xref: /linux/drivers/net/ethernet/intel/e1000e/netdev.c (revision c75c5ab575af7db707689cdbb5a5c458e9a034bb)
1 /*******************************************************************************
2 
3   Intel PRO/1000 Linux driver
4   Copyright(c) 1999 - 2013 Intel Corporation.
5 
6   This program is free software; you can redistribute it and/or modify it
7   under the terms and conditions of the GNU General Public License,
8   version 2, as published by the Free Software Foundation.
9 
10   This program is distributed in the hope it will be useful, but WITHOUT
11   ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12   FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13   more details.
14 
15   You should have received a copy of the GNU General Public License along with
16   this program; if not, write to the Free Software Foundation, Inc.,
17   51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
18 
19   The full GNU General Public License is included in this distribution in
20   the file called "COPYING".
21 
22   Contact Information:
23   Linux NICS <linux.nics@intel.com>
24   e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
25   Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
26 
27 *******************************************************************************/
28 
29 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
30 
31 #include <linux/module.h>
32 #include <linux/types.h>
33 #include <linux/init.h>
34 #include <linux/pci.h>
35 #include <linux/vmalloc.h>
36 #include <linux/pagemap.h>
37 #include <linux/delay.h>
38 #include <linux/netdevice.h>
39 #include <linux/interrupt.h>
40 #include <linux/tcp.h>
41 #include <linux/ipv6.h>
42 #include <linux/slab.h>
43 #include <net/checksum.h>
44 #include <net/ip6_checksum.h>
45 #include <linux/ethtool.h>
46 #include <linux/if_vlan.h>
47 #include <linux/cpu.h>
48 #include <linux/smp.h>
49 #include <linux/pm_qos.h>
50 #include <linux/pm_runtime.h>
51 #include <linux/aer.h>
52 #include <linux/prefetch.h>
53 
54 #include "e1000.h"
55 
56 #define DRV_EXTRAVERSION "-k"
57 
58 #define DRV_VERSION "2.2.14" DRV_EXTRAVERSION
59 char e1000e_driver_name[] = "e1000e";
60 const char e1000e_driver_version[] = DRV_VERSION;
61 
62 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
63 static int debug = -1;
64 module_param(debug, int, 0);
65 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
66 
67 static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state);
68 
69 static const struct e1000_info *e1000_info_tbl[] = {
70 	[board_82571]		= &e1000_82571_info,
71 	[board_82572]		= &e1000_82572_info,
72 	[board_82573]		= &e1000_82573_info,
73 	[board_82574]		= &e1000_82574_info,
74 	[board_82583]		= &e1000_82583_info,
75 	[board_80003es2lan]	= &e1000_es2_info,
76 	[board_ich8lan]		= &e1000_ich8_info,
77 	[board_ich9lan]		= &e1000_ich9_info,
78 	[board_ich10lan]	= &e1000_ich10_info,
79 	[board_pchlan]		= &e1000_pch_info,
80 	[board_pch2lan]		= &e1000_pch2_info,
81 	[board_pch_lpt]		= &e1000_pch_lpt_info,
82 };
83 
84 struct e1000_reg_info {
85 	u32 ofs;
86 	char *name;
87 };
88 
89 static const struct e1000_reg_info e1000_reg_info_tbl[] = {
90 	/* General Registers */
91 	{E1000_CTRL, "CTRL"},
92 	{E1000_STATUS, "STATUS"},
93 	{E1000_CTRL_EXT, "CTRL_EXT"},
94 
95 	/* Interrupt Registers */
96 	{E1000_ICR, "ICR"},
97 
98 	/* Rx Registers */
99 	{E1000_RCTL, "RCTL"},
100 	{E1000_RDLEN(0), "RDLEN"},
101 	{E1000_RDH(0), "RDH"},
102 	{E1000_RDT(0), "RDT"},
103 	{E1000_RDTR, "RDTR"},
104 	{E1000_RXDCTL(0), "RXDCTL"},
105 	{E1000_ERT, "ERT"},
106 	{E1000_RDBAL(0), "RDBAL"},
107 	{E1000_RDBAH(0), "RDBAH"},
108 	{E1000_RDFH, "RDFH"},
109 	{E1000_RDFT, "RDFT"},
110 	{E1000_RDFHS, "RDFHS"},
111 	{E1000_RDFTS, "RDFTS"},
112 	{E1000_RDFPC, "RDFPC"},
113 
114 	/* Tx Registers */
115 	{E1000_TCTL, "TCTL"},
116 	{E1000_TDBAL(0), "TDBAL"},
117 	{E1000_TDBAH(0), "TDBAH"},
118 	{E1000_TDLEN(0), "TDLEN"},
119 	{E1000_TDH(0), "TDH"},
120 	{E1000_TDT(0), "TDT"},
121 	{E1000_TIDV, "TIDV"},
122 	{E1000_TXDCTL(0), "TXDCTL"},
123 	{E1000_TADV, "TADV"},
124 	{E1000_TARC(0), "TARC"},
125 	{E1000_TDFH, "TDFH"},
126 	{E1000_TDFT, "TDFT"},
127 	{E1000_TDFHS, "TDFHS"},
128 	{E1000_TDFTS, "TDFTS"},
129 	{E1000_TDFPC, "TDFPC"},
130 
131 	/* List Terminator */
132 	{0, NULL}
133 };
134 
135 /**
136  * e1000_regdump - register printout routine
137  * @hw: pointer to the HW structure
138  * @reginfo: pointer to the register info table
139  **/
140 static void e1000_regdump(struct e1000_hw *hw, struct e1000_reg_info *reginfo)
141 {
142 	int n = 0;
143 	char rname[16];
144 	u32 regs[8];
145 
146 	switch (reginfo->ofs) {
147 	case E1000_RXDCTL(0):
148 		for (n = 0; n < 2; n++)
149 			regs[n] = __er32(hw, E1000_RXDCTL(n));
150 		break;
151 	case E1000_TXDCTL(0):
152 		for (n = 0; n < 2; n++)
153 			regs[n] = __er32(hw, E1000_TXDCTL(n));
154 		break;
155 	case E1000_TARC(0):
156 		for (n = 0; n < 2; n++)
157 			regs[n] = __er32(hw, E1000_TARC(n));
158 		break;
159 	default:
160 		pr_info("%-15s %08x\n",
161 			reginfo->name, __er32(hw, reginfo->ofs));
162 		return;
163 	}
164 
165 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-1]");
166 	pr_info("%-15s %08x %08x\n", rname, regs[0], regs[1]);
167 }
168 
169 static void e1000e_dump_ps_pages(struct e1000_adapter *adapter,
170 				 struct e1000_buffer *bi)
171 {
172 	int i;
173 	struct e1000_ps_page *ps_page;
174 
175 	for (i = 0; i < adapter->rx_ps_pages; i++) {
176 		ps_page = &bi->ps_pages[i];
177 
178 		if (ps_page->page) {
179 			pr_info("packet dump for ps_page %d:\n", i);
180 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
181 				       16, 1, page_address(ps_page->page),
182 				       PAGE_SIZE, true);
183 		}
184 	}
185 }
186 
187 /**
188  * e1000e_dump - Print registers, Tx-ring and Rx-ring
189  * @adapter: board private structure
190  **/
191 static void e1000e_dump(struct e1000_adapter *adapter)
192 {
193 	struct net_device *netdev = adapter->netdev;
194 	struct e1000_hw *hw = &adapter->hw;
195 	struct e1000_reg_info *reginfo;
196 	struct e1000_ring *tx_ring = adapter->tx_ring;
197 	struct e1000_tx_desc *tx_desc;
198 	struct my_u0 {
199 		__le64 a;
200 		__le64 b;
201 	} *u0;
202 	struct e1000_buffer *buffer_info;
203 	struct e1000_ring *rx_ring = adapter->rx_ring;
204 	union e1000_rx_desc_packet_split *rx_desc_ps;
205 	union e1000_rx_desc_extended *rx_desc;
206 	struct my_u1 {
207 		__le64 a;
208 		__le64 b;
209 		__le64 c;
210 		__le64 d;
211 	} *u1;
212 	u32 staterr;
213 	int i = 0;
214 
215 	if (!netif_msg_hw(adapter))
216 		return;
217 
218 	/* Print netdevice Info */
219 	if (netdev) {
220 		dev_info(&adapter->pdev->dev, "Net device Info\n");
221 		pr_info("Device Name     state            trans_start      last_rx\n");
222 		pr_info("%-15s %016lX %016lX %016lX\n",
223 			netdev->name, netdev->state, netdev->trans_start,
224 			netdev->last_rx);
225 	}
226 
227 	/* Print Registers */
228 	dev_info(&adapter->pdev->dev, "Register Dump\n");
229 	pr_info(" Register Name   Value\n");
230 	for (reginfo = (struct e1000_reg_info *)e1000_reg_info_tbl;
231 	     reginfo->name; reginfo++) {
232 		e1000_regdump(hw, reginfo);
233 	}
234 
235 	/* Print Tx Ring Summary */
236 	if (!netdev || !netif_running(netdev))
237 		return;
238 
239 	dev_info(&adapter->pdev->dev, "Tx Ring Summary\n");
240 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
241 	buffer_info = &tx_ring->buffer_info[tx_ring->next_to_clean];
242 	pr_info(" %5d %5X %5X %016llX %04X %3X %016llX\n",
243 		0, tx_ring->next_to_use, tx_ring->next_to_clean,
244 		(unsigned long long)buffer_info->dma,
245 		buffer_info->length,
246 		buffer_info->next_to_watch,
247 		(unsigned long long)buffer_info->time_stamp);
248 
249 	/* Print Tx Ring */
250 	if (!netif_msg_tx_done(adapter))
251 		goto rx_ring_summary;
252 
253 	dev_info(&adapter->pdev->dev, "Tx Ring Dump\n");
254 
255 	/* Transmit Descriptor Formats - DEXT[29] is 0 (Legacy) or 1 (Extended)
256 	 *
257 	 * Legacy Transmit Descriptor
258 	 *   +--------------------------------------------------------------+
259 	 * 0 |         Buffer Address [63:0] (Reserved on Write Back)       |
260 	 *   +--------------------------------------------------------------+
261 	 * 8 | Special  |    CSS     | Status |  CMD    |  CSO   |  Length  |
262 	 *   +--------------------------------------------------------------+
263 	 *   63       48 47        36 35    32 31     24 23    16 15        0
264 	 *
265 	 * Extended Context Descriptor (DTYP=0x0) for TSO or checksum offload
266 	 *   63      48 47    40 39       32 31             16 15    8 7      0
267 	 *   +----------------------------------------------------------------+
268 	 * 0 |  TUCSE  | TUCS0  |   TUCSS   |     IPCSE       | IPCS0 | IPCSS |
269 	 *   +----------------------------------------------------------------+
270 	 * 8 |   MSS   | HDRLEN | RSV | STA | TUCMD | DTYP |      PAYLEN      |
271 	 *   +----------------------------------------------------------------+
272 	 *   63      48 47    40 39 36 35 32 31   24 23  20 19                0
273 	 *
274 	 * Extended Data Descriptor (DTYP=0x1)
275 	 *   +----------------------------------------------------------------+
276 	 * 0 |                     Buffer Address [63:0]                      |
277 	 *   +----------------------------------------------------------------+
278 	 * 8 | VLAN tag |  POPTS  | Rsvd | Status | Command | DTYP |  DTALEN  |
279 	 *   +----------------------------------------------------------------+
280 	 *   63       48 47     40 39  36 35    32 31     24 23  20 19        0
281 	 */
282 	pr_info("Tl[desc]     [address 63:0  ] [SpeCssSCmCsLen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Legacy format\n");
283 	pr_info("Tc[desc]     [Ce CoCsIpceCoS] [MssHlRSCm0Plen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Context format\n");
284 	pr_info("Td[desc]     [address 63:0  ] [VlaPoRSCm1Dlen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Data format\n");
285 	for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
286 		const char *next_desc;
287 		tx_desc = E1000_TX_DESC(*tx_ring, i);
288 		buffer_info = &tx_ring->buffer_info[i];
289 		u0 = (struct my_u0 *)tx_desc;
290 		if (i == tx_ring->next_to_use && i == tx_ring->next_to_clean)
291 			next_desc = " NTC/U";
292 		else if (i == tx_ring->next_to_use)
293 			next_desc = " NTU";
294 		else if (i == tx_ring->next_to_clean)
295 			next_desc = " NTC";
296 		else
297 			next_desc = "";
298 		pr_info("T%c[0x%03X]    %016llX %016llX %016llX %04X  %3X %016llX %p%s\n",
299 			(!(le64_to_cpu(u0->b) & (1 << 29)) ? 'l' :
300 			 ((le64_to_cpu(u0->b) & (1 << 20)) ? 'd' : 'c')),
301 			i,
302 			(unsigned long long)le64_to_cpu(u0->a),
303 			(unsigned long long)le64_to_cpu(u0->b),
304 			(unsigned long long)buffer_info->dma,
305 			buffer_info->length, buffer_info->next_to_watch,
306 			(unsigned long long)buffer_info->time_stamp,
307 			buffer_info->skb, next_desc);
308 
309 		if (netif_msg_pktdata(adapter) && buffer_info->skb)
310 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
311 				       16, 1, buffer_info->skb->data,
312 				       buffer_info->skb->len, true);
313 	}
314 
315 	/* Print Rx Ring Summary */
316 rx_ring_summary:
317 	dev_info(&adapter->pdev->dev, "Rx Ring Summary\n");
318 	pr_info("Queue [NTU] [NTC]\n");
319 	pr_info(" %5d %5X %5X\n",
320 		0, rx_ring->next_to_use, rx_ring->next_to_clean);
321 
322 	/* Print Rx Ring */
323 	if (!netif_msg_rx_status(adapter))
324 		return;
325 
326 	dev_info(&adapter->pdev->dev, "Rx Ring Dump\n");
327 	switch (adapter->rx_ps_pages) {
328 	case 1:
329 	case 2:
330 	case 3:
331 		/* [Extended] Packet Split Receive Descriptor Format
332 		 *
333 		 *    +-----------------------------------------------------+
334 		 *  0 |                Buffer Address 0 [63:0]              |
335 		 *    +-----------------------------------------------------+
336 		 *  8 |                Buffer Address 1 [63:0]              |
337 		 *    +-----------------------------------------------------+
338 		 * 16 |                Buffer Address 2 [63:0]              |
339 		 *    +-----------------------------------------------------+
340 		 * 24 |                Buffer Address 3 [63:0]              |
341 		 *    +-----------------------------------------------------+
342 		 */
343 		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");
344 		/* [Extended] Receive Descriptor (Write-Back) Format
345 		 *
346 		 *   63       48 47    32 31     13 12    8 7    4 3        0
347 		 *   +------------------------------------------------------+
348 		 * 0 | Packet   | IP     |  Rsvd   | MRQ   | Rsvd | MRQ RSS |
349 		 *   | Checksum | Ident  |         | Queue |      |  Type   |
350 		 *   +------------------------------------------------------+
351 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
352 		 *   +------------------------------------------------------+
353 		 *   63       48 47    32 31            20 19               0
354 		 */
355 		pr_info("RWB[desc]      [ck ipid mrqhsh] [vl   l0 ee  es] [ l3  l2  l1 hs] [reserved      ] ---------------- [bi->skb] <-- Ext Rx Write-Back format\n");
356 		for (i = 0; i < rx_ring->count; i++) {
357 			const char *next_desc;
358 			buffer_info = &rx_ring->buffer_info[i];
359 			rx_desc_ps = E1000_RX_DESC_PS(*rx_ring, i);
360 			u1 = (struct my_u1 *)rx_desc_ps;
361 			staterr =
362 			    le32_to_cpu(rx_desc_ps->wb.middle.status_error);
363 
364 			if (i == rx_ring->next_to_use)
365 				next_desc = " NTU";
366 			else if (i == rx_ring->next_to_clean)
367 				next_desc = " NTC";
368 			else
369 				next_desc = "";
370 
371 			if (staterr & E1000_RXD_STAT_DD) {
372 				/* Descriptor Done */
373 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX ---------------- %p%s\n",
374 					"RWB", i,
375 					(unsigned long long)le64_to_cpu(u1->a),
376 					(unsigned long long)le64_to_cpu(u1->b),
377 					(unsigned long long)le64_to_cpu(u1->c),
378 					(unsigned long long)le64_to_cpu(u1->d),
379 					buffer_info->skb, next_desc);
380 			} else {
381 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX %016llX %p%s\n",
382 					"R  ", i,
383 					(unsigned long long)le64_to_cpu(u1->a),
384 					(unsigned long long)le64_to_cpu(u1->b),
385 					(unsigned long long)le64_to_cpu(u1->c),
386 					(unsigned long long)le64_to_cpu(u1->d),
387 					(unsigned long long)buffer_info->dma,
388 					buffer_info->skb, next_desc);
389 
390 				if (netif_msg_pktdata(adapter))
391 					e1000e_dump_ps_pages(adapter,
392 							     buffer_info);
393 			}
394 		}
395 		break;
396 	default:
397 	case 0:
398 		/* Extended Receive Descriptor (Read) Format
399 		 *
400 		 *   +-----------------------------------------------------+
401 		 * 0 |                Buffer Address [63:0]                |
402 		 *   +-----------------------------------------------------+
403 		 * 8 |                      Reserved                       |
404 		 *   +-----------------------------------------------------+
405 		 */
406 		pr_info("R  [desc]      [buf addr 63:0 ] [reserved 63:0 ] [bi->dma       ] [bi->skb] <-- Ext (Read) format\n");
407 		/* Extended Receive Descriptor (Write-Back) Format
408 		 *
409 		 *   63       48 47    32 31    24 23            4 3        0
410 		 *   +------------------------------------------------------+
411 		 *   |     RSS Hash      |        |               |         |
412 		 * 0 +-------------------+  Rsvd  |   Reserved    | MRQ RSS |
413 		 *   | Packet   | IP     |        |               |  Type   |
414 		 *   | Checksum | Ident  |        |               |         |
415 		 *   +------------------------------------------------------+
416 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
417 		 *   +------------------------------------------------------+
418 		 *   63       48 47    32 31            20 19               0
419 		 */
420 		pr_info("RWB[desc]      [cs ipid    mrq] [vt   ln xe  xs] [bi->skb] <-- Ext (Write-Back) format\n");
421 
422 		for (i = 0; i < rx_ring->count; i++) {
423 			const char *next_desc;
424 
425 			buffer_info = &rx_ring->buffer_info[i];
426 			rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
427 			u1 = (struct my_u1 *)rx_desc;
428 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
429 
430 			if (i == rx_ring->next_to_use)
431 				next_desc = " NTU";
432 			else if (i == rx_ring->next_to_clean)
433 				next_desc = " NTC";
434 			else
435 				next_desc = "";
436 
437 			if (staterr & E1000_RXD_STAT_DD) {
438 				/* Descriptor Done */
439 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %p%s\n",
440 					"RWB", i,
441 					(unsigned long long)le64_to_cpu(u1->a),
442 					(unsigned long long)le64_to_cpu(u1->b),
443 					buffer_info->skb, next_desc);
444 			} else {
445 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %p%s\n",
446 					"R  ", i,
447 					(unsigned long long)le64_to_cpu(u1->a),
448 					(unsigned long long)le64_to_cpu(u1->b),
449 					(unsigned long long)buffer_info->dma,
450 					buffer_info->skb, next_desc);
451 
452 				if (netif_msg_pktdata(adapter) &&
453 				    buffer_info->skb)
454 					print_hex_dump(KERN_INFO, "",
455 						       DUMP_PREFIX_ADDRESS, 16,
456 						       1,
457 						       buffer_info->skb->data,
458 						       adapter->rx_buffer_len,
459 						       true);
460 			}
461 		}
462 	}
463 }
464 
465 /**
466  * e1000_desc_unused - calculate if we have unused descriptors
467  **/
468 static int e1000_desc_unused(struct e1000_ring *ring)
469 {
470 	if (ring->next_to_clean > ring->next_to_use)
471 		return ring->next_to_clean - ring->next_to_use - 1;
472 
473 	return ring->count + ring->next_to_clean - ring->next_to_use - 1;
474 }
475 
476 /**
477  * e1000e_systim_to_hwtstamp - convert system time value to hw time stamp
478  * @adapter: board private structure
479  * @hwtstamps: time stamp structure to update
480  * @systim: unsigned 64bit system time value.
481  *
482  * Convert the system time value stored in the RX/TXSTMP registers into a
483  * hwtstamp which can be used by the upper level time stamping functions.
484  *
485  * The 'systim_lock' spinlock is used to protect the consistency of the
486  * system time value. This is needed because reading the 64 bit time
487  * value involves reading two 32 bit registers. The first read latches the
488  * value.
489  **/
490 static void e1000e_systim_to_hwtstamp(struct e1000_adapter *adapter,
491 				      struct skb_shared_hwtstamps *hwtstamps,
492 				      u64 systim)
493 {
494 	u64 ns;
495 	unsigned long flags;
496 
497 	spin_lock_irqsave(&adapter->systim_lock, flags);
498 	ns = timecounter_cyc2time(&adapter->tc, systim);
499 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
500 
501 	memset(hwtstamps, 0, sizeof(*hwtstamps));
502 	hwtstamps->hwtstamp = ns_to_ktime(ns);
503 }
504 
505 /**
506  * e1000e_rx_hwtstamp - utility function which checks for Rx time stamp
507  * @adapter: board private structure
508  * @status: descriptor extended error and status field
509  * @skb: particular skb to include time stamp
510  *
511  * If the time stamp is valid, convert it into the timecounter ns value
512  * and store that result into the shhwtstamps structure which is passed
513  * up the network stack.
514  **/
515 static void e1000e_rx_hwtstamp(struct e1000_adapter *adapter, u32 status,
516 			       struct sk_buff *skb)
517 {
518 	struct e1000_hw *hw = &adapter->hw;
519 	u64 rxstmp;
520 
521 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP) ||
522 	    !(status & E1000_RXDEXT_STATERR_TST) ||
523 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID))
524 		return;
525 
526 	/* The Rx time stamp registers contain the time stamp.  No other
527 	 * received packet will be time stamped until the Rx time stamp
528 	 * registers are read.  Because only one packet can be time stamped
529 	 * at a time, the register values must belong to this packet and
530 	 * therefore none of the other additional attributes need to be
531 	 * compared.
532 	 */
533 	rxstmp = (u64)er32(RXSTMPL);
534 	rxstmp |= (u64)er32(RXSTMPH) << 32;
535 	e1000e_systim_to_hwtstamp(adapter, skb_hwtstamps(skb), rxstmp);
536 
537 	adapter->flags2 &= ~FLAG2_CHECK_RX_HWTSTAMP;
538 }
539 
540 /**
541  * e1000_receive_skb - helper function to handle Rx indications
542  * @adapter: board private structure
543  * @staterr: descriptor extended error and status field as written by hardware
544  * @vlan: descriptor vlan field as written by hardware (no le/be conversion)
545  * @skb: pointer to sk_buff to be indicated to stack
546  **/
547 static void e1000_receive_skb(struct e1000_adapter *adapter,
548 			      struct net_device *netdev, struct sk_buff *skb,
549 			      u32 staterr, __le16 vlan)
550 {
551 	u16 tag = le16_to_cpu(vlan);
552 
553 	e1000e_rx_hwtstamp(adapter, staterr, skb);
554 
555 	skb->protocol = eth_type_trans(skb, netdev);
556 
557 	if (staterr & E1000_RXD_STAT_VP)
558 		__vlan_hwaccel_put_tag(skb, tag);
559 
560 	napi_gro_receive(&adapter->napi, skb);
561 }
562 
563 /**
564  * e1000_rx_checksum - Receive Checksum Offload
565  * @adapter: board private structure
566  * @status_err: receive descriptor status and error fields
567  * @csum: receive descriptor csum field
568  * @sk_buff: socket buffer with received data
569  **/
570 static void e1000_rx_checksum(struct e1000_adapter *adapter, u32 status_err,
571 			      struct sk_buff *skb)
572 {
573 	u16 status = (u16)status_err;
574 	u8 errors = (u8)(status_err >> 24);
575 
576 	skb_checksum_none_assert(skb);
577 
578 	/* Rx checksum disabled */
579 	if (!(adapter->netdev->features & NETIF_F_RXCSUM))
580 		return;
581 
582 	/* Ignore Checksum bit is set */
583 	if (status & E1000_RXD_STAT_IXSM)
584 		return;
585 
586 	/* TCP/UDP checksum error bit or IP checksum error bit is set */
587 	if (errors & (E1000_RXD_ERR_TCPE | E1000_RXD_ERR_IPE)) {
588 		/* let the stack verify checksum errors */
589 		adapter->hw_csum_err++;
590 		return;
591 	}
592 
593 	/* TCP/UDP Checksum has not been calculated */
594 	if (!(status & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)))
595 		return;
596 
597 	/* It must be a TCP or UDP packet with a valid checksum */
598 	skb->ip_summed = CHECKSUM_UNNECESSARY;
599 	adapter->hw_csum_good++;
600 }
601 
602 static void e1000e_update_rdt_wa(struct e1000_ring *rx_ring, unsigned int i)
603 {
604 	struct e1000_adapter *adapter = rx_ring->adapter;
605 	struct e1000_hw *hw = &adapter->hw;
606 	s32 ret_val = __ew32_prepare(hw);
607 
608 	writel(i, rx_ring->tail);
609 
610 	if (unlikely(!ret_val && (i != readl(rx_ring->tail)))) {
611 		u32 rctl = er32(RCTL);
612 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
613 		e_err("ME firmware caused invalid RDT - resetting\n");
614 		schedule_work(&adapter->reset_task);
615 	}
616 }
617 
618 static void e1000e_update_tdt_wa(struct e1000_ring *tx_ring, unsigned int i)
619 {
620 	struct e1000_adapter *adapter = tx_ring->adapter;
621 	struct e1000_hw *hw = &adapter->hw;
622 	s32 ret_val = __ew32_prepare(hw);
623 
624 	writel(i, tx_ring->tail);
625 
626 	if (unlikely(!ret_val && (i != readl(tx_ring->tail)))) {
627 		u32 tctl = er32(TCTL);
628 		ew32(TCTL, tctl & ~E1000_TCTL_EN);
629 		e_err("ME firmware caused invalid TDT - resetting\n");
630 		schedule_work(&adapter->reset_task);
631 	}
632 }
633 
634 /**
635  * e1000_alloc_rx_buffers - Replace used receive buffers
636  * @rx_ring: Rx descriptor ring
637  **/
638 static void e1000_alloc_rx_buffers(struct e1000_ring *rx_ring,
639 				   int cleaned_count, gfp_t gfp)
640 {
641 	struct e1000_adapter *adapter = rx_ring->adapter;
642 	struct net_device *netdev = adapter->netdev;
643 	struct pci_dev *pdev = adapter->pdev;
644 	union e1000_rx_desc_extended *rx_desc;
645 	struct e1000_buffer *buffer_info;
646 	struct sk_buff *skb;
647 	unsigned int i;
648 	unsigned int bufsz = adapter->rx_buffer_len;
649 
650 	i = rx_ring->next_to_use;
651 	buffer_info = &rx_ring->buffer_info[i];
652 
653 	while (cleaned_count--) {
654 		skb = buffer_info->skb;
655 		if (skb) {
656 			skb_trim(skb, 0);
657 			goto map_skb;
658 		}
659 
660 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
661 		if (!skb) {
662 			/* Better luck next round */
663 			adapter->alloc_rx_buff_failed++;
664 			break;
665 		}
666 
667 		buffer_info->skb = skb;
668 map_skb:
669 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
670 						  adapter->rx_buffer_len,
671 						  DMA_FROM_DEVICE);
672 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
673 			dev_err(&pdev->dev, "Rx DMA map failed\n");
674 			adapter->rx_dma_failed++;
675 			break;
676 		}
677 
678 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
679 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
680 
681 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
682 			/* Force memory writes to complete before letting h/w
683 			 * know there are new descriptors to fetch.  (Only
684 			 * applicable for weak-ordered memory model archs,
685 			 * such as IA-64).
686 			 */
687 			wmb();
688 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
689 				e1000e_update_rdt_wa(rx_ring, i);
690 			else
691 				writel(i, rx_ring->tail);
692 		}
693 		i++;
694 		if (i == rx_ring->count)
695 			i = 0;
696 		buffer_info = &rx_ring->buffer_info[i];
697 	}
698 
699 	rx_ring->next_to_use = i;
700 }
701 
702 /**
703  * e1000_alloc_rx_buffers_ps - Replace used receive buffers; packet split
704  * @rx_ring: Rx descriptor ring
705  **/
706 static void e1000_alloc_rx_buffers_ps(struct e1000_ring *rx_ring,
707 				      int cleaned_count, gfp_t gfp)
708 {
709 	struct e1000_adapter *adapter = rx_ring->adapter;
710 	struct net_device *netdev = adapter->netdev;
711 	struct pci_dev *pdev = adapter->pdev;
712 	union e1000_rx_desc_packet_split *rx_desc;
713 	struct e1000_buffer *buffer_info;
714 	struct e1000_ps_page *ps_page;
715 	struct sk_buff *skb;
716 	unsigned int i, j;
717 
718 	i = rx_ring->next_to_use;
719 	buffer_info = &rx_ring->buffer_info[i];
720 
721 	while (cleaned_count--) {
722 		rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
723 
724 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
725 			ps_page = &buffer_info->ps_pages[j];
726 			if (j >= adapter->rx_ps_pages) {
727 				/* all unused desc entries get hw null ptr */
728 				rx_desc->read.buffer_addr[j + 1] =
729 				    ~cpu_to_le64(0);
730 				continue;
731 			}
732 			if (!ps_page->page) {
733 				ps_page->page = alloc_page(gfp);
734 				if (!ps_page->page) {
735 					adapter->alloc_rx_buff_failed++;
736 					goto no_buffers;
737 				}
738 				ps_page->dma = dma_map_page(&pdev->dev,
739 							    ps_page->page,
740 							    0, PAGE_SIZE,
741 							    DMA_FROM_DEVICE);
742 				if (dma_mapping_error(&pdev->dev,
743 						      ps_page->dma)) {
744 					dev_err(&adapter->pdev->dev,
745 						"Rx DMA page map failed\n");
746 					adapter->rx_dma_failed++;
747 					goto no_buffers;
748 				}
749 			}
750 			/* Refresh the desc even if buffer_addrs
751 			 * didn't change because each write-back
752 			 * erases this info.
753 			 */
754 			rx_desc->read.buffer_addr[j + 1] =
755 			    cpu_to_le64(ps_page->dma);
756 		}
757 
758 		skb = __netdev_alloc_skb_ip_align(netdev,
759 						  adapter->rx_ps_bsize0,
760 						  gfp);
761 
762 		if (!skb) {
763 			adapter->alloc_rx_buff_failed++;
764 			break;
765 		}
766 
767 		buffer_info->skb = skb;
768 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
769 						  adapter->rx_ps_bsize0,
770 						  DMA_FROM_DEVICE);
771 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
772 			dev_err(&pdev->dev, "Rx DMA map failed\n");
773 			adapter->rx_dma_failed++;
774 			/* cleanup skb */
775 			dev_kfree_skb_any(skb);
776 			buffer_info->skb = NULL;
777 			break;
778 		}
779 
780 		rx_desc->read.buffer_addr[0] = cpu_to_le64(buffer_info->dma);
781 
782 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
783 			/* Force memory writes to complete before letting h/w
784 			 * know there are new descriptors to fetch.  (Only
785 			 * applicable for weak-ordered memory model archs,
786 			 * such as IA-64).
787 			 */
788 			wmb();
789 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
790 				e1000e_update_rdt_wa(rx_ring, i << 1);
791 			else
792 				writel(i << 1, rx_ring->tail);
793 		}
794 
795 		i++;
796 		if (i == rx_ring->count)
797 			i = 0;
798 		buffer_info = &rx_ring->buffer_info[i];
799 	}
800 
801 no_buffers:
802 	rx_ring->next_to_use = i;
803 }
804 
805 /**
806  * e1000_alloc_jumbo_rx_buffers - Replace used jumbo receive buffers
807  * @rx_ring: Rx descriptor ring
808  * @cleaned_count: number of buffers to allocate this pass
809  **/
810 
811 static void e1000_alloc_jumbo_rx_buffers(struct e1000_ring *rx_ring,
812 					 int cleaned_count, gfp_t gfp)
813 {
814 	struct e1000_adapter *adapter = rx_ring->adapter;
815 	struct net_device *netdev = adapter->netdev;
816 	struct pci_dev *pdev = adapter->pdev;
817 	union e1000_rx_desc_extended *rx_desc;
818 	struct e1000_buffer *buffer_info;
819 	struct sk_buff *skb;
820 	unsigned int i;
821 	unsigned int bufsz = 256 - 16;	/* for skb_reserve */
822 
823 	i = rx_ring->next_to_use;
824 	buffer_info = &rx_ring->buffer_info[i];
825 
826 	while (cleaned_count--) {
827 		skb = buffer_info->skb;
828 		if (skb) {
829 			skb_trim(skb, 0);
830 			goto check_page;
831 		}
832 
833 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
834 		if (unlikely(!skb)) {
835 			/* Better luck next round */
836 			adapter->alloc_rx_buff_failed++;
837 			break;
838 		}
839 
840 		buffer_info->skb = skb;
841 check_page:
842 		/* allocate a new page if necessary */
843 		if (!buffer_info->page) {
844 			buffer_info->page = alloc_page(gfp);
845 			if (unlikely(!buffer_info->page)) {
846 				adapter->alloc_rx_buff_failed++;
847 				break;
848 			}
849 		}
850 
851 		if (!buffer_info->dma) {
852 			buffer_info->dma = dma_map_page(&pdev->dev,
853 			                                buffer_info->page, 0,
854 			                                PAGE_SIZE,
855 							DMA_FROM_DEVICE);
856 			if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
857 				adapter->alloc_rx_buff_failed++;
858 				break;
859 			}
860 		}
861 
862 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
863 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
864 
865 		if (unlikely(++i == rx_ring->count))
866 			i = 0;
867 		buffer_info = &rx_ring->buffer_info[i];
868 	}
869 
870 	if (likely(rx_ring->next_to_use != i)) {
871 		rx_ring->next_to_use = i;
872 		if (unlikely(i-- == 0))
873 			i = (rx_ring->count - 1);
874 
875 		/* Force memory writes to complete before letting h/w
876 		 * know there are new descriptors to fetch.  (Only
877 		 * applicable for weak-ordered memory model archs,
878 		 * such as IA-64).
879 		 */
880 		wmb();
881 		if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
882 			e1000e_update_rdt_wa(rx_ring, i);
883 		else
884 			writel(i, rx_ring->tail);
885 	}
886 }
887 
888 static inline void e1000_rx_hash(struct net_device *netdev, __le32 rss,
889 				 struct sk_buff *skb)
890 {
891 	if (netdev->features & NETIF_F_RXHASH)
892 		skb->rxhash = le32_to_cpu(rss);
893 }
894 
895 /**
896  * e1000_clean_rx_irq - Send received data up the network stack
897  * @rx_ring: Rx descriptor ring
898  *
899  * the return value indicates whether actual cleaning was done, there
900  * is no guarantee that everything was cleaned
901  **/
902 static bool e1000_clean_rx_irq(struct e1000_ring *rx_ring, int *work_done,
903 			       int work_to_do)
904 {
905 	struct e1000_adapter *adapter = rx_ring->adapter;
906 	struct net_device *netdev = adapter->netdev;
907 	struct pci_dev *pdev = adapter->pdev;
908 	struct e1000_hw *hw = &adapter->hw;
909 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
910 	struct e1000_buffer *buffer_info, *next_buffer;
911 	u32 length, staterr;
912 	unsigned int i;
913 	int cleaned_count = 0;
914 	bool cleaned = false;
915 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
916 
917 	i = rx_ring->next_to_clean;
918 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
919 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
920 	buffer_info = &rx_ring->buffer_info[i];
921 
922 	while (staterr & E1000_RXD_STAT_DD) {
923 		struct sk_buff *skb;
924 
925 		if (*work_done >= work_to_do)
926 			break;
927 		(*work_done)++;
928 		rmb();	/* read descriptor and rx_buffer_info after status DD */
929 
930 		skb = buffer_info->skb;
931 		buffer_info->skb = NULL;
932 
933 		prefetch(skb->data - NET_IP_ALIGN);
934 
935 		i++;
936 		if (i == rx_ring->count)
937 			i = 0;
938 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
939 		prefetch(next_rxd);
940 
941 		next_buffer = &rx_ring->buffer_info[i];
942 
943 		cleaned = true;
944 		cleaned_count++;
945 		dma_unmap_single(&pdev->dev,
946 				 buffer_info->dma,
947 				 adapter->rx_buffer_len,
948 				 DMA_FROM_DEVICE);
949 		buffer_info->dma = 0;
950 
951 		length = le16_to_cpu(rx_desc->wb.upper.length);
952 
953 		/* !EOP means multiple descriptors were used to store a single
954 		 * packet, if that's the case we need to toss it.  In fact, we
955 		 * need to toss every packet with the EOP bit clear and the
956 		 * next frame that _does_ have the EOP bit set, as it is by
957 		 * definition only a frame fragment
958 		 */
959 		if (unlikely(!(staterr & E1000_RXD_STAT_EOP)))
960 			adapter->flags2 |= FLAG2_IS_DISCARDING;
961 
962 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
963 			/* All receives must fit into a single buffer */
964 			e_dbg("Receive packet consumed multiple buffers\n");
965 			/* recycle */
966 			buffer_info->skb = skb;
967 			if (staterr & E1000_RXD_STAT_EOP)
968 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
969 			goto next_desc;
970 		}
971 
972 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
973 			     !(netdev->features & NETIF_F_RXALL))) {
974 			/* recycle */
975 			buffer_info->skb = skb;
976 			goto next_desc;
977 		}
978 
979 		/* adjust length to remove Ethernet CRC */
980 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
981 			/* If configured to store CRC, don't subtract FCS,
982 			 * but keep the FCS bytes out of the total_rx_bytes
983 			 * counter
984 			 */
985 			if (netdev->features & NETIF_F_RXFCS)
986 				total_rx_bytes -= 4;
987 			else
988 				length -= 4;
989 		}
990 
991 		total_rx_bytes += length;
992 		total_rx_packets++;
993 
994 		/* code added for copybreak, this should improve
995 		 * performance for small packets with large amounts
996 		 * of reassembly being done in the stack
997 		 */
998 		if (length < copybreak) {
999 			struct sk_buff *new_skb =
1000 			    netdev_alloc_skb_ip_align(netdev, length);
1001 			if (new_skb) {
1002 				skb_copy_to_linear_data_offset(new_skb,
1003 							       -NET_IP_ALIGN,
1004 							       (skb->data -
1005 								NET_IP_ALIGN),
1006 							       (length +
1007 								NET_IP_ALIGN));
1008 				/* save the skb in buffer_info as good */
1009 				buffer_info->skb = skb;
1010 				skb = new_skb;
1011 			}
1012 			/* else just continue with the old one */
1013 		}
1014 		/* end copybreak code */
1015 		skb_put(skb, length);
1016 
1017 		/* Receive Checksum Offload */
1018 		e1000_rx_checksum(adapter, staterr, skb);
1019 
1020 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1021 
1022 		e1000_receive_skb(adapter, netdev, skb, staterr,
1023 				  rx_desc->wb.upper.vlan);
1024 
1025 next_desc:
1026 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1027 
1028 		/* return some buffers to hardware, one at a time is too slow */
1029 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1030 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1031 					      GFP_ATOMIC);
1032 			cleaned_count = 0;
1033 		}
1034 
1035 		/* use prefetched values */
1036 		rx_desc = next_rxd;
1037 		buffer_info = next_buffer;
1038 
1039 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1040 	}
1041 	rx_ring->next_to_clean = i;
1042 
1043 	cleaned_count = e1000_desc_unused(rx_ring);
1044 	if (cleaned_count)
1045 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1046 
1047 	adapter->total_rx_bytes += total_rx_bytes;
1048 	adapter->total_rx_packets += total_rx_packets;
1049 	return cleaned;
1050 }
1051 
1052 static void e1000_put_txbuf(struct e1000_ring *tx_ring,
1053 			    struct e1000_buffer *buffer_info)
1054 {
1055 	struct e1000_adapter *adapter = tx_ring->adapter;
1056 
1057 	if (buffer_info->dma) {
1058 		if (buffer_info->mapped_as_page)
1059 			dma_unmap_page(&adapter->pdev->dev, buffer_info->dma,
1060 				       buffer_info->length, DMA_TO_DEVICE);
1061 		else
1062 			dma_unmap_single(&adapter->pdev->dev, buffer_info->dma,
1063 					 buffer_info->length, DMA_TO_DEVICE);
1064 		buffer_info->dma = 0;
1065 	}
1066 	if (buffer_info->skb) {
1067 		dev_kfree_skb_any(buffer_info->skb);
1068 		buffer_info->skb = NULL;
1069 	}
1070 	buffer_info->time_stamp = 0;
1071 }
1072 
1073 static void e1000_print_hw_hang(struct work_struct *work)
1074 {
1075 	struct e1000_adapter *adapter = container_of(work,
1076 	                                             struct e1000_adapter,
1077 	                                             print_hang_task);
1078 	struct net_device *netdev = adapter->netdev;
1079 	struct e1000_ring *tx_ring = adapter->tx_ring;
1080 	unsigned int i = tx_ring->next_to_clean;
1081 	unsigned int eop = tx_ring->buffer_info[i].next_to_watch;
1082 	struct e1000_tx_desc *eop_desc = E1000_TX_DESC(*tx_ring, eop);
1083 	struct e1000_hw *hw = &adapter->hw;
1084 	u16 phy_status, phy_1000t_status, phy_ext_status;
1085 	u16 pci_status;
1086 
1087 	if (test_bit(__E1000_DOWN, &adapter->state))
1088 		return;
1089 
1090 	if (!adapter->tx_hang_recheck &&
1091 	    (adapter->flags2 & FLAG2_DMA_BURST)) {
1092 		/* May be block on write-back, flush and detect again
1093 		 * flush pending descriptor writebacks to memory
1094 		 */
1095 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1096 		/* execute the writes immediately */
1097 		e1e_flush();
1098 		/* Due to rare timing issues, write to TIDV again to ensure
1099 		 * the write is successful
1100 		 */
1101 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1102 		/* execute the writes immediately */
1103 		e1e_flush();
1104 		adapter->tx_hang_recheck = true;
1105 		return;
1106 	}
1107 	/* Real hang detected */
1108 	adapter->tx_hang_recheck = false;
1109 	netif_stop_queue(netdev);
1110 
1111 	e1e_rphy(hw, MII_BMSR, &phy_status);
1112 	e1e_rphy(hw, MII_STAT1000, &phy_1000t_status);
1113 	e1e_rphy(hw, MII_ESTATUS, &phy_ext_status);
1114 
1115 	pci_read_config_word(adapter->pdev, PCI_STATUS, &pci_status);
1116 
1117 	/* detected Hardware unit hang */
1118 	e_err("Detected Hardware Unit Hang:\n"
1119 	      "  TDH                  <%x>\n"
1120 	      "  TDT                  <%x>\n"
1121 	      "  next_to_use          <%x>\n"
1122 	      "  next_to_clean        <%x>\n"
1123 	      "buffer_info[next_to_clean]:\n"
1124 	      "  time_stamp           <%lx>\n"
1125 	      "  next_to_watch        <%x>\n"
1126 	      "  jiffies              <%lx>\n"
1127 	      "  next_to_watch.status <%x>\n"
1128 	      "MAC Status             <%x>\n"
1129 	      "PHY Status             <%x>\n"
1130 	      "PHY 1000BASE-T Status  <%x>\n"
1131 	      "PHY Extended Status    <%x>\n"
1132 	      "PCI Status             <%x>\n",
1133 	      readl(tx_ring->head),
1134 	      readl(tx_ring->tail),
1135 	      tx_ring->next_to_use,
1136 	      tx_ring->next_to_clean,
1137 	      tx_ring->buffer_info[eop].time_stamp,
1138 	      eop,
1139 	      jiffies,
1140 	      eop_desc->upper.fields.status,
1141 	      er32(STATUS),
1142 	      phy_status,
1143 	      phy_1000t_status,
1144 	      phy_ext_status,
1145 	      pci_status);
1146 
1147 	/* Suggest workaround for known h/w issue */
1148 	if ((hw->mac.type == e1000_pchlan) && (er32(CTRL) & E1000_CTRL_TFCE))
1149 		e_err("Try turning off Tx pause (flow control) via ethtool\n");
1150 }
1151 
1152 /**
1153  * e1000e_tx_hwtstamp_work - check for Tx time stamp
1154  * @work: pointer to work struct
1155  *
1156  * This work function polls the TSYNCTXCTL valid bit to determine when a
1157  * timestamp has been taken for the current stored skb.  The timestamp must
1158  * be for this skb because only one such packet is allowed in the queue.
1159  */
1160 static void e1000e_tx_hwtstamp_work(struct work_struct *work)
1161 {
1162 	struct e1000_adapter *adapter = container_of(work, struct e1000_adapter,
1163 						     tx_hwtstamp_work);
1164 	struct e1000_hw *hw = &adapter->hw;
1165 
1166 	if (!adapter->tx_hwtstamp_skb)
1167 		return;
1168 
1169 	if (er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID) {
1170 		struct skb_shared_hwtstamps shhwtstamps;
1171 		u64 txstmp;
1172 
1173 		txstmp = er32(TXSTMPL);
1174 		txstmp |= (u64)er32(TXSTMPH) << 32;
1175 
1176 		e1000e_systim_to_hwtstamp(adapter, &shhwtstamps, txstmp);
1177 
1178 		skb_tstamp_tx(adapter->tx_hwtstamp_skb, &shhwtstamps);
1179 		dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
1180 		adapter->tx_hwtstamp_skb = NULL;
1181 	} else {
1182 		/* reschedule to check later */
1183 		schedule_work(&adapter->tx_hwtstamp_work);
1184 	}
1185 }
1186 
1187 /**
1188  * e1000_clean_tx_irq - Reclaim resources after transmit completes
1189  * @tx_ring: Tx descriptor ring
1190  *
1191  * the return value indicates whether actual cleaning was done, there
1192  * is no guarantee that everything was cleaned
1193  **/
1194 static bool e1000_clean_tx_irq(struct e1000_ring *tx_ring)
1195 {
1196 	struct e1000_adapter *adapter = tx_ring->adapter;
1197 	struct net_device *netdev = adapter->netdev;
1198 	struct e1000_hw *hw = &adapter->hw;
1199 	struct e1000_tx_desc *tx_desc, *eop_desc;
1200 	struct e1000_buffer *buffer_info;
1201 	unsigned int i, eop;
1202 	unsigned int count = 0;
1203 	unsigned int total_tx_bytes = 0, total_tx_packets = 0;
1204 	unsigned int bytes_compl = 0, pkts_compl = 0;
1205 
1206 	i = tx_ring->next_to_clean;
1207 	eop = tx_ring->buffer_info[i].next_to_watch;
1208 	eop_desc = E1000_TX_DESC(*tx_ring, eop);
1209 
1210 	while ((eop_desc->upper.data & cpu_to_le32(E1000_TXD_STAT_DD)) &&
1211 	       (count < tx_ring->count)) {
1212 		bool cleaned = false;
1213 		rmb(); /* read buffer_info after eop_desc */
1214 		for (; !cleaned; count++) {
1215 			tx_desc = E1000_TX_DESC(*tx_ring, i);
1216 			buffer_info = &tx_ring->buffer_info[i];
1217 			cleaned = (i == eop);
1218 
1219 			if (cleaned) {
1220 				total_tx_packets += buffer_info->segs;
1221 				total_tx_bytes += buffer_info->bytecount;
1222 				if (buffer_info->skb) {
1223 					bytes_compl += buffer_info->skb->len;
1224 					pkts_compl++;
1225 				}
1226 			}
1227 
1228 			e1000_put_txbuf(tx_ring, buffer_info);
1229 			tx_desc->upper.data = 0;
1230 
1231 			i++;
1232 			if (i == tx_ring->count)
1233 				i = 0;
1234 		}
1235 
1236 		if (i == tx_ring->next_to_use)
1237 			break;
1238 		eop = tx_ring->buffer_info[i].next_to_watch;
1239 		eop_desc = E1000_TX_DESC(*tx_ring, eop);
1240 	}
1241 
1242 	tx_ring->next_to_clean = i;
1243 
1244 	netdev_completed_queue(netdev, pkts_compl, bytes_compl);
1245 
1246 #define TX_WAKE_THRESHOLD 32
1247 	if (count && netif_carrier_ok(netdev) &&
1248 	    e1000_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD) {
1249 		/* Make sure that anybody stopping the queue after this
1250 		 * sees the new next_to_clean.
1251 		 */
1252 		smp_mb();
1253 
1254 		if (netif_queue_stopped(netdev) &&
1255 		    !(test_bit(__E1000_DOWN, &adapter->state))) {
1256 			netif_wake_queue(netdev);
1257 			++adapter->restart_queue;
1258 		}
1259 	}
1260 
1261 	if (adapter->detect_tx_hung) {
1262 		/* Detect a transmit hang in hardware, this serializes the
1263 		 * check with the clearing of time_stamp and movement of i
1264 		 */
1265 		adapter->detect_tx_hung = false;
1266 		if (tx_ring->buffer_info[i].time_stamp &&
1267 		    time_after(jiffies, tx_ring->buffer_info[i].time_stamp
1268 			       + (adapter->tx_timeout_factor * HZ)) &&
1269 		    !(er32(STATUS) & E1000_STATUS_TXOFF))
1270 			schedule_work(&adapter->print_hang_task);
1271 		else
1272 			adapter->tx_hang_recheck = false;
1273 	}
1274 	adapter->total_tx_bytes += total_tx_bytes;
1275 	adapter->total_tx_packets += total_tx_packets;
1276 	return count < tx_ring->count;
1277 }
1278 
1279 /**
1280  * e1000_clean_rx_irq_ps - Send received data up the network stack; packet split
1281  * @rx_ring: Rx descriptor ring
1282  *
1283  * the return value indicates whether actual cleaning was done, there
1284  * is no guarantee that everything was cleaned
1285  **/
1286 static bool e1000_clean_rx_irq_ps(struct e1000_ring *rx_ring, int *work_done,
1287 				  int work_to_do)
1288 {
1289 	struct e1000_adapter *adapter = rx_ring->adapter;
1290 	struct e1000_hw *hw = &adapter->hw;
1291 	union e1000_rx_desc_packet_split *rx_desc, *next_rxd;
1292 	struct net_device *netdev = adapter->netdev;
1293 	struct pci_dev *pdev = adapter->pdev;
1294 	struct e1000_buffer *buffer_info, *next_buffer;
1295 	struct e1000_ps_page *ps_page;
1296 	struct sk_buff *skb;
1297 	unsigned int i, j;
1298 	u32 length, staterr;
1299 	int cleaned_count = 0;
1300 	bool cleaned = false;
1301 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1302 
1303 	i = rx_ring->next_to_clean;
1304 	rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
1305 	staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1306 	buffer_info = &rx_ring->buffer_info[i];
1307 
1308 	while (staterr & E1000_RXD_STAT_DD) {
1309 		if (*work_done >= work_to_do)
1310 			break;
1311 		(*work_done)++;
1312 		skb = buffer_info->skb;
1313 		rmb();	/* read descriptor and rx_buffer_info after status DD */
1314 
1315 		/* in the packet split case this is header only */
1316 		prefetch(skb->data - NET_IP_ALIGN);
1317 
1318 		i++;
1319 		if (i == rx_ring->count)
1320 			i = 0;
1321 		next_rxd = E1000_RX_DESC_PS(*rx_ring, i);
1322 		prefetch(next_rxd);
1323 
1324 		next_buffer = &rx_ring->buffer_info[i];
1325 
1326 		cleaned = true;
1327 		cleaned_count++;
1328 		dma_unmap_single(&pdev->dev, buffer_info->dma,
1329 				 adapter->rx_ps_bsize0, DMA_FROM_DEVICE);
1330 		buffer_info->dma = 0;
1331 
1332 		/* see !EOP comment in other Rx routine */
1333 		if (!(staterr & E1000_RXD_STAT_EOP))
1334 			adapter->flags2 |= FLAG2_IS_DISCARDING;
1335 
1336 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
1337 			e_dbg("Packet Split buffers didn't pick up the full packet\n");
1338 			dev_kfree_skb_irq(skb);
1339 			if (staterr & E1000_RXD_STAT_EOP)
1340 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1341 			goto next_desc;
1342 		}
1343 
1344 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1345 			     !(netdev->features & NETIF_F_RXALL))) {
1346 			dev_kfree_skb_irq(skb);
1347 			goto next_desc;
1348 		}
1349 
1350 		length = le16_to_cpu(rx_desc->wb.middle.length0);
1351 
1352 		if (!length) {
1353 			e_dbg("Last part of the packet spanning multiple descriptors\n");
1354 			dev_kfree_skb_irq(skb);
1355 			goto next_desc;
1356 		}
1357 
1358 		/* Good Receive */
1359 		skb_put(skb, length);
1360 
1361 		{
1362 			/* this looks ugly, but it seems compiler issues make
1363 			 * it more efficient than reusing j
1364 			 */
1365 			int l1 = le16_to_cpu(rx_desc->wb.upper.length[0]);
1366 
1367 			/* page alloc/put takes too long and effects small
1368 			 * packet throughput, so unsplit small packets and
1369 			 * save the alloc/put only valid in softirq (napi)
1370 			 * context to call kmap_*
1371 			 */
1372 			if (l1 && (l1 <= copybreak) &&
1373 			    ((length + l1) <= adapter->rx_ps_bsize0)) {
1374 				u8 *vaddr;
1375 
1376 				ps_page = &buffer_info->ps_pages[0];
1377 
1378 				/* there is no documentation about how to call
1379 				 * kmap_atomic, so we can't hold the mapping
1380 				 * very long
1381 				 */
1382 				dma_sync_single_for_cpu(&pdev->dev,
1383 							ps_page->dma,
1384 							PAGE_SIZE,
1385 							DMA_FROM_DEVICE);
1386 				vaddr = kmap_atomic(ps_page->page);
1387 				memcpy(skb_tail_pointer(skb), vaddr, l1);
1388 				kunmap_atomic(vaddr);
1389 				dma_sync_single_for_device(&pdev->dev,
1390 							   ps_page->dma,
1391 							   PAGE_SIZE,
1392 							   DMA_FROM_DEVICE);
1393 
1394 				/* remove the CRC */
1395 				if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1396 					if (!(netdev->features & NETIF_F_RXFCS))
1397 						l1 -= 4;
1398 				}
1399 
1400 				skb_put(skb, l1);
1401 				goto copydone;
1402 			} /* if */
1403 		}
1404 
1405 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1406 			length = le16_to_cpu(rx_desc->wb.upper.length[j]);
1407 			if (!length)
1408 				break;
1409 
1410 			ps_page = &buffer_info->ps_pages[j];
1411 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1412 				       DMA_FROM_DEVICE);
1413 			ps_page->dma = 0;
1414 			skb_fill_page_desc(skb, j, ps_page->page, 0, length);
1415 			ps_page->page = NULL;
1416 			skb->len += length;
1417 			skb->data_len += length;
1418 			skb->truesize += PAGE_SIZE;
1419 		}
1420 
1421 		/* strip the ethernet crc, problem is we're using pages now so
1422 		 * this whole operation can get a little cpu intensive
1423 		 */
1424 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1425 			if (!(netdev->features & NETIF_F_RXFCS))
1426 				pskb_trim(skb, skb->len - 4);
1427 		}
1428 
1429 copydone:
1430 		total_rx_bytes += skb->len;
1431 		total_rx_packets++;
1432 
1433 		e1000_rx_checksum(adapter, staterr, skb);
1434 
1435 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1436 
1437 		if (rx_desc->wb.upper.header_status &
1438 			   cpu_to_le16(E1000_RXDPS_HDRSTAT_HDRSP))
1439 			adapter->rx_hdr_split++;
1440 
1441 		e1000_receive_skb(adapter, netdev, skb, staterr,
1442 				  rx_desc->wb.middle.vlan);
1443 
1444 next_desc:
1445 		rx_desc->wb.middle.status_error &= cpu_to_le32(~0xFF);
1446 		buffer_info->skb = NULL;
1447 
1448 		/* return some buffers to hardware, one at a time is too slow */
1449 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1450 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1451 					      GFP_ATOMIC);
1452 			cleaned_count = 0;
1453 		}
1454 
1455 		/* use prefetched values */
1456 		rx_desc = next_rxd;
1457 		buffer_info = next_buffer;
1458 
1459 		staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1460 	}
1461 	rx_ring->next_to_clean = i;
1462 
1463 	cleaned_count = e1000_desc_unused(rx_ring);
1464 	if (cleaned_count)
1465 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1466 
1467 	adapter->total_rx_bytes += total_rx_bytes;
1468 	adapter->total_rx_packets += total_rx_packets;
1469 	return cleaned;
1470 }
1471 
1472 /**
1473  * e1000_consume_page - helper function
1474  **/
1475 static void e1000_consume_page(struct e1000_buffer *bi, struct sk_buff *skb,
1476                                u16 length)
1477 {
1478 	bi->page = NULL;
1479 	skb->len += length;
1480 	skb->data_len += length;
1481 	skb->truesize += PAGE_SIZE;
1482 }
1483 
1484 /**
1485  * e1000_clean_jumbo_rx_irq - Send received data up the network stack; legacy
1486  * @adapter: board private structure
1487  *
1488  * the return value indicates whether actual cleaning was done, there
1489  * is no guarantee that everything was cleaned
1490  **/
1491 static bool e1000_clean_jumbo_rx_irq(struct e1000_ring *rx_ring, int *work_done,
1492 				     int work_to_do)
1493 {
1494 	struct e1000_adapter *adapter = rx_ring->adapter;
1495 	struct net_device *netdev = adapter->netdev;
1496 	struct pci_dev *pdev = adapter->pdev;
1497 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
1498 	struct e1000_buffer *buffer_info, *next_buffer;
1499 	u32 length, staterr;
1500 	unsigned int i;
1501 	int cleaned_count = 0;
1502 	bool cleaned = false;
1503 	unsigned int total_rx_bytes=0, total_rx_packets=0;
1504 
1505 	i = rx_ring->next_to_clean;
1506 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
1507 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1508 	buffer_info = &rx_ring->buffer_info[i];
1509 
1510 	while (staterr & E1000_RXD_STAT_DD) {
1511 		struct sk_buff *skb;
1512 
1513 		if (*work_done >= work_to_do)
1514 			break;
1515 		(*work_done)++;
1516 		rmb();	/* read descriptor and rx_buffer_info after status DD */
1517 
1518 		skb = buffer_info->skb;
1519 		buffer_info->skb = NULL;
1520 
1521 		++i;
1522 		if (i == rx_ring->count)
1523 			i = 0;
1524 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
1525 		prefetch(next_rxd);
1526 
1527 		next_buffer = &rx_ring->buffer_info[i];
1528 
1529 		cleaned = true;
1530 		cleaned_count++;
1531 		dma_unmap_page(&pdev->dev, buffer_info->dma, PAGE_SIZE,
1532 			       DMA_FROM_DEVICE);
1533 		buffer_info->dma = 0;
1534 
1535 		length = le16_to_cpu(rx_desc->wb.upper.length);
1536 
1537 		/* errors is only valid for DD + EOP descriptors */
1538 		if (unlikely((staterr & E1000_RXD_STAT_EOP) &&
1539 			     ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1540 			      !(netdev->features & NETIF_F_RXALL)))) {
1541 			/* recycle both page and skb */
1542 			buffer_info->skb = skb;
1543 			/* an error means any chain goes out the window too */
1544 			if (rx_ring->rx_skb_top)
1545 				dev_kfree_skb_irq(rx_ring->rx_skb_top);
1546 			rx_ring->rx_skb_top = NULL;
1547 			goto next_desc;
1548 		}
1549 
1550 #define rxtop (rx_ring->rx_skb_top)
1551 		if (!(staterr & E1000_RXD_STAT_EOP)) {
1552 			/* this descriptor is only the beginning (or middle) */
1553 			if (!rxtop) {
1554 				/* this is the beginning of a chain */
1555 				rxtop = skb;
1556 				skb_fill_page_desc(rxtop, 0, buffer_info->page,
1557 				                   0, length);
1558 			} else {
1559 				/* this is the middle of a chain */
1560 				skb_fill_page_desc(rxtop,
1561 				    skb_shinfo(rxtop)->nr_frags,
1562 				    buffer_info->page, 0, length);
1563 				/* re-use the skb, only consumed the page */
1564 				buffer_info->skb = skb;
1565 			}
1566 			e1000_consume_page(buffer_info, rxtop, length);
1567 			goto next_desc;
1568 		} else {
1569 			if (rxtop) {
1570 				/* end of the chain */
1571 				skb_fill_page_desc(rxtop,
1572 				    skb_shinfo(rxtop)->nr_frags,
1573 				    buffer_info->page, 0, length);
1574 				/* re-use the current skb, we only consumed the
1575 				 * page
1576 				 */
1577 				buffer_info->skb = skb;
1578 				skb = rxtop;
1579 				rxtop = NULL;
1580 				e1000_consume_page(buffer_info, skb, length);
1581 			} else {
1582 				/* no chain, got EOP, this buf is the packet
1583 				 * copybreak to save the put_page/alloc_page
1584 				 */
1585 				if (length <= copybreak &&
1586 				    skb_tailroom(skb) >= length) {
1587 					u8 *vaddr;
1588 					vaddr = kmap_atomic(buffer_info->page);
1589 					memcpy(skb_tail_pointer(skb), vaddr,
1590 					       length);
1591 					kunmap_atomic(vaddr);
1592 					/* re-use the page, so don't erase
1593 					 * buffer_info->page
1594 					 */
1595 					skb_put(skb, length);
1596 				} else {
1597 					skb_fill_page_desc(skb, 0,
1598 					                   buffer_info->page, 0,
1599 				                           length);
1600 					e1000_consume_page(buffer_info, skb,
1601 					                   length);
1602 				}
1603 			}
1604 		}
1605 
1606 		/* Receive Checksum Offload */
1607 		e1000_rx_checksum(adapter, staterr, skb);
1608 
1609 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1610 
1611 		/* probably a little skewed due to removing CRC */
1612 		total_rx_bytes += skb->len;
1613 		total_rx_packets++;
1614 
1615 		/* eth type trans needs skb->data to point to something */
1616 		if (!pskb_may_pull(skb, ETH_HLEN)) {
1617 			e_err("pskb_may_pull failed.\n");
1618 			dev_kfree_skb_irq(skb);
1619 			goto next_desc;
1620 		}
1621 
1622 		e1000_receive_skb(adapter, netdev, skb, staterr,
1623 				  rx_desc->wb.upper.vlan);
1624 
1625 next_desc:
1626 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1627 
1628 		/* return some buffers to hardware, one at a time is too slow */
1629 		if (unlikely(cleaned_count >= E1000_RX_BUFFER_WRITE)) {
1630 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1631 					      GFP_ATOMIC);
1632 			cleaned_count = 0;
1633 		}
1634 
1635 		/* use prefetched values */
1636 		rx_desc = next_rxd;
1637 		buffer_info = next_buffer;
1638 
1639 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1640 	}
1641 	rx_ring->next_to_clean = i;
1642 
1643 	cleaned_count = e1000_desc_unused(rx_ring);
1644 	if (cleaned_count)
1645 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1646 
1647 	adapter->total_rx_bytes += total_rx_bytes;
1648 	adapter->total_rx_packets += total_rx_packets;
1649 	return cleaned;
1650 }
1651 
1652 /**
1653  * e1000_clean_rx_ring - Free Rx Buffers per Queue
1654  * @rx_ring: Rx descriptor ring
1655  **/
1656 static void e1000_clean_rx_ring(struct e1000_ring *rx_ring)
1657 {
1658 	struct e1000_adapter *adapter = rx_ring->adapter;
1659 	struct e1000_buffer *buffer_info;
1660 	struct e1000_ps_page *ps_page;
1661 	struct pci_dev *pdev = adapter->pdev;
1662 	unsigned int i, j;
1663 
1664 	/* Free all the Rx ring sk_buffs */
1665 	for (i = 0; i < rx_ring->count; i++) {
1666 		buffer_info = &rx_ring->buffer_info[i];
1667 		if (buffer_info->dma) {
1668 			if (adapter->clean_rx == e1000_clean_rx_irq)
1669 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1670 						 adapter->rx_buffer_len,
1671 						 DMA_FROM_DEVICE);
1672 			else if (adapter->clean_rx == e1000_clean_jumbo_rx_irq)
1673 				dma_unmap_page(&pdev->dev, buffer_info->dma,
1674 				               PAGE_SIZE,
1675 					       DMA_FROM_DEVICE);
1676 			else if (adapter->clean_rx == e1000_clean_rx_irq_ps)
1677 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1678 						 adapter->rx_ps_bsize0,
1679 						 DMA_FROM_DEVICE);
1680 			buffer_info->dma = 0;
1681 		}
1682 
1683 		if (buffer_info->page) {
1684 			put_page(buffer_info->page);
1685 			buffer_info->page = NULL;
1686 		}
1687 
1688 		if (buffer_info->skb) {
1689 			dev_kfree_skb(buffer_info->skb);
1690 			buffer_info->skb = NULL;
1691 		}
1692 
1693 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1694 			ps_page = &buffer_info->ps_pages[j];
1695 			if (!ps_page->page)
1696 				break;
1697 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1698 				       DMA_FROM_DEVICE);
1699 			ps_page->dma = 0;
1700 			put_page(ps_page->page);
1701 			ps_page->page = NULL;
1702 		}
1703 	}
1704 
1705 	/* there also may be some cached data from a chained receive */
1706 	if (rx_ring->rx_skb_top) {
1707 		dev_kfree_skb(rx_ring->rx_skb_top);
1708 		rx_ring->rx_skb_top = NULL;
1709 	}
1710 
1711 	/* Zero out the descriptor ring */
1712 	memset(rx_ring->desc, 0, rx_ring->size);
1713 
1714 	rx_ring->next_to_clean = 0;
1715 	rx_ring->next_to_use = 0;
1716 	adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1717 
1718 	writel(0, rx_ring->head);
1719 	if (rx_ring->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
1720 		e1000e_update_rdt_wa(rx_ring, 0);
1721 	else
1722 		writel(0, rx_ring->tail);
1723 }
1724 
1725 static void e1000e_downshift_workaround(struct work_struct *work)
1726 {
1727 	struct e1000_adapter *adapter = container_of(work,
1728 					struct e1000_adapter, downshift_task);
1729 
1730 	if (test_bit(__E1000_DOWN, &adapter->state))
1731 		return;
1732 
1733 	e1000e_gig_downshift_workaround_ich8lan(&adapter->hw);
1734 }
1735 
1736 /**
1737  * e1000_intr_msi - Interrupt Handler
1738  * @irq: interrupt number
1739  * @data: pointer to a network interface device structure
1740  **/
1741 static irqreturn_t e1000_intr_msi(int __always_unused irq, void *data)
1742 {
1743 	struct net_device *netdev = data;
1744 	struct e1000_adapter *adapter = netdev_priv(netdev);
1745 	struct e1000_hw *hw = &adapter->hw;
1746 	u32 icr = er32(ICR);
1747 
1748 	/* read ICR disables interrupts using IAM */
1749 	if (icr & E1000_ICR_LSC) {
1750 		hw->mac.get_link_status = true;
1751 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1752 		 * disconnect (LSC) before accessing any PHY registers
1753 		 */
1754 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1755 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1756 			schedule_work(&adapter->downshift_task);
1757 
1758 		/* 80003ES2LAN workaround-- For packet buffer work-around on
1759 		 * link down event; disable receives here in the ISR and reset
1760 		 * adapter in watchdog
1761 		 */
1762 		if (netif_carrier_ok(netdev) &&
1763 		    adapter->flags & FLAG_RX_NEEDS_RESTART) {
1764 			/* disable receives */
1765 			u32 rctl = er32(RCTL);
1766 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1767 			adapter->flags |= FLAG_RESTART_NOW;
1768 		}
1769 		/* guard against interrupt when we're going down */
1770 		if (!test_bit(__E1000_DOWN, &adapter->state))
1771 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1772 	}
1773 
1774 	/* Reset on uncorrectable ECC error */
1775 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type == e1000_pch_lpt)) {
1776 		u32 pbeccsts = er32(PBECCSTS);
1777 
1778 		adapter->corr_errors +=
1779 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1780 		adapter->uncorr_errors +=
1781 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1782 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1783 
1784 		/* Do the reset outside of interrupt context */
1785 		schedule_work(&adapter->reset_task);
1786 
1787 		/* return immediately since reset is imminent */
1788 		return IRQ_HANDLED;
1789 	}
1790 
1791 	if (napi_schedule_prep(&adapter->napi)) {
1792 		adapter->total_tx_bytes = 0;
1793 		adapter->total_tx_packets = 0;
1794 		adapter->total_rx_bytes = 0;
1795 		adapter->total_rx_packets = 0;
1796 		__napi_schedule(&adapter->napi);
1797 	}
1798 
1799 	return IRQ_HANDLED;
1800 }
1801 
1802 /**
1803  * e1000_intr - Interrupt Handler
1804  * @irq: interrupt number
1805  * @data: pointer to a network interface device structure
1806  **/
1807 static irqreturn_t e1000_intr(int __always_unused irq, void *data)
1808 {
1809 	struct net_device *netdev = data;
1810 	struct e1000_adapter *adapter = netdev_priv(netdev);
1811 	struct e1000_hw *hw = &adapter->hw;
1812 	u32 rctl, icr = er32(ICR);
1813 
1814 	if (!icr || test_bit(__E1000_DOWN, &adapter->state))
1815 		return IRQ_NONE;  /* Not our interrupt */
1816 
1817 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
1818 	 * not set, then the adapter didn't send an interrupt
1819 	 */
1820 	if (!(icr & E1000_ICR_INT_ASSERTED))
1821 		return IRQ_NONE;
1822 
1823 	/* Interrupt Auto-Mask...upon reading ICR,
1824 	 * interrupts are masked.  No need for the
1825 	 * IMC write
1826 	 */
1827 
1828 	if (icr & E1000_ICR_LSC) {
1829 		hw->mac.get_link_status = true;
1830 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1831 		 * disconnect (LSC) before accessing any PHY registers
1832 		 */
1833 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1834 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1835 			schedule_work(&adapter->downshift_task);
1836 
1837 		/* 80003ES2LAN workaround--
1838 		 * For packet buffer work-around on link down event;
1839 		 * disable receives here in the ISR and
1840 		 * reset adapter in watchdog
1841 		 */
1842 		if (netif_carrier_ok(netdev) &&
1843 		    (adapter->flags & FLAG_RX_NEEDS_RESTART)) {
1844 			/* disable receives */
1845 			rctl = er32(RCTL);
1846 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1847 			adapter->flags |= FLAG_RESTART_NOW;
1848 		}
1849 		/* guard against interrupt when we're going down */
1850 		if (!test_bit(__E1000_DOWN, &adapter->state))
1851 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1852 	}
1853 
1854 	/* Reset on uncorrectable ECC error */
1855 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type == e1000_pch_lpt)) {
1856 		u32 pbeccsts = er32(PBECCSTS);
1857 
1858 		adapter->corr_errors +=
1859 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1860 		adapter->uncorr_errors +=
1861 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1862 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1863 
1864 		/* Do the reset outside of interrupt context */
1865 		schedule_work(&adapter->reset_task);
1866 
1867 		/* return immediately since reset is imminent */
1868 		return IRQ_HANDLED;
1869 	}
1870 
1871 	if (napi_schedule_prep(&adapter->napi)) {
1872 		adapter->total_tx_bytes = 0;
1873 		adapter->total_tx_packets = 0;
1874 		adapter->total_rx_bytes = 0;
1875 		adapter->total_rx_packets = 0;
1876 		__napi_schedule(&adapter->napi);
1877 	}
1878 
1879 	return IRQ_HANDLED;
1880 }
1881 
1882 static irqreturn_t e1000_msix_other(int __always_unused irq, void *data)
1883 {
1884 	struct net_device *netdev = data;
1885 	struct e1000_adapter *adapter = netdev_priv(netdev);
1886 	struct e1000_hw *hw = &adapter->hw;
1887 	u32 icr = er32(ICR);
1888 
1889 	if (!(icr & E1000_ICR_INT_ASSERTED)) {
1890 		if (!test_bit(__E1000_DOWN, &adapter->state))
1891 			ew32(IMS, E1000_IMS_OTHER);
1892 		return IRQ_NONE;
1893 	}
1894 
1895 	if (icr & adapter->eiac_mask)
1896 		ew32(ICS, (icr & adapter->eiac_mask));
1897 
1898 	if (icr & E1000_ICR_OTHER) {
1899 		if (!(icr & E1000_ICR_LSC))
1900 			goto no_link_interrupt;
1901 		hw->mac.get_link_status = true;
1902 		/* guard against interrupt when we're going down */
1903 		if (!test_bit(__E1000_DOWN, &adapter->state))
1904 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1905 	}
1906 
1907 no_link_interrupt:
1908 	if (!test_bit(__E1000_DOWN, &adapter->state))
1909 		ew32(IMS, E1000_IMS_LSC | E1000_IMS_OTHER);
1910 
1911 	return IRQ_HANDLED;
1912 }
1913 
1914 static irqreturn_t e1000_intr_msix_tx(int __always_unused irq, void *data)
1915 {
1916 	struct net_device *netdev = data;
1917 	struct e1000_adapter *adapter = netdev_priv(netdev);
1918 	struct e1000_hw *hw = &adapter->hw;
1919 	struct e1000_ring *tx_ring = adapter->tx_ring;
1920 
1921 
1922 	adapter->total_tx_bytes = 0;
1923 	adapter->total_tx_packets = 0;
1924 
1925 	if (!e1000_clean_tx_irq(tx_ring))
1926 		/* Ring was not completely cleaned, so fire another interrupt */
1927 		ew32(ICS, tx_ring->ims_val);
1928 
1929 	return IRQ_HANDLED;
1930 }
1931 
1932 static irqreturn_t e1000_intr_msix_rx(int __always_unused irq, void *data)
1933 {
1934 	struct net_device *netdev = data;
1935 	struct e1000_adapter *adapter = netdev_priv(netdev);
1936 	struct e1000_ring *rx_ring = adapter->rx_ring;
1937 
1938 	/* Write the ITR value calculated at the end of the
1939 	 * previous interrupt.
1940 	 */
1941 	if (rx_ring->set_itr) {
1942 		writel(1000000000 / (rx_ring->itr_val * 256),
1943 		       rx_ring->itr_register);
1944 		rx_ring->set_itr = 0;
1945 	}
1946 
1947 	if (napi_schedule_prep(&adapter->napi)) {
1948 		adapter->total_rx_bytes = 0;
1949 		adapter->total_rx_packets = 0;
1950 		__napi_schedule(&adapter->napi);
1951 	}
1952 	return IRQ_HANDLED;
1953 }
1954 
1955 /**
1956  * e1000_configure_msix - Configure MSI-X hardware
1957  *
1958  * e1000_configure_msix sets up the hardware to properly
1959  * generate MSI-X interrupts.
1960  **/
1961 static void e1000_configure_msix(struct e1000_adapter *adapter)
1962 {
1963 	struct e1000_hw *hw = &adapter->hw;
1964 	struct e1000_ring *rx_ring = adapter->rx_ring;
1965 	struct e1000_ring *tx_ring = adapter->tx_ring;
1966 	int vector = 0;
1967 	u32 ctrl_ext, ivar = 0;
1968 
1969 	adapter->eiac_mask = 0;
1970 
1971 	/* Workaround issue with spurious interrupts on 82574 in MSI-X mode */
1972 	if (hw->mac.type == e1000_82574) {
1973 		u32 rfctl = er32(RFCTL);
1974 		rfctl |= E1000_RFCTL_ACK_DIS;
1975 		ew32(RFCTL, rfctl);
1976 	}
1977 
1978 #define E1000_IVAR_INT_ALLOC_VALID	0x8
1979 	/* Configure Rx vector */
1980 	rx_ring->ims_val = E1000_IMS_RXQ0;
1981 	adapter->eiac_mask |= rx_ring->ims_val;
1982 	if (rx_ring->itr_val)
1983 		writel(1000000000 / (rx_ring->itr_val * 256),
1984 		       rx_ring->itr_register);
1985 	else
1986 		writel(1, rx_ring->itr_register);
1987 	ivar = E1000_IVAR_INT_ALLOC_VALID | vector;
1988 
1989 	/* Configure Tx vector */
1990 	tx_ring->ims_val = E1000_IMS_TXQ0;
1991 	vector++;
1992 	if (tx_ring->itr_val)
1993 		writel(1000000000 / (tx_ring->itr_val * 256),
1994 		       tx_ring->itr_register);
1995 	else
1996 		writel(1, tx_ring->itr_register);
1997 	adapter->eiac_mask |= tx_ring->ims_val;
1998 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 8);
1999 
2000 	/* set vector for Other Causes, e.g. link changes */
2001 	vector++;
2002 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 16);
2003 	if (rx_ring->itr_val)
2004 		writel(1000000000 / (rx_ring->itr_val * 256),
2005 		       hw->hw_addr + E1000_EITR_82574(vector));
2006 	else
2007 		writel(1, hw->hw_addr + E1000_EITR_82574(vector));
2008 
2009 	/* Cause Tx interrupts on every write back */
2010 	ivar |= (1 << 31);
2011 
2012 	ew32(IVAR, ivar);
2013 
2014 	/* enable MSI-X PBA support */
2015 	ctrl_ext = er32(CTRL_EXT);
2016 	ctrl_ext |= E1000_CTRL_EXT_PBA_CLR;
2017 
2018 	/* Auto-Mask Other interrupts upon ICR read */
2019 	ew32(IAM, ~E1000_EIAC_MASK_82574 | E1000_IMS_OTHER);
2020 	ctrl_ext |= E1000_CTRL_EXT_EIAME;
2021 	ew32(CTRL_EXT, ctrl_ext);
2022 	e1e_flush();
2023 }
2024 
2025 void e1000e_reset_interrupt_capability(struct e1000_adapter *adapter)
2026 {
2027 	if (adapter->msix_entries) {
2028 		pci_disable_msix(adapter->pdev);
2029 		kfree(adapter->msix_entries);
2030 		adapter->msix_entries = NULL;
2031 	} else if (adapter->flags & FLAG_MSI_ENABLED) {
2032 		pci_disable_msi(adapter->pdev);
2033 		adapter->flags &= ~FLAG_MSI_ENABLED;
2034 	}
2035 }
2036 
2037 /**
2038  * e1000e_set_interrupt_capability - set MSI or MSI-X if supported
2039  *
2040  * Attempt to configure interrupts using the best available
2041  * capabilities of the hardware and kernel.
2042  **/
2043 void e1000e_set_interrupt_capability(struct e1000_adapter *adapter)
2044 {
2045 	int err;
2046 	int i;
2047 
2048 	switch (adapter->int_mode) {
2049 	case E1000E_INT_MODE_MSIX:
2050 		if (adapter->flags & FLAG_HAS_MSIX) {
2051 			adapter->num_vectors = 3; /* RxQ0, TxQ0 and other */
2052 			adapter->msix_entries = kcalloc(adapter->num_vectors,
2053 						      sizeof(struct msix_entry),
2054 						      GFP_KERNEL);
2055 			if (adapter->msix_entries) {
2056 				for (i = 0; i < adapter->num_vectors; i++)
2057 					adapter->msix_entries[i].entry = i;
2058 
2059 				err = pci_enable_msix(adapter->pdev,
2060 						      adapter->msix_entries,
2061 						      adapter->num_vectors);
2062 				if (err == 0)
2063 					return;
2064 			}
2065 			/* MSI-X failed, so fall through and try MSI */
2066 			e_err("Failed to initialize MSI-X interrupts.  Falling back to MSI interrupts.\n");
2067 			e1000e_reset_interrupt_capability(adapter);
2068 		}
2069 		adapter->int_mode = E1000E_INT_MODE_MSI;
2070 		/* Fall through */
2071 	case E1000E_INT_MODE_MSI:
2072 		if (!pci_enable_msi(adapter->pdev)) {
2073 			adapter->flags |= FLAG_MSI_ENABLED;
2074 		} else {
2075 			adapter->int_mode = E1000E_INT_MODE_LEGACY;
2076 			e_err("Failed to initialize MSI interrupts.  Falling back to legacy interrupts.\n");
2077 		}
2078 		/* Fall through */
2079 	case E1000E_INT_MODE_LEGACY:
2080 		/* Don't do anything; this is the system default */
2081 		break;
2082 	}
2083 
2084 	/* store the number of vectors being used */
2085 	adapter->num_vectors = 1;
2086 }
2087 
2088 /**
2089  * e1000_request_msix - Initialize MSI-X interrupts
2090  *
2091  * e1000_request_msix allocates MSI-X vectors and requests interrupts from the
2092  * kernel.
2093  **/
2094 static int e1000_request_msix(struct e1000_adapter *adapter)
2095 {
2096 	struct net_device *netdev = adapter->netdev;
2097 	int err = 0, vector = 0;
2098 
2099 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2100 		snprintf(adapter->rx_ring->name,
2101 			 sizeof(adapter->rx_ring->name) - 1,
2102 			 "%s-rx-0", netdev->name);
2103 	else
2104 		memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
2105 	err = request_irq(adapter->msix_entries[vector].vector,
2106 			  e1000_intr_msix_rx, 0, adapter->rx_ring->name,
2107 			  netdev);
2108 	if (err)
2109 		return err;
2110 	adapter->rx_ring->itr_register = adapter->hw.hw_addr +
2111 	    E1000_EITR_82574(vector);
2112 	adapter->rx_ring->itr_val = adapter->itr;
2113 	vector++;
2114 
2115 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2116 		snprintf(adapter->tx_ring->name,
2117 			 sizeof(adapter->tx_ring->name) - 1,
2118 			 "%s-tx-0", netdev->name);
2119 	else
2120 		memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
2121 	err = request_irq(adapter->msix_entries[vector].vector,
2122 			  e1000_intr_msix_tx, 0, adapter->tx_ring->name,
2123 			  netdev);
2124 	if (err)
2125 		return err;
2126 	adapter->tx_ring->itr_register = adapter->hw.hw_addr +
2127 	    E1000_EITR_82574(vector);
2128 	adapter->tx_ring->itr_val = adapter->itr;
2129 	vector++;
2130 
2131 	err = request_irq(adapter->msix_entries[vector].vector,
2132 			  e1000_msix_other, 0, netdev->name, netdev);
2133 	if (err)
2134 		return err;
2135 
2136 	e1000_configure_msix(adapter);
2137 
2138 	return 0;
2139 }
2140 
2141 /**
2142  * e1000_request_irq - initialize interrupts
2143  *
2144  * Attempts to configure interrupts using the best available
2145  * capabilities of the hardware and kernel.
2146  **/
2147 static int e1000_request_irq(struct e1000_adapter *adapter)
2148 {
2149 	struct net_device *netdev = adapter->netdev;
2150 	int err;
2151 
2152 	if (adapter->msix_entries) {
2153 		err = e1000_request_msix(adapter);
2154 		if (!err)
2155 			return err;
2156 		/* fall back to MSI */
2157 		e1000e_reset_interrupt_capability(adapter);
2158 		adapter->int_mode = E1000E_INT_MODE_MSI;
2159 		e1000e_set_interrupt_capability(adapter);
2160 	}
2161 	if (adapter->flags & FLAG_MSI_ENABLED) {
2162 		err = request_irq(adapter->pdev->irq, e1000_intr_msi, 0,
2163 				  netdev->name, netdev);
2164 		if (!err)
2165 			return err;
2166 
2167 		/* fall back to legacy interrupt */
2168 		e1000e_reset_interrupt_capability(adapter);
2169 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
2170 	}
2171 
2172 	err = request_irq(adapter->pdev->irq, e1000_intr, IRQF_SHARED,
2173 			  netdev->name, netdev);
2174 	if (err)
2175 		e_err("Unable to allocate interrupt, Error: %d\n", err);
2176 
2177 	return err;
2178 }
2179 
2180 static void e1000_free_irq(struct e1000_adapter *adapter)
2181 {
2182 	struct net_device *netdev = adapter->netdev;
2183 
2184 	if (adapter->msix_entries) {
2185 		int vector = 0;
2186 
2187 		free_irq(adapter->msix_entries[vector].vector, netdev);
2188 		vector++;
2189 
2190 		free_irq(adapter->msix_entries[vector].vector, netdev);
2191 		vector++;
2192 
2193 		/* Other Causes interrupt vector */
2194 		free_irq(adapter->msix_entries[vector].vector, netdev);
2195 		return;
2196 	}
2197 
2198 	free_irq(adapter->pdev->irq, netdev);
2199 }
2200 
2201 /**
2202  * e1000_irq_disable - Mask off interrupt generation on the NIC
2203  **/
2204 static void e1000_irq_disable(struct e1000_adapter *adapter)
2205 {
2206 	struct e1000_hw *hw = &adapter->hw;
2207 
2208 	ew32(IMC, ~0);
2209 	if (adapter->msix_entries)
2210 		ew32(EIAC_82574, 0);
2211 	e1e_flush();
2212 
2213 	if (adapter->msix_entries) {
2214 		int i;
2215 		for (i = 0; i < adapter->num_vectors; i++)
2216 			synchronize_irq(adapter->msix_entries[i].vector);
2217 	} else {
2218 		synchronize_irq(adapter->pdev->irq);
2219 	}
2220 }
2221 
2222 /**
2223  * e1000_irq_enable - Enable default interrupt generation settings
2224  **/
2225 static void e1000_irq_enable(struct e1000_adapter *adapter)
2226 {
2227 	struct e1000_hw *hw = &adapter->hw;
2228 
2229 	if (adapter->msix_entries) {
2230 		ew32(EIAC_82574, adapter->eiac_mask & E1000_EIAC_MASK_82574);
2231 		ew32(IMS, adapter->eiac_mask | E1000_IMS_OTHER | E1000_IMS_LSC);
2232 	} else if (hw->mac.type == e1000_pch_lpt) {
2233 		ew32(IMS, IMS_ENABLE_MASK | E1000_IMS_ECCER);
2234 	} else {
2235 		ew32(IMS, IMS_ENABLE_MASK);
2236 	}
2237 	e1e_flush();
2238 }
2239 
2240 /**
2241  * e1000e_get_hw_control - get control of the h/w from f/w
2242  * @adapter: address of board private structure
2243  *
2244  * e1000e_get_hw_control sets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2245  * For ASF and Pass Through versions of f/w this means that
2246  * the driver is loaded. For AMT version (only with 82573)
2247  * of the f/w this means that the network i/f is open.
2248  **/
2249 void e1000e_get_hw_control(struct e1000_adapter *adapter)
2250 {
2251 	struct e1000_hw *hw = &adapter->hw;
2252 	u32 ctrl_ext;
2253 	u32 swsm;
2254 
2255 	/* Let firmware know the driver has taken over */
2256 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2257 		swsm = er32(SWSM);
2258 		ew32(SWSM, swsm | E1000_SWSM_DRV_LOAD);
2259 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2260 		ctrl_ext = er32(CTRL_EXT);
2261 		ew32(CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2262 	}
2263 }
2264 
2265 /**
2266  * e1000e_release_hw_control - release control of the h/w to f/w
2267  * @adapter: address of board private structure
2268  *
2269  * e1000e_release_hw_control resets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2270  * For ASF and Pass Through versions of f/w this means that the
2271  * driver is no longer loaded. For AMT version (only with 82573) i
2272  * of the f/w this means that the network i/f is closed.
2273  *
2274  **/
2275 void e1000e_release_hw_control(struct e1000_adapter *adapter)
2276 {
2277 	struct e1000_hw *hw = &adapter->hw;
2278 	u32 ctrl_ext;
2279 	u32 swsm;
2280 
2281 	/* Let firmware taken over control of h/w */
2282 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2283 		swsm = er32(SWSM);
2284 		ew32(SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
2285 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2286 		ctrl_ext = er32(CTRL_EXT);
2287 		ew32(CTRL_EXT, ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2288 	}
2289 }
2290 
2291 /**
2292  * e1000_alloc_ring_dma - allocate memory for a ring structure
2293  **/
2294 static int e1000_alloc_ring_dma(struct e1000_adapter *adapter,
2295 				struct e1000_ring *ring)
2296 {
2297 	struct pci_dev *pdev = adapter->pdev;
2298 
2299 	ring->desc = dma_alloc_coherent(&pdev->dev, ring->size, &ring->dma,
2300 					GFP_KERNEL);
2301 	if (!ring->desc)
2302 		return -ENOMEM;
2303 
2304 	return 0;
2305 }
2306 
2307 /**
2308  * e1000e_setup_tx_resources - allocate Tx resources (Descriptors)
2309  * @tx_ring: Tx descriptor ring
2310  *
2311  * Return 0 on success, negative on failure
2312  **/
2313 int e1000e_setup_tx_resources(struct e1000_ring *tx_ring)
2314 {
2315 	struct e1000_adapter *adapter = tx_ring->adapter;
2316 	int err = -ENOMEM, size;
2317 
2318 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2319 	tx_ring->buffer_info = vzalloc(size);
2320 	if (!tx_ring->buffer_info)
2321 		goto err;
2322 
2323 	/* round up to nearest 4K */
2324 	tx_ring->size = tx_ring->count * sizeof(struct e1000_tx_desc);
2325 	tx_ring->size = ALIGN(tx_ring->size, 4096);
2326 
2327 	err = e1000_alloc_ring_dma(adapter, tx_ring);
2328 	if (err)
2329 		goto err;
2330 
2331 	tx_ring->next_to_use = 0;
2332 	tx_ring->next_to_clean = 0;
2333 
2334 	return 0;
2335 err:
2336 	vfree(tx_ring->buffer_info);
2337 	e_err("Unable to allocate memory for the transmit descriptor ring\n");
2338 	return err;
2339 }
2340 
2341 /**
2342  * e1000e_setup_rx_resources - allocate Rx resources (Descriptors)
2343  * @rx_ring: Rx descriptor ring
2344  *
2345  * Returns 0 on success, negative on failure
2346  **/
2347 int e1000e_setup_rx_resources(struct e1000_ring *rx_ring)
2348 {
2349 	struct e1000_adapter *adapter = rx_ring->adapter;
2350 	struct e1000_buffer *buffer_info;
2351 	int i, size, desc_len, err = -ENOMEM;
2352 
2353 	size = sizeof(struct e1000_buffer) * rx_ring->count;
2354 	rx_ring->buffer_info = vzalloc(size);
2355 	if (!rx_ring->buffer_info)
2356 		goto err;
2357 
2358 	for (i = 0; i < rx_ring->count; i++) {
2359 		buffer_info = &rx_ring->buffer_info[i];
2360 		buffer_info->ps_pages = kcalloc(PS_PAGE_BUFFERS,
2361 						sizeof(struct e1000_ps_page),
2362 						GFP_KERNEL);
2363 		if (!buffer_info->ps_pages)
2364 			goto err_pages;
2365 	}
2366 
2367 	desc_len = sizeof(union e1000_rx_desc_packet_split);
2368 
2369 	/* Round up to nearest 4K */
2370 	rx_ring->size = rx_ring->count * desc_len;
2371 	rx_ring->size = ALIGN(rx_ring->size, 4096);
2372 
2373 	err = e1000_alloc_ring_dma(adapter, rx_ring);
2374 	if (err)
2375 		goto err_pages;
2376 
2377 	rx_ring->next_to_clean = 0;
2378 	rx_ring->next_to_use = 0;
2379 	rx_ring->rx_skb_top = NULL;
2380 
2381 	return 0;
2382 
2383 err_pages:
2384 	for (i = 0; i < rx_ring->count; i++) {
2385 		buffer_info = &rx_ring->buffer_info[i];
2386 		kfree(buffer_info->ps_pages);
2387 	}
2388 err:
2389 	vfree(rx_ring->buffer_info);
2390 	e_err("Unable to allocate memory for the receive descriptor ring\n");
2391 	return err;
2392 }
2393 
2394 /**
2395  * e1000_clean_tx_ring - Free Tx Buffers
2396  * @tx_ring: Tx descriptor ring
2397  **/
2398 static void e1000_clean_tx_ring(struct e1000_ring *tx_ring)
2399 {
2400 	struct e1000_adapter *adapter = tx_ring->adapter;
2401 	struct e1000_buffer *buffer_info;
2402 	unsigned long size;
2403 	unsigned int i;
2404 
2405 	for (i = 0; i < tx_ring->count; i++) {
2406 		buffer_info = &tx_ring->buffer_info[i];
2407 		e1000_put_txbuf(tx_ring, buffer_info);
2408 	}
2409 
2410 	netdev_reset_queue(adapter->netdev);
2411 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2412 	memset(tx_ring->buffer_info, 0, size);
2413 
2414 	memset(tx_ring->desc, 0, tx_ring->size);
2415 
2416 	tx_ring->next_to_use = 0;
2417 	tx_ring->next_to_clean = 0;
2418 
2419 	writel(0, tx_ring->head);
2420 	if (tx_ring->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
2421 		e1000e_update_tdt_wa(tx_ring, 0);
2422 	else
2423 		writel(0, tx_ring->tail);
2424 }
2425 
2426 /**
2427  * e1000e_free_tx_resources - Free Tx Resources per Queue
2428  * @tx_ring: Tx descriptor ring
2429  *
2430  * Free all transmit software resources
2431  **/
2432 void e1000e_free_tx_resources(struct e1000_ring *tx_ring)
2433 {
2434 	struct e1000_adapter *adapter = tx_ring->adapter;
2435 	struct pci_dev *pdev = adapter->pdev;
2436 
2437 	e1000_clean_tx_ring(tx_ring);
2438 
2439 	vfree(tx_ring->buffer_info);
2440 	tx_ring->buffer_info = NULL;
2441 
2442 	dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
2443 			  tx_ring->dma);
2444 	tx_ring->desc = NULL;
2445 }
2446 
2447 /**
2448  * e1000e_free_rx_resources - Free Rx Resources
2449  * @rx_ring: Rx descriptor ring
2450  *
2451  * Free all receive software resources
2452  **/
2453 void e1000e_free_rx_resources(struct e1000_ring *rx_ring)
2454 {
2455 	struct e1000_adapter *adapter = rx_ring->adapter;
2456 	struct pci_dev *pdev = adapter->pdev;
2457 	int i;
2458 
2459 	e1000_clean_rx_ring(rx_ring);
2460 
2461 	for (i = 0; i < rx_ring->count; i++)
2462 		kfree(rx_ring->buffer_info[i].ps_pages);
2463 
2464 	vfree(rx_ring->buffer_info);
2465 	rx_ring->buffer_info = NULL;
2466 
2467 	dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
2468 			  rx_ring->dma);
2469 	rx_ring->desc = NULL;
2470 }
2471 
2472 /**
2473  * e1000_update_itr - update the dynamic ITR value based on statistics
2474  * @adapter: pointer to adapter
2475  * @itr_setting: current adapter->itr
2476  * @packets: the number of packets during this measurement interval
2477  * @bytes: the number of bytes during this measurement interval
2478  *
2479  *      Stores a new ITR value based on packets and byte
2480  *      counts during the last interrupt.  The advantage of per interrupt
2481  *      computation is faster updates and more accurate ITR for the current
2482  *      traffic pattern.  Constants in this function were computed
2483  *      based on theoretical maximum wire speed and thresholds were set based
2484  *      on testing data as well as attempting to minimize response time
2485  *      while increasing bulk throughput.  This functionality is controlled
2486  *      by the InterruptThrottleRate module parameter.
2487  **/
2488 static unsigned int e1000_update_itr(u16 itr_setting, int packets, int bytes)
2489 {
2490 	unsigned int retval = itr_setting;
2491 
2492 	if (packets == 0)
2493 		return itr_setting;
2494 
2495 	switch (itr_setting) {
2496 	case lowest_latency:
2497 		/* handle TSO and jumbo frames */
2498 		if (bytes/packets > 8000)
2499 			retval = bulk_latency;
2500 		else if ((packets < 5) && (bytes > 512))
2501 			retval = low_latency;
2502 		break;
2503 	case low_latency:  /* 50 usec aka 20000 ints/s */
2504 		if (bytes > 10000) {
2505 			/* this if handles the TSO accounting */
2506 			if (bytes/packets > 8000)
2507 				retval = bulk_latency;
2508 			else if ((packets < 10) || ((bytes/packets) > 1200))
2509 				retval = bulk_latency;
2510 			else if ((packets > 35))
2511 				retval = lowest_latency;
2512 		} else if (bytes/packets > 2000) {
2513 			retval = bulk_latency;
2514 		} else if (packets <= 2 && bytes < 512) {
2515 			retval = lowest_latency;
2516 		}
2517 		break;
2518 	case bulk_latency: /* 250 usec aka 4000 ints/s */
2519 		if (bytes > 25000) {
2520 			if (packets > 35)
2521 				retval = low_latency;
2522 		} else if (bytes < 6000) {
2523 			retval = low_latency;
2524 		}
2525 		break;
2526 	}
2527 
2528 	return retval;
2529 }
2530 
2531 static void e1000_set_itr(struct e1000_adapter *adapter)
2532 {
2533 	u16 current_itr;
2534 	u32 new_itr = adapter->itr;
2535 
2536 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
2537 	if (adapter->link_speed != SPEED_1000) {
2538 		current_itr = 0;
2539 		new_itr = 4000;
2540 		goto set_itr_now;
2541 	}
2542 
2543 	if (adapter->flags2 & FLAG2_DISABLE_AIM) {
2544 		new_itr = 0;
2545 		goto set_itr_now;
2546 	}
2547 
2548 	adapter->tx_itr = e1000_update_itr(adapter->tx_itr,
2549 					   adapter->total_tx_packets,
2550 					   adapter->total_tx_bytes);
2551 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2552 	if (adapter->itr_setting == 3 && adapter->tx_itr == lowest_latency)
2553 		adapter->tx_itr = low_latency;
2554 
2555 	adapter->rx_itr = e1000_update_itr(adapter->rx_itr,
2556 					   adapter->total_rx_packets,
2557 					   adapter->total_rx_bytes);
2558 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2559 	if (adapter->itr_setting == 3 && adapter->rx_itr == lowest_latency)
2560 		adapter->rx_itr = low_latency;
2561 
2562 	current_itr = max(adapter->rx_itr, adapter->tx_itr);
2563 
2564 	switch (current_itr) {
2565 	/* counts and packets in update_itr are dependent on these numbers */
2566 	case lowest_latency:
2567 		new_itr = 70000;
2568 		break;
2569 	case low_latency:
2570 		new_itr = 20000; /* aka hwitr = ~200 */
2571 		break;
2572 	case bulk_latency:
2573 		new_itr = 4000;
2574 		break;
2575 	default:
2576 		break;
2577 	}
2578 
2579 set_itr_now:
2580 	if (new_itr != adapter->itr) {
2581 		/* this attempts to bias the interrupt rate towards Bulk
2582 		 * by adding intermediate steps when interrupt rate is
2583 		 * increasing
2584 		 */
2585 		new_itr = new_itr > adapter->itr ?
2586 			     min(adapter->itr + (new_itr >> 2), new_itr) :
2587 			     new_itr;
2588 		adapter->itr = new_itr;
2589 		adapter->rx_ring->itr_val = new_itr;
2590 		if (adapter->msix_entries)
2591 			adapter->rx_ring->set_itr = 1;
2592 		else
2593 			e1000e_write_itr(adapter, new_itr);
2594 	}
2595 }
2596 
2597 /**
2598  * e1000e_write_itr - write the ITR value to the appropriate registers
2599  * @adapter: address of board private structure
2600  * @itr: new ITR value to program
2601  *
2602  * e1000e_write_itr determines if the adapter is in MSI-X mode
2603  * and, if so, writes the EITR registers with the ITR value.
2604  * Otherwise, it writes the ITR value into the ITR register.
2605  **/
2606 void e1000e_write_itr(struct e1000_adapter *adapter, u32 itr)
2607 {
2608 	struct e1000_hw *hw = &adapter->hw;
2609 	u32 new_itr = itr ? 1000000000 / (itr * 256) : 0;
2610 
2611 	if (adapter->msix_entries) {
2612 		int vector;
2613 
2614 		for (vector = 0; vector < adapter->num_vectors; vector++)
2615 			writel(new_itr, hw->hw_addr + E1000_EITR_82574(vector));
2616 	} else {
2617 		ew32(ITR, new_itr);
2618 	}
2619 }
2620 
2621 /**
2622  * e1000_alloc_queues - Allocate memory for all rings
2623  * @adapter: board private structure to initialize
2624  **/
2625 static int e1000_alloc_queues(struct e1000_adapter *adapter)
2626 {
2627 	int size = sizeof(struct e1000_ring);
2628 
2629 	adapter->tx_ring = kzalloc(size, GFP_KERNEL);
2630 	if (!adapter->tx_ring)
2631 		goto err;
2632 	adapter->tx_ring->count = adapter->tx_ring_count;
2633 	adapter->tx_ring->adapter = adapter;
2634 
2635 	adapter->rx_ring = kzalloc(size, GFP_KERNEL);
2636 	if (!adapter->rx_ring)
2637 		goto err;
2638 	adapter->rx_ring->count = adapter->rx_ring_count;
2639 	adapter->rx_ring->adapter = adapter;
2640 
2641 	return 0;
2642 err:
2643 	e_err("Unable to allocate memory for queues\n");
2644 	kfree(adapter->rx_ring);
2645 	kfree(adapter->tx_ring);
2646 	return -ENOMEM;
2647 }
2648 
2649 /**
2650  * e1000e_poll - NAPI Rx polling callback
2651  * @napi: struct associated with this polling callback
2652  * @weight: number of packets driver is allowed to process this poll
2653  **/
2654 static int e1000e_poll(struct napi_struct *napi, int weight)
2655 {
2656 	struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter,
2657 						     napi);
2658 	struct e1000_hw *hw = &adapter->hw;
2659 	struct net_device *poll_dev = adapter->netdev;
2660 	int tx_cleaned = 1, work_done = 0;
2661 
2662 	adapter = netdev_priv(poll_dev);
2663 
2664 	if (!adapter->msix_entries ||
2665 	    (adapter->rx_ring->ims_val & adapter->tx_ring->ims_val))
2666 		tx_cleaned = e1000_clean_tx_irq(adapter->tx_ring);
2667 
2668 	adapter->clean_rx(adapter->rx_ring, &work_done, weight);
2669 
2670 	if (!tx_cleaned)
2671 		work_done = weight;
2672 
2673 	/* If weight not fully consumed, exit the polling mode */
2674 	if (work_done < weight) {
2675 		if (adapter->itr_setting & 3)
2676 			e1000_set_itr(adapter);
2677 		napi_complete(napi);
2678 		if (!test_bit(__E1000_DOWN, &adapter->state)) {
2679 			if (adapter->msix_entries)
2680 				ew32(IMS, adapter->rx_ring->ims_val);
2681 			else
2682 				e1000_irq_enable(adapter);
2683 		}
2684 	}
2685 
2686 	return work_done;
2687 }
2688 
2689 static int e1000_vlan_rx_add_vid(struct net_device *netdev, u16 vid)
2690 {
2691 	struct e1000_adapter *adapter = netdev_priv(netdev);
2692 	struct e1000_hw *hw = &adapter->hw;
2693 	u32 vfta, index;
2694 
2695 	/* don't update vlan cookie if already programmed */
2696 	if ((adapter->hw.mng_cookie.status &
2697 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2698 	    (vid == adapter->mng_vlan_id))
2699 		return 0;
2700 
2701 	/* add VID to filter table */
2702 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2703 		index = (vid >> 5) & 0x7F;
2704 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2705 		vfta |= (1 << (vid & 0x1F));
2706 		hw->mac.ops.write_vfta(hw, index, vfta);
2707 	}
2708 
2709 	set_bit(vid, adapter->active_vlans);
2710 
2711 	return 0;
2712 }
2713 
2714 static int e1000_vlan_rx_kill_vid(struct net_device *netdev, u16 vid)
2715 {
2716 	struct e1000_adapter *adapter = netdev_priv(netdev);
2717 	struct e1000_hw *hw = &adapter->hw;
2718 	u32 vfta, index;
2719 
2720 	if ((adapter->hw.mng_cookie.status &
2721 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2722 	    (vid == adapter->mng_vlan_id)) {
2723 		/* release control to f/w */
2724 		e1000e_release_hw_control(adapter);
2725 		return 0;
2726 	}
2727 
2728 	/* remove VID from filter table */
2729 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2730 		index = (vid >> 5) & 0x7F;
2731 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2732 		vfta &= ~(1 << (vid & 0x1F));
2733 		hw->mac.ops.write_vfta(hw, index, vfta);
2734 	}
2735 
2736 	clear_bit(vid, adapter->active_vlans);
2737 
2738 	return 0;
2739 }
2740 
2741 /**
2742  * e1000e_vlan_filter_disable - helper to disable hw VLAN filtering
2743  * @adapter: board private structure to initialize
2744  **/
2745 static void e1000e_vlan_filter_disable(struct e1000_adapter *adapter)
2746 {
2747 	struct net_device *netdev = adapter->netdev;
2748 	struct e1000_hw *hw = &adapter->hw;
2749 	u32 rctl;
2750 
2751 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2752 		/* disable VLAN receive filtering */
2753 		rctl = er32(RCTL);
2754 		rctl &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
2755 		ew32(RCTL, rctl);
2756 
2757 		if (adapter->mng_vlan_id != (u16)E1000_MNG_VLAN_NONE) {
2758 			e1000_vlan_rx_kill_vid(netdev, adapter->mng_vlan_id);
2759 			adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
2760 		}
2761 	}
2762 }
2763 
2764 /**
2765  * e1000e_vlan_filter_enable - helper to enable HW VLAN filtering
2766  * @adapter: board private structure to initialize
2767  **/
2768 static void e1000e_vlan_filter_enable(struct e1000_adapter *adapter)
2769 {
2770 	struct e1000_hw *hw = &adapter->hw;
2771 	u32 rctl;
2772 
2773 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2774 		/* enable VLAN receive filtering */
2775 		rctl = er32(RCTL);
2776 		rctl |= E1000_RCTL_VFE;
2777 		rctl &= ~E1000_RCTL_CFIEN;
2778 		ew32(RCTL, rctl);
2779 	}
2780 }
2781 
2782 /**
2783  * e1000e_vlan_strip_enable - helper to disable HW VLAN stripping
2784  * @adapter: board private structure to initialize
2785  **/
2786 static void e1000e_vlan_strip_disable(struct e1000_adapter *adapter)
2787 {
2788 	struct e1000_hw *hw = &adapter->hw;
2789 	u32 ctrl;
2790 
2791 	/* disable VLAN tag insert/strip */
2792 	ctrl = er32(CTRL);
2793 	ctrl &= ~E1000_CTRL_VME;
2794 	ew32(CTRL, ctrl);
2795 }
2796 
2797 /**
2798  * e1000e_vlan_strip_enable - helper to enable HW VLAN stripping
2799  * @adapter: board private structure to initialize
2800  **/
2801 static void e1000e_vlan_strip_enable(struct e1000_adapter *adapter)
2802 {
2803 	struct e1000_hw *hw = &adapter->hw;
2804 	u32 ctrl;
2805 
2806 	/* enable VLAN tag insert/strip */
2807 	ctrl = er32(CTRL);
2808 	ctrl |= E1000_CTRL_VME;
2809 	ew32(CTRL, ctrl);
2810 }
2811 
2812 static void e1000_update_mng_vlan(struct e1000_adapter *adapter)
2813 {
2814 	struct net_device *netdev = adapter->netdev;
2815 	u16 vid = adapter->hw.mng_cookie.vlan_id;
2816 	u16 old_vid = adapter->mng_vlan_id;
2817 
2818 	if (adapter->hw.mng_cookie.status &
2819 	    E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
2820 		e1000_vlan_rx_add_vid(netdev, vid);
2821 		adapter->mng_vlan_id = vid;
2822 	}
2823 
2824 	if ((old_vid != (u16)E1000_MNG_VLAN_NONE) && (vid != old_vid))
2825 		e1000_vlan_rx_kill_vid(netdev, old_vid);
2826 }
2827 
2828 static void e1000_restore_vlan(struct e1000_adapter *adapter)
2829 {
2830 	u16 vid;
2831 
2832 	e1000_vlan_rx_add_vid(adapter->netdev, 0);
2833 
2834 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
2835 		e1000_vlan_rx_add_vid(adapter->netdev, vid);
2836 }
2837 
2838 static void e1000_init_manageability_pt(struct e1000_adapter *adapter)
2839 {
2840 	struct e1000_hw *hw = &adapter->hw;
2841 	u32 manc, manc2h, mdef, i, j;
2842 
2843 	if (!(adapter->flags & FLAG_MNG_PT_ENABLED))
2844 		return;
2845 
2846 	manc = er32(MANC);
2847 
2848 	/* enable receiving management packets to the host. this will probably
2849 	 * generate destination unreachable messages from the host OS, but
2850 	 * the packets will be handled on SMBUS
2851 	 */
2852 	manc |= E1000_MANC_EN_MNG2HOST;
2853 	manc2h = er32(MANC2H);
2854 
2855 	switch (hw->mac.type) {
2856 	default:
2857 		manc2h |= (E1000_MANC2H_PORT_623 | E1000_MANC2H_PORT_664);
2858 		break;
2859 	case e1000_82574:
2860 	case e1000_82583:
2861 		/* Check if IPMI pass-through decision filter already exists;
2862 		 * if so, enable it.
2863 		 */
2864 		for (i = 0, j = 0; i < 8; i++) {
2865 			mdef = er32(MDEF(i));
2866 
2867 			/* Ignore filters with anything other than IPMI ports */
2868 			if (mdef & ~(E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2869 				continue;
2870 
2871 			/* Enable this decision filter in MANC2H */
2872 			if (mdef)
2873 				manc2h |= (1 << i);
2874 
2875 			j |= mdef;
2876 		}
2877 
2878 		if (j == (E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2879 			break;
2880 
2881 		/* Create new decision filter in an empty filter */
2882 		for (i = 0, j = 0; i < 8; i++)
2883 			if (er32(MDEF(i)) == 0) {
2884 				ew32(MDEF(i), (E1000_MDEF_PORT_623 |
2885 					       E1000_MDEF_PORT_664));
2886 				manc2h |= (1 << 1);
2887 				j++;
2888 				break;
2889 			}
2890 
2891 		if (!j)
2892 			e_warn("Unable to create IPMI pass-through filter\n");
2893 		break;
2894 	}
2895 
2896 	ew32(MANC2H, manc2h);
2897 	ew32(MANC, manc);
2898 }
2899 
2900 /**
2901  * e1000_configure_tx - Configure Transmit Unit after Reset
2902  * @adapter: board private structure
2903  *
2904  * Configure the Tx unit of the MAC after a reset.
2905  **/
2906 static void e1000_configure_tx(struct e1000_adapter *adapter)
2907 {
2908 	struct e1000_hw *hw = &adapter->hw;
2909 	struct e1000_ring *tx_ring = adapter->tx_ring;
2910 	u64 tdba;
2911 	u32 tdlen, tarc;
2912 
2913 	/* Setup the HW Tx Head and Tail descriptor pointers */
2914 	tdba = tx_ring->dma;
2915 	tdlen = tx_ring->count * sizeof(struct e1000_tx_desc);
2916 	ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
2917 	ew32(TDBAH(0), (tdba >> 32));
2918 	ew32(TDLEN(0), tdlen);
2919 	ew32(TDH(0), 0);
2920 	ew32(TDT(0), 0);
2921 	tx_ring->head = adapter->hw.hw_addr + E1000_TDH(0);
2922 	tx_ring->tail = adapter->hw.hw_addr + E1000_TDT(0);
2923 
2924 	/* Set the Tx Interrupt Delay register */
2925 	ew32(TIDV, adapter->tx_int_delay);
2926 	/* Tx irq moderation */
2927 	ew32(TADV, adapter->tx_abs_int_delay);
2928 
2929 	if (adapter->flags2 & FLAG2_DMA_BURST) {
2930 		u32 txdctl = er32(TXDCTL(0));
2931 		txdctl &= ~(E1000_TXDCTL_PTHRESH | E1000_TXDCTL_HTHRESH |
2932 			    E1000_TXDCTL_WTHRESH);
2933 		/* set up some performance related parameters to encourage the
2934 		 * hardware to use the bus more efficiently in bursts, depends
2935 		 * on the tx_int_delay to be enabled,
2936 		 * wthresh = 1 ==> burst write is disabled to avoid Tx stalls
2937 		 * hthresh = 1 ==> prefetch when one or more available
2938 		 * pthresh = 0x1f ==> prefetch if internal cache 31 or less
2939 		 * BEWARE: this seems to work but should be considered first if
2940 		 * there are Tx hangs or other Tx related bugs
2941 		 */
2942 		txdctl |= E1000_TXDCTL_DMA_BURST_ENABLE;
2943 		ew32(TXDCTL(0), txdctl);
2944 	}
2945 	/* erratum work around: set txdctl the same for both queues */
2946 	ew32(TXDCTL(1), er32(TXDCTL(0)));
2947 
2948 	if (adapter->flags & FLAG_TARC_SPEED_MODE_BIT) {
2949 		tarc = er32(TARC(0));
2950 		/* set the speed mode bit, we'll clear it if we're not at
2951 		 * gigabit link later
2952 		 */
2953 #define SPEED_MODE_BIT (1 << 21)
2954 		tarc |= SPEED_MODE_BIT;
2955 		ew32(TARC(0), tarc);
2956 	}
2957 
2958 	/* errata: program both queues to unweighted RR */
2959 	if (adapter->flags & FLAG_TARC_SET_BIT_ZERO) {
2960 		tarc = er32(TARC(0));
2961 		tarc |= 1;
2962 		ew32(TARC(0), tarc);
2963 		tarc = er32(TARC(1));
2964 		tarc |= 1;
2965 		ew32(TARC(1), tarc);
2966 	}
2967 
2968 	/* Setup Transmit Descriptor Settings for eop descriptor */
2969 	adapter->txd_cmd = E1000_TXD_CMD_EOP | E1000_TXD_CMD_IFCS;
2970 
2971 	/* only set IDE if we are delaying interrupts using the timers */
2972 	if (adapter->tx_int_delay)
2973 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
2974 
2975 	/* enable Report Status bit */
2976 	adapter->txd_cmd |= E1000_TXD_CMD_RS;
2977 
2978 	hw->mac.ops.config_collision_dist(hw);
2979 }
2980 
2981 /**
2982  * e1000_setup_rctl - configure the receive control registers
2983  * @adapter: Board private structure
2984  **/
2985 #define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
2986 			   (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
2987 static void e1000_setup_rctl(struct e1000_adapter *adapter)
2988 {
2989 	struct e1000_hw *hw = &adapter->hw;
2990 	u32 rctl, rfctl;
2991 	u32 pages = 0;
2992 
2993 	/* Workaround Si errata on PCHx - configure jumbo frame flow */
2994 	if (hw->mac.type >= e1000_pch2lan) {
2995 		s32 ret_val;
2996 
2997 		if (adapter->netdev->mtu > ETH_DATA_LEN)
2998 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, true);
2999 		else
3000 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, false);
3001 
3002 		if (ret_val)
3003 			e_dbg("failed to enable jumbo frame workaround mode\n");
3004 	}
3005 
3006 	/* Program MC offset vector base */
3007 	rctl = er32(RCTL);
3008 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3009 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3010 		E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3011 		(adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3012 
3013 	/* Do not Store bad packets */
3014 	rctl &= ~E1000_RCTL_SBP;
3015 
3016 	/* Enable Long Packet receive */
3017 	if (adapter->netdev->mtu <= ETH_DATA_LEN)
3018 		rctl &= ~E1000_RCTL_LPE;
3019 	else
3020 		rctl |= E1000_RCTL_LPE;
3021 
3022 	/* Some systems expect that the CRC is included in SMBUS traffic. The
3023 	 * hardware strips the CRC before sending to both SMBUS (BMC) and to
3024 	 * host memory when this is enabled
3025 	 */
3026 	if (adapter->flags2 & FLAG2_CRC_STRIPPING)
3027 		rctl |= E1000_RCTL_SECRC;
3028 
3029 	/* Workaround Si errata on 82577 PHY - configure IPG for jumbos */
3030 	if ((hw->phy.type == e1000_phy_82577) && (rctl & E1000_RCTL_LPE)) {
3031 		u16 phy_data;
3032 
3033 		e1e_rphy(hw, PHY_REG(770, 26), &phy_data);
3034 		phy_data &= 0xfff8;
3035 		phy_data |= (1 << 2);
3036 		e1e_wphy(hw, PHY_REG(770, 26), phy_data);
3037 
3038 		e1e_rphy(hw, 22, &phy_data);
3039 		phy_data &= 0x0fff;
3040 		phy_data |= (1 << 14);
3041 		e1e_wphy(hw, 0x10, 0x2823);
3042 		e1e_wphy(hw, 0x11, 0x0003);
3043 		e1e_wphy(hw, 22, phy_data);
3044 	}
3045 
3046 	/* Setup buffer sizes */
3047 	rctl &= ~E1000_RCTL_SZ_4096;
3048 	rctl |= E1000_RCTL_BSEX;
3049 	switch (adapter->rx_buffer_len) {
3050 	case 2048:
3051 	default:
3052 		rctl |= E1000_RCTL_SZ_2048;
3053 		rctl &= ~E1000_RCTL_BSEX;
3054 		break;
3055 	case 4096:
3056 		rctl |= E1000_RCTL_SZ_4096;
3057 		break;
3058 	case 8192:
3059 		rctl |= E1000_RCTL_SZ_8192;
3060 		break;
3061 	case 16384:
3062 		rctl |= E1000_RCTL_SZ_16384;
3063 		break;
3064 	}
3065 
3066 	/* Enable Extended Status in all Receive Descriptors */
3067 	rfctl = er32(RFCTL);
3068 	rfctl |= E1000_RFCTL_EXTEN;
3069 	ew32(RFCTL, rfctl);
3070 
3071 	/* 82571 and greater support packet-split where the protocol
3072 	 * header is placed in skb->data and the packet data is
3073 	 * placed in pages hanging off of skb_shinfo(skb)->nr_frags.
3074 	 * In the case of a non-split, skb->data is linearly filled,
3075 	 * followed by the page buffers.  Therefore, skb->data is
3076 	 * sized to hold the largest protocol header.
3077 	 *
3078 	 * allocations using alloc_page take too long for regular MTU
3079 	 * so only enable packet split for jumbo frames
3080 	 *
3081 	 * Using pages when the page size is greater than 16k wastes
3082 	 * a lot of memory, since we allocate 3 pages at all times
3083 	 * per packet.
3084 	 */
3085 	pages = PAGE_USE_COUNT(adapter->netdev->mtu);
3086 	if ((pages <= 3) && (PAGE_SIZE <= 16384) && (rctl & E1000_RCTL_LPE))
3087 		adapter->rx_ps_pages = pages;
3088 	else
3089 		adapter->rx_ps_pages = 0;
3090 
3091 	if (adapter->rx_ps_pages) {
3092 		u32 psrctl = 0;
3093 
3094 		/* Enable Packet split descriptors */
3095 		rctl |= E1000_RCTL_DTYP_PS;
3096 
3097 		psrctl |= adapter->rx_ps_bsize0 >>
3098 			E1000_PSRCTL_BSIZE0_SHIFT;
3099 
3100 		switch (adapter->rx_ps_pages) {
3101 		case 3:
3102 			psrctl |= PAGE_SIZE <<
3103 				E1000_PSRCTL_BSIZE3_SHIFT;
3104 		case 2:
3105 			psrctl |= PAGE_SIZE <<
3106 				E1000_PSRCTL_BSIZE2_SHIFT;
3107 		case 1:
3108 			psrctl |= PAGE_SIZE >>
3109 				E1000_PSRCTL_BSIZE1_SHIFT;
3110 			break;
3111 		}
3112 
3113 		ew32(PSRCTL, psrctl);
3114 	}
3115 
3116 	/* This is useful for sniffing bad packets. */
3117 	if (adapter->netdev->features & NETIF_F_RXALL) {
3118 		/* UPE and MPE will be handled by normal PROMISC logic
3119 		 * in e1000e_set_rx_mode
3120 		 */
3121 		rctl |= (E1000_RCTL_SBP | /* Receive bad packets */
3122 			 E1000_RCTL_BAM | /* RX All Bcast Pkts */
3123 			 E1000_RCTL_PMCF); /* RX All MAC Ctrl Pkts */
3124 
3125 		rctl &= ~(E1000_RCTL_VFE | /* Disable VLAN filter */
3126 			  E1000_RCTL_DPF | /* Allow filtered pause */
3127 			  E1000_RCTL_CFIEN); /* Dis VLAN CFIEN Filter */
3128 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3129 		 * and that breaks VLANs.
3130 		 */
3131 	}
3132 
3133 	ew32(RCTL, rctl);
3134 	/* just started the receive unit, no need to restart */
3135 	adapter->flags &= ~FLAG_RESTART_NOW;
3136 }
3137 
3138 /**
3139  * e1000_configure_rx - Configure Receive Unit after Reset
3140  * @adapter: board private structure
3141  *
3142  * Configure the Rx unit of the MAC after a reset.
3143  **/
3144 static void e1000_configure_rx(struct e1000_adapter *adapter)
3145 {
3146 	struct e1000_hw *hw = &adapter->hw;
3147 	struct e1000_ring *rx_ring = adapter->rx_ring;
3148 	u64 rdba;
3149 	u32 rdlen, rctl, rxcsum, ctrl_ext;
3150 
3151 	if (adapter->rx_ps_pages) {
3152 		/* this is a 32 byte descriptor */
3153 		rdlen = rx_ring->count *
3154 		    sizeof(union e1000_rx_desc_packet_split);
3155 		adapter->clean_rx = e1000_clean_rx_irq_ps;
3156 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers_ps;
3157 	} else if (adapter->netdev->mtu > ETH_FRAME_LEN + ETH_FCS_LEN) {
3158 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3159 		adapter->clean_rx = e1000_clean_jumbo_rx_irq;
3160 		adapter->alloc_rx_buf = e1000_alloc_jumbo_rx_buffers;
3161 	} else {
3162 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3163 		adapter->clean_rx = e1000_clean_rx_irq;
3164 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers;
3165 	}
3166 
3167 	/* disable receives while setting up the descriptors */
3168 	rctl = er32(RCTL);
3169 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
3170 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
3171 	e1e_flush();
3172 	usleep_range(10000, 20000);
3173 
3174 	if (adapter->flags2 & FLAG2_DMA_BURST) {
3175 		/* set the writeback threshold (only takes effect if the RDTR
3176 		 * is set). set GRAN=1 and write back up to 0x4 worth, and
3177 		 * enable prefetching of 0x20 Rx descriptors
3178 		 * granularity = 01
3179 		 * wthresh = 04,
3180 		 * hthresh = 04,
3181 		 * pthresh = 0x20
3182 		 */
3183 		ew32(RXDCTL(0), E1000_RXDCTL_DMA_BURST_ENABLE);
3184 		ew32(RXDCTL(1), E1000_RXDCTL_DMA_BURST_ENABLE);
3185 
3186 		/* override the delay timers for enabling bursting, only if
3187 		 * the value was not set by the user via module options
3188 		 */
3189 		if (adapter->rx_int_delay == DEFAULT_RDTR)
3190 			adapter->rx_int_delay = BURST_RDTR;
3191 		if (adapter->rx_abs_int_delay == DEFAULT_RADV)
3192 			adapter->rx_abs_int_delay = BURST_RADV;
3193 	}
3194 
3195 	/* set the Receive Delay Timer Register */
3196 	ew32(RDTR, adapter->rx_int_delay);
3197 
3198 	/* irq moderation */
3199 	ew32(RADV, adapter->rx_abs_int_delay);
3200 	if ((adapter->itr_setting != 0) && (adapter->itr != 0))
3201 		e1000e_write_itr(adapter, adapter->itr);
3202 
3203 	ctrl_ext = er32(CTRL_EXT);
3204 	/* Auto-Mask interrupts upon ICR access */
3205 	ctrl_ext |= E1000_CTRL_EXT_IAME;
3206 	ew32(IAM, 0xffffffff);
3207 	ew32(CTRL_EXT, ctrl_ext);
3208 	e1e_flush();
3209 
3210 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
3211 	 * the Base and Length of the Rx Descriptor Ring
3212 	 */
3213 	rdba = rx_ring->dma;
3214 	ew32(RDBAL(0), (rdba & DMA_BIT_MASK(32)));
3215 	ew32(RDBAH(0), (rdba >> 32));
3216 	ew32(RDLEN(0), rdlen);
3217 	ew32(RDH(0), 0);
3218 	ew32(RDT(0), 0);
3219 	rx_ring->head = adapter->hw.hw_addr + E1000_RDH(0);
3220 	rx_ring->tail = adapter->hw.hw_addr + E1000_RDT(0);
3221 
3222 	/* Enable Receive Checksum Offload for TCP and UDP */
3223 	rxcsum = er32(RXCSUM);
3224 	if (adapter->netdev->features & NETIF_F_RXCSUM)
3225 		rxcsum |= E1000_RXCSUM_TUOFL;
3226 	else
3227 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3228 	ew32(RXCSUM, rxcsum);
3229 
3230 	/* With jumbo frames, excessive C-state transition latencies result
3231 	 * in dropped transactions.
3232 	 */
3233 	if (adapter->netdev->mtu > ETH_DATA_LEN) {
3234 		u32 lat =
3235 		    ((er32(PBA) & E1000_PBA_RXA_MASK) * 1024 -
3236 		     adapter->max_frame_size) * 8 / 1000;
3237 
3238 		if (adapter->flags & FLAG_IS_ICH) {
3239 			u32 rxdctl = er32(RXDCTL(0));
3240 			ew32(RXDCTL(0), rxdctl | 0x3);
3241 		}
3242 
3243 		pm_qos_update_request(&adapter->netdev->pm_qos_req, lat);
3244 	} else {
3245 		pm_qos_update_request(&adapter->netdev->pm_qos_req,
3246 				      PM_QOS_DEFAULT_VALUE);
3247 	}
3248 
3249 	/* Enable Receives */
3250 	ew32(RCTL, rctl);
3251 }
3252 
3253 /**
3254  * e1000e_write_mc_addr_list - write multicast addresses to MTA
3255  * @netdev: network interface device structure
3256  *
3257  * Writes multicast address list to the MTA hash table.
3258  * Returns: -ENOMEM on failure
3259  *                0 on no addresses written
3260  *                X on writing X addresses to MTA
3261  */
3262 static int e1000e_write_mc_addr_list(struct net_device *netdev)
3263 {
3264 	struct e1000_adapter *adapter = netdev_priv(netdev);
3265 	struct e1000_hw *hw = &adapter->hw;
3266 	struct netdev_hw_addr *ha;
3267 	u8 *mta_list;
3268 	int i;
3269 
3270 	if (netdev_mc_empty(netdev)) {
3271 		/* nothing to program, so clear mc list */
3272 		hw->mac.ops.update_mc_addr_list(hw, NULL, 0);
3273 		return 0;
3274 	}
3275 
3276 	mta_list = kzalloc(netdev_mc_count(netdev) * ETH_ALEN, GFP_ATOMIC);
3277 	if (!mta_list)
3278 		return -ENOMEM;
3279 
3280 	/* update_mc_addr_list expects a packed array of only addresses. */
3281 	i = 0;
3282 	netdev_for_each_mc_addr(ha, netdev)
3283 		memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3284 
3285 	hw->mac.ops.update_mc_addr_list(hw, mta_list, i);
3286 	kfree(mta_list);
3287 
3288 	return netdev_mc_count(netdev);
3289 }
3290 
3291 /**
3292  * e1000e_write_uc_addr_list - write unicast addresses to RAR table
3293  * @netdev: network interface device structure
3294  *
3295  * Writes unicast address list to the RAR table.
3296  * Returns: -ENOMEM on failure/insufficient address space
3297  *                0 on no addresses written
3298  *                X on writing X addresses to the RAR table
3299  **/
3300 static int e1000e_write_uc_addr_list(struct net_device *netdev)
3301 {
3302 	struct e1000_adapter *adapter = netdev_priv(netdev);
3303 	struct e1000_hw *hw = &adapter->hw;
3304 	unsigned int rar_entries = hw->mac.rar_entry_count;
3305 	int count = 0;
3306 
3307 	/* save a rar entry for our hardware address */
3308 	rar_entries--;
3309 
3310 	/* save a rar entry for the LAA workaround */
3311 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA)
3312 		rar_entries--;
3313 
3314 	/* return ENOMEM indicating insufficient memory for addresses */
3315 	if (netdev_uc_count(netdev) > rar_entries)
3316 		return -ENOMEM;
3317 
3318 	if (!netdev_uc_empty(netdev) && rar_entries) {
3319 		struct netdev_hw_addr *ha;
3320 
3321 		/* write the addresses in reverse order to avoid write
3322 		 * combining
3323 		 */
3324 		netdev_for_each_uc_addr(ha, netdev) {
3325 			if (!rar_entries)
3326 				break;
3327 			hw->mac.ops.rar_set(hw, ha->addr, rar_entries--);
3328 			count++;
3329 		}
3330 	}
3331 
3332 	/* zero out the remaining RAR entries not used above */
3333 	for (; rar_entries > 0; rar_entries--) {
3334 		ew32(RAH(rar_entries), 0);
3335 		ew32(RAL(rar_entries), 0);
3336 	}
3337 	e1e_flush();
3338 
3339 	return count;
3340 }
3341 
3342 /**
3343  * e1000e_set_rx_mode - secondary unicast, Multicast and Promiscuous mode set
3344  * @netdev: network interface device structure
3345  *
3346  * The ndo_set_rx_mode entry point is called whenever the unicast or multicast
3347  * address list or the network interface flags are updated.  This routine is
3348  * responsible for configuring the hardware for proper unicast, multicast,
3349  * promiscuous mode, and all-multi behavior.
3350  **/
3351 static void e1000e_set_rx_mode(struct net_device *netdev)
3352 {
3353 	struct e1000_adapter *adapter = netdev_priv(netdev);
3354 	struct e1000_hw *hw = &adapter->hw;
3355 	u32 rctl;
3356 
3357 	/* Check for Promiscuous and All Multicast modes */
3358 	rctl = er32(RCTL);
3359 
3360 	/* clear the affected bits */
3361 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
3362 
3363 	if (netdev->flags & IFF_PROMISC) {
3364 		rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
3365 		/* Do not hardware filter VLANs in promisc mode */
3366 		e1000e_vlan_filter_disable(adapter);
3367 	} else {
3368 		int count;
3369 
3370 		if (netdev->flags & IFF_ALLMULTI) {
3371 			rctl |= E1000_RCTL_MPE;
3372 		} else {
3373 			/* Write addresses to the MTA, if the attempt fails
3374 			 * then we should just turn on promiscuous mode so
3375 			 * that we can at least receive multicast traffic
3376 			 */
3377 			count = e1000e_write_mc_addr_list(netdev);
3378 			if (count < 0)
3379 				rctl |= E1000_RCTL_MPE;
3380 		}
3381 		e1000e_vlan_filter_enable(adapter);
3382 		/* Write addresses to available RAR registers, if there is not
3383 		 * sufficient space to store all the addresses then enable
3384 		 * unicast promiscuous mode
3385 		 */
3386 		count = e1000e_write_uc_addr_list(netdev);
3387 		if (count < 0)
3388 			rctl |= E1000_RCTL_UPE;
3389 	}
3390 
3391 	ew32(RCTL, rctl);
3392 
3393 	if (netdev->features & NETIF_F_HW_VLAN_RX)
3394 		e1000e_vlan_strip_enable(adapter);
3395 	else
3396 		e1000e_vlan_strip_disable(adapter);
3397 }
3398 
3399 static void e1000e_setup_rss_hash(struct e1000_adapter *adapter)
3400 {
3401 	struct e1000_hw *hw = &adapter->hw;
3402 	u32 mrqc, rxcsum;
3403 	int i;
3404 	static const u32 rsskey[10] = {
3405 		0xda565a6d, 0xc20e5b25, 0x3d256741, 0xb08fa343, 0xcb2bcad0,
3406 		0xb4307bae, 0xa32dcb77, 0x0cf23080, 0x3bb7426a, 0xfa01acbe
3407 	};
3408 
3409 	/* Fill out hash function seed */
3410 	for (i = 0; i < 10; i++)
3411 		ew32(RSSRK(i), rsskey[i]);
3412 
3413 	/* Direct all traffic to queue 0 */
3414 	for (i = 0; i < 32; i++)
3415 		ew32(RETA(i), 0);
3416 
3417 	/* Disable raw packet checksumming so that RSS hash is placed in
3418 	 * descriptor on writeback.
3419 	 */
3420 	rxcsum = er32(RXCSUM);
3421 	rxcsum |= E1000_RXCSUM_PCSD;
3422 
3423 	ew32(RXCSUM, rxcsum);
3424 
3425 	mrqc = (E1000_MRQC_RSS_FIELD_IPV4 |
3426 		E1000_MRQC_RSS_FIELD_IPV4_TCP |
3427 		E1000_MRQC_RSS_FIELD_IPV6 |
3428 		E1000_MRQC_RSS_FIELD_IPV6_TCP |
3429 		E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3430 
3431 	ew32(MRQC, mrqc);
3432 }
3433 
3434 /**
3435  * e1000e_get_base_timinca - get default SYSTIM time increment attributes
3436  * @adapter: board private structure
3437  * @timinca: pointer to returned time increment attributes
3438  *
3439  * Get attributes for incrementing the System Time Register SYSTIML/H at
3440  * the default base frequency, and set the cyclecounter shift value.
3441  **/
3442 s32 e1000e_get_base_timinca(struct e1000_adapter *adapter, u32 *timinca)
3443 {
3444 	struct e1000_hw *hw = &adapter->hw;
3445 	u32 incvalue, incperiod, shift;
3446 
3447 	/* Make sure clock is enabled on I217 before checking the frequency */
3448 	if ((hw->mac.type == e1000_pch_lpt) &&
3449 	    !(er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) &&
3450 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_ENABLED)) {
3451 		u32 fextnvm7 = er32(FEXTNVM7);
3452 
3453 		if (!(fextnvm7 & (1 << 0))) {
3454 			ew32(FEXTNVM7, fextnvm7 | (1 << 0));
3455 			e1e_flush();
3456 		}
3457 	}
3458 
3459 	switch (hw->mac.type) {
3460 	case e1000_pch2lan:
3461 	case e1000_pch_lpt:
3462 		/* On I217, the clock frequency is 25MHz or 96MHz as
3463 		 * indicated by the System Clock Frequency Indication
3464 		 */
3465 		if ((hw->mac.type != e1000_pch_lpt) ||
3466 		    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI)) {
3467 			/* Stable 96MHz frequency */
3468 			incperiod = INCPERIOD_96MHz;
3469 			incvalue = INCVALUE_96MHz;
3470 			shift = INCVALUE_SHIFT_96MHz;
3471 			adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHz;
3472 			break;
3473 		}
3474 		/* fall-through */
3475 	case e1000_82574:
3476 	case e1000_82583:
3477 		/* Stable 25MHz frequency */
3478 		incperiod = INCPERIOD_25MHz;
3479 		incvalue = INCVALUE_25MHz;
3480 		shift = INCVALUE_SHIFT_25MHz;
3481 		adapter->cc.shift = shift;
3482 		break;
3483 	default:
3484 		return -EINVAL;
3485 	}
3486 
3487 	*timinca = ((incperiod << E1000_TIMINCA_INCPERIOD_SHIFT) |
3488 		    ((incvalue << shift) & E1000_TIMINCA_INCVALUE_MASK));
3489 
3490 	return 0;
3491 }
3492 
3493 /**
3494  * e1000e_config_hwtstamp - configure the hwtstamp registers and enable/disable
3495  * @adapter: board private structure
3496  *
3497  * Outgoing time stamping can be enabled and disabled. Play nice and
3498  * disable it when requested, although it shouldn't cause any overhead
3499  * when no packet needs it. At most one packet in the queue may be
3500  * marked for time stamping, otherwise it would be impossible to tell
3501  * for sure to which packet the hardware time stamp belongs.
3502  *
3503  * Incoming time stamping has to be configured via the hardware filters.
3504  * Not all combinations are supported, in particular event type has to be
3505  * specified. Matching the kind of event packet is not supported, with the
3506  * exception of "all V2 events regardless of level 2 or 4".
3507  **/
3508 static int e1000e_config_hwtstamp(struct e1000_adapter *adapter)
3509 {
3510 	struct e1000_hw *hw = &adapter->hw;
3511 	struct hwtstamp_config *config = &adapter->hwtstamp_config;
3512 	u32 tsync_tx_ctl = E1000_TSYNCTXCTL_ENABLED;
3513 	u32 tsync_rx_ctl = E1000_TSYNCRXCTL_ENABLED;
3514 	u32 rxmtrl = 0;
3515 	u16 rxudp = 0;
3516 	bool is_l4 = false;
3517 	bool is_l2 = false;
3518 	u32 regval;
3519 	s32 ret_val;
3520 
3521 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3522 		return -EINVAL;
3523 
3524 	/* flags reserved for future extensions - must be zero */
3525 	if (config->flags)
3526 		return -EINVAL;
3527 
3528 	switch (config->tx_type) {
3529 	case HWTSTAMP_TX_OFF:
3530 		tsync_tx_ctl = 0;
3531 		break;
3532 	case HWTSTAMP_TX_ON:
3533 		break;
3534 	default:
3535 		return -ERANGE;
3536 	}
3537 
3538 	switch (config->rx_filter) {
3539 	case HWTSTAMP_FILTER_NONE:
3540 		tsync_rx_ctl = 0;
3541 		break;
3542 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3543 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3544 		rxmtrl = E1000_RXMTRL_PTP_V1_SYNC_MESSAGE;
3545 		is_l4 = true;
3546 		break;
3547 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3548 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3549 		rxmtrl = E1000_RXMTRL_PTP_V1_DELAY_REQ_MESSAGE;
3550 		is_l4 = true;
3551 		break;
3552 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3553 		/* Also time stamps V2 L2 Path Delay Request/Response */
3554 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3555 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3556 		is_l2 = true;
3557 		break;
3558 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3559 		/* Also time stamps V2 L2 Path Delay Request/Response. */
3560 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3561 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3562 		is_l2 = true;
3563 		break;
3564 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3565 		/* Hardware cannot filter just V2 L4 Sync messages;
3566 		 * fall-through to V2 (both L2 and L4) Sync.
3567 		 */
3568 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
3569 		/* Also time stamps V2 Path Delay Request/Response. */
3570 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3571 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3572 		is_l2 = true;
3573 		is_l4 = true;
3574 		break;
3575 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3576 		/* Hardware cannot filter just V2 L4 Delay Request messages;
3577 		 * fall-through to V2 (both L2 and L4) Delay Request.
3578 		 */
3579 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
3580 		/* Also time stamps V2 Path Delay Request/Response. */
3581 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3582 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3583 		is_l2 = true;
3584 		is_l4 = true;
3585 		break;
3586 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
3587 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
3588 		/* Hardware cannot filter just V2 L4 or L2 Event messages;
3589 		 * fall-through to all V2 (both L2 and L4) Events.
3590 		 */
3591 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
3592 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_EVENT_V2;
3593 		config->rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
3594 		is_l2 = true;
3595 		is_l4 = true;
3596 		break;
3597 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3598 		/* For V1, the hardware can only filter Sync messages or
3599 		 * Delay Request messages but not both so fall-through to
3600 		 * time stamp all packets.
3601 		 */
3602 	case HWTSTAMP_FILTER_ALL:
3603 		is_l2 = true;
3604 		is_l4 = true;
3605 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_ALL;
3606 		config->rx_filter = HWTSTAMP_FILTER_ALL;
3607 		break;
3608 	default:
3609 		return -ERANGE;
3610 	}
3611 
3612 	/* enable/disable Tx h/w time stamping */
3613 	regval = er32(TSYNCTXCTL);
3614 	regval &= ~E1000_TSYNCTXCTL_ENABLED;
3615 	regval |= tsync_tx_ctl;
3616 	ew32(TSYNCTXCTL, regval);
3617 	if ((er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) !=
3618 	    (regval & E1000_TSYNCTXCTL_ENABLED)) {
3619 		e_err("Timesync Tx Control register not set as expected\n");
3620 		return -EAGAIN;
3621 	}
3622 
3623 	/* enable/disable Rx h/w time stamping */
3624 	regval = er32(TSYNCRXCTL);
3625 	regval &= ~(E1000_TSYNCRXCTL_ENABLED | E1000_TSYNCRXCTL_TYPE_MASK);
3626 	regval |= tsync_rx_ctl;
3627 	ew32(TSYNCRXCTL, regval);
3628 	if ((er32(TSYNCRXCTL) & (E1000_TSYNCRXCTL_ENABLED |
3629 				 E1000_TSYNCRXCTL_TYPE_MASK)) !=
3630 	    (regval & (E1000_TSYNCRXCTL_ENABLED |
3631 		       E1000_TSYNCRXCTL_TYPE_MASK))) {
3632 		e_err("Timesync Rx Control register not set as expected\n");
3633 		return -EAGAIN;
3634 	}
3635 
3636 	/* L2: define ethertype filter for time stamped packets */
3637 	if (is_l2)
3638 		rxmtrl |= ETH_P_1588;
3639 
3640 	/* define which PTP packets get time stamped */
3641 	ew32(RXMTRL, rxmtrl);
3642 
3643 	/* Filter by destination port */
3644 	if (is_l4) {
3645 		rxudp = PTP_EV_PORT;
3646 		cpu_to_be16s(&rxudp);
3647 	}
3648 	ew32(RXUDP, rxudp);
3649 
3650 	e1e_flush();
3651 
3652 	/* Clear TSYNCRXCTL_VALID & TSYNCTXCTL_VALID bit */
3653 	er32(RXSTMPH);
3654 	er32(TXSTMPH);
3655 
3656 	/* Get and set the System Time Register SYSTIM base frequency */
3657 	ret_val = e1000e_get_base_timinca(adapter, &regval);
3658 	if (ret_val)
3659 		return ret_val;
3660 	ew32(TIMINCA, regval);
3661 
3662 	/* reset the ns time counter */
3663 	timecounter_init(&adapter->tc, &adapter->cc,
3664 			 ktime_to_ns(ktime_get_real()));
3665 
3666 	return 0;
3667 }
3668 
3669 /**
3670  * e1000_configure - configure the hardware for Rx and Tx
3671  * @adapter: private board structure
3672  **/
3673 static void e1000_configure(struct e1000_adapter *adapter)
3674 {
3675 	struct e1000_ring *rx_ring = adapter->rx_ring;
3676 
3677 	e1000e_set_rx_mode(adapter->netdev);
3678 
3679 	e1000_restore_vlan(adapter);
3680 	e1000_init_manageability_pt(adapter);
3681 
3682 	e1000_configure_tx(adapter);
3683 
3684 	if (adapter->netdev->features & NETIF_F_RXHASH)
3685 		e1000e_setup_rss_hash(adapter);
3686 	e1000_setup_rctl(adapter);
3687 	e1000_configure_rx(adapter);
3688 	adapter->alloc_rx_buf(rx_ring, e1000_desc_unused(rx_ring), GFP_KERNEL);
3689 }
3690 
3691 /**
3692  * e1000e_power_up_phy - restore link in case the phy was powered down
3693  * @adapter: address of board private structure
3694  *
3695  * The phy may be powered down to save power and turn off link when the
3696  * driver is unloaded and wake on lan is not enabled (among others)
3697  * *** this routine MUST be followed by a call to e1000e_reset ***
3698  **/
3699 void e1000e_power_up_phy(struct e1000_adapter *adapter)
3700 {
3701 	if (adapter->hw.phy.ops.power_up)
3702 		adapter->hw.phy.ops.power_up(&adapter->hw);
3703 
3704 	adapter->hw.mac.ops.setup_link(&adapter->hw);
3705 }
3706 
3707 /**
3708  * e1000_power_down_phy - Power down the PHY
3709  *
3710  * Power down the PHY so no link is implied when interface is down.
3711  * The PHY cannot be powered down if management or WoL is active.
3712  */
3713 static void e1000_power_down_phy(struct e1000_adapter *adapter)
3714 {
3715 	/* WoL is enabled */
3716 	if (adapter->wol)
3717 		return;
3718 
3719 	if (adapter->hw.phy.ops.power_down)
3720 		adapter->hw.phy.ops.power_down(&adapter->hw);
3721 }
3722 
3723 /**
3724  * e1000e_reset - bring the hardware into a known good state
3725  *
3726  * This function boots the hardware and enables some settings that
3727  * require a configuration cycle of the hardware - those cannot be
3728  * set/changed during runtime. After reset the device needs to be
3729  * properly configured for Rx, Tx etc.
3730  */
3731 void e1000e_reset(struct e1000_adapter *adapter)
3732 {
3733 	struct e1000_mac_info *mac = &adapter->hw.mac;
3734 	struct e1000_fc_info *fc = &adapter->hw.fc;
3735 	struct e1000_hw *hw = &adapter->hw;
3736 	u32 tx_space, min_tx_space, min_rx_space;
3737 	u32 pba = adapter->pba;
3738 	u16 hwm;
3739 
3740 	/* reset Packet Buffer Allocation to default */
3741 	ew32(PBA, pba);
3742 
3743 	if (adapter->max_frame_size > ETH_FRAME_LEN + ETH_FCS_LEN) {
3744 		/* To maintain wire speed transmits, the Tx FIFO should be
3745 		 * large enough to accommodate two full transmit packets,
3746 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
3747 		 * the Rx FIFO should be large enough to accommodate at least
3748 		 * one full receive packet and is similarly rounded up and
3749 		 * expressed in KB.
3750 		 */
3751 		pba = er32(PBA);
3752 		/* upper 16 bits has Tx packet buffer allocation size in KB */
3753 		tx_space = pba >> 16;
3754 		/* lower 16 bits has Rx packet buffer allocation size in KB */
3755 		pba &= 0xffff;
3756 		/* the Tx fifo also stores 16 bytes of information about the Tx
3757 		 * but don't include ethernet FCS because hardware appends it
3758 		 */
3759 		min_tx_space = (adapter->max_frame_size +
3760 				sizeof(struct e1000_tx_desc) -
3761 				ETH_FCS_LEN) * 2;
3762 		min_tx_space = ALIGN(min_tx_space, 1024);
3763 		min_tx_space >>= 10;
3764 		/* software strips receive CRC, so leave room for it */
3765 		min_rx_space = adapter->max_frame_size;
3766 		min_rx_space = ALIGN(min_rx_space, 1024);
3767 		min_rx_space >>= 10;
3768 
3769 		/* If current Tx allocation is less than the min Tx FIFO size,
3770 		 * and the min Tx FIFO size is less than the current Rx FIFO
3771 		 * allocation, take space away from current Rx allocation
3772 		 */
3773 		if ((tx_space < min_tx_space) &&
3774 		    ((min_tx_space - tx_space) < pba)) {
3775 			pba -= min_tx_space - tx_space;
3776 
3777 			/* if short on Rx space, Rx wins and must trump Tx
3778 			 * adjustment
3779 			 */
3780 			if (pba < min_rx_space)
3781 				pba = min_rx_space;
3782 		}
3783 
3784 		ew32(PBA, pba);
3785 	}
3786 
3787 	/* flow control settings
3788 	 *
3789 	 * The high water mark must be low enough to fit one full frame
3790 	 * (or the size used for early receive) above it in the Rx FIFO.
3791 	 * Set it to the lower of:
3792 	 * - 90% of the Rx FIFO size, and
3793 	 * - the full Rx FIFO size minus one full frame
3794 	 */
3795 	if (adapter->flags & FLAG_DISABLE_FC_PAUSE_TIME)
3796 		fc->pause_time = 0xFFFF;
3797 	else
3798 		fc->pause_time = E1000_FC_PAUSE_TIME;
3799 	fc->send_xon = true;
3800 	fc->current_mode = fc->requested_mode;
3801 
3802 	switch (hw->mac.type) {
3803 	case e1000_ich9lan:
3804 	case e1000_ich10lan:
3805 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
3806 			pba = 14;
3807 			ew32(PBA, pba);
3808 			fc->high_water = 0x2800;
3809 			fc->low_water = fc->high_water - 8;
3810 			break;
3811 		}
3812 		/* fall-through */
3813 	default:
3814 		hwm = min(((pba << 10) * 9 / 10),
3815 			  ((pba << 10) - adapter->max_frame_size));
3816 
3817 		fc->high_water = hwm & E1000_FCRTH_RTH; /* 8-byte granularity */
3818 		fc->low_water = fc->high_water - 8;
3819 		break;
3820 	case e1000_pchlan:
3821 		/* Workaround PCH LOM adapter hangs with certain network
3822 		 * loads.  If hangs persist, try disabling Tx flow control.
3823 		 */
3824 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
3825 			fc->high_water = 0x3500;
3826 			fc->low_water  = 0x1500;
3827 		} else {
3828 			fc->high_water = 0x5000;
3829 			fc->low_water  = 0x3000;
3830 		}
3831 		fc->refresh_time = 0x1000;
3832 		break;
3833 	case e1000_pch2lan:
3834 	case e1000_pch_lpt:
3835 		fc->refresh_time = 0x0400;
3836 
3837 		if (adapter->netdev->mtu <= ETH_DATA_LEN) {
3838 			fc->high_water = 0x05C20;
3839 			fc->low_water = 0x05048;
3840 			fc->pause_time = 0x0650;
3841 			break;
3842 		}
3843 
3844 		fc->high_water = ((pba << 10) * 9 / 10) & E1000_FCRTH_RTH;
3845 		fc->low_water = ((pba << 10) * 8 / 10) & E1000_FCRTL_RTL;
3846 		break;
3847 	}
3848 
3849 	/* Alignment of Tx data is on an arbitrary byte boundary with the
3850 	 * maximum size per Tx descriptor limited only to the transmit
3851 	 * allocation of the packet buffer minus 96 bytes with an upper
3852 	 * limit of 24KB due to receive synchronization limitations.
3853 	 */
3854 	adapter->tx_fifo_limit = min_t(u32, ((er32(PBA) >> 16) << 10) - 96,
3855 				       24 << 10);
3856 
3857 	/* Disable Adaptive Interrupt Moderation if 2 full packets cannot
3858 	 * fit in receive buffer.
3859 	 */
3860 	if (adapter->itr_setting & 0x3) {
3861 		if ((adapter->max_frame_size * 2) > (pba << 10)) {
3862 			if (!(adapter->flags2 & FLAG2_DISABLE_AIM)) {
3863 				dev_info(&adapter->pdev->dev,
3864 					"Interrupt Throttle Rate turned off\n");
3865 				adapter->flags2 |= FLAG2_DISABLE_AIM;
3866 				e1000e_write_itr(adapter, 0);
3867 			}
3868 		} else if (adapter->flags2 & FLAG2_DISABLE_AIM) {
3869 			dev_info(&adapter->pdev->dev,
3870 				 "Interrupt Throttle Rate turned on\n");
3871 			adapter->flags2 &= ~FLAG2_DISABLE_AIM;
3872 			adapter->itr = 20000;
3873 			e1000e_write_itr(adapter, adapter->itr);
3874 		}
3875 	}
3876 
3877 	/* Allow time for pending master requests to run */
3878 	mac->ops.reset_hw(hw);
3879 
3880 	/* For parts with AMT enabled, let the firmware know
3881 	 * that the network interface is in control
3882 	 */
3883 	if (adapter->flags & FLAG_HAS_AMT)
3884 		e1000e_get_hw_control(adapter);
3885 
3886 	ew32(WUC, 0);
3887 
3888 	if (mac->ops.init_hw(hw))
3889 		e_err("Hardware Error\n");
3890 
3891 	e1000_update_mng_vlan(adapter);
3892 
3893 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
3894 	ew32(VET, ETH_P_8021Q);
3895 
3896 	e1000e_reset_adaptive(hw);
3897 
3898 	/* initialize systim and reset the ns time counter */
3899 	e1000e_config_hwtstamp(adapter);
3900 
3901 	if (!netif_running(adapter->netdev) &&
3902 	    !test_bit(__E1000_TESTING, &adapter->state)) {
3903 		e1000_power_down_phy(adapter);
3904 		return;
3905 	}
3906 
3907 	e1000_get_phy_info(hw);
3908 
3909 	if ((adapter->flags & FLAG_HAS_SMART_POWER_DOWN) &&
3910 	    !(adapter->flags & FLAG_SMART_POWER_DOWN)) {
3911 		u16 phy_data = 0;
3912 		/* speed up time to link by disabling smart power down, ignore
3913 		 * the return value of this function because there is nothing
3914 		 * different we would do if it failed
3915 		 */
3916 		e1e_rphy(hw, IGP02E1000_PHY_POWER_MGMT, &phy_data);
3917 		phy_data &= ~IGP02E1000_PM_SPD;
3918 		e1e_wphy(hw, IGP02E1000_PHY_POWER_MGMT, phy_data);
3919 	}
3920 }
3921 
3922 int e1000e_up(struct e1000_adapter *adapter)
3923 {
3924 	struct e1000_hw *hw = &adapter->hw;
3925 
3926 	/* hardware has been reset, we need to reload some things */
3927 	e1000_configure(adapter);
3928 
3929 	clear_bit(__E1000_DOWN, &adapter->state);
3930 
3931 	if (adapter->msix_entries)
3932 		e1000_configure_msix(adapter);
3933 	e1000_irq_enable(adapter);
3934 
3935 	netif_start_queue(adapter->netdev);
3936 
3937 	/* fire a link change interrupt to start the watchdog */
3938 	if (adapter->msix_entries)
3939 		ew32(ICS, E1000_ICS_LSC | E1000_ICR_OTHER);
3940 	else
3941 		ew32(ICS, E1000_ICS_LSC);
3942 
3943 	return 0;
3944 }
3945 
3946 static void e1000e_flush_descriptors(struct e1000_adapter *adapter)
3947 {
3948 	struct e1000_hw *hw = &adapter->hw;
3949 
3950 	if (!(adapter->flags2 & FLAG2_DMA_BURST))
3951 		return;
3952 
3953 	/* flush pending descriptor writebacks to memory */
3954 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
3955 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
3956 
3957 	/* execute the writes immediately */
3958 	e1e_flush();
3959 
3960 	/* due to rare timing issues, write to TIDV/RDTR again to ensure the
3961 	 * write is successful
3962 	 */
3963 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
3964 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
3965 
3966 	/* execute the writes immediately */
3967 	e1e_flush();
3968 }
3969 
3970 static void e1000e_update_stats(struct e1000_adapter *adapter);
3971 
3972 void e1000e_down(struct e1000_adapter *adapter)
3973 {
3974 	struct net_device *netdev = adapter->netdev;
3975 	struct e1000_hw *hw = &adapter->hw;
3976 	u32 tctl, rctl;
3977 
3978 	/* signal that we're down so the interrupt handler does not
3979 	 * reschedule our watchdog timer
3980 	 */
3981 	set_bit(__E1000_DOWN, &adapter->state);
3982 
3983 	/* disable receives in the hardware */
3984 	rctl = er32(RCTL);
3985 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
3986 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
3987 	/* flush and sleep below */
3988 
3989 	netif_stop_queue(netdev);
3990 
3991 	/* disable transmits in the hardware */
3992 	tctl = er32(TCTL);
3993 	tctl &= ~E1000_TCTL_EN;
3994 	ew32(TCTL, tctl);
3995 
3996 	/* flush both disables and wait for them to finish */
3997 	e1e_flush();
3998 	usleep_range(10000, 20000);
3999 
4000 	e1000_irq_disable(adapter);
4001 
4002 	del_timer_sync(&adapter->watchdog_timer);
4003 	del_timer_sync(&adapter->phy_info_timer);
4004 
4005 	netif_carrier_off(netdev);
4006 
4007 	spin_lock(&adapter->stats64_lock);
4008 	e1000e_update_stats(adapter);
4009 	spin_unlock(&adapter->stats64_lock);
4010 
4011 	e1000e_flush_descriptors(adapter);
4012 	e1000_clean_tx_ring(adapter->tx_ring);
4013 	e1000_clean_rx_ring(adapter->rx_ring);
4014 
4015 	adapter->link_speed = 0;
4016 	adapter->link_duplex = 0;
4017 
4018 	if (!pci_channel_offline(adapter->pdev))
4019 		e1000e_reset(adapter);
4020 
4021 	/* TODO: for power management, we could drop the link and
4022 	 * pci_disable_device here.
4023 	 */
4024 }
4025 
4026 void e1000e_reinit_locked(struct e1000_adapter *adapter)
4027 {
4028 	might_sleep();
4029 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
4030 		usleep_range(1000, 2000);
4031 	e1000e_down(adapter);
4032 	e1000e_up(adapter);
4033 	clear_bit(__E1000_RESETTING, &adapter->state);
4034 }
4035 
4036 /**
4037  * e1000e_cyclecounter_read - read raw cycle counter (used by time counter)
4038  * @cc: cyclecounter structure
4039  **/
4040 static cycle_t e1000e_cyclecounter_read(const struct cyclecounter *cc)
4041 {
4042 	struct e1000_adapter *adapter = container_of(cc, struct e1000_adapter,
4043 						     cc);
4044 	struct e1000_hw *hw = &adapter->hw;
4045 	cycle_t systim;
4046 
4047 	/* latch SYSTIMH on read of SYSTIML */
4048 	systim = (cycle_t)er32(SYSTIML);
4049 	systim |= (cycle_t)er32(SYSTIMH) << 32;
4050 
4051 	return systim;
4052 }
4053 
4054 /**
4055  * e1000_sw_init - Initialize general software structures (struct e1000_adapter)
4056  * @adapter: board private structure to initialize
4057  *
4058  * e1000_sw_init initializes the Adapter private data structure.
4059  * Fields are initialized based on PCI device information and
4060  * OS network device settings (MTU size).
4061  **/
4062 static int e1000_sw_init(struct e1000_adapter *adapter)
4063 {
4064 	struct net_device *netdev = adapter->netdev;
4065 
4066 	adapter->rx_buffer_len = ETH_FRAME_LEN + VLAN_HLEN + ETH_FCS_LEN;
4067 	adapter->rx_ps_bsize0 = 128;
4068 	adapter->max_frame_size = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4069 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4070 	adapter->tx_ring_count = E1000_DEFAULT_TXD;
4071 	adapter->rx_ring_count = E1000_DEFAULT_RXD;
4072 
4073 	spin_lock_init(&adapter->stats64_lock);
4074 
4075 	e1000e_set_interrupt_capability(adapter);
4076 
4077 	if (e1000_alloc_queues(adapter))
4078 		return -ENOMEM;
4079 
4080 	/* Setup hardware time stamping cyclecounter */
4081 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
4082 		adapter->cc.read = e1000e_cyclecounter_read;
4083 		adapter->cc.mask = CLOCKSOURCE_MASK(64);
4084 		adapter->cc.mult = 1;
4085 		/* cc.shift set in e1000e_get_base_tininca() */
4086 
4087 		spin_lock_init(&adapter->systim_lock);
4088 		INIT_WORK(&adapter->tx_hwtstamp_work, e1000e_tx_hwtstamp_work);
4089 	}
4090 
4091 	/* Explicitly disable IRQ since the NIC can be in any state. */
4092 	e1000_irq_disable(adapter);
4093 
4094 	set_bit(__E1000_DOWN, &adapter->state);
4095 	return 0;
4096 }
4097 
4098 /**
4099  * e1000_intr_msi_test - Interrupt Handler
4100  * @irq: interrupt number
4101  * @data: pointer to a network interface device structure
4102  **/
4103 static irqreturn_t e1000_intr_msi_test(int __always_unused irq, void *data)
4104 {
4105 	struct net_device *netdev = data;
4106 	struct e1000_adapter *adapter = netdev_priv(netdev);
4107 	struct e1000_hw *hw = &adapter->hw;
4108 	u32 icr = er32(ICR);
4109 
4110 	e_dbg("icr is %08X\n", icr);
4111 	if (icr & E1000_ICR_RXSEQ) {
4112 		adapter->flags &= ~FLAG_MSI_TEST_FAILED;
4113 		/* Force memory writes to complete before acknowledging the
4114 		 * interrupt is handled.
4115 		 */
4116 		wmb();
4117 	}
4118 
4119 	return IRQ_HANDLED;
4120 }
4121 
4122 /**
4123  * e1000_test_msi_interrupt - Returns 0 for successful test
4124  * @adapter: board private struct
4125  *
4126  * code flow taken from tg3.c
4127  **/
4128 static int e1000_test_msi_interrupt(struct e1000_adapter *adapter)
4129 {
4130 	struct net_device *netdev = adapter->netdev;
4131 	struct e1000_hw *hw = &adapter->hw;
4132 	int err;
4133 
4134 	/* poll_enable hasn't been called yet, so don't need disable */
4135 	/* clear any pending events */
4136 	er32(ICR);
4137 
4138 	/* free the real vector and request a test handler */
4139 	e1000_free_irq(adapter);
4140 	e1000e_reset_interrupt_capability(adapter);
4141 
4142 	/* Assume that the test fails, if it succeeds then the test
4143 	 * MSI irq handler will unset this flag
4144 	 */
4145 	adapter->flags |= FLAG_MSI_TEST_FAILED;
4146 
4147 	err = pci_enable_msi(adapter->pdev);
4148 	if (err)
4149 		goto msi_test_failed;
4150 
4151 	err = request_irq(adapter->pdev->irq, e1000_intr_msi_test, 0,
4152 			  netdev->name, netdev);
4153 	if (err) {
4154 		pci_disable_msi(adapter->pdev);
4155 		goto msi_test_failed;
4156 	}
4157 
4158 	/* Force memory writes to complete before enabling and firing an
4159 	 * interrupt.
4160 	 */
4161 	wmb();
4162 
4163 	e1000_irq_enable(adapter);
4164 
4165 	/* fire an unusual interrupt on the test handler */
4166 	ew32(ICS, E1000_ICS_RXSEQ);
4167 	e1e_flush();
4168 	msleep(100);
4169 
4170 	e1000_irq_disable(adapter);
4171 
4172 	rmb();			/* read flags after interrupt has been fired */
4173 
4174 	if (adapter->flags & FLAG_MSI_TEST_FAILED) {
4175 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
4176 		e_info("MSI interrupt test failed, using legacy interrupt.\n");
4177 	} else {
4178 		e_dbg("MSI interrupt test succeeded!\n");
4179 	}
4180 
4181 	free_irq(adapter->pdev->irq, netdev);
4182 	pci_disable_msi(adapter->pdev);
4183 
4184 msi_test_failed:
4185 	e1000e_set_interrupt_capability(adapter);
4186 	return e1000_request_irq(adapter);
4187 }
4188 
4189 /**
4190  * e1000_test_msi - Returns 0 if MSI test succeeds or INTx mode is restored
4191  * @adapter: board private struct
4192  *
4193  * code flow taken from tg3.c, called with e1000 interrupts disabled.
4194  **/
4195 static int e1000_test_msi(struct e1000_adapter *adapter)
4196 {
4197 	int err;
4198 	u16 pci_cmd;
4199 
4200 	if (!(adapter->flags & FLAG_MSI_ENABLED))
4201 		return 0;
4202 
4203 	/* disable SERR in case the MSI write causes a master abort */
4204 	pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4205 	if (pci_cmd & PCI_COMMAND_SERR)
4206 		pci_write_config_word(adapter->pdev, PCI_COMMAND,
4207 				      pci_cmd & ~PCI_COMMAND_SERR);
4208 
4209 	err = e1000_test_msi_interrupt(adapter);
4210 
4211 	/* re-enable SERR */
4212 	if (pci_cmd & PCI_COMMAND_SERR) {
4213 		pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4214 		pci_cmd |= PCI_COMMAND_SERR;
4215 		pci_write_config_word(adapter->pdev, PCI_COMMAND, pci_cmd);
4216 	}
4217 
4218 	return err;
4219 }
4220 
4221 /**
4222  * e1000_open - Called when a network interface is made active
4223  * @netdev: network interface device structure
4224  *
4225  * Returns 0 on success, negative value on failure
4226  *
4227  * The open entry point is called when a network interface is made
4228  * active by the system (IFF_UP).  At this point all resources needed
4229  * for transmit and receive operations are allocated, the interrupt
4230  * handler is registered with the OS, the watchdog timer is started,
4231  * and the stack is notified that the interface is ready.
4232  **/
4233 static int e1000_open(struct net_device *netdev)
4234 {
4235 	struct e1000_adapter *adapter = netdev_priv(netdev);
4236 	struct e1000_hw *hw = &adapter->hw;
4237 	struct pci_dev *pdev = adapter->pdev;
4238 	int err;
4239 
4240 	/* disallow open during test */
4241 	if (test_bit(__E1000_TESTING, &adapter->state))
4242 		return -EBUSY;
4243 
4244 	pm_runtime_get_sync(&pdev->dev);
4245 
4246 	netif_carrier_off(netdev);
4247 
4248 	/* allocate transmit descriptors */
4249 	err = e1000e_setup_tx_resources(adapter->tx_ring);
4250 	if (err)
4251 		goto err_setup_tx;
4252 
4253 	/* allocate receive descriptors */
4254 	err = e1000e_setup_rx_resources(adapter->rx_ring);
4255 	if (err)
4256 		goto err_setup_rx;
4257 
4258 	/* If AMT is enabled, let the firmware know that the network
4259 	 * interface is now open and reset the part to a known state.
4260 	 */
4261 	if (adapter->flags & FLAG_HAS_AMT) {
4262 		e1000e_get_hw_control(adapter);
4263 		e1000e_reset(adapter);
4264 	}
4265 
4266 	e1000e_power_up_phy(adapter);
4267 
4268 	adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
4269 	if ((adapter->hw.mng_cookie.status &
4270 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
4271 		e1000_update_mng_vlan(adapter);
4272 
4273 	/* DMA latency requirement to workaround jumbo issue */
4274 	pm_qos_add_request(&adapter->netdev->pm_qos_req, PM_QOS_CPU_DMA_LATENCY,
4275 			   PM_QOS_DEFAULT_VALUE);
4276 
4277 	/* before we allocate an interrupt, we must be ready to handle it.
4278 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4279 	 * as soon as we call pci_request_irq, so we have to setup our
4280 	 * clean_rx handler before we do so.
4281 	 */
4282 	e1000_configure(adapter);
4283 
4284 	err = e1000_request_irq(adapter);
4285 	if (err)
4286 		goto err_req_irq;
4287 
4288 	/* Work around PCIe errata with MSI interrupts causing some chipsets to
4289 	 * ignore e1000e MSI messages, which means we need to test our MSI
4290 	 * interrupt now
4291 	 */
4292 	if (adapter->int_mode != E1000E_INT_MODE_LEGACY) {
4293 		err = e1000_test_msi(adapter);
4294 		if (err) {
4295 			e_err("Interrupt allocation failed\n");
4296 			goto err_req_irq;
4297 		}
4298 	}
4299 
4300 	/* From here on the code is the same as e1000e_up() */
4301 	clear_bit(__E1000_DOWN, &adapter->state);
4302 
4303 	napi_enable(&adapter->napi);
4304 
4305 	e1000_irq_enable(adapter);
4306 
4307 	adapter->tx_hang_recheck = false;
4308 	netif_start_queue(netdev);
4309 
4310 	adapter->idle_check = true;
4311 	hw->mac.get_link_status = true;
4312 	pm_runtime_put(&pdev->dev);
4313 
4314 	/* fire a link status change interrupt to start the watchdog */
4315 	if (adapter->msix_entries)
4316 		ew32(ICS, E1000_ICS_LSC | E1000_ICR_OTHER);
4317 	else
4318 		ew32(ICS, E1000_ICS_LSC);
4319 
4320 	return 0;
4321 
4322 err_req_irq:
4323 	e1000e_release_hw_control(adapter);
4324 	e1000_power_down_phy(adapter);
4325 	e1000e_free_rx_resources(adapter->rx_ring);
4326 err_setup_rx:
4327 	e1000e_free_tx_resources(adapter->tx_ring);
4328 err_setup_tx:
4329 	e1000e_reset(adapter);
4330 	pm_runtime_put_sync(&pdev->dev);
4331 
4332 	return err;
4333 }
4334 
4335 /**
4336  * e1000_close - Disables a network interface
4337  * @netdev: network interface device structure
4338  *
4339  * Returns 0, this is not allowed to fail
4340  *
4341  * The close entry point is called when an interface is de-activated
4342  * by the OS.  The hardware is still under the drivers control, but
4343  * needs to be disabled.  A global MAC reset is issued to stop the
4344  * hardware, and all transmit and receive resources are freed.
4345  **/
4346 static int e1000_close(struct net_device *netdev)
4347 {
4348 	struct e1000_adapter *adapter = netdev_priv(netdev);
4349 	struct pci_dev *pdev = adapter->pdev;
4350 	int count = E1000_CHECK_RESET_COUNT;
4351 
4352 	while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
4353 		usleep_range(10000, 20000);
4354 
4355 	WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
4356 
4357 	pm_runtime_get_sync(&pdev->dev);
4358 
4359 	napi_disable(&adapter->napi);
4360 
4361 	if (!test_bit(__E1000_DOWN, &adapter->state)) {
4362 		e1000e_down(adapter);
4363 		e1000_free_irq(adapter);
4364 	}
4365 	e1000_power_down_phy(adapter);
4366 
4367 	e1000e_free_tx_resources(adapter->tx_ring);
4368 	e1000e_free_rx_resources(adapter->rx_ring);
4369 
4370 	/* kill manageability vlan ID if supported, but not if a vlan with
4371 	 * the same ID is registered on the host OS (let 8021q kill it)
4372 	 */
4373 	if (adapter->hw.mng_cookie.status &
4374 	    E1000_MNG_DHCP_COOKIE_STATUS_VLAN)
4375 		e1000_vlan_rx_kill_vid(netdev, adapter->mng_vlan_id);
4376 
4377 	/* If AMT is enabled, let the firmware know that the network
4378 	 * interface is now closed
4379 	 */
4380 	if ((adapter->flags & FLAG_HAS_AMT) &&
4381 	    !test_bit(__E1000_TESTING, &adapter->state))
4382 		e1000e_release_hw_control(adapter);
4383 
4384 	pm_qos_remove_request(&adapter->netdev->pm_qos_req);
4385 
4386 	pm_runtime_put_sync(&pdev->dev);
4387 
4388 	return 0;
4389 }
4390 /**
4391  * e1000_set_mac - Change the Ethernet Address of the NIC
4392  * @netdev: network interface device structure
4393  * @p: pointer to an address structure
4394  *
4395  * Returns 0 on success, negative on failure
4396  **/
4397 static int e1000_set_mac(struct net_device *netdev, void *p)
4398 {
4399 	struct e1000_adapter *adapter = netdev_priv(netdev);
4400 	struct e1000_hw *hw = &adapter->hw;
4401 	struct sockaddr *addr = p;
4402 
4403 	if (!is_valid_ether_addr(addr->sa_data))
4404 		return -EADDRNOTAVAIL;
4405 
4406 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
4407 	memcpy(adapter->hw.mac.addr, addr->sa_data, netdev->addr_len);
4408 
4409 	hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
4410 
4411 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA) {
4412 		/* activate the work around */
4413 		e1000e_set_laa_state_82571(&adapter->hw, 1);
4414 
4415 		/* Hold a copy of the LAA in RAR[14] This is done so that
4416 		 * between the time RAR[0] gets clobbered  and the time it
4417 		 * gets fixed (in e1000_watchdog), the actual LAA is in one
4418 		 * of the RARs and no incoming packets directed to this port
4419 		 * are dropped. Eventually the LAA will be in RAR[0] and
4420 		 * RAR[14]
4421 		 */
4422 		hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr,
4423 				    adapter->hw.mac.rar_entry_count - 1);
4424 	}
4425 
4426 	return 0;
4427 }
4428 
4429 /**
4430  * e1000e_update_phy_task - work thread to update phy
4431  * @work: pointer to our work struct
4432  *
4433  * this worker thread exists because we must acquire a
4434  * semaphore to read the phy, which we could msleep while
4435  * waiting for it, and we can't msleep in a timer.
4436  **/
4437 static void e1000e_update_phy_task(struct work_struct *work)
4438 {
4439 	struct e1000_adapter *adapter = container_of(work,
4440 					struct e1000_adapter, update_phy_task);
4441 
4442 	if (test_bit(__E1000_DOWN, &adapter->state))
4443 		return;
4444 
4445 	e1000_get_phy_info(&adapter->hw);
4446 }
4447 
4448 /**
4449  * e1000_update_phy_info - timre call-back to update PHY info
4450  * @data: pointer to adapter cast into an unsigned long
4451  *
4452  * Need to wait a few seconds after link up to get diagnostic information from
4453  * the phy
4454  **/
4455 static void e1000_update_phy_info(unsigned long data)
4456 {
4457 	struct e1000_adapter *adapter = (struct e1000_adapter *) data;
4458 
4459 	if (test_bit(__E1000_DOWN, &adapter->state))
4460 		return;
4461 
4462 	schedule_work(&adapter->update_phy_task);
4463 }
4464 
4465 /**
4466  * e1000e_update_phy_stats - Update the PHY statistics counters
4467  * @adapter: board private structure
4468  *
4469  * Read/clear the upper 16-bit PHY registers and read/accumulate lower
4470  **/
4471 static void e1000e_update_phy_stats(struct e1000_adapter *adapter)
4472 {
4473 	struct e1000_hw *hw = &adapter->hw;
4474 	s32 ret_val;
4475 	u16 phy_data;
4476 
4477 	ret_val = hw->phy.ops.acquire(hw);
4478 	if (ret_val)
4479 		return;
4480 
4481 	/* A page set is expensive so check if already on desired page.
4482 	 * If not, set to the page with the PHY status registers.
4483 	 */
4484 	hw->phy.addr = 1;
4485 	ret_val = e1000e_read_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4486 					   &phy_data);
4487 	if (ret_val)
4488 		goto release;
4489 	if (phy_data != (HV_STATS_PAGE << IGP_PAGE_SHIFT)) {
4490 		ret_val = hw->phy.ops.set_page(hw,
4491 					       HV_STATS_PAGE << IGP_PAGE_SHIFT);
4492 		if (ret_val)
4493 			goto release;
4494 	}
4495 
4496 	/* Single Collision Count */
4497 	hw->phy.ops.read_reg_page(hw, HV_SCC_UPPER, &phy_data);
4498 	ret_val = hw->phy.ops.read_reg_page(hw, HV_SCC_LOWER, &phy_data);
4499 	if (!ret_val)
4500 		adapter->stats.scc += phy_data;
4501 
4502 	/* Excessive Collision Count */
4503 	hw->phy.ops.read_reg_page(hw, HV_ECOL_UPPER, &phy_data);
4504 	ret_val = hw->phy.ops.read_reg_page(hw, HV_ECOL_LOWER, &phy_data);
4505 	if (!ret_val)
4506 		adapter->stats.ecol += phy_data;
4507 
4508 	/* Multiple Collision Count */
4509 	hw->phy.ops.read_reg_page(hw, HV_MCC_UPPER, &phy_data);
4510 	ret_val = hw->phy.ops.read_reg_page(hw, HV_MCC_LOWER, &phy_data);
4511 	if (!ret_val)
4512 		adapter->stats.mcc += phy_data;
4513 
4514 	/* Late Collision Count */
4515 	hw->phy.ops.read_reg_page(hw, HV_LATECOL_UPPER, &phy_data);
4516 	ret_val = hw->phy.ops.read_reg_page(hw, HV_LATECOL_LOWER, &phy_data);
4517 	if (!ret_val)
4518 		adapter->stats.latecol += phy_data;
4519 
4520 	/* Collision Count - also used for adaptive IFS */
4521 	hw->phy.ops.read_reg_page(hw, HV_COLC_UPPER, &phy_data);
4522 	ret_val = hw->phy.ops.read_reg_page(hw, HV_COLC_LOWER, &phy_data);
4523 	if (!ret_val)
4524 		hw->mac.collision_delta = phy_data;
4525 
4526 	/* Defer Count */
4527 	hw->phy.ops.read_reg_page(hw, HV_DC_UPPER, &phy_data);
4528 	ret_val = hw->phy.ops.read_reg_page(hw, HV_DC_LOWER, &phy_data);
4529 	if (!ret_val)
4530 		adapter->stats.dc += phy_data;
4531 
4532 	/* Transmit with no CRS */
4533 	hw->phy.ops.read_reg_page(hw, HV_TNCRS_UPPER, &phy_data);
4534 	ret_val = hw->phy.ops.read_reg_page(hw, HV_TNCRS_LOWER, &phy_data);
4535 	if (!ret_val)
4536 		adapter->stats.tncrs += phy_data;
4537 
4538 release:
4539 	hw->phy.ops.release(hw);
4540 }
4541 
4542 /**
4543  * e1000e_update_stats - Update the board statistics counters
4544  * @adapter: board private structure
4545  **/
4546 static void e1000e_update_stats(struct e1000_adapter *adapter)
4547 {
4548 	struct net_device *netdev = adapter->netdev;
4549 	struct e1000_hw *hw = &adapter->hw;
4550 	struct pci_dev *pdev = adapter->pdev;
4551 
4552 	/* Prevent stats update while adapter is being reset, or if the pci
4553 	 * connection is down.
4554 	 */
4555 	if (adapter->link_speed == 0)
4556 		return;
4557 	if (pci_channel_offline(pdev))
4558 		return;
4559 
4560 	adapter->stats.crcerrs += er32(CRCERRS);
4561 	adapter->stats.gprc += er32(GPRC);
4562 	adapter->stats.gorc += er32(GORCL);
4563 	er32(GORCH); /* Clear gorc */
4564 	adapter->stats.bprc += er32(BPRC);
4565 	adapter->stats.mprc += er32(MPRC);
4566 	adapter->stats.roc += er32(ROC);
4567 
4568 	adapter->stats.mpc += er32(MPC);
4569 
4570 	/* Half-duplex statistics */
4571 	if (adapter->link_duplex == HALF_DUPLEX) {
4572 		if (adapter->flags2 & FLAG2_HAS_PHY_STATS) {
4573 			e1000e_update_phy_stats(adapter);
4574 		} else {
4575 			adapter->stats.scc += er32(SCC);
4576 			adapter->stats.ecol += er32(ECOL);
4577 			adapter->stats.mcc += er32(MCC);
4578 			adapter->stats.latecol += er32(LATECOL);
4579 			adapter->stats.dc += er32(DC);
4580 
4581 			hw->mac.collision_delta = er32(COLC);
4582 
4583 			if ((hw->mac.type != e1000_82574) &&
4584 			    (hw->mac.type != e1000_82583))
4585 				adapter->stats.tncrs += er32(TNCRS);
4586 		}
4587 		adapter->stats.colc += hw->mac.collision_delta;
4588 	}
4589 
4590 	adapter->stats.xonrxc += er32(XONRXC);
4591 	adapter->stats.xontxc += er32(XONTXC);
4592 	adapter->stats.xoffrxc += er32(XOFFRXC);
4593 	adapter->stats.xofftxc += er32(XOFFTXC);
4594 	adapter->stats.gptc += er32(GPTC);
4595 	adapter->stats.gotc += er32(GOTCL);
4596 	er32(GOTCH); /* Clear gotc */
4597 	adapter->stats.rnbc += er32(RNBC);
4598 	adapter->stats.ruc += er32(RUC);
4599 
4600 	adapter->stats.mptc += er32(MPTC);
4601 	adapter->stats.bptc += er32(BPTC);
4602 
4603 	/* used for adaptive IFS */
4604 
4605 	hw->mac.tx_packet_delta = er32(TPT);
4606 	adapter->stats.tpt += hw->mac.tx_packet_delta;
4607 
4608 	adapter->stats.algnerrc += er32(ALGNERRC);
4609 	adapter->stats.rxerrc += er32(RXERRC);
4610 	adapter->stats.cexterr += er32(CEXTERR);
4611 	adapter->stats.tsctc += er32(TSCTC);
4612 	adapter->stats.tsctfc += er32(TSCTFC);
4613 
4614 	/* Fill out the OS statistics structure */
4615 	netdev->stats.multicast = adapter->stats.mprc;
4616 	netdev->stats.collisions = adapter->stats.colc;
4617 
4618 	/* Rx Errors */
4619 
4620 	/* RLEC on some newer hardware can be incorrect so build
4621 	 * our own version based on RUC and ROC
4622 	 */
4623 	netdev->stats.rx_errors = adapter->stats.rxerrc +
4624 		adapter->stats.crcerrs + adapter->stats.algnerrc +
4625 		adapter->stats.ruc + adapter->stats.roc +
4626 		adapter->stats.cexterr;
4627 	netdev->stats.rx_length_errors = adapter->stats.ruc +
4628 					      adapter->stats.roc;
4629 	netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
4630 	netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
4631 	netdev->stats.rx_missed_errors = adapter->stats.mpc;
4632 
4633 	/* Tx Errors */
4634 	netdev->stats.tx_errors = adapter->stats.ecol +
4635 				       adapter->stats.latecol;
4636 	netdev->stats.tx_aborted_errors = adapter->stats.ecol;
4637 	netdev->stats.tx_window_errors = adapter->stats.latecol;
4638 	netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
4639 
4640 	/* Tx Dropped needs to be maintained elsewhere */
4641 
4642 	/* Management Stats */
4643 	adapter->stats.mgptc += er32(MGTPTC);
4644 	adapter->stats.mgprc += er32(MGTPRC);
4645 	adapter->stats.mgpdc += er32(MGTPDC);
4646 
4647 	/* Correctable ECC Errors */
4648 	if (hw->mac.type == e1000_pch_lpt) {
4649 		u32 pbeccsts = er32(PBECCSTS);
4650 		adapter->corr_errors +=
4651 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
4652 		adapter->uncorr_errors +=
4653 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
4654 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
4655 	}
4656 }
4657 
4658 /**
4659  * e1000_phy_read_status - Update the PHY register status snapshot
4660  * @adapter: board private structure
4661  **/
4662 static void e1000_phy_read_status(struct e1000_adapter *adapter)
4663 {
4664 	struct e1000_hw *hw = &adapter->hw;
4665 	struct e1000_phy_regs *phy = &adapter->phy_regs;
4666 
4667 	if ((er32(STATUS) & E1000_STATUS_LU) &&
4668 	    (adapter->hw.phy.media_type == e1000_media_type_copper)) {
4669 		int ret_val;
4670 
4671 		pm_runtime_get_sync(&adapter->pdev->dev);
4672 		ret_val = e1e_rphy(hw, MII_BMCR, &phy->bmcr);
4673 		ret_val |= e1e_rphy(hw, MII_BMSR, &phy->bmsr);
4674 		ret_val |= e1e_rphy(hw, MII_ADVERTISE, &phy->advertise);
4675 		ret_val |= e1e_rphy(hw, MII_LPA, &phy->lpa);
4676 		ret_val |= e1e_rphy(hw, MII_EXPANSION, &phy->expansion);
4677 		ret_val |= e1e_rphy(hw, MII_CTRL1000, &phy->ctrl1000);
4678 		ret_val |= e1e_rphy(hw, MII_STAT1000, &phy->stat1000);
4679 		ret_val |= e1e_rphy(hw, MII_ESTATUS, &phy->estatus);
4680 		if (ret_val)
4681 			e_warn("Error reading PHY register\n");
4682 		pm_runtime_put_sync(&adapter->pdev->dev);
4683 	} else {
4684 		/* Do not read PHY registers if link is not up
4685 		 * Set values to typical power-on defaults
4686 		 */
4687 		phy->bmcr = (BMCR_SPEED1000 | BMCR_ANENABLE | BMCR_FULLDPLX);
4688 		phy->bmsr = (BMSR_100FULL | BMSR_100HALF | BMSR_10FULL |
4689 			     BMSR_10HALF | BMSR_ESTATEN | BMSR_ANEGCAPABLE |
4690 			     BMSR_ERCAP);
4691 		phy->advertise = (ADVERTISE_PAUSE_ASYM | ADVERTISE_PAUSE_CAP |
4692 				  ADVERTISE_ALL | ADVERTISE_CSMA);
4693 		phy->lpa = 0;
4694 		phy->expansion = EXPANSION_ENABLENPAGE;
4695 		phy->ctrl1000 = ADVERTISE_1000FULL;
4696 		phy->stat1000 = 0;
4697 		phy->estatus = (ESTATUS_1000_TFULL | ESTATUS_1000_THALF);
4698 	}
4699 }
4700 
4701 static void e1000_print_link_info(struct e1000_adapter *adapter)
4702 {
4703 	struct e1000_hw *hw = &adapter->hw;
4704 	u32 ctrl = er32(CTRL);
4705 
4706 	/* Link status message must follow this format for user tools */
4707 	pr_info("%s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
4708 		adapter->netdev->name, adapter->link_speed,
4709 		adapter->link_duplex == FULL_DUPLEX ? "Full" : "Half",
4710 		(ctrl & E1000_CTRL_TFCE) && (ctrl & E1000_CTRL_RFCE) ? "Rx/Tx" :
4711 		(ctrl & E1000_CTRL_RFCE) ? "Rx" :
4712 		(ctrl & E1000_CTRL_TFCE) ? "Tx" : "None");
4713 }
4714 
4715 static bool e1000e_has_link(struct e1000_adapter *adapter)
4716 {
4717 	struct e1000_hw *hw = &adapter->hw;
4718 	bool link_active = false;
4719 	s32 ret_val = 0;
4720 
4721 	/* get_link_status is set on LSC (link status) interrupt or
4722 	 * Rx sequence error interrupt.  get_link_status will stay
4723 	 * false until the check_for_link establishes link
4724 	 * for copper adapters ONLY
4725 	 */
4726 	switch (hw->phy.media_type) {
4727 	case e1000_media_type_copper:
4728 		if (hw->mac.get_link_status) {
4729 			ret_val = hw->mac.ops.check_for_link(hw);
4730 			link_active = !hw->mac.get_link_status;
4731 		} else {
4732 			link_active = true;
4733 		}
4734 		break;
4735 	case e1000_media_type_fiber:
4736 		ret_val = hw->mac.ops.check_for_link(hw);
4737 		link_active = !!(er32(STATUS) & E1000_STATUS_LU);
4738 		break;
4739 	case e1000_media_type_internal_serdes:
4740 		ret_val = hw->mac.ops.check_for_link(hw);
4741 		link_active = adapter->hw.mac.serdes_has_link;
4742 		break;
4743 	default:
4744 	case e1000_media_type_unknown:
4745 		break;
4746 	}
4747 
4748 	if ((ret_val == E1000_ERR_PHY) && (hw->phy.type == e1000_phy_igp_3) &&
4749 	    (er32(CTRL) & E1000_PHY_CTRL_GBE_DISABLE)) {
4750 		/* See e1000_kmrn_lock_loss_workaround_ich8lan() */
4751 		e_info("Gigabit has been disabled, downgrading speed\n");
4752 	}
4753 
4754 	return link_active;
4755 }
4756 
4757 static void e1000e_enable_receives(struct e1000_adapter *adapter)
4758 {
4759 	/* make sure the receive unit is started */
4760 	if ((adapter->flags & FLAG_RX_NEEDS_RESTART) &&
4761 	    (adapter->flags & FLAG_RESTART_NOW)) {
4762 		struct e1000_hw *hw = &adapter->hw;
4763 		u32 rctl = er32(RCTL);
4764 		ew32(RCTL, rctl | E1000_RCTL_EN);
4765 		adapter->flags &= ~FLAG_RESTART_NOW;
4766 	}
4767 }
4768 
4769 static void e1000e_check_82574_phy_workaround(struct e1000_adapter *adapter)
4770 {
4771 	struct e1000_hw *hw = &adapter->hw;
4772 
4773 	/* With 82574 controllers, PHY needs to be checked periodically
4774 	 * for hung state and reset, if two calls return true
4775 	 */
4776 	if (e1000_check_phy_82574(hw))
4777 		adapter->phy_hang_count++;
4778 	else
4779 		adapter->phy_hang_count = 0;
4780 
4781 	if (adapter->phy_hang_count > 1) {
4782 		adapter->phy_hang_count = 0;
4783 		schedule_work(&adapter->reset_task);
4784 	}
4785 }
4786 
4787 /**
4788  * e1000_watchdog - Timer Call-back
4789  * @data: pointer to adapter cast into an unsigned long
4790  **/
4791 static void e1000_watchdog(unsigned long data)
4792 {
4793 	struct e1000_adapter *adapter = (struct e1000_adapter *) data;
4794 
4795 	/* Do the rest outside of interrupt context */
4796 	schedule_work(&adapter->watchdog_task);
4797 
4798 	/* TODO: make this use queue_delayed_work() */
4799 }
4800 
4801 static void e1000_watchdog_task(struct work_struct *work)
4802 {
4803 	struct e1000_adapter *adapter = container_of(work,
4804 					struct e1000_adapter, watchdog_task);
4805 	struct net_device *netdev = adapter->netdev;
4806 	struct e1000_mac_info *mac = &adapter->hw.mac;
4807 	struct e1000_phy_info *phy = &adapter->hw.phy;
4808 	struct e1000_ring *tx_ring = adapter->tx_ring;
4809 	struct e1000_hw *hw = &adapter->hw;
4810 	u32 link, tctl;
4811 
4812 	if (test_bit(__E1000_DOWN, &adapter->state))
4813 		return;
4814 
4815 	link = e1000e_has_link(adapter);
4816 	if ((netif_carrier_ok(netdev)) && link) {
4817 		/* Cancel scheduled suspend requests. */
4818 		pm_runtime_resume(netdev->dev.parent);
4819 
4820 		e1000e_enable_receives(adapter);
4821 		goto link_up;
4822 	}
4823 
4824 	if ((e1000e_enable_tx_pkt_filtering(hw)) &&
4825 	    (adapter->mng_vlan_id != adapter->hw.mng_cookie.vlan_id))
4826 		e1000_update_mng_vlan(adapter);
4827 
4828 	if (link) {
4829 		if (!netif_carrier_ok(netdev)) {
4830 			bool txb2b = true;
4831 
4832 			/* Cancel scheduled suspend requests. */
4833 			pm_runtime_resume(netdev->dev.parent);
4834 
4835 			/* update snapshot of PHY registers on LSC */
4836 			e1000_phy_read_status(adapter);
4837 			mac->ops.get_link_up_info(&adapter->hw,
4838 						   &adapter->link_speed,
4839 						   &adapter->link_duplex);
4840 			e1000_print_link_info(adapter);
4841 
4842 			/* check if SmartSpeed worked */
4843 			e1000e_check_downshift(hw);
4844 			if (phy->speed_downgraded)
4845 				netdev_warn(netdev,
4846 					    "Link Speed was downgraded by SmartSpeed\n");
4847 
4848 			/* On supported PHYs, check for duplex mismatch only
4849 			 * if link has autonegotiated at 10/100 half
4850 			 */
4851 			if ((hw->phy.type == e1000_phy_igp_3 ||
4852 			     hw->phy.type == e1000_phy_bm) &&
4853 			    (hw->mac.autoneg == true) &&
4854 			    (adapter->link_speed == SPEED_10 ||
4855 			     adapter->link_speed == SPEED_100) &&
4856 			    (adapter->link_duplex == HALF_DUPLEX)) {
4857 				u16 autoneg_exp;
4858 
4859 				e1e_rphy(hw, MII_EXPANSION, &autoneg_exp);
4860 
4861 				if (!(autoneg_exp & EXPANSION_NWAY))
4862 					e_info("Autonegotiated half duplex but link partner cannot autoneg.  Try forcing full duplex if link gets many collisions.\n");
4863 			}
4864 
4865 			/* adjust timeout factor according to speed/duplex */
4866 			adapter->tx_timeout_factor = 1;
4867 			switch (adapter->link_speed) {
4868 			case SPEED_10:
4869 				txb2b = false;
4870 				adapter->tx_timeout_factor = 16;
4871 				break;
4872 			case SPEED_100:
4873 				txb2b = false;
4874 				adapter->tx_timeout_factor = 10;
4875 				break;
4876 			}
4877 
4878 			/* workaround: re-program speed mode bit after
4879 			 * link-up event
4880 			 */
4881 			if ((adapter->flags & FLAG_TARC_SPEED_MODE_BIT) &&
4882 			    !txb2b) {
4883 				u32 tarc0;
4884 				tarc0 = er32(TARC(0));
4885 				tarc0 &= ~SPEED_MODE_BIT;
4886 				ew32(TARC(0), tarc0);
4887 			}
4888 
4889 			/* disable TSO for pcie and 10/100 speeds, to avoid
4890 			 * some hardware issues
4891 			 */
4892 			if (!(adapter->flags & FLAG_TSO_FORCE)) {
4893 				switch (adapter->link_speed) {
4894 				case SPEED_10:
4895 				case SPEED_100:
4896 					e_info("10/100 speed: disabling TSO\n");
4897 					netdev->features &= ~NETIF_F_TSO;
4898 					netdev->features &= ~NETIF_F_TSO6;
4899 					break;
4900 				case SPEED_1000:
4901 					netdev->features |= NETIF_F_TSO;
4902 					netdev->features |= NETIF_F_TSO6;
4903 					break;
4904 				default:
4905 					/* oops */
4906 					break;
4907 				}
4908 			}
4909 
4910 			/* enable transmits in the hardware, need to do this
4911 			 * after setting TARC(0)
4912 			 */
4913 			tctl = er32(TCTL);
4914 			tctl |= E1000_TCTL_EN;
4915 			ew32(TCTL, tctl);
4916 
4917 			/* Perform any post-link-up configuration before
4918 			 * reporting link up.
4919 			 */
4920 			if (phy->ops.cfg_on_link_up)
4921 				phy->ops.cfg_on_link_up(hw);
4922 
4923 			netif_carrier_on(netdev);
4924 
4925 			if (!test_bit(__E1000_DOWN, &adapter->state))
4926 				mod_timer(&adapter->phy_info_timer,
4927 					  round_jiffies(jiffies + 2 * HZ));
4928 		}
4929 	} else {
4930 		if (netif_carrier_ok(netdev)) {
4931 			adapter->link_speed = 0;
4932 			adapter->link_duplex = 0;
4933 			/* Link status message must follow this format */
4934 			pr_info("%s NIC Link is Down\n", adapter->netdev->name);
4935 			netif_carrier_off(netdev);
4936 			if (!test_bit(__E1000_DOWN, &adapter->state))
4937 				mod_timer(&adapter->phy_info_timer,
4938 					  round_jiffies(jiffies + 2 * HZ));
4939 
4940 			/* The link is lost so the controller stops DMA.
4941 			 * If there is queued Tx work that cannot be done
4942 			 * or if on an 8000ES2LAN which requires a Rx packet
4943 			 * buffer work-around on link down event, reset the
4944 			 * controller to flush the Tx/Rx packet buffers.
4945 			 * (Do the reset outside of interrupt context).
4946 			 */
4947 			if ((adapter->flags & FLAG_RX_NEEDS_RESTART) ||
4948 			    (e1000_desc_unused(tx_ring) + 1 < tx_ring->count))
4949 				adapter->flags |= FLAG_RESTART_NOW;
4950 			else
4951 				pm_schedule_suspend(netdev->dev.parent,
4952 							LINK_TIMEOUT);
4953 		}
4954 	}
4955 
4956 link_up:
4957 	spin_lock(&adapter->stats64_lock);
4958 	e1000e_update_stats(adapter);
4959 
4960 	mac->tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
4961 	adapter->tpt_old = adapter->stats.tpt;
4962 	mac->collision_delta = adapter->stats.colc - adapter->colc_old;
4963 	adapter->colc_old = adapter->stats.colc;
4964 
4965 	adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
4966 	adapter->gorc_old = adapter->stats.gorc;
4967 	adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
4968 	adapter->gotc_old = adapter->stats.gotc;
4969 	spin_unlock(&adapter->stats64_lock);
4970 
4971 	if (adapter->flags & FLAG_RESTART_NOW) {
4972 		schedule_work(&adapter->reset_task);
4973 		/* return immediately since reset is imminent */
4974 		return;
4975 	}
4976 
4977 	e1000e_update_adaptive(&adapter->hw);
4978 
4979 	/* Simple mode for Interrupt Throttle Rate (ITR) */
4980 	if (adapter->itr_setting == 4) {
4981 		/* Symmetric Tx/Rx gets a reduced ITR=2000;
4982 		 * Total asymmetrical Tx or Rx gets ITR=8000;
4983 		 * everyone else is between 2000-8000.
4984 		 */
4985 		u32 goc = (adapter->gotc + adapter->gorc) / 10000;
4986 		u32 dif = (adapter->gotc > adapter->gorc ?
4987 			    adapter->gotc - adapter->gorc :
4988 			    adapter->gorc - adapter->gotc) / 10000;
4989 		u32 itr = goc > 0 ? (dif * 6000 / goc + 2000) : 8000;
4990 
4991 		e1000e_write_itr(adapter, itr);
4992 	}
4993 
4994 	/* Cause software interrupt to ensure Rx ring is cleaned */
4995 	if (adapter->msix_entries)
4996 		ew32(ICS, adapter->rx_ring->ims_val);
4997 	else
4998 		ew32(ICS, E1000_ICS_RXDMT0);
4999 
5000 	/* flush pending descriptors to memory before detecting Tx hang */
5001 	e1000e_flush_descriptors(adapter);
5002 
5003 	/* Force detection of hung controller every watchdog period */
5004 	adapter->detect_tx_hung = true;
5005 
5006 	/* With 82571 controllers, LAA may be overwritten due to controller
5007 	 * reset from the other port. Set the appropriate LAA in RAR[0]
5008 	 */
5009 	if (e1000e_get_laa_state_82571(hw))
5010 		hw->mac.ops.rar_set(hw, adapter->hw.mac.addr, 0);
5011 
5012 	if (adapter->flags2 & FLAG2_CHECK_PHY_HANG)
5013 		e1000e_check_82574_phy_workaround(adapter);
5014 
5015 	/* Clear valid timestamp stuck in RXSTMPL/H due to a Rx error */
5016 	if (adapter->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE) {
5017 		if ((adapter->flags2 & FLAG2_CHECK_RX_HWTSTAMP) &&
5018 		    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID)) {
5019 			er32(RXSTMPH);
5020 			adapter->rx_hwtstamp_cleared++;
5021 		} else {
5022 			adapter->flags2 |= FLAG2_CHECK_RX_HWTSTAMP;
5023 		}
5024 	}
5025 
5026 	/* Reset the timer */
5027 	if (!test_bit(__E1000_DOWN, &adapter->state))
5028 		mod_timer(&adapter->watchdog_timer,
5029 			  round_jiffies(jiffies + 2 * HZ));
5030 }
5031 
5032 #define E1000_TX_FLAGS_CSUM		0x00000001
5033 #define E1000_TX_FLAGS_VLAN		0x00000002
5034 #define E1000_TX_FLAGS_TSO		0x00000004
5035 #define E1000_TX_FLAGS_IPV4		0x00000008
5036 #define E1000_TX_FLAGS_NO_FCS		0x00000010
5037 #define E1000_TX_FLAGS_HWTSTAMP		0x00000020
5038 #define E1000_TX_FLAGS_VLAN_MASK	0xffff0000
5039 #define E1000_TX_FLAGS_VLAN_SHIFT	16
5040 
5041 static int e1000_tso(struct e1000_ring *tx_ring, struct sk_buff *skb)
5042 {
5043 	struct e1000_context_desc *context_desc;
5044 	struct e1000_buffer *buffer_info;
5045 	unsigned int i;
5046 	u32 cmd_length = 0;
5047 	u16 ipcse = 0, mss;
5048 	u8 ipcss, ipcso, tucss, tucso, hdr_len;
5049 
5050 	if (!skb_is_gso(skb))
5051 		return 0;
5052 
5053 	if (skb_header_cloned(skb)) {
5054 		int err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
5055 
5056 		if (err)
5057 			return err;
5058 	}
5059 
5060 	hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5061 	mss = skb_shinfo(skb)->gso_size;
5062 	if (skb->protocol == htons(ETH_P_IP)) {
5063 		struct iphdr *iph = ip_hdr(skb);
5064 		iph->tot_len = 0;
5065 		iph->check = 0;
5066 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
5067 		                                         0, IPPROTO_TCP, 0);
5068 		cmd_length = E1000_TXD_CMD_IP;
5069 		ipcse = skb_transport_offset(skb) - 1;
5070 	} else if (skb_is_gso_v6(skb)) {
5071 		ipv6_hdr(skb)->payload_len = 0;
5072 		tcp_hdr(skb)->check = ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
5073 		                                       &ipv6_hdr(skb)->daddr,
5074 		                                       0, IPPROTO_TCP, 0);
5075 		ipcse = 0;
5076 	}
5077 	ipcss = skb_network_offset(skb);
5078 	ipcso = (void *)&(ip_hdr(skb)->check) - (void *)skb->data;
5079 	tucss = skb_transport_offset(skb);
5080 	tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
5081 
5082 	cmd_length |= (E1000_TXD_CMD_DEXT | E1000_TXD_CMD_TSE |
5083 	               E1000_TXD_CMD_TCP | (skb->len - (hdr_len)));
5084 
5085 	i = tx_ring->next_to_use;
5086 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5087 	buffer_info = &tx_ring->buffer_info[i];
5088 
5089 	context_desc->lower_setup.ip_fields.ipcss  = ipcss;
5090 	context_desc->lower_setup.ip_fields.ipcso  = ipcso;
5091 	context_desc->lower_setup.ip_fields.ipcse  = cpu_to_le16(ipcse);
5092 	context_desc->upper_setup.tcp_fields.tucss = tucss;
5093 	context_desc->upper_setup.tcp_fields.tucso = tucso;
5094 	context_desc->upper_setup.tcp_fields.tucse = 0;
5095 	context_desc->tcp_seg_setup.fields.mss     = cpu_to_le16(mss);
5096 	context_desc->tcp_seg_setup.fields.hdr_len = hdr_len;
5097 	context_desc->cmd_and_length = cpu_to_le32(cmd_length);
5098 
5099 	buffer_info->time_stamp = jiffies;
5100 	buffer_info->next_to_watch = i;
5101 
5102 	i++;
5103 	if (i == tx_ring->count)
5104 		i = 0;
5105 	tx_ring->next_to_use = i;
5106 
5107 	return 1;
5108 }
5109 
5110 static bool e1000_tx_csum(struct e1000_ring *tx_ring, struct sk_buff *skb)
5111 {
5112 	struct e1000_adapter *adapter = tx_ring->adapter;
5113 	struct e1000_context_desc *context_desc;
5114 	struct e1000_buffer *buffer_info;
5115 	unsigned int i;
5116 	u8 css;
5117 	u32 cmd_len = E1000_TXD_CMD_DEXT;
5118 	__be16 protocol;
5119 
5120 	if (skb->ip_summed != CHECKSUM_PARTIAL)
5121 		return 0;
5122 
5123 	if (skb->protocol == cpu_to_be16(ETH_P_8021Q))
5124 		protocol = vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
5125 	else
5126 		protocol = skb->protocol;
5127 
5128 	switch (protocol) {
5129 	case cpu_to_be16(ETH_P_IP):
5130 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
5131 			cmd_len |= E1000_TXD_CMD_TCP;
5132 		break;
5133 	case cpu_to_be16(ETH_P_IPV6):
5134 		/* XXX not handling all IPV6 headers */
5135 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
5136 			cmd_len |= E1000_TXD_CMD_TCP;
5137 		break;
5138 	default:
5139 		if (unlikely(net_ratelimit()))
5140 			e_warn("checksum_partial proto=%x!\n",
5141 			       be16_to_cpu(protocol));
5142 		break;
5143 	}
5144 
5145 	css = skb_checksum_start_offset(skb);
5146 
5147 	i = tx_ring->next_to_use;
5148 	buffer_info = &tx_ring->buffer_info[i];
5149 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5150 
5151 	context_desc->lower_setup.ip_config = 0;
5152 	context_desc->upper_setup.tcp_fields.tucss = css;
5153 	context_desc->upper_setup.tcp_fields.tucso =
5154 				css + skb->csum_offset;
5155 	context_desc->upper_setup.tcp_fields.tucse = 0;
5156 	context_desc->tcp_seg_setup.data = 0;
5157 	context_desc->cmd_and_length = cpu_to_le32(cmd_len);
5158 
5159 	buffer_info->time_stamp = jiffies;
5160 	buffer_info->next_to_watch = i;
5161 
5162 	i++;
5163 	if (i == tx_ring->count)
5164 		i = 0;
5165 	tx_ring->next_to_use = i;
5166 
5167 	return 1;
5168 }
5169 
5170 static int e1000_tx_map(struct e1000_ring *tx_ring, struct sk_buff *skb,
5171 			unsigned int first, unsigned int max_per_txd,
5172 			unsigned int nr_frags)
5173 {
5174 	struct e1000_adapter *adapter = tx_ring->adapter;
5175 	struct pci_dev *pdev = adapter->pdev;
5176 	struct e1000_buffer *buffer_info;
5177 	unsigned int len = skb_headlen(skb);
5178 	unsigned int offset = 0, size, count = 0, i;
5179 	unsigned int f, bytecount, segs;
5180 
5181 	i = tx_ring->next_to_use;
5182 
5183 	while (len) {
5184 		buffer_info = &tx_ring->buffer_info[i];
5185 		size = min(len, max_per_txd);
5186 
5187 		buffer_info->length = size;
5188 		buffer_info->time_stamp = jiffies;
5189 		buffer_info->next_to_watch = i;
5190 		buffer_info->dma = dma_map_single(&pdev->dev,
5191 						  skb->data + offset,
5192 						  size, DMA_TO_DEVICE);
5193 		buffer_info->mapped_as_page = false;
5194 		if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5195 			goto dma_error;
5196 
5197 		len -= size;
5198 		offset += size;
5199 		count++;
5200 
5201 		if (len) {
5202 			i++;
5203 			if (i == tx_ring->count)
5204 				i = 0;
5205 		}
5206 	}
5207 
5208 	for (f = 0; f < nr_frags; f++) {
5209 		const struct skb_frag_struct *frag;
5210 
5211 		frag = &skb_shinfo(skb)->frags[f];
5212 		len = skb_frag_size(frag);
5213 		offset = 0;
5214 
5215 		while (len) {
5216 			i++;
5217 			if (i == tx_ring->count)
5218 				i = 0;
5219 
5220 			buffer_info = &tx_ring->buffer_info[i];
5221 			size = min(len, max_per_txd);
5222 
5223 			buffer_info->length = size;
5224 			buffer_info->time_stamp = jiffies;
5225 			buffer_info->next_to_watch = i;
5226 			buffer_info->dma = skb_frag_dma_map(&pdev->dev, frag,
5227 						offset, size, DMA_TO_DEVICE);
5228 			buffer_info->mapped_as_page = true;
5229 			if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5230 				goto dma_error;
5231 
5232 			len -= size;
5233 			offset += size;
5234 			count++;
5235 		}
5236 	}
5237 
5238 	segs = skb_shinfo(skb)->gso_segs ? : 1;
5239 	/* multiply data chunks by size of headers */
5240 	bytecount = ((segs - 1) * skb_headlen(skb)) + skb->len;
5241 
5242 	tx_ring->buffer_info[i].skb = skb;
5243 	tx_ring->buffer_info[i].segs = segs;
5244 	tx_ring->buffer_info[i].bytecount = bytecount;
5245 	tx_ring->buffer_info[first].next_to_watch = i;
5246 
5247 	return count;
5248 
5249 dma_error:
5250 	dev_err(&pdev->dev, "Tx DMA map failed\n");
5251 	buffer_info->dma = 0;
5252 	if (count)
5253 		count--;
5254 
5255 	while (count--) {
5256 		if (i == 0)
5257 			i += tx_ring->count;
5258 		i--;
5259 		buffer_info = &tx_ring->buffer_info[i];
5260 		e1000_put_txbuf(tx_ring, buffer_info);
5261 	}
5262 
5263 	return 0;
5264 }
5265 
5266 static void e1000_tx_queue(struct e1000_ring *tx_ring, int tx_flags, int count)
5267 {
5268 	struct e1000_adapter *adapter = tx_ring->adapter;
5269 	struct e1000_tx_desc *tx_desc = NULL;
5270 	struct e1000_buffer *buffer_info;
5271 	u32 txd_upper = 0, txd_lower = E1000_TXD_CMD_IFCS;
5272 	unsigned int i;
5273 
5274 	if (tx_flags & E1000_TX_FLAGS_TSO) {
5275 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D |
5276 			     E1000_TXD_CMD_TSE;
5277 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5278 
5279 		if (tx_flags & E1000_TX_FLAGS_IPV4)
5280 			txd_upper |= E1000_TXD_POPTS_IXSM << 8;
5281 	}
5282 
5283 	if (tx_flags & E1000_TX_FLAGS_CSUM) {
5284 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5285 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5286 	}
5287 
5288 	if (tx_flags & E1000_TX_FLAGS_VLAN) {
5289 		txd_lower |= E1000_TXD_CMD_VLE;
5290 		txd_upper |= (tx_flags & E1000_TX_FLAGS_VLAN_MASK);
5291 	}
5292 
5293 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5294 		txd_lower &= ~(E1000_TXD_CMD_IFCS);
5295 
5296 	if (unlikely(tx_flags & E1000_TX_FLAGS_HWTSTAMP)) {
5297 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5298 		txd_upper |= E1000_TXD_EXTCMD_TSTAMP;
5299 	}
5300 
5301 	i = tx_ring->next_to_use;
5302 
5303 	do {
5304 		buffer_info = &tx_ring->buffer_info[i];
5305 		tx_desc = E1000_TX_DESC(*tx_ring, i);
5306 		tx_desc->buffer_addr = cpu_to_le64(buffer_info->dma);
5307 		tx_desc->lower.data =
5308 			cpu_to_le32(txd_lower | buffer_info->length);
5309 		tx_desc->upper.data = cpu_to_le32(txd_upper);
5310 
5311 		i++;
5312 		if (i == tx_ring->count)
5313 			i = 0;
5314 	} while (--count > 0);
5315 
5316 	tx_desc->lower.data |= cpu_to_le32(adapter->txd_cmd);
5317 
5318 	/* txd_cmd re-enables FCS, so we'll re-disable it here as desired. */
5319 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5320 		tx_desc->lower.data &= ~(cpu_to_le32(E1000_TXD_CMD_IFCS));
5321 
5322 	/* Force memory writes to complete before letting h/w
5323 	 * know there are new descriptors to fetch.  (Only
5324 	 * applicable for weak-ordered memory model archs,
5325 	 * such as IA-64).
5326 	 */
5327 	wmb();
5328 
5329 	tx_ring->next_to_use = i;
5330 
5331 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
5332 		e1000e_update_tdt_wa(tx_ring, i);
5333 	else
5334 		writel(i, tx_ring->tail);
5335 
5336 	/* we need this if more than one processor can write to our tail
5337 	 * at a time, it synchronizes IO on IA64/Altix systems
5338 	 */
5339 	mmiowb();
5340 }
5341 
5342 #define MINIMUM_DHCP_PACKET_SIZE 282
5343 static int e1000_transfer_dhcp_info(struct e1000_adapter *adapter,
5344 				    struct sk_buff *skb)
5345 {
5346 	struct e1000_hw *hw =  &adapter->hw;
5347 	u16 length, offset;
5348 
5349 	if (vlan_tx_tag_present(skb) &&
5350 	    !((vlan_tx_tag_get(skb) == adapter->hw.mng_cookie.vlan_id) &&
5351 	      (adapter->hw.mng_cookie.status &
5352 	       E1000_MNG_DHCP_COOKIE_STATUS_VLAN)))
5353 		return 0;
5354 
5355 	if (skb->len <= MINIMUM_DHCP_PACKET_SIZE)
5356 		return 0;
5357 
5358 	if (((struct ethhdr *) skb->data)->h_proto != htons(ETH_P_IP))
5359 		return 0;
5360 
5361 	{
5362 		const struct iphdr *ip = (struct iphdr *)((u8 *)skb->data+14);
5363 		struct udphdr *udp;
5364 
5365 		if (ip->protocol != IPPROTO_UDP)
5366 			return 0;
5367 
5368 		udp = (struct udphdr *)((u8 *)ip + (ip->ihl << 2));
5369 		if (ntohs(udp->dest) != 67)
5370 			return 0;
5371 
5372 		offset = (u8 *)udp + 8 - skb->data;
5373 		length = skb->len - offset;
5374 		return e1000e_mng_write_dhcp_info(hw, (u8 *)udp + 8, length);
5375 	}
5376 
5377 	return 0;
5378 }
5379 
5380 static int __e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5381 {
5382 	struct e1000_adapter *adapter = tx_ring->adapter;
5383 
5384 	netif_stop_queue(adapter->netdev);
5385 	/* Herbert's original patch had:
5386 	 *  smp_mb__after_netif_stop_queue();
5387 	 * but since that doesn't exist yet, just open code it.
5388 	 */
5389 	smp_mb();
5390 
5391 	/* We need to check again in a case another CPU has just
5392 	 * made room available.
5393 	 */
5394 	if (e1000_desc_unused(tx_ring) < size)
5395 		return -EBUSY;
5396 
5397 	/* A reprieve! */
5398 	netif_start_queue(adapter->netdev);
5399 	++adapter->restart_queue;
5400 	return 0;
5401 }
5402 
5403 static int e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5404 {
5405 	BUG_ON(size > tx_ring->count);
5406 
5407 	if (e1000_desc_unused(tx_ring) >= size)
5408 		return 0;
5409 	return __e1000_maybe_stop_tx(tx_ring, size);
5410 }
5411 
5412 static netdev_tx_t e1000_xmit_frame(struct sk_buff *skb,
5413 				    struct net_device *netdev)
5414 {
5415 	struct e1000_adapter *adapter = netdev_priv(netdev);
5416 	struct e1000_ring *tx_ring = adapter->tx_ring;
5417 	unsigned int first;
5418 	unsigned int tx_flags = 0;
5419 	unsigned int len = skb_headlen(skb);
5420 	unsigned int nr_frags;
5421 	unsigned int mss;
5422 	int count = 0;
5423 	int tso;
5424 	unsigned int f;
5425 
5426 	if (test_bit(__E1000_DOWN, &adapter->state)) {
5427 		dev_kfree_skb_any(skb);
5428 		return NETDEV_TX_OK;
5429 	}
5430 
5431 	if (skb->len <= 0) {
5432 		dev_kfree_skb_any(skb);
5433 		return NETDEV_TX_OK;
5434 	}
5435 
5436 	/* The minimum packet size with TCTL.PSP set is 17 bytes so
5437 	 * pad skb in order to meet this minimum size requirement
5438 	 */
5439 	if (unlikely(skb->len < 17)) {
5440 		if (skb_pad(skb, 17 - skb->len))
5441 			return NETDEV_TX_OK;
5442 		skb->len = 17;
5443 		skb_set_tail_pointer(skb, 17);
5444 	}
5445 
5446 	mss = skb_shinfo(skb)->gso_size;
5447 	if (mss) {
5448 		u8 hdr_len;
5449 
5450 		/* TSO Workaround for 82571/2/3 Controllers -- if skb->data
5451 		 * points to just header, pull a few bytes of payload from
5452 		 * frags into skb->data
5453 		 */
5454 		hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5455 		/* we do this workaround for ES2LAN, but it is un-necessary,
5456 		 * avoiding it could save a lot of cycles
5457 		 */
5458 		if (skb->data_len && (hdr_len == len)) {
5459 			unsigned int pull_size;
5460 
5461 			pull_size = min_t(unsigned int, 4, skb->data_len);
5462 			if (!__pskb_pull_tail(skb, pull_size)) {
5463 				e_err("__pskb_pull_tail failed.\n");
5464 				dev_kfree_skb_any(skb);
5465 				return NETDEV_TX_OK;
5466 			}
5467 			len = skb_headlen(skb);
5468 		}
5469 	}
5470 
5471 	/* reserve a descriptor for the offload context */
5472 	if ((mss) || (skb->ip_summed == CHECKSUM_PARTIAL))
5473 		count++;
5474 	count++;
5475 
5476 	count += DIV_ROUND_UP(len, adapter->tx_fifo_limit);
5477 
5478 	nr_frags = skb_shinfo(skb)->nr_frags;
5479 	for (f = 0; f < nr_frags; f++)
5480 		count += DIV_ROUND_UP(skb_frag_size(&skb_shinfo(skb)->frags[f]),
5481 				      adapter->tx_fifo_limit);
5482 
5483 	if (adapter->hw.mac.tx_pkt_filtering)
5484 		e1000_transfer_dhcp_info(adapter, skb);
5485 
5486 	/* need: count + 2 desc gap to keep tail from touching
5487 	 * head, otherwise try next time
5488 	 */
5489 	if (e1000_maybe_stop_tx(tx_ring, count + 2))
5490 		return NETDEV_TX_BUSY;
5491 
5492 	if (vlan_tx_tag_present(skb)) {
5493 		tx_flags |= E1000_TX_FLAGS_VLAN;
5494 		tx_flags |= (vlan_tx_tag_get(skb) << E1000_TX_FLAGS_VLAN_SHIFT);
5495 	}
5496 
5497 	first = tx_ring->next_to_use;
5498 
5499 	tso = e1000_tso(tx_ring, skb);
5500 	if (tso < 0) {
5501 		dev_kfree_skb_any(skb);
5502 		return NETDEV_TX_OK;
5503 	}
5504 
5505 	if (tso)
5506 		tx_flags |= E1000_TX_FLAGS_TSO;
5507 	else if (e1000_tx_csum(tx_ring, skb))
5508 		tx_flags |= E1000_TX_FLAGS_CSUM;
5509 
5510 	/* Old method was to assume IPv4 packet by default if TSO was enabled.
5511 	 * 82571 hardware supports TSO capabilities for IPv6 as well...
5512 	 * no longer assume, we must.
5513 	 */
5514 	if (skb->protocol == htons(ETH_P_IP))
5515 		tx_flags |= E1000_TX_FLAGS_IPV4;
5516 
5517 	if (unlikely(skb->no_fcs))
5518 		tx_flags |= E1000_TX_FLAGS_NO_FCS;
5519 
5520 	/* if count is 0 then mapping error has occurred */
5521 	count = e1000_tx_map(tx_ring, skb, first, adapter->tx_fifo_limit,
5522 			     nr_frags);
5523 	if (count) {
5524 		if (unlikely((skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
5525 			     !adapter->tx_hwtstamp_skb)) {
5526 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5527 			tx_flags |= E1000_TX_FLAGS_HWTSTAMP;
5528 			adapter->tx_hwtstamp_skb = skb_get(skb);
5529 			schedule_work(&adapter->tx_hwtstamp_work);
5530 		} else {
5531 			skb_tx_timestamp(skb);
5532 		}
5533 
5534 		netdev_sent_queue(netdev, skb->len);
5535 		e1000_tx_queue(tx_ring, tx_flags, count);
5536 		/* Make sure there is space in the ring for the next send. */
5537 		e1000_maybe_stop_tx(tx_ring,
5538 				    (MAX_SKB_FRAGS *
5539 				     DIV_ROUND_UP(PAGE_SIZE,
5540 						  adapter->tx_fifo_limit) + 2));
5541 	} else {
5542 		dev_kfree_skb_any(skb);
5543 		tx_ring->buffer_info[first].time_stamp = 0;
5544 		tx_ring->next_to_use = first;
5545 	}
5546 
5547 	return NETDEV_TX_OK;
5548 }
5549 
5550 /**
5551  * e1000_tx_timeout - Respond to a Tx Hang
5552  * @netdev: network interface device structure
5553  **/
5554 static void e1000_tx_timeout(struct net_device *netdev)
5555 {
5556 	struct e1000_adapter *adapter = netdev_priv(netdev);
5557 
5558 	/* Do the reset outside of interrupt context */
5559 	adapter->tx_timeout_count++;
5560 	schedule_work(&adapter->reset_task);
5561 }
5562 
5563 static void e1000_reset_task(struct work_struct *work)
5564 {
5565 	struct e1000_adapter *adapter;
5566 	adapter = container_of(work, struct e1000_adapter, reset_task);
5567 
5568 	/* don't run the task if already down */
5569 	if (test_bit(__E1000_DOWN, &adapter->state))
5570 		return;
5571 
5572 	if (!(adapter->flags & FLAG_RESTART_NOW)) {
5573 		e1000e_dump(adapter);
5574 		e_err("Reset adapter unexpectedly\n");
5575 	}
5576 	e1000e_reinit_locked(adapter);
5577 }
5578 
5579 /**
5580  * e1000_get_stats64 - Get System Network Statistics
5581  * @netdev: network interface device structure
5582  * @stats: rtnl_link_stats64 pointer
5583  *
5584  * Returns the address of the device statistics structure.
5585  **/
5586 struct rtnl_link_stats64 *e1000e_get_stats64(struct net_device *netdev,
5587                                              struct rtnl_link_stats64 *stats)
5588 {
5589 	struct e1000_adapter *adapter = netdev_priv(netdev);
5590 
5591 	memset(stats, 0, sizeof(struct rtnl_link_stats64));
5592 	spin_lock(&adapter->stats64_lock);
5593 	e1000e_update_stats(adapter);
5594 	/* Fill out the OS statistics structure */
5595 	stats->rx_bytes = adapter->stats.gorc;
5596 	stats->rx_packets = adapter->stats.gprc;
5597 	stats->tx_bytes = adapter->stats.gotc;
5598 	stats->tx_packets = adapter->stats.gptc;
5599 	stats->multicast = adapter->stats.mprc;
5600 	stats->collisions = adapter->stats.colc;
5601 
5602 	/* Rx Errors */
5603 
5604 	/* RLEC on some newer hardware can be incorrect so build
5605 	 * our own version based on RUC and ROC
5606 	 */
5607 	stats->rx_errors = adapter->stats.rxerrc +
5608 		adapter->stats.crcerrs + adapter->stats.algnerrc +
5609 		adapter->stats.ruc + adapter->stats.roc +
5610 		adapter->stats.cexterr;
5611 	stats->rx_length_errors = adapter->stats.ruc +
5612 					      adapter->stats.roc;
5613 	stats->rx_crc_errors = adapter->stats.crcerrs;
5614 	stats->rx_frame_errors = adapter->stats.algnerrc;
5615 	stats->rx_missed_errors = adapter->stats.mpc;
5616 
5617 	/* Tx Errors */
5618 	stats->tx_errors = adapter->stats.ecol +
5619 				       adapter->stats.latecol;
5620 	stats->tx_aborted_errors = adapter->stats.ecol;
5621 	stats->tx_window_errors = adapter->stats.latecol;
5622 	stats->tx_carrier_errors = adapter->stats.tncrs;
5623 
5624 	/* Tx Dropped needs to be maintained elsewhere */
5625 
5626 	spin_unlock(&adapter->stats64_lock);
5627 	return stats;
5628 }
5629 
5630 /**
5631  * e1000_change_mtu - Change the Maximum Transfer Unit
5632  * @netdev: network interface device structure
5633  * @new_mtu: new value for maximum frame size
5634  *
5635  * Returns 0 on success, negative on failure
5636  **/
5637 static int e1000_change_mtu(struct net_device *netdev, int new_mtu)
5638 {
5639 	struct e1000_adapter *adapter = netdev_priv(netdev);
5640 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
5641 
5642 	/* Jumbo frame support */
5643 	if ((max_frame > ETH_FRAME_LEN + ETH_FCS_LEN) &&
5644 	    !(adapter->flags & FLAG_HAS_JUMBO_FRAMES)) {
5645 		e_err("Jumbo Frames not supported.\n");
5646 		return -EINVAL;
5647 	}
5648 
5649 	/* Supported frame sizes */
5650 	if ((new_mtu < ETH_ZLEN + ETH_FCS_LEN + VLAN_HLEN) ||
5651 	    (max_frame > adapter->max_hw_frame_size)) {
5652 		e_err("Unsupported MTU setting\n");
5653 		return -EINVAL;
5654 	}
5655 
5656 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
5657 	if ((adapter->hw.mac.type >= e1000_pch2lan) &&
5658 	    !(adapter->flags2 & FLAG2_CRC_STRIPPING) &&
5659 	    (new_mtu > ETH_DATA_LEN)) {
5660 		e_err("Jumbo Frames not supported on this device when CRC stripping is disabled.\n");
5661 		return -EINVAL;
5662 	}
5663 
5664 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
5665 		usleep_range(1000, 2000);
5666 	/* e1000e_down -> e1000e_reset dependent on max_frame_size & mtu */
5667 	adapter->max_frame_size = max_frame;
5668 	e_info("changing MTU from %d to %d\n", netdev->mtu, new_mtu);
5669 	netdev->mtu = new_mtu;
5670 	if (netif_running(netdev))
5671 		e1000e_down(adapter);
5672 
5673 	/* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
5674 	 * means we reserve 2 more, this pushes us to allocate from the next
5675 	 * larger slab size.
5676 	 * i.e. RXBUFFER_2048 --> size-4096 slab
5677 	 * However with the new *_jumbo_rx* routines, jumbo receives will use
5678 	 * fragmented skbs
5679 	 */
5680 
5681 	if (max_frame <= 2048)
5682 		adapter->rx_buffer_len = 2048;
5683 	else
5684 		adapter->rx_buffer_len = 4096;
5685 
5686 	/* adjust allocation if LPE protects us, and we aren't using SBP */
5687 	if ((max_frame == ETH_FRAME_LEN + ETH_FCS_LEN) ||
5688 	     (max_frame == ETH_FRAME_LEN + VLAN_HLEN + ETH_FCS_LEN))
5689 		adapter->rx_buffer_len = ETH_FRAME_LEN + VLAN_HLEN
5690 					 + ETH_FCS_LEN;
5691 
5692 	if (netif_running(netdev))
5693 		e1000e_up(adapter);
5694 	else
5695 		e1000e_reset(adapter);
5696 
5697 	clear_bit(__E1000_RESETTING, &adapter->state);
5698 
5699 	return 0;
5700 }
5701 
5702 static int e1000_mii_ioctl(struct net_device *netdev, struct ifreq *ifr,
5703 			   int cmd)
5704 {
5705 	struct e1000_adapter *adapter = netdev_priv(netdev);
5706 	struct mii_ioctl_data *data = if_mii(ifr);
5707 
5708 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
5709 		return -EOPNOTSUPP;
5710 
5711 	switch (cmd) {
5712 	case SIOCGMIIPHY:
5713 		data->phy_id = adapter->hw.phy.addr;
5714 		break;
5715 	case SIOCGMIIREG:
5716 		e1000_phy_read_status(adapter);
5717 
5718 		switch (data->reg_num & 0x1F) {
5719 		case MII_BMCR:
5720 			data->val_out = adapter->phy_regs.bmcr;
5721 			break;
5722 		case MII_BMSR:
5723 			data->val_out = adapter->phy_regs.bmsr;
5724 			break;
5725 		case MII_PHYSID1:
5726 			data->val_out = (adapter->hw.phy.id >> 16);
5727 			break;
5728 		case MII_PHYSID2:
5729 			data->val_out = (adapter->hw.phy.id & 0xFFFF);
5730 			break;
5731 		case MII_ADVERTISE:
5732 			data->val_out = adapter->phy_regs.advertise;
5733 			break;
5734 		case MII_LPA:
5735 			data->val_out = adapter->phy_regs.lpa;
5736 			break;
5737 		case MII_EXPANSION:
5738 			data->val_out = adapter->phy_regs.expansion;
5739 			break;
5740 		case MII_CTRL1000:
5741 			data->val_out = adapter->phy_regs.ctrl1000;
5742 			break;
5743 		case MII_STAT1000:
5744 			data->val_out = adapter->phy_regs.stat1000;
5745 			break;
5746 		case MII_ESTATUS:
5747 			data->val_out = adapter->phy_regs.estatus;
5748 			break;
5749 		default:
5750 			return -EIO;
5751 		}
5752 		break;
5753 	case SIOCSMIIREG:
5754 	default:
5755 		return -EOPNOTSUPP;
5756 	}
5757 	return 0;
5758 }
5759 
5760 /**
5761  * e1000e_hwtstamp_ioctl - control hardware time stamping
5762  * @netdev: network interface device structure
5763  * @ifreq: interface request
5764  *
5765  * Outgoing time stamping can be enabled and disabled. Play nice and
5766  * disable it when requested, although it shouldn't cause any overhead
5767  * when no packet needs it. At most one packet in the queue may be
5768  * marked for time stamping, otherwise it would be impossible to tell
5769  * for sure to which packet the hardware time stamp belongs.
5770  *
5771  * Incoming time stamping has to be configured via the hardware filters.
5772  * Not all combinations are supported, in particular event type has to be
5773  * specified. Matching the kind of event packet is not supported, with the
5774  * exception of "all V2 events regardless of level 2 or 4".
5775  **/
5776 static int e1000e_hwtstamp_ioctl(struct net_device *netdev, struct ifreq *ifr)
5777 {
5778 	struct e1000_adapter *adapter = netdev_priv(netdev);
5779 	struct hwtstamp_config config;
5780 	int ret_val;
5781 
5782 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
5783 		return -EFAULT;
5784 
5785 	adapter->hwtstamp_config = config;
5786 
5787 	ret_val = e1000e_config_hwtstamp(adapter);
5788 	if (ret_val)
5789 		return ret_val;
5790 
5791 	config = adapter->hwtstamp_config;
5792 
5793 	switch (config.rx_filter) {
5794 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
5795 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
5796 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
5797 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
5798 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
5799 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
5800 		/* With V2 type filters which specify a Sync or Delay Request,
5801 		 * Path Delay Request/Response messages are also time stamped
5802 		 * by hardware so notify the caller the requested packets plus
5803 		 * some others are time stamped.
5804 		 */
5805 		config.rx_filter = HWTSTAMP_FILTER_SOME;
5806 		break;
5807 	default:
5808 		break;
5809 	}
5810 
5811 	return copy_to_user(ifr->ifr_data, &config,
5812 			    sizeof(config)) ? -EFAULT : 0;
5813 }
5814 
5815 static int e1000_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
5816 {
5817 	switch (cmd) {
5818 	case SIOCGMIIPHY:
5819 	case SIOCGMIIREG:
5820 	case SIOCSMIIREG:
5821 		return e1000_mii_ioctl(netdev, ifr, cmd);
5822 	case SIOCSHWTSTAMP:
5823 		return e1000e_hwtstamp_ioctl(netdev, ifr);
5824 	default:
5825 		return -EOPNOTSUPP;
5826 	}
5827 }
5828 
5829 static int e1000_init_phy_wakeup(struct e1000_adapter *adapter, u32 wufc)
5830 {
5831 	struct e1000_hw *hw = &adapter->hw;
5832 	u32 i, mac_reg;
5833 	u16 phy_reg, wuc_enable;
5834 	int retval;
5835 
5836 	/* copy MAC RARs to PHY RARs */
5837 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
5838 
5839 	retval = hw->phy.ops.acquire(hw);
5840 	if (retval) {
5841 		e_err("Could not acquire PHY\n");
5842 		return retval;
5843 	}
5844 
5845 	/* Enable access to wakeup registers on and set page to BM_WUC_PAGE */
5846 	retval = e1000_enable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
5847 	if (retval)
5848 		goto release;
5849 
5850 	/* copy MAC MTA to PHY MTA - only needed for pchlan */
5851 	for (i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
5852 		mac_reg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
5853 		hw->phy.ops.write_reg_page(hw, BM_MTA(i),
5854 					   (u16)(mac_reg & 0xFFFF));
5855 		hw->phy.ops.write_reg_page(hw, BM_MTA(i) + 1,
5856 					   (u16)((mac_reg >> 16) & 0xFFFF));
5857 	}
5858 
5859 	/* configure PHY Rx Control register */
5860 	hw->phy.ops.read_reg_page(&adapter->hw, BM_RCTL, &phy_reg);
5861 	mac_reg = er32(RCTL);
5862 	if (mac_reg & E1000_RCTL_UPE)
5863 		phy_reg |= BM_RCTL_UPE;
5864 	if (mac_reg & E1000_RCTL_MPE)
5865 		phy_reg |= BM_RCTL_MPE;
5866 	phy_reg &= ~(BM_RCTL_MO_MASK);
5867 	if (mac_reg & E1000_RCTL_MO_3)
5868 		phy_reg |= (((mac_reg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
5869 				<< BM_RCTL_MO_SHIFT);
5870 	if (mac_reg & E1000_RCTL_BAM)
5871 		phy_reg |= BM_RCTL_BAM;
5872 	if (mac_reg & E1000_RCTL_PMCF)
5873 		phy_reg |= BM_RCTL_PMCF;
5874 	mac_reg = er32(CTRL);
5875 	if (mac_reg & E1000_CTRL_RFCE)
5876 		phy_reg |= BM_RCTL_RFCE;
5877 	hw->phy.ops.write_reg_page(&adapter->hw, BM_RCTL, phy_reg);
5878 
5879 	/* enable PHY wakeup in MAC register */
5880 	ew32(WUFC, wufc);
5881 	ew32(WUC, E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN);
5882 
5883 	/* configure and enable PHY wakeup in PHY registers */
5884 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUFC, wufc);
5885 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUC, E1000_WUC_PME_EN);
5886 
5887 	/* activate PHY wakeup */
5888 	wuc_enable |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
5889 	retval = e1000_disable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
5890 	if (retval)
5891 		e_err("Could not set PHY Host Wakeup bit\n");
5892 release:
5893 	hw->phy.ops.release(hw);
5894 
5895 	return retval;
5896 }
5897 
5898 static int __e1000_shutdown(struct pci_dev *pdev, bool runtime)
5899 {
5900 	struct net_device *netdev = pci_get_drvdata(pdev);
5901 	struct e1000_adapter *adapter = netdev_priv(netdev);
5902 	struct e1000_hw *hw = &adapter->hw;
5903 	u32 ctrl, ctrl_ext, rctl, status;
5904 	/* Runtime suspend should only enable wakeup for link changes */
5905 	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
5906 	int retval = 0;
5907 
5908 	netif_device_detach(netdev);
5909 
5910 	if (netif_running(netdev)) {
5911 		int count = E1000_CHECK_RESET_COUNT;
5912 
5913 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
5914 			usleep_range(10000, 20000);
5915 
5916 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
5917 		e1000e_down(adapter);
5918 		e1000_free_irq(adapter);
5919 	}
5920 	e1000e_reset_interrupt_capability(adapter);
5921 
5922 	status = er32(STATUS);
5923 	if (status & E1000_STATUS_LU)
5924 		wufc &= ~E1000_WUFC_LNKC;
5925 
5926 	if (wufc) {
5927 		e1000_setup_rctl(adapter);
5928 		e1000e_set_rx_mode(netdev);
5929 
5930 		/* turn on all-multi mode if wake on multicast is enabled */
5931 		if (wufc & E1000_WUFC_MC) {
5932 			rctl = er32(RCTL);
5933 			rctl |= E1000_RCTL_MPE;
5934 			ew32(RCTL, rctl);
5935 		}
5936 
5937 		ctrl = er32(CTRL);
5938 		/* advertise wake from D3Cold */
5939 		#define E1000_CTRL_ADVD3WUC 0x00100000
5940 		/* phy power management enable */
5941 		#define E1000_CTRL_EN_PHY_PWR_MGMT 0x00200000
5942 		ctrl |= E1000_CTRL_ADVD3WUC;
5943 		if (!(adapter->flags2 & FLAG2_HAS_PHY_WAKEUP))
5944 			ctrl |= E1000_CTRL_EN_PHY_PWR_MGMT;
5945 		ew32(CTRL, ctrl);
5946 
5947 		if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
5948 		    adapter->hw.phy.media_type ==
5949 		    e1000_media_type_internal_serdes) {
5950 			/* keep the laser running in D3 */
5951 			ctrl_ext = er32(CTRL_EXT);
5952 			ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
5953 			ew32(CTRL_EXT, ctrl_ext);
5954 		}
5955 
5956 		if (adapter->flags & FLAG_IS_ICH)
5957 			e1000_suspend_workarounds_ich8lan(&adapter->hw);
5958 
5959 		/* Allow time for pending master requests to run */
5960 		e1000e_disable_pcie_master(&adapter->hw);
5961 
5962 		if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
5963 			/* enable wakeup by the PHY */
5964 			retval = e1000_init_phy_wakeup(adapter, wufc);
5965 			if (retval)
5966 				return retval;
5967 		} else {
5968 			/* enable wakeup by the MAC */
5969 			ew32(WUFC, wufc);
5970 			ew32(WUC, E1000_WUC_PME_EN);
5971 		}
5972 	} else {
5973 		ew32(WUC, 0);
5974 		ew32(WUFC, 0);
5975 	}
5976 
5977 	if (adapter->hw.phy.type == e1000_phy_igp_3)
5978 		e1000e_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
5979 
5980 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
5981 	 * would have already happened in close and is redundant.
5982 	 */
5983 	e1000e_release_hw_control(adapter);
5984 
5985 	pci_clear_master(pdev);
5986 
5987 	/* The pci-e switch on some quad port adapters will report a
5988 	 * correctable error when the MAC transitions from D0 to D3.  To
5989 	 * prevent this we need to mask off the correctable errors on the
5990 	 * downstream port of the pci-e switch.
5991 	 */
5992 	if (adapter->flags & FLAG_IS_QUAD_PORT) {
5993 		struct pci_dev *us_dev = pdev->bus->self;
5994 		u16 devctl;
5995 
5996 		pcie_capability_read_word(us_dev, PCI_EXP_DEVCTL, &devctl);
5997 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL,
5998 					   (devctl & ~PCI_EXP_DEVCTL_CERE));
5999 
6000 		pci_save_state(pdev);
6001 		pci_prepare_to_sleep(pdev);
6002 
6003 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL, devctl);
6004 	}
6005 
6006 	return 0;
6007 }
6008 
6009 #ifdef CONFIG_PCIEASPM
6010 static void __e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6011 {
6012 	pci_disable_link_state_locked(pdev, state);
6013 }
6014 #else
6015 static void __e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6016 {
6017 	u16 aspm_ctl = 0;
6018 
6019 	if (state & PCIE_LINK_STATE_L0S)
6020 		aspm_ctl |= PCI_EXP_LNKCTL_ASPM_L0S;
6021 	if (state & PCIE_LINK_STATE_L1)
6022 		aspm_ctl |= PCI_EXP_LNKCTL_ASPM_L1;
6023 
6024 	/* Both device and parent should have the same ASPM setting.
6025 	 * Disable ASPM in downstream component first and then upstream.
6026 	 */
6027 	pcie_capability_clear_word(pdev, PCI_EXP_LNKCTL, aspm_ctl);
6028 
6029 	if (pdev->bus->self)
6030 		pcie_capability_clear_word(pdev->bus->self, PCI_EXP_LNKCTL,
6031 					   aspm_ctl);
6032 }
6033 #endif
6034 static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6035 {
6036 	dev_info(&pdev->dev, "Disabling ASPM %s %s\n",
6037 		 (state & PCIE_LINK_STATE_L0S) ? "L0s" : "",
6038 		 (state & PCIE_LINK_STATE_L1) ? "L1" : "");
6039 
6040 	__e1000e_disable_aspm(pdev, state);
6041 }
6042 
6043 #ifdef CONFIG_PM
6044 static bool e1000e_pm_ready(struct e1000_adapter *adapter)
6045 {
6046 	return !!adapter->tx_ring->buffer_info;
6047 }
6048 
6049 static int __e1000_resume(struct pci_dev *pdev)
6050 {
6051 	struct net_device *netdev = pci_get_drvdata(pdev);
6052 	struct e1000_adapter *adapter = netdev_priv(netdev);
6053 	struct e1000_hw *hw = &adapter->hw;
6054 	u16 aspm_disable_flag = 0;
6055 	u32 err;
6056 
6057 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6058 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6059 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6060 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6061 	if (aspm_disable_flag)
6062 		e1000e_disable_aspm(pdev, aspm_disable_flag);
6063 
6064 	pci_set_master(pdev);
6065 
6066 	e1000e_set_interrupt_capability(adapter);
6067 	if (netif_running(netdev)) {
6068 		err = e1000_request_irq(adapter);
6069 		if (err)
6070 			return err;
6071 	}
6072 
6073 	if (hw->mac.type >= e1000_pch2lan)
6074 		e1000_resume_workarounds_pchlan(&adapter->hw);
6075 
6076 	e1000e_power_up_phy(adapter);
6077 
6078 	/* report the system wakeup cause from S3/S4 */
6079 	if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6080 		u16 phy_data;
6081 
6082 		e1e_rphy(&adapter->hw, BM_WUS, &phy_data);
6083 		if (phy_data) {
6084 			e_info("PHY Wakeup cause - %s\n",
6085 				phy_data & E1000_WUS_EX ? "Unicast Packet" :
6086 				phy_data & E1000_WUS_MC ? "Multicast Packet" :
6087 				phy_data & E1000_WUS_BC ? "Broadcast Packet" :
6088 				phy_data & E1000_WUS_MAG ? "Magic Packet" :
6089 				phy_data & E1000_WUS_LNKC ?
6090 				"Link Status Change" : "other");
6091 		}
6092 		e1e_wphy(&adapter->hw, BM_WUS, ~0);
6093 	} else {
6094 		u32 wus = er32(WUS);
6095 		if (wus) {
6096 			e_info("MAC Wakeup cause - %s\n",
6097 				wus & E1000_WUS_EX ? "Unicast Packet" :
6098 				wus & E1000_WUS_MC ? "Multicast Packet" :
6099 				wus & E1000_WUS_BC ? "Broadcast Packet" :
6100 				wus & E1000_WUS_MAG ? "Magic Packet" :
6101 				wus & E1000_WUS_LNKC ? "Link Status Change" :
6102 				"other");
6103 		}
6104 		ew32(WUS, ~0);
6105 	}
6106 
6107 	e1000e_reset(adapter);
6108 
6109 	e1000_init_manageability_pt(adapter);
6110 
6111 	if (netif_running(netdev))
6112 		e1000e_up(adapter);
6113 
6114 	netif_device_attach(netdev);
6115 
6116 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6117 	 * is up.  For all other cases, let the f/w know that the h/w is now
6118 	 * under the control of the driver.
6119 	 */
6120 	if (!(adapter->flags & FLAG_HAS_AMT))
6121 		e1000e_get_hw_control(adapter);
6122 
6123 	return 0;
6124 }
6125 
6126 #ifdef CONFIG_PM_SLEEP
6127 static int e1000_suspend(struct device *dev)
6128 {
6129 	struct pci_dev *pdev = to_pci_dev(dev);
6130 
6131 	return __e1000_shutdown(pdev, false);
6132 }
6133 
6134 static int e1000_resume(struct device *dev)
6135 {
6136 	struct pci_dev *pdev = to_pci_dev(dev);
6137 	struct net_device *netdev = pci_get_drvdata(pdev);
6138 	struct e1000_adapter *adapter = netdev_priv(netdev);
6139 
6140 	if (e1000e_pm_ready(adapter))
6141 		adapter->idle_check = true;
6142 
6143 	return __e1000_resume(pdev);
6144 }
6145 #endif /* CONFIG_PM_SLEEP */
6146 
6147 #ifdef CONFIG_PM_RUNTIME
6148 static int e1000_runtime_suspend(struct device *dev)
6149 {
6150 	struct pci_dev *pdev = to_pci_dev(dev);
6151 	struct net_device *netdev = pci_get_drvdata(pdev);
6152 	struct e1000_adapter *adapter = netdev_priv(netdev);
6153 
6154 	if (!e1000e_pm_ready(adapter))
6155 		return 0;
6156 
6157 	return __e1000_shutdown(pdev, true);
6158 }
6159 
6160 static int e1000_idle(struct device *dev)
6161 {
6162 	struct pci_dev *pdev = to_pci_dev(dev);
6163 	struct net_device *netdev = pci_get_drvdata(pdev);
6164 	struct e1000_adapter *adapter = netdev_priv(netdev);
6165 
6166 	if (!e1000e_pm_ready(adapter))
6167 		return 0;
6168 
6169 	if (adapter->idle_check) {
6170 		adapter->idle_check = false;
6171 		if (!e1000e_has_link(adapter))
6172 			pm_schedule_suspend(dev, MSEC_PER_SEC);
6173 	}
6174 
6175 	return -EBUSY;
6176 }
6177 
6178 static int e1000_runtime_resume(struct device *dev)
6179 {
6180 	struct pci_dev *pdev = to_pci_dev(dev);
6181 	struct net_device *netdev = pci_get_drvdata(pdev);
6182 	struct e1000_adapter *adapter = netdev_priv(netdev);
6183 
6184 	if (!e1000e_pm_ready(adapter))
6185 		return 0;
6186 
6187 	adapter->idle_check = !dev->power.runtime_auto;
6188 	return __e1000_resume(pdev);
6189 }
6190 #endif /* CONFIG_PM_RUNTIME */
6191 #endif /* CONFIG_PM */
6192 
6193 static void e1000_shutdown(struct pci_dev *pdev)
6194 {
6195 	__e1000_shutdown(pdev, false);
6196 }
6197 
6198 #ifdef CONFIG_NET_POLL_CONTROLLER
6199 
6200 static irqreturn_t e1000_intr_msix(int __always_unused irq, void *data)
6201 {
6202 	struct net_device *netdev = data;
6203 	struct e1000_adapter *adapter = netdev_priv(netdev);
6204 
6205 	if (adapter->msix_entries) {
6206 		int vector, msix_irq;
6207 
6208 		vector = 0;
6209 		msix_irq = adapter->msix_entries[vector].vector;
6210 		disable_irq(msix_irq);
6211 		e1000_intr_msix_rx(msix_irq, netdev);
6212 		enable_irq(msix_irq);
6213 
6214 		vector++;
6215 		msix_irq = adapter->msix_entries[vector].vector;
6216 		disable_irq(msix_irq);
6217 		e1000_intr_msix_tx(msix_irq, netdev);
6218 		enable_irq(msix_irq);
6219 
6220 		vector++;
6221 		msix_irq = adapter->msix_entries[vector].vector;
6222 		disable_irq(msix_irq);
6223 		e1000_msix_other(msix_irq, netdev);
6224 		enable_irq(msix_irq);
6225 	}
6226 
6227 	return IRQ_HANDLED;
6228 }
6229 
6230 /**
6231  * e1000_netpoll
6232  * @netdev: network interface device structure
6233  *
6234  * Polling 'interrupt' - used by things like netconsole to send skbs
6235  * without having to re-enable interrupts. It's not called while
6236  * the interrupt routine is executing.
6237  */
6238 static void e1000_netpoll(struct net_device *netdev)
6239 {
6240 	struct e1000_adapter *adapter = netdev_priv(netdev);
6241 
6242 	switch (adapter->int_mode) {
6243 	case E1000E_INT_MODE_MSIX:
6244 		e1000_intr_msix(adapter->pdev->irq, netdev);
6245 		break;
6246 	case E1000E_INT_MODE_MSI:
6247 		disable_irq(adapter->pdev->irq);
6248 		e1000_intr_msi(adapter->pdev->irq, netdev);
6249 		enable_irq(adapter->pdev->irq);
6250 		break;
6251 	default: /* E1000E_INT_MODE_LEGACY */
6252 		disable_irq(adapter->pdev->irq);
6253 		e1000_intr(adapter->pdev->irq, netdev);
6254 		enable_irq(adapter->pdev->irq);
6255 		break;
6256 	}
6257 }
6258 #endif
6259 
6260 /**
6261  * e1000_io_error_detected - called when PCI error is detected
6262  * @pdev: Pointer to PCI device
6263  * @state: The current pci connection state
6264  *
6265  * This function is called after a PCI bus error affecting
6266  * this device has been detected.
6267  */
6268 static pci_ers_result_t e1000_io_error_detected(struct pci_dev *pdev,
6269 						pci_channel_state_t state)
6270 {
6271 	struct net_device *netdev = pci_get_drvdata(pdev);
6272 	struct e1000_adapter *adapter = netdev_priv(netdev);
6273 
6274 	netif_device_detach(netdev);
6275 
6276 	if (state == pci_channel_io_perm_failure)
6277 		return PCI_ERS_RESULT_DISCONNECT;
6278 
6279 	if (netif_running(netdev))
6280 		e1000e_down(adapter);
6281 	pci_disable_device(pdev);
6282 
6283 	/* Request a slot slot reset. */
6284 	return PCI_ERS_RESULT_NEED_RESET;
6285 }
6286 
6287 /**
6288  * e1000_io_slot_reset - called after the pci bus has been reset.
6289  * @pdev: Pointer to PCI device
6290  *
6291  * Restart the card from scratch, as if from a cold-boot. Implementation
6292  * resembles the first-half of the e1000_resume routine.
6293  */
6294 static pci_ers_result_t e1000_io_slot_reset(struct pci_dev *pdev)
6295 {
6296 	struct net_device *netdev = pci_get_drvdata(pdev);
6297 	struct e1000_adapter *adapter = netdev_priv(netdev);
6298 	struct e1000_hw *hw = &adapter->hw;
6299 	u16 aspm_disable_flag = 0;
6300 	int err;
6301 	pci_ers_result_t result;
6302 
6303 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6304 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6305 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6306 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6307 	if (aspm_disable_flag)
6308 		e1000e_disable_aspm(pdev, aspm_disable_flag);
6309 
6310 	err = pci_enable_device_mem(pdev);
6311 	if (err) {
6312 		dev_err(&pdev->dev,
6313 			"Cannot re-enable PCI device after reset.\n");
6314 		result = PCI_ERS_RESULT_DISCONNECT;
6315 	} else {
6316 		pdev->state_saved = true;
6317 		pci_restore_state(pdev);
6318 		pci_set_master(pdev);
6319 
6320 		pci_enable_wake(pdev, PCI_D3hot, 0);
6321 		pci_enable_wake(pdev, PCI_D3cold, 0);
6322 
6323 		e1000e_reset(adapter);
6324 		ew32(WUS, ~0);
6325 		result = PCI_ERS_RESULT_RECOVERED;
6326 	}
6327 
6328 	pci_cleanup_aer_uncorrect_error_status(pdev);
6329 
6330 	return result;
6331 }
6332 
6333 /**
6334  * e1000_io_resume - called when traffic can start flowing again.
6335  * @pdev: Pointer to PCI device
6336  *
6337  * This callback is called when the error recovery driver tells us that
6338  * its OK to resume normal operation. Implementation resembles the
6339  * second-half of the e1000_resume routine.
6340  */
6341 static void e1000_io_resume(struct pci_dev *pdev)
6342 {
6343 	struct net_device *netdev = pci_get_drvdata(pdev);
6344 	struct e1000_adapter *adapter = netdev_priv(netdev);
6345 
6346 	e1000_init_manageability_pt(adapter);
6347 
6348 	if (netif_running(netdev)) {
6349 		if (e1000e_up(adapter)) {
6350 			dev_err(&pdev->dev,
6351 				"can't bring device back up after reset\n");
6352 			return;
6353 		}
6354 	}
6355 
6356 	netif_device_attach(netdev);
6357 
6358 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6359 	 * is up.  For all other cases, let the f/w know that the h/w is now
6360 	 * under the control of the driver.
6361 	 */
6362 	if (!(adapter->flags & FLAG_HAS_AMT))
6363 		e1000e_get_hw_control(adapter);
6364 }
6365 
6366 static void e1000_print_device_info(struct e1000_adapter *adapter)
6367 {
6368 	struct e1000_hw *hw = &adapter->hw;
6369 	struct net_device *netdev = adapter->netdev;
6370 	u32 ret_val;
6371 	u8 pba_str[E1000_PBANUM_LENGTH];
6372 
6373 	/* print bus type/speed/width info */
6374 	e_info("(PCI Express:2.5GT/s:%s) %pM\n",
6375 	       /* bus width */
6376 	       ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
6377 	        "Width x1"),
6378 	       /* MAC address */
6379 	       netdev->dev_addr);
6380 	e_info("Intel(R) PRO/%s Network Connection\n",
6381 	       (hw->phy.type == e1000_phy_ife) ? "10/100" : "1000");
6382 	ret_val = e1000_read_pba_string_generic(hw, pba_str,
6383 						E1000_PBANUM_LENGTH);
6384 	if (ret_val)
6385 		strlcpy((char *)pba_str, "Unknown", sizeof(pba_str));
6386 	e_info("MAC: %d, PHY: %d, PBA No: %s\n",
6387 	       hw->mac.type, hw->phy.type, pba_str);
6388 }
6389 
6390 static void e1000_eeprom_checks(struct e1000_adapter *adapter)
6391 {
6392 	struct e1000_hw *hw = &adapter->hw;
6393 	int ret_val;
6394 	u16 buf = 0;
6395 
6396 	if (hw->mac.type != e1000_82573)
6397 		return;
6398 
6399 	ret_val = e1000_read_nvm(hw, NVM_INIT_CONTROL2_REG, 1, &buf);
6400 	le16_to_cpus(&buf);
6401 	if (!ret_val && (!(buf & (1 << 0)))) {
6402 		/* Deep Smart Power Down (DSPD) */
6403 		dev_warn(&adapter->pdev->dev,
6404 			 "Warning: detected DSPD enabled in EEPROM\n");
6405 	}
6406 }
6407 
6408 static int e1000_set_features(struct net_device *netdev,
6409 			      netdev_features_t features)
6410 {
6411 	struct e1000_adapter *adapter = netdev_priv(netdev);
6412 	netdev_features_t changed = features ^ netdev->features;
6413 
6414 	if (changed & (NETIF_F_TSO | NETIF_F_TSO6))
6415 		adapter->flags |= FLAG_TSO_FORCE;
6416 
6417 	if (!(changed & (NETIF_F_HW_VLAN_RX | NETIF_F_HW_VLAN_TX |
6418 			 NETIF_F_RXCSUM | NETIF_F_RXHASH | NETIF_F_RXFCS |
6419 			 NETIF_F_RXALL)))
6420 		return 0;
6421 
6422 	if (changed & NETIF_F_RXFCS) {
6423 		if (features & NETIF_F_RXFCS) {
6424 			adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
6425 		} else {
6426 			/* We need to take it back to defaults, which might mean
6427 			 * stripping is still disabled at the adapter level.
6428 			 */
6429 			if (adapter->flags2 & FLAG2_DFLT_CRC_STRIPPING)
6430 				adapter->flags2 |= FLAG2_CRC_STRIPPING;
6431 			else
6432 				adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
6433 		}
6434 	}
6435 
6436 	netdev->features = features;
6437 
6438 	if (netif_running(netdev))
6439 		e1000e_reinit_locked(adapter);
6440 	else
6441 		e1000e_reset(adapter);
6442 
6443 	return 0;
6444 }
6445 
6446 static const struct net_device_ops e1000e_netdev_ops = {
6447 	.ndo_open		= e1000_open,
6448 	.ndo_stop		= e1000_close,
6449 	.ndo_start_xmit		= e1000_xmit_frame,
6450 	.ndo_get_stats64	= e1000e_get_stats64,
6451 	.ndo_set_rx_mode	= e1000e_set_rx_mode,
6452 	.ndo_set_mac_address	= e1000_set_mac,
6453 	.ndo_change_mtu		= e1000_change_mtu,
6454 	.ndo_do_ioctl		= e1000_ioctl,
6455 	.ndo_tx_timeout		= e1000_tx_timeout,
6456 	.ndo_validate_addr	= eth_validate_addr,
6457 
6458 	.ndo_vlan_rx_add_vid	= e1000_vlan_rx_add_vid,
6459 	.ndo_vlan_rx_kill_vid	= e1000_vlan_rx_kill_vid,
6460 #ifdef CONFIG_NET_POLL_CONTROLLER
6461 	.ndo_poll_controller	= e1000_netpoll,
6462 #endif
6463 	.ndo_set_features = e1000_set_features,
6464 };
6465 
6466 /**
6467  * e1000_probe - Device Initialization Routine
6468  * @pdev: PCI device information struct
6469  * @ent: entry in e1000_pci_tbl
6470  *
6471  * Returns 0 on success, negative on failure
6472  *
6473  * e1000_probe initializes an adapter identified by a pci_dev structure.
6474  * The OS initialization, configuring of the adapter private structure,
6475  * and a hardware reset occur.
6476  **/
6477 static int e1000_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
6478 {
6479 	struct net_device *netdev;
6480 	struct e1000_adapter *adapter;
6481 	struct e1000_hw *hw;
6482 	const struct e1000_info *ei = e1000_info_tbl[ent->driver_data];
6483 	resource_size_t mmio_start, mmio_len;
6484 	resource_size_t flash_start, flash_len;
6485 	static int cards_found;
6486 	u16 aspm_disable_flag = 0;
6487 	int i, err, pci_using_dac;
6488 	u16 eeprom_data = 0;
6489 	u16 eeprom_apme_mask = E1000_EEPROM_APME;
6490 
6491 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L0S)
6492 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6493 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L1)
6494 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6495 	if (aspm_disable_flag)
6496 		e1000e_disable_aspm(pdev, aspm_disable_flag);
6497 
6498 	err = pci_enable_device_mem(pdev);
6499 	if (err)
6500 		return err;
6501 
6502 	pci_using_dac = 0;
6503 	err = dma_set_mask(&pdev->dev, DMA_BIT_MASK(64));
6504 	if (!err) {
6505 		err = dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(64));
6506 		if (!err)
6507 			pci_using_dac = 1;
6508 	} else {
6509 		err = dma_set_mask(&pdev->dev, DMA_BIT_MASK(32));
6510 		if (err) {
6511 			err = dma_set_coherent_mask(&pdev->dev,
6512 						    DMA_BIT_MASK(32));
6513 			if (err) {
6514 				dev_err(&pdev->dev, "No usable DMA configuration, aborting\n");
6515 				goto err_dma;
6516 			}
6517 		}
6518 	}
6519 
6520 	err = pci_request_selected_regions_exclusive(pdev,
6521 					  pci_select_bars(pdev, IORESOURCE_MEM),
6522 					  e1000e_driver_name);
6523 	if (err)
6524 		goto err_pci_reg;
6525 
6526 	/* AER (Advanced Error Reporting) hooks */
6527 	pci_enable_pcie_error_reporting(pdev);
6528 
6529 	pci_set_master(pdev);
6530 	/* PCI config space info */
6531 	err = pci_save_state(pdev);
6532 	if (err)
6533 		goto err_alloc_etherdev;
6534 
6535 	err = -ENOMEM;
6536 	netdev = alloc_etherdev(sizeof(struct e1000_adapter));
6537 	if (!netdev)
6538 		goto err_alloc_etherdev;
6539 
6540 	SET_NETDEV_DEV(netdev, &pdev->dev);
6541 
6542 	netdev->irq = pdev->irq;
6543 
6544 	pci_set_drvdata(pdev, netdev);
6545 	adapter = netdev_priv(netdev);
6546 	hw = &adapter->hw;
6547 	adapter->netdev = netdev;
6548 	adapter->pdev = pdev;
6549 	adapter->ei = ei;
6550 	adapter->pba = ei->pba;
6551 	adapter->flags = ei->flags;
6552 	adapter->flags2 = ei->flags2;
6553 	adapter->hw.adapter = adapter;
6554 	adapter->hw.mac.type = ei->mac;
6555 	adapter->max_hw_frame_size = ei->max_hw_frame_size;
6556 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
6557 
6558 	mmio_start = pci_resource_start(pdev, 0);
6559 	mmio_len = pci_resource_len(pdev, 0);
6560 
6561 	err = -EIO;
6562 	adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
6563 	if (!adapter->hw.hw_addr)
6564 		goto err_ioremap;
6565 
6566 	if ((adapter->flags & FLAG_HAS_FLASH) &&
6567 	    (pci_resource_flags(pdev, 1) & IORESOURCE_MEM)) {
6568 		flash_start = pci_resource_start(pdev, 1);
6569 		flash_len = pci_resource_len(pdev, 1);
6570 		adapter->hw.flash_address = ioremap(flash_start, flash_len);
6571 		if (!adapter->hw.flash_address)
6572 			goto err_flashmap;
6573 	}
6574 
6575 	/* construct the net_device struct */
6576 	netdev->netdev_ops		= &e1000e_netdev_ops;
6577 	e1000e_set_ethtool_ops(netdev);
6578 	netdev->watchdog_timeo		= 5 * HZ;
6579 	netif_napi_add(netdev, &adapter->napi, e1000e_poll, 64);
6580 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
6581 
6582 	netdev->mem_start = mmio_start;
6583 	netdev->mem_end = mmio_start + mmio_len;
6584 
6585 	adapter->bd_number = cards_found++;
6586 
6587 	e1000e_check_options(adapter);
6588 
6589 	/* setup adapter struct */
6590 	err = e1000_sw_init(adapter);
6591 	if (err)
6592 		goto err_sw_init;
6593 
6594 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
6595 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
6596 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
6597 
6598 	err = ei->get_variants(adapter);
6599 	if (err)
6600 		goto err_hw_init;
6601 
6602 	if ((adapter->flags & FLAG_IS_ICH) &&
6603 	    (adapter->flags & FLAG_READ_ONLY_NVM))
6604 		e1000e_write_protect_nvm_ich8lan(&adapter->hw);
6605 
6606 	hw->mac.ops.get_bus_info(&adapter->hw);
6607 
6608 	adapter->hw.phy.autoneg_wait_to_complete = 0;
6609 
6610 	/* Copper options */
6611 	if (adapter->hw.phy.media_type == e1000_media_type_copper) {
6612 		adapter->hw.phy.mdix = AUTO_ALL_MODES;
6613 		adapter->hw.phy.disable_polarity_correction = 0;
6614 		adapter->hw.phy.ms_type = e1000_ms_hw_default;
6615 	}
6616 
6617 	if (hw->phy.ops.check_reset_block && hw->phy.ops.check_reset_block(hw))
6618 		dev_info(&pdev->dev,
6619 			 "PHY reset is blocked due to SOL/IDER session.\n");
6620 
6621 	/* Set initial default active device features */
6622 	netdev->features = (NETIF_F_SG |
6623 			    NETIF_F_HW_VLAN_RX |
6624 			    NETIF_F_HW_VLAN_TX |
6625 			    NETIF_F_TSO |
6626 			    NETIF_F_TSO6 |
6627 			    NETIF_F_RXHASH |
6628 			    NETIF_F_RXCSUM |
6629 			    NETIF_F_HW_CSUM);
6630 
6631 	/* Set user-changeable features (subset of all device features) */
6632 	netdev->hw_features = netdev->features;
6633 	netdev->hw_features |= NETIF_F_RXFCS;
6634 	netdev->priv_flags |= IFF_SUPP_NOFCS;
6635 	netdev->hw_features |= NETIF_F_RXALL;
6636 
6637 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER)
6638 		netdev->features |= NETIF_F_HW_VLAN_FILTER;
6639 
6640 	netdev->vlan_features |= (NETIF_F_SG |
6641 				  NETIF_F_TSO |
6642 				  NETIF_F_TSO6 |
6643 				  NETIF_F_HW_CSUM);
6644 
6645 	netdev->priv_flags |= IFF_UNICAST_FLT;
6646 
6647 	if (pci_using_dac) {
6648 		netdev->features |= NETIF_F_HIGHDMA;
6649 		netdev->vlan_features |= NETIF_F_HIGHDMA;
6650 	}
6651 
6652 	if (e1000e_enable_mng_pass_thru(&adapter->hw))
6653 		adapter->flags |= FLAG_MNG_PT_ENABLED;
6654 
6655 	/* before reading the NVM, reset the controller to
6656 	 * put the device in a known good starting state
6657 	 */
6658 	adapter->hw.mac.ops.reset_hw(&adapter->hw);
6659 
6660 	/* systems with ASPM and others may see the checksum fail on the first
6661 	 * attempt. Let's give it a few tries
6662 	 */
6663 	for (i = 0;; i++) {
6664 		if (e1000_validate_nvm_checksum(&adapter->hw) >= 0)
6665 			break;
6666 		if (i == 2) {
6667 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
6668 			err = -EIO;
6669 			goto err_eeprom;
6670 		}
6671 	}
6672 
6673 	e1000_eeprom_checks(adapter);
6674 
6675 	/* copy the MAC address */
6676 	if (e1000e_read_mac_addr(&adapter->hw))
6677 		dev_err(&pdev->dev,
6678 			"NVM Read Error while reading MAC address\n");
6679 
6680 	memcpy(netdev->dev_addr, adapter->hw.mac.addr, netdev->addr_len);
6681 
6682 	if (!is_valid_ether_addr(netdev->dev_addr)) {
6683 		dev_err(&pdev->dev, "Invalid MAC Address: %pM\n",
6684 			netdev->dev_addr);
6685 		err = -EIO;
6686 		goto err_eeprom;
6687 	}
6688 
6689 	init_timer(&adapter->watchdog_timer);
6690 	adapter->watchdog_timer.function = e1000_watchdog;
6691 	adapter->watchdog_timer.data = (unsigned long) adapter;
6692 
6693 	init_timer(&adapter->phy_info_timer);
6694 	adapter->phy_info_timer.function = e1000_update_phy_info;
6695 	adapter->phy_info_timer.data = (unsigned long) adapter;
6696 
6697 	INIT_WORK(&adapter->reset_task, e1000_reset_task);
6698 	INIT_WORK(&adapter->watchdog_task, e1000_watchdog_task);
6699 	INIT_WORK(&adapter->downshift_task, e1000e_downshift_workaround);
6700 	INIT_WORK(&adapter->update_phy_task, e1000e_update_phy_task);
6701 	INIT_WORK(&adapter->print_hang_task, e1000_print_hw_hang);
6702 
6703 	/* Initialize link parameters. User can change them with ethtool */
6704 	adapter->hw.mac.autoneg = 1;
6705 	adapter->fc_autoneg = true;
6706 	adapter->hw.fc.requested_mode = e1000_fc_default;
6707 	adapter->hw.fc.current_mode = e1000_fc_default;
6708 	adapter->hw.phy.autoneg_advertised = 0x2f;
6709 
6710 	/* ring size defaults */
6711 	adapter->rx_ring->count = E1000_DEFAULT_RXD;
6712 	adapter->tx_ring->count = E1000_DEFAULT_TXD;
6713 
6714 	/* Initial Wake on LAN setting - If APM wake is enabled in
6715 	 * the EEPROM, enable the ACPI Magic Packet filter
6716 	 */
6717 	if (adapter->flags & FLAG_APME_IN_WUC) {
6718 		/* APME bit in EEPROM is mapped to WUC.APME */
6719 		eeprom_data = er32(WUC);
6720 		eeprom_apme_mask = E1000_WUC_APME;
6721 		if ((hw->mac.type > e1000_ich10lan) &&
6722 		    (eeprom_data & E1000_WUC_PHY_WAKE))
6723 			adapter->flags2 |= FLAG2_HAS_PHY_WAKEUP;
6724 	} else if (adapter->flags & FLAG_APME_IN_CTRL3) {
6725 		if (adapter->flags & FLAG_APME_CHECK_PORT_B &&
6726 		    (adapter->hw.bus.func == 1))
6727 			e1000_read_nvm(&adapter->hw, NVM_INIT_CONTROL3_PORT_B,
6728 				       1, &eeprom_data);
6729 		else
6730 			e1000_read_nvm(&adapter->hw, NVM_INIT_CONTROL3_PORT_A,
6731 				       1, &eeprom_data);
6732 	}
6733 
6734 	/* fetch WoL from EEPROM */
6735 	if (eeprom_data & eeprom_apme_mask)
6736 		adapter->eeprom_wol |= E1000_WUFC_MAG;
6737 
6738 	/* now that we have the eeprom settings, apply the special cases
6739 	 * where the eeprom may be wrong or the board simply won't support
6740 	 * wake on lan on a particular port
6741 	 */
6742 	if (!(adapter->flags & FLAG_HAS_WOL))
6743 		adapter->eeprom_wol = 0;
6744 
6745 	/* initialize the wol settings based on the eeprom settings */
6746 	adapter->wol = adapter->eeprom_wol;
6747 
6748 	/* make sure adapter isn't asleep if manageability is enabled */
6749 	if (adapter->wol || (adapter->flags & FLAG_MNG_PT_ENABLED) ||
6750 	    (hw->mac.ops.check_mng_mode(hw)))
6751 		device_wakeup_enable(&pdev->dev);
6752 
6753 	/* save off EEPROM version number */
6754 	e1000_read_nvm(&adapter->hw, 5, 1, &adapter->eeprom_vers);
6755 
6756 	/* reset the hardware with the new settings */
6757 	e1000e_reset(adapter);
6758 
6759 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6760 	 * is up.  For all other cases, let the f/w know that the h/w is now
6761 	 * under the control of the driver.
6762 	 */
6763 	if (!(adapter->flags & FLAG_HAS_AMT))
6764 		e1000e_get_hw_control(adapter);
6765 
6766 	strlcpy(netdev->name, "eth%d", sizeof(netdev->name));
6767 	err = register_netdev(netdev);
6768 	if (err)
6769 		goto err_register;
6770 
6771 	/* carrier off reporting is important to ethtool even BEFORE open */
6772 	netif_carrier_off(netdev);
6773 
6774 	/* init PTP hardware clock */
6775 	e1000e_ptp_init(adapter);
6776 
6777 	e1000_print_device_info(adapter);
6778 
6779 	if (pci_dev_run_wake(pdev))
6780 		pm_runtime_put_noidle(&pdev->dev);
6781 
6782 	return 0;
6783 
6784 err_register:
6785 	if (!(adapter->flags & FLAG_HAS_AMT))
6786 		e1000e_release_hw_control(adapter);
6787 err_eeprom:
6788 	if (hw->phy.ops.check_reset_block && !hw->phy.ops.check_reset_block(hw))
6789 		e1000_phy_hw_reset(&adapter->hw);
6790 err_hw_init:
6791 	kfree(adapter->tx_ring);
6792 	kfree(adapter->rx_ring);
6793 err_sw_init:
6794 	if (adapter->hw.flash_address)
6795 		iounmap(adapter->hw.flash_address);
6796 	e1000e_reset_interrupt_capability(adapter);
6797 err_flashmap:
6798 	iounmap(adapter->hw.hw_addr);
6799 err_ioremap:
6800 	free_netdev(netdev);
6801 err_alloc_etherdev:
6802 	pci_release_selected_regions(pdev,
6803 	                             pci_select_bars(pdev, IORESOURCE_MEM));
6804 err_pci_reg:
6805 err_dma:
6806 	pci_disable_device(pdev);
6807 	return err;
6808 }
6809 
6810 /**
6811  * e1000_remove - Device Removal Routine
6812  * @pdev: PCI device information struct
6813  *
6814  * e1000_remove is called by the PCI subsystem to alert the driver
6815  * that it should release a PCI device.  The could be caused by a
6816  * Hot-Plug event, or because the driver is going to be removed from
6817  * memory.
6818  **/
6819 static void e1000_remove(struct pci_dev *pdev)
6820 {
6821 	struct net_device *netdev = pci_get_drvdata(pdev);
6822 	struct e1000_adapter *adapter = netdev_priv(netdev);
6823 	bool down = test_bit(__E1000_DOWN, &adapter->state);
6824 
6825 	e1000e_ptp_remove(adapter);
6826 
6827 	/* The timers may be rescheduled, so explicitly disable them
6828 	 * from being rescheduled.
6829 	 */
6830 	if (!down)
6831 		set_bit(__E1000_DOWN, &adapter->state);
6832 	del_timer_sync(&adapter->watchdog_timer);
6833 	del_timer_sync(&adapter->phy_info_timer);
6834 
6835 	cancel_work_sync(&adapter->reset_task);
6836 	cancel_work_sync(&adapter->watchdog_task);
6837 	cancel_work_sync(&adapter->downshift_task);
6838 	cancel_work_sync(&adapter->update_phy_task);
6839 	cancel_work_sync(&adapter->print_hang_task);
6840 
6841 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
6842 		cancel_work_sync(&adapter->tx_hwtstamp_work);
6843 		if (adapter->tx_hwtstamp_skb) {
6844 			dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
6845 			adapter->tx_hwtstamp_skb = NULL;
6846 		}
6847 	}
6848 
6849 	if (!(netdev->flags & IFF_UP))
6850 		e1000_power_down_phy(adapter);
6851 
6852 	/* Don't lie to e1000_close() down the road. */
6853 	if (!down)
6854 		clear_bit(__E1000_DOWN, &adapter->state);
6855 	unregister_netdev(netdev);
6856 
6857 	if (pci_dev_run_wake(pdev))
6858 		pm_runtime_get_noresume(&pdev->dev);
6859 
6860 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
6861 	 * would have already happened in close and is redundant.
6862 	 */
6863 	e1000e_release_hw_control(adapter);
6864 
6865 	e1000e_reset_interrupt_capability(adapter);
6866 	kfree(adapter->tx_ring);
6867 	kfree(adapter->rx_ring);
6868 
6869 	iounmap(adapter->hw.hw_addr);
6870 	if (adapter->hw.flash_address)
6871 		iounmap(adapter->hw.flash_address);
6872 	pci_release_selected_regions(pdev,
6873 	                             pci_select_bars(pdev, IORESOURCE_MEM));
6874 
6875 	free_netdev(netdev);
6876 
6877 	/* AER disable */
6878 	pci_disable_pcie_error_reporting(pdev);
6879 
6880 	pci_disable_device(pdev);
6881 }
6882 
6883 /* PCI Error Recovery (ERS) */
6884 static const struct pci_error_handlers e1000_err_handler = {
6885 	.error_detected = e1000_io_error_detected,
6886 	.slot_reset = e1000_io_slot_reset,
6887 	.resume = e1000_io_resume,
6888 };
6889 
6890 static DEFINE_PCI_DEVICE_TABLE(e1000_pci_tbl) = {
6891 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_COPPER), board_82571 },
6892 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_FIBER), board_82571 },
6893 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER), board_82571 },
6894 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER_LP), board_82571 },
6895 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_FIBER), board_82571 },
6896 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES), board_82571 },
6897 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_DUAL), board_82571 },
6898 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_QUAD), board_82571 },
6899 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571PT_QUAD_COPPER), board_82571 },
6900 
6901 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI), board_82572 },
6902 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_COPPER), board_82572 },
6903 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_FIBER), board_82572 },
6904 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_SERDES), board_82572 },
6905 
6906 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E), board_82573 },
6907 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E_IAMT), board_82573 },
6908 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573L), board_82573 },
6909 
6910 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574L), board_82574 },
6911 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574LA), board_82574 },
6912 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82583V), board_82583 },
6913 
6914 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_DPT),
6915 	  board_80003es2lan },
6916 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_SPT),
6917 	  board_80003es2lan },
6918 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_DPT),
6919 	  board_80003es2lan },
6920 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_SPT),
6921 	  board_80003es2lan },
6922 
6923 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE), board_ich8lan },
6924 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_G), board_ich8lan },
6925 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_GT), board_ich8lan },
6926 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_AMT), board_ich8lan },
6927 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_C), board_ich8lan },
6928 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M), board_ich8lan },
6929 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M_AMT), board_ich8lan },
6930 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_82567V_3), board_ich8lan },
6931 
6932 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE), board_ich9lan },
6933 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_G), board_ich9lan },
6934 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_GT), board_ich9lan },
6935 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_AMT), board_ich9lan },
6936 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_C), board_ich9lan },
6937 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_BM), board_ich9lan },
6938 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M), board_ich9lan },
6939 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_AMT), board_ich9lan },
6940 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_V), board_ich9lan },
6941 
6942 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LM), board_ich9lan },
6943 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LF), board_ich9lan },
6944 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_V), board_ich9lan },
6945 
6946 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LM), board_ich10lan },
6947 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LF), board_ich10lan },
6948 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_V), board_ich10lan },
6949 
6950 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LM), board_pchlan },
6951 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LC), board_pchlan },
6952 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DM), board_pchlan },
6953 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DC), board_pchlan },
6954 
6955 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_LM), board_pch2lan },
6956 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_V), board_pch2lan },
6957 
6958 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_LM), board_pch_lpt },
6959 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_V), board_pch_lpt },
6960 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_LM), board_pch_lpt },
6961 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_V), board_pch_lpt },
6962 
6963 	{ 0, 0, 0, 0, 0, 0, 0 }	/* terminate list */
6964 };
6965 MODULE_DEVICE_TABLE(pci, e1000_pci_tbl);
6966 
6967 #ifdef CONFIG_PM
6968 static const struct dev_pm_ops e1000_pm_ops = {
6969 	SET_SYSTEM_SLEEP_PM_OPS(e1000_suspend, e1000_resume)
6970 	SET_RUNTIME_PM_OPS(e1000_runtime_suspend,
6971 				e1000_runtime_resume, e1000_idle)
6972 };
6973 #endif
6974 
6975 /* PCI Device API Driver */
6976 static struct pci_driver e1000_driver = {
6977 	.name     = e1000e_driver_name,
6978 	.id_table = e1000_pci_tbl,
6979 	.probe    = e1000_probe,
6980 	.remove   = e1000_remove,
6981 #ifdef CONFIG_PM
6982 	.driver   = {
6983 		.pm = &e1000_pm_ops,
6984 	},
6985 #endif
6986 	.shutdown = e1000_shutdown,
6987 	.err_handler = &e1000_err_handler
6988 };
6989 
6990 /**
6991  * e1000_init_module - Driver Registration Routine
6992  *
6993  * e1000_init_module is the first routine called when the driver is
6994  * loaded. All it does is register with the PCI subsystem.
6995  **/
6996 static int __init e1000_init_module(void)
6997 {
6998 	int ret;
6999 	pr_info("Intel(R) PRO/1000 Network Driver - %s\n",
7000 		e1000e_driver_version);
7001 	pr_info("Copyright(c) 1999 - 2013 Intel Corporation.\n");
7002 	ret = pci_register_driver(&e1000_driver);
7003 
7004 	return ret;
7005 }
7006 module_init(e1000_init_module);
7007 
7008 /**
7009  * e1000_exit_module - Driver Exit Cleanup Routine
7010  *
7011  * e1000_exit_module is called just before the driver is removed
7012  * from memory.
7013  **/
7014 static void __exit e1000_exit_module(void)
7015 {
7016 	pci_unregister_driver(&e1000_driver);
7017 }
7018 module_exit(e1000_exit_module);
7019 
7020 
7021 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
7022 MODULE_DESCRIPTION("Intel(R) PRO/1000 Network Driver");
7023 MODULE_LICENSE("GPL");
7024 MODULE_VERSION(DRV_VERSION);
7025 
7026 /* netdev.c */
7027