xref: /linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision 0d456bad36d42d16022be045c8a53ddbb59ee478)
1 /*******************************************************************************
2 
3   Intel 10 Gigabit PCI Express Linux driver
4   Copyright(c) 1999 - 2012 Intel Corporation.
5 
6   This program is free software; you can redistribute it and/or modify it
7   under the terms and conditions of the GNU General Public License,
8   version 2, as published by the Free Software Foundation.
9 
10   This program is distributed in the hope it will be useful, but WITHOUT
11   ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12   FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13   more details.
14 
15   You should have received a copy of the GNU General Public License along with
16   this program; if not, write to the Free Software Foundation, Inc.,
17   51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
18 
19   The full GNU General Public License is included in this distribution in
20   the file called "COPYING".
21 
22   Contact Information:
23   e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
24   Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
25 
26 *******************************************************************************/
27 
28 #include <linux/types.h>
29 #include <linux/module.h>
30 #include <linux/pci.h>
31 #include <linux/netdevice.h>
32 #include <linux/vmalloc.h>
33 #include <linux/string.h>
34 #include <linux/in.h>
35 #include <linux/interrupt.h>
36 #include <linux/ip.h>
37 #include <linux/tcp.h>
38 #include <linux/sctp.h>
39 #include <linux/pkt_sched.h>
40 #include <linux/ipv6.h>
41 #include <linux/slab.h>
42 #include <net/checksum.h>
43 #include <net/ip6_checksum.h>
44 #include <linux/ethtool.h>
45 #include <linux/if.h>
46 #include <linux/if_vlan.h>
47 #include <linux/if_bridge.h>
48 #include <linux/prefetch.h>
49 #include <scsi/fc/fc_fcoe.h>
50 
51 #include "ixgbe.h"
52 #include "ixgbe_common.h"
53 #include "ixgbe_dcb_82599.h"
54 #include "ixgbe_sriov.h"
55 
56 char ixgbe_driver_name[] = "ixgbe";
57 static const char ixgbe_driver_string[] =
58 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
59 #ifdef IXGBE_FCOE
60 char ixgbe_default_device_descr[] =
61 			      "Intel(R) 10 Gigabit Network Connection";
62 #else
63 static char ixgbe_default_device_descr[] =
64 			      "Intel(R) 10 Gigabit Network Connection";
65 #endif
66 #define DRV_VERSION "3.11.33-k"
67 const char ixgbe_driver_version[] = DRV_VERSION;
68 static const char ixgbe_copyright[] =
69 				"Copyright (c) 1999-2012 Intel Corporation.";
70 
71 static const struct ixgbe_info *ixgbe_info_tbl[] = {
72 	[board_82598] = &ixgbe_82598_info,
73 	[board_82599] = &ixgbe_82599_info,
74 	[board_X540] = &ixgbe_X540_info,
75 };
76 
77 /* ixgbe_pci_tbl - PCI Device ID Table
78  *
79  * Wildcard entries (PCI_ANY_ID) should come last
80  * Last entry must be all 0s
81  *
82  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
83  *   Class, Class Mask, private data (not used) }
84  */
85 static DEFINE_PCI_DEVICE_TABLE(ixgbe_pci_tbl) = {
86 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
87 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
88 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
89 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
90 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
91 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
92 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
93 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
94 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
95 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
96 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
97 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
98 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
99 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
100 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
101 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
102 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
103 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
104 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
105 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
115 	/* required last entry */
116 	{0, }
117 };
118 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
119 
120 #ifdef CONFIG_IXGBE_DCA
121 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
122 			    void *p);
123 static struct notifier_block dca_notifier = {
124 	.notifier_call = ixgbe_notify_dca,
125 	.next          = NULL,
126 	.priority      = 0
127 };
128 #endif
129 
130 #ifdef CONFIG_PCI_IOV
131 static unsigned int max_vfs;
132 module_param(max_vfs, uint, 0);
133 MODULE_PARM_DESC(max_vfs,
134 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63");
135 #endif /* CONFIG_PCI_IOV */
136 
137 static unsigned int allow_unsupported_sfp;
138 module_param(allow_unsupported_sfp, uint, 0);
139 MODULE_PARM_DESC(allow_unsupported_sfp,
140 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
141 
142 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
143 static int debug = -1;
144 module_param(debug, int, 0);
145 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
146 
147 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
148 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
149 MODULE_LICENSE("GPL");
150 MODULE_VERSION(DRV_VERSION);
151 
152 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
153 {
154 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
155 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
156 		schedule_work(&adapter->service_task);
157 }
158 
159 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
160 {
161 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
162 
163 	/* flush memory to make sure state is correct before next watchdog */
164 	smp_mb__before_clear_bit();
165 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
166 }
167 
168 struct ixgbe_reg_info {
169 	u32 ofs;
170 	char *name;
171 };
172 
173 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
174 
175 	/* General Registers */
176 	{IXGBE_CTRL, "CTRL"},
177 	{IXGBE_STATUS, "STATUS"},
178 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
179 
180 	/* Interrupt Registers */
181 	{IXGBE_EICR, "EICR"},
182 
183 	/* RX Registers */
184 	{IXGBE_SRRCTL(0), "SRRCTL"},
185 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
186 	{IXGBE_RDLEN(0), "RDLEN"},
187 	{IXGBE_RDH(0), "RDH"},
188 	{IXGBE_RDT(0), "RDT"},
189 	{IXGBE_RXDCTL(0), "RXDCTL"},
190 	{IXGBE_RDBAL(0), "RDBAL"},
191 	{IXGBE_RDBAH(0), "RDBAH"},
192 
193 	/* TX Registers */
194 	{IXGBE_TDBAL(0), "TDBAL"},
195 	{IXGBE_TDBAH(0), "TDBAH"},
196 	{IXGBE_TDLEN(0), "TDLEN"},
197 	{IXGBE_TDH(0), "TDH"},
198 	{IXGBE_TDT(0), "TDT"},
199 	{IXGBE_TXDCTL(0), "TXDCTL"},
200 
201 	/* List Terminator */
202 	{}
203 };
204 
205 
206 /*
207  * ixgbe_regdump - register printout routine
208  */
209 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
210 {
211 	int i = 0, j = 0;
212 	char rname[16];
213 	u32 regs[64];
214 
215 	switch (reginfo->ofs) {
216 	case IXGBE_SRRCTL(0):
217 		for (i = 0; i < 64; i++)
218 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
219 		break;
220 	case IXGBE_DCA_RXCTRL(0):
221 		for (i = 0; i < 64; i++)
222 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
223 		break;
224 	case IXGBE_RDLEN(0):
225 		for (i = 0; i < 64; i++)
226 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
227 		break;
228 	case IXGBE_RDH(0):
229 		for (i = 0; i < 64; i++)
230 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
231 		break;
232 	case IXGBE_RDT(0):
233 		for (i = 0; i < 64; i++)
234 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
235 		break;
236 	case IXGBE_RXDCTL(0):
237 		for (i = 0; i < 64; i++)
238 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
239 		break;
240 	case IXGBE_RDBAL(0):
241 		for (i = 0; i < 64; i++)
242 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
243 		break;
244 	case IXGBE_RDBAH(0):
245 		for (i = 0; i < 64; i++)
246 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
247 		break;
248 	case IXGBE_TDBAL(0):
249 		for (i = 0; i < 64; i++)
250 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
251 		break;
252 	case IXGBE_TDBAH(0):
253 		for (i = 0; i < 64; i++)
254 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
255 		break;
256 	case IXGBE_TDLEN(0):
257 		for (i = 0; i < 64; i++)
258 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
259 		break;
260 	case IXGBE_TDH(0):
261 		for (i = 0; i < 64; i++)
262 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
263 		break;
264 	case IXGBE_TDT(0):
265 		for (i = 0; i < 64; i++)
266 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
267 		break;
268 	case IXGBE_TXDCTL(0):
269 		for (i = 0; i < 64; i++)
270 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
271 		break;
272 	default:
273 		pr_info("%-15s %08x\n", reginfo->name,
274 			IXGBE_READ_REG(hw, reginfo->ofs));
275 		return;
276 	}
277 
278 	for (i = 0; i < 8; i++) {
279 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i*8, i*8+7);
280 		pr_err("%-15s", rname);
281 		for (j = 0; j < 8; j++)
282 			pr_cont(" %08x", regs[i*8+j]);
283 		pr_cont("\n");
284 	}
285 
286 }
287 
288 /*
289  * ixgbe_dump - Print registers, tx-rings and rx-rings
290  */
291 static void ixgbe_dump(struct ixgbe_adapter *adapter)
292 {
293 	struct net_device *netdev = adapter->netdev;
294 	struct ixgbe_hw *hw = &adapter->hw;
295 	struct ixgbe_reg_info *reginfo;
296 	int n = 0;
297 	struct ixgbe_ring *tx_ring;
298 	struct ixgbe_tx_buffer *tx_buffer;
299 	union ixgbe_adv_tx_desc *tx_desc;
300 	struct my_u0 { u64 a; u64 b; } *u0;
301 	struct ixgbe_ring *rx_ring;
302 	union ixgbe_adv_rx_desc *rx_desc;
303 	struct ixgbe_rx_buffer *rx_buffer_info;
304 	u32 staterr;
305 	int i = 0;
306 
307 	if (!netif_msg_hw(adapter))
308 		return;
309 
310 	/* Print netdevice Info */
311 	if (netdev) {
312 		dev_info(&adapter->pdev->dev, "Net device Info\n");
313 		pr_info("Device Name     state            "
314 			"trans_start      last_rx\n");
315 		pr_info("%-15s %016lX %016lX %016lX\n",
316 			netdev->name,
317 			netdev->state,
318 			netdev->trans_start,
319 			netdev->last_rx);
320 	}
321 
322 	/* Print Registers */
323 	dev_info(&adapter->pdev->dev, "Register Dump\n");
324 	pr_info(" Register Name   Value\n");
325 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
326 	     reginfo->name; reginfo++) {
327 		ixgbe_regdump(hw, reginfo);
328 	}
329 
330 	/* Print TX Ring Summary */
331 	if (!netdev || !netif_running(netdev))
332 		goto exit;
333 
334 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
335 	pr_info(" %s     %s              %s        %s\n",
336 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
337 		"leng", "ntw", "timestamp");
338 	for (n = 0; n < adapter->num_tx_queues; n++) {
339 		tx_ring = adapter->tx_ring[n];
340 		tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
341 		pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
342 			   n, tx_ring->next_to_use, tx_ring->next_to_clean,
343 			   (u64)dma_unmap_addr(tx_buffer, dma),
344 			   dma_unmap_len(tx_buffer, len),
345 			   tx_buffer->next_to_watch,
346 			   (u64)tx_buffer->time_stamp);
347 	}
348 
349 	/* Print TX Rings */
350 	if (!netif_msg_tx_done(adapter))
351 		goto rx_ring_summary;
352 
353 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
354 
355 	/* Transmit Descriptor Formats
356 	 *
357 	 * 82598 Advanced Transmit Descriptor
358 	 *   +--------------------------------------------------------------+
359 	 * 0 |         Buffer Address [63:0]                                |
360 	 *   +--------------------------------------------------------------+
361 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
362 	 *   +--------------------------------------------------------------+
363 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
364 	 *
365 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
366 	 *   +--------------------------------------------------------------+
367 	 * 0 |                          RSV [63:0]                          |
368 	 *   +--------------------------------------------------------------+
369 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
370 	 *   +--------------------------------------------------------------+
371 	 *   63                       36 35   32 31                         0
372 	 *
373 	 * 82599+ Advanced Transmit Descriptor
374 	 *   +--------------------------------------------------------------+
375 	 * 0 |         Buffer Address [63:0]                                |
376 	 *   +--------------------------------------------------------------+
377 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
378 	 *   +--------------------------------------------------------------+
379 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
380 	 *
381 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
382 	 *   +--------------------------------------------------------------+
383 	 * 0 |                          RSV [63:0]                          |
384 	 *   +--------------------------------------------------------------+
385 	 * 8 |            RSV           |  STA  |           RSV             |
386 	 *   +--------------------------------------------------------------+
387 	 *   63                       36 35   32 31                         0
388 	 */
389 
390 	for (n = 0; n < adapter->num_tx_queues; n++) {
391 		tx_ring = adapter->tx_ring[n];
392 		pr_info("------------------------------------\n");
393 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
394 		pr_info("------------------------------------\n");
395 		pr_info("%s%s    %s              %s        %s          %s\n",
396 			"T [desc]     [address 63:0  ] ",
397 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
398 			"leng", "ntw", "timestamp", "bi->skb");
399 
400 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
401 			tx_desc = IXGBE_TX_DESC(tx_ring, i);
402 			tx_buffer = &tx_ring->tx_buffer_info[i];
403 			u0 = (struct my_u0 *)tx_desc;
404 			if (dma_unmap_len(tx_buffer, len) > 0) {
405 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p",
406 					i,
407 					le64_to_cpu(u0->a),
408 					le64_to_cpu(u0->b),
409 					(u64)dma_unmap_addr(tx_buffer, dma),
410 					dma_unmap_len(tx_buffer, len),
411 					tx_buffer->next_to_watch,
412 					(u64)tx_buffer->time_stamp,
413 					tx_buffer->skb);
414 				if (i == tx_ring->next_to_use &&
415 					i == tx_ring->next_to_clean)
416 					pr_cont(" NTC/U\n");
417 				else if (i == tx_ring->next_to_use)
418 					pr_cont(" NTU\n");
419 				else if (i == tx_ring->next_to_clean)
420 					pr_cont(" NTC\n");
421 				else
422 					pr_cont("\n");
423 
424 				if (netif_msg_pktdata(adapter) &&
425 				    tx_buffer->skb)
426 					print_hex_dump(KERN_INFO, "",
427 						DUMP_PREFIX_ADDRESS, 16, 1,
428 						tx_buffer->skb->data,
429 						dma_unmap_len(tx_buffer, len),
430 						true);
431 			}
432 		}
433 	}
434 
435 	/* Print RX Rings Summary */
436 rx_ring_summary:
437 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
438 	pr_info("Queue [NTU] [NTC]\n");
439 	for (n = 0; n < adapter->num_rx_queues; n++) {
440 		rx_ring = adapter->rx_ring[n];
441 		pr_info("%5d %5X %5X\n",
442 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
443 	}
444 
445 	/* Print RX Rings */
446 	if (!netif_msg_rx_status(adapter))
447 		goto exit;
448 
449 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
450 
451 	/* Receive Descriptor Formats
452 	 *
453 	 * 82598 Advanced Receive Descriptor (Read) Format
454 	 *    63                                           1        0
455 	 *    +-----------------------------------------------------+
456 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
457 	 *    +----------------------------------------------+------+
458 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
459 	 *    +-----------------------------------------------------+
460 	 *
461 	 *
462 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
463 	 *
464 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
465 	 *   +------------------------------------------------------+
466 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
467 	 *   | Packet   | IP     |   |          |     | Type | Type |
468 	 *   | Checksum | Ident  |   |          |     |      |      |
469 	 *   +------------------------------------------------------+
470 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
471 	 *   +------------------------------------------------------+
472 	 *   63       48 47    32 31            20 19               0
473 	 *
474 	 * 82599+ Advanced Receive Descriptor (Read) Format
475 	 *    63                                           1        0
476 	 *    +-----------------------------------------------------+
477 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
478 	 *    +----------------------------------------------+------+
479 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
480 	 *    +-----------------------------------------------------+
481 	 *
482 	 *
483 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
484 	 *
485 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
486 	 *   +------------------------------------------------------+
487 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
488 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
489 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
490 	 *   +------------------------------------------------------+
491 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
492 	 *   +------------------------------------------------------+
493 	 *   63       48 47    32 31          20 19                 0
494 	 */
495 
496 	for (n = 0; n < adapter->num_rx_queues; n++) {
497 		rx_ring = adapter->rx_ring[n];
498 		pr_info("------------------------------------\n");
499 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
500 		pr_info("------------------------------------\n");
501 		pr_info("%s%s%s",
502 			"R  [desc]      [ PktBuf     A0] ",
503 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
504 			"<-- Adv Rx Read format\n");
505 		pr_info("%s%s%s",
506 			"RWB[desc]      [PcsmIpSHl PtRs] ",
507 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
508 			"<-- Adv Rx Write-Back format\n");
509 
510 		for (i = 0; i < rx_ring->count; i++) {
511 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
512 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
513 			u0 = (struct my_u0 *)rx_desc;
514 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
515 			if (staterr & IXGBE_RXD_STAT_DD) {
516 				/* Descriptor Done */
517 				pr_info("RWB[0x%03X]     %016llX "
518 					"%016llX ---------------- %p", i,
519 					le64_to_cpu(u0->a),
520 					le64_to_cpu(u0->b),
521 					rx_buffer_info->skb);
522 			} else {
523 				pr_info("R  [0x%03X]     %016llX "
524 					"%016llX %016llX %p", i,
525 					le64_to_cpu(u0->a),
526 					le64_to_cpu(u0->b),
527 					(u64)rx_buffer_info->dma,
528 					rx_buffer_info->skb);
529 
530 				if (netif_msg_pktdata(adapter) &&
531 				    rx_buffer_info->dma) {
532 					print_hex_dump(KERN_INFO, "",
533 					   DUMP_PREFIX_ADDRESS, 16, 1,
534 					   page_address(rx_buffer_info->page) +
535 						    rx_buffer_info->page_offset,
536 					   ixgbe_rx_bufsz(rx_ring), true);
537 				}
538 			}
539 
540 			if (i == rx_ring->next_to_use)
541 				pr_cont(" NTU\n");
542 			else if (i == rx_ring->next_to_clean)
543 				pr_cont(" NTC\n");
544 			else
545 				pr_cont("\n");
546 
547 		}
548 	}
549 
550 exit:
551 	return;
552 }
553 
554 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
555 {
556 	u32 ctrl_ext;
557 
558 	/* Let firmware take over control of h/w */
559 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
560 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
561 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
562 }
563 
564 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
565 {
566 	u32 ctrl_ext;
567 
568 	/* Let firmware know the driver has taken over */
569 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
570 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
571 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
572 }
573 
574 /**
575  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
576  * @adapter: pointer to adapter struct
577  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
578  * @queue: queue to map the corresponding interrupt to
579  * @msix_vector: the vector to map to the corresponding queue
580  *
581  */
582 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
583 			   u8 queue, u8 msix_vector)
584 {
585 	u32 ivar, index;
586 	struct ixgbe_hw *hw = &adapter->hw;
587 	switch (hw->mac.type) {
588 	case ixgbe_mac_82598EB:
589 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
590 		if (direction == -1)
591 			direction = 0;
592 		index = (((direction * 64) + queue) >> 2) & 0x1F;
593 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
594 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
595 		ivar |= (msix_vector << (8 * (queue & 0x3)));
596 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
597 		break;
598 	case ixgbe_mac_82599EB:
599 	case ixgbe_mac_X540:
600 		if (direction == -1) {
601 			/* other causes */
602 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
603 			index = ((queue & 1) * 8);
604 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
605 			ivar &= ~(0xFF << index);
606 			ivar |= (msix_vector << index);
607 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
608 			break;
609 		} else {
610 			/* tx or rx causes */
611 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
612 			index = ((16 * (queue & 1)) + (8 * direction));
613 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
614 			ivar &= ~(0xFF << index);
615 			ivar |= (msix_vector << index);
616 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
617 			break;
618 		}
619 	default:
620 		break;
621 	}
622 }
623 
624 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
625 					  u64 qmask)
626 {
627 	u32 mask;
628 
629 	switch (adapter->hw.mac.type) {
630 	case ixgbe_mac_82598EB:
631 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
632 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
633 		break;
634 	case ixgbe_mac_82599EB:
635 	case ixgbe_mac_X540:
636 		mask = (qmask & 0xFFFFFFFF);
637 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
638 		mask = (qmask >> 32);
639 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
640 		break;
641 	default:
642 		break;
643 	}
644 }
645 
646 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring *ring,
647 				      struct ixgbe_tx_buffer *tx_buffer)
648 {
649 	if (tx_buffer->skb) {
650 		dev_kfree_skb_any(tx_buffer->skb);
651 		if (dma_unmap_len(tx_buffer, len))
652 			dma_unmap_single(ring->dev,
653 					 dma_unmap_addr(tx_buffer, dma),
654 					 dma_unmap_len(tx_buffer, len),
655 					 DMA_TO_DEVICE);
656 	} else if (dma_unmap_len(tx_buffer, len)) {
657 		dma_unmap_page(ring->dev,
658 			       dma_unmap_addr(tx_buffer, dma),
659 			       dma_unmap_len(tx_buffer, len),
660 			       DMA_TO_DEVICE);
661 	}
662 	tx_buffer->next_to_watch = NULL;
663 	tx_buffer->skb = NULL;
664 	dma_unmap_len_set(tx_buffer, len, 0);
665 	/* tx_buffer must be completely set up in the transmit path */
666 }
667 
668 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
669 {
670 	struct ixgbe_hw *hw = &adapter->hw;
671 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
672 	int i;
673 	u32 data;
674 
675 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
676 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
677 		return;
678 
679 	switch (hw->mac.type) {
680 	case ixgbe_mac_82598EB:
681 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
682 		break;
683 	default:
684 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
685 	}
686 	hwstats->lxoffrxc += data;
687 
688 	/* refill credits (no tx hang) if we received xoff */
689 	if (!data)
690 		return;
691 
692 	for (i = 0; i < adapter->num_tx_queues; i++)
693 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
694 			  &adapter->tx_ring[i]->state);
695 }
696 
697 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
698 {
699 	struct ixgbe_hw *hw = &adapter->hw;
700 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
701 	u32 xoff[8] = {0};
702 	u8 tc;
703 	int i;
704 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
705 
706 	if (adapter->ixgbe_ieee_pfc)
707 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
708 
709 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
710 		ixgbe_update_xoff_rx_lfc(adapter);
711 		return;
712 	}
713 
714 	/* update stats for each tc, only valid with PFC enabled */
715 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
716 		u32 pxoffrxc;
717 
718 		switch (hw->mac.type) {
719 		case ixgbe_mac_82598EB:
720 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
721 			break;
722 		default:
723 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
724 		}
725 		hwstats->pxoffrxc[i] += pxoffrxc;
726 		/* Get the TC for given UP */
727 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
728 		xoff[tc] += pxoffrxc;
729 	}
730 
731 	/* disarm tx queues that have received xoff frames */
732 	for (i = 0; i < adapter->num_tx_queues; i++) {
733 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
734 
735 		tc = tx_ring->dcb_tc;
736 		if (xoff[tc])
737 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
738 	}
739 }
740 
741 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
742 {
743 	return ring->stats.packets;
744 }
745 
746 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
747 {
748 	struct ixgbe_adapter *adapter = netdev_priv(ring->netdev);
749 	struct ixgbe_hw *hw = &adapter->hw;
750 
751 	u32 head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
752 	u32 tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
753 
754 	if (head != tail)
755 		return (head < tail) ?
756 			tail - head : (tail + ring->count - head);
757 
758 	return 0;
759 }
760 
761 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
762 {
763 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
764 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
765 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
766 	bool ret = false;
767 
768 	clear_check_for_tx_hang(tx_ring);
769 
770 	/*
771 	 * Check for a hung queue, but be thorough. This verifies
772 	 * that a transmit has been completed since the previous
773 	 * check AND there is at least one packet pending. The
774 	 * ARMED bit is set to indicate a potential hang. The
775 	 * bit is cleared if a pause frame is received to remove
776 	 * false hang detection due to PFC or 802.3x frames. By
777 	 * requiring this to fail twice we avoid races with
778 	 * pfc clearing the ARMED bit and conditions where we
779 	 * run the check_tx_hang logic with a transmit completion
780 	 * pending but without time to complete it yet.
781 	 */
782 	if ((tx_done_old == tx_done) && tx_pending) {
783 		/* make sure it is true for two checks in a row */
784 		ret = test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
785 				       &tx_ring->state);
786 	} else {
787 		/* update completed stats and continue */
788 		tx_ring->tx_stats.tx_done_old = tx_done;
789 		/* reset the countdown */
790 		clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
791 	}
792 
793 	return ret;
794 }
795 
796 /**
797  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
798  * @adapter: driver private struct
799  **/
800 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
801 {
802 
803 	/* Do the reset outside of interrupt context */
804 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
805 		adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
806 		ixgbe_service_event_schedule(adapter);
807 	}
808 }
809 
810 /**
811  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
812  * @q_vector: structure containing interrupt and ring information
813  * @tx_ring: tx ring to clean
814  **/
815 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
816 			       struct ixgbe_ring *tx_ring)
817 {
818 	struct ixgbe_adapter *adapter = q_vector->adapter;
819 	struct ixgbe_tx_buffer *tx_buffer;
820 	union ixgbe_adv_tx_desc *tx_desc;
821 	unsigned int total_bytes = 0, total_packets = 0;
822 	unsigned int budget = q_vector->tx.work_limit;
823 	unsigned int i = tx_ring->next_to_clean;
824 
825 	if (test_bit(__IXGBE_DOWN, &adapter->state))
826 		return true;
827 
828 	tx_buffer = &tx_ring->tx_buffer_info[i];
829 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
830 	i -= tx_ring->count;
831 
832 	do {
833 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
834 
835 		/* if next_to_watch is not set then there is no work pending */
836 		if (!eop_desc)
837 			break;
838 
839 		/* prevent any other reads prior to eop_desc */
840 		rmb();
841 
842 		/* if DD is not set pending work has not been completed */
843 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
844 			break;
845 
846 		/* clear next_to_watch to prevent false hangs */
847 		tx_buffer->next_to_watch = NULL;
848 
849 		/* update the statistics for this packet */
850 		total_bytes += tx_buffer->bytecount;
851 		total_packets += tx_buffer->gso_segs;
852 
853 		if (unlikely(tx_buffer->tx_flags & IXGBE_TX_FLAGS_TSTAMP))
854 			ixgbe_ptp_tx_hwtstamp(q_vector, tx_buffer->skb);
855 
856 		/* free the skb */
857 		dev_kfree_skb_any(tx_buffer->skb);
858 
859 		/* unmap skb header data */
860 		dma_unmap_single(tx_ring->dev,
861 				 dma_unmap_addr(tx_buffer, dma),
862 				 dma_unmap_len(tx_buffer, len),
863 				 DMA_TO_DEVICE);
864 
865 		/* clear tx_buffer data */
866 		tx_buffer->skb = NULL;
867 		dma_unmap_len_set(tx_buffer, len, 0);
868 
869 		/* unmap remaining buffers */
870 		while (tx_desc != eop_desc) {
871 			tx_buffer++;
872 			tx_desc++;
873 			i++;
874 			if (unlikely(!i)) {
875 				i -= tx_ring->count;
876 				tx_buffer = tx_ring->tx_buffer_info;
877 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
878 			}
879 
880 			/* unmap any remaining paged data */
881 			if (dma_unmap_len(tx_buffer, len)) {
882 				dma_unmap_page(tx_ring->dev,
883 					       dma_unmap_addr(tx_buffer, dma),
884 					       dma_unmap_len(tx_buffer, len),
885 					       DMA_TO_DEVICE);
886 				dma_unmap_len_set(tx_buffer, len, 0);
887 			}
888 		}
889 
890 		/* move us one more past the eop_desc for start of next pkt */
891 		tx_buffer++;
892 		tx_desc++;
893 		i++;
894 		if (unlikely(!i)) {
895 			i -= tx_ring->count;
896 			tx_buffer = tx_ring->tx_buffer_info;
897 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
898 		}
899 
900 		/* issue prefetch for next Tx descriptor */
901 		prefetch(tx_desc);
902 
903 		/* update budget accounting */
904 		budget--;
905 	} while (likely(budget));
906 
907 	i += tx_ring->count;
908 	tx_ring->next_to_clean = i;
909 	u64_stats_update_begin(&tx_ring->syncp);
910 	tx_ring->stats.bytes += total_bytes;
911 	tx_ring->stats.packets += total_packets;
912 	u64_stats_update_end(&tx_ring->syncp);
913 	q_vector->tx.total_bytes += total_bytes;
914 	q_vector->tx.total_packets += total_packets;
915 
916 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
917 		/* schedule immediate reset if we believe we hung */
918 		struct ixgbe_hw *hw = &adapter->hw;
919 		e_err(drv, "Detected Tx Unit Hang\n"
920 			"  Tx Queue             <%d>\n"
921 			"  TDH, TDT             <%x>, <%x>\n"
922 			"  next_to_use          <%x>\n"
923 			"  next_to_clean        <%x>\n"
924 			"tx_buffer_info[next_to_clean]\n"
925 			"  time_stamp           <%lx>\n"
926 			"  jiffies              <%lx>\n",
927 			tx_ring->queue_index,
928 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
929 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
930 			tx_ring->next_to_use, i,
931 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
932 
933 		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
934 
935 		e_info(probe,
936 		       "tx hang %d detected on queue %d, resetting adapter\n",
937 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
938 
939 		/* schedule immediate reset if we believe we hung */
940 		ixgbe_tx_timeout_reset(adapter);
941 
942 		/* the adapter is about to reset, no point in enabling stuff */
943 		return true;
944 	}
945 
946 	netdev_tx_completed_queue(txring_txq(tx_ring),
947 				  total_packets, total_bytes);
948 
949 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
950 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
951 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
952 		/* Make sure that anybody stopping the queue after this
953 		 * sees the new next_to_clean.
954 		 */
955 		smp_mb();
956 		if (__netif_subqueue_stopped(tx_ring->netdev,
957 					     tx_ring->queue_index)
958 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
959 			netif_wake_subqueue(tx_ring->netdev,
960 					    tx_ring->queue_index);
961 			++tx_ring->tx_stats.restart_queue;
962 		}
963 	}
964 
965 	return !!budget;
966 }
967 
968 #ifdef CONFIG_IXGBE_DCA
969 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
970 				struct ixgbe_ring *tx_ring,
971 				int cpu)
972 {
973 	struct ixgbe_hw *hw = &adapter->hw;
974 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
975 	u16 reg_offset;
976 
977 	switch (hw->mac.type) {
978 	case ixgbe_mac_82598EB:
979 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
980 		break;
981 	case ixgbe_mac_82599EB:
982 	case ixgbe_mac_X540:
983 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
984 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
985 		break;
986 	default:
987 		/* for unknown hardware do not write register */
988 		return;
989 	}
990 
991 	/*
992 	 * We can enable relaxed ordering for reads, but not writes when
993 	 * DCA is enabled.  This is due to a known issue in some chipsets
994 	 * which will cause the DCA tag to be cleared.
995 	 */
996 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
997 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
998 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
999 
1000 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1001 }
1002 
1003 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1004 				struct ixgbe_ring *rx_ring,
1005 				int cpu)
1006 {
1007 	struct ixgbe_hw *hw = &adapter->hw;
1008 	u32 rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1009 	u8 reg_idx = rx_ring->reg_idx;
1010 
1011 
1012 	switch (hw->mac.type) {
1013 	case ixgbe_mac_82599EB:
1014 	case ixgbe_mac_X540:
1015 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1016 		break;
1017 	default:
1018 		break;
1019 	}
1020 
1021 	/*
1022 	 * We can enable relaxed ordering for reads, but not writes when
1023 	 * DCA is enabled.  This is due to a known issue in some chipsets
1024 	 * which will cause the DCA tag to be cleared.
1025 	 */
1026 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1027 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1028 
1029 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1030 }
1031 
1032 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1033 {
1034 	struct ixgbe_adapter *adapter = q_vector->adapter;
1035 	struct ixgbe_ring *ring;
1036 	int cpu = get_cpu();
1037 
1038 	if (q_vector->cpu == cpu)
1039 		goto out_no_update;
1040 
1041 	ixgbe_for_each_ring(ring, q_vector->tx)
1042 		ixgbe_update_tx_dca(adapter, ring, cpu);
1043 
1044 	ixgbe_for_each_ring(ring, q_vector->rx)
1045 		ixgbe_update_rx_dca(adapter, ring, cpu);
1046 
1047 	q_vector->cpu = cpu;
1048 out_no_update:
1049 	put_cpu();
1050 }
1051 
1052 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1053 {
1054 	int i;
1055 
1056 	if (!(adapter->flags & IXGBE_FLAG_DCA_ENABLED))
1057 		return;
1058 
1059 	/* always use CB2 mode, difference is masked in the CB driver */
1060 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 2);
1061 
1062 	for (i = 0; i < adapter->num_q_vectors; i++) {
1063 		adapter->q_vector[i]->cpu = -1;
1064 		ixgbe_update_dca(adapter->q_vector[i]);
1065 	}
1066 }
1067 
1068 static int __ixgbe_notify_dca(struct device *dev, void *data)
1069 {
1070 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1071 	unsigned long event = *(unsigned long *)data;
1072 
1073 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1074 		return 0;
1075 
1076 	switch (event) {
1077 	case DCA_PROVIDER_ADD:
1078 		/* if we're already enabled, don't do it again */
1079 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1080 			break;
1081 		if (dca_add_requester(dev) == 0) {
1082 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1083 			ixgbe_setup_dca(adapter);
1084 			break;
1085 		}
1086 		/* Fall Through since DCA is disabled. */
1087 	case DCA_PROVIDER_REMOVE:
1088 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1089 			dca_remove_requester(dev);
1090 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1091 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
1092 		}
1093 		break;
1094 	}
1095 
1096 	return 0;
1097 }
1098 
1099 #endif /* CONFIG_IXGBE_DCA */
1100 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1101 				 union ixgbe_adv_rx_desc *rx_desc,
1102 				 struct sk_buff *skb)
1103 {
1104 	if (ring->netdev->features & NETIF_F_RXHASH)
1105 		skb->rxhash = le32_to_cpu(rx_desc->wb.lower.hi_dword.rss);
1106 }
1107 
1108 #ifdef IXGBE_FCOE
1109 /**
1110  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1111  * @ring: structure containing ring specific data
1112  * @rx_desc: advanced rx descriptor
1113  *
1114  * Returns : true if it is FCoE pkt
1115  */
1116 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1117 				    union ixgbe_adv_rx_desc *rx_desc)
1118 {
1119 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1120 
1121 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1122 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1123 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1124 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1125 }
1126 
1127 #endif /* IXGBE_FCOE */
1128 /**
1129  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1130  * @ring: structure containing ring specific data
1131  * @rx_desc: current Rx descriptor being processed
1132  * @skb: skb currently being received and modified
1133  **/
1134 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1135 				     union ixgbe_adv_rx_desc *rx_desc,
1136 				     struct sk_buff *skb)
1137 {
1138 	skb_checksum_none_assert(skb);
1139 
1140 	/* Rx csum disabled */
1141 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1142 		return;
1143 
1144 	/* if IP and error */
1145 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1146 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1147 		ring->rx_stats.csum_err++;
1148 		return;
1149 	}
1150 
1151 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1152 		return;
1153 
1154 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1155 		__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1156 
1157 		/*
1158 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1159 		 * checksum errors.
1160 		 */
1161 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1162 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1163 			return;
1164 
1165 		ring->rx_stats.csum_err++;
1166 		return;
1167 	}
1168 
1169 	/* It must be a TCP or UDP packet with a valid checksum */
1170 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1171 }
1172 
1173 static inline void ixgbe_release_rx_desc(struct ixgbe_ring *rx_ring, u32 val)
1174 {
1175 	rx_ring->next_to_use = val;
1176 
1177 	/* update next to alloc since we have filled the ring */
1178 	rx_ring->next_to_alloc = val;
1179 	/*
1180 	 * Force memory writes to complete before letting h/w
1181 	 * know there are new descriptors to fetch.  (Only
1182 	 * applicable for weak-ordered memory model archs,
1183 	 * such as IA-64).
1184 	 */
1185 	wmb();
1186 	writel(val, rx_ring->tail);
1187 }
1188 
1189 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1190 				    struct ixgbe_rx_buffer *bi)
1191 {
1192 	struct page *page = bi->page;
1193 	dma_addr_t dma = bi->dma;
1194 
1195 	/* since we are recycling buffers we should seldom need to alloc */
1196 	if (likely(dma))
1197 		return true;
1198 
1199 	/* alloc new page for storage */
1200 	if (likely(!page)) {
1201 		page = __skb_alloc_pages(GFP_ATOMIC | __GFP_COLD | __GFP_COMP,
1202 					 bi->skb, ixgbe_rx_pg_order(rx_ring));
1203 		if (unlikely(!page)) {
1204 			rx_ring->rx_stats.alloc_rx_page_failed++;
1205 			return false;
1206 		}
1207 		bi->page = page;
1208 	}
1209 
1210 	/* map page for use */
1211 	dma = dma_map_page(rx_ring->dev, page, 0,
1212 			   ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1213 
1214 	/*
1215 	 * if mapping failed free memory back to system since
1216 	 * there isn't much point in holding memory we can't use
1217 	 */
1218 	if (dma_mapping_error(rx_ring->dev, dma)) {
1219 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1220 		bi->page = NULL;
1221 
1222 		rx_ring->rx_stats.alloc_rx_page_failed++;
1223 		return false;
1224 	}
1225 
1226 	bi->dma = dma;
1227 	bi->page_offset = 0;
1228 
1229 	return true;
1230 }
1231 
1232 /**
1233  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1234  * @rx_ring: ring to place buffers on
1235  * @cleaned_count: number of buffers to replace
1236  **/
1237 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1238 {
1239 	union ixgbe_adv_rx_desc *rx_desc;
1240 	struct ixgbe_rx_buffer *bi;
1241 	u16 i = rx_ring->next_to_use;
1242 
1243 	/* nothing to do */
1244 	if (!cleaned_count)
1245 		return;
1246 
1247 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1248 	bi = &rx_ring->rx_buffer_info[i];
1249 	i -= rx_ring->count;
1250 
1251 	do {
1252 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1253 			break;
1254 
1255 		/*
1256 		 * Refresh the desc even if buffer_addrs didn't change
1257 		 * because each write-back erases this info.
1258 		 */
1259 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1260 
1261 		rx_desc++;
1262 		bi++;
1263 		i++;
1264 		if (unlikely(!i)) {
1265 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1266 			bi = rx_ring->rx_buffer_info;
1267 			i -= rx_ring->count;
1268 		}
1269 
1270 		/* clear the hdr_addr for the next_to_use descriptor */
1271 		rx_desc->read.hdr_addr = 0;
1272 
1273 		cleaned_count--;
1274 	} while (cleaned_count);
1275 
1276 	i += rx_ring->count;
1277 
1278 	if (rx_ring->next_to_use != i)
1279 		ixgbe_release_rx_desc(rx_ring, i);
1280 }
1281 
1282 /**
1283  * ixgbe_get_headlen - determine size of header for RSC/LRO/GRO/FCOE
1284  * @data: pointer to the start of the headers
1285  * @max_len: total length of section to find headers in
1286  *
1287  * This function is meant to determine the length of headers that will
1288  * be recognized by hardware for LRO, GRO, and RSC offloads.  The main
1289  * motivation of doing this is to only perform one pull for IPv4 TCP
1290  * packets so that we can do basic things like calculating the gso_size
1291  * based on the average data per packet.
1292  **/
1293 static unsigned int ixgbe_get_headlen(unsigned char *data,
1294 				      unsigned int max_len)
1295 {
1296 	union {
1297 		unsigned char *network;
1298 		/* l2 headers */
1299 		struct ethhdr *eth;
1300 		struct vlan_hdr *vlan;
1301 		/* l3 headers */
1302 		struct iphdr *ipv4;
1303 		struct ipv6hdr *ipv6;
1304 	} hdr;
1305 	__be16 protocol;
1306 	u8 nexthdr = 0;	/* default to not TCP */
1307 	u8 hlen;
1308 
1309 	/* this should never happen, but better safe than sorry */
1310 	if (max_len < ETH_HLEN)
1311 		return max_len;
1312 
1313 	/* initialize network frame pointer */
1314 	hdr.network = data;
1315 
1316 	/* set first protocol and move network header forward */
1317 	protocol = hdr.eth->h_proto;
1318 	hdr.network += ETH_HLEN;
1319 
1320 	/* handle any vlan tag if present */
1321 	if (protocol == __constant_htons(ETH_P_8021Q)) {
1322 		if ((hdr.network - data) > (max_len - VLAN_HLEN))
1323 			return max_len;
1324 
1325 		protocol = hdr.vlan->h_vlan_encapsulated_proto;
1326 		hdr.network += VLAN_HLEN;
1327 	}
1328 
1329 	/* handle L3 protocols */
1330 	if (protocol == __constant_htons(ETH_P_IP)) {
1331 		if ((hdr.network - data) > (max_len - sizeof(struct iphdr)))
1332 			return max_len;
1333 
1334 		/* access ihl as a u8 to avoid unaligned access on ia64 */
1335 		hlen = (hdr.network[0] & 0x0F) << 2;
1336 
1337 		/* verify hlen meets minimum size requirements */
1338 		if (hlen < sizeof(struct iphdr))
1339 			return hdr.network - data;
1340 
1341 		/* record next protocol if header is present */
1342 		if (!hdr.ipv4->frag_off)
1343 			nexthdr = hdr.ipv4->protocol;
1344 	} else if (protocol == __constant_htons(ETH_P_IPV6)) {
1345 		if ((hdr.network - data) > (max_len - sizeof(struct ipv6hdr)))
1346 			return max_len;
1347 
1348 		/* record next protocol */
1349 		nexthdr = hdr.ipv6->nexthdr;
1350 		hlen = sizeof(struct ipv6hdr);
1351 #ifdef IXGBE_FCOE
1352 	} else if (protocol == __constant_htons(ETH_P_FCOE)) {
1353 		if ((hdr.network - data) > (max_len - FCOE_HEADER_LEN))
1354 			return max_len;
1355 		hlen = FCOE_HEADER_LEN;
1356 #endif
1357 	} else {
1358 		return hdr.network - data;
1359 	}
1360 
1361 	/* relocate pointer to start of L4 header */
1362 	hdr.network += hlen;
1363 
1364 	/* finally sort out TCP/UDP */
1365 	if (nexthdr == IPPROTO_TCP) {
1366 		if ((hdr.network - data) > (max_len - sizeof(struct tcphdr)))
1367 			return max_len;
1368 
1369 		/* access doff as a u8 to avoid unaligned access on ia64 */
1370 		hlen = (hdr.network[12] & 0xF0) >> 2;
1371 
1372 		/* verify hlen meets minimum size requirements */
1373 		if (hlen < sizeof(struct tcphdr))
1374 			return hdr.network - data;
1375 
1376 		hdr.network += hlen;
1377 	} else if (nexthdr == IPPROTO_UDP) {
1378 		if ((hdr.network - data) > (max_len - sizeof(struct udphdr)))
1379 			return max_len;
1380 
1381 		hdr.network += sizeof(struct udphdr);
1382 	}
1383 
1384 	/*
1385 	 * If everything has gone correctly hdr.network should be the
1386 	 * data section of the packet and will be the end of the header.
1387 	 * If not then it probably represents the end of the last recognized
1388 	 * header.
1389 	 */
1390 	if ((hdr.network - data) < max_len)
1391 		return hdr.network - data;
1392 	else
1393 		return max_len;
1394 }
1395 
1396 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1397 				   struct sk_buff *skb)
1398 {
1399 	u16 hdr_len = skb_headlen(skb);
1400 
1401 	/* set gso_size to avoid messing up TCP MSS */
1402 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1403 						 IXGBE_CB(skb)->append_cnt);
1404 }
1405 
1406 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1407 				   struct sk_buff *skb)
1408 {
1409 	/* if append_cnt is 0 then frame is not RSC */
1410 	if (!IXGBE_CB(skb)->append_cnt)
1411 		return;
1412 
1413 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1414 	rx_ring->rx_stats.rsc_flush++;
1415 
1416 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1417 
1418 	/* gso_size is computed using append_cnt so always clear it last */
1419 	IXGBE_CB(skb)->append_cnt = 0;
1420 }
1421 
1422 /**
1423  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1424  * @rx_ring: rx descriptor ring packet is being transacted on
1425  * @rx_desc: pointer to the EOP Rx descriptor
1426  * @skb: pointer to current skb being populated
1427  *
1428  * This function checks the ring, descriptor, and packet information in
1429  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1430  * other fields within the skb.
1431  **/
1432 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1433 				     union ixgbe_adv_rx_desc *rx_desc,
1434 				     struct sk_buff *skb)
1435 {
1436 	struct net_device *dev = rx_ring->netdev;
1437 
1438 	ixgbe_update_rsc_stats(rx_ring, skb);
1439 
1440 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1441 
1442 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1443 
1444 	ixgbe_ptp_rx_hwtstamp(rx_ring->q_vector, rx_desc, skb);
1445 
1446 	if ((dev->features & NETIF_F_HW_VLAN_RX) &&
1447 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1448 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1449 		__vlan_hwaccel_put_tag(skb, vid);
1450 	}
1451 
1452 	skb_record_rx_queue(skb, rx_ring->queue_index);
1453 
1454 	skb->protocol = eth_type_trans(skb, dev);
1455 }
1456 
1457 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1458 			 struct sk_buff *skb)
1459 {
1460 	struct ixgbe_adapter *adapter = q_vector->adapter;
1461 
1462 	if (!(adapter->flags & IXGBE_FLAG_IN_NETPOLL))
1463 		napi_gro_receive(&q_vector->napi, skb);
1464 	else
1465 		netif_rx(skb);
1466 }
1467 
1468 /**
1469  * ixgbe_is_non_eop - process handling of non-EOP buffers
1470  * @rx_ring: Rx ring being processed
1471  * @rx_desc: Rx descriptor for current buffer
1472  * @skb: Current socket buffer containing buffer in progress
1473  *
1474  * This function updates next to clean.  If the buffer is an EOP buffer
1475  * this function exits returning false, otherwise it will place the
1476  * sk_buff in the next buffer to be chained and return true indicating
1477  * that this is in fact a non-EOP buffer.
1478  **/
1479 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1480 			     union ixgbe_adv_rx_desc *rx_desc,
1481 			     struct sk_buff *skb)
1482 {
1483 	u32 ntc = rx_ring->next_to_clean + 1;
1484 
1485 	/* fetch, update, and store next to clean */
1486 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1487 	rx_ring->next_to_clean = ntc;
1488 
1489 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1490 
1491 	/* update RSC append count if present */
1492 	if (ring_is_rsc_enabled(rx_ring)) {
1493 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1494 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1495 
1496 		if (unlikely(rsc_enabled)) {
1497 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1498 
1499 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1500 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1501 
1502 			/* update ntc based on RSC value */
1503 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1504 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1505 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1506 		}
1507 	}
1508 
1509 	/* if we are the last buffer then there is nothing else to do */
1510 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1511 		return false;
1512 
1513 	/* place skb in next buffer to be received */
1514 	rx_ring->rx_buffer_info[ntc].skb = skb;
1515 	rx_ring->rx_stats.non_eop_descs++;
1516 
1517 	return true;
1518 }
1519 
1520 /**
1521  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1522  * @rx_ring: rx descriptor ring packet is being transacted on
1523  * @skb: pointer to current skb being adjusted
1524  *
1525  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1526  * main difference between this version and the original function is that
1527  * this function can make several assumptions about the state of things
1528  * that allow for significant optimizations versus the standard function.
1529  * As a result we can do things like drop a frag and maintain an accurate
1530  * truesize for the skb.
1531  */
1532 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1533 			    struct sk_buff *skb)
1534 {
1535 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1536 	unsigned char *va;
1537 	unsigned int pull_len;
1538 
1539 	/*
1540 	 * it is valid to use page_address instead of kmap since we are
1541 	 * working with pages allocated out of the lomem pool per
1542 	 * alloc_page(GFP_ATOMIC)
1543 	 */
1544 	va = skb_frag_address(frag);
1545 
1546 	/*
1547 	 * we need the header to contain the greater of either ETH_HLEN or
1548 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1549 	 */
1550 	pull_len = ixgbe_get_headlen(va, IXGBE_RX_HDR_SIZE);
1551 
1552 	/* align pull length to size of long to optimize memcpy performance */
1553 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1554 
1555 	/* update all of the pointers */
1556 	skb_frag_size_sub(frag, pull_len);
1557 	frag->page_offset += pull_len;
1558 	skb->data_len -= pull_len;
1559 	skb->tail += pull_len;
1560 }
1561 
1562 /**
1563  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1564  * @rx_ring: rx descriptor ring packet is being transacted on
1565  * @skb: pointer to current skb being updated
1566  *
1567  * This function provides a basic DMA sync up for the first fragment of an
1568  * skb.  The reason for doing this is that the first fragment cannot be
1569  * unmapped until we have reached the end of packet descriptor for a buffer
1570  * chain.
1571  */
1572 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1573 				struct sk_buff *skb)
1574 {
1575 	/* if the page was released unmap it, else just sync our portion */
1576 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1577 		dma_unmap_page(rx_ring->dev, IXGBE_CB(skb)->dma,
1578 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1579 		IXGBE_CB(skb)->page_released = false;
1580 	} else {
1581 		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1582 
1583 		dma_sync_single_range_for_cpu(rx_ring->dev,
1584 					      IXGBE_CB(skb)->dma,
1585 					      frag->page_offset,
1586 					      ixgbe_rx_bufsz(rx_ring),
1587 					      DMA_FROM_DEVICE);
1588 	}
1589 	IXGBE_CB(skb)->dma = 0;
1590 }
1591 
1592 /**
1593  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1594  * @rx_ring: rx descriptor ring packet is being transacted on
1595  * @rx_desc: pointer to the EOP Rx descriptor
1596  * @skb: pointer to current skb being fixed
1597  *
1598  * Check for corrupted packet headers caused by senders on the local L2
1599  * embedded NIC switch not setting up their Tx Descriptors right.  These
1600  * should be very rare.
1601  *
1602  * Also address the case where we are pulling data in on pages only
1603  * and as such no data is present in the skb header.
1604  *
1605  * In addition if skb is not at least 60 bytes we need to pad it so that
1606  * it is large enough to qualify as a valid Ethernet frame.
1607  *
1608  * Returns true if an error was encountered and skb was freed.
1609  **/
1610 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1611 				  union ixgbe_adv_rx_desc *rx_desc,
1612 				  struct sk_buff *skb)
1613 {
1614 	struct net_device *netdev = rx_ring->netdev;
1615 
1616 	/* verify that the packet does not have any known errors */
1617 	if (unlikely(ixgbe_test_staterr(rx_desc,
1618 					IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1619 	    !(netdev->features & NETIF_F_RXALL))) {
1620 		dev_kfree_skb_any(skb);
1621 		return true;
1622 	}
1623 
1624 	/* place header in linear portion of buffer */
1625 	if (skb_is_nonlinear(skb))
1626 		ixgbe_pull_tail(rx_ring, skb);
1627 
1628 #ifdef IXGBE_FCOE
1629 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1630 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1631 		return false;
1632 
1633 #endif
1634 	/* if skb_pad returns an error the skb was freed */
1635 	if (unlikely(skb->len < 60)) {
1636 		int pad_len = 60 - skb->len;
1637 
1638 		if (skb_pad(skb, pad_len))
1639 			return true;
1640 		__skb_put(skb, pad_len);
1641 	}
1642 
1643 	return false;
1644 }
1645 
1646 /**
1647  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1648  * @rx_ring: rx descriptor ring to store buffers on
1649  * @old_buff: donor buffer to have page reused
1650  *
1651  * Synchronizes page for reuse by the adapter
1652  **/
1653 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1654 				struct ixgbe_rx_buffer *old_buff)
1655 {
1656 	struct ixgbe_rx_buffer *new_buff;
1657 	u16 nta = rx_ring->next_to_alloc;
1658 
1659 	new_buff = &rx_ring->rx_buffer_info[nta];
1660 
1661 	/* update, and store next to alloc */
1662 	nta++;
1663 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1664 
1665 	/* transfer page from old buffer to new buffer */
1666 	new_buff->page = old_buff->page;
1667 	new_buff->dma = old_buff->dma;
1668 	new_buff->page_offset = old_buff->page_offset;
1669 
1670 	/* sync the buffer for use by the device */
1671 	dma_sync_single_range_for_device(rx_ring->dev, new_buff->dma,
1672 					 new_buff->page_offset,
1673 					 ixgbe_rx_bufsz(rx_ring),
1674 					 DMA_FROM_DEVICE);
1675 }
1676 
1677 /**
1678  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1679  * @rx_ring: rx descriptor ring to transact packets on
1680  * @rx_buffer: buffer containing page to add
1681  * @rx_desc: descriptor containing length of buffer written by hardware
1682  * @skb: sk_buff to place the data into
1683  *
1684  * This function will add the data contained in rx_buffer->page to the skb.
1685  * This is done either through a direct copy if the data in the buffer is
1686  * less than the skb header size, otherwise it will just attach the page as
1687  * a frag to the skb.
1688  *
1689  * The function will then update the page offset if necessary and return
1690  * true if the buffer can be reused by the adapter.
1691  **/
1692 static bool ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1693 			      struct ixgbe_rx_buffer *rx_buffer,
1694 			      union ixgbe_adv_rx_desc *rx_desc,
1695 			      struct sk_buff *skb)
1696 {
1697 	struct page *page = rx_buffer->page;
1698 	unsigned int size = le16_to_cpu(rx_desc->wb.upper.length);
1699 #if (PAGE_SIZE < 8192)
1700 	unsigned int truesize = ixgbe_rx_bufsz(rx_ring);
1701 #else
1702 	unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
1703 	unsigned int last_offset = ixgbe_rx_pg_size(rx_ring) -
1704 				   ixgbe_rx_bufsz(rx_ring);
1705 #endif
1706 
1707 	if ((size <= IXGBE_RX_HDR_SIZE) && !skb_is_nonlinear(skb)) {
1708 		unsigned char *va = page_address(page) + rx_buffer->page_offset;
1709 
1710 		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
1711 
1712 		/* we can reuse buffer as-is, just make sure it is local */
1713 		if (likely(page_to_nid(page) == numa_node_id()))
1714 			return true;
1715 
1716 		/* this page cannot be reused so discard it */
1717 		put_page(page);
1718 		return false;
1719 	}
1720 
1721 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
1722 			rx_buffer->page_offset, size, truesize);
1723 
1724 	/* avoid re-using remote pages */
1725 	if (unlikely(page_to_nid(page) != numa_node_id()))
1726 		return false;
1727 
1728 #if (PAGE_SIZE < 8192)
1729 	/* if we are only owner of page we can reuse it */
1730 	if (unlikely(page_count(page) != 1))
1731 		return false;
1732 
1733 	/* flip page offset to other buffer */
1734 	rx_buffer->page_offset ^= truesize;
1735 
1736 	/*
1737 	 * since we are the only owner of the page and we need to
1738 	 * increment it, just set the value to 2 in order to avoid
1739 	 * an unecessary locked operation
1740 	 */
1741 	atomic_set(&page->_count, 2);
1742 #else
1743 	/* move offset up to the next cache line */
1744 	rx_buffer->page_offset += truesize;
1745 
1746 	if (rx_buffer->page_offset > last_offset)
1747 		return false;
1748 
1749 	/* bump ref count on page before it is given to the stack */
1750 	get_page(page);
1751 #endif
1752 
1753 	return true;
1754 }
1755 
1756 static struct sk_buff *ixgbe_fetch_rx_buffer(struct ixgbe_ring *rx_ring,
1757 					     union ixgbe_adv_rx_desc *rx_desc)
1758 {
1759 	struct ixgbe_rx_buffer *rx_buffer;
1760 	struct sk_buff *skb;
1761 	struct page *page;
1762 
1763 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
1764 	page = rx_buffer->page;
1765 	prefetchw(page);
1766 
1767 	skb = rx_buffer->skb;
1768 
1769 	if (likely(!skb)) {
1770 		void *page_addr = page_address(page) +
1771 				  rx_buffer->page_offset;
1772 
1773 		/* prefetch first cache line of first page */
1774 		prefetch(page_addr);
1775 #if L1_CACHE_BYTES < 128
1776 		prefetch(page_addr + L1_CACHE_BYTES);
1777 #endif
1778 
1779 		/* allocate a skb to store the frags */
1780 		skb = netdev_alloc_skb_ip_align(rx_ring->netdev,
1781 						IXGBE_RX_HDR_SIZE);
1782 		if (unlikely(!skb)) {
1783 			rx_ring->rx_stats.alloc_rx_buff_failed++;
1784 			return NULL;
1785 		}
1786 
1787 		/*
1788 		 * we will be copying header into skb->data in
1789 		 * pskb_may_pull so it is in our interest to prefetch
1790 		 * it now to avoid a possible cache miss
1791 		 */
1792 		prefetchw(skb->data);
1793 
1794 		/*
1795 		 * Delay unmapping of the first packet. It carries the
1796 		 * header information, HW may still access the header
1797 		 * after the writeback.  Only unmap it when EOP is
1798 		 * reached
1799 		 */
1800 		if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1801 			goto dma_sync;
1802 
1803 		IXGBE_CB(skb)->dma = rx_buffer->dma;
1804 	} else {
1805 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
1806 			ixgbe_dma_sync_frag(rx_ring, skb);
1807 
1808 dma_sync:
1809 		/* we are reusing so sync this buffer for CPU use */
1810 		dma_sync_single_range_for_cpu(rx_ring->dev,
1811 					      rx_buffer->dma,
1812 					      rx_buffer->page_offset,
1813 					      ixgbe_rx_bufsz(rx_ring),
1814 					      DMA_FROM_DEVICE);
1815 	}
1816 
1817 	/* pull page into skb */
1818 	if (ixgbe_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) {
1819 		/* hand second half of page back to the ring */
1820 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
1821 	} else if (IXGBE_CB(skb)->dma == rx_buffer->dma) {
1822 		/* the page has been released from the ring */
1823 		IXGBE_CB(skb)->page_released = true;
1824 	} else {
1825 		/* we are not reusing the buffer so unmap it */
1826 		dma_unmap_page(rx_ring->dev, rx_buffer->dma,
1827 			       ixgbe_rx_pg_size(rx_ring),
1828 			       DMA_FROM_DEVICE);
1829 	}
1830 
1831 	/* clear contents of buffer_info */
1832 	rx_buffer->skb = NULL;
1833 	rx_buffer->dma = 0;
1834 	rx_buffer->page = NULL;
1835 
1836 	return skb;
1837 }
1838 
1839 /**
1840  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
1841  * @q_vector: structure containing interrupt and ring information
1842  * @rx_ring: rx descriptor ring to transact packets on
1843  * @budget: Total limit on number of packets to process
1844  *
1845  * This function provides a "bounce buffer" approach to Rx interrupt
1846  * processing.  The advantage to this is that on systems that have
1847  * expensive overhead for IOMMU access this provides a means of avoiding
1848  * it by maintaining the mapping of the page to the syste.
1849  *
1850  * Returns true if all work is completed without reaching budget
1851  **/
1852 static bool ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
1853 			       struct ixgbe_ring *rx_ring,
1854 			       const int budget)
1855 {
1856 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1857 #ifdef IXGBE_FCOE
1858 	struct ixgbe_adapter *adapter = q_vector->adapter;
1859 	int ddp_bytes;
1860 	unsigned int mss = 0;
1861 #endif /* IXGBE_FCOE */
1862 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
1863 
1864 	do {
1865 		union ixgbe_adv_rx_desc *rx_desc;
1866 		struct sk_buff *skb;
1867 
1868 		/* return some buffers to hardware, one at a time is too slow */
1869 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
1870 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
1871 			cleaned_count = 0;
1872 		}
1873 
1874 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
1875 
1876 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_DD))
1877 			break;
1878 
1879 		/*
1880 		 * This memory barrier is needed to keep us from reading
1881 		 * any other fields out of the rx_desc until we know the
1882 		 * RXD_STAT_DD bit is set
1883 		 */
1884 		rmb();
1885 
1886 		/* retrieve a buffer from the ring */
1887 		skb = ixgbe_fetch_rx_buffer(rx_ring, rx_desc);
1888 
1889 		/* exit if we failed to retrieve a buffer */
1890 		if (!skb)
1891 			break;
1892 
1893 		cleaned_count++;
1894 
1895 		/* place incomplete frames back on ring for completion */
1896 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
1897 			continue;
1898 
1899 		/* verify the packet layout is correct */
1900 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
1901 			continue;
1902 
1903 		/* probably a little skewed due to removing CRC */
1904 		total_rx_bytes += skb->len;
1905 
1906 		/* populate checksum, timestamp, VLAN, and protocol */
1907 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
1908 
1909 #ifdef IXGBE_FCOE
1910 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
1911 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
1912 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
1913 			/* include DDPed FCoE data */
1914 			if (ddp_bytes > 0) {
1915 				if (!mss) {
1916 					mss = rx_ring->netdev->mtu -
1917 						sizeof(struct fcoe_hdr) -
1918 						sizeof(struct fc_frame_header) -
1919 						sizeof(struct fcoe_crc_eof);
1920 					if (mss > 512)
1921 						mss &= ~511;
1922 				}
1923 				total_rx_bytes += ddp_bytes;
1924 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
1925 								 mss);
1926 			}
1927 			if (!ddp_bytes) {
1928 				dev_kfree_skb_any(skb);
1929 				continue;
1930 			}
1931 		}
1932 
1933 #endif /* IXGBE_FCOE */
1934 		ixgbe_rx_skb(q_vector, skb);
1935 
1936 		/* update budget accounting */
1937 		total_rx_packets++;
1938 	} while (likely(total_rx_packets < budget));
1939 
1940 	u64_stats_update_begin(&rx_ring->syncp);
1941 	rx_ring->stats.packets += total_rx_packets;
1942 	rx_ring->stats.bytes += total_rx_bytes;
1943 	u64_stats_update_end(&rx_ring->syncp);
1944 	q_vector->rx.total_packets += total_rx_packets;
1945 	q_vector->rx.total_bytes += total_rx_bytes;
1946 
1947 	if (cleaned_count)
1948 		ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
1949 
1950 	return (total_rx_packets < budget);
1951 }
1952 
1953 /**
1954  * ixgbe_configure_msix - Configure MSI-X hardware
1955  * @adapter: board private structure
1956  *
1957  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
1958  * interrupts.
1959  **/
1960 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
1961 {
1962 	struct ixgbe_q_vector *q_vector;
1963 	int v_idx;
1964 	u32 mask;
1965 
1966 	/* Populate MSIX to EITR Select */
1967 	if (adapter->num_vfs > 32) {
1968 		u32 eitrsel = (1 << (adapter->num_vfs - 32)) - 1;
1969 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
1970 	}
1971 
1972 	/*
1973 	 * Populate the IVAR table and set the ITR values to the
1974 	 * corresponding register.
1975 	 */
1976 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
1977 		struct ixgbe_ring *ring;
1978 		q_vector = adapter->q_vector[v_idx];
1979 
1980 		ixgbe_for_each_ring(ring, q_vector->rx)
1981 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
1982 
1983 		ixgbe_for_each_ring(ring, q_vector->tx)
1984 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
1985 
1986 		ixgbe_write_eitr(q_vector);
1987 	}
1988 
1989 	switch (adapter->hw.mac.type) {
1990 	case ixgbe_mac_82598EB:
1991 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
1992 			       v_idx);
1993 		break;
1994 	case ixgbe_mac_82599EB:
1995 	case ixgbe_mac_X540:
1996 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
1997 		break;
1998 	default:
1999 		break;
2000 	}
2001 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2002 
2003 	/* set up to autoclear timer, and the vectors */
2004 	mask = IXGBE_EIMS_ENABLE_MASK;
2005 	mask &= ~(IXGBE_EIMS_OTHER |
2006 		  IXGBE_EIMS_MAILBOX |
2007 		  IXGBE_EIMS_LSC);
2008 
2009 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2010 }
2011 
2012 enum latency_range {
2013 	lowest_latency = 0,
2014 	low_latency = 1,
2015 	bulk_latency = 2,
2016 	latency_invalid = 255
2017 };
2018 
2019 /**
2020  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2021  * @q_vector: structure containing interrupt and ring information
2022  * @ring_container: structure containing ring performance data
2023  *
2024  *      Stores a new ITR value based on packets and byte
2025  *      counts during the last interrupt.  The advantage of per interrupt
2026  *      computation is faster updates and more accurate ITR for the current
2027  *      traffic pattern.  Constants in this function were computed
2028  *      based on theoretical maximum wire speed and thresholds were set based
2029  *      on testing data as well as attempting to minimize response time
2030  *      while increasing bulk throughput.
2031  *      this functionality is controlled by the InterruptThrottleRate module
2032  *      parameter (see ixgbe_param.c)
2033  **/
2034 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2035 			     struct ixgbe_ring_container *ring_container)
2036 {
2037 	int bytes = ring_container->total_bytes;
2038 	int packets = ring_container->total_packets;
2039 	u32 timepassed_us;
2040 	u64 bytes_perint;
2041 	u8 itr_setting = ring_container->itr;
2042 
2043 	if (packets == 0)
2044 		return;
2045 
2046 	/* simple throttlerate management
2047 	 *   0-10MB/s   lowest (100000 ints/s)
2048 	 *  10-20MB/s   low    (20000 ints/s)
2049 	 *  20-1249MB/s bulk   (8000 ints/s)
2050 	 */
2051 	/* what was last interrupt timeslice? */
2052 	timepassed_us = q_vector->itr >> 2;
2053 	bytes_perint = bytes / timepassed_us; /* bytes/usec */
2054 
2055 	switch (itr_setting) {
2056 	case lowest_latency:
2057 		if (bytes_perint > 10)
2058 			itr_setting = low_latency;
2059 		break;
2060 	case low_latency:
2061 		if (bytes_perint > 20)
2062 			itr_setting = bulk_latency;
2063 		else if (bytes_perint <= 10)
2064 			itr_setting = lowest_latency;
2065 		break;
2066 	case bulk_latency:
2067 		if (bytes_perint <= 20)
2068 			itr_setting = low_latency;
2069 		break;
2070 	}
2071 
2072 	/* clear work counters since we have the values we need */
2073 	ring_container->total_bytes = 0;
2074 	ring_container->total_packets = 0;
2075 
2076 	/* write updated itr to ring container */
2077 	ring_container->itr = itr_setting;
2078 }
2079 
2080 /**
2081  * ixgbe_write_eitr - write EITR register in hardware specific way
2082  * @q_vector: structure containing interrupt and ring information
2083  *
2084  * This function is made to be called by ethtool and by the driver
2085  * when it needs to update EITR registers at runtime.  Hardware
2086  * specific quirks/differences are taken care of here.
2087  */
2088 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2089 {
2090 	struct ixgbe_adapter *adapter = q_vector->adapter;
2091 	struct ixgbe_hw *hw = &adapter->hw;
2092 	int v_idx = q_vector->v_idx;
2093 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2094 
2095 	switch (adapter->hw.mac.type) {
2096 	case ixgbe_mac_82598EB:
2097 		/* must write high and low 16 bits to reset counter */
2098 		itr_reg |= (itr_reg << 16);
2099 		break;
2100 	case ixgbe_mac_82599EB:
2101 	case ixgbe_mac_X540:
2102 		/*
2103 		 * set the WDIS bit to not clear the timer bits and cause an
2104 		 * immediate assertion of the interrupt
2105 		 */
2106 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2107 		break;
2108 	default:
2109 		break;
2110 	}
2111 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2112 }
2113 
2114 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2115 {
2116 	u32 new_itr = q_vector->itr;
2117 	u8 current_itr;
2118 
2119 	ixgbe_update_itr(q_vector, &q_vector->tx);
2120 	ixgbe_update_itr(q_vector, &q_vector->rx);
2121 
2122 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
2123 
2124 	switch (current_itr) {
2125 	/* counts and packets in update_itr are dependent on these numbers */
2126 	case lowest_latency:
2127 		new_itr = IXGBE_100K_ITR;
2128 		break;
2129 	case low_latency:
2130 		new_itr = IXGBE_20K_ITR;
2131 		break;
2132 	case bulk_latency:
2133 		new_itr = IXGBE_8K_ITR;
2134 		break;
2135 	default:
2136 		break;
2137 	}
2138 
2139 	if (new_itr != q_vector->itr) {
2140 		/* do an exponential smoothing */
2141 		new_itr = (10 * new_itr * q_vector->itr) /
2142 			  ((9 * new_itr) + q_vector->itr);
2143 
2144 		/* save the algorithm value here */
2145 		q_vector->itr = new_itr;
2146 
2147 		ixgbe_write_eitr(q_vector);
2148 	}
2149 }
2150 
2151 /**
2152  * ixgbe_check_overtemp_subtask - check for over temperature
2153  * @adapter: pointer to adapter
2154  **/
2155 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2156 {
2157 	struct ixgbe_hw *hw = &adapter->hw;
2158 	u32 eicr = adapter->interrupt_event;
2159 
2160 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2161 		return;
2162 
2163 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) &&
2164 	    !(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2165 		return;
2166 
2167 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2168 
2169 	switch (hw->device_id) {
2170 	case IXGBE_DEV_ID_82599_T3_LOM:
2171 		/*
2172 		 * Since the warning interrupt is for both ports
2173 		 * we don't have to check if:
2174 		 *  - This interrupt wasn't for our port.
2175 		 *  - We may have missed the interrupt so always have to
2176 		 *    check if we  got a LSC
2177 		 */
2178 		if (!(eicr & IXGBE_EICR_GPI_SDP0) &&
2179 		    !(eicr & IXGBE_EICR_LSC))
2180 			return;
2181 
2182 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2183 			u32 autoneg;
2184 			bool link_up = false;
2185 
2186 			hw->mac.ops.check_link(hw, &autoneg, &link_up, false);
2187 
2188 			if (link_up)
2189 				return;
2190 		}
2191 
2192 		/* Check if this is not due to overtemp */
2193 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2194 			return;
2195 
2196 		break;
2197 	default:
2198 		if (!(eicr & IXGBE_EICR_GPI_SDP0))
2199 			return;
2200 		break;
2201 	}
2202 	e_crit(drv,
2203 	       "Network adapter has been stopped because it has over heated. "
2204 	       "Restart the computer. If the problem persists, "
2205 	       "power off the system and replace the adapter\n");
2206 
2207 	adapter->interrupt_event = 0;
2208 }
2209 
2210 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2211 {
2212 	struct ixgbe_hw *hw = &adapter->hw;
2213 
2214 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2215 	    (eicr & IXGBE_EICR_GPI_SDP1)) {
2216 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2217 		/* write to clear the interrupt */
2218 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2219 	}
2220 }
2221 
2222 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2223 {
2224 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2225 		return;
2226 
2227 	switch (adapter->hw.mac.type) {
2228 	case ixgbe_mac_82599EB:
2229 		/*
2230 		 * Need to check link state so complete overtemp check
2231 		 * on service task
2232 		 */
2233 		if (((eicr & IXGBE_EICR_GPI_SDP0) || (eicr & IXGBE_EICR_LSC)) &&
2234 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2235 			adapter->interrupt_event = eicr;
2236 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2237 			ixgbe_service_event_schedule(adapter);
2238 			return;
2239 		}
2240 		return;
2241 	case ixgbe_mac_X540:
2242 		if (!(eicr & IXGBE_EICR_TS))
2243 			return;
2244 		break;
2245 	default:
2246 		return;
2247 	}
2248 
2249 	e_crit(drv,
2250 	       "Network adapter has been stopped because it has over heated. "
2251 	       "Restart the computer. If the problem persists, "
2252 	       "power off the system and replace the adapter\n");
2253 }
2254 
2255 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2256 {
2257 	struct ixgbe_hw *hw = &adapter->hw;
2258 
2259 	if (eicr & IXGBE_EICR_GPI_SDP2) {
2260 		/* Clear the interrupt */
2261 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP2);
2262 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2263 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2264 			ixgbe_service_event_schedule(adapter);
2265 		}
2266 	}
2267 
2268 	if (eicr & IXGBE_EICR_GPI_SDP1) {
2269 		/* Clear the interrupt */
2270 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2271 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2272 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2273 			ixgbe_service_event_schedule(adapter);
2274 		}
2275 	}
2276 }
2277 
2278 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2279 {
2280 	struct ixgbe_hw *hw = &adapter->hw;
2281 
2282 	adapter->lsc_int++;
2283 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2284 	adapter->link_check_timeout = jiffies;
2285 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2286 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2287 		IXGBE_WRITE_FLUSH(hw);
2288 		ixgbe_service_event_schedule(adapter);
2289 	}
2290 }
2291 
2292 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2293 					   u64 qmask)
2294 {
2295 	u32 mask;
2296 	struct ixgbe_hw *hw = &adapter->hw;
2297 
2298 	switch (hw->mac.type) {
2299 	case ixgbe_mac_82598EB:
2300 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2301 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2302 		break;
2303 	case ixgbe_mac_82599EB:
2304 	case ixgbe_mac_X540:
2305 		mask = (qmask & 0xFFFFFFFF);
2306 		if (mask)
2307 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2308 		mask = (qmask >> 32);
2309 		if (mask)
2310 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2311 		break;
2312 	default:
2313 		break;
2314 	}
2315 	/* skip the flush */
2316 }
2317 
2318 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2319 					    u64 qmask)
2320 {
2321 	u32 mask;
2322 	struct ixgbe_hw *hw = &adapter->hw;
2323 
2324 	switch (hw->mac.type) {
2325 	case ixgbe_mac_82598EB:
2326 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2327 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2328 		break;
2329 	case ixgbe_mac_82599EB:
2330 	case ixgbe_mac_X540:
2331 		mask = (qmask & 0xFFFFFFFF);
2332 		if (mask)
2333 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2334 		mask = (qmask >> 32);
2335 		if (mask)
2336 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2337 		break;
2338 	default:
2339 		break;
2340 	}
2341 	/* skip the flush */
2342 }
2343 
2344 /**
2345  * ixgbe_irq_enable - Enable default interrupt generation settings
2346  * @adapter: board private structure
2347  **/
2348 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2349 				    bool flush)
2350 {
2351 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2352 
2353 	/* don't reenable LSC while waiting for link */
2354 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2355 		mask &= ~IXGBE_EIMS_LSC;
2356 
2357 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2358 		switch (adapter->hw.mac.type) {
2359 		case ixgbe_mac_82599EB:
2360 			mask |= IXGBE_EIMS_GPI_SDP0;
2361 			break;
2362 		case ixgbe_mac_X540:
2363 			mask |= IXGBE_EIMS_TS;
2364 			break;
2365 		default:
2366 			break;
2367 		}
2368 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2369 		mask |= IXGBE_EIMS_GPI_SDP1;
2370 	switch (adapter->hw.mac.type) {
2371 	case ixgbe_mac_82599EB:
2372 		mask |= IXGBE_EIMS_GPI_SDP1;
2373 		mask |= IXGBE_EIMS_GPI_SDP2;
2374 	case ixgbe_mac_X540:
2375 		mask |= IXGBE_EIMS_ECC;
2376 		mask |= IXGBE_EIMS_MAILBOX;
2377 		break;
2378 	default:
2379 		break;
2380 	}
2381 
2382 	if (adapter->hw.mac.type == ixgbe_mac_X540)
2383 		mask |= IXGBE_EIMS_TIMESYNC;
2384 
2385 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2386 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
2387 		mask |= IXGBE_EIMS_FLOW_DIR;
2388 
2389 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
2390 	if (queues)
2391 		ixgbe_irq_enable_queues(adapter, ~0);
2392 	if (flush)
2393 		IXGBE_WRITE_FLUSH(&adapter->hw);
2394 }
2395 
2396 static irqreturn_t ixgbe_msix_other(int irq, void *data)
2397 {
2398 	struct ixgbe_adapter *adapter = data;
2399 	struct ixgbe_hw *hw = &adapter->hw;
2400 	u32 eicr;
2401 
2402 	/*
2403 	 * Workaround for Silicon errata.  Use clear-by-write instead
2404 	 * of clear-by-read.  Reading with EICS will return the
2405 	 * interrupt causes without clearing, which later be done
2406 	 * with the write to EICR.
2407 	 */
2408 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2409 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2410 
2411 	if (eicr & IXGBE_EICR_LSC)
2412 		ixgbe_check_lsc(adapter);
2413 
2414 	if (eicr & IXGBE_EICR_MAILBOX)
2415 		ixgbe_msg_task(adapter);
2416 
2417 	switch (hw->mac.type) {
2418 	case ixgbe_mac_82599EB:
2419 	case ixgbe_mac_X540:
2420 		if (eicr & IXGBE_EICR_ECC)
2421 			e_info(link, "Received unrecoverable ECC Err, please "
2422 			       "reboot\n");
2423 		/* Handle Flow Director Full threshold interrupt */
2424 		if (eicr & IXGBE_EICR_FLOW_DIR) {
2425 			int reinit_count = 0;
2426 			int i;
2427 			for (i = 0; i < adapter->num_tx_queues; i++) {
2428 				struct ixgbe_ring *ring = adapter->tx_ring[i];
2429 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
2430 						       &ring->state))
2431 					reinit_count++;
2432 			}
2433 			if (reinit_count) {
2434 				/* no more flow director interrupts until after init */
2435 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
2436 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
2437 				ixgbe_service_event_schedule(adapter);
2438 			}
2439 		}
2440 		ixgbe_check_sfp_event(adapter, eicr);
2441 		ixgbe_check_overtemp_event(adapter, eicr);
2442 		break;
2443 	default:
2444 		break;
2445 	}
2446 
2447 	ixgbe_check_fan_failure(adapter, eicr);
2448 
2449 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2450 		ixgbe_ptp_check_pps_event(adapter, eicr);
2451 
2452 	/* re-enable the original interrupt state, no lsc, no queues */
2453 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2454 		ixgbe_irq_enable(adapter, false, false);
2455 
2456 	return IRQ_HANDLED;
2457 }
2458 
2459 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
2460 {
2461 	struct ixgbe_q_vector *q_vector = data;
2462 
2463 	/* EIAM disabled interrupts (on this vector) for us */
2464 
2465 	if (q_vector->rx.ring || q_vector->tx.ring)
2466 		napi_schedule(&q_vector->napi);
2467 
2468 	return IRQ_HANDLED;
2469 }
2470 
2471 /**
2472  * ixgbe_poll - NAPI Rx polling callback
2473  * @napi: structure for representing this polling device
2474  * @budget: how many packets driver is allowed to clean
2475  *
2476  * This function is used for legacy and MSI, NAPI mode
2477  **/
2478 int ixgbe_poll(struct napi_struct *napi, int budget)
2479 {
2480 	struct ixgbe_q_vector *q_vector =
2481 				container_of(napi, struct ixgbe_q_vector, napi);
2482 	struct ixgbe_adapter *adapter = q_vector->adapter;
2483 	struct ixgbe_ring *ring;
2484 	int per_ring_budget;
2485 	bool clean_complete = true;
2486 
2487 #ifdef CONFIG_IXGBE_DCA
2488 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
2489 		ixgbe_update_dca(q_vector);
2490 #endif
2491 
2492 	ixgbe_for_each_ring(ring, q_vector->tx)
2493 		clean_complete &= !!ixgbe_clean_tx_irq(q_vector, ring);
2494 
2495 	/* attempt to distribute budget to each queue fairly, but don't allow
2496 	 * the budget to go below 1 because we'll exit polling */
2497 	if (q_vector->rx.count > 1)
2498 		per_ring_budget = max(budget/q_vector->rx.count, 1);
2499 	else
2500 		per_ring_budget = budget;
2501 
2502 	ixgbe_for_each_ring(ring, q_vector->rx)
2503 		clean_complete &= ixgbe_clean_rx_irq(q_vector, ring,
2504 						     per_ring_budget);
2505 
2506 	/* If all work not completed, return budget and keep polling */
2507 	if (!clean_complete)
2508 		return budget;
2509 
2510 	/* all work done, exit the polling mode */
2511 	napi_complete(napi);
2512 	if (adapter->rx_itr_setting & 1)
2513 		ixgbe_set_itr(q_vector);
2514 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2515 		ixgbe_irq_enable_queues(adapter, ((u64)1 << q_vector->v_idx));
2516 
2517 	return 0;
2518 }
2519 
2520 /**
2521  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2522  * @adapter: board private structure
2523  *
2524  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2525  * interrupts from the kernel.
2526  **/
2527 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
2528 {
2529 	struct net_device *netdev = adapter->netdev;
2530 	int vector, err;
2531 	int ri = 0, ti = 0;
2532 
2533 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2534 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2535 		struct msix_entry *entry = &adapter->msix_entries[vector];
2536 
2537 		if (q_vector->tx.ring && q_vector->rx.ring) {
2538 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2539 				 "%s-%s-%d", netdev->name, "TxRx", ri++);
2540 			ti++;
2541 		} else if (q_vector->rx.ring) {
2542 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2543 				 "%s-%s-%d", netdev->name, "rx", ri++);
2544 		} else if (q_vector->tx.ring) {
2545 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2546 				 "%s-%s-%d", netdev->name, "tx", ti++);
2547 		} else {
2548 			/* skip this unused q_vector */
2549 			continue;
2550 		}
2551 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
2552 				  q_vector->name, q_vector);
2553 		if (err) {
2554 			e_err(probe, "request_irq failed for MSIX interrupt "
2555 			      "Error: %d\n", err);
2556 			goto free_queue_irqs;
2557 		}
2558 		/* If Flow Director is enabled, set interrupt affinity */
2559 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2560 			/* assign the mask for this irq */
2561 			irq_set_affinity_hint(entry->vector,
2562 					      &q_vector->affinity_mask);
2563 		}
2564 	}
2565 
2566 	err = request_irq(adapter->msix_entries[vector].vector,
2567 			  ixgbe_msix_other, 0, netdev->name, adapter);
2568 	if (err) {
2569 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
2570 		goto free_queue_irqs;
2571 	}
2572 
2573 	return 0;
2574 
2575 free_queue_irqs:
2576 	while (vector) {
2577 		vector--;
2578 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
2579 				      NULL);
2580 		free_irq(adapter->msix_entries[vector].vector,
2581 			 adapter->q_vector[vector]);
2582 	}
2583 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
2584 	pci_disable_msix(adapter->pdev);
2585 	kfree(adapter->msix_entries);
2586 	adapter->msix_entries = NULL;
2587 	return err;
2588 }
2589 
2590 /**
2591  * ixgbe_intr - legacy mode Interrupt Handler
2592  * @irq: interrupt number
2593  * @data: pointer to a network interface device structure
2594  **/
2595 static irqreturn_t ixgbe_intr(int irq, void *data)
2596 {
2597 	struct ixgbe_adapter *adapter = data;
2598 	struct ixgbe_hw *hw = &adapter->hw;
2599 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2600 	u32 eicr;
2601 
2602 	/*
2603 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
2604 	 * before the read of EICR.
2605 	 */
2606 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
2607 
2608 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2609 	 * therefore no explicit interrupt disable is necessary */
2610 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2611 	if (!eicr) {
2612 		/*
2613 		 * shared interrupt alert!
2614 		 * make sure interrupts are enabled because the read will
2615 		 * have disabled interrupts due to EIAM
2616 		 * finish the workaround of silicon errata on 82598.  Unmask
2617 		 * the interrupt that we masked before the EICR read.
2618 		 */
2619 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
2620 			ixgbe_irq_enable(adapter, true, true);
2621 		return IRQ_NONE;	/* Not our interrupt */
2622 	}
2623 
2624 	if (eicr & IXGBE_EICR_LSC)
2625 		ixgbe_check_lsc(adapter);
2626 
2627 	switch (hw->mac.type) {
2628 	case ixgbe_mac_82599EB:
2629 		ixgbe_check_sfp_event(adapter, eicr);
2630 		/* Fall through */
2631 	case ixgbe_mac_X540:
2632 		if (eicr & IXGBE_EICR_ECC)
2633 			e_info(link, "Received unrecoverable ECC err, please "
2634 				     "reboot\n");
2635 		ixgbe_check_overtemp_event(adapter, eicr);
2636 		break;
2637 	default:
2638 		break;
2639 	}
2640 
2641 	ixgbe_check_fan_failure(adapter, eicr);
2642 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2643 		ixgbe_ptp_check_pps_event(adapter, eicr);
2644 
2645 	/* would disable interrupts here but EIAM disabled it */
2646 	napi_schedule(&q_vector->napi);
2647 
2648 	/*
2649 	 * re-enable link(maybe) and non-queue interrupts, no flush.
2650 	 * ixgbe_poll will re-enable the queue interrupts
2651 	 */
2652 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2653 		ixgbe_irq_enable(adapter, false, false);
2654 
2655 	return IRQ_HANDLED;
2656 }
2657 
2658 /**
2659  * ixgbe_request_irq - initialize interrupts
2660  * @adapter: board private structure
2661  *
2662  * Attempts to configure interrupts using the best available
2663  * capabilities of the hardware and kernel.
2664  **/
2665 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
2666 {
2667 	struct net_device *netdev = adapter->netdev;
2668 	int err;
2669 
2670 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
2671 		err = ixgbe_request_msix_irqs(adapter);
2672 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
2673 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
2674 				  netdev->name, adapter);
2675 	else
2676 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
2677 				  netdev->name, adapter);
2678 
2679 	if (err)
2680 		e_err(probe, "request_irq failed, Error %d\n", err);
2681 
2682 	return err;
2683 }
2684 
2685 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
2686 {
2687 	int vector;
2688 
2689 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
2690 		free_irq(adapter->pdev->irq, adapter);
2691 		return;
2692 	}
2693 
2694 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2695 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2696 		struct msix_entry *entry = &adapter->msix_entries[vector];
2697 
2698 		/* free only the irqs that were actually requested */
2699 		if (!q_vector->rx.ring && !q_vector->tx.ring)
2700 			continue;
2701 
2702 		/* clear the affinity_mask in the IRQ descriptor */
2703 		irq_set_affinity_hint(entry->vector, NULL);
2704 
2705 		free_irq(entry->vector, q_vector);
2706 	}
2707 
2708 	free_irq(adapter->msix_entries[vector++].vector, adapter);
2709 }
2710 
2711 /**
2712  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
2713  * @adapter: board private structure
2714  **/
2715 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
2716 {
2717 	switch (adapter->hw.mac.type) {
2718 	case ixgbe_mac_82598EB:
2719 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
2720 		break;
2721 	case ixgbe_mac_82599EB:
2722 	case ixgbe_mac_X540:
2723 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
2724 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
2725 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
2726 		break;
2727 	default:
2728 		break;
2729 	}
2730 	IXGBE_WRITE_FLUSH(&adapter->hw);
2731 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
2732 		int vector;
2733 
2734 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
2735 			synchronize_irq(adapter->msix_entries[vector].vector);
2736 
2737 		synchronize_irq(adapter->msix_entries[vector++].vector);
2738 	} else {
2739 		synchronize_irq(adapter->pdev->irq);
2740 	}
2741 }
2742 
2743 /**
2744  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
2745  *
2746  **/
2747 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
2748 {
2749 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2750 
2751 	ixgbe_write_eitr(q_vector);
2752 
2753 	ixgbe_set_ivar(adapter, 0, 0, 0);
2754 	ixgbe_set_ivar(adapter, 1, 0, 0);
2755 
2756 	e_info(hw, "Legacy interrupt IVAR setup done\n");
2757 }
2758 
2759 /**
2760  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
2761  * @adapter: board private structure
2762  * @ring: structure containing ring specific data
2763  *
2764  * Configure the Tx descriptor ring after a reset.
2765  **/
2766 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
2767 			     struct ixgbe_ring *ring)
2768 {
2769 	struct ixgbe_hw *hw = &adapter->hw;
2770 	u64 tdba = ring->dma;
2771 	int wait_loop = 10;
2772 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
2773 	u8 reg_idx = ring->reg_idx;
2774 
2775 	/* disable queue to avoid issues while updating state */
2776 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
2777 	IXGBE_WRITE_FLUSH(hw);
2778 
2779 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
2780 			(tdba & DMA_BIT_MASK(32)));
2781 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
2782 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
2783 			ring->count * sizeof(union ixgbe_adv_tx_desc));
2784 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
2785 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
2786 	ring->tail = hw->hw_addr + IXGBE_TDT(reg_idx);
2787 
2788 	/*
2789 	 * set WTHRESH to encourage burst writeback, it should not be set
2790 	 * higher than 1 when ITR is 0 as it could cause false TX hangs
2791 	 *
2792 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
2793 	 * to or less than the number of on chip descriptors, which is
2794 	 * currently 40.
2795 	 */
2796 	if (!ring->q_vector || (ring->q_vector->itr < 8))
2797 		txdctl |= (1 << 16);	/* WTHRESH = 1 */
2798 	else
2799 		txdctl |= (8 << 16);	/* WTHRESH = 8 */
2800 
2801 	/*
2802 	 * Setting PTHRESH to 32 both improves performance
2803 	 * and avoids a TX hang with DFP enabled
2804 	 */
2805 	txdctl |= (1 << 8) |	/* HTHRESH = 1 */
2806 		   32;		/* PTHRESH = 32 */
2807 
2808 	/* reinitialize flowdirector state */
2809 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2810 		ring->atr_sample_rate = adapter->atr_sample_rate;
2811 		ring->atr_count = 0;
2812 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
2813 	} else {
2814 		ring->atr_sample_rate = 0;
2815 	}
2816 
2817 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
2818 
2819 	/* enable queue */
2820 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
2821 
2822 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
2823 	if (hw->mac.type == ixgbe_mac_82598EB &&
2824 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
2825 		return;
2826 
2827 	/* poll to verify queue is enabled */
2828 	do {
2829 		usleep_range(1000, 2000);
2830 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
2831 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
2832 	if (!wait_loop)
2833 		e_err(drv, "Could not enable Tx Queue %d\n", reg_idx);
2834 }
2835 
2836 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
2837 {
2838 	struct ixgbe_hw *hw = &adapter->hw;
2839 	u32 rttdcs, mtqc;
2840 	u8 tcs = netdev_get_num_tc(adapter->netdev);
2841 
2842 	if (hw->mac.type == ixgbe_mac_82598EB)
2843 		return;
2844 
2845 	/* disable the arbiter while setting MTQC */
2846 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
2847 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
2848 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
2849 
2850 	/* set transmit pool layout */
2851 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
2852 		mtqc = IXGBE_MTQC_VT_ENA;
2853 		if (tcs > 4)
2854 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
2855 		else if (tcs > 1)
2856 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
2857 		else if (adapter->ring_feature[RING_F_RSS].indices == 4)
2858 			mtqc |= IXGBE_MTQC_32VF;
2859 		else
2860 			mtqc |= IXGBE_MTQC_64VF;
2861 	} else {
2862 		if (tcs > 4)
2863 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
2864 		else if (tcs > 1)
2865 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
2866 		else
2867 			mtqc = IXGBE_MTQC_64Q_1PB;
2868 	}
2869 
2870 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
2871 
2872 	/* Enable Security TX Buffer IFG for multiple pb */
2873 	if (tcs) {
2874 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
2875 		sectx |= IXGBE_SECTX_DCB;
2876 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
2877 	}
2878 
2879 	/* re-enable the arbiter */
2880 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
2881 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
2882 }
2883 
2884 /**
2885  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
2886  * @adapter: board private structure
2887  *
2888  * Configure the Tx unit of the MAC after a reset.
2889  **/
2890 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
2891 {
2892 	struct ixgbe_hw *hw = &adapter->hw;
2893 	u32 dmatxctl;
2894 	u32 i;
2895 
2896 	ixgbe_setup_mtqc(adapter);
2897 
2898 	if (hw->mac.type != ixgbe_mac_82598EB) {
2899 		/* DMATXCTL.EN must be before Tx queues are enabled */
2900 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2901 		dmatxctl |= IXGBE_DMATXCTL_TE;
2902 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
2903 	}
2904 
2905 	/* Setup the HW Tx Head and Tail descriptor pointers */
2906 	for (i = 0; i < adapter->num_tx_queues; i++)
2907 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
2908 }
2909 
2910 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
2911 				 struct ixgbe_ring *ring)
2912 {
2913 	struct ixgbe_hw *hw = &adapter->hw;
2914 	u8 reg_idx = ring->reg_idx;
2915 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
2916 
2917 	srrctl |= IXGBE_SRRCTL_DROP_EN;
2918 
2919 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
2920 }
2921 
2922 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
2923 				  struct ixgbe_ring *ring)
2924 {
2925 	struct ixgbe_hw *hw = &adapter->hw;
2926 	u8 reg_idx = ring->reg_idx;
2927 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
2928 
2929 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
2930 
2931 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
2932 }
2933 
2934 #ifdef CONFIG_IXGBE_DCB
2935 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
2936 #else
2937 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
2938 #endif
2939 {
2940 	int i;
2941 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
2942 
2943 	if (adapter->ixgbe_ieee_pfc)
2944 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
2945 
2946 	/*
2947 	 * We should set the drop enable bit if:
2948 	 *  SR-IOV is enabled
2949 	 *   or
2950 	 *  Number of Rx queues > 1 and flow control is disabled
2951 	 *
2952 	 *  This allows us to avoid head of line blocking for security
2953 	 *  and performance reasons.
2954 	 */
2955 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
2956 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
2957 		for (i = 0; i < adapter->num_rx_queues; i++)
2958 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
2959 	} else {
2960 		for (i = 0; i < adapter->num_rx_queues; i++)
2961 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
2962 	}
2963 }
2964 
2965 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
2966 
2967 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
2968 				   struct ixgbe_ring *rx_ring)
2969 {
2970 	struct ixgbe_hw *hw = &adapter->hw;
2971 	u32 srrctl;
2972 	u8 reg_idx = rx_ring->reg_idx;
2973 
2974 	if (hw->mac.type == ixgbe_mac_82598EB) {
2975 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
2976 
2977 		/*
2978 		 * if VMDq is not active we must program one srrctl register
2979 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
2980 		 */
2981 		reg_idx &= mask;
2982 	}
2983 
2984 	/* configure header buffer length, needed for RSC */
2985 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
2986 
2987 	/* configure the packet buffer length */
2988 	srrctl |= ixgbe_rx_bufsz(rx_ring) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
2989 
2990 	/* configure descriptor type */
2991 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
2992 
2993 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
2994 }
2995 
2996 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
2997 {
2998 	struct ixgbe_hw *hw = &adapter->hw;
2999 	static const u32 seed[10] = { 0xE291D73D, 0x1805EC6C, 0x2A94B30D,
3000 			  0xA54F2BEC, 0xEA49AF7C, 0xE214AD3D, 0xB855AABE,
3001 			  0x6A3E67EA, 0x14364D17, 0x3BED200D};
3002 	u32 mrqc = 0, reta = 0;
3003 	u32 rxcsum;
3004 	int i, j;
3005 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3006 
3007 	/*
3008 	 * Program table for at least 2 queues w/ SR-IOV so that VFs can
3009 	 * make full use of any rings they may have.  We will use the
3010 	 * PSRTYPE register to control how many rings we use within the PF.
3011 	 */
3012 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 2))
3013 		rss_i = 2;
3014 
3015 	/* Fill out hash function seeds */
3016 	for (i = 0; i < 10; i++)
3017 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), seed[i]);
3018 
3019 	/* Fill out redirection table */
3020 	for (i = 0, j = 0; i < 128; i++, j++) {
3021 		if (j == rss_i)
3022 			j = 0;
3023 		/* reta = 4-byte sliding window of
3024 		 * 0x00..(indices-1)(indices-1)00..etc. */
3025 		reta = (reta << 8) | (j * 0x11);
3026 		if ((i & 3) == 3)
3027 			IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3028 	}
3029 
3030 	/* Disable indicating checksum in descriptor, enables RSS hash */
3031 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3032 	rxcsum |= IXGBE_RXCSUM_PCSD;
3033 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3034 
3035 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3036 		if (adapter->ring_feature[RING_F_RSS].mask)
3037 			mrqc = IXGBE_MRQC_RSSEN;
3038 	} else {
3039 		u8 tcs = netdev_get_num_tc(adapter->netdev);
3040 
3041 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3042 			if (tcs > 4)
3043 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3044 			else if (tcs > 1)
3045 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3046 			else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3047 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3048 			else
3049 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3050 		} else {
3051 			if (tcs > 4)
3052 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3053 			else if (tcs > 1)
3054 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3055 			else
3056 				mrqc = IXGBE_MRQC_RSSEN;
3057 		}
3058 	}
3059 
3060 	/* Perform hash on these packet types */
3061 	mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3062 		IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3063 		IXGBE_MRQC_RSS_FIELD_IPV6 |
3064 		IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3065 
3066 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3067 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3068 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3069 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3070 
3071 	IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3072 }
3073 
3074 /**
3075  * ixgbe_configure_rscctl - enable RSC for the indicated ring
3076  * @adapter:    address of board private structure
3077  * @index:      index of ring to set
3078  **/
3079 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
3080 				   struct ixgbe_ring *ring)
3081 {
3082 	struct ixgbe_hw *hw = &adapter->hw;
3083 	u32 rscctrl;
3084 	u8 reg_idx = ring->reg_idx;
3085 
3086 	if (!ring_is_rsc_enabled(ring))
3087 		return;
3088 
3089 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
3090 	rscctrl |= IXGBE_RSCCTL_RSCEN;
3091 	/*
3092 	 * we must limit the number of descriptors so that the
3093 	 * total size of max desc * buf_len is not greater
3094 	 * than 65536
3095 	 */
3096 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
3097 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
3098 }
3099 
3100 #define IXGBE_MAX_RX_DESC_POLL 10
3101 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
3102 				       struct ixgbe_ring *ring)
3103 {
3104 	struct ixgbe_hw *hw = &adapter->hw;
3105 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3106 	u32 rxdctl;
3107 	u8 reg_idx = ring->reg_idx;
3108 
3109 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3110 	if (hw->mac.type == ixgbe_mac_82598EB &&
3111 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3112 		return;
3113 
3114 	do {
3115 		usleep_range(1000, 2000);
3116 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3117 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
3118 
3119 	if (!wait_loop) {
3120 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
3121 		      "the polling period\n", reg_idx);
3122 	}
3123 }
3124 
3125 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
3126 			    struct ixgbe_ring *ring)
3127 {
3128 	struct ixgbe_hw *hw = &adapter->hw;
3129 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3130 	u32 rxdctl;
3131 	u8 reg_idx = ring->reg_idx;
3132 
3133 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3134 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
3135 
3136 	/* write value back with RXDCTL.ENABLE bit cleared */
3137 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3138 
3139 	if (hw->mac.type == ixgbe_mac_82598EB &&
3140 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3141 		return;
3142 
3143 	/* the hardware may take up to 100us to really disable the rx queue */
3144 	do {
3145 		udelay(10);
3146 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3147 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
3148 
3149 	if (!wait_loop) {
3150 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3151 		      "the polling period\n", reg_idx);
3152 	}
3153 }
3154 
3155 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
3156 			     struct ixgbe_ring *ring)
3157 {
3158 	struct ixgbe_hw *hw = &adapter->hw;
3159 	u64 rdba = ring->dma;
3160 	u32 rxdctl;
3161 	u8 reg_idx = ring->reg_idx;
3162 
3163 	/* disable queue to avoid issues while updating state */
3164 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3165 	ixgbe_disable_rx_queue(adapter, ring);
3166 
3167 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
3168 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
3169 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
3170 			ring->count * sizeof(union ixgbe_adv_rx_desc));
3171 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
3172 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
3173 	ring->tail = hw->hw_addr + IXGBE_RDT(reg_idx);
3174 
3175 	ixgbe_configure_srrctl(adapter, ring);
3176 	ixgbe_configure_rscctl(adapter, ring);
3177 
3178 	if (hw->mac.type == ixgbe_mac_82598EB) {
3179 		/*
3180 		 * enable cache line friendly hardware writes:
3181 		 * PTHRESH=32 descriptors (half the internal cache),
3182 		 * this also removes ugly rx_no_buffer_count increment
3183 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
3184 		 * WTHRESH=8 burst writeback up to two cache lines
3185 		 */
3186 		rxdctl &= ~0x3FFFFF;
3187 		rxdctl |=  0x080420;
3188 	}
3189 
3190 	/* enable receive descriptor ring */
3191 	rxdctl |= IXGBE_RXDCTL_ENABLE;
3192 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3193 
3194 	ixgbe_rx_desc_queue_enable(adapter, ring);
3195 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
3196 }
3197 
3198 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
3199 {
3200 	struct ixgbe_hw *hw = &adapter->hw;
3201 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
3202 	int p;
3203 
3204 	/* PSRTYPE must be initialized in non 82598 adapters */
3205 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
3206 		      IXGBE_PSRTYPE_UDPHDR |
3207 		      IXGBE_PSRTYPE_IPV4HDR |
3208 		      IXGBE_PSRTYPE_L2HDR |
3209 		      IXGBE_PSRTYPE_IPV6HDR;
3210 
3211 	if (hw->mac.type == ixgbe_mac_82598EB)
3212 		return;
3213 
3214 	if (rss_i > 3)
3215 		psrtype |= 2 << 29;
3216 	else if (rss_i > 1)
3217 		psrtype |= 1 << 29;
3218 
3219 	for (p = 0; p < adapter->num_rx_pools; p++)
3220 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(p)),
3221 				psrtype);
3222 }
3223 
3224 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
3225 {
3226 	struct ixgbe_hw *hw = &adapter->hw;
3227 	u32 reg_offset, vf_shift;
3228 	u32 gcr_ext, vmdctl;
3229 	int i;
3230 
3231 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3232 		return;
3233 
3234 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3235 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
3236 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
3237 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
3238 	vmdctl |= IXGBE_VT_CTL_REPLEN;
3239 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
3240 
3241 	vf_shift = VMDQ_P(0) % 32;
3242 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
3243 
3244 	/* Enable only the PF's pool for Tx/Rx */
3245 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), (~0) << vf_shift);
3246 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
3247 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), (~0) << vf_shift);
3248 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
3249 	if (adapter->flags2 & IXGBE_FLAG2_BRIDGE_MODE_VEB)
3250 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3251 
3252 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3253 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
3254 
3255 	/*
3256 	 * Set up VF register offsets for selected VT Mode,
3257 	 * i.e. 32 or 64 VFs for SR-IOV
3258 	 */
3259 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
3260 	case IXGBE_82599_VMDQ_8Q_MASK:
3261 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
3262 		break;
3263 	case IXGBE_82599_VMDQ_4Q_MASK:
3264 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
3265 		break;
3266 	default:
3267 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
3268 		break;
3269 	}
3270 
3271 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
3272 
3273 
3274 	/* Enable MAC Anti-Spoofing */
3275 	hw->mac.ops.set_mac_anti_spoofing(hw, (adapter->num_vfs != 0),
3276 					  adapter->num_vfs);
3277 	/* For VFs that have spoof checking turned off */
3278 	for (i = 0; i < adapter->num_vfs; i++) {
3279 		if (!adapter->vfinfo[i].spoofchk_enabled)
3280 			ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i, false);
3281 	}
3282 }
3283 
3284 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
3285 {
3286 	struct ixgbe_hw *hw = &adapter->hw;
3287 	struct net_device *netdev = adapter->netdev;
3288 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3289 	struct ixgbe_ring *rx_ring;
3290 	int i;
3291 	u32 mhadd, hlreg0;
3292 
3293 #ifdef IXGBE_FCOE
3294 	/* adjust max frame to be able to do baby jumbo for FCoE */
3295 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
3296 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
3297 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3298 
3299 #endif /* IXGBE_FCOE */
3300 
3301 	/* adjust max frame to be at least the size of a standard frame */
3302 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
3303 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
3304 
3305 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3306 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
3307 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3308 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
3309 
3310 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3311 	}
3312 
3313 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3314 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3315 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3316 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3317 
3318 	/*
3319 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3320 	 * the Base and Length of the Rx Descriptor Ring
3321 	 */
3322 	for (i = 0; i < adapter->num_rx_queues; i++) {
3323 		rx_ring = adapter->rx_ring[i];
3324 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
3325 			set_ring_rsc_enabled(rx_ring);
3326 		else
3327 			clear_ring_rsc_enabled(rx_ring);
3328 	}
3329 }
3330 
3331 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
3332 {
3333 	struct ixgbe_hw *hw = &adapter->hw;
3334 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
3335 
3336 	switch (hw->mac.type) {
3337 	case ixgbe_mac_82598EB:
3338 		/*
3339 		 * For VMDq support of different descriptor types or
3340 		 * buffer sizes through the use of multiple SRRCTL
3341 		 * registers, RDRXCTL.MVMEN must be set to 1
3342 		 *
3343 		 * also, the manual doesn't mention it clearly but DCA hints
3344 		 * will only use queue 0's tags unless this bit is set.  Side
3345 		 * effects of setting this bit are only that SRRCTL must be
3346 		 * fully programmed [0..15]
3347 		 */
3348 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
3349 		break;
3350 	case ixgbe_mac_82599EB:
3351 	case ixgbe_mac_X540:
3352 		/* Disable RSC for ACK packets */
3353 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
3354 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
3355 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
3356 		/* hardware requires some bits to be set by default */
3357 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
3358 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
3359 		break;
3360 	default:
3361 		/* We should do nothing since we don't know this hardware */
3362 		return;
3363 	}
3364 
3365 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
3366 }
3367 
3368 /**
3369  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3370  * @adapter: board private structure
3371  *
3372  * Configure the Rx unit of the MAC after a reset.
3373  **/
3374 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
3375 {
3376 	struct ixgbe_hw *hw = &adapter->hw;
3377 	int i;
3378 	u32 rxctrl;
3379 
3380 	/* disable receives while setting up the descriptors */
3381 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3382 	IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxctrl & ~IXGBE_RXCTRL_RXEN);
3383 
3384 	ixgbe_setup_psrtype(adapter);
3385 	ixgbe_setup_rdrxctl(adapter);
3386 
3387 	/* Program registers for the distribution of queues */
3388 	ixgbe_setup_mrqc(adapter);
3389 
3390 	/* set_rx_buffer_len must be called before ring initialization */
3391 	ixgbe_set_rx_buffer_len(adapter);
3392 
3393 	/*
3394 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3395 	 * the Base and Length of the Rx Descriptor Ring
3396 	 */
3397 	for (i = 0; i < adapter->num_rx_queues; i++)
3398 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
3399 
3400 	/* disable drop enable for 82598 parts */
3401 	if (hw->mac.type == ixgbe_mac_82598EB)
3402 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3403 
3404 	/* enable all receives */
3405 	rxctrl |= IXGBE_RXCTRL_RXEN;
3406 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
3407 }
3408 
3409 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev, u16 vid)
3410 {
3411 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3412 	struct ixgbe_hw *hw = &adapter->hw;
3413 
3414 	/* add VID to filter table */
3415 	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true);
3416 	set_bit(vid, adapter->active_vlans);
3417 
3418 	return 0;
3419 }
3420 
3421 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev, u16 vid)
3422 {
3423 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3424 	struct ixgbe_hw *hw = &adapter->hw;
3425 
3426 	/* remove VID from filter table */
3427 	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), false);
3428 	clear_bit(vid, adapter->active_vlans);
3429 
3430 	return 0;
3431 }
3432 
3433 /**
3434  * ixgbe_vlan_filter_disable - helper to disable hw vlan filtering
3435  * @adapter: driver data
3436  */
3437 static void ixgbe_vlan_filter_disable(struct ixgbe_adapter *adapter)
3438 {
3439 	struct ixgbe_hw *hw = &adapter->hw;
3440 	u32 vlnctrl;
3441 
3442 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3443 	vlnctrl &= ~(IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
3444 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3445 }
3446 
3447 /**
3448  * ixgbe_vlan_filter_enable - helper to enable hw vlan filtering
3449  * @adapter: driver data
3450  */
3451 static void ixgbe_vlan_filter_enable(struct ixgbe_adapter *adapter)
3452 {
3453 	struct ixgbe_hw *hw = &adapter->hw;
3454 	u32 vlnctrl;
3455 
3456 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3457 	vlnctrl |= IXGBE_VLNCTRL_VFE;
3458 	vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
3459 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3460 }
3461 
3462 /**
3463  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
3464  * @adapter: driver data
3465  */
3466 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
3467 {
3468 	struct ixgbe_hw *hw = &adapter->hw;
3469 	u32 vlnctrl;
3470 	int i, j;
3471 
3472 	switch (hw->mac.type) {
3473 	case ixgbe_mac_82598EB:
3474 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3475 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
3476 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3477 		break;
3478 	case ixgbe_mac_82599EB:
3479 	case ixgbe_mac_X540:
3480 		for (i = 0; i < adapter->num_rx_queues; i++) {
3481 			j = adapter->rx_ring[i]->reg_idx;
3482 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3483 			vlnctrl &= ~IXGBE_RXDCTL_VME;
3484 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3485 		}
3486 		break;
3487 	default:
3488 		break;
3489 	}
3490 }
3491 
3492 /**
3493  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
3494  * @adapter: driver data
3495  */
3496 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
3497 {
3498 	struct ixgbe_hw *hw = &adapter->hw;
3499 	u32 vlnctrl;
3500 	int i, j;
3501 
3502 	switch (hw->mac.type) {
3503 	case ixgbe_mac_82598EB:
3504 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3505 		vlnctrl |= IXGBE_VLNCTRL_VME;
3506 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3507 		break;
3508 	case ixgbe_mac_82599EB:
3509 	case ixgbe_mac_X540:
3510 		for (i = 0; i < adapter->num_rx_queues; i++) {
3511 			j = adapter->rx_ring[i]->reg_idx;
3512 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3513 			vlnctrl |= IXGBE_RXDCTL_VME;
3514 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3515 		}
3516 		break;
3517 	default:
3518 		break;
3519 	}
3520 }
3521 
3522 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
3523 {
3524 	u16 vid;
3525 
3526 	ixgbe_vlan_rx_add_vid(adapter->netdev, 0);
3527 
3528 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
3529 		ixgbe_vlan_rx_add_vid(adapter->netdev, vid);
3530 }
3531 
3532 /**
3533  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
3534  * @netdev: network interface device structure
3535  *
3536  * Writes unicast address list to the RAR table.
3537  * Returns: -ENOMEM on failure/insufficient address space
3538  *                0 on no addresses written
3539  *                X on writing X addresses to the RAR table
3540  **/
3541 static int ixgbe_write_uc_addr_list(struct net_device *netdev)
3542 {
3543 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3544 	struct ixgbe_hw *hw = &adapter->hw;
3545 	unsigned int rar_entries = hw->mac.num_rar_entries - 1;
3546 	int count = 0;
3547 
3548 	/* In SR-IOV mode significantly less RAR entries are available */
3549 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3550 		rar_entries = IXGBE_MAX_PF_MACVLANS - 1;
3551 
3552 	/* return ENOMEM indicating insufficient memory for addresses */
3553 	if (netdev_uc_count(netdev) > rar_entries)
3554 		return -ENOMEM;
3555 
3556 	if (!netdev_uc_empty(netdev)) {
3557 		struct netdev_hw_addr *ha;
3558 		/* return error if we do not support writing to RAR table */
3559 		if (!hw->mac.ops.set_rar)
3560 			return -ENOMEM;
3561 
3562 		netdev_for_each_uc_addr(ha, netdev) {
3563 			if (!rar_entries)
3564 				break;
3565 			hw->mac.ops.set_rar(hw, rar_entries--, ha->addr,
3566 					    VMDQ_P(0), IXGBE_RAH_AV);
3567 			count++;
3568 		}
3569 	}
3570 	/* write the addresses in reverse order to avoid write combining */
3571 	for (; rar_entries > 0 ; rar_entries--)
3572 		hw->mac.ops.clear_rar(hw, rar_entries);
3573 
3574 	return count;
3575 }
3576 
3577 /**
3578  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
3579  * @netdev: network interface device structure
3580  *
3581  * The set_rx_method entry point is called whenever the unicast/multicast
3582  * address list or the network interface flags are updated.  This routine is
3583  * responsible for configuring the hardware for proper unicast, multicast and
3584  * promiscuous mode.
3585  **/
3586 void ixgbe_set_rx_mode(struct net_device *netdev)
3587 {
3588 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3589 	struct ixgbe_hw *hw = &adapter->hw;
3590 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
3591 	int count;
3592 
3593 	/* Check for Promiscuous and All Multicast modes */
3594 
3595 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3596 
3597 	/* set all bits that we expect to always be set */
3598 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
3599 	fctrl |= IXGBE_FCTRL_BAM;
3600 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
3601 	fctrl |= IXGBE_FCTRL_PMCF;
3602 
3603 	/* clear the bits we are changing the status of */
3604 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3605 
3606 	if (netdev->flags & IFF_PROMISC) {
3607 		hw->addr_ctrl.user_set_promisc = true;
3608 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3609 		vmolr |= (IXGBE_VMOLR_ROPE | IXGBE_VMOLR_MPE);
3610 		/* don't hardware filter vlans in promisc mode */
3611 		ixgbe_vlan_filter_disable(adapter);
3612 	} else {
3613 		if (netdev->flags & IFF_ALLMULTI) {
3614 			fctrl |= IXGBE_FCTRL_MPE;
3615 			vmolr |= IXGBE_VMOLR_MPE;
3616 		} else {
3617 			/*
3618 			 * Write addresses to the MTA, if the attempt fails
3619 			 * then we should just turn on promiscuous mode so
3620 			 * that we can at least receive multicast traffic
3621 			 */
3622 			hw->mac.ops.update_mc_addr_list(hw, netdev);
3623 			vmolr |= IXGBE_VMOLR_ROMPE;
3624 		}
3625 		ixgbe_vlan_filter_enable(adapter);
3626 		hw->addr_ctrl.user_set_promisc = false;
3627 	}
3628 
3629 	/*
3630 	 * Write addresses to available RAR registers, if there is not
3631 	 * sufficient space to store all the addresses then enable
3632 	 * unicast promiscuous mode
3633 	 */
3634 	count = ixgbe_write_uc_addr_list(netdev);
3635 	if (count < 0) {
3636 		fctrl |= IXGBE_FCTRL_UPE;
3637 		vmolr |= IXGBE_VMOLR_ROPE;
3638 	}
3639 
3640 	if (adapter->num_vfs)
3641 		ixgbe_restore_vf_multicasts(adapter);
3642 
3643 	if (hw->mac.type != ixgbe_mac_82598EB) {
3644 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
3645 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
3646 			   IXGBE_VMOLR_ROPE);
3647 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
3648 	}
3649 
3650 	/* This is useful for sniffing bad packets. */
3651 	if (adapter->netdev->features & NETIF_F_RXALL) {
3652 		/* UPE and MPE will be handled by normal PROMISC logic
3653 		 * in e1000e_set_rx_mode */
3654 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
3655 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
3656 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
3657 
3658 		fctrl &= ~(IXGBE_FCTRL_DPF);
3659 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
3660 	}
3661 
3662 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
3663 
3664 	if (netdev->features & NETIF_F_HW_VLAN_RX)
3665 		ixgbe_vlan_strip_enable(adapter);
3666 	else
3667 		ixgbe_vlan_strip_disable(adapter);
3668 }
3669 
3670 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
3671 {
3672 	int q_idx;
3673 
3674 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
3675 		napi_enable(&adapter->q_vector[q_idx]->napi);
3676 }
3677 
3678 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
3679 {
3680 	int q_idx;
3681 
3682 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
3683 		napi_disable(&adapter->q_vector[q_idx]->napi);
3684 }
3685 
3686 #ifdef CONFIG_IXGBE_DCB
3687 /**
3688  * ixgbe_configure_dcb - Configure DCB hardware
3689  * @adapter: ixgbe adapter struct
3690  *
3691  * This is called by the driver on open to configure the DCB hardware.
3692  * This is also called by the gennetlink interface when reconfiguring
3693  * the DCB state.
3694  */
3695 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
3696 {
3697 	struct ixgbe_hw *hw = &adapter->hw;
3698 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3699 
3700 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
3701 		if (hw->mac.type == ixgbe_mac_82598EB)
3702 			netif_set_gso_max_size(adapter->netdev, 65536);
3703 		return;
3704 	}
3705 
3706 	if (hw->mac.type == ixgbe_mac_82598EB)
3707 		netif_set_gso_max_size(adapter->netdev, 32768);
3708 
3709 #ifdef IXGBE_FCOE
3710 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
3711 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
3712 #endif
3713 
3714 	/* reconfigure the hardware */
3715 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
3716 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
3717 						DCB_TX_CONFIG);
3718 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
3719 						DCB_RX_CONFIG);
3720 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
3721 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
3722 		ixgbe_dcb_hw_ets(&adapter->hw,
3723 				 adapter->ixgbe_ieee_ets,
3724 				 max_frame);
3725 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
3726 					adapter->ixgbe_ieee_pfc->pfc_en,
3727 					adapter->ixgbe_ieee_ets->prio_tc);
3728 	}
3729 
3730 	/* Enable RSS Hash per TC */
3731 	if (hw->mac.type != ixgbe_mac_82598EB) {
3732 		u32 msb = 0;
3733 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
3734 
3735 		while (rss_i) {
3736 			msb++;
3737 			rss_i >>= 1;
3738 		}
3739 
3740 		/* write msb to all 8 TCs in one write */
3741 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
3742 	}
3743 }
3744 #endif
3745 
3746 /* Additional bittime to account for IXGBE framing */
3747 #define IXGBE_ETH_FRAMING 20
3748 
3749 /**
3750  * ixgbe_hpbthresh - calculate high water mark for flow control
3751  *
3752  * @adapter: board private structure to calculate for
3753  * @pb: packet buffer to calculate
3754  */
3755 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
3756 {
3757 	struct ixgbe_hw *hw = &adapter->hw;
3758 	struct net_device *dev = adapter->netdev;
3759 	int link, tc, kb, marker;
3760 	u32 dv_id, rx_pba;
3761 
3762 	/* Calculate max LAN frame size */
3763 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
3764 
3765 #ifdef IXGBE_FCOE
3766 	/* FCoE traffic class uses FCOE jumbo frames */
3767 	if ((dev->features & NETIF_F_FCOE_MTU) &&
3768 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
3769 	    (pb == ixgbe_fcoe_get_tc(adapter)))
3770 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3771 
3772 #endif
3773 	/* Calculate delay value for device */
3774 	switch (hw->mac.type) {
3775 	case ixgbe_mac_X540:
3776 		dv_id = IXGBE_DV_X540(link, tc);
3777 		break;
3778 	default:
3779 		dv_id = IXGBE_DV(link, tc);
3780 		break;
3781 	}
3782 
3783 	/* Loopback switch introduces additional latency */
3784 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3785 		dv_id += IXGBE_B2BT(tc);
3786 
3787 	/* Delay value is calculated in bit times convert to KB */
3788 	kb = IXGBE_BT2KB(dv_id);
3789 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
3790 
3791 	marker = rx_pba - kb;
3792 
3793 	/* It is possible that the packet buffer is not large enough
3794 	 * to provide required headroom. In this case throw an error
3795 	 * to user and a do the best we can.
3796 	 */
3797 	if (marker < 0) {
3798 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
3799 			    "headroom to support flow control."
3800 			    "Decrease MTU or number of traffic classes\n", pb);
3801 		marker = tc + 1;
3802 	}
3803 
3804 	return marker;
3805 }
3806 
3807 /**
3808  * ixgbe_lpbthresh - calculate low water mark for for flow control
3809  *
3810  * @adapter: board private structure to calculate for
3811  * @pb: packet buffer to calculate
3812  */
3813 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter)
3814 {
3815 	struct ixgbe_hw *hw = &adapter->hw;
3816 	struct net_device *dev = adapter->netdev;
3817 	int tc;
3818 	u32 dv_id;
3819 
3820 	/* Calculate max LAN frame size */
3821 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
3822 
3823 	/* Calculate delay value for device */
3824 	switch (hw->mac.type) {
3825 	case ixgbe_mac_X540:
3826 		dv_id = IXGBE_LOW_DV_X540(tc);
3827 		break;
3828 	default:
3829 		dv_id = IXGBE_LOW_DV(tc);
3830 		break;
3831 	}
3832 
3833 	/* Delay value is calculated in bit times convert to KB */
3834 	return IXGBE_BT2KB(dv_id);
3835 }
3836 
3837 /*
3838  * ixgbe_pbthresh_setup - calculate and setup high low water marks
3839  */
3840 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
3841 {
3842 	struct ixgbe_hw *hw = &adapter->hw;
3843 	int num_tc = netdev_get_num_tc(adapter->netdev);
3844 	int i;
3845 
3846 	if (!num_tc)
3847 		num_tc = 1;
3848 
3849 	hw->fc.low_water = ixgbe_lpbthresh(adapter);
3850 
3851 	for (i = 0; i < num_tc; i++) {
3852 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
3853 
3854 		/* Low water marks must not be larger than high water marks */
3855 		if (hw->fc.low_water > hw->fc.high_water[i])
3856 			hw->fc.low_water = 0;
3857 	}
3858 }
3859 
3860 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
3861 {
3862 	struct ixgbe_hw *hw = &adapter->hw;
3863 	int hdrm;
3864 	u8 tc = netdev_get_num_tc(adapter->netdev);
3865 
3866 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
3867 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
3868 		hdrm = 32 << adapter->fdir_pballoc;
3869 	else
3870 		hdrm = 0;
3871 
3872 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
3873 	ixgbe_pbthresh_setup(adapter);
3874 }
3875 
3876 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
3877 {
3878 	struct ixgbe_hw *hw = &adapter->hw;
3879 	struct hlist_node *node, *node2;
3880 	struct ixgbe_fdir_filter *filter;
3881 
3882 	spin_lock(&adapter->fdir_perfect_lock);
3883 
3884 	if (!hlist_empty(&adapter->fdir_filter_list))
3885 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
3886 
3887 	hlist_for_each_entry_safe(filter, node, node2,
3888 				  &adapter->fdir_filter_list, fdir_node) {
3889 		ixgbe_fdir_write_perfect_filter_82599(hw,
3890 				&filter->filter,
3891 				filter->sw_idx,
3892 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
3893 				IXGBE_FDIR_DROP_QUEUE :
3894 				adapter->rx_ring[filter->action]->reg_idx);
3895 	}
3896 
3897 	spin_unlock(&adapter->fdir_perfect_lock);
3898 }
3899 
3900 static void ixgbe_configure(struct ixgbe_adapter *adapter)
3901 {
3902 	struct ixgbe_hw *hw = &adapter->hw;
3903 
3904 	ixgbe_configure_pb(adapter);
3905 #ifdef CONFIG_IXGBE_DCB
3906 	ixgbe_configure_dcb(adapter);
3907 #endif
3908 	/*
3909 	 * We must restore virtualization before VLANs or else
3910 	 * the VLVF registers will not be populated
3911 	 */
3912 	ixgbe_configure_virtualization(adapter);
3913 
3914 	ixgbe_set_rx_mode(adapter->netdev);
3915 	ixgbe_restore_vlan(adapter);
3916 
3917 	switch (hw->mac.type) {
3918 	case ixgbe_mac_82599EB:
3919 	case ixgbe_mac_X540:
3920 		hw->mac.ops.disable_rx_buff(hw);
3921 		break;
3922 	default:
3923 		break;
3924 	}
3925 
3926 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3927 		ixgbe_init_fdir_signature_82599(&adapter->hw,
3928 						adapter->fdir_pballoc);
3929 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
3930 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
3931 					      adapter->fdir_pballoc);
3932 		ixgbe_fdir_filter_restore(adapter);
3933 	}
3934 
3935 	switch (hw->mac.type) {
3936 	case ixgbe_mac_82599EB:
3937 	case ixgbe_mac_X540:
3938 		hw->mac.ops.enable_rx_buff(hw);
3939 		break;
3940 	default:
3941 		break;
3942 	}
3943 
3944 #ifdef IXGBE_FCOE
3945 	/* configure FCoE L2 filters, redirection table, and Rx control */
3946 	ixgbe_configure_fcoe(adapter);
3947 
3948 #endif /* IXGBE_FCOE */
3949 	ixgbe_configure_tx(adapter);
3950 	ixgbe_configure_rx(adapter);
3951 }
3952 
3953 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
3954 {
3955 	switch (hw->phy.type) {
3956 	case ixgbe_phy_sfp_avago:
3957 	case ixgbe_phy_sfp_ftl:
3958 	case ixgbe_phy_sfp_intel:
3959 	case ixgbe_phy_sfp_unknown:
3960 	case ixgbe_phy_sfp_passive_tyco:
3961 	case ixgbe_phy_sfp_passive_unknown:
3962 	case ixgbe_phy_sfp_active_unknown:
3963 	case ixgbe_phy_sfp_ftl_active:
3964 		return true;
3965 	case ixgbe_phy_nl:
3966 		if (hw->mac.type == ixgbe_mac_82598EB)
3967 			return true;
3968 	default:
3969 		return false;
3970 	}
3971 }
3972 
3973 /**
3974  * ixgbe_sfp_link_config - set up SFP+ link
3975  * @adapter: pointer to private adapter struct
3976  **/
3977 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
3978 {
3979 	/*
3980 	 * We are assuming the worst case scenario here, and that
3981 	 * is that an SFP was inserted/removed after the reset
3982 	 * but before SFP detection was enabled.  As such the best
3983 	 * solution is to just start searching as soon as we start
3984 	 */
3985 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3986 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
3987 
3988 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
3989 }
3990 
3991 /**
3992  * ixgbe_non_sfp_link_config - set up non-SFP+ link
3993  * @hw: pointer to private hardware struct
3994  *
3995  * Returns 0 on success, negative on failure
3996  **/
3997 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
3998 {
3999 	u32 autoneg;
4000 	bool negotiation, link_up = false;
4001 	u32 ret = IXGBE_ERR_LINK_SETUP;
4002 
4003 	if (hw->mac.ops.check_link)
4004 		ret = hw->mac.ops.check_link(hw, &autoneg, &link_up, false);
4005 
4006 	if (ret)
4007 		goto link_cfg_out;
4008 
4009 	autoneg = hw->phy.autoneg_advertised;
4010 	if ((!autoneg) && (hw->mac.ops.get_link_capabilities))
4011 		ret = hw->mac.ops.get_link_capabilities(hw, &autoneg,
4012 							&negotiation);
4013 	if (ret)
4014 		goto link_cfg_out;
4015 
4016 	if (hw->mac.ops.setup_link)
4017 		ret = hw->mac.ops.setup_link(hw, autoneg, negotiation, link_up);
4018 link_cfg_out:
4019 	return ret;
4020 }
4021 
4022 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
4023 {
4024 	struct ixgbe_hw *hw = &adapter->hw;
4025 	u32 gpie = 0;
4026 
4027 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
4028 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
4029 		       IXGBE_GPIE_OCD;
4030 		gpie |= IXGBE_GPIE_EIAME;
4031 		/*
4032 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
4033 		 * this saves a register write for every interrupt
4034 		 */
4035 		switch (hw->mac.type) {
4036 		case ixgbe_mac_82598EB:
4037 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4038 			break;
4039 		case ixgbe_mac_82599EB:
4040 		case ixgbe_mac_X540:
4041 		default:
4042 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
4043 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
4044 			break;
4045 		}
4046 	} else {
4047 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
4048 		 * specifically only auto mask tx and rx interrupts */
4049 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4050 	}
4051 
4052 	/* XXX: to interrupt immediately for EICS writes, enable this */
4053 	/* gpie |= IXGBE_GPIE_EIMEN; */
4054 
4055 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
4056 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
4057 
4058 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4059 		case IXGBE_82599_VMDQ_8Q_MASK:
4060 			gpie |= IXGBE_GPIE_VTMODE_16;
4061 			break;
4062 		case IXGBE_82599_VMDQ_4Q_MASK:
4063 			gpie |= IXGBE_GPIE_VTMODE_32;
4064 			break;
4065 		default:
4066 			gpie |= IXGBE_GPIE_VTMODE_64;
4067 			break;
4068 		}
4069 	}
4070 
4071 	/* Enable Thermal over heat sensor interrupt */
4072 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
4073 		switch (adapter->hw.mac.type) {
4074 		case ixgbe_mac_82599EB:
4075 			gpie |= IXGBE_SDP0_GPIEN;
4076 			break;
4077 		case ixgbe_mac_X540:
4078 			gpie |= IXGBE_EIMS_TS;
4079 			break;
4080 		default:
4081 			break;
4082 		}
4083 	}
4084 
4085 	/* Enable fan failure interrupt */
4086 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
4087 		gpie |= IXGBE_SDP1_GPIEN;
4088 
4089 	if (hw->mac.type == ixgbe_mac_82599EB) {
4090 		gpie |= IXGBE_SDP1_GPIEN;
4091 		gpie |= IXGBE_SDP2_GPIEN;
4092 	}
4093 
4094 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
4095 }
4096 
4097 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
4098 {
4099 	struct ixgbe_hw *hw = &adapter->hw;
4100 	int err;
4101 	u32 ctrl_ext;
4102 
4103 	ixgbe_get_hw_control(adapter);
4104 	ixgbe_setup_gpie(adapter);
4105 
4106 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
4107 		ixgbe_configure_msix(adapter);
4108 	else
4109 		ixgbe_configure_msi_and_legacy(adapter);
4110 
4111 	/* enable the optics for 82599 SFP+ fiber */
4112 	if (hw->mac.ops.enable_tx_laser)
4113 		hw->mac.ops.enable_tx_laser(hw);
4114 
4115 	clear_bit(__IXGBE_DOWN, &adapter->state);
4116 	ixgbe_napi_enable_all(adapter);
4117 
4118 	if (ixgbe_is_sfp(hw)) {
4119 		ixgbe_sfp_link_config(adapter);
4120 	} else {
4121 		err = ixgbe_non_sfp_link_config(hw);
4122 		if (err)
4123 			e_err(probe, "link_config FAILED %d\n", err);
4124 	}
4125 
4126 	/* clear any pending interrupts, may auto mask */
4127 	IXGBE_READ_REG(hw, IXGBE_EICR);
4128 	ixgbe_irq_enable(adapter, true, true);
4129 
4130 	/*
4131 	 * If this adapter has a fan, check to see if we had a failure
4132 	 * before we enabled the interrupt.
4133 	 */
4134 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
4135 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
4136 		if (esdp & IXGBE_ESDP_SDP1)
4137 			e_crit(drv, "Fan has stopped, replace the adapter\n");
4138 	}
4139 
4140 	/* enable transmits */
4141 	netif_tx_start_all_queues(adapter->netdev);
4142 
4143 	/* bring the link up in the watchdog, this could race with our first
4144 	 * link up interrupt but shouldn't be a problem */
4145 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4146 	adapter->link_check_timeout = jiffies;
4147 	mod_timer(&adapter->service_timer, jiffies);
4148 
4149 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
4150 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
4151 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
4152 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
4153 }
4154 
4155 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
4156 {
4157 	WARN_ON(in_interrupt());
4158 	/* put off any impending NetWatchDogTimeout */
4159 	adapter->netdev->trans_start = jiffies;
4160 
4161 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
4162 		usleep_range(1000, 2000);
4163 	ixgbe_down(adapter);
4164 	/*
4165 	 * If SR-IOV enabled then wait a bit before bringing the adapter
4166 	 * back up to give the VFs time to respond to the reset.  The
4167 	 * two second wait is based upon the watchdog timer cycle in
4168 	 * the VF driver.
4169 	 */
4170 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4171 		msleep(2000);
4172 	ixgbe_up(adapter);
4173 	clear_bit(__IXGBE_RESETTING, &adapter->state);
4174 }
4175 
4176 void ixgbe_up(struct ixgbe_adapter *adapter)
4177 {
4178 	/* hardware has been reset, we need to reload some things */
4179 	ixgbe_configure(adapter);
4180 
4181 	ixgbe_up_complete(adapter);
4182 }
4183 
4184 void ixgbe_reset(struct ixgbe_adapter *adapter)
4185 {
4186 	struct ixgbe_hw *hw = &adapter->hw;
4187 	int err;
4188 
4189 	/* lock SFP init bit to prevent race conditions with the watchdog */
4190 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
4191 		usleep_range(1000, 2000);
4192 
4193 	/* clear all SFP and link config related flags while holding SFP_INIT */
4194 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
4195 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
4196 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4197 
4198 	err = hw->mac.ops.init_hw(hw);
4199 	switch (err) {
4200 	case 0:
4201 	case IXGBE_ERR_SFP_NOT_PRESENT:
4202 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
4203 		break;
4204 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
4205 		e_dev_err("master disable timed out\n");
4206 		break;
4207 	case IXGBE_ERR_EEPROM_VERSION:
4208 		/* We are running on a pre-production device, log a warning */
4209 		e_dev_warn("This device is a pre-production adapter/LOM. "
4210 			   "Please be aware there may be issues associated with "
4211 			   "your hardware.  If you are experiencing problems "
4212 			   "please contact your Intel or hardware "
4213 			   "representative who provided you with this "
4214 			   "hardware.\n");
4215 		break;
4216 	default:
4217 		e_dev_err("Hardware Error: %d\n", err);
4218 	}
4219 
4220 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
4221 
4222 	/* reprogram the RAR[0] in case user changed it. */
4223 	hw->mac.ops.set_rar(hw, 0, hw->mac.addr, VMDQ_P(0), IXGBE_RAH_AV);
4224 
4225 	/* update SAN MAC vmdq pool selection */
4226 	if (hw->mac.san_mac_rar_index)
4227 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
4228 
4229 	if (adapter->flags2 & IXGBE_FLAG2_PTP_ENABLED)
4230 		ixgbe_ptp_reset(adapter);
4231 }
4232 
4233 /**
4234  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4235  * @rx_ring: ring to free buffers from
4236  **/
4237 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
4238 {
4239 	struct device *dev = rx_ring->dev;
4240 	unsigned long size;
4241 	u16 i;
4242 
4243 	/* ring already cleared, nothing to do */
4244 	if (!rx_ring->rx_buffer_info)
4245 		return;
4246 
4247 	/* Free all the Rx ring sk_buffs */
4248 	for (i = 0; i < rx_ring->count; i++) {
4249 		struct ixgbe_rx_buffer *rx_buffer;
4250 
4251 		rx_buffer = &rx_ring->rx_buffer_info[i];
4252 		if (rx_buffer->skb) {
4253 			struct sk_buff *skb = rx_buffer->skb;
4254 			if (IXGBE_CB(skb)->page_released) {
4255 				dma_unmap_page(dev,
4256 					       IXGBE_CB(skb)->dma,
4257 					       ixgbe_rx_bufsz(rx_ring),
4258 					       DMA_FROM_DEVICE);
4259 				IXGBE_CB(skb)->page_released = false;
4260 			}
4261 			dev_kfree_skb(skb);
4262 		}
4263 		rx_buffer->skb = NULL;
4264 		if (rx_buffer->dma)
4265 			dma_unmap_page(dev, rx_buffer->dma,
4266 				       ixgbe_rx_pg_size(rx_ring),
4267 				       DMA_FROM_DEVICE);
4268 		rx_buffer->dma = 0;
4269 		if (rx_buffer->page)
4270 			__free_pages(rx_buffer->page,
4271 				     ixgbe_rx_pg_order(rx_ring));
4272 		rx_buffer->page = NULL;
4273 	}
4274 
4275 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4276 	memset(rx_ring->rx_buffer_info, 0, size);
4277 
4278 	/* Zero out the descriptor ring */
4279 	memset(rx_ring->desc, 0, rx_ring->size);
4280 
4281 	rx_ring->next_to_alloc = 0;
4282 	rx_ring->next_to_clean = 0;
4283 	rx_ring->next_to_use = 0;
4284 }
4285 
4286 /**
4287  * ixgbe_clean_tx_ring - Free Tx Buffers
4288  * @tx_ring: ring to be cleaned
4289  **/
4290 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
4291 {
4292 	struct ixgbe_tx_buffer *tx_buffer_info;
4293 	unsigned long size;
4294 	u16 i;
4295 
4296 	/* ring already cleared, nothing to do */
4297 	if (!tx_ring->tx_buffer_info)
4298 		return;
4299 
4300 	/* Free all the Tx ring sk_buffs */
4301 	for (i = 0; i < tx_ring->count; i++) {
4302 		tx_buffer_info = &tx_ring->tx_buffer_info[i];
4303 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer_info);
4304 	}
4305 
4306 	netdev_tx_reset_queue(txring_txq(tx_ring));
4307 
4308 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
4309 	memset(tx_ring->tx_buffer_info, 0, size);
4310 
4311 	/* Zero out the descriptor ring */
4312 	memset(tx_ring->desc, 0, tx_ring->size);
4313 
4314 	tx_ring->next_to_use = 0;
4315 	tx_ring->next_to_clean = 0;
4316 }
4317 
4318 /**
4319  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
4320  * @adapter: board private structure
4321  **/
4322 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
4323 {
4324 	int i;
4325 
4326 	for (i = 0; i < adapter->num_rx_queues; i++)
4327 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
4328 }
4329 
4330 /**
4331  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
4332  * @adapter: board private structure
4333  **/
4334 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
4335 {
4336 	int i;
4337 
4338 	for (i = 0; i < adapter->num_tx_queues; i++)
4339 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
4340 }
4341 
4342 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
4343 {
4344 	struct hlist_node *node, *node2;
4345 	struct ixgbe_fdir_filter *filter;
4346 
4347 	spin_lock(&adapter->fdir_perfect_lock);
4348 
4349 	hlist_for_each_entry_safe(filter, node, node2,
4350 				  &adapter->fdir_filter_list, fdir_node) {
4351 		hlist_del(&filter->fdir_node);
4352 		kfree(filter);
4353 	}
4354 	adapter->fdir_filter_count = 0;
4355 
4356 	spin_unlock(&adapter->fdir_perfect_lock);
4357 }
4358 
4359 void ixgbe_down(struct ixgbe_adapter *adapter)
4360 {
4361 	struct net_device *netdev = adapter->netdev;
4362 	struct ixgbe_hw *hw = &adapter->hw;
4363 	u32 rxctrl;
4364 	int i;
4365 
4366 	/* signal that we are down to the interrupt handler */
4367 	set_bit(__IXGBE_DOWN, &adapter->state);
4368 
4369 	/* disable receives */
4370 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4371 	IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxctrl & ~IXGBE_RXCTRL_RXEN);
4372 
4373 	/* disable all enabled rx queues */
4374 	for (i = 0; i < adapter->num_rx_queues; i++)
4375 		/* this call also flushes the previous write */
4376 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
4377 
4378 	usleep_range(10000, 20000);
4379 
4380 	netif_tx_stop_all_queues(netdev);
4381 
4382 	/* call carrier off first to avoid false dev_watchdog timeouts */
4383 	netif_carrier_off(netdev);
4384 	netif_tx_disable(netdev);
4385 
4386 	ixgbe_irq_disable(adapter);
4387 
4388 	ixgbe_napi_disable_all(adapter);
4389 
4390 	adapter->flags2 &= ~(IXGBE_FLAG2_FDIR_REQUIRES_REINIT |
4391 			     IXGBE_FLAG2_RESET_REQUESTED);
4392 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4393 
4394 	del_timer_sync(&adapter->service_timer);
4395 
4396 	if (adapter->num_vfs) {
4397 		/* Clear EITR Select mapping */
4398 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
4399 
4400 		/* Mark all the VFs as inactive */
4401 		for (i = 0 ; i < adapter->num_vfs; i++)
4402 			adapter->vfinfo[i].clear_to_send = false;
4403 
4404 		/* ping all the active vfs to let them know we are going down */
4405 		ixgbe_ping_all_vfs(adapter);
4406 
4407 		/* Disable all VFTE/VFRE TX/RX */
4408 		ixgbe_disable_tx_rx(adapter);
4409 	}
4410 
4411 	/* disable transmits in the hardware now that interrupts are off */
4412 	for (i = 0; i < adapter->num_tx_queues; i++) {
4413 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
4414 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
4415 	}
4416 
4417 	/* Disable the Tx DMA engine on 82599 and X540 */
4418 	switch (hw->mac.type) {
4419 	case ixgbe_mac_82599EB:
4420 	case ixgbe_mac_X540:
4421 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
4422 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
4423 				 ~IXGBE_DMATXCTL_TE));
4424 		break;
4425 	default:
4426 		break;
4427 	}
4428 
4429 	if (!pci_channel_offline(adapter->pdev))
4430 		ixgbe_reset(adapter);
4431 
4432 	/* power down the optics for 82599 SFP+ fiber */
4433 	if (hw->mac.ops.disable_tx_laser)
4434 		hw->mac.ops.disable_tx_laser(hw);
4435 
4436 	ixgbe_clean_all_tx_rings(adapter);
4437 	ixgbe_clean_all_rx_rings(adapter);
4438 
4439 #ifdef CONFIG_IXGBE_DCA
4440 	/* since we reset the hardware DCA settings were cleared */
4441 	ixgbe_setup_dca(adapter);
4442 #endif
4443 }
4444 
4445 /**
4446  * ixgbe_tx_timeout - Respond to a Tx Hang
4447  * @netdev: network interface device structure
4448  **/
4449 static void ixgbe_tx_timeout(struct net_device *netdev)
4450 {
4451 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4452 
4453 	/* Do the reset outside of interrupt context */
4454 	ixgbe_tx_timeout_reset(adapter);
4455 }
4456 
4457 /**
4458  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
4459  * @adapter: board private structure to initialize
4460  *
4461  * ixgbe_sw_init initializes the Adapter private data structure.
4462  * Fields are initialized based on PCI device information and
4463  * OS network device settings (MTU size).
4464  **/
4465 static int ixgbe_sw_init(struct ixgbe_adapter *adapter)
4466 {
4467 	struct ixgbe_hw *hw = &adapter->hw;
4468 	struct pci_dev *pdev = adapter->pdev;
4469 	unsigned int rss;
4470 	u32 fwsm;
4471 #ifdef CONFIG_IXGBE_DCB
4472 	int j;
4473 	struct tc_configuration *tc;
4474 #endif
4475 
4476 	/* PCI config space info */
4477 
4478 	hw->vendor_id = pdev->vendor;
4479 	hw->device_id = pdev->device;
4480 	hw->revision_id = pdev->revision;
4481 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
4482 	hw->subsystem_device_id = pdev->subsystem_device;
4483 
4484 	/* Set capability flags */
4485 	rss = min_t(int, IXGBE_MAX_RSS_INDICES, num_online_cpus());
4486 	adapter->ring_feature[RING_F_RSS].limit = rss;
4487 	switch (hw->mac.type) {
4488 	case ixgbe_mac_82598EB:
4489 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
4490 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
4491 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
4492 		break;
4493 	case ixgbe_mac_X540:
4494 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM);
4495 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
4496 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
4497 	case ixgbe_mac_82599EB:
4498 		adapter->max_q_vectors = MAX_Q_VECTORS_82599;
4499 		adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
4500 		adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
4501 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
4502 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
4503 		/* Flow Director hash filters enabled */
4504 		adapter->atr_sample_rate = 20;
4505 		adapter->ring_feature[RING_F_FDIR].limit =
4506 							 IXGBE_MAX_FDIR_INDICES;
4507 		adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
4508 #ifdef IXGBE_FCOE
4509 		adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
4510 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
4511 #ifdef CONFIG_IXGBE_DCB
4512 		/* Default traffic class to use for FCoE */
4513 		adapter->fcoe.up = IXGBE_FCOE_DEFTC;
4514 #endif
4515 #endif /* IXGBE_FCOE */
4516 		break;
4517 	default:
4518 		break;
4519 	}
4520 
4521 #ifdef IXGBE_FCOE
4522 	/* FCoE support exists, always init the FCoE lock */
4523 	spin_lock_init(&adapter->fcoe.lock);
4524 
4525 #endif
4526 	/* n-tuple support exists, always init our spinlock */
4527 	spin_lock_init(&adapter->fdir_perfect_lock);
4528 
4529 #ifdef CONFIG_IXGBE_DCB
4530 	switch (hw->mac.type) {
4531 	case ixgbe_mac_X540:
4532 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
4533 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
4534 		break;
4535 	default:
4536 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
4537 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
4538 		break;
4539 	}
4540 
4541 	/* Configure DCB traffic classes */
4542 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
4543 		tc = &adapter->dcb_cfg.tc_config[j];
4544 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
4545 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
4546 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
4547 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
4548 		tc->dcb_pfc = pfc_disabled;
4549 	}
4550 
4551 	/* Initialize default user to priority mapping, UPx->TC0 */
4552 	tc = &adapter->dcb_cfg.tc_config[0];
4553 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
4554 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
4555 
4556 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
4557 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
4558 	adapter->dcb_cfg.pfc_mode_enable = false;
4559 	adapter->dcb_set_bitmap = 0x00;
4560 	adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
4561 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
4562 	       sizeof(adapter->temp_dcb_cfg));
4563 
4564 #endif
4565 
4566 	/* default flow control settings */
4567 	hw->fc.requested_mode = ixgbe_fc_full;
4568 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
4569 	ixgbe_pbthresh_setup(adapter);
4570 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
4571 	hw->fc.send_xon = true;
4572 	hw->fc.disable_fc_autoneg =
4573 		(ixgbe_device_supports_autoneg_fc(hw) == 0) ? false : true;
4574 
4575 #ifdef CONFIG_PCI_IOV
4576 	/* assign number of SR-IOV VFs */
4577 	if (hw->mac.type != ixgbe_mac_82598EB)
4578 		adapter->num_vfs = (max_vfs > 63) ? 0 : max_vfs;
4579 
4580 #endif
4581 	/* enable itr by default in dynamic mode */
4582 	adapter->rx_itr_setting = 1;
4583 	adapter->tx_itr_setting = 1;
4584 
4585 	/* set default ring sizes */
4586 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
4587 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
4588 
4589 	/* set default work limits */
4590 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
4591 
4592 	/* initialize eeprom parameters */
4593 	if (ixgbe_init_eeprom_params_generic(hw)) {
4594 		e_dev_err("EEPROM initialization failed\n");
4595 		return -EIO;
4596 	}
4597 
4598 	set_bit(__IXGBE_DOWN, &adapter->state);
4599 
4600 	return 0;
4601 }
4602 
4603 /**
4604  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
4605  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
4606  *
4607  * Return 0 on success, negative on failure
4608  **/
4609 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
4610 {
4611 	struct device *dev = tx_ring->dev;
4612 	int orig_node = dev_to_node(dev);
4613 	int numa_node = -1;
4614 	int size;
4615 
4616 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
4617 
4618 	if (tx_ring->q_vector)
4619 		numa_node = tx_ring->q_vector->numa_node;
4620 
4621 	tx_ring->tx_buffer_info = vzalloc_node(size, numa_node);
4622 	if (!tx_ring->tx_buffer_info)
4623 		tx_ring->tx_buffer_info = vzalloc(size);
4624 	if (!tx_ring->tx_buffer_info)
4625 		goto err;
4626 
4627 	/* round up to nearest 4K */
4628 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
4629 	tx_ring->size = ALIGN(tx_ring->size, 4096);
4630 
4631 	set_dev_node(dev, numa_node);
4632 	tx_ring->desc = dma_alloc_coherent(dev,
4633 					   tx_ring->size,
4634 					   &tx_ring->dma,
4635 					   GFP_KERNEL);
4636 	set_dev_node(dev, orig_node);
4637 	if (!tx_ring->desc)
4638 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
4639 						   &tx_ring->dma, GFP_KERNEL);
4640 	if (!tx_ring->desc)
4641 		goto err;
4642 
4643 	tx_ring->next_to_use = 0;
4644 	tx_ring->next_to_clean = 0;
4645 	return 0;
4646 
4647 err:
4648 	vfree(tx_ring->tx_buffer_info);
4649 	tx_ring->tx_buffer_info = NULL;
4650 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
4651 	return -ENOMEM;
4652 }
4653 
4654 /**
4655  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
4656  * @adapter: board private structure
4657  *
4658  * If this function returns with an error, then it's possible one or
4659  * more of the rings is populated (while the rest are not).  It is the
4660  * callers duty to clean those orphaned rings.
4661  *
4662  * Return 0 on success, negative on failure
4663  **/
4664 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
4665 {
4666 	int i, err = 0;
4667 
4668 	for (i = 0; i < adapter->num_tx_queues; i++) {
4669 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
4670 		if (!err)
4671 			continue;
4672 
4673 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
4674 		goto err_setup_tx;
4675 	}
4676 
4677 	return 0;
4678 err_setup_tx:
4679 	/* rewind the index freeing the rings as we go */
4680 	while (i--)
4681 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
4682 	return err;
4683 }
4684 
4685 /**
4686  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
4687  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
4688  *
4689  * Returns 0 on success, negative on failure
4690  **/
4691 int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
4692 {
4693 	struct device *dev = rx_ring->dev;
4694 	int orig_node = dev_to_node(dev);
4695 	int numa_node = -1;
4696 	int size;
4697 
4698 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4699 
4700 	if (rx_ring->q_vector)
4701 		numa_node = rx_ring->q_vector->numa_node;
4702 
4703 	rx_ring->rx_buffer_info = vzalloc_node(size, numa_node);
4704 	if (!rx_ring->rx_buffer_info)
4705 		rx_ring->rx_buffer_info = vzalloc(size);
4706 	if (!rx_ring->rx_buffer_info)
4707 		goto err;
4708 
4709 	/* Round up to nearest 4K */
4710 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
4711 	rx_ring->size = ALIGN(rx_ring->size, 4096);
4712 
4713 	set_dev_node(dev, numa_node);
4714 	rx_ring->desc = dma_alloc_coherent(dev,
4715 					   rx_ring->size,
4716 					   &rx_ring->dma,
4717 					   GFP_KERNEL);
4718 	set_dev_node(dev, orig_node);
4719 	if (!rx_ring->desc)
4720 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
4721 						   &rx_ring->dma, GFP_KERNEL);
4722 	if (!rx_ring->desc)
4723 		goto err;
4724 
4725 	rx_ring->next_to_clean = 0;
4726 	rx_ring->next_to_use = 0;
4727 
4728 	return 0;
4729 err:
4730 	vfree(rx_ring->rx_buffer_info);
4731 	rx_ring->rx_buffer_info = NULL;
4732 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
4733 	return -ENOMEM;
4734 }
4735 
4736 /**
4737  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
4738  * @adapter: board private structure
4739  *
4740  * If this function returns with an error, then it's possible one or
4741  * more of the rings is populated (while the rest are not).  It is the
4742  * callers duty to clean those orphaned rings.
4743  *
4744  * Return 0 on success, negative on failure
4745  **/
4746 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
4747 {
4748 	int i, err = 0;
4749 
4750 	for (i = 0; i < adapter->num_rx_queues; i++) {
4751 		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
4752 		if (!err)
4753 			continue;
4754 
4755 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
4756 		goto err_setup_rx;
4757 	}
4758 
4759 #ifdef IXGBE_FCOE
4760 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
4761 	if (!err)
4762 #endif
4763 		return 0;
4764 err_setup_rx:
4765 	/* rewind the index freeing the rings as we go */
4766 	while (i--)
4767 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
4768 	return err;
4769 }
4770 
4771 /**
4772  * ixgbe_free_tx_resources - Free Tx Resources per Queue
4773  * @tx_ring: Tx descriptor ring for a specific queue
4774  *
4775  * Free all transmit software resources
4776  **/
4777 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
4778 {
4779 	ixgbe_clean_tx_ring(tx_ring);
4780 
4781 	vfree(tx_ring->tx_buffer_info);
4782 	tx_ring->tx_buffer_info = NULL;
4783 
4784 	/* if not set, then don't free */
4785 	if (!tx_ring->desc)
4786 		return;
4787 
4788 	dma_free_coherent(tx_ring->dev, tx_ring->size,
4789 			  tx_ring->desc, tx_ring->dma);
4790 
4791 	tx_ring->desc = NULL;
4792 }
4793 
4794 /**
4795  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
4796  * @adapter: board private structure
4797  *
4798  * Free all transmit software resources
4799  **/
4800 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
4801 {
4802 	int i;
4803 
4804 	for (i = 0; i < adapter->num_tx_queues; i++)
4805 		if (adapter->tx_ring[i]->desc)
4806 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
4807 }
4808 
4809 /**
4810  * ixgbe_free_rx_resources - Free Rx Resources
4811  * @rx_ring: ring to clean the resources from
4812  *
4813  * Free all receive software resources
4814  **/
4815 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
4816 {
4817 	ixgbe_clean_rx_ring(rx_ring);
4818 
4819 	vfree(rx_ring->rx_buffer_info);
4820 	rx_ring->rx_buffer_info = NULL;
4821 
4822 	/* if not set, then don't free */
4823 	if (!rx_ring->desc)
4824 		return;
4825 
4826 	dma_free_coherent(rx_ring->dev, rx_ring->size,
4827 			  rx_ring->desc, rx_ring->dma);
4828 
4829 	rx_ring->desc = NULL;
4830 }
4831 
4832 /**
4833  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
4834  * @adapter: board private structure
4835  *
4836  * Free all receive software resources
4837  **/
4838 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
4839 {
4840 	int i;
4841 
4842 #ifdef IXGBE_FCOE
4843 	ixgbe_free_fcoe_ddp_resources(adapter);
4844 
4845 #endif
4846 	for (i = 0; i < adapter->num_rx_queues; i++)
4847 		if (adapter->rx_ring[i]->desc)
4848 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
4849 }
4850 
4851 /**
4852  * ixgbe_change_mtu - Change the Maximum Transfer Unit
4853  * @netdev: network interface device structure
4854  * @new_mtu: new value for maximum frame size
4855  *
4856  * Returns 0 on success, negative on failure
4857  **/
4858 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
4859 {
4860 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4861 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
4862 
4863 	/* MTU < 68 is an error and causes problems on some kernels */
4864 	if ((new_mtu < 68) || (max_frame > IXGBE_MAX_JUMBO_FRAME_SIZE))
4865 		return -EINVAL;
4866 
4867 	/*
4868 	 * For 82599EB we cannot allow legacy VFs to enable their receive
4869 	 * paths when MTU greater than 1500 is configured.  So display a
4870 	 * warning that legacy VFs will be disabled.
4871 	 */
4872 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
4873 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
4874 	    (max_frame > MAXIMUM_ETHERNET_VLAN_SIZE))
4875 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
4876 
4877 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
4878 
4879 	/* must set new MTU before calling down or up */
4880 	netdev->mtu = new_mtu;
4881 
4882 	if (netif_running(netdev))
4883 		ixgbe_reinit_locked(adapter);
4884 
4885 	return 0;
4886 }
4887 
4888 /**
4889  * ixgbe_open - Called when a network interface is made active
4890  * @netdev: network interface device structure
4891  *
4892  * Returns 0 on success, negative value on failure
4893  *
4894  * The open entry point is called when a network interface is made
4895  * active by the system (IFF_UP).  At this point all resources needed
4896  * for transmit and receive operations are allocated, the interrupt
4897  * handler is registered with the OS, the watchdog timer is started,
4898  * and the stack is notified that the interface is ready.
4899  **/
4900 static int ixgbe_open(struct net_device *netdev)
4901 {
4902 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4903 	int err;
4904 
4905 	/* disallow open during test */
4906 	if (test_bit(__IXGBE_TESTING, &adapter->state))
4907 		return -EBUSY;
4908 
4909 	netif_carrier_off(netdev);
4910 
4911 	/* allocate transmit descriptors */
4912 	err = ixgbe_setup_all_tx_resources(adapter);
4913 	if (err)
4914 		goto err_setup_tx;
4915 
4916 	/* allocate receive descriptors */
4917 	err = ixgbe_setup_all_rx_resources(adapter);
4918 	if (err)
4919 		goto err_setup_rx;
4920 
4921 	ixgbe_configure(adapter);
4922 
4923 	err = ixgbe_request_irq(adapter);
4924 	if (err)
4925 		goto err_req_irq;
4926 
4927 	/* Notify the stack of the actual queue counts. */
4928 	err = netif_set_real_num_tx_queues(netdev,
4929 					   adapter->num_rx_pools > 1 ? 1 :
4930 					   adapter->num_tx_queues);
4931 	if (err)
4932 		goto err_set_queues;
4933 
4934 
4935 	err = netif_set_real_num_rx_queues(netdev,
4936 					   adapter->num_rx_pools > 1 ? 1 :
4937 					   adapter->num_rx_queues);
4938 	if (err)
4939 		goto err_set_queues;
4940 
4941 	ixgbe_ptp_init(adapter);
4942 
4943 	ixgbe_up_complete(adapter);
4944 
4945 	return 0;
4946 
4947 err_set_queues:
4948 	ixgbe_free_irq(adapter);
4949 err_req_irq:
4950 	ixgbe_free_all_rx_resources(adapter);
4951 err_setup_rx:
4952 	ixgbe_free_all_tx_resources(adapter);
4953 err_setup_tx:
4954 	ixgbe_reset(adapter);
4955 
4956 	return err;
4957 }
4958 
4959 /**
4960  * ixgbe_close - Disables a network interface
4961  * @netdev: network interface device structure
4962  *
4963  * Returns 0, this is not allowed to fail
4964  *
4965  * The close entry point is called when an interface is de-activated
4966  * by the OS.  The hardware is still under the drivers control, but
4967  * needs to be disabled.  A global MAC reset is issued to stop the
4968  * hardware, and all transmit and receive resources are freed.
4969  **/
4970 static int ixgbe_close(struct net_device *netdev)
4971 {
4972 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4973 
4974 	ixgbe_ptp_stop(adapter);
4975 
4976 	ixgbe_down(adapter);
4977 	ixgbe_free_irq(adapter);
4978 
4979 	ixgbe_fdir_filter_exit(adapter);
4980 
4981 	ixgbe_free_all_tx_resources(adapter);
4982 	ixgbe_free_all_rx_resources(adapter);
4983 
4984 	ixgbe_release_hw_control(adapter);
4985 
4986 	return 0;
4987 }
4988 
4989 #ifdef CONFIG_PM
4990 static int ixgbe_resume(struct pci_dev *pdev)
4991 {
4992 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
4993 	struct net_device *netdev = adapter->netdev;
4994 	u32 err;
4995 
4996 	pci_set_power_state(pdev, PCI_D0);
4997 	pci_restore_state(pdev);
4998 	/*
4999 	 * pci_restore_state clears dev->state_saved so call
5000 	 * pci_save_state to restore it.
5001 	 */
5002 	pci_save_state(pdev);
5003 
5004 	err = pci_enable_device_mem(pdev);
5005 	if (err) {
5006 		e_dev_err("Cannot enable PCI device from suspend\n");
5007 		return err;
5008 	}
5009 	pci_set_master(pdev);
5010 
5011 	pci_wake_from_d3(pdev, false);
5012 
5013 	ixgbe_reset(adapter);
5014 
5015 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
5016 
5017 	rtnl_lock();
5018 	err = ixgbe_init_interrupt_scheme(adapter);
5019 	if (!err && netif_running(netdev))
5020 		err = ixgbe_open(netdev);
5021 
5022 	rtnl_unlock();
5023 
5024 	if (err)
5025 		return err;
5026 
5027 	netif_device_attach(netdev);
5028 
5029 	return 0;
5030 }
5031 #endif /* CONFIG_PM */
5032 
5033 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
5034 {
5035 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
5036 	struct net_device *netdev = adapter->netdev;
5037 	struct ixgbe_hw *hw = &adapter->hw;
5038 	u32 ctrl, fctrl;
5039 	u32 wufc = adapter->wol;
5040 #ifdef CONFIG_PM
5041 	int retval = 0;
5042 #endif
5043 
5044 	netif_device_detach(netdev);
5045 
5046 	if (netif_running(netdev)) {
5047 		rtnl_lock();
5048 		ixgbe_down(adapter);
5049 		ixgbe_free_irq(adapter);
5050 		ixgbe_free_all_tx_resources(adapter);
5051 		ixgbe_free_all_rx_resources(adapter);
5052 		rtnl_unlock();
5053 	}
5054 
5055 	ixgbe_clear_interrupt_scheme(adapter);
5056 
5057 #ifdef CONFIG_PM
5058 	retval = pci_save_state(pdev);
5059 	if (retval)
5060 		return retval;
5061 
5062 #endif
5063 	if (wufc) {
5064 		ixgbe_set_rx_mode(netdev);
5065 
5066 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
5067 		if (hw->mac.ops.enable_tx_laser)
5068 			hw->mac.ops.enable_tx_laser(hw);
5069 
5070 		/* turn on all-multi mode if wake on multicast is enabled */
5071 		if (wufc & IXGBE_WUFC_MC) {
5072 			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
5073 			fctrl |= IXGBE_FCTRL_MPE;
5074 			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
5075 		}
5076 
5077 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
5078 		ctrl |= IXGBE_CTRL_GIO_DIS;
5079 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
5080 
5081 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
5082 	} else {
5083 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
5084 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
5085 	}
5086 
5087 	switch (hw->mac.type) {
5088 	case ixgbe_mac_82598EB:
5089 		pci_wake_from_d3(pdev, false);
5090 		break;
5091 	case ixgbe_mac_82599EB:
5092 	case ixgbe_mac_X540:
5093 		pci_wake_from_d3(pdev, !!wufc);
5094 		break;
5095 	default:
5096 		break;
5097 	}
5098 
5099 	*enable_wake = !!wufc;
5100 
5101 	ixgbe_release_hw_control(adapter);
5102 
5103 	pci_disable_device(pdev);
5104 
5105 	return 0;
5106 }
5107 
5108 #ifdef CONFIG_PM
5109 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
5110 {
5111 	int retval;
5112 	bool wake;
5113 
5114 	retval = __ixgbe_shutdown(pdev, &wake);
5115 	if (retval)
5116 		return retval;
5117 
5118 	if (wake) {
5119 		pci_prepare_to_sleep(pdev);
5120 	} else {
5121 		pci_wake_from_d3(pdev, false);
5122 		pci_set_power_state(pdev, PCI_D3hot);
5123 	}
5124 
5125 	return 0;
5126 }
5127 #endif /* CONFIG_PM */
5128 
5129 static void ixgbe_shutdown(struct pci_dev *pdev)
5130 {
5131 	bool wake;
5132 
5133 	__ixgbe_shutdown(pdev, &wake);
5134 
5135 	if (system_state == SYSTEM_POWER_OFF) {
5136 		pci_wake_from_d3(pdev, wake);
5137 		pci_set_power_state(pdev, PCI_D3hot);
5138 	}
5139 }
5140 
5141 /**
5142  * ixgbe_update_stats - Update the board statistics counters.
5143  * @adapter: board private structure
5144  **/
5145 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
5146 {
5147 	struct net_device *netdev = adapter->netdev;
5148 	struct ixgbe_hw *hw = &adapter->hw;
5149 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
5150 	u64 total_mpc = 0;
5151 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
5152 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
5153 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
5154 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
5155 
5156 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5157 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5158 		return;
5159 
5160 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
5161 		u64 rsc_count = 0;
5162 		u64 rsc_flush = 0;
5163 		for (i = 0; i < adapter->num_rx_queues; i++) {
5164 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
5165 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
5166 		}
5167 		adapter->rsc_total_count = rsc_count;
5168 		adapter->rsc_total_flush = rsc_flush;
5169 	}
5170 
5171 	for (i = 0; i < adapter->num_rx_queues; i++) {
5172 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
5173 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
5174 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
5175 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
5176 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
5177 		bytes += rx_ring->stats.bytes;
5178 		packets += rx_ring->stats.packets;
5179 	}
5180 	adapter->non_eop_descs = non_eop_descs;
5181 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
5182 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
5183 	adapter->hw_csum_rx_error = hw_csum_rx_error;
5184 	netdev->stats.rx_bytes = bytes;
5185 	netdev->stats.rx_packets = packets;
5186 
5187 	bytes = 0;
5188 	packets = 0;
5189 	/* gather some stats to the adapter struct that are per queue */
5190 	for (i = 0; i < adapter->num_tx_queues; i++) {
5191 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
5192 		restart_queue += tx_ring->tx_stats.restart_queue;
5193 		tx_busy += tx_ring->tx_stats.tx_busy;
5194 		bytes += tx_ring->stats.bytes;
5195 		packets += tx_ring->stats.packets;
5196 	}
5197 	adapter->restart_queue = restart_queue;
5198 	adapter->tx_busy = tx_busy;
5199 	netdev->stats.tx_bytes = bytes;
5200 	netdev->stats.tx_packets = packets;
5201 
5202 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
5203 
5204 	/* 8 register reads */
5205 	for (i = 0; i < 8; i++) {
5206 		/* for packet buffers not used, the register should read 0 */
5207 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
5208 		missed_rx += mpc;
5209 		hwstats->mpc[i] += mpc;
5210 		total_mpc += hwstats->mpc[i];
5211 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
5212 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
5213 		switch (hw->mac.type) {
5214 		case ixgbe_mac_82598EB:
5215 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
5216 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
5217 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
5218 			hwstats->pxonrxc[i] +=
5219 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
5220 			break;
5221 		case ixgbe_mac_82599EB:
5222 		case ixgbe_mac_X540:
5223 			hwstats->pxonrxc[i] +=
5224 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
5225 			break;
5226 		default:
5227 			break;
5228 		}
5229 	}
5230 
5231 	/*16 register reads */
5232 	for (i = 0; i < 16; i++) {
5233 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
5234 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
5235 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
5236 		    (hw->mac.type == ixgbe_mac_X540)) {
5237 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
5238 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
5239 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
5240 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
5241 		}
5242 	}
5243 
5244 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
5245 	/* work around hardware counting issue */
5246 	hwstats->gprc -= missed_rx;
5247 
5248 	ixgbe_update_xoff_received(adapter);
5249 
5250 	/* 82598 hardware only has a 32 bit counter in the high register */
5251 	switch (hw->mac.type) {
5252 	case ixgbe_mac_82598EB:
5253 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
5254 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
5255 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
5256 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
5257 		break;
5258 	case ixgbe_mac_X540:
5259 		/* OS2BMC stats are X540 only*/
5260 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
5261 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
5262 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
5263 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
5264 	case ixgbe_mac_82599EB:
5265 		for (i = 0; i < 16; i++)
5266 			adapter->hw_rx_no_dma_resources +=
5267 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
5268 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
5269 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
5270 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
5271 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
5272 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
5273 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
5274 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
5275 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
5276 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
5277 #ifdef IXGBE_FCOE
5278 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
5279 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
5280 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
5281 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
5282 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
5283 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
5284 		/* Add up per cpu counters for total ddp aloc fail */
5285 		if (adapter->fcoe.ddp_pool) {
5286 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
5287 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
5288 			unsigned int cpu;
5289 			u64 noddp = 0, noddp_ext_buff = 0;
5290 			for_each_possible_cpu(cpu) {
5291 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
5292 				noddp += ddp_pool->noddp;
5293 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
5294 			}
5295 			hwstats->fcoe_noddp = noddp;
5296 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
5297 		}
5298 #endif /* IXGBE_FCOE */
5299 		break;
5300 	default:
5301 		break;
5302 	}
5303 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
5304 	hwstats->bprc += bprc;
5305 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
5306 	if (hw->mac.type == ixgbe_mac_82598EB)
5307 		hwstats->mprc -= bprc;
5308 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
5309 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
5310 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
5311 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
5312 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
5313 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
5314 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
5315 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
5316 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
5317 	hwstats->lxontxc += lxon;
5318 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
5319 	hwstats->lxofftxc += lxoff;
5320 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
5321 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
5322 	/*
5323 	 * 82598 errata - tx of flow control packets is included in tx counters
5324 	 */
5325 	xon_off_tot = lxon + lxoff;
5326 	hwstats->gptc -= xon_off_tot;
5327 	hwstats->mptc -= xon_off_tot;
5328 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
5329 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
5330 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
5331 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
5332 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
5333 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
5334 	hwstats->ptc64 -= xon_off_tot;
5335 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
5336 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
5337 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
5338 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
5339 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
5340 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
5341 
5342 	/* Fill out the OS statistics structure */
5343 	netdev->stats.multicast = hwstats->mprc;
5344 
5345 	/* Rx Errors */
5346 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
5347 	netdev->stats.rx_dropped = 0;
5348 	netdev->stats.rx_length_errors = hwstats->rlec;
5349 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
5350 	netdev->stats.rx_missed_errors = total_mpc;
5351 }
5352 
5353 /**
5354  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
5355  * @adapter: pointer to the device adapter structure
5356  **/
5357 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
5358 {
5359 	struct ixgbe_hw *hw = &adapter->hw;
5360 	int i;
5361 
5362 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
5363 		return;
5364 
5365 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
5366 
5367 	/* if interface is down do nothing */
5368 	if (test_bit(__IXGBE_DOWN, &adapter->state))
5369 		return;
5370 
5371 	/* do nothing if we are not using signature filters */
5372 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
5373 		return;
5374 
5375 	adapter->fdir_overflow++;
5376 
5377 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
5378 		for (i = 0; i < adapter->num_tx_queues; i++)
5379 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
5380 			        &(adapter->tx_ring[i]->state));
5381 		/* re-enable flow director interrupts */
5382 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
5383 	} else {
5384 		e_err(probe, "failed to finish FDIR re-initialization, "
5385 		      "ignored adding FDIR ATR filters\n");
5386 	}
5387 }
5388 
5389 /**
5390  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
5391  * @adapter: pointer to the device adapter structure
5392  *
5393  * This function serves two purposes.  First it strobes the interrupt lines
5394  * in order to make certain interrupts are occurring.  Secondly it sets the
5395  * bits needed to check for TX hangs.  As a result we should immediately
5396  * determine if a hang has occurred.
5397  */
5398 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
5399 {
5400 	struct ixgbe_hw *hw = &adapter->hw;
5401 	u64 eics = 0;
5402 	int i;
5403 
5404 	/* If we're down or resetting, just bail */
5405 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5406 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5407 		return;
5408 
5409 	/* Force detection of hung controller */
5410 	if (netif_carrier_ok(adapter->netdev)) {
5411 		for (i = 0; i < adapter->num_tx_queues; i++)
5412 			set_check_for_tx_hang(adapter->tx_ring[i]);
5413 	}
5414 
5415 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
5416 		/*
5417 		 * for legacy and MSI interrupts don't set any bits
5418 		 * that are enabled for EIAM, because this operation
5419 		 * would set *both* EIMS and EICS for any bit in EIAM
5420 		 */
5421 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
5422 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
5423 	} else {
5424 		/* get one bit for every active tx/rx interrupt vector */
5425 		for (i = 0; i < adapter->num_q_vectors; i++) {
5426 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
5427 			if (qv->rx.ring || qv->tx.ring)
5428 				eics |= ((u64)1 << i);
5429 		}
5430 	}
5431 
5432 	/* Cause software interrupt to ensure rings are cleaned */
5433 	ixgbe_irq_rearm_queues(adapter, eics);
5434 
5435 }
5436 
5437 /**
5438  * ixgbe_watchdog_update_link - update the link status
5439  * @adapter: pointer to the device adapter structure
5440  * @link_speed: pointer to a u32 to store the link_speed
5441  **/
5442 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
5443 {
5444 	struct ixgbe_hw *hw = &adapter->hw;
5445 	u32 link_speed = adapter->link_speed;
5446 	bool link_up = adapter->link_up;
5447 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
5448 
5449 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
5450 		return;
5451 
5452 	if (hw->mac.ops.check_link) {
5453 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
5454 	} else {
5455 		/* always assume link is up, if no check link function */
5456 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
5457 		link_up = true;
5458 	}
5459 
5460 	if (adapter->ixgbe_ieee_pfc)
5461 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
5462 
5463 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
5464 		hw->mac.ops.fc_enable(hw);
5465 		ixgbe_set_rx_drop_en(adapter);
5466 	}
5467 
5468 	if (link_up ||
5469 	    time_after(jiffies, (adapter->link_check_timeout +
5470 				 IXGBE_TRY_LINK_TIMEOUT))) {
5471 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5472 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
5473 		IXGBE_WRITE_FLUSH(hw);
5474 	}
5475 
5476 	adapter->link_up = link_up;
5477 	adapter->link_speed = link_speed;
5478 }
5479 
5480 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
5481 {
5482 #ifdef CONFIG_IXGBE_DCB
5483 	struct net_device *netdev = adapter->netdev;
5484 	struct dcb_app app = {
5485 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
5486 			      .protocol = 0,
5487 			     };
5488 	u8 up = 0;
5489 
5490 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
5491 		up = dcb_ieee_getapp_mask(netdev, &app);
5492 
5493 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
5494 #endif
5495 }
5496 
5497 /**
5498  * ixgbe_watchdog_link_is_up - update netif_carrier status and
5499  *                             print link up message
5500  * @adapter: pointer to the device adapter structure
5501  **/
5502 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
5503 {
5504 	struct net_device *netdev = adapter->netdev;
5505 	struct ixgbe_hw *hw = &adapter->hw;
5506 	u32 link_speed = adapter->link_speed;
5507 	bool flow_rx, flow_tx;
5508 
5509 	/* only continue if link was previously down */
5510 	if (netif_carrier_ok(netdev))
5511 		return;
5512 
5513 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
5514 
5515 	switch (hw->mac.type) {
5516 	case ixgbe_mac_82598EB: {
5517 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
5518 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
5519 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
5520 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
5521 	}
5522 		break;
5523 	case ixgbe_mac_X540:
5524 	case ixgbe_mac_82599EB: {
5525 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
5526 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
5527 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
5528 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
5529 	}
5530 		break;
5531 	default:
5532 		flow_tx = false;
5533 		flow_rx = false;
5534 		break;
5535 	}
5536 
5537 	if (adapter->flags2 & IXGBE_FLAG2_PTP_ENABLED)
5538 		ixgbe_ptp_start_cyclecounter(adapter);
5539 
5540 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n",
5541 	       (link_speed == IXGBE_LINK_SPEED_10GB_FULL ?
5542 	       "10 Gbps" :
5543 	       (link_speed == IXGBE_LINK_SPEED_1GB_FULL ?
5544 	       "1 Gbps" :
5545 	       (link_speed == IXGBE_LINK_SPEED_100_FULL ?
5546 	       "100 Mbps" :
5547 	       "unknown speed"))),
5548 	       ((flow_rx && flow_tx) ? "RX/TX" :
5549 	       (flow_rx ? "RX" :
5550 	       (flow_tx ? "TX" : "None"))));
5551 
5552 	netif_carrier_on(netdev);
5553 	ixgbe_check_vf_rate_limit(adapter);
5554 
5555 	/* update the default user priority for VFs */
5556 	ixgbe_update_default_up(adapter);
5557 
5558 	/* ping all the active vfs to let them know link has changed */
5559 	ixgbe_ping_all_vfs(adapter);
5560 }
5561 
5562 /**
5563  * ixgbe_watchdog_link_is_down - update netif_carrier status and
5564  *                               print link down message
5565  * @adapter: pointer to the adapter structure
5566  **/
5567 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
5568 {
5569 	struct net_device *netdev = adapter->netdev;
5570 	struct ixgbe_hw *hw = &adapter->hw;
5571 
5572 	adapter->link_up = false;
5573 	adapter->link_speed = 0;
5574 
5575 	/* only continue if link was up previously */
5576 	if (!netif_carrier_ok(netdev))
5577 		return;
5578 
5579 	/* poll for SFP+ cable when link is down */
5580 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
5581 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5582 
5583 	if (adapter->flags2 & IXGBE_FLAG2_PTP_ENABLED)
5584 		ixgbe_ptp_start_cyclecounter(adapter);
5585 
5586 	e_info(drv, "NIC Link is Down\n");
5587 	netif_carrier_off(netdev);
5588 
5589 	/* ping all the active vfs to let them know link has changed */
5590 	ixgbe_ping_all_vfs(adapter);
5591 }
5592 
5593 /**
5594  * ixgbe_watchdog_flush_tx - flush queues on link down
5595  * @adapter: pointer to the device adapter structure
5596  **/
5597 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
5598 {
5599 	int i;
5600 	int some_tx_pending = 0;
5601 
5602 	if (!netif_carrier_ok(adapter->netdev)) {
5603 		for (i = 0; i < adapter->num_tx_queues; i++) {
5604 			struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
5605 			if (tx_ring->next_to_use != tx_ring->next_to_clean) {
5606 				some_tx_pending = 1;
5607 				break;
5608 			}
5609 		}
5610 
5611 		if (some_tx_pending) {
5612 			/* We've lost link, so the controller stops DMA,
5613 			 * but we've got queued Tx work that's never going
5614 			 * to get done, so reset controller to flush Tx.
5615 			 * (Do the reset outside of interrupt context).
5616 			 */
5617 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
5618 		}
5619 	}
5620 }
5621 
5622 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
5623 {
5624 	u32 ssvpc;
5625 
5626 	/* Do not perform spoof check for 82598 or if not in IOV mode */
5627 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
5628 	    adapter->num_vfs == 0)
5629 		return;
5630 
5631 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
5632 
5633 	/*
5634 	 * ssvpc register is cleared on read, if zero then no
5635 	 * spoofed packets in the last interval.
5636 	 */
5637 	if (!ssvpc)
5638 		return;
5639 
5640 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
5641 }
5642 
5643 /**
5644  * ixgbe_watchdog_subtask - check and bring link up
5645  * @adapter: pointer to the device adapter structure
5646  **/
5647 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
5648 {
5649 	/* if interface is down do nothing */
5650 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5651 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5652 		return;
5653 
5654 	ixgbe_watchdog_update_link(adapter);
5655 
5656 	if (adapter->link_up)
5657 		ixgbe_watchdog_link_is_up(adapter);
5658 	else
5659 		ixgbe_watchdog_link_is_down(adapter);
5660 
5661 	ixgbe_spoof_check(adapter);
5662 	ixgbe_update_stats(adapter);
5663 
5664 	ixgbe_watchdog_flush_tx(adapter);
5665 }
5666 
5667 /**
5668  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
5669  * @adapter: the ixgbe adapter structure
5670  **/
5671 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
5672 {
5673 	struct ixgbe_hw *hw = &adapter->hw;
5674 	s32 err;
5675 
5676 	/* not searching for SFP so there is nothing to do here */
5677 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
5678 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
5679 		return;
5680 
5681 	/* someone else is in init, wait until next service event */
5682 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5683 		return;
5684 
5685 	err = hw->phy.ops.identify_sfp(hw);
5686 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
5687 		goto sfp_out;
5688 
5689 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
5690 		/* If no cable is present, then we need to reset
5691 		 * the next time we find a good cable. */
5692 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5693 	}
5694 
5695 	/* exit on error */
5696 	if (err)
5697 		goto sfp_out;
5698 
5699 	/* exit if reset not needed */
5700 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
5701 		goto sfp_out;
5702 
5703 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
5704 
5705 	/*
5706 	 * A module may be identified correctly, but the EEPROM may not have
5707 	 * support for that module.  setup_sfp() will fail in that case, so
5708 	 * we should not allow that module to load.
5709 	 */
5710 	if (hw->mac.type == ixgbe_mac_82598EB)
5711 		err = hw->phy.ops.reset(hw);
5712 	else
5713 		err = hw->mac.ops.setup_sfp(hw);
5714 
5715 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
5716 		goto sfp_out;
5717 
5718 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
5719 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
5720 
5721 sfp_out:
5722 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5723 
5724 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
5725 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
5726 		e_dev_err("failed to initialize because an unsupported "
5727 			  "SFP+ module type was detected.\n");
5728 		e_dev_err("Reload the driver after installing a "
5729 			  "supported module.\n");
5730 		unregister_netdev(adapter->netdev);
5731 	}
5732 }
5733 
5734 /**
5735  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
5736  * @adapter: the ixgbe adapter structure
5737  **/
5738 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
5739 {
5740 	struct ixgbe_hw *hw = &adapter->hw;
5741 	u32 autoneg;
5742 	bool negotiation;
5743 
5744 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
5745 		return;
5746 
5747 	/* someone else is in init, wait until next service event */
5748 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5749 		return;
5750 
5751 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5752 
5753 	autoneg = hw->phy.autoneg_advertised;
5754 	if ((!autoneg) && (hw->mac.ops.get_link_capabilities))
5755 		hw->mac.ops.get_link_capabilities(hw, &autoneg, &negotiation);
5756 	if (hw->mac.ops.setup_link)
5757 		hw->mac.ops.setup_link(hw, autoneg, negotiation, true);
5758 
5759 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5760 	adapter->link_check_timeout = jiffies;
5761 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5762 }
5763 
5764 #ifdef CONFIG_PCI_IOV
5765 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
5766 {
5767 	int vf;
5768 	struct ixgbe_hw *hw = &adapter->hw;
5769 	struct net_device *netdev = adapter->netdev;
5770 	u32 gpc;
5771 	u32 ciaa, ciad;
5772 
5773 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
5774 	if (gpc) /* If incrementing then no need for the check below */
5775 		return;
5776 	/*
5777 	 * Check to see if a bad DMA write target from an errant or
5778 	 * malicious VF has caused a PCIe error.  If so then we can
5779 	 * issue a VFLR to the offending VF(s) and then resume without
5780 	 * requesting a full slot reset.
5781 	 */
5782 
5783 	for (vf = 0; vf < adapter->num_vfs; vf++) {
5784 		ciaa = (vf << 16) | 0x80000000;
5785 		/* 32 bit read so align, we really want status at offset 6 */
5786 		ciaa |= PCI_COMMAND;
5787 		IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5788 		ciad = IXGBE_READ_REG(hw, IXGBE_CIAD_82599);
5789 		ciaa &= 0x7FFFFFFF;
5790 		/* disable debug mode asap after reading data */
5791 		IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5792 		/* Get the upper 16 bits which will be the PCI status reg */
5793 		ciad >>= 16;
5794 		if (ciad & PCI_STATUS_REC_MASTER_ABORT) {
5795 			netdev_err(netdev, "VF %d Hung DMA\n", vf);
5796 			/* Issue VFLR */
5797 			ciaa = (vf << 16) | 0x80000000;
5798 			ciaa |= 0xA8;
5799 			IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5800 			ciad = 0x00008000;  /* VFLR */
5801 			IXGBE_WRITE_REG(hw, IXGBE_CIAD_82599, ciad);
5802 			ciaa &= 0x7FFFFFFF;
5803 			IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5804 		}
5805 	}
5806 }
5807 
5808 #endif
5809 /**
5810  * ixgbe_service_timer - Timer Call-back
5811  * @data: pointer to adapter cast into an unsigned long
5812  **/
5813 static void ixgbe_service_timer(unsigned long data)
5814 {
5815 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
5816 	unsigned long next_event_offset;
5817 	bool ready = true;
5818 
5819 	/* poll faster when waiting for link */
5820 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
5821 		next_event_offset = HZ / 10;
5822 	else
5823 		next_event_offset = HZ * 2;
5824 
5825 #ifdef CONFIG_PCI_IOV
5826 	/*
5827 	 * don't bother with SR-IOV VF DMA hang check if there are
5828 	 * no VFs or the link is down
5829 	 */
5830 	if (!adapter->num_vfs ||
5831 	    (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
5832 		goto normal_timer_service;
5833 
5834 	/* If we have VFs allocated then we must check for DMA hangs */
5835 	ixgbe_check_for_bad_vf(adapter);
5836 	next_event_offset = HZ / 50;
5837 	adapter->timer_event_accumulator++;
5838 
5839 	if (adapter->timer_event_accumulator >= 100)
5840 		adapter->timer_event_accumulator = 0;
5841 	else
5842 		ready = false;
5843 
5844 normal_timer_service:
5845 #endif
5846 	/* Reset the timer */
5847 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
5848 
5849 	if (ready)
5850 		ixgbe_service_event_schedule(adapter);
5851 }
5852 
5853 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
5854 {
5855 	if (!(adapter->flags2 & IXGBE_FLAG2_RESET_REQUESTED))
5856 		return;
5857 
5858 	adapter->flags2 &= ~IXGBE_FLAG2_RESET_REQUESTED;
5859 
5860 	/* If we're already down or resetting, just bail */
5861 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5862 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5863 		return;
5864 
5865 	ixgbe_dump(adapter);
5866 	netdev_err(adapter->netdev, "Reset adapter\n");
5867 	adapter->tx_timeout_count++;
5868 
5869 	ixgbe_reinit_locked(adapter);
5870 }
5871 
5872 /**
5873  * ixgbe_service_task - manages and runs subtasks
5874  * @work: pointer to work_struct containing our data
5875  **/
5876 static void ixgbe_service_task(struct work_struct *work)
5877 {
5878 	struct ixgbe_adapter *adapter = container_of(work,
5879 						     struct ixgbe_adapter,
5880 						     service_task);
5881 
5882 	ixgbe_reset_subtask(adapter);
5883 	ixgbe_sfp_detection_subtask(adapter);
5884 	ixgbe_sfp_link_config_subtask(adapter);
5885 	ixgbe_check_overtemp_subtask(adapter);
5886 	ixgbe_watchdog_subtask(adapter);
5887 	ixgbe_fdir_reinit_subtask(adapter);
5888 	ixgbe_check_hang_subtask(adapter);
5889 	ixgbe_ptp_overflow_check(adapter);
5890 
5891 	ixgbe_service_event_complete(adapter);
5892 }
5893 
5894 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
5895 		     struct ixgbe_tx_buffer *first,
5896 		     u8 *hdr_len)
5897 {
5898 	struct sk_buff *skb = first->skb;
5899 	u32 vlan_macip_lens, type_tucmd;
5900 	u32 mss_l4len_idx, l4len;
5901 
5902 	if (!skb_is_gso(skb))
5903 		return 0;
5904 
5905 	if (skb_header_cloned(skb)) {
5906 		int err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
5907 		if (err)
5908 			return err;
5909 	}
5910 
5911 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
5912 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
5913 
5914 	if (first->protocol == __constant_htons(ETH_P_IP)) {
5915 		struct iphdr *iph = ip_hdr(skb);
5916 		iph->tot_len = 0;
5917 		iph->check = 0;
5918 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
5919 							 iph->daddr, 0,
5920 							 IPPROTO_TCP,
5921 							 0);
5922 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
5923 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
5924 				   IXGBE_TX_FLAGS_CSUM |
5925 				   IXGBE_TX_FLAGS_IPV4;
5926 	} else if (skb_is_gso_v6(skb)) {
5927 		ipv6_hdr(skb)->payload_len = 0;
5928 		tcp_hdr(skb)->check =
5929 		    ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
5930 				     &ipv6_hdr(skb)->daddr,
5931 				     0, IPPROTO_TCP, 0);
5932 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
5933 				   IXGBE_TX_FLAGS_CSUM;
5934 	}
5935 
5936 	/* compute header lengths */
5937 	l4len = tcp_hdrlen(skb);
5938 	*hdr_len = skb_transport_offset(skb) + l4len;
5939 
5940 	/* update gso size and bytecount with header size */
5941 	first->gso_segs = skb_shinfo(skb)->gso_segs;
5942 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
5943 
5944 	/* mss_l4len_id: use 1 as index for TSO */
5945 	mss_l4len_idx = l4len << IXGBE_ADVTXD_L4LEN_SHIFT;
5946 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
5947 	mss_l4len_idx |= 1 << IXGBE_ADVTXD_IDX_SHIFT;
5948 
5949 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
5950 	vlan_macip_lens = skb_network_header_len(skb);
5951 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
5952 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
5953 
5954 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
5955 			  mss_l4len_idx);
5956 
5957 	return 1;
5958 }
5959 
5960 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
5961 			  struct ixgbe_tx_buffer *first)
5962 {
5963 	struct sk_buff *skb = first->skb;
5964 	u32 vlan_macip_lens = 0;
5965 	u32 mss_l4len_idx = 0;
5966 	u32 type_tucmd = 0;
5967 
5968 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
5969 		if (!(first->tx_flags & IXGBE_TX_FLAGS_HW_VLAN)) {
5970 			if (unlikely(skb->no_fcs))
5971 				first->tx_flags |= IXGBE_TX_FLAGS_NO_IFCS;
5972 			if (!(first->tx_flags & IXGBE_TX_FLAGS_TXSW))
5973 				return;
5974 		}
5975 	} else {
5976 		u8 l4_hdr = 0;
5977 		switch (first->protocol) {
5978 		case __constant_htons(ETH_P_IP):
5979 			vlan_macip_lens |= skb_network_header_len(skb);
5980 			type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
5981 			l4_hdr = ip_hdr(skb)->protocol;
5982 			break;
5983 		case __constant_htons(ETH_P_IPV6):
5984 			vlan_macip_lens |= skb_network_header_len(skb);
5985 			l4_hdr = ipv6_hdr(skb)->nexthdr;
5986 			break;
5987 		default:
5988 			if (unlikely(net_ratelimit())) {
5989 				dev_warn(tx_ring->dev,
5990 				 "partial checksum but proto=%x!\n",
5991 				 first->protocol);
5992 			}
5993 			break;
5994 		}
5995 
5996 		switch (l4_hdr) {
5997 		case IPPROTO_TCP:
5998 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_TCP;
5999 			mss_l4len_idx = tcp_hdrlen(skb) <<
6000 					IXGBE_ADVTXD_L4LEN_SHIFT;
6001 			break;
6002 		case IPPROTO_SCTP:
6003 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_SCTP;
6004 			mss_l4len_idx = sizeof(struct sctphdr) <<
6005 					IXGBE_ADVTXD_L4LEN_SHIFT;
6006 			break;
6007 		case IPPROTO_UDP:
6008 			mss_l4len_idx = sizeof(struct udphdr) <<
6009 					IXGBE_ADVTXD_L4LEN_SHIFT;
6010 			break;
6011 		default:
6012 			if (unlikely(net_ratelimit())) {
6013 				dev_warn(tx_ring->dev,
6014 				 "partial checksum but l4 proto=%x!\n",
6015 				 l4_hdr);
6016 			}
6017 			break;
6018 		}
6019 
6020 		/* update TX checksum flag */
6021 		first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
6022 	}
6023 
6024 	/* vlan_macip_lens: MACLEN, VLAN tag */
6025 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
6026 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
6027 
6028 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0,
6029 			  type_tucmd, mss_l4len_idx);
6030 }
6031 
6032 static __le32 ixgbe_tx_cmd_type(u32 tx_flags)
6033 {
6034 	/* set type for advanced descriptor with frame checksum insertion */
6035 	__le32 cmd_type = cpu_to_le32(IXGBE_ADVTXD_DTYP_DATA |
6036 				      IXGBE_ADVTXD_DCMD_DEXT);
6037 
6038 	/* set HW vlan bit if vlan is present */
6039 	if (tx_flags & IXGBE_TX_FLAGS_HW_VLAN)
6040 		cmd_type |= cpu_to_le32(IXGBE_ADVTXD_DCMD_VLE);
6041 
6042 	if (tx_flags & IXGBE_TX_FLAGS_TSTAMP)
6043 		cmd_type |= cpu_to_le32(IXGBE_ADVTXD_MAC_TSTAMP);
6044 
6045 	/* set segmentation enable bits for TSO/FSO */
6046 #ifdef IXGBE_FCOE
6047 	if (tx_flags & (IXGBE_TX_FLAGS_TSO | IXGBE_TX_FLAGS_FSO))
6048 #else
6049 	if (tx_flags & IXGBE_TX_FLAGS_TSO)
6050 #endif
6051 		cmd_type |= cpu_to_le32(IXGBE_ADVTXD_DCMD_TSE);
6052 
6053 	/* insert frame checksum */
6054 	if (!(tx_flags & IXGBE_TX_FLAGS_NO_IFCS))
6055 		cmd_type |= cpu_to_le32(IXGBE_ADVTXD_DCMD_IFCS);
6056 
6057 	return cmd_type;
6058 }
6059 
6060 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
6061 				   u32 tx_flags, unsigned int paylen)
6062 {
6063 	__le32 olinfo_status = cpu_to_le32(paylen << IXGBE_ADVTXD_PAYLEN_SHIFT);
6064 
6065 	/* enable L4 checksum for TSO and TX checksum offload */
6066 	if (tx_flags & IXGBE_TX_FLAGS_CSUM)
6067 		olinfo_status |= cpu_to_le32(IXGBE_ADVTXD_POPTS_TXSM);
6068 
6069 	/* enble IPv4 checksum for TSO */
6070 	if (tx_flags & IXGBE_TX_FLAGS_IPV4)
6071 		olinfo_status |= cpu_to_le32(IXGBE_ADVTXD_POPTS_IXSM);
6072 
6073 	/* use index 1 context for TSO/FSO/FCOE */
6074 #ifdef IXGBE_FCOE
6075 	if (tx_flags & (IXGBE_TX_FLAGS_TSO | IXGBE_TX_FLAGS_FCOE))
6076 #else
6077 	if (tx_flags & IXGBE_TX_FLAGS_TSO)
6078 #endif
6079 		olinfo_status |= cpu_to_le32(1 << IXGBE_ADVTXD_IDX_SHIFT);
6080 
6081 	/*
6082 	 * Check Context must be set if Tx switch is enabled, which it
6083 	 * always is for case where virtual functions are running
6084 	 */
6085 #ifdef IXGBE_FCOE
6086 	if (tx_flags & (IXGBE_TX_FLAGS_TXSW | IXGBE_TX_FLAGS_FCOE))
6087 #else
6088 	if (tx_flags & IXGBE_TX_FLAGS_TXSW)
6089 #endif
6090 		olinfo_status |= cpu_to_le32(IXGBE_ADVTXD_CC);
6091 
6092 	tx_desc->read.olinfo_status = olinfo_status;
6093 }
6094 
6095 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
6096 		       IXGBE_TXD_CMD_RS)
6097 
6098 static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
6099 			 struct ixgbe_tx_buffer *first,
6100 			 const u8 hdr_len)
6101 {
6102 	dma_addr_t dma;
6103 	struct sk_buff *skb = first->skb;
6104 	struct ixgbe_tx_buffer *tx_buffer;
6105 	union ixgbe_adv_tx_desc *tx_desc;
6106 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
6107 	unsigned int data_len = skb->data_len;
6108 	unsigned int size = skb_headlen(skb);
6109 	unsigned int paylen = skb->len - hdr_len;
6110 	u32 tx_flags = first->tx_flags;
6111 	__le32 cmd_type;
6112 	u16 i = tx_ring->next_to_use;
6113 
6114 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
6115 
6116 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, paylen);
6117 	cmd_type = ixgbe_tx_cmd_type(tx_flags);
6118 
6119 #ifdef IXGBE_FCOE
6120 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
6121 		if (data_len < sizeof(struct fcoe_crc_eof)) {
6122 			size -= sizeof(struct fcoe_crc_eof) - data_len;
6123 			data_len = 0;
6124 		} else {
6125 			data_len -= sizeof(struct fcoe_crc_eof);
6126 		}
6127 	}
6128 
6129 #endif
6130 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
6131 	if (dma_mapping_error(tx_ring->dev, dma))
6132 		goto dma_error;
6133 
6134 	/* record length, and DMA address */
6135 	dma_unmap_len_set(first, len, size);
6136 	dma_unmap_addr_set(first, dma, dma);
6137 
6138 	tx_desc->read.buffer_addr = cpu_to_le64(dma);
6139 
6140 	for (;;) {
6141 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
6142 			tx_desc->read.cmd_type_len =
6143 				cmd_type | cpu_to_le32(IXGBE_MAX_DATA_PER_TXD);
6144 
6145 			i++;
6146 			tx_desc++;
6147 			if (i == tx_ring->count) {
6148 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6149 				i = 0;
6150 			}
6151 
6152 			dma += IXGBE_MAX_DATA_PER_TXD;
6153 			size -= IXGBE_MAX_DATA_PER_TXD;
6154 
6155 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
6156 			tx_desc->read.olinfo_status = 0;
6157 		}
6158 
6159 		if (likely(!data_len))
6160 			break;
6161 
6162 		tx_desc->read.cmd_type_len = cmd_type | cpu_to_le32(size);
6163 
6164 		i++;
6165 		tx_desc++;
6166 		if (i == tx_ring->count) {
6167 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6168 			i = 0;
6169 		}
6170 
6171 #ifdef IXGBE_FCOE
6172 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
6173 #else
6174 		size = skb_frag_size(frag);
6175 #endif
6176 		data_len -= size;
6177 
6178 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
6179 				       DMA_TO_DEVICE);
6180 		if (dma_mapping_error(tx_ring->dev, dma))
6181 			goto dma_error;
6182 
6183 		tx_buffer = &tx_ring->tx_buffer_info[i];
6184 		dma_unmap_len_set(tx_buffer, len, size);
6185 		dma_unmap_addr_set(tx_buffer, dma, dma);
6186 
6187 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
6188 		tx_desc->read.olinfo_status = 0;
6189 
6190 		frag++;
6191 	}
6192 
6193 	/* write last descriptor with RS and EOP bits */
6194 	cmd_type |= cpu_to_le32(size) | cpu_to_le32(IXGBE_TXD_CMD);
6195 	tx_desc->read.cmd_type_len = cmd_type;
6196 
6197 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
6198 
6199 	/* set the timestamp */
6200 	first->time_stamp = jiffies;
6201 
6202 	/*
6203 	 * Force memory writes to complete before letting h/w know there
6204 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
6205 	 * memory model archs, such as IA-64).
6206 	 *
6207 	 * We also need this memory barrier to make certain all of the
6208 	 * status bits have been updated before next_to_watch is written.
6209 	 */
6210 	wmb();
6211 
6212 	/* set next_to_watch value indicating a packet is present */
6213 	first->next_to_watch = tx_desc;
6214 
6215 	i++;
6216 	if (i == tx_ring->count)
6217 		i = 0;
6218 
6219 	tx_ring->next_to_use = i;
6220 
6221 	/* notify HW of packet */
6222 	writel(i, tx_ring->tail);
6223 
6224 	return;
6225 dma_error:
6226 	dev_err(tx_ring->dev, "TX DMA map failed\n");
6227 
6228 	/* clear dma mappings for failed tx_buffer_info map */
6229 	for (;;) {
6230 		tx_buffer = &tx_ring->tx_buffer_info[i];
6231 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer);
6232 		if (tx_buffer == first)
6233 			break;
6234 		if (i == 0)
6235 			i = tx_ring->count;
6236 		i--;
6237 	}
6238 
6239 	tx_ring->next_to_use = i;
6240 }
6241 
6242 static void ixgbe_atr(struct ixgbe_ring *ring,
6243 		      struct ixgbe_tx_buffer *first)
6244 {
6245 	struct ixgbe_q_vector *q_vector = ring->q_vector;
6246 	union ixgbe_atr_hash_dword input = { .dword = 0 };
6247 	union ixgbe_atr_hash_dword common = { .dword = 0 };
6248 	union {
6249 		unsigned char *network;
6250 		struct iphdr *ipv4;
6251 		struct ipv6hdr *ipv6;
6252 	} hdr;
6253 	struct tcphdr *th;
6254 	__be16 vlan_id;
6255 
6256 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
6257 	if (!q_vector)
6258 		return;
6259 
6260 	/* do nothing if sampling is disabled */
6261 	if (!ring->atr_sample_rate)
6262 		return;
6263 
6264 	ring->atr_count++;
6265 
6266 	/* snag network header to get L4 type and address */
6267 	hdr.network = skb_network_header(first->skb);
6268 
6269 	/* Currently only IPv4/IPv6 with TCP is supported */
6270 	if ((first->protocol != __constant_htons(ETH_P_IPV6) ||
6271 	     hdr.ipv6->nexthdr != IPPROTO_TCP) &&
6272 	    (first->protocol != __constant_htons(ETH_P_IP) ||
6273 	     hdr.ipv4->protocol != IPPROTO_TCP))
6274 		return;
6275 
6276 	th = tcp_hdr(first->skb);
6277 
6278 	/* skip this packet since it is invalid or the socket is closing */
6279 	if (!th || th->fin)
6280 		return;
6281 
6282 	/* sample on all syn packets or once every atr sample count */
6283 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
6284 		return;
6285 
6286 	/* reset sample count */
6287 	ring->atr_count = 0;
6288 
6289 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
6290 
6291 	/*
6292 	 * src and dst are inverted, think how the receiver sees them
6293 	 *
6294 	 * The input is broken into two sections, a non-compressed section
6295 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
6296 	 * is XORed together and stored in the compressed dword.
6297 	 */
6298 	input.formatted.vlan_id = vlan_id;
6299 
6300 	/*
6301 	 * since src port and flex bytes occupy the same word XOR them together
6302 	 * and write the value to source port portion of compressed dword
6303 	 */
6304 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
6305 		common.port.src ^= th->dest ^ __constant_htons(ETH_P_8021Q);
6306 	else
6307 		common.port.src ^= th->dest ^ first->protocol;
6308 	common.port.dst ^= th->source;
6309 
6310 	if (first->protocol == __constant_htons(ETH_P_IP)) {
6311 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
6312 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
6313 	} else {
6314 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
6315 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
6316 			     hdr.ipv6->saddr.s6_addr32[1] ^
6317 			     hdr.ipv6->saddr.s6_addr32[2] ^
6318 			     hdr.ipv6->saddr.s6_addr32[3] ^
6319 			     hdr.ipv6->daddr.s6_addr32[0] ^
6320 			     hdr.ipv6->daddr.s6_addr32[1] ^
6321 			     hdr.ipv6->daddr.s6_addr32[2] ^
6322 			     hdr.ipv6->daddr.s6_addr32[3];
6323 	}
6324 
6325 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
6326 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
6327 					      input, common, ring->queue_index);
6328 }
6329 
6330 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
6331 {
6332 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
6333 	/* Herbert's original patch had:
6334 	 *  smp_mb__after_netif_stop_queue();
6335 	 * but since that doesn't exist yet, just open code it. */
6336 	smp_mb();
6337 
6338 	/* We need to check again in a case another CPU has just
6339 	 * made room available. */
6340 	if (likely(ixgbe_desc_unused(tx_ring) < size))
6341 		return -EBUSY;
6342 
6343 	/* A reprieve! - use start_queue because it doesn't call schedule */
6344 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
6345 	++tx_ring->tx_stats.restart_queue;
6346 	return 0;
6347 }
6348 
6349 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
6350 {
6351 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
6352 		return 0;
6353 	return __ixgbe_maybe_stop_tx(tx_ring, size);
6354 }
6355 
6356 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb)
6357 {
6358 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6359 	int txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
6360 					       smp_processor_id();
6361 #ifdef IXGBE_FCOE
6362 	__be16 protocol = vlan_get_protocol(skb);
6363 
6364 	if (((protocol == htons(ETH_P_FCOE)) ||
6365 	    (protocol == htons(ETH_P_FIP))) &&
6366 	    (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)) {
6367 		struct ixgbe_ring_feature *f;
6368 
6369 		f = &adapter->ring_feature[RING_F_FCOE];
6370 
6371 		while (txq >= f->indices)
6372 			txq -= f->indices;
6373 		txq += adapter->ring_feature[RING_F_FCOE].offset;
6374 
6375 		return txq;
6376 	}
6377 #endif
6378 
6379 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
6380 		while (unlikely(txq >= dev->real_num_tx_queues))
6381 			txq -= dev->real_num_tx_queues;
6382 		return txq;
6383 	}
6384 
6385 	return skb_tx_hash(dev, skb);
6386 }
6387 
6388 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
6389 			  struct ixgbe_adapter *adapter,
6390 			  struct ixgbe_ring *tx_ring)
6391 {
6392 	struct ixgbe_tx_buffer *first;
6393 	int tso;
6394 	u32 tx_flags = 0;
6395 #if PAGE_SIZE > IXGBE_MAX_DATA_PER_TXD
6396 	unsigned short f;
6397 #endif
6398 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
6399 	__be16 protocol = skb->protocol;
6400 	u8 hdr_len = 0;
6401 
6402 	/*
6403 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
6404 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
6405 	 *       + 2 desc gap to keep tail from touching head,
6406 	 *       + 1 desc for context descriptor,
6407 	 * otherwise try next time
6408 	 */
6409 #if PAGE_SIZE > IXGBE_MAX_DATA_PER_TXD
6410 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
6411 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
6412 #else
6413 	count += skb_shinfo(skb)->nr_frags;
6414 #endif
6415 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
6416 		tx_ring->tx_stats.tx_busy++;
6417 		return NETDEV_TX_BUSY;
6418 	}
6419 
6420 	/* record the location of the first descriptor for this packet */
6421 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
6422 	first->skb = skb;
6423 	first->bytecount = skb->len;
6424 	first->gso_segs = 1;
6425 
6426 	/* if we have a HW VLAN tag being added default to the HW one */
6427 	if (vlan_tx_tag_present(skb)) {
6428 		tx_flags |= vlan_tx_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
6429 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
6430 	/* else if it is a SW VLAN check the next protocol and store the tag */
6431 	} else if (protocol == __constant_htons(ETH_P_8021Q)) {
6432 		struct vlan_hdr *vhdr, _vhdr;
6433 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
6434 		if (!vhdr)
6435 			goto out_drop;
6436 
6437 		protocol = vhdr->h_vlan_encapsulated_proto;
6438 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
6439 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
6440 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
6441 	}
6442 
6443 	skb_tx_timestamp(skb);
6444 
6445 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) {
6446 		skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
6447 		tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
6448 	}
6449 
6450 #ifdef CONFIG_PCI_IOV
6451 	/*
6452 	 * Use the l2switch_enable flag - would be false if the DMA
6453 	 * Tx switch had been disabled.
6454 	 */
6455 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
6456 		tx_flags |= IXGBE_TX_FLAGS_TXSW;
6457 
6458 #endif
6459 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
6460 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
6461 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
6462 	     (skb->priority != TC_PRIO_CONTROL))) {
6463 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
6464 		tx_flags |= (skb->priority & 0x7) <<
6465 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
6466 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
6467 			struct vlan_ethhdr *vhdr;
6468 			if (skb_header_cloned(skb) &&
6469 			    pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
6470 				goto out_drop;
6471 			vhdr = (struct vlan_ethhdr *)skb->data;
6472 			vhdr->h_vlan_TCI = htons(tx_flags >>
6473 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
6474 		} else {
6475 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
6476 		}
6477 	}
6478 
6479 	/* record initial flags and protocol */
6480 	first->tx_flags = tx_flags;
6481 	first->protocol = protocol;
6482 
6483 #ifdef IXGBE_FCOE
6484 	/* setup tx offload for FCoE */
6485 	if ((protocol == __constant_htons(ETH_P_FCOE)) &&
6486 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
6487 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
6488 		if (tso < 0)
6489 			goto out_drop;
6490 
6491 		goto xmit_fcoe;
6492 	}
6493 
6494 #endif /* IXGBE_FCOE */
6495 	tso = ixgbe_tso(tx_ring, first, &hdr_len);
6496 	if (tso < 0)
6497 		goto out_drop;
6498 	else if (!tso)
6499 		ixgbe_tx_csum(tx_ring, first);
6500 
6501 	/* add the ATR filter if ATR is on */
6502 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
6503 		ixgbe_atr(tx_ring, first);
6504 
6505 #ifdef IXGBE_FCOE
6506 xmit_fcoe:
6507 #endif /* IXGBE_FCOE */
6508 	ixgbe_tx_map(tx_ring, first, hdr_len);
6509 
6510 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
6511 
6512 	return NETDEV_TX_OK;
6513 
6514 out_drop:
6515 	dev_kfree_skb_any(first->skb);
6516 	first->skb = NULL;
6517 
6518 	return NETDEV_TX_OK;
6519 }
6520 
6521 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
6522 				    struct net_device *netdev)
6523 {
6524 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6525 	struct ixgbe_ring *tx_ring;
6526 
6527 	/*
6528 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
6529 	 * in order to meet this minimum size requirement.
6530 	 */
6531 	if (unlikely(skb->len < 17)) {
6532 		if (skb_pad(skb, 17 - skb->len))
6533 			return NETDEV_TX_OK;
6534 		skb->len = 17;
6535 		skb_set_tail_pointer(skb, 17);
6536 	}
6537 
6538 	tx_ring = adapter->tx_ring[skb->queue_mapping];
6539 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
6540 }
6541 
6542 /**
6543  * ixgbe_set_mac - Change the Ethernet Address of the NIC
6544  * @netdev: network interface device structure
6545  * @p: pointer to an address structure
6546  *
6547  * Returns 0 on success, negative on failure
6548  **/
6549 static int ixgbe_set_mac(struct net_device *netdev, void *p)
6550 {
6551 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6552 	struct ixgbe_hw *hw = &adapter->hw;
6553 	struct sockaddr *addr = p;
6554 
6555 	if (!is_valid_ether_addr(addr->sa_data))
6556 		return -EADDRNOTAVAIL;
6557 
6558 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
6559 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
6560 
6561 	hw->mac.ops.set_rar(hw, 0, hw->mac.addr, VMDQ_P(0), IXGBE_RAH_AV);
6562 
6563 	return 0;
6564 }
6565 
6566 static int
6567 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
6568 {
6569 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6570 	struct ixgbe_hw *hw = &adapter->hw;
6571 	u16 value;
6572 	int rc;
6573 
6574 	if (prtad != hw->phy.mdio.prtad)
6575 		return -EINVAL;
6576 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
6577 	if (!rc)
6578 		rc = value;
6579 	return rc;
6580 }
6581 
6582 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
6583 			    u16 addr, u16 value)
6584 {
6585 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6586 	struct ixgbe_hw *hw = &adapter->hw;
6587 
6588 	if (prtad != hw->phy.mdio.prtad)
6589 		return -EINVAL;
6590 	return hw->phy.ops.write_reg(hw, addr, devad, value);
6591 }
6592 
6593 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
6594 {
6595 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6596 
6597 	switch (cmd) {
6598 	case SIOCSHWTSTAMP:
6599 		return ixgbe_ptp_hwtstamp_ioctl(adapter, req, cmd);
6600 	default:
6601 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
6602 	}
6603 }
6604 
6605 /**
6606  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
6607  * netdev->dev_addrs
6608  * @netdev: network interface device structure
6609  *
6610  * Returns non-zero on failure
6611  **/
6612 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
6613 {
6614 	int err = 0;
6615 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6616 	struct ixgbe_hw *hw = &adapter->hw;
6617 
6618 	if (is_valid_ether_addr(hw->mac.san_addr)) {
6619 		rtnl_lock();
6620 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
6621 		rtnl_unlock();
6622 
6623 		/* update SAN MAC vmdq pool selection */
6624 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
6625 	}
6626 	return err;
6627 }
6628 
6629 /**
6630  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
6631  * netdev->dev_addrs
6632  * @netdev: network interface device structure
6633  *
6634  * Returns non-zero on failure
6635  **/
6636 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
6637 {
6638 	int err = 0;
6639 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6640 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
6641 
6642 	if (is_valid_ether_addr(mac->san_addr)) {
6643 		rtnl_lock();
6644 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
6645 		rtnl_unlock();
6646 	}
6647 	return err;
6648 }
6649 
6650 #ifdef CONFIG_NET_POLL_CONTROLLER
6651 /*
6652  * Polling 'interrupt' - used by things like netconsole to send skbs
6653  * without having to re-enable interrupts. It's not called while
6654  * the interrupt routine is executing.
6655  */
6656 static void ixgbe_netpoll(struct net_device *netdev)
6657 {
6658 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6659 	int i;
6660 
6661 	/* if interface is down do nothing */
6662 	if (test_bit(__IXGBE_DOWN, &adapter->state))
6663 		return;
6664 
6665 	adapter->flags |= IXGBE_FLAG_IN_NETPOLL;
6666 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
6667 		for (i = 0; i < adapter->num_q_vectors; i++)
6668 			ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
6669 	} else {
6670 		ixgbe_intr(adapter->pdev->irq, netdev);
6671 	}
6672 	adapter->flags &= ~IXGBE_FLAG_IN_NETPOLL;
6673 }
6674 
6675 #endif
6676 static struct rtnl_link_stats64 *ixgbe_get_stats64(struct net_device *netdev,
6677 						   struct rtnl_link_stats64 *stats)
6678 {
6679 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6680 	int i;
6681 
6682 	rcu_read_lock();
6683 	for (i = 0; i < adapter->num_rx_queues; i++) {
6684 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
6685 		u64 bytes, packets;
6686 		unsigned int start;
6687 
6688 		if (ring) {
6689 			do {
6690 				start = u64_stats_fetch_begin_bh(&ring->syncp);
6691 				packets = ring->stats.packets;
6692 				bytes   = ring->stats.bytes;
6693 			} while (u64_stats_fetch_retry_bh(&ring->syncp, start));
6694 			stats->rx_packets += packets;
6695 			stats->rx_bytes   += bytes;
6696 		}
6697 	}
6698 
6699 	for (i = 0; i < adapter->num_tx_queues; i++) {
6700 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
6701 		u64 bytes, packets;
6702 		unsigned int start;
6703 
6704 		if (ring) {
6705 			do {
6706 				start = u64_stats_fetch_begin_bh(&ring->syncp);
6707 				packets = ring->stats.packets;
6708 				bytes   = ring->stats.bytes;
6709 			} while (u64_stats_fetch_retry_bh(&ring->syncp, start));
6710 			stats->tx_packets += packets;
6711 			stats->tx_bytes   += bytes;
6712 		}
6713 	}
6714 	rcu_read_unlock();
6715 	/* following stats updated by ixgbe_watchdog_task() */
6716 	stats->multicast	= netdev->stats.multicast;
6717 	stats->rx_errors	= netdev->stats.rx_errors;
6718 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
6719 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
6720 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
6721 	return stats;
6722 }
6723 
6724 #ifdef CONFIG_IXGBE_DCB
6725 /**
6726  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
6727  * @adapter: pointer to ixgbe_adapter
6728  * @tc: number of traffic classes currently enabled
6729  *
6730  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
6731  * 802.1Q priority maps to a packet buffer that exists.
6732  */
6733 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
6734 {
6735 	struct ixgbe_hw *hw = &adapter->hw;
6736 	u32 reg, rsave;
6737 	int i;
6738 
6739 	/* 82598 have a static priority to TC mapping that can not
6740 	 * be changed so no validation is needed.
6741 	 */
6742 	if (hw->mac.type == ixgbe_mac_82598EB)
6743 		return;
6744 
6745 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
6746 	rsave = reg;
6747 
6748 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
6749 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
6750 
6751 		/* If up2tc is out of bounds default to zero */
6752 		if (up2tc > tc)
6753 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
6754 	}
6755 
6756 	if (reg != rsave)
6757 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
6758 
6759 	return;
6760 }
6761 
6762 /**
6763  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
6764  * @adapter: Pointer to adapter struct
6765  *
6766  * Populate the netdev user priority to tc map
6767  */
6768 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
6769 {
6770 	struct net_device *dev = adapter->netdev;
6771 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
6772 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
6773 	u8 prio;
6774 
6775 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
6776 		u8 tc = 0;
6777 
6778 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
6779 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
6780 		else if (ets)
6781 			tc = ets->prio_tc[prio];
6782 
6783 		netdev_set_prio_tc_map(dev, prio, tc);
6784 	}
6785 }
6786 
6787 /**
6788  * ixgbe_setup_tc - configure net_device for multiple traffic classes
6789  *
6790  * @netdev: net device to configure
6791  * @tc: number of traffic classes to enable
6792  */
6793 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
6794 {
6795 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6796 	struct ixgbe_hw *hw = &adapter->hw;
6797 
6798 	/* Hardware supports up to 8 traffic classes */
6799 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs ||
6800 	    (hw->mac.type == ixgbe_mac_82598EB &&
6801 	     tc < MAX_TRAFFIC_CLASS))
6802 		return -EINVAL;
6803 
6804 	/* Hardware has to reinitialize queues and interrupts to
6805 	 * match packet buffer alignment. Unfortunately, the
6806 	 * hardware is not flexible enough to do this dynamically.
6807 	 */
6808 	if (netif_running(dev))
6809 		ixgbe_close(dev);
6810 	ixgbe_clear_interrupt_scheme(adapter);
6811 
6812 	if (tc) {
6813 		netdev_set_num_tc(dev, tc);
6814 		ixgbe_set_prio_tc_map(adapter);
6815 
6816 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
6817 
6818 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
6819 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
6820 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
6821 		}
6822 	} else {
6823 		netdev_reset_tc(dev);
6824 
6825 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
6826 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
6827 
6828 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
6829 
6830 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
6831 		adapter->dcb_cfg.pfc_mode_enable = false;
6832 	}
6833 
6834 	ixgbe_init_interrupt_scheme(adapter);
6835 	ixgbe_validate_rtr(adapter, tc);
6836 	if (netif_running(dev))
6837 		ixgbe_open(dev);
6838 
6839 	return 0;
6840 }
6841 
6842 #endif /* CONFIG_IXGBE_DCB */
6843 void ixgbe_do_reset(struct net_device *netdev)
6844 {
6845 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6846 
6847 	if (netif_running(netdev))
6848 		ixgbe_reinit_locked(adapter);
6849 	else
6850 		ixgbe_reset(adapter);
6851 }
6852 
6853 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
6854 					    netdev_features_t features)
6855 {
6856 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6857 
6858 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
6859 	if (!(features & NETIF_F_RXCSUM))
6860 		features &= ~NETIF_F_LRO;
6861 
6862 	/* Turn off LRO if not RSC capable */
6863 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
6864 		features &= ~NETIF_F_LRO;
6865 
6866 	return features;
6867 }
6868 
6869 static int ixgbe_set_features(struct net_device *netdev,
6870 			      netdev_features_t features)
6871 {
6872 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6873 	netdev_features_t changed = netdev->features ^ features;
6874 	bool need_reset = false;
6875 
6876 	/* Make sure RSC matches LRO, reset if change */
6877 	if (!(features & NETIF_F_LRO)) {
6878 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
6879 			need_reset = true;
6880 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
6881 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
6882 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
6883 		if (adapter->rx_itr_setting == 1 ||
6884 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
6885 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
6886 			need_reset = true;
6887 		} else if ((changed ^ features) & NETIF_F_LRO) {
6888 			e_info(probe, "rx-usecs set too low, "
6889 			       "disabling RSC\n");
6890 		}
6891 	}
6892 
6893 	/*
6894 	 * Check if Flow Director n-tuple support was enabled or disabled.  If
6895 	 * the state changed, we need to reset.
6896 	 */
6897 	switch (features & NETIF_F_NTUPLE) {
6898 	case NETIF_F_NTUPLE:
6899 		/* turn off ATR, enable perfect filters and reset */
6900 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
6901 			need_reset = true;
6902 
6903 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
6904 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
6905 		break;
6906 	default:
6907 		/* turn off perfect filters, enable ATR and reset */
6908 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
6909 			need_reset = true;
6910 
6911 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
6912 
6913 		/* We cannot enable ATR if SR-IOV is enabled */
6914 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
6915 			break;
6916 
6917 		/* We cannot enable ATR if we have 2 or more traffic classes */
6918 		if (netdev_get_num_tc(netdev) > 1)
6919 			break;
6920 
6921 		/* We cannot enable ATR if RSS is disabled */
6922 		if (adapter->ring_feature[RING_F_RSS].limit <= 1)
6923 			break;
6924 
6925 		/* A sample rate of 0 indicates ATR disabled */
6926 		if (!adapter->atr_sample_rate)
6927 			break;
6928 
6929 		adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
6930 		break;
6931 	}
6932 
6933 	if (features & NETIF_F_HW_VLAN_RX)
6934 		ixgbe_vlan_strip_enable(adapter);
6935 	else
6936 		ixgbe_vlan_strip_disable(adapter);
6937 
6938 	if (changed & NETIF_F_RXALL)
6939 		need_reset = true;
6940 
6941 	netdev->features = features;
6942 	if (need_reset)
6943 		ixgbe_do_reset(netdev);
6944 
6945 	return 0;
6946 }
6947 
6948 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
6949 			     struct net_device *dev,
6950 			     const unsigned char *addr,
6951 			     u16 flags)
6952 {
6953 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6954 	int err;
6955 
6956 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
6957 		return -EOPNOTSUPP;
6958 
6959 	/* Hardware does not support aging addresses so if a
6960 	 * ndm_state is given only allow permanent addresses
6961 	 */
6962 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
6963 		pr_info("%s: FDB only supports static addresses\n",
6964 			ixgbe_driver_name);
6965 		return -EINVAL;
6966 	}
6967 
6968 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
6969 		u32 rar_uc_entries = IXGBE_MAX_PF_MACVLANS;
6970 
6971 		if (netdev_uc_count(dev) < rar_uc_entries)
6972 			err = dev_uc_add_excl(dev, addr);
6973 		else
6974 			err = -ENOMEM;
6975 	} else if (is_multicast_ether_addr(addr)) {
6976 		err = dev_mc_add_excl(dev, addr);
6977 	} else {
6978 		err = -EINVAL;
6979 	}
6980 
6981 	/* Only return duplicate errors if NLM_F_EXCL is set */
6982 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
6983 		err = 0;
6984 
6985 	return err;
6986 }
6987 
6988 static int ixgbe_ndo_fdb_del(struct ndmsg *ndm,
6989 			     struct net_device *dev,
6990 			     const unsigned char *addr)
6991 {
6992 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6993 	int err = -EOPNOTSUPP;
6994 
6995 	if (ndm->ndm_state & NUD_PERMANENT) {
6996 		pr_info("%s: FDB only supports static addresses\n",
6997 			ixgbe_driver_name);
6998 		return -EINVAL;
6999 	}
7000 
7001 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
7002 		if (is_unicast_ether_addr(addr))
7003 			err = dev_uc_del(dev, addr);
7004 		else if (is_multicast_ether_addr(addr))
7005 			err = dev_mc_del(dev, addr);
7006 		else
7007 			err = -EINVAL;
7008 	}
7009 
7010 	return err;
7011 }
7012 
7013 static int ixgbe_ndo_fdb_dump(struct sk_buff *skb,
7014 			      struct netlink_callback *cb,
7015 			      struct net_device *dev,
7016 			      int idx)
7017 {
7018 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7019 
7020 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7021 		idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
7022 
7023 	return idx;
7024 }
7025 
7026 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
7027 				    struct nlmsghdr *nlh)
7028 {
7029 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7030 	struct nlattr *attr, *br_spec;
7031 	int rem;
7032 
7033 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
7034 		return -EOPNOTSUPP;
7035 
7036 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
7037 
7038 	nla_for_each_nested(attr, br_spec, rem) {
7039 		__u16 mode;
7040 		u32 reg = 0;
7041 
7042 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
7043 			continue;
7044 
7045 		mode = nla_get_u16(attr);
7046 		if (mode == BRIDGE_MODE_VEPA) {
7047 			reg = 0;
7048 			adapter->flags2 &= ~IXGBE_FLAG2_BRIDGE_MODE_VEB;
7049 		} else if (mode == BRIDGE_MODE_VEB) {
7050 			reg = IXGBE_PFDTXGSWC_VT_LBEN;
7051 			adapter->flags2 |= IXGBE_FLAG2_BRIDGE_MODE_VEB;
7052 		} else
7053 			return -EINVAL;
7054 
7055 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, reg);
7056 
7057 		e_info(drv, "enabling bridge mode: %s\n",
7058 			mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
7059 	}
7060 
7061 	return 0;
7062 }
7063 
7064 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
7065 				    struct net_device *dev)
7066 {
7067 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7068 	u16 mode;
7069 
7070 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
7071 		return 0;
7072 
7073 	if (adapter->flags2 & IXGBE_FLAG2_BRIDGE_MODE_VEB)
7074 		mode = BRIDGE_MODE_VEB;
7075 	else
7076 		mode = BRIDGE_MODE_VEPA;
7077 
7078 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev, mode);
7079 }
7080 
7081 static const struct net_device_ops ixgbe_netdev_ops = {
7082 	.ndo_open		= ixgbe_open,
7083 	.ndo_stop		= ixgbe_close,
7084 	.ndo_start_xmit		= ixgbe_xmit_frame,
7085 	.ndo_select_queue	= ixgbe_select_queue,
7086 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
7087 	.ndo_validate_addr	= eth_validate_addr,
7088 	.ndo_set_mac_address	= ixgbe_set_mac,
7089 	.ndo_change_mtu		= ixgbe_change_mtu,
7090 	.ndo_tx_timeout		= ixgbe_tx_timeout,
7091 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
7092 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
7093 	.ndo_do_ioctl		= ixgbe_ioctl,
7094 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
7095 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
7096 	.ndo_set_vf_tx_rate	= ixgbe_ndo_set_vf_bw,
7097 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
7098 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
7099 	.ndo_get_stats64	= ixgbe_get_stats64,
7100 #ifdef CONFIG_IXGBE_DCB
7101 	.ndo_setup_tc		= ixgbe_setup_tc,
7102 #endif
7103 #ifdef CONFIG_NET_POLL_CONTROLLER
7104 	.ndo_poll_controller	= ixgbe_netpoll,
7105 #endif
7106 #ifdef IXGBE_FCOE
7107 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
7108 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
7109 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
7110 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
7111 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
7112 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
7113 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
7114 #endif /* IXGBE_FCOE */
7115 	.ndo_set_features = ixgbe_set_features,
7116 	.ndo_fix_features = ixgbe_fix_features,
7117 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
7118 	.ndo_fdb_del		= ixgbe_ndo_fdb_del,
7119 	.ndo_fdb_dump		= ixgbe_ndo_fdb_dump,
7120 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
7121 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
7122 };
7123 
7124 /**
7125  * ixgbe_wol_supported - Check whether device supports WoL
7126  * @hw: hw specific details
7127  * @device_id: the device ID
7128  * @subdev_id: the subsystem device ID
7129  *
7130  * This function is used by probe and ethtool to determine
7131  * which devices have WoL support
7132  *
7133  **/
7134 int ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
7135 			u16 subdevice_id)
7136 {
7137 	struct ixgbe_hw *hw = &adapter->hw;
7138 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
7139 	int is_wol_supported = 0;
7140 
7141 	switch (device_id) {
7142 	case IXGBE_DEV_ID_82599_SFP:
7143 		/* Only these subdevices could supports WOL */
7144 		switch (subdevice_id) {
7145 		case IXGBE_SUBDEV_ID_82599_560FLR:
7146 			/* only support first port */
7147 			if (hw->bus.func != 0)
7148 				break;
7149 		case IXGBE_SUBDEV_ID_82599_SFP:
7150 		case IXGBE_SUBDEV_ID_82599_RNDC:
7151 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
7152 			is_wol_supported = 1;
7153 			break;
7154 		}
7155 		break;
7156 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
7157 		/* All except this subdevice support WOL */
7158 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
7159 			is_wol_supported = 1;
7160 		break;
7161 	case IXGBE_DEV_ID_82599_KX4:
7162 		is_wol_supported = 1;
7163 		break;
7164 	case IXGBE_DEV_ID_X540T:
7165 	case IXGBE_DEV_ID_X540T1:
7166 		/* check eeprom to see if enabled wol */
7167 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
7168 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
7169 		     (hw->bus.func == 0))) {
7170 			is_wol_supported = 1;
7171 		}
7172 		break;
7173 	}
7174 
7175 	return is_wol_supported;
7176 }
7177 
7178 /**
7179  * ixgbe_probe - Device Initialization Routine
7180  * @pdev: PCI device information struct
7181  * @ent: entry in ixgbe_pci_tbl
7182  *
7183  * Returns 0 on success, negative on failure
7184  *
7185  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
7186  * The OS initialization, configuring of the adapter private structure,
7187  * and a hardware reset occur.
7188  **/
7189 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7190 {
7191 	struct net_device *netdev;
7192 	struct ixgbe_adapter *adapter = NULL;
7193 	struct ixgbe_hw *hw;
7194 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
7195 	static int cards_found;
7196 	int i, err, pci_using_dac;
7197 	u8 part_str[IXGBE_PBANUM_LENGTH];
7198 	unsigned int indices = num_possible_cpus();
7199 	unsigned int dcb_max = 0;
7200 #ifdef IXGBE_FCOE
7201 	u16 device_caps;
7202 #endif
7203 	u32 eec;
7204 
7205 	/* Catch broken hardware that put the wrong VF device ID in
7206 	 * the PCIe SR-IOV capability.
7207 	 */
7208 	if (pdev->is_virtfn) {
7209 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
7210 		     pci_name(pdev), pdev->vendor, pdev->device);
7211 		return -EINVAL;
7212 	}
7213 
7214 	err = pci_enable_device_mem(pdev);
7215 	if (err)
7216 		return err;
7217 
7218 	if (!dma_set_mask(&pdev->dev, DMA_BIT_MASK(64)) &&
7219 	    !dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(64))) {
7220 		pci_using_dac = 1;
7221 	} else {
7222 		err = dma_set_mask(&pdev->dev, DMA_BIT_MASK(32));
7223 		if (err) {
7224 			err = dma_set_coherent_mask(&pdev->dev,
7225 						    DMA_BIT_MASK(32));
7226 			if (err) {
7227 				dev_err(&pdev->dev,
7228 					"No usable DMA configuration, aborting\n");
7229 				goto err_dma;
7230 			}
7231 		}
7232 		pci_using_dac = 0;
7233 	}
7234 
7235 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
7236 					   IORESOURCE_MEM), ixgbe_driver_name);
7237 	if (err) {
7238 		dev_err(&pdev->dev,
7239 			"pci_request_selected_regions failed 0x%x\n", err);
7240 		goto err_pci_reg;
7241 	}
7242 
7243 	pci_enable_pcie_error_reporting(pdev);
7244 
7245 	pci_set_master(pdev);
7246 	pci_save_state(pdev);
7247 
7248 #ifdef CONFIG_IXGBE_DCB
7249 	if (ii->mac == ixgbe_mac_82598EB)
7250 		dcb_max = min_t(unsigned int, indices * MAX_TRAFFIC_CLASS,
7251 				IXGBE_MAX_RSS_INDICES);
7252 	else
7253 		dcb_max = min_t(unsigned int, indices * MAX_TRAFFIC_CLASS,
7254 				IXGBE_MAX_FDIR_INDICES);
7255 #endif
7256 
7257 	if (ii->mac == ixgbe_mac_82598EB)
7258 		indices = min_t(unsigned int, indices, IXGBE_MAX_RSS_INDICES);
7259 	else
7260 		indices = min_t(unsigned int, indices, IXGBE_MAX_FDIR_INDICES);
7261 
7262 #ifdef IXGBE_FCOE
7263 	indices += min_t(unsigned int, num_possible_cpus(),
7264 			 IXGBE_MAX_FCOE_INDICES);
7265 #endif
7266 	indices = max_t(unsigned int, dcb_max, indices);
7267 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
7268 	if (!netdev) {
7269 		err = -ENOMEM;
7270 		goto err_alloc_etherdev;
7271 	}
7272 
7273 	SET_NETDEV_DEV(netdev, &pdev->dev);
7274 
7275 	adapter = netdev_priv(netdev);
7276 	pci_set_drvdata(pdev, adapter);
7277 
7278 	adapter->netdev = netdev;
7279 	adapter->pdev = pdev;
7280 	hw = &adapter->hw;
7281 	hw->back = adapter;
7282 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
7283 
7284 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
7285 			      pci_resource_len(pdev, 0));
7286 	if (!hw->hw_addr) {
7287 		err = -EIO;
7288 		goto err_ioremap;
7289 	}
7290 
7291 	netdev->netdev_ops = &ixgbe_netdev_ops;
7292 	ixgbe_set_ethtool_ops(netdev);
7293 	netdev->watchdog_timeo = 5 * HZ;
7294 	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
7295 
7296 	adapter->bd_number = cards_found;
7297 
7298 	/* Setup hw api */
7299 	memcpy(&hw->mac.ops, ii->mac_ops, sizeof(hw->mac.ops));
7300 	hw->mac.type  = ii->mac;
7301 
7302 	/* EEPROM */
7303 	memcpy(&hw->eeprom.ops, ii->eeprom_ops, sizeof(hw->eeprom.ops));
7304 	eec = IXGBE_READ_REG(hw, IXGBE_EEC);
7305 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
7306 	if (!(eec & (1 << 8)))
7307 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
7308 
7309 	/* PHY */
7310 	memcpy(&hw->phy.ops, ii->phy_ops, sizeof(hw->phy.ops));
7311 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
7312 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
7313 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
7314 	hw->phy.mdio.mmds = 0;
7315 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
7316 	hw->phy.mdio.dev = netdev;
7317 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
7318 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
7319 
7320 	ii->get_invariants(hw);
7321 
7322 	/* setup the private structure */
7323 	err = ixgbe_sw_init(adapter);
7324 	if (err)
7325 		goto err_sw_init;
7326 
7327 	/* Make it possible the adapter to be woken up via WOL */
7328 	switch (adapter->hw.mac.type) {
7329 	case ixgbe_mac_82599EB:
7330 	case ixgbe_mac_X540:
7331 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
7332 		break;
7333 	default:
7334 		break;
7335 	}
7336 
7337 	/*
7338 	 * If there is a fan on this device and it has failed log the
7339 	 * failure.
7340 	 */
7341 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
7342 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
7343 		if (esdp & IXGBE_ESDP_SDP1)
7344 			e_crit(probe, "Fan has stopped, replace the adapter\n");
7345 	}
7346 
7347 	if (allow_unsupported_sfp)
7348 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
7349 
7350 	/* reset_hw fills in the perm_addr as well */
7351 	hw->phy.reset_if_overtemp = true;
7352 	err = hw->mac.ops.reset_hw(hw);
7353 	hw->phy.reset_if_overtemp = false;
7354 	if (err == IXGBE_ERR_SFP_NOT_PRESENT &&
7355 	    hw->mac.type == ixgbe_mac_82598EB) {
7356 		err = 0;
7357 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
7358 		e_dev_err("failed to load because an unsupported SFP+ "
7359 			  "module type was detected.\n");
7360 		e_dev_err("Reload the driver after installing a supported "
7361 			  "module.\n");
7362 		goto err_sw_init;
7363 	} else if (err) {
7364 		e_dev_err("HW Init failed: %d\n", err);
7365 		goto err_sw_init;
7366 	}
7367 
7368 #ifdef CONFIG_PCI_IOV
7369 	ixgbe_enable_sriov(adapter, ii);
7370 
7371 #endif
7372 	netdev->features = NETIF_F_SG |
7373 			   NETIF_F_IP_CSUM |
7374 			   NETIF_F_IPV6_CSUM |
7375 			   NETIF_F_HW_VLAN_TX |
7376 			   NETIF_F_HW_VLAN_RX |
7377 			   NETIF_F_HW_VLAN_FILTER |
7378 			   NETIF_F_TSO |
7379 			   NETIF_F_TSO6 |
7380 			   NETIF_F_RXHASH |
7381 			   NETIF_F_RXCSUM;
7382 
7383 	netdev->hw_features = netdev->features;
7384 
7385 	switch (adapter->hw.mac.type) {
7386 	case ixgbe_mac_82599EB:
7387 	case ixgbe_mac_X540:
7388 		netdev->features |= NETIF_F_SCTP_CSUM;
7389 		netdev->hw_features |= NETIF_F_SCTP_CSUM |
7390 				       NETIF_F_NTUPLE;
7391 		break;
7392 	default:
7393 		break;
7394 	}
7395 
7396 	netdev->hw_features |= NETIF_F_RXALL;
7397 
7398 	netdev->vlan_features |= NETIF_F_TSO;
7399 	netdev->vlan_features |= NETIF_F_TSO6;
7400 	netdev->vlan_features |= NETIF_F_IP_CSUM;
7401 	netdev->vlan_features |= NETIF_F_IPV6_CSUM;
7402 	netdev->vlan_features |= NETIF_F_SG;
7403 
7404 	netdev->priv_flags |= IFF_UNICAST_FLT;
7405 	netdev->priv_flags |= IFF_SUPP_NOFCS;
7406 
7407 #ifdef CONFIG_IXGBE_DCB
7408 	netdev->dcbnl_ops = &dcbnl_ops;
7409 #endif
7410 
7411 #ifdef IXGBE_FCOE
7412 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
7413 		if (hw->mac.ops.get_device_caps) {
7414 			hw->mac.ops.get_device_caps(hw, &device_caps);
7415 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
7416 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
7417 		}
7418 
7419 		adapter->ring_feature[RING_F_FCOE].limit = IXGBE_FCRETA_SIZE;
7420 
7421 		netdev->features |= NETIF_F_FSO |
7422 				    NETIF_F_FCOE_CRC;
7423 
7424 		netdev->vlan_features |= NETIF_F_FSO |
7425 					 NETIF_F_FCOE_CRC |
7426 					 NETIF_F_FCOE_MTU;
7427 	}
7428 #endif /* IXGBE_FCOE */
7429 	if (pci_using_dac) {
7430 		netdev->features |= NETIF_F_HIGHDMA;
7431 		netdev->vlan_features |= NETIF_F_HIGHDMA;
7432 	}
7433 
7434 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
7435 		netdev->hw_features |= NETIF_F_LRO;
7436 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
7437 		netdev->features |= NETIF_F_LRO;
7438 
7439 	/* make sure the EEPROM is good */
7440 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
7441 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
7442 		err = -EIO;
7443 		goto err_sw_init;
7444 	}
7445 
7446 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
7447 	memcpy(netdev->perm_addr, hw->mac.perm_addr, netdev->addr_len);
7448 
7449 	if (!is_valid_ether_addr(netdev->perm_addr)) {
7450 		e_dev_err("invalid MAC address\n");
7451 		err = -EIO;
7452 		goto err_sw_init;
7453 	}
7454 
7455 	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
7456 		    (unsigned long) adapter);
7457 
7458 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
7459 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
7460 
7461 	err = ixgbe_init_interrupt_scheme(adapter);
7462 	if (err)
7463 		goto err_sw_init;
7464 
7465 	/* WOL not supported for all devices */
7466 	adapter->wol = 0;
7467 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
7468 	if (ixgbe_wol_supported(adapter, pdev->device, pdev->subsystem_device))
7469 		adapter->wol = IXGBE_WUFC_MAG;
7470 
7471 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
7472 
7473 	/* save off EEPROM version number */
7474 	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
7475 	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
7476 
7477 	/* pick up the PCI bus settings for reporting later */
7478 	hw->mac.ops.get_bus_info(hw);
7479 
7480 	/* print bus type/speed/width info */
7481 	e_dev_info("(PCI Express:%s:%s) %pM\n",
7482 		   (hw->bus.speed == ixgbe_bus_speed_5000 ? "5.0GT/s" :
7483 		    hw->bus.speed == ixgbe_bus_speed_2500 ? "2.5GT/s" :
7484 		    "Unknown"),
7485 		   (hw->bus.width == ixgbe_bus_width_pcie_x8 ? "Width x8" :
7486 		    hw->bus.width == ixgbe_bus_width_pcie_x4 ? "Width x4" :
7487 		    hw->bus.width == ixgbe_bus_width_pcie_x1 ? "Width x1" :
7488 		    "Unknown"),
7489 		   netdev->dev_addr);
7490 
7491 	err = ixgbe_read_pba_string_generic(hw, part_str, IXGBE_PBANUM_LENGTH);
7492 	if (err)
7493 		strncpy(part_str, "Unknown", IXGBE_PBANUM_LENGTH);
7494 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
7495 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
7496 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
7497 		           part_str);
7498 	else
7499 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
7500 			   hw->mac.type, hw->phy.type, part_str);
7501 
7502 	if (hw->bus.width <= ixgbe_bus_width_pcie_x4) {
7503 		e_dev_warn("PCI-Express bandwidth available for this card is "
7504 			   "not sufficient for optimal performance.\n");
7505 		e_dev_warn("For optimal performance a x8 PCI-Express slot "
7506 			   "is required.\n");
7507 	}
7508 
7509 	/* reset the hardware with the new settings */
7510 	err = hw->mac.ops.start_hw(hw);
7511 	if (err == IXGBE_ERR_EEPROM_VERSION) {
7512 		/* We are running on a pre-production device, log a warning */
7513 		e_dev_warn("This device is a pre-production adapter/LOM. "
7514 			   "Please be aware there may be issues associated "
7515 			   "with your hardware.  If you are experiencing "
7516 			   "problems please contact your Intel or hardware "
7517 			   "representative who provided you with this "
7518 			   "hardware.\n");
7519 	}
7520 	strcpy(netdev->name, "eth%d");
7521 	err = register_netdev(netdev);
7522 	if (err)
7523 		goto err_register;
7524 
7525 	/* power down the optics for 82599 SFP+ fiber */
7526 	if (hw->mac.ops.disable_tx_laser)
7527 		hw->mac.ops.disable_tx_laser(hw);
7528 
7529 	/* carrier off reporting is important to ethtool even BEFORE open */
7530 	netif_carrier_off(netdev);
7531 
7532 #ifdef CONFIG_IXGBE_DCA
7533 	if (dca_add_requester(&pdev->dev) == 0) {
7534 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
7535 		ixgbe_setup_dca(adapter);
7536 	}
7537 #endif
7538 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
7539 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
7540 		for (i = 0; i < adapter->num_vfs; i++)
7541 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
7542 	}
7543 
7544 	/* firmware requires driver version to be 0xFFFFFFFF
7545 	 * since os does not support feature
7546 	 */
7547 	if (hw->mac.ops.set_fw_drv_ver)
7548 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF,
7549 					   0xFF);
7550 
7551 	/* add san mac addr to netdev */
7552 	ixgbe_add_sanmac_netdev(netdev);
7553 
7554 	e_dev_info("%s\n", ixgbe_default_device_descr);
7555 	cards_found++;
7556 
7557 #ifdef CONFIG_IXGBE_HWMON
7558 	if (ixgbe_sysfs_init(adapter))
7559 		e_err(probe, "failed to allocate sysfs resources\n");
7560 #endif /* CONFIG_IXGBE_HWMON */
7561 
7562 #ifdef CONFIG_DEBUG_FS
7563 	ixgbe_dbg_adapter_init(adapter);
7564 #endif /* CONFIG_DEBUG_FS */
7565 
7566 	return 0;
7567 
7568 err_register:
7569 	ixgbe_release_hw_control(adapter);
7570 	ixgbe_clear_interrupt_scheme(adapter);
7571 err_sw_init:
7572 	ixgbe_disable_sriov(adapter);
7573 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
7574 	iounmap(hw->hw_addr);
7575 err_ioremap:
7576 	free_netdev(netdev);
7577 err_alloc_etherdev:
7578 	pci_release_selected_regions(pdev,
7579 				     pci_select_bars(pdev, IORESOURCE_MEM));
7580 err_pci_reg:
7581 err_dma:
7582 	pci_disable_device(pdev);
7583 	return err;
7584 }
7585 
7586 /**
7587  * ixgbe_remove - Device Removal Routine
7588  * @pdev: PCI device information struct
7589  *
7590  * ixgbe_remove is called by the PCI subsystem to alert the driver
7591  * that it should release a PCI device.  The could be caused by a
7592  * Hot-Plug event, or because the driver is going to be removed from
7593  * memory.
7594  **/
7595 static void ixgbe_remove(struct pci_dev *pdev)
7596 {
7597 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7598 	struct net_device *netdev = adapter->netdev;
7599 
7600 #ifdef CONFIG_DEBUG_FS
7601 	ixgbe_dbg_adapter_exit(adapter);
7602 #endif /*CONFIG_DEBUG_FS */
7603 
7604 	set_bit(__IXGBE_DOWN, &adapter->state);
7605 	cancel_work_sync(&adapter->service_task);
7606 
7607 
7608 #ifdef CONFIG_IXGBE_DCA
7609 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
7610 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
7611 		dca_remove_requester(&pdev->dev);
7612 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
7613 	}
7614 
7615 #endif
7616 #ifdef CONFIG_IXGBE_HWMON
7617 	ixgbe_sysfs_exit(adapter);
7618 #endif /* CONFIG_IXGBE_HWMON */
7619 
7620 	/* remove the added san mac */
7621 	ixgbe_del_sanmac_netdev(netdev);
7622 
7623 	if (netdev->reg_state == NETREG_REGISTERED)
7624 		unregister_netdev(netdev);
7625 
7626 	ixgbe_disable_sriov(adapter);
7627 
7628 	ixgbe_clear_interrupt_scheme(adapter);
7629 
7630 	ixgbe_release_hw_control(adapter);
7631 
7632 #ifdef CONFIG_DCB
7633 	kfree(adapter->ixgbe_ieee_pfc);
7634 	kfree(adapter->ixgbe_ieee_ets);
7635 
7636 #endif
7637 	iounmap(adapter->hw.hw_addr);
7638 	pci_release_selected_regions(pdev, pci_select_bars(pdev,
7639 				     IORESOURCE_MEM));
7640 
7641 	e_dev_info("complete\n");
7642 
7643 	free_netdev(netdev);
7644 
7645 	pci_disable_pcie_error_reporting(pdev);
7646 
7647 	pci_disable_device(pdev);
7648 }
7649 
7650 /**
7651  * ixgbe_io_error_detected - called when PCI error is detected
7652  * @pdev: Pointer to PCI device
7653  * @state: The current pci connection state
7654  *
7655  * This function is called after a PCI bus error affecting
7656  * this device has been detected.
7657  */
7658 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
7659 						pci_channel_state_t state)
7660 {
7661 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7662 	struct net_device *netdev = adapter->netdev;
7663 
7664 #ifdef CONFIG_PCI_IOV
7665 	struct pci_dev *bdev, *vfdev;
7666 	u32 dw0, dw1, dw2, dw3;
7667 	int vf, pos;
7668 	u16 req_id, pf_func;
7669 
7670 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7671 	    adapter->num_vfs == 0)
7672 		goto skip_bad_vf_detection;
7673 
7674 	bdev = pdev->bus->self;
7675 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
7676 		bdev = bdev->bus->self;
7677 
7678 	if (!bdev)
7679 		goto skip_bad_vf_detection;
7680 
7681 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
7682 	if (!pos)
7683 		goto skip_bad_vf_detection;
7684 
7685 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG, &dw0);
7686 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG + 4, &dw1);
7687 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG + 8, &dw2);
7688 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG + 12, &dw3);
7689 
7690 	req_id = dw1 >> 16;
7691 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
7692 	if (!(req_id & 0x0080))
7693 		goto skip_bad_vf_detection;
7694 
7695 	pf_func = req_id & 0x01;
7696 	if ((pf_func & 1) == (pdev->devfn & 1)) {
7697 		unsigned int device_id;
7698 
7699 		vf = (req_id & 0x7F) >> 1;
7700 		e_dev_err("VF %d has caused a PCIe error\n", vf);
7701 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
7702 				"%8.8x\tdw3: %8.8x\n",
7703 		dw0, dw1, dw2, dw3);
7704 		switch (adapter->hw.mac.type) {
7705 		case ixgbe_mac_82599EB:
7706 			device_id = IXGBE_82599_VF_DEVICE_ID;
7707 			break;
7708 		case ixgbe_mac_X540:
7709 			device_id = IXGBE_X540_VF_DEVICE_ID;
7710 			break;
7711 		default:
7712 			device_id = 0;
7713 			break;
7714 		}
7715 
7716 		/* Find the pci device of the offending VF */
7717 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
7718 		while (vfdev) {
7719 			if (vfdev->devfn == (req_id & 0xFF))
7720 				break;
7721 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
7722 					       device_id, vfdev);
7723 		}
7724 		/*
7725 		 * There's a slim chance the VF could have been hot plugged,
7726 		 * so if it is no longer present we don't need to issue the
7727 		 * VFLR.  Just clean up the AER in that case.
7728 		 */
7729 		if (vfdev) {
7730 			e_dev_err("Issuing VFLR to VF %d\n", vf);
7731 			pci_write_config_dword(vfdev, 0xA8, 0x00008000);
7732 		}
7733 
7734 		pci_cleanup_aer_uncorrect_error_status(pdev);
7735 	}
7736 
7737 	/*
7738 	 * Even though the error may have occurred on the other port
7739 	 * we still need to increment the vf error reference count for
7740 	 * both ports because the I/O resume function will be called
7741 	 * for both of them.
7742 	 */
7743 	adapter->vferr_refcount++;
7744 
7745 	return PCI_ERS_RESULT_RECOVERED;
7746 
7747 skip_bad_vf_detection:
7748 #endif /* CONFIG_PCI_IOV */
7749 	netif_device_detach(netdev);
7750 
7751 	if (state == pci_channel_io_perm_failure)
7752 		return PCI_ERS_RESULT_DISCONNECT;
7753 
7754 	if (netif_running(netdev))
7755 		ixgbe_down(adapter);
7756 	pci_disable_device(pdev);
7757 
7758 	/* Request a slot reset. */
7759 	return PCI_ERS_RESULT_NEED_RESET;
7760 }
7761 
7762 /**
7763  * ixgbe_io_slot_reset - called after the pci bus has been reset.
7764  * @pdev: Pointer to PCI device
7765  *
7766  * Restart the card from scratch, as if from a cold-boot.
7767  */
7768 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
7769 {
7770 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7771 	pci_ers_result_t result;
7772 	int err;
7773 
7774 	if (pci_enable_device_mem(pdev)) {
7775 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
7776 		result = PCI_ERS_RESULT_DISCONNECT;
7777 	} else {
7778 		pci_set_master(pdev);
7779 		pci_restore_state(pdev);
7780 		pci_save_state(pdev);
7781 
7782 		pci_wake_from_d3(pdev, false);
7783 
7784 		ixgbe_reset(adapter);
7785 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
7786 		result = PCI_ERS_RESULT_RECOVERED;
7787 	}
7788 
7789 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
7790 	if (err) {
7791 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
7792 			  "failed 0x%0x\n", err);
7793 		/* non-fatal, continue */
7794 	}
7795 
7796 	return result;
7797 }
7798 
7799 /**
7800  * ixgbe_io_resume - called when traffic can start flowing again.
7801  * @pdev: Pointer to PCI device
7802  *
7803  * This callback is called when the error recovery driver tells us that
7804  * its OK to resume normal operation.
7805  */
7806 static void ixgbe_io_resume(struct pci_dev *pdev)
7807 {
7808 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7809 	struct net_device *netdev = adapter->netdev;
7810 
7811 #ifdef CONFIG_PCI_IOV
7812 	if (adapter->vferr_refcount) {
7813 		e_info(drv, "Resuming after VF err\n");
7814 		adapter->vferr_refcount--;
7815 		return;
7816 	}
7817 
7818 #endif
7819 	if (netif_running(netdev))
7820 		ixgbe_up(adapter);
7821 
7822 	netif_device_attach(netdev);
7823 }
7824 
7825 static const struct pci_error_handlers ixgbe_err_handler = {
7826 	.error_detected = ixgbe_io_error_detected,
7827 	.slot_reset = ixgbe_io_slot_reset,
7828 	.resume = ixgbe_io_resume,
7829 };
7830 
7831 static struct pci_driver ixgbe_driver = {
7832 	.name     = ixgbe_driver_name,
7833 	.id_table = ixgbe_pci_tbl,
7834 	.probe    = ixgbe_probe,
7835 	.remove   = ixgbe_remove,
7836 #ifdef CONFIG_PM
7837 	.suspend  = ixgbe_suspend,
7838 	.resume   = ixgbe_resume,
7839 #endif
7840 	.shutdown = ixgbe_shutdown,
7841 	.err_handler = &ixgbe_err_handler
7842 };
7843 
7844 /**
7845  * ixgbe_init_module - Driver Registration Routine
7846  *
7847  * ixgbe_init_module is the first routine called when the driver is
7848  * loaded. All it does is register with the PCI subsystem.
7849  **/
7850 static int __init ixgbe_init_module(void)
7851 {
7852 	int ret;
7853 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
7854 	pr_info("%s\n", ixgbe_copyright);
7855 
7856 #ifdef CONFIG_DEBUG_FS
7857 	ixgbe_dbg_init();
7858 #endif /* CONFIG_DEBUG_FS */
7859 
7860 #ifdef CONFIG_IXGBE_DCA
7861 	dca_register_notify(&dca_notifier);
7862 #endif
7863 
7864 	ret = pci_register_driver(&ixgbe_driver);
7865 	return ret;
7866 }
7867 
7868 module_init(ixgbe_init_module);
7869 
7870 /**
7871  * ixgbe_exit_module - Driver Exit Cleanup Routine
7872  *
7873  * ixgbe_exit_module is called just before the driver is removed
7874  * from memory.
7875  **/
7876 static void __exit ixgbe_exit_module(void)
7877 {
7878 #ifdef CONFIG_IXGBE_DCA
7879 	dca_unregister_notify(&dca_notifier);
7880 #endif
7881 	pci_unregister_driver(&ixgbe_driver);
7882 
7883 #ifdef CONFIG_DEBUG_FS
7884 	ixgbe_dbg_exit();
7885 #endif /* CONFIG_DEBUG_FS */
7886 
7887 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
7888 }
7889 
7890 #ifdef CONFIG_IXGBE_DCA
7891 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
7892 			    void *p)
7893 {
7894 	int ret_val;
7895 
7896 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
7897 					 __ixgbe_notify_dca);
7898 
7899 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
7900 }
7901 
7902 #endif /* CONFIG_IXGBE_DCA */
7903 
7904 module_exit(ixgbe_exit_module);
7905 
7906 /* ixgbe_main.c */
7907