1 /*******************************************************************************
2 
3   Intel 10 Gigabit PCI Express Linux driver
4   Copyright(c) 1999 - 2013 Intel Corporation.
5 
6   This program is free software; you can redistribute it and/or modify it
7   under the terms and conditions of the GNU General Public License,
8   version 2, as published by the Free Software Foundation.
9 
10   This program is distributed in the hope it will be useful, but WITHOUT
11   ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12   FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13   more details.
14 
15   You should have received a copy of the GNU General Public License along with
16   this program; if not, write to the Free Software Foundation, Inc.,
17   51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
18 
19   The full GNU General Public License is included in this distribution in
20   the file called "COPYING".
21 
22   Contact Information:
23   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 "3.13.10-k"
67 const char ixgbe_driver_version[] = DRV_VERSION;
68 static const char ixgbe_copyright[] =
69 				"Copyright (c) 1999-2013 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 int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
153 					  u32 reg, u16 *value)
154 {
155 	int pos = 0;
156 	struct pci_dev *parent_dev;
157 	struct pci_bus *parent_bus;
158 
159 	parent_bus = adapter->pdev->bus->parent;
160 	if (!parent_bus)
161 		return -1;
162 
163 	parent_dev = parent_bus->self;
164 	if (!parent_dev)
165 		return -1;
166 
167 	pos = pci_find_capability(parent_dev, PCI_CAP_ID_EXP);
168 	if (!pos)
169 		return -1;
170 
171 	pci_read_config_word(parent_dev, pos + reg, value);
172 	return 0;
173 }
174 
175 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
176 {
177 	struct ixgbe_hw *hw = &adapter->hw;
178 	u16 link_status = 0;
179 	int err;
180 
181 	hw->bus.type = ixgbe_bus_type_pci_express;
182 
183 	/* Get the negotiated link width and speed from PCI config space of the
184 	 * parent, as this device is behind a switch
185 	 */
186 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
187 
188 	/* assume caller will handle error case */
189 	if (err)
190 		return err;
191 
192 	hw->bus.width = ixgbe_convert_bus_width(link_status);
193 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
194 
195 	return 0;
196 }
197 
198 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
199 {
200 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
201 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
202 		schedule_work(&adapter->service_task);
203 }
204 
205 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
206 {
207 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
208 
209 	/* flush memory to make sure state is correct before next watchdog */
210 	smp_mb__before_clear_bit();
211 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
212 }
213 
214 struct ixgbe_reg_info {
215 	u32 ofs;
216 	char *name;
217 };
218 
219 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
220 
221 	/* General Registers */
222 	{IXGBE_CTRL, "CTRL"},
223 	{IXGBE_STATUS, "STATUS"},
224 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
225 
226 	/* Interrupt Registers */
227 	{IXGBE_EICR, "EICR"},
228 
229 	/* RX Registers */
230 	{IXGBE_SRRCTL(0), "SRRCTL"},
231 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
232 	{IXGBE_RDLEN(0), "RDLEN"},
233 	{IXGBE_RDH(0), "RDH"},
234 	{IXGBE_RDT(0), "RDT"},
235 	{IXGBE_RXDCTL(0), "RXDCTL"},
236 	{IXGBE_RDBAL(0), "RDBAL"},
237 	{IXGBE_RDBAH(0), "RDBAH"},
238 
239 	/* TX Registers */
240 	{IXGBE_TDBAL(0), "TDBAL"},
241 	{IXGBE_TDBAH(0), "TDBAH"},
242 	{IXGBE_TDLEN(0), "TDLEN"},
243 	{IXGBE_TDH(0), "TDH"},
244 	{IXGBE_TDT(0), "TDT"},
245 	{IXGBE_TXDCTL(0), "TXDCTL"},
246 
247 	/* List Terminator */
248 	{}
249 };
250 
251 
252 /*
253  * ixgbe_regdump - register printout routine
254  */
255 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
256 {
257 	int i = 0, j = 0;
258 	char rname[16];
259 	u32 regs[64];
260 
261 	switch (reginfo->ofs) {
262 	case IXGBE_SRRCTL(0):
263 		for (i = 0; i < 64; i++)
264 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
265 		break;
266 	case IXGBE_DCA_RXCTRL(0):
267 		for (i = 0; i < 64; i++)
268 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
269 		break;
270 	case IXGBE_RDLEN(0):
271 		for (i = 0; i < 64; i++)
272 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
273 		break;
274 	case IXGBE_RDH(0):
275 		for (i = 0; i < 64; i++)
276 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
277 		break;
278 	case IXGBE_RDT(0):
279 		for (i = 0; i < 64; i++)
280 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
281 		break;
282 	case IXGBE_RXDCTL(0):
283 		for (i = 0; i < 64; i++)
284 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
285 		break;
286 	case IXGBE_RDBAL(0):
287 		for (i = 0; i < 64; i++)
288 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
289 		break;
290 	case IXGBE_RDBAH(0):
291 		for (i = 0; i < 64; i++)
292 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
293 		break;
294 	case IXGBE_TDBAL(0):
295 		for (i = 0; i < 64; i++)
296 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
297 		break;
298 	case IXGBE_TDBAH(0):
299 		for (i = 0; i < 64; i++)
300 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
301 		break;
302 	case IXGBE_TDLEN(0):
303 		for (i = 0; i < 64; i++)
304 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
305 		break;
306 	case IXGBE_TDH(0):
307 		for (i = 0; i < 64; i++)
308 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
309 		break;
310 	case IXGBE_TDT(0):
311 		for (i = 0; i < 64; i++)
312 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
313 		break;
314 	case IXGBE_TXDCTL(0):
315 		for (i = 0; i < 64; i++)
316 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
317 		break;
318 	default:
319 		pr_info("%-15s %08x\n", reginfo->name,
320 			IXGBE_READ_REG(hw, reginfo->ofs));
321 		return;
322 	}
323 
324 	for (i = 0; i < 8; i++) {
325 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i*8, i*8+7);
326 		pr_err("%-15s", rname);
327 		for (j = 0; j < 8; j++)
328 			pr_cont(" %08x", regs[i*8+j]);
329 		pr_cont("\n");
330 	}
331 
332 }
333 
334 /*
335  * ixgbe_dump - Print registers, tx-rings and rx-rings
336  */
337 static void ixgbe_dump(struct ixgbe_adapter *adapter)
338 {
339 	struct net_device *netdev = adapter->netdev;
340 	struct ixgbe_hw *hw = &adapter->hw;
341 	struct ixgbe_reg_info *reginfo;
342 	int n = 0;
343 	struct ixgbe_ring *tx_ring;
344 	struct ixgbe_tx_buffer *tx_buffer;
345 	union ixgbe_adv_tx_desc *tx_desc;
346 	struct my_u0 { u64 a; u64 b; } *u0;
347 	struct ixgbe_ring *rx_ring;
348 	union ixgbe_adv_rx_desc *rx_desc;
349 	struct ixgbe_rx_buffer *rx_buffer_info;
350 	u32 staterr;
351 	int i = 0;
352 
353 	if (!netif_msg_hw(adapter))
354 		return;
355 
356 	/* Print netdevice Info */
357 	if (netdev) {
358 		dev_info(&adapter->pdev->dev, "Net device Info\n");
359 		pr_info("Device Name     state            "
360 			"trans_start      last_rx\n");
361 		pr_info("%-15s %016lX %016lX %016lX\n",
362 			netdev->name,
363 			netdev->state,
364 			netdev->trans_start,
365 			netdev->last_rx);
366 	}
367 
368 	/* Print Registers */
369 	dev_info(&adapter->pdev->dev, "Register Dump\n");
370 	pr_info(" Register Name   Value\n");
371 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
372 	     reginfo->name; reginfo++) {
373 		ixgbe_regdump(hw, reginfo);
374 	}
375 
376 	/* Print TX Ring Summary */
377 	if (!netdev || !netif_running(netdev))
378 		goto exit;
379 
380 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
381 	pr_info(" %s     %s              %s        %s\n",
382 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
383 		"leng", "ntw", "timestamp");
384 	for (n = 0; n < adapter->num_tx_queues; n++) {
385 		tx_ring = adapter->tx_ring[n];
386 		tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
387 		pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
388 			   n, tx_ring->next_to_use, tx_ring->next_to_clean,
389 			   (u64)dma_unmap_addr(tx_buffer, dma),
390 			   dma_unmap_len(tx_buffer, len),
391 			   tx_buffer->next_to_watch,
392 			   (u64)tx_buffer->time_stamp);
393 	}
394 
395 	/* Print TX Rings */
396 	if (!netif_msg_tx_done(adapter))
397 		goto rx_ring_summary;
398 
399 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
400 
401 	/* Transmit Descriptor Formats
402 	 *
403 	 * 82598 Advanced Transmit Descriptor
404 	 *   +--------------------------------------------------------------+
405 	 * 0 |         Buffer Address [63:0]                                |
406 	 *   +--------------------------------------------------------------+
407 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
408 	 *   +--------------------------------------------------------------+
409 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
410 	 *
411 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
412 	 *   +--------------------------------------------------------------+
413 	 * 0 |                          RSV [63:0]                          |
414 	 *   +--------------------------------------------------------------+
415 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
416 	 *   +--------------------------------------------------------------+
417 	 *   63                       36 35   32 31                         0
418 	 *
419 	 * 82599+ Advanced Transmit Descriptor
420 	 *   +--------------------------------------------------------------+
421 	 * 0 |         Buffer Address [63:0]                                |
422 	 *   +--------------------------------------------------------------+
423 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
424 	 *   +--------------------------------------------------------------+
425 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
426 	 *
427 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
428 	 *   +--------------------------------------------------------------+
429 	 * 0 |                          RSV [63:0]                          |
430 	 *   +--------------------------------------------------------------+
431 	 * 8 |            RSV           |  STA  |           RSV             |
432 	 *   +--------------------------------------------------------------+
433 	 *   63                       36 35   32 31                         0
434 	 */
435 
436 	for (n = 0; n < adapter->num_tx_queues; n++) {
437 		tx_ring = adapter->tx_ring[n];
438 		pr_info("------------------------------------\n");
439 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
440 		pr_info("------------------------------------\n");
441 		pr_info("%s%s    %s              %s        %s          %s\n",
442 			"T [desc]     [address 63:0  ] ",
443 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
444 			"leng", "ntw", "timestamp", "bi->skb");
445 
446 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
447 			tx_desc = IXGBE_TX_DESC(tx_ring, i);
448 			tx_buffer = &tx_ring->tx_buffer_info[i];
449 			u0 = (struct my_u0 *)tx_desc;
450 			if (dma_unmap_len(tx_buffer, len) > 0) {
451 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p",
452 					i,
453 					le64_to_cpu(u0->a),
454 					le64_to_cpu(u0->b),
455 					(u64)dma_unmap_addr(tx_buffer, dma),
456 					dma_unmap_len(tx_buffer, len),
457 					tx_buffer->next_to_watch,
458 					(u64)tx_buffer->time_stamp,
459 					tx_buffer->skb);
460 				if (i == tx_ring->next_to_use &&
461 					i == tx_ring->next_to_clean)
462 					pr_cont(" NTC/U\n");
463 				else if (i == tx_ring->next_to_use)
464 					pr_cont(" NTU\n");
465 				else if (i == tx_ring->next_to_clean)
466 					pr_cont(" NTC\n");
467 				else
468 					pr_cont("\n");
469 
470 				if (netif_msg_pktdata(adapter) &&
471 				    tx_buffer->skb)
472 					print_hex_dump(KERN_INFO, "",
473 						DUMP_PREFIX_ADDRESS, 16, 1,
474 						tx_buffer->skb->data,
475 						dma_unmap_len(tx_buffer, len),
476 						true);
477 			}
478 		}
479 	}
480 
481 	/* Print RX Rings Summary */
482 rx_ring_summary:
483 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
484 	pr_info("Queue [NTU] [NTC]\n");
485 	for (n = 0; n < adapter->num_rx_queues; n++) {
486 		rx_ring = adapter->rx_ring[n];
487 		pr_info("%5d %5X %5X\n",
488 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
489 	}
490 
491 	/* Print RX Rings */
492 	if (!netif_msg_rx_status(adapter))
493 		goto exit;
494 
495 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
496 
497 	/* Receive Descriptor Formats
498 	 *
499 	 * 82598 Advanced Receive Descriptor (Read) Format
500 	 *    63                                           1        0
501 	 *    +-----------------------------------------------------+
502 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
503 	 *    +----------------------------------------------+------+
504 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
505 	 *    +-----------------------------------------------------+
506 	 *
507 	 *
508 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
509 	 *
510 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
511 	 *   +------------------------------------------------------+
512 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
513 	 *   | Packet   | IP     |   |          |     | Type | Type |
514 	 *   | Checksum | Ident  |   |          |     |      |      |
515 	 *   +------------------------------------------------------+
516 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
517 	 *   +------------------------------------------------------+
518 	 *   63       48 47    32 31            20 19               0
519 	 *
520 	 * 82599+ Advanced Receive Descriptor (Read) Format
521 	 *    63                                           1        0
522 	 *    +-----------------------------------------------------+
523 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
524 	 *    +----------------------------------------------+------+
525 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
526 	 *    +-----------------------------------------------------+
527 	 *
528 	 *
529 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
530 	 *
531 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
532 	 *   +------------------------------------------------------+
533 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
534 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
535 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
536 	 *   +------------------------------------------------------+
537 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
538 	 *   +------------------------------------------------------+
539 	 *   63       48 47    32 31          20 19                 0
540 	 */
541 
542 	for (n = 0; n < adapter->num_rx_queues; n++) {
543 		rx_ring = adapter->rx_ring[n];
544 		pr_info("------------------------------------\n");
545 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
546 		pr_info("------------------------------------\n");
547 		pr_info("%s%s%s",
548 			"R  [desc]      [ PktBuf     A0] ",
549 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
550 			"<-- Adv Rx Read format\n");
551 		pr_info("%s%s%s",
552 			"RWB[desc]      [PcsmIpSHl PtRs] ",
553 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
554 			"<-- Adv Rx Write-Back format\n");
555 
556 		for (i = 0; i < rx_ring->count; i++) {
557 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
558 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
559 			u0 = (struct my_u0 *)rx_desc;
560 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
561 			if (staterr & IXGBE_RXD_STAT_DD) {
562 				/* Descriptor Done */
563 				pr_info("RWB[0x%03X]     %016llX "
564 					"%016llX ---------------- %p", i,
565 					le64_to_cpu(u0->a),
566 					le64_to_cpu(u0->b),
567 					rx_buffer_info->skb);
568 			} else {
569 				pr_info("R  [0x%03X]     %016llX "
570 					"%016llX %016llX %p", i,
571 					le64_to_cpu(u0->a),
572 					le64_to_cpu(u0->b),
573 					(u64)rx_buffer_info->dma,
574 					rx_buffer_info->skb);
575 
576 				if (netif_msg_pktdata(adapter) &&
577 				    rx_buffer_info->dma) {
578 					print_hex_dump(KERN_INFO, "",
579 					   DUMP_PREFIX_ADDRESS, 16, 1,
580 					   page_address(rx_buffer_info->page) +
581 						    rx_buffer_info->page_offset,
582 					   ixgbe_rx_bufsz(rx_ring), true);
583 				}
584 			}
585 
586 			if (i == rx_ring->next_to_use)
587 				pr_cont(" NTU\n");
588 			else if (i == rx_ring->next_to_clean)
589 				pr_cont(" NTC\n");
590 			else
591 				pr_cont("\n");
592 
593 		}
594 	}
595 
596 exit:
597 	return;
598 }
599 
600 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
601 {
602 	u32 ctrl_ext;
603 
604 	/* Let firmware take over control of h/w */
605 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
606 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
607 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
608 }
609 
610 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
611 {
612 	u32 ctrl_ext;
613 
614 	/* Let firmware know the driver has taken over */
615 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
616 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
617 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
618 }
619 
620 /**
621  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
622  * @adapter: pointer to adapter struct
623  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
624  * @queue: queue to map the corresponding interrupt to
625  * @msix_vector: the vector to map to the corresponding queue
626  *
627  */
628 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
629 			   u8 queue, u8 msix_vector)
630 {
631 	u32 ivar, index;
632 	struct ixgbe_hw *hw = &adapter->hw;
633 	switch (hw->mac.type) {
634 	case ixgbe_mac_82598EB:
635 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
636 		if (direction == -1)
637 			direction = 0;
638 		index = (((direction * 64) + queue) >> 2) & 0x1F;
639 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
640 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
641 		ivar |= (msix_vector << (8 * (queue & 0x3)));
642 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
643 		break;
644 	case ixgbe_mac_82599EB:
645 	case ixgbe_mac_X540:
646 		if (direction == -1) {
647 			/* other causes */
648 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
649 			index = ((queue & 1) * 8);
650 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
651 			ivar &= ~(0xFF << index);
652 			ivar |= (msix_vector << index);
653 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
654 			break;
655 		} else {
656 			/* tx or rx causes */
657 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
658 			index = ((16 * (queue & 1)) + (8 * direction));
659 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
660 			ivar &= ~(0xFF << index);
661 			ivar |= (msix_vector << index);
662 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
663 			break;
664 		}
665 	default:
666 		break;
667 	}
668 }
669 
670 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
671 					  u64 qmask)
672 {
673 	u32 mask;
674 
675 	switch (adapter->hw.mac.type) {
676 	case ixgbe_mac_82598EB:
677 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
678 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
679 		break;
680 	case ixgbe_mac_82599EB:
681 	case ixgbe_mac_X540:
682 		mask = (qmask & 0xFFFFFFFF);
683 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
684 		mask = (qmask >> 32);
685 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
686 		break;
687 	default:
688 		break;
689 	}
690 }
691 
692 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring *ring,
693 				      struct ixgbe_tx_buffer *tx_buffer)
694 {
695 	if (tx_buffer->skb) {
696 		dev_kfree_skb_any(tx_buffer->skb);
697 		if (dma_unmap_len(tx_buffer, len))
698 			dma_unmap_single(ring->dev,
699 					 dma_unmap_addr(tx_buffer, dma),
700 					 dma_unmap_len(tx_buffer, len),
701 					 DMA_TO_DEVICE);
702 	} else if (dma_unmap_len(tx_buffer, len)) {
703 		dma_unmap_page(ring->dev,
704 			       dma_unmap_addr(tx_buffer, dma),
705 			       dma_unmap_len(tx_buffer, len),
706 			       DMA_TO_DEVICE);
707 	}
708 	tx_buffer->next_to_watch = NULL;
709 	tx_buffer->skb = NULL;
710 	dma_unmap_len_set(tx_buffer, len, 0);
711 	/* tx_buffer must be completely set up in the transmit path */
712 }
713 
714 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
715 {
716 	struct ixgbe_hw *hw = &adapter->hw;
717 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
718 	int i;
719 	u32 data;
720 
721 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
722 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
723 		return;
724 
725 	switch (hw->mac.type) {
726 	case ixgbe_mac_82598EB:
727 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
728 		break;
729 	default:
730 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
731 	}
732 	hwstats->lxoffrxc += data;
733 
734 	/* refill credits (no tx hang) if we received xoff */
735 	if (!data)
736 		return;
737 
738 	for (i = 0; i < adapter->num_tx_queues; i++)
739 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
740 			  &adapter->tx_ring[i]->state);
741 }
742 
743 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
744 {
745 	struct ixgbe_hw *hw = &adapter->hw;
746 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
747 	u32 xoff[8] = {0};
748 	u8 tc;
749 	int i;
750 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
751 
752 	if (adapter->ixgbe_ieee_pfc)
753 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
754 
755 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
756 		ixgbe_update_xoff_rx_lfc(adapter);
757 		return;
758 	}
759 
760 	/* update stats for each tc, only valid with PFC enabled */
761 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
762 		u32 pxoffrxc;
763 
764 		switch (hw->mac.type) {
765 		case ixgbe_mac_82598EB:
766 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
767 			break;
768 		default:
769 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
770 		}
771 		hwstats->pxoffrxc[i] += pxoffrxc;
772 		/* Get the TC for given UP */
773 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
774 		xoff[tc] += pxoffrxc;
775 	}
776 
777 	/* disarm tx queues that have received xoff frames */
778 	for (i = 0; i < adapter->num_tx_queues; i++) {
779 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
780 
781 		tc = tx_ring->dcb_tc;
782 		if (xoff[tc])
783 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
784 	}
785 }
786 
787 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
788 {
789 	return ring->stats.packets;
790 }
791 
792 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
793 {
794 	struct ixgbe_adapter *adapter = netdev_priv(ring->netdev);
795 	struct ixgbe_hw *hw = &adapter->hw;
796 
797 	u32 head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
798 	u32 tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
799 
800 	if (head != tail)
801 		return (head < tail) ?
802 			tail - head : (tail + ring->count - head);
803 
804 	return 0;
805 }
806 
807 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
808 {
809 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
810 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
811 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
812 	bool ret = false;
813 
814 	clear_check_for_tx_hang(tx_ring);
815 
816 	/*
817 	 * Check for a hung queue, but be thorough. This verifies
818 	 * that a transmit has been completed since the previous
819 	 * check AND there is at least one packet pending. The
820 	 * ARMED bit is set to indicate a potential hang. The
821 	 * bit is cleared if a pause frame is received to remove
822 	 * false hang detection due to PFC or 802.3x frames. By
823 	 * requiring this to fail twice we avoid races with
824 	 * pfc clearing the ARMED bit and conditions where we
825 	 * run the check_tx_hang logic with a transmit completion
826 	 * pending but without time to complete it yet.
827 	 */
828 	if ((tx_done_old == tx_done) && tx_pending) {
829 		/* make sure it is true for two checks in a row */
830 		ret = test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
831 				       &tx_ring->state);
832 	} else {
833 		/* update completed stats and continue */
834 		tx_ring->tx_stats.tx_done_old = tx_done;
835 		/* reset the countdown */
836 		clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
837 	}
838 
839 	return ret;
840 }
841 
842 /**
843  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
844  * @adapter: driver private struct
845  **/
846 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
847 {
848 
849 	/* Do the reset outside of interrupt context */
850 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
851 		adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
852 		e_warn(drv, "initiating reset due to tx timeout\n");
853 		ixgbe_service_event_schedule(adapter);
854 	}
855 }
856 
857 /**
858  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
859  * @q_vector: structure containing interrupt and ring information
860  * @tx_ring: tx ring to clean
861  **/
862 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
863 			       struct ixgbe_ring *tx_ring)
864 {
865 	struct ixgbe_adapter *adapter = q_vector->adapter;
866 	struct ixgbe_tx_buffer *tx_buffer;
867 	union ixgbe_adv_tx_desc *tx_desc;
868 	unsigned int total_bytes = 0, total_packets = 0;
869 	unsigned int budget = q_vector->tx.work_limit;
870 	unsigned int i = tx_ring->next_to_clean;
871 
872 	if (test_bit(__IXGBE_DOWN, &adapter->state))
873 		return true;
874 
875 	tx_buffer = &tx_ring->tx_buffer_info[i];
876 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
877 	i -= tx_ring->count;
878 
879 	do {
880 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
881 
882 		/* if next_to_watch is not set then there is no work pending */
883 		if (!eop_desc)
884 			break;
885 
886 		/* prevent any other reads prior to eop_desc */
887 		read_barrier_depends();
888 
889 		/* if DD is not set pending work has not been completed */
890 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
891 			break;
892 
893 		/* clear next_to_watch to prevent false hangs */
894 		tx_buffer->next_to_watch = NULL;
895 
896 		/* update the statistics for this packet */
897 		total_bytes += tx_buffer->bytecount;
898 		total_packets += tx_buffer->gso_segs;
899 
900 		/* free the skb */
901 		dev_kfree_skb_any(tx_buffer->skb);
902 
903 		/* unmap skb header data */
904 		dma_unmap_single(tx_ring->dev,
905 				 dma_unmap_addr(tx_buffer, dma),
906 				 dma_unmap_len(tx_buffer, len),
907 				 DMA_TO_DEVICE);
908 
909 		/* clear tx_buffer data */
910 		tx_buffer->skb = NULL;
911 		dma_unmap_len_set(tx_buffer, len, 0);
912 
913 		/* unmap remaining buffers */
914 		while (tx_desc != eop_desc) {
915 			tx_buffer++;
916 			tx_desc++;
917 			i++;
918 			if (unlikely(!i)) {
919 				i -= tx_ring->count;
920 				tx_buffer = tx_ring->tx_buffer_info;
921 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
922 			}
923 
924 			/* unmap any remaining paged data */
925 			if (dma_unmap_len(tx_buffer, len)) {
926 				dma_unmap_page(tx_ring->dev,
927 					       dma_unmap_addr(tx_buffer, dma),
928 					       dma_unmap_len(tx_buffer, len),
929 					       DMA_TO_DEVICE);
930 				dma_unmap_len_set(tx_buffer, len, 0);
931 			}
932 		}
933 
934 		/* move us one more past the eop_desc for start of next pkt */
935 		tx_buffer++;
936 		tx_desc++;
937 		i++;
938 		if (unlikely(!i)) {
939 			i -= tx_ring->count;
940 			tx_buffer = tx_ring->tx_buffer_info;
941 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
942 		}
943 
944 		/* issue prefetch for next Tx descriptor */
945 		prefetch(tx_desc);
946 
947 		/* update budget accounting */
948 		budget--;
949 	} while (likely(budget));
950 
951 	i += tx_ring->count;
952 	tx_ring->next_to_clean = i;
953 	u64_stats_update_begin(&tx_ring->syncp);
954 	tx_ring->stats.bytes += total_bytes;
955 	tx_ring->stats.packets += total_packets;
956 	u64_stats_update_end(&tx_ring->syncp);
957 	q_vector->tx.total_bytes += total_bytes;
958 	q_vector->tx.total_packets += total_packets;
959 
960 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
961 		/* schedule immediate reset if we believe we hung */
962 		struct ixgbe_hw *hw = &adapter->hw;
963 		e_err(drv, "Detected Tx Unit Hang\n"
964 			"  Tx Queue             <%d>\n"
965 			"  TDH, TDT             <%x>, <%x>\n"
966 			"  next_to_use          <%x>\n"
967 			"  next_to_clean        <%x>\n"
968 			"tx_buffer_info[next_to_clean]\n"
969 			"  time_stamp           <%lx>\n"
970 			"  jiffies              <%lx>\n",
971 			tx_ring->queue_index,
972 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
973 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
974 			tx_ring->next_to_use, i,
975 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
976 
977 		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
978 
979 		e_info(probe,
980 		       "tx hang %d detected on queue %d, resetting adapter\n",
981 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
982 
983 		/* schedule immediate reset if we believe we hung */
984 		ixgbe_tx_timeout_reset(adapter);
985 
986 		/* the adapter is about to reset, no point in enabling stuff */
987 		return true;
988 	}
989 
990 	netdev_tx_completed_queue(txring_txq(tx_ring),
991 				  total_packets, total_bytes);
992 
993 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
994 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
995 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
996 		/* Make sure that anybody stopping the queue after this
997 		 * sees the new next_to_clean.
998 		 */
999 		smp_mb();
1000 		if (__netif_subqueue_stopped(tx_ring->netdev,
1001 					     tx_ring->queue_index)
1002 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1003 			netif_wake_subqueue(tx_ring->netdev,
1004 					    tx_ring->queue_index);
1005 			++tx_ring->tx_stats.restart_queue;
1006 		}
1007 	}
1008 
1009 	return !!budget;
1010 }
1011 
1012 #ifdef CONFIG_IXGBE_DCA
1013 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1014 				struct ixgbe_ring *tx_ring,
1015 				int cpu)
1016 {
1017 	struct ixgbe_hw *hw = &adapter->hw;
1018 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
1019 	u16 reg_offset;
1020 
1021 	switch (hw->mac.type) {
1022 	case ixgbe_mac_82598EB:
1023 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1024 		break;
1025 	case ixgbe_mac_82599EB:
1026 	case ixgbe_mac_X540:
1027 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1028 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1029 		break;
1030 	default:
1031 		/* for unknown hardware do not write register */
1032 		return;
1033 	}
1034 
1035 	/*
1036 	 * We can enable relaxed ordering for reads, but not writes when
1037 	 * DCA is enabled.  This is due to a known issue in some chipsets
1038 	 * which will cause the DCA tag to be cleared.
1039 	 */
1040 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1041 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1042 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1043 
1044 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1045 }
1046 
1047 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1048 				struct ixgbe_ring *rx_ring,
1049 				int cpu)
1050 {
1051 	struct ixgbe_hw *hw = &adapter->hw;
1052 	u32 rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1053 	u8 reg_idx = rx_ring->reg_idx;
1054 
1055 
1056 	switch (hw->mac.type) {
1057 	case ixgbe_mac_82599EB:
1058 	case ixgbe_mac_X540:
1059 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1060 		break;
1061 	default:
1062 		break;
1063 	}
1064 
1065 	/*
1066 	 * We can enable relaxed ordering for reads, but not writes when
1067 	 * DCA is enabled.  This is due to a known issue in some chipsets
1068 	 * which will cause the DCA tag to be cleared.
1069 	 */
1070 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1071 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1072 
1073 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1074 }
1075 
1076 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1077 {
1078 	struct ixgbe_adapter *adapter = q_vector->adapter;
1079 	struct ixgbe_ring *ring;
1080 	int cpu = get_cpu();
1081 
1082 	if (q_vector->cpu == cpu)
1083 		goto out_no_update;
1084 
1085 	ixgbe_for_each_ring(ring, q_vector->tx)
1086 		ixgbe_update_tx_dca(adapter, ring, cpu);
1087 
1088 	ixgbe_for_each_ring(ring, q_vector->rx)
1089 		ixgbe_update_rx_dca(adapter, ring, cpu);
1090 
1091 	q_vector->cpu = cpu;
1092 out_no_update:
1093 	put_cpu();
1094 }
1095 
1096 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1097 {
1098 	int i;
1099 
1100 	if (!(adapter->flags & IXGBE_FLAG_DCA_ENABLED))
1101 		return;
1102 
1103 	/* always use CB2 mode, difference is masked in the CB driver */
1104 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 2);
1105 
1106 	for (i = 0; i < adapter->num_q_vectors; i++) {
1107 		adapter->q_vector[i]->cpu = -1;
1108 		ixgbe_update_dca(adapter->q_vector[i]);
1109 	}
1110 }
1111 
1112 static int __ixgbe_notify_dca(struct device *dev, void *data)
1113 {
1114 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1115 	unsigned long event = *(unsigned long *)data;
1116 
1117 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1118 		return 0;
1119 
1120 	switch (event) {
1121 	case DCA_PROVIDER_ADD:
1122 		/* if we're already enabled, don't do it again */
1123 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1124 			break;
1125 		if (dca_add_requester(dev) == 0) {
1126 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1127 			ixgbe_setup_dca(adapter);
1128 			break;
1129 		}
1130 		/* Fall Through since DCA is disabled. */
1131 	case DCA_PROVIDER_REMOVE:
1132 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1133 			dca_remove_requester(dev);
1134 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1135 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
1136 		}
1137 		break;
1138 	}
1139 
1140 	return 0;
1141 }
1142 
1143 #endif /* CONFIG_IXGBE_DCA */
1144 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1145 				 union ixgbe_adv_rx_desc *rx_desc,
1146 				 struct sk_buff *skb)
1147 {
1148 	if (ring->netdev->features & NETIF_F_RXHASH)
1149 		skb->rxhash = le32_to_cpu(rx_desc->wb.lower.hi_dword.rss);
1150 }
1151 
1152 #ifdef IXGBE_FCOE
1153 /**
1154  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1155  * @ring: structure containing ring specific data
1156  * @rx_desc: advanced rx descriptor
1157  *
1158  * Returns : true if it is FCoE pkt
1159  */
1160 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1161 				    union ixgbe_adv_rx_desc *rx_desc)
1162 {
1163 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1164 
1165 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1166 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1167 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1168 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1169 }
1170 
1171 #endif /* IXGBE_FCOE */
1172 /**
1173  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1174  * @ring: structure containing ring specific data
1175  * @rx_desc: current Rx descriptor being processed
1176  * @skb: skb currently being received and modified
1177  **/
1178 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1179 				     union ixgbe_adv_rx_desc *rx_desc,
1180 				     struct sk_buff *skb)
1181 {
1182 	skb_checksum_none_assert(skb);
1183 
1184 	/* Rx csum disabled */
1185 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1186 		return;
1187 
1188 	/* if IP and error */
1189 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1190 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1191 		ring->rx_stats.csum_err++;
1192 		return;
1193 	}
1194 
1195 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1196 		return;
1197 
1198 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1199 		__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1200 
1201 		/*
1202 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1203 		 * checksum errors.
1204 		 */
1205 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1206 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1207 			return;
1208 
1209 		ring->rx_stats.csum_err++;
1210 		return;
1211 	}
1212 
1213 	/* It must be a TCP or UDP packet with a valid checksum */
1214 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1215 }
1216 
1217 static inline void ixgbe_release_rx_desc(struct ixgbe_ring *rx_ring, u32 val)
1218 {
1219 	rx_ring->next_to_use = val;
1220 
1221 	/* update next to alloc since we have filled the ring */
1222 	rx_ring->next_to_alloc = val;
1223 	/*
1224 	 * Force memory writes to complete before letting h/w
1225 	 * know there are new descriptors to fetch.  (Only
1226 	 * applicable for weak-ordered memory model archs,
1227 	 * such as IA-64).
1228 	 */
1229 	wmb();
1230 	writel(val, rx_ring->tail);
1231 }
1232 
1233 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1234 				    struct ixgbe_rx_buffer *bi)
1235 {
1236 	struct page *page = bi->page;
1237 	dma_addr_t dma = bi->dma;
1238 
1239 	/* since we are recycling buffers we should seldom need to alloc */
1240 	if (likely(dma))
1241 		return true;
1242 
1243 	/* alloc new page for storage */
1244 	if (likely(!page)) {
1245 		page = __skb_alloc_pages(GFP_ATOMIC | __GFP_COLD | __GFP_COMP,
1246 					 bi->skb, ixgbe_rx_pg_order(rx_ring));
1247 		if (unlikely(!page)) {
1248 			rx_ring->rx_stats.alloc_rx_page_failed++;
1249 			return false;
1250 		}
1251 		bi->page = page;
1252 	}
1253 
1254 	/* map page for use */
1255 	dma = dma_map_page(rx_ring->dev, page, 0,
1256 			   ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1257 
1258 	/*
1259 	 * if mapping failed free memory back to system since
1260 	 * there isn't much point in holding memory we can't use
1261 	 */
1262 	if (dma_mapping_error(rx_ring->dev, dma)) {
1263 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1264 		bi->page = NULL;
1265 
1266 		rx_ring->rx_stats.alloc_rx_page_failed++;
1267 		return false;
1268 	}
1269 
1270 	bi->dma = dma;
1271 	bi->page_offset = 0;
1272 
1273 	return true;
1274 }
1275 
1276 /**
1277  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1278  * @rx_ring: ring to place buffers on
1279  * @cleaned_count: number of buffers to replace
1280  **/
1281 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1282 {
1283 	union ixgbe_adv_rx_desc *rx_desc;
1284 	struct ixgbe_rx_buffer *bi;
1285 	u16 i = rx_ring->next_to_use;
1286 
1287 	/* nothing to do */
1288 	if (!cleaned_count)
1289 		return;
1290 
1291 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1292 	bi = &rx_ring->rx_buffer_info[i];
1293 	i -= rx_ring->count;
1294 
1295 	do {
1296 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1297 			break;
1298 
1299 		/*
1300 		 * Refresh the desc even if buffer_addrs didn't change
1301 		 * because each write-back erases this info.
1302 		 */
1303 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1304 
1305 		rx_desc++;
1306 		bi++;
1307 		i++;
1308 		if (unlikely(!i)) {
1309 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1310 			bi = rx_ring->rx_buffer_info;
1311 			i -= rx_ring->count;
1312 		}
1313 
1314 		/* clear the hdr_addr for the next_to_use descriptor */
1315 		rx_desc->read.hdr_addr = 0;
1316 
1317 		cleaned_count--;
1318 	} while (cleaned_count);
1319 
1320 	i += rx_ring->count;
1321 
1322 	if (rx_ring->next_to_use != i)
1323 		ixgbe_release_rx_desc(rx_ring, i);
1324 }
1325 
1326 /**
1327  * ixgbe_get_headlen - determine size of header for RSC/LRO/GRO/FCOE
1328  * @data: pointer to the start of the headers
1329  * @max_len: total length of section to find headers in
1330  *
1331  * This function is meant to determine the length of headers that will
1332  * be recognized by hardware for LRO, GRO, and RSC offloads.  The main
1333  * motivation of doing this is to only perform one pull for IPv4 TCP
1334  * packets so that we can do basic things like calculating the gso_size
1335  * based on the average data per packet.
1336  **/
1337 static unsigned int ixgbe_get_headlen(unsigned char *data,
1338 				      unsigned int max_len)
1339 {
1340 	union {
1341 		unsigned char *network;
1342 		/* l2 headers */
1343 		struct ethhdr *eth;
1344 		struct vlan_hdr *vlan;
1345 		/* l3 headers */
1346 		struct iphdr *ipv4;
1347 		struct ipv6hdr *ipv6;
1348 	} hdr;
1349 	__be16 protocol;
1350 	u8 nexthdr = 0;	/* default to not TCP */
1351 	u8 hlen;
1352 
1353 	/* this should never happen, but better safe than sorry */
1354 	if (max_len < ETH_HLEN)
1355 		return max_len;
1356 
1357 	/* initialize network frame pointer */
1358 	hdr.network = data;
1359 
1360 	/* set first protocol and move network header forward */
1361 	protocol = hdr.eth->h_proto;
1362 	hdr.network += ETH_HLEN;
1363 
1364 	/* handle any vlan tag if present */
1365 	if (protocol == __constant_htons(ETH_P_8021Q)) {
1366 		if ((hdr.network - data) > (max_len - VLAN_HLEN))
1367 			return max_len;
1368 
1369 		protocol = hdr.vlan->h_vlan_encapsulated_proto;
1370 		hdr.network += VLAN_HLEN;
1371 	}
1372 
1373 	/* handle L3 protocols */
1374 	if (protocol == __constant_htons(ETH_P_IP)) {
1375 		if ((hdr.network - data) > (max_len - sizeof(struct iphdr)))
1376 			return max_len;
1377 
1378 		/* access ihl as a u8 to avoid unaligned access on ia64 */
1379 		hlen = (hdr.network[0] & 0x0F) << 2;
1380 
1381 		/* verify hlen meets minimum size requirements */
1382 		if (hlen < sizeof(struct iphdr))
1383 			return hdr.network - data;
1384 
1385 		/* record next protocol if header is present */
1386 		if (!(hdr.ipv4->frag_off & htons(IP_OFFSET)))
1387 			nexthdr = hdr.ipv4->protocol;
1388 	} else if (protocol == __constant_htons(ETH_P_IPV6)) {
1389 		if ((hdr.network - data) > (max_len - sizeof(struct ipv6hdr)))
1390 			return max_len;
1391 
1392 		/* record next protocol */
1393 		nexthdr = hdr.ipv6->nexthdr;
1394 		hlen = sizeof(struct ipv6hdr);
1395 #ifdef IXGBE_FCOE
1396 	} else if (protocol == __constant_htons(ETH_P_FCOE)) {
1397 		if ((hdr.network - data) > (max_len - FCOE_HEADER_LEN))
1398 			return max_len;
1399 		hlen = FCOE_HEADER_LEN;
1400 #endif
1401 	} else {
1402 		return hdr.network - data;
1403 	}
1404 
1405 	/* relocate pointer to start of L4 header */
1406 	hdr.network += hlen;
1407 
1408 	/* finally sort out TCP/UDP */
1409 	if (nexthdr == IPPROTO_TCP) {
1410 		if ((hdr.network - data) > (max_len - sizeof(struct tcphdr)))
1411 			return max_len;
1412 
1413 		/* access doff as a u8 to avoid unaligned access on ia64 */
1414 		hlen = (hdr.network[12] & 0xF0) >> 2;
1415 
1416 		/* verify hlen meets minimum size requirements */
1417 		if (hlen < sizeof(struct tcphdr))
1418 			return hdr.network - data;
1419 
1420 		hdr.network += hlen;
1421 	} else if (nexthdr == IPPROTO_UDP) {
1422 		if ((hdr.network - data) > (max_len - sizeof(struct udphdr)))
1423 			return max_len;
1424 
1425 		hdr.network += sizeof(struct udphdr);
1426 	}
1427 
1428 	/*
1429 	 * If everything has gone correctly hdr.network should be the
1430 	 * data section of the packet and will be the end of the header.
1431 	 * If not then it probably represents the end of the last recognized
1432 	 * header.
1433 	 */
1434 	if ((hdr.network - data) < max_len)
1435 		return hdr.network - data;
1436 	else
1437 		return max_len;
1438 }
1439 
1440 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1441 				   struct sk_buff *skb)
1442 {
1443 	u16 hdr_len = skb_headlen(skb);
1444 
1445 	/* set gso_size to avoid messing up TCP MSS */
1446 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1447 						 IXGBE_CB(skb)->append_cnt);
1448 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1449 }
1450 
1451 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1452 				   struct sk_buff *skb)
1453 {
1454 	/* if append_cnt is 0 then frame is not RSC */
1455 	if (!IXGBE_CB(skb)->append_cnt)
1456 		return;
1457 
1458 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1459 	rx_ring->rx_stats.rsc_flush++;
1460 
1461 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1462 
1463 	/* gso_size is computed using append_cnt so always clear it last */
1464 	IXGBE_CB(skb)->append_cnt = 0;
1465 }
1466 
1467 /**
1468  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1469  * @rx_ring: rx descriptor ring packet is being transacted on
1470  * @rx_desc: pointer to the EOP Rx descriptor
1471  * @skb: pointer to current skb being populated
1472  *
1473  * This function checks the ring, descriptor, and packet information in
1474  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1475  * other fields within the skb.
1476  **/
1477 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1478 				     union ixgbe_adv_rx_desc *rx_desc,
1479 				     struct sk_buff *skb)
1480 {
1481 	struct net_device *dev = rx_ring->netdev;
1482 
1483 	ixgbe_update_rsc_stats(rx_ring, skb);
1484 
1485 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1486 
1487 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1488 
1489 	ixgbe_ptp_rx_hwtstamp(rx_ring, rx_desc, skb);
1490 
1491 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1492 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1493 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1494 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1495 	}
1496 
1497 	skb_record_rx_queue(skb, rx_ring->queue_index);
1498 
1499 	skb->protocol = eth_type_trans(skb, dev);
1500 }
1501 
1502 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1503 			 struct sk_buff *skb)
1504 {
1505 	struct ixgbe_adapter *adapter = q_vector->adapter;
1506 
1507 	if (ixgbe_qv_ll_polling(q_vector))
1508 		netif_receive_skb(skb);
1509 	else if (!(adapter->flags & IXGBE_FLAG_IN_NETPOLL))
1510 		napi_gro_receive(&q_vector->napi, skb);
1511 	else
1512 		netif_rx(skb);
1513 }
1514 
1515 /**
1516  * ixgbe_is_non_eop - process handling of non-EOP buffers
1517  * @rx_ring: Rx ring being processed
1518  * @rx_desc: Rx descriptor for current buffer
1519  * @skb: Current socket buffer containing buffer in progress
1520  *
1521  * This function updates next to clean.  If the buffer is an EOP buffer
1522  * this function exits returning false, otherwise it will place the
1523  * sk_buff in the next buffer to be chained and return true indicating
1524  * that this is in fact a non-EOP buffer.
1525  **/
1526 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1527 			     union ixgbe_adv_rx_desc *rx_desc,
1528 			     struct sk_buff *skb)
1529 {
1530 	u32 ntc = rx_ring->next_to_clean + 1;
1531 
1532 	/* fetch, update, and store next to clean */
1533 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1534 	rx_ring->next_to_clean = ntc;
1535 
1536 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1537 
1538 	/* update RSC append count if present */
1539 	if (ring_is_rsc_enabled(rx_ring)) {
1540 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1541 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1542 
1543 		if (unlikely(rsc_enabled)) {
1544 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1545 
1546 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1547 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1548 
1549 			/* update ntc based on RSC value */
1550 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1551 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1552 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1553 		}
1554 	}
1555 
1556 	/* if we are the last buffer then there is nothing else to do */
1557 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1558 		return false;
1559 
1560 	/* place skb in next buffer to be received */
1561 	rx_ring->rx_buffer_info[ntc].skb = skb;
1562 	rx_ring->rx_stats.non_eop_descs++;
1563 
1564 	return true;
1565 }
1566 
1567 /**
1568  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1569  * @rx_ring: rx descriptor ring packet is being transacted on
1570  * @skb: pointer to current skb being adjusted
1571  *
1572  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1573  * main difference between this version and the original function is that
1574  * this function can make several assumptions about the state of things
1575  * that allow for significant optimizations versus the standard function.
1576  * As a result we can do things like drop a frag and maintain an accurate
1577  * truesize for the skb.
1578  */
1579 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1580 			    struct sk_buff *skb)
1581 {
1582 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1583 	unsigned char *va;
1584 	unsigned int pull_len;
1585 
1586 	/*
1587 	 * it is valid to use page_address instead of kmap since we are
1588 	 * working with pages allocated out of the lomem pool per
1589 	 * alloc_page(GFP_ATOMIC)
1590 	 */
1591 	va = skb_frag_address(frag);
1592 
1593 	/*
1594 	 * we need the header to contain the greater of either ETH_HLEN or
1595 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1596 	 */
1597 	pull_len = ixgbe_get_headlen(va, IXGBE_RX_HDR_SIZE);
1598 
1599 	/* align pull length to size of long to optimize memcpy performance */
1600 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1601 
1602 	/* update all of the pointers */
1603 	skb_frag_size_sub(frag, pull_len);
1604 	frag->page_offset += pull_len;
1605 	skb->data_len -= pull_len;
1606 	skb->tail += pull_len;
1607 }
1608 
1609 /**
1610  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1611  * @rx_ring: rx descriptor ring packet is being transacted on
1612  * @skb: pointer to current skb being updated
1613  *
1614  * This function provides a basic DMA sync up for the first fragment of an
1615  * skb.  The reason for doing this is that the first fragment cannot be
1616  * unmapped until we have reached the end of packet descriptor for a buffer
1617  * chain.
1618  */
1619 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1620 				struct sk_buff *skb)
1621 {
1622 	/* if the page was released unmap it, else just sync our portion */
1623 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1624 		dma_unmap_page(rx_ring->dev, IXGBE_CB(skb)->dma,
1625 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1626 		IXGBE_CB(skb)->page_released = false;
1627 	} else {
1628 		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1629 
1630 		dma_sync_single_range_for_cpu(rx_ring->dev,
1631 					      IXGBE_CB(skb)->dma,
1632 					      frag->page_offset,
1633 					      ixgbe_rx_bufsz(rx_ring),
1634 					      DMA_FROM_DEVICE);
1635 	}
1636 	IXGBE_CB(skb)->dma = 0;
1637 }
1638 
1639 /**
1640  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1641  * @rx_ring: rx descriptor ring packet is being transacted on
1642  * @rx_desc: pointer to the EOP Rx descriptor
1643  * @skb: pointer to current skb being fixed
1644  *
1645  * Check for corrupted packet headers caused by senders on the local L2
1646  * embedded NIC switch not setting up their Tx Descriptors right.  These
1647  * should be very rare.
1648  *
1649  * Also address the case where we are pulling data in on pages only
1650  * and as such no data is present in the skb header.
1651  *
1652  * In addition if skb is not at least 60 bytes we need to pad it so that
1653  * it is large enough to qualify as a valid Ethernet frame.
1654  *
1655  * Returns true if an error was encountered and skb was freed.
1656  **/
1657 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1658 				  union ixgbe_adv_rx_desc *rx_desc,
1659 				  struct sk_buff *skb)
1660 {
1661 	struct net_device *netdev = rx_ring->netdev;
1662 
1663 	/* verify that the packet does not have any known errors */
1664 	if (unlikely(ixgbe_test_staterr(rx_desc,
1665 					IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1666 	    !(netdev->features & NETIF_F_RXALL))) {
1667 		dev_kfree_skb_any(skb);
1668 		return true;
1669 	}
1670 
1671 	/* place header in linear portion of buffer */
1672 	if (skb_is_nonlinear(skb))
1673 		ixgbe_pull_tail(rx_ring, skb);
1674 
1675 #ifdef IXGBE_FCOE
1676 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1677 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1678 		return false;
1679 
1680 #endif
1681 	/* if skb_pad returns an error the skb was freed */
1682 	if (unlikely(skb->len < 60)) {
1683 		int pad_len = 60 - skb->len;
1684 
1685 		if (skb_pad(skb, pad_len))
1686 			return true;
1687 		__skb_put(skb, pad_len);
1688 	}
1689 
1690 	return false;
1691 }
1692 
1693 /**
1694  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1695  * @rx_ring: rx descriptor ring to store buffers on
1696  * @old_buff: donor buffer to have page reused
1697  *
1698  * Synchronizes page for reuse by the adapter
1699  **/
1700 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1701 				struct ixgbe_rx_buffer *old_buff)
1702 {
1703 	struct ixgbe_rx_buffer *new_buff;
1704 	u16 nta = rx_ring->next_to_alloc;
1705 
1706 	new_buff = &rx_ring->rx_buffer_info[nta];
1707 
1708 	/* update, and store next to alloc */
1709 	nta++;
1710 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1711 
1712 	/* transfer page from old buffer to new buffer */
1713 	new_buff->page = old_buff->page;
1714 	new_buff->dma = old_buff->dma;
1715 	new_buff->page_offset = old_buff->page_offset;
1716 
1717 	/* sync the buffer for use by the device */
1718 	dma_sync_single_range_for_device(rx_ring->dev, new_buff->dma,
1719 					 new_buff->page_offset,
1720 					 ixgbe_rx_bufsz(rx_ring),
1721 					 DMA_FROM_DEVICE);
1722 }
1723 
1724 /**
1725  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1726  * @rx_ring: rx descriptor ring to transact packets on
1727  * @rx_buffer: buffer containing page to add
1728  * @rx_desc: descriptor containing length of buffer written by hardware
1729  * @skb: sk_buff to place the data into
1730  *
1731  * This function will add the data contained in rx_buffer->page to the skb.
1732  * This is done either through a direct copy if the data in the buffer is
1733  * less than the skb header size, otherwise it will just attach the page as
1734  * a frag to the skb.
1735  *
1736  * The function will then update the page offset if necessary and return
1737  * true if the buffer can be reused by the adapter.
1738  **/
1739 static bool ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1740 			      struct ixgbe_rx_buffer *rx_buffer,
1741 			      union ixgbe_adv_rx_desc *rx_desc,
1742 			      struct sk_buff *skb)
1743 {
1744 	struct page *page = rx_buffer->page;
1745 	unsigned int size = le16_to_cpu(rx_desc->wb.upper.length);
1746 #if (PAGE_SIZE < 8192)
1747 	unsigned int truesize = ixgbe_rx_bufsz(rx_ring);
1748 #else
1749 	unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
1750 	unsigned int last_offset = ixgbe_rx_pg_size(rx_ring) -
1751 				   ixgbe_rx_bufsz(rx_ring);
1752 #endif
1753 
1754 	if ((size <= IXGBE_RX_HDR_SIZE) && !skb_is_nonlinear(skb)) {
1755 		unsigned char *va = page_address(page) + rx_buffer->page_offset;
1756 
1757 		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
1758 
1759 		/* we can reuse buffer as-is, just make sure it is local */
1760 		if (likely(page_to_nid(page) == numa_node_id()))
1761 			return true;
1762 
1763 		/* this page cannot be reused so discard it */
1764 		put_page(page);
1765 		return false;
1766 	}
1767 
1768 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
1769 			rx_buffer->page_offset, size, truesize);
1770 
1771 	/* avoid re-using remote pages */
1772 	if (unlikely(page_to_nid(page) != numa_node_id()))
1773 		return false;
1774 
1775 #if (PAGE_SIZE < 8192)
1776 	/* if we are only owner of page we can reuse it */
1777 	if (unlikely(page_count(page) != 1))
1778 		return false;
1779 
1780 	/* flip page offset to other buffer */
1781 	rx_buffer->page_offset ^= truesize;
1782 
1783 	/*
1784 	 * since we are the only owner of the page and we need to
1785 	 * increment it, just set the value to 2 in order to avoid
1786 	 * an unecessary locked operation
1787 	 */
1788 	atomic_set(&page->_count, 2);
1789 #else
1790 	/* move offset up to the next cache line */
1791 	rx_buffer->page_offset += truesize;
1792 
1793 	if (rx_buffer->page_offset > last_offset)
1794 		return false;
1795 
1796 	/* bump ref count on page before it is given to the stack */
1797 	get_page(page);
1798 #endif
1799 
1800 	return true;
1801 }
1802 
1803 static struct sk_buff *ixgbe_fetch_rx_buffer(struct ixgbe_ring *rx_ring,
1804 					     union ixgbe_adv_rx_desc *rx_desc)
1805 {
1806 	struct ixgbe_rx_buffer *rx_buffer;
1807 	struct sk_buff *skb;
1808 	struct page *page;
1809 
1810 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
1811 	page = rx_buffer->page;
1812 	prefetchw(page);
1813 
1814 	skb = rx_buffer->skb;
1815 
1816 	if (likely(!skb)) {
1817 		void *page_addr = page_address(page) +
1818 				  rx_buffer->page_offset;
1819 
1820 		/* prefetch first cache line of first page */
1821 		prefetch(page_addr);
1822 #if L1_CACHE_BYTES < 128
1823 		prefetch(page_addr + L1_CACHE_BYTES);
1824 #endif
1825 
1826 		/* allocate a skb to store the frags */
1827 		skb = netdev_alloc_skb_ip_align(rx_ring->netdev,
1828 						IXGBE_RX_HDR_SIZE);
1829 		if (unlikely(!skb)) {
1830 			rx_ring->rx_stats.alloc_rx_buff_failed++;
1831 			return NULL;
1832 		}
1833 
1834 		/*
1835 		 * we will be copying header into skb->data in
1836 		 * pskb_may_pull so it is in our interest to prefetch
1837 		 * it now to avoid a possible cache miss
1838 		 */
1839 		prefetchw(skb->data);
1840 
1841 		/*
1842 		 * Delay unmapping of the first packet. It carries the
1843 		 * header information, HW may still access the header
1844 		 * after the writeback.  Only unmap it when EOP is
1845 		 * reached
1846 		 */
1847 		if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1848 			goto dma_sync;
1849 
1850 		IXGBE_CB(skb)->dma = rx_buffer->dma;
1851 	} else {
1852 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
1853 			ixgbe_dma_sync_frag(rx_ring, skb);
1854 
1855 dma_sync:
1856 		/* we are reusing so sync this buffer for CPU use */
1857 		dma_sync_single_range_for_cpu(rx_ring->dev,
1858 					      rx_buffer->dma,
1859 					      rx_buffer->page_offset,
1860 					      ixgbe_rx_bufsz(rx_ring),
1861 					      DMA_FROM_DEVICE);
1862 	}
1863 
1864 	/* pull page into skb */
1865 	if (ixgbe_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) {
1866 		/* hand second half of page back to the ring */
1867 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
1868 	} else if (IXGBE_CB(skb)->dma == rx_buffer->dma) {
1869 		/* the page has been released from the ring */
1870 		IXGBE_CB(skb)->page_released = true;
1871 	} else {
1872 		/* we are not reusing the buffer so unmap it */
1873 		dma_unmap_page(rx_ring->dev, rx_buffer->dma,
1874 			       ixgbe_rx_pg_size(rx_ring),
1875 			       DMA_FROM_DEVICE);
1876 	}
1877 
1878 	/* clear contents of buffer_info */
1879 	rx_buffer->skb = NULL;
1880 	rx_buffer->dma = 0;
1881 	rx_buffer->page = NULL;
1882 
1883 	return skb;
1884 }
1885 
1886 /**
1887  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
1888  * @q_vector: structure containing interrupt and ring information
1889  * @rx_ring: rx descriptor ring to transact packets on
1890  * @budget: Total limit on number of packets to process
1891  *
1892  * This function provides a "bounce buffer" approach to Rx interrupt
1893  * processing.  The advantage to this is that on systems that have
1894  * expensive overhead for IOMMU access this provides a means of avoiding
1895  * it by maintaining the mapping of the page to the syste.
1896  *
1897  * Returns amount of work completed
1898  **/
1899 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
1900 			       struct ixgbe_ring *rx_ring,
1901 			       const int budget)
1902 {
1903 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1904 #ifdef IXGBE_FCOE
1905 	struct ixgbe_adapter *adapter = q_vector->adapter;
1906 	int ddp_bytes;
1907 	unsigned int mss = 0;
1908 #endif /* IXGBE_FCOE */
1909 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
1910 
1911 	do {
1912 		union ixgbe_adv_rx_desc *rx_desc;
1913 		struct sk_buff *skb;
1914 
1915 		/* return some buffers to hardware, one at a time is too slow */
1916 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
1917 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
1918 			cleaned_count = 0;
1919 		}
1920 
1921 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
1922 
1923 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_DD))
1924 			break;
1925 
1926 		/*
1927 		 * This memory barrier is needed to keep us from reading
1928 		 * any other fields out of the rx_desc until we know the
1929 		 * RXD_STAT_DD bit is set
1930 		 */
1931 		rmb();
1932 
1933 		/* retrieve a buffer from the ring */
1934 		skb = ixgbe_fetch_rx_buffer(rx_ring, rx_desc);
1935 
1936 		/* exit if we failed to retrieve a buffer */
1937 		if (!skb)
1938 			break;
1939 
1940 		cleaned_count++;
1941 
1942 		/* place incomplete frames back on ring for completion */
1943 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
1944 			continue;
1945 
1946 		/* verify the packet layout is correct */
1947 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
1948 			continue;
1949 
1950 		/* probably a little skewed due to removing CRC */
1951 		total_rx_bytes += skb->len;
1952 
1953 		/* populate checksum, timestamp, VLAN, and protocol */
1954 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
1955 
1956 #ifdef IXGBE_FCOE
1957 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
1958 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
1959 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
1960 			/* include DDPed FCoE data */
1961 			if (ddp_bytes > 0) {
1962 				if (!mss) {
1963 					mss = rx_ring->netdev->mtu -
1964 						sizeof(struct fcoe_hdr) -
1965 						sizeof(struct fc_frame_header) -
1966 						sizeof(struct fcoe_crc_eof);
1967 					if (mss > 512)
1968 						mss &= ~511;
1969 				}
1970 				total_rx_bytes += ddp_bytes;
1971 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
1972 								 mss);
1973 			}
1974 			if (!ddp_bytes) {
1975 				dev_kfree_skb_any(skb);
1976 				continue;
1977 			}
1978 		}
1979 
1980 #endif /* IXGBE_FCOE */
1981 		skb_mark_napi_id(skb, &q_vector->napi);
1982 		ixgbe_rx_skb(q_vector, skb);
1983 
1984 		/* update budget accounting */
1985 		total_rx_packets++;
1986 	} while (likely(total_rx_packets < budget));
1987 
1988 	u64_stats_update_begin(&rx_ring->syncp);
1989 	rx_ring->stats.packets += total_rx_packets;
1990 	rx_ring->stats.bytes += total_rx_bytes;
1991 	u64_stats_update_end(&rx_ring->syncp);
1992 	q_vector->rx.total_packets += total_rx_packets;
1993 	q_vector->rx.total_bytes += total_rx_bytes;
1994 
1995 	if (cleaned_count)
1996 		ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
1997 
1998 	return total_rx_packets;
1999 }
2000 
2001 #ifdef CONFIG_NET_RX_BUSY_POLL
2002 /* must be called with local_bh_disable()d */
2003 static int ixgbe_low_latency_recv(struct napi_struct *napi)
2004 {
2005 	struct ixgbe_q_vector *q_vector =
2006 			container_of(napi, struct ixgbe_q_vector, napi);
2007 	struct ixgbe_adapter *adapter = q_vector->adapter;
2008 	struct ixgbe_ring  *ring;
2009 	int found = 0;
2010 
2011 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2012 		return LL_FLUSH_FAILED;
2013 
2014 	if (!ixgbe_qv_lock_poll(q_vector))
2015 		return LL_FLUSH_BUSY;
2016 
2017 	ixgbe_for_each_ring(ring, q_vector->rx) {
2018 		found = ixgbe_clean_rx_irq(q_vector, ring, 4);
2019 #ifdef LL_EXTENDED_STATS
2020 		if (found)
2021 			ring->stats.cleaned += found;
2022 		else
2023 			ring->stats.misses++;
2024 #endif
2025 		if (found)
2026 			break;
2027 	}
2028 
2029 	ixgbe_qv_unlock_poll(q_vector);
2030 
2031 	return found;
2032 }
2033 #endif	/* CONFIG_NET_RX_BUSY_POLL */
2034 
2035 /**
2036  * ixgbe_configure_msix - Configure MSI-X hardware
2037  * @adapter: board private structure
2038  *
2039  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2040  * interrupts.
2041  **/
2042 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2043 {
2044 	struct ixgbe_q_vector *q_vector;
2045 	int v_idx;
2046 	u32 mask;
2047 
2048 	/* Populate MSIX to EITR Select */
2049 	if (adapter->num_vfs > 32) {
2050 		u32 eitrsel = (1 << (adapter->num_vfs - 32)) - 1;
2051 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2052 	}
2053 
2054 	/*
2055 	 * Populate the IVAR table and set the ITR values to the
2056 	 * corresponding register.
2057 	 */
2058 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2059 		struct ixgbe_ring *ring;
2060 		q_vector = adapter->q_vector[v_idx];
2061 
2062 		ixgbe_for_each_ring(ring, q_vector->rx)
2063 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2064 
2065 		ixgbe_for_each_ring(ring, q_vector->tx)
2066 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2067 
2068 		ixgbe_write_eitr(q_vector);
2069 	}
2070 
2071 	switch (adapter->hw.mac.type) {
2072 	case ixgbe_mac_82598EB:
2073 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2074 			       v_idx);
2075 		break;
2076 	case ixgbe_mac_82599EB:
2077 	case ixgbe_mac_X540:
2078 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2079 		break;
2080 	default:
2081 		break;
2082 	}
2083 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2084 
2085 	/* set up to autoclear timer, and the vectors */
2086 	mask = IXGBE_EIMS_ENABLE_MASK;
2087 	mask &= ~(IXGBE_EIMS_OTHER |
2088 		  IXGBE_EIMS_MAILBOX |
2089 		  IXGBE_EIMS_LSC);
2090 
2091 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2092 }
2093 
2094 enum latency_range {
2095 	lowest_latency = 0,
2096 	low_latency = 1,
2097 	bulk_latency = 2,
2098 	latency_invalid = 255
2099 };
2100 
2101 /**
2102  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2103  * @q_vector: structure containing interrupt and ring information
2104  * @ring_container: structure containing ring performance data
2105  *
2106  *      Stores a new ITR value based on packets and byte
2107  *      counts during the last interrupt.  The advantage of per interrupt
2108  *      computation is faster updates and more accurate ITR for the current
2109  *      traffic pattern.  Constants in this function were computed
2110  *      based on theoretical maximum wire speed and thresholds were set based
2111  *      on testing data as well as attempting to minimize response time
2112  *      while increasing bulk throughput.
2113  *      this functionality is controlled by the InterruptThrottleRate module
2114  *      parameter (see ixgbe_param.c)
2115  **/
2116 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2117 			     struct ixgbe_ring_container *ring_container)
2118 {
2119 	int bytes = ring_container->total_bytes;
2120 	int packets = ring_container->total_packets;
2121 	u32 timepassed_us;
2122 	u64 bytes_perint;
2123 	u8 itr_setting = ring_container->itr;
2124 
2125 	if (packets == 0)
2126 		return;
2127 
2128 	/* simple throttlerate management
2129 	 *   0-10MB/s   lowest (100000 ints/s)
2130 	 *  10-20MB/s   low    (20000 ints/s)
2131 	 *  20-1249MB/s bulk   (8000 ints/s)
2132 	 */
2133 	/* what was last interrupt timeslice? */
2134 	timepassed_us = q_vector->itr >> 2;
2135 	if (timepassed_us == 0)
2136 		return;
2137 
2138 	bytes_perint = bytes / timepassed_us; /* bytes/usec */
2139 
2140 	switch (itr_setting) {
2141 	case lowest_latency:
2142 		if (bytes_perint > 10)
2143 			itr_setting = low_latency;
2144 		break;
2145 	case low_latency:
2146 		if (bytes_perint > 20)
2147 			itr_setting = bulk_latency;
2148 		else if (bytes_perint <= 10)
2149 			itr_setting = lowest_latency;
2150 		break;
2151 	case bulk_latency:
2152 		if (bytes_perint <= 20)
2153 			itr_setting = low_latency;
2154 		break;
2155 	}
2156 
2157 	/* clear work counters since we have the values we need */
2158 	ring_container->total_bytes = 0;
2159 	ring_container->total_packets = 0;
2160 
2161 	/* write updated itr to ring container */
2162 	ring_container->itr = itr_setting;
2163 }
2164 
2165 /**
2166  * ixgbe_write_eitr - write EITR register in hardware specific way
2167  * @q_vector: structure containing interrupt and ring information
2168  *
2169  * This function is made to be called by ethtool and by the driver
2170  * when it needs to update EITR registers at runtime.  Hardware
2171  * specific quirks/differences are taken care of here.
2172  */
2173 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2174 {
2175 	struct ixgbe_adapter *adapter = q_vector->adapter;
2176 	struct ixgbe_hw *hw = &adapter->hw;
2177 	int v_idx = q_vector->v_idx;
2178 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2179 
2180 	switch (adapter->hw.mac.type) {
2181 	case ixgbe_mac_82598EB:
2182 		/* must write high and low 16 bits to reset counter */
2183 		itr_reg |= (itr_reg << 16);
2184 		break;
2185 	case ixgbe_mac_82599EB:
2186 	case ixgbe_mac_X540:
2187 		/*
2188 		 * set the WDIS bit to not clear the timer bits and cause an
2189 		 * immediate assertion of the interrupt
2190 		 */
2191 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2192 		break;
2193 	default:
2194 		break;
2195 	}
2196 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2197 }
2198 
2199 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2200 {
2201 	u32 new_itr = q_vector->itr;
2202 	u8 current_itr;
2203 
2204 	ixgbe_update_itr(q_vector, &q_vector->tx);
2205 	ixgbe_update_itr(q_vector, &q_vector->rx);
2206 
2207 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
2208 
2209 	switch (current_itr) {
2210 	/* counts and packets in update_itr are dependent on these numbers */
2211 	case lowest_latency:
2212 		new_itr = IXGBE_100K_ITR;
2213 		break;
2214 	case low_latency:
2215 		new_itr = IXGBE_20K_ITR;
2216 		break;
2217 	case bulk_latency:
2218 		new_itr = IXGBE_8K_ITR;
2219 		break;
2220 	default:
2221 		break;
2222 	}
2223 
2224 	if (new_itr != q_vector->itr) {
2225 		/* do an exponential smoothing */
2226 		new_itr = (10 * new_itr * q_vector->itr) /
2227 			  ((9 * new_itr) + q_vector->itr);
2228 
2229 		/* save the algorithm value here */
2230 		q_vector->itr = new_itr;
2231 
2232 		ixgbe_write_eitr(q_vector);
2233 	}
2234 }
2235 
2236 /**
2237  * ixgbe_check_overtemp_subtask - check for over temperature
2238  * @adapter: pointer to adapter
2239  **/
2240 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2241 {
2242 	struct ixgbe_hw *hw = &adapter->hw;
2243 	u32 eicr = adapter->interrupt_event;
2244 
2245 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2246 		return;
2247 
2248 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) &&
2249 	    !(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2250 		return;
2251 
2252 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2253 
2254 	switch (hw->device_id) {
2255 	case IXGBE_DEV_ID_82599_T3_LOM:
2256 		/*
2257 		 * Since the warning interrupt is for both ports
2258 		 * we don't have to check if:
2259 		 *  - This interrupt wasn't for our port.
2260 		 *  - We may have missed the interrupt so always have to
2261 		 *    check if we  got a LSC
2262 		 */
2263 		if (!(eicr & IXGBE_EICR_GPI_SDP0) &&
2264 		    !(eicr & IXGBE_EICR_LSC))
2265 			return;
2266 
2267 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2268 			u32 speed;
2269 			bool link_up = false;
2270 
2271 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2272 
2273 			if (link_up)
2274 				return;
2275 		}
2276 
2277 		/* Check if this is not due to overtemp */
2278 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2279 			return;
2280 
2281 		break;
2282 	default:
2283 		if (!(eicr & IXGBE_EICR_GPI_SDP0))
2284 			return;
2285 		break;
2286 	}
2287 	e_crit(drv,
2288 	       "Network adapter has been stopped because it has over heated. "
2289 	       "Restart the computer. If the problem persists, "
2290 	       "power off the system and replace the adapter\n");
2291 
2292 	adapter->interrupt_event = 0;
2293 }
2294 
2295 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2296 {
2297 	struct ixgbe_hw *hw = &adapter->hw;
2298 
2299 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2300 	    (eicr & IXGBE_EICR_GPI_SDP1)) {
2301 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2302 		/* write to clear the interrupt */
2303 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2304 	}
2305 }
2306 
2307 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2308 {
2309 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2310 		return;
2311 
2312 	switch (adapter->hw.mac.type) {
2313 	case ixgbe_mac_82599EB:
2314 		/*
2315 		 * Need to check link state so complete overtemp check
2316 		 * on service task
2317 		 */
2318 		if (((eicr & IXGBE_EICR_GPI_SDP0) || (eicr & IXGBE_EICR_LSC)) &&
2319 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2320 			adapter->interrupt_event = eicr;
2321 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2322 			ixgbe_service_event_schedule(adapter);
2323 			return;
2324 		}
2325 		return;
2326 	case ixgbe_mac_X540:
2327 		if (!(eicr & IXGBE_EICR_TS))
2328 			return;
2329 		break;
2330 	default:
2331 		return;
2332 	}
2333 
2334 	e_crit(drv,
2335 	       "Network adapter has been stopped because it has over heated. "
2336 	       "Restart the computer. If the problem persists, "
2337 	       "power off the system and replace the adapter\n");
2338 }
2339 
2340 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2341 {
2342 	struct ixgbe_hw *hw = &adapter->hw;
2343 
2344 	if (eicr & IXGBE_EICR_GPI_SDP2) {
2345 		/* Clear the interrupt */
2346 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP2);
2347 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2348 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2349 			ixgbe_service_event_schedule(adapter);
2350 		}
2351 	}
2352 
2353 	if (eicr & IXGBE_EICR_GPI_SDP1) {
2354 		/* Clear the interrupt */
2355 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2356 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2357 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2358 			ixgbe_service_event_schedule(adapter);
2359 		}
2360 	}
2361 }
2362 
2363 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2364 {
2365 	struct ixgbe_hw *hw = &adapter->hw;
2366 
2367 	adapter->lsc_int++;
2368 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2369 	adapter->link_check_timeout = jiffies;
2370 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2371 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2372 		IXGBE_WRITE_FLUSH(hw);
2373 		ixgbe_service_event_schedule(adapter);
2374 	}
2375 }
2376 
2377 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2378 					   u64 qmask)
2379 {
2380 	u32 mask;
2381 	struct ixgbe_hw *hw = &adapter->hw;
2382 
2383 	switch (hw->mac.type) {
2384 	case ixgbe_mac_82598EB:
2385 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2386 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2387 		break;
2388 	case ixgbe_mac_82599EB:
2389 	case ixgbe_mac_X540:
2390 		mask = (qmask & 0xFFFFFFFF);
2391 		if (mask)
2392 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2393 		mask = (qmask >> 32);
2394 		if (mask)
2395 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2396 		break;
2397 	default:
2398 		break;
2399 	}
2400 	/* skip the flush */
2401 }
2402 
2403 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2404 					    u64 qmask)
2405 {
2406 	u32 mask;
2407 	struct ixgbe_hw *hw = &adapter->hw;
2408 
2409 	switch (hw->mac.type) {
2410 	case ixgbe_mac_82598EB:
2411 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2412 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2413 		break;
2414 	case ixgbe_mac_82599EB:
2415 	case ixgbe_mac_X540:
2416 		mask = (qmask & 0xFFFFFFFF);
2417 		if (mask)
2418 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2419 		mask = (qmask >> 32);
2420 		if (mask)
2421 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2422 		break;
2423 	default:
2424 		break;
2425 	}
2426 	/* skip the flush */
2427 }
2428 
2429 /**
2430  * ixgbe_irq_enable - Enable default interrupt generation settings
2431  * @adapter: board private structure
2432  **/
2433 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2434 				    bool flush)
2435 {
2436 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2437 
2438 	/* don't reenable LSC while waiting for link */
2439 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2440 		mask &= ~IXGBE_EIMS_LSC;
2441 
2442 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2443 		switch (adapter->hw.mac.type) {
2444 		case ixgbe_mac_82599EB:
2445 			mask |= IXGBE_EIMS_GPI_SDP0;
2446 			break;
2447 		case ixgbe_mac_X540:
2448 			mask |= IXGBE_EIMS_TS;
2449 			break;
2450 		default:
2451 			break;
2452 		}
2453 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2454 		mask |= IXGBE_EIMS_GPI_SDP1;
2455 	switch (adapter->hw.mac.type) {
2456 	case ixgbe_mac_82599EB:
2457 		mask |= IXGBE_EIMS_GPI_SDP1;
2458 		mask |= IXGBE_EIMS_GPI_SDP2;
2459 	case ixgbe_mac_X540:
2460 		mask |= IXGBE_EIMS_ECC;
2461 		mask |= IXGBE_EIMS_MAILBOX;
2462 		break;
2463 	default:
2464 		break;
2465 	}
2466 
2467 	if (adapter->hw.mac.type == ixgbe_mac_X540)
2468 		mask |= IXGBE_EIMS_TIMESYNC;
2469 
2470 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2471 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
2472 		mask |= IXGBE_EIMS_FLOW_DIR;
2473 
2474 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
2475 	if (queues)
2476 		ixgbe_irq_enable_queues(adapter, ~0);
2477 	if (flush)
2478 		IXGBE_WRITE_FLUSH(&adapter->hw);
2479 }
2480 
2481 static irqreturn_t ixgbe_msix_other(int irq, void *data)
2482 {
2483 	struct ixgbe_adapter *adapter = data;
2484 	struct ixgbe_hw *hw = &adapter->hw;
2485 	u32 eicr;
2486 
2487 	/*
2488 	 * Workaround for Silicon errata.  Use clear-by-write instead
2489 	 * of clear-by-read.  Reading with EICS will return the
2490 	 * interrupt causes without clearing, which later be done
2491 	 * with the write to EICR.
2492 	 */
2493 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2494 
2495 	/* The lower 16bits of the EICR register are for the queue interrupts
2496 	 * which should be masked here in order to not accidently clear them if
2497 	 * the bits are high when ixgbe_msix_other is called. There is a race
2498 	 * condition otherwise which results in possible performance loss
2499 	 * especially if the ixgbe_msix_other interrupt is triggering
2500 	 * consistently (as it would when PPS is turned on for the X540 device)
2501 	 */
2502 	eicr &= 0xFFFF0000;
2503 
2504 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2505 
2506 	if (eicr & IXGBE_EICR_LSC)
2507 		ixgbe_check_lsc(adapter);
2508 
2509 	if (eicr & IXGBE_EICR_MAILBOX)
2510 		ixgbe_msg_task(adapter);
2511 
2512 	switch (hw->mac.type) {
2513 	case ixgbe_mac_82599EB:
2514 	case ixgbe_mac_X540:
2515 		if (eicr & IXGBE_EICR_ECC)
2516 			e_info(link, "Received unrecoverable ECC Err, please "
2517 			       "reboot\n");
2518 		/* Handle Flow Director Full threshold interrupt */
2519 		if (eicr & IXGBE_EICR_FLOW_DIR) {
2520 			int reinit_count = 0;
2521 			int i;
2522 			for (i = 0; i < adapter->num_tx_queues; i++) {
2523 				struct ixgbe_ring *ring = adapter->tx_ring[i];
2524 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
2525 						       &ring->state))
2526 					reinit_count++;
2527 			}
2528 			if (reinit_count) {
2529 				/* no more flow director interrupts until after init */
2530 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
2531 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
2532 				ixgbe_service_event_schedule(adapter);
2533 			}
2534 		}
2535 		ixgbe_check_sfp_event(adapter, eicr);
2536 		ixgbe_check_overtemp_event(adapter, eicr);
2537 		break;
2538 	default:
2539 		break;
2540 	}
2541 
2542 	ixgbe_check_fan_failure(adapter, eicr);
2543 
2544 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2545 		ixgbe_ptp_check_pps_event(adapter, eicr);
2546 
2547 	/* re-enable the original interrupt state, no lsc, no queues */
2548 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2549 		ixgbe_irq_enable(adapter, false, false);
2550 
2551 	return IRQ_HANDLED;
2552 }
2553 
2554 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
2555 {
2556 	struct ixgbe_q_vector *q_vector = data;
2557 
2558 	/* EIAM disabled interrupts (on this vector) for us */
2559 
2560 	if (q_vector->rx.ring || q_vector->tx.ring)
2561 		napi_schedule(&q_vector->napi);
2562 
2563 	return IRQ_HANDLED;
2564 }
2565 
2566 /**
2567  * ixgbe_poll - NAPI Rx polling callback
2568  * @napi: structure for representing this polling device
2569  * @budget: how many packets driver is allowed to clean
2570  *
2571  * This function is used for legacy and MSI, NAPI mode
2572  **/
2573 int ixgbe_poll(struct napi_struct *napi, int budget)
2574 {
2575 	struct ixgbe_q_vector *q_vector =
2576 				container_of(napi, struct ixgbe_q_vector, napi);
2577 	struct ixgbe_adapter *adapter = q_vector->adapter;
2578 	struct ixgbe_ring *ring;
2579 	int per_ring_budget;
2580 	bool clean_complete = true;
2581 
2582 #ifdef CONFIG_IXGBE_DCA
2583 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
2584 		ixgbe_update_dca(q_vector);
2585 #endif
2586 
2587 	ixgbe_for_each_ring(ring, q_vector->tx)
2588 		clean_complete &= !!ixgbe_clean_tx_irq(q_vector, ring);
2589 
2590 	if (!ixgbe_qv_lock_napi(q_vector))
2591 		return budget;
2592 
2593 	/* attempt to distribute budget to each queue fairly, but don't allow
2594 	 * the budget to go below 1 because we'll exit polling */
2595 	if (q_vector->rx.count > 1)
2596 		per_ring_budget = max(budget/q_vector->rx.count, 1);
2597 	else
2598 		per_ring_budget = budget;
2599 
2600 	ixgbe_for_each_ring(ring, q_vector->rx)
2601 		clean_complete &= (ixgbe_clean_rx_irq(q_vector, ring,
2602 				   per_ring_budget) < per_ring_budget);
2603 
2604 	ixgbe_qv_unlock_napi(q_vector);
2605 	/* If all work not completed, return budget and keep polling */
2606 	if (!clean_complete)
2607 		return budget;
2608 
2609 	/* all work done, exit the polling mode */
2610 	napi_complete(napi);
2611 	if (adapter->rx_itr_setting & 1)
2612 		ixgbe_set_itr(q_vector);
2613 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2614 		ixgbe_irq_enable_queues(adapter, ((u64)1 << q_vector->v_idx));
2615 
2616 	return 0;
2617 }
2618 
2619 /**
2620  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2621  * @adapter: board private structure
2622  *
2623  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2624  * interrupts from the kernel.
2625  **/
2626 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
2627 {
2628 	struct net_device *netdev = adapter->netdev;
2629 	int vector, err;
2630 	int ri = 0, ti = 0;
2631 
2632 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2633 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2634 		struct msix_entry *entry = &adapter->msix_entries[vector];
2635 
2636 		if (q_vector->tx.ring && q_vector->rx.ring) {
2637 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2638 				 "%s-%s-%d", netdev->name, "TxRx", ri++);
2639 			ti++;
2640 		} else if (q_vector->rx.ring) {
2641 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2642 				 "%s-%s-%d", netdev->name, "rx", ri++);
2643 		} else if (q_vector->tx.ring) {
2644 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2645 				 "%s-%s-%d", netdev->name, "tx", ti++);
2646 		} else {
2647 			/* skip this unused q_vector */
2648 			continue;
2649 		}
2650 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
2651 				  q_vector->name, q_vector);
2652 		if (err) {
2653 			e_err(probe, "request_irq failed for MSIX interrupt "
2654 			      "Error: %d\n", err);
2655 			goto free_queue_irqs;
2656 		}
2657 		/* If Flow Director is enabled, set interrupt affinity */
2658 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2659 			/* assign the mask for this irq */
2660 			irq_set_affinity_hint(entry->vector,
2661 					      &q_vector->affinity_mask);
2662 		}
2663 	}
2664 
2665 	err = request_irq(adapter->msix_entries[vector].vector,
2666 			  ixgbe_msix_other, 0, netdev->name, adapter);
2667 	if (err) {
2668 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
2669 		goto free_queue_irqs;
2670 	}
2671 
2672 	return 0;
2673 
2674 free_queue_irqs:
2675 	while (vector) {
2676 		vector--;
2677 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
2678 				      NULL);
2679 		free_irq(adapter->msix_entries[vector].vector,
2680 			 adapter->q_vector[vector]);
2681 	}
2682 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
2683 	pci_disable_msix(adapter->pdev);
2684 	kfree(adapter->msix_entries);
2685 	adapter->msix_entries = NULL;
2686 	return err;
2687 }
2688 
2689 /**
2690  * ixgbe_intr - legacy mode Interrupt Handler
2691  * @irq: interrupt number
2692  * @data: pointer to a network interface device structure
2693  **/
2694 static irqreturn_t ixgbe_intr(int irq, void *data)
2695 {
2696 	struct ixgbe_adapter *adapter = data;
2697 	struct ixgbe_hw *hw = &adapter->hw;
2698 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2699 	u32 eicr;
2700 
2701 	/*
2702 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
2703 	 * before the read of EICR.
2704 	 */
2705 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
2706 
2707 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2708 	 * therefore no explicit interrupt disable is necessary */
2709 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2710 	if (!eicr) {
2711 		/*
2712 		 * shared interrupt alert!
2713 		 * make sure interrupts are enabled because the read will
2714 		 * have disabled interrupts due to EIAM
2715 		 * finish the workaround of silicon errata on 82598.  Unmask
2716 		 * the interrupt that we masked before the EICR read.
2717 		 */
2718 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
2719 			ixgbe_irq_enable(adapter, true, true);
2720 		return IRQ_NONE;	/* Not our interrupt */
2721 	}
2722 
2723 	if (eicr & IXGBE_EICR_LSC)
2724 		ixgbe_check_lsc(adapter);
2725 
2726 	switch (hw->mac.type) {
2727 	case ixgbe_mac_82599EB:
2728 		ixgbe_check_sfp_event(adapter, eicr);
2729 		/* Fall through */
2730 	case ixgbe_mac_X540:
2731 		if (eicr & IXGBE_EICR_ECC)
2732 			e_info(link, "Received unrecoverable ECC err, please "
2733 				     "reboot\n");
2734 		ixgbe_check_overtemp_event(adapter, eicr);
2735 		break;
2736 	default:
2737 		break;
2738 	}
2739 
2740 	ixgbe_check_fan_failure(adapter, eicr);
2741 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2742 		ixgbe_ptp_check_pps_event(adapter, eicr);
2743 
2744 	/* would disable interrupts here but EIAM disabled it */
2745 	napi_schedule(&q_vector->napi);
2746 
2747 	/*
2748 	 * re-enable link(maybe) and non-queue interrupts, no flush.
2749 	 * ixgbe_poll will re-enable the queue interrupts
2750 	 */
2751 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2752 		ixgbe_irq_enable(adapter, false, false);
2753 
2754 	return IRQ_HANDLED;
2755 }
2756 
2757 /**
2758  * ixgbe_request_irq - initialize interrupts
2759  * @adapter: board private structure
2760  *
2761  * Attempts to configure interrupts using the best available
2762  * capabilities of the hardware and kernel.
2763  **/
2764 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
2765 {
2766 	struct net_device *netdev = adapter->netdev;
2767 	int err;
2768 
2769 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
2770 		err = ixgbe_request_msix_irqs(adapter);
2771 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
2772 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
2773 				  netdev->name, adapter);
2774 	else
2775 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
2776 				  netdev->name, adapter);
2777 
2778 	if (err)
2779 		e_err(probe, "request_irq failed, Error %d\n", err);
2780 
2781 	return err;
2782 }
2783 
2784 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
2785 {
2786 	int vector;
2787 
2788 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
2789 		free_irq(adapter->pdev->irq, adapter);
2790 		return;
2791 	}
2792 
2793 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2794 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2795 		struct msix_entry *entry = &adapter->msix_entries[vector];
2796 
2797 		/* free only the irqs that were actually requested */
2798 		if (!q_vector->rx.ring && !q_vector->tx.ring)
2799 			continue;
2800 
2801 		/* clear the affinity_mask in the IRQ descriptor */
2802 		irq_set_affinity_hint(entry->vector, NULL);
2803 
2804 		free_irq(entry->vector, q_vector);
2805 	}
2806 
2807 	free_irq(adapter->msix_entries[vector++].vector, adapter);
2808 }
2809 
2810 /**
2811  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
2812  * @adapter: board private structure
2813  **/
2814 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
2815 {
2816 	switch (adapter->hw.mac.type) {
2817 	case ixgbe_mac_82598EB:
2818 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
2819 		break;
2820 	case ixgbe_mac_82599EB:
2821 	case ixgbe_mac_X540:
2822 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
2823 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
2824 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
2825 		break;
2826 	default:
2827 		break;
2828 	}
2829 	IXGBE_WRITE_FLUSH(&adapter->hw);
2830 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
2831 		int vector;
2832 
2833 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
2834 			synchronize_irq(adapter->msix_entries[vector].vector);
2835 
2836 		synchronize_irq(adapter->msix_entries[vector++].vector);
2837 	} else {
2838 		synchronize_irq(adapter->pdev->irq);
2839 	}
2840 }
2841 
2842 /**
2843  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
2844  *
2845  **/
2846 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
2847 {
2848 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2849 
2850 	ixgbe_write_eitr(q_vector);
2851 
2852 	ixgbe_set_ivar(adapter, 0, 0, 0);
2853 	ixgbe_set_ivar(adapter, 1, 0, 0);
2854 
2855 	e_info(hw, "Legacy interrupt IVAR setup done\n");
2856 }
2857 
2858 /**
2859  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
2860  * @adapter: board private structure
2861  * @ring: structure containing ring specific data
2862  *
2863  * Configure the Tx descriptor ring after a reset.
2864  **/
2865 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
2866 			     struct ixgbe_ring *ring)
2867 {
2868 	struct ixgbe_hw *hw = &adapter->hw;
2869 	u64 tdba = ring->dma;
2870 	int wait_loop = 10;
2871 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
2872 	u8 reg_idx = ring->reg_idx;
2873 
2874 	/* disable queue to avoid issues while updating state */
2875 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
2876 	IXGBE_WRITE_FLUSH(hw);
2877 
2878 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
2879 			(tdba & DMA_BIT_MASK(32)));
2880 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
2881 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
2882 			ring->count * sizeof(union ixgbe_adv_tx_desc));
2883 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
2884 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
2885 	ring->tail = hw->hw_addr + IXGBE_TDT(reg_idx);
2886 
2887 	/*
2888 	 * set WTHRESH to encourage burst writeback, it should not be set
2889 	 * higher than 1 when:
2890 	 * - ITR is 0 as it could cause false TX hangs
2891 	 * - ITR is set to > 100k int/sec and BQL is enabled
2892 	 *
2893 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
2894 	 * to or less than the number of on chip descriptors, which is
2895 	 * currently 40.
2896 	 */
2897 #if IS_ENABLED(CONFIG_BQL)
2898 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
2899 #else
2900 	if (!ring->q_vector || (ring->q_vector->itr < 8))
2901 #endif
2902 		txdctl |= (1 << 16);	/* WTHRESH = 1 */
2903 	else
2904 		txdctl |= (8 << 16);	/* WTHRESH = 8 */
2905 
2906 	/*
2907 	 * Setting PTHRESH to 32 both improves performance
2908 	 * and avoids a TX hang with DFP enabled
2909 	 */
2910 	txdctl |= (1 << 8) |	/* HTHRESH = 1 */
2911 		   32;		/* PTHRESH = 32 */
2912 
2913 	/* reinitialize flowdirector state */
2914 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2915 		ring->atr_sample_rate = adapter->atr_sample_rate;
2916 		ring->atr_count = 0;
2917 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
2918 	} else {
2919 		ring->atr_sample_rate = 0;
2920 	}
2921 
2922 	/* initialize XPS */
2923 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
2924 		struct ixgbe_q_vector *q_vector = ring->q_vector;
2925 
2926 		if (q_vector)
2927 			netif_set_xps_queue(adapter->netdev,
2928 					    &q_vector->affinity_mask,
2929 					    ring->queue_index);
2930 	}
2931 
2932 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
2933 
2934 	/* enable queue */
2935 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
2936 
2937 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
2938 	if (hw->mac.type == ixgbe_mac_82598EB &&
2939 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
2940 		return;
2941 
2942 	/* poll to verify queue is enabled */
2943 	do {
2944 		usleep_range(1000, 2000);
2945 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
2946 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
2947 	if (!wait_loop)
2948 		e_err(drv, "Could not enable Tx Queue %d\n", reg_idx);
2949 }
2950 
2951 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
2952 {
2953 	struct ixgbe_hw *hw = &adapter->hw;
2954 	u32 rttdcs, mtqc;
2955 	u8 tcs = netdev_get_num_tc(adapter->netdev);
2956 
2957 	if (hw->mac.type == ixgbe_mac_82598EB)
2958 		return;
2959 
2960 	/* disable the arbiter while setting MTQC */
2961 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
2962 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
2963 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
2964 
2965 	/* set transmit pool layout */
2966 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
2967 		mtqc = IXGBE_MTQC_VT_ENA;
2968 		if (tcs > 4)
2969 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
2970 		else if (tcs > 1)
2971 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
2972 		else if (adapter->ring_feature[RING_F_RSS].indices == 4)
2973 			mtqc |= IXGBE_MTQC_32VF;
2974 		else
2975 			mtqc |= IXGBE_MTQC_64VF;
2976 	} else {
2977 		if (tcs > 4)
2978 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
2979 		else if (tcs > 1)
2980 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
2981 		else
2982 			mtqc = IXGBE_MTQC_64Q_1PB;
2983 	}
2984 
2985 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
2986 
2987 	/* Enable Security TX Buffer IFG for multiple pb */
2988 	if (tcs) {
2989 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
2990 		sectx |= IXGBE_SECTX_DCB;
2991 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
2992 	}
2993 
2994 	/* re-enable the arbiter */
2995 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
2996 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
2997 }
2998 
2999 /**
3000  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3001  * @adapter: board private structure
3002  *
3003  * Configure the Tx unit of the MAC after a reset.
3004  **/
3005 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3006 {
3007 	struct ixgbe_hw *hw = &adapter->hw;
3008 	u32 dmatxctl;
3009 	u32 i;
3010 
3011 	ixgbe_setup_mtqc(adapter);
3012 
3013 	if (hw->mac.type != ixgbe_mac_82598EB) {
3014 		/* DMATXCTL.EN must be before Tx queues are enabled */
3015 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3016 		dmatxctl |= IXGBE_DMATXCTL_TE;
3017 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3018 	}
3019 
3020 	/* Setup the HW Tx Head and Tail descriptor pointers */
3021 	for (i = 0; i < adapter->num_tx_queues; i++)
3022 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3023 }
3024 
3025 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3026 				 struct ixgbe_ring *ring)
3027 {
3028 	struct ixgbe_hw *hw = &adapter->hw;
3029 	u8 reg_idx = ring->reg_idx;
3030 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3031 
3032 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3033 
3034 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3035 }
3036 
3037 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3038 				  struct ixgbe_ring *ring)
3039 {
3040 	struct ixgbe_hw *hw = &adapter->hw;
3041 	u8 reg_idx = ring->reg_idx;
3042 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3043 
3044 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3045 
3046 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3047 }
3048 
3049 #ifdef CONFIG_IXGBE_DCB
3050 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3051 #else
3052 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3053 #endif
3054 {
3055 	int i;
3056 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3057 
3058 	if (adapter->ixgbe_ieee_pfc)
3059 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3060 
3061 	/*
3062 	 * We should set the drop enable bit if:
3063 	 *  SR-IOV is enabled
3064 	 *   or
3065 	 *  Number of Rx queues > 1 and flow control is disabled
3066 	 *
3067 	 *  This allows us to avoid head of line blocking for security
3068 	 *  and performance reasons.
3069 	 */
3070 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3071 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3072 		for (i = 0; i < adapter->num_rx_queues; i++)
3073 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3074 	} else {
3075 		for (i = 0; i < adapter->num_rx_queues; i++)
3076 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3077 	}
3078 }
3079 
3080 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3081 
3082 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3083 				   struct ixgbe_ring *rx_ring)
3084 {
3085 	struct ixgbe_hw *hw = &adapter->hw;
3086 	u32 srrctl;
3087 	u8 reg_idx = rx_ring->reg_idx;
3088 
3089 	if (hw->mac.type == ixgbe_mac_82598EB) {
3090 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3091 
3092 		/*
3093 		 * if VMDq is not active we must program one srrctl register
3094 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3095 		 */
3096 		reg_idx &= mask;
3097 	}
3098 
3099 	/* configure header buffer length, needed for RSC */
3100 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3101 
3102 	/* configure the packet buffer length */
3103 	srrctl |= ixgbe_rx_bufsz(rx_ring) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3104 
3105 	/* configure descriptor type */
3106 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3107 
3108 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3109 }
3110 
3111 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3112 {
3113 	struct ixgbe_hw *hw = &adapter->hw;
3114 	static const u32 seed[10] = { 0xE291D73D, 0x1805EC6C, 0x2A94B30D,
3115 			  0xA54F2BEC, 0xEA49AF7C, 0xE214AD3D, 0xB855AABE,
3116 			  0x6A3E67EA, 0x14364D17, 0x3BED200D};
3117 	u32 mrqc = 0, reta = 0;
3118 	u32 rxcsum;
3119 	int i, j;
3120 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3121 
3122 	/*
3123 	 * Program table for at least 2 queues w/ SR-IOV so that VFs can
3124 	 * make full use of any rings they may have.  We will use the
3125 	 * PSRTYPE register to control how many rings we use within the PF.
3126 	 */
3127 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 2))
3128 		rss_i = 2;
3129 
3130 	/* Fill out hash function seeds */
3131 	for (i = 0; i < 10; i++)
3132 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), seed[i]);
3133 
3134 	/* Fill out redirection table */
3135 	for (i = 0, j = 0; i < 128; i++, j++) {
3136 		if (j == rss_i)
3137 			j = 0;
3138 		/* reta = 4-byte sliding window of
3139 		 * 0x00..(indices-1)(indices-1)00..etc. */
3140 		reta = (reta << 8) | (j * 0x11);
3141 		if ((i & 3) == 3)
3142 			IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3143 	}
3144 
3145 	/* Disable indicating checksum in descriptor, enables RSS hash */
3146 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3147 	rxcsum |= IXGBE_RXCSUM_PCSD;
3148 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3149 
3150 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3151 		if (adapter->ring_feature[RING_F_RSS].mask)
3152 			mrqc = IXGBE_MRQC_RSSEN;
3153 	} else {
3154 		u8 tcs = netdev_get_num_tc(adapter->netdev);
3155 
3156 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3157 			if (tcs > 4)
3158 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3159 			else if (tcs > 1)
3160 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3161 			else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3162 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3163 			else
3164 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3165 		} else {
3166 			if (tcs > 4)
3167 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3168 			else if (tcs > 1)
3169 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3170 			else
3171 				mrqc = IXGBE_MRQC_RSSEN;
3172 		}
3173 	}
3174 
3175 	/* Perform hash on these packet types */
3176 	mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3177 		IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3178 		IXGBE_MRQC_RSS_FIELD_IPV6 |
3179 		IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3180 
3181 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3182 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3183 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3184 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3185 
3186 	IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3187 }
3188 
3189 /**
3190  * ixgbe_configure_rscctl - enable RSC for the indicated ring
3191  * @adapter:    address of board private structure
3192  * @index:      index of ring to set
3193  **/
3194 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
3195 				   struct ixgbe_ring *ring)
3196 {
3197 	struct ixgbe_hw *hw = &adapter->hw;
3198 	u32 rscctrl;
3199 	u8 reg_idx = ring->reg_idx;
3200 
3201 	if (!ring_is_rsc_enabled(ring))
3202 		return;
3203 
3204 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
3205 	rscctrl |= IXGBE_RSCCTL_RSCEN;
3206 	/*
3207 	 * we must limit the number of descriptors so that the
3208 	 * total size of max desc * buf_len is not greater
3209 	 * than 65536
3210 	 */
3211 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
3212 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
3213 }
3214 
3215 #define IXGBE_MAX_RX_DESC_POLL 10
3216 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
3217 				       struct ixgbe_ring *ring)
3218 {
3219 	struct ixgbe_hw *hw = &adapter->hw;
3220 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3221 	u32 rxdctl;
3222 	u8 reg_idx = ring->reg_idx;
3223 
3224 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3225 	if (hw->mac.type == ixgbe_mac_82598EB &&
3226 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3227 		return;
3228 
3229 	do {
3230 		usleep_range(1000, 2000);
3231 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3232 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
3233 
3234 	if (!wait_loop) {
3235 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
3236 		      "the polling period\n", reg_idx);
3237 	}
3238 }
3239 
3240 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
3241 			    struct ixgbe_ring *ring)
3242 {
3243 	struct ixgbe_hw *hw = &adapter->hw;
3244 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3245 	u32 rxdctl;
3246 	u8 reg_idx = ring->reg_idx;
3247 
3248 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3249 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
3250 
3251 	/* write value back with RXDCTL.ENABLE bit cleared */
3252 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3253 
3254 	if (hw->mac.type == ixgbe_mac_82598EB &&
3255 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3256 		return;
3257 
3258 	/* the hardware may take up to 100us to really disable the rx queue */
3259 	do {
3260 		udelay(10);
3261 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3262 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
3263 
3264 	if (!wait_loop) {
3265 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3266 		      "the polling period\n", reg_idx);
3267 	}
3268 }
3269 
3270 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
3271 			     struct ixgbe_ring *ring)
3272 {
3273 	struct ixgbe_hw *hw = &adapter->hw;
3274 	u64 rdba = ring->dma;
3275 	u32 rxdctl;
3276 	u8 reg_idx = ring->reg_idx;
3277 
3278 	/* disable queue to avoid issues while updating state */
3279 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3280 	ixgbe_disable_rx_queue(adapter, ring);
3281 
3282 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
3283 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
3284 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
3285 			ring->count * sizeof(union ixgbe_adv_rx_desc));
3286 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
3287 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
3288 	ring->tail = hw->hw_addr + IXGBE_RDT(reg_idx);
3289 
3290 	ixgbe_configure_srrctl(adapter, ring);
3291 	ixgbe_configure_rscctl(adapter, ring);
3292 
3293 	if (hw->mac.type == ixgbe_mac_82598EB) {
3294 		/*
3295 		 * enable cache line friendly hardware writes:
3296 		 * PTHRESH=32 descriptors (half the internal cache),
3297 		 * this also removes ugly rx_no_buffer_count increment
3298 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
3299 		 * WTHRESH=8 burst writeback up to two cache lines
3300 		 */
3301 		rxdctl &= ~0x3FFFFF;
3302 		rxdctl |=  0x080420;
3303 	}
3304 
3305 	/* enable receive descriptor ring */
3306 	rxdctl |= IXGBE_RXDCTL_ENABLE;
3307 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3308 
3309 	ixgbe_rx_desc_queue_enable(adapter, ring);
3310 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
3311 }
3312 
3313 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
3314 {
3315 	struct ixgbe_hw *hw = &adapter->hw;
3316 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
3317 	int p;
3318 
3319 	/* PSRTYPE must be initialized in non 82598 adapters */
3320 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
3321 		      IXGBE_PSRTYPE_UDPHDR |
3322 		      IXGBE_PSRTYPE_IPV4HDR |
3323 		      IXGBE_PSRTYPE_L2HDR |
3324 		      IXGBE_PSRTYPE_IPV6HDR;
3325 
3326 	if (hw->mac.type == ixgbe_mac_82598EB)
3327 		return;
3328 
3329 	if (rss_i > 3)
3330 		psrtype |= 2 << 29;
3331 	else if (rss_i > 1)
3332 		psrtype |= 1 << 29;
3333 
3334 	for (p = 0; p < adapter->num_rx_pools; p++)
3335 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(p)),
3336 				psrtype);
3337 }
3338 
3339 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
3340 {
3341 	struct ixgbe_hw *hw = &adapter->hw;
3342 	u32 reg_offset, vf_shift;
3343 	u32 gcr_ext, vmdctl;
3344 	int i;
3345 
3346 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3347 		return;
3348 
3349 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3350 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
3351 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
3352 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
3353 	vmdctl |= IXGBE_VT_CTL_REPLEN;
3354 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
3355 
3356 	vf_shift = VMDQ_P(0) % 32;
3357 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
3358 
3359 	/* Enable only the PF's pool for Tx/Rx */
3360 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), (~0) << vf_shift);
3361 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
3362 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), (~0) << vf_shift);
3363 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
3364 	if (adapter->flags2 & IXGBE_FLAG2_BRIDGE_MODE_VEB)
3365 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3366 
3367 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3368 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
3369 
3370 	/*
3371 	 * Set up VF register offsets for selected VT Mode,
3372 	 * i.e. 32 or 64 VFs for SR-IOV
3373 	 */
3374 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
3375 	case IXGBE_82599_VMDQ_8Q_MASK:
3376 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
3377 		break;
3378 	case IXGBE_82599_VMDQ_4Q_MASK:
3379 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
3380 		break;
3381 	default:
3382 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
3383 		break;
3384 	}
3385 
3386 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
3387 
3388 
3389 	/* Enable MAC Anti-Spoofing */
3390 	hw->mac.ops.set_mac_anti_spoofing(hw, (adapter->num_vfs != 0),
3391 					  adapter->num_vfs);
3392 	/* For VFs that have spoof checking turned off */
3393 	for (i = 0; i < adapter->num_vfs; i++) {
3394 		if (!adapter->vfinfo[i].spoofchk_enabled)
3395 			ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i, false);
3396 	}
3397 }
3398 
3399 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
3400 {
3401 	struct ixgbe_hw *hw = &adapter->hw;
3402 	struct net_device *netdev = adapter->netdev;
3403 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3404 	struct ixgbe_ring *rx_ring;
3405 	int i;
3406 	u32 mhadd, hlreg0;
3407 
3408 #ifdef IXGBE_FCOE
3409 	/* adjust max frame to be able to do baby jumbo for FCoE */
3410 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
3411 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
3412 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3413 
3414 #endif /* IXGBE_FCOE */
3415 
3416 	/* adjust max frame to be at least the size of a standard frame */
3417 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
3418 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
3419 
3420 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3421 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
3422 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3423 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
3424 
3425 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3426 	}
3427 
3428 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3429 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3430 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3431 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3432 
3433 	/*
3434 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3435 	 * the Base and Length of the Rx Descriptor Ring
3436 	 */
3437 	for (i = 0; i < adapter->num_rx_queues; i++) {
3438 		rx_ring = adapter->rx_ring[i];
3439 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
3440 			set_ring_rsc_enabled(rx_ring);
3441 		else
3442 			clear_ring_rsc_enabled(rx_ring);
3443 	}
3444 }
3445 
3446 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
3447 {
3448 	struct ixgbe_hw *hw = &adapter->hw;
3449 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
3450 
3451 	switch (hw->mac.type) {
3452 	case ixgbe_mac_82598EB:
3453 		/*
3454 		 * For VMDq support of different descriptor types or
3455 		 * buffer sizes through the use of multiple SRRCTL
3456 		 * registers, RDRXCTL.MVMEN must be set to 1
3457 		 *
3458 		 * also, the manual doesn't mention it clearly but DCA hints
3459 		 * will only use queue 0's tags unless this bit is set.  Side
3460 		 * effects of setting this bit are only that SRRCTL must be
3461 		 * fully programmed [0..15]
3462 		 */
3463 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
3464 		break;
3465 	case ixgbe_mac_82599EB:
3466 	case ixgbe_mac_X540:
3467 		/* Disable RSC for ACK packets */
3468 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
3469 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
3470 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
3471 		/* hardware requires some bits to be set by default */
3472 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
3473 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
3474 		break;
3475 	default:
3476 		/* We should do nothing since we don't know this hardware */
3477 		return;
3478 	}
3479 
3480 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
3481 }
3482 
3483 /**
3484  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3485  * @adapter: board private structure
3486  *
3487  * Configure the Rx unit of the MAC after a reset.
3488  **/
3489 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
3490 {
3491 	struct ixgbe_hw *hw = &adapter->hw;
3492 	int i;
3493 	u32 rxctrl;
3494 
3495 	/* disable receives while setting up the descriptors */
3496 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3497 	IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxctrl & ~IXGBE_RXCTRL_RXEN);
3498 
3499 	ixgbe_setup_psrtype(adapter);
3500 	ixgbe_setup_rdrxctl(adapter);
3501 
3502 	/* Program registers for the distribution of queues */
3503 	ixgbe_setup_mrqc(adapter);
3504 
3505 	/* set_rx_buffer_len must be called before ring initialization */
3506 	ixgbe_set_rx_buffer_len(adapter);
3507 
3508 	/*
3509 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3510 	 * the Base and Length of the Rx Descriptor Ring
3511 	 */
3512 	for (i = 0; i < adapter->num_rx_queues; i++)
3513 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
3514 
3515 	/* disable drop enable for 82598 parts */
3516 	if (hw->mac.type == ixgbe_mac_82598EB)
3517 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3518 
3519 	/* enable all receives */
3520 	rxctrl |= IXGBE_RXCTRL_RXEN;
3521 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
3522 }
3523 
3524 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
3525 				 __be16 proto, u16 vid)
3526 {
3527 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3528 	struct ixgbe_hw *hw = &adapter->hw;
3529 
3530 	/* add VID to filter table */
3531 	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true);
3532 	set_bit(vid, adapter->active_vlans);
3533 
3534 	return 0;
3535 }
3536 
3537 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
3538 				  __be16 proto, u16 vid)
3539 {
3540 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3541 	struct ixgbe_hw *hw = &adapter->hw;
3542 
3543 	/* remove VID from filter table */
3544 	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), false);
3545 	clear_bit(vid, adapter->active_vlans);
3546 
3547 	return 0;
3548 }
3549 
3550 /**
3551  * ixgbe_vlan_filter_disable - helper to disable hw vlan filtering
3552  * @adapter: driver data
3553  */
3554 static void ixgbe_vlan_filter_disable(struct ixgbe_adapter *adapter)
3555 {
3556 	struct ixgbe_hw *hw = &adapter->hw;
3557 	u32 vlnctrl;
3558 
3559 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3560 	vlnctrl &= ~(IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
3561 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3562 }
3563 
3564 /**
3565  * ixgbe_vlan_filter_enable - helper to enable hw vlan filtering
3566  * @adapter: driver data
3567  */
3568 static void ixgbe_vlan_filter_enable(struct ixgbe_adapter *adapter)
3569 {
3570 	struct ixgbe_hw *hw = &adapter->hw;
3571 	u32 vlnctrl;
3572 
3573 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3574 	vlnctrl |= IXGBE_VLNCTRL_VFE;
3575 	vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
3576 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3577 }
3578 
3579 /**
3580  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
3581  * @adapter: driver data
3582  */
3583 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
3584 {
3585 	struct ixgbe_hw *hw = &adapter->hw;
3586 	u32 vlnctrl;
3587 	int i, j;
3588 
3589 	switch (hw->mac.type) {
3590 	case ixgbe_mac_82598EB:
3591 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3592 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
3593 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3594 		break;
3595 	case ixgbe_mac_82599EB:
3596 	case ixgbe_mac_X540:
3597 		for (i = 0; i < adapter->num_rx_queues; i++) {
3598 			j = adapter->rx_ring[i]->reg_idx;
3599 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3600 			vlnctrl &= ~IXGBE_RXDCTL_VME;
3601 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3602 		}
3603 		break;
3604 	default:
3605 		break;
3606 	}
3607 }
3608 
3609 /**
3610  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
3611  * @adapter: driver data
3612  */
3613 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
3614 {
3615 	struct ixgbe_hw *hw = &adapter->hw;
3616 	u32 vlnctrl;
3617 	int i, j;
3618 
3619 	switch (hw->mac.type) {
3620 	case ixgbe_mac_82598EB:
3621 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3622 		vlnctrl |= IXGBE_VLNCTRL_VME;
3623 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3624 		break;
3625 	case ixgbe_mac_82599EB:
3626 	case ixgbe_mac_X540:
3627 		for (i = 0; i < adapter->num_rx_queues; i++) {
3628 			j = adapter->rx_ring[i]->reg_idx;
3629 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3630 			vlnctrl |= IXGBE_RXDCTL_VME;
3631 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3632 		}
3633 		break;
3634 	default:
3635 		break;
3636 	}
3637 }
3638 
3639 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
3640 {
3641 	u16 vid;
3642 
3643 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
3644 
3645 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
3646 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
3647 }
3648 
3649 /**
3650  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
3651  * @netdev: network interface device structure
3652  *
3653  * Writes unicast address list to the RAR table.
3654  * Returns: -ENOMEM on failure/insufficient address space
3655  *                0 on no addresses written
3656  *                X on writing X addresses to the RAR table
3657  **/
3658 static int ixgbe_write_uc_addr_list(struct net_device *netdev)
3659 {
3660 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3661 	struct ixgbe_hw *hw = &adapter->hw;
3662 	unsigned int rar_entries = hw->mac.num_rar_entries - 1;
3663 	int count = 0;
3664 
3665 	/* In SR-IOV mode significantly less RAR entries are available */
3666 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3667 		rar_entries = IXGBE_MAX_PF_MACVLANS - 1;
3668 
3669 	/* return ENOMEM indicating insufficient memory for addresses */
3670 	if (netdev_uc_count(netdev) > rar_entries)
3671 		return -ENOMEM;
3672 
3673 	if (!netdev_uc_empty(netdev)) {
3674 		struct netdev_hw_addr *ha;
3675 		/* return error if we do not support writing to RAR table */
3676 		if (!hw->mac.ops.set_rar)
3677 			return -ENOMEM;
3678 
3679 		netdev_for_each_uc_addr(ha, netdev) {
3680 			if (!rar_entries)
3681 				break;
3682 			hw->mac.ops.set_rar(hw, rar_entries--, ha->addr,
3683 					    VMDQ_P(0), IXGBE_RAH_AV);
3684 			count++;
3685 		}
3686 	}
3687 	/* write the addresses in reverse order to avoid write combining */
3688 	for (; rar_entries > 0 ; rar_entries--)
3689 		hw->mac.ops.clear_rar(hw, rar_entries);
3690 
3691 	return count;
3692 }
3693 
3694 /**
3695  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
3696  * @netdev: network interface device structure
3697  *
3698  * The set_rx_method entry point is called whenever the unicast/multicast
3699  * address list or the network interface flags are updated.  This routine is
3700  * responsible for configuring the hardware for proper unicast, multicast and
3701  * promiscuous mode.
3702  **/
3703 void ixgbe_set_rx_mode(struct net_device *netdev)
3704 {
3705 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3706 	struct ixgbe_hw *hw = &adapter->hw;
3707 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
3708 	int count;
3709 
3710 	/* Check for Promiscuous and All Multicast modes */
3711 
3712 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3713 
3714 	/* set all bits that we expect to always be set */
3715 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
3716 	fctrl |= IXGBE_FCTRL_BAM;
3717 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
3718 	fctrl |= IXGBE_FCTRL_PMCF;
3719 
3720 	/* clear the bits we are changing the status of */
3721 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3722 
3723 	if (netdev->flags & IFF_PROMISC) {
3724 		hw->addr_ctrl.user_set_promisc = true;
3725 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3726 		vmolr |= (IXGBE_VMOLR_ROPE | IXGBE_VMOLR_MPE);
3727 		/* don't hardware filter vlans in promisc mode */
3728 		ixgbe_vlan_filter_disable(adapter);
3729 	} else {
3730 		if (netdev->flags & IFF_ALLMULTI) {
3731 			fctrl |= IXGBE_FCTRL_MPE;
3732 			vmolr |= IXGBE_VMOLR_MPE;
3733 		} else {
3734 			/*
3735 			 * Write addresses to the MTA, if the attempt fails
3736 			 * then we should just turn on promiscuous mode so
3737 			 * that we can at least receive multicast traffic
3738 			 */
3739 			hw->mac.ops.update_mc_addr_list(hw, netdev);
3740 			vmolr |= IXGBE_VMOLR_ROMPE;
3741 		}
3742 		ixgbe_vlan_filter_enable(adapter);
3743 		hw->addr_ctrl.user_set_promisc = false;
3744 	}
3745 
3746 	/*
3747 	 * Write addresses to available RAR registers, if there is not
3748 	 * sufficient space to store all the addresses then enable
3749 	 * unicast promiscuous mode
3750 	 */
3751 	count = ixgbe_write_uc_addr_list(netdev);
3752 	if (count < 0) {
3753 		fctrl |= IXGBE_FCTRL_UPE;
3754 		vmolr |= IXGBE_VMOLR_ROPE;
3755 	}
3756 
3757 	if (adapter->num_vfs)
3758 		ixgbe_restore_vf_multicasts(adapter);
3759 
3760 	if (hw->mac.type != ixgbe_mac_82598EB) {
3761 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
3762 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
3763 			   IXGBE_VMOLR_ROPE);
3764 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
3765 	}
3766 
3767 	/* This is useful for sniffing bad packets. */
3768 	if (adapter->netdev->features & NETIF_F_RXALL) {
3769 		/* UPE and MPE will be handled by normal PROMISC logic
3770 		 * in e1000e_set_rx_mode */
3771 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
3772 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
3773 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
3774 
3775 		fctrl &= ~(IXGBE_FCTRL_DPF);
3776 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
3777 	}
3778 
3779 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
3780 
3781 	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
3782 		ixgbe_vlan_strip_enable(adapter);
3783 	else
3784 		ixgbe_vlan_strip_disable(adapter);
3785 }
3786 
3787 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
3788 {
3789 	int q_idx;
3790 
3791 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
3792 		ixgbe_qv_init_lock(adapter->q_vector[q_idx]);
3793 		napi_enable(&adapter->q_vector[q_idx]->napi);
3794 	}
3795 }
3796 
3797 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
3798 {
3799 	int q_idx;
3800 
3801 	local_bh_disable(); /* for ixgbe_qv_lock_napi() */
3802 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
3803 		napi_disable(&adapter->q_vector[q_idx]->napi);
3804 		while (!ixgbe_qv_lock_napi(adapter->q_vector[q_idx])) {
3805 			pr_info("QV %d locked\n", q_idx);
3806 			mdelay(1);
3807 		}
3808 	}
3809 	local_bh_enable();
3810 }
3811 
3812 #ifdef CONFIG_IXGBE_DCB
3813 /**
3814  * ixgbe_configure_dcb - Configure DCB hardware
3815  * @adapter: ixgbe adapter struct
3816  *
3817  * This is called by the driver on open to configure the DCB hardware.
3818  * This is also called by the gennetlink interface when reconfiguring
3819  * the DCB state.
3820  */
3821 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
3822 {
3823 	struct ixgbe_hw *hw = &adapter->hw;
3824 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3825 
3826 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
3827 		if (hw->mac.type == ixgbe_mac_82598EB)
3828 			netif_set_gso_max_size(adapter->netdev, 65536);
3829 		return;
3830 	}
3831 
3832 	if (hw->mac.type == ixgbe_mac_82598EB)
3833 		netif_set_gso_max_size(adapter->netdev, 32768);
3834 
3835 #ifdef IXGBE_FCOE
3836 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
3837 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
3838 #endif
3839 
3840 	/* reconfigure the hardware */
3841 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
3842 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
3843 						DCB_TX_CONFIG);
3844 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
3845 						DCB_RX_CONFIG);
3846 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
3847 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
3848 		ixgbe_dcb_hw_ets(&adapter->hw,
3849 				 adapter->ixgbe_ieee_ets,
3850 				 max_frame);
3851 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
3852 					adapter->ixgbe_ieee_pfc->pfc_en,
3853 					adapter->ixgbe_ieee_ets->prio_tc);
3854 	}
3855 
3856 	/* Enable RSS Hash per TC */
3857 	if (hw->mac.type != ixgbe_mac_82598EB) {
3858 		u32 msb = 0;
3859 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
3860 
3861 		while (rss_i) {
3862 			msb++;
3863 			rss_i >>= 1;
3864 		}
3865 
3866 		/* write msb to all 8 TCs in one write */
3867 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
3868 	}
3869 }
3870 #endif
3871 
3872 /* Additional bittime to account for IXGBE framing */
3873 #define IXGBE_ETH_FRAMING 20
3874 
3875 /**
3876  * ixgbe_hpbthresh - calculate high water mark for flow control
3877  *
3878  * @adapter: board private structure to calculate for
3879  * @pb: packet buffer to calculate
3880  */
3881 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
3882 {
3883 	struct ixgbe_hw *hw = &adapter->hw;
3884 	struct net_device *dev = adapter->netdev;
3885 	int link, tc, kb, marker;
3886 	u32 dv_id, rx_pba;
3887 
3888 	/* Calculate max LAN frame size */
3889 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
3890 
3891 #ifdef IXGBE_FCOE
3892 	/* FCoE traffic class uses FCOE jumbo frames */
3893 	if ((dev->features & NETIF_F_FCOE_MTU) &&
3894 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
3895 	    (pb == ixgbe_fcoe_get_tc(adapter)))
3896 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3897 
3898 #endif
3899 	/* Calculate delay value for device */
3900 	switch (hw->mac.type) {
3901 	case ixgbe_mac_X540:
3902 		dv_id = IXGBE_DV_X540(link, tc);
3903 		break;
3904 	default:
3905 		dv_id = IXGBE_DV(link, tc);
3906 		break;
3907 	}
3908 
3909 	/* Loopback switch introduces additional latency */
3910 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3911 		dv_id += IXGBE_B2BT(tc);
3912 
3913 	/* Delay value is calculated in bit times convert to KB */
3914 	kb = IXGBE_BT2KB(dv_id);
3915 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
3916 
3917 	marker = rx_pba - kb;
3918 
3919 	/* It is possible that the packet buffer is not large enough
3920 	 * to provide required headroom. In this case throw an error
3921 	 * to user and a do the best we can.
3922 	 */
3923 	if (marker < 0) {
3924 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
3925 			    "headroom to support flow control."
3926 			    "Decrease MTU or number of traffic classes\n", pb);
3927 		marker = tc + 1;
3928 	}
3929 
3930 	return marker;
3931 }
3932 
3933 /**
3934  * ixgbe_lpbthresh - calculate low water mark for for flow control
3935  *
3936  * @adapter: board private structure to calculate for
3937  * @pb: packet buffer to calculate
3938  */
3939 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter)
3940 {
3941 	struct ixgbe_hw *hw = &adapter->hw;
3942 	struct net_device *dev = adapter->netdev;
3943 	int tc;
3944 	u32 dv_id;
3945 
3946 	/* Calculate max LAN frame size */
3947 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
3948 
3949 	/* Calculate delay value for device */
3950 	switch (hw->mac.type) {
3951 	case ixgbe_mac_X540:
3952 		dv_id = IXGBE_LOW_DV_X540(tc);
3953 		break;
3954 	default:
3955 		dv_id = IXGBE_LOW_DV(tc);
3956 		break;
3957 	}
3958 
3959 	/* Delay value is calculated in bit times convert to KB */
3960 	return IXGBE_BT2KB(dv_id);
3961 }
3962 
3963 /*
3964  * ixgbe_pbthresh_setup - calculate and setup high low water marks
3965  */
3966 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
3967 {
3968 	struct ixgbe_hw *hw = &adapter->hw;
3969 	int num_tc = netdev_get_num_tc(adapter->netdev);
3970 	int i;
3971 
3972 	if (!num_tc)
3973 		num_tc = 1;
3974 
3975 	hw->fc.low_water = ixgbe_lpbthresh(adapter);
3976 
3977 	for (i = 0; i < num_tc; i++) {
3978 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
3979 
3980 		/* Low water marks must not be larger than high water marks */
3981 		if (hw->fc.low_water > hw->fc.high_water[i])
3982 			hw->fc.low_water = 0;
3983 	}
3984 }
3985 
3986 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
3987 {
3988 	struct ixgbe_hw *hw = &adapter->hw;
3989 	int hdrm;
3990 	u8 tc = netdev_get_num_tc(adapter->netdev);
3991 
3992 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
3993 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
3994 		hdrm = 32 << adapter->fdir_pballoc;
3995 	else
3996 		hdrm = 0;
3997 
3998 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
3999 	ixgbe_pbthresh_setup(adapter);
4000 }
4001 
4002 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
4003 {
4004 	struct ixgbe_hw *hw = &adapter->hw;
4005 	struct hlist_node *node2;
4006 	struct ixgbe_fdir_filter *filter;
4007 
4008 	spin_lock(&adapter->fdir_perfect_lock);
4009 
4010 	if (!hlist_empty(&adapter->fdir_filter_list))
4011 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
4012 
4013 	hlist_for_each_entry_safe(filter, node2,
4014 				  &adapter->fdir_filter_list, fdir_node) {
4015 		ixgbe_fdir_write_perfect_filter_82599(hw,
4016 				&filter->filter,
4017 				filter->sw_idx,
4018 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
4019 				IXGBE_FDIR_DROP_QUEUE :
4020 				adapter->rx_ring[filter->action]->reg_idx);
4021 	}
4022 
4023 	spin_unlock(&adapter->fdir_perfect_lock);
4024 }
4025 
4026 static void ixgbe_configure(struct ixgbe_adapter *adapter)
4027 {
4028 	struct ixgbe_hw *hw = &adapter->hw;
4029 
4030 	ixgbe_configure_pb(adapter);
4031 #ifdef CONFIG_IXGBE_DCB
4032 	ixgbe_configure_dcb(adapter);
4033 #endif
4034 	/*
4035 	 * We must restore virtualization before VLANs or else
4036 	 * the VLVF registers will not be populated
4037 	 */
4038 	ixgbe_configure_virtualization(adapter);
4039 
4040 	ixgbe_set_rx_mode(adapter->netdev);
4041 	ixgbe_restore_vlan(adapter);
4042 
4043 	switch (hw->mac.type) {
4044 	case ixgbe_mac_82599EB:
4045 	case ixgbe_mac_X540:
4046 		hw->mac.ops.disable_rx_buff(hw);
4047 		break;
4048 	default:
4049 		break;
4050 	}
4051 
4052 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
4053 		ixgbe_init_fdir_signature_82599(&adapter->hw,
4054 						adapter->fdir_pballoc);
4055 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
4056 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
4057 					      adapter->fdir_pballoc);
4058 		ixgbe_fdir_filter_restore(adapter);
4059 	}
4060 
4061 	switch (hw->mac.type) {
4062 	case ixgbe_mac_82599EB:
4063 	case ixgbe_mac_X540:
4064 		hw->mac.ops.enable_rx_buff(hw);
4065 		break;
4066 	default:
4067 		break;
4068 	}
4069 
4070 #ifdef IXGBE_FCOE
4071 	/* configure FCoE L2 filters, redirection table, and Rx control */
4072 	ixgbe_configure_fcoe(adapter);
4073 
4074 #endif /* IXGBE_FCOE */
4075 	ixgbe_configure_tx(adapter);
4076 	ixgbe_configure_rx(adapter);
4077 }
4078 
4079 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
4080 {
4081 	switch (hw->phy.type) {
4082 	case ixgbe_phy_sfp_avago:
4083 	case ixgbe_phy_sfp_ftl:
4084 	case ixgbe_phy_sfp_intel:
4085 	case ixgbe_phy_sfp_unknown:
4086 	case ixgbe_phy_sfp_passive_tyco:
4087 	case ixgbe_phy_sfp_passive_unknown:
4088 	case ixgbe_phy_sfp_active_unknown:
4089 	case ixgbe_phy_sfp_ftl_active:
4090 		return true;
4091 	case ixgbe_phy_nl:
4092 		if (hw->mac.type == ixgbe_mac_82598EB)
4093 			return true;
4094 	default:
4095 		return false;
4096 	}
4097 }
4098 
4099 /**
4100  * ixgbe_sfp_link_config - set up SFP+ link
4101  * @adapter: pointer to private adapter struct
4102  **/
4103 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
4104 {
4105 	/*
4106 	 * We are assuming the worst case scenario here, and that
4107 	 * is that an SFP was inserted/removed after the reset
4108 	 * but before SFP detection was enabled.  As such the best
4109 	 * solution is to just start searching as soon as we start
4110 	 */
4111 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
4112 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
4113 
4114 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
4115 }
4116 
4117 /**
4118  * ixgbe_non_sfp_link_config - set up non-SFP+ link
4119  * @hw: pointer to private hardware struct
4120  *
4121  * Returns 0 on success, negative on failure
4122  **/
4123 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
4124 {
4125 	u32 speed;
4126 	bool autoneg, link_up = false;
4127 	u32 ret = IXGBE_ERR_LINK_SETUP;
4128 
4129 	if (hw->mac.ops.check_link)
4130 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
4131 
4132 	if (ret)
4133 		goto link_cfg_out;
4134 
4135 	speed = hw->phy.autoneg_advertised;
4136 	if ((!speed) && (hw->mac.ops.get_link_capabilities))
4137 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
4138 							&autoneg);
4139 	if (ret)
4140 		goto link_cfg_out;
4141 
4142 	if (hw->mac.ops.setup_link)
4143 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
4144 link_cfg_out:
4145 	return ret;
4146 }
4147 
4148 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
4149 {
4150 	struct ixgbe_hw *hw = &adapter->hw;
4151 	u32 gpie = 0;
4152 
4153 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
4154 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
4155 		       IXGBE_GPIE_OCD;
4156 		gpie |= IXGBE_GPIE_EIAME;
4157 		/*
4158 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
4159 		 * this saves a register write for every interrupt
4160 		 */
4161 		switch (hw->mac.type) {
4162 		case ixgbe_mac_82598EB:
4163 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4164 			break;
4165 		case ixgbe_mac_82599EB:
4166 		case ixgbe_mac_X540:
4167 		default:
4168 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
4169 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
4170 			break;
4171 		}
4172 	} else {
4173 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
4174 		 * specifically only auto mask tx and rx interrupts */
4175 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4176 	}
4177 
4178 	/* XXX: to interrupt immediately for EICS writes, enable this */
4179 	/* gpie |= IXGBE_GPIE_EIMEN; */
4180 
4181 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
4182 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
4183 
4184 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4185 		case IXGBE_82599_VMDQ_8Q_MASK:
4186 			gpie |= IXGBE_GPIE_VTMODE_16;
4187 			break;
4188 		case IXGBE_82599_VMDQ_4Q_MASK:
4189 			gpie |= IXGBE_GPIE_VTMODE_32;
4190 			break;
4191 		default:
4192 			gpie |= IXGBE_GPIE_VTMODE_64;
4193 			break;
4194 		}
4195 	}
4196 
4197 	/* Enable Thermal over heat sensor interrupt */
4198 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
4199 		switch (adapter->hw.mac.type) {
4200 		case ixgbe_mac_82599EB:
4201 			gpie |= IXGBE_SDP0_GPIEN;
4202 			break;
4203 		case ixgbe_mac_X540:
4204 			gpie |= IXGBE_EIMS_TS;
4205 			break;
4206 		default:
4207 			break;
4208 		}
4209 	}
4210 
4211 	/* Enable fan failure interrupt */
4212 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
4213 		gpie |= IXGBE_SDP1_GPIEN;
4214 
4215 	if (hw->mac.type == ixgbe_mac_82599EB) {
4216 		gpie |= IXGBE_SDP1_GPIEN;
4217 		gpie |= IXGBE_SDP2_GPIEN;
4218 	}
4219 
4220 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
4221 }
4222 
4223 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
4224 {
4225 	struct ixgbe_hw *hw = &adapter->hw;
4226 	int err;
4227 	u32 ctrl_ext;
4228 
4229 	ixgbe_get_hw_control(adapter);
4230 	ixgbe_setup_gpie(adapter);
4231 
4232 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
4233 		ixgbe_configure_msix(adapter);
4234 	else
4235 		ixgbe_configure_msi_and_legacy(adapter);
4236 
4237 	/* enable the optics for 82599 SFP+ fiber */
4238 	if (hw->mac.ops.enable_tx_laser)
4239 		hw->mac.ops.enable_tx_laser(hw);
4240 
4241 	clear_bit(__IXGBE_DOWN, &adapter->state);
4242 	ixgbe_napi_enable_all(adapter);
4243 
4244 	if (ixgbe_is_sfp(hw)) {
4245 		ixgbe_sfp_link_config(adapter);
4246 	} else {
4247 		err = ixgbe_non_sfp_link_config(hw);
4248 		if (err)
4249 			e_err(probe, "link_config FAILED %d\n", err);
4250 	}
4251 
4252 	/* clear any pending interrupts, may auto mask */
4253 	IXGBE_READ_REG(hw, IXGBE_EICR);
4254 	ixgbe_irq_enable(adapter, true, true);
4255 
4256 	/*
4257 	 * If this adapter has a fan, check to see if we had a failure
4258 	 * before we enabled the interrupt.
4259 	 */
4260 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
4261 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
4262 		if (esdp & IXGBE_ESDP_SDP1)
4263 			e_crit(drv, "Fan has stopped, replace the adapter\n");
4264 	}
4265 
4266 	/* enable transmits */
4267 	netif_tx_start_all_queues(adapter->netdev);
4268 
4269 	/* bring the link up in the watchdog, this could race with our first
4270 	 * link up interrupt but shouldn't be a problem */
4271 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4272 	adapter->link_check_timeout = jiffies;
4273 	mod_timer(&adapter->service_timer, jiffies);
4274 
4275 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
4276 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
4277 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
4278 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
4279 }
4280 
4281 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
4282 {
4283 	WARN_ON(in_interrupt());
4284 	/* put off any impending NetWatchDogTimeout */
4285 	adapter->netdev->trans_start = jiffies;
4286 
4287 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
4288 		usleep_range(1000, 2000);
4289 	ixgbe_down(adapter);
4290 	/*
4291 	 * If SR-IOV enabled then wait a bit before bringing the adapter
4292 	 * back up to give the VFs time to respond to the reset.  The
4293 	 * two second wait is based upon the watchdog timer cycle in
4294 	 * the VF driver.
4295 	 */
4296 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4297 		msleep(2000);
4298 	ixgbe_up(adapter);
4299 	clear_bit(__IXGBE_RESETTING, &adapter->state);
4300 }
4301 
4302 void ixgbe_up(struct ixgbe_adapter *adapter)
4303 {
4304 	/* hardware has been reset, we need to reload some things */
4305 	ixgbe_configure(adapter);
4306 
4307 	ixgbe_up_complete(adapter);
4308 }
4309 
4310 void ixgbe_reset(struct ixgbe_adapter *adapter)
4311 {
4312 	struct ixgbe_hw *hw = &adapter->hw;
4313 	int err;
4314 
4315 	/* lock SFP init bit to prevent race conditions with the watchdog */
4316 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
4317 		usleep_range(1000, 2000);
4318 
4319 	/* clear all SFP and link config related flags while holding SFP_INIT */
4320 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
4321 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
4322 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4323 
4324 	err = hw->mac.ops.init_hw(hw);
4325 	switch (err) {
4326 	case 0:
4327 	case IXGBE_ERR_SFP_NOT_PRESENT:
4328 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
4329 		break;
4330 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
4331 		e_dev_err("master disable timed out\n");
4332 		break;
4333 	case IXGBE_ERR_EEPROM_VERSION:
4334 		/* We are running on a pre-production device, log a warning */
4335 		e_dev_warn("This device is a pre-production adapter/LOM. "
4336 			   "Please be aware there may be issues associated with "
4337 			   "your hardware.  If you are experiencing problems "
4338 			   "please contact your Intel or hardware "
4339 			   "representative who provided you with this "
4340 			   "hardware.\n");
4341 		break;
4342 	default:
4343 		e_dev_err("Hardware Error: %d\n", err);
4344 	}
4345 
4346 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
4347 
4348 	/* reprogram the RAR[0] in case user changed it. */
4349 	hw->mac.ops.set_rar(hw, 0, hw->mac.addr, VMDQ_P(0), IXGBE_RAH_AV);
4350 
4351 	/* update SAN MAC vmdq pool selection */
4352 	if (hw->mac.san_mac_rar_index)
4353 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
4354 
4355 	if (adapter->flags2 & IXGBE_FLAG2_PTP_ENABLED)
4356 		ixgbe_ptp_reset(adapter);
4357 }
4358 
4359 /**
4360  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4361  * @rx_ring: ring to free buffers from
4362  **/
4363 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
4364 {
4365 	struct device *dev = rx_ring->dev;
4366 	unsigned long size;
4367 	u16 i;
4368 
4369 	/* ring already cleared, nothing to do */
4370 	if (!rx_ring->rx_buffer_info)
4371 		return;
4372 
4373 	/* Free all the Rx ring sk_buffs */
4374 	for (i = 0; i < rx_ring->count; i++) {
4375 		struct ixgbe_rx_buffer *rx_buffer;
4376 
4377 		rx_buffer = &rx_ring->rx_buffer_info[i];
4378 		if (rx_buffer->skb) {
4379 			struct sk_buff *skb = rx_buffer->skb;
4380 			if (IXGBE_CB(skb)->page_released) {
4381 				dma_unmap_page(dev,
4382 					       IXGBE_CB(skb)->dma,
4383 					       ixgbe_rx_bufsz(rx_ring),
4384 					       DMA_FROM_DEVICE);
4385 				IXGBE_CB(skb)->page_released = false;
4386 			}
4387 			dev_kfree_skb(skb);
4388 		}
4389 		rx_buffer->skb = NULL;
4390 		if (rx_buffer->dma)
4391 			dma_unmap_page(dev, rx_buffer->dma,
4392 				       ixgbe_rx_pg_size(rx_ring),
4393 				       DMA_FROM_DEVICE);
4394 		rx_buffer->dma = 0;
4395 		if (rx_buffer->page)
4396 			__free_pages(rx_buffer->page,
4397 				     ixgbe_rx_pg_order(rx_ring));
4398 		rx_buffer->page = NULL;
4399 	}
4400 
4401 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4402 	memset(rx_ring->rx_buffer_info, 0, size);
4403 
4404 	/* Zero out the descriptor ring */
4405 	memset(rx_ring->desc, 0, rx_ring->size);
4406 
4407 	rx_ring->next_to_alloc = 0;
4408 	rx_ring->next_to_clean = 0;
4409 	rx_ring->next_to_use = 0;
4410 }
4411 
4412 /**
4413  * ixgbe_clean_tx_ring - Free Tx Buffers
4414  * @tx_ring: ring to be cleaned
4415  **/
4416 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
4417 {
4418 	struct ixgbe_tx_buffer *tx_buffer_info;
4419 	unsigned long size;
4420 	u16 i;
4421 
4422 	/* ring already cleared, nothing to do */
4423 	if (!tx_ring->tx_buffer_info)
4424 		return;
4425 
4426 	/* Free all the Tx ring sk_buffs */
4427 	for (i = 0; i < tx_ring->count; i++) {
4428 		tx_buffer_info = &tx_ring->tx_buffer_info[i];
4429 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer_info);
4430 	}
4431 
4432 	netdev_tx_reset_queue(txring_txq(tx_ring));
4433 
4434 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
4435 	memset(tx_ring->tx_buffer_info, 0, size);
4436 
4437 	/* Zero out the descriptor ring */
4438 	memset(tx_ring->desc, 0, tx_ring->size);
4439 
4440 	tx_ring->next_to_use = 0;
4441 	tx_ring->next_to_clean = 0;
4442 }
4443 
4444 /**
4445  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
4446  * @adapter: board private structure
4447  **/
4448 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
4449 {
4450 	int i;
4451 
4452 	for (i = 0; i < adapter->num_rx_queues; i++)
4453 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
4454 }
4455 
4456 /**
4457  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
4458  * @adapter: board private structure
4459  **/
4460 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
4461 {
4462 	int i;
4463 
4464 	for (i = 0; i < adapter->num_tx_queues; i++)
4465 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
4466 }
4467 
4468 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
4469 {
4470 	struct hlist_node *node2;
4471 	struct ixgbe_fdir_filter *filter;
4472 
4473 	spin_lock(&adapter->fdir_perfect_lock);
4474 
4475 	hlist_for_each_entry_safe(filter, node2,
4476 				  &adapter->fdir_filter_list, fdir_node) {
4477 		hlist_del(&filter->fdir_node);
4478 		kfree(filter);
4479 	}
4480 	adapter->fdir_filter_count = 0;
4481 
4482 	spin_unlock(&adapter->fdir_perfect_lock);
4483 }
4484 
4485 void ixgbe_down(struct ixgbe_adapter *adapter)
4486 {
4487 	struct net_device *netdev = adapter->netdev;
4488 	struct ixgbe_hw *hw = &adapter->hw;
4489 	u32 rxctrl;
4490 	int i;
4491 
4492 	/* signal that we are down to the interrupt handler */
4493 	set_bit(__IXGBE_DOWN, &adapter->state);
4494 
4495 	/* disable receives */
4496 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4497 	IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxctrl & ~IXGBE_RXCTRL_RXEN);
4498 
4499 	/* disable all enabled rx queues */
4500 	for (i = 0; i < adapter->num_rx_queues; i++)
4501 		/* this call also flushes the previous write */
4502 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
4503 
4504 	usleep_range(10000, 20000);
4505 
4506 	netif_tx_stop_all_queues(netdev);
4507 
4508 	/* call carrier off first to avoid false dev_watchdog timeouts */
4509 	netif_carrier_off(netdev);
4510 	netif_tx_disable(netdev);
4511 
4512 	ixgbe_irq_disable(adapter);
4513 
4514 	ixgbe_napi_disable_all(adapter);
4515 
4516 	adapter->flags2 &= ~(IXGBE_FLAG2_FDIR_REQUIRES_REINIT |
4517 			     IXGBE_FLAG2_RESET_REQUESTED);
4518 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4519 
4520 	del_timer_sync(&adapter->service_timer);
4521 
4522 	if (adapter->num_vfs) {
4523 		/* Clear EITR Select mapping */
4524 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
4525 
4526 		/* Mark all the VFs as inactive */
4527 		for (i = 0 ; i < adapter->num_vfs; i++)
4528 			adapter->vfinfo[i].clear_to_send = false;
4529 
4530 		/* ping all the active vfs to let them know we are going down */
4531 		ixgbe_ping_all_vfs(adapter);
4532 
4533 		/* Disable all VFTE/VFRE TX/RX */
4534 		ixgbe_disable_tx_rx(adapter);
4535 	}
4536 
4537 	/* disable transmits in the hardware now that interrupts are off */
4538 	for (i = 0; i < adapter->num_tx_queues; i++) {
4539 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
4540 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
4541 	}
4542 
4543 	/* Disable the Tx DMA engine on 82599 and X540 */
4544 	switch (hw->mac.type) {
4545 	case ixgbe_mac_82599EB:
4546 	case ixgbe_mac_X540:
4547 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
4548 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
4549 				 ~IXGBE_DMATXCTL_TE));
4550 		break;
4551 	default:
4552 		break;
4553 	}
4554 
4555 	if (!pci_channel_offline(adapter->pdev))
4556 		ixgbe_reset(adapter);
4557 
4558 	/* power down the optics for 82599 SFP+ fiber */
4559 	if (hw->mac.ops.disable_tx_laser)
4560 		hw->mac.ops.disable_tx_laser(hw);
4561 
4562 	ixgbe_clean_all_tx_rings(adapter);
4563 	ixgbe_clean_all_rx_rings(adapter);
4564 
4565 #ifdef CONFIG_IXGBE_DCA
4566 	/* since we reset the hardware DCA settings were cleared */
4567 	ixgbe_setup_dca(adapter);
4568 #endif
4569 }
4570 
4571 /**
4572  * ixgbe_tx_timeout - Respond to a Tx Hang
4573  * @netdev: network interface device structure
4574  **/
4575 static void ixgbe_tx_timeout(struct net_device *netdev)
4576 {
4577 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4578 
4579 	/* Do the reset outside of interrupt context */
4580 	ixgbe_tx_timeout_reset(adapter);
4581 }
4582 
4583 /**
4584  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
4585  * @adapter: board private structure to initialize
4586  *
4587  * ixgbe_sw_init initializes the Adapter private data structure.
4588  * Fields are initialized based on PCI device information and
4589  * OS network device settings (MTU size).
4590  **/
4591 static int ixgbe_sw_init(struct ixgbe_adapter *adapter)
4592 {
4593 	struct ixgbe_hw *hw = &adapter->hw;
4594 	struct pci_dev *pdev = adapter->pdev;
4595 	unsigned int rss, fdir;
4596 	u32 fwsm;
4597 #ifdef CONFIG_IXGBE_DCB
4598 	int j;
4599 	struct tc_configuration *tc;
4600 #endif
4601 
4602 	/* PCI config space info */
4603 
4604 	hw->vendor_id = pdev->vendor;
4605 	hw->device_id = pdev->device;
4606 	hw->revision_id = pdev->revision;
4607 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
4608 	hw->subsystem_device_id = pdev->subsystem_device;
4609 
4610 	/* Set common capability flags and settings */
4611 	rss = min_t(int, IXGBE_MAX_RSS_INDICES, num_online_cpus());
4612 	adapter->ring_feature[RING_F_RSS].limit = rss;
4613 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
4614 	adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
4615 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
4616 	adapter->atr_sample_rate = 20;
4617 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
4618 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
4619 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
4620 #ifdef CONFIG_IXGBE_DCA
4621 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
4622 #endif
4623 #ifdef IXGBE_FCOE
4624 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
4625 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
4626 #ifdef CONFIG_IXGBE_DCB
4627 	/* Default traffic class to use for FCoE */
4628 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
4629 #endif /* CONFIG_IXGBE_DCB */
4630 #endif /* IXGBE_FCOE */
4631 
4632 	/* Set MAC specific capability flags and exceptions */
4633 	switch (hw->mac.type) {
4634 	case ixgbe_mac_82598EB:
4635 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
4636 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
4637 
4638 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
4639 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
4640 
4641 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
4642 		adapter->ring_feature[RING_F_FDIR].limit = 0;
4643 		adapter->atr_sample_rate = 0;
4644 		adapter->fdir_pballoc = 0;
4645 #ifdef IXGBE_FCOE
4646 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
4647 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
4648 #ifdef CONFIG_IXGBE_DCB
4649 		adapter->fcoe.up = 0;
4650 #endif /* IXGBE_DCB */
4651 #endif /* IXGBE_FCOE */
4652 		break;
4653 	case ixgbe_mac_82599EB:
4654 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
4655 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
4656 		break;
4657 	case ixgbe_mac_X540:
4658 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM);
4659 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
4660 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
4661 		break;
4662 	default:
4663 		break;
4664 	}
4665 
4666 #ifdef IXGBE_FCOE
4667 	/* FCoE support exists, always init the FCoE lock */
4668 	spin_lock_init(&adapter->fcoe.lock);
4669 
4670 #endif
4671 	/* n-tuple support exists, always init our spinlock */
4672 	spin_lock_init(&adapter->fdir_perfect_lock);
4673 
4674 #ifdef CONFIG_IXGBE_DCB
4675 	switch (hw->mac.type) {
4676 	case ixgbe_mac_X540:
4677 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
4678 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
4679 		break;
4680 	default:
4681 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
4682 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
4683 		break;
4684 	}
4685 
4686 	/* Configure DCB traffic classes */
4687 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
4688 		tc = &adapter->dcb_cfg.tc_config[j];
4689 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
4690 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
4691 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
4692 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
4693 		tc->dcb_pfc = pfc_disabled;
4694 	}
4695 
4696 	/* Initialize default user to priority mapping, UPx->TC0 */
4697 	tc = &adapter->dcb_cfg.tc_config[0];
4698 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
4699 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
4700 
4701 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
4702 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
4703 	adapter->dcb_cfg.pfc_mode_enable = false;
4704 	adapter->dcb_set_bitmap = 0x00;
4705 	adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
4706 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
4707 	       sizeof(adapter->temp_dcb_cfg));
4708 
4709 #endif
4710 
4711 	/* default flow control settings */
4712 	hw->fc.requested_mode = ixgbe_fc_full;
4713 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
4714 	ixgbe_pbthresh_setup(adapter);
4715 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
4716 	hw->fc.send_xon = true;
4717 	hw->fc.disable_fc_autoneg =
4718 		(ixgbe_device_supports_autoneg_fc(hw) == 0) ? false : true;
4719 
4720 #ifdef CONFIG_PCI_IOV
4721 	/* assign number of SR-IOV VFs */
4722 	if (hw->mac.type != ixgbe_mac_82598EB)
4723 		adapter->num_vfs = (max_vfs > 63) ? 0 : max_vfs;
4724 
4725 #endif
4726 	/* enable itr by default in dynamic mode */
4727 	adapter->rx_itr_setting = 1;
4728 	adapter->tx_itr_setting = 1;
4729 
4730 	/* set default ring sizes */
4731 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
4732 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
4733 
4734 	/* set default work limits */
4735 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
4736 
4737 	/* initialize eeprom parameters */
4738 	if (ixgbe_init_eeprom_params_generic(hw)) {
4739 		e_dev_err("EEPROM initialization failed\n");
4740 		return -EIO;
4741 	}
4742 
4743 	set_bit(__IXGBE_DOWN, &adapter->state);
4744 
4745 	return 0;
4746 }
4747 
4748 /**
4749  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
4750  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
4751  *
4752  * Return 0 on success, negative on failure
4753  **/
4754 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
4755 {
4756 	struct device *dev = tx_ring->dev;
4757 	int orig_node = dev_to_node(dev);
4758 	int numa_node = -1;
4759 	int size;
4760 
4761 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
4762 
4763 	if (tx_ring->q_vector)
4764 		numa_node = tx_ring->q_vector->numa_node;
4765 
4766 	tx_ring->tx_buffer_info = vzalloc_node(size, numa_node);
4767 	if (!tx_ring->tx_buffer_info)
4768 		tx_ring->tx_buffer_info = vzalloc(size);
4769 	if (!tx_ring->tx_buffer_info)
4770 		goto err;
4771 
4772 	/* round up to nearest 4K */
4773 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
4774 	tx_ring->size = ALIGN(tx_ring->size, 4096);
4775 
4776 	set_dev_node(dev, numa_node);
4777 	tx_ring->desc = dma_alloc_coherent(dev,
4778 					   tx_ring->size,
4779 					   &tx_ring->dma,
4780 					   GFP_KERNEL);
4781 	set_dev_node(dev, orig_node);
4782 	if (!tx_ring->desc)
4783 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
4784 						   &tx_ring->dma, GFP_KERNEL);
4785 	if (!tx_ring->desc)
4786 		goto err;
4787 
4788 	tx_ring->next_to_use = 0;
4789 	tx_ring->next_to_clean = 0;
4790 	return 0;
4791 
4792 err:
4793 	vfree(tx_ring->tx_buffer_info);
4794 	tx_ring->tx_buffer_info = NULL;
4795 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
4796 	return -ENOMEM;
4797 }
4798 
4799 /**
4800  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
4801  * @adapter: board private structure
4802  *
4803  * If this function returns with an error, then it's possible one or
4804  * more of the rings is populated (while the rest are not).  It is the
4805  * callers duty to clean those orphaned rings.
4806  *
4807  * Return 0 on success, negative on failure
4808  **/
4809 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
4810 {
4811 	int i, err = 0;
4812 
4813 	for (i = 0; i < adapter->num_tx_queues; i++) {
4814 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
4815 		if (!err)
4816 			continue;
4817 
4818 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
4819 		goto err_setup_tx;
4820 	}
4821 
4822 	return 0;
4823 err_setup_tx:
4824 	/* rewind the index freeing the rings as we go */
4825 	while (i--)
4826 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
4827 	return err;
4828 }
4829 
4830 /**
4831  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
4832  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
4833  *
4834  * Returns 0 on success, negative on failure
4835  **/
4836 int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
4837 {
4838 	struct device *dev = rx_ring->dev;
4839 	int orig_node = dev_to_node(dev);
4840 	int numa_node = -1;
4841 	int size;
4842 
4843 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4844 
4845 	if (rx_ring->q_vector)
4846 		numa_node = rx_ring->q_vector->numa_node;
4847 
4848 	rx_ring->rx_buffer_info = vzalloc_node(size, numa_node);
4849 	if (!rx_ring->rx_buffer_info)
4850 		rx_ring->rx_buffer_info = vzalloc(size);
4851 	if (!rx_ring->rx_buffer_info)
4852 		goto err;
4853 
4854 	/* Round up to nearest 4K */
4855 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
4856 	rx_ring->size = ALIGN(rx_ring->size, 4096);
4857 
4858 	set_dev_node(dev, numa_node);
4859 	rx_ring->desc = dma_alloc_coherent(dev,
4860 					   rx_ring->size,
4861 					   &rx_ring->dma,
4862 					   GFP_KERNEL);
4863 	set_dev_node(dev, orig_node);
4864 	if (!rx_ring->desc)
4865 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
4866 						   &rx_ring->dma, GFP_KERNEL);
4867 	if (!rx_ring->desc)
4868 		goto err;
4869 
4870 	rx_ring->next_to_clean = 0;
4871 	rx_ring->next_to_use = 0;
4872 
4873 	return 0;
4874 err:
4875 	vfree(rx_ring->rx_buffer_info);
4876 	rx_ring->rx_buffer_info = NULL;
4877 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
4878 	return -ENOMEM;
4879 }
4880 
4881 /**
4882  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
4883  * @adapter: board private structure
4884  *
4885  * If this function returns with an error, then it's possible one or
4886  * more of the rings is populated (while the rest are not).  It is the
4887  * callers duty to clean those orphaned rings.
4888  *
4889  * Return 0 on success, negative on failure
4890  **/
4891 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
4892 {
4893 	int i, err = 0;
4894 
4895 	for (i = 0; i < adapter->num_rx_queues; i++) {
4896 		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
4897 		if (!err)
4898 			continue;
4899 
4900 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
4901 		goto err_setup_rx;
4902 	}
4903 
4904 #ifdef IXGBE_FCOE
4905 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
4906 	if (!err)
4907 #endif
4908 		return 0;
4909 err_setup_rx:
4910 	/* rewind the index freeing the rings as we go */
4911 	while (i--)
4912 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
4913 	return err;
4914 }
4915 
4916 /**
4917  * ixgbe_free_tx_resources - Free Tx Resources per Queue
4918  * @tx_ring: Tx descriptor ring for a specific queue
4919  *
4920  * Free all transmit software resources
4921  **/
4922 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
4923 {
4924 	ixgbe_clean_tx_ring(tx_ring);
4925 
4926 	vfree(tx_ring->tx_buffer_info);
4927 	tx_ring->tx_buffer_info = NULL;
4928 
4929 	/* if not set, then don't free */
4930 	if (!tx_ring->desc)
4931 		return;
4932 
4933 	dma_free_coherent(tx_ring->dev, tx_ring->size,
4934 			  tx_ring->desc, tx_ring->dma);
4935 
4936 	tx_ring->desc = NULL;
4937 }
4938 
4939 /**
4940  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
4941  * @adapter: board private structure
4942  *
4943  * Free all transmit software resources
4944  **/
4945 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
4946 {
4947 	int i;
4948 
4949 	for (i = 0; i < adapter->num_tx_queues; i++)
4950 		if (adapter->tx_ring[i]->desc)
4951 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
4952 }
4953 
4954 /**
4955  * ixgbe_free_rx_resources - Free Rx Resources
4956  * @rx_ring: ring to clean the resources from
4957  *
4958  * Free all receive software resources
4959  **/
4960 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
4961 {
4962 	ixgbe_clean_rx_ring(rx_ring);
4963 
4964 	vfree(rx_ring->rx_buffer_info);
4965 	rx_ring->rx_buffer_info = NULL;
4966 
4967 	/* if not set, then don't free */
4968 	if (!rx_ring->desc)
4969 		return;
4970 
4971 	dma_free_coherent(rx_ring->dev, rx_ring->size,
4972 			  rx_ring->desc, rx_ring->dma);
4973 
4974 	rx_ring->desc = NULL;
4975 }
4976 
4977 /**
4978  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
4979  * @adapter: board private structure
4980  *
4981  * Free all receive software resources
4982  **/
4983 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
4984 {
4985 	int i;
4986 
4987 #ifdef IXGBE_FCOE
4988 	ixgbe_free_fcoe_ddp_resources(adapter);
4989 
4990 #endif
4991 	for (i = 0; i < adapter->num_rx_queues; i++)
4992 		if (adapter->rx_ring[i]->desc)
4993 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
4994 }
4995 
4996 /**
4997  * ixgbe_change_mtu - Change the Maximum Transfer Unit
4998  * @netdev: network interface device structure
4999  * @new_mtu: new value for maximum frame size
5000  *
5001  * Returns 0 on success, negative on failure
5002  **/
5003 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
5004 {
5005 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5006 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
5007 
5008 	/* MTU < 68 is an error and causes problems on some kernels */
5009 	if ((new_mtu < 68) || (max_frame > IXGBE_MAX_JUMBO_FRAME_SIZE))
5010 		return -EINVAL;
5011 
5012 	/*
5013 	 * For 82599EB we cannot allow legacy VFs to enable their receive
5014 	 * paths when MTU greater than 1500 is configured.  So display a
5015 	 * warning that legacy VFs will be disabled.
5016 	 */
5017 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
5018 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
5019 	    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
5020 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
5021 
5022 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
5023 
5024 	/* must set new MTU before calling down or up */
5025 	netdev->mtu = new_mtu;
5026 
5027 	if (netif_running(netdev))
5028 		ixgbe_reinit_locked(adapter);
5029 
5030 	return 0;
5031 }
5032 
5033 /**
5034  * ixgbe_open - Called when a network interface is made active
5035  * @netdev: network interface device structure
5036  *
5037  * Returns 0 on success, negative value on failure
5038  *
5039  * The open entry point is called when a network interface is made
5040  * active by the system (IFF_UP).  At this point all resources needed
5041  * for transmit and receive operations are allocated, the interrupt
5042  * handler is registered with the OS, the watchdog timer is started,
5043  * and the stack is notified that the interface is ready.
5044  **/
5045 static int ixgbe_open(struct net_device *netdev)
5046 {
5047 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5048 	int err;
5049 
5050 	/* disallow open during test */
5051 	if (test_bit(__IXGBE_TESTING, &adapter->state))
5052 		return -EBUSY;
5053 
5054 	netif_carrier_off(netdev);
5055 
5056 	/* allocate transmit descriptors */
5057 	err = ixgbe_setup_all_tx_resources(adapter);
5058 	if (err)
5059 		goto err_setup_tx;
5060 
5061 	/* allocate receive descriptors */
5062 	err = ixgbe_setup_all_rx_resources(adapter);
5063 	if (err)
5064 		goto err_setup_rx;
5065 
5066 	ixgbe_configure(adapter);
5067 
5068 	err = ixgbe_request_irq(adapter);
5069 	if (err)
5070 		goto err_req_irq;
5071 
5072 	/* Notify the stack of the actual queue counts. */
5073 	err = netif_set_real_num_tx_queues(netdev,
5074 					   adapter->num_rx_pools > 1 ? 1 :
5075 					   adapter->num_tx_queues);
5076 	if (err)
5077 		goto err_set_queues;
5078 
5079 
5080 	err = netif_set_real_num_rx_queues(netdev,
5081 					   adapter->num_rx_pools > 1 ? 1 :
5082 					   adapter->num_rx_queues);
5083 	if (err)
5084 		goto err_set_queues;
5085 
5086 	ixgbe_ptp_init(adapter);
5087 
5088 	ixgbe_up_complete(adapter);
5089 
5090 	return 0;
5091 
5092 err_set_queues:
5093 	ixgbe_free_irq(adapter);
5094 err_req_irq:
5095 	ixgbe_free_all_rx_resources(adapter);
5096 err_setup_rx:
5097 	ixgbe_free_all_tx_resources(adapter);
5098 err_setup_tx:
5099 	ixgbe_reset(adapter);
5100 
5101 	return err;
5102 }
5103 
5104 /**
5105  * ixgbe_close - Disables a network interface
5106  * @netdev: network interface device structure
5107  *
5108  * Returns 0, this is not allowed to fail
5109  *
5110  * The close entry point is called when an interface is de-activated
5111  * by the OS.  The hardware is still under the drivers control, but
5112  * needs to be disabled.  A global MAC reset is issued to stop the
5113  * hardware, and all transmit and receive resources are freed.
5114  **/
5115 static int ixgbe_close(struct net_device *netdev)
5116 {
5117 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5118 
5119 	ixgbe_ptp_stop(adapter);
5120 
5121 	ixgbe_down(adapter);
5122 	ixgbe_free_irq(adapter);
5123 
5124 	ixgbe_fdir_filter_exit(adapter);
5125 
5126 	ixgbe_free_all_tx_resources(adapter);
5127 	ixgbe_free_all_rx_resources(adapter);
5128 
5129 	ixgbe_release_hw_control(adapter);
5130 
5131 	return 0;
5132 }
5133 
5134 #ifdef CONFIG_PM
5135 static int ixgbe_resume(struct pci_dev *pdev)
5136 {
5137 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
5138 	struct net_device *netdev = adapter->netdev;
5139 	u32 err;
5140 
5141 	pci_set_power_state(pdev, PCI_D0);
5142 	pci_restore_state(pdev);
5143 	/*
5144 	 * pci_restore_state clears dev->state_saved so call
5145 	 * pci_save_state to restore it.
5146 	 */
5147 	pci_save_state(pdev);
5148 
5149 	err = pci_enable_device_mem(pdev);
5150 	if (err) {
5151 		e_dev_err("Cannot enable PCI device from suspend\n");
5152 		return err;
5153 	}
5154 	pci_set_master(pdev);
5155 
5156 	pci_wake_from_d3(pdev, false);
5157 
5158 	ixgbe_reset(adapter);
5159 
5160 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
5161 
5162 	rtnl_lock();
5163 	err = ixgbe_init_interrupt_scheme(adapter);
5164 	if (!err && netif_running(netdev))
5165 		err = ixgbe_open(netdev);
5166 
5167 	rtnl_unlock();
5168 
5169 	if (err)
5170 		return err;
5171 
5172 	netif_device_attach(netdev);
5173 
5174 	return 0;
5175 }
5176 #endif /* CONFIG_PM */
5177 
5178 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
5179 {
5180 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
5181 	struct net_device *netdev = adapter->netdev;
5182 	struct ixgbe_hw *hw = &adapter->hw;
5183 	u32 ctrl, fctrl;
5184 	u32 wufc = adapter->wol;
5185 #ifdef CONFIG_PM
5186 	int retval = 0;
5187 #endif
5188 
5189 	netif_device_detach(netdev);
5190 
5191 	rtnl_lock();
5192 	if (netif_running(netdev)) {
5193 		ixgbe_down(adapter);
5194 		ixgbe_free_irq(adapter);
5195 		ixgbe_free_all_tx_resources(adapter);
5196 		ixgbe_free_all_rx_resources(adapter);
5197 	}
5198 	rtnl_unlock();
5199 
5200 	ixgbe_clear_interrupt_scheme(adapter);
5201 
5202 #ifdef CONFIG_PM
5203 	retval = pci_save_state(pdev);
5204 	if (retval)
5205 		return retval;
5206 
5207 #endif
5208 	if (wufc) {
5209 		ixgbe_set_rx_mode(netdev);
5210 
5211 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
5212 		if (hw->mac.ops.enable_tx_laser)
5213 			hw->mac.ops.enable_tx_laser(hw);
5214 
5215 		/* turn on all-multi mode if wake on multicast is enabled */
5216 		if (wufc & IXGBE_WUFC_MC) {
5217 			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
5218 			fctrl |= IXGBE_FCTRL_MPE;
5219 			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
5220 		}
5221 
5222 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
5223 		ctrl |= IXGBE_CTRL_GIO_DIS;
5224 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
5225 
5226 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
5227 	} else {
5228 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
5229 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
5230 	}
5231 
5232 	switch (hw->mac.type) {
5233 	case ixgbe_mac_82598EB:
5234 		pci_wake_from_d3(pdev, false);
5235 		break;
5236 	case ixgbe_mac_82599EB:
5237 	case ixgbe_mac_X540:
5238 		pci_wake_from_d3(pdev, !!wufc);
5239 		break;
5240 	default:
5241 		break;
5242 	}
5243 
5244 	*enable_wake = !!wufc;
5245 
5246 	ixgbe_release_hw_control(adapter);
5247 
5248 	pci_disable_device(pdev);
5249 
5250 	return 0;
5251 }
5252 
5253 #ifdef CONFIG_PM
5254 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
5255 {
5256 	int retval;
5257 	bool wake;
5258 
5259 	retval = __ixgbe_shutdown(pdev, &wake);
5260 	if (retval)
5261 		return retval;
5262 
5263 	if (wake) {
5264 		pci_prepare_to_sleep(pdev);
5265 	} else {
5266 		pci_wake_from_d3(pdev, false);
5267 		pci_set_power_state(pdev, PCI_D3hot);
5268 	}
5269 
5270 	return 0;
5271 }
5272 #endif /* CONFIG_PM */
5273 
5274 static void ixgbe_shutdown(struct pci_dev *pdev)
5275 {
5276 	bool wake;
5277 
5278 	__ixgbe_shutdown(pdev, &wake);
5279 
5280 	if (system_state == SYSTEM_POWER_OFF) {
5281 		pci_wake_from_d3(pdev, wake);
5282 		pci_set_power_state(pdev, PCI_D3hot);
5283 	}
5284 }
5285 
5286 /**
5287  * ixgbe_update_stats - Update the board statistics counters.
5288  * @adapter: board private structure
5289  **/
5290 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
5291 {
5292 	struct net_device *netdev = adapter->netdev;
5293 	struct ixgbe_hw *hw = &adapter->hw;
5294 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
5295 	u64 total_mpc = 0;
5296 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
5297 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
5298 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
5299 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
5300 
5301 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5302 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5303 		return;
5304 
5305 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
5306 		u64 rsc_count = 0;
5307 		u64 rsc_flush = 0;
5308 		for (i = 0; i < adapter->num_rx_queues; i++) {
5309 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
5310 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
5311 		}
5312 		adapter->rsc_total_count = rsc_count;
5313 		adapter->rsc_total_flush = rsc_flush;
5314 	}
5315 
5316 	for (i = 0; i < adapter->num_rx_queues; i++) {
5317 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
5318 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
5319 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
5320 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
5321 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
5322 		bytes += rx_ring->stats.bytes;
5323 		packets += rx_ring->stats.packets;
5324 	}
5325 	adapter->non_eop_descs = non_eop_descs;
5326 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
5327 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
5328 	adapter->hw_csum_rx_error = hw_csum_rx_error;
5329 	netdev->stats.rx_bytes = bytes;
5330 	netdev->stats.rx_packets = packets;
5331 
5332 	bytes = 0;
5333 	packets = 0;
5334 	/* gather some stats to the adapter struct that are per queue */
5335 	for (i = 0; i < adapter->num_tx_queues; i++) {
5336 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
5337 		restart_queue += tx_ring->tx_stats.restart_queue;
5338 		tx_busy += tx_ring->tx_stats.tx_busy;
5339 		bytes += tx_ring->stats.bytes;
5340 		packets += tx_ring->stats.packets;
5341 	}
5342 	adapter->restart_queue = restart_queue;
5343 	adapter->tx_busy = tx_busy;
5344 	netdev->stats.tx_bytes = bytes;
5345 	netdev->stats.tx_packets = packets;
5346 
5347 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
5348 
5349 	/* 8 register reads */
5350 	for (i = 0; i < 8; i++) {
5351 		/* for packet buffers not used, the register should read 0 */
5352 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
5353 		missed_rx += mpc;
5354 		hwstats->mpc[i] += mpc;
5355 		total_mpc += hwstats->mpc[i];
5356 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
5357 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
5358 		switch (hw->mac.type) {
5359 		case ixgbe_mac_82598EB:
5360 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
5361 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
5362 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
5363 			hwstats->pxonrxc[i] +=
5364 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
5365 			break;
5366 		case ixgbe_mac_82599EB:
5367 		case ixgbe_mac_X540:
5368 			hwstats->pxonrxc[i] +=
5369 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
5370 			break;
5371 		default:
5372 			break;
5373 		}
5374 	}
5375 
5376 	/*16 register reads */
5377 	for (i = 0; i < 16; i++) {
5378 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
5379 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
5380 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
5381 		    (hw->mac.type == ixgbe_mac_X540)) {
5382 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
5383 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
5384 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
5385 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
5386 		}
5387 	}
5388 
5389 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
5390 	/* work around hardware counting issue */
5391 	hwstats->gprc -= missed_rx;
5392 
5393 	ixgbe_update_xoff_received(adapter);
5394 
5395 	/* 82598 hardware only has a 32 bit counter in the high register */
5396 	switch (hw->mac.type) {
5397 	case ixgbe_mac_82598EB:
5398 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
5399 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
5400 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
5401 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
5402 		break;
5403 	case ixgbe_mac_X540:
5404 		/* OS2BMC stats are X540 only*/
5405 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
5406 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
5407 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
5408 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
5409 	case ixgbe_mac_82599EB:
5410 		for (i = 0; i < 16; i++)
5411 			adapter->hw_rx_no_dma_resources +=
5412 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
5413 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
5414 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
5415 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
5416 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
5417 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
5418 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
5419 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
5420 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
5421 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
5422 #ifdef IXGBE_FCOE
5423 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
5424 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
5425 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
5426 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
5427 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
5428 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
5429 		/* Add up per cpu counters for total ddp aloc fail */
5430 		if (adapter->fcoe.ddp_pool) {
5431 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
5432 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
5433 			unsigned int cpu;
5434 			u64 noddp = 0, noddp_ext_buff = 0;
5435 			for_each_possible_cpu(cpu) {
5436 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
5437 				noddp += ddp_pool->noddp;
5438 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
5439 			}
5440 			hwstats->fcoe_noddp = noddp;
5441 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
5442 		}
5443 #endif /* IXGBE_FCOE */
5444 		break;
5445 	default:
5446 		break;
5447 	}
5448 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
5449 	hwstats->bprc += bprc;
5450 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
5451 	if (hw->mac.type == ixgbe_mac_82598EB)
5452 		hwstats->mprc -= bprc;
5453 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
5454 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
5455 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
5456 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
5457 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
5458 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
5459 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
5460 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
5461 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
5462 	hwstats->lxontxc += lxon;
5463 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
5464 	hwstats->lxofftxc += lxoff;
5465 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
5466 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
5467 	/*
5468 	 * 82598 errata - tx of flow control packets is included in tx counters
5469 	 */
5470 	xon_off_tot = lxon + lxoff;
5471 	hwstats->gptc -= xon_off_tot;
5472 	hwstats->mptc -= xon_off_tot;
5473 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
5474 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
5475 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
5476 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
5477 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
5478 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
5479 	hwstats->ptc64 -= xon_off_tot;
5480 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
5481 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
5482 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
5483 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
5484 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
5485 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
5486 
5487 	/* Fill out the OS statistics structure */
5488 	netdev->stats.multicast = hwstats->mprc;
5489 
5490 	/* Rx Errors */
5491 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
5492 	netdev->stats.rx_dropped = 0;
5493 	netdev->stats.rx_length_errors = hwstats->rlec;
5494 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
5495 	netdev->stats.rx_missed_errors = total_mpc;
5496 }
5497 
5498 /**
5499  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
5500  * @adapter: pointer to the device adapter structure
5501  **/
5502 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
5503 {
5504 	struct ixgbe_hw *hw = &adapter->hw;
5505 	int i;
5506 
5507 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
5508 		return;
5509 
5510 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
5511 
5512 	/* if interface is down do nothing */
5513 	if (test_bit(__IXGBE_DOWN, &adapter->state))
5514 		return;
5515 
5516 	/* do nothing if we are not using signature filters */
5517 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
5518 		return;
5519 
5520 	adapter->fdir_overflow++;
5521 
5522 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
5523 		for (i = 0; i < adapter->num_tx_queues; i++)
5524 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
5525 			        &(adapter->tx_ring[i]->state));
5526 		/* re-enable flow director interrupts */
5527 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
5528 	} else {
5529 		e_err(probe, "failed to finish FDIR re-initialization, "
5530 		      "ignored adding FDIR ATR filters\n");
5531 	}
5532 }
5533 
5534 /**
5535  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
5536  * @adapter: pointer to the device adapter structure
5537  *
5538  * This function serves two purposes.  First it strobes the interrupt lines
5539  * in order to make certain interrupts are occurring.  Secondly it sets the
5540  * bits needed to check for TX hangs.  As a result we should immediately
5541  * determine if a hang has occurred.
5542  */
5543 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
5544 {
5545 	struct ixgbe_hw *hw = &adapter->hw;
5546 	u64 eics = 0;
5547 	int i;
5548 
5549 	/* If we're down or resetting, just bail */
5550 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5551 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5552 		return;
5553 
5554 	/* Force detection of hung controller */
5555 	if (netif_carrier_ok(adapter->netdev)) {
5556 		for (i = 0; i < adapter->num_tx_queues; i++)
5557 			set_check_for_tx_hang(adapter->tx_ring[i]);
5558 	}
5559 
5560 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
5561 		/*
5562 		 * for legacy and MSI interrupts don't set any bits
5563 		 * that are enabled for EIAM, because this operation
5564 		 * would set *both* EIMS and EICS for any bit in EIAM
5565 		 */
5566 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
5567 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
5568 	} else {
5569 		/* get one bit for every active tx/rx interrupt vector */
5570 		for (i = 0; i < adapter->num_q_vectors; i++) {
5571 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
5572 			if (qv->rx.ring || qv->tx.ring)
5573 				eics |= ((u64)1 << i);
5574 		}
5575 	}
5576 
5577 	/* Cause software interrupt to ensure rings are cleaned */
5578 	ixgbe_irq_rearm_queues(adapter, eics);
5579 
5580 }
5581 
5582 /**
5583  * ixgbe_watchdog_update_link - update the link status
5584  * @adapter: pointer to the device adapter structure
5585  * @link_speed: pointer to a u32 to store the link_speed
5586  **/
5587 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
5588 {
5589 	struct ixgbe_hw *hw = &adapter->hw;
5590 	u32 link_speed = adapter->link_speed;
5591 	bool link_up = adapter->link_up;
5592 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
5593 
5594 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
5595 		return;
5596 
5597 	if (hw->mac.ops.check_link) {
5598 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
5599 	} else {
5600 		/* always assume link is up, if no check link function */
5601 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
5602 		link_up = true;
5603 	}
5604 
5605 	if (adapter->ixgbe_ieee_pfc)
5606 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
5607 
5608 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
5609 		hw->mac.ops.fc_enable(hw);
5610 		ixgbe_set_rx_drop_en(adapter);
5611 	}
5612 
5613 	if (link_up ||
5614 	    time_after(jiffies, (adapter->link_check_timeout +
5615 				 IXGBE_TRY_LINK_TIMEOUT))) {
5616 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5617 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
5618 		IXGBE_WRITE_FLUSH(hw);
5619 	}
5620 
5621 	adapter->link_up = link_up;
5622 	adapter->link_speed = link_speed;
5623 }
5624 
5625 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
5626 {
5627 #ifdef CONFIG_IXGBE_DCB
5628 	struct net_device *netdev = adapter->netdev;
5629 	struct dcb_app app = {
5630 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
5631 			      .protocol = 0,
5632 			     };
5633 	u8 up = 0;
5634 
5635 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
5636 		up = dcb_ieee_getapp_mask(netdev, &app);
5637 
5638 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
5639 #endif
5640 }
5641 
5642 /**
5643  * ixgbe_watchdog_link_is_up - update netif_carrier status and
5644  *                             print link up message
5645  * @adapter: pointer to the device adapter structure
5646  **/
5647 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
5648 {
5649 	struct net_device *netdev = adapter->netdev;
5650 	struct ixgbe_hw *hw = &adapter->hw;
5651 	u32 link_speed = adapter->link_speed;
5652 	bool flow_rx, flow_tx;
5653 
5654 	/* only continue if link was previously down */
5655 	if (netif_carrier_ok(netdev))
5656 		return;
5657 
5658 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
5659 
5660 	switch (hw->mac.type) {
5661 	case ixgbe_mac_82598EB: {
5662 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
5663 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
5664 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
5665 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
5666 	}
5667 		break;
5668 	case ixgbe_mac_X540:
5669 	case ixgbe_mac_82599EB: {
5670 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
5671 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
5672 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
5673 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
5674 	}
5675 		break;
5676 	default:
5677 		flow_tx = false;
5678 		flow_rx = false;
5679 		break;
5680 	}
5681 
5682 	adapter->last_rx_ptp_check = jiffies;
5683 
5684 	if (adapter->flags2 & IXGBE_FLAG2_PTP_ENABLED)
5685 		ixgbe_ptp_start_cyclecounter(adapter);
5686 
5687 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n",
5688 	       (link_speed == IXGBE_LINK_SPEED_10GB_FULL ?
5689 	       "10 Gbps" :
5690 	       (link_speed == IXGBE_LINK_SPEED_1GB_FULL ?
5691 	       "1 Gbps" :
5692 	       (link_speed == IXGBE_LINK_SPEED_100_FULL ?
5693 	       "100 Mbps" :
5694 	       "unknown speed"))),
5695 	       ((flow_rx && flow_tx) ? "RX/TX" :
5696 	       (flow_rx ? "RX" :
5697 	       (flow_tx ? "TX" : "None"))));
5698 
5699 	netif_carrier_on(netdev);
5700 	ixgbe_check_vf_rate_limit(adapter);
5701 
5702 	/* update the default user priority for VFs */
5703 	ixgbe_update_default_up(adapter);
5704 
5705 	/* ping all the active vfs to let them know link has changed */
5706 	ixgbe_ping_all_vfs(adapter);
5707 }
5708 
5709 /**
5710  * ixgbe_watchdog_link_is_down - update netif_carrier status and
5711  *                               print link down message
5712  * @adapter: pointer to the adapter structure
5713  **/
5714 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
5715 {
5716 	struct net_device *netdev = adapter->netdev;
5717 	struct ixgbe_hw *hw = &adapter->hw;
5718 
5719 	adapter->link_up = false;
5720 	adapter->link_speed = 0;
5721 
5722 	/* only continue if link was up previously */
5723 	if (!netif_carrier_ok(netdev))
5724 		return;
5725 
5726 	/* poll for SFP+ cable when link is down */
5727 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
5728 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5729 
5730 	if (adapter->flags2 & IXGBE_FLAG2_PTP_ENABLED)
5731 		ixgbe_ptp_start_cyclecounter(adapter);
5732 
5733 	e_info(drv, "NIC Link is Down\n");
5734 	netif_carrier_off(netdev);
5735 
5736 	/* ping all the active vfs to let them know link has changed */
5737 	ixgbe_ping_all_vfs(adapter);
5738 }
5739 
5740 /**
5741  * ixgbe_watchdog_flush_tx - flush queues on link down
5742  * @adapter: pointer to the device adapter structure
5743  **/
5744 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
5745 {
5746 	int i;
5747 	int some_tx_pending = 0;
5748 
5749 	if (!netif_carrier_ok(adapter->netdev)) {
5750 		for (i = 0; i < adapter->num_tx_queues; i++) {
5751 			struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
5752 			if (tx_ring->next_to_use != tx_ring->next_to_clean) {
5753 				some_tx_pending = 1;
5754 				break;
5755 			}
5756 		}
5757 
5758 		if (some_tx_pending) {
5759 			/* We've lost link, so the controller stops DMA,
5760 			 * but we've got queued Tx work that's never going
5761 			 * to get done, so reset controller to flush Tx.
5762 			 * (Do the reset outside of interrupt context).
5763 			 */
5764 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
5765 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
5766 		}
5767 	}
5768 }
5769 
5770 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
5771 {
5772 	u32 ssvpc;
5773 
5774 	/* Do not perform spoof check for 82598 or if not in IOV mode */
5775 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
5776 	    adapter->num_vfs == 0)
5777 		return;
5778 
5779 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
5780 
5781 	/*
5782 	 * ssvpc register is cleared on read, if zero then no
5783 	 * spoofed packets in the last interval.
5784 	 */
5785 	if (!ssvpc)
5786 		return;
5787 
5788 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
5789 }
5790 
5791 /**
5792  * ixgbe_watchdog_subtask - check and bring link up
5793  * @adapter: pointer to the device adapter structure
5794  **/
5795 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
5796 {
5797 	/* if interface is down do nothing */
5798 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5799 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5800 		return;
5801 
5802 	ixgbe_watchdog_update_link(adapter);
5803 
5804 	if (adapter->link_up)
5805 		ixgbe_watchdog_link_is_up(adapter);
5806 	else
5807 		ixgbe_watchdog_link_is_down(adapter);
5808 
5809 	ixgbe_spoof_check(adapter);
5810 	ixgbe_update_stats(adapter);
5811 
5812 	ixgbe_watchdog_flush_tx(adapter);
5813 }
5814 
5815 /**
5816  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
5817  * @adapter: the ixgbe adapter structure
5818  **/
5819 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
5820 {
5821 	struct ixgbe_hw *hw = &adapter->hw;
5822 	s32 err;
5823 
5824 	/* not searching for SFP so there is nothing to do here */
5825 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
5826 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
5827 		return;
5828 
5829 	/* concurent i2c reads are not supported */
5830 	if (test_bit(__IXGBE_READ_I2C, &adapter->state))
5831 		return;
5832 
5833 	/* someone else is in init, wait until next service event */
5834 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5835 		return;
5836 
5837 	err = hw->phy.ops.identify_sfp(hw);
5838 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
5839 		goto sfp_out;
5840 
5841 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
5842 		/* If no cable is present, then we need to reset
5843 		 * the next time we find a good cable. */
5844 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5845 	}
5846 
5847 	/* exit on error */
5848 	if (err)
5849 		goto sfp_out;
5850 
5851 	/* exit if reset not needed */
5852 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
5853 		goto sfp_out;
5854 
5855 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
5856 
5857 	/*
5858 	 * A module may be identified correctly, but the EEPROM may not have
5859 	 * support for that module.  setup_sfp() will fail in that case, so
5860 	 * we should not allow that module to load.
5861 	 */
5862 	if (hw->mac.type == ixgbe_mac_82598EB)
5863 		err = hw->phy.ops.reset(hw);
5864 	else
5865 		err = hw->mac.ops.setup_sfp(hw);
5866 
5867 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
5868 		goto sfp_out;
5869 
5870 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
5871 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
5872 
5873 sfp_out:
5874 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5875 
5876 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
5877 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
5878 		e_dev_err("failed to initialize because an unsupported "
5879 			  "SFP+ module type was detected.\n");
5880 		e_dev_err("Reload the driver after installing a "
5881 			  "supported module.\n");
5882 		unregister_netdev(adapter->netdev);
5883 	}
5884 }
5885 
5886 /**
5887  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
5888  * @adapter: the ixgbe adapter structure
5889  **/
5890 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
5891 {
5892 	struct ixgbe_hw *hw = &adapter->hw;
5893 	u32 speed;
5894 	bool autoneg = false;
5895 
5896 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
5897 		return;
5898 
5899 	/* someone else is in init, wait until next service event */
5900 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5901 		return;
5902 
5903 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5904 
5905 	speed = hw->phy.autoneg_advertised;
5906 	if ((!speed) && (hw->mac.ops.get_link_capabilities))
5907 		hw->mac.ops.get_link_capabilities(hw, &speed, &autoneg);
5908 	if (hw->mac.ops.setup_link)
5909 		hw->mac.ops.setup_link(hw, speed, true);
5910 
5911 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5912 	adapter->link_check_timeout = jiffies;
5913 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5914 }
5915 
5916 #ifdef CONFIG_PCI_IOV
5917 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
5918 {
5919 	int vf;
5920 	struct ixgbe_hw *hw = &adapter->hw;
5921 	struct net_device *netdev = adapter->netdev;
5922 	u32 gpc;
5923 	u32 ciaa, ciad;
5924 
5925 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
5926 	if (gpc) /* If incrementing then no need for the check below */
5927 		return;
5928 	/*
5929 	 * Check to see if a bad DMA write target from an errant or
5930 	 * malicious VF has caused a PCIe error.  If so then we can
5931 	 * issue a VFLR to the offending VF(s) and then resume without
5932 	 * requesting a full slot reset.
5933 	 */
5934 
5935 	for (vf = 0; vf < adapter->num_vfs; vf++) {
5936 		ciaa = (vf << 16) | 0x80000000;
5937 		/* 32 bit read so align, we really want status at offset 6 */
5938 		ciaa |= PCI_COMMAND;
5939 		IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5940 		ciad = IXGBE_READ_REG(hw, IXGBE_CIAD_82599);
5941 		ciaa &= 0x7FFFFFFF;
5942 		/* disable debug mode asap after reading data */
5943 		IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5944 		/* Get the upper 16 bits which will be the PCI status reg */
5945 		ciad >>= 16;
5946 		if (ciad & PCI_STATUS_REC_MASTER_ABORT) {
5947 			netdev_err(netdev, "VF %d Hung DMA\n", vf);
5948 			/* Issue VFLR */
5949 			ciaa = (vf << 16) | 0x80000000;
5950 			ciaa |= 0xA8;
5951 			IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5952 			ciad = 0x00008000;  /* VFLR */
5953 			IXGBE_WRITE_REG(hw, IXGBE_CIAD_82599, ciad);
5954 			ciaa &= 0x7FFFFFFF;
5955 			IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5956 		}
5957 	}
5958 }
5959 
5960 #endif
5961 /**
5962  * ixgbe_service_timer - Timer Call-back
5963  * @data: pointer to adapter cast into an unsigned long
5964  **/
5965 static void ixgbe_service_timer(unsigned long data)
5966 {
5967 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
5968 	unsigned long next_event_offset;
5969 	bool ready = true;
5970 
5971 	/* poll faster when waiting for link */
5972 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
5973 		next_event_offset = HZ / 10;
5974 	else
5975 		next_event_offset = HZ * 2;
5976 
5977 #ifdef CONFIG_PCI_IOV
5978 	/*
5979 	 * don't bother with SR-IOV VF DMA hang check if there are
5980 	 * no VFs or the link is down
5981 	 */
5982 	if (!adapter->num_vfs ||
5983 	    (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
5984 		goto normal_timer_service;
5985 
5986 	/* If we have VFs allocated then we must check for DMA hangs */
5987 	ixgbe_check_for_bad_vf(adapter);
5988 	next_event_offset = HZ / 50;
5989 	adapter->timer_event_accumulator++;
5990 
5991 	if (adapter->timer_event_accumulator >= 100)
5992 		adapter->timer_event_accumulator = 0;
5993 	else
5994 		ready = false;
5995 
5996 normal_timer_service:
5997 #endif
5998 	/* Reset the timer */
5999 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
6000 
6001 	if (ready)
6002 		ixgbe_service_event_schedule(adapter);
6003 }
6004 
6005 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
6006 {
6007 	if (!(adapter->flags2 & IXGBE_FLAG2_RESET_REQUESTED))
6008 		return;
6009 
6010 	adapter->flags2 &= ~IXGBE_FLAG2_RESET_REQUESTED;
6011 
6012 	/* If we're already down or resetting, just bail */
6013 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6014 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6015 		return;
6016 
6017 	ixgbe_dump(adapter);
6018 	netdev_err(adapter->netdev, "Reset adapter\n");
6019 	adapter->tx_timeout_count++;
6020 
6021 	ixgbe_reinit_locked(adapter);
6022 }
6023 
6024 /**
6025  * ixgbe_service_task - manages and runs subtasks
6026  * @work: pointer to work_struct containing our data
6027  **/
6028 static void ixgbe_service_task(struct work_struct *work)
6029 {
6030 	struct ixgbe_adapter *adapter = container_of(work,
6031 						     struct ixgbe_adapter,
6032 						     service_task);
6033 	ixgbe_reset_subtask(adapter);
6034 	ixgbe_sfp_detection_subtask(adapter);
6035 	ixgbe_sfp_link_config_subtask(adapter);
6036 	ixgbe_check_overtemp_subtask(adapter);
6037 	ixgbe_watchdog_subtask(adapter);
6038 	ixgbe_fdir_reinit_subtask(adapter);
6039 	ixgbe_check_hang_subtask(adapter);
6040 
6041 	if (adapter->flags2 & IXGBE_FLAG2_PTP_ENABLED) {
6042 		ixgbe_ptp_overflow_check(adapter);
6043 		ixgbe_ptp_rx_hang(adapter);
6044 	}
6045 
6046 	ixgbe_service_event_complete(adapter);
6047 }
6048 
6049 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
6050 		     struct ixgbe_tx_buffer *first,
6051 		     u8 *hdr_len)
6052 {
6053 	struct sk_buff *skb = first->skb;
6054 	u32 vlan_macip_lens, type_tucmd;
6055 	u32 mss_l4len_idx, l4len;
6056 
6057 	if (skb->ip_summed != CHECKSUM_PARTIAL)
6058 		return 0;
6059 
6060 	if (!skb_is_gso(skb))
6061 		return 0;
6062 
6063 	if (skb_header_cloned(skb)) {
6064 		int err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
6065 		if (err)
6066 			return err;
6067 	}
6068 
6069 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
6070 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
6071 
6072 	if (first->protocol == __constant_htons(ETH_P_IP)) {
6073 		struct iphdr *iph = ip_hdr(skb);
6074 		iph->tot_len = 0;
6075 		iph->check = 0;
6076 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
6077 							 iph->daddr, 0,
6078 							 IPPROTO_TCP,
6079 							 0);
6080 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
6081 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
6082 				   IXGBE_TX_FLAGS_CSUM |
6083 				   IXGBE_TX_FLAGS_IPV4;
6084 	} else if (skb_is_gso_v6(skb)) {
6085 		ipv6_hdr(skb)->payload_len = 0;
6086 		tcp_hdr(skb)->check =
6087 		    ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
6088 				     &ipv6_hdr(skb)->daddr,
6089 				     0, IPPROTO_TCP, 0);
6090 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
6091 				   IXGBE_TX_FLAGS_CSUM;
6092 	}
6093 
6094 	/* compute header lengths */
6095 	l4len = tcp_hdrlen(skb);
6096 	*hdr_len = skb_transport_offset(skb) + l4len;
6097 
6098 	/* update gso size and bytecount with header size */
6099 	first->gso_segs = skb_shinfo(skb)->gso_segs;
6100 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
6101 
6102 	/* mss_l4len_id: use 0 as index for TSO */
6103 	mss_l4len_idx = l4len << IXGBE_ADVTXD_L4LEN_SHIFT;
6104 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
6105 
6106 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
6107 	vlan_macip_lens = skb_network_header_len(skb);
6108 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
6109 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
6110 
6111 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
6112 			  mss_l4len_idx);
6113 
6114 	return 1;
6115 }
6116 
6117 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
6118 			  struct ixgbe_tx_buffer *first)
6119 {
6120 	struct sk_buff *skb = first->skb;
6121 	u32 vlan_macip_lens = 0;
6122 	u32 mss_l4len_idx = 0;
6123 	u32 type_tucmd = 0;
6124 
6125 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
6126 		if (!(first->tx_flags & IXGBE_TX_FLAGS_HW_VLAN) &&
6127 		    !(first->tx_flags & IXGBE_TX_FLAGS_CC))
6128 			return;
6129 	} else {
6130 		u8 l4_hdr = 0;
6131 		switch (first->protocol) {
6132 		case __constant_htons(ETH_P_IP):
6133 			vlan_macip_lens |= skb_network_header_len(skb);
6134 			type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
6135 			l4_hdr = ip_hdr(skb)->protocol;
6136 			break;
6137 		case __constant_htons(ETH_P_IPV6):
6138 			vlan_macip_lens |= skb_network_header_len(skb);
6139 			l4_hdr = ipv6_hdr(skb)->nexthdr;
6140 			break;
6141 		default:
6142 			if (unlikely(net_ratelimit())) {
6143 				dev_warn(tx_ring->dev,
6144 				 "partial checksum but proto=%x!\n",
6145 				 first->protocol);
6146 			}
6147 			break;
6148 		}
6149 
6150 		switch (l4_hdr) {
6151 		case IPPROTO_TCP:
6152 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_TCP;
6153 			mss_l4len_idx = tcp_hdrlen(skb) <<
6154 					IXGBE_ADVTXD_L4LEN_SHIFT;
6155 			break;
6156 		case IPPROTO_SCTP:
6157 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_SCTP;
6158 			mss_l4len_idx = sizeof(struct sctphdr) <<
6159 					IXGBE_ADVTXD_L4LEN_SHIFT;
6160 			break;
6161 		case IPPROTO_UDP:
6162 			mss_l4len_idx = sizeof(struct udphdr) <<
6163 					IXGBE_ADVTXD_L4LEN_SHIFT;
6164 			break;
6165 		default:
6166 			if (unlikely(net_ratelimit())) {
6167 				dev_warn(tx_ring->dev,
6168 				 "partial checksum but l4 proto=%x!\n",
6169 				 l4_hdr);
6170 			}
6171 			break;
6172 		}
6173 
6174 		/* update TX checksum flag */
6175 		first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
6176 	}
6177 
6178 	/* vlan_macip_lens: MACLEN, VLAN tag */
6179 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
6180 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
6181 
6182 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0,
6183 			  type_tucmd, mss_l4len_idx);
6184 }
6185 
6186 #define IXGBE_SET_FLAG(_input, _flag, _result) \
6187 	((_flag <= _result) ? \
6188 	 ((u32)(_input & _flag) * (_result / _flag)) : \
6189 	 ((u32)(_input & _flag) / (_flag / _result)))
6190 
6191 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
6192 {
6193 	/* set type for advanced descriptor with frame checksum insertion */
6194 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
6195 		       IXGBE_ADVTXD_DCMD_DEXT |
6196 		       IXGBE_ADVTXD_DCMD_IFCS;
6197 
6198 	/* set HW vlan bit if vlan is present */
6199 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
6200 				   IXGBE_ADVTXD_DCMD_VLE);
6201 
6202 	/* set segmentation enable bits for TSO/FSO */
6203 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
6204 				   IXGBE_ADVTXD_DCMD_TSE);
6205 
6206 	/* set timestamp bit if present */
6207 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
6208 				   IXGBE_ADVTXD_MAC_TSTAMP);
6209 
6210 	/* insert frame checksum */
6211 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
6212 
6213 	return cmd_type;
6214 }
6215 
6216 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
6217 				   u32 tx_flags, unsigned int paylen)
6218 {
6219 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
6220 
6221 	/* enable L4 checksum for TSO and TX checksum offload */
6222 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
6223 					IXGBE_TX_FLAGS_CSUM,
6224 					IXGBE_ADVTXD_POPTS_TXSM);
6225 
6226 	/* enble IPv4 checksum for TSO */
6227 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
6228 					IXGBE_TX_FLAGS_IPV4,
6229 					IXGBE_ADVTXD_POPTS_IXSM);
6230 
6231 	/*
6232 	 * Check Context must be set if Tx switch is enabled, which it
6233 	 * always is for case where virtual functions are running
6234 	 */
6235 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
6236 					IXGBE_TX_FLAGS_CC,
6237 					IXGBE_ADVTXD_CC);
6238 
6239 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
6240 }
6241 
6242 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
6243 		       IXGBE_TXD_CMD_RS)
6244 
6245 static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
6246 			 struct ixgbe_tx_buffer *first,
6247 			 const u8 hdr_len)
6248 {
6249 	struct sk_buff *skb = first->skb;
6250 	struct ixgbe_tx_buffer *tx_buffer;
6251 	union ixgbe_adv_tx_desc *tx_desc;
6252 	struct skb_frag_struct *frag;
6253 	dma_addr_t dma;
6254 	unsigned int data_len, size;
6255 	u32 tx_flags = first->tx_flags;
6256 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
6257 	u16 i = tx_ring->next_to_use;
6258 
6259 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
6260 
6261 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
6262 
6263 	size = skb_headlen(skb);
6264 	data_len = skb->data_len;
6265 
6266 #ifdef IXGBE_FCOE
6267 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
6268 		if (data_len < sizeof(struct fcoe_crc_eof)) {
6269 			size -= sizeof(struct fcoe_crc_eof) - data_len;
6270 			data_len = 0;
6271 		} else {
6272 			data_len -= sizeof(struct fcoe_crc_eof);
6273 		}
6274 	}
6275 
6276 #endif
6277 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
6278 
6279 	tx_buffer = first;
6280 
6281 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
6282 		if (dma_mapping_error(tx_ring->dev, dma))
6283 			goto dma_error;
6284 
6285 		/* record length, and DMA address */
6286 		dma_unmap_len_set(tx_buffer, len, size);
6287 		dma_unmap_addr_set(tx_buffer, dma, dma);
6288 
6289 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
6290 
6291 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
6292 			tx_desc->read.cmd_type_len =
6293 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
6294 
6295 			i++;
6296 			tx_desc++;
6297 			if (i == tx_ring->count) {
6298 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6299 				i = 0;
6300 			}
6301 			tx_desc->read.olinfo_status = 0;
6302 
6303 			dma += IXGBE_MAX_DATA_PER_TXD;
6304 			size -= IXGBE_MAX_DATA_PER_TXD;
6305 
6306 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
6307 		}
6308 
6309 		if (likely(!data_len))
6310 			break;
6311 
6312 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
6313 
6314 		i++;
6315 		tx_desc++;
6316 		if (i == tx_ring->count) {
6317 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6318 			i = 0;
6319 		}
6320 		tx_desc->read.olinfo_status = 0;
6321 
6322 #ifdef IXGBE_FCOE
6323 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
6324 #else
6325 		size = skb_frag_size(frag);
6326 #endif
6327 		data_len -= size;
6328 
6329 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
6330 				       DMA_TO_DEVICE);
6331 
6332 		tx_buffer = &tx_ring->tx_buffer_info[i];
6333 	}
6334 
6335 	/* write last descriptor with RS and EOP bits */
6336 	cmd_type |= size | IXGBE_TXD_CMD;
6337 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
6338 
6339 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
6340 
6341 	/* set the timestamp */
6342 	first->time_stamp = jiffies;
6343 
6344 	/*
6345 	 * Force memory writes to complete before letting h/w know there
6346 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
6347 	 * memory model archs, such as IA-64).
6348 	 *
6349 	 * We also need this memory barrier to make certain all of the
6350 	 * status bits have been updated before next_to_watch is written.
6351 	 */
6352 	wmb();
6353 
6354 	/* set next_to_watch value indicating a packet is present */
6355 	first->next_to_watch = tx_desc;
6356 
6357 	i++;
6358 	if (i == tx_ring->count)
6359 		i = 0;
6360 
6361 	tx_ring->next_to_use = i;
6362 
6363 	/* notify HW of packet */
6364 	writel(i, tx_ring->tail);
6365 
6366 	return;
6367 dma_error:
6368 	dev_err(tx_ring->dev, "TX DMA map failed\n");
6369 
6370 	/* clear dma mappings for failed tx_buffer_info map */
6371 	for (;;) {
6372 		tx_buffer = &tx_ring->tx_buffer_info[i];
6373 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer);
6374 		if (tx_buffer == first)
6375 			break;
6376 		if (i == 0)
6377 			i = tx_ring->count;
6378 		i--;
6379 	}
6380 
6381 	tx_ring->next_to_use = i;
6382 }
6383 
6384 static void ixgbe_atr(struct ixgbe_ring *ring,
6385 		      struct ixgbe_tx_buffer *first)
6386 {
6387 	struct ixgbe_q_vector *q_vector = ring->q_vector;
6388 	union ixgbe_atr_hash_dword input = { .dword = 0 };
6389 	union ixgbe_atr_hash_dword common = { .dword = 0 };
6390 	union {
6391 		unsigned char *network;
6392 		struct iphdr *ipv4;
6393 		struct ipv6hdr *ipv6;
6394 	} hdr;
6395 	struct tcphdr *th;
6396 	__be16 vlan_id;
6397 
6398 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
6399 	if (!q_vector)
6400 		return;
6401 
6402 	/* do nothing if sampling is disabled */
6403 	if (!ring->atr_sample_rate)
6404 		return;
6405 
6406 	ring->atr_count++;
6407 
6408 	/* snag network header to get L4 type and address */
6409 	hdr.network = skb_network_header(first->skb);
6410 
6411 	/* Currently only IPv4/IPv6 with TCP is supported */
6412 	if ((first->protocol != __constant_htons(ETH_P_IPV6) ||
6413 	     hdr.ipv6->nexthdr != IPPROTO_TCP) &&
6414 	    (first->protocol != __constant_htons(ETH_P_IP) ||
6415 	     hdr.ipv4->protocol != IPPROTO_TCP))
6416 		return;
6417 
6418 	th = tcp_hdr(first->skb);
6419 
6420 	/* skip this packet since it is invalid or the socket is closing */
6421 	if (!th || th->fin)
6422 		return;
6423 
6424 	/* sample on all syn packets or once every atr sample count */
6425 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
6426 		return;
6427 
6428 	/* reset sample count */
6429 	ring->atr_count = 0;
6430 
6431 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
6432 
6433 	/*
6434 	 * src and dst are inverted, think how the receiver sees them
6435 	 *
6436 	 * The input is broken into two sections, a non-compressed section
6437 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
6438 	 * is XORed together and stored in the compressed dword.
6439 	 */
6440 	input.formatted.vlan_id = vlan_id;
6441 
6442 	/*
6443 	 * since src port and flex bytes occupy the same word XOR them together
6444 	 * and write the value to source port portion of compressed dword
6445 	 */
6446 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
6447 		common.port.src ^= th->dest ^ __constant_htons(ETH_P_8021Q);
6448 	else
6449 		common.port.src ^= th->dest ^ first->protocol;
6450 	common.port.dst ^= th->source;
6451 
6452 	if (first->protocol == __constant_htons(ETH_P_IP)) {
6453 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
6454 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
6455 	} else {
6456 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
6457 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
6458 			     hdr.ipv6->saddr.s6_addr32[1] ^
6459 			     hdr.ipv6->saddr.s6_addr32[2] ^
6460 			     hdr.ipv6->saddr.s6_addr32[3] ^
6461 			     hdr.ipv6->daddr.s6_addr32[0] ^
6462 			     hdr.ipv6->daddr.s6_addr32[1] ^
6463 			     hdr.ipv6->daddr.s6_addr32[2] ^
6464 			     hdr.ipv6->daddr.s6_addr32[3];
6465 	}
6466 
6467 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
6468 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
6469 					      input, common, ring->queue_index);
6470 }
6471 
6472 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
6473 {
6474 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
6475 	/* Herbert's original patch had:
6476 	 *  smp_mb__after_netif_stop_queue();
6477 	 * but since that doesn't exist yet, just open code it. */
6478 	smp_mb();
6479 
6480 	/* We need to check again in a case another CPU has just
6481 	 * made room available. */
6482 	if (likely(ixgbe_desc_unused(tx_ring) < size))
6483 		return -EBUSY;
6484 
6485 	/* A reprieve! - use start_queue because it doesn't call schedule */
6486 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
6487 	++tx_ring->tx_stats.restart_queue;
6488 	return 0;
6489 }
6490 
6491 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
6492 {
6493 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
6494 		return 0;
6495 	return __ixgbe_maybe_stop_tx(tx_ring, size);
6496 }
6497 
6498 #ifdef IXGBE_FCOE
6499 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb)
6500 {
6501 	struct ixgbe_adapter *adapter;
6502 	struct ixgbe_ring_feature *f;
6503 	int txq;
6504 
6505 	/*
6506 	 * only execute the code below if protocol is FCoE
6507 	 * or FIP and we have FCoE enabled on the adapter
6508 	 */
6509 	switch (vlan_get_protocol(skb)) {
6510 	case __constant_htons(ETH_P_FCOE):
6511 	case __constant_htons(ETH_P_FIP):
6512 		adapter = netdev_priv(dev);
6513 
6514 		if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
6515 			break;
6516 	default:
6517 		return __netdev_pick_tx(dev, skb);
6518 	}
6519 
6520 	f = &adapter->ring_feature[RING_F_FCOE];
6521 
6522 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
6523 					   smp_processor_id();
6524 
6525 	while (txq >= f->indices)
6526 		txq -= f->indices;
6527 
6528 	return txq + f->offset;
6529 }
6530 
6531 #endif
6532 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
6533 			  struct ixgbe_adapter *adapter,
6534 			  struct ixgbe_ring *tx_ring)
6535 {
6536 	struct ixgbe_tx_buffer *first;
6537 	int tso;
6538 	u32 tx_flags = 0;
6539 	unsigned short f;
6540 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
6541 	__be16 protocol = skb->protocol;
6542 	u8 hdr_len = 0;
6543 
6544 	/*
6545 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
6546 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
6547 	 *       + 2 desc gap to keep tail from touching head,
6548 	 *       + 1 desc for context descriptor,
6549 	 * otherwise try next time
6550 	 */
6551 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
6552 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
6553 
6554 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
6555 		tx_ring->tx_stats.tx_busy++;
6556 		return NETDEV_TX_BUSY;
6557 	}
6558 
6559 	/* record the location of the first descriptor for this packet */
6560 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
6561 	first->skb = skb;
6562 	first->bytecount = skb->len;
6563 	first->gso_segs = 1;
6564 
6565 	/* if we have a HW VLAN tag being added default to the HW one */
6566 	if (vlan_tx_tag_present(skb)) {
6567 		tx_flags |= vlan_tx_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
6568 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
6569 	/* else if it is a SW VLAN check the next protocol and store the tag */
6570 	} else if (protocol == __constant_htons(ETH_P_8021Q)) {
6571 		struct vlan_hdr *vhdr, _vhdr;
6572 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
6573 		if (!vhdr)
6574 			goto out_drop;
6575 
6576 		protocol = vhdr->h_vlan_encapsulated_proto;
6577 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
6578 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
6579 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
6580 	}
6581 
6582 	skb_tx_timestamp(skb);
6583 
6584 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) {
6585 		skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
6586 		tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
6587 
6588 		/* schedule check for Tx timestamp */
6589 		adapter->ptp_tx_skb = skb_get(skb);
6590 		adapter->ptp_tx_start = jiffies;
6591 		schedule_work(&adapter->ptp_tx_work);
6592 	}
6593 
6594 #ifdef CONFIG_PCI_IOV
6595 	/*
6596 	 * Use the l2switch_enable flag - would be false if the DMA
6597 	 * Tx switch had been disabled.
6598 	 */
6599 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
6600 		tx_flags |= IXGBE_TX_FLAGS_CC;
6601 
6602 #endif
6603 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
6604 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
6605 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
6606 	     (skb->priority != TC_PRIO_CONTROL))) {
6607 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
6608 		tx_flags |= (skb->priority & 0x7) <<
6609 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
6610 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
6611 			struct vlan_ethhdr *vhdr;
6612 			if (skb_header_cloned(skb) &&
6613 			    pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
6614 				goto out_drop;
6615 			vhdr = (struct vlan_ethhdr *)skb->data;
6616 			vhdr->h_vlan_TCI = htons(tx_flags >>
6617 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
6618 		} else {
6619 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
6620 		}
6621 	}
6622 
6623 	/* record initial flags and protocol */
6624 	first->tx_flags = tx_flags;
6625 	first->protocol = protocol;
6626 
6627 #ifdef IXGBE_FCOE
6628 	/* setup tx offload for FCoE */
6629 	if ((protocol == __constant_htons(ETH_P_FCOE)) &&
6630 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
6631 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
6632 		if (tso < 0)
6633 			goto out_drop;
6634 
6635 		goto xmit_fcoe;
6636 	}
6637 
6638 #endif /* IXGBE_FCOE */
6639 	tso = ixgbe_tso(tx_ring, first, &hdr_len);
6640 	if (tso < 0)
6641 		goto out_drop;
6642 	else if (!tso)
6643 		ixgbe_tx_csum(tx_ring, first);
6644 
6645 	/* add the ATR filter if ATR is on */
6646 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
6647 		ixgbe_atr(tx_ring, first);
6648 
6649 #ifdef IXGBE_FCOE
6650 xmit_fcoe:
6651 #endif /* IXGBE_FCOE */
6652 	ixgbe_tx_map(tx_ring, first, hdr_len);
6653 
6654 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
6655 
6656 	return NETDEV_TX_OK;
6657 
6658 out_drop:
6659 	dev_kfree_skb_any(first->skb);
6660 	first->skb = NULL;
6661 
6662 	return NETDEV_TX_OK;
6663 }
6664 
6665 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
6666 				    struct net_device *netdev)
6667 {
6668 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6669 	struct ixgbe_ring *tx_ring;
6670 
6671 	/*
6672 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
6673 	 * in order to meet this minimum size requirement.
6674 	 */
6675 	if (unlikely(skb->len < 17)) {
6676 		if (skb_pad(skb, 17 - skb->len))
6677 			return NETDEV_TX_OK;
6678 		skb->len = 17;
6679 		skb_set_tail_pointer(skb, 17);
6680 	}
6681 
6682 	tx_ring = adapter->tx_ring[skb->queue_mapping];
6683 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
6684 }
6685 
6686 /**
6687  * ixgbe_set_mac - Change the Ethernet Address of the NIC
6688  * @netdev: network interface device structure
6689  * @p: pointer to an address structure
6690  *
6691  * Returns 0 on success, negative on failure
6692  **/
6693 static int ixgbe_set_mac(struct net_device *netdev, void *p)
6694 {
6695 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6696 	struct ixgbe_hw *hw = &adapter->hw;
6697 	struct sockaddr *addr = p;
6698 
6699 	if (!is_valid_ether_addr(addr->sa_data))
6700 		return -EADDRNOTAVAIL;
6701 
6702 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
6703 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
6704 
6705 	hw->mac.ops.set_rar(hw, 0, hw->mac.addr, VMDQ_P(0), IXGBE_RAH_AV);
6706 
6707 	return 0;
6708 }
6709 
6710 static int
6711 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
6712 {
6713 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6714 	struct ixgbe_hw *hw = &adapter->hw;
6715 	u16 value;
6716 	int rc;
6717 
6718 	if (prtad != hw->phy.mdio.prtad)
6719 		return -EINVAL;
6720 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
6721 	if (!rc)
6722 		rc = value;
6723 	return rc;
6724 }
6725 
6726 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
6727 			    u16 addr, u16 value)
6728 {
6729 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6730 	struct ixgbe_hw *hw = &adapter->hw;
6731 
6732 	if (prtad != hw->phy.mdio.prtad)
6733 		return -EINVAL;
6734 	return hw->phy.ops.write_reg(hw, addr, devad, value);
6735 }
6736 
6737 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
6738 {
6739 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6740 
6741 	switch (cmd) {
6742 	case SIOCSHWTSTAMP:
6743 		return ixgbe_ptp_hwtstamp_ioctl(adapter, req, cmd);
6744 	default:
6745 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
6746 	}
6747 }
6748 
6749 /**
6750  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
6751  * netdev->dev_addrs
6752  * @netdev: network interface device structure
6753  *
6754  * Returns non-zero on failure
6755  **/
6756 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
6757 {
6758 	int err = 0;
6759 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6760 	struct ixgbe_hw *hw = &adapter->hw;
6761 
6762 	if (is_valid_ether_addr(hw->mac.san_addr)) {
6763 		rtnl_lock();
6764 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
6765 		rtnl_unlock();
6766 
6767 		/* update SAN MAC vmdq pool selection */
6768 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
6769 	}
6770 	return err;
6771 }
6772 
6773 /**
6774  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
6775  * netdev->dev_addrs
6776  * @netdev: network interface device structure
6777  *
6778  * Returns non-zero on failure
6779  **/
6780 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
6781 {
6782 	int err = 0;
6783 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6784 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
6785 
6786 	if (is_valid_ether_addr(mac->san_addr)) {
6787 		rtnl_lock();
6788 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
6789 		rtnl_unlock();
6790 	}
6791 	return err;
6792 }
6793 
6794 #ifdef CONFIG_NET_POLL_CONTROLLER
6795 /*
6796  * Polling 'interrupt' - used by things like netconsole to send skbs
6797  * without having to re-enable interrupts. It's not called while
6798  * the interrupt routine is executing.
6799  */
6800 static void ixgbe_netpoll(struct net_device *netdev)
6801 {
6802 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6803 	int i;
6804 
6805 	/* if interface is down do nothing */
6806 	if (test_bit(__IXGBE_DOWN, &adapter->state))
6807 		return;
6808 
6809 	adapter->flags |= IXGBE_FLAG_IN_NETPOLL;
6810 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
6811 		for (i = 0; i < adapter->num_q_vectors; i++)
6812 			ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
6813 	} else {
6814 		ixgbe_intr(adapter->pdev->irq, netdev);
6815 	}
6816 	adapter->flags &= ~IXGBE_FLAG_IN_NETPOLL;
6817 }
6818 
6819 #endif
6820 static struct rtnl_link_stats64 *ixgbe_get_stats64(struct net_device *netdev,
6821 						   struct rtnl_link_stats64 *stats)
6822 {
6823 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6824 	int i;
6825 
6826 	rcu_read_lock();
6827 	for (i = 0; i < adapter->num_rx_queues; i++) {
6828 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
6829 		u64 bytes, packets;
6830 		unsigned int start;
6831 
6832 		if (ring) {
6833 			do {
6834 				start = u64_stats_fetch_begin_bh(&ring->syncp);
6835 				packets = ring->stats.packets;
6836 				bytes   = ring->stats.bytes;
6837 			} while (u64_stats_fetch_retry_bh(&ring->syncp, start));
6838 			stats->rx_packets += packets;
6839 			stats->rx_bytes   += bytes;
6840 		}
6841 	}
6842 
6843 	for (i = 0; i < adapter->num_tx_queues; i++) {
6844 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
6845 		u64 bytes, packets;
6846 		unsigned int start;
6847 
6848 		if (ring) {
6849 			do {
6850 				start = u64_stats_fetch_begin_bh(&ring->syncp);
6851 				packets = ring->stats.packets;
6852 				bytes   = ring->stats.bytes;
6853 			} while (u64_stats_fetch_retry_bh(&ring->syncp, start));
6854 			stats->tx_packets += packets;
6855 			stats->tx_bytes   += bytes;
6856 		}
6857 	}
6858 	rcu_read_unlock();
6859 	/* following stats updated by ixgbe_watchdog_task() */
6860 	stats->multicast	= netdev->stats.multicast;
6861 	stats->rx_errors	= netdev->stats.rx_errors;
6862 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
6863 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
6864 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
6865 	return stats;
6866 }
6867 
6868 #ifdef CONFIG_IXGBE_DCB
6869 /**
6870  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
6871  * @adapter: pointer to ixgbe_adapter
6872  * @tc: number of traffic classes currently enabled
6873  *
6874  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
6875  * 802.1Q priority maps to a packet buffer that exists.
6876  */
6877 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
6878 {
6879 	struct ixgbe_hw *hw = &adapter->hw;
6880 	u32 reg, rsave;
6881 	int i;
6882 
6883 	/* 82598 have a static priority to TC mapping that can not
6884 	 * be changed so no validation is needed.
6885 	 */
6886 	if (hw->mac.type == ixgbe_mac_82598EB)
6887 		return;
6888 
6889 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
6890 	rsave = reg;
6891 
6892 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
6893 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
6894 
6895 		/* If up2tc is out of bounds default to zero */
6896 		if (up2tc > tc)
6897 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
6898 	}
6899 
6900 	if (reg != rsave)
6901 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
6902 
6903 	return;
6904 }
6905 
6906 /**
6907  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
6908  * @adapter: Pointer to adapter struct
6909  *
6910  * Populate the netdev user priority to tc map
6911  */
6912 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
6913 {
6914 	struct net_device *dev = adapter->netdev;
6915 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
6916 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
6917 	u8 prio;
6918 
6919 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
6920 		u8 tc = 0;
6921 
6922 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
6923 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
6924 		else if (ets)
6925 			tc = ets->prio_tc[prio];
6926 
6927 		netdev_set_prio_tc_map(dev, prio, tc);
6928 	}
6929 }
6930 
6931 #endif /* CONFIG_IXGBE_DCB */
6932 /**
6933  * ixgbe_setup_tc - configure net_device for multiple traffic classes
6934  *
6935  * @netdev: net device to configure
6936  * @tc: number of traffic classes to enable
6937  */
6938 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
6939 {
6940 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6941 	struct ixgbe_hw *hw = &adapter->hw;
6942 
6943 	/* Hardware supports up to 8 traffic classes */
6944 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs ||
6945 	    (hw->mac.type == ixgbe_mac_82598EB &&
6946 	     tc < MAX_TRAFFIC_CLASS))
6947 		return -EINVAL;
6948 
6949 	/* Hardware has to reinitialize queues and interrupts to
6950 	 * match packet buffer alignment. Unfortunately, the
6951 	 * hardware is not flexible enough to do this dynamically.
6952 	 */
6953 	if (netif_running(dev))
6954 		ixgbe_close(dev);
6955 	ixgbe_clear_interrupt_scheme(adapter);
6956 
6957 #ifdef CONFIG_IXGBE_DCB
6958 	if (tc) {
6959 		netdev_set_num_tc(dev, tc);
6960 		ixgbe_set_prio_tc_map(adapter);
6961 
6962 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
6963 
6964 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
6965 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
6966 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
6967 		}
6968 	} else {
6969 		netdev_reset_tc(dev);
6970 
6971 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
6972 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
6973 
6974 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
6975 
6976 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
6977 		adapter->dcb_cfg.pfc_mode_enable = false;
6978 	}
6979 
6980 	ixgbe_validate_rtr(adapter, tc);
6981 
6982 #endif /* CONFIG_IXGBE_DCB */
6983 	ixgbe_init_interrupt_scheme(adapter);
6984 
6985 	if (netif_running(dev))
6986 		return ixgbe_open(dev);
6987 
6988 	return 0;
6989 }
6990 
6991 #ifdef CONFIG_PCI_IOV
6992 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
6993 {
6994 	struct net_device *netdev = adapter->netdev;
6995 
6996 	rtnl_lock();
6997 	ixgbe_setup_tc(netdev, netdev_get_num_tc(netdev));
6998 	rtnl_unlock();
6999 }
7000 
7001 #endif
7002 void ixgbe_do_reset(struct net_device *netdev)
7003 {
7004 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7005 
7006 	if (netif_running(netdev))
7007 		ixgbe_reinit_locked(adapter);
7008 	else
7009 		ixgbe_reset(adapter);
7010 }
7011 
7012 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
7013 					    netdev_features_t features)
7014 {
7015 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7016 
7017 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
7018 	if (!(features & NETIF_F_RXCSUM))
7019 		features &= ~NETIF_F_LRO;
7020 
7021 	/* Turn off LRO if not RSC capable */
7022 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
7023 		features &= ~NETIF_F_LRO;
7024 
7025 	return features;
7026 }
7027 
7028 static int ixgbe_set_features(struct net_device *netdev,
7029 			      netdev_features_t features)
7030 {
7031 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7032 	netdev_features_t changed = netdev->features ^ features;
7033 	bool need_reset = false;
7034 
7035 	/* Make sure RSC matches LRO, reset if change */
7036 	if (!(features & NETIF_F_LRO)) {
7037 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
7038 			need_reset = true;
7039 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
7040 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
7041 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
7042 		if (adapter->rx_itr_setting == 1 ||
7043 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
7044 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
7045 			need_reset = true;
7046 		} else if ((changed ^ features) & NETIF_F_LRO) {
7047 			e_info(probe, "rx-usecs set too low, "
7048 			       "disabling RSC\n");
7049 		}
7050 	}
7051 
7052 	/*
7053 	 * Check if Flow Director n-tuple support was enabled or disabled.  If
7054 	 * the state changed, we need to reset.
7055 	 */
7056 	switch (features & NETIF_F_NTUPLE) {
7057 	case NETIF_F_NTUPLE:
7058 		/* turn off ATR, enable perfect filters and reset */
7059 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
7060 			need_reset = true;
7061 
7062 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
7063 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
7064 		break;
7065 	default:
7066 		/* turn off perfect filters, enable ATR and reset */
7067 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
7068 			need_reset = true;
7069 
7070 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
7071 
7072 		/* We cannot enable ATR if SR-IOV is enabled */
7073 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7074 			break;
7075 
7076 		/* We cannot enable ATR if we have 2 or more traffic classes */
7077 		if (netdev_get_num_tc(netdev) > 1)
7078 			break;
7079 
7080 		/* We cannot enable ATR if RSS is disabled */
7081 		if (adapter->ring_feature[RING_F_RSS].limit <= 1)
7082 			break;
7083 
7084 		/* A sample rate of 0 indicates ATR disabled */
7085 		if (!adapter->atr_sample_rate)
7086 			break;
7087 
7088 		adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
7089 		break;
7090 	}
7091 
7092 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
7093 		ixgbe_vlan_strip_enable(adapter);
7094 	else
7095 		ixgbe_vlan_strip_disable(adapter);
7096 
7097 	if (changed & NETIF_F_RXALL)
7098 		need_reset = true;
7099 
7100 	netdev->features = features;
7101 	if (need_reset)
7102 		ixgbe_do_reset(netdev);
7103 
7104 	return 0;
7105 }
7106 
7107 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
7108 			     struct net_device *dev,
7109 			     const unsigned char *addr,
7110 			     u16 flags)
7111 {
7112 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7113 	int err;
7114 
7115 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
7116 		return ndo_dflt_fdb_add(ndm, tb, dev, addr, flags);
7117 
7118 	/* Hardware does not support aging addresses so if a
7119 	 * ndm_state is given only allow permanent addresses
7120 	 */
7121 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
7122 		pr_info("%s: FDB only supports static addresses\n",
7123 			ixgbe_driver_name);
7124 		return -EINVAL;
7125 	}
7126 
7127 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
7128 		u32 rar_uc_entries = IXGBE_MAX_PF_MACVLANS;
7129 
7130 		if (netdev_uc_count(dev) < rar_uc_entries)
7131 			err = dev_uc_add_excl(dev, addr);
7132 		else
7133 			err = -ENOMEM;
7134 	} else if (is_multicast_ether_addr(addr)) {
7135 		err = dev_mc_add_excl(dev, addr);
7136 	} else {
7137 		err = -EINVAL;
7138 	}
7139 
7140 	/* Only return duplicate errors if NLM_F_EXCL is set */
7141 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
7142 		err = 0;
7143 
7144 	return err;
7145 }
7146 
7147 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
7148 				    struct nlmsghdr *nlh)
7149 {
7150 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7151 	struct nlattr *attr, *br_spec;
7152 	int rem;
7153 
7154 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
7155 		return -EOPNOTSUPP;
7156 
7157 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
7158 
7159 	nla_for_each_nested(attr, br_spec, rem) {
7160 		__u16 mode;
7161 		u32 reg = 0;
7162 
7163 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
7164 			continue;
7165 
7166 		mode = nla_get_u16(attr);
7167 		if (mode == BRIDGE_MODE_VEPA) {
7168 			reg = 0;
7169 			adapter->flags2 &= ~IXGBE_FLAG2_BRIDGE_MODE_VEB;
7170 		} else if (mode == BRIDGE_MODE_VEB) {
7171 			reg = IXGBE_PFDTXGSWC_VT_LBEN;
7172 			adapter->flags2 |= IXGBE_FLAG2_BRIDGE_MODE_VEB;
7173 		} else
7174 			return -EINVAL;
7175 
7176 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, reg);
7177 
7178 		e_info(drv, "enabling bridge mode: %s\n",
7179 			mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
7180 	}
7181 
7182 	return 0;
7183 }
7184 
7185 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
7186 				    struct net_device *dev,
7187 				    u32 filter_mask)
7188 {
7189 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7190 	u16 mode;
7191 
7192 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
7193 		return 0;
7194 
7195 	if (adapter->flags2 & IXGBE_FLAG2_BRIDGE_MODE_VEB)
7196 		mode = BRIDGE_MODE_VEB;
7197 	else
7198 		mode = BRIDGE_MODE_VEPA;
7199 
7200 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev, mode);
7201 }
7202 
7203 static const struct net_device_ops ixgbe_netdev_ops = {
7204 	.ndo_open		= ixgbe_open,
7205 	.ndo_stop		= ixgbe_close,
7206 	.ndo_start_xmit		= ixgbe_xmit_frame,
7207 #ifdef IXGBE_FCOE
7208 	.ndo_select_queue	= ixgbe_select_queue,
7209 #endif
7210 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
7211 	.ndo_validate_addr	= eth_validate_addr,
7212 	.ndo_set_mac_address	= ixgbe_set_mac,
7213 	.ndo_change_mtu		= ixgbe_change_mtu,
7214 	.ndo_tx_timeout		= ixgbe_tx_timeout,
7215 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
7216 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
7217 	.ndo_do_ioctl		= ixgbe_ioctl,
7218 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
7219 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
7220 	.ndo_set_vf_tx_rate	= ixgbe_ndo_set_vf_bw,
7221 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
7222 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
7223 	.ndo_get_stats64	= ixgbe_get_stats64,
7224 #ifdef CONFIG_IXGBE_DCB
7225 	.ndo_setup_tc		= ixgbe_setup_tc,
7226 #endif
7227 #ifdef CONFIG_NET_POLL_CONTROLLER
7228 	.ndo_poll_controller	= ixgbe_netpoll,
7229 #endif
7230 #ifdef CONFIG_NET_RX_BUSY_POLL
7231 	.ndo_busy_poll		= ixgbe_low_latency_recv,
7232 #endif
7233 #ifdef IXGBE_FCOE
7234 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
7235 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
7236 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
7237 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
7238 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
7239 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
7240 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
7241 #endif /* IXGBE_FCOE */
7242 	.ndo_set_features = ixgbe_set_features,
7243 	.ndo_fix_features = ixgbe_fix_features,
7244 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
7245 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
7246 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
7247 };
7248 
7249 /**
7250  * ixgbe_wol_supported - Check whether device supports WoL
7251  * @hw: hw specific details
7252  * @device_id: the device ID
7253  * @subdev_id: the subsystem device ID
7254  *
7255  * This function is used by probe and ethtool to determine
7256  * which devices have WoL support
7257  *
7258  **/
7259 int ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
7260 			u16 subdevice_id)
7261 {
7262 	struct ixgbe_hw *hw = &adapter->hw;
7263 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
7264 	int is_wol_supported = 0;
7265 
7266 	switch (device_id) {
7267 	case IXGBE_DEV_ID_82599_SFP:
7268 		/* Only these subdevices could supports WOL */
7269 		switch (subdevice_id) {
7270 		case IXGBE_SUBDEV_ID_82599_560FLR:
7271 			/* only support first port */
7272 			if (hw->bus.func != 0)
7273 				break;
7274 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
7275 		case IXGBE_SUBDEV_ID_82599_SFP:
7276 		case IXGBE_SUBDEV_ID_82599_RNDC:
7277 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
7278 		case IXGBE_SUBDEV_ID_82599_LOM_SFP:
7279 			is_wol_supported = 1;
7280 			break;
7281 		}
7282 		break;
7283 	case IXGBE_DEV_ID_82599EN_SFP:
7284 		/* Only this subdevice supports WOL */
7285 		switch (subdevice_id) {
7286 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
7287 			is_wol_supported = 1;
7288 			break;
7289 		}
7290 		break;
7291 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
7292 		/* All except this subdevice support WOL */
7293 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
7294 			is_wol_supported = 1;
7295 		break;
7296 	case IXGBE_DEV_ID_82599_KX4:
7297 		is_wol_supported = 1;
7298 		break;
7299 	case IXGBE_DEV_ID_X540T:
7300 	case IXGBE_DEV_ID_X540T1:
7301 		/* check eeprom to see if enabled wol */
7302 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
7303 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
7304 		     (hw->bus.func == 0))) {
7305 			is_wol_supported = 1;
7306 		}
7307 		break;
7308 	}
7309 
7310 	return is_wol_supported;
7311 }
7312 
7313 /**
7314  * ixgbe_probe - Device Initialization Routine
7315  * @pdev: PCI device information struct
7316  * @ent: entry in ixgbe_pci_tbl
7317  *
7318  * Returns 0 on success, negative on failure
7319  *
7320  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
7321  * The OS initialization, configuring of the adapter private structure,
7322  * and a hardware reset occur.
7323  **/
7324 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7325 {
7326 	struct net_device *netdev;
7327 	struct ixgbe_adapter *adapter = NULL;
7328 	struct ixgbe_hw *hw;
7329 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
7330 	static int cards_found;
7331 	int i, err, pci_using_dac;
7332 	unsigned int indices = MAX_TX_QUEUES;
7333 	u8 part_str[IXGBE_PBANUM_LENGTH];
7334 #ifdef IXGBE_FCOE
7335 	u16 device_caps;
7336 #endif
7337 	u32 eec;
7338 
7339 	/* Catch broken hardware that put the wrong VF device ID in
7340 	 * the PCIe SR-IOV capability.
7341 	 */
7342 	if (pdev->is_virtfn) {
7343 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
7344 		     pci_name(pdev), pdev->vendor, pdev->device);
7345 		return -EINVAL;
7346 	}
7347 
7348 	err = pci_enable_device_mem(pdev);
7349 	if (err)
7350 		return err;
7351 
7352 	if (!dma_set_mask(&pdev->dev, DMA_BIT_MASK(64)) &&
7353 	    !dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(64))) {
7354 		pci_using_dac = 1;
7355 	} else {
7356 		err = dma_set_mask(&pdev->dev, DMA_BIT_MASK(32));
7357 		if (err) {
7358 			err = dma_set_coherent_mask(&pdev->dev,
7359 						    DMA_BIT_MASK(32));
7360 			if (err) {
7361 				dev_err(&pdev->dev,
7362 					"No usable DMA configuration, aborting\n");
7363 				goto err_dma;
7364 			}
7365 		}
7366 		pci_using_dac = 0;
7367 	}
7368 
7369 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
7370 					   IORESOURCE_MEM), ixgbe_driver_name);
7371 	if (err) {
7372 		dev_err(&pdev->dev,
7373 			"pci_request_selected_regions failed 0x%x\n", err);
7374 		goto err_pci_reg;
7375 	}
7376 
7377 	pci_enable_pcie_error_reporting(pdev);
7378 
7379 	pci_set_master(pdev);
7380 	pci_save_state(pdev);
7381 
7382 	if (ii->mac == ixgbe_mac_82598EB) {
7383 #ifdef CONFIG_IXGBE_DCB
7384 		/* 8 TC w/ 4 queues per TC */
7385 		indices = 4 * MAX_TRAFFIC_CLASS;
7386 #else
7387 		indices = IXGBE_MAX_RSS_INDICES;
7388 #endif
7389 	}
7390 
7391 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
7392 	if (!netdev) {
7393 		err = -ENOMEM;
7394 		goto err_alloc_etherdev;
7395 	}
7396 
7397 	SET_NETDEV_DEV(netdev, &pdev->dev);
7398 
7399 	adapter = netdev_priv(netdev);
7400 	pci_set_drvdata(pdev, adapter);
7401 
7402 	adapter->netdev = netdev;
7403 	adapter->pdev = pdev;
7404 	hw = &adapter->hw;
7405 	hw->back = adapter;
7406 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
7407 
7408 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
7409 			      pci_resource_len(pdev, 0));
7410 	if (!hw->hw_addr) {
7411 		err = -EIO;
7412 		goto err_ioremap;
7413 	}
7414 
7415 	netdev->netdev_ops = &ixgbe_netdev_ops;
7416 	ixgbe_set_ethtool_ops(netdev);
7417 	netdev->watchdog_timeo = 5 * HZ;
7418 	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
7419 
7420 	adapter->bd_number = cards_found;
7421 
7422 	/* Setup hw api */
7423 	memcpy(&hw->mac.ops, ii->mac_ops, sizeof(hw->mac.ops));
7424 	hw->mac.type  = ii->mac;
7425 
7426 	/* EEPROM */
7427 	memcpy(&hw->eeprom.ops, ii->eeprom_ops, sizeof(hw->eeprom.ops));
7428 	eec = IXGBE_READ_REG(hw, IXGBE_EEC);
7429 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
7430 	if (!(eec & (1 << 8)))
7431 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
7432 
7433 	/* PHY */
7434 	memcpy(&hw->phy.ops, ii->phy_ops, sizeof(hw->phy.ops));
7435 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
7436 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
7437 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
7438 	hw->phy.mdio.mmds = 0;
7439 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
7440 	hw->phy.mdio.dev = netdev;
7441 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
7442 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
7443 
7444 	ii->get_invariants(hw);
7445 
7446 	/* setup the private structure */
7447 	err = ixgbe_sw_init(adapter);
7448 	if (err)
7449 		goto err_sw_init;
7450 
7451 	/* Cache if MNG FW is up so we don't have to read the REG later */
7452 	if (hw->mac.ops.mng_fw_enabled)
7453 		hw->mng_fw_enabled = hw->mac.ops.mng_fw_enabled(hw);
7454 
7455 	/* Make it possible the adapter to be woken up via WOL */
7456 	switch (adapter->hw.mac.type) {
7457 	case ixgbe_mac_82599EB:
7458 	case ixgbe_mac_X540:
7459 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
7460 		break;
7461 	default:
7462 		break;
7463 	}
7464 
7465 	/*
7466 	 * If there is a fan on this device and it has failed log the
7467 	 * failure.
7468 	 */
7469 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
7470 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
7471 		if (esdp & IXGBE_ESDP_SDP1)
7472 			e_crit(probe, "Fan has stopped, replace the adapter\n");
7473 	}
7474 
7475 	if (allow_unsupported_sfp)
7476 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
7477 
7478 	/* reset_hw fills in the perm_addr as well */
7479 	hw->phy.reset_if_overtemp = true;
7480 	err = hw->mac.ops.reset_hw(hw);
7481 	hw->phy.reset_if_overtemp = false;
7482 	if (err == IXGBE_ERR_SFP_NOT_PRESENT &&
7483 	    hw->mac.type == ixgbe_mac_82598EB) {
7484 		err = 0;
7485 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
7486 		e_dev_err("failed to load because an unsupported SFP+ "
7487 			  "module type was detected.\n");
7488 		e_dev_err("Reload the driver after installing a supported "
7489 			  "module.\n");
7490 		goto err_sw_init;
7491 	} else if (err) {
7492 		e_dev_err("HW Init failed: %d\n", err);
7493 		goto err_sw_init;
7494 	}
7495 
7496 #ifdef CONFIG_PCI_IOV
7497 	/* SR-IOV not supported on the 82598 */
7498 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
7499 		goto skip_sriov;
7500 	/* Mailbox */
7501 	ixgbe_init_mbx_params_pf(hw);
7502 	memcpy(&hw->mbx.ops, ii->mbx_ops, sizeof(hw->mbx.ops));
7503 	ixgbe_enable_sriov(adapter);
7504 	pci_sriov_set_totalvfs(pdev, 63);
7505 skip_sriov:
7506 
7507 #endif
7508 	netdev->features = NETIF_F_SG |
7509 			   NETIF_F_IP_CSUM |
7510 			   NETIF_F_IPV6_CSUM |
7511 			   NETIF_F_HW_VLAN_CTAG_TX |
7512 			   NETIF_F_HW_VLAN_CTAG_RX |
7513 			   NETIF_F_HW_VLAN_CTAG_FILTER |
7514 			   NETIF_F_TSO |
7515 			   NETIF_F_TSO6 |
7516 			   NETIF_F_RXHASH |
7517 			   NETIF_F_RXCSUM;
7518 
7519 	netdev->hw_features = netdev->features;
7520 
7521 	switch (adapter->hw.mac.type) {
7522 	case ixgbe_mac_82599EB:
7523 	case ixgbe_mac_X540:
7524 		netdev->features |= NETIF_F_SCTP_CSUM;
7525 		netdev->hw_features |= NETIF_F_SCTP_CSUM |
7526 				       NETIF_F_NTUPLE;
7527 		break;
7528 	default:
7529 		break;
7530 	}
7531 
7532 	netdev->hw_features |= NETIF_F_RXALL;
7533 
7534 	netdev->vlan_features |= NETIF_F_TSO;
7535 	netdev->vlan_features |= NETIF_F_TSO6;
7536 	netdev->vlan_features |= NETIF_F_IP_CSUM;
7537 	netdev->vlan_features |= NETIF_F_IPV6_CSUM;
7538 	netdev->vlan_features |= NETIF_F_SG;
7539 
7540 	netdev->priv_flags |= IFF_UNICAST_FLT;
7541 	netdev->priv_flags |= IFF_SUPP_NOFCS;
7542 
7543 #ifdef CONFIG_IXGBE_DCB
7544 	netdev->dcbnl_ops = &dcbnl_ops;
7545 #endif
7546 
7547 #ifdef IXGBE_FCOE
7548 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
7549 		unsigned int fcoe_l;
7550 
7551 		if (hw->mac.ops.get_device_caps) {
7552 			hw->mac.ops.get_device_caps(hw, &device_caps);
7553 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
7554 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
7555 		}
7556 
7557 
7558 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
7559 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
7560 
7561 		netdev->features |= NETIF_F_FSO |
7562 				    NETIF_F_FCOE_CRC;
7563 
7564 		netdev->vlan_features |= NETIF_F_FSO |
7565 					 NETIF_F_FCOE_CRC |
7566 					 NETIF_F_FCOE_MTU;
7567 	}
7568 #endif /* IXGBE_FCOE */
7569 	if (pci_using_dac) {
7570 		netdev->features |= NETIF_F_HIGHDMA;
7571 		netdev->vlan_features |= NETIF_F_HIGHDMA;
7572 	}
7573 
7574 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
7575 		netdev->hw_features |= NETIF_F_LRO;
7576 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
7577 		netdev->features |= NETIF_F_LRO;
7578 
7579 	/* make sure the EEPROM is good */
7580 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
7581 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
7582 		err = -EIO;
7583 		goto err_sw_init;
7584 	}
7585 
7586 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
7587 
7588 	if (!is_valid_ether_addr(netdev->dev_addr)) {
7589 		e_dev_err("invalid MAC address\n");
7590 		err = -EIO;
7591 		goto err_sw_init;
7592 	}
7593 
7594 	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
7595 		    (unsigned long) adapter);
7596 
7597 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
7598 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
7599 
7600 	err = ixgbe_init_interrupt_scheme(adapter);
7601 	if (err)
7602 		goto err_sw_init;
7603 
7604 	/* WOL not supported for all devices */
7605 	adapter->wol = 0;
7606 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
7607 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
7608 						pdev->subsystem_device);
7609 	if (hw->wol_enabled)
7610 		adapter->wol = IXGBE_WUFC_MAG;
7611 
7612 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
7613 
7614 	/* save off EEPROM version number */
7615 	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
7616 	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
7617 
7618 	/* pick up the PCI bus settings for reporting later */
7619 	hw->mac.ops.get_bus_info(hw);
7620 	if (hw->device_id == IXGBE_DEV_ID_82599_SFP_SF_QP)
7621 		ixgbe_get_parent_bus_info(adapter);
7622 
7623 	/* print bus type/speed/width info */
7624 	e_dev_info("(PCI Express:%s:%s) %pM\n",
7625 		   (hw->bus.speed == ixgbe_bus_speed_8000 ? "8.0GT/s" :
7626 		    hw->bus.speed == ixgbe_bus_speed_5000 ? "5.0GT/s" :
7627 		    hw->bus.speed == ixgbe_bus_speed_2500 ? "2.5GT/s" :
7628 		    "Unknown"),
7629 		   (hw->bus.width == ixgbe_bus_width_pcie_x8 ? "Width x8" :
7630 		    hw->bus.width == ixgbe_bus_width_pcie_x4 ? "Width x4" :
7631 		    hw->bus.width == ixgbe_bus_width_pcie_x1 ? "Width x1" :
7632 		    "Unknown"),
7633 		   netdev->dev_addr);
7634 
7635 	err = ixgbe_read_pba_string_generic(hw, part_str, IXGBE_PBANUM_LENGTH);
7636 	if (err)
7637 		strncpy(part_str, "Unknown", IXGBE_PBANUM_LENGTH);
7638 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
7639 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
7640 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
7641 		           part_str);
7642 	else
7643 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
7644 			   hw->mac.type, hw->phy.type, part_str);
7645 
7646 	if (hw->bus.width <= ixgbe_bus_width_pcie_x4) {
7647 		e_dev_warn("PCI-Express bandwidth available for this card is "
7648 			   "not sufficient for optimal performance.\n");
7649 		e_dev_warn("For optimal performance a x8 PCI-Express slot "
7650 			   "is required.\n");
7651 	}
7652 
7653 	/* reset the hardware with the new settings */
7654 	err = hw->mac.ops.start_hw(hw);
7655 	if (err == IXGBE_ERR_EEPROM_VERSION) {
7656 		/* We are running on a pre-production device, log a warning */
7657 		e_dev_warn("This device is a pre-production adapter/LOM. "
7658 			   "Please be aware there may be issues associated "
7659 			   "with your hardware.  If you are experiencing "
7660 			   "problems please contact your Intel or hardware "
7661 			   "representative who provided you with this "
7662 			   "hardware.\n");
7663 	}
7664 	strcpy(netdev->name, "eth%d");
7665 	err = register_netdev(netdev);
7666 	if (err)
7667 		goto err_register;
7668 
7669 	/* power down the optics for 82599 SFP+ fiber */
7670 	if (hw->mac.ops.disable_tx_laser)
7671 		hw->mac.ops.disable_tx_laser(hw);
7672 
7673 	/* carrier off reporting is important to ethtool even BEFORE open */
7674 	netif_carrier_off(netdev);
7675 
7676 #ifdef CONFIG_IXGBE_DCA
7677 	if (dca_add_requester(&pdev->dev) == 0) {
7678 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
7679 		ixgbe_setup_dca(adapter);
7680 	}
7681 #endif
7682 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
7683 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
7684 		for (i = 0; i < adapter->num_vfs; i++)
7685 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
7686 	}
7687 
7688 	/* firmware requires driver version to be 0xFFFFFFFF
7689 	 * since os does not support feature
7690 	 */
7691 	if (hw->mac.ops.set_fw_drv_ver)
7692 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF,
7693 					   0xFF);
7694 
7695 	/* add san mac addr to netdev */
7696 	ixgbe_add_sanmac_netdev(netdev);
7697 
7698 	e_dev_info("%s\n", ixgbe_default_device_descr);
7699 	cards_found++;
7700 
7701 #ifdef CONFIG_IXGBE_HWMON
7702 	if (ixgbe_sysfs_init(adapter))
7703 		e_err(probe, "failed to allocate sysfs resources\n");
7704 #endif /* CONFIG_IXGBE_HWMON */
7705 
7706 	ixgbe_dbg_adapter_init(adapter);
7707 
7708 	/* Need link setup for MNG FW, else wait for IXGBE_UP */
7709 	if (hw->mng_fw_enabled && hw->mac.ops.setup_link)
7710 		hw->mac.ops.setup_link(hw,
7711 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
7712 			true);
7713 
7714 	return 0;
7715 
7716 err_register:
7717 	ixgbe_release_hw_control(adapter);
7718 	ixgbe_clear_interrupt_scheme(adapter);
7719 err_sw_init:
7720 	ixgbe_disable_sriov(adapter);
7721 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
7722 	iounmap(hw->hw_addr);
7723 err_ioremap:
7724 	free_netdev(netdev);
7725 err_alloc_etherdev:
7726 	pci_release_selected_regions(pdev,
7727 				     pci_select_bars(pdev, IORESOURCE_MEM));
7728 err_pci_reg:
7729 err_dma:
7730 	pci_disable_device(pdev);
7731 	return err;
7732 }
7733 
7734 /**
7735  * ixgbe_remove - Device Removal Routine
7736  * @pdev: PCI device information struct
7737  *
7738  * ixgbe_remove is called by the PCI subsystem to alert the driver
7739  * that it should release a PCI device.  The could be caused by a
7740  * Hot-Plug event, or because the driver is going to be removed from
7741  * memory.
7742  **/
7743 static void ixgbe_remove(struct pci_dev *pdev)
7744 {
7745 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7746 	struct net_device *netdev = adapter->netdev;
7747 
7748 	ixgbe_dbg_adapter_exit(adapter);
7749 
7750 	set_bit(__IXGBE_DOWN, &adapter->state);
7751 	cancel_work_sync(&adapter->service_task);
7752 
7753 
7754 #ifdef CONFIG_IXGBE_DCA
7755 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
7756 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
7757 		dca_remove_requester(&pdev->dev);
7758 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
7759 	}
7760 
7761 #endif
7762 #ifdef CONFIG_IXGBE_HWMON
7763 	ixgbe_sysfs_exit(adapter);
7764 #endif /* CONFIG_IXGBE_HWMON */
7765 
7766 	/* remove the added san mac */
7767 	ixgbe_del_sanmac_netdev(netdev);
7768 
7769 	if (netdev->reg_state == NETREG_REGISTERED)
7770 		unregister_netdev(netdev);
7771 
7772 #ifdef CONFIG_PCI_IOV
7773 	/*
7774 	 * Only disable SR-IOV on unload if the user specified the now
7775 	 * deprecated max_vfs module parameter.
7776 	 */
7777 	if (max_vfs)
7778 		ixgbe_disable_sriov(adapter);
7779 #endif
7780 	ixgbe_clear_interrupt_scheme(adapter);
7781 
7782 	ixgbe_release_hw_control(adapter);
7783 
7784 #ifdef CONFIG_DCB
7785 	kfree(adapter->ixgbe_ieee_pfc);
7786 	kfree(adapter->ixgbe_ieee_ets);
7787 
7788 #endif
7789 	iounmap(adapter->hw.hw_addr);
7790 	pci_release_selected_regions(pdev, pci_select_bars(pdev,
7791 				     IORESOURCE_MEM));
7792 
7793 	e_dev_info("complete\n");
7794 
7795 	free_netdev(netdev);
7796 
7797 	pci_disable_pcie_error_reporting(pdev);
7798 
7799 	pci_disable_device(pdev);
7800 }
7801 
7802 /**
7803  * ixgbe_io_error_detected - called when PCI error is detected
7804  * @pdev: Pointer to PCI device
7805  * @state: The current pci connection state
7806  *
7807  * This function is called after a PCI bus error affecting
7808  * this device has been detected.
7809  */
7810 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
7811 						pci_channel_state_t state)
7812 {
7813 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7814 	struct net_device *netdev = adapter->netdev;
7815 
7816 #ifdef CONFIG_PCI_IOV
7817 	struct pci_dev *bdev, *vfdev;
7818 	u32 dw0, dw1, dw2, dw3;
7819 	int vf, pos;
7820 	u16 req_id, pf_func;
7821 
7822 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7823 	    adapter->num_vfs == 0)
7824 		goto skip_bad_vf_detection;
7825 
7826 	bdev = pdev->bus->self;
7827 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
7828 		bdev = bdev->bus->self;
7829 
7830 	if (!bdev)
7831 		goto skip_bad_vf_detection;
7832 
7833 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
7834 	if (!pos)
7835 		goto skip_bad_vf_detection;
7836 
7837 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG, &dw0);
7838 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG + 4, &dw1);
7839 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG + 8, &dw2);
7840 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG + 12, &dw3);
7841 
7842 	req_id = dw1 >> 16;
7843 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
7844 	if (!(req_id & 0x0080))
7845 		goto skip_bad_vf_detection;
7846 
7847 	pf_func = req_id & 0x01;
7848 	if ((pf_func & 1) == (pdev->devfn & 1)) {
7849 		unsigned int device_id;
7850 
7851 		vf = (req_id & 0x7F) >> 1;
7852 		e_dev_err("VF %d has caused a PCIe error\n", vf);
7853 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
7854 				"%8.8x\tdw3: %8.8x\n",
7855 		dw0, dw1, dw2, dw3);
7856 		switch (adapter->hw.mac.type) {
7857 		case ixgbe_mac_82599EB:
7858 			device_id = IXGBE_82599_VF_DEVICE_ID;
7859 			break;
7860 		case ixgbe_mac_X540:
7861 			device_id = IXGBE_X540_VF_DEVICE_ID;
7862 			break;
7863 		default:
7864 			device_id = 0;
7865 			break;
7866 		}
7867 
7868 		/* Find the pci device of the offending VF */
7869 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
7870 		while (vfdev) {
7871 			if (vfdev->devfn == (req_id & 0xFF))
7872 				break;
7873 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
7874 					       device_id, vfdev);
7875 		}
7876 		/*
7877 		 * There's a slim chance the VF could have been hot plugged,
7878 		 * so if it is no longer present we don't need to issue the
7879 		 * VFLR.  Just clean up the AER in that case.
7880 		 */
7881 		if (vfdev) {
7882 			e_dev_err("Issuing VFLR to VF %d\n", vf);
7883 			pci_write_config_dword(vfdev, 0xA8, 0x00008000);
7884 			/* Free device reference count */
7885 			pci_dev_put(vfdev);
7886 		}
7887 
7888 		pci_cleanup_aer_uncorrect_error_status(pdev);
7889 	}
7890 
7891 	/*
7892 	 * Even though the error may have occurred on the other port
7893 	 * we still need to increment the vf error reference count for
7894 	 * both ports because the I/O resume function will be called
7895 	 * for both of them.
7896 	 */
7897 	adapter->vferr_refcount++;
7898 
7899 	return PCI_ERS_RESULT_RECOVERED;
7900 
7901 skip_bad_vf_detection:
7902 #endif /* CONFIG_PCI_IOV */
7903 	netif_device_detach(netdev);
7904 
7905 	if (state == pci_channel_io_perm_failure)
7906 		return PCI_ERS_RESULT_DISCONNECT;
7907 
7908 	if (netif_running(netdev))
7909 		ixgbe_down(adapter);
7910 	pci_disable_device(pdev);
7911 
7912 	/* Request a slot reset. */
7913 	return PCI_ERS_RESULT_NEED_RESET;
7914 }
7915 
7916 /**
7917  * ixgbe_io_slot_reset - called after the pci bus has been reset.
7918  * @pdev: Pointer to PCI device
7919  *
7920  * Restart the card from scratch, as if from a cold-boot.
7921  */
7922 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
7923 {
7924 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7925 	pci_ers_result_t result;
7926 	int err;
7927 
7928 	if (pci_enable_device_mem(pdev)) {
7929 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
7930 		result = PCI_ERS_RESULT_DISCONNECT;
7931 	} else {
7932 		pci_set_master(pdev);
7933 		pci_restore_state(pdev);
7934 		pci_save_state(pdev);
7935 
7936 		pci_wake_from_d3(pdev, false);
7937 
7938 		ixgbe_reset(adapter);
7939 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
7940 		result = PCI_ERS_RESULT_RECOVERED;
7941 	}
7942 
7943 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
7944 	if (err) {
7945 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
7946 			  "failed 0x%0x\n", err);
7947 		/* non-fatal, continue */
7948 	}
7949 
7950 	return result;
7951 }
7952 
7953 /**
7954  * ixgbe_io_resume - called when traffic can start flowing again.
7955  * @pdev: Pointer to PCI device
7956  *
7957  * This callback is called when the error recovery driver tells us that
7958  * its OK to resume normal operation.
7959  */
7960 static void ixgbe_io_resume(struct pci_dev *pdev)
7961 {
7962 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7963 	struct net_device *netdev = adapter->netdev;
7964 
7965 #ifdef CONFIG_PCI_IOV
7966 	if (adapter->vferr_refcount) {
7967 		e_info(drv, "Resuming after VF err\n");
7968 		adapter->vferr_refcount--;
7969 		return;
7970 	}
7971 
7972 #endif
7973 	if (netif_running(netdev))
7974 		ixgbe_up(adapter);
7975 
7976 	netif_device_attach(netdev);
7977 }
7978 
7979 static const struct pci_error_handlers ixgbe_err_handler = {
7980 	.error_detected = ixgbe_io_error_detected,
7981 	.slot_reset = ixgbe_io_slot_reset,
7982 	.resume = ixgbe_io_resume,
7983 };
7984 
7985 static struct pci_driver ixgbe_driver = {
7986 	.name     = ixgbe_driver_name,
7987 	.id_table = ixgbe_pci_tbl,
7988 	.probe    = ixgbe_probe,
7989 	.remove   = ixgbe_remove,
7990 #ifdef CONFIG_PM
7991 	.suspend  = ixgbe_suspend,
7992 	.resume   = ixgbe_resume,
7993 #endif
7994 	.shutdown = ixgbe_shutdown,
7995 	.sriov_configure = ixgbe_pci_sriov_configure,
7996 	.err_handler = &ixgbe_err_handler
7997 };
7998 
7999 /**
8000  * ixgbe_init_module - Driver Registration Routine
8001  *
8002  * ixgbe_init_module is the first routine called when the driver is
8003  * loaded. All it does is register with the PCI subsystem.
8004  **/
8005 static int __init ixgbe_init_module(void)
8006 {
8007 	int ret;
8008 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
8009 	pr_info("%s\n", ixgbe_copyright);
8010 
8011 	ixgbe_dbg_init();
8012 
8013 	ret = pci_register_driver(&ixgbe_driver);
8014 	if (ret) {
8015 		ixgbe_dbg_exit();
8016 		return ret;
8017 	}
8018 
8019 #ifdef CONFIG_IXGBE_DCA
8020 	dca_register_notify(&dca_notifier);
8021 #endif
8022 
8023 	return 0;
8024 }
8025 
8026 module_init(ixgbe_init_module);
8027 
8028 /**
8029  * ixgbe_exit_module - Driver Exit Cleanup Routine
8030  *
8031  * ixgbe_exit_module is called just before the driver is removed
8032  * from memory.
8033  **/
8034 static void __exit ixgbe_exit_module(void)
8035 {
8036 #ifdef CONFIG_IXGBE_DCA
8037 	dca_unregister_notify(&dca_notifier);
8038 #endif
8039 	pci_unregister_driver(&ixgbe_driver);
8040 
8041 	ixgbe_dbg_exit();
8042 
8043 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
8044 }
8045 
8046 #ifdef CONFIG_IXGBE_DCA
8047 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
8048 			    void *p)
8049 {
8050 	int ret_val;
8051 
8052 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
8053 					 __ixgbe_notify_dca);
8054 
8055 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
8056 }
8057 
8058 #endif /* CONFIG_IXGBE_DCA */
8059 
8060 module_exit(ixgbe_exit_module);
8061 
8062 /* ixgbe_main.c */
8063