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 };
75 
76 struct e1000_reg_info {
77 	u32 ofs;
78 	char *name;
79 };
80 
81 static const struct e1000_reg_info e1000_reg_info_tbl[] = {
82 	/* General Registers */
83 	{E1000_CTRL, "CTRL"},
84 	{E1000_STATUS, "STATUS"},
85 	{E1000_CTRL_EXT, "CTRL_EXT"},
86 
87 	/* Interrupt Registers */
88 	{E1000_ICR, "ICR"},
89 
90 	/* Rx Registers */
91 	{E1000_RCTL, "RCTL"},
92 	{E1000_RDLEN(0), "RDLEN"},
93 	{E1000_RDH(0), "RDH"},
94 	{E1000_RDT(0), "RDT"},
95 	{E1000_RDTR, "RDTR"},
96 	{E1000_RXDCTL(0), "RXDCTL"},
97 	{E1000_ERT, "ERT"},
98 	{E1000_RDBAL(0), "RDBAL"},
99 	{E1000_RDBAH(0), "RDBAH"},
100 	{E1000_RDFH, "RDFH"},
101 	{E1000_RDFT, "RDFT"},
102 	{E1000_RDFHS, "RDFHS"},
103 	{E1000_RDFTS, "RDFTS"},
104 	{E1000_RDFPC, "RDFPC"},
105 
106 	/* Tx Registers */
107 	{E1000_TCTL, "TCTL"},
108 	{E1000_TDBAL(0), "TDBAL"},
109 	{E1000_TDBAH(0), "TDBAH"},
110 	{E1000_TDLEN(0), "TDLEN"},
111 	{E1000_TDH(0), "TDH"},
112 	{E1000_TDT(0), "TDT"},
113 	{E1000_TIDV, "TIDV"},
114 	{E1000_TXDCTL(0), "TXDCTL"},
115 	{E1000_TADV, "TADV"},
116 	{E1000_TARC(0), "TARC"},
117 	{E1000_TDFH, "TDFH"},
118 	{E1000_TDFT, "TDFT"},
119 	{E1000_TDFHS, "TDFHS"},
120 	{E1000_TDFTS, "TDFTS"},
121 	{E1000_TDFPC, "TDFPC"},
122 
123 	/* List Terminator */
124 	{0, NULL}
125 };
126 
127 /**
128  * __ew32_prepare - prepare to write to MAC CSR register on certain parts
129  * @hw: pointer to the HW structure
130  *
131  * When updating the MAC CSR registers, the Manageability Engine (ME) could
132  * be accessing the registers at the same time.  Normally, this is handled in
133  * h/w by an arbiter but on some parts there is a bug that acknowledges Host
134  * accesses later than it should which could result in the register to have
135  * an incorrect value.  Workaround this by checking the FWSM register which
136  * has bit 24 set while ME is accessing MAC CSR registers, wait if it is set
137  * and try again a number of times.
138  **/
139 s32 __ew32_prepare(struct e1000_hw *hw)
140 {
141 	s32 i = E1000_ICH_FWSM_PCIM2PCI_COUNT;
142 
143 	while ((er32(FWSM) & E1000_ICH_FWSM_PCIM2PCI) && --i)
144 		udelay(50);
145 
146 	return i;
147 }
148 
149 void __ew32(struct e1000_hw *hw, unsigned long reg, u32 val)
150 {
151 	if (hw->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
152 		__ew32_prepare(hw);
153 
154 	writel(val, hw->hw_addr + reg);
155 }
156 
157 /**
158  * e1000_regdump - register printout routine
159  * @hw: pointer to the HW structure
160  * @reginfo: pointer to the register info table
161  **/
162 static void e1000_regdump(struct e1000_hw *hw, struct e1000_reg_info *reginfo)
163 {
164 	int n = 0;
165 	char rname[16];
166 	u32 regs[8];
167 
168 	switch (reginfo->ofs) {
169 	case E1000_RXDCTL(0):
170 		for (n = 0; n < 2; n++)
171 			regs[n] = __er32(hw, E1000_RXDCTL(n));
172 		break;
173 	case E1000_TXDCTL(0):
174 		for (n = 0; n < 2; n++)
175 			regs[n] = __er32(hw, E1000_TXDCTL(n));
176 		break;
177 	case E1000_TARC(0):
178 		for (n = 0; n < 2; n++)
179 			regs[n] = __er32(hw, E1000_TARC(n));
180 		break;
181 	default:
182 		pr_info("%-15s %08x\n",
183 			reginfo->name, __er32(hw, reginfo->ofs));
184 		return;
185 	}
186 
187 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-1]");
188 	pr_info("%-15s %08x %08x\n", rname, regs[0], regs[1]);
189 }
190 
191 static void e1000e_dump_ps_pages(struct e1000_adapter *adapter,
192 				 struct e1000_buffer *bi)
193 {
194 	int i;
195 	struct e1000_ps_page *ps_page;
196 
197 	for (i = 0; i < adapter->rx_ps_pages; i++) {
198 		ps_page = &bi->ps_pages[i];
199 
200 		if (ps_page->page) {
201 			pr_info("packet dump for ps_page %d:\n", i);
202 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
203 				       16, 1, page_address(ps_page->page),
204 				       PAGE_SIZE, true);
205 		}
206 	}
207 }
208 
209 /**
210  * e1000e_dump - Print registers, Tx-ring and Rx-ring
211  * @adapter: board private structure
212  **/
213 static void e1000e_dump(struct e1000_adapter *adapter)
214 {
215 	struct net_device *netdev = adapter->netdev;
216 	struct e1000_hw *hw = &adapter->hw;
217 	struct e1000_reg_info *reginfo;
218 	struct e1000_ring *tx_ring = adapter->tx_ring;
219 	struct e1000_tx_desc *tx_desc;
220 	struct my_u0 {
221 		__le64 a;
222 		__le64 b;
223 	} *u0;
224 	struct e1000_buffer *buffer_info;
225 	struct e1000_ring *rx_ring = adapter->rx_ring;
226 	union e1000_rx_desc_packet_split *rx_desc_ps;
227 	union e1000_rx_desc_extended *rx_desc;
228 	struct my_u1 {
229 		__le64 a;
230 		__le64 b;
231 		__le64 c;
232 		__le64 d;
233 	} *u1;
234 	u32 staterr;
235 	int i = 0;
236 
237 	if (!netif_msg_hw(adapter))
238 		return;
239 
240 	/* Print netdevice Info */
241 	if (netdev) {
242 		dev_info(&adapter->pdev->dev, "Net device Info\n");
243 		pr_info("Device Name     state            trans_start\n");
244 		pr_info("%-15s %016lX %016lX\n", netdev->name,
245 			netdev->state, dev_trans_start(netdev));
246 	}
247 
248 	/* Print Registers */
249 	dev_info(&adapter->pdev->dev, "Register Dump\n");
250 	pr_info(" Register Name   Value\n");
251 	for (reginfo = (struct e1000_reg_info *)e1000_reg_info_tbl;
252 	     reginfo->name; reginfo++) {
253 		e1000_regdump(hw, reginfo);
254 	}
255 
256 	/* Print Tx Ring Summary */
257 	if (!netdev || !netif_running(netdev))
258 		return;
259 
260 	dev_info(&adapter->pdev->dev, "Tx Ring Summary\n");
261 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
262 	buffer_info = &tx_ring->buffer_info[tx_ring->next_to_clean];
263 	pr_info(" %5d %5X %5X %016llX %04X %3X %016llX\n",
264 		0, tx_ring->next_to_use, tx_ring->next_to_clean,
265 		(unsigned long long)buffer_info->dma,
266 		buffer_info->length,
267 		buffer_info->next_to_watch,
268 		(unsigned long long)buffer_info->time_stamp);
269 
270 	/* Print Tx Ring */
271 	if (!netif_msg_tx_done(adapter))
272 		goto rx_ring_summary;
273 
274 	dev_info(&adapter->pdev->dev, "Tx Ring Dump\n");
275 
276 	/* Transmit Descriptor Formats - DEXT[29] is 0 (Legacy) or 1 (Extended)
277 	 *
278 	 * Legacy Transmit Descriptor
279 	 *   +--------------------------------------------------------------+
280 	 * 0 |         Buffer Address [63:0] (Reserved on Write Back)       |
281 	 *   +--------------------------------------------------------------+
282 	 * 8 | Special  |    CSS     | Status |  CMD    |  CSO   |  Length  |
283 	 *   +--------------------------------------------------------------+
284 	 *   63       48 47        36 35    32 31     24 23    16 15        0
285 	 *
286 	 * Extended Context Descriptor (DTYP=0x0) for TSO or checksum offload
287 	 *   63      48 47    40 39       32 31             16 15    8 7      0
288 	 *   +----------------------------------------------------------------+
289 	 * 0 |  TUCSE  | TUCS0  |   TUCSS   |     IPCSE       | IPCS0 | IPCSS |
290 	 *   +----------------------------------------------------------------+
291 	 * 8 |   MSS   | HDRLEN | RSV | STA | TUCMD | DTYP |      PAYLEN      |
292 	 *   +----------------------------------------------------------------+
293 	 *   63      48 47    40 39 36 35 32 31   24 23  20 19                0
294 	 *
295 	 * Extended Data Descriptor (DTYP=0x1)
296 	 *   +----------------------------------------------------------------+
297 	 * 0 |                     Buffer Address [63:0]                      |
298 	 *   +----------------------------------------------------------------+
299 	 * 8 | VLAN tag |  POPTS  | Rsvd | Status | Command | DTYP |  DTALEN  |
300 	 *   +----------------------------------------------------------------+
301 	 *   63       48 47     40 39  36 35    32 31     24 23  20 19        0
302 	 */
303 	pr_info("Tl[desc]     [address 63:0  ] [SpeCssSCmCsLen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Legacy format\n");
304 	pr_info("Tc[desc]     [Ce CoCsIpceCoS] [MssHlRSCm0Plen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Context format\n");
305 	pr_info("Td[desc]     [address 63:0  ] [VlaPoRSCm1Dlen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Data format\n");
306 	for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
307 		const char *next_desc;
308 		tx_desc = E1000_TX_DESC(*tx_ring, i);
309 		buffer_info = &tx_ring->buffer_info[i];
310 		u0 = (struct my_u0 *)tx_desc;
311 		if (i == tx_ring->next_to_use && i == tx_ring->next_to_clean)
312 			next_desc = " NTC/U";
313 		else if (i == tx_ring->next_to_use)
314 			next_desc = " NTU";
315 		else if (i == tx_ring->next_to_clean)
316 			next_desc = " NTC";
317 		else
318 			next_desc = "";
319 		pr_info("T%c[0x%03X]    %016llX %016llX %016llX %04X  %3X %016llX %p%s\n",
320 			(!(le64_to_cpu(u0->b) & BIT(29)) ? 'l' :
321 			 ((le64_to_cpu(u0->b) & BIT(20)) ? 'd' : 'c')),
322 			i,
323 			(unsigned long long)le64_to_cpu(u0->a),
324 			(unsigned long long)le64_to_cpu(u0->b),
325 			(unsigned long long)buffer_info->dma,
326 			buffer_info->length, buffer_info->next_to_watch,
327 			(unsigned long long)buffer_info->time_stamp,
328 			buffer_info->skb, next_desc);
329 
330 		if (netif_msg_pktdata(adapter) && buffer_info->skb)
331 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
332 				       16, 1, buffer_info->skb->data,
333 				       buffer_info->skb->len, true);
334 	}
335 
336 	/* Print Rx Ring Summary */
337 rx_ring_summary:
338 	dev_info(&adapter->pdev->dev, "Rx Ring Summary\n");
339 	pr_info("Queue [NTU] [NTC]\n");
340 	pr_info(" %5d %5X %5X\n",
341 		0, rx_ring->next_to_use, rx_ring->next_to_clean);
342 
343 	/* Print Rx Ring */
344 	if (!netif_msg_rx_status(adapter))
345 		return;
346 
347 	dev_info(&adapter->pdev->dev, "Rx Ring Dump\n");
348 	switch (adapter->rx_ps_pages) {
349 	case 1:
350 	case 2:
351 	case 3:
352 		/* [Extended] Packet Split Receive Descriptor Format
353 		 *
354 		 *    +-----------------------------------------------------+
355 		 *  0 |                Buffer Address 0 [63:0]              |
356 		 *    +-----------------------------------------------------+
357 		 *  8 |                Buffer Address 1 [63:0]              |
358 		 *    +-----------------------------------------------------+
359 		 * 16 |                Buffer Address 2 [63:0]              |
360 		 *    +-----------------------------------------------------+
361 		 * 24 |                Buffer Address 3 [63:0]              |
362 		 *    +-----------------------------------------------------+
363 		 */
364 		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");
365 		/* [Extended] Receive Descriptor (Write-Back) Format
366 		 *
367 		 *   63       48 47    32 31     13 12    8 7    4 3        0
368 		 *   +------------------------------------------------------+
369 		 * 0 | Packet   | IP     |  Rsvd   | MRQ   | Rsvd | MRQ RSS |
370 		 *   | Checksum | Ident  |         | Queue |      |  Type   |
371 		 *   +------------------------------------------------------+
372 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
373 		 *   +------------------------------------------------------+
374 		 *   63       48 47    32 31            20 19               0
375 		 */
376 		pr_info("RWB[desc]      [ck ipid mrqhsh] [vl   l0 ee  es] [ l3  l2  l1 hs] [reserved      ] ---------------- [bi->skb] <-- Ext Rx Write-Back format\n");
377 		for (i = 0; i < rx_ring->count; i++) {
378 			const char *next_desc;
379 			buffer_info = &rx_ring->buffer_info[i];
380 			rx_desc_ps = E1000_RX_DESC_PS(*rx_ring, i);
381 			u1 = (struct my_u1 *)rx_desc_ps;
382 			staterr =
383 			    le32_to_cpu(rx_desc_ps->wb.middle.status_error);
384 
385 			if (i == rx_ring->next_to_use)
386 				next_desc = " NTU";
387 			else if (i == rx_ring->next_to_clean)
388 				next_desc = " NTC";
389 			else
390 				next_desc = "";
391 
392 			if (staterr & E1000_RXD_STAT_DD) {
393 				/* Descriptor Done */
394 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX ---------------- %p%s\n",
395 					"RWB", i,
396 					(unsigned long long)le64_to_cpu(u1->a),
397 					(unsigned long long)le64_to_cpu(u1->b),
398 					(unsigned long long)le64_to_cpu(u1->c),
399 					(unsigned long long)le64_to_cpu(u1->d),
400 					buffer_info->skb, next_desc);
401 			} else {
402 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX %016llX %p%s\n",
403 					"R  ", i,
404 					(unsigned long long)le64_to_cpu(u1->a),
405 					(unsigned long long)le64_to_cpu(u1->b),
406 					(unsigned long long)le64_to_cpu(u1->c),
407 					(unsigned long long)le64_to_cpu(u1->d),
408 					(unsigned long long)buffer_info->dma,
409 					buffer_info->skb, next_desc);
410 
411 				if (netif_msg_pktdata(adapter))
412 					e1000e_dump_ps_pages(adapter,
413 							     buffer_info);
414 			}
415 		}
416 		break;
417 	default:
418 	case 0:
419 		/* Extended Receive Descriptor (Read) Format
420 		 *
421 		 *   +-----------------------------------------------------+
422 		 * 0 |                Buffer Address [63:0]                |
423 		 *   +-----------------------------------------------------+
424 		 * 8 |                      Reserved                       |
425 		 *   +-----------------------------------------------------+
426 		 */
427 		pr_info("R  [desc]      [buf addr 63:0 ] [reserved 63:0 ] [bi->dma       ] [bi->skb] <-- Ext (Read) format\n");
428 		/* Extended Receive Descriptor (Write-Back) Format
429 		 *
430 		 *   63       48 47    32 31    24 23            4 3        0
431 		 *   +------------------------------------------------------+
432 		 *   |     RSS Hash      |        |               |         |
433 		 * 0 +-------------------+  Rsvd  |   Reserved    | MRQ RSS |
434 		 *   | Packet   | IP     |        |               |  Type   |
435 		 *   | Checksum | Ident  |        |               |         |
436 		 *   +------------------------------------------------------+
437 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
438 		 *   +------------------------------------------------------+
439 		 *   63       48 47    32 31            20 19               0
440 		 */
441 		pr_info("RWB[desc]      [cs ipid    mrq] [vt   ln xe  xs] [bi->skb] <-- Ext (Write-Back) format\n");
442 
443 		for (i = 0; i < rx_ring->count; i++) {
444 			const char *next_desc;
445 
446 			buffer_info = &rx_ring->buffer_info[i];
447 			rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
448 			u1 = (struct my_u1 *)rx_desc;
449 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
450 
451 			if (i == rx_ring->next_to_use)
452 				next_desc = " NTU";
453 			else if (i == rx_ring->next_to_clean)
454 				next_desc = " NTC";
455 			else
456 				next_desc = "";
457 
458 			if (staterr & E1000_RXD_STAT_DD) {
459 				/* Descriptor Done */
460 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %p%s\n",
461 					"RWB", i,
462 					(unsigned long long)le64_to_cpu(u1->a),
463 					(unsigned long long)le64_to_cpu(u1->b),
464 					buffer_info->skb, next_desc);
465 			} else {
466 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %p%s\n",
467 					"R  ", i,
468 					(unsigned long long)le64_to_cpu(u1->a),
469 					(unsigned long long)le64_to_cpu(u1->b),
470 					(unsigned long long)buffer_info->dma,
471 					buffer_info->skb, next_desc);
472 
473 				if (netif_msg_pktdata(adapter) &&
474 				    buffer_info->skb)
475 					print_hex_dump(KERN_INFO, "",
476 						       DUMP_PREFIX_ADDRESS, 16,
477 						       1,
478 						       buffer_info->skb->data,
479 						       adapter->rx_buffer_len,
480 						       true);
481 			}
482 		}
483 	}
484 }
485 
486 /**
487  * e1000_desc_unused - calculate if we have unused descriptors
488  **/
489 static int e1000_desc_unused(struct e1000_ring *ring)
490 {
491 	if (ring->next_to_clean > ring->next_to_use)
492 		return ring->next_to_clean - ring->next_to_use - 1;
493 
494 	return ring->count + ring->next_to_clean - ring->next_to_use - 1;
495 }
496 
497 /**
498  * e1000e_systim_to_hwtstamp - convert system time value to hw time stamp
499  * @adapter: board private structure
500  * @hwtstamps: time stamp structure to update
501  * @systim: unsigned 64bit system time value.
502  *
503  * Convert the system time value stored in the RX/TXSTMP registers into a
504  * hwtstamp which can be used by the upper level time stamping functions.
505  *
506  * The 'systim_lock' spinlock is used to protect the consistency of the
507  * system time value. This is needed because reading the 64 bit time
508  * value involves reading two 32 bit registers. The first read latches the
509  * value.
510  **/
511 static void e1000e_systim_to_hwtstamp(struct e1000_adapter *adapter,
512 				      struct skb_shared_hwtstamps *hwtstamps,
513 				      u64 systim)
514 {
515 	u64 ns;
516 	unsigned long flags;
517 
518 	spin_lock_irqsave(&adapter->systim_lock, flags);
519 	ns = timecounter_cyc2time(&adapter->tc, systim);
520 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
521 
522 	memset(hwtstamps, 0, sizeof(*hwtstamps));
523 	hwtstamps->hwtstamp = ns_to_ktime(ns);
524 }
525 
526 /**
527  * e1000e_rx_hwtstamp - utility function which checks for Rx time stamp
528  * @adapter: board private structure
529  * @status: descriptor extended error and status field
530  * @skb: particular skb to include time stamp
531  *
532  * If the time stamp is valid, convert it into the timecounter ns value
533  * and store that result into the shhwtstamps structure which is passed
534  * up the network stack.
535  **/
536 static void e1000e_rx_hwtstamp(struct e1000_adapter *adapter, u32 status,
537 			       struct sk_buff *skb)
538 {
539 	struct e1000_hw *hw = &adapter->hw;
540 	u64 rxstmp;
541 
542 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP) ||
543 	    !(status & E1000_RXDEXT_STATERR_TST) ||
544 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID))
545 		return;
546 
547 	/* The Rx time stamp registers contain the time stamp.  No other
548 	 * received packet will be time stamped until the Rx time stamp
549 	 * registers are read.  Because only one packet can be time stamped
550 	 * at a time, the register values must belong to this packet and
551 	 * therefore none of the other additional attributes need to be
552 	 * compared.
553 	 */
554 	rxstmp = (u64)er32(RXSTMPL);
555 	rxstmp |= (u64)er32(RXSTMPH) << 32;
556 	e1000e_systim_to_hwtstamp(adapter, skb_hwtstamps(skb), rxstmp);
557 
558 	adapter->flags2 &= ~FLAG2_CHECK_RX_HWTSTAMP;
559 }
560 
561 /**
562  * e1000_receive_skb - helper function to handle Rx indications
563  * @adapter: board private structure
564  * @staterr: descriptor extended error and status field as written by hardware
565  * @vlan: descriptor vlan field as written by hardware (no le/be conversion)
566  * @skb: pointer to sk_buff to be indicated to stack
567  **/
568 static void e1000_receive_skb(struct e1000_adapter *adapter,
569 			      struct net_device *netdev, struct sk_buff *skb,
570 			      u32 staterr, __le16 vlan)
571 {
572 	u16 tag = le16_to_cpu(vlan);
573 
574 	e1000e_rx_hwtstamp(adapter, staterr, skb);
575 
576 	skb->protocol = eth_type_trans(skb, netdev);
577 
578 	if (staterr & E1000_RXD_STAT_VP)
579 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), tag);
580 
581 	napi_gro_receive(&adapter->napi, skb);
582 }
583 
584 /**
585  * e1000_rx_checksum - Receive Checksum Offload
586  * @adapter: board private structure
587  * @status_err: receive descriptor status and error fields
588  * @csum: receive descriptor csum field
589  * @sk_buff: socket buffer with received data
590  **/
591 static void e1000_rx_checksum(struct e1000_adapter *adapter, u32 status_err,
592 			      struct sk_buff *skb)
593 {
594 	u16 status = (u16)status_err;
595 	u8 errors = (u8)(status_err >> 24);
596 
597 	skb_checksum_none_assert(skb);
598 
599 	/* Rx checksum disabled */
600 	if (!(adapter->netdev->features & NETIF_F_RXCSUM))
601 		return;
602 
603 	/* Ignore Checksum bit is set */
604 	if (status & E1000_RXD_STAT_IXSM)
605 		return;
606 
607 	/* TCP/UDP checksum error bit or IP checksum error bit is set */
608 	if (errors & (E1000_RXD_ERR_TCPE | E1000_RXD_ERR_IPE)) {
609 		/* let the stack verify checksum errors */
610 		adapter->hw_csum_err++;
611 		return;
612 	}
613 
614 	/* TCP/UDP Checksum has not been calculated */
615 	if (!(status & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)))
616 		return;
617 
618 	/* It must be a TCP or UDP packet with a valid checksum */
619 	skb->ip_summed = CHECKSUM_UNNECESSARY;
620 	adapter->hw_csum_good++;
621 }
622 
623 static void e1000e_update_rdt_wa(struct e1000_ring *rx_ring, unsigned int i)
624 {
625 	struct e1000_adapter *adapter = rx_ring->adapter;
626 	struct e1000_hw *hw = &adapter->hw;
627 	s32 ret_val = __ew32_prepare(hw);
628 
629 	writel(i, rx_ring->tail);
630 
631 	if (unlikely(!ret_val && (i != readl(rx_ring->tail)))) {
632 		u32 rctl = er32(RCTL);
633 
634 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
635 		e_err("ME firmware caused invalid RDT - resetting\n");
636 		schedule_work(&adapter->reset_task);
637 	}
638 }
639 
640 static void e1000e_update_tdt_wa(struct e1000_ring *tx_ring, unsigned int i)
641 {
642 	struct e1000_adapter *adapter = tx_ring->adapter;
643 	struct e1000_hw *hw = &adapter->hw;
644 	s32 ret_val = __ew32_prepare(hw);
645 
646 	writel(i, tx_ring->tail);
647 
648 	if (unlikely(!ret_val && (i != readl(tx_ring->tail)))) {
649 		u32 tctl = er32(TCTL);
650 
651 		ew32(TCTL, tctl & ~E1000_TCTL_EN);
652 		e_err("ME firmware caused invalid TDT - resetting\n");
653 		schedule_work(&adapter->reset_task);
654 	}
655 }
656 
657 /**
658  * e1000_alloc_rx_buffers - Replace used receive buffers
659  * @rx_ring: Rx descriptor ring
660  **/
661 static void e1000_alloc_rx_buffers(struct e1000_ring *rx_ring,
662 				   int cleaned_count, gfp_t gfp)
663 {
664 	struct e1000_adapter *adapter = rx_ring->adapter;
665 	struct net_device *netdev = adapter->netdev;
666 	struct pci_dev *pdev = adapter->pdev;
667 	union e1000_rx_desc_extended *rx_desc;
668 	struct e1000_buffer *buffer_info;
669 	struct sk_buff *skb;
670 	unsigned int i;
671 	unsigned int bufsz = adapter->rx_buffer_len;
672 
673 	i = rx_ring->next_to_use;
674 	buffer_info = &rx_ring->buffer_info[i];
675 
676 	while (cleaned_count--) {
677 		skb = buffer_info->skb;
678 		if (skb) {
679 			skb_trim(skb, 0);
680 			goto map_skb;
681 		}
682 
683 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
684 		if (!skb) {
685 			/* Better luck next round */
686 			adapter->alloc_rx_buff_failed++;
687 			break;
688 		}
689 
690 		buffer_info->skb = skb;
691 map_skb:
692 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
693 						  adapter->rx_buffer_len,
694 						  DMA_FROM_DEVICE);
695 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
696 			dev_err(&pdev->dev, "Rx DMA map failed\n");
697 			adapter->rx_dma_failed++;
698 			break;
699 		}
700 
701 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
702 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
703 
704 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
705 			/* Force memory writes to complete before letting h/w
706 			 * know there are new descriptors to fetch.  (Only
707 			 * applicable for weak-ordered memory model archs,
708 			 * such as IA-64).
709 			 */
710 			wmb();
711 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
712 				e1000e_update_rdt_wa(rx_ring, i);
713 			else
714 				writel(i, rx_ring->tail);
715 		}
716 		i++;
717 		if (i == rx_ring->count)
718 			i = 0;
719 		buffer_info = &rx_ring->buffer_info[i];
720 	}
721 
722 	rx_ring->next_to_use = i;
723 }
724 
725 /**
726  * e1000_alloc_rx_buffers_ps - Replace used receive buffers; packet split
727  * @rx_ring: Rx descriptor ring
728  **/
729 static void e1000_alloc_rx_buffers_ps(struct e1000_ring *rx_ring,
730 				      int cleaned_count, gfp_t gfp)
731 {
732 	struct e1000_adapter *adapter = rx_ring->adapter;
733 	struct net_device *netdev = adapter->netdev;
734 	struct pci_dev *pdev = adapter->pdev;
735 	union e1000_rx_desc_packet_split *rx_desc;
736 	struct e1000_buffer *buffer_info;
737 	struct e1000_ps_page *ps_page;
738 	struct sk_buff *skb;
739 	unsigned int i, j;
740 
741 	i = rx_ring->next_to_use;
742 	buffer_info = &rx_ring->buffer_info[i];
743 
744 	while (cleaned_count--) {
745 		rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
746 
747 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
748 			ps_page = &buffer_info->ps_pages[j];
749 			if (j >= adapter->rx_ps_pages) {
750 				/* all unused desc entries get hw null ptr */
751 				rx_desc->read.buffer_addr[j + 1] =
752 				    ~cpu_to_le64(0);
753 				continue;
754 			}
755 			if (!ps_page->page) {
756 				ps_page->page = alloc_page(gfp);
757 				if (!ps_page->page) {
758 					adapter->alloc_rx_buff_failed++;
759 					goto no_buffers;
760 				}
761 				ps_page->dma = dma_map_page(&pdev->dev,
762 							    ps_page->page,
763 							    0, PAGE_SIZE,
764 							    DMA_FROM_DEVICE);
765 				if (dma_mapping_error(&pdev->dev,
766 						      ps_page->dma)) {
767 					dev_err(&adapter->pdev->dev,
768 						"Rx DMA page map failed\n");
769 					adapter->rx_dma_failed++;
770 					goto no_buffers;
771 				}
772 			}
773 			/* Refresh the desc even if buffer_addrs
774 			 * didn't change because each write-back
775 			 * erases this info.
776 			 */
777 			rx_desc->read.buffer_addr[j + 1] =
778 			    cpu_to_le64(ps_page->dma);
779 		}
780 
781 		skb = __netdev_alloc_skb_ip_align(netdev, adapter->rx_ps_bsize0,
782 						  gfp);
783 
784 		if (!skb) {
785 			adapter->alloc_rx_buff_failed++;
786 			break;
787 		}
788 
789 		buffer_info->skb = skb;
790 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
791 						  adapter->rx_ps_bsize0,
792 						  DMA_FROM_DEVICE);
793 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
794 			dev_err(&pdev->dev, "Rx DMA map failed\n");
795 			adapter->rx_dma_failed++;
796 			/* cleanup skb */
797 			dev_kfree_skb_any(skb);
798 			buffer_info->skb = NULL;
799 			break;
800 		}
801 
802 		rx_desc->read.buffer_addr[0] = cpu_to_le64(buffer_info->dma);
803 
804 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
805 			/* Force memory writes to complete before letting h/w
806 			 * know there are new descriptors to fetch.  (Only
807 			 * applicable for weak-ordered memory model archs,
808 			 * such as IA-64).
809 			 */
810 			wmb();
811 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
812 				e1000e_update_rdt_wa(rx_ring, i << 1);
813 			else
814 				writel(i << 1, rx_ring->tail);
815 		}
816 
817 		i++;
818 		if (i == rx_ring->count)
819 			i = 0;
820 		buffer_info = &rx_ring->buffer_info[i];
821 	}
822 
823 no_buffers:
824 	rx_ring->next_to_use = i;
825 }
826 
827 /**
828  * e1000_alloc_jumbo_rx_buffers - Replace used jumbo receive buffers
829  * @rx_ring: Rx descriptor ring
830  * @cleaned_count: number of buffers to allocate this pass
831  **/
832 
833 static void e1000_alloc_jumbo_rx_buffers(struct e1000_ring *rx_ring,
834 					 int cleaned_count, gfp_t gfp)
835 {
836 	struct e1000_adapter *adapter = rx_ring->adapter;
837 	struct net_device *netdev = adapter->netdev;
838 	struct pci_dev *pdev = adapter->pdev;
839 	union e1000_rx_desc_extended *rx_desc;
840 	struct e1000_buffer *buffer_info;
841 	struct sk_buff *skb;
842 	unsigned int i;
843 	unsigned int bufsz = 256 - 16;	/* for skb_reserve */
844 
845 	i = rx_ring->next_to_use;
846 	buffer_info = &rx_ring->buffer_info[i];
847 
848 	while (cleaned_count--) {
849 		skb = buffer_info->skb;
850 		if (skb) {
851 			skb_trim(skb, 0);
852 			goto check_page;
853 		}
854 
855 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
856 		if (unlikely(!skb)) {
857 			/* Better luck next round */
858 			adapter->alloc_rx_buff_failed++;
859 			break;
860 		}
861 
862 		buffer_info->skb = skb;
863 check_page:
864 		/* allocate a new page if necessary */
865 		if (!buffer_info->page) {
866 			buffer_info->page = alloc_page(gfp);
867 			if (unlikely(!buffer_info->page)) {
868 				adapter->alloc_rx_buff_failed++;
869 				break;
870 			}
871 		}
872 
873 		if (!buffer_info->dma) {
874 			buffer_info->dma = dma_map_page(&pdev->dev,
875 							buffer_info->page, 0,
876 							PAGE_SIZE,
877 							DMA_FROM_DEVICE);
878 			if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
879 				adapter->alloc_rx_buff_failed++;
880 				break;
881 			}
882 		}
883 
884 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
885 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
886 
887 		if (unlikely(++i == rx_ring->count))
888 			i = 0;
889 		buffer_info = &rx_ring->buffer_info[i];
890 	}
891 
892 	if (likely(rx_ring->next_to_use != i)) {
893 		rx_ring->next_to_use = i;
894 		if (unlikely(i-- == 0))
895 			i = (rx_ring->count - 1);
896 
897 		/* Force memory writes to complete before letting h/w
898 		 * know there are new descriptors to fetch.  (Only
899 		 * applicable for weak-ordered memory model archs,
900 		 * such as IA-64).
901 		 */
902 		wmb();
903 		if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
904 			e1000e_update_rdt_wa(rx_ring, i);
905 		else
906 			writel(i, rx_ring->tail);
907 	}
908 }
909 
910 static inline void e1000_rx_hash(struct net_device *netdev, __le32 rss,
911 				 struct sk_buff *skb)
912 {
913 	if (netdev->features & NETIF_F_RXHASH)
914 		skb_set_hash(skb, le32_to_cpu(rss), PKT_HASH_TYPE_L3);
915 }
916 
917 /**
918  * e1000_clean_rx_irq - Send received data up the network stack
919  * @rx_ring: Rx descriptor ring
920  *
921  * the return value indicates whether actual cleaning was done, there
922  * is no guarantee that everything was cleaned
923  **/
924 static bool e1000_clean_rx_irq(struct e1000_ring *rx_ring, int *work_done,
925 			       int work_to_do)
926 {
927 	struct e1000_adapter *adapter = rx_ring->adapter;
928 	struct net_device *netdev = adapter->netdev;
929 	struct pci_dev *pdev = adapter->pdev;
930 	struct e1000_hw *hw = &adapter->hw;
931 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
932 	struct e1000_buffer *buffer_info, *next_buffer;
933 	u32 length, staterr;
934 	unsigned int i;
935 	int cleaned_count = 0;
936 	bool cleaned = false;
937 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
938 
939 	i = rx_ring->next_to_clean;
940 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
941 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
942 	buffer_info = &rx_ring->buffer_info[i];
943 
944 	while (staterr & E1000_RXD_STAT_DD) {
945 		struct sk_buff *skb;
946 
947 		if (*work_done >= work_to_do)
948 			break;
949 		(*work_done)++;
950 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
951 
952 		skb = buffer_info->skb;
953 		buffer_info->skb = NULL;
954 
955 		prefetch(skb->data - NET_IP_ALIGN);
956 
957 		i++;
958 		if (i == rx_ring->count)
959 			i = 0;
960 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
961 		prefetch(next_rxd);
962 
963 		next_buffer = &rx_ring->buffer_info[i];
964 
965 		cleaned = true;
966 		cleaned_count++;
967 		dma_unmap_single(&pdev->dev, buffer_info->dma,
968 				 adapter->rx_buffer_len, DMA_FROM_DEVICE);
969 		buffer_info->dma = 0;
970 
971 		length = le16_to_cpu(rx_desc->wb.upper.length);
972 
973 		/* !EOP means multiple descriptors were used to store a single
974 		 * packet, if that's the case we need to toss it.  In fact, we
975 		 * need to toss every packet with the EOP bit clear and the
976 		 * next frame that _does_ have the EOP bit set, as it is by
977 		 * definition only a frame fragment
978 		 */
979 		if (unlikely(!(staterr & E1000_RXD_STAT_EOP)))
980 			adapter->flags2 |= FLAG2_IS_DISCARDING;
981 
982 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
983 			/* All receives must fit into a single buffer */
984 			e_dbg("Receive packet consumed multiple buffers\n");
985 			/* recycle */
986 			buffer_info->skb = skb;
987 			if (staterr & E1000_RXD_STAT_EOP)
988 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
989 			goto next_desc;
990 		}
991 
992 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
993 			     !(netdev->features & NETIF_F_RXALL))) {
994 			/* recycle */
995 			buffer_info->skb = skb;
996 			goto next_desc;
997 		}
998 
999 		/* adjust length to remove Ethernet CRC */
1000 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1001 			/* If configured to store CRC, don't subtract FCS,
1002 			 * but keep the FCS bytes out of the total_rx_bytes
1003 			 * counter
1004 			 */
1005 			if (netdev->features & NETIF_F_RXFCS)
1006 				total_rx_bytes -= 4;
1007 			else
1008 				length -= 4;
1009 		}
1010 
1011 		total_rx_bytes += length;
1012 		total_rx_packets++;
1013 
1014 		/* code added for copybreak, this should improve
1015 		 * performance for small packets with large amounts
1016 		 * of reassembly being done in the stack
1017 		 */
1018 		if (length < copybreak) {
1019 			struct sk_buff *new_skb =
1020 				napi_alloc_skb(&adapter->napi, length);
1021 			if (new_skb) {
1022 				skb_copy_to_linear_data_offset(new_skb,
1023 							       -NET_IP_ALIGN,
1024 							       (skb->data -
1025 								NET_IP_ALIGN),
1026 							       (length +
1027 								NET_IP_ALIGN));
1028 				/* save the skb in buffer_info as good */
1029 				buffer_info->skb = skb;
1030 				skb = new_skb;
1031 			}
1032 			/* else just continue with the old one */
1033 		}
1034 		/* end copybreak code */
1035 		skb_put(skb, length);
1036 
1037 		/* Receive Checksum Offload */
1038 		e1000_rx_checksum(adapter, staterr, skb);
1039 
1040 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1041 
1042 		e1000_receive_skb(adapter, netdev, skb, staterr,
1043 				  rx_desc->wb.upper.vlan);
1044 
1045 next_desc:
1046 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1047 
1048 		/* return some buffers to hardware, one at a time is too slow */
1049 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1050 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1051 					      GFP_ATOMIC);
1052 			cleaned_count = 0;
1053 		}
1054 
1055 		/* use prefetched values */
1056 		rx_desc = next_rxd;
1057 		buffer_info = next_buffer;
1058 
1059 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1060 	}
1061 	rx_ring->next_to_clean = i;
1062 
1063 	cleaned_count = e1000_desc_unused(rx_ring);
1064 	if (cleaned_count)
1065 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1066 
1067 	adapter->total_rx_bytes += total_rx_bytes;
1068 	adapter->total_rx_packets += total_rx_packets;
1069 	return cleaned;
1070 }
1071 
1072 static void e1000_put_txbuf(struct e1000_ring *tx_ring,
1073 			    struct e1000_buffer *buffer_info)
1074 {
1075 	struct e1000_adapter *adapter = tx_ring->adapter;
1076 
1077 	if (buffer_info->dma) {
1078 		if (buffer_info->mapped_as_page)
1079 			dma_unmap_page(&adapter->pdev->dev, buffer_info->dma,
1080 				       buffer_info->length, DMA_TO_DEVICE);
1081 		else
1082 			dma_unmap_single(&adapter->pdev->dev, buffer_info->dma,
1083 					 buffer_info->length, DMA_TO_DEVICE);
1084 		buffer_info->dma = 0;
1085 	}
1086 	if (buffer_info->skb) {
1087 		dev_kfree_skb_any(buffer_info->skb);
1088 		buffer_info->skb = NULL;
1089 	}
1090 	buffer_info->time_stamp = 0;
1091 }
1092 
1093 static void e1000_print_hw_hang(struct work_struct *work)
1094 {
1095 	struct e1000_adapter *adapter = container_of(work,
1096 						     struct e1000_adapter,
1097 						     print_hang_task);
1098 	struct net_device *netdev = adapter->netdev;
1099 	struct e1000_ring *tx_ring = adapter->tx_ring;
1100 	unsigned int i = tx_ring->next_to_clean;
1101 	unsigned int eop = tx_ring->buffer_info[i].next_to_watch;
1102 	struct e1000_tx_desc *eop_desc = E1000_TX_DESC(*tx_ring, eop);
1103 	struct e1000_hw *hw = &adapter->hw;
1104 	u16 phy_status, phy_1000t_status, phy_ext_status;
1105 	u16 pci_status;
1106 
1107 	if (test_bit(__E1000_DOWN, &adapter->state))
1108 		return;
1109 
1110 	if (!adapter->tx_hang_recheck && (adapter->flags2 & FLAG2_DMA_BURST)) {
1111 		/* May be block on write-back, flush and detect again
1112 		 * flush pending descriptor writebacks to memory
1113 		 */
1114 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1115 		/* execute the writes immediately */
1116 		e1e_flush();
1117 		/* Due to rare timing issues, write to TIDV again to ensure
1118 		 * the write is successful
1119 		 */
1120 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1121 		/* execute the writes immediately */
1122 		e1e_flush();
1123 		adapter->tx_hang_recheck = true;
1124 		return;
1125 	}
1126 	adapter->tx_hang_recheck = false;
1127 
1128 	if (er32(TDH(0)) == er32(TDT(0))) {
1129 		e_dbg("false hang detected, ignoring\n");
1130 		return;
1131 	}
1132 
1133 	/* Real hang detected */
1134 	netif_stop_queue(netdev);
1135 
1136 	e1e_rphy(hw, MII_BMSR, &phy_status);
1137 	e1e_rphy(hw, MII_STAT1000, &phy_1000t_status);
1138 	e1e_rphy(hw, MII_ESTATUS, &phy_ext_status);
1139 
1140 	pci_read_config_word(adapter->pdev, PCI_STATUS, &pci_status);
1141 
1142 	/* detected Hardware unit hang */
1143 	e_err("Detected Hardware Unit Hang:\n"
1144 	      "  TDH                  <%x>\n"
1145 	      "  TDT                  <%x>\n"
1146 	      "  next_to_use          <%x>\n"
1147 	      "  next_to_clean        <%x>\n"
1148 	      "buffer_info[next_to_clean]:\n"
1149 	      "  time_stamp           <%lx>\n"
1150 	      "  next_to_watch        <%x>\n"
1151 	      "  jiffies              <%lx>\n"
1152 	      "  next_to_watch.status <%x>\n"
1153 	      "MAC Status             <%x>\n"
1154 	      "PHY Status             <%x>\n"
1155 	      "PHY 1000BASE-T Status  <%x>\n"
1156 	      "PHY Extended Status    <%x>\n"
1157 	      "PCI Status             <%x>\n",
1158 	      readl(tx_ring->head), readl(tx_ring->tail), tx_ring->next_to_use,
1159 	      tx_ring->next_to_clean, tx_ring->buffer_info[eop].time_stamp,
1160 	      eop, jiffies, eop_desc->upper.fields.status, er32(STATUS),
1161 	      phy_status, phy_1000t_status, phy_ext_status, pci_status);
1162 
1163 	e1000e_dump(adapter);
1164 
1165 	/* Suggest workaround for known h/w issue */
1166 	if ((hw->mac.type == e1000_pchlan) && (er32(CTRL) & E1000_CTRL_TFCE))
1167 		e_err("Try turning off Tx pause (flow control) via ethtool\n");
1168 }
1169 
1170 /**
1171  * e1000e_tx_hwtstamp_work - check for Tx time stamp
1172  * @work: pointer to work struct
1173  *
1174  * This work function polls the TSYNCTXCTL valid bit to determine when a
1175  * timestamp has been taken for the current stored skb.  The timestamp must
1176  * be for this skb because only one such packet is allowed in the queue.
1177  */
1178 static void e1000e_tx_hwtstamp_work(struct work_struct *work)
1179 {
1180 	struct e1000_adapter *adapter = container_of(work, struct e1000_adapter,
1181 						     tx_hwtstamp_work);
1182 	struct e1000_hw *hw = &adapter->hw;
1183 
1184 	if (er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID) {
1185 		struct skb_shared_hwtstamps shhwtstamps;
1186 		u64 txstmp;
1187 
1188 		txstmp = er32(TXSTMPL);
1189 		txstmp |= (u64)er32(TXSTMPH) << 32;
1190 
1191 		e1000e_systim_to_hwtstamp(adapter, &shhwtstamps, txstmp);
1192 
1193 		skb_tstamp_tx(adapter->tx_hwtstamp_skb, &shhwtstamps);
1194 		dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
1195 		adapter->tx_hwtstamp_skb = NULL;
1196 	} else if (time_after(jiffies, adapter->tx_hwtstamp_start
1197 			      + adapter->tx_timeout_factor * HZ)) {
1198 		dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
1199 		adapter->tx_hwtstamp_skb = NULL;
1200 		adapter->tx_hwtstamp_timeouts++;
1201 		e_warn("clearing Tx timestamp hang\n");
1202 	} else {
1203 		/* reschedule to check later */
1204 		schedule_work(&adapter->tx_hwtstamp_work);
1205 	}
1206 }
1207 
1208 /**
1209  * e1000_clean_tx_irq - Reclaim resources after transmit completes
1210  * @tx_ring: Tx descriptor ring
1211  *
1212  * the return value indicates whether actual cleaning was done, there
1213  * is no guarantee that everything was cleaned
1214  **/
1215 static bool e1000_clean_tx_irq(struct e1000_ring *tx_ring)
1216 {
1217 	struct e1000_adapter *adapter = tx_ring->adapter;
1218 	struct net_device *netdev = adapter->netdev;
1219 	struct e1000_hw *hw = &adapter->hw;
1220 	struct e1000_tx_desc *tx_desc, *eop_desc;
1221 	struct e1000_buffer *buffer_info;
1222 	unsigned int i, eop;
1223 	unsigned int count = 0;
1224 	unsigned int total_tx_bytes = 0, total_tx_packets = 0;
1225 	unsigned int bytes_compl = 0, pkts_compl = 0;
1226 
1227 	i = tx_ring->next_to_clean;
1228 	eop = tx_ring->buffer_info[i].next_to_watch;
1229 	eop_desc = E1000_TX_DESC(*tx_ring, eop);
1230 
1231 	while ((eop_desc->upper.data & cpu_to_le32(E1000_TXD_STAT_DD)) &&
1232 	       (count < tx_ring->count)) {
1233 		bool cleaned = false;
1234 
1235 		dma_rmb();		/* read buffer_info after eop_desc */
1236 		for (; !cleaned; count++) {
1237 			tx_desc = E1000_TX_DESC(*tx_ring, i);
1238 			buffer_info = &tx_ring->buffer_info[i];
1239 			cleaned = (i == eop);
1240 
1241 			if (cleaned) {
1242 				total_tx_packets += buffer_info->segs;
1243 				total_tx_bytes += buffer_info->bytecount;
1244 				if (buffer_info->skb) {
1245 					bytes_compl += buffer_info->skb->len;
1246 					pkts_compl++;
1247 				}
1248 			}
1249 
1250 			e1000_put_txbuf(tx_ring, buffer_info);
1251 			tx_desc->upper.data = 0;
1252 
1253 			i++;
1254 			if (i == tx_ring->count)
1255 				i = 0;
1256 		}
1257 
1258 		if (i == tx_ring->next_to_use)
1259 			break;
1260 		eop = tx_ring->buffer_info[i].next_to_watch;
1261 		eop_desc = E1000_TX_DESC(*tx_ring, eop);
1262 	}
1263 
1264 	tx_ring->next_to_clean = i;
1265 
1266 	netdev_completed_queue(netdev, pkts_compl, bytes_compl);
1267 
1268 #define TX_WAKE_THRESHOLD 32
1269 	if (count && netif_carrier_ok(netdev) &&
1270 	    e1000_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD) {
1271 		/* Make sure that anybody stopping the queue after this
1272 		 * sees the new next_to_clean.
1273 		 */
1274 		smp_mb();
1275 
1276 		if (netif_queue_stopped(netdev) &&
1277 		    !(test_bit(__E1000_DOWN, &adapter->state))) {
1278 			netif_wake_queue(netdev);
1279 			++adapter->restart_queue;
1280 		}
1281 	}
1282 
1283 	if (adapter->detect_tx_hung) {
1284 		/* Detect a transmit hang in hardware, this serializes the
1285 		 * check with the clearing of time_stamp and movement of i
1286 		 */
1287 		adapter->detect_tx_hung = false;
1288 		if (tx_ring->buffer_info[i].time_stamp &&
1289 		    time_after(jiffies, tx_ring->buffer_info[i].time_stamp
1290 			       + (adapter->tx_timeout_factor * HZ)) &&
1291 		    !(er32(STATUS) & E1000_STATUS_TXOFF))
1292 			schedule_work(&adapter->print_hang_task);
1293 		else
1294 			adapter->tx_hang_recheck = false;
1295 	}
1296 	adapter->total_tx_bytes += total_tx_bytes;
1297 	adapter->total_tx_packets += total_tx_packets;
1298 	return count < tx_ring->count;
1299 }
1300 
1301 /**
1302  * e1000_clean_rx_irq_ps - Send received data up the network stack; packet split
1303  * @rx_ring: Rx descriptor ring
1304  *
1305  * the return value indicates whether actual cleaning was done, there
1306  * is no guarantee that everything was cleaned
1307  **/
1308 static bool e1000_clean_rx_irq_ps(struct e1000_ring *rx_ring, int *work_done,
1309 				  int work_to_do)
1310 {
1311 	struct e1000_adapter *adapter = rx_ring->adapter;
1312 	struct e1000_hw *hw = &adapter->hw;
1313 	union e1000_rx_desc_packet_split *rx_desc, *next_rxd;
1314 	struct net_device *netdev = adapter->netdev;
1315 	struct pci_dev *pdev = adapter->pdev;
1316 	struct e1000_buffer *buffer_info, *next_buffer;
1317 	struct e1000_ps_page *ps_page;
1318 	struct sk_buff *skb;
1319 	unsigned int i, j;
1320 	u32 length, staterr;
1321 	int cleaned_count = 0;
1322 	bool cleaned = false;
1323 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1324 
1325 	i = rx_ring->next_to_clean;
1326 	rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
1327 	staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1328 	buffer_info = &rx_ring->buffer_info[i];
1329 
1330 	while (staterr & E1000_RXD_STAT_DD) {
1331 		if (*work_done >= work_to_do)
1332 			break;
1333 		(*work_done)++;
1334 		skb = buffer_info->skb;
1335 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1336 
1337 		/* in the packet split case this is header only */
1338 		prefetch(skb->data - NET_IP_ALIGN);
1339 
1340 		i++;
1341 		if (i == rx_ring->count)
1342 			i = 0;
1343 		next_rxd = E1000_RX_DESC_PS(*rx_ring, i);
1344 		prefetch(next_rxd);
1345 
1346 		next_buffer = &rx_ring->buffer_info[i];
1347 
1348 		cleaned = true;
1349 		cleaned_count++;
1350 		dma_unmap_single(&pdev->dev, buffer_info->dma,
1351 				 adapter->rx_ps_bsize0, DMA_FROM_DEVICE);
1352 		buffer_info->dma = 0;
1353 
1354 		/* see !EOP comment in other Rx routine */
1355 		if (!(staterr & E1000_RXD_STAT_EOP))
1356 			adapter->flags2 |= FLAG2_IS_DISCARDING;
1357 
1358 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
1359 			e_dbg("Packet Split buffers didn't pick up the full packet\n");
1360 			dev_kfree_skb_irq(skb);
1361 			if (staterr & E1000_RXD_STAT_EOP)
1362 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1363 			goto next_desc;
1364 		}
1365 
1366 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1367 			     !(netdev->features & NETIF_F_RXALL))) {
1368 			dev_kfree_skb_irq(skb);
1369 			goto next_desc;
1370 		}
1371 
1372 		length = le16_to_cpu(rx_desc->wb.middle.length0);
1373 
1374 		if (!length) {
1375 			e_dbg("Last part of the packet spanning multiple descriptors\n");
1376 			dev_kfree_skb_irq(skb);
1377 			goto next_desc;
1378 		}
1379 
1380 		/* Good Receive */
1381 		skb_put(skb, length);
1382 
1383 		{
1384 			/* this looks ugly, but it seems compiler issues make
1385 			 * it more efficient than reusing j
1386 			 */
1387 			int l1 = le16_to_cpu(rx_desc->wb.upper.length[0]);
1388 
1389 			/* page alloc/put takes too long and effects small
1390 			 * packet throughput, so unsplit small packets and
1391 			 * save the alloc/put only valid in softirq (napi)
1392 			 * context to call kmap_*
1393 			 */
1394 			if (l1 && (l1 <= copybreak) &&
1395 			    ((length + l1) <= adapter->rx_ps_bsize0)) {
1396 				u8 *vaddr;
1397 
1398 				ps_page = &buffer_info->ps_pages[0];
1399 
1400 				/* there is no documentation about how to call
1401 				 * kmap_atomic, so we can't hold the mapping
1402 				 * very long
1403 				 */
1404 				dma_sync_single_for_cpu(&pdev->dev,
1405 							ps_page->dma,
1406 							PAGE_SIZE,
1407 							DMA_FROM_DEVICE);
1408 				vaddr = kmap_atomic(ps_page->page);
1409 				memcpy(skb_tail_pointer(skb), vaddr, l1);
1410 				kunmap_atomic(vaddr);
1411 				dma_sync_single_for_device(&pdev->dev,
1412 							   ps_page->dma,
1413 							   PAGE_SIZE,
1414 							   DMA_FROM_DEVICE);
1415 
1416 				/* remove the CRC */
1417 				if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1418 					if (!(netdev->features & NETIF_F_RXFCS))
1419 						l1 -= 4;
1420 				}
1421 
1422 				skb_put(skb, l1);
1423 				goto copydone;
1424 			}	/* if */
1425 		}
1426 
1427 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1428 			length = le16_to_cpu(rx_desc->wb.upper.length[j]);
1429 			if (!length)
1430 				break;
1431 
1432 			ps_page = &buffer_info->ps_pages[j];
1433 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1434 				       DMA_FROM_DEVICE);
1435 			ps_page->dma = 0;
1436 			skb_fill_page_desc(skb, j, ps_page->page, 0, length);
1437 			ps_page->page = NULL;
1438 			skb->len += length;
1439 			skb->data_len += length;
1440 			skb->truesize += PAGE_SIZE;
1441 		}
1442 
1443 		/* strip the ethernet crc, problem is we're using pages now so
1444 		 * this whole operation can get a little cpu intensive
1445 		 */
1446 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1447 			if (!(netdev->features & NETIF_F_RXFCS))
1448 				pskb_trim(skb, skb->len - 4);
1449 		}
1450 
1451 copydone:
1452 		total_rx_bytes += skb->len;
1453 		total_rx_packets++;
1454 
1455 		e1000_rx_checksum(adapter, staterr, skb);
1456 
1457 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1458 
1459 		if (rx_desc->wb.upper.header_status &
1460 		    cpu_to_le16(E1000_RXDPS_HDRSTAT_HDRSP))
1461 			adapter->rx_hdr_split++;
1462 
1463 		e1000_receive_skb(adapter, netdev, skb, staterr,
1464 				  rx_desc->wb.middle.vlan);
1465 
1466 next_desc:
1467 		rx_desc->wb.middle.status_error &= cpu_to_le32(~0xFF);
1468 		buffer_info->skb = NULL;
1469 
1470 		/* return some buffers to hardware, one at a time is too slow */
1471 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1472 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1473 					      GFP_ATOMIC);
1474 			cleaned_count = 0;
1475 		}
1476 
1477 		/* use prefetched values */
1478 		rx_desc = next_rxd;
1479 		buffer_info = next_buffer;
1480 
1481 		staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1482 	}
1483 	rx_ring->next_to_clean = i;
1484 
1485 	cleaned_count = e1000_desc_unused(rx_ring);
1486 	if (cleaned_count)
1487 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1488 
1489 	adapter->total_rx_bytes += total_rx_bytes;
1490 	adapter->total_rx_packets += total_rx_packets;
1491 	return cleaned;
1492 }
1493 
1494 /**
1495  * e1000_consume_page - helper function
1496  **/
1497 static void e1000_consume_page(struct e1000_buffer *bi, struct sk_buff *skb,
1498 			       u16 length)
1499 {
1500 	bi->page = NULL;
1501 	skb->len += length;
1502 	skb->data_len += length;
1503 	skb->truesize += PAGE_SIZE;
1504 }
1505 
1506 /**
1507  * e1000_clean_jumbo_rx_irq - Send received data up the network stack; legacy
1508  * @adapter: board private structure
1509  *
1510  * the return value indicates whether actual cleaning was done, there
1511  * is no guarantee that everything was cleaned
1512  **/
1513 static bool e1000_clean_jumbo_rx_irq(struct e1000_ring *rx_ring, int *work_done,
1514 				     int work_to_do)
1515 {
1516 	struct e1000_adapter *adapter = rx_ring->adapter;
1517 	struct net_device *netdev = adapter->netdev;
1518 	struct pci_dev *pdev = adapter->pdev;
1519 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
1520 	struct e1000_buffer *buffer_info, *next_buffer;
1521 	u32 length, staterr;
1522 	unsigned int i;
1523 	int cleaned_count = 0;
1524 	bool cleaned = false;
1525 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1526 	struct skb_shared_info *shinfo;
1527 
1528 	i = rx_ring->next_to_clean;
1529 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
1530 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1531 	buffer_info = &rx_ring->buffer_info[i];
1532 
1533 	while (staterr & E1000_RXD_STAT_DD) {
1534 		struct sk_buff *skb;
1535 
1536 		if (*work_done >= work_to_do)
1537 			break;
1538 		(*work_done)++;
1539 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1540 
1541 		skb = buffer_info->skb;
1542 		buffer_info->skb = NULL;
1543 
1544 		++i;
1545 		if (i == rx_ring->count)
1546 			i = 0;
1547 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
1548 		prefetch(next_rxd);
1549 
1550 		next_buffer = &rx_ring->buffer_info[i];
1551 
1552 		cleaned = true;
1553 		cleaned_count++;
1554 		dma_unmap_page(&pdev->dev, buffer_info->dma, PAGE_SIZE,
1555 			       DMA_FROM_DEVICE);
1556 		buffer_info->dma = 0;
1557 
1558 		length = le16_to_cpu(rx_desc->wb.upper.length);
1559 
1560 		/* errors is only valid for DD + EOP descriptors */
1561 		if (unlikely((staterr & E1000_RXD_STAT_EOP) &&
1562 			     ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1563 			      !(netdev->features & NETIF_F_RXALL)))) {
1564 			/* recycle both page and skb */
1565 			buffer_info->skb = skb;
1566 			/* an error means any chain goes out the window too */
1567 			if (rx_ring->rx_skb_top)
1568 				dev_kfree_skb_irq(rx_ring->rx_skb_top);
1569 			rx_ring->rx_skb_top = NULL;
1570 			goto next_desc;
1571 		}
1572 #define rxtop (rx_ring->rx_skb_top)
1573 		if (!(staterr & E1000_RXD_STAT_EOP)) {
1574 			/* this descriptor is only the beginning (or middle) */
1575 			if (!rxtop) {
1576 				/* this is the beginning of a chain */
1577 				rxtop = skb;
1578 				skb_fill_page_desc(rxtop, 0, buffer_info->page,
1579 						   0, length);
1580 			} else {
1581 				/* this is the middle of a chain */
1582 				shinfo = skb_shinfo(rxtop);
1583 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1584 						   buffer_info->page, 0,
1585 						   length);
1586 				/* re-use the skb, only consumed the page */
1587 				buffer_info->skb = skb;
1588 			}
1589 			e1000_consume_page(buffer_info, rxtop, length);
1590 			goto next_desc;
1591 		} else {
1592 			if (rxtop) {
1593 				/* end of the chain */
1594 				shinfo = skb_shinfo(rxtop);
1595 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1596 						   buffer_info->page, 0,
1597 						   length);
1598 				/* re-use the current skb, we only consumed the
1599 				 * page
1600 				 */
1601 				buffer_info->skb = skb;
1602 				skb = rxtop;
1603 				rxtop = NULL;
1604 				e1000_consume_page(buffer_info, skb, length);
1605 			} else {
1606 				/* no chain, got EOP, this buf is the packet
1607 				 * copybreak to save the put_page/alloc_page
1608 				 */
1609 				if (length <= copybreak &&
1610 				    skb_tailroom(skb) >= length) {
1611 					u8 *vaddr;
1612 					vaddr = kmap_atomic(buffer_info->page);
1613 					memcpy(skb_tail_pointer(skb), vaddr,
1614 					       length);
1615 					kunmap_atomic(vaddr);
1616 					/* re-use the page, so don't erase
1617 					 * buffer_info->page
1618 					 */
1619 					skb_put(skb, length);
1620 				} else {
1621 					skb_fill_page_desc(skb, 0,
1622 							   buffer_info->page, 0,
1623 							   length);
1624 					e1000_consume_page(buffer_info, skb,
1625 							   length);
1626 				}
1627 			}
1628 		}
1629 
1630 		/* Receive Checksum Offload */
1631 		e1000_rx_checksum(adapter, staterr, skb);
1632 
1633 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1634 
1635 		/* probably a little skewed due to removing CRC */
1636 		total_rx_bytes += skb->len;
1637 		total_rx_packets++;
1638 
1639 		/* eth type trans needs skb->data to point to something */
1640 		if (!pskb_may_pull(skb, ETH_HLEN)) {
1641 			e_err("pskb_may_pull failed.\n");
1642 			dev_kfree_skb_irq(skb);
1643 			goto next_desc;
1644 		}
1645 
1646 		e1000_receive_skb(adapter, netdev, skb, staterr,
1647 				  rx_desc->wb.upper.vlan);
1648 
1649 next_desc:
1650 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1651 
1652 		/* return some buffers to hardware, one at a time is too slow */
1653 		if (unlikely(cleaned_count >= E1000_RX_BUFFER_WRITE)) {
1654 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1655 					      GFP_ATOMIC);
1656 			cleaned_count = 0;
1657 		}
1658 
1659 		/* use prefetched values */
1660 		rx_desc = next_rxd;
1661 		buffer_info = next_buffer;
1662 
1663 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1664 	}
1665 	rx_ring->next_to_clean = i;
1666 
1667 	cleaned_count = e1000_desc_unused(rx_ring);
1668 	if (cleaned_count)
1669 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1670 
1671 	adapter->total_rx_bytes += total_rx_bytes;
1672 	adapter->total_rx_packets += total_rx_packets;
1673 	return cleaned;
1674 }
1675 
1676 /**
1677  * e1000_clean_rx_ring - Free Rx Buffers per Queue
1678  * @rx_ring: Rx descriptor ring
1679  **/
1680 static void e1000_clean_rx_ring(struct e1000_ring *rx_ring)
1681 {
1682 	struct e1000_adapter *adapter = rx_ring->adapter;
1683 	struct e1000_buffer *buffer_info;
1684 	struct e1000_ps_page *ps_page;
1685 	struct pci_dev *pdev = adapter->pdev;
1686 	unsigned int i, j;
1687 
1688 	/* Free all the Rx ring sk_buffs */
1689 	for (i = 0; i < rx_ring->count; i++) {
1690 		buffer_info = &rx_ring->buffer_info[i];
1691 		if (buffer_info->dma) {
1692 			if (adapter->clean_rx == e1000_clean_rx_irq)
1693 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1694 						 adapter->rx_buffer_len,
1695 						 DMA_FROM_DEVICE);
1696 			else if (adapter->clean_rx == e1000_clean_jumbo_rx_irq)
1697 				dma_unmap_page(&pdev->dev, buffer_info->dma,
1698 					       PAGE_SIZE, DMA_FROM_DEVICE);
1699 			else if (adapter->clean_rx == e1000_clean_rx_irq_ps)
1700 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1701 						 adapter->rx_ps_bsize0,
1702 						 DMA_FROM_DEVICE);
1703 			buffer_info->dma = 0;
1704 		}
1705 
1706 		if (buffer_info->page) {
1707 			put_page(buffer_info->page);
1708 			buffer_info->page = NULL;
1709 		}
1710 
1711 		if (buffer_info->skb) {
1712 			dev_kfree_skb(buffer_info->skb);
1713 			buffer_info->skb = NULL;
1714 		}
1715 
1716 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1717 			ps_page = &buffer_info->ps_pages[j];
1718 			if (!ps_page->page)
1719 				break;
1720 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1721 				       DMA_FROM_DEVICE);
1722 			ps_page->dma = 0;
1723 			put_page(ps_page->page);
1724 			ps_page->page = NULL;
1725 		}
1726 	}
1727 
1728 	/* there also may be some cached data from a chained receive */
1729 	if (rx_ring->rx_skb_top) {
1730 		dev_kfree_skb(rx_ring->rx_skb_top);
1731 		rx_ring->rx_skb_top = NULL;
1732 	}
1733 
1734 	/* Zero out the descriptor ring */
1735 	memset(rx_ring->desc, 0, rx_ring->size);
1736 
1737 	rx_ring->next_to_clean = 0;
1738 	rx_ring->next_to_use = 0;
1739 	adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1740 }
1741 
1742 static void e1000e_downshift_workaround(struct work_struct *work)
1743 {
1744 	struct e1000_adapter *adapter = container_of(work,
1745 						     struct e1000_adapter,
1746 						     downshift_task);
1747 
1748 	if (test_bit(__E1000_DOWN, &adapter->state))
1749 		return;
1750 
1751 	e1000e_gig_downshift_workaround_ich8lan(&adapter->hw);
1752 }
1753 
1754 /**
1755  * e1000_intr_msi - Interrupt Handler
1756  * @irq: interrupt number
1757  * @data: pointer to a network interface device structure
1758  **/
1759 static irqreturn_t e1000_intr_msi(int __always_unused irq, void *data)
1760 {
1761 	struct net_device *netdev = data;
1762 	struct e1000_adapter *adapter = netdev_priv(netdev);
1763 	struct e1000_hw *hw = &adapter->hw;
1764 	u32 icr = er32(ICR);
1765 
1766 	/* read ICR disables interrupts using IAM */
1767 	if (icr & E1000_ICR_LSC) {
1768 		hw->mac.get_link_status = true;
1769 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1770 		 * disconnect (LSC) before accessing any PHY registers
1771 		 */
1772 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1773 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1774 			schedule_work(&adapter->downshift_task);
1775 
1776 		/* 80003ES2LAN workaround-- For packet buffer work-around on
1777 		 * link down event; disable receives here in the ISR and reset
1778 		 * adapter in watchdog
1779 		 */
1780 		if (netif_carrier_ok(netdev) &&
1781 		    adapter->flags & FLAG_RX_NEEDS_RESTART) {
1782 			/* disable receives */
1783 			u32 rctl = er32(RCTL);
1784 
1785 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1786 			adapter->flags |= FLAG_RESTART_NOW;
1787 		}
1788 		/* guard against interrupt when we're going down */
1789 		if (!test_bit(__E1000_DOWN, &adapter->state))
1790 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1791 	}
1792 
1793 	/* Reset on uncorrectable ECC error */
1794 	if ((icr & E1000_ICR_ECCER) && ((hw->mac.type == e1000_pch_lpt) ||
1795 					(hw->mac.type == e1000_pch_spt))) {
1796 		u32 pbeccsts = er32(PBECCSTS);
1797 
1798 		adapter->corr_errors +=
1799 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1800 		adapter->uncorr_errors +=
1801 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1802 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1803 
1804 		/* Do the reset outside of interrupt context */
1805 		schedule_work(&adapter->reset_task);
1806 
1807 		/* return immediately since reset is imminent */
1808 		return IRQ_HANDLED;
1809 	}
1810 
1811 	if (napi_schedule_prep(&adapter->napi)) {
1812 		adapter->total_tx_bytes = 0;
1813 		adapter->total_tx_packets = 0;
1814 		adapter->total_rx_bytes = 0;
1815 		adapter->total_rx_packets = 0;
1816 		__napi_schedule(&adapter->napi);
1817 	}
1818 
1819 	return IRQ_HANDLED;
1820 }
1821 
1822 /**
1823  * e1000_intr - Interrupt Handler
1824  * @irq: interrupt number
1825  * @data: pointer to a network interface device structure
1826  **/
1827 static irqreturn_t e1000_intr(int __always_unused irq, void *data)
1828 {
1829 	struct net_device *netdev = data;
1830 	struct e1000_adapter *adapter = netdev_priv(netdev);
1831 	struct e1000_hw *hw = &adapter->hw;
1832 	u32 rctl, icr = er32(ICR);
1833 
1834 	if (!icr || test_bit(__E1000_DOWN, &adapter->state))
1835 		return IRQ_NONE;	/* Not our interrupt */
1836 
1837 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
1838 	 * not set, then the adapter didn't send an interrupt
1839 	 */
1840 	if (!(icr & E1000_ICR_INT_ASSERTED))
1841 		return IRQ_NONE;
1842 
1843 	/* Interrupt Auto-Mask...upon reading ICR,
1844 	 * interrupts are masked.  No need for the
1845 	 * IMC write
1846 	 */
1847 
1848 	if (icr & E1000_ICR_LSC) {
1849 		hw->mac.get_link_status = true;
1850 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1851 		 * disconnect (LSC) before accessing any PHY registers
1852 		 */
1853 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1854 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1855 			schedule_work(&adapter->downshift_task);
1856 
1857 		/* 80003ES2LAN workaround--
1858 		 * For packet buffer work-around on link down event;
1859 		 * disable receives here in the ISR and
1860 		 * reset adapter in watchdog
1861 		 */
1862 		if (netif_carrier_ok(netdev) &&
1863 		    (adapter->flags & FLAG_RX_NEEDS_RESTART)) {
1864 			/* disable receives */
1865 			rctl = er32(RCTL);
1866 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1867 			adapter->flags |= FLAG_RESTART_NOW;
1868 		}
1869 		/* guard against interrupt when we're going down */
1870 		if (!test_bit(__E1000_DOWN, &adapter->state))
1871 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1872 	}
1873 
1874 	/* Reset on uncorrectable ECC error */
1875 	if ((icr & E1000_ICR_ECCER) && ((hw->mac.type == e1000_pch_lpt) ||
1876 					(hw->mac.type == e1000_pch_spt))) {
1877 		u32 pbeccsts = er32(PBECCSTS);
1878 
1879 		adapter->corr_errors +=
1880 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1881 		adapter->uncorr_errors +=
1882 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1883 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1884 
1885 		/* Do the reset outside of interrupt context */
1886 		schedule_work(&adapter->reset_task);
1887 
1888 		/* return immediately since reset is imminent */
1889 		return IRQ_HANDLED;
1890 	}
1891 
1892 	if (napi_schedule_prep(&adapter->napi)) {
1893 		adapter->total_tx_bytes = 0;
1894 		adapter->total_tx_packets = 0;
1895 		adapter->total_rx_bytes = 0;
1896 		adapter->total_rx_packets = 0;
1897 		__napi_schedule(&adapter->napi);
1898 	}
1899 
1900 	return IRQ_HANDLED;
1901 }
1902 
1903 static irqreturn_t e1000_msix_other(int __always_unused irq, void *data)
1904 {
1905 	struct net_device *netdev = data;
1906 	struct e1000_adapter *adapter = netdev_priv(netdev);
1907 	struct e1000_hw *hw = &adapter->hw;
1908 
1909 	hw->mac.get_link_status = true;
1910 
1911 	/* guard against interrupt when we're going down */
1912 	if (!test_bit(__E1000_DOWN, &adapter->state)) {
1913 		mod_timer(&adapter->watchdog_timer, jiffies + 1);
1914 		ew32(IMS, E1000_IMS_OTHER);
1915 	}
1916 
1917 	return IRQ_HANDLED;
1918 }
1919 
1920 static irqreturn_t e1000_intr_msix_tx(int __always_unused irq, void *data)
1921 {
1922 	struct net_device *netdev = data;
1923 	struct e1000_adapter *adapter = netdev_priv(netdev);
1924 	struct e1000_hw *hw = &adapter->hw;
1925 	struct e1000_ring *tx_ring = adapter->tx_ring;
1926 
1927 	adapter->total_tx_bytes = 0;
1928 	adapter->total_tx_packets = 0;
1929 
1930 	if (!e1000_clean_tx_irq(tx_ring))
1931 		/* Ring was not completely cleaned, so fire another interrupt */
1932 		ew32(ICS, tx_ring->ims_val);
1933 
1934 	if (!test_bit(__E1000_DOWN, &adapter->state))
1935 		ew32(IMS, adapter->tx_ring->ims_val);
1936 
1937 	return IRQ_HANDLED;
1938 }
1939 
1940 static irqreturn_t e1000_intr_msix_rx(int __always_unused irq, void *data)
1941 {
1942 	struct net_device *netdev = data;
1943 	struct e1000_adapter *adapter = netdev_priv(netdev);
1944 	struct e1000_ring *rx_ring = adapter->rx_ring;
1945 
1946 	/* Write the ITR value calculated at the end of the
1947 	 * previous interrupt.
1948 	 */
1949 	if (rx_ring->set_itr) {
1950 		u32 itr = rx_ring->itr_val ?
1951 			  1000000000 / (rx_ring->itr_val * 256) : 0;
1952 
1953 		writel(itr, rx_ring->itr_register);
1954 		rx_ring->set_itr = 0;
1955 	}
1956 
1957 	if (napi_schedule_prep(&adapter->napi)) {
1958 		adapter->total_rx_bytes = 0;
1959 		adapter->total_rx_packets = 0;
1960 		__napi_schedule(&adapter->napi);
1961 	}
1962 	return IRQ_HANDLED;
1963 }
1964 
1965 /**
1966  * e1000_configure_msix - Configure MSI-X hardware
1967  *
1968  * e1000_configure_msix sets up the hardware to properly
1969  * generate MSI-X interrupts.
1970  **/
1971 static void e1000_configure_msix(struct e1000_adapter *adapter)
1972 {
1973 	struct e1000_hw *hw = &adapter->hw;
1974 	struct e1000_ring *rx_ring = adapter->rx_ring;
1975 	struct e1000_ring *tx_ring = adapter->tx_ring;
1976 	int vector = 0;
1977 	u32 ctrl_ext, ivar = 0;
1978 
1979 	adapter->eiac_mask = 0;
1980 
1981 	/* Workaround issue with spurious interrupts on 82574 in MSI-X mode */
1982 	if (hw->mac.type == e1000_82574) {
1983 		u32 rfctl = er32(RFCTL);
1984 
1985 		rfctl |= E1000_RFCTL_ACK_DIS;
1986 		ew32(RFCTL, rfctl);
1987 	}
1988 
1989 	/* Configure Rx vector */
1990 	rx_ring->ims_val = E1000_IMS_RXQ0;
1991 	adapter->eiac_mask |= rx_ring->ims_val;
1992 	if (rx_ring->itr_val)
1993 		writel(1000000000 / (rx_ring->itr_val * 256),
1994 		       rx_ring->itr_register);
1995 	else
1996 		writel(1, rx_ring->itr_register);
1997 	ivar = E1000_IVAR_INT_ALLOC_VALID | vector;
1998 
1999 	/* Configure Tx vector */
2000 	tx_ring->ims_val = E1000_IMS_TXQ0;
2001 	vector++;
2002 	if (tx_ring->itr_val)
2003 		writel(1000000000 / (tx_ring->itr_val * 256),
2004 		       tx_ring->itr_register);
2005 	else
2006 		writel(1, tx_ring->itr_register);
2007 	adapter->eiac_mask |= tx_ring->ims_val;
2008 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 8);
2009 
2010 	/* set vector for Other Causes, e.g. link changes */
2011 	vector++;
2012 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 16);
2013 	if (rx_ring->itr_val)
2014 		writel(1000000000 / (rx_ring->itr_val * 256),
2015 		       hw->hw_addr + E1000_EITR_82574(vector));
2016 	else
2017 		writel(1, hw->hw_addr + E1000_EITR_82574(vector));
2018 	adapter->eiac_mask |= E1000_IMS_OTHER;
2019 
2020 	/* Cause Tx interrupts on every write back */
2021 	ivar |= BIT(31);
2022 
2023 	ew32(IVAR, ivar);
2024 
2025 	/* enable MSI-X PBA support */
2026 	ctrl_ext = er32(CTRL_EXT) & ~E1000_CTRL_EXT_IAME;
2027 	ctrl_ext |= E1000_CTRL_EXT_PBA_CLR | E1000_CTRL_EXT_EIAME;
2028 	ew32(CTRL_EXT, ctrl_ext);
2029 	e1e_flush();
2030 }
2031 
2032 void e1000e_reset_interrupt_capability(struct e1000_adapter *adapter)
2033 {
2034 	if (adapter->msix_entries) {
2035 		pci_disable_msix(adapter->pdev);
2036 		kfree(adapter->msix_entries);
2037 		adapter->msix_entries = NULL;
2038 	} else if (adapter->flags & FLAG_MSI_ENABLED) {
2039 		pci_disable_msi(adapter->pdev);
2040 		adapter->flags &= ~FLAG_MSI_ENABLED;
2041 	}
2042 }
2043 
2044 /**
2045  * e1000e_set_interrupt_capability - set MSI or MSI-X if supported
2046  *
2047  * Attempt to configure interrupts using the best available
2048  * capabilities of the hardware and kernel.
2049  **/
2050 void e1000e_set_interrupt_capability(struct e1000_adapter *adapter)
2051 {
2052 	int err;
2053 	int i;
2054 
2055 	switch (adapter->int_mode) {
2056 	case E1000E_INT_MODE_MSIX:
2057 		if (adapter->flags & FLAG_HAS_MSIX) {
2058 			adapter->num_vectors = 3; /* RxQ0, TxQ0 and other */
2059 			adapter->msix_entries = kcalloc(adapter->num_vectors,
2060 							sizeof(struct
2061 							       msix_entry),
2062 							GFP_KERNEL);
2063 			if (adapter->msix_entries) {
2064 				struct e1000_adapter *a = adapter;
2065 
2066 				for (i = 0; i < adapter->num_vectors; i++)
2067 					adapter->msix_entries[i].entry = i;
2068 
2069 				err = pci_enable_msix_range(a->pdev,
2070 							    a->msix_entries,
2071 							    a->num_vectors,
2072 							    a->num_vectors);
2073 				if (err > 0)
2074 					return;
2075 			}
2076 			/* MSI-X failed, so fall through and try MSI */
2077 			e_err("Failed to initialize MSI-X interrupts.  Falling back to MSI interrupts.\n");
2078 			e1000e_reset_interrupt_capability(adapter);
2079 		}
2080 		adapter->int_mode = E1000E_INT_MODE_MSI;
2081 		/* Fall through */
2082 	case E1000E_INT_MODE_MSI:
2083 		if (!pci_enable_msi(adapter->pdev)) {
2084 			adapter->flags |= FLAG_MSI_ENABLED;
2085 		} else {
2086 			adapter->int_mode = E1000E_INT_MODE_LEGACY;
2087 			e_err("Failed to initialize MSI interrupts.  Falling back to legacy interrupts.\n");
2088 		}
2089 		/* Fall through */
2090 	case E1000E_INT_MODE_LEGACY:
2091 		/* Don't do anything; this is the system default */
2092 		break;
2093 	}
2094 
2095 	/* store the number of vectors being used */
2096 	adapter->num_vectors = 1;
2097 }
2098 
2099 /**
2100  * e1000_request_msix - Initialize MSI-X interrupts
2101  *
2102  * e1000_request_msix allocates MSI-X vectors and requests interrupts from the
2103  * kernel.
2104  **/
2105 static int e1000_request_msix(struct e1000_adapter *adapter)
2106 {
2107 	struct net_device *netdev = adapter->netdev;
2108 	int err = 0, vector = 0;
2109 
2110 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2111 		snprintf(adapter->rx_ring->name,
2112 			 sizeof(adapter->rx_ring->name) - 1,
2113 			 "%s-rx-0", netdev->name);
2114 	else
2115 		memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
2116 	err = request_irq(adapter->msix_entries[vector].vector,
2117 			  e1000_intr_msix_rx, 0, adapter->rx_ring->name,
2118 			  netdev);
2119 	if (err)
2120 		return err;
2121 	adapter->rx_ring->itr_register = adapter->hw.hw_addr +
2122 	    E1000_EITR_82574(vector);
2123 	adapter->rx_ring->itr_val = adapter->itr;
2124 	vector++;
2125 
2126 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2127 		snprintf(adapter->tx_ring->name,
2128 			 sizeof(adapter->tx_ring->name) - 1,
2129 			 "%s-tx-0", netdev->name);
2130 	else
2131 		memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
2132 	err = request_irq(adapter->msix_entries[vector].vector,
2133 			  e1000_intr_msix_tx, 0, adapter->tx_ring->name,
2134 			  netdev);
2135 	if (err)
2136 		return err;
2137 	adapter->tx_ring->itr_register = adapter->hw.hw_addr +
2138 	    E1000_EITR_82574(vector);
2139 	adapter->tx_ring->itr_val = adapter->itr;
2140 	vector++;
2141 
2142 	err = request_irq(adapter->msix_entries[vector].vector,
2143 			  e1000_msix_other, 0, netdev->name, netdev);
2144 	if (err)
2145 		return err;
2146 
2147 	e1000_configure_msix(adapter);
2148 
2149 	return 0;
2150 }
2151 
2152 /**
2153  * e1000_request_irq - initialize interrupts
2154  *
2155  * Attempts to configure interrupts using the best available
2156  * capabilities of the hardware and kernel.
2157  **/
2158 static int e1000_request_irq(struct e1000_adapter *adapter)
2159 {
2160 	struct net_device *netdev = adapter->netdev;
2161 	int err;
2162 
2163 	if (adapter->msix_entries) {
2164 		err = e1000_request_msix(adapter);
2165 		if (!err)
2166 			return err;
2167 		/* fall back to MSI */
2168 		e1000e_reset_interrupt_capability(adapter);
2169 		adapter->int_mode = E1000E_INT_MODE_MSI;
2170 		e1000e_set_interrupt_capability(adapter);
2171 	}
2172 	if (adapter->flags & FLAG_MSI_ENABLED) {
2173 		err = request_irq(adapter->pdev->irq, e1000_intr_msi, 0,
2174 				  netdev->name, netdev);
2175 		if (!err)
2176 			return err;
2177 
2178 		/* fall back to legacy interrupt */
2179 		e1000e_reset_interrupt_capability(adapter);
2180 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
2181 	}
2182 
2183 	err = request_irq(adapter->pdev->irq, e1000_intr, IRQF_SHARED,
2184 			  netdev->name, netdev);
2185 	if (err)
2186 		e_err("Unable to allocate interrupt, Error: %d\n", err);
2187 
2188 	return err;
2189 }
2190 
2191 static void e1000_free_irq(struct e1000_adapter *adapter)
2192 {
2193 	struct net_device *netdev = adapter->netdev;
2194 
2195 	if (adapter->msix_entries) {
2196 		int vector = 0;
2197 
2198 		free_irq(adapter->msix_entries[vector].vector, netdev);
2199 		vector++;
2200 
2201 		free_irq(adapter->msix_entries[vector].vector, netdev);
2202 		vector++;
2203 
2204 		/* Other Causes interrupt vector */
2205 		free_irq(adapter->msix_entries[vector].vector, netdev);
2206 		return;
2207 	}
2208 
2209 	free_irq(adapter->pdev->irq, netdev);
2210 }
2211 
2212 /**
2213  * e1000_irq_disable - Mask off interrupt generation on the NIC
2214  **/
2215 static void e1000_irq_disable(struct e1000_adapter *adapter)
2216 {
2217 	struct e1000_hw *hw = &adapter->hw;
2218 
2219 	ew32(IMC, ~0);
2220 	if (adapter->msix_entries)
2221 		ew32(EIAC_82574, 0);
2222 	e1e_flush();
2223 
2224 	if (adapter->msix_entries) {
2225 		int i;
2226 
2227 		for (i = 0; i < adapter->num_vectors; i++)
2228 			synchronize_irq(adapter->msix_entries[i].vector);
2229 	} else {
2230 		synchronize_irq(adapter->pdev->irq);
2231 	}
2232 }
2233 
2234 /**
2235  * e1000_irq_enable - Enable default interrupt generation settings
2236  **/
2237 static void e1000_irq_enable(struct e1000_adapter *adapter)
2238 {
2239 	struct e1000_hw *hw = &adapter->hw;
2240 
2241 	if (adapter->msix_entries) {
2242 		ew32(EIAC_82574, adapter->eiac_mask & E1000_EIAC_MASK_82574);
2243 		ew32(IMS, adapter->eiac_mask | E1000_IMS_LSC);
2244 	} else if ((hw->mac.type == e1000_pch_lpt) ||
2245 		   (hw->mac.type == e1000_pch_spt)) {
2246 		ew32(IMS, IMS_ENABLE_MASK | E1000_IMS_ECCER);
2247 	} else {
2248 		ew32(IMS, IMS_ENABLE_MASK);
2249 	}
2250 	e1e_flush();
2251 }
2252 
2253 /**
2254  * e1000e_get_hw_control - get control of the h/w from f/w
2255  * @adapter: address of board private structure
2256  *
2257  * e1000e_get_hw_control sets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2258  * For ASF and Pass Through versions of f/w this means that
2259  * the driver is loaded. For AMT version (only with 82573)
2260  * of the f/w this means that the network i/f is open.
2261  **/
2262 void e1000e_get_hw_control(struct e1000_adapter *adapter)
2263 {
2264 	struct e1000_hw *hw = &adapter->hw;
2265 	u32 ctrl_ext;
2266 	u32 swsm;
2267 
2268 	/* Let firmware know the driver has taken over */
2269 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2270 		swsm = er32(SWSM);
2271 		ew32(SWSM, swsm | E1000_SWSM_DRV_LOAD);
2272 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2273 		ctrl_ext = er32(CTRL_EXT);
2274 		ew32(CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2275 	}
2276 }
2277 
2278 /**
2279  * e1000e_release_hw_control - release control of the h/w to f/w
2280  * @adapter: address of board private structure
2281  *
2282  * e1000e_release_hw_control resets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2283  * For ASF and Pass Through versions of f/w this means that the
2284  * driver is no longer loaded. For AMT version (only with 82573) i
2285  * of the f/w this means that the network i/f is closed.
2286  *
2287  **/
2288 void e1000e_release_hw_control(struct e1000_adapter *adapter)
2289 {
2290 	struct e1000_hw *hw = &adapter->hw;
2291 	u32 ctrl_ext;
2292 	u32 swsm;
2293 
2294 	/* Let firmware taken over control of h/w */
2295 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2296 		swsm = er32(SWSM);
2297 		ew32(SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
2298 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2299 		ctrl_ext = er32(CTRL_EXT);
2300 		ew32(CTRL_EXT, ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2301 	}
2302 }
2303 
2304 /**
2305  * e1000_alloc_ring_dma - allocate memory for a ring structure
2306  **/
2307 static int e1000_alloc_ring_dma(struct e1000_adapter *adapter,
2308 				struct e1000_ring *ring)
2309 {
2310 	struct pci_dev *pdev = adapter->pdev;
2311 
2312 	ring->desc = dma_alloc_coherent(&pdev->dev, ring->size, &ring->dma,
2313 					GFP_KERNEL);
2314 	if (!ring->desc)
2315 		return -ENOMEM;
2316 
2317 	return 0;
2318 }
2319 
2320 /**
2321  * e1000e_setup_tx_resources - allocate Tx resources (Descriptors)
2322  * @tx_ring: Tx descriptor ring
2323  *
2324  * Return 0 on success, negative on failure
2325  **/
2326 int e1000e_setup_tx_resources(struct e1000_ring *tx_ring)
2327 {
2328 	struct e1000_adapter *adapter = tx_ring->adapter;
2329 	int err = -ENOMEM, size;
2330 
2331 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2332 	tx_ring->buffer_info = vzalloc(size);
2333 	if (!tx_ring->buffer_info)
2334 		goto err;
2335 
2336 	/* round up to nearest 4K */
2337 	tx_ring->size = tx_ring->count * sizeof(struct e1000_tx_desc);
2338 	tx_ring->size = ALIGN(tx_ring->size, 4096);
2339 
2340 	err = e1000_alloc_ring_dma(adapter, tx_ring);
2341 	if (err)
2342 		goto err;
2343 
2344 	tx_ring->next_to_use = 0;
2345 	tx_ring->next_to_clean = 0;
2346 
2347 	return 0;
2348 err:
2349 	vfree(tx_ring->buffer_info);
2350 	e_err("Unable to allocate memory for the transmit descriptor ring\n");
2351 	return err;
2352 }
2353 
2354 /**
2355  * e1000e_setup_rx_resources - allocate Rx resources (Descriptors)
2356  * @rx_ring: Rx descriptor ring
2357  *
2358  * Returns 0 on success, negative on failure
2359  **/
2360 int e1000e_setup_rx_resources(struct e1000_ring *rx_ring)
2361 {
2362 	struct e1000_adapter *adapter = rx_ring->adapter;
2363 	struct e1000_buffer *buffer_info;
2364 	int i, size, desc_len, err = -ENOMEM;
2365 
2366 	size = sizeof(struct e1000_buffer) * rx_ring->count;
2367 	rx_ring->buffer_info = vzalloc(size);
2368 	if (!rx_ring->buffer_info)
2369 		goto err;
2370 
2371 	for (i = 0; i < rx_ring->count; i++) {
2372 		buffer_info = &rx_ring->buffer_info[i];
2373 		buffer_info->ps_pages = kcalloc(PS_PAGE_BUFFERS,
2374 						sizeof(struct e1000_ps_page),
2375 						GFP_KERNEL);
2376 		if (!buffer_info->ps_pages)
2377 			goto err_pages;
2378 	}
2379 
2380 	desc_len = sizeof(union e1000_rx_desc_packet_split);
2381 
2382 	/* Round up to nearest 4K */
2383 	rx_ring->size = rx_ring->count * desc_len;
2384 	rx_ring->size = ALIGN(rx_ring->size, 4096);
2385 
2386 	err = e1000_alloc_ring_dma(adapter, rx_ring);
2387 	if (err)
2388 		goto err_pages;
2389 
2390 	rx_ring->next_to_clean = 0;
2391 	rx_ring->next_to_use = 0;
2392 	rx_ring->rx_skb_top = NULL;
2393 
2394 	return 0;
2395 
2396 err_pages:
2397 	for (i = 0; i < rx_ring->count; i++) {
2398 		buffer_info = &rx_ring->buffer_info[i];
2399 		kfree(buffer_info->ps_pages);
2400 	}
2401 err:
2402 	vfree(rx_ring->buffer_info);
2403 	e_err("Unable to allocate memory for the receive descriptor ring\n");
2404 	return err;
2405 }
2406 
2407 /**
2408  * e1000_clean_tx_ring - Free Tx Buffers
2409  * @tx_ring: Tx descriptor ring
2410  **/
2411 static void e1000_clean_tx_ring(struct e1000_ring *tx_ring)
2412 {
2413 	struct e1000_adapter *adapter = tx_ring->adapter;
2414 	struct e1000_buffer *buffer_info;
2415 	unsigned long size;
2416 	unsigned int i;
2417 
2418 	for (i = 0; i < tx_ring->count; i++) {
2419 		buffer_info = &tx_ring->buffer_info[i];
2420 		e1000_put_txbuf(tx_ring, buffer_info);
2421 	}
2422 
2423 	netdev_reset_queue(adapter->netdev);
2424 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2425 	memset(tx_ring->buffer_info, 0, size);
2426 
2427 	memset(tx_ring->desc, 0, tx_ring->size);
2428 
2429 	tx_ring->next_to_use = 0;
2430 	tx_ring->next_to_clean = 0;
2431 }
2432 
2433 /**
2434  * e1000e_free_tx_resources - Free Tx Resources per Queue
2435  * @tx_ring: Tx descriptor ring
2436  *
2437  * Free all transmit software resources
2438  **/
2439 void e1000e_free_tx_resources(struct e1000_ring *tx_ring)
2440 {
2441 	struct e1000_adapter *adapter = tx_ring->adapter;
2442 	struct pci_dev *pdev = adapter->pdev;
2443 
2444 	e1000_clean_tx_ring(tx_ring);
2445 
2446 	vfree(tx_ring->buffer_info);
2447 	tx_ring->buffer_info = NULL;
2448 
2449 	dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
2450 			  tx_ring->dma);
2451 	tx_ring->desc = NULL;
2452 }
2453 
2454 /**
2455  * e1000e_free_rx_resources - Free Rx Resources
2456  * @rx_ring: Rx descriptor ring
2457  *
2458  * Free all receive software resources
2459  **/
2460 void e1000e_free_rx_resources(struct e1000_ring *rx_ring)
2461 {
2462 	struct e1000_adapter *adapter = rx_ring->adapter;
2463 	struct pci_dev *pdev = adapter->pdev;
2464 	int i;
2465 
2466 	e1000_clean_rx_ring(rx_ring);
2467 
2468 	for (i = 0; i < rx_ring->count; i++)
2469 		kfree(rx_ring->buffer_info[i].ps_pages);
2470 
2471 	vfree(rx_ring->buffer_info);
2472 	rx_ring->buffer_info = NULL;
2473 
2474 	dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
2475 			  rx_ring->dma);
2476 	rx_ring->desc = NULL;
2477 }
2478 
2479 /**
2480  * e1000_update_itr - update the dynamic ITR value based on statistics
2481  * @adapter: pointer to adapter
2482  * @itr_setting: current adapter->itr
2483  * @packets: the number of packets during this measurement interval
2484  * @bytes: the number of bytes during this measurement interval
2485  *
2486  *      Stores a new ITR value based on packets and byte
2487  *      counts during the last interrupt.  The advantage of per interrupt
2488  *      computation is faster updates and more accurate ITR for the current
2489  *      traffic pattern.  Constants in this function were computed
2490  *      based on theoretical maximum wire speed and thresholds were set based
2491  *      on testing data as well as attempting to minimize response time
2492  *      while increasing bulk throughput.  This functionality is controlled
2493  *      by the InterruptThrottleRate module parameter.
2494  **/
2495 static unsigned int e1000_update_itr(u16 itr_setting, int packets, int bytes)
2496 {
2497 	unsigned int retval = itr_setting;
2498 
2499 	if (packets == 0)
2500 		return itr_setting;
2501 
2502 	switch (itr_setting) {
2503 	case lowest_latency:
2504 		/* handle TSO and jumbo frames */
2505 		if (bytes / packets > 8000)
2506 			retval = bulk_latency;
2507 		else if ((packets < 5) && (bytes > 512))
2508 			retval = low_latency;
2509 		break;
2510 	case low_latency:	/* 50 usec aka 20000 ints/s */
2511 		if (bytes > 10000) {
2512 			/* this if handles the TSO accounting */
2513 			if (bytes / packets > 8000)
2514 				retval = bulk_latency;
2515 			else if ((packets < 10) || ((bytes / packets) > 1200))
2516 				retval = bulk_latency;
2517 			else if ((packets > 35))
2518 				retval = lowest_latency;
2519 		} else if (bytes / packets > 2000) {
2520 			retval = bulk_latency;
2521 		} else if (packets <= 2 && bytes < 512) {
2522 			retval = lowest_latency;
2523 		}
2524 		break;
2525 	case bulk_latency:	/* 250 usec aka 4000 ints/s */
2526 		if (bytes > 25000) {
2527 			if (packets > 35)
2528 				retval = low_latency;
2529 		} else if (bytes < 6000) {
2530 			retval = low_latency;
2531 		}
2532 		break;
2533 	}
2534 
2535 	return retval;
2536 }
2537 
2538 static void e1000_set_itr(struct e1000_adapter *adapter)
2539 {
2540 	u16 current_itr;
2541 	u32 new_itr = adapter->itr;
2542 
2543 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
2544 	if (adapter->link_speed != SPEED_1000) {
2545 		current_itr = 0;
2546 		new_itr = 4000;
2547 		goto set_itr_now;
2548 	}
2549 
2550 	if (adapter->flags2 & FLAG2_DISABLE_AIM) {
2551 		new_itr = 0;
2552 		goto set_itr_now;
2553 	}
2554 
2555 	adapter->tx_itr = e1000_update_itr(adapter->tx_itr,
2556 					   adapter->total_tx_packets,
2557 					   adapter->total_tx_bytes);
2558 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2559 	if (adapter->itr_setting == 3 && adapter->tx_itr == lowest_latency)
2560 		adapter->tx_itr = low_latency;
2561 
2562 	adapter->rx_itr = e1000_update_itr(adapter->rx_itr,
2563 					   adapter->total_rx_packets,
2564 					   adapter->total_rx_bytes);
2565 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2566 	if (adapter->itr_setting == 3 && adapter->rx_itr == lowest_latency)
2567 		adapter->rx_itr = low_latency;
2568 
2569 	current_itr = max(adapter->rx_itr, adapter->tx_itr);
2570 
2571 	/* counts and packets in update_itr are dependent on these numbers */
2572 	switch (current_itr) {
2573 	case lowest_latency:
2574 		new_itr = 70000;
2575 		break;
2576 	case low_latency:
2577 		new_itr = 20000;	/* aka hwitr = ~200 */
2578 		break;
2579 	case bulk_latency:
2580 		new_itr = 4000;
2581 		break;
2582 	default:
2583 		break;
2584 	}
2585 
2586 set_itr_now:
2587 	if (new_itr != adapter->itr) {
2588 		/* this attempts to bias the interrupt rate towards Bulk
2589 		 * by adding intermediate steps when interrupt rate is
2590 		 * increasing
2591 		 */
2592 		new_itr = new_itr > adapter->itr ?
2593 		    min(adapter->itr + (new_itr >> 2), new_itr) : new_itr;
2594 		adapter->itr = new_itr;
2595 		adapter->rx_ring->itr_val = new_itr;
2596 		if (adapter->msix_entries)
2597 			adapter->rx_ring->set_itr = 1;
2598 		else
2599 			e1000e_write_itr(adapter, new_itr);
2600 	}
2601 }
2602 
2603 /**
2604  * e1000e_write_itr - write the ITR value to the appropriate registers
2605  * @adapter: address of board private structure
2606  * @itr: new ITR value to program
2607  *
2608  * e1000e_write_itr determines if the adapter is in MSI-X mode
2609  * and, if so, writes the EITR registers with the ITR value.
2610  * Otherwise, it writes the ITR value into the ITR register.
2611  **/
2612 void e1000e_write_itr(struct e1000_adapter *adapter, u32 itr)
2613 {
2614 	struct e1000_hw *hw = &adapter->hw;
2615 	u32 new_itr = itr ? 1000000000 / (itr * 256) : 0;
2616 
2617 	if (adapter->msix_entries) {
2618 		int vector;
2619 
2620 		for (vector = 0; vector < adapter->num_vectors; vector++)
2621 			writel(new_itr, hw->hw_addr + E1000_EITR_82574(vector));
2622 	} else {
2623 		ew32(ITR, new_itr);
2624 	}
2625 }
2626 
2627 /**
2628  * e1000_alloc_queues - Allocate memory for all rings
2629  * @adapter: board private structure to initialize
2630  **/
2631 static int e1000_alloc_queues(struct e1000_adapter *adapter)
2632 {
2633 	int size = sizeof(struct e1000_ring);
2634 
2635 	adapter->tx_ring = kzalloc(size, GFP_KERNEL);
2636 	if (!adapter->tx_ring)
2637 		goto err;
2638 	adapter->tx_ring->count = adapter->tx_ring_count;
2639 	adapter->tx_ring->adapter = adapter;
2640 
2641 	adapter->rx_ring = kzalloc(size, GFP_KERNEL);
2642 	if (!adapter->rx_ring)
2643 		goto err;
2644 	adapter->rx_ring->count = adapter->rx_ring_count;
2645 	adapter->rx_ring->adapter = adapter;
2646 
2647 	return 0;
2648 err:
2649 	e_err("Unable to allocate memory for queues\n");
2650 	kfree(adapter->rx_ring);
2651 	kfree(adapter->tx_ring);
2652 	return -ENOMEM;
2653 }
2654 
2655 /**
2656  * e1000e_poll - NAPI Rx polling callback
2657  * @napi: struct associated with this polling callback
2658  * @weight: number of packets driver is allowed to process this poll
2659  **/
2660 static int e1000e_poll(struct napi_struct *napi, int weight)
2661 {
2662 	struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter,
2663 						     napi);
2664 	struct e1000_hw *hw = &adapter->hw;
2665 	struct net_device *poll_dev = adapter->netdev;
2666 	int tx_cleaned = 1, work_done = 0;
2667 
2668 	adapter = netdev_priv(poll_dev);
2669 
2670 	if (!adapter->msix_entries ||
2671 	    (adapter->rx_ring->ims_val & adapter->tx_ring->ims_val))
2672 		tx_cleaned = e1000_clean_tx_irq(adapter->tx_ring);
2673 
2674 	adapter->clean_rx(adapter->rx_ring, &work_done, weight);
2675 
2676 	if (!tx_cleaned)
2677 		work_done = weight;
2678 
2679 	/* If weight not fully consumed, exit the polling mode */
2680 	if (work_done < weight) {
2681 		if (adapter->itr_setting & 3)
2682 			e1000_set_itr(adapter);
2683 		napi_complete_done(napi, work_done);
2684 		if (!test_bit(__E1000_DOWN, &adapter->state)) {
2685 			if (adapter->msix_entries)
2686 				ew32(IMS, adapter->rx_ring->ims_val);
2687 			else
2688 				e1000_irq_enable(adapter);
2689 		}
2690 	}
2691 
2692 	return work_done;
2693 }
2694 
2695 static int e1000_vlan_rx_add_vid(struct net_device *netdev,
2696 				 __always_unused __be16 proto, u16 vid)
2697 {
2698 	struct e1000_adapter *adapter = netdev_priv(netdev);
2699 	struct e1000_hw *hw = &adapter->hw;
2700 	u32 vfta, index;
2701 
2702 	/* don't update vlan cookie if already programmed */
2703 	if ((adapter->hw.mng_cookie.status &
2704 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2705 	    (vid == adapter->mng_vlan_id))
2706 		return 0;
2707 
2708 	/* add VID to filter table */
2709 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2710 		index = (vid >> 5) & 0x7F;
2711 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2712 		vfta |= BIT((vid & 0x1F));
2713 		hw->mac.ops.write_vfta(hw, index, vfta);
2714 	}
2715 
2716 	set_bit(vid, adapter->active_vlans);
2717 
2718 	return 0;
2719 }
2720 
2721 static int e1000_vlan_rx_kill_vid(struct net_device *netdev,
2722 				  __always_unused __be16 proto, u16 vid)
2723 {
2724 	struct e1000_adapter *adapter = netdev_priv(netdev);
2725 	struct e1000_hw *hw = &adapter->hw;
2726 	u32 vfta, index;
2727 
2728 	if ((adapter->hw.mng_cookie.status &
2729 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2730 	    (vid == adapter->mng_vlan_id)) {
2731 		/* release control to f/w */
2732 		e1000e_release_hw_control(adapter);
2733 		return 0;
2734 	}
2735 
2736 	/* remove VID from filter table */
2737 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2738 		index = (vid >> 5) & 0x7F;
2739 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2740 		vfta &= ~BIT((vid & 0x1F));
2741 		hw->mac.ops.write_vfta(hw, index, vfta);
2742 	}
2743 
2744 	clear_bit(vid, adapter->active_vlans);
2745 
2746 	return 0;
2747 }
2748 
2749 /**
2750  * e1000e_vlan_filter_disable - helper to disable hw VLAN filtering
2751  * @adapter: board private structure to initialize
2752  **/
2753 static void e1000e_vlan_filter_disable(struct e1000_adapter *adapter)
2754 {
2755 	struct net_device *netdev = adapter->netdev;
2756 	struct e1000_hw *hw = &adapter->hw;
2757 	u32 rctl;
2758 
2759 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2760 		/* disable VLAN receive filtering */
2761 		rctl = er32(RCTL);
2762 		rctl &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
2763 		ew32(RCTL, rctl);
2764 
2765 		if (adapter->mng_vlan_id != (u16)E1000_MNG_VLAN_NONE) {
2766 			e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
2767 					       adapter->mng_vlan_id);
2768 			adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
2769 		}
2770 	}
2771 }
2772 
2773 /**
2774  * e1000e_vlan_filter_enable - helper to enable HW VLAN filtering
2775  * @adapter: board private structure to initialize
2776  **/
2777 static void e1000e_vlan_filter_enable(struct e1000_adapter *adapter)
2778 {
2779 	struct e1000_hw *hw = &adapter->hw;
2780 	u32 rctl;
2781 
2782 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2783 		/* enable VLAN receive filtering */
2784 		rctl = er32(RCTL);
2785 		rctl |= E1000_RCTL_VFE;
2786 		rctl &= ~E1000_RCTL_CFIEN;
2787 		ew32(RCTL, rctl);
2788 	}
2789 }
2790 
2791 /**
2792  * e1000e_vlan_strip_disable - helper to disable HW VLAN stripping
2793  * @adapter: board private structure to initialize
2794  **/
2795 static void e1000e_vlan_strip_disable(struct e1000_adapter *adapter)
2796 {
2797 	struct e1000_hw *hw = &adapter->hw;
2798 	u32 ctrl;
2799 
2800 	/* disable VLAN tag insert/strip */
2801 	ctrl = er32(CTRL);
2802 	ctrl &= ~E1000_CTRL_VME;
2803 	ew32(CTRL, ctrl);
2804 }
2805 
2806 /**
2807  * e1000e_vlan_strip_enable - helper to enable HW VLAN stripping
2808  * @adapter: board private structure to initialize
2809  **/
2810 static void e1000e_vlan_strip_enable(struct e1000_adapter *adapter)
2811 {
2812 	struct e1000_hw *hw = &adapter->hw;
2813 	u32 ctrl;
2814 
2815 	/* enable VLAN tag insert/strip */
2816 	ctrl = er32(CTRL);
2817 	ctrl |= E1000_CTRL_VME;
2818 	ew32(CTRL, ctrl);
2819 }
2820 
2821 static void e1000_update_mng_vlan(struct e1000_adapter *adapter)
2822 {
2823 	struct net_device *netdev = adapter->netdev;
2824 	u16 vid = adapter->hw.mng_cookie.vlan_id;
2825 	u16 old_vid = adapter->mng_vlan_id;
2826 
2827 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
2828 		e1000_vlan_rx_add_vid(netdev, htons(ETH_P_8021Q), vid);
2829 		adapter->mng_vlan_id = vid;
2830 	}
2831 
2832 	if ((old_vid != (u16)E1000_MNG_VLAN_NONE) && (vid != old_vid))
2833 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q), old_vid);
2834 }
2835 
2836 static void e1000_restore_vlan(struct e1000_adapter *adapter)
2837 {
2838 	u16 vid;
2839 
2840 	e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
2841 
2842 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
2843 	    e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
2844 }
2845 
2846 static void e1000_init_manageability_pt(struct e1000_adapter *adapter)
2847 {
2848 	struct e1000_hw *hw = &adapter->hw;
2849 	u32 manc, manc2h, mdef, i, j;
2850 
2851 	if (!(adapter->flags & FLAG_MNG_PT_ENABLED))
2852 		return;
2853 
2854 	manc = er32(MANC);
2855 
2856 	/* enable receiving management packets to the host. this will probably
2857 	 * generate destination unreachable messages from the host OS, but
2858 	 * the packets will be handled on SMBUS
2859 	 */
2860 	manc |= E1000_MANC_EN_MNG2HOST;
2861 	manc2h = er32(MANC2H);
2862 
2863 	switch (hw->mac.type) {
2864 	default:
2865 		manc2h |= (E1000_MANC2H_PORT_623 | E1000_MANC2H_PORT_664);
2866 		break;
2867 	case e1000_82574:
2868 	case e1000_82583:
2869 		/* Check if IPMI pass-through decision filter already exists;
2870 		 * if so, enable it.
2871 		 */
2872 		for (i = 0, j = 0; i < 8; i++) {
2873 			mdef = er32(MDEF(i));
2874 
2875 			/* Ignore filters with anything other than IPMI ports */
2876 			if (mdef & ~(E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2877 				continue;
2878 
2879 			/* Enable this decision filter in MANC2H */
2880 			if (mdef)
2881 				manc2h |= BIT(i);
2882 
2883 			j |= mdef;
2884 		}
2885 
2886 		if (j == (E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2887 			break;
2888 
2889 		/* Create new decision filter in an empty filter */
2890 		for (i = 0, j = 0; i < 8; i++)
2891 			if (er32(MDEF(i)) == 0) {
2892 				ew32(MDEF(i), (E1000_MDEF_PORT_623 |
2893 					       E1000_MDEF_PORT_664));
2894 				manc2h |= BIT(1);
2895 				j++;
2896 				break;
2897 			}
2898 
2899 		if (!j)
2900 			e_warn("Unable to create IPMI pass-through filter\n");
2901 		break;
2902 	}
2903 
2904 	ew32(MANC2H, manc2h);
2905 	ew32(MANC, manc);
2906 }
2907 
2908 /**
2909  * e1000_configure_tx - Configure Transmit Unit after Reset
2910  * @adapter: board private structure
2911  *
2912  * Configure the Tx unit of the MAC after a reset.
2913  **/
2914 static void e1000_configure_tx(struct e1000_adapter *adapter)
2915 {
2916 	struct e1000_hw *hw = &adapter->hw;
2917 	struct e1000_ring *tx_ring = adapter->tx_ring;
2918 	u64 tdba;
2919 	u32 tdlen, tctl, tarc;
2920 
2921 	/* Setup the HW Tx Head and Tail descriptor pointers */
2922 	tdba = tx_ring->dma;
2923 	tdlen = tx_ring->count * sizeof(struct e1000_tx_desc);
2924 	ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
2925 	ew32(TDBAH(0), (tdba >> 32));
2926 	ew32(TDLEN(0), tdlen);
2927 	ew32(TDH(0), 0);
2928 	ew32(TDT(0), 0);
2929 	tx_ring->head = adapter->hw.hw_addr + E1000_TDH(0);
2930 	tx_ring->tail = adapter->hw.hw_addr + E1000_TDT(0);
2931 
2932 	writel(0, tx_ring->head);
2933 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
2934 		e1000e_update_tdt_wa(tx_ring, 0);
2935 	else
2936 		writel(0, tx_ring->tail);
2937 
2938 	/* Set the Tx Interrupt Delay register */
2939 	ew32(TIDV, adapter->tx_int_delay);
2940 	/* Tx irq moderation */
2941 	ew32(TADV, adapter->tx_abs_int_delay);
2942 
2943 	if (adapter->flags2 & FLAG2_DMA_BURST) {
2944 		u32 txdctl = er32(TXDCTL(0));
2945 
2946 		txdctl &= ~(E1000_TXDCTL_PTHRESH | E1000_TXDCTL_HTHRESH |
2947 			    E1000_TXDCTL_WTHRESH);
2948 		/* set up some performance related parameters to encourage the
2949 		 * hardware to use the bus more efficiently in bursts, depends
2950 		 * on the tx_int_delay to be enabled,
2951 		 * wthresh = 1 ==> burst write is disabled to avoid Tx stalls
2952 		 * hthresh = 1 ==> prefetch when one or more available
2953 		 * pthresh = 0x1f ==> prefetch if internal cache 31 or less
2954 		 * BEWARE: this seems to work but should be considered first if
2955 		 * there are Tx hangs or other Tx related bugs
2956 		 */
2957 		txdctl |= E1000_TXDCTL_DMA_BURST_ENABLE;
2958 		ew32(TXDCTL(0), txdctl);
2959 	}
2960 	/* erratum work around: set txdctl the same for both queues */
2961 	ew32(TXDCTL(1), er32(TXDCTL(0)));
2962 
2963 	/* Program the Transmit Control Register */
2964 	tctl = er32(TCTL);
2965 	tctl &= ~E1000_TCTL_CT;
2966 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
2967 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
2968 
2969 	if (adapter->flags & FLAG_TARC_SPEED_MODE_BIT) {
2970 		tarc = er32(TARC(0));
2971 		/* set the speed mode bit, we'll clear it if we're not at
2972 		 * gigabit link later
2973 		 */
2974 #define SPEED_MODE_BIT BIT(21)
2975 		tarc |= SPEED_MODE_BIT;
2976 		ew32(TARC(0), tarc);
2977 	}
2978 
2979 	/* errata: program both queues to unweighted RR */
2980 	if (adapter->flags & FLAG_TARC_SET_BIT_ZERO) {
2981 		tarc = er32(TARC(0));
2982 		tarc |= 1;
2983 		ew32(TARC(0), tarc);
2984 		tarc = er32(TARC(1));
2985 		tarc |= 1;
2986 		ew32(TARC(1), tarc);
2987 	}
2988 
2989 	/* Setup Transmit Descriptor Settings for eop descriptor */
2990 	adapter->txd_cmd = E1000_TXD_CMD_EOP | E1000_TXD_CMD_IFCS;
2991 
2992 	/* only set IDE if we are delaying interrupts using the timers */
2993 	if (adapter->tx_int_delay)
2994 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
2995 
2996 	/* enable Report Status bit */
2997 	adapter->txd_cmd |= E1000_TXD_CMD_RS;
2998 
2999 	ew32(TCTL, tctl);
3000 
3001 	hw->mac.ops.config_collision_dist(hw);
3002 
3003 	/* SPT Si errata workaround to avoid data corruption */
3004 	if (hw->mac.type == e1000_pch_spt) {
3005 		u32 reg_val;
3006 
3007 		reg_val = er32(IOSFPC);
3008 		reg_val |= E1000_RCTL_RDMTS_HEX;
3009 		ew32(IOSFPC, reg_val);
3010 
3011 		reg_val = er32(TARC(0));
3012 		reg_val |= E1000_TARC0_CB_MULTIQ_3_REQ;
3013 		ew32(TARC(0), reg_val);
3014 	}
3015 }
3016 
3017 /**
3018  * e1000_setup_rctl - configure the receive control registers
3019  * @adapter: Board private structure
3020  **/
3021 #define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
3022 			   (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
3023 static void e1000_setup_rctl(struct e1000_adapter *adapter)
3024 {
3025 	struct e1000_hw *hw = &adapter->hw;
3026 	u32 rctl, rfctl;
3027 	u32 pages = 0;
3028 
3029 	/* Workaround Si errata on PCHx - configure jumbo frame flow.
3030 	 * If jumbo frames not set, program related MAC/PHY registers
3031 	 * to h/w defaults
3032 	 */
3033 	if (hw->mac.type >= e1000_pch2lan) {
3034 		s32 ret_val;
3035 
3036 		if (adapter->netdev->mtu > ETH_DATA_LEN)
3037 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, true);
3038 		else
3039 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, false);
3040 
3041 		if (ret_val)
3042 			e_dbg("failed to enable|disable jumbo frame workaround mode\n");
3043 	}
3044 
3045 	/* Program MC offset vector base */
3046 	rctl = er32(RCTL);
3047 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3048 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3049 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3050 	    (adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3051 
3052 	/* Do not Store bad packets */
3053 	rctl &= ~E1000_RCTL_SBP;
3054 
3055 	/* Enable Long Packet receive */
3056 	if (adapter->netdev->mtu <= ETH_DATA_LEN)
3057 		rctl &= ~E1000_RCTL_LPE;
3058 	else
3059 		rctl |= E1000_RCTL_LPE;
3060 
3061 	/* Some systems expect that the CRC is included in SMBUS traffic. The
3062 	 * hardware strips the CRC before sending to both SMBUS (BMC) and to
3063 	 * host memory when this is enabled
3064 	 */
3065 	if (adapter->flags2 & FLAG2_CRC_STRIPPING)
3066 		rctl |= E1000_RCTL_SECRC;
3067 
3068 	/* Workaround Si errata on 82577 PHY - configure IPG for jumbos */
3069 	if ((hw->phy.type == e1000_phy_82577) && (rctl & E1000_RCTL_LPE)) {
3070 		u16 phy_data;
3071 
3072 		e1e_rphy(hw, PHY_REG(770, 26), &phy_data);
3073 		phy_data &= 0xfff8;
3074 		phy_data |= BIT(2);
3075 		e1e_wphy(hw, PHY_REG(770, 26), phy_data);
3076 
3077 		e1e_rphy(hw, 22, &phy_data);
3078 		phy_data &= 0x0fff;
3079 		phy_data |= BIT(14);
3080 		e1e_wphy(hw, 0x10, 0x2823);
3081 		e1e_wphy(hw, 0x11, 0x0003);
3082 		e1e_wphy(hw, 22, phy_data);
3083 	}
3084 
3085 	/* Setup buffer sizes */
3086 	rctl &= ~E1000_RCTL_SZ_4096;
3087 	rctl |= E1000_RCTL_BSEX;
3088 	switch (adapter->rx_buffer_len) {
3089 	case 2048:
3090 	default:
3091 		rctl |= E1000_RCTL_SZ_2048;
3092 		rctl &= ~E1000_RCTL_BSEX;
3093 		break;
3094 	case 4096:
3095 		rctl |= E1000_RCTL_SZ_4096;
3096 		break;
3097 	case 8192:
3098 		rctl |= E1000_RCTL_SZ_8192;
3099 		break;
3100 	case 16384:
3101 		rctl |= E1000_RCTL_SZ_16384;
3102 		break;
3103 	}
3104 
3105 	/* Enable Extended Status in all Receive Descriptors */
3106 	rfctl = er32(RFCTL);
3107 	rfctl |= E1000_RFCTL_EXTEN;
3108 	ew32(RFCTL, rfctl);
3109 
3110 	/* 82571 and greater support packet-split where the protocol
3111 	 * header is placed in skb->data and the packet data is
3112 	 * placed in pages hanging off of skb_shinfo(skb)->nr_frags.
3113 	 * In the case of a non-split, skb->data is linearly filled,
3114 	 * followed by the page buffers.  Therefore, skb->data is
3115 	 * sized to hold the largest protocol header.
3116 	 *
3117 	 * allocations using alloc_page take too long for regular MTU
3118 	 * so only enable packet split for jumbo frames
3119 	 *
3120 	 * Using pages when the page size is greater than 16k wastes
3121 	 * a lot of memory, since we allocate 3 pages at all times
3122 	 * per packet.
3123 	 */
3124 	pages = PAGE_USE_COUNT(adapter->netdev->mtu);
3125 	if ((pages <= 3) && (PAGE_SIZE <= 16384) && (rctl & E1000_RCTL_LPE))
3126 		adapter->rx_ps_pages = pages;
3127 	else
3128 		adapter->rx_ps_pages = 0;
3129 
3130 	if (adapter->rx_ps_pages) {
3131 		u32 psrctl = 0;
3132 
3133 		/* Enable Packet split descriptors */
3134 		rctl |= E1000_RCTL_DTYP_PS;
3135 
3136 		psrctl |= adapter->rx_ps_bsize0 >> E1000_PSRCTL_BSIZE0_SHIFT;
3137 
3138 		switch (adapter->rx_ps_pages) {
3139 		case 3:
3140 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE3_SHIFT;
3141 			/* fall-through */
3142 		case 2:
3143 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE2_SHIFT;
3144 			/* fall-through */
3145 		case 1:
3146 			psrctl |= PAGE_SIZE >> E1000_PSRCTL_BSIZE1_SHIFT;
3147 			break;
3148 		}
3149 
3150 		ew32(PSRCTL, psrctl);
3151 	}
3152 
3153 	/* This is useful for sniffing bad packets. */
3154 	if (adapter->netdev->features & NETIF_F_RXALL) {
3155 		/* UPE and MPE will be handled by normal PROMISC logic
3156 		 * in e1000e_set_rx_mode
3157 		 */
3158 		rctl |= (E1000_RCTL_SBP |	/* Receive bad packets */
3159 			 E1000_RCTL_BAM |	/* RX All Bcast Pkts */
3160 			 E1000_RCTL_PMCF);	/* RX All MAC Ctrl Pkts */
3161 
3162 		rctl &= ~(E1000_RCTL_VFE |	/* Disable VLAN filter */
3163 			  E1000_RCTL_DPF |	/* Allow filtered pause */
3164 			  E1000_RCTL_CFIEN);	/* Dis VLAN CFIEN Filter */
3165 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3166 		 * and that breaks VLANs.
3167 		 */
3168 	}
3169 
3170 	ew32(RCTL, rctl);
3171 	/* just started the receive unit, no need to restart */
3172 	adapter->flags &= ~FLAG_RESTART_NOW;
3173 }
3174 
3175 /**
3176  * e1000_configure_rx - Configure Receive Unit after Reset
3177  * @adapter: board private structure
3178  *
3179  * Configure the Rx unit of the MAC after a reset.
3180  **/
3181 static void e1000_configure_rx(struct e1000_adapter *adapter)
3182 {
3183 	struct e1000_hw *hw = &adapter->hw;
3184 	struct e1000_ring *rx_ring = adapter->rx_ring;
3185 	u64 rdba;
3186 	u32 rdlen, rctl, rxcsum, ctrl_ext;
3187 
3188 	if (adapter->rx_ps_pages) {
3189 		/* this is a 32 byte descriptor */
3190 		rdlen = rx_ring->count *
3191 		    sizeof(union e1000_rx_desc_packet_split);
3192 		adapter->clean_rx = e1000_clean_rx_irq_ps;
3193 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers_ps;
3194 	} else if (adapter->netdev->mtu > ETH_FRAME_LEN + ETH_FCS_LEN) {
3195 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3196 		adapter->clean_rx = e1000_clean_jumbo_rx_irq;
3197 		adapter->alloc_rx_buf = e1000_alloc_jumbo_rx_buffers;
3198 	} else {
3199 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3200 		adapter->clean_rx = e1000_clean_rx_irq;
3201 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers;
3202 	}
3203 
3204 	/* disable receives while setting up the descriptors */
3205 	rctl = er32(RCTL);
3206 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
3207 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
3208 	e1e_flush();
3209 	usleep_range(10000, 20000);
3210 
3211 	if (adapter->flags2 & FLAG2_DMA_BURST) {
3212 		/* set the writeback threshold (only takes effect if the RDTR
3213 		 * is set). set GRAN=1 and write back up to 0x4 worth, and
3214 		 * enable prefetching of 0x20 Rx descriptors
3215 		 * granularity = 01
3216 		 * wthresh = 04,
3217 		 * hthresh = 04,
3218 		 * pthresh = 0x20
3219 		 */
3220 		ew32(RXDCTL(0), E1000_RXDCTL_DMA_BURST_ENABLE);
3221 		ew32(RXDCTL(1), E1000_RXDCTL_DMA_BURST_ENABLE);
3222 
3223 		/* override the delay timers for enabling bursting, only if
3224 		 * the value was not set by the user via module options
3225 		 */
3226 		if (adapter->rx_int_delay == DEFAULT_RDTR)
3227 			adapter->rx_int_delay = BURST_RDTR;
3228 		if (adapter->rx_abs_int_delay == DEFAULT_RADV)
3229 			adapter->rx_abs_int_delay = BURST_RADV;
3230 	}
3231 
3232 	/* set the Receive Delay Timer Register */
3233 	ew32(RDTR, adapter->rx_int_delay);
3234 
3235 	/* irq moderation */
3236 	ew32(RADV, adapter->rx_abs_int_delay);
3237 	if ((adapter->itr_setting != 0) && (adapter->itr != 0))
3238 		e1000e_write_itr(adapter, adapter->itr);
3239 
3240 	ctrl_ext = er32(CTRL_EXT);
3241 	/* Auto-Mask interrupts upon ICR access */
3242 	ctrl_ext |= E1000_CTRL_EXT_IAME;
3243 	ew32(IAM, 0xffffffff);
3244 	ew32(CTRL_EXT, ctrl_ext);
3245 	e1e_flush();
3246 
3247 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
3248 	 * the Base and Length of the Rx Descriptor Ring
3249 	 */
3250 	rdba = rx_ring->dma;
3251 	ew32(RDBAL(0), (rdba & DMA_BIT_MASK(32)));
3252 	ew32(RDBAH(0), (rdba >> 32));
3253 	ew32(RDLEN(0), rdlen);
3254 	ew32(RDH(0), 0);
3255 	ew32(RDT(0), 0);
3256 	rx_ring->head = adapter->hw.hw_addr + E1000_RDH(0);
3257 	rx_ring->tail = adapter->hw.hw_addr + E1000_RDT(0);
3258 
3259 	writel(0, rx_ring->head);
3260 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
3261 		e1000e_update_rdt_wa(rx_ring, 0);
3262 	else
3263 		writel(0, rx_ring->tail);
3264 
3265 	/* Enable Receive Checksum Offload for TCP and UDP */
3266 	rxcsum = er32(RXCSUM);
3267 	if (adapter->netdev->features & NETIF_F_RXCSUM)
3268 		rxcsum |= E1000_RXCSUM_TUOFL;
3269 	else
3270 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3271 	ew32(RXCSUM, rxcsum);
3272 
3273 	/* With jumbo frames, excessive C-state transition latencies result
3274 	 * in dropped transactions.
3275 	 */
3276 	if (adapter->netdev->mtu > ETH_DATA_LEN) {
3277 		u32 lat =
3278 		    ((er32(PBA) & E1000_PBA_RXA_MASK) * 1024 -
3279 		     adapter->max_frame_size) * 8 / 1000;
3280 
3281 		if (adapter->flags & FLAG_IS_ICH) {
3282 			u32 rxdctl = er32(RXDCTL(0));
3283 
3284 			ew32(RXDCTL(0), rxdctl | 0x3);
3285 		}
3286 
3287 		pm_qos_update_request(&adapter->pm_qos_req, lat);
3288 	} else {
3289 		pm_qos_update_request(&adapter->pm_qos_req,
3290 				      PM_QOS_DEFAULT_VALUE);
3291 	}
3292 
3293 	/* Enable Receives */
3294 	ew32(RCTL, rctl);
3295 }
3296 
3297 /**
3298  * e1000e_write_mc_addr_list - write multicast addresses to MTA
3299  * @netdev: network interface device structure
3300  *
3301  * Writes multicast address list to the MTA hash table.
3302  * Returns: -ENOMEM on failure
3303  *                0 on no addresses written
3304  *                X on writing X addresses to MTA
3305  */
3306 static int e1000e_write_mc_addr_list(struct net_device *netdev)
3307 {
3308 	struct e1000_adapter *adapter = netdev_priv(netdev);
3309 	struct e1000_hw *hw = &adapter->hw;
3310 	struct netdev_hw_addr *ha;
3311 	u8 *mta_list;
3312 	int i;
3313 
3314 	if (netdev_mc_empty(netdev)) {
3315 		/* nothing to program, so clear mc list */
3316 		hw->mac.ops.update_mc_addr_list(hw, NULL, 0);
3317 		return 0;
3318 	}
3319 
3320 	mta_list = kzalloc(netdev_mc_count(netdev) * ETH_ALEN, GFP_ATOMIC);
3321 	if (!mta_list)
3322 		return -ENOMEM;
3323 
3324 	/* update_mc_addr_list expects a packed array of only addresses. */
3325 	i = 0;
3326 	netdev_for_each_mc_addr(ha, netdev)
3327 	    memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3328 
3329 	hw->mac.ops.update_mc_addr_list(hw, mta_list, i);
3330 	kfree(mta_list);
3331 
3332 	return netdev_mc_count(netdev);
3333 }
3334 
3335 /**
3336  * e1000e_write_uc_addr_list - write unicast addresses to RAR table
3337  * @netdev: network interface device structure
3338  *
3339  * Writes unicast address list to the RAR table.
3340  * Returns: -ENOMEM on failure/insufficient address space
3341  *                0 on no addresses written
3342  *                X on writing X addresses to the RAR table
3343  **/
3344 static int e1000e_write_uc_addr_list(struct net_device *netdev)
3345 {
3346 	struct e1000_adapter *adapter = netdev_priv(netdev);
3347 	struct e1000_hw *hw = &adapter->hw;
3348 	unsigned int rar_entries;
3349 	int count = 0;
3350 
3351 	rar_entries = hw->mac.ops.rar_get_count(hw);
3352 
3353 	/* save a rar entry for our hardware address */
3354 	rar_entries--;
3355 
3356 	/* save a rar entry for the LAA workaround */
3357 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA)
3358 		rar_entries--;
3359 
3360 	/* return ENOMEM indicating insufficient memory for addresses */
3361 	if (netdev_uc_count(netdev) > rar_entries)
3362 		return -ENOMEM;
3363 
3364 	if (!netdev_uc_empty(netdev) && rar_entries) {
3365 		struct netdev_hw_addr *ha;
3366 
3367 		/* write the addresses in reverse order to avoid write
3368 		 * combining
3369 		 */
3370 		netdev_for_each_uc_addr(ha, netdev) {
3371 			int ret_val;
3372 
3373 			if (!rar_entries)
3374 				break;
3375 			ret_val = hw->mac.ops.rar_set(hw, ha->addr, rar_entries--);
3376 			if (ret_val < 0)
3377 				return -ENOMEM;
3378 			count++;
3379 		}
3380 	}
3381 
3382 	/* zero out the remaining RAR entries not used above */
3383 	for (; rar_entries > 0; rar_entries--) {
3384 		ew32(RAH(rar_entries), 0);
3385 		ew32(RAL(rar_entries), 0);
3386 	}
3387 	e1e_flush();
3388 
3389 	return count;
3390 }
3391 
3392 /**
3393  * e1000e_set_rx_mode - secondary unicast, Multicast and Promiscuous mode set
3394  * @netdev: network interface device structure
3395  *
3396  * The ndo_set_rx_mode entry point is called whenever the unicast or multicast
3397  * address list or the network interface flags are updated.  This routine is
3398  * responsible for configuring the hardware for proper unicast, multicast,
3399  * promiscuous mode, and all-multi behavior.
3400  **/
3401 static void e1000e_set_rx_mode(struct net_device *netdev)
3402 {
3403 	struct e1000_adapter *adapter = netdev_priv(netdev);
3404 	struct e1000_hw *hw = &adapter->hw;
3405 	u32 rctl;
3406 
3407 	if (pm_runtime_suspended(netdev->dev.parent))
3408 		return;
3409 
3410 	/* Check for Promiscuous and All Multicast modes */
3411 	rctl = er32(RCTL);
3412 
3413 	/* clear the affected bits */
3414 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
3415 
3416 	if (netdev->flags & IFF_PROMISC) {
3417 		rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
3418 		/* Do not hardware filter VLANs in promisc mode */
3419 		e1000e_vlan_filter_disable(adapter);
3420 	} else {
3421 		int count;
3422 
3423 		if (netdev->flags & IFF_ALLMULTI) {
3424 			rctl |= E1000_RCTL_MPE;
3425 		} else {
3426 			/* Write addresses to the MTA, if the attempt fails
3427 			 * then we should just turn on promiscuous mode so
3428 			 * that we can at least receive multicast traffic
3429 			 */
3430 			count = e1000e_write_mc_addr_list(netdev);
3431 			if (count < 0)
3432 				rctl |= E1000_RCTL_MPE;
3433 		}
3434 		e1000e_vlan_filter_enable(adapter);
3435 		/* Write addresses to available RAR registers, if there is not
3436 		 * sufficient space to store all the addresses then enable
3437 		 * unicast promiscuous mode
3438 		 */
3439 		count = e1000e_write_uc_addr_list(netdev);
3440 		if (count < 0)
3441 			rctl |= E1000_RCTL_UPE;
3442 	}
3443 
3444 	ew32(RCTL, rctl);
3445 
3446 	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
3447 		e1000e_vlan_strip_enable(adapter);
3448 	else
3449 		e1000e_vlan_strip_disable(adapter);
3450 }
3451 
3452 static void e1000e_setup_rss_hash(struct e1000_adapter *adapter)
3453 {
3454 	struct e1000_hw *hw = &adapter->hw;
3455 	u32 mrqc, rxcsum;
3456 	u32 rss_key[10];
3457 	int i;
3458 
3459 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3460 	for (i = 0; i < 10; i++)
3461 		ew32(RSSRK(i), rss_key[i]);
3462 
3463 	/* Direct all traffic to queue 0 */
3464 	for (i = 0; i < 32; i++)
3465 		ew32(RETA(i), 0);
3466 
3467 	/* Disable raw packet checksumming so that RSS hash is placed in
3468 	 * descriptor on writeback.
3469 	 */
3470 	rxcsum = er32(RXCSUM);
3471 	rxcsum |= E1000_RXCSUM_PCSD;
3472 
3473 	ew32(RXCSUM, rxcsum);
3474 
3475 	mrqc = (E1000_MRQC_RSS_FIELD_IPV4 |
3476 		E1000_MRQC_RSS_FIELD_IPV4_TCP |
3477 		E1000_MRQC_RSS_FIELD_IPV6 |
3478 		E1000_MRQC_RSS_FIELD_IPV6_TCP |
3479 		E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3480 
3481 	ew32(MRQC, mrqc);
3482 }
3483 
3484 /**
3485  * e1000e_get_base_timinca - get default SYSTIM time increment attributes
3486  * @adapter: board private structure
3487  * @timinca: pointer to returned time increment attributes
3488  *
3489  * Get attributes for incrementing the System Time Register SYSTIML/H at
3490  * the default base frequency, and set the cyclecounter shift value.
3491  **/
3492 s32 e1000e_get_base_timinca(struct e1000_adapter *adapter, u32 *timinca)
3493 {
3494 	struct e1000_hw *hw = &adapter->hw;
3495 	u32 incvalue, incperiod, shift;
3496 
3497 	/* Make sure clock is enabled on I217/I218/I219  before checking
3498 	 * the frequency
3499 	 */
3500 	if (((hw->mac.type == e1000_pch_lpt) ||
3501 	     (hw->mac.type == e1000_pch_spt)) &&
3502 	    !(er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) &&
3503 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_ENABLED)) {
3504 		u32 fextnvm7 = er32(FEXTNVM7);
3505 
3506 		if (!(fextnvm7 & BIT(0))) {
3507 			ew32(FEXTNVM7, fextnvm7 | BIT(0));
3508 			e1e_flush();
3509 		}
3510 	}
3511 
3512 	switch (hw->mac.type) {
3513 	case e1000_pch2lan:
3514 	case e1000_pch_lpt:
3515 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3516 			/* Stable 96MHz frequency */
3517 			incperiod = INCPERIOD_96MHz;
3518 			incvalue = INCVALUE_96MHz;
3519 			shift = INCVALUE_SHIFT_96MHz;
3520 			adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHz;
3521 		} else {
3522 			/* Stable 25MHz frequency */
3523 			incperiod = INCPERIOD_25MHz;
3524 			incvalue = INCVALUE_25MHz;
3525 			shift = INCVALUE_SHIFT_25MHz;
3526 			adapter->cc.shift = shift;
3527 		}
3528 		break;
3529 	case e1000_pch_spt:
3530 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3531 			/* Stable 24MHz frequency */
3532 			incperiod = INCPERIOD_24MHz;
3533 			incvalue = INCVALUE_24MHz;
3534 			shift = INCVALUE_SHIFT_24MHz;
3535 			adapter->cc.shift = shift;
3536 			break;
3537 		}
3538 		return -EINVAL;
3539 	case e1000_82574:
3540 	case e1000_82583:
3541 		/* Stable 25MHz frequency */
3542 		incperiod = INCPERIOD_25MHz;
3543 		incvalue = INCVALUE_25MHz;
3544 		shift = INCVALUE_SHIFT_25MHz;
3545 		adapter->cc.shift = shift;
3546 		break;
3547 	default:
3548 		return -EINVAL;
3549 	}
3550 
3551 	*timinca = ((incperiod << E1000_TIMINCA_INCPERIOD_SHIFT) |
3552 		    ((incvalue << shift) & E1000_TIMINCA_INCVALUE_MASK));
3553 
3554 	return 0;
3555 }
3556 
3557 /**
3558  * e1000e_config_hwtstamp - configure the hwtstamp registers and enable/disable
3559  * @adapter: board private structure
3560  *
3561  * Outgoing time stamping can be enabled and disabled. Play nice and
3562  * disable it when requested, although it shouldn't cause any overhead
3563  * when no packet needs it. At most one packet in the queue may be
3564  * marked for time stamping, otherwise it would be impossible to tell
3565  * for sure to which packet the hardware time stamp belongs.
3566  *
3567  * Incoming time stamping has to be configured via the hardware filters.
3568  * Not all combinations are supported, in particular event type has to be
3569  * specified. Matching the kind of event packet is not supported, with the
3570  * exception of "all V2 events regardless of level 2 or 4".
3571  **/
3572 static int e1000e_config_hwtstamp(struct e1000_adapter *adapter,
3573 				  struct hwtstamp_config *config)
3574 {
3575 	struct e1000_hw *hw = &adapter->hw;
3576 	u32 tsync_tx_ctl = E1000_TSYNCTXCTL_ENABLED;
3577 	u32 tsync_rx_ctl = E1000_TSYNCRXCTL_ENABLED;
3578 	u32 rxmtrl = 0;
3579 	u16 rxudp = 0;
3580 	bool is_l4 = false;
3581 	bool is_l2 = false;
3582 	u32 regval;
3583 
3584 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3585 		return -EINVAL;
3586 
3587 	/* flags reserved for future extensions - must be zero */
3588 	if (config->flags)
3589 		return -EINVAL;
3590 
3591 	switch (config->tx_type) {
3592 	case HWTSTAMP_TX_OFF:
3593 		tsync_tx_ctl = 0;
3594 		break;
3595 	case HWTSTAMP_TX_ON:
3596 		break;
3597 	default:
3598 		return -ERANGE;
3599 	}
3600 
3601 	switch (config->rx_filter) {
3602 	case HWTSTAMP_FILTER_NONE:
3603 		tsync_rx_ctl = 0;
3604 		break;
3605 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3606 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3607 		rxmtrl = E1000_RXMTRL_PTP_V1_SYNC_MESSAGE;
3608 		is_l4 = true;
3609 		break;
3610 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3611 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3612 		rxmtrl = E1000_RXMTRL_PTP_V1_DELAY_REQ_MESSAGE;
3613 		is_l4 = true;
3614 		break;
3615 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3616 		/* Also time stamps V2 L2 Path Delay Request/Response */
3617 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3618 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3619 		is_l2 = true;
3620 		break;
3621 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3622 		/* Also time stamps V2 L2 Path Delay Request/Response. */
3623 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3624 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3625 		is_l2 = true;
3626 		break;
3627 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3628 		/* Hardware cannot filter just V2 L4 Sync messages;
3629 		 * fall-through to V2 (both L2 and L4) Sync.
3630 		 */
3631 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
3632 		/* Also time stamps V2 Path Delay Request/Response. */
3633 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3634 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3635 		is_l2 = true;
3636 		is_l4 = true;
3637 		break;
3638 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3639 		/* Hardware cannot filter just V2 L4 Delay Request messages;
3640 		 * fall-through to V2 (both L2 and L4) Delay Request.
3641 		 */
3642 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
3643 		/* Also time stamps V2 Path Delay Request/Response. */
3644 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3645 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3646 		is_l2 = true;
3647 		is_l4 = true;
3648 		break;
3649 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
3650 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
3651 		/* Hardware cannot filter just V2 L4 or L2 Event messages;
3652 		 * fall-through to all V2 (both L2 and L4) Events.
3653 		 */
3654 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
3655 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_EVENT_V2;
3656 		config->rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
3657 		is_l2 = true;
3658 		is_l4 = true;
3659 		break;
3660 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3661 		/* For V1, the hardware can only filter Sync messages or
3662 		 * Delay Request messages but not both so fall-through to
3663 		 * time stamp all packets.
3664 		 */
3665 	case HWTSTAMP_FILTER_ALL:
3666 		is_l2 = true;
3667 		is_l4 = true;
3668 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_ALL;
3669 		config->rx_filter = HWTSTAMP_FILTER_ALL;
3670 		break;
3671 	default:
3672 		return -ERANGE;
3673 	}
3674 
3675 	adapter->hwtstamp_config = *config;
3676 
3677 	/* enable/disable Tx h/w time stamping */
3678 	regval = er32(TSYNCTXCTL);
3679 	regval &= ~E1000_TSYNCTXCTL_ENABLED;
3680 	regval |= tsync_tx_ctl;
3681 	ew32(TSYNCTXCTL, regval);
3682 	if ((er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) !=
3683 	    (regval & E1000_TSYNCTXCTL_ENABLED)) {
3684 		e_err("Timesync Tx Control register not set as expected\n");
3685 		return -EAGAIN;
3686 	}
3687 
3688 	/* enable/disable Rx h/w time stamping */
3689 	regval = er32(TSYNCRXCTL);
3690 	regval &= ~(E1000_TSYNCRXCTL_ENABLED | E1000_TSYNCRXCTL_TYPE_MASK);
3691 	regval |= tsync_rx_ctl;
3692 	ew32(TSYNCRXCTL, regval);
3693 	if ((er32(TSYNCRXCTL) & (E1000_TSYNCRXCTL_ENABLED |
3694 				 E1000_TSYNCRXCTL_TYPE_MASK)) !=
3695 	    (regval & (E1000_TSYNCRXCTL_ENABLED |
3696 		       E1000_TSYNCRXCTL_TYPE_MASK))) {
3697 		e_err("Timesync Rx Control register not set as expected\n");
3698 		return -EAGAIN;
3699 	}
3700 
3701 	/* L2: define ethertype filter for time stamped packets */
3702 	if (is_l2)
3703 		rxmtrl |= ETH_P_1588;
3704 
3705 	/* define which PTP packets get time stamped */
3706 	ew32(RXMTRL, rxmtrl);
3707 
3708 	/* Filter by destination port */
3709 	if (is_l4) {
3710 		rxudp = PTP_EV_PORT;
3711 		cpu_to_be16s(&rxudp);
3712 	}
3713 	ew32(RXUDP, rxudp);
3714 
3715 	e1e_flush();
3716 
3717 	/* Clear TSYNCRXCTL_VALID & TSYNCTXCTL_VALID bit */
3718 	er32(RXSTMPH);
3719 	er32(TXSTMPH);
3720 
3721 	return 0;
3722 }
3723 
3724 /**
3725  * e1000_configure - configure the hardware for Rx and Tx
3726  * @adapter: private board structure
3727  **/
3728 static void e1000_configure(struct e1000_adapter *adapter)
3729 {
3730 	struct e1000_ring *rx_ring = adapter->rx_ring;
3731 
3732 	e1000e_set_rx_mode(adapter->netdev);
3733 
3734 	e1000_restore_vlan(adapter);
3735 	e1000_init_manageability_pt(adapter);
3736 
3737 	e1000_configure_tx(adapter);
3738 
3739 	if (adapter->netdev->features & NETIF_F_RXHASH)
3740 		e1000e_setup_rss_hash(adapter);
3741 	e1000_setup_rctl(adapter);
3742 	e1000_configure_rx(adapter);
3743 	adapter->alloc_rx_buf(rx_ring, e1000_desc_unused(rx_ring), GFP_KERNEL);
3744 }
3745 
3746 /**
3747  * e1000e_power_up_phy - restore link in case the phy was powered down
3748  * @adapter: address of board private structure
3749  *
3750  * The phy may be powered down to save power and turn off link when the
3751  * driver is unloaded and wake on lan is not enabled (among others)
3752  * *** this routine MUST be followed by a call to e1000e_reset ***
3753  **/
3754 void e1000e_power_up_phy(struct e1000_adapter *adapter)
3755 {
3756 	if (adapter->hw.phy.ops.power_up)
3757 		adapter->hw.phy.ops.power_up(&adapter->hw);
3758 
3759 	adapter->hw.mac.ops.setup_link(&adapter->hw);
3760 }
3761 
3762 /**
3763  * e1000_power_down_phy - Power down the PHY
3764  *
3765  * Power down the PHY so no link is implied when interface is down.
3766  * The PHY cannot be powered down if management or WoL is active.
3767  */
3768 static void e1000_power_down_phy(struct e1000_adapter *adapter)
3769 {
3770 	if (adapter->hw.phy.ops.power_down)
3771 		adapter->hw.phy.ops.power_down(&adapter->hw);
3772 }
3773 
3774 /**
3775  * e1000_flush_tx_ring - remove all descriptors from the tx_ring
3776  *
3777  * We want to clear all pending descriptors from the TX ring.
3778  * zeroing happens when the HW reads the regs. We  assign the ring itself as
3779  * the data of the next descriptor. We don't care about the data we are about
3780  * to reset the HW.
3781  */
3782 static void e1000_flush_tx_ring(struct e1000_adapter *adapter)
3783 {
3784 	struct e1000_hw *hw = &adapter->hw;
3785 	struct e1000_ring *tx_ring = adapter->tx_ring;
3786 	struct e1000_tx_desc *tx_desc = NULL;
3787 	u32 tdt, tctl, txd_lower = E1000_TXD_CMD_IFCS;
3788 	u16 size = 512;
3789 
3790 	tctl = er32(TCTL);
3791 	ew32(TCTL, tctl | E1000_TCTL_EN);
3792 	tdt = er32(TDT(0));
3793 	BUG_ON(tdt != tx_ring->next_to_use);
3794 	tx_desc =  E1000_TX_DESC(*tx_ring, tx_ring->next_to_use);
3795 	tx_desc->buffer_addr = tx_ring->dma;
3796 
3797 	tx_desc->lower.data = cpu_to_le32(txd_lower | size);
3798 	tx_desc->upper.data = 0;
3799 	/* flush descriptors to memory before notifying the HW */
3800 	wmb();
3801 	tx_ring->next_to_use++;
3802 	if (tx_ring->next_to_use == tx_ring->count)
3803 		tx_ring->next_to_use = 0;
3804 	ew32(TDT(0), tx_ring->next_to_use);
3805 	mmiowb();
3806 	usleep_range(200, 250);
3807 }
3808 
3809 /**
3810  * e1000_flush_rx_ring - remove all descriptors from the rx_ring
3811  *
3812  * Mark all descriptors in the RX ring as consumed and disable the rx ring
3813  */
3814 static void e1000_flush_rx_ring(struct e1000_adapter *adapter)
3815 {
3816 	u32 rctl, rxdctl;
3817 	struct e1000_hw *hw = &adapter->hw;
3818 
3819 	rctl = er32(RCTL);
3820 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3821 	e1e_flush();
3822 	usleep_range(100, 150);
3823 
3824 	rxdctl = er32(RXDCTL(0));
3825 	/* zero the lower 14 bits (prefetch and host thresholds) */
3826 	rxdctl &= 0xffffc000;
3827 
3828 	/* update thresholds: prefetch threshold to 31, host threshold to 1
3829 	 * and make sure the granularity is "descriptors" and not "cache lines"
3830 	 */
3831 	rxdctl |= (0x1F | BIT(8) | E1000_RXDCTL_THRESH_UNIT_DESC);
3832 
3833 	ew32(RXDCTL(0), rxdctl);
3834 	/* momentarily enable the RX ring for the changes to take effect */
3835 	ew32(RCTL, rctl | E1000_RCTL_EN);
3836 	e1e_flush();
3837 	usleep_range(100, 150);
3838 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3839 }
3840 
3841 /**
3842  * e1000_flush_desc_rings - remove all descriptors from the descriptor rings
3843  *
3844  * In i219, the descriptor rings must be emptied before resetting the HW
3845  * or before changing the device state to D3 during runtime (runtime PM).
3846  *
3847  * Failure to do this will cause the HW to enter a unit hang state which can
3848  * only be released by PCI reset on the device
3849  *
3850  */
3851 
3852 static void e1000_flush_desc_rings(struct e1000_adapter *adapter)
3853 {
3854 	u16 hang_state;
3855 	u32 fext_nvm11, tdlen;
3856 	struct e1000_hw *hw = &adapter->hw;
3857 
3858 	/* First, disable MULR fix in FEXTNVM11 */
3859 	fext_nvm11 = er32(FEXTNVM11);
3860 	fext_nvm11 |= E1000_FEXTNVM11_DISABLE_MULR_FIX;
3861 	ew32(FEXTNVM11, fext_nvm11);
3862 	/* do nothing if we're not in faulty state, or if the queue is empty */
3863 	tdlen = er32(TDLEN(0));
3864 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3865 			     &hang_state);
3866 	if (!(hang_state & FLUSH_DESC_REQUIRED) || !tdlen)
3867 		return;
3868 	e1000_flush_tx_ring(adapter);
3869 	/* recheck, maybe the fault is caused by the rx ring */
3870 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3871 			     &hang_state);
3872 	if (hang_state & FLUSH_DESC_REQUIRED)
3873 		e1000_flush_rx_ring(adapter);
3874 }
3875 
3876 /**
3877  * e1000e_systim_reset - reset the timesync registers after a hardware reset
3878  * @adapter: board private structure
3879  *
3880  * When the MAC is reset, all hardware bits for timesync will be reset to the
3881  * default values. This function will restore the settings last in place.
3882  * Since the clock SYSTIME registers are reset, we will simply restore the
3883  * cyclecounter to the kernel real clock time.
3884  **/
3885 static void e1000e_systim_reset(struct e1000_adapter *adapter)
3886 {
3887 	struct ptp_clock_info *info = &adapter->ptp_clock_info;
3888 	struct e1000_hw *hw = &adapter->hw;
3889 	unsigned long flags;
3890 	u32 timinca;
3891 	s32 ret_val;
3892 
3893 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3894 		return;
3895 
3896 	if (info->adjfreq) {
3897 		/* restore the previous ptp frequency delta */
3898 		ret_val = info->adjfreq(info, adapter->ptp_delta);
3899 	} else {
3900 		/* set the default base frequency if no adjustment possible */
3901 		ret_val = e1000e_get_base_timinca(adapter, &timinca);
3902 		if (!ret_val)
3903 			ew32(TIMINCA, timinca);
3904 	}
3905 
3906 	if (ret_val) {
3907 		dev_warn(&adapter->pdev->dev,
3908 			 "Failed to restore TIMINCA clock rate delta: %d\n",
3909 			 ret_val);
3910 		return;
3911 	}
3912 
3913 	/* reset the systim ns time counter */
3914 	spin_lock_irqsave(&adapter->systim_lock, flags);
3915 	timecounter_init(&adapter->tc, &adapter->cc,
3916 			 ktime_to_ns(ktime_get_real()));
3917 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
3918 
3919 	/* restore the previous hwtstamp configuration settings */
3920 	e1000e_config_hwtstamp(adapter, &adapter->hwtstamp_config);
3921 }
3922 
3923 /**
3924  * e1000e_reset - bring the hardware into a known good state
3925  *
3926  * This function boots the hardware and enables some settings that
3927  * require a configuration cycle of the hardware - those cannot be
3928  * set/changed during runtime. After reset the device needs to be
3929  * properly configured for Rx, Tx etc.
3930  */
3931 void e1000e_reset(struct e1000_adapter *adapter)
3932 {
3933 	struct e1000_mac_info *mac = &adapter->hw.mac;
3934 	struct e1000_fc_info *fc = &adapter->hw.fc;
3935 	struct e1000_hw *hw = &adapter->hw;
3936 	u32 tx_space, min_tx_space, min_rx_space;
3937 	u32 pba = adapter->pba;
3938 	u16 hwm;
3939 
3940 	/* reset Packet Buffer Allocation to default */
3941 	ew32(PBA, pba);
3942 
3943 	if (adapter->max_frame_size > (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN)) {
3944 		/* To maintain wire speed transmits, the Tx FIFO should be
3945 		 * large enough to accommodate two full transmit packets,
3946 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
3947 		 * the Rx FIFO should be large enough to accommodate at least
3948 		 * one full receive packet and is similarly rounded up and
3949 		 * expressed in KB.
3950 		 */
3951 		pba = er32(PBA);
3952 		/* upper 16 bits has Tx packet buffer allocation size in KB */
3953 		tx_space = pba >> 16;
3954 		/* lower 16 bits has Rx packet buffer allocation size in KB */
3955 		pba &= 0xffff;
3956 		/* the Tx fifo also stores 16 bytes of information about the Tx
3957 		 * but don't include ethernet FCS because hardware appends it
3958 		 */
3959 		min_tx_space = (adapter->max_frame_size +
3960 				sizeof(struct e1000_tx_desc) - ETH_FCS_LEN) * 2;
3961 		min_tx_space = ALIGN(min_tx_space, 1024);
3962 		min_tx_space >>= 10;
3963 		/* software strips receive CRC, so leave room for it */
3964 		min_rx_space = adapter->max_frame_size;
3965 		min_rx_space = ALIGN(min_rx_space, 1024);
3966 		min_rx_space >>= 10;
3967 
3968 		/* If current Tx allocation is less than the min Tx FIFO size,
3969 		 * and the min Tx FIFO size is less than the current Rx FIFO
3970 		 * allocation, take space away from current Rx allocation
3971 		 */
3972 		if ((tx_space < min_tx_space) &&
3973 		    ((min_tx_space - tx_space) < pba)) {
3974 			pba -= min_tx_space - tx_space;
3975 
3976 			/* if short on Rx space, Rx wins and must trump Tx
3977 			 * adjustment
3978 			 */
3979 			if (pba < min_rx_space)
3980 				pba = min_rx_space;
3981 		}
3982 
3983 		ew32(PBA, pba);
3984 	}
3985 
3986 	/* flow control settings
3987 	 *
3988 	 * The high water mark must be low enough to fit one full frame
3989 	 * (or the size used for early receive) above it in the Rx FIFO.
3990 	 * Set it to the lower of:
3991 	 * - 90% of the Rx FIFO size, and
3992 	 * - the full Rx FIFO size minus one full frame
3993 	 */
3994 	if (adapter->flags & FLAG_DISABLE_FC_PAUSE_TIME)
3995 		fc->pause_time = 0xFFFF;
3996 	else
3997 		fc->pause_time = E1000_FC_PAUSE_TIME;
3998 	fc->send_xon = true;
3999 	fc->current_mode = fc->requested_mode;
4000 
4001 	switch (hw->mac.type) {
4002 	case e1000_ich9lan:
4003 	case e1000_ich10lan:
4004 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4005 			pba = 14;
4006 			ew32(PBA, pba);
4007 			fc->high_water = 0x2800;
4008 			fc->low_water = fc->high_water - 8;
4009 			break;
4010 		}
4011 		/* fall-through */
4012 	default:
4013 		hwm = min(((pba << 10) * 9 / 10),
4014 			  ((pba << 10) - adapter->max_frame_size));
4015 
4016 		fc->high_water = hwm & E1000_FCRTH_RTH;	/* 8-byte granularity */
4017 		fc->low_water = fc->high_water - 8;
4018 		break;
4019 	case e1000_pchlan:
4020 		/* Workaround PCH LOM adapter hangs with certain network
4021 		 * loads.  If hangs persist, try disabling Tx flow control.
4022 		 */
4023 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4024 			fc->high_water = 0x3500;
4025 			fc->low_water = 0x1500;
4026 		} else {
4027 			fc->high_water = 0x5000;
4028 			fc->low_water = 0x3000;
4029 		}
4030 		fc->refresh_time = 0x1000;
4031 		break;
4032 	case e1000_pch2lan:
4033 	case e1000_pch_lpt:
4034 	case e1000_pch_spt:
4035 		fc->refresh_time = 0x0400;
4036 
4037 		if (adapter->netdev->mtu <= ETH_DATA_LEN) {
4038 			fc->high_water = 0x05C20;
4039 			fc->low_water = 0x05048;
4040 			fc->pause_time = 0x0650;
4041 			break;
4042 		}
4043 
4044 		pba = 14;
4045 		ew32(PBA, pba);
4046 		fc->high_water = ((pba << 10) * 9 / 10) & E1000_FCRTH_RTH;
4047 		fc->low_water = ((pba << 10) * 8 / 10) & E1000_FCRTL_RTL;
4048 		break;
4049 	}
4050 
4051 	/* Alignment of Tx data is on an arbitrary byte boundary with the
4052 	 * maximum size per Tx descriptor limited only to the transmit
4053 	 * allocation of the packet buffer minus 96 bytes with an upper
4054 	 * limit of 24KB due to receive synchronization limitations.
4055 	 */
4056 	adapter->tx_fifo_limit = min_t(u32, ((er32(PBA) >> 16) << 10) - 96,
4057 				       24 << 10);
4058 
4059 	/* Disable Adaptive Interrupt Moderation if 2 full packets cannot
4060 	 * fit in receive buffer.
4061 	 */
4062 	if (adapter->itr_setting & 0x3) {
4063 		if ((adapter->max_frame_size * 2) > (pba << 10)) {
4064 			if (!(adapter->flags2 & FLAG2_DISABLE_AIM)) {
4065 				dev_info(&adapter->pdev->dev,
4066 					 "Interrupt Throttle Rate off\n");
4067 				adapter->flags2 |= FLAG2_DISABLE_AIM;
4068 				e1000e_write_itr(adapter, 0);
4069 			}
4070 		} else if (adapter->flags2 & FLAG2_DISABLE_AIM) {
4071 			dev_info(&adapter->pdev->dev,
4072 				 "Interrupt Throttle Rate on\n");
4073 			adapter->flags2 &= ~FLAG2_DISABLE_AIM;
4074 			adapter->itr = 20000;
4075 			e1000e_write_itr(adapter, adapter->itr);
4076 		}
4077 	}
4078 
4079 	if (hw->mac.type == e1000_pch_spt)
4080 		e1000_flush_desc_rings(adapter);
4081 	/* Allow time for pending master requests to run */
4082 	mac->ops.reset_hw(hw);
4083 
4084 	/* For parts with AMT enabled, let the firmware know
4085 	 * that the network interface is in control
4086 	 */
4087 	if (adapter->flags & FLAG_HAS_AMT)
4088 		e1000e_get_hw_control(adapter);
4089 
4090 	ew32(WUC, 0);
4091 
4092 	if (mac->ops.init_hw(hw))
4093 		e_err("Hardware Error\n");
4094 
4095 	e1000_update_mng_vlan(adapter);
4096 
4097 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
4098 	ew32(VET, ETH_P_8021Q);
4099 
4100 	e1000e_reset_adaptive(hw);
4101 
4102 	/* restore systim and hwtstamp settings */
4103 	e1000e_systim_reset(adapter);
4104 
4105 	/* Set EEE advertisement as appropriate */
4106 	if (adapter->flags2 & FLAG2_HAS_EEE) {
4107 		s32 ret_val;
4108 		u16 adv_addr;
4109 
4110 		switch (hw->phy.type) {
4111 		case e1000_phy_82579:
4112 			adv_addr = I82579_EEE_ADVERTISEMENT;
4113 			break;
4114 		case e1000_phy_i217:
4115 			adv_addr = I217_EEE_ADVERTISEMENT;
4116 			break;
4117 		default:
4118 			dev_err(&adapter->pdev->dev,
4119 				"Invalid PHY type setting EEE advertisement\n");
4120 			return;
4121 		}
4122 
4123 		ret_val = hw->phy.ops.acquire(hw);
4124 		if (ret_val) {
4125 			dev_err(&adapter->pdev->dev,
4126 				"EEE advertisement - unable to acquire PHY\n");
4127 			return;
4128 		}
4129 
4130 		e1000_write_emi_reg_locked(hw, adv_addr,
4131 					   hw->dev_spec.ich8lan.eee_disable ?
4132 					   0 : adapter->eee_advert);
4133 
4134 		hw->phy.ops.release(hw);
4135 	}
4136 
4137 	if (!netif_running(adapter->netdev) &&
4138 	    !test_bit(__E1000_TESTING, &adapter->state))
4139 		e1000_power_down_phy(adapter);
4140 
4141 	e1000_get_phy_info(hw);
4142 
4143 	if ((adapter->flags & FLAG_HAS_SMART_POWER_DOWN) &&
4144 	    !(adapter->flags & FLAG_SMART_POWER_DOWN)) {
4145 		u16 phy_data = 0;
4146 		/* speed up time to link by disabling smart power down, ignore
4147 		 * the return value of this function because there is nothing
4148 		 * different we would do if it failed
4149 		 */
4150 		e1e_rphy(hw, IGP02E1000_PHY_POWER_MGMT, &phy_data);
4151 		phy_data &= ~IGP02E1000_PM_SPD;
4152 		e1e_wphy(hw, IGP02E1000_PHY_POWER_MGMT, phy_data);
4153 	}
4154 	if (hw->mac.type == e1000_pch_spt && adapter->int_mode == 0) {
4155 		u32 reg;
4156 
4157 		/* Fextnvm7 @ 0xe4[2] = 1 */
4158 		reg = er32(FEXTNVM7);
4159 		reg |= E1000_FEXTNVM7_SIDE_CLK_UNGATE;
4160 		ew32(FEXTNVM7, reg);
4161 		/* Fextnvm9 @ 0x5bb4[13:12] = 11 */
4162 		reg = er32(FEXTNVM9);
4163 		reg |= E1000_FEXTNVM9_IOSFSB_CLKGATE_DIS |
4164 		       E1000_FEXTNVM9_IOSFSB_CLKREQ_DIS;
4165 		ew32(FEXTNVM9, reg);
4166 	}
4167 
4168 }
4169 
4170 /**
4171  * e1000e_trigger_lsc - trigger an LSC interrupt
4172  * @adapter:
4173  *
4174  * Fire a link status change interrupt to start the watchdog.
4175  **/
4176 static void e1000e_trigger_lsc(struct e1000_adapter *adapter)
4177 {
4178 	struct e1000_hw *hw = &adapter->hw;
4179 
4180 	if (adapter->msix_entries)
4181 		ew32(ICS, E1000_ICS_OTHER);
4182 	else
4183 		ew32(ICS, E1000_ICS_LSC);
4184 }
4185 
4186 void e1000e_up(struct e1000_adapter *adapter)
4187 {
4188 	/* hardware has been reset, we need to reload some things */
4189 	e1000_configure(adapter);
4190 
4191 	clear_bit(__E1000_DOWN, &adapter->state);
4192 
4193 	if (adapter->msix_entries)
4194 		e1000_configure_msix(adapter);
4195 	e1000_irq_enable(adapter);
4196 
4197 	netif_start_queue(adapter->netdev);
4198 
4199 	e1000e_trigger_lsc(adapter);
4200 }
4201 
4202 static void e1000e_flush_descriptors(struct e1000_adapter *adapter)
4203 {
4204 	struct e1000_hw *hw = &adapter->hw;
4205 
4206 	if (!(adapter->flags2 & FLAG2_DMA_BURST))
4207 		return;
4208 
4209 	/* flush pending descriptor writebacks to memory */
4210 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4211 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4212 
4213 	/* execute the writes immediately */
4214 	e1e_flush();
4215 
4216 	/* due to rare timing issues, write to TIDV/RDTR again to ensure the
4217 	 * write is successful
4218 	 */
4219 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4220 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4221 
4222 	/* execute the writes immediately */
4223 	e1e_flush();
4224 }
4225 
4226 static void e1000e_update_stats(struct e1000_adapter *adapter);
4227 
4228 /**
4229  * e1000e_down - quiesce the device and optionally reset the hardware
4230  * @adapter: board private structure
4231  * @reset: boolean flag to reset the hardware or not
4232  */
4233 void e1000e_down(struct e1000_adapter *adapter, bool reset)
4234 {
4235 	struct net_device *netdev = adapter->netdev;
4236 	struct e1000_hw *hw = &adapter->hw;
4237 	u32 tctl, rctl;
4238 
4239 	/* signal that we're down so the interrupt handler does not
4240 	 * reschedule our watchdog timer
4241 	 */
4242 	set_bit(__E1000_DOWN, &adapter->state);
4243 
4244 	netif_carrier_off(netdev);
4245 
4246 	/* disable receives in the hardware */
4247 	rctl = er32(RCTL);
4248 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
4249 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
4250 	/* flush and sleep below */
4251 
4252 	netif_stop_queue(netdev);
4253 
4254 	/* disable transmits in the hardware */
4255 	tctl = er32(TCTL);
4256 	tctl &= ~E1000_TCTL_EN;
4257 	ew32(TCTL, tctl);
4258 
4259 	/* flush both disables and wait for them to finish */
4260 	e1e_flush();
4261 	usleep_range(10000, 20000);
4262 
4263 	e1000_irq_disable(adapter);
4264 
4265 	napi_synchronize(&adapter->napi);
4266 
4267 	del_timer_sync(&adapter->watchdog_timer);
4268 	del_timer_sync(&adapter->phy_info_timer);
4269 
4270 	spin_lock(&adapter->stats64_lock);
4271 	e1000e_update_stats(adapter);
4272 	spin_unlock(&adapter->stats64_lock);
4273 
4274 	e1000e_flush_descriptors(adapter);
4275 
4276 	adapter->link_speed = 0;
4277 	adapter->link_duplex = 0;
4278 
4279 	/* Disable Si errata workaround on PCHx for jumbo frame flow */
4280 	if ((hw->mac.type >= e1000_pch2lan) &&
4281 	    (adapter->netdev->mtu > ETH_DATA_LEN) &&
4282 	    e1000_lv_jumbo_workaround_ich8lan(hw, false))
4283 		e_dbg("failed to disable jumbo frame workaround mode\n");
4284 
4285 	if (!pci_channel_offline(adapter->pdev)) {
4286 		if (reset)
4287 			e1000e_reset(adapter);
4288 		else if (hw->mac.type == e1000_pch_spt)
4289 			e1000_flush_desc_rings(adapter);
4290 	}
4291 	e1000_clean_tx_ring(adapter->tx_ring);
4292 	e1000_clean_rx_ring(adapter->rx_ring);
4293 }
4294 
4295 void e1000e_reinit_locked(struct e1000_adapter *adapter)
4296 {
4297 	might_sleep();
4298 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
4299 		usleep_range(1000, 2000);
4300 	e1000e_down(adapter, true);
4301 	e1000e_up(adapter);
4302 	clear_bit(__E1000_RESETTING, &adapter->state);
4303 }
4304 
4305 /**
4306  * e1000e_sanitize_systim - sanitize raw cycle counter reads
4307  * @hw: pointer to the HW structure
4308  * @systim: time value read, sanitized and returned
4309  *
4310  * Errata for 82574/82583 possible bad bits read from SYSTIMH/L:
4311  * check to see that the time is incrementing at a reasonable
4312  * rate and is a multiple of incvalue.
4313  **/
4314 static u64 e1000e_sanitize_systim(struct e1000_hw *hw, u64 systim)
4315 {
4316 	u64 time_delta, rem, temp;
4317 	u64 systim_next;
4318 	u32 incvalue;
4319 	int i;
4320 
4321 	incvalue = er32(TIMINCA) & E1000_TIMINCA_INCVALUE_MASK;
4322 	for (i = 0; i < E1000_MAX_82574_SYSTIM_REREADS; i++) {
4323 		/* latch SYSTIMH on read of SYSTIML */
4324 		systim_next = (u64)er32(SYSTIML);
4325 		systim_next |= (u64)er32(SYSTIMH) << 32;
4326 
4327 		time_delta = systim_next - systim;
4328 		temp = time_delta;
4329 		/* VMWare users have seen incvalue of zero, don't div / 0 */
4330 		rem = incvalue ? do_div(temp, incvalue) : (time_delta != 0);
4331 
4332 		systim = systim_next;
4333 
4334 		if ((time_delta < E1000_82574_SYSTIM_EPSILON) && (rem == 0))
4335 			break;
4336 	}
4337 
4338 	return systim;
4339 }
4340 
4341 /**
4342  * e1000e_cyclecounter_read - read raw cycle counter (used by time counter)
4343  * @cc: cyclecounter structure
4344  **/
4345 static u64 e1000e_cyclecounter_read(const struct cyclecounter *cc)
4346 {
4347 	struct e1000_adapter *adapter = container_of(cc, struct e1000_adapter,
4348 						     cc);
4349 	struct e1000_hw *hw = &adapter->hw;
4350 	u32 systimel, systimeh;
4351 	u64 systim;
4352 	/* SYSTIMH latching upon SYSTIML read does not work well.
4353 	 * This means that if SYSTIML overflows after we read it but before
4354 	 * we read SYSTIMH, the value of SYSTIMH has been incremented and we
4355 	 * will experience a huge non linear increment in the systime value
4356 	 * to fix that we test for overflow and if true, we re-read systime.
4357 	 */
4358 	systimel = er32(SYSTIML);
4359 	systimeh = er32(SYSTIMH);
4360 	/* Is systimel is so large that overflow is possible? */
4361 	if (systimel >= (u32)0xffffffff - E1000_TIMINCA_INCVALUE_MASK) {
4362 		u32 systimel_2 = er32(SYSTIML);
4363 		if (systimel > systimel_2) {
4364 			/* There was an overflow, read again SYSTIMH, and use
4365 			 * systimel_2
4366 			 */
4367 			systimeh = er32(SYSTIMH);
4368 			systimel = systimel_2;
4369 		}
4370 	}
4371 	systim = (u64)systimel;
4372 	systim |= (u64)systimeh << 32;
4373 
4374 	if (adapter->flags2 & FLAG2_CHECK_SYSTIM_OVERFLOW)
4375 		systim = e1000e_sanitize_systim(hw, systim);
4376 
4377 	return systim;
4378 }
4379 
4380 /**
4381  * e1000_sw_init - Initialize general software structures (struct e1000_adapter)
4382  * @adapter: board private structure to initialize
4383  *
4384  * e1000_sw_init initializes the Adapter private data structure.
4385  * Fields are initialized based on PCI device information and
4386  * OS network device settings (MTU size).
4387  **/
4388 static int e1000_sw_init(struct e1000_adapter *adapter)
4389 {
4390 	struct net_device *netdev = adapter->netdev;
4391 
4392 	adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
4393 	adapter->rx_ps_bsize0 = 128;
4394 	adapter->max_frame_size = netdev->mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
4395 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4396 	adapter->tx_ring_count = E1000_DEFAULT_TXD;
4397 	adapter->rx_ring_count = E1000_DEFAULT_RXD;
4398 
4399 	spin_lock_init(&adapter->stats64_lock);
4400 
4401 	e1000e_set_interrupt_capability(adapter);
4402 
4403 	if (e1000_alloc_queues(adapter))
4404 		return -ENOMEM;
4405 
4406 	/* Setup hardware time stamping cyclecounter */
4407 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
4408 		adapter->cc.read = e1000e_cyclecounter_read;
4409 		adapter->cc.mask = CYCLECOUNTER_MASK(64);
4410 		adapter->cc.mult = 1;
4411 		/* cc.shift set in e1000e_get_base_tininca() */
4412 
4413 		spin_lock_init(&adapter->systim_lock);
4414 		INIT_WORK(&adapter->tx_hwtstamp_work, e1000e_tx_hwtstamp_work);
4415 	}
4416 
4417 	/* Explicitly disable IRQ since the NIC can be in any state. */
4418 	e1000_irq_disable(adapter);
4419 
4420 	set_bit(__E1000_DOWN, &adapter->state);
4421 	return 0;
4422 }
4423 
4424 /**
4425  * e1000_intr_msi_test - Interrupt Handler
4426  * @irq: interrupt number
4427  * @data: pointer to a network interface device structure
4428  **/
4429 static irqreturn_t e1000_intr_msi_test(int __always_unused irq, void *data)
4430 {
4431 	struct net_device *netdev = data;
4432 	struct e1000_adapter *adapter = netdev_priv(netdev);
4433 	struct e1000_hw *hw = &adapter->hw;
4434 	u32 icr = er32(ICR);
4435 
4436 	e_dbg("icr is %08X\n", icr);
4437 	if (icr & E1000_ICR_RXSEQ) {
4438 		adapter->flags &= ~FLAG_MSI_TEST_FAILED;
4439 		/* Force memory writes to complete before acknowledging the
4440 		 * interrupt is handled.
4441 		 */
4442 		wmb();
4443 	}
4444 
4445 	return IRQ_HANDLED;
4446 }
4447 
4448 /**
4449  * e1000_test_msi_interrupt - Returns 0 for successful test
4450  * @adapter: board private struct
4451  *
4452  * code flow taken from tg3.c
4453  **/
4454 static int e1000_test_msi_interrupt(struct e1000_adapter *adapter)
4455 {
4456 	struct net_device *netdev = adapter->netdev;
4457 	struct e1000_hw *hw = &adapter->hw;
4458 	int err;
4459 
4460 	/* poll_enable hasn't been called yet, so don't need disable */
4461 	/* clear any pending events */
4462 	er32(ICR);
4463 
4464 	/* free the real vector and request a test handler */
4465 	e1000_free_irq(adapter);
4466 	e1000e_reset_interrupt_capability(adapter);
4467 
4468 	/* Assume that the test fails, if it succeeds then the test
4469 	 * MSI irq handler will unset this flag
4470 	 */
4471 	adapter->flags |= FLAG_MSI_TEST_FAILED;
4472 
4473 	err = pci_enable_msi(adapter->pdev);
4474 	if (err)
4475 		goto msi_test_failed;
4476 
4477 	err = request_irq(adapter->pdev->irq, e1000_intr_msi_test, 0,
4478 			  netdev->name, netdev);
4479 	if (err) {
4480 		pci_disable_msi(adapter->pdev);
4481 		goto msi_test_failed;
4482 	}
4483 
4484 	/* Force memory writes to complete before enabling and firing an
4485 	 * interrupt.
4486 	 */
4487 	wmb();
4488 
4489 	e1000_irq_enable(adapter);
4490 
4491 	/* fire an unusual interrupt on the test handler */
4492 	ew32(ICS, E1000_ICS_RXSEQ);
4493 	e1e_flush();
4494 	msleep(100);
4495 
4496 	e1000_irq_disable(adapter);
4497 
4498 	rmb();			/* read flags after interrupt has been fired */
4499 
4500 	if (adapter->flags & FLAG_MSI_TEST_FAILED) {
4501 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
4502 		e_info("MSI interrupt test failed, using legacy interrupt.\n");
4503 	} else {
4504 		e_dbg("MSI interrupt test succeeded!\n");
4505 	}
4506 
4507 	free_irq(adapter->pdev->irq, netdev);
4508 	pci_disable_msi(adapter->pdev);
4509 
4510 msi_test_failed:
4511 	e1000e_set_interrupt_capability(adapter);
4512 	return e1000_request_irq(adapter);
4513 }
4514 
4515 /**
4516  * e1000_test_msi - Returns 0 if MSI test succeeds or INTx mode is restored
4517  * @adapter: board private struct
4518  *
4519  * code flow taken from tg3.c, called with e1000 interrupts disabled.
4520  **/
4521 static int e1000_test_msi(struct e1000_adapter *adapter)
4522 {
4523 	int err;
4524 	u16 pci_cmd;
4525 
4526 	if (!(adapter->flags & FLAG_MSI_ENABLED))
4527 		return 0;
4528 
4529 	/* disable SERR in case the MSI write causes a master abort */
4530 	pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4531 	if (pci_cmd & PCI_COMMAND_SERR)
4532 		pci_write_config_word(adapter->pdev, PCI_COMMAND,
4533 				      pci_cmd & ~PCI_COMMAND_SERR);
4534 
4535 	err = e1000_test_msi_interrupt(adapter);
4536 
4537 	/* re-enable SERR */
4538 	if (pci_cmd & PCI_COMMAND_SERR) {
4539 		pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4540 		pci_cmd |= PCI_COMMAND_SERR;
4541 		pci_write_config_word(adapter->pdev, PCI_COMMAND, pci_cmd);
4542 	}
4543 
4544 	return err;
4545 }
4546 
4547 /**
4548  * e1000e_open - Called when a network interface is made active
4549  * @netdev: network interface device structure
4550  *
4551  * Returns 0 on success, negative value on failure
4552  *
4553  * The open entry point is called when a network interface is made
4554  * active by the system (IFF_UP).  At this point all resources needed
4555  * for transmit and receive operations are allocated, the interrupt
4556  * handler is registered with the OS, the watchdog timer is started,
4557  * and the stack is notified that the interface is ready.
4558  **/
4559 int e1000e_open(struct net_device *netdev)
4560 {
4561 	struct e1000_adapter *adapter = netdev_priv(netdev);
4562 	struct e1000_hw *hw = &adapter->hw;
4563 	struct pci_dev *pdev = adapter->pdev;
4564 	int err;
4565 
4566 	/* disallow open during test */
4567 	if (test_bit(__E1000_TESTING, &adapter->state))
4568 		return -EBUSY;
4569 
4570 	pm_runtime_get_sync(&pdev->dev);
4571 
4572 	netif_carrier_off(netdev);
4573 
4574 	/* allocate transmit descriptors */
4575 	err = e1000e_setup_tx_resources(adapter->tx_ring);
4576 	if (err)
4577 		goto err_setup_tx;
4578 
4579 	/* allocate receive descriptors */
4580 	err = e1000e_setup_rx_resources(adapter->rx_ring);
4581 	if (err)
4582 		goto err_setup_rx;
4583 
4584 	/* If AMT is enabled, let the firmware know that the network
4585 	 * interface is now open and reset the part to a known state.
4586 	 */
4587 	if (adapter->flags & FLAG_HAS_AMT) {
4588 		e1000e_get_hw_control(adapter);
4589 		e1000e_reset(adapter);
4590 	}
4591 
4592 	e1000e_power_up_phy(adapter);
4593 
4594 	adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
4595 	if ((adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
4596 		e1000_update_mng_vlan(adapter);
4597 
4598 	/* DMA latency requirement to workaround jumbo issue */
4599 	pm_qos_add_request(&adapter->pm_qos_req, PM_QOS_CPU_DMA_LATENCY,
4600 			   PM_QOS_DEFAULT_VALUE);
4601 
4602 	/* before we allocate an interrupt, we must be ready to handle it.
4603 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4604 	 * as soon as we call pci_request_irq, so we have to setup our
4605 	 * clean_rx handler before we do so.
4606 	 */
4607 	e1000_configure(adapter);
4608 
4609 	err = e1000_request_irq(adapter);
4610 	if (err)
4611 		goto err_req_irq;
4612 
4613 	/* Work around PCIe errata with MSI interrupts causing some chipsets to
4614 	 * ignore e1000e MSI messages, which means we need to test our MSI
4615 	 * interrupt now
4616 	 */
4617 	if (adapter->int_mode != E1000E_INT_MODE_LEGACY) {
4618 		err = e1000_test_msi(adapter);
4619 		if (err) {
4620 			e_err("Interrupt allocation failed\n");
4621 			goto err_req_irq;
4622 		}
4623 	}
4624 
4625 	/* From here on the code is the same as e1000e_up() */
4626 	clear_bit(__E1000_DOWN, &adapter->state);
4627 
4628 	napi_enable(&adapter->napi);
4629 
4630 	e1000_irq_enable(adapter);
4631 
4632 	adapter->tx_hang_recheck = false;
4633 	netif_start_queue(netdev);
4634 
4635 	hw->mac.get_link_status = true;
4636 	pm_runtime_put(&pdev->dev);
4637 
4638 	e1000e_trigger_lsc(adapter);
4639 
4640 	return 0;
4641 
4642 err_req_irq:
4643 	pm_qos_remove_request(&adapter->pm_qos_req);
4644 	e1000e_release_hw_control(adapter);
4645 	e1000_power_down_phy(adapter);
4646 	e1000e_free_rx_resources(adapter->rx_ring);
4647 err_setup_rx:
4648 	e1000e_free_tx_resources(adapter->tx_ring);
4649 err_setup_tx:
4650 	e1000e_reset(adapter);
4651 	pm_runtime_put_sync(&pdev->dev);
4652 
4653 	return err;
4654 }
4655 
4656 /**
4657  * e1000e_close - Disables a network interface
4658  * @netdev: network interface device structure
4659  *
4660  * Returns 0, this is not allowed to fail
4661  *
4662  * The close entry point is called when an interface is de-activated
4663  * by the OS.  The hardware is still under the drivers control, but
4664  * needs to be disabled.  A global MAC reset is issued to stop the
4665  * hardware, and all transmit and receive resources are freed.
4666  **/
4667 int e1000e_close(struct net_device *netdev)
4668 {
4669 	struct e1000_adapter *adapter = netdev_priv(netdev);
4670 	struct pci_dev *pdev = adapter->pdev;
4671 	int count = E1000_CHECK_RESET_COUNT;
4672 
4673 	while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
4674 		usleep_range(10000, 20000);
4675 
4676 	WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
4677 
4678 	pm_runtime_get_sync(&pdev->dev);
4679 
4680 	if (!test_bit(__E1000_DOWN, &adapter->state)) {
4681 		e1000e_down(adapter, true);
4682 		e1000_free_irq(adapter);
4683 
4684 		/* Link status message must follow this format */
4685 		pr_info("%s NIC Link is Down\n", adapter->netdev->name);
4686 	}
4687 
4688 	napi_disable(&adapter->napi);
4689 
4690 	e1000e_free_tx_resources(adapter->tx_ring);
4691 	e1000e_free_rx_resources(adapter->rx_ring);
4692 
4693 	/* kill manageability vlan ID if supported, but not if a vlan with
4694 	 * the same ID is registered on the host OS (let 8021q kill it)
4695 	 */
4696 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN)
4697 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
4698 				       adapter->mng_vlan_id);
4699 
4700 	/* If AMT is enabled, let the firmware know that the network
4701 	 * interface is now closed
4702 	 */
4703 	if ((adapter->flags & FLAG_HAS_AMT) &&
4704 	    !test_bit(__E1000_TESTING, &adapter->state))
4705 		e1000e_release_hw_control(adapter);
4706 
4707 	pm_qos_remove_request(&adapter->pm_qos_req);
4708 
4709 	pm_runtime_put_sync(&pdev->dev);
4710 
4711 	return 0;
4712 }
4713 
4714 /**
4715  * e1000_set_mac - Change the Ethernet Address of the NIC
4716  * @netdev: network interface device structure
4717  * @p: pointer to an address structure
4718  *
4719  * Returns 0 on success, negative on failure
4720  **/
4721 static int e1000_set_mac(struct net_device *netdev, void *p)
4722 {
4723 	struct e1000_adapter *adapter = netdev_priv(netdev);
4724 	struct e1000_hw *hw = &adapter->hw;
4725 	struct sockaddr *addr = p;
4726 
4727 	if (!is_valid_ether_addr(addr->sa_data))
4728 		return -EADDRNOTAVAIL;
4729 
4730 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
4731 	memcpy(adapter->hw.mac.addr, addr->sa_data, netdev->addr_len);
4732 
4733 	hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
4734 
4735 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA) {
4736 		/* activate the work around */
4737 		e1000e_set_laa_state_82571(&adapter->hw, 1);
4738 
4739 		/* Hold a copy of the LAA in RAR[14] This is done so that
4740 		 * between the time RAR[0] gets clobbered  and the time it
4741 		 * gets fixed (in e1000_watchdog), the actual LAA is in one
4742 		 * of the RARs and no incoming packets directed to this port
4743 		 * are dropped. Eventually the LAA will be in RAR[0] and
4744 		 * RAR[14]
4745 		 */
4746 		hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr,
4747 				    adapter->hw.mac.rar_entry_count - 1);
4748 	}
4749 
4750 	return 0;
4751 }
4752 
4753 /**
4754  * e1000e_update_phy_task - work thread to update phy
4755  * @work: pointer to our work struct
4756  *
4757  * this worker thread exists because we must acquire a
4758  * semaphore to read the phy, which we could msleep while
4759  * waiting for it, and we can't msleep in a timer.
4760  **/
4761 static void e1000e_update_phy_task(struct work_struct *work)
4762 {
4763 	struct e1000_adapter *adapter = container_of(work,
4764 						     struct e1000_adapter,
4765 						     update_phy_task);
4766 	struct e1000_hw *hw = &adapter->hw;
4767 
4768 	if (test_bit(__E1000_DOWN, &adapter->state))
4769 		return;
4770 
4771 	e1000_get_phy_info(hw);
4772 
4773 	/* Enable EEE on 82579 after link up */
4774 	if (hw->phy.type >= e1000_phy_82579)
4775 		e1000_set_eee_pchlan(hw);
4776 }
4777 
4778 /**
4779  * e1000_update_phy_info - timre call-back to update PHY info
4780  * @data: pointer to adapter cast into an unsigned long
4781  *
4782  * Need to wait a few seconds after link up to get diagnostic information from
4783  * the phy
4784  **/
4785 static void e1000_update_phy_info(unsigned long data)
4786 {
4787 	struct e1000_adapter *adapter = (struct e1000_adapter *)data;
4788 
4789 	if (test_bit(__E1000_DOWN, &adapter->state))
4790 		return;
4791 
4792 	schedule_work(&adapter->update_phy_task);
4793 }
4794 
4795 /**
4796  * e1000e_update_phy_stats - Update the PHY statistics counters
4797  * @adapter: board private structure
4798  *
4799  * Read/clear the upper 16-bit PHY registers and read/accumulate lower
4800  **/
4801 static void e1000e_update_phy_stats(struct e1000_adapter *adapter)
4802 {
4803 	struct e1000_hw *hw = &adapter->hw;
4804 	s32 ret_val;
4805 	u16 phy_data;
4806 
4807 	ret_val = hw->phy.ops.acquire(hw);
4808 	if (ret_val)
4809 		return;
4810 
4811 	/* A page set is expensive so check if already on desired page.
4812 	 * If not, set to the page with the PHY status registers.
4813 	 */
4814 	hw->phy.addr = 1;
4815 	ret_val = e1000e_read_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4816 					   &phy_data);
4817 	if (ret_val)
4818 		goto release;
4819 	if (phy_data != (HV_STATS_PAGE << IGP_PAGE_SHIFT)) {
4820 		ret_val = hw->phy.ops.set_page(hw,
4821 					       HV_STATS_PAGE << IGP_PAGE_SHIFT);
4822 		if (ret_val)
4823 			goto release;
4824 	}
4825 
4826 	/* Single Collision Count */
4827 	hw->phy.ops.read_reg_page(hw, HV_SCC_UPPER, &phy_data);
4828 	ret_val = hw->phy.ops.read_reg_page(hw, HV_SCC_LOWER, &phy_data);
4829 	if (!ret_val)
4830 		adapter->stats.scc += phy_data;
4831 
4832 	/* Excessive Collision Count */
4833 	hw->phy.ops.read_reg_page(hw, HV_ECOL_UPPER, &phy_data);
4834 	ret_val = hw->phy.ops.read_reg_page(hw, HV_ECOL_LOWER, &phy_data);
4835 	if (!ret_val)
4836 		adapter->stats.ecol += phy_data;
4837 
4838 	/* Multiple Collision Count */
4839 	hw->phy.ops.read_reg_page(hw, HV_MCC_UPPER, &phy_data);
4840 	ret_val = hw->phy.ops.read_reg_page(hw, HV_MCC_LOWER, &phy_data);
4841 	if (!ret_val)
4842 		adapter->stats.mcc += phy_data;
4843 
4844 	/* Late Collision Count */
4845 	hw->phy.ops.read_reg_page(hw, HV_LATECOL_UPPER, &phy_data);
4846 	ret_val = hw->phy.ops.read_reg_page(hw, HV_LATECOL_LOWER, &phy_data);
4847 	if (!ret_val)
4848 		adapter->stats.latecol += phy_data;
4849 
4850 	/* Collision Count - also used for adaptive IFS */
4851 	hw->phy.ops.read_reg_page(hw, HV_COLC_UPPER, &phy_data);
4852 	ret_val = hw->phy.ops.read_reg_page(hw, HV_COLC_LOWER, &phy_data);
4853 	if (!ret_val)
4854 		hw->mac.collision_delta = phy_data;
4855 
4856 	/* Defer Count */
4857 	hw->phy.ops.read_reg_page(hw, HV_DC_UPPER, &phy_data);
4858 	ret_val = hw->phy.ops.read_reg_page(hw, HV_DC_LOWER, &phy_data);
4859 	if (!ret_val)
4860 		adapter->stats.dc += phy_data;
4861 
4862 	/* Transmit with no CRS */
4863 	hw->phy.ops.read_reg_page(hw, HV_TNCRS_UPPER, &phy_data);
4864 	ret_val = hw->phy.ops.read_reg_page(hw, HV_TNCRS_LOWER, &phy_data);
4865 	if (!ret_val)
4866 		adapter->stats.tncrs += phy_data;
4867 
4868 release:
4869 	hw->phy.ops.release(hw);
4870 }
4871 
4872 /**
4873  * e1000e_update_stats - Update the board statistics counters
4874  * @adapter: board private structure
4875  **/
4876 static void e1000e_update_stats(struct e1000_adapter *adapter)
4877 {
4878 	struct net_device *netdev = adapter->netdev;
4879 	struct e1000_hw *hw = &adapter->hw;
4880 	struct pci_dev *pdev = adapter->pdev;
4881 
4882 	/* Prevent stats update while adapter is being reset, or if the pci
4883 	 * connection is down.
4884 	 */
4885 	if (adapter->link_speed == 0)
4886 		return;
4887 	if (pci_channel_offline(pdev))
4888 		return;
4889 
4890 	adapter->stats.crcerrs += er32(CRCERRS);
4891 	adapter->stats.gprc += er32(GPRC);
4892 	adapter->stats.gorc += er32(GORCL);
4893 	er32(GORCH);		/* Clear gorc */
4894 	adapter->stats.bprc += er32(BPRC);
4895 	adapter->stats.mprc += er32(MPRC);
4896 	adapter->stats.roc += er32(ROC);
4897 
4898 	adapter->stats.mpc += er32(MPC);
4899 
4900 	/* Half-duplex statistics */
4901 	if (adapter->link_duplex == HALF_DUPLEX) {
4902 		if (adapter->flags2 & FLAG2_HAS_PHY_STATS) {
4903 			e1000e_update_phy_stats(adapter);
4904 		} else {
4905 			adapter->stats.scc += er32(SCC);
4906 			adapter->stats.ecol += er32(ECOL);
4907 			adapter->stats.mcc += er32(MCC);
4908 			adapter->stats.latecol += er32(LATECOL);
4909 			adapter->stats.dc += er32(DC);
4910 
4911 			hw->mac.collision_delta = er32(COLC);
4912 
4913 			if ((hw->mac.type != e1000_82574) &&
4914 			    (hw->mac.type != e1000_82583))
4915 				adapter->stats.tncrs += er32(TNCRS);
4916 		}
4917 		adapter->stats.colc += hw->mac.collision_delta;
4918 	}
4919 
4920 	adapter->stats.xonrxc += er32(XONRXC);
4921 	adapter->stats.xontxc += er32(XONTXC);
4922 	adapter->stats.xoffrxc += er32(XOFFRXC);
4923 	adapter->stats.xofftxc += er32(XOFFTXC);
4924 	adapter->stats.gptc += er32(GPTC);
4925 	adapter->stats.gotc += er32(GOTCL);
4926 	er32(GOTCH);		/* Clear gotc */
4927 	adapter->stats.rnbc += er32(RNBC);
4928 	adapter->stats.ruc += er32(RUC);
4929 
4930 	adapter->stats.mptc += er32(MPTC);
4931 	adapter->stats.bptc += er32(BPTC);
4932 
4933 	/* used for adaptive IFS */
4934 
4935 	hw->mac.tx_packet_delta = er32(TPT);
4936 	adapter->stats.tpt += hw->mac.tx_packet_delta;
4937 
4938 	adapter->stats.algnerrc += er32(ALGNERRC);
4939 	adapter->stats.rxerrc += er32(RXERRC);
4940 	adapter->stats.cexterr += er32(CEXTERR);
4941 	adapter->stats.tsctc += er32(TSCTC);
4942 	adapter->stats.tsctfc += er32(TSCTFC);
4943 
4944 	/* Fill out the OS statistics structure */
4945 	netdev->stats.multicast = adapter->stats.mprc;
4946 	netdev->stats.collisions = adapter->stats.colc;
4947 
4948 	/* Rx Errors */
4949 
4950 	/* RLEC on some newer hardware can be incorrect so build
4951 	 * our own version based on RUC and ROC
4952 	 */
4953 	netdev->stats.rx_errors = adapter->stats.rxerrc +
4954 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
4955 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
4956 	netdev->stats.rx_length_errors = adapter->stats.ruc +
4957 	    adapter->stats.roc;
4958 	netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
4959 	netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
4960 	netdev->stats.rx_missed_errors = adapter->stats.mpc;
4961 
4962 	/* Tx Errors */
4963 	netdev->stats.tx_errors = adapter->stats.ecol + adapter->stats.latecol;
4964 	netdev->stats.tx_aborted_errors = adapter->stats.ecol;
4965 	netdev->stats.tx_window_errors = adapter->stats.latecol;
4966 	netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
4967 
4968 	/* Tx Dropped needs to be maintained elsewhere */
4969 
4970 	/* Management Stats */
4971 	adapter->stats.mgptc += er32(MGTPTC);
4972 	adapter->stats.mgprc += er32(MGTPRC);
4973 	adapter->stats.mgpdc += er32(MGTPDC);
4974 
4975 	/* Correctable ECC Errors */
4976 	if ((hw->mac.type == e1000_pch_lpt) ||
4977 	    (hw->mac.type == e1000_pch_spt)) {
4978 		u32 pbeccsts = er32(PBECCSTS);
4979 
4980 		adapter->corr_errors +=
4981 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
4982 		adapter->uncorr_errors +=
4983 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
4984 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
4985 	}
4986 }
4987 
4988 /**
4989  * e1000_phy_read_status - Update the PHY register status snapshot
4990  * @adapter: board private structure
4991  **/
4992 static void e1000_phy_read_status(struct e1000_adapter *adapter)
4993 {
4994 	struct e1000_hw *hw = &adapter->hw;
4995 	struct e1000_phy_regs *phy = &adapter->phy_regs;
4996 
4997 	if (!pm_runtime_suspended((&adapter->pdev->dev)->parent) &&
4998 	    (er32(STATUS) & E1000_STATUS_LU) &&
4999 	    (adapter->hw.phy.media_type == e1000_media_type_copper)) {
5000 		int ret_val;
5001 
5002 		ret_val = e1e_rphy(hw, MII_BMCR, &phy->bmcr);
5003 		ret_val |= e1e_rphy(hw, MII_BMSR, &phy->bmsr);
5004 		ret_val |= e1e_rphy(hw, MII_ADVERTISE, &phy->advertise);
5005 		ret_val |= e1e_rphy(hw, MII_LPA, &phy->lpa);
5006 		ret_val |= e1e_rphy(hw, MII_EXPANSION, &phy->expansion);
5007 		ret_val |= e1e_rphy(hw, MII_CTRL1000, &phy->ctrl1000);
5008 		ret_val |= e1e_rphy(hw, MII_STAT1000, &phy->stat1000);
5009 		ret_val |= e1e_rphy(hw, MII_ESTATUS, &phy->estatus);
5010 		if (ret_val)
5011 			e_warn("Error reading PHY register\n");
5012 	} else {
5013 		/* Do not read PHY registers if link is not up
5014 		 * Set values to typical power-on defaults
5015 		 */
5016 		phy->bmcr = (BMCR_SPEED1000 | BMCR_ANENABLE | BMCR_FULLDPLX);
5017 		phy->bmsr = (BMSR_100FULL | BMSR_100HALF | BMSR_10FULL |
5018 			     BMSR_10HALF | BMSR_ESTATEN | BMSR_ANEGCAPABLE |
5019 			     BMSR_ERCAP);
5020 		phy->advertise = (ADVERTISE_PAUSE_ASYM | ADVERTISE_PAUSE_CAP |
5021 				  ADVERTISE_ALL | ADVERTISE_CSMA);
5022 		phy->lpa = 0;
5023 		phy->expansion = EXPANSION_ENABLENPAGE;
5024 		phy->ctrl1000 = ADVERTISE_1000FULL;
5025 		phy->stat1000 = 0;
5026 		phy->estatus = (ESTATUS_1000_TFULL | ESTATUS_1000_THALF);
5027 	}
5028 }
5029 
5030 static void e1000_print_link_info(struct e1000_adapter *adapter)
5031 {
5032 	struct e1000_hw *hw = &adapter->hw;
5033 	u32 ctrl = er32(CTRL);
5034 
5035 	/* Link status message must follow this format for user tools */
5036 	pr_info("%s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5037 		adapter->netdev->name, adapter->link_speed,
5038 		adapter->link_duplex == FULL_DUPLEX ? "Full" : "Half",
5039 		(ctrl & E1000_CTRL_TFCE) && (ctrl & E1000_CTRL_RFCE) ? "Rx/Tx" :
5040 		(ctrl & E1000_CTRL_RFCE) ? "Rx" :
5041 		(ctrl & E1000_CTRL_TFCE) ? "Tx" : "None");
5042 }
5043 
5044 static bool e1000e_has_link(struct e1000_adapter *adapter)
5045 {
5046 	struct e1000_hw *hw = &adapter->hw;
5047 	bool link_active = false;
5048 	s32 ret_val = 0;
5049 
5050 	/* get_link_status is set on LSC (link status) interrupt or
5051 	 * Rx sequence error interrupt.  get_link_status will stay
5052 	 * false until the check_for_link establishes link
5053 	 * for copper adapters ONLY
5054 	 */
5055 	switch (hw->phy.media_type) {
5056 	case e1000_media_type_copper:
5057 		if (hw->mac.get_link_status) {
5058 			ret_val = hw->mac.ops.check_for_link(hw);
5059 			link_active = !hw->mac.get_link_status;
5060 		} else {
5061 			link_active = true;
5062 		}
5063 		break;
5064 	case e1000_media_type_fiber:
5065 		ret_val = hw->mac.ops.check_for_link(hw);
5066 		link_active = !!(er32(STATUS) & E1000_STATUS_LU);
5067 		break;
5068 	case e1000_media_type_internal_serdes:
5069 		ret_val = hw->mac.ops.check_for_link(hw);
5070 		link_active = adapter->hw.mac.serdes_has_link;
5071 		break;
5072 	default:
5073 	case e1000_media_type_unknown:
5074 		break;
5075 	}
5076 
5077 	if ((ret_val == E1000_ERR_PHY) && (hw->phy.type == e1000_phy_igp_3) &&
5078 	    (er32(CTRL) & E1000_PHY_CTRL_GBE_DISABLE)) {
5079 		/* See e1000_kmrn_lock_loss_workaround_ich8lan() */
5080 		e_info("Gigabit has been disabled, downgrading speed\n");
5081 	}
5082 
5083 	return link_active;
5084 }
5085 
5086 static void e1000e_enable_receives(struct e1000_adapter *adapter)
5087 {
5088 	/* make sure the receive unit is started */
5089 	if ((adapter->flags & FLAG_RX_NEEDS_RESTART) &&
5090 	    (adapter->flags & FLAG_RESTART_NOW)) {
5091 		struct e1000_hw *hw = &adapter->hw;
5092 		u32 rctl = er32(RCTL);
5093 
5094 		ew32(RCTL, rctl | E1000_RCTL_EN);
5095 		adapter->flags &= ~FLAG_RESTART_NOW;
5096 	}
5097 }
5098 
5099 static void e1000e_check_82574_phy_workaround(struct e1000_adapter *adapter)
5100 {
5101 	struct e1000_hw *hw = &adapter->hw;
5102 
5103 	/* With 82574 controllers, PHY needs to be checked periodically
5104 	 * for hung state and reset, if two calls return true
5105 	 */
5106 	if (e1000_check_phy_82574(hw))
5107 		adapter->phy_hang_count++;
5108 	else
5109 		adapter->phy_hang_count = 0;
5110 
5111 	if (adapter->phy_hang_count > 1) {
5112 		adapter->phy_hang_count = 0;
5113 		e_dbg("PHY appears hung - resetting\n");
5114 		schedule_work(&adapter->reset_task);
5115 	}
5116 }
5117 
5118 /**
5119  * e1000_watchdog - Timer Call-back
5120  * @data: pointer to adapter cast into an unsigned long
5121  **/
5122 static void e1000_watchdog(unsigned long data)
5123 {
5124 	struct e1000_adapter *adapter = (struct e1000_adapter *)data;
5125 
5126 	/* Do the rest outside of interrupt context */
5127 	schedule_work(&adapter->watchdog_task);
5128 
5129 	/* TODO: make this use queue_delayed_work() */
5130 }
5131 
5132 static void e1000_watchdog_task(struct work_struct *work)
5133 {
5134 	struct e1000_adapter *adapter = container_of(work,
5135 						     struct e1000_adapter,
5136 						     watchdog_task);
5137 	struct net_device *netdev = adapter->netdev;
5138 	struct e1000_mac_info *mac = &adapter->hw.mac;
5139 	struct e1000_phy_info *phy = &adapter->hw.phy;
5140 	struct e1000_ring *tx_ring = adapter->tx_ring;
5141 	struct e1000_hw *hw = &adapter->hw;
5142 	u32 link, tctl;
5143 
5144 	if (test_bit(__E1000_DOWN, &adapter->state))
5145 		return;
5146 
5147 	link = e1000e_has_link(adapter);
5148 	if ((netif_carrier_ok(netdev)) && link) {
5149 		/* Cancel scheduled suspend requests. */
5150 		pm_runtime_resume(netdev->dev.parent);
5151 
5152 		e1000e_enable_receives(adapter);
5153 		goto link_up;
5154 	}
5155 
5156 	if ((e1000e_enable_tx_pkt_filtering(hw)) &&
5157 	    (adapter->mng_vlan_id != adapter->hw.mng_cookie.vlan_id))
5158 		e1000_update_mng_vlan(adapter);
5159 
5160 	if (link) {
5161 		if (!netif_carrier_ok(netdev)) {
5162 			bool txb2b = true;
5163 
5164 			/* Cancel scheduled suspend requests. */
5165 			pm_runtime_resume(netdev->dev.parent);
5166 
5167 			/* update snapshot of PHY registers on LSC */
5168 			e1000_phy_read_status(adapter);
5169 			mac->ops.get_link_up_info(&adapter->hw,
5170 						  &adapter->link_speed,
5171 						  &adapter->link_duplex);
5172 			e1000_print_link_info(adapter);
5173 
5174 			/* check if SmartSpeed worked */
5175 			e1000e_check_downshift(hw);
5176 			if (phy->speed_downgraded)
5177 				netdev_warn(netdev,
5178 					    "Link Speed was downgraded by SmartSpeed\n");
5179 
5180 			/* On supported PHYs, check for duplex mismatch only
5181 			 * if link has autonegotiated at 10/100 half
5182 			 */
5183 			if ((hw->phy.type == e1000_phy_igp_3 ||
5184 			     hw->phy.type == e1000_phy_bm) &&
5185 			    hw->mac.autoneg &&
5186 			    (adapter->link_speed == SPEED_10 ||
5187 			     adapter->link_speed == SPEED_100) &&
5188 			    (adapter->link_duplex == HALF_DUPLEX)) {
5189 				u16 autoneg_exp;
5190 
5191 				e1e_rphy(hw, MII_EXPANSION, &autoneg_exp);
5192 
5193 				if (!(autoneg_exp & EXPANSION_NWAY))
5194 					e_info("Autonegotiated half duplex but link partner cannot autoneg.  Try forcing full duplex if link gets many collisions.\n");
5195 			}
5196 
5197 			/* adjust timeout factor according to speed/duplex */
5198 			adapter->tx_timeout_factor = 1;
5199 			switch (adapter->link_speed) {
5200 			case SPEED_10:
5201 				txb2b = false;
5202 				adapter->tx_timeout_factor = 16;
5203 				break;
5204 			case SPEED_100:
5205 				txb2b = false;
5206 				adapter->tx_timeout_factor = 10;
5207 				break;
5208 			}
5209 
5210 			/* workaround: re-program speed mode bit after
5211 			 * link-up event
5212 			 */
5213 			if ((adapter->flags & FLAG_TARC_SPEED_MODE_BIT) &&
5214 			    !txb2b) {
5215 				u32 tarc0;
5216 
5217 				tarc0 = er32(TARC(0));
5218 				tarc0 &= ~SPEED_MODE_BIT;
5219 				ew32(TARC(0), tarc0);
5220 			}
5221 
5222 			/* disable TSO for pcie and 10/100 speeds, to avoid
5223 			 * some hardware issues
5224 			 */
5225 			if (!(adapter->flags & FLAG_TSO_FORCE)) {
5226 				switch (adapter->link_speed) {
5227 				case SPEED_10:
5228 				case SPEED_100:
5229 					e_info("10/100 speed: disabling TSO\n");
5230 					netdev->features &= ~NETIF_F_TSO;
5231 					netdev->features &= ~NETIF_F_TSO6;
5232 					break;
5233 				case SPEED_1000:
5234 					netdev->features |= NETIF_F_TSO;
5235 					netdev->features |= NETIF_F_TSO6;
5236 					break;
5237 				default:
5238 					/* oops */
5239 					break;
5240 				}
5241 			}
5242 
5243 			/* enable transmits in the hardware, need to do this
5244 			 * after setting TARC(0)
5245 			 */
5246 			tctl = er32(TCTL);
5247 			tctl |= E1000_TCTL_EN;
5248 			ew32(TCTL, tctl);
5249 
5250 			/* Perform any post-link-up configuration before
5251 			 * reporting link up.
5252 			 */
5253 			if (phy->ops.cfg_on_link_up)
5254 				phy->ops.cfg_on_link_up(hw);
5255 
5256 			netif_carrier_on(netdev);
5257 
5258 			if (!test_bit(__E1000_DOWN, &adapter->state))
5259 				mod_timer(&adapter->phy_info_timer,
5260 					  round_jiffies(jiffies + 2 * HZ));
5261 		}
5262 	} else {
5263 		if (netif_carrier_ok(netdev)) {
5264 			adapter->link_speed = 0;
5265 			adapter->link_duplex = 0;
5266 			/* Link status message must follow this format */
5267 			pr_info("%s NIC Link is Down\n", adapter->netdev->name);
5268 			netif_carrier_off(netdev);
5269 			if (!test_bit(__E1000_DOWN, &adapter->state))
5270 				mod_timer(&adapter->phy_info_timer,
5271 					  round_jiffies(jiffies + 2 * HZ));
5272 
5273 			/* 8000ES2LAN requires a Rx packet buffer work-around
5274 			 * on link down event; reset the controller to flush
5275 			 * the Rx packet buffer.
5276 			 */
5277 			if (adapter->flags & FLAG_RX_NEEDS_RESTART)
5278 				adapter->flags |= FLAG_RESTART_NOW;
5279 			else
5280 				pm_schedule_suspend(netdev->dev.parent,
5281 						    LINK_TIMEOUT);
5282 		}
5283 	}
5284 
5285 link_up:
5286 	spin_lock(&adapter->stats64_lock);
5287 	e1000e_update_stats(adapter);
5288 
5289 	mac->tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
5290 	adapter->tpt_old = adapter->stats.tpt;
5291 	mac->collision_delta = adapter->stats.colc - adapter->colc_old;
5292 	adapter->colc_old = adapter->stats.colc;
5293 
5294 	adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
5295 	adapter->gorc_old = adapter->stats.gorc;
5296 	adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
5297 	adapter->gotc_old = adapter->stats.gotc;
5298 	spin_unlock(&adapter->stats64_lock);
5299 
5300 	/* If the link is lost the controller stops DMA, but
5301 	 * if there is queued Tx work it cannot be done.  So
5302 	 * reset the controller to flush the Tx packet buffers.
5303 	 */
5304 	if (!netif_carrier_ok(netdev) &&
5305 	    (e1000_desc_unused(tx_ring) + 1 < tx_ring->count))
5306 		adapter->flags |= FLAG_RESTART_NOW;
5307 
5308 	/* If reset is necessary, do it outside of interrupt context. */
5309 	if (adapter->flags & FLAG_RESTART_NOW) {
5310 		schedule_work(&adapter->reset_task);
5311 		/* return immediately since reset is imminent */
5312 		return;
5313 	}
5314 
5315 	e1000e_update_adaptive(&adapter->hw);
5316 
5317 	/* Simple mode for Interrupt Throttle Rate (ITR) */
5318 	if (adapter->itr_setting == 4) {
5319 		/* Symmetric Tx/Rx gets a reduced ITR=2000;
5320 		 * Total asymmetrical Tx or Rx gets ITR=8000;
5321 		 * everyone else is between 2000-8000.
5322 		 */
5323 		u32 goc = (adapter->gotc + adapter->gorc) / 10000;
5324 		u32 dif = (adapter->gotc > adapter->gorc ?
5325 			   adapter->gotc - adapter->gorc :
5326 			   adapter->gorc - adapter->gotc) / 10000;
5327 		u32 itr = goc > 0 ? (dif * 6000 / goc + 2000) : 8000;
5328 
5329 		e1000e_write_itr(adapter, itr);
5330 	}
5331 
5332 	/* Cause software interrupt to ensure Rx ring is cleaned */
5333 	if (adapter->msix_entries)
5334 		ew32(ICS, adapter->rx_ring->ims_val);
5335 	else
5336 		ew32(ICS, E1000_ICS_RXDMT0);
5337 
5338 	/* flush pending descriptors to memory before detecting Tx hang */
5339 	e1000e_flush_descriptors(adapter);
5340 
5341 	/* Force detection of hung controller every watchdog period */
5342 	adapter->detect_tx_hung = true;
5343 
5344 	/* With 82571 controllers, LAA may be overwritten due to controller
5345 	 * reset from the other port. Set the appropriate LAA in RAR[0]
5346 	 */
5347 	if (e1000e_get_laa_state_82571(hw))
5348 		hw->mac.ops.rar_set(hw, adapter->hw.mac.addr, 0);
5349 
5350 	if (adapter->flags2 & FLAG2_CHECK_PHY_HANG)
5351 		e1000e_check_82574_phy_workaround(adapter);
5352 
5353 	/* Clear valid timestamp stuck in RXSTMPL/H due to a Rx error */
5354 	if (adapter->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE) {
5355 		if ((adapter->flags2 & FLAG2_CHECK_RX_HWTSTAMP) &&
5356 		    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID)) {
5357 			er32(RXSTMPH);
5358 			adapter->rx_hwtstamp_cleared++;
5359 		} else {
5360 			adapter->flags2 |= FLAG2_CHECK_RX_HWTSTAMP;
5361 		}
5362 	}
5363 
5364 	/* Reset the timer */
5365 	if (!test_bit(__E1000_DOWN, &adapter->state))
5366 		mod_timer(&adapter->watchdog_timer,
5367 			  round_jiffies(jiffies + 2 * HZ));
5368 }
5369 
5370 #define E1000_TX_FLAGS_CSUM		0x00000001
5371 #define E1000_TX_FLAGS_VLAN		0x00000002
5372 #define E1000_TX_FLAGS_TSO		0x00000004
5373 #define E1000_TX_FLAGS_IPV4		0x00000008
5374 #define E1000_TX_FLAGS_NO_FCS		0x00000010
5375 #define E1000_TX_FLAGS_HWTSTAMP		0x00000020
5376 #define E1000_TX_FLAGS_VLAN_MASK	0xffff0000
5377 #define E1000_TX_FLAGS_VLAN_SHIFT	16
5378 
5379 static int e1000_tso(struct e1000_ring *tx_ring, struct sk_buff *skb,
5380 		     __be16 protocol)
5381 {
5382 	struct e1000_context_desc *context_desc;
5383 	struct e1000_buffer *buffer_info;
5384 	unsigned int i;
5385 	u32 cmd_length = 0;
5386 	u16 ipcse = 0, mss;
5387 	u8 ipcss, ipcso, tucss, tucso, hdr_len;
5388 	int err;
5389 
5390 	if (!skb_is_gso(skb))
5391 		return 0;
5392 
5393 	err = skb_cow_head(skb, 0);
5394 	if (err < 0)
5395 		return err;
5396 
5397 	hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5398 	mss = skb_shinfo(skb)->gso_size;
5399 	if (protocol == htons(ETH_P_IP)) {
5400 		struct iphdr *iph = ip_hdr(skb);
5401 		iph->tot_len = 0;
5402 		iph->check = 0;
5403 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
5404 							 0, IPPROTO_TCP, 0);
5405 		cmd_length = E1000_TXD_CMD_IP;
5406 		ipcse = skb_transport_offset(skb) - 1;
5407 	} else if (skb_is_gso_v6(skb)) {
5408 		ipv6_hdr(skb)->payload_len = 0;
5409 		tcp_hdr(skb)->check = ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
5410 						       &ipv6_hdr(skb)->daddr,
5411 						       0, IPPROTO_TCP, 0);
5412 		ipcse = 0;
5413 	}
5414 	ipcss = skb_network_offset(skb);
5415 	ipcso = (void *)&(ip_hdr(skb)->check) - (void *)skb->data;
5416 	tucss = skb_transport_offset(skb);
5417 	tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
5418 
5419 	cmd_length |= (E1000_TXD_CMD_DEXT | E1000_TXD_CMD_TSE |
5420 		       E1000_TXD_CMD_TCP | (skb->len - (hdr_len)));
5421 
5422 	i = tx_ring->next_to_use;
5423 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5424 	buffer_info = &tx_ring->buffer_info[i];
5425 
5426 	context_desc->lower_setup.ip_fields.ipcss = ipcss;
5427 	context_desc->lower_setup.ip_fields.ipcso = ipcso;
5428 	context_desc->lower_setup.ip_fields.ipcse = cpu_to_le16(ipcse);
5429 	context_desc->upper_setup.tcp_fields.tucss = tucss;
5430 	context_desc->upper_setup.tcp_fields.tucso = tucso;
5431 	context_desc->upper_setup.tcp_fields.tucse = 0;
5432 	context_desc->tcp_seg_setup.fields.mss = cpu_to_le16(mss);
5433 	context_desc->tcp_seg_setup.fields.hdr_len = hdr_len;
5434 	context_desc->cmd_and_length = cpu_to_le32(cmd_length);
5435 
5436 	buffer_info->time_stamp = jiffies;
5437 	buffer_info->next_to_watch = i;
5438 
5439 	i++;
5440 	if (i == tx_ring->count)
5441 		i = 0;
5442 	tx_ring->next_to_use = i;
5443 
5444 	return 1;
5445 }
5446 
5447 static bool e1000_tx_csum(struct e1000_ring *tx_ring, struct sk_buff *skb,
5448 			  __be16 protocol)
5449 {
5450 	struct e1000_adapter *adapter = tx_ring->adapter;
5451 	struct e1000_context_desc *context_desc;
5452 	struct e1000_buffer *buffer_info;
5453 	unsigned int i;
5454 	u8 css;
5455 	u32 cmd_len = E1000_TXD_CMD_DEXT;
5456 
5457 	if (skb->ip_summed != CHECKSUM_PARTIAL)
5458 		return false;
5459 
5460 	switch (protocol) {
5461 	case cpu_to_be16(ETH_P_IP):
5462 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
5463 			cmd_len |= E1000_TXD_CMD_TCP;
5464 		break;
5465 	case cpu_to_be16(ETH_P_IPV6):
5466 		/* XXX not handling all IPV6 headers */
5467 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
5468 			cmd_len |= E1000_TXD_CMD_TCP;
5469 		break;
5470 	default:
5471 		if (unlikely(net_ratelimit()))
5472 			e_warn("checksum_partial proto=%x!\n",
5473 			       be16_to_cpu(protocol));
5474 		break;
5475 	}
5476 
5477 	css = skb_checksum_start_offset(skb);
5478 
5479 	i = tx_ring->next_to_use;
5480 	buffer_info = &tx_ring->buffer_info[i];
5481 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5482 
5483 	context_desc->lower_setup.ip_config = 0;
5484 	context_desc->upper_setup.tcp_fields.tucss = css;
5485 	context_desc->upper_setup.tcp_fields.tucso = css + skb->csum_offset;
5486 	context_desc->upper_setup.tcp_fields.tucse = 0;
5487 	context_desc->tcp_seg_setup.data = 0;
5488 	context_desc->cmd_and_length = cpu_to_le32(cmd_len);
5489 
5490 	buffer_info->time_stamp = jiffies;
5491 	buffer_info->next_to_watch = i;
5492 
5493 	i++;
5494 	if (i == tx_ring->count)
5495 		i = 0;
5496 	tx_ring->next_to_use = i;
5497 
5498 	return true;
5499 }
5500 
5501 static int e1000_tx_map(struct e1000_ring *tx_ring, struct sk_buff *skb,
5502 			unsigned int first, unsigned int max_per_txd,
5503 			unsigned int nr_frags)
5504 {
5505 	struct e1000_adapter *adapter = tx_ring->adapter;
5506 	struct pci_dev *pdev = adapter->pdev;
5507 	struct e1000_buffer *buffer_info;
5508 	unsigned int len = skb_headlen(skb);
5509 	unsigned int offset = 0, size, count = 0, i;
5510 	unsigned int f, bytecount, segs;
5511 
5512 	i = tx_ring->next_to_use;
5513 
5514 	while (len) {
5515 		buffer_info = &tx_ring->buffer_info[i];
5516 		size = min(len, max_per_txd);
5517 
5518 		buffer_info->length = size;
5519 		buffer_info->time_stamp = jiffies;
5520 		buffer_info->next_to_watch = i;
5521 		buffer_info->dma = dma_map_single(&pdev->dev,
5522 						  skb->data + offset,
5523 						  size, DMA_TO_DEVICE);
5524 		buffer_info->mapped_as_page = false;
5525 		if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5526 			goto dma_error;
5527 
5528 		len -= size;
5529 		offset += size;
5530 		count++;
5531 
5532 		if (len) {
5533 			i++;
5534 			if (i == tx_ring->count)
5535 				i = 0;
5536 		}
5537 	}
5538 
5539 	for (f = 0; f < nr_frags; f++) {
5540 		const struct skb_frag_struct *frag;
5541 
5542 		frag = &skb_shinfo(skb)->frags[f];
5543 		len = skb_frag_size(frag);
5544 		offset = 0;
5545 
5546 		while (len) {
5547 			i++;
5548 			if (i == tx_ring->count)
5549 				i = 0;
5550 
5551 			buffer_info = &tx_ring->buffer_info[i];
5552 			size = min(len, max_per_txd);
5553 
5554 			buffer_info->length = size;
5555 			buffer_info->time_stamp = jiffies;
5556 			buffer_info->next_to_watch = i;
5557 			buffer_info->dma = skb_frag_dma_map(&pdev->dev, frag,
5558 							    offset, size,
5559 							    DMA_TO_DEVICE);
5560 			buffer_info->mapped_as_page = true;
5561 			if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5562 				goto dma_error;
5563 
5564 			len -= size;
5565 			offset += size;
5566 			count++;
5567 		}
5568 	}
5569 
5570 	segs = skb_shinfo(skb)->gso_segs ? : 1;
5571 	/* multiply data chunks by size of headers */
5572 	bytecount = ((segs - 1) * skb_headlen(skb)) + skb->len;
5573 
5574 	tx_ring->buffer_info[i].skb = skb;
5575 	tx_ring->buffer_info[i].segs = segs;
5576 	tx_ring->buffer_info[i].bytecount = bytecount;
5577 	tx_ring->buffer_info[first].next_to_watch = i;
5578 
5579 	return count;
5580 
5581 dma_error:
5582 	dev_err(&pdev->dev, "Tx DMA map failed\n");
5583 	buffer_info->dma = 0;
5584 	if (count)
5585 		count--;
5586 
5587 	while (count--) {
5588 		if (i == 0)
5589 			i += tx_ring->count;
5590 		i--;
5591 		buffer_info = &tx_ring->buffer_info[i];
5592 		e1000_put_txbuf(tx_ring, buffer_info);
5593 	}
5594 
5595 	return 0;
5596 }
5597 
5598 static void e1000_tx_queue(struct e1000_ring *tx_ring, int tx_flags, int count)
5599 {
5600 	struct e1000_adapter *adapter = tx_ring->adapter;
5601 	struct e1000_tx_desc *tx_desc = NULL;
5602 	struct e1000_buffer *buffer_info;
5603 	u32 txd_upper = 0, txd_lower = E1000_TXD_CMD_IFCS;
5604 	unsigned int i;
5605 
5606 	if (tx_flags & E1000_TX_FLAGS_TSO) {
5607 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D |
5608 		    E1000_TXD_CMD_TSE;
5609 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5610 
5611 		if (tx_flags & E1000_TX_FLAGS_IPV4)
5612 			txd_upper |= E1000_TXD_POPTS_IXSM << 8;
5613 	}
5614 
5615 	if (tx_flags & E1000_TX_FLAGS_CSUM) {
5616 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5617 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5618 	}
5619 
5620 	if (tx_flags & E1000_TX_FLAGS_VLAN) {
5621 		txd_lower |= E1000_TXD_CMD_VLE;
5622 		txd_upper |= (tx_flags & E1000_TX_FLAGS_VLAN_MASK);
5623 	}
5624 
5625 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5626 		txd_lower &= ~(E1000_TXD_CMD_IFCS);
5627 
5628 	if (unlikely(tx_flags & E1000_TX_FLAGS_HWTSTAMP)) {
5629 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5630 		txd_upper |= E1000_TXD_EXTCMD_TSTAMP;
5631 	}
5632 
5633 	i = tx_ring->next_to_use;
5634 
5635 	do {
5636 		buffer_info = &tx_ring->buffer_info[i];
5637 		tx_desc = E1000_TX_DESC(*tx_ring, i);
5638 		tx_desc->buffer_addr = cpu_to_le64(buffer_info->dma);
5639 		tx_desc->lower.data = cpu_to_le32(txd_lower |
5640 						  buffer_info->length);
5641 		tx_desc->upper.data = cpu_to_le32(txd_upper);
5642 
5643 		i++;
5644 		if (i == tx_ring->count)
5645 			i = 0;
5646 	} while (--count > 0);
5647 
5648 	tx_desc->lower.data |= cpu_to_le32(adapter->txd_cmd);
5649 
5650 	/* txd_cmd re-enables FCS, so we'll re-disable it here as desired. */
5651 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5652 		tx_desc->lower.data &= ~(cpu_to_le32(E1000_TXD_CMD_IFCS));
5653 
5654 	/* Force memory writes to complete before letting h/w
5655 	 * know there are new descriptors to fetch.  (Only
5656 	 * applicable for weak-ordered memory model archs,
5657 	 * such as IA-64).
5658 	 */
5659 	wmb();
5660 
5661 	tx_ring->next_to_use = i;
5662 }
5663 
5664 #define MINIMUM_DHCP_PACKET_SIZE 282
5665 static int e1000_transfer_dhcp_info(struct e1000_adapter *adapter,
5666 				    struct sk_buff *skb)
5667 {
5668 	struct e1000_hw *hw = &adapter->hw;
5669 	u16 length, offset;
5670 
5671 	if (skb_vlan_tag_present(skb) &&
5672 	    !((skb_vlan_tag_get(skb) == adapter->hw.mng_cookie.vlan_id) &&
5673 	      (adapter->hw.mng_cookie.status &
5674 	       E1000_MNG_DHCP_COOKIE_STATUS_VLAN)))
5675 		return 0;
5676 
5677 	if (skb->len <= MINIMUM_DHCP_PACKET_SIZE)
5678 		return 0;
5679 
5680 	if (((struct ethhdr *)skb->data)->h_proto != htons(ETH_P_IP))
5681 		return 0;
5682 
5683 	{
5684 		const struct iphdr *ip = (struct iphdr *)((u8 *)skb->data + 14);
5685 		struct udphdr *udp;
5686 
5687 		if (ip->protocol != IPPROTO_UDP)
5688 			return 0;
5689 
5690 		udp = (struct udphdr *)((u8 *)ip + (ip->ihl << 2));
5691 		if (ntohs(udp->dest) != 67)
5692 			return 0;
5693 
5694 		offset = (u8 *)udp + 8 - skb->data;
5695 		length = skb->len - offset;
5696 		return e1000e_mng_write_dhcp_info(hw, (u8 *)udp + 8, length);
5697 	}
5698 
5699 	return 0;
5700 }
5701 
5702 static int __e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5703 {
5704 	struct e1000_adapter *adapter = tx_ring->adapter;
5705 
5706 	netif_stop_queue(adapter->netdev);
5707 	/* Herbert's original patch had:
5708 	 *  smp_mb__after_netif_stop_queue();
5709 	 * but since that doesn't exist yet, just open code it.
5710 	 */
5711 	smp_mb();
5712 
5713 	/* We need to check again in a case another CPU has just
5714 	 * made room available.
5715 	 */
5716 	if (e1000_desc_unused(tx_ring) < size)
5717 		return -EBUSY;
5718 
5719 	/* A reprieve! */
5720 	netif_start_queue(adapter->netdev);
5721 	++adapter->restart_queue;
5722 	return 0;
5723 }
5724 
5725 static int e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5726 {
5727 	BUG_ON(size > tx_ring->count);
5728 
5729 	if (e1000_desc_unused(tx_ring) >= size)
5730 		return 0;
5731 	return __e1000_maybe_stop_tx(tx_ring, size);
5732 }
5733 
5734 static netdev_tx_t e1000_xmit_frame(struct sk_buff *skb,
5735 				    struct net_device *netdev)
5736 {
5737 	struct e1000_adapter *adapter = netdev_priv(netdev);
5738 	struct e1000_ring *tx_ring = adapter->tx_ring;
5739 	unsigned int first;
5740 	unsigned int tx_flags = 0;
5741 	unsigned int len = skb_headlen(skb);
5742 	unsigned int nr_frags;
5743 	unsigned int mss;
5744 	int count = 0;
5745 	int tso;
5746 	unsigned int f;
5747 	__be16 protocol = vlan_get_protocol(skb);
5748 
5749 	if (test_bit(__E1000_DOWN, &adapter->state)) {
5750 		dev_kfree_skb_any(skb);
5751 		return NETDEV_TX_OK;
5752 	}
5753 
5754 	if (skb->len <= 0) {
5755 		dev_kfree_skb_any(skb);
5756 		return NETDEV_TX_OK;
5757 	}
5758 
5759 	/* The minimum packet size with TCTL.PSP set is 17 bytes so
5760 	 * pad skb in order to meet this minimum size requirement
5761 	 */
5762 	if (skb_put_padto(skb, 17))
5763 		return NETDEV_TX_OK;
5764 
5765 	mss = skb_shinfo(skb)->gso_size;
5766 	if (mss) {
5767 		u8 hdr_len;
5768 
5769 		/* TSO Workaround for 82571/2/3 Controllers -- if skb->data
5770 		 * points to just header, pull a few bytes of payload from
5771 		 * frags into skb->data
5772 		 */
5773 		hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5774 		/* we do this workaround for ES2LAN, but it is un-necessary,
5775 		 * avoiding it could save a lot of cycles
5776 		 */
5777 		if (skb->data_len && (hdr_len == len)) {
5778 			unsigned int pull_size;
5779 
5780 			pull_size = min_t(unsigned int, 4, skb->data_len);
5781 			if (!__pskb_pull_tail(skb, pull_size)) {
5782 				e_err("__pskb_pull_tail failed.\n");
5783 				dev_kfree_skb_any(skb);
5784 				return NETDEV_TX_OK;
5785 			}
5786 			len = skb_headlen(skb);
5787 		}
5788 	}
5789 
5790 	/* reserve a descriptor for the offload context */
5791 	if ((mss) || (skb->ip_summed == CHECKSUM_PARTIAL))
5792 		count++;
5793 	count++;
5794 
5795 	count += DIV_ROUND_UP(len, adapter->tx_fifo_limit);
5796 
5797 	nr_frags = skb_shinfo(skb)->nr_frags;
5798 	for (f = 0; f < nr_frags; f++)
5799 		count += DIV_ROUND_UP(skb_frag_size(&skb_shinfo(skb)->frags[f]),
5800 				      adapter->tx_fifo_limit);
5801 
5802 	if (adapter->hw.mac.tx_pkt_filtering)
5803 		e1000_transfer_dhcp_info(adapter, skb);
5804 
5805 	/* need: count + 2 desc gap to keep tail from touching
5806 	 * head, otherwise try next time
5807 	 */
5808 	if (e1000_maybe_stop_tx(tx_ring, count + 2))
5809 		return NETDEV_TX_BUSY;
5810 
5811 	if (skb_vlan_tag_present(skb)) {
5812 		tx_flags |= E1000_TX_FLAGS_VLAN;
5813 		tx_flags |= (skb_vlan_tag_get(skb) <<
5814 			     E1000_TX_FLAGS_VLAN_SHIFT);
5815 	}
5816 
5817 	first = tx_ring->next_to_use;
5818 
5819 	tso = e1000_tso(tx_ring, skb, protocol);
5820 	if (tso < 0) {
5821 		dev_kfree_skb_any(skb);
5822 		return NETDEV_TX_OK;
5823 	}
5824 
5825 	if (tso)
5826 		tx_flags |= E1000_TX_FLAGS_TSO;
5827 	else if (e1000_tx_csum(tx_ring, skb, protocol))
5828 		tx_flags |= E1000_TX_FLAGS_CSUM;
5829 
5830 	/* Old method was to assume IPv4 packet by default if TSO was enabled.
5831 	 * 82571 hardware supports TSO capabilities for IPv6 as well...
5832 	 * no longer assume, we must.
5833 	 */
5834 	if (protocol == htons(ETH_P_IP))
5835 		tx_flags |= E1000_TX_FLAGS_IPV4;
5836 
5837 	if (unlikely(skb->no_fcs))
5838 		tx_flags |= E1000_TX_FLAGS_NO_FCS;
5839 
5840 	/* if count is 0 then mapping error has occurred */
5841 	count = e1000_tx_map(tx_ring, skb, first, adapter->tx_fifo_limit,
5842 			     nr_frags);
5843 	if (count) {
5844 		if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
5845 		    (adapter->flags & FLAG_HAS_HW_TIMESTAMP) &&
5846 		    !adapter->tx_hwtstamp_skb) {
5847 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5848 			tx_flags |= E1000_TX_FLAGS_HWTSTAMP;
5849 			adapter->tx_hwtstamp_skb = skb_get(skb);
5850 			adapter->tx_hwtstamp_start = jiffies;
5851 			schedule_work(&adapter->tx_hwtstamp_work);
5852 		} else {
5853 			skb_tx_timestamp(skb);
5854 		}
5855 
5856 		netdev_sent_queue(netdev, skb->len);
5857 		e1000_tx_queue(tx_ring, tx_flags, count);
5858 		/* Make sure there is space in the ring for the next send. */
5859 		e1000_maybe_stop_tx(tx_ring,
5860 				    (MAX_SKB_FRAGS *
5861 				     DIV_ROUND_UP(PAGE_SIZE,
5862 						  adapter->tx_fifo_limit) + 2));
5863 
5864 		if (!skb->xmit_more ||
5865 		    netif_xmit_stopped(netdev_get_tx_queue(netdev, 0))) {
5866 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
5867 				e1000e_update_tdt_wa(tx_ring,
5868 						     tx_ring->next_to_use);
5869 			else
5870 				writel(tx_ring->next_to_use, tx_ring->tail);
5871 
5872 			/* we need this if more than one processor can write
5873 			 * to our tail at a time, it synchronizes IO on
5874 			 *IA64/Altix systems
5875 			 */
5876 			mmiowb();
5877 		}
5878 	} else {
5879 		dev_kfree_skb_any(skb);
5880 		tx_ring->buffer_info[first].time_stamp = 0;
5881 		tx_ring->next_to_use = first;
5882 	}
5883 
5884 	return NETDEV_TX_OK;
5885 }
5886 
5887 /**
5888  * e1000_tx_timeout - Respond to a Tx Hang
5889  * @netdev: network interface device structure
5890  **/
5891 static void e1000_tx_timeout(struct net_device *netdev)
5892 {
5893 	struct e1000_adapter *adapter = netdev_priv(netdev);
5894 
5895 	/* Do the reset outside of interrupt context */
5896 	adapter->tx_timeout_count++;
5897 	schedule_work(&adapter->reset_task);
5898 }
5899 
5900 static void e1000_reset_task(struct work_struct *work)
5901 {
5902 	struct e1000_adapter *adapter;
5903 	adapter = container_of(work, struct e1000_adapter, reset_task);
5904 
5905 	/* don't run the task if already down */
5906 	if (test_bit(__E1000_DOWN, &adapter->state))
5907 		return;
5908 
5909 	if (!(adapter->flags & FLAG_RESTART_NOW)) {
5910 		e1000e_dump(adapter);
5911 		e_err("Reset adapter unexpectedly\n");
5912 	}
5913 	e1000e_reinit_locked(adapter);
5914 }
5915 
5916 /**
5917  * e1000_get_stats64 - Get System Network Statistics
5918  * @netdev: network interface device structure
5919  * @stats: rtnl_link_stats64 pointer
5920  *
5921  * Returns the address of the device statistics structure.
5922  **/
5923 void e1000e_get_stats64(struct net_device *netdev,
5924 			struct rtnl_link_stats64 *stats)
5925 {
5926 	struct e1000_adapter *adapter = netdev_priv(netdev);
5927 
5928 	spin_lock(&adapter->stats64_lock);
5929 	e1000e_update_stats(adapter);
5930 	/* Fill out the OS statistics structure */
5931 	stats->rx_bytes = adapter->stats.gorc;
5932 	stats->rx_packets = adapter->stats.gprc;
5933 	stats->tx_bytes = adapter->stats.gotc;
5934 	stats->tx_packets = adapter->stats.gptc;
5935 	stats->multicast = adapter->stats.mprc;
5936 	stats->collisions = adapter->stats.colc;
5937 
5938 	/* Rx Errors */
5939 
5940 	/* RLEC on some newer hardware can be incorrect so build
5941 	 * our own version based on RUC and ROC
5942 	 */
5943 	stats->rx_errors = adapter->stats.rxerrc +
5944 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
5945 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5946 	stats->rx_length_errors = adapter->stats.ruc + adapter->stats.roc;
5947 	stats->rx_crc_errors = adapter->stats.crcerrs;
5948 	stats->rx_frame_errors = adapter->stats.algnerrc;
5949 	stats->rx_missed_errors = adapter->stats.mpc;
5950 
5951 	/* Tx Errors */
5952 	stats->tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5953 	stats->tx_aborted_errors = adapter->stats.ecol;
5954 	stats->tx_window_errors = adapter->stats.latecol;
5955 	stats->tx_carrier_errors = adapter->stats.tncrs;
5956 
5957 	/* Tx Dropped needs to be maintained elsewhere */
5958 
5959 	spin_unlock(&adapter->stats64_lock);
5960 }
5961 
5962 /**
5963  * e1000_change_mtu - Change the Maximum Transfer Unit
5964  * @netdev: network interface device structure
5965  * @new_mtu: new value for maximum frame size
5966  *
5967  * Returns 0 on success, negative on failure
5968  **/
5969 static int e1000_change_mtu(struct net_device *netdev, int new_mtu)
5970 {
5971 	struct e1000_adapter *adapter = netdev_priv(netdev);
5972 	int max_frame = new_mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
5973 
5974 	/* Jumbo frame support */
5975 	if ((new_mtu > ETH_DATA_LEN) &&
5976 	    !(adapter->flags & FLAG_HAS_JUMBO_FRAMES)) {
5977 		e_err("Jumbo Frames not supported.\n");
5978 		return -EINVAL;
5979 	}
5980 
5981 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
5982 	if ((adapter->hw.mac.type >= e1000_pch2lan) &&
5983 	    !(adapter->flags2 & FLAG2_CRC_STRIPPING) &&
5984 	    (new_mtu > ETH_DATA_LEN)) {
5985 		e_err("Jumbo Frames not supported on this device when CRC stripping is disabled.\n");
5986 		return -EINVAL;
5987 	}
5988 
5989 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
5990 		usleep_range(1000, 2000);
5991 	/* e1000e_down -> e1000e_reset dependent on max_frame_size & mtu */
5992 	adapter->max_frame_size = max_frame;
5993 	e_info("changing MTU from %d to %d\n", netdev->mtu, new_mtu);
5994 	netdev->mtu = new_mtu;
5995 
5996 	pm_runtime_get_sync(netdev->dev.parent);
5997 
5998 	if (netif_running(netdev))
5999 		e1000e_down(adapter, true);
6000 
6001 	/* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
6002 	 * means we reserve 2 more, this pushes us to allocate from the next
6003 	 * larger slab size.
6004 	 * i.e. RXBUFFER_2048 --> size-4096 slab
6005 	 * However with the new *_jumbo_rx* routines, jumbo receives will use
6006 	 * fragmented skbs
6007 	 */
6008 
6009 	if (max_frame <= 2048)
6010 		adapter->rx_buffer_len = 2048;
6011 	else
6012 		adapter->rx_buffer_len = 4096;
6013 
6014 	/* adjust allocation if LPE protects us, and we aren't using SBP */
6015 	if (max_frame <= (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN))
6016 		adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
6017 
6018 	if (netif_running(netdev))
6019 		e1000e_up(adapter);
6020 	else
6021 		e1000e_reset(adapter);
6022 
6023 	pm_runtime_put_sync(netdev->dev.parent);
6024 
6025 	clear_bit(__E1000_RESETTING, &adapter->state);
6026 
6027 	return 0;
6028 }
6029 
6030 static int e1000_mii_ioctl(struct net_device *netdev, struct ifreq *ifr,
6031 			   int cmd)
6032 {
6033 	struct e1000_adapter *adapter = netdev_priv(netdev);
6034 	struct mii_ioctl_data *data = if_mii(ifr);
6035 
6036 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
6037 		return -EOPNOTSUPP;
6038 
6039 	switch (cmd) {
6040 	case SIOCGMIIPHY:
6041 		data->phy_id = adapter->hw.phy.addr;
6042 		break;
6043 	case SIOCGMIIREG:
6044 		e1000_phy_read_status(adapter);
6045 
6046 		switch (data->reg_num & 0x1F) {
6047 		case MII_BMCR:
6048 			data->val_out = adapter->phy_regs.bmcr;
6049 			break;
6050 		case MII_BMSR:
6051 			data->val_out = adapter->phy_regs.bmsr;
6052 			break;
6053 		case MII_PHYSID1:
6054 			data->val_out = (adapter->hw.phy.id >> 16);
6055 			break;
6056 		case MII_PHYSID2:
6057 			data->val_out = (adapter->hw.phy.id & 0xFFFF);
6058 			break;
6059 		case MII_ADVERTISE:
6060 			data->val_out = adapter->phy_regs.advertise;
6061 			break;
6062 		case MII_LPA:
6063 			data->val_out = adapter->phy_regs.lpa;
6064 			break;
6065 		case MII_EXPANSION:
6066 			data->val_out = adapter->phy_regs.expansion;
6067 			break;
6068 		case MII_CTRL1000:
6069 			data->val_out = adapter->phy_regs.ctrl1000;
6070 			break;
6071 		case MII_STAT1000:
6072 			data->val_out = adapter->phy_regs.stat1000;
6073 			break;
6074 		case MII_ESTATUS:
6075 			data->val_out = adapter->phy_regs.estatus;
6076 			break;
6077 		default:
6078 			return -EIO;
6079 		}
6080 		break;
6081 	case SIOCSMIIREG:
6082 	default:
6083 		return -EOPNOTSUPP;
6084 	}
6085 	return 0;
6086 }
6087 
6088 /**
6089  * e1000e_hwtstamp_ioctl - control hardware time stamping
6090  * @netdev: network interface device structure
6091  * @ifreq: interface request
6092  *
6093  * Outgoing time stamping can be enabled and disabled. Play nice and
6094  * disable it when requested, although it shouldn't cause any overhead
6095  * when no packet needs it. At most one packet in the queue may be
6096  * marked for time stamping, otherwise it would be impossible to tell
6097  * for sure to which packet the hardware time stamp belongs.
6098  *
6099  * Incoming time stamping has to be configured via the hardware filters.
6100  * Not all combinations are supported, in particular event type has to be
6101  * specified. Matching the kind of event packet is not supported, with the
6102  * exception of "all V2 events regardless of level 2 or 4".
6103  **/
6104 static int e1000e_hwtstamp_set(struct net_device *netdev, struct ifreq *ifr)
6105 {
6106 	struct e1000_adapter *adapter = netdev_priv(netdev);
6107 	struct hwtstamp_config config;
6108 	int ret_val;
6109 
6110 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
6111 		return -EFAULT;
6112 
6113 	ret_val = e1000e_config_hwtstamp(adapter, &config);
6114 	if (ret_val)
6115 		return ret_val;
6116 
6117 	switch (config.rx_filter) {
6118 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
6119 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
6120 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
6121 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
6122 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
6123 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
6124 		/* With V2 type filters which specify a Sync or Delay Request,
6125 		 * Path Delay Request/Response messages are also time stamped
6126 		 * by hardware so notify the caller the requested packets plus
6127 		 * some others are time stamped.
6128 		 */
6129 		config.rx_filter = HWTSTAMP_FILTER_SOME;
6130 		break;
6131 	default:
6132 		break;
6133 	}
6134 
6135 	return copy_to_user(ifr->ifr_data, &config,
6136 			    sizeof(config)) ? -EFAULT : 0;
6137 }
6138 
6139 static int e1000e_hwtstamp_get(struct net_device *netdev, struct ifreq *ifr)
6140 {
6141 	struct e1000_adapter *adapter = netdev_priv(netdev);
6142 
6143 	return copy_to_user(ifr->ifr_data, &adapter->hwtstamp_config,
6144 			    sizeof(adapter->hwtstamp_config)) ? -EFAULT : 0;
6145 }
6146 
6147 static int e1000_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
6148 {
6149 	switch (cmd) {
6150 	case SIOCGMIIPHY:
6151 	case SIOCGMIIREG:
6152 	case SIOCSMIIREG:
6153 		return e1000_mii_ioctl(netdev, ifr, cmd);
6154 	case SIOCSHWTSTAMP:
6155 		return e1000e_hwtstamp_set(netdev, ifr);
6156 	case SIOCGHWTSTAMP:
6157 		return e1000e_hwtstamp_get(netdev, ifr);
6158 	default:
6159 		return -EOPNOTSUPP;
6160 	}
6161 }
6162 
6163 static int e1000_init_phy_wakeup(struct e1000_adapter *adapter, u32 wufc)
6164 {
6165 	struct e1000_hw *hw = &adapter->hw;
6166 	u32 i, mac_reg, wuc;
6167 	u16 phy_reg, wuc_enable;
6168 	int retval;
6169 
6170 	/* copy MAC RARs to PHY RARs */
6171 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
6172 
6173 	retval = hw->phy.ops.acquire(hw);
6174 	if (retval) {
6175 		e_err("Could not acquire PHY\n");
6176 		return retval;
6177 	}
6178 
6179 	/* Enable access to wakeup registers on and set page to BM_WUC_PAGE */
6180 	retval = e1000_enable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6181 	if (retval)
6182 		goto release;
6183 
6184 	/* copy MAC MTA to PHY MTA - only needed for pchlan */
6185 	for (i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
6186 		mac_reg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
6187 		hw->phy.ops.write_reg_page(hw, BM_MTA(i),
6188 					   (u16)(mac_reg & 0xFFFF));
6189 		hw->phy.ops.write_reg_page(hw, BM_MTA(i) + 1,
6190 					   (u16)((mac_reg >> 16) & 0xFFFF));
6191 	}
6192 
6193 	/* configure PHY Rx Control register */
6194 	hw->phy.ops.read_reg_page(&adapter->hw, BM_RCTL, &phy_reg);
6195 	mac_reg = er32(RCTL);
6196 	if (mac_reg & E1000_RCTL_UPE)
6197 		phy_reg |= BM_RCTL_UPE;
6198 	if (mac_reg & E1000_RCTL_MPE)
6199 		phy_reg |= BM_RCTL_MPE;
6200 	phy_reg &= ~(BM_RCTL_MO_MASK);
6201 	if (mac_reg & E1000_RCTL_MO_3)
6202 		phy_reg |= (((mac_reg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
6203 			    << BM_RCTL_MO_SHIFT);
6204 	if (mac_reg & E1000_RCTL_BAM)
6205 		phy_reg |= BM_RCTL_BAM;
6206 	if (mac_reg & E1000_RCTL_PMCF)
6207 		phy_reg |= BM_RCTL_PMCF;
6208 	mac_reg = er32(CTRL);
6209 	if (mac_reg & E1000_CTRL_RFCE)
6210 		phy_reg |= BM_RCTL_RFCE;
6211 	hw->phy.ops.write_reg_page(&adapter->hw, BM_RCTL, phy_reg);
6212 
6213 	wuc = E1000_WUC_PME_EN;
6214 	if (wufc & (E1000_WUFC_MAG | E1000_WUFC_LNKC))
6215 		wuc |= E1000_WUC_APME;
6216 
6217 	/* enable PHY wakeup in MAC register */
6218 	ew32(WUFC, wufc);
6219 	ew32(WUC, (E1000_WUC_PHY_WAKE | E1000_WUC_APMPME |
6220 		   E1000_WUC_PME_STATUS | wuc));
6221 
6222 	/* configure and enable PHY wakeup in PHY registers */
6223 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUFC, wufc);
6224 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUC, wuc);
6225 
6226 	/* activate PHY wakeup */
6227 	wuc_enable |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
6228 	retval = e1000_disable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6229 	if (retval)
6230 		e_err("Could not set PHY Host Wakeup bit\n");
6231 release:
6232 	hw->phy.ops.release(hw);
6233 
6234 	return retval;
6235 }
6236 
6237 static void e1000e_flush_lpic(struct pci_dev *pdev)
6238 {
6239 	struct net_device *netdev = pci_get_drvdata(pdev);
6240 	struct e1000_adapter *adapter = netdev_priv(netdev);
6241 	struct e1000_hw *hw = &adapter->hw;
6242 	u32 ret_val;
6243 
6244 	pm_runtime_get_sync(netdev->dev.parent);
6245 
6246 	ret_val = hw->phy.ops.acquire(hw);
6247 	if (ret_val)
6248 		goto fl_out;
6249 
6250 	pr_info("EEE TX LPI TIMER: %08X\n",
6251 		er32(LPIC) >> E1000_LPIC_LPIET_SHIFT);
6252 
6253 	hw->phy.ops.release(hw);
6254 
6255 fl_out:
6256 	pm_runtime_put_sync(netdev->dev.parent);
6257 }
6258 
6259 static int e1000e_pm_freeze(struct device *dev)
6260 {
6261 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6262 	struct e1000_adapter *adapter = netdev_priv(netdev);
6263 
6264 	netif_device_detach(netdev);
6265 
6266 	if (netif_running(netdev)) {
6267 		int count = E1000_CHECK_RESET_COUNT;
6268 
6269 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6270 			usleep_range(10000, 20000);
6271 
6272 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6273 
6274 		/* Quiesce the device without resetting the hardware */
6275 		e1000e_down(adapter, false);
6276 		e1000_free_irq(adapter);
6277 		e1000e_reset_interrupt_capability(adapter);
6278 	}
6279 
6280 	/* Allow time for pending master requests to run */
6281 	e1000e_disable_pcie_master(&adapter->hw);
6282 
6283 	return 0;
6284 }
6285 
6286 static int __e1000_shutdown(struct pci_dev *pdev, bool runtime)
6287 {
6288 	struct net_device *netdev = pci_get_drvdata(pdev);
6289 	struct e1000_adapter *adapter = netdev_priv(netdev);
6290 	struct e1000_hw *hw = &adapter->hw;
6291 	u32 ctrl, ctrl_ext, rctl, status;
6292 	/* Runtime suspend should only enable wakeup for link changes */
6293 	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
6294 	int retval = 0;
6295 
6296 	status = er32(STATUS);
6297 	if (status & E1000_STATUS_LU)
6298 		wufc &= ~E1000_WUFC_LNKC;
6299 
6300 	if (wufc) {
6301 		e1000_setup_rctl(adapter);
6302 		e1000e_set_rx_mode(netdev);
6303 
6304 		/* turn on all-multi mode if wake on multicast is enabled */
6305 		if (wufc & E1000_WUFC_MC) {
6306 			rctl = er32(RCTL);
6307 			rctl |= E1000_RCTL_MPE;
6308 			ew32(RCTL, rctl);
6309 		}
6310 
6311 		ctrl = er32(CTRL);
6312 		ctrl |= E1000_CTRL_ADVD3WUC;
6313 		if (!(adapter->flags2 & FLAG2_HAS_PHY_WAKEUP))
6314 			ctrl |= E1000_CTRL_EN_PHY_PWR_MGMT;
6315 		ew32(CTRL, ctrl);
6316 
6317 		if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
6318 		    adapter->hw.phy.media_type ==
6319 		    e1000_media_type_internal_serdes) {
6320 			/* keep the laser running in D3 */
6321 			ctrl_ext = er32(CTRL_EXT);
6322 			ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
6323 			ew32(CTRL_EXT, ctrl_ext);
6324 		}
6325 
6326 		if (!runtime)
6327 			e1000e_power_up_phy(adapter);
6328 
6329 		if (adapter->flags & FLAG_IS_ICH)
6330 			e1000_suspend_workarounds_ich8lan(&adapter->hw);
6331 
6332 		if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6333 			/* enable wakeup by the PHY */
6334 			retval = e1000_init_phy_wakeup(adapter, wufc);
6335 			if (retval)
6336 				return retval;
6337 		} else {
6338 			/* enable wakeup by the MAC */
6339 			ew32(WUFC, wufc);
6340 			ew32(WUC, E1000_WUC_PME_EN);
6341 		}
6342 	} else {
6343 		ew32(WUC, 0);
6344 		ew32(WUFC, 0);
6345 
6346 		e1000_power_down_phy(adapter);
6347 	}
6348 
6349 	if (adapter->hw.phy.type == e1000_phy_igp_3) {
6350 		e1000e_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
6351 	} else if ((hw->mac.type == e1000_pch_lpt) ||
6352 		   (hw->mac.type == e1000_pch_spt)) {
6353 		if (!(wufc & (E1000_WUFC_EX | E1000_WUFC_MC | E1000_WUFC_BC)))
6354 			/* ULP does not support wake from unicast, multicast
6355 			 * or broadcast.
6356 			 */
6357 			retval = e1000_enable_ulp_lpt_lp(hw, !runtime);
6358 
6359 		if (retval)
6360 			return retval;
6361 	}
6362 
6363 	/* Ensure that the appropriate bits are set in LPI_CTRL
6364 	 * for EEE in Sx
6365 	 */
6366 	if ((hw->phy.type >= e1000_phy_i217) &&
6367 	    adapter->eee_advert && hw->dev_spec.ich8lan.eee_lp_ability) {
6368 		u16 lpi_ctrl = 0;
6369 
6370 		retval = hw->phy.ops.acquire(hw);
6371 		if (!retval) {
6372 			retval = e1e_rphy_locked(hw, I82579_LPI_CTRL,
6373 						 &lpi_ctrl);
6374 			if (!retval) {
6375 				if (adapter->eee_advert &
6376 				    hw->dev_spec.ich8lan.eee_lp_ability &
6377 				    I82579_EEE_100_SUPPORTED)
6378 					lpi_ctrl |= I82579_LPI_CTRL_100_ENABLE;
6379 				if (adapter->eee_advert &
6380 				    hw->dev_spec.ich8lan.eee_lp_ability &
6381 				    I82579_EEE_1000_SUPPORTED)
6382 					lpi_ctrl |= I82579_LPI_CTRL_1000_ENABLE;
6383 
6384 				retval = e1e_wphy_locked(hw, I82579_LPI_CTRL,
6385 							 lpi_ctrl);
6386 			}
6387 		}
6388 		hw->phy.ops.release(hw);
6389 	}
6390 
6391 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
6392 	 * would have already happened in close and is redundant.
6393 	 */
6394 	e1000e_release_hw_control(adapter);
6395 
6396 	pci_clear_master(pdev);
6397 
6398 	/* The pci-e switch on some quad port adapters will report a
6399 	 * correctable error when the MAC transitions from D0 to D3.  To
6400 	 * prevent this we need to mask off the correctable errors on the
6401 	 * downstream port of the pci-e switch.
6402 	 *
6403 	 * We don't have the associated upstream bridge while assigning
6404 	 * the PCI device into guest. For example, the KVM on power is
6405 	 * one of the cases.
6406 	 */
6407 	if (adapter->flags & FLAG_IS_QUAD_PORT) {
6408 		struct pci_dev *us_dev = pdev->bus->self;
6409 		u16 devctl;
6410 
6411 		if (!us_dev)
6412 			return 0;
6413 
6414 		pcie_capability_read_word(us_dev, PCI_EXP_DEVCTL, &devctl);
6415 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL,
6416 					   (devctl & ~PCI_EXP_DEVCTL_CERE));
6417 
6418 		pci_save_state(pdev);
6419 		pci_prepare_to_sleep(pdev);
6420 
6421 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL, devctl);
6422 	}
6423 
6424 	return 0;
6425 }
6426 
6427 /**
6428  * __e1000e_disable_aspm - Disable ASPM states
6429  * @pdev: pointer to PCI device struct
6430  * @state: bit-mask of ASPM states to disable
6431  * @locked: indication if this context holds pci_bus_sem locked.
6432  *
6433  * Some devices *must* have certain ASPM states disabled per hardware errata.
6434  **/
6435 static void __e1000e_disable_aspm(struct pci_dev *pdev, u16 state, int locked)
6436 {
6437 	struct pci_dev *parent = pdev->bus->self;
6438 	u16 aspm_dis_mask = 0;
6439 	u16 pdev_aspmc, parent_aspmc;
6440 
6441 	switch (state) {
6442 	case PCIE_LINK_STATE_L0S:
6443 	case PCIE_LINK_STATE_L0S | PCIE_LINK_STATE_L1:
6444 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L0S;
6445 		/* fall-through - can't have L1 without L0s */
6446 	case PCIE_LINK_STATE_L1:
6447 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L1;
6448 		break;
6449 	default:
6450 		return;
6451 	}
6452 
6453 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6454 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6455 
6456 	if (parent) {
6457 		pcie_capability_read_word(parent, PCI_EXP_LNKCTL,
6458 					  &parent_aspmc);
6459 		parent_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6460 	}
6461 
6462 	/* Nothing to do if the ASPM states to be disabled already are */
6463 	if (!(pdev_aspmc & aspm_dis_mask) &&
6464 	    (!parent || !(parent_aspmc & aspm_dis_mask)))
6465 		return;
6466 
6467 	dev_info(&pdev->dev, "Disabling ASPM %s %s\n",
6468 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L0S) ?
6469 		 "L0s" : "",
6470 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L1) ?
6471 		 "L1" : "");
6472 
6473 #ifdef CONFIG_PCIEASPM
6474 	if (locked)
6475 		pci_disable_link_state_locked(pdev, state);
6476 	else
6477 		pci_disable_link_state(pdev, state);
6478 
6479 	/* Double-check ASPM control.  If not disabled by the above, the
6480 	 * BIOS is preventing that from happening (or CONFIG_PCIEASPM is
6481 	 * not enabled); override by writing PCI config space directly.
6482 	 */
6483 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6484 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6485 
6486 	if (!(aspm_dis_mask & pdev_aspmc))
6487 		return;
6488 #endif
6489 
6490 	/* Both device and parent should have the same ASPM setting.
6491 	 * Disable ASPM in downstream component first and then upstream.
6492 	 */
6493 	pcie_capability_clear_word(pdev, PCI_EXP_LNKCTL, aspm_dis_mask);
6494 
6495 	if (parent)
6496 		pcie_capability_clear_word(parent, PCI_EXP_LNKCTL,
6497 					   aspm_dis_mask);
6498 }
6499 
6500 /**
6501  * e1000e_disable_aspm - Disable ASPM states.
6502  * @pdev: pointer to PCI device struct
6503  * @state: bit-mask of ASPM states to disable
6504  *
6505  * This function acquires the pci_bus_sem!
6506  * Some devices *must* have certain ASPM states disabled per hardware errata.
6507  **/
6508 static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6509 {
6510 	__e1000e_disable_aspm(pdev, state, 0);
6511 }
6512 
6513 /**
6514  * e1000e_disable_aspm_locked   Disable ASPM states.
6515  * @pdev: pointer to PCI device struct
6516  * @state: bit-mask of ASPM states to disable
6517  *
6518  * This function must be called with pci_bus_sem acquired!
6519  * Some devices *must* have certain ASPM states disabled per hardware errata.
6520  **/
6521 static void e1000e_disable_aspm_locked(struct pci_dev *pdev, u16 state)
6522 {
6523 	__e1000e_disable_aspm(pdev, state, 1);
6524 }
6525 
6526 #ifdef CONFIG_PM
6527 static int __e1000_resume(struct pci_dev *pdev)
6528 {
6529 	struct net_device *netdev = pci_get_drvdata(pdev);
6530 	struct e1000_adapter *adapter = netdev_priv(netdev);
6531 	struct e1000_hw *hw = &adapter->hw;
6532 	u16 aspm_disable_flag = 0;
6533 
6534 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6535 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6536 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6537 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6538 	if (aspm_disable_flag)
6539 		e1000e_disable_aspm(pdev, aspm_disable_flag);
6540 
6541 	pci_set_master(pdev);
6542 
6543 	if (hw->mac.type >= e1000_pch2lan)
6544 		e1000_resume_workarounds_pchlan(&adapter->hw);
6545 
6546 	e1000e_power_up_phy(adapter);
6547 
6548 	/* report the system wakeup cause from S3/S4 */
6549 	if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6550 		u16 phy_data;
6551 
6552 		e1e_rphy(&adapter->hw, BM_WUS, &phy_data);
6553 		if (phy_data) {
6554 			e_info("PHY Wakeup cause - %s\n",
6555 			       phy_data & E1000_WUS_EX ? "Unicast Packet" :
6556 			       phy_data & E1000_WUS_MC ? "Multicast Packet" :
6557 			       phy_data & E1000_WUS_BC ? "Broadcast Packet" :
6558 			       phy_data & E1000_WUS_MAG ? "Magic Packet" :
6559 			       phy_data & E1000_WUS_LNKC ?
6560 			       "Link Status Change" : "other");
6561 		}
6562 		e1e_wphy(&adapter->hw, BM_WUS, ~0);
6563 	} else {
6564 		u32 wus = er32(WUS);
6565 
6566 		if (wus) {
6567 			e_info("MAC Wakeup cause - %s\n",
6568 			       wus & E1000_WUS_EX ? "Unicast Packet" :
6569 			       wus & E1000_WUS_MC ? "Multicast Packet" :
6570 			       wus & E1000_WUS_BC ? "Broadcast Packet" :
6571 			       wus & E1000_WUS_MAG ? "Magic Packet" :
6572 			       wus & E1000_WUS_LNKC ? "Link Status Change" :
6573 			       "other");
6574 		}
6575 		ew32(WUS, ~0);
6576 	}
6577 
6578 	e1000e_reset(adapter);
6579 
6580 	e1000_init_manageability_pt(adapter);
6581 
6582 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6583 	 * is up.  For all other cases, let the f/w know that the h/w is now
6584 	 * under the control of the driver.
6585 	 */
6586 	if (!(adapter->flags & FLAG_HAS_AMT))
6587 		e1000e_get_hw_control(adapter);
6588 
6589 	return 0;
6590 }
6591 
6592 #ifdef CONFIG_PM_SLEEP
6593 static int e1000e_pm_thaw(struct device *dev)
6594 {
6595 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6596 	struct e1000_adapter *adapter = netdev_priv(netdev);
6597 
6598 	e1000e_set_interrupt_capability(adapter);
6599 	if (netif_running(netdev)) {
6600 		u32 err = e1000_request_irq(adapter);
6601 
6602 		if (err)
6603 			return err;
6604 
6605 		e1000e_up(adapter);
6606 	}
6607 
6608 	netif_device_attach(netdev);
6609 
6610 	return 0;
6611 }
6612 
6613 static int e1000e_pm_suspend(struct device *dev)
6614 {
6615 	struct pci_dev *pdev = to_pci_dev(dev);
6616 
6617 	e1000e_flush_lpic(pdev);
6618 
6619 	e1000e_pm_freeze(dev);
6620 
6621 	return __e1000_shutdown(pdev, false);
6622 }
6623 
6624 static int e1000e_pm_resume(struct device *dev)
6625 {
6626 	struct pci_dev *pdev = to_pci_dev(dev);
6627 	int rc;
6628 
6629 	rc = __e1000_resume(pdev);
6630 	if (rc)
6631 		return rc;
6632 
6633 	return e1000e_pm_thaw(dev);
6634 }
6635 #endif /* CONFIG_PM_SLEEP */
6636 
6637 static int e1000e_pm_runtime_idle(struct device *dev)
6638 {
6639 	struct pci_dev *pdev = to_pci_dev(dev);
6640 	struct net_device *netdev = pci_get_drvdata(pdev);
6641 	struct e1000_adapter *adapter = netdev_priv(netdev);
6642 	u16 eee_lp;
6643 
6644 	eee_lp = adapter->hw.dev_spec.ich8lan.eee_lp_ability;
6645 
6646 	if (!e1000e_has_link(adapter)) {
6647 		adapter->hw.dev_spec.ich8lan.eee_lp_ability = eee_lp;
6648 		pm_schedule_suspend(dev, 5 * MSEC_PER_SEC);
6649 	}
6650 
6651 	return -EBUSY;
6652 }
6653 
6654 static int e1000e_pm_runtime_resume(struct device *dev)
6655 {
6656 	struct pci_dev *pdev = to_pci_dev(dev);
6657 	struct net_device *netdev = pci_get_drvdata(pdev);
6658 	struct e1000_adapter *adapter = netdev_priv(netdev);
6659 	int rc;
6660 
6661 	rc = __e1000_resume(pdev);
6662 	if (rc)
6663 		return rc;
6664 
6665 	if (netdev->flags & IFF_UP)
6666 		e1000e_up(adapter);
6667 
6668 	return rc;
6669 }
6670 
6671 static int e1000e_pm_runtime_suspend(struct device *dev)
6672 {
6673 	struct pci_dev *pdev = to_pci_dev(dev);
6674 	struct net_device *netdev = pci_get_drvdata(pdev);
6675 	struct e1000_adapter *adapter = netdev_priv(netdev);
6676 
6677 	if (netdev->flags & IFF_UP) {
6678 		int count = E1000_CHECK_RESET_COUNT;
6679 
6680 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6681 			usleep_range(10000, 20000);
6682 
6683 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6684 
6685 		/* Down the device without resetting the hardware */
6686 		e1000e_down(adapter, false);
6687 	}
6688 
6689 	if (__e1000_shutdown(pdev, true)) {
6690 		e1000e_pm_runtime_resume(dev);
6691 		return -EBUSY;
6692 	}
6693 
6694 	return 0;
6695 }
6696 #endif /* CONFIG_PM */
6697 
6698 static void e1000_shutdown(struct pci_dev *pdev)
6699 {
6700 	e1000e_flush_lpic(pdev);
6701 
6702 	e1000e_pm_freeze(&pdev->dev);
6703 
6704 	__e1000_shutdown(pdev, false);
6705 }
6706 
6707 #ifdef CONFIG_NET_POLL_CONTROLLER
6708 
6709 static irqreturn_t e1000_intr_msix(int __always_unused irq, void *data)
6710 {
6711 	struct net_device *netdev = data;
6712 	struct e1000_adapter *adapter = netdev_priv(netdev);
6713 
6714 	if (adapter->msix_entries) {
6715 		int vector, msix_irq;
6716 
6717 		vector = 0;
6718 		msix_irq = adapter->msix_entries[vector].vector;
6719 		disable_irq(msix_irq);
6720 		e1000_intr_msix_rx(msix_irq, netdev);
6721 		enable_irq(msix_irq);
6722 
6723 		vector++;
6724 		msix_irq = adapter->msix_entries[vector].vector;
6725 		disable_irq(msix_irq);
6726 		e1000_intr_msix_tx(msix_irq, netdev);
6727 		enable_irq(msix_irq);
6728 
6729 		vector++;
6730 		msix_irq = adapter->msix_entries[vector].vector;
6731 		disable_irq(msix_irq);
6732 		e1000_msix_other(msix_irq, netdev);
6733 		enable_irq(msix_irq);
6734 	}
6735 
6736 	return IRQ_HANDLED;
6737 }
6738 
6739 /**
6740  * e1000_netpoll
6741  * @netdev: network interface device structure
6742  *
6743  * Polling 'interrupt' - used by things like netconsole to send skbs
6744  * without having to re-enable interrupts. It's not called while
6745  * the interrupt routine is executing.
6746  */
6747 static void e1000_netpoll(struct net_device *netdev)
6748 {
6749 	struct e1000_adapter *adapter = netdev_priv(netdev);
6750 
6751 	switch (adapter->int_mode) {
6752 	case E1000E_INT_MODE_MSIX:
6753 		e1000_intr_msix(adapter->pdev->irq, netdev);
6754 		break;
6755 	case E1000E_INT_MODE_MSI:
6756 		if (disable_hardirq(adapter->pdev->irq))
6757 			e1000_intr_msi(adapter->pdev->irq, netdev);
6758 		enable_irq(adapter->pdev->irq);
6759 		break;
6760 	default:		/* E1000E_INT_MODE_LEGACY */
6761 		if (disable_hardirq(adapter->pdev->irq))
6762 			e1000_intr(adapter->pdev->irq, netdev);
6763 		enable_irq(adapter->pdev->irq);
6764 		break;
6765 	}
6766 }
6767 #endif
6768 
6769 /**
6770  * e1000_io_error_detected - called when PCI error is detected
6771  * @pdev: Pointer to PCI device
6772  * @state: The current pci connection state
6773  *
6774  * This function is called after a PCI bus error affecting
6775  * this device has been detected.
6776  */
6777 static pci_ers_result_t e1000_io_error_detected(struct pci_dev *pdev,
6778 						pci_channel_state_t state)
6779 {
6780 	struct net_device *netdev = pci_get_drvdata(pdev);
6781 	struct e1000_adapter *adapter = netdev_priv(netdev);
6782 
6783 	netif_device_detach(netdev);
6784 
6785 	if (state == pci_channel_io_perm_failure)
6786 		return PCI_ERS_RESULT_DISCONNECT;
6787 
6788 	if (netif_running(netdev))
6789 		e1000e_down(adapter, true);
6790 	pci_disable_device(pdev);
6791 
6792 	/* Request a slot slot reset. */
6793 	return PCI_ERS_RESULT_NEED_RESET;
6794 }
6795 
6796 /**
6797  * e1000_io_slot_reset - called after the pci bus has been reset.
6798  * @pdev: Pointer to PCI device
6799  *
6800  * Restart the card from scratch, as if from a cold-boot. Implementation
6801  * resembles the first-half of the e1000e_pm_resume routine.
6802  */
6803 static pci_ers_result_t e1000_io_slot_reset(struct pci_dev *pdev)
6804 {
6805 	struct net_device *netdev = pci_get_drvdata(pdev);
6806 	struct e1000_adapter *adapter = netdev_priv(netdev);
6807 	struct e1000_hw *hw = &adapter->hw;
6808 	u16 aspm_disable_flag = 0;
6809 	int err;
6810 	pci_ers_result_t result;
6811 
6812 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6813 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6814 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6815 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6816 	if (aspm_disable_flag)
6817 		e1000e_disable_aspm_locked(pdev, aspm_disable_flag);
6818 
6819 	err = pci_enable_device_mem(pdev);
6820 	if (err) {
6821 		dev_err(&pdev->dev,
6822 			"Cannot re-enable PCI device after reset.\n");
6823 		result = PCI_ERS_RESULT_DISCONNECT;
6824 	} else {
6825 		pdev->state_saved = true;
6826 		pci_restore_state(pdev);
6827 		pci_set_master(pdev);
6828 
6829 		pci_enable_wake(pdev, PCI_D3hot, 0);
6830 		pci_enable_wake(pdev, PCI_D3cold, 0);
6831 
6832 		e1000e_reset(adapter);
6833 		ew32(WUS, ~0);
6834 		result = PCI_ERS_RESULT_RECOVERED;
6835 	}
6836 
6837 	pci_cleanup_aer_uncorrect_error_status(pdev);
6838 
6839 	return result;
6840 }
6841 
6842 /**
6843  * e1000_io_resume - called when traffic can start flowing again.
6844  * @pdev: Pointer to PCI device
6845  *
6846  * This callback is called when the error recovery driver tells us that
6847  * its OK to resume normal operation. Implementation resembles the
6848  * second-half of the e1000e_pm_resume routine.
6849  */
6850 static void e1000_io_resume(struct pci_dev *pdev)
6851 {
6852 	struct net_device *netdev = pci_get_drvdata(pdev);
6853 	struct e1000_adapter *adapter = netdev_priv(netdev);
6854 
6855 	e1000_init_manageability_pt(adapter);
6856 
6857 	if (netif_running(netdev))
6858 		e1000e_up(adapter);
6859 
6860 	netif_device_attach(netdev);
6861 
6862 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6863 	 * is up.  For all other cases, let the f/w know that the h/w is now
6864 	 * under the control of the driver.
6865 	 */
6866 	if (!(adapter->flags & FLAG_HAS_AMT))
6867 		e1000e_get_hw_control(adapter);
6868 }
6869 
6870 static void e1000_print_device_info(struct e1000_adapter *adapter)
6871 {
6872 	struct e1000_hw *hw = &adapter->hw;
6873 	struct net_device *netdev = adapter->netdev;
6874 	u32 ret_val;
6875 	u8 pba_str[E1000_PBANUM_LENGTH];
6876 
6877 	/* print bus type/speed/width info */
6878 	e_info("(PCI Express:2.5GT/s:%s) %pM\n",
6879 	       /* bus width */
6880 	       ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
6881 		"Width x1"),
6882 	       /* MAC address */
6883 	       netdev->dev_addr);
6884 	e_info("Intel(R) PRO/%s Network Connection\n",
6885 	       (hw->phy.type == e1000_phy_ife) ? "10/100" : "1000");
6886 	ret_val = e1000_read_pba_string_generic(hw, pba_str,
6887 						E1000_PBANUM_LENGTH);
6888 	if (ret_val)
6889 		strlcpy((char *)pba_str, "Unknown", sizeof(pba_str));
6890 	e_info("MAC: %d, PHY: %d, PBA No: %s\n",
6891 	       hw->mac.type, hw->phy.type, pba_str);
6892 }
6893 
6894 static void e1000_eeprom_checks(struct e1000_adapter *adapter)
6895 {
6896 	struct e1000_hw *hw = &adapter->hw;
6897 	int ret_val;
6898 	u16 buf = 0;
6899 
6900 	if (hw->mac.type != e1000_82573)
6901 		return;
6902 
6903 	ret_val = e1000_read_nvm(hw, NVM_INIT_CONTROL2_REG, 1, &buf);
6904 	le16_to_cpus(&buf);
6905 	if (!ret_val && (!(buf & BIT(0)))) {
6906 		/* Deep Smart Power Down (DSPD) */
6907 		dev_warn(&adapter->pdev->dev,
6908 			 "Warning: detected DSPD enabled in EEPROM\n");
6909 	}
6910 }
6911 
6912 static netdev_features_t e1000_fix_features(struct net_device *netdev,
6913 					    netdev_features_t features)
6914 {
6915 	struct e1000_adapter *adapter = netdev_priv(netdev);
6916 	struct e1000_hw *hw = &adapter->hw;
6917 
6918 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
6919 	if ((hw->mac.type >= e1000_pch2lan) && (netdev->mtu > ETH_DATA_LEN))
6920 		features &= ~NETIF_F_RXFCS;
6921 
6922 	/* Since there is no support for separate Rx/Tx vlan accel
6923 	 * enable/disable make sure Tx flag is always in same state as Rx.
6924 	 */
6925 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
6926 		features |= NETIF_F_HW_VLAN_CTAG_TX;
6927 	else
6928 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
6929 
6930 	return features;
6931 }
6932 
6933 static int e1000_set_features(struct net_device *netdev,
6934 			      netdev_features_t features)
6935 {
6936 	struct e1000_adapter *adapter = netdev_priv(netdev);
6937 	netdev_features_t changed = features ^ netdev->features;
6938 
6939 	if (changed & (NETIF_F_TSO | NETIF_F_TSO6))
6940 		adapter->flags |= FLAG_TSO_FORCE;
6941 
6942 	if (!(changed & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX |
6943 			 NETIF_F_RXCSUM | NETIF_F_RXHASH | NETIF_F_RXFCS |
6944 			 NETIF_F_RXALL)))
6945 		return 0;
6946 
6947 	if (changed & NETIF_F_RXFCS) {
6948 		if (features & NETIF_F_RXFCS) {
6949 			adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
6950 		} else {
6951 			/* We need to take it back to defaults, which might mean
6952 			 * stripping is still disabled at the adapter level.
6953 			 */
6954 			if (adapter->flags2 & FLAG2_DFLT_CRC_STRIPPING)
6955 				adapter->flags2 |= FLAG2_CRC_STRIPPING;
6956 			else
6957 				adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
6958 		}
6959 	}
6960 
6961 	netdev->features = features;
6962 
6963 	if (netif_running(netdev))
6964 		e1000e_reinit_locked(adapter);
6965 	else
6966 		e1000e_reset(adapter);
6967 
6968 	return 0;
6969 }
6970 
6971 static const struct net_device_ops e1000e_netdev_ops = {
6972 	.ndo_open		= e1000e_open,
6973 	.ndo_stop		= e1000e_close,
6974 	.ndo_start_xmit		= e1000_xmit_frame,
6975 	.ndo_get_stats64	= e1000e_get_stats64,
6976 	.ndo_set_rx_mode	= e1000e_set_rx_mode,
6977 	.ndo_set_mac_address	= e1000_set_mac,
6978 	.ndo_change_mtu		= e1000_change_mtu,
6979 	.ndo_do_ioctl		= e1000_ioctl,
6980 	.ndo_tx_timeout		= e1000_tx_timeout,
6981 	.ndo_validate_addr	= eth_validate_addr,
6982 
6983 	.ndo_vlan_rx_add_vid	= e1000_vlan_rx_add_vid,
6984 	.ndo_vlan_rx_kill_vid	= e1000_vlan_rx_kill_vid,
6985 #ifdef CONFIG_NET_POLL_CONTROLLER
6986 	.ndo_poll_controller	= e1000_netpoll,
6987 #endif
6988 	.ndo_set_features = e1000_set_features,
6989 	.ndo_fix_features = e1000_fix_features,
6990 	.ndo_features_check	= passthru_features_check,
6991 };
6992 
6993 /**
6994  * e1000_probe - Device Initialization Routine
6995  * @pdev: PCI device information struct
6996  * @ent: entry in e1000_pci_tbl
6997  *
6998  * Returns 0 on success, negative on failure
6999  *
7000  * e1000_probe initializes an adapter identified by a pci_dev structure.
7001  * The OS initialization, configuring of the adapter private structure,
7002  * and a hardware reset occur.
7003  **/
7004 static int e1000_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7005 {
7006 	struct net_device *netdev;
7007 	struct e1000_adapter *adapter;
7008 	struct e1000_hw *hw;
7009 	const struct e1000_info *ei = e1000_info_tbl[ent->driver_data];
7010 	resource_size_t mmio_start, mmio_len;
7011 	resource_size_t flash_start, flash_len;
7012 	static int cards_found;
7013 	u16 aspm_disable_flag = 0;
7014 	int bars, i, err, pci_using_dac;
7015 	u16 eeprom_data = 0;
7016 	u16 eeprom_apme_mask = E1000_EEPROM_APME;
7017 	s32 ret_val = 0;
7018 
7019 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L0S)
7020 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7021 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L1)
7022 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7023 	if (aspm_disable_flag)
7024 		e1000e_disable_aspm(pdev, aspm_disable_flag);
7025 
7026 	err = pci_enable_device_mem(pdev);
7027 	if (err)
7028 		return err;
7029 
7030 	pci_using_dac = 0;
7031 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
7032 	if (!err) {
7033 		pci_using_dac = 1;
7034 	} else {
7035 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
7036 		if (err) {
7037 			dev_err(&pdev->dev,
7038 				"No usable DMA configuration, aborting\n");
7039 			goto err_dma;
7040 		}
7041 	}
7042 
7043 	bars = pci_select_bars(pdev, IORESOURCE_MEM);
7044 	err = pci_request_selected_regions_exclusive(pdev, bars,
7045 						     e1000e_driver_name);
7046 	if (err)
7047 		goto err_pci_reg;
7048 
7049 	/* AER (Advanced Error Reporting) hooks */
7050 	pci_enable_pcie_error_reporting(pdev);
7051 
7052 	pci_set_master(pdev);
7053 	/* PCI config space info */
7054 	err = pci_save_state(pdev);
7055 	if (err)
7056 		goto err_alloc_etherdev;
7057 
7058 	err = -ENOMEM;
7059 	netdev = alloc_etherdev(sizeof(struct e1000_adapter));
7060 	if (!netdev)
7061 		goto err_alloc_etherdev;
7062 
7063 	SET_NETDEV_DEV(netdev, &pdev->dev);
7064 
7065 	netdev->irq = pdev->irq;
7066 
7067 	pci_set_drvdata(pdev, netdev);
7068 	adapter = netdev_priv(netdev);
7069 	hw = &adapter->hw;
7070 	adapter->netdev = netdev;
7071 	adapter->pdev = pdev;
7072 	adapter->ei = ei;
7073 	adapter->pba = ei->pba;
7074 	adapter->flags = ei->flags;
7075 	adapter->flags2 = ei->flags2;
7076 	adapter->hw.adapter = adapter;
7077 	adapter->hw.mac.type = ei->mac;
7078 	adapter->max_hw_frame_size = ei->max_hw_frame_size;
7079 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
7080 
7081 	mmio_start = pci_resource_start(pdev, 0);
7082 	mmio_len = pci_resource_len(pdev, 0);
7083 
7084 	err = -EIO;
7085 	adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
7086 	if (!adapter->hw.hw_addr)
7087 		goto err_ioremap;
7088 
7089 	if ((adapter->flags & FLAG_HAS_FLASH) &&
7090 	    (pci_resource_flags(pdev, 1) & IORESOURCE_MEM) &&
7091 	    (hw->mac.type < e1000_pch_spt)) {
7092 		flash_start = pci_resource_start(pdev, 1);
7093 		flash_len = pci_resource_len(pdev, 1);
7094 		adapter->hw.flash_address = ioremap(flash_start, flash_len);
7095 		if (!adapter->hw.flash_address)
7096 			goto err_flashmap;
7097 	}
7098 
7099 	/* Set default EEE advertisement */
7100 	if (adapter->flags2 & FLAG2_HAS_EEE)
7101 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
7102 
7103 	/* construct the net_device struct */
7104 	netdev->netdev_ops = &e1000e_netdev_ops;
7105 	e1000e_set_ethtool_ops(netdev);
7106 	netdev->watchdog_timeo = 5 * HZ;
7107 	netif_napi_add(netdev, &adapter->napi, e1000e_poll, 64);
7108 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
7109 
7110 	netdev->mem_start = mmio_start;
7111 	netdev->mem_end = mmio_start + mmio_len;
7112 
7113 	adapter->bd_number = cards_found++;
7114 
7115 	e1000e_check_options(adapter);
7116 
7117 	/* setup adapter struct */
7118 	err = e1000_sw_init(adapter);
7119 	if (err)
7120 		goto err_sw_init;
7121 
7122 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
7123 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
7124 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
7125 
7126 	err = ei->get_variants(adapter);
7127 	if (err)
7128 		goto err_hw_init;
7129 
7130 	if ((adapter->flags & FLAG_IS_ICH) &&
7131 	    (adapter->flags & FLAG_READ_ONLY_NVM) &&
7132 	    (hw->mac.type < e1000_pch_spt))
7133 		e1000e_write_protect_nvm_ich8lan(&adapter->hw);
7134 
7135 	hw->mac.ops.get_bus_info(&adapter->hw);
7136 
7137 	adapter->hw.phy.autoneg_wait_to_complete = 0;
7138 
7139 	/* Copper options */
7140 	if (adapter->hw.phy.media_type == e1000_media_type_copper) {
7141 		adapter->hw.phy.mdix = AUTO_ALL_MODES;
7142 		adapter->hw.phy.disable_polarity_correction = 0;
7143 		adapter->hw.phy.ms_type = e1000_ms_hw_default;
7144 	}
7145 
7146 	if (hw->phy.ops.check_reset_block && hw->phy.ops.check_reset_block(hw))
7147 		dev_info(&pdev->dev,
7148 			 "PHY reset is blocked due to SOL/IDER session.\n");
7149 
7150 	/* Set initial default active device features */
7151 	netdev->features = (NETIF_F_SG |
7152 			    NETIF_F_HW_VLAN_CTAG_RX |
7153 			    NETIF_F_HW_VLAN_CTAG_TX |
7154 			    NETIF_F_TSO |
7155 			    NETIF_F_TSO6 |
7156 			    NETIF_F_RXHASH |
7157 			    NETIF_F_RXCSUM |
7158 			    NETIF_F_HW_CSUM);
7159 
7160 	/* Set user-changeable features (subset of all device features) */
7161 	netdev->hw_features = netdev->features;
7162 	netdev->hw_features |= NETIF_F_RXFCS;
7163 	netdev->priv_flags |= IFF_SUPP_NOFCS;
7164 	netdev->hw_features |= NETIF_F_RXALL;
7165 
7166 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER)
7167 		netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
7168 
7169 	netdev->vlan_features |= (NETIF_F_SG |
7170 				  NETIF_F_TSO |
7171 				  NETIF_F_TSO6 |
7172 				  NETIF_F_HW_CSUM);
7173 
7174 	netdev->priv_flags |= IFF_UNICAST_FLT;
7175 
7176 	if (pci_using_dac) {
7177 		netdev->features |= NETIF_F_HIGHDMA;
7178 		netdev->vlan_features |= NETIF_F_HIGHDMA;
7179 	}
7180 
7181 	/* MTU range: 68 - max_hw_frame_size */
7182 	netdev->min_mtu = ETH_MIN_MTU;
7183 	netdev->max_mtu = adapter->max_hw_frame_size -
7184 			  (VLAN_ETH_HLEN + ETH_FCS_LEN);
7185 
7186 	if (e1000e_enable_mng_pass_thru(&adapter->hw))
7187 		adapter->flags |= FLAG_MNG_PT_ENABLED;
7188 
7189 	/* before reading the NVM, reset the controller to
7190 	 * put the device in a known good starting state
7191 	 */
7192 	adapter->hw.mac.ops.reset_hw(&adapter->hw);
7193 
7194 	/* systems with ASPM and others may see the checksum fail on the first
7195 	 * attempt. Let's give it a few tries
7196 	 */
7197 	for (i = 0;; i++) {
7198 		if (e1000_validate_nvm_checksum(&adapter->hw) >= 0)
7199 			break;
7200 		if (i == 2) {
7201 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
7202 			err = -EIO;
7203 			goto err_eeprom;
7204 		}
7205 	}
7206 
7207 	e1000_eeprom_checks(adapter);
7208 
7209 	/* copy the MAC address */
7210 	if (e1000e_read_mac_addr(&adapter->hw))
7211 		dev_err(&pdev->dev,
7212 			"NVM Read Error while reading MAC address\n");
7213 
7214 	memcpy(netdev->dev_addr, adapter->hw.mac.addr, netdev->addr_len);
7215 
7216 	if (!is_valid_ether_addr(netdev->dev_addr)) {
7217 		dev_err(&pdev->dev, "Invalid MAC Address: %pM\n",
7218 			netdev->dev_addr);
7219 		err = -EIO;
7220 		goto err_eeprom;
7221 	}
7222 
7223 	init_timer(&adapter->watchdog_timer);
7224 	adapter->watchdog_timer.function = e1000_watchdog;
7225 	adapter->watchdog_timer.data = (unsigned long)adapter;
7226 
7227 	init_timer(&adapter->phy_info_timer);
7228 	adapter->phy_info_timer.function = e1000_update_phy_info;
7229 	adapter->phy_info_timer.data = (unsigned long)adapter;
7230 
7231 	INIT_WORK(&adapter->reset_task, e1000_reset_task);
7232 	INIT_WORK(&adapter->watchdog_task, e1000_watchdog_task);
7233 	INIT_WORK(&adapter->downshift_task, e1000e_downshift_workaround);
7234 	INIT_WORK(&adapter->update_phy_task, e1000e_update_phy_task);
7235 	INIT_WORK(&adapter->print_hang_task, e1000_print_hw_hang);
7236 
7237 	/* Initialize link parameters. User can change them with ethtool */
7238 	adapter->hw.mac.autoneg = 1;
7239 	adapter->fc_autoneg = true;
7240 	adapter->hw.fc.requested_mode = e1000_fc_default;
7241 	adapter->hw.fc.current_mode = e1000_fc_default;
7242 	adapter->hw.phy.autoneg_advertised = 0x2f;
7243 
7244 	/* Initial Wake on LAN setting - If APM wake is enabled in
7245 	 * the EEPROM, enable the ACPI Magic Packet filter
7246 	 */
7247 	if (adapter->flags & FLAG_APME_IN_WUC) {
7248 		/* APME bit in EEPROM is mapped to WUC.APME */
7249 		eeprom_data = er32(WUC);
7250 		eeprom_apme_mask = E1000_WUC_APME;
7251 		if ((hw->mac.type > e1000_ich10lan) &&
7252 		    (eeprom_data & E1000_WUC_PHY_WAKE))
7253 			adapter->flags2 |= FLAG2_HAS_PHY_WAKEUP;
7254 	} else if (adapter->flags & FLAG_APME_IN_CTRL3) {
7255 		if (adapter->flags & FLAG_APME_CHECK_PORT_B &&
7256 		    (adapter->hw.bus.func == 1))
7257 			ret_val = e1000_read_nvm(&adapter->hw,
7258 					      NVM_INIT_CONTROL3_PORT_B,
7259 					      1, &eeprom_data);
7260 		else
7261 			ret_val = e1000_read_nvm(&adapter->hw,
7262 					      NVM_INIT_CONTROL3_PORT_A,
7263 					      1, &eeprom_data);
7264 	}
7265 
7266 	/* fetch WoL from EEPROM */
7267 	if (ret_val)
7268 		e_dbg("NVM read error getting WoL initial values: %d\n", ret_val);
7269 	else if (eeprom_data & eeprom_apme_mask)
7270 		adapter->eeprom_wol |= E1000_WUFC_MAG;
7271 
7272 	/* now that we have the eeprom settings, apply the special cases
7273 	 * where the eeprom may be wrong or the board simply won't support
7274 	 * wake on lan on a particular port
7275 	 */
7276 	if (!(adapter->flags & FLAG_HAS_WOL))
7277 		adapter->eeprom_wol = 0;
7278 
7279 	/* initialize the wol settings based on the eeprom settings */
7280 	adapter->wol = adapter->eeprom_wol;
7281 
7282 	/* make sure adapter isn't asleep if manageability is enabled */
7283 	if (adapter->wol || (adapter->flags & FLAG_MNG_PT_ENABLED) ||
7284 	    (hw->mac.ops.check_mng_mode(hw)))
7285 		device_wakeup_enable(&pdev->dev);
7286 
7287 	/* save off EEPROM version number */
7288 	ret_val = e1000_read_nvm(&adapter->hw, 5, 1, &adapter->eeprom_vers);
7289 
7290 	if (ret_val) {
7291 		e_dbg("NVM read error getting EEPROM version: %d\n", ret_val);
7292 		adapter->eeprom_vers = 0;
7293 	}
7294 
7295 	/* init PTP hardware clock */
7296 	e1000e_ptp_init(adapter);
7297 
7298 	/* reset the hardware with the new settings */
7299 	e1000e_reset(adapter);
7300 
7301 	/* If the controller has AMT, do not set DRV_LOAD until the interface
7302 	 * is up.  For all other cases, let the f/w know that the h/w is now
7303 	 * under the control of the driver.
7304 	 */
7305 	if (!(adapter->flags & FLAG_HAS_AMT))
7306 		e1000e_get_hw_control(adapter);
7307 
7308 	strlcpy(netdev->name, "eth%d", sizeof(netdev->name));
7309 	err = register_netdev(netdev);
7310 	if (err)
7311 		goto err_register;
7312 
7313 	/* carrier off reporting is important to ethtool even BEFORE open */
7314 	netif_carrier_off(netdev);
7315 
7316 	e1000_print_device_info(adapter);
7317 
7318 	if (pci_dev_run_wake(pdev))
7319 		pm_runtime_put_noidle(&pdev->dev);
7320 
7321 	return 0;
7322 
7323 err_register:
7324 	if (!(adapter->flags & FLAG_HAS_AMT))
7325 		e1000e_release_hw_control(adapter);
7326 err_eeprom:
7327 	if (hw->phy.ops.check_reset_block && !hw->phy.ops.check_reset_block(hw))
7328 		e1000_phy_hw_reset(&adapter->hw);
7329 err_hw_init:
7330 	kfree(adapter->tx_ring);
7331 	kfree(adapter->rx_ring);
7332 err_sw_init:
7333 	if ((adapter->hw.flash_address) && (hw->mac.type < e1000_pch_spt))
7334 		iounmap(adapter->hw.flash_address);
7335 	e1000e_reset_interrupt_capability(adapter);
7336 err_flashmap:
7337 	iounmap(adapter->hw.hw_addr);
7338 err_ioremap:
7339 	free_netdev(netdev);
7340 err_alloc_etherdev:
7341 	pci_release_mem_regions(pdev);
7342 err_pci_reg:
7343 err_dma:
7344 	pci_disable_device(pdev);
7345 	return err;
7346 }
7347 
7348 /**
7349  * e1000_remove - Device Removal Routine
7350  * @pdev: PCI device information struct
7351  *
7352  * e1000_remove is called by the PCI subsystem to alert the driver
7353  * that it should release a PCI device.  The could be caused by a
7354  * Hot-Plug event, or because the driver is going to be removed from
7355  * memory.
7356  **/
7357 static void e1000_remove(struct pci_dev *pdev)
7358 {
7359 	struct net_device *netdev = pci_get_drvdata(pdev);
7360 	struct e1000_adapter *adapter = netdev_priv(netdev);
7361 	bool down = test_bit(__E1000_DOWN, &adapter->state);
7362 
7363 	e1000e_ptp_remove(adapter);
7364 
7365 	/* The timers may be rescheduled, so explicitly disable them
7366 	 * from being rescheduled.
7367 	 */
7368 	if (!down)
7369 		set_bit(__E1000_DOWN, &adapter->state);
7370 	del_timer_sync(&adapter->watchdog_timer);
7371 	del_timer_sync(&adapter->phy_info_timer);
7372 
7373 	cancel_work_sync(&adapter->reset_task);
7374 	cancel_work_sync(&adapter->watchdog_task);
7375 	cancel_work_sync(&adapter->downshift_task);
7376 	cancel_work_sync(&adapter->update_phy_task);
7377 	cancel_work_sync(&adapter->print_hang_task);
7378 
7379 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
7380 		cancel_work_sync(&adapter->tx_hwtstamp_work);
7381 		if (adapter->tx_hwtstamp_skb) {
7382 			dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
7383 			adapter->tx_hwtstamp_skb = NULL;
7384 		}
7385 	}
7386 
7387 	/* Don't lie to e1000_close() down the road. */
7388 	if (!down)
7389 		clear_bit(__E1000_DOWN, &adapter->state);
7390 	unregister_netdev(netdev);
7391 
7392 	if (pci_dev_run_wake(pdev))
7393 		pm_runtime_get_noresume(&pdev->dev);
7394 
7395 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
7396 	 * would have already happened in close and is redundant.
7397 	 */
7398 	e1000e_release_hw_control(adapter);
7399 
7400 	e1000e_reset_interrupt_capability(adapter);
7401 	kfree(adapter->tx_ring);
7402 	kfree(adapter->rx_ring);
7403 
7404 	iounmap(adapter->hw.hw_addr);
7405 	if ((adapter->hw.flash_address) &&
7406 	    (adapter->hw.mac.type < e1000_pch_spt))
7407 		iounmap(adapter->hw.flash_address);
7408 	pci_release_mem_regions(pdev);
7409 
7410 	free_netdev(netdev);
7411 
7412 	/* AER disable */
7413 	pci_disable_pcie_error_reporting(pdev);
7414 
7415 	pci_disable_device(pdev);
7416 }
7417 
7418 /* PCI Error Recovery (ERS) */
7419 static const struct pci_error_handlers e1000_err_handler = {
7420 	.error_detected = e1000_io_error_detected,
7421 	.slot_reset = e1000_io_slot_reset,
7422 	.resume = e1000_io_resume,
7423 };
7424 
7425 static const struct pci_device_id e1000_pci_tbl[] = {
7426 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_COPPER), board_82571 },
7427 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_FIBER), board_82571 },
7428 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER), board_82571 },
7429 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER_LP),
7430 	  board_82571 },
7431 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_FIBER), board_82571 },
7432 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES), board_82571 },
7433 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_DUAL), board_82571 },
7434 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_QUAD), board_82571 },
7435 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571PT_QUAD_COPPER), board_82571 },
7436 
7437 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI), board_82572 },
7438 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_COPPER), board_82572 },
7439 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_FIBER), board_82572 },
7440 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_SERDES), board_82572 },
7441 
7442 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E), board_82573 },
7443 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E_IAMT), board_82573 },
7444 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573L), board_82573 },
7445 
7446 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574L), board_82574 },
7447 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574LA), board_82574 },
7448 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82583V), board_82583 },
7449 
7450 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_DPT),
7451 	  board_80003es2lan },
7452 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_SPT),
7453 	  board_80003es2lan },
7454 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_DPT),
7455 	  board_80003es2lan },
7456 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_SPT),
7457 	  board_80003es2lan },
7458 
7459 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE), board_ich8lan },
7460 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_G), board_ich8lan },
7461 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_GT), board_ich8lan },
7462 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_AMT), board_ich8lan },
7463 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_C), board_ich8lan },
7464 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M), board_ich8lan },
7465 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M_AMT), board_ich8lan },
7466 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_82567V_3), board_ich8lan },
7467 
7468 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE), board_ich9lan },
7469 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_G), board_ich9lan },
7470 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_GT), board_ich9lan },
7471 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_AMT), board_ich9lan },
7472 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_C), board_ich9lan },
7473 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_BM), board_ich9lan },
7474 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M), board_ich9lan },
7475 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_AMT), board_ich9lan },
7476 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_V), board_ich9lan },
7477 
7478 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LM), board_ich9lan },
7479 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LF), board_ich9lan },
7480 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_V), board_ich9lan },
7481 
7482 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LM), board_ich10lan },
7483 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LF), board_ich10lan },
7484 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_V), board_ich10lan },
7485 
7486 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LM), board_pchlan },
7487 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LC), board_pchlan },
7488 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DM), board_pchlan },
7489 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DC), board_pchlan },
7490 
7491 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_LM), board_pch2lan },
7492 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_V), board_pch2lan },
7493 
7494 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_LM), board_pch_lpt },
7495 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_V), board_pch_lpt },
7496 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_LM), board_pch_lpt },
7497 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_V), board_pch_lpt },
7498 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM2), board_pch_lpt },
7499 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V2), board_pch_lpt },
7500 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM3), board_pch_lpt },
7501 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V3), board_pch_lpt },
7502 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM), board_pch_spt },
7503 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V), board_pch_spt },
7504 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM2), board_pch_spt },
7505 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V2), board_pch_spt },
7506 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LBG_I219_LM3), board_pch_spt },
7507 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM4), board_pch_spt },
7508 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V4), board_pch_spt },
7509 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM5), board_pch_spt },
7510 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V5), board_pch_spt },
7511 
7512 	{ 0, 0, 0, 0, 0, 0, 0 }	/* terminate list */
7513 };
7514 MODULE_DEVICE_TABLE(pci, e1000_pci_tbl);
7515 
7516 static const struct dev_pm_ops e1000_pm_ops = {
7517 #ifdef CONFIG_PM_SLEEP
7518 	.suspend	= e1000e_pm_suspend,
7519 	.resume		= e1000e_pm_resume,
7520 	.freeze		= e1000e_pm_freeze,
7521 	.thaw		= e1000e_pm_thaw,
7522 	.poweroff	= e1000e_pm_suspend,
7523 	.restore	= e1000e_pm_resume,
7524 #endif
7525 	SET_RUNTIME_PM_OPS(e1000e_pm_runtime_suspend, e1000e_pm_runtime_resume,
7526 			   e1000e_pm_runtime_idle)
7527 };
7528 
7529 /* PCI Device API Driver */
7530 static struct pci_driver e1000_driver = {
7531 	.name     = e1000e_driver_name,
7532 	.id_table = e1000_pci_tbl,
7533 	.probe    = e1000_probe,
7534 	.remove   = e1000_remove,
7535 	.driver   = {
7536 		.pm = &e1000_pm_ops,
7537 	},
7538 	.shutdown = e1000_shutdown,
7539 	.err_handler = &e1000_err_handler
7540 };
7541 
7542 /**
7543  * e1000_init_module - Driver Registration Routine
7544  *
7545  * e1000_init_module is the first routine called when the driver is
7546  * loaded. All it does is register with the PCI subsystem.
7547  **/
7548 static int __init e1000_init_module(void)
7549 {
7550 	pr_info("Intel(R) PRO/1000 Network Driver - %s\n",
7551 		e1000e_driver_version);
7552 	pr_info("Copyright(c) 1999 - 2015 Intel Corporation.\n");
7553 
7554 	return pci_register_driver(&e1000_driver);
7555 }
7556 module_init(e1000_init_module);
7557 
7558 /**
7559  * e1000_exit_module - Driver Exit Cleanup Routine
7560  *
7561  * e1000_exit_module is called just before the driver is removed
7562  * from memory.
7563  **/
7564 static void __exit e1000_exit_module(void)
7565 {
7566 	pci_unregister_driver(&e1000_driver);
7567 }
7568 module_exit(e1000_exit_module);
7569 
7570 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
7571 MODULE_DESCRIPTION("Intel(R) PRO/1000 Network Driver");
7572 MODULE_LICENSE("GPL");
7573 MODULE_VERSION(DRV_VERSION);
7574 
7575 /* netdev.c */
7576