * e1000e_vlan_strip_disable - helper to disable HW VLAN stripping
2818  * @adapter: board private structure to initialize
2819  **/
2820 static void e1000e_vlan_strip_disable(struct e1000_adapter *adapter)
2821 {
2822 	struct e1000_hw *hw = &adapter->hw;
2823 	u32 ctrl;
2824 
2825 	/* disable VLAN tag insert/strip */
2826 	ctrl = er32(CTRL);
2827 	ctrl &= ~E1000_CTRL_VME;
2828 	ew32(CTRL, ctrl);
2829 }
2830 
2831 /**
2832  * e1000e_vlan_strip_enable - helper to enable HW VLAN stripping
2833  * @adapter: board private structure to initialize
2834  **/
2835 static void e1000e_vlan_strip_enable(struct e1000_adapter *adapter)
2836 {
2837 	struct e1000_hw *hw = &adapter->hw;
2838 	u32 ctrl;
2839 
2840 	/* enable VLAN tag insert/strip */
2841 	ctrl = er32(CTRL);
2842 	ctrl |= E1000_CTRL_VME;
2843 	ew32(CTRL, ctrl);
2844 }
2845 
2846 static void e1000_update_mng_vlan(struct e1000_adapter *adapter)
2847 {
2848 	struct net_device *netdev = adapter->netdev;
2849 	u16 vid = adapter->hw.mng_cookie.vlan_id;
2850 	u16 old_vid = adapter->mng_vlan_id;
2851 
2852 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
2853 		e1000_vlan_rx_add_vid(netdev, htons(ETH_P_8021Q), vid);
2854 		adapter->mng_vlan_id = vid;
2855 	}
2856 
2857 	if ((old_vid != (u16)E1000_MNG_VLAN_NONE) && (vid != old_vid))
2858 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q), old_vid);
2859 }
2860 
2861 static void e1000_restore_vlan(struct e1000_adapter *adapter)
2862 {
2863 	u16 vid;
2864 
2865 	e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
2866 
2867 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
2868 	    e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
2869 }
2870 
2871 static void e1000_init_manageability_pt(struct e1000_adapter *adapter)
2872 {
2873 	struct e1000_hw *hw = &adapter->hw;
2874 	u32 manc, manc2h, mdef, i, j;
2875 
2876 	if (!(adapter->flags & FLAG_MNG_PT_ENABLED))
2877 		return;
2878 
2879 	manc = er32(MANC);
2880 
2881 	/* enable receiving management packets to the host. this will probably
2882 	 * generate destination unreachable messages from the host OS, but
2883 	 * the packets will be handled on SMBUS
2884 	 */
2885 	manc |= E1000_MANC_EN_MNG2HOST;
2886 	manc2h = er32(MANC2H);
2887 
2888 	switch (hw->mac.type) {
2889 	default:
2890 		manc2h |= (E1000_MANC2H_PORT_623 | E1000_MANC2H_PORT_664);
2891 		break;
2892 	case e1000_82574:
2893 	case e1000_82583:
2894 		/* Check if IPMI pass-through decision filter already exists;
2895 		 * if so, enable it.
2896 		 */
2897 		for (i = 0, j = 0; i < 8; i++) {
2898 			mdef = er32(MDEF(i));
2899 
2900 			/* Ignore filters with anything other than IPMI ports */
2901 			if (mdef & ~(E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2902 				continue;
2903 
2904 			/* Enable this decision filter in MANC2H */
2905 			if (mdef)
2906 				manc2h |= BIT(i);
2907 
2908 			j |= mdef;
2909 		}
2910 
2911 		if (j == (E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2912 			break;
2913 
2914 		/* Create new decision filter in an empty filter */
2915 		for (i = 0, j = 0; i < 8; i++)
2916 			if (er32(MDEF(i)) == 0) {
2917 				ew32(MDEF(i), (E1000_MDEF_PORT_623 |
2918 					       E1000_MDEF_PORT_664));
2919 				manc2h |= BIT(1);
2920 				j++;
2921 				break;
2922 			}
2923 
2924 		if (!j)
2925 			e_warn("Unable to create IPMI pass-through filter\n");
2926 		break;
2927 	}
2928 
2929 	ew32(MANC2H, manc2h);
2930 	ew32(MANC, manc);
2931 }
2932 
2933 /**
2934  * e1000_configure_tx - Configure Transmit Unit after Reset
2935  * @adapter: board private structure
2936  *
2937  * Configure the Tx unit of the MAC after a reset.
2938  **/
2939 static void e1000_configure_tx(struct e1000_adapter *adapter)
2940 {
2941 	struct e1000_hw *hw = &adapter->hw;
2942 	struct e1000_ring *tx_ring = adapter->tx_ring;
2943 	u64 tdba;
2944 	u32 tdlen, tctl, tarc;
2945 
2946 	/* Setup the HW Tx Head and Tail descriptor pointers */
2947 	tdba = tx_ring->dma;
2948 	tdlen = tx_ring->count * sizeof(struct e1000_tx_desc);
2949 	ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
2950 	ew32(TDBAH(0), (tdba >> 32));
2951 	ew32(TDLEN(0), tdlen);
2952 	ew32(TDH(0), 0);
2953 	ew32(TDT(0), 0);
2954 	tx_ring->head = adapter->hw.hw_addr + E1000_TDH(0);
2955 	tx_ring->tail = adapter->hw.hw_addr + E1000_TDT(0);
2956 
2957 	writel(0, tx_ring->head);
2958 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
2959 		e1000e_update_tdt_wa(tx_ring, 0);
2960 	else
2961 		writel(0, tx_ring->tail);
2962 
2963 	/* Set the Tx Interrupt Delay register */
2964 	ew32(TIDV, adapter->tx_int_delay);
2965 	/* Tx irq moderation */
2966 	ew32(TADV, adapter->tx_abs_int_delay);
2967 
2968 	if (adapter->flags2 & FLAG2_DMA_BURST) {
2969 		u32 txdctl = er32(TXDCTL(0));
2970 
2971 		txdctl &= ~(E1000_TXDCTL_PTHRESH | E1000_TXDCTL_HTHRESH |
2972 			    E1000_TXDCTL_WTHRESH);
2973 		/* set up some performance related parameters to encourage the
2974 		 * hardware to use the bus more efficiently in bursts, depends
2975 		 * on the tx_int_delay to be enabled,
2976 		 * wthresh = 1 ==> burst write is disabled to avoid Tx stalls
2977 		 * hthresh = 1 ==> prefetch when one or more available
2978 		 * pthresh = 0x1f ==> prefetch if internal cache 31 or less
2979 		 * BEWARE: this seems to work but should be considered first if
2980 		 * there are Tx hangs or other Tx related bugs
2981 		 */
2982 		txdctl |= E1000_TXDCTL_DMA_BURST_ENABLE;
2983 		ew32(TXDCTL(0), txdctl);
2984 	}
2985 	/* erratum work around: set txdctl the same for both queues */
2986 	ew32(TXDCTL(1), er32(TXDCTL(0)));
2987 
2988 	/* Program the Transmit Control Register */
2989 	tctl = er32(TCTL);
2990 	tctl &= ~E1000_TCTL_CT;
2991 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
2992 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
2993 
2994 	if (adapter->flags & FLAG_TARC_SPEED_MODE_BIT) {
2995 		tarc = er32(TARC(0));
2996 		/* set the speed mode bit, we'll clear it if we're not at
2997 		 * gigabit link later
2998 		 */
2999 #define SPEED_MODE_BIT BIT(21)
3000 		tarc |= SPEED_MODE_BIT;
3001 		ew32(TARC(0), tarc);
3002 	}
3003 
3004 	/* errata: program both queues to unweighted RR */
3005 	if (adapter->flags & FLAG_TARC_SET_BIT_ZERO) {
3006 		tarc = er32(TARC(0));
3007 		tarc |= 1;
3008 		ew32(TARC(0), tarc);
3009 		tarc = er32(TARC(1));
3010 		tarc |= 1;
3011 		ew32(TARC(1), tarc);
3012 	}
3013 
3014 	/* Setup Transmit Descriptor Settings for eop descriptor */
3015 	adapter->txd_cmd = E1000_TXD_CMD_EOP | E1000_TXD_CMD_IFCS;
3016 
3017 	/* only set IDE if we are delaying interrupts using the timers */
3018 	if (adapter->tx_int_delay)
3019 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
3020 
3021 	/* enable Report Status bit */
3022 	adapter->txd_cmd |= E1000_TXD_CMD_RS;
3023 
3024 	ew32(TCTL, tctl);
3025 
3026 	hw->mac.ops.config_collision_dist(hw);
3027 
3028 	/* SPT and KBL Si errata workaround to avoid data corruption */
3029 	if (hw->mac.type == e1000_pch_spt) {
3030 		u32 reg_val;
3031 
3032 		reg_val = er32(IOSFPC);
3033 		reg_val |= E1000_RCTL_RDMTS_HEX;
3034 		ew32(IOSFPC, reg_val);
3035 
3036 		reg_val = er32(TARC(0));
3037 		/* SPT and KBL Si errata workaround to avoid Tx hang.
3038 		 * Dropping the number of outstanding requests from
3039 		 * 3 to 2 in order to avoid a buffer overrun.
3040 		 */
3041 		reg_val &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3042 		reg_val |= E1000_TARC0_CB_MULTIQ_2_REQ;
3043 		ew32(TARC(0), reg_val);
3044 	}
3045 }
3046 
3047 /**
3048  * e1000_setup_rctl - configure the receive control registers
3049  * @adapter: Board private structure
3050  **/
3051 #define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
3052 			   (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
3053 static void e1000_setup_rctl(struct e1000_adapter *adapter)
3054 {
3055 	struct e1000_hw *hw = &adapter->hw;
3056 	u32 rctl, rfctl;
3057 	u32 pages = 0;
3058 
3059 	/* Workaround Si errata on PCHx - configure jumbo frame flow.
3060 	 * If jumbo frames not set, program related MAC/PHY registers
3061 	 * to h/w defaults
3062 	 */
3063 	if (hw->mac.type >= e1000_pch2lan) {
3064 		s32 ret_val;
3065 
3066 		if (adapter->netdev->mtu > ETH_DATA_LEN)
3067 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, true);
3068 		else
3069 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, false);
3070 
3071 		if (ret_val)
3072 			e_dbg("failed to enable|disable jumbo frame workaround mode\n");
3073 	}
3074 
3075 	/* Program MC offset vector base */
3076 	rctl = er32(RCTL);
3077 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3078 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3079 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3080 	    (adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3081 
3082 	/* Do not Store bad packets */
3083 	rctl &= ~E1000_RCTL_SBP;
3084 
3085 	/* Enable Long Packet receive */
3086 	if (adapter->netdev->mtu <= ETH_DATA_LEN)
3087 		rctl &= ~E1000_RCTL_LPE;
3088 	else
3089 		rctl |= E1000_RCTL_LPE;
3090 
3091 	/* Some systems expect that the CRC is included in SMBUS traffic. The
3092 	 * hardware strips the CRC before sending to both SMBUS (BMC) and to
3093 	 * host memory when this is enabled
3094 	 */
3095 	if (adapter->flags2 & FLAG2_CRC_STRIPPING)
3096 		rctl |= E1000_RCTL_SECRC;
3097 
3098 	/* Workaround Si errata on 82577 PHY - configure IPG for jumbos */
3099 	if ((hw->phy.type == e1000_phy_82577) && (rctl & E1000_RCTL_LPE)) {
3100 		u16 phy_data;
3101 
3102 		e1e_rphy(hw, PHY_REG(770, 26), &phy_data);
3103 		phy_data &= 0xfff8;
3104 		phy_data |= BIT(2);
3105 		e1e_wphy(hw, PHY_REG(770, 26), phy_data);
3106 
3107 		e1e_rphy(hw, 22, &phy_data);
3108 		phy_data &= 0x0fff;
3109 		phy_data |= BIT(14);
3110 		e1e_wphy(hw, 0x10, 0x2823);
3111 		e1e_wphy(hw, 0x11, 0x0003);
3112 		e1e_wphy(hw, 22, phy_data);
3113 	}
3114 
3115 	/* Setup buffer sizes */
3116 	rctl &= ~E1000_RCTL_SZ_4096;
3117 	rctl |= E1000_RCTL_BSEX;
3118 	switch (adapter->rx_buffer_len) {
3119 	case 2048:
3120 	default:
3121 		rctl |= E1000_RCTL_SZ_2048;
3122 		rctl &= ~E1000_RCTL_BSEX;
3123 		break;
3124 	case 4096:
3125 		rctl |= E1000_RCTL_SZ_4096;
3126 		break;
3127 	case 8192:
3128 		rctl |= E1000_RCTL_SZ_8192;
3129 		break;
3130 	case 16384:
3131 		rctl |= E1000_RCTL_SZ_16384;
3132 		break;
3133 	}
3134 
3135 	/* Enable Extended Status in all Receive Descriptors */
3136 	rfctl = er32(RFCTL);
3137 	rfctl |= E1000_RFCTL_EXTEN;
3138 	ew32(RFCTL, rfctl);
3139 
3140 	/* 82571 and greater support packet-split where the protocol
3141 	 * header is placed in skb->data and the packet data is
3142 	 * placed in pages hanging off of skb_shinfo(skb)->nr_frags.
3143 	 * In the case of a non-split, skb->data is linearly filled,
3144 	 * followed by the page buffers.  Therefore, skb->data is
3145 	 * sized to hold the largest protocol header.
3146 	 *
3147 	 * allocations using alloc_page take too long for regular MTU
3148 	 * so only enable packet split for jumbo frames
3149 	 *
3150 	 * Using pages when the page size is greater than 16k wastes
3151 	 * a lot of memory, since we allocate 3 pages at all times
3152 	 * per packet.
3153 	 */
3154 	pages = PAGE_USE_COUNT(adapter->netdev->mtu);
3155 	if ((pages <= 3) && (PAGE_SIZE <= 16384) && (rctl & E1000_RCTL_LPE))
3156 		adapter->rx_ps_pages = pages;
3157 	else
3158 		adapter->rx_ps_pages = 0;
3159 
3160 	if (adapter->rx_ps_pages) {
3161 		u32 psrctl = 0;
3162 
3163 		/* Enable Packet split descriptors */
3164 		rctl |= E1000_RCTL_DTYP_PS;
3165 
3166 		psrctl |= adapter->rx_ps_bsize0 >> E1000_PSRCTL_BSIZE0_SHIFT;
3167 
3168 		switch (adapter->rx_ps_pages) {
3169 		case 3:
3170 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE3_SHIFT;
3171 			/* fall-through */
3172 		case 2:
3173 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE2_SHIFT;
3174 			/* fall-through */
3175 		case 1:
3176 			psrctl |= PAGE_SIZE >> E1000_PSRCTL_BSIZE1_SHIFT;
3177 			break;
3178 		}
3179 
3180 		ew32(PSRCTL, psrctl);
3181 	}
3182 
3183 	/* This is useful for sniffing bad packets. */
3184 	if (adapter->netdev->features & NETIF_F_RXALL) {
3185 		/* UPE and MPE will be handled by normal PROMISC logic
3186 		 * in e1000e_set_rx_mode
3187 		 */
3188 		rctl |= (E1000_RCTL_SBP |	/* Receive bad packets */
3189 			 E1000_RCTL_BAM |	/* RX All Bcast Pkts */
3190 			 E1000_RCTL_PMCF);	/* RX All MAC Ctrl Pkts */
3191 
3192 		rctl &= ~(E1000_RCTL_VFE |	/* Disable VLAN filter */
3193 			  E1000_RCTL_DPF |	/* Allow filtered pause */
3194 			  E1000_RCTL_CFIEN);	/* Dis VLAN CFIEN Filter */
3195 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3196 		 * and that breaks VLANs.
3197 		 */
3198 	}
3199 
3200 	ew32(RCTL, rctl);
3201 	/* just started the receive unit, no need to restart */
3202 	adapter->flags &= ~FLAG_RESTART_NOW;
3203 }
3204 
3205 /**
3206  * e1000_configure_rx - Configure Receive Unit after Reset
3207  * @adapter: board private structure
3208  *
3209  * Configure the Rx unit of the MAC after a reset.
3210  **/
3211 static void e1000_configure_rx(struct e1000_adapter *adapter)
3212 {
3213 	struct e1000_hw *hw = &adapter->hw;
3214 	struct e1000_ring *rx_ring = adapter->rx_ring;
3215 	u64 rdba;
3216 	u32 rdlen, rctl, rxcsum, ctrl_ext;
3217 
3218 	if (adapter->rx_ps_pages) {
3219 		/* this is a 32 byte descriptor */
3220 		rdlen = rx_ring->count *
3221 		    sizeof(union e1000_rx_desc_packet_split);
3222 		adapter->clean_rx = e1000_clean_rx_irq_ps;
3223 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers_ps;
3224 	} else if (adapter->netdev->mtu > ETH_FRAME_LEN + ETH_FCS_LEN) {
3225 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3226 		adapter->clean_rx = e1000_clean_jumbo_rx_irq;
3227 		adapter->alloc_rx_buf = e1000_alloc_jumbo_rx_buffers;
3228 	} else {
3229 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3230 		adapter->clean_rx = e1000_clean_rx_irq;
3231 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers;
3232 	}
3233 
3234 	/* disable receives while setting up the descriptors */
3235 	rctl = er32(RCTL);
3236 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
3237 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
3238 	e1e_flush();
3239 	usleep_range(10000, 20000);
3240 
3241 	if (adapter->flags2 & FLAG2_DMA_BURST) {
3242 		/* set the writeback threshold (only takes effect if the RDTR
3243 		 * is set). set GRAN=1 and write back up to 0x4 worth, and
3244 		 * enable prefetching of 0x20 Rx descriptors
3245 		 * granularity = 01
3246 		 * wthresh = 04,
3247 		 * hthresh = 04,
3248 		 * pthresh = 0x20
3249 		 */
3250 		ew32(RXDCTL(0), E1000_RXDCTL_DMA_BURST_ENABLE);
3251 		ew32(RXDCTL(1), E1000_RXDCTL_DMA_BURST_ENABLE);
3252 	}
3253 
3254 	/* set the Receive Delay Timer Register */
3255 	ew32(RDTR, adapter->rx_int_delay);
3256 
3257 	/* irq moderation */
3258 	ew32(RADV, adapter->rx_abs_int_delay);
3259 	if ((adapter->itr_setting != 0) && (adapter->itr != 0))
3260 		e1000e_write_itr(adapter, adapter->itr);
3261 
3262 	ctrl_ext = er32(CTRL_EXT);
3263 	/* Auto-Mask interrupts upon ICR access */
3264 	ctrl_ext |= E1000_CTRL_EXT_IAME;
3265 	ew32(IAM, 0xffffffff);
3266 	ew32(CTRL_EXT, ctrl_ext);
3267 	e1e_flush();
3268 
3269 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
3270 	 * the Base and Length of the Rx Descriptor Ring
3271 	 */
3272 	rdba = rx_ring->dma;
3273 	ew32(RDBAL(0), (rdba & DMA_BIT_MASK(32)));
3274 	ew32(RDBAH(0), (rdba >> 32));
3275 	ew32(RDLEN(0), rdlen);
3276 	ew32(RDH(0), 0);
3277 	ew32(RDT(0), 0);
3278 	rx_ring->head = adapter->hw.hw_addr + E1000_RDH(0);
3279 	rx_ring->tail = adapter->hw.hw_addr + E1000_RDT(0);
3280 
3281 	writel(0, rx_ring->head);
3282 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
3283 		e1000e_update_rdt_wa(rx_ring, 0);
3284 	else
3285 		writel(0, rx_ring->tail);
3286 
3287 	/* Enable Receive Checksum Offload for TCP and UDP */
3288 	rxcsum = er32(RXCSUM);
3289 	if (adapter->netdev->features & NETIF_F_RXCSUM)
3290 		rxcsum |= E1000_RXCSUM_TUOFL;
3291 	else
3292 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3293 	ew32(RXCSUM, rxcsum);
3294 
3295 	/* With jumbo frames, excessive C-state transition latencies result
3296 	 * in dropped transactions.
3297 	 */
3298 	if (adapter->netdev->mtu > ETH_DATA_LEN) {
3299 		u32 lat =
3300 		    ((er32(PBA) & E1000_PBA_RXA_MASK) * 1024 -
3301 		     adapter->max_frame_size) * 8 / 1000;
3302 
3303 		if (adapter->flags & FLAG_IS_ICH) {
3304 			u32 rxdctl = er32(RXDCTL(0));
3305 
3306 			ew32(RXDCTL(0), rxdctl | 0x3 | BIT(8));
3307 		}
3308 
3309 		dev_info(&adapter->pdev->dev,
3310 			 "Some CPU C-states have been disabled in order to enable jumbo frames\n");
3311 		pm_qos_update_request(&adapter->pm_qos_req, lat);
3312 	} else {
3313 		pm_qos_update_request(&adapter->pm_qos_req,
3314 				      PM_QOS_DEFAULT_VALUE);
3315 	}
3316 
3317 	/* Enable Receives */
3318 	ew32(RCTL, rctl);
3319 }
3320 
3321 /**
3322  * e1000e_write_mc_addr_list - write multicast addresses to MTA
3323  * @netdev: network interface device structure
3324  *
3325  * Writes multicast address list to the MTA hash table.
3326  * Returns: -ENOMEM on failure
3327  *                0 on no addresses written
3328  *                X on writing X addresses to MTA
3329  */
3330 static int e1000e_write_mc_addr_list(struct net_device *netdev)
3331 {
3332 	struct e1000_adapter *adapter = netdev_priv(netdev);
3333 	struct e1000_hw *hw = &adapter->hw;
3334 	struct netdev_hw_addr *ha;
3335 	u8 *mta_list;
3336 	int i;
3337 
3338 	if (netdev_mc_empty(netdev)) {
3339 		/* nothing to program, so clear mc list */
3340 		hw->mac.ops.update_mc_addr_list(hw, NULL, 0);
3341 		return 0;
3342 	}
3343 
3344 	mta_list = kzalloc(netdev_mc_count(netdev) * ETH_ALEN, GFP_ATOMIC);
3345 	if (!mta_list)
3346 		return -ENOMEM;
3347 
3348 	/* update_mc_addr_list expects a packed array of only addresses. */
3349 	i = 0;
3350 	netdev_for_each_mc_addr(ha, netdev)
3351 	    memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3352 
3353 	hw->mac.ops.update_mc_addr_list(hw, mta_list, i);
3354 	kfree(mta_list);
3355 
3356 	return netdev_mc_count(netdev);
3357 }
3358 
3359 /**
3360  * e1000e_write_uc_addr_list - write unicast addresses to RAR table
3361  * @netdev: network interface device structure
3362  *
3363  * Writes unicast address list to the RAR table.
3364  * Returns: -ENOMEM on failure/insufficient address space
3365  *                0 on no addresses written
3366  *                X on writing X addresses to the RAR table
3367  **/
3368 static int e1000e_write_uc_addr_list(struct net_device *netdev)
3369 {
3370 	struct e1000_adapter *adapter = netdev_priv(netdev);
3371 	struct e1000_hw *hw = &adapter->hw;
3372 	unsigned int rar_entries;
3373 	int count = 0;
3374 
3375 	rar_entries = hw->mac.ops.rar_get_count(hw);
3376 
3377 	/* save a rar entry for our hardware address */
3378 	rar_entries--;
3379 
3380 	/* save a rar entry for the LAA workaround */
3381 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA)
3382 		rar_entries--;
3383 
3384 	/* return ENOMEM indicating insufficient memory for addresses */
3385 	if (netdev_uc_count(netdev) > rar_entries)
3386 		return -ENOMEM;
3387 
3388 	if (!netdev_uc_empty(netdev) && rar_entries) {
3389 		struct netdev_hw_addr *ha;
3390 
3391 		/* write the addresses in reverse order to avoid write
3392 		 * combining
3393 		 */
3394 		netdev_for_each_uc_addr(ha, netdev) {
3395 			int ret_val;
3396 
3397 			if (!rar_entries)
3398 				break;
3399 			ret_val = hw->mac.ops.rar_set(hw, ha->addr, rar_entries--);
3400 			if (ret_val < 0)
3401 				return -ENOMEM;
3402 			count++;
3403 		}
3404 	}
3405 
3406 	/* zero out the remaining RAR entries not used above */
3407 	for (; rar_entries > 0; rar_entries--) {
3408 		ew32(RAH(rar_entries), 0);
3409 		ew32(RAL(rar_entries), 0);
3410 	}
3411 	e1e_flush();
3412 
3413 	return count;
3414 }
3415 
3416 /**
3417  * e1000e_set_rx_mode - secondary unicast, Multicast and Promiscuous mode set
3418  * @netdev: network interface device structure
3419  *
3420  * The ndo_set_rx_mode entry point is called whenever the unicast or multicast
3421  * address list or the network interface flags are updated.  This routine is
3422  * responsible for configuring the hardware for proper unicast, multicast,
3423  * promiscuous mode, and all-multi behavior.
3424  **/
3425 static void e1000e_set_rx_mode(struct net_device *netdev)
3426 {
3427 	struct e1000_adapter *adapter = netdev_priv(netdev);
3428 	struct e1000_hw *hw = &adapter->hw;
3429 	u32 rctl;
3430 
3431 	if (pm_runtime_suspended(netdev->dev.parent))
3432 		return;
3433 
3434 	/* Check for Promiscuous and All Multicast modes */
3435 	rctl = er32(RCTL);
3436 
3437 	/* clear the affected bits */
3438 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
3439 
3440 	if (netdev->flags & IFF_PROMISC) {
3441 		rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
3442 		/* Do not hardware filter VLANs in promisc mode */
3443 		e1000e_vlan_filter_disable(adapter);
3444 	} else {
3445 		int count;
3446 
3447 		if (netdev->flags & IFF_ALLMULTI) {
3448 			rctl |= E1000_RCTL_MPE;
3449 		} else {
3450 			/* Write addresses to the MTA, if the attempt fails
3451 			 * then we should just turn on promiscuous mode so
3452 			 * that we can at least receive multicast traffic
3453 			 */
3454 			count = e1000e_write_mc_addr_list(netdev);
3455 			if (count < 0)
3456 				rctl |= E1000_RCTL_MPE;
3457 		}
3458 		e1000e_vlan_filter_enable(adapter);
3459 		/* Write addresses to available RAR registers, if there is not
3460 		 * sufficient space to store all the addresses then enable
3461 		 * unicast promiscuous mode
3462 		 */
3463 		count = e1000e_write_uc_addr_list(netdev);
3464 		if (count < 0)
3465 			rctl |= E1000_RCTL_UPE;
3466 	}
3467 
3468 	ew32(RCTL, rctl);
3469 
3470 	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
3471 		e1000e_vlan_strip_enable(adapter);
3472 	else
3473 		e1000e_vlan_strip_disable(adapter);
3474 }
3475 
3476 static void e1000e_setup_rss_hash(struct e1000_adapter *adapter)
3477 {
3478 	struct e1000_hw *hw = &adapter->hw;
3479 	u32 mrqc, rxcsum;
3480 	u32 rss_key[10];
3481 	int i;
3482 
3483 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3484 	for (i = 0; i < 10; i++)
3485 		ew32(RSSRK(i), rss_key[i]);
3486 
3487 	/* Direct all traffic to queue 0 */
3488 	for (i = 0; i < 32; i++)
3489 		ew32(RETA(i), 0);
3490 
3491 	/* Disable raw packet checksumming so that RSS hash is placed in
3492 	 * descriptor on writeback.
3493 	 */
3494 	rxcsum = er32(RXCSUM);
3495 	rxcsum |= E1000_RXCSUM_PCSD;
3496 
3497 	ew32(RXCSUM, rxcsum);
3498 
3499 	mrqc = (E1000_MRQC_RSS_FIELD_IPV4 |
3500 		E1000_MRQC_RSS_FIELD_IPV4_TCP |
3501 		E1000_MRQC_RSS_FIELD_IPV6 |
3502 		E1000_MRQC_RSS_FIELD_IPV6_TCP |
3503 		E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3504 
3505 	ew32(MRQC, mrqc);
3506 }
3507 
3508 /**
3509  * e1000e_get_base_timinca - get default SYSTIM time increment attributes
3510  * @adapter: board private structure
3511  * @timinca: pointer to returned time increment attributes
3512  *
3513  * Get attributes for incrementing the System Time Register SYSTIML/H at
3514  * the default base frequency, and set the cyclecounter shift value.
3515  **/
3516 s32 e1000e_get_base_timinca(struct e1000_adapter *adapter, u32 *timinca)
3517 {
3518 	struct e1000_hw *hw = &adapter->hw;
3519 	u32 incvalue, incperiod, shift;
3520 
3521 	/* Make sure clock is enabled on I217/I218/I219  before checking
3522 	 * the frequency
3523 	 */
3524 	if ((hw->mac.type >= e1000_pch_lpt) &&
3525 	    !(er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) &&
3526 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_ENABLED)) {
3527 		u32 fextnvm7 = er32(FEXTNVM7);
3528 
3529 		if (!(fextnvm7 & BIT(0))) {
3530 			ew32(FEXTNVM7, fextnvm7 | BIT(0));
3531 			e1e_flush();
3532 		}
3533 	}
3534 
3535 	switch (hw->mac.type) {
3536 	case e1000_pch2lan:
3537 		/* Stable 96MHz frequency */
3538 		incperiod = INCPERIOD_96MHZ;
3539 		incvalue = INCVALUE_96MHZ;
3540 		shift = INCVALUE_SHIFT_96MHZ;
3541 		adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3542 		break;
3543 	case e1000_pch_lpt:
3544 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3545 			/* Stable 96MHz frequency */
3546 			incperiod = INCPERIOD_96MHZ;
3547 			incvalue = INCVALUE_96MHZ;
3548 			shift = INCVALUE_SHIFT_96MHZ;
3549 			adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3550 		} else {
3551 			/* Stable 25MHz frequency */
3552 			incperiod = INCPERIOD_25MHZ;
3553 			incvalue = INCVALUE_25MHZ;
3554 			shift = INCVALUE_SHIFT_25MHZ;
3555 			adapter->cc.shift = shift;
3556 		}
3557 		break;
3558 	case e1000_pch_spt:
3559 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3560 			/* Stable 24MHz frequency */
3561 			incperiod = INCPERIOD_24MHZ;
3562 			incvalue = INCVALUE_24MHZ;
3563 			shift = INCVALUE_SHIFT_24MHZ;
3564 			adapter->cc.shift = shift;
3565 			break;
3566 		}
3567 		return -EINVAL;
3568 	case e1000_pch_cnp:
3569 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3570 			/* Stable 24MHz frequency */
3571 			incperiod = INCPERIOD_24MHZ;
3572 			incvalue = INCVALUE_24MHZ;
3573 			shift = INCVALUE_SHIFT_24MHZ;
3574 			adapter->cc.shift = shift;
3575 		} else {
3576 			/* Stable 38400KHz frequency */
3577 			incperiod = INCPERIOD_38400KHZ;
3578 			incvalue = INCVALUE_38400KHZ;
3579 			shift = INCVALUE_SHIFT_38400KHZ;
3580 			adapter->cc.shift = shift;
3581 		}
3582 		break;
3583 	case e1000_82574:
3584 	case e1000_82583:
3585 		/* Stable 25MHz frequency */
3586 		incperiod = INCPERIOD_25MHZ;
3587 		incvalue = INCVALUE_25MHZ;
3588 		shift = INCVALUE_SHIFT_25MHZ;
3589 		adapter->cc.shift = shift;
3590 		break;
3591 	default:
3592 		return -EINVAL;
3593 	}
3594 
3595 	*timinca = ((incperiod << E1000_TIMINCA_INCPERIOD_SHIFT) |
3596 		    ((incvalue << shift) & E1000_TIMINCA_INCVALUE_MASK));
3597 
3598 	return 0;
3599 }
3600 
3601 /**
3602  * e1000e_config_hwtstamp - configure the hwtstamp registers and enable/disable
3603  * @adapter: board private structure
3604  *
3605  * Outgoing time stamping can be enabled and disabled. Play nice and
3606  * disable it when requested, although it shouldn't cause any overhead
3607  * when no packet needs it. At most one packet in the queue may be
3608  * marked for time stamping, otherwise it would be impossible to tell
3609  * for sure to which packet the hardware time stamp belongs.
3610  *
3611  * Incoming time stamping has to be configured via the hardware filters.
3612  * Not all combinations are supported, in particular event type has to be
3613  * specified. Matching the kind of event packet is not supported, with the
3614  * exception of "all V2 events regardless of level 2 or 4".
3615  **/
3616 static int e1000e_config_hwtstamp(struct e1000_adapter *adapter,
3617 				  struct hwtstamp_config *config)
3618 {
3619 	struct e1000_hw *hw = &adapter->hw;
3620 	u32 tsync_tx_ctl = E1000_TSYNCTXCTL_ENABLED;
3621 	u32 tsync_rx_ctl = E1000_TSYNCRXCTL_ENABLED;
3622 	u32 rxmtrl = 0;
3623 	u16 rxudp = 0;
3624 	bool is_l4 = false;
3625 	bool is_l2 = false;
3626 	u32 regval;
3627 
3628 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3629 		return -EINVAL;
3630 
3631 	/* flags reserved for future extensions - must be zero */
3632 	if (config->flags)
3633 		return -EINVAL;
3634 
3635 	switch (config->tx_type) {
3636 	case HWTSTAMP_TX_OFF:
3637 		tsync_tx_ctl = 0;
3638 		break;
3639 	case HWTSTAMP_TX_ON:
3640 		break;
3641 	default:
3642 		return -ERANGE;
3643 	}
3644 
3645 	switch (config->rx_filter) {
3646 	case HWTSTAMP_FILTER_NONE:
3647 		tsync_rx_ctl = 0;
3648 		break;
3649 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3650 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3651 		rxmtrl = E1000_RXMTRL_PTP_V1_SYNC_MESSAGE;
3652 		is_l4 = true;
3653 		break;
3654 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3655 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3656 		rxmtrl = E1000_RXMTRL_PTP_V1_DELAY_REQ_MESSAGE;
3657 		is_l4 = true;
3658 		break;
3659 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3660 		/* Also time stamps V2 L2 Path Delay Request/Response */
3661 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3662 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3663 		is_l2 = true;
3664 		break;
3665 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3666 		/* Also time stamps V2 L2 Path Delay Request/Response. */
3667 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3668 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3669 		is_l2 = true;
3670 		break;
3671 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3672 		/* Hardware cannot filter just V2 L4 Sync messages;
3673 		 * fall-through to V2 (both L2 and L4) Sync.
3674 		 */
3675 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
3676 		/* Also time stamps V2 Path Delay Request/Response. */
3677 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3678 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3679 		is_l2 = true;
3680 		is_l4 = true;
3681 		break;
3682 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3683 		/* Hardware cannot filter just V2 L4 Delay Request messages;
3684 		 * fall-through to V2 (both L2 and L4) Delay Request.
3685 		 */
3686 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
3687 		/* Also time stamps V2 Path Delay Request/Response. */
3688 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3689 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3690 		is_l2 = true;
3691 		is_l4 = true;
3692 		break;
3693 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
3694 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
3695 		/* Hardware cannot filter just V2 L4 or L2 Event messages;
3696 		 * fall-through to all V2 (both L2 and L4) Events.
3697 		 */
3698 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
3699 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_EVENT_V2;
3700 		config->rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
3701 		is_l2 = true;
3702 		is_l4 = true;
3703 		break;
3704 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3705 		/* For V1, the hardware can only filter Sync messages or
3706 		 * Delay Request messages but not both so fall-through to
3707 		 * time stamp all packets.
3708 		 */
3709 	case HWTSTAMP_FILTER_NTP_ALL:
3710 	case HWTSTAMP_FILTER_ALL:
3711 		is_l2 = true;
3712 		is_l4 = true;
3713 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_ALL;
3714 		config->rx_filter = HWTSTAMP_FILTER_ALL;
3715 		break;
3716 	default:
3717 		return -ERANGE;
3718 	}
3719 
3720 	adapter->hwtstamp_config = *config;
3721 
3722 	/* enable/disable Tx h/w time stamping */
3723 	regval = er32(TSYNCTXCTL);
3724 	regval &= ~E1000_TSYNCTXCTL_ENABLED;
3725 	regval |= tsync_tx_ctl;
3726 	ew32(TSYNCTXCTL, regval);
3727 	if ((er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) !=
3728 	    (regval & E1000_TSYNCTXCTL_ENABLED)) {
3729 		e_err("Timesync Tx Control register not set as expected\n");
3730 		return -EAGAIN;
3731 	}
3732 
3733 	/* enable/disable Rx h/w time stamping */
3734 	regval = er32(TSYNCRXCTL);
3735 	regval &= ~(E1000_TSYNCRXCTL_ENABLED | E1000_TSYNCRXCTL_TYPE_MASK);
3736 	regval |= tsync_rx_ctl;
3737 	ew32(TSYNCRXCTL, regval);
3738 	if ((er32(TSYNCRXCTL) & (E1000_TSYNCRXCTL_ENABLED |
3739 				 E1000_TSYNCRXCTL_TYPE_MASK)) !=
3740 	    (regval & (E1000_TSYNCRXCTL_ENABLED |
3741 		       E1000_TSYNCRXCTL_TYPE_MASK))) {
3742 		e_err("Timesync Rx Control register not set as expected\n");
3743 		return -EAGAIN;
3744 	}
3745 
3746 	/* L2: define ethertype filter for time stamped packets */
3747 	if (is_l2)
3748 		rxmtrl |= ETH_P_1588;
3749 
3750 	/* define which PTP packets get time stamped */
3751 	ew32(RXMTRL, rxmtrl);
3752 
3753 	/* Filter by destination port */
3754 	if (is_l4) {
3755 		rxudp = PTP_EV_PORT;
3756 		cpu_to_be16s(&rxudp);
3757 	}
3758 	ew32(RXUDP, rxudp);
3759 
3760 	e1e_flush();
3761 
3762 	/* Clear TSYNCRXCTL_VALID & TSYNCTXCTL_VALID bit */
3763 	er32(RXSTMPH);
3764 	er32(TXSTMPH);
3765 
3766 	return 0;
3767 }
3768 
3769 /**
3770  * e1000_configure - configure the hardware for Rx and Tx
3771  * @adapter: private board structure
3772  **/
3773 static void e1000_configure(struct e1000_adapter *adapter)
3774 {
3775 	struct e1000_ring *rx_ring = adapter->rx_ring;
3776 
3777 	e1000e_set_rx_mode(adapter->netdev);
3778 
3779 	e1000_restore_vlan(adapter);
3780 	e1000_init_manageability_pt(adapter);
3781 
3782 	e1000_configure_tx(adapter);
3783 
3784 	if (adapter->netdev->features & NETIF_F_RXHASH)
3785 		e1000e_setup_rss_hash(adapter);
3786 	e1000_setup_rctl(adapter);
3787 	e1000_configure_rx(adapter);
3788 	adapter->alloc_rx_buf(rx_ring, e1000_desc_unused(rx_ring), GFP_KERNEL);
3789 }
3790 
3791 /**
3792  * e1000e_power_up_phy - restore link in case the phy was powered down
3793  * @adapter: address of board private structure
3794  *
3795  * The phy may be powered down to save power and turn off link when the
3796  * driver is unloaded and wake on lan is not enabled (among others)
3797  * *** this routine MUST be followed by a call to e1000e_reset ***
3798  **/
3799 void e1000e_power_up_phy(struct e1000_adapter *adapter)
3800 {
3801 	if (adapter->hw.phy.ops.power_up)
3802 		adapter->hw.phy.ops.power_up(&adapter->hw);
3803 
3804 	adapter->hw.mac.ops.setup_link(&adapter->hw);
3805 }
3806 
3807 /**
3808  * e1000_power_down_phy - Power down the PHY
3809  *
3810  * Power down the PHY so no link is implied when interface is down.
3811  * The PHY cannot be powered down if management or WoL is active.
3812  */
3813 static void e1000_power_down_phy(struct e1000_adapter *adapter)
3814 {
3815 	if (adapter->hw.phy.ops.power_down)
3816 		adapter->hw.phy.ops.power_down(&adapter->hw);
3817 }
3818 
3819 /**
3820  * e1000_flush_tx_ring - remove all descriptors from the tx_ring
3821  *
3822  * We want to clear all pending descriptors from the TX ring.
3823  * zeroing happens when the HW reads the regs. We  assign the ring itself as
3824  * the data of the next descriptor. We don't care about the data we are about
3825  * to reset the HW.
3826  */
3827 static void e1000_flush_tx_ring(struct e1000_adapter *adapter)
3828 {
3829 	struct e1000_hw *hw = &adapter->hw;
3830 	struct e1000_ring *tx_ring = adapter->tx_ring;
3831 	struct e1000_tx_desc *tx_desc = NULL;
3832 	u32 tdt, tctl, txd_lower = E1000_TXD_CMD_IFCS;
3833 	u16 size = 512;
3834 
3835 	tctl = er32(TCTL);
3836 	ew32(TCTL, tctl | E1000_TCTL_EN);
3837 	tdt = er32(TDT(0));
3838 	BUG_ON(tdt != tx_ring->next_to_use);
3839 	tx_desc =  E1000_TX_DESC(*tx_ring, tx_ring->next_to_use);
3840 	tx_desc->buffer_addr = tx_ring->dma;
3841 
3842 	tx_desc->lower.data = cpu_to_le32(txd_lower | size);
3843 	tx_desc->upper.data = 0;
3844 	/* flush descriptors to memory before notifying the HW */
3845 	wmb();
3846 	tx_ring->next_to_use++;
3847 	if (tx_ring->next_to_use == tx_ring->count)
3848 		tx_ring->next_to_use = 0;
3849 	ew32(TDT(0), tx_ring->next_to_use);
3850 	mmiowb();
3851 	usleep_range(200, 250);
3852 }
3853 
3854 /**
3855  * e1000_flush_rx_ring - remove all descriptors from the rx_ring
3856  *
3857  * Mark all descriptors in the RX ring as consumed and disable the rx ring
3858  */
3859 static void e1000_flush_rx_ring(struct e1000_adapter *adapter)
3860 {
3861 	u32 rctl, rxdctl;
3862 	struct e1000_hw *hw = &adapter->hw;
3863 
3864 	rctl = er32(RCTL);
3865 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3866 	e1e_flush();
3867 	usleep_range(100, 150);
3868 
3869 	rxdctl = er32(RXDCTL(0));
3870 	/* zero the lower 14 bits (prefetch and host thresholds) */
3871 	rxdctl &= 0xffffc000;
3872 
3873 	/* update thresholds: prefetch threshold to 31, host threshold to 1
3874 	 * and make sure the granularity is "descriptors" and not "cache lines"
3875 	 */
3876 	rxdctl |= (0x1F | BIT(8) | E1000_RXDCTL_THRESH_UNIT_DESC);
3877 
3878 	ew32(RXDCTL(0), rxdctl);
3879 	/* momentarily enable the RX ring for the changes to take effect */
3880 	ew32(RCTL, rctl | E1000_RCTL_EN);
3881 	e1e_flush();
3882 	usleep_range(100, 150);
3883 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3884 }
3885 
3886 /**
3887  * e1000_flush_desc_rings - remove all descriptors from the descriptor rings
3888  *
3889  * In i219, the descriptor rings must be emptied before resetting the HW
3890  * or before changing the device state to D3 during runtime (runtime PM).
3891  *
3892  * Failure to do this will cause the HW to enter a unit hang state which can
3893  * only be released by PCI reset on the device
3894  *
3895  */
3896 
3897 static void e1000_flush_desc_rings(struct e1000_adapter *adapter)
3898 {
3899 	u16 hang_state;
3900 	u32 fext_nvm11, tdlen;
3901 	struct e1000_hw *hw = &adapter->hw;
3902 
3903 	/* First, disable MULR fix in FEXTNVM11 */
3904 	fext_nvm11 = er32(FEXTNVM11);
3905 	fext_nvm11 |= E1000_FEXTNVM11_DISABLE_MULR_FIX;
3906 	ew32(FEXTNVM11, fext_nvm11);
3907 	/* do nothing if we're not in faulty state, or if the queue is empty */
3908 	tdlen = er32(TDLEN(0));
3909 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3910 			     &hang_state);
3911 	if (!(hang_state & FLUSH_DESC_REQUIRED) || !tdlen)
3912 		return;
3913 	e1000_flush_tx_ring(adapter);
3914 	/* recheck, maybe the fault is caused by the rx ring */
3915 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3916 			     &hang_state);
3917 	if (hang_state & FLUSH_DESC_REQUIRED)
3918 		e1000_flush_rx_ring(adapter);
3919 }
3920 
3921 /**
3922  * e1000e_systim_reset - reset the timesync registers after a hardware reset
3923  * @adapter: board private structure
3924  *
3925  * When the MAC is reset, all hardware bits for timesync will be reset to the
3926  * default values. This function will restore the settings last in place.
3927  * Since the clock SYSTIME registers are reset, we will simply restore the
3928  * cyclecounter to the kernel real clock time.
3929  **/
3930 static void e1000e_systim_reset(struct e1000_adapter *adapter)
3931 {
3932 	struct ptp_clock_info *info = &adapter->ptp_clock_info;
3933 	struct e1000_hw *hw = &adapter->hw;
3934 	unsigned long flags;
3935 	u32 timinca;
3936 	s32 ret_val;
3937 
3938 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3939 		return;
3940 
3941 	if (info->adjfreq) {
3942 		/* restore the previous ptp frequency delta */
3943 		ret_val = info->adjfreq(info, adapter->ptp_delta);
3944 	} else {
3945 		/* set the default base frequency if no adjustment possible */
3946 		ret_val = e1000e_get_base_timinca(adapter, &timinca);
3947 		if (!ret_val)
3948 			ew32(TIMINCA, timinca);
3949 	}
3950 
3951 	if (ret_val) {
3952 		dev_warn(&adapter->pdev->dev,
3953 			 "Failed to restore TIMINCA clock rate delta: %d\n",
3954 			 ret_val);
3955 		return;
3956 	}
3957 
3958 	/* reset the systim ns time counter */
3959 	spin_lock_irqsave(&adapter->systim_lock, flags);
3960 	timecounter_init(&adapter->tc, &adapter->cc,
3961 			 ktime_to_ns(ktime_get_real()));
3962 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
3963 
3964 	/* restore the previous hwtstamp configuration settings */
3965 	e1000e_config_hwtstamp(adapter, &adapter->hwtstamp_config);
3966 }
3967 
3968 /**
3969  * e1000e_reset - bring the hardware into a known good state
3970  *
3971  * This function boots the hardware and enables some settings that
3972  * require a configuration cycle of the hardware - those cannot be
3973  * set/changed during runtime. After reset the device needs to be
3974  * properly configured for Rx, Tx etc.
3975  */
3976 void e1000e_reset(struct e1000_adapter *adapter)
3977 {
3978 	struct e1000_mac_info *mac = &adapter->hw.mac;
3979 	struct e1000_fc_info *fc = &adapter->hw.fc;
3980 	struct e1000_hw *hw = &adapter->hw;
3981 	u32 tx_space, min_tx_space, min_rx_space;
3982 	u32 pba = adapter->pba;
3983 	u16 hwm;
3984 
3985 	/* reset Packet Buffer Allocation to default */
3986 	ew32(PBA, pba);
3987 
3988 	if (adapter->max_frame_size > (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN)) {
3989 		/* To maintain wire speed transmits, the Tx FIFO should be
3990 		 * large enough to accommodate two full transmit packets,
3991 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
3992 		 * the Rx FIFO should be large enough to accommodate at least
3993 		 * one full receive packet and is similarly rounded up and
3994 		 * expressed in KB.
3995 		 */
3996 		pba = er32(PBA);
3997 		/* upper 16 bits has Tx packet buffer allocation size in KB */
3998 		tx_space = pba >> 16;
3999 		/* lower 16 bits has Rx packet buffer allocation size in KB */
4000 		pba &= 0xffff;
4001 		/* the Tx fifo also stores 16 bytes of information about the Tx
4002 		 * but don't include ethernet FCS because hardware appends it
4003 		 */
4004 		min_tx_space = (adapter->max_frame_size +
4005 				sizeof(struct e1000_tx_desc) - ETH_FCS_LEN) * 2;
4006 		min_tx_space = ALIGN(min_tx_space, 1024);
4007 		min_tx_space >>= 10;
4008 		/* software strips receive CRC, so leave room for it */
4009 		min_rx_space = adapter->max_frame_size;
4010 		min_rx_space = ALIGN(min_rx_space, 1024);
4011 		min_rx_space >>= 10;
4012 
4013 		/* If current Tx allocation is less than the min Tx FIFO size,
4014 		 * and the min Tx FIFO size is less than the current Rx FIFO
4015 		 * allocation, take space away from current Rx allocation
4016 		 */
4017 		if ((tx_space < min_tx_space) &&
4018 		    ((min_tx_space - tx_space) < pba)) {
4019 			pba -= min_tx_space - tx_space;
4020 
4021 			/* if short on Rx space, Rx wins and must trump Tx
4022 			 * adjustment
4023 			 */
4024 			if (pba < min_rx_space)
4025 				pba = min_rx_space;
4026 		}
4027 
4028 		ew32(PBA, pba);
4029 	}
4030 
4031 	/* flow control settings
4032 	 *
4033 	 * The high water mark must be low enough to fit one full frame
4034 	 * (or the size used for early receive) above it in the Rx FIFO.
4035 	 * Set it to the lower of:
4036 	 * - 90% of the Rx FIFO size, and
4037 	 * - the full Rx FIFO size minus one full frame
4038 	 */
4039 	if (adapter->flags & FLAG_DISABLE_FC_PAUSE_TIME)
4040 		fc->pause_time = 0xFFFF;
4041 	else
4042 		fc->pause_time = E1000_FC_PAUSE_TIME;
4043 	fc->send_xon = true;
4044 	fc->current_mode = fc->requested_mode;
4045 
4046 	switch (hw->mac.type) {
4047 	case e1000_ich9lan:
4048 	case e1000_ich10lan:
4049 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4050 			pba = 14;
4051 			ew32(PBA, pba);
4052 			fc->high_water = 0x2800;
4053 			fc->low_water = fc->high_water - 8;
4054 			break;
4055 		}
4056 		/* fall-through */
4057 	default:
4058 		hwm = min(((pba << 10) * 9 / 10),
4059 			  ((pba << 10) - adapter->max_frame_size));
4060 
4061 		fc->high_water = hwm & E1000_FCRTH_RTH;	/* 8-byte granularity */
4062 		fc->low_water = fc->high_water - 8;
4063 		break;
4064 	case e1000_pchlan:
4065 		/* Workaround PCH LOM adapter hangs with certain network
4066 		 * loads.  If hangs persist, try disabling Tx flow control.
4067 		 */
4068 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4069 			fc->high_water = 0x3500;
4070 			fc->low_water = 0x1500;
4071 		} else {
4072 			fc->high_water = 0x5000;
4073 			fc->low_water = 0x3000;
4074 		}
4075 		fc->refresh_time = 0x1000;
4076 		break;
4077 	case e1000_pch2lan:
4078 	case e1000_pch_lpt:
4079 	case e1000_pch_spt:
4080 	case e1000_pch_cnp:
4081 		fc->refresh_time = 0x0400;
4082 
4083 		if (adapter->netdev->mtu <= ETH_DATA_LEN) {
4084 			fc->high_water = 0x05C20;
4085 			fc->low_water = 0x05048;
4086 			fc->pause_time = 0x0650;
4087 			break;
4088 		}
4089 
4090 		pba = 14;
4091 		ew32(PBA, pba);
4092 		fc->high_water = ((pba << 10) * 9 / 10) & E1000_FCRTH_RTH;
4093 		fc->low_water = ((pba << 10) * 8 / 10) & E1000_FCRTL_RTL;
4094 		break;
4095 	}
4096 
4097 	/* Alignment of Tx data is on an arbitrary byte boundary with the
4098 	 * maximum size per Tx descriptor limited only to the transmit
4099 	 * allocation of the packet buffer minus 96 bytes with an upper
4100 	 * limit of 24KB due to receive synchronization limitations.
4101 	 */
4102 	adapter->tx_fifo_limit = min_t(u32, ((er32(PBA) >> 16) << 10) - 96,
4103 				       24 << 10);
4104 
4105 	/* Disable Adaptive Interrupt Moderation if 2 full packets cannot
4106 	 * fit in receive buffer.
4107 	 */
4108 	if (adapter->itr_setting & 0x3) {
4109 		if ((adapter->max_frame_size * 2) > (pba << 10)) {
4110 			if (!(adapter->flags2 & FLAG2_DISABLE_AIM)) {
4111 				dev_info(&adapter->pdev->dev,
4112 					 "Interrupt Throttle Rate off\n");
4113 				adapter->flags2 |= FLAG2_DISABLE_AIM;
4114 				e1000e_write_itr(adapter, 0);
4115 			}
4116 		} else if (adapter->flags2 & FLAG2_DISABLE_AIM) {
4117 			dev_info(&adapter->pdev->dev,
4118 				 "Interrupt Throttle Rate on\n");
4119 			adapter->flags2 &= ~FLAG2_DISABLE_AIM;
4120 			adapter->itr = 20000;
4121 			e1000e_write_itr(adapter, adapter->itr);
4122 		}
4123 	}
4124 
4125 	if (hw->mac.type >= e1000_pch_spt)
4126 		e1000_flush_desc_rings(adapter);
4127 	/* Allow time for pending master requests to run */
4128 	mac->ops.reset_hw(hw);
4129 
4130 	/* For parts with AMT enabled, let the firmware know
4131 	 * that the network interface is in control
4132 	 */
4133 	if (adapter->flags & FLAG_HAS_AMT)
4134 		e1000e_get_hw_control(adapter);
4135 
4136 	ew32(WUC, 0);
4137 
4138 	if (mac->ops.init_hw(hw))
4139 		e_err("Hardware Error\n");
4140 
4141 	e1000_update_mng_vlan(adapter);
4142 
4143 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
4144 	ew32(VET, ETH_P_8021Q);
4145 
4146 	e1000e_reset_adaptive(hw);
4147 
4148 	/* restore systim and hwtstamp settings */
4149 	e1000e_systim_reset(adapter);
4150 
4151 	/* Set EEE advertisement as appropriate */
4152 	if (adapter->flags2 & FLAG2_HAS_EEE) {
4153 		s32 ret_val;
4154 		u16 adv_addr;
4155 
4156 		switch (hw->phy.type) {
4157 		case e1000_phy_82579:
4158 			adv_addr = I82579_EEE_ADVERTISEMENT;
4159 			break;
4160 		case e1000_phy_i217:
4161 			adv_addr = I217_EEE_ADVERTISEMENT;
4162 			break;
4163 		default:
4164 			dev_err(&adapter->pdev->dev,
4165 				"Invalid PHY type setting EEE advertisement\n");
4166 			return;
4167 		}
4168 
4169 		ret_val = hw->phy.ops.acquire(hw);
4170 		if (ret_val) {
4171 			dev_err(&adapter->pdev->dev,
4172 				"EEE advertisement - unable to acquire PHY\n");
4173 			return;
4174 		}
4175 
4176 		e1000_write_emi_reg_locked(hw, adv_addr,
4177 					   hw->dev_spec.ich8lan.eee_disable ?
4178 					   0 : adapter->eee_advert);
4179 
4180 		hw->phy.ops.release(hw);
4181 	}
4182 
4183 	if (!netif_running(adapter->netdev) &&
4184 	    !test_bit(__E1000_TESTING, &adapter->state))
4185 		e1000_power_down_phy(adapter);
4186 
4187 	e1000_get_phy_info(hw);
4188 
4189 	if ((adapter->flags & FLAG_HAS_SMART_POWER_DOWN) &&
4190 	    !(adapter->flags & FLAG_SMART_POWER_DOWN)) {
4191 		u16 phy_data = 0;
4192 		/* speed up time to link by disabling smart power down, ignore
4193 		 * the return value of this function because there is nothing
4194 		 * different we would do if it failed
4195 		 */
4196 		e1e_rphy(hw, IGP02E1000_PHY_POWER_MGMT, &phy_data);
4197 		phy_data &= ~IGP02E1000_PM_SPD;
4198 		e1e_wphy(hw, IGP02E1000_PHY_POWER_MGMT, phy_data);
4199 	}
4200 	if (hw->mac.type >= e1000_pch_spt && adapter->int_mode == 0) {
4201 		u32 reg;
4202 
4203 		/* Fextnvm7 @ 0xe4[2] = 1 */
4204 		reg = er32(FEXTNVM7);
4205 		reg |= E1000_FEXTNVM7_SIDE_CLK_UNGATE;
4206 		ew32(FEXTNVM7, reg);
4207 		/* Fextnvm9 @ 0x5bb4[13:12] = 11 */
4208 		reg = er32(FEXTNVM9);
4209 		reg |= E1000_FEXTNVM9_IOSFSB_CLKGATE_DIS |
4210 		       E1000_FEXTNVM9_IOSFSB_CLKREQ_DIS;
4211 		ew32(FEXTNVM9, reg);
4212 	}
4213 
4214 }
4215 
4216 /**
4217  * e1000e_trigger_lsc - trigger an LSC interrupt
4218  * @adapter:
4219  *
4220  * Fire a link status change interrupt to start the watchdog.
4221  **/
4222 static void e1000e_trigger_lsc(struct e1000_adapter *adapter)
4223 {
4224 	struct e1000_hw *hw = &adapter->hw;
4225 
4226 	if (adapter->msix_entries)
4227 		ew32(ICS, E1000_ICS_LSC | E1000_ICS_OTHER);
4228 	else
4229 		ew32(ICS, E1000_ICS_LSC);
4230 }
4231 
4232 void e1000e_up(struct e1000_adapter *adapter)
4233 {
4234 	/* hardware has been reset, we need to reload some things */
4235 	e1000_configure(adapter);
4236 
4237 	clear_bit(__E1000_DOWN, &adapter->state);
4238 
4239 	if (adapter->msix_entries)
4240 		e1000_configure_msix(adapter);
4241 	e1000_irq_enable(adapter);
4242 
4243 	netif_start_queue(adapter->netdev);
4244 
4245 	e1000e_trigger_lsc(adapter);
4246 }
4247 
4248 static void e1000e_flush_descriptors(struct e1000_adapter *adapter)
4249 {
4250 	struct e1000_hw *hw = &adapter->hw;
4251 
4252 	if (!(adapter->flags2 & FLAG2_DMA_BURST))
4253 		return;
4254 
4255 	/* flush pending descriptor writebacks to memory */
4256 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4257 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4258 
4259 	/* execute the writes immediately */
4260 	e1e_flush();
4261 
4262 	/* due to rare timing issues, write to TIDV/RDTR again to ensure the
4263 	 * write is successful
4264 	 */
4265 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4266 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4267 
4268 	/* execute the writes immediately */
4269 	e1e_flush();
4270 }
4271 
4272 static void e1000e_update_stats(struct e1000_adapter *adapter);
4273 
4274 /**
4275  * e1000e_down - quiesce the device and optionally reset the hardware
4276  * @adapter: board private structure
4277  * @reset: boolean flag to reset the hardware or not
4278  */
4279 void e1000e_down(struct e1000_adapter *adapter, bool reset)
4280 {
4281 	struct net_device *netdev = adapter->netdev;
4282 	struct e1000_hw *hw = &adapter->hw;
4283 	u32 tctl, rctl;
4284 
4285 	/* signal that we're down so the interrupt handler does not
4286 	 * reschedule our watchdog timer
4287 	 */
4288 	set_bit(__E1000_DOWN, &adapter->state);
4289 
4290 	netif_carrier_off(netdev);
4291 
4292 	/* disable receives in the hardware */
4293 	rctl = er32(RCTL);
4294 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
4295 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
4296 	/* flush and sleep below */
4297 
4298 	netif_stop_queue(netdev);
4299 
4300 	/* disable transmits in the hardware */
4301 	tctl = er32(TCTL);
4302 	tctl &= ~E1000_TCTL_EN;
4303 	ew32(TCTL, tctl);
4304 
4305 	/* flush both disables and wait for them to finish */
4306 	e1e_flush();
4307 	usleep_range(10000, 20000);
4308 
4309 	e1000_irq_disable(adapter);
4310 
4311 	napi_synchronize(&adapter->napi);
4312 
4313 	del_timer_sync(&adapter->watchdog_timer);
4314 	del_timer_sync(&adapter->phy_info_timer);
4315 
4316 	spin_lock(&adapter->stats64_lock);
4317 	e1000e_update_stats(adapter);
4318 	spin_unlock(&adapter->stats64_lock);
4319 
4320 	e1000e_flush_descriptors(adapter);
4321 
4322 	adapter->link_speed = 0;
4323 	adapter->link_duplex = 0;
4324 
4325 	/* Disable Si errata workaround on PCHx for jumbo frame flow */
4326 	if ((hw->mac.type >= e1000_pch2lan) &&
4327 	    (adapter->netdev->mtu > ETH_DATA_LEN) &&
4328 	    e1000_lv_jumbo_workaround_ich8lan(hw, false))
4329 		e_dbg("failed to disable jumbo frame workaround mode\n");
4330 
4331 	if (!pci_channel_offline(adapter->pdev)) {
4332 		if (reset)
4333 			e1000e_reset(adapter);
4334 		else if (hw->mac.type >= e1000_pch_spt)
4335 			e1000_flush_desc_rings(adapter);
4336 	}
4337 	e1000_clean_tx_ring(adapter->tx_ring);
4338 	e1000_clean_rx_ring(adapter->rx_ring);
4339 }
4340 
4341 void e1000e_reinit_locked(struct e1000_adapter *adapter)
4342 {
4343 	might_sleep();
4344 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
4345 		usleep_range(1000, 2000);
4346 	e1000e_down(adapter, true);
4347 	e1000e_up(adapter);
4348 	clear_bit(__E1000_RESETTING, &adapter->state);
4349 }
4350 
4351 /**
4352  * e1000e_sanitize_systim - sanitize raw cycle counter reads
4353  * @hw: pointer to the HW structure
4354  * @systim: time value read, sanitized and returned
4355  *
4356  * Errata for 82574/82583 possible bad bits read from SYSTIMH/L:
4357  * check to see that the time is incrementing at a reasonable
4358  * rate and is a multiple of incvalue.
4359  **/
4360 static u64 e1000e_sanitize_systim(struct e1000_hw *hw, u64 systim)
4361 {
4362 	u64 time_delta, rem, temp;
4363 	u64 systim_next;
4364 	u32 incvalue;
4365 	int i;
4366 
4367 	incvalue = er32(TIMINCA) & E1000_TIMINCA_INCVALUE_MASK;
4368 	for (i = 0; i < E1000_MAX_82574_SYSTIM_REREADS; i++) {
4369 		/* latch SYSTIMH on read of SYSTIML */
4370 		systim_next = (u64)er32(SYSTIML);
4371 		systim_next |= (u64)er32(SYSTIMH) << 32;
4372 
4373 		time_delta = systim_next - systim;
4374 		temp = time_delta;
4375 		/* VMWare users have seen incvalue of zero, don't div / 0 */
4376 		rem = incvalue ? do_div(temp, incvalue) : (time_delta != 0);
4377 
4378 		systim = systim_next;
4379 
4380 		if ((time_delta < E1000_82574_SYSTIM_EPSILON) && (rem == 0))
4381 			break;
4382 	}
4383 
4384 	return systim;
4385 }
4386 
4387 /**
4388  * e1000e_cyclecounter_read - read raw cycle counter (used by time counter)
4389  * @cc: cyclecounter structure
4390  **/
4391 static u64 e1000e_cyclecounter_read(const struct cyclecounter *cc)
4392 {
4393 	struct e1000_adapter *adapter = container_of(cc, struct e1000_adapter,
4394 						     cc);
4395 	struct e1000_hw *hw = &adapter->hw;
4396 	u32 systimel, systimeh;
4397 	u64 systim;
4398 	/* SYSTIMH latching upon SYSTIML read does not work well.
4399 	 * This means that if SYSTIML overflows after we read it but before
4400 	 * we read SYSTIMH, the value of SYSTIMH has been incremented and we
4401 	 * will experience a huge non linear increment in the systime value
4402 	 * to fix that we test for overflow and if true, we re-read systime.
4403 	 */
4404 	systimel = er32(SYSTIML);
4405 	systimeh = er32(SYSTIMH);
4406 	/* Is systimel is so large that overflow is possible? */
4407 	if (systimel >= (u32)0xffffffff - E1000_TIMINCA_INCVALUE_MASK) {
4408 		u32 systimel_2 = er32(SYSTIML);
4409 		if (systimel > systimel_2) {
4410 			/* There was an overflow, read again SYSTIMH, and use
4411 			 * systimel_2
4412 			 */
4413 			systimeh = er32(SYSTIMH);
4414 			systimel = systimel_2;
4415 		}
4416 	}
4417 	systim = (u64)systimel;
4418 	systim |= (u64)systimeh << 32;
4419 
4420 	if (adapter->flags2 & FLAG2_CHECK_SYSTIM_OVERFLOW)
4421 		systim = e1000e_sanitize_systim(hw, systim);
4422 
4423 	return systim;
4424 }
4425 
4426 /**
4427  * e1000_sw_init - Initialize general software structures (struct e1000_adapter)
4428  * @adapter: board private structure to initialize
4429  *
4430  * e1000_sw_init initializes the Adapter private data structure.
4431  * Fields are initialized based on PCI device information and
4432  * OS network device settings (MTU size).
4433  **/
4434 static int e1000_sw_init(struct e1000_adapter *adapter)
4435 {
4436 	struct net_device *netdev = adapter->netdev;
4437 
4438 	adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
4439 	adapter->rx_ps_bsize0 = 128;
4440 	adapter->max_frame_size = netdev->mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
4441 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4442 	adapter->tx_ring_count = E1000_DEFAULT_TXD;
4443 	adapter->rx_ring_count = E1000_DEFAULT_RXD;
4444 
4445 	spin_lock_init(&adapter->stats64_lock);
4446 
4447 	e1000e_set_interrupt_capability(adapter);
4448 
4449 	if (e1000_alloc_queues(adapter))
4450 		return -ENOMEM;
4451 
4452 	/* Setup hardware time stamping cyclecounter */
4453 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
4454 		adapter->cc.read = e1000e_cyclecounter_read;
4455 		adapter->cc.mask = CYCLECOUNTER_MASK(64);
4456 		adapter->cc.mult = 1;
4457 		/* cc.shift set in e1000e_get_base_tininca() */
4458 
4459 		spin_lock_init(&adapter->systim_lock);
4460 		INIT_WORK(&adapter->tx_hwtstamp_work, e1000e_tx_hwtstamp_work);
4461 	}
4462 
4463 	/* Explicitly disable IRQ since the NIC can be in any state. */
4464 	e1000_irq_disable(adapter);
4465 
4466 	set_bit(__E1000_DOWN, &adapter->state);
4467 	return 0;
4468 }
4469 
4470 /**
4471  * e1000_intr_msi_test - Interrupt Handler
4472  * @irq: interrupt number
4473  * @data: pointer to a network interface device structure
4474  **/
4475 static irqreturn_t e1000_intr_msi_test(int __always_unused irq, void *data)
4476 {
4477 	struct net_device *netdev = data;
4478 	struct e1000_adapter *adapter = netdev_priv(netdev);
4479 	struct e1000_hw *hw = &adapter->hw;
4480 	u32 icr = er32(ICR);
4481 
4482 	e_dbg("icr is %08X\n", icr);
4483 	if (icr & E1000_ICR_RXSEQ) {
4484 		adapter->flags &= ~FLAG_MSI_TEST_FAILED;
4485 		/* Force memory writes to complete before acknowledging the
4486 		 * interrupt is handled.
4487 		 */
4488 		wmb();
4489 	}
4490 
4491 	return IRQ_HANDLED;
4492 }
4493 
4494 /**
4495  * e1000_test_msi_interrupt - Returns 0 for successful test
4496  * @adapter: board private struct
4497  *
4498  * code flow taken from tg3.c
4499  **/
4500 static int e1000_test_msi_interrupt(struct e1000_adapter *adapter)
4501 {
4502 	struct net_device *netdev = adapter->netdev;
4503 	struct e1000_hw *hw = &adapter->hw;
4504 	int err;
4505 
4506 	/* poll_enable hasn't been called yet, so don't need disable */
4507 	/* clear any pending events */
4508 	er32(ICR);
4509 
4510 	/* free the real vector and request a test handler */
4511 	e1000_free_irq(adapter);
4512 	e1000e_reset_interrupt_capability(adapter);
4513 
4514 	/* Assume that the test fails, if it succeeds then the test
4515 	 * MSI irq handler will unset this flag
4516 	 */
4517 	adapter->flags |= FLAG_MSI_TEST_FAILED;
4518 
4519 	err = pci_enable_msi(adapter->pdev);
4520 	if (err)
4521 		goto msi_test_failed;
4522 
4523 	err = request_irq(adapter->pdev->irq, e1000_intr_msi_test, 0,
4524 			  netdev->name, netdev);
4525 	if (err) {
4526 		pci_disable_msi(adapter->pdev);
4527 		goto msi_test_failed;
4528 	}
4529 
4530 	/* Force memory writes to complete before enabling and firing an
4531 	 * interrupt.
4532 	 */
4533 	wmb();
4534 
4535 	e1000_irq_enable(adapter);
4536 
4537 	/* fire an unusual interrupt on the test handler */
4538 	ew32(ICS, E1000_ICS_RXSEQ);
4539 	e1e_flush();
4540 	msleep(100);
4541 
4542 	e1000_irq_disable(adapter);
4543 
4544 	rmb();			/* read flags after interrupt has been fired */
4545 
4546 	if (adapter->flags & FLAG_MSI_TEST_FAILED) {
4547 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
4548 		e_info("MSI interrupt test failed, using legacy interrupt.\n");
4549 	} else {
4550 		e_dbg("MSI interrupt test succeeded!\n");
4551 	}
4552 
4553 	free_irq(adapter->pdev->irq, netdev);
4554 	pci_disable_msi(adapter->pdev);
4555 
4556 msi_test_failed:
4557 	e1000e_set_interrupt_capability(adapter);
4558 	return e1000_request_irq(adapter);
4559 }
4560 
4561 /**
4562  * e1000_test_msi - Returns 0 if MSI test succeeds or INTx mode is restored
4563  * @adapter: board private struct
4564  *
4565  * code flow taken from tg3.c, called with e1000 interrupts disabled.
4566  **/
4567 static int e1000_test_msi(struct e1000_adapter *adapter)
4568 {
4569 	int err;
4570 	u16 pci_cmd;
4571 
4572 	if (!(adapter->flags & FLAG_MSI_ENABLED))
4573 		return 0;
4574 
4575 	/* disable SERR in case the MSI write causes a master abort */
4576 	pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4577 	if (pci_cmd & PCI_COMMAND_SERR)
4578 		pci_write_config_word(adapter->pdev, PCI_COMMAND,
4579 				      pci_cmd & ~PCI_COMMAND_SERR);
4580 
4581 	err = e1000_test_msi_interrupt(adapter);
4582 
4583 	/* re-enable SERR */
4584 	if (pci_cmd & PCI_COMMAND_SERR) {
4585 		pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4586 		pci_cmd |= PCI_COMMAND_SERR;
4587 		pci_write_config_word(adapter->pdev, PCI_COMMAND, pci_cmd);
4588 	}
4589 
4590 	return err;
4591 }
4592 
4593 /**
4594  * e1000e_open - Called when a network interface is made active
4595  * @netdev: network interface device structure
4596  *
4597  * Returns 0 on success, negative value on failure
4598  *
4599  * The open entry point is called when a network interface is made
4600  * active by the system (IFF_UP).  At this point all resources needed
4601  * for transmit and receive operations are allocated, the interrupt
4602  * handler is registered with the OS, the watchdog timer is started,
4603  * and the stack is notified that the interface is ready.
4604  **/
4605 int e1000e_open(struct net_device *netdev)
4606 {
4607 	struct e1000_adapter *adapter = netdev_priv(netdev);
4608 	struct e1000_hw *hw = &adapter->hw;
4609 	struct pci_dev *pdev = adapter->pdev;
4610 	int err;
4611 
4612 	/* disallow open during test */
4613 	if (test_bit(__E1000_TESTING, &adapter->state))
4614 		return -EBUSY;
4615 
4616 	pm_runtime_get_sync(&pdev->dev);
4617 
4618 	netif_carrier_off(netdev);
4619 
4620 	/* allocate transmit descriptors */
4621 	err = e1000e_setup_tx_resources(adapter->tx_ring);
4622 	if (err)
4623 		goto err_setup_tx;
4624 
4625 	/* allocate receive descriptors */
4626 	err = e1000e_setup_rx_resources(adapter->rx_ring);
4627 	if (err)
4628 		goto err_setup_rx;
4629 
4630 	/* If AMT is enabled, let the firmware know that the network
4631 	 * interface is now open and reset the part to a known state.
4632 	 */
4633 	if (adapter->flags & FLAG_HAS_AMT) {
4634 		e1000e_get_hw_control(adapter);
4635 		e1000e_reset(adapter);
4636 	}
4637 
4638 	e1000e_power_up_phy(adapter);
4639 
4640 	adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
4641 	if ((adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
4642 		e1000_update_mng_vlan(adapter);
4643 
4644 	/* DMA latency requirement to workaround jumbo issue */
4645 	pm_qos_add_request(&adapter->pm_qos_req, PM_QOS_CPU_DMA_LATENCY,
4646 			   PM_QOS_DEFAULT_VALUE);
4647 
4648 	/* before we allocate an interrupt, we must be ready to handle it.
4649 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4650 	 * as soon as we call pci_request_irq, so we have to setup our
4651 	 * clean_rx handler before we do so.
4652 	 */
4653 	e1000_configure(adapter);
4654 
4655 	err = e1000_request_irq(adapter);
4656 	if (err)
4657 		goto err_req_irq;
4658 
4659 	/* Work around PCIe errata with MSI interrupts causing some chipsets to
4660 	 * ignore e1000e MSI messages, which means we need to test our MSI
4661 	 * interrupt now
4662 	 */
4663 	if (adapter->int_mode != E1000E_INT_MODE_LEGACY) {
4664 		err = e1000_test_msi(adapter);
4665 		if (err) {
4666 			e_err("Interrupt allocation failed\n");
4667 			goto err_req_irq;
4668 		}
4669 	}
4670 
4671 	/* From here on the code is the same as e1000e_up() */
4672 	clear_bit(__E1000_DOWN, &adapter->state);
4673 
4674 	napi_enable(&adapter->napi);
4675 
4676 	e1000_irq_enable(adapter);
4677 
4678 	adapter->tx_hang_recheck = false;
4679 	netif_start_queue(netdev);
4680 
4681 	hw->mac.get_link_status = true;
4682 	pm_runtime_put(&pdev->dev);
4683 
4684 	e1000e_trigger_lsc(adapter);
4685 
4686 	return 0;
4687 
4688 err_req_irq:
4689 	pm_qos_remove_request(&adapter->pm_qos_req);
4690 	e1000e_release_hw_control(adapter);
4691 	e1000_power_down_phy(adapter);
4692 	e1000e_free_rx_resources(adapter->rx_ring);
4693 err_setup_rx:
4694 	e1000e_free_tx_resources(adapter->tx_ring);
4695 err_setup_tx:
4696 	e1000e_reset(adapter);
4697 	pm_runtime_put_sync(&pdev->dev);
4698 
4699 	return err;
4700 }
4701 
4702 /**
4703  * e1000e_close - Disables a network interface
4704  * @netdev: network interface device structure
4705  *
4706  * Returns 0, this is not allowed to fail
4707  *
4708  * The close entry point is called when an interface is de-activated
4709  * by the OS.  The hardware is still under the drivers control, but
4710  * needs to be disabled.  A global MAC reset is issued to stop the
4711  * hardware, and all transmit and receive resources are freed.
4712  **/
4713 int e1000e_close(struct net_device *netdev)
4714 {
4715 	struct e1000_adapter *adapter = netdev_priv(netdev);
4716 	struct pci_dev *pdev = adapter->pdev;
4717 	int count = E1000_CHECK_RESET_COUNT;
4718 
4719 	while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
4720 		usleep_range(10000, 20000);
4721 
4722 	WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
4723 
4724 	pm_runtime_get_sync(&pdev->dev);
4725 
4726 	if (!test_bit(__E1000_DOWN, &adapter->state)) {
4727 		e1000e_down(adapter, true);
4728 		e1000_free_irq(adapter);
4729 
4730 		/* Link status message must follow this format */
4731 		pr_info("%s NIC Link is Down\n", adapter->netdev->name);
4732 	}
4733 
4734 	napi_disable(&adapter->napi);
4735 
4736 	e1000e_free_tx_resources(adapter->tx_ring);
4737 	e1000e_free_rx_resources(adapter->rx_ring);
4738 
4739 	/* kill manageability vlan ID if supported, but not if a vlan with
4740 	 * the same ID is registered on the host OS (let 8021q kill it)
4741 	 */
4742 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN)
4743 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
4744 				       adapter->mng_vlan_id);
4745 
4746 	/* If AMT is enabled, let the firmware know that the network
4747 	 * interface is now closed
4748 	 */
4749 	if ((adapter->flags & FLAG_HAS_AMT) &&
4750 	    !test_bit(__E1000_TESTING, &adapter->state))
4751 		e1000e_release_hw_control(adapter);
4752 
4753 	pm_qos_remove_request(&adapter->pm_qos_req);
4754 
4755 	pm_runtime_put_sync(&pdev->dev);
4756 
4757 	return 0;
4758 }
4759 
4760 /**
4761  * e1000_set_mac - Change the Ethernet Address of the NIC
4762  * @netdev: network interface device structure
4763  * @p: pointer to an address structure
4764  *
4765  * Returns 0 on success, negative on failure
4766  **/
4767 static int e1000_set_mac(struct net_device *netdev, void *p)
4768 {
4769 	struct e1000_adapter *adapter = netdev_priv(netdev);
4770 	struct e1000_hw *hw = &adapter->hw;
4771 	struct sockaddr *addr = p;
4772 
4773 	if (!is_valid_ether_addr(addr->sa_data))
4774 		return -EADDRNOTAVAIL;
4775 
4776 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
4777 	memcpy(adapter->hw.mac.addr, addr->sa_data, netdev->addr_len);
4778 
4779 	hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
4780 
4781 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA) {
4782 		/* activate the work around */
4783 		e1000e_set_laa_state_82571(&adapter->hw, 1);
4784 
4785 		/* Hold a copy of the LAA in RAR[14] This is done so that
4786 		 * between the time RAR[0] gets clobbered  and the time it
4787 		 * gets fixed (in e1000_watchdog), the actual LAA is in one
4788 		 * of the RARs and no incoming packets directed to this port
4789 		 * are dropped. Eventually the LAA will be in RAR[0] and
4790 		 * RAR[14]
4791 		 */
4792 		hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr,
4793 				    adapter->hw.mac.rar_entry_count - 1);
4794 	}
4795 
4796 	return 0;
4797 }
4798 
4799 /**
4800  * e1000e_update_phy_task - work thread to update phy
4801  * @work: pointer to our work struct
4802  *
4803  * this worker thread exists because we must acquire a
4804  * semaphore to read the phy, which we could msleep while
4805  * waiting for it, and we can't msleep in a timer.
4806  **/
4807 static void e1000e_update_phy_task(struct work_struct *work)
4808 {
4809 	struct e1000_adapter *adapter = container_of(work,
4810 						     struct e1000_adapter,
4811 						     update_phy_task);
4812 	struct e1000_hw *hw = &adapter->hw;
4813 
4814 	if (test_bit(__E1000_DOWN, &adapter->state))
4815 		return;
4816 
4817 	e1000_get_phy_info(hw);
4818 
4819 	/* Enable EEE on 82579 after link up */
4820 	if (hw->phy.type >= e1000_phy_82579)
4821 		e1000_set_eee_pchlan(hw);
4822 }
4823 
4824 /**
4825  * e1000_update_phy_info - timre call-back to update PHY info
4826  * @data: pointer to adapter cast into an unsigned long
4827  *
4828  * Need to wait a few seconds after link up to get diagnostic information from
4829  * the phy
4830  **/
4831 static void e1000_update_phy_info(struct timer_list *t)
4832 {
4833 	struct e1000_adapter *adapter = from_timer(adapter, t, phy_info_timer);
4834 
4835 	if (test_bit(__E1000_DOWN, &adapter->state))
4836 		return;
4837 
4838 	schedule_work(&adapter->update_phy_task);
4839 }
4840 
4841 /**
4842  * e1000e_update_phy_stats - Update the PHY statistics counters
4843  * @adapter: board private structure
4844  *
4845  * Read/clear the upper 16-bit PHY registers and read/accumulate lower
4846  **/
4847 static void e1000e_update_phy_stats(struct e1000_adapter *adapter)
4848 {
4849 	struct e1000_hw *hw = &adapter->hw;
4850 	s32 ret_val;
4851 	u16 phy_data;
4852 
4853 	ret_val = hw->phy.ops.acquire(hw);
4854 	if (ret_val)
4855 		return;
4856 
4857 	/* A page set is expensive so check if already on desired page.
4858 	 * If not, set to the page with the PHY status registers.
4859 	 */
4860 	hw->phy.addr = 1;
4861 	ret_val = e1000e_read_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4862 					   &phy_data);
4863 	if (ret_val)
4864 		goto release;
4865 	if (phy_data != (HV_STATS_PAGE << IGP_PAGE_SHIFT)) {
4866 		ret_val = hw->phy.ops.set_page(hw,
4867 					       HV_STATS_PAGE << IGP_PAGE_SHIFT);
4868 		if (ret_val)
4869 			goto release;
4870 	}
4871 
4872 	/* Single Collision Count */
4873 	hw->phy.ops.read_reg_page(hw, HV_SCC_UPPER, &phy_data);
4874 	ret_val = hw->phy.ops.read_reg_page(hw, HV_SCC_LOWER, &phy_data);
4875 	if (!ret_val)
4876 		adapter->stats.scc += phy_data;
4877 
4878 	/* Excessive Collision Count */
4879 	hw->phy.ops.read_reg_page(hw, HV_ECOL_UPPER, &phy_data);
4880 	ret_val = hw->phy.ops.read_reg_page(hw, HV_ECOL_LOWER, &phy_data);
4881 	if (!ret_val)
4882 		adapter->stats.ecol += phy_data;
4883 
4884 	/* Multiple Collision Count */
4885 	hw->phy.ops.read_reg_page(hw, HV_MCC_UPPER, &phy_data);
4886 	ret_val = hw->phy.ops.read_reg_page(hw, HV_MCC_LOWER, &phy_data);
4887 	if (!ret_val)
4888 		adapter->stats.mcc += phy_data;
4889 
4890 	/* Late Collision Count */
4891 	hw->phy.ops.read_reg_page(hw, HV_LATECOL_UPPER, &phy_data);
4892 	ret_val = hw->phy.ops.read_reg_page(hw, HV_LATECOL_LOWER, &phy_data);
4893 	if (!ret_val)
4894 		adapter->stats.latecol += phy_data;
4895 
4896 	/* Collision Count - also used for adaptive IFS */
4897 	hw->phy.ops.read_reg_page(hw, HV_COLC_UPPER, &phy_data);
4898 	ret_val = hw->phy.ops.read_reg_page(hw, HV_COLC_LOWER, &phy_data);
4899 	if (!ret_val)
4900 		hw->mac.collision_delta = phy_data;
4901 
4902 	/* Defer Count */
4903 	hw->phy.ops.read_reg_page(hw, HV_DC_UPPER, &phy_data);
4904 	ret_val = hw->phy.ops.read_reg_page(hw, HV_DC_LOWER, &phy_data);
4905 	if (!ret_val)
4906 		adapter->stats.dc += phy_data;
4907 
4908 	/* Transmit with no CRS */
4909 	hw->phy.ops.read_reg_page(hw, HV_TNCRS_UPPER, &phy_data);
4910 	ret_val = hw->phy.ops.read_reg_page(hw, HV_TNCRS_LOWER, &phy_data);
4911 	if (!ret_val)
4912 		adapter->stats.tncrs += phy_data;
4913 
4914 release:
4915 	hw->phy.ops.release(hw);
4916 }
4917 
4918 /**
4919  * e1000e_update_stats - Update the board statistics counters
4920  * @adapter: board private structure
4921  **/
4922 static void e1000e_update_stats(struct e1000_adapter *adapter)
4923 {
4924 	struct net_device *netdev = adapter->netdev;
4925 	struct e1000_hw *hw = &adapter->hw;
4926 	struct pci_dev *pdev = adapter->pdev;
4927 
4928 	/* Prevent stats update while adapter is being reset, or if the pci
4929 	 * connection is down.
4930 	 */
4931 	if (adapter->link_speed == 0)
4932 		return;
4933 	if (pci_channel_offline(pdev))
4934 		return;
4935 
4936 	adapter->stats.crcerrs += er32(CRCERRS);
4937 	adapter->stats.gprc += er32(GPRC);
4938 	adapter->stats.gorc += er32(GORCL);
4939 	er32(GORCH);		/* Clear gorc */
4940 	adapter->stats.bprc += er32(BPRC);
4941 	adapter->stats.mprc += er32(MPRC);
4942 	adapter->stats.roc += er32(ROC);
4943 
4944 	adapter->stats.mpc += er32(MPC);
4945 
4946 	/* Half-duplex statistics */
4947 	if (adapter->link_duplex == HALF_DUPLEX) {
4948 		if (adapter->flags2 & FLAG2_HAS_PHY_STATS) {
4949 			e1000e_update_phy_stats(adapter);
4950 		} else {
4951 			adapter->stats.scc += er32(SCC);
4952 			adapter->stats.ecol += er32(ECOL);
4953 			adapter->stats.mcc += er32(MCC);
4954 			adapter->stats.latecol += er32(LATECOL);
4955 			adapter->stats.dc += er32(DC);
4956 
4957 			hw->mac.collision_delta = er32(COLC);
4958 
4959 			if ((hw->mac.type != e1000_82574) &&
4960 			    (hw->mac.type != e1000_82583))
4961 				adapter->stats.tncrs += er32(TNCRS);
4962 		}
4963 		adapter->stats.colc += hw->mac.collision_delta;
4964 	}
4965 
4966 	adapter->stats.xonrxc += er32(XONRXC);
4967 	adapter->stats.xontxc += er32(XONTXC);
4968 	adapter->stats.xoffrxc += er32(XOFFRXC);
4969 	adapter->stats.xofftxc += er32(XOFFTXC);
4970 	adapter->stats.gptc += er32(GPTC);
4971 	adapter->stats.gotc += er32(GOTCL);
4972 	er32(GOTCH);		/* Clear gotc */
4973 	adapter->stats.rnbc += er32(RNBC);
4974 	adapter->stats.ruc += er32(RUC);
4975 
4976 	adapter->stats.mptc += er32(MPTC);
4977 	adapter->stats.bptc += er32(BPTC);
4978 
4979 	/* used for adaptive IFS */
4980 
4981 	hw->mac.tx_packet_delta = er32(TPT);
4982 	adapter->stats.tpt += hw->mac.tx_packet_delta;
4983 
4984 	adapter->stats.algnerrc += er32(ALGNERRC);
4985 	adapter->stats.rxerrc += er32(RXERRC);
4986 	adapter->stats.cexterr += er32(CEXTERR);
4987 	adapter->stats.tsctc += er32(TSCTC);
4988 	adapter->stats.tsctfc += er32(TSCTFC);
4989 
4990 	/* Fill out the OS statistics structure */
4991 	netdev->stats.multicast = adapter->stats.mprc;
4992 	netdev->stats.collisions = adapter->stats.colc;
4993 
4994 	/* Rx Errors */
4995 
4996 	/* RLEC on some newer hardware can be incorrect so build
4997 	 * our own version based on RUC and ROC
4998 	 */
4999 	netdev->stats.rx_errors = adapter->stats.rxerrc +
5000 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
5001 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5002 	netdev->stats.rx_length_errors = adapter->stats.ruc +
5003 	    adapter->stats.roc;
5004 	netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
5005 	netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
5006 	netdev->stats.rx_missed_errors = adapter->stats.mpc;
5007 
5008 	/* Tx Errors */
5009 	netdev->stats.tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5010 	netdev->stats.tx_aborted_errors = adapter->stats.ecol;
5011 	netdev->stats.tx_window_errors = adapter->stats.latecol;
5012 	netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
5013 
5014 	/* Tx Dropped needs to be maintained elsewhere */
5015 
5016 	/* Management Stats */
5017 	adapter->stats.mgptc += er32(MGTPTC);
5018 	adapter->stats.mgprc += er32(MGTPRC);
5019 	adapter->stats.mgpdc += er32(MGTPDC);
5020 
5021 	/* Correctable ECC Errors */
5022 	if (hw->mac.type >= e1000_pch_lpt) {
5023 		u32 pbeccsts = er32(PBECCSTS);
5024 
5025 		adapter->corr_errors +=
5026 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
5027 		adapter->uncorr_errors +=
5028 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
5029 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
5030 	}
5031 }
5032 
5033 /**
5034  * e1000_phy_read_status - Update the PHY register status snapshot
5035  * @adapter: board private structure
5036  **/
5037 static void e1000_phy_read_status(struct e1000_adapter *adapter)
5038 {
5039 	struct e1000_hw *hw = &adapter->hw;
5040 	struct e1000_phy_regs *phy = &adapter->phy_regs;
5041 
5042 	if (!pm_runtime_suspended((&adapter->pdev->dev)->parent) &&
5043 	    (er32(STATUS) & E1000_STATUS_LU) &&
5044 	    (adapter->hw.phy.media_type == e1000_media_type_copper)) {
5045 		int ret_val;
5046 
5047 		ret_val = e1e_rphy(hw, MII_BMCR, &phy->bmcr);
5048 		ret_val |= e1e_rphy(hw, MII_BMSR, &phy->bmsr);
5049 		ret_val |= e1e_rphy(hw, MII_ADVERTISE, &phy->advertise);
5050 		ret_val |= e1e_rphy(hw, MII_LPA, &phy->lpa);
5051 		ret_val |= e1e_rphy(hw, MII_EXPANSION, &phy->expansion);
5052 		ret_val |= e1e_rphy(hw, MII_CTRL1000, &phy->ctrl1000);
5053 		ret_val |= e1e_rphy(hw, MII_STAT1000, &phy->stat1000);
5054 		ret_val |= e1e_rphy(hw, MII_ESTATUS, &phy->estatus);
5055 		if (ret_val)
5056 			e_warn("Error reading PHY register\n");
5057 	} else {
5058 		/* Do not read PHY registers if link is not up
5059 		 * Set values to typical power-on defaults
5060 		 */
5061 		phy->bmcr = (BMCR_SPEED1000 | BMCR_ANENABLE | BMCR_FULLDPLX);
5062 		phy->bmsr = (BMSR_100FULL | BMSR_100HALF | BMSR_10FULL |
5063 			     BMSR_10HALF | BMSR_ESTATEN | BMSR_ANEGCAPABLE |
5064 			     BMSR_ERCAP);
5065 		phy->advertise = (ADVERTISE_PAUSE_ASYM | ADVERTISE_PAUSE_CAP |
5066 				  ADVERTISE_ALL | ADVERTISE_CSMA);
5067 		phy->lpa = 0;
5068 		phy->expansion = EXPANSION_ENABLENPAGE;
5069 		phy->ctrl1000 = ADVERTISE_1000FULL;
5070 		phy->stat1000 = 0;
5071 		phy->estatus = (ESTATUS_1000_TFULL | ESTATUS_1000_THALF);
5072 	}
5073 }
5074 
5075 static void e1000_print_link_info(struct e1000_adapter *adapter)
5076 {
5077 	struct e1000_hw *hw = &adapter->hw;
5078 	u32 ctrl = er32(CTRL);
5079 
5080 	/* Link status message must follow this format for user tools */
5081 	pr_info("%s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5082 		adapter->netdev->name, adapter->link_speed,
5083 		adapter->link_duplex == FULL_DUPLEX ? "Full" : "Half",
5084 		(ctrl & E1000_CTRL_TFCE) && (ctrl & E1000_CTRL_RFCE) ? "Rx/Tx" :
5085 		(ctrl & E1000_CTRL_RFCE) ? "Rx" :
5086 		(ctrl & E1000_CTRL_TFCE) ? "Tx" : "None");
5087 }
5088 
5089 static bool e1000e_has_link(struct e1000_adapter *adapter)
5090 {
5091 	struct e1000_hw *hw = &adapter->hw;
5092 	bool link_active = false;
5093 	s32 ret_val = 0;
5094 
5095 	/* get_link_status is set on LSC (link status) interrupt or
5096 	 * Rx sequence error interrupt.  get_link_status will stay
5097 	 * true until the check_for_link establishes link
5098 	 * for copper adapters ONLY
5099 	 */
5100 	switch (hw->phy.media_type) {
5101 	case e1000_media_type_copper:
5102 		if (hw->mac.get_link_status) {
5103 			ret_val = hw->mac.ops.check_for_link(hw);
5104 			link_active = ret_val > 0;
5105 		} else {
5106 			link_active = true;
5107 		}
5108 		break;
5109 	case e1000_media_type_fiber:
5110 		ret_val = hw->mac.ops.check_for_link(hw);
5111 		link_active = !!(er32(STATUS) & E1000_STATUS_LU);
5112 		break;
5113 	case e1000_media_type_internal_serdes:
5114 		ret_val = hw->mac.ops.check_for_link(hw);
5115 		link_active = hw->mac.serdes_has_link;
5116 		break;
5117 	default:
5118 	case e1000_media_type_unknown:
5119 		break;
5120 	}
5121 
5122 	if ((ret_val == -E1000_ERR_PHY) && (hw->phy.type == e1000_phy_igp_3) &&
5123 	    (er32(CTRL) & E1000_PHY_CTRL_GBE_DISABLE)) {
5124 		/* See e1000_kmrn_lock_loss_workaround_ich8lan() */
5125 		e_info("Gigabit has been disabled, downgrading speed\n");
5126 	}
5127 
5128 	return link_active;
5129 }
5130 
5131 static void e1000e_enable_receives(struct e1000_adapter *adapter)
5132 {
5133 	/* make sure the receive unit is started */
5134 	if ((adapter->flags & FLAG_RX_NEEDS_RESTART) &&
5135 	    (adapter->flags & FLAG_RESTART_NOW)) {
5136 		struct e1000_hw *hw = &adapter->hw;
5137 		u32 rctl = er32(RCTL);
5138 
5139 		ew32(RCTL, rctl | E1000_RCTL_EN);
5140 		adapter->flags &= ~FLAG_RESTART_NOW;
5141 	}
5142 }
5143 
5144 static void e1000e_check_82574_phy_workaround(struct e1000_adapter *adapter)
5145 {
5146 	struct e1000_hw *hw = &adapter->hw;
5147 
5148 	/* With 82574 controllers, PHY needs to be checked periodically
5149 	 * for hung state and reset, if two calls return true
5150 	 */
5151 	if (e1000_check_phy_82574(hw))
5152 		adapter->phy_hang_count++;
5153 	else
5154 		adapter->phy_hang_count = 0;
5155 
5156 	if (adapter->phy_hang_count > 1) {
5157 		adapter->phy_hang_count = 0;
5158 		e_dbg("PHY appears hung - resetting\n");
5159 		schedule_work(&adapter->reset_task);
5160 	}
5161 }
5162 
5163 /**
5164  * e1000_watchdog - Timer Call-back
5165  * @data: pointer to adapter cast into an unsigned long
5166  **/
5167 static void e1000_watchdog(struct timer_list *t)
5168 {
5169 	struct e1000_adapter *adapter = from_timer(adapter, t, watchdog_timer);
5170 
5171 	/* Do the rest outside of interrupt context */
5172 	schedule_work(&adapter->watchdog_task);
5173 
5174 	/* TODO: make this use queue_delayed_work() */
5175 }
5176 
5177 static void e1000_watchdog_task(struct work_struct *work)
5178 {
5179 	struct e1000_adapter *adapter = container_of(work,
5180 						     struct e1000_adapter,
5181 						     watchdog_task);
5182 	struct net_device *netdev = adapter->netdev;
5183 	struct e1000_mac_info *mac = &adapter->hw.mac;
5184 	struct e1000_phy_info *phy = &adapter->hw.phy;
5185 	struct e1000_ring *tx_ring = adapter->tx_ring;
5186 	struct e1000_hw *hw = &adapter->hw;
5187 	u32 link, tctl;
5188 
5189 	if (test_bit(__E1000_DOWN, &adapter->state))
5190 		return;
5191 
5192 	link = e1000e_has_link(adapter);
5193 	if ((netif_carrier_ok(netdev)) && link) {
5194 		/* Cancel scheduled suspend requests. */
5195 		pm_runtime_resume(netdev->dev.parent);
5196 
5197 		e1000e_enable_receives(adapter);
5198 		goto link_up;
5199 	}
5200 
5201 	if ((e1000e_enable_tx_pkt_filtering(hw)) &&
5202 	    (adapter->mng_vlan_id != adapter->hw.mng_cookie.vlan_id))
5203 		e1000_update_mng_vlan(adapter);
5204 
5205 	if (link) {
5206 		if (!netif_carrier_ok(netdev)) {
5207 			bool txb2b = true;
5208 
5209 			/* Cancel scheduled suspend requests. */
5210 			pm_runtime_resume(netdev->dev.parent);
5211 
5212 			/* update snapshot of PHY registers on LSC */
5213 			e1000_phy_read_status(adapter);
5214 			mac->ops.get_link_up_info(&adapter->hw,
5215 						  &adapter->link_speed,
5216 						  &adapter->link_duplex);
5217 			e1000_print_link_info(adapter);
5218 
5219 			/* check if SmartSpeed worked */
5220 			e1000e_check_downshift(hw);
5221 			if (phy->speed_downgraded)
5222 				netdev_warn(netdev,
5223 					    "Link Speed was downgraded by SmartSpeed\n");
5224 
5225 			/* On supported PHYs, check for duplex mismatch only
5226 			 * if link has autonegotiated at 10/100 half
5227 			 */
5228 			if ((hw->phy.type == e1000_phy_igp_3 ||
5229 			     hw->phy.type == e1000_phy_bm) &&
5230 			    hw->mac.autoneg &&
5231 			    (adapter->link_speed == SPEED_10 ||
5232 			     adapter->link_speed == SPEED_100) &&
5233 			    (adapter->link_duplex == HALF_DUPLEX)) {
5234 				u16 autoneg_exp;
5235 
5236 				e1e_rphy(hw, MII_EXPANSION, &autoneg_exp);
5237 
5238 				if (!(autoneg_exp & EXPANSION_NWAY))
5239 					e_info("Autonegotiated half duplex but link partner cannot autoneg.  Try forcing full duplex if link gets many collisions.\n");
5240 			}
5241 
5242 			/* adjust timeout factor according to speed/duplex */
5243 			adapter->tx_timeout_factor = 1;
5244 			switch (adapter->link_speed) {
5245 			case SPEED_10:
5246 				txb2b = false;
5247 				adapter->tx_timeout_factor = 16;
5248 				break;
5249 			case SPEED_100:
5250 				txb2b = false;
5251 				adapter->tx_timeout_factor = 10;
5252 				break;
5253 			}
5254 
5255 			/* workaround: re-program speed mode bit after
5256 			 * link-up event
5257 			 */
5258 			if ((adapter->flags & FLAG_TARC_SPEED_MODE_BIT) &&
5259 			    !txb2b) {
5260 				u32 tarc0;
5261 
5262 				tarc0 = er32(TARC(0));
5263 				tarc0 &= ~SPEED_MODE_BIT;
5264 				ew32(TARC(0), tarc0);
5265 			}
5266 
5267 			/* disable TSO for pcie and 10/100 speeds, to avoid
5268 			 * some hardware issues
5269 			 */
5270 			if (!(adapter->flags & FLAG_TSO_FORCE)) {
5271 				switch (adapter->link_speed) {
5272 				case SPEED_10:
5273 				case SPEED_100:
5274 					e_info("10/100 speed: disabling TSO\n");
5275 					netdev->features &= ~NETIF_F_TSO;
5276 					netdev->features &= ~NETIF_F_TSO6;
5277 					break;
5278 				case SPEED_1000:
5279 					netdev->features |= NETIF_F_TSO;
5280 					netdev->features |= NETIF_F_TSO6;
5281 					break;
5282 				default:
5283 					/* oops */
5284 					break;
5285 				}
5286 			}
5287 
5288 			/* enable transmits in the hardware, need to do this
5289 			 * after setting TARC(0)
5290 			 */
5291 			tctl = er32(TCTL);
5292 			tctl |= E1000_TCTL_EN;
5293 			ew32(TCTL, tctl);
5294 
5295 			/* Perform any post-link-up configuration before
5296 			 * reporting link up.
5297 			 */
5298 			if (phy->ops.cfg_on_link_up)
5299 				phy->ops.cfg_on_link_up(hw);
5300 
5301 			netif_carrier_on(netdev);
5302 
5303 			if (!test_bit(__E1000_DOWN, &adapter->state))
5304 				mod_timer(&adapter->phy_info_timer,
5305 					  round_jiffies(jiffies + 2 * HZ));
5306 		}
5307 	} else {
5308 		if (netif_carrier_ok(netdev)) {
5309 			adapter->link_speed = 0;
5310 			adapter->link_duplex = 0;
5311 			/* Link status message must follow this format */
5312 			pr_info("%s NIC Link is Down\n", adapter->netdev->name);
5313 			netif_carrier_off(netdev);
5314 			if (!test_bit(__E1000_DOWN, &adapter->state))
5315 				mod_timer(&adapter->phy_info_timer,
5316 					  round_jiffies(jiffies + 2 * HZ));
5317 
5318 			/* 8000ES2LAN requires a Rx packet buffer work-around
5319 			 * on link down event; reset the controller to flush
5320 			 * the Rx packet buffer.
5321 			 */
5322 			if (adapter->flags & FLAG_RX_NEEDS_RESTART)
5323 				adapter->flags |= FLAG_RESTART_NOW;
5324 			else
5325 				pm_schedule_suspend(netdev->dev.parent,
5326 						    LINK_TIMEOUT);
5327 		}
5328 	}
5329 
5330 link_up:
5331 	spin_lock(&adapter->stats64_lock);
5332 	e1000e_update_stats(adapter);
5333 
5334 	mac->tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
5335 	adapter->tpt_old = adapter->stats.tpt;
5336 	mac->collision_delta = adapter->stats.colc - adapter->colc_old;
5337 	adapter->colc_old = adapter->stats.colc;
5338 
5339 	adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
5340 	adapter->gorc_old = adapter->stats.gorc;
5341 	adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
5342 	adapter->gotc_old = adapter->stats.gotc;
5343 	spin_unlock(&adapter->stats64_lock);
5344 
5345 	/* If the link is lost the controller stops DMA, but
5346 	 * if there is queued Tx work it cannot be done.  So
5347 	 * reset the controller to flush the Tx packet buffers.
5348 	 */
5349 	if (!netif_carrier_ok(netdev) &&
5350 	    (e1000_desc_unused(tx_ring) + 1 < tx_ring->count))
5351 		adapter->flags |= FLAG_RESTART_NOW;
5352 
5353 	/* If reset is necessary, do it outside of interrupt context. */
5354 	if (adapter->flags & FLAG_RESTART_NOW) {
5355 		schedule_work(&adapter->reset_task);
5356 		/* return immediately since reset is imminent */
5357 		return;
5358 	}
5359 
5360 	e1000e_update_adaptive(&adapter->hw);
5361 
5362 	/* Simple mode for Interrupt Throttle Rate (ITR) */
5363 	if (adapter->itr_setting == 4) {
5364 		/* Symmetric Tx/Rx gets a reduced ITR=2000;
5365 		 * Total asymmetrical Tx or Rx gets ITR=8000;
5366 		 * everyone else is between 2000-8000.
5367 		 */
5368 		u32 goc = (adapter->gotc + adapter->gorc) / 10000;
5369 		u32 dif = (adapter->gotc > adapter->gorc ?
5370 			   adapter->gotc - adapter->gorc :
5371 			   adapter->gorc - adapter->gotc) / 10000;
5372 		u32 itr = goc > 0 ? (dif * 6000 / goc + 2000) : 8000;
5373 
5374 		e1000e_write_itr(adapter, itr);
5375 	}
5376 
5377 	/* Cause software interrupt to ensure Rx ring is cleaned */
5378 	if (adapter->msix_entries)
5379 		ew32(ICS, adapter->rx_ring->ims_val);
5380 	else
5381 		ew32(ICS, E1000_ICS_RXDMT0);
5382 
5383 	/* flush pending descriptors to memory before detecting Tx hang */
5384 	e1000e_flush_descriptors(adapter);
5385 
5386 	/* Force detection of hung controller every watchdog period */
5387 	adapter->detect_tx_hung = true;
5388 
5389 	/* With 82571 controllers, LAA may be overwritten due to controller
5390 	 * reset from the other port. Set the appropriate LAA in RAR[0]
5391 	 */
5392 	if (e1000e_get_laa_state_82571(hw))
5393 		hw->mac.ops.rar_set(hw, adapter->hw.mac.addr, 0);
5394 
5395 	if (adapter->flags2 & FLAG2_CHECK_PHY_HANG)
5396 		e1000e_check_82574_phy_workaround(adapter);
5397 
5398 	/* Clear valid timestamp stuck in RXSTMPL/H due to a Rx error */
5399 	if (adapter->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE) {
5400 		if ((adapter->flags2 & FLAG2_CHECK_RX_HWTSTAMP) &&
5401 		    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID)) {
5402 			er32(RXSTMPH);
5403 			adapter->rx_hwtstamp_cleared++;
5404 		} else {
5405 			adapter->flags2 |= FLAG2_CHECK_RX_HWTSTAMP;
5406 		}
5407 	}
5408 
5409 	/* Reset the timer */
5410 	if (!test_bit(__E1000_DOWN, &adapter->state))
5411 		mod_timer(&adapter->watchdog_timer,
5412 			  round_jiffies(jiffies + 2 * HZ));
5413 }
5414 
5415 #define E1000_TX_FLAGS_CSUM		0x00000001
5416 #define E1000_TX_FLAGS_VLAN		0x00000002
5417 #define E1000_TX_FLAGS_TSO		0x00000004
5418 #define E1000_TX_FLAGS_IPV4		0x00000008
5419 #define E1000_TX_FLAGS_NO_FCS		0x00000010
5420 #define E1000_TX_FLAGS_HWTSTAMP		0x00000020
5421 #define E1000_TX_FLAGS_VLAN_MASK	0xffff0000
5422 #define E1000_TX_FLAGS_VLAN_SHIFT	16
5423 
5424 static int e1000_tso(struct e1000_ring *tx_ring, struct sk_buff *skb,
5425 		     __be16 protocol)
5426 {
5427 	struct e1000_context_desc *context_desc;
5428 	struct e1000_buffer *buffer_info;
5429 	unsigned int i;
5430 	u32 cmd_length = 0;
5431 	u16 ipcse = 0, mss;
5432 	u8 ipcss, ipcso, tucss, tucso, hdr_len;
5433 	int err;
5434 
5435 	if (!skb_is_gso(skb))
5436 		return 0;
5437 
5438 	err = skb_cow_head(skb, 0);
5439 	if (err < 0)
5440 		return err;
5441 
5442 	hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5443 	mss = skb_shinfo(skb)->gso_size;
5444 	if (protocol == htons(ETH_P_IP)) {
5445 		struct iphdr *iph = ip_hdr(skb);
5446 		iph->tot_len = 0;
5447 		iph->check = 0;
5448 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
5449 							 0, IPPROTO_TCP, 0);
5450 		cmd_length = E1000_TXD_CMD_IP;
5451 		ipcse = skb_transport_offset(skb) - 1;
5452 	} else if (skb_is_gso_v6(skb)) {
5453 		ipv6_hdr(skb)->payload_len = 0;
5454 		tcp_hdr(skb)->check = ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
5455 						       &ipv6_hdr(skb)->daddr,
5456 						       0, IPPROTO_TCP, 0);
5457 		ipcse = 0;
5458 	}
5459 	ipcss = skb_network_offset(skb);
5460 	ipcso = (void *)&(ip_hdr(skb)->check) - (void *)skb->data;
5461 	tucss = skb_transport_offset(skb);
5462 	tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
5463 
5464 	cmd_length |= (E1000_TXD_CMD_DEXT | E1000_TXD_CMD_TSE |
5465 		       E1000_TXD_CMD_TCP | (skb->len - (hdr_len)));
5466 
5467 	i = tx_ring->next_to_use;
5468 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5469 	buffer_info = &tx_ring->buffer_info[i];
5470 
5471 	context_desc->lower_setup.ip_fields.ipcss = ipcss;
5472 	context_desc->lower_setup.ip_fields.ipcso = ipcso;
5473 	context_desc->lower_setup.ip_fields.ipcse = cpu_to_le16(ipcse);
5474 	context_desc->upper_setup.tcp_fields.tucss = tucss;
5475 	context_desc->upper_setup.tcp_fields.tucso = tucso;
5476 	context_desc->upper_setup.tcp_fields.tucse = 0;
5477 	context_desc->tcp_seg_setup.fields.mss = cpu_to_le16(mss);
5478 	context_desc->tcp_seg_setup.fields.hdr_len = hdr_len;
5479 	context_desc->cmd_and_length = cpu_to_le32(cmd_length);
5480 
5481 	buffer_info->time_stamp = jiffies;
5482 	buffer_info->next_to_watch = i;
5483 
5484 	i++;
5485 	if (i == tx_ring->count)
5486 		i = 0;
5487 	tx_ring->next_to_use = i;
5488 
5489 	return 1;
5490 }
5491 
5492 static bool e1000_tx_csum(struct e1000_ring *tx_ring, struct sk_buff *skb,
5493 			  __be16 protocol)
5494 {
5495 	struct e1000_adapter *adapter = tx_ring->adapter;
5496 	struct e1000_context_desc *context_desc;
5497 	struct e1000_buffer *buffer_info;
5498 	unsigned int i;
5499 	u8 css;
5500 	u32 cmd_len = E1000_TXD_CMD_DEXT;
5501 
5502 	if (skb->ip_summed != CHECKSUM_PARTIAL)
5503 		return false;
5504 
5505 	switch (protocol) {
5506 	case cpu_to_be16(ETH_P_IP):
5507 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
5508 			cmd_len |= E1000_TXD_CMD_TCP;
5509 		break;
5510 	case cpu_to_be16(ETH_P_IPV6):
5511 		/* XXX not handling all IPV6 headers */
5512 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
5513 			cmd_len |= E1000_TXD_CMD_TCP;
5514 		break;
5515 	default:
5516 		if (unlikely(net_ratelimit()))
5517 			e_warn("checksum_partial proto=%x!\n",
5518 			       be16_to_cpu(protocol));
5519 		break;
5520 	}
5521 
5522 	css = skb_checksum_start_offset(skb);
5523 
5524 	i = tx_ring->next_to_use;
5525 	buffer_info = &tx_ring->buffer_info[i];
5526 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5527 
5528 	context_desc->lower_setup.ip_config = 0;
5529 	context_desc->upper_setup.tcp_fields.tucss = css;
5530 	context_desc->upper_setup.tcp_fields.tucso = css + skb->csum_offset;
5531 	context_desc->upper_setup.tcp_fields.tucse = 0;
5532 	context_desc->tcp_seg_setup.data = 0;
5533 	context_desc->cmd_and_length = cpu_to_le32(cmd_len);
5534 
5535 	buffer_info->time_stamp = jiffies;
5536 	buffer_info->next_to_watch = i;
5537 
5538 	i++;
5539 	if (i == tx_ring->count)
5540 		i = 0;
5541 	tx_ring->next_to_use = i;
5542 
5543 	return true;
5544 }
5545 
5546 static int e1000_tx_map(struct e1000_ring *tx_ring, struct sk_buff *skb,
5547 			unsigned int first, unsigned int max_per_txd,
5548 			unsigned int nr_frags)
5549 {
5550 	struct e1000_adapter *adapter = tx_ring->adapter;
5551 	struct pci_dev *pdev = adapter->pdev;
5552 	struct e1000_buffer *buffer_info;
5553 	unsigned int len = skb_headlen(skb);
5554 	unsigned int offset = 0, size, count = 0, i;
5555 	unsigned int f, bytecount, segs;
5556 
5557 	i = tx_ring->next_to_use;
5558 
5559 	while (len) {
5560 		buffer_info = &tx_ring->buffer_info[i];
5561 		size = min(len, max_per_txd);
5562 
5563 		buffer_info->length = size;
5564 		buffer_info->time_stamp = jiffies;
5565 		buffer_info->next_to_watch = i;
5566 		buffer_info->dma = dma_map_single(&pdev->dev,
5567 						  skb->data + offset,
5568 						  size, DMA_TO_DEVICE);
5569 		buffer_info->mapped_as_page = false;
5570 		if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5571 			goto dma_error;
5572 
5573 		len -= size;
5574 		offset += size;
5575 		count++;
5576 
5577 		if (len) {
5578 			i++;
5579 			if (i == tx_ring->count)
5580 				i = 0;
5581 		}
5582 	}
5583 
5584 	for (f = 0; f < nr_frags; f++) {
5585 		const struct skb_frag_struct *frag;
5586 
5587 		frag = &skb_shinfo(skb)->frags[f];
5588 		len = skb_frag_size(frag);
5589 		offset = 0;
5590 
5591 		while (len) {
5592 			i++;
5593 			if (i == tx_ring->count)
5594 				i = 0;
5595 
5596 			buffer_info = &tx_ring->buffer_info[i];
5597 			size = min(len, max_per_txd);
5598 
5599 			buffer_info->length = size;
5600 			buffer_info->time_stamp = jiffies;
5601 			buffer_info->next_to_watch = i;
5602 			buffer_info->dma = skb_frag_dma_map(&pdev->dev, frag,
5603 							    offset, size,
5604 							    DMA_TO_DEVICE);
5605 			buffer_info->mapped_as_page = true;
5606 			if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5607 				goto dma_error;
5608 
5609 			len -= size;
5610 			offset += size;
5611 			count++;
5612 		}
5613 	}
5614 
5615 	segs = skb_shinfo(skb)->gso_segs ? : 1;
5616 	/* multiply data chunks by size of headers */
5617 	bytecount = ((segs - 1) * skb_headlen(skb)) + skb->len;
5618 
5619 	tx_ring->buffer_info[i].skb = skb;
5620 	tx_ring->buffer_info[i].segs = segs;
5621 	tx_ring->buffer_info[i].bytecount = bytecount;
5622 	tx_ring->buffer_info[first].next_to_watch = i;
5623 
5624 	return count;
5625 
5626 dma_error:
5627 	dev_err(&pdev->dev, "Tx DMA map failed\n");
5628 	buffer_info->dma = 0;
5629 	if (count)
5630 		count--;
5631 
5632 	while (count--) {
5633 		if (i == 0)
5634 			i += tx_ring->count;
5635 		i--;
5636 		buffer_info = &tx_ring->buffer_info[i];
5637 		e1000_put_txbuf(tx_ring, buffer_info, true);
5638 	}
5639 
5640 	return 0;
5641 }
5642 
5643 static void e1000_tx_queue(struct e1000_ring *tx_ring, int tx_flags, int count)
5644 {
5645 	struct e1000_adapter *adapter = tx_ring->adapter;
5646 	struct e1000_tx_desc *tx_desc = NULL;
5647 	struct e1000_buffer *buffer_info;
5648 	u32 txd_upper = 0, txd_lower = E1000_TXD_CMD_IFCS;
5649 	unsigned int i;
5650 
5651 	if (tx_flags & E1000_TX_FLAGS_TSO) {
5652 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D |
5653 		    E1000_TXD_CMD_TSE;
5654 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5655 
5656 		if (tx_flags & E1000_TX_FLAGS_IPV4)
5657 			txd_upper |= E1000_TXD_POPTS_IXSM << 8;
5658 	}
5659 
5660 	if (tx_flags & E1000_TX_FLAGS_CSUM) {
5661 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5662 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5663 	}
5664 
5665 	if (tx_flags & E1000_TX_FLAGS_VLAN) {
5666 		txd_lower |= E1000_TXD_CMD_VLE;
5667 		txd_upper |= (tx_flags & E1000_TX_FLAGS_VLAN_MASK);
5668 	}
5669 
5670 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5671 		txd_lower &= ~(E1000_TXD_CMD_IFCS);
5672 
5673 	if (unlikely(tx_flags & E1000_TX_FLAGS_HWTSTAMP)) {
5674 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5675 		txd_upper |= E1000_TXD_EXTCMD_TSTAMP;
5676 	}
5677 
5678 	i = tx_ring->next_to_use;
5679 
5680 	do {
5681 		buffer_info = &tx_ring->buffer_info[i];
5682 		tx_desc = E1000_TX_DESC(*tx_ring, i);
5683 		tx_desc->buffer_addr = cpu_to_le64(buffer_info->dma);
5684 		tx_desc->lower.data = cpu_to_le32(txd_lower |
5685 						  buffer_info->length);
5686 		tx_desc->upper.data = cpu_to_le32(txd_upper);
5687 
5688 		i++;
5689 		if (i == tx_ring->count)
5690 			i = 0;
5691 	} while (--count > 0);
5692 
5693 	tx_desc->lower.data |= cpu_to_le32(adapter->txd_cmd);
5694 
5695 	/* txd_cmd re-enables FCS, so we'll re-disable it here as desired. */
5696 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5697 		tx_desc->lower.data &= ~(cpu_to_le32(E1000_TXD_CMD_IFCS));
5698 
5699 	/* Force memory writes to complete before letting h/w
5700 	 * know there are new descriptors to fetch.  (Only
5701 	 * applicable for weak-ordered memory model archs,
5702 	 * such as IA-64).
5703 	 */
5704 	wmb();
5705 
5706 	tx_ring->next_to_use = i;
5707 }
5708 
5709 #define MINIMUM_DHCP_PACKET_SIZE 282
5710 static int e1000_transfer_dhcp_info(struct e1000_adapter *adapter,
5711 				    struct sk_buff *skb)
5712 {
5713 	struct e1000_hw *hw = &adapter->hw;
5714 	u16 length, offset;
5715 
5716 	if (skb_vlan_tag_present(skb) &&
5717 	    !((skb_vlan_tag_get(skb) == adapter->hw.mng_cookie.vlan_id) &&
5718 	      (adapter->hw.mng_cookie.status &
5719 	       E1000_MNG_DHCP_COOKIE_STATUS_VLAN)))
5720 		return 0;
5721 
5722 	if (skb->len <= MINIMUM_DHCP_PACKET_SIZE)
5723 		return 0;
5724 
5725 	if (((struct ethhdr *)skb->data)->h_proto != htons(ETH_P_IP))
5726 		return 0;
5727 
5728 	{
5729 		const struct iphdr *ip = (struct iphdr *)((u8 *)skb->data + 14);
5730 		struct udphdr *udp;
5731 
5732 		if (ip->protocol != IPPROTO_UDP)
5733 			return 0;
5734 
5735 		udp = (struct udphdr *)((u8 *)ip + (ip->ihl << 2));
5736 		if (ntohs(udp->dest) != 67)
5737 			return 0;
5738 
5739 		offset = (u8 *)udp + 8 - skb->data;
5740 		length = skb->len - offset;
5741 		return e1000e_mng_write_dhcp_info(hw, (u8 *)udp + 8, length);
5742 	}
5743 
5744 	return 0;
5745 }
5746 
5747 static int __e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5748 {
5749 	struct e1000_adapter *adapter = tx_ring->adapter;
5750 
5751 	netif_stop_queue(adapter->netdev);
5752 	/* Herbert's original patch had:
5753 	 *  smp_mb__after_netif_stop_queue();
5754 	 * but since that doesn't exist yet, just open code it.
5755 	 */
5756 	smp_mb();
5757 
5758 	/* We need to check again in a case another CPU has just
5759 	 * made room available.
5760 	 */
5761 	if (e1000_desc_unused(tx_ring) < size)
5762 		return -EBUSY;
5763 
5764 	/* A reprieve! */
5765 	netif_start_queue(adapter->netdev);
5766 	++adapter->restart_queue;
5767 	return 0;
5768 }
5769 
5770 static int e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5771 {
5772 	BUG_ON(size > tx_ring->count);
5773 
5774 	if (e1000_desc_unused(tx_ring) >= size)
5775 		return 0;
5776 	return __e1000_maybe_stop_tx(tx_ring, size);
5777 }
5778 
5779 static netdev_tx_t e1000_xmit_frame(struct sk_buff *skb,
5780 				    struct net_device *netdev)
5781 {
5782 	struct e1000_adapter *adapter = netdev_priv(netdev);
5783 	struct e1000_ring *tx_ring = adapter->tx_ring;
5784 	unsigned int first;
5785 	unsigned int tx_flags = 0;
5786 	unsigned int len = skb_headlen(skb);
5787 	unsigned int nr_frags;
5788 	unsigned int mss;
5789 	int count = 0;
5790 	int tso;
5791 	unsigned int f;
5792 	__be16 protocol = vlan_get_protocol(skb);
5793 
5794 	if (test_bit(__E1000_DOWN, &adapter->state)) {
5795 		dev_kfree_skb_any(skb);
5796 		return NETDEV_TX_OK;
5797 	}
5798 
5799 	if (skb->len <= 0) {
5800 		dev_kfree_skb_any(skb);
5801 		return NETDEV_TX_OK;
5802 	}
5803 
5804 	/* The minimum packet size with TCTL.PSP set is 17 bytes so
5805 	 * pad skb in order to meet this minimum size requirement
5806 	 */
5807 	if (skb_put_padto(skb, 17))
5808 		return NETDEV_TX_OK;
5809 
5810 	mss = skb_shinfo(skb)->gso_size;
5811 	if (mss) {
5812 		u8 hdr_len;
5813 
5814 		/* TSO Workaround for 82571/2/3 Controllers -- if skb->data
5815 		 * points to just header, pull a few bytes of payload from
5816 		 * frags into skb->data
5817 		 */
5818 		hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5819 		/* we do this workaround for ES2LAN, but it is un-necessary,
5820 		 * avoiding it could save a lot of cycles
5821 		 */
5822 		if (skb->data_len && (hdr_len == len)) {
5823 			unsigned int pull_size;
5824 
5825 			pull_size = min_t(unsigned int, 4, skb->data_len);
5826 			if (!__pskb_pull_tail(skb, pull_size)) {
5827 				e_err("__pskb_pull_tail failed.\n");
5828 				dev_kfree_skb_any(skb);
5829 				return NETDEV_TX_OK;
5830 			}
5831 			len = skb_headlen(skb);
5832 		}
5833 	}
5834 
5835 	/* reserve a descriptor for the offload context */
5836 	if ((mss) || (skb->ip_summed == CHECKSUM_PARTIAL))
5837 		count++;
5838 	count++;
5839 
5840 	count += DIV_ROUND_UP(len, adapter->tx_fifo_limit);
5841 
5842 	nr_frags = skb_shinfo(skb)->nr_frags;
5843 	for (f = 0; f < nr_frags; f++)
5844 		count += DIV_ROUND_UP(skb_frag_size(&skb_shinfo(skb)->frags[f]),
5845 				      adapter->tx_fifo_limit);
5846 
5847 	if (adapter->hw.mac.tx_pkt_filtering)
5848 		e1000_transfer_dhcp_info(adapter, skb);
5849 
5850 	/* need: count + 2 desc gap to keep tail from touching
5851 	 * head, otherwise try next time
5852 	 */
5853 	if (e1000_maybe_stop_tx(tx_ring, count + 2))
5854 		return NETDEV_TX_BUSY;
5855 
5856 	if (skb_vlan_tag_present(skb)) {
5857 		tx_flags |= E1000_TX_FLAGS_VLAN;
5858 		tx_flags |= (skb_vlan_tag_get(skb) <<
5859 			     E1000_TX_FLAGS_VLAN_SHIFT);
5860 	}
5861 
5862 	first = tx_ring->next_to_use;
5863 
5864 	tso = e1000_tso(tx_ring, skb, protocol);
5865 	if (tso < 0) {
5866 		dev_kfree_skb_any(skb);
5867 		return NETDEV_TX_OK;
5868 	}
5869 
5870 	if (tso)
5871 		tx_flags |= E1000_TX_FLAGS_TSO;
5872 	else if (e1000_tx_csum(tx_ring, skb, protocol))
5873 		tx_flags |= E1000_TX_FLAGS_CSUM;
5874 
5875 	/* Old method was to assume IPv4 packet by default if TSO was enabled.
5876 	 * 82571 hardware supports TSO capabilities for IPv6 as well...
5877 	 * no longer assume, we must.
5878 	 */
5879 	if (protocol == htons(ETH_P_IP))
5880 		tx_flags |= E1000_TX_FLAGS_IPV4;
5881 
5882 	if (unlikely(skb->no_fcs))
5883 		tx_flags |= E1000_TX_FLAGS_NO_FCS;
5884 
5885 	/* if count is 0 then mapping error has occurred */
5886 	count = e1000_tx_map(tx_ring, skb, first, adapter->tx_fifo_limit,
5887 			     nr_frags);
5888 	if (count) {
5889 		if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
5890 		    (adapter->flags & FLAG_HAS_HW_TIMESTAMP)) {
5891 			if (!adapter->tx_hwtstamp_skb) {
5892 				skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5893 				tx_flags |= E1000_TX_FLAGS_HWTSTAMP;
5894 				adapter->tx_hwtstamp_skb = skb_get(skb);
5895 				adapter->tx_hwtstamp_start = jiffies;
5896 				schedule_work(&adapter->tx_hwtstamp_work);
5897 			} else {
5898 				adapter->tx_hwtstamp_skipped++;
5899 			}
5900 		}
5901 
5902 		skb_tx_timestamp(skb);
5903 
5904 		netdev_sent_queue(netdev, skb->len);
5905 		e1000_tx_queue(tx_ring, tx_flags, count);
5906 		/* Make sure there is space in the ring for the next send. */
5907 		e1000_maybe_stop_tx(tx_ring,
5908 				    (MAX_SKB_FRAGS *
5909 				     DIV_ROUND_UP(PAGE_SIZE,
5910 						  adapter->tx_fifo_limit) + 2));
5911 
5912 		if (!skb->xmit_more ||
5913 		    netif_xmit_stopped(netdev_get_tx_queue(netdev, 0))) {
5914 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
5915 				e1000e_update_tdt_wa(tx_ring,
5916 						     tx_ring->next_to_use);
5917 			else
5918 				writel(tx_ring->next_to_use, tx_ring->tail);
5919 
5920 			/* we need this if more than one processor can write
5921 			 * to our tail at a time, it synchronizes IO on
5922 			 *IA64/Altix systems
5923 			 */
5924 			mmiowb();
5925 		}
5926 	} else {
5927 		dev_kfree_skb_any(skb);
5928 		tx_ring->buffer_info[first].time_stamp = 0;
5929 		tx_ring->next_to_use = first;
5930 	}
5931 
5932 	return NETDEV_TX_OK;
5933 }
5934 
5935 /**
5936  * e1000_tx_timeout - Respond to a Tx Hang
5937  * @netdev: network interface device structure
5938  **/
5939 static void e1000_tx_timeout(struct net_device *netdev)
5940 {
5941 	struct e1000_adapter *adapter = netdev_priv(netdev);
5942 
5943 	/* Do the reset outside of interrupt context */
5944 	adapter->tx_timeout_count++;
5945 	schedule_work(&adapter->reset_task);
5946 }
5947 
5948 static void e1000_reset_task(struct work_struct *work)
5949 {
5950 	struct e1000_adapter *adapter;
5951 	adapter = container_of(work, struct e1000_adapter, reset_task);
5952 
5953 	/* don't run the task if already down */
5954 	if (test_bit(__E1000_DOWN, &adapter->state))
5955 		return;
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 }
5963 
5964 /**
5965  * e1000_get_stats64 - Get System Network Statistics
5966  * @netdev: network interface device structure
5967  * @stats: rtnl_link_stats64 pointer
5968  *
5969  * Returns the address of the device statistics structure.
5970  **/
5971 void e1000e_get_stats64(struct net_device *netdev,
5972 			struct rtnl_link_stats64 *stats)
5973 {
5974 	struct e1000_adapter *adapter = netdev_priv(netdev);
5975 
5976 	spin_lock(&adapter->stats64_lock);
5977 	e1000e_update_stats(adapter);
5978 	/* Fill out the OS statistics structure */
5979 	stats->rx_bytes = adapter->stats.gorc;
5980 	stats->rx_packets = adapter->stats.gprc;
5981 	stats->tx_bytes = adapter->stats.gotc;
5982 	stats->tx_packets = adapter->stats.gptc;
5983 	stats->multicast = adapter->stats.mprc;
5984 	stats->collisions = adapter->stats.colc;
5985 
5986 	/* Rx Errors */
5987 
5988 	/* RLEC on some newer hardware can be incorrect so build
5989 	 * our own version based on RUC and ROC
5990 	 */
5991 	stats->rx_errors = adapter->stats.rxerrc +
5992 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
5993 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5994 	stats->rx_length_errors = adapter->stats.ruc + adapter->stats.roc;
5995 	stats->rx_crc_errors = adapter->stats.crcerrs;
5996 	stats->rx_frame_errors = adapter->stats.algnerrc;
5997 	stats->rx_missed_errors = adapter->stats.mpc;
5998 
5999 	/* Tx Errors */
6000 	stats->tx_errors = adapter->stats.ecol + adapter->stats.latecol;
6001 	stats->tx_aborted_errors = adapter->stats.ecol;
6002 	stats->tx_window_errors = adapter->stats.latecol;
6003 	stats->tx_carrier_errors = adapter->stats.tncrs;
6004 
6005 	/* Tx Dropped needs to be maintained elsewhere */
6006 
6007 	spin_unlock(&adapter->stats64_lock);
6008 }
6009 
6010 /**
6011  * e1000_change_mtu - Change the Maximum Transfer Unit
6012  * @netdev: network interface device structure
6013  * @new_mtu: new value for maximum frame size
6014  *
6015  * Returns 0 on success, negative on failure
6016  **/
6017 static int e1000_change_mtu(struct net_device *netdev, int new_mtu)
6018 {
6019 	struct e1000_adapter *adapter = netdev_priv(netdev);
6020 	int max_frame = new_mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
6021 
6022 	/* Jumbo frame support */
6023 	if ((new_mtu > ETH_DATA_LEN) &&
6024 	    !(adapter->flags & FLAG_HAS_JUMBO_FRAMES)) {
6025 		e_err("Jumbo Frames not supported.\n");
6026 		return -EINVAL;
6027 	}
6028 
6029 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
6030 	if ((adapter->hw.mac.type >= e1000_pch2lan) &&
6031 	    !(adapter->flags2 & FLAG2_CRC_STRIPPING) &&
6032 	    (new_mtu > ETH_DATA_LEN)) {
6033 		e_err("Jumbo Frames not supported on this device when CRC stripping is disabled.\n");
6034 		return -EINVAL;
6035 	}
6036 
6037 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
6038 		usleep_range(1000, 2000);
6039 	/* e1000e_down -> e1000e_reset dependent on max_frame_size & mtu */
6040 	adapter->max_frame_size = max_frame;
6041 	e_info("changing MTU from %d to %d\n", netdev->mtu, new_mtu);
6042 	netdev->mtu = new_mtu;
6043 
6044 	pm_runtime_get_sync(netdev->dev.parent);
6045 
6046 	if (netif_running(netdev))
6047 		e1000e_down(adapter, true);
6048 
6049 	/* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
6050 	 * means we reserve 2 more, this pushes us to allocate from the next
6051 	 * larger slab size.
6052 	 * i.e. RXBUFFER_2048 --> size-4096 slab
6053 	 * However with the new *_jumbo_rx* routines, jumbo receives will use
6054 	 * fragmented skbs
6055 	 */
6056 
6057 	if (max_frame <= 2048)
6058 		adapter->rx_buffer_len = 2048;
6059 	else
6060 		adapter->rx_buffer_len = 4096;
6061 
6062 	/* adjust allocation if LPE protects us, and we aren't using SBP */
6063 	if (max_frame <= (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN))
6064 		adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
6065 
6066 	if (netif_running(netdev))
6067 		e1000e_up(adapter);
6068 	else
6069 		e1000e_reset(adapter);
6070 
6071 	pm_runtime_put_sync(netdev->dev.parent);
6072 
6073 	clear_bit(__E1000_RESETTING, &adapter->state);
6074 
6075 	return 0;
6076 }
6077 
6078 static int e1000_mii_ioctl(struct net_device *netdev, struct ifreq *ifr,
6079 			   int cmd)
6080 {
6081 	struct e1000_adapter *adapter = netdev_priv(netdev);
6082 	struct mii_ioctl_data *data = if_mii(ifr);
6083 
6084 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
6085 		return -EOPNOTSUPP;
6086 
6087 	switch (cmd) {
6088 	case SIOCGMIIPHY:
6089 		data->phy_id = adapter->hw.phy.addr;
6090 		break;
6091 	case SIOCGMIIREG:
6092 		e1000_phy_read_status(adapter);
6093 
6094 		switch (data->reg_num & 0x1F) {
6095 		case MII_BMCR:
6096 			data->val_out = adapter->phy_regs.bmcr;
6097 			break;
6098 		case MII_BMSR:
6099 			data->val_out = adapter->phy_regs.bmsr;
6100 			break;
6101 		case MII_PHYSID1:
6102 			data->val_out = (adapter->hw.phy.id >> 16);
6103 			break;
6104 		case MII_PHYSID2:
6105 			data->val_out = (adapter->hw.phy.id & 0xFFFF);
6106 			break;
6107 		case MII_ADVERTISE:
6108 			data->val_out = adapter->phy_regs.advertise;
6109 			break;
6110 		case MII_LPA:
6111 			data->val_out = adapter->phy_regs.lpa;
6112 			break;
6113 		case MII_EXPANSION:
6114 			data->val_out = adapter->phy_regs.expansion;
6115 			break;
6116 		case MII_CTRL1000:
6117 			data->val_out = adapter->phy_regs.ctrl1000;
6118 			break;
6119 		case MII_STAT1000:
6120 			data->val_out = adapter->phy_regs.stat1000;
6121 			break;
6122 		case MII_ESTATUS:
6123 			data->val_out = adapter->phy_regs.estatus;
6124 			break;
6125 		default:
6126 			return -EIO;
6127 		}
6128 		break;
6129 	case SIOCSMIIREG:
6130 	default:
6131 		return -EOPNOTSUPP;
6132 	}
6133 	return 0;
6134 }
6135 
6136 /**
6137  * e1000e_hwtstamp_ioctl - control hardware time stamping
6138  * @netdev: network interface device structure
6139  * @ifreq: interface request
6140  *
6141  * Outgoing time stamping can be enabled and disabled. Play nice and
6142  * disable it when requested, although it shouldn't cause any overhead
6143  * when no packet needs it. At most one packet in the queue may be
6144  * marked for time stamping, otherwise it would be impossible to tell
6145  * for sure to which packet the hardware time stamp belongs.
6146  *
6147  * Incoming time stamping has to be configured via the hardware filters.
6148  * Not all combinations are supported, in particular event type has to be
6149  * specified. Matching the kind of event packet is not supported, with the
6150  * exception of "all V2 events regardless of level 2 or 4".
6151  **/
6152 static int e1000e_hwtstamp_set(struct net_device *netdev, struct ifreq *ifr)
6153 {
6154 	struct e1000_adapter *adapter = netdev_priv(netdev);
6155 	struct hwtstamp_config config;
6156 	int ret_val;
6157 
6158 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
6159 		return -EFAULT;
6160 
6161 	ret_val = e1000e_config_hwtstamp(adapter, &config);
6162 	if (ret_val)
6163 		return ret_val;
6164 
6165 	switch (config.rx_filter) {
6166 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
6167 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
6168 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
6169 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
6170 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
6171 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
6172 		/* With V2 type filters which specify a Sync or Delay Request,
6173 		 * Path Delay Request/Response messages are also time stamped
6174 		 * by hardware so notify the caller the requested packets plus
6175 		 * some others are time stamped.
6176 		 */
6177 		config.rx_filter = HWTSTAMP_FILTER_SOME;
6178 		break;
6179 	default:
6180 		break;
6181 	}
6182 
6183 	return copy_to_user(ifr->ifr_data, &config,
6184 			    sizeof(config)) ? -EFAULT : 0;
6185 }
6186 
6187 static int e1000e_hwtstamp_get(struct net_device *netdev, struct ifreq *ifr)
6188 {
6189 	struct e1000_adapter *adapter = netdev_priv(netdev);
6190 
6191 	return copy_to_user(ifr->ifr_data, &adapter->hwtstamp_config,
6192 			    sizeof(adapter->hwtstamp_config)) ? -EFAULT : 0;
6193 }
6194 
6195 static int e1000_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
6196 {
6197 	switch (cmd) {
6198 	case SIOCGMIIPHY:
6199 	case SIOCGMIIREG:
6200 	case SIOCSMIIREG:
6201 		return e1000_mii_ioctl(netdev, ifr, cmd);
6202 	case SIOCSHWTSTAMP:
6203 		return e1000e_hwtstamp_set(netdev, ifr);
6204 	case SIOCGHWTSTAMP:
6205 		return e1000e_hwtstamp_get(netdev, ifr);
6206 	default:
6207 		return -EOPNOTSUPP;
6208 	}
6209 }
6210 
6211 static int e1000_init_phy_wakeup(struct e1000_adapter *adapter, u32 wufc)
6212 {
6213 	struct e1000_hw *hw = &adapter->hw;
6214 	u32 i, mac_reg, wuc;
6215 	u16 phy_reg, wuc_enable;
6216 	int retval;
6217 
6218 	/* copy MAC RARs to PHY RARs */
6219 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
6220 
6221 	retval = hw->phy.ops.acquire(hw);
6222 	if (retval) {
6223 		e_err("Could not acquire PHY\n");
6224 		return retval;
6225 	}
6226 
6227 	/* Enable access to wakeup registers on and set page to BM_WUC_PAGE */
6228 	retval = e1000_enable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6229 	if (retval)
6230 		goto release;
6231 
6232 	/* copy MAC MTA to PHY MTA - only needed for pchlan */
6233 	for (i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
6234 		mac_reg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
6235 		hw->phy.ops.write_reg_page(hw, BM_MTA(i),
6236 					   (u16)(mac_reg & 0xFFFF));
6237 		hw->phy.ops.write_reg_page(hw, BM_MTA(i) + 1,
6238 					   (u16)((mac_reg >> 16) & 0xFFFF));
6239 	}
6240 
6241 	/* configure PHY Rx Control register */
6242 	hw->phy.ops.read_reg_page(&adapter->hw, BM_RCTL, &phy_reg);
6243 	mac_reg = er32(RCTL);
6244 	if (mac_reg & E1000_RCTL_UPE)
6245 		phy_reg |= BM_RCTL_UPE;
6246 	if (mac_reg & E1000_RCTL_MPE)
6247 		phy_reg |= BM_RCTL_MPE;
6248 	phy_reg &= ~(BM_RCTL_MO_MASK);
6249 	if (mac_reg & E1000_RCTL_MO_3)
6250 		phy_reg |= (((mac_reg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
6251 			    << BM_RCTL_MO_SHIFT);
6252 	if (mac_reg & E1000_RCTL_BAM)
6253 		phy_reg |= BM_RCTL_BAM;
6254 	if (mac_reg & E1000_RCTL_PMCF)
6255 		phy_reg |= BM_RCTL_PMCF;
6256 	mac_reg = er32(CTRL);
6257 	if (mac_reg & E1000_CTRL_RFCE)
6258 		phy_reg |= BM_RCTL_RFCE;
6259 	hw->phy.ops.write_reg_page(&adapter->hw, BM_RCTL, phy_reg);
6260 
6261 	wuc = E1000_WUC_PME_EN;
6262 	if (wufc & (E1000_WUFC_MAG | E1000_WUFC_LNKC))
6263 		wuc |= E1000_WUC_APME;
6264 
6265 	/* enable PHY wakeup in MAC register */
6266 	ew32(WUFC, wufc);
6267 	ew32(WUC, (E1000_WUC_PHY_WAKE | E1000_WUC_APMPME |
6268 		   E1000_WUC_PME_STATUS | wuc));
6269 
6270 	/* configure and enable PHY wakeup in PHY registers */
6271 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUFC, wufc);
6272 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUC, wuc);
6273 
6274 	/* activate PHY wakeup */
6275 	wuc_enable |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
6276 	retval = e1000_disable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6277 	if (retval)
6278 		e_err("Could not set PHY Host Wakeup bit\n");
6279 release:
6280 	hw->phy.ops.release(hw);
6281 
6282 	return retval;
6283 }
6284 
6285 static void e1000e_flush_lpic(struct pci_dev *pdev)
6286 {
6287 	struct net_device *netdev = pci_get_drvdata(pdev);
6288 	struct e1000_adapter *adapter = netdev_priv(netdev);
6289 	struct e1000_hw *hw = &adapter->hw;
6290 	u32 ret_val;
6291 
6292 	pm_runtime_get_sync(netdev->dev.parent);
6293 
6294 	ret_val = hw->phy.ops.acquire(hw);
6295 	if (ret_val)
6296 		goto fl_out;
6297 
6298 	pr_info("EEE TX LPI TIMER: %08X\n",
6299 		er32(LPIC) >> E1000_LPIC_LPIET_SHIFT);
6300 
6301 	hw->phy.ops.release(hw);
6302 
6303 fl_out:
6304 	pm_runtime_put_sync(netdev->dev.parent);
6305 }
6306 
6307 static int e1000e_pm_freeze(struct device *dev)
6308 {
6309 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6310 	struct e1000_adapter *adapter = netdev_priv(netdev);
6311 
6312 	netif_device_detach(netdev);
6313 
6314 	if (netif_running(netdev)) {
6315 		int count = E1000_CHECK_RESET_COUNT;
6316 
6317 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6318 			usleep_range(10000, 20000);
6319 
6320 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6321 
6322 		/* Quiesce the device without resetting the hardware */
6323 		e1000e_down(adapter, false);
6324 		e1000_free_irq(adapter);
6325 	}
6326 	e1000e_reset_interrupt_capability(adapter);
6327 
6328 	/* Allow time for pending master requests to run */
6329 	e1000e_disable_pcie_master(&adapter->hw);
6330 
6331 	return 0;
6332 }
6333 
6334 static int __e1000_shutdown(struct pci_dev *pdev, bool runtime)
6335 {
6336 	struct net_device *netdev = pci_get_drvdata(pdev);
6337 	struct e1000_adapter *adapter = netdev_priv(netdev);
6338 	struct e1000_hw *hw = &adapter->hw;
6339 	u32 ctrl, ctrl_ext, rctl, status;
6340 	/* Runtime suspend should only enable wakeup for link changes */
6341 	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
6342 	int retval = 0;
6343 
6344 	status = er32(STATUS);
6345 	if (status & E1000_STATUS_LU)
6346 		wufc &= ~E1000_WUFC_LNKC;
6347 
6348 	if (wufc) {
6349 		e1000_setup_rctl(adapter);
6350 		e1000e_set_rx_mode(netdev);
6351 
6352 		/* turn on all-multi mode if wake on multicast is enabled */
6353 		if (wufc & E1000_WUFC_MC) {
6354 			rctl = er32(RCTL);
6355 			rctl |= E1000_RCTL_MPE;
6356 			ew32(RCTL, rctl);
6357 		}
6358 
6359 		ctrl = er32(CTRL);
6360 		ctrl |= E1000_CTRL_ADVD3WUC;
6361 		if (!(adapter->flags2 & FLAG2_HAS_PHY_WAKEUP))
6362 			ctrl |= E1000_CTRL_EN_PHY_PWR_MGMT;
6363 		ew32(CTRL, ctrl);
6364 
6365 		if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
6366 		    adapter->hw.phy.media_type ==
6367 		    e1000_media_type_internal_serdes) {
6368 			/* keep the laser running in D3 */
6369 			ctrl_ext = er32(CTRL_EXT);
6370 			ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
6371 			ew32(CTRL_EXT, ctrl_ext);
6372 		}
6373 
6374 		if (!runtime)
6375 			e1000e_power_up_phy(adapter);
6376 
6377 		if (adapter->flags & FLAG_IS_ICH)
6378 			e1000_suspend_workarounds_ich8lan(&adapter->hw);
6379 
6380 		if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6381 			/* enable wakeup by the PHY */
6382 			retval = e1000_init_phy_wakeup(adapter, wufc);
6383 			if (retval)
6384 				return retval;
6385 		} else {
6386 			/* enable wakeup by the MAC */
6387 			ew32(WUFC, wufc);
6388 			ew32(WUC, E1000_WUC_PME_EN);
6389 		}
6390 	} else {
6391 		ew32(WUC, 0);
6392 		ew32(WUFC, 0);
6393 
6394 		e1000_power_down_phy(adapter);
6395 	}
6396 
6397 	if (adapter->hw.phy.type == e1000_phy_igp_3) {
6398 		e1000e_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
6399 	} else if (hw->mac.type >= e1000_pch_lpt) {
6400 		if (!(wufc & (E1000_WUFC_EX | E1000_WUFC_MC | E1000_WUFC_BC)))
6401 			/* ULP does not support wake from unicast, multicast
6402 			 * or broadcast.
6403 			 */
6404 			retval = e1000_enable_ulp_lpt_lp(hw, !runtime);
6405 
6406 		if (retval)
6407 			return retval;
6408 	}
6409 
6410 	/* Ensure that the appropriate bits are set in LPI_CTRL
6411 	 * for EEE in Sx
6412 	 */
6413 	if ((hw->phy.type >= e1000_phy_i217) &&
6414 	    adapter->eee_advert && hw->dev_spec.ich8lan.eee_lp_ability) {
6415 		u16 lpi_ctrl = 0;
6416 
6417 		retval = hw->phy.ops.acquire(hw);
6418 		if (!retval) {
6419 			retval = e1e_rphy_locked(hw, I82579_LPI_CTRL,
6420 						 &lpi_ctrl);
6421 			if (!retval) {
6422 				if (adapter->eee_advert &
6423 				    hw->dev_spec.ich8lan.eee_lp_ability &
6424 				    I82579_EEE_100_SUPPORTED)
6425 					lpi_ctrl |= I82579_LPI_CTRL_100_ENABLE;
6426 				if (adapter->eee_advert &
6427 				    hw->dev_spec.ich8lan.eee_lp_ability &
6428 				    I82579_EEE_1000_SUPPORTED)
6429 					lpi_ctrl |= I82579_LPI_CTRL_1000_ENABLE;
6430 
6431 				retval = e1e_wphy_locked(hw, I82579_LPI_CTRL,
6432 							 lpi_ctrl);
6433 			}
6434 		}
6435 		hw->phy.ops.release(hw);
6436 	}
6437 
6438 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
6439 	 * would have already happened in close and is redundant.
6440 	 */
6441 	e1000e_release_hw_control(adapter);
6442 
6443 	pci_clear_master(pdev);
6444 
6445 	/* The pci-e switch on some quad port adapters will report a
6446 	 * correctable error when the MAC transitions from D0 to D3.  To
6447 	 * prevent this we need to mask off the correctable errors on the
6448 	 * downstream port of the pci-e switch.
6449 	 *
6450 	 * We don't have the associated upstream bridge while assigning
6451 	 * the PCI device into guest. For example, the KVM on power is
6452 	 * one of the cases.
6453 	 */
6454 	if (adapter->flags & FLAG_IS_QUAD_PORT) {
6455 		struct pci_dev *us_dev = pdev->bus->self;
6456 		u16 devctl;
6457 
6458 		if (!us_dev)
6459 			return 0;
6460 
6461 		pcie_capability_read_word(us_dev, PCI_EXP_DEVCTL, &devctl);
6462 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL,
6463 					   (devctl & ~PCI_EXP_DEVCTL_CERE));
6464 
6465 		pci_save_state(pdev);
6466 		pci_prepare_to_sleep(pdev);
6467 
6468 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL, devctl);
6469 	}
6470 
6471 	return 0;
6472 }
6473 
6474 /**
6475  * __e1000e_disable_aspm - Disable ASPM states
6476  * @pdev: pointer to PCI device struct
6477  * @state: bit-mask of ASPM states to disable
6478  * @locked: indication if this context holds pci_bus_sem locked.
6479  *
6480  * Some devices *must* have certain ASPM states disabled per hardware errata.
6481  **/
6482 static void __e1000e_disable_aspm(struct pci_dev *pdev, u16 state, int locked)
6483 {
6484 	struct pci_dev *parent = pdev->bus->self;
6485 	u16 aspm_dis_mask = 0;
6486 	u16 pdev_aspmc, parent_aspmc;
6487 
6488 	switch (state) {
6489 	case PCIE_LINK_STATE_L0S:
6490 	case PCIE_LINK_STATE_L0S | PCIE_LINK_STATE_L1:
6491 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L0S;
6492 		/* fall-through - can't have L1 without L0s */
6493 	case PCIE_LINK_STATE_L1:
6494 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L1;
6495 		break;
6496 	default:
6497 		return;
6498 	}
6499 
6500 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6501 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6502 
6503 	if (parent) {
6504 		pcie_capability_read_word(parent, PCI_EXP_LNKCTL,
6505 					  &parent_aspmc);
6506 		parent_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6507 	}
6508 
6509 	/* Nothing to do if the ASPM states to be disabled already are */
6510 	if (!(pdev_aspmc & aspm_dis_mask) &&
6511 	    (!parent || !(parent_aspmc & aspm_dis_mask)))
6512 		return;
6513 
6514 	dev_info(&pdev->dev, "Disabling ASPM %s %s\n",
6515 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L0S) ?
6516 		 "L0s" : "",
6517 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L1) ?
6518 		 "L1" : "");
6519 
6520 #ifdef CONFIG_PCIEASPM
6521 	if (locked)
6522 		pci_disable_link_state_locked(pdev, state);
6523 	else
6524 		pci_disable_link_state(pdev, state);
6525 
6526 	/* Double-check ASPM control.  If not disabled by the above, the
6527 	 * BIOS is preventing that from happening (or CONFIG_PCIEASPM is
6528 	 * not enabled); override by writing PCI config space directly.
6529 	 */
6530 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6531 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6532 
6533 	if (!(aspm_dis_mask & pdev_aspmc))
6534 		return;
6535 #endif
6536 
6537 	/* Both device and parent should have the same ASPM setting.
6538 	 * Disable ASPM in downstream component first and then upstream.
6539 	 */
6540 	pcie_capability_clear_word(pdev, PCI_EXP_LNKCTL, aspm_dis_mask);
6541 
6542 	if (parent)
6543 		pcie_capability_clear_word(parent, PCI_EXP_LNKCTL,
6544 					   aspm_dis_mask);
6545 }
6546 
6547 /**
6548  * e1000e_disable_aspm - Disable ASPM states.
6549  * @pdev: pointer to PCI device struct
6550  * @state: bit-mask of ASPM states to disable
6551  *
6552  * This function acquires the pci_bus_sem!
6553  * Some devices *must* have certain ASPM states disabled per hardware errata.
6554  **/
6555 static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6556 {
6557 	__e1000e_disable_aspm(pdev, state, 0);
6558 }
6559 
6560 /**
6561  * e1000e_disable_aspm_locked   Disable ASPM states.
6562  * @pdev: pointer to PCI device struct
6563  * @state: bit-mask of ASPM states to disable
6564  *
6565  * This function must be called with pci_bus_sem acquired!
6566  * Some devices *must* have certain ASPM states disabled per hardware errata.
6567  **/
6568 static void e1000e_disable_aspm_locked(struct pci_dev *pdev, u16 state)
6569 {
6570 	__e1000e_disable_aspm(pdev, state, 1);
6571 }
6572 
6573 #ifdef CONFIG_PM
6574 static int __e1000_resume(struct pci_dev *pdev)
6575 {
6576 	struct net_device *netdev = pci_get_drvdata(pdev);
6577 	struct e1000_adapter *adapter = netdev_priv(netdev);
6578 	struct e1000_hw *hw = &adapter->hw;
6579 	u16 aspm_disable_flag = 0;
6580 
6581 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6582 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6583 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6584 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6585 	if (aspm_disable_flag)
6586 		e1000e_disable_aspm(pdev, aspm_disable_flag);
6587 
6588 	pci_set_master(pdev);
6589 
6590 	if (hw->mac.type >= e1000_pch2lan)
6591 		e1000_resume_workarounds_pchlan(&adapter->hw);
6592 
6593 	e1000e_power_up_phy(adapter);
6594 
6595 	/* report the system wakeup cause from S3/S4 */
6596 	if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6597 		u16 phy_data;
6598 
6599 		e1e_rphy(&adapter->hw, BM_WUS, &phy_data);
6600 		if (phy_data) {
6601 			e_info("PHY Wakeup cause - %s\n",
6602 			       phy_data & E1000_WUS_EX ? "Unicast Packet" :
6603 			       phy_data & E1000_WUS_MC ? "Multicast Packet" :
6604 			       phy_data & E1000_WUS_BC ? "Broadcast Packet" :
6605 			       phy_data & E1000_WUS_MAG ? "Magic Packet" :
6606 			       phy_data & E1000_WUS_LNKC ?
6607 			       "Link Status Change" : "other");
6608 		}
6609 		e1e_wphy(&adapter->hw, BM_WUS, ~0);
6610 	} else {
6611 		u32 wus = er32(WUS);
6612 
6613 		if (wus) {
6614 			e_info("MAC Wakeup cause - %s\n",
6615 			       wus & E1000_WUS_EX ? "Unicast Packet" :
6616 			       wus & E1000_WUS_MC ? "Multicast Packet" :
6617 			       wus & E1000_WUS_BC ? "Broadcast Packet" :
6618 			       wus & E1000_WUS_MAG ? "Magic Packet" :
6619 			       wus & E1000_WUS_LNKC ? "Link Status Change" :
6620 			       "other");
6621 		}
6622 		ew32(WUS, ~0);
6623 	}
6624 
6625 	e1000e_reset(adapter);
6626 
6627 	e1000_init_manageability_pt(adapter);
6628 
6629 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6630 	 * is up.  For all other cases, let the f/w know that the h/w is now
6631 	 * under the control of the driver.
6632 	 */
6633 	if (!(adapter->flags & FLAG_HAS_AMT))
6634 		e1000e_get_hw_control(adapter);
6635 
6636 	return 0;
6637 }
6638 
6639 #ifdef CONFIG_PM_SLEEP
6640 static int e1000e_pm_thaw(struct device *dev)
6641 {
6642 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6643 	struct e1000_adapter *adapter = netdev_priv(netdev);
6644 
6645 	e1000e_set_interrupt_capability(adapter);
6646 	if (netif_running(netdev)) {
6647 		u32 err = e1000_request_irq(adapter);
6648 
6649 		if (err)
6650 			return err;
6651 
6652 		e1000e_up(adapter);
6653 	}
6654 
6655 	netif_device_attach(netdev);
6656 
6657 	return 0;
6658 }
6659 
6660 static int e1000e_pm_suspend(struct device *dev)
6661 {
6662 	struct pci_dev *pdev = to_pci_dev(dev);
6663 	int rc;
6664 
6665 	e1000e_flush_lpic(pdev);
6666 
6667 	e1000e_pm_freeze(dev);
6668 
6669 	rc = __e1000_shutdown(pdev, false);
6670 	if (rc)
6671 		e1000e_pm_thaw(dev);
6672 
6673 	return rc;
6674 }
6675 
6676 static int e1000e_pm_resume(struct device *dev)
6677 {
6678 	struct pci_dev *pdev = to_pci_dev(dev);
6679 	int rc;
6680 
6681 	rc = __e1000_resume(pdev);
6682 	if (rc)
6683 		return rc;
6684 
6685 	return e1000e_pm_thaw(dev);
6686 }
6687 #endif /* CONFIG_PM_SLEEP */
6688 
6689 static int e1000e_pm_runtime_idle(struct device *dev)
6690 {
6691 	struct pci_dev *pdev = to_pci_dev(dev);
6692 	struct net_device *netdev = pci_get_drvdata(pdev);
6693 	struct e1000_adapter *adapter = netdev_priv(netdev);
6694 	u16 eee_lp;
6695 
6696 	eee_lp = adapter->hw.dev_spec.ich8lan.eee_lp_ability;
6697 
6698 	if (!e1000e_has_link(adapter)) {
6699 		adapter->hw.dev_spec.ich8lan.eee_lp_ability = eee_lp;
6700 		pm_schedule_suspend(dev, 5 * MSEC_PER_SEC);
6701 	}
6702 
6703 	return -EBUSY;
6704 }
6705 
6706 static int e1000e_pm_runtime_resume(struct device *dev)
6707 {
6708 	struct pci_dev *pdev = to_pci_dev(dev);
6709 	struct net_device *netdev = pci_get_drvdata(pdev);
6710 	struct e1000_adapter *adapter = netdev_priv(netdev);
6711 	int rc;
6712 
6713 	rc = __e1000_resume(pdev);
6714 	if (rc)
6715 		return rc;
6716 
6717 	if (netdev->flags & IFF_UP)
6718 		e1000e_up(adapter);
6719 
6720 	return rc;
6721 }
6722 
6723 static int e1000e_pm_runtime_suspend(struct device *dev)
6724 {
6725 	struct pci_dev *pdev = to_pci_dev(dev);
6726 	struct net_device *netdev = pci_get_drvdata(pdev);
6727 	struct e1000_adapter *adapter = netdev_priv(netdev);
6728 
6729 	if (netdev->flags & IFF_UP) {
6730 		int count = E1000_CHECK_RESET_COUNT;
6731 
6732 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6733 			usleep_range(10000, 20000);
6734 
6735 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6736 
6737 		/* Down the device without resetting the hardware */
6738 		e1000e_down(adapter, false);
6739 	}
6740 
6741 	if (__e1000_shutdown(pdev, true)) {
6742 		e1000e_pm_runtime_resume(dev);
6743 		return -EBUSY;
6744 	}
6745 
6746 	return 0;
6747 }
6748 #endif /* CONFIG_PM */
6749 
6750 static void e1000_shutdown(struct pci_dev *pdev)
6751 {
6752 	e1000e_flush_lpic(pdev);
6753 
6754 	e1000e_pm_freeze(&pdev->dev);
6755 
6756 	__e1000_shutdown(pdev, false);
6757 }
6758 
6759 #ifdef CONFIG_NET_POLL_CONTROLLER
6760 
6761 static irqreturn_t e1000_intr_msix(int __always_unused irq, void *data)
6762 {
6763 	struct net_device *netdev = data;
6764 	struct e1000_adapter *adapter = netdev_priv(netdev);
6765 
6766 	if (adapter->msix_entries) {
6767 		int vector, msix_irq;
6768 
6769 		vector = 0;
6770 		msix_irq = adapter->msix_entries[vector].vector;
6771 		if (disable_hardirq(msix_irq))
6772 			e1000_intr_msix_rx(msix_irq, netdev);
6773 		enable_irq(msix_irq);
6774 
6775 		vector++;
6776 		msix_irq = adapter->msix_entries[vector].vector;
6777 		if (disable_hardirq(msix_irq))
6778 			e1000_intr_msix_tx(msix_irq, netdev);
6779 		enable_irq(msix_irq);
6780 
6781 		vector++;
6782 		msix_irq = adapter->msix_entries[vector].vector;
6783 		if (disable_hardirq(msix_irq))
6784 			e1000_msix_other(msix_irq, netdev);
6785 		enable_irq(msix_irq);
6786 	}
6787 
6788 	return IRQ_HANDLED;
6789 }
6790 
6791 /**
6792  * e1000_netpoll
6793  * @netdev: network interface device structure
6794  *
6795  * Polling 'interrupt' - used by things like netconsole to send skbs
6796  * without having to re-enable interrupts. It's not called while
6797  * the interrupt routine is executing.
6798  */
6799 static void e1000_netpoll(struct net_device *netdev)
6800 {
6801 	struct e1000_adapter *adapter = netdev_priv(netdev);
6802 
6803 	switch (adapter->int_mode) {
6804 	case E1000E_INT_MODE_MSIX:
6805 		e1000_intr_msix(adapter->pdev->irq, netdev);
6806 		break;
6807 	case E1000E_INT_MODE_MSI:
6808 		if (disable_hardirq(adapter->pdev->irq))
6809 			e1000_intr_msi(adapter->pdev->irq, netdev);
6810 		enable_irq(adapter->pdev->irq);
6811 		break;
6812 	default:		/* E1000E_INT_MODE_LEGACY */
6813 		if (disable_hardirq(adapter->pdev->irq))
6814 			e1000_intr(adapter->pdev->irq, netdev);
6815 		enable_irq(adapter->pdev->irq);
6816 		break;
6817 	}
6818 }
6819 #endif
6820 
6821 /**
6822  * e1000_io_error_detected - called when PCI error is detected
6823  * @pdev: Pointer to PCI device
6824  * @state: The current pci connection state
6825  *
6826  * This function is called after a PCI bus error affecting
6827  * this device has been detected.
6828  */
6829 static pci_ers_result_t e1000_io_error_detected(struct pci_dev *pdev,
6830 						pci_channel_state_t state)
6831 {
6832 	struct net_device *netdev = pci_get_drvdata(pdev);
6833 	struct e1000_adapter *adapter = netdev_priv(netdev);
6834 
6835 	netif_device_detach(netdev);
6836 
6837 	if (state == pci_channel_io_perm_failure)
6838 		return PCI_ERS_RESULT_DISCONNECT;
6839 
6840 	if (netif_running(netdev))
6841 		e1000e_down(adapter, true);
6842 	pci_disable_device(pdev);
6843 
6844 	/* Request a slot slot reset. */
6845 	return PCI_ERS_RESULT_NEED_RESET;
6846 }
6847 
6848 /**
6849  * e1000_io_slot_reset - called after the pci bus has been reset.
6850  * @pdev: Pointer to PCI device
6851  *
6852  * Restart the card from scratch, as if from a cold-boot. Implementation
6853  * resembles the first-half of the e1000e_pm_resume routine.
6854  */
6855 static pci_ers_result_t e1000_io_slot_reset(struct pci_dev *pdev)
6856 {
6857 	struct net_device *netdev = pci_get_drvdata(pdev);
6858 	struct e1000_adapter *adapter = netdev_priv(netdev);
6859 	struct e1000_hw *hw = &adapter->hw;
6860 	u16 aspm_disable_flag = 0;
6861 	int err;
6862 	pci_ers_result_t result;
6863 
6864 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6865 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6866 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6867 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6868 	if (aspm_disable_flag)
6869 		e1000e_disable_aspm_locked(pdev, aspm_disable_flag);
6870 
6871 	err = pci_enable_device_mem(pdev);
6872 	if (err) {
6873 		dev_err(&pdev->dev,
6874 			"Cannot re-enable PCI device after reset.\n");
6875 		result = PCI_ERS_RESULT_DISCONNECT;
6876 	} else {
6877 		pdev->state_saved = true;
6878 		pci_restore_state(pdev);
6879 		pci_set_master(pdev);
6880 
6881 		pci_enable_wake(pdev, PCI_D3hot, 0);
6882 		pci_enable_wake(pdev, PCI_D3cold, 0);
6883 
6884 		e1000e_reset(adapter);
6885 		ew32(WUS, ~0);
6886 		result = PCI_ERS_RESULT_RECOVERED;
6887 	}
6888 
6889 	pci_cleanup_aer_uncorrect_error_status(pdev);
6890 
6891 	return result;
6892 }
6893 
6894 /**
6895  * e1000_io_resume - called when traffic can start flowing again.
6896  * @pdev: Pointer to PCI device
6897  *
6898  * This callback is called when the error recovery driver tells us that
6899  * its OK to resume normal operation. Implementation resembles the
6900  * second-half of the e1000e_pm_resume routine.
6901  */
6902 static void e1000_io_resume(struct pci_dev *pdev)
6903 {
6904 	struct net_device *netdev = pci_get_drvdata(pdev);
6905 	struct e1000_adapter *adapter = netdev_priv(netdev);
6906 
6907 	e1000_init_manageability_pt(adapter);
6908 
6909 	if (netif_running(netdev))
6910 		e1000e_up(adapter);
6911 
6912 	netif_device_attach(netdev);
6913 
6914 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6915 	 * is up.  For all other cases, let the f/w know that the h/w is now
6916 	 * under the control of the driver.
6917 	 */
6918 	if (!(adapter->flags & FLAG_HAS_AMT))
6919 		e1000e_get_hw_control(adapter);
6920 }
6921 
6922 static void e1000_print_device_info(struct e1000_adapter *adapter)
6923 {
6924 	struct e1000_hw *hw = &adapter->hw;
6925 	struct net_device *netdev = adapter->netdev;
6926 	u32 ret_val;
6927 	u8 pba_str[E1000_PBANUM_LENGTH];
6928 
6929 	/* print bus type/speed/width info */
6930 	e_info("(PCI Express:2.5GT/s:%s) %pM\n",
6931 	       /* bus width */
6932 	       ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
6933 		"Width x1"),
6934 	       /* MAC address */
6935 	       netdev->dev_addr);
6936 	e_info("Intel(R) PRO/%s Network Connection\n",
6937 	       (hw->phy.type == e1000_phy_ife) ? "10/100" : "1000");
6938 	ret_val = e1000_read_pba_string_generic(hw, pba_str,
6939 						E1000_PBANUM_LENGTH);
6940 	if (ret_val)
6941 		strlcpy((char *)pba_str, "Unknown", sizeof(pba_str));
6942 	e_info("MAC: %d, PHY: %d, PBA No: %s\n",
6943 	       hw->mac.type, hw->phy.type, pba_str);
6944 }
6945 
6946 static void e1000_eeprom_checks(struct e1000_adapter *adapter)
6947 {
6948 	struct e1000_hw *hw = &adapter->hw;
6949 	int ret_val;
6950 	u16 buf = 0;
6951 
6952 	if (hw->mac.type != e1000_82573)
6953 		return;
6954 
6955 	ret_val = e1000_read_nvm(hw, NVM_INIT_CONTROL2_REG, 1, &buf);
6956 	le16_to_cpus(&buf);
6957 	if (!ret_val && (!(buf & BIT(0)))) {
6958 		/* Deep Smart Power Down (DSPD) */
6959 		dev_warn(&adapter->pdev->dev,
6960 			 "Warning: detected DSPD enabled in EEPROM\n");
6961 	}
6962 }
6963 
6964 static netdev_features_t e1000_fix_features(struct net_device *netdev,
6965 					    netdev_features_t features)
6966 {
6967 	struct e1000_adapter *adapter = netdev_priv(netdev);
6968 	struct e1000_hw *hw = &adapter->hw;
6969 
6970 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
6971 	if ((hw->mac.type >= e1000_pch2lan) && (netdev->mtu > ETH_DATA_LEN))
6972 		features &= ~NETIF_F_RXFCS;
6973 
6974 	/* Since there is no support for separate Rx/Tx vlan accel
6975 	 * enable/disable make sure Tx flag is always in same state as Rx.
6976 	 */
6977 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
6978 		features |= NETIF_F_HW_VLAN_CTAG_TX;
6979 	else
6980 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
6981 
6982 	return features;
6983 }
6984 
6985 static int e1000_set_features(struct net_device *netdev,
6986 			      netdev_features_t features)
6987 {
6988 	struct e1000_adapter *adapter = netdev_priv(netdev);
6989 	netdev_features_t changed = features ^ netdev->features;
6990 
6991 	if (changed & (NETIF_F_TSO | NETIF_F_TSO6))
6992 		adapter->flags |= FLAG_TSO_FORCE;
6993 
6994 	if (!(changed & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX |
6995 			 NETIF_F_RXCSUM | NETIF_F_RXHASH | NETIF_F_RXFCS |
6996 			 NETIF_F_RXALL)))
6997 		return 0;
6998 
6999 	if (changed & NETIF_F_RXFCS) {
7000 		if (features & NETIF_F_RXFCS) {
7001 			adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7002 		} else {
7003 			/* We need to take it back to defaults, which might mean
7004 			 * stripping is still disabled at the adapter level.
7005 			 */
7006 			if (adapter->flags2 & FLAG2_DFLT_CRC_STRIPPING)
7007 				adapter->flags2 |= FLAG2_CRC_STRIPPING;
7008 			else
7009 				adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7010 		}
7011 	}
7012 
7013 	netdev->features = features;
7014 
7015 	if (netif_running(netdev))
7016 		e1000e_reinit_locked(adapter);
7017 	else
7018 		e1000e_reset(adapter);
7019 
7020 	return 0;
7021 }
7022 
7023 static const struct net_device_ops e1000e_netdev_ops = {
7024 	.ndo_open		= e1000e_open,
7025 	.ndo_stop		= e1000e_close,
7026 	.ndo_start_xmit		= e1000_xmit_frame,
7027 	.ndo_get_stats64	= e1000e_get_stats64,
7028 	.ndo_set_rx_mode	= e1000e_set_rx_mode,
7029 	.ndo_set_mac_address	= e1000_set_mac,
7030 	.ndo_change_mtu		= e1000_change_mtu,
7031 	.ndo_do_ioctl		= e1000_ioctl,
7032 	.ndo_tx_timeout		= e1000_tx_timeout,
7033 	.ndo_validate_addr	= eth_validate_addr,
7034 
7035 	.ndo_vlan_rx_add_vid	= e1000_vlan_rx_add_vid,
7036 	.ndo_vlan_rx_kill_vid	= e1000_vlan_rx_kill_vid,
7037 #ifdef CONFIG_NET_POLL_CONTROLLER
7038 	.ndo_poll_controller	= e1000_netpoll,
7039 #endif
7040 	.ndo_set_features = e1000_set_features,
7041 	.ndo_fix_features = e1000_fix_features,
7042 	.ndo_features_check	= passthru_features_check,
7043 };
7044 
7045 /**
7046  * e1000_probe - Device Initialization Routine
7047  * @pdev: PCI device information struct
7048  * @ent: entry in e1000_pci_tbl
7049  *
7050  * Returns 0 on success, negative on failure
7051  *
7052  * e1000_probe initializes an adapter identified by a pci_dev structure.
7053  * The OS initialization, configuring of the adapter private structure,
7054  * and a hardware reset occur.
7055  **/
7056 static int e1000_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7057 {
7058 	struct net_device *netdev;
7059 	struct e1000_adapter *adapter;
7060 	struct e1000_hw *hw;
7061 	const struct e1000_info *ei = e1000_info_tbl[ent->driver_data];
7062 	resource_size_t mmio_start, mmio_len;
7063 	resource_size_t flash_start, flash_len;
7064 	static int cards_found;
7065 	u16 aspm_disable_flag = 0;
7066 	int bars, i, err, pci_using_dac;
7067 	u16 eeprom_data = 0;
7068 	u16 eeprom_apme_mask = E1000_EEPROM_APME;
7069 	s32 ret_val = 0;
7070 
7071 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L0S)
7072 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7073 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L1)
7074 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7075 	if (aspm_disable_flag)
7076 		e1000e_disable_aspm(pdev, aspm_disable_flag);
7077 
7078 	err = pci_enable_device_mem(pdev);
7079 	if (err)
7080 		return err;
7081 
7082 	pci_using_dac = 0;
7083 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
7084 	if (!err) {
7085 		pci_using_dac = 1;
7086 	} else {
7087 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
7088 		if (err) {
7089 			dev_err(&pdev->dev,
7090 				"No usable DMA configuration, aborting\n");
7091 			goto err_dma;
7092 		}
7093 	}
7094 
7095 	bars = pci_select_bars(pdev, IORESOURCE_MEM);
7096 	err = pci_request_selected_regions_exclusive(pdev, bars,
7097 						     e1000e_driver_name);
7098 	if (err)
7099 		goto err_pci_reg;
7100 
7101 	/* AER (Advanced Error Reporting) hooks */
7102 	pci_enable_pcie_error_reporting(pdev);
7103 
7104 	pci_set_master(pdev);
7105 	/* PCI config space info */
7106 	err = pci_save_state(pdev);
7107 	if (err)
7108 		goto err_alloc_etherdev;
7109 
7110 	err = -ENOMEM;
7111 	netdev = alloc_etherdev(sizeof(struct e1000_adapter));
7112 	if (!netdev)
7113 		goto err_alloc_etherdev;
7114 
7115 	SET_NETDEV_DEV(netdev, &pdev->dev);
7116 
7117 	netdev->irq = pdev->irq;
7118 
7119 	pci_set_drvdata(pdev, netdev);
7120 	adapter = netdev_priv(netdev);
7121 	hw = &adapter->hw;
7122 	adapter->netdev = netdev;
7123 	adapter->pdev = pdev;
7124 	adapter->ei = ei;
7125 	adapter->pba = ei->pba;
7126 	adapter->flags = ei->flags;
7127 	adapter->flags2 = ei->flags2;
7128 	adapter->hw.adapter = adapter;
7129 	adapter->hw.mac.type = ei->mac;
7130 	adapter->max_hw_frame_size = ei->max_hw_frame_size;
7131 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
7132 
7133 	mmio_start = pci_resource_start(pdev, 0);
7134 	mmio_len = pci_resource_len(pdev, 0);
7135 
7136 	err = -EIO;
7137 	adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
7138 	if (!adapter->hw.hw_addr)
7139 		goto err_ioremap;
7140 
7141 	if ((adapter->flags & FLAG_HAS_FLASH) &&
7142 	    (pci_resource_flags(pdev, 1) & IORESOURCE_MEM) &&
7143 	    (hw->mac.type < e1000_pch_spt)) {
7144 		flash_start = pci_resource_start(pdev, 1);
7145 		flash_len = pci_resource_len(pdev, 1);
7146 		adapter->hw.flash_address = ioremap(flash_start, flash_len);
7147 		if (!adapter->hw.flash_address)
7148 			goto err_flashmap;
7149 	}
7150 
7151 	/* Set default EEE advertisement */
7152 	if (adapter->flags2 & FLAG2_HAS_EEE)
7153 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
7154 
7155 	/* construct the net_device struct */
7156 	netdev->netdev_ops = &e1000e_netdev_ops;
7157 	e1000e_set_ethtool_ops(netdev);
7158 	netdev->watchdog_timeo = 5 * HZ;
7159 	netif_napi_add(netdev, &adapter->napi, e1000e_poll, 64);
7160 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
7161 
7162 	netdev->mem_start = mmio_start;
7163 	netdev->mem_end = mmio_start + mmio_len;
7164 
7165 	adapter->bd_number = cards_found++;
7166 
7167 	e1000e_check_options(adapter);
7168 
7169 	/* setup adapter struct */
7170 	err = e1000_sw_init(adapter);
7171 	if (err)
7172 		goto err_sw_init;
7173 
7174 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
7175 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
7176 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
7177 
7178 	err = ei->get_variants(adapter);
7179 	if (err)
7180 		goto err_hw_init;
7181 
7182 	if ((adapter->flags & FLAG_IS_ICH) &&
7183 	    (adapter->flags & FLAG_READ_ONLY_NVM) &&
7184 	    (hw->mac.type < e1000_pch_spt))
7185 		e1000e_write_protect_nvm_ich8lan(&adapter->hw);
7186 
7187 	hw->mac.ops.get_bus_info(&adapter->hw);
7188 
7189 	adapter->hw.phy.autoneg_wait_to_complete = 0;
7190 
7191 	/* Copper options */
7192 	if (adapter->hw.phy.media_type == e1000_media_type_copper) {
7193 		adapter->hw.phy.mdix = AUTO_ALL_MODES;
7194 		adapter->hw.phy.disable_polarity_correction = 0;
7195 		adapter->hw.phy.ms_type = e1000_ms_hw_default;
7196 	}
7197 
7198 	if (hw->phy.ops.check_reset_block && hw->phy.ops.check_reset_block(hw))
7199 		dev_info(&pdev->dev,
7200 			 "PHY reset is blocked due to SOL/IDER session.\n");
7201 
7202 	/* Set initial default active device features */
7203 	netdev->features = (NETIF_F_SG |
7204 			    NETIF_F_HW_VLAN_CTAG_RX |
7205 			    NETIF_F_HW_VLAN_CTAG_TX |
7206 			    NETIF_F_TSO |
7207 			    NETIF_F_TSO6 |
7208 			    NETIF_F_RXHASH |
7209 			    NETIF_F_RXCSUM |
7210 			    NETIF_F_HW_CSUM);
7211 
7212 	/* Set user-changeable features (subset of all device features) */
7213 	netdev->hw_features = netdev->features;
7214 	netdev->hw_features |= NETIF_F_RXFCS;
7215 	netdev->priv_flags |= IFF_SUPP_NOFCS;
7216 	netdev->hw_features |= NETIF_F_RXALL;
7217 
7218 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER)
7219 		netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
7220 
7221 	netdev->vlan_features |= (NETIF_F_SG |
7222 				  NETIF_F_TSO |
7223 				  NETIF_F_TSO6 |
7224 				  NETIF_F_HW_CSUM);
7225 
7226 	netdev->priv_flags |= IFF_UNICAST_FLT;
7227 
7228 	if (pci_using_dac) {
7229 		netdev->features |= NETIF_F_HIGHDMA;
7230 		netdev->vlan_features |= NETIF_F_HIGHDMA;
7231 	}
7232 
7233 	/* MTU range: 68 - max_hw_frame_size */
7234 	netdev->min_mtu = ETH_MIN_MTU;
7235 	netdev->max_mtu = adapter->max_hw_frame_size -
7236 			  (VLAN_ETH_HLEN + ETH_FCS_LEN);
7237 
7238 	if (e1000e_enable_mng_pass_thru(&adapter->hw))
7239 		adapter->flags |= FLAG_MNG_PT_ENABLED;
7240 
7241 	/* before reading the NVM, reset the controller to
7242 	 * put the device in a known good starting state
7243 	 */
7244 	adapter->hw.mac.ops.reset_hw(&adapter->hw);
7245 
7246 	/* systems with ASPM and others may see the checksum fail on the first
7247 	 * attempt. Let's give it a few tries
7248 	 */
7249 	for (i = 0;; i++) {
7250 		if (e1000_validate_nvm_checksum(&adapter->hw) >= 0)
7251 			break;
7252 		if (i == 2) {
7253 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
7254 			err = -EIO;
7255 			goto err_eeprom;
7256 		}
7257 	}
7258 
7259 	e1000_eeprom_checks(adapter);
7260 
7261 	/* copy the MAC address */
7262 	if (e1000e_read_mac_addr(&adapter->hw))
7263 		dev_err(&pdev->dev,
7264 			"NVM Read Error while reading MAC address\n");
7265 
7266 	memcpy(netdev->dev_addr, adapter->hw.mac.addr, netdev->addr_len);
7267 
7268 	if (!is_valid_ether_addr(netdev->dev_addr)) {
7269 		dev_err(&pdev->dev, "Invalid MAC Address: %pM\n",
7270 			netdev->dev_addr);
7271 		err = -EIO;
7272 		goto err_eeprom;
7273 	}
7274 
7275 	timer_setup(&adapter->watchdog_timer, e1000_watchdog, 0);
7276 	timer_setup(&adapter->phy_info_timer, e1000_update_phy_info, 0);
7277 
7278 	INIT_WORK(&adapter->reset_task, e1000_reset_task);
7279 	INIT_WORK(&adapter->watchdog_task, e1000_watchdog_task);
7280 	INIT_WORK(&adapter->downshift_task, e1000e_downshift_workaround);
7281 	INIT_WORK(&adapter->update_phy_task, e1000e_update_phy_task);
7282 	INIT_WORK(&adapter->print_hang_task, e1000_print_hw_hang);
7283 
7284 	/* Initialize link parameters. User can change them with ethtool */
7285 	adapter->hw.mac.autoneg = 1;
7286 	adapter->fc_autoneg = true;
7287 	adapter->hw.fc.requested_mode = e1000_fc_default;
7288 	adapter->hw.fc.current_mode = e1000_fc_default;
7289 	adapter->hw.phy.autoneg_advertised = 0x2f;
7290 
7291 	/* Initial Wake on LAN setting - If APM wake is enabled in
7292 	 * the EEPROM, enable the ACPI Magic Packet filter
7293 	 */
7294 	if (adapter->flags & FLAG_APME_IN_WUC) {
7295 		/* APME bit in EEPROM is mapped to WUC.APME */
7296 		eeprom_data = er32(WUC);
7297 		eeprom_apme_mask = E1000_WUC_APME;
7298 		if ((hw->mac.type > e1000_ich10lan) &&
7299 		    (eeprom_data & E1000_WUC_PHY_WAKE))
7300 			adapter->flags2 |= FLAG2_HAS_PHY_WAKEUP;
7301 	} else if (adapter->flags & FLAG_APME_IN_CTRL3) {
7302 		if (adapter->flags & FLAG_APME_CHECK_PORT_B &&
7303 		    (adapter->hw.bus.func == 1))
7304 			ret_val = e1000_read_nvm(&adapter->hw,
7305 					      NVM_INIT_CONTROL3_PORT_B,
7306 					      1, &eeprom_data);
7307 		else
7308 			ret_val = e1000_read_nvm(&adapter->hw,
7309 					      NVM_INIT_CONTROL3_PORT_A,
7310 					      1, &eeprom_data);
7311 	}
7312 
7313 	/* fetch WoL from EEPROM */
7314 	if (ret_val)
7315 		e_dbg("NVM read error getting WoL initial values: %d\n", ret_val);
7316 	else if (eeprom_data & eeprom_apme_mask)
7317 		adapter->eeprom_wol |= E1000_WUFC_MAG;
7318 
7319 	/* now that we have the eeprom settings, apply the special cases
7320 	 * where the eeprom may be wrong or the board simply won't support
7321 	 * wake on lan on a particular port
7322 	 */
7323 	if (!(adapter->flags & FLAG_HAS_WOL))
7324 		adapter->eeprom_wol = 0;
7325 
7326 	/* initialize the wol settings based on the eeprom settings */
7327 	adapter->wol = adapter->eeprom_wol;
7328 
7329 	/* make sure adapter isn't asleep if manageability is enabled */
7330 	if (adapter->wol || (adapter->flags & FLAG_MNG_PT_ENABLED) ||
7331 	    (hw->mac.ops.check_mng_mode(hw)))
7332 		device_wakeup_enable(&pdev->dev);
7333 
7334 	/* save off EEPROM version number */
7335 	ret_val = e1000_read_nvm(&adapter->hw, 5, 1, &adapter->eeprom_vers);
7336 
7337 	if (ret_val) {
7338 		e_dbg("NVM read error getting EEPROM version: %d\n", ret_val);
7339 		adapter->eeprom_vers = 0;
7340 	}
7341 
7342 	/* init PTP hardware clock */
7343 	e1000e_ptp_init(adapter);
7344 
7345 	/* reset the hardware with the new settings */
7346 	e1000e_reset(adapter);
7347 
7348 	/* If the controller has AMT, do not set DRV_LOAD until the interface
7349 	 * is up.  For all other cases, let the f/w know that the h/w is now
7350 	 * under the control of the driver.
7351 	 */
7352 	if (!(adapter->flags & FLAG_HAS_AMT))
7353 		e1000e_get_hw_control(adapter);
7354 
7355 	strlcpy(netdev->name, "eth%d", sizeof(netdev->name));
7356 	err = register_netdev(netdev);
7357 	if (err)
7358 		goto err_register;
7359 
7360 	/* carrier off reporting is important to ethtool even BEFORE open */
7361 	netif_carrier_off(netdev);
7362 
7363 	e1000_print_device_info(adapter);
7364 
7365 	if (pci_dev_run_wake(pdev))
7366 		pm_runtime_put_noidle(&pdev->dev);
7367 
7368 	return 0;
7369 
7370 err_register:
7371 	if (!(adapter->flags & FLAG_HAS_AMT))
7372 		e1000e_release_hw_control(adapter);
7373 err_eeprom:
7374 	if (hw->phy.ops.check_reset_block && !hw->phy.ops.check_reset_block(hw))
7375 		e1000_phy_hw_reset(&adapter->hw);
7376 err_hw_init:
7377 	kfree(adapter->tx_ring);
7378 	kfree(adapter->rx_ring);
7379 err_sw_init:
7380 	if ((adapter->hw.flash_address) && (hw->mac.type < e1000_pch_spt))
7381 		iounmap(adapter->hw.flash_address);
7382 	e1000e_reset_interrupt_capability(adapter);
7383 err_flashmap:
7384 	iounmap(adapter->hw.hw_addr);
7385 err_ioremap:
7386 	free_netdev(netdev);
7387 err_alloc_etherdev:
7388 	pci_release_mem_regions(pdev);
7389 err_pci_reg:
7390 err_dma:
7391 	pci_disable_device(pdev);
7392 	return err;
7393 }
7394 
7395 /**
7396  * e1000_remove - Device Removal Routine
7397  * @pdev: PCI device information struct
7398  *
7399  * e1000_remove is called by the PCI subsystem to alert the driver
7400  * that it should release a PCI device.  The could be caused by a
7401  * Hot-Plug event, or because the driver is going to be removed from
7402  * memory.
7403  **/
7404 static void e1000_remove(struct pci_dev *pdev)
7405 {
7406 	struct net_device *netdev = pci_get_drvdata(pdev);
7407 	struct e1000_adapter *adapter = netdev_priv(netdev);
7408 	bool down = test_bit(__E1000_DOWN, &adapter->state);
7409 
7410 	e1000e_ptp_remove(adapter);
7411 
7412 	/* The timers may be rescheduled, so explicitly disable them
7413 	 * from being rescheduled.
7414 	 */
7415 	if (!down)
7416 		set_bit(__E1000_DOWN, &adapter->state);
7417 	del_timer_sync(&adapter->watchdog_timer);
7418 	del_timer_sync(&adapter->phy_info_timer);
7419 
7420 	cancel_work_sync(&adapter->reset_task);
7421 	cancel_work_sync(&adapter->watchdog_task);
7422 	cancel_work_sync(&adapter->downshift_task);
7423 	cancel_work_sync(&adapter->update_phy_task);
7424 	cancel_work_sync(&adapter->print_hang_task);
7425 
7426 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
7427 		cancel_work_sync(&adapter->tx_hwtstamp_work);
7428 		if (adapter->tx_hwtstamp_skb) {
7429 			dev_consume_skb_any(adapter->tx_hwtstamp_skb);
7430 			adapter->tx_hwtstamp_skb = NULL;
7431 		}
7432 	}
7433 
7434 	/* Don't lie to e1000_close() down the road. */
7435 	if (!down)
7436 		clear_bit(__E1000_DOWN, &adapter->state);
7437 	unregister_netdev(netdev);
7438 
7439 	if (pci_dev_run_wake(pdev))
7440 		pm_runtime_get_noresume(&pdev->dev);
7441 
7442 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
7443 	 * would have already happened in close and is redundant.
7444 	 */
7445 	e1000e_release_hw_control(adapter);
7446 
7447 	e1000e_reset_interrupt_capability(adapter);
7448 	kfree(adapter->tx_ring);
7449 	kfree(adapter->rx_ring);
7450 
7451 	iounmap(adapter->hw.hw_addr);
7452 	if ((adapter->hw.flash_address) &&
7453 	    (adapter->hw.mac.type < e1000_pch_spt))
7454 		iounmap(adapter->hw.flash_address);
7455 	pci_release_mem_regions(pdev);
7456 
7457 	free_netdev(netdev);
7458 
7459 	/* AER disable */
7460 	pci_disable_pcie_error_reporting(pdev);
7461 
7462 	pci_disable_device(pdev);
7463 }
7464 
7465 /* PCI Error Recovery (ERS) */
7466 static const struct pci_error_handlers e1000_err_handler = {
7467 	.error_detected = e1000_io_error_detected,
7468 	.slot_reset = e1000_io_slot_reset,
7469 	.resume = e1000_io_resume,
7470 };
7471 
7472 static const struct pci_device_id e1000_pci_tbl[] = {
7473 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_COPPER), board_82571 },
7474 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_FIBER), board_82571 },
7475 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER), board_82571 },
7476 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER_LP),
7477 	  board_82571 },
7478 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_FIBER), board_82571 },
7479 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES), board_82571 },
7480 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_DUAL), board_82571 },
7481 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_QUAD), board_82571 },
7482 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571PT_QUAD_COPPER), board_82571 },
7483 
7484 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI), board_82572 },
7485 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_COPPER), board_82572 },
7486 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_FIBER), board_82572 },
7487 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_SERDES), board_82572 },
7488 
7489 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E), board_82573 },
7490 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E_IAMT), board_82573 },
7491 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573L), board_82573 },
7492 
7493 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574L), board_82574 },
7494 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574LA), board_82574 },
7495 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82583V), board_82583 },
7496 
7497 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_DPT),
7498 	  board_80003es2lan },
7499 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_SPT),
7500 	  board_80003es2lan },
7501 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_DPT),
7502 	  board_80003es2lan },
7503 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_SPT),
7504 	  board_80003es2lan },
7505 
7506 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE), board_ich8lan },
7507 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_G), board_ich8lan },
7508 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_GT), board_ich8lan },
7509 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_AMT), board_ich8lan },
7510 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_C), board_ich8lan },
7511 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M), board_ich8lan },
7512 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M_AMT), board_ich8lan },
7513 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_82567V_3), board_ich8lan },
7514 
7515 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE), board_ich9lan },
7516 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_G), board_ich9lan },
7517 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_GT), board_ich9lan },
7518 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_AMT), board_ich9lan },
7519 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_C), board_ich9lan },
7520 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_BM), board_ich9lan },
7521 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M), board_ich9lan },
7522 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_AMT), board_ich9lan },
7523 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_V), board_ich9lan },
7524 
7525 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LM), board_ich9lan },
7526 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LF), board_ich9lan },
7527 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_V), board_ich9lan },
7528 
7529 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LM), board_ich10lan },
7530 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LF), board_ich10lan },
7531 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_V), board_ich10lan },
7532 
7533 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LM), board_pchlan },
7534 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LC), board_pchlan },
7535 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DM), board_pchlan },
7536 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DC), board_pchlan },
7537 
7538 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_LM), board_pch2lan },
7539 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_V), board_pch2lan },
7540 
7541 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_LM), board_pch_lpt },
7542 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_V), board_pch_lpt },
7543 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_LM), board_pch_lpt },
7544 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_V), board_pch_lpt },
7545 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM2), board_pch_lpt },
7546 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V2), board_pch_lpt },
7547 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM3), board_pch_lpt },
7548 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V3), board_pch_lpt },
7549 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM), board_pch_spt },
7550 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V), board_pch_spt },
7551 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM2), board_pch_spt },
7552 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V2), board_pch_spt },
7553 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LBG_I219_LM3), board_pch_spt },
7554 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM4), board_pch_spt },
7555 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V4), board_pch_spt },
7556 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM5), board_pch_spt },
7557 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V5), board_pch_spt },
7558 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM6), board_pch_cnp },
7559 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V6), board_pch_cnp },
7560 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM7), board_pch_cnp },
7561 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V7), board_pch_cnp },
7562 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM8), board_pch_cnp },
7563 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V8), board_pch_cnp },
7564 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM9), board_pch_cnp },
7565 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V9), board_pch_cnp },
7566 
7567 	{ 0, 0, 0, 0, 0, 0, 0 }	/* terminate list */
7568 };
7569 MODULE_DEVICE_TABLE(pci, e1000_pci_tbl);
7570 
7571 static const struct dev_pm_ops e1000_pm_ops = {
7572 #ifdef CONFIG_PM_SLEEP
7573 	.suspend	= e1000e_pm_suspend,
7574 	.resume		= e1000e_pm_resume,
7575 	.freeze		= e1000e_pm_freeze,
7576 	.thaw		= e1000e_pm_thaw,
7577 	.poweroff	= e1000e_pm_suspend,
7578 	.restore	= e1000e_pm_resume,
7579 #endif
7580 	SET_RUNTIME_PM_OPS(e1000e_pm_runtime_suspend, e1000e_pm_runtime_resume,
7581 			   e1000e_pm_runtime_idle)
7582 };
7583 
7584 /* PCI Device API Driver */
7585 static struct pci_driver e1000_driver = {
7586 	.name     = e1000e_driver_name,
7587 	.id_table = e1000_pci_tbl,
7588 	.probe    = e1000_probe,
7589 	.remove   = e1000_remove,
7590 	.driver   = {
7591 		.pm = &e1000_pm_ops,
7592 	},
7593 	.shutdown = e1000_shutdown,
7594 	.err_handler = &e1000_err_handler
7595 };
7596 
7597 /**
7598  * e1000_init_module - Driver Registration Routine
7599  *
7600  * e1000_init_module is the first routine called when the driver is
7601  * loaded. All it does is register with the PCI subsystem.
7602  **/
7603 static int __init e1000_init_module(void)
7604 {
7605 	pr_info("Intel(R) PRO/1000 Network Driver - %s\n",
7606 		e1000e_driver_version);
7607 	pr_info("Copyright(c) 1999 - 2015 Intel Corporation.\n");
7608 
7609 	return pci_register_driver(&e1000_driver);
7610 }
7611 module_init(e1000_init_module);
7612 
7613 /**
7614  * e1000_exit_module - Driver Exit Cleanup Routine
7615  *
7616  * e1000_exit_module is called just before the driver is removed
7617  * from memory.
7618  **/
7619 static void __exit e1000_exit_module(void)
7620 {
7621 	pci_unregister_driver(&e1000_driver);
7622 }
7623 module_exit(e1000_exit_module);
7624 
7625 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
7626 MODULE_DESCRIPTION("Intel(R) PRO/1000 Network Driver");
7627 MODULE_LICENSE("GPL");
7628 MODULE_VERSION(DRV_VERSION);
7629 
7630 /* netdev.c */
7631