xref: /linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision d0b73b488c55df905ea8faaad079f8535629ed26)
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/if_bridge.h>
48 #include <linux/prefetch.h>
49 #include <scsi/fc/fc_fcoe.h>
50 
51 #include "ixgbe.h"
52 #include "ixgbe_common.h"
53 #include "ixgbe_dcb_82599.h"
54 #include "ixgbe_sriov.h"
55 
56 char ixgbe_driver_name[] = "ixgbe";
57 static const char ixgbe_driver_string[] =
58 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
59 #ifdef IXGBE_FCOE
60 char ixgbe_default_device_descr[] =
61 			      "Intel(R) 10 Gigabit Network Connection";
62 #else
63 static char ixgbe_default_device_descr[] =
64 			      "Intel(R) 10 Gigabit Network Connection";
65 #endif
66 #define DRV_VERSION "3.11.33-k"
67 const char ixgbe_driver_version[] = DRV_VERSION;
68 static const char ixgbe_copyright[] =
69 				"Copyright (c) 1999-2012 Intel Corporation.";
70 
71 static const struct ixgbe_info *ixgbe_info_tbl[] = {
72 	[board_82598] = &ixgbe_82598_info,
73 	[board_82599] = &ixgbe_82599_info,
74 	[board_X540] = &ixgbe_X540_info,
75 };
76 
77 /* ixgbe_pci_tbl - PCI Device ID Table
78  *
79  * Wildcard entries (PCI_ANY_ID) should come last
80  * Last entry must be all 0s
81  *
82  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
83  *   Class, Class Mask, private data (not used) }
84  */
85 static DEFINE_PCI_DEVICE_TABLE(ixgbe_pci_tbl) = {
86 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
87 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
88 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
89 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
90 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
91 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
92 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
93 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
94 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
95 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
96 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
97 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
98 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
99 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
100 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
101 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
102 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
103 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
104 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
105 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
115 	/* required last entry */
116 	{0, }
117 };
118 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
119 
120 #ifdef CONFIG_IXGBE_DCA
121 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
122 			    void *p);
123 static struct notifier_block dca_notifier = {
124 	.notifier_call = ixgbe_notify_dca,
125 	.next          = NULL,
126 	.priority      = 0
127 };
128 #endif
129 
130 #ifdef CONFIG_PCI_IOV
131 static unsigned int max_vfs;
132 module_param(max_vfs, uint, 0);
133 MODULE_PARM_DESC(max_vfs,
134 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63");
135 #endif /* CONFIG_PCI_IOV */
136 
137 static unsigned int allow_unsupported_sfp;
138 module_param(allow_unsupported_sfp, uint, 0);
139 MODULE_PARM_DESC(allow_unsupported_sfp,
140 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
141 
142 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
143 static int debug = -1;
144 module_param(debug, int, 0);
145 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
146 
147 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
148 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
149 MODULE_LICENSE("GPL");
150 MODULE_VERSION(DRV_VERSION);
151 
152 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
153 {
154 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
155 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
156 		schedule_work(&adapter->service_task);
157 }
158 
159 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
160 {
161 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
162 
163 	/* flush memory to make sure state is correct before next watchdog */
164 	smp_mb__before_clear_bit();
165 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
166 }
167 
168 struct ixgbe_reg_info {
169 	u32 ofs;
170 	char *name;
171 };
172 
173 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
174 
175 	/* General Registers */
176 	{IXGBE_CTRL, "CTRL"},
177 	{IXGBE_STATUS, "STATUS"},
178 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
179 
180 	/* Interrupt Registers */
181 	{IXGBE_EICR, "EICR"},
182 
183 	/* RX Registers */
184 	{IXGBE_SRRCTL(0), "SRRCTL"},
185 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
186 	{IXGBE_RDLEN(0), "RDLEN"},
187 	{IXGBE_RDH(0), "RDH"},
188 	{IXGBE_RDT(0), "RDT"},
189 	{IXGBE_RXDCTL(0), "RXDCTL"},
190 	{IXGBE_RDBAL(0), "RDBAL"},
191 	{IXGBE_RDBAH(0), "RDBAH"},
192 
193 	/* TX Registers */
194 	{IXGBE_TDBAL(0), "TDBAL"},
195 	{IXGBE_TDBAH(0), "TDBAH"},
196 	{IXGBE_TDLEN(0), "TDLEN"},
197 	{IXGBE_TDH(0), "TDH"},
198 	{IXGBE_TDT(0), "TDT"},
199 	{IXGBE_TXDCTL(0), "TXDCTL"},
200 
201 	/* List Terminator */
202 	{}
203 };
204 
205 
206 /*
207  * ixgbe_regdump - register printout routine
208  */
209 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
210 {
211 	int i = 0, j = 0;
212 	char rname[16];
213 	u32 regs[64];
214 
215 	switch (reginfo->ofs) {
216 	case IXGBE_SRRCTL(0):
217 		for (i = 0; i < 64; i++)
218 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
219 		break;
220 	case IXGBE_DCA_RXCTRL(0):
221 		for (i = 0; i < 64; i++)
222 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
223 		break;
224 	case IXGBE_RDLEN(0):
225 		for (i = 0; i < 64; i++)
226 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
227 		break;
228 	case IXGBE_RDH(0):
229 		for (i = 0; i < 64; i++)
230 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
231 		break;
232 	case IXGBE_RDT(0):
233 		for (i = 0; i < 64; i++)
234 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
235 		break;
236 	case IXGBE_RXDCTL(0):
237 		for (i = 0; i < 64; i++)
238 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
239 		break;
240 	case IXGBE_RDBAL(0):
241 		for (i = 0; i < 64; i++)
242 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
243 		break;
244 	case IXGBE_RDBAH(0):
245 		for (i = 0; i < 64; i++)
246 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
247 		break;
248 	case IXGBE_TDBAL(0):
249 		for (i = 0; i < 64; i++)
250 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
251 		break;
252 	case IXGBE_TDBAH(0):
253 		for (i = 0; i < 64; i++)
254 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
255 		break;
256 	case IXGBE_TDLEN(0):
257 		for (i = 0; i < 64; i++)
258 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
259 		break;
260 	case IXGBE_TDH(0):
261 		for (i = 0; i < 64; i++)
262 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
263 		break;
264 	case IXGBE_TDT(0):
265 		for (i = 0; i < 64; i++)
266 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
267 		break;
268 	case IXGBE_TXDCTL(0):
269 		for (i = 0; i < 64; i++)
270 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
271 		break;
272 	default:
273 		pr_info("%-15s %08x\n", reginfo->name,
274 			IXGBE_READ_REG(hw, reginfo->ofs));
275 		return;
276 	}
277 
278 	for (i = 0; i < 8; i++) {
279 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i*8, i*8+7);
280 		pr_err("%-15s", rname);
281 		for (j = 0; j < 8; j++)
282 			pr_cont(" %08x", regs[i*8+j]);
283 		pr_cont("\n");
284 	}
285 
286 }
287 
288 /*
289  * ixgbe_dump - Print registers, tx-rings and rx-rings
290  */
291 static void ixgbe_dump(struct ixgbe_adapter *adapter)
292 {
293 	struct net_device *netdev = adapter->netdev;
294 	struct ixgbe_hw *hw = &adapter->hw;
295 	struct ixgbe_reg_info *reginfo;
296 	int n = 0;
297 	struct ixgbe_ring *tx_ring;
298 	struct ixgbe_tx_buffer *tx_buffer;
299 	union ixgbe_adv_tx_desc *tx_desc;
300 	struct my_u0 { u64 a; u64 b; } *u0;
301 	struct ixgbe_ring *rx_ring;
302 	union ixgbe_adv_rx_desc *rx_desc;
303 	struct ixgbe_rx_buffer *rx_buffer_info;
304 	u32 staterr;
305 	int i = 0;
306 
307 	if (!netif_msg_hw(adapter))
308 		return;
309 
310 	/* Print netdevice Info */
311 	if (netdev) {
312 		dev_info(&adapter->pdev->dev, "Net device Info\n");
313 		pr_info("Device Name     state            "
314 			"trans_start      last_rx\n");
315 		pr_info("%-15s %016lX %016lX %016lX\n",
316 			netdev->name,
317 			netdev->state,
318 			netdev->trans_start,
319 			netdev->last_rx);
320 	}
321 
322 	/* Print Registers */
323 	dev_info(&adapter->pdev->dev, "Register Dump\n");
324 	pr_info(" Register Name   Value\n");
325 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
326 	     reginfo->name; reginfo++) {
327 		ixgbe_regdump(hw, reginfo);
328 	}
329 
330 	/* Print TX Ring Summary */
331 	if (!netdev || !netif_running(netdev))
332 		goto exit;
333 
334 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
335 	pr_info(" %s     %s              %s        %s\n",
336 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
337 		"leng", "ntw", "timestamp");
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 %08X %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 	 * 82598 Advanced Transmit Descriptor
358 	 *   +--------------------------------------------------------------+
359 	 * 0 |         Buffer Address [63:0]                                |
360 	 *   +--------------------------------------------------------------+
361 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
362 	 *   +--------------------------------------------------------------+
363 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
364 	 *
365 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
366 	 *   +--------------------------------------------------------------+
367 	 * 0 |                          RSV [63:0]                          |
368 	 *   +--------------------------------------------------------------+
369 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
370 	 *   +--------------------------------------------------------------+
371 	 *   63                       36 35   32 31                         0
372 	 *
373 	 * 82599+ Advanced Transmit Descriptor
374 	 *   +--------------------------------------------------------------+
375 	 * 0 |         Buffer Address [63:0]                                |
376 	 *   +--------------------------------------------------------------+
377 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
378 	 *   +--------------------------------------------------------------+
379 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
380 	 *
381 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
382 	 *   +--------------------------------------------------------------+
383 	 * 0 |                          RSV [63:0]                          |
384 	 *   +--------------------------------------------------------------+
385 	 * 8 |            RSV           |  STA  |           RSV             |
386 	 *   +--------------------------------------------------------------+
387 	 *   63                       36 35   32 31                         0
388 	 */
389 
390 	for (n = 0; n < adapter->num_tx_queues; n++) {
391 		tx_ring = adapter->tx_ring[n];
392 		pr_info("------------------------------------\n");
393 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
394 		pr_info("------------------------------------\n");
395 		pr_info("%s%s    %s              %s        %s          %s\n",
396 			"T [desc]     [address 63:0  ] ",
397 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
398 			"leng", "ntw", "timestamp", "bi->skb");
399 
400 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
401 			tx_desc = IXGBE_TX_DESC(tx_ring, i);
402 			tx_buffer = &tx_ring->tx_buffer_info[i];
403 			u0 = (struct my_u0 *)tx_desc;
404 			if (dma_unmap_len(tx_buffer, len) > 0) {
405 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p",
406 					i,
407 					le64_to_cpu(u0->a),
408 					le64_to_cpu(u0->b),
409 					(u64)dma_unmap_addr(tx_buffer, dma),
410 					dma_unmap_len(tx_buffer, len),
411 					tx_buffer->next_to_watch,
412 					(u64)tx_buffer->time_stamp,
413 					tx_buffer->skb);
414 				if (i == tx_ring->next_to_use &&
415 					i == tx_ring->next_to_clean)
416 					pr_cont(" NTC/U\n");
417 				else if (i == tx_ring->next_to_use)
418 					pr_cont(" NTU\n");
419 				else if (i == tx_ring->next_to_clean)
420 					pr_cont(" NTC\n");
421 				else
422 					pr_cont("\n");
423 
424 				if (netif_msg_pktdata(adapter) &&
425 				    tx_buffer->skb)
426 					print_hex_dump(KERN_INFO, "",
427 						DUMP_PREFIX_ADDRESS, 16, 1,
428 						tx_buffer->skb->data,
429 						dma_unmap_len(tx_buffer, len),
430 						true);
431 			}
432 		}
433 	}
434 
435 	/* Print RX Rings Summary */
436 rx_ring_summary:
437 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
438 	pr_info("Queue [NTU] [NTC]\n");
439 	for (n = 0; n < adapter->num_rx_queues; n++) {
440 		rx_ring = adapter->rx_ring[n];
441 		pr_info("%5d %5X %5X\n",
442 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
443 	}
444 
445 	/* Print RX Rings */
446 	if (!netif_msg_rx_status(adapter))
447 		goto exit;
448 
449 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
450 
451 	/* Receive Descriptor Formats
452 	 *
453 	 * 82598 Advanced Receive Descriptor (Read) Format
454 	 *    63                                           1        0
455 	 *    +-----------------------------------------------------+
456 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
457 	 *    +----------------------------------------------+------+
458 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
459 	 *    +-----------------------------------------------------+
460 	 *
461 	 *
462 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
463 	 *
464 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
465 	 *   +------------------------------------------------------+
466 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
467 	 *   | Packet   | IP     |   |          |     | Type | Type |
468 	 *   | Checksum | Ident  |   |          |     |      |      |
469 	 *   +------------------------------------------------------+
470 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
471 	 *   +------------------------------------------------------+
472 	 *   63       48 47    32 31            20 19               0
473 	 *
474 	 * 82599+ Advanced Receive Descriptor (Read) Format
475 	 *    63                                           1        0
476 	 *    +-----------------------------------------------------+
477 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
478 	 *    +----------------------------------------------+------+
479 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
480 	 *    +-----------------------------------------------------+
481 	 *
482 	 *
483 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
484 	 *
485 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
486 	 *   +------------------------------------------------------+
487 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
488 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
489 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
490 	 *   +------------------------------------------------------+
491 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
492 	 *   +------------------------------------------------------+
493 	 *   63       48 47    32 31          20 19                 0
494 	 */
495 
496 	for (n = 0; n < adapter->num_rx_queues; n++) {
497 		rx_ring = adapter->rx_ring[n];
498 		pr_info("------------------------------------\n");
499 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
500 		pr_info("------------------------------------\n");
501 		pr_info("%s%s%s",
502 			"R  [desc]      [ PktBuf     A0] ",
503 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
504 			"<-- Adv Rx Read format\n");
505 		pr_info("%s%s%s",
506 			"RWB[desc]      [PcsmIpSHl PtRs] ",
507 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
508 			"<-- Adv Rx Write-Back format\n");
509 
510 		for (i = 0; i < rx_ring->count; i++) {
511 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
512 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
513 			u0 = (struct my_u0 *)rx_desc;
514 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
515 			if (staterr & IXGBE_RXD_STAT_DD) {
516 				/* Descriptor Done */
517 				pr_info("RWB[0x%03X]     %016llX "
518 					"%016llX ---------------- %p", i,
519 					le64_to_cpu(u0->a),
520 					le64_to_cpu(u0->b),
521 					rx_buffer_info->skb);
522 			} else {
523 				pr_info("R  [0x%03X]     %016llX "
524 					"%016llX %016llX %p", i,
525 					le64_to_cpu(u0->a),
526 					le64_to_cpu(u0->b),
527 					(u64)rx_buffer_info->dma,
528 					rx_buffer_info->skb);
529 
530 				if (netif_msg_pktdata(adapter) &&
531 				    rx_buffer_info->dma) {
532 					print_hex_dump(KERN_INFO, "",
533 					   DUMP_PREFIX_ADDRESS, 16, 1,
534 					   page_address(rx_buffer_info->page) +
535 						    rx_buffer_info->page_offset,
536 					   ixgbe_rx_bufsz(rx_ring), true);
537 				}
538 			}
539 
540 			if (i == rx_ring->next_to_use)
541 				pr_cont(" NTU\n");
542 			else if (i == rx_ring->next_to_clean)
543 				pr_cont(" NTC\n");
544 			else
545 				pr_cont("\n");
546 
547 		}
548 	}
549 
550 exit:
551 	return;
552 }
553 
554 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
555 {
556 	u32 ctrl_ext;
557 
558 	/* Let firmware take over control of h/w */
559 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
560 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
561 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
562 }
563 
564 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
565 {
566 	u32 ctrl_ext;
567 
568 	/* Let firmware know the driver has taken over */
569 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
570 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
571 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
572 }
573 
574 /**
575  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
576  * @adapter: pointer to adapter struct
577  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
578  * @queue: queue to map the corresponding interrupt to
579  * @msix_vector: the vector to map to the corresponding queue
580  *
581  */
582 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
583 			   u8 queue, u8 msix_vector)
584 {
585 	u32 ivar, index;
586 	struct ixgbe_hw *hw = &adapter->hw;
587 	switch (hw->mac.type) {
588 	case ixgbe_mac_82598EB:
589 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
590 		if (direction == -1)
591 			direction = 0;
592 		index = (((direction * 64) + queue) >> 2) & 0x1F;
593 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
594 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
595 		ivar |= (msix_vector << (8 * (queue & 0x3)));
596 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
597 		break;
598 	case ixgbe_mac_82599EB:
599 	case ixgbe_mac_X540:
600 		if (direction == -1) {
601 			/* other causes */
602 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
603 			index = ((queue & 1) * 8);
604 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
605 			ivar &= ~(0xFF << index);
606 			ivar |= (msix_vector << index);
607 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
608 			break;
609 		} else {
610 			/* tx or rx causes */
611 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
612 			index = ((16 * (queue & 1)) + (8 * direction));
613 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
614 			ivar &= ~(0xFF << index);
615 			ivar |= (msix_vector << index);
616 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
617 			break;
618 		}
619 	default:
620 		break;
621 	}
622 }
623 
624 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
625 					  u64 qmask)
626 {
627 	u32 mask;
628 
629 	switch (adapter->hw.mac.type) {
630 	case ixgbe_mac_82598EB:
631 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
632 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
633 		break;
634 	case ixgbe_mac_82599EB:
635 	case ixgbe_mac_X540:
636 		mask = (qmask & 0xFFFFFFFF);
637 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
638 		mask = (qmask >> 32);
639 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
640 		break;
641 	default:
642 		break;
643 	}
644 }
645 
646 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring *ring,
647 				      struct ixgbe_tx_buffer *tx_buffer)
648 {
649 	if (tx_buffer->skb) {
650 		dev_kfree_skb_any(tx_buffer->skb);
651 		if (dma_unmap_len(tx_buffer, len))
652 			dma_unmap_single(ring->dev,
653 					 dma_unmap_addr(tx_buffer, dma),
654 					 dma_unmap_len(tx_buffer, len),
655 					 DMA_TO_DEVICE);
656 	} else if (dma_unmap_len(tx_buffer, len)) {
657 		dma_unmap_page(ring->dev,
658 			       dma_unmap_addr(tx_buffer, dma),
659 			       dma_unmap_len(tx_buffer, len),
660 			       DMA_TO_DEVICE);
661 	}
662 	tx_buffer->next_to_watch = NULL;
663 	tx_buffer->skb = NULL;
664 	dma_unmap_len_set(tx_buffer, len, 0);
665 	/* tx_buffer must be completely set up in the transmit path */
666 }
667 
668 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
669 {
670 	struct ixgbe_hw *hw = &adapter->hw;
671 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
672 	int i;
673 	u32 data;
674 
675 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
676 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
677 		return;
678 
679 	switch (hw->mac.type) {
680 	case ixgbe_mac_82598EB:
681 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
682 		break;
683 	default:
684 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
685 	}
686 	hwstats->lxoffrxc += data;
687 
688 	/* refill credits (no tx hang) if we received xoff */
689 	if (!data)
690 		return;
691 
692 	for (i = 0; i < adapter->num_tx_queues; i++)
693 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
694 			  &adapter->tx_ring[i]->state);
695 }
696 
697 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
698 {
699 	struct ixgbe_hw *hw = &adapter->hw;
700 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
701 	u32 xoff[8] = {0};
702 	u8 tc;
703 	int i;
704 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
705 
706 	if (adapter->ixgbe_ieee_pfc)
707 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
708 
709 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
710 		ixgbe_update_xoff_rx_lfc(adapter);
711 		return;
712 	}
713 
714 	/* update stats for each tc, only valid with PFC enabled */
715 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
716 		u32 pxoffrxc;
717 
718 		switch (hw->mac.type) {
719 		case ixgbe_mac_82598EB:
720 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
721 			break;
722 		default:
723 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
724 		}
725 		hwstats->pxoffrxc[i] += pxoffrxc;
726 		/* Get the TC for given UP */
727 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
728 		xoff[tc] += pxoffrxc;
729 	}
730 
731 	/* disarm tx queues that have received xoff frames */
732 	for (i = 0; i < adapter->num_tx_queues; i++) {
733 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
734 
735 		tc = tx_ring->dcb_tc;
736 		if (xoff[tc])
737 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
738 	}
739 }
740 
741 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
742 {
743 	return ring->stats.packets;
744 }
745 
746 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
747 {
748 	struct ixgbe_adapter *adapter = netdev_priv(ring->netdev);
749 	struct ixgbe_hw *hw = &adapter->hw;
750 
751 	u32 head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
752 	u32 tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
753 
754 	if (head != tail)
755 		return (head < tail) ?
756 			tail - head : (tail + ring->count - head);
757 
758 	return 0;
759 }
760 
761 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
762 {
763 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
764 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
765 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
766 	bool ret = false;
767 
768 	clear_check_for_tx_hang(tx_ring);
769 
770 	/*
771 	 * Check for a hung queue, but be thorough. This verifies
772 	 * that a transmit has been completed since the previous
773 	 * check AND there is at least one packet pending. The
774 	 * ARMED bit is set to indicate a potential hang. The
775 	 * bit is cleared if a pause frame is received to remove
776 	 * false hang detection due to PFC or 802.3x frames. By
777 	 * requiring this to fail twice we avoid races with
778 	 * pfc clearing the ARMED bit and conditions where we
779 	 * run the check_tx_hang logic with a transmit completion
780 	 * pending but without time to complete it yet.
781 	 */
782 	if ((tx_done_old == tx_done) && tx_pending) {
783 		/* make sure it is true for two checks in a row */
784 		ret = test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
785 				       &tx_ring->state);
786 	} else {
787 		/* update completed stats and continue */
788 		tx_ring->tx_stats.tx_done_old = tx_done;
789 		/* reset the countdown */
790 		clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
791 	}
792 
793 	return ret;
794 }
795 
796 /**
797  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
798  * @adapter: driver private struct
799  **/
800 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
801 {
802 
803 	/* Do the reset outside of interrupt context */
804 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
805 		adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
806 		ixgbe_service_event_schedule(adapter);
807 	}
808 }
809 
810 /**
811  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
812  * @q_vector: structure containing interrupt and ring information
813  * @tx_ring: tx ring to clean
814  **/
815 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
816 			       struct ixgbe_ring *tx_ring)
817 {
818 	struct ixgbe_adapter *adapter = q_vector->adapter;
819 	struct ixgbe_tx_buffer *tx_buffer;
820 	union ixgbe_adv_tx_desc *tx_desc;
821 	unsigned int total_bytes = 0, total_packets = 0;
822 	unsigned int budget = q_vector->tx.work_limit;
823 	unsigned int i = tx_ring->next_to_clean;
824 
825 	if (test_bit(__IXGBE_DOWN, &adapter->state))
826 		return true;
827 
828 	tx_buffer = &tx_ring->tx_buffer_info[i];
829 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
830 	i -= tx_ring->count;
831 
832 	do {
833 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
834 
835 		/* if next_to_watch is not set then there is no work pending */
836 		if (!eop_desc)
837 			break;
838 
839 		/* prevent any other reads prior to eop_desc */
840 		rmb();
841 
842 		/* if DD is not set pending work has not been completed */
843 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
844 			break;
845 
846 		/* clear next_to_watch to prevent false hangs */
847 		tx_buffer->next_to_watch = NULL;
848 
849 		/* update the statistics for this packet */
850 		total_bytes += tx_buffer->bytecount;
851 		total_packets += tx_buffer->gso_segs;
852 
853 		if (unlikely(tx_buffer->tx_flags & IXGBE_TX_FLAGS_TSTAMP))
854 			ixgbe_ptp_tx_hwtstamp(q_vector, tx_buffer->skb);
855 
856 		/* free the skb */
857 		dev_kfree_skb_any(tx_buffer->skb);
858 
859 		/* unmap skb header data */
860 		dma_unmap_single(tx_ring->dev,
861 				 dma_unmap_addr(tx_buffer, dma),
862 				 dma_unmap_len(tx_buffer, len),
863 				 DMA_TO_DEVICE);
864 
865 		/* clear tx_buffer data */
866 		tx_buffer->skb = NULL;
867 		dma_unmap_len_set(tx_buffer, len, 0);
868 
869 		/* unmap remaining buffers */
870 		while (tx_desc != eop_desc) {
871 			tx_buffer++;
872 			tx_desc++;
873 			i++;
874 			if (unlikely(!i)) {
875 				i -= tx_ring->count;
876 				tx_buffer = tx_ring->tx_buffer_info;
877 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
878 			}
879 
880 			/* unmap any remaining paged data */
881 			if (dma_unmap_len(tx_buffer, len)) {
882 				dma_unmap_page(tx_ring->dev,
883 					       dma_unmap_addr(tx_buffer, dma),
884 					       dma_unmap_len(tx_buffer, len),
885 					       DMA_TO_DEVICE);
886 				dma_unmap_len_set(tx_buffer, len, 0);
887 			}
888 		}
889 
890 		/* move us one more past the eop_desc for start of next pkt */
891 		tx_buffer++;
892 		tx_desc++;
893 		i++;
894 		if (unlikely(!i)) {
895 			i -= tx_ring->count;
896 			tx_buffer = tx_ring->tx_buffer_info;
897 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
898 		}
899 
900 		/* issue prefetch for next Tx descriptor */
901 		prefetch(tx_desc);
902 
903 		/* update budget accounting */
904 		budget--;
905 	} while (likely(budget));
906 
907 	i += tx_ring->count;
908 	tx_ring->next_to_clean = i;
909 	u64_stats_update_begin(&tx_ring->syncp);
910 	tx_ring->stats.bytes += total_bytes;
911 	tx_ring->stats.packets += total_packets;
912 	u64_stats_update_end(&tx_ring->syncp);
913 	q_vector->tx.total_bytes += total_bytes;
914 	q_vector->tx.total_packets += total_packets;
915 
916 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
917 		/* schedule immediate reset if we believe we hung */
918 		struct ixgbe_hw *hw = &adapter->hw;
919 		e_err(drv, "Detected Tx Unit Hang\n"
920 			"  Tx Queue             <%d>\n"
921 			"  TDH, TDT             <%x>, <%x>\n"
922 			"  next_to_use          <%x>\n"
923 			"  next_to_clean        <%x>\n"
924 			"tx_buffer_info[next_to_clean]\n"
925 			"  time_stamp           <%lx>\n"
926 			"  jiffies              <%lx>\n",
927 			tx_ring->queue_index,
928 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
929 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
930 			tx_ring->next_to_use, i,
931 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
932 
933 		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
934 
935 		e_info(probe,
936 		       "tx hang %d detected on queue %d, resetting adapter\n",
937 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
938 
939 		/* schedule immediate reset if we believe we hung */
940 		ixgbe_tx_timeout_reset(adapter);
941 
942 		/* the adapter is about to reset, no point in enabling stuff */
943 		return true;
944 	}
945 
946 	netdev_tx_completed_queue(txring_txq(tx_ring),
947 				  total_packets, total_bytes);
948 
949 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
950 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
951 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
952 		/* Make sure that anybody stopping the queue after this
953 		 * sees the new next_to_clean.
954 		 */
955 		smp_mb();
956 		if (__netif_subqueue_stopped(tx_ring->netdev,
957 					     tx_ring->queue_index)
958 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
959 			netif_wake_subqueue(tx_ring->netdev,
960 					    tx_ring->queue_index);
961 			++tx_ring->tx_stats.restart_queue;
962 		}
963 	}
964 
965 	return !!budget;
966 }
967 
968 #ifdef CONFIG_IXGBE_DCA
969 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
970 				struct ixgbe_ring *tx_ring,
971 				int cpu)
972 {
973 	struct ixgbe_hw *hw = &adapter->hw;
974 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
975 	u16 reg_offset;
976 
977 	switch (hw->mac.type) {
978 	case ixgbe_mac_82598EB:
979 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
980 		break;
981 	case ixgbe_mac_82599EB:
982 	case ixgbe_mac_X540:
983 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
984 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
985 		break;
986 	default:
987 		/* for unknown hardware do not write register */
988 		return;
989 	}
990 
991 	/*
992 	 * We can enable relaxed ordering for reads, but not writes when
993 	 * DCA is enabled.  This is due to a known issue in some chipsets
994 	 * which will cause the DCA tag to be cleared.
995 	 */
996 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
997 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
998 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
999 
1000 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1001 }
1002 
1003 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1004 				struct ixgbe_ring *rx_ring,
1005 				int cpu)
1006 {
1007 	struct ixgbe_hw *hw = &adapter->hw;
1008 	u32 rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1009 	u8 reg_idx = rx_ring->reg_idx;
1010 
1011 
1012 	switch (hw->mac.type) {
1013 	case ixgbe_mac_82599EB:
1014 	case ixgbe_mac_X540:
1015 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1016 		break;
1017 	default:
1018 		break;
1019 	}
1020 
1021 	/*
1022 	 * We can enable relaxed ordering for reads, but not writes when
1023 	 * DCA is enabled.  This is due to a known issue in some chipsets
1024 	 * which will cause the DCA tag to be cleared.
1025 	 */
1026 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1027 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1028 
1029 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1030 }
1031 
1032 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1033 {
1034 	struct ixgbe_adapter *adapter = q_vector->adapter;
1035 	struct ixgbe_ring *ring;
1036 	int cpu = get_cpu();
1037 
1038 	if (q_vector->cpu == cpu)
1039 		goto out_no_update;
1040 
1041 	ixgbe_for_each_ring(ring, q_vector->tx)
1042 		ixgbe_update_tx_dca(adapter, ring, cpu);
1043 
1044 	ixgbe_for_each_ring(ring, q_vector->rx)
1045 		ixgbe_update_rx_dca(adapter, ring, cpu);
1046 
1047 	q_vector->cpu = cpu;
1048 out_no_update:
1049 	put_cpu();
1050 }
1051 
1052 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1053 {
1054 	int i;
1055 
1056 	if (!(adapter->flags & IXGBE_FLAG_DCA_ENABLED))
1057 		return;
1058 
1059 	/* always use CB2 mode, difference is masked in the CB driver */
1060 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 2);
1061 
1062 	for (i = 0; i < adapter->num_q_vectors; i++) {
1063 		adapter->q_vector[i]->cpu = -1;
1064 		ixgbe_update_dca(adapter->q_vector[i]);
1065 	}
1066 }
1067 
1068 static int __ixgbe_notify_dca(struct device *dev, void *data)
1069 {
1070 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1071 	unsigned long event = *(unsigned long *)data;
1072 
1073 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1074 		return 0;
1075 
1076 	switch (event) {
1077 	case DCA_PROVIDER_ADD:
1078 		/* if we're already enabled, don't do it again */
1079 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1080 			break;
1081 		if (dca_add_requester(dev) == 0) {
1082 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1083 			ixgbe_setup_dca(adapter);
1084 			break;
1085 		}
1086 		/* Fall Through since DCA is disabled. */
1087 	case DCA_PROVIDER_REMOVE:
1088 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1089 			dca_remove_requester(dev);
1090 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1091 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
1092 		}
1093 		break;
1094 	}
1095 
1096 	return 0;
1097 }
1098 
1099 #endif /* CONFIG_IXGBE_DCA */
1100 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1101 				 union ixgbe_adv_rx_desc *rx_desc,
1102 				 struct sk_buff *skb)
1103 {
1104 	if (ring->netdev->features & NETIF_F_RXHASH)
1105 		skb->rxhash = le32_to_cpu(rx_desc->wb.lower.hi_dword.rss);
1106 }
1107 
1108 #ifdef IXGBE_FCOE
1109 /**
1110  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1111  * @ring: structure containing ring specific data
1112  * @rx_desc: advanced rx descriptor
1113  *
1114  * Returns : true if it is FCoE pkt
1115  */
1116 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1117 				    union ixgbe_adv_rx_desc *rx_desc)
1118 {
1119 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1120 
1121 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1122 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1123 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1124 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1125 }
1126 
1127 #endif /* IXGBE_FCOE */
1128 /**
1129  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1130  * @ring: structure containing ring specific data
1131  * @rx_desc: current Rx descriptor being processed
1132  * @skb: skb currently being received and modified
1133  **/
1134 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1135 				     union ixgbe_adv_rx_desc *rx_desc,
1136 				     struct sk_buff *skb)
1137 {
1138 	skb_checksum_none_assert(skb);
1139 
1140 	/* Rx csum disabled */
1141 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1142 		return;
1143 
1144 	/* if IP and error */
1145 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1146 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1147 		ring->rx_stats.csum_err++;
1148 		return;
1149 	}
1150 
1151 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1152 		return;
1153 
1154 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1155 		__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1156 
1157 		/*
1158 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1159 		 * checksum errors.
1160 		 */
1161 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1162 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1163 			return;
1164 
1165 		ring->rx_stats.csum_err++;
1166 		return;
1167 	}
1168 
1169 	/* It must be a TCP or UDP packet with a valid checksum */
1170 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1171 }
1172 
1173 static inline void ixgbe_release_rx_desc(struct ixgbe_ring *rx_ring, u32 val)
1174 {
1175 	rx_ring->next_to_use = val;
1176 
1177 	/* update next to alloc since we have filled the ring */
1178 	rx_ring->next_to_alloc = val;
1179 	/*
1180 	 * Force memory writes to complete before letting h/w
1181 	 * know there are new descriptors to fetch.  (Only
1182 	 * applicable for weak-ordered memory model archs,
1183 	 * such as IA-64).
1184 	 */
1185 	wmb();
1186 	writel(val, rx_ring->tail);
1187 }
1188 
1189 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1190 				    struct ixgbe_rx_buffer *bi)
1191 {
1192 	struct page *page = bi->page;
1193 	dma_addr_t dma = bi->dma;
1194 
1195 	/* since we are recycling buffers we should seldom need to alloc */
1196 	if (likely(dma))
1197 		return true;
1198 
1199 	/* alloc new page for storage */
1200 	if (likely(!page)) {
1201 		page = __skb_alloc_pages(GFP_ATOMIC | __GFP_COLD | __GFP_COMP,
1202 					 bi->skb, ixgbe_rx_pg_order(rx_ring));
1203 		if (unlikely(!page)) {
1204 			rx_ring->rx_stats.alloc_rx_page_failed++;
1205 			return false;
1206 		}
1207 		bi->page = page;
1208 	}
1209 
1210 	/* map page for use */
1211 	dma = dma_map_page(rx_ring->dev, page, 0,
1212 			   ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1213 
1214 	/*
1215 	 * if mapping failed free memory back to system since
1216 	 * there isn't much point in holding memory we can't use
1217 	 */
1218 	if (dma_mapping_error(rx_ring->dev, dma)) {
1219 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1220 		bi->page = NULL;
1221 
1222 		rx_ring->rx_stats.alloc_rx_page_failed++;
1223 		return false;
1224 	}
1225 
1226 	bi->dma = dma;
1227 	bi->page_offset = 0;
1228 
1229 	return true;
1230 }
1231 
1232 /**
1233  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1234  * @rx_ring: ring to place buffers on
1235  * @cleaned_count: number of buffers to replace
1236  **/
1237 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1238 {
1239 	union ixgbe_adv_rx_desc *rx_desc;
1240 	struct ixgbe_rx_buffer *bi;
1241 	u16 i = rx_ring->next_to_use;
1242 
1243 	/* nothing to do */
1244 	if (!cleaned_count)
1245 		return;
1246 
1247 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1248 	bi = &rx_ring->rx_buffer_info[i];
1249 	i -= rx_ring->count;
1250 
1251 	do {
1252 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1253 			break;
1254 
1255 		/*
1256 		 * Refresh the desc even if buffer_addrs didn't change
1257 		 * because each write-back erases this info.
1258 		 */
1259 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1260 
1261 		rx_desc++;
1262 		bi++;
1263 		i++;
1264 		if (unlikely(!i)) {
1265 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1266 			bi = rx_ring->rx_buffer_info;
1267 			i -= rx_ring->count;
1268 		}
1269 
1270 		/* clear the hdr_addr for the next_to_use descriptor */
1271 		rx_desc->read.hdr_addr = 0;
1272 
1273 		cleaned_count--;
1274 	} while (cleaned_count);
1275 
1276 	i += rx_ring->count;
1277 
1278 	if (rx_ring->next_to_use != i)
1279 		ixgbe_release_rx_desc(rx_ring, i);
1280 }
1281 
1282 /**
1283  * ixgbe_get_headlen - determine size of header for RSC/LRO/GRO/FCOE
1284  * @data: pointer to the start of the headers
1285  * @max_len: total length of section to find headers in
1286  *
1287  * This function is meant to determine the length of headers that will
1288  * be recognized by hardware for LRO, GRO, and RSC offloads.  The main
1289  * motivation of doing this is to only perform one pull for IPv4 TCP
1290  * packets so that we can do basic things like calculating the gso_size
1291  * based on the average data per packet.
1292  **/
1293 static unsigned int ixgbe_get_headlen(unsigned char *data,
1294 				      unsigned int max_len)
1295 {
1296 	union {
1297 		unsigned char *network;
1298 		/* l2 headers */
1299 		struct ethhdr *eth;
1300 		struct vlan_hdr *vlan;
1301 		/* l3 headers */
1302 		struct iphdr *ipv4;
1303 		struct ipv6hdr *ipv6;
1304 	} hdr;
1305 	__be16 protocol;
1306 	u8 nexthdr = 0;	/* default to not TCP */
1307 	u8 hlen;
1308 
1309 	/* this should never happen, but better safe than sorry */
1310 	if (max_len < ETH_HLEN)
1311 		return max_len;
1312 
1313 	/* initialize network frame pointer */
1314 	hdr.network = data;
1315 
1316 	/* set first protocol and move network header forward */
1317 	protocol = hdr.eth->h_proto;
1318 	hdr.network += ETH_HLEN;
1319 
1320 	/* handle any vlan tag if present */
1321 	if (protocol == __constant_htons(ETH_P_8021Q)) {
1322 		if ((hdr.network - data) > (max_len - VLAN_HLEN))
1323 			return max_len;
1324 
1325 		protocol = hdr.vlan->h_vlan_encapsulated_proto;
1326 		hdr.network += VLAN_HLEN;
1327 	}
1328 
1329 	/* handle L3 protocols */
1330 	if (protocol == __constant_htons(ETH_P_IP)) {
1331 		if ((hdr.network - data) > (max_len - sizeof(struct iphdr)))
1332 			return max_len;
1333 
1334 		/* access ihl as a u8 to avoid unaligned access on ia64 */
1335 		hlen = (hdr.network[0] & 0x0F) << 2;
1336 
1337 		/* verify hlen meets minimum size requirements */
1338 		if (hlen < sizeof(struct iphdr))
1339 			return hdr.network - data;
1340 
1341 		/* record next protocol if header is present */
1342 		if (!hdr.ipv4->frag_off)
1343 			nexthdr = hdr.ipv4->protocol;
1344 	} else if (protocol == __constant_htons(ETH_P_IPV6)) {
1345 		if ((hdr.network - data) > (max_len - sizeof(struct ipv6hdr)))
1346 			return max_len;
1347 
1348 		/* record next protocol */
1349 		nexthdr = hdr.ipv6->nexthdr;
1350 		hlen = sizeof(struct ipv6hdr);
1351 #ifdef IXGBE_FCOE
1352 	} else if (protocol == __constant_htons(ETH_P_FCOE)) {
1353 		if ((hdr.network - data) > (max_len - FCOE_HEADER_LEN))
1354 			return max_len;
1355 		hlen = FCOE_HEADER_LEN;
1356 #endif
1357 	} else {
1358 		return hdr.network - data;
1359 	}
1360 
1361 	/* relocate pointer to start of L4 header */
1362 	hdr.network += hlen;
1363 
1364 	/* finally sort out TCP/UDP */
1365 	if (nexthdr == IPPROTO_TCP) {
1366 		if ((hdr.network - data) > (max_len - sizeof(struct tcphdr)))
1367 			return max_len;
1368 
1369 		/* access doff as a u8 to avoid unaligned access on ia64 */
1370 		hlen = (hdr.network[12] & 0xF0) >> 2;
1371 
1372 		/* verify hlen meets minimum size requirements */
1373 		if (hlen < sizeof(struct tcphdr))
1374 			return hdr.network - data;
1375 
1376 		hdr.network += hlen;
1377 	} else if (nexthdr == IPPROTO_UDP) {
1378 		if ((hdr.network - data) > (max_len - sizeof(struct udphdr)))
1379 			return max_len;
1380 
1381 		hdr.network += sizeof(struct udphdr);
1382 	}
1383 
1384 	/*
1385 	 * If everything has gone correctly hdr.network should be the
1386 	 * data section of the packet and will be the end of the header.
1387 	 * If not then it probably represents the end of the last recognized
1388 	 * header.
1389 	 */
1390 	if ((hdr.network - data) < max_len)
1391 		return hdr.network - data;
1392 	else
1393 		return max_len;
1394 }
1395 
1396 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1397 				   struct sk_buff *skb)
1398 {
1399 	u16 hdr_len = skb_headlen(skb);
1400 
1401 	/* set gso_size to avoid messing up TCP MSS */
1402 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1403 						 IXGBE_CB(skb)->append_cnt);
1404 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1405 }
1406 
1407 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1408 				   struct sk_buff *skb)
1409 {
1410 	/* if append_cnt is 0 then frame is not RSC */
1411 	if (!IXGBE_CB(skb)->append_cnt)
1412 		return;
1413 
1414 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1415 	rx_ring->rx_stats.rsc_flush++;
1416 
1417 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1418 
1419 	/* gso_size is computed using append_cnt so always clear it last */
1420 	IXGBE_CB(skb)->append_cnt = 0;
1421 }
1422 
1423 /**
1424  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1425  * @rx_ring: rx descriptor ring packet is being transacted on
1426  * @rx_desc: pointer to the EOP Rx descriptor
1427  * @skb: pointer to current skb being populated
1428  *
1429  * This function checks the ring, descriptor, and packet information in
1430  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1431  * other fields within the skb.
1432  **/
1433 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1434 				     union ixgbe_adv_rx_desc *rx_desc,
1435 				     struct sk_buff *skb)
1436 {
1437 	struct net_device *dev = rx_ring->netdev;
1438 
1439 	ixgbe_update_rsc_stats(rx_ring, skb);
1440 
1441 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1442 
1443 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1444 
1445 	ixgbe_ptp_rx_hwtstamp(rx_ring->q_vector, rx_desc, skb);
1446 
1447 	if ((dev->features & NETIF_F_HW_VLAN_RX) &&
1448 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1449 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1450 		__vlan_hwaccel_put_tag(skb, vid);
1451 	}
1452 
1453 	skb_record_rx_queue(skb, rx_ring->queue_index);
1454 
1455 	skb->protocol = eth_type_trans(skb, dev);
1456 }
1457 
1458 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1459 			 struct sk_buff *skb)
1460 {
1461 	struct ixgbe_adapter *adapter = q_vector->adapter;
1462 
1463 	if (!(adapter->flags & IXGBE_FLAG_IN_NETPOLL))
1464 		napi_gro_receive(&q_vector->napi, skb);
1465 	else
1466 		netif_rx(skb);
1467 }
1468 
1469 /**
1470  * ixgbe_is_non_eop - process handling of non-EOP buffers
1471  * @rx_ring: Rx ring being processed
1472  * @rx_desc: Rx descriptor for current buffer
1473  * @skb: Current socket buffer containing buffer in progress
1474  *
1475  * This function updates next to clean.  If the buffer is an EOP buffer
1476  * this function exits returning false, otherwise it will place the
1477  * sk_buff in the next buffer to be chained and return true indicating
1478  * that this is in fact a non-EOP buffer.
1479  **/
1480 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1481 			     union ixgbe_adv_rx_desc *rx_desc,
1482 			     struct sk_buff *skb)
1483 {
1484 	u32 ntc = rx_ring->next_to_clean + 1;
1485 
1486 	/* fetch, update, and store next to clean */
1487 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1488 	rx_ring->next_to_clean = ntc;
1489 
1490 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1491 
1492 	/* update RSC append count if present */
1493 	if (ring_is_rsc_enabled(rx_ring)) {
1494 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1495 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1496 
1497 		if (unlikely(rsc_enabled)) {
1498 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1499 
1500 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1501 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1502 
1503 			/* update ntc based on RSC value */
1504 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1505 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1506 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1507 		}
1508 	}
1509 
1510 	/* if we are the last buffer then there is nothing else to do */
1511 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1512 		return false;
1513 
1514 	/* place skb in next buffer to be received */
1515 	rx_ring->rx_buffer_info[ntc].skb = skb;
1516 	rx_ring->rx_stats.non_eop_descs++;
1517 
1518 	return true;
1519 }
1520 
1521 /**
1522  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1523  * @rx_ring: rx descriptor ring packet is being transacted on
1524  * @skb: pointer to current skb being adjusted
1525  *
1526  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1527  * main difference between this version and the original function is that
1528  * this function can make several assumptions about the state of things
1529  * that allow for significant optimizations versus the standard function.
1530  * As a result we can do things like drop a frag and maintain an accurate
1531  * truesize for the skb.
1532  */
1533 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1534 			    struct sk_buff *skb)
1535 {
1536 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1537 	unsigned char *va;
1538 	unsigned int pull_len;
1539 
1540 	/*
1541 	 * it is valid to use page_address instead of kmap since we are
1542 	 * working with pages allocated out of the lomem pool per
1543 	 * alloc_page(GFP_ATOMIC)
1544 	 */
1545 	va = skb_frag_address(frag);
1546 
1547 	/*
1548 	 * we need the header to contain the greater of either ETH_HLEN or
1549 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1550 	 */
1551 	pull_len = ixgbe_get_headlen(va, IXGBE_RX_HDR_SIZE);
1552 
1553 	/* align pull length to size of long to optimize memcpy performance */
1554 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1555 
1556 	/* update all of the pointers */
1557 	skb_frag_size_sub(frag, pull_len);
1558 	frag->page_offset += pull_len;
1559 	skb->data_len -= pull_len;
1560 	skb->tail += pull_len;
1561 }
1562 
1563 /**
1564  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1565  * @rx_ring: rx descriptor ring packet is being transacted on
1566  * @skb: pointer to current skb being updated
1567  *
1568  * This function provides a basic DMA sync up for the first fragment of an
1569  * skb.  The reason for doing this is that the first fragment cannot be
1570  * unmapped until we have reached the end of packet descriptor for a buffer
1571  * chain.
1572  */
1573 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1574 				struct sk_buff *skb)
1575 {
1576 	/* if the page was released unmap it, else just sync our portion */
1577 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1578 		dma_unmap_page(rx_ring->dev, IXGBE_CB(skb)->dma,
1579 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1580 		IXGBE_CB(skb)->page_released = false;
1581 	} else {
1582 		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1583 
1584 		dma_sync_single_range_for_cpu(rx_ring->dev,
1585 					      IXGBE_CB(skb)->dma,
1586 					      frag->page_offset,
1587 					      ixgbe_rx_bufsz(rx_ring),
1588 					      DMA_FROM_DEVICE);
1589 	}
1590 	IXGBE_CB(skb)->dma = 0;
1591 }
1592 
1593 /**
1594  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1595  * @rx_ring: rx descriptor ring packet is being transacted on
1596  * @rx_desc: pointer to the EOP Rx descriptor
1597  * @skb: pointer to current skb being fixed
1598  *
1599  * Check for corrupted packet headers caused by senders on the local L2
1600  * embedded NIC switch not setting up their Tx Descriptors right.  These
1601  * should be very rare.
1602  *
1603  * Also address the case where we are pulling data in on pages only
1604  * and as such no data is present in the skb header.
1605  *
1606  * In addition if skb is not at least 60 bytes we need to pad it so that
1607  * it is large enough to qualify as a valid Ethernet frame.
1608  *
1609  * Returns true if an error was encountered and skb was freed.
1610  **/
1611 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1612 				  union ixgbe_adv_rx_desc *rx_desc,
1613 				  struct sk_buff *skb)
1614 {
1615 	struct net_device *netdev = rx_ring->netdev;
1616 
1617 	/* verify that the packet does not have any known errors */
1618 	if (unlikely(ixgbe_test_staterr(rx_desc,
1619 					IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1620 	    !(netdev->features & NETIF_F_RXALL))) {
1621 		dev_kfree_skb_any(skb);
1622 		return true;
1623 	}
1624 
1625 	/* place header in linear portion of buffer */
1626 	if (skb_is_nonlinear(skb))
1627 		ixgbe_pull_tail(rx_ring, skb);
1628 
1629 #ifdef IXGBE_FCOE
1630 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1631 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1632 		return false;
1633 
1634 #endif
1635 	/* if skb_pad returns an error the skb was freed */
1636 	if (unlikely(skb->len < 60)) {
1637 		int pad_len = 60 - skb->len;
1638 
1639 		if (skb_pad(skb, pad_len))
1640 			return true;
1641 		__skb_put(skb, pad_len);
1642 	}
1643 
1644 	return false;
1645 }
1646 
1647 /**
1648  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1649  * @rx_ring: rx descriptor ring to store buffers on
1650  * @old_buff: donor buffer to have page reused
1651  *
1652  * Synchronizes page for reuse by the adapter
1653  **/
1654 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1655 				struct ixgbe_rx_buffer *old_buff)
1656 {
1657 	struct ixgbe_rx_buffer *new_buff;
1658 	u16 nta = rx_ring->next_to_alloc;
1659 
1660 	new_buff = &rx_ring->rx_buffer_info[nta];
1661 
1662 	/* update, and store next to alloc */
1663 	nta++;
1664 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1665 
1666 	/* transfer page from old buffer to new buffer */
1667 	new_buff->page = old_buff->page;
1668 	new_buff->dma = old_buff->dma;
1669 	new_buff->page_offset = old_buff->page_offset;
1670 
1671 	/* sync the buffer for use by the device */
1672 	dma_sync_single_range_for_device(rx_ring->dev, new_buff->dma,
1673 					 new_buff->page_offset,
1674 					 ixgbe_rx_bufsz(rx_ring),
1675 					 DMA_FROM_DEVICE);
1676 }
1677 
1678 /**
1679  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1680  * @rx_ring: rx descriptor ring to transact packets on
1681  * @rx_buffer: buffer containing page to add
1682  * @rx_desc: descriptor containing length of buffer written by hardware
1683  * @skb: sk_buff to place the data into
1684  *
1685  * This function will add the data contained in rx_buffer->page to the skb.
1686  * This is done either through a direct copy if the data in the buffer is
1687  * less than the skb header size, otherwise it will just attach the page as
1688  * a frag to the skb.
1689  *
1690  * The function will then update the page offset if necessary and return
1691  * true if the buffer can be reused by the adapter.
1692  **/
1693 static bool ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1694 			      struct ixgbe_rx_buffer *rx_buffer,
1695 			      union ixgbe_adv_rx_desc *rx_desc,
1696 			      struct sk_buff *skb)
1697 {
1698 	struct page *page = rx_buffer->page;
1699 	unsigned int size = le16_to_cpu(rx_desc->wb.upper.length);
1700 #if (PAGE_SIZE < 8192)
1701 	unsigned int truesize = ixgbe_rx_bufsz(rx_ring);
1702 #else
1703 	unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
1704 	unsigned int last_offset = ixgbe_rx_pg_size(rx_ring) -
1705 				   ixgbe_rx_bufsz(rx_ring);
1706 #endif
1707 
1708 	if ((size <= IXGBE_RX_HDR_SIZE) && !skb_is_nonlinear(skb)) {
1709 		unsigned char *va = page_address(page) + rx_buffer->page_offset;
1710 
1711 		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
1712 
1713 		/* we can reuse buffer as-is, just make sure it is local */
1714 		if (likely(page_to_nid(page) == numa_node_id()))
1715 			return true;
1716 
1717 		/* this page cannot be reused so discard it */
1718 		put_page(page);
1719 		return false;
1720 	}
1721 
1722 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
1723 			rx_buffer->page_offset, size, truesize);
1724 
1725 	/* avoid re-using remote pages */
1726 	if (unlikely(page_to_nid(page) != numa_node_id()))
1727 		return false;
1728 
1729 #if (PAGE_SIZE < 8192)
1730 	/* if we are only owner of page we can reuse it */
1731 	if (unlikely(page_count(page) != 1))
1732 		return false;
1733 
1734 	/* flip page offset to other buffer */
1735 	rx_buffer->page_offset ^= truesize;
1736 
1737 	/*
1738 	 * since we are the only owner of the page and we need to
1739 	 * increment it, just set the value to 2 in order to avoid
1740 	 * an unecessary locked operation
1741 	 */
1742 	atomic_set(&page->_count, 2);
1743 #else
1744 	/* move offset up to the next cache line */
1745 	rx_buffer->page_offset += truesize;
1746 
1747 	if (rx_buffer->page_offset > last_offset)
1748 		return false;
1749 
1750 	/* bump ref count on page before it is given to the stack */
1751 	get_page(page);
1752 #endif
1753 
1754 	return true;
1755 }
1756 
1757 static struct sk_buff *ixgbe_fetch_rx_buffer(struct ixgbe_ring *rx_ring,
1758 					     union ixgbe_adv_rx_desc *rx_desc)
1759 {
1760 	struct ixgbe_rx_buffer *rx_buffer;
1761 	struct sk_buff *skb;
1762 	struct page *page;
1763 
1764 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
1765 	page = rx_buffer->page;
1766 	prefetchw(page);
1767 
1768 	skb = rx_buffer->skb;
1769 
1770 	if (likely(!skb)) {
1771 		void *page_addr = page_address(page) +
1772 				  rx_buffer->page_offset;
1773 
1774 		/* prefetch first cache line of first page */
1775 		prefetch(page_addr);
1776 #if L1_CACHE_BYTES < 128
1777 		prefetch(page_addr + L1_CACHE_BYTES);
1778 #endif
1779 
1780 		/* allocate a skb to store the frags */
1781 		skb = netdev_alloc_skb_ip_align(rx_ring->netdev,
1782 						IXGBE_RX_HDR_SIZE);
1783 		if (unlikely(!skb)) {
1784 			rx_ring->rx_stats.alloc_rx_buff_failed++;
1785 			return NULL;
1786 		}
1787 
1788 		/*
1789 		 * we will be copying header into skb->data in
1790 		 * pskb_may_pull so it is in our interest to prefetch
1791 		 * it now to avoid a possible cache miss
1792 		 */
1793 		prefetchw(skb->data);
1794 
1795 		/*
1796 		 * Delay unmapping of the first packet. It carries the
1797 		 * header information, HW may still access the header
1798 		 * after the writeback.  Only unmap it when EOP is
1799 		 * reached
1800 		 */
1801 		if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1802 			goto dma_sync;
1803 
1804 		IXGBE_CB(skb)->dma = rx_buffer->dma;
1805 	} else {
1806 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
1807 			ixgbe_dma_sync_frag(rx_ring, skb);
1808 
1809 dma_sync:
1810 		/* we are reusing so sync this buffer for CPU use */
1811 		dma_sync_single_range_for_cpu(rx_ring->dev,
1812 					      rx_buffer->dma,
1813 					      rx_buffer->page_offset,
1814 					      ixgbe_rx_bufsz(rx_ring),
1815 					      DMA_FROM_DEVICE);
1816 	}
1817 
1818 	/* pull page into skb */
1819 	if (ixgbe_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) {
1820 		/* hand second half of page back to the ring */
1821 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
1822 	} else if (IXGBE_CB(skb)->dma == rx_buffer->dma) {
1823 		/* the page has been released from the ring */
1824 		IXGBE_CB(skb)->page_released = true;
1825 	} else {
1826 		/* we are not reusing the buffer so unmap it */
1827 		dma_unmap_page(rx_ring->dev, rx_buffer->dma,
1828 			       ixgbe_rx_pg_size(rx_ring),
1829 			       DMA_FROM_DEVICE);
1830 	}
1831 
1832 	/* clear contents of buffer_info */
1833 	rx_buffer->skb = NULL;
1834 	rx_buffer->dma = 0;
1835 	rx_buffer->page = NULL;
1836 
1837 	return skb;
1838 }
1839 
1840 /**
1841  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
1842  * @q_vector: structure containing interrupt and ring information
1843  * @rx_ring: rx descriptor ring to transact packets on
1844  * @budget: Total limit on number of packets to process
1845  *
1846  * This function provides a "bounce buffer" approach to Rx interrupt
1847  * processing.  The advantage to this is that on systems that have
1848  * expensive overhead for IOMMU access this provides a means of avoiding
1849  * it by maintaining the mapping of the page to the syste.
1850  *
1851  * Returns true if all work is completed without reaching budget
1852  **/
1853 static bool ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
1854 			       struct ixgbe_ring *rx_ring,
1855 			       const int budget)
1856 {
1857 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1858 #ifdef IXGBE_FCOE
1859 	struct ixgbe_adapter *adapter = q_vector->adapter;
1860 	int ddp_bytes;
1861 	unsigned int mss = 0;
1862 #endif /* IXGBE_FCOE */
1863 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
1864 
1865 	do {
1866 		union ixgbe_adv_rx_desc *rx_desc;
1867 		struct sk_buff *skb;
1868 
1869 		/* return some buffers to hardware, one at a time is too slow */
1870 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
1871 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
1872 			cleaned_count = 0;
1873 		}
1874 
1875 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
1876 
1877 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_DD))
1878 			break;
1879 
1880 		/*
1881 		 * This memory barrier is needed to keep us from reading
1882 		 * any other fields out of the rx_desc until we know the
1883 		 * RXD_STAT_DD bit is set
1884 		 */
1885 		rmb();
1886 
1887 		/* retrieve a buffer from the ring */
1888 		skb = ixgbe_fetch_rx_buffer(rx_ring, rx_desc);
1889 
1890 		/* exit if we failed to retrieve a buffer */
1891 		if (!skb)
1892 			break;
1893 
1894 		cleaned_count++;
1895 
1896 		/* place incomplete frames back on ring for completion */
1897 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
1898 			continue;
1899 
1900 		/* verify the packet layout is correct */
1901 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
1902 			continue;
1903 
1904 		/* probably a little skewed due to removing CRC */
1905 		total_rx_bytes += skb->len;
1906 
1907 		/* populate checksum, timestamp, VLAN, and protocol */
1908 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
1909 
1910 #ifdef IXGBE_FCOE
1911 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
1912 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
1913 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
1914 			/* include DDPed FCoE data */
1915 			if (ddp_bytes > 0) {
1916 				if (!mss) {
1917 					mss = rx_ring->netdev->mtu -
1918 						sizeof(struct fcoe_hdr) -
1919 						sizeof(struct fc_frame_header) -
1920 						sizeof(struct fcoe_crc_eof);
1921 					if (mss > 512)
1922 						mss &= ~511;
1923 				}
1924 				total_rx_bytes += ddp_bytes;
1925 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
1926 								 mss);
1927 			}
1928 			if (!ddp_bytes) {
1929 				dev_kfree_skb_any(skb);
1930 				continue;
1931 			}
1932 		}
1933 
1934 #endif /* IXGBE_FCOE */
1935 		ixgbe_rx_skb(q_vector, skb);
1936 
1937 		/* update budget accounting */
1938 		total_rx_packets++;
1939 	} while (likely(total_rx_packets < budget));
1940 
1941 	u64_stats_update_begin(&rx_ring->syncp);
1942 	rx_ring->stats.packets += total_rx_packets;
1943 	rx_ring->stats.bytes += total_rx_bytes;
1944 	u64_stats_update_end(&rx_ring->syncp);
1945 	q_vector->rx.total_packets += total_rx_packets;
1946 	q_vector->rx.total_bytes += total_rx_bytes;
1947 
1948 	if (cleaned_count)
1949 		ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
1950 
1951 	return (total_rx_packets < budget);
1952 }
1953 
1954 /**
1955  * ixgbe_configure_msix - Configure MSI-X hardware
1956  * @adapter: board private structure
1957  *
1958  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
1959  * interrupts.
1960  **/
1961 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
1962 {
1963 	struct ixgbe_q_vector *q_vector;
1964 	int v_idx;
1965 	u32 mask;
1966 
1967 	/* Populate MSIX to EITR Select */
1968 	if (adapter->num_vfs > 32) {
1969 		u32 eitrsel = (1 << (adapter->num_vfs - 32)) - 1;
1970 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
1971 	}
1972 
1973 	/*
1974 	 * Populate the IVAR table and set the ITR values to the
1975 	 * corresponding register.
1976 	 */
1977 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
1978 		struct ixgbe_ring *ring;
1979 		q_vector = adapter->q_vector[v_idx];
1980 
1981 		ixgbe_for_each_ring(ring, q_vector->rx)
1982 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
1983 
1984 		ixgbe_for_each_ring(ring, q_vector->tx)
1985 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
1986 
1987 		ixgbe_write_eitr(q_vector);
1988 	}
1989 
1990 	switch (adapter->hw.mac.type) {
1991 	case ixgbe_mac_82598EB:
1992 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
1993 			       v_idx);
1994 		break;
1995 	case ixgbe_mac_82599EB:
1996 	case ixgbe_mac_X540:
1997 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
1998 		break;
1999 	default:
2000 		break;
2001 	}
2002 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2003 
2004 	/* set up to autoclear timer, and the vectors */
2005 	mask = IXGBE_EIMS_ENABLE_MASK;
2006 	mask &= ~(IXGBE_EIMS_OTHER |
2007 		  IXGBE_EIMS_MAILBOX |
2008 		  IXGBE_EIMS_LSC);
2009 
2010 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2011 }
2012 
2013 enum latency_range {
2014 	lowest_latency = 0,
2015 	low_latency = 1,
2016 	bulk_latency = 2,
2017 	latency_invalid = 255
2018 };
2019 
2020 /**
2021  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2022  * @q_vector: structure containing interrupt and ring information
2023  * @ring_container: structure containing ring performance data
2024  *
2025  *      Stores a new ITR value based on packets and byte
2026  *      counts during the last interrupt.  The advantage of per interrupt
2027  *      computation is faster updates and more accurate ITR for the current
2028  *      traffic pattern.  Constants in this function were computed
2029  *      based on theoretical maximum wire speed and thresholds were set based
2030  *      on testing data as well as attempting to minimize response time
2031  *      while increasing bulk throughput.
2032  *      this functionality is controlled by the InterruptThrottleRate module
2033  *      parameter (see ixgbe_param.c)
2034  **/
2035 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2036 			     struct ixgbe_ring_container *ring_container)
2037 {
2038 	int bytes = ring_container->total_bytes;
2039 	int packets = ring_container->total_packets;
2040 	u32 timepassed_us;
2041 	u64 bytes_perint;
2042 	u8 itr_setting = ring_container->itr;
2043 
2044 	if (packets == 0)
2045 		return;
2046 
2047 	/* simple throttlerate management
2048 	 *   0-10MB/s   lowest (100000 ints/s)
2049 	 *  10-20MB/s   low    (20000 ints/s)
2050 	 *  20-1249MB/s bulk   (8000 ints/s)
2051 	 */
2052 	/* what was last interrupt timeslice? */
2053 	timepassed_us = q_vector->itr >> 2;
2054 	bytes_perint = bytes / timepassed_us; /* bytes/usec */
2055 
2056 	switch (itr_setting) {
2057 	case lowest_latency:
2058 		if (bytes_perint > 10)
2059 			itr_setting = low_latency;
2060 		break;
2061 	case low_latency:
2062 		if (bytes_perint > 20)
2063 			itr_setting = bulk_latency;
2064 		else if (bytes_perint <= 10)
2065 			itr_setting = lowest_latency;
2066 		break;
2067 	case bulk_latency:
2068 		if (bytes_perint <= 20)
2069 			itr_setting = low_latency;
2070 		break;
2071 	}
2072 
2073 	/* clear work counters since we have the values we need */
2074 	ring_container->total_bytes = 0;
2075 	ring_container->total_packets = 0;
2076 
2077 	/* write updated itr to ring container */
2078 	ring_container->itr = itr_setting;
2079 }
2080 
2081 /**
2082  * ixgbe_write_eitr - write EITR register in hardware specific way
2083  * @q_vector: structure containing interrupt and ring information
2084  *
2085  * This function is made to be called by ethtool and by the driver
2086  * when it needs to update EITR registers at runtime.  Hardware
2087  * specific quirks/differences are taken care of here.
2088  */
2089 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2090 {
2091 	struct ixgbe_adapter *adapter = q_vector->adapter;
2092 	struct ixgbe_hw *hw = &adapter->hw;
2093 	int v_idx = q_vector->v_idx;
2094 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2095 
2096 	switch (adapter->hw.mac.type) {
2097 	case ixgbe_mac_82598EB:
2098 		/* must write high and low 16 bits to reset counter */
2099 		itr_reg |= (itr_reg << 16);
2100 		break;
2101 	case ixgbe_mac_82599EB:
2102 	case ixgbe_mac_X540:
2103 		/*
2104 		 * set the WDIS bit to not clear the timer bits and cause an
2105 		 * immediate assertion of the interrupt
2106 		 */
2107 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2108 		break;
2109 	default:
2110 		break;
2111 	}
2112 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2113 }
2114 
2115 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2116 {
2117 	u32 new_itr = q_vector->itr;
2118 	u8 current_itr;
2119 
2120 	ixgbe_update_itr(q_vector, &q_vector->tx);
2121 	ixgbe_update_itr(q_vector, &q_vector->rx);
2122 
2123 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
2124 
2125 	switch (current_itr) {
2126 	/* counts and packets in update_itr are dependent on these numbers */
2127 	case lowest_latency:
2128 		new_itr = IXGBE_100K_ITR;
2129 		break;
2130 	case low_latency:
2131 		new_itr = IXGBE_20K_ITR;
2132 		break;
2133 	case bulk_latency:
2134 		new_itr = IXGBE_8K_ITR;
2135 		break;
2136 	default:
2137 		break;
2138 	}
2139 
2140 	if (new_itr != q_vector->itr) {
2141 		/* do an exponential smoothing */
2142 		new_itr = (10 * new_itr * q_vector->itr) /
2143 			  ((9 * new_itr) + q_vector->itr);
2144 
2145 		/* save the algorithm value here */
2146 		q_vector->itr = new_itr;
2147 
2148 		ixgbe_write_eitr(q_vector);
2149 	}
2150 }
2151 
2152 /**
2153  * ixgbe_check_overtemp_subtask - check for over temperature
2154  * @adapter: pointer to adapter
2155  **/
2156 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2157 {
2158 	struct ixgbe_hw *hw = &adapter->hw;
2159 	u32 eicr = adapter->interrupt_event;
2160 
2161 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2162 		return;
2163 
2164 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) &&
2165 	    !(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2166 		return;
2167 
2168 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2169 
2170 	switch (hw->device_id) {
2171 	case IXGBE_DEV_ID_82599_T3_LOM:
2172 		/*
2173 		 * Since the warning interrupt is for both ports
2174 		 * we don't have to check if:
2175 		 *  - This interrupt wasn't for our port.
2176 		 *  - We may have missed the interrupt so always have to
2177 		 *    check if we  got a LSC
2178 		 */
2179 		if (!(eicr & IXGBE_EICR_GPI_SDP0) &&
2180 		    !(eicr & IXGBE_EICR_LSC))
2181 			return;
2182 
2183 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2184 			u32 autoneg;
2185 			bool link_up = false;
2186 
2187 			hw->mac.ops.check_link(hw, &autoneg, &link_up, false);
2188 
2189 			if (link_up)
2190 				return;
2191 		}
2192 
2193 		/* Check if this is not due to overtemp */
2194 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2195 			return;
2196 
2197 		break;
2198 	default:
2199 		if (!(eicr & IXGBE_EICR_GPI_SDP0))
2200 			return;
2201 		break;
2202 	}
2203 	e_crit(drv,
2204 	       "Network adapter has been stopped because it has over heated. "
2205 	       "Restart the computer. If the problem persists, "
2206 	       "power off the system and replace the adapter\n");
2207 
2208 	adapter->interrupt_event = 0;
2209 }
2210 
2211 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2212 {
2213 	struct ixgbe_hw *hw = &adapter->hw;
2214 
2215 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2216 	    (eicr & IXGBE_EICR_GPI_SDP1)) {
2217 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2218 		/* write to clear the interrupt */
2219 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2220 	}
2221 }
2222 
2223 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2224 {
2225 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2226 		return;
2227 
2228 	switch (adapter->hw.mac.type) {
2229 	case ixgbe_mac_82599EB:
2230 		/*
2231 		 * Need to check link state so complete overtemp check
2232 		 * on service task
2233 		 */
2234 		if (((eicr & IXGBE_EICR_GPI_SDP0) || (eicr & IXGBE_EICR_LSC)) &&
2235 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2236 			adapter->interrupt_event = eicr;
2237 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2238 			ixgbe_service_event_schedule(adapter);
2239 			return;
2240 		}
2241 		return;
2242 	case ixgbe_mac_X540:
2243 		if (!(eicr & IXGBE_EICR_TS))
2244 			return;
2245 		break;
2246 	default:
2247 		return;
2248 	}
2249 
2250 	e_crit(drv,
2251 	       "Network adapter has been stopped because it has over heated. "
2252 	       "Restart the computer. If the problem persists, "
2253 	       "power off the system and replace the adapter\n");
2254 }
2255 
2256 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2257 {
2258 	struct ixgbe_hw *hw = &adapter->hw;
2259 
2260 	if (eicr & IXGBE_EICR_GPI_SDP2) {
2261 		/* Clear the interrupt */
2262 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP2);
2263 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2264 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2265 			ixgbe_service_event_schedule(adapter);
2266 		}
2267 	}
2268 
2269 	if (eicr & IXGBE_EICR_GPI_SDP1) {
2270 		/* Clear the interrupt */
2271 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2272 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2273 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2274 			ixgbe_service_event_schedule(adapter);
2275 		}
2276 	}
2277 }
2278 
2279 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2280 {
2281 	struct ixgbe_hw *hw = &adapter->hw;
2282 
2283 	adapter->lsc_int++;
2284 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2285 	adapter->link_check_timeout = jiffies;
2286 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2287 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2288 		IXGBE_WRITE_FLUSH(hw);
2289 		ixgbe_service_event_schedule(adapter);
2290 	}
2291 }
2292 
2293 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2294 					   u64 qmask)
2295 {
2296 	u32 mask;
2297 	struct ixgbe_hw *hw = &adapter->hw;
2298 
2299 	switch (hw->mac.type) {
2300 	case ixgbe_mac_82598EB:
2301 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2302 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2303 		break;
2304 	case ixgbe_mac_82599EB:
2305 	case ixgbe_mac_X540:
2306 		mask = (qmask & 0xFFFFFFFF);
2307 		if (mask)
2308 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2309 		mask = (qmask >> 32);
2310 		if (mask)
2311 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2312 		break;
2313 	default:
2314 		break;
2315 	}
2316 	/* skip the flush */
2317 }
2318 
2319 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2320 					    u64 qmask)
2321 {
2322 	u32 mask;
2323 	struct ixgbe_hw *hw = &adapter->hw;
2324 
2325 	switch (hw->mac.type) {
2326 	case ixgbe_mac_82598EB:
2327 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2328 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2329 		break;
2330 	case ixgbe_mac_82599EB:
2331 	case ixgbe_mac_X540:
2332 		mask = (qmask & 0xFFFFFFFF);
2333 		if (mask)
2334 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2335 		mask = (qmask >> 32);
2336 		if (mask)
2337 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2338 		break;
2339 	default:
2340 		break;
2341 	}
2342 	/* skip the flush */
2343 }
2344 
2345 /**
2346  * ixgbe_irq_enable - Enable default interrupt generation settings
2347  * @adapter: board private structure
2348  **/
2349 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2350 				    bool flush)
2351 {
2352 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2353 
2354 	/* don't reenable LSC while waiting for link */
2355 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2356 		mask &= ~IXGBE_EIMS_LSC;
2357 
2358 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2359 		switch (adapter->hw.mac.type) {
2360 		case ixgbe_mac_82599EB:
2361 			mask |= IXGBE_EIMS_GPI_SDP0;
2362 			break;
2363 		case ixgbe_mac_X540:
2364 			mask |= IXGBE_EIMS_TS;
2365 			break;
2366 		default:
2367 			break;
2368 		}
2369 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2370 		mask |= IXGBE_EIMS_GPI_SDP1;
2371 	switch (adapter->hw.mac.type) {
2372 	case ixgbe_mac_82599EB:
2373 		mask |= IXGBE_EIMS_GPI_SDP1;
2374 		mask |= IXGBE_EIMS_GPI_SDP2;
2375 	case ixgbe_mac_X540:
2376 		mask |= IXGBE_EIMS_ECC;
2377 		mask |= IXGBE_EIMS_MAILBOX;
2378 		break;
2379 	default:
2380 		break;
2381 	}
2382 
2383 	if (adapter->hw.mac.type == ixgbe_mac_X540)
2384 		mask |= IXGBE_EIMS_TIMESYNC;
2385 
2386 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2387 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
2388 		mask |= IXGBE_EIMS_FLOW_DIR;
2389 
2390 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
2391 	if (queues)
2392 		ixgbe_irq_enable_queues(adapter, ~0);
2393 	if (flush)
2394 		IXGBE_WRITE_FLUSH(&adapter->hw);
2395 }
2396 
2397 static irqreturn_t ixgbe_msix_other(int irq, void *data)
2398 {
2399 	struct ixgbe_adapter *adapter = data;
2400 	struct ixgbe_hw *hw = &adapter->hw;
2401 	u32 eicr;
2402 
2403 	/*
2404 	 * Workaround for Silicon errata.  Use clear-by-write instead
2405 	 * of clear-by-read.  Reading with EICS will return the
2406 	 * interrupt causes without clearing, which later be done
2407 	 * with the write to EICR.
2408 	 */
2409 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2410 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2411 
2412 	if (eicr & IXGBE_EICR_LSC)
2413 		ixgbe_check_lsc(adapter);
2414 
2415 	if (eicr & IXGBE_EICR_MAILBOX)
2416 		ixgbe_msg_task(adapter);
2417 
2418 	switch (hw->mac.type) {
2419 	case ixgbe_mac_82599EB:
2420 	case ixgbe_mac_X540:
2421 		if (eicr & IXGBE_EICR_ECC)
2422 			e_info(link, "Received unrecoverable ECC Err, please "
2423 			       "reboot\n");
2424 		/* Handle Flow Director Full threshold interrupt */
2425 		if (eicr & IXGBE_EICR_FLOW_DIR) {
2426 			int reinit_count = 0;
2427 			int i;
2428 			for (i = 0; i < adapter->num_tx_queues; i++) {
2429 				struct ixgbe_ring *ring = adapter->tx_ring[i];
2430 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
2431 						       &ring->state))
2432 					reinit_count++;
2433 			}
2434 			if (reinit_count) {
2435 				/* no more flow director interrupts until after init */
2436 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
2437 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
2438 				ixgbe_service_event_schedule(adapter);
2439 			}
2440 		}
2441 		ixgbe_check_sfp_event(adapter, eicr);
2442 		ixgbe_check_overtemp_event(adapter, eicr);
2443 		break;
2444 	default:
2445 		break;
2446 	}
2447 
2448 	ixgbe_check_fan_failure(adapter, eicr);
2449 
2450 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2451 		ixgbe_ptp_check_pps_event(adapter, eicr);
2452 
2453 	/* re-enable the original interrupt state, no lsc, no queues */
2454 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2455 		ixgbe_irq_enable(adapter, false, false);
2456 
2457 	return IRQ_HANDLED;
2458 }
2459 
2460 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
2461 {
2462 	struct ixgbe_q_vector *q_vector = data;
2463 
2464 	/* EIAM disabled interrupts (on this vector) for us */
2465 
2466 	if (q_vector->rx.ring || q_vector->tx.ring)
2467 		napi_schedule(&q_vector->napi);
2468 
2469 	return IRQ_HANDLED;
2470 }
2471 
2472 /**
2473  * ixgbe_poll - NAPI Rx polling callback
2474  * @napi: structure for representing this polling device
2475  * @budget: how many packets driver is allowed to clean
2476  *
2477  * This function is used for legacy and MSI, NAPI mode
2478  **/
2479 int ixgbe_poll(struct napi_struct *napi, int budget)
2480 {
2481 	struct ixgbe_q_vector *q_vector =
2482 				container_of(napi, struct ixgbe_q_vector, napi);
2483 	struct ixgbe_adapter *adapter = q_vector->adapter;
2484 	struct ixgbe_ring *ring;
2485 	int per_ring_budget;
2486 	bool clean_complete = true;
2487 
2488 #ifdef CONFIG_IXGBE_DCA
2489 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
2490 		ixgbe_update_dca(q_vector);
2491 #endif
2492 
2493 	ixgbe_for_each_ring(ring, q_vector->tx)
2494 		clean_complete &= !!ixgbe_clean_tx_irq(q_vector, ring);
2495 
2496 	/* attempt to distribute budget to each queue fairly, but don't allow
2497 	 * the budget to go below 1 because we'll exit polling */
2498 	if (q_vector->rx.count > 1)
2499 		per_ring_budget = max(budget/q_vector->rx.count, 1);
2500 	else
2501 		per_ring_budget = budget;
2502 
2503 	ixgbe_for_each_ring(ring, q_vector->rx)
2504 		clean_complete &= ixgbe_clean_rx_irq(q_vector, ring,
2505 						     per_ring_budget);
2506 
2507 	/* If all work not completed, return budget and keep polling */
2508 	if (!clean_complete)
2509 		return budget;
2510 
2511 	/* all work done, exit the polling mode */
2512 	napi_complete(napi);
2513 	if (adapter->rx_itr_setting & 1)
2514 		ixgbe_set_itr(q_vector);
2515 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2516 		ixgbe_irq_enable_queues(adapter, ((u64)1 << q_vector->v_idx));
2517 
2518 	return 0;
2519 }
2520 
2521 /**
2522  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2523  * @adapter: board private structure
2524  *
2525  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2526  * interrupts from the kernel.
2527  **/
2528 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
2529 {
2530 	struct net_device *netdev = adapter->netdev;
2531 	int vector, err;
2532 	int ri = 0, ti = 0;
2533 
2534 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2535 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2536 		struct msix_entry *entry = &adapter->msix_entries[vector];
2537 
2538 		if (q_vector->tx.ring && q_vector->rx.ring) {
2539 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2540 				 "%s-%s-%d", netdev->name, "TxRx", ri++);
2541 			ti++;
2542 		} else if (q_vector->rx.ring) {
2543 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2544 				 "%s-%s-%d", netdev->name, "rx", ri++);
2545 		} else if (q_vector->tx.ring) {
2546 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2547 				 "%s-%s-%d", netdev->name, "tx", ti++);
2548 		} else {
2549 			/* skip this unused q_vector */
2550 			continue;
2551 		}
2552 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
2553 				  q_vector->name, q_vector);
2554 		if (err) {
2555 			e_err(probe, "request_irq failed for MSIX interrupt "
2556 			      "Error: %d\n", err);
2557 			goto free_queue_irqs;
2558 		}
2559 		/* If Flow Director is enabled, set interrupt affinity */
2560 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2561 			/* assign the mask for this irq */
2562 			irq_set_affinity_hint(entry->vector,
2563 					      &q_vector->affinity_mask);
2564 		}
2565 	}
2566 
2567 	err = request_irq(adapter->msix_entries[vector].vector,
2568 			  ixgbe_msix_other, 0, netdev->name, adapter);
2569 	if (err) {
2570 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
2571 		goto free_queue_irqs;
2572 	}
2573 
2574 	return 0;
2575 
2576 free_queue_irqs:
2577 	while (vector) {
2578 		vector--;
2579 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
2580 				      NULL);
2581 		free_irq(adapter->msix_entries[vector].vector,
2582 			 adapter->q_vector[vector]);
2583 	}
2584 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
2585 	pci_disable_msix(adapter->pdev);
2586 	kfree(adapter->msix_entries);
2587 	adapter->msix_entries = NULL;
2588 	return err;
2589 }
2590 
2591 /**
2592  * ixgbe_intr - legacy mode Interrupt Handler
2593  * @irq: interrupt number
2594  * @data: pointer to a network interface device structure
2595  **/
2596 static irqreturn_t ixgbe_intr(int irq, void *data)
2597 {
2598 	struct ixgbe_adapter *adapter = data;
2599 	struct ixgbe_hw *hw = &adapter->hw;
2600 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2601 	u32 eicr;
2602 
2603 	/*
2604 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
2605 	 * before the read of EICR.
2606 	 */
2607 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
2608 
2609 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2610 	 * therefore no explicit interrupt disable is necessary */
2611 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2612 	if (!eicr) {
2613 		/*
2614 		 * shared interrupt alert!
2615 		 * make sure interrupts are enabled because the read will
2616 		 * have disabled interrupts due to EIAM
2617 		 * finish the workaround of silicon errata on 82598.  Unmask
2618 		 * the interrupt that we masked before the EICR read.
2619 		 */
2620 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
2621 			ixgbe_irq_enable(adapter, true, true);
2622 		return IRQ_NONE;	/* Not our interrupt */
2623 	}
2624 
2625 	if (eicr & IXGBE_EICR_LSC)
2626 		ixgbe_check_lsc(adapter);
2627 
2628 	switch (hw->mac.type) {
2629 	case ixgbe_mac_82599EB:
2630 		ixgbe_check_sfp_event(adapter, eicr);
2631 		/* Fall through */
2632 	case ixgbe_mac_X540:
2633 		if (eicr & IXGBE_EICR_ECC)
2634 			e_info(link, "Received unrecoverable ECC err, please "
2635 				     "reboot\n");
2636 		ixgbe_check_overtemp_event(adapter, eicr);
2637 		break;
2638 	default:
2639 		break;
2640 	}
2641 
2642 	ixgbe_check_fan_failure(adapter, eicr);
2643 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2644 		ixgbe_ptp_check_pps_event(adapter, eicr);
2645 
2646 	/* would disable interrupts here but EIAM disabled it */
2647 	napi_schedule(&q_vector->napi);
2648 
2649 	/*
2650 	 * re-enable link(maybe) and non-queue interrupts, no flush.
2651 	 * ixgbe_poll will re-enable the queue interrupts
2652 	 */
2653 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2654 		ixgbe_irq_enable(adapter, false, false);
2655 
2656 	return IRQ_HANDLED;
2657 }
2658 
2659 /**
2660  * ixgbe_request_irq - initialize interrupts
2661  * @adapter: board private structure
2662  *
2663  * Attempts to configure interrupts using the best available
2664  * capabilities of the hardware and kernel.
2665  **/
2666 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
2667 {
2668 	struct net_device *netdev = adapter->netdev;
2669 	int err;
2670 
2671 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
2672 		err = ixgbe_request_msix_irqs(adapter);
2673 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
2674 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
2675 				  netdev->name, adapter);
2676 	else
2677 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
2678 				  netdev->name, adapter);
2679 
2680 	if (err)
2681 		e_err(probe, "request_irq failed, Error %d\n", err);
2682 
2683 	return err;
2684 }
2685 
2686 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
2687 {
2688 	int vector;
2689 
2690 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
2691 		free_irq(adapter->pdev->irq, adapter);
2692 		return;
2693 	}
2694 
2695 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2696 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2697 		struct msix_entry *entry = &adapter->msix_entries[vector];
2698 
2699 		/* free only the irqs that were actually requested */
2700 		if (!q_vector->rx.ring && !q_vector->tx.ring)
2701 			continue;
2702 
2703 		/* clear the affinity_mask in the IRQ descriptor */
2704 		irq_set_affinity_hint(entry->vector, NULL);
2705 
2706 		free_irq(entry->vector, q_vector);
2707 	}
2708 
2709 	free_irq(adapter->msix_entries[vector++].vector, adapter);
2710 }
2711 
2712 /**
2713  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
2714  * @adapter: board private structure
2715  **/
2716 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
2717 {
2718 	switch (adapter->hw.mac.type) {
2719 	case ixgbe_mac_82598EB:
2720 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
2721 		break;
2722 	case ixgbe_mac_82599EB:
2723 	case ixgbe_mac_X540:
2724 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
2725 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
2726 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
2727 		break;
2728 	default:
2729 		break;
2730 	}
2731 	IXGBE_WRITE_FLUSH(&adapter->hw);
2732 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
2733 		int vector;
2734 
2735 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
2736 			synchronize_irq(adapter->msix_entries[vector].vector);
2737 
2738 		synchronize_irq(adapter->msix_entries[vector++].vector);
2739 	} else {
2740 		synchronize_irq(adapter->pdev->irq);
2741 	}
2742 }
2743 
2744 /**
2745  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
2746  *
2747  **/
2748 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
2749 {
2750 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2751 
2752 	ixgbe_write_eitr(q_vector);
2753 
2754 	ixgbe_set_ivar(adapter, 0, 0, 0);
2755 	ixgbe_set_ivar(adapter, 1, 0, 0);
2756 
2757 	e_info(hw, "Legacy interrupt IVAR setup done\n");
2758 }
2759 
2760 /**
2761  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
2762  * @adapter: board private structure
2763  * @ring: structure containing ring specific data
2764  *
2765  * Configure the Tx descriptor ring after a reset.
2766  **/
2767 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
2768 			     struct ixgbe_ring *ring)
2769 {
2770 	struct ixgbe_hw *hw = &adapter->hw;
2771 	u64 tdba = ring->dma;
2772 	int wait_loop = 10;
2773 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
2774 	u8 reg_idx = ring->reg_idx;
2775 
2776 	/* disable queue to avoid issues while updating state */
2777 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
2778 	IXGBE_WRITE_FLUSH(hw);
2779 
2780 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
2781 			(tdba & DMA_BIT_MASK(32)));
2782 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
2783 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
2784 			ring->count * sizeof(union ixgbe_adv_tx_desc));
2785 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
2786 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
2787 	ring->tail = hw->hw_addr + IXGBE_TDT(reg_idx);
2788 
2789 	/*
2790 	 * set WTHRESH to encourage burst writeback, it should not be set
2791 	 * higher than 1 when ITR is 0 as it could cause false TX hangs
2792 	 *
2793 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
2794 	 * to or less than the number of on chip descriptors, which is
2795 	 * currently 40.
2796 	 */
2797 	if (!ring->q_vector || (ring->q_vector->itr < 8))
2798 		txdctl |= (1 << 16);	/* WTHRESH = 1 */
2799 	else
2800 		txdctl |= (8 << 16);	/* WTHRESH = 8 */
2801 
2802 	/*
2803 	 * Setting PTHRESH to 32 both improves performance
2804 	 * and avoids a TX hang with DFP enabled
2805 	 */
2806 	txdctl |= (1 << 8) |	/* HTHRESH = 1 */
2807 		   32;		/* PTHRESH = 32 */
2808 
2809 	/* reinitialize flowdirector state */
2810 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2811 		ring->atr_sample_rate = adapter->atr_sample_rate;
2812 		ring->atr_count = 0;
2813 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
2814 	} else {
2815 		ring->atr_sample_rate = 0;
2816 	}
2817 
2818 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
2819 
2820 	/* enable queue */
2821 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
2822 
2823 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
2824 	if (hw->mac.type == ixgbe_mac_82598EB &&
2825 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
2826 		return;
2827 
2828 	/* poll to verify queue is enabled */
2829 	do {
2830 		usleep_range(1000, 2000);
2831 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
2832 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
2833 	if (!wait_loop)
2834 		e_err(drv, "Could not enable Tx Queue %d\n", reg_idx);
2835 }
2836 
2837 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
2838 {
2839 	struct ixgbe_hw *hw = &adapter->hw;
2840 	u32 rttdcs, mtqc;
2841 	u8 tcs = netdev_get_num_tc(adapter->netdev);
2842 
2843 	if (hw->mac.type == ixgbe_mac_82598EB)
2844 		return;
2845 
2846 	/* disable the arbiter while setting MTQC */
2847 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
2848 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
2849 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
2850 
2851 	/* set transmit pool layout */
2852 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
2853 		mtqc = IXGBE_MTQC_VT_ENA;
2854 		if (tcs > 4)
2855 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
2856 		else if (tcs > 1)
2857 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
2858 		else if (adapter->ring_feature[RING_F_RSS].indices == 4)
2859 			mtqc |= IXGBE_MTQC_32VF;
2860 		else
2861 			mtqc |= IXGBE_MTQC_64VF;
2862 	} else {
2863 		if (tcs > 4)
2864 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
2865 		else if (tcs > 1)
2866 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
2867 		else
2868 			mtqc = IXGBE_MTQC_64Q_1PB;
2869 	}
2870 
2871 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
2872 
2873 	/* Enable Security TX Buffer IFG for multiple pb */
2874 	if (tcs) {
2875 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
2876 		sectx |= IXGBE_SECTX_DCB;
2877 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
2878 	}
2879 
2880 	/* re-enable the arbiter */
2881 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
2882 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
2883 }
2884 
2885 /**
2886  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
2887  * @adapter: board private structure
2888  *
2889  * Configure the Tx unit of the MAC after a reset.
2890  **/
2891 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
2892 {
2893 	struct ixgbe_hw *hw = &adapter->hw;
2894 	u32 dmatxctl;
2895 	u32 i;
2896 
2897 	ixgbe_setup_mtqc(adapter);
2898 
2899 	if (hw->mac.type != ixgbe_mac_82598EB) {
2900 		/* DMATXCTL.EN must be before Tx queues are enabled */
2901 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2902 		dmatxctl |= IXGBE_DMATXCTL_TE;
2903 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
2904 	}
2905 
2906 	/* Setup the HW Tx Head and Tail descriptor pointers */
2907 	for (i = 0; i < adapter->num_tx_queues; i++)
2908 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
2909 }
2910 
2911 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
2912 				 struct ixgbe_ring *ring)
2913 {
2914 	struct ixgbe_hw *hw = &adapter->hw;
2915 	u8 reg_idx = ring->reg_idx;
2916 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
2917 
2918 	srrctl |= IXGBE_SRRCTL_DROP_EN;
2919 
2920 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
2921 }
2922 
2923 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
2924 				  struct ixgbe_ring *ring)
2925 {
2926 	struct ixgbe_hw *hw = &adapter->hw;
2927 	u8 reg_idx = ring->reg_idx;
2928 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
2929 
2930 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
2931 
2932 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
2933 }
2934 
2935 #ifdef CONFIG_IXGBE_DCB
2936 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
2937 #else
2938 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
2939 #endif
2940 {
2941 	int i;
2942 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
2943 
2944 	if (adapter->ixgbe_ieee_pfc)
2945 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
2946 
2947 	/*
2948 	 * We should set the drop enable bit if:
2949 	 *  SR-IOV is enabled
2950 	 *   or
2951 	 *  Number of Rx queues > 1 and flow control is disabled
2952 	 *
2953 	 *  This allows us to avoid head of line blocking for security
2954 	 *  and performance reasons.
2955 	 */
2956 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
2957 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
2958 		for (i = 0; i < adapter->num_rx_queues; i++)
2959 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
2960 	} else {
2961 		for (i = 0; i < adapter->num_rx_queues; i++)
2962 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
2963 	}
2964 }
2965 
2966 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
2967 
2968 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
2969 				   struct ixgbe_ring *rx_ring)
2970 {
2971 	struct ixgbe_hw *hw = &adapter->hw;
2972 	u32 srrctl;
2973 	u8 reg_idx = rx_ring->reg_idx;
2974 
2975 	if (hw->mac.type == ixgbe_mac_82598EB) {
2976 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
2977 
2978 		/*
2979 		 * if VMDq is not active we must program one srrctl register
2980 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
2981 		 */
2982 		reg_idx &= mask;
2983 	}
2984 
2985 	/* configure header buffer length, needed for RSC */
2986 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
2987 
2988 	/* configure the packet buffer length */
2989 	srrctl |= ixgbe_rx_bufsz(rx_ring) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
2990 
2991 	/* configure descriptor type */
2992 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
2993 
2994 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
2995 }
2996 
2997 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
2998 {
2999 	struct ixgbe_hw *hw = &adapter->hw;
3000 	static const u32 seed[10] = { 0xE291D73D, 0x1805EC6C, 0x2A94B30D,
3001 			  0xA54F2BEC, 0xEA49AF7C, 0xE214AD3D, 0xB855AABE,
3002 			  0x6A3E67EA, 0x14364D17, 0x3BED200D};
3003 	u32 mrqc = 0, reta = 0;
3004 	u32 rxcsum;
3005 	int i, j;
3006 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3007 
3008 	/*
3009 	 * Program table for at least 2 queues w/ SR-IOV so that VFs can
3010 	 * make full use of any rings they may have.  We will use the
3011 	 * PSRTYPE register to control how many rings we use within the PF.
3012 	 */
3013 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 2))
3014 		rss_i = 2;
3015 
3016 	/* Fill out hash function seeds */
3017 	for (i = 0; i < 10; i++)
3018 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), seed[i]);
3019 
3020 	/* Fill out redirection table */
3021 	for (i = 0, j = 0; i < 128; i++, j++) {
3022 		if (j == rss_i)
3023 			j = 0;
3024 		/* reta = 4-byte sliding window of
3025 		 * 0x00..(indices-1)(indices-1)00..etc. */
3026 		reta = (reta << 8) | (j * 0x11);
3027 		if ((i & 3) == 3)
3028 			IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3029 	}
3030 
3031 	/* Disable indicating checksum in descriptor, enables RSS hash */
3032 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3033 	rxcsum |= IXGBE_RXCSUM_PCSD;
3034 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3035 
3036 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3037 		if (adapter->ring_feature[RING_F_RSS].mask)
3038 			mrqc = IXGBE_MRQC_RSSEN;
3039 	} else {
3040 		u8 tcs = netdev_get_num_tc(adapter->netdev);
3041 
3042 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3043 			if (tcs > 4)
3044 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3045 			else if (tcs > 1)
3046 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3047 			else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3048 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3049 			else
3050 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3051 		} else {
3052 			if (tcs > 4)
3053 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3054 			else if (tcs > 1)
3055 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3056 			else
3057 				mrqc = IXGBE_MRQC_RSSEN;
3058 		}
3059 	}
3060 
3061 	/* Perform hash on these packet types */
3062 	mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3063 		IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3064 		IXGBE_MRQC_RSS_FIELD_IPV6 |
3065 		IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3066 
3067 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3068 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3069 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3070 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3071 
3072 	IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3073 }
3074 
3075 /**
3076  * ixgbe_configure_rscctl - enable RSC for the indicated ring
3077  * @adapter:    address of board private structure
3078  * @index:      index of ring to set
3079  **/
3080 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
3081 				   struct ixgbe_ring *ring)
3082 {
3083 	struct ixgbe_hw *hw = &adapter->hw;
3084 	u32 rscctrl;
3085 	u8 reg_idx = ring->reg_idx;
3086 
3087 	if (!ring_is_rsc_enabled(ring))
3088 		return;
3089 
3090 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
3091 	rscctrl |= IXGBE_RSCCTL_RSCEN;
3092 	/*
3093 	 * we must limit the number of descriptors so that the
3094 	 * total size of max desc * buf_len is not greater
3095 	 * than 65536
3096 	 */
3097 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
3098 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
3099 }
3100 
3101 #define IXGBE_MAX_RX_DESC_POLL 10
3102 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
3103 				       struct ixgbe_ring *ring)
3104 {
3105 	struct ixgbe_hw *hw = &adapter->hw;
3106 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3107 	u32 rxdctl;
3108 	u8 reg_idx = ring->reg_idx;
3109 
3110 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3111 	if (hw->mac.type == ixgbe_mac_82598EB &&
3112 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3113 		return;
3114 
3115 	do {
3116 		usleep_range(1000, 2000);
3117 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3118 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
3119 
3120 	if (!wait_loop) {
3121 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
3122 		      "the polling period\n", reg_idx);
3123 	}
3124 }
3125 
3126 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
3127 			    struct ixgbe_ring *ring)
3128 {
3129 	struct ixgbe_hw *hw = &adapter->hw;
3130 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3131 	u32 rxdctl;
3132 	u8 reg_idx = ring->reg_idx;
3133 
3134 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3135 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
3136 
3137 	/* write value back with RXDCTL.ENABLE bit cleared */
3138 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3139 
3140 	if (hw->mac.type == ixgbe_mac_82598EB &&
3141 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3142 		return;
3143 
3144 	/* the hardware may take up to 100us to really disable the rx queue */
3145 	do {
3146 		udelay(10);
3147 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3148 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
3149 
3150 	if (!wait_loop) {
3151 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3152 		      "the polling period\n", reg_idx);
3153 	}
3154 }
3155 
3156 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
3157 			     struct ixgbe_ring *ring)
3158 {
3159 	struct ixgbe_hw *hw = &adapter->hw;
3160 	u64 rdba = ring->dma;
3161 	u32 rxdctl;
3162 	u8 reg_idx = ring->reg_idx;
3163 
3164 	/* disable queue to avoid issues while updating state */
3165 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3166 	ixgbe_disable_rx_queue(adapter, ring);
3167 
3168 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
3169 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
3170 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
3171 			ring->count * sizeof(union ixgbe_adv_rx_desc));
3172 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
3173 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
3174 	ring->tail = hw->hw_addr + IXGBE_RDT(reg_idx);
3175 
3176 	ixgbe_configure_srrctl(adapter, ring);
3177 	ixgbe_configure_rscctl(adapter, ring);
3178 
3179 	if (hw->mac.type == ixgbe_mac_82598EB) {
3180 		/*
3181 		 * enable cache line friendly hardware writes:
3182 		 * PTHRESH=32 descriptors (half the internal cache),
3183 		 * this also removes ugly rx_no_buffer_count increment
3184 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
3185 		 * WTHRESH=8 burst writeback up to two cache lines
3186 		 */
3187 		rxdctl &= ~0x3FFFFF;
3188 		rxdctl |=  0x080420;
3189 	}
3190 
3191 	/* enable receive descriptor ring */
3192 	rxdctl |= IXGBE_RXDCTL_ENABLE;
3193 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3194 
3195 	ixgbe_rx_desc_queue_enable(adapter, ring);
3196 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
3197 }
3198 
3199 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
3200 {
3201 	struct ixgbe_hw *hw = &adapter->hw;
3202 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
3203 	int p;
3204 
3205 	/* PSRTYPE must be initialized in non 82598 adapters */
3206 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
3207 		      IXGBE_PSRTYPE_UDPHDR |
3208 		      IXGBE_PSRTYPE_IPV4HDR |
3209 		      IXGBE_PSRTYPE_L2HDR |
3210 		      IXGBE_PSRTYPE_IPV6HDR;
3211 
3212 	if (hw->mac.type == ixgbe_mac_82598EB)
3213 		return;
3214 
3215 	if (rss_i > 3)
3216 		psrtype |= 2 << 29;
3217 	else if (rss_i > 1)
3218 		psrtype |= 1 << 29;
3219 
3220 	for (p = 0; p < adapter->num_rx_pools; p++)
3221 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(p)),
3222 				psrtype);
3223 }
3224 
3225 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
3226 {
3227 	struct ixgbe_hw *hw = &adapter->hw;
3228 	u32 reg_offset, vf_shift;
3229 	u32 gcr_ext, vmdctl;
3230 	int i;
3231 
3232 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3233 		return;
3234 
3235 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3236 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
3237 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
3238 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
3239 	vmdctl |= IXGBE_VT_CTL_REPLEN;
3240 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
3241 
3242 	vf_shift = VMDQ_P(0) % 32;
3243 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
3244 
3245 	/* Enable only the PF's pool for Tx/Rx */
3246 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), (~0) << vf_shift);
3247 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
3248 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), (~0) << vf_shift);
3249 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
3250 	if (adapter->flags2 & IXGBE_FLAG2_BRIDGE_MODE_VEB)
3251 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3252 
3253 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3254 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
3255 
3256 	/*
3257 	 * Set up VF register offsets for selected VT Mode,
3258 	 * i.e. 32 or 64 VFs for SR-IOV
3259 	 */
3260 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
3261 	case IXGBE_82599_VMDQ_8Q_MASK:
3262 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
3263 		break;
3264 	case IXGBE_82599_VMDQ_4Q_MASK:
3265 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
3266 		break;
3267 	default:
3268 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
3269 		break;
3270 	}
3271 
3272 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
3273 
3274 
3275 	/* Enable MAC Anti-Spoofing */
3276 	hw->mac.ops.set_mac_anti_spoofing(hw, (adapter->num_vfs != 0),
3277 					  adapter->num_vfs);
3278 	/* For VFs that have spoof checking turned off */
3279 	for (i = 0; i < adapter->num_vfs; i++) {
3280 		if (!adapter->vfinfo[i].spoofchk_enabled)
3281 			ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i, false);
3282 	}
3283 }
3284 
3285 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
3286 {
3287 	struct ixgbe_hw *hw = &adapter->hw;
3288 	struct net_device *netdev = adapter->netdev;
3289 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3290 	struct ixgbe_ring *rx_ring;
3291 	int i;
3292 	u32 mhadd, hlreg0;
3293 
3294 #ifdef IXGBE_FCOE
3295 	/* adjust max frame to be able to do baby jumbo for FCoE */
3296 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
3297 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
3298 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3299 
3300 #endif /* IXGBE_FCOE */
3301 
3302 	/* adjust max frame to be at least the size of a standard frame */
3303 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
3304 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
3305 
3306 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3307 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
3308 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3309 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
3310 
3311 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3312 	}
3313 
3314 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3315 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3316 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3317 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3318 
3319 	/*
3320 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3321 	 * the Base and Length of the Rx Descriptor Ring
3322 	 */
3323 	for (i = 0; i < adapter->num_rx_queues; i++) {
3324 		rx_ring = adapter->rx_ring[i];
3325 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
3326 			set_ring_rsc_enabled(rx_ring);
3327 		else
3328 			clear_ring_rsc_enabled(rx_ring);
3329 	}
3330 }
3331 
3332 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
3333 {
3334 	struct ixgbe_hw *hw = &adapter->hw;
3335 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
3336 
3337 	switch (hw->mac.type) {
3338 	case ixgbe_mac_82598EB:
3339 		/*
3340 		 * For VMDq support of different descriptor types or
3341 		 * buffer sizes through the use of multiple SRRCTL
3342 		 * registers, RDRXCTL.MVMEN must be set to 1
3343 		 *
3344 		 * also, the manual doesn't mention it clearly but DCA hints
3345 		 * will only use queue 0's tags unless this bit is set.  Side
3346 		 * effects of setting this bit are only that SRRCTL must be
3347 		 * fully programmed [0..15]
3348 		 */
3349 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
3350 		break;
3351 	case ixgbe_mac_82599EB:
3352 	case ixgbe_mac_X540:
3353 		/* Disable RSC for ACK packets */
3354 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
3355 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
3356 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
3357 		/* hardware requires some bits to be set by default */
3358 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
3359 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
3360 		break;
3361 	default:
3362 		/* We should do nothing since we don't know this hardware */
3363 		return;
3364 	}
3365 
3366 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
3367 }
3368 
3369 /**
3370  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3371  * @adapter: board private structure
3372  *
3373  * Configure the Rx unit of the MAC after a reset.
3374  **/
3375 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
3376 {
3377 	struct ixgbe_hw *hw = &adapter->hw;
3378 	int i;
3379 	u32 rxctrl;
3380 
3381 	/* disable receives while setting up the descriptors */
3382 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3383 	IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxctrl & ~IXGBE_RXCTRL_RXEN);
3384 
3385 	ixgbe_setup_psrtype(adapter);
3386 	ixgbe_setup_rdrxctl(adapter);
3387 
3388 	/* Program registers for the distribution of queues */
3389 	ixgbe_setup_mrqc(adapter);
3390 
3391 	/* set_rx_buffer_len must be called before ring initialization */
3392 	ixgbe_set_rx_buffer_len(adapter);
3393 
3394 	/*
3395 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3396 	 * the Base and Length of the Rx Descriptor Ring
3397 	 */
3398 	for (i = 0; i < adapter->num_rx_queues; i++)
3399 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
3400 
3401 	/* disable drop enable for 82598 parts */
3402 	if (hw->mac.type == ixgbe_mac_82598EB)
3403 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3404 
3405 	/* enable all receives */
3406 	rxctrl |= IXGBE_RXCTRL_RXEN;
3407 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
3408 }
3409 
3410 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev, u16 vid)
3411 {
3412 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3413 	struct ixgbe_hw *hw = &adapter->hw;
3414 
3415 	/* add VID to filter table */
3416 	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true);
3417 	set_bit(vid, adapter->active_vlans);
3418 
3419 	return 0;
3420 }
3421 
3422 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev, u16 vid)
3423 {
3424 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3425 	struct ixgbe_hw *hw = &adapter->hw;
3426 
3427 	/* remove VID from filter table */
3428 	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), false);
3429 	clear_bit(vid, adapter->active_vlans);
3430 
3431 	return 0;
3432 }
3433 
3434 /**
3435  * ixgbe_vlan_filter_disable - helper to disable hw vlan filtering
3436  * @adapter: driver data
3437  */
3438 static void ixgbe_vlan_filter_disable(struct ixgbe_adapter *adapter)
3439 {
3440 	struct ixgbe_hw *hw = &adapter->hw;
3441 	u32 vlnctrl;
3442 
3443 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3444 	vlnctrl &= ~(IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
3445 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3446 }
3447 
3448 /**
3449  * ixgbe_vlan_filter_enable - helper to enable hw vlan filtering
3450  * @adapter: driver data
3451  */
3452 static void ixgbe_vlan_filter_enable(struct ixgbe_adapter *adapter)
3453 {
3454 	struct ixgbe_hw *hw = &adapter->hw;
3455 	u32 vlnctrl;
3456 
3457 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3458 	vlnctrl |= IXGBE_VLNCTRL_VFE;
3459 	vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
3460 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3461 }
3462 
3463 /**
3464  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
3465  * @adapter: driver data
3466  */
3467 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
3468 {
3469 	struct ixgbe_hw *hw = &adapter->hw;
3470 	u32 vlnctrl;
3471 	int i, j;
3472 
3473 	switch (hw->mac.type) {
3474 	case ixgbe_mac_82598EB:
3475 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3476 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
3477 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3478 		break;
3479 	case ixgbe_mac_82599EB:
3480 	case ixgbe_mac_X540:
3481 		for (i = 0; i < adapter->num_rx_queues; i++) {
3482 			j = adapter->rx_ring[i]->reg_idx;
3483 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3484 			vlnctrl &= ~IXGBE_RXDCTL_VME;
3485 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3486 		}
3487 		break;
3488 	default:
3489 		break;
3490 	}
3491 }
3492 
3493 /**
3494  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
3495  * @adapter: driver data
3496  */
3497 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
3498 {
3499 	struct ixgbe_hw *hw = &adapter->hw;
3500 	u32 vlnctrl;
3501 	int i, j;
3502 
3503 	switch (hw->mac.type) {
3504 	case ixgbe_mac_82598EB:
3505 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3506 		vlnctrl |= IXGBE_VLNCTRL_VME;
3507 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3508 		break;
3509 	case ixgbe_mac_82599EB:
3510 	case ixgbe_mac_X540:
3511 		for (i = 0; i < adapter->num_rx_queues; i++) {
3512 			j = adapter->rx_ring[i]->reg_idx;
3513 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3514 			vlnctrl |= IXGBE_RXDCTL_VME;
3515 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3516 		}
3517 		break;
3518 	default:
3519 		break;
3520 	}
3521 }
3522 
3523 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
3524 {
3525 	u16 vid;
3526 
3527 	ixgbe_vlan_rx_add_vid(adapter->netdev, 0);
3528 
3529 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
3530 		ixgbe_vlan_rx_add_vid(adapter->netdev, vid);
3531 }
3532 
3533 /**
3534  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
3535  * @netdev: network interface device structure
3536  *
3537  * Writes unicast address list to the RAR table.
3538  * Returns: -ENOMEM on failure/insufficient address space
3539  *                0 on no addresses written
3540  *                X on writing X addresses to the RAR table
3541  **/
3542 static int ixgbe_write_uc_addr_list(struct net_device *netdev)
3543 {
3544 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3545 	struct ixgbe_hw *hw = &adapter->hw;
3546 	unsigned int rar_entries = hw->mac.num_rar_entries - 1;
3547 	int count = 0;
3548 
3549 	/* In SR-IOV mode significantly less RAR entries are available */
3550 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3551 		rar_entries = IXGBE_MAX_PF_MACVLANS - 1;
3552 
3553 	/* return ENOMEM indicating insufficient memory for addresses */
3554 	if (netdev_uc_count(netdev) > rar_entries)
3555 		return -ENOMEM;
3556 
3557 	if (!netdev_uc_empty(netdev)) {
3558 		struct netdev_hw_addr *ha;
3559 		/* return error if we do not support writing to RAR table */
3560 		if (!hw->mac.ops.set_rar)
3561 			return -ENOMEM;
3562 
3563 		netdev_for_each_uc_addr(ha, netdev) {
3564 			if (!rar_entries)
3565 				break;
3566 			hw->mac.ops.set_rar(hw, rar_entries--, ha->addr,
3567 					    VMDQ_P(0), IXGBE_RAH_AV);
3568 			count++;
3569 		}
3570 	}
3571 	/* write the addresses in reverse order to avoid write combining */
3572 	for (; rar_entries > 0 ; rar_entries--)
3573 		hw->mac.ops.clear_rar(hw, rar_entries);
3574 
3575 	return count;
3576 }
3577 
3578 /**
3579  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
3580  * @netdev: network interface device structure
3581  *
3582  * The set_rx_method entry point is called whenever the unicast/multicast
3583  * address list or the network interface flags are updated.  This routine is
3584  * responsible for configuring the hardware for proper unicast, multicast and
3585  * promiscuous mode.
3586  **/
3587 void ixgbe_set_rx_mode(struct net_device *netdev)
3588 {
3589 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3590 	struct ixgbe_hw *hw = &adapter->hw;
3591 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
3592 	int count;
3593 
3594 	/* Check for Promiscuous and All Multicast modes */
3595 
3596 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3597 
3598 	/* set all bits that we expect to always be set */
3599 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
3600 	fctrl |= IXGBE_FCTRL_BAM;
3601 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
3602 	fctrl |= IXGBE_FCTRL_PMCF;
3603 
3604 	/* clear the bits we are changing the status of */
3605 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3606 
3607 	if (netdev->flags & IFF_PROMISC) {
3608 		hw->addr_ctrl.user_set_promisc = true;
3609 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3610 		vmolr |= (IXGBE_VMOLR_ROPE | IXGBE_VMOLR_MPE);
3611 		/* don't hardware filter vlans in promisc mode */
3612 		ixgbe_vlan_filter_disable(adapter);
3613 	} else {
3614 		if (netdev->flags & IFF_ALLMULTI) {
3615 			fctrl |= IXGBE_FCTRL_MPE;
3616 			vmolr |= IXGBE_VMOLR_MPE;
3617 		} else {
3618 			/*
3619 			 * Write addresses to the MTA, if the attempt fails
3620 			 * then we should just turn on promiscuous mode so
3621 			 * that we can at least receive multicast traffic
3622 			 */
3623 			hw->mac.ops.update_mc_addr_list(hw, netdev);
3624 			vmolr |= IXGBE_VMOLR_ROMPE;
3625 		}
3626 		ixgbe_vlan_filter_enable(adapter);
3627 		hw->addr_ctrl.user_set_promisc = false;
3628 	}
3629 
3630 	/*
3631 	 * Write addresses to available RAR registers, if there is not
3632 	 * sufficient space to store all the addresses then enable
3633 	 * unicast promiscuous mode
3634 	 */
3635 	count = ixgbe_write_uc_addr_list(netdev);
3636 	if (count < 0) {
3637 		fctrl |= IXGBE_FCTRL_UPE;
3638 		vmolr |= IXGBE_VMOLR_ROPE;
3639 	}
3640 
3641 	if (adapter->num_vfs)
3642 		ixgbe_restore_vf_multicasts(adapter);
3643 
3644 	if (hw->mac.type != ixgbe_mac_82598EB) {
3645 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
3646 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
3647 			   IXGBE_VMOLR_ROPE);
3648 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
3649 	}
3650 
3651 	/* This is useful for sniffing bad packets. */
3652 	if (adapter->netdev->features & NETIF_F_RXALL) {
3653 		/* UPE and MPE will be handled by normal PROMISC logic
3654 		 * in e1000e_set_rx_mode */
3655 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
3656 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
3657 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
3658 
3659 		fctrl &= ~(IXGBE_FCTRL_DPF);
3660 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
3661 	}
3662 
3663 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
3664 
3665 	if (netdev->features & NETIF_F_HW_VLAN_RX)
3666 		ixgbe_vlan_strip_enable(adapter);
3667 	else
3668 		ixgbe_vlan_strip_disable(adapter);
3669 }
3670 
3671 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
3672 {
3673 	int q_idx;
3674 
3675 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
3676 		napi_enable(&adapter->q_vector[q_idx]->napi);
3677 }
3678 
3679 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
3680 {
3681 	int q_idx;
3682 
3683 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
3684 		napi_disable(&adapter->q_vector[q_idx]->napi);
3685 }
3686 
3687 #ifdef CONFIG_IXGBE_DCB
3688 /**
3689  * ixgbe_configure_dcb - Configure DCB hardware
3690  * @adapter: ixgbe adapter struct
3691  *
3692  * This is called by the driver on open to configure the DCB hardware.
3693  * This is also called by the gennetlink interface when reconfiguring
3694  * the DCB state.
3695  */
3696 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
3697 {
3698 	struct ixgbe_hw *hw = &adapter->hw;
3699 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3700 
3701 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
3702 		if (hw->mac.type == ixgbe_mac_82598EB)
3703 			netif_set_gso_max_size(adapter->netdev, 65536);
3704 		return;
3705 	}
3706 
3707 	if (hw->mac.type == ixgbe_mac_82598EB)
3708 		netif_set_gso_max_size(adapter->netdev, 32768);
3709 
3710 #ifdef IXGBE_FCOE
3711 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
3712 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
3713 #endif
3714 
3715 	/* reconfigure the hardware */
3716 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
3717 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
3718 						DCB_TX_CONFIG);
3719 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
3720 						DCB_RX_CONFIG);
3721 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
3722 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
3723 		ixgbe_dcb_hw_ets(&adapter->hw,
3724 				 adapter->ixgbe_ieee_ets,
3725 				 max_frame);
3726 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
3727 					adapter->ixgbe_ieee_pfc->pfc_en,
3728 					adapter->ixgbe_ieee_ets->prio_tc);
3729 	}
3730 
3731 	/* Enable RSS Hash per TC */
3732 	if (hw->mac.type != ixgbe_mac_82598EB) {
3733 		u32 msb = 0;
3734 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
3735 
3736 		while (rss_i) {
3737 			msb++;
3738 			rss_i >>= 1;
3739 		}
3740 
3741 		/* write msb to all 8 TCs in one write */
3742 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
3743 	}
3744 }
3745 #endif
3746 
3747 /* Additional bittime to account for IXGBE framing */
3748 #define IXGBE_ETH_FRAMING 20
3749 
3750 /**
3751  * ixgbe_hpbthresh - calculate high water mark for flow control
3752  *
3753  * @adapter: board private structure to calculate for
3754  * @pb: packet buffer to calculate
3755  */
3756 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
3757 {
3758 	struct ixgbe_hw *hw = &adapter->hw;
3759 	struct net_device *dev = adapter->netdev;
3760 	int link, tc, kb, marker;
3761 	u32 dv_id, rx_pba;
3762 
3763 	/* Calculate max LAN frame size */
3764 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
3765 
3766 #ifdef IXGBE_FCOE
3767 	/* FCoE traffic class uses FCOE jumbo frames */
3768 	if ((dev->features & NETIF_F_FCOE_MTU) &&
3769 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
3770 	    (pb == ixgbe_fcoe_get_tc(adapter)))
3771 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3772 
3773 #endif
3774 	/* Calculate delay value for device */
3775 	switch (hw->mac.type) {
3776 	case ixgbe_mac_X540:
3777 		dv_id = IXGBE_DV_X540(link, tc);
3778 		break;
3779 	default:
3780 		dv_id = IXGBE_DV(link, tc);
3781 		break;
3782 	}
3783 
3784 	/* Loopback switch introduces additional latency */
3785 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3786 		dv_id += IXGBE_B2BT(tc);
3787 
3788 	/* Delay value is calculated in bit times convert to KB */
3789 	kb = IXGBE_BT2KB(dv_id);
3790 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
3791 
3792 	marker = rx_pba - kb;
3793 
3794 	/* It is possible that the packet buffer is not large enough
3795 	 * to provide required headroom. In this case throw an error
3796 	 * to user and a do the best we can.
3797 	 */
3798 	if (marker < 0) {
3799 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
3800 			    "headroom to support flow control."
3801 			    "Decrease MTU or number of traffic classes\n", pb);
3802 		marker = tc + 1;
3803 	}
3804 
3805 	return marker;
3806 }
3807 
3808 /**
3809  * ixgbe_lpbthresh - calculate low water mark for for flow control
3810  *
3811  * @adapter: board private structure to calculate for
3812  * @pb: packet buffer to calculate
3813  */
3814 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter)
3815 {
3816 	struct ixgbe_hw *hw = &adapter->hw;
3817 	struct net_device *dev = adapter->netdev;
3818 	int tc;
3819 	u32 dv_id;
3820 
3821 	/* Calculate max LAN frame size */
3822 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
3823 
3824 	/* Calculate delay value for device */
3825 	switch (hw->mac.type) {
3826 	case ixgbe_mac_X540:
3827 		dv_id = IXGBE_LOW_DV_X540(tc);
3828 		break;
3829 	default:
3830 		dv_id = IXGBE_LOW_DV(tc);
3831 		break;
3832 	}
3833 
3834 	/* Delay value is calculated in bit times convert to KB */
3835 	return IXGBE_BT2KB(dv_id);
3836 }
3837 
3838 /*
3839  * ixgbe_pbthresh_setup - calculate and setup high low water marks
3840  */
3841 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
3842 {
3843 	struct ixgbe_hw *hw = &adapter->hw;
3844 	int num_tc = netdev_get_num_tc(adapter->netdev);
3845 	int i;
3846 
3847 	if (!num_tc)
3848 		num_tc = 1;
3849 
3850 	hw->fc.low_water = ixgbe_lpbthresh(adapter);
3851 
3852 	for (i = 0; i < num_tc; i++) {
3853 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
3854 
3855 		/* Low water marks must not be larger than high water marks */
3856 		if (hw->fc.low_water > hw->fc.high_water[i])
3857 			hw->fc.low_water = 0;
3858 	}
3859 }
3860 
3861 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
3862 {
3863 	struct ixgbe_hw *hw = &adapter->hw;
3864 	int hdrm;
3865 	u8 tc = netdev_get_num_tc(adapter->netdev);
3866 
3867 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
3868 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
3869 		hdrm = 32 << adapter->fdir_pballoc;
3870 	else
3871 		hdrm = 0;
3872 
3873 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
3874 	ixgbe_pbthresh_setup(adapter);
3875 }
3876 
3877 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
3878 {
3879 	struct ixgbe_hw *hw = &adapter->hw;
3880 	struct hlist_node *node, *node2;
3881 	struct ixgbe_fdir_filter *filter;
3882 
3883 	spin_lock(&adapter->fdir_perfect_lock);
3884 
3885 	if (!hlist_empty(&adapter->fdir_filter_list))
3886 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
3887 
3888 	hlist_for_each_entry_safe(filter, node, node2,
3889 				  &adapter->fdir_filter_list, fdir_node) {
3890 		ixgbe_fdir_write_perfect_filter_82599(hw,
3891 				&filter->filter,
3892 				filter->sw_idx,
3893 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
3894 				IXGBE_FDIR_DROP_QUEUE :
3895 				adapter->rx_ring[filter->action]->reg_idx);
3896 	}
3897 
3898 	spin_unlock(&adapter->fdir_perfect_lock);
3899 }
3900 
3901 static void ixgbe_configure(struct ixgbe_adapter *adapter)
3902 {
3903 	struct ixgbe_hw *hw = &adapter->hw;
3904 
3905 	ixgbe_configure_pb(adapter);
3906 #ifdef CONFIG_IXGBE_DCB
3907 	ixgbe_configure_dcb(adapter);
3908 #endif
3909 	/*
3910 	 * We must restore virtualization before VLANs or else
3911 	 * the VLVF registers will not be populated
3912 	 */
3913 	ixgbe_configure_virtualization(adapter);
3914 
3915 	ixgbe_set_rx_mode(adapter->netdev);
3916 	ixgbe_restore_vlan(adapter);
3917 
3918 	switch (hw->mac.type) {
3919 	case ixgbe_mac_82599EB:
3920 	case ixgbe_mac_X540:
3921 		hw->mac.ops.disable_rx_buff(hw);
3922 		break;
3923 	default:
3924 		break;
3925 	}
3926 
3927 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3928 		ixgbe_init_fdir_signature_82599(&adapter->hw,
3929 						adapter->fdir_pballoc);
3930 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
3931 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
3932 					      adapter->fdir_pballoc);
3933 		ixgbe_fdir_filter_restore(adapter);
3934 	}
3935 
3936 	switch (hw->mac.type) {
3937 	case ixgbe_mac_82599EB:
3938 	case ixgbe_mac_X540:
3939 		hw->mac.ops.enable_rx_buff(hw);
3940 		break;
3941 	default:
3942 		break;
3943 	}
3944 
3945 #ifdef IXGBE_FCOE
3946 	/* configure FCoE L2 filters, redirection table, and Rx control */
3947 	ixgbe_configure_fcoe(adapter);
3948 
3949 #endif /* IXGBE_FCOE */
3950 	ixgbe_configure_tx(adapter);
3951 	ixgbe_configure_rx(adapter);
3952 }
3953 
3954 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
3955 {
3956 	switch (hw->phy.type) {
3957 	case ixgbe_phy_sfp_avago:
3958 	case ixgbe_phy_sfp_ftl:
3959 	case ixgbe_phy_sfp_intel:
3960 	case ixgbe_phy_sfp_unknown:
3961 	case ixgbe_phy_sfp_passive_tyco:
3962 	case ixgbe_phy_sfp_passive_unknown:
3963 	case ixgbe_phy_sfp_active_unknown:
3964 	case ixgbe_phy_sfp_ftl_active:
3965 		return true;
3966 	case ixgbe_phy_nl:
3967 		if (hw->mac.type == ixgbe_mac_82598EB)
3968 			return true;
3969 	default:
3970 		return false;
3971 	}
3972 }
3973 
3974 /**
3975  * ixgbe_sfp_link_config - set up SFP+ link
3976  * @adapter: pointer to private adapter struct
3977  **/
3978 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
3979 {
3980 	/*
3981 	 * We are assuming the worst case scenario here, and that
3982 	 * is that an SFP was inserted/removed after the reset
3983 	 * but before SFP detection was enabled.  As such the best
3984 	 * solution is to just start searching as soon as we start
3985 	 */
3986 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3987 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
3988 
3989 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
3990 }
3991 
3992 /**
3993  * ixgbe_non_sfp_link_config - set up non-SFP+ link
3994  * @hw: pointer to private hardware struct
3995  *
3996  * Returns 0 on success, negative on failure
3997  **/
3998 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
3999 {
4000 	u32 autoneg;
4001 	bool negotiation, link_up = false;
4002 	u32 ret = IXGBE_ERR_LINK_SETUP;
4003 
4004 	if (hw->mac.ops.check_link)
4005 		ret = hw->mac.ops.check_link(hw, &autoneg, &link_up, false);
4006 
4007 	if (ret)
4008 		goto link_cfg_out;
4009 
4010 	autoneg = hw->phy.autoneg_advertised;
4011 	if ((!autoneg) && (hw->mac.ops.get_link_capabilities))
4012 		ret = hw->mac.ops.get_link_capabilities(hw, &autoneg,
4013 							&negotiation);
4014 	if (ret)
4015 		goto link_cfg_out;
4016 
4017 	if (hw->mac.ops.setup_link)
4018 		ret = hw->mac.ops.setup_link(hw, autoneg, negotiation, link_up);
4019 link_cfg_out:
4020 	return ret;
4021 }
4022 
4023 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
4024 {
4025 	struct ixgbe_hw *hw = &adapter->hw;
4026 	u32 gpie = 0;
4027 
4028 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
4029 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
4030 		       IXGBE_GPIE_OCD;
4031 		gpie |= IXGBE_GPIE_EIAME;
4032 		/*
4033 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
4034 		 * this saves a register write for every interrupt
4035 		 */
4036 		switch (hw->mac.type) {
4037 		case ixgbe_mac_82598EB:
4038 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4039 			break;
4040 		case ixgbe_mac_82599EB:
4041 		case ixgbe_mac_X540:
4042 		default:
4043 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
4044 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
4045 			break;
4046 		}
4047 	} else {
4048 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
4049 		 * specifically only auto mask tx and rx interrupts */
4050 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4051 	}
4052 
4053 	/* XXX: to interrupt immediately for EICS writes, enable this */
4054 	/* gpie |= IXGBE_GPIE_EIMEN; */
4055 
4056 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
4057 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
4058 
4059 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4060 		case IXGBE_82599_VMDQ_8Q_MASK:
4061 			gpie |= IXGBE_GPIE_VTMODE_16;
4062 			break;
4063 		case IXGBE_82599_VMDQ_4Q_MASK:
4064 			gpie |= IXGBE_GPIE_VTMODE_32;
4065 			break;
4066 		default:
4067 			gpie |= IXGBE_GPIE_VTMODE_64;
4068 			break;
4069 		}
4070 	}
4071 
4072 	/* Enable Thermal over heat sensor interrupt */
4073 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
4074 		switch (adapter->hw.mac.type) {
4075 		case ixgbe_mac_82599EB:
4076 			gpie |= IXGBE_SDP0_GPIEN;
4077 			break;
4078 		case ixgbe_mac_X540:
4079 			gpie |= IXGBE_EIMS_TS;
4080 			break;
4081 		default:
4082 			break;
4083 		}
4084 	}
4085 
4086 	/* Enable fan failure interrupt */
4087 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
4088 		gpie |= IXGBE_SDP1_GPIEN;
4089 
4090 	if (hw->mac.type == ixgbe_mac_82599EB) {
4091 		gpie |= IXGBE_SDP1_GPIEN;
4092 		gpie |= IXGBE_SDP2_GPIEN;
4093 	}
4094 
4095 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
4096 }
4097 
4098 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
4099 {
4100 	struct ixgbe_hw *hw = &adapter->hw;
4101 	int err;
4102 	u32 ctrl_ext;
4103 
4104 	ixgbe_get_hw_control(adapter);
4105 	ixgbe_setup_gpie(adapter);
4106 
4107 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
4108 		ixgbe_configure_msix(adapter);
4109 	else
4110 		ixgbe_configure_msi_and_legacy(adapter);
4111 
4112 	/* enable the optics for 82599 SFP+ fiber */
4113 	if (hw->mac.ops.enable_tx_laser)
4114 		hw->mac.ops.enable_tx_laser(hw);
4115 
4116 	clear_bit(__IXGBE_DOWN, &adapter->state);
4117 	ixgbe_napi_enable_all(adapter);
4118 
4119 	if (ixgbe_is_sfp(hw)) {
4120 		ixgbe_sfp_link_config(adapter);
4121 	} else {
4122 		err = ixgbe_non_sfp_link_config(hw);
4123 		if (err)
4124 			e_err(probe, "link_config FAILED %d\n", err);
4125 	}
4126 
4127 	/* clear any pending interrupts, may auto mask */
4128 	IXGBE_READ_REG(hw, IXGBE_EICR);
4129 	ixgbe_irq_enable(adapter, true, true);
4130 
4131 	/*
4132 	 * If this adapter has a fan, check to see if we had a failure
4133 	 * before we enabled the interrupt.
4134 	 */
4135 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
4136 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
4137 		if (esdp & IXGBE_ESDP_SDP1)
4138 			e_crit(drv, "Fan has stopped, replace the adapter\n");
4139 	}
4140 
4141 	/* enable transmits */
4142 	netif_tx_start_all_queues(adapter->netdev);
4143 
4144 	/* bring the link up in the watchdog, this could race with our first
4145 	 * link up interrupt but shouldn't be a problem */
4146 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4147 	adapter->link_check_timeout = jiffies;
4148 	mod_timer(&adapter->service_timer, jiffies);
4149 
4150 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
4151 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
4152 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
4153 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
4154 }
4155 
4156 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
4157 {
4158 	WARN_ON(in_interrupt());
4159 	/* put off any impending NetWatchDogTimeout */
4160 	adapter->netdev->trans_start = jiffies;
4161 
4162 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
4163 		usleep_range(1000, 2000);
4164 	ixgbe_down(adapter);
4165 	/*
4166 	 * If SR-IOV enabled then wait a bit before bringing the adapter
4167 	 * back up to give the VFs time to respond to the reset.  The
4168 	 * two second wait is based upon the watchdog timer cycle in
4169 	 * the VF driver.
4170 	 */
4171 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4172 		msleep(2000);
4173 	ixgbe_up(adapter);
4174 	clear_bit(__IXGBE_RESETTING, &adapter->state);
4175 }
4176 
4177 void ixgbe_up(struct ixgbe_adapter *adapter)
4178 {
4179 	/* hardware has been reset, we need to reload some things */
4180 	ixgbe_configure(adapter);
4181 
4182 	ixgbe_up_complete(adapter);
4183 }
4184 
4185 void ixgbe_reset(struct ixgbe_adapter *adapter)
4186 {
4187 	struct ixgbe_hw *hw = &adapter->hw;
4188 	int err;
4189 
4190 	/* lock SFP init bit to prevent race conditions with the watchdog */
4191 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
4192 		usleep_range(1000, 2000);
4193 
4194 	/* clear all SFP and link config related flags while holding SFP_INIT */
4195 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
4196 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
4197 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4198 
4199 	err = hw->mac.ops.init_hw(hw);
4200 	switch (err) {
4201 	case 0:
4202 	case IXGBE_ERR_SFP_NOT_PRESENT:
4203 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
4204 		break;
4205 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
4206 		e_dev_err("master disable timed out\n");
4207 		break;
4208 	case IXGBE_ERR_EEPROM_VERSION:
4209 		/* We are running on a pre-production device, log a warning */
4210 		e_dev_warn("This device is a pre-production adapter/LOM. "
4211 			   "Please be aware there may be issues associated with "
4212 			   "your hardware.  If you are experiencing problems "
4213 			   "please contact your Intel or hardware "
4214 			   "representative who provided you with this "
4215 			   "hardware.\n");
4216 		break;
4217 	default:
4218 		e_dev_err("Hardware Error: %d\n", err);
4219 	}
4220 
4221 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
4222 
4223 	/* reprogram the RAR[0] in case user changed it. */
4224 	hw->mac.ops.set_rar(hw, 0, hw->mac.addr, VMDQ_P(0), IXGBE_RAH_AV);
4225 
4226 	/* update SAN MAC vmdq pool selection */
4227 	if (hw->mac.san_mac_rar_index)
4228 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
4229 
4230 	if (adapter->flags2 & IXGBE_FLAG2_PTP_ENABLED)
4231 		ixgbe_ptp_reset(adapter);
4232 }
4233 
4234 /**
4235  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4236  * @rx_ring: ring to free buffers from
4237  **/
4238 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
4239 {
4240 	struct device *dev = rx_ring->dev;
4241 	unsigned long size;
4242 	u16 i;
4243 
4244 	/* ring already cleared, nothing to do */
4245 	if (!rx_ring->rx_buffer_info)
4246 		return;
4247 
4248 	/* Free all the Rx ring sk_buffs */
4249 	for (i = 0; i < rx_ring->count; i++) {
4250 		struct ixgbe_rx_buffer *rx_buffer;
4251 
4252 		rx_buffer = &rx_ring->rx_buffer_info[i];
4253 		if (rx_buffer->skb) {
4254 			struct sk_buff *skb = rx_buffer->skb;
4255 			if (IXGBE_CB(skb)->page_released) {
4256 				dma_unmap_page(dev,
4257 					       IXGBE_CB(skb)->dma,
4258 					       ixgbe_rx_bufsz(rx_ring),
4259 					       DMA_FROM_DEVICE);
4260 				IXGBE_CB(skb)->page_released = false;
4261 			}
4262 			dev_kfree_skb(skb);
4263 		}
4264 		rx_buffer->skb = NULL;
4265 		if (rx_buffer->dma)
4266 			dma_unmap_page(dev, rx_buffer->dma,
4267 				       ixgbe_rx_pg_size(rx_ring),
4268 				       DMA_FROM_DEVICE);
4269 		rx_buffer->dma = 0;
4270 		if (rx_buffer->page)
4271 			__free_pages(rx_buffer->page,
4272 				     ixgbe_rx_pg_order(rx_ring));
4273 		rx_buffer->page = NULL;
4274 	}
4275 
4276 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4277 	memset(rx_ring->rx_buffer_info, 0, size);
4278 
4279 	/* Zero out the descriptor ring */
4280 	memset(rx_ring->desc, 0, rx_ring->size);
4281 
4282 	rx_ring->next_to_alloc = 0;
4283 	rx_ring->next_to_clean = 0;
4284 	rx_ring->next_to_use = 0;
4285 }
4286 
4287 /**
4288  * ixgbe_clean_tx_ring - Free Tx Buffers
4289  * @tx_ring: ring to be cleaned
4290  **/
4291 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
4292 {
4293 	struct ixgbe_tx_buffer *tx_buffer_info;
4294 	unsigned long size;
4295 	u16 i;
4296 
4297 	/* ring already cleared, nothing to do */
4298 	if (!tx_ring->tx_buffer_info)
4299 		return;
4300 
4301 	/* Free all the Tx ring sk_buffs */
4302 	for (i = 0; i < tx_ring->count; i++) {
4303 		tx_buffer_info = &tx_ring->tx_buffer_info[i];
4304 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer_info);
4305 	}
4306 
4307 	netdev_tx_reset_queue(txring_txq(tx_ring));
4308 
4309 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
4310 	memset(tx_ring->tx_buffer_info, 0, size);
4311 
4312 	/* Zero out the descriptor ring */
4313 	memset(tx_ring->desc, 0, tx_ring->size);
4314 
4315 	tx_ring->next_to_use = 0;
4316 	tx_ring->next_to_clean = 0;
4317 }
4318 
4319 /**
4320  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
4321  * @adapter: board private structure
4322  **/
4323 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
4324 {
4325 	int i;
4326 
4327 	for (i = 0; i < adapter->num_rx_queues; i++)
4328 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
4329 }
4330 
4331 /**
4332  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
4333  * @adapter: board private structure
4334  **/
4335 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
4336 {
4337 	int i;
4338 
4339 	for (i = 0; i < adapter->num_tx_queues; i++)
4340 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
4341 }
4342 
4343 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
4344 {
4345 	struct hlist_node *node, *node2;
4346 	struct ixgbe_fdir_filter *filter;
4347 
4348 	spin_lock(&adapter->fdir_perfect_lock);
4349 
4350 	hlist_for_each_entry_safe(filter, node, node2,
4351 				  &adapter->fdir_filter_list, fdir_node) {
4352 		hlist_del(&filter->fdir_node);
4353 		kfree(filter);
4354 	}
4355 	adapter->fdir_filter_count = 0;
4356 
4357 	spin_unlock(&adapter->fdir_perfect_lock);
4358 }
4359 
4360 void ixgbe_down(struct ixgbe_adapter *adapter)
4361 {
4362 	struct net_device *netdev = adapter->netdev;
4363 	struct ixgbe_hw *hw = &adapter->hw;
4364 	u32 rxctrl;
4365 	int i;
4366 
4367 	/* signal that we are down to the interrupt handler */
4368 	set_bit(__IXGBE_DOWN, &adapter->state);
4369 
4370 	/* disable receives */
4371 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4372 	IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxctrl & ~IXGBE_RXCTRL_RXEN);
4373 
4374 	/* disable all enabled rx queues */
4375 	for (i = 0; i < adapter->num_rx_queues; i++)
4376 		/* this call also flushes the previous write */
4377 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
4378 
4379 	usleep_range(10000, 20000);
4380 
4381 	netif_tx_stop_all_queues(netdev);
4382 
4383 	/* call carrier off first to avoid false dev_watchdog timeouts */
4384 	netif_carrier_off(netdev);
4385 	netif_tx_disable(netdev);
4386 
4387 	ixgbe_irq_disable(adapter);
4388 
4389 	ixgbe_napi_disable_all(adapter);
4390 
4391 	adapter->flags2 &= ~(IXGBE_FLAG2_FDIR_REQUIRES_REINIT |
4392 			     IXGBE_FLAG2_RESET_REQUESTED);
4393 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4394 
4395 	del_timer_sync(&adapter->service_timer);
4396 
4397 	if (adapter->num_vfs) {
4398 		/* Clear EITR Select mapping */
4399 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
4400 
4401 		/* Mark all the VFs as inactive */
4402 		for (i = 0 ; i < adapter->num_vfs; i++)
4403 			adapter->vfinfo[i].clear_to_send = false;
4404 
4405 		/* ping all the active vfs to let them know we are going down */
4406 		ixgbe_ping_all_vfs(adapter);
4407 
4408 		/* Disable all VFTE/VFRE TX/RX */
4409 		ixgbe_disable_tx_rx(adapter);
4410 	}
4411 
4412 	/* disable transmits in the hardware now that interrupts are off */
4413 	for (i = 0; i < adapter->num_tx_queues; i++) {
4414 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
4415 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
4416 	}
4417 
4418 	/* Disable the Tx DMA engine on 82599 and X540 */
4419 	switch (hw->mac.type) {
4420 	case ixgbe_mac_82599EB:
4421 	case ixgbe_mac_X540:
4422 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
4423 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
4424 				 ~IXGBE_DMATXCTL_TE));
4425 		break;
4426 	default:
4427 		break;
4428 	}
4429 
4430 	if (!pci_channel_offline(adapter->pdev))
4431 		ixgbe_reset(adapter);
4432 
4433 	/* power down the optics for 82599 SFP+ fiber */
4434 	if (hw->mac.ops.disable_tx_laser)
4435 		hw->mac.ops.disable_tx_laser(hw);
4436 
4437 	ixgbe_clean_all_tx_rings(adapter);
4438 	ixgbe_clean_all_rx_rings(adapter);
4439 
4440 #ifdef CONFIG_IXGBE_DCA
4441 	/* since we reset the hardware DCA settings were cleared */
4442 	ixgbe_setup_dca(adapter);
4443 #endif
4444 }
4445 
4446 /**
4447  * ixgbe_tx_timeout - Respond to a Tx Hang
4448  * @netdev: network interface device structure
4449  **/
4450 static void ixgbe_tx_timeout(struct net_device *netdev)
4451 {
4452 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4453 
4454 	/* Do the reset outside of interrupt context */
4455 	ixgbe_tx_timeout_reset(adapter);
4456 }
4457 
4458 /**
4459  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
4460  * @adapter: board private structure to initialize
4461  *
4462  * ixgbe_sw_init initializes the Adapter private data structure.
4463  * Fields are initialized based on PCI device information and
4464  * OS network device settings (MTU size).
4465  **/
4466 static int ixgbe_sw_init(struct ixgbe_adapter *adapter)
4467 {
4468 	struct ixgbe_hw *hw = &adapter->hw;
4469 	struct pci_dev *pdev = adapter->pdev;
4470 	unsigned int rss;
4471 	u32 fwsm;
4472 #ifdef CONFIG_IXGBE_DCB
4473 	int j;
4474 	struct tc_configuration *tc;
4475 #endif
4476 
4477 	/* PCI config space info */
4478 
4479 	hw->vendor_id = pdev->vendor;
4480 	hw->device_id = pdev->device;
4481 	hw->revision_id = pdev->revision;
4482 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
4483 	hw->subsystem_device_id = pdev->subsystem_device;
4484 
4485 	/* Set capability flags */
4486 	rss = min_t(int, IXGBE_MAX_RSS_INDICES, num_online_cpus());
4487 	adapter->ring_feature[RING_F_RSS].limit = rss;
4488 	switch (hw->mac.type) {
4489 	case ixgbe_mac_82598EB:
4490 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
4491 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
4492 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
4493 		break;
4494 	case ixgbe_mac_X540:
4495 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM);
4496 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
4497 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
4498 	case ixgbe_mac_82599EB:
4499 		adapter->max_q_vectors = MAX_Q_VECTORS_82599;
4500 		adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
4501 		adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
4502 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
4503 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
4504 		/* Flow Director hash filters enabled */
4505 		adapter->atr_sample_rate = 20;
4506 		adapter->ring_feature[RING_F_FDIR].limit =
4507 							 IXGBE_MAX_FDIR_INDICES;
4508 		adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
4509 #ifdef IXGBE_FCOE
4510 		adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
4511 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
4512 #ifdef CONFIG_IXGBE_DCB
4513 		/* Default traffic class to use for FCoE */
4514 		adapter->fcoe.up = IXGBE_FCOE_DEFTC;
4515 #endif
4516 #endif /* IXGBE_FCOE */
4517 		break;
4518 	default:
4519 		break;
4520 	}
4521 
4522 #ifdef IXGBE_FCOE
4523 	/* FCoE support exists, always init the FCoE lock */
4524 	spin_lock_init(&adapter->fcoe.lock);
4525 
4526 #endif
4527 	/* n-tuple support exists, always init our spinlock */
4528 	spin_lock_init(&adapter->fdir_perfect_lock);
4529 
4530 #ifdef CONFIG_IXGBE_DCB
4531 	switch (hw->mac.type) {
4532 	case ixgbe_mac_X540:
4533 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
4534 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
4535 		break;
4536 	default:
4537 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
4538 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
4539 		break;
4540 	}
4541 
4542 	/* Configure DCB traffic classes */
4543 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
4544 		tc = &adapter->dcb_cfg.tc_config[j];
4545 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
4546 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
4547 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
4548 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
4549 		tc->dcb_pfc = pfc_disabled;
4550 	}
4551 
4552 	/* Initialize default user to priority mapping, UPx->TC0 */
4553 	tc = &adapter->dcb_cfg.tc_config[0];
4554 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
4555 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
4556 
4557 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
4558 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
4559 	adapter->dcb_cfg.pfc_mode_enable = false;
4560 	adapter->dcb_set_bitmap = 0x00;
4561 	adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
4562 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
4563 	       sizeof(adapter->temp_dcb_cfg));
4564 
4565 #endif
4566 
4567 	/* default flow control settings */
4568 	hw->fc.requested_mode = ixgbe_fc_full;
4569 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
4570 	ixgbe_pbthresh_setup(adapter);
4571 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
4572 	hw->fc.send_xon = true;
4573 	hw->fc.disable_fc_autoneg =
4574 		(ixgbe_device_supports_autoneg_fc(hw) == 0) ? false : true;
4575 
4576 #ifdef CONFIG_PCI_IOV
4577 	/* assign number of SR-IOV VFs */
4578 	if (hw->mac.type != ixgbe_mac_82598EB)
4579 		adapter->num_vfs = (max_vfs > 63) ? 0 : max_vfs;
4580 
4581 #endif
4582 	/* enable itr by default in dynamic mode */
4583 	adapter->rx_itr_setting = 1;
4584 	adapter->tx_itr_setting = 1;
4585 
4586 	/* set default ring sizes */
4587 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
4588 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
4589 
4590 	/* set default work limits */
4591 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
4592 
4593 	/* initialize eeprom parameters */
4594 	if (ixgbe_init_eeprom_params_generic(hw)) {
4595 		e_dev_err("EEPROM initialization failed\n");
4596 		return -EIO;
4597 	}
4598 
4599 	set_bit(__IXGBE_DOWN, &adapter->state);
4600 
4601 	return 0;
4602 }
4603 
4604 /**
4605  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
4606  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
4607  *
4608  * Return 0 on success, negative on failure
4609  **/
4610 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
4611 {
4612 	struct device *dev = tx_ring->dev;
4613 	int orig_node = dev_to_node(dev);
4614 	int numa_node = -1;
4615 	int size;
4616 
4617 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
4618 
4619 	if (tx_ring->q_vector)
4620 		numa_node = tx_ring->q_vector->numa_node;
4621 
4622 	tx_ring->tx_buffer_info = vzalloc_node(size, numa_node);
4623 	if (!tx_ring->tx_buffer_info)
4624 		tx_ring->tx_buffer_info = vzalloc(size);
4625 	if (!tx_ring->tx_buffer_info)
4626 		goto err;
4627 
4628 	/* round up to nearest 4K */
4629 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
4630 	tx_ring->size = ALIGN(tx_ring->size, 4096);
4631 
4632 	set_dev_node(dev, numa_node);
4633 	tx_ring->desc = dma_alloc_coherent(dev,
4634 					   tx_ring->size,
4635 					   &tx_ring->dma,
4636 					   GFP_KERNEL);
4637 	set_dev_node(dev, orig_node);
4638 	if (!tx_ring->desc)
4639 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
4640 						   &tx_ring->dma, GFP_KERNEL);
4641 	if (!tx_ring->desc)
4642 		goto err;
4643 
4644 	tx_ring->next_to_use = 0;
4645 	tx_ring->next_to_clean = 0;
4646 	return 0;
4647 
4648 err:
4649 	vfree(tx_ring->tx_buffer_info);
4650 	tx_ring->tx_buffer_info = NULL;
4651 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
4652 	return -ENOMEM;
4653 }
4654 
4655 /**
4656  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
4657  * @adapter: board private structure
4658  *
4659  * If this function returns with an error, then it's possible one or
4660  * more of the rings is populated (while the rest are not).  It is the
4661  * callers duty to clean those orphaned rings.
4662  *
4663  * Return 0 on success, negative on failure
4664  **/
4665 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
4666 {
4667 	int i, err = 0;
4668 
4669 	for (i = 0; i < adapter->num_tx_queues; i++) {
4670 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
4671 		if (!err)
4672 			continue;
4673 
4674 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
4675 		goto err_setup_tx;
4676 	}
4677 
4678 	return 0;
4679 err_setup_tx:
4680 	/* rewind the index freeing the rings as we go */
4681 	while (i--)
4682 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
4683 	return err;
4684 }
4685 
4686 /**
4687  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
4688  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
4689  *
4690  * Returns 0 on success, negative on failure
4691  **/
4692 int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
4693 {
4694 	struct device *dev = rx_ring->dev;
4695 	int orig_node = dev_to_node(dev);
4696 	int numa_node = -1;
4697 	int size;
4698 
4699 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4700 
4701 	if (rx_ring->q_vector)
4702 		numa_node = rx_ring->q_vector->numa_node;
4703 
4704 	rx_ring->rx_buffer_info = vzalloc_node(size, numa_node);
4705 	if (!rx_ring->rx_buffer_info)
4706 		rx_ring->rx_buffer_info = vzalloc(size);
4707 	if (!rx_ring->rx_buffer_info)
4708 		goto err;
4709 
4710 	/* Round up to nearest 4K */
4711 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
4712 	rx_ring->size = ALIGN(rx_ring->size, 4096);
4713 
4714 	set_dev_node(dev, numa_node);
4715 	rx_ring->desc = dma_alloc_coherent(dev,
4716 					   rx_ring->size,
4717 					   &rx_ring->dma,
4718 					   GFP_KERNEL);
4719 	set_dev_node(dev, orig_node);
4720 	if (!rx_ring->desc)
4721 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
4722 						   &rx_ring->dma, GFP_KERNEL);
4723 	if (!rx_ring->desc)
4724 		goto err;
4725 
4726 	rx_ring->next_to_clean = 0;
4727 	rx_ring->next_to_use = 0;
4728 
4729 	return 0;
4730 err:
4731 	vfree(rx_ring->rx_buffer_info);
4732 	rx_ring->rx_buffer_info = NULL;
4733 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
4734 	return -ENOMEM;
4735 }
4736 
4737 /**
4738  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
4739  * @adapter: board private structure
4740  *
4741  * If this function returns with an error, then it's possible one or
4742  * more of the rings is populated (while the rest are not).  It is the
4743  * callers duty to clean those orphaned rings.
4744  *
4745  * Return 0 on success, negative on failure
4746  **/
4747 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
4748 {
4749 	int i, err = 0;
4750 
4751 	for (i = 0; i < adapter->num_rx_queues; i++) {
4752 		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
4753 		if (!err)
4754 			continue;
4755 
4756 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
4757 		goto err_setup_rx;
4758 	}
4759 
4760 #ifdef IXGBE_FCOE
4761 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
4762 	if (!err)
4763 #endif
4764 		return 0;
4765 err_setup_rx:
4766 	/* rewind the index freeing the rings as we go */
4767 	while (i--)
4768 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
4769 	return err;
4770 }
4771 
4772 /**
4773  * ixgbe_free_tx_resources - Free Tx Resources per Queue
4774  * @tx_ring: Tx descriptor ring for a specific queue
4775  *
4776  * Free all transmit software resources
4777  **/
4778 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
4779 {
4780 	ixgbe_clean_tx_ring(tx_ring);
4781 
4782 	vfree(tx_ring->tx_buffer_info);
4783 	tx_ring->tx_buffer_info = NULL;
4784 
4785 	/* if not set, then don't free */
4786 	if (!tx_ring->desc)
4787 		return;
4788 
4789 	dma_free_coherent(tx_ring->dev, tx_ring->size,
4790 			  tx_ring->desc, tx_ring->dma);
4791 
4792 	tx_ring->desc = NULL;
4793 }
4794 
4795 /**
4796  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
4797  * @adapter: board private structure
4798  *
4799  * Free all transmit software resources
4800  **/
4801 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
4802 {
4803 	int i;
4804 
4805 	for (i = 0; i < adapter->num_tx_queues; i++)
4806 		if (adapter->tx_ring[i]->desc)
4807 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
4808 }
4809 
4810 /**
4811  * ixgbe_free_rx_resources - Free Rx Resources
4812  * @rx_ring: ring to clean the resources from
4813  *
4814  * Free all receive software resources
4815  **/
4816 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
4817 {
4818 	ixgbe_clean_rx_ring(rx_ring);
4819 
4820 	vfree(rx_ring->rx_buffer_info);
4821 	rx_ring->rx_buffer_info = NULL;
4822 
4823 	/* if not set, then don't free */
4824 	if (!rx_ring->desc)
4825 		return;
4826 
4827 	dma_free_coherent(rx_ring->dev, rx_ring->size,
4828 			  rx_ring->desc, rx_ring->dma);
4829 
4830 	rx_ring->desc = NULL;
4831 }
4832 
4833 /**
4834  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
4835  * @adapter: board private structure
4836  *
4837  * Free all receive software resources
4838  **/
4839 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
4840 {
4841 	int i;
4842 
4843 #ifdef IXGBE_FCOE
4844 	ixgbe_free_fcoe_ddp_resources(adapter);
4845 
4846 #endif
4847 	for (i = 0; i < adapter->num_rx_queues; i++)
4848 		if (adapter->rx_ring[i]->desc)
4849 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
4850 }
4851 
4852 /**
4853  * ixgbe_change_mtu - Change the Maximum Transfer Unit
4854  * @netdev: network interface device structure
4855  * @new_mtu: new value for maximum frame size
4856  *
4857  * Returns 0 on success, negative on failure
4858  **/
4859 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
4860 {
4861 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4862 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
4863 
4864 	/* MTU < 68 is an error and causes problems on some kernels */
4865 	if ((new_mtu < 68) || (max_frame > IXGBE_MAX_JUMBO_FRAME_SIZE))
4866 		return -EINVAL;
4867 
4868 	/*
4869 	 * For 82599EB we cannot allow legacy VFs to enable their receive
4870 	 * paths when MTU greater than 1500 is configured.  So display a
4871 	 * warning that legacy VFs will be disabled.
4872 	 */
4873 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
4874 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
4875 	    (max_frame > MAXIMUM_ETHERNET_VLAN_SIZE))
4876 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
4877 
4878 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
4879 
4880 	/* must set new MTU before calling down or up */
4881 	netdev->mtu = new_mtu;
4882 
4883 	if (netif_running(netdev))
4884 		ixgbe_reinit_locked(adapter);
4885 
4886 	return 0;
4887 }
4888 
4889 /**
4890  * ixgbe_open - Called when a network interface is made active
4891  * @netdev: network interface device structure
4892  *
4893  * Returns 0 on success, negative value on failure
4894  *
4895  * The open entry point is called when a network interface is made
4896  * active by the system (IFF_UP).  At this point all resources needed
4897  * for transmit and receive operations are allocated, the interrupt
4898  * handler is registered with the OS, the watchdog timer is started,
4899  * and the stack is notified that the interface is ready.
4900  **/
4901 static int ixgbe_open(struct net_device *netdev)
4902 {
4903 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4904 	int err;
4905 
4906 	/* disallow open during test */
4907 	if (test_bit(__IXGBE_TESTING, &adapter->state))
4908 		return -EBUSY;
4909 
4910 	netif_carrier_off(netdev);
4911 
4912 	/* allocate transmit descriptors */
4913 	err = ixgbe_setup_all_tx_resources(adapter);
4914 	if (err)
4915 		goto err_setup_tx;
4916 
4917 	/* allocate receive descriptors */
4918 	err = ixgbe_setup_all_rx_resources(adapter);
4919 	if (err)
4920 		goto err_setup_rx;
4921 
4922 	ixgbe_configure(adapter);
4923 
4924 	err = ixgbe_request_irq(adapter);
4925 	if (err)
4926 		goto err_req_irq;
4927 
4928 	/* Notify the stack of the actual queue counts. */
4929 	err = netif_set_real_num_tx_queues(netdev,
4930 					   adapter->num_rx_pools > 1 ? 1 :
4931 					   adapter->num_tx_queues);
4932 	if (err)
4933 		goto err_set_queues;
4934 
4935 
4936 	err = netif_set_real_num_rx_queues(netdev,
4937 					   adapter->num_rx_pools > 1 ? 1 :
4938 					   adapter->num_rx_queues);
4939 	if (err)
4940 		goto err_set_queues;
4941 
4942 	ixgbe_ptp_init(adapter);
4943 
4944 	ixgbe_up_complete(adapter);
4945 
4946 	return 0;
4947 
4948 err_set_queues:
4949 	ixgbe_free_irq(adapter);
4950 err_req_irq:
4951 	ixgbe_free_all_rx_resources(adapter);
4952 err_setup_rx:
4953 	ixgbe_free_all_tx_resources(adapter);
4954 err_setup_tx:
4955 	ixgbe_reset(adapter);
4956 
4957 	return err;
4958 }
4959 
4960 /**
4961  * ixgbe_close - Disables a network interface
4962  * @netdev: network interface device structure
4963  *
4964  * Returns 0, this is not allowed to fail
4965  *
4966  * The close entry point is called when an interface is de-activated
4967  * by the OS.  The hardware is still under the drivers control, but
4968  * needs to be disabled.  A global MAC reset is issued to stop the
4969  * hardware, and all transmit and receive resources are freed.
4970  **/
4971 static int ixgbe_close(struct net_device *netdev)
4972 {
4973 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4974 
4975 	ixgbe_ptp_stop(adapter);
4976 
4977 	ixgbe_down(adapter);
4978 	ixgbe_free_irq(adapter);
4979 
4980 	ixgbe_fdir_filter_exit(adapter);
4981 
4982 	ixgbe_free_all_tx_resources(adapter);
4983 	ixgbe_free_all_rx_resources(adapter);
4984 
4985 	ixgbe_release_hw_control(adapter);
4986 
4987 	return 0;
4988 }
4989 
4990 #ifdef CONFIG_PM
4991 static int ixgbe_resume(struct pci_dev *pdev)
4992 {
4993 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
4994 	struct net_device *netdev = adapter->netdev;
4995 	u32 err;
4996 
4997 	pci_set_power_state(pdev, PCI_D0);
4998 	pci_restore_state(pdev);
4999 	/*
5000 	 * pci_restore_state clears dev->state_saved so call
5001 	 * pci_save_state to restore it.
5002 	 */
5003 	pci_save_state(pdev);
5004 
5005 	err = pci_enable_device_mem(pdev);
5006 	if (err) {
5007 		e_dev_err("Cannot enable PCI device from suspend\n");
5008 		return err;
5009 	}
5010 	pci_set_master(pdev);
5011 
5012 	pci_wake_from_d3(pdev, false);
5013 
5014 	ixgbe_reset(adapter);
5015 
5016 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
5017 
5018 	rtnl_lock();
5019 	err = ixgbe_init_interrupt_scheme(adapter);
5020 	if (!err && netif_running(netdev))
5021 		err = ixgbe_open(netdev);
5022 
5023 	rtnl_unlock();
5024 
5025 	if (err)
5026 		return err;
5027 
5028 	netif_device_attach(netdev);
5029 
5030 	return 0;
5031 }
5032 #endif /* CONFIG_PM */
5033 
5034 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
5035 {
5036 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
5037 	struct net_device *netdev = adapter->netdev;
5038 	struct ixgbe_hw *hw = &adapter->hw;
5039 	u32 ctrl, fctrl;
5040 	u32 wufc = adapter->wol;
5041 #ifdef CONFIG_PM
5042 	int retval = 0;
5043 #endif
5044 
5045 	netif_device_detach(netdev);
5046 
5047 	if (netif_running(netdev)) {
5048 		rtnl_lock();
5049 		ixgbe_down(adapter);
5050 		ixgbe_free_irq(adapter);
5051 		ixgbe_free_all_tx_resources(adapter);
5052 		ixgbe_free_all_rx_resources(adapter);
5053 		rtnl_unlock();
5054 	}
5055 
5056 	ixgbe_clear_interrupt_scheme(adapter);
5057 
5058 #ifdef CONFIG_PM
5059 	retval = pci_save_state(pdev);
5060 	if (retval)
5061 		return retval;
5062 
5063 #endif
5064 	if (wufc) {
5065 		ixgbe_set_rx_mode(netdev);
5066 
5067 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
5068 		if (hw->mac.ops.enable_tx_laser)
5069 			hw->mac.ops.enable_tx_laser(hw);
5070 
5071 		/* turn on all-multi mode if wake on multicast is enabled */
5072 		if (wufc & IXGBE_WUFC_MC) {
5073 			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
5074 			fctrl |= IXGBE_FCTRL_MPE;
5075 			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
5076 		}
5077 
5078 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
5079 		ctrl |= IXGBE_CTRL_GIO_DIS;
5080 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
5081 
5082 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
5083 	} else {
5084 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
5085 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
5086 	}
5087 
5088 	switch (hw->mac.type) {
5089 	case ixgbe_mac_82598EB:
5090 		pci_wake_from_d3(pdev, false);
5091 		break;
5092 	case ixgbe_mac_82599EB:
5093 	case ixgbe_mac_X540:
5094 		pci_wake_from_d3(pdev, !!wufc);
5095 		break;
5096 	default:
5097 		break;
5098 	}
5099 
5100 	*enable_wake = !!wufc;
5101 
5102 	ixgbe_release_hw_control(adapter);
5103 
5104 	pci_disable_device(pdev);
5105 
5106 	return 0;
5107 }
5108 
5109 #ifdef CONFIG_PM
5110 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
5111 {
5112 	int retval;
5113 	bool wake;
5114 
5115 	retval = __ixgbe_shutdown(pdev, &wake);
5116 	if (retval)
5117 		return retval;
5118 
5119 	if (wake) {
5120 		pci_prepare_to_sleep(pdev);
5121 	} else {
5122 		pci_wake_from_d3(pdev, false);
5123 		pci_set_power_state(pdev, PCI_D3hot);
5124 	}
5125 
5126 	return 0;
5127 }
5128 #endif /* CONFIG_PM */
5129 
5130 static void ixgbe_shutdown(struct pci_dev *pdev)
5131 {
5132 	bool wake;
5133 
5134 	__ixgbe_shutdown(pdev, &wake);
5135 
5136 	if (system_state == SYSTEM_POWER_OFF) {
5137 		pci_wake_from_d3(pdev, wake);
5138 		pci_set_power_state(pdev, PCI_D3hot);
5139 	}
5140 }
5141 
5142 /**
5143  * ixgbe_update_stats - Update the board statistics counters.
5144  * @adapter: board private structure
5145  **/
5146 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
5147 {
5148 	struct net_device *netdev = adapter->netdev;
5149 	struct ixgbe_hw *hw = &adapter->hw;
5150 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
5151 	u64 total_mpc = 0;
5152 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
5153 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
5154 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
5155 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
5156 
5157 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5158 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5159 		return;
5160 
5161 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
5162 		u64 rsc_count = 0;
5163 		u64 rsc_flush = 0;
5164 		for (i = 0; i < adapter->num_rx_queues; i++) {
5165 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
5166 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
5167 		}
5168 		adapter->rsc_total_count = rsc_count;
5169 		adapter->rsc_total_flush = rsc_flush;
5170 	}
5171 
5172 	for (i = 0; i < adapter->num_rx_queues; i++) {
5173 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
5174 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
5175 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
5176 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
5177 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
5178 		bytes += rx_ring->stats.bytes;
5179 		packets += rx_ring->stats.packets;
5180 	}
5181 	adapter->non_eop_descs = non_eop_descs;
5182 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
5183 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
5184 	adapter->hw_csum_rx_error = hw_csum_rx_error;
5185 	netdev->stats.rx_bytes = bytes;
5186 	netdev->stats.rx_packets = packets;
5187 
5188 	bytes = 0;
5189 	packets = 0;
5190 	/* gather some stats to the adapter struct that are per queue */
5191 	for (i = 0; i < adapter->num_tx_queues; i++) {
5192 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
5193 		restart_queue += tx_ring->tx_stats.restart_queue;
5194 		tx_busy += tx_ring->tx_stats.tx_busy;
5195 		bytes += tx_ring->stats.bytes;
5196 		packets += tx_ring->stats.packets;
5197 	}
5198 	adapter->restart_queue = restart_queue;
5199 	adapter->tx_busy = tx_busy;
5200 	netdev->stats.tx_bytes = bytes;
5201 	netdev->stats.tx_packets = packets;
5202 
5203 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
5204 
5205 	/* 8 register reads */
5206 	for (i = 0; i < 8; i++) {
5207 		/* for packet buffers not used, the register should read 0 */
5208 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
5209 		missed_rx += mpc;
5210 		hwstats->mpc[i] += mpc;
5211 		total_mpc += hwstats->mpc[i];
5212 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
5213 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
5214 		switch (hw->mac.type) {
5215 		case ixgbe_mac_82598EB:
5216 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
5217 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
5218 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
5219 			hwstats->pxonrxc[i] +=
5220 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
5221 			break;
5222 		case ixgbe_mac_82599EB:
5223 		case ixgbe_mac_X540:
5224 			hwstats->pxonrxc[i] +=
5225 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
5226 			break;
5227 		default:
5228 			break;
5229 		}
5230 	}
5231 
5232 	/*16 register reads */
5233 	for (i = 0; i < 16; i++) {
5234 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
5235 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
5236 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
5237 		    (hw->mac.type == ixgbe_mac_X540)) {
5238 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
5239 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
5240 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
5241 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
5242 		}
5243 	}
5244 
5245 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
5246 	/* work around hardware counting issue */
5247 	hwstats->gprc -= missed_rx;
5248 
5249 	ixgbe_update_xoff_received(adapter);
5250 
5251 	/* 82598 hardware only has a 32 bit counter in the high register */
5252 	switch (hw->mac.type) {
5253 	case ixgbe_mac_82598EB:
5254 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
5255 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
5256 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
5257 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
5258 		break;
5259 	case ixgbe_mac_X540:
5260 		/* OS2BMC stats are X540 only*/
5261 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
5262 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
5263 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
5264 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
5265 	case ixgbe_mac_82599EB:
5266 		for (i = 0; i < 16; i++)
5267 			adapter->hw_rx_no_dma_resources +=
5268 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
5269 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
5270 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
5271 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
5272 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
5273 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
5274 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
5275 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
5276 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
5277 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
5278 #ifdef IXGBE_FCOE
5279 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
5280 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
5281 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
5282 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
5283 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
5284 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
5285 		/* Add up per cpu counters for total ddp aloc fail */
5286 		if (adapter->fcoe.ddp_pool) {
5287 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
5288 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
5289 			unsigned int cpu;
5290 			u64 noddp = 0, noddp_ext_buff = 0;
5291 			for_each_possible_cpu(cpu) {
5292 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
5293 				noddp += ddp_pool->noddp;
5294 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
5295 			}
5296 			hwstats->fcoe_noddp = noddp;
5297 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
5298 		}
5299 #endif /* IXGBE_FCOE */
5300 		break;
5301 	default:
5302 		break;
5303 	}
5304 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
5305 	hwstats->bprc += bprc;
5306 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
5307 	if (hw->mac.type == ixgbe_mac_82598EB)
5308 		hwstats->mprc -= bprc;
5309 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
5310 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
5311 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
5312 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
5313 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
5314 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
5315 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
5316 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
5317 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
5318 	hwstats->lxontxc += lxon;
5319 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
5320 	hwstats->lxofftxc += lxoff;
5321 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
5322 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
5323 	/*
5324 	 * 82598 errata - tx of flow control packets is included in tx counters
5325 	 */
5326 	xon_off_tot = lxon + lxoff;
5327 	hwstats->gptc -= xon_off_tot;
5328 	hwstats->mptc -= xon_off_tot;
5329 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
5330 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
5331 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
5332 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
5333 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
5334 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
5335 	hwstats->ptc64 -= xon_off_tot;
5336 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
5337 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
5338 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
5339 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
5340 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
5341 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
5342 
5343 	/* Fill out the OS statistics structure */
5344 	netdev->stats.multicast = hwstats->mprc;
5345 
5346 	/* Rx Errors */
5347 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
5348 	netdev->stats.rx_dropped = 0;
5349 	netdev->stats.rx_length_errors = hwstats->rlec;
5350 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
5351 	netdev->stats.rx_missed_errors = total_mpc;
5352 }
5353 
5354 /**
5355  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
5356  * @adapter: pointer to the device adapter structure
5357  **/
5358 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
5359 {
5360 	struct ixgbe_hw *hw = &adapter->hw;
5361 	int i;
5362 
5363 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
5364 		return;
5365 
5366 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
5367 
5368 	/* if interface is down do nothing */
5369 	if (test_bit(__IXGBE_DOWN, &adapter->state))
5370 		return;
5371 
5372 	/* do nothing if we are not using signature filters */
5373 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
5374 		return;
5375 
5376 	adapter->fdir_overflow++;
5377 
5378 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
5379 		for (i = 0; i < adapter->num_tx_queues; i++)
5380 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
5381 			        &(adapter->tx_ring[i]->state));
5382 		/* re-enable flow director interrupts */
5383 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
5384 	} else {
5385 		e_err(probe, "failed to finish FDIR re-initialization, "
5386 		      "ignored adding FDIR ATR filters\n");
5387 	}
5388 }
5389 
5390 /**
5391  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
5392  * @adapter: pointer to the device adapter structure
5393  *
5394  * This function serves two purposes.  First it strobes the interrupt lines
5395  * in order to make certain interrupts are occurring.  Secondly it sets the
5396  * bits needed to check for TX hangs.  As a result we should immediately
5397  * determine if a hang has occurred.
5398  */
5399 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
5400 {
5401 	struct ixgbe_hw *hw = &adapter->hw;
5402 	u64 eics = 0;
5403 	int i;
5404 
5405 	/* If we're down or resetting, just bail */
5406 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5407 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5408 		return;
5409 
5410 	/* Force detection of hung controller */
5411 	if (netif_carrier_ok(adapter->netdev)) {
5412 		for (i = 0; i < adapter->num_tx_queues; i++)
5413 			set_check_for_tx_hang(adapter->tx_ring[i]);
5414 	}
5415 
5416 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
5417 		/*
5418 		 * for legacy and MSI interrupts don't set any bits
5419 		 * that are enabled for EIAM, because this operation
5420 		 * would set *both* EIMS and EICS for any bit in EIAM
5421 		 */
5422 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
5423 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
5424 	} else {
5425 		/* get one bit for every active tx/rx interrupt vector */
5426 		for (i = 0; i < adapter->num_q_vectors; i++) {
5427 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
5428 			if (qv->rx.ring || qv->tx.ring)
5429 				eics |= ((u64)1 << i);
5430 		}
5431 	}
5432 
5433 	/* Cause software interrupt to ensure rings are cleaned */
5434 	ixgbe_irq_rearm_queues(adapter, eics);
5435 
5436 }
5437 
5438 /**
5439  * ixgbe_watchdog_update_link - update the link status
5440  * @adapter: pointer to the device adapter structure
5441  * @link_speed: pointer to a u32 to store the link_speed
5442  **/
5443 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
5444 {
5445 	struct ixgbe_hw *hw = &adapter->hw;
5446 	u32 link_speed = adapter->link_speed;
5447 	bool link_up = adapter->link_up;
5448 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
5449 
5450 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
5451 		return;
5452 
5453 	if (hw->mac.ops.check_link) {
5454 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
5455 	} else {
5456 		/* always assume link is up, if no check link function */
5457 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
5458 		link_up = true;
5459 	}
5460 
5461 	if (adapter->ixgbe_ieee_pfc)
5462 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
5463 
5464 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
5465 		hw->mac.ops.fc_enable(hw);
5466 		ixgbe_set_rx_drop_en(adapter);
5467 	}
5468 
5469 	if (link_up ||
5470 	    time_after(jiffies, (adapter->link_check_timeout +
5471 				 IXGBE_TRY_LINK_TIMEOUT))) {
5472 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5473 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
5474 		IXGBE_WRITE_FLUSH(hw);
5475 	}
5476 
5477 	adapter->link_up = link_up;
5478 	adapter->link_speed = link_speed;
5479 }
5480 
5481 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
5482 {
5483 #ifdef CONFIG_IXGBE_DCB
5484 	struct net_device *netdev = adapter->netdev;
5485 	struct dcb_app app = {
5486 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
5487 			      .protocol = 0,
5488 			     };
5489 	u8 up = 0;
5490 
5491 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
5492 		up = dcb_ieee_getapp_mask(netdev, &app);
5493 
5494 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
5495 #endif
5496 }
5497 
5498 /**
5499  * ixgbe_watchdog_link_is_up - update netif_carrier status and
5500  *                             print link up message
5501  * @adapter: pointer to the device adapter structure
5502  **/
5503 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
5504 {
5505 	struct net_device *netdev = adapter->netdev;
5506 	struct ixgbe_hw *hw = &adapter->hw;
5507 	u32 link_speed = adapter->link_speed;
5508 	bool flow_rx, flow_tx;
5509 
5510 	/* only continue if link was previously down */
5511 	if (netif_carrier_ok(netdev))
5512 		return;
5513 
5514 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
5515 
5516 	switch (hw->mac.type) {
5517 	case ixgbe_mac_82598EB: {
5518 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
5519 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
5520 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
5521 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
5522 	}
5523 		break;
5524 	case ixgbe_mac_X540:
5525 	case ixgbe_mac_82599EB: {
5526 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
5527 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
5528 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
5529 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
5530 	}
5531 		break;
5532 	default:
5533 		flow_tx = false;
5534 		flow_rx = false;
5535 		break;
5536 	}
5537 
5538 	if (adapter->flags2 & IXGBE_FLAG2_PTP_ENABLED)
5539 		ixgbe_ptp_start_cyclecounter(adapter);
5540 
5541 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n",
5542 	       (link_speed == IXGBE_LINK_SPEED_10GB_FULL ?
5543 	       "10 Gbps" :
5544 	       (link_speed == IXGBE_LINK_SPEED_1GB_FULL ?
5545 	       "1 Gbps" :
5546 	       (link_speed == IXGBE_LINK_SPEED_100_FULL ?
5547 	       "100 Mbps" :
5548 	       "unknown speed"))),
5549 	       ((flow_rx && flow_tx) ? "RX/TX" :
5550 	       (flow_rx ? "RX" :
5551 	       (flow_tx ? "TX" : "None"))));
5552 
5553 	netif_carrier_on(netdev);
5554 	ixgbe_check_vf_rate_limit(adapter);
5555 
5556 	/* update the default user priority for VFs */
5557 	ixgbe_update_default_up(adapter);
5558 
5559 	/* ping all the active vfs to let them know link has changed */
5560 	ixgbe_ping_all_vfs(adapter);
5561 }
5562 
5563 /**
5564  * ixgbe_watchdog_link_is_down - update netif_carrier status and
5565  *                               print link down message
5566  * @adapter: pointer to the adapter structure
5567  **/
5568 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
5569 {
5570 	struct net_device *netdev = adapter->netdev;
5571 	struct ixgbe_hw *hw = &adapter->hw;
5572 
5573 	adapter->link_up = false;
5574 	adapter->link_speed = 0;
5575 
5576 	/* only continue if link was up previously */
5577 	if (!netif_carrier_ok(netdev))
5578 		return;
5579 
5580 	/* poll for SFP+ cable when link is down */
5581 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
5582 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5583 
5584 	if (adapter->flags2 & IXGBE_FLAG2_PTP_ENABLED)
5585 		ixgbe_ptp_start_cyclecounter(adapter);
5586 
5587 	e_info(drv, "NIC Link is Down\n");
5588 	netif_carrier_off(netdev);
5589 
5590 	/* ping all the active vfs to let them know link has changed */
5591 	ixgbe_ping_all_vfs(adapter);
5592 }
5593 
5594 /**
5595  * ixgbe_watchdog_flush_tx - flush queues on link down
5596  * @adapter: pointer to the device adapter structure
5597  **/
5598 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
5599 {
5600 	int i;
5601 	int some_tx_pending = 0;
5602 
5603 	if (!netif_carrier_ok(adapter->netdev)) {
5604 		for (i = 0; i < adapter->num_tx_queues; i++) {
5605 			struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
5606 			if (tx_ring->next_to_use != tx_ring->next_to_clean) {
5607 				some_tx_pending = 1;
5608 				break;
5609 			}
5610 		}
5611 
5612 		if (some_tx_pending) {
5613 			/* We've lost link, so the controller stops DMA,
5614 			 * but we've got queued Tx work that's never going
5615 			 * to get done, so reset controller to flush Tx.
5616 			 * (Do the reset outside of interrupt context).
5617 			 */
5618 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
5619 		}
5620 	}
5621 }
5622 
5623 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
5624 {
5625 	u32 ssvpc;
5626 
5627 	/* Do not perform spoof check for 82598 or if not in IOV mode */
5628 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
5629 	    adapter->num_vfs == 0)
5630 		return;
5631 
5632 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
5633 
5634 	/*
5635 	 * ssvpc register is cleared on read, if zero then no
5636 	 * spoofed packets in the last interval.
5637 	 */
5638 	if (!ssvpc)
5639 		return;
5640 
5641 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
5642 }
5643 
5644 /**
5645  * ixgbe_watchdog_subtask - check and bring link up
5646  * @adapter: pointer to the device adapter structure
5647  **/
5648 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
5649 {
5650 	/* if interface is down do nothing */
5651 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5652 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5653 		return;
5654 
5655 	ixgbe_watchdog_update_link(adapter);
5656 
5657 	if (adapter->link_up)
5658 		ixgbe_watchdog_link_is_up(adapter);
5659 	else
5660 		ixgbe_watchdog_link_is_down(adapter);
5661 
5662 	ixgbe_spoof_check(adapter);
5663 	ixgbe_update_stats(adapter);
5664 
5665 	ixgbe_watchdog_flush_tx(adapter);
5666 }
5667 
5668 /**
5669  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
5670  * @adapter: the ixgbe adapter structure
5671  **/
5672 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
5673 {
5674 	struct ixgbe_hw *hw = &adapter->hw;
5675 	s32 err;
5676 
5677 	/* not searching for SFP so there is nothing to do here */
5678 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
5679 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
5680 		return;
5681 
5682 	/* someone else is in init, wait until next service event */
5683 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5684 		return;
5685 
5686 	err = hw->phy.ops.identify_sfp(hw);
5687 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
5688 		goto sfp_out;
5689 
5690 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
5691 		/* If no cable is present, then we need to reset
5692 		 * the next time we find a good cable. */
5693 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5694 	}
5695 
5696 	/* exit on error */
5697 	if (err)
5698 		goto sfp_out;
5699 
5700 	/* exit if reset not needed */
5701 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
5702 		goto sfp_out;
5703 
5704 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
5705 
5706 	/*
5707 	 * A module may be identified correctly, but the EEPROM may not have
5708 	 * support for that module.  setup_sfp() will fail in that case, so
5709 	 * we should not allow that module to load.
5710 	 */
5711 	if (hw->mac.type == ixgbe_mac_82598EB)
5712 		err = hw->phy.ops.reset(hw);
5713 	else
5714 		err = hw->mac.ops.setup_sfp(hw);
5715 
5716 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
5717 		goto sfp_out;
5718 
5719 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
5720 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
5721 
5722 sfp_out:
5723 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5724 
5725 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
5726 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
5727 		e_dev_err("failed to initialize because an unsupported "
5728 			  "SFP+ module type was detected.\n");
5729 		e_dev_err("Reload the driver after installing a "
5730 			  "supported module.\n");
5731 		unregister_netdev(adapter->netdev);
5732 	}
5733 }
5734 
5735 /**
5736  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
5737  * @adapter: the ixgbe adapter structure
5738  **/
5739 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
5740 {
5741 	struct ixgbe_hw *hw = &adapter->hw;
5742 	u32 autoneg;
5743 	bool negotiation;
5744 
5745 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
5746 		return;
5747 
5748 	/* someone else is in init, wait until next service event */
5749 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5750 		return;
5751 
5752 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5753 
5754 	autoneg = hw->phy.autoneg_advertised;
5755 	if ((!autoneg) && (hw->mac.ops.get_link_capabilities))
5756 		hw->mac.ops.get_link_capabilities(hw, &autoneg, &negotiation);
5757 	if (hw->mac.ops.setup_link)
5758 		hw->mac.ops.setup_link(hw, autoneg, negotiation, true);
5759 
5760 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5761 	adapter->link_check_timeout = jiffies;
5762 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5763 }
5764 
5765 #ifdef CONFIG_PCI_IOV
5766 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
5767 {
5768 	int vf;
5769 	struct ixgbe_hw *hw = &adapter->hw;
5770 	struct net_device *netdev = adapter->netdev;
5771 	u32 gpc;
5772 	u32 ciaa, ciad;
5773 
5774 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
5775 	if (gpc) /* If incrementing then no need for the check below */
5776 		return;
5777 	/*
5778 	 * Check to see if a bad DMA write target from an errant or
5779 	 * malicious VF has caused a PCIe error.  If so then we can
5780 	 * issue a VFLR to the offending VF(s) and then resume without
5781 	 * requesting a full slot reset.
5782 	 */
5783 
5784 	for (vf = 0; vf < adapter->num_vfs; vf++) {
5785 		ciaa = (vf << 16) | 0x80000000;
5786 		/* 32 bit read so align, we really want status at offset 6 */
5787 		ciaa |= PCI_COMMAND;
5788 		IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5789 		ciad = IXGBE_READ_REG(hw, IXGBE_CIAD_82599);
5790 		ciaa &= 0x7FFFFFFF;
5791 		/* disable debug mode asap after reading data */
5792 		IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5793 		/* Get the upper 16 bits which will be the PCI status reg */
5794 		ciad >>= 16;
5795 		if (ciad & PCI_STATUS_REC_MASTER_ABORT) {
5796 			netdev_err(netdev, "VF %d Hung DMA\n", vf);
5797 			/* Issue VFLR */
5798 			ciaa = (vf << 16) | 0x80000000;
5799 			ciaa |= 0xA8;
5800 			IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5801 			ciad = 0x00008000;  /* VFLR */
5802 			IXGBE_WRITE_REG(hw, IXGBE_CIAD_82599, ciad);
5803 			ciaa &= 0x7FFFFFFF;
5804 			IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5805 		}
5806 	}
5807 }
5808 
5809 #endif
5810 /**
5811  * ixgbe_service_timer - Timer Call-back
5812  * @data: pointer to adapter cast into an unsigned long
5813  **/
5814 static void ixgbe_service_timer(unsigned long data)
5815 {
5816 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
5817 	unsigned long next_event_offset;
5818 	bool ready = true;
5819 
5820 	/* poll faster when waiting for link */
5821 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
5822 		next_event_offset = HZ / 10;
5823 	else
5824 		next_event_offset = HZ * 2;
5825 
5826 #ifdef CONFIG_PCI_IOV
5827 	/*
5828 	 * don't bother with SR-IOV VF DMA hang check if there are
5829 	 * no VFs or the link is down
5830 	 */
5831 	if (!adapter->num_vfs ||
5832 	    (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
5833 		goto normal_timer_service;
5834 
5835 	/* If we have VFs allocated then we must check for DMA hangs */
5836 	ixgbe_check_for_bad_vf(adapter);
5837 	next_event_offset = HZ / 50;
5838 	adapter->timer_event_accumulator++;
5839 
5840 	if (adapter->timer_event_accumulator >= 100)
5841 		adapter->timer_event_accumulator = 0;
5842 	else
5843 		ready = false;
5844 
5845 normal_timer_service:
5846 #endif
5847 	/* Reset the timer */
5848 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
5849 
5850 	if (ready)
5851 		ixgbe_service_event_schedule(adapter);
5852 }
5853 
5854 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
5855 {
5856 	if (!(adapter->flags2 & IXGBE_FLAG2_RESET_REQUESTED))
5857 		return;
5858 
5859 	adapter->flags2 &= ~IXGBE_FLAG2_RESET_REQUESTED;
5860 
5861 	/* If we're already down or resetting, just bail */
5862 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5863 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5864 		return;
5865 
5866 	ixgbe_dump(adapter);
5867 	netdev_err(adapter->netdev, "Reset adapter\n");
5868 	adapter->tx_timeout_count++;
5869 
5870 	ixgbe_reinit_locked(adapter);
5871 }
5872 
5873 /**
5874  * ixgbe_service_task - manages and runs subtasks
5875  * @work: pointer to work_struct containing our data
5876  **/
5877 static void ixgbe_service_task(struct work_struct *work)
5878 {
5879 	struct ixgbe_adapter *adapter = container_of(work,
5880 						     struct ixgbe_adapter,
5881 						     service_task);
5882 
5883 	ixgbe_reset_subtask(adapter);
5884 	ixgbe_sfp_detection_subtask(adapter);
5885 	ixgbe_sfp_link_config_subtask(adapter);
5886 	ixgbe_check_overtemp_subtask(adapter);
5887 	ixgbe_watchdog_subtask(adapter);
5888 	ixgbe_fdir_reinit_subtask(adapter);
5889 	ixgbe_check_hang_subtask(adapter);
5890 	ixgbe_ptp_overflow_check(adapter);
5891 
5892 	ixgbe_service_event_complete(adapter);
5893 }
5894 
5895 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
5896 		     struct ixgbe_tx_buffer *first,
5897 		     u8 *hdr_len)
5898 {
5899 	struct sk_buff *skb = first->skb;
5900 	u32 vlan_macip_lens, type_tucmd;
5901 	u32 mss_l4len_idx, l4len;
5902 
5903 	if (!skb_is_gso(skb))
5904 		return 0;
5905 
5906 	if (skb_header_cloned(skb)) {
5907 		int err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
5908 		if (err)
5909 			return err;
5910 	}
5911 
5912 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
5913 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
5914 
5915 	if (first->protocol == __constant_htons(ETH_P_IP)) {
5916 		struct iphdr *iph = ip_hdr(skb);
5917 		iph->tot_len = 0;
5918 		iph->check = 0;
5919 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
5920 							 iph->daddr, 0,
5921 							 IPPROTO_TCP,
5922 							 0);
5923 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
5924 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
5925 				   IXGBE_TX_FLAGS_CSUM |
5926 				   IXGBE_TX_FLAGS_IPV4;
5927 	} else if (skb_is_gso_v6(skb)) {
5928 		ipv6_hdr(skb)->payload_len = 0;
5929 		tcp_hdr(skb)->check =
5930 		    ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
5931 				     &ipv6_hdr(skb)->daddr,
5932 				     0, IPPROTO_TCP, 0);
5933 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
5934 				   IXGBE_TX_FLAGS_CSUM;
5935 	}
5936 
5937 	/* compute header lengths */
5938 	l4len = tcp_hdrlen(skb);
5939 	*hdr_len = skb_transport_offset(skb) + l4len;
5940 
5941 	/* update gso size and bytecount with header size */
5942 	first->gso_segs = skb_shinfo(skb)->gso_segs;
5943 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
5944 
5945 	/* mss_l4len_id: use 1 as index for TSO */
5946 	mss_l4len_idx = l4len << IXGBE_ADVTXD_L4LEN_SHIFT;
5947 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
5948 	mss_l4len_idx |= 1 << IXGBE_ADVTXD_IDX_SHIFT;
5949 
5950 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
5951 	vlan_macip_lens = skb_network_header_len(skb);
5952 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
5953 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
5954 
5955 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
5956 			  mss_l4len_idx);
5957 
5958 	return 1;
5959 }
5960 
5961 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
5962 			  struct ixgbe_tx_buffer *first)
5963 {
5964 	struct sk_buff *skb = first->skb;
5965 	u32 vlan_macip_lens = 0;
5966 	u32 mss_l4len_idx = 0;
5967 	u32 type_tucmd = 0;
5968 
5969 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
5970 		if (!(first->tx_flags & IXGBE_TX_FLAGS_HW_VLAN)) {
5971 			if (unlikely(skb->no_fcs))
5972 				first->tx_flags |= IXGBE_TX_FLAGS_NO_IFCS;
5973 			if (!(first->tx_flags & IXGBE_TX_FLAGS_TXSW))
5974 				return;
5975 		}
5976 	} else {
5977 		u8 l4_hdr = 0;
5978 		switch (first->protocol) {
5979 		case __constant_htons(ETH_P_IP):
5980 			vlan_macip_lens |= skb_network_header_len(skb);
5981 			type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
5982 			l4_hdr = ip_hdr(skb)->protocol;
5983 			break;
5984 		case __constant_htons(ETH_P_IPV6):
5985 			vlan_macip_lens |= skb_network_header_len(skb);
5986 			l4_hdr = ipv6_hdr(skb)->nexthdr;
5987 			break;
5988 		default:
5989 			if (unlikely(net_ratelimit())) {
5990 				dev_warn(tx_ring->dev,
5991 				 "partial checksum but proto=%x!\n",
5992 				 first->protocol);
5993 			}
5994 			break;
5995 		}
5996 
5997 		switch (l4_hdr) {
5998 		case IPPROTO_TCP:
5999 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_TCP;
6000 			mss_l4len_idx = tcp_hdrlen(skb) <<
6001 					IXGBE_ADVTXD_L4LEN_SHIFT;
6002 			break;
6003 		case IPPROTO_SCTP:
6004 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_SCTP;
6005 			mss_l4len_idx = sizeof(struct sctphdr) <<
6006 					IXGBE_ADVTXD_L4LEN_SHIFT;
6007 			break;
6008 		case IPPROTO_UDP:
6009 			mss_l4len_idx = sizeof(struct udphdr) <<
6010 					IXGBE_ADVTXD_L4LEN_SHIFT;
6011 			break;
6012 		default:
6013 			if (unlikely(net_ratelimit())) {
6014 				dev_warn(tx_ring->dev,
6015 				 "partial checksum but l4 proto=%x!\n",
6016 				 l4_hdr);
6017 			}
6018 			break;
6019 		}
6020 
6021 		/* update TX checksum flag */
6022 		first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
6023 	}
6024 
6025 	/* vlan_macip_lens: MACLEN, VLAN tag */
6026 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
6027 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
6028 
6029 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0,
6030 			  type_tucmd, mss_l4len_idx);
6031 }
6032 
6033 static __le32 ixgbe_tx_cmd_type(u32 tx_flags)
6034 {
6035 	/* set type for advanced descriptor with frame checksum insertion */
6036 	__le32 cmd_type = cpu_to_le32(IXGBE_ADVTXD_DTYP_DATA |
6037 				      IXGBE_ADVTXD_DCMD_DEXT);
6038 
6039 	/* set HW vlan bit if vlan is present */
6040 	if (tx_flags & IXGBE_TX_FLAGS_HW_VLAN)
6041 		cmd_type |= cpu_to_le32(IXGBE_ADVTXD_DCMD_VLE);
6042 
6043 	if (tx_flags & IXGBE_TX_FLAGS_TSTAMP)
6044 		cmd_type |= cpu_to_le32(IXGBE_ADVTXD_MAC_TSTAMP);
6045 
6046 	/* set segmentation enable bits for TSO/FSO */
6047 #ifdef IXGBE_FCOE
6048 	if (tx_flags & (IXGBE_TX_FLAGS_TSO | IXGBE_TX_FLAGS_FSO))
6049 #else
6050 	if (tx_flags & IXGBE_TX_FLAGS_TSO)
6051 #endif
6052 		cmd_type |= cpu_to_le32(IXGBE_ADVTXD_DCMD_TSE);
6053 
6054 	/* insert frame checksum */
6055 	if (!(tx_flags & IXGBE_TX_FLAGS_NO_IFCS))
6056 		cmd_type |= cpu_to_le32(IXGBE_ADVTXD_DCMD_IFCS);
6057 
6058 	return cmd_type;
6059 }
6060 
6061 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
6062 				   u32 tx_flags, unsigned int paylen)
6063 {
6064 	__le32 olinfo_status = cpu_to_le32(paylen << IXGBE_ADVTXD_PAYLEN_SHIFT);
6065 
6066 	/* enable L4 checksum for TSO and TX checksum offload */
6067 	if (tx_flags & IXGBE_TX_FLAGS_CSUM)
6068 		olinfo_status |= cpu_to_le32(IXGBE_ADVTXD_POPTS_TXSM);
6069 
6070 	/* enble IPv4 checksum for TSO */
6071 	if (tx_flags & IXGBE_TX_FLAGS_IPV4)
6072 		olinfo_status |= cpu_to_le32(IXGBE_ADVTXD_POPTS_IXSM);
6073 
6074 	/* use index 1 context for TSO/FSO/FCOE */
6075 #ifdef IXGBE_FCOE
6076 	if (tx_flags & (IXGBE_TX_FLAGS_TSO | IXGBE_TX_FLAGS_FCOE))
6077 #else
6078 	if (tx_flags & IXGBE_TX_FLAGS_TSO)
6079 #endif
6080 		olinfo_status |= cpu_to_le32(1 << IXGBE_ADVTXD_IDX_SHIFT);
6081 
6082 	/*
6083 	 * Check Context must be set if Tx switch is enabled, which it
6084 	 * always is for case where virtual functions are running
6085 	 */
6086 #ifdef IXGBE_FCOE
6087 	if (tx_flags & (IXGBE_TX_FLAGS_TXSW | IXGBE_TX_FLAGS_FCOE))
6088 #else
6089 	if (tx_flags & IXGBE_TX_FLAGS_TXSW)
6090 #endif
6091 		olinfo_status |= cpu_to_le32(IXGBE_ADVTXD_CC);
6092 
6093 	tx_desc->read.olinfo_status = olinfo_status;
6094 }
6095 
6096 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
6097 		       IXGBE_TXD_CMD_RS)
6098 
6099 static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
6100 			 struct ixgbe_tx_buffer *first,
6101 			 const u8 hdr_len)
6102 {
6103 	dma_addr_t dma;
6104 	struct sk_buff *skb = first->skb;
6105 	struct ixgbe_tx_buffer *tx_buffer;
6106 	union ixgbe_adv_tx_desc *tx_desc;
6107 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
6108 	unsigned int data_len = skb->data_len;
6109 	unsigned int size = skb_headlen(skb);
6110 	unsigned int paylen = skb->len - hdr_len;
6111 	u32 tx_flags = first->tx_flags;
6112 	__le32 cmd_type;
6113 	u16 i = tx_ring->next_to_use;
6114 
6115 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
6116 
6117 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, paylen);
6118 	cmd_type = ixgbe_tx_cmd_type(tx_flags);
6119 
6120 #ifdef IXGBE_FCOE
6121 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
6122 		if (data_len < sizeof(struct fcoe_crc_eof)) {
6123 			size -= sizeof(struct fcoe_crc_eof) - data_len;
6124 			data_len = 0;
6125 		} else {
6126 			data_len -= sizeof(struct fcoe_crc_eof);
6127 		}
6128 	}
6129 
6130 #endif
6131 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
6132 	if (dma_mapping_error(tx_ring->dev, dma))
6133 		goto dma_error;
6134 
6135 	/* record length, and DMA address */
6136 	dma_unmap_len_set(first, len, size);
6137 	dma_unmap_addr_set(first, dma, dma);
6138 
6139 	tx_desc->read.buffer_addr = cpu_to_le64(dma);
6140 
6141 	for (;;) {
6142 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
6143 			tx_desc->read.cmd_type_len =
6144 				cmd_type | cpu_to_le32(IXGBE_MAX_DATA_PER_TXD);
6145 
6146 			i++;
6147 			tx_desc++;
6148 			if (i == tx_ring->count) {
6149 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6150 				i = 0;
6151 			}
6152 
6153 			dma += IXGBE_MAX_DATA_PER_TXD;
6154 			size -= IXGBE_MAX_DATA_PER_TXD;
6155 
6156 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
6157 			tx_desc->read.olinfo_status = 0;
6158 		}
6159 
6160 		if (likely(!data_len))
6161 			break;
6162 
6163 		tx_desc->read.cmd_type_len = cmd_type | cpu_to_le32(size);
6164 
6165 		i++;
6166 		tx_desc++;
6167 		if (i == tx_ring->count) {
6168 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6169 			i = 0;
6170 		}
6171 
6172 #ifdef IXGBE_FCOE
6173 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
6174 #else
6175 		size = skb_frag_size(frag);
6176 #endif
6177 		data_len -= size;
6178 
6179 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
6180 				       DMA_TO_DEVICE);
6181 		if (dma_mapping_error(tx_ring->dev, dma))
6182 			goto dma_error;
6183 
6184 		tx_buffer = &tx_ring->tx_buffer_info[i];
6185 		dma_unmap_len_set(tx_buffer, len, size);
6186 		dma_unmap_addr_set(tx_buffer, dma, dma);
6187 
6188 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
6189 		tx_desc->read.olinfo_status = 0;
6190 
6191 		frag++;
6192 	}
6193 
6194 	/* write last descriptor with RS and EOP bits */
6195 	cmd_type |= cpu_to_le32(size) | cpu_to_le32(IXGBE_TXD_CMD);
6196 	tx_desc->read.cmd_type_len = cmd_type;
6197 
6198 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
6199 
6200 	/* set the timestamp */
6201 	first->time_stamp = jiffies;
6202 
6203 	/*
6204 	 * Force memory writes to complete before letting h/w know there
6205 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
6206 	 * memory model archs, such as IA-64).
6207 	 *
6208 	 * We also need this memory barrier to make certain all of the
6209 	 * status bits have been updated before next_to_watch is written.
6210 	 */
6211 	wmb();
6212 
6213 	/* set next_to_watch value indicating a packet is present */
6214 	first->next_to_watch = tx_desc;
6215 
6216 	i++;
6217 	if (i == tx_ring->count)
6218 		i = 0;
6219 
6220 	tx_ring->next_to_use = i;
6221 
6222 	/* notify HW of packet */
6223 	writel(i, tx_ring->tail);
6224 
6225 	return;
6226 dma_error:
6227 	dev_err(tx_ring->dev, "TX DMA map failed\n");
6228 
6229 	/* clear dma mappings for failed tx_buffer_info map */
6230 	for (;;) {
6231 		tx_buffer = &tx_ring->tx_buffer_info[i];
6232 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer);
6233 		if (tx_buffer == first)
6234 			break;
6235 		if (i == 0)
6236 			i = tx_ring->count;
6237 		i--;
6238 	}
6239 
6240 	tx_ring->next_to_use = i;
6241 }
6242 
6243 static void ixgbe_atr(struct ixgbe_ring *ring,
6244 		      struct ixgbe_tx_buffer *first)
6245 {
6246 	struct ixgbe_q_vector *q_vector = ring->q_vector;
6247 	union ixgbe_atr_hash_dword input = { .dword = 0 };
6248 	union ixgbe_atr_hash_dword common = { .dword = 0 };
6249 	union {
6250 		unsigned char *network;
6251 		struct iphdr *ipv4;
6252 		struct ipv6hdr *ipv6;
6253 	} hdr;
6254 	struct tcphdr *th;
6255 	__be16 vlan_id;
6256 
6257 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
6258 	if (!q_vector)
6259 		return;
6260 
6261 	/* do nothing if sampling is disabled */
6262 	if (!ring->atr_sample_rate)
6263 		return;
6264 
6265 	ring->atr_count++;
6266 
6267 	/* snag network header to get L4 type and address */
6268 	hdr.network = skb_network_header(first->skb);
6269 
6270 	/* Currently only IPv4/IPv6 with TCP is supported */
6271 	if ((first->protocol != __constant_htons(ETH_P_IPV6) ||
6272 	     hdr.ipv6->nexthdr != IPPROTO_TCP) &&
6273 	    (first->protocol != __constant_htons(ETH_P_IP) ||
6274 	     hdr.ipv4->protocol != IPPROTO_TCP))
6275 		return;
6276 
6277 	th = tcp_hdr(first->skb);
6278 
6279 	/* skip this packet since it is invalid or the socket is closing */
6280 	if (!th || th->fin)
6281 		return;
6282 
6283 	/* sample on all syn packets or once every atr sample count */
6284 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
6285 		return;
6286 
6287 	/* reset sample count */
6288 	ring->atr_count = 0;
6289 
6290 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
6291 
6292 	/*
6293 	 * src and dst are inverted, think how the receiver sees them
6294 	 *
6295 	 * The input is broken into two sections, a non-compressed section
6296 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
6297 	 * is XORed together and stored in the compressed dword.
6298 	 */
6299 	input.formatted.vlan_id = vlan_id;
6300 
6301 	/*
6302 	 * since src port and flex bytes occupy the same word XOR them together
6303 	 * and write the value to source port portion of compressed dword
6304 	 */
6305 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
6306 		common.port.src ^= th->dest ^ __constant_htons(ETH_P_8021Q);
6307 	else
6308 		common.port.src ^= th->dest ^ first->protocol;
6309 	common.port.dst ^= th->source;
6310 
6311 	if (first->protocol == __constant_htons(ETH_P_IP)) {
6312 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
6313 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
6314 	} else {
6315 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
6316 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
6317 			     hdr.ipv6->saddr.s6_addr32[1] ^
6318 			     hdr.ipv6->saddr.s6_addr32[2] ^
6319 			     hdr.ipv6->saddr.s6_addr32[3] ^
6320 			     hdr.ipv6->daddr.s6_addr32[0] ^
6321 			     hdr.ipv6->daddr.s6_addr32[1] ^
6322 			     hdr.ipv6->daddr.s6_addr32[2] ^
6323 			     hdr.ipv6->daddr.s6_addr32[3];
6324 	}
6325 
6326 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
6327 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
6328 					      input, common, ring->queue_index);
6329 }
6330 
6331 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
6332 {
6333 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
6334 	/* Herbert's original patch had:
6335 	 *  smp_mb__after_netif_stop_queue();
6336 	 * but since that doesn't exist yet, just open code it. */
6337 	smp_mb();
6338 
6339 	/* We need to check again in a case another CPU has just
6340 	 * made room available. */
6341 	if (likely(ixgbe_desc_unused(tx_ring) < size))
6342 		return -EBUSY;
6343 
6344 	/* A reprieve! - use start_queue because it doesn't call schedule */
6345 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
6346 	++tx_ring->tx_stats.restart_queue;
6347 	return 0;
6348 }
6349 
6350 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
6351 {
6352 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
6353 		return 0;
6354 	return __ixgbe_maybe_stop_tx(tx_ring, size);
6355 }
6356 
6357 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb)
6358 {
6359 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6360 	int txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
6361 					       smp_processor_id();
6362 #ifdef IXGBE_FCOE
6363 	__be16 protocol = vlan_get_protocol(skb);
6364 
6365 	if (((protocol == htons(ETH_P_FCOE)) ||
6366 	    (protocol == htons(ETH_P_FIP))) &&
6367 	    (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)) {
6368 		struct ixgbe_ring_feature *f;
6369 
6370 		f = &adapter->ring_feature[RING_F_FCOE];
6371 
6372 		while (txq >= f->indices)
6373 			txq -= f->indices;
6374 		txq += adapter->ring_feature[RING_F_FCOE].offset;
6375 
6376 		return txq;
6377 	}
6378 #endif
6379 
6380 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
6381 		while (unlikely(txq >= dev->real_num_tx_queues))
6382 			txq -= dev->real_num_tx_queues;
6383 		return txq;
6384 	}
6385 
6386 	return skb_tx_hash(dev, skb);
6387 }
6388 
6389 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
6390 			  struct ixgbe_adapter *adapter,
6391 			  struct ixgbe_ring *tx_ring)
6392 {
6393 	struct ixgbe_tx_buffer *first;
6394 	int tso;
6395 	u32 tx_flags = 0;
6396 #if PAGE_SIZE > IXGBE_MAX_DATA_PER_TXD
6397 	unsigned short f;
6398 #endif
6399 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
6400 	__be16 protocol = skb->protocol;
6401 	u8 hdr_len = 0;
6402 
6403 	/*
6404 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
6405 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
6406 	 *       + 2 desc gap to keep tail from touching head,
6407 	 *       + 1 desc for context descriptor,
6408 	 * otherwise try next time
6409 	 */
6410 #if PAGE_SIZE > IXGBE_MAX_DATA_PER_TXD
6411 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
6412 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
6413 #else
6414 	count += skb_shinfo(skb)->nr_frags;
6415 #endif
6416 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
6417 		tx_ring->tx_stats.tx_busy++;
6418 		return NETDEV_TX_BUSY;
6419 	}
6420 
6421 	/* record the location of the first descriptor for this packet */
6422 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
6423 	first->skb = skb;
6424 	first->bytecount = skb->len;
6425 	first->gso_segs = 1;
6426 
6427 	/* if we have a HW VLAN tag being added default to the HW one */
6428 	if (vlan_tx_tag_present(skb)) {
6429 		tx_flags |= vlan_tx_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
6430 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
6431 	/* else if it is a SW VLAN check the next protocol and store the tag */
6432 	} else if (protocol == __constant_htons(ETH_P_8021Q)) {
6433 		struct vlan_hdr *vhdr, _vhdr;
6434 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
6435 		if (!vhdr)
6436 			goto out_drop;
6437 
6438 		protocol = vhdr->h_vlan_encapsulated_proto;
6439 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
6440 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
6441 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
6442 	}
6443 
6444 	skb_tx_timestamp(skb);
6445 
6446 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) {
6447 		skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
6448 		tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
6449 	}
6450 
6451 #ifdef CONFIG_PCI_IOV
6452 	/*
6453 	 * Use the l2switch_enable flag - would be false if the DMA
6454 	 * Tx switch had been disabled.
6455 	 */
6456 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
6457 		tx_flags |= IXGBE_TX_FLAGS_TXSW;
6458 
6459 #endif
6460 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
6461 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
6462 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
6463 	     (skb->priority != TC_PRIO_CONTROL))) {
6464 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
6465 		tx_flags |= (skb->priority & 0x7) <<
6466 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
6467 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
6468 			struct vlan_ethhdr *vhdr;
6469 			if (skb_header_cloned(skb) &&
6470 			    pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
6471 				goto out_drop;
6472 			vhdr = (struct vlan_ethhdr *)skb->data;
6473 			vhdr->h_vlan_TCI = htons(tx_flags >>
6474 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
6475 		} else {
6476 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
6477 		}
6478 	}
6479 
6480 	/* record initial flags and protocol */
6481 	first->tx_flags = tx_flags;
6482 	first->protocol = protocol;
6483 
6484 #ifdef IXGBE_FCOE
6485 	/* setup tx offload for FCoE */
6486 	if ((protocol == __constant_htons(ETH_P_FCOE)) &&
6487 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
6488 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
6489 		if (tso < 0)
6490 			goto out_drop;
6491 
6492 		goto xmit_fcoe;
6493 	}
6494 
6495 #endif /* IXGBE_FCOE */
6496 	tso = ixgbe_tso(tx_ring, first, &hdr_len);
6497 	if (tso < 0)
6498 		goto out_drop;
6499 	else if (!tso)
6500 		ixgbe_tx_csum(tx_ring, first);
6501 
6502 	/* add the ATR filter if ATR is on */
6503 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
6504 		ixgbe_atr(tx_ring, first);
6505 
6506 #ifdef IXGBE_FCOE
6507 xmit_fcoe:
6508 #endif /* IXGBE_FCOE */
6509 	ixgbe_tx_map(tx_ring, first, hdr_len);
6510 
6511 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
6512 
6513 	return NETDEV_TX_OK;
6514 
6515 out_drop:
6516 	dev_kfree_skb_any(first->skb);
6517 	first->skb = NULL;
6518 
6519 	return NETDEV_TX_OK;
6520 }
6521 
6522 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
6523 				    struct net_device *netdev)
6524 {
6525 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6526 	struct ixgbe_ring *tx_ring;
6527 
6528 	/*
6529 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
6530 	 * in order to meet this minimum size requirement.
6531 	 */
6532 	if (unlikely(skb->len < 17)) {
6533 		if (skb_pad(skb, 17 - skb->len))
6534 			return NETDEV_TX_OK;
6535 		skb->len = 17;
6536 		skb_set_tail_pointer(skb, 17);
6537 	}
6538 
6539 	tx_ring = adapter->tx_ring[skb->queue_mapping];
6540 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
6541 }
6542 
6543 /**
6544  * ixgbe_set_mac - Change the Ethernet Address of the NIC
6545  * @netdev: network interface device structure
6546  * @p: pointer to an address structure
6547  *
6548  * Returns 0 on success, negative on failure
6549  **/
6550 static int ixgbe_set_mac(struct net_device *netdev, void *p)
6551 {
6552 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6553 	struct ixgbe_hw *hw = &adapter->hw;
6554 	struct sockaddr *addr = p;
6555 
6556 	if (!is_valid_ether_addr(addr->sa_data))
6557 		return -EADDRNOTAVAIL;
6558 
6559 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
6560 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
6561 
6562 	hw->mac.ops.set_rar(hw, 0, hw->mac.addr, VMDQ_P(0), IXGBE_RAH_AV);
6563 
6564 	return 0;
6565 }
6566 
6567 static int
6568 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
6569 {
6570 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6571 	struct ixgbe_hw *hw = &adapter->hw;
6572 	u16 value;
6573 	int rc;
6574 
6575 	if (prtad != hw->phy.mdio.prtad)
6576 		return -EINVAL;
6577 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
6578 	if (!rc)
6579 		rc = value;
6580 	return rc;
6581 }
6582 
6583 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
6584 			    u16 addr, u16 value)
6585 {
6586 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6587 	struct ixgbe_hw *hw = &adapter->hw;
6588 
6589 	if (prtad != hw->phy.mdio.prtad)
6590 		return -EINVAL;
6591 	return hw->phy.ops.write_reg(hw, addr, devad, value);
6592 }
6593 
6594 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
6595 {
6596 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6597 
6598 	switch (cmd) {
6599 	case SIOCSHWTSTAMP:
6600 		return ixgbe_ptp_hwtstamp_ioctl(adapter, req, cmd);
6601 	default:
6602 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
6603 	}
6604 }
6605 
6606 /**
6607  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
6608  * netdev->dev_addrs
6609  * @netdev: network interface device structure
6610  *
6611  * Returns non-zero on failure
6612  **/
6613 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
6614 {
6615 	int err = 0;
6616 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6617 	struct ixgbe_hw *hw = &adapter->hw;
6618 
6619 	if (is_valid_ether_addr(hw->mac.san_addr)) {
6620 		rtnl_lock();
6621 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
6622 		rtnl_unlock();
6623 
6624 		/* update SAN MAC vmdq pool selection */
6625 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
6626 	}
6627 	return err;
6628 }
6629 
6630 /**
6631  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
6632  * netdev->dev_addrs
6633  * @netdev: network interface device structure
6634  *
6635  * Returns non-zero on failure
6636  **/
6637 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
6638 {
6639 	int err = 0;
6640 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6641 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
6642 
6643 	if (is_valid_ether_addr(mac->san_addr)) {
6644 		rtnl_lock();
6645 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
6646 		rtnl_unlock();
6647 	}
6648 	return err;
6649 }
6650 
6651 #ifdef CONFIG_NET_POLL_CONTROLLER
6652 /*
6653  * Polling 'interrupt' - used by things like netconsole to send skbs
6654  * without having to re-enable interrupts. It's not called while
6655  * the interrupt routine is executing.
6656  */
6657 static void ixgbe_netpoll(struct net_device *netdev)
6658 {
6659 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6660 	int i;
6661 
6662 	/* if interface is down do nothing */
6663 	if (test_bit(__IXGBE_DOWN, &adapter->state))
6664 		return;
6665 
6666 	adapter->flags |= IXGBE_FLAG_IN_NETPOLL;
6667 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
6668 		for (i = 0; i < adapter->num_q_vectors; i++)
6669 			ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
6670 	} else {
6671 		ixgbe_intr(adapter->pdev->irq, netdev);
6672 	}
6673 	adapter->flags &= ~IXGBE_FLAG_IN_NETPOLL;
6674 }
6675 
6676 #endif
6677 static struct rtnl_link_stats64 *ixgbe_get_stats64(struct net_device *netdev,
6678 						   struct rtnl_link_stats64 *stats)
6679 {
6680 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6681 	int i;
6682 
6683 	rcu_read_lock();
6684 	for (i = 0; i < adapter->num_rx_queues; i++) {
6685 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
6686 		u64 bytes, packets;
6687 		unsigned int start;
6688 
6689 		if (ring) {
6690 			do {
6691 				start = u64_stats_fetch_begin_bh(&ring->syncp);
6692 				packets = ring->stats.packets;
6693 				bytes   = ring->stats.bytes;
6694 			} while (u64_stats_fetch_retry_bh(&ring->syncp, start));
6695 			stats->rx_packets += packets;
6696 			stats->rx_bytes   += bytes;
6697 		}
6698 	}
6699 
6700 	for (i = 0; i < adapter->num_tx_queues; i++) {
6701 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
6702 		u64 bytes, packets;
6703 		unsigned int start;
6704 
6705 		if (ring) {
6706 			do {
6707 				start = u64_stats_fetch_begin_bh(&ring->syncp);
6708 				packets = ring->stats.packets;
6709 				bytes   = ring->stats.bytes;
6710 			} while (u64_stats_fetch_retry_bh(&ring->syncp, start));
6711 			stats->tx_packets += packets;
6712 			stats->tx_bytes   += bytes;
6713 		}
6714 	}
6715 	rcu_read_unlock();
6716 	/* following stats updated by ixgbe_watchdog_task() */
6717 	stats->multicast	= netdev->stats.multicast;
6718 	stats->rx_errors	= netdev->stats.rx_errors;
6719 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
6720 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
6721 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
6722 	return stats;
6723 }
6724 
6725 #ifdef CONFIG_IXGBE_DCB
6726 /**
6727  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
6728  * @adapter: pointer to ixgbe_adapter
6729  * @tc: number of traffic classes currently enabled
6730  *
6731  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
6732  * 802.1Q priority maps to a packet buffer that exists.
6733  */
6734 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
6735 {
6736 	struct ixgbe_hw *hw = &adapter->hw;
6737 	u32 reg, rsave;
6738 	int i;
6739 
6740 	/* 82598 have a static priority to TC mapping that can not
6741 	 * be changed so no validation is needed.
6742 	 */
6743 	if (hw->mac.type == ixgbe_mac_82598EB)
6744 		return;
6745 
6746 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
6747 	rsave = reg;
6748 
6749 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
6750 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
6751 
6752 		/* If up2tc is out of bounds default to zero */
6753 		if (up2tc > tc)
6754 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
6755 	}
6756 
6757 	if (reg != rsave)
6758 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
6759 
6760 	return;
6761 }
6762 
6763 /**
6764  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
6765  * @adapter: Pointer to adapter struct
6766  *
6767  * Populate the netdev user priority to tc map
6768  */
6769 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
6770 {
6771 	struct net_device *dev = adapter->netdev;
6772 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
6773 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
6774 	u8 prio;
6775 
6776 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
6777 		u8 tc = 0;
6778 
6779 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
6780 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
6781 		else if (ets)
6782 			tc = ets->prio_tc[prio];
6783 
6784 		netdev_set_prio_tc_map(dev, prio, tc);
6785 	}
6786 }
6787 
6788 /**
6789  * ixgbe_setup_tc - configure net_device for multiple traffic classes
6790  *
6791  * @netdev: net device to configure
6792  * @tc: number of traffic classes to enable
6793  */
6794 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
6795 {
6796 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6797 	struct ixgbe_hw *hw = &adapter->hw;
6798 
6799 	/* Hardware supports up to 8 traffic classes */
6800 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs ||
6801 	    (hw->mac.type == ixgbe_mac_82598EB &&
6802 	     tc < MAX_TRAFFIC_CLASS))
6803 		return -EINVAL;
6804 
6805 	/* Hardware has to reinitialize queues and interrupts to
6806 	 * match packet buffer alignment. Unfortunately, the
6807 	 * hardware is not flexible enough to do this dynamically.
6808 	 */
6809 	if (netif_running(dev))
6810 		ixgbe_close(dev);
6811 	ixgbe_clear_interrupt_scheme(adapter);
6812 
6813 	if (tc) {
6814 		netdev_set_num_tc(dev, tc);
6815 		ixgbe_set_prio_tc_map(adapter);
6816 
6817 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
6818 
6819 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
6820 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
6821 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
6822 		}
6823 	} else {
6824 		netdev_reset_tc(dev);
6825 
6826 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
6827 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
6828 
6829 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
6830 
6831 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
6832 		adapter->dcb_cfg.pfc_mode_enable = false;
6833 	}
6834 
6835 	ixgbe_init_interrupt_scheme(adapter);
6836 	ixgbe_validate_rtr(adapter, tc);
6837 	if (netif_running(dev))
6838 		ixgbe_open(dev);
6839 
6840 	return 0;
6841 }
6842 
6843 #endif /* CONFIG_IXGBE_DCB */
6844 void ixgbe_do_reset(struct net_device *netdev)
6845 {
6846 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6847 
6848 	if (netif_running(netdev))
6849 		ixgbe_reinit_locked(adapter);
6850 	else
6851 		ixgbe_reset(adapter);
6852 }
6853 
6854 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
6855 					    netdev_features_t features)
6856 {
6857 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6858 
6859 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
6860 	if (!(features & NETIF_F_RXCSUM))
6861 		features &= ~NETIF_F_LRO;
6862 
6863 	/* Turn off LRO if not RSC capable */
6864 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
6865 		features &= ~NETIF_F_LRO;
6866 
6867 	return features;
6868 }
6869 
6870 static int ixgbe_set_features(struct net_device *netdev,
6871 			      netdev_features_t features)
6872 {
6873 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6874 	netdev_features_t changed = netdev->features ^ features;
6875 	bool need_reset = false;
6876 
6877 	/* Make sure RSC matches LRO, reset if change */
6878 	if (!(features & NETIF_F_LRO)) {
6879 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
6880 			need_reset = true;
6881 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
6882 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
6883 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
6884 		if (adapter->rx_itr_setting == 1 ||
6885 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
6886 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
6887 			need_reset = true;
6888 		} else if ((changed ^ features) & NETIF_F_LRO) {
6889 			e_info(probe, "rx-usecs set too low, "
6890 			       "disabling RSC\n");
6891 		}
6892 	}
6893 
6894 	/*
6895 	 * Check if Flow Director n-tuple support was enabled or disabled.  If
6896 	 * the state changed, we need to reset.
6897 	 */
6898 	switch (features & NETIF_F_NTUPLE) {
6899 	case NETIF_F_NTUPLE:
6900 		/* turn off ATR, enable perfect filters and reset */
6901 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
6902 			need_reset = true;
6903 
6904 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
6905 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
6906 		break;
6907 	default:
6908 		/* turn off perfect filters, enable ATR and reset */
6909 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
6910 			need_reset = true;
6911 
6912 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
6913 
6914 		/* We cannot enable ATR if SR-IOV is enabled */
6915 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
6916 			break;
6917 
6918 		/* We cannot enable ATR if we have 2 or more traffic classes */
6919 		if (netdev_get_num_tc(netdev) > 1)
6920 			break;
6921 
6922 		/* We cannot enable ATR if RSS is disabled */
6923 		if (adapter->ring_feature[RING_F_RSS].limit <= 1)
6924 			break;
6925 
6926 		/* A sample rate of 0 indicates ATR disabled */
6927 		if (!adapter->atr_sample_rate)
6928 			break;
6929 
6930 		adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
6931 		break;
6932 	}
6933 
6934 	if (features & NETIF_F_HW_VLAN_RX)
6935 		ixgbe_vlan_strip_enable(adapter);
6936 	else
6937 		ixgbe_vlan_strip_disable(adapter);
6938 
6939 	if (changed & NETIF_F_RXALL)
6940 		need_reset = true;
6941 
6942 	netdev->features = features;
6943 	if (need_reset)
6944 		ixgbe_do_reset(netdev);
6945 
6946 	return 0;
6947 }
6948 
6949 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
6950 			     struct net_device *dev,
6951 			     const unsigned char *addr,
6952 			     u16 flags)
6953 {
6954 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6955 	int err;
6956 
6957 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
6958 		return -EOPNOTSUPP;
6959 
6960 	/* Hardware does not support aging addresses so if a
6961 	 * ndm_state is given only allow permanent addresses
6962 	 */
6963 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
6964 		pr_info("%s: FDB only supports static addresses\n",
6965 			ixgbe_driver_name);
6966 		return -EINVAL;
6967 	}
6968 
6969 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
6970 		u32 rar_uc_entries = IXGBE_MAX_PF_MACVLANS;
6971 
6972 		if (netdev_uc_count(dev) < rar_uc_entries)
6973 			err = dev_uc_add_excl(dev, addr);
6974 		else
6975 			err = -ENOMEM;
6976 	} else if (is_multicast_ether_addr(addr)) {
6977 		err = dev_mc_add_excl(dev, addr);
6978 	} else {
6979 		err = -EINVAL;
6980 	}
6981 
6982 	/* Only return duplicate errors if NLM_F_EXCL is set */
6983 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
6984 		err = 0;
6985 
6986 	return err;
6987 }
6988 
6989 static int ixgbe_ndo_fdb_del(struct ndmsg *ndm,
6990 			     struct net_device *dev,
6991 			     const unsigned char *addr)
6992 {
6993 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6994 	int err = -EOPNOTSUPP;
6995 
6996 	if (ndm->ndm_state & NUD_PERMANENT) {
6997 		pr_info("%s: FDB only supports static addresses\n",
6998 			ixgbe_driver_name);
6999 		return -EINVAL;
7000 	}
7001 
7002 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
7003 		if (is_unicast_ether_addr(addr))
7004 			err = dev_uc_del(dev, addr);
7005 		else if (is_multicast_ether_addr(addr))
7006 			err = dev_mc_del(dev, addr);
7007 		else
7008 			err = -EINVAL;
7009 	}
7010 
7011 	return err;
7012 }
7013 
7014 static int ixgbe_ndo_fdb_dump(struct sk_buff *skb,
7015 			      struct netlink_callback *cb,
7016 			      struct net_device *dev,
7017 			      int idx)
7018 {
7019 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7020 
7021 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7022 		idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
7023 
7024 	return idx;
7025 }
7026 
7027 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
7028 				    struct nlmsghdr *nlh)
7029 {
7030 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7031 	struct nlattr *attr, *br_spec;
7032 	int rem;
7033 
7034 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
7035 		return -EOPNOTSUPP;
7036 
7037 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
7038 
7039 	nla_for_each_nested(attr, br_spec, rem) {
7040 		__u16 mode;
7041 		u32 reg = 0;
7042 
7043 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
7044 			continue;
7045 
7046 		mode = nla_get_u16(attr);
7047 		if (mode == BRIDGE_MODE_VEPA) {
7048 			reg = 0;
7049 			adapter->flags2 &= ~IXGBE_FLAG2_BRIDGE_MODE_VEB;
7050 		} else if (mode == BRIDGE_MODE_VEB) {
7051 			reg = IXGBE_PFDTXGSWC_VT_LBEN;
7052 			adapter->flags2 |= IXGBE_FLAG2_BRIDGE_MODE_VEB;
7053 		} else
7054 			return -EINVAL;
7055 
7056 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, reg);
7057 
7058 		e_info(drv, "enabling bridge mode: %s\n",
7059 			mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
7060 	}
7061 
7062 	return 0;
7063 }
7064 
7065 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
7066 				    struct net_device *dev)
7067 {
7068 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7069 	u16 mode;
7070 
7071 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
7072 		return 0;
7073 
7074 	if (adapter->flags2 & IXGBE_FLAG2_BRIDGE_MODE_VEB)
7075 		mode = BRIDGE_MODE_VEB;
7076 	else
7077 		mode = BRIDGE_MODE_VEPA;
7078 
7079 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev, mode);
7080 }
7081 
7082 static const struct net_device_ops ixgbe_netdev_ops = {
7083 	.ndo_open		= ixgbe_open,
7084 	.ndo_stop		= ixgbe_close,
7085 	.ndo_start_xmit		= ixgbe_xmit_frame,
7086 	.ndo_select_queue	= ixgbe_select_queue,
7087 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
7088 	.ndo_validate_addr	= eth_validate_addr,
7089 	.ndo_set_mac_address	= ixgbe_set_mac,
7090 	.ndo_change_mtu		= ixgbe_change_mtu,
7091 	.ndo_tx_timeout		= ixgbe_tx_timeout,
7092 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
7093 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
7094 	.ndo_do_ioctl		= ixgbe_ioctl,
7095 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
7096 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
7097 	.ndo_set_vf_tx_rate	= ixgbe_ndo_set_vf_bw,
7098 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
7099 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
7100 	.ndo_get_stats64	= ixgbe_get_stats64,
7101 #ifdef CONFIG_IXGBE_DCB
7102 	.ndo_setup_tc		= ixgbe_setup_tc,
7103 #endif
7104 #ifdef CONFIG_NET_POLL_CONTROLLER
7105 	.ndo_poll_controller	= ixgbe_netpoll,
7106 #endif
7107 #ifdef IXGBE_FCOE
7108 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
7109 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
7110 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
7111 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
7112 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
7113 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
7114 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
7115 #endif /* IXGBE_FCOE */
7116 	.ndo_set_features = ixgbe_set_features,
7117 	.ndo_fix_features = ixgbe_fix_features,
7118 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
7119 	.ndo_fdb_del		= ixgbe_ndo_fdb_del,
7120 	.ndo_fdb_dump		= ixgbe_ndo_fdb_dump,
7121 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
7122 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
7123 };
7124 
7125 /**
7126  * ixgbe_wol_supported - Check whether device supports WoL
7127  * @hw: hw specific details
7128  * @device_id: the device ID
7129  * @subdev_id: the subsystem device ID
7130  *
7131  * This function is used by probe and ethtool to determine
7132  * which devices have WoL support
7133  *
7134  **/
7135 int ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
7136 			u16 subdevice_id)
7137 {
7138 	struct ixgbe_hw *hw = &adapter->hw;
7139 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
7140 	int is_wol_supported = 0;
7141 
7142 	switch (device_id) {
7143 	case IXGBE_DEV_ID_82599_SFP:
7144 		/* Only these subdevices could supports WOL */
7145 		switch (subdevice_id) {
7146 		case IXGBE_SUBDEV_ID_82599_560FLR:
7147 			/* only support first port */
7148 			if (hw->bus.func != 0)
7149 				break;
7150 		case IXGBE_SUBDEV_ID_82599_SFP:
7151 		case IXGBE_SUBDEV_ID_82599_RNDC:
7152 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
7153 			is_wol_supported = 1;
7154 			break;
7155 		}
7156 		break;
7157 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
7158 		/* All except this subdevice support WOL */
7159 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
7160 			is_wol_supported = 1;
7161 		break;
7162 	case IXGBE_DEV_ID_82599_KX4:
7163 		is_wol_supported = 1;
7164 		break;
7165 	case IXGBE_DEV_ID_X540T:
7166 	case IXGBE_DEV_ID_X540T1:
7167 		/* check eeprom to see if enabled wol */
7168 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
7169 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
7170 		     (hw->bus.func == 0))) {
7171 			is_wol_supported = 1;
7172 		}
7173 		break;
7174 	}
7175 
7176 	return is_wol_supported;
7177 }
7178 
7179 /**
7180  * ixgbe_probe - Device Initialization Routine
7181  * @pdev: PCI device information struct
7182  * @ent: entry in ixgbe_pci_tbl
7183  *
7184  * Returns 0 on success, negative on failure
7185  *
7186  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
7187  * The OS initialization, configuring of the adapter private structure,
7188  * and a hardware reset occur.
7189  **/
7190 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7191 {
7192 	struct net_device *netdev;
7193 	struct ixgbe_adapter *adapter = NULL;
7194 	struct ixgbe_hw *hw;
7195 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
7196 	static int cards_found;
7197 	int i, err, pci_using_dac;
7198 	u8 part_str[IXGBE_PBANUM_LENGTH];
7199 	unsigned int indices = num_possible_cpus();
7200 	unsigned int dcb_max = 0;
7201 #ifdef IXGBE_FCOE
7202 	u16 device_caps;
7203 #endif
7204 	u32 eec;
7205 
7206 	/* Catch broken hardware that put the wrong VF device ID in
7207 	 * the PCIe SR-IOV capability.
7208 	 */
7209 	if (pdev->is_virtfn) {
7210 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
7211 		     pci_name(pdev), pdev->vendor, pdev->device);
7212 		return -EINVAL;
7213 	}
7214 
7215 	err = pci_enable_device_mem(pdev);
7216 	if (err)
7217 		return err;
7218 
7219 	if (!dma_set_mask(&pdev->dev, DMA_BIT_MASK(64)) &&
7220 	    !dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(64))) {
7221 		pci_using_dac = 1;
7222 	} else {
7223 		err = dma_set_mask(&pdev->dev, DMA_BIT_MASK(32));
7224 		if (err) {
7225 			err = dma_set_coherent_mask(&pdev->dev,
7226 						    DMA_BIT_MASK(32));
7227 			if (err) {
7228 				dev_err(&pdev->dev,
7229 					"No usable DMA configuration, aborting\n");
7230 				goto err_dma;
7231 			}
7232 		}
7233 		pci_using_dac = 0;
7234 	}
7235 
7236 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
7237 					   IORESOURCE_MEM), ixgbe_driver_name);
7238 	if (err) {
7239 		dev_err(&pdev->dev,
7240 			"pci_request_selected_regions failed 0x%x\n", err);
7241 		goto err_pci_reg;
7242 	}
7243 
7244 	pci_enable_pcie_error_reporting(pdev);
7245 
7246 	pci_set_master(pdev);
7247 	pci_save_state(pdev);
7248 
7249 #ifdef CONFIG_IXGBE_DCB
7250 	if (ii->mac == ixgbe_mac_82598EB)
7251 		dcb_max = min_t(unsigned int, indices * MAX_TRAFFIC_CLASS,
7252 				IXGBE_MAX_RSS_INDICES);
7253 	else
7254 		dcb_max = min_t(unsigned int, indices * MAX_TRAFFIC_CLASS,
7255 				IXGBE_MAX_FDIR_INDICES);
7256 #endif
7257 
7258 	if (ii->mac == ixgbe_mac_82598EB)
7259 		indices = min_t(unsigned int, indices, IXGBE_MAX_RSS_INDICES);
7260 	else
7261 		indices = min_t(unsigned int, indices, IXGBE_MAX_FDIR_INDICES);
7262 
7263 #ifdef IXGBE_FCOE
7264 	indices += min_t(unsigned int, num_possible_cpus(),
7265 			 IXGBE_MAX_FCOE_INDICES);
7266 #endif
7267 	indices = max_t(unsigned int, dcb_max, indices);
7268 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
7269 	if (!netdev) {
7270 		err = -ENOMEM;
7271 		goto err_alloc_etherdev;
7272 	}
7273 
7274 	SET_NETDEV_DEV(netdev, &pdev->dev);
7275 
7276 	adapter = netdev_priv(netdev);
7277 	pci_set_drvdata(pdev, adapter);
7278 
7279 	adapter->netdev = netdev;
7280 	adapter->pdev = pdev;
7281 	hw = &adapter->hw;
7282 	hw->back = adapter;
7283 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
7284 
7285 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
7286 			      pci_resource_len(pdev, 0));
7287 	if (!hw->hw_addr) {
7288 		err = -EIO;
7289 		goto err_ioremap;
7290 	}
7291 
7292 	netdev->netdev_ops = &ixgbe_netdev_ops;
7293 	ixgbe_set_ethtool_ops(netdev);
7294 	netdev->watchdog_timeo = 5 * HZ;
7295 	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
7296 
7297 	adapter->bd_number = cards_found;
7298 
7299 	/* Setup hw api */
7300 	memcpy(&hw->mac.ops, ii->mac_ops, sizeof(hw->mac.ops));
7301 	hw->mac.type  = ii->mac;
7302 
7303 	/* EEPROM */
7304 	memcpy(&hw->eeprom.ops, ii->eeprom_ops, sizeof(hw->eeprom.ops));
7305 	eec = IXGBE_READ_REG(hw, IXGBE_EEC);
7306 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
7307 	if (!(eec & (1 << 8)))
7308 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
7309 
7310 	/* PHY */
7311 	memcpy(&hw->phy.ops, ii->phy_ops, sizeof(hw->phy.ops));
7312 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
7313 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
7314 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
7315 	hw->phy.mdio.mmds = 0;
7316 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
7317 	hw->phy.mdio.dev = netdev;
7318 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
7319 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
7320 
7321 	ii->get_invariants(hw);
7322 
7323 	/* setup the private structure */
7324 	err = ixgbe_sw_init(adapter);
7325 	if (err)
7326 		goto err_sw_init;
7327 
7328 	/* Make it possible the adapter to be woken up via WOL */
7329 	switch (adapter->hw.mac.type) {
7330 	case ixgbe_mac_82599EB:
7331 	case ixgbe_mac_X540:
7332 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
7333 		break;
7334 	default:
7335 		break;
7336 	}
7337 
7338 	/*
7339 	 * If there is a fan on this device and it has failed log the
7340 	 * failure.
7341 	 */
7342 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
7343 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
7344 		if (esdp & IXGBE_ESDP_SDP1)
7345 			e_crit(probe, "Fan has stopped, replace the adapter\n");
7346 	}
7347 
7348 	if (allow_unsupported_sfp)
7349 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
7350 
7351 	/* reset_hw fills in the perm_addr as well */
7352 	hw->phy.reset_if_overtemp = true;
7353 	err = hw->mac.ops.reset_hw(hw);
7354 	hw->phy.reset_if_overtemp = false;
7355 	if (err == IXGBE_ERR_SFP_NOT_PRESENT &&
7356 	    hw->mac.type == ixgbe_mac_82598EB) {
7357 		err = 0;
7358 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
7359 		e_dev_err("failed to load because an unsupported SFP+ "
7360 			  "module type was detected.\n");
7361 		e_dev_err("Reload the driver after installing a supported "
7362 			  "module.\n");
7363 		goto err_sw_init;
7364 	} else if (err) {
7365 		e_dev_err("HW Init failed: %d\n", err);
7366 		goto err_sw_init;
7367 	}
7368 
7369 #ifdef CONFIG_PCI_IOV
7370 	ixgbe_enable_sriov(adapter, ii);
7371 
7372 #endif
7373 	netdev->features = NETIF_F_SG |
7374 			   NETIF_F_IP_CSUM |
7375 			   NETIF_F_IPV6_CSUM |
7376 			   NETIF_F_HW_VLAN_TX |
7377 			   NETIF_F_HW_VLAN_RX |
7378 			   NETIF_F_HW_VLAN_FILTER |
7379 			   NETIF_F_TSO |
7380 			   NETIF_F_TSO6 |
7381 			   NETIF_F_RXHASH |
7382 			   NETIF_F_RXCSUM;
7383 
7384 	netdev->hw_features = netdev->features;
7385 
7386 	switch (adapter->hw.mac.type) {
7387 	case ixgbe_mac_82599EB:
7388 	case ixgbe_mac_X540:
7389 		netdev->features |= NETIF_F_SCTP_CSUM;
7390 		netdev->hw_features |= NETIF_F_SCTP_CSUM |
7391 				       NETIF_F_NTUPLE;
7392 		break;
7393 	default:
7394 		break;
7395 	}
7396 
7397 	netdev->hw_features |= NETIF_F_RXALL;
7398 
7399 	netdev->vlan_features |= NETIF_F_TSO;
7400 	netdev->vlan_features |= NETIF_F_TSO6;
7401 	netdev->vlan_features |= NETIF_F_IP_CSUM;
7402 	netdev->vlan_features |= NETIF_F_IPV6_CSUM;
7403 	netdev->vlan_features |= NETIF_F_SG;
7404 
7405 	netdev->priv_flags |= IFF_UNICAST_FLT;
7406 	netdev->priv_flags |= IFF_SUPP_NOFCS;
7407 
7408 #ifdef CONFIG_IXGBE_DCB
7409 	netdev->dcbnl_ops = &dcbnl_ops;
7410 #endif
7411 
7412 #ifdef IXGBE_FCOE
7413 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
7414 		if (hw->mac.ops.get_device_caps) {
7415 			hw->mac.ops.get_device_caps(hw, &device_caps);
7416 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
7417 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
7418 		}
7419 
7420 		adapter->ring_feature[RING_F_FCOE].limit = IXGBE_FCRETA_SIZE;
7421 
7422 		netdev->features |= NETIF_F_FSO |
7423 				    NETIF_F_FCOE_CRC;
7424 
7425 		netdev->vlan_features |= NETIF_F_FSO |
7426 					 NETIF_F_FCOE_CRC |
7427 					 NETIF_F_FCOE_MTU;
7428 	}
7429 #endif /* IXGBE_FCOE */
7430 	if (pci_using_dac) {
7431 		netdev->features |= NETIF_F_HIGHDMA;
7432 		netdev->vlan_features |= NETIF_F_HIGHDMA;
7433 	}
7434 
7435 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
7436 		netdev->hw_features |= NETIF_F_LRO;
7437 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
7438 		netdev->features |= NETIF_F_LRO;
7439 
7440 	/* make sure the EEPROM is good */
7441 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
7442 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
7443 		err = -EIO;
7444 		goto err_sw_init;
7445 	}
7446 
7447 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
7448 	memcpy(netdev->perm_addr, hw->mac.perm_addr, netdev->addr_len);
7449 
7450 	if (!is_valid_ether_addr(netdev->perm_addr)) {
7451 		e_dev_err("invalid MAC address\n");
7452 		err = -EIO;
7453 		goto err_sw_init;
7454 	}
7455 
7456 	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
7457 		    (unsigned long) adapter);
7458 
7459 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
7460 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
7461 
7462 	err = ixgbe_init_interrupt_scheme(adapter);
7463 	if (err)
7464 		goto err_sw_init;
7465 
7466 	/* WOL not supported for all devices */
7467 	adapter->wol = 0;
7468 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
7469 	if (ixgbe_wol_supported(adapter, pdev->device, pdev->subsystem_device))
7470 		adapter->wol = IXGBE_WUFC_MAG;
7471 
7472 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
7473 
7474 	/* save off EEPROM version number */
7475 	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
7476 	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
7477 
7478 	/* pick up the PCI bus settings for reporting later */
7479 	hw->mac.ops.get_bus_info(hw);
7480 
7481 	/* print bus type/speed/width info */
7482 	e_dev_info("(PCI Express:%s:%s) %pM\n",
7483 		   (hw->bus.speed == ixgbe_bus_speed_5000 ? "5.0GT/s" :
7484 		    hw->bus.speed == ixgbe_bus_speed_2500 ? "2.5GT/s" :
7485 		    "Unknown"),
7486 		   (hw->bus.width == ixgbe_bus_width_pcie_x8 ? "Width x8" :
7487 		    hw->bus.width == ixgbe_bus_width_pcie_x4 ? "Width x4" :
7488 		    hw->bus.width == ixgbe_bus_width_pcie_x1 ? "Width x1" :
7489 		    "Unknown"),
7490 		   netdev->dev_addr);
7491 
7492 	err = ixgbe_read_pba_string_generic(hw, part_str, IXGBE_PBANUM_LENGTH);
7493 	if (err)
7494 		strncpy(part_str, "Unknown", IXGBE_PBANUM_LENGTH);
7495 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
7496 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
7497 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
7498 		           part_str);
7499 	else
7500 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
7501 			   hw->mac.type, hw->phy.type, part_str);
7502 
7503 	if (hw->bus.width <= ixgbe_bus_width_pcie_x4) {
7504 		e_dev_warn("PCI-Express bandwidth available for this card is "
7505 			   "not sufficient for optimal performance.\n");
7506 		e_dev_warn("For optimal performance a x8 PCI-Express slot "
7507 			   "is required.\n");
7508 	}
7509 
7510 	/* reset the hardware with the new settings */
7511 	err = hw->mac.ops.start_hw(hw);
7512 	if (err == IXGBE_ERR_EEPROM_VERSION) {
7513 		/* We are running on a pre-production device, log a warning */
7514 		e_dev_warn("This device is a pre-production adapter/LOM. "
7515 			   "Please be aware there may be issues associated "
7516 			   "with your hardware.  If you are experiencing "
7517 			   "problems please contact your Intel or hardware "
7518 			   "representative who provided you with this "
7519 			   "hardware.\n");
7520 	}
7521 	strcpy(netdev->name, "eth%d");
7522 	err = register_netdev(netdev);
7523 	if (err)
7524 		goto err_register;
7525 
7526 	/* power down the optics for 82599 SFP+ fiber */
7527 	if (hw->mac.ops.disable_tx_laser)
7528 		hw->mac.ops.disable_tx_laser(hw);
7529 
7530 	/* carrier off reporting is important to ethtool even BEFORE open */
7531 	netif_carrier_off(netdev);
7532 
7533 #ifdef CONFIG_IXGBE_DCA
7534 	if (dca_add_requester(&pdev->dev) == 0) {
7535 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
7536 		ixgbe_setup_dca(adapter);
7537 	}
7538 #endif
7539 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
7540 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
7541 		for (i = 0; i < adapter->num_vfs; i++)
7542 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
7543 	}
7544 
7545 	/* firmware requires driver version to be 0xFFFFFFFF
7546 	 * since os does not support feature
7547 	 */
7548 	if (hw->mac.ops.set_fw_drv_ver)
7549 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF,
7550 					   0xFF);
7551 
7552 	/* add san mac addr to netdev */
7553 	ixgbe_add_sanmac_netdev(netdev);
7554 
7555 	e_dev_info("%s\n", ixgbe_default_device_descr);
7556 	cards_found++;
7557 
7558 #ifdef CONFIG_IXGBE_HWMON
7559 	if (ixgbe_sysfs_init(adapter))
7560 		e_err(probe, "failed to allocate sysfs resources\n");
7561 #endif /* CONFIG_IXGBE_HWMON */
7562 
7563 #ifdef CONFIG_DEBUG_FS
7564 	ixgbe_dbg_adapter_init(adapter);
7565 #endif /* CONFIG_DEBUG_FS */
7566 
7567 	return 0;
7568 
7569 err_register:
7570 	ixgbe_release_hw_control(adapter);
7571 	ixgbe_clear_interrupt_scheme(adapter);
7572 err_sw_init:
7573 	ixgbe_disable_sriov(adapter);
7574 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
7575 	iounmap(hw->hw_addr);
7576 err_ioremap:
7577 	free_netdev(netdev);
7578 err_alloc_etherdev:
7579 	pci_release_selected_regions(pdev,
7580 				     pci_select_bars(pdev, IORESOURCE_MEM));
7581 err_pci_reg:
7582 err_dma:
7583 	pci_disable_device(pdev);
7584 	return err;
7585 }
7586 
7587 /**
7588  * ixgbe_remove - Device Removal Routine
7589  * @pdev: PCI device information struct
7590  *
7591  * ixgbe_remove is called by the PCI subsystem to alert the driver
7592  * that it should release a PCI device.  The could be caused by a
7593  * Hot-Plug event, or because the driver is going to be removed from
7594  * memory.
7595  **/
7596 static void ixgbe_remove(struct pci_dev *pdev)
7597 {
7598 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7599 	struct net_device *netdev = adapter->netdev;
7600 
7601 #ifdef CONFIG_DEBUG_FS
7602 	ixgbe_dbg_adapter_exit(adapter);
7603 #endif /*CONFIG_DEBUG_FS */
7604 
7605 	set_bit(__IXGBE_DOWN, &adapter->state);
7606 	cancel_work_sync(&adapter->service_task);
7607 
7608 
7609 #ifdef CONFIG_IXGBE_DCA
7610 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
7611 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
7612 		dca_remove_requester(&pdev->dev);
7613 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
7614 	}
7615 
7616 #endif
7617 #ifdef CONFIG_IXGBE_HWMON
7618 	ixgbe_sysfs_exit(adapter);
7619 #endif /* CONFIG_IXGBE_HWMON */
7620 
7621 	/* remove the added san mac */
7622 	ixgbe_del_sanmac_netdev(netdev);
7623 
7624 	if (netdev->reg_state == NETREG_REGISTERED)
7625 		unregister_netdev(netdev);
7626 
7627 	ixgbe_disable_sriov(adapter);
7628 
7629 	ixgbe_clear_interrupt_scheme(adapter);
7630 
7631 	ixgbe_release_hw_control(adapter);
7632 
7633 #ifdef CONFIG_DCB
7634 	kfree(adapter->ixgbe_ieee_pfc);
7635 	kfree(adapter->ixgbe_ieee_ets);
7636 
7637 #endif
7638 	iounmap(adapter->hw.hw_addr);
7639 	pci_release_selected_regions(pdev, pci_select_bars(pdev,
7640 				     IORESOURCE_MEM));
7641 
7642 	e_dev_info("complete\n");
7643 
7644 	free_netdev(netdev);
7645 
7646 	pci_disable_pcie_error_reporting(pdev);
7647 
7648 	pci_disable_device(pdev);
7649 }
7650 
7651 /**
7652  * ixgbe_io_error_detected - called when PCI error is detected
7653  * @pdev: Pointer to PCI device
7654  * @state: The current pci connection state
7655  *
7656  * This function is called after a PCI bus error affecting
7657  * this device has been detected.
7658  */
7659 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
7660 						pci_channel_state_t state)
7661 {
7662 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7663 	struct net_device *netdev = adapter->netdev;
7664 
7665 #ifdef CONFIG_PCI_IOV
7666 	struct pci_dev *bdev, *vfdev;
7667 	u32 dw0, dw1, dw2, dw3;
7668 	int vf, pos;
7669 	u16 req_id, pf_func;
7670 
7671 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7672 	    adapter->num_vfs == 0)
7673 		goto skip_bad_vf_detection;
7674 
7675 	bdev = pdev->bus->self;
7676 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
7677 		bdev = bdev->bus->self;
7678 
7679 	if (!bdev)
7680 		goto skip_bad_vf_detection;
7681 
7682 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
7683 	if (!pos)
7684 		goto skip_bad_vf_detection;
7685 
7686 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG, &dw0);
7687 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG + 4, &dw1);
7688 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG + 8, &dw2);
7689 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG + 12, &dw3);
7690 
7691 	req_id = dw1 >> 16;
7692 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
7693 	if (!(req_id & 0x0080))
7694 		goto skip_bad_vf_detection;
7695 
7696 	pf_func = req_id & 0x01;
7697 	if ((pf_func & 1) == (pdev->devfn & 1)) {
7698 		unsigned int device_id;
7699 
7700 		vf = (req_id & 0x7F) >> 1;
7701 		e_dev_err("VF %d has caused a PCIe error\n", vf);
7702 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
7703 				"%8.8x\tdw3: %8.8x\n",
7704 		dw0, dw1, dw2, dw3);
7705 		switch (adapter->hw.mac.type) {
7706 		case ixgbe_mac_82599EB:
7707 			device_id = IXGBE_82599_VF_DEVICE_ID;
7708 			break;
7709 		case ixgbe_mac_X540:
7710 			device_id = IXGBE_X540_VF_DEVICE_ID;
7711 			break;
7712 		default:
7713 			device_id = 0;
7714 			break;
7715 		}
7716 
7717 		/* Find the pci device of the offending VF */
7718 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
7719 		while (vfdev) {
7720 			if (vfdev->devfn == (req_id & 0xFF))
7721 				break;
7722 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
7723 					       device_id, vfdev);
7724 		}
7725 		/*
7726 		 * There's a slim chance the VF could have been hot plugged,
7727 		 * so if it is no longer present we don't need to issue the
7728 		 * VFLR.  Just clean up the AER in that case.
7729 		 */
7730 		if (vfdev) {
7731 			e_dev_err("Issuing VFLR to VF %d\n", vf);
7732 			pci_write_config_dword(vfdev, 0xA8, 0x00008000);
7733 		}
7734 
7735 		pci_cleanup_aer_uncorrect_error_status(pdev);
7736 	}
7737 
7738 	/*
7739 	 * Even though the error may have occurred on the other port
7740 	 * we still need to increment the vf error reference count for
7741 	 * both ports because the I/O resume function will be called
7742 	 * for both of them.
7743 	 */
7744 	adapter->vferr_refcount++;
7745 
7746 	return PCI_ERS_RESULT_RECOVERED;
7747 
7748 skip_bad_vf_detection:
7749 #endif /* CONFIG_PCI_IOV */
7750 	netif_device_detach(netdev);
7751 
7752 	if (state == pci_channel_io_perm_failure)
7753 		return PCI_ERS_RESULT_DISCONNECT;
7754 
7755 	if (netif_running(netdev))
7756 		ixgbe_down(adapter);
7757 	pci_disable_device(pdev);
7758 
7759 	/* Request a slot reset. */
7760 	return PCI_ERS_RESULT_NEED_RESET;
7761 }
7762 
7763 /**
7764  * ixgbe_io_slot_reset - called after the pci bus has been reset.
7765  * @pdev: Pointer to PCI device
7766  *
7767  * Restart the card from scratch, as if from a cold-boot.
7768  */
7769 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
7770 {
7771 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7772 	pci_ers_result_t result;
7773 	int err;
7774 
7775 	if (pci_enable_device_mem(pdev)) {
7776 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
7777 		result = PCI_ERS_RESULT_DISCONNECT;
7778 	} else {
7779 		pci_set_master(pdev);
7780 		pci_restore_state(pdev);
7781 		pci_save_state(pdev);
7782 
7783 		pci_wake_from_d3(pdev, false);
7784 
7785 		ixgbe_reset(adapter);
7786 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
7787 		result = PCI_ERS_RESULT_RECOVERED;
7788 	}
7789 
7790 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
7791 	if (err) {
7792 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
7793 			  "failed 0x%0x\n", err);
7794 		/* non-fatal, continue */
7795 	}
7796 
7797 	return result;
7798 }
7799 
7800 /**
7801  * ixgbe_io_resume - called when traffic can start flowing again.
7802  * @pdev: Pointer to PCI device
7803  *
7804  * This callback is called when the error recovery driver tells us that
7805  * its OK to resume normal operation.
7806  */
7807 static void ixgbe_io_resume(struct pci_dev *pdev)
7808 {
7809 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7810 	struct net_device *netdev = adapter->netdev;
7811 
7812 #ifdef CONFIG_PCI_IOV
7813 	if (adapter->vferr_refcount) {
7814 		e_info(drv, "Resuming after VF err\n");
7815 		adapter->vferr_refcount--;
7816 		return;
7817 	}
7818 
7819 #endif
7820 	if (netif_running(netdev))
7821 		ixgbe_up(adapter);
7822 
7823 	netif_device_attach(netdev);
7824 }
7825 
7826 static const struct pci_error_handlers ixgbe_err_handler = {
7827 	.error_detected = ixgbe_io_error_detected,
7828 	.slot_reset = ixgbe_io_slot_reset,
7829 	.resume = ixgbe_io_resume,
7830 };
7831 
7832 static struct pci_driver ixgbe_driver = {
7833 	.name     = ixgbe_driver_name,
7834 	.id_table = ixgbe_pci_tbl,
7835 	.probe    = ixgbe_probe,
7836 	.remove   = ixgbe_remove,
7837 #ifdef CONFIG_PM
7838 	.suspend  = ixgbe_suspend,
7839 	.resume   = ixgbe_resume,
7840 #endif
7841 	.shutdown = ixgbe_shutdown,
7842 	.err_handler = &ixgbe_err_handler
7843 };
7844 
7845 /**
7846  * ixgbe_init_module - Driver Registration Routine
7847  *
7848  * ixgbe_init_module is the first routine called when the driver is
7849  * loaded. All it does is register with the PCI subsystem.
7850  **/
7851 static int __init ixgbe_init_module(void)
7852 {
7853 	int ret;
7854 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
7855 	pr_info("%s\n", ixgbe_copyright);
7856 
7857 #ifdef CONFIG_DEBUG_FS
7858 	ixgbe_dbg_init();
7859 #endif /* CONFIG_DEBUG_FS */
7860 
7861 #ifdef CONFIG_IXGBE_DCA
7862 	dca_register_notify(&dca_notifier);
7863 #endif
7864 
7865 	ret = pci_register_driver(&ixgbe_driver);
7866 	return ret;
7867 }
7868 
7869 module_init(ixgbe_init_module);
7870 
7871 /**
7872  * ixgbe_exit_module - Driver Exit Cleanup Routine
7873  *
7874  * ixgbe_exit_module is called just before the driver is removed
7875  * from memory.
7876  **/
7877 static void __exit ixgbe_exit_module(void)
7878 {
7879 #ifdef CONFIG_IXGBE_DCA
7880 	dca_unregister_notify(&dca_notifier);
7881 #endif
7882 	pci_unregister_driver(&ixgbe_driver);
7883 
7884 #ifdef CONFIG_DEBUG_FS
7885 	ixgbe_dbg_exit();
7886 #endif /* CONFIG_DEBUG_FS */
7887 
7888 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
7889 }
7890 
7891 #ifdef CONFIG_IXGBE_DCA
7892 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
7893 			    void *p)
7894 {
7895 	int ret_val;
7896 
7897 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
7898 					 __ixgbe_notify_dca);
7899 
7900 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
7901 }
7902 
7903 #endif /* CONFIG_IXGBE_DCA */
7904 
7905 module_exit(ixgbe_exit_module);
7906 
7907 /* ixgbe_main.c */
7908