xref: /linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision 2dbf708448c836754d25fe6108c5bfe1f5697c95)
1 /*******************************************************************************
2 
3   Intel 10 Gigabit PCI Express Linux driver
4   Copyright(c) 1999 - 2012 Intel Corporation.
5 
6   This program is free software; you can redistribute it and/or modify it
7   under the terms and conditions of the GNU General Public License,
8   version 2, as published by the Free Software Foundation.
9 
10   This program is distributed in the hope it will be useful, but WITHOUT
11   ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12   FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13   more details.
14 
15   You should have received a copy of the GNU General Public License along with
16   this program; if not, write to the Free Software Foundation, Inc.,
17   51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
18 
19   The full GNU General Public License is included in this distribution in
20   the file called "COPYING".
21 
22   Contact Information:
23   e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
24   Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
25 
26 *******************************************************************************/
27 
28 #include <linux/types.h>
29 #include <linux/module.h>
30 #include <linux/pci.h>
31 #include <linux/netdevice.h>
32 #include <linux/vmalloc.h>
33 #include <linux/string.h>
34 #include <linux/in.h>
35 #include <linux/interrupt.h>
36 #include <linux/ip.h>
37 #include <linux/tcp.h>
38 #include <linux/sctp.h>
39 #include <linux/pkt_sched.h>
40 #include <linux/ipv6.h>
41 #include <linux/slab.h>
42 #include <net/checksum.h>
43 #include <net/ip6_checksum.h>
44 #include <linux/ethtool.h>
45 #include <linux/if.h>
46 #include <linux/if_vlan.h>
47 #include <linux/prefetch.h>
48 #include <scsi/fc/fc_fcoe.h>
49 
50 #include "ixgbe.h"
51 #include "ixgbe_common.h"
52 #include "ixgbe_dcb_82599.h"
53 #include "ixgbe_sriov.h"
54 
55 char ixgbe_driver_name[] = "ixgbe";
56 static const char ixgbe_driver_string[] =
57 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
58 #ifdef IXGBE_FCOE
59 char ixgbe_default_device_descr[] =
60 			      "Intel(R) 10 Gigabit Network Connection";
61 #else
62 static char ixgbe_default_device_descr[] =
63 			      "Intel(R) 10 Gigabit Network Connection";
64 #endif
65 #define MAJ 3
66 #define MIN 8
67 #define BUILD 21
68 #define DRV_VERSION __stringify(MAJ) "." __stringify(MIN) "." \
69 	__stringify(BUILD) "-k"
70 const char ixgbe_driver_version[] = DRV_VERSION;
71 static const char ixgbe_copyright[] =
72 				"Copyright (c) 1999-2012 Intel Corporation.";
73 
74 static const struct ixgbe_info *ixgbe_info_tbl[] = {
75 	[board_82598] = &ixgbe_82598_info,
76 	[board_82599] = &ixgbe_82599_info,
77 	[board_X540] = &ixgbe_X540_info,
78 };
79 
80 /* ixgbe_pci_tbl - PCI Device ID Table
81  *
82  * Wildcard entries (PCI_ANY_ID) should come last
83  * Last entry must be all 0s
84  *
85  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
86  *   Class, Class Mask, private data (not used) }
87  */
88 static DEFINE_PCI_DEVICE_TABLE(ixgbe_pci_tbl) = {
89 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
90 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
91 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
92 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
93 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
94 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
95 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
96 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
97 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
98 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
99 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
100 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
101 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
102 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
103 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
104 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
105 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
115 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
116 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
117 	/* required last entry */
118 	{0, }
119 };
120 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
121 
122 #ifdef CONFIG_IXGBE_DCA
123 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
124 			    void *p);
125 static struct notifier_block dca_notifier = {
126 	.notifier_call = ixgbe_notify_dca,
127 	.next          = NULL,
128 	.priority      = 0
129 };
130 #endif
131 
132 #ifdef CONFIG_PCI_IOV
133 static unsigned int max_vfs;
134 module_param(max_vfs, uint, 0);
135 MODULE_PARM_DESC(max_vfs,
136 		 "Maximum number of virtual functions to allocate per physical function");
137 #endif /* CONFIG_PCI_IOV */
138 
139 static unsigned int allow_unsupported_sfp;
140 module_param(allow_unsupported_sfp, uint, 0);
141 MODULE_PARM_DESC(allow_unsupported_sfp,
142 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
143 
144 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
145 static int debug = -1;
146 module_param(debug, int, 0);
147 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
148 
149 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
150 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
151 MODULE_LICENSE("GPL");
152 MODULE_VERSION(DRV_VERSION);
153 
154 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
155 {
156 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
157 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
158 		schedule_work(&adapter->service_task);
159 }
160 
161 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
162 {
163 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
164 
165 	/* flush memory to make sure state is correct before next watchdog */
166 	smp_mb__before_clear_bit();
167 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
168 }
169 
170 struct ixgbe_reg_info {
171 	u32 ofs;
172 	char *name;
173 };
174 
175 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
176 
177 	/* General Registers */
178 	{IXGBE_CTRL, "CTRL"},
179 	{IXGBE_STATUS, "STATUS"},
180 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
181 
182 	/* Interrupt Registers */
183 	{IXGBE_EICR, "EICR"},
184 
185 	/* RX Registers */
186 	{IXGBE_SRRCTL(0), "SRRCTL"},
187 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
188 	{IXGBE_RDLEN(0), "RDLEN"},
189 	{IXGBE_RDH(0), "RDH"},
190 	{IXGBE_RDT(0), "RDT"},
191 	{IXGBE_RXDCTL(0), "RXDCTL"},
192 	{IXGBE_RDBAL(0), "RDBAL"},
193 	{IXGBE_RDBAH(0), "RDBAH"},
194 
195 	/* TX Registers */
196 	{IXGBE_TDBAL(0), "TDBAL"},
197 	{IXGBE_TDBAH(0), "TDBAH"},
198 	{IXGBE_TDLEN(0), "TDLEN"},
199 	{IXGBE_TDH(0), "TDH"},
200 	{IXGBE_TDT(0), "TDT"},
201 	{IXGBE_TXDCTL(0), "TXDCTL"},
202 
203 	/* List Terminator */
204 	{}
205 };
206 
207 
208 /*
209  * ixgbe_regdump - register printout routine
210  */
211 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
212 {
213 	int i = 0, j = 0;
214 	char rname[16];
215 	u32 regs[64];
216 
217 	switch (reginfo->ofs) {
218 	case IXGBE_SRRCTL(0):
219 		for (i = 0; i < 64; i++)
220 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
221 		break;
222 	case IXGBE_DCA_RXCTRL(0):
223 		for (i = 0; i < 64; i++)
224 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
225 		break;
226 	case IXGBE_RDLEN(0):
227 		for (i = 0; i < 64; i++)
228 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
229 		break;
230 	case IXGBE_RDH(0):
231 		for (i = 0; i < 64; i++)
232 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
233 		break;
234 	case IXGBE_RDT(0):
235 		for (i = 0; i < 64; i++)
236 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
237 		break;
238 	case IXGBE_RXDCTL(0):
239 		for (i = 0; i < 64; i++)
240 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
241 		break;
242 	case IXGBE_RDBAL(0):
243 		for (i = 0; i < 64; i++)
244 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
245 		break;
246 	case IXGBE_RDBAH(0):
247 		for (i = 0; i < 64; i++)
248 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
249 		break;
250 	case IXGBE_TDBAL(0):
251 		for (i = 0; i < 64; i++)
252 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
253 		break;
254 	case IXGBE_TDBAH(0):
255 		for (i = 0; i < 64; i++)
256 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
257 		break;
258 	case IXGBE_TDLEN(0):
259 		for (i = 0; i < 64; i++)
260 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
261 		break;
262 	case IXGBE_TDH(0):
263 		for (i = 0; i < 64; i++)
264 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
265 		break;
266 	case IXGBE_TDT(0):
267 		for (i = 0; i < 64; i++)
268 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
269 		break;
270 	case IXGBE_TXDCTL(0):
271 		for (i = 0; i < 64; i++)
272 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
273 		break;
274 	default:
275 		pr_info("%-15s %08x\n", reginfo->name,
276 			IXGBE_READ_REG(hw, reginfo->ofs));
277 		return;
278 	}
279 
280 	for (i = 0; i < 8; i++) {
281 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i*8, i*8+7);
282 		pr_err("%-15s", rname);
283 		for (j = 0; j < 8; j++)
284 			pr_cont(" %08x", regs[i*8+j]);
285 		pr_cont("\n");
286 	}
287 
288 }
289 
290 /*
291  * ixgbe_dump - Print registers, tx-rings and rx-rings
292  */
293 static void ixgbe_dump(struct ixgbe_adapter *adapter)
294 {
295 	struct net_device *netdev = adapter->netdev;
296 	struct ixgbe_hw *hw = &adapter->hw;
297 	struct ixgbe_reg_info *reginfo;
298 	int n = 0;
299 	struct ixgbe_ring *tx_ring;
300 	struct ixgbe_tx_buffer *tx_buffer;
301 	union ixgbe_adv_tx_desc *tx_desc;
302 	struct my_u0 { u64 a; u64 b; } *u0;
303 	struct ixgbe_ring *rx_ring;
304 	union ixgbe_adv_rx_desc *rx_desc;
305 	struct ixgbe_rx_buffer *rx_buffer_info;
306 	u32 staterr;
307 	int i = 0;
308 
309 	if (!netif_msg_hw(adapter))
310 		return;
311 
312 	/* Print netdevice Info */
313 	if (netdev) {
314 		dev_info(&adapter->pdev->dev, "Net device Info\n");
315 		pr_info("Device Name     state            "
316 			"trans_start      last_rx\n");
317 		pr_info("%-15s %016lX %016lX %016lX\n",
318 			netdev->name,
319 			netdev->state,
320 			netdev->trans_start,
321 			netdev->last_rx);
322 	}
323 
324 	/* Print Registers */
325 	dev_info(&adapter->pdev->dev, "Register Dump\n");
326 	pr_info(" Register Name   Value\n");
327 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
328 	     reginfo->name; reginfo++) {
329 		ixgbe_regdump(hw, reginfo);
330 	}
331 
332 	/* Print TX Ring Summary */
333 	if (!netdev || !netif_running(netdev))
334 		goto exit;
335 
336 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
337 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
338 	for (n = 0; n < adapter->num_tx_queues; n++) {
339 		tx_ring = adapter->tx_ring[n];
340 		tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
341 		pr_info(" %5d %5X %5X %016llX %04X %p %016llX\n",
342 			   n, tx_ring->next_to_use, tx_ring->next_to_clean,
343 			   (u64)dma_unmap_addr(tx_buffer, dma),
344 			   dma_unmap_len(tx_buffer, len),
345 			   tx_buffer->next_to_watch,
346 			   (u64)tx_buffer->time_stamp);
347 	}
348 
349 	/* Print TX Rings */
350 	if (!netif_msg_tx_done(adapter))
351 		goto rx_ring_summary;
352 
353 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
354 
355 	/* Transmit Descriptor Formats
356 	 *
357 	 * Advanced Transmit Descriptor
358 	 *   +--------------------------------------------------------------+
359 	 * 0 |         Buffer Address [63:0]                                |
360 	 *   +--------------------------------------------------------------+
361 	 * 8 |  PAYLEN  | PORTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
362 	 *   +--------------------------------------------------------------+
363 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
364 	 */
365 
366 	for (n = 0; n < adapter->num_tx_queues; n++) {
367 		tx_ring = adapter->tx_ring[n];
368 		pr_info("------------------------------------\n");
369 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
370 		pr_info("------------------------------------\n");
371 		pr_info("T [desc]     [address 63:0  ] "
372 			"[PlPOIdStDDt Ln] [bi->dma       ] "
373 			"leng  ntw timestamp        bi->skb\n");
374 
375 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
376 			tx_desc = IXGBE_TX_DESC(tx_ring, i);
377 			tx_buffer = &tx_ring->tx_buffer_info[i];
378 			u0 = (struct my_u0 *)tx_desc;
379 			pr_info("T [0x%03X]    %016llX %016llX %016llX"
380 				" %04X  %p %016llX %p", i,
381 				le64_to_cpu(u0->a),
382 				le64_to_cpu(u0->b),
383 				(u64)dma_unmap_addr(tx_buffer, dma),
384 				dma_unmap_len(tx_buffer, len),
385 				tx_buffer->next_to_watch,
386 				(u64)tx_buffer->time_stamp,
387 				tx_buffer->skb);
388 			if (i == tx_ring->next_to_use &&
389 				i == tx_ring->next_to_clean)
390 				pr_cont(" NTC/U\n");
391 			else if (i == tx_ring->next_to_use)
392 				pr_cont(" NTU\n");
393 			else if (i == tx_ring->next_to_clean)
394 				pr_cont(" NTC\n");
395 			else
396 				pr_cont("\n");
397 
398 			if (netif_msg_pktdata(adapter) &&
399 			    dma_unmap_len(tx_buffer, len) != 0)
400 				print_hex_dump(KERN_INFO, "",
401 					DUMP_PREFIX_ADDRESS, 16, 1,
402 					phys_to_virt(dma_unmap_addr(tx_buffer,
403 								    dma)),
404 					dma_unmap_len(tx_buffer, len),
405 					true);
406 		}
407 	}
408 
409 	/* Print RX Rings Summary */
410 rx_ring_summary:
411 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
412 	pr_info("Queue [NTU] [NTC]\n");
413 	for (n = 0; n < adapter->num_rx_queues; n++) {
414 		rx_ring = adapter->rx_ring[n];
415 		pr_info("%5d %5X %5X\n",
416 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
417 	}
418 
419 	/* Print RX Rings */
420 	if (!netif_msg_rx_status(adapter))
421 		goto exit;
422 
423 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
424 
425 	/* Advanced Receive Descriptor (Read) Format
426 	 *    63                                           1        0
427 	 *    +-----------------------------------------------------+
428 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
429 	 *    +----------------------------------------------+------+
430 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
431 	 *    +-----------------------------------------------------+
432 	 *
433 	 *
434 	 * Advanced Receive Descriptor (Write-Back) Format
435 	 *
436 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
437 	 *   +------------------------------------------------------+
438 	 * 0 | Packet     IP     |SPH| HDR_LEN   | RSV|Packet|  RSS |
439 	 *   | Checksum   Ident  |   |           |    | Type | Type |
440 	 *   +------------------------------------------------------+
441 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
442 	 *   +------------------------------------------------------+
443 	 *   63       48 47    32 31            20 19               0
444 	 */
445 	for (n = 0; n < adapter->num_rx_queues; n++) {
446 		rx_ring = adapter->rx_ring[n];
447 		pr_info("------------------------------------\n");
448 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
449 		pr_info("------------------------------------\n");
450 		pr_info("R  [desc]      [ PktBuf     A0] "
451 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb] "
452 			"<-- Adv Rx Read format\n");
453 		pr_info("RWB[desc]      [PcsmIpSHl PtRs] "
454 			"[vl er S cks ln] ---------------- [bi->skb] "
455 			"<-- Adv Rx Write-Back format\n");
456 
457 		for (i = 0; i < rx_ring->count; i++) {
458 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
459 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
460 			u0 = (struct my_u0 *)rx_desc;
461 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
462 			if (staterr & IXGBE_RXD_STAT_DD) {
463 				/* Descriptor Done */
464 				pr_info("RWB[0x%03X]     %016llX "
465 					"%016llX ---------------- %p", i,
466 					le64_to_cpu(u0->a),
467 					le64_to_cpu(u0->b),
468 					rx_buffer_info->skb);
469 			} else {
470 				pr_info("R  [0x%03X]     %016llX "
471 					"%016llX %016llX %p", i,
472 					le64_to_cpu(u0->a),
473 					le64_to_cpu(u0->b),
474 					(u64)rx_buffer_info->dma,
475 					rx_buffer_info->skb);
476 
477 				if (netif_msg_pktdata(adapter)) {
478 					print_hex_dump(KERN_INFO, "",
479 					   DUMP_PREFIX_ADDRESS, 16, 1,
480 					   phys_to_virt(rx_buffer_info->dma),
481 					   ixgbe_rx_bufsz(rx_ring), true);
482 				}
483 			}
484 
485 			if (i == rx_ring->next_to_use)
486 				pr_cont(" NTU\n");
487 			else if (i == rx_ring->next_to_clean)
488 				pr_cont(" NTC\n");
489 			else
490 				pr_cont("\n");
491 
492 		}
493 	}
494 
495 exit:
496 	return;
497 }
498 
499 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
500 {
501 	u32 ctrl_ext;
502 
503 	/* Let firmware take over control of h/w */
504 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
505 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
506 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
507 }
508 
509 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
510 {
511 	u32 ctrl_ext;
512 
513 	/* Let firmware know the driver has taken over */
514 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
515 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
516 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
517 }
518 
519 /*
520  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
521  * @adapter: pointer to adapter struct
522  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
523  * @queue: queue to map the corresponding interrupt to
524  * @msix_vector: the vector to map to the corresponding queue
525  *
526  */
527 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
528 			   u8 queue, u8 msix_vector)
529 {
530 	u32 ivar, index;
531 	struct ixgbe_hw *hw = &adapter->hw;
532 	switch (hw->mac.type) {
533 	case ixgbe_mac_82598EB:
534 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
535 		if (direction == -1)
536 			direction = 0;
537 		index = (((direction * 64) + queue) >> 2) & 0x1F;
538 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
539 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
540 		ivar |= (msix_vector << (8 * (queue & 0x3)));
541 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
542 		break;
543 	case ixgbe_mac_82599EB:
544 	case ixgbe_mac_X540:
545 		if (direction == -1) {
546 			/* other causes */
547 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
548 			index = ((queue & 1) * 8);
549 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
550 			ivar &= ~(0xFF << index);
551 			ivar |= (msix_vector << index);
552 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
553 			break;
554 		} else {
555 			/* tx or rx causes */
556 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
557 			index = ((16 * (queue & 1)) + (8 * direction));
558 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
559 			ivar &= ~(0xFF << index);
560 			ivar |= (msix_vector << index);
561 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
562 			break;
563 		}
564 	default:
565 		break;
566 	}
567 }
568 
569 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
570 					  u64 qmask)
571 {
572 	u32 mask;
573 
574 	switch (adapter->hw.mac.type) {
575 	case ixgbe_mac_82598EB:
576 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
577 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
578 		break;
579 	case ixgbe_mac_82599EB:
580 	case ixgbe_mac_X540:
581 		mask = (qmask & 0xFFFFFFFF);
582 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
583 		mask = (qmask >> 32);
584 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
585 		break;
586 	default:
587 		break;
588 	}
589 }
590 
591 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring *ring,
592 				      struct ixgbe_tx_buffer *tx_buffer)
593 {
594 	if (tx_buffer->skb) {
595 		dev_kfree_skb_any(tx_buffer->skb);
596 		if (dma_unmap_len(tx_buffer, len))
597 			dma_unmap_single(ring->dev,
598 					 dma_unmap_addr(tx_buffer, dma),
599 					 dma_unmap_len(tx_buffer, len),
600 					 DMA_TO_DEVICE);
601 	} else if (dma_unmap_len(tx_buffer, len)) {
602 		dma_unmap_page(ring->dev,
603 			       dma_unmap_addr(tx_buffer, dma),
604 			       dma_unmap_len(tx_buffer, len),
605 			       DMA_TO_DEVICE);
606 	}
607 	tx_buffer->next_to_watch = NULL;
608 	tx_buffer->skb = NULL;
609 	dma_unmap_len_set(tx_buffer, len, 0);
610 	/* tx_buffer must be completely set up in the transmit path */
611 }
612 
613 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
614 {
615 	struct ixgbe_hw *hw = &adapter->hw;
616 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
617 	u32 data = 0;
618 	u32 xoff[8] = {0};
619 	int i;
620 
621 	if ((hw->fc.current_mode == ixgbe_fc_full) ||
622 	    (hw->fc.current_mode == ixgbe_fc_rx_pause)) {
623 		switch (hw->mac.type) {
624 		case ixgbe_mac_82598EB:
625 			data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
626 			break;
627 		default:
628 			data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
629 		}
630 		hwstats->lxoffrxc += data;
631 
632 		/* refill credits (no tx hang) if we received xoff */
633 		if (!data)
634 			return;
635 
636 		for (i = 0; i < adapter->num_tx_queues; i++)
637 			clear_bit(__IXGBE_HANG_CHECK_ARMED,
638 				  &adapter->tx_ring[i]->state);
639 		return;
640 	} else if (!(adapter->dcb_cfg.pfc_mode_enable))
641 		return;
642 
643 	/* update stats for each tc, only valid with PFC enabled */
644 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
645 		switch (hw->mac.type) {
646 		case ixgbe_mac_82598EB:
647 			xoff[i] = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
648 			break;
649 		default:
650 			xoff[i] = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
651 		}
652 		hwstats->pxoffrxc[i] += xoff[i];
653 	}
654 
655 	/* disarm tx queues that have received xoff frames */
656 	for (i = 0; i < adapter->num_tx_queues; i++) {
657 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
658 		u8 tc = tx_ring->dcb_tc;
659 
660 		if (xoff[tc])
661 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
662 	}
663 }
664 
665 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
666 {
667 	return ring->stats.packets;
668 }
669 
670 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
671 {
672 	struct ixgbe_adapter *adapter = netdev_priv(ring->netdev);
673 	struct ixgbe_hw *hw = &adapter->hw;
674 
675 	u32 head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
676 	u32 tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
677 
678 	if (head != tail)
679 		return (head < tail) ?
680 			tail - head : (tail + ring->count - head);
681 
682 	return 0;
683 }
684 
685 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
686 {
687 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
688 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
689 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
690 	bool ret = false;
691 
692 	clear_check_for_tx_hang(tx_ring);
693 
694 	/*
695 	 * Check for a hung queue, but be thorough. This verifies
696 	 * that a transmit has been completed since the previous
697 	 * check AND there is at least one packet pending. The
698 	 * ARMED bit is set to indicate a potential hang. The
699 	 * bit is cleared if a pause frame is received to remove
700 	 * false hang detection due to PFC or 802.3x frames. By
701 	 * requiring this to fail twice we avoid races with
702 	 * pfc clearing the ARMED bit and conditions where we
703 	 * run the check_tx_hang logic with a transmit completion
704 	 * pending but without time to complete it yet.
705 	 */
706 	if ((tx_done_old == tx_done) && tx_pending) {
707 		/* make sure it is true for two checks in a row */
708 		ret = test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
709 				       &tx_ring->state);
710 	} else {
711 		/* update completed stats and continue */
712 		tx_ring->tx_stats.tx_done_old = tx_done;
713 		/* reset the countdown */
714 		clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
715 	}
716 
717 	return ret;
718 }
719 
720 /**
721  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
722  * @adapter: driver private struct
723  **/
724 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
725 {
726 
727 	/* Do the reset outside of interrupt context */
728 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
729 		adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
730 		ixgbe_service_event_schedule(adapter);
731 	}
732 }
733 
734 /**
735  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
736  * @q_vector: structure containing interrupt and ring information
737  * @tx_ring: tx ring to clean
738  **/
739 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
740 			       struct ixgbe_ring *tx_ring)
741 {
742 	struct ixgbe_adapter *adapter = q_vector->adapter;
743 	struct ixgbe_tx_buffer *tx_buffer;
744 	union ixgbe_adv_tx_desc *tx_desc;
745 	unsigned int total_bytes = 0, total_packets = 0;
746 	unsigned int budget = q_vector->tx.work_limit;
747 	unsigned int i = tx_ring->next_to_clean;
748 
749 	if (test_bit(__IXGBE_DOWN, &adapter->state))
750 		return true;
751 
752 	tx_buffer = &tx_ring->tx_buffer_info[i];
753 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
754 	i -= tx_ring->count;
755 
756 	do {
757 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
758 
759 		/* if next_to_watch is not set then there is no work pending */
760 		if (!eop_desc)
761 			break;
762 
763 		/* prevent any other reads prior to eop_desc */
764 		rmb();
765 
766 		/* if DD is not set pending work has not been completed */
767 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
768 			break;
769 
770 		/* clear next_to_watch to prevent false hangs */
771 		tx_buffer->next_to_watch = NULL;
772 
773 		/* update the statistics for this packet */
774 		total_bytes += tx_buffer->bytecount;
775 		total_packets += tx_buffer->gso_segs;
776 
777 		/* free the skb */
778 		dev_kfree_skb_any(tx_buffer->skb);
779 
780 		/* unmap skb header data */
781 		dma_unmap_single(tx_ring->dev,
782 				 dma_unmap_addr(tx_buffer, dma),
783 				 dma_unmap_len(tx_buffer, len),
784 				 DMA_TO_DEVICE);
785 
786 		/* clear tx_buffer data */
787 		tx_buffer->skb = NULL;
788 		dma_unmap_len_set(tx_buffer, len, 0);
789 
790 		/* unmap remaining buffers */
791 		while (tx_desc != eop_desc) {
792 			tx_buffer++;
793 			tx_desc++;
794 			i++;
795 			if (unlikely(!i)) {
796 				i -= tx_ring->count;
797 				tx_buffer = tx_ring->tx_buffer_info;
798 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
799 			}
800 
801 			/* unmap any remaining paged data */
802 			if (dma_unmap_len(tx_buffer, len)) {
803 				dma_unmap_page(tx_ring->dev,
804 					       dma_unmap_addr(tx_buffer, dma),
805 					       dma_unmap_len(tx_buffer, len),
806 					       DMA_TO_DEVICE);
807 				dma_unmap_len_set(tx_buffer, len, 0);
808 			}
809 		}
810 
811 		/* move us one more past the eop_desc for start of next pkt */
812 		tx_buffer++;
813 		tx_desc++;
814 		i++;
815 		if (unlikely(!i)) {
816 			i -= tx_ring->count;
817 			tx_buffer = tx_ring->tx_buffer_info;
818 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
819 		}
820 
821 		/* issue prefetch for next Tx descriptor */
822 		prefetch(tx_desc);
823 
824 		/* update budget accounting */
825 		budget--;
826 	} while (likely(budget));
827 
828 	i += tx_ring->count;
829 	tx_ring->next_to_clean = i;
830 	u64_stats_update_begin(&tx_ring->syncp);
831 	tx_ring->stats.bytes += total_bytes;
832 	tx_ring->stats.packets += total_packets;
833 	u64_stats_update_end(&tx_ring->syncp);
834 	q_vector->tx.total_bytes += total_bytes;
835 	q_vector->tx.total_packets += total_packets;
836 
837 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
838 		/* schedule immediate reset if we believe we hung */
839 		struct ixgbe_hw *hw = &adapter->hw;
840 		e_err(drv, "Detected Tx Unit Hang\n"
841 			"  Tx Queue             <%d>\n"
842 			"  TDH, TDT             <%x>, <%x>\n"
843 			"  next_to_use          <%x>\n"
844 			"  next_to_clean        <%x>\n"
845 			"tx_buffer_info[next_to_clean]\n"
846 			"  time_stamp           <%lx>\n"
847 			"  jiffies              <%lx>\n",
848 			tx_ring->queue_index,
849 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
850 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
851 			tx_ring->next_to_use, i,
852 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
853 
854 		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
855 
856 		e_info(probe,
857 		       "tx hang %d detected on queue %d, resetting adapter\n",
858 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
859 
860 		/* schedule immediate reset if we believe we hung */
861 		ixgbe_tx_timeout_reset(adapter);
862 
863 		/* the adapter is about to reset, no point in enabling stuff */
864 		return true;
865 	}
866 
867 	netdev_tx_completed_queue(txring_txq(tx_ring),
868 				  total_packets, total_bytes);
869 
870 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
871 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
872 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
873 		/* Make sure that anybody stopping the queue after this
874 		 * sees the new next_to_clean.
875 		 */
876 		smp_mb();
877 		if (__netif_subqueue_stopped(tx_ring->netdev,
878 					     tx_ring->queue_index)
879 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
880 			netif_wake_subqueue(tx_ring->netdev,
881 					    tx_ring->queue_index);
882 			++tx_ring->tx_stats.restart_queue;
883 		}
884 	}
885 
886 	return !!budget;
887 }
888 
889 #ifdef CONFIG_IXGBE_DCA
890 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
891 				struct ixgbe_ring *tx_ring,
892 				int cpu)
893 {
894 	struct ixgbe_hw *hw = &adapter->hw;
895 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
896 	u16 reg_offset;
897 
898 	switch (hw->mac.type) {
899 	case ixgbe_mac_82598EB:
900 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
901 		break;
902 	case ixgbe_mac_82599EB:
903 	case ixgbe_mac_X540:
904 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
905 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
906 		break;
907 	default:
908 		/* for unknown hardware do not write register */
909 		return;
910 	}
911 
912 	/*
913 	 * We can enable relaxed ordering for reads, but not writes when
914 	 * DCA is enabled.  This is due to a known issue in some chipsets
915 	 * which will cause the DCA tag to be cleared.
916 	 */
917 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
918 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
919 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
920 
921 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
922 }
923 
924 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
925 				struct ixgbe_ring *rx_ring,
926 				int cpu)
927 {
928 	struct ixgbe_hw *hw = &adapter->hw;
929 	u32 rxctrl = dca3_get_tag(rx_ring->dev, cpu);
930 	u8 reg_idx = rx_ring->reg_idx;
931 
932 
933 	switch (hw->mac.type) {
934 	case ixgbe_mac_82599EB:
935 	case ixgbe_mac_X540:
936 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
937 		break;
938 	default:
939 		break;
940 	}
941 
942 	/*
943 	 * We can enable relaxed ordering for reads, but not writes when
944 	 * DCA is enabled.  This is due to a known issue in some chipsets
945 	 * which will cause the DCA tag to be cleared.
946 	 */
947 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
948 		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
949 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
950 
951 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
952 }
953 
954 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
955 {
956 	struct ixgbe_adapter *adapter = q_vector->adapter;
957 	struct ixgbe_ring *ring;
958 	int cpu = get_cpu();
959 
960 	if (q_vector->cpu == cpu)
961 		goto out_no_update;
962 
963 	ixgbe_for_each_ring(ring, q_vector->tx)
964 		ixgbe_update_tx_dca(adapter, ring, cpu);
965 
966 	ixgbe_for_each_ring(ring, q_vector->rx)
967 		ixgbe_update_rx_dca(adapter, ring, cpu);
968 
969 	q_vector->cpu = cpu;
970 out_no_update:
971 	put_cpu();
972 }
973 
974 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
975 {
976 	int num_q_vectors;
977 	int i;
978 
979 	if (!(adapter->flags & IXGBE_FLAG_DCA_ENABLED))
980 		return;
981 
982 	/* always use CB2 mode, difference is masked in the CB driver */
983 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 2);
984 
985 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
986 		num_q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
987 	else
988 		num_q_vectors = 1;
989 
990 	for (i = 0; i < num_q_vectors; i++) {
991 		adapter->q_vector[i]->cpu = -1;
992 		ixgbe_update_dca(adapter->q_vector[i]);
993 	}
994 }
995 
996 static int __ixgbe_notify_dca(struct device *dev, void *data)
997 {
998 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
999 	unsigned long event = *(unsigned long *)data;
1000 
1001 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1002 		return 0;
1003 
1004 	switch (event) {
1005 	case DCA_PROVIDER_ADD:
1006 		/* if we're already enabled, don't do it again */
1007 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1008 			break;
1009 		if (dca_add_requester(dev) == 0) {
1010 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1011 			ixgbe_setup_dca(adapter);
1012 			break;
1013 		}
1014 		/* Fall Through since DCA is disabled. */
1015 	case DCA_PROVIDER_REMOVE:
1016 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1017 			dca_remove_requester(dev);
1018 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1019 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
1020 		}
1021 		break;
1022 	}
1023 
1024 	return 0;
1025 }
1026 
1027 #endif /* CONFIG_IXGBE_DCA */
1028 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1029 				 union ixgbe_adv_rx_desc *rx_desc,
1030 				 struct sk_buff *skb)
1031 {
1032 	if (ring->netdev->features & NETIF_F_RXHASH)
1033 		skb->rxhash = le32_to_cpu(rx_desc->wb.lower.hi_dword.rss);
1034 }
1035 
1036 #ifdef IXGBE_FCOE
1037 /**
1038  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1039  * @adapter: address of board private structure
1040  * @rx_desc: advanced rx descriptor
1041  *
1042  * Returns : true if it is FCoE pkt
1043  */
1044 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_adapter *adapter,
1045 				    union ixgbe_adv_rx_desc *rx_desc)
1046 {
1047 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1048 
1049 	return (adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
1050 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1051 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1052 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1053 }
1054 
1055 #endif /* IXGBE_FCOE */
1056 /**
1057  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1058  * @ring: structure containing ring specific data
1059  * @rx_desc: current Rx descriptor being processed
1060  * @skb: skb currently being received and modified
1061  **/
1062 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1063 				     union ixgbe_adv_rx_desc *rx_desc,
1064 				     struct sk_buff *skb)
1065 {
1066 	skb_checksum_none_assert(skb);
1067 
1068 	/* Rx csum disabled */
1069 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1070 		return;
1071 
1072 	/* if IP and error */
1073 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1074 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1075 		ring->rx_stats.csum_err++;
1076 		return;
1077 	}
1078 
1079 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1080 		return;
1081 
1082 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1083 		__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1084 
1085 		/*
1086 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1087 		 * checksum errors.
1088 		 */
1089 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1090 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1091 			return;
1092 
1093 		ring->rx_stats.csum_err++;
1094 		return;
1095 	}
1096 
1097 	/* It must be a TCP or UDP packet with a valid checksum */
1098 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1099 }
1100 
1101 static inline void ixgbe_release_rx_desc(struct ixgbe_ring *rx_ring, u32 val)
1102 {
1103 	rx_ring->next_to_use = val;
1104 
1105 	/* update next to alloc since we have filled the ring */
1106 	rx_ring->next_to_alloc = val;
1107 	/*
1108 	 * Force memory writes to complete before letting h/w
1109 	 * know there are new descriptors to fetch.  (Only
1110 	 * applicable for weak-ordered memory model archs,
1111 	 * such as IA-64).
1112 	 */
1113 	wmb();
1114 	writel(val, rx_ring->tail);
1115 }
1116 
1117 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1118 				    struct ixgbe_rx_buffer *bi)
1119 {
1120 	struct page *page = bi->page;
1121 	dma_addr_t dma = bi->dma;
1122 
1123 	/* since we are recycling buffers we should seldom need to alloc */
1124 	if (likely(dma))
1125 		return true;
1126 
1127 	/* alloc new page for storage */
1128 	if (likely(!page)) {
1129 		page = alloc_pages(GFP_ATOMIC | __GFP_COLD,
1130 				   ixgbe_rx_pg_order(rx_ring));
1131 		if (unlikely(!page)) {
1132 			rx_ring->rx_stats.alloc_rx_page_failed++;
1133 			return false;
1134 		}
1135 		bi->page = page;
1136 	}
1137 
1138 	/* map page for use */
1139 	dma = dma_map_page(rx_ring->dev, page, 0,
1140 			   ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1141 
1142 	/*
1143 	 * if mapping failed free memory back to system since
1144 	 * there isn't much point in holding memory we can't use
1145 	 */
1146 	if (dma_mapping_error(rx_ring->dev, dma)) {
1147 		put_page(page);
1148 		bi->page = NULL;
1149 
1150 		rx_ring->rx_stats.alloc_rx_page_failed++;
1151 		return false;
1152 	}
1153 
1154 	bi->dma = dma;
1155 	bi->page_offset ^= ixgbe_rx_bufsz(rx_ring);
1156 
1157 	return true;
1158 }
1159 
1160 /**
1161  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1162  * @rx_ring: ring to place buffers on
1163  * @cleaned_count: number of buffers to replace
1164  **/
1165 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1166 {
1167 	union ixgbe_adv_rx_desc *rx_desc;
1168 	struct ixgbe_rx_buffer *bi;
1169 	u16 i = rx_ring->next_to_use;
1170 
1171 	/* nothing to do */
1172 	if (!cleaned_count)
1173 		return;
1174 
1175 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1176 	bi = &rx_ring->rx_buffer_info[i];
1177 	i -= rx_ring->count;
1178 
1179 	do {
1180 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1181 			break;
1182 
1183 		/*
1184 		 * Refresh the desc even if buffer_addrs didn't change
1185 		 * because each write-back erases this info.
1186 		 */
1187 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1188 
1189 		rx_desc++;
1190 		bi++;
1191 		i++;
1192 		if (unlikely(!i)) {
1193 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1194 			bi = rx_ring->rx_buffer_info;
1195 			i -= rx_ring->count;
1196 		}
1197 
1198 		/* clear the hdr_addr for the next_to_use descriptor */
1199 		rx_desc->read.hdr_addr = 0;
1200 
1201 		cleaned_count--;
1202 	} while (cleaned_count);
1203 
1204 	i += rx_ring->count;
1205 
1206 	if (rx_ring->next_to_use != i)
1207 		ixgbe_release_rx_desc(rx_ring, i);
1208 }
1209 
1210 /**
1211  * ixgbe_get_headlen - determine size of header for RSC/LRO/GRO/FCOE
1212  * @data: pointer to the start of the headers
1213  * @max_len: total length of section to find headers in
1214  *
1215  * This function is meant to determine the length of headers that will
1216  * be recognized by hardware for LRO, GRO, and RSC offloads.  The main
1217  * motivation of doing this is to only perform one pull for IPv4 TCP
1218  * packets so that we can do basic things like calculating the gso_size
1219  * based on the average data per packet.
1220  **/
1221 static unsigned int ixgbe_get_headlen(unsigned char *data,
1222 				      unsigned int max_len)
1223 {
1224 	union {
1225 		unsigned char *network;
1226 		/* l2 headers */
1227 		struct ethhdr *eth;
1228 		struct vlan_hdr *vlan;
1229 		/* l3 headers */
1230 		struct iphdr *ipv4;
1231 	} hdr;
1232 	__be16 protocol;
1233 	u8 nexthdr = 0;	/* default to not TCP */
1234 	u8 hlen;
1235 
1236 	/* this should never happen, but better safe than sorry */
1237 	if (max_len < ETH_HLEN)
1238 		return max_len;
1239 
1240 	/* initialize network frame pointer */
1241 	hdr.network = data;
1242 
1243 	/* set first protocol and move network header forward */
1244 	protocol = hdr.eth->h_proto;
1245 	hdr.network += ETH_HLEN;
1246 
1247 	/* handle any vlan tag if present */
1248 	if (protocol == __constant_htons(ETH_P_8021Q)) {
1249 		if ((hdr.network - data) > (max_len - VLAN_HLEN))
1250 			return max_len;
1251 
1252 		protocol = hdr.vlan->h_vlan_encapsulated_proto;
1253 		hdr.network += VLAN_HLEN;
1254 	}
1255 
1256 	/* handle L3 protocols */
1257 	if (protocol == __constant_htons(ETH_P_IP)) {
1258 		if ((hdr.network - data) > (max_len - sizeof(struct iphdr)))
1259 			return max_len;
1260 
1261 		/* access ihl as a u8 to avoid unaligned access on ia64 */
1262 		hlen = (hdr.network[0] & 0x0F) << 2;
1263 
1264 		/* verify hlen meets minimum size requirements */
1265 		if (hlen < sizeof(struct iphdr))
1266 			return hdr.network - data;
1267 
1268 		/* record next protocol */
1269 		nexthdr = hdr.ipv4->protocol;
1270 		hdr.network += hlen;
1271 #ifdef IXGBE_FCOE
1272 	} else if (protocol == __constant_htons(ETH_P_FCOE)) {
1273 		if ((hdr.network - data) > (max_len - FCOE_HEADER_LEN))
1274 			return max_len;
1275 		hdr.network += FCOE_HEADER_LEN;
1276 #endif
1277 	} else {
1278 		return hdr.network - data;
1279 	}
1280 
1281 	/* finally sort out TCP */
1282 	if (nexthdr == IPPROTO_TCP) {
1283 		if ((hdr.network - data) > (max_len - sizeof(struct tcphdr)))
1284 			return max_len;
1285 
1286 		/* access doff as a u8 to avoid unaligned access on ia64 */
1287 		hlen = (hdr.network[12] & 0xF0) >> 2;
1288 
1289 		/* verify hlen meets minimum size requirements */
1290 		if (hlen < sizeof(struct tcphdr))
1291 			return hdr.network - data;
1292 
1293 		hdr.network += hlen;
1294 	}
1295 
1296 	/*
1297 	 * If everything has gone correctly hdr.network should be the
1298 	 * data section of the packet and will be the end of the header.
1299 	 * If not then it probably represents the end of the last recognized
1300 	 * header.
1301 	 */
1302 	if ((hdr.network - data) < max_len)
1303 		return hdr.network - data;
1304 	else
1305 		return max_len;
1306 }
1307 
1308 static void ixgbe_get_rsc_cnt(struct ixgbe_ring *rx_ring,
1309 			      union ixgbe_adv_rx_desc *rx_desc,
1310 			      struct sk_buff *skb)
1311 {
1312 	__le32 rsc_enabled;
1313 	u32 rsc_cnt;
1314 
1315 	if (!ring_is_rsc_enabled(rx_ring))
1316 		return;
1317 
1318 	rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1319 		      cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1320 
1321 	/* If this is an RSC frame rsc_cnt should be non-zero */
1322 	if (!rsc_enabled)
1323 		return;
1324 
1325 	rsc_cnt = le32_to_cpu(rsc_enabled);
1326 	rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1327 
1328 	IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1329 }
1330 
1331 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1332 				   struct sk_buff *skb)
1333 {
1334 	u16 hdr_len = skb_headlen(skb);
1335 
1336 	/* set gso_size to avoid messing up TCP MSS */
1337 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1338 						 IXGBE_CB(skb)->append_cnt);
1339 }
1340 
1341 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1342 				   struct sk_buff *skb)
1343 {
1344 	/* if append_cnt is 0 then frame is not RSC */
1345 	if (!IXGBE_CB(skb)->append_cnt)
1346 		return;
1347 
1348 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1349 	rx_ring->rx_stats.rsc_flush++;
1350 
1351 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1352 
1353 	/* gso_size is computed using append_cnt so always clear it last */
1354 	IXGBE_CB(skb)->append_cnt = 0;
1355 }
1356 
1357 /**
1358  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1359  * @rx_ring: rx descriptor ring packet is being transacted on
1360  * @rx_desc: pointer to the EOP Rx descriptor
1361  * @skb: pointer to current skb being populated
1362  *
1363  * This function checks the ring, descriptor, and packet information in
1364  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1365  * other fields within the skb.
1366  **/
1367 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1368 				     union ixgbe_adv_rx_desc *rx_desc,
1369 				     struct sk_buff *skb)
1370 {
1371 	ixgbe_update_rsc_stats(rx_ring, skb);
1372 
1373 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1374 
1375 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1376 
1377 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1378 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1379 		__vlan_hwaccel_put_tag(skb, vid);
1380 	}
1381 
1382 	skb_record_rx_queue(skb, rx_ring->queue_index);
1383 
1384 	skb->protocol = eth_type_trans(skb, rx_ring->netdev);
1385 }
1386 
1387 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1388 			 struct sk_buff *skb)
1389 {
1390 	struct ixgbe_adapter *adapter = q_vector->adapter;
1391 
1392 	if (!(adapter->flags & IXGBE_FLAG_IN_NETPOLL))
1393 		napi_gro_receive(&q_vector->napi, skb);
1394 	else
1395 		netif_rx(skb);
1396 }
1397 
1398 /**
1399  * ixgbe_is_non_eop - process handling of non-EOP buffers
1400  * @rx_ring: Rx ring being processed
1401  * @rx_desc: Rx descriptor for current buffer
1402  * @skb: Current socket buffer containing buffer in progress
1403  *
1404  * This function updates next to clean.  If the buffer is an EOP buffer
1405  * this function exits returning false, otherwise it will place the
1406  * sk_buff in the next buffer to be chained and return true indicating
1407  * that this is in fact a non-EOP buffer.
1408  **/
1409 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1410 			     union ixgbe_adv_rx_desc *rx_desc,
1411 			     struct sk_buff *skb)
1412 {
1413 	u32 ntc = rx_ring->next_to_clean + 1;
1414 
1415 	/* fetch, update, and store next to clean */
1416 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1417 	rx_ring->next_to_clean = ntc;
1418 
1419 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1420 
1421 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1422 		return false;
1423 
1424 	/* append_cnt indicates packet is RSC, if so fetch nextp */
1425 	if (IXGBE_CB(skb)->append_cnt) {
1426 		ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1427 		ntc &= IXGBE_RXDADV_NEXTP_MASK;
1428 		ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1429 	}
1430 
1431 	/* place skb in next buffer to be received */
1432 	rx_ring->rx_buffer_info[ntc].skb = skb;
1433 	rx_ring->rx_stats.non_eop_descs++;
1434 
1435 	return true;
1436 }
1437 
1438 /**
1439  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1440  * @rx_ring: rx descriptor ring packet is being transacted on
1441  * @rx_desc: pointer to the EOP Rx descriptor
1442  * @skb: pointer to current skb being fixed
1443  *
1444  * Check for corrupted packet headers caused by senders on the local L2
1445  * embedded NIC switch not setting up their Tx Descriptors right.  These
1446  * should be very rare.
1447  *
1448  * Also address the case where we are pulling data in on pages only
1449  * and as such no data is present in the skb header.
1450  *
1451  * In addition if skb is not at least 60 bytes we need to pad it so that
1452  * it is large enough to qualify as a valid Ethernet frame.
1453  *
1454  * Returns true if an error was encountered and skb was freed.
1455  **/
1456 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1457 				  union ixgbe_adv_rx_desc *rx_desc,
1458 				  struct sk_buff *skb)
1459 {
1460 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1461 	struct net_device *netdev = rx_ring->netdev;
1462 	unsigned char *va;
1463 	unsigned int pull_len;
1464 
1465 	/* if the page was released unmap it, else just sync our portion */
1466 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1467 		dma_unmap_page(rx_ring->dev, IXGBE_CB(skb)->dma,
1468 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1469 		IXGBE_CB(skb)->page_released = false;
1470 	} else {
1471 		dma_sync_single_range_for_cpu(rx_ring->dev,
1472 					      IXGBE_CB(skb)->dma,
1473 					      frag->page_offset,
1474 					      ixgbe_rx_bufsz(rx_ring),
1475 					      DMA_FROM_DEVICE);
1476 	}
1477 	IXGBE_CB(skb)->dma = 0;
1478 
1479 	/* verify that the packet does not have any known errors */
1480 	if (unlikely(ixgbe_test_staterr(rx_desc,
1481 					IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1482 	    !(netdev->features & NETIF_F_RXALL))) {
1483 		dev_kfree_skb_any(skb);
1484 		return true;
1485 	}
1486 
1487 	/*
1488 	 * it is valid to use page_address instead of kmap since we are
1489 	 * working with pages allocated out of the lomem pool per
1490 	 * alloc_page(GFP_ATOMIC)
1491 	 */
1492 	va = skb_frag_address(frag);
1493 
1494 	/*
1495 	 * we need the header to contain the greater of either ETH_HLEN or
1496 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1497 	 */
1498 	pull_len = skb_frag_size(frag);
1499 	if (pull_len > 256)
1500 		pull_len = ixgbe_get_headlen(va, pull_len);
1501 
1502 	/* align pull length to size of long to optimize memcpy performance */
1503 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1504 
1505 	/* update all of the pointers */
1506 	skb_frag_size_sub(frag, pull_len);
1507 	frag->page_offset += pull_len;
1508 	skb->data_len -= pull_len;
1509 	skb->tail += pull_len;
1510 
1511 	/*
1512 	 * if we sucked the frag empty then we should free it,
1513 	 * if there are other frags here something is screwed up in hardware
1514 	 */
1515 	if (skb_frag_size(frag) == 0) {
1516 		BUG_ON(skb_shinfo(skb)->nr_frags != 1);
1517 		skb_shinfo(skb)->nr_frags = 0;
1518 		__skb_frag_unref(frag);
1519 		skb->truesize -= ixgbe_rx_bufsz(rx_ring);
1520 	}
1521 
1522 	/* if skb_pad returns an error the skb was freed */
1523 	if (unlikely(skb->len < 60)) {
1524 		int pad_len = 60 - skb->len;
1525 
1526 		if (skb_pad(skb, pad_len))
1527 			return true;
1528 		__skb_put(skb, pad_len);
1529 	}
1530 
1531 	return false;
1532 }
1533 
1534 /**
1535  * ixgbe_can_reuse_page - determine if we can reuse a page
1536  * @rx_buffer: pointer to rx_buffer containing the page we want to reuse
1537  *
1538  * Returns true if page can be reused in another Rx buffer
1539  **/
1540 static inline bool ixgbe_can_reuse_page(struct ixgbe_rx_buffer *rx_buffer)
1541 {
1542 	struct page *page = rx_buffer->page;
1543 
1544 	/* if we are only owner of page and it is local we can reuse it */
1545 	return likely(page_count(page) == 1) &&
1546 	       likely(page_to_nid(page) == numa_node_id());
1547 }
1548 
1549 /**
1550  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1551  * @rx_ring: rx descriptor ring to store buffers on
1552  * @old_buff: donor buffer to have page reused
1553  *
1554  * Syncronizes page for reuse by the adapter
1555  **/
1556 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1557 				struct ixgbe_rx_buffer *old_buff)
1558 {
1559 	struct ixgbe_rx_buffer *new_buff;
1560 	u16 nta = rx_ring->next_to_alloc;
1561 	u16 bufsz = ixgbe_rx_bufsz(rx_ring);
1562 
1563 	new_buff = &rx_ring->rx_buffer_info[nta];
1564 
1565 	/* update, and store next to alloc */
1566 	nta++;
1567 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1568 
1569 	/* transfer page from old buffer to new buffer */
1570 	new_buff->page = old_buff->page;
1571 	new_buff->dma = old_buff->dma;
1572 
1573 	/* flip page offset to other buffer and store to new_buff */
1574 	new_buff->page_offset = old_buff->page_offset ^ bufsz;
1575 
1576 	/* sync the buffer for use by the device */
1577 	dma_sync_single_range_for_device(rx_ring->dev, new_buff->dma,
1578 					 new_buff->page_offset, bufsz,
1579 					 DMA_FROM_DEVICE);
1580 
1581 	/* bump ref count on page before it is given to the stack */
1582 	get_page(new_buff->page);
1583 }
1584 
1585 /**
1586  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1587  * @rx_ring: rx descriptor ring to transact packets on
1588  * @rx_buffer: buffer containing page to add
1589  * @rx_desc: descriptor containing length of buffer written by hardware
1590  * @skb: sk_buff to place the data into
1591  *
1592  * This function is based on skb_add_rx_frag.  I would have used that
1593  * function however it doesn't handle the truesize case correctly since we
1594  * are allocating more memory than might be used for a single receive.
1595  **/
1596 static void ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1597 			      struct ixgbe_rx_buffer *rx_buffer,
1598 			      struct sk_buff *skb, int size)
1599 {
1600 	skb_fill_page_desc(skb, skb_shinfo(skb)->nr_frags,
1601 			   rx_buffer->page, rx_buffer->page_offset,
1602 			   size);
1603 	skb->len += size;
1604 	skb->data_len += size;
1605 	skb->truesize += ixgbe_rx_bufsz(rx_ring);
1606 }
1607 
1608 /**
1609  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
1610  * @q_vector: structure containing interrupt and ring information
1611  * @rx_ring: rx descriptor ring to transact packets on
1612  * @budget: Total limit on number of packets to process
1613  *
1614  * This function provides a "bounce buffer" approach to Rx interrupt
1615  * processing.  The advantage to this is that on systems that have
1616  * expensive overhead for IOMMU access this provides a means of avoiding
1617  * it by maintaining the mapping of the page to the syste.
1618  *
1619  * Returns true if all work is completed without reaching budget
1620  **/
1621 static bool ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
1622 			       struct ixgbe_ring *rx_ring,
1623 			       int budget)
1624 {
1625 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1626 #ifdef IXGBE_FCOE
1627 	struct ixgbe_adapter *adapter = q_vector->adapter;
1628 	int ddp_bytes = 0;
1629 #endif /* IXGBE_FCOE */
1630 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
1631 
1632 	do {
1633 		struct ixgbe_rx_buffer *rx_buffer;
1634 		union ixgbe_adv_rx_desc *rx_desc;
1635 		struct sk_buff *skb;
1636 		struct page *page;
1637 		u16 ntc;
1638 
1639 		/* return some buffers to hardware, one at a time is too slow */
1640 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
1641 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
1642 			cleaned_count = 0;
1643 		}
1644 
1645 		ntc = rx_ring->next_to_clean;
1646 		rx_desc = IXGBE_RX_DESC(rx_ring, ntc);
1647 		rx_buffer = &rx_ring->rx_buffer_info[ntc];
1648 
1649 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_DD))
1650 			break;
1651 
1652 		/*
1653 		 * This memory barrier is needed to keep us from reading
1654 		 * any other fields out of the rx_desc until we know the
1655 		 * RXD_STAT_DD bit is set
1656 		 */
1657 		rmb();
1658 
1659 		page = rx_buffer->page;
1660 		prefetchw(page);
1661 
1662 		skb = rx_buffer->skb;
1663 
1664 		if (likely(!skb)) {
1665 			void *page_addr = page_address(page) +
1666 					  rx_buffer->page_offset;
1667 
1668 			/* prefetch first cache line of first page */
1669 			prefetch(page_addr);
1670 #if L1_CACHE_BYTES < 128
1671 			prefetch(page_addr + L1_CACHE_BYTES);
1672 #endif
1673 
1674 			/* allocate a skb to store the frags */
1675 			skb = netdev_alloc_skb_ip_align(rx_ring->netdev,
1676 							IXGBE_RX_HDR_SIZE);
1677 			if (unlikely(!skb)) {
1678 				rx_ring->rx_stats.alloc_rx_buff_failed++;
1679 				break;
1680 			}
1681 
1682 			/*
1683 			 * we will be copying header into skb->data in
1684 			 * pskb_may_pull so it is in our interest to prefetch
1685 			 * it now to avoid a possible cache miss
1686 			 */
1687 			prefetchw(skb->data);
1688 
1689 			/*
1690 			 * Delay unmapping of the first packet. It carries the
1691 			 * header information, HW may still access the header
1692 			 * after the writeback.  Only unmap it when EOP is
1693 			 * reached
1694 			 */
1695 			IXGBE_CB(skb)->dma = rx_buffer->dma;
1696 		} else {
1697 			/* we are reusing so sync this buffer for CPU use */
1698 			dma_sync_single_range_for_cpu(rx_ring->dev,
1699 						      rx_buffer->dma,
1700 						      rx_buffer->page_offset,
1701 						      ixgbe_rx_bufsz(rx_ring),
1702 						      DMA_FROM_DEVICE);
1703 		}
1704 
1705 		/* pull page into skb */
1706 		ixgbe_add_rx_frag(rx_ring, rx_buffer, skb,
1707 				  le16_to_cpu(rx_desc->wb.upper.length));
1708 
1709 		if (ixgbe_can_reuse_page(rx_buffer)) {
1710 			/* hand second half of page back to the ring */
1711 			ixgbe_reuse_rx_page(rx_ring, rx_buffer);
1712 		} else if (IXGBE_CB(skb)->dma == rx_buffer->dma) {
1713 			/* the page has been released from the ring */
1714 			IXGBE_CB(skb)->page_released = true;
1715 		} else {
1716 			/* we are not reusing the buffer so unmap it */
1717 			dma_unmap_page(rx_ring->dev, rx_buffer->dma,
1718 				       ixgbe_rx_pg_size(rx_ring),
1719 				       DMA_FROM_DEVICE);
1720 		}
1721 
1722 		/* clear contents of buffer_info */
1723 		rx_buffer->skb = NULL;
1724 		rx_buffer->dma = 0;
1725 		rx_buffer->page = NULL;
1726 
1727 		ixgbe_get_rsc_cnt(rx_ring, rx_desc, skb);
1728 
1729 		cleaned_count++;
1730 
1731 		/* place incomplete frames back on ring for completion */
1732 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
1733 			continue;
1734 
1735 		/* verify the packet layout is correct */
1736 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
1737 			continue;
1738 
1739 		/* probably a little skewed due to removing CRC */
1740 		total_rx_bytes += skb->len;
1741 		total_rx_packets++;
1742 
1743 		/* populate checksum, timestamp, VLAN, and protocol */
1744 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
1745 
1746 #ifdef IXGBE_FCOE
1747 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
1748 		if (ixgbe_rx_is_fcoe(adapter, rx_desc)) {
1749 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
1750 			if (!ddp_bytes) {
1751 				dev_kfree_skb_any(skb);
1752 				continue;
1753 			}
1754 		}
1755 
1756 #endif /* IXGBE_FCOE */
1757 		ixgbe_rx_skb(q_vector, skb);
1758 
1759 		/* update budget accounting */
1760 		budget--;
1761 	} while (likely(budget));
1762 
1763 #ifdef IXGBE_FCOE
1764 	/* include DDPed FCoE data */
1765 	if (ddp_bytes > 0) {
1766 		unsigned int mss;
1767 
1768 		mss = rx_ring->netdev->mtu - sizeof(struct fcoe_hdr) -
1769 			sizeof(struct fc_frame_header) -
1770 			sizeof(struct fcoe_crc_eof);
1771 		if (mss > 512)
1772 			mss &= ~511;
1773 		total_rx_bytes += ddp_bytes;
1774 		total_rx_packets += DIV_ROUND_UP(ddp_bytes, mss);
1775 	}
1776 
1777 #endif /* IXGBE_FCOE */
1778 	u64_stats_update_begin(&rx_ring->syncp);
1779 	rx_ring->stats.packets += total_rx_packets;
1780 	rx_ring->stats.bytes += total_rx_bytes;
1781 	u64_stats_update_end(&rx_ring->syncp);
1782 	q_vector->rx.total_packets += total_rx_packets;
1783 	q_vector->rx.total_bytes += total_rx_bytes;
1784 
1785 	if (cleaned_count)
1786 		ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
1787 
1788 	return !!budget;
1789 }
1790 
1791 /**
1792  * ixgbe_configure_msix - Configure MSI-X hardware
1793  * @adapter: board private structure
1794  *
1795  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
1796  * interrupts.
1797  **/
1798 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
1799 {
1800 	struct ixgbe_q_vector *q_vector;
1801 	int q_vectors, v_idx;
1802 	u32 mask;
1803 
1804 	q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
1805 
1806 	/* Populate MSIX to EITR Select */
1807 	if (adapter->num_vfs > 32) {
1808 		u32 eitrsel = (1 << (adapter->num_vfs - 32)) - 1;
1809 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
1810 	}
1811 
1812 	/*
1813 	 * Populate the IVAR table and set the ITR values to the
1814 	 * corresponding register.
1815 	 */
1816 	for (v_idx = 0; v_idx < q_vectors; v_idx++) {
1817 		struct ixgbe_ring *ring;
1818 		q_vector = adapter->q_vector[v_idx];
1819 
1820 		ixgbe_for_each_ring(ring, q_vector->rx)
1821 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
1822 
1823 		ixgbe_for_each_ring(ring, q_vector->tx)
1824 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
1825 
1826 		if (q_vector->tx.ring && !q_vector->rx.ring) {
1827 			/* tx only vector */
1828 			if (adapter->tx_itr_setting == 1)
1829 				q_vector->itr = IXGBE_10K_ITR;
1830 			else
1831 				q_vector->itr = adapter->tx_itr_setting;
1832 		} else {
1833 			/* rx or rx/tx vector */
1834 			if (adapter->rx_itr_setting == 1)
1835 				q_vector->itr = IXGBE_20K_ITR;
1836 			else
1837 				q_vector->itr = adapter->rx_itr_setting;
1838 		}
1839 
1840 		ixgbe_write_eitr(q_vector);
1841 	}
1842 
1843 	switch (adapter->hw.mac.type) {
1844 	case ixgbe_mac_82598EB:
1845 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
1846 			       v_idx);
1847 		break;
1848 	case ixgbe_mac_82599EB:
1849 	case ixgbe_mac_X540:
1850 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
1851 		break;
1852 	default:
1853 		break;
1854 	}
1855 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
1856 
1857 	/* set up to autoclear timer, and the vectors */
1858 	mask = IXGBE_EIMS_ENABLE_MASK;
1859 	mask &= ~(IXGBE_EIMS_OTHER |
1860 		  IXGBE_EIMS_MAILBOX |
1861 		  IXGBE_EIMS_LSC);
1862 
1863 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
1864 }
1865 
1866 enum latency_range {
1867 	lowest_latency = 0,
1868 	low_latency = 1,
1869 	bulk_latency = 2,
1870 	latency_invalid = 255
1871 };
1872 
1873 /**
1874  * ixgbe_update_itr - update the dynamic ITR value based on statistics
1875  * @q_vector: structure containing interrupt and ring information
1876  * @ring_container: structure containing ring performance data
1877  *
1878  *      Stores a new ITR value based on packets and byte
1879  *      counts during the last interrupt.  The advantage of per interrupt
1880  *      computation is faster updates and more accurate ITR for the current
1881  *      traffic pattern.  Constants in this function were computed
1882  *      based on theoretical maximum wire speed and thresholds were set based
1883  *      on testing data as well as attempting to minimize response time
1884  *      while increasing bulk throughput.
1885  *      this functionality is controlled by the InterruptThrottleRate module
1886  *      parameter (see ixgbe_param.c)
1887  **/
1888 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
1889 			     struct ixgbe_ring_container *ring_container)
1890 {
1891 	int bytes = ring_container->total_bytes;
1892 	int packets = ring_container->total_packets;
1893 	u32 timepassed_us;
1894 	u64 bytes_perint;
1895 	u8 itr_setting = ring_container->itr;
1896 
1897 	if (packets == 0)
1898 		return;
1899 
1900 	/* simple throttlerate management
1901 	 *   0-10MB/s   lowest (100000 ints/s)
1902 	 *  10-20MB/s   low    (20000 ints/s)
1903 	 *  20-1249MB/s bulk   (8000 ints/s)
1904 	 */
1905 	/* what was last interrupt timeslice? */
1906 	timepassed_us = q_vector->itr >> 2;
1907 	bytes_perint = bytes / timepassed_us; /* bytes/usec */
1908 
1909 	switch (itr_setting) {
1910 	case lowest_latency:
1911 		if (bytes_perint > 10)
1912 			itr_setting = low_latency;
1913 		break;
1914 	case low_latency:
1915 		if (bytes_perint > 20)
1916 			itr_setting = bulk_latency;
1917 		else if (bytes_perint <= 10)
1918 			itr_setting = lowest_latency;
1919 		break;
1920 	case bulk_latency:
1921 		if (bytes_perint <= 20)
1922 			itr_setting = low_latency;
1923 		break;
1924 	}
1925 
1926 	/* clear work counters since we have the values we need */
1927 	ring_container->total_bytes = 0;
1928 	ring_container->total_packets = 0;
1929 
1930 	/* write updated itr to ring container */
1931 	ring_container->itr = itr_setting;
1932 }
1933 
1934 /**
1935  * ixgbe_write_eitr - write EITR register in hardware specific way
1936  * @q_vector: structure containing interrupt and ring information
1937  *
1938  * This function is made to be called by ethtool and by the driver
1939  * when it needs to update EITR registers at runtime.  Hardware
1940  * specific quirks/differences are taken care of here.
1941  */
1942 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
1943 {
1944 	struct ixgbe_adapter *adapter = q_vector->adapter;
1945 	struct ixgbe_hw *hw = &adapter->hw;
1946 	int v_idx = q_vector->v_idx;
1947 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
1948 
1949 	switch (adapter->hw.mac.type) {
1950 	case ixgbe_mac_82598EB:
1951 		/* must write high and low 16 bits to reset counter */
1952 		itr_reg |= (itr_reg << 16);
1953 		break;
1954 	case ixgbe_mac_82599EB:
1955 	case ixgbe_mac_X540:
1956 		/*
1957 		 * set the WDIS bit to not clear the timer bits and cause an
1958 		 * immediate assertion of the interrupt
1959 		 */
1960 		itr_reg |= IXGBE_EITR_CNT_WDIS;
1961 		break;
1962 	default:
1963 		break;
1964 	}
1965 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
1966 }
1967 
1968 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
1969 {
1970 	u32 new_itr = q_vector->itr;
1971 	u8 current_itr;
1972 
1973 	ixgbe_update_itr(q_vector, &q_vector->tx);
1974 	ixgbe_update_itr(q_vector, &q_vector->rx);
1975 
1976 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
1977 
1978 	switch (current_itr) {
1979 	/* counts and packets in update_itr are dependent on these numbers */
1980 	case lowest_latency:
1981 		new_itr = IXGBE_100K_ITR;
1982 		break;
1983 	case low_latency:
1984 		new_itr = IXGBE_20K_ITR;
1985 		break;
1986 	case bulk_latency:
1987 		new_itr = IXGBE_8K_ITR;
1988 		break;
1989 	default:
1990 		break;
1991 	}
1992 
1993 	if (new_itr != q_vector->itr) {
1994 		/* do an exponential smoothing */
1995 		new_itr = (10 * new_itr * q_vector->itr) /
1996 			  ((9 * new_itr) + q_vector->itr);
1997 
1998 		/* save the algorithm value here */
1999 		q_vector->itr = new_itr;
2000 
2001 		ixgbe_write_eitr(q_vector);
2002 	}
2003 }
2004 
2005 /**
2006  * ixgbe_check_overtemp_subtask - check for over temperature
2007  * @adapter: pointer to adapter
2008  **/
2009 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2010 {
2011 	struct ixgbe_hw *hw = &adapter->hw;
2012 	u32 eicr = adapter->interrupt_event;
2013 
2014 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2015 		return;
2016 
2017 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) &&
2018 	    !(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2019 		return;
2020 
2021 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2022 
2023 	switch (hw->device_id) {
2024 	case IXGBE_DEV_ID_82599_T3_LOM:
2025 		/*
2026 		 * Since the warning interrupt is for both ports
2027 		 * we don't have to check if:
2028 		 *  - This interrupt wasn't for our port.
2029 		 *  - We may have missed the interrupt so always have to
2030 		 *    check if we  got a LSC
2031 		 */
2032 		if (!(eicr & IXGBE_EICR_GPI_SDP0) &&
2033 		    !(eicr & IXGBE_EICR_LSC))
2034 			return;
2035 
2036 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2037 			u32 autoneg;
2038 			bool link_up = false;
2039 
2040 			hw->mac.ops.check_link(hw, &autoneg, &link_up, false);
2041 
2042 			if (link_up)
2043 				return;
2044 		}
2045 
2046 		/* Check if this is not due to overtemp */
2047 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2048 			return;
2049 
2050 		break;
2051 	default:
2052 		if (!(eicr & IXGBE_EICR_GPI_SDP0))
2053 			return;
2054 		break;
2055 	}
2056 	e_crit(drv,
2057 	       "Network adapter has been stopped because it has over heated. "
2058 	       "Restart the computer. If the problem persists, "
2059 	       "power off the system and replace the adapter\n");
2060 
2061 	adapter->interrupt_event = 0;
2062 }
2063 
2064 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2065 {
2066 	struct ixgbe_hw *hw = &adapter->hw;
2067 
2068 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2069 	    (eicr & IXGBE_EICR_GPI_SDP1)) {
2070 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2071 		/* write to clear the interrupt */
2072 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2073 	}
2074 }
2075 
2076 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2077 {
2078 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2079 		return;
2080 
2081 	switch (adapter->hw.mac.type) {
2082 	case ixgbe_mac_82599EB:
2083 		/*
2084 		 * Need to check link state so complete overtemp check
2085 		 * on service task
2086 		 */
2087 		if (((eicr & IXGBE_EICR_GPI_SDP0) || (eicr & IXGBE_EICR_LSC)) &&
2088 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2089 			adapter->interrupt_event = eicr;
2090 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2091 			ixgbe_service_event_schedule(adapter);
2092 			return;
2093 		}
2094 		return;
2095 	case ixgbe_mac_X540:
2096 		if (!(eicr & IXGBE_EICR_TS))
2097 			return;
2098 		break;
2099 	default:
2100 		return;
2101 	}
2102 
2103 	e_crit(drv,
2104 	       "Network adapter has been stopped because it has over heated. "
2105 	       "Restart the computer. If the problem persists, "
2106 	       "power off the system and replace the adapter\n");
2107 }
2108 
2109 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2110 {
2111 	struct ixgbe_hw *hw = &adapter->hw;
2112 
2113 	if (eicr & IXGBE_EICR_GPI_SDP2) {
2114 		/* Clear the interrupt */
2115 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP2);
2116 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2117 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2118 			ixgbe_service_event_schedule(adapter);
2119 		}
2120 	}
2121 
2122 	if (eicr & IXGBE_EICR_GPI_SDP1) {
2123 		/* Clear the interrupt */
2124 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2125 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2126 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2127 			ixgbe_service_event_schedule(adapter);
2128 		}
2129 	}
2130 }
2131 
2132 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2133 {
2134 	struct ixgbe_hw *hw = &adapter->hw;
2135 
2136 	adapter->lsc_int++;
2137 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2138 	adapter->link_check_timeout = jiffies;
2139 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2140 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2141 		IXGBE_WRITE_FLUSH(hw);
2142 		ixgbe_service_event_schedule(adapter);
2143 	}
2144 }
2145 
2146 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2147 					   u64 qmask)
2148 {
2149 	u32 mask;
2150 	struct ixgbe_hw *hw = &adapter->hw;
2151 
2152 	switch (hw->mac.type) {
2153 	case ixgbe_mac_82598EB:
2154 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2155 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2156 		break;
2157 	case ixgbe_mac_82599EB:
2158 	case ixgbe_mac_X540:
2159 		mask = (qmask & 0xFFFFFFFF);
2160 		if (mask)
2161 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2162 		mask = (qmask >> 32);
2163 		if (mask)
2164 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2165 		break;
2166 	default:
2167 		break;
2168 	}
2169 	/* skip the flush */
2170 }
2171 
2172 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2173 					    u64 qmask)
2174 {
2175 	u32 mask;
2176 	struct ixgbe_hw *hw = &adapter->hw;
2177 
2178 	switch (hw->mac.type) {
2179 	case ixgbe_mac_82598EB:
2180 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2181 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2182 		break;
2183 	case ixgbe_mac_82599EB:
2184 	case ixgbe_mac_X540:
2185 		mask = (qmask & 0xFFFFFFFF);
2186 		if (mask)
2187 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2188 		mask = (qmask >> 32);
2189 		if (mask)
2190 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2191 		break;
2192 	default:
2193 		break;
2194 	}
2195 	/* skip the flush */
2196 }
2197 
2198 /**
2199  * ixgbe_irq_enable - Enable default interrupt generation settings
2200  * @adapter: board private structure
2201  **/
2202 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2203 				    bool flush)
2204 {
2205 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2206 
2207 	/* don't reenable LSC while waiting for link */
2208 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2209 		mask &= ~IXGBE_EIMS_LSC;
2210 
2211 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2212 		switch (adapter->hw.mac.type) {
2213 		case ixgbe_mac_82599EB:
2214 			mask |= IXGBE_EIMS_GPI_SDP0;
2215 			break;
2216 		case ixgbe_mac_X540:
2217 			mask |= IXGBE_EIMS_TS;
2218 			break;
2219 		default:
2220 			break;
2221 		}
2222 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2223 		mask |= IXGBE_EIMS_GPI_SDP1;
2224 	switch (adapter->hw.mac.type) {
2225 	case ixgbe_mac_82599EB:
2226 		mask |= IXGBE_EIMS_GPI_SDP1;
2227 		mask |= IXGBE_EIMS_GPI_SDP2;
2228 	case ixgbe_mac_X540:
2229 		mask |= IXGBE_EIMS_ECC;
2230 		mask |= IXGBE_EIMS_MAILBOX;
2231 		break;
2232 	default:
2233 		break;
2234 	}
2235 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2236 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
2237 		mask |= IXGBE_EIMS_FLOW_DIR;
2238 
2239 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
2240 	if (queues)
2241 		ixgbe_irq_enable_queues(adapter, ~0);
2242 	if (flush)
2243 		IXGBE_WRITE_FLUSH(&adapter->hw);
2244 }
2245 
2246 static irqreturn_t ixgbe_msix_other(int irq, void *data)
2247 {
2248 	struct ixgbe_adapter *adapter = data;
2249 	struct ixgbe_hw *hw = &adapter->hw;
2250 	u32 eicr;
2251 
2252 	/*
2253 	 * Workaround for Silicon errata.  Use clear-by-write instead
2254 	 * of clear-by-read.  Reading with EICS will return the
2255 	 * interrupt causes without clearing, which later be done
2256 	 * with the write to EICR.
2257 	 */
2258 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2259 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2260 
2261 	if (eicr & IXGBE_EICR_LSC)
2262 		ixgbe_check_lsc(adapter);
2263 
2264 	if (eicr & IXGBE_EICR_MAILBOX)
2265 		ixgbe_msg_task(adapter);
2266 
2267 	switch (hw->mac.type) {
2268 	case ixgbe_mac_82599EB:
2269 	case ixgbe_mac_X540:
2270 		if (eicr & IXGBE_EICR_ECC)
2271 			e_info(link, "Received unrecoverable ECC Err, please "
2272 			       "reboot\n");
2273 		/* Handle Flow Director Full threshold interrupt */
2274 		if (eicr & IXGBE_EICR_FLOW_DIR) {
2275 			int reinit_count = 0;
2276 			int i;
2277 			for (i = 0; i < adapter->num_tx_queues; i++) {
2278 				struct ixgbe_ring *ring = adapter->tx_ring[i];
2279 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
2280 						       &ring->state))
2281 					reinit_count++;
2282 			}
2283 			if (reinit_count) {
2284 				/* no more flow director interrupts until after init */
2285 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
2286 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
2287 				ixgbe_service_event_schedule(adapter);
2288 			}
2289 		}
2290 		ixgbe_check_sfp_event(adapter, eicr);
2291 		ixgbe_check_overtemp_event(adapter, eicr);
2292 		break;
2293 	default:
2294 		break;
2295 	}
2296 
2297 	ixgbe_check_fan_failure(adapter, eicr);
2298 
2299 	/* re-enable the original interrupt state, no lsc, no queues */
2300 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2301 		ixgbe_irq_enable(adapter, false, false);
2302 
2303 	return IRQ_HANDLED;
2304 }
2305 
2306 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
2307 {
2308 	struct ixgbe_q_vector *q_vector = data;
2309 
2310 	/* EIAM disabled interrupts (on this vector) for us */
2311 
2312 	if (q_vector->rx.ring || q_vector->tx.ring)
2313 		napi_schedule(&q_vector->napi);
2314 
2315 	return IRQ_HANDLED;
2316 }
2317 
2318 /**
2319  * ixgbe_poll - NAPI Rx polling callback
2320  * @napi: structure for representing this polling device
2321  * @budget: how many packets driver is allowed to clean
2322  *
2323  * This function is used for legacy and MSI, NAPI mode
2324  **/
2325 int ixgbe_poll(struct napi_struct *napi, int budget)
2326 {
2327 	struct ixgbe_q_vector *q_vector =
2328 				container_of(napi, struct ixgbe_q_vector, napi);
2329 	struct ixgbe_adapter *adapter = q_vector->adapter;
2330 	struct ixgbe_ring *ring;
2331 	int per_ring_budget;
2332 	bool clean_complete = true;
2333 
2334 #ifdef CONFIG_IXGBE_DCA
2335 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
2336 		ixgbe_update_dca(q_vector);
2337 #endif
2338 
2339 	ixgbe_for_each_ring(ring, q_vector->tx)
2340 		clean_complete &= !!ixgbe_clean_tx_irq(q_vector, ring);
2341 
2342 	/* attempt to distribute budget to each queue fairly, but don't allow
2343 	 * the budget to go below 1 because we'll exit polling */
2344 	if (q_vector->rx.count > 1)
2345 		per_ring_budget = max(budget/q_vector->rx.count, 1);
2346 	else
2347 		per_ring_budget = budget;
2348 
2349 	ixgbe_for_each_ring(ring, q_vector->rx)
2350 		clean_complete &= ixgbe_clean_rx_irq(q_vector, ring,
2351 						     per_ring_budget);
2352 
2353 	/* If all work not completed, return budget and keep polling */
2354 	if (!clean_complete)
2355 		return budget;
2356 
2357 	/* all work done, exit the polling mode */
2358 	napi_complete(napi);
2359 	if (adapter->rx_itr_setting & 1)
2360 		ixgbe_set_itr(q_vector);
2361 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2362 		ixgbe_irq_enable_queues(adapter, ((u64)1 << q_vector->v_idx));
2363 
2364 	return 0;
2365 }
2366 
2367 /**
2368  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2369  * @adapter: board private structure
2370  *
2371  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2372  * interrupts from the kernel.
2373  **/
2374 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
2375 {
2376 	struct net_device *netdev = adapter->netdev;
2377 	int q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
2378 	int vector, err;
2379 	int ri = 0, ti = 0;
2380 
2381 	for (vector = 0; vector < q_vectors; vector++) {
2382 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2383 		struct msix_entry *entry = &adapter->msix_entries[vector];
2384 
2385 		if (q_vector->tx.ring && q_vector->rx.ring) {
2386 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2387 				 "%s-%s-%d", netdev->name, "TxRx", ri++);
2388 			ti++;
2389 		} else if (q_vector->rx.ring) {
2390 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2391 				 "%s-%s-%d", netdev->name, "rx", ri++);
2392 		} else if (q_vector->tx.ring) {
2393 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2394 				 "%s-%s-%d", netdev->name, "tx", ti++);
2395 		} else {
2396 			/* skip this unused q_vector */
2397 			continue;
2398 		}
2399 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
2400 				  q_vector->name, q_vector);
2401 		if (err) {
2402 			e_err(probe, "request_irq failed for MSIX interrupt "
2403 			      "Error: %d\n", err);
2404 			goto free_queue_irqs;
2405 		}
2406 		/* If Flow Director is enabled, set interrupt affinity */
2407 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2408 			/* assign the mask for this irq */
2409 			irq_set_affinity_hint(entry->vector,
2410 					      &q_vector->affinity_mask);
2411 		}
2412 	}
2413 
2414 	err = request_irq(adapter->msix_entries[vector].vector,
2415 			  ixgbe_msix_other, 0, netdev->name, adapter);
2416 	if (err) {
2417 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
2418 		goto free_queue_irqs;
2419 	}
2420 
2421 	return 0;
2422 
2423 free_queue_irqs:
2424 	while (vector) {
2425 		vector--;
2426 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
2427 				      NULL);
2428 		free_irq(adapter->msix_entries[vector].vector,
2429 			 adapter->q_vector[vector]);
2430 	}
2431 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
2432 	pci_disable_msix(adapter->pdev);
2433 	kfree(adapter->msix_entries);
2434 	adapter->msix_entries = NULL;
2435 	return err;
2436 }
2437 
2438 /**
2439  * ixgbe_intr - legacy mode Interrupt Handler
2440  * @irq: interrupt number
2441  * @data: pointer to a network interface device structure
2442  **/
2443 static irqreturn_t ixgbe_intr(int irq, void *data)
2444 {
2445 	struct ixgbe_adapter *adapter = data;
2446 	struct ixgbe_hw *hw = &adapter->hw;
2447 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2448 	u32 eicr;
2449 
2450 	/*
2451 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
2452 	 * before the read of EICR.
2453 	 */
2454 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
2455 
2456 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2457 	 * therefore no explicit interrupt disable is necessary */
2458 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2459 	if (!eicr) {
2460 		/*
2461 		 * shared interrupt alert!
2462 		 * make sure interrupts are enabled because the read will
2463 		 * have disabled interrupts due to EIAM
2464 		 * finish the workaround of silicon errata on 82598.  Unmask
2465 		 * the interrupt that we masked before the EICR read.
2466 		 */
2467 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
2468 			ixgbe_irq_enable(adapter, true, true);
2469 		return IRQ_NONE;	/* Not our interrupt */
2470 	}
2471 
2472 	if (eicr & IXGBE_EICR_LSC)
2473 		ixgbe_check_lsc(adapter);
2474 
2475 	switch (hw->mac.type) {
2476 	case ixgbe_mac_82599EB:
2477 		ixgbe_check_sfp_event(adapter, eicr);
2478 		/* Fall through */
2479 	case ixgbe_mac_X540:
2480 		if (eicr & IXGBE_EICR_ECC)
2481 			e_info(link, "Received unrecoverable ECC err, please "
2482 				     "reboot\n");
2483 		ixgbe_check_overtemp_event(adapter, eicr);
2484 		break;
2485 	default:
2486 		break;
2487 	}
2488 
2489 	ixgbe_check_fan_failure(adapter, eicr);
2490 
2491 	/* would disable interrupts here but EIAM disabled it */
2492 	napi_schedule(&q_vector->napi);
2493 
2494 	/*
2495 	 * re-enable link(maybe) and non-queue interrupts, no flush.
2496 	 * ixgbe_poll will re-enable the queue interrupts
2497 	 */
2498 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2499 		ixgbe_irq_enable(adapter, false, false);
2500 
2501 	return IRQ_HANDLED;
2502 }
2503 
2504 /**
2505  * ixgbe_request_irq - initialize interrupts
2506  * @adapter: board private structure
2507  *
2508  * Attempts to configure interrupts using the best available
2509  * capabilities of the hardware and kernel.
2510  **/
2511 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
2512 {
2513 	struct net_device *netdev = adapter->netdev;
2514 	int err;
2515 
2516 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
2517 		err = ixgbe_request_msix_irqs(adapter);
2518 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
2519 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
2520 				  netdev->name, adapter);
2521 	else
2522 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
2523 				  netdev->name, adapter);
2524 
2525 	if (err)
2526 		e_err(probe, "request_irq failed, Error %d\n", err);
2527 
2528 	return err;
2529 }
2530 
2531 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
2532 {
2533 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
2534 		int i, q_vectors;
2535 
2536 		q_vectors = adapter->num_msix_vectors;
2537 		i = q_vectors - 1;
2538 		free_irq(adapter->msix_entries[i].vector, adapter);
2539 		i--;
2540 
2541 		for (; i >= 0; i--) {
2542 			/* free only the irqs that were actually requested */
2543 			if (!adapter->q_vector[i]->rx.ring &&
2544 			    !adapter->q_vector[i]->tx.ring)
2545 				continue;
2546 
2547 			/* clear the affinity_mask in the IRQ descriptor */
2548 			irq_set_affinity_hint(adapter->msix_entries[i].vector,
2549 					      NULL);
2550 
2551 			free_irq(adapter->msix_entries[i].vector,
2552 				 adapter->q_vector[i]);
2553 		}
2554 	} else {
2555 		free_irq(adapter->pdev->irq, adapter);
2556 	}
2557 }
2558 
2559 /**
2560  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
2561  * @adapter: board private structure
2562  **/
2563 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
2564 {
2565 	switch (adapter->hw.mac.type) {
2566 	case ixgbe_mac_82598EB:
2567 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
2568 		break;
2569 	case ixgbe_mac_82599EB:
2570 	case ixgbe_mac_X540:
2571 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
2572 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
2573 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
2574 		break;
2575 	default:
2576 		break;
2577 	}
2578 	IXGBE_WRITE_FLUSH(&adapter->hw);
2579 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
2580 		int i;
2581 		for (i = 0; i < adapter->num_msix_vectors; i++)
2582 			synchronize_irq(adapter->msix_entries[i].vector);
2583 	} else {
2584 		synchronize_irq(adapter->pdev->irq);
2585 	}
2586 }
2587 
2588 /**
2589  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
2590  *
2591  **/
2592 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
2593 {
2594 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2595 
2596 	/* rx/tx vector */
2597 	if (adapter->rx_itr_setting == 1)
2598 		q_vector->itr = IXGBE_20K_ITR;
2599 	else
2600 		q_vector->itr = adapter->rx_itr_setting;
2601 
2602 	ixgbe_write_eitr(q_vector);
2603 
2604 	ixgbe_set_ivar(adapter, 0, 0, 0);
2605 	ixgbe_set_ivar(adapter, 1, 0, 0);
2606 
2607 	e_info(hw, "Legacy interrupt IVAR setup done\n");
2608 }
2609 
2610 /**
2611  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
2612  * @adapter: board private structure
2613  * @ring: structure containing ring specific data
2614  *
2615  * Configure the Tx descriptor ring after a reset.
2616  **/
2617 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
2618 			     struct ixgbe_ring *ring)
2619 {
2620 	struct ixgbe_hw *hw = &adapter->hw;
2621 	u64 tdba = ring->dma;
2622 	int wait_loop = 10;
2623 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
2624 	u8 reg_idx = ring->reg_idx;
2625 
2626 	/* disable queue to avoid issues while updating state */
2627 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
2628 	IXGBE_WRITE_FLUSH(hw);
2629 
2630 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
2631 			(tdba & DMA_BIT_MASK(32)));
2632 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
2633 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
2634 			ring->count * sizeof(union ixgbe_adv_tx_desc));
2635 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
2636 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
2637 	ring->tail = hw->hw_addr + IXGBE_TDT(reg_idx);
2638 
2639 	/*
2640 	 * set WTHRESH to encourage burst writeback, it should not be set
2641 	 * higher than 1 when ITR is 0 as it could cause false TX hangs
2642 	 *
2643 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
2644 	 * to or less than the number of on chip descriptors, which is
2645 	 * currently 40.
2646 	 */
2647 	if (!ring->q_vector || (ring->q_vector->itr < 8))
2648 		txdctl |= (1 << 16);	/* WTHRESH = 1 */
2649 	else
2650 		txdctl |= (8 << 16);	/* WTHRESH = 8 */
2651 
2652 	/*
2653 	 * Setting PTHRESH to 32 both improves performance
2654 	 * and avoids a TX hang with DFP enabled
2655 	 */
2656 	txdctl |= (1 << 8) |	/* HTHRESH = 1 */
2657 		   32;		/* PTHRESH = 32 */
2658 
2659 	/* reinitialize flowdirector state */
2660 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2661 	    adapter->atr_sample_rate) {
2662 		ring->atr_sample_rate = adapter->atr_sample_rate;
2663 		ring->atr_count = 0;
2664 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
2665 	} else {
2666 		ring->atr_sample_rate = 0;
2667 	}
2668 
2669 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
2670 
2671 	/* enable queue */
2672 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
2673 
2674 	netdev_tx_reset_queue(txring_txq(ring));
2675 
2676 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
2677 	if (hw->mac.type == ixgbe_mac_82598EB &&
2678 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
2679 		return;
2680 
2681 	/* poll to verify queue is enabled */
2682 	do {
2683 		usleep_range(1000, 2000);
2684 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
2685 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
2686 	if (!wait_loop)
2687 		e_err(drv, "Could not enable Tx Queue %d\n", reg_idx);
2688 }
2689 
2690 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
2691 {
2692 	struct ixgbe_hw *hw = &adapter->hw;
2693 	u32 rttdcs;
2694 	u32 reg;
2695 	u8 tcs = netdev_get_num_tc(adapter->netdev);
2696 
2697 	if (hw->mac.type == ixgbe_mac_82598EB)
2698 		return;
2699 
2700 	/* disable the arbiter while setting MTQC */
2701 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
2702 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
2703 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
2704 
2705 	/* set transmit pool layout */
2706 	switch (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
2707 	case (IXGBE_FLAG_SRIOV_ENABLED):
2708 		IXGBE_WRITE_REG(hw, IXGBE_MTQC,
2709 				(IXGBE_MTQC_VT_ENA | IXGBE_MTQC_64VF));
2710 		break;
2711 	default:
2712 		if (!tcs)
2713 			reg = IXGBE_MTQC_64Q_1PB;
2714 		else if (tcs <= 4)
2715 			reg = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
2716 		else
2717 			reg = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
2718 
2719 		IXGBE_WRITE_REG(hw, IXGBE_MTQC, reg);
2720 
2721 		/* Enable Security TX Buffer IFG for multiple pb */
2722 		if (tcs) {
2723 			reg = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
2724 			reg |= IXGBE_SECTX_DCB;
2725 			IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, reg);
2726 		}
2727 		break;
2728 	}
2729 
2730 	/* re-enable the arbiter */
2731 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
2732 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
2733 }
2734 
2735 /**
2736  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
2737  * @adapter: board private structure
2738  *
2739  * Configure the Tx unit of the MAC after a reset.
2740  **/
2741 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
2742 {
2743 	struct ixgbe_hw *hw = &adapter->hw;
2744 	u32 dmatxctl;
2745 	u32 i;
2746 
2747 	ixgbe_setup_mtqc(adapter);
2748 
2749 	if (hw->mac.type != ixgbe_mac_82598EB) {
2750 		/* DMATXCTL.EN must be before Tx queues are enabled */
2751 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2752 		dmatxctl |= IXGBE_DMATXCTL_TE;
2753 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
2754 	}
2755 
2756 	/* Setup the HW Tx Head and Tail descriptor pointers */
2757 	for (i = 0; i < adapter->num_tx_queues; i++)
2758 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
2759 }
2760 
2761 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
2762 
2763 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
2764 				   struct ixgbe_ring *rx_ring)
2765 {
2766 	u32 srrctl;
2767 	u8 reg_idx = rx_ring->reg_idx;
2768 
2769 	switch (adapter->hw.mac.type) {
2770 	case ixgbe_mac_82598EB: {
2771 		struct ixgbe_ring_feature *feature = adapter->ring_feature;
2772 		const int mask = feature[RING_F_RSS].mask;
2773 		reg_idx = reg_idx & mask;
2774 	}
2775 		break;
2776 	case ixgbe_mac_82599EB:
2777 	case ixgbe_mac_X540:
2778 	default:
2779 		break;
2780 	}
2781 
2782 	srrctl = IXGBE_READ_REG(&adapter->hw, IXGBE_SRRCTL(reg_idx));
2783 
2784 	srrctl &= ~IXGBE_SRRCTL_BSIZEHDR_MASK;
2785 	srrctl &= ~IXGBE_SRRCTL_BSIZEPKT_MASK;
2786 	if (adapter->num_vfs)
2787 		srrctl |= IXGBE_SRRCTL_DROP_EN;
2788 
2789 	srrctl |= (IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT) &
2790 		  IXGBE_SRRCTL_BSIZEHDR_MASK;
2791 
2792 #if PAGE_SIZE > IXGBE_MAX_RXBUFFER
2793 	srrctl |= IXGBE_MAX_RXBUFFER >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
2794 #else
2795 	srrctl |= ixgbe_rx_bufsz(rx_ring) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
2796 #endif
2797 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
2798 
2799 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_SRRCTL(reg_idx), srrctl);
2800 }
2801 
2802 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
2803 {
2804 	struct ixgbe_hw *hw = &adapter->hw;
2805 	static const u32 seed[10] = { 0xE291D73D, 0x1805EC6C, 0x2A94B30D,
2806 			  0xA54F2BEC, 0xEA49AF7C, 0xE214AD3D, 0xB855AABE,
2807 			  0x6A3E67EA, 0x14364D17, 0x3BED200D};
2808 	u32 mrqc = 0, reta = 0;
2809 	u32 rxcsum;
2810 	int i, j;
2811 	u8 tcs = netdev_get_num_tc(adapter->netdev);
2812 	int maxq = adapter->ring_feature[RING_F_RSS].indices;
2813 
2814 	if (tcs)
2815 		maxq = min(maxq, adapter->num_tx_queues / tcs);
2816 
2817 	/* Fill out hash function seeds */
2818 	for (i = 0; i < 10; i++)
2819 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), seed[i]);
2820 
2821 	/* Fill out redirection table */
2822 	for (i = 0, j = 0; i < 128; i++, j++) {
2823 		if (j == maxq)
2824 			j = 0;
2825 		/* reta = 4-byte sliding window of
2826 		 * 0x00..(indices-1)(indices-1)00..etc. */
2827 		reta = (reta << 8) | (j * 0x11);
2828 		if ((i & 3) == 3)
2829 			IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
2830 	}
2831 
2832 	/* Disable indicating checksum in descriptor, enables RSS hash */
2833 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
2834 	rxcsum |= IXGBE_RXCSUM_PCSD;
2835 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
2836 
2837 	if (adapter->hw.mac.type == ixgbe_mac_82598EB &&
2838 	    (adapter->flags & IXGBE_FLAG_RSS_ENABLED)) {
2839 		mrqc = IXGBE_MRQC_RSSEN;
2840 	} else {
2841 		int mask = adapter->flags & (IXGBE_FLAG_RSS_ENABLED
2842 					     | IXGBE_FLAG_SRIOV_ENABLED);
2843 
2844 		switch (mask) {
2845 		case (IXGBE_FLAG_RSS_ENABLED):
2846 			if (!tcs)
2847 				mrqc = IXGBE_MRQC_RSSEN;
2848 			else if (tcs <= 4)
2849 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
2850 			else
2851 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
2852 			break;
2853 		case (IXGBE_FLAG_SRIOV_ENABLED):
2854 			mrqc = IXGBE_MRQC_VMDQEN;
2855 			break;
2856 		default:
2857 			break;
2858 		}
2859 	}
2860 
2861 	/* Perform hash on these packet types */
2862 	mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4
2863 	      | IXGBE_MRQC_RSS_FIELD_IPV4_TCP
2864 	      | IXGBE_MRQC_RSS_FIELD_IPV6
2865 	      | IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
2866 
2867 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
2868 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
2869 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
2870 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
2871 
2872 	IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
2873 }
2874 
2875 /**
2876  * ixgbe_configure_rscctl - enable RSC for the indicated ring
2877  * @adapter:    address of board private structure
2878  * @index:      index of ring to set
2879  **/
2880 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
2881 				   struct ixgbe_ring *ring)
2882 {
2883 	struct ixgbe_hw *hw = &adapter->hw;
2884 	u32 rscctrl;
2885 	u8 reg_idx = ring->reg_idx;
2886 
2887 	if (!ring_is_rsc_enabled(ring))
2888 		return;
2889 
2890 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
2891 	rscctrl |= IXGBE_RSCCTL_RSCEN;
2892 	/*
2893 	 * we must limit the number of descriptors so that the
2894 	 * total size of max desc * buf_len is not greater
2895 	 * than 65536
2896 	 */
2897 #if (PAGE_SIZE <= 8192)
2898 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
2899 #elif (PAGE_SIZE <= 16384)
2900 	rscctrl |= IXGBE_RSCCTL_MAXDESC_8;
2901 #else
2902 	rscctrl |= IXGBE_RSCCTL_MAXDESC_4;
2903 #endif
2904 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
2905 }
2906 
2907 /**
2908  *  ixgbe_set_uta - Set unicast filter table address
2909  *  @adapter: board private structure
2910  *
2911  *  The unicast table address is a register array of 32-bit registers.
2912  *  The table is meant to be used in a way similar to how the MTA is used
2913  *  however due to certain limitations in the hardware it is necessary to
2914  *  set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscuous
2915  *  enable bit to allow vlan tag stripping when promiscuous mode is enabled
2916  **/
2917 static void ixgbe_set_uta(struct ixgbe_adapter *adapter)
2918 {
2919 	struct ixgbe_hw *hw = &adapter->hw;
2920 	int i;
2921 
2922 	/* The UTA table only exists on 82599 hardware and newer */
2923 	if (hw->mac.type < ixgbe_mac_82599EB)
2924 		return;
2925 
2926 	/* we only need to do this if VMDq is enabled */
2927 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
2928 		return;
2929 
2930 	for (i = 0; i < 128; i++)
2931 		IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
2932 }
2933 
2934 #define IXGBE_MAX_RX_DESC_POLL 10
2935 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
2936 				       struct ixgbe_ring *ring)
2937 {
2938 	struct ixgbe_hw *hw = &adapter->hw;
2939 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
2940 	u32 rxdctl;
2941 	u8 reg_idx = ring->reg_idx;
2942 
2943 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
2944 	if (hw->mac.type == ixgbe_mac_82598EB &&
2945 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
2946 		return;
2947 
2948 	do {
2949 		usleep_range(1000, 2000);
2950 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
2951 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
2952 
2953 	if (!wait_loop) {
2954 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
2955 		      "the polling period\n", reg_idx);
2956 	}
2957 }
2958 
2959 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
2960 			    struct ixgbe_ring *ring)
2961 {
2962 	struct ixgbe_hw *hw = &adapter->hw;
2963 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
2964 	u32 rxdctl;
2965 	u8 reg_idx = ring->reg_idx;
2966 
2967 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
2968 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
2969 
2970 	/* write value back with RXDCTL.ENABLE bit cleared */
2971 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
2972 
2973 	if (hw->mac.type == ixgbe_mac_82598EB &&
2974 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
2975 		return;
2976 
2977 	/* the hardware may take up to 100us to really disable the rx queue */
2978 	do {
2979 		udelay(10);
2980 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
2981 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
2982 
2983 	if (!wait_loop) {
2984 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
2985 		      "the polling period\n", reg_idx);
2986 	}
2987 }
2988 
2989 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
2990 			     struct ixgbe_ring *ring)
2991 {
2992 	struct ixgbe_hw *hw = &adapter->hw;
2993 	u64 rdba = ring->dma;
2994 	u32 rxdctl;
2995 	u8 reg_idx = ring->reg_idx;
2996 
2997 	/* disable queue to avoid issues while updating state */
2998 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
2999 	ixgbe_disable_rx_queue(adapter, ring);
3000 
3001 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
3002 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
3003 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
3004 			ring->count * sizeof(union ixgbe_adv_rx_desc));
3005 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
3006 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
3007 	ring->tail = hw->hw_addr + IXGBE_RDT(reg_idx);
3008 
3009 	ixgbe_configure_srrctl(adapter, ring);
3010 	ixgbe_configure_rscctl(adapter, ring);
3011 
3012 	/* If operating in IOV mode set RLPML for X540 */
3013 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
3014 	    hw->mac.type == ixgbe_mac_X540) {
3015 		rxdctl &= ~IXGBE_RXDCTL_RLPMLMASK;
3016 		rxdctl |= ((ring->netdev->mtu + ETH_HLEN +
3017 			    ETH_FCS_LEN + VLAN_HLEN) | IXGBE_RXDCTL_RLPML_EN);
3018 	}
3019 
3020 	if (hw->mac.type == ixgbe_mac_82598EB) {
3021 		/*
3022 		 * enable cache line friendly hardware writes:
3023 		 * PTHRESH=32 descriptors (half the internal cache),
3024 		 * this also removes ugly rx_no_buffer_count increment
3025 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
3026 		 * WTHRESH=8 burst writeback up to two cache lines
3027 		 */
3028 		rxdctl &= ~0x3FFFFF;
3029 		rxdctl |=  0x080420;
3030 	}
3031 
3032 	/* enable receive descriptor ring */
3033 	rxdctl |= IXGBE_RXDCTL_ENABLE;
3034 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3035 
3036 	ixgbe_rx_desc_queue_enable(adapter, ring);
3037 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
3038 }
3039 
3040 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
3041 {
3042 	struct ixgbe_hw *hw = &adapter->hw;
3043 	int p;
3044 
3045 	/* PSRTYPE must be initialized in non 82598 adapters */
3046 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
3047 		      IXGBE_PSRTYPE_UDPHDR |
3048 		      IXGBE_PSRTYPE_IPV4HDR |
3049 		      IXGBE_PSRTYPE_L2HDR |
3050 		      IXGBE_PSRTYPE_IPV6HDR;
3051 
3052 	if (hw->mac.type == ixgbe_mac_82598EB)
3053 		return;
3054 
3055 	if (adapter->flags & IXGBE_FLAG_RSS_ENABLED)
3056 		psrtype |= (adapter->num_rx_queues_per_pool << 29);
3057 
3058 	for (p = 0; p < adapter->num_rx_pools; p++)
3059 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(adapter->num_vfs + p),
3060 				psrtype);
3061 }
3062 
3063 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
3064 {
3065 	struct ixgbe_hw *hw = &adapter->hw;
3066 	u32 gcr_ext;
3067 	u32 vt_reg_bits;
3068 	u32 reg_offset, vf_shift;
3069 	u32 vmdctl;
3070 	int i;
3071 
3072 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3073 		return;
3074 
3075 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3076 	vt_reg_bits = IXGBE_VMD_CTL_VMDQ_EN | IXGBE_VT_CTL_REPLEN;
3077 	vt_reg_bits |= (adapter->num_vfs << IXGBE_VT_CTL_POOL_SHIFT);
3078 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl | vt_reg_bits);
3079 
3080 	vf_shift = adapter->num_vfs % 32;
3081 	reg_offset = (adapter->num_vfs >= 32) ? 1 : 0;
3082 
3083 	/* Enable only the PF's pool for Tx/Rx */
3084 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), (1 << vf_shift));
3085 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), 0);
3086 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), (1 << vf_shift));
3087 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), 0);
3088 	IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3089 
3090 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3091 	hw->mac.ops.set_vmdq(hw, 0, adapter->num_vfs);
3092 
3093 	/*
3094 	 * Set up VF register offsets for selected VT Mode,
3095 	 * i.e. 32 or 64 VFs for SR-IOV
3096 	 */
3097 	gcr_ext = IXGBE_READ_REG(hw, IXGBE_GCR_EXT);
3098 	gcr_ext |= IXGBE_GCR_EXT_MSIX_EN;
3099 	gcr_ext |= IXGBE_GCR_EXT_VT_MODE_64;
3100 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
3101 
3102 	/* enable Tx loopback for VF/PF communication */
3103 	IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3104 	/* Enable MAC Anti-Spoofing */
3105 	hw->mac.ops.set_mac_anti_spoofing(hw,
3106 					   (adapter->num_vfs != 0),
3107 					  adapter->num_vfs);
3108 	/* For VFs that have spoof checking turned off */
3109 	for (i = 0; i < adapter->num_vfs; i++) {
3110 		if (!adapter->vfinfo[i].spoofchk_enabled)
3111 			ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i, false);
3112 	}
3113 }
3114 
3115 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
3116 {
3117 	struct ixgbe_hw *hw = &adapter->hw;
3118 	struct net_device *netdev = adapter->netdev;
3119 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3120 	struct ixgbe_ring *rx_ring;
3121 	int i;
3122 	u32 mhadd, hlreg0;
3123 
3124 #ifdef IXGBE_FCOE
3125 	/* adjust max frame to be able to do baby jumbo for FCoE */
3126 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
3127 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
3128 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3129 
3130 #endif /* IXGBE_FCOE */
3131 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3132 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
3133 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3134 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
3135 
3136 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3137 	}
3138 
3139 	/* MHADD will allow an extra 4 bytes past for vlan tagged frames */
3140 	max_frame += VLAN_HLEN;
3141 
3142 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3143 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3144 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3145 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3146 
3147 	/*
3148 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3149 	 * the Base and Length of the Rx Descriptor Ring
3150 	 */
3151 	for (i = 0; i < adapter->num_rx_queues; i++) {
3152 		rx_ring = adapter->rx_ring[i];
3153 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
3154 			set_ring_rsc_enabled(rx_ring);
3155 		else
3156 			clear_ring_rsc_enabled(rx_ring);
3157 	}
3158 }
3159 
3160 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
3161 {
3162 	struct ixgbe_hw *hw = &adapter->hw;
3163 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
3164 
3165 	switch (hw->mac.type) {
3166 	case ixgbe_mac_82598EB:
3167 		/*
3168 		 * For VMDq support of different descriptor types or
3169 		 * buffer sizes through the use of multiple SRRCTL
3170 		 * registers, RDRXCTL.MVMEN must be set to 1
3171 		 *
3172 		 * also, the manual doesn't mention it clearly but DCA hints
3173 		 * will only use queue 0's tags unless this bit is set.  Side
3174 		 * effects of setting this bit are only that SRRCTL must be
3175 		 * fully programmed [0..15]
3176 		 */
3177 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
3178 		break;
3179 	case ixgbe_mac_82599EB:
3180 	case ixgbe_mac_X540:
3181 		/* Disable RSC for ACK packets */
3182 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
3183 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
3184 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
3185 		/* hardware requires some bits to be set by default */
3186 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
3187 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
3188 		break;
3189 	default:
3190 		/* We should do nothing since we don't know this hardware */
3191 		return;
3192 	}
3193 
3194 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
3195 }
3196 
3197 /**
3198  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3199  * @adapter: board private structure
3200  *
3201  * Configure the Rx unit of the MAC after a reset.
3202  **/
3203 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
3204 {
3205 	struct ixgbe_hw *hw = &adapter->hw;
3206 	int i;
3207 	u32 rxctrl;
3208 
3209 	/* disable receives while setting up the descriptors */
3210 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3211 	IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxctrl & ~IXGBE_RXCTRL_RXEN);
3212 
3213 	ixgbe_setup_psrtype(adapter);
3214 	ixgbe_setup_rdrxctl(adapter);
3215 
3216 	/* Program registers for the distribution of queues */
3217 	ixgbe_setup_mrqc(adapter);
3218 
3219 	ixgbe_set_uta(adapter);
3220 
3221 	/* set_rx_buffer_len must be called before ring initialization */
3222 	ixgbe_set_rx_buffer_len(adapter);
3223 
3224 	/*
3225 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3226 	 * the Base and Length of the Rx Descriptor Ring
3227 	 */
3228 	for (i = 0; i < adapter->num_rx_queues; i++)
3229 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
3230 
3231 	/* disable drop enable for 82598 parts */
3232 	if (hw->mac.type == ixgbe_mac_82598EB)
3233 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3234 
3235 	/* enable all receives */
3236 	rxctrl |= IXGBE_RXCTRL_RXEN;
3237 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
3238 }
3239 
3240 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev, u16 vid)
3241 {
3242 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3243 	struct ixgbe_hw *hw = &adapter->hw;
3244 	int pool_ndx = adapter->num_vfs;
3245 
3246 	/* add VID to filter table */
3247 	hw->mac.ops.set_vfta(&adapter->hw, vid, pool_ndx, true);
3248 	set_bit(vid, adapter->active_vlans);
3249 
3250 	return 0;
3251 }
3252 
3253 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev, u16 vid)
3254 {
3255 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3256 	struct ixgbe_hw *hw = &adapter->hw;
3257 	int pool_ndx = adapter->num_vfs;
3258 
3259 	/* remove VID from filter table */
3260 	hw->mac.ops.set_vfta(&adapter->hw, vid, pool_ndx, false);
3261 	clear_bit(vid, adapter->active_vlans);
3262 
3263 	return 0;
3264 }
3265 
3266 /**
3267  * ixgbe_vlan_filter_disable - helper to disable hw vlan filtering
3268  * @adapter: driver data
3269  */
3270 static void ixgbe_vlan_filter_disable(struct ixgbe_adapter *adapter)
3271 {
3272 	struct ixgbe_hw *hw = &adapter->hw;
3273 	u32 vlnctrl;
3274 
3275 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3276 	vlnctrl &= ~(IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
3277 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3278 }
3279 
3280 /**
3281  * ixgbe_vlan_filter_enable - helper to enable hw vlan filtering
3282  * @adapter: driver data
3283  */
3284 static void ixgbe_vlan_filter_enable(struct ixgbe_adapter *adapter)
3285 {
3286 	struct ixgbe_hw *hw = &adapter->hw;
3287 	u32 vlnctrl;
3288 
3289 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3290 	vlnctrl |= IXGBE_VLNCTRL_VFE;
3291 	vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
3292 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3293 }
3294 
3295 /**
3296  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
3297  * @adapter: driver data
3298  */
3299 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
3300 {
3301 	struct ixgbe_hw *hw = &adapter->hw;
3302 	u32 vlnctrl;
3303 	int i, j;
3304 
3305 	switch (hw->mac.type) {
3306 	case ixgbe_mac_82598EB:
3307 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3308 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
3309 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3310 		break;
3311 	case ixgbe_mac_82599EB:
3312 	case ixgbe_mac_X540:
3313 		for (i = 0; i < adapter->num_rx_queues; i++) {
3314 			j = adapter->rx_ring[i]->reg_idx;
3315 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3316 			vlnctrl &= ~IXGBE_RXDCTL_VME;
3317 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3318 		}
3319 		break;
3320 	default:
3321 		break;
3322 	}
3323 }
3324 
3325 /**
3326  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
3327  * @adapter: driver data
3328  */
3329 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
3330 {
3331 	struct ixgbe_hw *hw = &adapter->hw;
3332 	u32 vlnctrl;
3333 	int i, j;
3334 
3335 	switch (hw->mac.type) {
3336 	case ixgbe_mac_82598EB:
3337 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3338 		vlnctrl |= IXGBE_VLNCTRL_VME;
3339 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3340 		break;
3341 	case ixgbe_mac_82599EB:
3342 	case ixgbe_mac_X540:
3343 		for (i = 0; i < adapter->num_rx_queues; i++) {
3344 			j = adapter->rx_ring[i]->reg_idx;
3345 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3346 			vlnctrl |= IXGBE_RXDCTL_VME;
3347 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3348 		}
3349 		break;
3350 	default:
3351 		break;
3352 	}
3353 }
3354 
3355 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
3356 {
3357 	u16 vid;
3358 
3359 	ixgbe_vlan_rx_add_vid(adapter->netdev, 0);
3360 
3361 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
3362 		ixgbe_vlan_rx_add_vid(adapter->netdev, vid);
3363 }
3364 
3365 /**
3366  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
3367  * @netdev: network interface device structure
3368  *
3369  * Writes unicast address list to the RAR table.
3370  * Returns: -ENOMEM on failure/insufficient address space
3371  *                0 on no addresses written
3372  *                X on writing X addresses to the RAR table
3373  **/
3374 static int ixgbe_write_uc_addr_list(struct net_device *netdev)
3375 {
3376 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3377 	struct ixgbe_hw *hw = &adapter->hw;
3378 	unsigned int vfn = adapter->num_vfs;
3379 	unsigned int rar_entries = IXGBE_MAX_PF_MACVLANS;
3380 	int count = 0;
3381 
3382 	/* return ENOMEM indicating insufficient memory for addresses */
3383 	if (netdev_uc_count(netdev) > rar_entries)
3384 		return -ENOMEM;
3385 
3386 	if (!netdev_uc_empty(netdev) && rar_entries) {
3387 		struct netdev_hw_addr *ha;
3388 		/* return error if we do not support writing to RAR table */
3389 		if (!hw->mac.ops.set_rar)
3390 			return -ENOMEM;
3391 
3392 		netdev_for_each_uc_addr(ha, netdev) {
3393 			if (!rar_entries)
3394 				break;
3395 			hw->mac.ops.set_rar(hw, rar_entries--, ha->addr,
3396 					    vfn, IXGBE_RAH_AV);
3397 			count++;
3398 		}
3399 	}
3400 	/* write the addresses in reverse order to avoid write combining */
3401 	for (; rar_entries > 0 ; rar_entries--)
3402 		hw->mac.ops.clear_rar(hw, rar_entries);
3403 
3404 	return count;
3405 }
3406 
3407 /**
3408  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
3409  * @netdev: network interface device structure
3410  *
3411  * The set_rx_method entry point is called whenever the unicast/multicast
3412  * address list or the network interface flags are updated.  This routine is
3413  * responsible for configuring the hardware for proper unicast, multicast and
3414  * promiscuous mode.
3415  **/
3416 void ixgbe_set_rx_mode(struct net_device *netdev)
3417 {
3418 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3419 	struct ixgbe_hw *hw = &adapter->hw;
3420 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
3421 	int count;
3422 
3423 	/* Check for Promiscuous and All Multicast modes */
3424 
3425 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3426 
3427 	/* set all bits that we expect to always be set */
3428 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
3429 	fctrl |= IXGBE_FCTRL_BAM;
3430 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
3431 	fctrl |= IXGBE_FCTRL_PMCF;
3432 
3433 	/* clear the bits we are changing the status of */
3434 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3435 
3436 	if (netdev->flags & IFF_PROMISC) {
3437 		hw->addr_ctrl.user_set_promisc = true;
3438 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3439 		vmolr |= (IXGBE_VMOLR_ROPE | IXGBE_VMOLR_MPE);
3440 		/* don't hardware filter vlans in promisc mode */
3441 		ixgbe_vlan_filter_disable(adapter);
3442 	} else {
3443 		if (netdev->flags & IFF_ALLMULTI) {
3444 			fctrl |= IXGBE_FCTRL_MPE;
3445 			vmolr |= IXGBE_VMOLR_MPE;
3446 		} else {
3447 			/*
3448 			 * Write addresses to the MTA, if the attempt fails
3449 			 * then we should just turn on promiscuous mode so
3450 			 * that we can at least receive multicast traffic
3451 			 */
3452 			hw->mac.ops.update_mc_addr_list(hw, netdev);
3453 			vmolr |= IXGBE_VMOLR_ROMPE;
3454 		}
3455 		ixgbe_vlan_filter_enable(adapter);
3456 		hw->addr_ctrl.user_set_promisc = false;
3457 		/*
3458 		 * Write addresses to available RAR registers, if there is not
3459 		 * sufficient space to store all the addresses then enable
3460 		 * unicast promiscuous mode
3461 		 */
3462 		count = ixgbe_write_uc_addr_list(netdev);
3463 		if (count < 0) {
3464 			fctrl |= IXGBE_FCTRL_UPE;
3465 			vmolr |= IXGBE_VMOLR_ROPE;
3466 		}
3467 	}
3468 
3469 	if (adapter->num_vfs) {
3470 		ixgbe_restore_vf_multicasts(adapter);
3471 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(adapter->num_vfs)) &
3472 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
3473 			   IXGBE_VMOLR_ROPE);
3474 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(adapter->num_vfs), vmolr);
3475 	}
3476 
3477 	/* This is useful for sniffing bad packets. */
3478 	if (adapter->netdev->features & NETIF_F_RXALL) {
3479 		/* UPE and MPE will be handled by normal PROMISC logic
3480 		 * in e1000e_set_rx_mode */
3481 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
3482 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
3483 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
3484 
3485 		fctrl &= ~(IXGBE_FCTRL_DPF);
3486 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
3487 	}
3488 
3489 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
3490 
3491 	if (netdev->features & NETIF_F_HW_VLAN_RX)
3492 		ixgbe_vlan_strip_enable(adapter);
3493 	else
3494 		ixgbe_vlan_strip_disable(adapter);
3495 }
3496 
3497 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
3498 {
3499 	int q_idx;
3500 	struct ixgbe_q_vector *q_vector;
3501 	int q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
3502 
3503 	/* legacy and MSI only use one vector */
3504 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED))
3505 		q_vectors = 1;
3506 
3507 	for (q_idx = 0; q_idx < q_vectors; q_idx++) {
3508 		q_vector = adapter->q_vector[q_idx];
3509 		napi_enable(&q_vector->napi);
3510 	}
3511 }
3512 
3513 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
3514 {
3515 	int q_idx;
3516 	struct ixgbe_q_vector *q_vector;
3517 	int q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
3518 
3519 	/* legacy and MSI only use one vector */
3520 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED))
3521 		q_vectors = 1;
3522 
3523 	for (q_idx = 0; q_idx < q_vectors; q_idx++) {
3524 		q_vector = adapter->q_vector[q_idx];
3525 		napi_disable(&q_vector->napi);
3526 	}
3527 }
3528 
3529 #ifdef CONFIG_IXGBE_DCB
3530 /*
3531  * ixgbe_configure_dcb - Configure DCB hardware
3532  * @adapter: ixgbe adapter struct
3533  *
3534  * This is called by the driver on open to configure the DCB hardware.
3535  * This is also called by the gennetlink interface when reconfiguring
3536  * the DCB state.
3537  */
3538 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
3539 {
3540 	struct ixgbe_hw *hw = &adapter->hw;
3541 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3542 
3543 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
3544 		if (hw->mac.type == ixgbe_mac_82598EB)
3545 			netif_set_gso_max_size(adapter->netdev, 65536);
3546 		return;
3547 	}
3548 
3549 	if (hw->mac.type == ixgbe_mac_82598EB)
3550 		netif_set_gso_max_size(adapter->netdev, 32768);
3551 
3552 
3553 	/* Enable VLAN tag insert/strip */
3554 	adapter->netdev->features |= NETIF_F_HW_VLAN_RX;
3555 
3556 	hw->mac.ops.set_vfta(&adapter->hw, 0, 0, true);
3557 
3558 #ifdef IXGBE_FCOE
3559 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
3560 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
3561 #endif
3562 
3563 	/* reconfigure the hardware */
3564 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
3565 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
3566 						DCB_TX_CONFIG);
3567 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
3568 						DCB_RX_CONFIG);
3569 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
3570 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
3571 		ixgbe_dcb_hw_ets(&adapter->hw,
3572 				 adapter->ixgbe_ieee_ets,
3573 				 max_frame);
3574 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
3575 					adapter->ixgbe_ieee_pfc->pfc_en,
3576 					adapter->ixgbe_ieee_ets->prio_tc);
3577 	}
3578 
3579 	/* Enable RSS Hash per TC */
3580 	if (hw->mac.type != ixgbe_mac_82598EB) {
3581 		int i;
3582 		u32 reg = 0;
3583 
3584 		for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
3585 			u8 msb = 0;
3586 			u8 cnt = adapter->netdev->tc_to_txq[i].count;
3587 
3588 			while (cnt >>= 1)
3589 				msb++;
3590 
3591 			reg |= msb << IXGBE_RQTC_SHIFT_TC(i);
3592 		}
3593 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, reg);
3594 	}
3595 }
3596 #endif
3597 
3598 /* Additional bittime to account for IXGBE framing */
3599 #define IXGBE_ETH_FRAMING 20
3600 
3601 /*
3602  * ixgbe_hpbthresh - calculate high water mark for flow control
3603  *
3604  * @adapter: board private structure to calculate for
3605  * @pb - packet buffer to calculate
3606  */
3607 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
3608 {
3609 	struct ixgbe_hw *hw = &adapter->hw;
3610 	struct net_device *dev = adapter->netdev;
3611 	int link, tc, kb, marker;
3612 	u32 dv_id, rx_pba;
3613 
3614 	/* Calculate max LAN frame size */
3615 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
3616 
3617 #ifdef IXGBE_FCOE
3618 	/* FCoE traffic class uses FCOE jumbo frames */
3619 	if (dev->features & NETIF_F_FCOE_MTU) {
3620 		int fcoe_pb = 0;
3621 
3622 #ifdef CONFIG_IXGBE_DCB
3623 		fcoe_pb = netdev_get_prio_tc_map(dev, adapter->fcoe.up);
3624 
3625 #endif
3626 		if (fcoe_pb == pb && tc < IXGBE_FCOE_JUMBO_FRAME_SIZE)
3627 			tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3628 	}
3629 #endif
3630 
3631 	/* Calculate delay value for device */
3632 	switch (hw->mac.type) {
3633 	case ixgbe_mac_X540:
3634 		dv_id = IXGBE_DV_X540(link, tc);
3635 		break;
3636 	default:
3637 		dv_id = IXGBE_DV(link, tc);
3638 		break;
3639 	}
3640 
3641 	/* Loopback switch introduces additional latency */
3642 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3643 		dv_id += IXGBE_B2BT(tc);
3644 
3645 	/* Delay value is calculated in bit times convert to KB */
3646 	kb = IXGBE_BT2KB(dv_id);
3647 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
3648 
3649 	marker = rx_pba - kb;
3650 
3651 	/* It is possible that the packet buffer is not large enough
3652 	 * to provide required headroom. In this case throw an error
3653 	 * to user and a do the best we can.
3654 	 */
3655 	if (marker < 0) {
3656 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
3657 			    "headroom to support flow control."
3658 			    "Decrease MTU or number of traffic classes\n", pb);
3659 		marker = tc + 1;
3660 	}
3661 
3662 	return marker;
3663 }
3664 
3665 /*
3666  * ixgbe_lpbthresh - calculate low water mark for for flow control
3667  *
3668  * @adapter: board private structure to calculate for
3669  * @pb - packet buffer to calculate
3670  */
3671 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter)
3672 {
3673 	struct ixgbe_hw *hw = &adapter->hw;
3674 	struct net_device *dev = adapter->netdev;
3675 	int tc;
3676 	u32 dv_id;
3677 
3678 	/* Calculate max LAN frame size */
3679 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
3680 
3681 	/* Calculate delay value for device */
3682 	switch (hw->mac.type) {
3683 	case ixgbe_mac_X540:
3684 		dv_id = IXGBE_LOW_DV_X540(tc);
3685 		break;
3686 	default:
3687 		dv_id = IXGBE_LOW_DV(tc);
3688 		break;
3689 	}
3690 
3691 	/* Delay value is calculated in bit times convert to KB */
3692 	return IXGBE_BT2KB(dv_id);
3693 }
3694 
3695 /*
3696  * ixgbe_pbthresh_setup - calculate and setup high low water marks
3697  */
3698 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
3699 {
3700 	struct ixgbe_hw *hw = &adapter->hw;
3701 	int num_tc = netdev_get_num_tc(adapter->netdev);
3702 	int i;
3703 
3704 	if (!num_tc)
3705 		num_tc = 1;
3706 
3707 	hw->fc.low_water = ixgbe_lpbthresh(adapter);
3708 
3709 	for (i = 0; i < num_tc; i++) {
3710 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
3711 
3712 		/* Low water marks must not be larger than high water marks */
3713 		if (hw->fc.low_water > hw->fc.high_water[i])
3714 			hw->fc.low_water = 0;
3715 	}
3716 }
3717 
3718 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
3719 {
3720 	struct ixgbe_hw *hw = &adapter->hw;
3721 	int hdrm;
3722 	u8 tc = netdev_get_num_tc(adapter->netdev);
3723 
3724 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
3725 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
3726 		hdrm = 32 << adapter->fdir_pballoc;
3727 	else
3728 		hdrm = 0;
3729 
3730 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
3731 	ixgbe_pbthresh_setup(adapter);
3732 }
3733 
3734 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
3735 {
3736 	struct ixgbe_hw *hw = &adapter->hw;
3737 	struct hlist_node *node, *node2;
3738 	struct ixgbe_fdir_filter *filter;
3739 
3740 	spin_lock(&adapter->fdir_perfect_lock);
3741 
3742 	if (!hlist_empty(&adapter->fdir_filter_list))
3743 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
3744 
3745 	hlist_for_each_entry_safe(filter, node, node2,
3746 				  &adapter->fdir_filter_list, fdir_node) {
3747 		ixgbe_fdir_write_perfect_filter_82599(hw,
3748 				&filter->filter,
3749 				filter->sw_idx,
3750 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
3751 				IXGBE_FDIR_DROP_QUEUE :
3752 				adapter->rx_ring[filter->action]->reg_idx);
3753 	}
3754 
3755 	spin_unlock(&adapter->fdir_perfect_lock);
3756 }
3757 
3758 static void ixgbe_configure(struct ixgbe_adapter *adapter)
3759 {
3760 	struct ixgbe_hw *hw = &adapter->hw;
3761 
3762 	ixgbe_configure_pb(adapter);
3763 #ifdef CONFIG_IXGBE_DCB
3764 	ixgbe_configure_dcb(adapter);
3765 #endif
3766 
3767 	ixgbe_set_rx_mode(adapter->netdev);
3768 	ixgbe_restore_vlan(adapter);
3769 
3770 #ifdef IXGBE_FCOE
3771 	if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
3772 		ixgbe_configure_fcoe(adapter);
3773 
3774 #endif /* IXGBE_FCOE */
3775 
3776 	switch (hw->mac.type) {
3777 	case ixgbe_mac_82599EB:
3778 	case ixgbe_mac_X540:
3779 		hw->mac.ops.disable_rx_buff(hw);
3780 		break;
3781 	default:
3782 		break;
3783 	}
3784 
3785 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3786 		ixgbe_init_fdir_signature_82599(&adapter->hw,
3787 						adapter->fdir_pballoc);
3788 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
3789 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
3790 					      adapter->fdir_pballoc);
3791 		ixgbe_fdir_filter_restore(adapter);
3792 	}
3793 
3794 	switch (hw->mac.type) {
3795 	case ixgbe_mac_82599EB:
3796 	case ixgbe_mac_X540:
3797 		hw->mac.ops.enable_rx_buff(hw);
3798 		break;
3799 	default:
3800 		break;
3801 	}
3802 
3803 	ixgbe_configure_virtualization(adapter);
3804 
3805 	ixgbe_configure_tx(adapter);
3806 	ixgbe_configure_rx(adapter);
3807 }
3808 
3809 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
3810 {
3811 	switch (hw->phy.type) {
3812 	case ixgbe_phy_sfp_avago:
3813 	case ixgbe_phy_sfp_ftl:
3814 	case ixgbe_phy_sfp_intel:
3815 	case ixgbe_phy_sfp_unknown:
3816 	case ixgbe_phy_sfp_passive_tyco:
3817 	case ixgbe_phy_sfp_passive_unknown:
3818 	case ixgbe_phy_sfp_active_unknown:
3819 	case ixgbe_phy_sfp_ftl_active:
3820 		return true;
3821 	case ixgbe_phy_nl:
3822 		if (hw->mac.type == ixgbe_mac_82598EB)
3823 			return true;
3824 	default:
3825 		return false;
3826 	}
3827 }
3828 
3829 /**
3830  * ixgbe_sfp_link_config - set up SFP+ link
3831  * @adapter: pointer to private adapter struct
3832  **/
3833 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
3834 {
3835 	/*
3836 	 * We are assuming the worst case scenario here, and that
3837 	 * is that an SFP was inserted/removed after the reset
3838 	 * but before SFP detection was enabled.  As such the best
3839 	 * solution is to just start searching as soon as we start
3840 	 */
3841 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3842 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
3843 
3844 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
3845 }
3846 
3847 /**
3848  * ixgbe_non_sfp_link_config - set up non-SFP+ link
3849  * @hw: pointer to private hardware struct
3850  *
3851  * Returns 0 on success, negative on failure
3852  **/
3853 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
3854 {
3855 	u32 autoneg;
3856 	bool negotiation, link_up = false;
3857 	u32 ret = IXGBE_ERR_LINK_SETUP;
3858 
3859 	if (hw->mac.ops.check_link)
3860 		ret = hw->mac.ops.check_link(hw, &autoneg, &link_up, false);
3861 
3862 	if (ret)
3863 		goto link_cfg_out;
3864 
3865 	autoneg = hw->phy.autoneg_advertised;
3866 	if ((!autoneg) && (hw->mac.ops.get_link_capabilities))
3867 		ret = hw->mac.ops.get_link_capabilities(hw, &autoneg,
3868 							&negotiation);
3869 	if (ret)
3870 		goto link_cfg_out;
3871 
3872 	if (hw->mac.ops.setup_link)
3873 		ret = hw->mac.ops.setup_link(hw, autoneg, negotiation, link_up);
3874 link_cfg_out:
3875 	return ret;
3876 }
3877 
3878 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
3879 {
3880 	struct ixgbe_hw *hw = &adapter->hw;
3881 	u32 gpie = 0;
3882 
3883 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3884 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
3885 		       IXGBE_GPIE_OCD;
3886 		gpie |= IXGBE_GPIE_EIAME;
3887 		/*
3888 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
3889 		 * this saves a register write for every interrupt
3890 		 */
3891 		switch (hw->mac.type) {
3892 		case ixgbe_mac_82598EB:
3893 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
3894 			break;
3895 		case ixgbe_mac_82599EB:
3896 		case ixgbe_mac_X540:
3897 		default:
3898 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
3899 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
3900 			break;
3901 		}
3902 	} else {
3903 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
3904 		 * specifically only auto mask tx and rx interrupts */
3905 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
3906 	}
3907 
3908 	/* XXX: to interrupt immediately for EICS writes, enable this */
3909 	/* gpie |= IXGBE_GPIE_EIMEN; */
3910 
3911 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3912 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
3913 		gpie |= IXGBE_GPIE_VTMODE_64;
3914 	}
3915 
3916 	/* Enable Thermal over heat sensor interrupt */
3917 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
3918 		switch (adapter->hw.mac.type) {
3919 		case ixgbe_mac_82599EB:
3920 			gpie |= IXGBE_SDP0_GPIEN;
3921 			break;
3922 		case ixgbe_mac_X540:
3923 			gpie |= IXGBE_EIMS_TS;
3924 			break;
3925 		default:
3926 			break;
3927 		}
3928 	}
3929 
3930 	/* Enable fan failure interrupt */
3931 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
3932 		gpie |= IXGBE_SDP1_GPIEN;
3933 
3934 	if (hw->mac.type == ixgbe_mac_82599EB) {
3935 		gpie |= IXGBE_SDP1_GPIEN;
3936 		gpie |= IXGBE_SDP2_GPIEN;
3937 	}
3938 
3939 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
3940 }
3941 
3942 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
3943 {
3944 	struct ixgbe_hw *hw = &adapter->hw;
3945 	int err;
3946 	u32 ctrl_ext;
3947 
3948 	ixgbe_get_hw_control(adapter);
3949 	ixgbe_setup_gpie(adapter);
3950 
3951 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
3952 		ixgbe_configure_msix(adapter);
3953 	else
3954 		ixgbe_configure_msi_and_legacy(adapter);
3955 
3956 	/* enable the optics for both mult-speed fiber and 82599 SFP+ fiber */
3957 	if (hw->mac.ops.enable_tx_laser &&
3958 	    ((hw->phy.multispeed_fiber) ||
3959 	     ((hw->mac.ops.get_media_type(hw) == ixgbe_media_type_fiber) &&
3960 	      (hw->mac.type == ixgbe_mac_82599EB))))
3961 		hw->mac.ops.enable_tx_laser(hw);
3962 
3963 	clear_bit(__IXGBE_DOWN, &adapter->state);
3964 	ixgbe_napi_enable_all(adapter);
3965 
3966 	if (ixgbe_is_sfp(hw)) {
3967 		ixgbe_sfp_link_config(adapter);
3968 	} else {
3969 		err = ixgbe_non_sfp_link_config(hw);
3970 		if (err)
3971 			e_err(probe, "link_config FAILED %d\n", err);
3972 	}
3973 
3974 	/* clear any pending interrupts, may auto mask */
3975 	IXGBE_READ_REG(hw, IXGBE_EICR);
3976 	ixgbe_irq_enable(adapter, true, true);
3977 
3978 	/*
3979 	 * If this adapter has a fan, check to see if we had a failure
3980 	 * before we enabled the interrupt.
3981 	 */
3982 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
3983 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
3984 		if (esdp & IXGBE_ESDP_SDP1)
3985 			e_crit(drv, "Fan has stopped, replace the adapter\n");
3986 	}
3987 
3988 	/* enable transmits */
3989 	netif_tx_start_all_queues(adapter->netdev);
3990 
3991 	/* bring the link up in the watchdog, this could race with our first
3992 	 * link up interrupt but shouldn't be a problem */
3993 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
3994 	adapter->link_check_timeout = jiffies;
3995 	mod_timer(&adapter->service_timer, jiffies);
3996 
3997 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
3998 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
3999 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
4000 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
4001 }
4002 
4003 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
4004 {
4005 	WARN_ON(in_interrupt());
4006 	/* put off any impending NetWatchDogTimeout */
4007 	adapter->netdev->trans_start = jiffies;
4008 
4009 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
4010 		usleep_range(1000, 2000);
4011 	ixgbe_down(adapter);
4012 	/*
4013 	 * If SR-IOV enabled then wait a bit before bringing the adapter
4014 	 * back up to give the VFs time to respond to the reset.  The
4015 	 * two second wait is based upon the watchdog timer cycle in
4016 	 * the VF driver.
4017 	 */
4018 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4019 		msleep(2000);
4020 	ixgbe_up(adapter);
4021 	clear_bit(__IXGBE_RESETTING, &adapter->state);
4022 }
4023 
4024 void ixgbe_up(struct ixgbe_adapter *adapter)
4025 {
4026 	/* hardware has been reset, we need to reload some things */
4027 	ixgbe_configure(adapter);
4028 
4029 	ixgbe_up_complete(adapter);
4030 }
4031 
4032 void ixgbe_reset(struct ixgbe_adapter *adapter)
4033 {
4034 	struct ixgbe_hw *hw = &adapter->hw;
4035 	int err;
4036 
4037 	/* lock SFP init bit to prevent race conditions with the watchdog */
4038 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
4039 		usleep_range(1000, 2000);
4040 
4041 	/* clear all SFP and link config related flags while holding SFP_INIT */
4042 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
4043 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
4044 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4045 
4046 	err = hw->mac.ops.init_hw(hw);
4047 	switch (err) {
4048 	case 0:
4049 	case IXGBE_ERR_SFP_NOT_PRESENT:
4050 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
4051 		break;
4052 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
4053 		e_dev_err("master disable timed out\n");
4054 		break;
4055 	case IXGBE_ERR_EEPROM_VERSION:
4056 		/* We are running on a pre-production device, log a warning */
4057 		e_dev_warn("This device is a pre-production adapter/LOM. "
4058 			   "Please be aware there may be issues associated with "
4059 			   "your hardware.  If you are experiencing problems "
4060 			   "please contact your Intel or hardware "
4061 			   "representative who provided you with this "
4062 			   "hardware.\n");
4063 		break;
4064 	default:
4065 		e_dev_err("Hardware Error: %d\n", err);
4066 	}
4067 
4068 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
4069 
4070 	/* reprogram the RAR[0] in case user changed it. */
4071 	hw->mac.ops.set_rar(hw, 0, hw->mac.addr, adapter->num_vfs,
4072 			    IXGBE_RAH_AV);
4073 }
4074 
4075 /**
4076  * ixgbe_init_rx_page_offset - initialize page offset values for Rx buffers
4077  * @rx_ring: ring to setup
4078  *
4079  * On many IA platforms the L1 cache has a critical stride of 4K, this
4080  * results in each receive buffer starting in the same cache set.  To help
4081  * reduce the pressure on this cache set we can interleave the offsets so
4082  * that only every other buffer will be in the same cache set.
4083  **/
4084 static void ixgbe_init_rx_page_offset(struct ixgbe_ring *rx_ring)
4085 {
4086 	struct ixgbe_rx_buffer *rx_buffer = rx_ring->rx_buffer_info;
4087 	u16 i;
4088 
4089 	for (i = 0; i < rx_ring->count; i += 2) {
4090 		rx_buffer[0].page_offset = 0;
4091 		rx_buffer[1].page_offset = ixgbe_rx_bufsz(rx_ring);
4092 		rx_buffer = &rx_buffer[2];
4093 	}
4094 }
4095 
4096 /**
4097  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4098  * @rx_ring: ring to free buffers from
4099  **/
4100 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
4101 {
4102 	struct device *dev = rx_ring->dev;
4103 	unsigned long size;
4104 	u16 i;
4105 
4106 	/* ring already cleared, nothing to do */
4107 	if (!rx_ring->rx_buffer_info)
4108 		return;
4109 
4110 	/* Free all the Rx ring sk_buffs */
4111 	for (i = 0; i < rx_ring->count; i++) {
4112 		struct ixgbe_rx_buffer *rx_buffer;
4113 
4114 		rx_buffer = &rx_ring->rx_buffer_info[i];
4115 		if (rx_buffer->skb) {
4116 			struct sk_buff *skb = rx_buffer->skb;
4117 			if (IXGBE_CB(skb)->page_released) {
4118 				dma_unmap_page(dev,
4119 					       IXGBE_CB(skb)->dma,
4120 					       ixgbe_rx_bufsz(rx_ring),
4121 					       DMA_FROM_DEVICE);
4122 				IXGBE_CB(skb)->page_released = false;
4123 			}
4124 			dev_kfree_skb(skb);
4125 		}
4126 		rx_buffer->skb = NULL;
4127 		if (rx_buffer->dma)
4128 			dma_unmap_page(dev, rx_buffer->dma,
4129 				       ixgbe_rx_pg_size(rx_ring),
4130 				       DMA_FROM_DEVICE);
4131 		rx_buffer->dma = 0;
4132 		if (rx_buffer->page)
4133 			put_page(rx_buffer->page);
4134 		rx_buffer->page = NULL;
4135 	}
4136 
4137 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4138 	memset(rx_ring->rx_buffer_info, 0, size);
4139 
4140 	ixgbe_init_rx_page_offset(rx_ring);
4141 
4142 	/* Zero out the descriptor ring */
4143 	memset(rx_ring->desc, 0, rx_ring->size);
4144 
4145 	rx_ring->next_to_alloc = 0;
4146 	rx_ring->next_to_clean = 0;
4147 	rx_ring->next_to_use = 0;
4148 }
4149 
4150 /**
4151  * ixgbe_clean_tx_ring - Free Tx Buffers
4152  * @tx_ring: ring to be cleaned
4153  **/
4154 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
4155 {
4156 	struct ixgbe_tx_buffer *tx_buffer_info;
4157 	unsigned long size;
4158 	u16 i;
4159 
4160 	/* ring already cleared, nothing to do */
4161 	if (!tx_ring->tx_buffer_info)
4162 		return;
4163 
4164 	/* Free all the Tx ring sk_buffs */
4165 	for (i = 0; i < tx_ring->count; i++) {
4166 		tx_buffer_info = &tx_ring->tx_buffer_info[i];
4167 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer_info);
4168 	}
4169 
4170 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
4171 	memset(tx_ring->tx_buffer_info, 0, size);
4172 
4173 	/* Zero out the descriptor ring */
4174 	memset(tx_ring->desc, 0, tx_ring->size);
4175 
4176 	tx_ring->next_to_use = 0;
4177 	tx_ring->next_to_clean = 0;
4178 }
4179 
4180 /**
4181  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
4182  * @adapter: board private structure
4183  **/
4184 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
4185 {
4186 	int i;
4187 
4188 	for (i = 0; i < adapter->num_rx_queues; i++)
4189 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
4190 }
4191 
4192 /**
4193  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
4194  * @adapter: board private structure
4195  **/
4196 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
4197 {
4198 	int i;
4199 
4200 	for (i = 0; i < adapter->num_tx_queues; i++)
4201 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
4202 }
4203 
4204 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
4205 {
4206 	struct hlist_node *node, *node2;
4207 	struct ixgbe_fdir_filter *filter;
4208 
4209 	spin_lock(&adapter->fdir_perfect_lock);
4210 
4211 	hlist_for_each_entry_safe(filter, node, node2,
4212 				  &adapter->fdir_filter_list, fdir_node) {
4213 		hlist_del(&filter->fdir_node);
4214 		kfree(filter);
4215 	}
4216 	adapter->fdir_filter_count = 0;
4217 
4218 	spin_unlock(&adapter->fdir_perfect_lock);
4219 }
4220 
4221 void ixgbe_down(struct ixgbe_adapter *adapter)
4222 {
4223 	struct net_device *netdev = adapter->netdev;
4224 	struct ixgbe_hw *hw = &adapter->hw;
4225 	u32 rxctrl;
4226 	int i;
4227 
4228 	/* signal that we are down to the interrupt handler */
4229 	set_bit(__IXGBE_DOWN, &adapter->state);
4230 
4231 	/* disable receives */
4232 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4233 	IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxctrl & ~IXGBE_RXCTRL_RXEN);
4234 
4235 	/* disable all enabled rx queues */
4236 	for (i = 0; i < adapter->num_rx_queues; i++)
4237 		/* this call also flushes the previous write */
4238 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
4239 
4240 	usleep_range(10000, 20000);
4241 
4242 	netif_tx_stop_all_queues(netdev);
4243 
4244 	/* call carrier off first to avoid false dev_watchdog timeouts */
4245 	netif_carrier_off(netdev);
4246 	netif_tx_disable(netdev);
4247 
4248 	ixgbe_irq_disable(adapter);
4249 
4250 	ixgbe_napi_disable_all(adapter);
4251 
4252 	adapter->flags2 &= ~(IXGBE_FLAG2_FDIR_REQUIRES_REINIT |
4253 			     IXGBE_FLAG2_RESET_REQUESTED);
4254 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4255 
4256 	del_timer_sync(&adapter->service_timer);
4257 
4258 	if (adapter->num_vfs) {
4259 		/* Clear EITR Select mapping */
4260 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
4261 
4262 		/* Mark all the VFs as inactive */
4263 		for (i = 0 ; i < adapter->num_vfs; i++)
4264 			adapter->vfinfo[i].clear_to_send = false;
4265 
4266 		/* ping all the active vfs to let them know we are going down */
4267 		ixgbe_ping_all_vfs(adapter);
4268 
4269 		/* Disable all VFTE/VFRE TX/RX */
4270 		ixgbe_disable_tx_rx(adapter);
4271 	}
4272 
4273 	/* disable transmits in the hardware now that interrupts are off */
4274 	for (i = 0; i < adapter->num_tx_queues; i++) {
4275 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
4276 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
4277 	}
4278 
4279 	/* Disable the Tx DMA engine on 82599 and X540 */
4280 	switch (hw->mac.type) {
4281 	case ixgbe_mac_82599EB:
4282 	case ixgbe_mac_X540:
4283 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
4284 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
4285 				 ~IXGBE_DMATXCTL_TE));
4286 		break;
4287 	default:
4288 		break;
4289 	}
4290 
4291 	if (!pci_channel_offline(adapter->pdev))
4292 		ixgbe_reset(adapter);
4293 
4294 	/* power down the optics for multispeed fiber and 82599 SFP+ fiber */
4295 	if (hw->mac.ops.disable_tx_laser &&
4296 	    ((hw->phy.multispeed_fiber) ||
4297 	     ((hw->mac.ops.get_media_type(hw) == ixgbe_media_type_fiber) &&
4298 	      (hw->mac.type == ixgbe_mac_82599EB))))
4299 		hw->mac.ops.disable_tx_laser(hw);
4300 
4301 	ixgbe_clean_all_tx_rings(adapter);
4302 	ixgbe_clean_all_rx_rings(adapter);
4303 
4304 #ifdef CONFIG_IXGBE_DCA
4305 	/* since we reset the hardware DCA settings were cleared */
4306 	ixgbe_setup_dca(adapter);
4307 #endif
4308 }
4309 
4310 /**
4311  * ixgbe_tx_timeout - Respond to a Tx Hang
4312  * @netdev: network interface device structure
4313  **/
4314 static void ixgbe_tx_timeout(struct net_device *netdev)
4315 {
4316 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4317 
4318 	/* Do the reset outside of interrupt context */
4319 	ixgbe_tx_timeout_reset(adapter);
4320 }
4321 
4322 /**
4323  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
4324  * @adapter: board private structure to initialize
4325  *
4326  * ixgbe_sw_init initializes the Adapter private data structure.
4327  * Fields are initialized based on PCI device information and
4328  * OS network device settings (MTU size).
4329  **/
4330 static int __devinit ixgbe_sw_init(struct ixgbe_adapter *adapter)
4331 {
4332 	struct ixgbe_hw *hw = &adapter->hw;
4333 	struct pci_dev *pdev = adapter->pdev;
4334 	unsigned int rss;
4335 #ifdef CONFIG_IXGBE_DCB
4336 	int j;
4337 	struct tc_configuration *tc;
4338 #endif
4339 
4340 	/* PCI config space info */
4341 
4342 	hw->vendor_id = pdev->vendor;
4343 	hw->device_id = pdev->device;
4344 	hw->revision_id = pdev->revision;
4345 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
4346 	hw->subsystem_device_id = pdev->subsystem_device;
4347 
4348 	/* Set capability flags */
4349 	rss = min_t(int, IXGBE_MAX_RSS_INDICES, num_online_cpus());
4350 	adapter->ring_feature[RING_F_RSS].indices = rss;
4351 	adapter->flags |= IXGBE_FLAG_RSS_ENABLED;
4352 	switch (hw->mac.type) {
4353 	case ixgbe_mac_82598EB:
4354 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
4355 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
4356 		adapter->max_msix_q_vectors = MAX_MSIX_Q_VECTORS_82598;
4357 		break;
4358 	case ixgbe_mac_X540:
4359 		adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
4360 	case ixgbe_mac_82599EB:
4361 		adapter->max_msix_q_vectors = MAX_MSIX_Q_VECTORS_82599;
4362 		adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
4363 		adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
4364 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
4365 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
4366 		/* Flow Director hash filters enabled */
4367 		adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
4368 		adapter->atr_sample_rate = 20;
4369 		adapter->ring_feature[RING_F_FDIR].indices =
4370 							 IXGBE_MAX_FDIR_INDICES;
4371 		adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
4372 #ifdef IXGBE_FCOE
4373 		adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
4374 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
4375 		adapter->ring_feature[RING_F_FCOE].indices = 0;
4376 #ifdef CONFIG_IXGBE_DCB
4377 		/* Default traffic class to use for FCoE */
4378 		adapter->fcoe.up = IXGBE_FCOE_DEFTC;
4379 #endif
4380 #endif /* IXGBE_FCOE */
4381 		break;
4382 	default:
4383 		break;
4384 	}
4385 
4386 	/* n-tuple support exists, always init our spinlock */
4387 	spin_lock_init(&adapter->fdir_perfect_lock);
4388 
4389 #ifdef CONFIG_IXGBE_DCB
4390 	switch (hw->mac.type) {
4391 	case ixgbe_mac_X540:
4392 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
4393 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
4394 		break;
4395 	default:
4396 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
4397 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
4398 		break;
4399 	}
4400 
4401 	/* Configure DCB traffic classes */
4402 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
4403 		tc = &adapter->dcb_cfg.tc_config[j];
4404 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
4405 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
4406 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
4407 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
4408 		tc->dcb_pfc = pfc_disabled;
4409 	}
4410 
4411 	/* Initialize default user to priority mapping, UPx->TC0 */
4412 	tc = &adapter->dcb_cfg.tc_config[0];
4413 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
4414 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
4415 
4416 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
4417 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
4418 	adapter->dcb_cfg.pfc_mode_enable = false;
4419 	adapter->dcb_set_bitmap = 0x00;
4420 	adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
4421 	ixgbe_copy_dcb_cfg(&adapter->dcb_cfg, &adapter->temp_dcb_cfg,
4422 			   MAX_TRAFFIC_CLASS);
4423 
4424 #endif
4425 
4426 	/* default flow control settings */
4427 	hw->fc.requested_mode = ixgbe_fc_full;
4428 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
4429 #ifdef CONFIG_DCB
4430 	adapter->last_lfc_mode = hw->fc.current_mode;
4431 #endif
4432 	ixgbe_pbthresh_setup(adapter);
4433 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
4434 	hw->fc.send_xon = true;
4435 	hw->fc.disable_fc_autoneg = false;
4436 
4437 	/* enable itr by default in dynamic mode */
4438 	adapter->rx_itr_setting = 1;
4439 	adapter->tx_itr_setting = 1;
4440 
4441 	/* set default ring sizes */
4442 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
4443 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
4444 
4445 	/* set default work limits */
4446 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
4447 
4448 	/* initialize eeprom parameters */
4449 	if (ixgbe_init_eeprom_params_generic(hw)) {
4450 		e_dev_err("EEPROM initialization failed\n");
4451 		return -EIO;
4452 	}
4453 
4454 	set_bit(__IXGBE_DOWN, &adapter->state);
4455 
4456 	return 0;
4457 }
4458 
4459 /**
4460  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
4461  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
4462  *
4463  * Return 0 on success, negative on failure
4464  **/
4465 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
4466 {
4467 	struct device *dev = tx_ring->dev;
4468 	int orig_node = dev_to_node(dev);
4469 	int numa_node = -1;
4470 	int size;
4471 
4472 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
4473 
4474 	if (tx_ring->q_vector)
4475 		numa_node = tx_ring->q_vector->numa_node;
4476 
4477 	tx_ring->tx_buffer_info = vzalloc_node(size, numa_node);
4478 	if (!tx_ring->tx_buffer_info)
4479 		tx_ring->tx_buffer_info = vzalloc(size);
4480 	if (!tx_ring->tx_buffer_info)
4481 		goto err;
4482 
4483 	/* round up to nearest 4K */
4484 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
4485 	tx_ring->size = ALIGN(tx_ring->size, 4096);
4486 
4487 	set_dev_node(dev, numa_node);
4488 	tx_ring->desc = dma_alloc_coherent(dev,
4489 					   tx_ring->size,
4490 					   &tx_ring->dma,
4491 					   GFP_KERNEL);
4492 	set_dev_node(dev, orig_node);
4493 	if (!tx_ring->desc)
4494 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
4495 						   &tx_ring->dma, GFP_KERNEL);
4496 	if (!tx_ring->desc)
4497 		goto err;
4498 
4499 	tx_ring->next_to_use = 0;
4500 	tx_ring->next_to_clean = 0;
4501 	return 0;
4502 
4503 err:
4504 	vfree(tx_ring->tx_buffer_info);
4505 	tx_ring->tx_buffer_info = NULL;
4506 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
4507 	return -ENOMEM;
4508 }
4509 
4510 /**
4511  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
4512  * @adapter: board private structure
4513  *
4514  * If this function returns with an error, then it's possible one or
4515  * more of the rings is populated (while the rest are not).  It is the
4516  * callers duty to clean those orphaned rings.
4517  *
4518  * Return 0 on success, negative on failure
4519  **/
4520 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
4521 {
4522 	int i, err = 0;
4523 
4524 	for (i = 0; i < adapter->num_tx_queues; i++) {
4525 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
4526 		if (!err)
4527 			continue;
4528 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
4529 		break;
4530 	}
4531 
4532 	return err;
4533 }
4534 
4535 /**
4536  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
4537  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
4538  *
4539  * Returns 0 on success, negative on failure
4540  **/
4541 int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
4542 {
4543 	struct device *dev = rx_ring->dev;
4544 	int orig_node = dev_to_node(dev);
4545 	int numa_node = -1;
4546 	int size;
4547 
4548 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4549 
4550 	if (rx_ring->q_vector)
4551 		numa_node = rx_ring->q_vector->numa_node;
4552 
4553 	rx_ring->rx_buffer_info = vzalloc_node(size, numa_node);
4554 	if (!rx_ring->rx_buffer_info)
4555 		rx_ring->rx_buffer_info = vzalloc(size);
4556 	if (!rx_ring->rx_buffer_info)
4557 		goto err;
4558 
4559 	/* Round up to nearest 4K */
4560 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
4561 	rx_ring->size = ALIGN(rx_ring->size, 4096);
4562 
4563 	set_dev_node(dev, numa_node);
4564 	rx_ring->desc = dma_alloc_coherent(dev,
4565 					   rx_ring->size,
4566 					   &rx_ring->dma,
4567 					   GFP_KERNEL);
4568 	set_dev_node(dev, orig_node);
4569 	if (!rx_ring->desc)
4570 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
4571 						   &rx_ring->dma, GFP_KERNEL);
4572 	if (!rx_ring->desc)
4573 		goto err;
4574 
4575 	rx_ring->next_to_clean = 0;
4576 	rx_ring->next_to_use = 0;
4577 
4578 	ixgbe_init_rx_page_offset(rx_ring);
4579 
4580 	return 0;
4581 err:
4582 	vfree(rx_ring->rx_buffer_info);
4583 	rx_ring->rx_buffer_info = NULL;
4584 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
4585 	return -ENOMEM;
4586 }
4587 
4588 /**
4589  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
4590  * @adapter: board private structure
4591  *
4592  * If this function returns with an error, then it's possible one or
4593  * more of the rings is populated (while the rest are not).  It is the
4594  * callers duty to clean those orphaned rings.
4595  *
4596  * Return 0 on success, negative on failure
4597  **/
4598 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
4599 {
4600 	int i, err = 0;
4601 
4602 	for (i = 0; i < adapter->num_rx_queues; i++) {
4603 		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
4604 		if (!err)
4605 			continue;
4606 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
4607 		break;
4608 	}
4609 
4610 	return err;
4611 }
4612 
4613 /**
4614  * ixgbe_free_tx_resources - Free Tx Resources per Queue
4615  * @tx_ring: Tx descriptor ring for a specific queue
4616  *
4617  * Free all transmit software resources
4618  **/
4619 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
4620 {
4621 	ixgbe_clean_tx_ring(tx_ring);
4622 
4623 	vfree(tx_ring->tx_buffer_info);
4624 	tx_ring->tx_buffer_info = NULL;
4625 
4626 	/* if not set, then don't free */
4627 	if (!tx_ring->desc)
4628 		return;
4629 
4630 	dma_free_coherent(tx_ring->dev, tx_ring->size,
4631 			  tx_ring->desc, tx_ring->dma);
4632 
4633 	tx_ring->desc = NULL;
4634 }
4635 
4636 /**
4637  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
4638  * @adapter: board private structure
4639  *
4640  * Free all transmit software resources
4641  **/
4642 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
4643 {
4644 	int i;
4645 
4646 	for (i = 0; i < adapter->num_tx_queues; i++)
4647 		if (adapter->tx_ring[i]->desc)
4648 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
4649 }
4650 
4651 /**
4652  * ixgbe_free_rx_resources - Free Rx Resources
4653  * @rx_ring: ring to clean the resources from
4654  *
4655  * Free all receive software resources
4656  **/
4657 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
4658 {
4659 	ixgbe_clean_rx_ring(rx_ring);
4660 
4661 	vfree(rx_ring->rx_buffer_info);
4662 	rx_ring->rx_buffer_info = NULL;
4663 
4664 	/* if not set, then don't free */
4665 	if (!rx_ring->desc)
4666 		return;
4667 
4668 	dma_free_coherent(rx_ring->dev, rx_ring->size,
4669 			  rx_ring->desc, rx_ring->dma);
4670 
4671 	rx_ring->desc = NULL;
4672 }
4673 
4674 /**
4675  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
4676  * @adapter: board private structure
4677  *
4678  * Free all receive software resources
4679  **/
4680 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
4681 {
4682 	int i;
4683 
4684 	for (i = 0; i < adapter->num_rx_queues; i++)
4685 		if (adapter->rx_ring[i]->desc)
4686 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
4687 }
4688 
4689 /**
4690  * ixgbe_change_mtu - Change the Maximum Transfer Unit
4691  * @netdev: network interface device structure
4692  * @new_mtu: new value for maximum frame size
4693  *
4694  * Returns 0 on success, negative on failure
4695  **/
4696 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
4697 {
4698 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4699 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
4700 
4701 	/* MTU < 68 is an error and causes problems on some kernels */
4702 	if ((new_mtu < 68) || (max_frame > IXGBE_MAX_JUMBO_FRAME_SIZE))
4703 		return -EINVAL;
4704 
4705 	/*
4706 	 * For 82599EB we cannot allow PF to change MTU greater than 1500
4707 	 * in SR-IOV mode as it may cause buffer overruns in guest VFs that
4708 	 * don't allocate and chain buffers correctly.
4709 	 */
4710 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
4711 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
4712 	    (max_frame > MAXIMUM_ETHERNET_VLAN_SIZE))
4713 			return -EINVAL;
4714 
4715 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
4716 
4717 	/* must set new MTU before calling down or up */
4718 	netdev->mtu = new_mtu;
4719 
4720 	if (netif_running(netdev))
4721 		ixgbe_reinit_locked(adapter);
4722 
4723 	return 0;
4724 }
4725 
4726 /**
4727  * ixgbe_open - Called when a network interface is made active
4728  * @netdev: network interface device structure
4729  *
4730  * Returns 0 on success, negative value on failure
4731  *
4732  * The open entry point is called when a network interface is made
4733  * active by the system (IFF_UP).  At this point all resources needed
4734  * for transmit and receive operations are allocated, the interrupt
4735  * handler is registered with the OS, the watchdog timer is started,
4736  * and the stack is notified that the interface is ready.
4737  **/
4738 static int ixgbe_open(struct net_device *netdev)
4739 {
4740 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4741 	int err;
4742 
4743 	/* disallow open during test */
4744 	if (test_bit(__IXGBE_TESTING, &adapter->state))
4745 		return -EBUSY;
4746 
4747 	netif_carrier_off(netdev);
4748 
4749 	/* allocate transmit descriptors */
4750 	err = ixgbe_setup_all_tx_resources(adapter);
4751 	if (err)
4752 		goto err_setup_tx;
4753 
4754 	/* allocate receive descriptors */
4755 	err = ixgbe_setup_all_rx_resources(adapter);
4756 	if (err)
4757 		goto err_setup_rx;
4758 
4759 	ixgbe_configure(adapter);
4760 
4761 	err = ixgbe_request_irq(adapter);
4762 	if (err)
4763 		goto err_req_irq;
4764 
4765 	ixgbe_up_complete(adapter);
4766 
4767 	return 0;
4768 
4769 err_req_irq:
4770 err_setup_rx:
4771 	ixgbe_free_all_rx_resources(adapter);
4772 err_setup_tx:
4773 	ixgbe_free_all_tx_resources(adapter);
4774 	ixgbe_reset(adapter);
4775 
4776 	return err;
4777 }
4778 
4779 /**
4780  * ixgbe_close - Disables a network interface
4781  * @netdev: network interface device structure
4782  *
4783  * Returns 0, this is not allowed to fail
4784  *
4785  * The close entry point is called when an interface is de-activated
4786  * by the OS.  The hardware is still under the drivers control, but
4787  * needs to be disabled.  A global MAC reset is issued to stop the
4788  * hardware, and all transmit and receive resources are freed.
4789  **/
4790 static int ixgbe_close(struct net_device *netdev)
4791 {
4792 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4793 
4794 	ixgbe_down(adapter);
4795 	ixgbe_free_irq(adapter);
4796 
4797 	ixgbe_fdir_filter_exit(adapter);
4798 
4799 	ixgbe_free_all_tx_resources(adapter);
4800 	ixgbe_free_all_rx_resources(adapter);
4801 
4802 	ixgbe_release_hw_control(adapter);
4803 
4804 	return 0;
4805 }
4806 
4807 #ifdef CONFIG_PM
4808 static int ixgbe_resume(struct pci_dev *pdev)
4809 {
4810 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
4811 	struct net_device *netdev = adapter->netdev;
4812 	u32 err;
4813 
4814 	pci_set_power_state(pdev, PCI_D0);
4815 	pci_restore_state(pdev);
4816 	/*
4817 	 * pci_restore_state clears dev->state_saved so call
4818 	 * pci_save_state to restore it.
4819 	 */
4820 	pci_save_state(pdev);
4821 
4822 	err = pci_enable_device_mem(pdev);
4823 	if (err) {
4824 		e_dev_err("Cannot enable PCI device from suspend\n");
4825 		return err;
4826 	}
4827 	pci_set_master(pdev);
4828 
4829 	pci_wake_from_d3(pdev, false);
4830 
4831 	rtnl_lock();
4832 	err = ixgbe_init_interrupt_scheme(adapter);
4833 	rtnl_unlock();
4834 	if (err) {
4835 		e_dev_err("Cannot initialize interrupts for device\n");
4836 		return err;
4837 	}
4838 
4839 	ixgbe_reset(adapter);
4840 
4841 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
4842 
4843 	if (netif_running(netdev)) {
4844 		err = ixgbe_open(netdev);
4845 		if (err)
4846 			return err;
4847 	}
4848 
4849 	netif_device_attach(netdev);
4850 
4851 	return 0;
4852 }
4853 #endif /* CONFIG_PM */
4854 
4855 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
4856 {
4857 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
4858 	struct net_device *netdev = adapter->netdev;
4859 	struct ixgbe_hw *hw = &adapter->hw;
4860 	u32 ctrl, fctrl;
4861 	u32 wufc = adapter->wol;
4862 #ifdef CONFIG_PM
4863 	int retval = 0;
4864 #endif
4865 
4866 	netif_device_detach(netdev);
4867 
4868 	if (netif_running(netdev)) {
4869 		ixgbe_down(adapter);
4870 		ixgbe_free_irq(adapter);
4871 		ixgbe_free_all_tx_resources(adapter);
4872 		ixgbe_free_all_rx_resources(adapter);
4873 	}
4874 
4875 	ixgbe_clear_interrupt_scheme(adapter);
4876 #ifdef CONFIG_DCB
4877 	kfree(adapter->ixgbe_ieee_pfc);
4878 	kfree(adapter->ixgbe_ieee_ets);
4879 #endif
4880 
4881 #ifdef CONFIG_PM
4882 	retval = pci_save_state(pdev);
4883 	if (retval)
4884 		return retval;
4885 
4886 #endif
4887 	if (wufc) {
4888 		ixgbe_set_rx_mode(netdev);
4889 
4890 		/*
4891 		 * enable the optics for both mult-speed fiber and
4892 		 * 82599 SFP+ fiber as we can WoL.
4893 		 */
4894 		if (hw->mac.ops.enable_tx_laser &&
4895 		    (hw->phy.multispeed_fiber ||
4896 		    (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_fiber &&
4897 		     hw->mac.type == ixgbe_mac_82599EB)))
4898 			hw->mac.ops.enable_tx_laser(hw);
4899 
4900 		/* turn on all-multi mode if wake on multicast is enabled */
4901 		if (wufc & IXGBE_WUFC_MC) {
4902 			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4903 			fctrl |= IXGBE_FCTRL_MPE;
4904 			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4905 		}
4906 
4907 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
4908 		ctrl |= IXGBE_CTRL_GIO_DIS;
4909 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
4910 
4911 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
4912 	} else {
4913 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
4914 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
4915 	}
4916 
4917 	switch (hw->mac.type) {
4918 	case ixgbe_mac_82598EB:
4919 		pci_wake_from_d3(pdev, false);
4920 		break;
4921 	case ixgbe_mac_82599EB:
4922 	case ixgbe_mac_X540:
4923 		pci_wake_from_d3(pdev, !!wufc);
4924 		break;
4925 	default:
4926 		break;
4927 	}
4928 
4929 	*enable_wake = !!wufc;
4930 
4931 	ixgbe_release_hw_control(adapter);
4932 
4933 	pci_disable_device(pdev);
4934 
4935 	return 0;
4936 }
4937 
4938 #ifdef CONFIG_PM
4939 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
4940 {
4941 	int retval;
4942 	bool wake;
4943 
4944 	retval = __ixgbe_shutdown(pdev, &wake);
4945 	if (retval)
4946 		return retval;
4947 
4948 	if (wake) {
4949 		pci_prepare_to_sleep(pdev);
4950 	} else {
4951 		pci_wake_from_d3(pdev, false);
4952 		pci_set_power_state(pdev, PCI_D3hot);
4953 	}
4954 
4955 	return 0;
4956 }
4957 #endif /* CONFIG_PM */
4958 
4959 static void ixgbe_shutdown(struct pci_dev *pdev)
4960 {
4961 	bool wake;
4962 
4963 	__ixgbe_shutdown(pdev, &wake);
4964 
4965 	if (system_state == SYSTEM_POWER_OFF) {
4966 		pci_wake_from_d3(pdev, wake);
4967 		pci_set_power_state(pdev, PCI_D3hot);
4968 	}
4969 }
4970 
4971 /**
4972  * ixgbe_update_stats - Update the board statistics counters.
4973  * @adapter: board private structure
4974  **/
4975 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
4976 {
4977 	struct net_device *netdev = adapter->netdev;
4978 	struct ixgbe_hw *hw = &adapter->hw;
4979 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
4980 	u64 total_mpc = 0;
4981 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
4982 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
4983 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
4984 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
4985 #ifdef IXGBE_FCOE
4986 	struct ixgbe_fcoe *fcoe = &adapter->fcoe;
4987 	unsigned int cpu;
4988 	u64 fcoe_noddp_counts_sum = 0, fcoe_noddp_ext_buff_counts_sum = 0;
4989 #endif /* IXGBE_FCOE */
4990 
4991 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
4992 	    test_bit(__IXGBE_RESETTING, &adapter->state))
4993 		return;
4994 
4995 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
4996 		u64 rsc_count = 0;
4997 		u64 rsc_flush = 0;
4998 		for (i = 0; i < 16; i++)
4999 			adapter->hw_rx_no_dma_resources +=
5000 				IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
5001 		for (i = 0; i < adapter->num_rx_queues; i++) {
5002 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
5003 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
5004 		}
5005 		adapter->rsc_total_count = rsc_count;
5006 		adapter->rsc_total_flush = rsc_flush;
5007 	}
5008 
5009 	for (i = 0; i < adapter->num_rx_queues; i++) {
5010 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
5011 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
5012 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
5013 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
5014 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
5015 		bytes += rx_ring->stats.bytes;
5016 		packets += rx_ring->stats.packets;
5017 	}
5018 	adapter->non_eop_descs = non_eop_descs;
5019 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
5020 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
5021 	adapter->hw_csum_rx_error = hw_csum_rx_error;
5022 	netdev->stats.rx_bytes = bytes;
5023 	netdev->stats.rx_packets = packets;
5024 
5025 	bytes = 0;
5026 	packets = 0;
5027 	/* gather some stats to the adapter struct that are per queue */
5028 	for (i = 0; i < adapter->num_tx_queues; i++) {
5029 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
5030 		restart_queue += tx_ring->tx_stats.restart_queue;
5031 		tx_busy += tx_ring->tx_stats.tx_busy;
5032 		bytes += tx_ring->stats.bytes;
5033 		packets += tx_ring->stats.packets;
5034 	}
5035 	adapter->restart_queue = restart_queue;
5036 	adapter->tx_busy = tx_busy;
5037 	netdev->stats.tx_bytes = bytes;
5038 	netdev->stats.tx_packets = packets;
5039 
5040 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
5041 
5042 	/* 8 register reads */
5043 	for (i = 0; i < 8; i++) {
5044 		/* for packet buffers not used, the register should read 0 */
5045 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
5046 		missed_rx += mpc;
5047 		hwstats->mpc[i] += mpc;
5048 		total_mpc += hwstats->mpc[i];
5049 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
5050 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
5051 		switch (hw->mac.type) {
5052 		case ixgbe_mac_82598EB:
5053 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
5054 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
5055 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
5056 			hwstats->pxonrxc[i] +=
5057 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
5058 			break;
5059 		case ixgbe_mac_82599EB:
5060 		case ixgbe_mac_X540:
5061 			hwstats->pxonrxc[i] +=
5062 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
5063 			break;
5064 		default:
5065 			break;
5066 		}
5067 	}
5068 
5069 	/*16 register reads */
5070 	for (i = 0; i < 16; i++) {
5071 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
5072 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
5073 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
5074 		    (hw->mac.type == ixgbe_mac_X540)) {
5075 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
5076 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
5077 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
5078 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
5079 		}
5080 	}
5081 
5082 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
5083 	/* work around hardware counting issue */
5084 	hwstats->gprc -= missed_rx;
5085 
5086 	ixgbe_update_xoff_received(adapter);
5087 
5088 	/* 82598 hardware only has a 32 bit counter in the high register */
5089 	switch (hw->mac.type) {
5090 	case ixgbe_mac_82598EB:
5091 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
5092 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
5093 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
5094 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
5095 		break;
5096 	case ixgbe_mac_X540:
5097 		/* OS2BMC stats are X540 only*/
5098 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
5099 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
5100 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
5101 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
5102 	case ixgbe_mac_82599EB:
5103 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
5104 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
5105 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
5106 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
5107 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
5108 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
5109 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
5110 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
5111 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
5112 #ifdef IXGBE_FCOE
5113 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
5114 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
5115 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
5116 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
5117 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
5118 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
5119 		/* Add up per cpu counters for total ddp aloc fail */
5120 		if (fcoe->pcpu_noddp && fcoe->pcpu_noddp_ext_buff) {
5121 			for_each_possible_cpu(cpu) {
5122 				fcoe_noddp_counts_sum +=
5123 					*per_cpu_ptr(fcoe->pcpu_noddp, cpu);
5124 				fcoe_noddp_ext_buff_counts_sum +=
5125 					*per_cpu_ptr(fcoe->
5126 						pcpu_noddp_ext_buff, cpu);
5127 			}
5128 		}
5129 		hwstats->fcoe_noddp = fcoe_noddp_counts_sum;
5130 		hwstats->fcoe_noddp_ext_buff = fcoe_noddp_ext_buff_counts_sum;
5131 #endif /* IXGBE_FCOE */
5132 		break;
5133 	default:
5134 		break;
5135 	}
5136 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
5137 	hwstats->bprc += bprc;
5138 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
5139 	if (hw->mac.type == ixgbe_mac_82598EB)
5140 		hwstats->mprc -= bprc;
5141 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
5142 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
5143 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
5144 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
5145 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
5146 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
5147 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
5148 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
5149 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
5150 	hwstats->lxontxc += lxon;
5151 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
5152 	hwstats->lxofftxc += lxoff;
5153 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
5154 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
5155 	/*
5156 	 * 82598 errata - tx of flow control packets is included in tx counters
5157 	 */
5158 	xon_off_tot = lxon + lxoff;
5159 	hwstats->gptc -= xon_off_tot;
5160 	hwstats->mptc -= xon_off_tot;
5161 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
5162 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
5163 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
5164 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
5165 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
5166 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
5167 	hwstats->ptc64 -= xon_off_tot;
5168 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
5169 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
5170 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
5171 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
5172 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
5173 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
5174 
5175 	/* Fill out the OS statistics structure */
5176 	netdev->stats.multicast = hwstats->mprc;
5177 
5178 	/* Rx Errors */
5179 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
5180 	netdev->stats.rx_dropped = 0;
5181 	netdev->stats.rx_length_errors = hwstats->rlec;
5182 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
5183 	netdev->stats.rx_missed_errors = total_mpc;
5184 }
5185 
5186 /**
5187  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
5188  * @adapter - pointer to the device adapter structure
5189  **/
5190 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
5191 {
5192 	struct ixgbe_hw *hw = &adapter->hw;
5193 	int i;
5194 
5195 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
5196 		return;
5197 
5198 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
5199 
5200 	/* if interface is down do nothing */
5201 	if (test_bit(__IXGBE_DOWN, &adapter->state))
5202 		return;
5203 
5204 	/* do nothing if we are not using signature filters */
5205 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
5206 		return;
5207 
5208 	adapter->fdir_overflow++;
5209 
5210 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
5211 		for (i = 0; i < adapter->num_tx_queues; i++)
5212 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
5213 			        &(adapter->tx_ring[i]->state));
5214 		/* re-enable flow director interrupts */
5215 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
5216 	} else {
5217 		e_err(probe, "failed to finish FDIR re-initialization, "
5218 		      "ignored adding FDIR ATR filters\n");
5219 	}
5220 }
5221 
5222 /**
5223  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
5224  * @adapter - pointer to the device adapter structure
5225  *
5226  * This function serves two purposes.  First it strobes the interrupt lines
5227  * in order to make certain interrupts are occurring.  Secondly it sets the
5228  * bits needed to check for TX hangs.  As a result we should immediately
5229  * determine if a hang has occurred.
5230  */
5231 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
5232 {
5233 	struct ixgbe_hw *hw = &adapter->hw;
5234 	u64 eics = 0;
5235 	int i;
5236 
5237 	/* If we're down or resetting, just bail */
5238 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5239 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5240 		return;
5241 
5242 	/* Force detection of hung controller */
5243 	if (netif_carrier_ok(adapter->netdev)) {
5244 		for (i = 0; i < adapter->num_tx_queues; i++)
5245 			set_check_for_tx_hang(adapter->tx_ring[i]);
5246 	}
5247 
5248 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
5249 		/*
5250 		 * for legacy and MSI interrupts don't set any bits
5251 		 * that are enabled for EIAM, because this operation
5252 		 * would set *both* EIMS and EICS for any bit in EIAM
5253 		 */
5254 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
5255 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
5256 	} else {
5257 		/* get one bit for every active tx/rx interrupt vector */
5258 		for (i = 0; i < adapter->num_msix_vectors - NON_Q_VECTORS; i++) {
5259 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
5260 			if (qv->rx.ring || qv->tx.ring)
5261 				eics |= ((u64)1 << i);
5262 		}
5263 	}
5264 
5265 	/* Cause software interrupt to ensure rings are cleaned */
5266 	ixgbe_irq_rearm_queues(adapter, eics);
5267 
5268 }
5269 
5270 /**
5271  * ixgbe_watchdog_update_link - update the link status
5272  * @adapter - pointer to the device adapter structure
5273  * @link_speed - pointer to a u32 to store the link_speed
5274  **/
5275 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
5276 {
5277 	struct ixgbe_hw *hw = &adapter->hw;
5278 	u32 link_speed = adapter->link_speed;
5279 	bool link_up = adapter->link_up;
5280 	int i;
5281 
5282 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
5283 		return;
5284 
5285 	if (hw->mac.ops.check_link) {
5286 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
5287 	} else {
5288 		/* always assume link is up, if no check link function */
5289 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
5290 		link_up = true;
5291 	}
5292 	if (link_up) {
5293 		if (adapter->flags & IXGBE_FLAG_DCB_ENABLED) {
5294 			for (i = 0; i < MAX_TRAFFIC_CLASS; i++)
5295 				hw->mac.ops.fc_enable(hw, i);
5296 		} else {
5297 			hw->mac.ops.fc_enable(hw, 0);
5298 		}
5299 	}
5300 
5301 	if (link_up ||
5302 	    time_after(jiffies, (adapter->link_check_timeout +
5303 				 IXGBE_TRY_LINK_TIMEOUT))) {
5304 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5305 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
5306 		IXGBE_WRITE_FLUSH(hw);
5307 	}
5308 
5309 	adapter->link_up = link_up;
5310 	adapter->link_speed = link_speed;
5311 }
5312 
5313 /**
5314  * ixgbe_watchdog_link_is_up - update netif_carrier status and
5315  *                             print link up message
5316  * @adapter - pointer to the device adapter structure
5317  **/
5318 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
5319 {
5320 	struct net_device *netdev = adapter->netdev;
5321 	struct ixgbe_hw *hw = &adapter->hw;
5322 	u32 link_speed = adapter->link_speed;
5323 	bool flow_rx, flow_tx;
5324 
5325 	/* only continue if link was previously down */
5326 	if (netif_carrier_ok(netdev))
5327 		return;
5328 
5329 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
5330 
5331 	switch (hw->mac.type) {
5332 	case ixgbe_mac_82598EB: {
5333 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
5334 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
5335 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
5336 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
5337 	}
5338 		break;
5339 	case ixgbe_mac_X540:
5340 	case ixgbe_mac_82599EB: {
5341 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
5342 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
5343 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
5344 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
5345 	}
5346 		break;
5347 	default:
5348 		flow_tx = false;
5349 		flow_rx = false;
5350 		break;
5351 	}
5352 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n",
5353 	       (link_speed == IXGBE_LINK_SPEED_10GB_FULL ?
5354 	       "10 Gbps" :
5355 	       (link_speed == IXGBE_LINK_SPEED_1GB_FULL ?
5356 	       "1 Gbps" :
5357 	       (link_speed == IXGBE_LINK_SPEED_100_FULL ?
5358 	       "100 Mbps" :
5359 	       "unknown speed"))),
5360 	       ((flow_rx && flow_tx) ? "RX/TX" :
5361 	       (flow_rx ? "RX" :
5362 	       (flow_tx ? "TX" : "None"))));
5363 
5364 	netif_carrier_on(netdev);
5365 	ixgbe_check_vf_rate_limit(adapter);
5366 }
5367 
5368 /**
5369  * ixgbe_watchdog_link_is_down - update netif_carrier status and
5370  *                               print link down message
5371  * @adapter - pointer to the adapter structure
5372  **/
5373 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
5374 {
5375 	struct net_device *netdev = adapter->netdev;
5376 	struct ixgbe_hw *hw = &adapter->hw;
5377 
5378 	adapter->link_up = false;
5379 	adapter->link_speed = 0;
5380 
5381 	/* only continue if link was up previously */
5382 	if (!netif_carrier_ok(netdev))
5383 		return;
5384 
5385 	/* poll for SFP+ cable when link is down */
5386 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
5387 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5388 
5389 	e_info(drv, "NIC Link is Down\n");
5390 	netif_carrier_off(netdev);
5391 }
5392 
5393 /**
5394  * ixgbe_watchdog_flush_tx - flush queues on link down
5395  * @adapter - pointer to the device adapter structure
5396  **/
5397 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
5398 {
5399 	int i;
5400 	int some_tx_pending = 0;
5401 
5402 	if (!netif_carrier_ok(adapter->netdev)) {
5403 		for (i = 0; i < adapter->num_tx_queues; i++) {
5404 			struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
5405 			if (tx_ring->next_to_use != tx_ring->next_to_clean) {
5406 				some_tx_pending = 1;
5407 				break;
5408 			}
5409 		}
5410 
5411 		if (some_tx_pending) {
5412 			/* We've lost link, so the controller stops DMA,
5413 			 * but we've got queued Tx work that's never going
5414 			 * to get done, so reset controller to flush Tx.
5415 			 * (Do the reset outside of interrupt context).
5416 			 */
5417 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
5418 		}
5419 	}
5420 }
5421 
5422 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
5423 {
5424 	u32 ssvpc;
5425 
5426 	/* Do not perform spoof check for 82598 */
5427 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5428 		return;
5429 
5430 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
5431 
5432 	/*
5433 	 * ssvpc register is cleared on read, if zero then no
5434 	 * spoofed packets in the last interval.
5435 	 */
5436 	if (!ssvpc)
5437 		return;
5438 
5439 	e_warn(drv, "%d Spoofed packets detected\n", ssvpc);
5440 }
5441 
5442 /**
5443  * ixgbe_watchdog_subtask - check and bring link up
5444  * @adapter - pointer to the device adapter structure
5445  **/
5446 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
5447 {
5448 	/* if interface is down do nothing */
5449 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5450 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5451 		return;
5452 
5453 	ixgbe_watchdog_update_link(adapter);
5454 
5455 	if (adapter->link_up)
5456 		ixgbe_watchdog_link_is_up(adapter);
5457 	else
5458 		ixgbe_watchdog_link_is_down(adapter);
5459 
5460 	ixgbe_spoof_check(adapter);
5461 	ixgbe_update_stats(adapter);
5462 
5463 	ixgbe_watchdog_flush_tx(adapter);
5464 }
5465 
5466 /**
5467  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
5468  * @adapter - the ixgbe adapter structure
5469  **/
5470 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
5471 {
5472 	struct ixgbe_hw *hw = &adapter->hw;
5473 	s32 err;
5474 
5475 	/* not searching for SFP so there is nothing to do here */
5476 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
5477 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
5478 		return;
5479 
5480 	/* someone else is in init, wait until next service event */
5481 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5482 		return;
5483 
5484 	err = hw->phy.ops.identify_sfp(hw);
5485 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
5486 		goto sfp_out;
5487 
5488 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
5489 		/* If no cable is present, then we need to reset
5490 		 * the next time we find a good cable. */
5491 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5492 	}
5493 
5494 	/* exit on error */
5495 	if (err)
5496 		goto sfp_out;
5497 
5498 	/* exit if reset not needed */
5499 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
5500 		goto sfp_out;
5501 
5502 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
5503 
5504 	/*
5505 	 * A module may be identified correctly, but the EEPROM may not have
5506 	 * support for that module.  setup_sfp() will fail in that case, so
5507 	 * we should not allow that module to load.
5508 	 */
5509 	if (hw->mac.type == ixgbe_mac_82598EB)
5510 		err = hw->phy.ops.reset(hw);
5511 	else
5512 		err = hw->mac.ops.setup_sfp(hw);
5513 
5514 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
5515 		goto sfp_out;
5516 
5517 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
5518 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
5519 
5520 sfp_out:
5521 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5522 
5523 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
5524 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
5525 		e_dev_err("failed to initialize because an unsupported "
5526 			  "SFP+ module type was detected.\n");
5527 		e_dev_err("Reload the driver after installing a "
5528 			  "supported module.\n");
5529 		unregister_netdev(adapter->netdev);
5530 	}
5531 }
5532 
5533 /**
5534  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
5535  * @adapter - the ixgbe adapter structure
5536  **/
5537 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
5538 {
5539 	struct ixgbe_hw *hw = &adapter->hw;
5540 	u32 autoneg;
5541 	bool negotiation;
5542 
5543 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
5544 		return;
5545 
5546 	/* someone else is in init, wait until next service event */
5547 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5548 		return;
5549 
5550 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5551 
5552 	autoneg = hw->phy.autoneg_advertised;
5553 	if ((!autoneg) && (hw->mac.ops.get_link_capabilities))
5554 		hw->mac.ops.get_link_capabilities(hw, &autoneg, &negotiation);
5555 	if (hw->mac.ops.setup_link)
5556 		hw->mac.ops.setup_link(hw, autoneg, negotiation, true);
5557 
5558 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5559 	adapter->link_check_timeout = jiffies;
5560 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5561 }
5562 
5563 #ifdef CONFIG_PCI_IOV
5564 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
5565 {
5566 	int vf;
5567 	struct ixgbe_hw *hw = &adapter->hw;
5568 	struct net_device *netdev = adapter->netdev;
5569 	u32 gpc;
5570 	u32 ciaa, ciad;
5571 
5572 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
5573 	if (gpc) /* If incrementing then no need for the check below */
5574 		return;
5575 	/*
5576 	 * Check to see if a bad DMA write target from an errant or
5577 	 * malicious VF has caused a PCIe error.  If so then we can
5578 	 * issue a VFLR to the offending VF(s) and then resume without
5579 	 * requesting a full slot reset.
5580 	 */
5581 
5582 	for (vf = 0; vf < adapter->num_vfs; vf++) {
5583 		ciaa = (vf << 16) | 0x80000000;
5584 		/* 32 bit read so align, we really want status at offset 6 */
5585 		ciaa |= PCI_COMMAND;
5586 		IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5587 		ciad = IXGBE_READ_REG(hw, IXGBE_CIAD_82599);
5588 		ciaa &= 0x7FFFFFFF;
5589 		/* disable debug mode asap after reading data */
5590 		IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5591 		/* Get the upper 16 bits which will be the PCI status reg */
5592 		ciad >>= 16;
5593 		if (ciad & PCI_STATUS_REC_MASTER_ABORT) {
5594 			netdev_err(netdev, "VF %d Hung DMA\n", vf);
5595 			/* Issue VFLR */
5596 			ciaa = (vf << 16) | 0x80000000;
5597 			ciaa |= 0xA8;
5598 			IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5599 			ciad = 0x00008000;  /* VFLR */
5600 			IXGBE_WRITE_REG(hw, IXGBE_CIAD_82599, ciad);
5601 			ciaa &= 0x7FFFFFFF;
5602 			IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
5603 		}
5604 	}
5605 }
5606 
5607 #endif
5608 /**
5609  * ixgbe_service_timer - Timer Call-back
5610  * @data: pointer to adapter cast into an unsigned long
5611  **/
5612 static void ixgbe_service_timer(unsigned long data)
5613 {
5614 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
5615 	unsigned long next_event_offset;
5616 	bool ready = true;
5617 
5618 	/* poll faster when waiting for link */
5619 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
5620 		next_event_offset = HZ / 10;
5621 	else
5622 		next_event_offset = HZ * 2;
5623 
5624 #ifdef CONFIG_PCI_IOV
5625 	/*
5626 	 * don't bother with SR-IOV VF DMA hang check if there are
5627 	 * no VFs or the link is down
5628 	 */
5629 	if (!adapter->num_vfs ||
5630 	    (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
5631 		goto normal_timer_service;
5632 
5633 	/* If we have VFs allocated then we must check for DMA hangs */
5634 	ixgbe_check_for_bad_vf(adapter);
5635 	next_event_offset = HZ / 50;
5636 	adapter->timer_event_accumulator++;
5637 
5638 	if (adapter->timer_event_accumulator >= 100)
5639 		adapter->timer_event_accumulator = 0;
5640 	else
5641 		ready = false;
5642 
5643 normal_timer_service:
5644 #endif
5645 	/* Reset the timer */
5646 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
5647 
5648 	if (ready)
5649 		ixgbe_service_event_schedule(adapter);
5650 }
5651 
5652 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
5653 {
5654 	if (!(adapter->flags2 & IXGBE_FLAG2_RESET_REQUESTED))
5655 		return;
5656 
5657 	adapter->flags2 &= ~IXGBE_FLAG2_RESET_REQUESTED;
5658 
5659 	/* If we're already down or resetting, just bail */
5660 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5661 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5662 		return;
5663 
5664 	ixgbe_dump(adapter);
5665 	netdev_err(adapter->netdev, "Reset adapter\n");
5666 	adapter->tx_timeout_count++;
5667 
5668 	ixgbe_reinit_locked(adapter);
5669 }
5670 
5671 /**
5672  * ixgbe_service_task - manages and runs subtasks
5673  * @work: pointer to work_struct containing our data
5674  **/
5675 static void ixgbe_service_task(struct work_struct *work)
5676 {
5677 	struct ixgbe_adapter *adapter = container_of(work,
5678 						     struct ixgbe_adapter,
5679 						     service_task);
5680 
5681 	ixgbe_reset_subtask(adapter);
5682 	ixgbe_sfp_detection_subtask(adapter);
5683 	ixgbe_sfp_link_config_subtask(adapter);
5684 	ixgbe_check_overtemp_subtask(adapter);
5685 	ixgbe_watchdog_subtask(adapter);
5686 	ixgbe_fdir_reinit_subtask(adapter);
5687 	ixgbe_check_hang_subtask(adapter);
5688 
5689 	ixgbe_service_event_complete(adapter);
5690 }
5691 
5692 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
5693 		     struct ixgbe_tx_buffer *first,
5694 		     u8 *hdr_len)
5695 {
5696 	struct sk_buff *skb = first->skb;
5697 	u32 vlan_macip_lens, type_tucmd;
5698 	u32 mss_l4len_idx, l4len;
5699 
5700 	if (!skb_is_gso(skb))
5701 		return 0;
5702 
5703 	if (skb_header_cloned(skb)) {
5704 		int err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
5705 		if (err)
5706 			return err;
5707 	}
5708 
5709 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
5710 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
5711 
5712 	if (first->protocol == __constant_htons(ETH_P_IP)) {
5713 		struct iphdr *iph = ip_hdr(skb);
5714 		iph->tot_len = 0;
5715 		iph->check = 0;
5716 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
5717 							 iph->daddr, 0,
5718 							 IPPROTO_TCP,
5719 							 0);
5720 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
5721 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
5722 				   IXGBE_TX_FLAGS_CSUM |
5723 				   IXGBE_TX_FLAGS_IPV4;
5724 	} else if (skb_is_gso_v6(skb)) {
5725 		ipv6_hdr(skb)->payload_len = 0;
5726 		tcp_hdr(skb)->check =
5727 		    ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
5728 				     &ipv6_hdr(skb)->daddr,
5729 				     0, IPPROTO_TCP, 0);
5730 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
5731 				   IXGBE_TX_FLAGS_CSUM;
5732 	}
5733 
5734 	/* compute header lengths */
5735 	l4len = tcp_hdrlen(skb);
5736 	*hdr_len = skb_transport_offset(skb) + l4len;
5737 
5738 	/* update gso size and bytecount with header size */
5739 	first->gso_segs = skb_shinfo(skb)->gso_segs;
5740 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
5741 
5742 	/* mss_l4len_id: use 1 as index for TSO */
5743 	mss_l4len_idx = l4len << IXGBE_ADVTXD_L4LEN_SHIFT;
5744 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
5745 	mss_l4len_idx |= 1 << IXGBE_ADVTXD_IDX_SHIFT;
5746 
5747 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
5748 	vlan_macip_lens = skb_network_header_len(skb);
5749 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
5750 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
5751 
5752 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
5753 			  mss_l4len_idx);
5754 
5755 	return 1;
5756 }
5757 
5758 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
5759 			  struct ixgbe_tx_buffer *first)
5760 {
5761 	struct sk_buff *skb = first->skb;
5762 	u32 vlan_macip_lens = 0;
5763 	u32 mss_l4len_idx = 0;
5764 	u32 type_tucmd = 0;
5765 
5766 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
5767 		if (!(first->tx_flags & IXGBE_TX_FLAGS_HW_VLAN) &&
5768 		    !(first->tx_flags & IXGBE_TX_FLAGS_TXSW))
5769 			return;
5770 	} else {
5771 		u8 l4_hdr = 0;
5772 		switch (first->protocol) {
5773 		case __constant_htons(ETH_P_IP):
5774 			vlan_macip_lens |= skb_network_header_len(skb);
5775 			type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
5776 			l4_hdr = ip_hdr(skb)->protocol;
5777 			break;
5778 		case __constant_htons(ETH_P_IPV6):
5779 			vlan_macip_lens |= skb_network_header_len(skb);
5780 			l4_hdr = ipv6_hdr(skb)->nexthdr;
5781 			break;
5782 		default:
5783 			if (unlikely(net_ratelimit())) {
5784 				dev_warn(tx_ring->dev,
5785 				 "partial checksum but proto=%x!\n",
5786 				 first->protocol);
5787 			}
5788 			break;
5789 		}
5790 
5791 		switch (l4_hdr) {
5792 		case IPPROTO_TCP:
5793 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_TCP;
5794 			mss_l4len_idx = tcp_hdrlen(skb) <<
5795 					IXGBE_ADVTXD_L4LEN_SHIFT;
5796 			break;
5797 		case IPPROTO_SCTP:
5798 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_SCTP;
5799 			mss_l4len_idx = sizeof(struct sctphdr) <<
5800 					IXGBE_ADVTXD_L4LEN_SHIFT;
5801 			break;
5802 		case IPPROTO_UDP:
5803 			mss_l4len_idx = sizeof(struct udphdr) <<
5804 					IXGBE_ADVTXD_L4LEN_SHIFT;
5805 			break;
5806 		default:
5807 			if (unlikely(net_ratelimit())) {
5808 				dev_warn(tx_ring->dev,
5809 				 "partial checksum but l4 proto=%x!\n",
5810 				 l4_hdr);
5811 			}
5812 			break;
5813 		}
5814 
5815 		/* update TX checksum flag */
5816 		first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
5817 	}
5818 
5819 	/* vlan_macip_lens: MACLEN, VLAN tag */
5820 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
5821 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
5822 
5823 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0,
5824 			  type_tucmd, mss_l4len_idx);
5825 }
5826 
5827 static __le32 ixgbe_tx_cmd_type(u32 tx_flags)
5828 {
5829 	/* set type for advanced descriptor with frame checksum insertion */
5830 	__le32 cmd_type = cpu_to_le32(IXGBE_ADVTXD_DTYP_DATA |
5831 				      IXGBE_ADVTXD_DCMD_IFCS |
5832 				      IXGBE_ADVTXD_DCMD_DEXT);
5833 
5834 	/* set HW vlan bit if vlan is present */
5835 	if (tx_flags & IXGBE_TX_FLAGS_HW_VLAN)
5836 		cmd_type |= cpu_to_le32(IXGBE_ADVTXD_DCMD_VLE);
5837 
5838 	/* set segmentation enable bits for TSO/FSO */
5839 #ifdef IXGBE_FCOE
5840 	if (tx_flags & (IXGBE_TX_FLAGS_TSO | IXGBE_TX_FLAGS_FSO))
5841 #else
5842 	if (tx_flags & IXGBE_TX_FLAGS_TSO)
5843 #endif
5844 		cmd_type |= cpu_to_le32(IXGBE_ADVTXD_DCMD_TSE);
5845 
5846 	return cmd_type;
5847 }
5848 
5849 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
5850 				   u32 tx_flags, unsigned int paylen)
5851 {
5852 	__le32 olinfo_status = cpu_to_le32(paylen << IXGBE_ADVTXD_PAYLEN_SHIFT);
5853 
5854 	/* enable L4 checksum for TSO and TX checksum offload */
5855 	if (tx_flags & IXGBE_TX_FLAGS_CSUM)
5856 		olinfo_status |= cpu_to_le32(IXGBE_ADVTXD_POPTS_TXSM);
5857 
5858 	/* enble IPv4 checksum for TSO */
5859 	if (tx_flags & IXGBE_TX_FLAGS_IPV4)
5860 		olinfo_status |= cpu_to_le32(IXGBE_ADVTXD_POPTS_IXSM);
5861 
5862 	/* use index 1 context for TSO/FSO/FCOE */
5863 #ifdef IXGBE_FCOE
5864 	if (tx_flags & (IXGBE_TX_FLAGS_TSO | IXGBE_TX_FLAGS_FCOE))
5865 #else
5866 	if (tx_flags & IXGBE_TX_FLAGS_TSO)
5867 #endif
5868 		olinfo_status |= cpu_to_le32(1 << IXGBE_ADVTXD_IDX_SHIFT);
5869 
5870 	/*
5871 	 * Check Context must be set if Tx switch is enabled, which it
5872 	 * always is for case where virtual functions are running
5873 	 */
5874 #ifdef IXGBE_FCOE
5875 	if (tx_flags & (IXGBE_TX_FLAGS_TXSW | IXGBE_TX_FLAGS_FCOE))
5876 #else
5877 	if (tx_flags & IXGBE_TX_FLAGS_TXSW)
5878 #endif
5879 		olinfo_status |= cpu_to_le32(IXGBE_ADVTXD_CC);
5880 
5881 	tx_desc->read.olinfo_status = olinfo_status;
5882 }
5883 
5884 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
5885 		       IXGBE_TXD_CMD_RS)
5886 
5887 static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
5888 			 struct ixgbe_tx_buffer *first,
5889 			 const u8 hdr_len)
5890 {
5891 	dma_addr_t dma;
5892 	struct sk_buff *skb = first->skb;
5893 	struct ixgbe_tx_buffer *tx_buffer;
5894 	union ixgbe_adv_tx_desc *tx_desc;
5895 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
5896 	unsigned int data_len = skb->data_len;
5897 	unsigned int size = skb_headlen(skb);
5898 	unsigned int paylen = skb->len - hdr_len;
5899 	u32 tx_flags = first->tx_flags;
5900 	__le32 cmd_type;
5901 	u16 i = tx_ring->next_to_use;
5902 
5903 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
5904 
5905 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, paylen);
5906 	cmd_type = ixgbe_tx_cmd_type(tx_flags);
5907 
5908 #ifdef IXGBE_FCOE
5909 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
5910 		if (data_len < sizeof(struct fcoe_crc_eof)) {
5911 			size -= sizeof(struct fcoe_crc_eof) - data_len;
5912 			data_len = 0;
5913 		} else {
5914 			data_len -= sizeof(struct fcoe_crc_eof);
5915 		}
5916 	}
5917 
5918 #endif
5919 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
5920 	if (dma_mapping_error(tx_ring->dev, dma))
5921 		goto dma_error;
5922 
5923 	/* record length, and DMA address */
5924 	dma_unmap_len_set(first, len, size);
5925 	dma_unmap_addr_set(first, dma, dma);
5926 
5927 	tx_desc->read.buffer_addr = cpu_to_le64(dma);
5928 
5929 	for (;;) {
5930 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
5931 			tx_desc->read.cmd_type_len =
5932 				cmd_type | cpu_to_le32(IXGBE_MAX_DATA_PER_TXD);
5933 
5934 			i++;
5935 			tx_desc++;
5936 			if (i == tx_ring->count) {
5937 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
5938 				i = 0;
5939 			}
5940 
5941 			dma += IXGBE_MAX_DATA_PER_TXD;
5942 			size -= IXGBE_MAX_DATA_PER_TXD;
5943 
5944 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
5945 			tx_desc->read.olinfo_status = 0;
5946 		}
5947 
5948 		if (likely(!data_len))
5949 			break;
5950 
5951 		if (unlikely(skb->no_fcs))
5952 			cmd_type &= ~(cpu_to_le32(IXGBE_ADVTXD_DCMD_IFCS));
5953 		tx_desc->read.cmd_type_len = cmd_type | cpu_to_le32(size);
5954 
5955 		i++;
5956 		tx_desc++;
5957 		if (i == tx_ring->count) {
5958 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
5959 			i = 0;
5960 		}
5961 
5962 #ifdef IXGBE_FCOE
5963 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
5964 #else
5965 		size = skb_frag_size(frag);
5966 #endif
5967 		data_len -= size;
5968 
5969 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
5970 				       DMA_TO_DEVICE);
5971 		if (dma_mapping_error(tx_ring->dev, dma))
5972 			goto dma_error;
5973 
5974 		tx_buffer = &tx_ring->tx_buffer_info[i];
5975 		dma_unmap_len_set(tx_buffer, len, size);
5976 		dma_unmap_addr_set(tx_buffer, dma, dma);
5977 
5978 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
5979 		tx_desc->read.olinfo_status = 0;
5980 
5981 		frag++;
5982 	}
5983 
5984 	/* write last descriptor with RS and EOP bits */
5985 	cmd_type |= cpu_to_le32(size) | cpu_to_le32(IXGBE_TXD_CMD);
5986 	tx_desc->read.cmd_type_len = cmd_type;
5987 
5988 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
5989 
5990 	/* set the timestamp */
5991 	first->time_stamp = jiffies;
5992 
5993 	/*
5994 	 * Force memory writes to complete before letting h/w know there
5995 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
5996 	 * memory model archs, such as IA-64).
5997 	 *
5998 	 * We also need this memory barrier to make certain all of the
5999 	 * status bits have been updated before next_to_watch is written.
6000 	 */
6001 	wmb();
6002 
6003 	/* set next_to_watch value indicating a packet is present */
6004 	first->next_to_watch = tx_desc;
6005 
6006 	i++;
6007 	if (i == tx_ring->count)
6008 		i = 0;
6009 
6010 	tx_ring->next_to_use = i;
6011 
6012 	/* notify HW of packet */
6013 	writel(i, tx_ring->tail);
6014 
6015 	return;
6016 dma_error:
6017 	dev_err(tx_ring->dev, "TX DMA map failed\n");
6018 
6019 	/* clear dma mappings for failed tx_buffer_info map */
6020 	for (;;) {
6021 		tx_buffer = &tx_ring->tx_buffer_info[i];
6022 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer);
6023 		if (tx_buffer == first)
6024 			break;
6025 		if (i == 0)
6026 			i = tx_ring->count;
6027 		i--;
6028 	}
6029 
6030 	tx_ring->next_to_use = i;
6031 }
6032 
6033 static void ixgbe_atr(struct ixgbe_ring *ring,
6034 		      struct ixgbe_tx_buffer *first)
6035 {
6036 	struct ixgbe_q_vector *q_vector = ring->q_vector;
6037 	union ixgbe_atr_hash_dword input = { .dword = 0 };
6038 	union ixgbe_atr_hash_dword common = { .dword = 0 };
6039 	union {
6040 		unsigned char *network;
6041 		struct iphdr *ipv4;
6042 		struct ipv6hdr *ipv6;
6043 	} hdr;
6044 	struct tcphdr *th;
6045 	__be16 vlan_id;
6046 
6047 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
6048 	if (!q_vector)
6049 		return;
6050 
6051 	/* do nothing if sampling is disabled */
6052 	if (!ring->atr_sample_rate)
6053 		return;
6054 
6055 	ring->atr_count++;
6056 
6057 	/* snag network header to get L4 type and address */
6058 	hdr.network = skb_network_header(first->skb);
6059 
6060 	/* Currently only IPv4/IPv6 with TCP is supported */
6061 	if ((first->protocol != __constant_htons(ETH_P_IPV6) ||
6062 	     hdr.ipv6->nexthdr != IPPROTO_TCP) &&
6063 	    (first->protocol != __constant_htons(ETH_P_IP) ||
6064 	     hdr.ipv4->protocol != IPPROTO_TCP))
6065 		return;
6066 
6067 	th = tcp_hdr(first->skb);
6068 
6069 	/* skip this packet since it is invalid or the socket is closing */
6070 	if (!th || th->fin)
6071 		return;
6072 
6073 	/* sample on all syn packets or once every atr sample count */
6074 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
6075 		return;
6076 
6077 	/* reset sample count */
6078 	ring->atr_count = 0;
6079 
6080 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
6081 
6082 	/*
6083 	 * src and dst are inverted, think how the receiver sees them
6084 	 *
6085 	 * The input is broken into two sections, a non-compressed section
6086 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
6087 	 * is XORed together and stored in the compressed dword.
6088 	 */
6089 	input.formatted.vlan_id = vlan_id;
6090 
6091 	/*
6092 	 * since src port and flex bytes occupy the same word XOR them together
6093 	 * and write the value to source port portion of compressed dword
6094 	 */
6095 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
6096 		common.port.src ^= th->dest ^ __constant_htons(ETH_P_8021Q);
6097 	else
6098 		common.port.src ^= th->dest ^ first->protocol;
6099 	common.port.dst ^= th->source;
6100 
6101 	if (first->protocol == __constant_htons(ETH_P_IP)) {
6102 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
6103 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
6104 	} else {
6105 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
6106 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
6107 			     hdr.ipv6->saddr.s6_addr32[1] ^
6108 			     hdr.ipv6->saddr.s6_addr32[2] ^
6109 			     hdr.ipv6->saddr.s6_addr32[3] ^
6110 			     hdr.ipv6->daddr.s6_addr32[0] ^
6111 			     hdr.ipv6->daddr.s6_addr32[1] ^
6112 			     hdr.ipv6->daddr.s6_addr32[2] ^
6113 			     hdr.ipv6->daddr.s6_addr32[3];
6114 	}
6115 
6116 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
6117 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
6118 					      input, common, ring->queue_index);
6119 }
6120 
6121 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
6122 {
6123 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
6124 	/* Herbert's original patch had:
6125 	 *  smp_mb__after_netif_stop_queue();
6126 	 * but since that doesn't exist yet, just open code it. */
6127 	smp_mb();
6128 
6129 	/* We need to check again in a case another CPU has just
6130 	 * made room available. */
6131 	if (likely(ixgbe_desc_unused(tx_ring) < size))
6132 		return -EBUSY;
6133 
6134 	/* A reprieve! - use start_queue because it doesn't call schedule */
6135 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
6136 	++tx_ring->tx_stats.restart_queue;
6137 	return 0;
6138 }
6139 
6140 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
6141 {
6142 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
6143 		return 0;
6144 	return __ixgbe_maybe_stop_tx(tx_ring, size);
6145 }
6146 
6147 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb)
6148 {
6149 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6150 	int txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
6151 					       smp_processor_id();
6152 #ifdef IXGBE_FCOE
6153 	__be16 protocol = vlan_get_protocol(skb);
6154 
6155 	if (((protocol == htons(ETH_P_FCOE)) ||
6156 	    (protocol == htons(ETH_P_FIP))) &&
6157 	    (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)) {
6158 		txq &= (adapter->ring_feature[RING_F_FCOE].indices - 1);
6159 		txq += adapter->ring_feature[RING_F_FCOE].mask;
6160 		return txq;
6161 	}
6162 #endif
6163 
6164 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
6165 		while (unlikely(txq >= dev->real_num_tx_queues))
6166 			txq -= dev->real_num_tx_queues;
6167 		return txq;
6168 	}
6169 
6170 	return skb_tx_hash(dev, skb);
6171 }
6172 
6173 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
6174 			  struct ixgbe_adapter *adapter,
6175 			  struct ixgbe_ring *tx_ring)
6176 {
6177 	struct ixgbe_tx_buffer *first;
6178 	int tso;
6179 	u32 tx_flags = 0;
6180 #if PAGE_SIZE > IXGBE_MAX_DATA_PER_TXD
6181 	unsigned short f;
6182 #endif
6183 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
6184 	__be16 protocol = skb->protocol;
6185 	u8 hdr_len = 0;
6186 
6187 	/*
6188 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
6189 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
6190 	 *       + 2 desc gap to keep tail from touching head,
6191 	 *       + 1 desc for context descriptor,
6192 	 * otherwise try next time
6193 	 */
6194 #if PAGE_SIZE > IXGBE_MAX_DATA_PER_TXD
6195 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
6196 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
6197 #else
6198 	count += skb_shinfo(skb)->nr_frags;
6199 #endif
6200 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
6201 		tx_ring->tx_stats.tx_busy++;
6202 		return NETDEV_TX_BUSY;
6203 	}
6204 
6205 	/* record the location of the first descriptor for this packet */
6206 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
6207 	first->skb = skb;
6208 	first->bytecount = skb->len;
6209 	first->gso_segs = 1;
6210 
6211 	/* if we have a HW VLAN tag being added default to the HW one */
6212 	if (vlan_tx_tag_present(skb)) {
6213 		tx_flags |= vlan_tx_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
6214 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
6215 	/* else if it is a SW VLAN check the next protocol and store the tag */
6216 	} else if (protocol == __constant_htons(ETH_P_8021Q)) {
6217 		struct vlan_hdr *vhdr, _vhdr;
6218 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
6219 		if (!vhdr)
6220 			goto out_drop;
6221 
6222 		protocol = vhdr->h_vlan_encapsulated_proto;
6223 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
6224 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
6225 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
6226 	}
6227 
6228 #ifdef CONFIG_PCI_IOV
6229 	/*
6230 	 * Use the l2switch_enable flag - would be false if the DMA
6231 	 * Tx switch had been disabled.
6232 	 */
6233 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
6234 		tx_flags |= IXGBE_TX_FLAGS_TXSW;
6235 
6236 #endif
6237 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
6238 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
6239 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
6240 	     (skb->priority != TC_PRIO_CONTROL))) {
6241 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
6242 		tx_flags |= (skb->priority & 0x7) <<
6243 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
6244 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
6245 			struct vlan_ethhdr *vhdr;
6246 			if (skb_header_cloned(skb) &&
6247 			    pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
6248 				goto out_drop;
6249 			vhdr = (struct vlan_ethhdr *)skb->data;
6250 			vhdr->h_vlan_TCI = htons(tx_flags >>
6251 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
6252 		} else {
6253 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
6254 		}
6255 	}
6256 
6257 	/* record initial flags and protocol */
6258 	first->tx_flags = tx_flags;
6259 	first->protocol = protocol;
6260 
6261 #ifdef IXGBE_FCOE
6262 	/* setup tx offload for FCoE */
6263 	if ((protocol == __constant_htons(ETH_P_FCOE)) &&
6264 	    (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)) {
6265 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
6266 		if (tso < 0)
6267 			goto out_drop;
6268 
6269 		goto xmit_fcoe;
6270 	}
6271 
6272 #endif /* IXGBE_FCOE */
6273 	tso = ixgbe_tso(tx_ring, first, &hdr_len);
6274 	if (tso < 0)
6275 		goto out_drop;
6276 	else if (!tso)
6277 		ixgbe_tx_csum(tx_ring, first);
6278 
6279 	/* add the ATR filter if ATR is on */
6280 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
6281 		ixgbe_atr(tx_ring, first);
6282 
6283 #ifdef IXGBE_FCOE
6284 xmit_fcoe:
6285 #endif /* IXGBE_FCOE */
6286 	ixgbe_tx_map(tx_ring, first, hdr_len);
6287 
6288 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
6289 
6290 	return NETDEV_TX_OK;
6291 
6292 out_drop:
6293 	dev_kfree_skb_any(first->skb);
6294 	first->skb = NULL;
6295 
6296 	return NETDEV_TX_OK;
6297 }
6298 
6299 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
6300 				    struct net_device *netdev)
6301 {
6302 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6303 	struct ixgbe_ring *tx_ring;
6304 
6305 	if (skb->len <= 0) {
6306 		dev_kfree_skb_any(skb);
6307 		return NETDEV_TX_OK;
6308 	}
6309 
6310 	/*
6311 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
6312 	 * in order to meet this minimum size requirement.
6313 	 */
6314 	if (skb->len < 17) {
6315 		if (skb_padto(skb, 17))
6316 			return NETDEV_TX_OK;
6317 		skb->len = 17;
6318 	}
6319 
6320 	tx_ring = adapter->tx_ring[skb->queue_mapping];
6321 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
6322 }
6323 
6324 /**
6325  * ixgbe_set_mac - Change the Ethernet Address of the NIC
6326  * @netdev: network interface device structure
6327  * @p: pointer to an address structure
6328  *
6329  * Returns 0 on success, negative on failure
6330  **/
6331 static int ixgbe_set_mac(struct net_device *netdev, void *p)
6332 {
6333 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6334 	struct ixgbe_hw *hw = &adapter->hw;
6335 	struct sockaddr *addr = p;
6336 
6337 	if (!is_valid_ether_addr(addr->sa_data))
6338 		return -EADDRNOTAVAIL;
6339 
6340 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
6341 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
6342 
6343 	hw->mac.ops.set_rar(hw, 0, hw->mac.addr, adapter->num_vfs,
6344 			    IXGBE_RAH_AV);
6345 
6346 	return 0;
6347 }
6348 
6349 static int
6350 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
6351 {
6352 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6353 	struct ixgbe_hw *hw = &adapter->hw;
6354 	u16 value;
6355 	int rc;
6356 
6357 	if (prtad != hw->phy.mdio.prtad)
6358 		return -EINVAL;
6359 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
6360 	if (!rc)
6361 		rc = value;
6362 	return rc;
6363 }
6364 
6365 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
6366 			    u16 addr, u16 value)
6367 {
6368 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6369 	struct ixgbe_hw *hw = &adapter->hw;
6370 
6371 	if (prtad != hw->phy.mdio.prtad)
6372 		return -EINVAL;
6373 	return hw->phy.ops.write_reg(hw, addr, devad, value);
6374 }
6375 
6376 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
6377 {
6378 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6379 
6380 	return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
6381 }
6382 
6383 /**
6384  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
6385  * netdev->dev_addrs
6386  * @netdev: network interface device structure
6387  *
6388  * Returns non-zero on failure
6389  **/
6390 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
6391 {
6392 	int err = 0;
6393 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6394 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
6395 
6396 	if (is_valid_ether_addr(mac->san_addr)) {
6397 		rtnl_lock();
6398 		err = dev_addr_add(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
6399 		rtnl_unlock();
6400 	}
6401 	return err;
6402 }
6403 
6404 /**
6405  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
6406  * netdev->dev_addrs
6407  * @netdev: network interface device structure
6408  *
6409  * Returns non-zero on failure
6410  **/
6411 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
6412 {
6413 	int err = 0;
6414 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6415 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
6416 
6417 	if (is_valid_ether_addr(mac->san_addr)) {
6418 		rtnl_lock();
6419 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
6420 		rtnl_unlock();
6421 	}
6422 	return err;
6423 }
6424 
6425 #ifdef CONFIG_NET_POLL_CONTROLLER
6426 /*
6427  * Polling 'interrupt' - used by things like netconsole to send skbs
6428  * without having to re-enable interrupts. It's not called while
6429  * the interrupt routine is executing.
6430  */
6431 static void ixgbe_netpoll(struct net_device *netdev)
6432 {
6433 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6434 	int i;
6435 
6436 	/* if interface is down do nothing */
6437 	if (test_bit(__IXGBE_DOWN, &adapter->state))
6438 		return;
6439 
6440 	adapter->flags |= IXGBE_FLAG_IN_NETPOLL;
6441 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
6442 		int num_q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
6443 		for (i = 0; i < num_q_vectors; i++) {
6444 			struct ixgbe_q_vector *q_vector = adapter->q_vector[i];
6445 			ixgbe_msix_clean_rings(0, q_vector);
6446 		}
6447 	} else {
6448 		ixgbe_intr(adapter->pdev->irq, netdev);
6449 	}
6450 	adapter->flags &= ~IXGBE_FLAG_IN_NETPOLL;
6451 }
6452 
6453 #endif
6454 static struct rtnl_link_stats64 *ixgbe_get_stats64(struct net_device *netdev,
6455 						   struct rtnl_link_stats64 *stats)
6456 {
6457 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6458 	int i;
6459 
6460 	rcu_read_lock();
6461 	for (i = 0; i < adapter->num_rx_queues; i++) {
6462 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
6463 		u64 bytes, packets;
6464 		unsigned int start;
6465 
6466 		if (ring) {
6467 			do {
6468 				start = u64_stats_fetch_begin_bh(&ring->syncp);
6469 				packets = ring->stats.packets;
6470 				bytes   = ring->stats.bytes;
6471 			} while (u64_stats_fetch_retry_bh(&ring->syncp, start));
6472 			stats->rx_packets += packets;
6473 			stats->rx_bytes   += bytes;
6474 		}
6475 	}
6476 
6477 	for (i = 0; i < adapter->num_tx_queues; i++) {
6478 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
6479 		u64 bytes, packets;
6480 		unsigned int start;
6481 
6482 		if (ring) {
6483 			do {
6484 				start = u64_stats_fetch_begin_bh(&ring->syncp);
6485 				packets = ring->stats.packets;
6486 				bytes   = ring->stats.bytes;
6487 			} while (u64_stats_fetch_retry_bh(&ring->syncp, start));
6488 			stats->tx_packets += packets;
6489 			stats->tx_bytes   += bytes;
6490 		}
6491 	}
6492 	rcu_read_unlock();
6493 	/* following stats updated by ixgbe_watchdog_task() */
6494 	stats->multicast	= netdev->stats.multicast;
6495 	stats->rx_errors	= netdev->stats.rx_errors;
6496 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
6497 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
6498 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
6499 	return stats;
6500 }
6501 
6502 #ifdef CONFIG_IXGBE_DCB
6503 /* ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
6504  * #adapter: pointer to ixgbe_adapter
6505  * @tc: number of traffic classes currently enabled
6506  *
6507  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
6508  * 802.1Q priority maps to a packet buffer that exists.
6509  */
6510 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
6511 {
6512 	struct ixgbe_hw *hw = &adapter->hw;
6513 	u32 reg, rsave;
6514 	int i;
6515 
6516 	/* 82598 have a static priority to TC mapping that can not
6517 	 * be changed so no validation is needed.
6518 	 */
6519 	if (hw->mac.type == ixgbe_mac_82598EB)
6520 		return;
6521 
6522 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
6523 	rsave = reg;
6524 
6525 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
6526 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
6527 
6528 		/* If up2tc is out of bounds default to zero */
6529 		if (up2tc > tc)
6530 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
6531 	}
6532 
6533 	if (reg != rsave)
6534 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
6535 
6536 	return;
6537 }
6538 
6539 /* ixgbe_setup_tc - routine to configure net_device for multiple traffic
6540  * classes.
6541  *
6542  * @netdev: net device to configure
6543  * @tc: number of traffic classes to enable
6544  */
6545 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
6546 {
6547 	struct ixgbe_adapter *adapter = netdev_priv(dev);
6548 	struct ixgbe_hw *hw = &adapter->hw;
6549 
6550 	/* Multiple traffic classes requires multiple queues */
6551 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
6552 		e_err(drv, "Enable failed, needs MSI-X\n");
6553 		return -EINVAL;
6554 	}
6555 
6556 	/* Hardware supports up to 8 traffic classes */
6557 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs ||
6558 	    (hw->mac.type == ixgbe_mac_82598EB &&
6559 	     tc < MAX_TRAFFIC_CLASS))
6560 		return -EINVAL;
6561 
6562 	/* Hardware has to reinitialize queues and interrupts to
6563 	 * match packet buffer alignment. Unfortunately, the
6564 	 * hardware is not flexible enough to do this dynamically.
6565 	 */
6566 	if (netif_running(dev))
6567 		ixgbe_close(dev);
6568 	ixgbe_clear_interrupt_scheme(adapter);
6569 
6570 	if (tc) {
6571 		netdev_set_num_tc(dev, tc);
6572 		adapter->last_lfc_mode = adapter->hw.fc.current_mode;
6573 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
6574 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
6575 
6576 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
6577 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
6578 	} else {
6579 		netdev_reset_tc(dev);
6580 		adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
6581 
6582 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
6583 		adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
6584 
6585 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
6586 		adapter->dcb_cfg.pfc_mode_enable = false;
6587 	}
6588 
6589 	ixgbe_init_interrupt_scheme(adapter);
6590 	ixgbe_validate_rtr(adapter, tc);
6591 	if (netif_running(dev))
6592 		ixgbe_open(dev);
6593 
6594 	return 0;
6595 }
6596 
6597 #endif /* CONFIG_IXGBE_DCB */
6598 void ixgbe_do_reset(struct net_device *netdev)
6599 {
6600 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6601 
6602 	if (netif_running(netdev))
6603 		ixgbe_reinit_locked(adapter);
6604 	else
6605 		ixgbe_reset(adapter);
6606 }
6607 
6608 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
6609 					    netdev_features_t features)
6610 {
6611 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6612 
6613 #ifdef CONFIG_DCB
6614 	if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
6615 		features &= ~NETIF_F_HW_VLAN_RX;
6616 #endif
6617 
6618 	/* return error if RXHASH is being enabled when RSS is not supported */
6619 	if (!(adapter->flags & IXGBE_FLAG_RSS_ENABLED))
6620 		features &= ~NETIF_F_RXHASH;
6621 
6622 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
6623 	if (!(features & NETIF_F_RXCSUM))
6624 		features &= ~NETIF_F_LRO;
6625 
6626 	/* Turn off LRO if not RSC capable */
6627 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
6628 		features &= ~NETIF_F_LRO;
6629 
6630 
6631 	return features;
6632 }
6633 
6634 static int ixgbe_set_features(struct net_device *netdev,
6635 			      netdev_features_t features)
6636 {
6637 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6638 	netdev_features_t changed = netdev->features ^ features;
6639 	bool need_reset = false;
6640 
6641 	/* Make sure RSC matches LRO, reset if change */
6642 	if (!(features & NETIF_F_LRO)) {
6643 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
6644 			need_reset = true;
6645 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
6646 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
6647 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
6648 		if (adapter->rx_itr_setting == 1 ||
6649 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
6650 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
6651 			need_reset = true;
6652 		} else if ((changed ^ features) & NETIF_F_LRO) {
6653 			e_info(probe, "rx-usecs set too low, "
6654 			       "disabling RSC\n");
6655 		}
6656 	}
6657 
6658 	/*
6659 	 * Check if Flow Director n-tuple support was enabled or disabled.  If
6660 	 * the state changed, we need to reset.
6661 	 */
6662 	if (!(features & NETIF_F_NTUPLE)) {
6663 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
6664 			/* turn off Flow Director, set ATR and reset */
6665 			if ((adapter->flags & IXGBE_FLAG_RSS_ENABLED) &&
6666 			    !(adapter->flags & IXGBE_FLAG_DCB_ENABLED))
6667 				adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
6668 			need_reset = true;
6669 		}
6670 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
6671 	} else if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)) {
6672 		/* turn off ATR, enable perfect filters and reset */
6673 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
6674 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
6675 		need_reset = true;
6676 	}
6677 
6678 	if (changed & NETIF_F_RXALL)
6679 		need_reset = true;
6680 
6681 	netdev->features = features;
6682 	if (need_reset)
6683 		ixgbe_do_reset(netdev);
6684 
6685 	return 0;
6686 }
6687 
6688 static const struct net_device_ops ixgbe_netdev_ops = {
6689 	.ndo_open		= ixgbe_open,
6690 	.ndo_stop		= ixgbe_close,
6691 	.ndo_start_xmit		= ixgbe_xmit_frame,
6692 	.ndo_select_queue	= ixgbe_select_queue,
6693 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
6694 	.ndo_validate_addr	= eth_validate_addr,
6695 	.ndo_set_mac_address	= ixgbe_set_mac,
6696 	.ndo_change_mtu		= ixgbe_change_mtu,
6697 	.ndo_tx_timeout		= ixgbe_tx_timeout,
6698 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
6699 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
6700 	.ndo_do_ioctl		= ixgbe_ioctl,
6701 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
6702 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
6703 	.ndo_set_vf_tx_rate	= ixgbe_ndo_set_vf_bw,
6704 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
6705 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
6706 	.ndo_get_stats64	= ixgbe_get_stats64,
6707 #ifdef CONFIG_IXGBE_DCB
6708 	.ndo_setup_tc		= ixgbe_setup_tc,
6709 #endif
6710 #ifdef CONFIG_NET_POLL_CONTROLLER
6711 	.ndo_poll_controller	= ixgbe_netpoll,
6712 #endif
6713 #ifdef IXGBE_FCOE
6714 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
6715 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
6716 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
6717 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
6718 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
6719 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
6720 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
6721 #endif /* IXGBE_FCOE */
6722 	.ndo_set_features = ixgbe_set_features,
6723 	.ndo_fix_features = ixgbe_fix_features,
6724 };
6725 
6726 static void __devinit ixgbe_probe_vf(struct ixgbe_adapter *adapter,
6727 				     const struct ixgbe_info *ii)
6728 {
6729 #ifdef CONFIG_PCI_IOV
6730 	struct ixgbe_hw *hw = &adapter->hw;
6731 
6732 	if (hw->mac.type == ixgbe_mac_82598EB)
6733 		return;
6734 
6735 	/* The 82599 supports up to 64 VFs per physical function
6736 	 * but this implementation limits allocation to 63 so that
6737 	 * basic networking resources are still available to the
6738 	 * physical function
6739 	 */
6740 	adapter->num_vfs = (max_vfs > 63) ? 63 : max_vfs;
6741 	ixgbe_enable_sriov(adapter, ii);
6742 #endif /* CONFIG_PCI_IOV */
6743 }
6744 
6745 /**
6746  * ixgbe_probe - Device Initialization Routine
6747  * @pdev: PCI device information struct
6748  * @ent: entry in ixgbe_pci_tbl
6749  *
6750  * Returns 0 on success, negative on failure
6751  *
6752  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
6753  * The OS initialization, configuring of the adapter private structure,
6754  * and a hardware reset occur.
6755  **/
6756 static int __devinit ixgbe_probe(struct pci_dev *pdev,
6757 				 const struct pci_device_id *ent)
6758 {
6759 	struct net_device *netdev;
6760 	struct ixgbe_adapter *adapter = NULL;
6761 	struct ixgbe_hw *hw;
6762 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
6763 	static int cards_found;
6764 	int i, err, pci_using_dac;
6765 	u8 part_str[IXGBE_PBANUM_LENGTH];
6766 	unsigned int indices = num_possible_cpus();
6767 #ifdef IXGBE_FCOE
6768 	u16 device_caps;
6769 #endif
6770 	u32 eec;
6771 	u16 wol_cap;
6772 
6773 	/* Catch broken hardware that put the wrong VF device ID in
6774 	 * the PCIe SR-IOV capability.
6775 	 */
6776 	if (pdev->is_virtfn) {
6777 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
6778 		     pci_name(pdev), pdev->vendor, pdev->device);
6779 		return -EINVAL;
6780 	}
6781 
6782 	err = pci_enable_device_mem(pdev);
6783 	if (err)
6784 		return err;
6785 
6786 	if (!dma_set_mask(&pdev->dev, DMA_BIT_MASK(64)) &&
6787 	    !dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(64))) {
6788 		pci_using_dac = 1;
6789 	} else {
6790 		err = dma_set_mask(&pdev->dev, DMA_BIT_MASK(32));
6791 		if (err) {
6792 			err = dma_set_coherent_mask(&pdev->dev,
6793 						    DMA_BIT_MASK(32));
6794 			if (err) {
6795 				dev_err(&pdev->dev,
6796 					"No usable DMA configuration, aborting\n");
6797 				goto err_dma;
6798 			}
6799 		}
6800 		pci_using_dac = 0;
6801 	}
6802 
6803 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
6804 					   IORESOURCE_MEM), ixgbe_driver_name);
6805 	if (err) {
6806 		dev_err(&pdev->dev,
6807 			"pci_request_selected_regions failed 0x%x\n", err);
6808 		goto err_pci_reg;
6809 	}
6810 
6811 	pci_enable_pcie_error_reporting(pdev);
6812 
6813 	pci_set_master(pdev);
6814 	pci_save_state(pdev);
6815 
6816 #ifdef CONFIG_IXGBE_DCB
6817 	indices *= MAX_TRAFFIC_CLASS;
6818 #endif
6819 
6820 	if (ii->mac == ixgbe_mac_82598EB)
6821 		indices = min_t(unsigned int, indices, IXGBE_MAX_RSS_INDICES);
6822 	else
6823 		indices = min_t(unsigned int, indices, IXGBE_MAX_FDIR_INDICES);
6824 
6825 #ifdef IXGBE_FCOE
6826 	indices += min_t(unsigned int, num_possible_cpus(),
6827 			 IXGBE_MAX_FCOE_INDICES);
6828 #endif
6829 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
6830 	if (!netdev) {
6831 		err = -ENOMEM;
6832 		goto err_alloc_etherdev;
6833 	}
6834 
6835 	SET_NETDEV_DEV(netdev, &pdev->dev);
6836 
6837 	adapter = netdev_priv(netdev);
6838 	pci_set_drvdata(pdev, adapter);
6839 
6840 	adapter->netdev = netdev;
6841 	adapter->pdev = pdev;
6842 	hw = &adapter->hw;
6843 	hw->back = adapter;
6844 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
6845 
6846 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
6847 			      pci_resource_len(pdev, 0));
6848 	if (!hw->hw_addr) {
6849 		err = -EIO;
6850 		goto err_ioremap;
6851 	}
6852 
6853 	for (i = 1; i <= 5; i++) {
6854 		if (pci_resource_len(pdev, i) == 0)
6855 			continue;
6856 	}
6857 
6858 	netdev->netdev_ops = &ixgbe_netdev_ops;
6859 	ixgbe_set_ethtool_ops(netdev);
6860 	netdev->watchdog_timeo = 5 * HZ;
6861 	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
6862 
6863 	adapter->bd_number = cards_found;
6864 
6865 	/* Setup hw api */
6866 	memcpy(&hw->mac.ops, ii->mac_ops, sizeof(hw->mac.ops));
6867 	hw->mac.type  = ii->mac;
6868 
6869 	/* EEPROM */
6870 	memcpy(&hw->eeprom.ops, ii->eeprom_ops, sizeof(hw->eeprom.ops));
6871 	eec = IXGBE_READ_REG(hw, IXGBE_EEC);
6872 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
6873 	if (!(eec & (1 << 8)))
6874 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
6875 
6876 	/* PHY */
6877 	memcpy(&hw->phy.ops, ii->phy_ops, sizeof(hw->phy.ops));
6878 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
6879 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
6880 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
6881 	hw->phy.mdio.mmds = 0;
6882 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
6883 	hw->phy.mdio.dev = netdev;
6884 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
6885 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
6886 
6887 	ii->get_invariants(hw);
6888 
6889 	/* setup the private structure */
6890 	err = ixgbe_sw_init(adapter);
6891 	if (err)
6892 		goto err_sw_init;
6893 
6894 	/* Make it possible the adapter to be woken up via WOL */
6895 	switch (adapter->hw.mac.type) {
6896 	case ixgbe_mac_82599EB:
6897 	case ixgbe_mac_X540:
6898 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
6899 		break;
6900 	default:
6901 		break;
6902 	}
6903 
6904 	/*
6905 	 * If there is a fan on this device and it has failed log the
6906 	 * failure.
6907 	 */
6908 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
6909 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
6910 		if (esdp & IXGBE_ESDP_SDP1)
6911 			e_crit(probe, "Fan has stopped, replace the adapter\n");
6912 	}
6913 
6914 	if (allow_unsupported_sfp)
6915 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
6916 
6917 	/* reset_hw fills in the perm_addr as well */
6918 	hw->phy.reset_if_overtemp = true;
6919 	err = hw->mac.ops.reset_hw(hw);
6920 	hw->phy.reset_if_overtemp = false;
6921 	if (err == IXGBE_ERR_SFP_NOT_PRESENT &&
6922 	    hw->mac.type == ixgbe_mac_82598EB) {
6923 		err = 0;
6924 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
6925 		e_dev_err("failed to load because an unsupported SFP+ "
6926 			  "module type was detected.\n");
6927 		e_dev_err("Reload the driver after installing a supported "
6928 			  "module.\n");
6929 		goto err_sw_init;
6930 	} else if (err) {
6931 		e_dev_err("HW Init failed: %d\n", err);
6932 		goto err_sw_init;
6933 	}
6934 
6935 	ixgbe_probe_vf(adapter, ii);
6936 
6937 	netdev->features = NETIF_F_SG |
6938 			   NETIF_F_IP_CSUM |
6939 			   NETIF_F_IPV6_CSUM |
6940 			   NETIF_F_HW_VLAN_TX |
6941 			   NETIF_F_HW_VLAN_RX |
6942 			   NETIF_F_HW_VLAN_FILTER |
6943 			   NETIF_F_TSO |
6944 			   NETIF_F_TSO6 |
6945 			   NETIF_F_RXHASH |
6946 			   NETIF_F_RXCSUM;
6947 
6948 	netdev->hw_features = netdev->features;
6949 
6950 	switch (adapter->hw.mac.type) {
6951 	case ixgbe_mac_82599EB:
6952 	case ixgbe_mac_X540:
6953 		netdev->features |= NETIF_F_SCTP_CSUM;
6954 		netdev->hw_features |= NETIF_F_SCTP_CSUM |
6955 				       NETIF_F_NTUPLE;
6956 		break;
6957 	default:
6958 		break;
6959 	}
6960 
6961 	netdev->hw_features |= NETIF_F_RXALL;
6962 
6963 	netdev->vlan_features |= NETIF_F_TSO;
6964 	netdev->vlan_features |= NETIF_F_TSO6;
6965 	netdev->vlan_features |= NETIF_F_IP_CSUM;
6966 	netdev->vlan_features |= NETIF_F_IPV6_CSUM;
6967 	netdev->vlan_features |= NETIF_F_SG;
6968 
6969 	netdev->priv_flags |= IFF_UNICAST_FLT;
6970 	netdev->priv_flags |= IFF_SUPP_NOFCS;
6971 
6972 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
6973 		adapter->flags &= ~(IXGBE_FLAG_RSS_ENABLED |
6974 				    IXGBE_FLAG_DCB_ENABLED);
6975 
6976 #ifdef CONFIG_IXGBE_DCB
6977 	netdev->dcbnl_ops = &dcbnl_ops;
6978 #endif
6979 
6980 #ifdef IXGBE_FCOE
6981 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
6982 		if (hw->mac.ops.get_device_caps) {
6983 			hw->mac.ops.get_device_caps(hw, &device_caps);
6984 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
6985 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6986 		}
6987 	}
6988 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
6989 		netdev->vlan_features |= NETIF_F_FCOE_CRC;
6990 		netdev->vlan_features |= NETIF_F_FSO;
6991 		netdev->vlan_features |= NETIF_F_FCOE_MTU;
6992 	}
6993 #endif /* IXGBE_FCOE */
6994 	if (pci_using_dac) {
6995 		netdev->features |= NETIF_F_HIGHDMA;
6996 		netdev->vlan_features |= NETIF_F_HIGHDMA;
6997 	}
6998 
6999 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
7000 		netdev->hw_features |= NETIF_F_LRO;
7001 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
7002 		netdev->features |= NETIF_F_LRO;
7003 
7004 	/* make sure the EEPROM is good */
7005 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
7006 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
7007 		err = -EIO;
7008 		goto err_sw_init;
7009 	}
7010 
7011 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
7012 	memcpy(netdev->perm_addr, hw->mac.perm_addr, netdev->addr_len);
7013 
7014 	if (ixgbe_validate_mac_addr(netdev->perm_addr)) {
7015 		e_dev_err("invalid MAC address\n");
7016 		err = -EIO;
7017 		goto err_sw_init;
7018 	}
7019 
7020 	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
7021 		    (unsigned long) adapter);
7022 
7023 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
7024 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
7025 
7026 	err = ixgbe_init_interrupt_scheme(adapter);
7027 	if (err)
7028 		goto err_sw_init;
7029 
7030 	if (!(adapter->flags & IXGBE_FLAG_RSS_ENABLED)) {
7031 		netdev->hw_features &= ~NETIF_F_RXHASH;
7032 		netdev->features &= ~NETIF_F_RXHASH;
7033 	}
7034 
7035 	/* WOL not supported for all but the following */
7036 	adapter->wol = 0;
7037 	switch (pdev->device) {
7038 	case IXGBE_DEV_ID_82599_SFP:
7039 		/* Only these subdevice supports WOL */
7040 		switch (pdev->subsystem_device) {
7041 		case IXGBE_SUBDEV_ID_82599_560FLR:
7042 			/* only support first port */
7043 			if (hw->bus.func != 0)
7044 				break;
7045 		case IXGBE_SUBDEV_ID_82599_SFP:
7046 			adapter->wol = IXGBE_WUFC_MAG;
7047 			break;
7048 		}
7049 		break;
7050 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
7051 		/* All except this subdevice support WOL */
7052 		if (pdev->subsystem_device != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
7053 			adapter->wol = IXGBE_WUFC_MAG;
7054 		break;
7055 	case IXGBE_DEV_ID_82599_KX4:
7056 		adapter->wol = IXGBE_WUFC_MAG;
7057 		break;
7058 	case IXGBE_DEV_ID_X540T:
7059 		/* Check eeprom to see if it is enabled */
7060 		hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
7061 		wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
7062 
7063 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
7064 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
7065 		     (hw->bus.func == 0)))
7066 			adapter->wol = IXGBE_WUFC_MAG;
7067 		break;
7068 	}
7069 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
7070 
7071 	/* save off EEPROM version number */
7072 	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
7073 	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
7074 
7075 	/* pick up the PCI bus settings for reporting later */
7076 	hw->mac.ops.get_bus_info(hw);
7077 
7078 	/* print bus type/speed/width info */
7079 	e_dev_info("(PCI Express:%s:%s) %pM\n",
7080 		   (hw->bus.speed == ixgbe_bus_speed_5000 ? "5.0GT/s" :
7081 		    hw->bus.speed == ixgbe_bus_speed_2500 ? "2.5GT/s" :
7082 		    "Unknown"),
7083 		   (hw->bus.width == ixgbe_bus_width_pcie_x8 ? "Width x8" :
7084 		    hw->bus.width == ixgbe_bus_width_pcie_x4 ? "Width x4" :
7085 		    hw->bus.width == ixgbe_bus_width_pcie_x1 ? "Width x1" :
7086 		    "Unknown"),
7087 		   netdev->dev_addr);
7088 
7089 	err = ixgbe_read_pba_string_generic(hw, part_str, IXGBE_PBANUM_LENGTH);
7090 	if (err)
7091 		strncpy(part_str, "Unknown", IXGBE_PBANUM_LENGTH);
7092 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
7093 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
7094 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
7095 		           part_str);
7096 	else
7097 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
7098 			   hw->mac.type, hw->phy.type, part_str);
7099 
7100 	if (hw->bus.width <= ixgbe_bus_width_pcie_x4) {
7101 		e_dev_warn("PCI-Express bandwidth available for this card is "
7102 			   "not sufficient for optimal performance.\n");
7103 		e_dev_warn("For optimal performance a x8 PCI-Express slot "
7104 			   "is required.\n");
7105 	}
7106 
7107 	/* reset the hardware with the new settings */
7108 	err = hw->mac.ops.start_hw(hw);
7109 	if (err == IXGBE_ERR_EEPROM_VERSION) {
7110 		/* We are running on a pre-production device, log a warning */
7111 		e_dev_warn("This device is a pre-production adapter/LOM. "
7112 			   "Please be aware there may be issues associated "
7113 			   "with your hardware.  If you are experiencing "
7114 			   "problems please contact your Intel or hardware "
7115 			   "representative who provided you with this "
7116 			   "hardware.\n");
7117 	}
7118 	strcpy(netdev->name, "eth%d");
7119 	err = register_netdev(netdev);
7120 	if (err)
7121 		goto err_register;
7122 
7123 	/* power down the optics for multispeed fiber and 82599 SFP+ fiber */
7124 	if (hw->mac.ops.disable_tx_laser &&
7125 	    ((hw->phy.multispeed_fiber) ||
7126 	     ((hw->mac.ops.get_media_type(hw) == ixgbe_media_type_fiber) &&
7127 	      (hw->mac.type == ixgbe_mac_82599EB))))
7128 		hw->mac.ops.disable_tx_laser(hw);
7129 
7130 	/* carrier off reporting is important to ethtool even BEFORE open */
7131 	netif_carrier_off(netdev);
7132 
7133 #ifdef CONFIG_IXGBE_DCA
7134 	if (dca_add_requester(&pdev->dev) == 0) {
7135 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
7136 		ixgbe_setup_dca(adapter);
7137 	}
7138 #endif
7139 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
7140 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
7141 		for (i = 0; i < adapter->num_vfs; i++)
7142 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
7143 	}
7144 
7145 	/* firmware requires driver version to be 0xFFFFFFFF
7146 	 * since os does not support feature
7147 	 */
7148 	if (hw->mac.ops.set_fw_drv_ver)
7149 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF,
7150 					   0xFF);
7151 
7152 	/* add san mac addr to netdev */
7153 	ixgbe_add_sanmac_netdev(netdev);
7154 
7155 	e_dev_info("%s\n", ixgbe_default_device_descr);
7156 	cards_found++;
7157 	return 0;
7158 
7159 err_register:
7160 	ixgbe_release_hw_control(adapter);
7161 	ixgbe_clear_interrupt_scheme(adapter);
7162 err_sw_init:
7163 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7164 		ixgbe_disable_sriov(adapter);
7165 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
7166 	iounmap(hw->hw_addr);
7167 err_ioremap:
7168 	free_netdev(netdev);
7169 err_alloc_etherdev:
7170 	pci_release_selected_regions(pdev,
7171 				     pci_select_bars(pdev, IORESOURCE_MEM));
7172 err_pci_reg:
7173 err_dma:
7174 	pci_disable_device(pdev);
7175 	return err;
7176 }
7177 
7178 /**
7179  * ixgbe_remove - Device Removal Routine
7180  * @pdev: PCI device information struct
7181  *
7182  * ixgbe_remove is called by the PCI subsystem to alert the driver
7183  * that it should release a PCI device.  The could be caused by a
7184  * Hot-Plug event, or because the driver is going to be removed from
7185  * memory.
7186  **/
7187 static void __devexit ixgbe_remove(struct pci_dev *pdev)
7188 {
7189 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7190 	struct net_device *netdev = adapter->netdev;
7191 
7192 	set_bit(__IXGBE_DOWN, &adapter->state);
7193 	cancel_work_sync(&adapter->service_task);
7194 
7195 #ifdef CONFIG_IXGBE_DCA
7196 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
7197 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
7198 		dca_remove_requester(&pdev->dev);
7199 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
7200 	}
7201 
7202 #endif
7203 #ifdef IXGBE_FCOE
7204 	if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
7205 		ixgbe_cleanup_fcoe(adapter);
7206 
7207 #endif /* IXGBE_FCOE */
7208 
7209 	/* remove the added san mac */
7210 	ixgbe_del_sanmac_netdev(netdev);
7211 
7212 	if (netdev->reg_state == NETREG_REGISTERED)
7213 		unregister_netdev(netdev);
7214 
7215 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
7216 		if (!(ixgbe_check_vf_assignment(adapter)))
7217 			ixgbe_disable_sriov(adapter);
7218 		else
7219 			e_dev_warn("Unloading driver while VFs are assigned "
7220 				   "- VFs will not be deallocated\n");
7221 	}
7222 
7223 	ixgbe_clear_interrupt_scheme(adapter);
7224 
7225 	ixgbe_release_hw_control(adapter);
7226 
7227 	iounmap(adapter->hw.hw_addr);
7228 	pci_release_selected_regions(pdev, pci_select_bars(pdev,
7229 				     IORESOURCE_MEM));
7230 
7231 	e_dev_info("complete\n");
7232 
7233 	free_netdev(netdev);
7234 
7235 	pci_disable_pcie_error_reporting(pdev);
7236 
7237 	pci_disable_device(pdev);
7238 }
7239 
7240 /**
7241  * ixgbe_io_error_detected - called when PCI error is detected
7242  * @pdev: Pointer to PCI device
7243  * @state: The current pci connection state
7244  *
7245  * This function is called after a PCI bus error affecting
7246  * this device has been detected.
7247  */
7248 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
7249 						pci_channel_state_t state)
7250 {
7251 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7252 	struct net_device *netdev = adapter->netdev;
7253 
7254 #ifdef CONFIG_PCI_IOV
7255 	struct pci_dev *bdev, *vfdev;
7256 	u32 dw0, dw1, dw2, dw3;
7257 	int vf, pos;
7258 	u16 req_id, pf_func;
7259 
7260 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7261 	    adapter->num_vfs == 0)
7262 		goto skip_bad_vf_detection;
7263 
7264 	bdev = pdev->bus->self;
7265 	while (bdev && (bdev->pcie_type != PCI_EXP_TYPE_ROOT_PORT))
7266 		bdev = bdev->bus->self;
7267 
7268 	if (!bdev)
7269 		goto skip_bad_vf_detection;
7270 
7271 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
7272 	if (!pos)
7273 		goto skip_bad_vf_detection;
7274 
7275 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG, &dw0);
7276 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG + 4, &dw1);
7277 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG + 8, &dw2);
7278 	pci_read_config_dword(bdev, pos + PCI_ERR_HEADER_LOG + 12, &dw3);
7279 
7280 	req_id = dw1 >> 16;
7281 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
7282 	if (!(req_id & 0x0080))
7283 		goto skip_bad_vf_detection;
7284 
7285 	pf_func = req_id & 0x01;
7286 	if ((pf_func & 1) == (pdev->devfn & 1)) {
7287 		unsigned int device_id;
7288 
7289 		vf = (req_id & 0x7F) >> 1;
7290 		e_dev_err("VF %d has caused a PCIe error\n", vf);
7291 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
7292 				"%8.8x\tdw3: %8.8x\n",
7293 		dw0, dw1, dw2, dw3);
7294 		switch (adapter->hw.mac.type) {
7295 		case ixgbe_mac_82599EB:
7296 			device_id = IXGBE_82599_VF_DEVICE_ID;
7297 			break;
7298 		case ixgbe_mac_X540:
7299 			device_id = IXGBE_X540_VF_DEVICE_ID;
7300 			break;
7301 		default:
7302 			device_id = 0;
7303 			break;
7304 		}
7305 
7306 		/* Find the pci device of the offending VF */
7307 		vfdev = pci_get_device(IXGBE_INTEL_VENDOR_ID, device_id, NULL);
7308 		while (vfdev) {
7309 			if (vfdev->devfn == (req_id & 0xFF))
7310 				break;
7311 			vfdev = pci_get_device(IXGBE_INTEL_VENDOR_ID,
7312 					       device_id, vfdev);
7313 		}
7314 		/*
7315 		 * There's a slim chance the VF could have been hot plugged,
7316 		 * so if it is no longer present we don't need to issue the
7317 		 * VFLR.  Just clean up the AER in that case.
7318 		 */
7319 		if (vfdev) {
7320 			e_dev_err("Issuing VFLR to VF %d\n", vf);
7321 			pci_write_config_dword(vfdev, 0xA8, 0x00008000);
7322 		}
7323 
7324 		pci_cleanup_aer_uncorrect_error_status(pdev);
7325 	}
7326 
7327 	/*
7328 	 * Even though the error may have occurred on the other port
7329 	 * we still need to increment the vf error reference count for
7330 	 * both ports because the I/O resume function will be called
7331 	 * for both of them.
7332 	 */
7333 	adapter->vferr_refcount++;
7334 
7335 	return PCI_ERS_RESULT_RECOVERED;
7336 
7337 skip_bad_vf_detection:
7338 #endif /* CONFIG_PCI_IOV */
7339 	netif_device_detach(netdev);
7340 
7341 	if (state == pci_channel_io_perm_failure)
7342 		return PCI_ERS_RESULT_DISCONNECT;
7343 
7344 	if (netif_running(netdev))
7345 		ixgbe_down(adapter);
7346 	pci_disable_device(pdev);
7347 
7348 	/* Request a slot reset. */
7349 	return PCI_ERS_RESULT_NEED_RESET;
7350 }
7351 
7352 /**
7353  * ixgbe_io_slot_reset - called after the pci bus has been reset.
7354  * @pdev: Pointer to PCI device
7355  *
7356  * Restart the card from scratch, as if from a cold-boot.
7357  */
7358 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
7359 {
7360 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7361 	pci_ers_result_t result;
7362 	int err;
7363 
7364 	if (pci_enable_device_mem(pdev)) {
7365 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
7366 		result = PCI_ERS_RESULT_DISCONNECT;
7367 	} else {
7368 		pci_set_master(pdev);
7369 		pci_restore_state(pdev);
7370 		pci_save_state(pdev);
7371 
7372 		pci_wake_from_d3(pdev, false);
7373 
7374 		ixgbe_reset(adapter);
7375 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
7376 		result = PCI_ERS_RESULT_RECOVERED;
7377 	}
7378 
7379 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
7380 	if (err) {
7381 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
7382 			  "failed 0x%0x\n", err);
7383 		/* non-fatal, continue */
7384 	}
7385 
7386 	return result;
7387 }
7388 
7389 /**
7390  * ixgbe_io_resume - called when traffic can start flowing again.
7391  * @pdev: Pointer to PCI device
7392  *
7393  * This callback is called when the error recovery driver tells us that
7394  * its OK to resume normal operation.
7395  */
7396 static void ixgbe_io_resume(struct pci_dev *pdev)
7397 {
7398 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7399 	struct net_device *netdev = adapter->netdev;
7400 
7401 #ifdef CONFIG_PCI_IOV
7402 	if (adapter->vferr_refcount) {
7403 		e_info(drv, "Resuming after VF err\n");
7404 		adapter->vferr_refcount--;
7405 		return;
7406 	}
7407 
7408 #endif
7409 	if (netif_running(netdev))
7410 		ixgbe_up(adapter);
7411 
7412 	netif_device_attach(netdev);
7413 }
7414 
7415 static struct pci_error_handlers ixgbe_err_handler = {
7416 	.error_detected = ixgbe_io_error_detected,
7417 	.slot_reset = ixgbe_io_slot_reset,
7418 	.resume = ixgbe_io_resume,
7419 };
7420 
7421 static struct pci_driver ixgbe_driver = {
7422 	.name     = ixgbe_driver_name,
7423 	.id_table = ixgbe_pci_tbl,
7424 	.probe    = ixgbe_probe,
7425 	.remove   = __devexit_p(ixgbe_remove),
7426 #ifdef CONFIG_PM
7427 	.suspend  = ixgbe_suspend,
7428 	.resume   = ixgbe_resume,
7429 #endif
7430 	.shutdown = ixgbe_shutdown,
7431 	.err_handler = &ixgbe_err_handler
7432 };
7433 
7434 /**
7435  * ixgbe_init_module - Driver Registration Routine
7436  *
7437  * ixgbe_init_module is the first routine called when the driver is
7438  * loaded. All it does is register with the PCI subsystem.
7439  **/
7440 static int __init ixgbe_init_module(void)
7441 {
7442 	int ret;
7443 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
7444 	pr_info("%s\n", ixgbe_copyright);
7445 
7446 #ifdef CONFIG_IXGBE_DCA
7447 	dca_register_notify(&dca_notifier);
7448 #endif
7449 
7450 	ret = pci_register_driver(&ixgbe_driver);
7451 	return ret;
7452 }
7453 
7454 module_init(ixgbe_init_module);
7455 
7456 /**
7457  * ixgbe_exit_module - Driver Exit Cleanup Routine
7458  *
7459  * ixgbe_exit_module is called just before the driver is removed
7460  * from memory.
7461  **/
7462 static void __exit ixgbe_exit_module(void)
7463 {
7464 #ifdef CONFIG_IXGBE_DCA
7465 	dca_unregister_notify(&dca_notifier);
7466 #endif
7467 	pci_unregister_driver(&ixgbe_driver);
7468 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
7469 }
7470 
7471 #ifdef CONFIG_IXGBE_DCA
7472 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
7473 			    void *p)
7474 {
7475 	int ret_val;
7476 
7477 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
7478 					 __ixgbe_notify_dca);
7479 
7480 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
7481 }
7482 
7483 #endif /* CONFIG_IXGBE_DCA */
7484 
7485 module_exit(ixgbe_exit_module);
7486 
7487 /* ixgbe_main.c */
7488