xref: /linux/drivers/net/ethernet/intel/e1000e/netdev.c (revision b0d5c81e872ed21de1e56feb0fa6e4161da7be61)
1 /* Intel PRO/1000 Linux driver
2  * Copyright(c) 1999 - 2015 Intel Corporation.
3  *
4  * This program is free software; you can redistribute it and/or modify it
5  * under the terms and conditions of the GNU General Public License,
6  * version 2, as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope it will be useful, but WITHOUT
9  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
11  * more details.
12  *
13  * The full GNU General Public License is included in this distribution in
14  * the file called "COPYING".
15  *
16  * Contact Information:
17  * Linux NICS <linux.nics@intel.com>
18  * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
19  * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
20  */
21 
22 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
23 
24 #include <linux/module.h>
25 #include <linux/types.h>
26 #include <linux/init.h>
27 #include <linux/pci.h>
28 #include <linux/vmalloc.h>
29 #include <linux/pagemap.h>
30 #include <linux/delay.h>
31 #include <linux/netdevice.h>
32 #include <linux/interrupt.h>
33 #include <linux/tcp.h>
34 #include <linux/ipv6.h>
35 #include <linux/slab.h>
36 #include <net/checksum.h>
37 #include <net/ip6_checksum.h>
38 #include <linux/ethtool.h>
39 #include <linux/if_vlan.h>
40 #include <linux/cpu.h>
41 #include <linux/smp.h>
42 #include <linux/pm_qos.h>
43 #include <linux/pm_runtime.h>
44 #include <linux/aer.h>
45 #include <linux/prefetch.h>
46 
47 #include "e1000.h"
48 
49 #define DRV_EXTRAVERSION "-k"
50 
51 #define DRV_VERSION "3.2.6" DRV_EXTRAVERSION
52 char e1000e_driver_name[] = "e1000e";
53 const char e1000e_driver_version[] = DRV_VERSION;
54 
55 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
56 static int debug = -1;
57 module_param(debug, int, 0);
58 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
59 
60 static const struct e1000_info *e1000_info_tbl[] = {
61 	[board_82571]		= &e1000_82571_info,
62 	[board_82572]		= &e1000_82572_info,
63 	[board_82573]		= &e1000_82573_info,
64 	[board_82574]		= &e1000_82574_info,
65 	[board_82583]		= &e1000_82583_info,
66 	[board_80003es2lan]	= &e1000_es2_info,
67 	[board_ich8lan]		= &e1000_ich8_info,
68 	[board_ich9lan]		= &e1000_ich9_info,
69 	[board_ich10lan]	= &e1000_ich10_info,
70 	[board_pchlan]		= &e1000_pch_info,
71 	[board_pch2lan]		= &e1000_pch2_info,
72 	[board_pch_lpt]		= &e1000_pch_lpt_info,
73 	[board_pch_spt]		= &e1000_pch_spt_info,
74 	[board_pch_cnp]		= &e1000_pch_cnp_info,
75 };
76 
77 struct e1000_reg_info {
78 	u32 ofs;
79 	char *name;
80 };
81 
82 static const struct e1000_reg_info e1000_reg_info_tbl[] = {
83 	/* General Registers */
84 	{E1000_CTRL, "CTRL"},
85 	{E1000_STATUS, "STATUS"},
86 	{E1000_CTRL_EXT, "CTRL_EXT"},
87 
88 	/* Interrupt Registers */
89 	{E1000_ICR, "ICR"},
90 
91 	/* Rx Registers */
92 	{E1000_RCTL, "RCTL"},
93 	{E1000_RDLEN(0), "RDLEN"},
94 	{E1000_RDH(0), "RDH"},
95 	{E1000_RDT(0), "RDT"},
96 	{E1000_RDTR, "RDTR"},
97 	{E1000_RXDCTL(0), "RXDCTL"},
98 	{E1000_ERT, "ERT"},
99 	{E1000_RDBAL(0), "RDBAL"},
100 	{E1000_RDBAH(0), "RDBAH"},
101 	{E1000_RDFH, "RDFH"},
102 	{E1000_RDFT, "RDFT"},
103 	{E1000_RDFHS, "RDFHS"},
104 	{E1000_RDFTS, "RDFTS"},
105 	{E1000_RDFPC, "RDFPC"},
106 
107 	/* Tx Registers */
108 	{E1000_TCTL, "TCTL"},
109 	{E1000_TDBAL(0), "TDBAL"},
110 	{E1000_TDBAH(0), "TDBAH"},
111 	{E1000_TDLEN(0), "TDLEN"},
112 	{E1000_TDH(0), "TDH"},
113 	{E1000_TDT(0), "TDT"},
114 	{E1000_TIDV, "TIDV"},
115 	{E1000_TXDCTL(0), "TXDCTL"},
116 	{E1000_TADV, "TADV"},
117 	{E1000_TARC(0), "TARC"},
118 	{E1000_TDFH, "TDFH"},
119 	{E1000_TDFT, "TDFT"},
120 	{E1000_TDFHS, "TDFHS"},
121 	{E1000_TDFTS, "TDFTS"},
122 	{E1000_TDFPC, "TDFPC"},
123 
124 	/* List Terminator */
125 	{0, NULL}
126 };
127 
128 /**
129  * __ew32_prepare - prepare to write to MAC CSR register on certain parts
130  * @hw: pointer to the HW structure
131  *
132  * When updating the MAC CSR registers, the Manageability Engine (ME) could
133  * be accessing the registers at the same time.  Normally, this is handled in
134  * h/w by an arbiter but on some parts there is a bug that acknowledges Host
135  * accesses later than it should which could result in the register to have
136  * an incorrect value.  Workaround this by checking the FWSM register which
137  * has bit 24 set while ME is accessing MAC CSR registers, wait if it is set
138  * and try again a number of times.
139  **/
140 s32 __ew32_prepare(struct e1000_hw *hw)
141 {
142 	s32 i = E1000_ICH_FWSM_PCIM2PCI_COUNT;
143 
144 	while ((er32(FWSM) & E1000_ICH_FWSM_PCIM2PCI) && --i)
145 		udelay(50);
146 
147 	return i;
148 }
149 
150 void __ew32(struct e1000_hw *hw, unsigned long reg, u32 val)
151 {
152 	if (hw->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
153 		__ew32_prepare(hw);
154 
155 	writel(val, hw->hw_addr + reg);
156 }
157 
158 /**
159  * e1000_regdump - register printout routine
160  * @hw: pointer to the HW structure
161  * @reginfo: pointer to the register info table
162  **/
163 static void e1000_regdump(struct e1000_hw *hw, struct e1000_reg_info *reginfo)
164 {
165 	int n = 0;
166 	char rname[16];
167 	u32 regs[8];
168 
169 	switch (reginfo->ofs) {
170 	case E1000_RXDCTL(0):
171 		for (n = 0; n < 2; n++)
172 			regs[n] = __er32(hw, E1000_RXDCTL(n));
173 		break;
174 	case E1000_TXDCTL(0):
175 		for (n = 0; n < 2; n++)
176 			regs[n] = __er32(hw, E1000_TXDCTL(n));
177 		break;
178 	case E1000_TARC(0):
179 		for (n = 0; n < 2; n++)
180 			regs[n] = __er32(hw, E1000_TARC(n));
181 		break;
182 	default:
183 		pr_info("%-15s %08x\n",
184 			reginfo->name, __er32(hw, reginfo->ofs));
185 		return;
186 	}
187 
188 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-1]");
189 	pr_info("%-15s %08x %08x\n", rname, regs[0], regs[1]);
190 }
191 
192 static void e1000e_dump_ps_pages(struct e1000_adapter *adapter,
193 				 struct e1000_buffer *bi)
194 {
195 	int i;
196 	struct e1000_ps_page *ps_page;
197 
198 	for (i = 0; i < adapter->rx_ps_pages; i++) {
199 		ps_page = &bi->ps_pages[i];
200 
201 		if (ps_page->page) {
202 			pr_info("packet dump for ps_page %d:\n", i);
203 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
204 				       16, 1, page_address(ps_page->page),
205 				       PAGE_SIZE, true);
206 		}
207 	}
208 }
209 
210 /**
211  * e1000e_dump - Print registers, Tx-ring and Rx-ring
212  * @adapter: board private structure
213  **/
214 static void e1000e_dump(struct e1000_adapter *adapter)
215 {
216 	struct net_device *netdev = adapter->netdev;
217 	struct e1000_hw *hw = &adapter->hw;
218 	struct e1000_reg_info *reginfo;
219 	struct e1000_ring *tx_ring = adapter->tx_ring;
220 	struct e1000_tx_desc *tx_desc;
221 	struct my_u0 {
222 		__le64 a;
223 		__le64 b;
224 	} *u0;
225 	struct e1000_buffer *buffer_info;
226 	struct e1000_ring *rx_ring = adapter->rx_ring;
227 	union e1000_rx_desc_packet_split *rx_desc_ps;
228 	union e1000_rx_desc_extended *rx_desc;
229 	struct my_u1 {
230 		__le64 a;
231 		__le64 b;
232 		__le64 c;
233 		__le64 d;
234 	} *u1;
235 	u32 staterr;
236 	int i = 0;
237 
238 	if (!netif_msg_hw(adapter))
239 		return;
240 
241 	/* Print netdevice Info */
242 	if (netdev) {
243 		dev_info(&adapter->pdev->dev, "Net device Info\n");
244 		pr_info("Device Name     state            trans_start\n");
245 		pr_info("%-15s %016lX %016lX\n", netdev->name,
246 			netdev->state, dev_trans_start(netdev));
247 	}
248 
249 	/* Print Registers */
250 	dev_info(&adapter->pdev->dev, "Register Dump\n");
251 	pr_info(" Register Name   Value\n");
252 	for (reginfo = (struct e1000_reg_info *)e1000_reg_info_tbl;
253 	     reginfo->name; reginfo++) {
254 		e1000_regdump(hw, reginfo);
255 	}
256 
257 	/* Print Tx Ring Summary */
258 	if (!netdev || !netif_running(netdev))
259 		return;
260 
261 	dev_info(&adapter->pdev->dev, "Tx Ring Summary\n");
262 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
263 	buffer_info = &tx_ring->buffer_info[tx_ring->next_to_clean];
264 	pr_info(" %5d %5X %5X %016llX %04X %3X %016llX\n",
265 		0, tx_ring->next_to_use, tx_ring->next_to_clean,
266 		(unsigned long long)buffer_info->dma,
267 		buffer_info->length,
268 		buffer_info->next_to_watch,
269 		(unsigned long long)buffer_info->time_stamp);
270 
271 	/* Print Tx Ring */
272 	if (!netif_msg_tx_done(adapter))
273 		goto rx_ring_summary;
274 
275 	dev_info(&adapter->pdev->dev, "Tx Ring Dump\n");
276 
277 	/* Transmit Descriptor Formats - DEXT[29] is 0 (Legacy) or 1 (Extended)
278 	 *
279 	 * Legacy Transmit Descriptor
280 	 *   +--------------------------------------------------------------+
281 	 * 0 |         Buffer Address [63:0] (Reserved on Write Back)       |
282 	 *   +--------------------------------------------------------------+
283 	 * 8 | Special  |    CSS     | Status |  CMD    |  CSO   |  Length  |
284 	 *   +--------------------------------------------------------------+
285 	 *   63       48 47        36 35    32 31     24 23    16 15        0
286 	 *
287 	 * Extended Context Descriptor (DTYP=0x0) for TSO or checksum offload
288 	 *   63      48 47    40 39       32 31             16 15    8 7      0
289 	 *   +----------------------------------------------------------------+
290 	 * 0 |  TUCSE  | TUCS0  |   TUCSS   |     IPCSE       | IPCS0 | IPCSS |
291 	 *   +----------------------------------------------------------------+
292 	 * 8 |   MSS   | HDRLEN | RSV | STA | TUCMD | DTYP |      PAYLEN      |
293 	 *   +----------------------------------------------------------------+
294 	 *   63      48 47    40 39 36 35 32 31   24 23  20 19                0
295 	 *
296 	 * Extended Data Descriptor (DTYP=0x1)
297 	 *   +----------------------------------------------------------------+
298 	 * 0 |                     Buffer Address [63:0]                      |
299 	 *   +----------------------------------------------------------------+
300 	 * 8 | VLAN tag |  POPTS  | Rsvd | Status | Command | DTYP |  DTALEN  |
301 	 *   +----------------------------------------------------------------+
302 	 *   63       48 47     40 39  36 35    32 31     24 23  20 19        0
303 	 */
304 	pr_info("Tl[desc]     [address 63:0  ] [SpeCssSCmCsLen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Legacy format\n");
305 	pr_info("Tc[desc]     [Ce CoCsIpceCoS] [MssHlRSCm0Plen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Context format\n");
306 	pr_info("Td[desc]     [address 63:0  ] [VlaPoRSCm1Dlen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Data format\n");
307 	for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
308 		const char *next_desc;
309 		tx_desc = E1000_TX_DESC(*tx_ring, i);
310 		buffer_info = &tx_ring->buffer_info[i];
311 		u0 = (struct my_u0 *)tx_desc;
312 		if (i == tx_ring->next_to_use && i == tx_ring->next_to_clean)
313 			next_desc = " NTC/U";
314 		else if (i == tx_ring->next_to_use)
315 			next_desc = " NTU";
316 		else if (i == tx_ring->next_to_clean)
317 			next_desc = " NTC";
318 		else
319 			next_desc = "";
320 		pr_info("T%c[0x%03X]    %016llX %016llX %016llX %04X  %3X %016llX %p%s\n",
321 			(!(le64_to_cpu(u0->b) & BIT(29)) ? 'l' :
322 			 ((le64_to_cpu(u0->b) & BIT(20)) ? 'd' : 'c')),
323 			i,
324 			(unsigned long long)le64_to_cpu(u0->a),
325 			(unsigned long long)le64_to_cpu(u0->b),
326 			(unsigned long long)buffer_info->dma,
327 			buffer_info->length, buffer_info->next_to_watch,
328 			(unsigned long long)buffer_info->time_stamp,
329 			buffer_info->skb, next_desc);
330 
331 		if (netif_msg_pktdata(adapter) && buffer_info->skb)
332 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
333 				       16, 1, buffer_info->skb->data,
334 				       buffer_info->skb->len, true);
335 	}
336 
337 	/* Print Rx Ring Summary */
338 rx_ring_summary:
339 	dev_info(&adapter->pdev->dev, "Rx Ring Summary\n");
340 	pr_info("Queue [NTU] [NTC]\n");
341 	pr_info(" %5d %5X %5X\n",
342 		0, rx_ring->next_to_use, rx_ring->next_to_clean);
343 
344 	/* Print Rx Ring */
345 	if (!netif_msg_rx_status(adapter))
346 		return;
347 
348 	dev_info(&adapter->pdev->dev, "Rx Ring Dump\n");
349 	switch (adapter->rx_ps_pages) {
350 	case 1:
351 	case 2:
352 	case 3:
353 		/* [Extended] Packet Split Receive Descriptor Format
354 		 *
355 		 *    +-----------------------------------------------------+
356 		 *  0 |                Buffer Address 0 [63:0]              |
357 		 *    +-----------------------------------------------------+
358 		 *  8 |                Buffer Address 1 [63:0]              |
359 		 *    +-----------------------------------------------------+
360 		 * 16 |                Buffer Address 2 [63:0]              |
361 		 *    +-----------------------------------------------------+
362 		 * 24 |                Buffer Address 3 [63:0]              |
363 		 *    +-----------------------------------------------------+
364 		 */
365 		pr_info("R  [desc]      [buffer 0 63:0 ] [buffer 1 63:0 ] [buffer 2 63:0 ] [buffer 3 63:0 ] [bi->dma       ] [bi->skb] <-- Ext Pkt Split format\n");
366 		/* [Extended] Receive Descriptor (Write-Back) Format
367 		 *
368 		 *   63       48 47    32 31     13 12    8 7    4 3        0
369 		 *   +------------------------------------------------------+
370 		 * 0 | Packet   | IP     |  Rsvd   | MRQ   | Rsvd | MRQ RSS |
371 		 *   | Checksum | Ident  |         | Queue |      |  Type   |
372 		 *   +------------------------------------------------------+
373 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
374 		 *   +------------------------------------------------------+
375 		 *   63       48 47    32 31            20 19               0
376 		 */
377 		pr_info("RWB[desc]      [ck ipid mrqhsh] [vl   l0 ee  es] [ l3  l2  l1 hs] [reserved      ] ---------------- [bi->skb] <-- Ext Rx Write-Back format\n");
378 		for (i = 0; i < rx_ring->count; i++) {
379 			const char *next_desc;
380 			buffer_info = &rx_ring->buffer_info[i];
381 			rx_desc_ps = E1000_RX_DESC_PS(*rx_ring, i);
382 			u1 = (struct my_u1 *)rx_desc_ps;
383 			staterr =
384 			    le32_to_cpu(rx_desc_ps->wb.middle.status_error);
385 
386 			if (i == rx_ring->next_to_use)
387 				next_desc = " NTU";
388 			else if (i == rx_ring->next_to_clean)
389 				next_desc = " NTC";
390 			else
391 				next_desc = "";
392 
393 			if (staterr & E1000_RXD_STAT_DD) {
394 				/* Descriptor Done */
395 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX ---------------- %p%s\n",
396 					"RWB", i,
397 					(unsigned long long)le64_to_cpu(u1->a),
398 					(unsigned long long)le64_to_cpu(u1->b),
399 					(unsigned long long)le64_to_cpu(u1->c),
400 					(unsigned long long)le64_to_cpu(u1->d),
401 					buffer_info->skb, next_desc);
402 			} else {
403 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX %016llX %p%s\n",
404 					"R  ", i,
405 					(unsigned long long)le64_to_cpu(u1->a),
406 					(unsigned long long)le64_to_cpu(u1->b),
407 					(unsigned long long)le64_to_cpu(u1->c),
408 					(unsigned long long)le64_to_cpu(u1->d),
409 					(unsigned long long)buffer_info->dma,
410 					buffer_info->skb, next_desc);
411 
412 				if (netif_msg_pktdata(adapter))
413 					e1000e_dump_ps_pages(adapter,
414 							     buffer_info);
415 			}
416 		}
417 		break;
418 	default:
419 	case 0:
420 		/* Extended Receive Descriptor (Read) Format
421 		 *
422 		 *   +-----------------------------------------------------+
423 		 * 0 |                Buffer Address [63:0]                |
424 		 *   +-----------------------------------------------------+
425 		 * 8 |                      Reserved                       |
426 		 *   +-----------------------------------------------------+
427 		 */
428 		pr_info("R  [desc]      [buf addr 63:0 ] [reserved 63:0 ] [bi->dma       ] [bi->skb] <-- Ext (Read) format\n");
429 		/* Extended Receive Descriptor (Write-Back) Format
430 		 *
431 		 *   63       48 47    32 31    24 23            4 3        0
432 		 *   +------------------------------------------------------+
433 		 *   |     RSS Hash      |        |               |         |
434 		 * 0 +-------------------+  Rsvd  |   Reserved    | MRQ RSS |
435 		 *   | Packet   | IP     |        |               |  Type   |
436 		 *   | Checksum | Ident  |        |               |         |
437 		 *   +------------------------------------------------------+
438 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
439 		 *   +------------------------------------------------------+
440 		 *   63       48 47    32 31            20 19               0
441 		 */
442 		pr_info("RWB[desc]      [cs ipid    mrq] [vt   ln xe  xs] [bi->skb] <-- Ext (Write-Back) format\n");
443 
444 		for (i = 0; i < rx_ring->count; i++) {
445 			const char *next_desc;
446 
447 			buffer_info = &rx_ring->buffer_info[i];
448 			rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
449 			u1 = (struct my_u1 *)rx_desc;
450 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
451 
452 			if (i == rx_ring->next_to_use)
453 				next_desc = " NTU";
454 			else if (i == rx_ring->next_to_clean)
455 				next_desc = " NTC";
456 			else
457 				next_desc = "";
458 
459 			if (staterr & E1000_RXD_STAT_DD) {
460 				/* Descriptor Done */
461 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %p%s\n",
462 					"RWB", i,
463 					(unsigned long long)le64_to_cpu(u1->a),
464 					(unsigned long long)le64_to_cpu(u1->b),
465 					buffer_info->skb, next_desc);
466 			} else {
467 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %p%s\n",
468 					"R  ", i,
469 					(unsigned long long)le64_to_cpu(u1->a),
470 					(unsigned long long)le64_to_cpu(u1->b),
471 					(unsigned long long)buffer_info->dma,
472 					buffer_info->skb, next_desc);
473 
474 				if (netif_msg_pktdata(adapter) &&
475 				    buffer_info->skb)
476 					print_hex_dump(KERN_INFO, "",
477 						       DUMP_PREFIX_ADDRESS, 16,
478 						       1,
479 						       buffer_info->skb->data,
480 						       adapter->rx_buffer_len,
481 						       true);
482 			}
483 		}
484 	}
485 }
486 
487 /**
488  * e1000_desc_unused - calculate if we have unused descriptors
489  **/
490 static int e1000_desc_unused(struct e1000_ring *ring)
491 {
492 	if (ring->next_to_clean > ring->next_to_use)
493 		return ring->next_to_clean - ring->next_to_use - 1;
494 
495 	return ring->count + ring->next_to_clean - ring->next_to_use - 1;
496 }
497 
498 /**
499  * e1000e_systim_to_hwtstamp - convert system time value to hw time stamp
500  * @adapter: board private structure
501  * @hwtstamps: time stamp structure to update
502  * @systim: unsigned 64bit system time value.
503  *
504  * Convert the system time value stored in the RX/TXSTMP registers into a
505  * hwtstamp which can be used by the upper level time stamping functions.
506  *
507  * The 'systim_lock' spinlock is used to protect the consistency of the
508  * system time value. This is needed because reading the 64 bit time
509  * value involves reading two 32 bit registers. The first read latches the
510  * value.
511  **/
512 static void e1000e_systim_to_hwtstamp(struct e1000_adapter *adapter,
513 				      struct skb_shared_hwtstamps *hwtstamps,
514 				      u64 systim)
515 {
516 	u64 ns;
517 	unsigned long flags;
518 
519 	spin_lock_irqsave(&adapter->systim_lock, flags);
520 	ns = timecounter_cyc2time(&adapter->tc, systim);
521 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
522 
523 	memset(hwtstamps, 0, sizeof(*hwtstamps));
524 	hwtstamps->hwtstamp = ns_to_ktime(ns);
525 }
526 
527 /**
528  * e1000e_rx_hwtstamp - utility function which checks for Rx time stamp
529  * @adapter: board private structure
530  * @status: descriptor extended error and status field
531  * @skb: particular skb to include time stamp
532  *
533  * If the time stamp is valid, convert it into the timecounter ns value
534  * and store that result into the shhwtstamps structure which is passed
535  * up the network stack.
536  **/
537 static void e1000e_rx_hwtstamp(struct e1000_adapter *adapter, u32 status,
538 			       struct sk_buff *skb)
539 {
540 	struct e1000_hw *hw = &adapter->hw;
541 	u64 rxstmp;
542 
543 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP) ||
544 	    !(status & E1000_RXDEXT_STATERR_TST) ||
545 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID))
546 		return;
547 
548 	/* The Rx time stamp registers contain the time stamp.  No other
549 	 * received packet will be time stamped until the Rx time stamp
550 	 * registers are read.  Because only one packet can be time stamped
551 	 * at a time, the register values must belong to this packet and
552 	 * therefore none of the other additional attributes need to be
553 	 * compared.
554 	 */
555 	rxstmp = (u64)er32(RXSTMPL);
556 	rxstmp |= (u64)er32(RXSTMPH) << 32;
557 	e1000e_systim_to_hwtstamp(adapter, skb_hwtstamps(skb), rxstmp);
558 
559 	adapter->flags2 &= ~FLAG2_CHECK_RX_HWTSTAMP;
560 }
561 
562 /**
563  * e1000_receive_skb - helper function to handle Rx indications
564  * @adapter: board private structure
565  * @staterr: descriptor extended error and status field as written by hardware
566  * @vlan: descriptor vlan field as written by hardware (no le/be conversion)
567  * @skb: pointer to sk_buff to be indicated to stack
568  **/
569 static void e1000_receive_skb(struct e1000_adapter *adapter,
570 			      struct net_device *netdev, struct sk_buff *skb,
571 			      u32 staterr, __le16 vlan)
572 {
573 	u16 tag = le16_to_cpu(vlan);
574 
575 	e1000e_rx_hwtstamp(adapter, staterr, skb);
576 
577 	skb->protocol = eth_type_trans(skb, netdev);
578 
579 	if (staterr & E1000_RXD_STAT_VP)
580 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), tag);
581 
582 	napi_gro_receive(&adapter->napi, skb);
583 }
584 
585 /**
586  * e1000_rx_checksum - Receive Checksum Offload
587  * @adapter: board private structure
588  * @status_err: receive descriptor status and error fields
589  * @csum: receive descriptor csum field
590  * @sk_buff: socket buffer with received data
591  **/
592 static void e1000_rx_checksum(struct e1000_adapter *adapter, u32 status_err,
593 			      struct sk_buff *skb)
594 {
595 	u16 status = (u16)status_err;
596 	u8 errors = (u8)(status_err >> 24);
597 
598 	skb_checksum_none_assert(skb);
599 
600 	/* Rx checksum disabled */
601 	if (!(adapter->netdev->features & NETIF_F_RXCSUM))
602 		return;
603 
604 	/* Ignore Checksum bit is set */
605 	if (status & E1000_RXD_STAT_IXSM)
606 		return;
607 
608 	/* TCP/UDP checksum error bit or IP checksum error bit is set */
609 	if (errors & (E1000_RXD_ERR_TCPE | E1000_RXD_ERR_IPE)) {
610 		/* let the stack verify checksum errors */
611 		adapter->hw_csum_err++;
612 		return;
613 	}
614 
615 	/* TCP/UDP Checksum has not been calculated */
616 	if (!(status & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)))
617 		return;
618 
619 	/* It must be a TCP or UDP packet with a valid checksum */
620 	skb->ip_summed = CHECKSUM_UNNECESSARY;
621 	adapter->hw_csum_good++;
622 }
623 
624 static void e1000e_update_rdt_wa(struct e1000_ring *rx_ring, unsigned int i)
625 {
626 	struct e1000_adapter *adapter = rx_ring->adapter;
627 	struct e1000_hw *hw = &adapter->hw;
628 	s32 ret_val = __ew32_prepare(hw);
629 
630 	writel(i, rx_ring->tail);
631 
632 	if (unlikely(!ret_val && (i != readl(rx_ring->tail)))) {
633 		u32 rctl = er32(RCTL);
634 
635 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
636 		e_err("ME firmware caused invalid RDT - resetting\n");
637 		schedule_work(&adapter->reset_task);
638 	}
639 }
640 
641 static void e1000e_update_tdt_wa(struct e1000_ring *tx_ring, unsigned int i)
642 {
643 	struct e1000_adapter *adapter = tx_ring->adapter;
644 	struct e1000_hw *hw = &adapter->hw;
645 	s32 ret_val = __ew32_prepare(hw);
646 
647 	writel(i, tx_ring->tail);
648 
649 	if (unlikely(!ret_val && (i != readl(tx_ring->tail)))) {
650 		u32 tctl = er32(TCTL);
651 
652 		ew32(TCTL, tctl & ~E1000_TCTL_EN);
653 		e_err("ME firmware caused invalid TDT - resetting\n");
654 		schedule_work(&adapter->reset_task);
655 	}
656 }
657 
658 /**
659  * e1000_alloc_rx_buffers - Replace used receive buffers
660  * @rx_ring: Rx descriptor ring
661  **/
662 static void e1000_alloc_rx_buffers(struct e1000_ring *rx_ring,
663 				   int cleaned_count, gfp_t gfp)
664 {
665 	struct e1000_adapter *adapter = rx_ring->adapter;
666 	struct net_device *netdev = adapter->netdev;
667 	struct pci_dev *pdev = adapter->pdev;
668 	union e1000_rx_desc_extended *rx_desc;
669 	struct e1000_buffer *buffer_info;
670 	struct sk_buff *skb;
671 	unsigned int i;
672 	unsigned int bufsz = adapter->rx_buffer_len;
673 
674 	i = rx_ring->next_to_use;
675 	buffer_info = &rx_ring->buffer_info[i];
676 
677 	while (cleaned_count--) {
678 		skb = buffer_info->skb;
679 		if (skb) {
680 			skb_trim(skb, 0);
681 			goto map_skb;
682 		}
683 
684 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
685 		if (!skb) {
686 			/* Better luck next round */
687 			adapter->alloc_rx_buff_failed++;
688 			break;
689 		}
690 
691 		buffer_info->skb = skb;
692 map_skb:
693 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
694 						  adapter->rx_buffer_len,
695 						  DMA_FROM_DEVICE);
696 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
697 			dev_err(&pdev->dev, "Rx DMA map failed\n");
698 			adapter->rx_dma_failed++;
699 			break;
700 		}
701 
702 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
703 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
704 
705 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
706 			/* Force memory writes to complete before letting h/w
707 			 * know there are new descriptors to fetch.  (Only
708 			 * applicable for weak-ordered memory model archs,
709 			 * such as IA-64).
710 			 */
711 			wmb();
712 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
713 				e1000e_update_rdt_wa(rx_ring, i);
714 			else
715 				writel(i, rx_ring->tail);
716 		}
717 		i++;
718 		if (i == rx_ring->count)
719 			i = 0;
720 		buffer_info = &rx_ring->buffer_info[i];
721 	}
722 
723 	rx_ring->next_to_use = i;
724 }
725 
726 /**
727  * e1000_alloc_rx_buffers_ps - Replace used receive buffers; packet split
728  * @rx_ring: Rx descriptor ring
729  **/
730 static void e1000_alloc_rx_buffers_ps(struct e1000_ring *rx_ring,
731 				      int cleaned_count, gfp_t gfp)
732 {
733 	struct e1000_adapter *adapter = rx_ring->adapter;
734 	struct net_device *netdev = adapter->netdev;
735 	struct pci_dev *pdev = adapter->pdev;
736 	union e1000_rx_desc_packet_split *rx_desc;
737 	struct e1000_buffer *buffer_info;
738 	struct e1000_ps_page *ps_page;
739 	struct sk_buff *skb;
740 	unsigned int i, j;
741 
742 	i = rx_ring->next_to_use;
743 	buffer_info = &rx_ring->buffer_info[i];
744 
745 	while (cleaned_count--) {
746 		rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
747 
748 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
749 			ps_page = &buffer_info->ps_pages[j];
750 			if (j >= adapter->rx_ps_pages) {
751 				/* all unused desc entries get hw null ptr */
752 				rx_desc->read.buffer_addr[j + 1] =
753 				    ~cpu_to_le64(0);
754 				continue;
755 			}
756 			if (!ps_page->page) {
757 				ps_page->page = alloc_page(gfp);
758 				if (!ps_page->page) {
759 					adapter->alloc_rx_buff_failed++;
760 					goto no_buffers;
761 				}
762 				ps_page->dma = dma_map_page(&pdev->dev,
763 							    ps_page->page,
764 							    0, PAGE_SIZE,
765 							    DMA_FROM_DEVICE);
766 				if (dma_mapping_error(&pdev->dev,
767 						      ps_page->dma)) {
768 					dev_err(&adapter->pdev->dev,
769 						"Rx DMA page map failed\n");
770 					adapter->rx_dma_failed++;
771 					goto no_buffers;
772 				}
773 			}
774 			/* Refresh the desc even if buffer_addrs
775 			 * didn't change because each write-back
776 			 * erases this info.
777 			 */
778 			rx_desc->read.buffer_addr[j + 1] =
779 			    cpu_to_le64(ps_page->dma);
780 		}
781 
782 		skb = __netdev_alloc_skb_ip_align(netdev, adapter->rx_ps_bsize0,
783 						  gfp);
784 
785 		if (!skb) {
786 			adapter->alloc_rx_buff_failed++;
787 			break;
788 		}
789 
790 		buffer_info->skb = skb;
791 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
792 						  adapter->rx_ps_bsize0,
793 						  DMA_FROM_DEVICE);
794 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
795 			dev_err(&pdev->dev, "Rx DMA map failed\n");
796 			adapter->rx_dma_failed++;
797 			/* cleanup skb */
798 			dev_kfree_skb_any(skb);
799 			buffer_info->skb = NULL;
800 			break;
801 		}
802 
803 		rx_desc->read.buffer_addr[0] = cpu_to_le64(buffer_info->dma);
804 
805 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
806 			/* Force memory writes to complete before letting h/w
807 			 * know there are new descriptors to fetch.  (Only
808 			 * applicable for weak-ordered memory model archs,
809 			 * such as IA-64).
810 			 */
811 			wmb();
812 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
813 				e1000e_update_rdt_wa(rx_ring, i << 1);
814 			else
815 				writel(i << 1, rx_ring->tail);
816 		}
817 
818 		i++;
819 		if (i == rx_ring->count)
820 			i = 0;
821 		buffer_info = &rx_ring->buffer_info[i];
822 	}
823 
824 no_buffers:
825 	rx_ring->next_to_use = i;
826 }
827 
828 /**
829  * e1000_alloc_jumbo_rx_buffers - Replace used jumbo receive buffers
830  * @rx_ring: Rx descriptor ring
831  * @cleaned_count: number of buffers to allocate this pass
832  **/
833 
834 static void e1000_alloc_jumbo_rx_buffers(struct e1000_ring *rx_ring,
835 					 int cleaned_count, gfp_t gfp)
836 {
837 	struct e1000_adapter *adapter = rx_ring->adapter;
838 	struct net_device *netdev = adapter->netdev;
839 	struct pci_dev *pdev = adapter->pdev;
840 	union e1000_rx_desc_extended *rx_desc;
841 	struct e1000_buffer *buffer_info;
842 	struct sk_buff *skb;
843 	unsigned int i;
844 	unsigned int bufsz = 256 - 16;	/* for skb_reserve */
845 
846 	i = rx_ring->next_to_use;
847 	buffer_info = &rx_ring->buffer_info[i];
848 
849 	while (cleaned_count--) {
850 		skb = buffer_info->skb;
851 		if (skb) {
852 			skb_trim(skb, 0);
853 			goto check_page;
854 		}
855 
856 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
857 		if (unlikely(!skb)) {
858 			/* Better luck next round */
859 			adapter->alloc_rx_buff_failed++;
860 			break;
861 		}
862 
863 		buffer_info->skb = skb;
864 check_page:
865 		/* allocate a new page if necessary */
866 		if (!buffer_info->page) {
867 			buffer_info->page = alloc_page(gfp);
868 			if (unlikely(!buffer_info->page)) {
869 				adapter->alloc_rx_buff_failed++;
870 				break;
871 			}
872 		}
873 
874 		if (!buffer_info->dma) {
875 			buffer_info->dma = dma_map_page(&pdev->dev,
876 							buffer_info->page, 0,
877 							PAGE_SIZE,
878 							DMA_FROM_DEVICE);
879 			if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
880 				adapter->alloc_rx_buff_failed++;
881 				break;
882 			}
883 		}
884 
885 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
886 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
887 
888 		if (unlikely(++i == rx_ring->count))
889 			i = 0;
890 		buffer_info = &rx_ring->buffer_info[i];
891 	}
892 
893 	if (likely(rx_ring->next_to_use != i)) {
894 		rx_ring->next_to_use = i;
895 		if (unlikely(i-- == 0))
896 			i = (rx_ring->count - 1);
897 
898 		/* Force memory writes to complete before letting h/w
899 		 * know there are new descriptors to fetch.  (Only
900 		 * applicable for weak-ordered memory model archs,
901 		 * such as IA-64).
902 		 */
903 		wmb();
904 		if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
905 			e1000e_update_rdt_wa(rx_ring, i);
906 		else
907 			writel(i, rx_ring->tail);
908 	}
909 }
910 
911 static inline void e1000_rx_hash(struct net_device *netdev, __le32 rss,
912 				 struct sk_buff *skb)
913 {
914 	if (netdev->features & NETIF_F_RXHASH)
915 		skb_set_hash(skb, le32_to_cpu(rss), PKT_HASH_TYPE_L3);
916 }
917 
918 /**
919  * e1000_clean_rx_irq - Send received data up the network stack
920  * @rx_ring: Rx descriptor ring
921  *
922  * the return value indicates whether actual cleaning was done, there
923  * is no guarantee that everything was cleaned
924  **/
925 static bool e1000_clean_rx_irq(struct e1000_ring *rx_ring, int *work_done,
926 			       int work_to_do)
927 {
928 	struct e1000_adapter *adapter = rx_ring->adapter;
929 	struct net_device *netdev = adapter->netdev;
930 	struct pci_dev *pdev = adapter->pdev;
931 	struct e1000_hw *hw = &adapter->hw;
932 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
933 	struct e1000_buffer *buffer_info, *next_buffer;
934 	u32 length, staterr;
935 	unsigned int i;
936 	int cleaned_count = 0;
937 	bool cleaned = false;
938 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
939 
940 	i = rx_ring->next_to_clean;
941 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
942 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
943 	buffer_info = &rx_ring->buffer_info[i];
944 
945 	while (staterr & E1000_RXD_STAT_DD) {
946 		struct sk_buff *skb;
947 
948 		if (*work_done >= work_to_do)
949 			break;
950 		(*work_done)++;
951 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
952 
953 		skb = buffer_info->skb;
954 		buffer_info->skb = NULL;
955 
956 		prefetch(skb->data - NET_IP_ALIGN);
957 
958 		i++;
959 		if (i == rx_ring->count)
960 			i = 0;
961 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
962 		prefetch(next_rxd);
963 
964 		next_buffer = &rx_ring->buffer_info[i];
965 
966 		cleaned = true;
967 		cleaned_count++;
968 		dma_unmap_single(&pdev->dev, buffer_info->dma,
969 				 adapter->rx_buffer_len, DMA_FROM_DEVICE);
970 		buffer_info->dma = 0;
971 
972 		length = le16_to_cpu(rx_desc->wb.upper.length);
973 
974 		/* !EOP means multiple descriptors were used to store a single
975 		 * packet, if that's the case we need to toss it.  In fact, we
976 		 * need to toss every packet with the EOP bit clear and the
977 		 * next frame that _does_ have the EOP bit set, as it is by
978 		 * definition only a frame fragment
979 		 */
980 		if (unlikely(!(staterr & E1000_RXD_STAT_EOP)))
981 			adapter->flags2 |= FLAG2_IS_DISCARDING;
982 
983 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
984 			/* All receives must fit into a single buffer */
985 			e_dbg("Receive packet consumed multiple buffers\n");
986 			/* recycle */
987 			buffer_info->skb = skb;
988 			if (staterr & E1000_RXD_STAT_EOP)
989 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
990 			goto next_desc;
991 		}
992 
993 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
994 			     !(netdev->features & NETIF_F_RXALL))) {
995 			/* recycle */
996 			buffer_info->skb = skb;
997 			goto next_desc;
998 		}
999 
1000 		/* adjust length to remove Ethernet CRC */
1001 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1002 			/* If configured to store CRC, don't subtract FCS,
1003 			 * but keep the FCS bytes out of the total_rx_bytes
1004 			 * counter
1005 			 */
1006 			if (netdev->features & NETIF_F_RXFCS)
1007 				total_rx_bytes -= 4;
1008 			else
1009 				length -= 4;
1010 		}
1011 
1012 		total_rx_bytes += length;
1013 		total_rx_packets++;
1014 
1015 		/* code added for copybreak, this should improve
1016 		 * performance for small packets with large amounts
1017 		 * of reassembly being done in the stack
1018 		 */
1019 		if (length < copybreak) {
1020 			struct sk_buff *new_skb =
1021 				napi_alloc_skb(&adapter->napi, length);
1022 			if (new_skb) {
1023 				skb_copy_to_linear_data_offset(new_skb,
1024 							       -NET_IP_ALIGN,
1025 							       (skb->data -
1026 								NET_IP_ALIGN),
1027 							       (length +
1028 								NET_IP_ALIGN));
1029 				/* save the skb in buffer_info as good */
1030 				buffer_info->skb = skb;
1031 				skb = new_skb;
1032 			}
1033 			/* else just continue with the old one */
1034 		}
1035 		/* end copybreak code */
1036 		skb_put(skb, length);
1037 
1038 		/* Receive Checksum Offload */
1039 		e1000_rx_checksum(adapter, staterr, skb);
1040 
1041 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1042 
1043 		e1000_receive_skb(adapter, netdev, skb, staterr,
1044 				  rx_desc->wb.upper.vlan);
1045 
1046 next_desc:
1047 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1048 
1049 		/* return some buffers to hardware, one at a time is too slow */
1050 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1051 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1052 					      GFP_ATOMIC);
1053 			cleaned_count = 0;
1054 		}
1055 
1056 		/* use prefetched values */
1057 		rx_desc = next_rxd;
1058 		buffer_info = next_buffer;
1059 
1060 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1061 	}
1062 	rx_ring->next_to_clean = i;
1063 
1064 	cleaned_count = e1000_desc_unused(rx_ring);
1065 	if (cleaned_count)
1066 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1067 
1068 	adapter->total_rx_bytes += total_rx_bytes;
1069 	adapter->total_rx_packets += total_rx_packets;
1070 	return cleaned;
1071 }
1072 
1073 static void e1000_put_txbuf(struct e1000_ring *tx_ring,
1074 			    struct e1000_buffer *buffer_info,
1075 			    bool drop)
1076 {
1077 	struct e1000_adapter *adapter = tx_ring->adapter;
1078 
1079 	if (buffer_info->dma) {
1080 		if (buffer_info->mapped_as_page)
1081 			dma_unmap_page(&adapter->pdev->dev, buffer_info->dma,
1082 				       buffer_info->length, DMA_TO_DEVICE);
1083 		else
1084 			dma_unmap_single(&adapter->pdev->dev, buffer_info->dma,
1085 					 buffer_info->length, DMA_TO_DEVICE);
1086 		buffer_info->dma = 0;
1087 	}
1088 	if (buffer_info->skb) {
1089 		if (drop)
1090 			dev_kfree_skb_any(buffer_info->skb);
1091 		else
1092 			dev_consume_skb_any(buffer_info->skb);
1093 		buffer_info->skb = NULL;
1094 	}
1095 	buffer_info->time_stamp = 0;
1096 }
1097 
1098 static void e1000_print_hw_hang(struct work_struct *work)
1099 {
1100 	struct e1000_adapter *adapter = container_of(work,
1101 						     struct e1000_adapter,
1102 						     print_hang_task);
1103 	struct net_device *netdev = adapter->netdev;
1104 	struct e1000_ring *tx_ring = adapter->tx_ring;
1105 	unsigned int i = tx_ring->next_to_clean;
1106 	unsigned int eop = tx_ring->buffer_info[i].next_to_watch;
1107 	struct e1000_tx_desc *eop_desc = E1000_TX_DESC(*tx_ring, eop);
1108 	struct e1000_hw *hw = &adapter->hw;
1109 	u16 phy_status, phy_1000t_status, phy_ext_status;
1110 	u16 pci_status;
1111 
1112 	if (test_bit(__E1000_DOWN, &adapter->state))
1113 		return;
1114 
1115 	if (!adapter->tx_hang_recheck && (adapter->flags2 & FLAG2_DMA_BURST)) {
1116 		/* May be block on write-back, flush and detect again
1117 		 * flush pending descriptor writebacks to memory
1118 		 */
1119 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1120 		/* execute the writes immediately */
1121 		e1e_flush();
1122 		/* Due to rare timing issues, write to TIDV again to ensure
1123 		 * the write is successful
1124 		 */
1125 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1126 		/* execute the writes immediately */
1127 		e1e_flush();
1128 		adapter->tx_hang_recheck = true;
1129 		return;
1130 	}
1131 	adapter->tx_hang_recheck = false;
1132 
1133 	if (er32(TDH(0)) == er32(TDT(0))) {
1134 		e_dbg("false hang detected, ignoring\n");
1135 		return;
1136 	}
1137 
1138 	/* Real hang detected */
1139 	netif_stop_queue(netdev);
1140 
1141 	e1e_rphy(hw, MII_BMSR, &phy_status);
1142 	e1e_rphy(hw, MII_STAT1000, &phy_1000t_status);
1143 	e1e_rphy(hw, MII_ESTATUS, &phy_ext_status);
1144 
1145 	pci_read_config_word(adapter->pdev, PCI_STATUS, &pci_status);
1146 
1147 	/* detected Hardware unit hang */
1148 	e_err("Detected Hardware Unit Hang:\n"
1149 	      "  TDH                  <%x>\n"
1150 	      "  TDT                  <%x>\n"
1151 	      "  next_to_use          <%x>\n"
1152 	      "  next_to_clean        <%x>\n"
1153 	      "buffer_info[next_to_clean]:\n"
1154 	      "  time_stamp           <%lx>\n"
1155 	      "  next_to_watch        <%x>\n"
1156 	      "  jiffies              <%lx>\n"
1157 	      "  next_to_watch.status <%x>\n"
1158 	      "MAC Status             <%x>\n"
1159 	      "PHY Status             <%x>\n"
1160 	      "PHY 1000BASE-T Status  <%x>\n"
1161 	      "PHY Extended Status    <%x>\n"
1162 	      "PCI Status             <%x>\n",
1163 	      readl(tx_ring->head), readl(tx_ring->tail), tx_ring->next_to_use,
1164 	      tx_ring->next_to_clean, tx_ring->buffer_info[eop].time_stamp,
1165 	      eop, jiffies, eop_desc->upper.fields.status, er32(STATUS),
1166 	      phy_status, phy_1000t_status, phy_ext_status, pci_status);
1167 
1168 	e1000e_dump(adapter);
1169 
1170 	/* Suggest workaround for known h/w issue */
1171 	if ((hw->mac.type == e1000_pchlan) && (er32(CTRL) & E1000_CTRL_TFCE))
1172 		e_err("Try turning off Tx pause (flow control) via ethtool\n");
1173 }
1174 
1175 /**
1176  * e1000e_tx_hwtstamp_work - check for Tx time stamp
1177  * @work: pointer to work struct
1178  *
1179  * This work function polls the TSYNCTXCTL valid bit to determine when a
1180  * timestamp has been taken for the current stored skb.  The timestamp must
1181  * be for this skb because only one such packet is allowed in the queue.
1182  */
1183 static void e1000e_tx_hwtstamp_work(struct work_struct *work)
1184 {
1185 	struct e1000_adapter *adapter = container_of(work, struct e1000_adapter,
1186 						     tx_hwtstamp_work);
1187 	struct e1000_hw *hw = &adapter->hw;
1188 
1189 	if (er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID) {
1190 		struct sk_buff *skb = adapter->tx_hwtstamp_skb;
1191 		struct skb_shared_hwtstamps shhwtstamps;
1192 		u64 txstmp;
1193 
1194 		txstmp = er32(TXSTMPL);
1195 		txstmp |= (u64)er32(TXSTMPH) << 32;
1196 
1197 		e1000e_systim_to_hwtstamp(adapter, &shhwtstamps, txstmp);
1198 
1199 		/* Clear the global tx_hwtstamp_skb pointer and force writes
1200 		 * prior to notifying the stack of a Tx timestamp.
1201 		 */
1202 		adapter->tx_hwtstamp_skb = NULL;
1203 		wmb(); /* force write prior to skb_tstamp_tx */
1204 
1205 		skb_tstamp_tx(skb, &shhwtstamps);
1206 		dev_consume_skb_any(skb);
1207 	} else if (time_after(jiffies, adapter->tx_hwtstamp_start
1208 			      + adapter->tx_timeout_factor * HZ)) {
1209 		dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
1210 		adapter->tx_hwtstamp_skb = NULL;
1211 		adapter->tx_hwtstamp_timeouts++;
1212 		e_warn("clearing Tx timestamp hang\n");
1213 	} else {
1214 		/* reschedule to check later */
1215 		schedule_work(&adapter->tx_hwtstamp_work);
1216 	}
1217 }
1218 
1219 /**
1220  * e1000_clean_tx_irq - Reclaim resources after transmit completes
1221  * @tx_ring: Tx descriptor ring
1222  *
1223  * the return value indicates whether actual cleaning was done, there
1224  * is no guarantee that everything was cleaned
1225  **/
1226 static bool e1000_clean_tx_irq(struct e1000_ring *tx_ring)
1227 {
1228 	struct e1000_adapter *adapter = tx_ring->adapter;
1229 	struct net_device *netdev = adapter->netdev;
1230 	struct e1000_hw *hw = &adapter->hw;
1231 	struct e1000_tx_desc *tx_desc, *eop_desc;
1232 	struct e1000_buffer *buffer_info;
1233 	unsigned int i, eop;
1234 	unsigned int count = 0;
1235 	unsigned int total_tx_bytes = 0, total_tx_packets = 0;
1236 	unsigned int bytes_compl = 0, pkts_compl = 0;
1237 
1238 	i = tx_ring->next_to_clean;
1239 	eop = tx_ring->buffer_info[i].next_to_watch;
1240 	eop_desc = E1000_TX_DESC(*tx_ring, eop);
1241 
1242 	while ((eop_desc->upper.data & cpu_to_le32(E1000_TXD_STAT_DD)) &&
1243 	       (count < tx_ring->count)) {
1244 		bool cleaned = false;
1245 
1246 		dma_rmb();		/* read buffer_info after eop_desc */
1247 		for (; !cleaned; count++) {
1248 			tx_desc = E1000_TX_DESC(*tx_ring, i);
1249 			buffer_info = &tx_ring->buffer_info[i];
1250 			cleaned = (i == eop);
1251 
1252 			if (cleaned) {
1253 				total_tx_packets += buffer_info->segs;
1254 				total_tx_bytes += buffer_info->bytecount;
1255 				if (buffer_info->skb) {
1256 					bytes_compl += buffer_info->skb->len;
1257 					pkts_compl++;
1258 				}
1259 			}
1260 
1261 			e1000_put_txbuf(tx_ring, buffer_info, false);
1262 			tx_desc->upper.data = 0;
1263 
1264 			i++;
1265 			if (i == tx_ring->count)
1266 				i = 0;
1267 		}
1268 
1269 		if (i == tx_ring->next_to_use)
1270 			break;
1271 		eop = tx_ring->buffer_info[i].next_to_watch;
1272 		eop_desc = E1000_TX_DESC(*tx_ring, eop);
1273 	}
1274 
1275 	tx_ring->next_to_clean = i;
1276 
1277 	netdev_completed_queue(netdev, pkts_compl, bytes_compl);
1278 
1279 #define TX_WAKE_THRESHOLD 32
1280 	if (count && netif_carrier_ok(netdev) &&
1281 	    e1000_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD) {
1282 		/* Make sure that anybody stopping the queue after this
1283 		 * sees the new next_to_clean.
1284 		 */
1285 		smp_mb();
1286 
1287 		if (netif_queue_stopped(netdev) &&
1288 		    !(test_bit(__E1000_DOWN, &adapter->state))) {
1289 			netif_wake_queue(netdev);
1290 			++adapter->restart_queue;
1291 		}
1292 	}
1293 
1294 	if (adapter->detect_tx_hung) {
1295 		/* Detect a transmit hang in hardware, this serializes the
1296 		 * check with the clearing of time_stamp and movement of i
1297 		 */
1298 		adapter->detect_tx_hung = false;
1299 		if (tx_ring->buffer_info[i].time_stamp &&
1300 		    time_after(jiffies, tx_ring->buffer_info[i].time_stamp
1301 			       + (adapter->tx_timeout_factor * HZ)) &&
1302 		    !(er32(STATUS) & E1000_STATUS_TXOFF))
1303 			schedule_work(&adapter->print_hang_task);
1304 		else
1305 			adapter->tx_hang_recheck = false;
1306 	}
1307 	adapter->total_tx_bytes += total_tx_bytes;
1308 	adapter->total_tx_packets += total_tx_packets;
1309 	return count < tx_ring->count;
1310 }
1311 
1312 /**
1313  * e1000_clean_rx_irq_ps - Send received data up the network stack; packet split
1314  * @rx_ring: Rx descriptor ring
1315  *
1316  * the return value indicates whether actual cleaning was done, there
1317  * is no guarantee that everything was cleaned
1318  **/
1319 static bool e1000_clean_rx_irq_ps(struct e1000_ring *rx_ring, int *work_done,
1320 				  int work_to_do)
1321 {
1322 	struct e1000_adapter *adapter = rx_ring->adapter;
1323 	struct e1000_hw *hw = &adapter->hw;
1324 	union e1000_rx_desc_packet_split *rx_desc, *next_rxd;
1325 	struct net_device *netdev = adapter->netdev;
1326 	struct pci_dev *pdev = adapter->pdev;
1327 	struct e1000_buffer *buffer_info, *next_buffer;
1328 	struct e1000_ps_page *ps_page;
1329 	struct sk_buff *skb;
1330 	unsigned int i, j;
1331 	u32 length, staterr;
1332 	int cleaned_count = 0;
1333 	bool cleaned = false;
1334 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1335 
1336 	i = rx_ring->next_to_clean;
1337 	rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
1338 	staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1339 	buffer_info = &rx_ring->buffer_info[i];
1340 
1341 	while (staterr & E1000_RXD_STAT_DD) {
1342 		if (*work_done >= work_to_do)
1343 			break;
1344 		(*work_done)++;
1345 		skb = buffer_info->skb;
1346 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1347 
1348 		/* in the packet split case this is header only */
1349 		prefetch(skb->data - NET_IP_ALIGN);
1350 
1351 		i++;
1352 		if (i == rx_ring->count)
1353 			i = 0;
1354 		next_rxd = E1000_RX_DESC_PS(*rx_ring, i);
1355 		prefetch(next_rxd);
1356 
1357 		next_buffer = &rx_ring->buffer_info[i];
1358 
1359 		cleaned = true;
1360 		cleaned_count++;
1361 		dma_unmap_single(&pdev->dev, buffer_info->dma,
1362 				 adapter->rx_ps_bsize0, DMA_FROM_DEVICE);
1363 		buffer_info->dma = 0;
1364 
1365 		/* see !EOP comment in other Rx routine */
1366 		if (!(staterr & E1000_RXD_STAT_EOP))
1367 			adapter->flags2 |= FLAG2_IS_DISCARDING;
1368 
1369 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
1370 			e_dbg("Packet Split buffers didn't pick up the full packet\n");
1371 			dev_kfree_skb_irq(skb);
1372 			if (staterr & E1000_RXD_STAT_EOP)
1373 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1374 			goto next_desc;
1375 		}
1376 
1377 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1378 			     !(netdev->features & NETIF_F_RXALL))) {
1379 			dev_kfree_skb_irq(skb);
1380 			goto next_desc;
1381 		}
1382 
1383 		length = le16_to_cpu(rx_desc->wb.middle.length0);
1384 
1385 		if (!length) {
1386 			e_dbg("Last part of the packet spanning multiple descriptors\n");
1387 			dev_kfree_skb_irq(skb);
1388 			goto next_desc;
1389 		}
1390 
1391 		/* Good Receive */
1392 		skb_put(skb, length);
1393 
1394 		{
1395 			/* this looks ugly, but it seems compiler issues make
1396 			 * it more efficient than reusing j
1397 			 */
1398 			int l1 = le16_to_cpu(rx_desc->wb.upper.length[0]);
1399 
1400 			/* page alloc/put takes too long and effects small
1401 			 * packet throughput, so unsplit small packets and
1402 			 * save the alloc/put only valid in softirq (napi)
1403 			 * context to call kmap_*
1404 			 */
1405 			if (l1 && (l1 <= copybreak) &&
1406 			    ((length + l1) <= adapter->rx_ps_bsize0)) {
1407 				u8 *vaddr;
1408 
1409 				ps_page = &buffer_info->ps_pages[0];
1410 
1411 				/* there is no documentation about how to call
1412 				 * kmap_atomic, so we can't hold the mapping
1413 				 * very long
1414 				 */
1415 				dma_sync_single_for_cpu(&pdev->dev,
1416 							ps_page->dma,
1417 							PAGE_SIZE,
1418 							DMA_FROM_DEVICE);
1419 				vaddr = kmap_atomic(ps_page->page);
1420 				memcpy(skb_tail_pointer(skb), vaddr, l1);
1421 				kunmap_atomic(vaddr);
1422 				dma_sync_single_for_device(&pdev->dev,
1423 							   ps_page->dma,
1424 							   PAGE_SIZE,
1425 							   DMA_FROM_DEVICE);
1426 
1427 				/* remove the CRC */
1428 				if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1429 					if (!(netdev->features & NETIF_F_RXFCS))
1430 						l1 -= 4;
1431 				}
1432 
1433 				skb_put(skb, l1);
1434 				goto copydone;
1435 			}	/* if */
1436 		}
1437 
1438 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1439 			length = le16_to_cpu(rx_desc->wb.upper.length[j]);
1440 			if (!length)
1441 				break;
1442 
1443 			ps_page = &buffer_info->ps_pages[j];
1444 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1445 				       DMA_FROM_DEVICE);
1446 			ps_page->dma = 0;
1447 			skb_fill_page_desc(skb, j, ps_page->page, 0, length);
1448 			ps_page->page = NULL;
1449 			skb->len += length;
1450 			skb->data_len += length;
1451 			skb->truesize += PAGE_SIZE;
1452 		}
1453 
1454 		/* strip the ethernet crc, problem is we're using pages now so
1455 		 * this whole operation can get a little cpu intensive
1456 		 */
1457 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1458 			if (!(netdev->features & NETIF_F_RXFCS))
1459 				pskb_trim(skb, skb->len - 4);
1460 		}
1461 
1462 copydone:
1463 		total_rx_bytes += skb->len;
1464 		total_rx_packets++;
1465 
1466 		e1000_rx_checksum(adapter, staterr, skb);
1467 
1468 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1469 
1470 		if (rx_desc->wb.upper.header_status &
1471 		    cpu_to_le16(E1000_RXDPS_HDRSTAT_HDRSP))
1472 			adapter->rx_hdr_split++;
1473 
1474 		e1000_receive_skb(adapter, netdev, skb, staterr,
1475 				  rx_desc->wb.middle.vlan);
1476 
1477 next_desc:
1478 		rx_desc->wb.middle.status_error &= cpu_to_le32(~0xFF);
1479 		buffer_info->skb = NULL;
1480 
1481 		/* return some buffers to hardware, one at a time is too slow */
1482 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1483 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1484 					      GFP_ATOMIC);
1485 			cleaned_count = 0;
1486 		}
1487 
1488 		/* use prefetched values */
1489 		rx_desc = next_rxd;
1490 		buffer_info = next_buffer;
1491 
1492 		staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1493 	}
1494 	rx_ring->next_to_clean = i;
1495 
1496 	cleaned_count = e1000_desc_unused(rx_ring);
1497 	if (cleaned_count)
1498 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1499 
1500 	adapter->total_rx_bytes += total_rx_bytes;
1501 	adapter->total_rx_packets += total_rx_packets;
1502 	return cleaned;
1503 }
1504 
1505 /**
1506  * e1000_consume_page - helper function
1507  **/
1508 static void e1000_consume_page(struct e1000_buffer *bi, struct sk_buff *skb,
1509 			       u16 length)
1510 {
1511 	bi->page = NULL;
1512 	skb->len += length;
1513 	skb->data_len += length;
1514 	skb->truesize += PAGE_SIZE;
1515 }
1516 
1517 /**
1518  * e1000_clean_jumbo_rx_irq - Send received data up the network stack; legacy
1519  * @adapter: board private structure
1520  *
1521  * the return value indicates whether actual cleaning was done, there
1522  * is no guarantee that everything was cleaned
1523  **/
1524 static bool e1000_clean_jumbo_rx_irq(struct e1000_ring *rx_ring, int *work_done,
1525 				     int work_to_do)
1526 {
1527 	struct e1000_adapter *adapter = rx_ring->adapter;
1528 	struct net_device *netdev = adapter->netdev;
1529 	struct pci_dev *pdev = adapter->pdev;
1530 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
1531 	struct e1000_buffer *buffer_info, *next_buffer;
1532 	u32 length, staterr;
1533 	unsigned int i;
1534 	int cleaned_count = 0;
1535 	bool cleaned = false;
1536 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1537 	struct skb_shared_info *shinfo;
1538 
1539 	i = rx_ring->next_to_clean;
1540 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
1541 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1542 	buffer_info = &rx_ring->buffer_info[i];
1543 
1544 	while (staterr & E1000_RXD_STAT_DD) {
1545 		struct sk_buff *skb;
1546 
1547 		if (*work_done >= work_to_do)
1548 			break;
1549 		(*work_done)++;
1550 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1551 
1552 		skb = buffer_info->skb;
1553 		buffer_info->skb = NULL;
1554 
1555 		++i;
1556 		if (i == rx_ring->count)
1557 			i = 0;
1558 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
1559 		prefetch(next_rxd);
1560 
1561 		next_buffer = &rx_ring->buffer_info[i];
1562 
1563 		cleaned = true;
1564 		cleaned_count++;
1565 		dma_unmap_page(&pdev->dev, buffer_info->dma, PAGE_SIZE,
1566 			       DMA_FROM_DEVICE);
1567 		buffer_info->dma = 0;
1568 
1569 		length = le16_to_cpu(rx_desc->wb.upper.length);
1570 
1571 		/* errors is only valid for DD + EOP descriptors */
1572 		if (unlikely((staterr & E1000_RXD_STAT_EOP) &&
1573 			     ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1574 			      !(netdev->features & NETIF_F_RXALL)))) {
1575 			/* recycle both page and skb */
1576 			buffer_info->skb = skb;
1577 			/* an error means any chain goes out the window too */
1578 			if (rx_ring->rx_skb_top)
1579 				dev_kfree_skb_irq(rx_ring->rx_skb_top);
1580 			rx_ring->rx_skb_top = NULL;
1581 			goto next_desc;
1582 		}
1583 #define rxtop (rx_ring->rx_skb_top)
1584 		if (!(staterr & E1000_RXD_STAT_EOP)) {
1585 			/* this descriptor is only the beginning (or middle) */
1586 			if (!rxtop) {
1587 				/* this is the beginning of a chain */
1588 				rxtop = skb;
1589 				skb_fill_page_desc(rxtop, 0, buffer_info->page,
1590 						   0, length);
1591 			} else {
1592 				/* this is the middle of a chain */
1593 				shinfo = skb_shinfo(rxtop);
1594 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1595 						   buffer_info->page, 0,
1596 						   length);
1597 				/* re-use the skb, only consumed the page */
1598 				buffer_info->skb = skb;
1599 			}
1600 			e1000_consume_page(buffer_info, rxtop, length);
1601 			goto next_desc;
1602 		} else {
1603 			if (rxtop) {
1604 				/* end of the chain */
1605 				shinfo = skb_shinfo(rxtop);
1606 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1607 						   buffer_info->page, 0,
1608 						   length);
1609 				/* re-use the current skb, we only consumed the
1610 				 * page
1611 				 */
1612 				buffer_info->skb = skb;
1613 				skb = rxtop;
1614 				rxtop = NULL;
1615 				e1000_consume_page(buffer_info, skb, length);
1616 			} else {
1617 				/* no chain, got EOP, this buf is the packet
1618 				 * copybreak to save the put_page/alloc_page
1619 				 */
1620 				if (length <= copybreak &&
1621 				    skb_tailroom(skb) >= length) {
1622 					u8 *vaddr;
1623 					vaddr = kmap_atomic(buffer_info->page);
1624 					memcpy(skb_tail_pointer(skb), vaddr,
1625 					       length);
1626 					kunmap_atomic(vaddr);
1627 					/* re-use the page, so don't erase
1628 					 * buffer_info->page
1629 					 */
1630 					skb_put(skb, length);
1631 				} else {
1632 					skb_fill_page_desc(skb, 0,
1633 							   buffer_info->page, 0,
1634 							   length);
1635 					e1000_consume_page(buffer_info, skb,
1636 							   length);
1637 				}
1638 			}
1639 		}
1640 
1641 		/* Receive Checksum Offload */
1642 		e1000_rx_checksum(adapter, staterr, skb);
1643 
1644 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1645 
1646 		/* probably a little skewed due to removing CRC */
1647 		total_rx_bytes += skb->len;
1648 		total_rx_packets++;
1649 
1650 		/* eth type trans needs skb->data to point to something */
1651 		if (!pskb_may_pull(skb, ETH_HLEN)) {
1652 			e_err("pskb_may_pull failed.\n");
1653 			dev_kfree_skb_irq(skb);
1654 			goto next_desc;
1655 		}
1656 
1657 		e1000_receive_skb(adapter, netdev, skb, staterr,
1658 				  rx_desc->wb.upper.vlan);
1659 
1660 next_desc:
1661 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1662 
1663 		/* return some buffers to hardware, one at a time is too slow */
1664 		if (unlikely(cleaned_count >= E1000_RX_BUFFER_WRITE)) {
1665 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1666 					      GFP_ATOMIC);
1667 			cleaned_count = 0;
1668 		}
1669 
1670 		/* use prefetched values */
1671 		rx_desc = next_rxd;
1672 		buffer_info = next_buffer;
1673 
1674 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1675 	}
1676 	rx_ring->next_to_clean = i;
1677 
1678 	cleaned_count = e1000_desc_unused(rx_ring);
1679 	if (cleaned_count)
1680 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1681 
1682 	adapter->total_rx_bytes += total_rx_bytes;
1683 	adapter->total_rx_packets += total_rx_packets;
1684 	return cleaned;
1685 }
1686 
1687 /**
1688  * e1000_clean_rx_ring - Free Rx Buffers per Queue
1689  * @rx_ring: Rx descriptor ring
1690  **/
1691 static void e1000_clean_rx_ring(struct e1000_ring *rx_ring)
1692 {
1693 	struct e1000_adapter *adapter = rx_ring->adapter;
1694 	struct e1000_buffer *buffer_info;
1695 	struct e1000_ps_page *ps_page;
1696 	struct pci_dev *pdev = adapter->pdev;
1697 	unsigned int i, j;
1698 
1699 	/* Free all the Rx ring sk_buffs */
1700 	for (i = 0; i < rx_ring->count; i++) {
1701 		buffer_info = &rx_ring->buffer_info[i];
1702 		if (buffer_info->dma) {
1703 			if (adapter->clean_rx == e1000_clean_rx_irq)
1704 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1705 						 adapter->rx_buffer_len,
1706 						 DMA_FROM_DEVICE);
1707 			else if (adapter->clean_rx == e1000_clean_jumbo_rx_irq)
1708 				dma_unmap_page(&pdev->dev, buffer_info->dma,
1709 					       PAGE_SIZE, DMA_FROM_DEVICE);
1710 			else if (adapter->clean_rx == e1000_clean_rx_irq_ps)
1711 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1712 						 adapter->rx_ps_bsize0,
1713 						 DMA_FROM_DEVICE);
1714 			buffer_info->dma = 0;
1715 		}
1716 
1717 		if (buffer_info->page) {
1718 			put_page(buffer_info->page);
1719 			buffer_info->page = NULL;
1720 		}
1721 
1722 		if (buffer_info->skb) {
1723 			dev_kfree_skb(buffer_info->skb);
1724 			buffer_info->skb = NULL;
1725 		}
1726 
1727 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1728 			ps_page = &buffer_info->ps_pages[j];
1729 			if (!ps_page->page)
1730 				break;
1731 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1732 				       DMA_FROM_DEVICE);
1733 			ps_page->dma = 0;
1734 			put_page(ps_page->page);
1735 			ps_page->page = NULL;
1736 		}
1737 	}
1738 
1739 	/* there also may be some cached data from a chained receive */
1740 	if (rx_ring->rx_skb_top) {
1741 		dev_kfree_skb(rx_ring->rx_skb_top);
1742 		rx_ring->rx_skb_top = NULL;
1743 	}
1744 
1745 	/* Zero out the descriptor ring */
1746 	memset(rx_ring->desc, 0, rx_ring->size);
1747 
1748 	rx_ring->next_to_clean = 0;
1749 	rx_ring->next_to_use = 0;
1750 	adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1751 }
1752 
1753 static void e1000e_downshift_workaround(struct work_struct *work)
1754 {
1755 	struct e1000_adapter *adapter = container_of(work,
1756 						     struct e1000_adapter,
1757 						     downshift_task);
1758 
1759 	if (test_bit(__E1000_DOWN, &adapter->state))
1760 		return;
1761 
1762 	e1000e_gig_downshift_workaround_ich8lan(&adapter->hw);
1763 }
1764 
1765 /**
1766  * e1000_intr_msi - Interrupt Handler
1767  * @irq: interrupt number
1768  * @data: pointer to a network interface device structure
1769  **/
1770 static irqreturn_t e1000_intr_msi(int __always_unused irq, void *data)
1771 {
1772 	struct net_device *netdev = data;
1773 	struct e1000_adapter *adapter = netdev_priv(netdev);
1774 	struct e1000_hw *hw = &adapter->hw;
1775 	u32 icr = er32(ICR);
1776 
1777 	/* read ICR disables interrupts using IAM */
1778 	if (icr & E1000_ICR_LSC) {
1779 		hw->mac.get_link_status = true;
1780 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1781 		 * disconnect (LSC) before accessing any PHY registers
1782 		 */
1783 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1784 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1785 			schedule_work(&adapter->downshift_task);
1786 
1787 		/* 80003ES2LAN workaround-- For packet buffer work-around on
1788 		 * link down event; disable receives here in the ISR and reset
1789 		 * adapter in watchdog
1790 		 */
1791 		if (netif_carrier_ok(netdev) &&
1792 		    adapter->flags & FLAG_RX_NEEDS_RESTART) {
1793 			/* disable receives */
1794 			u32 rctl = er32(RCTL);
1795 
1796 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1797 			adapter->flags |= FLAG_RESTART_NOW;
1798 		}
1799 		/* guard against interrupt when we're going down */
1800 		if (!test_bit(__E1000_DOWN, &adapter->state))
1801 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1802 	}
1803 
1804 	/* Reset on uncorrectable ECC error */
1805 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1806 		u32 pbeccsts = er32(PBECCSTS);
1807 
1808 		adapter->corr_errors +=
1809 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1810 		adapter->uncorr_errors +=
1811 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1812 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1813 
1814 		/* Do the reset outside of interrupt context */
1815 		schedule_work(&adapter->reset_task);
1816 
1817 		/* return immediately since reset is imminent */
1818 		return IRQ_HANDLED;
1819 	}
1820 
1821 	if (napi_schedule_prep(&adapter->napi)) {
1822 		adapter->total_tx_bytes = 0;
1823 		adapter->total_tx_packets = 0;
1824 		adapter->total_rx_bytes = 0;
1825 		adapter->total_rx_packets = 0;
1826 		__napi_schedule(&adapter->napi);
1827 	}
1828 
1829 	return IRQ_HANDLED;
1830 }
1831 
1832 /**
1833  * e1000_intr - Interrupt Handler
1834  * @irq: interrupt number
1835  * @data: pointer to a network interface device structure
1836  **/
1837 static irqreturn_t e1000_intr(int __always_unused irq, void *data)
1838 {
1839 	struct net_device *netdev = data;
1840 	struct e1000_adapter *adapter = netdev_priv(netdev);
1841 	struct e1000_hw *hw = &adapter->hw;
1842 	u32 rctl, icr = er32(ICR);
1843 
1844 	if (!icr || test_bit(__E1000_DOWN, &adapter->state))
1845 		return IRQ_NONE;	/* Not our interrupt */
1846 
1847 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
1848 	 * not set, then the adapter didn't send an interrupt
1849 	 */
1850 	if (!(icr & E1000_ICR_INT_ASSERTED))
1851 		return IRQ_NONE;
1852 
1853 	/* Interrupt Auto-Mask...upon reading ICR,
1854 	 * interrupts are masked.  No need for the
1855 	 * IMC write
1856 	 */
1857 
1858 	if (icr & E1000_ICR_LSC) {
1859 		hw->mac.get_link_status = true;
1860 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1861 		 * disconnect (LSC) before accessing any PHY registers
1862 		 */
1863 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1864 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1865 			schedule_work(&adapter->downshift_task);
1866 
1867 		/* 80003ES2LAN workaround--
1868 		 * For packet buffer work-around on link down event;
1869 		 * disable receives here in the ISR and
1870 		 * reset adapter in watchdog
1871 		 */
1872 		if (netif_carrier_ok(netdev) &&
1873 		    (adapter->flags & FLAG_RX_NEEDS_RESTART)) {
1874 			/* disable receives */
1875 			rctl = er32(RCTL);
1876 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1877 			adapter->flags |= FLAG_RESTART_NOW;
1878 		}
1879 		/* guard against interrupt when we're going down */
1880 		if (!test_bit(__E1000_DOWN, &adapter->state))
1881 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1882 	}
1883 
1884 	/* Reset on uncorrectable ECC error */
1885 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1886 		u32 pbeccsts = er32(PBECCSTS);
1887 
1888 		adapter->corr_errors +=
1889 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1890 		adapter->uncorr_errors +=
1891 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1892 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1893 
1894 		/* Do the reset outside of interrupt context */
1895 		schedule_work(&adapter->reset_task);
1896 
1897 		/* return immediately since reset is imminent */
1898 		return IRQ_HANDLED;
1899 	}
1900 
1901 	if (napi_schedule_prep(&adapter->napi)) {
1902 		adapter->total_tx_bytes = 0;
1903 		adapter->total_tx_packets = 0;
1904 		adapter->total_rx_bytes = 0;
1905 		adapter->total_rx_packets = 0;
1906 		__napi_schedule(&adapter->napi);
1907 	}
1908 
1909 	return IRQ_HANDLED;
1910 }
1911 
1912 static irqreturn_t e1000_msix_other(int __always_unused irq, void *data)
1913 {
1914 	struct net_device *netdev = data;
1915 	struct e1000_adapter *adapter = netdev_priv(netdev);
1916 	struct e1000_hw *hw = &adapter->hw;
1917 	u32 icr = er32(ICR);
1918 
1919 	if (icr & adapter->eiac_mask)
1920 		ew32(ICS, (icr & adapter->eiac_mask));
1921 
1922 	if (icr & E1000_ICR_LSC) {
1923 		hw->mac.get_link_status = true;
1924 		/* guard against interrupt when we're going down */
1925 		if (!test_bit(__E1000_DOWN, &adapter->state))
1926 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1927 	}
1928 
1929 	if (!test_bit(__E1000_DOWN, &adapter->state))
1930 		ew32(IMS, E1000_IMS_OTHER | IMS_OTHER_MASK);
1931 
1932 	return IRQ_HANDLED;
1933 }
1934 
1935 static irqreturn_t e1000_intr_msix_tx(int __always_unused irq, void *data)
1936 {
1937 	struct net_device *netdev = data;
1938 	struct e1000_adapter *adapter = netdev_priv(netdev);
1939 	struct e1000_hw *hw = &adapter->hw;
1940 	struct e1000_ring *tx_ring = adapter->tx_ring;
1941 
1942 	adapter->total_tx_bytes = 0;
1943 	adapter->total_tx_packets = 0;
1944 
1945 	if (!e1000_clean_tx_irq(tx_ring))
1946 		/* Ring was not completely cleaned, so fire another interrupt */
1947 		ew32(ICS, tx_ring->ims_val);
1948 
1949 	if (!test_bit(__E1000_DOWN, &adapter->state))
1950 		ew32(IMS, adapter->tx_ring->ims_val);
1951 
1952 	return IRQ_HANDLED;
1953 }
1954 
1955 static irqreturn_t e1000_intr_msix_rx(int __always_unused irq, void *data)
1956 {
1957 	struct net_device *netdev = data;
1958 	struct e1000_adapter *adapter = netdev_priv(netdev);
1959 	struct e1000_ring *rx_ring = adapter->rx_ring;
1960 
1961 	/* Write the ITR value calculated at the end of the
1962 	 * previous interrupt.
1963 	 */
1964 	if (rx_ring->set_itr) {
1965 		u32 itr = rx_ring->itr_val ?
1966 			  1000000000 / (rx_ring->itr_val * 256) : 0;
1967 
1968 		writel(itr, rx_ring->itr_register);
1969 		rx_ring->set_itr = 0;
1970 	}
1971 
1972 	if (napi_schedule_prep(&adapter->napi)) {
1973 		adapter->total_rx_bytes = 0;
1974 		adapter->total_rx_packets = 0;
1975 		__napi_schedule(&adapter->napi);
1976 	}
1977 	return IRQ_HANDLED;
1978 }
1979 
1980 /**
1981  * e1000_configure_msix - Configure MSI-X hardware
1982  *
1983  * e1000_configure_msix sets up the hardware to properly
1984  * generate MSI-X interrupts.
1985  **/
1986 static void e1000_configure_msix(struct e1000_adapter *adapter)
1987 {
1988 	struct e1000_hw *hw = &adapter->hw;
1989 	struct e1000_ring *rx_ring = adapter->rx_ring;
1990 	struct e1000_ring *tx_ring = adapter->tx_ring;
1991 	int vector = 0;
1992 	u32 ctrl_ext, ivar = 0;
1993 
1994 	adapter->eiac_mask = 0;
1995 
1996 	/* Workaround issue with spurious interrupts on 82574 in MSI-X mode */
1997 	if (hw->mac.type == e1000_82574) {
1998 		u32 rfctl = er32(RFCTL);
1999 
2000 		rfctl |= E1000_RFCTL_ACK_DIS;
2001 		ew32(RFCTL, rfctl);
2002 	}
2003 
2004 	/* Configure Rx vector */
2005 	rx_ring->ims_val = E1000_IMS_RXQ0;
2006 	adapter->eiac_mask |= rx_ring->ims_val;
2007 	if (rx_ring->itr_val)
2008 		writel(1000000000 / (rx_ring->itr_val * 256),
2009 		       rx_ring->itr_register);
2010 	else
2011 		writel(1, rx_ring->itr_register);
2012 	ivar = E1000_IVAR_INT_ALLOC_VALID | vector;
2013 
2014 	/* Configure Tx vector */
2015 	tx_ring->ims_val = E1000_IMS_TXQ0;
2016 	vector++;
2017 	if (tx_ring->itr_val)
2018 		writel(1000000000 / (tx_ring->itr_val * 256),
2019 		       tx_ring->itr_register);
2020 	else
2021 		writel(1, tx_ring->itr_register);
2022 	adapter->eiac_mask |= tx_ring->ims_val;
2023 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 8);
2024 
2025 	/* set vector for Other Causes, e.g. link changes */
2026 	vector++;
2027 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 16);
2028 	if (rx_ring->itr_val)
2029 		writel(1000000000 / (rx_ring->itr_val * 256),
2030 		       hw->hw_addr + E1000_EITR_82574(vector));
2031 	else
2032 		writel(1, hw->hw_addr + E1000_EITR_82574(vector));
2033 
2034 	/* Cause Tx interrupts on every write back */
2035 	ivar |= BIT(31);
2036 
2037 	ew32(IVAR, ivar);
2038 
2039 	/* enable MSI-X PBA support */
2040 	ctrl_ext = er32(CTRL_EXT) & ~E1000_CTRL_EXT_IAME;
2041 	ctrl_ext |= E1000_CTRL_EXT_PBA_CLR | E1000_CTRL_EXT_EIAME;
2042 	ew32(CTRL_EXT, ctrl_ext);
2043 	e1e_flush();
2044 }
2045 
2046 void e1000e_reset_interrupt_capability(struct e1000_adapter *adapter)
2047 {
2048 	if (adapter->msix_entries) {
2049 		pci_disable_msix(adapter->pdev);
2050 		kfree(adapter->msix_entries);
2051 		adapter->msix_entries = NULL;
2052 	} else if (adapter->flags & FLAG_MSI_ENABLED) {
2053 		pci_disable_msi(adapter->pdev);
2054 		adapter->flags &= ~FLAG_MSI_ENABLED;
2055 	}
2056 }
2057 
2058 /**
2059  * e1000e_set_interrupt_capability - set MSI or MSI-X if supported
2060  *
2061  * Attempt to configure interrupts using the best available
2062  * capabilities of the hardware and kernel.
2063  **/
2064 void e1000e_set_interrupt_capability(struct e1000_adapter *adapter)
2065 {
2066 	int err;
2067 	int i;
2068 
2069 	switch (adapter->int_mode) {
2070 	case E1000E_INT_MODE_MSIX:
2071 		if (adapter->flags & FLAG_HAS_MSIX) {
2072 			adapter->num_vectors = 3; /* RxQ0, TxQ0 and other */
2073 			adapter->msix_entries = kcalloc(adapter->num_vectors,
2074 							sizeof(struct
2075 							       msix_entry),
2076 							GFP_KERNEL);
2077 			if (adapter->msix_entries) {
2078 				struct e1000_adapter *a = adapter;
2079 
2080 				for (i = 0; i < adapter->num_vectors; i++)
2081 					adapter->msix_entries[i].entry = i;
2082 
2083 				err = pci_enable_msix_range(a->pdev,
2084 							    a->msix_entries,
2085 							    a->num_vectors,
2086 							    a->num_vectors);
2087 				if (err > 0)
2088 					return;
2089 			}
2090 			/* MSI-X failed, so fall through and try MSI */
2091 			e_err("Failed to initialize MSI-X interrupts.  Falling back to MSI interrupts.\n");
2092 			e1000e_reset_interrupt_capability(adapter);
2093 		}
2094 		adapter->int_mode = E1000E_INT_MODE_MSI;
2095 		/* Fall through */
2096 	case E1000E_INT_MODE_MSI:
2097 		if (!pci_enable_msi(adapter->pdev)) {
2098 			adapter->flags |= FLAG_MSI_ENABLED;
2099 		} else {
2100 			adapter->int_mode = E1000E_INT_MODE_LEGACY;
2101 			e_err("Failed to initialize MSI interrupts.  Falling back to legacy interrupts.\n");
2102 		}
2103 		/* Fall through */
2104 	case E1000E_INT_MODE_LEGACY:
2105 		/* Don't do anything; this is the system default */
2106 		break;
2107 	}
2108 
2109 	/* store the number of vectors being used */
2110 	adapter->num_vectors = 1;
2111 }
2112 
2113 /**
2114  * e1000_request_msix - Initialize MSI-X interrupts
2115  *
2116  * e1000_request_msix allocates MSI-X vectors and requests interrupts from the
2117  * kernel.
2118  **/
2119 static int e1000_request_msix(struct e1000_adapter *adapter)
2120 {
2121 	struct net_device *netdev = adapter->netdev;
2122 	int err = 0, vector = 0;
2123 
2124 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2125 		snprintf(adapter->rx_ring->name,
2126 			 sizeof(adapter->rx_ring->name) - 1,
2127 			 "%s-rx-0", netdev->name);
2128 	else
2129 		memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
2130 	err = request_irq(adapter->msix_entries[vector].vector,
2131 			  e1000_intr_msix_rx, 0, adapter->rx_ring->name,
2132 			  netdev);
2133 	if (err)
2134 		return err;
2135 	adapter->rx_ring->itr_register = adapter->hw.hw_addr +
2136 	    E1000_EITR_82574(vector);
2137 	adapter->rx_ring->itr_val = adapter->itr;
2138 	vector++;
2139 
2140 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2141 		snprintf(adapter->tx_ring->name,
2142 			 sizeof(adapter->tx_ring->name) - 1,
2143 			 "%s-tx-0", netdev->name);
2144 	else
2145 		memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
2146 	err = request_irq(adapter->msix_entries[vector].vector,
2147 			  e1000_intr_msix_tx, 0, adapter->tx_ring->name,
2148 			  netdev);
2149 	if (err)
2150 		return err;
2151 	adapter->tx_ring->itr_register = adapter->hw.hw_addr +
2152 	    E1000_EITR_82574(vector);
2153 	adapter->tx_ring->itr_val = adapter->itr;
2154 	vector++;
2155 
2156 	err = request_irq(adapter->msix_entries[vector].vector,
2157 			  e1000_msix_other, 0, netdev->name, netdev);
2158 	if (err)
2159 		return err;
2160 
2161 	e1000_configure_msix(adapter);
2162 
2163 	return 0;
2164 }
2165 
2166 /**
2167  * e1000_request_irq - initialize interrupts
2168  *
2169  * Attempts to configure interrupts using the best available
2170  * capabilities of the hardware and kernel.
2171  **/
2172 static int e1000_request_irq(struct e1000_adapter *adapter)
2173 {
2174 	struct net_device *netdev = adapter->netdev;
2175 	int err;
2176 
2177 	if (adapter->msix_entries) {
2178 		err = e1000_request_msix(adapter);
2179 		if (!err)
2180 			return err;
2181 		/* fall back to MSI */
2182 		e1000e_reset_interrupt_capability(adapter);
2183 		adapter->int_mode = E1000E_INT_MODE_MSI;
2184 		e1000e_set_interrupt_capability(adapter);
2185 	}
2186 	if (adapter->flags & FLAG_MSI_ENABLED) {
2187 		err = request_irq(adapter->pdev->irq, e1000_intr_msi, 0,
2188 				  netdev->name, netdev);
2189 		if (!err)
2190 			return err;
2191 
2192 		/* fall back to legacy interrupt */
2193 		e1000e_reset_interrupt_capability(adapter);
2194 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
2195 	}
2196 
2197 	err = request_irq(adapter->pdev->irq, e1000_intr, IRQF_SHARED,
2198 			  netdev->name, netdev);
2199 	if (err)
2200 		e_err("Unable to allocate interrupt, Error: %d\n", err);
2201 
2202 	return err;
2203 }
2204 
2205 static void e1000_free_irq(struct e1000_adapter *adapter)
2206 {
2207 	struct net_device *netdev = adapter->netdev;
2208 
2209 	if (adapter->msix_entries) {
2210 		int vector = 0;
2211 
2212 		free_irq(adapter->msix_entries[vector].vector, netdev);
2213 		vector++;
2214 
2215 		free_irq(adapter->msix_entries[vector].vector, netdev);
2216 		vector++;
2217 
2218 		/* Other Causes interrupt vector */
2219 		free_irq(adapter->msix_entries[vector].vector, netdev);
2220 		return;
2221 	}
2222 
2223 	free_irq(adapter->pdev->irq, netdev);
2224 }
2225 
2226 /**
2227  * e1000_irq_disable - Mask off interrupt generation on the NIC
2228  **/
2229 static void e1000_irq_disable(struct e1000_adapter *adapter)
2230 {
2231 	struct e1000_hw *hw = &adapter->hw;
2232 
2233 	ew32(IMC, ~0);
2234 	if (adapter->msix_entries)
2235 		ew32(EIAC_82574, 0);
2236 	e1e_flush();
2237 
2238 	if (adapter->msix_entries) {
2239 		int i;
2240 
2241 		for (i = 0; i < adapter->num_vectors; i++)
2242 			synchronize_irq(adapter->msix_entries[i].vector);
2243 	} else {
2244 		synchronize_irq(adapter->pdev->irq);
2245 	}
2246 }
2247 
2248 /**
2249  * e1000_irq_enable - Enable default interrupt generation settings
2250  **/
2251 static void e1000_irq_enable(struct e1000_adapter *adapter)
2252 {
2253 	struct e1000_hw *hw = &adapter->hw;
2254 
2255 	if (adapter->msix_entries) {
2256 		ew32(EIAC_82574, adapter->eiac_mask & E1000_EIAC_MASK_82574);
2257 		ew32(IMS, adapter->eiac_mask | E1000_IMS_OTHER |
2258 		     IMS_OTHER_MASK);
2259 	} else if (hw->mac.type >= e1000_pch_lpt) {
2260 		ew32(IMS, IMS_ENABLE_MASK | E1000_IMS_ECCER);
2261 	} else {
2262 		ew32(IMS, IMS_ENABLE_MASK);
2263 	}
2264 	e1e_flush();
2265 }
2266 
2267 /**
2268  * e1000e_get_hw_control - get control of the h/w from f/w
2269  * @adapter: address of board private structure
2270  *
2271  * e1000e_get_hw_control sets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2272  * For ASF and Pass Through versions of f/w this means that
2273  * the driver is loaded. For AMT version (only with 82573)
2274  * of the f/w this means that the network i/f is open.
2275  **/
2276 void e1000e_get_hw_control(struct e1000_adapter *adapter)
2277 {
2278 	struct e1000_hw *hw = &adapter->hw;
2279 	u32 ctrl_ext;
2280 	u32 swsm;
2281 
2282 	/* Let firmware know the driver has taken over */
2283 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2284 		swsm = er32(SWSM);
2285 		ew32(SWSM, swsm | E1000_SWSM_DRV_LOAD);
2286 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2287 		ctrl_ext = er32(CTRL_EXT);
2288 		ew32(CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2289 	}
2290 }
2291 
2292 /**
2293  * e1000e_release_hw_control - release control of the h/w to f/w
2294  * @adapter: address of board private structure
2295  *
2296  * e1000e_release_hw_control resets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2297  * For ASF and Pass Through versions of f/w this means that the
2298  * driver is no longer loaded. For AMT version (only with 82573) i
2299  * of the f/w this means that the network i/f is closed.
2300  *
2301  **/
2302 void e1000e_release_hw_control(struct e1000_adapter *adapter)
2303 {
2304 	struct e1000_hw *hw = &adapter->hw;
2305 	u32 ctrl_ext;
2306 	u32 swsm;
2307 
2308 	/* Let firmware taken over control of h/w */
2309 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2310 		swsm = er32(SWSM);
2311 		ew32(SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
2312 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2313 		ctrl_ext = er32(CTRL_EXT);
2314 		ew32(CTRL_EXT, ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2315 	}
2316 }
2317 
2318 /**
2319  * e1000_alloc_ring_dma - allocate memory for a ring structure
2320  **/
2321 static int e1000_alloc_ring_dma(struct e1000_adapter *adapter,
2322 				struct e1000_ring *ring)
2323 {
2324 	struct pci_dev *pdev = adapter->pdev;
2325 
2326 	ring->desc = dma_zalloc_coherent(&pdev->dev, ring->size, &ring->dma,
2327 					 GFP_KERNEL);
2328 	if (!ring->desc)
2329 		return -ENOMEM;
2330 
2331 	return 0;
2332 }
2333 
2334 /**
2335  * e1000e_setup_tx_resources - allocate Tx resources (Descriptors)
2336  * @tx_ring: Tx descriptor ring
2337  *
2338  * Return 0 on success, negative on failure
2339  **/
2340 int e1000e_setup_tx_resources(struct e1000_ring *tx_ring)
2341 {
2342 	struct e1000_adapter *adapter = tx_ring->adapter;
2343 	int err = -ENOMEM, size;
2344 
2345 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2346 	tx_ring->buffer_info = vzalloc(size);
2347 	if (!tx_ring->buffer_info)
2348 		goto err;
2349 
2350 	/* round up to nearest 4K */
2351 	tx_ring->size = tx_ring->count * sizeof(struct e1000_tx_desc);
2352 	tx_ring->size = ALIGN(tx_ring->size, 4096);
2353 
2354 	err = e1000_alloc_ring_dma(adapter, tx_ring);
2355 	if (err)
2356 		goto err;
2357 
2358 	tx_ring->next_to_use = 0;
2359 	tx_ring->next_to_clean = 0;
2360 
2361 	return 0;
2362 err:
2363 	vfree(tx_ring->buffer_info);
2364 	e_err("Unable to allocate memory for the transmit descriptor ring\n");
2365 	return err;
2366 }
2367 
2368 /**
2369  * e1000e_setup_rx_resources - allocate Rx resources (Descriptors)
2370  * @rx_ring: Rx descriptor ring
2371  *
2372  * Returns 0 on success, negative on failure
2373  **/
2374 int e1000e_setup_rx_resources(struct e1000_ring *rx_ring)
2375 {
2376 	struct e1000_adapter *adapter = rx_ring->adapter;
2377 	struct e1000_buffer *buffer_info;
2378 	int i, size, desc_len, err = -ENOMEM;
2379 
2380 	size = sizeof(struct e1000_buffer) * rx_ring->count;
2381 	rx_ring->buffer_info = vzalloc(size);
2382 	if (!rx_ring->buffer_info)
2383 		goto err;
2384 
2385 	for (i = 0; i < rx_ring->count; i++) {
2386 		buffer_info = &rx_ring->buffer_info[i];
2387 		buffer_info->ps_pages = kcalloc(PS_PAGE_BUFFERS,
2388 						sizeof(struct e1000_ps_page),
2389 						GFP_KERNEL);
2390 		if (!buffer_info->ps_pages)
2391 			goto err_pages;
2392 	}
2393 
2394 	desc_len = sizeof(union e1000_rx_desc_packet_split);
2395 
2396 	/* Round up to nearest 4K */
2397 	rx_ring->size = rx_ring->count * desc_len;
2398 	rx_ring->size = ALIGN(rx_ring->size, 4096);
2399 
2400 	err = e1000_alloc_ring_dma(adapter, rx_ring);
2401 	if (err)
2402 		goto err_pages;
2403 
2404 	rx_ring->next_to_clean = 0;
2405 	rx_ring->next_to_use = 0;
2406 	rx_ring->rx_skb_top = NULL;
2407 
2408 	return 0;
2409 
2410 err_pages:
2411 	for (i = 0; i < rx_ring->count; i++) {
2412 		buffer_info = &rx_ring->buffer_info[i];
2413 		kfree(buffer_info->ps_pages);
2414 	}
2415 err:
2416 	vfree(rx_ring->buffer_info);
2417 	e_err("Unable to allocate memory for the receive descriptor ring\n");
2418 	return err;
2419 }
2420 
2421 /**
2422  * e1000_clean_tx_ring - Free Tx Buffers
2423  * @tx_ring: Tx descriptor ring
2424  **/
2425 static void e1000_clean_tx_ring(struct e1000_ring *tx_ring)
2426 {
2427 	struct e1000_adapter *adapter = tx_ring->adapter;
2428 	struct e1000_buffer *buffer_info;
2429 	unsigned long size;
2430 	unsigned int i;
2431 
2432 	for (i = 0; i < tx_ring->count; i++) {
2433 		buffer_info = &tx_ring->buffer_info[i];
2434 		e1000_put_txbuf(tx_ring, buffer_info, false);
2435 	}
2436 
2437 	netdev_reset_queue(adapter->netdev);
2438 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2439 	memset(tx_ring->buffer_info, 0, size);
2440 
2441 	memset(tx_ring->desc, 0, tx_ring->size);
2442 
2443 	tx_ring->next_to_use = 0;
2444 	tx_ring->next_to_clean = 0;
2445 }
2446 
2447 /**
2448  * e1000e_free_tx_resources - Free Tx Resources per Queue
2449  * @tx_ring: Tx descriptor ring
2450  *
2451  * Free all transmit software resources
2452  **/
2453 void e1000e_free_tx_resources(struct e1000_ring *tx_ring)
2454 {
2455 	struct e1000_adapter *adapter = tx_ring->adapter;
2456 	struct pci_dev *pdev = adapter->pdev;
2457 
2458 	e1000_clean_tx_ring(tx_ring);
2459 
2460 	vfree(tx_ring->buffer_info);
2461 	tx_ring->buffer_info = NULL;
2462 
2463 	dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
2464 			  tx_ring->dma);
2465 	tx_ring->desc = NULL;
2466 }
2467 
2468 /**
2469  * e1000e_free_rx_resources - Free Rx Resources
2470  * @rx_ring: Rx descriptor ring
2471  *
2472  * Free all receive software resources
2473  **/
2474 void e1000e_free_rx_resources(struct e1000_ring *rx_ring)
2475 {
2476 	struct e1000_adapter *adapter = rx_ring->adapter;
2477 	struct pci_dev *pdev = adapter->pdev;
2478 	int i;
2479 
2480 	e1000_clean_rx_ring(rx_ring);
2481 
2482 	for (i = 0; i < rx_ring->count; i++)
2483 		kfree(rx_ring->buffer_info[i].ps_pages);
2484 
2485 	vfree(rx_ring->buffer_info);
2486 	rx_ring->buffer_info = NULL;
2487 
2488 	dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
2489 			  rx_ring->dma);
2490 	rx_ring->desc = NULL;
2491 }
2492 
2493 /**
2494  * e1000_update_itr - update the dynamic ITR value based on statistics
2495  * @adapter: pointer to adapter
2496  * @itr_setting: current adapter->itr
2497  * @packets: the number of packets during this measurement interval
2498  * @bytes: the number of bytes during this measurement interval
2499  *
2500  *      Stores a new ITR value based on packets and byte
2501  *      counts during the last interrupt.  The advantage of per interrupt
2502  *      computation is faster updates and more accurate ITR for the current
2503  *      traffic pattern.  Constants in this function were computed
2504  *      based on theoretical maximum wire speed and thresholds were set based
2505  *      on testing data as well as attempting to minimize response time
2506  *      while increasing bulk throughput.  This functionality is controlled
2507  *      by the InterruptThrottleRate module parameter.
2508  **/
2509 static unsigned int e1000_update_itr(u16 itr_setting, int packets, int bytes)
2510 {
2511 	unsigned int retval = itr_setting;
2512 
2513 	if (packets == 0)
2514 		return itr_setting;
2515 
2516 	switch (itr_setting) {
2517 	case lowest_latency:
2518 		/* handle TSO and jumbo frames */
2519 		if (bytes / packets > 8000)
2520 			retval = bulk_latency;
2521 		else if ((packets < 5) && (bytes > 512))
2522 			retval = low_latency;
2523 		break;
2524 	case low_latency:	/* 50 usec aka 20000 ints/s */
2525 		if (bytes > 10000) {
2526 			/* this if handles the TSO accounting */
2527 			if (bytes / packets > 8000)
2528 				retval = bulk_latency;
2529 			else if ((packets < 10) || ((bytes / packets) > 1200))
2530 				retval = bulk_latency;
2531 			else if ((packets > 35))
2532 				retval = lowest_latency;
2533 		} else if (bytes / packets > 2000) {
2534 			retval = bulk_latency;
2535 		} else if (packets <= 2 && bytes < 512) {
2536 			retval = lowest_latency;
2537 		}
2538 		break;
2539 	case bulk_latency:	/* 250 usec aka 4000 ints/s */
2540 		if (bytes > 25000) {
2541 			if (packets > 35)
2542 				retval = low_latency;
2543 		} else if (bytes < 6000) {
2544 			retval = low_latency;
2545 		}
2546 		break;
2547 	}
2548 
2549 	return retval;
2550 }
2551 
2552 static void e1000_set_itr(struct e1000_adapter *adapter)
2553 {
2554 	u16 current_itr;
2555 	u32 new_itr = adapter->itr;
2556 
2557 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
2558 	if (adapter->link_speed != SPEED_1000) {
2559 		current_itr = 0;
2560 		new_itr = 4000;
2561 		goto set_itr_now;
2562 	}
2563 
2564 	if (adapter->flags2 & FLAG2_DISABLE_AIM) {
2565 		new_itr = 0;
2566 		goto set_itr_now;
2567 	}
2568 
2569 	adapter->tx_itr = e1000_update_itr(adapter->tx_itr,
2570 					   adapter->total_tx_packets,
2571 					   adapter->total_tx_bytes);
2572 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2573 	if (adapter->itr_setting == 3 && adapter->tx_itr == lowest_latency)
2574 		adapter->tx_itr = low_latency;
2575 
2576 	adapter->rx_itr = e1000_update_itr(adapter->rx_itr,
2577 					   adapter->total_rx_packets,
2578 					   adapter->total_rx_bytes);
2579 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2580 	if (adapter->itr_setting == 3 && adapter->rx_itr == lowest_latency)
2581 		adapter->rx_itr = low_latency;
2582 
2583 	current_itr = max(adapter->rx_itr, adapter->tx_itr);
2584 
2585 	/* counts and packets in update_itr are dependent on these numbers */
2586 	switch (current_itr) {
2587 	case lowest_latency:
2588 		new_itr = 70000;
2589 		break;
2590 	case low_latency:
2591 		new_itr = 20000;	/* aka hwitr = ~200 */
2592 		break;
2593 	case bulk_latency:
2594 		new_itr = 4000;
2595 		break;
2596 	default:
2597 		break;
2598 	}
2599 
2600 set_itr_now:
2601 	if (new_itr != adapter->itr) {
2602 		/* this attempts to bias the interrupt rate towards Bulk
2603 		 * by adding intermediate steps when interrupt rate is
2604 		 * increasing
2605 		 */
2606 		new_itr = new_itr > adapter->itr ?
2607 		    min(adapter->itr + (new_itr >> 2), new_itr) : new_itr;
2608 		adapter->itr = new_itr;
2609 		adapter->rx_ring->itr_val = new_itr;
2610 		if (adapter->msix_entries)
2611 			adapter->rx_ring->set_itr = 1;
2612 		else
2613 			e1000e_write_itr(adapter, new_itr);
2614 	}
2615 }
2616 
2617 /**
2618  * e1000e_write_itr - write the ITR value to the appropriate registers
2619  * @adapter: address of board private structure
2620  * @itr: new ITR value to program
2621  *
2622  * e1000e_write_itr determines if the adapter is in MSI-X mode
2623  * and, if so, writes the EITR registers with the ITR value.
2624  * Otherwise, it writes the ITR value into the ITR register.
2625  **/
2626 void e1000e_write_itr(struct e1000_adapter *adapter, u32 itr)
2627 {
2628 	struct e1000_hw *hw = &adapter->hw;
2629 	u32 new_itr = itr ? 1000000000 / (itr * 256) : 0;
2630 
2631 	if (adapter->msix_entries) {
2632 		int vector;
2633 
2634 		for (vector = 0; vector < adapter->num_vectors; vector++)
2635 			writel(new_itr, hw->hw_addr + E1000_EITR_82574(vector));
2636 	} else {
2637 		ew32(ITR, new_itr);
2638 	}
2639 }
2640 
2641 /**
2642  * e1000_alloc_queues - Allocate memory for all rings
2643  * @adapter: board private structure to initialize
2644  **/
2645 static int e1000_alloc_queues(struct e1000_adapter *adapter)
2646 {
2647 	int size = sizeof(struct e1000_ring);
2648 
2649 	adapter->tx_ring = kzalloc(size, GFP_KERNEL);
2650 	if (!adapter->tx_ring)
2651 		goto err;
2652 	adapter->tx_ring->count = adapter->tx_ring_count;
2653 	adapter->tx_ring->adapter = adapter;
2654 
2655 	adapter->rx_ring = kzalloc(size, GFP_KERNEL);
2656 	if (!adapter->rx_ring)
2657 		goto err;
2658 	adapter->rx_ring->count = adapter->rx_ring_count;
2659 	adapter->rx_ring->adapter = adapter;
2660 
2661 	return 0;
2662 err:
2663 	e_err("Unable to allocate memory for queues\n");
2664 	kfree(adapter->rx_ring);
2665 	kfree(adapter->tx_ring);
2666 	return -ENOMEM;
2667 }
2668 
2669 /**
2670  * e1000e_poll - NAPI Rx polling callback
2671  * @napi: struct associated with this polling callback
2672  * @weight: number of packets driver is allowed to process this poll
2673  **/
2674 static int e1000e_poll(struct napi_struct *napi, int weight)
2675 {
2676 	struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter,
2677 						     napi);
2678 	struct e1000_hw *hw = &adapter->hw;
2679 	struct net_device *poll_dev = adapter->netdev;
2680 	int tx_cleaned = 1, work_done = 0;
2681 
2682 	adapter = netdev_priv(poll_dev);
2683 
2684 	if (!adapter->msix_entries ||
2685 	    (adapter->rx_ring->ims_val & adapter->tx_ring->ims_val))
2686 		tx_cleaned = e1000_clean_tx_irq(adapter->tx_ring);
2687 
2688 	adapter->clean_rx(adapter->rx_ring, &work_done, weight);
2689 
2690 	if (!tx_cleaned)
2691 		work_done = weight;
2692 
2693 	/* If weight not fully consumed, exit the polling mode */
2694 	if (work_done < weight) {
2695 		if (adapter->itr_setting & 3)
2696 			e1000_set_itr(adapter);
2697 		napi_complete_done(napi, work_done);
2698 		if (!test_bit(__E1000_DOWN, &adapter->state)) {
2699 			if (adapter->msix_entries)
2700 				ew32(IMS, adapter->rx_ring->ims_val);
2701 			else
2702 				e1000_irq_enable(adapter);
2703 		}
2704 	}
2705 
2706 	return work_done;
2707 }
2708 
2709 static int e1000_vlan_rx_add_vid(struct net_device *netdev,
2710 				 __always_unused __be16 proto, u16 vid)
2711 {
2712 	struct e1000_adapter *adapter = netdev_priv(netdev);
2713 	struct e1000_hw *hw = &adapter->hw;
2714 	u32 vfta, index;
2715 
2716 	/* don't update vlan cookie if already programmed */
2717 	if ((adapter->hw.mng_cookie.status &
2718 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2719 	    (vid == adapter->mng_vlan_id))
2720 		return 0;
2721 
2722 	/* add VID to filter table */
2723 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2724 		index = (vid >> 5) & 0x7F;
2725 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2726 		vfta |= BIT((vid & 0x1F));
2727 		hw->mac.ops.write_vfta(hw, index, vfta);
2728 	}
2729 
2730 	set_bit(vid, adapter->active_vlans);
2731 
2732 	return 0;
2733 }
2734 
2735 static int e1000_vlan_rx_kill_vid(struct net_device *netdev,
2736 				  __always_unused __be16 proto, u16 vid)
2737 {
2738 	struct e1000_adapter *adapter = netdev_priv(netdev);
2739 	struct e1000_hw *hw = &adapter->hw;
2740 	u32 vfta, index;
2741 
2742 	if ((adapter->hw.mng_cookie.status &
2743 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2744 	    (vid == adapter->mng_vlan_id)) {
2745 		/* release control to f/w */
2746 		e1000e_release_hw_control(adapter);
2747 		return 0;
2748 	}
2749 
2750 	/* remove VID from filter table */
2751 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2752 		index = (vid >> 5) & 0x7F;
2753 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2754 		vfta &= ~BIT((vid & 0x1F));
2755 		hw->mac.ops.write_vfta(hw, index, vfta);
2756 	}
2757 
2758 	clear_bit(vid, adapter->active_vlans);
2759 
2760 	return 0;
2761 }
2762 
2763 /**
2764  * e1000e_vlan_filter_disable - helper to disable hw VLAN filtering
2765  * @adapter: board private structure to initialize
2766  **/
2767 static void e1000e_vlan_filter_disable(struct e1000_adapter *adapter)
2768 {
2769 	struct net_device *netdev = adapter->netdev;
2770 	struct e1000_hw *hw = &adapter->hw;
2771 	u32 rctl;
2772 
2773 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2774 		/* disable VLAN receive filtering */
2775 		rctl = er32(RCTL);
2776 		rctl &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
2777 		ew32(RCTL, rctl);
2778 
2779 		if (adapter->mng_vlan_id != (u16)E1000_MNG_VLAN_NONE) {
2780 			e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
2781 					       adapter->mng_vlan_id);
2782 			adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
2783 		}
2784 	}
2785 }
2786 
2787 /**
2788  * e1000e_vlan_filter_enable - helper to enable HW VLAN filtering
2789  * @adapter: board private structure to initialize
2790  **/
2791 static void e1000e_vlan_filter_enable(struct e1000_adapter *adapter)
2792 {
2793 	struct e1000_hw *hw = &adapter->hw;
2794 	u32 rctl;
2795 
2796 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2797 		/* enable VLAN receive filtering */
2798 		rctl = er32(RCTL);
2799 		rctl |= E1000_RCTL_VFE;
2800 		rctl &= ~E1000_RCTL_CFIEN;
2801 		ew32(RCTL, rctl);
2802 	}
2803 }
2804 
2805 /**
2806  * e1000e_vlan_strip_disable - helper to disable HW VLAN stripping
2807  * @adapter: board private structure to initialize
2808  **/
2809 static void e1000e_vlan_strip_disable(struct e1000_adapter *adapter)
2810 {
2811 	struct e1000_hw *hw = &adapter->hw;
2812 	u32 ctrl;
2813 
2814 	/* disable VLAN tag insert/strip */
2815 	ctrl = er32(CTRL);
2816 	ctrl &= ~E1000_CTRL_VME;
2817 	ew32(CTRL, ctrl);
2818 }
2819 
2820 /**
2821  * e1000e_vlan_strip_enable - helper to enable HW VLAN stripping
2822  * @adapter: board private structure to initialize
2823  **/
2824 static void e1000e_vlan_strip_enable(struct e1000_adapter *adapter)
2825 {
2826 	struct e1000_hw *hw = &adapter->hw;
2827 	u32 ctrl;
2828 
2829 	/* enable VLAN tag insert/strip */
2830 	ctrl = er32(CTRL);
2831 	ctrl |= E1000_CTRL_VME;
2832 	ew32(CTRL, ctrl);
2833 }
2834 
2835 static void e1000_update_mng_vlan(struct e1000_adapter *adapter)
2836 {
2837 	struct net_device *netdev = adapter->netdev;
2838 	u16 vid = adapter->hw.mng_cookie.vlan_id;
2839 	u16 old_vid = adapter->mng_vlan_id;
2840 
2841 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
2842 		e1000_vlan_rx_add_vid(netdev, htons(ETH_P_8021Q), vid);
2843 		adapter->mng_vlan_id = vid;
2844 	}
2845 
2846 	if ((old_vid != (u16)E1000_MNG_VLAN_NONE) && (vid != old_vid))
2847 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q), old_vid);
2848 }
2849 
2850 static void e1000_restore_vlan(struct e1000_adapter *adapter)
2851 {
2852 	u16 vid;
2853 
2854 	e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
2855 
2856 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
2857 	    e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
2858 }
2859 
2860 static void e1000_init_manageability_pt(struct e1000_adapter *adapter)
2861 {
2862 	struct e1000_hw *hw = &adapter->hw;
2863 	u32 manc, manc2h, mdef, i, j;
2864 
2865 	if (!(adapter->flags & FLAG_MNG_PT_ENABLED))
2866 		return;
2867 
2868 	manc = er32(MANC);
2869 
2870 	/* enable receiving management packets to the host. this will probably
2871 	 * generate destination unreachable messages from the host OS, but
2872 	 * the packets will be handled on SMBUS
2873 	 */
2874 	manc |= E1000_MANC_EN_MNG2HOST;
2875 	manc2h = er32(MANC2H);
2876 
2877 	switch (hw->mac.type) {
2878 	default:
2879 		manc2h |= (E1000_MANC2H_PORT_623 | E1000_MANC2H_PORT_664);
2880 		break;
2881 	case e1000_82574:
2882 	case e1000_82583:
2883 		/* Check if IPMI pass-through decision filter already exists;
2884 		 * if so, enable it.
2885 		 */
2886 		for (i = 0, j = 0; i < 8; i++) {
2887 			mdef = er32(MDEF(i));
2888 
2889 			/* Ignore filters with anything other than IPMI ports */
2890 			if (mdef & ~(E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2891 				continue;
2892 
2893 			/* Enable this decision filter in MANC2H */
2894 			if (mdef)
2895 				manc2h |= BIT(i);
2896 
2897 			j |= mdef;
2898 		}
2899 
2900 		if (j == (E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2901 			break;
2902 
2903 		/* Create new decision filter in an empty filter */
2904 		for (i = 0, j = 0; i < 8; i++)
2905 			if (er32(MDEF(i)) == 0) {
2906 				ew32(MDEF(i), (E1000_MDEF_PORT_623 |
2907 					       E1000_MDEF_PORT_664));
2908 				manc2h |= BIT(1);
2909 				j++;
2910 				break;
2911 			}
2912 
2913 		if (!j)
2914 			e_warn("Unable to create IPMI pass-through filter\n");
2915 		break;
2916 	}
2917 
2918 	ew32(MANC2H, manc2h);
2919 	ew32(MANC, manc);
2920 }
2921 
2922 /**
2923  * e1000_configure_tx - Configure Transmit Unit after Reset
2924  * @adapter: board private structure
2925  *
2926  * Configure the Tx unit of the MAC after a reset.
2927  **/
2928 static void e1000_configure_tx(struct e1000_adapter *adapter)
2929 {
2930 	struct e1000_hw *hw = &adapter->hw;
2931 	struct e1000_ring *tx_ring = adapter->tx_ring;
2932 	u64 tdba;
2933 	u32 tdlen, tctl, tarc;
2934 
2935 	/* Setup the HW Tx Head and Tail descriptor pointers */
2936 	tdba = tx_ring->dma;
2937 	tdlen = tx_ring->count * sizeof(struct e1000_tx_desc);
2938 	ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
2939 	ew32(TDBAH(0), (tdba >> 32));
2940 	ew32(TDLEN(0), tdlen);
2941 	ew32(TDH(0), 0);
2942 	ew32(TDT(0), 0);
2943 	tx_ring->head = adapter->hw.hw_addr + E1000_TDH(0);
2944 	tx_ring->tail = adapter->hw.hw_addr + E1000_TDT(0);
2945 
2946 	writel(0, tx_ring->head);
2947 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
2948 		e1000e_update_tdt_wa(tx_ring, 0);
2949 	else
2950 		writel(0, tx_ring->tail);
2951 
2952 	/* Set the Tx Interrupt Delay register */
2953 	ew32(TIDV, adapter->tx_int_delay);
2954 	/* Tx irq moderation */
2955 	ew32(TADV, adapter->tx_abs_int_delay);
2956 
2957 	if (adapter->flags2 & FLAG2_DMA_BURST) {
2958 		u32 txdctl = er32(TXDCTL(0));
2959 
2960 		txdctl &= ~(E1000_TXDCTL_PTHRESH | E1000_TXDCTL_HTHRESH |
2961 			    E1000_TXDCTL_WTHRESH);
2962 		/* set up some performance related parameters to encourage the
2963 		 * hardware to use the bus more efficiently in bursts, depends
2964 		 * on the tx_int_delay to be enabled,
2965 		 * wthresh = 1 ==> burst write is disabled to avoid Tx stalls
2966 		 * hthresh = 1 ==> prefetch when one or more available
2967 		 * pthresh = 0x1f ==> prefetch if internal cache 31 or less
2968 		 * BEWARE: this seems to work but should be considered first if
2969 		 * there are Tx hangs or other Tx related bugs
2970 		 */
2971 		txdctl |= E1000_TXDCTL_DMA_BURST_ENABLE;
2972 		ew32(TXDCTL(0), txdctl);
2973 	}
2974 	/* erratum work around: set txdctl the same for both queues */
2975 	ew32(TXDCTL(1), er32(TXDCTL(0)));
2976 
2977 	/* Program the Transmit Control Register */
2978 	tctl = er32(TCTL);
2979 	tctl &= ~E1000_TCTL_CT;
2980 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
2981 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
2982 
2983 	if (adapter->flags & FLAG_TARC_SPEED_MODE_BIT) {
2984 		tarc = er32(TARC(0));
2985 		/* set the speed mode bit, we'll clear it if we're not at
2986 		 * gigabit link later
2987 		 */
2988 #define SPEED_MODE_BIT BIT(21)
2989 		tarc |= SPEED_MODE_BIT;
2990 		ew32(TARC(0), tarc);
2991 	}
2992 
2993 	/* errata: program both queues to unweighted RR */
2994 	if (adapter->flags & FLAG_TARC_SET_BIT_ZERO) {
2995 		tarc = er32(TARC(0));
2996 		tarc |= 1;
2997 		ew32(TARC(0), tarc);
2998 		tarc = er32(TARC(1));
2999 		tarc |= 1;
3000 		ew32(TARC(1), tarc);
3001 	}
3002 
3003 	/* Setup Transmit Descriptor Settings for eop descriptor */
3004 	adapter->txd_cmd = E1000_TXD_CMD_EOP | E1000_TXD_CMD_IFCS;
3005 
3006 	/* only set IDE if we are delaying interrupts using the timers */
3007 	if (adapter->tx_int_delay)
3008 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
3009 
3010 	/* enable Report Status bit */
3011 	adapter->txd_cmd |= E1000_TXD_CMD_RS;
3012 
3013 	ew32(TCTL, tctl);
3014 
3015 	hw->mac.ops.config_collision_dist(hw);
3016 
3017 	/* SPT and KBL Si errata workaround to avoid data corruption */
3018 	if (hw->mac.type == e1000_pch_spt) {
3019 		u32 reg_val;
3020 
3021 		reg_val = er32(IOSFPC);
3022 		reg_val |= E1000_RCTL_RDMTS_HEX;
3023 		ew32(IOSFPC, reg_val);
3024 
3025 		reg_val = er32(TARC(0));
3026 		/* SPT and KBL Si errata workaround to avoid Tx hang.
3027 		 * Dropping the number of outstanding requests from
3028 		 * 3 to 2 in order to avoid a buffer overrun.
3029 		 */
3030 		reg_val &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3031 		reg_val |= E1000_TARC0_CB_MULTIQ_2_REQ;
3032 		ew32(TARC(0), reg_val);
3033 	}
3034 }
3035 
3036 /**
3037  * e1000_setup_rctl - configure the receive control registers
3038  * @adapter: Board private structure
3039  **/
3040 #define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
3041 			   (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
3042 static void e1000_setup_rctl(struct e1000_adapter *adapter)
3043 {
3044 	struct e1000_hw *hw = &adapter->hw;
3045 	u32 rctl, rfctl;
3046 	u32 pages = 0;
3047 
3048 	/* Workaround Si errata on PCHx - configure jumbo frame flow.
3049 	 * If jumbo frames not set, program related MAC/PHY registers
3050 	 * to h/w defaults
3051 	 */
3052 	if (hw->mac.type >= e1000_pch2lan) {
3053 		s32 ret_val;
3054 
3055 		if (adapter->netdev->mtu > ETH_DATA_LEN)
3056 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, true);
3057 		else
3058 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, false);
3059 
3060 		if (ret_val)
3061 			e_dbg("failed to enable|disable jumbo frame workaround mode\n");
3062 	}
3063 
3064 	/* Program MC offset vector base */
3065 	rctl = er32(RCTL);
3066 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3067 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3068 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3069 	    (adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3070 
3071 	/* Do not Store bad packets */
3072 	rctl &= ~E1000_RCTL_SBP;
3073 
3074 	/* Enable Long Packet receive */
3075 	if (adapter->netdev->mtu <= ETH_DATA_LEN)
3076 		rctl &= ~E1000_RCTL_LPE;
3077 	else
3078 		rctl |= E1000_RCTL_LPE;
3079 
3080 	/* Some systems expect that the CRC is included in SMBUS traffic. The
3081 	 * hardware strips the CRC before sending to both SMBUS (BMC) and to
3082 	 * host memory when this is enabled
3083 	 */
3084 	if (adapter->flags2 & FLAG2_CRC_STRIPPING)
3085 		rctl |= E1000_RCTL_SECRC;
3086 
3087 	/* Workaround Si errata on 82577 PHY - configure IPG for jumbos */
3088 	if ((hw->phy.type == e1000_phy_82577) && (rctl & E1000_RCTL_LPE)) {
3089 		u16 phy_data;
3090 
3091 		e1e_rphy(hw, PHY_REG(770, 26), &phy_data);
3092 		phy_data &= 0xfff8;
3093 		phy_data |= BIT(2);
3094 		e1e_wphy(hw, PHY_REG(770, 26), phy_data);
3095 
3096 		e1e_rphy(hw, 22, &phy_data);
3097 		phy_data &= 0x0fff;
3098 		phy_data |= BIT(14);
3099 		e1e_wphy(hw, 0x10, 0x2823);
3100 		e1e_wphy(hw, 0x11, 0x0003);
3101 		e1e_wphy(hw, 22, phy_data);
3102 	}
3103 
3104 	/* Setup buffer sizes */
3105 	rctl &= ~E1000_RCTL_SZ_4096;
3106 	rctl |= E1000_RCTL_BSEX;
3107 	switch (adapter->rx_buffer_len) {
3108 	case 2048:
3109 	default:
3110 		rctl |= E1000_RCTL_SZ_2048;
3111 		rctl &= ~E1000_RCTL_BSEX;
3112 		break;
3113 	case 4096:
3114 		rctl |= E1000_RCTL_SZ_4096;
3115 		break;
3116 	case 8192:
3117 		rctl |= E1000_RCTL_SZ_8192;
3118 		break;
3119 	case 16384:
3120 		rctl |= E1000_RCTL_SZ_16384;
3121 		break;
3122 	}
3123 
3124 	/* Enable Extended Status in all Receive Descriptors */
3125 	rfctl = er32(RFCTL);
3126 	rfctl |= E1000_RFCTL_EXTEN;
3127 	ew32(RFCTL, rfctl);
3128 
3129 	/* 82571 and greater support packet-split where the protocol
3130 	 * header is placed in skb->data and the packet data is
3131 	 * placed in pages hanging off of skb_shinfo(skb)->nr_frags.
3132 	 * In the case of a non-split, skb->data is linearly filled,
3133 	 * followed by the page buffers.  Therefore, skb->data is
3134 	 * sized to hold the largest protocol header.
3135 	 *
3136 	 * allocations using alloc_page take too long for regular MTU
3137 	 * so only enable packet split for jumbo frames
3138 	 *
3139 	 * Using pages when the page size is greater than 16k wastes
3140 	 * a lot of memory, since we allocate 3 pages at all times
3141 	 * per packet.
3142 	 */
3143 	pages = PAGE_USE_COUNT(adapter->netdev->mtu);
3144 	if ((pages <= 3) && (PAGE_SIZE <= 16384) && (rctl & E1000_RCTL_LPE))
3145 		adapter->rx_ps_pages = pages;
3146 	else
3147 		adapter->rx_ps_pages = 0;
3148 
3149 	if (adapter->rx_ps_pages) {
3150 		u32 psrctl = 0;
3151 
3152 		/* Enable Packet split descriptors */
3153 		rctl |= E1000_RCTL_DTYP_PS;
3154 
3155 		psrctl |= adapter->rx_ps_bsize0 >> E1000_PSRCTL_BSIZE0_SHIFT;
3156 
3157 		switch (adapter->rx_ps_pages) {
3158 		case 3:
3159 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE3_SHIFT;
3160 			/* fall-through */
3161 		case 2:
3162 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE2_SHIFT;
3163 			/* fall-through */
3164 		case 1:
3165 			psrctl |= PAGE_SIZE >> E1000_PSRCTL_BSIZE1_SHIFT;
3166 			break;
3167 		}
3168 
3169 		ew32(PSRCTL, psrctl);
3170 	}
3171 
3172 	/* This is useful for sniffing bad packets. */
3173 	if (adapter->netdev->features & NETIF_F_RXALL) {
3174 		/* UPE and MPE will be handled by normal PROMISC logic
3175 		 * in e1000e_set_rx_mode
3176 		 */
3177 		rctl |= (E1000_RCTL_SBP |	/* Receive bad packets */
3178 			 E1000_RCTL_BAM |	/* RX All Bcast Pkts */
3179 			 E1000_RCTL_PMCF);	/* RX All MAC Ctrl Pkts */
3180 
3181 		rctl &= ~(E1000_RCTL_VFE |	/* Disable VLAN filter */
3182 			  E1000_RCTL_DPF |	/* Allow filtered pause */
3183 			  E1000_RCTL_CFIEN);	/* Dis VLAN CFIEN Filter */
3184 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3185 		 * and that breaks VLANs.
3186 		 */
3187 	}
3188 
3189 	ew32(RCTL, rctl);
3190 	/* just started the receive unit, no need to restart */
3191 	adapter->flags &= ~FLAG_RESTART_NOW;
3192 }
3193 
3194 /**
3195  * e1000_configure_rx - Configure Receive Unit after Reset
3196  * @adapter: board private structure
3197  *
3198  * Configure the Rx unit of the MAC after a reset.
3199  **/
3200 static void e1000_configure_rx(struct e1000_adapter *adapter)
3201 {
3202 	struct e1000_hw *hw = &adapter->hw;
3203 	struct e1000_ring *rx_ring = adapter->rx_ring;
3204 	u64 rdba;
3205 	u32 rdlen, rctl, rxcsum, ctrl_ext;
3206 
3207 	if (adapter->rx_ps_pages) {
3208 		/* this is a 32 byte descriptor */
3209 		rdlen = rx_ring->count *
3210 		    sizeof(union e1000_rx_desc_packet_split);
3211 		adapter->clean_rx = e1000_clean_rx_irq_ps;
3212 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers_ps;
3213 	} else if (adapter->netdev->mtu > ETH_FRAME_LEN + ETH_FCS_LEN) {
3214 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3215 		adapter->clean_rx = e1000_clean_jumbo_rx_irq;
3216 		adapter->alloc_rx_buf = e1000_alloc_jumbo_rx_buffers;
3217 	} else {
3218 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3219 		adapter->clean_rx = e1000_clean_rx_irq;
3220 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers;
3221 	}
3222 
3223 	/* disable receives while setting up the descriptors */
3224 	rctl = er32(RCTL);
3225 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
3226 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
3227 	e1e_flush();
3228 	usleep_range(10000, 20000);
3229 
3230 	if (adapter->flags2 & FLAG2_DMA_BURST) {
3231 		/* set the writeback threshold (only takes effect if the RDTR
3232 		 * is set). set GRAN=1 and write back up to 0x4 worth, and
3233 		 * enable prefetching of 0x20 Rx descriptors
3234 		 * granularity = 01
3235 		 * wthresh = 04,
3236 		 * hthresh = 04,
3237 		 * pthresh = 0x20
3238 		 */
3239 		ew32(RXDCTL(0), E1000_RXDCTL_DMA_BURST_ENABLE);
3240 		ew32(RXDCTL(1), E1000_RXDCTL_DMA_BURST_ENABLE);
3241 	}
3242 
3243 	/* set the Receive Delay Timer Register */
3244 	ew32(RDTR, adapter->rx_int_delay);
3245 
3246 	/* irq moderation */
3247 	ew32(RADV, adapter->rx_abs_int_delay);
3248 	if ((adapter->itr_setting != 0) && (adapter->itr != 0))
3249 		e1000e_write_itr(adapter, adapter->itr);
3250 
3251 	ctrl_ext = er32(CTRL_EXT);
3252 	/* Auto-Mask interrupts upon ICR access */
3253 	ctrl_ext |= E1000_CTRL_EXT_IAME;
3254 	ew32(IAM, 0xffffffff);
3255 	ew32(CTRL_EXT, ctrl_ext);
3256 	e1e_flush();
3257 
3258 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
3259 	 * the Base and Length of the Rx Descriptor Ring
3260 	 */
3261 	rdba = rx_ring->dma;
3262 	ew32(RDBAL(0), (rdba & DMA_BIT_MASK(32)));
3263 	ew32(RDBAH(0), (rdba >> 32));
3264 	ew32(RDLEN(0), rdlen);
3265 	ew32(RDH(0), 0);
3266 	ew32(RDT(0), 0);
3267 	rx_ring->head = adapter->hw.hw_addr + E1000_RDH(0);
3268 	rx_ring->tail = adapter->hw.hw_addr + E1000_RDT(0);
3269 
3270 	writel(0, rx_ring->head);
3271 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
3272 		e1000e_update_rdt_wa(rx_ring, 0);
3273 	else
3274 		writel(0, rx_ring->tail);
3275 
3276 	/* Enable Receive Checksum Offload for TCP and UDP */
3277 	rxcsum = er32(RXCSUM);
3278 	if (adapter->netdev->features & NETIF_F_RXCSUM)
3279 		rxcsum |= E1000_RXCSUM_TUOFL;
3280 	else
3281 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3282 	ew32(RXCSUM, rxcsum);
3283 
3284 	/* With jumbo frames, excessive C-state transition latencies result
3285 	 * in dropped transactions.
3286 	 */
3287 	if (adapter->netdev->mtu > ETH_DATA_LEN) {
3288 		u32 lat =
3289 		    ((er32(PBA) & E1000_PBA_RXA_MASK) * 1024 -
3290 		     adapter->max_frame_size) * 8 / 1000;
3291 
3292 		if (adapter->flags & FLAG_IS_ICH) {
3293 			u32 rxdctl = er32(RXDCTL(0));
3294 
3295 			ew32(RXDCTL(0), rxdctl | 0x3 | BIT(8));
3296 		}
3297 
3298 		dev_info(&adapter->pdev->dev,
3299 			 "Some CPU C-states have been disabled in order to enable jumbo frames\n");
3300 		pm_qos_update_request(&adapter->pm_qos_req, lat);
3301 	} else {
3302 		pm_qos_update_request(&adapter->pm_qos_req,
3303 				      PM_QOS_DEFAULT_VALUE);
3304 	}
3305 
3306 	/* Enable Receives */
3307 	ew32(RCTL, rctl);
3308 }
3309 
3310 /**
3311  * e1000e_write_mc_addr_list - write multicast addresses to MTA
3312  * @netdev: network interface device structure
3313  *
3314  * Writes multicast address list to the MTA hash table.
3315  * Returns: -ENOMEM on failure
3316  *                0 on no addresses written
3317  *                X on writing X addresses to MTA
3318  */
3319 static int e1000e_write_mc_addr_list(struct net_device *netdev)
3320 {
3321 	struct e1000_adapter *adapter = netdev_priv(netdev);
3322 	struct e1000_hw *hw = &adapter->hw;
3323 	struct netdev_hw_addr *ha;
3324 	u8 *mta_list;
3325 	int i;
3326 
3327 	if (netdev_mc_empty(netdev)) {
3328 		/* nothing to program, so clear mc list */
3329 		hw->mac.ops.update_mc_addr_list(hw, NULL, 0);
3330 		return 0;
3331 	}
3332 
3333 	mta_list = kzalloc(netdev_mc_count(netdev) * ETH_ALEN, GFP_ATOMIC);
3334 	if (!mta_list)
3335 		return -ENOMEM;
3336 
3337 	/* update_mc_addr_list expects a packed array of only addresses. */
3338 	i = 0;
3339 	netdev_for_each_mc_addr(ha, netdev)
3340 	    memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3341 
3342 	hw->mac.ops.update_mc_addr_list(hw, mta_list, i);
3343 	kfree(mta_list);
3344 
3345 	return netdev_mc_count(netdev);
3346 }
3347 
3348 /**
3349  * e1000e_write_uc_addr_list - write unicast addresses to RAR table
3350  * @netdev: network interface device structure
3351  *
3352  * Writes unicast address list to the RAR table.
3353  * Returns: -ENOMEM on failure/insufficient address space
3354  *                0 on no addresses written
3355  *                X on writing X addresses to the RAR table
3356  **/
3357 static int e1000e_write_uc_addr_list(struct net_device *netdev)
3358 {
3359 	struct e1000_adapter *adapter = netdev_priv(netdev);
3360 	struct e1000_hw *hw = &adapter->hw;
3361 	unsigned int rar_entries;
3362 	int count = 0;
3363 
3364 	rar_entries = hw->mac.ops.rar_get_count(hw);
3365 
3366 	/* save a rar entry for our hardware address */
3367 	rar_entries--;
3368 
3369 	/* save a rar entry for the LAA workaround */
3370 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA)
3371 		rar_entries--;
3372 
3373 	/* return ENOMEM indicating insufficient memory for addresses */
3374 	if (netdev_uc_count(netdev) > rar_entries)
3375 		return -ENOMEM;
3376 
3377 	if (!netdev_uc_empty(netdev) && rar_entries) {
3378 		struct netdev_hw_addr *ha;
3379 
3380 		/* write the addresses in reverse order to avoid write
3381 		 * combining
3382 		 */
3383 		netdev_for_each_uc_addr(ha, netdev) {
3384 			int ret_val;
3385 
3386 			if (!rar_entries)
3387 				break;
3388 			ret_val = hw->mac.ops.rar_set(hw, ha->addr, rar_entries--);
3389 			if (ret_val < 0)
3390 				return -ENOMEM;
3391 			count++;
3392 		}
3393 	}
3394 
3395 	/* zero out the remaining RAR entries not used above */
3396 	for (; rar_entries > 0; rar_entries--) {
3397 		ew32(RAH(rar_entries), 0);
3398 		ew32(RAL(rar_entries), 0);
3399 	}
3400 	e1e_flush();
3401 
3402 	return count;
3403 }
3404 
3405 /**
3406  * e1000e_set_rx_mode - secondary unicast, Multicast and Promiscuous mode set
3407  * @netdev: network interface device structure
3408  *
3409  * The ndo_set_rx_mode entry point is called whenever the unicast or multicast
3410  * address list or the network interface flags are updated.  This routine is
3411  * responsible for configuring the hardware for proper unicast, multicast,
3412  * promiscuous mode, and all-multi behavior.
3413  **/
3414 static void e1000e_set_rx_mode(struct net_device *netdev)
3415 {
3416 	struct e1000_adapter *adapter = netdev_priv(netdev);
3417 	struct e1000_hw *hw = &adapter->hw;
3418 	u32 rctl;
3419 
3420 	if (pm_runtime_suspended(netdev->dev.parent))
3421 		return;
3422 
3423 	/* Check for Promiscuous and All Multicast modes */
3424 	rctl = er32(RCTL);
3425 
3426 	/* clear the affected bits */
3427 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
3428 
3429 	if (netdev->flags & IFF_PROMISC) {
3430 		rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
3431 		/* Do not hardware filter VLANs in promisc mode */
3432 		e1000e_vlan_filter_disable(adapter);
3433 	} else {
3434 		int count;
3435 
3436 		if (netdev->flags & IFF_ALLMULTI) {
3437 			rctl |= E1000_RCTL_MPE;
3438 		} else {
3439 			/* Write addresses to the MTA, if the attempt fails
3440 			 * then we should just turn on promiscuous mode so
3441 			 * that we can at least receive multicast traffic
3442 			 */
3443 			count = e1000e_write_mc_addr_list(netdev);
3444 			if (count < 0)
3445 				rctl |= E1000_RCTL_MPE;
3446 		}
3447 		e1000e_vlan_filter_enable(adapter);
3448 		/* Write addresses to available RAR registers, if there is not
3449 		 * sufficient space to store all the addresses then enable
3450 		 * unicast promiscuous mode
3451 		 */
3452 		count = e1000e_write_uc_addr_list(netdev);
3453 		if (count < 0)
3454 			rctl |= E1000_RCTL_UPE;
3455 	}
3456 
3457 	ew32(RCTL, rctl);
3458 
3459 	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
3460 		e1000e_vlan_strip_enable(adapter);
3461 	else
3462 		e1000e_vlan_strip_disable(adapter);
3463 }
3464 
3465 static void e1000e_setup_rss_hash(struct e1000_adapter *adapter)
3466 {
3467 	struct e1000_hw *hw = &adapter->hw;
3468 	u32 mrqc, rxcsum;
3469 	u32 rss_key[10];
3470 	int i;
3471 
3472 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3473 	for (i = 0; i < 10; i++)
3474 		ew32(RSSRK(i), rss_key[i]);
3475 
3476 	/* Direct all traffic to queue 0 */
3477 	for (i = 0; i < 32; i++)
3478 		ew32(RETA(i), 0);
3479 
3480 	/* Disable raw packet checksumming so that RSS hash is placed in
3481 	 * descriptor on writeback.
3482 	 */
3483 	rxcsum = er32(RXCSUM);
3484 	rxcsum |= E1000_RXCSUM_PCSD;
3485 
3486 	ew32(RXCSUM, rxcsum);
3487 
3488 	mrqc = (E1000_MRQC_RSS_FIELD_IPV4 |
3489 		E1000_MRQC_RSS_FIELD_IPV4_TCP |
3490 		E1000_MRQC_RSS_FIELD_IPV6 |
3491 		E1000_MRQC_RSS_FIELD_IPV6_TCP |
3492 		E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3493 
3494 	ew32(MRQC, mrqc);
3495 }
3496 
3497 /**
3498  * e1000e_get_base_timinca - get default SYSTIM time increment attributes
3499  * @adapter: board private structure
3500  * @timinca: pointer to returned time increment attributes
3501  *
3502  * Get attributes for incrementing the System Time Register SYSTIML/H at
3503  * the default base frequency, and set the cyclecounter shift value.
3504  **/
3505 s32 e1000e_get_base_timinca(struct e1000_adapter *adapter, u32 *timinca)
3506 {
3507 	struct e1000_hw *hw = &adapter->hw;
3508 	u32 incvalue, incperiod, shift;
3509 
3510 	/* Make sure clock is enabled on I217/I218/I219  before checking
3511 	 * the frequency
3512 	 */
3513 	if ((hw->mac.type >= e1000_pch_lpt) &&
3514 	    !(er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) &&
3515 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_ENABLED)) {
3516 		u32 fextnvm7 = er32(FEXTNVM7);
3517 
3518 		if (!(fextnvm7 & BIT(0))) {
3519 			ew32(FEXTNVM7, fextnvm7 | BIT(0));
3520 			e1e_flush();
3521 		}
3522 	}
3523 
3524 	switch (hw->mac.type) {
3525 	case e1000_pch2lan:
3526 		/* Stable 96MHz frequency */
3527 		incperiod = INCPERIOD_96MHZ;
3528 		incvalue = INCVALUE_96MHZ;
3529 		shift = INCVALUE_SHIFT_96MHZ;
3530 		adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3531 		break;
3532 	case e1000_pch_lpt:
3533 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3534 			/* Stable 96MHz frequency */
3535 			incperiod = INCPERIOD_96MHZ;
3536 			incvalue = INCVALUE_96MHZ;
3537 			shift = INCVALUE_SHIFT_96MHZ;
3538 			adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3539 		} else {
3540 			/* Stable 25MHz frequency */
3541 			incperiod = INCPERIOD_25MHZ;
3542 			incvalue = INCVALUE_25MHZ;
3543 			shift = INCVALUE_SHIFT_25MHZ;
3544 			adapter->cc.shift = shift;
3545 		}
3546 		break;
3547 	case e1000_pch_spt:
3548 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3549 			/* Stable 24MHz frequency */
3550 			incperiod = INCPERIOD_24MHZ;
3551 			incvalue = INCVALUE_24MHZ;
3552 			shift = INCVALUE_SHIFT_24MHZ;
3553 			adapter->cc.shift = shift;
3554 			break;
3555 		}
3556 		return -EINVAL;
3557 	case e1000_pch_cnp:
3558 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3559 			/* Stable 24MHz frequency */
3560 			incperiod = INCPERIOD_24MHZ;
3561 			incvalue = INCVALUE_24MHZ;
3562 			shift = INCVALUE_SHIFT_24MHZ;
3563 			adapter->cc.shift = shift;
3564 		} else {
3565 			/* Stable 38400KHz frequency */
3566 			incperiod = INCPERIOD_38400KHZ;
3567 			incvalue = INCVALUE_38400KHZ;
3568 			shift = INCVALUE_SHIFT_38400KHZ;
3569 			adapter->cc.shift = shift;
3570 		}
3571 		break;
3572 	case e1000_82574:
3573 	case e1000_82583:
3574 		/* Stable 25MHz frequency */
3575 		incperiod = INCPERIOD_25MHZ;
3576 		incvalue = INCVALUE_25MHZ;
3577 		shift = INCVALUE_SHIFT_25MHZ;
3578 		adapter->cc.shift = shift;
3579 		break;
3580 	default:
3581 		return -EINVAL;
3582 	}
3583 
3584 	*timinca = ((incperiod << E1000_TIMINCA_INCPERIOD_SHIFT) |
3585 		    ((incvalue << shift) & E1000_TIMINCA_INCVALUE_MASK));
3586 
3587 	return 0;
3588 }
3589 
3590 /**
3591  * e1000e_config_hwtstamp - configure the hwtstamp registers and enable/disable
3592  * @adapter: board private structure
3593  *
3594  * Outgoing time stamping can be enabled and disabled. Play nice and
3595  * disable it when requested, although it shouldn't cause any overhead
3596  * when no packet needs it. At most one packet in the queue may be
3597  * marked for time stamping, otherwise it would be impossible to tell
3598  * for sure to which packet the hardware time stamp belongs.
3599  *
3600  * Incoming time stamping has to be configured via the hardware filters.
3601  * Not all combinations are supported, in particular event type has to be
3602  * specified. Matching the kind of event packet is not supported, with the
3603  * exception of "all V2 events regardless of level 2 or 4".
3604  **/
3605 static int e1000e_config_hwtstamp(struct e1000_adapter *adapter,
3606 				  struct hwtstamp_config *config)
3607 {
3608 	struct e1000_hw *hw = &adapter->hw;
3609 	u32 tsync_tx_ctl = E1000_TSYNCTXCTL_ENABLED;
3610 	u32 tsync_rx_ctl = E1000_TSYNCRXCTL_ENABLED;
3611 	u32 rxmtrl = 0;
3612 	u16 rxudp = 0;
3613 	bool is_l4 = false;
3614 	bool is_l2 = false;
3615 	u32 regval;
3616 
3617 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3618 		return -EINVAL;
3619 
3620 	/* flags reserved for future extensions - must be zero */
3621 	if (config->flags)
3622 		return -EINVAL;
3623 
3624 	switch (config->tx_type) {
3625 	case HWTSTAMP_TX_OFF:
3626 		tsync_tx_ctl = 0;
3627 		break;
3628 	case HWTSTAMP_TX_ON:
3629 		break;
3630 	default:
3631 		return -ERANGE;
3632 	}
3633 
3634 	switch (config->rx_filter) {
3635 	case HWTSTAMP_FILTER_NONE:
3636 		tsync_rx_ctl = 0;
3637 		break;
3638 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3639 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3640 		rxmtrl = E1000_RXMTRL_PTP_V1_SYNC_MESSAGE;
3641 		is_l4 = true;
3642 		break;
3643 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3644 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3645 		rxmtrl = E1000_RXMTRL_PTP_V1_DELAY_REQ_MESSAGE;
3646 		is_l4 = true;
3647 		break;
3648 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3649 		/* Also time stamps V2 L2 Path Delay Request/Response */
3650 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3651 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3652 		is_l2 = true;
3653 		break;
3654 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3655 		/* Also time stamps V2 L2 Path Delay Request/Response. */
3656 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3657 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3658 		is_l2 = true;
3659 		break;
3660 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3661 		/* Hardware cannot filter just V2 L4 Sync messages;
3662 		 * fall-through to V2 (both L2 and L4) Sync.
3663 		 */
3664 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
3665 		/* Also time stamps V2 Path Delay Request/Response. */
3666 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3667 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3668 		is_l2 = true;
3669 		is_l4 = true;
3670 		break;
3671 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3672 		/* Hardware cannot filter just V2 L4 Delay Request messages;
3673 		 * fall-through to V2 (both L2 and L4) Delay Request.
3674 		 */
3675 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
3676 		/* Also time stamps V2 Path Delay Request/Response. */
3677 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3678 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3679 		is_l2 = true;
3680 		is_l4 = true;
3681 		break;
3682 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
3683 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
3684 		/* Hardware cannot filter just V2 L4 or L2 Event messages;
3685 		 * fall-through to all V2 (both L2 and L4) Events.
3686 		 */
3687 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
3688 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_EVENT_V2;
3689 		config->rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
3690 		is_l2 = true;
3691 		is_l4 = true;
3692 		break;
3693 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3694 		/* For V1, the hardware can only filter Sync messages or
3695 		 * Delay Request messages but not both so fall-through to
3696 		 * time stamp all packets.
3697 		 */
3698 	case HWTSTAMP_FILTER_NTP_ALL:
3699 	case HWTSTAMP_FILTER_ALL:
3700 		is_l2 = true;
3701 		is_l4 = true;
3702 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_ALL;
3703 		config->rx_filter = HWTSTAMP_FILTER_ALL;
3704 		break;
3705 	default:
3706 		return -ERANGE;
3707 	}
3708 
3709 	adapter->hwtstamp_config = *config;
3710 
3711 	/* enable/disable Tx h/w time stamping */
3712 	regval = er32(TSYNCTXCTL);
3713 	regval &= ~E1000_TSYNCTXCTL_ENABLED;
3714 	regval |= tsync_tx_ctl;
3715 	ew32(TSYNCTXCTL, regval);
3716 	if ((er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) !=
3717 	    (regval & E1000_TSYNCTXCTL_ENABLED)) {
3718 		e_err("Timesync Tx Control register not set as expected\n");
3719 		return -EAGAIN;
3720 	}
3721 
3722 	/* enable/disable Rx h/w time stamping */
3723 	regval = er32(TSYNCRXCTL);
3724 	regval &= ~(E1000_TSYNCRXCTL_ENABLED | E1000_TSYNCRXCTL_TYPE_MASK);
3725 	regval |= tsync_rx_ctl;
3726 	ew32(TSYNCRXCTL, regval);
3727 	if ((er32(TSYNCRXCTL) & (E1000_TSYNCRXCTL_ENABLED |
3728 				 E1000_TSYNCRXCTL_TYPE_MASK)) !=
3729 	    (regval & (E1000_TSYNCRXCTL_ENABLED |
3730 		       E1000_TSYNCRXCTL_TYPE_MASK))) {
3731 		e_err("Timesync Rx Control register not set as expected\n");
3732 		return -EAGAIN;
3733 	}
3734 
3735 	/* L2: define ethertype filter for time stamped packets */
3736 	if (is_l2)
3737 		rxmtrl |= ETH_P_1588;
3738 
3739 	/* define which PTP packets get time stamped */
3740 	ew32(RXMTRL, rxmtrl);
3741 
3742 	/* Filter by destination port */
3743 	if (is_l4) {
3744 		rxudp = PTP_EV_PORT;
3745 		cpu_to_be16s(&rxudp);
3746 	}
3747 	ew32(RXUDP, rxudp);
3748 
3749 	e1e_flush();
3750 
3751 	/* Clear TSYNCRXCTL_VALID & TSYNCTXCTL_VALID bit */
3752 	er32(RXSTMPH);
3753 	er32(TXSTMPH);
3754 
3755 	return 0;
3756 }
3757 
3758 /**
3759  * e1000_configure - configure the hardware for Rx and Tx
3760  * @adapter: private board structure
3761  **/
3762 static void e1000_configure(struct e1000_adapter *adapter)
3763 {
3764 	struct e1000_ring *rx_ring = adapter->rx_ring;
3765 
3766 	e1000e_set_rx_mode(adapter->netdev);
3767 
3768 	e1000_restore_vlan(adapter);
3769 	e1000_init_manageability_pt(adapter);
3770 
3771 	e1000_configure_tx(adapter);
3772 
3773 	if (adapter->netdev->features & NETIF_F_RXHASH)
3774 		e1000e_setup_rss_hash(adapter);
3775 	e1000_setup_rctl(adapter);
3776 	e1000_configure_rx(adapter);
3777 	adapter->alloc_rx_buf(rx_ring, e1000_desc_unused(rx_ring), GFP_KERNEL);
3778 }
3779 
3780 /**
3781  * e1000e_power_up_phy - restore link in case the phy was powered down
3782  * @adapter: address of board private structure
3783  *
3784  * The phy may be powered down to save power and turn off link when the
3785  * driver is unloaded and wake on lan is not enabled (among others)
3786  * *** this routine MUST be followed by a call to e1000e_reset ***
3787  **/
3788 void e1000e_power_up_phy(struct e1000_adapter *adapter)
3789 {
3790 	if (adapter->hw.phy.ops.power_up)
3791 		adapter->hw.phy.ops.power_up(&adapter->hw);
3792 
3793 	adapter->hw.mac.ops.setup_link(&adapter->hw);
3794 }
3795 
3796 /**
3797  * e1000_power_down_phy - Power down the PHY
3798  *
3799  * Power down the PHY so no link is implied when interface is down.
3800  * The PHY cannot be powered down if management or WoL is active.
3801  */
3802 static void e1000_power_down_phy(struct e1000_adapter *adapter)
3803 {
3804 	if (adapter->hw.phy.ops.power_down)
3805 		adapter->hw.phy.ops.power_down(&adapter->hw);
3806 }
3807 
3808 /**
3809  * e1000_flush_tx_ring - remove all descriptors from the tx_ring
3810  *
3811  * We want to clear all pending descriptors from the TX ring.
3812  * zeroing happens when the HW reads the regs. We  assign the ring itself as
3813  * the data of the next descriptor. We don't care about the data we are about
3814  * to reset the HW.
3815  */
3816 static void e1000_flush_tx_ring(struct e1000_adapter *adapter)
3817 {
3818 	struct e1000_hw *hw = &adapter->hw;
3819 	struct e1000_ring *tx_ring = adapter->tx_ring;
3820 	struct e1000_tx_desc *tx_desc = NULL;
3821 	u32 tdt, tctl, txd_lower = E1000_TXD_CMD_IFCS;
3822 	u16 size = 512;
3823 
3824 	tctl = er32(TCTL);
3825 	ew32(TCTL, tctl | E1000_TCTL_EN);
3826 	tdt = er32(TDT(0));
3827 	BUG_ON(tdt != tx_ring->next_to_use);
3828 	tx_desc =  E1000_TX_DESC(*tx_ring, tx_ring->next_to_use);
3829 	tx_desc->buffer_addr = tx_ring->dma;
3830 
3831 	tx_desc->lower.data = cpu_to_le32(txd_lower | size);
3832 	tx_desc->upper.data = 0;
3833 	/* flush descriptors to memory before notifying the HW */
3834 	wmb();
3835 	tx_ring->next_to_use++;
3836 	if (tx_ring->next_to_use == tx_ring->count)
3837 		tx_ring->next_to_use = 0;
3838 	ew32(TDT(0), tx_ring->next_to_use);
3839 	mmiowb();
3840 	usleep_range(200, 250);
3841 }
3842 
3843 /**
3844  * e1000_flush_rx_ring - remove all descriptors from the rx_ring
3845  *
3846  * Mark all descriptors in the RX ring as consumed and disable the rx ring
3847  */
3848 static void e1000_flush_rx_ring(struct e1000_adapter *adapter)
3849 {
3850 	u32 rctl, rxdctl;
3851 	struct e1000_hw *hw = &adapter->hw;
3852 
3853 	rctl = er32(RCTL);
3854 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3855 	e1e_flush();
3856 	usleep_range(100, 150);
3857 
3858 	rxdctl = er32(RXDCTL(0));
3859 	/* zero the lower 14 bits (prefetch and host thresholds) */
3860 	rxdctl &= 0xffffc000;
3861 
3862 	/* update thresholds: prefetch threshold to 31, host threshold to 1
3863 	 * and make sure the granularity is "descriptors" and not "cache lines"
3864 	 */
3865 	rxdctl |= (0x1F | BIT(8) | E1000_RXDCTL_THRESH_UNIT_DESC);
3866 
3867 	ew32(RXDCTL(0), rxdctl);
3868 	/* momentarily enable the RX ring for the changes to take effect */
3869 	ew32(RCTL, rctl | E1000_RCTL_EN);
3870 	e1e_flush();
3871 	usleep_range(100, 150);
3872 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3873 }
3874 
3875 /**
3876  * e1000_flush_desc_rings - remove all descriptors from the descriptor rings
3877  *
3878  * In i219, the descriptor rings must be emptied before resetting the HW
3879  * or before changing the device state to D3 during runtime (runtime PM).
3880  *
3881  * Failure to do this will cause the HW to enter a unit hang state which can
3882  * only be released by PCI reset on the device
3883  *
3884  */
3885 
3886 static void e1000_flush_desc_rings(struct e1000_adapter *adapter)
3887 {
3888 	u16 hang_state;
3889 	u32 fext_nvm11, tdlen;
3890 	struct e1000_hw *hw = &adapter->hw;
3891 
3892 	/* First, disable MULR fix in FEXTNVM11 */
3893 	fext_nvm11 = er32(FEXTNVM11);
3894 	fext_nvm11 |= E1000_FEXTNVM11_DISABLE_MULR_FIX;
3895 	ew32(FEXTNVM11, fext_nvm11);
3896 	/* do nothing if we're not in faulty state, or if the queue is empty */
3897 	tdlen = er32(TDLEN(0));
3898 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3899 			     &hang_state);
3900 	if (!(hang_state & FLUSH_DESC_REQUIRED) || !tdlen)
3901 		return;
3902 	e1000_flush_tx_ring(adapter);
3903 	/* recheck, maybe the fault is caused by the rx ring */
3904 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3905 			     &hang_state);
3906 	if (hang_state & FLUSH_DESC_REQUIRED)
3907 		e1000_flush_rx_ring(adapter);
3908 }
3909 
3910 /**
3911  * e1000e_systim_reset - reset the timesync registers after a hardware reset
3912  * @adapter: board private structure
3913  *
3914  * When the MAC is reset, all hardware bits for timesync will be reset to the
3915  * default values. This function will restore the settings last in place.
3916  * Since the clock SYSTIME registers are reset, we will simply restore the
3917  * cyclecounter to the kernel real clock time.
3918  **/
3919 static void e1000e_systim_reset(struct e1000_adapter *adapter)
3920 {
3921 	struct ptp_clock_info *info = &adapter->ptp_clock_info;
3922 	struct e1000_hw *hw = &adapter->hw;
3923 	unsigned long flags;
3924 	u32 timinca;
3925 	s32 ret_val;
3926 
3927 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3928 		return;
3929 
3930 	if (info->adjfreq) {
3931 		/* restore the previous ptp frequency delta */
3932 		ret_val = info->adjfreq(info, adapter->ptp_delta);
3933 	} else {
3934 		/* set the default base frequency if no adjustment possible */
3935 		ret_val = e1000e_get_base_timinca(adapter, &timinca);
3936 		if (!ret_val)
3937 			ew32(TIMINCA, timinca);
3938 	}
3939 
3940 	if (ret_val) {
3941 		dev_warn(&adapter->pdev->dev,
3942 			 "Failed to restore TIMINCA clock rate delta: %d\n",
3943 			 ret_val);
3944 		return;
3945 	}
3946 
3947 	/* reset the systim ns time counter */
3948 	spin_lock_irqsave(&adapter->systim_lock, flags);
3949 	timecounter_init(&adapter->tc, &adapter->cc,
3950 			 ktime_to_ns(ktime_get_real()));
3951 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
3952 
3953 	/* restore the previous hwtstamp configuration settings */
3954 	e1000e_config_hwtstamp(adapter, &adapter->hwtstamp_config);
3955 }
3956 
3957 /**
3958  * e1000e_reset - bring the hardware into a known good state
3959  *
3960  * This function boots the hardware and enables some settings that
3961  * require a configuration cycle of the hardware - those cannot be
3962  * set/changed during runtime. After reset the device needs to be
3963  * properly configured for Rx, Tx etc.
3964  */
3965 void e1000e_reset(struct e1000_adapter *adapter)
3966 {
3967 	struct e1000_mac_info *mac = &adapter->hw.mac;
3968 	struct e1000_fc_info *fc = &adapter->hw.fc;
3969 	struct e1000_hw *hw = &adapter->hw;
3970 	u32 tx_space, min_tx_space, min_rx_space;
3971 	u32 pba = adapter->pba;
3972 	u16 hwm;
3973 
3974 	/* reset Packet Buffer Allocation to default */
3975 	ew32(PBA, pba);
3976 
3977 	if (adapter->max_frame_size > (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN)) {
3978 		/* To maintain wire speed transmits, the Tx FIFO should be
3979 		 * large enough to accommodate two full transmit packets,
3980 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
3981 		 * the Rx FIFO should be large enough to accommodate at least
3982 		 * one full receive packet and is similarly rounded up and
3983 		 * expressed in KB.
3984 		 */
3985 		pba = er32(PBA);
3986 		/* upper 16 bits has Tx packet buffer allocation size in KB */
3987 		tx_space = pba >> 16;
3988 		/* lower 16 bits has Rx packet buffer allocation size in KB */
3989 		pba &= 0xffff;
3990 		/* the Tx fifo also stores 16 bytes of information about the Tx
3991 		 * but don't include ethernet FCS because hardware appends it
3992 		 */
3993 		min_tx_space = (adapter->max_frame_size +
3994 				sizeof(struct e1000_tx_desc) - ETH_FCS_LEN) * 2;
3995 		min_tx_space = ALIGN(min_tx_space, 1024);
3996 		min_tx_space >>= 10;
3997 		/* software strips receive CRC, so leave room for it */
3998 		min_rx_space = adapter->max_frame_size;
3999 		min_rx_space = ALIGN(min_rx_space, 1024);
4000 		min_rx_space >>= 10;
4001 
4002 		/* If current Tx allocation is less than the min Tx FIFO size,
4003 		 * and the min Tx FIFO size is less than the current Rx FIFO
4004 		 * allocation, take space away from current Rx allocation
4005 		 */
4006 		if ((tx_space < min_tx_space) &&
4007 		    ((min_tx_space - tx_space) < pba)) {
4008 			pba -= min_tx_space - tx_space;
4009 
4010 			/* if short on Rx space, Rx wins and must trump Tx
4011 			 * adjustment
4012 			 */
4013 			if (pba < min_rx_space)
4014 				pba = min_rx_space;
4015 		}
4016 
4017 		ew32(PBA, pba);
4018 	}
4019 
4020 	/* flow control settings
4021 	 *
4022 	 * The high water mark must be low enough to fit one full frame
4023 	 * (or the size used for early receive) above it in the Rx FIFO.
4024 	 * Set it to the lower of:
4025 	 * - 90% of the Rx FIFO size, and
4026 	 * - the full Rx FIFO size minus one full frame
4027 	 */
4028 	if (adapter->flags & FLAG_DISABLE_FC_PAUSE_TIME)
4029 		fc->pause_time = 0xFFFF;
4030 	else
4031 		fc->pause_time = E1000_FC_PAUSE_TIME;
4032 	fc->send_xon = true;
4033 	fc->current_mode = fc->requested_mode;
4034 
4035 	switch (hw->mac.type) {
4036 	case e1000_ich9lan:
4037 	case e1000_ich10lan:
4038 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4039 			pba = 14;
4040 			ew32(PBA, pba);
4041 			fc->high_water = 0x2800;
4042 			fc->low_water = fc->high_water - 8;
4043 			break;
4044 		}
4045 		/* fall-through */
4046 	default:
4047 		hwm = min(((pba << 10) * 9 / 10),
4048 			  ((pba << 10) - adapter->max_frame_size));
4049 
4050 		fc->high_water = hwm & E1000_FCRTH_RTH;	/* 8-byte granularity */
4051 		fc->low_water = fc->high_water - 8;
4052 		break;
4053 	case e1000_pchlan:
4054 		/* Workaround PCH LOM adapter hangs with certain network
4055 		 * loads.  If hangs persist, try disabling Tx flow control.
4056 		 */
4057 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4058 			fc->high_water = 0x3500;
4059 			fc->low_water = 0x1500;
4060 		} else {
4061 			fc->high_water = 0x5000;
4062 			fc->low_water = 0x3000;
4063 		}
4064 		fc->refresh_time = 0x1000;
4065 		break;
4066 	case e1000_pch2lan:
4067 	case e1000_pch_lpt:
4068 	case e1000_pch_spt:
4069 	case e1000_pch_cnp:
4070 		fc->refresh_time = 0x0400;
4071 
4072 		if (adapter->netdev->mtu <= ETH_DATA_LEN) {
4073 			fc->high_water = 0x05C20;
4074 			fc->low_water = 0x05048;
4075 			fc->pause_time = 0x0650;
4076 			break;
4077 		}
4078 
4079 		pba = 14;
4080 		ew32(PBA, pba);
4081 		fc->high_water = ((pba << 10) * 9 / 10) & E1000_FCRTH_RTH;
4082 		fc->low_water = ((pba << 10) * 8 / 10) & E1000_FCRTL_RTL;
4083 		break;
4084 	}
4085 
4086 	/* Alignment of Tx data is on an arbitrary byte boundary with the
4087 	 * maximum size per Tx descriptor limited only to the transmit
4088 	 * allocation of the packet buffer minus 96 bytes with an upper
4089 	 * limit of 24KB due to receive synchronization limitations.
4090 	 */
4091 	adapter->tx_fifo_limit = min_t(u32, ((er32(PBA) >> 16) << 10) - 96,
4092 				       24 << 10);
4093 
4094 	/* Disable Adaptive Interrupt Moderation if 2 full packets cannot
4095 	 * fit in receive buffer.
4096 	 */
4097 	if (adapter->itr_setting & 0x3) {
4098 		if ((adapter->max_frame_size * 2) > (pba << 10)) {
4099 			if (!(adapter->flags2 & FLAG2_DISABLE_AIM)) {
4100 				dev_info(&adapter->pdev->dev,
4101 					 "Interrupt Throttle Rate off\n");
4102 				adapter->flags2 |= FLAG2_DISABLE_AIM;
4103 				e1000e_write_itr(adapter, 0);
4104 			}
4105 		} else if (adapter->flags2 & FLAG2_DISABLE_AIM) {
4106 			dev_info(&adapter->pdev->dev,
4107 				 "Interrupt Throttle Rate on\n");
4108 			adapter->flags2 &= ~FLAG2_DISABLE_AIM;
4109 			adapter->itr = 20000;
4110 			e1000e_write_itr(adapter, adapter->itr);
4111 		}
4112 	}
4113 
4114 	if (hw->mac.type >= e1000_pch_spt)
4115 		e1000_flush_desc_rings(adapter);
4116 	/* Allow time for pending master requests to run */
4117 	mac->ops.reset_hw(hw);
4118 
4119 	/* For parts with AMT enabled, let the firmware know
4120 	 * that the network interface is in control
4121 	 */
4122 	if (adapter->flags & FLAG_HAS_AMT)
4123 		e1000e_get_hw_control(adapter);
4124 
4125 	ew32(WUC, 0);
4126 
4127 	if (mac->ops.init_hw(hw))
4128 		e_err("Hardware Error\n");
4129 
4130 	e1000_update_mng_vlan(adapter);
4131 
4132 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
4133 	ew32(VET, ETH_P_8021Q);
4134 
4135 	e1000e_reset_adaptive(hw);
4136 
4137 	/* restore systim and hwtstamp settings */
4138 	e1000e_systim_reset(adapter);
4139 
4140 	/* Set EEE advertisement as appropriate */
4141 	if (adapter->flags2 & FLAG2_HAS_EEE) {
4142 		s32 ret_val;
4143 		u16 adv_addr;
4144 
4145 		switch (hw->phy.type) {
4146 		case e1000_phy_82579:
4147 			adv_addr = I82579_EEE_ADVERTISEMENT;
4148 			break;
4149 		case e1000_phy_i217:
4150 			adv_addr = I217_EEE_ADVERTISEMENT;
4151 			break;
4152 		default:
4153 			dev_err(&adapter->pdev->dev,
4154 				"Invalid PHY type setting EEE advertisement\n");
4155 			return;
4156 		}
4157 
4158 		ret_val = hw->phy.ops.acquire(hw);
4159 		if (ret_val) {
4160 			dev_err(&adapter->pdev->dev,
4161 				"EEE advertisement - unable to acquire PHY\n");
4162 			return;
4163 		}
4164 
4165 		e1000_write_emi_reg_locked(hw, adv_addr,
4166 					   hw->dev_spec.ich8lan.eee_disable ?
4167 					   0 : adapter->eee_advert);
4168 
4169 		hw->phy.ops.release(hw);
4170 	}
4171 
4172 	if (!netif_running(adapter->netdev) &&
4173 	    !test_bit(__E1000_TESTING, &adapter->state))
4174 		e1000_power_down_phy(adapter);
4175 
4176 	e1000_get_phy_info(hw);
4177 
4178 	if ((adapter->flags & FLAG_HAS_SMART_POWER_DOWN) &&
4179 	    !(adapter->flags & FLAG_SMART_POWER_DOWN)) {
4180 		u16 phy_data = 0;
4181 		/* speed up time to link by disabling smart power down, ignore
4182 		 * the return value of this function because there is nothing
4183 		 * different we would do if it failed
4184 		 */
4185 		e1e_rphy(hw, IGP02E1000_PHY_POWER_MGMT, &phy_data);
4186 		phy_data &= ~IGP02E1000_PM_SPD;
4187 		e1e_wphy(hw, IGP02E1000_PHY_POWER_MGMT, phy_data);
4188 	}
4189 	if (hw->mac.type >= e1000_pch_spt && adapter->int_mode == 0) {
4190 		u32 reg;
4191 
4192 		/* Fextnvm7 @ 0xe4[2] = 1 */
4193 		reg = er32(FEXTNVM7);
4194 		reg |= E1000_FEXTNVM7_SIDE_CLK_UNGATE;
4195 		ew32(FEXTNVM7, reg);
4196 		/* Fextnvm9 @ 0x5bb4[13:12] = 11 */
4197 		reg = er32(FEXTNVM9);
4198 		reg |= E1000_FEXTNVM9_IOSFSB_CLKGATE_DIS |
4199 		       E1000_FEXTNVM9_IOSFSB_CLKREQ_DIS;
4200 		ew32(FEXTNVM9, reg);
4201 	}
4202 
4203 }
4204 
4205 /**
4206  * e1000e_trigger_lsc - trigger an LSC interrupt
4207  * @adapter:
4208  *
4209  * Fire a link status change interrupt to start the watchdog.
4210  **/
4211 static void e1000e_trigger_lsc(struct e1000_adapter *adapter)
4212 {
4213 	struct e1000_hw *hw = &adapter->hw;
4214 
4215 	if (adapter->msix_entries)
4216 		ew32(ICS, E1000_ICS_LSC | E1000_ICS_OTHER);
4217 	else
4218 		ew32(ICS, E1000_ICS_LSC);
4219 }
4220 
4221 void e1000e_up(struct e1000_adapter *adapter)
4222 {
4223 	/* hardware has been reset, we need to reload some things */
4224 	e1000_configure(adapter);
4225 
4226 	clear_bit(__E1000_DOWN, &adapter->state);
4227 
4228 	if (adapter->msix_entries)
4229 		e1000_configure_msix(adapter);
4230 	e1000_irq_enable(adapter);
4231 
4232 	netif_start_queue(adapter->netdev);
4233 
4234 	e1000e_trigger_lsc(adapter);
4235 }
4236 
4237 static void e1000e_flush_descriptors(struct e1000_adapter *adapter)
4238 {
4239 	struct e1000_hw *hw = &adapter->hw;
4240 
4241 	if (!(adapter->flags2 & FLAG2_DMA_BURST))
4242 		return;
4243 
4244 	/* flush pending descriptor writebacks to memory */
4245 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4246 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4247 
4248 	/* execute the writes immediately */
4249 	e1e_flush();
4250 
4251 	/* due to rare timing issues, write to TIDV/RDTR again to ensure the
4252 	 * write is successful
4253 	 */
4254 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4255 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4256 
4257 	/* execute the writes immediately */
4258 	e1e_flush();
4259 }
4260 
4261 static void e1000e_update_stats(struct e1000_adapter *adapter);
4262 
4263 /**
4264  * e1000e_down - quiesce the device and optionally reset the hardware
4265  * @adapter: board private structure
4266  * @reset: boolean flag to reset the hardware or not
4267  */
4268 void e1000e_down(struct e1000_adapter *adapter, bool reset)
4269 {
4270 	struct net_device *netdev = adapter->netdev;
4271 	struct e1000_hw *hw = &adapter->hw;
4272 	u32 tctl, rctl;
4273 
4274 	/* signal that we're down so the interrupt handler does not
4275 	 * reschedule our watchdog timer
4276 	 */
4277 	set_bit(__E1000_DOWN, &adapter->state);
4278 
4279 	netif_carrier_off(netdev);
4280 
4281 	/* disable receives in the hardware */
4282 	rctl = er32(RCTL);
4283 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
4284 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
4285 	/* flush and sleep below */
4286 
4287 	netif_stop_queue(netdev);
4288 
4289 	/* disable transmits in the hardware */
4290 	tctl = er32(TCTL);
4291 	tctl &= ~E1000_TCTL_EN;
4292 	ew32(TCTL, tctl);
4293 
4294 	/* flush both disables and wait for them to finish */
4295 	e1e_flush();
4296 	usleep_range(10000, 20000);
4297 
4298 	e1000_irq_disable(adapter);
4299 
4300 	napi_synchronize(&adapter->napi);
4301 
4302 	del_timer_sync(&adapter->watchdog_timer);
4303 	del_timer_sync(&adapter->phy_info_timer);
4304 
4305 	spin_lock(&adapter->stats64_lock);
4306 	e1000e_update_stats(adapter);
4307 	spin_unlock(&adapter->stats64_lock);
4308 
4309 	e1000e_flush_descriptors(adapter);
4310 
4311 	adapter->link_speed = 0;
4312 	adapter->link_duplex = 0;
4313 
4314 	/* Disable Si errata workaround on PCHx for jumbo frame flow */
4315 	if ((hw->mac.type >= e1000_pch2lan) &&
4316 	    (adapter->netdev->mtu > ETH_DATA_LEN) &&
4317 	    e1000_lv_jumbo_workaround_ich8lan(hw, false))
4318 		e_dbg("failed to disable jumbo frame workaround mode\n");
4319 
4320 	if (!pci_channel_offline(adapter->pdev)) {
4321 		if (reset)
4322 			e1000e_reset(adapter);
4323 		else if (hw->mac.type >= e1000_pch_spt)
4324 			e1000_flush_desc_rings(adapter);
4325 	}
4326 	e1000_clean_tx_ring(adapter->tx_ring);
4327 	e1000_clean_rx_ring(adapter->rx_ring);
4328 }
4329 
4330 void e1000e_reinit_locked(struct e1000_adapter *adapter)
4331 {
4332 	might_sleep();
4333 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
4334 		usleep_range(1000, 2000);
4335 	e1000e_down(adapter, true);
4336 	e1000e_up(adapter);
4337 	clear_bit(__E1000_RESETTING, &adapter->state);
4338 }
4339 
4340 /**
4341  * e1000e_sanitize_systim - sanitize raw cycle counter reads
4342  * @hw: pointer to the HW structure
4343  * @systim: time value read, sanitized and returned
4344  *
4345  * Errata for 82574/82583 possible bad bits read from SYSTIMH/L:
4346  * check to see that the time is incrementing at a reasonable
4347  * rate and is a multiple of incvalue.
4348  **/
4349 static u64 e1000e_sanitize_systim(struct e1000_hw *hw, u64 systim)
4350 {
4351 	u64 time_delta, rem, temp;
4352 	u64 systim_next;
4353 	u32 incvalue;
4354 	int i;
4355 
4356 	incvalue = er32(TIMINCA) & E1000_TIMINCA_INCVALUE_MASK;
4357 	for (i = 0; i < E1000_MAX_82574_SYSTIM_REREADS; i++) {
4358 		/* latch SYSTIMH on read of SYSTIML */
4359 		systim_next = (u64)er32(SYSTIML);
4360 		systim_next |= (u64)er32(SYSTIMH) << 32;
4361 
4362 		time_delta = systim_next - systim;
4363 		temp = time_delta;
4364 		/* VMWare users have seen incvalue of zero, don't div / 0 */
4365 		rem = incvalue ? do_div(temp, incvalue) : (time_delta != 0);
4366 
4367 		systim = systim_next;
4368 
4369 		if ((time_delta < E1000_82574_SYSTIM_EPSILON) && (rem == 0))
4370 			break;
4371 	}
4372 
4373 	return systim;
4374 }
4375 
4376 /**
4377  * e1000e_cyclecounter_read - read raw cycle counter (used by time counter)
4378  * @cc: cyclecounter structure
4379  **/
4380 static u64 e1000e_cyclecounter_read(const struct cyclecounter *cc)
4381 {
4382 	struct e1000_adapter *adapter = container_of(cc, struct e1000_adapter,
4383 						     cc);
4384 	struct e1000_hw *hw = &adapter->hw;
4385 	u32 systimel, systimeh;
4386 	u64 systim;
4387 	/* SYSTIMH latching upon SYSTIML read does not work well.
4388 	 * This means that if SYSTIML overflows after we read it but before
4389 	 * we read SYSTIMH, the value of SYSTIMH has been incremented and we
4390 	 * will experience a huge non linear increment in the systime value
4391 	 * to fix that we test for overflow and if true, we re-read systime.
4392 	 */
4393 	systimel = er32(SYSTIML);
4394 	systimeh = er32(SYSTIMH);
4395 	/* Is systimel is so large that overflow is possible? */
4396 	if (systimel >= (u32)0xffffffff - E1000_TIMINCA_INCVALUE_MASK) {
4397 		u32 systimel_2 = er32(SYSTIML);
4398 		if (systimel > systimel_2) {
4399 			/* There was an overflow, read again SYSTIMH, and use
4400 			 * systimel_2
4401 			 */
4402 			systimeh = er32(SYSTIMH);
4403 			systimel = systimel_2;
4404 		}
4405 	}
4406 	systim = (u64)systimel;
4407 	systim |= (u64)systimeh << 32;
4408 
4409 	if (adapter->flags2 & FLAG2_CHECK_SYSTIM_OVERFLOW)
4410 		systim = e1000e_sanitize_systim(hw, systim);
4411 
4412 	return systim;
4413 }
4414 
4415 /**
4416  * e1000_sw_init - Initialize general software structures (struct e1000_adapter)
4417  * @adapter: board private structure to initialize
4418  *
4419  * e1000_sw_init initializes the Adapter private data structure.
4420  * Fields are initialized based on PCI device information and
4421  * OS network device settings (MTU size).
4422  **/
4423 static int e1000_sw_init(struct e1000_adapter *adapter)
4424 {
4425 	struct net_device *netdev = adapter->netdev;
4426 
4427 	adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
4428 	adapter->rx_ps_bsize0 = 128;
4429 	adapter->max_frame_size = netdev->mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
4430 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4431 	adapter->tx_ring_count = E1000_DEFAULT_TXD;
4432 	adapter->rx_ring_count = E1000_DEFAULT_RXD;
4433 
4434 	spin_lock_init(&adapter->stats64_lock);
4435 
4436 	e1000e_set_interrupt_capability(adapter);
4437 
4438 	if (e1000_alloc_queues(adapter))
4439 		return -ENOMEM;
4440 
4441 	/* Setup hardware time stamping cyclecounter */
4442 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
4443 		adapter->cc.read = e1000e_cyclecounter_read;
4444 		adapter->cc.mask = CYCLECOUNTER_MASK(64);
4445 		adapter->cc.mult = 1;
4446 		/* cc.shift set in e1000e_get_base_tininca() */
4447 
4448 		spin_lock_init(&adapter->systim_lock);
4449 		INIT_WORK(&adapter->tx_hwtstamp_work, e1000e_tx_hwtstamp_work);
4450 	}
4451 
4452 	/* Explicitly disable IRQ since the NIC can be in any state. */
4453 	e1000_irq_disable(adapter);
4454 
4455 	set_bit(__E1000_DOWN, &adapter->state);
4456 	return 0;
4457 }
4458 
4459 /**
4460  * e1000_intr_msi_test - Interrupt Handler
4461  * @irq: interrupt number
4462  * @data: pointer to a network interface device structure
4463  **/
4464 static irqreturn_t e1000_intr_msi_test(int __always_unused irq, void *data)
4465 {
4466 	struct net_device *netdev = data;
4467 	struct e1000_adapter *adapter = netdev_priv(netdev);
4468 	struct e1000_hw *hw = &adapter->hw;
4469 	u32 icr = er32(ICR);
4470 
4471 	e_dbg("icr is %08X\n", icr);
4472 	if (icr & E1000_ICR_RXSEQ) {
4473 		adapter->flags &= ~FLAG_MSI_TEST_FAILED;
4474 		/* Force memory writes to complete before acknowledging the
4475 		 * interrupt is handled.
4476 		 */
4477 		wmb();
4478 	}
4479 
4480 	return IRQ_HANDLED;
4481 }
4482 
4483 /**
4484  * e1000_test_msi_interrupt - Returns 0 for successful test
4485  * @adapter: board private struct
4486  *
4487  * code flow taken from tg3.c
4488  **/
4489 static int e1000_test_msi_interrupt(struct e1000_adapter *adapter)
4490 {
4491 	struct net_device *netdev = adapter->netdev;
4492 	struct e1000_hw *hw = &adapter->hw;
4493 	int err;
4494 
4495 	/* poll_enable hasn't been called yet, so don't need disable */
4496 	/* clear any pending events */
4497 	er32(ICR);
4498 
4499 	/* free the real vector and request a test handler */
4500 	e1000_free_irq(adapter);
4501 	e1000e_reset_interrupt_capability(adapter);
4502 
4503 	/* Assume that the test fails, if it succeeds then the test
4504 	 * MSI irq handler will unset this flag
4505 	 */
4506 	adapter->flags |= FLAG_MSI_TEST_FAILED;
4507 
4508 	err = pci_enable_msi(adapter->pdev);
4509 	if (err)
4510 		goto msi_test_failed;
4511 
4512 	err = request_irq(adapter->pdev->irq, e1000_intr_msi_test, 0,
4513 			  netdev->name, netdev);
4514 	if (err) {
4515 		pci_disable_msi(adapter->pdev);
4516 		goto msi_test_failed;
4517 	}
4518 
4519 	/* Force memory writes to complete before enabling and firing an
4520 	 * interrupt.
4521 	 */
4522 	wmb();
4523 
4524 	e1000_irq_enable(adapter);
4525 
4526 	/* fire an unusual interrupt on the test handler */
4527 	ew32(ICS, E1000_ICS_RXSEQ);
4528 	e1e_flush();
4529 	msleep(100);
4530 
4531 	e1000_irq_disable(adapter);
4532 
4533 	rmb();			/* read flags after interrupt has been fired */
4534 
4535 	if (adapter->flags & FLAG_MSI_TEST_FAILED) {
4536 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
4537 		e_info("MSI interrupt test failed, using legacy interrupt.\n");
4538 	} else {
4539 		e_dbg("MSI interrupt test succeeded!\n");
4540 	}
4541 
4542 	free_irq(adapter->pdev->irq, netdev);
4543 	pci_disable_msi(adapter->pdev);
4544 
4545 msi_test_failed:
4546 	e1000e_set_interrupt_capability(adapter);
4547 	return e1000_request_irq(adapter);
4548 }
4549 
4550 /**
4551  * e1000_test_msi - Returns 0 if MSI test succeeds or INTx mode is restored
4552  * @adapter: board private struct
4553  *
4554  * code flow taken from tg3.c, called with e1000 interrupts disabled.
4555  **/
4556 static int e1000_test_msi(struct e1000_adapter *adapter)
4557 {
4558 	int err;
4559 	u16 pci_cmd;
4560 
4561 	if (!(adapter->flags & FLAG_MSI_ENABLED))
4562 		return 0;
4563 
4564 	/* disable SERR in case the MSI write causes a master abort */
4565 	pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4566 	if (pci_cmd & PCI_COMMAND_SERR)
4567 		pci_write_config_word(adapter->pdev, PCI_COMMAND,
4568 				      pci_cmd & ~PCI_COMMAND_SERR);
4569 
4570 	err = e1000_test_msi_interrupt(adapter);
4571 
4572 	/* re-enable SERR */
4573 	if (pci_cmd & PCI_COMMAND_SERR) {
4574 		pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4575 		pci_cmd |= PCI_COMMAND_SERR;
4576 		pci_write_config_word(adapter->pdev, PCI_COMMAND, pci_cmd);
4577 	}
4578 
4579 	return err;
4580 }
4581 
4582 /**
4583  * e1000e_open - Called when a network interface is made active
4584  * @netdev: network interface device structure
4585  *
4586  * Returns 0 on success, negative value on failure
4587  *
4588  * The open entry point is called when a network interface is made
4589  * active by the system (IFF_UP).  At this point all resources needed
4590  * for transmit and receive operations are allocated, the interrupt
4591  * handler is registered with the OS, the watchdog timer is started,
4592  * and the stack is notified that the interface is ready.
4593  **/
4594 int e1000e_open(struct net_device *netdev)
4595 {
4596 	struct e1000_adapter *adapter = netdev_priv(netdev);
4597 	struct e1000_hw *hw = &adapter->hw;
4598 	struct pci_dev *pdev = adapter->pdev;
4599 	int err;
4600 
4601 	/* disallow open during test */
4602 	if (test_bit(__E1000_TESTING, &adapter->state))
4603 		return -EBUSY;
4604 
4605 	pm_runtime_get_sync(&pdev->dev);
4606 
4607 	netif_carrier_off(netdev);
4608 
4609 	/* allocate transmit descriptors */
4610 	err = e1000e_setup_tx_resources(adapter->tx_ring);
4611 	if (err)
4612 		goto err_setup_tx;
4613 
4614 	/* allocate receive descriptors */
4615 	err = e1000e_setup_rx_resources(adapter->rx_ring);
4616 	if (err)
4617 		goto err_setup_rx;
4618 
4619 	/* If AMT is enabled, let the firmware know that the network
4620 	 * interface is now open and reset the part to a known state.
4621 	 */
4622 	if (adapter->flags & FLAG_HAS_AMT) {
4623 		e1000e_get_hw_control(adapter);
4624 		e1000e_reset(adapter);
4625 	}
4626 
4627 	e1000e_power_up_phy(adapter);
4628 
4629 	adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
4630 	if ((adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
4631 		e1000_update_mng_vlan(adapter);
4632 
4633 	/* DMA latency requirement to workaround jumbo issue */
4634 	pm_qos_add_request(&adapter->pm_qos_req, PM_QOS_CPU_DMA_LATENCY,
4635 			   PM_QOS_DEFAULT_VALUE);
4636 
4637 	/* before we allocate an interrupt, we must be ready to handle it.
4638 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4639 	 * as soon as we call pci_request_irq, so we have to setup our
4640 	 * clean_rx handler before we do so.
4641 	 */
4642 	e1000_configure(adapter);
4643 
4644 	err = e1000_request_irq(adapter);
4645 	if (err)
4646 		goto err_req_irq;
4647 
4648 	/* Work around PCIe errata with MSI interrupts causing some chipsets to
4649 	 * ignore e1000e MSI messages, which means we need to test our MSI
4650 	 * interrupt now
4651 	 */
4652 	if (adapter->int_mode != E1000E_INT_MODE_LEGACY) {
4653 		err = e1000_test_msi(adapter);
4654 		if (err) {
4655 			e_err("Interrupt allocation failed\n");
4656 			goto err_req_irq;
4657 		}
4658 	}
4659 
4660 	/* From here on the code is the same as e1000e_up() */
4661 	clear_bit(__E1000_DOWN, &adapter->state);
4662 
4663 	napi_enable(&adapter->napi);
4664 
4665 	e1000_irq_enable(adapter);
4666 
4667 	adapter->tx_hang_recheck = false;
4668 	netif_start_queue(netdev);
4669 
4670 	hw->mac.get_link_status = true;
4671 	pm_runtime_put(&pdev->dev);
4672 
4673 	e1000e_trigger_lsc(adapter);
4674 
4675 	return 0;
4676 
4677 err_req_irq:
4678 	pm_qos_remove_request(&adapter->pm_qos_req);
4679 	e1000e_release_hw_control(adapter);
4680 	e1000_power_down_phy(adapter);
4681 	e1000e_free_rx_resources(adapter->rx_ring);
4682 err_setup_rx:
4683 	e1000e_free_tx_resources(adapter->tx_ring);
4684 err_setup_tx:
4685 	e1000e_reset(adapter);
4686 	pm_runtime_put_sync(&pdev->dev);
4687 
4688 	return err;
4689 }
4690 
4691 /**
4692  * e1000e_close - Disables a network interface
4693  * @netdev: network interface device structure
4694  *
4695  * Returns 0, this is not allowed to fail
4696  *
4697  * The close entry point is called when an interface is de-activated
4698  * by the OS.  The hardware is still under the drivers control, but
4699  * needs to be disabled.  A global MAC reset is issued to stop the
4700  * hardware, and all transmit and receive resources are freed.
4701  **/
4702 int e1000e_close(struct net_device *netdev)
4703 {
4704 	struct e1000_adapter *adapter = netdev_priv(netdev);
4705 	struct pci_dev *pdev = adapter->pdev;
4706 	int count = E1000_CHECK_RESET_COUNT;
4707 
4708 	while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
4709 		usleep_range(10000, 20000);
4710 
4711 	WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
4712 
4713 	pm_runtime_get_sync(&pdev->dev);
4714 
4715 	if (!test_bit(__E1000_DOWN, &adapter->state)) {
4716 		e1000e_down(adapter, true);
4717 		e1000_free_irq(adapter);
4718 
4719 		/* Link status message must follow this format */
4720 		pr_info("%s NIC Link is Down\n", adapter->netdev->name);
4721 	}
4722 
4723 	napi_disable(&adapter->napi);
4724 
4725 	e1000e_free_tx_resources(adapter->tx_ring);
4726 	e1000e_free_rx_resources(adapter->rx_ring);
4727 
4728 	/* kill manageability vlan ID if supported, but not if a vlan with
4729 	 * the same ID is registered on the host OS (let 8021q kill it)
4730 	 */
4731 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN)
4732 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
4733 				       adapter->mng_vlan_id);
4734 
4735 	/* If AMT is enabled, let the firmware know that the network
4736 	 * interface is now closed
4737 	 */
4738 	if ((adapter->flags & FLAG_HAS_AMT) &&
4739 	    !test_bit(__E1000_TESTING, &adapter->state))
4740 		e1000e_release_hw_control(adapter);
4741 
4742 	pm_qos_remove_request(&adapter->pm_qos_req);
4743 
4744 	pm_runtime_put_sync(&pdev->dev);
4745 
4746 	return 0;
4747 }
4748 
4749 /**
4750  * e1000_set_mac - Change the Ethernet Address of the NIC
4751  * @netdev: network interface device structure
4752  * @p: pointer to an address structure
4753  *
4754  * Returns 0 on success, negative on failure
4755  **/
4756 static int e1000_set_mac(struct net_device *netdev, void *p)
4757 {
4758 	struct e1000_adapter *adapter = netdev_priv(netdev);
4759 	struct e1000_hw *hw = &adapter->hw;
4760 	struct sockaddr *addr = p;
4761 
4762 	if (!is_valid_ether_addr(addr->sa_data))
4763 		return -EADDRNOTAVAIL;
4764 
4765 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
4766 	memcpy(adapter->hw.mac.addr, addr->sa_data, netdev->addr_len);
4767 
4768 	hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
4769 
4770 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA) {
4771 		/* activate the work around */
4772 		e1000e_set_laa_state_82571(&adapter->hw, 1);
4773 
4774 		/* Hold a copy of the LAA in RAR[14] This is done so that
4775 		 * between the time RAR[0] gets clobbered  and the time it
4776 		 * gets fixed (in e1000_watchdog), the actual LAA is in one
4777 		 * of the RARs and no incoming packets directed to this port
4778 		 * are dropped. Eventually the LAA will be in RAR[0] and
4779 		 * RAR[14]
4780 		 */
4781 		hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr,
4782 				    adapter->hw.mac.rar_entry_count - 1);
4783 	}
4784 
4785 	return 0;
4786 }
4787 
4788 /**
4789  * e1000e_update_phy_task - work thread to update phy
4790  * @work: pointer to our work struct
4791  *
4792  * this worker thread exists because we must acquire a
4793  * semaphore to read the phy, which we could msleep while
4794  * waiting for it, and we can't msleep in a timer.
4795  **/
4796 static void e1000e_update_phy_task(struct work_struct *work)
4797 {
4798 	struct e1000_adapter *adapter = container_of(work,
4799 						     struct e1000_adapter,
4800 						     update_phy_task);
4801 	struct e1000_hw *hw = &adapter->hw;
4802 
4803 	if (test_bit(__E1000_DOWN, &adapter->state))
4804 		return;
4805 
4806 	e1000_get_phy_info(hw);
4807 
4808 	/* Enable EEE on 82579 after link up */
4809 	if (hw->phy.type >= e1000_phy_82579)
4810 		e1000_set_eee_pchlan(hw);
4811 }
4812 
4813 /**
4814  * e1000_update_phy_info - timre call-back to update PHY info
4815  * @data: pointer to adapter cast into an unsigned long
4816  *
4817  * Need to wait a few seconds after link up to get diagnostic information from
4818  * the phy
4819  **/
4820 static void e1000_update_phy_info(struct timer_list *t)
4821 {
4822 	struct e1000_adapter *adapter = from_timer(adapter, t, phy_info_timer);
4823 
4824 	if (test_bit(__E1000_DOWN, &adapter->state))
4825 		return;
4826 
4827 	schedule_work(&adapter->update_phy_task);
4828 }
4829 
4830 /**
4831  * e1000e_update_phy_stats - Update the PHY statistics counters
4832  * @adapter: board private structure
4833  *
4834  * Read/clear the upper 16-bit PHY registers and read/accumulate lower
4835  **/
4836 static void e1000e_update_phy_stats(struct e1000_adapter *adapter)
4837 {
4838 	struct e1000_hw *hw = &adapter->hw;
4839 	s32 ret_val;
4840 	u16 phy_data;
4841 
4842 	ret_val = hw->phy.ops.acquire(hw);
4843 	if (ret_val)
4844 		return;
4845 
4846 	/* A page set is expensive so check if already on desired page.
4847 	 * If not, set to the page with the PHY status registers.
4848 	 */
4849 	hw->phy.addr = 1;
4850 	ret_val = e1000e_read_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4851 					   &phy_data);
4852 	if (ret_val)
4853 		goto release;
4854 	if (phy_data != (HV_STATS_PAGE << IGP_PAGE_SHIFT)) {
4855 		ret_val = hw->phy.ops.set_page(hw,
4856 					       HV_STATS_PAGE << IGP_PAGE_SHIFT);
4857 		if (ret_val)
4858 			goto release;
4859 	}
4860 
4861 	/* Single Collision Count */
4862 	hw->phy.ops.read_reg_page(hw, HV_SCC_UPPER, &phy_data);
4863 	ret_val = hw->phy.ops.read_reg_page(hw, HV_SCC_LOWER, &phy_data);
4864 	if (!ret_val)
4865 		adapter->stats.scc += phy_data;
4866 
4867 	/* Excessive Collision Count */
4868 	hw->phy.ops.read_reg_page(hw, HV_ECOL_UPPER, &phy_data);
4869 	ret_val = hw->phy.ops.read_reg_page(hw, HV_ECOL_LOWER, &phy_data);
4870 	if (!ret_val)
4871 		adapter->stats.ecol += phy_data;
4872 
4873 	/* Multiple Collision Count */
4874 	hw->phy.ops.read_reg_page(hw, HV_MCC_UPPER, &phy_data);
4875 	ret_val = hw->phy.ops.read_reg_page(hw, HV_MCC_LOWER, &phy_data);
4876 	if (!ret_val)
4877 		adapter->stats.mcc += phy_data;
4878 
4879 	/* Late Collision Count */
4880 	hw->phy.ops.read_reg_page(hw, HV_LATECOL_UPPER, &phy_data);
4881 	ret_val = hw->phy.ops.read_reg_page(hw, HV_LATECOL_LOWER, &phy_data);
4882 	if (!ret_val)
4883 		adapter->stats.latecol += phy_data;
4884 
4885 	/* Collision Count - also used for adaptive IFS */
4886 	hw->phy.ops.read_reg_page(hw, HV_COLC_UPPER, &phy_data);
4887 	ret_val = hw->phy.ops.read_reg_page(hw, HV_COLC_LOWER, &phy_data);
4888 	if (!ret_val)
4889 		hw->mac.collision_delta = phy_data;
4890 
4891 	/* Defer Count */
4892 	hw->phy.ops.read_reg_page(hw, HV_DC_UPPER, &phy_data);
4893 	ret_val = hw->phy.ops.read_reg_page(hw, HV_DC_LOWER, &phy_data);
4894 	if (!ret_val)
4895 		adapter->stats.dc += phy_data;
4896 
4897 	/* Transmit with no CRS */
4898 	hw->phy.ops.read_reg_page(hw, HV_TNCRS_UPPER, &phy_data);
4899 	ret_val = hw->phy.ops.read_reg_page(hw, HV_TNCRS_LOWER, &phy_data);
4900 	if (!ret_val)
4901 		adapter->stats.tncrs += phy_data;
4902 
4903 release:
4904 	hw->phy.ops.release(hw);
4905 }
4906 
4907 /**
4908  * e1000e_update_stats - Update the board statistics counters
4909  * @adapter: board private structure
4910  **/
4911 static void e1000e_update_stats(struct e1000_adapter *adapter)
4912 {
4913 	struct net_device *netdev = adapter->netdev;
4914 	struct e1000_hw *hw = &adapter->hw;
4915 	struct pci_dev *pdev = adapter->pdev;
4916 
4917 	/* Prevent stats update while adapter is being reset, or if the pci
4918 	 * connection is down.
4919 	 */
4920 	if (adapter->link_speed == 0)
4921 		return;
4922 	if (pci_channel_offline(pdev))
4923 		return;
4924 
4925 	adapter->stats.crcerrs += er32(CRCERRS);
4926 	adapter->stats.gprc += er32(GPRC);
4927 	adapter->stats.gorc += er32(GORCL);
4928 	er32(GORCH);		/* Clear gorc */
4929 	adapter->stats.bprc += er32(BPRC);
4930 	adapter->stats.mprc += er32(MPRC);
4931 	adapter->stats.roc += er32(ROC);
4932 
4933 	adapter->stats.mpc += er32(MPC);
4934 
4935 	/* Half-duplex statistics */
4936 	if (adapter->link_duplex == HALF_DUPLEX) {
4937 		if (adapter->flags2 & FLAG2_HAS_PHY_STATS) {
4938 			e1000e_update_phy_stats(adapter);
4939 		} else {
4940 			adapter->stats.scc += er32(SCC);
4941 			adapter->stats.ecol += er32(ECOL);
4942 			adapter->stats.mcc += er32(MCC);
4943 			adapter->stats.latecol += er32(LATECOL);
4944 			adapter->stats.dc += er32(DC);
4945 
4946 			hw->mac.collision_delta = er32(COLC);
4947 
4948 			if ((hw->mac.type != e1000_82574) &&
4949 			    (hw->mac.type != e1000_82583))
4950 				adapter->stats.tncrs += er32(TNCRS);
4951 		}
4952 		adapter->stats.colc += hw->mac.collision_delta;
4953 	}
4954 
4955 	adapter->stats.xonrxc += er32(XONRXC);
4956 	adapter->stats.xontxc += er32(XONTXC);
4957 	adapter->stats.xoffrxc += er32(XOFFRXC);
4958 	adapter->stats.xofftxc += er32(XOFFTXC);
4959 	adapter->stats.gptc += er32(GPTC);
4960 	adapter->stats.gotc += er32(GOTCL);
4961 	er32(GOTCH);		/* Clear gotc */
4962 	adapter->stats.rnbc += er32(RNBC);
4963 	adapter->stats.ruc += er32(RUC);
4964 
4965 	adapter->stats.mptc += er32(MPTC);
4966 	adapter->stats.bptc += er32(BPTC);
4967 
4968 	/* used for adaptive IFS */
4969 
4970 	hw->mac.tx_packet_delta = er32(TPT);
4971 	adapter->stats.tpt += hw->mac.tx_packet_delta;
4972 
4973 	adapter->stats.algnerrc += er32(ALGNERRC);
4974 	adapter->stats.rxerrc += er32(RXERRC);
4975 	adapter->stats.cexterr += er32(CEXTERR);
4976 	adapter->stats.tsctc += er32(TSCTC);
4977 	adapter->stats.tsctfc += er32(TSCTFC);
4978 
4979 	/* Fill out the OS statistics structure */
4980 	netdev->stats.multicast = adapter->stats.mprc;
4981 	netdev->stats.collisions = adapter->stats.colc;
4982 
4983 	/* Rx Errors */
4984 
4985 	/* RLEC on some newer hardware can be incorrect so build
4986 	 * our own version based on RUC and ROC
4987 	 */
4988 	netdev->stats.rx_errors = adapter->stats.rxerrc +
4989 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
4990 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
4991 	netdev->stats.rx_length_errors = adapter->stats.ruc +
4992 	    adapter->stats.roc;
4993 	netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
4994 	netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
4995 	netdev->stats.rx_missed_errors = adapter->stats.mpc;
4996 
4997 	/* Tx Errors */
4998 	netdev->stats.tx_errors = adapter->stats.ecol + adapter->stats.latecol;
4999 	netdev->stats.tx_aborted_errors = adapter->stats.ecol;
5000 	netdev->stats.tx_window_errors = adapter->stats.latecol;
5001 	netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
5002 
5003 	/* Tx Dropped needs to be maintained elsewhere */
5004 
5005 	/* Management Stats */
5006 	adapter->stats.mgptc += er32(MGTPTC);
5007 	adapter->stats.mgprc += er32(MGTPRC);
5008 	adapter->stats.mgpdc += er32(MGTPDC);
5009 
5010 	/* Correctable ECC Errors */
5011 	if (hw->mac.type >= e1000_pch_lpt) {
5012 		u32 pbeccsts = er32(PBECCSTS);
5013 
5014 		adapter->corr_errors +=
5015 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
5016 		adapter->uncorr_errors +=
5017 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
5018 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
5019 	}
5020 }
5021 
5022 /**
5023  * e1000_phy_read_status - Update the PHY register status snapshot
5024  * @adapter: board private structure
5025  **/
5026 static void e1000_phy_read_status(struct e1000_adapter *adapter)
5027 {
5028 	struct e1000_hw *hw = &adapter->hw;
5029 	struct e1000_phy_regs *phy = &adapter->phy_regs;
5030 
5031 	if (!pm_runtime_suspended((&adapter->pdev->dev)->parent) &&
5032 	    (er32(STATUS) & E1000_STATUS_LU) &&
5033 	    (adapter->hw.phy.media_type == e1000_media_type_copper)) {
5034 		int ret_val;
5035 
5036 		ret_val = e1e_rphy(hw, MII_BMCR, &phy->bmcr);
5037 		ret_val |= e1e_rphy(hw, MII_BMSR, &phy->bmsr);
5038 		ret_val |= e1e_rphy(hw, MII_ADVERTISE, &phy->advertise);
5039 		ret_val |= e1e_rphy(hw, MII_LPA, &phy->lpa);
5040 		ret_val |= e1e_rphy(hw, MII_EXPANSION, &phy->expansion);
5041 		ret_val |= e1e_rphy(hw, MII_CTRL1000, &phy->ctrl1000);
5042 		ret_val |= e1e_rphy(hw, MII_STAT1000, &phy->stat1000);
5043 		ret_val |= e1e_rphy(hw, MII_ESTATUS, &phy->estatus);
5044 		if (ret_val)
5045 			e_warn("Error reading PHY register\n");
5046 	} else {
5047 		/* Do not read PHY registers if link is not up
5048 		 * Set values to typical power-on defaults
5049 		 */
5050 		phy->bmcr = (BMCR_SPEED1000 | BMCR_ANENABLE | BMCR_FULLDPLX);
5051 		phy->bmsr = (BMSR_100FULL | BMSR_100HALF | BMSR_10FULL |
5052 			     BMSR_10HALF | BMSR_ESTATEN | BMSR_ANEGCAPABLE |
5053 			     BMSR_ERCAP);
5054 		phy->advertise = (ADVERTISE_PAUSE_ASYM | ADVERTISE_PAUSE_CAP |
5055 				  ADVERTISE_ALL | ADVERTISE_CSMA);
5056 		phy->lpa = 0;
5057 		phy->expansion = EXPANSION_ENABLENPAGE;
5058 		phy->ctrl1000 = ADVERTISE_1000FULL;
5059 		phy->stat1000 = 0;
5060 		phy->estatus = (ESTATUS_1000_TFULL | ESTATUS_1000_THALF);
5061 	}
5062 }
5063 
5064 static void e1000_print_link_info(struct e1000_adapter *adapter)
5065 {
5066 	struct e1000_hw *hw = &adapter->hw;
5067 	u32 ctrl = er32(CTRL);
5068 
5069 	/* Link status message must follow this format for user tools */
5070 	pr_info("%s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5071 		adapter->netdev->name, adapter->link_speed,
5072 		adapter->link_duplex == FULL_DUPLEX ? "Full" : "Half",
5073 		(ctrl & E1000_CTRL_TFCE) && (ctrl & E1000_CTRL_RFCE) ? "Rx/Tx" :
5074 		(ctrl & E1000_CTRL_RFCE) ? "Rx" :
5075 		(ctrl & E1000_CTRL_TFCE) ? "Tx" : "None");
5076 }
5077 
5078 static bool e1000e_has_link(struct e1000_adapter *adapter)
5079 {
5080 	struct e1000_hw *hw = &adapter->hw;
5081 	bool link_active = false;
5082 	s32 ret_val = 0;
5083 
5084 	/* get_link_status is set on LSC (link status) interrupt or
5085 	 * Rx sequence error interrupt.  get_link_status will stay
5086 	 * true until the check_for_link establishes link
5087 	 * for copper adapters ONLY
5088 	 */
5089 	switch (hw->phy.media_type) {
5090 	case e1000_media_type_copper:
5091 		if (hw->mac.get_link_status) {
5092 			ret_val = hw->mac.ops.check_for_link(hw);
5093 			link_active = !hw->mac.get_link_status;
5094 		} else {
5095 			link_active = true;
5096 		}
5097 		break;
5098 	case e1000_media_type_fiber:
5099 		ret_val = hw->mac.ops.check_for_link(hw);
5100 		link_active = !!(er32(STATUS) & E1000_STATUS_LU);
5101 		break;
5102 	case e1000_media_type_internal_serdes:
5103 		ret_val = hw->mac.ops.check_for_link(hw);
5104 		link_active = hw->mac.serdes_has_link;
5105 		break;
5106 	default:
5107 	case e1000_media_type_unknown:
5108 		break;
5109 	}
5110 
5111 	if ((ret_val == -E1000_ERR_PHY) && (hw->phy.type == e1000_phy_igp_3) &&
5112 	    (er32(CTRL) & E1000_PHY_CTRL_GBE_DISABLE)) {
5113 		/* See e1000_kmrn_lock_loss_workaround_ich8lan() */
5114 		e_info("Gigabit has been disabled, downgrading speed\n");
5115 	}
5116 
5117 	return link_active;
5118 }
5119 
5120 static void e1000e_enable_receives(struct e1000_adapter *adapter)
5121 {
5122 	/* make sure the receive unit is started */
5123 	if ((adapter->flags & FLAG_RX_NEEDS_RESTART) &&
5124 	    (adapter->flags & FLAG_RESTART_NOW)) {
5125 		struct e1000_hw *hw = &adapter->hw;
5126 		u32 rctl = er32(RCTL);
5127 
5128 		ew32(RCTL, rctl | E1000_RCTL_EN);
5129 		adapter->flags &= ~FLAG_RESTART_NOW;
5130 	}
5131 }
5132 
5133 static void e1000e_check_82574_phy_workaround(struct e1000_adapter *adapter)
5134 {
5135 	struct e1000_hw *hw = &adapter->hw;
5136 
5137 	/* With 82574 controllers, PHY needs to be checked periodically
5138 	 * for hung state and reset, if two calls return true
5139 	 */
5140 	if (e1000_check_phy_82574(hw))
5141 		adapter->phy_hang_count++;
5142 	else
5143 		adapter->phy_hang_count = 0;
5144 
5145 	if (adapter->phy_hang_count > 1) {
5146 		adapter->phy_hang_count = 0;
5147 		e_dbg("PHY appears hung - resetting\n");
5148 		schedule_work(&adapter->reset_task);
5149 	}
5150 }
5151 
5152 /**
5153  * e1000_watchdog - Timer Call-back
5154  * @data: pointer to adapter cast into an unsigned long
5155  **/
5156 static void e1000_watchdog(struct timer_list *t)
5157 {
5158 	struct e1000_adapter *adapter = from_timer(adapter, t, watchdog_timer);
5159 
5160 	/* Do the rest outside of interrupt context */
5161 	schedule_work(&adapter->watchdog_task);
5162 
5163 	/* TODO: make this use queue_delayed_work() */
5164 }
5165 
5166 static void e1000_watchdog_task(struct work_struct *work)
5167 {
5168 	struct e1000_adapter *adapter = container_of(work,
5169 						     struct e1000_adapter,
5170 						     watchdog_task);
5171 	struct net_device *netdev = adapter->netdev;
5172 	struct e1000_mac_info *mac = &adapter->hw.mac;
5173 	struct e1000_phy_info *phy = &adapter->hw.phy;
5174 	struct e1000_ring *tx_ring = adapter->tx_ring;
5175 	struct e1000_hw *hw = &adapter->hw;
5176 	u32 link, tctl;
5177 
5178 	if (test_bit(__E1000_DOWN, &adapter->state))
5179 		return;
5180 
5181 	link = e1000e_has_link(adapter);
5182 	if ((netif_carrier_ok(netdev)) && link) {
5183 		/* Cancel scheduled suspend requests. */
5184 		pm_runtime_resume(netdev->dev.parent);
5185 
5186 		e1000e_enable_receives(adapter);
5187 		goto link_up;
5188 	}
5189 
5190 	if ((e1000e_enable_tx_pkt_filtering(hw)) &&
5191 	    (adapter->mng_vlan_id != adapter->hw.mng_cookie.vlan_id))
5192 		e1000_update_mng_vlan(adapter);
5193 
5194 	if (link) {
5195 		if (!netif_carrier_ok(netdev)) {
5196 			bool txb2b = true;
5197 
5198 			/* Cancel scheduled suspend requests. */
5199 			pm_runtime_resume(netdev->dev.parent);
5200 
5201 			/* update snapshot of PHY registers on LSC */
5202 			e1000_phy_read_status(adapter);
5203 			mac->ops.get_link_up_info(&adapter->hw,
5204 						  &adapter->link_speed,
5205 						  &adapter->link_duplex);
5206 			e1000_print_link_info(adapter);
5207 
5208 			/* check if SmartSpeed worked */
5209 			e1000e_check_downshift(hw);
5210 			if (phy->speed_downgraded)
5211 				netdev_warn(netdev,
5212 					    "Link Speed was downgraded by SmartSpeed\n");
5213 
5214 			/* On supported PHYs, check for duplex mismatch only
5215 			 * if link has autonegotiated at 10/100 half
5216 			 */
5217 			if ((hw->phy.type == e1000_phy_igp_3 ||
5218 			     hw->phy.type == e1000_phy_bm) &&
5219 			    hw->mac.autoneg &&
5220 			    (adapter->link_speed == SPEED_10 ||
5221 			     adapter->link_speed == SPEED_100) &&
5222 			    (adapter->link_duplex == HALF_DUPLEX)) {
5223 				u16 autoneg_exp;
5224 
5225 				e1e_rphy(hw, MII_EXPANSION, &autoneg_exp);
5226 
5227 				if (!(autoneg_exp & EXPANSION_NWAY))
5228 					e_info("Autonegotiated half duplex but link partner cannot autoneg.  Try forcing full duplex if link gets many collisions.\n");
5229 			}
5230 
5231 			/* adjust timeout factor according to speed/duplex */
5232 			adapter->tx_timeout_factor = 1;
5233 			switch (adapter->link_speed) {
5234 			case SPEED_10:
5235 				txb2b = false;
5236 				adapter->tx_timeout_factor = 16;
5237 				break;
5238 			case SPEED_100:
5239 				txb2b = false;
5240 				adapter->tx_timeout_factor = 10;
5241 				break;
5242 			}
5243 
5244 			/* workaround: re-program speed mode bit after
5245 			 * link-up event
5246 			 */
5247 			if ((adapter->flags & FLAG_TARC_SPEED_MODE_BIT) &&
5248 			    !txb2b) {
5249 				u32 tarc0;
5250 
5251 				tarc0 = er32(TARC(0));
5252 				tarc0 &= ~SPEED_MODE_BIT;
5253 				ew32(TARC(0), tarc0);
5254 			}
5255 
5256 			/* disable TSO for pcie and 10/100 speeds, to avoid
5257 			 * some hardware issues
5258 			 */
5259 			if (!(adapter->flags & FLAG_TSO_FORCE)) {
5260 				switch (adapter->link_speed) {
5261 				case SPEED_10:
5262 				case SPEED_100:
5263 					e_info("10/100 speed: disabling TSO\n");
5264 					netdev->features &= ~NETIF_F_TSO;
5265 					netdev->features &= ~NETIF_F_TSO6;
5266 					break;
5267 				case SPEED_1000:
5268 					netdev->features |= NETIF_F_TSO;
5269 					netdev->features |= NETIF_F_TSO6;
5270 					break;
5271 				default:
5272 					/* oops */
5273 					break;
5274 				}
5275 			}
5276 
5277 			/* enable transmits in the hardware, need to do this
5278 			 * after setting TARC(0)
5279 			 */
5280 			tctl = er32(TCTL);
5281 			tctl |= E1000_TCTL_EN;
5282 			ew32(TCTL, tctl);
5283 
5284 			/* Perform any post-link-up configuration before
5285 			 * reporting link up.
5286 			 */
5287 			if (phy->ops.cfg_on_link_up)
5288 				phy->ops.cfg_on_link_up(hw);
5289 
5290 			netif_carrier_on(netdev);
5291 
5292 			if (!test_bit(__E1000_DOWN, &adapter->state))
5293 				mod_timer(&adapter->phy_info_timer,
5294 					  round_jiffies(jiffies + 2 * HZ));
5295 		}
5296 	} else {
5297 		if (netif_carrier_ok(netdev)) {
5298 			adapter->link_speed = 0;
5299 			adapter->link_duplex = 0;
5300 			/* Link status message must follow this format */
5301 			pr_info("%s NIC Link is Down\n", adapter->netdev->name);
5302 			netif_carrier_off(netdev);
5303 			if (!test_bit(__E1000_DOWN, &adapter->state))
5304 				mod_timer(&adapter->phy_info_timer,
5305 					  round_jiffies(jiffies + 2 * HZ));
5306 
5307 			/* 8000ES2LAN requires a Rx packet buffer work-around
5308 			 * on link down event; reset the controller to flush
5309 			 * the Rx packet buffer.
5310 			 */
5311 			if (adapter->flags & FLAG_RX_NEEDS_RESTART)
5312 				adapter->flags |= FLAG_RESTART_NOW;
5313 			else
5314 				pm_schedule_suspend(netdev->dev.parent,
5315 						    LINK_TIMEOUT);
5316 		}
5317 	}
5318 
5319 link_up:
5320 	spin_lock(&adapter->stats64_lock);
5321 	e1000e_update_stats(adapter);
5322 
5323 	mac->tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
5324 	adapter->tpt_old = adapter->stats.tpt;
5325 	mac->collision_delta = adapter->stats.colc - adapter->colc_old;
5326 	adapter->colc_old = adapter->stats.colc;
5327 
5328 	adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
5329 	adapter->gorc_old = adapter->stats.gorc;
5330 	adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
5331 	adapter->gotc_old = adapter->stats.gotc;
5332 	spin_unlock(&adapter->stats64_lock);
5333 
5334 	/* If the link is lost the controller stops DMA, but
5335 	 * if there is queued Tx work it cannot be done.  So
5336 	 * reset the controller to flush the Tx packet buffers.
5337 	 */
5338 	if (!netif_carrier_ok(netdev) &&
5339 	    (e1000_desc_unused(tx_ring) + 1 < tx_ring->count))
5340 		adapter->flags |= FLAG_RESTART_NOW;
5341 
5342 	/* If reset is necessary, do it outside of interrupt context. */
5343 	if (adapter->flags & FLAG_RESTART_NOW) {
5344 		schedule_work(&adapter->reset_task);
5345 		/* return immediately since reset is imminent */
5346 		return;
5347 	}
5348 
5349 	e1000e_update_adaptive(&adapter->hw);
5350 
5351 	/* Simple mode for Interrupt Throttle Rate (ITR) */
5352 	if (adapter->itr_setting == 4) {
5353 		/* Symmetric Tx/Rx gets a reduced ITR=2000;
5354 		 * Total asymmetrical Tx or Rx gets ITR=8000;
5355 		 * everyone else is between 2000-8000.
5356 		 */
5357 		u32 goc = (adapter->gotc + adapter->gorc) / 10000;
5358 		u32 dif = (adapter->gotc > adapter->gorc ?
5359 			   adapter->gotc - adapter->gorc :
5360 			   adapter->gorc - adapter->gotc) / 10000;
5361 		u32 itr = goc > 0 ? (dif * 6000 / goc + 2000) : 8000;
5362 
5363 		e1000e_write_itr(adapter, itr);
5364 	}
5365 
5366 	/* Cause software interrupt to ensure Rx ring is cleaned */
5367 	if (adapter->msix_entries)
5368 		ew32(ICS, adapter->rx_ring->ims_val);
5369 	else
5370 		ew32(ICS, E1000_ICS_RXDMT0);
5371 
5372 	/* flush pending descriptors to memory before detecting Tx hang */
5373 	e1000e_flush_descriptors(adapter);
5374 
5375 	/* Force detection of hung controller every watchdog period */
5376 	adapter->detect_tx_hung = true;
5377 
5378 	/* With 82571 controllers, LAA may be overwritten due to controller
5379 	 * reset from the other port. Set the appropriate LAA in RAR[0]
5380 	 */
5381 	if (e1000e_get_laa_state_82571(hw))
5382 		hw->mac.ops.rar_set(hw, adapter->hw.mac.addr, 0);
5383 
5384 	if (adapter->flags2 & FLAG2_CHECK_PHY_HANG)
5385 		e1000e_check_82574_phy_workaround(adapter);
5386 
5387 	/* Clear valid timestamp stuck in RXSTMPL/H due to a Rx error */
5388 	if (adapter->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE) {
5389 		if ((adapter->flags2 & FLAG2_CHECK_RX_HWTSTAMP) &&
5390 		    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID)) {
5391 			er32(RXSTMPH);
5392 			adapter->rx_hwtstamp_cleared++;
5393 		} else {
5394 			adapter->flags2 |= FLAG2_CHECK_RX_HWTSTAMP;
5395 		}
5396 	}
5397 
5398 	/* Reset the timer */
5399 	if (!test_bit(__E1000_DOWN, &adapter->state))
5400 		mod_timer(&adapter->watchdog_timer,
5401 			  round_jiffies(jiffies + 2 * HZ));
5402 }
5403 
5404 #define E1000_TX_FLAGS_CSUM		0x00000001
5405 #define E1000_TX_FLAGS_VLAN		0x00000002
5406 #define E1000_TX_FLAGS_TSO		0x00000004
5407 #define E1000_TX_FLAGS_IPV4		0x00000008
5408 #define E1000_TX_FLAGS_NO_FCS		0x00000010
5409 #define E1000_TX_FLAGS_HWTSTAMP		0x00000020
5410 #define E1000_TX_FLAGS_VLAN_MASK	0xffff0000
5411 #define E1000_TX_FLAGS_VLAN_SHIFT	16
5412 
5413 static int e1000_tso(struct e1000_ring *tx_ring, struct sk_buff *skb,
5414 		     __be16 protocol)
5415 {
5416 	struct e1000_context_desc *context_desc;
5417 	struct e1000_buffer *buffer_info;
5418 	unsigned int i;
5419 	u32 cmd_length = 0;
5420 	u16 ipcse = 0, mss;
5421 	u8 ipcss, ipcso, tucss, tucso, hdr_len;
5422 	int err;
5423 
5424 	if (!skb_is_gso(skb))
5425 		return 0;
5426 
5427 	err = skb_cow_head(skb, 0);
5428 	if (err < 0)
5429 		return err;
5430 
5431 	hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5432 	mss = skb_shinfo(skb)->gso_size;
5433 	if (protocol == htons(ETH_P_IP)) {
5434 		struct iphdr *iph = ip_hdr(skb);
5435 		iph->tot_len = 0;
5436 		iph->check = 0;
5437 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
5438 							 0, IPPROTO_TCP, 0);
5439 		cmd_length = E1000_TXD_CMD_IP;
5440 		ipcse = skb_transport_offset(skb) - 1;
5441 	} else if (skb_is_gso_v6(skb)) {
5442 		ipv6_hdr(skb)->payload_len = 0;
5443 		tcp_hdr(skb)->check = ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
5444 						       &ipv6_hdr(skb)->daddr,
5445 						       0, IPPROTO_TCP, 0);
5446 		ipcse = 0;
5447 	}
5448 	ipcss = skb_network_offset(skb);
5449 	ipcso = (void *)&(ip_hdr(skb)->check) - (void *)skb->data;
5450 	tucss = skb_transport_offset(skb);
5451 	tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
5452 
5453 	cmd_length |= (E1000_TXD_CMD_DEXT | E1000_TXD_CMD_TSE |
5454 		       E1000_TXD_CMD_TCP | (skb->len - (hdr_len)));
5455 
5456 	i = tx_ring->next_to_use;
5457 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5458 	buffer_info = &tx_ring->buffer_info[i];
5459 
5460 	context_desc->lower_setup.ip_fields.ipcss = ipcss;
5461 	context_desc->lower_setup.ip_fields.ipcso = ipcso;
5462 	context_desc->lower_setup.ip_fields.ipcse = cpu_to_le16(ipcse);
5463 	context_desc->upper_setup.tcp_fields.tucss = tucss;
5464 	context_desc->upper_setup.tcp_fields.tucso = tucso;
5465 	context_desc->upper_setup.tcp_fields.tucse = 0;
5466 	context_desc->tcp_seg_setup.fields.mss = cpu_to_le16(mss);
5467 	context_desc->tcp_seg_setup.fields.hdr_len = hdr_len;
5468 	context_desc->cmd_and_length = cpu_to_le32(cmd_length);
5469 
5470 	buffer_info->time_stamp = jiffies;
5471 	buffer_info->next_to_watch = i;
5472 
5473 	i++;
5474 	if (i == tx_ring->count)
5475 		i = 0;
5476 	tx_ring->next_to_use = i;
5477 
5478 	return 1;
5479 }
5480 
5481 static bool e1000_tx_csum(struct e1000_ring *tx_ring, struct sk_buff *skb,
5482 			  __be16 protocol)
5483 {
5484 	struct e1000_adapter *adapter = tx_ring->adapter;
5485 	struct e1000_context_desc *context_desc;
5486 	struct e1000_buffer *buffer_info;
5487 	unsigned int i;
5488 	u8 css;
5489 	u32 cmd_len = E1000_TXD_CMD_DEXT;
5490 
5491 	if (skb->ip_summed != CHECKSUM_PARTIAL)
5492 		return false;
5493 
5494 	switch (protocol) {
5495 	case cpu_to_be16(ETH_P_IP):
5496 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
5497 			cmd_len |= E1000_TXD_CMD_TCP;
5498 		break;
5499 	case cpu_to_be16(ETH_P_IPV6):
5500 		/* XXX not handling all IPV6 headers */
5501 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
5502 			cmd_len |= E1000_TXD_CMD_TCP;
5503 		break;
5504 	default:
5505 		if (unlikely(net_ratelimit()))
5506 			e_warn("checksum_partial proto=%x!\n",
5507 			       be16_to_cpu(protocol));
5508 		break;
5509 	}
5510 
5511 	css = skb_checksum_start_offset(skb);
5512 
5513 	i = tx_ring->next_to_use;
5514 	buffer_info = &tx_ring->buffer_info[i];
5515 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5516 
5517 	context_desc->lower_setup.ip_config = 0;
5518 	context_desc->upper_setup.tcp_fields.tucss = css;
5519 	context_desc->upper_setup.tcp_fields.tucso = css + skb->csum_offset;
5520 	context_desc->upper_setup.tcp_fields.tucse = 0;
5521 	context_desc->tcp_seg_setup.data = 0;
5522 	context_desc->cmd_and_length = cpu_to_le32(cmd_len);
5523 
5524 	buffer_info->time_stamp = jiffies;
5525 	buffer_info->next_to_watch = i;
5526 
5527 	i++;
5528 	if (i == tx_ring->count)
5529 		i = 0;
5530 	tx_ring->next_to_use = i;
5531 
5532 	return true;
5533 }
5534 
5535 static int e1000_tx_map(struct e1000_ring *tx_ring, struct sk_buff *skb,
5536 			unsigned int first, unsigned int max_per_txd,
5537 			unsigned int nr_frags)
5538 {
5539 	struct e1000_adapter *adapter = tx_ring->adapter;
5540 	struct pci_dev *pdev = adapter->pdev;
5541 	struct e1000_buffer *buffer_info;
5542 	unsigned int len = skb_headlen(skb);
5543 	unsigned int offset = 0, size, count = 0, i;
5544 	unsigned int f, bytecount, segs;
5545 
5546 	i = tx_ring->next_to_use;
5547 
5548 	while (len) {
5549 		buffer_info = &tx_ring->buffer_info[i];
5550 		size = min(len, max_per_txd);
5551 
5552 		buffer_info->length = size;
5553 		buffer_info->time_stamp = jiffies;
5554 		buffer_info->next_to_watch = i;
5555 		buffer_info->dma = dma_map_single(&pdev->dev,
5556 						  skb->data + offset,
5557 						  size, DMA_TO_DEVICE);
5558 		buffer_info->mapped_as_page = false;
5559 		if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5560 			goto dma_error;
5561 
5562 		len -= size;
5563 		offset += size;
5564 		count++;
5565 
5566 		if (len) {
5567 			i++;
5568 			if (i == tx_ring->count)
5569 				i = 0;
5570 		}
5571 	}
5572 
5573 	for (f = 0; f < nr_frags; f++) {
5574 		const struct skb_frag_struct *frag;
5575 
5576 		frag = &skb_shinfo(skb)->frags[f];
5577 		len = skb_frag_size(frag);
5578 		offset = 0;
5579 
5580 		while (len) {
5581 			i++;
5582 			if (i == tx_ring->count)
5583 				i = 0;
5584 
5585 			buffer_info = &tx_ring->buffer_info[i];
5586 			size = min(len, max_per_txd);
5587 
5588 			buffer_info->length = size;
5589 			buffer_info->time_stamp = jiffies;
5590 			buffer_info->next_to_watch = i;
5591 			buffer_info->dma = skb_frag_dma_map(&pdev->dev, frag,
5592 							    offset, size,
5593 							    DMA_TO_DEVICE);
5594 			buffer_info->mapped_as_page = true;
5595 			if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5596 				goto dma_error;
5597 
5598 			len -= size;
5599 			offset += size;
5600 			count++;
5601 		}
5602 	}
5603 
5604 	segs = skb_shinfo(skb)->gso_segs ? : 1;
5605 	/* multiply data chunks by size of headers */
5606 	bytecount = ((segs - 1) * skb_headlen(skb)) + skb->len;
5607 
5608 	tx_ring->buffer_info[i].skb = skb;
5609 	tx_ring->buffer_info[i].segs = segs;
5610 	tx_ring->buffer_info[i].bytecount = bytecount;
5611 	tx_ring->buffer_info[first].next_to_watch = i;
5612 
5613 	return count;
5614 
5615 dma_error:
5616 	dev_err(&pdev->dev, "Tx DMA map failed\n");
5617 	buffer_info->dma = 0;
5618 	if (count)
5619 		count--;
5620 
5621 	while (count--) {
5622 		if (i == 0)
5623 			i += tx_ring->count;
5624 		i--;
5625 		buffer_info = &tx_ring->buffer_info[i];
5626 		e1000_put_txbuf(tx_ring, buffer_info, true);
5627 	}
5628 
5629 	return 0;
5630 }
5631 
5632 static void e1000_tx_queue(struct e1000_ring *tx_ring, int tx_flags, int count)
5633 {
5634 	struct e1000_adapter *adapter = tx_ring->adapter;
5635 	struct e1000_tx_desc *tx_desc = NULL;
5636 	struct e1000_buffer *buffer_info;
5637 	u32 txd_upper = 0, txd_lower = E1000_TXD_CMD_IFCS;
5638 	unsigned int i;
5639 
5640 	if (tx_flags & E1000_TX_FLAGS_TSO) {
5641 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D |
5642 		    E1000_TXD_CMD_TSE;
5643 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5644 
5645 		if (tx_flags & E1000_TX_FLAGS_IPV4)
5646 			txd_upper |= E1000_TXD_POPTS_IXSM << 8;
5647 	}
5648 
5649 	if (tx_flags & E1000_TX_FLAGS_CSUM) {
5650 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5651 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5652 	}
5653 
5654 	if (tx_flags & E1000_TX_FLAGS_VLAN) {
5655 		txd_lower |= E1000_TXD_CMD_VLE;
5656 		txd_upper |= (tx_flags & E1000_TX_FLAGS_VLAN_MASK);
5657 	}
5658 
5659 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5660 		txd_lower &= ~(E1000_TXD_CMD_IFCS);
5661 
5662 	if (unlikely(tx_flags & E1000_TX_FLAGS_HWTSTAMP)) {
5663 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5664 		txd_upper |= E1000_TXD_EXTCMD_TSTAMP;
5665 	}
5666 
5667 	i = tx_ring->next_to_use;
5668 
5669 	do {
5670 		buffer_info = &tx_ring->buffer_info[i];
5671 		tx_desc = E1000_TX_DESC(*tx_ring, i);
5672 		tx_desc->buffer_addr = cpu_to_le64(buffer_info->dma);
5673 		tx_desc->lower.data = cpu_to_le32(txd_lower |
5674 						  buffer_info->length);
5675 		tx_desc->upper.data = cpu_to_le32(txd_upper);
5676 
5677 		i++;
5678 		if (i == tx_ring->count)
5679 			i = 0;
5680 	} while (--count > 0);
5681 
5682 	tx_desc->lower.data |= cpu_to_le32(adapter->txd_cmd);
5683 
5684 	/* txd_cmd re-enables FCS, so we'll re-disable it here as desired. */
5685 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5686 		tx_desc->lower.data &= ~(cpu_to_le32(E1000_TXD_CMD_IFCS));
5687 
5688 	/* Force memory writes to complete before letting h/w
5689 	 * know there are new descriptors to fetch.  (Only
5690 	 * applicable for weak-ordered memory model archs,
5691 	 * such as IA-64).
5692 	 */
5693 	wmb();
5694 
5695 	tx_ring->next_to_use = i;
5696 }
5697 
5698 #define MINIMUM_DHCP_PACKET_SIZE 282
5699 static int e1000_transfer_dhcp_info(struct e1000_adapter *adapter,
5700 				    struct sk_buff *skb)
5701 {
5702 	struct e1000_hw *hw = &adapter->hw;
5703 	u16 length, offset;
5704 
5705 	if (skb_vlan_tag_present(skb) &&
5706 	    !((skb_vlan_tag_get(skb) == adapter->hw.mng_cookie.vlan_id) &&
5707 	      (adapter->hw.mng_cookie.status &
5708 	       E1000_MNG_DHCP_COOKIE_STATUS_VLAN)))
5709 		return 0;
5710 
5711 	if (skb->len <= MINIMUM_DHCP_PACKET_SIZE)
5712 		return 0;
5713 
5714 	if (((struct ethhdr *)skb->data)->h_proto != htons(ETH_P_IP))
5715 		return 0;
5716 
5717 	{
5718 		const struct iphdr *ip = (struct iphdr *)((u8 *)skb->data + 14);
5719 		struct udphdr *udp;
5720 
5721 		if (ip->protocol != IPPROTO_UDP)
5722 			return 0;
5723 
5724 		udp = (struct udphdr *)((u8 *)ip + (ip->ihl << 2));
5725 		if (ntohs(udp->dest) != 67)
5726 			return 0;
5727 
5728 		offset = (u8 *)udp + 8 - skb->data;
5729 		length = skb->len - offset;
5730 		return e1000e_mng_write_dhcp_info(hw, (u8 *)udp + 8, length);
5731 	}
5732 
5733 	return 0;
5734 }
5735 
5736 static int __e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5737 {
5738 	struct e1000_adapter *adapter = tx_ring->adapter;
5739 
5740 	netif_stop_queue(adapter->netdev);
5741 	/* Herbert's original patch had:
5742 	 *  smp_mb__after_netif_stop_queue();
5743 	 * but since that doesn't exist yet, just open code it.
5744 	 */
5745 	smp_mb();
5746 
5747 	/* We need to check again in a case another CPU has just
5748 	 * made room available.
5749 	 */
5750 	if (e1000_desc_unused(tx_ring) < size)
5751 		return -EBUSY;
5752 
5753 	/* A reprieve! */
5754 	netif_start_queue(adapter->netdev);
5755 	++adapter->restart_queue;
5756 	return 0;
5757 }
5758 
5759 static int e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5760 {
5761 	BUG_ON(size > tx_ring->count);
5762 
5763 	if (e1000_desc_unused(tx_ring) >= size)
5764 		return 0;
5765 	return __e1000_maybe_stop_tx(tx_ring, size);
5766 }
5767 
5768 static netdev_tx_t e1000_xmit_frame(struct sk_buff *skb,
5769 				    struct net_device *netdev)
5770 {
5771 	struct e1000_adapter *adapter = netdev_priv(netdev);
5772 	struct e1000_ring *tx_ring = adapter->tx_ring;
5773 	unsigned int first;
5774 	unsigned int tx_flags = 0;
5775 	unsigned int len = skb_headlen(skb);
5776 	unsigned int nr_frags;
5777 	unsigned int mss;
5778 	int count = 0;
5779 	int tso;
5780 	unsigned int f;
5781 	__be16 protocol = vlan_get_protocol(skb);
5782 
5783 	if (test_bit(__E1000_DOWN, &adapter->state)) {
5784 		dev_kfree_skb_any(skb);
5785 		return NETDEV_TX_OK;
5786 	}
5787 
5788 	if (skb->len <= 0) {
5789 		dev_kfree_skb_any(skb);
5790 		return NETDEV_TX_OK;
5791 	}
5792 
5793 	/* The minimum packet size with TCTL.PSP set is 17 bytes so
5794 	 * pad skb in order to meet this minimum size requirement
5795 	 */
5796 	if (skb_put_padto(skb, 17))
5797 		return NETDEV_TX_OK;
5798 
5799 	mss = skb_shinfo(skb)->gso_size;
5800 	if (mss) {
5801 		u8 hdr_len;
5802 
5803 		/* TSO Workaround for 82571/2/3 Controllers -- if skb->data
5804 		 * points to just header, pull a few bytes of payload from
5805 		 * frags into skb->data
5806 		 */
5807 		hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5808 		/* we do this workaround for ES2LAN, but it is un-necessary,
5809 		 * avoiding it could save a lot of cycles
5810 		 */
5811 		if (skb->data_len && (hdr_len == len)) {
5812 			unsigned int pull_size;
5813 
5814 			pull_size = min_t(unsigned int, 4, skb->data_len);
5815 			if (!__pskb_pull_tail(skb, pull_size)) {
5816 				e_err("__pskb_pull_tail failed.\n");
5817 				dev_kfree_skb_any(skb);
5818 				return NETDEV_TX_OK;
5819 			}
5820 			len = skb_headlen(skb);
5821 		}
5822 	}
5823 
5824 	/* reserve a descriptor for the offload context */
5825 	if ((mss) || (skb->ip_summed == CHECKSUM_PARTIAL))
5826 		count++;
5827 	count++;
5828 
5829 	count += DIV_ROUND_UP(len, adapter->tx_fifo_limit);
5830 
5831 	nr_frags = skb_shinfo(skb)->nr_frags;
5832 	for (f = 0; f < nr_frags; f++)
5833 		count += DIV_ROUND_UP(skb_frag_size(&skb_shinfo(skb)->frags[f]),
5834 				      adapter->tx_fifo_limit);
5835 
5836 	if (adapter->hw.mac.tx_pkt_filtering)
5837 		e1000_transfer_dhcp_info(adapter, skb);
5838 
5839 	/* need: count + 2 desc gap to keep tail from touching
5840 	 * head, otherwise try next time
5841 	 */
5842 	if (e1000_maybe_stop_tx(tx_ring, count + 2))
5843 		return NETDEV_TX_BUSY;
5844 
5845 	if (skb_vlan_tag_present(skb)) {
5846 		tx_flags |= E1000_TX_FLAGS_VLAN;
5847 		tx_flags |= (skb_vlan_tag_get(skb) <<
5848 			     E1000_TX_FLAGS_VLAN_SHIFT);
5849 	}
5850 
5851 	first = tx_ring->next_to_use;
5852 
5853 	tso = e1000_tso(tx_ring, skb, protocol);
5854 	if (tso < 0) {
5855 		dev_kfree_skb_any(skb);
5856 		return NETDEV_TX_OK;
5857 	}
5858 
5859 	if (tso)
5860 		tx_flags |= E1000_TX_FLAGS_TSO;
5861 	else if (e1000_tx_csum(tx_ring, skb, protocol))
5862 		tx_flags |= E1000_TX_FLAGS_CSUM;
5863 
5864 	/* Old method was to assume IPv4 packet by default if TSO was enabled.
5865 	 * 82571 hardware supports TSO capabilities for IPv6 as well...
5866 	 * no longer assume, we must.
5867 	 */
5868 	if (protocol == htons(ETH_P_IP))
5869 		tx_flags |= E1000_TX_FLAGS_IPV4;
5870 
5871 	if (unlikely(skb->no_fcs))
5872 		tx_flags |= E1000_TX_FLAGS_NO_FCS;
5873 
5874 	/* if count is 0 then mapping error has occurred */
5875 	count = e1000_tx_map(tx_ring, skb, first, adapter->tx_fifo_limit,
5876 			     nr_frags);
5877 	if (count) {
5878 		if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
5879 		    (adapter->flags & FLAG_HAS_HW_TIMESTAMP)) {
5880 			if (!adapter->tx_hwtstamp_skb) {
5881 				skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5882 				tx_flags |= E1000_TX_FLAGS_HWTSTAMP;
5883 				adapter->tx_hwtstamp_skb = skb_get(skb);
5884 				adapter->tx_hwtstamp_start = jiffies;
5885 				schedule_work(&adapter->tx_hwtstamp_work);
5886 			} else {
5887 				adapter->tx_hwtstamp_skipped++;
5888 			}
5889 		}
5890 
5891 		skb_tx_timestamp(skb);
5892 
5893 		netdev_sent_queue(netdev, skb->len);
5894 		e1000_tx_queue(tx_ring, tx_flags, count);
5895 		/* Make sure there is space in the ring for the next send. */
5896 		e1000_maybe_stop_tx(tx_ring,
5897 				    (MAX_SKB_FRAGS *
5898 				     DIV_ROUND_UP(PAGE_SIZE,
5899 						  adapter->tx_fifo_limit) + 2));
5900 
5901 		if (!skb->xmit_more ||
5902 		    netif_xmit_stopped(netdev_get_tx_queue(netdev, 0))) {
5903 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
5904 				e1000e_update_tdt_wa(tx_ring,
5905 						     tx_ring->next_to_use);
5906 			else
5907 				writel(tx_ring->next_to_use, tx_ring->tail);
5908 
5909 			/* we need this if more than one processor can write
5910 			 * to our tail at a time, it synchronizes IO on
5911 			 *IA64/Altix systems
5912 			 */
5913 			mmiowb();
5914 		}
5915 	} else {
5916 		dev_kfree_skb_any(skb);
5917 		tx_ring->buffer_info[first].time_stamp = 0;
5918 		tx_ring->next_to_use = first;
5919 	}
5920 
5921 	return NETDEV_TX_OK;
5922 }
5923 
5924 /**
5925  * e1000_tx_timeout - Respond to a Tx Hang
5926  * @netdev: network interface device structure
5927  **/
5928 static void e1000_tx_timeout(struct net_device *netdev)
5929 {
5930 	struct e1000_adapter *adapter = netdev_priv(netdev);
5931 
5932 	/* Do the reset outside of interrupt context */
5933 	adapter->tx_timeout_count++;
5934 	schedule_work(&adapter->reset_task);
5935 }
5936 
5937 static void e1000_reset_task(struct work_struct *work)
5938 {
5939 	struct e1000_adapter *adapter;
5940 	adapter = container_of(work, struct e1000_adapter, reset_task);
5941 
5942 	/* don't run the task if already down */
5943 	if (test_bit(__E1000_DOWN, &adapter->state))
5944 		return;
5945 
5946 	if (!(adapter->flags & FLAG_RESTART_NOW)) {
5947 		e1000e_dump(adapter);
5948 		e_err("Reset adapter unexpectedly\n");
5949 	}
5950 	e1000e_reinit_locked(adapter);
5951 }
5952 
5953 /**
5954  * e1000_get_stats64 - Get System Network Statistics
5955  * @netdev: network interface device structure
5956  * @stats: rtnl_link_stats64 pointer
5957  *
5958  * Returns the address of the device statistics structure.
5959  **/
5960 void e1000e_get_stats64(struct net_device *netdev,
5961 			struct rtnl_link_stats64 *stats)
5962 {
5963 	struct e1000_adapter *adapter = netdev_priv(netdev);
5964 
5965 	spin_lock(&adapter->stats64_lock);
5966 	e1000e_update_stats(adapter);
5967 	/* Fill out the OS statistics structure */
5968 	stats->rx_bytes = adapter->stats.gorc;
5969 	stats->rx_packets = adapter->stats.gprc;
5970 	stats->tx_bytes = adapter->stats.gotc;
5971 	stats->tx_packets = adapter->stats.gptc;
5972 	stats->multicast = adapter->stats.mprc;
5973 	stats->collisions = adapter->stats.colc;
5974 
5975 	/* Rx Errors */
5976 
5977 	/* RLEC on some newer hardware can be incorrect so build
5978 	 * our own version based on RUC and ROC
5979 	 */
5980 	stats->rx_errors = adapter->stats.rxerrc +
5981 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
5982 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5983 	stats->rx_length_errors = adapter->stats.ruc + adapter->stats.roc;
5984 	stats->rx_crc_errors = adapter->stats.crcerrs;
5985 	stats->rx_frame_errors = adapter->stats.algnerrc;
5986 	stats->rx_missed_errors = adapter->stats.mpc;
5987 
5988 	/* Tx Errors */
5989 	stats->tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5990 	stats->tx_aborted_errors = adapter->stats.ecol;
5991 	stats->tx_window_errors = adapter->stats.latecol;
5992 	stats->tx_carrier_errors = adapter->stats.tncrs;
5993 
5994 	/* Tx Dropped needs to be maintained elsewhere */
5995 
5996 	spin_unlock(&adapter->stats64_lock);
5997 }
5998 
5999 /**
6000  * e1000_change_mtu - Change the Maximum Transfer Unit
6001  * @netdev: network interface device structure
6002  * @new_mtu: new value for maximum frame size
6003  *
6004  * Returns 0 on success, negative on failure
6005  **/
6006 static int e1000_change_mtu(struct net_device *netdev, int new_mtu)
6007 {
6008 	struct e1000_adapter *adapter = netdev_priv(netdev);
6009 	int max_frame = new_mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
6010 
6011 	/* Jumbo frame support */
6012 	if ((new_mtu > ETH_DATA_LEN) &&
6013 	    !(adapter->flags & FLAG_HAS_JUMBO_FRAMES)) {
6014 		e_err("Jumbo Frames not supported.\n");
6015 		return -EINVAL;
6016 	}
6017 
6018 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
6019 	if ((adapter->hw.mac.type >= e1000_pch2lan) &&
6020 	    !(adapter->flags2 & FLAG2_CRC_STRIPPING) &&
6021 	    (new_mtu > ETH_DATA_LEN)) {
6022 		e_err("Jumbo Frames not supported on this device when CRC stripping is disabled.\n");
6023 		return -EINVAL;
6024 	}
6025 
6026 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
6027 		usleep_range(1000, 2000);
6028 	/* e1000e_down -> e1000e_reset dependent on max_frame_size & mtu */
6029 	adapter->max_frame_size = max_frame;
6030 	e_info("changing MTU from %d to %d\n", netdev->mtu, new_mtu);
6031 	netdev->mtu = new_mtu;
6032 
6033 	pm_runtime_get_sync(netdev->dev.parent);
6034 
6035 	if (netif_running(netdev))
6036 		e1000e_down(adapter, true);
6037 
6038 	/* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
6039 	 * means we reserve 2 more, this pushes us to allocate from the next
6040 	 * larger slab size.
6041 	 * i.e. RXBUFFER_2048 --> size-4096 slab
6042 	 * However with the new *_jumbo_rx* routines, jumbo receives will use
6043 	 * fragmented skbs
6044 	 */
6045 
6046 	if (max_frame <= 2048)
6047 		adapter->rx_buffer_len = 2048;
6048 	else
6049 		adapter->rx_buffer_len = 4096;
6050 
6051 	/* adjust allocation if LPE protects us, and we aren't using SBP */
6052 	if (max_frame <= (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN))
6053 		adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
6054 
6055 	if (netif_running(netdev))
6056 		e1000e_up(adapter);
6057 	else
6058 		e1000e_reset(adapter);
6059 
6060 	pm_runtime_put_sync(netdev->dev.parent);
6061 
6062 	clear_bit(__E1000_RESETTING, &adapter->state);
6063 
6064 	return 0;
6065 }
6066 
6067 static int e1000_mii_ioctl(struct net_device *netdev, struct ifreq *ifr,
6068 			   int cmd)
6069 {
6070 	struct e1000_adapter *adapter = netdev_priv(netdev);
6071 	struct mii_ioctl_data *data = if_mii(ifr);
6072 
6073 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
6074 		return -EOPNOTSUPP;
6075 
6076 	switch (cmd) {
6077 	case SIOCGMIIPHY:
6078 		data->phy_id = adapter->hw.phy.addr;
6079 		break;
6080 	case SIOCGMIIREG:
6081 		e1000_phy_read_status(adapter);
6082 
6083 		switch (data->reg_num & 0x1F) {
6084 		case MII_BMCR:
6085 			data->val_out = adapter->phy_regs.bmcr;
6086 			break;
6087 		case MII_BMSR:
6088 			data->val_out = adapter->phy_regs.bmsr;
6089 			break;
6090 		case MII_PHYSID1:
6091 			data->val_out = (adapter->hw.phy.id >> 16);
6092 			break;
6093 		case MII_PHYSID2:
6094 			data->val_out = (adapter->hw.phy.id & 0xFFFF);
6095 			break;
6096 		case MII_ADVERTISE:
6097 			data->val_out = adapter->phy_regs.advertise;
6098 			break;
6099 		case MII_LPA:
6100 			data->val_out = adapter->phy_regs.lpa;
6101 			break;
6102 		case MII_EXPANSION:
6103 			data->val_out = adapter->phy_regs.expansion;
6104 			break;
6105 		case MII_CTRL1000:
6106 			data->val_out = adapter->phy_regs.ctrl1000;
6107 			break;
6108 		case MII_STAT1000:
6109 			data->val_out = adapter->phy_regs.stat1000;
6110 			break;
6111 		case MII_ESTATUS:
6112 			data->val_out = adapter->phy_regs.estatus;
6113 			break;
6114 		default:
6115 			return -EIO;
6116 		}
6117 		break;
6118 	case SIOCSMIIREG:
6119 	default:
6120 		return -EOPNOTSUPP;
6121 	}
6122 	return 0;
6123 }
6124 
6125 /**
6126  * e1000e_hwtstamp_ioctl - control hardware time stamping
6127  * @netdev: network interface device structure
6128  * @ifreq: interface request
6129  *
6130  * Outgoing time stamping can be enabled and disabled. Play nice and
6131  * disable it when requested, although it shouldn't cause any overhead
6132  * when no packet needs it. At most one packet in the queue may be
6133  * marked for time stamping, otherwise it would be impossible to tell
6134  * for sure to which packet the hardware time stamp belongs.
6135  *
6136  * Incoming time stamping has to be configured via the hardware filters.
6137  * Not all combinations are supported, in particular event type has to be
6138  * specified. Matching the kind of event packet is not supported, with the
6139  * exception of "all V2 events regardless of level 2 or 4".
6140  **/
6141 static int e1000e_hwtstamp_set(struct net_device *netdev, struct ifreq *ifr)
6142 {
6143 	struct e1000_adapter *adapter = netdev_priv(netdev);
6144 	struct hwtstamp_config config;
6145 	int ret_val;
6146 
6147 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
6148 		return -EFAULT;
6149 
6150 	ret_val = e1000e_config_hwtstamp(adapter, &config);
6151 	if (ret_val)
6152 		return ret_val;
6153 
6154 	switch (config.rx_filter) {
6155 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
6156 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
6157 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
6158 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
6159 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
6160 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
6161 		/* With V2 type filters which specify a Sync or Delay Request,
6162 		 * Path Delay Request/Response messages are also time stamped
6163 		 * by hardware so notify the caller the requested packets plus
6164 		 * some others are time stamped.
6165 		 */
6166 		config.rx_filter = HWTSTAMP_FILTER_SOME;
6167 		break;
6168 	default:
6169 		break;
6170 	}
6171 
6172 	return copy_to_user(ifr->ifr_data, &config,
6173 			    sizeof(config)) ? -EFAULT : 0;
6174 }
6175 
6176 static int e1000e_hwtstamp_get(struct net_device *netdev, struct ifreq *ifr)
6177 {
6178 	struct e1000_adapter *adapter = netdev_priv(netdev);
6179 
6180 	return copy_to_user(ifr->ifr_data, &adapter->hwtstamp_config,
6181 			    sizeof(adapter->hwtstamp_config)) ? -EFAULT : 0;
6182 }
6183 
6184 static int e1000_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
6185 {
6186 	switch (cmd) {
6187 	case SIOCGMIIPHY:
6188 	case SIOCGMIIREG:
6189 	case SIOCSMIIREG:
6190 		return e1000_mii_ioctl(netdev, ifr, cmd);
6191 	case SIOCSHWTSTAMP:
6192 		return e1000e_hwtstamp_set(netdev, ifr);
6193 	case SIOCGHWTSTAMP:
6194 		return e1000e_hwtstamp_get(netdev, ifr);
6195 	default:
6196 		return -EOPNOTSUPP;
6197 	}
6198 }
6199 
6200 static int e1000_init_phy_wakeup(struct e1000_adapter *adapter, u32 wufc)
6201 {
6202 	struct e1000_hw *hw = &adapter->hw;
6203 	u32 i, mac_reg, wuc;
6204 	u16 phy_reg, wuc_enable;
6205 	int retval;
6206 
6207 	/* copy MAC RARs to PHY RARs */
6208 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
6209 
6210 	retval = hw->phy.ops.acquire(hw);
6211 	if (retval) {
6212 		e_err("Could not acquire PHY\n");
6213 		return retval;
6214 	}
6215 
6216 	/* Enable access to wakeup registers on and set page to BM_WUC_PAGE */
6217 	retval = e1000_enable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6218 	if (retval)
6219 		goto release;
6220 
6221 	/* copy MAC MTA to PHY MTA - only needed for pchlan */
6222 	for (i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
6223 		mac_reg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
6224 		hw->phy.ops.write_reg_page(hw, BM_MTA(i),
6225 					   (u16)(mac_reg & 0xFFFF));
6226 		hw->phy.ops.write_reg_page(hw, BM_MTA(i) + 1,
6227 					   (u16)((mac_reg >> 16) & 0xFFFF));
6228 	}
6229 
6230 	/* configure PHY Rx Control register */
6231 	hw->phy.ops.read_reg_page(&adapter->hw, BM_RCTL, &phy_reg);
6232 	mac_reg = er32(RCTL);
6233 	if (mac_reg & E1000_RCTL_UPE)
6234 		phy_reg |= BM_RCTL_UPE;
6235 	if (mac_reg & E1000_RCTL_MPE)
6236 		phy_reg |= BM_RCTL_MPE;
6237 	phy_reg &= ~(BM_RCTL_MO_MASK);
6238 	if (mac_reg & E1000_RCTL_MO_3)
6239 		phy_reg |= (((mac_reg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
6240 			    << BM_RCTL_MO_SHIFT);
6241 	if (mac_reg & E1000_RCTL_BAM)
6242 		phy_reg |= BM_RCTL_BAM;
6243 	if (mac_reg & E1000_RCTL_PMCF)
6244 		phy_reg |= BM_RCTL_PMCF;
6245 	mac_reg = er32(CTRL);
6246 	if (mac_reg & E1000_CTRL_RFCE)
6247 		phy_reg |= BM_RCTL_RFCE;
6248 	hw->phy.ops.write_reg_page(&adapter->hw, BM_RCTL, phy_reg);
6249 
6250 	wuc = E1000_WUC_PME_EN;
6251 	if (wufc & (E1000_WUFC_MAG | E1000_WUFC_LNKC))
6252 		wuc |= E1000_WUC_APME;
6253 
6254 	/* enable PHY wakeup in MAC register */
6255 	ew32(WUFC, wufc);
6256 	ew32(WUC, (E1000_WUC_PHY_WAKE | E1000_WUC_APMPME |
6257 		   E1000_WUC_PME_STATUS | wuc));
6258 
6259 	/* configure and enable PHY wakeup in PHY registers */
6260 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUFC, wufc);
6261 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUC, wuc);
6262 
6263 	/* activate PHY wakeup */
6264 	wuc_enable |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
6265 	retval = e1000_disable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6266 	if (retval)
6267 		e_err("Could not set PHY Host Wakeup bit\n");
6268 release:
6269 	hw->phy.ops.release(hw);
6270 
6271 	return retval;
6272 }
6273 
6274 static void e1000e_flush_lpic(struct pci_dev *pdev)
6275 {
6276 	struct net_device *netdev = pci_get_drvdata(pdev);
6277 	struct e1000_adapter *adapter = netdev_priv(netdev);
6278 	struct e1000_hw *hw = &adapter->hw;
6279 	u32 ret_val;
6280 
6281 	pm_runtime_get_sync(netdev->dev.parent);
6282 
6283 	ret_val = hw->phy.ops.acquire(hw);
6284 	if (ret_val)
6285 		goto fl_out;
6286 
6287 	pr_info("EEE TX LPI TIMER: %08X\n",
6288 		er32(LPIC) >> E1000_LPIC_LPIET_SHIFT);
6289 
6290 	hw->phy.ops.release(hw);
6291 
6292 fl_out:
6293 	pm_runtime_put_sync(netdev->dev.parent);
6294 }
6295 
6296 static int e1000e_pm_freeze(struct device *dev)
6297 {
6298 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6299 	struct e1000_adapter *adapter = netdev_priv(netdev);
6300 
6301 	netif_device_detach(netdev);
6302 
6303 	if (netif_running(netdev)) {
6304 		int count = E1000_CHECK_RESET_COUNT;
6305 
6306 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6307 			usleep_range(10000, 20000);
6308 
6309 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6310 
6311 		/* Quiesce the device without resetting the hardware */
6312 		e1000e_down(adapter, false);
6313 		e1000_free_irq(adapter);
6314 	}
6315 	e1000e_reset_interrupt_capability(adapter);
6316 
6317 	/* Allow time for pending master requests to run */
6318 	e1000e_disable_pcie_master(&adapter->hw);
6319 
6320 	return 0;
6321 }
6322 
6323 static int __e1000_shutdown(struct pci_dev *pdev, bool runtime)
6324 {
6325 	struct net_device *netdev = pci_get_drvdata(pdev);
6326 	struct e1000_adapter *adapter = netdev_priv(netdev);
6327 	struct e1000_hw *hw = &adapter->hw;
6328 	u32 ctrl, ctrl_ext, rctl, status;
6329 	/* Runtime suspend should only enable wakeup for link changes */
6330 	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
6331 	int retval = 0;
6332 
6333 	status = er32(STATUS);
6334 	if (status & E1000_STATUS_LU)
6335 		wufc &= ~E1000_WUFC_LNKC;
6336 
6337 	if (wufc) {
6338 		e1000_setup_rctl(adapter);
6339 		e1000e_set_rx_mode(netdev);
6340 
6341 		/* turn on all-multi mode if wake on multicast is enabled */
6342 		if (wufc & E1000_WUFC_MC) {
6343 			rctl = er32(RCTL);
6344 			rctl |= E1000_RCTL_MPE;
6345 			ew32(RCTL, rctl);
6346 		}
6347 
6348 		ctrl = er32(CTRL);
6349 		ctrl |= E1000_CTRL_ADVD3WUC;
6350 		if (!(adapter->flags2 & FLAG2_HAS_PHY_WAKEUP))
6351 			ctrl |= E1000_CTRL_EN_PHY_PWR_MGMT;
6352 		ew32(CTRL, ctrl);
6353 
6354 		if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
6355 		    adapter->hw.phy.media_type ==
6356 		    e1000_media_type_internal_serdes) {
6357 			/* keep the laser running in D3 */
6358 			ctrl_ext = er32(CTRL_EXT);
6359 			ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
6360 			ew32(CTRL_EXT, ctrl_ext);
6361 		}
6362 
6363 		if (!runtime)
6364 			e1000e_power_up_phy(adapter);
6365 
6366 		if (adapter->flags & FLAG_IS_ICH)
6367 			e1000_suspend_workarounds_ich8lan(&adapter->hw);
6368 
6369 		if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6370 			/* enable wakeup by the PHY */
6371 			retval = e1000_init_phy_wakeup(adapter, wufc);
6372 			if (retval)
6373 				return retval;
6374 		} else {
6375 			/* enable wakeup by the MAC */
6376 			ew32(WUFC, wufc);
6377 			ew32(WUC, E1000_WUC_PME_EN);
6378 		}
6379 	} else {
6380 		ew32(WUC, 0);
6381 		ew32(WUFC, 0);
6382 
6383 		e1000_power_down_phy(adapter);
6384 	}
6385 
6386 	if (adapter->hw.phy.type == e1000_phy_igp_3) {
6387 		e1000e_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
6388 	} else if (hw->mac.type >= e1000_pch_lpt) {
6389 		if (!(wufc & (E1000_WUFC_EX | E1000_WUFC_MC | E1000_WUFC_BC)))
6390 			/* ULP does not support wake from unicast, multicast
6391 			 * or broadcast.
6392 			 */
6393 			retval = e1000_enable_ulp_lpt_lp(hw, !runtime);
6394 
6395 		if (retval)
6396 			return retval;
6397 	}
6398 
6399 	/* Ensure that the appropriate bits are set in LPI_CTRL
6400 	 * for EEE in Sx
6401 	 */
6402 	if ((hw->phy.type >= e1000_phy_i217) &&
6403 	    adapter->eee_advert && hw->dev_spec.ich8lan.eee_lp_ability) {
6404 		u16 lpi_ctrl = 0;
6405 
6406 		retval = hw->phy.ops.acquire(hw);
6407 		if (!retval) {
6408 			retval = e1e_rphy_locked(hw, I82579_LPI_CTRL,
6409 						 &lpi_ctrl);
6410 			if (!retval) {
6411 				if (adapter->eee_advert &
6412 				    hw->dev_spec.ich8lan.eee_lp_ability &
6413 				    I82579_EEE_100_SUPPORTED)
6414 					lpi_ctrl |= I82579_LPI_CTRL_100_ENABLE;
6415 				if (adapter->eee_advert &
6416 				    hw->dev_spec.ich8lan.eee_lp_ability &
6417 				    I82579_EEE_1000_SUPPORTED)
6418 					lpi_ctrl |= I82579_LPI_CTRL_1000_ENABLE;
6419 
6420 				retval = e1e_wphy_locked(hw, I82579_LPI_CTRL,
6421 							 lpi_ctrl);
6422 			}
6423 		}
6424 		hw->phy.ops.release(hw);
6425 	}
6426 
6427 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
6428 	 * would have already happened in close and is redundant.
6429 	 */
6430 	e1000e_release_hw_control(adapter);
6431 
6432 	pci_clear_master(pdev);
6433 
6434 	/* The pci-e switch on some quad port adapters will report a
6435 	 * correctable error when the MAC transitions from D0 to D3.  To
6436 	 * prevent this we need to mask off the correctable errors on the
6437 	 * downstream port of the pci-e switch.
6438 	 *
6439 	 * We don't have the associated upstream bridge while assigning
6440 	 * the PCI device into guest. For example, the KVM on power is
6441 	 * one of the cases.
6442 	 */
6443 	if (adapter->flags & FLAG_IS_QUAD_PORT) {
6444 		struct pci_dev *us_dev = pdev->bus->self;
6445 		u16 devctl;
6446 
6447 		if (!us_dev)
6448 			return 0;
6449 
6450 		pcie_capability_read_word(us_dev, PCI_EXP_DEVCTL, &devctl);
6451 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL,
6452 					   (devctl & ~PCI_EXP_DEVCTL_CERE));
6453 
6454 		pci_save_state(pdev);
6455 		pci_prepare_to_sleep(pdev);
6456 
6457 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL, devctl);
6458 	}
6459 
6460 	return 0;
6461 }
6462 
6463 /**
6464  * __e1000e_disable_aspm - Disable ASPM states
6465  * @pdev: pointer to PCI device struct
6466  * @state: bit-mask of ASPM states to disable
6467  * @locked: indication if this context holds pci_bus_sem locked.
6468  *
6469  * Some devices *must* have certain ASPM states disabled per hardware errata.
6470  **/
6471 static void __e1000e_disable_aspm(struct pci_dev *pdev, u16 state, int locked)
6472 {
6473 	struct pci_dev *parent = pdev->bus->self;
6474 	u16 aspm_dis_mask = 0;
6475 	u16 pdev_aspmc, parent_aspmc;
6476 
6477 	switch (state) {
6478 	case PCIE_LINK_STATE_L0S:
6479 	case PCIE_LINK_STATE_L0S | PCIE_LINK_STATE_L1:
6480 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L0S;
6481 		/* fall-through - can't have L1 without L0s */
6482 	case PCIE_LINK_STATE_L1:
6483 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L1;
6484 		break;
6485 	default:
6486 		return;
6487 	}
6488 
6489 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6490 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6491 
6492 	if (parent) {
6493 		pcie_capability_read_word(parent, PCI_EXP_LNKCTL,
6494 					  &parent_aspmc);
6495 		parent_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6496 	}
6497 
6498 	/* Nothing to do if the ASPM states to be disabled already are */
6499 	if (!(pdev_aspmc & aspm_dis_mask) &&
6500 	    (!parent || !(parent_aspmc & aspm_dis_mask)))
6501 		return;
6502 
6503 	dev_info(&pdev->dev, "Disabling ASPM %s %s\n",
6504 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L0S) ?
6505 		 "L0s" : "",
6506 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L1) ?
6507 		 "L1" : "");
6508 
6509 #ifdef CONFIG_PCIEASPM
6510 	if (locked)
6511 		pci_disable_link_state_locked(pdev, state);
6512 	else
6513 		pci_disable_link_state(pdev, state);
6514 
6515 	/* Double-check ASPM control.  If not disabled by the above, the
6516 	 * BIOS is preventing that from happening (or CONFIG_PCIEASPM is
6517 	 * not enabled); override by writing PCI config space directly.
6518 	 */
6519 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6520 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6521 
6522 	if (!(aspm_dis_mask & pdev_aspmc))
6523 		return;
6524 #endif
6525 
6526 	/* Both device and parent should have the same ASPM setting.
6527 	 * Disable ASPM in downstream component first and then upstream.
6528 	 */
6529 	pcie_capability_clear_word(pdev, PCI_EXP_LNKCTL, aspm_dis_mask);
6530 
6531 	if (parent)
6532 		pcie_capability_clear_word(parent, PCI_EXP_LNKCTL,
6533 					   aspm_dis_mask);
6534 }
6535 
6536 /**
6537  * e1000e_disable_aspm - Disable ASPM states.
6538  * @pdev: pointer to PCI device struct
6539  * @state: bit-mask of ASPM states to disable
6540  *
6541  * This function acquires the pci_bus_sem!
6542  * Some devices *must* have certain ASPM states disabled per hardware errata.
6543  **/
6544 static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6545 {
6546 	__e1000e_disable_aspm(pdev, state, 0);
6547 }
6548 
6549 /**
6550  * e1000e_disable_aspm_locked   Disable ASPM states.
6551  * @pdev: pointer to PCI device struct
6552  * @state: bit-mask of ASPM states to disable
6553  *
6554  * This function must be called with pci_bus_sem acquired!
6555  * Some devices *must* have certain ASPM states disabled per hardware errata.
6556  **/
6557 static void e1000e_disable_aspm_locked(struct pci_dev *pdev, u16 state)
6558 {
6559 	__e1000e_disable_aspm(pdev, state, 1);
6560 }
6561 
6562 #ifdef CONFIG_PM
6563 static int __e1000_resume(struct pci_dev *pdev)
6564 {
6565 	struct net_device *netdev = pci_get_drvdata(pdev);
6566 	struct e1000_adapter *adapter = netdev_priv(netdev);
6567 	struct e1000_hw *hw = &adapter->hw;
6568 	u16 aspm_disable_flag = 0;
6569 
6570 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6571 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6572 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6573 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6574 	if (aspm_disable_flag)
6575 		e1000e_disable_aspm(pdev, aspm_disable_flag);
6576 
6577 	pci_set_master(pdev);
6578 
6579 	if (hw->mac.type >= e1000_pch2lan)
6580 		e1000_resume_workarounds_pchlan(&adapter->hw);
6581 
6582 	e1000e_power_up_phy(adapter);
6583 
6584 	/* report the system wakeup cause from S3/S4 */
6585 	if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6586 		u16 phy_data;
6587 
6588 		e1e_rphy(&adapter->hw, BM_WUS, &phy_data);
6589 		if (phy_data) {
6590 			e_info("PHY Wakeup cause - %s\n",
6591 			       phy_data & E1000_WUS_EX ? "Unicast Packet" :
6592 			       phy_data & E1000_WUS_MC ? "Multicast Packet" :
6593 			       phy_data & E1000_WUS_BC ? "Broadcast Packet" :
6594 			       phy_data & E1000_WUS_MAG ? "Magic Packet" :
6595 			       phy_data & E1000_WUS_LNKC ?
6596 			       "Link Status Change" : "other");
6597 		}
6598 		e1e_wphy(&adapter->hw, BM_WUS, ~0);
6599 	} else {
6600 		u32 wus = er32(WUS);
6601 
6602 		if (wus) {
6603 			e_info("MAC Wakeup cause - %s\n",
6604 			       wus & E1000_WUS_EX ? "Unicast Packet" :
6605 			       wus & E1000_WUS_MC ? "Multicast Packet" :
6606 			       wus & E1000_WUS_BC ? "Broadcast Packet" :
6607 			       wus & E1000_WUS_MAG ? "Magic Packet" :
6608 			       wus & E1000_WUS_LNKC ? "Link Status Change" :
6609 			       "other");
6610 		}
6611 		ew32(WUS, ~0);
6612 	}
6613 
6614 	e1000e_reset(adapter);
6615 
6616 	e1000_init_manageability_pt(adapter);
6617 
6618 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6619 	 * is up.  For all other cases, let the f/w know that the h/w is now
6620 	 * under the control of the driver.
6621 	 */
6622 	if (!(adapter->flags & FLAG_HAS_AMT))
6623 		e1000e_get_hw_control(adapter);
6624 
6625 	return 0;
6626 }
6627 
6628 #ifdef CONFIG_PM_SLEEP
6629 static int e1000e_pm_thaw(struct device *dev)
6630 {
6631 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6632 	struct e1000_adapter *adapter = netdev_priv(netdev);
6633 
6634 	e1000e_set_interrupt_capability(adapter);
6635 	if (netif_running(netdev)) {
6636 		u32 err = e1000_request_irq(adapter);
6637 
6638 		if (err)
6639 			return err;
6640 
6641 		e1000e_up(adapter);
6642 	}
6643 
6644 	netif_device_attach(netdev);
6645 
6646 	return 0;
6647 }
6648 
6649 static int e1000e_pm_suspend(struct device *dev)
6650 {
6651 	struct pci_dev *pdev = to_pci_dev(dev);
6652 	int rc;
6653 
6654 	e1000e_flush_lpic(pdev);
6655 
6656 	e1000e_pm_freeze(dev);
6657 
6658 	rc = __e1000_shutdown(pdev, false);
6659 	if (rc)
6660 		e1000e_pm_thaw(dev);
6661 
6662 	return rc;
6663 }
6664 
6665 static int e1000e_pm_resume(struct device *dev)
6666 {
6667 	struct pci_dev *pdev = to_pci_dev(dev);
6668 	int rc;
6669 
6670 	rc = __e1000_resume(pdev);
6671 	if (rc)
6672 		return rc;
6673 
6674 	return e1000e_pm_thaw(dev);
6675 }
6676 #endif /* CONFIG_PM_SLEEP */
6677 
6678 static int e1000e_pm_runtime_idle(struct device *dev)
6679 {
6680 	struct pci_dev *pdev = to_pci_dev(dev);
6681 	struct net_device *netdev = pci_get_drvdata(pdev);
6682 	struct e1000_adapter *adapter = netdev_priv(netdev);
6683 	u16 eee_lp;
6684 
6685 	eee_lp = adapter->hw.dev_spec.ich8lan.eee_lp_ability;
6686 
6687 	if (!e1000e_has_link(adapter)) {
6688 		adapter->hw.dev_spec.ich8lan.eee_lp_ability = eee_lp;
6689 		pm_schedule_suspend(dev, 5 * MSEC_PER_SEC);
6690 	}
6691 
6692 	return -EBUSY;
6693 }
6694 
6695 static int e1000e_pm_runtime_resume(struct device *dev)
6696 {
6697 	struct pci_dev *pdev = to_pci_dev(dev);
6698 	struct net_device *netdev = pci_get_drvdata(pdev);
6699 	struct e1000_adapter *adapter = netdev_priv(netdev);
6700 	int rc;
6701 
6702 	rc = __e1000_resume(pdev);
6703 	if (rc)
6704 		return rc;
6705 
6706 	if (netdev->flags & IFF_UP)
6707 		e1000e_up(adapter);
6708 
6709 	return rc;
6710 }
6711 
6712 static int e1000e_pm_runtime_suspend(struct device *dev)
6713 {
6714 	struct pci_dev *pdev = to_pci_dev(dev);
6715 	struct net_device *netdev = pci_get_drvdata(pdev);
6716 	struct e1000_adapter *adapter = netdev_priv(netdev);
6717 
6718 	if (netdev->flags & IFF_UP) {
6719 		int count = E1000_CHECK_RESET_COUNT;
6720 
6721 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6722 			usleep_range(10000, 20000);
6723 
6724 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6725 
6726 		/* Down the device without resetting the hardware */
6727 		e1000e_down(adapter, false);
6728 	}
6729 
6730 	if (__e1000_shutdown(pdev, true)) {
6731 		e1000e_pm_runtime_resume(dev);
6732 		return -EBUSY;
6733 	}
6734 
6735 	return 0;
6736 }
6737 #endif /* CONFIG_PM */
6738 
6739 static void e1000_shutdown(struct pci_dev *pdev)
6740 {
6741 	e1000e_flush_lpic(pdev);
6742 
6743 	e1000e_pm_freeze(&pdev->dev);
6744 
6745 	__e1000_shutdown(pdev, false);
6746 }
6747 
6748 #ifdef CONFIG_NET_POLL_CONTROLLER
6749 
6750 static irqreturn_t e1000_intr_msix(int __always_unused irq, void *data)
6751 {
6752 	struct net_device *netdev = data;
6753 	struct e1000_adapter *adapter = netdev_priv(netdev);
6754 
6755 	if (adapter->msix_entries) {
6756 		int vector, msix_irq;
6757 
6758 		vector = 0;
6759 		msix_irq = adapter->msix_entries[vector].vector;
6760 		if (disable_hardirq(msix_irq))
6761 			e1000_intr_msix_rx(msix_irq, netdev);
6762 		enable_irq(msix_irq);
6763 
6764 		vector++;
6765 		msix_irq = adapter->msix_entries[vector].vector;
6766 		if (disable_hardirq(msix_irq))
6767 			e1000_intr_msix_tx(msix_irq, netdev);
6768 		enable_irq(msix_irq);
6769 
6770 		vector++;
6771 		msix_irq = adapter->msix_entries[vector].vector;
6772 		if (disable_hardirq(msix_irq))
6773 			e1000_msix_other(msix_irq, netdev);
6774 		enable_irq(msix_irq);
6775 	}
6776 
6777 	return IRQ_HANDLED;
6778 }
6779 
6780 /**
6781  * e1000_netpoll
6782  * @netdev: network interface device structure
6783  *
6784  * Polling 'interrupt' - used by things like netconsole to send skbs
6785  * without having to re-enable interrupts. It's not called while
6786  * the interrupt routine is executing.
6787  */
6788 static void e1000_netpoll(struct net_device *netdev)
6789 {
6790 	struct e1000_adapter *adapter = netdev_priv(netdev);
6791 
6792 	switch (adapter->int_mode) {
6793 	case E1000E_INT_MODE_MSIX:
6794 		e1000_intr_msix(adapter->pdev->irq, netdev);
6795 		break;
6796 	case E1000E_INT_MODE_MSI:
6797 		if (disable_hardirq(adapter->pdev->irq))
6798 			e1000_intr_msi(adapter->pdev->irq, netdev);
6799 		enable_irq(adapter->pdev->irq);
6800 		break;
6801 	default:		/* E1000E_INT_MODE_LEGACY */
6802 		if (disable_hardirq(adapter->pdev->irq))
6803 			e1000_intr(adapter->pdev->irq, netdev);
6804 		enable_irq(adapter->pdev->irq);
6805 		break;
6806 	}
6807 }
6808 #endif
6809 
6810 /**
6811  * e1000_io_error_detected - called when PCI error is detected
6812  * @pdev: Pointer to PCI device
6813  * @state: The current pci connection state
6814  *
6815  * This function is called after a PCI bus error affecting
6816  * this device has been detected.
6817  */
6818 static pci_ers_result_t e1000_io_error_detected(struct pci_dev *pdev,
6819 						pci_channel_state_t state)
6820 {
6821 	struct net_device *netdev = pci_get_drvdata(pdev);
6822 	struct e1000_adapter *adapter = netdev_priv(netdev);
6823 
6824 	netif_device_detach(netdev);
6825 
6826 	if (state == pci_channel_io_perm_failure)
6827 		return PCI_ERS_RESULT_DISCONNECT;
6828 
6829 	if (netif_running(netdev))
6830 		e1000e_down(adapter, true);
6831 	pci_disable_device(pdev);
6832 
6833 	/* Request a slot slot reset. */
6834 	return PCI_ERS_RESULT_NEED_RESET;
6835 }
6836 
6837 /**
6838  * e1000_io_slot_reset - called after the pci bus has been reset.
6839  * @pdev: Pointer to PCI device
6840  *
6841  * Restart the card from scratch, as if from a cold-boot. Implementation
6842  * resembles the first-half of the e1000e_pm_resume routine.
6843  */
6844 static pci_ers_result_t e1000_io_slot_reset(struct pci_dev *pdev)
6845 {
6846 	struct net_device *netdev = pci_get_drvdata(pdev);
6847 	struct e1000_adapter *adapter = netdev_priv(netdev);
6848 	struct e1000_hw *hw = &adapter->hw;
6849 	u16 aspm_disable_flag = 0;
6850 	int err;
6851 	pci_ers_result_t result;
6852 
6853 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6854 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6855 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6856 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6857 	if (aspm_disable_flag)
6858 		e1000e_disable_aspm_locked(pdev, aspm_disable_flag);
6859 
6860 	err = pci_enable_device_mem(pdev);
6861 	if (err) {
6862 		dev_err(&pdev->dev,
6863 			"Cannot re-enable PCI device after reset.\n");
6864 		result = PCI_ERS_RESULT_DISCONNECT;
6865 	} else {
6866 		pdev->state_saved = true;
6867 		pci_restore_state(pdev);
6868 		pci_set_master(pdev);
6869 
6870 		pci_enable_wake(pdev, PCI_D3hot, 0);
6871 		pci_enable_wake(pdev, PCI_D3cold, 0);
6872 
6873 		e1000e_reset(adapter);
6874 		ew32(WUS, ~0);
6875 		result = PCI_ERS_RESULT_RECOVERED;
6876 	}
6877 
6878 	pci_cleanup_aer_uncorrect_error_status(pdev);
6879 
6880 	return result;
6881 }
6882 
6883 /**
6884  * e1000_io_resume - called when traffic can start flowing again.
6885  * @pdev: Pointer to PCI device
6886  *
6887  * This callback is called when the error recovery driver tells us that
6888  * its OK to resume normal operation. Implementation resembles the
6889  * second-half of the e1000e_pm_resume routine.
6890  */
6891 static void e1000_io_resume(struct pci_dev *pdev)
6892 {
6893 	struct net_device *netdev = pci_get_drvdata(pdev);
6894 	struct e1000_adapter *adapter = netdev_priv(netdev);
6895 
6896 	e1000_init_manageability_pt(adapter);
6897 
6898 	if (netif_running(netdev))
6899 		e1000e_up(adapter);
6900 
6901 	netif_device_attach(netdev);
6902 
6903 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6904 	 * is up.  For all other cases, let the f/w know that the h/w is now
6905 	 * under the control of the driver.
6906 	 */
6907 	if (!(adapter->flags & FLAG_HAS_AMT))
6908 		e1000e_get_hw_control(adapter);
6909 }
6910 
6911 static void e1000_print_device_info(struct e1000_adapter *adapter)
6912 {
6913 	struct e1000_hw *hw = &adapter->hw;
6914 	struct net_device *netdev = adapter->netdev;
6915 	u32 ret_val;
6916 	u8 pba_str[E1000_PBANUM_LENGTH];
6917 
6918 	/* print bus type/speed/width info */
6919 	e_info("(PCI Express:2.5GT/s:%s) %pM\n",
6920 	       /* bus width */
6921 	       ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
6922 		"Width x1"),
6923 	       /* MAC address */
6924 	       netdev->dev_addr);
6925 	e_info("Intel(R) PRO/%s Network Connection\n",
6926 	       (hw->phy.type == e1000_phy_ife) ? "10/100" : "1000");
6927 	ret_val = e1000_read_pba_string_generic(hw, pba_str,
6928 						E1000_PBANUM_LENGTH);
6929 	if (ret_val)
6930 		strlcpy((char *)pba_str, "Unknown", sizeof(pba_str));
6931 	e_info("MAC: %d, PHY: %d, PBA No: %s\n",
6932 	       hw->mac.type, hw->phy.type, pba_str);
6933 }
6934 
6935 static void e1000_eeprom_checks(struct e1000_adapter *adapter)
6936 {
6937 	struct e1000_hw *hw = &adapter->hw;
6938 	int ret_val;
6939 	u16 buf = 0;
6940 
6941 	if (hw->mac.type != e1000_82573)
6942 		return;
6943 
6944 	ret_val = e1000_read_nvm(hw, NVM_INIT_CONTROL2_REG, 1, &buf);
6945 	le16_to_cpus(&buf);
6946 	if (!ret_val && (!(buf & BIT(0)))) {
6947 		/* Deep Smart Power Down (DSPD) */
6948 		dev_warn(&adapter->pdev->dev,
6949 			 "Warning: detected DSPD enabled in EEPROM\n");
6950 	}
6951 }
6952 
6953 static netdev_features_t e1000_fix_features(struct net_device *netdev,
6954 					    netdev_features_t features)
6955 {
6956 	struct e1000_adapter *adapter = netdev_priv(netdev);
6957 	struct e1000_hw *hw = &adapter->hw;
6958 
6959 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
6960 	if ((hw->mac.type >= e1000_pch2lan) && (netdev->mtu > ETH_DATA_LEN))
6961 		features &= ~NETIF_F_RXFCS;
6962 
6963 	/* Since there is no support for separate Rx/Tx vlan accel
6964 	 * enable/disable make sure Tx flag is always in same state as Rx.
6965 	 */
6966 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
6967 		features |= NETIF_F_HW_VLAN_CTAG_TX;
6968 	else
6969 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
6970 
6971 	return features;
6972 }
6973 
6974 static int e1000_set_features(struct net_device *netdev,
6975 			      netdev_features_t features)
6976 {
6977 	struct e1000_adapter *adapter = netdev_priv(netdev);
6978 	netdev_features_t changed = features ^ netdev->features;
6979 
6980 	if (changed & (NETIF_F_TSO | NETIF_F_TSO6))
6981 		adapter->flags |= FLAG_TSO_FORCE;
6982 
6983 	if (!(changed & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX |
6984 			 NETIF_F_RXCSUM | NETIF_F_RXHASH | NETIF_F_RXFCS |
6985 			 NETIF_F_RXALL)))
6986 		return 0;
6987 
6988 	if (changed & NETIF_F_RXFCS) {
6989 		if (features & NETIF_F_RXFCS) {
6990 			adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
6991 		} else {
6992 			/* We need to take it back to defaults, which might mean
6993 			 * stripping is still disabled at the adapter level.
6994 			 */
6995 			if (adapter->flags2 & FLAG2_DFLT_CRC_STRIPPING)
6996 				adapter->flags2 |= FLAG2_CRC_STRIPPING;
6997 			else
6998 				adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
6999 		}
7000 	}
7001 
7002 	netdev->features = features;
7003 
7004 	if (netif_running(netdev))
7005 		e1000e_reinit_locked(adapter);
7006 	else
7007 		e1000e_reset(adapter);
7008 
7009 	return 0;
7010 }
7011 
7012 static const struct net_device_ops e1000e_netdev_ops = {
7013 	.ndo_open		= e1000e_open,
7014 	.ndo_stop		= e1000e_close,
7015 	.ndo_start_xmit		= e1000_xmit_frame,
7016 	.ndo_get_stats64	= e1000e_get_stats64,
7017 	.ndo_set_rx_mode	= e1000e_set_rx_mode,
7018 	.ndo_set_mac_address	= e1000_set_mac,
7019 	.ndo_change_mtu		= e1000_change_mtu,
7020 	.ndo_do_ioctl		= e1000_ioctl,
7021 	.ndo_tx_timeout		= e1000_tx_timeout,
7022 	.ndo_validate_addr	= eth_validate_addr,
7023 
7024 	.ndo_vlan_rx_add_vid	= e1000_vlan_rx_add_vid,
7025 	.ndo_vlan_rx_kill_vid	= e1000_vlan_rx_kill_vid,
7026 #ifdef CONFIG_NET_POLL_CONTROLLER
7027 	.ndo_poll_controller	= e1000_netpoll,
7028 #endif
7029 	.ndo_set_features = e1000_set_features,
7030 	.ndo_fix_features = e1000_fix_features,
7031 	.ndo_features_check	= passthru_features_check,
7032 };
7033 
7034 /**
7035  * e1000_probe - Device Initialization Routine
7036  * @pdev: PCI device information struct
7037  * @ent: entry in e1000_pci_tbl
7038  *
7039  * Returns 0 on success, negative on failure
7040  *
7041  * e1000_probe initializes an adapter identified by a pci_dev structure.
7042  * The OS initialization, configuring of the adapter private structure,
7043  * and a hardware reset occur.
7044  **/
7045 static int e1000_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7046 {
7047 	struct net_device *netdev;
7048 	struct e1000_adapter *adapter;
7049 	struct e1000_hw *hw;
7050 	const struct e1000_info *ei = e1000_info_tbl[ent->driver_data];
7051 	resource_size_t mmio_start, mmio_len;
7052 	resource_size_t flash_start, flash_len;
7053 	static int cards_found;
7054 	u16 aspm_disable_flag = 0;
7055 	int bars, i, err, pci_using_dac;
7056 	u16 eeprom_data = 0;
7057 	u16 eeprom_apme_mask = E1000_EEPROM_APME;
7058 	s32 ret_val = 0;
7059 
7060 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L0S)
7061 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7062 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L1)
7063 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7064 	if (aspm_disable_flag)
7065 		e1000e_disable_aspm(pdev, aspm_disable_flag);
7066 
7067 	err = pci_enable_device_mem(pdev);
7068 	if (err)
7069 		return err;
7070 
7071 	pci_using_dac = 0;
7072 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
7073 	if (!err) {
7074 		pci_using_dac = 1;
7075 	} else {
7076 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
7077 		if (err) {
7078 			dev_err(&pdev->dev,
7079 				"No usable DMA configuration, aborting\n");
7080 			goto err_dma;
7081 		}
7082 	}
7083 
7084 	bars = pci_select_bars(pdev, IORESOURCE_MEM);
7085 	err = pci_request_selected_regions_exclusive(pdev, bars,
7086 						     e1000e_driver_name);
7087 	if (err)
7088 		goto err_pci_reg;
7089 
7090 	/* AER (Advanced Error Reporting) hooks */
7091 	pci_enable_pcie_error_reporting(pdev);
7092 
7093 	pci_set_master(pdev);
7094 	/* PCI config space info */
7095 	err = pci_save_state(pdev);
7096 	if (err)
7097 		goto err_alloc_etherdev;
7098 
7099 	err = -ENOMEM;
7100 	netdev = alloc_etherdev(sizeof(struct e1000_adapter));
7101 	if (!netdev)
7102 		goto err_alloc_etherdev;
7103 
7104 	SET_NETDEV_DEV(netdev, &pdev->dev);
7105 
7106 	netdev->irq = pdev->irq;
7107 
7108 	pci_set_drvdata(pdev, netdev);
7109 	adapter = netdev_priv(netdev);
7110 	hw = &adapter->hw;
7111 	adapter->netdev = netdev;
7112 	adapter->pdev = pdev;
7113 	adapter->ei = ei;
7114 	adapter->pba = ei->pba;
7115 	adapter->flags = ei->flags;
7116 	adapter->flags2 = ei->flags2;
7117 	adapter->hw.adapter = adapter;
7118 	adapter->hw.mac.type = ei->mac;
7119 	adapter->max_hw_frame_size = ei->max_hw_frame_size;
7120 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
7121 
7122 	mmio_start = pci_resource_start(pdev, 0);
7123 	mmio_len = pci_resource_len(pdev, 0);
7124 
7125 	err = -EIO;
7126 	adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
7127 	if (!adapter->hw.hw_addr)
7128 		goto err_ioremap;
7129 
7130 	if ((adapter->flags & FLAG_HAS_FLASH) &&
7131 	    (pci_resource_flags(pdev, 1) & IORESOURCE_MEM) &&
7132 	    (hw->mac.type < e1000_pch_spt)) {
7133 		flash_start = pci_resource_start(pdev, 1);
7134 		flash_len = pci_resource_len(pdev, 1);
7135 		adapter->hw.flash_address = ioremap(flash_start, flash_len);
7136 		if (!adapter->hw.flash_address)
7137 			goto err_flashmap;
7138 	}
7139 
7140 	/* Set default EEE advertisement */
7141 	if (adapter->flags2 & FLAG2_HAS_EEE)
7142 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
7143 
7144 	/* construct the net_device struct */
7145 	netdev->netdev_ops = &e1000e_netdev_ops;
7146 	e1000e_set_ethtool_ops(netdev);
7147 	netdev->watchdog_timeo = 5 * HZ;
7148 	netif_napi_add(netdev, &adapter->napi, e1000e_poll, 64);
7149 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
7150 
7151 	netdev->mem_start = mmio_start;
7152 	netdev->mem_end = mmio_start + mmio_len;
7153 
7154 	adapter->bd_number = cards_found++;
7155 
7156 	e1000e_check_options(adapter);
7157 
7158 	/* setup adapter struct */
7159 	err = e1000_sw_init(adapter);
7160 	if (err)
7161 		goto err_sw_init;
7162 
7163 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
7164 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
7165 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
7166 
7167 	err = ei->get_variants(adapter);
7168 	if (err)
7169 		goto err_hw_init;
7170 
7171 	if ((adapter->flags & FLAG_IS_ICH) &&
7172 	    (adapter->flags & FLAG_READ_ONLY_NVM) &&
7173 	    (hw->mac.type < e1000_pch_spt))
7174 		e1000e_write_protect_nvm_ich8lan(&adapter->hw);
7175 
7176 	hw->mac.ops.get_bus_info(&adapter->hw);
7177 
7178 	adapter->hw.phy.autoneg_wait_to_complete = 0;
7179 
7180 	/* Copper options */
7181 	if (adapter->hw.phy.media_type == e1000_media_type_copper) {
7182 		adapter->hw.phy.mdix = AUTO_ALL_MODES;
7183 		adapter->hw.phy.disable_polarity_correction = 0;
7184 		adapter->hw.phy.ms_type = e1000_ms_hw_default;
7185 	}
7186 
7187 	if (hw->phy.ops.check_reset_block && hw->phy.ops.check_reset_block(hw))
7188 		dev_info(&pdev->dev,
7189 			 "PHY reset is blocked due to SOL/IDER session.\n");
7190 
7191 	/* Set initial default active device features */
7192 	netdev->features = (NETIF_F_SG |
7193 			    NETIF_F_HW_VLAN_CTAG_RX |
7194 			    NETIF_F_HW_VLAN_CTAG_TX |
7195 			    NETIF_F_TSO |
7196 			    NETIF_F_TSO6 |
7197 			    NETIF_F_RXHASH |
7198 			    NETIF_F_RXCSUM |
7199 			    NETIF_F_HW_CSUM);
7200 
7201 	/* Set user-changeable features (subset of all device features) */
7202 	netdev->hw_features = netdev->features;
7203 	netdev->hw_features |= NETIF_F_RXFCS;
7204 	netdev->priv_flags |= IFF_SUPP_NOFCS;
7205 	netdev->hw_features |= NETIF_F_RXALL;
7206 
7207 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER)
7208 		netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
7209 
7210 	netdev->vlan_features |= (NETIF_F_SG |
7211 				  NETIF_F_TSO |
7212 				  NETIF_F_TSO6 |
7213 				  NETIF_F_HW_CSUM);
7214 
7215 	netdev->priv_flags |= IFF_UNICAST_FLT;
7216 
7217 	if (pci_using_dac) {
7218 		netdev->features |= NETIF_F_HIGHDMA;
7219 		netdev->vlan_features |= NETIF_F_HIGHDMA;
7220 	}
7221 
7222 	/* MTU range: 68 - max_hw_frame_size */
7223 	netdev->min_mtu = ETH_MIN_MTU;
7224 	netdev->max_mtu = adapter->max_hw_frame_size -
7225 			  (VLAN_ETH_HLEN + ETH_FCS_LEN);
7226 
7227 	if (e1000e_enable_mng_pass_thru(&adapter->hw))
7228 		adapter->flags |= FLAG_MNG_PT_ENABLED;
7229 
7230 	/* before reading the NVM, reset the controller to
7231 	 * put the device in a known good starting state
7232 	 */
7233 	adapter->hw.mac.ops.reset_hw(&adapter->hw);
7234 
7235 	/* systems with ASPM and others may see the checksum fail on the first
7236 	 * attempt. Let's give it a few tries
7237 	 */
7238 	for (i = 0;; i++) {
7239 		if (e1000_validate_nvm_checksum(&adapter->hw) >= 0)
7240 			break;
7241 		if (i == 2) {
7242 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
7243 			err = -EIO;
7244 			goto err_eeprom;
7245 		}
7246 	}
7247 
7248 	e1000_eeprom_checks(adapter);
7249 
7250 	/* copy the MAC address */
7251 	if (e1000e_read_mac_addr(&adapter->hw))
7252 		dev_err(&pdev->dev,
7253 			"NVM Read Error while reading MAC address\n");
7254 
7255 	memcpy(netdev->dev_addr, adapter->hw.mac.addr, netdev->addr_len);
7256 
7257 	if (!is_valid_ether_addr(netdev->dev_addr)) {
7258 		dev_err(&pdev->dev, "Invalid MAC Address: %pM\n",
7259 			netdev->dev_addr);
7260 		err = -EIO;
7261 		goto err_eeprom;
7262 	}
7263 
7264 	timer_setup(&adapter->watchdog_timer, e1000_watchdog, 0);
7265 	timer_setup(&adapter->phy_info_timer, e1000_update_phy_info, 0);
7266 
7267 	INIT_WORK(&adapter->reset_task, e1000_reset_task);
7268 	INIT_WORK(&adapter->watchdog_task, e1000_watchdog_task);
7269 	INIT_WORK(&adapter->downshift_task, e1000e_downshift_workaround);
7270 	INIT_WORK(&adapter->update_phy_task, e1000e_update_phy_task);
7271 	INIT_WORK(&adapter->print_hang_task, e1000_print_hw_hang);
7272 
7273 	/* Initialize link parameters. User can change them with ethtool */
7274 	adapter->hw.mac.autoneg = 1;
7275 	adapter->fc_autoneg = true;
7276 	adapter->hw.fc.requested_mode = e1000_fc_default;
7277 	adapter->hw.fc.current_mode = e1000_fc_default;
7278 	adapter->hw.phy.autoneg_advertised = 0x2f;
7279 
7280 	/* Initial Wake on LAN setting - If APM wake is enabled in
7281 	 * the EEPROM, enable the ACPI Magic Packet filter
7282 	 */
7283 	if (adapter->flags & FLAG_APME_IN_WUC) {
7284 		/* APME bit in EEPROM is mapped to WUC.APME */
7285 		eeprom_data = er32(WUC);
7286 		eeprom_apme_mask = E1000_WUC_APME;
7287 		if ((hw->mac.type > e1000_ich10lan) &&
7288 		    (eeprom_data & E1000_WUC_PHY_WAKE))
7289 			adapter->flags2 |= FLAG2_HAS_PHY_WAKEUP;
7290 	} else if (adapter->flags & FLAG_APME_IN_CTRL3) {
7291 		if (adapter->flags & FLAG_APME_CHECK_PORT_B &&
7292 		    (adapter->hw.bus.func == 1))
7293 			ret_val = e1000_read_nvm(&adapter->hw,
7294 					      NVM_INIT_CONTROL3_PORT_B,
7295 					      1, &eeprom_data);
7296 		else
7297 			ret_val = e1000_read_nvm(&adapter->hw,
7298 					      NVM_INIT_CONTROL3_PORT_A,
7299 					      1, &eeprom_data);
7300 	}
7301 
7302 	/* fetch WoL from EEPROM */
7303 	if (ret_val)
7304 		e_dbg("NVM read error getting WoL initial values: %d\n", ret_val);
7305 	else if (eeprom_data & eeprom_apme_mask)
7306 		adapter->eeprom_wol |= E1000_WUFC_MAG;
7307 
7308 	/* now that we have the eeprom settings, apply the special cases
7309 	 * where the eeprom may be wrong or the board simply won't support
7310 	 * wake on lan on a particular port
7311 	 */
7312 	if (!(adapter->flags & FLAG_HAS_WOL))
7313 		adapter->eeprom_wol = 0;
7314 
7315 	/* initialize the wol settings based on the eeprom settings */
7316 	adapter->wol = adapter->eeprom_wol;
7317 
7318 	/* make sure adapter isn't asleep if manageability is enabled */
7319 	if (adapter->wol || (adapter->flags & FLAG_MNG_PT_ENABLED) ||
7320 	    (hw->mac.ops.check_mng_mode(hw)))
7321 		device_wakeup_enable(&pdev->dev);
7322 
7323 	/* save off EEPROM version number */
7324 	ret_val = e1000_read_nvm(&adapter->hw, 5, 1, &adapter->eeprom_vers);
7325 
7326 	if (ret_val) {
7327 		e_dbg("NVM read error getting EEPROM version: %d\n", ret_val);
7328 		adapter->eeprom_vers = 0;
7329 	}
7330 
7331 	/* init PTP hardware clock */
7332 	e1000e_ptp_init(adapter);
7333 
7334 	/* reset the hardware with the new settings */
7335 	e1000e_reset(adapter);
7336 
7337 	/* If the controller has AMT, do not set DRV_LOAD until the interface
7338 	 * is up.  For all other cases, let the f/w know that the h/w is now
7339 	 * under the control of the driver.
7340 	 */
7341 	if (!(adapter->flags & FLAG_HAS_AMT))
7342 		e1000e_get_hw_control(adapter);
7343 
7344 	strlcpy(netdev->name, "eth%d", sizeof(netdev->name));
7345 	err = register_netdev(netdev);
7346 	if (err)
7347 		goto err_register;
7348 
7349 	/* carrier off reporting is important to ethtool even BEFORE open */
7350 	netif_carrier_off(netdev);
7351 
7352 	e1000_print_device_info(adapter);
7353 
7354 	if (pci_dev_run_wake(pdev))
7355 		pm_runtime_put_noidle(&pdev->dev);
7356 
7357 	return 0;
7358 
7359 err_register:
7360 	if (!(adapter->flags & FLAG_HAS_AMT))
7361 		e1000e_release_hw_control(adapter);
7362 err_eeprom:
7363 	if (hw->phy.ops.check_reset_block && !hw->phy.ops.check_reset_block(hw))
7364 		e1000_phy_hw_reset(&adapter->hw);
7365 err_hw_init:
7366 	kfree(adapter->tx_ring);
7367 	kfree(adapter->rx_ring);
7368 err_sw_init:
7369 	if ((adapter->hw.flash_address) && (hw->mac.type < e1000_pch_spt))
7370 		iounmap(adapter->hw.flash_address);
7371 	e1000e_reset_interrupt_capability(adapter);
7372 err_flashmap:
7373 	iounmap(adapter->hw.hw_addr);
7374 err_ioremap:
7375 	free_netdev(netdev);
7376 err_alloc_etherdev:
7377 	pci_release_mem_regions(pdev);
7378 err_pci_reg:
7379 err_dma:
7380 	pci_disable_device(pdev);
7381 	return err;
7382 }
7383 
7384 /**
7385  * e1000_remove - Device Removal Routine
7386  * @pdev: PCI device information struct
7387  *
7388  * e1000_remove is called by the PCI subsystem to alert the driver
7389  * that it should release a PCI device.  The could be caused by a
7390  * Hot-Plug event, or because the driver is going to be removed from
7391  * memory.
7392  **/
7393 static void e1000_remove(struct pci_dev *pdev)
7394 {
7395 	struct net_device *netdev = pci_get_drvdata(pdev);
7396 	struct e1000_adapter *adapter = netdev_priv(netdev);
7397 	bool down = test_bit(__E1000_DOWN, &adapter->state);
7398 
7399 	e1000e_ptp_remove(adapter);
7400 
7401 	/* The timers may be rescheduled, so explicitly disable them
7402 	 * from being rescheduled.
7403 	 */
7404 	if (!down)
7405 		set_bit(__E1000_DOWN, &adapter->state);
7406 	del_timer_sync(&adapter->watchdog_timer);
7407 	del_timer_sync(&adapter->phy_info_timer);
7408 
7409 	cancel_work_sync(&adapter->reset_task);
7410 	cancel_work_sync(&adapter->watchdog_task);
7411 	cancel_work_sync(&adapter->downshift_task);
7412 	cancel_work_sync(&adapter->update_phy_task);
7413 	cancel_work_sync(&adapter->print_hang_task);
7414 
7415 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
7416 		cancel_work_sync(&adapter->tx_hwtstamp_work);
7417 		if (adapter->tx_hwtstamp_skb) {
7418 			dev_consume_skb_any(adapter->tx_hwtstamp_skb);
7419 			adapter->tx_hwtstamp_skb = NULL;
7420 		}
7421 	}
7422 
7423 	/* Don't lie to e1000_close() down the road. */
7424 	if (!down)
7425 		clear_bit(__E1000_DOWN, &adapter->state);
7426 	unregister_netdev(netdev);
7427 
7428 	if (pci_dev_run_wake(pdev))
7429 		pm_runtime_get_noresume(&pdev->dev);
7430 
7431 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
7432 	 * would have already happened in close and is redundant.
7433 	 */
7434 	e1000e_release_hw_control(adapter);
7435 
7436 	e1000e_reset_interrupt_capability(adapter);
7437 	kfree(adapter->tx_ring);
7438 	kfree(adapter->rx_ring);
7439 
7440 	iounmap(adapter->hw.hw_addr);
7441 	if ((adapter->hw.flash_address) &&
7442 	    (adapter->hw.mac.type < e1000_pch_spt))
7443 		iounmap(adapter->hw.flash_address);
7444 	pci_release_mem_regions(pdev);
7445 
7446 	free_netdev(netdev);
7447 
7448 	/* AER disable */
7449 	pci_disable_pcie_error_reporting(pdev);
7450 
7451 	pci_disable_device(pdev);
7452 }
7453 
7454 /* PCI Error Recovery (ERS) */
7455 static const struct pci_error_handlers e1000_err_handler = {
7456 	.error_detected = e1000_io_error_detected,
7457 	.slot_reset = e1000_io_slot_reset,
7458 	.resume = e1000_io_resume,
7459 };
7460 
7461 static const struct pci_device_id e1000_pci_tbl[] = {
7462 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_COPPER), board_82571 },
7463 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_FIBER), board_82571 },
7464 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER), board_82571 },
7465 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER_LP),
7466 	  board_82571 },
7467 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_FIBER), board_82571 },
7468 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES), board_82571 },
7469 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_DUAL), board_82571 },
7470 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_QUAD), board_82571 },
7471 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571PT_QUAD_COPPER), board_82571 },
7472 
7473 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI), board_82572 },
7474 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_COPPER), board_82572 },
7475 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_FIBER), board_82572 },
7476 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_SERDES), board_82572 },
7477 
7478 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E), board_82573 },
7479 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E_IAMT), board_82573 },
7480 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573L), board_82573 },
7481 
7482 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574L), board_82574 },
7483 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574LA), board_82574 },
7484 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82583V), board_82583 },
7485 
7486 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_DPT),
7487 	  board_80003es2lan },
7488 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_SPT),
7489 	  board_80003es2lan },
7490 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_DPT),
7491 	  board_80003es2lan },
7492 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_SPT),
7493 	  board_80003es2lan },
7494 
7495 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE), board_ich8lan },
7496 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_G), board_ich8lan },
7497 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_GT), board_ich8lan },
7498 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_AMT), board_ich8lan },
7499 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_C), board_ich8lan },
7500 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M), board_ich8lan },
7501 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M_AMT), board_ich8lan },
7502 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_82567V_3), board_ich8lan },
7503 
7504 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE), board_ich9lan },
7505 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_G), board_ich9lan },
7506 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_GT), board_ich9lan },
7507 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_AMT), board_ich9lan },
7508 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_C), board_ich9lan },
7509 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_BM), board_ich9lan },
7510 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M), board_ich9lan },
7511 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_AMT), board_ich9lan },
7512 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_V), board_ich9lan },
7513 
7514 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LM), board_ich9lan },
7515 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LF), board_ich9lan },
7516 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_V), board_ich9lan },
7517 
7518 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LM), board_ich10lan },
7519 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LF), board_ich10lan },
7520 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_V), board_ich10lan },
7521 
7522 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LM), board_pchlan },
7523 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LC), board_pchlan },
7524 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DM), board_pchlan },
7525 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DC), board_pchlan },
7526 
7527 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_LM), board_pch2lan },
7528 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_V), board_pch2lan },
7529 
7530 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_LM), board_pch_lpt },
7531 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_V), board_pch_lpt },
7532 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_LM), board_pch_lpt },
7533 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_V), board_pch_lpt },
7534 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM2), board_pch_lpt },
7535 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V2), board_pch_lpt },
7536 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM3), board_pch_lpt },
7537 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V3), board_pch_lpt },
7538 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM), board_pch_spt },
7539 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V), board_pch_spt },
7540 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM2), board_pch_spt },
7541 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V2), board_pch_spt },
7542 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LBG_I219_LM3), board_pch_spt },
7543 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM4), board_pch_spt },
7544 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V4), board_pch_spt },
7545 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM5), board_pch_spt },
7546 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V5), board_pch_spt },
7547 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM6), board_pch_cnp },
7548 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V6), board_pch_cnp },
7549 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM7), board_pch_cnp },
7550 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V7), board_pch_cnp },
7551 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM8), board_pch_cnp },
7552 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V8), board_pch_cnp },
7553 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM9), board_pch_cnp },
7554 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V9), board_pch_cnp },
7555 
7556 	{ 0, 0, 0, 0, 0, 0, 0 }	/* terminate list */
7557 };
7558 MODULE_DEVICE_TABLE(pci, e1000_pci_tbl);
7559 
7560 static const struct dev_pm_ops e1000_pm_ops = {
7561 #ifdef CONFIG_PM_SLEEP
7562 	.suspend	= e1000e_pm_suspend,
7563 	.resume		= e1000e_pm_resume,
7564 	.freeze		= e1000e_pm_freeze,
7565 	.thaw		= e1000e_pm_thaw,
7566 	.poweroff	= e1000e_pm_suspend,
7567 	.restore	= e1000e_pm_resume,
7568 #endif
7569 	SET_RUNTIME_PM_OPS(e1000e_pm_runtime_suspend, e1000e_pm_runtime_resume,
7570 			   e1000e_pm_runtime_idle)
7571 };
7572 
7573 /* PCI Device API Driver */
7574 static struct pci_driver e1000_driver = {
7575 	.name     = e1000e_driver_name,
7576 	.id_table = e1000_pci_tbl,
7577 	.probe    = e1000_probe,
7578 	.remove   = e1000_remove,
7579 	.driver   = {
7580 		.pm = &e1000_pm_ops,
7581 	},
7582 	.shutdown = e1000_shutdown,
7583 	.err_handler = &e1000_err_handler
7584 };
7585 
7586 /**
7587  * e1000_init_module - Driver Registration Routine
7588  *
7589  * e1000_init_module is the first routine called when the driver is
7590  * loaded. All it does is register with the PCI subsystem.
7591  **/
7592 static int __init e1000_init_module(void)
7593 {
7594 	pr_info("Intel(R) PRO/1000 Network Driver - %s\n",
7595 		e1000e_driver_version);
7596 	pr_info("Copyright(c) 1999 - 2015 Intel Corporation.\n");
7597 
7598 	return pci_register_driver(&e1000_driver);
7599 }
7600 module_init(e1000_init_module);
7601 
7602 /**
7603  * e1000_exit_module - Driver Exit Cleanup Routine
7604  *
7605  * e1000_exit_module is called just before the driver is removed
7606  * from memory.
7607  **/
7608 static void __exit e1000_exit_module(void)
7609 {
7610 	pci_unregister_driver(&e1000_driver);
7611 }
7612 module_exit(e1000_exit_module);
7613 
7614 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
7615 MODULE_DESCRIPTION("Intel(R) PRO/1000 Network Driver");
7616 MODULE_LICENSE("GPL");
7617 MODULE_VERSION(DRV_VERSION);
7618 
7619 /* netdev.c */
7620