xref: /openbmc/linux/drivers/net/ethernet/intel/e1000e/netdev.c (revision 5ef12cb4a3a78ffb331c03a795a15eea4ae35155)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Intel PRO/1000 Linux driver
3  * Copyright(c) 1999 - 2015 Intel Corporation.
4  *
5  * This program is free software; you can redistribute it and/or modify it
6  * under the terms and conditions of the GNU General Public License,
7  * version 2, as published by the Free Software Foundation.
8  *
9  * This program is distributed in the hope it will be useful, but WITHOUT
10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
12  * more details.
13  *
14  * The full GNU General Public License is included in this distribution in
15  * the file called "COPYING".
16  *
17  * Contact Information:
18  * Linux NICS <linux.nics@intel.com>
19  * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
20  * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
21  */
22 
23 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
24 
25 #include <linux/module.h>
26 #include <linux/types.h>
27 #include <linux/init.h>
28 #include <linux/pci.h>
29 #include <linux/vmalloc.h>
30 #include <linux/pagemap.h>
31 #include <linux/delay.h>
32 #include <linux/netdevice.h>
33 #include <linux/interrupt.h>
34 #include <linux/tcp.h>
35 #include <linux/ipv6.h>
36 #include <linux/slab.h>
37 #include <net/checksum.h>
38 #include <net/ip6_checksum.h>
39 #include <linux/ethtool.h>
40 #include <linux/if_vlan.h>
41 #include <linux/cpu.h>
42 #include <linux/smp.h>
43 #include <linux/pm_qos.h>
44 #include <linux/pm_runtime.h>
45 #include <linux/aer.h>
46 #include <linux/prefetch.h>
47 
48 #include "e1000.h"
49 
50 #define DRV_EXTRAVERSION "-k"
51 
52 #define DRV_VERSION "3.2.6" DRV_EXTRAVERSION
53 char e1000e_driver_name[] = "e1000e";
54 const char e1000e_driver_version[] = DRV_VERSION;
55 
56 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
57 static int debug = -1;
58 module_param(debug, int, 0);
59 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
60 
61 static const struct e1000_info *e1000_info_tbl[] = {
62 	[board_82571]		= &e1000_82571_info,
63 	[board_82572]		= &e1000_82572_info,
64 	[board_82573]		= &e1000_82573_info,
65 	[board_82574]		= &e1000_82574_info,
66 	[board_82583]		= &e1000_82583_info,
67 	[board_80003es2lan]	= &e1000_es2_info,
68 	[board_ich8lan]		= &e1000_ich8_info,
69 	[board_ich9lan]		= &e1000_ich9_info,
70 	[board_ich10lan]	= &e1000_ich10_info,
71 	[board_pchlan]		= &e1000_pch_info,
72 	[board_pch2lan]		= &e1000_pch2_info,
73 	[board_pch_lpt]		= &e1000_pch_lpt_info,
74 	[board_pch_spt]		= &e1000_pch_spt_info,
75 	[board_pch_cnp]		= &e1000_pch_cnp_info,
76 };
77 
78 struct e1000_reg_info {
79 	u32 ofs;
80 	char *name;
81 };
82 
83 static const struct e1000_reg_info e1000_reg_info_tbl[] = {
84 	/* General Registers */
85 	{E1000_CTRL, "CTRL"},
86 	{E1000_STATUS, "STATUS"},
87 	{E1000_CTRL_EXT, "CTRL_EXT"},
88 
89 	/* Interrupt Registers */
90 	{E1000_ICR, "ICR"},
91 
92 	/* Rx Registers */
93 	{E1000_RCTL, "RCTL"},
94 	{E1000_RDLEN(0), "RDLEN"},
95 	{E1000_RDH(0), "RDH"},
96 	{E1000_RDT(0), "RDT"},
97 	{E1000_RDTR, "RDTR"},
98 	{E1000_RXDCTL(0), "RXDCTL"},
99 	{E1000_ERT, "ERT"},
100 	{E1000_RDBAL(0), "RDBAL"},
101 	{E1000_RDBAH(0), "RDBAH"},
102 	{E1000_RDFH, "RDFH"},
103 	{E1000_RDFT, "RDFT"},
104 	{E1000_RDFHS, "RDFHS"},
105 	{E1000_RDFTS, "RDFTS"},
106 	{E1000_RDFPC, "RDFPC"},
107 
108 	/* Tx Registers */
109 	{E1000_TCTL, "TCTL"},
110 	{E1000_TDBAL(0), "TDBAL"},
111 	{E1000_TDBAH(0), "TDBAH"},
112 	{E1000_TDLEN(0), "TDLEN"},
113 	{E1000_TDH(0), "TDH"},
114 	{E1000_TDT(0), "TDT"},
115 	{E1000_TIDV, "TIDV"},
116 	{E1000_TXDCTL(0), "TXDCTL"},
117 	{E1000_TADV, "TADV"},
118 	{E1000_TARC(0), "TARC"},
119 	{E1000_TDFH, "TDFH"},
120 	{E1000_TDFT, "TDFT"},
121 	{E1000_TDFHS, "TDFHS"},
122 	{E1000_TDFTS, "TDFTS"},
123 	{E1000_TDFPC, "TDFPC"},
124 
125 	/* List Terminator */
126 	{0, NULL}
127 };
128 
129 /**
130  * __ew32_prepare - prepare to write to MAC CSR register on certain parts
131  * @hw: pointer to the HW structure
132  *
133  * When updating the MAC CSR registers, the Manageability Engine (ME) could
134  * be accessing the registers at the same time.  Normally, this is handled in
135  * h/w by an arbiter but on some parts there is a bug that acknowledges Host
136  * accesses later than it should which could result in the register to have
137  * an incorrect value.  Workaround this by checking the FWSM register which
138  * has bit 24 set while ME is accessing MAC CSR registers, wait if it is set
139  * and try again a number of times.
140  **/
141 s32 __ew32_prepare(struct e1000_hw *hw)
142 {
143 	s32 i = E1000_ICH_FWSM_PCIM2PCI_COUNT;
144 
145 	while ((er32(FWSM) & E1000_ICH_FWSM_PCIM2PCI) && --i)
146 		udelay(50);
147 
148 	return i;
149 }
150 
151 void __ew32(struct e1000_hw *hw, unsigned long reg, u32 val)
152 {
153 	if (hw->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
154 		__ew32_prepare(hw);
155 
156 	writel(val, hw->hw_addr + reg);
157 }
158 
159 /**
160  * e1000_regdump - register printout routine
161  * @hw: pointer to the HW structure
162  * @reginfo: pointer to the register info table
163  **/
164 static void e1000_regdump(struct e1000_hw *hw, struct e1000_reg_info *reginfo)
165 {
166 	int n = 0;
167 	char rname[16];
168 	u32 regs[8];
169 
170 	switch (reginfo->ofs) {
171 	case E1000_RXDCTL(0):
172 		for (n = 0; n < 2; n++)
173 			regs[n] = __er32(hw, E1000_RXDCTL(n));
174 		break;
175 	case E1000_TXDCTL(0):
176 		for (n = 0; n < 2; n++)
177 			regs[n] = __er32(hw, E1000_TXDCTL(n));
178 		break;
179 	case E1000_TARC(0):
180 		for (n = 0; n < 2; n++)
181 			regs[n] = __er32(hw, E1000_TARC(n));
182 		break;
183 	default:
184 		pr_info("%-15s %08x\n",
185 			reginfo->name, __er32(hw, reginfo->ofs));
186 		return;
187 	}
188 
189 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-1]");
190 	pr_info("%-15s %08x %08x\n", rname, regs[0], regs[1]);
191 }
192 
193 static void e1000e_dump_ps_pages(struct e1000_adapter *adapter,
194 				 struct e1000_buffer *bi)
195 {
196 	int i;
197 	struct e1000_ps_page *ps_page;
198 
199 	for (i = 0; i < adapter->rx_ps_pages; i++) {
200 		ps_page = &bi->ps_pages[i];
201 
202 		if (ps_page->page) {
203 			pr_info("packet dump for ps_page %d:\n", i);
204 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
205 				       16, 1, page_address(ps_page->page),
206 				       PAGE_SIZE, true);
207 		}
208 	}
209 }
210 
211 /**
212  * e1000e_dump - Print registers, Tx-ring and Rx-ring
213  * @adapter: board private structure
214  **/
215 static void e1000e_dump(struct e1000_adapter *adapter)
216 {
217 	struct net_device *netdev = adapter->netdev;
218 	struct e1000_hw *hw = &adapter->hw;
219 	struct e1000_reg_info *reginfo;
220 	struct e1000_ring *tx_ring = adapter->tx_ring;
221 	struct e1000_tx_desc *tx_desc;
222 	struct my_u0 {
223 		__le64 a;
224 		__le64 b;
225 	} *u0;
226 	struct e1000_buffer *buffer_info;
227 	struct e1000_ring *rx_ring = adapter->rx_ring;
228 	union e1000_rx_desc_packet_split *rx_desc_ps;
229 	union e1000_rx_desc_extended *rx_desc;
230 	struct my_u1 {
231 		__le64 a;
232 		__le64 b;
233 		__le64 c;
234 		__le64 d;
235 	} *u1;
236 	u32 staterr;
237 	int i = 0;
238 
239 	if (!netif_msg_hw(adapter))
240 		return;
241 
242 	/* Print netdevice Info */
243 	if (netdev) {
244 		dev_info(&adapter->pdev->dev, "Net device Info\n");
245 		pr_info("Device Name     state            trans_start\n");
246 		pr_info("%-15s %016lX %016lX\n", netdev->name,
247 			netdev->state, dev_trans_start(netdev));
248 	}
249 
250 	/* Print Registers */
251 	dev_info(&adapter->pdev->dev, "Register Dump\n");
252 	pr_info(" Register Name   Value\n");
253 	for (reginfo = (struct e1000_reg_info *)e1000_reg_info_tbl;
254 	     reginfo->name; reginfo++) {
255 		e1000_regdump(hw, reginfo);
256 	}
257 
258 	/* Print Tx Ring Summary */
259 	if (!netdev || !netif_running(netdev))
260 		return;
261 
262 	dev_info(&adapter->pdev->dev, "Tx Ring Summary\n");
263 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
264 	buffer_info = &tx_ring->buffer_info[tx_ring->next_to_clean];
265 	pr_info(" %5d %5X %5X %016llX %04X %3X %016llX\n",
266 		0, tx_ring->next_to_use, tx_ring->next_to_clean,
267 		(unsigned long long)buffer_info->dma,
268 		buffer_info->length,
269 		buffer_info->next_to_watch,
270 		(unsigned long long)buffer_info->time_stamp);
271 
272 	/* Print Tx Ring */
273 	if (!netif_msg_tx_done(adapter))
274 		goto rx_ring_summary;
275 
276 	dev_info(&adapter->pdev->dev, "Tx Ring Dump\n");
277 
278 	/* Transmit Descriptor Formats - DEXT[29] is 0 (Legacy) or 1 (Extended)
279 	 *
280 	 * Legacy Transmit Descriptor
281 	 *   +--------------------------------------------------------------+
282 	 * 0 |         Buffer Address [63:0] (Reserved on Write Back)       |
283 	 *   +--------------------------------------------------------------+
284 	 * 8 | Special  |    CSS     | Status |  CMD    |  CSO   |  Length  |
285 	 *   +--------------------------------------------------------------+
286 	 *   63       48 47        36 35    32 31     24 23    16 15        0
287 	 *
288 	 * Extended Context Descriptor (DTYP=0x0) for TSO or checksum offload
289 	 *   63      48 47    40 39       32 31             16 15    8 7      0
290 	 *   +----------------------------------------------------------------+
291 	 * 0 |  TUCSE  | TUCS0  |   TUCSS   |     IPCSE       | IPCS0 | IPCSS |
292 	 *   +----------------------------------------------------------------+
293 	 * 8 |   MSS   | HDRLEN | RSV | STA | TUCMD | DTYP |      PAYLEN      |
294 	 *   +----------------------------------------------------------------+
295 	 *   63      48 47    40 39 36 35 32 31   24 23  20 19                0
296 	 *
297 	 * Extended Data Descriptor (DTYP=0x1)
298 	 *   +----------------------------------------------------------------+
299 	 * 0 |                     Buffer Address [63:0]                      |
300 	 *   +----------------------------------------------------------------+
301 	 * 8 | VLAN tag |  POPTS  | Rsvd | Status | Command | DTYP |  DTALEN  |
302 	 *   +----------------------------------------------------------------+
303 	 *   63       48 47     40 39  36 35    32 31     24 23  20 19        0
304 	 */
305 	pr_info("Tl[desc]     [address 63:0  ] [SpeCssSCmCsLen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Legacy format\n");
306 	pr_info("Tc[desc]     [Ce CoCsIpceCoS] [MssHlRSCm0Plen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Context format\n");
307 	pr_info("Td[desc]     [address 63:0  ] [VlaPoRSCm1Dlen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Data format\n");
308 	for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
309 		const char *next_desc;
310 		tx_desc = E1000_TX_DESC(*tx_ring, i);
311 		buffer_info = &tx_ring->buffer_info[i];
312 		u0 = (struct my_u0 *)tx_desc;
313 		if (i == tx_ring->next_to_use && i == tx_ring->next_to_clean)
314 			next_desc = " NTC/U";
315 		else if (i == tx_ring->next_to_use)
316 			next_desc = " NTU";
317 		else if (i == tx_ring->next_to_clean)
318 			next_desc = " NTC";
319 		else
320 			next_desc = "";
321 		pr_info("T%c[0x%03X]    %016llX %016llX %016llX %04X  %3X %016llX %p%s\n",
322 			(!(le64_to_cpu(u0->b) & BIT(29)) ? 'l' :
323 			 ((le64_to_cpu(u0->b) & BIT(20)) ? 'd' : 'c')),
324 			i,
325 			(unsigned long long)le64_to_cpu(u0->a),
326 			(unsigned long long)le64_to_cpu(u0->b),
327 			(unsigned long long)buffer_info->dma,
328 			buffer_info->length, buffer_info->next_to_watch,
329 			(unsigned long long)buffer_info->time_stamp,
330 			buffer_info->skb, next_desc);
331 
332 		if (netif_msg_pktdata(adapter) && buffer_info->skb)
333 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
334 				       16, 1, buffer_info->skb->data,
335 				       buffer_info->skb->len, true);
336 	}
337 
338 	/* Print Rx Ring Summary */
339 rx_ring_summary:
340 	dev_info(&adapter->pdev->dev, "Rx Ring Summary\n");
341 	pr_info("Queue [NTU] [NTC]\n");
342 	pr_info(" %5d %5X %5X\n",
343 		0, rx_ring->next_to_use, rx_ring->next_to_clean);
344 
345 	/* Print Rx Ring */
346 	if (!netif_msg_rx_status(adapter))
347 		return;
348 
349 	dev_info(&adapter->pdev->dev, "Rx Ring Dump\n");
350 	switch (adapter->rx_ps_pages) {
351 	case 1:
352 	case 2:
353 	case 3:
354 		/* [Extended] Packet Split Receive Descriptor Format
355 		 *
356 		 *    +-----------------------------------------------------+
357 		 *  0 |                Buffer Address 0 [63:0]              |
358 		 *    +-----------------------------------------------------+
359 		 *  8 |                Buffer Address 1 [63:0]              |
360 		 *    +-----------------------------------------------------+
361 		 * 16 |                Buffer Address 2 [63:0]              |
362 		 *    +-----------------------------------------------------+
363 		 * 24 |                Buffer Address 3 [63:0]              |
364 		 *    +-----------------------------------------------------+
365 		 */
366 		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");
367 		/* [Extended] Receive Descriptor (Write-Back) Format
368 		 *
369 		 *   63       48 47    32 31     13 12    8 7    4 3        0
370 		 *   +------------------------------------------------------+
371 		 * 0 | Packet   | IP     |  Rsvd   | MRQ   | Rsvd | MRQ RSS |
372 		 *   | Checksum | Ident  |         | Queue |      |  Type   |
373 		 *   +------------------------------------------------------+
374 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
375 		 *   +------------------------------------------------------+
376 		 *   63       48 47    32 31            20 19               0
377 		 */
378 		pr_info("RWB[desc]      [ck ipid mrqhsh] [vl   l0 ee  es] [ l3  l2  l1 hs] [reserved      ] ---------------- [bi->skb] <-- Ext Rx Write-Back format\n");
379 		for (i = 0; i < rx_ring->count; i++) {
380 			const char *next_desc;
381 			buffer_info = &rx_ring->buffer_info[i];
382 			rx_desc_ps = E1000_RX_DESC_PS(*rx_ring, i);
383 			u1 = (struct my_u1 *)rx_desc_ps;
384 			staterr =
385 			    le32_to_cpu(rx_desc_ps->wb.middle.status_error);
386 
387 			if (i == rx_ring->next_to_use)
388 				next_desc = " NTU";
389 			else if (i == rx_ring->next_to_clean)
390 				next_desc = " NTC";
391 			else
392 				next_desc = "";
393 
394 			if (staterr & E1000_RXD_STAT_DD) {
395 				/* Descriptor Done */
396 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX ---------------- %p%s\n",
397 					"RWB", i,
398 					(unsigned long long)le64_to_cpu(u1->a),
399 					(unsigned long long)le64_to_cpu(u1->b),
400 					(unsigned long long)le64_to_cpu(u1->c),
401 					(unsigned long long)le64_to_cpu(u1->d),
402 					buffer_info->skb, next_desc);
403 			} else {
404 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX %016llX %p%s\n",
405 					"R  ", i,
406 					(unsigned long long)le64_to_cpu(u1->a),
407 					(unsigned long long)le64_to_cpu(u1->b),
408 					(unsigned long long)le64_to_cpu(u1->c),
409 					(unsigned long long)le64_to_cpu(u1->d),
410 					(unsigned long long)buffer_info->dma,
411 					buffer_info->skb, next_desc);
412 
413 				if (netif_msg_pktdata(adapter))
414 					e1000e_dump_ps_pages(adapter,
415 							     buffer_info);
416 			}
417 		}
418 		break;
419 	default:
420 	case 0:
421 		/* Extended Receive Descriptor (Read) Format
422 		 *
423 		 *   +-----------------------------------------------------+
424 		 * 0 |                Buffer Address [63:0]                |
425 		 *   +-----------------------------------------------------+
426 		 * 8 |                      Reserved                       |
427 		 *   +-----------------------------------------------------+
428 		 */
429 		pr_info("R  [desc]      [buf addr 63:0 ] [reserved 63:0 ] [bi->dma       ] [bi->skb] <-- Ext (Read) format\n");
430 		/* Extended Receive Descriptor (Write-Back) Format
431 		 *
432 		 *   63       48 47    32 31    24 23            4 3        0
433 		 *   +------------------------------------------------------+
434 		 *   |     RSS Hash      |        |               |         |
435 		 * 0 +-------------------+  Rsvd  |   Reserved    | MRQ RSS |
436 		 *   | Packet   | IP     |        |               |  Type   |
437 		 *   | Checksum | Ident  |        |               |         |
438 		 *   +------------------------------------------------------+
439 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
440 		 *   +------------------------------------------------------+
441 		 *   63       48 47    32 31            20 19               0
442 		 */
443 		pr_info("RWB[desc]      [cs ipid    mrq] [vt   ln xe  xs] [bi->skb] <-- Ext (Write-Back) format\n");
444 
445 		for (i = 0; i < rx_ring->count; i++) {
446 			const char *next_desc;
447 
448 			buffer_info = &rx_ring->buffer_info[i];
449 			rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
450 			u1 = (struct my_u1 *)rx_desc;
451 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
452 
453 			if (i == rx_ring->next_to_use)
454 				next_desc = " NTU";
455 			else if (i == rx_ring->next_to_clean)
456 				next_desc = " NTC";
457 			else
458 				next_desc = "";
459 
460 			if (staterr & E1000_RXD_STAT_DD) {
461 				/* Descriptor Done */
462 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %p%s\n",
463 					"RWB", i,
464 					(unsigned long long)le64_to_cpu(u1->a),
465 					(unsigned long long)le64_to_cpu(u1->b),
466 					buffer_info->skb, next_desc);
467 			} else {
468 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %p%s\n",
469 					"R  ", i,
470 					(unsigned long long)le64_to_cpu(u1->a),
471 					(unsigned long long)le64_to_cpu(u1->b),
472 					(unsigned long long)buffer_info->dma,
473 					buffer_info->skb, next_desc);
474 
475 				if (netif_msg_pktdata(adapter) &&
476 				    buffer_info->skb)
477 					print_hex_dump(KERN_INFO, "",
478 						       DUMP_PREFIX_ADDRESS, 16,
479 						       1,
480 						       buffer_info->skb->data,
481 						       adapter->rx_buffer_len,
482 						       true);
483 			}
484 		}
485 	}
486 }
487 
488 /**
489  * e1000_desc_unused - calculate if we have unused descriptors
490  **/
491 static int e1000_desc_unused(struct e1000_ring *ring)
492 {
493 	if (ring->next_to_clean > ring->next_to_use)
494 		return ring->next_to_clean - ring->next_to_use - 1;
495 
496 	return ring->count + ring->next_to_clean - ring->next_to_use - 1;
497 }
498 
499 /**
500  * e1000e_systim_to_hwtstamp - convert system time value to hw time stamp
501  * @adapter: board private structure
502  * @hwtstamps: time stamp structure to update
503  * @systim: unsigned 64bit system time value.
504  *
505  * Convert the system time value stored in the RX/TXSTMP registers into a
506  * hwtstamp which can be used by the upper level time stamping functions.
507  *
508  * The 'systim_lock' spinlock is used to protect the consistency of the
509  * system time value. This is needed because reading the 64 bit time
510  * value involves reading two 32 bit registers. The first read latches the
511  * value.
512  **/
513 static void e1000e_systim_to_hwtstamp(struct e1000_adapter *adapter,
514 				      struct skb_shared_hwtstamps *hwtstamps,
515 				      u64 systim)
516 {
517 	u64 ns;
518 	unsigned long flags;
519 
520 	spin_lock_irqsave(&adapter->systim_lock, flags);
521 	ns = timecounter_cyc2time(&adapter->tc, systim);
522 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
523 
524 	memset(hwtstamps, 0, sizeof(*hwtstamps));
525 	hwtstamps->hwtstamp = ns_to_ktime(ns);
526 }
527 
528 /**
529  * e1000e_rx_hwtstamp - utility function which checks for Rx time stamp
530  * @adapter: board private structure
531  * @status: descriptor extended error and status field
532  * @skb: particular skb to include time stamp
533  *
534  * If the time stamp is valid, convert it into the timecounter ns value
535  * and store that result into the shhwtstamps structure which is passed
536  * up the network stack.
537  **/
538 static void e1000e_rx_hwtstamp(struct e1000_adapter *adapter, u32 status,
539 			       struct sk_buff *skb)
540 {
541 	struct e1000_hw *hw = &adapter->hw;
542 	u64 rxstmp;
543 
544 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP) ||
545 	    !(status & E1000_RXDEXT_STATERR_TST) ||
546 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID))
547 		return;
548 
549 	/* The Rx time stamp registers contain the time stamp.  No other
550 	 * received packet will be time stamped until the Rx time stamp
551 	 * registers are read.  Because only one packet can be time stamped
552 	 * at a time, the register values must belong to this packet and
553 	 * therefore none of the other additional attributes need to be
554 	 * compared.
555 	 */
556 	rxstmp = (u64)er32(RXSTMPL);
557 	rxstmp |= (u64)er32(RXSTMPH) << 32;
558 	e1000e_systim_to_hwtstamp(adapter, skb_hwtstamps(skb), rxstmp);
559 
560 	adapter->flags2 &= ~FLAG2_CHECK_RX_HWTSTAMP;
561 }
562 
563 /**
564  * e1000_receive_skb - helper function to handle Rx indications
565  * @adapter: board private structure
566  * @staterr: descriptor extended error and status field as written by hardware
567  * @vlan: descriptor vlan field as written by hardware (no le/be conversion)
568  * @skb: pointer to sk_buff to be indicated to stack
569  **/
570 static void e1000_receive_skb(struct e1000_adapter *adapter,
571 			      struct net_device *netdev, struct sk_buff *skb,
572 			      u32 staterr, __le16 vlan)
573 {
574 	u16 tag = le16_to_cpu(vlan);
575 
576 	e1000e_rx_hwtstamp(adapter, staterr, skb);
577 
578 	skb->protocol = eth_type_trans(skb, netdev);
579 
580 	if (staterr & E1000_RXD_STAT_VP)
581 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), tag);
582 
583 	napi_gro_receive(&adapter->napi, skb);
584 }
585 
586 /**
587  * e1000_rx_checksum - Receive Checksum Offload
588  * @adapter: board private structure
589  * @status_err: receive descriptor status and error fields
590  * @csum: receive descriptor csum field
591  * @sk_buff: socket buffer with received data
592  **/
593 static void e1000_rx_checksum(struct e1000_adapter *adapter, u32 status_err,
594 			      struct sk_buff *skb)
595 {
596 	u16 status = (u16)status_err;
597 	u8 errors = (u8)(status_err >> 24);
598 
599 	skb_checksum_none_assert(skb);
600 
601 	/* Rx checksum disabled */
602 	if (!(adapter->netdev->features & NETIF_F_RXCSUM))
603 		return;
604 
605 	/* Ignore Checksum bit is set */
606 	if (status & E1000_RXD_STAT_IXSM)
607 		return;
608 
609 	/* TCP/UDP checksum error bit or IP checksum error bit is set */
610 	if (errors & (E1000_RXD_ERR_TCPE | E1000_RXD_ERR_IPE)) {
611 		/* let the stack verify checksum errors */
612 		adapter->hw_csum_err++;
613 		return;
614 	}
615 
616 	/* TCP/UDP Checksum has not been calculated */
617 	if (!(status & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)))
618 		return;
619 
620 	/* It must be a TCP or UDP packet with a valid checksum */
621 	skb->ip_summed = CHECKSUM_UNNECESSARY;
622 	adapter->hw_csum_good++;
623 }
624 
625 static void e1000e_update_rdt_wa(struct e1000_ring *rx_ring, unsigned int i)
626 {
627 	struct e1000_adapter *adapter = rx_ring->adapter;
628 	struct e1000_hw *hw = &adapter->hw;
629 	s32 ret_val = __ew32_prepare(hw);
630 
631 	writel(i, rx_ring->tail);
632 
633 	if (unlikely(!ret_val && (i != readl(rx_ring->tail)))) {
634 		u32 rctl = er32(RCTL);
635 
636 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
637 		e_err("ME firmware caused invalid RDT - resetting\n");
638 		schedule_work(&adapter->reset_task);
639 	}
640 }
641 
642 static void e1000e_update_tdt_wa(struct e1000_ring *tx_ring, unsigned int i)
643 {
644 	struct e1000_adapter *adapter = tx_ring->adapter;
645 	struct e1000_hw *hw = &adapter->hw;
646 	s32 ret_val = __ew32_prepare(hw);
647 
648 	writel(i, tx_ring->tail);
649 
650 	if (unlikely(!ret_val && (i != readl(tx_ring->tail)))) {
651 		u32 tctl = er32(TCTL);
652 
653 		ew32(TCTL, tctl & ~E1000_TCTL_EN);
654 		e_err("ME firmware caused invalid TDT - resetting\n");
655 		schedule_work(&adapter->reset_task);
656 	}
657 }
658 
659 /**
660  * e1000_alloc_rx_buffers - Replace used receive buffers
661  * @rx_ring: Rx descriptor ring
662  **/
663 static void e1000_alloc_rx_buffers(struct e1000_ring *rx_ring,
664 				   int cleaned_count, gfp_t gfp)
665 {
666 	struct e1000_adapter *adapter = rx_ring->adapter;
667 	struct net_device *netdev = adapter->netdev;
668 	struct pci_dev *pdev = adapter->pdev;
669 	union e1000_rx_desc_extended *rx_desc;
670 	struct e1000_buffer *buffer_info;
671 	struct sk_buff *skb;
672 	unsigned int i;
673 	unsigned int bufsz = adapter->rx_buffer_len;
674 
675 	i = rx_ring->next_to_use;
676 	buffer_info = &rx_ring->buffer_info[i];
677 
678 	while (cleaned_count--) {
679 		skb = buffer_info->skb;
680 		if (skb) {
681 			skb_trim(skb, 0);
682 			goto map_skb;
683 		}
684 
685 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
686 		if (!skb) {
687 			/* Better luck next round */
688 			adapter->alloc_rx_buff_failed++;
689 			break;
690 		}
691 
692 		buffer_info->skb = skb;
693 map_skb:
694 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
695 						  adapter->rx_buffer_len,
696 						  DMA_FROM_DEVICE);
697 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
698 			dev_err(&pdev->dev, "Rx DMA map failed\n");
699 			adapter->rx_dma_failed++;
700 			break;
701 		}
702 
703 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
704 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
705 
706 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
707 			/* Force memory writes to complete before letting h/w
708 			 * know there are new descriptors to fetch.  (Only
709 			 * applicable for weak-ordered memory model archs,
710 			 * such as IA-64).
711 			 */
712 			wmb();
713 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
714 				e1000e_update_rdt_wa(rx_ring, i);
715 			else
716 				writel(i, rx_ring->tail);
717 		}
718 		i++;
719 		if (i == rx_ring->count)
720 			i = 0;
721 		buffer_info = &rx_ring->buffer_info[i];
722 	}
723 
724 	rx_ring->next_to_use = i;
725 }
726 
727 /**
728  * e1000_alloc_rx_buffers_ps - Replace used receive buffers; packet split
729  * @rx_ring: Rx descriptor ring
730  **/
731 static void e1000_alloc_rx_buffers_ps(struct e1000_ring *rx_ring,
732 				      int cleaned_count, gfp_t gfp)
733 {
734 	struct e1000_adapter *adapter = rx_ring->adapter;
735 	struct net_device *netdev = adapter->netdev;
736 	struct pci_dev *pdev = adapter->pdev;
737 	union e1000_rx_desc_packet_split *rx_desc;
738 	struct e1000_buffer *buffer_info;
739 	struct e1000_ps_page *ps_page;
740 	struct sk_buff *skb;
741 	unsigned int i, j;
742 
743 	i = rx_ring->next_to_use;
744 	buffer_info = &rx_ring->buffer_info[i];
745 
746 	while (cleaned_count--) {
747 		rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
748 
749 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
750 			ps_page = &buffer_info->ps_pages[j];
751 			if (j >= adapter->rx_ps_pages) {
752 				/* all unused desc entries get hw null ptr */
753 				rx_desc->read.buffer_addr[j + 1] =
754 				    ~cpu_to_le64(0);
755 				continue;
756 			}
757 			if (!ps_page->page) {
758 				ps_page->page = alloc_page(gfp);
759 				if (!ps_page->page) {
760 					adapter->alloc_rx_buff_failed++;
761 					goto no_buffers;
762 				}
763 				ps_page->dma = dma_map_page(&pdev->dev,
764 							    ps_page->page,
765 							    0, PAGE_SIZE,
766 							    DMA_FROM_DEVICE);
767 				if (dma_mapping_error(&pdev->dev,
768 						      ps_page->dma)) {
769 					dev_err(&adapter->pdev->dev,
770 						"Rx DMA page map failed\n");
771 					adapter->rx_dma_failed++;
772 					goto no_buffers;
773 				}
774 			}
775 			/* Refresh the desc even if buffer_addrs
776 			 * didn't change because each write-back
777 			 * erases this info.
778 			 */
779 			rx_desc->read.buffer_addr[j + 1] =
780 			    cpu_to_le64(ps_page->dma);
781 		}
782 
783 		skb = __netdev_alloc_skb_ip_align(netdev, adapter->rx_ps_bsize0,
784 						  gfp);
785 
786 		if (!skb) {
787 			adapter->alloc_rx_buff_failed++;
788 			break;
789 		}
790 
791 		buffer_info->skb = skb;
792 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
793 						  adapter->rx_ps_bsize0,
794 						  DMA_FROM_DEVICE);
795 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
796 			dev_err(&pdev->dev, "Rx DMA map failed\n");
797 			adapter->rx_dma_failed++;
798 			/* cleanup skb */
799 			dev_kfree_skb_any(skb);
800 			buffer_info->skb = NULL;
801 			break;
802 		}
803 
804 		rx_desc->read.buffer_addr[0] = cpu_to_le64(buffer_info->dma);
805 
806 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
807 			/* Force memory writes to complete before letting h/w
808 			 * know there are new descriptors to fetch.  (Only
809 			 * applicable for weak-ordered memory model archs,
810 			 * such as IA-64).
811 			 */
812 			wmb();
813 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
814 				e1000e_update_rdt_wa(rx_ring, i << 1);
815 			else
816 				writel(i << 1, rx_ring->tail);
817 		}
818 
819 		i++;
820 		if (i == rx_ring->count)
821 			i = 0;
822 		buffer_info = &rx_ring->buffer_info[i];
823 	}
824 
825 no_buffers:
826 	rx_ring->next_to_use = i;
827 }
828 
829 /**
830  * e1000_alloc_jumbo_rx_buffers - Replace used jumbo receive buffers
831  * @rx_ring: Rx descriptor ring
832  * @cleaned_count: number of buffers to allocate this pass
833  **/
834 
835 static void e1000_alloc_jumbo_rx_buffers(struct e1000_ring *rx_ring,
836 					 int cleaned_count, gfp_t gfp)
837 {
838 	struct e1000_adapter *adapter = rx_ring->adapter;
839 	struct net_device *netdev = adapter->netdev;
840 	struct pci_dev *pdev = adapter->pdev;
841 	union e1000_rx_desc_extended *rx_desc;
842 	struct e1000_buffer *buffer_info;
843 	struct sk_buff *skb;
844 	unsigned int i;
845 	unsigned int bufsz = 256 - 16;	/* for skb_reserve */
846 
847 	i = rx_ring->next_to_use;
848 	buffer_info = &rx_ring->buffer_info[i];
849 
850 	while (cleaned_count--) {
851 		skb = buffer_info->skb;
852 		if (skb) {
853 			skb_trim(skb, 0);
854 			goto check_page;
855 		}
856 
857 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
858 		if (unlikely(!skb)) {
859 			/* Better luck next round */
860 			adapter->alloc_rx_buff_failed++;
861 			break;
862 		}
863 
864 		buffer_info->skb = skb;
865 check_page:
866 		/* allocate a new page if necessary */
867 		if (!buffer_info->page) {
868 			buffer_info->page = alloc_page(gfp);
869 			if (unlikely(!buffer_info->page)) {
870 				adapter->alloc_rx_buff_failed++;
871 				break;
872 			}
873 		}
874 
875 		if (!buffer_info->dma) {
876 			buffer_info->dma = dma_map_page(&pdev->dev,
877 							buffer_info->page, 0,
878 							PAGE_SIZE,
879 							DMA_FROM_DEVICE);
880 			if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
881 				adapter->alloc_rx_buff_failed++;
882 				break;
883 			}
884 		}
885 
886 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
887 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
888 
889 		if (unlikely(++i == rx_ring->count))
890 			i = 0;
891 		buffer_info = &rx_ring->buffer_info[i];
892 	}
893 
894 	if (likely(rx_ring->next_to_use != i)) {
895 		rx_ring->next_to_use = i;
896 		if (unlikely(i-- == 0))
897 			i = (rx_ring->count - 1);
898 
899 		/* Force memory writes to complete before letting h/w
900 		 * know there are new descriptors to fetch.  (Only
901 		 * applicable for weak-ordered memory model archs,
902 		 * such as IA-64).
903 		 */
904 		wmb();
905 		if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
906 			e1000e_update_rdt_wa(rx_ring, i);
907 		else
908 			writel(i, rx_ring->tail);
909 	}
910 }
911 
912 static inline void e1000_rx_hash(struct net_device *netdev, __le32 rss,
913 				 struct sk_buff *skb)
914 {
915 	if (netdev->features & NETIF_F_RXHASH)
916 		skb_set_hash(skb, le32_to_cpu(rss), PKT_HASH_TYPE_L3);
917 }
918 
919 /**
920  * e1000_clean_rx_irq - Send received data up the network stack
921  * @rx_ring: Rx descriptor ring
922  *
923  * the return value indicates whether actual cleaning was done, there
924  * is no guarantee that everything was cleaned
925  **/
926 static bool e1000_clean_rx_irq(struct e1000_ring *rx_ring, int *work_done,
927 			       int work_to_do)
928 {
929 	struct e1000_adapter *adapter = rx_ring->adapter;
930 	struct net_device *netdev = adapter->netdev;
931 	struct pci_dev *pdev = adapter->pdev;
932 	struct e1000_hw *hw = &adapter->hw;
933 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
934 	struct e1000_buffer *buffer_info, *next_buffer;
935 	u32 length, staterr;
936 	unsigned int i;
937 	int cleaned_count = 0;
938 	bool cleaned = false;
939 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
940 
941 	i = rx_ring->next_to_clean;
942 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
943 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
944 	buffer_info = &rx_ring->buffer_info[i];
945 
946 	while (staterr & E1000_RXD_STAT_DD) {
947 		struct sk_buff *skb;
948 
949 		if (*work_done >= work_to_do)
950 			break;
951 		(*work_done)++;
952 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
953 
954 		skb = buffer_info->skb;
955 		buffer_info->skb = NULL;
956 
957 		prefetch(skb->data - NET_IP_ALIGN);
958 
959 		i++;
960 		if (i == rx_ring->count)
961 			i = 0;
962 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
963 		prefetch(next_rxd);
964 
965 		next_buffer = &rx_ring->buffer_info[i];
966 
967 		cleaned = true;
968 		cleaned_count++;
969 		dma_unmap_single(&pdev->dev, buffer_info->dma,
970 				 adapter->rx_buffer_len, DMA_FROM_DEVICE);
971 		buffer_info->dma = 0;
972 
973 		length = le16_to_cpu(rx_desc->wb.upper.length);
974 
975 		/* !EOP means multiple descriptors were used to store a single
976 		 * packet, if that's the case we need to toss it.  In fact, we
977 		 * need to toss every packet with the EOP bit clear and the
978 		 * next frame that _does_ have the EOP bit set, as it is by
979 		 * definition only a frame fragment
980 		 */
981 		if (unlikely(!(staterr & E1000_RXD_STAT_EOP)))
982 			adapter->flags2 |= FLAG2_IS_DISCARDING;
983 
984 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
985 			/* All receives must fit into a single buffer */
986 			e_dbg("Receive packet consumed multiple buffers\n");
987 			/* recycle */
988 			buffer_info->skb = skb;
989 			if (staterr & E1000_RXD_STAT_EOP)
990 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
991 			goto next_desc;
992 		}
993 
994 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
995 			     !(netdev->features & NETIF_F_RXALL))) {
996 			/* recycle */
997 			buffer_info->skb = skb;
998 			goto next_desc;
999 		}
1000 
1001 		/* adjust length to remove Ethernet CRC */
1002 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1003 			/* If configured to store CRC, don't subtract FCS,
1004 			 * but keep the FCS bytes out of the total_rx_bytes
1005 			 * counter
1006 			 */
1007 			if (netdev->features & NETIF_F_RXFCS)
1008 				total_rx_bytes -= 4;
1009 			else
1010 				length -= 4;
1011 		}
1012 
1013 		total_rx_bytes += length;
1014 		total_rx_packets++;
1015 
1016 		/* code added for copybreak, this should improve
1017 		 * performance for small packets with large amounts
1018 		 * of reassembly being done in the stack
1019 		 */
1020 		if (length < copybreak) {
1021 			struct sk_buff *new_skb =
1022 				napi_alloc_skb(&adapter->napi, length);
1023 			if (new_skb) {
1024 				skb_copy_to_linear_data_offset(new_skb,
1025 							       -NET_IP_ALIGN,
1026 							       (skb->data -
1027 								NET_IP_ALIGN),
1028 							       (length +
1029 								NET_IP_ALIGN));
1030 				/* save the skb in buffer_info as good */
1031 				buffer_info->skb = skb;
1032 				skb = new_skb;
1033 			}
1034 			/* else just continue with the old one */
1035 		}
1036 		/* end copybreak code */
1037 		skb_put(skb, length);
1038 
1039 		/* Receive Checksum Offload */
1040 		e1000_rx_checksum(adapter, staterr, skb);
1041 
1042 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1043 
1044 		e1000_receive_skb(adapter, netdev, skb, staterr,
1045 				  rx_desc->wb.upper.vlan);
1046 
1047 next_desc:
1048 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1049 
1050 		/* return some buffers to hardware, one at a time is too slow */
1051 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1052 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1053 					      GFP_ATOMIC);
1054 			cleaned_count = 0;
1055 		}
1056 
1057 		/* use prefetched values */
1058 		rx_desc = next_rxd;
1059 		buffer_info = next_buffer;
1060 
1061 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1062 	}
1063 	rx_ring->next_to_clean = i;
1064 
1065 	cleaned_count = e1000_desc_unused(rx_ring);
1066 	if (cleaned_count)
1067 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1068 
1069 	adapter->total_rx_bytes += total_rx_bytes;
1070 	adapter->total_rx_packets += total_rx_packets;
1071 	return cleaned;
1072 }
1073 
1074 static void e1000_put_txbuf(struct e1000_ring *tx_ring,
1075 			    struct e1000_buffer *buffer_info,
1076 			    bool drop)
1077 {
1078 	struct e1000_adapter *adapter = tx_ring->adapter;
1079 
1080 	if (buffer_info->dma) {
1081 		if (buffer_info->mapped_as_page)
1082 			dma_unmap_page(&adapter->pdev->dev, buffer_info->dma,
1083 				       buffer_info->length, DMA_TO_DEVICE);
1084 		else
1085 			dma_unmap_single(&adapter->pdev->dev, buffer_info->dma,
1086 					 buffer_info->length, DMA_TO_DEVICE);
1087 		buffer_info->dma = 0;
1088 	}
1089 	if (buffer_info->skb) {
1090 		if (drop)
1091 			dev_kfree_skb_any(buffer_info->skb);
1092 		else
1093 			dev_consume_skb_any(buffer_info->skb);
1094 		buffer_info->skb = NULL;
1095 	}
1096 	buffer_info->time_stamp = 0;
1097 }
1098 
1099 static void e1000_print_hw_hang(struct work_struct *work)
1100 {
1101 	struct e1000_adapter *adapter = container_of(work,
1102 						     struct e1000_adapter,
1103 						     print_hang_task);
1104 	struct net_device *netdev = adapter->netdev;
1105 	struct e1000_ring *tx_ring = adapter->tx_ring;
1106 	unsigned int i = tx_ring->next_to_clean;
1107 	unsigned int eop = tx_ring->buffer_info[i].next_to_watch;
1108 	struct e1000_tx_desc *eop_desc = E1000_TX_DESC(*tx_ring, eop);
1109 	struct e1000_hw *hw = &adapter->hw;
1110 	u16 phy_status, phy_1000t_status, phy_ext_status;
1111 	u16 pci_status;
1112 
1113 	if (test_bit(__E1000_DOWN, &adapter->state))
1114 		return;
1115 
1116 	if (!adapter->tx_hang_recheck && (adapter->flags2 & FLAG2_DMA_BURST)) {
1117 		/* May be block on write-back, flush and detect again
1118 		 * flush pending descriptor writebacks to memory
1119 		 */
1120 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1121 		/* execute the writes immediately */
1122 		e1e_flush();
1123 		/* Due to rare timing issues, write to TIDV again to ensure
1124 		 * the write is successful
1125 		 */
1126 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1127 		/* execute the writes immediately */
1128 		e1e_flush();
1129 		adapter->tx_hang_recheck = true;
1130 		return;
1131 	}
1132 	adapter->tx_hang_recheck = false;
1133 
1134 	if (er32(TDH(0)) == er32(TDT(0))) {
1135 		e_dbg("false hang detected, ignoring\n");
1136 		return;
1137 	}
1138 
1139 	/* Real hang detected */
1140 	netif_stop_queue(netdev);
1141 
1142 	e1e_rphy(hw, MII_BMSR, &phy_status);
1143 	e1e_rphy(hw, MII_STAT1000, &phy_1000t_status);
1144 	e1e_rphy(hw, MII_ESTATUS, &phy_ext_status);
1145 
1146 	pci_read_config_word(adapter->pdev, PCI_STATUS, &pci_status);
1147 
1148 	/* detected Hardware unit hang */
1149 	e_err("Detected Hardware Unit Hang:\n"
1150 	      "  TDH                  <%x>\n"
1151 	      "  TDT                  <%x>\n"
1152 	      "  next_to_use          <%x>\n"
1153 	      "  next_to_clean        <%x>\n"
1154 	      "buffer_info[next_to_clean]:\n"
1155 	      "  time_stamp           <%lx>\n"
1156 	      "  next_to_watch        <%x>\n"
1157 	      "  jiffies              <%lx>\n"
1158 	      "  next_to_watch.status <%x>\n"
1159 	      "MAC Status             <%x>\n"
1160 	      "PHY Status             <%x>\n"
1161 	      "PHY 1000BASE-T Status  <%x>\n"
1162 	      "PHY Extended Status    <%x>\n"
1163 	      "PCI Status             <%x>\n",
1164 	      readl(tx_ring->head), readl(tx_ring->tail), tx_ring->next_to_use,
1165 	      tx_ring->next_to_clean, tx_ring->buffer_info[eop].time_stamp,
1166 	      eop, jiffies, eop_desc->upper.fields.status, er32(STATUS),
1167 	      phy_status, phy_1000t_status, phy_ext_status, pci_status);
1168 
1169 	e1000e_dump(adapter);
1170 
1171 	/* Suggest workaround for known h/w issue */
1172 	if ((hw->mac.type == e1000_pchlan) && (er32(CTRL) & E1000_CTRL_TFCE))
1173 		e_err("Try turning off Tx pause (flow control) via ethtool\n");
1174 }
1175 
1176 /**
1177  * e1000e_tx_hwtstamp_work - check for Tx time stamp
1178  * @work: pointer to work struct
1179  *
1180  * This work function polls the TSYNCTXCTL valid bit to determine when a
1181  * timestamp has been taken for the current stored skb.  The timestamp must
1182  * be for this skb because only one such packet is allowed in the queue.
1183  */
1184 static void e1000e_tx_hwtstamp_work(struct work_struct *work)
1185 {
1186 	struct e1000_adapter *adapter = container_of(work, struct e1000_adapter,
1187 						     tx_hwtstamp_work);
1188 	struct e1000_hw *hw = &adapter->hw;
1189 
1190 	if (er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID) {
1191 		struct sk_buff *skb = adapter->tx_hwtstamp_skb;
1192 		struct skb_shared_hwtstamps shhwtstamps;
1193 		u64 txstmp;
1194 
1195 		txstmp = er32(TXSTMPL);
1196 		txstmp |= (u64)er32(TXSTMPH) << 32;
1197 
1198 		e1000e_systim_to_hwtstamp(adapter, &shhwtstamps, txstmp);
1199 
1200 		/* Clear the global tx_hwtstamp_skb pointer and force writes
1201 		 * prior to notifying the stack of a Tx timestamp.
1202 		 */
1203 		adapter->tx_hwtstamp_skb = NULL;
1204 		wmb(); /* force write prior to skb_tstamp_tx */
1205 
1206 		skb_tstamp_tx(skb, &shhwtstamps);
1207 		dev_consume_skb_any(skb);
1208 	} else if (time_after(jiffies, adapter->tx_hwtstamp_start
1209 			      + adapter->tx_timeout_factor * HZ)) {
1210 		dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
1211 		adapter->tx_hwtstamp_skb = NULL;
1212 		adapter->tx_hwtstamp_timeouts++;
1213 		e_warn("clearing Tx timestamp hang\n");
1214 	} else {
1215 		/* reschedule to check later */
1216 		schedule_work(&adapter->tx_hwtstamp_work);
1217 	}
1218 }
1219 
1220 /**
1221  * e1000_clean_tx_irq - Reclaim resources after transmit completes
1222  * @tx_ring: Tx descriptor ring
1223  *
1224  * the return value indicates whether actual cleaning was done, there
1225  * is no guarantee that everything was cleaned
1226  **/
1227 static bool e1000_clean_tx_irq(struct e1000_ring *tx_ring)
1228 {
1229 	struct e1000_adapter *adapter = tx_ring->adapter;
1230 	struct net_device *netdev = adapter->netdev;
1231 	struct e1000_hw *hw = &adapter->hw;
1232 	struct e1000_tx_desc *tx_desc, *eop_desc;
1233 	struct e1000_buffer *buffer_info;
1234 	unsigned int i, eop;
1235 	unsigned int count = 0;
1236 	unsigned int total_tx_bytes = 0, total_tx_packets = 0;
1237 	unsigned int bytes_compl = 0, pkts_compl = 0;
1238 
1239 	i = tx_ring->next_to_clean;
1240 	eop = tx_ring->buffer_info[i].next_to_watch;
1241 	eop_desc = E1000_TX_DESC(*tx_ring, eop);
1242 
1243 	while ((eop_desc->upper.data & cpu_to_le32(E1000_TXD_STAT_DD)) &&
1244 	       (count < tx_ring->count)) {
1245 		bool cleaned = false;
1246 
1247 		dma_rmb();		/* read buffer_info after eop_desc */
1248 		for (; !cleaned; count++) {
1249 			tx_desc = E1000_TX_DESC(*tx_ring, i);
1250 			buffer_info = &tx_ring->buffer_info[i];
1251 			cleaned = (i == eop);
1252 
1253 			if (cleaned) {
1254 				total_tx_packets += buffer_info->segs;
1255 				total_tx_bytes += buffer_info->bytecount;
1256 				if (buffer_info->skb) {
1257 					bytes_compl += buffer_info->skb->len;
1258 					pkts_compl++;
1259 				}
1260 			}
1261 
1262 			e1000_put_txbuf(tx_ring, buffer_info, false);
1263 			tx_desc->upper.data = 0;
1264 
1265 			i++;
1266 			if (i == tx_ring->count)
1267 				i = 0;
1268 		}
1269 
1270 		if (i == tx_ring->next_to_use)
1271 			break;
1272 		eop = tx_ring->buffer_info[i].next_to_watch;
1273 		eop_desc = E1000_TX_DESC(*tx_ring, eop);
1274 	}
1275 
1276 	tx_ring->next_to_clean = i;
1277 
1278 	netdev_completed_queue(netdev, pkts_compl, bytes_compl);
1279 
1280 #define TX_WAKE_THRESHOLD 32
1281 	if (count && netif_carrier_ok(netdev) &&
1282 	    e1000_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD) {
1283 		/* Make sure that anybody stopping the queue after this
1284 		 * sees the new next_to_clean.
1285 		 */
1286 		smp_mb();
1287 
1288 		if (netif_queue_stopped(netdev) &&
1289 		    !(test_bit(__E1000_DOWN, &adapter->state))) {
1290 			netif_wake_queue(netdev);
1291 			++adapter->restart_queue;
1292 		}
1293 	}
1294 
1295 	if (adapter->detect_tx_hung) {
1296 		/* Detect a transmit hang in hardware, this serializes the
1297 		 * check with the clearing of time_stamp and movement of i
1298 		 */
1299 		adapter->detect_tx_hung = false;
1300 		if (tx_ring->buffer_info[i].time_stamp &&
1301 		    time_after(jiffies, tx_ring->buffer_info[i].time_stamp
1302 			       + (adapter->tx_timeout_factor * HZ)) &&
1303 		    !(er32(STATUS) & E1000_STATUS_TXOFF))
1304 			schedule_work(&adapter->print_hang_task);
1305 		else
1306 			adapter->tx_hang_recheck = false;
1307 	}
1308 	adapter->total_tx_bytes += total_tx_bytes;
1309 	adapter->total_tx_packets += total_tx_packets;
1310 	return count < tx_ring->count;
1311 }
1312 
1313 /**
1314  * e1000_clean_rx_irq_ps - Send received data up the network stack; packet split
1315  * @rx_ring: Rx descriptor ring
1316  *
1317  * the return value indicates whether actual cleaning was done, there
1318  * is no guarantee that everything was cleaned
1319  **/
1320 static bool e1000_clean_rx_irq_ps(struct e1000_ring *rx_ring, int *work_done,
1321 				  int work_to_do)
1322 {
1323 	struct e1000_adapter *adapter = rx_ring->adapter;
1324 	struct e1000_hw *hw = &adapter->hw;
1325 	union e1000_rx_desc_packet_split *rx_desc, *next_rxd;
1326 	struct net_device *netdev = adapter->netdev;
1327 	struct pci_dev *pdev = adapter->pdev;
1328 	struct e1000_buffer *buffer_info, *next_buffer;
1329 	struct e1000_ps_page *ps_page;
1330 	struct sk_buff *skb;
1331 	unsigned int i, j;
1332 	u32 length, staterr;
1333 	int cleaned_count = 0;
1334 	bool cleaned = false;
1335 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1336 
1337 	i = rx_ring->next_to_clean;
1338 	rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
1339 	staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1340 	buffer_info = &rx_ring->buffer_info[i];
1341 
1342 	while (staterr & E1000_RXD_STAT_DD) {
1343 		if (*work_done >= work_to_do)
1344 			break;
1345 		(*work_done)++;
1346 		skb = buffer_info->skb;
1347 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1348 
1349 		/* in the packet split case this is header only */
1350 		prefetch(skb->data - NET_IP_ALIGN);
1351 
1352 		i++;
1353 		if (i == rx_ring->count)
1354 			i = 0;
1355 		next_rxd = E1000_RX_DESC_PS(*rx_ring, i);
1356 		prefetch(next_rxd);
1357 
1358 		next_buffer = &rx_ring->buffer_info[i];
1359 
1360 		cleaned = true;
1361 		cleaned_count++;
1362 		dma_unmap_single(&pdev->dev, buffer_info->dma,
1363 				 adapter->rx_ps_bsize0, DMA_FROM_DEVICE);
1364 		buffer_info->dma = 0;
1365 
1366 		/* see !EOP comment in other Rx routine */
1367 		if (!(staterr & E1000_RXD_STAT_EOP))
1368 			adapter->flags2 |= FLAG2_IS_DISCARDING;
1369 
1370 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
1371 			e_dbg("Packet Split buffers didn't pick up the full packet\n");
1372 			dev_kfree_skb_irq(skb);
1373 			if (staterr & E1000_RXD_STAT_EOP)
1374 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1375 			goto next_desc;
1376 		}
1377 
1378 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1379 			     !(netdev->features & NETIF_F_RXALL))) {
1380 			dev_kfree_skb_irq(skb);
1381 			goto next_desc;
1382 		}
1383 
1384 		length = le16_to_cpu(rx_desc->wb.middle.length0);
1385 
1386 		if (!length) {
1387 			e_dbg("Last part of the packet spanning multiple descriptors\n");
1388 			dev_kfree_skb_irq(skb);
1389 			goto next_desc;
1390 		}
1391 
1392 		/* Good Receive */
1393 		skb_put(skb, length);
1394 
1395 		{
1396 			/* this looks ugly, but it seems compiler issues make
1397 			 * it more efficient than reusing j
1398 			 */
1399 			int l1 = le16_to_cpu(rx_desc->wb.upper.length[0]);
1400 
1401 			/* page alloc/put takes too long and effects small
1402 			 * packet throughput, so unsplit small packets and
1403 			 * save the alloc/put only valid in softirq (napi)
1404 			 * context to call kmap_*
1405 			 */
1406 			if (l1 && (l1 <= copybreak) &&
1407 			    ((length + l1) <= adapter->rx_ps_bsize0)) {
1408 				u8 *vaddr;
1409 
1410 				ps_page = &buffer_info->ps_pages[0];
1411 
1412 				/* there is no documentation about how to call
1413 				 * kmap_atomic, so we can't hold the mapping
1414 				 * very long
1415 				 */
1416 				dma_sync_single_for_cpu(&pdev->dev,
1417 							ps_page->dma,
1418 							PAGE_SIZE,
1419 							DMA_FROM_DEVICE);
1420 				vaddr = kmap_atomic(ps_page->page);
1421 				memcpy(skb_tail_pointer(skb), vaddr, l1);
1422 				kunmap_atomic(vaddr);
1423 				dma_sync_single_for_device(&pdev->dev,
1424 							   ps_page->dma,
1425 							   PAGE_SIZE,
1426 							   DMA_FROM_DEVICE);
1427 
1428 				/* remove the CRC */
1429 				if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1430 					if (!(netdev->features & NETIF_F_RXFCS))
1431 						l1 -= 4;
1432 				}
1433 
1434 				skb_put(skb, l1);
1435 				goto copydone;
1436 			}	/* if */
1437 		}
1438 
1439 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1440 			length = le16_to_cpu(rx_desc->wb.upper.length[j]);
1441 			if (!length)
1442 				break;
1443 
1444 			ps_page = &buffer_info->ps_pages[j];
1445 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1446 				       DMA_FROM_DEVICE);
1447 			ps_page->dma = 0;
1448 			skb_fill_page_desc(skb, j, ps_page->page, 0, length);
1449 			ps_page->page = NULL;
1450 			skb->len += length;
1451 			skb->data_len += length;
1452 			skb->truesize += PAGE_SIZE;
1453 		}
1454 
1455 		/* strip the ethernet crc, problem is we're using pages now so
1456 		 * this whole operation can get a little cpu intensive
1457 		 */
1458 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1459 			if (!(netdev->features & NETIF_F_RXFCS))
1460 				pskb_trim(skb, skb->len - 4);
1461 		}
1462 
1463 copydone:
1464 		total_rx_bytes += skb->len;
1465 		total_rx_packets++;
1466 
1467 		e1000_rx_checksum(adapter, staterr, skb);
1468 
1469 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1470 
1471 		if (rx_desc->wb.upper.header_status &
1472 		    cpu_to_le16(E1000_RXDPS_HDRSTAT_HDRSP))
1473 			adapter->rx_hdr_split++;
1474 
1475 		e1000_receive_skb(adapter, netdev, skb, staterr,
1476 				  rx_desc->wb.middle.vlan);
1477 
1478 next_desc:
1479 		rx_desc->wb.middle.status_error &= cpu_to_le32(~0xFF);
1480 		buffer_info->skb = NULL;
1481 
1482 		/* return some buffers to hardware, one at a time is too slow */
1483 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1484 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1485 					      GFP_ATOMIC);
1486 			cleaned_count = 0;
1487 		}
1488 
1489 		/* use prefetched values */
1490 		rx_desc = next_rxd;
1491 		buffer_info = next_buffer;
1492 
1493 		staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1494 	}
1495 	rx_ring->next_to_clean = i;
1496 
1497 	cleaned_count = e1000_desc_unused(rx_ring);
1498 	if (cleaned_count)
1499 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1500 
1501 	adapter->total_rx_bytes += total_rx_bytes;
1502 	adapter->total_rx_packets += total_rx_packets;
1503 	return cleaned;
1504 }
1505 
1506 /**
1507  * e1000_consume_page - helper function
1508  **/
1509 static void e1000_consume_page(struct e1000_buffer *bi, struct sk_buff *skb,
1510 			       u16 length)
1511 {
1512 	bi->page = NULL;
1513 	skb->len += length;
1514 	skb->data_len += length;
1515 	skb->truesize += PAGE_SIZE;
1516 }
1517 
1518 /**
1519  * e1000_clean_jumbo_rx_irq - Send received data up the network stack; legacy
1520  * @adapter: board private structure
1521  *
1522  * the return value indicates whether actual cleaning was done, there
1523  * is no guarantee that everything was cleaned
1524  **/
1525 static bool e1000_clean_jumbo_rx_irq(struct e1000_ring *rx_ring, int *work_done,
1526 				     int work_to_do)
1527 {
1528 	struct e1000_adapter *adapter = rx_ring->adapter;
1529 	struct net_device *netdev = adapter->netdev;
1530 	struct pci_dev *pdev = adapter->pdev;
1531 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
1532 	struct e1000_buffer *buffer_info, *next_buffer;
1533 	u32 length, staterr;
1534 	unsigned int i;
1535 	int cleaned_count = 0;
1536 	bool cleaned = false;
1537 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1538 	struct skb_shared_info *shinfo;
1539 
1540 	i = rx_ring->next_to_clean;
1541 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
1542 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1543 	buffer_info = &rx_ring->buffer_info[i];
1544 
1545 	while (staterr & E1000_RXD_STAT_DD) {
1546 		struct sk_buff *skb;
1547 
1548 		if (*work_done >= work_to_do)
1549 			break;
1550 		(*work_done)++;
1551 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1552 
1553 		skb = buffer_info->skb;
1554 		buffer_info->skb = NULL;
1555 
1556 		++i;
1557 		if (i == rx_ring->count)
1558 			i = 0;
1559 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
1560 		prefetch(next_rxd);
1561 
1562 		next_buffer = &rx_ring->buffer_info[i];
1563 
1564 		cleaned = true;
1565 		cleaned_count++;
1566 		dma_unmap_page(&pdev->dev, buffer_info->dma, PAGE_SIZE,
1567 			       DMA_FROM_DEVICE);
1568 		buffer_info->dma = 0;
1569 
1570 		length = le16_to_cpu(rx_desc->wb.upper.length);
1571 
1572 		/* errors is only valid for DD + EOP descriptors */
1573 		if (unlikely((staterr & E1000_RXD_STAT_EOP) &&
1574 			     ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1575 			      !(netdev->features & NETIF_F_RXALL)))) {
1576 			/* recycle both page and skb */
1577 			buffer_info->skb = skb;
1578 			/* an error means any chain goes out the window too */
1579 			if (rx_ring->rx_skb_top)
1580 				dev_kfree_skb_irq(rx_ring->rx_skb_top);
1581 			rx_ring->rx_skb_top = NULL;
1582 			goto next_desc;
1583 		}
1584 #define rxtop (rx_ring->rx_skb_top)
1585 		if (!(staterr & E1000_RXD_STAT_EOP)) {
1586 			/* this descriptor is only the beginning (or middle) */
1587 			if (!rxtop) {
1588 				/* this is the beginning of a chain */
1589 				rxtop = skb;
1590 				skb_fill_page_desc(rxtop, 0, buffer_info->page,
1591 						   0, length);
1592 			} else {
1593 				/* this is the middle of a chain */
1594 				shinfo = skb_shinfo(rxtop);
1595 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1596 						   buffer_info->page, 0,
1597 						   length);
1598 				/* re-use the skb, only consumed the page */
1599 				buffer_info->skb = skb;
1600 			}
1601 			e1000_consume_page(buffer_info, rxtop, length);
1602 			goto next_desc;
1603 		} else {
1604 			if (rxtop) {
1605 				/* end of the chain */
1606 				shinfo = skb_shinfo(rxtop);
1607 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1608 						   buffer_info->page, 0,
1609 						   length);
1610 				/* re-use the current skb, we only consumed the
1611 				 * page
1612 				 */
1613 				buffer_info->skb = skb;
1614 				skb = rxtop;
1615 				rxtop = NULL;
1616 				e1000_consume_page(buffer_info, skb, length);
1617 			} else {
1618 				/* no chain, got EOP, this buf is the packet
1619 				 * copybreak to save the put_page/alloc_page
1620 				 */
1621 				if (length <= copybreak &&
1622 				    skb_tailroom(skb) >= length) {
1623 					u8 *vaddr;
1624 					vaddr = kmap_atomic(buffer_info->page);
1625 					memcpy(skb_tail_pointer(skb), vaddr,
1626 					       length);
1627 					kunmap_atomic(vaddr);
1628 					/* re-use the page, so don't erase
1629 					 * buffer_info->page
1630 					 */
1631 					skb_put(skb, length);
1632 				} else {
1633 					skb_fill_page_desc(skb, 0,
1634 							   buffer_info->page, 0,
1635 							   length);
1636 					e1000_consume_page(buffer_info, skb,
1637 							   length);
1638 				}
1639 			}
1640 		}
1641 
1642 		/* Receive Checksum Offload */
1643 		e1000_rx_checksum(adapter, staterr, skb);
1644 
1645 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1646 
1647 		/* probably a little skewed due to removing CRC */
1648 		total_rx_bytes += skb->len;
1649 		total_rx_packets++;
1650 
1651 		/* eth type trans needs skb->data to point to something */
1652 		if (!pskb_may_pull(skb, ETH_HLEN)) {
1653 			e_err("pskb_may_pull failed.\n");
1654 			dev_kfree_skb_irq(skb);
1655 			goto next_desc;
1656 		}
1657 
1658 		e1000_receive_skb(adapter, netdev, skb, staterr,
1659 				  rx_desc->wb.upper.vlan);
1660 
1661 next_desc:
1662 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1663 
1664 		/* return some buffers to hardware, one at a time is too slow */
1665 		if (unlikely(cleaned_count >= E1000_RX_BUFFER_WRITE)) {
1666 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1667 					      GFP_ATOMIC);
1668 			cleaned_count = 0;
1669 		}
1670 
1671 		/* use prefetched values */
1672 		rx_desc = next_rxd;
1673 		buffer_info = next_buffer;
1674 
1675 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1676 	}
1677 	rx_ring->next_to_clean = i;
1678 
1679 	cleaned_count = e1000_desc_unused(rx_ring);
1680 	if (cleaned_count)
1681 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1682 
1683 	adapter->total_rx_bytes += total_rx_bytes;
1684 	adapter->total_rx_packets += total_rx_packets;
1685 	return cleaned;
1686 }
1687 
1688 /**
1689  * e1000_clean_rx_ring - Free Rx Buffers per Queue
1690  * @rx_ring: Rx descriptor ring
1691  **/
1692 static void e1000_clean_rx_ring(struct e1000_ring *rx_ring)
1693 {
1694 	struct e1000_adapter *adapter = rx_ring->adapter;
1695 	struct e1000_buffer *buffer_info;
1696 	struct e1000_ps_page *ps_page;
1697 	struct pci_dev *pdev = adapter->pdev;
1698 	unsigned int i, j;
1699 
1700 	/* Free all the Rx ring sk_buffs */
1701 	for (i = 0; i < rx_ring->count; i++) {
1702 		buffer_info = &rx_ring->buffer_info[i];
1703 		if (buffer_info->dma) {
1704 			if (adapter->clean_rx == e1000_clean_rx_irq)
1705 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1706 						 adapter->rx_buffer_len,
1707 						 DMA_FROM_DEVICE);
1708 			else if (adapter->clean_rx == e1000_clean_jumbo_rx_irq)
1709 				dma_unmap_page(&pdev->dev, buffer_info->dma,
1710 					       PAGE_SIZE, DMA_FROM_DEVICE);
1711 			else if (adapter->clean_rx == e1000_clean_rx_irq_ps)
1712 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1713 						 adapter->rx_ps_bsize0,
1714 						 DMA_FROM_DEVICE);
1715 			buffer_info->dma = 0;
1716 		}
1717 
1718 		if (buffer_info->page) {
1719 			put_page(buffer_info->page);
1720 			buffer_info->page = NULL;
1721 		}
1722 
1723 		if (buffer_info->skb) {
1724 			dev_kfree_skb(buffer_info->skb);
1725 			buffer_info->skb = NULL;
1726 		}
1727 
1728 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1729 			ps_page = &buffer_info->ps_pages[j];
1730 			if (!ps_page->page)
1731 				break;
1732 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1733 				       DMA_FROM_DEVICE);
1734 			ps_page->dma = 0;
1735 			put_page(ps_page->page);
1736 			ps_page->page = NULL;
1737 		}
1738 	}
1739 
1740 	/* there also may be some cached data from a chained receive */
1741 	if (rx_ring->rx_skb_top) {
1742 		dev_kfree_skb(rx_ring->rx_skb_top);
1743 		rx_ring->rx_skb_top = NULL;
1744 	}
1745 
1746 	/* Zero out the descriptor ring */
1747 	memset(rx_ring->desc, 0, rx_ring->size);
1748 
1749 	rx_ring->next_to_clean = 0;
1750 	rx_ring->next_to_use = 0;
1751 	adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1752 }
1753 
1754 static void e1000e_downshift_workaround(struct work_struct *work)
1755 {
1756 	struct e1000_adapter *adapter = container_of(work,
1757 						     struct e1000_adapter,
1758 						     downshift_task);
1759 
1760 	if (test_bit(__E1000_DOWN, &adapter->state))
1761 		return;
1762 
1763 	e1000e_gig_downshift_workaround_ich8lan(&adapter->hw);
1764 }
1765 
1766 /**
1767  * e1000_intr_msi - Interrupt Handler
1768  * @irq: interrupt number
1769  * @data: pointer to a network interface device structure
1770  **/
1771 static irqreturn_t e1000_intr_msi(int __always_unused irq, void *data)
1772 {
1773 	struct net_device *netdev = data;
1774 	struct e1000_adapter *adapter = netdev_priv(netdev);
1775 	struct e1000_hw *hw = &adapter->hw;
1776 	u32 icr = er32(ICR);
1777 
1778 	/* read ICR disables interrupts using IAM */
1779 	if (icr & E1000_ICR_LSC) {
1780 		hw->mac.get_link_status = true;
1781 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1782 		 * disconnect (LSC) before accessing any PHY registers
1783 		 */
1784 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1785 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1786 			schedule_work(&adapter->downshift_task);
1787 
1788 		/* 80003ES2LAN workaround-- For packet buffer work-around on
1789 		 * link down event; disable receives here in the ISR and reset
1790 		 * adapter in watchdog
1791 		 */
1792 		if (netif_carrier_ok(netdev) &&
1793 		    adapter->flags & FLAG_RX_NEEDS_RESTART) {
1794 			/* disable receives */
1795 			u32 rctl = er32(RCTL);
1796 
1797 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1798 			adapter->flags |= FLAG_RESTART_NOW;
1799 		}
1800 		/* guard against interrupt when we're going down */
1801 		if (!test_bit(__E1000_DOWN, &adapter->state))
1802 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1803 	}
1804 
1805 	/* Reset on uncorrectable ECC error */
1806 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1807 		u32 pbeccsts = er32(PBECCSTS);
1808 
1809 		adapter->corr_errors +=
1810 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1811 		adapter->uncorr_errors +=
1812 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1813 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1814 
1815 		/* Do the reset outside of interrupt context */
1816 		schedule_work(&adapter->reset_task);
1817 
1818 		/* return immediately since reset is imminent */
1819 		return IRQ_HANDLED;
1820 	}
1821 
1822 	if (napi_schedule_prep(&adapter->napi)) {
1823 		adapter->total_tx_bytes = 0;
1824 		adapter->total_tx_packets = 0;
1825 		adapter->total_rx_bytes = 0;
1826 		adapter->total_rx_packets = 0;
1827 		__napi_schedule(&adapter->napi);
1828 	}
1829 
1830 	return IRQ_HANDLED;
1831 }
1832 
1833 /**
1834  * e1000_intr - Interrupt Handler
1835  * @irq: interrupt number
1836  * @data: pointer to a network interface device structure
1837  **/
1838 static irqreturn_t e1000_intr(int __always_unused irq, void *data)
1839 {
1840 	struct net_device *netdev = data;
1841 	struct e1000_adapter *adapter = netdev_priv(netdev);
1842 	struct e1000_hw *hw = &adapter->hw;
1843 	u32 rctl, icr = er32(ICR);
1844 
1845 	if (!icr || test_bit(__E1000_DOWN, &adapter->state))
1846 		return IRQ_NONE;	/* Not our interrupt */
1847 
1848 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
1849 	 * not set, then the adapter didn't send an interrupt
1850 	 */
1851 	if (!(icr & E1000_ICR_INT_ASSERTED))
1852 		return IRQ_NONE;
1853 
1854 	/* Interrupt Auto-Mask...upon reading ICR,
1855 	 * interrupts are masked.  No need for the
1856 	 * IMC write
1857 	 */
1858 
1859 	if (icr & E1000_ICR_LSC) {
1860 		hw->mac.get_link_status = true;
1861 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1862 		 * disconnect (LSC) before accessing any PHY registers
1863 		 */
1864 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1865 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1866 			schedule_work(&adapter->downshift_task);
1867 
1868 		/* 80003ES2LAN workaround--
1869 		 * For packet buffer work-around on link down event;
1870 		 * disable receives here in the ISR and
1871 		 * reset adapter in watchdog
1872 		 */
1873 		if (netif_carrier_ok(netdev) &&
1874 		    (adapter->flags & FLAG_RX_NEEDS_RESTART)) {
1875 			/* disable receives */
1876 			rctl = er32(RCTL);
1877 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1878 			adapter->flags |= FLAG_RESTART_NOW;
1879 		}
1880 		/* guard against interrupt when we're going down */
1881 		if (!test_bit(__E1000_DOWN, &adapter->state))
1882 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1883 	}
1884 
1885 	/* Reset on uncorrectable ECC error */
1886 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1887 		u32 pbeccsts = er32(PBECCSTS);
1888 
1889 		adapter->corr_errors +=
1890 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1891 		adapter->uncorr_errors +=
1892 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1893 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1894 
1895 		/* Do the reset outside of interrupt context */
1896 		schedule_work(&adapter->reset_task);
1897 
1898 		/* return immediately since reset is imminent */
1899 		return IRQ_HANDLED;
1900 	}
1901 
1902 	if (napi_schedule_prep(&adapter->napi)) {
1903 		adapter->total_tx_bytes = 0;
1904 		adapter->total_tx_packets = 0;
1905 		adapter->total_rx_bytes = 0;
1906 		adapter->total_rx_packets = 0;
1907 		__napi_schedule(&adapter->napi);
1908 	}
1909 
1910 	return IRQ_HANDLED;
1911 }
1912 
1913 static irqreturn_t e1000_msix_other(int __always_unused irq, void *data)
1914 {
1915 	struct net_device *netdev = data;
1916 	struct e1000_adapter *adapter = netdev_priv(netdev);
1917 	struct e1000_hw *hw = &adapter->hw;
1918 	u32 icr = er32(ICR);
1919 
1920 	if (icr & adapter->eiac_mask)
1921 		ew32(ICS, (icr & adapter->eiac_mask));
1922 
1923 	if (icr & E1000_ICR_LSC) {
1924 		hw->mac.get_link_status = true;
1925 		/* guard against interrupt when we're going down */
1926 		if (!test_bit(__E1000_DOWN, &adapter->state))
1927 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1928 	}
1929 
1930 	if (!test_bit(__E1000_DOWN, &adapter->state))
1931 		ew32(IMS, E1000_IMS_OTHER | IMS_OTHER_MASK);
1932 
1933 	return IRQ_HANDLED;
1934 }
1935 
1936 static irqreturn_t e1000_intr_msix_tx(int __always_unused irq, void *data)
1937 {
1938 	struct net_device *netdev = data;
1939 	struct e1000_adapter *adapter = netdev_priv(netdev);
1940 	struct e1000_hw *hw = &adapter->hw;
1941 	struct e1000_ring *tx_ring = adapter->tx_ring;
1942 
1943 	adapter->total_tx_bytes = 0;
1944 	adapter->total_tx_packets = 0;
1945 
1946 	if (!e1000_clean_tx_irq(tx_ring))
1947 		/* Ring was not completely cleaned, so fire another interrupt */
1948 		ew32(ICS, tx_ring->ims_val);
1949 
1950 	if (!test_bit(__E1000_DOWN, &adapter->state))
1951 		ew32(IMS, adapter->tx_ring->ims_val);
1952 
1953 	return IRQ_HANDLED;
1954 }
1955 
1956 static irqreturn_t e1000_intr_msix_rx(int __always_unused irq, void *data)
1957 {
1958 	struct net_device *netdev = data;
1959 	struct e1000_adapter *adapter = netdev_priv(netdev);
1960 	struct e1000_ring *rx_ring = adapter->rx_ring;
1961 
1962 	/* Write the ITR value calculated at the end of the
1963 	 * previous interrupt.
1964 	 */
1965 	if (rx_ring->set_itr) {
1966 		u32 itr = rx_ring->itr_val ?
1967 			  1000000000 / (rx_ring->itr_val * 256) : 0;
1968 
1969 		writel(itr, rx_ring->itr_register);
1970 		rx_ring->set_itr = 0;
1971 	}
1972 
1973 	if (napi_schedule_prep(&adapter->napi)) {
1974 		adapter->total_rx_bytes = 0;
1975 		adapter->total_rx_packets = 0;
1976 		__napi_schedule(&adapter->napi);
1977 	}
1978 	return IRQ_HANDLED;
1979 }
1980 
1981 /**
1982  * e1000_configure_msix - Configure MSI-X hardware
1983  *
1984  * e1000_configure_msix sets up the hardware to properly
1985  * generate MSI-X interrupts.
1986  **/
1987 static void e1000_configure_msix(struct e1000_adapter *adapter)
1988 {
1989 	struct e1000_hw *hw = &adapter->hw;
1990 	struct e1000_ring *rx_ring = adapter->rx_ring;
1991 	struct e1000_ring *tx_ring = adapter->tx_ring;
1992 	int vector = 0;
1993 	u32 ctrl_ext, ivar = 0;
1994 
1995 	adapter->eiac_mask = 0;
1996 
1997 	/* Workaround issue with spurious interrupts on 82574 in MSI-X mode */
1998 	if (hw->mac.type == e1000_82574) {
1999 		u32 rfctl = er32(RFCTL);
2000 
2001 		rfctl |= E1000_RFCTL_ACK_DIS;
2002 		ew32(RFCTL, rfctl);
2003 	}
2004 
2005 	/* Configure Rx vector */
2006 	rx_ring->ims_val = E1000_IMS_RXQ0;
2007 	adapter->eiac_mask |= rx_ring->ims_val;
2008 	if (rx_ring->itr_val)
2009 		writel(1000000000 / (rx_ring->itr_val * 256),
2010 		       rx_ring->itr_register);
2011 	else
2012 		writel(1, rx_ring->itr_register);
2013 	ivar = E1000_IVAR_INT_ALLOC_VALID | vector;
2014 
2015 	/* Configure Tx vector */
2016 	tx_ring->ims_val = E1000_IMS_TXQ0;
2017 	vector++;
2018 	if (tx_ring->itr_val)
2019 		writel(1000000000 / (tx_ring->itr_val * 256),
2020 		       tx_ring->itr_register);
2021 	else
2022 		writel(1, tx_ring->itr_register);
2023 	adapter->eiac_mask |= tx_ring->ims_val;
2024 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 8);
2025 
2026 	/* set vector for Other Causes, e.g. link changes */
2027 	vector++;
2028 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 16);
2029 	if (rx_ring->itr_val)
2030 		writel(1000000000 / (rx_ring->itr_val * 256),
2031 		       hw->hw_addr + E1000_EITR_82574(vector));
2032 	else
2033 		writel(1, hw->hw_addr + E1000_EITR_82574(vector));
2034 
2035 	/* Cause Tx interrupts on every write back */
2036 	ivar |= BIT(31);
2037 
2038 	ew32(IVAR, ivar);
2039 
2040 	/* enable MSI-X PBA support */
2041 	ctrl_ext = er32(CTRL_EXT) & ~E1000_CTRL_EXT_IAME;
2042 	ctrl_ext |= E1000_CTRL_EXT_PBA_CLR | E1000_CTRL_EXT_EIAME;
2043 	ew32(CTRL_EXT, ctrl_ext);
2044 	e1e_flush();
2045 }
2046 
2047 void e1000e_reset_interrupt_capability(struct e1000_adapter *adapter)
2048 {
2049 	if (adapter->msix_entries) {
2050 		pci_disable_msix(adapter->pdev);
2051 		kfree(adapter->msix_entries);
2052 		adapter->msix_entries = NULL;
2053 	} else if (adapter->flags & FLAG_MSI_ENABLED) {
2054 		pci_disable_msi(adapter->pdev);
2055 		adapter->flags &= ~FLAG_MSI_ENABLED;
2056 	}
2057 }
2058 
2059 /**
2060  * e1000e_set_interrupt_capability - set MSI or MSI-X if supported
2061  *
2062  * Attempt to configure interrupts using the best available
2063  * capabilities of the hardware and kernel.
2064  **/
2065 void e1000e_set_interrupt_capability(struct e1000_adapter *adapter)
2066 {
2067 	int err;
2068 	int i;
2069 
2070 	switch (adapter->int_mode) {
2071 	case E1000E_INT_MODE_MSIX:
2072 		if (adapter->flags & FLAG_HAS_MSIX) {
2073 			adapter->num_vectors = 3; /* RxQ0, TxQ0 and other */
2074 			adapter->msix_entries = kcalloc(adapter->num_vectors,
2075 							sizeof(struct
2076 							       msix_entry),
2077 							GFP_KERNEL);
2078 			if (adapter->msix_entries) {
2079 				struct e1000_adapter *a = adapter;
2080 
2081 				for (i = 0; i < adapter->num_vectors; i++)
2082 					adapter->msix_entries[i].entry = i;
2083 
2084 				err = pci_enable_msix_range(a->pdev,
2085 							    a->msix_entries,
2086 							    a->num_vectors,
2087 							    a->num_vectors);
2088 				if (err > 0)
2089 					return;
2090 			}
2091 			/* MSI-X failed, so fall through and try MSI */
2092 			e_err("Failed to initialize MSI-X interrupts.  Falling back to MSI interrupts.\n");
2093 			e1000e_reset_interrupt_capability(adapter);
2094 		}
2095 		adapter->int_mode = E1000E_INT_MODE_MSI;
2096 		/* Fall through */
2097 	case E1000E_INT_MODE_MSI:
2098 		if (!pci_enable_msi(adapter->pdev)) {
2099 			adapter->flags |= FLAG_MSI_ENABLED;
2100 		} else {
2101 			adapter->int_mode = E1000E_INT_MODE_LEGACY;
2102 			e_err("Failed to initialize MSI interrupts.  Falling back to legacy interrupts.\n");
2103 		}
2104 		/* Fall through */
2105 	case E1000E_INT_MODE_LEGACY:
2106 		/* Don't do anything; this is the system default */
2107 		break;
2108 	}
2109 
2110 	/* store the number of vectors being used */
2111 	adapter->num_vectors = 1;
2112 }
2113 
2114 /**
2115  * e1000_request_msix - Initialize MSI-X interrupts
2116  *
2117  * e1000_request_msix allocates MSI-X vectors and requests interrupts from the
2118  * kernel.
2119  **/
2120 static int e1000_request_msix(struct e1000_adapter *adapter)
2121 {
2122 	struct net_device *netdev = adapter->netdev;
2123 	int err = 0, vector = 0;
2124 
2125 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2126 		snprintf(adapter->rx_ring->name,
2127 			 sizeof(adapter->rx_ring->name) - 1,
2128 			 "%s-rx-0", netdev->name);
2129 	else
2130 		memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
2131 	err = request_irq(adapter->msix_entries[vector].vector,
2132 			  e1000_intr_msix_rx, 0, adapter->rx_ring->name,
2133 			  netdev);
2134 	if (err)
2135 		return err;
2136 	adapter->rx_ring->itr_register = adapter->hw.hw_addr +
2137 	    E1000_EITR_82574(vector);
2138 	adapter->rx_ring->itr_val = adapter->itr;
2139 	vector++;
2140 
2141 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2142 		snprintf(adapter->tx_ring->name,
2143 			 sizeof(adapter->tx_ring->name) - 1,
2144 			 "%s-tx-0", netdev->name);
2145 	else
2146 		memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
2147 	err = request_irq(adapter->msix_entries[vector].vector,
2148 			  e1000_intr_msix_tx, 0, adapter->tx_ring->name,
2149 			  netdev);
2150 	if (err)
2151 		return err;
2152 	adapter->tx_ring->itr_register = adapter->hw.hw_addr +
2153 	    E1000_EITR_82574(vector);
2154 	adapter->tx_ring->itr_val = adapter->itr;
2155 	vector++;
2156 
2157 	err = request_irq(adapter->msix_entries[vector].vector,
2158 			  e1000_msix_other, 0, netdev->name, netdev);
2159 	if (err)
2160 		return err;
2161 
2162 	e1000_configure_msix(adapter);
2163 
2164 	return 0;
2165 }
2166 
2167 /**
2168  * e1000_request_irq - initialize interrupts
2169  *
2170  * Attempts to configure interrupts using the best available
2171  * capabilities of the hardware and kernel.
2172  **/
2173 static int e1000_request_irq(struct e1000_adapter *adapter)
2174 {
2175 	struct net_device *netdev = adapter->netdev;
2176 	int err;
2177 
2178 	if (adapter->msix_entries) {
2179 		err = e1000_request_msix(adapter);
2180 		if (!err)
2181 			return err;
2182 		/* fall back to MSI */
2183 		e1000e_reset_interrupt_capability(adapter);
2184 		adapter->int_mode = E1000E_INT_MODE_MSI;
2185 		e1000e_set_interrupt_capability(adapter);
2186 	}
2187 	if (adapter->flags & FLAG_MSI_ENABLED) {
2188 		err = request_irq(adapter->pdev->irq, e1000_intr_msi, 0,
2189 				  netdev->name, netdev);
2190 		if (!err)
2191 			return err;
2192 
2193 		/* fall back to legacy interrupt */
2194 		e1000e_reset_interrupt_capability(adapter);
2195 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
2196 	}
2197 
2198 	err = request_irq(adapter->pdev->irq, e1000_intr, IRQF_SHARED,
2199 			  netdev->name, netdev);
2200 	if (err)
2201 		e_err("Unable to allocate interrupt, Error: %d\n", err);
2202 
2203 	return err;
2204 }
2205 
2206 static void e1000_free_irq(struct e1000_adapter *adapter)
2207 {
2208 	struct net_device *netdev = adapter->netdev;
2209 
2210 	if (adapter->msix_entries) {
2211 		int vector = 0;
2212 
2213 		free_irq(adapter->msix_entries[vector].vector, netdev);
2214 		vector++;
2215 
2216 		free_irq(adapter->msix_entries[vector].vector, netdev);
2217 		vector++;
2218 
2219 		/* Other Causes interrupt vector */
2220 		free_irq(adapter->msix_entries[vector].vector, netdev);
2221 		return;
2222 	}
2223 
2224 	free_irq(adapter->pdev->irq, netdev);
2225 }
2226 
2227 /**
2228  * e1000_irq_disable - Mask off interrupt generation on the NIC
2229  **/
2230 static void e1000_irq_disable(struct e1000_adapter *adapter)
2231 {
2232 	struct e1000_hw *hw = &adapter->hw;
2233 
2234 	ew32(IMC, ~0);
2235 	if (adapter->msix_entries)
2236 		ew32(EIAC_82574, 0);
2237 	e1e_flush();
2238 
2239 	if (adapter->msix_entries) {
2240 		int i;
2241 
2242 		for (i = 0; i < adapter->num_vectors; i++)
2243 			synchronize_irq(adapter->msix_entries[i].vector);
2244 	} else {
2245 		synchronize_irq(adapter->pdev->irq);
2246 	}
2247 }
2248 
2249 /**
2250  * e1000_irq_enable - Enable default interrupt generation settings
2251  **/
2252 static void e1000_irq_enable(struct e1000_adapter *adapter)
2253 {
2254 	struct e1000_hw *hw = &adapter->hw;
2255 
2256 	if (adapter->msix_entries) {
2257 		ew32(EIAC_82574, adapter->eiac_mask & E1000_EIAC_MASK_82574);
2258 		ew32(IMS, adapter->eiac_mask | E1000_IMS_OTHER |
2259 		     IMS_OTHER_MASK);
2260 	} else if (hw->mac.type >= e1000_pch_lpt) {
2261 		ew32(IMS, IMS_ENABLE_MASK | E1000_IMS_ECCER);
2262 	} else {
2263 		ew32(IMS, IMS_ENABLE_MASK);
2264 	}
2265 	e1e_flush();
2266 }
2267 
2268 /**
2269  * e1000e_get_hw_control - get control of the h/w from f/w
2270  * @adapter: address of board private structure
2271  *
2272  * e1000e_get_hw_control sets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2273  * For ASF and Pass Through versions of f/w this means that
2274  * the driver is loaded. For AMT version (only with 82573)
2275  * of the f/w this means that the network i/f is open.
2276  **/
2277 void e1000e_get_hw_control(struct e1000_adapter *adapter)
2278 {
2279 	struct e1000_hw *hw = &adapter->hw;
2280 	u32 ctrl_ext;
2281 	u32 swsm;
2282 
2283 	/* Let firmware know the driver has taken over */
2284 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2285 		swsm = er32(SWSM);
2286 		ew32(SWSM, swsm | E1000_SWSM_DRV_LOAD);
2287 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2288 		ctrl_ext = er32(CTRL_EXT);
2289 		ew32(CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2290 	}
2291 }
2292 
2293 /**
2294  * e1000e_release_hw_control - release control of the h/w to f/w
2295  * @adapter: address of board private structure
2296  *
2297  * e1000e_release_hw_control resets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2298  * For ASF and Pass Through versions of f/w this means that the
2299  * driver is no longer loaded. For AMT version (only with 82573) i
2300  * of the f/w this means that the network i/f is closed.
2301  *
2302  **/
2303 void e1000e_release_hw_control(struct e1000_adapter *adapter)
2304 {
2305 	struct e1000_hw *hw = &adapter->hw;
2306 	u32 ctrl_ext;
2307 	u32 swsm;
2308 
2309 	/* Let firmware taken over control of h/w */
2310 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2311 		swsm = er32(SWSM);
2312 		ew32(SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
2313 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2314 		ctrl_ext = er32(CTRL_EXT);
2315 		ew32(CTRL_EXT, ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2316 	}
2317 }
2318 
2319 /**
2320  * e1000_alloc_ring_dma - allocate memory for a ring structure
2321  **/
2322 static int e1000_alloc_ring_dma(struct e1000_adapter *adapter,
2323 				struct e1000_ring *ring)
2324 {
2325 	struct pci_dev *pdev = adapter->pdev;
2326 
2327 	ring->desc = dma_zalloc_coherent(&pdev->dev, ring->size, &ring->dma,
2328 					 GFP_KERNEL);
2329 	if (!ring->desc)
2330 		return -ENOMEM;
2331 
2332 	return 0;
2333 }
2334 
2335 /**
2336  * e1000e_setup_tx_resources - allocate Tx resources (Descriptors)
2337  * @tx_ring: Tx descriptor ring
2338  *
2339  * Return 0 on success, negative on failure
2340  **/
2341 int e1000e_setup_tx_resources(struct e1000_ring *tx_ring)
2342 {
2343 	struct e1000_adapter *adapter = tx_ring->adapter;
2344 	int err = -ENOMEM, size;
2345 
2346 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2347 	tx_ring->buffer_info = vzalloc(size);
2348 	if (!tx_ring->buffer_info)
2349 		goto err;
2350 
2351 	/* round up to nearest 4K */
2352 	tx_ring->size = tx_ring->count * sizeof(struct e1000_tx_desc);
2353 	tx_ring->size = ALIGN(tx_ring->size, 4096);
2354 
2355 	err = e1000_alloc_ring_dma(adapter, tx_ring);
2356 	if (err)
2357 		goto err;
2358 
2359 	tx_ring->next_to_use = 0;
2360 	tx_ring->next_to_clean = 0;
2361 
2362 	return 0;
2363 err:
2364 	vfree(tx_ring->buffer_info);
2365 	e_err("Unable to allocate memory for the transmit descriptor ring\n");
2366 	return err;
2367 }
2368 
2369 /**
2370  * e1000e_setup_rx_resources - allocate Rx resources (Descriptors)
2371  * @rx_ring: Rx descriptor ring
2372  *
2373  * Returns 0 on success, negative on failure
2374  **/
2375 int e1000e_setup_rx_resources(struct e1000_ring *rx_ring)
2376 {
2377 	struct e1000_adapter *adapter = rx_ring->adapter;
2378 	struct e1000_buffer *buffer_info;
2379 	int i, size, desc_len, err = -ENOMEM;
2380 
2381 	size = sizeof(struct e1000_buffer) * rx_ring->count;
2382 	rx_ring->buffer_info = vzalloc(size);
2383 	if (!rx_ring->buffer_info)
2384 		goto err;
2385 
2386 	for (i = 0; i < rx_ring->count; i++) {
2387 		buffer_info = &rx_ring->buffer_info[i];
2388 		buffer_info->ps_pages = kcalloc(PS_PAGE_BUFFERS,
2389 						sizeof(struct e1000_ps_page),
2390 						GFP_KERNEL);
2391 		if (!buffer_info->ps_pages)
2392 			goto err_pages;
2393 	}
2394 
2395 	desc_len = sizeof(union e1000_rx_desc_packet_split);
2396 
2397 	/* Round up to nearest 4K */
2398 	rx_ring->size = rx_ring->count * desc_len;
2399 	rx_ring->size = ALIGN(rx_ring->size, 4096);
2400 
2401 	err = e1000_alloc_ring_dma(adapter, rx_ring);
2402 	if (err)
2403 		goto err_pages;
2404 
2405 	rx_ring->next_to_clean = 0;
2406 	rx_ring->next_to_use = 0;
2407 	rx_ring->rx_skb_top = NULL;
2408 
2409 	return 0;
2410 
2411 err_pages:
2412 	for (i = 0; i < rx_ring->count; i++) {
2413 		buffer_info = &rx_ring->buffer_info[i];
2414 		kfree(buffer_info->ps_pages);
2415 	}
2416 err:
2417 	vfree(rx_ring->buffer_info);
2418 	e_err("Unable to allocate memory for the receive descriptor ring\n");
2419 	return err;
2420 }
2421 
2422 /**
2423  * e1000_clean_tx_ring - Free Tx Buffers
2424  * @tx_ring: Tx descriptor ring
2425  **/
2426 static void e1000_clean_tx_ring(struct e1000_ring *tx_ring)
2427 {
2428 	struct e1000_adapter *adapter = tx_ring->adapter;
2429 	struct e1000_buffer *buffer_info;
2430 	unsigned long size;
2431 	unsigned int i;
2432 
2433 	for (i = 0; i < tx_ring->count; i++) {
2434 		buffer_info = &tx_ring->buffer_info[i];
2435 		e1000_put_txbuf(tx_ring, buffer_info, false);
2436 	}
2437 
2438 	netdev_reset_queue(adapter->netdev);
2439 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2440 	memset(tx_ring->buffer_info, 0, size);
2441 
2442 	memset(tx_ring->desc, 0, tx_ring->size);
2443 
2444 	tx_ring->next_to_use = 0;
2445 	tx_ring->next_to_clean = 0;
2446 }
2447 
2448 /**
2449  * e1000e_free_tx_resources - Free Tx Resources per Queue
2450  * @tx_ring: Tx descriptor ring
2451  *
2452  * Free all transmit software resources
2453  **/
2454 void e1000e_free_tx_resources(struct e1000_ring *tx_ring)
2455 {
2456 	struct e1000_adapter *adapter = tx_ring->adapter;
2457 	struct pci_dev *pdev = adapter->pdev;
2458 
2459 	e1000_clean_tx_ring(tx_ring);
2460 
2461 	vfree(tx_ring->buffer_info);
2462 	tx_ring->buffer_info = NULL;
2463 
2464 	dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
2465 			  tx_ring->dma);
2466 	tx_ring->desc = NULL;
2467 }
2468 
2469 /**
2470  * e1000e_free_rx_resources - Free Rx Resources
2471  * @rx_ring: Rx descriptor ring
2472  *
2473  * Free all receive software resources
2474  **/
2475 void e1000e_free_rx_resources(struct e1000_ring *rx_ring)
2476 {
2477 	struct e1000_adapter *adapter = rx_ring->adapter;
2478 	struct pci_dev *pdev = adapter->pdev;
2479 	int i;
2480 
2481 	e1000_clean_rx_ring(rx_ring);
2482 
2483 	for (i = 0; i < rx_ring->count; i++)
2484 		kfree(rx_ring->buffer_info[i].ps_pages);
2485 
2486 	vfree(rx_ring->buffer_info);
2487 	rx_ring->buffer_info = NULL;
2488 
2489 	dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
2490 			  rx_ring->dma);
2491 	rx_ring->desc = NULL;
2492 }
2493 
2494 /**
2495  * e1000_update_itr - update the dynamic ITR value based on statistics
2496  * @adapter: pointer to adapter
2497  * @itr_setting: current adapter->itr
2498  * @packets: the number of packets during this measurement interval
2499  * @bytes: the number of bytes during this measurement interval
2500  *
2501  *      Stores a new ITR value based on packets and byte
2502  *      counts during the last interrupt.  The advantage of per interrupt
2503  *      computation is faster updates and more accurate ITR for the current
2504  *      traffic pattern.  Constants in this function were computed
2505  *      based on theoretical maximum wire speed and thresholds were set based
2506  *      on testing data as well as attempting to minimize response time
2507  *      while increasing bulk throughput.  This functionality is controlled
2508  *      by the InterruptThrottleRate module parameter.
2509  **/
2510 static unsigned int e1000_update_itr(u16 itr_setting, int packets, int bytes)
2511 {
2512 	unsigned int retval = itr_setting;
2513 
2514 	if (packets == 0)
2515 		return itr_setting;
2516 
2517 	switch (itr_setting) {
2518 	case lowest_latency:
2519 		/* handle TSO and jumbo frames */
2520 		if (bytes / packets > 8000)
2521 			retval = bulk_latency;
2522 		else if ((packets < 5) && (bytes > 512))
2523 			retval = low_latency;
2524 		break;
2525 	case low_latency:	/* 50 usec aka 20000 ints/s */
2526 		if (bytes > 10000) {
2527 			/* this if handles the TSO accounting */
2528 			if (bytes / packets > 8000)
2529 				retval = bulk_latency;
2530 			else if ((packets < 10) || ((bytes / packets) > 1200))
2531 				retval = bulk_latency;
2532 			else if ((packets > 35))
2533 				retval = lowest_latency;
2534 		} else if (bytes / packets > 2000) {
2535 			retval = bulk_latency;
2536 		} else if (packets <= 2 && bytes < 512) {
2537 			retval = lowest_latency;
2538 		}
2539 		break;
2540 	case bulk_latency:	/* 250 usec aka 4000 ints/s */
2541 		if (bytes > 25000) {
2542 			if (packets > 35)
2543 				retval = low_latency;
2544 		} else if (bytes < 6000) {
2545 			retval = low_latency;
2546 		}
2547 		break;
2548 	}
2549 
2550 	return retval;
2551 }
2552 
2553 static void e1000_set_itr(struct e1000_adapter *adapter)
2554 {
2555 	u16 current_itr;
2556 	u32 new_itr = adapter->itr;
2557 
2558 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
2559 	if (adapter->link_speed != SPEED_1000) {
2560 		current_itr = 0;
2561 		new_itr = 4000;
2562 		goto set_itr_now;
2563 	}
2564 
2565 	if (adapter->flags2 & FLAG2_DISABLE_AIM) {
2566 		new_itr = 0;
2567 		goto set_itr_now;
2568 	}
2569 
2570 	adapter->tx_itr = e1000_update_itr(adapter->tx_itr,
2571 					   adapter->total_tx_packets,
2572 					   adapter->total_tx_bytes);
2573 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2574 	if (adapter->itr_setting == 3 && adapter->tx_itr == lowest_latency)
2575 		adapter->tx_itr = low_latency;
2576 
2577 	adapter->rx_itr = e1000_update_itr(adapter->rx_itr,
2578 					   adapter->total_rx_packets,
2579 					   adapter->total_rx_bytes);
2580 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2581 	if (adapter->itr_setting == 3 && adapter->rx_itr == lowest_latency)
2582 		adapter->rx_itr = low_latency;
2583 
2584 	current_itr = max(adapter->rx_itr, adapter->tx_itr);
2585 
2586 	/* counts and packets in update_itr are dependent on these numbers */
2587 	switch (current_itr) {
2588 	case lowest_latency:
2589 		new_itr = 70000;
2590 		break;
2591 	case low_latency:
2592 		new_itr = 20000;	/* aka hwitr = ~200 */
2593 		break;
2594 	case bulk_latency:
2595 		new_itr = 4000;
2596 		break;
2597 	default:
2598 		break;
2599 	}
2600 
2601 set_itr_now:
2602 	if (new_itr != adapter->itr) {
2603 		/* this attempts to bias the interrupt rate towards Bulk
2604 		 * by adding intermediate steps when interrupt rate is
2605 		 * increasing
2606 		 */
2607 		new_itr = new_itr > adapter->itr ?
2608 		    min(adapter->itr + (new_itr >> 2), new_itr) : new_itr;
2609 		adapter->itr = new_itr;
2610 		adapter->rx_ring->itr_val = new_itr;
2611 		if (adapter->msix_entries)
2612 			adapter->rx_ring->set_itr = 1;
2613 		else
2614 			e1000e_write_itr(adapter, new_itr);
2615 	}
2616 }
2617 
2618 /**
2619  * e1000e_write_itr - write the ITR value to the appropriate registers
2620  * @adapter: address of board private structure
2621  * @itr: new ITR value to program
2622  *
2623  * e1000e_write_itr determines if the adapter is in MSI-X mode
2624  * and, if so, writes the EITR registers with the ITR value.
2625  * Otherwise, it writes the ITR value into the ITR register.
2626  **/
2627 void e1000e_write_itr(struct e1000_adapter *adapter, u32 itr)
2628 {
2629 	struct e1000_hw *hw = &adapter->hw;
2630 	u32 new_itr = itr ? 1000000000 / (itr * 256) : 0;
2631 
2632 	if (adapter->msix_entries) {
2633 		int vector;
2634 
2635 		for (vector = 0; vector < adapter->num_vectors; vector++)
2636 			writel(new_itr, hw->hw_addr + E1000_EITR_82574(vector));
2637 	} else {
2638 		ew32(ITR, new_itr);
2639 	}
2640 }
2641 
2642 /**
2643  * e1000_alloc_queues - Allocate memory for all rings
2644  * @adapter: board private structure to initialize
2645  **/
2646 static int e1000_alloc_queues(struct e1000_adapter *adapter)
2647 {
2648 	int size = sizeof(struct e1000_ring);
2649 
2650 	adapter->tx_ring = kzalloc(size, GFP_KERNEL);
2651 	if (!adapter->tx_ring)
2652 		goto err;
2653 	adapter->tx_ring->count = adapter->tx_ring_count;
2654 	adapter->tx_ring->adapter = adapter;
2655 
2656 	adapter->rx_ring = kzalloc(size, GFP_KERNEL);
2657 	if (!adapter->rx_ring)
2658 		goto err;
2659 	adapter->rx_ring->count = adapter->rx_ring_count;
2660 	adapter->rx_ring->adapter = adapter;
2661 
2662 	return 0;
2663 err:
2664 	e_err("Unable to allocate memory for queues\n");
2665 	kfree(adapter->rx_ring);
2666 	kfree(adapter->tx_ring);
2667 	return -ENOMEM;
2668 }
2669 
2670 /**
2671  * e1000e_poll - NAPI Rx polling callback
2672  * @napi: struct associated with this polling callback
2673  * @weight: number of packets driver is allowed to process this poll
2674  **/
2675 static int e1000e_poll(struct napi_struct *napi, int weight)
2676 {
2677 	struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter,
2678 						     napi);
2679 	struct e1000_hw *hw = &adapter->hw;
2680 	struct net_device *poll_dev = adapter->netdev;
2681 	int tx_cleaned = 1, work_done = 0;
2682 
2683 	adapter = netdev_priv(poll_dev);
2684 
2685 	if (!adapter->msix_entries ||
2686 	    (adapter->rx_ring->ims_val & adapter->tx_ring->ims_val))
2687 		tx_cleaned = e1000_clean_tx_irq(adapter->tx_ring);
2688 
2689 	adapter->clean_rx(adapter->rx_ring, &work_done, weight);
2690 
2691 	if (!tx_cleaned)
2692 		work_done = weight;
2693 
2694 	/* If weight not fully consumed, exit the polling mode */
2695 	if (work_done < weight) {
2696 		if (adapter->itr_setting & 3)
2697 			e1000_set_itr(adapter);
2698 		napi_complete_done(napi, work_done);
2699 		if (!test_bit(__E1000_DOWN, &adapter->state)) {
2700 			if (adapter->msix_entries)
2701 				ew32(IMS, adapter->rx_ring->ims_val);
2702 			else
2703 				e1000_irq_enable(adapter);
2704 		}
2705 	}
2706 
2707 	return work_done;
2708 }
2709 
2710 static int e1000_vlan_rx_add_vid(struct net_device *netdev,
2711 				 __always_unused __be16 proto, u16 vid)
2712 {
2713 	struct e1000_adapter *adapter = netdev_priv(netdev);
2714 	struct e1000_hw *hw = &adapter->hw;
2715 	u32 vfta, index;
2716 
2717 	/* don't update vlan cookie if already programmed */
2718 	if ((adapter->hw.mng_cookie.status &
2719 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2720 	    (vid == adapter->mng_vlan_id))
2721 		return 0;
2722 
2723 	/* add VID to filter table */
2724 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2725 		index = (vid >> 5) & 0x7F;
2726 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2727 		vfta |= BIT((vid & 0x1F));
2728 		hw->mac.ops.write_vfta(hw, index, vfta);
2729 	}
2730 
2731 	set_bit(vid, adapter->active_vlans);
2732 
2733 	return 0;
2734 }
2735 
2736 static int e1000_vlan_rx_kill_vid(struct net_device *netdev,
2737 				  __always_unused __be16 proto, u16 vid)
2738 {
2739 	struct e1000_adapter *adapter = netdev_priv(netdev);
2740 	struct e1000_hw *hw = &adapter->hw;
2741 	u32 vfta, index;
2742 
2743 	if ((adapter->hw.mng_cookie.status &
2744 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2745 	    (vid == adapter->mng_vlan_id)) {
2746 		/* release control to f/w */
2747 		e1000e_release_hw_control(adapter);
2748 		return 0;
2749 	}
2750 
2751 	/* remove VID from filter table */
2752 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2753 		index = (vid >> 5) & 0x7F;
2754 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2755 		vfta &= ~BIT((vid & 0x1F));
2756 		hw->mac.ops.write_vfta(hw, index, vfta);
2757 	}
2758 
2759 	clear_bit(vid, adapter->active_vlans);
2760 
2761 	return 0;
2762 }
2763 
2764 /**
2765  * e1000e_vlan_filter_disable - helper to disable hw VLAN filtering
2766  * @adapter: board private structure to initialize
2767  **/
2768 static void e1000e_vlan_filter_disable(struct e1000_adapter *adapter)
2769 {
2770 	struct net_device *netdev = adapter->netdev;
2771 	struct e1000_hw *hw = &adapter->hw;
2772 	u32 rctl;
2773 
2774 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2775 		/* disable VLAN receive filtering */
2776 		rctl = er32(RCTL);
2777 		rctl &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
2778 		ew32(RCTL, rctl);
2779 
2780 		if (adapter->mng_vlan_id != (u16)E1000_MNG_VLAN_NONE) {
2781 			e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
2782 					       adapter->mng_vlan_id);
2783 			adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
2784 		}
2785 	}
2786 }
2787 
2788 /**
2789  * e1000e_vlan_filter_enable - helper to enable HW VLAN filtering
2790  * @adapter: board private structure to initialize
2791  **/
2792 static void e1000e_vlan_filter_enable(struct e1000_adapter *adapter)
2793 {
2794 	struct e1000_hw *hw = &adapter->hw;
2795 	u32 rctl;
2796 
2797 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2798 		/* enable VLAN receive filtering */
2799 		rctl = er32(RCTL);
2800 		rctl |= E1000_RCTL_VFE;
2801 		rctl &= ~E1000_RCTL_CFIEN;
2802 		ew32(RCTL, rctl);
2803 	}
2804 }
2805 
2806 /**
2807  * e1000e_vlan_strip_disable - helper to disable HW VLAN stripping
2808  * @adapter: board private structure to initialize
2809  **/
2810 static void e1000e_vlan_strip_disable(struct e1000_adapter *adapter)
2811 {
2812 	struct e1000_hw *hw = &adapter->hw;
2813 	u32 ctrl;
2814 
2815 	/* disable VLAN tag insert/strip */
2816 	ctrl = er32(CTRL);
2817 	ctrl &= ~E1000_CTRL_VME;
2818 	ew32(CTRL, ctrl);
2819 }
2820 
2821 /**
2822  * e1000e_vlan_strip_enable - helper to enable HW VLAN stripping
2823  * @adapter: board private structure to initialize
2824  **/
2825 static void e1000e_vlan_strip_enable(struct e1000_adapter *adapter)
2826 {
2827 	struct e1000_hw *hw = &adapter->hw;
2828 	u32 ctrl;
2829 
2830 	/* enable VLAN tag insert/strip */
2831 	ctrl = er32(CTRL);
2832 	ctrl |= E1000_CTRL_VME;
2833 	ew32(CTRL, ctrl);
2834 }
2835 
2836 static void e1000_update_mng_vlan(struct e1000_adapter *adapter)
2837 {
2838 	struct net_device *netdev = adapter->netdev;
2839 	u16 vid = adapter->hw.mng_cookie.vlan_id;
2840 	u16 old_vid = adapter->mng_vlan_id;
2841 
2842 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
2843 		e1000_vlan_rx_add_vid(netdev, htons(ETH_P_8021Q), vid);
2844 		adapter->mng_vlan_id = vid;
2845 	}
2846 
2847 	if ((old_vid != (u16)E1000_MNG_VLAN_NONE) && (vid != old_vid))
2848 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q), old_vid);
2849 }
2850 
2851 static void e1000_restore_vlan(struct e1000_adapter *adapter)
2852 {
2853 	u16 vid;
2854 
2855 	e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
2856 
2857 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
2858 	    e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
2859 }
2860 
2861 static void e1000_init_manageability_pt(struct e1000_adapter *adapter)
2862 {
2863 	struct e1000_hw *hw = &adapter->hw;
2864 	u32 manc, manc2h, mdef, i, j;
2865 
2866 	if (!(adapter->flags & FLAG_MNG_PT_ENABLED))
2867 		return;
2868 
2869 	manc = er32(MANC);
2870 
2871 	/* enable receiving management packets to the host. this will probably
2872 	 * generate destination unreachable messages from the host OS, but
2873 	 * the packets will be handled on SMBUS
2874 	 */
2875 	manc |= E1000_MANC_EN_MNG2HOST;
2876 	manc2h = er32(MANC2H);
2877 
2878 	switch (hw->mac.type) {
2879 	default:
2880 		manc2h |= (E1000_MANC2H_PORT_623 | E1000_MANC2H_PORT_664);
2881 		break;
2882 	case e1000_82574:
2883 	case e1000_82583:
2884 		/* Check if IPMI pass-through decision filter already exists;
2885 		 * if so, enable it.
2886 		 */
2887 		for (i = 0, j = 0; i < 8; i++) {
2888 			mdef = er32(MDEF(i));
2889 
2890 			/* Ignore filters with anything other than IPMI ports */
2891 			if (mdef & ~(E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2892 				continue;
2893 
2894 			/* Enable this decision filter in MANC2H */
2895 			if (mdef)
2896 				manc2h |= BIT(i);
2897 
2898 			j |= mdef;
2899 		}
2900 
2901 		if (j == (E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2902 			break;
2903 
2904 		/* Create new decision filter in an empty filter */
2905 		for (i = 0, j = 0; i < 8; i++)
2906 			if (er32(MDEF(i)) == 0) {
2907 				ew32(MDEF(i), (E1000_MDEF_PORT_623 |
2908 					       E1000_MDEF_PORT_664));
2909 				manc2h |= BIT(1);
2910 				j++;
2911 				break;
2912 			}
2913 
2914 		if (!j)
2915 			e_warn("Unable to create IPMI pass-through filter\n");
2916 		break;
2917 	}
2918 
2919 	ew32(MANC2H, manc2h);
2920 	ew32(MANC, manc);
2921 }
2922 
2923 /**
2924  * e1000_configure_tx - Configure Transmit Unit after Reset
2925  * @adapter: board private structure
2926  *
2927  * Configure the Tx unit of the MAC after a reset.
2928  **/
2929 static void e1000_configure_tx(struct e1000_adapter *adapter)
2930 {
2931 	struct e1000_hw *hw = &adapter->hw;
2932 	struct e1000_ring *tx_ring = adapter->tx_ring;
2933 	u64 tdba;
2934 	u32 tdlen, tctl, tarc;
2935 
2936 	/* Setup the HW Tx Head and Tail descriptor pointers */
2937 	tdba = tx_ring->dma;
2938 	tdlen = tx_ring->count * sizeof(struct e1000_tx_desc);
2939 	ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
2940 	ew32(TDBAH(0), (tdba >> 32));
2941 	ew32(TDLEN(0), tdlen);
2942 	ew32(TDH(0), 0);
2943 	ew32(TDT(0), 0);
2944 	tx_ring->head = adapter->hw.hw_addr + E1000_TDH(0);
2945 	tx_ring->tail = adapter->hw.hw_addr + E1000_TDT(0);
2946 
2947 	writel(0, tx_ring->head);
2948 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
2949 		e1000e_update_tdt_wa(tx_ring, 0);
2950 	else
2951 		writel(0, tx_ring->tail);
2952 
2953 	/* Set the Tx Interrupt Delay register */
2954 	ew32(TIDV, adapter->tx_int_delay);
2955 	/* Tx irq moderation */
2956 	ew32(TADV, adapter->tx_abs_int_delay);
2957 
2958 	if (adapter->flags2 & FLAG2_DMA_BURST) {
2959 		u32 txdctl = er32(TXDCTL(0));
2960 
2961 		txdctl &= ~(E1000_TXDCTL_PTHRESH | E1000_TXDCTL_HTHRESH |
2962 			    E1000_TXDCTL_WTHRESH);
2963 		/* set up some performance related parameters to encourage the
2964 		 * hardware to use the bus more efficiently in bursts, depends
2965 		 * on the tx_int_delay to be enabled,
2966 		 * wthresh = 1 ==> burst write is disabled to avoid Tx stalls
2967 		 * hthresh = 1 ==> prefetch when one or more available
2968 		 * pthresh = 0x1f ==> prefetch if internal cache 31 or less
2969 		 * BEWARE: this seems to work but should be considered first if
2970 		 * there are Tx hangs or other Tx related bugs
2971 		 */
2972 		txdctl |= E1000_TXDCTL_DMA_BURST_ENABLE;
2973 		ew32(TXDCTL(0), txdctl);
2974 	}
2975 	/* erratum work around: set txdctl the same for both queues */
2976 	ew32(TXDCTL(1), er32(TXDCTL(0)));
2977 
2978 	/* Program the Transmit Control Register */
2979 	tctl = er32(TCTL);
2980 	tctl &= ~E1000_TCTL_CT;
2981 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
2982 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
2983 
2984 	if (adapter->flags & FLAG_TARC_SPEED_MODE_BIT) {
2985 		tarc = er32(TARC(0));
2986 		/* set the speed mode bit, we'll clear it if we're not at
2987 		 * gigabit link later
2988 		 */
2989 #define SPEED_MODE_BIT BIT(21)
2990 		tarc |= SPEED_MODE_BIT;
2991 		ew32(TARC(0), tarc);
2992 	}
2993 
2994 	/* errata: program both queues to unweighted RR */
2995 	if (adapter->flags & FLAG_TARC_SET_BIT_ZERO) {
2996 		tarc = er32(TARC(0));
2997 		tarc |= 1;
2998 		ew32(TARC(0), tarc);
2999 		tarc = er32(TARC(1));
3000 		tarc |= 1;
3001 		ew32(TARC(1), tarc);
3002 	}
3003 
3004 	/* Setup Transmit Descriptor Settings for eop descriptor */
3005 	adapter->txd_cmd = E1000_TXD_CMD_EOP | E1000_TXD_CMD_IFCS;
3006 
3007 	/* only set IDE if we are delaying interrupts using the timers */
3008 	if (adapter->tx_int_delay)
3009 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
3010 
3011 	/* enable Report Status bit */
3012 	adapter->txd_cmd |= E1000_TXD_CMD_RS;
3013 
3014 	ew32(TCTL, tctl);
3015 
3016 	hw->mac.ops.config_collision_dist(hw);
3017 
3018 	/* SPT and KBL Si errata workaround to avoid data corruption */
3019 	if (hw->mac.type == e1000_pch_spt) {
3020 		u32 reg_val;
3021 
3022 		reg_val = er32(IOSFPC);
3023 		reg_val |= E1000_RCTL_RDMTS_HEX;
3024 		ew32(IOSFPC, reg_val);
3025 
3026 		reg_val = er32(TARC(0));
3027 		/* SPT and KBL Si errata workaround to avoid Tx hang.
3028 		 * Dropping the number of outstanding requests from
3029 		 * 3 to 2 in order to avoid a buffer overrun.
3030 		 */
3031 		reg_val &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3032 		reg_val |= E1000_TARC0_CB_MULTIQ_2_REQ;
3033 		ew32(TARC(0), reg_val);
3034 	}
3035 }
3036 
3037 /**
3038  * e1000_setup_rctl - configure the receive control registers
3039  * @adapter: Board private structure
3040  **/
3041 #define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
3042 			   (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
3043 static void e1000_setup_rctl(struct e1000_adapter *adapter)
3044 {
3045 	struct e1000_hw *hw = &adapter->hw;
3046 	u32 rctl, rfctl;
3047 	u32 pages = 0;
3048 
3049 	/* Workaround Si errata on PCHx - configure jumbo frame flow.
3050 	 * If jumbo frames not set, program related MAC/PHY registers
3051 	 * to h/w defaults
3052 	 */
3053 	if (hw->mac.type >= e1000_pch2lan) {
3054 		s32 ret_val;
3055 
3056 		if (adapter->netdev->mtu > ETH_DATA_LEN)
3057 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, true);
3058 		else
3059 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, false);
3060 
3061 		if (ret_val)
3062 			e_dbg("failed to enable|disable jumbo frame workaround mode\n");
3063 	}
3064 
3065 	/* Program MC offset vector base */
3066 	rctl = er32(RCTL);
3067 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3068 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3069 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3070 	    (adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3071 
3072 	/* Do not Store bad packets */
3073 	rctl &= ~E1000_RCTL_SBP;
3074 
3075 	/* Enable Long Packet receive */
3076 	if (adapter->netdev->mtu <= ETH_DATA_LEN)
3077 		rctl &= ~E1000_RCTL_LPE;
3078 	else
3079 		rctl |= E1000_RCTL_LPE;
3080 
3081 	/* Some systems expect that the CRC is included in SMBUS traffic. The
3082 	 * hardware strips the CRC before sending to both SMBUS (BMC) and to
3083 	 * host memory when this is enabled
3084 	 */
3085 	if (adapter->flags2 & FLAG2_CRC_STRIPPING)
3086 		rctl |= E1000_RCTL_SECRC;
3087 
3088 	/* Workaround Si errata on 82577 PHY - configure IPG for jumbos */
3089 	if ((hw->phy.type == e1000_phy_82577) && (rctl & E1000_RCTL_LPE)) {
3090 		u16 phy_data;
3091 
3092 		e1e_rphy(hw, PHY_REG(770, 26), &phy_data);
3093 		phy_data &= 0xfff8;
3094 		phy_data |= BIT(2);
3095 		e1e_wphy(hw, PHY_REG(770, 26), phy_data);
3096 
3097 		e1e_rphy(hw, 22, &phy_data);
3098 		phy_data &= 0x0fff;
3099 		phy_data |= BIT(14);
3100 		e1e_wphy(hw, 0x10, 0x2823);
3101 		e1e_wphy(hw, 0x11, 0x0003);
3102 		e1e_wphy(hw, 22, phy_data);
3103 	}
3104 
3105 	/* Setup buffer sizes */
3106 	rctl &= ~E1000_RCTL_SZ_4096;
3107 	rctl |= E1000_RCTL_BSEX;
3108 	switch (adapter->rx_buffer_len) {
3109 	case 2048:
3110 	default:
3111 		rctl |= E1000_RCTL_SZ_2048;
3112 		rctl &= ~E1000_RCTL_BSEX;
3113 		break;
3114 	case 4096:
3115 		rctl |= E1000_RCTL_SZ_4096;
3116 		break;
3117 	case 8192:
3118 		rctl |= E1000_RCTL_SZ_8192;
3119 		break;
3120 	case 16384:
3121 		rctl |= E1000_RCTL_SZ_16384;
3122 		break;
3123 	}
3124 
3125 	/* Enable Extended Status in all Receive Descriptors */
3126 	rfctl = er32(RFCTL);
3127 	rfctl |= E1000_RFCTL_EXTEN;
3128 	ew32(RFCTL, rfctl);
3129 
3130 	/* 82571 and greater support packet-split where the protocol
3131 	 * header is placed in skb->data and the packet data is
3132 	 * placed in pages hanging off of skb_shinfo(skb)->nr_frags.
3133 	 * In the case of a non-split, skb->data is linearly filled,
3134 	 * followed by the page buffers.  Therefore, skb->data is
3135 	 * sized to hold the largest protocol header.
3136 	 *
3137 	 * allocations using alloc_page take too long for regular MTU
3138 	 * so only enable packet split for jumbo frames
3139 	 *
3140 	 * Using pages when the page size is greater than 16k wastes
3141 	 * a lot of memory, since we allocate 3 pages at all times
3142 	 * per packet.
3143 	 */
3144 	pages = PAGE_USE_COUNT(adapter->netdev->mtu);
3145 	if ((pages <= 3) && (PAGE_SIZE <= 16384) && (rctl & E1000_RCTL_LPE))
3146 		adapter->rx_ps_pages = pages;
3147 	else
3148 		adapter->rx_ps_pages = 0;
3149 
3150 	if (adapter->rx_ps_pages) {
3151 		u32 psrctl = 0;
3152 
3153 		/* Enable Packet split descriptors */
3154 		rctl |= E1000_RCTL_DTYP_PS;
3155 
3156 		psrctl |= adapter->rx_ps_bsize0 >> E1000_PSRCTL_BSIZE0_SHIFT;
3157 
3158 		switch (adapter->rx_ps_pages) {
3159 		case 3:
3160 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE3_SHIFT;
3161 			/* fall-through */
3162 		case 2:
3163 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE2_SHIFT;
3164 			/* fall-through */
3165 		case 1:
3166 			psrctl |= PAGE_SIZE >> E1000_PSRCTL_BSIZE1_SHIFT;
3167 			break;
3168 		}
3169 
3170 		ew32(PSRCTL, psrctl);
3171 	}
3172 
3173 	/* This is useful for sniffing bad packets. */
3174 	if (adapter->netdev->features & NETIF_F_RXALL) {
3175 		/* UPE and MPE will be handled by normal PROMISC logic
3176 		 * in e1000e_set_rx_mode
3177 		 */
3178 		rctl |= (E1000_RCTL_SBP |	/* Receive bad packets */
3179 			 E1000_RCTL_BAM |	/* RX All Bcast Pkts */
3180 			 E1000_RCTL_PMCF);	/* RX All MAC Ctrl Pkts */
3181 
3182 		rctl &= ~(E1000_RCTL_VFE |	/* Disable VLAN filter */
3183 			  E1000_RCTL_DPF |	/* Allow filtered pause */
3184 			  E1000_RCTL_CFIEN);	/* Dis VLAN CFIEN Filter */
3185 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3186 		 * and that breaks VLANs.
3187 		 */
3188 	}
3189 
3190 	ew32(RCTL, rctl);
3191 	/* just started the receive unit, no need to restart */
3192 	adapter->flags &= ~FLAG_RESTART_NOW;
3193 }
3194 
3195 /**
3196  * e1000_configure_rx - Configure Receive Unit after Reset
3197  * @adapter: board private structure
3198  *
3199  * Configure the Rx unit of the MAC after a reset.
3200  **/
3201 static void e1000_configure_rx(struct e1000_adapter *adapter)
3202 {
3203 	struct e1000_hw *hw = &adapter->hw;
3204 	struct e1000_ring *rx_ring = adapter->rx_ring;
3205 	u64 rdba;
3206 	u32 rdlen, rctl, rxcsum, ctrl_ext;
3207 
3208 	if (adapter->rx_ps_pages) {
3209 		/* this is a 32 byte descriptor */
3210 		rdlen = rx_ring->count *
3211 		    sizeof(union e1000_rx_desc_packet_split);
3212 		adapter->clean_rx = e1000_clean_rx_irq_ps;
3213 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers_ps;
3214 	} else if (adapter->netdev->mtu > ETH_FRAME_LEN + ETH_FCS_LEN) {
3215 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3216 		adapter->clean_rx = e1000_clean_jumbo_rx_irq;
3217 		adapter->alloc_rx_buf = e1000_alloc_jumbo_rx_buffers;
3218 	} else {
3219 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3220 		adapter->clean_rx = e1000_clean_rx_irq;
3221 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers;
3222 	}
3223 
3224 	/* disable receives while setting up the descriptors */
3225 	rctl = er32(RCTL);
3226 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
3227 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
3228 	e1e_flush();
3229 	usleep_range(10000, 20000);
3230 
3231 	if (adapter->flags2 & FLAG2_DMA_BURST) {
3232 		/* set the writeback threshold (only takes effect if the RDTR
3233 		 * is set). set GRAN=1 and write back up to 0x4 worth, and
3234 		 * enable prefetching of 0x20 Rx descriptors
3235 		 * granularity = 01
3236 		 * wthresh = 04,
3237 		 * hthresh = 04,
3238 		 * pthresh = 0x20
3239 		 */
3240 		ew32(RXDCTL(0), E1000_RXDCTL_DMA_BURST_ENABLE);
3241 		ew32(RXDCTL(1), E1000_RXDCTL_DMA_BURST_ENABLE);
3242 	}
3243 
3244 	/* set the Receive Delay Timer Register */
3245 	ew32(RDTR, adapter->rx_int_delay);
3246 
3247 	/* irq moderation */
3248 	ew32(RADV, adapter->rx_abs_int_delay);
3249 	if ((adapter->itr_setting != 0) && (adapter->itr != 0))
3250 		e1000e_write_itr(adapter, adapter->itr);
3251 
3252 	ctrl_ext = er32(CTRL_EXT);
3253 	/* Auto-Mask interrupts upon ICR access */
3254 	ctrl_ext |= E1000_CTRL_EXT_IAME;
3255 	ew32(IAM, 0xffffffff);
3256 	ew32(CTRL_EXT, ctrl_ext);
3257 	e1e_flush();
3258 
3259 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
3260 	 * the Base and Length of the Rx Descriptor Ring
3261 	 */
3262 	rdba = rx_ring->dma;
3263 	ew32(RDBAL(0), (rdba & DMA_BIT_MASK(32)));
3264 	ew32(RDBAH(0), (rdba >> 32));
3265 	ew32(RDLEN(0), rdlen);
3266 	ew32(RDH(0), 0);
3267 	ew32(RDT(0), 0);
3268 	rx_ring->head = adapter->hw.hw_addr + E1000_RDH(0);
3269 	rx_ring->tail = adapter->hw.hw_addr + E1000_RDT(0);
3270 
3271 	writel(0, rx_ring->head);
3272 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
3273 		e1000e_update_rdt_wa(rx_ring, 0);
3274 	else
3275 		writel(0, rx_ring->tail);
3276 
3277 	/* Enable Receive Checksum Offload for TCP and UDP */
3278 	rxcsum = er32(RXCSUM);
3279 	if (adapter->netdev->features & NETIF_F_RXCSUM)
3280 		rxcsum |= E1000_RXCSUM_TUOFL;
3281 	else
3282 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3283 	ew32(RXCSUM, rxcsum);
3284 
3285 	/* With jumbo frames, excessive C-state transition latencies result
3286 	 * in dropped transactions.
3287 	 */
3288 	if (adapter->netdev->mtu > ETH_DATA_LEN) {
3289 		u32 lat =
3290 		    ((er32(PBA) & E1000_PBA_RXA_MASK) * 1024 -
3291 		     adapter->max_frame_size) * 8 / 1000;
3292 
3293 		if (adapter->flags & FLAG_IS_ICH) {
3294 			u32 rxdctl = er32(RXDCTL(0));
3295 
3296 			ew32(RXDCTL(0), rxdctl | 0x3 | BIT(8));
3297 		}
3298 
3299 		dev_info(&adapter->pdev->dev,
3300 			 "Some CPU C-states have been disabled in order to enable jumbo frames\n");
3301 		pm_qos_update_request(&adapter->pm_qos_req, lat);
3302 	} else {
3303 		pm_qos_update_request(&adapter->pm_qos_req,
3304 				      PM_QOS_DEFAULT_VALUE);
3305 	}
3306 
3307 	/* Enable Receives */
3308 	ew32(RCTL, rctl);
3309 }
3310 
3311 /**
3312  * e1000e_write_mc_addr_list - write multicast addresses to MTA
3313  * @netdev: network interface device structure
3314  *
3315  * Writes multicast address list to the MTA hash table.
3316  * Returns: -ENOMEM on failure
3317  *                0 on no addresses written
3318  *                X on writing X addresses to MTA
3319  */
3320 static int e1000e_write_mc_addr_list(struct net_device *netdev)
3321 {
3322 	struct e1000_adapter *adapter = netdev_priv(netdev);
3323 	struct e1000_hw *hw = &adapter->hw;
3324 	struct netdev_hw_addr *ha;
3325 	u8 *mta_list;
3326 	int i;
3327 
3328 	if (netdev_mc_empty(netdev)) {
3329 		/* nothing to program, so clear mc list */
3330 		hw->mac.ops.update_mc_addr_list(hw, NULL, 0);
3331 		return 0;
3332 	}
3333 
3334 	mta_list = kzalloc(netdev_mc_count(netdev) * ETH_ALEN, GFP_ATOMIC);
3335 	if (!mta_list)
3336 		return -ENOMEM;
3337 
3338 	/* update_mc_addr_list expects a packed array of only addresses. */
3339 	i = 0;
3340 	netdev_for_each_mc_addr(ha, netdev)
3341 	    memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3342 
3343 	hw->mac.ops.update_mc_addr_list(hw, mta_list, i);
3344 	kfree(mta_list);
3345 
3346 	return netdev_mc_count(netdev);
3347 }
3348 
3349 /**
3350  * e1000e_write_uc_addr_list - write unicast addresses to RAR table
3351  * @netdev: network interface device structure
3352  *
3353  * Writes unicast address list to the RAR table.
3354  * Returns: -ENOMEM on failure/insufficient address space
3355  *                0 on no addresses written
3356  *                X on writing X addresses to the RAR table
3357  **/
3358 static int e1000e_write_uc_addr_list(struct net_device *netdev)
3359 {
3360 	struct e1000_adapter *adapter = netdev_priv(netdev);
3361 	struct e1000_hw *hw = &adapter->hw;
3362 	unsigned int rar_entries;
3363 	int count = 0;
3364 
3365 	rar_entries = hw->mac.ops.rar_get_count(hw);
3366 
3367 	/* save a rar entry for our hardware address */
3368 	rar_entries--;
3369 
3370 	/* save a rar entry for the LAA workaround */
3371 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA)
3372 		rar_entries--;
3373 
3374 	/* return ENOMEM indicating insufficient memory for addresses */
3375 	if (netdev_uc_count(netdev) > rar_entries)
3376 		return -ENOMEM;
3377 
3378 	if (!netdev_uc_empty(netdev) && rar_entries) {
3379 		struct netdev_hw_addr *ha;
3380 
3381 		/* write the addresses in reverse order to avoid write
3382 		 * combining
3383 		 */
3384 		netdev_for_each_uc_addr(ha, netdev) {
3385 			int ret_val;
3386 
3387 			if (!rar_entries)
3388 				break;
3389 			ret_val = hw->mac.ops.rar_set(hw, ha->addr, rar_entries--);
3390 			if (ret_val < 0)
3391 				return -ENOMEM;
3392 			count++;
3393 		}
3394 	}
3395 
3396 	/* zero out the remaining RAR entries not used above */
3397 	for (; rar_entries > 0; rar_entries--) {
3398 		ew32(RAH(rar_entries), 0);
3399 		ew32(RAL(rar_entries), 0);
3400 	}
3401 	e1e_flush();
3402 
3403 	return count;
3404 }
3405 
3406 /**
3407  * e1000e_set_rx_mode - secondary unicast, Multicast and Promiscuous mode set
3408  * @netdev: network interface device structure
3409  *
3410  * The ndo_set_rx_mode entry point is called whenever the unicast or multicast
3411  * address list or the network interface flags are updated.  This routine is
3412  * responsible for configuring the hardware for proper unicast, multicast,
3413  * promiscuous mode, and all-multi behavior.
3414  **/
3415 static void e1000e_set_rx_mode(struct net_device *netdev)
3416 {
3417 	struct e1000_adapter *adapter = netdev_priv(netdev);
3418 	struct e1000_hw *hw = &adapter->hw;
3419 	u32 rctl;
3420 
3421 	if (pm_runtime_suspended(netdev->dev.parent))
3422 		return;
3423 
3424 	/* Check for Promiscuous and All Multicast modes */
3425 	rctl = er32(RCTL);
3426 
3427 	/* clear the affected bits */
3428 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
3429 
3430 	if (netdev->flags & IFF_PROMISC) {
3431 		rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
3432 		/* Do not hardware filter VLANs in promisc mode */
3433 		e1000e_vlan_filter_disable(adapter);
3434 	} else {
3435 		int count;
3436 
3437 		if (netdev->flags & IFF_ALLMULTI) {
3438 			rctl |= E1000_RCTL_MPE;
3439 		} else {
3440 			/* Write addresses to the MTA, if the attempt fails
3441 			 * then we should just turn on promiscuous mode so
3442 			 * that we can at least receive multicast traffic
3443 			 */
3444 			count = e1000e_write_mc_addr_list(netdev);
3445 			if (count < 0)
3446 				rctl |= E1000_RCTL_MPE;
3447 		}
3448 		e1000e_vlan_filter_enable(adapter);
3449 		/* Write addresses to available RAR registers, if there is not
3450 		 * sufficient space to store all the addresses then enable
3451 		 * unicast promiscuous mode
3452 		 */
3453 		count = e1000e_write_uc_addr_list(netdev);
3454 		if (count < 0)
3455 			rctl |= E1000_RCTL_UPE;
3456 	}
3457 
3458 	ew32(RCTL, rctl);
3459 
3460 	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
3461 		e1000e_vlan_strip_enable(adapter);
3462 	else
3463 		e1000e_vlan_strip_disable(adapter);
3464 }
3465 
3466 static void e1000e_setup_rss_hash(struct e1000_adapter *adapter)
3467 {
3468 	struct e1000_hw *hw = &adapter->hw;
3469 	u32 mrqc, rxcsum;
3470 	u32 rss_key[10];
3471 	int i;
3472 
3473 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3474 	for (i = 0; i < 10; i++)
3475 		ew32(RSSRK(i), rss_key[i]);
3476 
3477 	/* Direct all traffic to queue 0 */
3478 	for (i = 0; i < 32; i++)
3479 		ew32(RETA(i), 0);
3480 
3481 	/* Disable raw packet checksumming so that RSS hash is placed in
3482 	 * descriptor on writeback.
3483 	 */
3484 	rxcsum = er32(RXCSUM);
3485 	rxcsum |= E1000_RXCSUM_PCSD;
3486 
3487 	ew32(RXCSUM, rxcsum);
3488 
3489 	mrqc = (E1000_MRQC_RSS_FIELD_IPV4 |
3490 		E1000_MRQC_RSS_FIELD_IPV4_TCP |
3491 		E1000_MRQC_RSS_FIELD_IPV6 |
3492 		E1000_MRQC_RSS_FIELD_IPV6_TCP |
3493 		E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3494 
3495 	ew32(MRQC, mrqc);
3496 }
3497 
3498 /**
3499  * e1000e_get_base_timinca - get default SYSTIM time increment attributes
3500  * @adapter: board private structure
3501  * @timinca: pointer to returned time increment attributes
3502  *
3503  * Get attributes for incrementing the System Time Register SYSTIML/H at
3504  * the default base frequency, and set the cyclecounter shift value.
3505  **/
3506 s32 e1000e_get_base_timinca(struct e1000_adapter *adapter, u32 *timinca)
3507 {
3508 	struct e1000_hw *hw = &adapter->hw;
3509 	u32 incvalue, incperiod, shift;
3510 
3511 	/* Make sure clock is enabled on I217/I218/I219  before checking
3512 	 * the frequency
3513 	 */
3514 	if ((hw->mac.type >= e1000_pch_lpt) &&
3515 	    !(er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) &&
3516 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_ENABLED)) {
3517 		u32 fextnvm7 = er32(FEXTNVM7);
3518 
3519 		if (!(fextnvm7 & BIT(0))) {
3520 			ew32(FEXTNVM7, fextnvm7 | BIT(0));
3521 			e1e_flush();
3522 		}
3523 	}
3524 
3525 	switch (hw->mac.type) {
3526 	case e1000_pch2lan:
3527 		/* Stable 96MHz frequency */
3528 		incperiod = INCPERIOD_96MHZ;
3529 		incvalue = INCVALUE_96MHZ;
3530 		shift = INCVALUE_SHIFT_96MHZ;
3531 		adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3532 		break;
3533 	case e1000_pch_lpt:
3534 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3535 			/* Stable 96MHz frequency */
3536 			incperiod = INCPERIOD_96MHZ;
3537 			incvalue = INCVALUE_96MHZ;
3538 			shift = INCVALUE_SHIFT_96MHZ;
3539 			adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3540 		} else {
3541 			/* Stable 25MHz frequency */
3542 			incperiod = INCPERIOD_25MHZ;
3543 			incvalue = INCVALUE_25MHZ;
3544 			shift = INCVALUE_SHIFT_25MHZ;
3545 			adapter->cc.shift = shift;
3546 		}
3547 		break;
3548 	case e1000_pch_spt:
3549 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3550 			/* Stable 24MHz frequency */
3551 			incperiod = INCPERIOD_24MHZ;
3552 			incvalue = INCVALUE_24MHZ;
3553 			shift = INCVALUE_SHIFT_24MHZ;
3554 			adapter->cc.shift = shift;
3555 			break;
3556 		}
3557 		return -EINVAL;
3558 	case e1000_pch_cnp:
3559 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3560 			/* Stable 24MHz frequency */
3561 			incperiod = INCPERIOD_24MHZ;
3562 			incvalue = INCVALUE_24MHZ;
3563 			shift = INCVALUE_SHIFT_24MHZ;
3564 			adapter->cc.shift = shift;
3565 		} else {
3566 			/* Stable 38400KHz frequency */
3567 			incperiod = INCPERIOD_38400KHZ;
3568 			incvalue = INCVALUE_38400KHZ;
3569 			shift = INCVALUE_SHIFT_38400KHZ;
3570 			adapter->cc.shift = shift;
3571 		}
3572 		break;
3573 	case e1000_82574:
3574 	case e1000_82583:
3575 		/* Stable 25MHz frequency */
3576 		incperiod = INCPERIOD_25MHZ;
3577 		incvalue = INCVALUE_25MHZ;
3578 		shift = INCVALUE_SHIFT_25MHZ;
3579 		adapter->cc.shift = shift;
3580 		break;
3581 	default:
3582 		return -EINVAL;
3583 	}
3584 
3585 	*timinca = ((incperiod << E1000_TIMINCA_INCPERIOD_SHIFT) |
3586 		    ((incvalue << shift) & E1000_TIMINCA_INCVALUE_MASK));
3587 
3588 	return 0;
3589 }
3590 
3591 /**
3592  * e1000e_config_hwtstamp - configure the hwtstamp registers and enable/disable
3593  * @adapter: board private structure
3594  *
3595  * Outgoing time stamping can be enabled and disabled. Play nice and
3596  * disable it when requested, although it shouldn't cause any overhead
3597  * when no packet needs it. At most one packet in the queue may be
3598  * marked for time stamping, otherwise it would be impossible to tell
3599  * for sure to which packet the hardware time stamp belongs.
3600  *
3601  * Incoming time stamping has to be configured via the hardware filters.
3602  * Not all combinations are supported, in particular event type has to be
3603  * specified. Matching the kind of event packet is not supported, with the
3604  * exception of "all V2 events regardless of level 2 or 4".
3605  **/
3606 static int e1000e_config_hwtstamp(struct e1000_adapter *adapter,
3607 				  struct hwtstamp_config *config)
3608 {
3609 	struct e1000_hw *hw = &adapter->hw;
3610 	u32 tsync_tx_ctl = E1000_TSYNCTXCTL_ENABLED;
3611 	u32 tsync_rx_ctl = E1000_TSYNCRXCTL_ENABLED;
3612 	u32 rxmtrl = 0;
3613 	u16 rxudp = 0;
3614 	bool is_l4 = false;
3615 	bool is_l2 = false;
3616 	u32 regval;
3617 
3618 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3619 		return -EINVAL;
3620 
3621 	/* flags reserved for future extensions - must be zero */
3622 	if (config->flags)
3623 		return -EINVAL;
3624 
3625 	switch (config->tx_type) {
3626 	case HWTSTAMP_TX_OFF:
3627 		tsync_tx_ctl = 0;
3628 		break;
3629 	case HWTSTAMP_TX_ON:
3630 		break;
3631 	default:
3632 		return -ERANGE;
3633 	}
3634 
3635 	switch (config->rx_filter) {
3636 	case HWTSTAMP_FILTER_NONE:
3637 		tsync_rx_ctl = 0;
3638 		break;
3639 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3640 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3641 		rxmtrl = E1000_RXMTRL_PTP_V1_SYNC_MESSAGE;
3642 		is_l4 = true;
3643 		break;
3644 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3645 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3646 		rxmtrl = E1000_RXMTRL_PTP_V1_DELAY_REQ_MESSAGE;
3647 		is_l4 = true;
3648 		break;
3649 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3650 		/* Also time stamps V2 L2 Path Delay Request/Response */
3651 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3652 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3653 		is_l2 = true;
3654 		break;
3655 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3656 		/* Also time stamps V2 L2 Path Delay Request/Response. */
3657 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3658 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3659 		is_l2 = true;
3660 		break;
3661 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3662 		/* Hardware cannot filter just V2 L4 Sync messages;
3663 		 * fall-through to V2 (both L2 and L4) Sync.
3664 		 */
3665 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
3666 		/* Also time stamps V2 Path Delay Request/Response. */
3667 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3668 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3669 		is_l2 = true;
3670 		is_l4 = true;
3671 		break;
3672 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3673 		/* Hardware cannot filter just V2 L4 Delay Request messages;
3674 		 * fall-through to V2 (both L2 and L4) Delay Request.
3675 		 */
3676 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
3677 		/* Also time stamps V2 Path Delay Request/Response. */
3678 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3679 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3680 		is_l2 = true;
3681 		is_l4 = true;
3682 		break;
3683 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
3684 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
3685 		/* Hardware cannot filter just V2 L4 or L2 Event messages;
3686 		 * fall-through to all V2 (both L2 and L4) Events.
3687 		 */
3688 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
3689 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_EVENT_V2;
3690 		config->rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
3691 		is_l2 = true;
3692 		is_l4 = true;
3693 		break;
3694 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3695 		/* For V1, the hardware can only filter Sync messages or
3696 		 * Delay Request messages but not both so fall-through to
3697 		 * time stamp all packets.
3698 		 */
3699 	case HWTSTAMP_FILTER_NTP_ALL:
3700 	case HWTSTAMP_FILTER_ALL:
3701 		is_l2 = true;
3702 		is_l4 = true;
3703 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_ALL;
3704 		config->rx_filter = HWTSTAMP_FILTER_ALL;
3705 		break;
3706 	default:
3707 		return -ERANGE;
3708 	}
3709 
3710 	adapter->hwtstamp_config = *config;
3711 
3712 	/* enable/disable Tx h/w time stamping */
3713 	regval = er32(TSYNCTXCTL);
3714 	regval &= ~E1000_TSYNCTXCTL_ENABLED;
3715 	regval |= tsync_tx_ctl;
3716 	ew32(TSYNCTXCTL, regval);
3717 	if ((er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) !=
3718 	    (regval & E1000_TSYNCTXCTL_ENABLED)) {
3719 		e_err("Timesync Tx Control register not set as expected\n");
3720 		return -EAGAIN;
3721 	}
3722 
3723 	/* enable/disable Rx h/w time stamping */
3724 	regval = er32(TSYNCRXCTL);
3725 	regval &= ~(E1000_TSYNCRXCTL_ENABLED | E1000_TSYNCRXCTL_TYPE_MASK);
3726 	regval |= tsync_rx_ctl;
3727 	ew32(TSYNCRXCTL, regval);
3728 	if ((er32(TSYNCRXCTL) & (E1000_TSYNCRXCTL_ENABLED |
3729 				 E1000_TSYNCRXCTL_TYPE_MASK)) !=
3730 	    (regval & (E1000_TSYNCRXCTL_ENABLED |
3731 		       E1000_TSYNCRXCTL_TYPE_MASK))) {
3732 		e_err("Timesync Rx Control register not set as expected\n");
3733 		return -EAGAIN;
3734 	}
3735 
3736 	/* L2: define ethertype filter for time stamped packets */
3737 	if (is_l2)
3738 		rxmtrl |= ETH_P_1588;
3739 
3740 	/* define which PTP packets get time stamped */
3741 	ew32(RXMTRL, rxmtrl);
3742 
3743 	/* Filter by destination port */
3744 	if (is_l4) {
3745 		rxudp = PTP_EV_PORT;
3746 		cpu_to_be16s(&rxudp);
3747 	}
3748 	ew32(RXUDP, rxudp);
3749 
3750 	e1e_flush();
3751 
3752 	/* Clear TSYNCRXCTL_VALID & TSYNCTXCTL_VALID bit */
3753 	er32(RXSTMPH);
3754 	er32(TXSTMPH);
3755 
3756 	return 0;
3757 }
3758 
3759 /**
3760  * e1000_configure - configure the hardware for Rx and Tx
3761  * @adapter: private board structure
3762  **/
3763 static void e1000_configure(struct e1000_adapter *adapter)
3764 {
3765 	struct e1000_ring *rx_ring = adapter->rx_ring;
3766 
3767 	e1000e_set_rx_mode(adapter->netdev);
3768 
3769 	e1000_restore_vlan(adapter);
3770 	e1000_init_manageability_pt(adapter);
3771 
3772 	e1000_configure_tx(adapter);
3773 
3774 	if (adapter->netdev->features & NETIF_F_RXHASH)
3775 		e1000e_setup_rss_hash(adapter);
3776 	e1000_setup_rctl(adapter);
3777 	e1000_configure_rx(adapter);
3778 	adapter->alloc_rx_buf(rx_ring, e1000_desc_unused(rx_ring), GFP_KERNEL);
3779 }
3780 
3781 /**
3782  * e1000e_power_up_phy - restore link in case the phy was powered down
3783  * @adapter: address of board private structure
3784  *
3785  * The phy may be powered down to save power and turn off link when the
3786  * driver is unloaded and wake on lan is not enabled (among others)
3787  * *** this routine MUST be followed by a call to e1000e_reset ***
3788  **/
3789 void e1000e_power_up_phy(struct e1000_adapter *adapter)
3790 {
3791 	if (adapter->hw.phy.ops.power_up)
3792 		adapter->hw.phy.ops.power_up(&adapter->hw);
3793 
3794 	adapter->hw.mac.ops.setup_link(&adapter->hw);
3795 }
3796 
3797 /**
3798  * e1000_power_down_phy - Power down the PHY
3799  *
3800  * Power down the PHY so no link is implied when interface is down.
3801  * The PHY cannot be powered down if management or WoL is active.
3802  */
3803 static void e1000_power_down_phy(struct e1000_adapter *adapter)
3804 {
3805 	if (adapter->hw.phy.ops.power_down)
3806 		adapter->hw.phy.ops.power_down(&adapter->hw);
3807 }
3808 
3809 /**
3810  * e1000_flush_tx_ring - remove all descriptors from the tx_ring
3811  *
3812  * We want to clear all pending descriptors from the TX ring.
3813  * zeroing happens when the HW reads the regs. We  assign the ring itself as
3814  * the data of the next descriptor. We don't care about the data we are about
3815  * to reset the HW.
3816  */
3817 static void e1000_flush_tx_ring(struct e1000_adapter *adapter)
3818 {
3819 	struct e1000_hw *hw = &adapter->hw;
3820 	struct e1000_ring *tx_ring = adapter->tx_ring;
3821 	struct e1000_tx_desc *tx_desc = NULL;
3822 	u32 tdt, tctl, txd_lower = E1000_TXD_CMD_IFCS;
3823 	u16 size = 512;
3824 
3825 	tctl = er32(TCTL);
3826 	ew32(TCTL, tctl | E1000_TCTL_EN);
3827 	tdt = er32(TDT(0));
3828 	BUG_ON(tdt != tx_ring->next_to_use);
3829 	tx_desc =  E1000_TX_DESC(*tx_ring, tx_ring->next_to_use);
3830 	tx_desc->buffer_addr = tx_ring->dma;
3831 
3832 	tx_desc->lower.data = cpu_to_le32(txd_lower | size);
3833 	tx_desc->upper.data = 0;
3834 	/* flush descriptors to memory before notifying the HW */
3835 	wmb();
3836 	tx_ring->next_to_use++;
3837 	if (tx_ring->next_to_use == tx_ring->count)
3838 		tx_ring->next_to_use = 0;
3839 	ew32(TDT(0), tx_ring->next_to_use);
3840 	mmiowb();
3841 	usleep_range(200, 250);
3842 }
3843 
3844 /**
3845  * e1000_flush_rx_ring - remove all descriptors from the rx_ring
3846  *
3847  * Mark all descriptors in the RX ring as consumed and disable the rx ring
3848  */
3849 static void e1000_flush_rx_ring(struct e1000_adapter *adapter)
3850 {
3851 	u32 rctl, rxdctl;
3852 	struct e1000_hw *hw = &adapter->hw;
3853 
3854 	rctl = er32(RCTL);
3855 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3856 	e1e_flush();
3857 	usleep_range(100, 150);
3858 
3859 	rxdctl = er32(RXDCTL(0));
3860 	/* zero the lower 14 bits (prefetch and host thresholds) */
3861 	rxdctl &= 0xffffc000;
3862 
3863 	/* update thresholds: prefetch threshold to 31, host threshold to 1
3864 	 * and make sure the granularity is "descriptors" and not "cache lines"
3865 	 */
3866 	rxdctl |= (0x1F | BIT(8) | E1000_RXDCTL_THRESH_UNIT_DESC);
3867 
3868 	ew32(RXDCTL(0), rxdctl);
3869 	/* momentarily enable the RX ring for the changes to take effect */
3870 	ew32(RCTL, rctl | E1000_RCTL_EN);
3871 	e1e_flush();
3872 	usleep_range(100, 150);
3873 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3874 }
3875 
3876 /**
3877  * e1000_flush_desc_rings - remove all descriptors from the descriptor rings
3878  *
3879  * In i219, the descriptor rings must be emptied before resetting the HW
3880  * or before changing the device state to D3 during runtime (runtime PM).
3881  *
3882  * Failure to do this will cause the HW to enter a unit hang state which can
3883  * only be released by PCI reset on the device
3884  *
3885  */
3886 
3887 static void e1000_flush_desc_rings(struct e1000_adapter *adapter)
3888 {
3889 	u16 hang_state;
3890 	u32 fext_nvm11, tdlen;
3891 	struct e1000_hw *hw = &adapter->hw;
3892 
3893 	/* First, disable MULR fix in FEXTNVM11 */
3894 	fext_nvm11 = er32(FEXTNVM11);
3895 	fext_nvm11 |= E1000_FEXTNVM11_DISABLE_MULR_FIX;
3896 	ew32(FEXTNVM11, fext_nvm11);
3897 	/* do nothing if we're not in faulty state, or if the queue is empty */
3898 	tdlen = er32(TDLEN(0));
3899 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3900 			     &hang_state);
3901 	if (!(hang_state & FLUSH_DESC_REQUIRED) || !tdlen)
3902 		return;
3903 	e1000_flush_tx_ring(adapter);
3904 	/* recheck, maybe the fault is caused by the rx ring */
3905 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3906 			     &hang_state);
3907 	if (hang_state & FLUSH_DESC_REQUIRED)
3908 		e1000_flush_rx_ring(adapter);
3909 }
3910 
3911 /**
3912  * e1000e_systim_reset - reset the timesync registers after a hardware reset
3913  * @adapter: board private structure
3914  *
3915  * When the MAC is reset, all hardware bits for timesync will be reset to the
3916  * default values. This function will restore the settings last in place.
3917  * Since the clock SYSTIME registers are reset, we will simply restore the
3918  * cyclecounter to the kernel real clock time.
3919  **/
3920 static void e1000e_systim_reset(struct e1000_adapter *adapter)
3921 {
3922 	struct ptp_clock_info *info = &adapter->ptp_clock_info;
3923 	struct e1000_hw *hw = &adapter->hw;
3924 	unsigned long flags;
3925 	u32 timinca;
3926 	s32 ret_val;
3927 
3928 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3929 		return;
3930 
3931 	if (info->adjfreq) {
3932 		/* restore the previous ptp frequency delta */
3933 		ret_val = info->adjfreq(info, adapter->ptp_delta);
3934 	} else {
3935 		/* set the default base frequency if no adjustment possible */
3936 		ret_val = e1000e_get_base_timinca(adapter, &timinca);
3937 		if (!ret_val)
3938 			ew32(TIMINCA, timinca);
3939 	}
3940 
3941 	if (ret_val) {
3942 		dev_warn(&adapter->pdev->dev,
3943 			 "Failed to restore TIMINCA clock rate delta: %d\n",
3944 			 ret_val);
3945 		return;
3946 	}
3947 
3948 	/* reset the systim ns time counter */
3949 	spin_lock_irqsave(&adapter->systim_lock, flags);
3950 	timecounter_init(&adapter->tc, &adapter->cc,
3951 			 ktime_to_ns(ktime_get_real()));
3952 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
3953 
3954 	/* restore the previous hwtstamp configuration settings */
3955 	e1000e_config_hwtstamp(adapter, &adapter->hwtstamp_config);
3956 }
3957 
3958 /**
3959  * e1000e_reset - bring the hardware into a known good state
3960  *
3961  * This function boots the hardware and enables some settings that
3962  * require a configuration cycle of the hardware - those cannot be
3963  * set/changed during runtime. After reset the device needs to be
3964  * properly configured for Rx, Tx etc.
3965  */
3966 void e1000e_reset(struct e1000_adapter *adapter)
3967 {
3968 	struct e1000_mac_info *mac = &adapter->hw.mac;
3969 	struct e1000_fc_info *fc = &adapter->hw.fc;
3970 	struct e1000_hw *hw = &adapter->hw;
3971 	u32 tx_space, min_tx_space, min_rx_space;
3972 	u32 pba = adapter->pba;
3973 	u16 hwm;
3974 
3975 	/* reset Packet Buffer Allocation to default */
3976 	ew32(PBA, pba);
3977 
3978 	if (adapter->max_frame_size > (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN)) {
3979 		/* To maintain wire speed transmits, the Tx FIFO should be
3980 		 * large enough to accommodate two full transmit packets,
3981 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
3982 		 * the Rx FIFO should be large enough to accommodate at least
3983 		 * one full receive packet and is similarly rounded up and
3984 		 * expressed in KB.
3985 		 */
3986 		pba = er32(PBA);
3987 		/* upper 16 bits has Tx packet buffer allocation size in KB */
3988 		tx_space = pba >> 16;
3989 		/* lower 16 bits has Rx packet buffer allocation size in KB */
3990 		pba &= 0xffff;
3991 		/* the Tx fifo also stores 16 bytes of information about the Tx
3992 		 * but don't include ethernet FCS because hardware appends it
3993 		 */
3994 		min_tx_space = (adapter->max_frame_size +
3995 				sizeof(struct e1000_tx_desc) - ETH_FCS_LEN) * 2;
3996 		min_tx_space = ALIGN(min_tx_space, 1024);
3997 		min_tx_space >>= 10;
3998 		/* software strips receive CRC, so leave room for it */
3999 		min_rx_space = adapter->max_frame_size;
4000 		min_rx_space = ALIGN(min_rx_space, 1024);
4001 		min_rx_space >>= 10;
4002 
4003 		/* If current Tx allocation is less than the min Tx FIFO size,
4004 		 * and the min Tx FIFO size is less than the current Rx FIFO
4005 		 * allocation, take space away from current Rx allocation
4006 		 */
4007 		if ((tx_space < min_tx_space) &&
4008 		    ((min_tx_space - tx_space) < pba)) {
4009 			pba -= min_tx_space - tx_space;
4010 
4011 			/* if short on Rx space, Rx wins and must trump Tx
4012 			 * adjustment
4013 			 */
4014 			if (pba < min_rx_space)
4015 				pba = min_rx_space;
4016 		}
4017 
4018 		ew32(PBA, pba);
4019 	}
4020 
4021 	/* flow control settings
4022 	 *
4023 	 * The high water mark must be low enough to fit one full frame
4024 	 * (or the size used for early receive) above it in the Rx FIFO.
4025 	 * Set it to the lower of:
4026 	 * - 90% of the Rx FIFO size, and
4027 	 * - the full Rx FIFO size minus one full frame
4028 	 */
4029 	if (adapter->flags & FLAG_DISABLE_FC_PAUSE_TIME)
4030 		fc->pause_time = 0xFFFF;
4031 	else
4032 		fc->pause_time = E1000_FC_PAUSE_TIME;
4033 	fc->send_xon = true;
4034 	fc->current_mode = fc->requested_mode;
4035 
4036 	switch (hw->mac.type) {
4037 	case e1000_ich9lan:
4038 	case e1000_ich10lan:
4039 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4040 			pba = 14;
4041 			ew32(PBA, pba);
4042 			fc->high_water = 0x2800;
4043 			fc->low_water = fc->high_water - 8;
4044 			break;
4045 		}
4046 		/* fall-through */
4047 	default:
4048 		hwm = min(((pba << 10) * 9 / 10),
4049 			  ((pba << 10) - adapter->max_frame_size));
4050 
4051 		fc->high_water = hwm & E1000_FCRTH_RTH;	/* 8-byte granularity */
4052 		fc->low_water = fc->high_water - 8;
4053 		break;
4054 	case e1000_pchlan:
4055 		/* Workaround PCH LOM adapter hangs with certain network
4056 		 * loads.  If hangs persist, try disabling Tx flow control.
4057 		 */
4058 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4059 			fc->high_water = 0x3500;
4060 			fc->low_water = 0x1500;
4061 		} else {
4062 			fc->high_water = 0x5000;
4063 			fc->low_water = 0x3000;
4064 		}
4065 		fc->refresh_time = 0x1000;
4066 		break;
4067 	case e1000_pch2lan:
4068 	case e1000_pch_lpt:
4069 	case e1000_pch_spt:
4070 	case e1000_pch_cnp:
4071 		fc->refresh_time = 0x0400;
4072 
4073 		if (adapter->netdev->mtu <= ETH_DATA_LEN) {
4074 			fc->high_water = 0x05C20;
4075 			fc->low_water = 0x05048;
4076 			fc->pause_time = 0x0650;
4077 			break;
4078 		}
4079 
4080 		pba = 14;
4081 		ew32(PBA, pba);
4082 		fc->high_water = ((pba << 10) * 9 / 10) & E1000_FCRTH_RTH;
4083 		fc->low_water = ((pba << 10) * 8 / 10) & E1000_FCRTL_RTL;
4084 		break;
4085 	}
4086 
4087 	/* Alignment of Tx data is on an arbitrary byte boundary with the
4088 	 * maximum size per Tx descriptor limited only to the transmit
4089 	 * allocation of the packet buffer minus 96 bytes with an upper
4090 	 * limit of 24KB due to receive synchronization limitations.
4091 	 */
4092 	adapter->tx_fifo_limit = min_t(u32, ((er32(PBA) >> 16) << 10) - 96,
4093 				       24 << 10);
4094 
4095 	/* Disable Adaptive Interrupt Moderation if 2 full packets cannot
4096 	 * fit in receive buffer.
4097 	 */
4098 	if (adapter->itr_setting & 0x3) {
4099 		if ((adapter->max_frame_size * 2) > (pba << 10)) {
4100 			if (!(adapter->flags2 & FLAG2_DISABLE_AIM)) {
4101 				dev_info(&adapter->pdev->dev,
4102 					 "Interrupt Throttle Rate off\n");
4103 				adapter->flags2 |= FLAG2_DISABLE_AIM;
4104 				e1000e_write_itr(adapter, 0);
4105 			}
4106 		} else if (adapter->flags2 & FLAG2_DISABLE_AIM) {
4107 			dev_info(&adapter->pdev->dev,
4108 				 "Interrupt Throttle Rate on\n");
4109 			adapter->flags2 &= ~FLAG2_DISABLE_AIM;
4110 			adapter->itr = 20000;
4111 			e1000e_write_itr(adapter, adapter->itr);
4112 		}
4113 	}
4114 
4115 	if (hw->mac.type >= e1000_pch_spt)
4116 		e1000_flush_desc_rings(adapter);
4117 	/* Allow time for pending master requests to run */
4118 	mac->ops.reset_hw(hw);
4119 
4120 	/* For parts with AMT enabled, let the firmware know
4121 	 * that the network interface is in control
4122 	 */
4123 	if (adapter->flags & FLAG_HAS_AMT)
4124 		e1000e_get_hw_control(adapter);
4125 
4126 	ew32(WUC, 0);
4127 
4128 	if (mac->ops.init_hw(hw))
4129 		e_err("Hardware Error\n");
4130 
4131 	e1000_update_mng_vlan(adapter);
4132 
4133 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
4134 	ew32(VET, ETH_P_8021Q);
4135 
4136 	e1000e_reset_adaptive(hw);
4137 
4138 	/* restore systim and hwtstamp settings */
4139 	e1000e_systim_reset(adapter);
4140 
4141 	/* Set EEE advertisement as appropriate */
4142 	if (adapter->flags2 & FLAG2_HAS_EEE) {
4143 		s32 ret_val;
4144 		u16 adv_addr;
4145 
4146 		switch (hw->phy.type) {
4147 		case e1000_phy_82579:
4148 			adv_addr = I82579_EEE_ADVERTISEMENT;
4149 			break;
4150 		case e1000_phy_i217:
4151 			adv_addr = I217_EEE_ADVERTISEMENT;
4152 			break;
4153 		default:
4154 			dev_err(&adapter->pdev->dev,
4155 				"Invalid PHY type setting EEE advertisement\n");
4156 			return;
4157 		}
4158 
4159 		ret_val = hw->phy.ops.acquire(hw);
4160 		if (ret_val) {
4161 			dev_err(&adapter->pdev->dev,
4162 				"EEE advertisement - unable to acquire PHY\n");
4163 			return;
4164 		}
4165 
4166 		e1000_write_emi_reg_locked(hw, adv_addr,
4167 					   hw->dev_spec.ich8lan.eee_disable ?
4168 					   0 : adapter->eee_advert);
4169 
4170 		hw->phy.ops.release(hw);
4171 	}
4172 
4173 	if (!netif_running(adapter->netdev) &&
4174 	    !test_bit(__E1000_TESTING, &adapter->state))
4175 		e1000_power_down_phy(adapter);
4176 
4177 	e1000_get_phy_info(hw);
4178 
4179 	if ((adapter->flags & FLAG_HAS_SMART_POWER_DOWN) &&
4180 	    !(adapter->flags & FLAG_SMART_POWER_DOWN)) {
4181 		u16 phy_data = 0;
4182 		/* speed up time to link by disabling smart power down, ignore
4183 		 * the return value of this function because there is nothing
4184 		 * different we would do if it failed
4185 		 */
4186 		e1e_rphy(hw, IGP02E1000_PHY_POWER_MGMT, &phy_data);
4187 		phy_data &= ~IGP02E1000_PM_SPD;
4188 		e1e_wphy(hw, IGP02E1000_PHY_POWER_MGMT, phy_data);
4189 	}
4190 	if (hw->mac.type >= e1000_pch_spt && adapter->int_mode == 0) {
4191 		u32 reg;
4192 
4193 		/* Fextnvm7 @ 0xe4[2] = 1 */
4194 		reg = er32(FEXTNVM7);
4195 		reg |= E1000_FEXTNVM7_SIDE_CLK_UNGATE;
4196 		ew32(FEXTNVM7, reg);
4197 		/* Fextnvm9 @ 0x5bb4[13:12] = 11 */
4198 		reg = er32(FEXTNVM9);
4199 		reg |= E1000_FEXTNVM9_IOSFSB_CLKGATE_DIS |
4200 		       E1000_FEXTNVM9_IOSFSB_CLKREQ_DIS;
4201 		ew32(FEXTNVM9, reg);
4202 	}
4203 
4204 }
4205 
4206 /**
4207  * e1000e_trigger_lsc - trigger an LSC interrupt
4208  * @adapter:
4209  *
4210  * Fire a link status change interrupt to start the watchdog.
4211  **/
4212 static void e1000e_trigger_lsc(struct e1000_adapter *adapter)
4213 {
4214 	struct e1000_hw *hw = &adapter->hw;
4215 
4216 	if (adapter->msix_entries)
4217 		ew32(ICS, E1000_ICS_LSC | E1000_ICS_OTHER);
4218 	else
4219 		ew32(ICS, E1000_ICS_LSC);
4220 }
4221 
4222 void e1000e_up(struct e1000_adapter *adapter)
4223 {
4224 	/* hardware has been reset, we need to reload some things */
4225 	e1000_configure(adapter);
4226 
4227 	clear_bit(__E1000_DOWN, &adapter->state);
4228 
4229 	if (adapter->msix_entries)
4230 		e1000_configure_msix(adapter);
4231 	e1000_irq_enable(adapter);
4232 
4233 	netif_start_queue(adapter->netdev);
4234 
4235 	e1000e_trigger_lsc(adapter);
4236 }
4237 
4238 static void e1000e_flush_descriptors(struct e1000_adapter *adapter)
4239 {
4240 	struct e1000_hw *hw = &adapter->hw;
4241 
4242 	if (!(adapter->flags2 & FLAG2_DMA_BURST))
4243 		return;
4244 
4245 	/* flush pending descriptor writebacks to memory */
4246 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4247 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4248 
4249 	/* execute the writes immediately */
4250 	e1e_flush();
4251 
4252 	/* due to rare timing issues, write to TIDV/RDTR again to ensure the
4253 	 * write is successful
4254 	 */
4255 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4256 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4257 
4258 	/* execute the writes immediately */
4259 	e1e_flush();
4260 }
4261 
4262 static void e1000e_update_stats(struct e1000_adapter *adapter);
4263 
4264 /**
4265  * e1000e_down - quiesce the device and optionally reset the hardware
4266  * @adapter: board private structure
4267  * @reset: boolean flag to reset the hardware or not
4268  */
4269 void e1000e_down(struct e1000_adapter *adapter, bool reset)
4270 {
4271 	struct net_device *netdev = adapter->netdev;
4272 	struct e1000_hw *hw = &adapter->hw;
4273 	u32 tctl, rctl;
4274 
4275 	/* signal that we're down so the interrupt handler does not
4276 	 * reschedule our watchdog timer
4277 	 */
4278 	set_bit(__E1000_DOWN, &adapter->state);
4279 
4280 	netif_carrier_off(netdev);
4281 
4282 	/* disable receives in the hardware */
4283 	rctl = er32(RCTL);
4284 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
4285 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
4286 	/* flush and sleep below */
4287 
4288 	netif_stop_queue(netdev);
4289 
4290 	/* disable transmits in the hardware */
4291 	tctl = er32(TCTL);
4292 	tctl &= ~E1000_TCTL_EN;
4293 	ew32(TCTL, tctl);
4294 
4295 	/* flush both disables and wait for them to finish */
4296 	e1e_flush();
4297 	usleep_range(10000, 20000);
4298 
4299 	e1000_irq_disable(adapter);
4300 
4301 	napi_synchronize(&adapter->napi);
4302 
4303 	del_timer_sync(&adapter->watchdog_timer);
4304 	del_timer_sync(&adapter->phy_info_timer);
4305 
4306 	spin_lock(&adapter->stats64_lock);
4307 	e1000e_update_stats(adapter);
4308 	spin_unlock(&adapter->stats64_lock);
4309 
4310 	e1000e_flush_descriptors(adapter);
4311 
4312 	adapter->link_speed = 0;
4313 	adapter->link_duplex = 0;
4314 
4315 	/* Disable Si errata workaround on PCHx for jumbo frame flow */
4316 	if ((hw->mac.type >= e1000_pch2lan) &&
4317 	    (adapter->netdev->mtu > ETH_DATA_LEN) &&
4318 	    e1000_lv_jumbo_workaround_ich8lan(hw, false))
4319 		e_dbg("failed to disable jumbo frame workaround mode\n");
4320 
4321 	if (!pci_channel_offline(adapter->pdev)) {
4322 		if (reset)
4323 			e1000e_reset(adapter);
4324 		else if (hw->mac.type >= e1000_pch_spt)
4325 			e1000_flush_desc_rings(adapter);
4326 	}
4327 	e1000_clean_tx_ring(adapter->tx_ring);
4328 	e1000_clean_rx_ring(adapter->rx_ring);
4329 }
4330 
4331 void e1000e_reinit_locked(struct e1000_adapter *adapter)
4332 {
4333 	might_sleep();
4334 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
4335 		usleep_range(1000, 2000);
4336 	e1000e_down(adapter, true);
4337 	e1000e_up(adapter);
4338 	clear_bit(__E1000_RESETTING, &adapter->state);
4339 }
4340 
4341 /**
4342  * e1000e_sanitize_systim - sanitize raw cycle counter reads
4343  * @hw: pointer to the HW structure
4344  * @systim: time value read, sanitized and returned
4345  *
4346  * Errata for 82574/82583 possible bad bits read from SYSTIMH/L:
4347  * check to see that the time is incrementing at a reasonable
4348  * rate and is a multiple of incvalue.
4349  **/
4350 static u64 e1000e_sanitize_systim(struct e1000_hw *hw, u64 systim)
4351 {
4352 	u64 time_delta, rem, temp;
4353 	u64 systim_next;
4354 	u32 incvalue;
4355 	int i;
4356 
4357 	incvalue = er32(TIMINCA) & E1000_TIMINCA_INCVALUE_MASK;
4358 	for (i = 0; i < E1000_MAX_82574_SYSTIM_REREADS; i++) {
4359 		/* latch SYSTIMH on read of SYSTIML */
4360 		systim_next = (u64)er32(SYSTIML);
4361 		systim_next |= (u64)er32(SYSTIMH) << 32;
4362 
4363 		time_delta = systim_next - systim;
4364 		temp = time_delta;
4365 		/* VMWare users have seen incvalue of zero, don't div / 0 */
4366 		rem = incvalue ? do_div(temp, incvalue) : (time_delta != 0);
4367 
4368 		systim = systim_next;
4369 
4370 		if ((time_delta < E1000_82574_SYSTIM_EPSILON) && (rem == 0))
4371 			break;
4372 	}
4373 
4374 	return systim;
4375 }
4376 
4377 /**
4378  * e1000e_cyclecounter_read - read raw cycle counter (used by time counter)
4379  * @cc: cyclecounter structure
4380  **/
4381 static u64 e1000e_cyclecounter_read(const struct cyclecounter *cc)
4382 {
4383 	struct e1000_adapter *adapter = container_of(cc, struct e1000_adapter,
4384 						     cc);
4385 	struct e1000_hw *hw = &adapter->hw;
4386 	u32 systimel, systimeh;
4387 	u64 systim;
4388 	/* SYSTIMH latching upon SYSTIML read does not work well.
4389 	 * This means that if SYSTIML overflows after we read it but before
4390 	 * we read SYSTIMH, the value of SYSTIMH has been incremented and we
4391 	 * will experience a huge non linear increment in the systime value
4392 	 * to fix that we test for overflow and if true, we re-read systime.
4393 	 */
4394 	systimel = er32(SYSTIML);
4395 	systimeh = er32(SYSTIMH);
4396 	/* Is systimel is so large that overflow is possible? */
4397 	if (systimel >= (u32)0xffffffff - E1000_TIMINCA_INCVALUE_MASK) {
4398 		u32 systimel_2 = er32(SYSTIML);
4399 		if (systimel > systimel_2) {
4400 			/* There was an overflow, read again SYSTIMH, and use
4401 			 * systimel_2
4402 			 */
4403 			systimeh = er32(SYSTIMH);
4404 			systimel = systimel_2;
4405 		}
4406 	}
4407 	systim = (u64)systimel;
4408 	systim |= (u64)systimeh << 32;
4409 
4410 	if (adapter->flags2 & FLAG2_CHECK_SYSTIM_OVERFLOW)
4411 		systim = e1000e_sanitize_systim(hw, systim);
4412 
4413 	return systim;
4414 }
4415 
4416 /**
4417  * e1000_sw_init - Initialize general software structures (struct e1000_adapter)
4418  * @adapter: board private structure to initialize
4419  *
4420  * e1000_sw_init initializes the Adapter private data structure.
4421  * Fields are initialized based on PCI device information and
4422  * OS network device settings (MTU size).
4423  **/
4424 static int e1000_sw_init(struct e1000_adapter *adapter)
4425 {
4426 	struct net_device *netdev = adapter->netdev;
4427 
4428 	adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
4429 	adapter->rx_ps_bsize0 = 128;
4430 	adapter->max_frame_size = netdev->mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
4431 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4432 	adapter->tx_ring_count = E1000_DEFAULT_TXD;
4433 	adapter->rx_ring_count = E1000_DEFAULT_RXD;
4434 
4435 	spin_lock_init(&adapter->stats64_lock);
4436 
4437 	e1000e_set_interrupt_capability(adapter);
4438 
4439 	if (e1000_alloc_queues(adapter))
4440 		return -ENOMEM;
4441 
4442 	/* Setup hardware time stamping cyclecounter */
4443 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
4444 		adapter->cc.read = e1000e_cyclecounter_read;
4445 		adapter->cc.mask = CYCLECOUNTER_MASK(64);
4446 		adapter->cc.mult = 1;
4447 		/* cc.shift set in e1000e_get_base_tininca() */
4448 
4449 		spin_lock_init(&adapter->systim_lock);
4450 		INIT_WORK(&adapter->tx_hwtstamp_work, e1000e_tx_hwtstamp_work);
4451 	}
4452 
4453 	/* Explicitly disable IRQ since the NIC can be in any state. */
4454 	e1000_irq_disable(adapter);
4455 
4456 	set_bit(__E1000_DOWN, &adapter->state);
4457 	return 0;
4458 }
4459 
4460 /**
4461  * e1000_intr_msi_test - Interrupt Handler
4462  * @irq: interrupt number
4463  * @data: pointer to a network interface device structure
4464  **/
4465 static irqreturn_t e1000_intr_msi_test(int __always_unused irq, void *data)
4466 {
4467 	struct net_device *netdev = data;
4468 	struct e1000_adapter *adapter = netdev_priv(netdev);
4469 	struct e1000_hw *hw = &adapter->hw;
4470 	u32 icr = er32(ICR);
4471 
4472 	e_dbg("icr is %08X\n", icr);
4473 	if (icr & E1000_ICR_RXSEQ) {
4474 		adapter->flags &= ~FLAG_MSI_TEST_FAILED;
4475 		/* Force memory writes to complete before acknowledging the
4476 		 * interrupt is handled.
4477 		 */
4478 		wmb();
4479 	}
4480 
4481 	return IRQ_HANDLED;
4482 }
4483 
4484 /**
4485  * e1000_test_msi_interrupt - Returns 0 for successful test
4486  * @adapter: board private struct
4487  *
4488  * code flow taken from tg3.c
4489  **/
4490 static int e1000_test_msi_interrupt(struct e1000_adapter *adapter)
4491 {
4492 	struct net_device *netdev = adapter->netdev;
4493 	struct e1000_hw *hw = &adapter->hw;
4494 	int err;
4495 
4496 	/* poll_enable hasn't been called yet, so don't need disable */
4497 	/* clear any pending events */
4498 	er32(ICR);
4499 
4500 	/* free the real vector and request a test handler */
4501 	e1000_free_irq(adapter);
4502 	e1000e_reset_interrupt_capability(adapter);
4503 
4504 	/* Assume that the test fails, if it succeeds then the test
4505 	 * MSI irq handler will unset this flag
4506 	 */
4507 	adapter->flags |= FLAG_MSI_TEST_FAILED;
4508 
4509 	err = pci_enable_msi(adapter->pdev);
4510 	if (err)
4511 		goto msi_test_failed;
4512 
4513 	err = request_irq(adapter->pdev->irq, e1000_intr_msi_test, 0,
4514 			  netdev->name, netdev);
4515 	if (err) {
4516 		pci_disable_msi(adapter->pdev);
4517 		goto msi_test_failed;
4518 	}
4519 
4520 	/* Force memory writes to complete before enabling and firing an
4521 	 * interrupt.
4522 	 */
4523 	wmb();
4524 
4525 	e1000_irq_enable(adapter);
4526 
4527 	/* fire an unusual interrupt on the test handler */
4528 	ew32(ICS, E1000_ICS_RXSEQ);
4529 	e1e_flush();
4530 	msleep(100);
4531 
4532 	e1000_irq_disable(adapter);
4533 
4534 	rmb();			/* read flags after interrupt has been fired */
4535 
4536 	if (adapter->flags & FLAG_MSI_TEST_FAILED) {
4537 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
4538 		e_info("MSI interrupt test failed, using legacy interrupt.\n");
4539 	} else {
4540 		e_dbg("MSI interrupt test succeeded!\n");
4541 	}
4542 
4543 	free_irq(adapter->pdev->irq, netdev);
4544 	pci_disable_msi(adapter->pdev);
4545 
4546 msi_test_failed:
4547 	e1000e_set_interrupt_capability(adapter);
4548 	return e1000_request_irq(adapter);
4549 }
4550 
4551 /**
4552  * e1000_test_msi - Returns 0 if MSI test succeeds or INTx mode is restored
4553  * @adapter: board private struct
4554  *
4555  * code flow taken from tg3.c, called with e1000 interrupts disabled.
4556  **/
4557 static int e1000_test_msi(struct e1000_adapter *adapter)
4558 {
4559 	int err;
4560 	u16 pci_cmd;
4561 
4562 	if (!(adapter->flags & FLAG_MSI_ENABLED))
4563 		return 0;
4564 
4565 	/* disable SERR in case the MSI write causes a master abort */
4566 	pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4567 	if (pci_cmd & PCI_COMMAND_SERR)
4568 		pci_write_config_word(adapter->pdev, PCI_COMMAND,
4569 				      pci_cmd & ~PCI_COMMAND_SERR);
4570 
4571 	err = e1000_test_msi_interrupt(adapter);
4572 
4573 	/* re-enable SERR */
4574 	if (pci_cmd & PCI_COMMAND_SERR) {
4575 		pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4576 		pci_cmd |= PCI_COMMAND_SERR;
4577 		pci_write_config_word(adapter->pdev, PCI_COMMAND, pci_cmd);
4578 	}
4579 
4580 	return err;
4581 }
4582 
4583 /**
4584  * e1000e_open - Called when a network interface is made active
4585  * @netdev: network interface device structure
4586  *
4587  * Returns 0 on success, negative value on failure
4588  *
4589  * The open entry point is called when a network interface is made
4590  * active by the system (IFF_UP).  At this point all resources needed
4591  * for transmit and receive operations are allocated, the interrupt
4592  * handler is registered with the OS, the watchdog timer is started,
4593  * and the stack is notified that the interface is ready.
4594  **/
4595 int e1000e_open(struct net_device *netdev)
4596 {
4597 	struct e1000_adapter *adapter = netdev_priv(netdev);
4598 	struct e1000_hw *hw = &adapter->hw;
4599 	struct pci_dev *pdev = adapter->pdev;
4600 	int err;
4601 
4602 	/* disallow open during test */
4603 	if (test_bit(__E1000_TESTING, &adapter->state))
4604 		return -EBUSY;
4605 
4606 	pm_runtime_get_sync(&pdev->dev);
4607 
4608 	netif_carrier_off(netdev);
4609 
4610 	/* allocate transmit descriptors */
4611 	err = e1000e_setup_tx_resources(adapter->tx_ring);
4612 	if (err)
4613 		goto err_setup_tx;
4614 
4615 	/* allocate receive descriptors */
4616 	err = e1000e_setup_rx_resources(adapter->rx_ring);
4617 	if (err)
4618 		goto err_setup_rx;
4619 
4620 	/* If AMT is enabled, let the firmware know that the network
4621 	 * interface is now open and reset the part to a known state.
4622 	 */
4623 	if (adapter->flags & FLAG_HAS_AMT) {
4624 		e1000e_get_hw_control(adapter);
4625 		e1000e_reset(adapter);
4626 	}
4627 
4628 	e1000e_power_up_phy(adapter);
4629 
4630 	adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
4631 	if ((adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
4632 		e1000_update_mng_vlan(adapter);
4633 
4634 	/* DMA latency requirement to workaround jumbo issue */
4635 	pm_qos_add_request(&adapter->pm_qos_req, PM_QOS_CPU_DMA_LATENCY,
4636 			   PM_QOS_DEFAULT_VALUE);
4637 
4638 	/* before we allocate an interrupt, we must be ready to handle it.
4639 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4640 	 * as soon as we call pci_request_irq, so we have to setup our
4641 	 * clean_rx handler before we do so.
4642 	 */
4643 	e1000_configure(adapter);
4644 
4645 	err = e1000_request_irq(adapter);
4646 	if (err)
4647 		goto err_req_irq;
4648 
4649 	/* Work around PCIe errata with MSI interrupts causing some chipsets to
4650 	 * ignore e1000e MSI messages, which means we need to test our MSI
4651 	 * interrupt now
4652 	 */
4653 	if (adapter->int_mode != E1000E_INT_MODE_LEGACY) {
4654 		err = e1000_test_msi(adapter);
4655 		if (err) {
4656 			e_err("Interrupt allocation failed\n");
4657 			goto err_req_irq;
4658 		}
4659 	}
4660 
4661 	/* From here on the code is the same as e1000e_up() */
4662 	clear_bit(__E1000_DOWN, &adapter->state);
4663 
4664 	napi_enable(&adapter->napi);
4665 
4666 	e1000_irq_enable(adapter);
4667 
4668 	adapter->tx_hang_recheck = false;
4669 	netif_start_queue(netdev);
4670 
4671 	hw->mac.get_link_status = true;
4672 	pm_runtime_put(&pdev->dev);
4673 
4674 	e1000e_trigger_lsc(adapter);
4675 
4676 	return 0;
4677 
4678 err_req_irq:
4679 	pm_qos_remove_request(&adapter->pm_qos_req);
4680 	e1000e_release_hw_control(adapter);
4681 	e1000_power_down_phy(adapter);
4682 	e1000e_free_rx_resources(adapter->rx_ring);
4683 err_setup_rx:
4684 	e1000e_free_tx_resources(adapter->tx_ring);
4685 err_setup_tx:
4686 	e1000e_reset(adapter);
4687 	pm_runtime_put_sync(&pdev->dev);
4688 
4689 	return err;
4690 }
4691 
4692 /**
4693  * e1000e_close - Disables a network interface
4694  * @netdev: network interface device structure
4695  *
4696  * Returns 0, this is not allowed to fail
4697  *
4698  * The close entry point is called when an interface is de-activated
4699  * by the OS.  The hardware is still under the drivers control, but
4700  * needs to be disabled.  A global MAC reset is issued to stop the
4701  * hardware, and all transmit and receive resources are freed.
4702  **/
4703 int e1000e_close(struct net_device *netdev)
4704 {
4705 	struct e1000_adapter *adapter = netdev_priv(netdev);
4706 	struct pci_dev *pdev = adapter->pdev;
4707 	int count = E1000_CHECK_RESET_COUNT;
4708 
4709 	while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
4710 		usleep_range(10000, 20000);
4711 
4712 	WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
4713 
4714 	pm_runtime_get_sync(&pdev->dev);
4715 
4716 	if (!test_bit(__E1000_DOWN, &adapter->state)) {
4717 		e1000e_down(adapter, true);
4718 		e1000_free_irq(adapter);
4719 
4720 		/* Link status message must follow this format */
4721 		pr_info("%s NIC Link is Down\n", adapter->netdev->name);
4722 	}
4723 
4724 	napi_disable(&adapter->napi);
4725 
4726 	e1000e_free_tx_resources(adapter->tx_ring);
4727 	e1000e_free_rx_resources(adapter->rx_ring);
4728 
4729 	/* kill manageability vlan ID if supported, but not if a vlan with
4730 	 * the same ID is registered on the host OS (let 8021q kill it)
4731 	 */
4732 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN)
4733 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
4734 				       adapter->mng_vlan_id);
4735 
4736 	/* If AMT is enabled, let the firmware know that the network
4737 	 * interface is now closed
4738 	 */
4739 	if ((adapter->flags & FLAG_HAS_AMT) &&
4740 	    !test_bit(__E1000_TESTING, &adapter->state))
4741 		e1000e_release_hw_control(adapter);
4742 
4743 	pm_qos_remove_request(&adapter->pm_qos_req);
4744 
4745 	pm_runtime_put_sync(&pdev->dev);
4746 
4747 	return 0;
4748 }
4749 
4750 /**
4751  * e1000_set_mac - Change the Ethernet Address of the NIC
4752  * @netdev: network interface device structure
4753  * @p: pointer to an address structure
4754  *
4755  * Returns 0 on success, negative on failure
4756  **/
4757 static int e1000_set_mac(struct net_device *netdev, void *p)
4758 {
4759 	struct e1000_adapter *adapter = netdev_priv(netdev);
4760 	struct e1000_hw *hw = &adapter->hw;
4761 	struct sockaddr *addr = p;
4762 
4763 	if (!is_valid_ether_addr(addr->sa_data))
4764 		return -EADDRNOTAVAIL;
4765 
4766 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
4767 	memcpy(adapter->hw.mac.addr, addr->sa_data, netdev->addr_len);
4768 
4769 	hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
4770 
4771 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA) {
4772 		/* activate the work around */
4773 		e1000e_set_laa_state_82571(&adapter->hw, 1);
4774 
4775 		/* Hold a copy of the LAA in RAR[14] This is done so that
4776 		 * between the time RAR[0] gets clobbered  and the time it
4777 		 * gets fixed (in e1000_watchdog), the actual LAA is in one
4778 		 * of the RARs and no incoming packets directed to this port
4779 		 * are dropped. Eventually the LAA will be in RAR[0] and
4780 		 * RAR[14]
4781 		 */
4782 		hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr,
4783 				    adapter->hw.mac.rar_entry_count - 1);
4784 	}
4785 
4786 	return 0;
4787 }
4788 
4789 /**
4790  * e1000e_update_phy_task - work thread to update phy
4791  * @work: pointer to our work struct
4792  *
4793  * this worker thread exists because we must acquire a
4794  * semaphore to read the phy, which we could msleep while
4795  * waiting for it, and we can't msleep in a timer.
4796  **/
4797 static void e1000e_update_phy_task(struct work_struct *work)
4798 {
4799 	struct e1000_adapter *adapter = container_of(work,
4800 						     struct e1000_adapter,
4801 						     update_phy_task);
4802 	struct e1000_hw *hw = &adapter->hw;
4803 
4804 	if (test_bit(__E1000_DOWN, &adapter->state))
4805 		return;
4806 
4807 	e1000_get_phy_info(hw);
4808 
4809 	/* Enable EEE on 82579 after link up */
4810 	if (hw->phy.type >= e1000_phy_82579)
4811 		e1000_set_eee_pchlan(hw);
4812 }
4813 
4814 /**
4815  * e1000_update_phy_info - timre call-back to update PHY info
4816  * @data: pointer to adapter cast into an unsigned long
4817  *
4818  * Need to wait a few seconds after link up to get diagnostic information from
4819  * the phy
4820  **/
4821 static void e1000_update_phy_info(struct timer_list *t)
4822 {
4823 	struct e1000_adapter *adapter = from_timer(adapter, t, phy_info_timer);
4824 
4825 	if (test_bit(__E1000_DOWN, &adapter->state))
4826 		return;
4827 
4828 	schedule_work(&adapter->update_phy_task);
4829 }
4830 
4831 /**
4832  * e1000e_update_phy_stats - Update the PHY statistics counters
4833  * @adapter: board private structure
4834  *
4835  * Read/clear the upper 16-bit PHY registers and read/accumulate lower
4836  **/
4837 static void e1000e_update_phy_stats(struct e1000_adapter *adapter)
4838 {
4839 	struct e1000_hw *hw = &adapter->hw;
4840 	s32 ret_val;
4841 	u16 phy_data;
4842 
4843 	ret_val = hw->phy.ops.acquire(hw);
4844 	if (ret_val)
4845 		return;
4846 
4847 	/* A page set is expensive so check if already on desired page.
4848 	 * If not, set to the page with the PHY status registers.
4849 	 */
4850 	hw->phy.addr = 1;
4851 	ret_val = e1000e_read_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4852 					   &phy_data);
4853 	if (ret_val)
4854 		goto release;
4855 	if (phy_data != (HV_STATS_PAGE << IGP_PAGE_SHIFT)) {
4856 		ret_val = hw->phy.ops.set_page(hw,
4857 					       HV_STATS_PAGE << IGP_PAGE_SHIFT);
4858 		if (ret_val)
4859 			goto release;
4860 	}
4861 
4862 	/* Single Collision Count */
4863 	hw->phy.ops.read_reg_page(hw, HV_SCC_UPPER, &phy_data);
4864 	ret_val = hw->phy.ops.read_reg_page(hw, HV_SCC_LOWER, &phy_data);
4865 	if (!ret_val)
4866 		adapter->stats.scc += phy_data;
4867 
4868 	/* Excessive Collision Count */
4869 	hw->phy.ops.read_reg_page(hw, HV_ECOL_UPPER, &phy_data);
4870 	ret_val = hw->phy.ops.read_reg_page(hw, HV_ECOL_LOWER, &phy_data);
4871 	if (!ret_val)
4872 		adapter->stats.ecol += phy_data;
4873 
4874 	/* Multiple Collision Count */
4875 	hw->phy.ops.read_reg_page(hw, HV_MCC_UPPER, &phy_data);
4876 	ret_val = hw->phy.ops.read_reg_page(hw, HV_MCC_LOWER, &phy_data);
4877 	if (!ret_val)
4878 		adapter->stats.mcc += phy_data;
4879 
4880 	/* Late Collision Count */
4881 	hw->phy.ops.read_reg_page(hw, HV_LATECOL_UPPER, &phy_data);
4882 	ret_val = hw->phy.ops.read_reg_page(hw, HV_LATECOL_LOWER, &phy_data);
4883 	if (!ret_val)
4884 		adapter->stats.latecol += phy_data;
4885 
4886 	/* Collision Count - also used for adaptive IFS */
4887 	hw->phy.ops.read_reg_page(hw, HV_COLC_UPPER, &phy_data);
4888 	ret_val = hw->phy.ops.read_reg_page(hw, HV_COLC_LOWER, &phy_data);
4889 	if (!ret_val)
4890 		hw->mac.collision_delta = phy_data;
4891 
4892 	/* Defer Count */
4893 	hw->phy.ops.read_reg_page(hw, HV_DC_UPPER, &phy_data);
4894 	ret_val = hw->phy.ops.read_reg_page(hw, HV_DC_LOWER, &phy_data);
4895 	if (!ret_val)
4896 		adapter->stats.dc += phy_data;
4897 
4898 	/* Transmit with no CRS */
4899 	hw->phy.ops.read_reg_page(hw, HV_TNCRS_UPPER, &phy_data);
4900 	ret_val = hw->phy.ops.read_reg_page(hw, HV_TNCRS_LOWER, &phy_data);
4901 	if (!ret_val)
4902 		adapter->stats.tncrs += phy_data;
4903 
4904 release:
4905 	hw->phy.ops.release(hw);
4906 }
4907 
4908 /**
4909  * e1000e_update_stats - Update the board statistics counters
4910  * @adapter: board private structure
4911  **/
4912 static void e1000e_update_stats(struct e1000_adapter *adapter)
4913 {
4914 	struct net_device *netdev = adapter->netdev;
4915 	struct e1000_hw *hw = &adapter->hw;
4916 	struct pci_dev *pdev = adapter->pdev;
4917 
4918 	/* Prevent stats update while adapter is being reset, or if the pci
4919 	 * connection is down.
4920 	 */
4921 	if (adapter->link_speed == 0)
4922 		return;
4923 	if (pci_channel_offline(pdev))
4924 		return;
4925 
4926 	adapter->stats.crcerrs += er32(CRCERRS);
4927 	adapter->stats.gprc += er32(GPRC);
4928 	adapter->stats.gorc += er32(GORCL);
4929 	er32(GORCH);		/* Clear gorc */
4930 	adapter->stats.bprc += er32(BPRC);
4931 	adapter->stats.mprc += er32(MPRC);
4932 	adapter->stats.roc += er32(ROC);
4933 
4934 	adapter->stats.mpc += er32(MPC);
4935 
4936 	/* Half-duplex statistics */
4937 	if (adapter->link_duplex == HALF_DUPLEX) {
4938 		if (adapter->flags2 & FLAG2_HAS_PHY_STATS) {
4939 			e1000e_update_phy_stats(adapter);
4940 		} else {
4941 			adapter->stats.scc += er32(SCC);
4942 			adapter->stats.ecol += er32(ECOL);
4943 			adapter->stats.mcc += er32(MCC);
4944 			adapter->stats.latecol += er32(LATECOL);
4945 			adapter->stats.dc += er32(DC);
4946 
4947 			hw->mac.collision_delta = er32(COLC);
4948 
4949 			if ((hw->mac.type != e1000_82574) &&
4950 			    (hw->mac.type != e1000_82583))
4951 				adapter->stats.tncrs += er32(TNCRS);
4952 		}
4953 		adapter->stats.colc += hw->mac.collision_delta;
4954 	}
4955 
4956 	adapter->stats.xonrxc += er32(XONRXC);
4957 	adapter->stats.xontxc += er32(XONTXC);
4958 	adapter->stats.xoffrxc += er32(XOFFRXC);
4959 	adapter->stats.xofftxc += er32(XOFFTXC);
4960 	adapter->stats.gptc += er32(GPTC);
4961 	adapter->stats.gotc += er32(GOTCL);
4962 	er32(GOTCH);		/* Clear gotc */
4963 	adapter->stats.rnbc += er32(RNBC);
4964 	adapter->stats.ruc += er32(RUC);
4965 
4966 	adapter->stats.mptc += er32(MPTC);
4967 	adapter->stats.bptc += er32(BPTC);
4968 
4969 	/* used for adaptive IFS */
4970 
4971 	hw->mac.tx_packet_delta = er32(TPT);
4972 	adapter->stats.tpt += hw->mac.tx_packet_delta;
4973 
4974 	adapter->stats.algnerrc += er32(ALGNERRC);
4975 	adapter->stats.rxerrc += er32(RXERRC);
4976 	adapter->stats.cexterr += er32(CEXTERR);
4977 	adapter->stats.tsctc += er32(TSCTC);
4978 	adapter->stats.tsctfc += er32(TSCTFC);
4979 
4980 	/* Fill out the OS statistics structure */
4981 	netdev->stats.multicast = adapter->stats.mprc;
4982 	netdev->stats.collisions = adapter->stats.colc;
4983 
4984 	/* Rx Errors */
4985 
4986 	/* RLEC on some newer hardware can be incorrect so build
4987 	 * our own version based on RUC and ROC
4988 	 */
4989 	netdev->stats.rx_errors = adapter->stats.rxerrc +
4990 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
4991 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
4992 	netdev->stats.rx_length_errors = adapter->stats.ruc +
4993 	    adapter->stats.roc;
4994 	netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
4995 	netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
4996 	netdev->stats.rx_missed_errors = adapter->stats.mpc;
4997 
4998 	/* Tx Errors */
4999 	netdev->stats.tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5000 	netdev->stats.tx_aborted_errors = adapter->stats.ecol;
5001 	netdev->stats.tx_window_errors = adapter->stats.latecol;
5002 	netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
5003 
5004 	/* Tx Dropped needs to be maintained elsewhere */
5005 
5006 	/* Management Stats */
5007 	adapter->stats.mgptc += er32(MGTPTC);
5008 	adapter->stats.mgprc += er32(MGTPRC);
5009 	adapter->stats.mgpdc += er32(MGTPDC);
5010 
5011 	/* Correctable ECC Errors */
5012 	if (hw->mac.type >= e1000_pch_lpt) {
5013 		u32 pbeccsts = er32(PBECCSTS);
5014 
5015 		adapter->corr_errors +=
5016 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
5017 		adapter->uncorr_errors +=
5018 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
5019 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
5020 	}
5021 }
5022 
5023 /**
5024  * e1000_phy_read_status - Update the PHY register status snapshot
5025  * @adapter: board private structure
5026  **/
5027 static void e1000_phy_read_status(struct e1000_adapter *adapter)
5028 {
5029 	struct e1000_hw *hw = &adapter->hw;
5030 	struct e1000_phy_regs *phy = &adapter->phy_regs;
5031 
5032 	if (!pm_runtime_suspended((&adapter->pdev->dev)->parent) &&
5033 	    (er32(STATUS) & E1000_STATUS_LU) &&
5034 	    (adapter->hw.phy.media_type == e1000_media_type_copper)) {
5035 		int ret_val;
5036 
5037 		ret_val = e1e_rphy(hw, MII_BMCR, &phy->bmcr);
5038 		ret_val |= e1e_rphy(hw, MII_BMSR, &phy->bmsr);
5039 		ret_val |= e1e_rphy(hw, MII_ADVERTISE, &phy->advertise);
5040 		ret_val |= e1e_rphy(hw, MII_LPA, &phy->lpa);
5041 		ret_val |= e1e_rphy(hw, MII_EXPANSION, &phy->expansion);
5042 		ret_val |= e1e_rphy(hw, MII_CTRL1000, &phy->ctrl1000);
5043 		ret_val |= e1e_rphy(hw, MII_STAT1000, &phy->stat1000);
5044 		ret_val |= e1e_rphy(hw, MII_ESTATUS, &phy->estatus);
5045 		if (ret_val)
5046 			e_warn("Error reading PHY register\n");
5047 	} else {
5048 		/* Do not read PHY registers if link is not up
5049 		 * Set values to typical power-on defaults
5050 		 */
5051 		phy->bmcr = (BMCR_SPEED1000 | BMCR_ANENABLE | BMCR_FULLDPLX);
5052 		phy->bmsr = (BMSR_100FULL | BMSR_100HALF | BMSR_10FULL |
5053 			     BMSR_10HALF | BMSR_ESTATEN | BMSR_ANEGCAPABLE |
5054 			     BMSR_ERCAP);
5055 		phy->advertise = (ADVERTISE_PAUSE_ASYM | ADVERTISE_PAUSE_CAP |
5056 				  ADVERTISE_ALL | ADVERTISE_CSMA);
5057 		phy->lpa = 0;
5058 		phy->expansion = EXPANSION_ENABLENPAGE;
5059 		phy->ctrl1000 = ADVERTISE_1000FULL;
5060 		phy->stat1000 = 0;
5061 		phy->estatus = (ESTATUS_1000_TFULL | ESTATUS_1000_THALF);
5062 	}
5063 }
5064 
5065 static void e1000_print_link_info(struct e1000_adapter *adapter)
5066 {
5067 	struct e1000_hw *hw = &adapter->hw;
5068 	u32 ctrl = er32(CTRL);
5069 
5070 	/* Link status message must follow this format for user tools */
5071 	pr_info("%s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5072 		adapter->netdev->name, adapter->link_speed,
5073 		adapter->link_duplex == FULL_DUPLEX ? "Full" : "Half",
5074 		(ctrl & E1000_CTRL_TFCE) && (ctrl & E1000_CTRL_RFCE) ? "Rx/Tx" :
5075 		(ctrl & E1000_CTRL_RFCE) ? "Rx" :
5076 		(ctrl & E1000_CTRL_TFCE) ? "Tx" : "None");
5077 }
5078 
5079 static bool e1000e_has_link(struct e1000_adapter *adapter)
5080 {
5081 	struct e1000_hw *hw = &adapter->hw;
5082 	bool link_active = false;
5083 	s32 ret_val = 0;
5084 
5085 	/* get_link_status is set on LSC (link status) interrupt or
5086 	 * Rx sequence error interrupt.  get_link_status will stay
5087 	 * true until the check_for_link establishes link
5088 	 * for copper adapters ONLY
5089 	 */
5090 	switch (hw->phy.media_type) {
5091 	case e1000_media_type_copper:
5092 		if (hw->mac.get_link_status) {
5093 			ret_val = hw->mac.ops.check_for_link(hw);
5094 			link_active = !hw->mac.get_link_status;
5095 		} else {
5096 			link_active = true;
5097 		}
5098 		break;
5099 	case e1000_media_type_fiber:
5100 		ret_val = hw->mac.ops.check_for_link(hw);
5101 		link_active = !!(er32(STATUS) & E1000_STATUS_LU);
5102 		break;
5103 	case e1000_media_type_internal_serdes:
5104 		ret_val = hw->mac.ops.check_for_link(hw);
5105 		link_active = hw->mac.serdes_has_link;
5106 		break;
5107 	default:
5108 	case e1000_media_type_unknown:
5109 		break;
5110 	}
5111 
5112 	if ((ret_val == -E1000_ERR_PHY) && (hw->phy.type == e1000_phy_igp_3) &&
5113 	    (er32(CTRL) & E1000_PHY_CTRL_GBE_DISABLE)) {
5114 		/* See e1000_kmrn_lock_loss_workaround_ich8lan() */
5115 		e_info("Gigabit has been disabled, downgrading speed\n");
5116 	}
5117 
5118 	return link_active;
5119 }
5120 
5121 static void e1000e_enable_receives(struct e1000_adapter *adapter)
5122 {
5123 	/* make sure the receive unit is started */
5124 	if ((adapter->flags & FLAG_RX_NEEDS_RESTART) &&
5125 	    (adapter->flags & FLAG_RESTART_NOW)) {
5126 		struct e1000_hw *hw = &adapter->hw;
5127 		u32 rctl = er32(RCTL);
5128 
5129 		ew32(RCTL, rctl | E1000_RCTL_EN);
5130 		adapter->flags &= ~FLAG_RESTART_NOW;
5131 	}
5132 }
5133 
5134 static void e1000e_check_82574_phy_workaround(struct e1000_adapter *adapter)
5135 {
5136 	struct e1000_hw *hw = &adapter->hw;
5137 
5138 	/* With 82574 controllers, PHY needs to be checked periodically
5139 	 * for hung state and reset, if two calls return true
5140 	 */
5141 	if (e1000_check_phy_82574(hw))
5142 		adapter->phy_hang_count++;
5143 	else
5144 		adapter->phy_hang_count = 0;
5145 
5146 	if (adapter->phy_hang_count > 1) {
5147 		adapter->phy_hang_count = 0;
5148 		e_dbg("PHY appears hung - resetting\n");
5149 		schedule_work(&adapter->reset_task);
5150 	}
5151 }
5152 
5153 /**
5154  * e1000_watchdog - Timer Call-back
5155  * @data: pointer to adapter cast into an unsigned long
5156  **/
5157 static void e1000_watchdog(struct timer_list *t)
5158 {
5159 	struct e1000_adapter *adapter = from_timer(adapter, t, watchdog_timer);
5160 
5161 	/* Do the rest outside of interrupt context */
5162 	schedule_work(&adapter->watchdog_task);
5163 
5164 	/* TODO: make this use queue_delayed_work() */
5165 }
5166 
5167 static void e1000_watchdog_task(struct work_struct *work)
5168 {
5169 	struct e1000_adapter *adapter = container_of(work,
5170 						     struct e1000_adapter,
5171 						     watchdog_task);
5172 	struct net_device *netdev = adapter->netdev;
5173 	struct e1000_mac_info *mac = &adapter->hw.mac;
5174 	struct e1000_phy_info *phy = &adapter->hw.phy;
5175 	struct e1000_ring *tx_ring = adapter->tx_ring;
5176 	struct e1000_hw *hw = &adapter->hw;
5177 	u32 link, tctl;
5178 
5179 	if (test_bit(__E1000_DOWN, &adapter->state))
5180 		return;
5181 
5182 	link = e1000e_has_link(adapter);
5183 	if ((netif_carrier_ok(netdev)) && link) {
5184 		/* Cancel scheduled suspend requests. */
5185 		pm_runtime_resume(netdev->dev.parent);
5186 
5187 		e1000e_enable_receives(adapter);
5188 		goto link_up;
5189 	}
5190 
5191 	if ((e1000e_enable_tx_pkt_filtering(hw)) &&
5192 	    (adapter->mng_vlan_id != adapter->hw.mng_cookie.vlan_id))
5193 		e1000_update_mng_vlan(adapter);
5194 
5195 	if (link) {
5196 		if (!netif_carrier_ok(netdev)) {
5197 			bool txb2b = true;
5198 
5199 			/* Cancel scheduled suspend requests. */
5200 			pm_runtime_resume(netdev->dev.parent);
5201 
5202 			/* update snapshot of PHY registers on LSC */
5203 			e1000_phy_read_status(adapter);
5204 			mac->ops.get_link_up_info(&adapter->hw,
5205 						  &adapter->link_speed,
5206 						  &adapter->link_duplex);
5207 			e1000_print_link_info(adapter);
5208 
5209 			/* check if SmartSpeed worked */
5210 			e1000e_check_downshift(hw);
5211 			if (phy->speed_downgraded)
5212 				netdev_warn(netdev,
5213 					    "Link Speed was downgraded by SmartSpeed\n");
5214 
5215 			/* On supported PHYs, check for duplex mismatch only
5216 			 * if link has autonegotiated at 10/100 half
5217 			 */
5218 			if ((hw->phy.type == e1000_phy_igp_3 ||
5219 			     hw->phy.type == e1000_phy_bm) &&
5220 			    hw->mac.autoneg &&
5221 			    (adapter->link_speed == SPEED_10 ||
5222 			     adapter->link_speed == SPEED_100) &&
5223 			    (adapter->link_duplex == HALF_DUPLEX)) {
5224 				u16 autoneg_exp;
5225 
5226 				e1e_rphy(hw, MII_EXPANSION, &autoneg_exp);
5227 
5228 				if (!(autoneg_exp & EXPANSION_NWAY))
5229 					e_info("Autonegotiated half duplex but link partner cannot autoneg.  Try forcing full duplex if link gets many collisions.\n");
5230 			}
5231 
5232 			/* adjust timeout factor according to speed/duplex */
5233 			adapter->tx_timeout_factor = 1;
5234 			switch (adapter->link_speed) {
5235 			case SPEED_10:
5236 				txb2b = false;
5237 				adapter->tx_timeout_factor = 16;
5238 				break;
5239 			case SPEED_100:
5240 				txb2b = false;
5241 				adapter->tx_timeout_factor = 10;
5242 				break;
5243 			}
5244 
5245 			/* workaround: re-program speed mode bit after
5246 			 * link-up event
5247 			 */
5248 			if ((adapter->flags & FLAG_TARC_SPEED_MODE_BIT) &&
5249 			    !txb2b) {
5250 				u32 tarc0;
5251 
5252 				tarc0 = er32(TARC(0));
5253 				tarc0 &= ~SPEED_MODE_BIT;
5254 				ew32(TARC(0), tarc0);
5255 			}
5256 
5257 			/* disable TSO for pcie and 10/100 speeds, to avoid
5258 			 * some hardware issues
5259 			 */
5260 			if (!(adapter->flags & FLAG_TSO_FORCE)) {
5261 				switch (adapter->link_speed) {
5262 				case SPEED_10:
5263 				case SPEED_100:
5264 					e_info("10/100 speed: disabling TSO\n");
5265 					netdev->features &= ~NETIF_F_TSO;
5266 					netdev->features &= ~NETIF_F_TSO6;
5267 					break;
5268 				case SPEED_1000:
5269 					netdev->features |= NETIF_F_TSO;
5270 					netdev->features |= NETIF_F_TSO6;
5271 					break;
5272 				default:
5273 					/* oops */
5274 					break;
5275 				}
5276 			}
5277 
5278 			/* enable transmits in the hardware, need to do this
5279 			 * after setting TARC(0)
5280 			 */
5281 			tctl = er32(TCTL);
5282 			tctl |= E1000_TCTL_EN;
5283 			ew32(TCTL, tctl);
5284 
5285 			/* Perform any post-link-up configuration before
5286 			 * reporting link up.
5287 			 */
5288 			if (phy->ops.cfg_on_link_up)
5289 				phy->ops.cfg_on_link_up(hw);
5290 
5291 			netif_carrier_on(netdev);
5292 
5293 			if (!test_bit(__E1000_DOWN, &adapter->state))
5294 				mod_timer(&adapter->phy_info_timer,
5295 					  round_jiffies(jiffies + 2 * HZ));
5296 		}
5297 	} else {
5298 		if (netif_carrier_ok(netdev)) {
5299 			adapter->link_speed = 0;
5300 			adapter->link_duplex = 0;
5301 			/* Link status message must follow this format */
5302 			pr_info("%s NIC Link is Down\n", adapter->netdev->name);
5303 			netif_carrier_off(netdev);
5304 			if (!test_bit(__E1000_DOWN, &adapter->state))
5305 				mod_timer(&adapter->phy_info_timer,
5306 					  round_jiffies(jiffies + 2 * HZ));
5307 
5308 			/* 8000ES2LAN requires a Rx packet buffer work-around
5309 			 * on link down event; reset the controller to flush
5310 			 * the Rx packet buffer.
5311 			 */
5312 			if (adapter->flags & FLAG_RX_NEEDS_RESTART)
5313 				adapter->flags |= FLAG_RESTART_NOW;
5314 			else
5315 				pm_schedule_suspend(netdev->dev.parent,
5316 						    LINK_TIMEOUT);
5317 		}
5318 	}
5319 
5320 link_up:
5321 	spin_lock(&adapter->stats64_lock);
5322 	e1000e_update_stats(adapter);
5323 
5324 	mac->tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
5325 	adapter->tpt_old = adapter->stats.tpt;
5326 	mac->collision_delta = adapter->stats.colc - adapter->colc_old;
5327 	adapter->colc_old = adapter->stats.colc;
5328 
5329 	adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
5330 	adapter->gorc_old = adapter->stats.gorc;
5331 	adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
5332 	adapter->gotc_old = adapter->stats.gotc;
5333 	spin_unlock(&adapter->stats64_lock);
5334 
5335 	/* If the link is lost the controller stops DMA, but
5336 	 * if there is queued Tx work it cannot be done.  So
5337 	 * reset the controller to flush the Tx packet buffers.
5338 	 */
5339 	if (!netif_carrier_ok(netdev) &&
5340 	    (e1000_desc_unused(tx_ring) + 1 < tx_ring->count))
5341 		adapter->flags |= FLAG_RESTART_NOW;
5342 
5343 	/* If reset is necessary, do it outside of interrupt context. */
5344 	if (adapter->flags & FLAG_RESTART_NOW) {
5345 		schedule_work(&adapter->reset_task);
5346 		/* return immediately since reset is imminent */
5347 		return;
5348 	}
5349 
5350 	e1000e_update_adaptive(&adapter->hw);
5351 
5352 	/* Simple mode for Interrupt Throttle Rate (ITR) */
5353 	if (adapter->itr_setting == 4) {
5354 		/* Symmetric Tx/Rx gets a reduced ITR=2000;
5355 		 * Total asymmetrical Tx or Rx gets ITR=8000;
5356 		 * everyone else is between 2000-8000.
5357 		 */
5358 		u32 goc = (adapter->gotc + adapter->gorc) / 10000;
5359 		u32 dif = (adapter->gotc > adapter->gorc ?
5360 			   adapter->gotc - adapter->gorc :
5361 			   adapter->gorc - adapter->gotc) / 10000;
5362 		u32 itr = goc > 0 ? (dif * 6000 / goc + 2000) : 8000;
5363 
5364 		e1000e_write_itr(adapter, itr);
5365 	}
5366 
5367 	/* Cause software interrupt to ensure Rx ring is cleaned */
5368 	if (adapter->msix_entries)
5369 		ew32(ICS, adapter->rx_ring->ims_val);
5370 	else
5371 		ew32(ICS, E1000_ICS_RXDMT0);
5372 
5373 	/* flush pending descriptors to memory before detecting Tx hang */
5374 	e1000e_flush_descriptors(adapter);
5375 
5376 	/* Force detection of hung controller every watchdog period */
5377 	adapter->detect_tx_hung = true;
5378 
5379 	/* With 82571 controllers, LAA may be overwritten due to controller
5380 	 * reset from the other port. Set the appropriate LAA in RAR[0]
5381 	 */
5382 	if (e1000e_get_laa_state_82571(hw))
5383 		hw->mac.ops.rar_set(hw, adapter->hw.mac.addr, 0);
5384 
5385 	if (adapter->flags2 & FLAG2_CHECK_PHY_HANG)
5386 		e1000e_check_82574_phy_workaround(adapter);
5387 
5388 	/* Clear valid timestamp stuck in RXSTMPL/H due to a Rx error */
5389 	if (adapter->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE) {
5390 		if ((adapter->flags2 & FLAG2_CHECK_RX_HWTSTAMP) &&
5391 		    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID)) {
5392 			er32(RXSTMPH);
5393 			adapter->rx_hwtstamp_cleared++;
5394 		} else {
5395 			adapter->flags2 |= FLAG2_CHECK_RX_HWTSTAMP;
5396 		}
5397 	}
5398 
5399 	/* Reset the timer */
5400 	if (!test_bit(__E1000_DOWN, &adapter->state))
5401 		mod_timer(&adapter->watchdog_timer,
5402 			  round_jiffies(jiffies + 2 * HZ));
5403 }
5404 
5405 #define E1000_TX_FLAGS_CSUM		0x00000001
5406 #define E1000_TX_FLAGS_VLAN		0x00000002
5407 #define E1000_TX_FLAGS_TSO		0x00000004
5408 #define E1000_TX_FLAGS_IPV4		0x00000008
5409 #define E1000_TX_FLAGS_NO_FCS		0x00000010
5410 #define E1000_TX_FLAGS_HWTSTAMP		0x00000020
5411 #define E1000_TX_FLAGS_VLAN_MASK	0xffff0000
5412 #define E1000_TX_FLAGS_VLAN_SHIFT	16
5413 
5414 static int e1000_tso(struct e1000_ring *tx_ring, struct sk_buff *skb,
5415 		     __be16 protocol)
5416 {
5417 	struct e1000_context_desc *context_desc;
5418 	struct e1000_buffer *buffer_info;
5419 	unsigned int i;
5420 	u32 cmd_length = 0;
5421 	u16 ipcse = 0, mss;
5422 	u8 ipcss, ipcso, tucss, tucso, hdr_len;
5423 	int err;
5424 
5425 	if (!skb_is_gso(skb))
5426 		return 0;
5427 
5428 	err = skb_cow_head(skb, 0);
5429 	if (err < 0)
5430 		return err;
5431 
5432 	hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5433 	mss = skb_shinfo(skb)->gso_size;
5434 	if (protocol == htons(ETH_P_IP)) {
5435 		struct iphdr *iph = ip_hdr(skb);
5436 		iph->tot_len = 0;
5437 		iph->check = 0;
5438 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
5439 							 0, IPPROTO_TCP, 0);
5440 		cmd_length = E1000_TXD_CMD_IP;
5441 		ipcse = skb_transport_offset(skb) - 1;
5442 	} else if (skb_is_gso_v6(skb)) {
5443 		ipv6_hdr(skb)->payload_len = 0;
5444 		tcp_hdr(skb)->check = ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
5445 						       &ipv6_hdr(skb)->daddr,
5446 						       0, IPPROTO_TCP, 0);
5447 		ipcse = 0;
5448 	}
5449 	ipcss = skb_network_offset(skb);
5450 	ipcso = (void *)&(ip_hdr(skb)->check) - (void *)skb->data;
5451 	tucss = skb_transport_offset(skb);
5452 	tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
5453 
5454 	cmd_length |= (E1000_TXD_CMD_DEXT | E1000_TXD_CMD_TSE |
5455 		       E1000_TXD_CMD_TCP | (skb->len - (hdr_len)));
5456 
5457 	i = tx_ring->next_to_use;
5458 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5459 	buffer_info = &tx_ring->buffer_info[i];
5460 
5461 	context_desc->lower_setup.ip_fields.ipcss = ipcss;
5462 	context_desc->lower_setup.ip_fields.ipcso = ipcso;
5463 	context_desc->lower_setup.ip_fields.ipcse = cpu_to_le16(ipcse);
5464 	context_desc->upper_setup.tcp_fields.tucss = tucss;
5465 	context_desc->upper_setup.tcp_fields.tucso = tucso;
5466 	context_desc->upper_setup.tcp_fields.tucse = 0;
5467 	context_desc->tcp_seg_setup.fields.mss = cpu_to_le16(mss);
5468 	context_desc->tcp_seg_setup.fields.hdr_len = hdr_len;
5469 	context_desc->cmd_and_length = cpu_to_le32(cmd_length);
5470 
5471 	buffer_info->time_stamp = jiffies;
5472 	buffer_info->next_to_watch = i;
5473 
5474 	i++;
5475 	if (i == tx_ring->count)
5476 		i = 0;
5477 	tx_ring->next_to_use = i;
5478 
5479 	return 1;
5480 }
5481 
5482 static bool e1000_tx_csum(struct e1000_ring *tx_ring, struct sk_buff *skb,
5483 			  __be16 protocol)
5484 {
5485 	struct e1000_adapter *adapter = tx_ring->adapter;
5486 	struct e1000_context_desc *context_desc;
5487 	struct e1000_buffer *buffer_info;
5488 	unsigned int i;
5489 	u8 css;
5490 	u32 cmd_len = E1000_TXD_CMD_DEXT;
5491 
5492 	if (skb->ip_summed != CHECKSUM_PARTIAL)
5493 		return false;
5494 
5495 	switch (protocol) {
5496 	case cpu_to_be16(ETH_P_IP):
5497 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
5498 			cmd_len |= E1000_TXD_CMD_TCP;
5499 		break;
5500 	case cpu_to_be16(ETH_P_IPV6):
5501 		/* XXX not handling all IPV6 headers */
5502 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
5503 			cmd_len |= E1000_TXD_CMD_TCP;
5504 		break;
5505 	default:
5506 		if (unlikely(net_ratelimit()))
5507 			e_warn("checksum_partial proto=%x!\n",
5508 			       be16_to_cpu(protocol));
5509 		break;
5510 	}
5511 
5512 	css = skb_checksum_start_offset(skb);
5513 
5514 	i = tx_ring->next_to_use;
5515 	buffer_info = &tx_ring->buffer_info[i];
5516 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5517 
5518 	context_desc->lower_setup.ip_config = 0;
5519 	context_desc->upper_setup.tcp_fields.tucss = css;
5520 	context_desc->upper_setup.tcp_fields.tucso = css + skb->csum_offset;
5521 	context_desc->upper_setup.tcp_fields.tucse = 0;
5522 	context_desc->tcp_seg_setup.data = 0;
5523 	context_desc->cmd_and_length = cpu_to_le32(cmd_len);
5524 
5525 	buffer_info->time_stamp = jiffies;
5526 	buffer_info->next_to_watch = i;
5527 
5528 	i++;
5529 	if (i == tx_ring->count)
5530 		i = 0;
5531 	tx_ring->next_to_use = i;
5532 
5533 	return true;
5534 }
5535 
5536 static int e1000_tx_map(struct e1000_ring *tx_ring, struct sk_buff *skb,
5537 			unsigned int first, unsigned int max_per_txd,
5538 			unsigned int nr_frags)
5539 {
5540 	struct e1000_adapter *adapter = tx_ring->adapter;
5541 	struct pci_dev *pdev = adapter->pdev;
5542 	struct e1000_buffer *buffer_info;
5543 	unsigned int len = skb_headlen(skb);
5544 	unsigned int offset = 0, size, count = 0, i;
5545 	unsigned int f, bytecount, segs;
5546 
5547 	i = tx_ring->next_to_use;
5548 
5549 	while (len) {
5550 		buffer_info = &tx_ring->buffer_info[i];
5551 		size = min(len, max_per_txd);
5552 
5553 		buffer_info->length = size;
5554 		buffer_info->time_stamp = jiffies;
5555 		buffer_info->next_to_watch = i;
5556 		buffer_info->dma = dma_map_single(&pdev->dev,
5557 						  skb->data + offset,
5558 						  size, DMA_TO_DEVICE);
5559 		buffer_info->mapped_as_page = false;
5560 		if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5561 			goto dma_error;
5562 
5563 		len -= size;
5564 		offset += size;
5565 		count++;
5566 
5567 		if (len) {
5568 			i++;
5569 			if (i == tx_ring->count)
5570 				i = 0;
5571 		}
5572 	}
5573 
5574 	for (f = 0; f < nr_frags; f++) {
5575 		const struct skb_frag_struct *frag;
5576 
5577 		frag = &skb_shinfo(skb)->frags[f];
5578 		len = skb_frag_size(frag);
5579 		offset = 0;
5580 
5581 		while (len) {
5582 			i++;
5583 			if (i == tx_ring->count)
5584 				i = 0;
5585 
5586 			buffer_info = &tx_ring->buffer_info[i];
5587 			size = min(len, max_per_txd);
5588 
5589 			buffer_info->length = size;
5590 			buffer_info->time_stamp = jiffies;
5591 			buffer_info->next_to_watch = i;
5592 			buffer_info->dma = skb_frag_dma_map(&pdev->dev, frag,
5593 							    offset, size,
5594 							    DMA_TO_DEVICE);
5595 			buffer_info->mapped_as_page = true;
5596 			if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5597 				goto dma_error;
5598 
5599 			len -= size;
5600 			offset += size;
5601 			count++;
5602 		}
5603 	}
5604 
5605 	segs = skb_shinfo(skb)->gso_segs ? : 1;
5606 	/* multiply data chunks by size of headers */
5607 	bytecount = ((segs - 1) * skb_headlen(skb)) + skb->len;
5608 
5609 	tx_ring->buffer_info[i].skb = skb;
5610 	tx_ring->buffer_info[i].segs = segs;
5611 	tx_ring->buffer_info[i].bytecount = bytecount;
5612 	tx_ring->buffer_info[first].next_to_watch = i;
5613 
5614 	return count;
5615 
5616 dma_error:
5617 	dev_err(&pdev->dev, "Tx DMA map failed\n");
5618 	buffer_info->dma = 0;
5619 	if (count)
5620 		count--;
5621 
5622 	while (count--) {
5623 		if (i == 0)
5624 			i += tx_ring->count;
5625 		i--;
5626 		buffer_info = &tx_ring->buffer_info[i];
5627 		e1000_put_txbuf(tx_ring, buffer_info, true);
5628 	}
5629 
5630 	return 0;
5631 }
5632 
5633 static void e1000_tx_queue(struct e1000_ring *tx_ring, int tx_flags, int count)
5634 {
5635 	struct e1000_adapter *adapter = tx_ring->adapter;
5636 	struct e1000_tx_desc *tx_desc = NULL;
5637 	struct e1000_buffer *buffer_info;
5638 	u32 txd_upper = 0, txd_lower = E1000_TXD_CMD_IFCS;
5639 	unsigned int i;
5640 
5641 	if (tx_flags & E1000_TX_FLAGS_TSO) {
5642 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D |
5643 		    E1000_TXD_CMD_TSE;
5644 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5645 
5646 		if (tx_flags & E1000_TX_FLAGS_IPV4)
5647 			txd_upper |= E1000_TXD_POPTS_IXSM << 8;
5648 	}
5649 
5650 	if (tx_flags & E1000_TX_FLAGS_CSUM) {
5651 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5652 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5653 	}
5654 
5655 	if (tx_flags & E1000_TX_FLAGS_VLAN) {
5656 		txd_lower |= E1000_TXD_CMD_VLE;
5657 		txd_upper |= (tx_flags & E1000_TX_FLAGS_VLAN_MASK);
5658 	}
5659 
5660 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5661 		txd_lower &= ~(E1000_TXD_CMD_IFCS);
5662 
5663 	if (unlikely(tx_flags & E1000_TX_FLAGS_HWTSTAMP)) {
5664 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5665 		txd_upper |= E1000_TXD_EXTCMD_TSTAMP;
5666 	}
5667 
5668 	i = tx_ring->next_to_use;
5669 
5670 	do {
5671 		buffer_info = &tx_ring->buffer_info[i];
5672 		tx_desc = E1000_TX_DESC(*tx_ring, i);
5673 		tx_desc->buffer_addr = cpu_to_le64(buffer_info->dma);
5674 		tx_desc->lower.data = cpu_to_le32(txd_lower |
5675 						  buffer_info->length);
5676 		tx_desc->upper.data = cpu_to_le32(txd_upper);
5677 
5678 		i++;
5679 		if (i == tx_ring->count)
5680 			i = 0;
5681 	} while (--count > 0);
5682 
5683 	tx_desc->lower.data |= cpu_to_le32(adapter->txd_cmd);
5684 
5685 	/* txd_cmd re-enables FCS, so we'll re-disable it here as desired. */
5686 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5687 		tx_desc->lower.data &= ~(cpu_to_le32(E1000_TXD_CMD_IFCS));
5688 
5689 	/* Force memory writes to complete before letting h/w
5690 	 * know there are new descriptors to fetch.  (Only
5691 	 * applicable for weak-ordered memory model archs,
5692 	 * such as IA-64).
5693 	 */
5694 	wmb();
5695 
5696 	tx_ring->next_to_use = i;
5697 }
5698 
5699 #define MINIMUM_DHCP_PACKET_SIZE 282
5700 static int e1000_transfer_dhcp_info(struct e1000_adapter *adapter,
5701 				    struct sk_buff *skb)
5702 {
5703 	struct e1000_hw *hw = &adapter->hw;
5704 	u16 length, offset;
5705 
5706 	if (skb_vlan_tag_present(skb) &&
5707 	    !((skb_vlan_tag_get(skb) == adapter->hw.mng_cookie.vlan_id) &&
5708 	      (adapter->hw.mng_cookie.status &
5709 	       E1000_MNG_DHCP_COOKIE_STATUS_VLAN)))
5710 		return 0;
5711 
5712 	if (skb->len <= MINIMUM_DHCP_PACKET_SIZE)
5713 		return 0;
5714 
5715 	if (((struct ethhdr *)skb->data)->h_proto != htons(ETH_P_IP))
5716 		return 0;
5717 
5718 	{
5719 		const struct iphdr *ip = (struct iphdr *)((u8 *)skb->data + 14);
5720 		struct udphdr *udp;
5721 
5722 		if (ip->protocol != IPPROTO_UDP)
5723 			return 0;
5724 
5725 		udp = (struct udphdr *)((u8 *)ip + (ip->ihl << 2));
5726 		if (ntohs(udp->dest) != 67)
5727 			return 0;
5728 
5729 		offset = (u8 *)udp + 8 - skb->data;
5730 		length = skb->len - offset;
5731 		return e1000e_mng_write_dhcp_info(hw, (u8 *)udp + 8, length);
5732 	}
5733 
5734 	return 0;
5735 }
5736 
5737 static int __e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5738 {
5739 	struct e1000_adapter *adapter = tx_ring->adapter;
5740 
5741 	netif_stop_queue(adapter->netdev);
5742 	/* Herbert's original patch had:
5743 	 *  smp_mb__after_netif_stop_queue();
5744 	 * but since that doesn't exist yet, just open code it.
5745 	 */
5746 	smp_mb();
5747 
5748 	/* We need to check again in a case another CPU has just
5749 	 * made room available.
5750 	 */
5751 	if (e1000_desc_unused(tx_ring) < size)
5752 		return -EBUSY;
5753 
5754 	/* A reprieve! */
5755 	netif_start_queue(adapter->netdev);
5756 	++adapter->restart_queue;
5757 	return 0;
5758 }
5759 
5760 static int e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5761 {
5762 	BUG_ON(size > tx_ring->count);
5763 
5764 	if (e1000_desc_unused(tx_ring) >= size)
5765 		return 0;
5766 	return __e1000_maybe_stop_tx(tx_ring, size);
5767 }
5768 
5769 static netdev_tx_t e1000_xmit_frame(struct sk_buff *skb,
5770 				    struct net_device *netdev)
5771 {
5772 	struct e1000_adapter *adapter = netdev_priv(netdev);
5773 	struct e1000_ring *tx_ring = adapter->tx_ring;
5774 	unsigned int first;
5775 	unsigned int tx_flags = 0;
5776 	unsigned int len = skb_headlen(skb);
5777 	unsigned int nr_frags;
5778 	unsigned int mss;
5779 	int count = 0;
5780 	int tso;
5781 	unsigned int f;
5782 	__be16 protocol = vlan_get_protocol(skb);
5783 
5784 	if (test_bit(__E1000_DOWN, &adapter->state)) {
5785 		dev_kfree_skb_any(skb);
5786 		return NETDEV_TX_OK;
5787 	}
5788 
5789 	if (skb->len <= 0) {
5790 		dev_kfree_skb_any(skb);
5791 		return NETDEV_TX_OK;
5792 	}
5793 
5794 	/* The minimum packet size with TCTL.PSP set is 17 bytes so
5795 	 * pad skb in order to meet this minimum size requirement
5796 	 */
5797 	if (skb_put_padto(skb, 17))
5798 		return NETDEV_TX_OK;
5799 
5800 	mss = skb_shinfo(skb)->gso_size;
5801 	if (mss) {
5802 		u8 hdr_len;
5803 
5804 		/* TSO Workaround for 82571/2/3 Controllers -- if skb->data
5805 		 * points to just header, pull a few bytes of payload from
5806 		 * frags into skb->data
5807 		 */
5808 		hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5809 		/* we do this workaround for ES2LAN, but it is un-necessary,
5810 		 * avoiding it could save a lot of cycles
5811 		 */
5812 		if (skb->data_len && (hdr_len == len)) {
5813 			unsigned int pull_size;
5814 
5815 			pull_size = min_t(unsigned int, 4, skb->data_len);
5816 			if (!__pskb_pull_tail(skb, pull_size)) {
5817 				e_err("__pskb_pull_tail failed.\n");
5818 				dev_kfree_skb_any(skb);
5819 				return NETDEV_TX_OK;
5820 			}
5821 			len = skb_headlen(skb);
5822 		}
5823 	}
5824 
5825 	/* reserve a descriptor for the offload context */
5826 	if ((mss) || (skb->ip_summed == CHECKSUM_PARTIAL))
5827 		count++;
5828 	count++;
5829 
5830 	count += DIV_ROUND_UP(len, adapter->tx_fifo_limit);
5831 
5832 	nr_frags = skb_shinfo(skb)->nr_frags;
5833 	for (f = 0; f < nr_frags; f++)
5834 		count += DIV_ROUND_UP(skb_frag_size(&skb_shinfo(skb)->frags[f]),
5835 				      adapter->tx_fifo_limit);
5836 
5837 	if (adapter->hw.mac.tx_pkt_filtering)
5838 		e1000_transfer_dhcp_info(adapter, skb);
5839 
5840 	/* need: count + 2 desc gap to keep tail from touching
5841 	 * head, otherwise try next time
5842 	 */
5843 	if (e1000_maybe_stop_tx(tx_ring, count + 2))
5844 		return NETDEV_TX_BUSY;
5845 
5846 	if (skb_vlan_tag_present(skb)) {
5847 		tx_flags |= E1000_TX_FLAGS_VLAN;
5848 		tx_flags |= (skb_vlan_tag_get(skb) <<
5849 			     E1000_TX_FLAGS_VLAN_SHIFT);
5850 	}
5851 
5852 	first = tx_ring->next_to_use;
5853 
5854 	tso = e1000_tso(tx_ring, skb, protocol);
5855 	if (tso < 0) {
5856 		dev_kfree_skb_any(skb);
5857 		return NETDEV_TX_OK;
5858 	}
5859 
5860 	if (tso)
5861 		tx_flags |= E1000_TX_FLAGS_TSO;
5862 	else if (e1000_tx_csum(tx_ring, skb, protocol))
5863 		tx_flags |= E1000_TX_FLAGS_CSUM;
5864 
5865 	/* Old method was to assume IPv4 packet by default if TSO was enabled.
5866 	 * 82571 hardware supports TSO capabilities for IPv6 as well...
5867 	 * no longer assume, we must.
5868 	 */
5869 	if (protocol == htons(ETH_P_IP))
5870 		tx_flags |= E1000_TX_FLAGS_IPV4;
5871 
5872 	if (unlikely(skb->no_fcs))
5873 		tx_flags |= E1000_TX_FLAGS_NO_FCS;
5874 
5875 	/* if count is 0 then mapping error has occurred */
5876 	count = e1000_tx_map(tx_ring, skb, first, adapter->tx_fifo_limit,
5877 			     nr_frags);
5878 	if (count) {
5879 		if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
5880 		    (adapter->flags & FLAG_HAS_HW_TIMESTAMP)) {
5881 			if (!adapter->tx_hwtstamp_skb) {
5882 				skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5883 				tx_flags |= E1000_TX_FLAGS_HWTSTAMP;
5884 				adapter->tx_hwtstamp_skb = skb_get(skb);
5885 				adapter->tx_hwtstamp_start = jiffies;
5886 				schedule_work(&adapter->tx_hwtstamp_work);
5887 			} else {
5888 				adapter->tx_hwtstamp_skipped++;
5889 			}
5890 		}
5891 
5892 		skb_tx_timestamp(skb);
5893 
5894 		netdev_sent_queue(netdev, skb->len);
5895 		e1000_tx_queue(tx_ring, tx_flags, count);
5896 		/* Make sure there is space in the ring for the next send. */
5897 		e1000_maybe_stop_tx(tx_ring,
5898 				    (MAX_SKB_FRAGS *
5899 				     DIV_ROUND_UP(PAGE_SIZE,
5900 						  adapter->tx_fifo_limit) + 2));
5901 
5902 		if (!skb->xmit_more ||
5903 		    netif_xmit_stopped(netdev_get_tx_queue(netdev, 0))) {
5904 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
5905 				e1000e_update_tdt_wa(tx_ring,
5906 						     tx_ring->next_to_use);
5907 			else
5908 				writel(tx_ring->next_to_use, tx_ring->tail);
5909 
5910 			/* we need this if more than one processor can write
5911 			 * to our tail at a time, it synchronizes IO on
5912 			 *IA64/Altix systems
5913 			 */
5914 			mmiowb();
5915 		}
5916 	} else {
5917 		dev_kfree_skb_any(skb);
5918 		tx_ring->buffer_info[first].time_stamp = 0;
5919 		tx_ring->next_to_use = first;
5920 	}
5921 
5922 	return NETDEV_TX_OK;
5923 }
5924 
5925 /**
5926  * e1000_tx_timeout - Respond to a Tx Hang
5927  * @netdev: network interface device structure
5928  **/
5929 static void e1000_tx_timeout(struct net_device *netdev)
5930 {
5931 	struct e1000_adapter *adapter = netdev_priv(netdev);
5932 
5933 	/* Do the reset outside of interrupt context */
5934 	adapter->tx_timeout_count++;
5935 	schedule_work(&adapter->reset_task);
5936 }
5937 
5938 static void e1000_reset_task(struct work_struct *work)
5939 {
5940 	struct e1000_adapter *adapter;
5941 	adapter = container_of(work, struct e1000_adapter, reset_task);
5942 
5943 	/* don't run the task if already down */
5944 	if (test_bit(__E1000_DOWN, &adapter->state))
5945 		return;
5946 
5947 	if (!(adapter->flags & FLAG_RESTART_NOW)) {
5948 		e1000e_dump(adapter);
5949 		e_err("Reset adapter unexpectedly\n");
5950 	}
5951 	e1000e_reinit_locked(adapter);
5952 }
5953 
5954 /**
5955  * e1000_get_stats64 - Get System Network Statistics
5956  * @netdev: network interface device structure
5957  * @stats: rtnl_link_stats64 pointer
5958  *
5959  * Returns the address of the device statistics structure.
5960  **/
5961 void e1000e_get_stats64(struct net_device *netdev,
5962 			struct rtnl_link_stats64 *stats)
5963 {
5964 	struct e1000_adapter *adapter = netdev_priv(netdev);
5965 
5966 	spin_lock(&adapter->stats64_lock);
5967 	e1000e_update_stats(adapter);
5968 	/* Fill out the OS statistics structure */
5969 	stats->rx_bytes = adapter->stats.gorc;
5970 	stats->rx_packets = adapter->stats.gprc;
5971 	stats->tx_bytes = adapter->stats.gotc;
5972 	stats->tx_packets = adapter->stats.gptc;
5973 	stats->multicast = adapter->stats.mprc;
5974 	stats->collisions = adapter->stats.colc;
5975 
5976 	/* Rx Errors */
5977 
5978 	/* RLEC on some newer hardware can be incorrect so build
5979 	 * our own version based on RUC and ROC
5980 	 */
5981 	stats->rx_errors = adapter->stats.rxerrc +
5982 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
5983 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5984 	stats->rx_length_errors = adapter->stats.ruc + adapter->stats.roc;
5985 	stats->rx_crc_errors = adapter->stats.crcerrs;
5986 	stats->rx_frame_errors = adapter->stats.algnerrc;
5987 	stats->rx_missed_errors = adapter->stats.mpc;
5988 
5989 	/* Tx Errors */
5990 	stats->tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5991 	stats->tx_aborted_errors = adapter->stats.ecol;
5992 	stats->tx_window_errors = adapter->stats.latecol;
5993 	stats->tx_carrier_errors = adapter->stats.tncrs;
5994 
5995 	/* Tx Dropped needs to be maintained elsewhere */
5996 
5997 	spin_unlock(&adapter->stats64_lock);
5998 }
5999 
6000 /**
6001  * e1000_change_mtu - Change the Maximum Transfer Unit
6002  * @netdev: network interface device structure
6003  * @new_mtu: new value for maximum frame size
6004  *
6005  * Returns 0 on success, negative on failure
6006  **/
6007 static int e1000_change_mtu(struct net_device *netdev, int new_mtu)
6008 {
6009 	struct e1000_adapter *adapter = netdev_priv(netdev);
6010 	int max_frame = new_mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
6011 
6012 	/* Jumbo frame support */
6013 	if ((new_mtu > ETH_DATA_LEN) &&
6014 	    !(adapter->flags & FLAG_HAS_JUMBO_FRAMES)) {
6015 		e_err("Jumbo Frames not supported.\n");
6016 		return -EINVAL;
6017 	}
6018 
6019 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
6020 	if ((adapter->hw.mac.type >= e1000_pch2lan) &&
6021 	    !(adapter->flags2 & FLAG2_CRC_STRIPPING) &&
6022 	    (new_mtu > ETH_DATA_LEN)) {
6023 		e_err("Jumbo Frames not supported on this device when CRC stripping is disabled.\n");
6024 		return -EINVAL;
6025 	}
6026 
6027 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
6028 		usleep_range(1000, 2000);
6029 	/* e1000e_down -> e1000e_reset dependent on max_frame_size & mtu */
6030 	adapter->max_frame_size = max_frame;
6031 	e_info("changing MTU from %d to %d\n", netdev->mtu, new_mtu);
6032 	netdev->mtu = new_mtu;
6033 
6034 	pm_runtime_get_sync(netdev->dev.parent);
6035 
6036 	if (netif_running(netdev))
6037 		e1000e_down(adapter, true);
6038 
6039 	/* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
6040 	 * means we reserve 2 more, this pushes us to allocate from the next
6041 	 * larger slab size.
6042 	 * i.e. RXBUFFER_2048 --> size-4096 slab
6043 	 * However with the new *_jumbo_rx* routines, jumbo receives will use
6044 	 * fragmented skbs
6045 	 */
6046 
6047 	if (max_frame <= 2048)
6048 		adapter->rx_buffer_len = 2048;
6049 	else
6050 		adapter->rx_buffer_len = 4096;
6051 
6052 	/* adjust allocation if LPE protects us, and we aren't using SBP */
6053 	if (max_frame <= (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN))
6054 		adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
6055 
6056 	if (netif_running(netdev))
6057 		e1000e_up(adapter);
6058 	else
6059 		e1000e_reset(adapter);
6060 
6061 	pm_runtime_put_sync(netdev->dev.parent);
6062 
6063 	clear_bit(__E1000_RESETTING, &adapter->state);
6064 
6065 	return 0;
6066 }
6067 
6068 static int e1000_mii_ioctl(struct net_device *netdev, struct ifreq *ifr,
6069 			   int cmd)
6070 {
6071 	struct e1000_adapter *adapter = netdev_priv(netdev);
6072 	struct mii_ioctl_data *data = if_mii(ifr);
6073 
6074 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
6075 		return -EOPNOTSUPP;
6076 
6077 	switch (cmd) {
6078 	case SIOCGMIIPHY:
6079 		data->phy_id = adapter->hw.phy.addr;
6080 		break;
6081 	case SIOCGMIIREG:
6082 		e1000_phy_read_status(adapter);
6083 
6084 		switch (data->reg_num & 0x1F) {
6085 		case MII_BMCR:
6086 			data->val_out = adapter->phy_regs.bmcr;
6087 			break;
6088 		case MII_BMSR:
6089 			data->val_out = adapter->phy_regs.bmsr;
6090 			break;
6091 		case MII_PHYSID1:
6092 			data->val_out = (adapter->hw.phy.id >> 16);
6093 			break;
6094 		case MII_PHYSID2:
6095 			data->val_out = (adapter->hw.phy.id & 0xFFFF);
6096 			break;
6097 		case MII_ADVERTISE:
6098 			data->val_out = adapter->phy_regs.advertise;
6099 			break;
6100 		case MII_LPA:
6101 			data->val_out = adapter->phy_regs.lpa;
6102 			break;
6103 		case MII_EXPANSION:
6104 			data->val_out = adapter->phy_regs.expansion;
6105 			break;
6106 		case MII_CTRL1000:
6107 			data->val_out = adapter->phy_regs.ctrl1000;
6108 			break;
6109 		case MII_STAT1000:
6110 			data->val_out = adapter->phy_regs.stat1000;
6111 			break;
6112 		case MII_ESTATUS:
6113 			data->val_out = adapter->phy_regs.estatus;
6114 			break;
6115 		default:
6116 			return -EIO;
6117 		}
6118 		break;
6119 	case SIOCSMIIREG:
6120 	default:
6121 		return -EOPNOTSUPP;
6122 	}
6123 	return 0;
6124 }
6125 
6126 /**
6127  * e1000e_hwtstamp_ioctl - control hardware time stamping
6128  * @netdev: network interface device structure
6129  * @ifreq: interface request
6130  *
6131  * Outgoing time stamping can be enabled and disabled. Play nice and
6132  * disable it when requested, although it shouldn't cause any overhead
6133  * when no packet needs it. At most one packet in the queue may be
6134  * marked for time stamping, otherwise it would be impossible to tell
6135  * for sure to which packet the hardware time stamp belongs.
6136  *
6137  * Incoming time stamping has to be configured via the hardware filters.
6138  * Not all combinations are supported, in particular event type has to be
6139  * specified. Matching the kind of event packet is not supported, with the
6140  * exception of "all V2 events regardless of level 2 or 4".
6141  **/
6142 static int e1000e_hwtstamp_set(struct net_device *netdev, struct ifreq *ifr)
6143 {
6144 	struct e1000_adapter *adapter = netdev_priv(netdev);
6145 	struct hwtstamp_config config;
6146 	int ret_val;
6147 
6148 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
6149 		return -EFAULT;
6150 
6151 	ret_val = e1000e_config_hwtstamp(adapter, &config);
6152 	if (ret_val)
6153 		return ret_val;
6154 
6155 	switch (config.rx_filter) {
6156 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
6157 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
6158 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
6159 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
6160 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
6161 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
6162 		/* With V2 type filters which specify a Sync or Delay Request,
6163 		 * Path Delay Request/Response messages are also time stamped
6164 		 * by hardware so notify the caller the requested packets plus
6165 		 * some others are time stamped.
6166 		 */
6167 		config.rx_filter = HWTSTAMP_FILTER_SOME;
6168 		break;
6169 	default:
6170 		break;
6171 	}
6172 
6173 	return copy_to_user(ifr->ifr_data, &config,
6174 			    sizeof(config)) ? -EFAULT : 0;
6175 }
6176 
6177 static int e1000e_hwtstamp_get(struct net_device *netdev, struct ifreq *ifr)
6178 {
6179 	struct e1000_adapter *adapter = netdev_priv(netdev);
6180 
6181 	return copy_to_user(ifr->ifr_data, &adapter->hwtstamp_config,
6182 			    sizeof(adapter->hwtstamp_config)) ? -EFAULT : 0;
6183 }
6184 
6185 static int e1000_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
6186 {
6187 	switch (cmd) {
6188 	case SIOCGMIIPHY:
6189 	case SIOCGMIIREG:
6190 	case SIOCSMIIREG:
6191 		return e1000_mii_ioctl(netdev, ifr, cmd);
6192 	case SIOCSHWTSTAMP:
6193 		return e1000e_hwtstamp_set(netdev, ifr);
6194 	case SIOCGHWTSTAMP:
6195 		return e1000e_hwtstamp_get(netdev, ifr);
6196 	default:
6197 		return -EOPNOTSUPP;
6198 	}
6199 }
6200 
6201 static int e1000_init_phy_wakeup(struct e1000_adapter *adapter, u32 wufc)
6202 {
6203 	struct e1000_hw *hw = &adapter->hw;
6204 	u32 i, mac_reg, wuc;
6205 	u16 phy_reg, wuc_enable;
6206 	int retval;
6207 
6208 	/* copy MAC RARs to PHY RARs */
6209 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
6210 
6211 	retval = hw->phy.ops.acquire(hw);
6212 	if (retval) {
6213 		e_err("Could not acquire PHY\n");
6214 		return retval;
6215 	}
6216 
6217 	/* Enable access to wakeup registers on and set page to BM_WUC_PAGE */
6218 	retval = e1000_enable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6219 	if (retval)
6220 		goto release;
6221 
6222 	/* copy MAC MTA to PHY MTA - only needed for pchlan */
6223 	for (i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
6224 		mac_reg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
6225 		hw->phy.ops.write_reg_page(hw, BM_MTA(i),
6226 					   (u16)(mac_reg & 0xFFFF));
6227 		hw->phy.ops.write_reg_page(hw, BM_MTA(i) + 1,
6228 					   (u16)((mac_reg >> 16) & 0xFFFF));
6229 	}
6230 
6231 	/* configure PHY Rx Control register */
6232 	hw->phy.ops.read_reg_page(&adapter->hw, BM_RCTL, &phy_reg);
6233 	mac_reg = er32(RCTL);
6234 	if (mac_reg & E1000_RCTL_UPE)
6235 		phy_reg |= BM_RCTL_UPE;
6236 	if (mac_reg & E1000_RCTL_MPE)
6237 		phy_reg |= BM_RCTL_MPE;
6238 	phy_reg &= ~(BM_RCTL_MO_MASK);
6239 	if (mac_reg & E1000_RCTL_MO_3)
6240 		phy_reg |= (((mac_reg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
6241 			    << BM_RCTL_MO_SHIFT);
6242 	if (mac_reg & E1000_RCTL_BAM)
6243 		phy_reg |= BM_RCTL_BAM;
6244 	if (mac_reg & E1000_RCTL_PMCF)
6245 		phy_reg |= BM_RCTL_PMCF;
6246 	mac_reg = er32(CTRL);
6247 	if (mac_reg & E1000_CTRL_RFCE)
6248 		phy_reg |= BM_RCTL_RFCE;
6249 	hw->phy.ops.write_reg_page(&adapter->hw, BM_RCTL, phy_reg);
6250 
6251 	wuc = E1000_WUC_PME_EN;
6252 	if (wufc & (E1000_WUFC_MAG | E1000_WUFC_LNKC))
6253 		wuc |= E1000_WUC_APME;
6254 
6255 	/* enable PHY wakeup in MAC register */
6256 	ew32(WUFC, wufc);
6257 	ew32(WUC, (E1000_WUC_PHY_WAKE | E1000_WUC_APMPME |
6258 		   E1000_WUC_PME_STATUS | wuc));
6259 
6260 	/* configure and enable PHY wakeup in PHY registers */
6261 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUFC, wufc);
6262 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUC, wuc);
6263 
6264 	/* activate PHY wakeup */
6265 	wuc_enable |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
6266 	retval = e1000_disable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6267 	if (retval)
6268 		e_err("Could not set PHY Host Wakeup bit\n");
6269 release:
6270 	hw->phy.ops.release(hw);
6271 
6272 	return retval;
6273 }
6274 
6275 static void e1000e_flush_lpic(struct pci_dev *pdev)
6276 {
6277 	struct net_device *netdev = pci_get_drvdata(pdev);
6278 	struct e1000_adapter *adapter = netdev_priv(netdev);
6279 	struct e1000_hw *hw = &adapter->hw;
6280 	u32 ret_val;
6281 
6282 	pm_runtime_get_sync(netdev->dev.parent);
6283 
6284 	ret_val = hw->phy.ops.acquire(hw);
6285 	if (ret_val)
6286 		goto fl_out;
6287 
6288 	pr_info("EEE TX LPI TIMER: %08X\n",
6289 		er32(LPIC) >> E1000_LPIC_LPIET_SHIFT);
6290 
6291 	hw->phy.ops.release(hw);
6292 
6293 fl_out:
6294 	pm_runtime_put_sync(netdev->dev.parent);
6295 }
6296 
6297 static int e1000e_pm_freeze(struct device *dev)
6298 {
6299 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6300 	struct e1000_adapter *adapter = netdev_priv(netdev);
6301 
6302 	netif_device_detach(netdev);
6303 
6304 	if (netif_running(netdev)) {
6305 		int count = E1000_CHECK_RESET_COUNT;
6306 
6307 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6308 			usleep_range(10000, 20000);
6309 
6310 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6311 
6312 		/* Quiesce the device without resetting the hardware */
6313 		e1000e_down(adapter, false);
6314 		e1000_free_irq(adapter);
6315 	}
6316 	e1000e_reset_interrupt_capability(adapter);
6317 
6318 	/* Allow time for pending master requests to run */
6319 	e1000e_disable_pcie_master(&adapter->hw);
6320 
6321 	return 0;
6322 }
6323 
6324 static int __e1000_shutdown(struct pci_dev *pdev, bool runtime)
6325 {
6326 	struct net_device *netdev = pci_get_drvdata(pdev);
6327 	struct e1000_adapter *adapter = netdev_priv(netdev);
6328 	struct e1000_hw *hw = &adapter->hw;
6329 	u32 ctrl, ctrl_ext, rctl, status;
6330 	/* Runtime suspend should only enable wakeup for link changes */
6331 	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
6332 	int retval = 0;
6333 
6334 	status = er32(STATUS);
6335 	if (status & E1000_STATUS_LU)
6336 		wufc &= ~E1000_WUFC_LNKC;
6337 
6338 	if (wufc) {
6339 		e1000_setup_rctl(adapter);
6340 		e1000e_set_rx_mode(netdev);
6341 
6342 		/* turn on all-multi mode if wake on multicast is enabled */
6343 		if (wufc & E1000_WUFC_MC) {
6344 			rctl = er32(RCTL);
6345 			rctl |= E1000_RCTL_MPE;
6346 			ew32(RCTL, rctl);
6347 		}
6348 
6349 		ctrl = er32(CTRL);
6350 		ctrl |= E1000_CTRL_ADVD3WUC;
6351 		if (!(adapter->flags2 & FLAG2_HAS_PHY_WAKEUP))
6352 			ctrl |= E1000_CTRL_EN_PHY_PWR_MGMT;
6353 		ew32(CTRL, ctrl);
6354 
6355 		if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
6356 		    adapter->hw.phy.media_type ==
6357 		    e1000_media_type_internal_serdes) {
6358 			/* keep the laser running in D3 */
6359 			ctrl_ext = er32(CTRL_EXT);
6360 			ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
6361 			ew32(CTRL_EXT, ctrl_ext);
6362 		}
6363 
6364 		if (!runtime)
6365 			e1000e_power_up_phy(adapter);
6366 
6367 		if (adapter->flags & FLAG_IS_ICH)
6368 			e1000_suspend_workarounds_ich8lan(&adapter->hw);
6369 
6370 		if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6371 			/* enable wakeup by the PHY */
6372 			retval = e1000_init_phy_wakeup(adapter, wufc);
6373 			if (retval)
6374 				return retval;
6375 		} else {
6376 			/* enable wakeup by the MAC */
6377 			ew32(WUFC, wufc);
6378 			ew32(WUC, E1000_WUC_PME_EN);
6379 		}
6380 	} else {
6381 		ew32(WUC, 0);
6382 		ew32(WUFC, 0);
6383 
6384 		e1000_power_down_phy(adapter);
6385 	}
6386 
6387 	if (adapter->hw.phy.type == e1000_phy_igp_3) {
6388 		e1000e_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
6389 	} else if (hw->mac.type >= e1000_pch_lpt) {
6390 		if (!(wufc & (E1000_WUFC_EX | E1000_WUFC_MC | E1000_WUFC_BC)))
6391 			/* ULP does not support wake from unicast, multicast
6392 			 * or broadcast.
6393 			 */
6394 			retval = e1000_enable_ulp_lpt_lp(hw, !runtime);
6395 
6396 		if (retval)
6397 			return retval;
6398 	}
6399 
6400 	/* Ensure that the appropriate bits are set in LPI_CTRL
6401 	 * for EEE in Sx
6402 	 */
6403 	if ((hw->phy.type >= e1000_phy_i217) &&
6404 	    adapter->eee_advert && hw->dev_spec.ich8lan.eee_lp_ability) {
6405 		u16 lpi_ctrl = 0;
6406 
6407 		retval = hw->phy.ops.acquire(hw);
6408 		if (!retval) {
6409 			retval = e1e_rphy_locked(hw, I82579_LPI_CTRL,
6410 						 &lpi_ctrl);
6411 			if (!retval) {
6412 				if (adapter->eee_advert &
6413 				    hw->dev_spec.ich8lan.eee_lp_ability &
6414 				    I82579_EEE_100_SUPPORTED)
6415 					lpi_ctrl |= I82579_LPI_CTRL_100_ENABLE;
6416 				if (adapter->eee_advert &
6417 				    hw->dev_spec.ich8lan.eee_lp_ability &
6418 				    I82579_EEE_1000_SUPPORTED)
6419 					lpi_ctrl |= I82579_LPI_CTRL_1000_ENABLE;
6420 
6421 				retval = e1e_wphy_locked(hw, I82579_LPI_CTRL,
6422 							 lpi_ctrl);
6423 			}
6424 		}
6425 		hw->phy.ops.release(hw);
6426 	}
6427 
6428 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
6429 	 * would have already happened in close and is redundant.
6430 	 */
6431 	e1000e_release_hw_control(adapter);
6432 
6433 	pci_clear_master(pdev);
6434 
6435 	/* The pci-e switch on some quad port adapters will report a
6436 	 * correctable error when the MAC transitions from D0 to D3.  To
6437 	 * prevent this we need to mask off the correctable errors on the
6438 	 * downstream port of the pci-e switch.
6439 	 *
6440 	 * We don't have the associated upstream bridge while assigning
6441 	 * the PCI device into guest. For example, the KVM on power is
6442 	 * one of the cases.
6443 	 */
6444 	if (adapter->flags & FLAG_IS_QUAD_PORT) {
6445 		struct pci_dev *us_dev = pdev->bus->self;
6446 		u16 devctl;
6447 
6448 		if (!us_dev)
6449 			return 0;
6450 
6451 		pcie_capability_read_word(us_dev, PCI_EXP_DEVCTL, &devctl);
6452 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL,
6453 					   (devctl & ~PCI_EXP_DEVCTL_CERE));
6454 
6455 		pci_save_state(pdev);
6456 		pci_prepare_to_sleep(pdev);
6457 
6458 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL, devctl);
6459 	}
6460 
6461 	return 0;
6462 }
6463 
6464 /**
6465  * __e1000e_disable_aspm - Disable ASPM states
6466  * @pdev: pointer to PCI device struct
6467  * @state: bit-mask of ASPM states to disable
6468  * @locked: indication if this context holds pci_bus_sem locked.
6469  *
6470  * Some devices *must* have certain ASPM states disabled per hardware errata.
6471  **/
6472 static void __e1000e_disable_aspm(struct pci_dev *pdev, u16 state, int locked)
6473 {
6474 	struct pci_dev *parent = pdev->bus->self;
6475 	u16 aspm_dis_mask = 0;
6476 	u16 pdev_aspmc, parent_aspmc;
6477 
6478 	switch (state) {
6479 	case PCIE_LINK_STATE_L0S:
6480 	case PCIE_LINK_STATE_L0S | PCIE_LINK_STATE_L1:
6481 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L0S;
6482 		/* fall-through - can't have L1 without L0s */
6483 	case PCIE_LINK_STATE_L1:
6484 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L1;
6485 		break;
6486 	default:
6487 		return;
6488 	}
6489 
6490 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6491 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6492 
6493 	if (parent) {
6494 		pcie_capability_read_word(parent, PCI_EXP_LNKCTL,
6495 					  &parent_aspmc);
6496 		parent_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6497 	}
6498 
6499 	/* Nothing to do if the ASPM states to be disabled already are */
6500 	if (!(pdev_aspmc & aspm_dis_mask) &&
6501 	    (!parent || !(parent_aspmc & aspm_dis_mask)))
6502 		return;
6503 
6504 	dev_info(&pdev->dev, "Disabling ASPM %s %s\n",
6505 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L0S) ?
6506 		 "L0s" : "",
6507 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L1) ?
6508 		 "L1" : "");
6509 
6510 #ifdef CONFIG_PCIEASPM
6511 	if (locked)
6512 		pci_disable_link_state_locked(pdev, state);
6513 	else
6514 		pci_disable_link_state(pdev, state);
6515 
6516 	/* Double-check ASPM control.  If not disabled by the above, the
6517 	 * BIOS is preventing that from happening (or CONFIG_PCIEASPM is
6518 	 * not enabled); override by writing PCI config space directly.
6519 	 */
6520 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6521 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6522 
6523 	if (!(aspm_dis_mask & pdev_aspmc))
6524 		return;
6525 #endif
6526 
6527 	/* Both device and parent should have the same ASPM setting.
6528 	 * Disable ASPM in downstream component first and then upstream.
6529 	 */
6530 	pcie_capability_clear_word(pdev, PCI_EXP_LNKCTL, aspm_dis_mask);
6531 
6532 	if (parent)
6533 		pcie_capability_clear_word(parent, PCI_EXP_LNKCTL,
6534 					   aspm_dis_mask);
6535 }
6536 
6537 /**
6538  * e1000e_disable_aspm - Disable ASPM states.
6539  * @pdev: pointer to PCI device struct
6540  * @state: bit-mask of ASPM states to disable
6541  *
6542  * This function acquires the pci_bus_sem!
6543  * Some devices *must* have certain ASPM states disabled per hardware errata.
6544  **/
6545 static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6546 {
6547 	__e1000e_disable_aspm(pdev, state, 0);
6548 }
6549 
6550 /**
6551  * e1000e_disable_aspm_locked   Disable ASPM states.
6552  * @pdev: pointer to PCI device struct
6553  * @state: bit-mask of ASPM states to disable
6554  *
6555  * This function must be called with pci_bus_sem acquired!
6556  * Some devices *must* have certain ASPM states disabled per hardware errata.
6557  **/
6558 static void e1000e_disable_aspm_locked(struct pci_dev *pdev, u16 state)
6559 {
6560 	__e1000e_disable_aspm(pdev, state, 1);
6561 }
6562 
6563 #ifdef CONFIG_PM
6564 static int __e1000_resume(struct pci_dev *pdev)
6565 {
6566 	struct net_device *netdev = pci_get_drvdata(pdev);
6567 	struct e1000_adapter *adapter = netdev_priv(netdev);
6568 	struct e1000_hw *hw = &adapter->hw;
6569 	u16 aspm_disable_flag = 0;
6570 
6571 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6572 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6573 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6574 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6575 	if (aspm_disable_flag)
6576 		e1000e_disable_aspm(pdev, aspm_disable_flag);
6577 
6578 	pci_set_master(pdev);
6579 
6580 	if (hw->mac.type >= e1000_pch2lan)
6581 		e1000_resume_workarounds_pchlan(&adapter->hw);
6582 
6583 	e1000e_power_up_phy(adapter);
6584 
6585 	/* report the system wakeup cause from S3/S4 */
6586 	if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6587 		u16 phy_data;
6588 
6589 		e1e_rphy(&adapter->hw, BM_WUS, &phy_data);
6590 		if (phy_data) {
6591 			e_info("PHY Wakeup cause - %s\n",
6592 			       phy_data & E1000_WUS_EX ? "Unicast Packet" :
6593 			       phy_data & E1000_WUS_MC ? "Multicast Packet" :
6594 			       phy_data & E1000_WUS_BC ? "Broadcast Packet" :
6595 			       phy_data & E1000_WUS_MAG ? "Magic Packet" :
6596 			       phy_data & E1000_WUS_LNKC ?
6597 			       "Link Status Change" : "other");
6598 		}
6599 		e1e_wphy(&adapter->hw, BM_WUS, ~0);
6600 	} else {
6601 		u32 wus = er32(WUS);
6602 
6603 		if (wus) {
6604 			e_info("MAC Wakeup cause - %s\n",
6605 			       wus & E1000_WUS_EX ? "Unicast Packet" :
6606 			       wus & E1000_WUS_MC ? "Multicast Packet" :
6607 			       wus & E1000_WUS_BC ? "Broadcast Packet" :
6608 			       wus & E1000_WUS_MAG ? "Magic Packet" :
6609 			       wus & E1000_WUS_LNKC ? "Link Status Change" :
6610 			       "other");
6611 		}
6612 		ew32(WUS, ~0);
6613 	}
6614 
6615 	e1000e_reset(adapter);
6616 
6617 	e1000_init_manageability_pt(adapter);
6618 
6619 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6620 	 * is up.  For all other cases, let the f/w know that the h/w is now
6621 	 * under the control of the driver.
6622 	 */
6623 	if (!(adapter->flags & FLAG_HAS_AMT))
6624 		e1000e_get_hw_control(adapter);
6625 
6626 	return 0;
6627 }
6628 
6629 #ifdef CONFIG_PM_SLEEP
6630 static int e1000e_pm_thaw(struct device *dev)
6631 {
6632 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6633 	struct e1000_adapter *adapter = netdev_priv(netdev);
6634 
6635 	e1000e_set_interrupt_capability(adapter);
6636 	if (netif_running(netdev)) {
6637 		u32 err = e1000_request_irq(adapter);
6638 
6639 		if (err)
6640 			return err;
6641 
6642 		e1000e_up(adapter);
6643 	}
6644 
6645 	netif_device_attach(netdev);
6646 
6647 	return 0;
6648 }
6649 
6650 static int e1000e_pm_suspend(struct device *dev)
6651 {
6652 	struct pci_dev *pdev = to_pci_dev(dev);
6653 	int rc;
6654 
6655 	e1000e_flush_lpic(pdev);
6656 
6657 	e1000e_pm_freeze(dev);
6658 
6659 	rc = __e1000_shutdown(pdev, false);
6660 	if (rc)
6661 		e1000e_pm_thaw(dev);
6662 
6663 	return rc;
6664 }
6665 
6666 static int e1000e_pm_resume(struct device *dev)
6667 {
6668 	struct pci_dev *pdev = to_pci_dev(dev);
6669 	int rc;
6670 
6671 	rc = __e1000_resume(pdev);
6672 	if (rc)
6673 		return rc;
6674 
6675 	return e1000e_pm_thaw(dev);
6676 }
6677 #endif /* CONFIG_PM_SLEEP */
6678 
6679 static int e1000e_pm_runtime_idle(struct device *dev)
6680 {
6681 	struct pci_dev *pdev = to_pci_dev(dev);
6682 	struct net_device *netdev = pci_get_drvdata(pdev);
6683 	struct e1000_adapter *adapter = netdev_priv(netdev);
6684 	u16 eee_lp;
6685 
6686 	eee_lp = adapter->hw.dev_spec.ich8lan.eee_lp_ability;
6687 
6688 	if (!e1000e_has_link(adapter)) {
6689 		adapter->hw.dev_spec.ich8lan.eee_lp_ability = eee_lp;
6690 		pm_schedule_suspend(dev, 5 * MSEC_PER_SEC);
6691 	}
6692 
6693 	return -EBUSY;
6694 }
6695 
6696 static int e1000e_pm_runtime_resume(struct device *dev)
6697 {
6698 	struct pci_dev *pdev = to_pci_dev(dev);
6699 	struct net_device *netdev = pci_get_drvdata(pdev);
6700 	struct e1000_adapter *adapter = netdev_priv(netdev);
6701 	int rc;
6702 
6703 	rc = __e1000_resume(pdev);
6704 	if (rc)
6705 		return rc;
6706 
6707 	if (netdev->flags & IFF_UP)
6708 		e1000e_up(adapter);
6709 
6710 	return rc;
6711 }
6712 
6713 static int e1000e_pm_runtime_suspend(struct device *dev)
6714 {
6715 	struct pci_dev *pdev = to_pci_dev(dev);
6716 	struct net_device *netdev = pci_get_drvdata(pdev);
6717 	struct e1000_adapter *adapter = netdev_priv(netdev);
6718 
6719 	if (netdev->flags & IFF_UP) {
6720 		int count = E1000_CHECK_RESET_COUNT;
6721 
6722 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6723 			usleep_range(10000, 20000);
6724 
6725 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6726 
6727 		/* Down the device without resetting the hardware */
6728 		e1000e_down(adapter, false);
6729 	}
6730 
6731 	if (__e1000_shutdown(pdev, true)) {
6732 		e1000e_pm_runtime_resume(dev);
6733 		return -EBUSY;
6734 	}
6735 
6736 	return 0;
6737 }
6738 #endif /* CONFIG_PM */
6739 
6740 static void e1000_shutdown(struct pci_dev *pdev)
6741 {
6742 	e1000e_flush_lpic(pdev);
6743 
6744 	e1000e_pm_freeze(&pdev->dev);
6745 
6746 	__e1000_shutdown(pdev, false);
6747 }
6748 
6749 #ifdef CONFIG_NET_POLL_CONTROLLER
6750 
6751 static irqreturn_t e1000_intr_msix(int __always_unused irq, void *data)
6752 {
6753 	struct net_device *netdev = data;
6754 	struct e1000_adapter *adapter = netdev_priv(netdev);
6755 
6756 	if (adapter->msix_entries) {
6757 		int vector, msix_irq;
6758 
6759 		vector = 0;
6760 		msix_irq = adapter->msix_entries[vector].vector;
6761 		if (disable_hardirq(msix_irq))
6762 			e1000_intr_msix_rx(msix_irq, netdev);
6763 		enable_irq(msix_irq);
6764 
6765 		vector++;
6766 		msix_irq = adapter->msix_entries[vector].vector;
6767 		if (disable_hardirq(msix_irq))
6768 			e1000_intr_msix_tx(msix_irq, netdev);
6769 		enable_irq(msix_irq);
6770 
6771 		vector++;
6772 		msix_irq = adapter->msix_entries[vector].vector;
6773 		if (disable_hardirq(msix_irq))
6774 			e1000_msix_other(msix_irq, netdev);
6775 		enable_irq(msix_irq);
6776 	}
6777 
6778 	return IRQ_HANDLED;
6779 }
6780 
6781 /**
6782  * e1000_netpoll
6783  * @netdev: network interface device structure
6784  *
6785  * Polling 'interrupt' - used by things like netconsole to send skbs
6786  * without having to re-enable interrupts. It's not called while
6787  * the interrupt routine is executing.
6788  */
6789 static void e1000_netpoll(struct net_device *netdev)
6790 {
6791 	struct e1000_adapter *adapter = netdev_priv(netdev);
6792 
6793 	switch (adapter->int_mode) {
6794 	case E1000E_INT_MODE_MSIX:
6795 		e1000_intr_msix(adapter->pdev->irq, netdev);
6796 		break;
6797 	case E1000E_INT_MODE_MSI:
6798 		if (disable_hardirq(adapter->pdev->irq))
6799 			e1000_intr_msi(adapter->pdev->irq, netdev);
6800 		enable_irq(adapter->pdev->irq);
6801 		break;
6802 	default:		/* E1000E_INT_MODE_LEGACY */
6803 		if (disable_hardirq(adapter->pdev->irq))
6804 			e1000_intr(adapter->pdev->irq, netdev);
6805 		enable_irq(adapter->pdev->irq);
6806 		break;
6807 	}
6808 }
6809 #endif
6810 
6811 /**
6812  * e1000_io_error_detected - called when PCI error is detected
6813  * @pdev: Pointer to PCI device
6814  * @state: The current pci connection state
6815  *
6816  * This function is called after a PCI bus error affecting
6817  * this device has been detected.
6818  */
6819 static pci_ers_result_t e1000_io_error_detected(struct pci_dev *pdev,
6820 						pci_channel_state_t state)
6821 {
6822 	struct net_device *netdev = pci_get_drvdata(pdev);
6823 	struct e1000_adapter *adapter = netdev_priv(netdev);
6824 
6825 	netif_device_detach(netdev);
6826 
6827 	if (state == pci_channel_io_perm_failure)
6828 		return PCI_ERS_RESULT_DISCONNECT;
6829 
6830 	if (netif_running(netdev))
6831 		e1000e_down(adapter, true);
6832 	pci_disable_device(pdev);
6833 
6834 	/* Request a slot slot reset. */
6835 	return PCI_ERS_RESULT_NEED_RESET;
6836 }
6837 
6838 /**
6839  * e1000_io_slot_reset - called after the pci bus has been reset.
6840  * @pdev: Pointer to PCI device
6841  *
6842  * Restart the card from scratch, as if from a cold-boot. Implementation
6843  * resembles the first-half of the e1000e_pm_resume routine.
6844  */
6845 static pci_ers_result_t e1000_io_slot_reset(struct pci_dev *pdev)
6846 {
6847 	struct net_device *netdev = pci_get_drvdata(pdev);
6848 	struct e1000_adapter *adapter = netdev_priv(netdev);
6849 	struct e1000_hw *hw = &adapter->hw;
6850 	u16 aspm_disable_flag = 0;
6851 	int err;
6852 	pci_ers_result_t result;
6853 
6854 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6855 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6856 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6857 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6858 	if (aspm_disable_flag)
6859 		e1000e_disable_aspm_locked(pdev, aspm_disable_flag);
6860 
6861 	err = pci_enable_device_mem(pdev);
6862 	if (err) {
6863 		dev_err(&pdev->dev,
6864 			"Cannot re-enable PCI device after reset.\n");
6865 		result = PCI_ERS_RESULT_DISCONNECT;
6866 	} else {
6867 		pdev->state_saved = true;
6868 		pci_restore_state(pdev);
6869 		pci_set_master(pdev);
6870 
6871 		pci_enable_wake(pdev, PCI_D3hot, 0);
6872 		pci_enable_wake(pdev, PCI_D3cold, 0);
6873 
6874 		e1000e_reset(adapter);
6875 		ew32(WUS, ~0);
6876 		result = PCI_ERS_RESULT_RECOVERED;
6877 	}
6878 
6879 	pci_cleanup_aer_uncorrect_error_status(pdev);
6880 
6881 	return result;
6882 }
6883 
6884 /**
6885  * e1000_io_resume - called when traffic can start flowing again.
6886  * @pdev: Pointer to PCI device
6887  *
6888  * This callback is called when the error recovery driver tells us that
6889  * its OK to resume normal operation. Implementation resembles the
6890  * second-half of the e1000e_pm_resume routine.
6891  */
6892 static void e1000_io_resume(struct pci_dev *pdev)
6893 {
6894 	struct net_device *netdev = pci_get_drvdata(pdev);
6895 	struct e1000_adapter *adapter = netdev_priv(netdev);
6896 
6897 	e1000_init_manageability_pt(adapter);
6898 
6899 	if (netif_running(netdev))
6900 		e1000e_up(adapter);
6901 
6902 	netif_device_attach(netdev);
6903 
6904 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6905 	 * is up.  For all other cases, let the f/w know that the h/w is now
6906 	 * under the control of the driver.
6907 	 */
6908 	if (!(adapter->flags & FLAG_HAS_AMT))
6909 		e1000e_get_hw_control(adapter);
6910 }
6911 
6912 static void e1000_print_device_info(struct e1000_adapter *adapter)
6913 {
6914 	struct e1000_hw *hw = &adapter->hw;
6915 	struct net_device *netdev = adapter->netdev;
6916 	u32 ret_val;
6917 	u8 pba_str[E1000_PBANUM_LENGTH];
6918 
6919 	/* print bus type/speed/width info */
6920 	e_info("(PCI Express:2.5GT/s:%s) %pM\n",
6921 	       /* bus width */
6922 	       ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
6923 		"Width x1"),
6924 	       /* MAC address */
6925 	       netdev->dev_addr);
6926 	e_info("Intel(R) PRO/%s Network Connection\n",
6927 	       (hw->phy.type == e1000_phy_ife) ? "10/100" : "1000");
6928 	ret_val = e1000_read_pba_string_generic(hw, pba_str,
6929 						E1000_PBANUM_LENGTH);
6930 	if (ret_val)
6931 		strlcpy((char *)pba_str, "Unknown", sizeof(pba_str));
6932 	e_info("MAC: %d, PHY: %d, PBA No: %s\n",
6933 	       hw->mac.type, hw->phy.type, pba_str);
6934 }
6935 
6936 static void e1000_eeprom_checks(struct e1000_adapter *adapter)
6937 {
6938 	struct e1000_hw *hw = &adapter->hw;
6939 	int ret_val;
6940 	u16 buf = 0;
6941 
6942 	if (hw->mac.type != e1000_82573)
6943 		return;
6944 
6945 	ret_val = e1000_read_nvm(hw, NVM_INIT_CONTROL2_REG, 1, &buf);
6946 	le16_to_cpus(&buf);
6947 	if (!ret_val && (!(buf & BIT(0)))) {
6948 		/* Deep Smart Power Down (DSPD) */
6949 		dev_warn(&adapter->pdev->dev,
6950 			 "Warning: detected DSPD enabled in EEPROM\n");
6951 	}
6952 }
6953 
6954 static netdev_features_t e1000_fix_features(struct net_device *netdev,
6955 					    netdev_features_t features)
6956 {
6957 	struct e1000_adapter *adapter = netdev_priv(netdev);
6958 	struct e1000_hw *hw = &adapter->hw;
6959 
6960 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
6961 	if ((hw->mac.type >= e1000_pch2lan) && (netdev->mtu > ETH_DATA_LEN))
6962 		features &= ~NETIF_F_RXFCS;
6963 
6964 	/* Since there is no support for separate Rx/Tx vlan accel
6965 	 * enable/disable make sure Tx flag is always in same state as Rx.
6966 	 */
6967 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
6968 		features |= NETIF_F_HW_VLAN_CTAG_TX;
6969 	else
6970 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
6971 
6972 	return features;
6973 }
6974 
6975 static int e1000_set_features(struct net_device *netdev,
6976 			      netdev_features_t features)
6977 {
6978 	struct e1000_adapter *adapter = netdev_priv(netdev);
6979 	netdev_features_t changed = features ^ netdev->features;
6980 
6981 	if (changed & (NETIF_F_TSO | NETIF_F_TSO6))
6982 		adapter->flags |= FLAG_TSO_FORCE;
6983 
6984 	if (!(changed & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX |
6985 			 NETIF_F_RXCSUM | NETIF_F_RXHASH | NETIF_F_RXFCS |
6986 			 NETIF_F_RXALL)))
6987 		return 0;
6988 
6989 	if (changed & NETIF_F_RXFCS) {
6990 		if (features & NETIF_F_RXFCS) {
6991 			adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
6992 		} else {
6993 			/* We need to take it back to defaults, which might mean
6994 			 * stripping is still disabled at the adapter level.
6995 			 */
6996 			if (adapter->flags2 & FLAG2_DFLT_CRC_STRIPPING)
6997 				adapter->flags2 |= FLAG2_CRC_STRIPPING;
6998 			else
6999 				adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7000 		}
7001 	}
7002 
7003 	netdev->features = features;
7004 
7005 	if (netif_running(netdev))
7006 		e1000e_reinit_locked(adapter);
7007 	else
7008 		e1000e_reset(adapter);
7009 
7010 	return 0;
7011 }
7012 
7013 static const struct net_device_ops e1000e_netdev_ops = {
7014 	.ndo_open		= e1000e_open,
7015 	.ndo_stop		= e1000e_close,
7016 	.ndo_start_xmit		= e1000_xmit_frame,
7017 	.ndo_get_stats64	= e1000e_get_stats64,
7018 	.ndo_set_rx_mode	= e1000e_set_rx_mode,
7019 	.ndo_set_mac_address	= e1000_set_mac,
7020 	.ndo_change_mtu		= e1000_change_mtu,
7021 	.ndo_do_ioctl		= e1000_ioctl,
7022 	.ndo_tx_timeout		= e1000_tx_timeout,
7023 	.ndo_validate_addr	= eth_validate_addr,
7024 
7025 	.ndo_vlan_rx_add_vid	= e1000_vlan_rx_add_vid,
7026 	.ndo_vlan_rx_kill_vid	= e1000_vlan_rx_kill_vid,
7027 #ifdef CONFIG_NET_POLL_CONTROLLER
7028 	.ndo_poll_controller	= e1000_netpoll,
7029 #endif
7030 	.ndo_set_features = e1000_set_features,
7031 	.ndo_fix_features = e1000_fix_features,
7032 	.ndo_features_check	= passthru_features_check,
7033 };
7034 
7035 /**
7036  * e1000_probe - Device Initialization Routine
7037  * @pdev: PCI device information struct
7038  * @ent: entry in e1000_pci_tbl
7039  *
7040  * Returns 0 on success, negative on failure
7041  *
7042  * e1000_probe initializes an adapter identified by a pci_dev structure.
7043  * The OS initialization, configuring of the adapter private structure,
7044  * and a hardware reset occur.
7045  **/
7046 static int e1000_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7047 {
7048 	struct net_device *netdev;
7049 	struct e1000_adapter *adapter;
7050 	struct e1000_hw *hw;
7051 	const struct e1000_info *ei = e1000_info_tbl[ent->driver_data];
7052 	resource_size_t mmio_start, mmio_len;
7053 	resource_size_t flash_start, flash_len;
7054 	static int cards_found;
7055 	u16 aspm_disable_flag = 0;
7056 	int bars, i, err, pci_using_dac;
7057 	u16 eeprom_data = 0;
7058 	u16 eeprom_apme_mask = E1000_EEPROM_APME;
7059 	s32 ret_val = 0;
7060 
7061 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L0S)
7062 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7063 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L1)
7064 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7065 	if (aspm_disable_flag)
7066 		e1000e_disable_aspm(pdev, aspm_disable_flag);
7067 
7068 	err = pci_enable_device_mem(pdev);
7069 	if (err)
7070 		return err;
7071 
7072 	pci_using_dac = 0;
7073 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
7074 	if (!err) {
7075 		pci_using_dac = 1;
7076 	} else {
7077 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
7078 		if (err) {
7079 			dev_err(&pdev->dev,
7080 				"No usable DMA configuration, aborting\n");
7081 			goto err_dma;
7082 		}
7083 	}
7084 
7085 	bars = pci_select_bars(pdev, IORESOURCE_MEM);
7086 	err = pci_request_selected_regions_exclusive(pdev, bars,
7087 						     e1000e_driver_name);
7088 	if (err)
7089 		goto err_pci_reg;
7090 
7091 	/* AER (Advanced Error Reporting) hooks */
7092 	pci_enable_pcie_error_reporting(pdev);
7093 
7094 	pci_set_master(pdev);
7095 	/* PCI config space info */
7096 	err = pci_save_state(pdev);
7097 	if (err)
7098 		goto err_alloc_etherdev;
7099 
7100 	err = -ENOMEM;
7101 	netdev = alloc_etherdev(sizeof(struct e1000_adapter));
7102 	if (!netdev)
7103 		goto err_alloc_etherdev;
7104 
7105 	SET_NETDEV_DEV(netdev, &pdev->dev);
7106 
7107 	netdev->irq = pdev->irq;
7108 
7109 	pci_set_drvdata(pdev, netdev);
7110 	adapter = netdev_priv(netdev);
7111 	hw = &adapter->hw;
7112 	adapter->netdev = netdev;
7113 	adapter->pdev = pdev;
7114 	adapter->ei = ei;
7115 	adapter->pba = ei->pba;
7116 	adapter->flags = ei->flags;
7117 	adapter->flags2 = ei->flags2;
7118 	adapter->hw.adapter = adapter;
7119 	adapter->hw.mac.type = ei->mac;
7120 	adapter->max_hw_frame_size = ei->max_hw_frame_size;
7121 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
7122 
7123 	mmio_start = pci_resource_start(pdev, 0);
7124 	mmio_len = pci_resource_len(pdev, 0);
7125 
7126 	err = -EIO;
7127 	adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
7128 	if (!adapter->hw.hw_addr)
7129 		goto err_ioremap;
7130 
7131 	if ((adapter->flags & FLAG_HAS_FLASH) &&
7132 	    (pci_resource_flags(pdev, 1) & IORESOURCE_MEM) &&
7133 	    (hw->mac.type < e1000_pch_spt)) {
7134 		flash_start = pci_resource_start(pdev, 1);
7135 		flash_len = pci_resource_len(pdev, 1);
7136 		adapter->hw.flash_address = ioremap(flash_start, flash_len);
7137 		if (!adapter->hw.flash_address)
7138 			goto err_flashmap;
7139 	}
7140 
7141 	/* Set default EEE advertisement */
7142 	if (adapter->flags2 & FLAG2_HAS_EEE)
7143 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
7144 
7145 	/* construct the net_device struct */
7146 	netdev->netdev_ops = &e1000e_netdev_ops;
7147 	e1000e_set_ethtool_ops(netdev);
7148 	netdev->watchdog_timeo = 5 * HZ;
7149 	netif_napi_add(netdev, &adapter->napi, e1000e_poll, 64);
7150 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
7151 
7152 	netdev->mem_start = mmio_start;
7153 	netdev->mem_end = mmio_start + mmio_len;
7154 
7155 	adapter->bd_number = cards_found++;
7156 
7157 	e1000e_check_options(adapter);
7158 
7159 	/* setup adapter struct */
7160 	err = e1000_sw_init(adapter);
7161 	if (err)
7162 		goto err_sw_init;
7163 
7164 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
7165 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
7166 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
7167 
7168 	err = ei->get_variants(adapter);
7169 	if (err)
7170 		goto err_hw_init;
7171 
7172 	if ((adapter->flags & FLAG_IS_ICH) &&
7173 	    (adapter->flags & FLAG_READ_ONLY_NVM) &&
7174 	    (hw->mac.type < e1000_pch_spt))
7175 		e1000e_write_protect_nvm_ich8lan(&adapter->hw);
7176 
7177 	hw->mac.ops.get_bus_info(&adapter->hw);
7178 
7179 	adapter->hw.phy.autoneg_wait_to_complete = 0;
7180 
7181 	/* Copper options */
7182 	if (adapter->hw.phy.media_type == e1000_media_type_copper) {
7183 		adapter->hw.phy.mdix = AUTO_ALL_MODES;
7184 		adapter->hw.phy.disable_polarity_correction = 0;
7185 		adapter->hw.phy.ms_type = e1000_ms_hw_default;
7186 	}
7187 
7188 	if (hw->phy.ops.check_reset_block && hw->phy.ops.check_reset_block(hw))
7189 		dev_info(&pdev->dev,
7190 			 "PHY reset is blocked due to SOL/IDER session.\n");
7191 
7192 	/* Set initial default active device features */
7193 	netdev->features = (NETIF_F_SG |
7194 			    NETIF_F_HW_VLAN_CTAG_RX |
7195 			    NETIF_F_HW_VLAN_CTAG_TX |
7196 			    NETIF_F_TSO |
7197 			    NETIF_F_TSO6 |
7198 			    NETIF_F_RXHASH |
7199 			    NETIF_F_RXCSUM |
7200 			    NETIF_F_HW_CSUM);
7201 
7202 	/* Set user-changeable features (subset of all device features) */
7203 	netdev->hw_features = netdev->features;
7204 	netdev->hw_features |= NETIF_F_RXFCS;
7205 	netdev->priv_flags |= IFF_SUPP_NOFCS;
7206 	netdev->hw_features |= NETIF_F_RXALL;
7207 
7208 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER)
7209 		netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
7210 
7211 	netdev->vlan_features |= (NETIF_F_SG |
7212 				  NETIF_F_TSO |
7213 				  NETIF_F_TSO6 |
7214 				  NETIF_F_HW_CSUM);
7215 
7216 	netdev->priv_flags |= IFF_UNICAST_FLT;
7217 
7218 	if (pci_using_dac) {
7219 		netdev->features |= NETIF_F_HIGHDMA;
7220 		netdev->vlan_features |= NETIF_F_HIGHDMA;
7221 	}
7222 
7223 	/* MTU range: 68 - max_hw_frame_size */
7224 	netdev->min_mtu = ETH_MIN_MTU;
7225 	netdev->max_mtu = adapter->max_hw_frame_size -
7226 			  (VLAN_ETH_HLEN + ETH_FCS_LEN);
7227 
7228 	if (e1000e_enable_mng_pass_thru(&adapter->hw))
7229 		adapter->flags |= FLAG_MNG_PT_ENABLED;
7230 
7231 	/* before reading the NVM, reset the controller to
7232 	 * put the device in a known good starting state
7233 	 */
7234 	adapter->hw.mac.ops.reset_hw(&adapter->hw);
7235 
7236 	/* systems with ASPM and others may see the checksum fail on the first
7237 	 * attempt. Let's give it a few tries
7238 	 */
7239 	for (i = 0;; i++) {
7240 		if (e1000_validate_nvm_checksum(&adapter->hw) >= 0)
7241 			break;
7242 		if (i == 2) {
7243 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
7244 			err = -EIO;
7245 			goto err_eeprom;
7246 		}
7247 	}
7248 
7249 	e1000_eeprom_checks(adapter);
7250 
7251 	/* copy the MAC address */
7252 	if (e1000e_read_mac_addr(&adapter->hw))
7253 		dev_err(&pdev->dev,
7254 			"NVM Read Error while reading MAC address\n");
7255 
7256 	memcpy(netdev->dev_addr, adapter->hw.mac.addr, netdev->addr_len);
7257 
7258 	if (!is_valid_ether_addr(netdev->dev_addr)) {
7259 		dev_err(&pdev->dev, "Invalid MAC Address: %pM\n",
7260 			netdev->dev_addr);
7261 		err = -EIO;
7262 		goto err_eeprom;
7263 	}
7264 
7265 	timer_setup(&adapter->watchdog_timer, e1000_watchdog, 0);
7266 	timer_setup(&adapter->phy_info_timer, e1000_update_phy_info, 0);
7267 
7268 	INIT_WORK(&adapter->reset_task, e1000_reset_task);
7269 	INIT_WORK(&adapter->watchdog_task, e1000_watchdog_task);
7270 	INIT_WORK(&adapter->downshift_task, e1000e_downshift_workaround);
7271 	INIT_WORK(&adapter->update_phy_task, e1000e_update_phy_task);
7272 	INIT_WORK(&adapter->print_hang_task, e1000_print_hw_hang);
7273 
7274 	/* Initialize link parameters. User can change them with ethtool */
7275 	adapter->hw.mac.autoneg = 1;
7276 	adapter->fc_autoneg = true;
7277 	adapter->hw.fc.requested_mode = e1000_fc_default;
7278 	adapter->hw.fc.current_mode = e1000_fc_default;
7279 	adapter->hw.phy.autoneg_advertised = 0x2f;
7280 
7281 	/* Initial Wake on LAN setting - If APM wake is enabled in
7282 	 * the EEPROM, enable the ACPI Magic Packet filter
7283 	 */
7284 	if (adapter->flags & FLAG_APME_IN_WUC) {
7285 		/* APME bit in EEPROM is mapped to WUC.APME */
7286 		eeprom_data = er32(WUC);
7287 		eeprom_apme_mask = E1000_WUC_APME;
7288 		if ((hw->mac.type > e1000_ich10lan) &&
7289 		    (eeprom_data & E1000_WUC_PHY_WAKE))
7290 			adapter->flags2 |= FLAG2_HAS_PHY_WAKEUP;
7291 	} else if (adapter->flags & FLAG_APME_IN_CTRL3) {
7292 		if (adapter->flags & FLAG_APME_CHECK_PORT_B &&
7293 		    (adapter->hw.bus.func == 1))
7294 			ret_val = e1000_read_nvm(&adapter->hw,
7295 					      NVM_INIT_CONTROL3_PORT_B,
7296 					      1, &eeprom_data);
7297 		else
7298 			ret_val = e1000_read_nvm(&adapter->hw,
7299 					      NVM_INIT_CONTROL3_PORT_A,
7300 					      1, &eeprom_data);
7301 	}
7302 
7303 	/* fetch WoL from EEPROM */
7304 	if (ret_val)
7305 		e_dbg("NVM read error getting WoL initial values: %d\n", ret_val);
7306 	else if (eeprom_data & eeprom_apme_mask)
7307 		adapter->eeprom_wol |= E1000_WUFC_MAG;
7308 
7309 	/* now that we have the eeprom settings, apply the special cases
7310 	 * where the eeprom may be wrong or the board simply won't support
7311 	 * wake on lan on a particular port
7312 	 */
7313 	if (!(adapter->flags & FLAG_HAS_WOL))
7314 		adapter->eeprom_wol = 0;
7315 
7316 	/* initialize the wol settings based on the eeprom settings */
7317 	adapter->wol = adapter->eeprom_wol;
7318 
7319 	/* make sure adapter isn't asleep if manageability is enabled */
7320 	if (adapter->wol || (adapter->flags & FLAG_MNG_PT_ENABLED) ||
7321 	    (hw->mac.ops.check_mng_mode(hw)))
7322 		device_wakeup_enable(&pdev->dev);
7323 
7324 	/* save off EEPROM version number */
7325 	ret_val = e1000_read_nvm(&adapter->hw, 5, 1, &adapter->eeprom_vers);
7326 
7327 	if (ret_val) {
7328 		e_dbg("NVM read error getting EEPROM version: %d\n", ret_val);
7329 		adapter->eeprom_vers = 0;
7330 	}
7331 
7332 	/* init PTP hardware clock */
7333 	e1000e_ptp_init(adapter);
7334 
7335 	/* reset the hardware with the new settings */
7336 	e1000e_reset(adapter);
7337 
7338 	/* If the controller has AMT, do not set DRV_LOAD until the interface
7339 	 * is up.  For all other cases, let the f/w know that the h/w is now
7340 	 * under the control of the driver.
7341 	 */
7342 	if (!(adapter->flags & FLAG_HAS_AMT))
7343 		e1000e_get_hw_control(adapter);
7344 
7345 	strlcpy(netdev->name, "eth%d", sizeof(netdev->name));
7346 	err = register_netdev(netdev);
7347 	if (err)
7348 		goto err_register;
7349 
7350 	/* carrier off reporting is important to ethtool even BEFORE open */
7351 	netif_carrier_off(netdev);
7352 
7353 	e1000_print_device_info(adapter);
7354 
7355 	if (pci_dev_run_wake(pdev))
7356 		pm_runtime_put_noidle(&pdev->dev);
7357 
7358 	return 0;
7359 
7360 err_register:
7361 	if (!(adapter->flags & FLAG_HAS_AMT))
7362 		e1000e_release_hw_control(adapter);
7363 err_eeprom:
7364 	if (hw->phy.ops.check_reset_block && !hw->phy.ops.check_reset_block(hw))
7365 		e1000_phy_hw_reset(&adapter->hw);
7366 err_hw_init:
7367 	kfree(adapter->tx_ring);
7368 	kfree(adapter->rx_ring);
7369 err_sw_init:
7370 	if ((adapter->hw.flash_address) && (hw->mac.type < e1000_pch_spt))
7371 		iounmap(adapter->hw.flash_address);
7372 	e1000e_reset_interrupt_capability(adapter);
7373 err_flashmap:
7374 	iounmap(adapter->hw.hw_addr);
7375 err_ioremap:
7376 	free_netdev(netdev);
7377 err_alloc_etherdev:
7378 	pci_release_mem_regions(pdev);
7379 err_pci_reg:
7380 err_dma:
7381 	pci_disable_device(pdev);
7382 	return err;
7383 }
7384 
7385 /**
7386  * e1000_remove - Device Removal Routine
7387  * @pdev: PCI device information struct
7388  *
7389  * e1000_remove is called by the PCI subsystem to alert the driver
7390  * that it should release a PCI device.  The could be caused by a
7391  * Hot-Plug event, or because the driver is going to be removed from
7392  * memory.
7393  **/
7394 static void e1000_remove(struct pci_dev *pdev)
7395 {
7396 	struct net_device *netdev = pci_get_drvdata(pdev);
7397 	struct e1000_adapter *adapter = netdev_priv(netdev);
7398 	bool down = test_bit(__E1000_DOWN, &adapter->state);
7399 
7400 	e1000e_ptp_remove(adapter);
7401 
7402 	/* The timers may be rescheduled, so explicitly disable them
7403 	 * from being rescheduled.
7404 	 */
7405 	if (!down)
7406 		set_bit(__E1000_DOWN, &adapter->state);
7407 	del_timer_sync(&adapter->watchdog_timer);
7408 	del_timer_sync(&adapter->phy_info_timer);
7409 
7410 	cancel_work_sync(&adapter->reset_task);
7411 	cancel_work_sync(&adapter->watchdog_task);
7412 	cancel_work_sync(&adapter->downshift_task);
7413 	cancel_work_sync(&adapter->update_phy_task);
7414 	cancel_work_sync(&adapter->print_hang_task);
7415 
7416 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
7417 		cancel_work_sync(&adapter->tx_hwtstamp_work);
7418 		if (adapter->tx_hwtstamp_skb) {
7419 			dev_consume_skb_any(adapter->tx_hwtstamp_skb);
7420 			adapter->tx_hwtstamp_skb = NULL;
7421 		}
7422 	}
7423 
7424 	/* Don't lie to e1000_close() down the road. */
7425 	if (!down)
7426 		clear_bit(__E1000_DOWN, &adapter->state);
7427 	unregister_netdev(netdev);
7428 
7429 	if (pci_dev_run_wake(pdev))
7430 		pm_runtime_get_noresume(&pdev->dev);
7431 
7432 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
7433 	 * would have already happened in close and is redundant.
7434 	 */
7435 	e1000e_release_hw_control(adapter);
7436 
7437 	e1000e_reset_interrupt_capability(adapter);
7438 	kfree(adapter->tx_ring);
7439 	kfree(adapter->rx_ring);
7440 
7441 	iounmap(adapter->hw.hw_addr);
7442 	if ((adapter->hw.flash_address) &&
7443 	    (adapter->hw.mac.type < e1000_pch_spt))
7444 		iounmap(adapter->hw.flash_address);
7445 	pci_release_mem_regions(pdev);
7446 
7447 	free_netdev(netdev);
7448 
7449 	/* AER disable */
7450 	pci_disable_pcie_error_reporting(pdev);
7451 
7452 	pci_disable_device(pdev);
7453 }
7454 
7455 /* PCI Error Recovery (ERS) */
7456 static const struct pci_error_handlers e1000_err_handler = {
7457 	.error_detected = e1000_io_error_detected,
7458 	.slot_reset = e1000_io_slot_reset,
7459 	.resume = e1000_io_resume,
7460 };
7461 
7462 static const struct pci_device_id e1000_pci_tbl[] = {
7463 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_COPPER), board_82571 },
7464 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_FIBER), board_82571 },
7465 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER), board_82571 },
7466 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER_LP),
7467 	  board_82571 },
7468 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_FIBER), board_82571 },
7469 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES), board_82571 },
7470 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_DUAL), board_82571 },
7471 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_QUAD), board_82571 },
7472 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571PT_QUAD_COPPER), board_82571 },
7473 
7474 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI), board_82572 },
7475 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_COPPER), board_82572 },
7476 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_FIBER), board_82572 },
7477 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_SERDES), board_82572 },
7478 
7479 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E), board_82573 },
7480 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E_IAMT), board_82573 },
7481 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573L), board_82573 },
7482 
7483 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574L), board_82574 },
7484 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574LA), board_82574 },
7485 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82583V), board_82583 },
7486 
7487 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_DPT),
7488 	  board_80003es2lan },
7489 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_SPT),
7490 	  board_80003es2lan },
7491 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_DPT),
7492 	  board_80003es2lan },
7493 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_SPT),
7494 	  board_80003es2lan },
7495 
7496 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE), board_ich8lan },
7497 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_G), board_ich8lan },
7498 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_GT), board_ich8lan },
7499 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_AMT), board_ich8lan },
7500 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_C), board_ich8lan },
7501 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M), board_ich8lan },
7502 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M_AMT), board_ich8lan },
7503 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_82567V_3), board_ich8lan },
7504 
7505 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE), board_ich9lan },
7506 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_G), board_ich9lan },
7507 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_GT), board_ich9lan },
7508 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_AMT), board_ich9lan },
7509 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_C), board_ich9lan },
7510 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_BM), board_ich9lan },
7511 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M), board_ich9lan },
7512 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_AMT), board_ich9lan },
7513 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_V), board_ich9lan },
7514 
7515 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LM), board_ich9lan },
7516 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LF), board_ich9lan },
7517 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_V), board_ich9lan },
7518 
7519 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LM), board_ich10lan },
7520 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LF), board_ich10lan },
7521 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_V), board_ich10lan },
7522 
7523 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LM), board_pchlan },
7524 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LC), board_pchlan },
7525 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DM), board_pchlan },
7526 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DC), board_pchlan },
7527 
7528 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_LM), board_pch2lan },
7529 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_V), board_pch2lan },
7530 
7531 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_LM), board_pch_lpt },
7532 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_V), board_pch_lpt },
7533 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_LM), board_pch_lpt },
7534 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_V), board_pch_lpt },
7535 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM2), board_pch_lpt },
7536 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V2), board_pch_lpt },
7537 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM3), board_pch_lpt },
7538 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V3), board_pch_lpt },
7539 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM), board_pch_spt },
7540 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V), board_pch_spt },
7541 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM2), board_pch_spt },
7542 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V2), board_pch_spt },
7543 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LBG_I219_LM3), board_pch_spt },
7544 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM4), board_pch_spt },
7545 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V4), board_pch_spt },
7546 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM5), board_pch_spt },
7547 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V5), board_pch_spt },
7548 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM6), board_pch_cnp },
7549 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V6), board_pch_cnp },
7550 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM7), board_pch_cnp },
7551 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V7), board_pch_cnp },
7552 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM8), board_pch_cnp },
7553 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V8), board_pch_cnp },
7554 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM9), board_pch_cnp },
7555 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V9), board_pch_cnp },
7556 
7557 	{ 0, 0, 0, 0, 0, 0, 0 }	/* terminate list */
7558 };
7559 MODULE_DEVICE_TABLE(pci, e1000_pci_tbl);
7560 
7561 static const struct dev_pm_ops e1000_pm_ops = {
7562 #ifdef CONFIG_PM_SLEEP
7563 	.suspend	= e1000e_pm_suspend,
7564 	.resume		= e1000e_pm_resume,
7565 	.freeze		= e1000e_pm_freeze,
7566 	.thaw		= e1000e_pm_thaw,
7567 	.poweroff	= e1000e_pm_suspend,
7568 	.restore	= e1000e_pm_resume,
7569 #endif
7570 	SET_RUNTIME_PM_OPS(e1000e_pm_runtime_suspend, e1000e_pm_runtime_resume,
7571 			   e1000e_pm_runtime_idle)
7572 };
7573 
7574 /* PCI Device API Driver */
7575 static struct pci_driver e1000_driver = {
7576 	.name     = e1000e_driver_name,
7577 	.id_table = e1000_pci_tbl,
7578 	.probe    = e1000_probe,
7579 	.remove   = e1000_remove,
7580 	.driver   = {
7581 		.pm = &e1000_pm_ops,
7582 	},
7583 	.shutdown = e1000_shutdown,
7584 	.err_handler = &e1000_err_handler
7585 };
7586 
7587 /**
7588  * e1000_init_module - Driver Registration Routine
7589  *
7590  * e1000_init_module is the first routine called when the driver is
7591  * loaded. All it does is register with the PCI subsystem.
7592  **/
7593 static int __init e1000_init_module(void)
7594 {
7595 	pr_info("Intel(R) PRO/1000 Network Driver - %s\n",
7596 		e1000e_driver_version);
7597 	pr_info("Copyright(c) 1999 - 2015 Intel Corporation.\n");
7598 
7599 	return pci_register_driver(&e1000_driver);
7600 }
7601 module_init(e1000_init_module);
7602 
7603 /**
7604  * e1000_exit_module - Driver Exit Cleanup Routine
7605  *
7606  * e1000_exit_module is called just before the driver is removed
7607  * from memory.
7608  **/
7609 static void __exit e1000_exit_module(void)
7610 {
7611 	pci_unregister_driver(&e1000_driver);
7612 }
7613 module_exit(e1000_exit_module);
7614 
7615 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
7616 MODULE_DESCRIPTION("Intel(R) PRO/1000 Network Driver");
7617 MODULE_LICENSE("GPL");
7618 MODULE_VERSION(DRV_VERSION);
7619 
7620 /* netdev.c */
7621