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 "2.3.2" 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 const struct e1000_info *e1000_info_tbl[] = {
68 	[board_82571]		= &e1000_82571_info,
69 	[board_82572]		= &e1000_82572_info,
70 	[board_82573]		= &e1000_82573_info,
71 	[board_82574]		= &e1000_82574_info,
72 	[board_82583]		= &e1000_82583_info,
73 	[board_80003es2lan]	= &e1000_es2_info,
74 	[board_ich8lan]		= &e1000_ich8_info,
75 	[board_ich9lan]		= &e1000_ich9_info,
76 	[board_ich10lan]	= &e1000_ich10_info,
77 	[board_pchlan]		= &e1000_pch_info,
78 	[board_pch2lan]		= &e1000_pch2_info,
79 	[board_pch_lpt]		= &e1000_pch_lpt_info,
80 };
81 
82 struct e1000_reg_info {
83 	u32 ofs;
84 	char *name;
85 };
86 
87 static const struct e1000_reg_info e1000_reg_info_tbl[] = {
88 	/* General Registers */
89 	{E1000_CTRL, "CTRL"},
90 	{E1000_STATUS, "STATUS"},
91 	{E1000_CTRL_EXT, "CTRL_EXT"},
92 
93 	/* Interrupt Registers */
94 	{E1000_ICR, "ICR"},
95 
96 	/* Rx Registers */
97 	{E1000_RCTL, "RCTL"},
98 	{E1000_RDLEN(0), "RDLEN"},
99 	{E1000_RDH(0), "RDH"},
100 	{E1000_RDT(0), "RDT"},
101 	{E1000_RDTR, "RDTR"},
102 	{E1000_RXDCTL(0), "RXDCTL"},
103 	{E1000_ERT, "ERT"},
104 	{E1000_RDBAL(0), "RDBAL"},
105 	{E1000_RDBAH(0), "RDBAH"},
106 	{E1000_RDFH, "RDFH"},
107 	{E1000_RDFT, "RDFT"},
108 	{E1000_RDFHS, "RDFHS"},
109 	{E1000_RDFTS, "RDFTS"},
110 	{E1000_RDFPC, "RDFPC"},
111 
112 	/* Tx Registers */
113 	{E1000_TCTL, "TCTL"},
114 	{E1000_TDBAL(0), "TDBAL"},
115 	{E1000_TDBAH(0), "TDBAH"},
116 	{E1000_TDLEN(0), "TDLEN"},
117 	{E1000_TDH(0), "TDH"},
118 	{E1000_TDT(0), "TDT"},
119 	{E1000_TIDV, "TIDV"},
120 	{E1000_TXDCTL(0), "TXDCTL"},
121 	{E1000_TADV, "TADV"},
122 	{E1000_TARC(0), "TARC"},
123 	{E1000_TDFH, "TDFH"},
124 	{E1000_TDFT, "TDFT"},
125 	{E1000_TDFHS, "TDFHS"},
126 	{E1000_TDFTS, "TDFTS"},
127 	{E1000_TDFPC, "TDFPC"},
128 
129 	/* List Terminator */
130 	{0, NULL}
131 };
132 
133 /**
134  * e1000_regdump - register printout routine
135  * @hw: pointer to the HW structure
136  * @reginfo: pointer to the register info table
137  **/
138 static void e1000_regdump(struct e1000_hw *hw, struct e1000_reg_info *reginfo)
139 {
140 	int n = 0;
141 	char rname[16];
142 	u32 regs[8];
143 
144 	switch (reginfo->ofs) {
145 	case E1000_RXDCTL(0):
146 		for (n = 0; n < 2; n++)
147 			regs[n] = __er32(hw, E1000_RXDCTL(n));
148 		break;
149 	case E1000_TXDCTL(0):
150 		for (n = 0; n < 2; n++)
151 			regs[n] = __er32(hw, E1000_TXDCTL(n));
152 		break;
153 	case E1000_TARC(0):
154 		for (n = 0; n < 2; n++)
155 			regs[n] = __er32(hw, E1000_TARC(n));
156 		break;
157 	default:
158 		pr_info("%-15s %08x\n",
159 			reginfo->name, __er32(hw, reginfo->ofs));
160 		return;
161 	}
162 
163 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-1]");
164 	pr_info("%-15s %08x %08x\n", rname, regs[0], regs[1]);
165 }
166 
167 static void e1000e_dump_ps_pages(struct e1000_adapter *adapter,
168 				 struct e1000_buffer *bi)
169 {
170 	int i;
171 	struct e1000_ps_page *ps_page;
172 
173 	for (i = 0; i < adapter->rx_ps_pages; i++) {
174 		ps_page = &bi->ps_pages[i];
175 
176 		if (ps_page->page) {
177 			pr_info("packet dump for ps_page %d:\n", i);
178 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
179 				       16, 1, page_address(ps_page->page),
180 				       PAGE_SIZE, true);
181 		}
182 	}
183 }
184 
185 /**
186  * e1000e_dump - Print registers, Tx-ring and Rx-ring
187  * @adapter: board private structure
188  **/
189 static void e1000e_dump(struct e1000_adapter *adapter)
190 {
191 	struct net_device *netdev = adapter->netdev;
192 	struct e1000_hw *hw = &adapter->hw;
193 	struct e1000_reg_info *reginfo;
194 	struct e1000_ring *tx_ring = adapter->tx_ring;
195 	struct e1000_tx_desc *tx_desc;
196 	struct my_u0 {
197 		__le64 a;
198 		__le64 b;
199 	} *u0;
200 	struct e1000_buffer *buffer_info;
201 	struct e1000_ring *rx_ring = adapter->rx_ring;
202 	union e1000_rx_desc_packet_split *rx_desc_ps;
203 	union e1000_rx_desc_extended *rx_desc;
204 	struct my_u1 {
205 		__le64 a;
206 		__le64 b;
207 		__le64 c;
208 		__le64 d;
209 	} *u1;
210 	u32 staterr;
211 	int i = 0;
212 
213 	if (!netif_msg_hw(adapter))
214 		return;
215 
216 	/* Print netdevice Info */
217 	if (netdev) {
218 		dev_info(&adapter->pdev->dev, "Net device Info\n");
219 		pr_info("Device Name     state            trans_start      last_rx\n");
220 		pr_info("%-15s %016lX %016lX %016lX\n", netdev->name,
221 			netdev->state, netdev->trans_start, netdev->last_rx);
222 	}
223 
224 	/* Print Registers */
225 	dev_info(&adapter->pdev->dev, "Register Dump\n");
226 	pr_info(" Register Name   Value\n");
227 	for (reginfo = (struct e1000_reg_info *)e1000_reg_info_tbl;
228 	     reginfo->name; reginfo++) {
229 		e1000_regdump(hw, reginfo);
230 	}
231 
232 	/* Print Tx Ring Summary */
233 	if (!netdev || !netif_running(netdev))
234 		return;
235 
236 	dev_info(&adapter->pdev->dev, "Tx Ring Summary\n");
237 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
238 	buffer_info = &tx_ring->buffer_info[tx_ring->next_to_clean];
239 	pr_info(" %5d %5X %5X %016llX %04X %3X %016llX\n",
240 		0, tx_ring->next_to_use, tx_ring->next_to_clean,
241 		(unsigned long long)buffer_info->dma,
242 		buffer_info->length,
243 		buffer_info->next_to_watch,
244 		(unsigned long long)buffer_info->time_stamp);
245 
246 	/* Print Tx Ring */
247 	if (!netif_msg_tx_done(adapter))
248 		goto rx_ring_summary;
249 
250 	dev_info(&adapter->pdev->dev, "Tx Ring Dump\n");
251 
252 	/* Transmit Descriptor Formats - DEXT[29] is 0 (Legacy) or 1 (Extended)
253 	 *
254 	 * Legacy Transmit Descriptor
255 	 *   +--------------------------------------------------------------+
256 	 * 0 |         Buffer Address [63:0] (Reserved on Write Back)       |
257 	 *   +--------------------------------------------------------------+
258 	 * 8 | Special  |    CSS     | Status |  CMD    |  CSO   |  Length  |
259 	 *   +--------------------------------------------------------------+
260 	 *   63       48 47        36 35    32 31     24 23    16 15        0
261 	 *
262 	 * Extended Context Descriptor (DTYP=0x0) for TSO or checksum offload
263 	 *   63      48 47    40 39       32 31             16 15    8 7      0
264 	 *   +----------------------------------------------------------------+
265 	 * 0 |  TUCSE  | TUCS0  |   TUCSS   |     IPCSE       | IPCS0 | IPCSS |
266 	 *   +----------------------------------------------------------------+
267 	 * 8 |   MSS   | HDRLEN | RSV | STA | TUCMD | DTYP |      PAYLEN      |
268 	 *   +----------------------------------------------------------------+
269 	 *   63      48 47    40 39 36 35 32 31   24 23  20 19                0
270 	 *
271 	 * Extended Data Descriptor (DTYP=0x1)
272 	 *   +----------------------------------------------------------------+
273 	 * 0 |                     Buffer Address [63:0]                      |
274 	 *   +----------------------------------------------------------------+
275 	 * 8 | VLAN tag |  POPTS  | Rsvd | Status | Command | DTYP |  DTALEN  |
276 	 *   +----------------------------------------------------------------+
277 	 *   63       48 47     40 39  36 35    32 31     24 23  20 19        0
278 	 */
279 	pr_info("Tl[desc]     [address 63:0  ] [SpeCssSCmCsLen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Legacy format\n");
280 	pr_info("Tc[desc]     [Ce CoCsIpceCoS] [MssHlRSCm0Plen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Context format\n");
281 	pr_info("Td[desc]     [address 63:0  ] [VlaPoRSCm1Dlen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Data format\n");
282 	for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
283 		const char *next_desc;
284 		tx_desc = E1000_TX_DESC(*tx_ring, i);
285 		buffer_info = &tx_ring->buffer_info[i];
286 		u0 = (struct my_u0 *)tx_desc;
287 		if (i == tx_ring->next_to_use && i == tx_ring->next_to_clean)
288 			next_desc = " NTC/U";
289 		else if (i == tx_ring->next_to_use)
290 			next_desc = " NTU";
291 		else if (i == tx_ring->next_to_clean)
292 			next_desc = " NTC";
293 		else
294 			next_desc = "";
295 		pr_info("T%c[0x%03X]    %016llX %016llX %016llX %04X  %3X %016llX %p%s\n",
296 			(!(le64_to_cpu(u0->b) & (1 << 29)) ? 'l' :
297 			 ((le64_to_cpu(u0->b) & (1 << 20)) ? 'd' : 'c')),
298 			i,
299 			(unsigned long long)le64_to_cpu(u0->a),
300 			(unsigned long long)le64_to_cpu(u0->b),
301 			(unsigned long long)buffer_info->dma,
302 			buffer_info->length, buffer_info->next_to_watch,
303 			(unsigned long long)buffer_info->time_stamp,
304 			buffer_info->skb, next_desc);
305 
306 		if (netif_msg_pktdata(adapter) && buffer_info->skb)
307 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
308 				       16, 1, buffer_info->skb->data,
309 				       buffer_info->skb->len, true);
310 	}
311 
312 	/* Print Rx Ring Summary */
313 rx_ring_summary:
314 	dev_info(&adapter->pdev->dev, "Rx Ring Summary\n");
315 	pr_info("Queue [NTU] [NTC]\n");
316 	pr_info(" %5d %5X %5X\n",
317 		0, rx_ring->next_to_use, rx_ring->next_to_clean);
318 
319 	/* Print Rx Ring */
320 	if (!netif_msg_rx_status(adapter))
321 		return;
322 
323 	dev_info(&adapter->pdev->dev, "Rx Ring Dump\n");
324 	switch (adapter->rx_ps_pages) {
325 	case 1:
326 	case 2:
327 	case 3:
328 		/* [Extended] Packet Split Receive Descriptor Format
329 		 *
330 		 *    +-----------------------------------------------------+
331 		 *  0 |                Buffer Address 0 [63:0]              |
332 		 *    +-----------------------------------------------------+
333 		 *  8 |                Buffer Address 1 [63:0]              |
334 		 *    +-----------------------------------------------------+
335 		 * 16 |                Buffer Address 2 [63:0]              |
336 		 *    +-----------------------------------------------------+
337 		 * 24 |                Buffer Address 3 [63:0]              |
338 		 *    +-----------------------------------------------------+
339 		 */
340 		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");
341 		/* [Extended] Receive Descriptor (Write-Back) Format
342 		 *
343 		 *   63       48 47    32 31     13 12    8 7    4 3        0
344 		 *   +------------------------------------------------------+
345 		 * 0 | Packet   | IP     |  Rsvd   | MRQ   | Rsvd | MRQ RSS |
346 		 *   | Checksum | Ident  |         | Queue |      |  Type   |
347 		 *   +------------------------------------------------------+
348 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
349 		 *   +------------------------------------------------------+
350 		 *   63       48 47    32 31            20 19               0
351 		 */
352 		pr_info("RWB[desc]      [ck ipid mrqhsh] [vl   l0 ee  es] [ l3  l2  l1 hs] [reserved      ] ---------------- [bi->skb] <-- Ext Rx Write-Back format\n");
353 		for (i = 0; i < rx_ring->count; i++) {
354 			const char *next_desc;
355 			buffer_info = &rx_ring->buffer_info[i];
356 			rx_desc_ps = E1000_RX_DESC_PS(*rx_ring, i);
357 			u1 = (struct my_u1 *)rx_desc_ps;
358 			staterr =
359 			    le32_to_cpu(rx_desc_ps->wb.middle.status_error);
360 
361 			if (i == rx_ring->next_to_use)
362 				next_desc = " NTU";
363 			else if (i == rx_ring->next_to_clean)
364 				next_desc = " NTC";
365 			else
366 				next_desc = "";
367 
368 			if (staterr & E1000_RXD_STAT_DD) {
369 				/* Descriptor Done */
370 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX ---------------- %p%s\n",
371 					"RWB", i,
372 					(unsigned long long)le64_to_cpu(u1->a),
373 					(unsigned long long)le64_to_cpu(u1->b),
374 					(unsigned long long)le64_to_cpu(u1->c),
375 					(unsigned long long)le64_to_cpu(u1->d),
376 					buffer_info->skb, next_desc);
377 			} else {
378 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX %016llX %p%s\n",
379 					"R  ", i,
380 					(unsigned long long)le64_to_cpu(u1->a),
381 					(unsigned long long)le64_to_cpu(u1->b),
382 					(unsigned long long)le64_to_cpu(u1->c),
383 					(unsigned long long)le64_to_cpu(u1->d),
384 					(unsigned long long)buffer_info->dma,
385 					buffer_info->skb, next_desc);
386 
387 				if (netif_msg_pktdata(adapter))
388 					e1000e_dump_ps_pages(adapter,
389 							     buffer_info);
390 			}
391 		}
392 		break;
393 	default:
394 	case 0:
395 		/* Extended Receive Descriptor (Read) Format
396 		 *
397 		 *   +-----------------------------------------------------+
398 		 * 0 |                Buffer Address [63:0]                |
399 		 *   +-----------------------------------------------------+
400 		 * 8 |                      Reserved                       |
401 		 *   +-----------------------------------------------------+
402 		 */
403 		pr_info("R  [desc]      [buf addr 63:0 ] [reserved 63:0 ] [bi->dma       ] [bi->skb] <-- Ext (Read) format\n");
404 		/* Extended Receive Descriptor (Write-Back) Format
405 		 *
406 		 *   63       48 47    32 31    24 23            4 3        0
407 		 *   +------------------------------------------------------+
408 		 *   |     RSS Hash      |        |               |         |
409 		 * 0 +-------------------+  Rsvd  |   Reserved    | MRQ RSS |
410 		 *   | Packet   | IP     |        |               |  Type   |
411 		 *   | Checksum | Ident  |        |               |         |
412 		 *   +------------------------------------------------------+
413 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
414 		 *   +------------------------------------------------------+
415 		 *   63       48 47    32 31            20 19               0
416 		 */
417 		pr_info("RWB[desc]      [cs ipid    mrq] [vt   ln xe  xs] [bi->skb] <-- Ext (Write-Back) format\n");
418 
419 		for (i = 0; i < rx_ring->count; i++) {
420 			const char *next_desc;
421 
422 			buffer_info = &rx_ring->buffer_info[i];
423 			rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
424 			u1 = (struct my_u1 *)rx_desc;
425 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
426 
427 			if (i == rx_ring->next_to_use)
428 				next_desc = " NTU";
429 			else if (i == rx_ring->next_to_clean)
430 				next_desc = " NTC";
431 			else
432 				next_desc = "";
433 
434 			if (staterr & E1000_RXD_STAT_DD) {
435 				/* Descriptor Done */
436 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %p%s\n",
437 					"RWB", i,
438 					(unsigned long long)le64_to_cpu(u1->a),
439 					(unsigned long long)le64_to_cpu(u1->b),
440 					buffer_info->skb, next_desc);
441 			} else {
442 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %p%s\n",
443 					"R  ", i,
444 					(unsigned long long)le64_to_cpu(u1->a),
445 					(unsigned long long)le64_to_cpu(u1->b),
446 					(unsigned long long)buffer_info->dma,
447 					buffer_info->skb, next_desc);
448 
449 				if (netif_msg_pktdata(adapter) &&
450 				    buffer_info->skb)
451 					print_hex_dump(KERN_INFO, "",
452 						       DUMP_PREFIX_ADDRESS, 16,
453 						       1,
454 						       buffer_info->skb->data,
455 						       adapter->rx_buffer_len,
456 						       true);
457 			}
458 		}
459 	}
460 }
461 
462 /**
463  * e1000_desc_unused - calculate if we have unused descriptors
464  **/
465 static int e1000_desc_unused(struct e1000_ring *ring)
466 {
467 	if (ring->next_to_clean > ring->next_to_use)
468 		return ring->next_to_clean - ring->next_to_use - 1;
469 
470 	return ring->count + ring->next_to_clean - ring->next_to_use - 1;
471 }
472 
473 /**
474  * e1000e_systim_to_hwtstamp - convert system time value to hw time stamp
475  * @adapter: board private structure
476  * @hwtstamps: time stamp structure to update
477  * @systim: unsigned 64bit system time value.
478  *
479  * Convert the system time value stored in the RX/TXSTMP registers into a
480  * hwtstamp which can be used by the upper level time stamping functions.
481  *
482  * The 'systim_lock' spinlock is used to protect the consistency of the
483  * system time value. This is needed because reading the 64 bit time
484  * value involves reading two 32 bit registers. The first read latches the
485  * value.
486  **/
487 static void e1000e_systim_to_hwtstamp(struct e1000_adapter *adapter,
488 				      struct skb_shared_hwtstamps *hwtstamps,
489 				      u64 systim)
490 {
491 	u64 ns;
492 	unsigned long flags;
493 
494 	spin_lock_irqsave(&adapter->systim_lock, flags);
495 	ns = timecounter_cyc2time(&adapter->tc, systim);
496 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
497 
498 	memset(hwtstamps, 0, sizeof(*hwtstamps));
499 	hwtstamps->hwtstamp = ns_to_ktime(ns);
500 }
501 
502 /**
503  * e1000e_rx_hwtstamp - utility function which checks for Rx time stamp
504  * @adapter: board private structure
505  * @status: descriptor extended error and status field
506  * @skb: particular skb to include time stamp
507  *
508  * If the time stamp is valid, convert it into the timecounter ns value
509  * and store that result into the shhwtstamps structure which is passed
510  * up the network stack.
511  **/
512 static void e1000e_rx_hwtstamp(struct e1000_adapter *adapter, u32 status,
513 			       struct sk_buff *skb)
514 {
515 	struct e1000_hw *hw = &adapter->hw;
516 	u64 rxstmp;
517 
518 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP) ||
519 	    !(status & E1000_RXDEXT_STATERR_TST) ||
520 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID))
521 		return;
522 
523 	/* The Rx time stamp registers contain the time stamp.  No other
524 	 * received packet will be time stamped until the Rx time stamp
525 	 * registers are read.  Because only one packet can be time stamped
526 	 * at a time, the register values must belong to this packet and
527 	 * therefore none of the other additional attributes need to be
528 	 * compared.
529 	 */
530 	rxstmp = (u64)er32(RXSTMPL);
531 	rxstmp |= (u64)er32(RXSTMPH) << 32;
532 	e1000e_systim_to_hwtstamp(adapter, skb_hwtstamps(skb), rxstmp);
533 
534 	adapter->flags2 &= ~FLAG2_CHECK_RX_HWTSTAMP;
535 }
536 
537 /**
538  * e1000_receive_skb - helper function to handle Rx indications
539  * @adapter: board private structure
540  * @staterr: descriptor extended error and status field as written by hardware
541  * @vlan: descriptor vlan field as written by hardware (no le/be conversion)
542  * @skb: pointer to sk_buff to be indicated to stack
543  **/
544 static void e1000_receive_skb(struct e1000_adapter *adapter,
545 			      struct net_device *netdev, struct sk_buff *skb,
546 			      u32 staterr, __le16 vlan)
547 {
548 	u16 tag = le16_to_cpu(vlan);
549 
550 	e1000e_rx_hwtstamp(adapter, staterr, skb);
551 
552 	skb->protocol = eth_type_trans(skb, netdev);
553 
554 	if (staterr & E1000_RXD_STAT_VP)
555 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), tag);
556 
557 	napi_gro_receive(&adapter->napi, skb);
558 }
559 
560 /**
561  * e1000_rx_checksum - Receive Checksum Offload
562  * @adapter: board private structure
563  * @status_err: receive descriptor status and error fields
564  * @csum: receive descriptor csum field
565  * @sk_buff: socket buffer with received data
566  **/
567 static void e1000_rx_checksum(struct e1000_adapter *adapter, u32 status_err,
568 			      struct sk_buff *skb)
569 {
570 	u16 status = (u16)status_err;
571 	u8 errors = (u8)(status_err >> 24);
572 
573 	skb_checksum_none_assert(skb);
574 
575 	/* Rx checksum disabled */
576 	if (!(adapter->netdev->features & NETIF_F_RXCSUM))
577 		return;
578 
579 	/* Ignore Checksum bit is set */
580 	if (status & E1000_RXD_STAT_IXSM)
581 		return;
582 
583 	/* TCP/UDP checksum error bit or IP checksum error bit is set */
584 	if (errors & (E1000_RXD_ERR_TCPE | E1000_RXD_ERR_IPE)) {
585 		/* let the stack verify checksum errors */
586 		adapter->hw_csum_err++;
587 		return;
588 	}
589 
590 	/* TCP/UDP Checksum has not been calculated */
591 	if (!(status & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)))
592 		return;
593 
594 	/* It must be a TCP or UDP packet with a valid checksum */
595 	skb->ip_summed = CHECKSUM_UNNECESSARY;
596 	adapter->hw_csum_good++;
597 }
598 
599 static void e1000e_update_rdt_wa(struct e1000_ring *rx_ring, unsigned int i)
600 {
601 	struct e1000_adapter *adapter = rx_ring->adapter;
602 	struct e1000_hw *hw = &adapter->hw;
603 	s32 ret_val = __ew32_prepare(hw);
604 
605 	writel(i, rx_ring->tail);
606 
607 	if (unlikely(!ret_val && (i != readl(rx_ring->tail)))) {
608 		u32 rctl = er32(RCTL);
609 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
610 		e_err("ME firmware caused invalid RDT - resetting\n");
611 		schedule_work(&adapter->reset_task);
612 	}
613 }
614 
615 static void e1000e_update_tdt_wa(struct e1000_ring *tx_ring, unsigned int i)
616 {
617 	struct e1000_adapter *adapter = tx_ring->adapter;
618 	struct e1000_hw *hw = &adapter->hw;
619 	s32 ret_val = __ew32_prepare(hw);
620 
621 	writel(i, tx_ring->tail);
622 
623 	if (unlikely(!ret_val && (i != readl(tx_ring->tail)))) {
624 		u32 tctl = er32(TCTL);
625 		ew32(TCTL, tctl & ~E1000_TCTL_EN);
626 		e_err("ME firmware caused invalid TDT - resetting\n");
627 		schedule_work(&adapter->reset_task);
628 	}
629 }
630 
631 /**
632  * e1000_alloc_rx_buffers - Replace used receive buffers
633  * @rx_ring: Rx descriptor ring
634  **/
635 static void e1000_alloc_rx_buffers(struct e1000_ring *rx_ring,
636 				   int cleaned_count, gfp_t gfp)
637 {
638 	struct e1000_adapter *adapter = rx_ring->adapter;
639 	struct net_device *netdev = adapter->netdev;
640 	struct pci_dev *pdev = adapter->pdev;
641 	union e1000_rx_desc_extended *rx_desc;
642 	struct e1000_buffer *buffer_info;
643 	struct sk_buff *skb;
644 	unsigned int i;
645 	unsigned int bufsz = adapter->rx_buffer_len;
646 
647 	i = rx_ring->next_to_use;
648 	buffer_info = &rx_ring->buffer_info[i];
649 
650 	while (cleaned_count--) {
651 		skb = buffer_info->skb;
652 		if (skb) {
653 			skb_trim(skb, 0);
654 			goto map_skb;
655 		}
656 
657 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
658 		if (!skb) {
659 			/* Better luck next round */
660 			adapter->alloc_rx_buff_failed++;
661 			break;
662 		}
663 
664 		buffer_info->skb = skb;
665 map_skb:
666 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
667 						  adapter->rx_buffer_len,
668 						  DMA_FROM_DEVICE);
669 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
670 			dev_err(&pdev->dev, "Rx DMA map failed\n");
671 			adapter->rx_dma_failed++;
672 			break;
673 		}
674 
675 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
676 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
677 
678 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
679 			/* Force memory writes to complete before letting h/w
680 			 * know there are new descriptors to fetch.  (Only
681 			 * applicable for weak-ordered memory model archs,
682 			 * such as IA-64).
683 			 */
684 			wmb();
685 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
686 				e1000e_update_rdt_wa(rx_ring, i);
687 			else
688 				writel(i, rx_ring->tail);
689 		}
690 		i++;
691 		if (i == rx_ring->count)
692 			i = 0;
693 		buffer_info = &rx_ring->buffer_info[i];
694 	}
695 
696 	rx_ring->next_to_use = i;
697 }
698 
699 /**
700  * e1000_alloc_rx_buffers_ps - Replace used receive buffers; packet split
701  * @rx_ring: Rx descriptor ring
702  **/
703 static void e1000_alloc_rx_buffers_ps(struct e1000_ring *rx_ring,
704 				      int cleaned_count, gfp_t gfp)
705 {
706 	struct e1000_adapter *adapter = rx_ring->adapter;
707 	struct net_device *netdev = adapter->netdev;
708 	struct pci_dev *pdev = adapter->pdev;
709 	union e1000_rx_desc_packet_split *rx_desc;
710 	struct e1000_buffer *buffer_info;
711 	struct e1000_ps_page *ps_page;
712 	struct sk_buff *skb;
713 	unsigned int i, j;
714 
715 	i = rx_ring->next_to_use;
716 	buffer_info = &rx_ring->buffer_info[i];
717 
718 	while (cleaned_count--) {
719 		rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
720 
721 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
722 			ps_page = &buffer_info->ps_pages[j];
723 			if (j >= adapter->rx_ps_pages) {
724 				/* all unused desc entries get hw null ptr */
725 				rx_desc->read.buffer_addr[j + 1] =
726 				    ~cpu_to_le64(0);
727 				continue;
728 			}
729 			if (!ps_page->page) {
730 				ps_page->page = alloc_page(gfp);
731 				if (!ps_page->page) {
732 					adapter->alloc_rx_buff_failed++;
733 					goto no_buffers;
734 				}
735 				ps_page->dma = dma_map_page(&pdev->dev,
736 							    ps_page->page,
737 							    0, PAGE_SIZE,
738 							    DMA_FROM_DEVICE);
739 				if (dma_mapping_error(&pdev->dev,
740 						      ps_page->dma)) {
741 					dev_err(&adapter->pdev->dev,
742 						"Rx DMA page map failed\n");
743 					adapter->rx_dma_failed++;
744 					goto no_buffers;
745 				}
746 			}
747 			/* Refresh the desc even if buffer_addrs
748 			 * didn't change because each write-back
749 			 * erases this info.
750 			 */
751 			rx_desc->read.buffer_addr[j + 1] =
752 			    cpu_to_le64(ps_page->dma);
753 		}
754 
755 		skb = __netdev_alloc_skb_ip_align(netdev, adapter->rx_ps_bsize0,
756 						  gfp);
757 
758 		if (!skb) {
759 			adapter->alloc_rx_buff_failed++;
760 			break;
761 		}
762 
763 		buffer_info->skb = skb;
764 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
765 						  adapter->rx_ps_bsize0,
766 						  DMA_FROM_DEVICE);
767 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
768 			dev_err(&pdev->dev, "Rx DMA map failed\n");
769 			adapter->rx_dma_failed++;
770 			/* cleanup skb */
771 			dev_kfree_skb_any(skb);
772 			buffer_info->skb = NULL;
773 			break;
774 		}
775 
776 		rx_desc->read.buffer_addr[0] = cpu_to_le64(buffer_info->dma);
777 
778 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
779 			/* Force memory writes to complete before letting h/w
780 			 * know there are new descriptors to fetch.  (Only
781 			 * applicable for weak-ordered memory model archs,
782 			 * such as IA-64).
783 			 */
784 			wmb();
785 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
786 				e1000e_update_rdt_wa(rx_ring, i << 1);
787 			else
788 				writel(i << 1, rx_ring->tail);
789 		}
790 
791 		i++;
792 		if (i == rx_ring->count)
793 			i = 0;
794 		buffer_info = &rx_ring->buffer_info[i];
795 	}
796 
797 no_buffers:
798 	rx_ring->next_to_use = i;
799 }
800 
801 /**
802  * e1000_alloc_jumbo_rx_buffers - Replace used jumbo receive buffers
803  * @rx_ring: Rx descriptor ring
804  * @cleaned_count: number of buffers to allocate this pass
805  **/
806 
807 static void e1000_alloc_jumbo_rx_buffers(struct e1000_ring *rx_ring,
808 					 int cleaned_count, gfp_t gfp)
809 {
810 	struct e1000_adapter *adapter = rx_ring->adapter;
811 	struct net_device *netdev = adapter->netdev;
812 	struct pci_dev *pdev = adapter->pdev;
813 	union e1000_rx_desc_extended *rx_desc;
814 	struct e1000_buffer *buffer_info;
815 	struct sk_buff *skb;
816 	unsigned int i;
817 	unsigned int bufsz = 256 - 16;	/* for skb_reserve */
818 
819 	i = rx_ring->next_to_use;
820 	buffer_info = &rx_ring->buffer_info[i];
821 
822 	while (cleaned_count--) {
823 		skb = buffer_info->skb;
824 		if (skb) {
825 			skb_trim(skb, 0);
826 			goto check_page;
827 		}
828 
829 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
830 		if (unlikely(!skb)) {
831 			/* Better luck next round */
832 			adapter->alloc_rx_buff_failed++;
833 			break;
834 		}
835 
836 		buffer_info->skb = skb;
837 check_page:
838 		/* allocate a new page if necessary */
839 		if (!buffer_info->page) {
840 			buffer_info->page = alloc_page(gfp);
841 			if (unlikely(!buffer_info->page)) {
842 				adapter->alloc_rx_buff_failed++;
843 				break;
844 			}
845 		}
846 
847 		if (!buffer_info->dma) {
848 			buffer_info->dma = dma_map_page(&pdev->dev,
849 							buffer_info->page, 0,
850 							PAGE_SIZE,
851 							DMA_FROM_DEVICE);
852 			if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
853 				adapter->alloc_rx_buff_failed++;
854 				break;
855 			}
856 		}
857 
858 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
859 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
860 
861 		if (unlikely(++i == rx_ring->count))
862 			i = 0;
863 		buffer_info = &rx_ring->buffer_info[i];
864 	}
865 
866 	if (likely(rx_ring->next_to_use != i)) {
867 		rx_ring->next_to_use = i;
868 		if (unlikely(i-- == 0))
869 			i = (rx_ring->count - 1);
870 
871 		/* Force memory writes to complete before letting h/w
872 		 * know there are new descriptors to fetch.  (Only
873 		 * applicable for weak-ordered memory model archs,
874 		 * such as IA-64).
875 		 */
876 		wmb();
877 		if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
878 			e1000e_update_rdt_wa(rx_ring, i);
879 		else
880 			writel(i, rx_ring->tail);
881 	}
882 }
883 
884 static inline void e1000_rx_hash(struct net_device *netdev, __le32 rss,
885 				 struct sk_buff *skb)
886 {
887 	if (netdev->features & NETIF_F_RXHASH)
888 		skb->rxhash = le32_to_cpu(rss);
889 }
890 
891 /**
892  * e1000_clean_rx_irq - Send received data up the network stack
893  * @rx_ring: Rx descriptor ring
894  *
895  * the return value indicates whether actual cleaning was done, there
896  * is no guarantee that everything was cleaned
897  **/
898 static bool e1000_clean_rx_irq(struct e1000_ring *rx_ring, int *work_done,
899 			       int work_to_do)
900 {
901 	struct e1000_adapter *adapter = rx_ring->adapter;
902 	struct net_device *netdev = adapter->netdev;
903 	struct pci_dev *pdev = adapter->pdev;
904 	struct e1000_hw *hw = &adapter->hw;
905 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
906 	struct e1000_buffer *buffer_info, *next_buffer;
907 	u32 length, staterr;
908 	unsigned int i;
909 	int cleaned_count = 0;
910 	bool cleaned = false;
911 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
912 
913 	i = rx_ring->next_to_clean;
914 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
915 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
916 	buffer_info = &rx_ring->buffer_info[i];
917 
918 	while (staterr & E1000_RXD_STAT_DD) {
919 		struct sk_buff *skb;
920 
921 		if (*work_done >= work_to_do)
922 			break;
923 		(*work_done)++;
924 		rmb();	/* read descriptor and rx_buffer_info after status DD */
925 
926 		skb = buffer_info->skb;
927 		buffer_info->skb = NULL;
928 
929 		prefetch(skb->data - NET_IP_ALIGN);
930 
931 		i++;
932 		if (i == rx_ring->count)
933 			i = 0;
934 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
935 		prefetch(next_rxd);
936 
937 		next_buffer = &rx_ring->buffer_info[i];
938 
939 		cleaned = true;
940 		cleaned_count++;
941 		dma_unmap_single(&pdev->dev, buffer_info->dma,
942 				 adapter->rx_buffer_len, DMA_FROM_DEVICE);
943 		buffer_info->dma = 0;
944 
945 		length = le16_to_cpu(rx_desc->wb.upper.length);
946 
947 		/* !EOP means multiple descriptors were used to store a single
948 		 * packet, if that's the case we need to toss it.  In fact, we
949 		 * need to toss every packet with the EOP bit clear and the
950 		 * next frame that _does_ have the EOP bit set, as it is by
951 		 * definition only a frame fragment
952 		 */
953 		if (unlikely(!(staterr & E1000_RXD_STAT_EOP)))
954 			adapter->flags2 |= FLAG2_IS_DISCARDING;
955 
956 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
957 			/* All receives must fit into a single buffer */
958 			e_dbg("Receive packet consumed multiple buffers\n");
959 			/* recycle */
960 			buffer_info->skb = skb;
961 			if (staterr & E1000_RXD_STAT_EOP)
962 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
963 			goto next_desc;
964 		}
965 
966 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
967 			     !(netdev->features & NETIF_F_RXALL))) {
968 			/* recycle */
969 			buffer_info->skb = skb;
970 			goto next_desc;
971 		}
972 
973 		/* adjust length to remove Ethernet CRC */
974 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
975 			/* If configured to store CRC, don't subtract FCS,
976 			 * but keep the FCS bytes out of the total_rx_bytes
977 			 * counter
978 			 */
979 			if (netdev->features & NETIF_F_RXFCS)
980 				total_rx_bytes -= 4;
981 			else
982 				length -= 4;
983 		}
984 
985 		total_rx_bytes += length;
986 		total_rx_packets++;
987 
988 		/* code added for copybreak, this should improve
989 		 * performance for small packets with large amounts
990 		 * of reassembly being done in the stack
991 		 */
992 		if (length < copybreak) {
993 			struct sk_buff *new_skb =
994 			    netdev_alloc_skb_ip_align(netdev, length);
995 			if (new_skb) {
996 				skb_copy_to_linear_data_offset(new_skb,
997 							       -NET_IP_ALIGN,
998 							       (skb->data -
999 								NET_IP_ALIGN),
1000 							       (length +
1001 								NET_IP_ALIGN));
1002 				/* save the skb in buffer_info as good */
1003 				buffer_info->skb = skb;
1004 				skb = new_skb;
1005 			}
1006 			/* else just continue with the old one */
1007 		}
1008 		/* end copybreak code */
1009 		skb_put(skb, length);
1010 
1011 		/* Receive Checksum Offload */
1012 		e1000_rx_checksum(adapter, staterr, skb);
1013 
1014 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1015 
1016 		e1000_receive_skb(adapter, netdev, skb, staterr,
1017 				  rx_desc->wb.upper.vlan);
1018 
1019 next_desc:
1020 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1021 
1022 		/* return some buffers to hardware, one at a time is too slow */
1023 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1024 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1025 					      GFP_ATOMIC);
1026 			cleaned_count = 0;
1027 		}
1028 
1029 		/* use prefetched values */
1030 		rx_desc = next_rxd;
1031 		buffer_info = next_buffer;
1032 
1033 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1034 	}
1035 	rx_ring->next_to_clean = i;
1036 
1037 	cleaned_count = e1000_desc_unused(rx_ring);
1038 	if (cleaned_count)
1039 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1040 
1041 	adapter->total_rx_bytes += total_rx_bytes;
1042 	adapter->total_rx_packets += total_rx_packets;
1043 	return cleaned;
1044 }
1045 
1046 static void e1000_put_txbuf(struct e1000_ring *tx_ring,
1047 			    struct e1000_buffer *buffer_info)
1048 {
1049 	struct e1000_adapter *adapter = tx_ring->adapter;
1050 
1051 	if (buffer_info->dma) {
1052 		if (buffer_info->mapped_as_page)
1053 			dma_unmap_page(&adapter->pdev->dev, buffer_info->dma,
1054 				       buffer_info->length, DMA_TO_DEVICE);
1055 		else
1056 			dma_unmap_single(&adapter->pdev->dev, buffer_info->dma,
1057 					 buffer_info->length, DMA_TO_DEVICE);
1058 		buffer_info->dma = 0;
1059 	}
1060 	if (buffer_info->skb) {
1061 		dev_kfree_skb_any(buffer_info->skb);
1062 		buffer_info->skb = NULL;
1063 	}
1064 	buffer_info->time_stamp = 0;
1065 }
1066 
1067 static void e1000_print_hw_hang(struct work_struct *work)
1068 {
1069 	struct e1000_adapter *adapter = container_of(work,
1070 						     struct e1000_adapter,
1071 						     print_hang_task);
1072 	struct net_device *netdev = adapter->netdev;
1073 	struct e1000_ring *tx_ring = adapter->tx_ring;
1074 	unsigned int i = tx_ring->next_to_clean;
1075 	unsigned int eop = tx_ring->buffer_info[i].next_to_watch;
1076 	struct e1000_tx_desc *eop_desc = E1000_TX_DESC(*tx_ring, eop);
1077 	struct e1000_hw *hw = &adapter->hw;
1078 	u16 phy_status, phy_1000t_status, phy_ext_status;
1079 	u16 pci_status;
1080 
1081 	if (test_bit(__E1000_DOWN, &adapter->state))
1082 		return;
1083 
1084 	if (!adapter->tx_hang_recheck && (adapter->flags2 & FLAG2_DMA_BURST)) {
1085 		/* May be block on write-back, flush and detect again
1086 		 * flush pending descriptor writebacks to memory
1087 		 */
1088 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1089 		/* execute the writes immediately */
1090 		e1e_flush();
1091 		/* Due to rare timing issues, write to TIDV again to ensure
1092 		 * the write is successful
1093 		 */
1094 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1095 		/* execute the writes immediately */
1096 		e1e_flush();
1097 		adapter->tx_hang_recheck = true;
1098 		return;
1099 	}
1100 	/* Real hang detected */
1101 	adapter->tx_hang_recheck = false;
1102 	netif_stop_queue(netdev);
1103 
1104 	e1e_rphy(hw, MII_BMSR, &phy_status);
1105 	e1e_rphy(hw, MII_STAT1000, &phy_1000t_status);
1106 	e1e_rphy(hw, MII_ESTATUS, &phy_ext_status);
1107 
1108 	pci_read_config_word(adapter->pdev, PCI_STATUS, &pci_status);
1109 
1110 	/* detected Hardware unit hang */
1111 	e_err("Detected Hardware Unit Hang:\n"
1112 	      "  TDH                  <%x>\n"
1113 	      "  TDT                  <%x>\n"
1114 	      "  next_to_use          <%x>\n"
1115 	      "  next_to_clean        <%x>\n"
1116 	      "buffer_info[next_to_clean]:\n"
1117 	      "  time_stamp           <%lx>\n"
1118 	      "  next_to_watch        <%x>\n"
1119 	      "  jiffies              <%lx>\n"
1120 	      "  next_to_watch.status <%x>\n"
1121 	      "MAC Status             <%x>\n"
1122 	      "PHY Status             <%x>\n"
1123 	      "PHY 1000BASE-T Status  <%x>\n"
1124 	      "PHY Extended Status    <%x>\n"
1125 	      "PCI Status             <%x>\n",
1126 	      readl(tx_ring->head), readl(tx_ring->tail), tx_ring->next_to_use,
1127 	      tx_ring->next_to_clean, tx_ring->buffer_info[eop].time_stamp,
1128 	      eop, jiffies, eop_desc->upper.fields.status, er32(STATUS),
1129 	      phy_status, phy_1000t_status, phy_ext_status, pci_status);
1130 
1131 	/* Suggest workaround for known h/w issue */
1132 	if ((hw->mac.type == e1000_pchlan) && (er32(CTRL) & E1000_CTRL_TFCE))
1133 		e_err("Try turning off Tx pause (flow control) via ethtool\n");
1134 }
1135 
1136 /**
1137  * e1000e_tx_hwtstamp_work - check for Tx time stamp
1138  * @work: pointer to work struct
1139  *
1140  * This work function polls the TSYNCTXCTL valid bit to determine when a
1141  * timestamp has been taken for the current stored skb.  The timestamp must
1142  * be for this skb because only one such packet is allowed in the queue.
1143  */
1144 static void e1000e_tx_hwtstamp_work(struct work_struct *work)
1145 {
1146 	struct e1000_adapter *adapter = container_of(work, struct e1000_adapter,
1147 						     tx_hwtstamp_work);
1148 	struct e1000_hw *hw = &adapter->hw;
1149 
1150 	if (!adapter->tx_hwtstamp_skb)
1151 		return;
1152 
1153 	if (er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID) {
1154 		struct skb_shared_hwtstamps shhwtstamps;
1155 		u64 txstmp;
1156 
1157 		txstmp = er32(TXSTMPL);
1158 		txstmp |= (u64)er32(TXSTMPH) << 32;
1159 
1160 		e1000e_systim_to_hwtstamp(adapter, &shhwtstamps, txstmp);
1161 
1162 		skb_tstamp_tx(adapter->tx_hwtstamp_skb, &shhwtstamps);
1163 		dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
1164 		adapter->tx_hwtstamp_skb = NULL;
1165 	} else {
1166 		/* reschedule to check later */
1167 		schedule_work(&adapter->tx_hwtstamp_work);
1168 	}
1169 }
1170 
1171 /**
1172  * e1000_clean_tx_irq - Reclaim resources after transmit completes
1173  * @tx_ring: Tx descriptor ring
1174  *
1175  * the return value indicates whether actual cleaning was done, there
1176  * is no guarantee that everything was cleaned
1177  **/
1178 static bool e1000_clean_tx_irq(struct e1000_ring *tx_ring)
1179 {
1180 	struct e1000_adapter *adapter = tx_ring->adapter;
1181 	struct net_device *netdev = adapter->netdev;
1182 	struct e1000_hw *hw = &adapter->hw;
1183 	struct e1000_tx_desc *tx_desc, *eop_desc;
1184 	struct e1000_buffer *buffer_info;
1185 	unsigned int i, eop;
1186 	unsigned int count = 0;
1187 	unsigned int total_tx_bytes = 0, total_tx_packets = 0;
1188 	unsigned int bytes_compl = 0, pkts_compl = 0;
1189 
1190 	i = tx_ring->next_to_clean;
1191 	eop = tx_ring->buffer_info[i].next_to_watch;
1192 	eop_desc = E1000_TX_DESC(*tx_ring, eop);
1193 
1194 	while ((eop_desc->upper.data & cpu_to_le32(E1000_TXD_STAT_DD)) &&
1195 	       (count < tx_ring->count)) {
1196 		bool cleaned = false;
1197 		rmb();		/* read buffer_info after eop_desc */
1198 		for (; !cleaned; count++) {
1199 			tx_desc = E1000_TX_DESC(*tx_ring, i);
1200 			buffer_info = &tx_ring->buffer_info[i];
1201 			cleaned = (i == eop);
1202 
1203 			if (cleaned) {
1204 				total_tx_packets += buffer_info->segs;
1205 				total_tx_bytes += buffer_info->bytecount;
1206 				if (buffer_info->skb) {
1207 					bytes_compl += buffer_info->skb->len;
1208 					pkts_compl++;
1209 				}
1210 			}
1211 
1212 			e1000_put_txbuf(tx_ring, buffer_info);
1213 			tx_desc->upper.data = 0;
1214 
1215 			i++;
1216 			if (i == tx_ring->count)
1217 				i = 0;
1218 		}
1219 
1220 		if (i == tx_ring->next_to_use)
1221 			break;
1222 		eop = tx_ring->buffer_info[i].next_to_watch;
1223 		eop_desc = E1000_TX_DESC(*tx_ring, eop);
1224 	}
1225 
1226 	tx_ring->next_to_clean = i;
1227 
1228 	netdev_completed_queue(netdev, pkts_compl, bytes_compl);
1229 
1230 #define TX_WAKE_THRESHOLD 32
1231 	if (count && netif_carrier_ok(netdev) &&
1232 	    e1000_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD) {
1233 		/* Make sure that anybody stopping the queue after this
1234 		 * sees the new next_to_clean.
1235 		 */
1236 		smp_mb();
1237 
1238 		if (netif_queue_stopped(netdev) &&
1239 		    !(test_bit(__E1000_DOWN, &adapter->state))) {
1240 			netif_wake_queue(netdev);
1241 			++adapter->restart_queue;
1242 		}
1243 	}
1244 
1245 	if (adapter->detect_tx_hung) {
1246 		/* Detect a transmit hang in hardware, this serializes the
1247 		 * check with the clearing of time_stamp and movement of i
1248 		 */
1249 		adapter->detect_tx_hung = false;
1250 		if (tx_ring->buffer_info[i].time_stamp &&
1251 		    time_after(jiffies, tx_ring->buffer_info[i].time_stamp
1252 			       + (adapter->tx_timeout_factor * HZ)) &&
1253 		    !(er32(STATUS) & E1000_STATUS_TXOFF))
1254 			schedule_work(&adapter->print_hang_task);
1255 		else
1256 			adapter->tx_hang_recheck = false;
1257 	}
1258 	adapter->total_tx_bytes += total_tx_bytes;
1259 	adapter->total_tx_packets += total_tx_packets;
1260 	return count < tx_ring->count;
1261 }
1262 
1263 /**
1264  * e1000_clean_rx_irq_ps - Send received data up the network stack; packet split
1265  * @rx_ring: Rx descriptor ring
1266  *
1267  * the return value indicates whether actual cleaning was done, there
1268  * is no guarantee that everything was cleaned
1269  **/
1270 static bool e1000_clean_rx_irq_ps(struct e1000_ring *rx_ring, int *work_done,
1271 				  int work_to_do)
1272 {
1273 	struct e1000_adapter *adapter = rx_ring->adapter;
1274 	struct e1000_hw *hw = &adapter->hw;
1275 	union e1000_rx_desc_packet_split *rx_desc, *next_rxd;
1276 	struct net_device *netdev = adapter->netdev;
1277 	struct pci_dev *pdev = adapter->pdev;
1278 	struct e1000_buffer *buffer_info, *next_buffer;
1279 	struct e1000_ps_page *ps_page;
1280 	struct sk_buff *skb;
1281 	unsigned int i, j;
1282 	u32 length, staterr;
1283 	int cleaned_count = 0;
1284 	bool cleaned = false;
1285 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1286 
1287 	i = rx_ring->next_to_clean;
1288 	rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
1289 	staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1290 	buffer_info = &rx_ring->buffer_info[i];
1291 
1292 	while (staterr & E1000_RXD_STAT_DD) {
1293 		if (*work_done >= work_to_do)
1294 			break;
1295 		(*work_done)++;
1296 		skb = buffer_info->skb;
1297 		rmb();	/* read descriptor and rx_buffer_info after status DD */
1298 
1299 		/* in the packet split case this is header only */
1300 		prefetch(skb->data - NET_IP_ALIGN);
1301 
1302 		i++;
1303 		if (i == rx_ring->count)
1304 			i = 0;
1305 		next_rxd = E1000_RX_DESC_PS(*rx_ring, i);
1306 		prefetch(next_rxd);
1307 
1308 		next_buffer = &rx_ring->buffer_info[i];
1309 
1310 		cleaned = true;
1311 		cleaned_count++;
1312 		dma_unmap_single(&pdev->dev, buffer_info->dma,
1313 				 adapter->rx_ps_bsize0, DMA_FROM_DEVICE);
1314 		buffer_info->dma = 0;
1315 
1316 		/* see !EOP comment in other Rx routine */
1317 		if (!(staterr & E1000_RXD_STAT_EOP))
1318 			adapter->flags2 |= FLAG2_IS_DISCARDING;
1319 
1320 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
1321 			e_dbg("Packet Split buffers didn't pick up the full packet\n");
1322 			dev_kfree_skb_irq(skb);
1323 			if (staterr & E1000_RXD_STAT_EOP)
1324 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1325 			goto next_desc;
1326 		}
1327 
1328 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1329 			     !(netdev->features & NETIF_F_RXALL))) {
1330 			dev_kfree_skb_irq(skb);
1331 			goto next_desc;
1332 		}
1333 
1334 		length = le16_to_cpu(rx_desc->wb.middle.length0);
1335 
1336 		if (!length) {
1337 			e_dbg("Last part of the packet spanning multiple descriptors\n");
1338 			dev_kfree_skb_irq(skb);
1339 			goto next_desc;
1340 		}
1341 
1342 		/* Good Receive */
1343 		skb_put(skb, length);
1344 
1345 		{
1346 			/* this looks ugly, but it seems compiler issues make
1347 			 * it more efficient than reusing j
1348 			 */
1349 			int l1 = le16_to_cpu(rx_desc->wb.upper.length[0]);
1350 
1351 			/* page alloc/put takes too long and effects small
1352 			 * packet throughput, so unsplit small packets and
1353 			 * save the alloc/put only valid in softirq (napi)
1354 			 * context to call kmap_*
1355 			 */
1356 			if (l1 && (l1 <= copybreak) &&
1357 			    ((length + l1) <= adapter->rx_ps_bsize0)) {
1358 				u8 *vaddr;
1359 
1360 				ps_page = &buffer_info->ps_pages[0];
1361 
1362 				/* there is no documentation about how to call
1363 				 * kmap_atomic, so we can't hold the mapping
1364 				 * very long
1365 				 */
1366 				dma_sync_single_for_cpu(&pdev->dev,
1367 							ps_page->dma,
1368 							PAGE_SIZE,
1369 							DMA_FROM_DEVICE);
1370 				vaddr = kmap_atomic(ps_page->page);
1371 				memcpy(skb_tail_pointer(skb), vaddr, l1);
1372 				kunmap_atomic(vaddr);
1373 				dma_sync_single_for_device(&pdev->dev,
1374 							   ps_page->dma,
1375 							   PAGE_SIZE,
1376 							   DMA_FROM_DEVICE);
1377 
1378 				/* remove the CRC */
1379 				if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1380 					if (!(netdev->features & NETIF_F_RXFCS))
1381 						l1 -= 4;
1382 				}
1383 
1384 				skb_put(skb, l1);
1385 				goto copydone;
1386 			}	/* if */
1387 		}
1388 
1389 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1390 			length = le16_to_cpu(rx_desc->wb.upper.length[j]);
1391 			if (!length)
1392 				break;
1393 
1394 			ps_page = &buffer_info->ps_pages[j];
1395 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1396 				       DMA_FROM_DEVICE);
1397 			ps_page->dma = 0;
1398 			skb_fill_page_desc(skb, j, ps_page->page, 0, length);
1399 			ps_page->page = NULL;
1400 			skb->len += length;
1401 			skb->data_len += length;
1402 			skb->truesize += PAGE_SIZE;
1403 		}
1404 
1405 		/* strip the ethernet crc, problem is we're using pages now so
1406 		 * this whole operation can get a little cpu intensive
1407 		 */
1408 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1409 			if (!(netdev->features & NETIF_F_RXFCS))
1410 				pskb_trim(skb, skb->len - 4);
1411 		}
1412 
1413 copydone:
1414 		total_rx_bytes += skb->len;
1415 		total_rx_packets++;
1416 
1417 		e1000_rx_checksum(adapter, staterr, skb);
1418 
1419 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1420 
1421 		if (rx_desc->wb.upper.header_status &
1422 		    cpu_to_le16(E1000_RXDPS_HDRSTAT_HDRSP))
1423 			adapter->rx_hdr_split++;
1424 
1425 		e1000_receive_skb(adapter, netdev, skb, staterr,
1426 				  rx_desc->wb.middle.vlan);
1427 
1428 next_desc:
1429 		rx_desc->wb.middle.status_error &= cpu_to_le32(~0xFF);
1430 		buffer_info->skb = NULL;
1431 
1432 		/* return some buffers to hardware, one at a time is too slow */
1433 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1434 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1435 					      GFP_ATOMIC);
1436 			cleaned_count = 0;
1437 		}
1438 
1439 		/* use prefetched values */
1440 		rx_desc = next_rxd;
1441 		buffer_info = next_buffer;
1442 
1443 		staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1444 	}
1445 	rx_ring->next_to_clean = i;
1446 
1447 	cleaned_count = e1000_desc_unused(rx_ring);
1448 	if (cleaned_count)
1449 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1450 
1451 	adapter->total_rx_bytes += total_rx_bytes;
1452 	adapter->total_rx_packets += total_rx_packets;
1453 	return cleaned;
1454 }
1455 
1456 /**
1457  * e1000_consume_page - helper function
1458  **/
1459 static void e1000_consume_page(struct e1000_buffer *bi, struct sk_buff *skb,
1460 			       u16 length)
1461 {
1462 	bi->page = NULL;
1463 	skb->len += length;
1464 	skb->data_len += length;
1465 	skb->truesize += PAGE_SIZE;
1466 }
1467 
1468 /**
1469  * e1000_clean_jumbo_rx_irq - Send received data up the network stack; legacy
1470  * @adapter: board private structure
1471  *
1472  * the return value indicates whether actual cleaning was done, there
1473  * is no guarantee that everything was cleaned
1474  **/
1475 static bool e1000_clean_jumbo_rx_irq(struct e1000_ring *rx_ring, int *work_done,
1476 				     int work_to_do)
1477 {
1478 	struct e1000_adapter *adapter = rx_ring->adapter;
1479 	struct net_device *netdev = adapter->netdev;
1480 	struct pci_dev *pdev = adapter->pdev;
1481 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
1482 	struct e1000_buffer *buffer_info, *next_buffer;
1483 	u32 length, staterr;
1484 	unsigned int i;
1485 	int cleaned_count = 0;
1486 	bool cleaned = false;
1487 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1488 	struct skb_shared_info *shinfo;
1489 
1490 	i = rx_ring->next_to_clean;
1491 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
1492 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1493 	buffer_info = &rx_ring->buffer_info[i];
1494 
1495 	while (staterr & E1000_RXD_STAT_DD) {
1496 		struct sk_buff *skb;
1497 
1498 		if (*work_done >= work_to_do)
1499 			break;
1500 		(*work_done)++;
1501 		rmb();	/* read descriptor and rx_buffer_info after status DD */
1502 
1503 		skb = buffer_info->skb;
1504 		buffer_info->skb = NULL;
1505 
1506 		++i;
1507 		if (i == rx_ring->count)
1508 			i = 0;
1509 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
1510 		prefetch(next_rxd);
1511 
1512 		next_buffer = &rx_ring->buffer_info[i];
1513 
1514 		cleaned = true;
1515 		cleaned_count++;
1516 		dma_unmap_page(&pdev->dev, buffer_info->dma, PAGE_SIZE,
1517 			       DMA_FROM_DEVICE);
1518 		buffer_info->dma = 0;
1519 
1520 		length = le16_to_cpu(rx_desc->wb.upper.length);
1521 
1522 		/* errors is only valid for DD + EOP descriptors */
1523 		if (unlikely((staterr & E1000_RXD_STAT_EOP) &&
1524 			     ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1525 			      !(netdev->features & NETIF_F_RXALL)))) {
1526 			/* recycle both page and skb */
1527 			buffer_info->skb = skb;
1528 			/* an error means any chain goes out the window too */
1529 			if (rx_ring->rx_skb_top)
1530 				dev_kfree_skb_irq(rx_ring->rx_skb_top);
1531 			rx_ring->rx_skb_top = NULL;
1532 			goto next_desc;
1533 		}
1534 #define rxtop (rx_ring->rx_skb_top)
1535 		if (!(staterr & E1000_RXD_STAT_EOP)) {
1536 			/* this descriptor is only the beginning (or middle) */
1537 			if (!rxtop) {
1538 				/* this is the beginning of a chain */
1539 				rxtop = skb;
1540 				skb_fill_page_desc(rxtop, 0, buffer_info->page,
1541 						   0, length);
1542 			} else {
1543 				/* this is the middle of a chain */
1544 				shinfo = skb_shinfo(rxtop);
1545 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1546 						   buffer_info->page, 0,
1547 						   length);
1548 				/* re-use the skb, only consumed the page */
1549 				buffer_info->skb = skb;
1550 			}
1551 			e1000_consume_page(buffer_info, rxtop, length);
1552 			goto next_desc;
1553 		} else {
1554 			if (rxtop) {
1555 				/* end of the chain */
1556 				shinfo = skb_shinfo(rxtop);
1557 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1558 						   buffer_info->page, 0,
1559 						   length);
1560 				/* re-use the current skb, we only consumed the
1561 				 * page
1562 				 */
1563 				buffer_info->skb = skb;
1564 				skb = rxtop;
1565 				rxtop = NULL;
1566 				e1000_consume_page(buffer_info, skb, length);
1567 			} else {
1568 				/* no chain, got EOP, this buf is the packet
1569 				 * copybreak to save the put_page/alloc_page
1570 				 */
1571 				if (length <= copybreak &&
1572 				    skb_tailroom(skb) >= length) {
1573 					u8 *vaddr;
1574 					vaddr = kmap_atomic(buffer_info->page);
1575 					memcpy(skb_tail_pointer(skb), vaddr,
1576 					       length);
1577 					kunmap_atomic(vaddr);
1578 					/* re-use the page, so don't erase
1579 					 * buffer_info->page
1580 					 */
1581 					skb_put(skb, length);
1582 				} else {
1583 					skb_fill_page_desc(skb, 0,
1584 							   buffer_info->page, 0,
1585 							   length);
1586 					e1000_consume_page(buffer_info, skb,
1587 							   length);
1588 				}
1589 			}
1590 		}
1591 
1592 		/* Receive Checksum Offload */
1593 		e1000_rx_checksum(adapter, staterr, skb);
1594 
1595 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1596 
1597 		/* probably a little skewed due to removing CRC */
1598 		total_rx_bytes += skb->len;
1599 		total_rx_packets++;
1600 
1601 		/* eth type trans needs skb->data to point to something */
1602 		if (!pskb_may_pull(skb, ETH_HLEN)) {
1603 			e_err("pskb_may_pull failed.\n");
1604 			dev_kfree_skb_irq(skb);
1605 			goto next_desc;
1606 		}
1607 
1608 		e1000_receive_skb(adapter, netdev, skb, staterr,
1609 				  rx_desc->wb.upper.vlan);
1610 
1611 next_desc:
1612 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1613 
1614 		/* return some buffers to hardware, one at a time is too slow */
1615 		if (unlikely(cleaned_count >= E1000_RX_BUFFER_WRITE)) {
1616 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1617 					      GFP_ATOMIC);
1618 			cleaned_count = 0;
1619 		}
1620 
1621 		/* use prefetched values */
1622 		rx_desc = next_rxd;
1623 		buffer_info = next_buffer;
1624 
1625 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1626 	}
1627 	rx_ring->next_to_clean = i;
1628 
1629 	cleaned_count = e1000_desc_unused(rx_ring);
1630 	if (cleaned_count)
1631 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1632 
1633 	adapter->total_rx_bytes += total_rx_bytes;
1634 	adapter->total_rx_packets += total_rx_packets;
1635 	return cleaned;
1636 }
1637 
1638 /**
1639  * e1000_clean_rx_ring - Free Rx Buffers per Queue
1640  * @rx_ring: Rx descriptor ring
1641  **/
1642 static void e1000_clean_rx_ring(struct e1000_ring *rx_ring)
1643 {
1644 	struct e1000_adapter *adapter = rx_ring->adapter;
1645 	struct e1000_buffer *buffer_info;
1646 	struct e1000_ps_page *ps_page;
1647 	struct pci_dev *pdev = adapter->pdev;
1648 	unsigned int i, j;
1649 
1650 	/* Free all the Rx ring sk_buffs */
1651 	for (i = 0; i < rx_ring->count; i++) {
1652 		buffer_info = &rx_ring->buffer_info[i];
1653 		if (buffer_info->dma) {
1654 			if (adapter->clean_rx == e1000_clean_rx_irq)
1655 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1656 						 adapter->rx_buffer_len,
1657 						 DMA_FROM_DEVICE);
1658 			else if (adapter->clean_rx == e1000_clean_jumbo_rx_irq)
1659 				dma_unmap_page(&pdev->dev, buffer_info->dma,
1660 					       PAGE_SIZE, DMA_FROM_DEVICE);
1661 			else if (adapter->clean_rx == e1000_clean_rx_irq_ps)
1662 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1663 						 adapter->rx_ps_bsize0,
1664 						 DMA_FROM_DEVICE);
1665 			buffer_info->dma = 0;
1666 		}
1667 
1668 		if (buffer_info->page) {
1669 			put_page(buffer_info->page);
1670 			buffer_info->page = NULL;
1671 		}
1672 
1673 		if (buffer_info->skb) {
1674 			dev_kfree_skb(buffer_info->skb);
1675 			buffer_info->skb = NULL;
1676 		}
1677 
1678 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1679 			ps_page = &buffer_info->ps_pages[j];
1680 			if (!ps_page->page)
1681 				break;
1682 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1683 				       DMA_FROM_DEVICE);
1684 			ps_page->dma = 0;
1685 			put_page(ps_page->page);
1686 			ps_page->page = NULL;
1687 		}
1688 	}
1689 
1690 	/* there also may be some cached data from a chained receive */
1691 	if (rx_ring->rx_skb_top) {
1692 		dev_kfree_skb(rx_ring->rx_skb_top);
1693 		rx_ring->rx_skb_top = NULL;
1694 	}
1695 
1696 	/* Zero out the descriptor ring */
1697 	memset(rx_ring->desc, 0, rx_ring->size);
1698 
1699 	rx_ring->next_to_clean = 0;
1700 	rx_ring->next_to_use = 0;
1701 	adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1702 
1703 	writel(0, rx_ring->head);
1704 	if (rx_ring->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
1705 		e1000e_update_rdt_wa(rx_ring, 0);
1706 	else
1707 		writel(0, rx_ring->tail);
1708 }
1709 
1710 static void e1000e_downshift_workaround(struct work_struct *work)
1711 {
1712 	struct e1000_adapter *adapter = container_of(work,
1713 						     struct e1000_adapter,
1714 						     downshift_task);
1715 
1716 	if (test_bit(__E1000_DOWN, &adapter->state))
1717 		return;
1718 
1719 	e1000e_gig_downshift_workaround_ich8lan(&adapter->hw);
1720 }
1721 
1722 /**
1723  * e1000_intr_msi - Interrupt Handler
1724  * @irq: interrupt number
1725  * @data: pointer to a network interface device structure
1726  **/
1727 static irqreturn_t e1000_intr_msi(int __always_unused irq, void *data)
1728 {
1729 	struct net_device *netdev = data;
1730 	struct e1000_adapter *adapter = netdev_priv(netdev);
1731 	struct e1000_hw *hw = &adapter->hw;
1732 	u32 icr = er32(ICR);
1733 
1734 	/* read ICR disables interrupts using IAM */
1735 	if (icr & E1000_ICR_LSC) {
1736 		hw->mac.get_link_status = true;
1737 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1738 		 * disconnect (LSC) before accessing any PHY registers
1739 		 */
1740 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1741 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1742 			schedule_work(&adapter->downshift_task);
1743 
1744 		/* 80003ES2LAN workaround-- For packet buffer work-around on
1745 		 * link down event; disable receives here in the ISR and reset
1746 		 * adapter in watchdog
1747 		 */
1748 		if (netif_carrier_ok(netdev) &&
1749 		    adapter->flags & FLAG_RX_NEEDS_RESTART) {
1750 			/* disable receives */
1751 			u32 rctl = er32(RCTL);
1752 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1753 			adapter->flags |= FLAG_RESTART_NOW;
1754 		}
1755 		/* guard against interrupt when we're going down */
1756 		if (!test_bit(__E1000_DOWN, &adapter->state))
1757 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1758 	}
1759 
1760 	/* Reset on uncorrectable ECC error */
1761 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type == e1000_pch_lpt)) {
1762 		u32 pbeccsts = er32(PBECCSTS);
1763 
1764 		adapter->corr_errors +=
1765 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1766 		adapter->uncorr_errors +=
1767 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1768 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1769 
1770 		/* Do the reset outside of interrupt context */
1771 		schedule_work(&adapter->reset_task);
1772 
1773 		/* return immediately since reset is imminent */
1774 		return IRQ_HANDLED;
1775 	}
1776 
1777 	if (napi_schedule_prep(&adapter->napi)) {
1778 		adapter->total_tx_bytes = 0;
1779 		adapter->total_tx_packets = 0;
1780 		adapter->total_rx_bytes = 0;
1781 		adapter->total_rx_packets = 0;
1782 		__napi_schedule(&adapter->napi);
1783 	}
1784 
1785 	return IRQ_HANDLED;
1786 }
1787 
1788 /**
1789  * e1000_intr - Interrupt Handler
1790  * @irq: interrupt number
1791  * @data: pointer to a network interface device structure
1792  **/
1793 static irqreturn_t e1000_intr(int __always_unused irq, void *data)
1794 {
1795 	struct net_device *netdev = data;
1796 	struct e1000_adapter *adapter = netdev_priv(netdev);
1797 	struct e1000_hw *hw = &adapter->hw;
1798 	u32 rctl, icr = er32(ICR);
1799 
1800 	if (!icr || test_bit(__E1000_DOWN, &adapter->state))
1801 		return IRQ_NONE;	/* Not our interrupt */
1802 
1803 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
1804 	 * not set, then the adapter didn't send an interrupt
1805 	 */
1806 	if (!(icr & E1000_ICR_INT_ASSERTED))
1807 		return IRQ_NONE;
1808 
1809 	/* Interrupt Auto-Mask...upon reading ICR,
1810 	 * interrupts are masked.  No need for the
1811 	 * IMC write
1812 	 */
1813 
1814 	if (icr & E1000_ICR_LSC) {
1815 		hw->mac.get_link_status = true;
1816 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1817 		 * disconnect (LSC) before accessing any PHY registers
1818 		 */
1819 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1820 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1821 			schedule_work(&adapter->downshift_task);
1822 
1823 		/* 80003ES2LAN workaround--
1824 		 * For packet buffer work-around on link down event;
1825 		 * disable receives here in the ISR and
1826 		 * reset adapter in watchdog
1827 		 */
1828 		if (netif_carrier_ok(netdev) &&
1829 		    (adapter->flags & FLAG_RX_NEEDS_RESTART)) {
1830 			/* disable receives */
1831 			rctl = er32(RCTL);
1832 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1833 			adapter->flags |= FLAG_RESTART_NOW;
1834 		}
1835 		/* guard against interrupt when we're going down */
1836 		if (!test_bit(__E1000_DOWN, &adapter->state))
1837 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1838 	}
1839 
1840 	/* Reset on uncorrectable ECC error */
1841 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type == e1000_pch_lpt)) {
1842 		u32 pbeccsts = er32(PBECCSTS);
1843 
1844 		adapter->corr_errors +=
1845 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1846 		adapter->uncorr_errors +=
1847 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1848 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1849 
1850 		/* Do the reset outside of interrupt context */
1851 		schedule_work(&adapter->reset_task);
1852 
1853 		/* return immediately since reset is imminent */
1854 		return IRQ_HANDLED;
1855 	}
1856 
1857 	if (napi_schedule_prep(&adapter->napi)) {
1858 		adapter->total_tx_bytes = 0;
1859 		adapter->total_tx_packets = 0;
1860 		adapter->total_rx_bytes = 0;
1861 		adapter->total_rx_packets = 0;
1862 		__napi_schedule(&adapter->napi);
1863 	}
1864 
1865 	return IRQ_HANDLED;
1866 }
1867 
1868 static irqreturn_t e1000_msix_other(int __always_unused irq, void *data)
1869 {
1870 	struct net_device *netdev = data;
1871 	struct e1000_adapter *adapter = netdev_priv(netdev);
1872 	struct e1000_hw *hw = &adapter->hw;
1873 	u32 icr = er32(ICR);
1874 
1875 	if (!(icr & E1000_ICR_INT_ASSERTED)) {
1876 		if (!test_bit(__E1000_DOWN, &adapter->state))
1877 			ew32(IMS, E1000_IMS_OTHER);
1878 		return IRQ_NONE;
1879 	}
1880 
1881 	if (icr & adapter->eiac_mask)
1882 		ew32(ICS, (icr & adapter->eiac_mask));
1883 
1884 	if (icr & E1000_ICR_OTHER) {
1885 		if (!(icr & E1000_ICR_LSC))
1886 			goto no_link_interrupt;
1887 		hw->mac.get_link_status = true;
1888 		/* guard against interrupt when we're going down */
1889 		if (!test_bit(__E1000_DOWN, &adapter->state))
1890 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1891 	}
1892 
1893 no_link_interrupt:
1894 	if (!test_bit(__E1000_DOWN, &adapter->state))
1895 		ew32(IMS, E1000_IMS_LSC | E1000_IMS_OTHER);
1896 
1897 	return IRQ_HANDLED;
1898 }
1899 
1900 static irqreturn_t e1000_intr_msix_tx(int __always_unused irq, void *data)
1901 {
1902 	struct net_device *netdev = data;
1903 	struct e1000_adapter *adapter = netdev_priv(netdev);
1904 	struct e1000_hw *hw = &adapter->hw;
1905 	struct e1000_ring *tx_ring = adapter->tx_ring;
1906 
1907 	adapter->total_tx_bytes = 0;
1908 	adapter->total_tx_packets = 0;
1909 
1910 	if (!e1000_clean_tx_irq(tx_ring))
1911 		/* Ring was not completely cleaned, so fire another interrupt */
1912 		ew32(ICS, tx_ring->ims_val);
1913 
1914 	return IRQ_HANDLED;
1915 }
1916 
1917 static irqreturn_t e1000_intr_msix_rx(int __always_unused irq, void *data)
1918 {
1919 	struct net_device *netdev = data;
1920 	struct e1000_adapter *adapter = netdev_priv(netdev);
1921 	struct e1000_ring *rx_ring = adapter->rx_ring;
1922 
1923 	/* Write the ITR value calculated at the end of the
1924 	 * previous interrupt.
1925 	 */
1926 	if (rx_ring->set_itr) {
1927 		writel(1000000000 / (rx_ring->itr_val * 256),
1928 		       rx_ring->itr_register);
1929 		rx_ring->set_itr = 0;
1930 	}
1931 
1932 	if (napi_schedule_prep(&adapter->napi)) {
1933 		adapter->total_rx_bytes = 0;
1934 		adapter->total_rx_packets = 0;
1935 		__napi_schedule(&adapter->napi);
1936 	}
1937 	return IRQ_HANDLED;
1938 }
1939 
1940 /**
1941  * e1000_configure_msix - Configure MSI-X hardware
1942  *
1943  * e1000_configure_msix sets up the hardware to properly
1944  * generate MSI-X interrupts.
1945  **/
1946 static void e1000_configure_msix(struct e1000_adapter *adapter)
1947 {
1948 	struct e1000_hw *hw = &adapter->hw;
1949 	struct e1000_ring *rx_ring = adapter->rx_ring;
1950 	struct e1000_ring *tx_ring = adapter->tx_ring;
1951 	int vector = 0;
1952 	u32 ctrl_ext, ivar = 0;
1953 
1954 	adapter->eiac_mask = 0;
1955 
1956 	/* Workaround issue with spurious interrupts on 82574 in MSI-X mode */
1957 	if (hw->mac.type == e1000_82574) {
1958 		u32 rfctl = er32(RFCTL);
1959 		rfctl |= E1000_RFCTL_ACK_DIS;
1960 		ew32(RFCTL, rfctl);
1961 	}
1962 
1963 	/* Configure Rx vector */
1964 	rx_ring->ims_val = E1000_IMS_RXQ0;
1965 	adapter->eiac_mask |= rx_ring->ims_val;
1966 	if (rx_ring->itr_val)
1967 		writel(1000000000 / (rx_ring->itr_val * 256),
1968 		       rx_ring->itr_register);
1969 	else
1970 		writel(1, rx_ring->itr_register);
1971 	ivar = E1000_IVAR_INT_ALLOC_VALID | vector;
1972 
1973 	/* Configure Tx vector */
1974 	tx_ring->ims_val = E1000_IMS_TXQ0;
1975 	vector++;
1976 	if (tx_ring->itr_val)
1977 		writel(1000000000 / (tx_ring->itr_val * 256),
1978 		       tx_ring->itr_register);
1979 	else
1980 		writel(1, tx_ring->itr_register);
1981 	adapter->eiac_mask |= tx_ring->ims_val;
1982 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 8);
1983 
1984 	/* set vector for Other Causes, e.g. link changes */
1985 	vector++;
1986 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 16);
1987 	if (rx_ring->itr_val)
1988 		writel(1000000000 / (rx_ring->itr_val * 256),
1989 		       hw->hw_addr + E1000_EITR_82574(vector));
1990 	else
1991 		writel(1, hw->hw_addr + E1000_EITR_82574(vector));
1992 
1993 	/* Cause Tx interrupts on every write back */
1994 	ivar |= (1 << 31);
1995 
1996 	ew32(IVAR, ivar);
1997 
1998 	/* enable MSI-X PBA support */
1999 	ctrl_ext = er32(CTRL_EXT);
2000 	ctrl_ext |= E1000_CTRL_EXT_PBA_CLR;
2001 
2002 	/* Auto-Mask Other interrupts upon ICR read */
2003 	ew32(IAM, ~E1000_EIAC_MASK_82574 | E1000_IMS_OTHER);
2004 	ctrl_ext |= E1000_CTRL_EXT_EIAME;
2005 	ew32(CTRL_EXT, ctrl_ext);
2006 	e1e_flush();
2007 }
2008 
2009 void e1000e_reset_interrupt_capability(struct e1000_adapter *adapter)
2010 {
2011 	if (adapter->msix_entries) {
2012 		pci_disable_msix(adapter->pdev);
2013 		kfree(adapter->msix_entries);
2014 		adapter->msix_entries = NULL;
2015 	} else if (adapter->flags & FLAG_MSI_ENABLED) {
2016 		pci_disable_msi(adapter->pdev);
2017 		adapter->flags &= ~FLAG_MSI_ENABLED;
2018 	}
2019 }
2020 
2021 /**
2022  * e1000e_set_interrupt_capability - set MSI or MSI-X if supported
2023  *
2024  * Attempt to configure interrupts using the best available
2025  * capabilities of the hardware and kernel.
2026  **/
2027 void e1000e_set_interrupt_capability(struct e1000_adapter *adapter)
2028 {
2029 	int err;
2030 	int i;
2031 
2032 	switch (adapter->int_mode) {
2033 	case E1000E_INT_MODE_MSIX:
2034 		if (adapter->flags & FLAG_HAS_MSIX) {
2035 			adapter->num_vectors = 3; /* RxQ0, TxQ0 and other */
2036 			adapter->msix_entries = kcalloc(adapter->num_vectors,
2037 							sizeof(struct
2038 							       msix_entry),
2039 							GFP_KERNEL);
2040 			if (adapter->msix_entries) {
2041 				for (i = 0; i < adapter->num_vectors; i++)
2042 					adapter->msix_entries[i].entry = i;
2043 
2044 				err = pci_enable_msix(adapter->pdev,
2045 						      adapter->msix_entries,
2046 						      adapter->num_vectors);
2047 				if (err == 0)
2048 					return;
2049 			}
2050 			/* MSI-X failed, so fall through and try MSI */
2051 			e_err("Failed to initialize MSI-X interrupts.  Falling back to MSI interrupts.\n");
2052 			e1000e_reset_interrupt_capability(adapter);
2053 		}
2054 		adapter->int_mode = E1000E_INT_MODE_MSI;
2055 		/* Fall through */
2056 	case E1000E_INT_MODE_MSI:
2057 		if (!pci_enable_msi(adapter->pdev)) {
2058 			adapter->flags |= FLAG_MSI_ENABLED;
2059 		} else {
2060 			adapter->int_mode = E1000E_INT_MODE_LEGACY;
2061 			e_err("Failed to initialize MSI interrupts.  Falling back to legacy interrupts.\n");
2062 		}
2063 		/* Fall through */
2064 	case E1000E_INT_MODE_LEGACY:
2065 		/* Don't do anything; this is the system default */
2066 		break;
2067 	}
2068 
2069 	/* store the number of vectors being used */
2070 	adapter->num_vectors = 1;
2071 }
2072 
2073 /**
2074  * e1000_request_msix - Initialize MSI-X interrupts
2075  *
2076  * e1000_request_msix allocates MSI-X vectors and requests interrupts from the
2077  * kernel.
2078  **/
2079 static int e1000_request_msix(struct e1000_adapter *adapter)
2080 {
2081 	struct net_device *netdev = adapter->netdev;
2082 	int err = 0, vector = 0;
2083 
2084 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2085 		snprintf(adapter->rx_ring->name,
2086 			 sizeof(adapter->rx_ring->name) - 1,
2087 			 "%s-rx-0", netdev->name);
2088 	else
2089 		memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
2090 	err = request_irq(adapter->msix_entries[vector].vector,
2091 			  e1000_intr_msix_rx, 0, adapter->rx_ring->name,
2092 			  netdev);
2093 	if (err)
2094 		return err;
2095 	adapter->rx_ring->itr_register = adapter->hw.hw_addr +
2096 	    E1000_EITR_82574(vector);
2097 	adapter->rx_ring->itr_val = adapter->itr;
2098 	vector++;
2099 
2100 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2101 		snprintf(adapter->tx_ring->name,
2102 			 sizeof(adapter->tx_ring->name) - 1,
2103 			 "%s-tx-0", netdev->name);
2104 	else
2105 		memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
2106 	err = request_irq(adapter->msix_entries[vector].vector,
2107 			  e1000_intr_msix_tx, 0, adapter->tx_ring->name,
2108 			  netdev);
2109 	if (err)
2110 		return err;
2111 	adapter->tx_ring->itr_register = adapter->hw.hw_addr +
2112 	    E1000_EITR_82574(vector);
2113 	adapter->tx_ring->itr_val = adapter->itr;
2114 	vector++;
2115 
2116 	err = request_irq(adapter->msix_entries[vector].vector,
2117 			  e1000_msix_other, 0, netdev->name, netdev);
2118 	if (err)
2119 		return err;
2120 
2121 	e1000_configure_msix(adapter);
2122 
2123 	return 0;
2124 }
2125 
2126 /**
2127  * e1000_request_irq - initialize interrupts
2128  *
2129  * Attempts to configure interrupts using the best available
2130  * capabilities of the hardware and kernel.
2131  **/
2132 static int e1000_request_irq(struct e1000_adapter *adapter)
2133 {
2134 	struct net_device *netdev = adapter->netdev;
2135 	int err;
2136 
2137 	if (adapter->msix_entries) {
2138 		err = e1000_request_msix(adapter);
2139 		if (!err)
2140 			return err;
2141 		/* fall back to MSI */
2142 		e1000e_reset_interrupt_capability(adapter);
2143 		adapter->int_mode = E1000E_INT_MODE_MSI;
2144 		e1000e_set_interrupt_capability(adapter);
2145 	}
2146 	if (adapter->flags & FLAG_MSI_ENABLED) {
2147 		err = request_irq(adapter->pdev->irq, e1000_intr_msi, 0,
2148 				  netdev->name, netdev);
2149 		if (!err)
2150 			return err;
2151 
2152 		/* fall back to legacy interrupt */
2153 		e1000e_reset_interrupt_capability(adapter);
2154 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
2155 	}
2156 
2157 	err = request_irq(adapter->pdev->irq, e1000_intr, IRQF_SHARED,
2158 			  netdev->name, netdev);
2159 	if (err)
2160 		e_err("Unable to allocate interrupt, Error: %d\n", err);
2161 
2162 	return err;
2163 }
2164 
2165 static void e1000_free_irq(struct e1000_adapter *adapter)
2166 {
2167 	struct net_device *netdev = adapter->netdev;
2168 
2169 	if (adapter->msix_entries) {
2170 		int vector = 0;
2171 
2172 		free_irq(adapter->msix_entries[vector].vector, netdev);
2173 		vector++;
2174 
2175 		free_irq(adapter->msix_entries[vector].vector, netdev);
2176 		vector++;
2177 
2178 		/* Other Causes interrupt vector */
2179 		free_irq(adapter->msix_entries[vector].vector, netdev);
2180 		return;
2181 	}
2182 
2183 	free_irq(adapter->pdev->irq, netdev);
2184 }
2185 
2186 /**
2187  * e1000_irq_disable - Mask off interrupt generation on the NIC
2188  **/
2189 static void e1000_irq_disable(struct e1000_adapter *adapter)
2190 {
2191 	struct e1000_hw *hw = &adapter->hw;
2192 
2193 	ew32(IMC, ~0);
2194 	if (adapter->msix_entries)
2195 		ew32(EIAC_82574, 0);
2196 	e1e_flush();
2197 
2198 	if (adapter->msix_entries) {
2199 		int i;
2200 		for (i = 0; i < adapter->num_vectors; i++)
2201 			synchronize_irq(adapter->msix_entries[i].vector);
2202 	} else {
2203 		synchronize_irq(adapter->pdev->irq);
2204 	}
2205 }
2206 
2207 /**
2208  * e1000_irq_enable - Enable default interrupt generation settings
2209  **/
2210 static void e1000_irq_enable(struct e1000_adapter *adapter)
2211 {
2212 	struct e1000_hw *hw = &adapter->hw;
2213 
2214 	if (adapter->msix_entries) {
2215 		ew32(EIAC_82574, adapter->eiac_mask & E1000_EIAC_MASK_82574);
2216 		ew32(IMS, adapter->eiac_mask | E1000_IMS_OTHER | E1000_IMS_LSC);
2217 	} else if (hw->mac.type == e1000_pch_lpt) {
2218 		ew32(IMS, IMS_ENABLE_MASK | E1000_IMS_ECCER);
2219 	} else {
2220 		ew32(IMS, IMS_ENABLE_MASK);
2221 	}
2222 	e1e_flush();
2223 }
2224 
2225 /**
2226  * e1000e_get_hw_control - get control of the h/w from f/w
2227  * @adapter: address of board private structure
2228  *
2229  * e1000e_get_hw_control sets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2230  * For ASF and Pass Through versions of f/w this means that
2231  * the driver is loaded. For AMT version (only with 82573)
2232  * of the f/w this means that the network i/f is open.
2233  **/
2234 void e1000e_get_hw_control(struct e1000_adapter *adapter)
2235 {
2236 	struct e1000_hw *hw = &adapter->hw;
2237 	u32 ctrl_ext;
2238 	u32 swsm;
2239 
2240 	/* Let firmware know the driver has taken over */
2241 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2242 		swsm = er32(SWSM);
2243 		ew32(SWSM, swsm | E1000_SWSM_DRV_LOAD);
2244 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2245 		ctrl_ext = er32(CTRL_EXT);
2246 		ew32(CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2247 	}
2248 }
2249 
2250 /**
2251  * e1000e_release_hw_control - release control of the h/w to f/w
2252  * @adapter: address of board private structure
2253  *
2254  * e1000e_release_hw_control resets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2255  * For ASF and Pass Through versions of f/w this means that the
2256  * driver is no longer loaded. For AMT version (only with 82573) i
2257  * of the f/w this means that the network i/f is closed.
2258  *
2259  **/
2260 void e1000e_release_hw_control(struct e1000_adapter *adapter)
2261 {
2262 	struct e1000_hw *hw = &adapter->hw;
2263 	u32 ctrl_ext;
2264 	u32 swsm;
2265 
2266 	/* Let firmware taken over control of h/w */
2267 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2268 		swsm = er32(SWSM);
2269 		ew32(SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
2270 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2271 		ctrl_ext = er32(CTRL_EXT);
2272 		ew32(CTRL_EXT, ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2273 	}
2274 }
2275 
2276 /**
2277  * e1000_alloc_ring_dma - allocate memory for a ring structure
2278  **/
2279 static int e1000_alloc_ring_dma(struct e1000_adapter *adapter,
2280 				struct e1000_ring *ring)
2281 {
2282 	struct pci_dev *pdev = adapter->pdev;
2283 
2284 	ring->desc = dma_alloc_coherent(&pdev->dev, ring->size, &ring->dma,
2285 					GFP_KERNEL);
2286 	if (!ring->desc)
2287 		return -ENOMEM;
2288 
2289 	return 0;
2290 }
2291 
2292 /**
2293  * e1000e_setup_tx_resources - allocate Tx resources (Descriptors)
2294  * @tx_ring: Tx descriptor ring
2295  *
2296  * Return 0 on success, negative on failure
2297  **/
2298 int e1000e_setup_tx_resources(struct e1000_ring *tx_ring)
2299 {
2300 	struct e1000_adapter *adapter = tx_ring->adapter;
2301 	int err = -ENOMEM, size;
2302 
2303 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2304 	tx_ring->buffer_info = vzalloc(size);
2305 	if (!tx_ring->buffer_info)
2306 		goto err;
2307 
2308 	/* round up to nearest 4K */
2309 	tx_ring->size = tx_ring->count * sizeof(struct e1000_tx_desc);
2310 	tx_ring->size = ALIGN(tx_ring->size, 4096);
2311 
2312 	err = e1000_alloc_ring_dma(adapter, tx_ring);
2313 	if (err)
2314 		goto err;
2315 
2316 	tx_ring->next_to_use = 0;
2317 	tx_ring->next_to_clean = 0;
2318 
2319 	return 0;
2320 err:
2321 	vfree(tx_ring->buffer_info);
2322 	e_err("Unable to allocate memory for the transmit descriptor ring\n");
2323 	return err;
2324 }
2325 
2326 /**
2327  * e1000e_setup_rx_resources - allocate Rx resources (Descriptors)
2328  * @rx_ring: Rx descriptor ring
2329  *
2330  * Returns 0 on success, negative on failure
2331  **/
2332 int e1000e_setup_rx_resources(struct e1000_ring *rx_ring)
2333 {
2334 	struct e1000_adapter *adapter = rx_ring->adapter;
2335 	struct e1000_buffer *buffer_info;
2336 	int i, size, desc_len, err = -ENOMEM;
2337 
2338 	size = sizeof(struct e1000_buffer) * rx_ring->count;
2339 	rx_ring->buffer_info = vzalloc(size);
2340 	if (!rx_ring->buffer_info)
2341 		goto err;
2342 
2343 	for (i = 0; i < rx_ring->count; i++) {
2344 		buffer_info = &rx_ring->buffer_info[i];
2345 		buffer_info->ps_pages = kcalloc(PS_PAGE_BUFFERS,
2346 						sizeof(struct e1000_ps_page),
2347 						GFP_KERNEL);
2348 		if (!buffer_info->ps_pages)
2349 			goto err_pages;
2350 	}
2351 
2352 	desc_len = sizeof(union e1000_rx_desc_packet_split);
2353 
2354 	/* Round up to nearest 4K */
2355 	rx_ring->size = rx_ring->count * desc_len;
2356 	rx_ring->size = ALIGN(rx_ring->size, 4096);
2357 
2358 	err = e1000_alloc_ring_dma(adapter, rx_ring);
2359 	if (err)
2360 		goto err_pages;
2361 
2362 	rx_ring->next_to_clean = 0;
2363 	rx_ring->next_to_use = 0;
2364 	rx_ring->rx_skb_top = NULL;
2365 
2366 	return 0;
2367 
2368 err_pages:
2369 	for (i = 0; i < rx_ring->count; i++) {
2370 		buffer_info = &rx_ring->buffer_info[i];
2371 		kfree(buffer_info->ps_pages);
2372 	}
2373 err:
2374 	vfree(rx_ring->buffer_info);
2375 	e_err("Unable to allocate memory for the receive descriptor ring\n");
2376 	return err;
2377 }
2378 
2379 /**
2380  * e1000_clean_tx_ring - Free Tx Buffers
2381  * @tx_ring: Tx descriptor ring
2382  **/
2383 static void e1000_clean_tx_ring(struct e1000_ring *tx_ring)
2384 {
2385 	struct e1000_adapter *adapter = tx_ring->adapter;
2386 	struct e1000_buffer *buffer_info;
2387 	unsigned long size;
2388 	unsigned int i;
2389 
2390 	for (i = 0; i < tx_ring->count; i++) {
2391 		buffer_info = &tx_ring->buffer_info[i];
2392 		e1000_put_txbuf(tx_ring, buffer_info);
2393 	}
2394 
2395 	netdev_reset_queue(adapter->netdev);
2396 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2397 	memset(tx_ring->buffer_info, 0, size);
2398 
2399 	memset(tx_ring->desc, 0, tx_ring->size);
2400 
2401 	tx_ring->next_to_use = 0;
2402 	tx_ring->next_to_clean = 0;
2403 
2404 	writel(0, tx_ring->head);
2405 	if (tx_ring->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
2406 		e1000e_update_tdt_wa(tx_ring, 0);
2407 	else
2408 		writel(0, tx_ring->tail);
2409 }
2410 
2411 /**
2412  * e1000e_free_tx_resources - Free Tx Resources per Queue
2413  * @tx_ring: Tx descriptor ring
2414  *
2415  * Free all transmit software resources
2416  **/
2417 void e1000e_free_tx_resources(struct e1000_ring *tx_ring)
2418 {
2419 	struct e1000_adapter *adapter = tx_ring->adapter;
2420 	struct pci_dev *pdev = adapter->pdev;
2421 
2422 	e1000_clean_tx_ring(tx_ring);
2423 
2424 	vfree(tx_ring->buffer_info);
2425 	tx_ring->buffer_info = NULL;
2426 
2427 	dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
2428 			  tx_ring->dma);
2429 	tx_ring->desc = NULL;
2430 }
2431 
2432 /**
2433  * e1000e_free_rx_resources - Free Rx Resources
2434  * @rx_ring: Rx descriptor ring
2435  *
2436  * Free all receive software resources
2437  **/
2438 void e1000e_free_rx_resources(struct e1000_ring *rx_ring)
2439 {
2440 	struct e1000_adapter *adapter = rx_ring->adapter;
2441 	struct pci_dev *pdev = adapter->pdev;
2442 	int i;
2443 
2444 	e1000_clean_rx_ring(rx_ring);
2445 
2446 	for (i = 0; i < rx_ring->count; i++)
2447 		kfree(rx_ring->buffer_info[i].ps_pages);
2448 
2449 	vfree(rx_ring->buffer_info);
2450 	rx_ring->buffer_info = NULL;
2451 
2452 	dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
2453 			  rx_ring->dma);
2454 	rx_ring->desc = NULL;
2455 }
2456 
2457 /**
2458  * e1000_update_itr - update the dynamic ITR value based on statistics
2459  * @adapter: pointer to adapter
2460  * @itr_setting: current adapter->itr
2461  * @packets: the number of packets during this measurement interval
2462  * @bytes: the number of bytes during this measurement interval
2463  *
2464  *      Stores a new ITR value based on packets and byte
2465  *      counts during the last interrupt.  The advantage of per interrupt
2466  *      computation is faster updates and more accurate ITR for the current
2467  *      traffic pattern.  Constants in this function were computed
2468  *      based on theoretical maximum wire speed and thresholds were set based
2469  *      on testing data as well as attempting to minimize response time
2470  *      while increasing bulk throughput.  This functionality is controlled
2471  *      by the InterruptThrottleRate module parameter.
2472  **/
2473 static unsigned int e1000_update_itr(u16 itr_setting, int packets, int bytes)
2474 {
2475 	unsigned int retval = itr_setting;
2476 
2477 	if (packets == 0)
2478 		return itr_setting;
2479 
2480 	switch (itr_setting) {
2481 	case lowest_latency:
2482 		/* handle TSO and jumbo frames */
2483 		if (bytes / packets > 8000)
2484 			retval = bulk_latency;
2485 		else if ((packets < 5) && (bytes > 512))
2486 			retval = low_latency;
2487 		break;
2488 	case low_latency:	/* 50 usec aka 20000 ints/s */
2489 		if (bytes > 10000) {
2490 			/* this if handles the TSO accounting */
2491 			if (bytes / packets > 8000)
2492 				retval = bulk_latency;
2493 			else if ((packets < 10) || ((bytes / packets) > 1200))
2494 				retval = bulk_latency;
2495 			else if ((packets > 35))
2496 				retval = lowest_latency;
2497 		} else if (bytes / packets > 2000) {
2498 			retval = bulk_latency;
2499 		} else if (packets <= 2 && bytes < 512) {
2500 			retval = lowest_latency;
2501 		}
2502 		break;
2503 	case bulk_latency:	/* 250 usec aka 4000 ints/s */
2504 		if (bytes > 25000) {
2505 			if (packets > 35)
2506 				retval = low_latency;
2507 		} else if (bytes < 6000) {
2508 			retval = low_latency;
2509 		}
2510 		break;
2511 	}
2512 
2513 	return retval;
2514 }
2515 
2516 static void e1000_set_itr(struct e1000_adapter *adapter)
2517 {
2518 	u16 current_itr;
2519 	u32 new_itr = adapter->itr;
2520 
2521 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
2522 	if (adapter->link_speed != SPEED_1000) {
2523 		current_itr = 0;
2524 		new_itr = 4000;
2525 		goto set_itr_now;
2526 	}
2527 
2528 	if (adapter->flags2 & FLAG2_DISABLE_AIM) {
2529 		new_itr = 0;
2530 		goto set_itr_now;
2531 	}
2532 
2533 	adapter->tx_itr = e1000_update_itr(adapter->tx_itr,
2534 					   adapter->total_tx_packets,
2535 					   adapter->total_tx_bytes);
2536 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2537 	if (adapter->itr_setting == 3 && adapter->tx_itr == lowest_latency)
2538 		adapter->tx_itr = low_latency;
2539 
2540 	adapter->rx_itr = e1000_update_itr(adapter->rx_itr,
2541 					   adapter->total_rx_packets,
2542 					   adapter->total_rx_bytes);
2543 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2544 	if (adapter->itr_setting == 3 && adapter->rx_itr == lowest_latency)
2545 		adapter->rx_itr = low_latency;
2546 
2547 	current_itr = max(adapter->rx_itr, adapter->tx_itr);
2548 
2549 	/* counts and packets in update_itr are dependent on these numbers */
2550 	switch (current_itr) {
2551 	case lowest_latency:
2552 		new_itr = 70000;
2553 		break;
2554 	case low_latency:
2555 		new_itr = 20000;	/* aka hwitr = ~200 */
2556 		break;
2557 	case bulk_latency:
2558 		new_itr = 4000;
2559 		break;
2560 	default:
2561 		break;
2562 	}
2563 
2564 set_itr_now:
2565 	if (new_itr != adapter->itr) {
2566 		/* this attempts to bias the interrupt rate towards Bulk
2567 		 * by adding intermediate steps when interrupt rate is
2568 		 * increasing
2569 		 */
2570 		new_itr = new_itr > adapter->itr ?
2571 		    min(adapter->itr + (new_itr >> 2), new_itr) : new_itr;
2572 		adapter->itr = new_itr;
2573 		adapter->rx_ring->itr_val = new_itr;
2574 		if (adapter->msix_entries)
2575 			adapter->rx_ring->set_itr = 1;
2576 		else
2577 			e1000e_write_itr(adapter, new_itr);
2578 	}
2579 }
2580 
2581 /**
2582  * e1000e_write_itr - write the ITR value to the appropriate registers
2583  * @adapter: address of board private structure
2584  * @itr: new ITR value to program
2585  *
2586  * e1000e_write_itr determines if the adapter is in MSI-X mode
2587  * and, if so, writes the EITR registers with the ITR value.
2588  * Otherwise, it writes the ITR value into the ITR register.
2589  **/
2590 void e1000e_write_itr(struct e1000_adapter *adapter, u32 itr)
2591 {
2592 	struct e1000_hw *hw = &adapter->hw;
2593 	u32 new_itr = itr ? 1000000000 / (itr * 256) : 0;
2594 
2595 	if (adapter->msix_entries) {
2596 		int vector;
2597 
2598 		for (vector = 0; vector < adapter->num_vectors; vector++)
2599 			writel(new_itr, hw->hw_addr + E1000_EITR_82574(vector));
2600 	} else {
2601 		ew32(ITR, new_itr);
2602 	}
2603 }
2604 
2605 /**
2606  * e1000_alloc_queues - Allocate memory for all rings
2607  * @adapter: board private structure to initialize
2608  **/
2609 static int e1000_alloc_queues(struct e1000_adapter *adapter)
2610 {
2611 	int size = sizeof(struct e1000_ring);
2612 
2613 	adapter->tx_ring = kzalloc(size, GFP_KERNEL);
2614 	if (!adapter->tx_ring)
2615 		goto err;
2616 	adapter->tx_ring->count = adapter->tx_ring_count;
2617 	adapter->tx_ring->adapter = adapter;
2618 
2619 	adapter->rx_ring = kzalloc(size, GFP_KERNEL);
2620 	if (!adapter->rx_ring)
2621 		goto err;
2622 	adapter->rx_ring->count = adapter->rx_ring_count;
2623 	adapter->rx_ring->adapter = adapter;
2624 
2625 	return 0;
2626 err:
2627 	e_err("Unable to allocate memory for queues\n");
2628 	kfree(adapter->rx_ring);
2629 	kfree(adapter->tx_ring);
2630 	return -ENOMEM;
2631 }
2632 
2633 /**
2634  * e1000e_poll - NAPI Rx polling callback
2635  * @napi: struct associated with this polling callback
2636  * @weight: number of packets driver is allowed to process this poll
2637  **/
2638 static int e1000e_poll(struct napi_struct *napi, int weight)
2639 {
2640 	struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter,
2641 						     napi);
2642 	struct e1000_hw *hw = &adapter->hw;
2643 	struct net_device *poll_dev = adapter->netdev;
2644 	int tx_cleaned = 1, work_done = 0;
2645 
2646 	adapter = netdev_priv(poll_dev);
2647 
2648 	if (!adapter->msix_entries ||
2649 	    (adapter->rx_ring->ims_val & adapter->tx_ring->ims_val))
2650 		tx_cleaned = e1000_clean_tx_irq(adapter->tx_ring);
2651 
2652 	adapter->clean_rx(adapter->rx_ring, &work_done, weight);
2653 
2654 	if (!tx_cleaned)
2655 		work_done = weight;
2656 
2657 	/* If weight not fully consumed, exit the polling mode */
2658 	if (work_done < weight) {
2659 		if (adapter->itr_setting & 3)
2660 			e1000_set_itr(adapter);
2661 		napi_complete(napi);
2662 		if (!test_bit(__E1000_DOWN, &adapter->state)) {
2663 			if (adapter->msix_entries)
2664 				ew32(IMS, adapter->rx_ring->ims_val);
2665 			else
2666 				e1000_irq_enable(adapter);
2667 		}
2668 	}
2669 
2670 	return work_done;
2671 }
2672 
2673 static int e1000_vlan_rx_add_vid(struct net_device *netdev,
2674 				 __always_unused __be16 proto, u16 vid)
2675 {
2676 	struct e1000_adapter *adapter = netdev_priv(netdev);
2677 	struct e1000_hw *hw = &adapter->hw;
2678 	u32 vfta, index;
2679 
2680 	/* don't update vlan cookie if already programmed */
2681 	if ((adapter->hw.mng_cookie.status &
2682 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2683 	    (vid == adapter->mng_vlan_id))
2684 		return 0;
2685 
2686 	/* add VID to filter table */
2687 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2688 		index = (vid >> 5) & 0x7F;
2689 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2690 		vfta |= (1 << (vid & 0x1F));
2691 		hw->mac.ops.write_vfta(hw, index, vfta);
2692 	}
2693 
2694 	set_bit(vid, adapter->active_vlans);
2695 
2696 	return 0;
2697 }
2698 
2699 static int e1000_vlan_rx_kill_vid(struct net_device *netdev,
2700 				  __always_unused __be16 proto, u16 vid)
2701 {
2702 	struct e1000_adapter *adapter = netdev_priv(netdev);
2703 	struct e1000_hw *hw = &adapter->hw;
2704 	u32 vfta, index;
2705 
2706 	if ((adapter->hw.mng_cookie.status &
2707 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2708 	    (vid == adapter->mng_vlan_id)) {
2709 		/* release control to f/w */
2710 		e1000e_release_hw_control(adapter);
2711 		return 0;
2712 	}
2713 
2714 	/* remove VID from filter table */
2715 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2716 		index = (vid >> 5) & 0x7F;
2717 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2718 		vfta &= ~(1 << (vid & 0x1F));
2719 		hw->mac.ops.write_vfta(hw, index, vfta);
2720 	}
2721 
2722 	clear_bit(vid, adapter->active_vlans);
2723 
2724 	return 0;
2725 }
2726 
2727 /**
2728  * e1000e_vlan_filter_disable - helper to disable hw VLAN filtering
2729  * @adapter: board private structure to initialize
2730  **/
2731 static void e1000e_vlan_filter_disable(struct e1000_adapter *adapter)
2732 {
2733 	struct net_device *netdev = adapter->netdev;
2734 	struct e1000_hw *hw = &adapter->hw;
2735 	u32 rctl;
2736 
2737 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2738 		/* disable VLAN receive filtering */
2739 		rctl = er32(RCTL);
2740 		rctl &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
2741 		ew32(RCTL, rctl);
2742 
2743 		if (adapter->mng_vlan_id != (u16)E1000_MNG_VLAN_NONE) {
2744 			e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
2745 					       adapter->mng_vlan_id);
2746 			adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
2747 		}
2748 	}
2749 }
2750 
2751 /**
2752  * e1000e_vlan_filter_enable - helper to enable HW VLAN filtering
2753  * @adapter: board private structure to initialize
2754  **/
2755 static void e1000e_vlan_filter_enable(struct e1000_adapter *adapter)
2756 {
2757 	struct e1000_hw *hw = &adapter->hw;
2758 	u32 rctl;
2759 
2760 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2761 		/* enable VLAN receive filtering */
2762 		rctl = er32(RCTL);
2763 		rctl |= E1000_RCTL_VFE;
2764 		rctl &= ~E1000_RCTL_CFIEN;
2765 		ew32(RCTL, rctl);
2766 	}
2767 }
2768 
2769 /**
2770  * e1000e_vlan_strip_enable - helper to disable HW VLAN stripping
2771  * @adapter: board private structure to initialize
2772  **/
2773 static void e1000e_vlan_strip_disable(struct e1000_adapter *adapter)
2774 {
2775 	struct e1000_hw *hw = &adapter->hw;
2776 	u32 ctrl;
2777 
2778 	/* disable VLAN tag insert/strip */
2779 	ctrl = er32(CTRL);
2780 	ctrl &= ~E1000_CTRL_VME;
2781 	ew32(CTRL, ctrl);
2782 }
2783 
2784 /**
2785  * e1000e_vlan_strip_enable - helper to enable HW VLAN stripping
2786  * @adapter: board private structure to initialize
2787  **/
2788 static void e1000e_vlan_strip_enable(struct e1000_adapter *adapter)
2789 {
2790 	struct e1000_hw *hw = &adapter->hw;
2791 	u32 ctrl;
2792 
2793 	/* enable VLAN tag insert/strip */
2794 	ctrl = er32(CTRL);
2795 	ctrl |= E1000_CTRL_VME;
2796 	ew32(CTRL, ctrl);
2797 }
2798 
2799 static void e1000_update_mng_vlan(struct e1000_adapter *adapter)
2800 {
2801 	struct net_device *netdev = adapter->netdev;
2802 	u16 vid = adapter->hw.mng_cookie.vlan_id;
2803 	u16 old_vid = adapter->mng_vlan_id;
2804 
2805 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
2806 		e1000_vlan_rx_add_vid(netdev, htons(ETH_P_8021Q), vid);
2807 		adapter->mng_vlan_id = vid;
2808 	}
2809 
2810 	if ((old_vid != (u16)E1000_MNG_VLAN_NONE) && (vid != old_vid))
2811 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q), old_vid);
2812 }
2813 
2814 static void e1000_restore_vlan(struct e1000_adapter *adapter)
2815 {
2816 	u16 vid;
2817 
2818 	e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
2819 
2820 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
2821 	    e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
2822 }
2823 
2824 static void e1000_init_manageability_pt(struct e1000_adapter *adapter)
2825 {
2826 	struct e1000_hw *hw = &adapter->hw;
2827 	u32 manc, manc2h, mdef, i, j;
2828 
2829 	if (!(adapter->flags & FLAG_MNG_PT_ENABLED))
2830 		return;
2831 
2832 	manc = er32(MANC);
2833 
2834 	/* enable receiving management packets to the host. this will probably
2835 	 * generate destination unreachable messages from the host OS, but
2836 	 * the packets will be handled on SMBUS
2837 	 */
2838 	manc |= E1000_MANC_EN_MNG2HOST;
2839 	manc2h = er32(MANC2H);
2840 
2841 	switch (hw->mac.type) {
2842 	default:
2843 		manc2h |= (E1000_MANC2H_PORT_623 | E1000_MANC2H_PORT_664);
2844 		break;
2845 	case e1000_82574:
2846 	case e1000_82583:
2847 		/* Check if IPMI pass-through decision filter already exists;
2848 		 * if so, enable it.
2849 		 */
2850 		for (i = 0, j = 0; i < 8; i++) {
2851 			mdef = er32(MDEF(i));
2852 
2853 			/* Ignore filters with anything other than IPMI ports */
2854 			if (mdef & ~(E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2855 				continue;
2856 
2857 			/* Enable this decision filter in MANC2H */
2858 			if (mdef)
2859 				manc2h |= (1 << i);
2860 
2861 			j |= mdef;
2862 		}
2863 
2864 		if (j == (E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2865 			break;
2866 
2867 		/* Create new decision filter in an empty filter */
2868 		for (i = 0, j = 0; i < 8; i++)
2869 			if (er32(MDEF(i)) == 0) {
2870 				ew32(MDEF(i), (E1000_MDEF_PORT_623 |
2871 					       E1000_MDEF_PORT_664));
2872 				manc2h |= (1 << 1);
2873 				j++;
2874 				break;
2875 			}
2876 
2877 		if (!j)
2878 			e_warn("Unable to create IPMI pass-through filter\n");
2879 		break;
2880 	}
2881 
2882 	ew32(MANC2H, manc2h);
2883 	ew32(MANC, manc);
2884 }
2885 
2886 /**
2887  * e1000_configure_tx - Configure Transmit Unit after Reset
2888  * @adapter: board private structure
2889  *
2890  * Configure the Tx unit of the MAC after a reset.
2891  **/
2892 static void e1000_configure_tx(struct e1000_adapter *adapter)
2893 {
2894 	struct e1000_hw *hw = &adapter->hw;
2895 	struct e1000_ring *tx_ring = adapter->tx_ring;
2896 	u64 tdba;
2897 	u32 tdlen, tarc;
2898 
2899 	/* Setup the HW Tx Head and Tail descriptor pointers */
2900 	tdba = tx_ring->dma;
2901 	tdlen = tx_ring->count * sizeof(struct e1000_tx_desc);
2902 	ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
2903 	ew32(TDBAH(0), (tdba >> 32));
2904 	ew32(TDLEN(0), tdlen);
2905 	ew32(TDH(0), 0);
2906 	ew32(TDT(0), 0);
2907 	tx_ring->head = adapter->hw.hw_addr + E1000_TDH(0);
2908 	tx_ring->tail = adapter->hw.hw_addr + E1000_TDT(0);
2909 
2910 	/* Set the Tx Interrupt Delay register */
2911 	ew32(TIDV, adapter->tx_int_delay);
2912 	/* Tx irq moderation */
2913 	ew32(TADV, adapter->tx_abs_int_delay);
2914 
2915 	if (adapter->flags2 & FLAG2_DMA_BURST) {
2916 		u32 txdctl = er32(TXDCTL(0));
2917 		txdctl &= ~(E1000_TXDCTL_PTHRESH | E1000_TXDCTL_HTHRESH |
2918 			    E1000_TXDCTL_WTHRESH);
2919 		/* set up some performance related parameters to encourage the
2920 		 * hardware to use the bus more efficiently in bursts, depends
2921 		 * on the tx_int_delay to be enabled,
2922 		 * wthresh = 1 ==> burst write is disabled to avoid Tx stalls
2923 		 * hthresh = 1 ==> prefetch when one or more available
2924 		 * pthresh = 0x1f ==> prefetch if internal cache 31 or less
2925 		 * BEWARE: this seems to work but should be considered first if
2926 		 * there are Tx hangs or other Tx related bugs
2927 		 */
2928 		txdctl |= E1000_TXDCTL_DMA_BURST_ENABLE;
2929 		ew32(TXDCTL(0), txdctl);
2930 	}
2931 	/* erratum work around: set txdctl the same for both queues */
2932 	ew32(TXDCTL(1), er32(TXDCTL(0)));
2933 
2934 	if (adapter->flags & FLAG_TARC_SPEED_MODE_BIT) {
2935 		tarc = er32(TARC(0));
2936 		/* set the speed mode bit, we'll clear it if we're not at
2937 		 * gigabit link later
2938 		 */
2939 #define SPEED_MODE_BIT (1 << 21)
2940 		tarc |= SPEED_MODE_BIT;
2941 		ew32(TARC(0), tarc);
2942 	}
2943 
2944 	/* errata: program both queues to unweighted RR */
2945 	if (adapter->flags & FLAG_TARC_SET_BIT_ZERO) {
2946 		tarc = er32(TARC(0));
2947 		tarc |= 1;
2948 		ew32(TARC(0), tarc);
2949 		tarc = er32(TARC(1));
2950 		tarc |= 1;
2951 		ew32(TARC(1), tarc);
2952 	}
2953 
2954 	/* Setup Transmit Descriptor Settings for eop descriptor */
2955 	adapter->txd_cmd = E1000_TXD_CMD_EOP | E1000_TXD_CMD_IFCS;
2956 
2957 	/* only set IDE if we are delaying interrupts using the timers */
2958 	if (adapter->tx_int_delay)
2959 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
2960 
2961 	/* enable Report Status bit */
2962 	adapter->txd_cmd |= E1000_TXD_CMD_RS;
2963 
2964 	hw->mac.ops.config_collision_dist(hw);
2965 }
2966 
2967 /**
2968  * e1000_setup_rctl - configure the receive control registers
2969  * @adapter: Board private structure
2970  **/
2971 #define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
2972 			   (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
2973 static void e1000_setup_rctl(struct e1000_adapter *adapter)
2974 {
2975 	struct e1000_hw *hw = &adapter->hw;
2976 	u32 rctl, rfctl;
2977 	u32 pages = 0;
2978 
2979 	/* Workaround Si errata on PCHx - configure jumbo frame flow */
2980 	if ((hw->mac.type >= e1000_pch2lan) &&
2981 	    (adapter->netdev->mtu > ETH_DATA_LEN) &&
2982 	    e1000_lv_jumbo_workaround_ich8lan(hw, true))
2983 		e_dbg("failed to enable jumbo frame workaround mode\n");
2984 
2985 	/* Program MC offset vector base */
2986 	rctl = er32(RCTL);
2987 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
2988 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
2989 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
2990 	    (adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
2991 
2992 	/* Do not Store bad packets */
2993 	rctl &= ~E1000_RCTL_SBP;
2994 
2995 	/* Enable Long Packet receive */
2996 	if (adapter->netdev->mtu <= ETH_DATA_LEN)
2997 		rctl &= ~E1000_RCTL_LPE;
2998 	else
2999 		rctl |= E1000_RCTL_LPE;
3000 
3001 	/* Some systems expect that the CRC is included in SMBUS traffic. The
3002 	 * hardware strips the CRC before sending to both SMBUS (BMC) and to
3003 	 * host memory when this is enabled
3004 	 */
3005 	if (adapter->flags2 & FLAG2_CRC_STRIPPING)
3006 		rctl |= E1000_RCTL_SECRC;
3007 
3008 	/* Workaround Si errata on 82577 PHY - configure IPG for jumbos */
3009 	if ((hw->phy.type == e1000_phy_82577) && (rctl & E1000_RCTL_LPE)) {
3010 		u16 phy_data;
3011 
3012 		e1e_rphy(hw, PHY_REG(770, 26), &phy_data);
3013 		phy_data &= 0xfff8;
3014 		phy_data |= (1 << 2);
3015 		e1e_wphy(hw, PHY_REG(770, 26), phy_data);
3016 
3017 		e1e_rphy(hw, 22, &phy_data);
3018 		phy_data &= 0x0fff;
3019 		phy_data |= (1 << 14);
3020 		e1e_wphy(hw, 0x10, 0x2823);
3021 		e1e_wphy(hw, 0x11, 0x0003);
3022 		e1e_wphy(hw, 22, phy_data);
3023 	}
3024 
3025 	/* Setup buffer sizes */
3026 	rctl &= ~E1000_RCTL_SZ_4096;
3027 	rctl |= E1000_RCTL_BSEX;
3028 	switch (adapter->rx_buffer_len) {
3029 	case 2048:
3030 	default:
3031 		rctl |= E1000_RCTL_SZ_2048;
3032 		rctl &= ~E1000_RCTL_BSEX;
3033 		break;
3034 	case 4096:
3035 		rctl |= E1000_RCTL_SZ_4096;
3036 		break;
3037 	case 8192:
3038 		rctl |= E1000_RCTL_SZ_8192;
3039 		break;
3040 	case 16384:
3041 		rctl |= E1000_RCTL_SZ_16384;
3042 		break;
3043 	}
3044 
3045 	/* Enable Extended Status in all Receive Descriptors */
3046 	rfctl = er32(RFCTL);
3047 	rfctl |= E1000_RFCTL_EXTEN;
3048 	ew32(RFCTL, rfctl);
3049 
3050 	/* 82571 and greater support packet-split where the protocol
3051 	 * header is placed in skb->data and the packet data is
3052 	 * placed in pages hanging off of skb_shinfo(skb)->nr_frags.
3053 	 * In the case of a non-split, skb->data is linearly filled,
3054 	 * followed by the page buffers.  Therefore, skb->data is
3055 	 * sized to hold the largest protocol header.
3056 	 *
3057 	 * allocations using alloc_page take too long for regular MTU
3058 	 * so only enable packet split for jumbo frames
3059 	 *
3060 	 * Using pages when the page size is greater than 16k wastes
3061 	 * a lot of memory, since we allocate 3 pages at all times
3062 	 * per packet.
3063 	 */
3064 	pages = PAGE_USE_COUNT(adapter->netdev->mtu);
3065 	if ((pages <= 3) && (PAGE_SIZE <= 16384) && (rctl & E1000_RCTL_LPE))
3066 		adapter->rx_ps_pages = pages;
3067 	else
3068 		adapter->rx_ps_pages = 0;
3069 
3070 	if (adapter->rx_ps_pages) {
3071 		u32 psrctl = 0;
3072 
3073 		/* Enable Packet split descriptors */
3074 		rctl |= E1000_RCTL_DTYP_PS;
3075 
3076 		psrctl |= adapter->rx_ps_bsize0 >> E1000_PSRCTL_BSIZE0_SHIFT;
3077 
3078 		switch (adapter->rx_ps_pages) {
3079 		case 3:
3080 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE3_SHIFT;
3081 			/* fall-through */
3082 		case 2:
3083 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE2_SHIFT;
3084 			/* fall-through */
3085 		case 1:
3086 			psrctl |= PAGE_SIZE >> E1000_PSRCTL_BSIZE1_SHIFT;
3087 			break;
3088 		}
3089 
3090 		ew32(PSRCTL, psrctl);
3091 	}
3092 
3093 	/* This is useful for sniffing bad packets. */
3094 	if (adapter->netdev->features & NETIF_F_RXALL) {
3095 		/* UPE and MPE will be handled by normal PROMISC logic
3096 		 * in e1000e_set_rx_mode
3097 		 */
3098 		rctl |= (E1000_RCTL_SBP |	/* Receive bad packets */
3099 			 E1000_RCTL_BAM |	/* RX All Bcast Pkts */
3100 			 E1000_RCTL_PMCF);	/* RX All MAC Ctrl Pkts */
3101 
3102 		rctl &= ~(E1000_RCTL_VFE |	/* Disable VLAN filter */
3103 			  E1000_RCTL_DPF |	/* Allow filtered pause */
3104 			  E1000_RCTL_CFIEN);	/* Dis VLAN CFIEN Filter */
3105 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3106 		 * and that breaks VLANs.
3107 		 */
3108 	}
3109 
3110 	ew32(RCTL, rctl);
3111 	/* just started the receive unit, no need to restart */
3112 	adapter->flags &= ~FLAG_RESTART_NOW;
3113 }
3114 
3115 /**
3116  * e1000_configure_rx - Configure Receive Unit after Reset
3117  * @adapter: board private structure
3118  *
3119  * Configure the Rx unit of the MAC after a reset.
3120  **/
3121 static void e1000_configure_rx(struct e1000_adapter *adapter)
3122 {
3123 	struct e1000_hw *hw = &adapter->hw;
3124 	struct e1000_ring *rx_ring = adapter->rx_ring;
3125 	u64 rdba;
3126 	u32 rdlen, rctl, rxcsum, ctrl_ext;
3127 
3128 	if (adapter->rx_ps_pages) {
3129 		/* this is a 32 byte descriptor */
3130 		rdlen = rx_ring->count *
3131 		    sizeof(union e1000_rx_desc_packet_split);
3132 		adapter->clean_rx = e1000_clean_rx_irq_ps;
3133 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers_ps;
3134 	} else if (adapter->netdev->mtu > ETH_FRAME_LEN + ETH_FCS_LEN) {
3135 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3136 		adapter->clean_rx = e1000_clean_jumbo_rx_irq;
3137 		adapter->alloc_rx_buf = e1000_alloc_jumbo_rx_buffers;
3138 	} else {
3139 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3140 		adapter->clean_rx = e1000_clean_rx_irq;
3141 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers;
3142 	}
3143 
3144 	/* disable receives while setting up the descriptors */
3145 	rctl = er32(RCTL);
3146 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
3147 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
3148 	e1e_flush();
3149 	usleep_range(10000, 20000);
3150 
3151 	if (adapter->flags2 & FLAG2_DMA_BURST) {
3152 		/* set the writeback threshold (only takes effect if the RDTR
3153 		 * is set). set GRAN=1 and write back up to 0x4 worth, and
3154 		 * enable prefetching of 0x20 Rx descriptors
3155 		 * granularity = 01
3156 		 * wthresh = 04,
3157 		 * hthresh = 04,
3158 		 * pthresh = 0x20
3159 		 */
3160 		ew32(RXDCTL(0), E1000_RXDCTL_DMA_BURST_ENABLE);
3161 		ew32(RXDCTL(1), E1000_RXDCTL_DMA_BURST_ENABLE);
3162 
3163 		/* override the delay timers for enabling bursting, only if
3164 		 * the value was not set by the user via module options
3165 		 */
3166 		if (adapter->rx_int_delay == DEFAULT_RDTR)
3167 			adapter->rx_int_delay = BURST_RDTR;
3168 		if (adapter->rx_abs_int_delay == DEFAULT_RADV)
3169 			adapter->rx_abs_int_delay = BURST_RADV;
3170 	}
3171 
3172 	/* set the Receive Delay Timer Register */
3173 	ew32(RDTR, adapter->rx_int_delay);
3174 
3175 	/* irq moderation */
3176 	ew32(RADV, adapter->rx_abs_int_delay);
3177 	if ((adapter->itr_setting != 0) && (adapter->itr != 0))
3178 		e1000e_write_itr(adapter, adapter->itr);
3179 
3180 	ctrl_ext = er32(CTRL_EXT);
3181 	/* Auto-Mask interrupts upon ICR access */
3182 	ctrl_ext |= E1000_CTRL_EXT_IAME;
3183 	ew32(IAM, 0xffffffff);
3184 	ew32(CTRL_EXT, ctrl_ext);
3185 	e1e_flush();
3186 
3187 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
3188 	 * the Base and Length of the Rx Descriptor Ring
3189 	 */
3190 	rdba = rx_ring->dma;
3191 	ew32(RDBAL(0), (rdba & DMA_BIT_MASK(32)));
3192 	ew32(RDBAH(0), (rdba >> 32));
3193 	ew32(RDLEN(0), rdlen);
3194 	ew32(RDH(0), 0);
3195 	ew32(RDT(0), 0);
3196 	rx_ring->head = adapter->hw.hw_addr + E1000_RDH(0);
3197 	rx_ring->tail = adapter->hw.hw_addr + E1000_RDT(0);
3198 
3199 	/* Enable Receive Checksum Offload for TCP and UDP */
3200 	rxcsum = er32(RXCSUM);
3201 	if (adapter->netdev->features & NETIF_F_RXCSUM)
3202 		rxcsum |= E1000_RXCSUM_TUOFL;
3203 	else
3204 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3205 	ew32(RXCSUM, rxcsum);
3206 
3207 	/* With jumbo frames, excessive C-state transition latencies result
3208 	 * in dropped transactions.
3209 	 */
3210 	if (adapter->netdev->mtu > ETH_DATA_LEN) {
3211 		u32 lat =
3212 		    ((er32(PBA) & E1000_PBA_RXA_MASK) * 1024 -
3213 		     adapter->max_frame_size) * 8 / 1000;
3214 
3215 		if (adapter->flags & FLAG_IS_ICH) {
3216 			u32 rxdctl = er32(RXDCTL(0));
3217 			ew32(RXDCTL(0), rxdctl | 0x3);
3218 		}
3219 
3220 		pm_qos_update_request(&adapter->netdev->pm_qos_req, lat);
3221 	} else {
3222 		pm_qos_update_request(&adapter->netdev->pm_qos_req,
3223 				      PM_QOS_DEFAULT_VALUE);
3224 	}
3225 
3226 	/* Enable Receives */
3227 	ew32(RCTL, rctl);
3228 }
3229 
3230 /**
3231  * e1000e_write_mc_addr_list - write multicast addresses to MTA
3232  * @netdev: network interface device structure
3233  *
3234  * Writes multicast address list to the MTA hash table.
3235  * Returns: -ENOMEM on failure
3236  *                0 on no addresses written
3237  *                X on writing X addresses to MTA
3238  */
3239 static int e1000e_write_mc_addr_list(struct net_device *netdev)
3240 {
3241 	struct e1000_adapter *adapter = netdev_priv(netdev);
3242 	struct e1000_hw *hw = &adapter->hw;
3243 	struct netdev_hw_addr *ha;
3244 	u8 *mta_list;
3245 	int i;
3246 
3247 	if (netdev_mc_empty(netdev)) {
3248 		/* nothing to program, so clear mc list */
3249 		hw->mac.ops.update_mc_addr_list(hw, NULL, 0);
3250 		return 0;
3251 	}
3252 
3253 	mta_list = kzalloc(netdev_mc_count(netdev) * ETH_ALEN, GFP_ATOMIC);
3254 	if (!mta_list)
3255 		return -ENOMEM;
3256 
3257 	/* update_mc_addr_list expects a packed array of only addresses. */
3258 	i = 0;
3259 	netdev_for_each_mc_addr(ha, netdev)
3260 	    memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3261 
3262 	hw->mac.ops.update_mc_addr_list(hw, mta_list, i);
3263 	kfree(mta_list);
3264 
3265 	return netdev_mc_count(netdev);
3266 }
3267 
3268 /**
3269  * e1000e_write_uc_addr_list - write unicast addresses to RAR table
3270  * @netdev: network interface device structure
3271  *
3272  * Writes unicast address list to the RAR table.
3273  * Returns: -ENOMEM on failure/insufficient address space
3274  *                0 on no addresses written
3275  *                X on writing X addresses to the RAR table
3276  **/
3277 static int e1000e_write_uc_addr_list(struct net_device *netdev)
3278 {
3279 	struct e1000_adapter *adapter = netdev_priv(netdev);
3280 	struct e1000_hw *hw = &adapter->hw;
3281 	unsigned int rar_entries = hw->mac.rar_entry_count;
3282 	int count = 0;
3283 
3284 	/* save a rar entry for our hardware address */
3285 	rar_entries--;
3286 
3287 	/* save a rar entry for the LAA workaround */
3288 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA)
3289 		rar_entries--;
3290 
3291 	/* return ENOMEM indicating insufficient memory for addresses */
3292 	if (netdev_uc_count(netdev) > rar_entries)
3293 		return -ENOMEM;
3294 
3295 	if (!netdev_uc_empty(netdev) && rar_entries) {
3296 		struct netdev_hw_addr *ha;
3297 
3298 		/* write the addresses in reverse order to avoid write
3299 		 * combining
3300 		 */
3301 		netdev_for_each_uc_addr(ha, netdev) {
3302 			if (!rar_entries)
3303 				break;
3304 			hw->mac.ops.rar_set(hw, ha->addr, rar_entries--);
3305 			count++;
3306 		}
3307 	}
3308 
3309 	/* zero out the remaining RAR entries not used above */
3310 	for (; rar_entries > 0; rar_entries--) {
3311 		ew32(RAH(rar_entries), 0);
3312 		ew32(RAL(rar_entries), 0);
3313 	}
3314 	e1e_flush();
3315 
3316 	return count;
3317 }
3318 
3319 /**
3320  * e1000e_set_rx_mode - secondary unicast, Multicast and Promiscuous mode set
3321  * @netdev: network interface device structure
3322  *
3323  * The ndo_set_rx_mode entry point is called whenever the unicast or multicast
3324  * address list or the network interface flags are updated.  This routine is
3325  * responsible for configuring the hardware for proper unicast, multicast,
3326  * promiscuous mode, and all-multi behavior.
3327  **/
3328 static void e1000e_set_rx_mode(struct net_device *netdev)
3329 {
3330 	struct e1000_adapter *adapter = netdev_priv(netdev);
3331 	struct e1000_hw *hw = &adapter->hw;
3332 	u32 rctl;
3333 
3334 	/* Check for Promiscuous and All Multicast modes */
3335 	rctl = er32(RCTL);
3336 
3337 	/* clear the affected bits */
3338 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
3339 
3340 	if (netdev->flags & IFF_PROMISC) {
3341 		rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
3342 		/* Do not hardware filter VLANs in promisc mode */
3343 		e1000e_vlan_filter_disable(adapter);
3344 	} else {
3345 		int count;
3346 
3347 		if (netdev->flags & IFF_ALLMULTI) {
3348 			rctl |= E1000_RCTL_MPE;
3349 		} else {
3350 			/* Write addresses to the MTA, if the attempt fails
3351 			 * then we should just turn on promiscuous mode so
3352 			 * that we can at least receive multicast traffic
3353 			 */
3354 			count = e1000e_write_mc_addr_list(netdev);
3355 			if (count < 0)
3356 				rctl |= E1000_RCTL_MPE;
3357 		}
3358 		e1000e_vlan_filter_enable(adapter);
3359 		/* Write addresses to available RAR registers, if there is not
3360 		 * sufficient space to store all the addresses then enable
3361 		 * unicast promiscuous mode
3362 		 */
3363 		count = e1000e_write_uc_addr_list(netdev);
3364 		if (count < 0)
3365 			rctl |= E1000_RCTL_UPE;
3366 	}
3367 
3368 	ew32(RCTL, rctl);
3369 
3370 	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
3371 		e1000e_vlan_strip_enable(adapter);
3372 	else
3373 		e1000e_vlan_strip_disable(adapter);
3374 }
3375 
3376 static void e1000e_setup_rss_hash(struct e1000_adapter *adapter)
3377 {
3378 	struct e1000_hw *hw = &adapter->hw;
3379 	u32 mrqc, rxcsum;
3380 	int i;
3381 	static const u32 rsskey[10] = {
3382 		0xda565a6d, 0xc20e5b25, 0x3d256741, 0xb08fa343, 0xcb2bcad0,
3383 		0xb4307bae, 0xa32dcb77, 0x0cf23080, 0x3bb7426a, 0xfa01acbe
3384 	};
3385 
3386 	/* Fill out hash function seed */
3387 	for (i = 0; i < 10; i++)
3388 		ew32(RSSRK(i), rsskey[i]);
3389 
3390 	/* Direct all traffic to queue 0 */
3391 	for (i = 0; i < 32; i++)
3392 		ew32(RETA(i), 0);
3393 
3394 	/* Disable raw packet checksumming so that RSS hash is placed in
3395 	 * descriptor on writeback.
3396 	 */
3397 	rxcsum = er32(RXCSUM);
3398 	rxcsum |= E1000_RXCSUM_PCSD;
3399 
3400 	ew32(RXCSUM, rxcsum);
3401 
3402 	mrqc = (E1000_MRQC_RSS_FIELD_IPV4 |
3403 		E1000_MRQC_RSS_FIELD_IPV4_TCP |
3404 		E1000_MRQC_RSS_FIELD_IPV6 |
3405 		E1000_MRQC_RSS_FIELD_IPV6_TCP |
3406 		E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3407 
3408 	ew32(MRQC, mrqc);
3409 }
3410 
3411 /**
3412  * e1000e_get_base_timinca - get default SYSTIM time increment attributes
3413  * @adapter: board private structure
3414  * @timinca: pointer to returned time increment attributes
3415  *
3416  * Get attributes for incrementing the System Time Register SYSTIML/H at
3417  * the default base frequency, and set the cyclecounter shift value.
3418  **/
3419 s32 e1000e_get_base_timinca(struct e1000_adapter *adapter, u32 *timinca)
3420 {
3421 	struct e1000_hw *hw = &adapter->hw;
3422 	u32 incvalue, incperiod, shift;
3423 
3424 	/* Make sure clock is enabled on I217 before checking the frequency */
3425 	if ((hw->mac.type == e1000_pch_lpt) &&
3426 	    !(er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) &&
3427 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_ENABLED)) {
3428 		u32 fextnvm7 = er32(FEXTNVM7);
3429 
3430 		if (!(fextnvm7 & (1 << 0))) {
3431 			ew32(FEXTNVM7, fextnvm7 | (1 << 0));
3432 			e1e_flush();
3433 		}
3434 	}
3435 
3436 	switch (hw->mac.type) {
3437 	case e1000_pch2lan:
3438 	case e1000_pch_lpt:
3439 		/* On I217, the clock frequency is 25MHz or 96MHz as
3440 		 * indicated by the System Clock Frequency Indication
3441 		 */
3442 		if ((hw->mac.type != e1000_pch_lpt) ||
3443 		    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI)) {
3444 			/* Stable 96MHz frequency */
3445 			incperiod = INCPERIOD_96MHz;
3446 			incvalue = INCVALUE_96MHz;
3447 			shift = INCVALUE_SHIFT_96MHz;
3448 			adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHz;
3449 			break;
3450 		}
3451 		/* fall-through */
3452 	case e1000_82574:
3453 	case e1000_82583:
3454 		/* Stable 25MHz frequency */
3455 		incperiod = INCPERIOD_25MHz;
3456 		incvalue = INCVALUE_25MHz;
3457 		shift = INCVALUE_SHIFT_25MHz;
3458 		adapter->cc.shift = shift;
3459 		break;
3460 	default:
3461 		return -EINVAL;
3462 	}
3463 
3464 	*timinca = ((incperiod << E1000_TIMINCA_INCPERIOD_SHIFT) |
3465 		    ((incvalue << shift) & E1000_TIMINCA_INCVALUE_MASK));
3466 
3467 	return 0;
3468 }
3469 
3470 /**
3471  * e1000e_config_hwtstamp - configure the hwtstamp registers and enable/disable
3472  * @adapter: board private structure
3473  *
3474  * Outgoing time stamping can be enabled and disabled. Play nice and
3475  * disable it when requested, although it shouldn't cause any overhead
3476  * when no packet needs it. At most one packet in the queue may be
3477  * marked for time stamping, otherwise it would be impossible to tell
3478  * for sure to which packet the hardware time stamp belongs.
3479  *
3480  * Incoming time stamping has to be configured via the hardware filters.
3481  * Not all combinations are supported, in particular event type has to be
3482  * specified. Matching the kind of event packet is not supported, with the
3483  * exception of "all V2 events regardless of level 2 or 4".
3484  **/
3485 static int e1000e_config_hwtstamp(struct e1000_adapter *adapter,
3486 				  struct hwtstamp_config *config)
3487 {
3488 	struct e1000_hw *hw = &adapter->hw;
3489 	u32 tsync_tx_ctl = E1000_TSYNCTXCTL_ENABLED;
3490 	u32 tsync_rx_ctl = E1000_TSYNCRXCTL_ENABLED;
3491 	u32 rxmtrl = 0;
3492 	u16 rxudp = 0;
3493 	bool is_l4 = false;
3494 	bool is_l2 = false;
3495 	u32 regval;
3496 	s32 ret_val;
3497 
3498 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3499 		return -EINVAL;
3500 
3501 	/* flags reserved for future extensions - must be zero */
3502 	if (config->flags)
3503 		return -EINVAL;
3504 
3505 	switch (config->tx_type) {
3506 	case HWTSTAMP_TX_OFF:
3507 		tsync_tx_ctl = 0;
3508 		break;
3509 	case HWTSTAMP_TX_ON:
3510 		break;
3511 	default:
3512 		return -ERANGE;
3513 	}
3514 
3515 	switch (config->rx_filter) {
3516 	case HWTSTAMP_FILTER_NONE:
3517 		tsync_rx_ctl = 0;
3518 		break;
3519 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3520 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3521 		rxmtrl = E1000_RXMTRL_PTP_V1_SYNC_MESSAGE;
3522 		is_l4 = true;
3523 		break;
3524 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3525 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3526 		rxmtrl = E1000_RXMTRL_PTP_V1_DELAY_REQ_MESSAGE;
3527 		is_l4 = true;
3528 		break;
3529 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3530 		/* Also time stamps V2 L2 Path Delay Request/Response */
3531 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3532 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3533 		is_l2 = true;
3534 		break;
3535 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3536 		/* Also time stamps V2 L2 Path Delay Request/Response. */
3537 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3538 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3539 		is_l2 = true;
3540 		break;
3541 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3542 		/* Hardware cannot filter just V2 L4 Sync messages;
3543 		 * fall-through to V2 (both L2 and L4) Sync.
3544 		 */
3545 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
3546 		/* Also time stamps V2 Path Delay Request/Response. */
3547 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3548 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3549 		is_l2 = true;
3550 		is_l4 = true;
3551 		break;
3552 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3553 		/* Hardware cannot filter just V2 L4 Delay Request messages;
3554 		 * fall-through to V2 (both L2 and L4) Delay Request.
3555 		 */
3556 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
3557 		/* Also time stamps V2 Path Delay Request/Response. */
3558 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3559 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3560 		is_l2 = true;
3561 		is_l4 = true;
3562 		break;
3563 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
3564 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
3565 		/* Hardware cannot filter just V2 L4 or L2 Event messages;
3566 		 * fall-through to all V2 (both L2 and L4) Events.
3567 		 */
3568 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
3569 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_EVENT_V2;
3570 		config->rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
3571 		is_l2 = true;
3572 		is_l4 = true;
3573 		break;
3574 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3575 		/* For V1, the hardware can only filter Sync messages or
3576 		 * Delay Request messages but not both so fall-through to
3577 		 * time stamp all packets.
3578 		 */
3579 	case HWTSTAMP_FILTER_ALL:
3580 		is_l2 = true;
3581 		is_l4 = true;
3582 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_ALL;
3583 		config->rx_filter = HWTSTAMP_FILTER_ALL;
3584 		break;
3585 	default:
3586 		return -ERANGE;
3587 	}
3588 
3589 	adapter->hwtstamp_config = *config;
3590 
3591 	/* enable/disable Tx h/w time stamping */
3592 	regval = er32(TSYNCTXCTL);
3593 	regval &= ~E1000_TSYNCTXCTL_ENABLED;
3594 	regval |= tsync_tx_ctl;
3595 	ew32(TSYNCTXCTL, regval);
3596 	if ((er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) !=
3597 	    (regval & E1000_TSYNCTXCTL_ENABLED)) {
3598 		e_err("Timesync Tx Control register not set as expected\n");
3599 		return -EAGAIN;
3600 	}
3601 
3602 	/* enable/disable Rx h/w time stamping */
3603 	regval = er32(TSYNCRXCTL);
3604 	regval &= ~(E1000_TSYNCRXCTL_ENABLED | E1000_TSYNCRXCTL_TYPE_MASK);
3605 	regval |= tsync_rx_ctl;
3606 	ew32(TSYNCRXCTL, regval);
3607 	if ((er32(TSYNCRXCTL) & (E1000_TSYNCRXCTL_ENABLED |
3608 				 E1000_TSYNCRXCTL_TYPE_MASK)) !=
3609 	    (regval & (E1000_TSYNCRXCTL_ENABLED |
3610 		       E1000_TSYNCRXCTL_TYPE_MASK))) {
3611 		e_err("Timesync Rx Control register not set as expected\n");
3612 		return -EAGAIN;
3613 	}
3614 
3615 	/* L2: define ethertype filter for time stamped packets */
3616 	if (is_l2)
3617 		rxmtrl |= ETH_P_1588;
3618 
3619 	/* define which PTP packets get time stamped */
3620 	ew32(RXMTRL, rxmtrl);
3621 
3622 	/* Filter by destination port */
3623 	if (is_l4) {
3624 		rxudp = PTP_EV_PORT;
3625 		cpu_to_be16s(&rxudp);
3626 	}
3627 	ew32(RXUDP, rxudp);
3628 
3629 	e1e_flush();
3630 
3631 	/* Clear TSYNCRXCTL_VALID & TSYNCTXCTL_VALID bit */
3632 	er32(RXSTMPH);
3633 	er32(TXSTMPH);
3634 
3635 	/* Get and set the System Time Register SYSTIM base frequency */
3636 	ret_val = e1000e_get_base_timinca(adapter, &regval);
3637 	if (ret_val)
3638 		return ret_val;
3639 	ew32(TIMINCA, regval);
3640 
3641 	/* reset the ns time counter */
3642 	timecounter_init(&adapter->tc, &adapter->cc,
3643 			 ktime_to_ns(ktime_get_real()));
3644 
3645 	return 0;
3646 }
3647 
3648 /**
3649  * e1000_configure - configure the hardware for Rx and Tx
3650  * @adapter: private board structure
3651  **/
3652 static void e1000_configure(struct e1000_adapter *adapter)
3653 {
3654 	struct e1000_ring *rx_ring = adapter->rx_ring;
3655 
3656 	e1000e_set_rx_mode(adapter->netdev);
3657 
3658 	e1000_restore_vlan(adapter);
3659 	e1000_init_manageability_pt(adapter);
3660 
3661 	e1000_configure_tx(adapter);
3662 
3663 	if (adapter->netdev->features & NETIF_F_RXHASH)
3664 		e1000e_setup_rss_hash(adapter);
3665 	e1000_setup_rctl(adapter);
3666 	e1000_configure_rx(adapter);
3667 	adapter->alloc_rx_buf(rx_ring, e1000_desc_unused(rx_ring), GFP_KERNEL);
3668 }
3669 
3670 /**
3671  * e1000e_power_up_phy - restore link in case the phy was powered down
3672  * @adapter: address of board private structure
3673  *
3674  * The phy may be powered down to save power and turn off link when the
3675  * driver is unloaded and wake on lan is not enabled (among others)
3676  * *** this routine MUST be followed by a call to e1000e_reset ***
3677  **/
3678 void e1000e_power_up_phy(struct e1000_adapter *adapter)
3679 {
3680 	if (adapter->hw.phy.ops.power_up)
3681 		adapter->hw.phy.ops.power_up(&adapter->hw);
3682 
3683 	adapter->hw.mac.ops.setup_link(&adapter->hw);
3684 }
3685 
3686 /**
3687  * e1000_power_down_phy - Power down the PHY
3688  *
3689  * Power down the PHY so no link is implied when interface is down.
3690  * The PHY cannot be powered down if management or WoL is active.
3691  */
3692 static void e1000_power_down_phy(struct e1000_adapter *adapter)
3693 {
3694 	/* WoL is enabled */
3695 	if (adapter->wol)
3696 		return;
3697 
3698 	if (adapter->hw.phy.ops.power_down)
3699 		adapter->hw.phy.ops.power_down(&adapter->hw);
3700 }
3701 
3702 /**
3703  * e1000e_reset - bring the hardware into a known good state
3704  *
3705  * This function boots the hardware and enables some settings that
3706  * require a configuration cycle of the hardware - those cannot be
3707  * set/changed during runtime. After reset the device needs to be
3708  * properly configured for Rx, Tx etc.
3709  */
3710 void e1000e_reset(struct e1000_adapter *adapter)
3711 {
3712 	struct e1000_mac_info *mac = &adapter->hw.mac;
3713 	struct e1000_fc_info *fc = &adapter->hw.fc;
3714 	struct e1000_hw *hw = &adapter->hw;
3715 	u32 tx_space, min_tx_space, min_rx_space;
3716 	u32 pba = adapter->pba;
3717 	u16 hwm;
3718 
3719 	/* reset Packet Buffer Allocation to default */
3720 	ew32(PBA, pba);
3721 
3722 	if (adapter->max_frame_size > ETH_FRAME_LEN + ETH_FCS_LEN) {
3723 		/* To maintain wire speed transmits, the Tx FIFO should be
3724 		 * large enough to accommodate two full transmit packets,
3725 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
3726 		 * the Rx FIFO should be large enough to accommodate at least
3727 		 * one full receive packet and is similarly rounded up and
3728 		 * expressed in KB.
3729 		 */
3730 		pba = er32(PBA);
3731 		/* upper 16 bits has Tx packet buffer allocation size in KB */
3732 		tx_space = pba >> 16;
3733 		/* lower 16 bits has Rx packet buffer allocation size in KB */
3734 		pba &= 0xffff;
3735 		/* the Tx fifo also stores 16 bytes of information about the Tx
3736 		 * but don't include ethernet FCS because hardware appends it
3737 		 */
3738 		min_tx_space = (adapter->max_frame_size +
3739 				sizeof(struct e1000_tx_desc) - ETH_FCS_LEN) * 2;
3740 		min_tx_space = ALIGN(min_tx_space, 1024);
3741 		min_tx_space >>= 10;
3742 		/* software strips receive CRC, so leave room for it */
3743 		min_rx_space = adapter->max_frame_size;
3744 		min_rx_space = ALIGN(min_rx_space, 1024);
3745 		min_rx_space >>= 10;
3746 
3747 		/* If current Tx allocation is less than the min Tx FIFO size,
3748 		 * and the min Tx FIFO size is less than the current Rx FIFO
3749 		 * allocation, take space away from current Rx allocation
3750 		 */
3751 		if ((tx_space < min_tx_space) &&
3752 		    ((min_tx_space - tx_space) < pba)) {
3753 			pba -= min_tx_space - tx_space;
3754 
3755 			/* if short on Rx space, Rx wins and must trump Tx
3756 			 * adjustment
3757 			 */
3758 			if (pba < min_rx_space)
3759 				pba = min_rx_space;
3760 		}
3761 
3762 		ew32(PBA, pba);
3763 	}
3764 
3765 	/* flow control settings
3766 	 *
3767 	 * The high water mark must be low enough to fit one full frame
3768 	 * (or the size used for early receive) above it in the Rx FIFO.
3769 	 * Set it to the lower of:
3770 	 * - 90% of the Rx FIFO size, and
3771 	 * - the full Rx FIFO size minus one full frame
3772 	 */
3773 	if (adapter->flags & FLAG_DISABLE_FC_PAUSE_TIME)
3774 		fc->pause_time = 0xFFFF;
3775 	else
3776 		fc->pause_time = E1000_FC_PAUSE_TIME;
3777 	fc->send_xon = true;
3778 	fc->current_mode = fc->requested_mode;
3779 
3780 	switch (hw->mac.type) {
3781 	case e1000_ich9lan:
3782 	case e1000_ich10lan:
3783 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
3784 			pba = 14;
3785 			ew32(PBA, pba);
3786 			fc->high_water = 0x2800;
3787 			fc->low_water = fc->high_water - 8;
3788 			break;
3789 		}
3790 		/* fall-through */
3791 	default:
3792 		hwm = min(((pba << 10) * 9 / 10),
3793 			  ((pba << 10) - adapter->max_frame_size));
3794 
3795 		fc->high_water = hwm & E1000_FCRTH_RTH;	/* 8-byte granularity */
3796 		fc->low_water = fc->high_water - 8;
3797 		break;
3798 	case e1000_pchlan:
3799 		/* Workaround PCH LOM adapter hangs with certain network
3800 		 * loads.  If hangs persist, try disabling Tx flow control.
3801 		 */
3802 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
3803 			fc->high_water = 0x3500;
3804 			fc->low_water = 0x1500;
3805 		} else {
3806 			fc->high_water = 0x5000;
3807 			fc->low_water = 0x3000;
3808 		}
3809 		fc->refresh_time = 0x1000;
3810 		break;
3811 	case e1000_pch2lan:
3812 	case e1000_pch_lpt:
3813 		fc->refresh_time = 0x0400;
3814 
3815 		if (adapter->netdev->mtu <= ETH_DATA_LEN) {
3816 			fc->high_water = 0x05C20;
3817 			fc->low_water = 0x05048;
3818 			fc->pause_time = 0x0650;
3819 			break;
3820 		}
3821 
3822 		pba = 14;
3823 		ew32(PBA, pba);
3824 		fc->high_water = ((pba << 10) * 9 / 10) & E1000_FCRTH_RTH;
3825 		fc->low_water = ((pba << 10) * 8 / 10) & E1000_FCRTL_RTL;
3826 		break;
3827 	}
3828 
3829 	/* Alignment of Tx data is on an arbitrary byte boundary with the
3830 	 * maximum size per Tx descriptor limited only to the transmit
3831 	 * allocation of the packet buffer minus 96 bytes with an upper
3832 	 * limit of 24KB due to receive synchronization limitations.
3833 	 */
3834 	adapter->tx_fifo_limit = min_t(u32, ((er32(PBA) >> 16) << 10) - 96,
3835 				       24 << 10);
3836 
3837 	/* Disable Adaptive Interrupt Moderation if 2 full packets cannot
3838 	 * fit in receive buffer.
3839 	 */
3840 	if (adapter->itr_setting & 0x3) {
3841 		if ((adapter->max_frame_size * 2) > (pba << 10)) {
3842 			if (!(adapter->flags2 & FLAG2_DISABLE_AIM)) {
3843 				dev_info(&adapter->pdev->dev,
3844 					 "Interrupt Throttle Rate off\n");
3845 				adapter->flags2 |= FLAG2_DISABLE_AIM;
3846 				e1000e_write_itr(adapter, 0);
3847 			}
3848 		} else if (adapter->flags2 & FLAG2_DISABLE_AIM) {
3849 			dev_info(&adapter->pdev->dev,
3850 				 "Interrupt Throttle Rate on\n");
3851 			adapter->flags2 &= ~FLAG2_DISABLE_AIM;
3852 			adapter->itr = 20000;
3853 			e1000e_write_itr(adapter, adapter->itr);
3854 		}
3855 	}
3856 
3857 	/* Allow time for pending master requests to run */
3858 	mac->ops.reset_hw(hw);
3859 
3860 	/* For parts with AMT enabled, let the firmware know
3861 	 * that the network interface is in control
3862 	 */
3863 	if (adapter->flags & FLAG_HAS_AMT)
3864 		e1000e_get_hw_control(adapter);
3865 
3866 	ew32(WUC, 0);
3867 
3868 	if (mac->ops.init_hw(hw))
3869 		e_err("Hardware Error\n");
3870 
3871 	e1000_update_mng_vlan(adapter);
3872 
3873 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
3874 	ew32(VET, ETH_P_8021Q);
3875 
3876 	e1000e_reset_adaptive(hw);
3877 
3878 	/* initialize systim and reset the ns time counter */
3879 	e1000e_config_hwtstamp(adapter, &adapter->hwtstamp_config);
3880 
3881 	/* Set EEE advertisement as appropriate */
3882 	if (adapter->flags2 & FLAG2_HAS_EEE) {
3883 		s32 ret_val;
3884 		u16 adv_addr;
3885 
3886 		switch (hw->phy.type) {
3887 		case e1000_phy_82579:
3888 			adv_addr = I82579_EEE_ADVERTISEMENT;
3889 			break;
3890 		case e1000_phy_i217:
3891 			adv_addr = I217_EEE_ADVERTISEMENT;
3892 			break;
3893 		default:
3894 			dev_err(&adapter->pdev->dev,
3895 				"Invalid PHY type setting EEE advertisement\n");
3896 			return;
3897 		}
3898 
3899 		ret_val = hw->phy.ops.acquire(hw);
3900 		if (ret_val) {
3901 			dev_err(&adapter->pdev->dev,
3902 				"EEE advertisement - unable to acquire PHY\n");
3903 			return;
3904 		}
3905 
3906 		e1000_write_emi_reg_locked(hw, adv_addr,
3907 					   hw->dev_spec.ich8lan.eee_disable ?
3908 					   0 : adapter->eee_advert);
3909 
3910 		hw->phy.ops.release(hw);
3911 	}
3912 
3913 	if (!netif_running(adapter->netdev) &&
3914 	    !test_bit(__E1000_TESTING, &adapter->state)) {
3915 		e1000_power_down_phy(adapter);
3916 		return;
3917 	}
3918 
3919 	e1000_get_phy_info(hw);
3920 
3921 	if ((adapter->flags & FLAG_HAS_SMART_POWER_DOWN) &&
3922 	    !(adapter->flags & FLAG_SMART_POWER_DOWN)) {
3923 		u16 phy_data = 0;
3924 		/* speed up time to link by disabling smart power down, ignore
3925 		 * the return value of this function because there is nothing
3926 		 * different we would do if it failed
3927 		 */
3928 		e1e_rphy(hw, IGP02E1000_PHY_POWER_MGMT, &phy_data);
3929 		phy_data &= ~IGP02E1000_PM_SPD;
3930 		e1e_wphy(hw, IGP02E1000_PHY_POWER_MGMT, phy_data);
3931 	}
3932 }
3933 
3934 int e1000e_up(struct e1000_adapter *adapter)
3935 {
3936 	struct e1000_hw *hw = &adapter->hw;
3937 
3938 	/* hardware has been reset, we need to reload some things */
3939 	e1000_configure(adapter);
3940 
3941 	clear_bit(__E1000_DOWN, &adapter->state);
3942 
3943 	if (adapter->msix_entries)
3944 		e1000_configure_msix(adapter);
3945 	e1000_irq_enable(adapter);
3946 
3947 	netif_start_queue(adapter->netdev);
3948 
3949 	/* fire a link change interrupt to start the watchdog */
3950 	if (adapter->msix_entries)
3951 		ew32(ICS, E1000_ICS_LSC | E1000_ICR_OTHER);
3952 	else
3953 		ew32(ICS, E1000_ICS_LSC);
3954 
3955 	return 0;
3956 }
3957 
3958 static void e1000e_flush_descriptors(struct e1000_adapter *adapter)
3959 {
3960 	struct e1000_hw *hw = &adapter->hw;
3961 
3962 	if (!(adapter->flags2 & FLAG2_DMA_BURST))
3963 		return;
3964 
3965 	/* flush pending descriptor writebacks to memory */
3966 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
3967 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
3968 
3969 	/* execute the writes immediately */
3970 	e1e_flush();
3971 
3972 	/* due to rare timing issues, write to TIDV/RDTR again to ensure the
3973 	 * write is successful
3974 	 */
3975 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
3976 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
3977 
3978 	/* execute the writes immediately */
3979 	e1e_flush();
3980 }
3981 
3982 static void e1000e_update_stats(struct e1000_adapter *adapter);
3983 
3984 void e1000e_down(struct e1000_adapter *adapter)
3985 {
3986 	struct net_device *netdev = adapter->netdev;
3987 	struct e1000_hw *hw = &adapter->hw;
3988 	u32 tctl, rctl;
3989 
3990 	/* signal that we're down so the interrupt handler does not
3991 	 * reschedule our watchdog timer
3992 	 */
3993 	set_bit(__E1000_DOWN, &adapter->state);
3994 
3995 	/* disable receives in the hardware */
3996 	rctl = er32(RCTL);
3997 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
3998 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
3999 	/* flush and sleep below */
4000 
4001 	netif_stop_queue(netdev);
4002 
4003 	/* disable transmits in the hardware */
4004 	tctl = er32(TCTL);
4005 	tctl &= ~E1000_TCTL_EN;
4006 	ew32(TCTL, tctl);
4007 
4008 	/* flush both disables and wait for them to finish */
4009 	e1e_flush();
4010 	usleep_range(10000, 20000);
4011 
4012 	e1000_irq_disable(adapter);
4013 
4014 	napi_synchronize(&adapter->napi);
4015 
4016 	del_timer_sync(&adapter->watchdog_timer);
4017 	del_timer_sync(&adapter->phy_info_timer);
4018 
4019 	netif_carrier_off(netdev);
4020 
4021 	spin_lock(&adapter->stats64_lock);
4022 	e1000e_update_stats(adapter);
4023 	spin_unlock(&adapter->stats64_lock);
4024 
4025 	e1000e_flush_descriptors(adapter);
4026 	e1000_clean_tx_ring(adapter->tx_ring);
4027 	e1000_clean_rx_ring(adapter->rx_ring);
4028 
4029 	adapter->link_speed = 0;
4030 	adapter->link_duplex = 0;
4031 
4032 	/* Disable Si errata workaround on PCHx for jumbo frame flow */
4033 	if ((hw->mac.type >= e1000_pch2lan) &&
4034 	    (adapter->netdev->mtu > ETH_DATA_LEN) &&
4035 	    e1000_lv_jumbo_workaround_ich8lan(hw, false))
4036 		e_dbg("failed to disable jumbo frame workaround mode\n");
4037 
4038 	if (!pci_channel_offline(adapter->pdev))
4039 		e1000e_reset(adapter);
4040 
4041 	/* TODO: for power management, we could drop the link and
4042 	 * pci_disable_device here.
4043 	 */
4044 }
4045 
4046 void e1000e_reinit_locked(struct e1000_adapter *adapter)
4047 {
4048 	might_sleep();
4049 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
4050 		usleep_range(1000, 2000);
4051 	e1000e_down(adapter);
4052 	e1000e_up(adapter);
4053 	clear_bit(__E1000_RESETTING, &adapter->state);
4054 }
4055 
4056 /**
4057  * e1000e_cyclecounter_read - read raw cycle counter (used by time counter)
4058  * @cc: cyclecounter structure
4059  **/
4060 static cycle_t e1000e_cyclecounter_read(const struct cyclecounter *cc)
4061 {
4062 	struct e1000_adapter *adapter = container_of(cc, struct e1000_adapter,
4063 						     cc);
4064 	struct e1000_hw *hw = &adapter->hw;
4065 	cycle_t systim;
4066 
4067 	/* latch SYSTIMH on read of SYSTIML */
4068 	systim = (cycle_t)er32(SYSTIML);
4069 	systim |= (cycle_t)er32(SYSTIMH) << 32;
4070 
4071 	return systim;
4072 }
4073 
4074 /**
4075  * e1000_sw_init - Initialize general software structures (struct e1000_adapter)
4076  * @adapter: board private structure to initialize
4077  *
4078  * e1000_sw_init initializes the Adapter private data structure.
4079  * Fields are initialized based on PCI device information and
4080  * OS network device settings (MTU size).
4081  **/
4082 static int e1000_sw_init(struct e1000_adapter *adapter)
4083 {
4084 	struct net_device *netdev = adapter->netdev;
4085 
4086 	adapter->rx_buffer_len = ETH_FRAME_LEN + VLAN_HLEN + ETH_FCS_LEN;
4087 	adapter->rx_ps_bsize0 = 128;
4088 	adapter->max_frame_size = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4089 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4090 	adapter->tx_ring_count = E1000_DEFAULT_TXD;
4091 	adapter->rx_ring_count = E1000_DEFAULT_RXD;
4092 
4093 	spin_lock_init(&adapter->stats64_lock);
4094 
4095 	e1000e_set_interrupt_capability(adapter);
4096 
4097 	if (e1000_alloc_queues(adapter))
4098 		return -ENOMEM;
4099 
4100 	/* Setup hardware time stamping cyclecounter */
4101 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
4102 		adapter->cc.read = e1000e_cyclecounter_read;
4103 		adapter->cc.mask = CLOCKSOURCE_MASK(64);
4104 		adapter->cc.mult = 1;
4105 		/* cc.shift set in e1000e_get_base_tininca() */
4106 
4107 		spin_lock_init(&adapter->systim_lock);
4108 		INIT_WORK(&adapter->tx_hwtstamp_work, e1000e_tx_hwtstamp_work);
4109 	}
4110 
4111 	/* Explicitly disable IRQ since the NIC can be in any state. */
4112 	e1000_irq_disable(adapter);
4113 
4114 	set_bit(__E1000_DOWN, &adapter->state);
4115 	return 0;
4116 }
4117 
4118 /**
4119  * e1000_intr_msi_test - Interrupt Handler
4120  * @irq: interrupt number
4121  * @data: pointer to a network interface device structure
4122  **/
4123 static irqreturn_t e1000_intr_msi_test(int __always_unused irq, void *data)
4124 {
4125 	struct net_device *netdev = data;
4126 	struct e1000_adapter *adapter = netdev_priv(netdev);
4127 	struct e1000_hw *hw = &adapter->hw;
4128 	u32 icr = er32(ICR);
4129 
4130 	e_dbg("icr is %08X\n", icr);
4131 	if (icr & E1000_ICR_RXSEQ) {
4132 		adapter->flags &= ~FLAG_MSI_TEST_FAILED;
4133 		/* Force memory writes to complete before acknowledging the
4134 		 * interrupt is handled.
4135 		 */
4136 		wmb();
4137 	}
4138 
4139 	return IRQ_HANDLED;
4140 }
4141 
4142 /**
4143  * e1000_test_msi_interrupt - Returns 0 for successful test
4144  * @adapter: board private struct
4145  *
4146  * code flow taken from tg3.c
4147  **/
4148 static int e1000_test_msi_interrupt(struct e1000_adapter *adapter)
4149 {
4150 	struct net_device *netdev = adapter->netdev;
4151 	struct e1000_hw *hw = &adapter->hw;
4152 	int err;
4153 
4154 	/* poll_enable hasn't been called yet, so don't need disable */
4155 	/* clear any pending events */
4156 	er32(ICR);
4157 
4158 	/* free the real vector and request a test handler */
4159 	e1000_free_irq(adapter);
4160 	e1000e_reset_interrupt_capability(adapter);
4161 
4162 	/* Assume that the test fails, if it succeeds then the test
4163 	 * MSI irq handler will unset this flag
4164 	 */
4165 	adapter->flags |= FLAG_MSI_TEST_FAILED;
4166 
4167 	err = pci_enable_msi(adapter->pdev);
4168 	if (err)
4169 		goto msi_test_failed;
4170 
4171 	err = request_irq(adapter->pdev->irq, e1000_intr_msi_test, 0,
4172 			  netdev->name, netdev);
4173 	if (err) {
4174 		pci_disable_msi(adapter->pdev);
4175 		goto msi_test_failed;
4176 	}
4177 
4178 	/* Force memory writes to complete before enabling and firing an
4179 	 * interrupt.
4180 	 */
4181 	wmb();
4182 
4183 	e1000_irq_enable(adapter);
4184 
4185 	/* fire an unusual interrupt on the test handler */
4186 	ew32(ICS, E1000_ICS_RXSEQ);
4187 	e1e_flush();
4188 	msleep(100);
4189 
4190 	e1000_irq_disable(adapter);
4191 
4192 	rmb();			/* read flags after interrupt has been fired */
4193 
4194 	if (adapter->flags & FLAG_MSI_TEST_FAILED) {
4195 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
4196 		e_info("MSI interrupt test failed, using legacy interrupt.\n");
4197 	} else {
4198 		e_dbg("MSI interrupt test succeeded!\n");
4199 	}
4200 
4201 	free_irq(adapter->pdev->irq, netdev);
4202 	pci_disable_msi(adapter->pdev);
4203 
4204 msi_test_failed:
4205 	e1000e_set_interrupt_capability(adapter);
4206 	return e1000_request_irq(adapter);
4207 }
4208 
4209 /**
4210  * e1000_test_msi - Returns 0 if MSI test succeeds or INTx mode is restored
4211  * @adapter: board private struct
4212  *
4213  * code flow taken from tg3.c, called with e1000 interrupts disabled.
4214  **/
4215 static int e1000_test_msi(struct e1000_adapter *adapter)
4216 {
4217 	int err;
4218 	u16 pci_cmd;
4219 
4220 	if (!(adapter->flags & FLAG_MSI_ENABLED))
4221 		return 0;
4222 
4223 	/* disable SERR in case the MSI write causes a master abort */
4224 	pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4225 	if (pci_cmd & PCI_COMMAND_SERR)
4226 		pci_write_config_word(adapter->pdev, PCI_COMMAND,
4227 				      pci_cmd & ~PCI_COMMAND_SERR);
4228 
4229 	err = e1000_test_msi_interrupt(adapter);
4230 
4231 	/* re-enable SERR */
4232 	if (pci_cmd & PCI_COMMAND_SERR) {
4233 		pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4234 		pci_cmd |= PCI_COMMAND_SERR;
4235 		pci_write_config_word(adapter->pdev, PCI_COMMAND, pci_cmd);
4236 	}
4237 
4238 	return err;
4239 }
4240 
4241 /**
4242  * e1000_open - Called when a network interface is made active
4243  * @netdev: network interface device structure
4244  *
4245  * Returns 0 on success, negative value on failure
4246  *
4247  * The open entry point is called when a network interface is made
4248  * active by the system (IFF_UP).  At this point all resources needed
4249  * for transmit and receive operations are allocated, the interrupt
4250  * handler is registered with the OS, the watchdog timer is started,
4251  * and the stack is notified that the interface is ready.
4252  **/
4253 static int e1000_open(struct net_device *netdev)
4254 {
4255 	struct e1000_adapter *adapter = netdev_priv(netdev);
4256 	struct e1000_hw *hw = &adapter->hw;
4257 	struct pci_dev *pdev = adapter->pdev;
4258 	int err;
4259 
4260 	/* disallow open during test */
4261 	if (test_bit(__E1000_TESTING, &adapter->state))
4262 		return -EBUSY;
4263 
4264 	pm_runtime_get_sync(&pdev->dev);
4265 
4266 	netif_carrier_off(netdev);
4267 
4268 	/* allocate transmit descriptors */
4269 	err = e1000e_setup_tx_resources(adapter->tx_ring);
4270 	if (err)
4271 		goto err_setup_tx;
4272 
4273 	/* allocate receive descriptors */
4274 	err = e1000e_setup_rx_resources(adapter->rx_ring);
4275 	if (err)
4276 		goto err_setup_rx;
4277 
4278 	/* If AMT is enabled, let the firmware know that the network
4279 	 * interface is now open and reset the part to a known state.
4280 	 */
4281 	if (adapter->flags & FLAG_HAS_AMT) {
4282 		e1000e_get_hw_control(adapter);
4283 		e1000e_reset(adapter);
4284 	}
4285 
4286 	e1000e_power_up_phy(adapter);
4287 
4288 	adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
4289 	if ((adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
4290 		e1000_update_mng_vlan(adapter);
4291 
4292 	/* DMA latency requirement to workaround jumbo issue */
4293 	pm_qos_add_request(&adapter->netdev->pm_qos_req, PM_QOS_CPU_DMA_LATENCY,
4294 			   PM_QOS_DEFAULT_VALUE);
4295 
4296 	/* before we allocate an interrupt, we must be ready to handle it.
4297 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4298 	 * as soon as we call pci_request_irq, so we have to setup our
4299 	 * clean_rx handler before we do so.
4300 	 */
4301 	e1000_configure(adapter);
4302 
4303 	err = e1000_request_irq(adapter);
4304 	if (err)
4305 		goto err_req_irq;
4306 
4307 	/* Work around PCIe errata with MSI interrupts causing some chipsets to
4308 	 * ignore e1000e MSI messages, which means we need to test our MSI
4309 	 * interrupt now
4310 	 */
4311 	if (adapter->int_mode != E1000E_INT_MODE_LEGACY) {
4312 		err = e1000_test_msi(adapter);
4313 		if (err) {
4314 			e_err("Interrupt allocation failed\n");
4315 			goto err_req_irq;
4316 		}
4317 	}
4318 
4319 	/* From here on the code is the same as e1000e_up() */
4320 	clear_bit(__E1000_DOWN, &adapter->state);
4321 
4322 	napi_enable(&adapter->napi);
4323 
4324 	e1000_irq_enable(adapter);
4325 
4326 	adapter->tx_hang_recheck = false;
4327 	netif_start_queue(netdev);
4328 
4329 	adapter->idle_check = true;
4330 	hw->mac.get_link_status = true;
4331 	pm_runtime_put(&pdev->dev);
4332 
4333 	/* fire a link status change interrupt to start the watchdog */
4334 	if (adapter->msix_entries)
4335 		ew32(ICS, E1000_ICS_LSC | E1000_ICR_OTHER);
4336 	else
4337 		ew32(ICS, E1000_ICS_LSC);
4338 
4339 	return 0;
4340 
4341 err_req_irq:
4342 	e1000e_release_hw_control(adapter);
4343 	e1000_power_down_phy(adapter);
4344 	e1000e_free_rx_resources(adapter->rx_ring);
4345 err_setup_rx:
4346 	e1000e_free_tx_resources(adapter->tx_ring);
4347 err_setup_tx:
4348 	e1000e_reset(adapter);
4349 	pm_runtime_put_sync(&pdev->dev);
4350 
4351 	return err;
4352 }
4353 
4354 /**
4355  * e1000_close - Disables a network interface
4356  * @netdev: network interface device structure
4357  *
4358  * Returns 0, this is not allowed to fail
4359  *
4360  * The close entry point is called when an interface is de-activated
4361  * by the OS.  The hardware is still under the drivers control, but
4362  * needs to be disabled.  A global MAC reset is issued to stop the
4363  * hardware, and all transmit and receive resources are freed.
4364  **/
4365 static int e1000_close(struct net_device *netdev)
4366 {
4367 	struct e1000_adapter *adapter = netdev_priv(netdev);
4368 	struct pci_dev *pdev = adapter->pdev;
4369 	int count = E1000_CHECK_RESET_COUNT;
4370 
4371 	while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
4372 		usleep_range(10000, 20000);
4373 
4374 	WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
4375 
4376 	pm_runtime_get_sync(&pdev->dev);
4377 
4378 	if (!test_bit(__E1000_DOWN, &adapter->state)) {
4379 		e1000e_down(adapter);
4380 		e1000_free_irq(adapter);
4381 	}
4382 
4383 	napi_disable(&adapter->napi);
4384 
4385 	e1000_power_down_phy(adapter);
4386 
4387 	e1000e_free_tx_resources(adapter->tx_ring);
4388 	e1000e_free_rx_resources(adapter->rx_ring);
4389 
4390 	/* kill manageability vlan ID if supported, but not if a vlan with
4391 	 * the same ID is registered on the host OS (let 8021q kill it)
4392 	 */
4393 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN)
4394 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
4395 				       adapter->mng_vlan_id);
4396 
4397 	/* If AMT is enabled, let the firmware know that the network
4398 	 * interface is now closed
4399 	 */
4400 	if ((adapter->flags & FLAG_HAS_AMT) &&
4401 	    !test_bit(__E1000_TESTING, &adapter->state))
4402 		e1000e_release_hw_control(adapter);
4403 
4404 	pm_qos_remove_request(&adapter->netdev->pm_qos_req);
4405 
4406 	pm_runtime_put_sync(&pdev->dev);
4407 
4408 	return 0;
4409 }
4410 
4411 /**
4412  * e1000_set_mac - Change the Ethernet Address of the NIC
4413  * @netdev: network interface device structure
4414  * @p: pointer to an address structure
4415  *
4416  * Returns 0 on success, negative on failure
4417  **/
4418 static int e1000_set_mac(struct net_device *netdev, void *p)
4419 {
4420 	struct e1000_adapter *adapter = netdev_priv(netdev);
4421 	struct e1000_hw *hw = &adapter->hw;
4422 	struct sockaddr *addr = p;
4423 
4424 	if (!is_valid_ether_addr(addr->sa_data))
4425 		return -EADDRNOTAVAIL;
4426 
4427 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
4428 	memcpy(adapter->hw.mac.addr, addr->sa_data, netdev->addr_len);
4429 
4430 	hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
4431 
4432 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA) {
4433 		/* activate the work around */
4434 		e1000e_set_laa_state_82571(&adapter->hw, 1);
4435 
4436 		/* Hold a copy of the LAA in RAR[14] This is done so that
4437 		 * between the time RAR[0] gets clobbered  and the time it
4438 		 * gets fixed (in e1000_watchdog), the actual LAA is in one
4439 		 * of the RARs and no incoming packets directed to this port
4440 		 * are dropped. Eventually the LAA will be in RAR[0] and
4441 		 * RAR[14]
4442 		 */
4443 		hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr,
4444 				    adapter->hw.mac.rar_entry_count - 1);
4445 	}
4446 
4447 	return 0;
4448 }
4449 
4450 /**
4451  * e1000e_update_phy_task - work thread to update phy
4452  * @work: pointer to our work struct
4453  *
4454  * this worker thread exists because we must acquire a
4455  * semaphore to read the phy, which we could msleep while
4456  * waiting for it, and we can't msleep in a timer.
4457  **/
4458 static void e1000e_update_phy_task(struct work_struct *work)
4459 {
4460 	struct e1000_adapter *adapter = container_of(work,
4461 						     struct e1000_adapter,
4462 						     update_phy_task);
4463 
4464 	if (test_bit(__E1000_DOWN, &adapter->state))
4465 		return;
4466 
4467 	e1000_get_phy_info(&adapter->hw);
4468 }
4469 
4470 /**
4471  * e1000_update_phy_info - timre call-back to update PHY info
4472  * @data: pointer to adapter cast into an unsigned long
4473  *
4474  * Need to wait a few seconds after link up to get diagnostic information from
4475  * the phy
4476  **/
4477 static void e1000_update_phy_info(unsigned long data)
4478 {
4479 	struct e1000_adapter *adapter = (struct e1000_adapter *)data;
4480 
4481 	if (test_bit(__E1000_DOWN, &adapter->state))
4482 		return;
4483 
4484 	schedule_work(&adapter->update_phy_task);
4485 }
4486 
4487 /**
4488  * e1000e_update_phy_stats - Update the PHY statistics counters
4489  * @adapter: board private structure
4490  *
4491  * Read/clear the upper 16-bit PHY registers and read/accumulate lower
4492  **/
4493 static void e1000e_update_phy_stats(struct e1000_adapter *adapter)
4494 {
4495 	struct e1000_hw *hw = &adapter->hw;
4496 	s32 ret_val;
4497 	u16 phy_data;
4498 
4499 	ret_val = hw->phy.ops.acquire(hw);
4500 	if (ret_val)
4501 		return;
4502 
4503 	/* A page set is expensive so check if already on desired page.
4504 	 * If not, set to the page with the PHY status registers.
4505 	 */
4506 	hw->phy.addr = 1;
4507 	ret_val = e1000e_read_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4508 					   &phy_data);
4509 	if (ret_val)
4510 		goto release;
4511 	if (phy_data != (HV_STATS_PAGE << IGP_PAGE_SHIFT)) {
4512 		ret_val = hw->phy.ops.set_page(hw,
4513 					       HV_STATS_PAGE << IGP_PAGE_SHIFT);
4514 		if (ret_val)
4515 			goto release;
4516 	}
4517 
4518 	/* Single Collision Count */
4519 	hw->phy.ops.read_reg_page(hw, HV_SCC_UPPER, &phy_data);
4520 	ret_val = hw->phy.ops.read_reg_page(hw, HV_SCC_LOWER, &phy_data);
4521 	if (!ret_val)
4522 		adapter->stats.scc += phy_data;
4523 
4524 	/* Excessive Collision Count */
4525 	hw->phy.ops.read_reg_page(hw, HV_ECOL_UPPER, &phy_data);
4526 	ret_val = hw->phy.ops.read_reg_page(hw, HV_ECOL_LOWER, &phy_data);
4527 	if (!ret_val)
4528 		adapter->stats.ecol += phy_data;
4529 
4530 	/* Multiple Collision Count */
4531 	hw->phy.ops.read_reg_page(hw, HV_MCC_UPPER, &phy_data);
4532 	ret_val = hw->phy.ops.read_reg_page(hw, HV_MCC_LOWER, &phy_data);
4533 	if (!ret_val)
4534 		adapter->stats.mcc += phy_data;
4535 
4536 	/* Late Collision Count */
4537 	hw->phy.ops.read_reg_page(hw, HV_LATECOL_UPPER, &phy_data);
4538 	ret_val = hw->phy.ops.read_reg_page(hw, HV_LATECOL_LOWER, &phy_data);
4539 	if (!ret_val)
4540 		adapter->stats.latecol += phy_data;
4541 
4542 	/* Collision Count - also used for adaptive IFS */
4543 	hw->phy.ops.read_reg_page(hw, HV_COLC_UPPER, &phy_data);
4544 	ret_val = hw->phy.ops.read_reg_page(hw, HV_COLC_LOWER, &phy_data);
4545 	if (!ret_val)
4546 		hw->mac.collision_delta = phy_data;
4547 
4548 	/* Defer Count */
4549 	hw->phy.ops.read_reg_page(hw, HV_DC_UPPER, &phy_data);
4550 	ret_val = hw->phy.ops.read_reg_page(hw, HV_DC_LOWER, &phy_data);
4551 	if (!ret_val)
4552 		adapter->stats.dc += phy_data;
4553 
4554 	/* Transmit with no CRS */
4555 	hw->phy.ops.read_reg_page(hw, HV_TNCRS_UPPER, &phy_data);
4556 	ret_val = hw->phy.ops.read_reg_page(hw, HV_TNCRS_LOWER, &phy_data);
4557 	if (!ret_val)
4558 		adapter->stats.tncrs += phy_data;
4559 
4560 release:
4561 	hw->phy.ops.release(hw);
4562 }
4563 
4564 /**
4565  * e1000e_update_stats - Update the board statistics counters
4566  * @adapter: board private structure
4567  **/
4568 static void e1000e_update_stats(struct e1000_adapter *adapter)
4569 {
4570 	struct net_device *netdev = adapter->netdev;
4571 	struct e1000_hw *hw = &adapter->hw;
4572 	struct pci_dev *pdev = adapter->pdev;
4573 
4574 	/* Prevent stats update while adapter is being reset, or if the pci
4575 	 * connection is down.
4576 	 */
4577 	if (adapter->link_speed == 0)
4578 		return;
4579 	if (pci_channel_offline(pdev))
4580 		return;
4581 
4582 	adapter->stats.crcerrs += er32(CRCERRS);
4583 	adapter->stats.gprc += er32(GPRC);
4584 	adapter->stats.gorc += er32(GORCL);
4585 	er32(GORCH);		/* Clear gorc */
4586 	adapter->stats.bprc += er32(BPRC);
4587 	adapter->stats.mprc += er32(MPRC);
4588 	adapter->stats.roc += er32(ROC);
4589 
4590 	adapter->stats.mpc += er32(MPC);
4591 
4592 	/* Half-duplex statistics */
4593 	if (adapter->link_duplex == HALF_DUPLEX) {
4594 		if (adapter->flags2 & FLAG2_HAS_PHY_STATS) {
4595 			e1000e_update_phy_stats(adapter);
4596 		} else {
4597 			adapter->stats.scc += er32(SCC);
4598 			adapter->stats.ecol += er32(ECOL);
4599 			adapter->stats.mcc += er32(MCC);
4600 			adapter->stats.latecol += er32(LATECOL);
4601 			adapter->stats.dc += er32(DC);
4602 
4603 			hw->mac.collision_delta = er32(COLC);
4604 
4605 			if ((hw->mac.type != e1000_82574) &&
4606 			    (hw->mac.type != e1000_82583))
4607 				adapter->stats.tncrs += er32(TNCRS);
4608 		}
4609 		adapter->stats.colc += hw->mac.collision_delta;
4610 	}
4611 
4612 	adapter->stats.xonrxc += er32(XONRXC);
4613 	adapter->stats.xontxc += er32(XONTXC);
4614 	adapter->stats.xoffrxc += er32(XOFFRXC);
4615 	adapter->stats.xofftxc += er32(XOFFTXC);
4616 	adapter->stats.gptc += er32(GPTC);
4617 	adapter->stats.gotc += er32(GOTCL);
4618 	er32(GOTCH);		/* Clear gotc */
4619 	adapter->stats.rnbc += er32(RNBC);
4620 	adapter->stats.ruc += er32(RUC);
4621 
4622 	adapter->stats.mptc += er32(MPTC);
4623 	adapter->stats.bptc += er32(BPTC);
4624 
4625 	/* used for adaptive IFS */
4626 
4627 	hw->mac.tx_packet_delta = er32(TPT);
4628 	adapter->stats.tpt += hw->mac.tx_packet_delta;
4629 
4630 	adapter->stats.algnerrc += er32(ALGNERRC);
4631 	adapter->stats.rxerrc += er32(RXERRC);
4632 	adapter->stats.cexterr += er32(CEXTERR);
4633 	adapter->stats.tsctc += er32(TSCTC);
4634 	adapter->stats.tsctfc += er32(TSCTFC);
4635 
4636 	/* Fill out the OS statistics structure */
4637 	netdev->stats.multicast = adapter->stats.mprc;
4638 	netdev->stats.collisions = adapter->stats.colc;
4639 
4640 	/* Rx Errors */
4641 
4642 	/* RLEC on some newer hardware can be incorrect so build
4643 	 * our own version based on RUC and ROC
4644 	 */
4645 	netdev->stats.rx_errors = adapter->stats.rxerrc +
4646 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
4647 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
4648 	netdev->stats.rx_length_errors = adapter->stats.ruc +
4649 	    adapter->stats.roc;
4650 	netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
4651 	netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
4652 	netdev->stats.rx_missed_errors = adapter->stats.mpc;
4653 
4654 	/* Tx Errors */
4655 	netdev->stats.tx_errors = adapter->stats.ecol + adapter->stats.latecol;
4656 	netdev->stats.tx_aborted_errors = adapter->stats.ecol;
4657 	netdev->stats.tx_window_errors = adapter->stats.latecol;
4658 	netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
4659 
4660 	/* Tx Dropped needs to be maintained elsewhere */
4661 
4662 	/* Management Stats */
4663 	adapter->stats.mgptc += er32(MGTPTC);
4664 	adapter->stats.mgprc += er32(MGTPRC);
4665 	adapter->stats.mgpdc += er32(MGTPDC);
4666 
4667 	/* Correctable ECC Errors */
4668 	if (hw->mac.type == e1000_pch_lpt) {
4669 		u32 pbeccsts = er32(PBECCSTS);
4670 		adapter->corr_errors +=
4671 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
4672 		adapter->uncorr_errors +=
4673 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
4674 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
4675 	}
4676 }
4677 
4678 /**
4679  * e1000_phy_read_status - Update the PHY register status snapshot
4680  * @adapter: board private structure
4681  **/
4682 static void e1000_phy_read_status(struct e1000_adapter *adapter)
4683 {
4684 	struct e1000_hw *hw = &adapter->hw;
4685 	struct e1000_phy_regs *phy = &adapter->phy_regs;
4686 
4687 	if (!pm_runtime_suspended((&adapter->pdev->dev)->parent) &&
4688 	    (er32(STATUS) & E1000_STATUS_LU) &&
4689 	    (adapter->hw.phy.media_type == e1000_media_type_copper)) {
4690 		int ret_val;
4691 
4692 		ret_val = e1e_rphy(hw, MII_BMCR, &phy->bmcr);
4693 		ret_val |= e1e_rphy(hw, MII_BMSR, &phy->bmsr);
4694 		ret_val |= e1e_rphy(hw, MII_ADVERTISE, &phy->advertise);
4695 		ret_val |= e1e_rphy(hw, MII_LPA, &phy->lpa);
4696 		ret_val |= e1e_rphy(hw, MII_EXPANSION, &phy->expansion);
4697 		ret_val |= e1e_rphy(hw, MII_CTRL1000, &phy->ctrl1000);
4698 		ret_val |= e1e_rphy(hw, MII_STAT1000, &phy->stat1000);
4699 		ret_val |= e1e_rphy(hw, MII_ESTATUS, &phy->estatus);
4700 		if (ret_val)
4701 			e_warn("Error reading PHY register\n");
4702 	} else {
4703 		/* Do not read PHY registers if link is not up
4704 		 * Set values to typical power-on defaults
4705 		 */
4706 		phy->bmcr = (BMCR_SPEED1000 | BMCR_ANENABLE | BMCR_FULLDPLX);
4707 		phy->bmsr = (BMSR_100FULL | BMSR_100HALF | BMSR_10FULL |
4708 			     BMSR_10HALF | BMSR_ESTATEN | BMSR_ANEGCAPABLE |
4709 			     BMSR_ERCAP);
4710 		phy->advertise = (ADVERTISE_PAUSE_ASYM | ADVERTISE_PAUSE_CAP |
4711 				  ADVERTISE_ALL | ADVERTISE_CSMA);
4712 		phy->lpa = 0;
4713 		phy->expansion = EXPANSION_ENABLENPAGE;
4714 		phy->ctrl1000 = ADVERTISE_1000FULL;
4715 		phy->stat1000 = 0;
4716 		phy->estatus = (ESTATUS_1000_TFULL | ESTATUS_1000_THALF);
4717 	}
4718 }
4719 
4720 static void e1000_print_link_info(struct e1000_adapter *adapter)
4721 {
4722 	struct e1000_hw *hw = &adapter->hw;
4723 	u32 ctrl = er32(CTRL);
4724 
4725 	/* Link status message must follow this format for user tools */
4726 	pr_info("%s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
4727 		adapter->netdev->name, adapter->link_speed,
4728 		adapter->link_duplex == FULL_DUPLEX ? "Full" : "Half",
4729 		(ctrl & E1000_CTRL_TFCE) && (ctrl & E1000_CTRL_RFCE) ? "Rx/Tx" :
4730 		(ctrl & E1000_CTRL_RFCE) ? "Rx" :
4731 		(ctrl & E1000_CTRL_TFCE) ? "Tx" : "None");
4732 }
4733 
4734 static bool e1000e_has_link(struct e1000_adapter *adapter)
4735 {
4736 	struct e1000_hw *hw = &adapter->hw;
4737 	bool link_active = false;
4738 	s32 ret_val = 0;
4739 
4740 	/* get_link_status is set on LSC (link status) interrupt or
4741 	 * Rx sequence error interrupt.  get_link_status will stay
4742 	 * false until the check_for_link establishes link
4743 	 * for copper adapters ONLY
4744 	 */
4745 	switch (hw->phy.media_type) {
4746 	case e1000_media_type_copper:
4747 		if (hw->mac.get_link_status) {
4748 			ret_val = hw->mac.ops.check_for_link(hw);
4749 			link_active = !hw->mac.get_link_status;
4750 		} else {
4751 			link_active = true;
4752 		}
4753 		break;
4754 	case e1000_media_type_fiber:
4755 		ret_val = hw->mac.ops.check_for_link(hw);
4756 		link_active = !!(er32(STATUS) & E1000_STATUS_LU);
4757 		break;
4758 	case e1000_media_type_internal_serdes:
4759 		ret_val = hw->mac.ops.check_for_link(hw);
4760 		link_active = adapter->hw.mac.serdes_has_link;
4761 		break;
4762 	default:
4763 	case e1000_media_type_unknown:
4764 		break;
4765 	}
4766 
4767 	if ((ret_val == E1000_ERR_PHY) && (hw->phy.type == e1000_phy_igp_3) &&
4768 	    (er32(CTRL) & E1000_PHY_CTRL_GBE_DISABLE)) {
4769 		/* See e1000_kmrn_lock_loss_workaround_ich8lan() */
4770 		e_info("Gigabit has been disabled, downgrading speed\n");
4771 	}
4772 
4773 	return link_active;
4774 }
4775 
4776 static void e1000e_enable_receives(struct e1000_adapter *adapter)
4777 {
4778 	/* make sure the receive unit is started */
4779 	if ((adapter->flags & FLAG_RX_NEEDS_RESTART) &&
4780 	    (adapter->flags & FLAG_RESTART_NOW)) {
4781 		struct e1000_hw *hw = &adapter->hw;
4782 		u32 rctl = er32(RCTL);
4783 		ew32(RCTL, rctl | E1000_RCTL_EN);
4784 		adapter->flags &= ~FLAG_RESTART_NOW;
4785 	}
4786 }
4787 
4788 static void e1000e_check_82574_phy_workaround(struct e1000_adapter *adapter)
4789 {
4790 	struct e1000_hw *hw = &adapter->hw;
4791 
4792 	/* With 82574 controllers, PHY needs to be checked periodically
4793 	 * for hung state and reset, if two calls return true
4794 	 */
4795 	if (e1000_check_phy_82574(hw))
4796 		adapter->phy_hang_count++;
4797 	else
4798 		adapter->phy_hang_count = 0;
4799 
4800 	if (adapter->phy_hang_count > 1) {
4801 		adapter->phy_hang_count = 0;
4802 		schedule_work(&adapter->reset_task);
4803 	}
4804 }
4805 
4806 /**
4807  * e1000_watchdog - Timer Call-back
4808  * @data: pointer to adapter cast into an unsigned long
4809  **/
4810 static void e1000_watchdog(unsigned long data)
4811 {
4812 	struct e1000_adapter *adapter = (struct e1000_adapter *)data;
4813 
4814 	/* Do the rest outside of interrupt context */
4815 	schedule_work(&adapter->watchdog_task);
4816 
4817 	/* TODO: make this use queue_delayed_work() */
4818 }
4819 
4820 static void e1000_watchdog_task(struct work_struct *work)
4821 {
4822 	struct e1000_adapter *adapter = container_of(work,
4823 						     struct e1000_adapter,
4824 						     watchdog_task);
4825 	struct net_device *netdev = adapter->netdev;
4826 	struct e1000_mac_info *mac = &adapter->hw.mac;
4827 	struct e1000_phy_info *phy = &adapter->hw.phy;
4828 	struct e1000_ring *tx_ring = adapter->tx_ring;
4829 	struct e1000_hw *hw = &adapter->hw;
4830 	u32 link, tctl;
4831 
4832 	if (test_bit(__E1000_DOWN, &adapter->state))
4833 		return;
4834 
4835 	link = e1000e_has_link(adapter);
4836 	if ((netif_carrier_ok(netdev)) && link) {
4837 		/* Cancel scheduled suspend requests. */
4838 		pm_runtime_resume(netdev->dev.parent);
4839 
4840 		e1000e_enable_receives(adapter);
4841 		goto link_up;
4842 	}
4843 
4844 	if ((e1000e_enable_tx_pkt_filtering(hw)) &&
4845 	    (adapter->mng_vlan_id != adapter->hw.mng_cookie.vlan_id))
4846 		e1000_update_mng_vlan(adapter);
4847 
4848 	if (link) {
4849 		if (!netif_carrier_ok(netdev)) {
4850 			bool txb2b = true;
4851 
4852 			/* Cancel scheduled suspend requests. */
4853 			pm_runtime_resume(netdev->dev.parent);
4854 
4855 			/* update snapshot of PHY registers on LSC */
4856 			e1000_phy_read_status(adapter);
4857 			mac->ops.get_link_up_info(&adapter->hw,
4858 						  &adapter->link_speed,
4859 						  &adapter->link_duplex);
4860 			e1000_print_link_info(adapter);
4861 
4862 			/* check if SmartSpeed worked */
4863 			e1000e_check_downshift(hw);
4864 			if (phy->speed_downgraded)
4865 				netdev_warn(netdev,
4866 					    "Link Speed was downgraded by SmartSpeed\n");
4867 
4868 			/* On supported PHYs, check for duplex mismatch only
4869 			 * if link has autonegotiated at 10/100 half
4870 			 */
4871 			if ((hw->phy.type == e1000_phy_igp_3 ||
4872 			     hw->phy.type == e1000_phy_bm) &&
4873 			    hw->mac.autoneg &&
4874 			    (adapter->link_speed == SPEED_10 ||
4875 			     adapter->link_speed == SPEED_100) &&
4876 			    (adapter->link_duplex == HALF_DUPLEX)) {
4877 				u16 autoneg_exp;
4878 
4879 				e1e_rphy(hw, MII_EXPANSION, &autoneg_exp);
4880 
4881 				if (!(autoneg_exp & EXPANSION_NWAY))
4882 					e_info("Autonegotiated half duplex but link partner cannot autoneg.  Try forcing full duplex if link gets many collisions.\n");
4883 			}
4884 
4885 			/* adjust timeout factor according to speed/duplex */
4886 			adapter->tx_timeout_factor = 1;
4887 			switch (adapter->link_speed) {
4888 			case SPEED_10:
4889 				txb2b = false;
4890 				adapter->tx_timeout_factor = 16;
4891 				break;
4892 			case SPEED_100:
4893 				txb2b = false;
4894 				adapter->tx_timeout_factor = 10;
4895 				break;
4896 			}
4897 
4898 			/* workaround: re-program speed mode bit after
4899 			 * link-up event
4900 			 */
4901 			if ((adapter->flags & FLAG_TARC_SPEED_MODE_BIT) &&
4902 			    !txb2b) {
4903 				u32 tarc0;
4904 				tarc0 = er32(TARC(0));
4905 				tarc0 &= ~SPEED_MODE_BIT;
4906 				ew32(TARC(0), tarc0);
4907 			}
4908 
4909 			/* disable TSO for pcie and 10/100 speeds, to avoid
4910 			 * some hardware issues
4911 			 */
4912 			if (!(adapter->flags & FLAG_TSO_FORCE)) {
4913 				switch (adapter->link_speed) {
4914 				case SPEED_10:
4915 				case SPEED_100:
4916 					e_info("10/100 speed: disabling TSO\n");
4917 					netdev->features &= ~NETIF_F_TSO;
4918 					netdev->features &= ~NETIF_F_TSO6;
4919 					break;
4920 				case SPEED_1000:
4921 					netdev->features |= NETIF_F_TSO;
4922 					netdev->features |= NETIF_F_TSO6;
4923 					break;
4924 				default:
4925 					/* oops */
4926 					break;
4927 				}
4928 			}
4929 
4930 			/* enable transmits in the hardware, need to do this
4931 			 * after setting TARC(0)
4932 			 */
4933 			tctl = er32(TCTL);
4934 			tctl |= E1000_TCTL_EN;
4935 			ew32(TCTL, tctl);
4936 
4937 			/* Perform any post-link-up configuration before
4938 			 * reporting link up.
4939 			 */
4940 			if (phy->ops.cfg_on_link_up)
4941 				phy->ops.cfg_on_link_up(hw);
4942 
4943 			netif_carrier_on(netdev);
4944 
4945 			if (!test_bit(__E1000_DOWN, &adapter->state))
4946 				mod_timer(&adapter->phy_info_timer,
4947 					  round_jiffies(jiffies + 2 * HZ));
4948 		}
4949 	} else {
4950 		if (netif_carrier_ok(netdev)) {
4951 			adapter->link_speed = 0;
4952 			adapter->link_duplex = 0;
4953 			/* Link status message must follow this format */
4954 			pr_info("%s NIC Link is Down\n", adapter->netdev->name);
4955 			netif_carrier_off(netdev);
4956 			if (!test_bit(__E1000_DOWN, &adapter->state))
4957 				mod_timer(&adapter->phy_info_timer,
4958 					  round_jiffies(jiffies + 2 * HZ));
4959 
4960 			/* The link is lost so the controller stops DMA.
4961 			 * If there is queued Tx work that cannot be done
4962 			 * or if on an 8000ES2LAN which requires a Rx packet
4963 			 * buffer work-around on link down event, reset the
4964 			 * controller to flush the Tx/Rx packet buffers.
4965 			 * (Do the reset outside of interrupt context).
4966 			 */
4967 			if ((adapter->flags & FLAG_RX_NEEDS_RESTART) ||
4968 			    (e1000_desc_unused(tx_ring) + 1 < tx_ring->count))
4969 				adapter->flags |= FLAG_RESTART_NOW;
4970 			else
4971 				pm_schedule_suspend(netdev->dev.parent,
4972 						    LINK_TIMEOUT);
4973 		}
4974 	}
4975 
4976 link_up:
4977 	spin_lock(&adapter->stats64_lock);
4978 	e1000e_update_stats(adapter);
4979 
4980 	mac->tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
4981 	adapter->tpt_old = adapter->stats.tpt;
4982 	mac->collision_delta = adapter->stats.colc - adapter->colc_old;
4983 	adapter->colc_old = adapter->stats.colc;
4984 
4985 	adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
4986 	adapter->gorc_old = adapter->stats.gorc;
4987 	adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
4988 	adapter->gotc_old = adapter->stats.gotc;
4989 	spin_unlock(&adapter->stats64_lock);
4990 
4991 	if (adapter->flags & FLAG_RESTART_NOW) {
4992 		schedule_work(&adapter->reset_task);
4993 		/* return immediately since reset is imminent */
4994 		return;
4995 	}
4996 
4997 	e1000e_update_adaptive(&adapter->hw);
4998 
4999 	/* Simple mode for Interrupt Throttle Rate (ITR) */
5000 	if (adapter->itr_setting == 4) {
5001 		/* Symmetric Tx/Rx gets a reduced ITR=2000;
5002 		 * Total asymmetrical Tx or Rx gets ITR=8000;
5003 		 * everyone else is between 2000-8000.
5004 		 */
5005 		u32 goc = (adapter->gotc + adapter->gorc) / 10000;
5006 		u32 dif = (adapter->gotc > adapter->gorc ?
5007 			   adapter->gotc - adapter->gorc :
5008 			   adapter->gorc - adapter->gotc) / 10000;
5009 		u32 itr = goc > 0 ? (dif * 6000 / goc + 2000) : 8000;
5010 
5011 		e1000e_write_itr(adapter, itr);
5012 	}
5013 
5014 	/* Cause software interrupt to ensure Rx ring is cleaned */
5015 	if (adapter->msix_entries)
5016 		ew32(ICS, adapter->rx_ring->ims_val);
5017 	else
5018 		ew32(ICS, E1000_ICS_RXDMT0);
5019 
5020 	/* flush pending descriptors to memory before detecting Tx hang */
5021 	e1000e_flush_descriptors(adapter);
5022 
5023 	/* Force detection of hung controller every watchdog period */
5024 	adapter->detect_tx_hung = true;
5025 
5026 	/* With 82571 controllers, LAA may be overwritten due to controller
5027 	 * reset from the other port. Set the appropriate LAA in RAR[0]
5028 	 */
5029 	if (e1000e_get_laa_state_82571(hw))
5030 		hw->mac.ops.rar_set(hw, adapter->hw.mac.addr, 0);
5031 
5032 	if (adapter->flags2 & FLAG2_CHECK_PHY_HANG)
5033 		e1000e_check_82574_phy_workaround(adapter);
5034 
5035 	/* Clear valid timestamp stuck in RXSTMPL/H due to a Rx error */
5036 	if (adapter->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE) {
5037 		if ((adapter->flags2 & FLAG2_CHECK_RX_HWTSTAMP) &&
5038 		    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID)) {
5039 			er32(RXSTMPH);
5040 			adapter->rx_hwtstamp_cleared++;
5041 		} else {
5042 			adapter->flags2 |= FLAG2_CHECK_RX_HWTSTAMP;
5043 		}
5044 	}
5045 
5046 	/* Reset the timer */
5047 	if (!test_bit(__E1000_DOWN, &adapter->state))
5048 		mod_timer(&adapter->watchdog_timer,
5049 			  round_jiffies(jiffies + 2 * HZ));
5050 }
5051 
5052 #define E1000_TX_FLAGS_CSUM		0x00000001
5053 #define E1000_TX_FLAGS_VLAN		0x00000002
5054 #define E1000_TX_FLAGS_TSO		0x00000004
5055 #define E1000_TX_FLAGS_IPV4		0x00000008
5056 #define E1000_TX_FLAGS_NO_FCS		0x00000010
5057 #define E1000_TX_FLAGS_HWTSTAMP		0x00000020
5058 #define E1000_TX_FLAGS_VLAN_MASK	0xffff0000
5059 #define E1000_TX_FLAGS_VLAN_SHIFT	16
5060 
5061 static int e1000_tso(struct e1000_ring *tx_ring, struct sk_buff *skb)
5062 {
5063 	struct e1000_context_desc *context_desc;
5064 	struct e1000_buffer *buffer_info;
5065 	unsigned int i;
5066 	u32 cmd_length = 0;
5067 	u16 ipcse = 0, mss;
5068 	u8 ipcss, ipcso, tucss, tucso, hdr_len;
5069 
5070 	if (!skb_is_gso(skb))
5071 		return 0;
5072 
5073 	if (skb_header_cloned(skb)) {
5074 		int err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
5075 
5076 		if (err)
5077 			return err;
5078 	}
5079 
5080 	hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5081 	mss = skb_shinfo(skb)->gso_size;
5082 	if (skb->protocol == htons(ETH_P_IP)) {
5083 		struct iphdr *iph = ip_hdr(skb);
5084 		iph->tot_len = 0;
5085 		iph->check = 0;
5086 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
5087 							 0, IPPROTO_TCP, 0);
5088 		cmd_length = E1000_TXD_CMD_IP;
5089 		ipcse = skb_transport_offset(skb) - 1;
5090 	} else if (skb_is_gso_v6(skb)) {
5091 		ipv6_hdr(skb)->payload_len = 0;
5092 		tcp_hdr(skb)->check = ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
5093 						       &ipv6_hdr(skb)->daddr,
5094 						       0, IPPROTO_TCP, 0);
5095 		ipcse = 0;
5096 	}
5097 	ipcss = skb_network_offset(skb);
5098 	ipcso = (void *)&(ip_hdr(skb)->check) - (void *)skb->data;
5099 	tucss = skb_transport_offset(skb);
5100 	tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
5101 
5102 	cmd_length |= (E1000_TXD_CMD_DEXT | E1000_TXD_CMD_TSE |
5103 		       E1000_TXD_CMD_TCP | (skb->len - (hdr_len)));
5104 
5105 	i = tx_ring->next_to_use;
5106 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5107 	buffer_info = &tx_ring->buffer_info[i];
5108 
5109 	context_desc->lower_setup.ip_fields.ipcss = ipcss;
5110 	context_desc->lower_setup.ip_fields.ipcso = ipcso;
5111 	context_desc->lower_setup.ip_fields.ipcse = cpu_to_le16(ipcse);
5112 	context_desc->upper_setup.tcp_fields.tucss = tucss;
5113 	context_desc->upper_setup.tcp_fields.tucso = tucso;
5114 	context_desc->upper_setup.tcp_fields.tucse = 0;
5115 	context_desc->tcp_seg_setup.fields.mss = cpu_to_le16(mss);
5116 	context_desc->tcp_seg_setup.fields.hdr_len = hdr_len;
5117 	context_desc->cmd_and_length = cpu_to_le32(cmd_length);
5118 
5119 	buffer_info->time_stamp = jiffies;
5120 	buffer_info->next_to_watch = i;
5121 
5122 	i++;
5123 	if (i == tx_ring->count)
5124 		i = 0;
5125 	tx_ring->next_to_use = i;
5126 
5127 	return 1;
5128 }
5129 
5130 static bool e1000_tx_csum(struct e1000_ring *tx_ring, struct sk_buff *skb)
5131 {
5132 	struct e1000_adapter *adapter = tx_ring->adapter;
5133 	struct e1000_context_desc *context_desc;
5134 	struct e1000_buffer *buffer_info;
5135 	unsigned int i;
5136 	u8 css;
5137 	u32 cmd_len = E1000_TXD_CMD_DEXT;
5138 	__be16 protocol;
5139 
5140 	if (skb->ip_summed != CHECKSUM_PARTIAL)
5141 		return 0;
5142 
5143 	if (skb->protocol == cpu_to_be16(ETH_P_8021Q))
5144 		protocol = vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
5145 	else
5146 		protocol = skb->protocol;
5147 
5148 	switch (protocol) {
5149 	case cpu_to_be16(ETH_P_IP):
5150 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
5151 			cmd_len |= E1000_TXD_CMD_TCP;
5152 		break;
5153 	case cpu_to_be16(ETH_P_IPV6):
5154 		/* XXX not handling all IPV6 headers */
5155 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
5156 			cmd_len |= E1000_TXD_CMD_TCP;
5157 		break;
5158 	default:
5159 		if (unlikely(net_ratelimit()))
5160 			e_warn("checksum_partial proto=%x!\n",
5161 			       be16_to_cpu(protocol));
5162 		break;
5163 	}
5164 
5165 	css = skb_checksum_start_offset(skb);
5166 
5167 	i = tx_ring->next_to_use;
5168 	buffer_info = &tx_ring->buffer_info[i];
5169 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5170 
5171 	context_desc->lower_setup.ip_config = 0;
5172 	context_desc->upper_setup.tcp_fields.tucss = css;
5173 	context_desc->upper_setup.tcp_fields.tucso = css + skb->csum_offset;
5174 	context_desc->upper_setup.tcp_fields.tucse = 0;
5175 	context_desc->tcp_seg_setup.data = 0;
5176 	context_desc->cmd_and_length = cpu_to_le32(cmd_len);
5177 
5178 	buffer_info->time_stamp = jiffies;
5179 	buffer_info->next_to_watch = i;
5180 
5181 	i++;
5182 	if (i == tx_ring->count)
5183 		i = 0;
5184 	tx_ring->next_to_use = i;
5185 
5186 	return 1;
5187 }
5188 
5189 static int e1000_tx_map(struct e1000_ring *tx_ring, struct sk_buff *skb,
5190 			unsigned int first, unsigned int max_per_txd,
5191 			unsigned int nr_frags)
5192 {
5193 	struct e1000_adapter *adapter = tx_ring->adapter;
5194 	struct pci_dev *pdev = adapter->pdev;
5195 	struct e1000_buffer *buffer_info;
5196 	unsigned int len = skb_headlen(skb);
5197 	unsigned int offset = 0, size, count = 0, i;
5198 	unsigned int f, bytecount, segs;
5199 
5200 	i = tx_ring->next_to_use;
5201 
5202 	while (len) {
5203 		buffer_info = &tx_ring->buffer_info[i];
5204 		size = min(len, max_per_txd);
5205 
5206 		buffer_info->length = size;
5207 		buffer_info->time_stamp = jiffies;
5208 		buffer_info->next_to_watch = i;
5209 		buffer_info->dma = dma_map_single(&pdev->dev,
5210 						  skb->data + offset,
5211 						  size, DMA_TO_DEVICE);
5212 		buffer_info->mapped_as_page = false;
5213 		if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5214 			goto dma_error;
5215 
5216 		len -= size;
5217 		offset += size;
5218 		count++;
5219 
5220 		if (len) {
5221 			i++;
5222 			if (i == tx_ring->count)
5223 				i = 0;
5224 		}
5225 	}
5226 
5227 	for (f = 0; f < nr_frags; f++) {
5228 		const struct skb_frag_struct *frag;
5229 
5230 		frag = &skb_shinfo(skb)->frags[f];
5231 		len = skb_frag_size(frag);
5232 		offset = 0;
5233 
5234 		while (len) {
5235 			i++;
5236 			if (i == tx_ring->count)
5237 				i = 0;
5238 
5239 			buffer_info = &tx_ring->buffer_info[i];
5240 			size = min(len, max_per_txd);
5241 
5242 			buffer_info->length = size;
5243 			buffer_info->time_stamp = jiffies;
5244 			buffer_info->next_to_watch = i;
5245 			buffer_info->dma = skb_frag_dma_map(&pdev->dev, frag,
5246 							    offset, size,
5247 							    DMA_TO_DEVICE);
5248 			buffer_info->mapped_as_page = true;
5249 			if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5250 				goto dma_error;
5251 
5252 			len -= size;
5253 			offset += size;
5254 			count++;
5255 		}
5256 	}
5257 
5258 	segs = skb_shinfo(skb)->gso_segs ? : 1;
5259 	/* multiply data chunks by size of headers */
5260 	bytecount = ((segs - 1) * skb_headlen(skb)) + skb->len;
5261 
5262 	tx_ring->buffer_info[i].skb = skb;
5263 	tx_ring->buffer_info[i].segs = segs;
5264 	tx_ring->buffer_info[i].bytecount = bytecount;
5265 	tx_ring->buffer_info[first].next_to_watch = i;
5266 
5267 	return count;
5268 
5269 dma_error:
5270 	dev_err(&pdev->dev, "Tx DMA map failed\n");
5271 	buffer_info->dma = 0;
5272 	if (count)
5273 		count--;
5274 
5275 	while (count--) {
5276 		if (i == 0)
5277 			i += tx_ring->count;
5278 		i--;
5279 		buffer_info = &tx_ring->buffer_info[i];
5280 		e1000_put_txbuf(tx_ring, buffer_info);
5281 	}
5282 
5283 	return 0;
5284 }
5285 
5286 static void e1000_tx_queue(struct e1000_ring *tx_ring, int tx_flags, int count)
5287 {
5288 	struct e1000_adapter *adapter = tx_ring->adapter;
5289 	struct e1000_tx_desc *tx_desc = NULL;
5290 	struct e1000_buffer *buffer_info;
5291 	u32 txd_upper = 0, txd_lower = E1000_TXD_CMD_IFCS;
5292 	unsigned int i;
5293 
5294 	if (tx_flags & E1000_TX_FLAGS_TSO) {
5295 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D |
5296 		    E1000_TXD_CMD_TSE;
5297 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5298 
5299 		if (tx_flags & E1000_TX_FLAGS_IPV4)
5300 			txd_upper |= E1000_TXD_POPTS_IXSM << 8;
5301 	}
5302 
5303 	if (tx_flags & E1000_TX_FLAGS_CSUM) {
5304 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5305 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5306 	}
5307 
5308 	if (tx_flags & E1000_TX_FLAGS_VLAN) {
5309 		txd_lower |= E1000_TXD_CMD_VLE;
5310 		txd_upper |= (tx_flags & E1000_TX_FLAGS_VLAN_MASK);
5311 	}
5312 
5313 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5314 		txd_lower &= ~(E1000_TXD_CMD_IFCS);
5315 
5316 	if (unlikely(tx_flags & E1000_TX_FLAGS_HWTSTAMP)) {
5317 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5318 		txd_upper |= E1000_TXD_EXTCMD_TSTAMP;
5319 	}
5320 
5321 	i = tx_ring->next_to_use;
5322 
5323 	do {
5324 		buffer_info = &tx_ring->buffer_info[i];
5325 		tx_desc = E1000_TX_DESC(*tx_ring, i);
5326 		tx_desc->buffer_addr = cpu_to_le64(buffer_info->dma);
5327 		tx_desc->lower.data = cpu_to_le32(txd_lower |
5328 						  buffer_info->length);
5329 		tx_desc->upper.data = cpu_to_le32(txd_upper);
5330 
5331 		i++;
5332 		if (i == tx_ring->count)
5333 			i = 0;
5334 	} while (--count > 0);
5335 
5336 	tx_desc->lower.data |= cpu_to_le32(adapter->txd_cmd);
5337 
5338 	/* txd_cmd re-enables FCS, so we'll re-disable it here as desired. */
5339 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5340 		tx_desc->lower.data &= ~(cpu_to_le32(E1000_TXD_CMD_IFCS));
5341 
5342 	/* Force memory writes to complete before letting h/w
5343 	 * know there are new descriptors to fetch.  (Only
5344 	 * applicable for weak-ordered memory model archs,
5345 	 * such as IA-64).
5346 	 */
5347 	wmb();
5348 
5349 	tx_ring->next_to_use = i;
5350 
5351 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
5352 		e1000e_update_tdt_wa(tx_ring, i);
5353 	else
5354 		writel(i, tx_ring->tail);
5355 
5356 	/* we need this if more than one processor can write to our tail
5357 	 * at a time, it synchronizes IO on IA64/Altix systems
5358 	 */
5359 	mmiowb();
5360 }
5361 
5362 #define MINIMUM_DHCP_PACKET_SIZE 282
5363 static int e1000_transfer_dhcp_info(struct e1000_adapter *adapter,
5364 				    struct sk_buff *skb)
5365 {
5366 	struct e1000_hw *hw = &adapter->hw;
5367 	u16 length, offset;
5368 
5369 	if (vlan_tx_tag_present(skb) &&
5370 	    !((vlan_tx_tag_get(skb) == adapter->hw.mng_cookie.vlan_id) &&
5371 	      (adapter->hw.mng_cookie.status &
5372 	       E1000_MNG_DHCP_COOKIE_STATUS_VLAN)))
5373 		return 0;
5374 
5375 	if (skb->len <= MINIMUM_DHCP_PACKET_SIZE)
5376 		return 0;
5377 
5378 	if (((struct ethhdr *)skb->data)->h_proto != htons(ETH_P_IP))
5379 		return 0;
5380 
5381 	{
5382 		const struct iphdr *ip = (struct iphdr *)((u8 *)skb->data + 14);
5383 		struct udphdr *udp;
5384 
5385 		if (ip->protocol != IPPROTO_UDP)
5386 			return 0;
5387 
5388 		udp = (struct udphdr *)((u8 *)ip + (ip->ihl << 2));
5389 		if (ntohs(udp->dest) != 67)
5390 			return 0;
5391 
5392 		offset = (u8 *)udp + 8 - skb->data;
5393 		length = skb->len - offset;
5394 		return e1000e_mng_write_dhcp_info(hw, (u8 *)udp + 8, length);
5395 	}
5396 
5397 	return 0;
5398 }
5399 
5400 static int __e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5401 {
5402 	struct e1000_adapter *adapter = tx_ring->adapter;
5403 
5404 	netif_stop_queue(adapter->netdev);
5405 	/* Herbert's original patch had:
5406 	 *  smp_mb__after_netif_stop_queue();
5407 	 * but since that doesn't exist yet, just open code it.
5408 	 */
5409 	smp_mb();
5410 
5411 	/* We need to check again in a case another CPU has just
5412 	 * made room available.
5413 	 */
5414 	if (e1000_desc_unused(tx_ring) < size)
5415 		return -EBUSY;
5416 
5417 	/* A reprieve! */
5418 	netif_start_queue(adapter->netdev);
5419 	++adapter->restart_queue;
5420 	return 0;
5421 }
5422 
5423 static int e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5424 {
5425 	BUG_ON(size > tx_ring->count);
5426 
5427 	if (e1000_desc_unused(tx_ring) >= size)
5428 		return 0;
5429 	return __e1000_maybe_stop_tx(tx_ring, size);
5430 }
5431 
5432 static netdev_tx_t e1000_xmit_frame(struct sk_buff *skb,
5433 				    struct net_device *netdev)
5434 {
5435 	struct e1000_adapter *adapter = netdev_priv(netdev);
5436 	struct e1000_ring *tx_ring = adapter->tx_ring;
5437 	unsigned int first;
5438 	unsigned int tx_flags = 0;
5439 	unsigned int len = skb_headlen(skb);
5440 	unsigned int nr_frags;
5441 	unsigned int mss;
5442 	int count = 0;
5443 	int tso;
5444 	unsigned int f;
5445 
5446 	if (test_bit(__E1000_DOWN, &adapter->state)) {
5447 		dev_kfree_skb_any(skb);
5448 		return NETDEV_TX_OK;
5449 	}
5450 
5451 	if (skb->len <= 0) {
5452 		dev_kfree_skb_any(skb);
5453 		return NETDEV_TX_OK;
5454 	}
5455 
5456 	/* The minimum packet size with TCTL.PSP set is 17 bytes so
5457 	 * pad skb in order to meet this minimum size requirement
5458 	 */
5459 	if (unlikely(skb->len < 17)) {
5460 		if (skb_pad(skb, 17 - skb->len))
5461 			return NETDEV_TX_OK;
5462 		skb->len = 17;
5463 		skb_set_tail_pointer(skb, 17);
5464 	}
5465 
5466 	mss = skb_shinfo(skb)->gso_size;
5467 	if (mss) {
5468 		u8 hdr_len;
5469 
5470 		/* TSO Workaround for 82571/2/3 Controllers -- if skb->data
5471 		 * points to just header, pull a few bytes of payload from
5472 		 * frags into skb->data
5473 		 */
5474 		hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5475 		/* we do this workaround for ES2LAN, but it is un-necessary,
5476 		 * avoiding it could save a lot of cycles
5477 		 */
5478 		if (skb->data_len && (hdr_len == len)) {
5479 			unsigned int pull_size;
5480 
5481 			pull_size = min_t(unsigned int, 4, skb->data_len);
5482 			if (!__pskb_pull_tail(skb, pull_size)) {
5483 				e_err("__pskb_pull_tail failed.\n");
5484 				dev_kfree_skb_any(skb);
5485 				return NETDEV_TX_OK;
5486 			}
5487 			len = skb_headlen(skb);
5488 		}
5489 	}
5490 
5491 	/* reserve a descriptor for the offload context */
5492 	if ((mss) || (skb->ip_summed == CHECKSUM_PARTIAL))
5493 		count++;
5494 	count++;
5495 
5496 	count += DIV_ROUND_UP(len, adapter->tx_fifo_limit);
5497 
5498 	nr_frags = skb_shinfo(skb)->nr_frags;
5499 	for (f = 0; f < nr_frags; f++)
5500 		count += DIV_ROUND_UP(skb_frag_size(&skb_shinfo(skb)->frags[f]),
5501 				      adapter->tx_fifo_limit);
5502 
5503 	if (adapter->hw.mac.tx_pkt_filtering)
5504 		e1000_transfer_dhcp_info(adapter, skb);
5505 
5506 	/* need: count + 2 desc gap to keep tail from touching
5507 	 * head, otherwise try next time
5508 	 */
5509 	if (e1000_maybe_stop_tx(tx_ring, count + 2))
5510 		return NETDEV_TX_BUSY;
5511 
5512 	if (vlan_tx_tag_present(skb)) {
5513 		tx_flags |= E1000_TX_FLAGS_VLAN;
5514 		tx_flags |= (vlan_tx_tag_get(skb) << E1000_TX_FLAGS_VLAN_SHIFT);
5515 	}
5516 
5517 	first = tx_ring->next_to_use;
5518 
5519 	tso = e1000_tso(tx_ring, skb);
5520 	if (tso < 0) {
5521 		dev_kfree_skb_any(skb);
5522 		return NETDEV_TX_OK;
5523 	}
5524 
5525 	if (tso)
5526 		tx_flags |= E1000_TX_FLAGS_TSO;
5527 	else if (e1000_tx_csum(tx_ring, skb))
5528 		tx_flags |= E1000_TX_FLAGS_CSUM;
5529 
5530 	/* Old method was to assume IPv4 packet by default if TSO was enabled.
5531 	 * 82571 hardware supports TSO capabilities for IPv6 as well...
5532 	 * no longer assume, we must.
5533 	 */
5534 	if (skb->protocol == htons(ETH_P_IP))
5535 		tx_flags |= E1000_TX_FLAGS_IPV4;
5536 
5537 	if (unlikely(skb->no_fcs))
5538 		tx_flags |= E1000_TX_FLAGS_NO_FCS;
5539 
5540 	/* if count is 0 then mapping error has occurred */
5541 	count = e1000_tx_map(tx_ring, skb, first, adapter->tx_fifo_limit,
5542 			     nr_frags);
5543 	if (count) {
5544 		if (unlikely((skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
5545 			     !adapter->tx_hwtstamp_skb)) {
5546 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5547 			tx_flags |= E1000_TX_FLAGS_HWTSTAMP;
5548 			adapter->tx_hwtstamp_skb = skb_get(skb);
5549 			schedule_work(&adapter->tx_hwtstamp_work);
5550 		} else {
5551 			skb_tx_timestamp(skb);
5552 		}
5553 
5554 		netdev_sent_queue(netdev, skb->len);
5555 		e1000_tx_queue(tx_ring, tx_flags, count);
5556 		/* Make sure there is space in the ring for the next send. */
5557 		e1000_maybe_stop_tx(tx_ring,
5558 				    (MAX_SKB_FRAGS *
5559 				     DIV_ROUND_UP(PAGE_SIZE,
5560 						  adapter->tx_fifo_limit) + 2));
5561 	} else {
5562 		dev_kfree_skb_any(skb);
5563 		tx_ring->buffer_info[first].time_stamp = 0;
5564 		tx_ring->next_to_use = first;
5565 	}
5566 
5567 	return NETDEV_TX_OK;
5568 }
5569 
5570 /**
5571  * e1000_tx_timeout - Respond to a Tx Hang
5572  * @netdev: network interface device structure
5573  **/
5574 static void e1000_tx_timeout(struct net_device *netdev)
5575 {
5576 	struct e1000_adapter *adapter = netdev_priv(netdev);
5577 
5578 	/* Do the reset outside of interrupt context */
5579 	adapter->tx_timeout_count++;
5580 	schedule_work(&adapter->reset_task);
5581 }
5582 
5583 static void e1000_reset_task(struct work_struct *work)
5584 {
5585 	struct e1000_adapter *adapter;
5586 	adapter = container_of(work, struct e1000_adapter, reset_task);
5587 
5588 	/* don't run the task if already down */
5589 	if (test_bit(__E1000_DOWN, &adapter->state))
5590 		return;
5591 
5592 	if (!(adapter->flags & FLAG_RESTART_NOW)) {
5593 		e1000e_dump(adapter);
5594 		e_err("Reset adapter unexpectedly\n");
5595 	}
5596 	e1000e_reinit_locked(adapter);
5597 }
5598 
5599 /**
5600  * e1000_get_stats64 - Get System Network Statistics
5601  * @netdev: network interface device structure
5602  * @stats: rtnl_link_stats64 pointer
5603  *
5604  * Returns the address of the device statistics structure.
5605  **/
5606 struct rtnl_link_stats64 *e1000e_get_stats64(struct net_device *netdev,
5607 					     struct rtnl_link_stats64 *stats)
5608 {
5609 	struct e1000_adapter *adapter = netdev_priv(netdev);
5610 
5611 	memset(stats, 0, sizeof(struct rtnl_link_stats64));
5612 	spin_lock(&adapter->stats64_lock);
5613 	e1000e_update_stats(adapter);
5614 	/* Fill out the OS statistics structure */
5615 	stats->rx_bytes = adapter->stats.gorc;
5616 	stats->rx_packets = adapter->stats.gprc;
5617 	stats->tx_bytes = adapter->stats.gotc;
5618 	stats->tx_packets = adapter->stats.gptc;
5619 	stats->multicast = adapter->stats.mprc;
5620 	stats->collisions = adapter->stats.colc;
5621 
5622 	/* Rx Errors */
5623 
5624 	/* RLEC on some newer hardware can be incorrect so build
5625 	 * our own version based on RUC and ROC
5626 	 */
5627 	stats->rx_errors = adapter->stats.rxerrc +
5628 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
5629 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5630 	stats->rx_length_errors = adapter->stats.ruc + adapter->stats.roc;
5631 	stats->rx_crc_errors = adapter->stats.crcerrs;
5632 	stats->rx_frame_errors = adapter->stats.algnerrc;
5633 	stats->rx_missed_errors = adapter->stats.mpc;
5634 
5635 	/* Tx Errors */
5636 	stats->tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5637 	stats->tx_aborted_errors = adapter->stats.ecol;
5638 	stats->tx_window_errors = adapter->stats.latecol;
5639 	stats->tx_carrier_errors = adapter->stats.tncrs;
5640 
5641 	/* Tx Dropped needs to be maintained elsewhere */
5642 
5643 	spin_unlock(&adapter->stats64_lock);
5644 	return stats;
5645 }
5646 
5647 /**
5648  * e1000_change_mtu - Change the Maximum Transfer Unit
5649  * @netdev: network interface device structure
5650  * @new_mtu: new value for maximum frame size
5651  *
5652  * Returns 0 on success, negative on failure
5653  **/
5654 static int e1000_change_mtu(struct net_device *netdev, int new_mtu)
5655 {
5656 	struct e1000_adapter *adapter = netdev_priv(netdev);
5657 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
5658 
5659 	/* Jumbo frame support */
5660 	if ((max_frame > ETH_FRAME_LEN + ETH_FCS_LEN) &&
5661 	    !(adapter->flags & FLAG_HAS_JUMBO_FRAMES)) {
5662 		e_err("Jumbo Frames not supported.\n");
5663 		return -EINVAL;
5664 	}
5665 
5666 	/* Supported frame sizes */
5667 	if ((new_mtu < ETH_ZLEN + ETH_FCS_LEN + VLAN_HLEN) ||
5668 	    (max_frame > adapter->max_hw_frame_size)) {
5669 		e_err("Unsupported MTU setting\n");
5670 		return -EINVAL;
5671 	}
5672 
5673 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
5674 	if ((adapter->hw.mac.type >= e1000_pch2lan) &&
5675 	    !(adapter->flags2 & FLAG2_CRC_STRIPPING) &&
5676 	    (new_mtu > ETH_DATA_LEN)) {
5677 		e_err("Jumbo Frames not supported on this device when CRC stripping is disabled.\n");
5678 		return -EINVAL;
5679 	}
5680 
5681 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
5682 		usleep_range(1000, 2000);
5683 	/* e1000e_down -> e1000e_reset dependent on max_frame_size & mtu */
5684 	adapter->max_frame_size = max_frame;
5685 	e_info("changing MTU from %d to %d\n", netdev->mtu, new_mtu);
5686 	netdev->mtu = new_mtu;
5687 	if (netif_running(netdev))
5688 		e1000e_down(adapter);
5689 
5690 	/* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
5691 	 * means we reserve 2 more, this pushes us to allocate from the next
5692 	 * larger slab size.
5693 	 * i.e. RXBUFFER_2048 --> size-4096 slab
5694 	 * However with the new *_jumbo_rx* routines, jumbo receives will use
5695 	 * fragmented skbs
5696 	 */
5697 
5698 	if (max_frame <= 2048)
5699 		adapter->rx_buffer_len = 2048;
5700 	else
5701 		adapter->rx_buffer_len = 4096;
5702 
5703 	/* adjust allocation if LPE protects us, and we aren't using SBP */
5704 	if ((max_frame == ETH_FRAME_LEN + ETH_FCS_LEN) ||
5705 	    (max_frame == ETH_FRAME_LEN + VLAN_HLEN + ETH_FCS_LEN))
5706 		adapter->rx_buffer_len = ETH_FRAME_LEN + VLAN_HLEN
5707 		    + ETH_FCS_LEN;
5708 
5709 	if (netif_running(netdev))
5710 		e1000e_up(adapter);
5711 	else
5712 		e1000e_reset(adapter);
5713 
5714 	clear_bit(__E1000_RESETTING, &adapter->state);
5715 
5716 	return 0;
5717 }
5718 
5719 static int e1000_mii_ioctl(struct net_device *netdev, struct ifreq *ifr,
5720 			   int cmd)
5721 {
5722 	struct e1000_adapter *adapter = netdev_priv(netdev);
5723 	struct mii_ioctl_data *data = if_mii(ifr);
5724 
5725 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
5726 		return -EOPNOTSUPP;
5727 
5728 	switch (cmd) {
5729 	case SIOCGMIIPHY:
5730 		data->phy_id = adapter->hw.phy.addr;
5731 		break;
5732 	case SIOCGMIIREG:
5733 		e1000_phy_read_status(adapter);
5734 
5735 		switch (data->reg_num & 0x1F) {
5736 		case MII_BMCR:
5737 			data->val_out = adapter->phy_regs.bmcr;
5738 			break;
5739 		case MII_BMSR:
5740 			data->val_out = adapter->phy_regs.bmsr;
5741 			break;
5742 		case MII_PHYSID1:
5743 			data->val_out = (adapter->hw.phy.id >> 16);
5744 			break;
5745 		case MII_PHYSID2:
5746 			data->val_out = (adapter->hw.phy.id & 0xFFFF);
5747 			break;
5748 		case MII_ADVERTISE:
5749 			data->val_out = adapter->phy_regs.advertise;
5750 			break;
5751 		case MII_LPA:
5752 			data->val_out = adapter->phy_regs.lpa;
5753 			break;
5754 		case MII_EXPANSION:
5755 			data->val_out = adapter->phy_regs.expansion;
5756 			break;
5757 		case MII_CTRL1000:
5758 			data->val_out = adapter->phy_regs.ctrl1000;
5759 			break;
5760 		case MII_STAT1000:
5761 			data->val_out = adapter->phy_regs.stat1000;
5762 			break;
5763 		case MII_ESTATUS:
5764 			data->val_out = adapter->phy_regs.estatus;
5765 			break;
5766 		default:
5767 			return -EIO;
5768 		}
5769 		break;
5770 	case SIOCSMIIREG:
5771 	default:
5772 		return -EOPNOTSUPP;
5773 	}
5774 	return 0;
5775 }
5776 
5777 /**
5778  * e1000e_hwtstamp_ioctl - control hardware time stamping
5779  * @netdev: network interface device structure
5780  * @ifreq: interface request
5781  *
5782  * Outgoing time stamping can be enabled and disabled. Play nice and
5783  * disable it when requested, although it shouldn't cause any overhead
5784  * when no packet needs it. At most one packet in the queue may be
5785  * marked for time stamping, otherwise it would be impossible to tell
5786  * for sure to which packet the hardware time stamp belongs.
5787  *
5788  * Incoming time stamping has to be configured via the hardware filters.
5789  * Not all combinations are supported, in particular event type has to be
5790  * specified. Matching the kind of event packet is not supported, with the
5791  * exception of "all V2 events regardless of level 2 or 4".
5792  **/
5793 static int e1000e_hwtstamp_ioctl(struct net_device *netdev, struct ifreq *ifr)
5794 {
5795 	struct e1000_adapter *adapter = netdev_priv(netdev);
5796 	struct hwtstamp_config config;
5797 	int ret_val;
5798 
5799 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
5800 		return -EFAULT;
5801 
5802 	ret_val = e1000e_config_hwtstamp(adapter, &config);
5803 	if (ret_val)
5804 		return ret_val;
5805 
5806 	switch (config.rx_filter) {
5807 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
5808 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
5809 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
5810 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
5811 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
5812 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
5813 		/* With V2 type filters which specify a Sync or Delay Request,
5814 		 * Path Delay Request/Response messages are also time stamped
5815 		 * by hardware so notify the caller the requested packets plus
5816 		 * some others are time stamped.
5817 		 */
5818 		config.rx_filter = HWTSTAMP_FILTER_SOME;
5819 		break;
5820 	default:
5821 		break;
5822 	}
5823 
5824 	return copy_to_user(ifr->ifr_data, &config,
5825 			    sizeof(config)) ? -EFAULT : 0;
5826 }
5827 
5828 static int e1000_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
5829 {
5830 	switch (cmd) {
5831 	case SIOCGMIIPHY:
5832 	case SIOCGMIIREG:
5833 	case SIOCSMIIREG:
5834 		return e1000_mii_ioctl(netdev, ifr, cmd);
5835 	case SIOCSHWTSTAMP:
5836 		return e1000e_hwtstamp_ioctl(netdev, ifr);
5837 	default:
5838 		return -EOPNOTSUPP;
5839 	}
5840 }
5841 
5842 static int e1000_init_phy_wakeup(struct e1000_adapter *adapter, u32 wufc)
5843 {
5844 	struct e1000_hw *hw = &adapter->hw;
5845 	u32 i, mac_reg;
5846 	u16 phy_reg, wuc_enable;
5847 	int retval;
5848 
5849 	/* copy MAC RARs to PHY RARs */
5850 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
5851 
5852 	retval = hw->phy.ops.acquire(hw);
5853 	if (retval) {
5854 		e_err("Could not acquire PHY\n");
5855 		return retval;
5856 	}
5857 
5858 	/* Enable access to wakeup registers on and set page to BM_WUC_PAGE */
5859 	retval = e1000_enable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
5860 	if (retval)
5861 		goto release;
5862 
5863 	/* copy MAC MTA to PHY MTA - only needed for pchlan */
5864 	for (i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
5865 		mac_reg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
5866 		hw->phy.ops.write_reg_page(hw, BM_MTA(i),
5867 					   (u16)(mac_reg & 0xFFFF));
5868 		hw->phy.ops.write_reg_page(hw, BM_MTA(i) + 1,
5869 					   (u16)((mac_reg >> 16) & 0xFFFF));
5870 	}
5871 
5872 	/* configure PHY Rx Control register */
5873 	hw->phy.ops.read_reg_page(&adapter->hw, BM_RCTL, &phy_reg);
5874 	mac_reg = er32(RCTL);
5875 	if (mac_reg & E1000_RCTL_UPE)
5876 		phy_reg |= BM_RCTL_UPE;
5877 	if (mac_reg & E1000_RCTL_MPE)
5878 		phy_reg |= BM_RCTL_MPE;
5879 	phy_reg &= ~(BM_RCTL_MO_MASK);
5880 	if (mac_reg & E1000_RCTL_MO_3)
5881 		phy_reg |= (((mac_reg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
5882 			    << BM_RCTL_MO_SHIFT);
5883 	if (mac_reg & E1000_RCTL_BAM)
5884 		phy_reg |= BM_RCTL_BAM;
5885 	if (mac_reg & E1000_RCTL_PMCF)
5886 		phy_reg |= BM_RCTL_PMCF;
5887 	mac_reg = er32(CTRL);
5888 	if (mac_reg & E1000_CTRL_RFCE)
5889 		phy_reg |= BM_RCTL_RFCE;
5890 	hw->phy.ops.write_reg_page(&adapter->hw, BM_RCTL, phy_reg);
5891 
5892 	/* enable PHY wakeup in MAC register */
5893 	ew32(WUFC, wufc);
5894 	ew32(WUC, E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN);
5895 
5896 	/* configure and enable PHY wakeup in PHY registers */
5897 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUFC, wufc);
5898 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUC, E1000_WUC_PME_EN);
5899 
5900 	/* activate PHY wakeup */
5901 	wuc_enable |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
5902 	retval = e1000_disable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
5903 	if (retval)
5904 		e_err("Could not set PHY Host Wakeup bit\n");
5905 release:
5906 	hw->phy.ops.release(hw);
5907 
5908 	return retval;
5909 }
5910 
5911 static int __e1000_shutdown(struct pci_dev *pdev, bool runtime)
5912 {
5913 	struct net_device *netdev = pci_get_drvdata(pdev);
5914 	struct e1000_adapter *adapter = netdev_priv(netdev);
5915 	struct e1000_hw *hw = &adapter->hw;
5916 	u32 ctrl, ctrl_ext, rctl, status;
5917 	/* Runtime suspend should only enable wakeup for link changes */
5918 	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
5919 	int retval = 0;
5920 
5921 	netif_device_detach(netdev);
5922 
5923 	if (netif_running(netdev)) {
5924 		int count = E1000_CHECK_RESET_COUNT;
5925 
5926 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
5927 			usleep_range(10000, 20000);
5928 
5929 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
5930 		e1000e_down(adapter);
5931 		e1000_free_irq(adapter);
5932 	}
5933 	e1000e_reset_interrupt_capability(adapter);
5934 
5935 	status = er32(STATUS);
5936 	if (status & E1000_STATUS_LU)
5937 		wufc &= ~E1000_WUFC_LNKC;
5938 
5939 	if (wufc) {
5940 		e1000_setup_rctl(adapter);
5941 		e1000e_set_rx_mode(netdev);
5942 
5943 		/* turn on all-multi mode if wake on multicast is enabled */
5944 		if (wufc & E1000_WUFC_MC) {
5945 			rctl = er32(RCTL);
5946 			rctl |= E1000_RCTL_MPE;
5947 			ew32(RCTL, rctl);
5948 		}
5949 
5950 		ctrl = er32(CTRL);
5951 		ctrl |= E1000_CTRL_ADVD3WUC;
5952 		if (!(adapter->flags2 & FLAG2_HAS_PHY_WAKEUP))
5953 			ctrl |= E1000_CTRL_EN_PHY_PWR_MGMT;
5954 		ew32(CTRL, ctrl);
5955 
5956 		if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
5957 		    adapter->hw.phy.media_type ==
5958 		    e1000_media_type_internal_serdes) {
5959 			/* keep the laser running in D3 */
5960 			ctrl_ext = er32(CTRL_EXT);
5961 			ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
5962 			ew32(CTRL_EXT, ctrl_ext);
5963 		}
5964 
5965 		if (adapter->flags & FLAG_IS_ICH)
5966 			e1000_suspend_workarounds_ich8lan(&adapter->hw);
5967 
5968 		/* Allow time for pending master requests to run */
5969 		e1000e_disable_pcie_master(&adapter->hw);
5970 
5971 		if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
5972 			/* enable wakeup by the PHY */
5973 			retval = e1000_init_phy_wakeup(adapter, wufc);
5974 			if (retval)
5975 				return retval;
5976 		} else {
5977 			/* enable wakeup by the MAC */
5978 			ew32(WUFC, wufc);
5979 			ew32(WUC, E1000_WUC_PME_EN);
5980 		}
5981 	} else {
5982 		ew32(WUC, 0);
5983 		ew32(WUFC, 0);
5984 	}
5985 
5986 	if (adapter->hw.phy.type == e1000_phy_igp_3)
5987 		e1000e_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
5988 
5989 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
5990 	 * would have already happened in close and is redundant.
5991 	 */
5992 	e1000e_release_hw_control(adapter);
5993 
5994 	pci_clear_master(pdev);
5995 
5996 	/* The pci-e switch on some quad port adapters will report a
5997 	 * correctable error when the MAC transitions from D0 to D3.  To
5998 	 * prevent this we need to mask off the correctable errors on the
5999 	 * downstream port of the pci-e switch.
6000 	 *
6001 	 * We don't have the associated upstream bridge while assigning
6002 	 * the PCI device into guest. For example, the KVM on power is
6003 	 * one of the cases.
6004 	 */
6005 	if (adapter->flags & FLAG_IS_QUAD_PORT) {
6006 		struct pci_dev *us_dev = pdev->bus->self;
6007 		u16 devctl;
6008 
6009 		if (!us_dev)
6010 			return 0;
6011 
6012 		pcie_capability_read_word(us_dev, PCI_EXP_DEVCTL, &devctl);
6013 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL,
6014 					   (devctl & ~PCI_EXP_DEVCTL_CERE));
6015 
6016 		pci_save_state(pdev);
6017 		pci_prepare_to_sleep(pdev);
6018 
6019 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL, devctl);
6020 	}
6021 
6022 	return 0;
6023 }
6024 
6025 /**
6026  * e1000e_disable_aspm - Disable ASPM states
6027  * @pdev: pointer to PCI device struct
6028  * @state: bit-mask of ASPM states to disable
6029  *
6030  * Some devices *must* have certain ASPM states disabled per hardware errata.
6031  **/
6032 static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6033 {
6034 	struct pci_dev *parent = pdev->bus->self;
6035 	u16 aspm_dis_mask = 0;
6036 	u16 pdev_aspmc, parent_aspmc;
6037 
6038 	switch (state) {
6039 	case PCIE_LINK_STATE_L0S:
6040 	case PCIE_LINK_STATE_L0S | PCIE_LINK_STATE_L1:
6041 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L0S;
6042 		/* fall-through - can't have L1 without L0s */
6043 	case PCIE_LINK_STATE_L1:
6044 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L1;
6045 		break;
6046 	default:
6047 		return;
6048 	}
6049 
6050 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6051 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6052 
6053 	if (parent) {
6054 		pcie_capability_read_word(parent, PCI_EXP_LNKCTL,
6055 					  &parent_aspmc);
6056 		parent_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6057 	}
6058 
6059 	/* Nothing to do if the ASPM states to be disabled already are */
6060 	if (!(pdev_aspmc & aspm_dis_mask) &&
6061 	    (!parent || !(parent_aspmc & aspm_dis_mask)))
6062 		return;
6063 
6064 	dev_info(&pdev->dev, "Disabling ASPM %s %s\n",
6065 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L0S) ?
6066 		 "L0s" : "",
6067 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L1) ?
6068 		 "L1" : "");
6069 
6070 #ifdef CONFIG_PCIEASPM
6071 	pci_disable_link_state_locked(pdev, state);
6072 
6073 	/* Double-check ASPM control.  If not disabled by the above, the
6074 	 * BIOS is preventing that from happening (or CONFIG_PCIEASPM is
6075 	 * not enabled); override by writing PCI config space directly.
6076 	 */
6077 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6078 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6079 
6080 	if (!(aspm_dis_mask & pdev_aspmc))
6081 		return;
6082 #endif
6083 
6084 	/* Both device and parent should have the same ASPM setting.
6085 	 * Disable ASPM in downstream component first and then upstream.
6086 	 */
6087 	pcie_capability_clear_word(pdev, PCI_EXP_LNKCTL, aspm_dis_mask);
6088 
6089 	if (parent)
6090 		pcie_capability_clear_word(parent, PCI_EXP_LNKCTL,
6091 					   aspm_dis_mask);
6092 }
6093 
6094 #ifdef CONFIG_PM
6095 static bool e1000e_pm_ready(struct e1000_adapter *adapter)
6096 {
6097 	return !!adapter->tx_ring->buffer_info;
6098 }
6099 
6100 static int __e1000_resume(struct pci_dev *pdev)
6101 {
6102 	struct net_device *netdev = pci_get_drvdata(pdev);
6103 	struct e1000_adapter *adapter = netdev_priv(netdev);
6104 	struct e1000_hw *hw = &adapter->hw;
6105 	u16 aspm_disable_flag = 0;
6106 	u32 err;
6107 
6108 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6109 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6110 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6111 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6112 	if (aspm_disable_flag)
6113 		e1000e_disable_aspm(pdev, aspm_disable_flag);
6114 
6115 	pci_set_master(pdev);
6116 
6117 	e1000e_set_interrupt_capability(adapter);
6118 	if (netif_running(netdev)) {
6119 		err = e1000_request_irq(adapter);
6120 		if (err)
6121 			return err;
6122 	}
6123 
6124 	if (hw->mac.type >= e1000_pch2lan)
6125 		e1000_resume_workarounds_pchlan(&adapter->hw);
6126 
6127 	e1000e_power_up_phy(adapter);
6128 
6129 	/* report the system wakeup cause from S3/S4 */
6130 	if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6131 		u16 phy_data;
6132 
6133 		e1e_rphy(&adapter->hw, BM_WUS, &phy_data);
6134 		if (phy_data) {
6135 			e_info("PHY Wakeup cause - %s\n",
6136 			       phy_data & E1000_WUS_EX ? "Unicast Packet" :
6137 			       phy_data & E1000_WUS_MC ? "Multicast Packet" :
6138 			       phy_data & E1000_WUS_BC ? "Broadcast Packet" :
6139 			       phy_data & E1000_WUS_MAG ? "Magic Packet" :
6140 			       phy_data & E1000_WUS_LNKC ?
6141 			       "Link Status Change" : "other");
6142 		}
6143 		e1e_wphy(&adapter->hw, BM_WUS, ~0);
6144 	} else {
6145 		u32 wus = er32(WUS);
6146 		if (wus) {
6147 			e_info("MAC Wakeup cause - %s\n",
6148 			       wus & E1000_WUS_EX ? "Unicast Packet" :
6149 			       wus & E1000_WUS_MC ? "Multicast Packet" :
6150 			       wus & E1000_WUS_BC ? "Broadcast Packet" :
6151 			       wus & E1000_WUS_MAG ? "Magic Packet" :
6152 			       wus & E1000_WUS_LNKC ? "Link Status Change" :
6153 			       "other");
6154 		}
6155 		ew32(WUS, ~0);
6156 	}
6157 
6158 	e1000e_reset(adapter);
6159 
6160 	e1000_init_manageability_pt(adapter);
6161 
6162 	if (netif_running(netdev))
6163 		e1000e_up(adapter);
6164 
6165 	netif_device_attach(netdev);
6166 
6167 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6168 	 * is up.  For all other cases, let the f/w know that the h/w is now
6169 	 * under the control of the driver.
6170 	 */
6171 	if (!(adapter->flags & FLAG_HAS_AMT))
6172 		e1000e_get_hw_control(adapter);
6173 
6174 	return 0;
6175 }
6176 
6177 #ifdef CONFIG_PM
6178 static int e1000_suspend(struct device *dev)
6179 {
6180 	struct pci_dev *pdev = to_pci_dev(dev);
6181 
6182 	return __e1000_shutdown(pdev, false);
6183 }
6184 
6185 static int e1000_resume(struct device *dev)
6186 {
6187 	struct pci_dev *pdev = to_pci_dev(dev);
6188 	struct net_device *netdev = pci_get_drvdata(pdev);
6189 	struct e1000_adapter *adapter = netdev_priv(netdev);
6190 
6191 	if (e1000e_pm_ready(adapter))
6192 		adapter->idle_check = true;
6193 
6194 	return __e1000_resume(pdev);
6195 }
6196 #endif /* CONFIG_PM */
6197 
6198 #ifdef CONFIG_PM_RUNTIME
6199 static int e1000_runtime_suspend(struct device *dev)
6200 {
6201 	struct pci_dev *pdev = to_pci_dev(dev);
6202 	struct net_device *netdev = pci_get_drvdata(pdev);
6203 	struct e1000_adapter *adapter = netdev_priv(netdev);
6204 
6205 	if (!e1000e_pm_ready(adapter))
6206 		return 0;
6207 
6208 	return __e1000_shutdown(pdev, true);
6209 }
6210 
6211 static int e1000_idle(struct device *dev)
6212 {
6213 	struct pci_dev *pdev = to_pci_dev(dev);
6214 	struct net_device *netdev = pci_get_drvdata(pdev);
6215 	struct e1000_adapter *adapter = netdev_priv(netdev);
6216 
6217 	if (!e1000e_pm_ready(adapter))
6218 		return 0;
6219 
6220 	if (adapter->idle_check) {
6221 		adapter->idle_check = false;
6222 		if (!e1000e_has_link(adapter))
6223 			pm_schedule_suspend(dev, MSEC_PER_SEC);
6224 	}
6225 
6226 	return -EBUSY;
6227 }
6228 
6229 static int e1000_runtime_resume(struct device *dev)
6230 {
6231 	struct pci_dev *pdev = to_pci_dev(dev);
6232 	struct net_device *netdev = pci_get_drvdata(pdev);
6233 	struct e1000_adapter *adapter = netdev_priv(netdev);
6234 
6235 	if (!e1000e_pm_ready(adapter))
6236 		return 0;
6237 
6238 	adapter->idle_check = !dev->power.runtime_auto;
6239 	return __e1000_resume(pdev);
6240 }
6241 #endif /* CONFIG_PM_RUNTIME */
6242 #endif /* CONFIG_PM */
6243 
6244 static void e1000_shutdown(struct pci_dev *pdev)
6245 {
6246 	__e1000_shutdown(pdev, false);
6247 }
6248 
6249 #ifdef CONFIG_NET_POLL_CONTROLLER
6250 
6251 static irqreturn_t e1000_intr_msix(int __always_unused irq, void *data)
6252 {
6253 	struct net_device *netdev = data;
6254 	struct e1000_adapter *adapter = netdev_priv(netdev);
6255 
6256 	if (adapter->msix_entries) {
6257 		int vector, msix_irq;
6258 
6259 		vector = 0;
6260 		msix_irq = adapter->msix_entries[vector].vector;
6261 		disable_irq(msix_irq);
6262 		e1000_intr_msix_rx(msix_irq, netdev);
6263 		enable_irq(msix_irq);
6264 
6265 		vector++;
6266 		msix_irq = adapter->msix_entries[vector].vector;
6267 		disable_irq(msix_irq);
6268 		e1000_intr_msix_tx(msix_irq, netdev);
6269 		enable_irq(msix_irq);
6270 
6271 		vector++;
6272 		msix_irq = adapter->msix_entries[vector].vector;
6273 		disable_irq(msix_irq);
6274 		e1000_msix_other(msix_irq, netdev);
6275 		enable_irq(msix_irq);
6276 	}
6277 
6278 	return IRQ_HANDLED;
6279 }
6280 
6281 /**
6282  * e1000_netpoll
6283  * @netdev: network interface device structure
6284  *
6285  * Polling 'interrupt' - used by things like netconsole to send skbs
6286  * without having to re-enable interrupts. It's not called while
6287  * the interrupt routine is executing.
6288  */
6289 static void e1000_netpoll(struct net_device *netdev)
6290 {
6291 	struct e1000_adapter *adapter = netdev_priv(netdev);
6292 
6293 	switch (adapter->int_mode) {
6294 	case E1000E_INT_MODE_MSIX:
6295 		e1000_intr_msix(adapter->pdev->irq, netdev);
6296 		break;
6297 	case E1000E_INT_MODE_MSI:
6298 		disable_irq(adapter->pdev->irq);
6299 		e1000_intr_msi(adapter->pdev->irq, netdev);
6300 		enable_irq(adapter->pdev->irq);
6301 		break;
6302 	default:		/* E1000E_INT_MODE_LEGACY */
6303 		disable_irq(adapter->pdev->irq);
6304 		e1000_intr(adapter->pdev->irq, netdev);
6305 		enable_irq(adapter->pdev->irq);
6306 		break;
6307 	}
6308 }
6309 #endif
6310 
6311 /**
6312  * e1000_io_error_detected - called when PCI error is detected
6313  * @pdev: Pointer to PCI device
6314  * @state: The current pci connection state
6315  *
6316  * This function is called after a PCI bus error affecting
6317  * this device has been detected.
6318  */
6319 static pci_ers_result_t e1000_io_error_detected(struct pci_dev *pdev,
6320 						pci_channel_state_t state)
6321 {
6322 	struct net_device *netdev = pci_get_drvdata(pdev);
6323 	struct e1000_adapter *adapter = netdev_priv(netdev);
6324 
6325 	netif_device_detach(netdev);
6326 
6327 	if (state == pci_channel_io_perm_failure)
6328 		return PCI_ERS_RESULT_DISCONNECT;
6329 
6330 	if (netif_running(netdev))
6331 		e1000e_down(adapter);
6332 	pci_disable_device(pdev);
6333 
6334 	/* Request a slot slot reset. */
6335 	return PCI_ERS_RESULT_NEED_RESET;
6336 }
6337 
6338 /**
6339  * e1000_io_slot_reset - called after the pci bus has been reset.
6340  * @pdev: Pointer to PCI device
6341  *
6342  * Restart the card from scratch, as if from a cold-boot. Implementation
6343  * resembles the first-half of the e1000_resume routine.
6344  */
6345 static pci_ers_result_t e1000_io_slot_reset(struct pci_dev *pdev)
6346 {
6347 	struct net_device *netdev = pci_get_drvdata(pdev);
6348 	struct e1000_adapter *adapter = netdev_priv(netdev);
6349 	struct e1000_hw *hw = &adapter->hw;
6350 	u16 aspm_disable_flag = 0;
6351 	int err;
6352 	pci_ers_result_t result;
6353 
6354 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6355 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6356 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6357 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6358 	if (aspm_disable_flag)
6359 		e1000e_disable_aspm(pdev, aspm_disable_flag);
6360 
6361 	err = pci_enable_device_mem(pdev);
6362 	if (err) {
6363 		dev_err(&pdev->dev,
6364 			"Cannot re-enable PCI device after reset.\n");
6365 		result = PCI_ERS_RESULT_DISCONNECT;
6366 	} else {
6367 		pdev->state_saved = true;
6368 		pci_restore_state(pdev);
6369 		pci_set_master(pdev);
6370 
6371 		pci_enable_wake(pdev, PCI_D3hot, 0);
6372 		pci_enable_wake(pdev, PCI_D3cold, 0);
6373 
6374 		e1000e_reset(adapter);
6375 		ew32(WUS, ~0);
6376 		result = PCI_ERS_RESULT_RECOVERED;
6377 	}
6378 
6379 	pci_cleanup_aer_uncorrect_error_status(pdev);
6380 
6381 	return result;
6382 }
6383 
6384 /**
6385  * e1000_io_resume - called when traffic can start flowing again.
6386  * @pdev: Pointer to PCI device
6387  *
6388  * This callback is called when the error recovery driver tells us that
6389  * its OK to resume normal operation. Implementation resembles the
6390  * second-half of the e1000_resume routine.
6391  */
6392 static void e1000_io_resume(struct pci_dev *pdev)
6393 {
6394 	struct net_device *netdev = pci_get_drvdata(pdev);
6395 	struct e1000_adapter *adapter = netdev_priv(netdev);
6396 
6397 	e1000_init_manageability_pt(adapter);
6398 
6399 	if (netif_running(netdev)) {
6400 		if (e1000e_up(adapter)) {
6401 			dev_err(&pdev->dev,
6402 				"can't bring device back up after reset\n");
6403 			return;
6404 		}
6405 	}
6406 
6407 	netif_device_attach(netdev);
6408 
6409 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6410 	 * is up.  For all other cases, let the f/w know that the h/w is now
6411 	 * under the control of the driver.
6412 	 */
6413 	if (!(adapter->flags & FLAG_HAS_AMT))
6414 		e1000e_get_hw_control(adapter);
6415 }
6416 
6417 static void e1000_print_device_info(struct e1000_adapter *adapter)
6418 {
6419 	struct e1000_hw *hw = &adapter->hw;
6420 	struct net_device *netdev = adapter->netdev;
6421 	u32 ret_val;
6422 	u8 pba_str[E1000_PBANUM_LENGTH];
6423 
6424 	/* print bus type/speed/width info */
6425 	e_info("(PCI Express:2.5GT/s:%s) %pM\n",
6426 	       /* bus width */
6427 	       ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
6428 		"Width x1"),
6429 	       /* MAC address */
6430 	       netdev->dev_addr);
6431 	e_info("Intel(R) PRO/%s Network Connection\n",
6432 	       (hw->phy.type == e1000_phy_ife) ? "10/100" : "1000");
6433 	ret_val = e1000_read_pba_string_generic(hw, pba_str,
6434 						E1000_PBANUM_LENGTH);
6435 	if (ret_val)
6436 		strlcpy((char *)pba_str, "Unknown", sizeof(pba_str));
6437 	e_info("MAC: %d, PHY: %d, PBA No: %s\n",
6438 	       hw->mac.type, hw->phy.type, pba_str);
6439 }
6440 
6441 static void e1000_eeprom_checks(struct e1000_adapter *adapter)
6442 {
6443 	struct e1000_hw *hw = &adapter->hw;
6444 	int ret_val;
6445 	u16 buf = 0;
6446 
6447 	if (hw->mac.type != e1000_82573)
6448 		return;
6449 
6450 	ret_val = e1000_read_nvm(hw, NVM_INIT_CONTROL2_REG, 1, &buf);
6451 	le16_to_cpus(&buf);
6452 	if (!ret_val && (!(buf & (1 << 0)))) {
6453 		/* Deep Smart Power Down (DSPD) */
6454 		dev_warn(&adapter->pdev->dev,
6455 			 "Warning: detected DSPD enabled in EEPROM\n");
6456 	}
6457 }
6458 
6459 static int e1000_set_features(struct net_device *netdev,
6460 			      netdev_features_t features)
6461 {
6462 	struct e1000_adapter *adapter = netdev_priv(netdev);
6463 	netdev_features_t changed = features ^ netdev->features;
6464 
6465 	if (changed & (NETIF_F_TSO | NETIF_F_TSO6))
6466 		adapter->flags |= FLAG_TSO_FORCE;
6467 
6468 	if (!(changed & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX |
6469 			 NETIF_F_RXCSUM | NETIF_F_RXHASH | NETIF_F_RXFCS |
6470 			 NETIF_F_RXALL)))
6471 		return 0;
6472 
6473 	if (changed & NETIF_F_RXFCS) {
6474 		if (features & NETIF_F_RXFCS) {
6475 			adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
6476 		} else {
6477 			/* We need to take it back to defaults, which might mean
6478 			 * stripping is still disabled at the adapter level.
6479 			 */
6480 			if (adapter->flags2 & FLAG2_DFLT_CRC_STRIPPING)
6481 				adapter->flags2 |= FLAG2_CRC_STRIPPING;
6482 			else
6483 				adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
6484 		}
6485 	}
6486 
6487 	netdev->features = features;
6488 
6489 	if (netif_running(netdev))
6490 		e1000e_reinit_locked(adapter);
6491 	else
6492 		e1000e_reset(adapter);
6493 
6494 	return 0;
6495 }
6496 
6497 static const struct net_device_ops e1000e_netdev_ops = {
6498 	.ndo_open		= e1000_open,
6499 	.ndo_stop		= e1000_close,
6500 	.ndo_start_xmit		= e1000_xmit_frame,
6501 	.ndo_get_stats64	= e1000e_get_stats64,
6502 	.ndo_set_rx_mode	= e1000e_set_rx_mode,
6503 	.ndo_set_mac_address	= e1000_set_mac,
6504 	.ndo_change_mtu		= e1000_change_mtu,
6505 	.ndo_do_ioctl		= e1000_ioctl,
6506 	.ndo_tx_timeout		= e1000_tx_timeout,
6507 	.ndo_validate_addr	= eth_validate_addr,
6508 
6509 	.ndo_vlan_rx_add_vid	= e1000_vlan_rx_add_vid,
6510 	.ndo_vlan_rx_kill_vid	= e1000_vlan_rx_kill_vid,
6511 #ifdef CONFIG_NET_POLL_CONTROLLER
6512 	.ndo_poll_controller	= e1000_netpoll,
6513 #endif
6514 	.ndo_set_features = e1000_set_features,
6515 };
6516 
6517 /**
6518  * e1000_probe - Device Initialization Routine
6519  * @pdev: PCI device information struct
6520  * @ent: entry in e1000_pci_tbl
6521  *
6522  * Returns 0 on success, negative on failure
6523  *
6524  * e1000_probe initializes an adapter identified by a pci_dev structure.
6525  * The OS initialization, configuring of the adapter private structure,
6526  * and a hardware reset occur.
6527  **/
6528 static int e1000_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
6529 {
6530 	struct net_device *netdev;
6531 	struct e1000_adapter *adapter;
6532 	struct e1000_hw *hw;
6533 	const struct e1000_info *ei = e1000_info_tbl[ent->driver_data];
6534 	resource_size_t mmio_start, mmio_len;
6535 	resource_size_t flash_start, flash_len;
6536 	static int cards_found;
6537 	u16 aspm_disable_flag = 0;
6538 	int bars, i, err, pci_using_dac;
6539 	u16 eeprom_data = 0;
6540 	u16 eeprom_apme_mask = E1000_EEPROM_APME;
6541 
6542 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L0S)
6543 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6544 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L1)
6545 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6546 	if (aspm_disable_flag)
6547 		e1000e_disable_aspm(pdev, aspm_disable_flag);
6548 
6549 	err = pci_enable_device_mem(pdev);
6550 	if (err)
6551 		return err;
6552 
6553 	pci_using_dac = 0;
6554 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
6555 	if (!err) {
6556 		pci_using_dac = 1;
6557 	} else {
6558 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
6559 		if (err) {
6560 			dev_err(&pdev->dev,
6561 				"No usable DMA configuration, aborting\n");
6562 			goto err_dma;
6563 		}
6564 	}
6565 
6566 	bars = pci_select_bars(pdev, IORESOURCE_MEM);
6567 	err = pci_request_selected_regions_exclusive(pdev, bars,
6568 						     e1000e_driver_name);
6569 	if (err)
6570 		goto err_pci_reg;
6571 
6572 	/* AER (Advanced Error Reporting) hooks */
6573 	pci_enable_pcie_error_reporting(pdev);
6574 
6575 	pci_set_master(pdev);
6576 	/* PCI config space info */
6577 	err = pci_save_state(pdev);
6578 	if (err)
6579 		goto err_alloc_etherdev;
6580 
6581 	err = -ENOMEM;
6582 	netdev = alloc_etherdev(sizeof(struct e1000_adapter));
6583 	if (!netdev)
6584 		goto err_alloc_etherdev;
6585 
6586 	SET_NETDEV_DEV(netdev, &pdev->dev);
6587 
6588 	netdev->irq = pdev->irq;
6589 
6590 	pci_set_drvdata(pdev, netdev);
6591 	adapter = netdev_priv(netdev);
6592 	hw = &adapter->hw;
6593 	adapter->netdev = netdev;
6594 	adapter->pdev = pdev;
6595 	adapter->ei = ei;
6596 	adapter->pba = ei->pba;
6597 	adapter->flags = ei->flags;
6598 	adapter->flags2 = ei->flags2;
6599 	adapter->hw.adapter = adapter;
6600 	adapter->hw.mac.type = ei->mac;
6601 	adapter->max_hw_frame_size = ei->max_hw_frame_size;
6602 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
6603 
6604 	mmio_start = pci_resource_start(pdev, 0);
6605 	mmio_len = pci_resource_len(pdev, 0);
6606 
6607 	err = -EIO;
6608 	adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
6609 	if (!adapter->hw.hw_addr)
6610 		goto err_ioremap;
6611 
6612 	if ((adapter->flags & FLAG_HAS_FLASH) &&
6613 	    (pci_resource_flags(pdev, 1) & IORESOURCE_MEM)) {
6614 		flash_start = pci_resource_start(pdev, 1);
6615 		flash_len = pci_resource_len(pdev, 1);
6616 		adapter->hw.flash_address = ioremap(flash_start, flash_len);
6617 		if (!adapter->hw.flash_address)
6618 			goto err_flashmap;
6619 	}
6620 
6621 	/* Set default EEE advertisement */
6622 	if (adapter->flags2 & FLAG2_HAS_EEE)
6623 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
6624 
6625 	/* construct the net_device struct */
6626 	netdev->netdev_ops = &e1000e_netdev_ops;
6627 	e1000e_set_ethtool_ops(netdev);
6628 	netdev->watchdog_timeo = 5 * HZ;
6629 	netif_napi_add(netdev, &adapter->napi, e1000e_poll, 64);
6630 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
6631 
6632 	netdev->mem_start = mmio_start;
6633 	netdev->mem_end = mmio_start + mmio_len;
6634 
6635 	adapter->bd_number = cards_found++;
6636 
6637 	e1000e_check_options(adapter);
6638 
6639 	/* setup adapter struct */
6640 	err = e1000_sw_init(adapter);
6641 	if (err)
6642 		goto err_sw_init;
6643 
6644 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
6645 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
6646 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
6647 
6648 	err = ei->get_variants(adapter);
6649 	if (err)
6650 		goto err_hw_init;
6651 
6652 	if ((adapter->flags & FLAG_IS_ICH) &&
6653 	    (adapter->flags & FLAG_READ_ONLY_NVM))
6654 		e1000e_write_protect_nvm_ich8lan(&adapter->hw);
6655 
6656 	hw->mac.ops.get_bus_info(&adapter->hw);
6657 
6658 	adapter->hw.phy.autoneg_wait_to_complete = 0;
6659 
6660 	/* Copper options */
6661 	if (adapter->hw.phy.media_type == e1000_media_type_copper) {
6662 		adapter->hw.phy.mdix = AUTO_ALL_MODES;
6663 		adapter->hw.phy.disable_polarity_correction = 0;
6664 		adapter->hw.phy.ms_type = e1000_ms_hw_default;
6665 	}
6666 
6667 	if (hw->phy.ops.check_reset_block && hw->phy.ops.check_reset_block(hw))
6668 		dev_info(&pdev->dev,
6669 			 "PHY reset is blocked due to SOL/IDER session.\n");
6670 
6671 	/* Set initial default active device features */
6672 	netdev->features = (NETIF_F_SG |
6673 			    NETIF_F_HW_VLAN_CTAG_RX |
6674 			    NETIF_F_HW_VLAN_CTAG_TX |
6675 			    NETIF_F_TSO |
6676 			    NETIF_F_TSO6 |
6677 			    NETIF_F_RXHASH |
6678 			    NETIF_F_RXCSUM |
6679 			    NETIF_F_HW_CSUM);
6680 
6681 	/* Set user-changeable features (subset of all device features) */
6682 	netdev->hw_features = netdev->features;
6683 	netdev->hw_features |= NETIF_F_RXFCS;
6684 	netdev->priv_flags |= IFF_SUPP_NOFCS;
6685 	netdev->hw_features |= NETIF_F_RXALL;
6686 
6687 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER)
6688 		netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
6689 
6690 	netdev->vlan_features |= (NETIF_F_SG |
6691 				  NETIF_F_TSO |
6692 				  NETIF_F_TSO6 |
6693 				  NETIF_F_HW_CSUM);
6694 
6695 	netdev->priv_flags |= IFF_UNICAST_FLT;
6696 
6697 	if (pci_using_dac) {
6698 		netdev->features |= NETIF_F_HIGHDMA;
6699 		netdev->vlan_features |= NETIF_F_HIGHDMA;
6700 	}
6701 
6702 	if (e1000e_enable_mng_pass_thru(&adapter->hw))
6703 		adapter->flags |= FLAG_MNG_PT_ENABLED;
6704 
6705 	/* before reading the NVM, reset the controller to
6706 	 * put the device in a known good starting state
6707 	 */
6708 	adapter->hw.mac.ops.reset_hw(&adapter->hw);
6709 
6710 	/* systems with ASPM and others may see the checksum fail on the first
6711 	 * attempt. Let's give it a few tries
6712 	 */
6713 	for (i = 0;; i++) {
6714 		if (e1000_validate_nvm_checksum(&adapter->hw) >= 0)
6715 			break;
6716 		if (i == 2) {
6717 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
6718 			err = -EIO;
6719 			goto err_eeprom;
6720 		}
6721 	}
6722 
6723 	e1000_eeprom_checks(adapter);
6724 
6725 	/* copy the MAC address */
6726 	if (e1000e_read_mac_addr(&adapter->hw))
6727 		dev_err(&pdev->dev,
6728 			"NVM Read Error while reading MAC address\n");
6729 
6730 	memcpy(netdev->dev_addr, adapter->hw.mac.addr, netdev->addr_len);
6731 
6732 	if (!is_valid_ether_addr(netdev->dev_addr)) {
6733 		dev_err(&pdev->dev, "Invalid MAC Address: %pM\n",
6734 			netdev->dev_addr);
6735 		err = -EIO;
6736 		goto err_eeprom;
6737 	}
6738 
6739 	init_timer(&adapter->watchdog_timer);
6740 	adapter->watchdog_timer.function = e1000_watchdog;
6741 	adapter->watchdog_timer.data = (unsigned long)adapter;
6742 
6743 	init_timer(&adapter->phy_info_timer);
6744 	adapter->phy_info_timer.function = e1000_update_phy_info;
6745 	adapter->phy_info_timer.data = (unsigned long)adapter;
6746 
6747 	INIT_WORK(&adapter->reset_task, e1000_reset_task);
6748 	INIT_WORK(&adapter->watchdog_task, e1000_watchdog_task);
6749 	INIT_WORK(&adapter->downshift_task, e1000e_downshift_workaround);
6750 	INIT_WORK(&adapter->update_phy_task, e1000e_update_phy_task);
6751 	INIT_WORK(&adapter->print_hang_task, e1000_print_hw_hang);
6752 
6753 	/* Initialize link parameters. User can change them with ethtool */
6754 	adapter->hw.mac.autoneg = 1;
6755 	adapter->fc_autoneg = true;
6756 	adapter->hw.fc.requested_mode = e1000_fc_default;
6757 	adapter->hw.fc.current_mode = e1000_fc_default;
6758 	adapter->hw.phy.autoneg_advertised = 0x2f;
6759 
6760 	/* Initial Wake on LAN setting - If APM wake is enabled in
6761 	 * the EEPROM, enable the ACPI Magic Packet filter
6762 	 */
6763 	if (adapter->flags & FLAG_APME_IN_WUC) {
6764 		/* APME bit in EEPROM is mapped to WUC.APME */
6765 		eeprom_data = er32(WUC);
6766 		eeprom_apme_mask = E1000_WUC_APME;
6767 		if ((hw->mac.type > e1000_ich10lan) &&
6768 		    (eeprom_data & E1000_WUC_PHY_WAKE))
6769 			adapter->flags2 |= FLAG2_HAS_PHY_WAKEUP;
6770 	} else if (adapter->flags & FLAG_APME_IN_CTRL3) {
6771 		if (adapter->flags & FLAG_APME_CHECK_PORT_B &&
6772 		    (adapter->hw.bus.func == 1))
6773 			e1000_read_nvm(&adapter->hw, NVM_INIT_CONTROL3_PORT_B,
6774 				       1, &eeprom_data);
6775 		else
6776 			e1000_read_nvm(&adapter->hw, NVM_INIT_CONTROL3_PORT_A,
6777 				       1, &eeprom_data);
6778 	}
6779 
6780 	/* fetch WoL from EEPROM */
6781 	if (eeprom_data & eeprom_apme_mask)
6782 		adapter->eeprom_wol |= E1000_WUFC_MAG;
6783 
6784 	/* now that we have the eeprom settings, apply the special cases
6785 	 * where the eeprom may be wrong or the board simply won't support
6786 	 * wake on lan on a particular port
6787 	 */
6788 	if (!(adapter->flags & FLAG_HAS_WOL))
6789 		adapter->eeprom_wol = 0;
6790 
6791 	/* initialize the wol settings based on the eeprom settings */
6792 	adapter->wol = adapter->eeprom_wol;
6793 
6794 	/* make sure adapter isn't asleep if manageability is enabled */
6795 	if (adapter->wol || (adapter->flags & FLAG_MNG_PT_ENABLED) ||
6796 	    (hw->mac.ops.check_mng_mode(hw)))
6797 		device_wakeup_enable(&pdev->dev);
6798 
6799 	/* save off EEPROM version number */
6800 	e1000_read_nvm(&adapter->hw, 5, 1, &adapter->eeprom_vers);
6801 
6802 	/* reset the hardware with the new settings */
6803 	e1000e_reset(adapter);
6804 
6805 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6806 	 * is up.  For all other cases, let the f/w know that the h/w is now
6807 	 * under the control of the driver.
6808 	 */
6809 	if (!(adapter->flags & FLAG_HAS_AMT))
6810 		e1000e_get_hw_control(adapter);
6811 
6812 	strlcpy(netdev->name, "eth%d", sizeof(netdev->name));
6813 	err = register_netdev(netdev);
6814 	if (err)
6815 		goto err_register;
6816 
6817 	/* carrier off reporting is important to ethtool even BEFORE open */
6818 	netif_carrier_off(netdev);
6819 
6820 	/* init PTP hardware clock */
6821 	e1000e_ptp_init(adapter);
6822 
6823 	e1000_print_device_info(adapter);
6824 
6825 	if (pci_dev_run_wake(pdev))
6826 		pm_runtime_put_noidle(&pdev->dev);
6827 
6828 	return 0;
6829 
6830 err_register:
6831 	if (!(adapter->flags & FLAG_HAS_AMT))
6832 		e1000e_release_hw_control(adapter);
6833 err_eeprom:
6834 	if (hw->phy.ops.check_reset_block && !hw->phy.ops.check_reset_block(hw))
6835 		e1000_phy_hw_reset(&adapter->hw);
6836 err_hw_init:
6837 	kfree(adapter->tx_ring);
6838 	kfree(adapter->rx_ring);
6839 err_sw_init:
6840 	if (adapter->hw.flash_address)
6841 		iounmap(adapter->hw.flash_address);
6842 	e1000e_reset_interrupt_capability(adapter);
6843 err_flashmap:
6844 	iounmap(adapter->hw.hw_addr);
6845 err_ioremap:
6846 	free_netdev(netdev);
6847 err_alloc_etherdev:
6848 	pci_release_selected_regions(pdev,
6849 				     pci_select_bars(pdev, IORESOURCE_MEM));
6850 err_pci_reg:
6851 err_dma:
6852 	pci_disable_device(pdev);
6853 	return err;
6854 }
6855 
6856 /**
6857  * e1000_remove - Device Removal Routine
6858  * @pdev: PCI device information struct
6859  *
6860  * e1000_remove is called by the PCI subsystem to alert the driver
6861  * that it should release a PCI device.  The could be caused by a
6862  * Hot-Plug event, or because the driver is going to be removed from
6863  * memory.
6864  **/
6865 static void e1000_remove(struct pci_dev *pdev)
6866 {
6867 	struct net_device *netdev = pci_get_drvdata(pdev);
6868 	struct e1000_adapter *adapter = netdev_priv(netdev);
6869 	bool down = test_bit(__E1000_DOWN, &adapter->state);
6870 
6871 	e1000e_ptp_remove(adapter);
6872 
6873 	/* The timers may be rescheduled, so explicitly disable them
6874 	 * from being rescheduled.
6875 	 */
6876 	if (!down)
6877 		set_bit(__E1000_DOWN, &adapter->state);
6878 	del_timer_sync(&adapter->watchdog_timer);
6879 	del_timer_sync(&adapter->phy_info_timer);
6880 
6881 	cancel_work_sync(&adapter->reset_task);
6882 	cancel_work_sync(&adapter->watchdog_task);
6883 	cancel_work_sync(&adapter->downshift_task);
6884 	cancel_work_sync(&adapter->update_phy_task);
6885 	cancel_work_sync(&adapter->print_hang_task);
6886 
6887 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
6888 		cancel_work_sync(&adapter->tx_hwtstamp_work);
6889 		if (adapter->tx_hwtstamp_skb) {
6890 			dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
6891 			adapter->tx_hwtstamp_skb = NULL;
6892 		}
6893 	}
6894 
6895 	if (!(netdev->flags & IFF_UP))
6896 		e1000_power_down_phy(adapter);
6897 
6898 	/* Don't lie to e1000_close() down the road. */
6899 	if (!down)
6900 		clear_bit(__E1000_DOWN, &adapter->state);
6901 	unregister_netdev(netdev);
6902 
6903 	if (pci_dev_run_wake(pdev))
6904 		pm_runtime_get_noresume(&pdev->dev);
6905 
6906 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
6907 	 * would have already happened in close and is redundant.
6908 	 */
6909 	e1000e_release_hw_control(adapter);
6910 
6911 	e1000e_reset_interrupt_capability(adapter);
6912 	kfree(adapter->tx_ring);
6913 	kfree(adapter->rx_ring);
6914 
6915 	iounmap(adapter->hw.hw_addr);
6916 	if (adapter->hw.flash_address)
6917 		iounmap(adapter->hw.flash_address);
6918 	pci_release_selected_regions(pdev,
6919 				     pci_select_bars(pdev, IORESOURCE_MEM));
6920 
6921 	free_netdev(netdev);
6922 
6923 	/* AER disable */
6924 	pci_disable_pcie_error_reporting(pdev);
6925 
6926 	pci_disable_device(pdev);
6927 }
6928 
6929 /* PCI Error Recovery (ERS) */
6930 static const struct pci_error_handlers e1000_err_handler = {
6931 	.error_detected = e1000_io_error_detected,
6932 	.slot_reset = e1000_io_slot_reset,
6933 	.resume = e1000_io_resume,
6934 };
6935 
6936 static DEFINE_PCI_DEVICE_TABLE(e1000_pci_tbl) = {
6937 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_COPPER), board_82571 },
6938 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_FIBER), board_82571 },
6939 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER), board_82571 },
6940 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER_LP),
6941 	  board_82571 },
6942 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_FIBER), board_82571 },
6943 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES), board_82571 },
6944 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_DUAL), board_82571 },
6945 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_QUAD), board_82571 },
6946 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571PT_QUAD_COPPER), board_82571 },
6947 
6948 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI), board_82572 },
6949 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_COPPER), board_82572 },
6950 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_FIBER), board_82572 },
6951 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_SERDES), board_82572 },
6952 
6953 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E), board_82573 },
6954 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E_IAMT), board_82573 },
6955 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573L), board_82573 },
6956 
6957 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574L), board_82574 },
6958 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574LA), board_82574 },
6959 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82583V), board_82583 },
6960 
6961 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_DPT),
6962 	  board_80003es2lan },
6963 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_SPT),
6964 	  board_80003es2lan },
6965 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_DPT),
6966 	  board_80003es2lan },
6967 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_SPT),
6968 	  board_80003es2lan },
6969 
6970 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE), board_ich8lan },
6971 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_G), board_ich8lan },
6972 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_GT), board_ich8lan },
6973 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_AMT), board_ich8lan },
6974 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_C), board_ich8lan },
6975 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M), board_ich8lan },
6976 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M_AMT), board_ich8lan },
6977 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_82567V_3), board_ich8lan },
6978 
6979 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE), board_ich9lan },
6980 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_G), board_ich9lan },
6981 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_GT), board_ich9lan },
6982 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_AMT), board_ich9lan },
6983 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_C), board_ich9lan },
6984 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_BM), board_ich9lan },
6985 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M), board_ich9lan },
6986 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_AMT), board_ich9lan },
6987 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_V), board_ich9lan },
6988 
6989 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LM), board_ich9lan },
6990 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LF), board_ich9lan },
6991 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_V), board_ich9lan },
6992 
6993 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LM), board_ich10lan },
6994 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LF), board_ich10lan },
6995 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_V), board_ich10lan },
6996 
6997 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LM), board_pchlan },
6998 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LC), board_pchlan },
6999 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DM), board_pchlan },
7000 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DC), board_pchlan },
7001 
7002 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_LM), board_pch2lan },
7003 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_V), board_pch2lan },
7004 
7005 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_LM), board_pch_lpt },
7006 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_V), board_pch_lpt },
7007 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_LM), board_pch_lpt },
7008 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_V), board_pch_lpt },
7009 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM2), board_pch_lpt },
7010 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V2), board_pch_lpt },
7011 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM3), board_pch_lpt },
7012 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V3), board_pch_lpt },
7013 
7014 	{ 0, 0, 0, 0, 0, 0, 0 }	/* terminate list */
7015 };
7016 MODULE_DEVICE_TABLE(pci, e1000_pci_tbl);
7017 
7018 #ifdef CONFIG_PM
7019 static const struct dev_pm_ops e1000_pm_ops = {
7020 	SET_SYSTEM_SLEEP_PM_OPS(e1000_suspend, e1000_resume)
7021 	SET_RUNTIME_PM_OPS(e1000_runtime_suspend, e1000_runtime_resume,
7022 			   e1000_idle)
7023 };
7024 #endif
7025 
7026 /* PCI Device API Driver */
7027 static struct pci_driver e1000_driver = {
7028 	.name     = e1000e_driver_name,
7029 	.id_table = e1000_pci_tbl,
7030 	.probe    = e1000_probe,
7031 	.remove   = e1000_remove,
7032 #ifdef CONFIG_PM
7033 	.driver   = {
7034 		.pm = &e1000_pm_ops,
7035 	},
7036 #endif
7037 	.shutdown = e1000_shutdown,
7038 	.err_handler = &e1000_err_handler
7039 };
7040 
7041 /**
7042  * e1000_init_module - Driver Registration Routine
7043  *
7044  * e1000_init_module is the first routine called when the driver is
7045  * loaded. All it does is register with the PCI subsystem.
7046  **/
7047 static int __init e1000_init_module(void)
7048 {
7049 	int ret;
7050 	pr_info("Intel(R) PRO/1000 Network Driver - %s\n",
7051 		e1000e_driver_version);
7052 	pr_info("Copyright(c) 1999 - 2013 Intel Corporation.\n");
7053 	ret = pci_register_driver(&e1000_driver);
7054 
7055 	return ret;
7056 }
7057 module_init(e1000_init_module);
7058 
7059 /**
7060  * e1000_exit_module - Driver Exit Cleanup Routine
7061  *
7062  * e1000_exit_module is called just before the driver is removed
7063  * from memory.
7064  **/
7065 static void __exit e1000_exit_module(void)
7066 {
7067 	pci_unregister_driver(&e1000_driver);
7068 }
7069 module_exit(e1000_exit_module);
7070 
7071 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
7072 MODULE_DESCRIPTION("Intel(R) PRO/1000 Network Driver");
7073 MODULE_LICENSE("GPL");
7074 MODULE_VERSION(DRV_VERSION);
7075 
7076 /* netdev.c */
7077