xref: /linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision a508da6cc0093171833efb8376b00473f24221b9)
1 /*******************************************************************************
2 
3   Intel 10 Gigabit PCI Express Linux driver
4   Copyright(c) 1999 - 2012 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   e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
24   Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
25 
26 *******************************************************************************/
27 
28 #include <linux/types.h>
29 #include <linux/module.h>
30 #include <linux/pci.h>
31 #include <linux/netdevice.h>
32 #include <linux/vmalloc.h>
33 #include <linux/string.h>
34 #include <linux/in.h>
35 #include <linux/interrupt.h>
36 #include <linux/ip.h>
37 #include <linux/tcp.h>
38 #include <linux/sctp.h>
39 #include <linux/pkt_sched.h>
40 #include <linux/ipv6.h>
41 #include <linux/slab.h>
42 #include <net/checksum.h>
43 #include <net/ip6_checksum.h>
44 #include <linux/ethtool.h>
45 #include <linux/if.h>
46 #include <linux/if_vlan.h>
47 #include <linux/prefetch.h>
48 #include <scsi/fc/fc_fcoe.h>
49 
50 #include "ixgbe.h"
51 #include "ixgbe_common.h"
52 #include "ixgbe_dcb_82599.h"
53 #include "ixgbe_sriov.h"
54 
55 char ixgbe_driver_name[] = "ixgbe";
56 static const char ixgbe_driver_string[] =
57 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
58 #ifdef IXGBE_FCOE
59 char ixgbe_default_device_descr[] =
60 			      "Intel(R) 10 Gigabit Network Connection";
61 #else
62 static char ixgbe_default_device_descr[] =
63 			      "Intel(R) 10 Gigabit Network Connection";
64 #endif
65 #define MAJ 3
66 #define MIN 9
67 #define BUILD 15
68 #define DRV_VERSION __stringify(MAJ) "." __stringify(MIN) "." \
69 	__stringify(BUILD) "-k"
70 const char ixgbe_driver_version[] = DRV_VERSION;
71 static const char ixgbe_copyright[] =
72 				"Copyright (c) 1999-2012 Intel Corporation.";
73 
74 static const struct ixgbe_info *ixgbe_info_tbl[] = {
75 	[board_82598] = &ixgbe_82598_info,
76 	[board_82599] = &ixgbe_82599_info,
77 	[board_X540] = &ixgbe_X540_info,
78 };
79 
80 /* ixgbe_pci_tbl - PCI Device ID Table
81  *
82  * Wildcard entries (PCI_ANY_ID) should come last
83  * Last entry must be all 0s
84  *
85  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
86  *   Class, Class Mask, private data (not used) }
87  */
88 static DEFINE_PCI_DEVICE_TABLE(ixgbe_pci_tbl) = {
89 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
90 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
91 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
92 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
93 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
94 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
95 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
96 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
97 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
98 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
99 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
100 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
101 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
102 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
103 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
104 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
105 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
115 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
116 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
117 	/* required last entry */
118 	{0, }
119 };
120 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
121 
122 #ifdef CONFIG_IXGBE_DCA
123 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
124 			    void *p);
125 static struct notifier_block dca_notifier = {
126 	.notifier_call = ixgbe_notify_dca,
127 	.next          = NULL,
128 	.priority      = 0
129 };
130 #endif
131 
132 #ifdef CONFIG_PCI_IOV
133 static unsigned int max_vfs;
134 module_param(max_vfs, uint, 0);
135 MODULE_PARM_DESC(max_vfs,
136 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63");
137 #endif /* CONFIG_PCI_IOV */
138 
139 static unsigned int allow_unsupported_sfp;
140 module_param(allow_unsupported_sfp, uint, 0);
141 MODULE_PARM_DESC(allow_unsupported_sfp,
142 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
143 
144 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
145 static int debug = -1;
146 module_param(debug, int, 0);
147 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
148 
149 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
150 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
151 MODULE_LICENSE("GPL");
152 MODULE_VERSION(DRV_VERSION);
153 
154 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
155 {
156 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
157 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
158 		schedule_work(&adapter->service_task);
159 }
160 
161 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
162 {
163 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
164 
165 	/* flush memory to make sure state is correct before next watchdog */
166 	smp_mb__before_clear_bit();
167 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
168 }
169 
170 struct ixgbe_reg_info {
171 	u32 ofs;
172 	char *name;
173 };
174 
175 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
176 
177 	/* General Registers */
178 	{IXGBE_CTRL, "CTRL"},
179 	{IXGBE_STATUS, "STATUS"},
180 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
181 
182 	/* Interrupt Registers */
183 	{IXGBE_EICR, "EICR"},
184 
185 	/* RX Registers */
186 	{IXGBE_SRRCTL(0), "SRRCTL"},
187 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
188 	{IXGBE_RDLEN(0), "RDLEN"},
189 	{IXGBE_RDH(0), "RDH"},
190 	{IXGBE_RDT(0), "RDT"},
191 	{IXGBE_RXDCTL(0), "RXDCTL"},
192 	{IXGBE_RDBAL(0), "RDBAL"},
193 	{IXGBE_RDBAH(0), "RDBAH"},
194 
195 	/* TX Registers */
196 	{IXGBE_TDBAL(0), "TDBAL"},
197 	{IXGBE_TDBAH(0), "TDBAH"},
198 	{IXGBE_TDLEN(0), "TDLEN"},
199 	{IXGBE_TDH(0), "TDH"},
200 	{IXGBE_TDT(0), "TDT"},
201 	{IXGBE_TXDCTL(0), "TXDCTL"},
202 
203 	/* List Terminator */
204 	{}
205 };
206 
207 
208 /*
209  * ixgbe_regdump - register printout routine
210  */
211 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
212 {
213 	int i = 0, j = 0;
214 	char rname[16];
215 	u32 regs[64];
216 
217 	switch (reginfo->ofs) {
218 	case IXGBE_SRRCTL(0):
219 		for (i = 0; i < 64; i++)
220 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
221 		break;
222 	case IXGBE_DCA_RXCTRL(0):
223 		for (i = 0; i < 64; i++)
224 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
225 		break;
226 	case IXGBE_RDLEN(0):
227 		for (i = 0; i < 64; i++)
228 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
229 		break;
230 	case IXGBE_RDH(0):
231 		for (i = 0; i < 64; i++)
232 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
233 		break;
234 	case IXGBE_RDT(0):
235 		for (i = 0; i < 64; i++)
236 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
237 		break;
238 	case IXGBE_RXDCTL(0):
239 		for (i = 0; i < 64; i++)
240 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
241 		break;
242 	case IXGBE_RDBAL(0):
243 		for (i = 0; i < 64; i++)
244 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
245 		break;
246 	case IXGBE_RDBAH(0):
247 		for (i = 0; i < 64; i++)
248 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
249 		break;
250 	case IXGBE_TDBAL(0):
251 		for (i = 0; i < 64; i++)
252 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
253 		break;
254 	case IXGBE_TDBAH(0):
255 		for (i = 0; i < 64; i++)
256 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
257 		break;
258 	case IXGBE_TDLEN(0):
259 		for (i = 0; i < 64; i++)
260 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
261 		break;
262 	case IXGBE_TDH(0):
263 		for (i = 0; i < 64; i++)
264 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
265 		break;
266 	case IXGBE_TDT(0):
267 		for (i = 0; i < 64; i++)
268 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
269 		break;
270 	case IXGBE_TXDCTL(0):
271 		for (i = 0; i < 64; i++)
272 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
273 		break;
274 	default:
275 		pr_info("%-15s %08x\n", reginfo->name,
276 			IXGBE_READ_REG(hw, reginfo->ofs));
277 		return;
278 	}
279 
280 	for (i = 0; i < 8; i++) {
281 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i*8, i*8+7);
282 		pr_err("%-15s", rname);
283 		for (j = 0; j < 8; j++)
284 			pr_cont(" %08x", regs[i*8+j]);
285 		pr_cont("\n");
286 	}
287 
288 }
289 
290 /*
291  * ixgbe_dump - Print registers, tx-rings and rx-rings
292  */
293 static void ixgbe_dump(struct ixgbe_adapter *adapter)
294 {
295 	struct net_device *netdev = adapter->netdev;
296 	struct ixgbe_hw *hw = &adapter->hw;
297 	struct ixgbe_reg_info *reginfo;
298 	int n = 0;
299 	struct ixgbe_ring *tx_ring;
300 	struct ixgbe_tx_buffer *tx_buffer;
301 	union ixgbe_adv_tx_desc *tx_desc;
302 	struct my_u0 { u64 a; u64 b; } *u0;
303 	struct ixgbe_ring *rx_ring;
304 	union ixgbe_adv_rx_desc *rx_desc;
305 	struct ixgbe_rx_buffer *rx_buffer_info;
306 	u32 staterr;
307 	int i = 0;
308 
309 	if (!netif_msg_hw(adapter))
310 		return;
311 
312 	/* Print netdevice Info */
313 	if (netdev) {
314 		dev_info(&adapter->pdev->dev, "Net device Info\n");
315 		pr_info("Device Name     state            "
316 			"trans_start      last_rx\n");
317 		pr_info("%-15s %016lX %016lX %016lX\n",
318 			netdev->name,
319 			netdev->state,
320 			netdev->trans_start,
321 			netdev->last_rx);
322 	}
323 
324 	/* Print Registers */
325 	dev_info(&adapter->pdev->dev, "Register Dump\n");
326 	pr_info(" Register Name   Value\n");
327 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
328 	     reginfo->name; reginfo++) {
329 		ixgbe_regdump(hw, reginfo);
330 	}
331 
332 	/* Print TX Ring Summary */
333 	if (!netdev || !netif_running(netdev))
334 		goto exit;
335 
336 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
337 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
338 	for (n = 0; n < adapter->num_tx_queues; n++) {
339 		tx_ring = adapter->tx_ring[n];
340 		tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
341 		pr_info(" %5d %5X %5X %016llX %04X %p %016llX\n",
342 			   n, tx_ring->next_to_use, tx_ring->next_to_clean,
343 			   (u64)dma_unmap_addr(tx_buffer, dma),
344 			   dma_unmap_len(tx_buffer, len),
345 			   tx_buffer->next_to_watch,
346 			   (u64)tx_buffer->time_stamp);
347 	}
348 
349 	/* Print TX Rings */
350 	if (!netif_msg_tx_done(adapter))
351 		goto rx_ring_summary;
352 
353 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
354 
355 	/* Transmit Descriptor Formats
356 	 *
357 	 * Advanced Transmit Descriptor
358 	 *   +--------------------------------------------------------------+
359 	 * 0 |         Buffer Address [63:0]                                |
360 	 *   +--------------------------------------------------------------+
361 	 * 8 |  PAYLEN  | PORTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
362 	 *   +--------------------------------------------------------------+
363 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
364 	 */
365 
366 	for (n = 0; n < adapter->num_tx_queues; n++) {
367 		tx_ring = adapter->tx_ring[n];
368 		pr_info("------------------------------------\n");
369 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
370 		pr_info("------------------------------------\n");
371 		pr_info("T [desc]     [address 63:0  ] "
372 			"[PlPOIdStDDt Ln] [bi->dma       ] "
373 			"leng  ntw timestamp        bi->skb\n");
374 
375 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
376 			tx_desc = IXGBE_TX_DESC(tx_ring, i);
377 			tx_buffer = &tx_ring->tx_buffer_info[i];
378 			u0 = (struct my_u0 *)tx_desc;
379 			pr_info("T [0x%03X]    %016llX %016llX %016llX"
380 				" %04X  %p %016llX %p", i,
381 				le64_to_cpu(u0->a),
382 				le64_to_cpu(u0->b),
383 				(u64)dma_unmap_addr(tx_buffer, dma),
384 				dma_unmap_len(tx_buffer, len),
385 				tx_buffer->next_to_watch,
386 				(u64)tx_buffer->time_stamp,
387 				tx_buffer->skb);
388 			if (i == tx_ring->next_to_use &&
389 				i == tx_ring->next_to_clean)
390 				pr_cont(" NTC/U\n");
391 			else if (i == tx_ring->next_to_use)
392 				pr_cont(" NTU\n");
393 			else if (i == tx_ring->next_to_clean)
394 				pr_cont(" NTC\n");
395 			else
396 				pr_cont("\n");
397 
398 			if (netif_msg_pktdata(adapter) &&
399 			    dma_unmap_len(tx_buffer, len) != 0)
400 				print_hex_dump(KERN_INFO, "",
401 					DUMP_PREFIX_ADDRESS, 16, 1,
402 					phys_to_virt(dma_unmap_addr(tx_buffer,
403 								    dma)),
404 					dma_unmap_len(tx_buffer, len),
405 					true);
406 		}
407 	}
408 
409 	/* Print RX Rings Summary */
410 rx_ring_summary:
411 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
412 	pr_info("Queue [NTU] [NTC]\n");
413 	for (n = 0; n < adapter->num_rx_queues; n++) {
414 		rx_ring = adapter->rx_ring[n];
415 		pr_info("%5d %5X %5X\n",
416 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
417 	}
418 
419 	/* Print RX Rings */
420 	if (!netif_msg_rx_status(adapter))
421 		goto exit;
422 
423 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
424 
425 	/* Advanced Receive Descriptor (Read) Format
426 	 *    63                                           1        0
427 	 *    +-----------------------------------------------------+
428 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
429 	 *    +----------------------------------------------+------+
430 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
431 	 *    +-----------------------------------------------------+
432 	 *
433 	 *
434 	 * Advanced Receive Descriptor (Write-Back) Format
435 	 *
436 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
437 	 *   +------------------------------------------------------+
438 	 * 0 | Packet     IP     |SPH| HDR_LEN   | RSV|Packet|  RSS |
439 	 *   | Checksum   Ident  |   |           |    | Type | Type |
440 	 *   +------------------------------------------------------+
441 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
442 	 *   +------------------------------------------------------+
443 	 *   63       48 47    32 31            20 19               0
444 	 */
445 	for (n = 0; n < adapter->num_rx_queues; n++) {
446 		rx_ring = adapter->rx_ring[n];
447 		pr_info("------------------------------------\n");
448 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
449 		pr_info("------------------------------------\n");
450 		pr_info("R  [desc]      [ PktBuf     A0] "
451 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb] "
452 			"<-- Adv Rx Read format\n");
453 		pr_info("RWB[desc]      [PcsmIpSHl PtRs] "
454 			"[vl er S cks ln] ---------------- [bi->skb] "
455 			"<-- Adv Rx Write-Back format\n");
456 
457 		for (i = 0; i < rx_ring->count; i++) {
458 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
459 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
460 			u0 = (struct my_u0 *)rx_desc;
461 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
462 			if (staterr & IXGBE_RXD_STAT_DD) {
463 				/* Descriptor Done */
464 				pr_info("RWB[0x%03X]     %016llX "
465 					"%016llX ---------------- %p", i,
466 					le64_to_cpu(u0->a),
467 					le64_to_cpu(u0->b),
468 					rx_buffer_info->skb);
469 			} else {
470 				pr_info("R  [0x%03X]     %016llX "
471 					"%016llX %016llX %p", i,
472 					le64_to_cpu(u0->a),
473 					le64_to_cpu(u0->b),
474 					(u64)rx_buffer_info->dma,
475 					rx_buffer_info->skb);
476 
477 				if (netif_msg_pktdata(adapter)) {
478 					print_hex_dump(KERN_INFO, "",
479 					   DUMP_PREFIX_ADDRESS, 16, 1,
480 					   phys_to_virt(rx_buffer_info->dma),
481 					   ixgbe_rx_bufsz(rx_ring), true);
482 				}
483 			}
484 
485 			if (i == rx_ring->next_to_use)
486 				pr_cont(" NTU\n");
487 			else if (i == rx_ring->next_to_clean)
488 				pr_cont(" NTC\n");
489 			else
490 				pr_cont("\n");
491 
492 		}
493 	}
494 
495 exit:
496 	return;
497 }
498 
499 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
500 {
501 	u32 ctrl_ext;
502 
503 	/* Let firmware take over control of h/w */
504 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
505 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
506 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
507 }
508 
509 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
510 {
511 	u32 ctrl_ext;
512 
513 	/* Let firmware know the driver has taken over */
514 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
515 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
516 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
517 }
518 
519 /*
520  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
521  * @adapter: pointer to adapter struct
522  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
523  * @queue: queue to map the corresponding interrupt to
524  * @msix_vector: the vector to map to the corresponding queue
525  *
526  */
527 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
528 			   u8 queue, u8 msix_vector)
529 {
530 	u32 ivar, index;
531 	struct ixgbe_hw *hw = &adapter->hw;
532 	switch (hw->mac.type) {
533 	case ixgbe_mac_82598EB:
534 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
535 		if (direction == -1)
536 			direction = 0;
537 		index = (((direction * 64) + queue) >> 2) & 0x1F;
538 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
539 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
540 		ivar |= (msix_vector << (8 * (queue & 0x3)));
541 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
542 		break;
543 	case ixgbe_mac_82599EB:
544 	case ixgbe_mac_X540:
545 		if (direction == -1) {
546 			/* other causes */
547 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
548 			index = ((queue & 1) * 8);
549 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
550 			ivar &= ~(0xFF << index);
551 			ivar |= (msix_vector << index);
552 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
553 			break;
554 		} else {
555 			/* tx or rx causes */
556 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
557 			index = ((16 * (queue & 1)) + (8 * direction));
558 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
559 			ivar &= ~(0xFF << index);
560 			ivar |= (msix_vector << index);
561 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
562 			break;
563 		}
564 	default:
565 		break;
566 	}
567 }
568 
569 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
570 					  u64 qmask)
571 {
572 	u32 mask;
573 
574 	switch (adapter->hw.mac.type) {
575 	case ixgbe_mac_82598EB:
576 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
577 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
578 		break;
579 	case ixgbe_mac_82599EB:
580 	case ixgbe_mac_X540:
581 		mask = (qmask & 0xFFFFFFFF);
582 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
583 		mask = (qmask >> 32);
584 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
585 		break;
586 	default:
587 		break;
588 	}
589 }
590 
591 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring *ring,
592 				      struct ixgbe_tx_buffer *tx_buffer)
593 {
594 	if (tx_buffer->skb) {
595 		dev_kfree_skb_any(tx_buffer->skb);
596 		if (dma_unmap_len(tx_buffer, len))
597 			dma_unmap_single(ring->dev,
598 					 dma_unmap_addr(tx_buffer, dma),
599 					 dma_unmap_len(tx_buffer, len),
600 					 DMA_TO_DEVICE);
601 	} else if (dma_unmap_len(tx_buffer, len)) {
602 		dma_unmap_page(ring->dev,
603 			       dma_unmap_addr(tx_buffer, dma),
604 			       dma_unmap_len(tx_buffer, len),
605 			       DMA_TO_DEVICE);
606 	}
607 	tx_buffer->next_to_watch = NULL;
608 	tx_buffer->skb = NULL;
609 	dma_unmap_len_set(tx_buffer, len, 0);
610 	/* tx_buffer must be completely set up in the transmit path */
611 }
612 
613 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
614 {
615 	struct ixgbe_hw *hw = &adapter->hw;
616 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
617 	int i;
618 	u32 data;
619 
620 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
621 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
622 		return;
623 
624 	switch (hw->mac.type) {
625 	case ixgbe_mac_82598EB:
626 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
627 		break;
628 	default:
629 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
630 	}
631 	hwstats->lxoffrxc += data;
632 
633 	/* refill credits (no tx hang) if we received xoff */
634 	if (!data)
635 		return;
636 
637 	for (i = 0; i < adapter->num_tx_queues; i++)
638 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
639 			  &adapter->tx_ring[i]->state);
640 }
641 
642 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
643 {
644 	struct ixgbe_hw *hw = &adapter->hw;
645 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
646 	u32 xoff[8] = {0};
647 	int i;
648 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
649 
650 	if (adapter->ixgbe_ieee_pfc)
651 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
652 
653 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
654 		ixgbe_update_xoff_rx_lfc(adapter);
655 		return;
656 	}
657 
658 	/* update stats for each tc, only valid with PFC enabled */
659 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
660 		switch (hw->mac.type) {
661 		case ixgbe_mac_82598EB:
662 			xoff[i] = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
663 			break;
664 		default:
665 			xoff[i] = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
666 		}
667 		hwstats->pxoffrxc[i] += xoff[i];
668 	}
669 
670 	/* disarm tx queues that have received xoff frames */
671 	for (i = 0; i < adapter->num_tx_queues; i++) {
672 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
673 		u8 tc = tx_ring->dcb_tc;
674 
675 		if (xoff[tc])
676 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
677 	}
678 }
679 
680 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
681 {
682 	return ring->stats.packets;
683 }
684 
685 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
686 {
687 	struct ixgbe_adapter *adapter = netdev_priv(ring->netdev);
688 	struct ixgbe_hw *hw = &adapter->hw;
689 
690 	u32 head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
691 	u32 tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
692 
693 	if (head != tail)
694 		return (head < tail) ?
695 			tail - head : (tail + ring->count - head);
696 
697 	return 0;
698 }
699 
700 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
701 {
702 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
703 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
704 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
705 	bool ret = false;
706 
707 	clear_check_for_tx_hang(tx_ring);
708 
709 	/*
710 	 * Check for a hung queue, but be thorough. This verifies
711 	 * that a transmit has been completed since the previous
712 	 * check AND there is at least one packet pending. The
713 	 * ARMED bit is set to indicate a potential hang. The
714 	 * bit is cleared if a pause frame is received to remove
715 	 * false hang detection due to PFC or 802.3x frames. By
716 	 * requiring this to fail twice we avoid races with
717 	 * pfc clearing the ARMED bit and conditions where we
718 	 * run the check_tx_hang logic with a transmit completion
719 	 * pending but without time to complete it yet.
720 	 */
721 	if ((tx_done_old == tx_done) && tx_pending) {
722 		/* make sure it is true for two checks in a row */
723 		ret = test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
724 				       &tx_ring->state);
725 	} else {
726 		/* update completed stats and continue */
727 		tx_ring->tx_stats.tx_done_old = tx_done;
728 		/* reset the countdown */
729 		clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
730 	}
731 
732 	return ret;
733 }
734 
735 /**
736  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
737  * @adapter: driver private struct
738  **/
739 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
740 {
741 
742 	/* Do the reset outside of interrupt context */
743 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
744 		adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
745 		ixgbe_service_event_schedule(adapter);
746 	}
747 }
748 
749 /**
750  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
751  * @q_vector: structure containing interrupt and ring information
752  * @tx_ring: tx ring to clean
753  **/
754 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
755 			       struct ixgbe_ring *tx_ring)
756 {
757 	struct ixgbe_adapter *adapter = q_vector->adapter;
758 	struct ixgbe_tx_buffer *tx_buffer;
759 	union ixgbe_adv_tx_desc *tx_desc;
760 	unsigned int total_bytes = 0, total_packets = 0;
761 	unsigned int budget = q_vector->tx.work_limit;
762 	unsigned int i = tx_ring->next_to_clean;
763 
764 	if (test_bit(__IXGBE_DOWN, &adapter->state))
765 		return true;
766 
767 	tx_buffer = &tx_ring->tx_buffer_info[i];
768 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
769 	i -= tx_ring->count;
770 
771 	do {
772 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
773 
774 		/* if next_to_watch is not set then there is no work pending */
775 		if (!eop_desc)
776 			break;
777 
778 		/* prevent any other reads prior to eop_desc */
779 		rmb();
780 
781 		/* if DD is not set pending work has not been completed */
782 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
783 			break;
784 
785 		/* clear next_to_watch to prevent false hangs */
786 		tx_buffer->next_to_watch = NULL;
787 
788 		/* update the statistics for this packet */
789 		total_bytes += tx_buffer->bytecount;
790 		total_packets += tx_buffer->gso_segs;
791 
792 #ifdef CONFIG_IXGBE_PTP
793 		if (unlikely(tx_buffer->tx_flags &
794 			     IXGBE_TX_FLAGS_TSTAMP))
795 			ixgbe_ptp_tx_hwtstamp(q_vector,
796 					      tx_buffer->skb);
797 
798 #endif
799 		/* free the skb */
800 		dev_kfree_skb_any(tx_buffer->skb);
801 
802 		/* unmap skb header data */
803 		dma_unmap_single(tx_ring->dev,
804 				 dma_unmap_addr(tx_buffer, dma),
805 				 dma_unmap_len(tx_buffer, len),
806 				 DMA_TO_DEVICE);
807 
808 		/* clear tx_buffer data */
809 		tx_buffer->skb = NULL;
810 		dma_unmap_len_set(tx_buffer, len, 0);
811 
812 		/* unmap remaining buffers */
813 		while (tx_desc != eop_desc) {
814 			tx_buffer++;
815 			tx_desc++;
816 			i++;
817 			if (unlikely(!i)) {
818 				i -= tx_ring->count;
819 				tx_buffer = tx_ring->tx_buffer_info;
820 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
821 			}
822 
823 			/* unmap any remaining paged data */
824 			if (dma_unmap_len(tx_buffer, len)) {
825 				dma_unmap_page(tx_ring->dev,
826 					       dma_unmap_addr(tx_buffer, dma),
827 					       dma_unmap_len(tx_buffer, len),
828 					       DMA_TO_DEVICE);
829 				dma_unmap_len_set(tx_buffer, len, 0);
830 			}
831 		}
832 
833 		/* move us one more past the eop_desc for start of next pkt */
834 		tx_buffer++;
835 		tx_desc++;
836 		i++;
837 		if (unlikely(!i)) {
838 			i -= tx_ring->count;
839 			tx_buffer = tx_ring->tx_buffer_info;
840 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
841 		}
842 
843 		/* issue prefetch for next Tx descriptor */
844 		prefetch(tx_desc);
845 
846 		/* update budget accounting */
847 		budget--;
848 	} while (likely(budget));
849 
850 	i += tx_ring->count;
851 	tx_ring->next_to_clean = i;
852 	u64_stats_update_begin(&tx_ring->syncp);
853 	tx_ring->stats.bytes += total_bytes;
854 	tx_ring->stats.packets += total_packets;
855 	u64_stats_update_end(&tx_ring->syncp);
856 	q_vector->tx.total_bytes += total_bytes;
857 	q_vector->tx.total_packets += total_packets;
858 
859 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
860 		/* schedule immediate reset if we believe we hung */
861 		struct ixgbe_hw *hw = &adapter->hw;
862 		e_err(drv, "Detected Tx Unit Hang\n"
863 			"  Tx Queue             <%d>\n"
864 			"  TDH, TDT             <%x>, <%x>\n"
865 			"  next_to_use          <%x>\n"
866 			"  next_to_clean        <%x>\n"
867 			"tx_buffer_info[next_to_clean]\n"
868 			"  time_stamp           <%lx>\n"
869 			"  jiffies              <%lx>\n",
870 			tx_ring->queue_index,
871 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
872 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
873 			tx_ring->next_to_use, i,
874 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
875 
876 		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
877 
878 		e_info(probe,
879 		       "tx hang %d detected on queue %d, resetting adapter\n",
880 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
881 
882 		/* schedule immediate reset if we believe we hung */
883 		ixgbe_tx_timeout_reset(adapter);
884 
885 		/* the adapter is about to reset, no point in enabling stuff */
886 		return true;
887 	}
888 
889 	netdev_tx_completed_queue(txring_txq(tx_ring),
890 				  total_packets, total_bytes);
891 
892 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
893 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
894 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
895 		/* Make sure that anybody stopping the queue after this
896 		 * sees the new next_to_clean.
897 		 */
898 		smp_mb();
899 		if (__netif_subqueue_stopped(tx_ring->netdev,
900 					     tx_ring->queue_index)
901 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
902 			netif_wake_subqueue(tx_ring->netdev,
903 					    tx_ring->queue_index);
904 			++tx_ring->tx_stats.restart_queue;
905 		}
906 	}
907 
908 	return !!budget;
909 }
910 
911 #ifdef CONFIG_IXGBE_DCA
912 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
913 				struct ixgbe_ring *tx_ring,
914 				int cpu)
915 {
916 	struct ixgbe_hw *hw = &adapter->hw;
917 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
918 	u16 reg_offset;
919 
920 	switch (hw->mac.type) {
921 	case ixgbe_mac_82598EB:
922 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
923 		break;
924 	case ixgbe_mac_82599EB:
925 	case ixgbe_mac_X540:
926 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
927 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
928 		break;
929 	default:
930 		/* for unknown hardware do not write register */
931 		return;
932 	}
933 
934 	/*
935 	 * We can enable relaxed ordering for reads, but not writes when
936 	 * DCA is enabled.  This is due to a known issue in some chipsets
937 	 * which will cause the DCA tag to be cleared.
938 	 */
939 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
940 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
941 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
942 
943 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
944 }
945 
946 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
947 				struct ixgbe_ring *rx_ring,
948 				int cpu)
949 {
950 	struct ixgbe_hw *hw = &adapter->hw;
951 	u32 rxctrl = dca3_get_tag(rx_ring->dev, cpu);
952 	u8 reg_idx = rx_ring->reg_idx;
953 
954 
955 	switch (hw->mac.type) {
956 	case ixgbe_mac_82599EB:
957 	case ixgbe_mac_X540:
958 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
959 		break;
960 	default:
961 		break;
962 	}
963 
964 	/*
965 	 * We can enable relaxed ordering for reads, but not writes when
966 	 * DCA is enabled.  This is due to a known issue in some chipsets
967 	 * which will cause the DCA tag to be cleared.
968 	 */
969 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
970 		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
971 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
972 
973 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
974 }
975 
976 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
977 {
978 	struct ixgbe_adapter *adapter = q_vector->adapter;
979 	struct ixgbe_ring *ring;
980 	int cpu = get_cpu();
981 
982 	if (q_vector->cpu == cpu)
983 		goto out_no_update;
984 
985 	ixgbe_for_each_ring(ring, q_vector->tx)
986 		ixgbe_update_tx_dca(adapter, ring, cpu);
987 
988 	ixgbe_for_each_ring(ring, q_vector->rx)
989 		ixgbe_update_rx_dca(adapter, ring, cpu);
990 
991 	q_vector->cpu = cpu;
992 out_no_update:
993 	put_cpu();
994 }
995 
996 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
997 {
998 	int num_q_vectors;
999 	int i;
1000 
1001 	if (!(adapter->flags & IXGBE_FLAG_DCA_ENABLED))
1002 		return;
1003 
1004 	/* always use CB2 mode, difference is masked in the CB driver */
1005 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 2);
1006 
1007 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
1008 		num_q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
1009 	else
1010 		num_q_vectors = 1;
1011 
1012 	for (i = 0; i < num_q_vectors; i++) {
1013 		adapter->q_vector[i]->cpu = -1;
1014 		ixgbe_update_dca(adapter->q_vector[i]);
1015 	}
1016 }
1017 
1018 static int __ixgbe_notify_dca(struct device *dev, void *data)
1019 {
1020 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1021 	unsigned long event = *(unsigned long *)data;
1022 
1023 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1024 		return 0;
1025 
1026 	switch (event) {
1027 	case DCA_PROVIDER_ADD:
1028 		/* if we're already enabled, don't do it again */
1029 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1030 			break;
1031 		if (dca_add_requester(dev) == 0) {
1032 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1033 			ixgbe_setup_dca(adapter);
1034 			break;
1035 		}
1036 		/* Fall Through since DCA is disabled. */
1037 	case DCA_PROVIDER_REMOVE:
1038 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1039 			dca_remove_requester(dev);
1040 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1041 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
1042 		}
1043 		break;
1044 	}
1045 
1046 	return 0;
1047 }
1048 
1049 #endif /* CONFIG_IXGBE_DCA */
1050 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1051 				 union ixgbe_adv_rx_desc *rx_desc,
1052 				 struct sk_buff *skb)
1053 {
1054 	if (ring->netdev->features & NETIF_F_RXHASH)
1055 		skb->rxhash = le32_to_cpu(rx_desc->wb.lower.hi_dword.rss);
1056 }
1057 
1058 #ifdef IXGBE_FCOE
1059 /**
1060  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1061  * @adapter: address of board private structure
1062  * @rx_desc: advanced rx descriptor
1063  *
1064  * Returns : true if it is FCoE pkt
1065  */
1066 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_adapter *adapter,
1067 				    union ixgbe_adv_rx_desc *rx_desc)
1068 {
1069 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1070 
1071 	return (adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
1072 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1073 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1074 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1075 }
1076 
1077 #endif /* IXGBE_FCOE */
1078 /**
1079  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1080  * @ring: structure containing ring specific data
1081  * @rx_desc: current Rx descriptor being processed
1082  * @skb: skb currently being received and modified
1083  **/
1084 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1085 				     union ixgbe_adv_rx_desc *rx_desc,
1086 				     struct sk_buff *skb)
1087 {
1088 	skb_checksum_none_assert(skb);
1089 
1090 	/* Rx csum disabled */
1091 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1092 		return;
1093 
1094 	/* if IP and error */
1095 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1096 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1097 		ring->rx_stats.csum_err++;
1098 		return;
1099 	}
1100 
1101 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1102 		return;
1103 
1104 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1105 		__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1106 
1107 		/*
1108 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1109 		 * checksum errors.
1110 		 */
1111 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1112 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1113 			return;
1114 
1115 		ring->rx_stats.csum_err++;
1116 		return;
1117 	}
1118 
1119 	/* It must be a TCP or UDP packet with a valid checksum */
1120 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1121 }
1122 
1123 static inline void ixgbe_release_rx_desc(struct ixgbe_ring *rx_ring, u32 val)
1124 {
1125 	rx_ring->next_to_use = val;
1126 
1127 	/* update next to alloc since we have filled the ring */
1128 	rx_ring->next_to_alloc = val;
1129 	/*
1130 	 * Force memory writes to complete before letting h/w
1131 	 * know there are new descriptors to fetch.  (Only
1132 	 * applicable for weak-ordered memory model archs,
1133 	 * such as IA-64).
1134 	 */
1135 	wmb();
1136 	writel(val, rx_ring->tail);
1137 }
1138 
1139 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1140 				    struct ixgbe_rx_buffer *bi)
1141 {
1142 	struct page *page = bi->page;
1143 	dma_addr_t dma = bi->dma;
1144 
1145 	/* since we are recycling buffers we should seldom need to alloc */
1146 	if (likely(dma))
1147 		return true;
1148 
1149 	/* alloc new page for storage */
1150 	if (likely(!page)) {
1151 		page = alloc_pages(GFP_ATOMIC | __GFP_COLD,
1152 				   ixgbe_rx_pg_order(rx_ring));
1153 		if (unlikely(!page)) {
1154 			rx_ring->rx_stats.alloc_rx_page_failed++;
1155 			return false;
1156 		}
1157 		bi->page = page;
1158 	}
1159 
1160 	/* map page for use */
1161 	dma = dma_map_page(rx_ring->dev, page, 0,
1162 			   ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1163 
1164 	/*
1165 	 * if mapping failed free memory back to system since
1166 	 * there isn't much point in holding memory we can't use
1167 	 */
1168 	if (dma_mapping_error(rx_ring->dev, dma)) {
1169 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1170 		bi->page = NULL;
1171 
1172 		rx_ring->rx_stats.alloc_rx_page_failed++;
1173 		return false;
1174 	}
1175 
1176 	bi->dma = dma;
1177 	bi->page_offset ^= ixgbe_rx_bufsz(rx_ring);
1178 
1179 	return true;
1180 }
1181 
1182 /**
1183  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1184  * @rx_ring: ring to place buffers on
1185  * @cleaned_count: number of buffers to replace
1186  **/
1187 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1188 {
1189 	union ixgbe_adv_rx_desc *rx_desc;
1190 	struct ixgbe_rx_buffer *bi;
1191 	u16 i = rx_ring->next_to_use;
1192 
1193 	/* nothing to do */
1194 	if (!cleaned_count)
1195 		return;
1196 
1197 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1198 	bi = &rx_ring->rx_buffer_info[i];
1199 	i -= rx_ring->count;
1200 
1201 	do {
1202 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1203 			break;
1204 
1205 		/*
1206 		 * Refresh the desc even if buffer_addrs didn't change
1207 		 * because each write-back erases this info.
1208 		 */
1209 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1210 
1211 		rx_desc++;
1212 		bi++;
1213 		i++;
1214 		if (unlikely(!i)) {
1215 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1216 			bi = rx_ring->rx_buffer_info;
1217 			i -= rx_ring->count;
1218 		}
1219 
1220 		/* clear the hdr_addr for the next_to_use descriptor */
1221 		rx_desc->read.hdr_addr = 0;
1222 
1223 		cleaned_count--;
1224 	} while (cleaned_count);
1225 
1226 	i += rx_ring->count;
1227 
1228 	if (rx_ring->next_to_use != i)
1229 		ixgbe_release_rx_desc(rx_ring, i);
1230 }
1231 
1232 /**
1233  * ixgbe_get_headlen - determine size of header for RSC/LRO/GRO/FCOE
1234  * @data: pointer to the start of the headers
1235  * @max_len: total length of section to find headers in
1236  *
1237  * This function is meant to determine the length of headers that will
1238  * be recognized by hardware for LRO, GRO, and RSC offloads.  The main
1239  * motivation of doing this is to only perform one pull for IPv4 TCP
1240  * packets so that we can do basic things like calculating the gso_size
1241  * based on the average data per packet.
1242  **/
1243 static unsigned int ixgbe_get_headlen(unsigned char *data,
1244 				      unsigned int max_len)
1245 {
1246 	union {
1247 		unsigned char *network;
1248 		/* l2 headers */
1249 		struct ethhdr *eth;
1250 		struct vlan_hdr *vlan;
1251 		/* l3 headers */
1252 		struct iphdr *ipv4;
1253 	} hdr;
1254 	__be16 protocol;
1255 	u8 nexthdr = 0;	/* default to not TCP */
1256 	u8 hlen;
1257 
1258 	/* this should never happen, but better safe than sorry */
1259 	if (max_len < ETH_HLEN)
1260 		return max_len;
1261 
1262 	/* initialize network frame pointer */
1263 	hdr.network = data;
1264 
1265 	/* set first protocol and move network header forward */
1266 	protocol = hdr.eth->h_proto;
1267 	hdr.network += ETH_HLEN;
1268 
1269 	/* handle any vlan tag if present */
1270 	if (protocol == __constant_htons(ETH_P_8021Q)) {
1271 		if ((hdr.network - data) > (max_len - VLAN_HLEN))
1272 			return max_len;
1273 
1274 		protocol = hdr.vlan->h_vlan_encapsulated_proto;
1275 		hdr.network += VLAN_HLEN;
1276 	}
1277 
1278 	/* handle L3 protocols */
1279 	if (protocol == __constant_htons(ETH_P_IP)) {
1280 		if ((hdr.network - data) > (max_len - sizeof(struct iphdr)))
1281 			return max_len;
1282 
1283 		/* access ihl as a u8 to avoid unaligned access on ia64 */
1284 		hlen = (hdr.network[0] & 0x0F) << 2;
1285 
1286 		/* verify hlen meets minimum size requirements */
1287 		if (hlen < sizeof(struct iphdr))
1288 			return hdr.network - data;
1289 
1290 		/* record next protocol */
1291 		nexthdr = hdr.ipv4->protocol;
1292 		hdr.network += hlen;
1293 #ifdef IXGBE_FCOE
1294 	} else if (protocol == __constant_htons(ETH_P_FCOE)) {
1295 		if ((hdr.network - data) > (max_len - FCOE_HEADER_LEN))
1296 			return max_len;
1297 		hdr.network += FCOE_HEADER_LEN;
1298 #endif
1299 	} else {
1300 		return hdr.network - data;
1301 	}
1302 
1303 	/* finally sort out TCP */
1304 	if (nexthdr == IPPROTO_TCP) {
1305 		if ((hdr.network - data) > (max_len - sizeof(struct tcphdr)))
1306 			return max_len;
1307 
1308 		/* access doff as a u8 to avoid unaligned access on ia64 */
1309 		hlen = (hdr.network[12] & 0xF0) >> 2;
1310 
1311 		/* verify hlen meets minimum size requirements */
1312 		if (hlen < sizeof(struct tcphdr))
1313 			return hdr.network - data;
1314 
1315 		hdr.network += hlen;
1316 	}
1317 
1318 	/*
1319 	 * If everything has gone correctly hdr.network should be the
1320 	 * data section of the packet and will be the end of the header.
1321 	 * If not then it probably represents the end of the last recognized
1322 	 * header.
1323 	 */
1324 	if ((hdr.network - data) < max_len)
1325 		return hdr.network - data;
1326 	else
1327 		return max_len;
1328 }
1329 
1330 static void ixgbe_get_rsc_cnt(struct ixgbe_ring *rx_ring,
1331 			      union ixgbe_adv_rx_desc *rx_desc,
1332 			      struct sk_buff *skb)
1333 {
1334 	__le32 rsc_enabled;
1335 	u32 rsc_cnt;
1336 
1337 	if (!ring_is_rsc_enabled(rx_ring))
1338 		return;
1339 
1340 	rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1341 		      cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1342 
1343 	/* If this is an RSC frame rsc_cnt should be non-zero */
1344 	if (!rsc_enabled)
1345 		return;
1346 
1347 	rsc_cnt = le32_to_cpu(rsc_enabled);
1348 	rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1349 
1350 	IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1351 }
1352 
1353 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1354 				   struct sk_buff *skb)
1355 {
1356 	u16 hdr_len = skb_headlen(skb);
1357 
1358 	/* set gso_size to avoid messing up TCP MSS */
1359 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1360 						 IXGBE_CB(skb)->append_cnt);
1361 }
1362 
1363 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1364 				   struct sk_buff *skb)
1365 {
1366 	/* if append_cnt is 0 then frame is not RSC */
1367 	if (!IXGBE_CB(skb)->append_cnt)
1368 		return;
1369 
1370 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1371 	rx_ring->rx_stats.rsc_flush++;
1372 
1373 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1374 
1375 	/* gso_size is computed using append_cnt so always clear it last */
1376 	IXGBE_CB(skb)->append_cnt = 0;
1377 }
1378 
1379 /**
1380  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1381  * @rx_ring: rx descriptor ring packet is being transacted on
1382  * @rx_desc: pointer to the EOP Rx descriptor
1383  * @skb: pointer to current skb being populated
1384  *
1385  * This function checks the ring, descriptor, and packet information in
1386  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1387  * other fields within the skb.
1388  **/
1389 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1390 				     union ixgbe_adv_rx_desc *rx_desc,
1391 				     struct sk_buff *skb)
1392 {
1393 	ixgbe_update_rsc_stats(rx_ring, skb);
1394 
1395 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1396 
1397 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1398 
1399 #ifdef CONFIG_IXGBE_PTP
1400 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_TS))
1401 		ixgbe_ptp_rx_hwtstamp(rx_ring->q_vector, skb);
1402 #endif
1403 
1404 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1405 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1406 		__vlan_hwaccel_put_tag(skb, vid);
1407 	}
1408 
1409 	skb_record_rx_queue(skb, rx_ring->queue_index);
1410 
1411 	skb->protocol = eth_type_trans(skb, rx_ring->netdev);
1412 }
1413 
1414 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1415 			 struct sk_buff *skb)
1416 {
1417 	struct ixgbe_adapter *adapter = q_vector->adapter;
1418 
1419 	if (!(adapter->flags & IXGBE_FLAG_IN_NETPOLL))
1420 		napi_gro_receive(&q_vector->napi, skb);
1421 	else
1422 		netif_rx(skb);
1423 }
1424 
1425 /**
1426  * ixgbe_is_non_eop - process handling of non-EOP buffers
1427  * @rx_ring: Rx ring being processed
1428  * @rx_desc: Rx descriptor for current buffer
1429  * @skb: Current socket buffer containing buffer in progress
1430  *
1431  * This function updates next to clean.  If the buffer is an EOP buffer
1432  * this function exits returning false, otherwise it will place the
1433  * sk_buff in the next buffer to be chained and return true indicating
1434  * that this is in fact a non-EOP buffer.
1435  **/
1436 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1437 			     union ixgbe_adv_rx_desc *rx_desc,
1438 			     struct sk_buff *skb)
1439 {
1440 	u32 ntc = rx_ring->next_to_clean + 1;
1441 
1442 	/* fetch, update, and store next to clean */
1443 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1444 	rx_ring->next_to_clean = ntc;
1445 
1446 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1447 
1448 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1449 		return false;
1450 
1451 	/* append_cnt indicates packet is RSC, if so fetch nextp */
1452 	if (IXGBE_CB(skb)->append_cnt) {
1453 		ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1454 		ntc &= IXGBE_RXDADV_NEXTP_MASK;
1455 		ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1456 	}
1457 
1458 	/* place skb in next buffer to be received */
1459 	rx_ring->rx_buffer_info[ntc].skb = skb;
1460 	rx_ring->rx_stats.non_eop_descs++;
1461 
1462 	return true;
1463 }
1464 
1465 /**
1466  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1467  * @rx_ring: rx descriptor ring packet is being transacted on
1468  * @rx_desc: pointer to the EOP Rx descriptor
1469  * @skb: pointer to current skb being fixed
1470  *
1471  * Check for corrupted packet headers caused by senders on the local L2
1472  * embedded NIC switch not setting up their Tx Descriptors right.  These
1473  * should be very rare.
1474  *
1475  * Also address the case where we are pulling data in on pages only
1476  * and as such no data is present in the skb header.
1477  *
1478  * In addition if skb is not at least 60 bytes we need to pad it so that
1479  * it is large enough to qualify as a valid Ethernet frame.
1480  *
1481  * Returns true if an error was encountered and skb was freed.
1482  **/
1483 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1484 				  union ixgbe_adv_rx_desc *rx_desc,
1485 				  struct sk_buff *skb)
1486 {
1487 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1488 	struct net_device *netdev = rx_ring->netdev;
1489 	unsigned char *va;
1490 	unsigned int pull_len;
1491 
1492 	/* if the page was released unmap it, else just sync our portion */
1493 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1494 		dma_unmap_page(rx_ring->dev, IXGBE_CB(skb)->dma,
1495 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1496 		IXGBE_CB(skb)->page_released = false;
1497 	} else {
1498 		dma_sync_single_range_for_cpu(rx_ring->dev,
1499 					      IXGBE_CB(skb)->dma,
1500 					      frag->page_offset,
1501 					      ixgbe_rx_bufsz(rx_ring),
1502 					      DMA_FROM_DEVICE);
1503 	}
1504 	IXGBE_CB(skb)->dma = 0;
1505 
1506 	/* verify that the packet does not have any known errors */
1507 	if (unlikely(ixgbe_test_staterr(rx_desc,
1508 					IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1509 	    !(netdev->features & NETIF_F_RXALL))) {
1510 		dev_kfree_skb_any(skb);
1511 		return true;
1512 	}
1513 
1514 	/*
1515 	 * it is valid to use page_address instead of kmap since we are
1516 	 * working with pages allocated out of the lomem pool per
1517 	 * alloc_page(GFP_ATOMIC)
1518 	 */
1519 	va = skb_frag_address(frag);
1520 
1521 	/*
1522 	 * we need the header to contain the greater of either ETH_HLEN or
1523 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1524 	 */
1525 	pull_len = skb_frag_size(frag);
1526 	if (pull_len > 256)
1527 		pull_len = ixgbe_get_headlen(va, pull_len);
1528 
1529 	/* align pull length to size of long to optimize memcpy performance */
1530 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1531 
1532 	/* update all of the pointers */
1533 	skb_frag_size_sub(frag, pull_len);
1534 	frag->page_offset += pull_len;
1535 	skb->data_len -= pull_len;
1536 	skb->tail += pull_len;
1537 
1538 	/*
1539 	 * if we sucked the frag empty then we should free it,
1540 	 * if there are other frags here something is screwed up in hardware
1541 	 */
1542 	if (skb_frag_size(frag) == 0) {
1543 		BUG_ON(skb_shinfo(skb)->nr_frags != 1);
1544 		skb_shinfo(skb)->nr_frags = 0;
1545 		__skb_frag_unref(frag);
1546 		skb->truesize -= ixgbe_rx_bufsz(rx_ring);
1547 	}
1548 
1549 	/* if skb_pad returns an error the skb was freed */
1550 	if (unlikely(skb->len < 60)) {
1551 		int pad_len = 60 - skb->len;
1552 
1553 		if (skb_pad(skb, pad_len))
1554 			return true;
1555 		__skb_put(skb, pad_len);
1556 	}
1557 
1558 	return false;
1559 }
1560 
1561 /**
1562  * ixgbe_can_reuse_page - determine if we can reuse a page
1563  * @rx_buffer: pointer to rx_buffer containing the page we want to reuse
1564  *
1565  * Returns true if page can be reused in another Rx buffer
1566  **/
1567 static inline bool ixgbe_can_reuse_page(struct ixgbe_rx_buffer *rx_buffer)
1568 {
1569 	struct page *page = rx_buffer->page;
1570 
1571 	/* if we are only owner of page and it is local we can reuse it */
1572 	return likely(page_count(page) == 1) &&
1573 	       likely(page_to_nid(page) == numa_node_id());
1574 }
1575 
1576 /**
1577  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1578  * @rx_ring: rx descriptor ring to store buffers on
1579  * @old_buff: donor buffer to have page reused
1580  *
1581  * Syncronizes page for reuse by the adapter
1582  **/
1583 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1584 				struct ixgbe_rx_buffer *old_buff)
1585 {
1586 	struct ixgbe_rx_buffer *new_buff;
1587 	u16 nta = rx_ring->next_to_alloc;
1588 	u16 bufsz = ixgbe_rx_bufsz(rx_ring);
1589 
1590 	new_buff = &rx_ring->rx_buffer_info[nta];
1591 
1592 	/* update, and store next to alloc */
1593 	nta++;
1594 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1595 
1596 	/* transfer page from old buffer to new buffer */
1597 	new_buff->page = old_buff->page;
1598 	new_buff->dma = old_buff->dma;
1599 
1600 	/* flip page offset to other buffer and store to new_buff */
1601 	new_buff->page_offset = old_buff->page_offset ^ bufsz;
1602 
1603 	/* sync the buffer for use by the device */
1604 	dma_sync_single_range_for_device(rx_ring->dev, new_buff->dma,
1605 					 new_buff->page_offset, bufsz,
1606 					 DMA_FROM_DEVICE);
1607 
1608 	/* bump ref count on page before it is given to the stack */
1609 	get_page(new_buff->page);
1610 }
1611 
1612 /**
1613  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1614  * @rx_ring: rx descriptor ring to transact packets on
1615  * @rx_buffer: buffer containing page to add
1616  * @rx_desc: descriptor containing length of buffer written by hardware
1617  * @skb: sk_buff to place the data into
1618  *
1619  * This function is based on skb_add_rx_frag.  I would have used that
1620  * function however it doesn't handle the truesize case correctly since we
1621  * are allocating more memory than might be used for a single receive.
1622  **/
1623 static void ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1624 			      struct ixgbe_rx_buffer *rx_buffer,
1625 			      struct sk_buff *skb, int size)
1626 {
1627 	skb_fill_page_desc(skb, skb_shinfo(skb)->nr_frags,
1628 			   rx_buffer->page, rx_buffer->page_offset,
1629 			   size);
1630 	skb->len += size;
1631 	skb->data_len += size;
1632 	skb->truesize += ixgbe_rx_bufsz(rx_ring);
1633 }
1634 
1635 /**
1636  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
1637  * @q_vector: structure containing interrupt and ring information
1638  * @rx_ring: rx descriptor ring to transact packets on
1639  * @budget: Total limit on number of packets to process
1640  *
1641  * This function provides a "bounce buffer" approach to Rx interrupt
1642  * processing.  The advantage to this is that on systems that have
1643  * expensive overhead for IOMMU access this provides a means of avoiding
1644  * it by maintaining the mapping of the page to the syste.
1645  *
1646  * Returns true if all work is completed without reaching budget
1647  **/
1648 static bool ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
1649 			       struct ixgbe_ring *rx_ring,
1650 			       int budget)
1651 {
1652 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1653 #ifdef IXGBE_FCOE
1654 	struct ixgbe_adapter *adapter = q_vector->adapter;
1655 	int ddp_bytes = 0;
1656 #endif /* IXGBE_FCOE */
1657 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
1658 
1659 	do {
1660 		struct ixgbe_rx_buffer *rx_buffer;
1661 		union ixgbe_adv_rx_desc *rx_desc;
1662 		struct sk_buff *skb;
1663 		struct page *page;
1664 		u16 ntc;
1665 
1666 		/* return some buffers to hardware, one at a time is too slow */
1667 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
1668 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
1669 			cleaned_count = 0;
1670 		}
1671 
1672 		ntc = rx_ring->next_to_clean;
1673 		rx_desc = IXGBE_RX_DESC(rx_ring, ntc);
1674 		rx_buffer = &rx_ring->rx_buffer_info[ntc];
1675 
1676 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_DD))
1677 			break;
1678 
1679 		/*
1680 		 * This memory barrier is needed to keep us from reading
1681 		 * any other fields out of the rx_desc until we know the
1682 		 * RXD_STAT_DD bit is set
1683 		 */
1684 		rmb();
1685 
1686 		page = rx_buffer->page;
1687 		prefetchw(page);
1688 
1689 		skb = rx_buffer->skb;
1690 
1691 		if (likely(!skb)) {
1692 			void *page_addr = page_address(page) +
1693 					  rx_buffer->page_offset;
1694 
1695 			/* prefetch first cache line of first page */
1696 			prefetch(page_addr);
1697 #if L1_CACHE_BYTES < 128
1698 			prefetch(page_addr + L1_CACHE_BYTES);
1699 #endif
1700 
1701 			/* allocate a skb to store the frags */
1702 			skb = netdev_alloc_skb_ip_align(rx_ring->netdev,
1703 							IXGBE_RX_HDR_SIZE);
1704 			if (unlikely(!skb)) {
1705 				rx_ring->rx_stats.alloc_rx_buff_failed++;
1706 				break;
1707 			}
1708 
1709 			/*
1710 			 * we will be copying header into skb->data in
1711 			 * pskb_may_pull so it is in our interest to prefetch
1712 			 * it now to avoid a possible cache miss
1713 			 */
1714 			prefetchw(skb->data);
1715 
1716 			/*
1717 			 * Delay unmapping of the first packet. It carries the
1718 			 * header information, HW may still access the header
1719 			 * after the writeback.  Only unmap it when EOP is
1720 			 * reached
1721 			 */
1722 			IXGBE_CB(skb)->dma = rx_buffer->dma;
1723 		} else {
1724 			/* we are reusing so sync this buffer for CPU use */
1725 			dma_sync_single_range_for_cpu(rx_ring->dev,
1726 						      rx_buffer->dma,
1727 						      rx_buffer->page_offset,
1728 						      ixgbe_rx_bufsz(rx_ring),
1729 						      DMA_FROM_DEVICE);
1730 		}
1731 
1732 		/* pull page into skb */
1733 		ixgbe_add_rx_frag(rx_ring, rx_buffer, skb,
1734 				  le16_to_cpu(rx_desc->wb.upper.length));
1735 
1736 		if (ixgbe_can_reuse_page(rx_buffer)) {
1737 			/* hand second half of page back to the ring */
1738 			ixgbe_reuse_rx_page(rx_ring, rx_buffer);
1739 		} else if (IXGBE_CB(skb)->dma == rx_buffer->dma) {
1740 			/* the page has been released from the ring */
1741 			IXGBE_CB(skb)->page_released = true;
1742 		} else {
1743 			/* we are not reusing the buffer so unmap it */
1744 			dma_unmap_page(rx_ring->dev, rx_buffer->dma,
1745 				       ixgbe_rx_pg_size(rx_ring),
1746 				       DMA_FROM_DEVICE);
1747 		}
1748 
1749 		/* clear contents of buffer_info */
1750 		rx_buffer->skb = NULL;
1751 		rx_buffer->dma = 0;
1752 		rx_buffer->page = NULL;
1753 
1754 		ixgbe_get_rsc_cnt(rx_ring, rx_desc, skb);
1755 
1756 		cleaned_count++;
1757 
1758 		/* place incomplete frames back on ring for completion */
1759 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
1760 			continue;
1761 
1762 		/* verify the packet layout is correct */
1763 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
1764 			continue;
1765 
1766 		/* probably a little skewed due to removing CRC */
1767 		total_rx_bytes += skb->len;
1768 		total_rx_packets++;
1769 
1770 		/* populate checksum, timestamp, VLAN, and protocol */
1771 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
1772 
1773 #ifdef IXGBE_FCOE
1774 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
1775 		if (ixgbe_rx_is_fcoe(adapter, rx_desc)) {
1776 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
1777 			if (!ddp_bytes) {
1778 				dev_kfree_skb_any(skb);
1779 				continue;
1780 			}
1781 		}
1782 
1783 #endif /* IXGBE_FCOE */
1784 		ixgbe_rx_skb(q_vector, skb);
1785 
1786 		/* update budget accounting */
1787 		budget--;
1788 	} while (likely(budget));
1789 
1790 #ifdef IXGBE_FCOE
1791 	/* include DDPed FCoE data */
1792 	if (ddp_bytes > 0) {
1793 		unsigned int mss;
1794 
1795 		mss = rx_ring->netdev->mtu - sizeof(struct fcoe_hdr) -
1796 			sizeof(struct fc_frame_header) -
1797 			sizeof(struct fcoe_crc_eof);
1798 		if (mss > 512)
1799 			mss &= ~511;
1800 		total_rx_bytes += ddp_bytes;
1801 		total_rx_packets += DIV_ROUND_UP(ddp_bytes, mss);
1802 	}
1803 
1804 #endif /* IXGBE_FCOE */
1805 	u64_stats_update_begin(&rx_ring->syncp);
1806 	rx_ring->stats.packets += total_rx_packets;
1807 	rx_ring->stats.bytes += total_rx_bytes;
1808 	u64_stats_update_end(&rx_ring->syncp);
1809 	q_vector->rx.total_packets += total_rx_packets;
1810 	q_vector->rx.total_bytes += total_rx_bytes;
1811 
1812 	if (cleaned_count)
1813 		ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
1814 
1815 	return !!budget;
1816 }
1817 
1818 /**
1819  * ixgbe_configure_msix - Configure MSI-X hardware
1820  * @adapter: board private structure
1821  *
1822  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
1823  * interrupts.
1824  **/
1825 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
1826 {
1827 	struct ixgbe_q_vector *q_vector;
1828 	int q_vectors, v_idx;
1829 	u32 mask;
1830 
1831 	q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
1832 
1833 	/* Populate MSIX to EITR Select */
1834 	if (adapter->num_vfs > 32) {
1835 		u32 eitrsel = (1 << (adapter->num_vfs - 32)) - 1;
1836 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
1837 	}
1838 
1839 	/*
1840 	 * Populate the IVAR table and set the ITR values to the
1841 	 * corresponding register.
1842 	 */
1843 	for (v_idx = 0; v_idx < q_vectors; v_idx++) {
1844 		struct ixgbe_ring *ring;
1845 		q_vector = adapter->q_vector[v_idx];
1846 
1847 		ixgbe_for_each_ring(ring, q_vector->rx)
1848 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
1849 
1850 		ixgbe_for_each_ring(ring, q_vector->tx)
1851 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
1852 
1853 		if (q_vector->tx.ring && !q_vector->rx.ring) {
1854 			/* tx only vector */
1855 			if (adapter->tx_itr_setting == 1)
1856 				q_vector->itr = IXGBE_10K_ITR;
1857 			else
1858 				q_vector->itr = adapter->tx_itr_setting;
1859 		} else {
1860 			/* rx or rx/tx vector */
1861 			if (adapter->rx_itr_setting == 1)
1862 				q_vector->itr = IXGBE_20K_ITR;
1863 			else
1864 				q_vector->itr = adapter->rx_itr_setting;
1865 		}
1866 
1867 		ixgbe_write_eitr(q_vector);
1868 	}
1869 
1870 	switch (adapter->hw.mac.type) {
1871 	case ixgbe_mac_82598EB:
1872 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
1873 			       v_idx);
1874 		break;
1875 	case ixgbe_mac_82599EB:
1876 	case ixgbe_mac_X540:
1877 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
1878 		break;
1879 	default:
1880 		break;
1881 	}
1882 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
1883 
1884 	/* set up to autoclear timer, and the vectors */
1885 	mask = IXGBE_EIMS_ENABLE_MASK;
1886 	mask &= ~(IXGBE_EIMS_OTHER |
1887 		  IXGBE_EIMS_MAILBOX |
1888 		  IXGBE_EIMS_LSC);
1889 
1890 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
1891 }
1892 
1893 enum latency_range {
1894 	lowest_latency = 0,
1895 	low_latency = 1,
1896 	bulk_latency = 2,
1897 	latency_invalid = 255
1898 };
1899 
1900 /**
1901  * ixgbe_update_itr - update the dynamic ITR value based on statistics
1902  * @q_vector: structure containing interrupt and ring information
1903  * @ring_container: structure containing ring performance data
1904  *
1905  *      Stores a new ITR value based on packets and byte
1906  *      counts during the last interrupt.  The advantage of per interrupt
1907  *      computation is faster updates and more accurate ITR for the current
1908  *      traffic pattern.  Constants in this function were computed
1909  *      based on theoretical maximum wire speed and thresholds were set based
1910  *      on testing data as well as attempting to minimize response time
1911  *      while increasing bulk throughput.
1912  *      this functionality is controlled by the InterruptThrottleRate module
1913  *      parameter (see ixgbe_param.c)
1914  **/
1915 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
1916 			     struct ixgbe_ring_container *ring_container)
1917 {
1918 	int bytes = ring_container->total_bytes;
1919 	int packets = ring_container->total_packets;
1920 	u32 timepassed_us;
1921 	u64 bytes_perint;
1922 	u8 itr_setting = ring_container->itr;
1923 
1924 	if (packets == 0)
1925 		return;
1926 
1927 	/* simple throttlerate management
1928 	 *   0-10MB/s   lowest (100000 ints/s)
1929 	 *  10-20MB/s   low    (20000 ints/s)
1930 	 *  20-1249MB/s bulk   (8000 ints/s)
1931 	 */
1932 	/* what was last interrupt timeslice? */
1933 	timepassed_us = q_vector->itr >> 2;
1934 	bytes_perint = bytes / timepassed_us; /* bytes/usec */
1935 
1936 	switch (itr_setting) {
1937 	case lowest_latency:
1938 		if (bytes_perint > 10)
1939 			itr_setting = low_latency;
1940 		break;
1941 	case low_latency:
1942 		if (bytes_perint > 20)
1943 			itr_setting = bulk_latency;
1944 		else if (bytes_perint <= 10)
1945 			itr_setting = lowest_latency;
1946 		break;
1947 	case bulk_latency:
1948 		if (bytes_perint <= 20)
1949 			itr_setting = low_latency;
1950 		break;
1951 	}
1952 
1953 	/* clear work counters since we have the values we need */
1954 	ring_container->total_bytes = 0;
1955 	ring_container->total_packets = 0;
1956 
1957 	/* write updated itr to ring container */
1958 	ring_container->itr = itr_setting;
1959 }
1960 
1961 /**
1962  * ixgbe_write_eitr - write EITR register in hardware specific way
1963  * @q_vector: structure containing interrupt and ring information
1964  *
1965  * This function is made to be called by ethtool and by the driver
1966  * when it needs to update EITR registers at runtime.  Hardware
1967  * specific quirks/differences are taken care of here.
1968  */
1969 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
1970 {
1971 	struct ixgbe_adapter *adapter = q_vector->adapter;
1972 	struct ixgbe_hw *hw = &adapter->hw;
1973 	int v_idx = q_vector->v_idx;
1974 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
1975 
1976 	switch (adapter->hw.mac.type) {
1977 	case ixgbe_mac_82598EB:
1978 		/* must write high and low 16 bits to reset counter */
1979 		itr_reg |= (itr_reg << 16);
1980 		break;
1981 	case ixgbe_mac_82599EB:
1982 	case ixgbe_mac_X540:
1983 		/*
1984 		 * set the WDIS bit to not clear the timer bits and cause an
1985 		 * immediate assertion of the interrupt
1986 		 */
1987 		itr_reg |= IXGBE_EITR_CNT_WDIS;
1988 		break;
1989 	default:
1990 		break;
1991 	}
1992 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
1993 }
1994 
1995 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
1996 {
1997 	u32 new_itr = q_vector->itr;
1998 	u8 current_itr;
1999 
2000 	ixgbe_update_itr(q_vector, &q_vector->tx);
2001 	ixgbe_update_itr(q_vector, &q_vector->rx);
2002 
2003 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
2004 
2005 	switch (current_itr) {
2006 	/* counts and packets in update_itr are dependent on these numbers */
2007 	case lowest_latency:
2008 		new_itr = IXGBE_100K_ITR;
2009 		break;
2010 	case low_latency:
2011 		new_itr = IXGBE_20K_ITR;
2012 		break;
2013 	case bulk_latency:
2014 		new_itr = IXGBE_8K_ITR;
2015 		break;
2016 	default:
2017 		break;
2018 	}
2019 
2020 	if (new_itr != q_vector->itr) {
2021 		/* do an exponential smoothing */
2022 		new_itr = (10 * new_itr * q_vector->itr) /
2023 			  ((9 * new_itr) + q_vector->itr);
2024 
2025 		/* save the algorithm value here */
2026 		q_vector->itr = new_itr;
2027 
2028 		ixgbe_write_eitr(q_vector);
2029 	}
2030 }
2031 
2032 /**
2033  * ixgbe_check_overtemp_subtask - check for over temperature
2034  * @adapter: pointer to adapter
2035  **/
2036 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2037 {
2038 	struct ixgbe_hw *hw = &adapter->hw;
2039 	u32 eicr = adapter->interrupt_event;
2040 
2041 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2042 		return;
2043 
2044 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) &&
2045 	    !(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2046 		return;
2047 
2048 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2049 
2050 	switch (hw->device_id) {
2051 	case IXGBE_DEV_ID_82599_T3_LOM:
2052 		/*
2053 		 * Since the warning interrupt is for both ports
2054 		 * we don't have to check if:
2055 		 *  - This interrupt wasn't for our port.
2056 		 *  - We may have missed the interrupt so always have to
2057 		 *    check if we  got a LSC
2058 		 */
2059 		if (!(eicr & IXGBE_EICR_GPI_SDP0) &&
2060 		    !(eicr & IXGBE_EICR_LSC))
2061 			return;
2062 
2063 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2064 			u32 autoneg;
2065 			bool link_up = false;
2066 
2067 			hw->mac.ops.check_link(hw, &autoneg, &link_up, false);
2068 
2069 			if (link_up)
2070 				return;
2071 		}
2072 
2073 		/* Check if this is not due to overtemp */
2074 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2075 			return;
2076 
2077 		break;
2078 	default:
2079 		if (!(eicr & IXGBE_EICR_GPI_SDP0))
2080 			return;
2081 		break;
2082 	}
2083 	e_crit(drv,
2084 	       "Network adapter has been stopped because it has over heated. "
2085 	       "Restart the computer. If the problem persists, "
2086 	       "power off the system and replace the adapter\n");
2087 
2088 	adapter->interrupt_event = 0;
2089 }
2090 
2091 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2092 {
2093 	struct ixgbe_hw *hw = &adapter->hw;
2094 
2095 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2096 	    (eicr & IXGBE_EICR_GPI_SDP1)) {
2097 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2098 		/* write to clear the interrupt */
2099 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2100 	}
2101 }
2102 
2103 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2104 {
2105 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2106 		return;
2107 
2108 	switch (adapter->hw.mac.type) {
2109 	case ixgbe_mac_82599EB:
2110 		/*
2111 		 * Need to check link state so complete overtemp check
2112 		 * on service task
2113 		 */
2114 		if (((eicr & IXGBE_EICR_GPI_SDP0) || (eicr & IXGBE_EICR_LSC)) &&
2115 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2116 			adapter->interrupt_event = eicr;
2117 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2118 			ixgbe_service_event_schedule(adapter);
2119 			return;
2120 		}
2121 		return;
2122 	case ixgbe_mac_X540:
2123 		if (!(eicr & IXGBE_EICR_TS))
2124 			return;
2125 		break;
2126 	default:
2127 		return;
2128 	}
2129 
2130 	e_crit(drv,
2131 	       "Network adapter has been stopped because it has over heated. "
2132 	       "Restart the computer. If the problem persists, "
2133 	       "power off the system and replace the adapter\n");
2134 }
2135 
2136 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2137 {
2138 	struct ixgbe_hw *hw = &adapter->hw;
2139 
2140 	if (eicr & IXGBE_EICR_GPI_SDP2) {
2141 		/* Clear the interrupt */
2142 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP2);
2143 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2144 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2145 			ixgbe_service_event_schedule(adapter);
2146 		}
2147 	}
2148 
2149 	if (eicr & IXGBE_EICR_GPI_SDP1) {
2150 		/* Clear the interrupt */
2151 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2152 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2153 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2154 			ixgbe_service_event_schedule(adapter);
2155 		}
2156 	}
2157 }
2158 
2159 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2160 {
2161 	struct ixgbe_hw *hw = &adapter->hw;
2162 
2163 	adapter->lsc_int++;
2164 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2165 	adapter->link_check_timeout = jiffies;
2166 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2167 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2168 		IXGBE_WRITE_FLUSH(hw);
2169 		ixgbe_service_event_schedule(adapter);
2170 	}
2171 }
2172 
2173 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2174 					   u64 qmask)
2175 {
2176 	u32 mask;
2177 	struct ixgbe_hw *hw = &adapter->hw;
2178 
2179 	switch (hw->mac.type) {
2180 	case ixgbe_mac_82598EB:
2181 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2182 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2183 		break;
2184 	case ixgbe_mac_82599EB:
2185 	case ixgbe_mac_X540:
2186 		mask = (qmask & 0xFFFFFFFF);
2187 		if (mask)
2188 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2189 		mask = (qmask >> 32);
2190 		if (mask)
2191 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2192 		break;
2193 	default:
2194 		break;
2195 	}
2196 	/* skip the flush */
2197 }
2198 
2199 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2200 					    u64 qmask)
2201 {
2202 	u32 mask;
2203 	struct ixgbe_hw *hw = &adapter->hw;
2204 
2205 	switch (hw->mac.type) {
2206 	case ixgbe_mac_82598EB:
2207 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2208 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2209 		break;
2210 	case ixgbe_mac_82599EB:
2211 	case ixgbe_mac_X540:
2212 		mask = (qmask & 0xFFFFFFFF);
2213 		if (mask)
2214 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2215 		mask = (qmask >> 32);
2216 		if (mask)
2217 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2218 		break;
2219 	default:
2220 		break;
2221 	}
2222 	/* skip the flush */
2223 }
2224 
2225 /**
2226  * ixgbe_irq_enable - Enable default interrupt generation settings
2227  * @adapter: board private structure
2228  **/
2229 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2230 				    bool flush)
2231 {
2232 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2233 
2234 	/* don't reenable LSC while waiting for link */
2235 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2236 		mask &= ~IXGBE_EIMS_LSC;
2237 
2238 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2239 		switch (adapter->hw.mac.type) {
2240 		case ixgbe_mac_82599EB:
2241 			mask |= IXGBE_EIMS_GPI_SDP0;
2242 			break;
2243 		case ixgbe_mac_X540:
2244 			mask |= IXGBE_EIMS_TS;
2245 			break;
2246 		default:
2247 			break;
2248 		}
2249 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2250 		mask |= IXGBE_EIMS_GPI_SDP1;
2251 	switch (adapter->hw.mac.type) {
2252 	case ixgbe_mac_82599EB:
2253 		mask |= IXGBE_EIMS_GPI_SDP1;
2254 		mask |= IXGBE_EIMS_GPI_SDP2;
2255 	case ixgbe_mac_X540:
2256 		mask |= IXGBE_EIMS_ECC;
2257 		mask |= IXGBE_EIMS_MAILBOX;
2258 		break;
2259 	default:
2260 		break;
2261 	}
2262 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2263 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
2264 		mask |= IXGBE_EIMS_FLOW_DIR;
2265 
2266 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
2267 	if (queues)
2268 		ixgbe_irq_enable_queues(adapter, ~0);
2269 	if (flush)
2270 		IXGBE_WRITE_FLUSH(&adapter->hw);
2271 }
2272 
2273 static irqreturn_t ixgbe_msix_other(int irq, void *data)
2274 {
2275 	struct ixgbe_adapter *adapter = data;
2276 	struct ixgbe_hw *hw = &adapter->hw;
2277 	u32 eicr;
2278 
2279 	/*
2280 	 * Workaround for Silicon errata.  Use clear-by-write instead
2281 	 * of clear-by-read.  Reading with EICS will return the
2282 	 * interrupt causes without clearing, which later be done
2283 	 * with the write to EICR.
2284 	 */
2285 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2286 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2287 
2288 	if (eicr & IXGBE_EICR_LSC)
2289 		ixgbe_check_lsc(adapter);
2290 
2291 	if (eicr & IXGBE_EICR_MAILBOX)
2292 		ixgbe_msg_task(adapter);
2293 
2294 	switch (hw->mac.type) {
2295 	case ixgbe_mac_82599EB:
2296 	case ixgbe_mac_X540:
2297 		if (eicr & IXGBE_EICR_ECC)
2298 			e_info(link, "Received unrecoverable ECC Err, please "
2299 			       "reboot\n");
2300 		/* Handle Flow Director Full threshold interrupt */
2301 		if (eicr & IXGBE_EICR_FLOW_DIR) {
2302 			int reinit_count = 0;
2303 			int i;
2304 			for (i = 0; i < adapter->num_tx_queues; i++) {
2305 				struct ixgbe_ring *ring = adapter->tx_ring[i];
2306 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
2307 						       &ring->state))
2308 					reinit_count++;
2309 			}
2310 			if (reinit_count) {
2311 				/* no more flow director interrupts until after init */
2312 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
2313 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
2314 				ixgbe_service_event_schedule(adapter);
2315 			}
2316 		}
2317 		ixgbe_check_sfp_event(adapter, eicr);
2318 		ixgbe_check_overtemp_event(adapter, eicr);
2319 		break;
2320 	default:
2321 		break;
2322 	}
2323 
2324 	ixgbe_check_fan_failure(adapter, eicr);
2325 #ifdef CONFIG_IXGBE_PTP
2326 	ixgbe_ptp_check_pps_event(adapter, eicr);
2327 #endif
2328 
2329 	/* re-enable the original interrupt state, no lsc, no queues */
2330 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2331 		ixgbe_irq_enable(adapter, false, false);
2332 
2333 	return IRQ_HANDLED;
2334 }
2335 
2336 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
2337 {
2338 	struct ixgbe_q_vector *q_vector = data;
2339 
2340 	/* EIAM disabled interrupts (on this vector) for us */
2341 
2342 	if (q_vector->rx.ring || q_vector->tx.ring)
2343 		napi_schedule(&q_vector->napi);
2344 
2345 	return IRQ_HANDLED;
2346 }
2347 
2348 /**
2349  * ixgbe_poll - NAPI Rx polling callback
2350  * @napi: structure for representing this polling device
2351  * @budget: how many packets driver is allowed to clean
2352  *
2353  * This function is used for legacy and MSI, NAPI mode
2354  **/
2355 int ixgbe_poll(struct napi_struct *napi, int budget)
2356 {
2357 	struct ixgbe_q_vector *q_vector =
2358 				container_of(napi, struct ixgbe_q_vector, napi);
2359 	struct ixgbe_adapter *adapter = q_vector->adapter;
2360 	struct ixgbe_ring *ring;
2361 	int per_ring_budget;
2362 	bool clean_complete = true;
2363 
2364 #ifdef CONFIG_IXGBE_DCA
2365 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
2366 		ixgbe_update_dca(q_vector);
2367 #endif
2368 
2369 	ixgbe_for_each_ring(ring, q_vector->tx)
2370 		clean_complete &= !!ixgbe_clean_tx_irq(q_vector, ring);
2371 
2372 	/* attempt to distribute budget to each queue fairly, but don't allow
2373 	 * the budget to go below 1 because we'll exit polling */
2374 	if (q_vector->rx.count > 1)
2375 		per_ring_budget = max(budget/q_vector->rx.count, 1);
2376 	else
2377 		per_ring_budget = budget;
2378 
2379 	ixgbe_for_each_ring(ring, q_vector->rx)
2380 		clean_complete &= ixgbe_clean_rx_irq(q_vector, ring,
2381 						     per_ring_budget);
2382 
2383 	/* If all work not completed, return budget and keep polling */
2384 	if (!clean_complete)
2385 		return budget;
2386 
2387 	/* all work done, exit the polling mode */
2388 	napi_complete(napi);
2389 	if (adapter->rx_itr_setting & 1)
2390 		ixgbe_set_itr(q_vector);
2391 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2392 		ixgbe_irq_enable_queues(adapter, ((u64)1 << q_vector->v_idx));
2393 
2394 	return 0;
2395 }
2396 
2397 /**
2398  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2399  * @adapter: board private structure
2400  *
2401  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2402  * interrupts from the kernel.
2403  **/
2404 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
2405 {
2406 	struct net_device *netdev = adapter->netdev;
2407 	int q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
2408 	int vector, err;
2409 	int ri = 0, ti = 0;
2410 
2411 	for (vector = 0; vector < q_vectors; vector++) {
2412 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2413 		struct msix_entry *entry = &adapter->msix_entries[vector];
2414 
2415 		if (q_vector->tx.ring && q_vector->rx.ring) {
2416 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2417 				 "%s-%s-%d", netdev->name, "TxRx", ri++);
2418 			ti++;
2419 		} else if (q_vector->rx.ring) {
2420 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2421 				 "%s-%s-%d", netdev->name, "rx", ri++);
2422 		} else if (q_vector->tx.ring) {
2423 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2424 				 "%s-%s-%d", netdev->name, "tx", ti++);
2425 		} else {
2426 			/* skip this unused q_vector */
2427 			continue;
2428 		}
2429 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
2430 				  q_vector->name, q_vector);
2431 		if (err) {
2432 			e_err(probe, "request_irq failed for MSIX interrupt "
2433 			      "Error: %d\n", err);
2434 			goto free_queue_irqs;
2435 		}
2436 		/* If Flow Director is enabled, set interrupt affinity */
2437 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2438 			/* assign the mask for this irq */
2439 			irq_set_affinity_hint(entry->vector,
2440 					      &q_vector->affinity_mask);
2441 		}
2442 	}
2443 
2444 	err = request_irq(adapter->msix_entries[vector].vector,
2445 			  ixgbe_msix_other, 0, netdev->name, adapter);
2446 	if (err) {
2447 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
2448 		goto free_queue_irqs;
2449 	}
2450 
2451 	return 0;
2452 
2453 free_queue_irqs:
2454 	while (vector) {
2455 		vector--;
2456 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
2457 				      NULL);
2458 		free_irq(adapter->msix_entries[vector].vector,
2459 			 adapter->q_vector[vector]);
2460 	}
2461 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
2462 	pci_disable_msix(adapter->pdev);
2463 	kfree(adapter->msix_entries);
2464 	adapter->msix_entries = NULL;
2465 	return err;
2466 }
2467 
2468 /**
2469  * ixgbe_intr - legacy mode Interrupt Handler
2470  * @irq: interrupt number
2471  * @data: pointer to a network interface device structure
2472  **/
2473 static irqreturn_t ixgbe_intr(int irq, void *data)
2474 {
2475 	struct ixgbe_adapter *adapter = data;
2476 	struct ixgbe_hw *hw = &adapter->hw;
2477 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2478 	u32 eicr;
2479 
2480 	/*
2481 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
2482 	 * before the read of EICR.
2483 	 */
2484 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
2485 
2486 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2487 	 * therefore no explicit interrupt disable is necessary */
2488 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2489 	if (!eicr) {
2490 		/*
2491 		 * shared interrupt alert!
2492 		 * make sure interrupts are enabled because the read will
2493 		 * have disabled interrupts due to EIAM
2494 		 * finish the workaround of silicon errata on 82598.  Unmask
2495 		 * the interrupt that we masked before the EICR read.
2496 		 */
2497 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
2498 			ixgbe_irq_enable(adapter, true, true);
2499 		return IRQ_NONE;	/* Not our interrupt */
2500 	}
2501 
2502 	if (eicr & IXGBE_EICR_LSC)
2503 		ixgbe_check_lsc(adapter);
2504 
2505 	switch (hw->mac.type) {
2506 	case ixgbe_mac_82599EB:
2507 		ixgbe_check_sfp_event(adapter, eicr);
2508 		/* Fall through */
2509 	case ixgbe_mac_X540:
2510 		if (eicr & IXGBE_EICR_ECC)
2511 			e_info(link, "Received unrecoverable ECC err, please "
2512 				     "reboot\n");
2513 		ixgbe_check_overtemp_event(adapter, eicr);
2514 		break;
2515 	default:
2516 		break;
2517 	}
2518 
2519 	ixgbe_check_fan_failure(adapter, eicr);
2520 #ifdef CONFIG_IXGBE_PTP
2521 	ixgbe_ptp_check_pps_event(adapter, eicr);
2522 #endif
2523 
2524 	/* would disable interrupts here but EIAM disabled it */
2525 	napi_schedule(&q_vector->napi);
2526 
2527 	/*
2528 	 * re-enable link(maybe) and non-queue interrupts, no flush.
2529 	 * ixgbe_poll will re-enable the queue interrupts
2530 	 */
2531 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2532 		ixgbe_irq_enable(adapter, false, false);
2533 
2534 	return IRQ_HANDLED;
2535 }
2536 
2537 /**
2538  * ixgbe_request_irq - initialize interrupts
2539  * @adapter: board private structure
2540  *
2541  * Attempts to configure interrupts using the best available
2542  * capabilities of the hardware and kernel.
2543  **/
2544 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
2545 {
2546 	struct net_device *netdev = adapter->netdev;
2547 	int err;
2548 
2549 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
2550 		err = ixgbe_request_msix_irqs(adapter);
2551 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
2552 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
2553 				  netdev->name, adapter);
2554 	else
2555 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
2556 				  netdev->name, adapter);
2557 
2558 	if (err)
2559 		e_err(probe, "request_irq failed, Error %d\n", err);
2560 
2561 	return err;
2562 }
2563 
2564 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
2565 {
2566 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
2567 		int i, q_vectors;
2568 
2569 		q_vectors = adapter->num_msix_vectors;
2570 		i = q_vectors - 1;
2571 		free_irq(adapter->msix_entries[i].vector, adapter);
2572 		i--;
2573 
2574 		for (; i >= 0; i--) {
2575 			/* free only the irqs that were actually requested */
2576 			if (!adapter->q_vector[i]->rx.ring &&
2577 			    !adapter->q_vector[i]->tx.ring)
2578 				continue;
2579 
2580 			/* clear the affinity_mask in the IRQ descriptor */
2581 			irq_set_affinity_hint(adapter->msix_entries[i].vector,
2582 					      NULL);
2583 
2584 			free_irq(adapter->msix_entries[i].vector,
2585 				 adapter->q_vector[i]);
2586 		}
2587 	} else {
2588 		free_irq(adapter->pdev->irq, adapter);
2589 	}
2590 }
2591 
2592 /**
2593  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
2594  * @adapter: board private structure
2595  **/
2596 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
2597 {
2598 	switch (adapter->hw.mac.type) {
2599 	case ixgbe_mac_82598EB:
2600 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
2601 		break;
2602 	case ixgbe_mac_82599EB:
2603 	case ixgbe_mac_X540:
2604 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
2605 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
2606 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
2607 		break;
2608 	default:
2609 		break;
2610 	}
2611 	IXGBE_WRITE_FLUSH(&adapter->hw);
2612 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
2613 		int i;
2614 		for (i = 0; i < adapter->num_msix_vectors; i++)
2615 			synchronize_irq(adapter->msix_entries[i].vector);
2616 	} else {
2617 		synchronize_irq(adapter->pdev->irq);
2618 	}
2619 }
2620 
2621 /**
2622  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
2623  *
2624  **/
2625 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
2626 {
2627 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2628 
2629 	/* rx/tx vector */
2630 	if (adapter->rx_itr_setting == 1)
2631 		q_vector->itr = IXGBE_20K_ITR;
2632 	else
2633 		q_vector->itr = adapter->rx_itr_setting;
2634 
2635 	ixgbe_write_eitr(q_vector);
2636 
2637 	ixgbe_set_ivar(adapter, 0, 0, 0);
2638 	ixgbe_set_ivar(adapter, 1, 0, 0);
2639 
2640 	e_info(hw, "Legacy interrupt IVAR setup done\n");
2641 }
2642 
2643 /**
2644  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
2645  * @adapter: board private structure
2646  * @ring: structure containing ring specific data
2647  *
2648  * Configure the Tx descriptor ring after a reset.
2649  **/
2650 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
2651 			     struct ixgbe_ring *ring)
2652 {
2653 	struct ixgbe_hw *hw = &adapter->hw;
2654 	u64 tdba = ring->dma;
2655 	int wait_loop = 10;
2656 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
2657 	u8 reg_idx = ring->reg_idx;
2658 
2659 	/* disable queue to avoid issues while updating state */
2660 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
2661 	IXGBE_WRITE_FLUSH(hw);
2662 
2663 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
2664 			(tdba & DMA_BIT_MASK(32)));
2665 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
2666 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
2667 			ring->count * sizeof(union ixgbe_adv_tx_desc));
2668 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
2669 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
2670 	ring->tail = hw->hw_addr + IXGBE_TDT(reg_idx);
2671 
2672 	/*
2673 	 * set WTHRESH to encourage burst writeback, it should not be set
2674 	 * higher than 1 when ITR is 0 as it could cause false TX hangs
2675 	 *
2676 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
2677 	 * to or less than the number of on chip descriptors, which is
2678 	 * currently 40.
2679 	 */
2680 	if (!ring->q_vector || (ring->q_vector->itr < 8))
2681 		txdctl |= (1 << 16);	/* WTHRESH = 1 */
2682 	else
2683 		txdctl |= (8 << 16);	/* WTHRESH = 8 */
2684 
2685 	/*
2686 	 * Setting PTHRESH to 32 both improves performance
2687 	 * and avoids a TX hang with DFP enabled
2688 	 */
2689 	txdctl |= (1 << 8) |	/* HTHRESH = 1 */
2690 		   32;		/* PTHRESH = 32 */
2691 
2692 	/* reinitialize flowdirector state */
2693 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2694 	    adapter->atr_sample_rate) {
2695 		ring->atr_sample_rate = adapter->atr_sample_rate;
2696 		ring->atr_count = 0;
2697 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
2698 	} else {
2699 		ring->atr_sample_rate = 0;
2700 	}
2701 
2702 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
2703 
2704 	/* enable queue */
2705 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
2706 
2707 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
2708 	if (hw->mac.type == ixgbe_mac_82598EB &&
2709 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
2710 		return;
2711 
2712 	/* poll to verify queue is enabled */
2713 	do {
2714 		usleep_range(1000, 2000);
2715 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
2716 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
2717 	if (!wait_loop)
2718 		e_err(drv, "Could not enable Tx Queue %d\n", reg_idx);
2719 }
2720 
2721 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
2722 {
2723 	struct ixgbe_hw *hw = &adapter->hw;
2724 	u32 rttdcs;
2725 	u32 reg;
2726 	u8 tcs = netdev_get_num_tc(adapter->netdev);
2727 
2728 	if (hw->mac.type == ixgbe_mac_82598EB)
2729 		return;
2730 
2731 	/* disable the arbiter while setting MTQC */
2732 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
2733 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
2734 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
2735 
2736 	/* set transmit pool layout */
2737 	switch (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
2738 	case (IXGBE_FLAG_SRIOV_ENABLED):
2739 		IXGBE_WRITE_REG(hw, IXGBE_MTQC,
2740 				(IXGBE_MTQC_VT_ENA | IXGBE_MTQC_64VF));
2741 		break;
2742 	default:
2743 		if (!tcs)
2744 			reg = IXGBE_MTQC_64Q_1PB;
2745 		else if (tcs <= 4)
2746 			reg = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
2747 		else
2748 			reg = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
2749 
2750 		IXGBE_WRITE_REG(hw, IXGBE_MTQC, reg);
2751 
2752 		/* Enable Security TX Buffer IFG for multiple pb */
2753 		if (tcs) {
2754 			reg = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
2755 			reg |= IXGBE_SECTX_DCB;
2756 			IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, reg);
2757 		}
2758 		break;
2759 	}
2760 
2761 	/* re-enable the arbiter */
2762 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
2763 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
2764 }
2765 
2766 /**
2767  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
2768  * @adapter: board private structure
2769  *
2770  * Configure the Tx unit of the MAC after a reset.
2771  **/
2772 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
2773 {
2774 	struct ixgbe_hw *hw = &adapter->hw;
2775 	u32 dmatxctl;
2776 	u32 i;
2777 
2778 	ixgbe_setup_mtqc(adapter);
2779 
2780 	if (hw->mac.type != ixgbe_mac_82598EB) {
2781 		/* DMATXCTL.EN must be before Tx queues are enabled */
2782 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2783 		dmatxctl |= IXGBE_DMATXCTL_TE;
2784 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
2785 	}
2786 
2787 	/* Setup the HW Tx Head and Tail descriptor pointers */
2788 	for (i = 0; i < adapter->num_tx_queues; i++)
2789 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
2790 }
2791 
2792 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
2793 				 struct ixgbe_ring *ring)
2794 {
2795 	struct ixgbe_hw *hw = &adapter->hw;
2796 	u8 reg_idx = ring->reg_idx;
2797 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
2798 
2799 	srrctl |= IXGBE_SRRCTL_DROP_EN;
2800 
2801 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
2802 }
2803 
2804 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
2805 				  struct ixgbe_ring *ring)
2806 {
2807 	struct ixgbe_hw *hw = &adapter->hw;
2808 	u8 reg_idx = ring->reg_idx;
2809 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
2810 
2811 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
2812 
2813 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
2814 }
2815 
2816 #ifdef CONFIG_IXGBE_DCB
2817 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
2818 #else
2819 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
2820 #endif
2821 {
2822 	int i;
2823 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
2824 
2825 	if (adapter->ixgbe_ieee_pfc)
2826 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
2827 
2828 	/*
2829 	 * We should set the drop enable bit if:
2830 	 *  SR-IOV is enabled
2831 	 *   or
2832 	 *  Number of Rx queues > 1 and flow control is disabled
2833 	 *
2834 	 *  This allows us to avoid head of line blocking for security
2835 	 *  and performance reasons.
2836 	 */
2837 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
2838 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
2839 		for (i = 0; i < adapter->num_rx_queues; i++)
2840 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
2841 	} else {
2842 		for (i = 0; i < adapter->num_rx_queues; i++)
2843 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
2844 	}
2845 }
2846 
2847 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
2848 
2849 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
2850 				   struct ixgbe_ring *rx_ring)
2851 {
2852 	u32 srrctl;
2853 	u8 reg_idx = rx_ring->reg_idx;
2854 
2855 	switch (adapter->hw.mac.type) {
2856 	case ixgbe_mac_82598EB: {
2857 		struct ixgbe_ring_feature *feature = adapter->ring_feature;
2858 		const int mask = feature[RING_F_RSS].mask;
2859 		reg_idx = reg_idx & mask;
2860 	}
2861 		break;
2862 	case ixgbe_mac_82599EB:
2863 	case ixgbe_mac_X540:
2864 	default:
2865 		break;
2866 	}
2867 
2868 	srrctl = IXGBE_READ_REG(&adapter->hw, IXGBE_SRRCTL(reg_idx));
2869 
2870 	srrctl &= ~IXGBE_SRRCTL_BSIZEHDR_MASK;
2871 	srrctl &= ~IXGBE_SRRCTL_BSIZEPKT_MASK;
2872 	if (adapter->num_vfs)
2873 		srrctl |= IXGBE_SRRCTL_DROP_EN;
2874 
2875 	srrctl |= (IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT) &
2876 		  IXGBE_SRRCTL_BSIZEHDR_MASK;
2877 
2878 #if PAGE_SIZE > IXGBE_MAX_RXBUFFER
2879 	srrctl |= IXGBE_MAX_RXBUFFER >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
2880 #else
2881 	srrctl |= ixgbe_rx_bufsz(rx_ring) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
2882 #endif
2883 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
2884 
2885 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_SRRCTL(reg_idx), srrctl);
2886 }
2887 
2888 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
2889 {
2890 	struct ixgbe_hw *hw = &adapter->hw;
2891 	static const u32 seed[10] = { 0xE291D73D, 0x1805EC6C, 0x2A94B30D,
2892 			  0xA54F2BEC, 0xEA49AF7C, 0xE214AD3D, 0xB855AABE,
2893 			  0x6A3E67EA, 0x14364D17, 0x3BED200D};
2894 	u32 mrqc = 0, reta = 0;
2895 	u32 rxcsum;
2896 	int i, j;
2897 	u8 tcs = netdev_get_num_tc(adapter->netdev);
2898 	int maxq = adapter->ring_feature[RING_F_RSS].indices;
2899 
2900 	if (tcs)
2901 		maxq = min(maxq, adapter->num_tx_queues / tcs);
2902 
2903 	/* Fill out hash function seeds */
2904 	for (i = 0; i < 10; i++)
2905 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), seed[i]);
2906 
2907 	/* Fill out redirection table */
2908 	for (i = 0, j = 0; i < 128; i++, j++) {
2909 		if (j == maxq)
2910 			j = 0;
2911 		/* reta = 4-byte sliding window of
2912 		 * 0x00..(indices-1)(indices-1)00..etc. */
2913 		reta = (reta << 8) | (j * 0x11);
2914 		if ((i & 3) == 3)
2915 			IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
2916 	}
2917 
2918 	/* Disable indicating checksum in descriptor, enables RSS hash */
2919 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
2920 	rxcsum |= IXGBE_RXCSUM_PCSD;
2921 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
2922 
2923 	if (adapter->hw.mac.type == ixgbe_mac_82598EB &&
2924 	    (adapter->flags & IXGBE_FLAG_RSS_ENABLED)) {
2925 		mrqc = IXGBE_MRQC_RSSEN;
2926 	} else {
2927 		int mask = adapter->flags & (IXGBE_FLAG_RSS_ENABLED
2928 					     | IXGBE_FLAG_SRIOV_ENABLED);
2929 
2930 		switch (mask) {
2931 		case (IXGBE_FLAG_RSS_ENABLED):
2932 			if (!tcs)
2933 				mrqc = IXGBE_MRQC_RSSEN;
2934 			else if (tcs <= 4)
2935 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
2936 			else
2937 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
2938 			break;
2939 		case (IXGBE_FLAG_SRIOV_ENABLED):
2940 			mrqc = IXGBE_MRQC_VMDQEN;
2941 			break;
2942 		default:
2943 			break;
2944 		}
2945 	}
2946 
2947 	/* Perform hash on these packet types */
2948 	mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4
2949 	      | IXGBE_MRQC_RSS_FIELD_IPV4_TCP
2950 	      | IXGBE_MRQC_RSS_FIELD_IPV6
2951 	      | IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
2952 
2953 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
2954 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
2955 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
2956 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
2957 
2958 	IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
2959 }
2960 
2961 /**
2962  * ixgbe_configure_rscctl - enable RSC for the indicated ring
2963  * @adapter:    address of board private structure
2964  * @index:      index of ring to set
2965  **/
2966 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
2967 				   struct ixgbe_ring *ring)
2968 {
2969 	struct ixgbe_hw *hw = &adapter->hw;
2970 	u32 rscctrl;
2971 	u8 reg_idx = ring->reg_idx;
2972 
2973 	if (!ring_is_rsc_enabled(ring))
2974 		return;
2975 
2976 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
2977 	rscctrl |= IXGBE_RSCCTL_RSCEN;
2978 	/*
2979 	 * we must limit the number of descriptors so that the
2980 	 * total size of max desc * buf_len is not greater
2981 	 * than 65536
2982 	 */
2983 #if (PAGE_SIZE <= 8192)
2984 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
2985 #elif (PAGE_SIZE <= 16384)
2986 	rscctrl |= IXGBE_RSCCTL_MAXDESC_8;
2987 #else
2988 	rscctrl |= IXGBE_RSCCTL_MAXDESC_4;
2989 #endif
2990 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
2991 }
2992 
2993 #define IXGBE_MAX_RX_DESC_POLL 10
2994 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
2995 				       struct ixgbe_ring *ring)
2996 {
2997 	struct ixgbe_hw *hw = &adapter->hw;
2998 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
2999 	u32 rxdctl;
3000 	u8 reg_idx = ring->reg_idx;
3001 
3002 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3003 	if (hw->mac.type == ixgbe_mac_82598EB &&
3004 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3005 		return;
3006 
3007 	do {
3008 		usleep_range(1000, 2000);
3009 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3010 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
3011 
3012 	if (!wait_loop) {
3013 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
3014 		      "the polling period\n", reg_idx);
3015 	}
3016 }
3017 
3018 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
3019 			    struct ixgbe_ring *ring)
3020 {
3021 	struct ixgbe_hw *hw = &adapter->hw;
3022 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3023 	u32 rxdctl;
3024 	u8 reg_idx = ring->reg_idx;
3025 
3026 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3027 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
3028 
3029 	/* write value back with RXDCTL.ENABLE bit cleared */
3030 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3031 
3032 	if (hw->mac.type == ixgbe_mac_82598EB &&
3033 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3034 		return;
3035 
3036 	/* the hardware may take up to 100us to really disable the rx queue */
3037 	do {
3038 		udelay(10);
3039 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3040 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
3041 
3042 	if (!wait_loop) {
3043 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3044 		      "the polling period\n", reg_idx);
3045 	}
3046 }
3047 
3048 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
3049 			     struct ixgbe_ring *ring)
3050 {
3051 	struct ixgbe_hw *hw = &adapter->hw;
3052 	u64 rdba = ring->dma;
3053 	u32 rxdctl;
3054 	u8 reg_idx = ring->reg_idx;
3055 
3056 	/* disable queue to avoid issues while updating state */
3057 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3058 	ixgbe_disable_rx_queue(adapter, ring);
3059 
3060 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
3061 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
3062 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
3063 			ring->count * sizeof(union ixgbe_adv_rx_desc));
3064 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
3065 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
3066 	ring->tail = hw->hw_addr + IXGBE_RDT(reg_idx);
3067 
3068 	ixgbe_configure_srrctl(adapter, ring);
3069 	ixgbe_configure_rscctl(adapter, ring);
3070 
3071 	/* If operating in IOV mode set RLPML for X540 */
3072 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
3073 	    hw->mac.type == ixgbe_mac_X540) {
3074 		rxdctl &= ~IXGBE_RXDCTL_RLPMLMASK;
3075 		rxdctl |= ((ring->netdev->mtu + ETH_HLEN +
3076 			    ETH_FCS_LEN + VLAN_HLEN) | IXGBE_RXDCTL_RLPML_EN);
3077 	}
3078 
3079 	if (hw->mac.type == ixgbe_mac_82598EB) {
3080 		/*
3081 		 * enable cache line friendly hardware writes:
3082 		 * PTHRESH=32 descriptors (half the internal cache),
3083 		 * this also removes ugly rx_no_buffer_count increment
3084 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
3085 		 * WTHRESH=8 burst writeback up to two cache lines
3086 		 */
3087 		rxdctl &= ~0x3FFFFF;
3088 		rxdctl |=  0x080420;
3089 	}
3090 
3091 	/* enable receive descriptor ring */
3092 	rxdctl |= IXGBE_RXDCTL_ENABLE;
3093 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3094 
3095 	ixgbe_rx_desc_queue_enable(adapter, ring);
3096 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
3097 }
3098 
3099 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
3100 {
3101 	struct ixgbe_hw *hw = &adapter->hw;
3102 	int p;
3103 
3104 	/* PSRTYPE must be initialized in non 82598 adapters */
3105 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
3106 		      IXGBE_PSRTYPE_UDPHDR |
3107 		      IXGBE_PSRTYPE_IPV4HDR |
3108 		      IXGBE_PSRTYPE_L2HDR |
3109 		      IXGBE_PSRTYPE_IPV6HDR;
3110 
3111 	if (hw->mac.type == ixgbe_mac_82598EB)
3112 		return;
3113 
3114 	if (adapter->flags & IXGBE_FLAG_RSS_ENABLED)
3115 		psrtype |= (adapter->num_rx_queues_per_pool << 29);
3116 
3117 	for (p = 0; p < adapter->num_rx_pools; p++)
3118 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(adapter->num_vfs + p),
3119 				psrtype);
3120 }
3121 
3122 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
3123 {
3124 	struct ixgbe_hw *hw = &adapter->hw;
3125 	u32 gcr_ext;
3126 	u32 vt_reg_bits;
3127 	u32 reg_offset, vf_shift;
3128 	u32 vmdctl;
3129 	int i;
3130 
3131 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3132 		return;
3133 
3134 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3135 	vt_reg_bits = IXGBE_VMD_CTL_VMDQ_EN | IXGBE_VT_CTL_REPLEN;
3136 	vt_reg_bits |= (adapter->num_vfs << IXGBE_VT_CTL_POOL_SHIFT);
3137 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl | vt_reg_bits);
3138 
3139 	vf_shift = adapter->num_vfs % 32;
3140 	reg_offset = (adapter->num_vfs >= 32) ? 1 : 0;
3141 
3142 	/* Enable only the PF's pool for Tx/Rx */
3143 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), (1 << vf_shift));
3144 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), 0);
3145 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), (1 << vf_shift));
3146 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), 0);
3147 	IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3148 
3149 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3150 	hw->mac.ops.set_vmdq(hw, 0, adapter->num_vfs);
3151 
3152 	/*
3153 	 * Set up VF register offsets for selected VT Mode,
3154 	 * i.e. 32 or 64 VFs for SR-IOV
3155 	 */
3156 	gcr_ext = IXGBE_READ_REG(hw, IXGBE_GCR_EXT);
3157 	gcr_ext |= IXGBE_GCR_EXT_MSIX_EN;
3158 	gcr_ext |= IXGBE_GCR_EXT_VT_MODE_64;
3159 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
3160 
3161 	/* enable Tx loopback for VF/PF communication */
3162 	IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3163 	/* Enable MAC Anti-Spoofing */
3164 	hw->mac.ops.set_mac_anti_spoofing(hw,
3165 					   (adapter->num_vfs != 0),
3166 					  adapter->num_vfs);
3167 	/* For VFs that have spoof checking turned off */
3168 	for (i = 0; i < adapter->num_vfs; i++) {
3169 		if (!adapter->vfinfo[i].spoofchk_enabled)
3170 			ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i, false);
3171 	}
3172 }
3173 
3174 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
3175 {
3176 	struct ixgbe_hw *hw = &adapter->hw;
3177 	struct net_device *netdev = adapter->netdev;
3178 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3179 	struct ixgbe_ring *rx_ring;
3180 	int i;
3181 	u32 mhadd, hlreg0;
3182 
3183 #ifdef IXGBE_FCOE
3184 	/* adjust max frame to be able to do baby jumbo for FCoE */
3185 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
3186 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
3187 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3188 
3189 #endif /* IXGBE_FCOE */
3190 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3191 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
3192 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3193 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
3194 
3195 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3196 	}
3197 
3198 	/* MHADD will allow an extra 4 bytes past for vlan tagged frames */
3199 	max_frame += VLAN_HLEN;
3200 
3201 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3202 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3203 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3204 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3205 
3206 	/*
3207 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3208 	 * the Base and Length of the Rx Descriptor Ring
3209 	 */
3210 	for (i = 0; i < adapter->num_rx_queues; i++) {
3211 		rx_ring = adapter->rx_ring[i];
3212 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
3213 			set_ring_rsc_enabled(rx_ring);
3214 		else
3215 			clear_ring_rsc_enabled(rx_ring);
3216 	}
3217 }
3218 
3219 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
3220 {
3221 	struct ixgbe_hw *hw = &adapter->hw;
3222 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
3223 
3224 	switch (hw->mac.type) {
3225 	case ixgbe_mac_82598EB:
3226 		/*
3227 		 * For VMDq support of different descriptor types or
3228 		 * buffer sizes through the use of multiple SRRCTL
3229 		 * registers, RDRXCTL.MVMEN must be set to 1
3230 		 *
3231 		 * also, the manual doesn't mention it clearly but DCA hints
3232 		 * will only use queue 0's tags unless this bit is set.  Side
3233 		 * effects of setting this bit are only that SRRCTL must be
3234 		 * fully programmed [0..15]
3235 		 */
3236 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
3237 		break;
3238 	case ixgbe_mac_82599EB:
3239 	case ixgbe_mac_X540:
3240 		/* Disable RSC for ACK packets */
3241 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
3242 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
3243 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
3244 		/* hardware requires some bits to be set by default */
3245 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
3246 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
3247 		break;
3248 	default:
3249 		/* We should do nothing since we don't know this hardware */
3250 		return;
3251 	}
3252 
3253 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
3254 }
3255 
3256 /**
3257  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3258  * @adapter: board private structure
3259  *
3260  * Configure the Rx unit of the MAC after a reset.
3261  **/
3262 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
3263 {
3264 	struct ixgbe_hw *hw = &adapter->hw;
3265 	int i;
3266 	u32 rxctrl;
3267 
3268 	/* disable receives while setting up the descriptors */
3269 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3270 	IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxctrl & ~IXGBE_RXCTRL_RXEN);
3271 
3272 	ixgbe_setup_psrtype(adapter);
3273 	ixgbe_setup_rdrxctl(adapter);
3274 
3275 	/* Program registers for the distribution of queues */
3276 	ixgbe_setup_mrqc(adapter);
3277 
3278 	/* set_rx_buffer_len must be called before ring initialization */
3279 	ixgbe_set_rx_buffer_len(adapter);
3280 
3281 	/*
3282 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3283 	 * the Base and Length of the Rx Descriptor Ring
3284 	 */
3285 	for (i = 0; i < adapter->num_rx_queues; i++)
3286 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
3287 
3288 	/* disable drop enable for 82598 parts */
3289 	if (hw->mac.type == ixgbe_mac_82598EB)
3290 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3291 
3292 	/* enable all receives */
3293 	rxctrl |= IXGBE_RXCTRL_RXEN;
3294 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
3295 }
3296 
3297 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev, u16 vid)
3298 {
3299 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3300 	struct ixgbe_hw *hw = &adapter->hw;
3301 	int pool_ndx = adapter->num_vfs;
3302 
3303 	/* add VID to filter table */
3304 	hw->mac.ops.set_vfta(&adapter->hw, vid, pool_ndx, true);
3305 	set_bit(vid, adapter->active_vlans);
3306 
3307 	return 0;
3308 }
3309 
3310 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev, u16 vid)
3311 {
3312 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3313 	struct ixgbe_hw *hw = &adapter->hw;
3314 	int pool_ndx = adapter->num_vfs;
3315 
3316 	/* remove VID from filter table */
3317 	hw->mac.ops.set_vfta(&adapter->hw, vid, pool_ndx, false);
3318 	clear_bit(vid, adapter->active_vlans);
3319 
3320 	return 0;
3321 }
3322 
3323 /**
3324  * ixgbe_vlan_filter_disable - helper to disable hw vlan filtering
3325  * @adapter: driver data
3326  */
3327 static void ixgbe_vlan_filter_disable(struct ixgbe_adapter *adapter)
3328 {
3329 	struct ixgbe_hw *hw = &adapter->hw;
3330 	u32 vlnctrl;
3331 
3332 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3333 	vlnctrl &= ~(IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
3334 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3335 }
3336 
3337 /**
3338  * ixgbe_vlan_filter_enable - helper to enable hw vlan filtering
3339  * @adapter: driver data
3340  */
3341 static void ixgbe_vlan_filter_enable(struct ixgbe_adapter *adapter)
3342 {
3343 	struct ixgbe_hw *hw = &adapter->hw;
3344 	u32 vlnctrl;
3345 
3346 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3347 	vlnctrl |= IXGBE_VLNCTRL_VFE;
3348 	vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
3349 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3350 }
3351 
3352 /**
3353  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
3354  * @adapter: driver data
3355  */
3356 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
3357 {
3358 	struct ixgbe_hw *hw = &adapter->hw;
3359 	u32 vlnctrl;
3360 	int i, j;
3361 
3362 	switch (hw->mac.type) {
3363 	case ixgbe_mac_82598EB:
3364 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3365 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
3366 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3367 		break;
3368 	case ixgbe_mac_82599EB:
3369 	case ixgbe_mac_X540:
3370 		for (i = 0; i < adapter->num_rx_queues; i++) {
3371 			j = adapter->rx_ring[i]->reg_idx;
3372 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3373 			vlnctrl &= ~IXGBE_RXDCTL_VME;
3374 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3375 		}
3376 		break;
3377 	default:
3378 		break;
3379 	}
3380 }
3381 
3382 /**
3383  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
3384  * @adapter: driver data
3385  */
3386 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
3387 {
3388 	struct ixgbe_hw *hw = &adapter->hw;
3389 	u32 vlnctrl;
3390 	int i, j;
3391 
3392 	switch (hw->mac.type) {
3393 	case ixgbe_mac_82598EB:
3394 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3395 		vlnctrl |= IXGBE_VLNCTRL_VME;
3396 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3397 		break;
3398 	case ixgbe_mac_82599EB:
3399 	case ixgbe_mac_X540:
3400 		for (i = 0; i < adapter->num_rx_queues; i++) {
3401 			j = adapter->rx_ring[i]->reg_idx;
3402 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3403 			vlnctrl |= IXGBE_RXDCTL_VME;
3404 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3405 		}
3406 		break;
3407 	default:
3408 		break;
3409 	}
3410 }
3411 
3412 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
3413 {
3414 	u16 vid;
3415 
3416 	ixgbe_vlan_rx_add_vid(adapter->netdev, 0);
3417 
3418 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
3419 		ixgbe_vlan_rx_add_vid(adapter->netdev, vid);
3420 }
3421 
3422 /**
3423  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
3424  * @netdev: network interface device structure
3425  *
3426  * Writes unicast address list to the RAR table.
3427  * Returns: -ENOMEM on failure/insufficient address space
3428  *                0 on no addresses written
3429  *                X on writing X addresses to the RAR table
3430  **/
3431 static int ixgbe_write_uc_addr_list(struct net_device *netdev)
3432 {
3433 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3434 	struct ixgbe_hw *hw = &adapter->hw;
3435 	unsigned int vfn = adapter->num_vfs;
3436 	unsigned int rar_entries = IXGBE_MAX_PF_MACVLANS;
3437 	int count = 0;
3438 
3439 	/* return ENOMEM indicating insufficient memory for addresses */
3440 	if (netdev_uc_count(netdev) > rar_entries)
3441 		return -ENOMEM;
3442 
3443 	if (!netdev_uc_empty(netdev) && rar_entries) {
3444 		struct netdev_hw_addr *ha;
3445 		/* return error if we do not support writing to RAR table */
3446 		if (!hw->mac.ops.set_rar)
3447 			return -ENOMEM;
3448 
3449 		netdev_for_each_uc_addr(ha, netdev) {
3450 			if (!rar_entries)
3451 				break;
3452 			hw->mac.ops.set_rar(hw, rar_entries--, ha->addr,
3453 					    vfn, IXGBE_RAH_AV);
3454 			count++;
3455 		}
3456 	}
3457 	/* write the addresses in reverse order to avoid write combining */
3458 	for (; rar_entries > 0 ; rar_entries--)
3459 		hw->mac.ops.clear_rar(hw, rar_entries);
3460 
3461 	return count;
3462 }
3463 
3464 /**
3465  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
3466  * @netdev: network interface device structure
3467  *
3468  * The set_rx_method entry point is called whenever the unicast/multicast
3469  * address list or the network interface flags are updated.  This routine is
3470  * responsible for configuring the hardware for proper unicast, multicast and
3471  * promiscuous mode.
3472  **/
3473 void ixgbe_set_rx_mode(struct net_device *netdev)
3474 {
3475 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3476 	struct ixgbe_hw *hw = &adapter->hw;
3477 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
3478 	int count;
3479 
3480 	/* Check for Promiscuous and All Multicast modes */
3481 
3482 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3483 
3484 	/* set all bits that we expect to always be set */
3485 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
3486 	fctrl |= IXGBE_FCTRL_BAM;
3487 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
3488 	fctrl |= IXGBE_FCTRL_PMCF;
3489 
3490 	/* clear the bits we are changing the status of */
3491 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3492 
3493 	if (netdev->flags & IFF_PROMISC) {
3494 		hw->addr_ctrl.user_set_promisc = true;
3495 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3496 		vmolr |= (IXGBE_VMOLR_ROPE | IXGBE_VMOLR_MPE);
3497 		/* don't hardware filter vlans in promisc mode */
3498 		ixgbe_vlan_filter_disable(adapter);
3499 	} else {
3500 		if (netdev->flags & IFF_ALLMULTI) {
3501 			fctrl |= IXGBE_FCTRL_MPE;
3502 			vmolr |= IXGBE_VMOLR_MPE;
3503 		} else {
3504 			/*
3505 			 * Write addresses to the MTA, if the attempt fails
3506 			 * then we should just turn on promiscuous mode so
3507 			 * that we can at least receive multicast traffic
3508 			 */
3509 			hw->mac.ops.update_mc_addr_list(hw, netdev);
3510 			vmolr |= IXGBE_VMOLR_ROMPE;
3511 		}
3512 		ixgbe_vlan_filter_enable(adapter);
3513 		hw->addr_ctrl.user_set_promisc = false;
3514 	}
3515 
3516 	/*
3517 	 * Write addresses to available RAR registers, if there is not
3518 	 * sufficient space to store all the addresses then enable
3519 	 * unicast promiscuous mode
3520 	 */
3521 	count = ixgbe_write_uc_addr_list(netdev);
3522 	if (count < 0) {
3523 		fctrl |= IXGBE_FCTRL_UPE;
3524 		vmolr |= IXGBE_VMOLR_ROPE;
3525 	}
3526 
3527 	if (adapter->num_vfs) {
3528 		ixgbe_restore_vf_multicasts(adapter);
3529 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(adapter->num_vfs)) &
3530 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
3531 			   IXGBE_VMOLR_ROPE);
3532 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(adapter->num_vfs), vmolr);
3533 	}
3534 
3535 	/* This is useful for sniffing bad packets. */
3536 	if (adapter->netdev->features & NETIF_F_RXALL) {
3537 		/* UPE and MPE will be handled by normal PROMISC logic
3538 		 * in e1000e_set_rx_mode */
3539 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
3540 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
3541 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
3542 
3543 		fctrl &= ~(IXGBE_FCTRL_DPF);
3544 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
3545 	}
3546 
3547 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
3548 
3549 	if (netdev->features & NETIF_F_HW_VLAN_RX)
3550 		ixgbe_vlan_strip_enable(adapter);
3551 	else
3552 		ixgbe_vlan_strip_disable(adapter);
3553 }
3554 
3555 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
3556 {
3557 	int q_idx;
3558 	struct ixgbe_q_vector *q_vector;
3559 	int q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
3560 
3561 	/* legacy and MSI only use one vector */
3562 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED))
3563 		q_vectors = 1;
3564 
3565 	for (q_idx = 0; q_idx < q_vectors; q_idx++) {
3566 		q_vector = adapter->q_vector[q_idx];
3567 		napi_enable(&q_vector->napi);
3568 	}
3569 }
3570 
3571 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
3572 {
3573 	int q_idx;
3574 	struct ixgbe_q_vector *q_vector;
3575 	int q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
3576 
3577 	/* legacy and MSI only use one vector */
3578 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED))
3579 		q_vectors = 1;
3580 
3581 	for (q_idx = 0; q_idx < q_vectors; q_idx++) {
3582 		q_vector = adapter->q_vector[q_idx];
3583 		napi_disable(&q_vector->napi);
3584 	}
3585 }
3586 
3587 #ifdef CONFIG_IXGBE_DCB
3588 /*
3589  * ixgbe_configure_dcb - Configure DCB hardware
3590  * @adapter: ixgbe adapter struct
3591  *
3592  * This is called by the driver on open to configure the DCB hardware.
3593  * This is also called by the gennetlink interface when reconfiguring
3594  * the DCB state.
3595  */
3596 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
3597 {
3598 	struct ixgbe_hw *hw = &adapter->hw;
3599 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3600 
3601 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
3602 		if (hw->mac.type == ixgbe_mac_82598EB)
3603 			netif_set_gso_max_size(adapter->netdev, 65536);
3604 		return;
3605 	}
3606 
3607 	if (hw->mac.type == ixgbe_mac_82598EB)
3608 		netif_set_gso_max_size(adapter->netdev, 32768);
3609 
3610 
3611 	/* Enable VLAN tag insert/strip */
3612 	adapter->netdev->features |= NETIF_F_HW_VLAN_RX;
3613 
3614 	hw->mac.ops.set_vfta(&adapter->hw, 0, 0, true);
3615 
3616 #ifdef IXGBE_FCOE
3617 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
3618 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
3619 #endif
3620 
3621 	/* reconfigure the hardware */
3622 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
3623 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
3624 						DCB_TX_CONFIG);
3625 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
3626 						DCB_RX_CONFIG);
3627 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
3628 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
3629 		ixgbe_dcb_hw_ets(&adapter->hw,
3630 				 adapter->ixgbe_ieee_ets,
3631 				 max_frame);
3632 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
3633 					adapter->ixgbe_ieee_pfc->pfc_en,
3634 					adapter->ixgbe_ieee_ets->prio_tc);
3635 	}
3636 
3637 	/* Enable RSS Hash per TC */
3638 	if (hw->mac.type != ixgbe_mac_82598EB) {
3639 		int i;
3640 		u32 reg = 0;
3641 
3642 		for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
3643 			u8 msb = 0;
3644 			u8 cnt = adapter->netdev->tc_to_txq[i].count;
3645 
3646 			while (cnt >>= 1)
3647 				msb++;
3648 
3649 			reg |= msb << IXGBE_RQTC_SHIFT_TC(i);
3650 		}
3651 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, reg);
3652 	}
3653 }
3654 #endif
3655 
3656 /* Additional bittime to account for IXGBE framing */
3657 #define IXGBE_ETH_FRAMING 20
3658 
3659 /*
3660  * ixgbe_hpbthresh - calculate high water mark for flow control
3661  *
3662  * @adapter: board private structure to calculate for
3663  * @pb - packet buffer to calculate
3664  */
3665 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
3666 {
3667 	struct ixgbe_hw *hw = &adapter->hw;
3668 	struct net_device *dev = adapter->netdev;
3669 	int link, tc, kb, marker;
3670 	u32 dv_id, rx_pba;
3671 
3672 	/* Calculate max LAN frame size */
3673 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
3674 
3675 #ifdef IXGBE_FCOE
3676 	/* FCoE traffic class uses FCOE jumbo frames */
3677 	if (dev->features & NETIF_F_FCOE_MTU) {
3678 		int fcoe_pb = 0;
3679 
3680 #ifdef CONFIG_IXGBE_DCB
3681 		fcoe_pb = netdev_get_prio_tc_map(dev, adapter->fcoe.up);
3682 
3683 #endif
3684 		if (fcoe_pb == pb && tc < IXGBE_FCOE_JUMBO_FRAME_SIZE)
3685 			tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3686 	}
3687 #endif
3688 
3689 	/* Calculate delay value for device */
3690 	switch (hw->mac.type) {
3691 	case ixgbe_mac_X540:
3692 		dv_id = IXGBE_DV_X540(link, tc);
3693 		break;
3694 	default:
3695 		dv_id = IXGBE_DV(link, tc);
3696 		break;
3697 	}
3698 
3699 	/* Loopback switch introduces additional latency */
3700 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3701 		dv_id += IXGBE_B2BT(tc);
3702 
3703 	/* Delay value is calculated in bit times convert to KB */
3704 	kb = IXGBE_BT2KB(dv_id);
3705 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
3706 
3707 	marker = rx_pba - kb;
3708 
3709 	/* It is possible that the packet buffer is not large enough
3710 	 * to provide required headroom. In this case throw an error
3711 	 * to user and a do the best we can.
3712 	 */
3713 	if (marker < 0) {
3714 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
3715 			    "headroom to support flow control."
3716 			    "Decrease MTU or number of traffic classes\n", pb);
3717 		marker = tc + 1;
3718 	}
3719 
3720 	return marker;
3721 }
3722 
3723 /*
3724  * ixgbe_lpbthresh - calculate low water mark for for flow control
3725  *
3726  * @adapter: board private structure to calculate for
3727  * @pb - packet buffer to calculate
3728  */
3729 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter)
3730 {
3731 	struct ixgbe_hw *hw = &adapter->hw;
3732 	struct net_device *dev = adapter->netdev;
3733 	int tc;
3734 	u32 dv_id;
3735 
3736 	/* Calculate max LAN frame size */
3737 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
3738 
3739 	/* Calculate delay value for device */
3740 	switch (hw->mac.type) {
3741 	case ixgbe_mac_X540:
3742 		dv_id = IXGBE_LOW_DV_X540(tc);
3743 		break;
3744 	default:
3745 		dv_id = IXGBE_LOW_DV(tc);
3746 		break;
3747 	}
3748 
3749 	/* Delay value is calculated in bit times convert to KB */
3750 	return IXGBE_BT2KB(dv_id);
3751 }
3752 
3753 /*
3754  * ixgbe_pbthresh_setup - calculate and setup high low water marks
3755  */
3756 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
3757 {
3758 	struct ixgbe_hw *hw = &adapter->hw;
3759 	int num_tc = netdev_get_num_tc(adapter->netdev);
3760 	int i;
3761 
3762 	if (!num_tc)
3763 		num_tc = 1;
3764 
3765 	hw->fc.low_water = ixgbe_lpbthresh(adapter);
3766 
3767 	for (i = 0; i < num_tc; i++) {
3768 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
3769 
3770 		/* Low water marks must not be larger than high water marks */
3771 		if (hw->fc.low_water > hw->fc.high_water[i])
3772 			hw->fc.low_water = 0;
3773 	}
3774 }
3775 
3776 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
3777 {
3778 	struct ixgbe_hw *hw = &adapter->hw;
3779 	int hdrm;
3780 	u8 tc = netdev_get_num_tc(adapter->netdev);
3781 
3782 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
3783 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
3784 		hdrm = 32 << adapter->fdir_pballoc;
3785 	else
3786 		hdrm = 0;
3787 
3788 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
3789 	ixgbe_pbthresh_setup(adapter);
3790 }
3791 
3792 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
3793 {
3794 	struct ixgbe_hw *hw = &adapter->hw;
3795 	struct hlist_node *node, *node2;
3796 	struct ixgbe_fdir_filter *filter;
3797 
3798 	spin_lock(&adapter->fdir_perfect_lock);
3799 
3800 	if (!hlist_empty(&adapter->fdir_filter_list))
3801 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
3802 
3803 	hlist_for_each_entry_safe(filter, node, node2,
3804 				  &adapter->fdir_filter_list, fdir_node) {
3805 		ixgbe_fdir_write_perfect_filter_82599(hw,
3806 				&filter->filter,
3807 				filter->sw_idx,
3808 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
3809 				IXGBE_FDIR_DROP_QUEUE :
3810 				adapter->rx_ring[filter->action]->reg_idx);
3811 	}
3812 
3813 	spin_unlock(&adapter->fdir_perfect_lock);
3814 }
3815 
3816 static void ixgbe_configure(struct ixgbe_adapter *adapter)
3817 {
3818 	struct ixgbe_hw *hw = &adapter->hw;
3819 
3820 	ixgbe_configure_pb(adapter);
3821 #ifdef CONFIG_IXGBE_DCB
3822 	ixgbe_configure_dcb(adapter);
3823 #endif
3824 
3825 	ixgbe_set_rx_mode(adapter->netdev);
3826 	ixgbe_restore_vlan(adapter);
3827 
3828 #ifdef IXGBE_FCOE
3829 	if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
3830 		ixgbe_configure_fcoe(adapter);
3831 
3832 #endif /* IXGBE_FCOE */
3833 
3834 	switch (hw->mac.type) {
3835 	case ixgbe_mac_82599EB:
3836 	case ixgbe_mac_X540:
3837 		hw->mac.ops.disable_rx_buff(hw);
3838 		break;
3839 	default:
3840 		break;
3841 	}
3842 
3843 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3844 		ixgbe_init_fdir_signature_82599(&adapter->hw,
3845 						adapter->fdir_pballoc);
3846 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
3847 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
3848 					      adapter->fdir_pballoc);
3849 		ixgbe_fdir_filter_restore(adapter);
3850 	}
3851 
3852 	switch (hw->mac.type) {
3853 	case ixgbe_mac_82599EB:
3854 	case ixgbe_mac_X540:
3855 		hw->mac.ops.enable_rx_buff(hw);
3856 		break;
3857 	default:
3858 		break;
3859 	}
3860 
3861 	ixgbe_configure_virtualization(adapter);
3862 
3863 	ixgbe_configure_tx(adapter);
3864 	ixgbe_configure_rx(adapter);
3865 }
3866 
3867 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
3868 {
3869 	switch (hw->phy.type) {
3870 	case ixgbe_phy_sfp_avago:
3871 	case ixgbe_phy_sfp_ftl:
3872 	case ixgbe_phy_sfp_intel:
3873 	case ixgbe_phy_sfp_unknown:
3874 	case ixgbe_phy_sfp_passive_tyco:
3875 	case ixgbe_phy_sfp_passive_unknown:
3876 	case ixgbe_phy_sfp_active_unknown:
3877 	case ixgbe_phy_sfp_ftl_active:
3878 		return true;
3879 	case ixgbe_phy_nl:
3880 		if (hw->mac.type == ixgbe_mac_82598EB)
3881 			return true;
3882 	default:
3883 		return false;
3884 	}
3885 }
3886 
3887 /**
3888  * ixgbe_sfp_link_config - set up SFP+ link
3889  * @adapter: pointer to private adapter struct
3890  **/
3891 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
3892 {
3893 	/*
3894 	 * We are assuming the worst case scenario here, and that
3895 	 * is that an SFP was inserted/removed after the reset
3896 	 * but before SFP detection was enabled.  As such the best
3897 	 * solution is to just start searching as soon as we start
3898 	 */
3899 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3900 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
3901 
3902 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
3903 }
3904 
3905 /**
3906  * ixgbe_non_sfp_link_config - set up non-SFP+ link
3907  * @hw: pointer to private hardware struct
3908  *
3909  * Returns 0 on success, negative on failure
3910  **/
3911 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
3912 {
3913 	u32 autoneg;
3914 	bool negotiation, link_up = false;
3915 	u32 ret = IXGBE_ERR_LINK_SETUP;
3916 
3917 	if (hw->mac.ops.check_link)
3918 		ret = hw->mac.ops.check_link(hw, &autoneg, &link_up, false);
3919 
3920 	if (ret)
3921 		goto link_cfg_out;
3922 
3923 	autoneg = hw->phy.autoneg_advertised;
3924 	if ((!autoneg) && (hw->mac.ops.get_link_capabilities))
3925 		ret = hw->mac.ops.get_link_capabilities(hw, &autoneg,
3926 							&negotiation);
3927 	if (ret)
3928 		goto link_cfg_out;
3929 
3930 	if (hw->mac.ops.setup_link)
3931 		ret = hw->mac.ops.setup_link(hw, autoneg, negotiation, link_up);
3932 link_cfg_out:
3933 	return ret;
3934 }
3935 
3936 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
3937 {
3938 	struct ixgbe_hw *hw = &adapter->hw;
3939 	u32 gpie = 0;
3940 
3941 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3942 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
3943 		       IXGBE_GPIE_OCD;
3944 		gpie |= IXGBE_GPIE_EIAME;
3945 		/*
3946 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
3947 		 * this saves a register write for every interrupt
3948 		 */
3949 		switch (hw->mac.type) {
3950 		case ixgbe_mac_82598EB:
3951 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
3952 			break;
3953 		case ixgbe_mac_82599EB:
3954 		case ixgbe_mac_X540:
3955 		default:
3956 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
3957 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
3958 			break;
3959 		}
3960 	} else {
3961 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
3962 		 * specifically only auto mask tx and rx interrupts */
3963 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
3964 	}
3965 
3966 	/* XXX: to interrupt immediately for EICS writes, enable this */
3967 	/* gpie |= IXGBE_GPIE_EIMEN; */
3968 
3969 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3970 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
3971 		gpie |= IXGBE_GPIE_VTMODE_64;
3972 	}
3973 
3974 	/* Enable Thermal over heat sensor interrupt */
3975 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
3976 		switch (adapter->hw.mac.type) {
3977 		case ixgbe_mac_82599EB:
3978 			gpie |= IXGBE_SDP0_GPIEN;
3979 			break;
3980 		case ixgbe_mac_X540:
3981 			gpie |= IXGBE_EIMS_TS;
3982 			break;
3983 		default:
3984 			break;
3985 		}
3986 	}
3987 
3988 	/* Enable fan failure interrupt */
3989 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
3990 		gpie |= IXGBE_SDP1_GPIEN;
3991 
3992 	if (hw->mac.type == ixgbe_mac_82599EB) {
3993 		gpie |= IXGBE_SDP1_GPIEN;
3994 		gpie |= IXGBE_SDP2_GPIEN;
3995 	}
3996 
3997 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
3998 }
3999 
4000 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
4001 {
4002 	struct ixgbe_hw *hw = &adapter->hw;
4003 	int err;
4004 	u32 ctrl_ext;
4005 
4006 	ixgbe_get_hw_control(adapter);
4007 	ixgbe_setup_gpie(adapter);
4008 
4009 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
4010 		ixgbe_configure_msix(adapter);
4011 	else
4012 		ixgbe_configure_msi_and_legacy(adapter);
4013 
4014 	/* enable the optics for both mult-speed fiber and 82599 SFP+ fiber */
4015 	if (hw->mac.ops.enable_tx_laser &&
4016 	    ((hw->phy.multispeed_fiber) ||
4017 	     ((hw->mac.ops.get_media_type(hw) == ixgbe_media_type_fiber) &&
4018 	      (hw->mac.type == ixgbe_mac_82599EB))))
4019 		hw->mac.ops.enable_tx_laser(hw);
4020 
4021 	clear_bit(__IXGBE_DOWN, &adapter->state);
4022 	ixgbe_napi_enable_all(adapter);
4023 
4024 	if (ixgbe_is_sfp(hw)) {
4025 		ixgbe_sfp_link_config(adapter);
4026 	} else {
4027 		err = ixgbe_non_sfp_link_config(hw);
4028 		if (err)
4029 			e_err(probe, "link_config FAILED %d\n", err);
4030 	}
4031 
4032 	/* clear any pending interrupts, may auto mask */
4033 	IXGBE_READ_REG(hw, IXGBE_EICR);
4034 	ixgbe_irq_enable(adapter, true, true);
4035 
4036 	/*
4037 	 * If this adapter has a fan, check to see if we had a failure
4038 	 * before we enabled the interrupt.
4039 	 */
4040 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
4041 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
4042 		if (esdp & IXGBE_ESDP_SDP1)
4043 			e_crit(drv, "Fan has stopped, replace the adapter\n");
4044 	}
4045 
4046 	/* enable transmits */
4047 	netif_tx_start_all_queues(adapter->netdev);
4048 
4049 	/* bring the link up in the watchdog, this could race with our first
4050 	 * link up interrupt but shouldn't be a problem */
4051 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4052 	adapter->link_check_timeout = jiffies;
4053 	mod_timer(&adapter->service_timer, jiffies);
4054 
4055 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
4056 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
4057 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
4058 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
4059 }
4060 
4061 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
4062 {
4063 	WARN_ON(in_interrupt());
4064 	/* put off any impending NetWatchDogTimeout */
4065 	adapter->netdev->trans_start = jiffies;
4066 
4067 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
4068 		usleep_range(1000, 2000);
4069 	ixgbe_down(adapter);
4070 	/*
4071 	 * If SR-IOV enabled then wait a bit before bringing the adapter
4072 	 * back up to give the VFs time to respond to the reset.  The
4073 	 * two second wait is based upon the watchdog timer cycle in
4074 	 * the VF driver.
4075 	 */
4076 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4077 		msleep(2000);
4078 	ixgbe_up(adapter);
4079 	clear_bit(__IXGBE_RESETTING, &adapter->state);
4080 }
4081 
4082 void ixgbe_up(struct ixgbe_adapter *adapter)
4083 {
4084 	/* hardware has been reset, we need to reload some things */
4085 	ixgbe_configure(adapter);
4086 
4087 	ixgbe_up_complete(adapter);
4088 }
4089 
4090 void ixgbe_reset(struct ixgbe_adapter *adapter)
4091 {
4092 	struct ixgbe_hw *hw = &adapter->hw;
4093 	int err;
4094 
4095 	/* lock SFP init bit to prevent race conditions with the watchdog */
4096 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
4097 		usleep_range(1000, 2000);
4098 
4099 	/* clear all SFP and link config related flags while holding SFP_INIT */
4100 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
4101 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
4102 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4103 
4104 	err = hw->mac.ops.init_hw(hw);
4105 	switch (err) {
4106 	case 0:
4107 	case IXGBE_ERR_SFP_NOT_PRESENT:
4108 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
4109 		break;
4110 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
4111 		e_dev_err("master disable timed out\n");
4112 		break;
4113 	case IXGBE_ERR_EEPROM_VERSION:
4114 		/* We are running on a pre-production device, log a warning */
4115 		e_dev_warn("This device is a pre-production adapter/LOM. "
4116 			   "Please be aware there may be issues associated with "
4117 			   "your hardware.  If you are experiencing problems "
4118 			   "please contact your Intel or hardware "
4119 			   "representative who provided you with this "
4120 			   "hardware.\n");
4121 		break;
4122 	default:
4123 		e_dev_err("Hardware Error: %d\n", err);
4124 	}
4125 
4126 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
4127 
4128 	/* reprogram the RAR[0] in case user changed it. */
4129 	hw->mac.ops.set_rar(hw, 0, hw->mac.addr, adapter->num_vfs,
4130 			    IXGBE_RAH_AV);
4131 }
4132 
4133 /**
4134  * ixgbe_init_rx_page_offset - initialize page offset values for Rx buffers
4135  * @rx_ring: ring to setup
4136  *
4137  * On many IA platforms the L1 cache has a critical stride of 4K, this
4138  * results in each receive buffer starting in the same cache set.  To help
4139  * reduce the pressure on this cache set we can interleave the offsets so
4140  * that only every other buffer will be in the same cache set.
4141  **/
4142 static void ixgbe_init_rx_page_offset(struct ixgbe_ring *rx_ring)
4143 {
4144 	struct ixgbe_rx_buffer *rx_buffer = rx_ring->rx_buffer_info;
4145 	u16 i;
4146 
4147 	for (i = 0; i < rx_ring->count; i += 2) {
4148 		rx_buffer[0].page_offset = 0;
4149 		rx_buffer[1].page_offset = ixgbe_rx_bufsz(rx_ring);
4150 		rx_buffer = &rx_buffer[2];
4151 	}
4152 }
4153 
4154 /**
4155  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4156  * @rx_ring: ring to free buffers from
4157  **/
4158 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
4159 {
4160 	struct device *dev = rx_ring->dev;
4161 	unsigned long size;
4162 	u16 i;
4163 
4164 	/* ring already cleared, nothing to do */
4165 	if (!rx_ring->rx_buffer_info)
4166 		return;
4167 
4168 	/* Free all the Rx ring sk_buffs */
4169 	for (i = 0; i < rx_ring->count; i++) {
4170 		struct ixgbe_rx_buffer *rx_buffer;
4171 
4172 		rx_buffer = &rx_ring->rx_buffer_info[i];
4173 		if (rx_buffer->skb) {
4174 			struct sk_buff *skb = rx_buffer->skb;
4175 			if (IXGBE_CB(skb)->page_released) {
4176 				dma_unmap_page(dev,
4177 					       IXGBE_CB(skb)->dma,
4178 					       ixgbe_rx_bufsz(rx_ring),
4179 					       DMA_FROM_DEVICE);
4180 				IXGBE_CB(skb)->page_released = false;
4181 			}
4182 			dev_kfree_skb(skb);
4183 		}
4184 		rx_buffer->skb = NULL;
4185 		if (rx_buffer->dma)
4186 			dma_unmap_page(dev, rx_buffer->dma,
4187 				       ixgbe_rx_pg_size(rx_ring),
4188 				       DMA_FROM_DEVICE);
4189 		rx_buffer->dma = 0;
4190 		if (rx_buffer->page)
4191 			__free_pages(rx_buffer->page,
4192 				     ixgbe_rx_pg_order(rx_ring));
4193 		rx_buffer->page = NULL;
4194 	}
4195 
4196 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4197 	memset(rx_ring->rx_buffer_info, 0, size);
4198 
4199 	ixgbe_init_rx_page_offset(rx_ring);
4200 
4201 	/* Zero out the descriptor ring */
4202 	memset(rx_ring->desc, 0, rx_ring->size);
4203 
4204 	rx_ring->next_to_alloc = 0;
4205 	rx_ring->next_to_clean = 0;
4206 	rx_ring->next_to_use = 0;
4207 }
4208 
4209 /**
4210  * ixgbe_clean_tx_ring - Free Tx Buffers
4211  * @tx_ring: ring to be cleaned
4212  **/
4213 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
4214 {
4215 	struct ixgbe_tx_buffer *tx_buffer_info;
4216 	unsigned long size;
4217 	u16 i;
4218 
4219 	/* ring already cleared, nothing to do */
4220 	if (!tx_ring->tx_buffer_info)
4221 		return;
4222 
4223 	/* Free all the Tx ring sk_buffs */
4224 	for (i = 0; i < tx_ring->count; i++) {
4225 		tx_buffer_info = &tx_ring->tx_buffer_info[i];
4226 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer_info);
4227 	}
4228 
4229 	netdev_tx_reset_queue(txring_txq(tx_ring));
4230 
4231 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
4232 	memset(tx_ring->tx_buffer_info, 0, size);
4233 
4234 	/* Zero out the descriptor ring */
4235 	memset(tx_ring->desc, 0, tx_ring->size);
4236 
4237 	tx_ring->next_to_use = 0;
4238 	tx_ring->next_to_clean = 0;
4239 }
4240 
4241 /**
4242  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
4243  * @adapter: board private structure
4244  **/
4245 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
4246 {
4247 	int i;
4248 
4249 	for (i = 0; i < adapter->num_rx_queues; i++)
4250 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
4251 }
4252 
4253 /**
4254  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
4255  * @adapter: board private structure
4256  **/
4257 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
4258 {
4259 	int i;
4260 
4261 	for (i = 0; i < adapter->num_tx_queues; i++)
4262 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
4263 }
4264 
4265 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
4266 {
4267 	struct hlist_node *node, *node2;
4268 	struct ixgbe_fdir_filter *filter;
4269 
4270 	spin_lock(&adapter->fdir_perfect_lock);
4271 
4272 	hlist_for_each_entry_safe(filter, node, node2,
4273 				  &adapter->fdir_filter_list, fdir_node) {
4274 		hlist_del(&filter->fdir_node);
4275 		kfree(filter);
4276 	}
4277 	adapter->fdir_filter_count = 0;
4278 
4279 	spin_unlock(&adapter->fdir_perfect_lock);
4280 }
4281 
4282 void ixgbe_down(struct ixgbe_adapter *adapter)
4283 {
4284 	struct net_device *netdev = adapter->netdev;
4285 	struct ixgbe_hw *hw = &adapter->hw;
4286 	u32 rxctrl;
4287 	int i;
4288 
4289 	/* signal that we are down to the interrupt handler */
4290 	set_bit(__IXGBE_DOWN, &adapter->state);
4291 
4292 	/* disable receives */
4293 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4294 	IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxctrl & ~IXGBE_RXCTRL_RXEN);
4295 
4296 	/* disable all enabled rx queues */
4297 	for (i = 0; i < adapter->num_rx_queues; i++)
4298 		/* this call also flushes the previous write */
4299 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
4300 
4301 	usleep_range(10000, 20000);
4302 
4303 	netif_tx_stop_all_queues(netdev);
4304 
4305 	/* call carrier off first to avoid false dev_watchdog timeouts */
4306 	netif_carrier_off(netdev);
4307 	netif_tx_disable(netdev);
4308 
4309 	ixgbe_irq_disable(adapter);
4310 
4311 	ixgbe_napi_disable_all(adapter);
4312 
4313 	adapter->flags2 &= ~(IXGBE_FLAG2_FDIR_REQUIRES_REINIT |
4314 			     IXGBE_FLAG2_RESET_REQUESTED);
4315 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4316 
4317 	del_timer_sync(&adapter->service_timer);
4318 
4319 	if (adapter->num_vfs) {
4320 		/* Clear EITR Select mapping */
4321 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
4322 
4323 		/* Mark all the VFs as inactive */
4324 		for (i = 0 ; i < adapter->num_vfs; i++)
4325 			adapter->vfinfo[i].clear_to_send = false;
4326 
4327 		/* ping all the active vfs to let them know we are going down */
4328 		ixgbe_ping_all_vfs(adapter);
4329 
4330 		/* Disable all VFTE/VFRE TX/RX */
4331 		ixgbe_disable_tx_rx(adapter);
4332 	}
4333 
4334 	/* disable transmits in the hardware now that interrupts are off */
4335 	for (i = 0; i < adapter->num_tx_queues; i++) {
4336 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
4337 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
4338 	}
4339 
4340 	/* Disable the Tx DMA engine on 82599 and X540 */
4341 	switch (hw->mac.type) {
4342 	case ixgbe_mac_82599EB:
4343 	case ixgbe_mac_X540:
4344 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
4345 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
4346 				 ~IXGBE_DMATXCTL_TE));
4347 		break;
4348 	default:
4349 		break;
4350 	}
4351 
4352 	if (!pci_channel_offline(adapter->pdev))
4353 		ixgbe_reset(adapter);
4354 
4355 	/* power down the optics for multispeed fiber and 82599 SFP+ fiber */
4356 	if (hw->mac.ops.disable_tx_laser &&
4357 	    ((hw->phy.multispeed_fiber) ||
4358 	     ((hw->mac.ops.get_media_type(hw) == ixgbe_media_type_fiber) &&
4359 	      (hw->mac.type == ixgbe_mac_82599EB))))
4360 		hw->mac.ops.disable_tx_laser(hw);
4361 
4362 	ixgbe_clean_all_tx_rings(adapter);
4363 	ixgbe_clean_all_rx_rings(adapter);
4364 
4365 #ifdef CONFIG_IXGBE_DCA
4366 	/* since we reset the hardware DCA settings were cleared */
4367 	ixgbe_setup_dca(adapter);
4368 #endif
4369 }
4370 
4371 /**
4372  * ixgbe_tx_timeout - Respond to a Tx Hang
4373  * @netdev: network interface device structure
4374  **/
4375 static void ixgbe_tx_timeout(struct net_device *netdev)
4376 {
4377 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4378 
4379 	/* Do the reset outside of interrupt context */
4380 	ixgbe_tx_timeout_reset(adapter);
4381 }
4382 
4383 /**
4384  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
4385  * @adapter: board private structure to initialize
4386  *
4387  * ixgbe_sw_init initializes the Adapter private data structure.
4388  * Fields are initialized based on PCI device information and
4389  * OS network device settings (MTU size).
4390  **/
4391 static int __devinit ixgbe_sw_init(struct ixgbe_adapter *adapter)
4392 {
4393 	struct ixgbe_hw *hw = &adapter->hw;
4394 	struct pci_dev *pdev = adapter->pdev;
4395 	unsigned int rss;
4396 #ifdef CONFIG_IXGBE_DCB
4397 	int j;
4398 	struct tc_configuration *tc;
4399 #endif
4400 
4401 	/* PCI config space info */
4402 
4403 	hw->vendor_id = pdev->vendor;
4404 	hw->device_id = pdev->device;
4405 	hw->revision_id = pdev->revision;
4406 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
4407 	hw->subsystem_device_id = pdev->subsystem_device;
4408 
4409 	/* Set capability flags */
4410 	rss = min_t(int, IXGBE_MAX_RSS_INDICES, num_online_cpus());
4411 	adapter->ring_feature[RING_F_RSS].indices = rss;
4412 	adapter->flags |= IXGBE_FLAG_RSS_ENABLED;
4413 	switch (hw->mac.type) {
4414 	case ixgbe_mac_82598EB:
4415 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
4416 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
4417 		adapter->max_msix_q_vectors = MAX_MSIX_Q_VECTORS_82598;
4418 		break;
4419 	case ixgbe_mac_X540:
4420 		adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
4421 	case ixgbe_mac_82599EB:
4422 		adapter->max_msix_q_vectors = MAX_MSIX_Q_VECTORS_82599;
4423 		adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
4424 		adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
4425 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
4426 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
4427 		/* Flow Director hash filters enabled */
4428 		adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
4429 		adapter->atr_sample_rate = 20;
4430 		adapter->ring_feature[RING_F_FDIR].indices =
4431 							 IXGBE_MAX_FDIR_INDICES;
4432 		adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
4433 #ifdef IXGBE_FCOE
4434 		adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
4435 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
4436 		adapter->ring_feature[RING_F_FCOE].indices = 0;
4437 #ifdef CONFIG_IXGBE_DCB
4438 		/* Default traffic class to use for FCoE */
4439 		adapter->fcoe.up = IXGBE_FCOE_DEFTC;
4440 #endif
4441 #endif /* IXGBE_FCOE */
4442 		break;
4443 	default:
4444 		break;
4445 	}
4446 
4447 	/* n-tuple support exists, always init our spinlock */
4448 	spin_lock_init(&adapter->fdir_perfect_lock);
4449 
4450 #ifdef CONFIG_IXGBE_DCB
4451 	switch (hw->mac.type) {
4452 	case ixgbe_mac_X540:
4453 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
4454 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
4455 		break;
4456 	default:
4457 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
4458 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
4459 		break;
4460 	}
4461 
4462 	/* Configure DCB traffic classes */
4463 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
4464 		tc = &adapter->dcb_cfg.tc_config[j];
4465 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
4466 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
4467 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
4468 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
4469 		tc->dcb_pfc = pfc_disabled;
4470 	}
4471 
4472 	/* Initialize default user to priority mapping, UPx->TC0 */
4473 	tc = &adapter->dcb_cfg.tc_config[0];
4474 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
4475 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
4476 
4477 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
4478 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
4479 	adapter->dcb_cfg.pfc_mode_enable = false;
4480 	adapter->dcb_set_bitmap = 0x00;
4481 	adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
4482 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
4483 	       sizeof(adapter->temp_dcb_cfg));
4484 
4485 #endif
4486 
4487 	/* default flow control settings */
4488 	hw->fc.requested_mode = ixgbe_fc_full;
4489 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
4490 	ixgbe_pbthresh_setup(adapter);
4491 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
4492 	hw->fc.send_xon = true;
4493 	hw->fc.disable_fc_autoneg = false;
4494 
4495 	/* enable itr by default in dynamic mode */
4496 	adapter->rx_itr_setting = 1;
4497 	adapter->tx_itr_setting = 1;
4498 
4499 	/* set default ring sizes */
4500 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
4501 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
4502 
4503 	/* set default work limits */
4504 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
4505 
4506 	/* initialize eeprom parameters */
4507 	if (ixgbe_init_eeprom_params_generic(hw)) {
4508 		e_dev_err("EEPROM initialization failed\n");
4509 		return -EIO;
4510 	}
4511 
4512 	set_bit(__IXGBE_DOWN, &adapter->state);
4513 
4514 	return 0;
4515 }
4516 
4517 /**
4518  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
4519  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
4520  *
4521  * Return 0 on success, negative on failure
4522  **/
4523 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
4524 {
4525 	struct device *dev = tx_ring->dev;
4526 	int orig_node = dev_to_node(dev);
4527 	int numa_node = -1;
4528 	int size;
4529 
4530 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
4531 
4532 	if (tx_ring->q_vector)
4533 		numa_node = tx_ring->q_vector->numa_node;
4534 
4535 	tx_ring->tx_buffer_info = vzalloc_node(size, numa_node);
4536 	if (!tx_ring->tx_buffer_info)
4537 		tx_ring->tx_buffer_info = vzalloc(size);
4538 	if (!tx_ring->tx_buffer_info)
4539 		goto err;
4540 
4541 	/* round up to nearest 4K */
4542 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
4543 	tx_ring->size = ALIGN(tx_ring->size, 4096);
4544 
4545 	set_dev_node(dev, numa_node);
4546 	tx_ring->desc = dma_alloc_coherent(dev,
4547 					   tx_ring->size,
4548 					   &tx_ring->dma,
4549 					   GFP_KERNEL);
4550 	set_dev_node(dev, orig_node);
4551 	if (!tx_ring->desc)
4552 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
4553 						   &tx_ring->dma, GFP_KERNEL);
4554 	if (!tx_ring->desc)
4555 		goto err;
4556 
4557 	tx_ring->next_to_use = 0;
4558 	tx_ring->next_to_clean = 0;
4559 	return 0;
4560 
4561 err:
4562 	vfree(tx_ring->tx_buffer_info);
4563 	tx_ring->tx_buffer_info = NULL;
4564 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
4565 	return -ENOMEM;
4566 }
4567 
4568 /**
4569  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
4570  * @adapter: board private structure
4571  *
4572  * If this function returns with an error, then it's possible one or
4573  * more of the rings is populated (while the rest are not).  It is the
4574  * callers duty to clean those orphaned rings.
4575  *
4576  * Return 0 on success, negative on failure
4577  **/
4578 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
4579 {
4580 	int i, err = 0;
4581 
4582 	for (i = 0; i < adapter->num_tx_queues; i++) {
4583 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
4584 		if (!err)
4585 			continue;
4586 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
4587 		break;
4588 	}
4589 
4590 	return err;
4591 }
4592 
4593 /**
4594  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
4595  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
4596  *
4597  * Returns 0 on success, negative on failure
4598  **/
4599 int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
4600 {
4601 	struct device *dev = rx_ring->dev;
4602 	int orig_node = dev_to_node(dev);
4603 	int numa_node = -1;
4604 	int size;
4605 
4606 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4607 
4608 	if (rx_ring->q_vector)
4609 		numa_node = rx_ring->q_vector->numa_node;
4610 
4611 	rx_ring->rx_buffer_info = vzalloc_node(size, numa_node);
4612 	if (!rx_ring->rx_buffer_info)
4613 		rx_ring->rx_buffer_info = vzalloc(size);
4614 	if (!rx_ring->rx_buffer_info)
4615 		goto err;
4616 
4617 	/* Round up to nearest 4K */
4618 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
4619 	rx_ring->size = ALIGN(rx_ring->size, 4096);
4620 
4621 	set_dev_node(dev, numa_node);
4622 	rx_ring->desc = dma_alloc_coherent(dev,
4623 					   rx_ring->size,
4624 					   &rx_ring->dma,
4625 					   GFP_KERNEL);
4626 	set_dev_node(dev, orig_node);
4627 	if (!rx_ring->desc)
4628 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
4629 						   &rx_ring->dma, GFP_KERNEL);
4630 	if (!rx_ring->desc)
4631 		goto err;
4632 
4633 	rx_ring->next_to_clean = 0;
4634 	rx_ring->next_to_use = 0;
4635 
4636 	ixgbe_init_rx_page_offset(rx_ring);
4637 
4638 	return 0;
4639 err:
4640 	vfree(rx_ring->rx_buffer_info);
4641 	rx_ring->rx_buffer_info = NULL;
4642 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
4643 	return -ENOMEM;
4644 }
4645 
4646 /**
4647  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
4648  * @adapter: board private structure
4649  *
4650  * If this function returns with an error, then it's possible one or
4651  * more of the rings is populated (while the rest are not).  It is the
4652  * callers duty to clean those orphaned rings.
4653  *
4654  * Return 0 on success, negative on failure
4655  **/
4656 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
4657 {
4658 	int i, err = 0;
4659 
4660 	for (i = 0; i < adapter->num_rx_queues; i++) {
4661 		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
4662 		if (!err)
4663 			continue;
4664 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
4665 		break;
4666 	}
4667 
4668 	return err;
4669 }
4670 
4671 /**
4672  * ixgbe_free_tx_resources - Free Tx Resources per Queue
4673  * @tx_ring: Tx descriptor ring for a specific queue
4674  *
4675  * Free all transmit software resources
4676  **/
4677 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
4678 {
4679 	ixgbe_clean_tx_ring(tx_ring);
4680 
4681 	vfree(tx_ring->tx_buffer_info);
4682 	tx_ring->tx_buffer_info = NULL;
4683 
4684 	/* if not set, then don't free */
4685 	if (!tx_ring->desc)
4686 		return;
4687 
4688 	dma_free_coherent(tx_ring->dev, tx_ring->size,
4689 			  tx_ring->desc, tx_ring->dma);
4690 
4691 	tx_ring->desc = NULL;
4692 }
4693 
4694 /**
4695  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
4696  * @adapter: board private structure
4697  *
4698  * Free all transmit software resources
4699  **/
4700 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
4701 {
4702 	int i;
4703 
4704 	for (i = 0; i < adapter->num_tx_queues; i++)
4705 		if (adapter->tx_ring[i]->desc)
4706 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
4707 }
4708 
4709 /**
4710  * ixgbe_free_rx_resources - Free Rx Resources
4711  * @rx_ring: ring to clean the resources from
4712  *
4713  * Free all receive software resources
4714  **/
4715 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
4716 {
4717 	ixgbe_clean_rx_ring(rx_ring);
4718 
4719 	vfree(rx_ring->rx_buffer_info);
4720 	rx_ring->rx_buffer_info = NULL;
4721 
4722 	/* if not set, then don't free */
4723 	if (!rx_ring->desc)
4724 		return;
4725 
4726 	dma_free_coherent(rx_ring->dev, rx_ring->size,
4727 			  rx_ring->desc, rx_ring->dma);
4728 
4729 	rx_ring->desc = NULL;
4730 }
4731 
4732 /**
4733  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
4734  * @adapter: board private structure
4735  *
4736  * Free all receive software resources
4737  **/
4738 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
4739 {
4740 	int i;
4741 
4742 	for (i = 0; i < adapter->num_rx_queues; i++)
4743 		if (adapter->rx_ring[i]->desc)
4744 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
4745 }
4746 
4747 /**
4748  * ixgbe_change_mtu - Change the Maximum Transfer Unit
4749  * @netdev: network interface device structure
4750  * @new_mtu: new value for maximum frame size
4751  *
4752  * Returns 0 on success, negative on failure
4753  **/
4754 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
4755 {
4756 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4757 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
4758 
4759 	/* MTU < 68 is an error and causes problems on some kernels */
4760 	if ((new_mtu < 68) || (max_frame > IXGBE_MAX_JUMBO_FRAME_SIZE))
4761 		return -EINVAL;
4762 
4763 	/*
4764 	 * For 82599EB we cannot allow PF to change MTU greater than 1500
4765 	 * in SR-IOV mode as it may cause buffer overruns in guest VFs that
4766 	 * don't allocate and chain buffers correctly.
4767 	 */
4768 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
4769 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
4770 	    (max_frame > MAXIMUM_ETHERNET_VLAN_SIZE))
4771 			return -EINVAL;
4772 
4773 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
4774 
4775 	/* must set new MTU before calling down or up */
4776 	netdev->mtu = new_mtu;
4777 
4778 	if (netif_running(netdev))
4779 		ixgbe_reinit_locked(adapter);
4780 
4781 	return 0;
4782 }
4783 
4784 /**
4785  * ixgbe_open - Called when a network interface is made active
4786  * @netdev: network interface device structure
4787  *
4788  * Returns 0 on success, negative value on failure
4789  *
4790  * The open entry point is called when a network interface is made
4791  * active by the system (IFF_UP).  At this point all resources needed
4792  * for transmit and receive operations are allocated, the interrupt
4793  * handler is registered with the OS, the watchdog timer is started,
4794  * and the stack is notified that the interface is ready.
4795  **/
4796 static int ixgbe_open(struct net_device *netdev)
4797 {
4798 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4799 	int err;
4800 
4801 	/* disallow open during test */
4802 	if (test_bit(__IXGBE_TESTING, &adapter->state))
4803 		return -EBUSY;
4804 
4805 	netif_carrier_off(netdev);
4806 
4807 	/* allocate transmit descriptors */
4808 	err = ixgbe_setup_all_tx_resources(adapter);
4809 	if (err)
4810 		goto err_setup_tx;
4811 
4812 	/* allocate receive descriptors */
4813 	err = ixgbe_setup_all_rx_resources(adapter);
4814 	if (err)
4815 		goto err_setup_rx;
4816 
4817 	ixgbe_configure(adapter);
4818 
4819 	err = ixgbe_request_irq(adapter);
4820 	if (err)
4821 		goto err_req_irq;
4822 
4823 	ixgbe_up_complete(adapter);
4824 
4825 	return 0;
4826 
4827 err_req_irq:
4828 err_setup_rx:
4829 	ixgbe_free_all_rx_resources(adapter);
4830 err_setup_tx:
4831 	ixgbe_free_all_tx_resources(adapter);
4832 	ixgbe_reset(adapter);
4833 
4834 	return err;
4835 }
4836 
4837 /**
4838  * ixgbe_close - Disables a network interface
4839  * @netdev: network interface device structure
4840  *
4841  * Returns 0, this is not allowed to fail
4842  *
4843  * The close entry point is called when an interface is de-activated
4844  * by the OS.  The hardware is still under the drivers control, but
4845  * needs to be disabled.  A global MAC reset is issued to stop the
4846  * hardware, and all transmit and receive resources are freed.
4847  **/
4848 static int ixgbe_close(struct net_device *netdev)
4849 {
4850 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4851 
4852 	ixgbe_down(adapter);
4853 	ixgbe_free_irq(adapter);
4854 
4855 	ixgbe_fdir_filter_exit(adapter);
4856 
4857 	ixgbe_free_all_tx_resources(adapter);
4858 	ixgbe_free_all_rx_resources(adapter);
4859 
4860 	ixgbe_release_hw_control(adapter);
4861 
4862 	return 0;
4863 }
4864 
4865 #ifdef CONFIG_PM
4866 static int ixgbe_resume(struct pci_dev *pdev)
4867 {
4868 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
4869 	struct net_device *netdev = adapter->netdev;
4870 	u32 err;
4871 
4872 	pci_set_power_state(pdev, PCI_D0);
4873 	pci_restore_state(pdev);
4874 	/*
4875 	 * pci_restore_state clears dev->state_saved so call
4876 	 * pci_save_state to restore it.
4877 	 */
4878 	pci_save_state(pdev);
4879 
4880 	err = pci_enable_device_mem(pdev);
4881 	if (err) {
4882 		e_dev_err("Cannot enable PCI device from suspend\n");
4883 		return err;
4884 	}
4885 	pci_set_master(pdev);
4886 
4887 	pci_wake_from_d3(pdev, false);
4888 
4889 	rtnl_lock();
4890 	err = ixgbe_init_interrupt_scheme(adapter);
4891 	rtnl_unlock();
4892 	if (err) {
4893 		e_dev_err("Cannot initialize interrupts for device\n");
4894 		return err;
4895 	}
4896 
4897 	ixgbe_reset(adapter);
4898 
4899 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
4900 
4901 	if (netif_running(netdev)) {
4902 		err = ixgbe_open(netdev);
4903 		if (err)
4904 			return err;
4905 	}
4906 
4907 	netif_device_attach(netdev);
4908 
4909 	return 0;
4910 }
4911 #endif /* CONFIG_PM */
4912 
4913 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
4914 {
4915 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
4916 	struct net_device *netdev = adapter->netdev;
4917 	struct ixgbe_hw *hw = &adapter->hw;
4918 	u32 ctrl, fctrl;
4919 	u32 wufc = adapter->wol;
4920 #ifdef CONFIG_PM
4921 	int retval = 0;
4922 #endif
4923 
4924 	netif_device_detach(netdev);
4925 
4926 	if (netif_running(netdev)) {
4927 		rtnl_lock();
4928 		ixgbe_down(adapter);
4929 		ixgbe_free_irq(adapter);
4930 		ixgbe_free_all_tx_resources(adapter);
4931 		ixgbe_free_all_rx_resources(adapter);
4932 		rtnl_unlock();
4933 	}
4934 
4935 	ixgbe_clear_interrupt_scheme(adapter);
4936 
4937 #ifdef CONFIG_PM
4938 	retval = pci_save_state(pdev);
4939 	if (retval)
4940 		return retval;
4941 
4942 #endif
4943 	if (wufc) {
4944 		ixgbe_set_rx_mode(netdev);
4945 
4946 		/*
4947 		 * enable the optics for both mult-speed fiber and
4948 		 * 82599 SFP+ fiber as we can WoL.
4949 		 */
4950 		if (hw->mac.ops.enable_tx_laser &&
4951 		    (hw->phy.multispeed_fiber ||
4952 		    (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_fiber &&
4953 		     hw->mac.type == ixgbe_mac_82599EB)))
4954 			hw->mac.ops.enable_tx_laser(hw);
4955 
4956 		/* turn on all-multi mode if wake on multicast is enabled */
4957 		if (wufc & IXGBE_WUFC_MC) {
4958 			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4959 			fctrl |= IXGBE_FCTRL_MPE;
4960 			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4961 		}
4962 
4963 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
4964 		ctrl |= IXGBE_CTRL_GIO_DIS;
4965 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
4966 
4967 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
4968 	} else {
4969 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
4970 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
4971 	}
4972 
4973 	switch (hw->mac.type) {
4974 	case ixgbe_mac_82598EB:
4975 		pci_wake_from_d3(pdev, false);
4976 		break;
4977 	case ixgbe_mac_82599EB:
4978 	case ixgbe_mac_X540:
4979 		pci_wake_from_d3(pdev, !!wufc);
4980 		break;
4981 	default:
4982 		break;
4983 	}
4984 
4985 	*enable_wake = !!wufc;
4986 
4987 	ixgbe_release_hw_control(adapter);
4988 
4989 	pci_disable_device(pdev);
4990 
4991 	return 0;
4992 }
4993 
4994 #ifdef CONFIG_PM
4995 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
4996 {
4997 	int retval;
4998 	bool wake;
4999 
5000 	retval = __ixgbe_shutdown(pdev, &wake);
5001 	if (retval)
5002 		return retval;
5003 
5004 	if (wake) {
5005 		pci_prepare_to_sleep(pdev);
5006 	} else {
5007 		pci_wake_from_d3(pdev, false);
5008 		pci_set_power_state(pdev, PCI_D3hot);
5009 	}
5010 
5011 	return 0;
5012 }
5013 #endif /* CONFIG_PM */
5014 
5015 static void ixgbe_shutdown(struct pci_dev *pdev)
5016 {
5017 	bool wake;
5018 
5019 	__ixgbe_shutdown(pdev, &wake);
5020 
5021 	if (system_state == SYSTEM_POWER_OFF) {
5022 		pci_wake_from_d3(pdev, wake);
5023 		pci_set_power_state(pdev, PCI_D3hot);
5024 	}
5025 }
5026 
5027 /**
5028  * ixgbe_update_stats - Update the board statistics counters.
5029  * @adapter: board private structure
5030  **/
5031 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
5032 {
5033 	struct net_device *netdev = adapter->netdev;
5034 	struct ixgbe_hw *hw = &adapter->hw;
5035 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
5036 	u64 total_mpc = 0;
5037 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
5038 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
5039 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
5040 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
5041 #ifdef IXGBE_FCOE
5042 	struct ixgbe_fcoe *fcoe = &adapter->fcoe;
5043 	unsigned int cpu;
5044 	u64 fcoe_noddp_counts_sum = 0, fcoe_noddp_ext_buff_counts_sum = 0;
5045 #endif /* IXGBE_FCOE */
5046 
5047 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5048 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5049 		return;
5050 
5051 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
5052 		u64 rsc_count = 0;
5053 		u64 rsc_flush = 0;
5054 		for (i = 0; i < adapter->num_rx_queues; i++) {
5055 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
5056 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
5057 		}
5058 		adapter->rsc_total_count = rsc_count;
5059 		adapter->rsc_total_flush = rsc_flush;
5060 	}
5061 
5062 	for (i = 0; i < adapter->num_rx_queues; i++) {
5063 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
5064 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
5065 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
5066 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
5067 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
5068 		bytes += rx_ring->stats.bytes;
5069 		packets += rx_ring->stats.packets;
5070 	}
5071 	adapter->non_eop_descs = non_eop_descs;
5072 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
5073 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
5074 	adapter->hw_csum_rx_error = hw_csum_rx_error;
5075 	netdev->stats.rx_bytes = bytes;
5076 	netdev->stats.rx_packets = packets;
5077 
5078 	bytes = 0;
5079 	packets = 0;
5080 	/* gather some stats to the adapter struct that are per queue */
5081 	for (i = 0; i < adapter->num_tx_queues; i++) {
5082 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
5083 		restart_queue += tx_ring->tx_stats.restart_queue;
5084 		tx_busy += tx_ring->tx_stats.tx_busy;
5085 		bytes += tx_ring->stats.bytes;
5086 		packets += tx_ring->stats.packets;
5087 	}
5088 	adapter->restart_queue = restart_queue;
5089 	adapter->tx_busy = tx_busy;
5090 	netdev->stats.tx_bytes = bytes;
5091 	netdev->stats.tx_packets = packets;
5092 
5093 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
5094 
5095 	/* 8 register reads */
5096 	for (i = 0; i < 8; i++) {
5097 		/* for packet buffers not used, the register should read 0 */
5098 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
5099 		missed_rx += mpc;
5100 		hwstats->mpc[i] += mpc;
5101 		total_mpc += hwstats->mpc[i];
5102 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
5103 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
5104 		switch (hw->mac.type) {
5105 		case ixgbe_mac_82598EB:
5106 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
5107 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
5108 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
5109 			hwstats->pxonrxc[i] +=
5110 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
5111 			break;
5112 		case ixgbe_mac_82599EB:
5113 		case ixgbe_mac_X540:
5114 			hwstats->pxonrxc[i] +=
5115 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
5116 			break;
5117 		default:
5118 			break;
5119 		}
5120 	}
5121 
5122 	/*16 register reads */
5123 	for (i = 0; i < 16; i++) {
5124 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
5125 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
5126 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
5127 		    (hw->mac.type == ixgbe_mac_X540)) {
5128 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
5129 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
5130 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
5131 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
5132 		}
5133 	}
5134 
5135 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
5136 	/* work around hardware counting issue */
5137 	hwstats->gprc -= missed_rx;
5138 
5139 	ixgbe_update_xoff_received(adapter);
5140 
5141 	/* 82598 hardware only has a 32 bit counter in the high register */
5142 	switch (hw->mac.type) {
5143 	case ixgbe_mac_82598EB:
5144 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
5145 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
5146 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
5147 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
5148 		break;
5149 	case ixgbe_mac_X540:
5150 		/* OS2BMC stats are X540 only*/
5151 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
5152 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
5153 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
5154 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
5155 	case ixgbe_mac_82599EB:
5156 		for (i = 0; i < 16; i++)
5157 			adapter->hw_rx_no_dma_resources +=
5158 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
5159 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
5160 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
5161 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
5162 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
5163 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
5164 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
5165 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
5166 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
5167 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
5168 #ifdef IXGBE_FCOE
5169 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
5170 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
5171 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
5172 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
5173 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
5174 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
5175 		/* Add up per cpu counters for total ddp aloc fail */
5176 		if (fcoe->pcpu_noddp && fcoe->pcpu_noddp_ext_buff) {
5177 			for_each_possible_cpu(cpu) {
5178 				fcoe_noddp_counts_sum +=
5179 					*per_cpu_ptr(fcoe->pcpu_noddp, cpu);
5180 				fcoe_noddp_ext_buff_counts_sum +=
5181 					*per_cpu_ptr(fcoe->
5182 						pcpu_noddp_ext_buff, cpu);
5183 			}
5184 		}
5185 		hwstats->fcoe_noddp = fcoe_noddp_counts_sum;
5186 		hwstats->fcoe_noddp_ext_buff = fcoe_noddp_ext_buff_counts_sum;
5187 #endif /* IXGBE_FCOE */
5188 		break;
5189 	default:
5190 		break;
5191 	}
5192 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
5193 	hwstats->bprc += bprc;
5194 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
5195 	if (hw->mac.type == ixgbe_mac_82598EB)
5196 		hwstats->mprc -= bprc;
5197 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
5198 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
5199 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
5200 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
5201 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
5202 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
5203 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
5204 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
5205 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
5206 	hwstats->lxontxc += lxon;
5207 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
5208 	hwstats->lxofftxc += lxoff;
5209 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
5210 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
5211 	/*
5212 	 * 82598 errata - tx of flow control packets is included in tx counters
5213 	 */
5214 	xon_off_tot = lxon + lxoff;
5215 	hwstats->gptc -= xon_off_tot;
5216 	hwstats->mptc -= xon_off_tot;
5217 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
5218 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
5219 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
5220 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
5221 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
5222 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
5223 	hwstats->ptc64 -= xon_off_tot;
5224 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
5225 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
5226 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
5227 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
5228 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
5229 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
5230 
5231 	/* Fill out the OS statistics structure */
5232 	netdev->stats.multicast = hwstats->mprc;
5233 
5234 	/* Rx Errors */
5235 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
5236 	netdev->stats.rx_dropped = 0;
5237 	netdev->stats.rx_length_errors = hwstats->rlec;
5238 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
5239 	netdev->stats.rx_missed_errors = total_mpc;
5240 }
5241 
5242 /**
5243  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
5244  * @adapter - pointer to the device adapter structure
5245  **/
5246 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
5247 {
5248 	struct ixgbe_hw *hw = &adapter->hw;
5249 	int i;
5250 
5251 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
5252 		return;
5253 
5254 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
5255 
5256 	/* if interface is down do nothing */
5257 	if (test_bit(__IXGBE_DOWN, &adapter->state))
5258 		return;
5259 
5260 	/* do nothing if we are not using signature filters */
5261 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
5262 		return;
5263 
5264 	adapter->fdir_overflow++;
5265 
5266 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
5267 		for (i = 0; i < adapter->num_tx_queues; i++)
5268 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
5269 			        &(adapter->tx_ring[i]->state));
5270 		/* re-enable flow director interrupts */
5271 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
5272 	} else {
5273 		e_err(probe, "failed to finish FDIR re-initialization, "
5274 		      "ignored adding FDIR ATR filters\n");
5275 	}
5276 }
5277 
5278 /**
5279  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
5280  * @adapter - pointer to the device adapter structure
5281  *
5282  * This function serves two purposes.  First it strobes the interrupt lines
5283  * in order to make certain interrupts are occurring.  Secondly it sets the
5284  * bits needed to check for TX hangs.  As a result we should immediately
5285  * determine if a hang has occurred.
5286  */
5287 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
5288 {
5289 	struct ixgbe_hw *hw = &adapter->hw;
5290 	u64 eics = 0;
5291 	int i;
5292 
5293 	/* If we're down or resetting, just bail */
5294 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5295 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5296 		return;
5297 
5298 	/* Force detection of hung controller */
5299 	if (netif_carrier_ok(adapter->netdev)) {
5300 		for (i = 0; i < adapter->num_tx_queues; i++)
5301 			set_check_for_tx_hang(adapter->tx_ring[i]);
5302 	}
5303 
5304 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
5305 		/*
5306 		 * for legacy and MSI interrupts don't set any bits
5307 		 * that are enabled for EIAM, because this operation
5308 		 * would set *both* EIMS and EICS for any bit in EIAM
5309 		 */
5310 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
5311 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
5312 	} else {
5313 		/* get one bit for every active tx/rx interrupt vector */
5314 		for (i = 0; i < adapter->num_msix_vectors - NON_Q_VECTORS; i++) {
5315 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
5316 			if (qv->rx.ring || qv->tx.ring)
5317 				eics |= ((u64)1 << i);
5318 		}
5319 	}
5320 
5321 	/* Cause software interrupt to ensure rings are cleaned */
5322 	ixgbe_irq_rearm_queues(adapter, eics);
5323 
5324 }
5325 
5326 /**
5327  * ixgbe_watchdog_update_link - update the link status
5328  * @adapter - pointer to the device adapter structure
5329  * @link_speed - pointer to a u32 to store the link_speed
5330  **/
5331 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
5332 {
5333 	struct ixgbe_hw *hw = &adapter->hw;
5334 	u32 link_speed = adapter->link_speed;
5335 	bool link_up = adapter->link_up;
5336 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
5337 
5338 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
5339 		return;
5340 
5341 	if (hw->mac.ops.check_link) {
5342 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
5343 	} else {
5344 		/* always assume link is up, if no check link function */
5345 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
5346 		link_up = true;
5347 	}
5348 
5349 	if (adapter->ixgbe_ieee_pfc)
5350 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
5351 
5352 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
5353 		hw->mac.ops.fc_enable(hw);
5354 		ixgbe_set_rx_drop_en(adapter);
5355 	}
5356 
5357 	if (link_up ||
5358 	    time_after(jiffies, (adapter->link_check_timeout +
5359 				 IXGBE_TRY_LINK_TIMEOUT))) {
5360 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5361 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
5362 		IXGBE_WRITE_FLUSH(hw);
5363 	}
5364 
5365 	adapter->link_up = link_up;
5366 	adapter->link_speed = link_speed;
5367 }
5368 
5369 /**
5370  * ixgbe_watchdog_link_is_up - update netif_carrier status and
5371  *                             print link up message
5372  * @adapter - pointer to the device adapter structure
5373  **/
5374 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
5375 {
5376 	struct net_device *netdev = adapter->netdev;
5377 	struct ixgbe_hw *hw = &adapter->hw;
5378 	u32 link_speed = adapter->link_speed;
5379 	bool flow_rx, flow_tx;
5380 
5381 	/* only continue if link was previously down */
5382 	if (netif_carrier_ok(netdev))
5383 		return;
5384 
5385 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
5386 
5387 	switch (hw->mac.type) {
5388 	case ixgbe_mac_82598EB: {
5389 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
5390 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
5391 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
5392 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
5393 	}
5394 		break;
5395 	case ixgbe_mac_X540:
5396 	case ixgbe_mac_82599EB: {
5397 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
5398 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
5399 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
5400 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
5401 	}
5402 		break;
5403 	default:
5404 		flow_tx = false;
5405 		flow_rx = false;
5406 		break;
5407 	}
5408 
5409 #ifdef CONFIG_IXGBE_PTP
5410 	ixgbe_ptp_start_cyclecounter(adapter);
5411 #endif
5412 
5413 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n",
5414 	       (link_speed == IXGBE_LINK_SPEED_10GB_FULL ?
5415 	       "10 Gbps" :
5416 	       (link_speed == IXGBE_LINK_SPEED_1GB_FULL ?
5417 	       "1 Gbps" :
5418 	       (link_speed == IXGBE_LINK_SPEED_100_FULL ?
5419 	       "100 Mbps" :
5420 	       "unknown speed"))),
5421 	       ((flow_rx && flow_tx) ? "RX/TX" :
5422 	       (flow_rx ? "RX" :
5423 	       (flow_tx ? "TX" : "None"))));
5424 
5425 	netif_carrier_on(netdev);
5426 	ixgbe_check_vf_rate_limit(adapter);
5427 }
5428 
5429 /**
5430  * ixgbe_watchdog_link_is_down - update netif_carrier status and
5431  *                               print link down message
5432  * @adapter - pointer to the adapter structure
5433  **/
5434 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
5435 {
5436 	struct net_device *netdev = adapter->netdev;
5437 	struct ixgbe_hw *hw = &adapter->hw;
5438 
5439 	adapter->link_up = false;
5440 	adapter->link_speed = 0;
5441 
5442 	/* only continue if link was up previously */
5443 	if (!netif_carrier_ok(netdev))
5444 		return;
5445 
5446 	/* poll for SFP+ cable when link is down */
5447 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
5448 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5449 
5450 #ifdef CONFIG_IXGBE_PTP
5451 	ixgbe_ptp_start_cyclecounter(adapter);
5452 #endif
5453 
5454 	e_info(drv, "NIC Link is Down\n");
5455 	netif_carrier_off(netdev);
5456 }
5457 
5458 /**
5459  * ixgbe_watchdog_flush_tx - flush queues on link down
5460  * @adapter - pointer to the device adapter structure
5461  **/
5462 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
5463 {
5464 	int i;
5465 	int some_tx_pending = 0;
5466 
5467 	if (!netif_carrier_ok(adapter->netdev)) {
5468 		for (i = 0; i < adapter->num_tx_queues; i++) {
5469 			struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
5470 			if (tx_ring->next_to_use != tx_ring->next_to_clean) {
5471 				some_tx_pending = 1;
5472 				break;
5473 			}
5474 		}
5475 
5476 		if (some_tx_pending) {
5477 			/* We've lost link, so the controller stops DMA,
5478 			 * but we've got queued Tx work that's never going
5479 			 * to get done, so reset controller to flush Tx.
5480 			 * (Do the reset outside of interrupt context).
5481 			 */
5482 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
5483 		}
5484 	}
5485 }
5486 
5487 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
5488 {
5489 	u32 ssvpc;
5490 
5491 	/* Do not perform spoof check for 82598 */
5492 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5493 		return;
5494 
5495 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
5496 
5497 	/*
5498 	 * ssvpc register is cleared on read, if zero then no
5499 	 * spoofed packets in the last interval.
5500 	 */
5501 	if (!ssvpc)
5502 		return;
5503 
5504 	e_warn(drv, "%d Spoofed packets detected\n", ssvpc);
5505 }
5506 
5507 /**
5508  * ixgbe_watchdog_subtask - check and bring link up
5509  * @adapter - pointer to the device adapter structure
5510  **/
5511 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
5512 {
5513 	/* if interface is down do nothing */
5514 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5515 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5516 		return;
5517 
5518 	ixgbe_watchdog_update_link(adapter);
5519 
5520 	if (adapter->link_up)
5521 		ixgbe_watchdog_link_is_up(adapter);
5522 	else
5523 		ixgbe_watchdog_link_is_down(adapter);
5524 
5525 	ixgbe_spoof_check(adapter);
5526 	ixgbe_update_stats(adapter);
5527 
5528 	ixgbe_watchdog_flush_tx(adapter);
5529 }
5530 
5531 /**
5532  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
5533  * @adapter - the ixgbe adapter structure
5534  **/
5535 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
5536 {
5537 	struct ixgbe_hw *hw = &adapter->hw;
5538 	s32 err;
5539 
5540 	/* not searching for SFP so there is nothing to do here */
5541 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
5542 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
5543 		return;
5544 
5545 	/* someone else is in init, wait until next service event */
5546 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5547 		return;
5548 
5549 	err = hw->phy.ops.identify_sfp(hw);
5550 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
5551 		goto sfp_out;
5552 
5553 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
5554 		/* If no cable is present, then we need to reset
5555 		 * the next time we find a good cable. */
5556 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5557 	}
5558 
5559 	/* exit on error */
5560 	if (err)
5561 		goto sfp_out;
5562 
5563 	/* exit if reset not needed */
5564 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
5565 		goto sfp_out;
5566 
5567 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
5568 
5569 	/*
5570 	 * A module may be identified correctly, but the EEPROM may not have
5571 	 * support for that module.  setup_sfp() will fail in that case, so
5572 	 * we should not allow that module to load.
5573 	 */
5574 	if (hw->mac.type == ixgbe_mac_82598EB)
5575 		err = hw->phy.ops.reset(hw);
5576 	else
5577 		err = hw->mac.ops.setup_sfp(hw);
5578 
5579 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
5580 		goto sfp_out;
5581 
5582 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
5583 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
5584 
5585 sfp_out:
5586 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5587 
5588 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
5589 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
5590 		e_dev_err("failed to initialize because an unsupported "
5591 			  "SFP+ module type was detected.\n");
5592 		e_dev_err("Reload the driver after installing a "
5593 			  "supported module.\n");
5594 		unregister_netdev(adapter->netdev);
5595 	}
5596 }
5597 
5598 /**
5599  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
5600  * @adapter - the ixgbe adapter structure
5601  **/
5602 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
5603 {
5604 	struct ixgbe_hw *hw = &adapter->hw;
5605 	u32 autoneg;
5606 	bool negotiation;
5607 
5608 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
5609 		return;
5610 
5611 	/* someone else is in init, wait until next service event */
5612 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5613 		return;
5614 
5615 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5616 
5617 	autoneg = hw->phy.autoneg_advertised;
5618 	if ((!autoneg) && (hw->mac.ops.get_link_capabilities))
5619 		hw->mac.ops.get_link_capabilities(hw, &autoneg, &negotiation);
5620 	if (hw->mac.ops.setup_link)
5621 		hw->mac.ops.setup_link(hw, autoneg, negotiation, true);
5622 
5623 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5624 	adapter->link_check_timeout = jiffies;
5625 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5626 }
5627 
5628 #ifdef CONFIG_PCI_IOV
5629 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
5630 {
5631 	int vf;
5632 	struct ixgbe_hw *hw = &adapter->hw;
5633 	struct net_device *netdev = adapter->netdev;
5634 	u32 gpc;
5635 	u32 ciaa, ciad;
5636 
5637 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
5638 	if (gpc) /* If incrementing then no need for the check below */
5639 		return;
5640 	/*
5641 	 * Check to see if a bad DMA write target from an errant or
5642 	 * malicious VF has caused a PCIe error.  If so then we can
5643 	 * issue a VFLR to the offending VF(s) and then resume without
5644 	 * requesting a full slot reset.
5645 	 */
5646 
5647 	for (vf = 0; vf < adapter->num_vfs; vf++) {
5648 		ciaa = (vf << 16) | 0x80000000;
5649 		/* 32 bit read so align, we really want status at offset 6 */
5650 		ciaa |= PCI_COMMAND;
5651 		IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5652 		ciad = IXGBE_READ_REG(hw, IXGBE_CIAD_82599);
5653 		ciaa &= 0x7FFFFFFF;
5654 		/* disable debug mode asap after reading data */
5655 		IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5656 		/* Get the upper 16 bits which will be the PCI status reg */
5657 		ciad >>= 16;
5658 		if (ciad & PCI_STATUS_REC_MASTER_ABORT) {
5659 			netdev_err(netdev, "VF %d Hung DMA\n", vf);
5660 			/* Issue VFLR */
5661 			ciaa = (vf << 16) | 0x80000000;
5662 			ciaa |= 0xA8;
5663 			IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5664 			ciad = 0x00008000;  /* VFLR */
5665 			IXGBE_WRITE_REG(hw, IXGBE_CIAD_82599, ciad);
5666 			ciaa &= 0x7FFFFFFF;
5667 			IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5668 		}
5669 	}
5670 }
5671 
5672 #endif
5673 /**
5674  * ixgbe_service_timer - Timer Call-back
5675  * @data: pointer to adapter cast into an unsigned long
5676  **/
5677 static void ixgbe_service_timer(unsigned long data)
5678 {
5679 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
5680 	unsigned long next_event_offset;
5681 	bool ready = true;
5682 
5683 	/* poll faster when waiting for link */
5684 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
5685 		next_event_offset = HZ / 10;
5686 	else
5687 		next_event_offset = HZ * 2;
5688 
5689 #ifdef CONFIG_PCI_IOV
5690 	/*
5691 	 * don't bother with SR-IOV VF DMA hang check if there are
5692 	 * no VFs or the link is down
5693 	 */
5694 	if (!adapter->num_vfs ||
5695 	    (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
5696 		goto normal_timer_service;
5697 
5698 	/* If we have VFs allocated then we must check for DMA hangs */
5699 	ixgbe_check_for_bad_vf(adapter);
5700 	next_event_offset = HZ / 50;
5701 	adapter->timer_event_accumulator++;
5702 
5703 	if (adapter->timer_event_accumulator >= 100)
5704 		adapter->timer_event_accumulator = 0;
5705 	else
5706 		ready = false;
5707 
5708 normal_timer_service:
5709 #endif
5710 	/* Reset the timer */
5711 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
5712 
5713 	if (ready)
5714 		ixgbe_service_event_schedule(adapter);
5715 }
5716 
5717 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
5718 {
5719 	if (!(adapter->flags2 & IXGBE_FLAG2_RESET_REQUESTED))
5720 		return;
5721 
5722 	adapter->flags2 &= ~IXGBE_FLAG2_RESET_REQUESTED;
5723 
5724 	/* If we're already down or resetting, just bail */
5725 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5726 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5727 		return;
5728 
5729 	ixgbe_dump(adapter);
5730 	netdev_err(adapter->netdev, "Reset adapter\n");
5731 	adapter->tx_timeout_count++;
5732 
5733 	ixgbe_reinit_locked(adapter);
5734 }
5735 
5736 /**
5737  * ixgbe_service_task - manages and runs subtasks
5738  * @work: pointer to work_struct containing our data
5739  **/
5740 static void ixgbe_service_task(struct work_struct *work)
5741 {
5742 	struct ixgbe_adapter *adapter = container_of(work,
5743 						     struct ixgbe_adapter,
5744 						     service_task);
5745 
5746 	ixgbe_reset_subtask(adapter);
5747 	ixgbe_sfp_detection_subtask(adapter);
5748 	ixgbe_sfp_link_config_subtask(adapter);
5749 	ixgbe_check_overtemp_subtask(adapter);
5750 	ixgbe_watchdog_subtask(adapter);
5751 	ixgbe_fdir_reinit_subtask(adapter);
5752 	ixgbe_check_hang_subtask(adapter);
5753 #ifdef CONFIG_IXGBE_PTP
5754 	ixgbe_ptp_overflow_check(adapter);
5755 #endif
5756 
5757 	ixgbe_service_event_complete(adapter);
5758 }
5759 
5760 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
5761 		     struct ixgbe_tx_buffer *first,
5762 		     u8 *hdr_len)
5763 {
5764 	struct sk_buff *skb = first->skb;
5765 	u32 vlan_macip_lens, type_tucmd;
5766 	u32 mss_l4len_idx, l4len;
5767 
5768 	if (!skb_is_gso(skb))
5769 		return 0;
5770 
5771 	if (skb_header_cloned(skb)) {
5772 		int err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
5773 		if (err)
5774 			return err;
5775 	}
5776 
5777 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
5778 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
5779 
5780 	if (first->protocol == __constant_htons(ETH_P_IP)) {
5781 		struct iphdr *iph = ip_hdr(skb);
5782 		iph->tot_len = 0;
5783 		iph->check = 0;
5784 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
5785 							 iph->daddr, 0,
5786 							 IPPROTO_TCP,
5787 							 0);
5788 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
5789 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
5790 				   IXGBE_TX_FLAGS_CSUM |
5791 				   IXGBE_TX_FLAGS_IPV4;
5792 	} else if (skb_is_gso_v6(skb)) {
5793 		ipv6_hdr(skb)->payload_len = 0;
5794 		tcp_hdr(skb)->check =
5795 		    ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
5796 				     &ipv6_hdr(skb)->daddr,
5797 				     0, IPPROTO_TCP, 0);
5798 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
5799 				   IXGBE_TX_FLAGS_CSUM;
5800 	}
5801 
5802 	/* compute header lengths */
5803 	l4len = tcp_hdrlen(skb);
5804 	*hdr_len = skb_transport_offset(skb) + l4len;
5805 
5806 	/* update gso size and bytecount with header size */
5807 	first->gso_segs = skb_shinfo(skb)->gso_segs;
5808 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
5809 
5810 	/* mss_l4len_id: use 1 as index for TSO */
5811 	mss_l4len_idx = l4len << IXGBE_ADVTXD_L4LEN_SHIFT;
5812 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
5813 	mss_l4len_idx |= 1 << IXGBE_ADVTXD_IDX_SHIFT;
5814 
5815 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
5816 	vlan_macip_lens = skb_network_header_len(skb);
5817 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
5818 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
5819 
5820 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
5821 			  mss_l4len_idx);
5822 
5823 	return 1;
5824 }
5825 
5826 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
5827 			  struct ixgbe_tx_buffer *first)
5828 {
5829 	struct sk_buff *skb = first->skb;
5830 	u32 vlan_macip_lens = 0;
5831 	u32 mss_l4len_idx = 0;
5832 	u32 type_tucmd = 0;
5833 
5834 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
5835 		if (!(first->tx_flags & IXGBE_TX_FLAGS_HW_VLAN) &&
5836 		    !(first->tx_flags & IXGBE_TX_FLAGS_TXSW))
5837 			return;
5838 	} else {
5839 		u8 l4_hdr = 0;
5840 		switch (first->protocol) {
5841 		case __constant_htons(ETH_P_IP):
5842 			vlan_macip_lens |= skb_network_header_len(skb);
5843 			type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
5844 			l4_hdr = ip_hdr(skb)->protocol;
5845 			break;
5846 		case __constant_htons(ETH_P_IPV6):
5847 			vlan_macip_lens |= skb_network_header_len(skb);
5848 			l4_hdr = ipv6_hdr(skb)->nexthdr;
5849 			break;
5850 		default:
5851 			if (unlikely(net_ratelimit())) {
5852 				dev_warn(tx_ring->dev,
5853 				 "partial checksum but proto=%x!\n",
5854 				 first->protocol);
5855 			}
5856 			break;
5857 		}
5858 
5859 		switch (l4_hdr) {
5860 		case IPPROTO_TCP:
5861 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_TCP;
5862 			mss_l4len_idx = tcp_hdrlen(skb) <<
5863 					IXGBE_ADVTXD_L4LEN_SHIFT;
5864 			break;
5865 		case IPPROTO_SCTP:
5866 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_SCTP;
5867 			mss_l4len_idx = sizeof(struct sctphdr) <<
5868 					IXGBE_ADVTXD_L4LEN_SHIFT;
5869 			break;
5870 		case IPPROTO_UDP:
5871 			mss_l4len_idx = sizeof(struct udphdr) <<
5872 					IXGBE_ADVTXD_L4LEN_SHIFT;
5873 			break;
5874 		default:
5875 			if (unlikely(net_ratelimit())) {
5876 				dev_warn(tx_ring->dev,
5877 				 "partial checksum but l4 proto=%x!\n",
5878 				 l4_hdr);
5879 			}
5880 			break;
5881 		}
5882 
5883 		/* update TX checksum flag */
5884 		first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
5885 	}
5886 
5887 	/* vlan_macip_lens: MACLEN, VLAN tag */
5888 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
5889 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
5890 
5891 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0,
5892 			  type_tucmd, mss_l4len_idx);
5893 }
5894 
5895 static __le32 ixgbe_tx_cmd_type(u32 tx_flags)
5896 {
5897 	/* set type for advanced descriptor with frame checksum insertion */
5898 	__le32 cmd_type = cpu_to_le32(IXGBE_ADVTXD_DTYP_DATA |
5899 				      IXGBE_ADVTXD_DCMD_IFCS |
5900 				      IXGBE_ADVTXD_DCMD_DEXT);
5901 
5902 	/* set HW vlan bit if vlan is present */
5903 	if (tx_flags & IXGBE_TX_FLAGS_HW_VLAN)
5904 		cmd_type |= cpu_to_le32(IXGBE_ADVTXD_DCMD_VLE);
5905 
5906 #ifdef CONFIG_IXGBE_PTP
5907 	if (tx_flags & IXGBE_TX_FLAGS_TSTAMP)
5908 		cmd_type |= cpu_to_le32(IXGBE_ADVTXD_MAC_TSTAMP);
5909 #endif
5910 
5911 	/* set segmentation enable bits for TSO/FSO */
5912 #ifdef IXGBE_FCOE
5913 	if (tx_flags & (IXGBE_TX_FLAGS_TSO | IXGBE_TX_FLAGS_FSO))
5914 #else
5915 	if (tx_flags & IXGBE_TX_FLAGS_TSO)
5916 #endif
5917 		cmd_type |= cpu_to_le32(IXGBE_ADVTXD_DCMD_TSE);
5918 
5919 	return cmd_type;
5920 }
5921 
5922 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
5923 				   u32 tx_flags, unsigned int paylen)
5924 {
5925 	__le32 olinfo_status = cpu_to_le32(paylen << IXGBE_ADVTXD_PAYLEN_SHIFT);
5926 
5927 	/* enable L4 checksum for TSO and TX checksum offload */
5928 	if (tx_flags & IXGBE_TX_FLAGS_CSUM)
5929 		olinfo_status |= cpu_to_le32(IXGBE_ADVTXD_POPTS_TXSM);
5930 
5931 	/* enble IPv4 checksum for TSO */
5932 	if (tx_flags & IXGBE_TX_FLAGS_IPV4)
5933 		olinfo_status |= cpu_to_le32(IXGBE_ADVTXD_POPTS_IXSM);
5934 
5935 	/* use index 1 context for TSO/FSO/FCOE */
5936 #ifdef IXGBE_FCOE
5937 	if (tx_flags & (IXGBE_TX_FLAGS_TSO | IXGBE_TX_FLAGS_FCOE))
5938 #else
5939 	if (tx_flags & IXGBE_TX_FLAGS_TSO)
5940 #endif
5941 		olinfo_status |= cpu_to_le32(1 << IXGBE_ADVTXD_IDX_SHIFT);
5942 
5943 	/*
5944 	 * Check Context must be set if Tx switch is enabled, which it
5945 	 * always is for case where virtual functions are running
5946 	 */
5947 #ifdef IXGBE_FCOE
5948 	if (tx_flags & (IXGBE_TX_FLAGS_TXSW | IXGBE_TX_FLAGS_FCOE))
5949 #else
5950 	if (tx_flags & IXGBE_TX_FLAGS_TXSW)
5951 #endif
5952 		olinfo_status |= cpu_to_le32(IXGBE_ADVTXD_CC);
5953 
5954 	tx_desc->read.olinfo_status = olinfo_status;
5955 }
5956 
5957 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
5958 		       IXGBE_TXD_CMD_RS)
5959 
5960 static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
5961 			 struct ixgbe_tx_buffer *first,
5962 			 const u8 hdr_len)
5963 {
5964 	dma_addr_t dma;
5965 	struct sk_buff *skb = first->skb;
5966 	struct ixgbe_tx_buffer *tx_buffer;
5967 	union ixgbe_adv_tx_desc *tx_desc;
5968 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
5969 	unsigned int data_len = skb->data_len;
5970 	unsigned int size = skb_headlen(skb);
5971 	unsigned int paylen = skb->len - hdr_len;
5972 	u32 tx_flags = first->tx_flags;
5973 	__le32 cmd_type;
5974 	u16 i = tx_ring->next_to_use;
5975 
5976 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
5977 
5978 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, paylen);
5979 	cmd_type = ixgbe_tx_cmd_type(tx_flags);
5980 
5981 #ifdef IXGBE_FCOE
5982 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
5983 		if (data_len < sizeof(struct fcoe_crc_eof)) {
5984 			size -= sizeof(struct fcoe_crc_eof) - data_len;
5985 			data_len = 0;
5986 		} else {
5987 			data_len -= sizeof(struct fcoe_crc_eof);
5988 		}
5989 	}
5990 
5991 #endif
5992 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
5993 	if (dma_mapping_error(tx_ring->dev, dma))
5994 		goto dma_error;
5995 
5996 	/* record length, and DMA address */
5997 	dma_unmap_len_set(first, len, size);
5998 	dma_unmap_addr_set(first, dma, dma);
5999 
6000 	tx_desc->read.buffer_addr = cpu_to_le64(dma);
6001 
6002 	for (;;) {
6003 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
6004 			tx_desc->read.cmd_type_len =
6005 				cmd_type | cpu_to_le32(IXGBE_MAX_DATA_PER_TXD);
6006 
6007 			i++;
6008 			tx_desc++;
6009 			if (i == tx_ring->count) {
6010 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6011 				i = 0;
6012 			}
6013 
6014 			dma += IXGBE_MAX_DATA_PER_TXD;
6015 			size -= IXGBE_MAX_DATA_PER_TXD;
6016 
6017 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
6018 			tx_desc->read.olinfo_status = 0;
6019 		}
6020 
6021 		if (likely(!data_len))
6022 			break;
6023 
6024 		if (unlikely(skb->no_fcs))
6025 			cmd_type &= ~(cpu_to_le32(IXGBE_ADVTXD_DCMD_IFCS));
6026 		tx_desc->read.cmd_type_len = cmd_type | cpu_to_le32(size);
6027 
6028 		i++;
6029 		tx_desc++;
6030 		if (i == tx_ring->count) {
6031 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6032 			i = 0;
6033 		}
6034 
6035 #ifdef IXGBE_FCOE
6036 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
6037 #else
6038 		size = skb_frag_size(frag);
6039 #endif
6040 		data_len -= size;
6041 
6042 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
6043 				       DMA_TO_DEVICE);
6044 		if (dma_mapping_error(tx_ring->dev, dma))
6045 			goto dma_error;
6046 
6047 		tx_buffer = &tx_ring->tx_buffer_info[i];
6048 		dma_unmap_len_set(tx_buffer, len, size);
6049 		dma_unmap_addr_set(tx_buffer, dma, dma);
6050 
6051 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
6052 		tx_desc->read.olinfo_status = 0;
6053 
6054 		frag++;
6055 	}
6056 
6057 	/* write last descriptor with RS and EOP bits */
6058 	cmd_type |= cpu_to_le32(size) | cpu_to_le32(IXGBE_TXD_CMD);
6059 	tx_desc->read.cmd_type_len = cmd_type;
6060 
6061 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
6062 
6063 	/* set the timestamp */
6064 	first->time_stamp = jiffies;
6065 
6066 	/*
6067 	 * Force memory writes to complete before letting h/w know there
6068 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
6069 	 * memory model archs, such as IA-64).
6070 	 *
6071 	 * We also need this memory barrier to make certain all of the
6072 	 * status bits have been updated before next_to_watch is written.
6073 	 */
6074 	wmb();
6075 
6076 	/* set next_to_watch value indicating a packet is present */
6077 	first->next_to_watch = tx_desc;
6078 
6079 	i++;
6080 	if (i == tx_ring->count)
6081 		i = 0;
6082 
6083 	tx_ring->next_to_use = i;
6084 
6085 	/* notify HW of packet */
6086 	writel(i, tx_ring->tail);
6087 
6088 	return;
6089 dma_error:
6090 	dev_err(tx_ring->dev, "TX DMA map failed\n");
6091 
6092 	/* clear dma mappings for failed tx_buffer_info map */
6093 	for (;;) {
6094 		tx_buffer = &tx_ring->tx_buffer_info[i];
6095 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer);
6096 		if (tx_buffer == first)
6097 			break;
6098 		if (i == 0)
6099 			i = tx_ring->count;
6100 		i--;
6101 	}
6102 
6103 	tx_ring->next_to_use = i;
6104 }
6105 
6106 static void ixgbe_atr(struct ixgbe_ring *ring,
6107 		      struct ixgbe_tx_buffer *first)
6108 {
6109 	struct ixgbe_q_vector *q_vector = ring->q_vector;
6110 	union ixgbe_atr_hash_dword input = { .dword = 0 };
6111 	union ixgbe_atr_hash_dword common = { .dword = 0 };
6112 	union {
6113 		unsigned char *network;
6114 		struct iphdr *ipv4;
6115 		struct ipv6hdr *ipv6;
6116 	} hdr;
6117 	struct tcphdr *th;
6118 	__be16 vlan_id;
6119 
6120 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
6121 	if (!q_vector)
6122 		return;
6123 
6124 	/* do nothing if sampling is disabled */
6125 	if (!ring->atr_sample_rate)
6126 		return;
6127 
6128 	ring->atr_count++;
6129 
6130 	/* snag network header to get L4 type and address */
6131 	hdr.network = skb_network_header(first->skb);
6132 
6133 	/* Currently only IPv4/IPv6 with TCP is supported */
6134 	if ((first->protocol != __constant_htons(ETH_P_IPV6) ||
6135 	     hdr.ipv6->nexthdr != IPPROTO_TCP) &&
6136 	    (first->protocol != __constant_htons(ETH_P_IP) ||
6137 	     hdr.ipv4->protocol != IPPROTO_TCP))
6138 		return;
6139 
6140 	th = tcp_hdr(first->skb);
6141 
6142 	/* skip this packet since it is invalid or the socket is closing */
6143 	if (!th || th->fin)
6144 		return;
6145 
6146 	/* sample on all syn packets or once every atr sample count */
6147 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
6148 		return;
6149 
6150 	/* reset sample count */
6151 	ring->atr_count = 0;
6152 
6153 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
6154 
6155 	/*
6156 	 * src and dst are inverted, think how the receiver sees them
6157 	 *
6158 	 * The input is broken into two sections, a non-compressed section
6159 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
6160 	 * is XORed together and stored in the compressed dword.
6161 	 */
6162 	input.formatted.vlan_id = vlan_id;
6163 
6164 	/*
6165 	 * since src port and flex bytes occupy the same word XOR them together
6166 	 * and write the value to source port portion of compressed dword
6167 	 */
6168 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
6169 		common.port.src ^= th->dest ^ __constant_htons(ETH_P_8021Q);
6170 	else
6171 		common.port.src ^= th->dest ^ first->protocol;
6172 	common.port.dst ^= th->source;
6173 
6174 	if (first->protocol == __constant_htons(ETH_P_IP)) {
6175 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
6176 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
6177 	} else {
6178 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
6179 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
6180 			     hdr.ipv6->saddr.s6_addr32[1] ^
6181 			     hdr.ipv6->saddr.s6_addr32[2] ^
6182 			     hdr.ipv6->saddr.s6_addr32[3] ^
6183 			     hdr.ipv6->daddr.s6_addr32[0] ^
6184 			     hdr.ipv6->daddr.s6_addr32[1] ^
6185 			     hdr.ipv6->daddr.s6_addr32[2] ^
6186 			     hdr.ipv6->daddr.s6_addr32[3];
6187 	}
6188 
6189 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
6190 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
6191 					      input, common, ring->queue_index);
6192 }
6193 
6194 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
6195 {
6196 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
6197 	/* Herbert's original patch had:
6198 	 *  smp_mb__after_netif_stop_queue();
6199 	 * but since that doesn't exist yet, just open code it. */
6200 	smp_mb();
6201 
6202 	/* We need to check again in a case another CPU has just
6203 	 * made room available. */
6204 	if (likely(ixgbe_desc_unused(tx_ring) < size))
6205 		return -EBUSY;
6206 
6207 	/* A reprieve! - use start_queue because it doesn't call schedule */
6208 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
6209 	++tx_ring->tx_stats.restart_queue;
6210 	return 0;
6211 }
6212 
6213 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
6214 {
6215 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
6216 		return 0;
6217 	return __ixgbe_maybe_stop_tx(tx_ring, size);
6218 }
6219 
6220 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb)
6221 {
6222 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6223 	int txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
6224 					       smp_processor_id();
6225 #ifdef IXGBE_FCOE
6226 	__be16 protocol = vlan_get_protocol(skb);
6227 
6228 	if (((protocol == htons(ETH_P_FCOE)) ||
6229 	    (protocol == htons(ETH_P_FIP))) &&
6230 	    (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)) {
6231 		txq &= (adapter->ring_feature[RING_F_FCOE].indices - 1);
6232 		txq += adapter->ring_feature[RING_F_FCOE].mask;
6233 		return txq;
6234 	}
6235 #endif
6236 
6237 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
6238 		while (unlikely(txq >= dev->real_num_tx_queues))
6239 			txq -= dev->real_num_tx_queues;
6240 		return txq;
6241 	}
6242 
6243 	return skb_tx_hash(dev, skb);
6244 }
6245 
6246 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
6247 			  struct ixgbe_adapter *adapter,
6248 			  struct ixgbe_ring *tx_ring)
6249 {
6250 	struct ixgbe_tx_buffer *first;
6251 	int tso;
6252 	u32 tx_flags = 0;
6253 #if PAGE_SIZE > IXGBE_MAX_DATA_PER_TXD
6254 	unsigned short f;
6255 #endif
6256 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
6257 	__be16 protocol = skb->protocol;
6258 	u8 hdr_len = 0;
6259 
6260 	/*
6261 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
6262 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
6263 	 *       + 2 desc gap to keep tail from touching head,
6264 	 *       + 1 desc for context descriptor,
6265 	 * otherwise try next time
6266 	 */
6267 #if PAGE_SIZE > IXGBE_MAX_DATA_PER_TXD
6268 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
6269 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
6270 #else
6271 	count += skb_shinfo(skb)->nr_frags;
6272 #endif
6273 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
6274 		tx_ring->tx_stats.tx_busy++;
6275 		return NETDEV_TX_BUSY;
6276 	}
6277 
6278 	/* record the location of the first descriptor for this packet */
6279 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
6280 	first->skb = skb;
6281 	first->bytecount = skb->len;
6282 	first->gso_segs = 1;
6283 
6284 	/* if we have a HW VLAN tag being added default to the HW one */
6285 	if (vlan_tx_tag_present(skb)) {
6286 		tx_flags |= vlan_tx_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
6287 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
6288 	/* else if it is a SW VLAN check the next protocol and store the tag */
6289 	} else if (protocol == __constant_htons(ETH_P_8021Q)) {
6290 		struct vlan_hdr *vhdr, _vhdr;
6291 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
6292 		if (!vhdr)
6293 			goto out_drop;
6294 
6295 		protocol = vhdr->h_vlan_encapsulated_proto;
6296 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
6297 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
6298 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
6299 	}
6300 
6301 	skb_tx_timestamp(skb);
6302 
6303 #ifdef CONFIG_IXGBE_PTP
6304 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) {
6305 		skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
6306 		tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
6307 	}
6308 #endif
6309 
6310 #ifdef CONFIG_PCI_IOV
6311 	/*
6312 	 * Use the l2switch_enable flag - would be false if the DMA
6313 	 * Tx switch had been disabled.
6314 	 */
6315 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
6316 		tx_flags |= IXGBE_TX_FLAGS_TXSW;
6317 
6318 #endif
6319 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
6320 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
6321 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
6322 	     (skb->priority != TC_PRIO_CONTROL))) {
6323 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
6324 		tx_flags |= (skb->priority & 0x7) <<
6325 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
6326 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
6327 			struct vlan_ethhdr *vhdr;
6328 			if (skb_header_cloned(skb) &&
6329 			    pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
6330 				goto out_drop;
6331 			vhdr = (struct vlan_ethhdr *)skb->data;
6332 			vhdr->h_vlan_TCI = htons(tx_flags >>
6333 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
6334 		} else {
6335 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
6336 		}
6337 	}
6338 
6339 	/* record initial flags and protocol */
6340 	first->tx_flags = tx_flags;
6341 	first->protocol = protocol;
6342 
6343 #ifdef IXGBE_FCOE
6344 	/* setup tx offload for FCoE */
6345 	if ((protocol == __constant_htons(ETH_P_FCOE)) &&
6346 	    (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)) {
6347 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
6348 		if (tso < 0)
6349 			goto out_drop;
6350 
6351 		goto xmit_fcoe;
6352 	}
6353 
6354 #endif /* IXGBE_FCOE */
6355 	tso = ixgbe_tso(tx_ring, first, &hdr_len);
6356 	if (tso < 0)
6357 		goto out_drop;
6358 	else if (!tso)
6359 		ixgbe_tx_csum(tx_ring, first);
6360 
6361 	/* add the ATR filter if ATR is on */
6362 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
6363 		ixgbe_atr(tx_ring, first);
6364 
6365 #ifdef IXGBE_FCOE
6366 xmit_fcoe:
6367 #endif /* IXGBE_FCOE */
6368 	ixgbe_tx_map(tx_ring, first, hdr_len);
6369 
6370 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
6371 
6372 	return NETDEV_TX_OK;
6373 
6374 out_drop:
6375 	dev_kfree_skb_any(first->skb);
6376 	first->skb = NULL;
6377 
6378 	return NETDEV_TX_OK;
6379 }
6380 
6381 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
6382 				    struct net_device *netdev)
6383 {
6384 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6385 	struct ixgbe_ring *tx_ring;
6386 
6387 	if (skb->len <= 0) {
6388 		dev_kfree_skb_any(skb);
6389 		return NETDEV_TX_OK;
6390 	}
6391 
6392 	/*
6393 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
6394 	 * in order to meet this minimum size requirement.
6395 	 */
6396 	if (skb->len < 17) {
6397 		if (skb_padto(skb, 17))
6398 			return NETDEV_TX_OK;
6399 		skb->len = 17;
6400 	}
6401 
6402 	tx_ring = adapter->tx_ring[skb->queue_mapping];
6403 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
6404 }
6405 
6406 /**
6407  * ixgbe_set_mac - Change the Ethernet Address of the NIC
6408  * @netdev: network interface device structure
6409  * @p: pointer to an address structure
6410  *
6411  * Returns 0 on success, negative on failure
6412  **/
6413 static int ixgbe_set_mac(struct net_device *netdev, void *p)
6414 {
6415 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6416 	struct ixgbe_hw *hw = &adapter->hw;
6417 	struct sockaddr *addr = p;
6418 
6419 	if (!is_valid_ether_addr(addr->sa_data))
6420 		return -EADDRNOTAVAIL;
6421 
6422 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
6423 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
6424 
6425 	hw->mac.ops.set_rar(hw, 0, hw->mac.addr, adapter->num_vfs,
6426 			    IXGBE_RAH_AV);
6427 
6428 	return 0;
6429 }
6430 
6431 static int
6432 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
6433 {
6434 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6435 	struct ixgbe_hw *hw = &adapter->hw;
6436 	u16 value;
6437 	int rc;
6438 
6439 	if (prtad != hw->phy.mdio.prtad)
6440 		return -EINVAL;
6441 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
6442 	if (!rc)
6443 		rc = value;
6444 	return rc;
6445 }
6446 
6447 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
6448 			    u16 addr, u16 value)
6449 {
6450 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6451 	struct ixgbe_hw *hw = &adapter->hw;
6452 
6453 	if (prtad != hw->phy.mdio.prtad)
6454 		return -EINVAL;
6455 	return hw->phy.ops.write_reg(hw, addr, devad, value);
6456 }
6457 
6458 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
6459 {
6460 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6461 
6462 	switch (cmd) {
6463 #ifdef CONFIG_IXGBE_PTP
6464 	case SIOCSHWTSTAMP:
6465 		return ixgbe_ptp_hwtstamp_ioctl(adapter, req, cmd);
6466 #endif
6467 	default:
6468 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
6469 	}
6470 }
6471 
6472 /**
6473  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
6474  * netdev->dev_addrs
6475  * @netdev: network interface device structure
6476  *
6477  * Returns non-zero on failure
6478  **/
6479 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
6480 {
6481 	int err = 0;
6482 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6483 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
6484 
6485 	if (is_valid_ether_addr(mac->san_addr)) {
6486 		rtnl_lock();
6487 		err = dev_addr_add(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
6488 		rtnl_unlock();
6489 	}
6490 	return err;
6491 }
6492 
6493 /**
6494  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
6495  * netdev->dev_addrs
6496  * @netdev: network interface device structure
6497  *
6498  * Returns non-zero on failure
6499  **/
6500 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
6501 {
6502 	int err = 0;
6503 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6504 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
6505 
6506 	if (is_valid_ether_addr(mac->san_addr)) {
6507 		rtnl_lock();
6508 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
6509 		rtnl_unlock();
6510 	}
6511 	return err;
6512 }
6513 
6514 #ifdef CONFIG_NET_POLL_CONTROLLER
6515 /*
6516  * Polling 'interrupt' - used by things like netconsole to send skbs
6517  * without having to re-enable interrupts. It's not called while
6518  * the interrupt routine is executing.
6519  */
6520 static void ixgbe_netpoll(struct net_device *netdev)
6521 {
6522 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6523 	int i;
6524 
6525 	/* if interface is down do nothing */
6526 	if (test_bit(__IXGBE_DOWN, &adapter->state))
6527 		return;
6528 
6529 	adapter->flags |= IXGBE_FLAG_IN_NETPOLL;
6530 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
6531 		int num_q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
6532 		for (i = 0; i < num_q_vectors; i++) {
6533 			struct ixgbe_q_vector *q_vector = adapter->q_vector[i];
6534 			ixgbe_msix_clean_rings(0, q_vector);
6535 		}
6536 	} else {
6537 		ixgbe_intr(adapter->pdev->irq, netdev);
6538 	}
6539 	adapter->flags &= ~IXGBE_FLAG_IN_NETPOLL;
6540 }
6541 
6542 #endif
6543 static struct rtnl_link_stats64 *ixgbe_get_stats64(struct net_device *netdev,
6544 						   struct rtnl_link_stats64 *stats)
6545 {
6546 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6547 	int i;
6548 
6549 	rcu_read_lock();
6550 	for (i = 0; i < adapter->num_rx_queues; i++) {
6551 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
6552 		u64 bytes, packets;
6553 		unsigned int start;
6554 
6555 		if (ring) {
6556 			do {
6557 				start = u64_stats_fetch_begin_bh(&ring->syncp);
6558 				packets = ring->stats.packets;
6559 				bytes   = ring->stats.bytes;
6560 			} while (u64_stats_fetch_retry_bh(&ring->syncp, start));
6561 			stats->rx_packets += packets;
6562 			stats->rx_bytes   += bytes;
6563 		}
6564 	}
6565 
6566 	for (i = 0; i < adapter->num_tx_queues; i++) {
6567 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
6568 		u64 bytes, packets;
6569 		unsigned int start;
6570 
6571 		if (ring) {
6572 			do {
6573 				start = u64_stats_fetch_begin_bh(&ring->syncp);
6574 				packets = ring->stats.packets;
6575 				bytes   = ring->stats.bytes;
6576 			} while (u64_stats_fetch_retry_bh(&ring->syncp, start));
6577 			stats->tx_packets += packets;
6578 			stats->tx_bytes   += bytes;
6579 		}
6580 	}
6581 	rcu_read_unlock();
6582 	/* following stats updated by ixgbe_watchdog_task() */
6583 	stats->multicast	= netdev->stats.multicast;
6584 	stats->rx_errors	= netdev->stats.rx_errors;
6585 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
6586 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
6587 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
6588 	return stats;
6589 }
6590 
6591 #ifdef CONFIG_IXGBE_DCB
6592 /* ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
6593  * #adapter: pointer to ixgbe_adapter
6594  * @tc: number of traffic classes currently enabled
6595  *
6596  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
6597  * 802.1Q priority maps to a packet buffer that exists.
6598  */
6599 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
6600 {
6601 	struct ixgbe_hw *hw = &adapter->hw;
6602 	u32 reg, rsave;
6603 	int i;
6604 
6605 	/* 82598 have a static priority to TC mapping that can not
6606 	 * be changed so no validation is needed.
6607 	 */
6608 	if (hw->mac.type == ixgbe_mac_82598EB)
6609 		return;
6610 
6611 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
6612 	rsave = reg;
6613 
6614 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
6615 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
6616 
6617 		/* If up2tc is out of bounds default to zero */
6618 		if (up2tc > tc)
6619 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
6620 	}
6621 
6622 	if (reg != rsave)
6623 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
6624 
6625 	return;
6626 }
6627 
6628 /* ixgbe_setup_tc - routine to configure net_device for multiple traffic
6629  * classes.
6630  *
6631  * @netdev: net device to configure
6632  * @tc: number of traffic classes to enable
6633  */
6634 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
6635 {
6636 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6637 	struct ixgbe_hw *hw = &adapter->hw;
6638 
6639 	/* Multiple traffic classes requires multiple queues */
6640 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
6641 		e_err(drv, "Enable failed, needs MSI-X\n");
6642 		return -EINVAL;
6643 	}
6644 
6645 	/* Hardware supports up to 8 traffic classes */
6646 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs ||
6647 	    (hw->mac.type == ixgbe_mac_82598EB &&
6648 	     tc < MAX_TRAFFIC_CLASS))
6649 		return -EINVAL;
6650 
6651 	/* Hardware has to reinitialize queues and interrupts to
6652 	 * match packet buffer alignment. Unfortunately, the
6653 	 * hardware is not flexible enough to do this dynamically.
6654 	 */
6655 	if (netif_running(dev))
6656 		ixgbe_close(dev);
6657 	ixgbe_clear_interrupt_scheme(adapter);
6658 
6659 	if (tc) {
6660 		netdev_set_num_tc(dev, tc);
6661 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
6662 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
6663 
6664 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
6665 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
6666 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
6667 		}
6668 	} else {
6669 		netdev_reset_tc(dev);
6670 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
6671 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
6672 
6673 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
6674 		adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
6675 
6676 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
6677 		adapter->dcb_cfg.pfc_mode_enable = false;
6678 	}
6679 
6680 	ixgbe_init_interrupt_scheme(adapter);
6681 	ixgbe_validate_rtr(adapter, tc);
6682 	if (netif_running(dev))
6683 		ixgbe_open(dev);
6684 
6685 	return 0;
6686 }
6687 
6688 #endif /* CONFIG_IXGBE_DCB */
6689 void ixgbe_do_reset(struct net_device *netdev)
6690 {
6691 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6692 
6693 	if (netif_running(netdev))
6694 		ixgbe_reinit_locked(adapter);
6695 	else
6696 		ixgbe_reset(adapter);
6697 }
6698 
6699 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
6700 					    netdev_features_t features)
6701 {
6702 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6703 
6704 #ifdef CONFIG_DCB
6705 	if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
6706 		features &= ~NETIF_F_HW_VLAN_RX;
6707 #endif
6708 
6709 	/* return error if RXHASH is being enabled when RSS is not supported */
6710 	if (!(adapter->flags & IXGBE_FLAG_RSS_ENABLED))
6711 		features &= ~NETIF_F_RXHASH;
6712 
6713 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
6714 	if (!(features & NETIF_F_RXCSUM))
6715 		features &= ~NETIF_F_LRO;
6716 
6717 	/* Turn off LRO if not RSC capable */
6718 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
6719 		features &= ~NETIF_F_LRO;
6720 
6721 
6722 	return features;
6723 }
6724 
6725 static int ixgbe_set_features(struct net_device *netdev,
6726 			      netdev_features_t features)
6727 {
6728 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6729 	netdev_features_t changed = netdev->features ^ features;
6730 	bool need_reset = false;
6731 
6732 	/* Make sure RSC matches LRO, reset if change */
6733 	if (!(features & NETIF_F_LRO)) {
6734 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
6735 			need_reset = true;
6736 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
6737 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
6738 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
6739 		if (adapter->rx_itr_setting == 1 ||
6740 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
6741 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
6742 			need_reset = true;
6743 		} else if ((changed ^ features) & NETIF_F_LRO) {
6744 			e_info(probe, "rx-usecs set too low, "
6745 			       "disabling RSC\n");
6746 		}
6747 	}
6748 
6749 	/*
6750 	 * Check if Flow Director n-tuple support was enabled or disabled.  If
6751 	 * the state changed, we need to reset.
6752 	 */
6753 	if (!(features & NETIF_F_NTUPLE)) {
6754 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
6755 			/* turn off Flow Director, set ATR and reset */
6756 			if ((adapter->flags & IXGBE_FLAG_RSS_ENABLED) &&
6757 			    !(adapter->flags & IXGBE_FLAG_DCB_ENABLED))
6758 				adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
6759 			need_reset = true;
6760 		}
6761 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
6762 	} else if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)) {
6763 		/* turn off ATR, enable perfect filters and reset */
6764 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
6765 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
6766 		need_reset = true;
6767 	}
6768 
6769 	if (changed & NETIF_F_RXALL)
6770 		need_reset = true;
6771 
6772 	netdev->features = features;
6773 	if (need_reset)
6774 		ixgbe_do_reset(netdev);
6775 
6776 	return 0;
6777 }
6778 
6779 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm,
6780 			     struct net_device *dev,
6781 			     unsigned char *addr,
6782 			     u16 flags)
6783 {
6784 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6785 	int err = -EOPNOTSUPP;
6786 
6787 	if (ndm->ndm_state & NUD_PERMANENT) {
6788 		pr_info("%s: FDB only supports static addresses\n",
6789 			ixgbe_driver_name);
6790 		return -EINVAL;
6791 	}
6792 
6793 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
6794 		if (is_unicast_ether_addr(addr))
6795 			err = dev_uc_add_excl(dev, addr);
6796 		else if (is_multicast_ether_addr(addr))
6797 			err = dev_mc_add_excl(dev, addr);
6798 		else
6799 			err = -EINVAL;
6800 	}
6801 
6802 	/* Only return duplicate errors if NLM_F_EXCL is set */
6803 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
6804 		err = 0;
6805 
6806 	return err;
6807 }
6808 
6809 static int ixgbe_ndo_fdb_del(struct ndmsg *ndm,
6810 			     struct net_device *dev,
6811 			     unsigned char *addr)
6812 {
6813 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6814 	int err = -EOPNOTSUPP;
6815 
6816 	if (ndm->ndm_state & NUD_PERMANENT) {
6817 		pr_info("%s: FDB only supports static addresses\n",
6818 			ixgbe_driver_name);
6819 		return -EINVAL;
6820 	}
6821 
6822 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
6823 		if (is_unicast_ether_addr(addr))
6824 			err = dev_uc_del(dev, addr);
6825 		else if (is_multicast_ether_addr(addr))
6826 			err = dev_mc_del(dev, addr);
6827 		else
6828 			err = -EINVAL;
6829 	}
6830 
6831 	return err;
6832 }
6833 
6834 static int ixgbe_ndo_fdb_dump(struct sk_buff *skb,
6835 			      struct netlink_callback *cb,
6836 			      struct net_device *dev,
6837 			      int idx)
6838 {
6839 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6840 
6841 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
6842 		idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
6843 
6844 	return idx;
6845 }
6846 
6847 static const struct net_device_ops ixgbe_netdev_ops = {
6848 	.ndo_open		= ixgbe_open,
6849 	.ndo_stop		= ixgbe_close,
6850 	.ndo_start_xmit		= ixgbe_xmit_frame,
6851 	.ndo_select_queue	= ixgbe_select_queue,
6852 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
6853 	.ndo_validate_addr	= eth_validate_addr,
6854 	.ndo_set_mac_address	= ixgbe_set_mac,
6855 	.ndo_change_mtu		= ixgbe_change_mtu,
6856 	.ndo_tx_timeout		= ixgbe_tx_timeout,
6857 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
6858 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
6859 	.ndo_do_ioctl		= ixgbe_ioctl,
6860 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
6861 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
6862 	.ndo_set_vf_tx_rate	= ixgbe_ndo_set_vf_bw,
6863 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
6864 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
6865 	.ndo_get_stats64	= ixgbe_get_stats64,
6866 #ifdef CONFIG_IXGBE_DCB
6867 	.ndo_setup_tc		= ixgbe_setup_tc,
6868 #endif
6869 #ifdef CONFIG_NET_POLL_CONTROLLER
6870 	.ndo_poll_controller	= ixgbe_netpoll,
6871 #endif
6872 #ifdef IXGBE_FCOE
6873 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
6874 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
6875 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
6876 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
6877 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
6878 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
6879 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
6880 #endif /* IXGBE_FCOE */
6881 	.ndo_set_features = ixgbe_set_features,
6882 	.ndo_fix_features = ixgbe_fix_features,
6883 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
6884 	.ndo_fdb_del		= ixgbe_ndo_fdb_del,
6885 	.ndo_fdb_dump		= ixgbe_ndo_fdb_dump,
6886 };
6887 
6888 static void __devinit ixgbe_probe_vf(struct ixgbe_adapter *adapter,
6889 				     const struct ixgbe_info *ii)
6890 {
6891 #ifdef CONFIG_PCI_IOV
6892 	struct ixgbe_hw *hw = &adapter->hw;
6893 
6894 	if (hw->mac.type == ixgbe_mac_82598EB)
6895 		return;
6896 
6897 	/* The 82599 supports up to 64 VFs per physical function
6898 	 * but this implementation limits allocation to 63 so that
6899 	 * basic networking resources are still available to the
6900 	 * physical function.  If the user requests greater thn
6901 	 * 63 VFs then it is an error - reset to default of zero.
6902 	 */
6903 	adapter->num_vfs = (max_vfs > 63) ? 0 : max_vfs;
6904 	ixgbe_enable_sriov(adapter, ii);
6905 #endif /* CONFIG_PCI_IOV */
6906 }
6907 
6908 /**
6909  * ixgbe_wol_supported - Check whether device supports WoL
6910  * @hw: hw specific details
6911  * @device_id: the device ID
6912  * @subdev_id: the subsystem device ID
6913  *
6914  * This function is used by probe and ethtool to determine
6915  * which devices have WoL support
6916  *
6917  **/
6918 int ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
6919 			u16 subdevice_id)
6920 {
6921 	struct ixgbe_hw *hw = &adapter->hw;
6922 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
6923 	int is_wol_supported = 0;
6924 
6925 	switch (device_id) {
6926 	case IXGBE_DEV_ID_82599_SFP:
6927 		/* Only these subdevices could supports WOL */
6928 		switch (subdevice_id) {
6929 		case IXGBE_SUBDEV_ID_82599_560FLR:
6930 			/* only support first port */
6931 			if (hw->bus.func != 0)
6932 				break;
6933 		case IXGBE_SUBDEV_ID_82599_SFP:
6934 			is_wol_supported = 1;
6935 			break;
6936 		}
6937 		break;
6938 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
6939 		/* All except this subdevice support WOL */
6940 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
6941 			is_wol_supported = 1;
6942 		break;
6943 	case IXGBE_DEV_ID_82599_KX4:
6944 		is_wol_supported = 1;
6945 		break;
6946 	case IXGBE_DEV_ID_X540T:
6947 		/* check eeprom to see if enabled wol */
6948 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
6949 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
6950 		     (hw->bus.func == 0))) {
6951 			is_wol_supported = 1;
6952 		}
6953 		break;
6954 	}
6955 
6956 	return is_wol_supported;
6957 }
6958 
6959 /**
6960  * ixgbe_probe - Device Initialization Routine
6961  * @pdev: PCI device information struct
6962  * @ent: entry in ixgbe_pci_tbl
6963  *
6964  * Returns 0 on success, negative on failure
6965  *
6966  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
6967  * The OS initialization, configuring of the adapter private structure,
6968  * and a hardware reset occur.
6969  **/
6970 static int __devinit ixgbe_probe(struct pci_dev *pdev,
6971 				 const struct pci_device_id *ent)
6972 {
6973 	struct net_device *netdev;
6974 	struct ixgbe_adapter *adapter = NULL;
6975 	struct ixgbe_hw *hw;
6976 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
6977 	static int cards_found;
6978 	int i, err, pci_using_dac;
6979 	u8 part_str[IXGBE_PBANUM_LENGTH];
6980 	unsigned int indices = num_possible_cpus();
6981 #ifdef IXGBE_FCOE
6982 	u16 device_caps;
6983 #endif
6984 	u32 eec;
6985 
6986 	/* Catch broken hardware that put the wrong VF device ID in
6987 	 * the PCIe SR-IOV capability.
6988 	 */
6989 	if (pdev->is_virtfn) {
6990 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
6991 		     pci_name(pdev), pdev->vendor, pdev->device);
6992 		return -EINVAL;
6993 	}
6994 
6995 	err = pci_enable_device_mem(pdev);
6996 	if (err)
6997 		return err;
6998 
6999 	if (!dma_set_mask(&pdev->dev, DMA_BIT_MASK(64)) &&
7000 	    !dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(64))) {
7001 		pci_using_dac = 1;
7002 	} else {
7003 		err = dma_set_mask(&pdev->dev, DMA_BIT_MASK(32));
7004 		if (err) {
7005 			err = dma_set_coherent_mask(&pdev->dev,
7006 						    DMA_BIT_MASK(32));
7007 			if (err) {
7008 				dev_err(&pdev->dev,
7009 					"No usable DMA configuration, aborting\n");
7010 				goto err_dma;
7011 			}
7012 		}
7013 		pci_using_dac = 0;
7014 	}
7015 
7016 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
7017 					   IORESOURCE_MEM), ixgbe_driver_name);
7018 	if (err) {
7019 		dev_err(&pdev->dev,
7020 			"pci_request_selected_regions failed 0x%x\n", err);
7021 		goto err_pci_reg;
7022 	}
7023 
7024 	pci_enable_pcie_error_reporting(pdev);
7025 
7026 	pci_set_master(pdev);
7027 	pci_save_state(pdev);
7028 
7029 #ifdef CONFIG_IXGBE_DCB
7030 	indices *= MAX_TRAFFIC_CLASS;
7031 #endif
7032 
7033 	if (ii->mac == ixgbe_mac_82598EB)
7034 		indices = min_t(unsigned int, indices, IXGBE_MAX_RSS_INDICES);
7035 	else
7036 		indices = min_t(unsigned int, indices, IXGBE_MAX_FDIR_INDICES);
7037 
7038 #ifdef IXGBE_FCOE
7039 	indices += min_t(unsigned int, num_possible_cpus(),
7040 			 IXGBE_MAX_FCOE_INDICES);
7041 #endif
7042 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
7043 	if (!netdev) {
7044 		err = -ENOMEM;
7045 		goto err_alloc_etherdev;
7046 	}
7047 
7048 	SET_NETDEV_DEV(netdev, &pdev->dev);
7049 
7050 	adapter = netdev_priv(netdev);
7051 	pci_set_drvdata(pdev, adapter);
7052 
7053 	adapter->netdev = netdev;
7054 	adapter->pdev = pdev;
7055 	hw = &adapter->hw;
7056 	hw->back = adapter;
7057 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
7058 
7059 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
7060 			      pci_resource_len(pdev, 0));
7061 	if (!hw->hw_addr) {
7062 		err = -EIO;
7063 		goto err_ioremap;
7064 	}
7065 
7066 	for (i = 1; i <= 5; i++) {
7067 		if (pci_resource_len(pdev, i) == 0)
7068 			continue;
7069 	}
7070 
7071 	netdev->netdev_ops = &ixgbe_netdev_ops;
7072 	ixgbe_set_ethtool_ops(netdev);
7073 	netdev->watchdog_timeo = 5 * HZ;
7074 	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
7075 
7076 	adapter->bd_number = cards_found;
7077 
7078 	/* Setup hw api */
7079 	memcpy(&hw->mac.ops, ii->mac_ops, sizeof(hw->mac.ops));
7080 	hw->mac.type  = ii->mac;
7081 
7082 	/* EEPROM */
7083 	memcpy(&hw->eeprom.ops, ii->eeprom_ops, sizeof(hw->eeprom.ops));
7084 	eec = IXGBE_READ_REG(hw, IXGBE_EEC);
7085 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
7086 	if (!(eec & (1 << 8)))
7087 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
7088 
7089 	/* PHY */
7090 	memcpy(&hw->phy.ops, ii->phy_ops, sizeof(hw->phy.ops));
7091 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
7092 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
7093 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
7094 	hw->phy.mdio.mmds = 0;
7095 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
7096 	hw->phy.mdio.dev = netdev;
7097 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
7098 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
7099 
7100 	ii->get_invariants(hw);
7101 
7102 	/* setup the private structure */
7103 	err = ixgbe_sw_init(adapter);
7104 	if (err)
7105 		goto err_sw_init;
7106 
7107 	/* Make it possible the adapter to be woken up via WOL */
7108 	switch (adapter->hw.mac.type) {
7109 	case ixgbe_mac_82599EB:
7110 	case ixgbe_mac_X540:
7111 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
7112 		break;
7113 	default:
7114 		break;
7115 	}
7116 
7117 	/*
7118 	 * If there is a fan on this device and it has failed log the
7119 	 * failure.
7120 	 */
7121 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
7122 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
7123 		if (esdp & IXGBE_ESDP_SDP1)
7124 			e_crit(probe, "Fan has stopped, replace the adapter\n");
7125 	}
7126 
7127 	if (allow_unsupported_sfp)
7128 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
7129 
7130 	/* reset_hw fills in the perm_addr as well */
7131 	hw->phy.reset_if_overtemp = true;
7132 	err = hw->mac.ops.reset_hw(hw);
7133 	hw->phy.reset_if_overtemp = false;
7134 	if (err == IXGBE_ERR_SFP_NOT_PRESENT &&
7135 	    hw->mac.type == ixgbe_mac_82598EB) {
7136 		err = 0;
7137 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
7138 		e_dev_err("failed to load because an unsupported SFP+ "
7139 			  "module type was detected.\n");
7140 		e_dev_err("Reload the driver after installing a supported "
7141 			  "module.\n");
7142 		goto err_sw_init;
7143 	} else if (err) {
7144 		e_dev_err("HW Init failed: %d\n", err);
7145 		goto err_sw_init;
7146 	}
7147 
7148 	ixgbe_probe_vf(adapter, ii);
7149 
7150 	netdev->features = NETIF_F_SG |
7151 			   NETIF_F_IP_CSUM |
7152 			   NETIF_F_IPV6_CSUM |
7153 			   NETIF_F_HW_VLAN_TX |
7154 			   NETIF_F_HW_VLAN_RX |
7155 			   NETIF_F_HW_VLAN_FILTER |
7156 			   NETIF_F_TSO |
7157 			   NETIF_F_TSO6 |
7158 			   NETIF_F_RXHASH |
7159 			   NETIF_F_RXCSUM;
7160 
7161 	netdev->hw_features = netdev->features;
7162 
7163 	switch (adapter->hw.mac.type) {
7164 	case ixgbe_mac_82599EB:
7165 	case ixgbe_mac_X540:
7166 		netdev->features |= NETIF_F_SCTP_CSUM;
7167 		netdev->hw_features |= NETIF_F_SCTP_CSUM |
7168 				       NETIF_F_NTUPLE;
7169 		break;
7170 	default:
7171 		break;
7172 	}
7173 
7174 	netdev->hw_features |= NETIF_F_RXALL;
7175 
7176 	netdev->vlan_features |= NETIF_F_TSO;
7177 	netdev->vlan_features |= NETIF_F_TSO6;
7178 	netdev->vlan_features |= NETIF_F_IP_CSUM;
7179 	netdev->vlan_features |= NETIF_F_IPV6_CSUM;
7180 	netdev->vlan_features |= NETIF_F_SG;
7181 
7182 	netdev->priv_flags |= IFF_UNICAST_FLT;
7183 	netdev->priv_flags |= IFF_SUPP_NOFCS;
7184 
7185 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7186 		adapter->flags &= ~(IXGBE_FLAG_RSS_ENABLED |
7187 				    IXGBE_FLAG_DCB_ENABLED);
7188 
7189 #ifdef CONFIG_IXGBE_DCB
7190 	netdev->dcbnl_ops = &dcbnl_ops;
7191 #endif
7192 
7193 #ifdef IXGBE_FCOE
7194 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
7195 		if (hw->mac.ops.get_device_caps) {
7196 			hw->mac.ops.get_device_caps(hw, &device_caps);
7197 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
7198 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
7199 		}
7200 	}
7201 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
7202 		netdev->vlan_features |= NETIF_F_FCOE_CRC;
7203 		netdev->vlan_features |= NETIF_F_FSO;
7204 		netdev->vlan_features |= NETIF_F_FCOE_MTU;
7205 	}
7206 #endif /* IXGBE_FCOE */
7207 	if (pci_using_dac) {
7208 		netdev->features |= NETIF_F_HIGHDMA;
7209 		netdev->vlan_features |= NETIF_F_HIGHDMA;
7210 	}
7211 
7212 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
7213 		netdev->hw_features |= NETIF_F_LRO;
7214 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
7215 		netdev->features |= NETIF_F_LRO;
7216 
7217 	/* make sure the EEPROM is good */
7218 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
7219 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
7220 		err = -EIO;
7221 		goto err_sw_init;
7222 	}
7223 
7224 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
7225 	memcpy(netdev->perm_addr, hw->mac.perm_addr, netdev->addr_len);
7226 
7227 	if (ixgbe_validate_mac_addr(netdev->perm_addr)) {
7228 		e_dev_err("invalid MAC address\n");
7229 		err = -EIO;
7230 		goto err_sw_init;
7231 	}
7232 
7233 	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
7234 		    (unsigned long) adapter);
7235 
7236 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
7237 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
7238 
7239 	err = ixgbe_init_interrupt_scheme(adapter);
7240 	if (err)
7241 		goto err_sw_init;
7242 
7243 	if (!(adapter->flags & IXGBE_FLAG_RSS_ENABLED)) {
7244 		netdev->hw_features &= ~NETIF_F_RXHASH;
7245 		netdev->features &= ~NETIF_F_RXHASH;
7246 	}
7247 
7248 	/* WOL not supported for all devices */
7249 	adapter->wol = 0;
7250 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
7251 	if (ixgbe_wol_supported(adapter, pdev->device, pdev->subsystem_device))
7252 		adapter->wol = IXGBE_WUFC_MAG;
7253 
7254 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
7255 
7256 #ifdef CONFIG_IXGBE_PTP
7257 	ixgbe_ptp_init(adapter);
7258 #endif /* CONFIG_IXGBE_PTP*/
7259 
7260 	/* save off EEPROM version number */
7261 	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
7262 	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
7263 
7264 	/* pick up the PCI bus settings for reporting later */
7265 	hw->mac.ops.get_bus_info(hw);
7266 
7267 	/* print bus type/speed/width info */
7268 	e_dev_info("(PCI Express:%s:%s) %pM\n",
7269 		   (hw->bus.speed == ixgbe_bus_speed_5000 ? "5.0GT/s" :
7270 		    hw->bus.speed == ixgbe_bus_speed_2500 ? "2.5GT/s" :
7271 		    "Unknown"),
7272 		   (hw->bus.width == ixgbe_bus_width_pcie_x8 ? "Width x8" :
7273 		    hw->bus.width == ixgbe_bus_width_pcie_x4 ? "Width x4" :
7274 		    hw->bus.width == ixgbe_bus_width_pcie_x1 ? "Width x1" :
7275 		    "Unknown"),
7276 		   netdev->dev_addr);
7277 
7278 	err = ixgbe_read_pba_string_generic(hw, part_str, IXGBE_PBANUM_LENGTH);
7279 	if (err)
7280 		strncpy(part_str, "Unknown", IXGBE_PBANUM_LENGTH);
7281 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
7282 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
7283 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
7284 		           part_str);
7285 	else
7286 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
7287 			   hw->mac.type, hw->phy.type, part_str);
7288 
7289 	if (hw->bus.width <= ixgbe_bus_width_pcie_x4) {
7290 		e_dev_warn("PCI-Express bandwidth available for this card is "
7291 			   "not sufficient for optimal performance.\n");
7292 		e_dev_warn("For optimal performance a x8 PCI-Express slot "
7293 			   "is required.\n");
7294 	}
7295 
7296 	/* reset the hardware with the new settings */
7297 	err = hw->mac.ops.start_hw(hw);
7298 	if (err == IXGBE_ERR_EEPROM_VERSION) {
7299 		/* We are running on a pre-production device, log a warning */
7300 		e_dev_warn("This device is a pre-production adapter/LOM. "
7301 			   "Please be aware there may be issues associated "
7302 			   "with your hardware.  If you are experiencing "
7303 			   "problems please contact your Intel or hardware "
7304 			   "representative who provided you with this "
7305 			   "hardware.\n");
7306 	}
7307 	strcpy(netdev->name, "eth%d");
7308 	err = register_netdev(netdev);
7309 	if (err)
7310 		goto err_register;
7311 
7312 	/* power down the optics for multispeed fiber and 82599 SFP+ fiber */
7313 	if (hw->mac.ops.disable_tx_laser &&
7314 	    ((hw->phy.multispeed_fiber) ||
7315 	     ((hw->mac.ops.get_media_type(hw) == ixgbe_media_type_fiber) &&
7316 	      (hw->mac.type == ixgbe_mac_82599EB))))
7317 		hw->mac.ops.disable_tx_laser(hw);
7318 
7319 	/* carrier off reporting is important to ethtool even BEFORE open */
7320 	netif_carrier_off(netdev);
7321 
7322 #ifdef CONFIG_IXGBE_DCA
7323 	if (dca_add_requester(&pdev->dev) == 0) {
7324 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
7325 		ixgbe_setup_dca(adapter);
7326 	}
7327 #endif
7328 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
7329 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
7330 		for (i = 0; i < adapter->num_vfs; i++)
7331 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
7332 	}
7333 
7334 	/* firmware requires driver version to be 0xFFFFFFFF
7335 	 * since os does not support feature
7336 	 */
7337 	if (hw->mac.ops.set_fw_drv_ver)
7338 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF,
7339 					   0xFF);
7340 
7341 	/* add san mac addr to netdev */
7342 	ixgbe_add_sanmac_netdev(netdev);
7343 
7344 	e_dev_info("%s\n", ixgbe_default_device_descr);
7345 	cards_found++;
7346 
7347 #ifdef CONFIG_IXGBE_HWMON
7348 	if (ixgbe_sysfs_init(adapter))
7349 		e_err(probe, "failed to allocate sysfs resources\n");
7350 #endif /* CONFIG_IXGBE_HWMON */
7351 
7352 	return 0;
7353 
7354 err_register:
7355 	ixgbe_release_hw_control(adapter);
7356 	ixgbe_clear_interrupt_scheme(adapter);
7357 err_sw_init:
7358 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7359 		ixgbe_disable_sriov(adapter);
7360 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
7361 	iounmap(hw->hw_addr);
7362 err_ioremap:
7363 	free_netdev(netdev);
7364 err_alloc_etherdev:
7365 	pci_release_selected_regions(pdev,
7366 				     pci_select_bars(pdev, IORESOURCE_MEM));
7367 err_pci_reg:
7368 err_dma:
7369 	pci_disable_device(pdev);
7370 	return err;
7371 }
7372 
7373 /**
7374  * ixgbe_remove - Device Removal Routine
7375  * @pdev: PCI device information struct
7376  *
7377  * ixgbe_remove is called by the PCI subsystem to alert the driver
7378  * that it should release a PCI device.  The could be caused by a
7379  * Hot-Plug event, or because the driver is going to be removed from
7380  * memory.
7381  **/
7382 static void __devexit ixgbe_remove(struct pci_dev *pdev)
7383 {
7384 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7385 	struct net_device *netdev = adapter->netdev;
7386 
7387 	set_bit(__IXGBE_DOWN, &adapter->state);
7388 	cancel_work_sync(&adapter->service_task);
7389 
7390 #ifdef CONFIG_IXGBE_PTP
7391 	ixgbe_ptp_stop(adapter);
7392 #endif
7393 
7394 #ifdef CONFIG_IXGBE_DCA
7395 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
7396 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
7397 		dca_remove_requester(&pdev->dev);
7398 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
7399 	}
7400 
7401 #endif
7402 #ifdef CONFIG_IXGBE_HWMON
7403 	ixgbe_sysfs_exit(adapter);
7404 #endif /* CONFIG_IXGBE_HWMON */
7405 
7406 #ifdef IXGBE_FCOE
7407 	if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
7408 		ixgbe_cleanup_fcoe(adapter);
7409 
7410 #endif /* IXGBE_FCOE */
7411 
7412 	/* remove the added san mac */
7413 	ixgbe_del_sanmac_netdev(netdev);
7414 
7415 	if (netdev->reg_state == NETREG_REGISTERED)
7416 		unregister_netdev(netdev);
7417 
7418 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
7419 		if (!(ixgbe_check_vf_assignment(adapter)))
7420 			ixgbe_disable_sriov(adapter);
7421 		else
7422 			e_dev_warn("Unloading driver while VFs are assigned "
7423 				   "- VFs will not be deallocated\n");
7424 	}
7425 
7426 	ixgbe_clear_interrupt_scheme(adapter);
7427 
7428 	ixgbe_release_hw_control(adapter);
7429 
7430 #ifdef CONFIG_DCB
7431 	kfree(adapter->ixgbe_ieee_pfc);
7432 	kfree(adapter->ixgbe_ieee_ets);
7433 
7434 #endif
7435 	iounmap(adapter->hw.hw_addr);
7436 	pci_release_selected_regions(pdev, pci_select_bars(pdev,
7437 				     IORESOURCE_MEM));
7438 
7439 	e_dev_info("complete\n");
7440 
7441 	free_netdev(netdev);
7442 
7443 	pci_disable_pcie_error_reporting(pdev);
7444 
7445 	pci_disable_device(pdev);
7446 }
7447 
7448 /**
7449  * ixgbe_io_error_detected - called when PCI error is detected
7450  * @pdev: Pointer to PCI device
7451  * @state: The current pci connection state
7452  *
7453  * This function is called after a PCI bus error affecting
7454  * this device has been detected.
7455  */
7456 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
7457 						pci_channel_state_t state)
7458 {
7459 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7460 	struct net_device *netdev = adapter->netdev;
7461 
7462 #ifdef CONFIG_PCI_IOV
7463 	struct pci_dev *bdev, *vfdev;
7464 	u32 dw0, dw1, dw2, dw3;
7465 	int vf, pos;
7466 	u16 req_id, pf_func;
7467 
7468 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7469 	    adapter->num_vfs == 0)
7470 		goto skip_bad_vf_detection;
7471 
7472 	bdev = pdev->bus->self;
7473 	while (bdev && (bdev->pcie_type != PCI_EXP_TYPE_ROOT_PORT))
7474 		bdev = bdev->bus->self;
7475 
7476 	if (!bdev)
7477 		goto skip_bad_vf_detection;
7478 
7479 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
7480 	if (!pos)
7481 		goto skip_bad_vf_detection;
7482 
7483 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG, &dw0);
7484 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG + 4, &dw1);
7485 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG + 8, &dw2);
7486 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG + 12, &dw3);
7487 
7488 	req_id = dw1 >> 16;
7489 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
7490 	if (!(req_id & 0x0080))
7491 		goto skip_bad_vf_detection;
7492 
7493 	pf_func = req_id & 0x01;
7494 	if ((pf_func & 1) == (pdev->devfn & 1)) {
7495 		unsigned int device_id;
7496 
7497 		vf = (req_id & 0x7F) >> 1;
7498 		e_dev_err("VF %d has caused a PCIe error\n", vf);
7499 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
7500 				"%8.8x\tdw3: %8.8x\n",
7501 		dw0, dw1, dw2, dw3);
7502 		switch (adapter->hw.mac.type) {
7503 		case ixgbe_mac_82599EB:
7504 			device_id = IXGBE_82599_VF_DEVICE_ID;
7505 			break;
7506 		case ixgbe_mac_X540:
7507 			device_id = IXGBE_X540_VF_DEVICE_ID;
7508 			break;
7509 		default:
7510 			device_id = 0;
7511 			break;
7512 		}
7513 
7514 		/* Find the pci device of the offending VF */
7515 		vfdev = pci_get_device(IXGBE_INTEL_VENDOR_ID, device_id, NULL);
7516 		while (vfdev) {
7517 			if (vfdev->devfn == (req_id & 0xFF))
7518 				break;
7519 			vfdev = pci_get_device(IXGBE_INTEL_VENDOR_ID,
7520 					       device_id, vfdev);
7521 		}
7522 		/*
7523 		 * There's a slim chance the VF could have been hot plugged,
7524 		 * so if it is no longer present we don't need to issue the
7525 		 * VFLR.  Just clean up the AER in that case.
7526 		 */
7527 		if (vfdev) {
7528 			e_dev_err("Issuing VFLR to VF %d\n", vf);
7529 			pci_write_config_dword(vfdev, 0xA8, 0x00008000);
7530 		}
7531 
7532 		pci_cleanup_aer_uncorrect_error_status(pdev);
7533 	}
7534 
7535 	/*
7536 	 * Even though the error may have occurred on the other port
7537 	 * we still need to increment the vf error reference count for
7538 	 * both ports because the I/O resume function will be called
7539 	 * for both of them.
7540 	 */
7541 	adapter->vferr_refcount++;
7542 
7543 	return PCI_ERS_RESULT_RECOVERED;
7544 
7545 skip_bad_vf_detection:
7546 #endif /* CONFIG_PCI_IOV */
7547 	netif_device_detach(netdev);
7548 
7549 	if (state == pci_channel_io_perm_failure)
7550 		return PCI_ERS_RESULT_DISCONNECT;
7551 
7552 	if (netif_running(netdev))
7553 		ixgbe_down(adapter);
7554 	pci_disable_device(pdev);
7555 
7556 	/* Request a slot reset. */
7557 	return PCI_ERS_RESULT_NEED_RESET;
7558 }
7559 
7560 /**
7561  * ixgbe_io_slot_reset - called after the pci bus has been reset.
7562  * @pdev: Pointer to PCI device
7563  *
7564  * Restart the card from scratch, as if from a cold-boot.
7565  */
7566 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
7567 {
7568 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7569 	pci_ers_result_t result;
7570 	int err;
7571 
7572 	if (pci_enable_device_mem(pdev)) {
7573 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
7574 		result = PCI_ERS_RESULT_DISCONNECT;
7575 	} else {
7576 		pci_set_master(pdev);
7577 		pci_restore_state(pdev);
7578 		pci_save_state(pdev);
7579 
7580 		pci_wake_from_d3(pdev, false);
7581 
7582 		ixgbe_reset(adapter);
7583 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
7584 		result = PCI_ERS_RESULT_RECOVERED;
7585 	}
7586 
7587 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
7588 	if (err) {
7589 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
7590 			  "failed 0x%0x\n", err);
7591 		/* non-fatal, continue */
7592 	}
7593 
7594 	return result;
7595 }
7596 
7597 /**
7598  * ixgbe_io_resume - called when traffic can start flowing again.
7599  * @pdev: Pointer to PCI device
7600  *
7601  * This callback is called when the error recovery driver tells us that
7602  * its OK to resume normal operation.
7603  */
7604 static void ixgbe_io_resume(struct pci_dev *pdev)
7605 {
7606 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7607 	struct net_device *netdev = adapter->netdev;
7608 
7609 #ifdef CONFIG_PCI_IOV
7610 	if (adapter->vferr_refcount) {
7611 		e_info(drv, "Resuming after VF err\n");
7612 		adapter->vferr_refcount--;
7613 		return;
7614 	}
7615 
7616 #endif
7617 	if (netif_running(netdev))
7618 		ixgbe_up(adapter);
7619 
7620 	netif_device_attach(netdev);
7621 }
7622 
7623 static struct pci_error_handlers ixgbe_err_handler = {
7624 	.error_detected = ixgbe_io_error_detected,
7625 	.slot_reset = ixgbe_io_slot_reset,
7626 	.resume = ixgbe_io_resume,
7627 };
7628 
7629 static struct pci_driver ixgbe_driver = {
7630 	.name     = ixgbe_driver_name,
7631 	.id_table = ixgbe_pci_tbl,
7632 	.probe    = ixgbe_probe,
7633 	.remove   = __devexit_p(ixgbe_remove),
7634 #ifdef CONFIG_PM
7635 	.suspend  = ixgbe_suspend,
7636 	.resume   = ixgbe_resume,
7637 #endif
7638 	.shutdown = ixgbe_shutdown,
7639 	.err_handler = &ixgbe_err_handler
7640 };
7641 
7642 /**
7643  * ixgbe_init_module - Driver Registration Routine
7644  *
7645  * ixgbe_init_module is the first routine called when the driver is
7646  * loaded. All it does is register with the PCI subsystem.
7647  **/
7648 static int __init ixgbe_init_module(void)
7649 {
7650 	int ret;
7651 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
7652 	pr_info("%s\n", ixgbe_copyright);
7653 
7654 #ifdef CONFIG_IXGBE_DCA
7655 	dca_register_notify(&dca_notifier);
7656 #endif
7657 
7658 	ret = pci_register_driver(&ixgbe_driver);
7659 	return ret;
7660 }
7661 
7662 module_init(ixgbe_init_module);
7663 
7664 /**
7665  * ixgbe_exit_module - Driver Exit Cleanup Routine
7666  *
7667  * ixgbe_exit_module is called just before the driver is removed
7668  * from memory.
7669  **/
7670 static void __exit ixgbe_exit_module(void)
7671 {
7672 #ifdef CONFIG_IXGBE_DCA
7673 	dca_unregister_notify(&dca_notifier);
7674 #endif
7675 	pci_unregister_driver(&ixgbe_driver);
7676 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
7677 }
7678 
7679 #ifdef CONFIG_IXGBE_DCA
7680 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
7681 			    void *p)
7682 {
7683 	int ret_val;
7684 
7685 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
7686 					 __ixgbe_notify_dca);
7687 
7688 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
7689 }
7690 
7691 #endif /* CONFIG_IXGBE_DCA */
7692 
7693 module_exit(ixgbe_exit_module);
7694 
7695 /* ixgbe_main.c */
7696