xref: /openbmc/linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision 206e8c00752fbe9cc463184236ac64b2a532cda5)
1 /*******************************************************************************
2 
3   Intel 10 Gigabit PCI Express Linux driver
4   Copyright(c) 1999 - 2015 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   Linux NICS <linux.nics@intel.com>
24   e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
25   Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
26 
27 *******************************************************************************/
28 
29 #include <linux/types.h>
30 #include <linux/module.h>
31 #include <linux/pci.h>
32 #include <linux/netdevice.h>
33 #include <linux/vmalloc.h>
34 #include <linux/string.h>
35 #include <linux/in.h>
36 #include <linux/interrupt.h>
37 #include <linux/ip.h>
38 #include <linux/tcp.h>
39 #include <linux/sctp.h>
40 #include <linux/pkt_sched.h>
41 #include <linux/ipv6.h>
42 #include <linux/slab.h>
43 #include <net/checksum.h>
44 #include <net/ip6_checksum.h>
45 #include <linux/etherdevice.h>
46 #include <linux/ethtool.h>
47 #include <linux/if.h>
48 #include <linux/if_vlan.h>
49 #include <linux/if_macvlan.h>
50 #include <linux/if_bridge.h>
51 #include <linux/prefetch.h>
52 #include <scsi/fc/fc_fcoe.h>
53 #include <net/vxlan.h>
54 
55 #ifdef CONFIG_OF
56 #include <linux/of_net.h>
57 #endif
58 
59 #ifdef CONFIG_SPARC
60 #include <asm/idprom.h>
61 #include <asm/prom.h>
62 #endif
63 
64 #include "ixgbe.h"
65 #include "ixgbe_common.h"
66 #include "ixgbe_dcb_82599.h"
67 #include "ixgbe_sriov.h"
68 #ifdef CONFIG_IXGBE_VXLAN
69 #include <net/vxlan.h>
70 #endif
71 
72 char ixgbe_driver_name[] = "ixgbe";
73 static const char ixgbe_driver_string[] =
74 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
75 #ifdef IXGBE_FCOE
76 char ixgbe_default_device_descr[] =
77 			      "Intel(R) 10 Gigabit Network Connection";
78 #else
79 static char ixgbe_default_device_descr[] =
80 			      "Intel(R) 10 Gigabit Network Connection";
81 #endif
82 #define DRV_VERSION "4.0.1-k"
83 const char ixgbe_driver_version[] = DRV_VERSION;
84 static const char ixgbe_copyright[] =
85 				"Copyright (c) 1999-2015 Intel Corporation.";
86 
87 static const char ixgbe_overheat_msg[] = "Network adapter has been stopped because it has over heated. Restart the computer. If the problem persists, power off the system and replace the adapter";
88 
89 static const struct ixgbe_info *ixgbe_info_tbl[] = {
90 	[board_82598]		= &ixgbe_82598_info,
91 	[board_82599]		= &ixgbe_82599_info,
92 	[board_X540]		= &ixgbe_X540_info,
93 	[board_X550]		= &ixgbe_X550_info,
94 	[board_X550EM_x]	= &ixgbe_X550EM_x_info,
95 };
96 
97 /* ixgbe_pci_tbl - PCI Device ID Table
98  *
99  * Wildcard entries (PCI_ANY_ID) should come last
100  * Last entry must be all 0s
101  *
102  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
103  *   Class, Class Mask, private data (not used) }
104  */
105 static const struct pci_device_id ixgbe_pci_tbl[] = {
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
115 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
116 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
117 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
118 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
119 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
120 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
121 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
122 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
123 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
124 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
125 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
126 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
127 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
128 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
129 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
130 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
131 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
132 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
133 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
134 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
135 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
136 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
137 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
138 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
139 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_10G_T), board_X550EM_x},
140 	/* required last entry */
141 	{0, }
142 };
143 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
144 
145 #ifdef CONFIG_IXGBE_DCA
146 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
147 			    void *p);
148 static struct notifier_block dca_notifier = {
149 	.notifier_call = ixgbe_notify_dca,
150 	.next          = NULL,
151 	.priority      = 0
152 };
153 #endif
154 
155 #ifdef CONFIG_PCI_IOV
156 static unsigned int max_vfs;
157 module_param(max_vfs, uint, 0);
158 MODULE_PARM_DESC(max_vfs,
159 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
160 #endif /* CONFIG_PCI_IOV */
161 
162 static unsigned int allow_unsupported_sfp;
163 module_param(allow_unsupported_sfp, uint, 0);
164 MODULE_PARM_DESC(allow_unsupported_sfp,
165 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
166 
167 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
168 static int debug = -1;
169 module_param(debug, int, 0);
170 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
171 
172 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
173 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
174 MODULE_LICENSE("GPL");
175 MODULE_VERSION(DRV_VERSION);
176 
177 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
178 
179 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
180 					  u32 reg, u16 *value)
181 {
182 	struct pci_dev *parent_dev;
183 	struct pci_bus *parent_bus;
184 
185 	parent_bus = adapter->pdev->bus->parent;
186 	if (!parent_bus)
187 		return -1;
188 
189 	parent_dev = parent_bus->self;
190 	if (!parent_dev)
191 		return -1;
192 
193 	if (!pci_is_pcie(parent_dev))
194 		return -1;
195 
196 	pcie_capability_read_word(parent_dev, reg, value);
197 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
198 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
199 		return -1;
200 	return 0;
201 }
202 
203 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
204 {
205 	struct ixgbe_hw *hw = &adapter->hw;
206 	u16 link_status = 0;
207 	int err;
208 
209 	hw->bus.type = ixgbe_bus_type_pci_express;
210 
211 	/* Get the negotiated link width and speed from PCI config space of the
212 	 * parent, as this device is behind a switch
213 	 */
214 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
215 
216 	/* assume caller will handle error case */
217 	if (err)
218 		return err;
219 
220 	hw->bus.width = ixgbe_convert_bus_width(link_status);
221 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
222 
223 	return 0;
224 }
225 
226 /**
227  * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
228  * @hw: hw specific details
229  *
230  * This function is used by probe to determine whether a device's PCI-Express
231  * bandwidth details should be gathered from the parent bus instead of from the
232  * device. Used to ensure that various locations all have the correct device ID
233  * checks.
234  */
235 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
236 {
237 	switch (hw->device_id) {
238 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
239 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
240 		return true;
241 	default:
242 		return false;
243 	}
244 }
245 
246 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
247 				     int expected_gts)
248 {
249 	struct ixgbe_hw *hw = &adapter->hw;
250 	int max_gts = 0;
251 	enum pci_bus_speed speed = PCI_SPEED_UNKNOWN;
252 	enum pcie_link_width width = PCIE_LNK_WIDTH_UNKNOWN;
253 	struct pci_dev *pdev;
254 
255 	/* Some devices are not connected over PCIe and thus do not negotiate
256 	 * speed. These devices do not have valid bus info, and thus any report
257 	 * we generate may not be correct.
258 	 */
259 	if (hw->bus.type == ixgbe_bus_type_internal)
260 		return;
261 
262 	/* determine whether to use the parent device */
263 	if (ixgbe_pcie_from_parent(&adapter->hw))
264 		pdev = adapter->pdev->bus->parent->self;
265 	else
266 		pdev = adapter->pdev;
267 
268 	if (pcie_get_minimum_link(pdev, &speed, &width) ||
269 	    speed == PCI_SPEED_UNKNOWN || width == PCIE_LNK_WIDTH_UNKNOWN) {
270 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
271 		return;
272 	}
273 
274 	switch (speed) {
275 	case PCIE_SPEED_2_5GT:
276 		/* 8b/10b encoding reduces max throughput by 20% */
277 		max_gts = 2 * width;
278 		break;
279 	case PCIE_SPEED_5_0GT:
280 		/* 8b/10b encoding reduces max throughput by 20% */
281 		max_gts = 4 * width;
282 		break;
283 	case PCIE_SPEED_8_0GT:
284 		/* 128b/130b encoding reduces throughput by less than 2% */
285 		max_gts = 8 * width;
286 		break;
287 	default:
288 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
289 		return;
290 	}
291 
292 	e_dev_info("PCI Express bandwidth of %dGT/s available\n",
293 		   max_gts);
294 	e_dev_info("(Speed:%s, Width: x%d, Encoding Loss:%s)\n",
295 		   (speed == PCIE_SPEED_8_0GT ? "8.0GT/s" :
296 		    speed == PCIE_SPEED_5_0GT ? "5.0GT/s" :
297 		    speed == PCIE_SPEED_2_5GT ? "2.5GT/s" :
298 		    "Unknown"),
299 		   width,
300 		   (speed == PCIE_SPEED_2_5GT ? "20%" :
301 		    speed == PCIE_SPEED_5_0GT ? "20%" :
302 		    speed == PCIE_SPEED_8_0GT ? "<2%" :
303 		    "Unknown"));
304 
305 	if (max_gts < expected_gts) {
306 		e_dev_warn("This is not sufficient for optimal performance of this card.\n");
307 		e_dev_warn("For optimal performance, at least %dGT/s of bandwidth is required.\n",
308 			expected_gts);
309 		e_dev_warn("A slot with more lanes and/or higher speed is suggested.\n");
310 	}
311 }
312 
313 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
314 {
315 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
316 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
317 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
318 		schedule_work(&adapter->service_task);
319 }
320 
321 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
322 {
323 	struct ixgbe_adapter *adapter = hw->back;
324 
325 	if (!hw->hw_addr)
326 		return;
327 	hw->hw_addr = NULL;
328 	e_dev_err("Adapter removed\n");
329 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
330 		ixgbe_service_event_schedule(adapter);
331 }
332 
333 static void ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
334 {
335 	u32 value;
336 
337 	/* The following check not only optimizes a bit by not
338 	 * performing a read on the status register when the
339 	 * register just read was a status register read that
340 	 * returned IXGBE_FAILED_READ_REG. It also blocks any
341 	 * potential recursion.
342 	 */
343 	if (reg == IXGBE_STATUS) {
344 		ixgbe_remove_adapter(hw);
345 		return;
346 	}
347 	value = ixgbe_read_reg(hw, IXGBE_STATUS);
348 	if (value == IXGBE_FAILED_READ_REG)
349 		ixgbe_remove_adapter(hw);
350 }
351 
352 /**
353  * ixgbe_read_reg - Read from device register
354  * @hw: hw specific details
355  * @reg: offset of register to read
356  *
357  * Returns : value read or IXGBE_FAILED_READ_REG if removed
358  *
359  * This function is used to read device registers. It checks for device
360  * removal by confirming any read that returns all ones by checking the
361  * status register value for all ones. This function avoids reading from
362  * the hardware if a removal was previously detected in which case it
363  * returns IXGBE_FAILED_READ_REG (all ones).
364  */
365 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
366 {
367 	u8 __iomem *reg_addr = ACCESS_ONCE(hw->hw_addr);
368 	u32 value;
369 
370 	if (ixgbe_removed(reg_addr))
371 		return IXGBE_FAILED_READ_REG;
372 	value = readl(reg_addr + reg);
373 	if (unlikely(value == IXGBE_FAILED_READ_REG))
374 		ixgbe_check_remove(hw, reg);
375 	return value;
376 }
377 
378 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
379 {
380 	u16 value;
381 
382 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
383 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
384 		ixgbe_remove_adapter(hw);
385 		return true;
386 	}
387 	return false;
388 }
389 
390 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
391 {
392 	struct ixgbe_adapter *adapter = hw->back;
393 	u16 value;
394 
395 	if (ixgbe_removed(hw->hw_addr))
396 		return IXGBE_FAILED_READ_CFG_WORD;
397 	pci_read_config_word(adapter->pdev, reg, &value);
398 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
399 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
400 		return IXGBE_FAILED_READ_CFG_WORD;
401 	return value;
402 }
403 
404 #ifdef CONFIG_PCI_IOV
405 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
406 {
407 	struct ixgbe_adapter *adapter = hw->back;
408 	u32 value;
409 
410 	if (ixgbe_removed(hw->hw_addr))
411 		return IXGBE_FAILED_READ_CFG_DWORD;
412 	pci_read_config_dword(adapter->pdev, reg, &value);
413 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
414 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
415 		return IXGBE_FAILED_READ_CFG_DWORD;
416 	return value;
417 }
418 #endif /* CONFIG_PCI_IOV */
419 
420 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
421 {
422 	struct ixgbe_adapter *adapter = hw->back;
423 
424 	if (ixgbe_removed(hw->hw_addr))
425 		return;
426 	pci_write_config_word(adapter->pdev, reg, value);
427 }
428 
429 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
430 {
431 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
432 
433 	/* flush memory to make sure state is correct before next watchdog */
434 	smp_mb__before_atomic();
435 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
436 }
437 
438 struct ixgbe_reg_info {
439 	u32 ofs;
440 	char *name;
441 };
442 
443 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
444 
445 	/* General Registers */
446 	{IXGBE_CTRL, "CTRL"},
447 	{IXGBE_STATUS, "STATUS"},
448 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
449 
450 	/* Interrupt Registers */
451 	{IXGBE_EICR, "EICR"},
452 
453 	/* RX Registers */
454 	{IXGBE_SRRCTL(0), "SRRCTL"},
455 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
456 	{IXGBE_RDLEN(0), "RDLEN"},
457 	{IXGBE_RDH(0), "RDH"},
458 	{IXGBE_RDT(0), "RDT"},
459 	{IXGBE_RXDCTL(0), "RXDCTL"},
460 	{IXGBE_RDBAL(0), "RDBAL"},
461 	{IXGBE_RDBAH(0), "RDBAH"},
462 
463 	/* TX Registers */
464 	{IXGBE_TDBAL(0), "TDBAL"},
465 	{IXGBE_TDBAH(0), "TDBAH"},
466 	{IXGBE_TDLEN(0), "TDLEN"},
467 	{IXGBE_TDH(0), "TDH"},
468 	{IXGBE_TDT(0), "TDT"},
469 	{IXGBE_TXDCTL(0), "TXDCTL"},
470 
471 	/* List Terminator */
472 	{ .name = NULL }
473 };
474 
475 
476 /*
477  * ixgbe_regdump - register printout routine
478  */
479 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
480 {
481 	int i = 0, j = 0;
482 	char rname[16];
483 	u32 regs[64];
484 
485 	switch (reginfo->ofs) {
486 	case IXGBE_SRRCTL(0):
487 		for (i = 0; i < 64; i++)
488 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
489 		break;
490 	case IXGBE_DCA_RXCTRL(0):
491 		for (i = 0; i < 64; i++)
492 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
493 		break;
494 	case IXGBE_RDLEN(0):
495 		for (i = 0; i < 64; i++)
496 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
497 		break;
498 	case IXGBE_RDH(0):
499 		for (i = 0; i < 64; i++)
500 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
501 		break;
502 	case IXGBE_RDT(0):
503 		for (i = 0; i < 64; i++)
504 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
505 		break;
506 	case IXGBE_RXDCTL(0):
507 		for (i = 0; i < 64; i++)
508 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
509 		break;
510 	case IXGBE_RDBAL(0):
511 		for (i = 0; i < 64; i++)
512 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
513 		break;
514 	case IXGBE_RDBAH(0):
515 		for (i = 0; i < 64; i++)
516 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
517 		break;
518 	case IXGBE_TDBAL(0):
519 		for (i = 0; i < 64; i++)
520 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
521 		break;
522 	case IXGBE_TDBAH(0):
523 		for (i = 0; i < 64; i++)
524 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
525 		break;
526 	case IXGBE_TDLEN(0):
527 		for (i = 0; i < 64; i++)
528 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
529 		break;
530 	case IXGBE_TDH(0):
531 		for (i = 0; i < 64; i++)
532 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
533 		break;
534 	case IXGBE_TDT(0):
535 		for (i = 0; i < 64; i++)
536 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
537 		break;
538 	case IXGBE_TXDCTL(0):
539 		for (i = 0; i < 64; i++)
540 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
541 		break;
542 	default:
543 		pr_info("%-15s %08x\n", reginfo->name,
544 			IXGBE_READ_REG(hw, reginfo->ofs));
545 		return;
546 	}
547 
548 	for (i = 0; i < 8; i++) {
549 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i*8, i*8+7);
550 		pr_err("%-15s", rname);
551 		for (j = 0; j < 8; j++)
552 			pr_cont(" %08x", regs[i*8+j]);
553 		pr_cont("\n");
554 	}
555 
556 }
557 
558 /*
559  * ixgbe_dump - Print registers, tx-rings and rx-rings
560  */
561 static void ixgbe_dump(struct ixgbe_adapter *adapter)
562 {
563 	struct net_device *netdev = adapter->netdev;
564 	struct ixgbe_hw *hw = &adapter->hw;
565 	struct ixgbe_reg_info *reginfo;
566 	int n = 0;
567 	struct ixgbe_ring *tx_ring;
568 	struct ixgbe_tx_buffer *tx_buffer;
569 	union ixgbe_adv_tx_desc *tx_desc;
570 	struct my_u0 { u64 a; u64 b; } *u0;
571 	struct ixgbe_ring *rx_ring;
572 	union ixgbe_adv_rx_desc *rx_desc;
573 	struct ixgbe_rx_buffer *rx_buffer_info;
574 	u32 staterr;
575 	int i = 0;
576 
577 	if (!netif_msg_hw(adapter))
578 		return;
579 
580 	/* Print netdevice Info */
581 	if (netdev) {
582 		dev_info(&adapter->pdev->dev, "Net device Info\n");
583 		pr_info("Device Name     state            "
584 			"trans_start      last_rx\n");
585 		pr_info("%-15s %016lX %016lX %016lX\n",
586 			netdev->name,
587 			netdev->state,
588 			netdev->trans_start,
589 			netdev->last_rx);
590 	}
591 
592 	/* Print Registers */
593 	dev_info(&adapter->pdev->dev, "Register Dump\n");
594 	pr_info(" Register Name   Value\n");
595 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
596 	     reginfo->name; reginfo++) {
597 		ixgbe_regdump(hw, reginfo);
598 	}
599 
600 	/* Print TX Ring Summary */
601 	if (!netdev || !netif_running(netdev))
602 		return;
603 
604 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
605 	pr_info(" %s     %s              %s        %s\n",
606 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
607 		"leng", "ntw", "timestamp");
608 	for (n = 0; n < adapter->num_tx_queues; n++) {
609 		tx_ring = adapter->tx_ring[n];
610 		tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
611 		pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
612 			   n, tx_ring->next_to_use, tx_ring->next_to_clean,
613 			   (u64)dma_unmap_addr(tx_buffer, dma),
614 			   dma_unmap_len(tx_buffer, len),
615 			   tx_buffer->next_to_watch,
616 			   (u64)tx_buffer->time_stamp);
617 	}
618 
619 	/* Print TX Rings */
620 	if (!netif_msg_tx_done(adapter))
621 		goto rx_ring_summary;
622 
623 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
624 
625 	/* Transmit Descriptor Formats
626 	 *
627 	 * 82598 Advanced Transmit Descriptor
628 	 *   +--------------------------------------------------------------+
629 	 * 0 |         Buffer Address [63:0]                                |
630 	 *   +--------------------------------------------------------------+
631 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
632 	 *   +--------------------------------------------------------------+
633 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
634 	 *
635 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
636 	 *   +--------------------------------------------------------------+
637 	 * 0 |                          RSV [63:0]                          |
638 	 *   +--------------------------------------------------------------+
639 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
640 	 *   +--------------------------------------------------------------+
641 	 *   63                       36 35   32 31                         0
642 	 *
643 	 * 82599+ Advanced Transmit Descriptor
644 	 *   +--------------------------------------------------------------+
645 	 * 0 |         Buffer Address [63:0]                                |
646 	 *   +--------------------------------------------------------------+
647 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
648 	 *   +--------------------------------------------------------------+
649 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
650 	 *
651 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
652 	 *   +--------------------------------------------------------------+
653 	 * 0 |                          RSV [63:0]                          |
654 	 *   +--------------------------------------------------------------+
655 	 * 8 |            RSV           |  STA  |           RSV             |
656 	 *   +--------------------------------------------------------------+
657 	 *   63                       36 35   32 31                         0
658 	 */
659 
660 	for (n = 0; n < adapter->num_tx_queues; n++) {
661 		tx_ring = adapter->tx_ring[n];
662 		pr_info("------------------------------------\n");
663 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
664 		pr_info("------------------------------------\n");
665 		pr_info("%s%s    %s              %s        %s          %s\n",
666 			"T [desc]     [address 63:0  ] ",
667 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
668 			"leng", "ntw", "timestamp", "bi->skb");
669 
670 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
671 			tx_desc = IXGBE_TX_DESC(tx_ring, i);
672 			tx_buffer = &tx_ring->tx_buffer_info[i];
673 			u0 = (struct my_u0 *)tx_desc;
674 			if (dma_unmap_len(tx_buffer, len) > 0) {
675 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p",
676 					i,
677 					le64_to_cpu(u0->a),
678 					le64_to_cpu(u0->b),
679 					(u64)dma_unmap_addr(tx_buffer, dma),
680 					dma_unmap_len(tx_buffer, len),
681 					tx_buffer->next_to_watch,
682 					(u64)tx_buffer->time_stamp,
683 					tx_buffer->skb);
684 				if (i == tx_ring->next_to_use &&
685 					i == tx_ring->next_to_clean)
686 					pr_cont(" NTC/U\n");
687 				else if (i == tx_ring->next_to_use)
688 					pr_cont(" NTU\n");
689 				else if (i == tx_ring->next_to_clean)
690 					pr_cont(" NTC\n");
691 				else
692 					pr_cont("\n");
693 
694 				if (netif_msg_pktdata(adapter) &&
695 				    tx_buffer->skb)
696 					print_hex_dump(KERN_INFO, "",
697 						DUMP_PREFIX_ADDRESS, 16, 1,
698 						tx_buffer->skb->data,
699 						dma_unmap_len(tx_buffer, len),
700 						true);
701 			}
702 		}
703 	}
704 
705 	/* Print RX Rings Summary */
706 rx_ring_summary:
707 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
708 	pr_info("Queue [NTU] [NTC]\n");
709 	for (n = 0; n < adapter->num_rx_queues; n++) {
710 		rx_ring = adapter->rx_ring[n];
711 		pr_info("%5d %5X %5X\n",
712 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
713 	}
714 
715 	/* Print RX Rings */
716 	if (!netif_msg_rx_status(adapter))
717 		return;
718 
719 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
720 
721 	/* Receive Descriptor Formats
722 	 *
723 	 * 82598 Advanced Receive Descriptor (Read) Format
724 	 *    63                                           1        0
725 	 *    +-----------------------------------------------------+
726 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
727 	 *    +----------------------------------------------+------+
728 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
729 	 *    +-----------------------------------------------------+
730 	 *
731 	 *
732 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
733 	 *
734 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
735 	 *   +------------------------------------------------------+
736 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
737 	 *   | Packet   | IP     |   |          |     | Type | Type |
738 	 *   | Checksum | Ident  |   |          |     |      |      |
739 	 *   +------------------------------------------------------+
740 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
741 	 *   +------------------------------------------------------+
742 	 *   63       48 47    32 31            20 19               0
743 	 *
744 	 * 82599+ Advanced Receive Descriptor (Read) Format
745 	 *    63                                           1        0
746 	 *    +-----------------------------------------------------+
747 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
748 	 *    +----------------------------------------------+------+
749 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
750 	 *    +-----------------------------------------------------+
751 	 *
752 	 *
753 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
754 	 *
755 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
756 	 *   +------------------------------------------------------+
757 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
758 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
759 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
760 	 *   +------------------------------------------------------+
761 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
762 	 *   +------------------------------------------------------+
763 	 *   63       48 47    32 31          20 19                 0
764 	 */
765 
766 	for (n = 0; n < adapter->num_rx_queues; n++) {
767 		rx_ring = adapter->rx_ring[n];
768 		pr_info("------------------------------------\n");
769 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
770 		pr_info("------------------------------------\n");
771 		pr_info("%s%s%s",
772 			"R  [desc]      [ PktBuf     A0] ",
773 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
774 			"<-- Adv Rx Read format\n");
775 		pr_info("%s%s%s",
776 			"RWB[desc]      [PcsmIpSHl PtRs] ",
777 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
778 			"<-- Adv Rx Write-Back format\n");
779 
780 		for (i = 0; i < rx_ring->count; i++) {
781 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
782 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
783 			u0 = (struct my_u0 *)rx_desc;
784 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
785 			if (staterr & IXGBE_RXD_STAT_DD) {
786 				/* Descriptor Done */
787 				pr_info("RWB[0x%03X]     %016llX "
788 					"%016llX ---------------- %p", i,
789 					le64_to_cpu(u0->a),
790 					le64_to_cpu(u0->b),
791 					rx_buffer_info->skb);
792 			} else {
793 				pr_info("R  [0x%03X]     %016llX "
794 					"%016llX %016llX %p", i,
795 					le64_to_cpu(u0->a),
796 					le64_to_cpu(u0->b),
797 					(u64)rx_buffer_info->dma,
798 					rx_buffer_info->skb);
799 
800 				if (netif_msg_pktdata(adapter) &&
801 				    rx_buffer_info->dma) {
802 					print_hex_dump(KERN_INFO, "",
803 					   DUMP_PREFIX_ADDRESS, 16, 1,
804 					   page_address(rx_buffer_info->page) +
805 						    rx_buffer_info->page_offset,
806 					   ixgbe_rx_bufsz(rx_ring), true);
807 				}
808 			}
809 
810 			if (i == rx_ring->next_to_use)
811 				pr_cont(" NTU\n");
812 			else if (i == rx_ring->next_to_clean)
813 				pr_cont(" NTC\n");
814 			else
815 				pr_cont("\n");
816 
817 		}
818 	}
819 }
820 
821 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
822 {
823 	u32 ctrl_ext;
824 
825 	/* Let firmware take over control of h/w */
826 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
827 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
828 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
829 }
830 
831 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
832 {
833 	u32 ctrl_ext;
834 
835 	/* Let firmware know the driver has taken over */
836 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
837 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
838 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
839 }
840 
841 /**
842  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
843  * @adapter: pointer to adapter struct
844  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
845  * @queue: queue to map the corresponding interrupt to
846  * @msix_vector: the vector to map to the corresponding queue
847  *
848  */
849 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
850 			   u8 queue, u8 msix_vector)
851 {
852 	u32 ivar, index;
853 	struct ixgbe_hw *hw = &adapter->hw;
854 	switch (hw->mac.type) {
855 	case ixgbe_mac_82598EB:
856 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
857 		if (direction == -1)
858 			direction = 0;
859 		index = (((direction * 64) + queue) >> 2) & 0x1F;
860 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
861 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
862 		ivar |= (msix_vector << (8 * (queue & 0x3)));
863 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
864 		break;
865 	case ixgbe_mac_82599EB:
866 	case ixgbe_mac_X540:
867 	case ixgbe_mac_X550:
868 	case ixgbe_mac_X550EM_x:
869 		if (direction == -1) {
870 			/* other causes */
871 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
872 			index = ((queue & 1) * 8);
873 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
874 			ivar &= ~(0xFF << index);
875 			ivar |= (msix_vector << index);
876 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
877 			break;
878 		} else {
879 			/* tx or rx causes */
880 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
881 			index = ((16 * (queue & 1)) + (8 * direction));
882 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
883 			ivar &= ~(0xFF << index);
884 			ivar |= (msix_vector << index);
885 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
886 			break;
887 		}
888 	default:
889 		break;
890 	}
891 }
892 
893 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
894 					  u64 qmask)
895 {
896 	u32 mask;
897 
898 	switch (adapter->hw.mac.type) {
899 	case ixgbe_mac_82598EB:
900 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
901 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
902 		break;
903 	case ixgbe_mac_82599EB:
904 	case ixgbe_mac_X540:
905 	case ixgbe_mac_X550:
906 	case ixgbe_mac_X550EM_x:
907 		mask = (qmask & 0xFFFFFFFF);
908 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
909 		mask = (qmask >> 32);
910 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
911 		break;
912 	default:
913 		break;
914 	}
915 }
916 
917 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring *ring,
918 				      struct ixgbe_tx_buffer *tx_buffer)
919 {
920 	if (tx_buffer->skb) {
921 		dev_kfree_skb_any(tx_buffer->skb);
922 		if (dma_unmap_len(tx_buffer, len))
923 			dma_unmap_single(ring->dev,
924 					 dma_unmap_addr(tx_buffer, dma),
925 					 dma_unmap_len(tx_buffer, len),
926 					 DMA_TO_DEVICE);
927 	} else if (dma_unmap_len(tx_buffer, len)) {
928 		dma_unmap_page(ring->dev,
929 			       dma_unmap_addr(tx_buffer, dma),
930 			       dma_unmap_len(tx_buffer, len),
931 			       DMA_TO_DEVICE);
932 	}
933 	tx_buffer->next_to_watch = NULL;
934 	tx_buffer->skb = NULL;
935 	dma_unmap_len_set(tx_buffer, len, 0);
936 	/* tx_buffer must be completely set up in the transmit path */
937 }
938 
939 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
940 {
941 	struct ixgbe_hw *hw = &adapter->hw;
942 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
943 	int i;
944 	u32 data;
945 
946 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
947 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
948 		return;
949 
950 	switch (hw->mac.type) {
951 	case ixgbe_mac_82598EB:
952 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
953 		break;
954 	default:
955 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
956 	}
957 	hwstats->lxoffrxc += data;
958 
959 	/* refill credits (no tx hang) if we received xoff */
960 	if (!data)
961 		return;
962 
963 	for (i = 0; i < adapter->num_tx_queues; i++)
964 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
965 			  &adapter->tx_ring[i]->state);
966 }
967 
968 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
969 {
970 	struct ixgbe_hw *hw = &adapter->hw;
971 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
972 	u32 xoff[8] = {0};
973 	u8 tc;
974 	int i;
975 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
976 
977 	if (adapter->ixgbe_ieee_pfc)
978 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
979 
980 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
981 		ixgbe_update_xoff_rx_lfc(adapter);
982 		return;
983 	}
984 
985 	/* update stats for each tc, only valid with PFC enabled */
986 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
987 		u32 pxoffrxc;
988 
989 		switch (hw->mac.type) {
990 		case ixgbe_mac_82598EB:
991 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
992 			break;
993 		default:
994 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
995 		}
996 		hwstats->pxoffrxc[i] += pxoffrxc;
997 		/* Get the TC for given UP */
998 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
999 		xoff[tc] += pxoffrxc;
1000 	}
1001 
1002 	/* disarm tx queues that have received xoff frames */
1003 	for (i = 0; i < adapter->num_tx_queues; i++) {
1004 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
1005 
1006 		tc = tx_ring->dcb_tc;
1007 		if (xoff[tc])
1008 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1009 	}
1010 }
1011 
1012 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1013 {
1014 	return ring->stats.packets;
1015 }
1016 
1017 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1018 {
1019 	struct ixgbe_adapter *adapter;
1020 	struct ixgbe_hw *hw;
1021 	u32 head, tail;
1022 
1023 	if (ring->l2_accel_priv)
1024 		adapter = ring->l2_accel_priv->real_adapter;
1025 	else
1026 		adapter = netdev_priv(ring->netdev);
1027 
1028 	hw = &adapter->hw;
1029 	head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
1030 	tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
1031 
1032 	if (head != tail)
1033 		return (head < tail) ?
1034 			tail - head : (tail + ring->count - head);
1035 
1036 	return 0;
1037 }
1038 
1039 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1040 {
1041 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1042 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1043 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1044 
1045 	clear_check_for_tx_hang(tx_ring);
1046 
1047 	/*
1048 	 * Check for a hung queue, but be thorough. This verifies
1049 	 * that a transmit has been completed since the previous
1050 	 * check AND there is at least one packet pending. The
1051 	 * ARMED bit is set to indicate a potential hang. The
1052 	 * bit is cleared if a pause frame is received to remove
1053 	 * false hang detection due to PFC or 802.3x frames. By
1054 	 * requiring this to fail twice we avoid races with
1055 	 * pfc clearing the ARMED bit and conditions where we
1056 	 * run the check_tx_hang logic with a transmit completion
1057 	 * pending but without time to complete it yet.
1058 	 */
1059 	if (tx_done_old == tx_done && tx_pending)
1060 		/* make sure it is true for two checks in a row */
1061 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1062 					&tx_ring->state);
1063 	/* update completed stats and continue */
1064 	tx_ring->tx_stats.tx_done_old = tx_done;
1065 	/* reset the countdown */
1066 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1067 
1068 	return false;
1069 }
1070 
1071 /**
1072  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1073  * @adapter: driver private struct
1074  **/
1075 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1076 {
1077 
1078 	/* Do the reset outside of interrupt context */
1079 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1080 		adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
1081 		e_warn(drv, "initiating reset due to tx timeout\n");
1082 		ixgbe_service_event_schedule(adapter);
1083 	}
1084 }
1085 
1086 /**
1087  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1088  * @q_vector: structure containing interrupt and ring information
1089  * @tx_ring: tx ring to clean
1090  **/
1091 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1092 			       struct ixgbe_ring *tx_ring)
1093 {
1094 	struct ixgbe_adapter *adapter = q_vector->adapter;
1095 	struct ixgbe_tx_buffer *tx_buffer;
1096 	union ixgbe_adv_tx_desc *tx_desc;
1097 	unsigned int total_bytes = 0, total_packets = 0;
1098 	unsigned int budget = q_vector->tx.work_limit;
1099 	unsigned int i = tx_ring->next_to_clean;
1100 
1101 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1102 		return true;
1103 
1104 	tx_buffer = &tx_ring->tx_buffer_info[i];
1105 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1106 	i -= tx_ring->count;
1107 
1108 	do {
1109 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1110 
1111 		/* if next_to_watch is not set then there is no work pending */
1112 		if (!eop_desc)
1113 			break;
1114 
1115 		/* prevent any other reads prior to eop_desc */
1116 		read_barrier_depends();
1117 
1118 		/* if DD is not set pending work has not been completed */
1119 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1120 			break;
1121 
1122 		/* clear next_to_watch to prevent false hangs */
1123 		tx_buffer->next_to_watch = NULL;
1124 
1125 		/* update the statistics for this packet */
1126 		total_bytes += tx_buffer->bytecount;
1127 		total_packets += tx_buffer->gso_segs;
1128 
1129 		/* free the skb */
1130 		dev_consume_skb_any(tx_buffer->skb);
1131 
1132 		/* unmap skb header data */
1133 		dma_unmap_single(tx_ring->dev,
1134 				 dma_unmap_addr(tx_buffer, dma),
1135 				 dma_unmap_len(tx_buffer, len),
1136 				 DMA_TO_DEVICE);
1137 
1138 		/* clear tx_buffer data */
1139 		tx_buffer->skb = NULL;
1140 		dma_unmap_len_set(tx_buffer, len, 0);
1141 
1142 		/* unmap remaining buffers */
1143 		while (tx_desc != eop_desc) {
1144 			tx_buffer++;
1145 			tx_desc++;
1146 			i++;
1147 			if (unlikely(!i)) {
1148 				i -= tx_ring->count;
1149 				tx_buffer = tx_ring->tx_buffer_info;
1150 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1151 			}
1152 
1153 			/* unmap any remaining paged data */
1154 			if (dma_unmap_len(tx_buffer, len)) {
1155 				dma_unmap_page(tx_ring->dev,
1156 					       dma_unmap_addr(tx_buffer, dma),
1157 					       dma_unmap_len(tx_buffer, len),
1158 					       DMA_TO_DEVICE);
1159 				dma_unmap_len_set(tx_buffer, len, 0);
1160 			}
1161 		}
1162 
1163 		/* move us one more past the eop_desc for start of next pkt */
1164 		tx_buffer++;
1165 		tx_desc++;
1166 		i++;
1167 		if (unlikely(!i)) {
1168 			i -= tx_ring->count;
1169 			tx_buffer = tx_ring->tx_buffer_info;
1170 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1171 		}
1172 
1173 		/* issue prefetch for next Tx descriptor */
1174 		prefetch(tx_desc);
1175 
1176 		/* update budget accounting */
1177 		budget--;
1178 	} while (likely(budget));
1179 
1180 	i += tx_ring->count;
1181 	tx_ring->next_to_clean = i;
1182 	u64_stats_update_begin(&tx_ring->syncp);
1183 	tx_ring->stats.bytes += total_bytes;
1184 	tx_ring->stats.packets += total_packets;
1185 	u64_stats_update_end(&tx_ring->syncp);
1186 	q_vector->tx.total_bytes += total_bytes;
1187 	q_vector->tx.total_packets += total_packets;
1188 
1189 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1190 		/* schedule immediate reset if we believe we hung */
1191 		struct ixgbe_hw *hw = &adapter->hw;
1192 		e_err(drv, "Detected Tx Unit Hang\n"
1193 			"  Tx Queue             <%d>\n"
1194 			"  TDH, TDT             <%x>, <%x>\n"
1195 			"  next_to_use          <%x>\n"
1196 			"  next_to_clean        <%x>\n"
1197 			"tx_buffer_info[next_to_clean]\n"
1198 			"  time_stamp           <%lx>\n"
1199 			"  jiffies              <%lx>\n",
1200 			tx_ring->queue_index,
1201 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1202 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1203 			tx_ring->next_to_use, i,
1204 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1205 
1206 		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
1207 
1208 		e_info(probe,
1209 		       "tx hang %d detected on queue %d, resetting adapter\n",
1210 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1211 
1212 		/* schedule immediate reset if we believe we hung */
1213 		ixgbe_tx_timeout_reset(adapter);
1214 
1215 		/* the adapter is about to reset, no point in enabling stuff */
1216 		return true;
1217 	}
1218 
1219 	netdev_tx_completed_queue(txring_txq(tx_ring),
1220 				  total_packets, total_bytes);
1221 
1222 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1223 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1224 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1225 		/* Make sure that anybody stopping the queue after this
1226 		 * sees the new next_to_clean.
1227 		 */
1228 		smp_mb();
1229 		if (__netif_subqueue_stopped(tx_ring->netdev,
1230 					     tx_ring->queue_index)
1231 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1232 			netif_wake_subqueue(tx_ring->netdev,
1233 					    tx_ring->queue_index);
1234 			++tx_ring->tx_stats.restart_queue;
1235 		}
1236 	}
1237 
1238 	return !!budget;
1239 }
1240 
1241 #ifdef CONFIG_IXGBE_DCA
1242 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1243 				struct ixgbe_ring *tx_ring,
1244 				int cpu)
1245 {
1246 	struct ixgbe_hw *hw = &adapter->hw;
1247 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
1248 	u16 reg_offset;
1249 
1250 	switch (hw->mac.type) {
1251 	case ixgbe_mac_82598EB:
1252 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1253 		break;
1254 	case ixgbe_mac_82599EB:
1255 	case ixgbe_mac_X540:
1256 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1257 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1258 		break;
1259 	default:
1260 		/* for unknown hardware do not write register */
1261 		return;
1262 	}
1263 
1264 	/*
1265 	 * We can enable relaxed ordering for reads, but not writes when
1266 	 * DCA is enabled.  This is due to a known issue in some chipsets
1267 	 * which will cause the DCA tag to be cleared.
1268 	 */
1269 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1270 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1271 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1272 
1273 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1274 }
1275 
1276 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1277 				struct ixgbe_ring *rx_ring,
1278 				int cpu)
1279 {
1280 	struct ixgbe_hw *hw = &adapter->hw;
1281 	u32 rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1282 	u8 reg_idx = rx_ring->reg_idx;
1283 
1284 
1285 	switch (hw->mac.type) {
1286 	case ixgbe_mac_82599EB:
1287 	case ixgbe_mac_X540:
1288 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1289 		break;
1290 	default:
1291 		break;
1292 	}
1293 
1294 	/*
1295 	 * We can enable relaxed ordering for reads, but not writes when
1296 	 * DCA is enabled.  This is due to a known issue in some chipsets
1297 	 * which will cause the DCA tag to be cleared.
1298 	 */
1299 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1300 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1301 
1302 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1303 }
1304 
1305 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1306 {
1307 	struct ixgbe_adapter *adapter = q_vector->adapter;
1308 	struct ixgbe_ring *ring;
1309 	int cpu = get_cpu();
1310 
1311 	if (q_vector->cpu == cpu)
1312 		goto out_no_update;
1313 
1314 	ixgbe_for_each_ring(ring, q_vector->tx)
1315 		ixgbe_update_tx_dca(adapter, ring, cpu);
1316 
1317 	ixgbe_for_each_ring(ring, q_vector->rx)
1318 		ixgbe_update_rx_dca(adapter, ring, cpu);
1319 
1320 	q_vector->cpu = cpu;
1321 out_no_update:
1322 	put_cpu();
1323 }
1324 
1325 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1326 {
1327 	int i;
1328 
1329 	if (!(adapter->flags & IXGBE_FLAG_DCA_ENABLED))
1330 		return;
1331 
1332 	/* always use CB2 mode, difference is masked in the CB driver */
1333 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 2);
1334 
1335 	for (i = 0; i < adapter->num_q_vectors; i++) {
1336 		adapter->q_vector[i]->cpu = -1;
1337 		ixgbe_update_dca(adapter->q_vector[i]);
1338 	}
1339 }
1340 
1341 static int __ixgbe_notify_dca(struct device *dev, void *data)
1342 {
1343 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1344 	unsigned long event = *(unsigned long *)data;
1345 
1346 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1347 		return 0;
1348 
1349 	switch (event) {
1350 	case DCA_PROVIDER_ADD:
1351 		/* if we're already enabled, don't do it again */
1352 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1353 			break;
1354 		if (dca_add_requester(dev) == 0) {
1355 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1356 			ixgbe_setup_dca(adapter);
1357 			break;
1358 		}
1359 		/* Fall Through since DCA is disabled. */
1360 	case DCA_PROVIDER_REMOVE:
1361 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1362 			dca_remove_requester(dev);
1363 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1364 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
1365 		}
1366 		break;
1367 	}
1368 
1369 	return 0;
1370 }
1371 
1372 #endif /* CONFIG_IXGBE_DCA */
1373 
1374 #define IXGBE_RSS_L4_TYPES_MASK \
1375 	((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1376 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1377 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1378 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1379 
1380 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1381 				 union ixgbe_adv_rx_desc *rx_desc,
1382 				 struct sk_buff *skb)
1383 {
1384 	u16 rss_type;
1385 
1386 	if (!(ring->netdev->features & NETIF_F_RXHASH))
1387 		return;
1388 
1389 	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1390 		   IXGBE_RXDADV_RSSTYPE_MASK;
1391 
1392 	if (!rss_type)
1393 		return;
1394 
1395 	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1396 		     (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1397 		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1398 }
1399 
1400 #ifdef IXGBE_FCOE
1401 /**
1402  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1403  * @ring: structure containing ring specific data
1404  * @rx_desc: advanced rx descriptor
1405  *
1406  * Returns : true if it is FCoE pkt
1407  */
1408 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1409 				    union ixgbe_adv_rx_desc *rx_desc)
1410 {
1411 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1412 
1413 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1414 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1415 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1416 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1417 }
1418 
1419 #endif /* IXGBE_FCOE */
1420 /**
1421  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1422  * @ring: structure containing ring specific data
1423  * @rx_desc: current Rx descriptor being processed
1424  * @skb: skb currently being received and modified
1425  **/
1426 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1427 				     union ixgbe_adv_rx_desc *rx_desc,
1428 				     struct sk_buff *skb)
1429 {
1430 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1431 	__le16 hdr_info = rx_desc->wb.lower.lo_dword.hs_rss.hdr_info;
1432 	bool encap_pkt = false;
1433 
1434 	skb_checksum_none_assert(skb);
1435 
1436 	/* Rx csum disabled */
1437 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1438 		return;
1439 
1440 	if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) &&
1441 	    (hdr_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_TUNNEL >> 16))) {
1442 		encap_pkt = true;
1443 		skb->encapsulation = 1;
1444 	}
1445 
1446 	/* if IP and error */
1447 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1448 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1449 		ring->rx_stats.csum_err++;
1450 		return;
1451 	}
1452 
1453 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1454 		return;
1455 
1456 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1457 		/*
1458 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1459 		 * checksum errors.
1460 		 */
1461 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1462 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1463 			return;
1464 
1465 		ring->rx_stats.csum_err++;
1466 		return;
1467 	}
1468 
1469 	/* It must be a TCP or UDP packet with a valid checksum */
1470 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1471 	if (encap_pkt) {
1472 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1473 			return;
1474 
1475 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1476 			ring->rx_stats.csum_err++;
1477 			return;
1478 		}
1479 		/* If we checked the outer header let the stack know */
1480 		skb->csum_level = 1;
1481 	}
1482 }
1483 
1484 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1485 				    struct ixgbe_rx_buffer *bi)
1486 {
1487 	struct page *page = bi->page;
1488 	dma_addr_t dma;
1489 
1490 	/* since we are recycling buffers we should seldom need to alloc */
1491 	if (likely(page))
1492 		return true;
1493 
1494 	/* alloc new page for storage */
1495 	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1496 	if (unlikely(!page)) {
1497 		rx_ring->rx_stats.alloc_rx_page_failed++;
1498 		return false;
1499 	}
1500 
1501 	/* map page for use */
1502 	dma = dma_map_page(rx_ring->dev, page, 0,
1503 			   ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1504 
1505 	/*
1506 	 * if mapping failed free memory back to system since
1507 	 * there isn't much point in holding memory we can't use
1508 	 */
1509 	if (dma_mapping_error(rx_ring->dev, dma)) {
1510 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1511 
1512 		rx_ring->rx_stats.alloc_rx_page_failed++;
1513 		return false;
1514 	}
1515 
1516 	bi->dma = dma;
1517 	bi->page = page;
1518 	bi->page_offset = 0;
1519 
1520 	return true;
1521 }
1522 
1523 /**
1524  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1525  * @rx_ring: ring to place buffers on
1526  * @cleaned_count: number of buffers to replace
1527  **/
1528 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1529 {
1530 	union ixgbe_adv_rx_desc *rx_desc;
1531 	struct ixgbe_rx_buffer *bi;
1532 	u16 i = rx_ring->next_to_use;
1533 
1534 	/* nothing to do */
1535 	if (!cleaned_count)
1536 		return;
1537 
1538 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1539 	bi = &rx_ring->rx_buffer_info[i];
1540 	i -= rx_ring->count;
1541 
1542 	do {
1543 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1544 			break;
1545 
1546 		/*
1547 		 * Refresh the desc even if buffer_addrs didn't change
1548 		 * because each write-back erases this info.
1549 		 */
1550 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1551 
1552 		rx_desc++;
1553 		bi++;
1554 		i++;
1555 		if (unlikely(!i)) {
1556 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1557 			bi = rx_ring->rx_buffer_info;
1558 			i -= rx_ring->count;
1559 		}
1560 
1561 		/* clear the status bits for the next_to_use descriptor */
1562 		rx_desc->wb.upper.status_error = 0;
1563 
1564 		cleaned_count--;
1565 	} while (cleaned_count);
1566 
1567 	i += rx_ring->count;
1568 
1569 	if (rx_ring->next_to_use != i) {
1570 		rx_ring->next_to_use = i;
1571 
1572 		/* update next to alloc since we have filled the ring */
1573 		rx_ring->next_to_alloc = i;
1574 
1575 		/* Force memory writes to complete before letting h/w
1576 		 * know there are new descriptors to fetch.  (Only
1577 		 * applicable for weak-ordered memory model archs,
1578 		 * such as IA-64).
1579 		 */
1580 		wmb();
1581 		writel(i, rx_ring->tail);
1582 	}
1583 }
1584 
1585 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1586 				   struct sk_buff *skb)
1587 {
1588 	u16 hdr_len = skb_headlen(skb);
1589 
1590 	/* set gso_size to avoid messing up TCP MSS */
1591 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1592 						 IXGBE_CB(skb)->append_cnt);
1593 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1594 }
1595 
1596 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1597 				   struct sk_buff *skb)
1598 {
1599 	/* if append_cnt is 0 then frame is not RSC */
1600 	if (!IXGBE_CB(skb)->append_cnt)
1601 		return;
1602 
1603 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1604 	rx_ring->rx_stats.rsc_flush++;
1605 
1606 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1607 
1608 	/* gso_size is computed using append_cnt so always clear it last */
1609 	IXGBE_CB(skb)->append_cnt = 0;
1610 }
1611 
1612 /**
1613  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1614  * @rx_ring: rx descriptor ring packet is being transacted on
1615  * @rx_desc: pointer to the EOP Rx descriptor
1616  * @skb: pointer to current skb being populated
1617  *
1618  * This function checks the ring, descriptor, and packet information in
1619  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1620  * other fields within the skb.
1621  **/
1622 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1623 				     union ixgbe_adv_rx_desc *rx_desc,
1624 				     struct sk_buff *skb)
1625 {
1626 	struct net_device *dev = rx_ring->netdev;
1627 
1628 	ixgbe_update_rsc_stats(rx_ring, skb);
1629 
1630 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1631 
1632 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1633 
1634 	if (unlikely(ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_TS)))
1635 		ixgbe_ptp_rx_hwtstamp(rx_ring->q_vector->adapter, skb);
1636 
1637 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1638 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1639 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1640 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1641 	}
1642 
1643 	skb_record_rx_queue(skb, rx_ring->queue_index);
1644 
1645 	skb->protocol = eth_type_trans(skb, dev);
1646 }
1647 
1648 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1649 			 struct sk_buff *skb)
1650 {
1651 	if (ixgbe_qv_busy_polling(q_vector))
1652 		netif_receive_skb(skb);
1653 	else
1654 		napi_gro_receive(&q_vector->napi, skb);
1655 }
1656 
1657 /**
1658  * ixgbe_is_non_eop - process handling of non-EOP buffers
1659  * @rx_ring: Rx ring being processed
1660  * @rx_desc: Rx descriptor for current buffer
1661  * @skb: Current socket buffer containing buffer in progress
1662  *
1663  * This function updates next to clean.  If the buffer is an EOP buffer
1664  * this function exits returning false, otherwise it will place the
1665  * sk_buff in the next buffer to be chained and return true indicating
1666  * that this is in fact a non-EOP buffer.
1667  **/
1668 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1669 			     union ixgbe_adv_rx_desc *rx_desc,
1670 			     struct sk_buff *skb)
1671 {
1672 	u32 ntc = rx_ring->next_to_clean + 1;
1673 
1674 	/* fetch, update, and store next to clean */
1675 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1676 	rx_ring->next_to_clean = ntc;
1677 
1678 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1679 
1680 	/* update RSC append count if present */
1681 	if (ring_is_rsc_enabled(rx_ring)) {
1682 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1683 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1684 
1685 		if (unlikely(rsc_enabled)) {
1686 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1687 
1688 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1689 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1690 
1691 			/* update ntc based on RSC value */
1692 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1693 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1694 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1695 		}
1696 	}
1697 
1698 	/* if we are the last buffer then there is nothing else to do */
1699 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1700 		return false;
1701 
1702 	/* place skb in next buffer to be received */
1703 	rx_ring->rx_buffer_info[ntc].skb = skb;
1704 	rx_ring->rx_stats.non_eop_descs++;
1705 
1706 	return true;
1707 }
1708 
1709 /**
1710  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1711  * @rx_ring: rx descriptor ring packet is being transacted on
1712  * @skb: pointer to current skb being adjusted
1713  *
1714  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1715  * main difference between this version and the original function is that
1716  * this function can make several assumptions about the state of things
1717  * that allow for significant optimizations versus the standard function.
1718  * As a result we can do things like drop a frag and maintain an accurate
1719  * truesize for the skb.
1720  */
1721 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1722 			    struct sk_buff *skb)
1723 {
1724 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1725 	unsigned char *va;
1726 	unsigned int pull_len;
1727 
1728 	/*
1729 	 * it is valid to use page_address instead of kmap since we are
1730 	 * working with pages allocated out of the lomem pool per
1731 	 * alloc_page(GFP_ATOMIC)
1732 	 */
1733 	va = skb_frag_address(frag);
1734 
1735 	/*
1736 	 * we need the header to contain the greater of either ETH_HLEN or
1737 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1738 	 */
1739 	pull_len = eth_get_headlen(va, IXGBE_RX_HDR_SIZE);
1740 
1741 	/* align pull length to size of long to optimize memcpy performance */
1742 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1743 
1744 	/* update all of the pointers */
1745 	skb_frag_size_sub(frag, pull_len);
1746 	frag->page_offset += pull_len;
1747 	skb->data_len -= pull_len;
1748 	skb->tail += pull_len;
1749 }
1750 
1751 /**
1752  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1753  * @rx_ring: rx descriptor ring packet is being transacted on
1754  * @skb: pointer to current skb being updated
1755  *
1756  * This function provides a basic DMA sync up for the first fragment of an
1757  * skb.  The reason for doing this is that the first fragment cannot be
1758  * unmapped until we have reached the end of packet descriptor for a buffer
1759  * chain.
1760  */
1761 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1762 				struct sk_buff *skb)
1763 {
1764 	/* if the page was released unmap it, else just sync our portion */
1765 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1766 		dma_unmap_page(rx_ring->dev, IXGBE_CB(skb)->dma,
1767 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1768 		IXGBE_CB(skb)->page_released = false;
1769 	} else {
1770 		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1771 
1772 		dma_sync_single_range_for_cpu(rx_ring->dev,
1773 					      IXGBE_CB(skb)->dma,
1774 					      frag->page_offset,
1775 					      ixgbe_rx_bufsz(rx_ring),
1776 					      DMA_FROM_DEVICE);
1777 	}
1778 	IXGBE_CB(skb)->dma = 0;
1779 }
1780 
1781 /**
1782  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1783  * @rx_ring: rx descriptor ring packet is being transacted on
1784  * @rx_desc: pointer to the EOP Rx descriptor
1785  * @skb: pointer to current skb being fixed
1786  *
1787  * Check for corrupted packet headers caused by senders on the local L2
1788  * embedded NIC switch not setting up their Tx Descriptors right.  These
1789  * should be very rare.
1790  *
1791  * Also address the case where we are pulling data in on pages only
1792  * and as such no data is present in the skb header.
1793  *
1794  * In addition if skb is not at least 60 bytes we need to pad it so that
1795  * it is large enough to qualify as a valid Ethernet frame.
1796  *
1797  * Returns true if an error was encountered and skb was freed.
1798  **/
1799 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1800 				  union ixgbe_adv_rx_desc *rx_desc,
1801 				  struct sk_buff *skb)
1802 {
1803 	struct net_device *netdev = rx_ring->netdev;
1804 
1805 	/* verify that the packet does not have any known errors */
1806 	if (unlikely(ixgbe_test_staterr(rx_desc,
1807 					IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1808 	    !(netdev->features & NETIF_F_RXALL))) {
1809 		dev_kfree_skb_any(skb);
1810 		return true;
1811 	}
1812 
1813 	/* place header in linear portion of buffer */
1814 	if (skb_is_nonlinear(skb))
1815 		ixgbe_pull_tail(rx_ring, skb);
1816 
1817 #ifdef IXGBE_FCOE
1818 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1819 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1820 		return false;
1821 
1822 #endif
1823 	/* if eth_skb_pad returns an error the skb was freed */
1824 	if (eth_skb_pad(skb))
1825 		return true;
1826 
1827 	return false;
1828 }
1829 
1830 /**
1831  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1832  * @rx_ring: rx descriptor ring to store buffers on
1833  * @old_buff: donor buffer to have page reused
1834  *
1835  * Synchronizes page for reuse by the adapter
1836  **/
1837 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1838 				struct ixgbe_rx_buffer *old_buff)
1839 {
1840 	struct ixgbe_rx_buffer *new_buff;
1841 	u16 nta = rx_ring->next_to_alloc;
1842 
1843 	new_buff = &rx_ring->rx_buffer_info[nta];
1844 
1845 	/* update, and store next to alloc */
1846 	nta++;
1847 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1848 
1849 	/* transfer page from old buffer to new buffer */
1850 	*new_buff = *old_buff;
1851 
1852 	/* sync the buffer for use by the device */
1853 	dma_sync_single_range_for_device(rx_ring->dev, new_buff->dma,
1854 					 new_buff->page_offset,
1855 					 ixgbe_rx_bufsz(rx_ring),
1856 					 DMA_FROM_DEVICE);
1857 }
1858 
1859 static inline bool ixgbe_page_is_reserved(struct page *page)
1860 {
1861 	return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
1862 }
1863 
1864 /**
1865  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1866  * @rx_ring: rx descriptor ring to transact packets on
1867  * @rx_buffer: buffer containing page to add
1868  * @rx_desc: descriptor containing length of buffer written by hardware
1869  * @skb: sk_buff to place the data into
1870  *
1871  * This function will add the data contained in rx_buffer->page to the skb.
1872  * This is done either through a direct copy if the data in the buffer is
1873  * less than the skb header size, otherwise it will just attach the page as
1874  * a frag to the skb.
1875  *
1876  * The function will then update the page offset if necessary and return
1877  * true if the buffer can be reused by the adapter.
1878  **/
1879 static bool ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1880 			      struct ixgbe_rx_buffer *rx_buffer,
1881 			      union ixgbe_adv_rx_desc *rx_desc,
1882 			      struct sk_buff *skb)
1883 {
1884 	struct page *page = rx_buffer->page;
1885 	unsigned int size = le16_to_cpu(rx_desc->wb.upper.length);
1886 #if (PAGE_SIZE < 8192)
1887 	unsigned int truesize = ixgbe_rx_bufsz(rx_ring);
1888 #else
1889 	unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
1890 	unsigned int last_offset = ixgbe_rx_pg_size(rx_ring) -
1891 				   ixgbe_rx_bufsz(rx_ring);
1892 #endif
1893 
1894 	if ((size <= IXGBE_RX_HDR_SIZE) && !skb_is_nonlinear(skb)) {
1895 		unsigned char *va = page_address(page) + rx_buffer->page_offset;
1896 
1897 		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
1898 
1899 		/* page is not reserved, we can reuse buffer as-is */
1900 		if (likely(!ixgbe_page_is_reserved(page)))
1901 			return true;
1902 
1903 		/* this page cannot be reused so discard it */
1904 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1905 		return false;
1906 	}
1907 
1908 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
1909 			rx_buffer->page_offset, size, truesize);
1910 
1911 	/* avoid re-using remote pages */
1912 	if (unlikely(ixgbe_page_is_reserved(page)))
1913 		return false;
1914 
1915 #if (PAGE_SIZE < 8192)
1916 	/* if we are only owner of page we can reuse it */
1917 	if (unlikely(page_count(page) != 1))
1918 		return false;
1919 
1920 	/* flip page offset to other buffer */
1921 	rx_buffer->page_offset ^= truesize;
1922 #else
1923 	/* move offset up to the next cache line */
1924 	rx_buffer->page_offset += truesize;
1925 
1926 	if (rx_buffer->page_offset > last_offset)
1927 		return false;
1928 #endif
1929 
1930 	/* Even if we own the page, we are not allowed to use atomic_set()
1931 	 * This would break get_page_unless_zero() users.
1932 	 */
1933 	atomic_inc(&page->_count);
1934 
1935 	return true;
1936 }
1937 
1938 static struct sk_buff *ixgbe_fetch_rx_buffer(struct ixgbe_ring *rx_ring,
1939 					     union ixgbe_adv_rx_desc *rx_desc)
1940 {
1941 	struct ixgbe_rx_buffer *rx_buffer;
1942 	struct sk_buff *skb;
1943 	struct page *page;
1944 
1945 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
1946 	page = rx_buffer->page;
1947 	prefetchw(page);
1948 
1949 	skb = rx_buffer->skb;
1950 
1951 	if (likely(!skb)) {
1952 		void *page_addr = page_address(page) +
1953 				  rx_buffer->page_offset;
1954 
1955 		/* prefetch first cache line of first page */
1956 		prefetch(page_addr);
1957 #if L1_CACHE_BYTES < 128
1958 		prefetch(page_addr + L1_CACHE_BYTES);
1959 #endif
1960 
1961 		/* allocate a skb to store the frags */
1962 		skb = napi_alloc_skb(&rx_ring->q_vector->napi,
1963 				     IXGBE_RX_HDR_SIZE);
1964 		if (unlikely(!skb)) {
1965 			rx_ring->rx_stats.alloc_rx_buff_failed++;
1966 			return NULL;
1967 		}
1968 
1969 		/*
1970 		 * we will be copying header into skb->data in
1971 		 * pskb_may_pull so it is in our interest to prefetch
1972 		 * it now to avoid a possible cache miss
1973 		 */
1974 		prefetchw(skb->data);
1975 
1976 		/*
1977 		 * Delay unmapping of the first packet. It carries the
1978 		 * header information, HW may still access the header
1979 		 * after the writeback.  Only unmap it when EOP is
1980 		 * reached
1981 		 */
1982 		if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1983 			goto dma_sync;
1984 
1985 		IXGBE_CB(skb)->dma = rx_buffer->dma;
1986 	} else {
1987 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
1988 			ixgbe_dma_sync_frag(rx_ring, skb);
1989 
1990 dma_sync:
1991 		/* we are reusing so sync this buffer for CPU use */
1992 		dma_sync_single_range_for_cpu(rx_ring->dev,
1993 					      rx_buffer->dma,
1994 					      rx_buffer->page_offset,
1995 					      ixgbe_rx_bufsz(rx_ring),
1996 					      DMA_FROM_DEVICE);
1997 
1998 		rx_buffer->skb = NULL;
1999 	}
2000 
2001 	/* pull page into skb */
2002 	if (ixgbe_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) {
2003 		/* hand second half of page back to the ring */
2004 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2005 	} else if (IXGBE_CB(skb)->dma == rx_buffer->dma) {
2006 		/* the page has been released from the ring */
2007 		IXGBE_CB(skb)->page_released = true;
2008 	} else {
2009 		/* we are not reusing the buffer so unmap it */
2010 		dma_unmap_page(rx_ring->dev, rx_buffer->dma,
2011 			       ixgbe_rx_pg_size(rx_ring),
2012 			       DMA_FROM_DEVICE);
2013 	}
2014 
2015 	/* clear contents of buffer_info */
2016 	rx_buffer->page = NULL;
2017 
2018 	return skb;
2019 }
2020 
2021 /**
2022  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2023  * @q_vector: structure containing interrupt and ring information
2024  * @rx_ring: rx descriptor ring to transact packets on
2025  * @budget: Total limit on number of packets to process
2026  *
2027  * This function provides a "bounce buffer" approach to Rx interrupt
2028  * processing.  The advantage to this is that on systems that have
2029  * expensive overhead for IOMMU access this provides a means of avoiding
2030  * it by maintaining the mapping of the page to the syste.
2031  *
2032  * Returns amount of work completed
2033  **/
2034 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2035 			       struct ixgbe_ring *rx_ring,
2036 			       const int budget)
2037 {
2038 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
2039 #ifdef IXGBE_FCOE
2040 	struct ixgbe_adapter *adapter = q_vector->adapter;
2041 	int ddp_bytes;
2042 	unsigned int mss = 0;
2043 #endif /* IXGBE_FCOE */
2044 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2045 
2046 	while (likely(total_rx_packets < budget)) {
2047 		union ixgbe_adv_rx_desc *rx_desc;
2048 		struct sk_buff *skb;
2049 
2050 		/* return some buffers to hardware, one at a time is too slow */
2051 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2052 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2053 			cleaned_count = 0;
2054 		}
2055 
2056 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2057 
2058 		if (!rx_desc->wb.upper.status_error)
2059 			break;
2060 
2061 		/* This memory barrier is needed to keep us from reading
2062 		 * any other fields out of the rx_desc until we know the
2063 		 * descriptor has been written back
2064 		 */
2065 		dma_rmb();
2066 
2067 		/* retrieve a buffer from the ring */
2068 		skb = ixgbe_fetch_rx_buffer(rx_ring, rx_desc);
2069 
2070 		/* exit if we failed to retrieve a buffer */
2071 		if (!skb)
2072 			break;
2073 
2074 		cleaned_count++;
2075 
2076 		/* place incomplete frames back on ring for completion */
2077 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2078 			continue;
2079 
2080 		/* verify the packet layout is correct */
2081 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2082 			continue;
2083 
2084 		/* probably a little skewed due to removing CRC */
2085 		total_rx_bytes += skb->len;
2086 
2087 		/* populate checksum, timestamp, VLAN, and protocol */
2088 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2089 
2090 #ifdef IXGBE_FCOE
2091 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2092 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2093 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2094 			/* include DDPed FCoE data */
2095 			if (ddp_bytes > 0) {
2096 				if (!mss) {
2097 					mss = rx_ring->netdev->mtu -
2098 						sizeof(struct fcoe_hdr) -
2099 						sizeof(struct fc_frame_header) -
2100 						sizeof(struct fcoe_crc_eof);
2101 					if (mss > 512)
2102 						mss &= ~511;
2103 				}
2104 				total_rx_bytes += ddp_bytes;
2105 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2106 								 mss);
2107 			}
2108 			if (!ddp_bytes) {
2109 				dev_kfree_skb_any(skb);
2110 				continue;
2111 			}
2112 		}
2113 
2114 #endif /* IXGBE_FCOE */
2115 		skb_mark_napi_id(skb, &q_vector->napi);
2116 		ixgbe_rx_skb(q_vector, skb);
2117 
2118 		/* update budget accounting */
2119 		total_rx_packets++;
2120 	}
2121 
2122 	u64_stats_update_begin(&rx_ring->syncp);
2123 	rx_ring->stats.packets += total_rx_packets;
2124 	rx_ring->stats.bytes += total_rx_bytes;
2125 	u64_stats_update_end(&rx_ring->syncp);
2126 	q_vector->rx.total_packets += total_rx_packets;
2127 	q_vector->rx.total_bytes += total_rx_bytes;
2128 
2129 	return total_rx_packets;
2130 }
2131 
2132 #ifdef CONFIG_NET_RX_BUSY_POLL
2133 /* must be called with local_bh_disable()d */
2134 static int ixgbe_low_latency_recv(struct napi_struct *napi)
2135 {
2136 	struct ixgbe_q_vector *q_vector =
2137 			container_of(napi, struct ixgbe_q_vector, napi);
2138 	struct ixgbe_adapter *adapter = q_vector->adapter;
2139 	struct ixgbe_ring  *ring;
2140 	int found = 0;
2141 
2142 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2143 		return LL_FLUSH_FAILED;
2144 
2145 	if (!ixgbe_qv_lock_poll(q_vector))
2146 		return LL_FLUSH_BUSY;
2147 
2148 	ixgbe_for_each_ring(ring, q_vector->rx) {
2149 		found = ixgbe_clean_rx_irq(q_vector, ring, 4);
2150 #ifdef BP_EXTENDED_STATS
2151 		if (found)
2152 			ring->stats.cleaned += found;
2153 		else
2154 			ring->stats.misses++;
2155 #endif
2156 		if (found)
2157 			break;
2158 	}
2159 
2160 	ixgbe_qv_unlock_poll(q_vector);
2161 
2162 	return found;
2163 }
2164 #endif	/* CONFIG_NET_RX_BUSY_POLL */
2165 
2166 /**
2167  * ixgbe_configure_msix - Configure MSI-X hardware
2168  * @adapter: board private structure
2169  *
2170  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2171  * interrupts.
2172  **/
2173 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2174 {
2175 	struct ixgbe_q_vector *q_vector;
2176 	int v_idx;
2177 	u32 mask;
2178 
2179 	/* Populate MSIX to EITR Select */
2180 	if (adapter->num_vfs > 32) {
2181 		u32 eitrsel = (1 << (adapter->num_vfs - 32)) - 1;
2182 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2183 	}
2184 
2185 	/*
2186 	 * Populate the IVAR table and set the ITR values to the
2187 	 * corresponding register.
2188 	 */
2189 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2190 		struct ixgbe_ring *ring;
2191 		q_vector = adapter->q_vector[v_idx];
2192 
2193 		ixgbe_for_each_ring(ring, q_vector->rx)
2194 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2195 
2196 		ixgbe_for_each_ring(ring, q_vector->tx)
2197 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2198 
2199 		ixgbe_write_eitr(q_vector);
2200 	}
2201 
2202 	switch (adapter->hw.mac.type) {
2203 	case ixgbe_mac_82598EB:
2204 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2205 			       v_idx);
2206 		break;
2207 	case ixgbe_mac_82599EB:
2208 	case ixgbe_mac_X540:
2209 	case ixgbe_mac_X550:
2210 	case ixgbe_mac_X550EM_x:
2211 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2212 		break;
2213 	default:
2214 		break;
2215 	}
2216 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2217 
2218 	/* set up to autoclear timer, and the vectors */
2219 	mask = IXGBE_EIMS_ENABLE_MASK;
2220 	mask &= ~(IXGBE_EIMS_OTHER |
2221 		  IXGBE_EIMS_MAILBOX |
2222 		  IXGBE_EIMS_LSC);
2223 
2224 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2225 }
2226 
2227 enum latency_range {
2228 	lowest_latency = 0,
2229 	low_latency = 1,
2230 	bulk_latency = 2,
2231 	latency_invalid = 255
2232 };
2233 
2234 /**
2235  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2236  * @q_vector: structure containing interrupt and ring information
2237  * @ring_container: structure containing ring performance data
2238  *
2239  *      Stores a new ITR value based on packets and byte
2240  *      counts during the last interrupt.  The advantage of per interrupt
2241  *      computation is faster updates and more accurate ITR for the current
2242  *      traffic pattern.  Constants in this function were computed
2243  *      based on theoretical maximum wire speed and thresholds were set based
2244  *      on testing data as well as attempting to minimize response time
2245  *      while increasing bulk throughput.
2246  *      this functionality is controlled by the InterruptThrottleRate module
2247  *      parameter (see ixgbe_param.c)
2248  **/
2249 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2250 			     struct ixgbe_ring_container *ring_container)
2251 {
2252 	int bytes = ring_container->total_bytes;
2253 	int packets = ring_container->total_packets;
2254 	u32 timepassed_us;
2255 	u64 bytes_perint;
2256 	u8 itr_setting = ring_container->itr;
2257 
2258 	if (packets == 0)
2259 		return;
2260 
2261 	/* simple throttlerate management
2262 	 *   0-10MB/s   lowest (100000 ints/s)
2263 	 *  10-20MB/s   low    (20000 ints/s)
2264 	 *  20-1249MB/s bulk   (12000 ints/s)
2265 	 */
2266 	/* what was last interrupt timeslice? */
2267 	timepassed_us = q_vector->itr >> 2;
2268 	if (timepassed_us == 0)
2269 		return;
2270 
2271 	bytes_perint = bytes / timepassed_us; /* bytes/usec */
2272 
2273 	switch (itr_setting) {
2274 	case lowest_latency:
2275 		if (bytes_perint > 10)
2276 			itr_setting = low_latency;
2277 		break;
2278 	case low_latency:
2279 		if (bytes_perint > 20)
2280 			itr_setting = bulk_latency;
2281 		else if (bytes_perint <= 10)
2282 			itr_setting = lowest_latency;
2283 		break;
2284 	case bulk_latency:
2285 		if (bytes_perint <= 20)
2286 			itr_setting = low_latency;
2287 		break;
2288 	}
2289 
2290 	/* clear work counters since we have the values we need */
2291 	ring_container->total_bytes = 0;
2292 	ring_container->total_packets = 0;
2293 
2294 	/* write updated itr to ring container */
2295 	ring_container->itr = itr_setting;
2296 }
2297 
2298 /**
2299  * ixgbe_write_eitr - write EITR register in hardware specific way
2300  * @q_vector: structure containing interrupt and ring information
2301  *
2302  * This function is made to be called by ethtool and by the driver
2303  * when it needs to update EITR registers at runtime.  Hardware
2304  * specific quirks/differences are taken care of here.
2305  */
2306 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2307 {
2308 	struct ixgbe_adapter *adapter = q_vector->adapter;
2309 	struct ixgbe_hw *hw = &adapter->hw;
2310 	int v_idx = q_vector->v_idx;
2311 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2312 
2313 	switch (adapter->hw.mac.type) {
2314 	case ixgbe_mac_82598EB:
2315 		/* must write high and low 16 bits to reset counter */
2316 		itr_reg |= (itr_reg << 16);
2317 		break;
2318 	case ixgbe_mac_82599EB:
2319 	case ixgbe_mac_X540:
2320 	case ixgbe_mac_X550:
2321 	case ixgbe_mac_X550EM_x:
2322 		/*
2323 		 * set the WDIS bit to not clear the timer bits and cause an
2324 		 * immediate assertion of the interrupt
2325 		 */
2326 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2327 		break;
2328 	default:
2329 		break;
2330 	}
2331 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2332 }
2333 
2334 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2335 {
2336 	u32 new_itr = q_vector->itr;
2337 	u8 current_itr;
2338 
2339 	ixgbe_update_itr(q_vector, &q_vector->tx);
2340 	ixgbe_update_itr(q_vector, &q_vector->rx);
2341 
2342 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
2343 
2344 	switch (current_itr) {
2345 	/* counts and packets in update_itr are dependent on these numbers */
2346 	case lowest_latency:
2347 		new_itr = IXGBE_100K_ITR;
2348 		break;
2349 	case low_latency:
2350 		new_itr = IXGBE_20K_ITR;
2351 		break;
2352 	case bulk_latency:
2353 		new_itr = IXGBE_12K_ITR;
2354 		break;
2355 	default:
2356 		break;
2357 	}
2358 
2359 	if (new_itr != q_vector->itr) {
2360 		/* do an exponential smoothing */
2361 		new_itr = (10 * new_itr * q_vector->itr) /
2362 			  ((9 * new_itr) + q_vector->itr);
2363 
2364 		/* save the algorithm value here */
2365 		q_vector->itr = new_itr;
2366 
2367 		ixgbe_write_eitr(q_vector);
2368 	}
2369 }
2370 
2371 /**
2372  * ixgbe_check_overtemp_subtask - check for over temperature
2373  * @adapter: pointer to adapter
2374  **/
2375 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2376 {
2377 	struct ixgbe_hw *hw = &adapter->hw;
2378 	u32 eicr = adapter->interrupt_event;
2379 
2380 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2381 		return;
2382 
2383 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) &&
2384 	    !(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2385 		return;
2386 
2387 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2388 
2389 	switch (hw->device_id) {
2390 	case IXGBE_DEV_ID_82599_T3_LOM:
2391 		/*
2392 		 * Since the warning interrupt is for both ports
2393 		 * we don't have to check if:
2394 		 *  - This interrupt wasn't for our port.
2395 		 *  - We may have missed the interrupt so always have to
2396 		 *    check if we  got a LSC
2397 		 */
2398 		if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2399 		    !(eicr & IXGBE_EICR_LSC))
2400 			return;
2401 
2402 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2403 			u32 speed;
2404 			bool link_up = false;
2405 
2406 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2407 
2408 			if (link_up)
2409 				return;
2410 		}
2411 
2412 		/* Check if this is not due to overtemp */
2413 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2414 			return;
2415 
2416 		break;
2417 	default:
2418 		if (adapter->hw.mac.type >= ixgbe_mac_X540)
2419 			return;
2420 		if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
2421 			return;
2422 		break;
2423 	}
2424 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2425 
2426 	adapter->interrupt_event = 0;
2427 }
2428 
2429 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2430 {
2431 	struct ixgbe_hw *hw = &adapter->hw;
2432 
2433 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2434 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2435 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2436 		/* write to clear the interrupt */
2437 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2438 	}
2439 }
2440 
2441 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2442 {
2443 	struct ixgbe_hw *hw = &adapter->hw;
2444 
2445 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2446 		return;
2447 
2448 	switch (adapter->hw.mac.type) {
2449 	case ixgbe_mac_82599EB:
2450 		/*
2451 		 * Need to check link state so complete overtemp check
2452 		 * on service task
2453 		 */
2454 		if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
2455 		     (eicr & IXGBE_EICR_LSC)) &&
2456 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2457 			adapter->interrupt_event = eicr;
2458 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2459 			ixgbe_service_event_schedule(adapter);
2460 			return;
2461 		}
2462 		return;
2463 	case ixgbe_mac_X540:
2464 		if (!(eicr & IXGBE_EICR_TS))
2465 			return;
2466 		break;
2467 	default:
2468 		return;
2469 	}
2470 
2471 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2472 }
2473 
2474 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
2475 {
2476 	switch (hw->mac.type) {
2477 	case ixgbe_mac_82598EB:
2478 		if (hw->phy.type == ixgbe_phy_nl)
2479 			return true;
2480 		return false;
2481 	case ixgbe_mac_82599EB:
2482 	case ixgbe_mac_X550EM_x:
2483 		switch (hw->mac.ops.get_media_type(hw)) {
2484 		case ixgbe_media_type_fiber:
2485 		case ixgbe_media_type_fiber_qsfp:
2486 			return true;
2487 		default:
2488 			return false;
2489 		}
2490 	default:
2491 		return false;
2492 	}
2493 }
2494 
2495 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2496 {
2497 	struct ixgbe_hw *hw = &adapter->hw;
2498 	u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
2499 
2500 	if (!ixgbe_is_sfp(hw))
2501 		return;
2502 
2503 	/* Later MAC's use different SDP */
2504 	if (hw->mac.type >= ixgbe_mac_X540)
2505 		eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2506 
2507 	if (eicr & eicr_mask) {
2508 		/* Clear the interrupt */
2509 		IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2510 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2511 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2512 			ixgbe_service_event_schedule(adapter);
2513 		}
2514 	}
2515 
2516 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
2517 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2518 		/* Clear the interrupt */
2519 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2520 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2521 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2522 			ixgbe_service_event_schedule(adapter);
2523 		}
2524 	}
2525 }
2526 
2527 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2528 {
2529 	struct ixgbe_hw *hw = &adapter->hw;
2530 
2531 	adapter->lsc_int++;
2532 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2533 	adapter->link_check_timeout = jiffies;
2534 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2535 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2536 		IXGBE_WRITE_FLUSH(hw);
2537 		ixgbe_service_event_schedule(adapter);
2538 	}
2539 }
2540 
2541 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2542 					   u64 qmask)
2543 {
2544 	u32 mask;
2545 	struct ixgbe_hw *hw = &adapter->hw;
2546 
2547 	switch (hw->mac.type) {
2548 	case ixgbe_mac_82598EB:
2549 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2550 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2551 		break;
2552 	case ixgbe_mac_82599EB:
2553 	case ixgbe_mac_X540:
2554 	case ixgbe_mac_X550:
2555 	case ixgbe_mac_X550EM_x:
2556 		mask = (qmask & 0xFFFFFFFF);
2557 		if (mask)
2558 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2559 		mask = (qmask >> 32);
2560 		if (mask)
2561 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2562 		break;
2563 	default:
2564 		break;
2565 	}
2566 	/* skip the flush */
2567 }
2568 
2569 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2570 					    u64 qmask)
2571 {
2572 	u32 mask;
2573 	struct ixgbe_hw *hw = &adapter->hw;
2574 
2575 	switch (hw->mac.type) {
2576 	case ixgbe_mac_82598EB:
2577 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2578 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2579 		break;
2580 	case ixgbe_mac_82599EB:
2581 	case ixgbe_mac_X540:
2582 	case ixgbe_mac_X550:
2583 	case ixgbe_mac_X550EM_x:
2584 		mask = (qmask & 0xFFFFFFFF);
2585 		if (mask)
2586 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2587 		mask = (qmask >> 32);
2588 		if (mask)
2589 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2590 		break;
2591 	default:
2592 		break;
2593 	}
2594 	/* skip the flush */
2595 }
2596 
2597 /**
2598  * ixgbe_irq_enable - Enable default interrupt generation settings
2599  * @adapter: board private structure
2600  **/
2601 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2602 				    bool flush)
2603 {
2604 	struct ixgbe_hw *hw = &adapter->hw;
2605 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2606 
2607 	/* don't reenable LSC while waiting for link */
2608 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2609 		mask &= ~IXGBE_EIMS_LSC;
2610 
2611 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2612 		switch (adapter->hw.mac.type) {
2613 		case ixgbe_mac_82599EB:
2614 			mask |= IXGBE_EIMS_GPI_SDP0(hw);
2615 			break;
2616 		case ixgbe_mac_X540:
2617 		case ixgbe_mac_X550:
2618 		case ixgbe_mac_X550EM_x:
2619 			mask |= IXGBE_EIMS_TS;
2620 			break;
2621 		default:
2622 			break;
2623 		}
2624 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2625 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
2626 	switch (adapter->hw.mac.type) {
2627 	case ixgbe_mac_82599EB:
2628 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
2629 		mask |= IXGBE_EIMS_GPI_SDP2(hw);
2630 		/* fall through */
2631 	case ixgbe_mac_X540:
2632 	case ixgbe_mac_X550:
2633 	case ixgbe_mac_X550EM_x:
2634 		if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
2635 			mask |= IXGBE_EICR_GPI_SDP0_X540;
2636 		mask |= IXGBE_EIMS_ECC;
2637 		mask |= IXGBE_EIMS_MAILBOX;
2638 		break;
2639 	default:
2640 		break;
2641 	}
2642 
2643 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2644 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
2645 		mask |= IXGBE_EIMS_FLOW_DIR;
2646 
2647 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
2648 	if (queues)
2649 		ixgbe_irq_enable_queues(adapter, ~0);
2650 	if (flush)
2651 		IXGBE_WRITE_FLUSH(&adapter->hw);
2652 }
2653 
2654 static irqreturn_t ixgbe_msix_other(int irq, void *data)
2655 {
2656 	struct ixgbe_adapter *adapter = data;
2657 	struct ixgbe_hw *hw = &adapter->hw;
2658 	u32 eicr;
2659 
2660 	/*
2661 	 * Workaround for Silicon errata.  Use clear-by-write instead
2662 	 * of clear-by-read.  Reading with EICS will return the
2663 	 * interrupt causes without clearing, which later be done
2664 	 * with the write to EICR.
2665 	 */
2666 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2667 
2668 	/* The lower 16bits of the EICR register are for the queue interrupts
2669 	 * which should be masked here in order to not accidentally clear them if
2670 	 * the bits are high when ixgbe_msix_other is called. There is a race
2671 	 * condition otherwise which results in possible performance loss
2672 	 * especially if the ixgbe_msix_other interrupt is triggering
2673 	 * consistently (as it would when PPS is turned on for the X540 device)
2674 	 */
2675 	eicr &= 0xFFFF0000;
2676 
2677 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2678 
2679 	if (eicr & IXGBE_EICR_LSC)
2680 		ixgbe_check_lsc(adapter);
2681 
2682 	if (eicr & IXGBE_EICR_MAILBOX)
2683 		ixgbe_msg_task(adapter);
2684 
2685 	switch (hw->mac.type) {
2686 	case ixgbe_mac_82599EB:
2687 	case ixgbe_mac_X540:
2688 	case ixgbe_mac_X550:
2689 	case ixgbe_mac_X550EM_x:
2690 		if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
2691 		    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
2692 			adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
2693 			ixgbe_service_event_schedule(adapter);
2694 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
2695 					IXGBE_EICR_GPI_SDP0_X540);
2696 		}
2697 		if (eicr & IXGBE_EICR_ECC) {
2698 			e_info(link, "Received ECC Err, initiating reset\n");
2699 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
2700 			ixgbe_service_event_schedule(adapter);
2701 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2702 		}
2703 		/* Handle Flow Director Full threshold interrupt */
2704 		if (eicr & IXGBE_EICR_FLOW_DIR) {
2705 			int reinit_count = 0;
2706 			int i;
2707 			for (i = 0; i < adapter->num_tx_queues; i++) {
2708 				struct ixgbe_ring *ring = adapter->tx_ring[i];
2709 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
2710 						       &ring->state))
2711 					reinit_count++;
2712 			}
2713 			if (reinit_count) {
2714 				/* no more flow director interrupts until after init */
2715 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
2716 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
2717 				ixgbe_service_event_schedule(adapter);
2718 			}
2719 		}
2720 		ixgbe_check_sfp_event(adapter, eicr);
2721 		ixgbe_check_overtemp_event(adapter, eicr);
2722 		break;
2723 	default:
2724 		break;
2725 	}
2726 
2727 	ixgbe_check_fan_failure(adapter, eicr);
2728 
2729 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2730 		ixgbe_ptp_check_pps_event(adapter, eicr);
2731 
2732 	/* re-enable the original interrupt state, no lsc, no queues */
2733 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2734 		ixgbe_irq_enable(adapter, false, false);
2735 
2736 	return IRQ_HANDLED;
2737 }
2738 
2739 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
2740 {
2741 	struct ixgbe_q_vector *q_vector = data;
2742 
2743 	/* EIAM disabled interrupts (on this vector) for us */
2744 
2745 	if (q_vector->rx.ring || q_vector->tx.ring)
2746 		napi_schedule(&q_vector->napi);
2747 
2748 	return IRQ_HANDLED;
2749 }
2750 
2751 /**
2752  * ixgbe_poll - NAPI Rx polling callback
2753  * @napi: structure for representing this polling device
2754  * @budget: how many packets driver is allowed to clean
2755  *
2756  * This function is used for legacy and MSI, NAPI mode
2757  **/
2758 int ixgbe_poll(struct napi_struct *napi, int budget)
2759 {
2760 	struct ixgbe_q_vector *q_vector =
2761 				container_of(napi, struct ixgbe_q_vector, napi);
2762 	struct ixgbe_adapter *adapter = q_vector->adapter;
2763 	struct ixgbe_ring *ring;
2764 	int per_ring_budget;
2765 	bool clean_complete = true;
2766 
2767 #ifdef CONFIG_IXGBE_DCA
2768 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
2769 		ixgbe_update_dca(q_vector);
2770 #endif
2771 
2772 	ixgbe_for_each_ring(ring, q_vector->tx)
2773 		clean_complete &= !!ixgbe_clean_tx_irq(q_vector, ring);
2774 
2775 	if (!ixgbe_qv_lock_napi(q_vector))
2776 		return budget;
2777 
2778 	/* attempt to distribute budget to each queue fairly, but don't allow
2779 	 * the budget to go below 1 because we'll exit polling */
2780 	if (q_vector->rx.count > 1)
2781 		per_ring_budget = max(budget/q_vector->rx.count, 1);
2782 	else
2783 		per_ring_budget = budget;
2784 
2785 	ixgbe_for_each_ring(ring, q_vector->rx)
2786 		clean_complete &= (ixgbe_clean_rx_irq(q_vector, ring,
2787 				   per_ring_budget) < per_ring_budget);
2788 
2789 	ixgbe_qv_unlock_napi(q_vector);
2790 	/* If all work not completed, return budget and keep polling */
2791 	if (!clean_complete)
2792 		return budget;
2793 
2794 	/* all work done, exit the polling mode */
2795 	napi_complete(napi);
2796 	if (adapter->rx_itr_setting & 1)
2797 		ixgbe_set_itr(q_vector);
2798 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2799 		ixgbe_irq_enable_queues(adapter, ((u64)1 << q_vector->v_idx));
2800 
2801 	return 0;
2802 }
2803 
2804 /**
2805  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2806  * @adapter: board private structure
2807  *
2808  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2809  * interrupts from the kernel.
2810  **/
2811 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
2812 {
2813 	struct net_device *netdev = adapter->netdev;
2814 	int vector, err;
2815 	int ri = 0, ti = 0;
2816 
2817 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2818 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2819 		struct msix_entry *entry = &adapter->msix_entries[vector];
2820 
2821 		if (q_vector->tx.ring && q_vector->rx.ring) {
2822 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2823 				 "%s-%s-%d", netdev->name, "TxRx", ri++);
2824 			ti++;
2825 		} else if (q_vector->rx.ring) {
2826 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2827 				 "%s-%s-%d", netdev->name, "rx", ri++);
2828 		} else if (q_vector->tx.ring) {
2829 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2830 				 "%s-%s-%d", netdev->name, "tx", ti++);
2831 		} else {
2832 			/* skip this unused q_vector */
2833 			continue;
2834 		}
2835 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
2836 				  q_vector->name, q_vector);
2837 		if (err) {
2838 			e_err(probe, "request_irq failed for MSIX interrupt "
2839 			      "Error: %d\n", err);
2840 			goto free_queue_irqs;
2841 		}
2842 		/* If Flow Director is enabled, set interrupt affinity */
2843 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2844 			/* assign the mask for this irq */
2845 			irq_set_affinity_hint(entry->vector,
2846 					      &q_vector->affinity_mask);
2847 		}
2848 	}
2849 
2850 	err = request_irq(adapter->msix_entries[vector].vector,
2851 			  ixgbe_msix_other, 0, netdev->name, adapter);
2852 	if (err) {
2853 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
2854 		goto free_queue_irqs;
2855 	}
2856 
2857 	return 0;
2858 
2859 free_queue_irqs:
2860 	while (vector) {
2861 		vector--;
2862 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
2863 				      NULL);
2864 		free_irq(adapter->msix_entries[vector].vector,
2865 			 adapter->q_vector[vector]);
2866 	}
2867 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
2868 	pci_disable_msix(adapter->pdev);
2869 	kfree(adapter->msix_entries);
2870 	adapter->msix_entries = NULL;
2871 	return err;
2872 }
2873 
2874 /**
2875  * ixgbe_intr - legacy mode Interrupt Handler
2876  * @irq: interrupt number
2877  * @data: pointer to a network interface device structure
2878  **/
2879 static irqreturn_t ixgbe_intr(int irq, void *data)
2880 {
2881 	struct ixgbe_adapter *adapter = data;
2882 	struct ixgbe_hw *hw = &adapter->hw;
2883 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2884 	u32 eicr;
2885 
2886 	/*
2887 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
2888 	 * before the read of EICR.
2889 	 */
2890 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
2891 
2892 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2893 	 * therefore no explicit interrupt disable is necessary */
2894 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2895 	if (!eicr) {
2896 		/*
2897 		 * shared interrupt alert!
2898 		 * make sure interrupts are enabled because the read will
2899 		 * have disabled interrupts due to EIAM
2900 		 * finish the workaround of silicon errata on 82598.  Unmask
2901 		 * the interrupt that we masked before the EICR read.
2902 		 */
2903 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
2904 			ixgbe_irq_enable(adapter, true, true);
2905 		return IRQ_NONE;	/* Not our interrupt */
2906 	}
2907 
2908 	if (eicr & IXGBE_EICR_LSC)
2909 		ixgbe_check_lsc(adapter);
2910 
2911 	switch (hw->mac.type) {
2912 	case ixgbe_mac_82599EB:
2913 		ixgbe_check_sfp_event(adapter, eicr);
2914 		/* Fall through */
2915 	case ixgbe_mac_X540:
2916 	case ixgbe_mac_X550:
2917 	case ixgbe_mac_X550EM_x:
2918 		if (eicr & IXGBE_EICR_ECC) {
2919 			e_info(link, "Received ECC Err, initiating reset\n");
2920 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
2921 			ixgbe_service_event_schedule(adapter);
2922 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2923 		}
2924 		ixgbe_check_overtemp_event(adapter, eicr);
2925 		break;
2926 	default:
2927 		break;
2928 	}
2929 
2930 	ixgbe_check_fan_failure(adapter, eicr);
2931 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2932 		ixgbe_ptp_check_pps_event(adapter, eicr);
2933 
2934 	/* would disable interrupts here but EIAM disabled it */
2935 	napi_schedule(&q_vector->napi);
2936 
2937 	/*
2938 	 * re-enable link(maybe) and non-queue interrupts, no flush.
2939 	 * ixgbe_poll will re-enable the queue interrupts
2940 	 */
2941 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2942 		ixgbe_irq_enable(adapter, false, false);
2943 
2944 	return IRQ_HANDLED;
2945 }
2946 
2947 /**
2948  * ixgbe_request_irq - initialize interrupts
2949  * @adapter: board private structure
2950  *
2951  * Attempts to configure interrupts using the best available
2952  * capabilities of the hardware and kernel.
2953  **/
2954 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
2955 {
2956 	struct net_device *netdev = adapter->netdev;
2957 	int err;
2958 
2959 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
2960 		err = ixgbe_request_msix_irqs(adapter);
2961 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
2962 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
2963 				  netdev->name, adapter);
2964 	else
2965 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
2966 				  netdev->name, adapter);
2967 
2968 	if (err)
2969 		e_err(probe, "request_irq failed, Error %d\n", err);
2970 
2971 	return err;
2972 }
2973 
2974 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
2975 {
2976 	int vector;
2977 
2978 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
2979 		free_irq(adapter->pdev->irq, adapter);
2980 		return;
2981 	}
2982 
2983 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2984 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2985 		struct msix_entry *entry = &adapter->msix_entries[vector];
2986 
2987 		/* free only the irqs that were actually requested */
2988 		if (!q_vector->rx.ring && !q_vector->tx.ring)
2989 			continue;
2990 
2991 		/* clear the affinity_mask in the IRQ descriptor */
2992 		irq_set_affinity_hint(entry->vector, NULL);
2993 
2994 		free_irq(entry->vector, q_vector);
2995 	}
2996 
2997 	free_irq(adapter->msix_entries[vector++].vector, adapter);
2998 }
2999 
3000 /**
3001  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3002  * @adapter: board private structure
3003  **/
3004 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3005 {
3006 	switch (adapter->hw.mac.type) {
3007 	case ixgbe_mac_82598EB:
3008 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3009 		break;
3010 	case ixgbe_mac_82599EB:
3011 	case ixgbe_mac_X540:
3012 	case ixgbe_mac_X550:
3013 	case ixgbe_mac_X550EM_x:
3014 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3015 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3016 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3017 		break;
3018 	default:
3019 		break;
3020 	}
3021 	IXGBE_WRITE_FLUSH(&adapter->hw);
3022 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3023 		int vector;
3024 
3025 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
3026 			synchronize_irq(adapter->msix_entries[vector].vector);
3027 
3028 		synchronize_irq(adapter->msix_entries[vector++].vector);
3029 	} else {
3030 		synchronize_irq(adapter->pdev->irq);
3031 	}
3032 }
3033 
3034 /**
3035  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3036  *
3037  **/
3038 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3039 {
3040 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3041 
3042 	ixgbe_write_eitr(q_vector);
3043 
3044 	ixgbe_set_ivar(adapter, 0, 0, 0);
3045 	ixgbe_set_ivar(adapter, 1, 0, 0);
3046 
3047 	e_info(hw, "Legacy interrupt IVAR setup done\n");
3048 }
3049 
3050 /**
3051  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3052  * @adapter: board private structure
3053  * @ring: structure containing ring specific data
3054  *
3055  * Configure the Tx descriptor ring after a reset.
3056  **/
3057 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3058 			     struct ixgbe_ring *ring)
3059 {
3060 	struct ixgbe_hw *hw = &adapter->hw;
3061 	u64 tdba = ring->dma;
3062 	int wait_loop = 10;
3063 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
3064 	u8 reg_idx = ring->reg_idx;
3065 
3066 	/* disable queue to avoid issues while updating state */
3067 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3068 	IXGBE_WRITE_FLUSH(hw);
3069 
3070 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3071 			(tdba & DMA_BIT_MASK(32)));
3072 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3073 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3074 			ring->count * sizeof(union ixgbe_adv_tx_desc));
3075 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3076 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3077 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3078 
3079 	/*
3080 	 * set WTHRESH to encourage burst writeback, it should not be set
3081 	 * higher than 1 when:
3082 	 * - ITR is 0 as it could cause false TX hangs
3083 	 * - ITR is set to > 100k int/sec and BQL is enabled
3084 	 *
3085 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3086 	 * to or less than the number of on chip descriptors, which is
3087 	 * currently 40.
3088 	 */
3089 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3090 		txdctl |= (1 << 16);	/* WTHRESH = 1 */
3091 	else
3092 		txdctl |= (8 << 16);	/* WTHRESH = 8 */
3093 
3094 	/*
3095 	 * Setting PTHRESH to 32 both improves performance
3096 	 * and avoids a TX hang with DFP enabled
3097 	 */
3098 	txdctl |= (1 << 8) |	/* HTHRESH = 1 */
3099 		   32;		/* PTHRESH = 32 */
3100 
3101 	/* reinitialize flowdirector state */
3102 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3103 		ring->atr_sample_rate = adapter->atr_sample_rate;
3104 		ring->atr_count = 0;
3105 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3106 	} else {
3107 		ring->atr_sample_rate = 0;
3108 	}
3109 
3110 	/* initialize XPS */
3111 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3112 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3113 
3114 		if (q_vector)
3115 			netif_set_xps_queue(ring->netdev,
3116 					    &q_vector->affinity_mask,
3117 					    ring->queue_index);
3118 	}
3119 
3120 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3121 
3122 	/* enable queue */
3123 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3124 
3125 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3126 	if (hw->mac.type == ixgbe_mac_82598EB &&
3127 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3128 		return;
3129 
3130 	/* poll to verify queue is enabled */
3131 	do {
3132 		usleep_range(1000, 2000);
3133 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3134 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3135 	if (!wait_loop)
3136 		e_err(drv, "Could not enable Tx Queue %d\n", reg_idx);
3137 }
3138 
3139 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3140 {
3141 	struct ixgbe_hw *hw = &adapter->hw;
3142 	u32 rttdcs, mtqc;
3143 	u8 tcs = netdev_get_num_tc(adapter->netdev);
3144 
3145 	if (hw->mac.type == ixgbe_mac_82598EB)
3146 		return;
3147 
3148 	/* disable the arbiter while setting MTQC */
3149 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3150 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3151 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3152 
3153 	/* set transmit pool layout */
3154 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3155 		mtqc = IXGBE_MTQC_VT_ENA;
3156 		if (tcs > 4)
3157 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3158 		else if (tcs > 1)
3159 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3160 		else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3161 			mtqc |= IXGBE_MTQC_32VF;
3162 		else
3163 			mtqc |= IXGBE_MTQC_64VF;
3164 	} else {
3165 		if (tcs > 4)
3166 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3167 		else if (tcs > 1)
3168 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3169 		else
3170 			mtqc = IXGBE_MTQC_64Q_1PB;
3171 	}
3172 
3173 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3174 
3175 	/* Enable Security TX Buffer IFG for multiple pb */
3176 	if (tcs) {
3177 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3178 		sectx |= IXGBE_SECTX_DCB;
3179 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3180 	}
3181 
3182 	/* re-enable the arbiter */
3183 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3184 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3185 }
3186 
3187 /**
3188  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3189  * @adapter: board private structure
3190  *
3191  * Configure the Tx unit of the MAC after a reset.
3192  **/
3193 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3194 {
3195 	struct ixgbe_hw *hw = &adapter->hw;
3196 	u32 dmatxctl;
3197 	u32 i;
3198 
3199 	ixgbe_setup_mtqc(adapter);
3200 
3201 	if (hw->mac.type != ixgbe_mac_82598EB) {
3202 		/* DMATXCTL.EN must be before Tx queues are enabled */
3203 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3204 		dmatxctl |= IXGBE_DMATXCTL_TE;
3205 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3206 	}
3207 
3208 	/* Setup the HW Tx Head and Tail descriptor pointers */
3209 	for (i = 0; i < adapter->num_tx_queues; i++)
3210 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3211 }
3212 
3213 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3214 				 struct ixgbe_ring *ring)
3215 {
3216 	struct ixgbe_hw *hw = &adapter->hw;
3217 	u8 reg_idx = ring->reg_idx;
3218 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3219 
3220 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3221 
3222 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3223 }
3224 
3225 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3226 				  struct ixgbe_ring *ring)
3227 {
3228 	struct ixgbe_hw *hw = &adapter->hw;
3229 	u8 reg_idx = ring->reg_idx;
3230 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3231 
3232 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3233 
3234 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3235 }
3236 
3237 #ifdef CONFIG_IXGBE_DCB
3238 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3239 #else
3240 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3241 #endif
3242 {
3243 	int i;
3244 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3245 
3246 	if (adapter->ixgbe_ieee_pfc)
3247 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3248 
3249 	/*
3250 	 * We should set the drop enable bit if:
3251 	 *  SR-IOV is enabled
3252 	 *   or
3253 	 *  Number of Rx queues > 1 and flow control is disabled
3254 	 *
3255 	 *  This allows us to avoid head of line blocking for security
3256 	 *  and performance reasons.
3257 	 */
3258 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3259 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3260 		for (i = 0; i < adapter->num_rx_queues; i++)
3261 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3262 	} else {
3263 		for (i = 0; i < adapter->num_rx_queues; i++)
3264 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3265 	}
3266 }
3267 
3268 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3269 
3270 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3271 				   struct ixgbe_ring *rx_ring)
3272 {
3273 	struct ixgbe_hw *hw = &adapter->hw;
3274 	u32 srrctl;
3275 	u8 reg_idx = rx_ring->reg_idx;
3276 
3277 	if (hw->mac.type == ixgbe_mac_82598EB) {
3278 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3279 
3280 		/*
3281 		 * if VMDq is not active we must program one srrctl register
3282 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3283 		 */
3284 		reg_idx &= mask;
3285 	}
3286 
3287 	/* configure header buffer length, needed for RSC */
3288 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3289 
3290 	/* configure the packet buffer length */
3291 	srrctl |= ixgbe_rx_bufsz(rx_ring) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3292 
3293 	/* configure descriptor type */
3294 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3295 
3296 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3297 }
3298 
3299 /**
3300  * Return a number of entries in the RSS indirection table
3301  *
3302  * @adapter: device handle
3303  *
3304  *  - 82598/82599/X540:     128
3305  *  - X550(non-SRIOV mode): 512
3306  *  - X550(SRIOV mode):     64
3307  */
3308 u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3309 {
3310 	if (adapter->hw.mac.type < ixgbe_mac_X550)
3311 		return 128;
3312 	else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3313 		return 64;
3314 	else
3315 		return 512;
3316 }
3317 
3318 /**
3319  * Write the RETA table to HW
3320  *
3321  * @adapter: device handle
3322  *
3323  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3324  */
3325 void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3326 {
3327 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3328 	struct ixgbe_hw *hw = &adapter->hw;
3329 	u32 reta = 0;
3330 	u32 indices_multi;
3331 	u8 *indir_tbl = adapter->rss_indir_tbl;
3332 
3333 	/* Fill out the redirection table as follows:
3334 	 *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3335 	 *    indices.
3336 	 *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3337 	 *  - X550:       8 bit wide entries containing 6 bit RSS index
3338 	 */
3339 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3340 		indices_multi = 0x11;
3341 	else
3342 		indices_multi = 0x1;
3343 
3344 	/* Write redirection table to HW */
3345 	for (i = 0; i < reta_entries; i++) {
3346 		reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3347 		if ((i & 3) == 3) {
3348 			if (i < 128)
3349 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3350 			else
3351 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3352 						reta);
3353 			reta = 0;
3354 		}
3355 	}
3356 }
3357 
3358 /**
3359  * Write the RETA table to HW (for x550 devices in SRIOV mode)
3360  *
3361  * @adapter: device handle
3362  *
3363  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3364  */
3365 static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3366 {
3367 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3368 	struct ixgbe_hw *hw = &adapter->hw;
3369 	u32 vfreta = 0;
3370 	unsigned int pf_pool = adapter->num_vfs;
3371 
3372 	/* Write redirection table to HW */
3373 	for (i = 0; i < reta_entries; i++) {
3374 		vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3375 		if ((i & 3) == 3) {
3376 			IXGBE_WRITE_REG(hw, IXGBE_PFVFRETA(i >> 2, pf_pool),
3377 					vfreta);
3378 			vfreta = 0;
3379 		}
3380 	}
3381 }
3382 
3383 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3384 {
3385 	struct ixgbe_hw *hw = &adapter->hw;
3386 	u32 i, j;
3387 	u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3388 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3389 
3390 	/* Program table for at least 2 queues w/ SR-IOV so that VFs can
3391 	 * make full use of any rings they may have.  We will use the
3392 	 * PSRTYPE register to control how many rings we use within the PF.
3393 	 */
3394 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 2))
3395 		rss_i = 2;
3396 
3397 	/* Fill out hash function seeds */
3398 	for (i = 0; i < 10; i++)
3399 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3400 
3401 	/* Fill out redirection table */
3402 	memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3403 
3404 	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3405 		if (j == rss_i)
3406 			j = 0;
3407 
3408 		adapter->rss_indir_tbl[i] = j;
3409 	}
3410 
3411 	ixgbe_store_reta(adapter);
3412 }
3413 
3414 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3415 {
3416 	struct ixgbe_hw *hw = &adapter->hw;
3417 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3418 	unsigned int pf_pool = adapter->num_vfs;
3419 	int i, j;
3420 
3421 	/* Fill out hash function seeds */
3422 	for (i = 0; i < 10; i++)
3423 		IXGBE_WRITE_REG(hw, IXGBE_PFVFRSSRK(i, pf_pool),
3424 				adapter->rss_key[i]);
3425 
3426 	/* Fill out the redirection table */
3427 	for (i = 0, j = 0; i < 64; i++, j++) {
3428 		if (j == rss_i)
3429 			j = 0;
3430 
3431 		adapter->rss_indir_tbl[i] = j;
3432 	}
3433 
3434 	ixgbe_store_vfreta(adapter);
3435 }
3436 
3437 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3438 {
3439 	struct ixgbe_hw *hw = &adapter->hw;
3440 	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3441 	u32 rxcsum;
3442 
3443 	/* Disable indicating checksum in descriptor, enables RSS hash */
3444 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3445 	rxcsum |= IXGBE_RXCSUM_PCSD;
3446 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3447 
3448 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3449 		if (adapter->ring_feature[RING_F_RSS].mask)
3450 			mrqc = IXGBE_MRQC_RSSEN;
3451 	} else {
3452 		u8 tcs = netdev_get_num_tc(adapter->netdev);
3453 
3454 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3455 			if (tcs > 4)
3456 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3457 			else if (tcs > 1)
3458 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3459 			else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3460 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3461 			else
3462 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3463 		} else {
3464 			if (tcs > 4)
3465 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3466 			else if (tcs > 1)
3467 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3468 			else
3469 				mrqc = IXGBE_MRQC_RSSEN;
3470 		}
3471 	}
3472 
3473 	/* Perform hash on these packet types */
3474 	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3475 		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3476 		     IXGBE_MRQC_RSS_FIELD_IPV6 |
3477 		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3478 
3479 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3480 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3481 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3482 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3483 
3484 	netdev_rss_key_fill(adapter->rss_key, sizeof(adapter->rss_key));
3485 	if ((hw->mac.type >= ixgbe_mac_X550) &&
3486 	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
3487 		unsigned int pf_pool = adapter->num_vfs;
3488 
3489 		/* Enable VF RSS mode */
3490 		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
3491 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3492 
3493 		/* Setup RSS through the VF registers */
3494 		ixgbe_setup_vfreta(adapter);
3495 		vfmrqc = IXGBE_MRQC_RSSEN;
3496 		vfmrqc |= rss_field;
3497 		IXGBE_WRITE_REG(hw, IXGBE_PFVFMRQC(pf_pool), vfmrqc);
3498 	} else {
3499 		ixgbe_setup_reta(adapter);
3500 		mrqc |= rss_field;
3501 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3502 	}
3503 }
3504 
3505 /**
3506  * ixgbe_configure_rscctl - enable RSC for the indicated ring
3507  * @adapter:    address of board private structure
3508  * @index:      index of ring to set
3509  **/
3510 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
3511 				   struct ixgbe_ring *ring)
3512 {
3513 	struct ixgbe_hw *hw = &adapter->hw;
3514 	u32 rscctrl;
3515 	u8 reg_idx = ring->reg_idx;
3516 
3517 	if (!ring_is_rsc_enabled(ring))
3518 		return;
3519 
3520 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
3521 	rscctrl |= IXGBE_RSCCTL_RSCEN;
3522 	/*
3523 	 * we must limit the number of descriptors so that the
3524 	 * total size of max desc * buf_len is not greater
3525 	 * than 65536
3526 	 */
3527 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
3528 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
3529 }
3530 
3531 #define IXGBE_MAX_RX_DESC_POLL 10
3532 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
3533 				       struct ixgbe_ring *ring)
3534 {
3535 	struct ixgbe_hw *hw = &adapter->hw;
3536 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3537 	u32 rxdctl;
3538 	u8 reg_idx = ring->reg_idx;
3539 
3540 	if (ixgbe_removed(hw->hw_addr))
3541 		return;
3542 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3543 	if (hw->mac.type == ixgbe_mac_82598EB &&
3544 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3545 		return;
3546 
3547 	do {
3548 		usleep_range(1000, 2000);
3549 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3550 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
3551 
3552 	if (!wait_loop) {
3553 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
3554 		      "the polling period\n", reg_idx);
3555 	}
3556 }
3557 
3558 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
3559 			    struct ixgbe_ring *ring)
3560 {
3561 	struct ixgbe_hw *hw = &adapter->hw;
3562 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3563 	u32 rxdctl;
3564 	u8 reg_idx = ring->reg_idx;
3565 
3566 	if (ixgbe_removed(hw->hw_addr))
3567 		return;
3568 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3569 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
3570 
3571 	/* write value back with RXDCTL.ENABLE bit cleared */
3572 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3573 
3574 	if (hw->mac.type == ixgbe_mac_82598EB &&
3575 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3576 		return;
3577 
3578 	/* the hardware may take up to 100us to really disable the rx queue */
3579 	do {
3580 		udelay(10);
3581 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3582 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
3583 
3584 	if (!wait_loop) {
3585 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3586 		      "the polling period\n", reg_idx);
3587 	}
3588 }
3589 
3590 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
3591 			     struct ixgbe_ring *ring)
3592 {
3593 	struct ixgbe_hw *hw = &adapter->hw;
3594 	u64 rdba = ring->dma;
3595 	u32 rxdctl;
3596 	u8 reg_idx = ring->reg_idx;
3597 
3598 	/* disable queue to avoid issues while updating state */
3599 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3600 	ixgbe_disable_rx_queue(adapter, ring);
3601 
3602 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
3603 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
3604 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
3605 			ring->count * sizeof(union ixgbe_adv_rx_desc));
3606 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
3607 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
3608 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
3609 
3610 	ixgbe_configure_srrctl(adapter, ring);
3611 	ixgbe_configure_rscctl(adapter, ring);
3612 
3613 	if (hw->mac.type == ixgbe_mac_82598EB) {
3614 		/*
3615 		 * enable cache line friendly hardware writes:
3616 		 * PTHRESH=32 descriptors (half the internal cache),
3617 		 * this also removes ugly rx_no_buffer_count increment
3618 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
3619 		 * WTHRESH=8 burst writeback up to two cache lines
3620 		 */
3621 		rxdctl &= ~0x3FFFFF;
3622 		rxdctl |=  0x080420;
3623 	}
3624 
3625 	/* enable receive descriptor ring */
3626 	rxdctl |= IXGBE_RXDCTL_ENABLE;
3627 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3628 
3629 	ixgbe_rx_desc_queue_enable(adapter, ring);
3630 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
3631 }
3632 
3633 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
3634 {
3635 	struct ixgbe_hw *hw = &adapter->hw;
3636 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
3637 	u16 pool;
3638 
3639 	/* PSRTYPE must be initialized in non 82598 adapters */
3640 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
3641 		      IXGBE_PSRTYPE_UDPHDR |
3642 		      IXGBE_PSRTYPE_IPV4HDR |
3643 		      IXGBE_PSRTYPE_L2HDR |
3644 		      IXGBE_PSRTYPE_IPV6HDR;
3645 
3646 	if (hw->mac.type == ixgbe_mac_82598EB)
3647 		return;
3648 
3649 	if (rss_i > 3)
3650 		psrtype |= 2 << 29;
3651 	else if (rss_i > 1)
3652 		psrtype |= 1 << 29;
3653 
3654 	for_each_set_bit(pool, &adapter->fwd_bitmask, 32)
3655 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
3656 }
3657 
3658 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
3659 {
3660 	struct ixgbe_hw *hw = &adapter->hw;
3661 	u32 reg_offset, vf_shift;
3662 	u32 gcr_ext, vmdctl;
3663 	int i;
3664 
3665 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3666 		return;
3667 
3668 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3669 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
3670 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
3671 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
3672 	vmdctl |= IXGBE_VT_CTL_REPLEN;
3673 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
3674 
3675 	vf_shift = VMDQ_P(0) % 32;
3676 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
3677 
3678 	/* Enable only the PF's pool for Tx/Rx */
3679 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), (~0) << vf_shift);
3680 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
3681 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), (~0) << vf_shift);
3682 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
3683 	if (adapter->bridge_mode == BRIDGE_MODE_VEB)
3684 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3685 
3686 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3687 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
3688 
3689 	/*
3690 	 * Set up VF register offsets for selected VT Mode,
3691 	 * i.e. 32 or 64 VFs for SR-IOV
3692 	 */
3693 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
3694 	case IXGBE_82599_VMDQ_8Q_MASK:
3695 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
3696 		break;
3697 	case IXGBE_82599_VMDQ_4Q_MASK:
3698 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
3699 		break;
3700 	default:
3701 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
3702 		break;
3703 	}
3704 
3705 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
3706 
3707 
3708 	/* Enable MAC Anti-Spoofing */
3709 	hw->mac.ops.set_mac_anti_spoofing(hw, (adapter->num_vfs != 0),
3710 					  adapter->num_vfs);
3711 
3712 	/* Ensure LLDP is set for Ethertype Antispoofing if we will be
3713 	 * calling set_ethertype_anti_spoofing for each VF in loop below
3714 	 */
3715 	if (hw->mac.ops.set_ethertype_anti_spoofing)
3716 		IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_LLDP),
3717 				(IXGBE_ETQF_FILTER_EN    | /* enable filter */
3718 				 IXGBE_ETQF_TX_ANTISPOOF | /* tx antispoof */
3719 				 IXGBE_ETH_P_LLDP));	   /* LLDP eth type */
3720 
3721 	/* For VFs that have spoof checking turned off */
3722 	for (i = 0; i < adapter->num_vfs; i++) {
3723 		if (!adapter->vfinfo[i].spoofchk_enabled)
3724 			ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i, false);
3725 
3726 		/* enable ethertype anti spoofing if hw supports it */
3727 		if (hw->mac.ops.set_ethertype_anti_spoofing)
3728 			hw->mac.ops.set_ethertype_anti_spoofing(hw, true, i);
3729 
3730 		/* Enable/Disable RSS query feature  */
3731 		ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
3732 					  adapter->vfinfo[i].rss_query_enabled);
3733 	}
3734 }
3735 
3736 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
3737 {
3738 	struct ixgbe_hw *hw = &adapter->hw;
3739 	struct net_device *netdev = adapter->netdev;
3740 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3741 	struct ixgbe_ring *rx_ring;
3742 	int i;
3743 	u32 mhadd, hlreg0;
3744 
3745 #ifdef IXGBE_FCOE
3746 	/* adjust max frame to be able to do baby jumbo for FCoE */
3747 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
3748 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
3749 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3750 
3751 #endif /* IXGBE_FCOE */
3752 
3753 	/* adjust max frame to be at least the size of a standard frame */
3754 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
3755 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
3756 
3757 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3758 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
3759 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3760 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
3761 
3762 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3763 	}
3764 
3765 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3766 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3767 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3768 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3769 
3770 	/*
3771 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3772 	 * the Base and Length of the Rx Descriptor Ring
3773 	 */
3774 	for (i = 0; i < adapter->num_rx_queues; i++) {
3775 		rx_ring = adapter->rx_ring[i];
3776 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
3777 			set_ring_rsc_enabled(rx_ring);
3778 		else
3779 			clear_ring_rsc_enabled(rx_ring);
3780 	}
3781 }
3782 
3783 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
3784 {
3785 	struct ixgbe_hw *hw = &adapter->hw;
3786 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
3787 
3788 	switch (hw->mac.type) {
3789 	case ixgbe_mac_X550:
3790 	case ixgbe_mac_X550EM_x:
3791 	case ixgbe_mac_82598EB:
3792 		/*
3793 		 * For VMDq support of different descriptor types or
3794 		 * buffer sizes through the use of multiple SRRCTL
3795 		 * registers, RDRXCTL.MVMEN must be set to 1
3796 		 *
3797 		 * also, the manual doesn't mention it clearly but DCA hints
3798 		 * will only use queue 0's tags unless this bit is set.  Side
3799 		 * effects of setting this bit are only that SRRCTL must be
3800 		 * fully programmed [0..15]
3801 		 */
3802 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
3803 		break;
3804 	case ixgbe_mac_82599EB:
3805 	case ixgbe_mac_X540:
3806 		/* Disable RSC for ACK packets */
3807 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
3808 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
3809 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
3810 		/* hardware requires some bits to be set by default */
3811 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
3812 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
3813 		break;
3814 	default:
3815 		/* We should do nothing since we don't know this hardware */
3816 		return;
3817 	}
3818 
3819 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
3820 }
3821 
3822 /**
3823  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3824  * @adapter: board private structure
3825  *
3826  * Configure the Rx unit of the MAC after a reset.
3827  **/
3828 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
3829 {
3830 	struct ixgbe_hw *hw = &adapter->hw;
3831 	int i;
3832 	u32 rxctrl, rfctl;
3833 
3834 	/* disable receives while setting up the descriptors */
3835 	hw->mac.ops.disable_rx(hw);
3836 
3837 	ixgbe_setup_psrtype(adapter);
3838 	ixgbe_setup_rdrxctl(adapter);
3839 
3840 	/* RSC Setup */
3841 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
3842 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
3843 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
3844 		rfctl |= IXGBE_RFCTL_RSC_DIS;
3845 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
3846 
3847 	/* Program registers for the distribution of queues */
3848 	ixgbe_setup_mrqc(adapter);
3849 
3850 	/* set_rx_buffer_len must be called before ring initialization */
3851 	ixgbe_set_rx_buffer_len(adapter);
3852 
3853 	/*
3854 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3855 	 * the Base and Length of the Rx Descriptor Ring
3856 	 */
3857 	for (i = 0; i < adapter->num_rx_queues; i++)
3858 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
3859 
3860 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3861 	/* disable drop enable for 82598 parts */
3862 	if (hw->mac.type == ixgbe_mac_82598EB)
3863 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3864 
3865 	/* enable all receives */
3866 	rxctrl |= IXGBE_RXCTRL_RXEN;
3867 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
3868 }
3869 
3870 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
3871 				 __be16 proto, u16 vid)
3872 {
3873 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3874 	struct ixgbe_hw *hw = &adapter->hw;
3875 
3876 	/* add VID to filter table */
3877 	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true);
3878 	set_bit(vid, adapter->active_vlans);
3879 
3880 	return 0;
3881 }
3882 
3883 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
3884 				  __be16 proto, u16 vid)
3885 {
3886 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3887 	struct ixgbe_hw *hw = &adapter->hw;
3888 
3889 	/* remove VID from filter table */
3890 	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), false);
3891 	clear_bit(vid, adapter->active_vlans);
3892 
3893 	return 0;
3894 }
3895 
3896 /**
3897  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
3898  * @adapter: driver data
3899  */
3900 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
3901 {
3902 	struct ixgbe_hw *hw = &adapter->hw;
3903 	u32 vlnctrl;
3904 	int i, j;
3905 
3906 	switch (hw->mac.type) {
3907 	case ixgbe_mac_82598EB:
3908 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3909 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
3910 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3911 		break;
3912 	case ixgbe_mac_82599EB:
3913 	case ixgbe_mac_X540:
3914 	case ixgbe_mac_X550:
3915 	case ixgbe_mac_X550EM_x:
3916 		for (i = 0; i < adapter->num_rx_queues; i++) {
3917 			struct ixgbe_ring *ring = adapter->rx_ring[i];
3918 
3919 			if (ring->l2_accel_priv)
3920 				continue;
3921 			j = ring->reg_idx;
3922 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3923 			vlnctrl &= ~IXGBE_RXDCTL_VME;
3924 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3925 		}
3926 		break;
3927 	default:
3928 		break;
3929 	}
3930 }
3931 
3932 /**
3933  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
3934  * @adapter: driver data
3935  */
3936 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
3937 {
3938 	struct ixgbe_hw *hw = &adapter->hw;
3939 	u32 vlnctrl;
3940 	int i, j;
3941 
3942 	switch (hw->mac.type) {
3943 	case ixgbe_mac_82598EB:
3944 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3945 		vlnctrl |= IXGBE_VLNCTRL_VME;
3946 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3947 		break;
3948 	case ixgbe_mac_82599EB:
3949 	case ixgbe_mac_X540:
3950 	case ixgbe_mac_X550:
3951 	case ixgbe_mac_X550EM_x:
3952 		for (i = 0; i < adapter->num_rx_queues; i++) {
3953 			struct ixgbe_ring *ring = adapter->rx_ring[i];
3954 
3955 			if (ring->l2_accel_priv)
3956 				continue;
3957 			j = ring->reg_idx;
3958 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3959 			vlnctrl |= IXGBE_RXDCTL_VME;
3960 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3961 		}
3962 		break;
3963 	default:
3964 		break;
3965 	}
3966 }
3967 
3968 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
3969 {
3970 	u16 vid;
3971 
3972 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
3973 
3974 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
3975 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
3976 }
3977 
3978 /**
3979  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
3980  * @netdev: network interface device structure
3981  *
3982  * Writes multicast address list to the MTA hash table.
3983  * Returns: -ENOMEM on failure
3984  *                0 on no addresses written
3985  *                X on writing X addresses to MTA
3986  **/
3987 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
3988 {
3989 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3990 	struct ixgbe_hw *hw = &adapter->hw;
3991 
3992 	if (!netif_running(netdev))
3993 		return 0;
3994 
3995 	if (hw->mac.ops.update_mc_addr_list)
3996 		hw->mac.ops.update_mc_addr_list(hw, netdev);
3997 	else
3998 		return -ENOMEM;
3999 
4000 #ifdef CONFIG_PCI_IOV
4001 	ixgbe_restore_vf_multicasts(adapter);
4002 #endif
4003 
4004 	return netdev_mc_count(netdev);
4005 }
4006 
4007 #ifdef CONFIG_PCI_IOV
4008 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4009 {
4010 	struct ixgbe_hw *hw = &adapter->hw;
4011 	int i;
4012 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
4013 		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_IN_USE)
4014 			hw->mac.ops.set_rar(hw, i, adapter->mac_table[i].addr,
4015 					    adapter->mac_table[i].queue,
4016 					    IXGBE_RAH_AV);
4017 		else
4018 			hw->mac.ops.clear_rar(hw, i);
4019 
4020 		adapter->mac_table[i].state &= ~(IXGBE_MAC_STATE_MODIFIED);
4021 	}
4022 }
4023 #endif
4024 
4025 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4026 {
4027 	struct ixgbe_hw *hw = &adapter->hw;
4028 	int i;
4029 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
4030 		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_MODIFIED) {
4031 			if (adapter->mac_table[i].state &
4032 			    IXGBE_MAC_STATE_IN_USE)
4033 				hw->mac.ops.set_rar(hw, i,
4034 						adapter->mac_table[i].addr,
4035 						adapter->mac_table[i].queue,
4036 						IXGBE_RAH_AV);
4037 			else
4038 				hw->mac.ops.clear_rar(hw, i);
4039 
4040 			adapter->mac_table[i].state &=
4041 						~(IXGBE_MAC_STATE_MODIFIED);
4042 		}
4043 	}
4044 }
4045 
4046 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4047 {
4048 	int i;
4049 	struct ixgbe_hw *hw = &adapter->hw;
4050 
4051 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
4052 		adapter->mac_table[i].state |= IXGBE_MAC_STATE_MODIFIED;
4053 		adapter->mac_table[i].state &= ~IXGBE_MAC_STATE_IN_USE;
4054 		eth_zero_addr(adapter->mac_table[i].addr);
4055 		adapter->mac_table[i].queue = 0;
4056 	}
4057 	ixgbe_sync_mac_table(adapter);
4058 }
4059 
4060 static int ixgbe_available_rars(struct ixgbe_adapter *adapter)
4061 {
4062 	struct ixgbe_hw *hw = &adapter->hw;
4063 	int i, count = 0;
4064 
4065 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
4066 		if (adapter->mac_table[i].state == 0)
4067 			count++;
4068 	}
4069 	return count;
4070 }
4071 
4072 /* this function destroys the first RAR entry */
4073 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter,
4074 					 u8 *addr)
4075 {
4076 	struct ixgbe_hw *hw = &adapter->hw;
4077 
4078 	memcpy(&adapter->mac_table[0].addr, addr, ETH_ALEN);
4079 	adapter->mac_table[0].queue = VMDQ_P(0);
4080 	adapter->mac_table[0].state = (IXGBE_MAC_STATE_DEFAULT |
4081 				       IXGBE_MAC_STATE_IN_USE);
4082 	hw->mac.ops.set_rar(hw, 0, adapter->mac_table[0].addr,
4083 			    adapter->mac_table[0].queue,
4084 			    IXGBE_RAH_AV);
4085 }
4086 
4087 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter, u8 *addr, u16 queue)
4088 {
4089 	struct ixgbe_hw *hw = &adapter->hw;
4090 	int i;
4091 
4092 	if (is_zero_ether_addr(addr))
4093 		return -EINVAL;
4094 
4095 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
4096 		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_IN_USE)
4097 			continue;
4098 		adapter->mac_table[i].state |= (IXGBE_MAC_STATE_MODIFIED |
4099 						IXGBE_MAC_STATE_IN_USE);
4100 		ether_addr_copy(adapter->mac_table[i].addr, addr);
4101 		adapter->mac_table[i].queue = queue;
4102 		ixgbe_sync_mac_table(adapter);
4103 		return i;
4104 	}
4105 	return -ENOMEM;
4106 }
4107 
4108 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter, u8 *addr, u16 queue)
4109 {
4110 	/* search table for addr, if found, set to 0 and sync */
4111 	int i;
4112 	struct ixgbe_hw *hw = &adapter->hw;
4113 
4114 	if (is_zero_ether_addr(addr))
4115 		return -EINVAL;
4116 
4117 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
4118 		if (ether_addr_equal(addr, adapter->mac_table[i].addr) &&
4119 		    adapter->mac_table[i].queue == queue) {
4120 			adapter->mac_table[i].state |= IXGBE_MAC_STATE_MODIFIED;
4121 			adapter->mac_table[i].state &= ~IXGBE_MAC_STATE_IN_USE;
4122 			eth_zero_addr(adapter->mac_table[i].addr);
4123 			adapter->mac_table[i].queue = 0;
4124 			ixgbe_sync_mac_table(adapter);
4125 			return 0;
4126 		}
4127 	}
4128 	return -ENOMEM;
4129 }
4130 /**
4131  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
4132  * @netdev: network interface device structure
4133  *
4134  * Writes unicast address list to the RAR table.
4135  * Returns: -ENOMEM on failure/insufficient address space
4136  *                0 on no addresses written
4137  *                X on writing X addresses to the RAR table
4138  **/
4139 static int ixgbe_write_uc_addr_list(struct net_device *netdev, int vfn)
4140 {
4141 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4142 	int count = 0;
4143 
4144 	/* return ENOMEM indicating insufficient memory for addresses */
4145 	if (netdev_uc_count(netdev) > ixgbe_available_rars(adapter))
4146 		return -ENOMEM;
4147 
4148 	if (!netdev_uc_empty(netdev)) {
4149 		struct netdev_hw_addr *ha;
4150 		netdev_for_each_uc_addr(ha, netdev) {
4151 			ixgbe_del_mac_filter(adapter, ha->addr, vfn);
4152 			ixgbe_add_mac_filter(adapter, ha->addr, vfn);
4153 			count++;
4154 		}
4155 	}
4156 	return count;
4157 }
4158 
4159 /**
4160  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4161  * @netdev: network interface device structure
4162  *
4163  * The set_rx_method entry point is called whenever the unicast/multicast
4164  * address list or the network interface flags are updated.  This routine is
4165  * responsible for configuring the hardware for proper unicast, multicast and
4166  * promiscuous mode.
4167  **/
4168 void ixgbe_set_rx_mode(struct net_device *netdev)
4169 {
4170 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4171 	struct ixgbe_hw *hw = &adapter->hw;
4172 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4173 	u32 vlnctrl;
4174 	int count;
4175 
4176 	/* Check for Promiscuous and All Multicast modes */
4177 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4178 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4179 
4180 	/* set all bits that we expect to always be set */
4181 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4182 	fctrl |= IXGBE_FCTRL_BAM;
4183 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4184 	fctrl |= IXGBE_FCTRL_PMCF;
4185 
4186 	/* clear the bits we are changing the status of */
4187 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4188 	vlnctrl &= ~(IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
4189 	if (netdev->flags & IFF_PROMISC) {
4190 		hw->addr_ctrl.user_set_promisc = true;
4191 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4192 		vmolr |= IXGBE_VMOLR_MPE;
4193 		/* Only disable hardware filter vlans in promiscuous mode
4194 		 * if SR-IOV and VMDQ are disabled - otherwise ensure
4195 		 * that hardware VLAN filters remain enabled.
4196 		 */
4197 		if (adapter->flags & (IXGBE_FLAG_VMDQ_ENABLED |
4198 				      IXGBE_FLAG_SRIOV_ENABLED))
4199 			vlnctrl |= (IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
4200 	} else {
4201 		if (netdev->flags & IFF_ALLMULTI) {
4202 			fctrl |= IXGBE_FCTRL_MPE;
4203 			vmolr |= IXGBE_VMOLR_MPE;
4204 		}
4205 		vlnctrl |= IXGBE_VLNCTRL_VFE;
4206 		hw->addr_ctrl.user_set_promisc = false;
4207 	}
4208 
4209 	/*
4210 	 * Write addresses to available RAR registers, if there is not
4211 	 * sufficient space to store all the addresses then enable
4212 	 * unicast promiscuous mode
4213 	 */
4214 	count = ixgbe_write_uc_addr_list(netdev, VMDQ_P(0));
4215 	if (count < 0) {
4216 		fctrl |= IXGBE_FCTRL_UPE;
4217 		vmolr |= IXGBE_VMOLR_ROPE;
4218 	}
4219 
4220 	/* Write addresses to the MTA, if the attempt fails
4221 	 * then we should just turn on promiscuous mode so
4222 	 * that we can at least receive multicast traffic
4223 	 */
4224 	count = ixgbe_write_mc_addr_list(netdev);
4225 	if (count < 0) {
4226 		fctrl |= IXGBE_FCTRL_MPE;
4227 		vmolr |= IXGBE_VMOLR_MPE;
4228 	} else if (count) {
4229 		vmolr |= IXGBE_VMOLR_ROMPE;
4230 	}
4231 
4232 	if (hw->mac.type != ixgbe_mac_82598EB) {
4233 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4234 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4235 			   IXGBE_VMOLR_ROPE);
4236 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4237 	}
4238 
4239 	/* This is useful for sniffing bad packets. */
4240 	if (adapter->netdev->features & NETIF_F_RXALL) {
4241 		/* UPE and MPE will be handled by normal PROMISC logic
4242 		 * in e1000e_set_rx_mode */
4243 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4244 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4245 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4246 
4247 		fctrl &= ~(IXGBE_FCTRL_DPF);
4248 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
4249 	}
4250 
4251 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4252 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4253 
4254 	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
4255 		ixgbe_vlan_strip_enable(adapter);
4256 	else
4257 		ixgbe_vlan_strip_disable(adapter);
4258 }
4259 
4260 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4261 {
4262 	int q_idx;
4263 
4264 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4265 		ixgbe_qv_init_lock(adapter->q_vector[q_idx]);
4266 		napi_enable(&adapter->q_vector[q_idx]->napi);
4267 	}
4268 }
4269 
4270 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
4271 {
4272 	int q_idx;
4273 
4274 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4275 		napi_disable(&adapter->q_vector[q_idx]->napi);
4276 		while (!ixgbe_qv_disable(adapter->q_vector[q_idx])) {
4277 			pr_info("QV %d locked\n", q_idx);
4278 			usleep_range(1000, 20000);
4279 		}
4280 	}
4281 }
4282 
4283 static void ixgbe_clear_vxlan_port(struct ixgbe_adapter *adapter)
4284 {
4285 	switch (adapter->hw.mac.type) {
4286 	case ixgbe_mac_X550:
4287 	case ixgbe_mac_X550EM_x:
4288 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_VXLANCTRL, 0);
4289 #ifdef CONFIG_IXGBE_VXLAN
4290 		adapter->vxlan_port = 0;
4291 #endif
4292 		break;
4293 	default:
4294 		break;
4295 	}
4296 }
4297 
4298 #ifdef CONFIG_IXGBE_DCB
4299 /**
4300  * ixgbe_configure_dcb - Configure DCB hardware
4301  * @adapter: ixgbe adapter struct
4302  *
4303  * This is called by the driver on open to configure the DCB hardware.
4304  * This is also called by the gennetlink interface when reconfiguring
4305  * the DCB state.
4306  */
4307 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
4308 {
4309 	struct ixgbe_hw *hw = &adapter->hw;
4310 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4311 
4312 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
4313 		if (hw->mac.type == ixgbe_mac_82598EB)
4314 			netif_set_gso_max_size(adapter->netdev, 65536);
4315 		return;
4316 	}
4317 
4318 	if (hw->mac.type == ixgbe_mac_82598EB)
4319 		netif_set_gso_max_size(adapter->netdev, 32768);
4320 
4321 #ifdef IXGBE_FCOE
4322 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
4323 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
4324 #endif
4325 
4326 	/* reconfigure the hardware */
4327 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
4328 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4329 						DCB_TX_CONFIG);
4330 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4331 						DCB_RX_CONFIG);
4332 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
4333 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
4334 		ixgbe_dcb_hw_ets(&adapter->hw,
4335 				 adapter->ixgbe_ieee_ets,
4336 				 max_frame);
4337 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
4338 					adapter->ixgbe_ieee_pfc->pfc_en,
4339 					adapter->ixgbe_ieee_ets->prio_tc);
4340 	}
4341 
4342 	/* Enable RSS Hash per TC */
4343 	if (hw->mac.type != ixgbe_mac_82598EB) {
4344 		u32 msb = 0;
4345 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
4346 
4347 		while (rss_i) {
4348 			msb++;
4349 			rss_i >>= 1;
4350 		}
4351 
4352 		/* write msb to all 8 TCs in one write */
4353 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
4354 	}
4355 }
4356 #endif
4357 
4358 /* Additional bittime to account for IXGBE framing */
4359 #define IXGBE_ETH_FRAMING 20
4360 
4361 /**
4362  * ixgbe_hpbthresh - calculate high water mark for flow control
4363  *
4364  * @adapter: board private structure to calculate for
4365  * @pb: packet buffer to calculate
4366  */
4367 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
4368 {
4369 	struct ixgbe_hw *hw = &adapter->hw;
4370 	struct net_device *dev = adapter->netdev;
4371 	int link, tc, kb, marker;
4372 	u32 dv_id, rx_pba;
4373 
4374 	/* Calculate max LAN frame size */
4375 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
4376 
4377 #ifdef IXGBE_FCOE
4378 	/* FCoE traffic class uses FCOE jumbo frames */
4379 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4380 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4381 	    (pb == ixgbe_fcoe_get_tc(adapter)))
4382 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4383 #endif
4384 
4385 	/* Calculate delay value for device */
4386 	switch (hw->mac.type) {
4387 	case ixgbe_mac_X540:
4388 	case ixgbe_mac_X550:
4389 	case ixgbe_mac_X550EM_x:
4390 		dv_id = IXGBE_DV_X540(link, tc);
4391 		break;
4392 	default:
4393 		dv_id = IXGBE_DV(link, tc);
4394 		break;
4395 	}
4396 
4397 	/* Loopback switch introduces additional latency */
4398 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4399 		dv_id += IXGBE_B2BT(tc);
4400 
4401 	/* Delay value is calculated in bit times convert to KB */
4402 	kb = IXGBE_BT2KB(dv_id);
4403 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
4404 
4405 	marker = rx_pba - kb;
4406 
4407 	/* It is possible that the packet buffer is not large enough
4408 	 * to provide required headroom. In this case throw an error
4409 	 * to user and a do the best we can.
4410 	 */
4411 	if (marker < 0) {
4412 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
4413 			    "headroom to support flow control."
4414 			    "Decrease MTU or number of traffic classes\n", pb);
4415 		marker = tc + 1;
4416 	}
4417 
4418 	return marker;
4419 }
4420 
4421 /**
4422  * ixgbe_lpbthresh - calculate low water mark for for flow control
4423  *
4424  * @adapter: board private structure to calculate for
4425  * @pb: packet buffer to calculate
4426  */
4427 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
4428 {
4429 	struct ixgbe_hw *hw = &adapter->hw;
4430 	struct net_device *dev = adapter->netdev;
4431 	int tc;
4432 	u32 dv_id;
4433 
4434 	/* Calculate max LAN frame size */
4435 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
4436 
4437 #ifdef IXGBE_FCOE
4438 	/* FCoE traffic class uses FCOE jumbo frames */
4439 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4440 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4441 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
4442 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4443 #endif
4444 
4445 	/* Calculate delay value for device */
4446 	switch (hw->mac.type) {
4447 	case ixgbe_mac_X540:
4448 	case ixgbe_mac_X550:
4449 	case ixgbe_mac_X550EM_x:
4450 		dv_id = IXGBE_LOW_DV_X540(tc);
4451 		break;
4452 	default:
4453 		dv_id = IXGBE_LOW_DV(tc);
4454 		break;
4455 	}
4456 
4457 	/* Delay value is calculated in bit times convert to KB */
4458 	return IXGBE_BT2KB(dv_id);
4459 }
4460 
4461 /*
4462  * ixgbe_pbthresh_setup - calculate and setup high low water marks
4463  */
4464 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
4465 {
4466 	struct ixgbe_hw *hw = &adapter->hw;
4467 	int num_tc = netdev_get_num_tc(adapter->netdev);
4468 	int i;
4469 
4470 	if (!num_tc)
4471 		num_tc = 1;
4472 
4473 	for (i = 0; i < num_tc; i++) {
4474 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
4475 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
4476 
4477 		/* Low water marks must not be larger than high water marks */
4478 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
4479 			hw->fc.low_water[i] = 0;
4480 	}
4481 
4482 	for (; i < MAX_TRAFFIC_CLASS; i++)
4483 		hw->fc.high_water[i] = 0;
4484 }
4485 
4486 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
4487 {
4488 	struct ixgbe_hw *hw = &adapter->hw;
4489 	int hdrm;
4490 	u8 tc = netdev_get_num_tc(adapter->netdev);
4491 
4492 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
4493 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
4494 		hdrm = 32 << adapter->fdir_pballoc;
4495 	else
4496 		hdrm = 0;
4497 
4498 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
4499 	ixgbe_pbthresh_setup(adapter);
4500 }
4501 
4502 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
4503 {
4504 	struct ixgbe_hw *hw = &adapter->hw;
4505 	struct hlist_node *node2;
4506 	struct ixgbe_fdir_filter *filter;
4507 
4508 	spin_lock(&adapter->fdir_perfect_lock);
4509 
4510 	if (!hlist_empty(&adapter->fdir_filter_list))
4511 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
4512 
4513 	hlist_for_each_entry_safe(filter, node2,
4514 				  &adapter->fdir_filter_list, fdir_node) {
4515 		ixgbe_fdir_write_perfect_filter_82599(hw,
4516 				&filter->filter,
4517 				filter->sw_idx,
4518 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
4519 				IXGBE_FDIR_DROP_QUEUE :
4520 				adapter->rx_ring[filter->action]->reg_idx);
4521 	}
4522 
4523 	spin_unlock(&adapter->fdir_perfect_lock);
4524 }
4525 
4526 static void ixgbe_macvlan_set_rx_mode(struct net_device *dev, unsigned int pool,
4527 				      struct ixgbe_adapter *adapter)
4528 {
4529 	struct ixgbe_hw *hw = &adapter->hw;
4530 	u32 vmolr;
4531 
4532 	/* No unicast promiscuous support for VMDQ devices. */
4533 	vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
4534 	vmolr |= (IXGBE_VMOLR_ROMPE | IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE);
4535 
4536 	/* clear the affected bit */
4537 	vmolr &= ~IXGBE_VMOLR_MPE;
4538 
4539 	if (dev->flags & IFF_ALLMULTI) {
4540 		vmolr |= IXGBE_VMOLR_MPE;
4541 	} else {
4542 		vmolr |= IXGBE_VMOLR_ROMPE;
4543 		hw->mac.ops.update_mc_addr_list(hw, dev);
4544 	}
4545 	ixgbe_write_uc_addr_list(adapter->netdev, pool);
4546 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
4547 }
4548 
4549 static void ixgbe_fwd_psrtype(struct ixgbe_fwd_adapter *vadapter)
4550 {
4551 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4552 	int rss_i = adapter->num_rx_queues_per_pool;
4553 	struct ixgbe_hw *hw = &adapter->hw;
4554 	u16 pool = vadapter->pool;
4555 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4556 		      IXGBE_PSRTYPE_UDPHDR |
4557 		      IXGBE_PSRTYPE_IPV4HDR |
4558 		      IXGBE_PSRTYPE_L2HDR |
4559 		      IXGBE_PSRTYPE_IPV6HDR;
4560 
4561 	if (hw->mac.type == ixgbe_mac_82598EB)
4562 		return;
4563 
4564 	if (rss_i > 3)
4565 		psrtype |= 2 << 29;
4566 	else if (rss_i > 1)
4567 		psrtype |= 1 << 29;
4568 
4569 	IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4570 }
4571 
4572 /**
4573  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4574  * @rx_ring: ring to free buffers from
4575  **/
4576 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
4577 {
4578 	struct device *dev = rx_ring->dev;
4579 	unsigned long size;
4580 	u16 i;
4581 
4582 	/* ring already cleared, nothing to do */
4583 	if (!rx_ring->rx_buffer_info)
4584 		return;
4585 
4586 	/* Free all the Rx ring sk_buffs */
4587 	for (i = 0; i < rx_ring->count; i++) {
4588 		struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
4589 
4590 		if (rx_buffer->skb) {
4591 			struct sk_buff *skb = rx_buffer->skb;
4592 			if (IXGBE_CB(skb)->page_released)
4593 				dma_unmap_page(dev,
4594 					       IXGBE_CB(skb)->dma,
4595 					       ixgbe_rx_bufsz(rx_ring),
4596 					       DMA_FROM_DEVICE);
4597 			dev_kfree_skb(skb);
4598 			rx_buffer->skb = NULL;
4599 		}
4600 
4601 		if (!rx_buffer->page)
4602 			continue;
4603 
4604 		dma_unmap_page(dev, rx_buffer->dma,
4605 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
4606 		__free_pages(rx_buffer->page, ixgbe_rx_pg_order(rx_ring));
4607 
4608 		rx_buffer->page = NULL;
4609 	}
4610 
4611 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4612 	memset(rx_ring->rx_buffer_info, 0, size);
4613 
4614 	/* Zero out the descriptor ring */
4615 	memset(rx_ring->desc, 0, rx_ring->size);
4616 
4617 	rx_ring->next_to_alloc = 0;
4618 	rx_ring->next_to_clean = 0;
4619 	rx_ring->next_to_use = 0;
4620 }
4621 
4622 static void ixgbe_disable_fwd_ring(struct ixgbe_fwd_adapter *vadapter,
4623 				   struct ixgbe_ring *rx_ring)
4624 {
4625 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4626 	int index = rx_ring->queue_index + vadapter->rx_base_queue;
4627 
4628 	/* shutdown specific queue receive and wait for dma to settle */
4629 	ixgbe_disable_rx_queue(adapter, rx_ring);
4630 	usleep_range(10000, 20000);
4631 	ixgbe_irq_disable_queues(adapter, ((u64)1 << index));
4632 	ixgbe_clean_rx_ring(rx_ring);
4633 	rx_ring->l2_accel_priv = NULL;
4634 }
4635 
4636 static int ixgbe_fwd_ring_down(struct net_device *vdev,
4637 			       struct ixgbe_fwd_adapter *accel)
4638 {
4639 	struct ixgbe_adapter *adapter = accel->real_adapter;
4640 	unsigned int rxbase = accel->rx_base_queue;
4641 	unsigned int txbase = accel->tx_base_queue;
4642 	int i;
4643 
4644 	netif_tx_stop_all_queues(vdev);
4645 
4646 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4647 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4648 		adapter->rx_ring[rxbase + i]->netdev = adapter->netdev;
4649 	}
4650 
4651 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4652 		adapter->tx_ring[txbase + i]->l2_accel_priv = NULL;
4653 		adapter->tx_ring[txbase + i]->netdev = adapter->netdev;
4654 	}
4655 
4656 
4657 	return 0;
4658 }
4659 
4660 static int ixgbe_fwd_ring_up(struct net_device *vdev,
4661 			     struct ixgbe_fwd_adapter *accel)
4662 {
4663 	struct ixgbe_adapter *adapter = accel->real_adapter;
4664 	unsigned int rxbase, txbase, queues;
4665 	int i, baseq, err = 0;
4666 
4667 	if (!test_bit(accel->pool, &adapter->fwd_bitmask))
4668 		return 0;
4669 
4670 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
4671 	netdev_dbg(vdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
4672 		   accel->pool, adapter->num_rx_pools,
4673 		   baseq, baseq + adapter->num_rx_queues_per_pool,
4674 		   adapter->fwd_bitmask);
4675 
4676 	accel->netdev = vdev;
4677 	accel->rx_base_queue = rxbase = baseq;
4678 	accel->tx_base_queue = txbase = baseq;
4679 
4680 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
4681 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4682 
4683 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4684 		adapter->rx_ring[rxbase + i]->netdev = vdev;
4685 		adapter->rx_ring[rxbase + i]->l2_accel_priv = accel;
4686 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[rxbase + i]);
4687 	}
4688 
4689 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4690 		adapter->tx_ring[txbase + i]->netdev = vdev;
4691 		adapter->tx_ring[txbase + i]->l2_accel_priv = accel;
4692 	}
4693 
4694 	queues = min_t(unsigned int,
4695 		       adapter->num_rx_queues_per_pool, vdev->num_tx_queues);
4696 	err = netif_set_real_num_tx_queues(vdev, queues);
4697 	if (err)
4698 		goto fwd_queue_err;
4699 
4700 	err = netif_set_real_num_rx_queues(vdev, queues);
4701 	if (err)
4702 		goto fwd_queue_err;
4703 
4704 	if (is_valid_ether_addr(vdev->dev_addr))
4705 		ixgbe_add_mac_filter(adapter, vdev->dev_addr, accel->pool);
4706 
4707 	ixgbe_fwd_psrtype(accel);
4708 	ixgbe_macvlan_set_rx_mode(vdev, accel->pool, adapter);
4709 	return err;
4710 fwd_queue_err:
4711 	ixgbe_fwd_ring_down(vdev, accel);
4712 	return err;
4713 }
4714 
4715 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
4716 {
4717 	struct net_device *upper;
4718 	struct list_head *iter;
4719 	int err;
4720 
4721 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
4722 		if (netif_is_macvlan(upper)) {
4723 			struct macvlan_dev *dfwd = netdev_priv(upper);
4724 			struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
4725 
4726 			if (dfwd->fwd_priv) {
4727 				err = ixgbe_fwd_ring_up(upper, vadapter);
4728 				if (err)
4729 					continue;
4730 			}
4731 		}
4732 	}
4733 }
4734 
4735 static void ixgbe_configure(struct ixgbe_adapter *adapter)
4736 {
4737 	struct ixgbe_hw *hw = &adapter->hw;
4738 
4739 	ixgbe_configure_pb(adapter);
4740 #ifdef CONFIG_IXGBE_DCB
4741 	ixgbe_configure_dcb(adapter);
4742 #endif
4743 	/*
4744 	 * We must restore virtualization before VLANs or else
4745 	 * the VLVF registers will not be populated
4746 	 */
4747 	ixgbe_configure_virtualization(adapter);
4748 
4749 	ixgbe_set_rx_mode(adapter->netdev);
4750 	ixgbe_restore_vlan(adapter);
4751 
4752 	switch (hw->mac.type) {
4753 	case ixgbe_mac_82599EB:
4754 	case ixgbe_mac_X540:
4755 		hw->mac.ops.disable_rx_buff(hw);
4756 		break;
4757 	default:
4758 		break;
4759 	}
4760 
4761 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
4762 		ixgbe_init_fdir_signature_82599(&adapter->hw,
4763 						adapter->fdir_pballoc);
4764 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
4765 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
4766 					      adapter->fdir_pballoc);
4767 		ixgbe_fdir_filter_restore(adapter);
4768 	}
4769 
4770 	switch (hw->mac.type) {
4771 	case ixgbe_mac_82599EB:
4772 	case ixgbe_mac_X540:
4773 		hw->mac.ops.enable_rx_buff(hw);
4774 		break;
4775 	default:
4776 		break;
4777 	}
4778 
4779 #ifdef IXGBE_FCOE
4780 	/* configure FCoE L2 filters, redirection table, and Rx control */
4781 	ixgbe_configure_fcoe(adapter);
4782 
4783 #endif /* IXGBE_FCOE */
4784 	ixgbe_configure_tx(adapter);
4785 	ixgbe_configure_rx(adapter);
4786 	ixgbe_configure_dfwd(adapter);
4787 }
4788 
4789 /**
4790  * ixgbe_sfp_link_config - set up SFP+ link
4791  * @adapter: pointer to private adapter struct
4792  **/
4793 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
4794 {
4795 	/*
4796 	 * We are assuming the worst case scenario here, and that
4797 	 * is that an SFP was inserted/removed after the reset
4798 	 * but before SFP detection was enabled.  As such the best
4799 	 * solution is to just start searching as soon as we start
4800 	 */
4801 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
4802 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
4803 
4804 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
4805 }
4806 
4807 /**
4808  * ixgbe_non_sfp_link_config - set up non-SFP+ link
4809  * @hw: pointer to private hardware struct
4810  *
4811  * Returns 0 on success, negative on failure
4812  **/
4813 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
4814 {
4815 	u32 speed;
4816 	bool autoneg, link_up = false;
4817 	int ret = IXGBE_ERR_LINK_SETUP;
4818 
4819 	if (hw->mac.ops.check_link)
4820 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
4821 
4822 	if (ret)
4823 		return ret;
4824 
4825 	speed = hw->phy.autoneg_advertised;
4826 	if ((!speed) && (hw->mac.ops.get_link_capabilities))
4827 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
4828 							&autoneg);
4829 	if (ret)
4830 		return ret;
4831 
4832 	if (hw->mac.ops.setup_link)
4833 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
4834 
4835 	return ret;
4836 }
4837 
4838 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
4839 {
4840 	struct ixgbe_hw *hw = &adapter->hw;
4841 	u32 gpie = 0;
4842 
4843 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
4844 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
4845 		       IXGBE_GPIE_OCD;
4846 		gpie |= IXGBE_GPIE_EIAME;
4847 		/*
4848 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
4849 		 * this saves a register write for every interrupt
4850 		 */
4851 		switch (hw->mac.type) {
4852 		case ixgbe_mac_82598EB:
4853 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4854 			break;
4855 		case ixgbe_mac_82599EB:
4856 		case ixgbe_mac_X540:
4857 		case ixgbe_mac_X550:
4858 		case ixgbe_mac_X550EM_x:
4859 		default:
4860 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
4861 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
4862 			break;
4863 		}
4864 	} else {
4865 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
4866 		 * specifically only auto mask tx and rx interrupts */
4867 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4868 	}
4869 
4870 	/* XXX: to interrupt immediately for EICS writes, enable this */
4871 	/* gpie |= IXGBE_GPIE_EIMEN; */
4872 
4873 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
4874 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
4875 
4876 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4877 		case IXGBE_82599_VMDQ_8Q_MASK:
4878 			gpie |= IXGBE_GPIE_VTMODE_16;
4879 			break;
4880 		case IXGBE_82599_VMDQ_4Q_MASK:
4881 			gpie |= IXGBE_GPIE_VTMODE_32;
4882 			break;
4883 		default:
4884 			gpie |= IXGBE_GPIE_VTMODE_64;
4885 			break;
4886 		}
4887 	}
4888 
4889 	/* Enable Thermal over heat sensor interrupt */
4890 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
4891 		switch (adapter->hw.mac.type) {
4892 		case ixgbe_mac_82599EB:
4893 			gpie |= IXGBE_SDP0_GPIEN_8259X;
4894 			break;
4895 		case ixgbe_mac_X540:
4896 			gpie |= IXGBE_EIMS_TS;
4897 			break;
4898 		default:
4899 			break;
4900 		}
4901 	}
4902 
4903 	/* Enable fan failure interrupt */
4904 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
4905 		gpie |= IXGBE_SDP1_GPIEN(hw);
4906 
4907 	if (hw->mac.type == ixgbe_mac_82599EB) {
4908 		gpie |= IXGBE_SDP1_GPIEN_8259X;
4909 		gpie |= IXGBE_SDP2_GPIEN_8259X;
4910 	}
4911 
4912 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
4913 }
4914 
4915 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
4916 {
4917 	struct ixgbe_hw *hw = &adapter->hw;
4918 	int err;
4919 	u32 ctrl_ext;
4920 
4921 	ixgbe_get_hw_control(adapter);
4922 	ixgbe_setup_gpie(adapter);
4923 
4924 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
4925 		ixgbe_configure_msix(adapter);
4926 	else
4927 		ixgbe_configure_msi_and_legacy(adapter);
4928 
4929 	/* enable the optics for 82599 SFP+ fiber */
4930 	if (hw->mac.ops.enable_tx_laser)
4931 		hw->mac.ops.enable_tx_laser(hw);
4932 
4933 	if (hw->phy.ops.set_phy_power)
4934 		hw->phy.ops.set_phy_power(hw, true);
4935 
4936 	smp_mb__before_atomic();
4937 	clear_bit(__IXGBE_DOWN, &adapter->state);
4938 	ixgbe_napi_enable_all(adapter);
4939 
4940 	if (ixgbe_is_sfp(hw)) {
4941 		ixgbe_sfp_link_config(adapter);
4942 	} else {
4943 		err = ixgbe_non_sfp_link_config(hw);
4944 		if (err)
4945 			e_err(probe, "link_config FAILED %d\n", err);
4946 	}
4947 
4948 	/* clear any pending interrupts, may auto mask */
4949 	IXGBE_READ_REG(hw, IXGBE_EICR);
4950 	ixgbe_irq_enable(adapter, true, true);
4951 
4952 	/*
4953 	 * If this adapter has a fan, check to see if we had a failure
4954 	 * before we enabled the interrupt.
4955 	 */
4956 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
4957 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
4958 		if (esdp & IXGBE_ESDP_SDP1)
4959 			e_crit(drv, "Fan has stopped, replace the adapter\n");
4960 	}
4961 
4962 	/* bring the link up in the watchdog, this could race with our first
4963 	 * link up interrupt but shouldn't be a problem */
4964 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4965 	adapter->link_check_timeout = jiffies;
4966 	mod_timer(&adapter->service_timer, jiffies);
4967 
4968 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
4969 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
4970 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
4971 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
4972 }
4973 
4974 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
4975 {
4976 	WARN_ON(in_interrupt());
4977 	/* put off any impending NetWatchDogTimeout */
4978 	adapter->netdev->trans_start = jiffies;
4979 
4980 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
4981 		usleep_range(1000, 2000);
4982 	ixgbe_down(adapter);
4983 	/*
4984 	 * If SR-IOV enabled then wait a bit before bringing the adapter
4985 	 * back up to give the VFs time to respond to the reset.  The
4986 	 * two second wait is based upon the watchdog timer cycle in
4987 	 * the VF driver.
4988 	 */
4989 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4990 		msleep(2000);
4991 	ixgbe_up(adapter);
4992 	clear_bit(__IXGBE_RESETTING, &adapter->state);
4993 }
4994 
4995 void ixgbe_up(struct ixgbe_adapter *adapter)
4996 {
4997 	/* hardware has been reset, we need to reload some things */
4998 	ixgbe_configure(adapter);
4999 
5000 	ixgbe_up_complete(adapter);
5001 }
5002 
5003 void ixgbe_reset(struct ixgbe_adapter *adapter)
5004 {
5005 	struct ixgbe_hw *hw = &adapter->hw;
5006 	struct net_device *netdev = adapter->netdev;
5007 	int err;
5008 	u8 old_addr[ETH_ALEN];
5009 
5010 	if (ixgbe_removed(hw->hw_addr))
5011 		return;
5012 	/* lock SFP init bit to prevent race conditions with the watchdog */
5013 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5014 		usleep_range(1000, 2000);
5015 
5016 	/* clear all SFP and link config related flags while holding SFP_INIT */
5017 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
5018 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
5019 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5020 
5021 	err = hw->mac.ops.init_hw(hw);
5022 	switch (err) {
5023 	case 0:
5024 	case IXGBE_ERR_SFP_NOT_PRESENT:
5025 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
5026 		break;
5027 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
5028 		e_dev_err("master disable timed out\n");
5029 		break;
5030 	case IXGBE_ERR_EEPROM_VERSION:
5031 		/* We are running on a pre-production device, log a warning */
5032 		e_dev_warn("This device is a pre-production adapter/LOM. "
5033 			   "Please be aware there may be issues associated with "
5034 			   "your hardware.  If you are experiencing problems "
5035 			   "please contact your Intel or hardware "
5036 			   "representative who provided you with this "
5037 			   "hardware.\n");
5038 		break;
5039 	default:
5040 		e_dev_err("Hardware Error: %d\n", err);
5041 	}
5042 
5043 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5044 	/* do not flush user set addresses */
5045 	memcpy(old_addr, &adapter->mac_table[0].addr, netdev->addr_len);
5046 	ixgbe_flush_sw_mac_table(adapter);
5047 	ixgbe_mac_set_default_filter(adapter, old_addr);
5048 
5049 	/* update SAN MAC vmdq pool selection */
5050 	if (hw->mac.san_mac_rar_index)
5051 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
5052 
5053 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
5054 		ixgbe_ptp_reset(adapter);
5055 
5056 	if (hw->phy.ops.set_phy_power) {
5057 		if (!netif_running(adapter->netdev) && !adapter->wol)
5058 			hw->phy.ops.set_phy_power(hw, false);
5059 		else
5060 			hw->phy.ops.set_phy_power(hw, true);
5061 	}
5062 }
5063 
5064 /**
5065  * ixgbe_clean_tx_ring - Free Tx Buffers
5066  * @tx_ring: ring to be cleaned
5067  **/
5068 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
5069 {
5070 	struct ixgbe_tx_buffer *tx_buffer_info;
5071 	unsigned long size;
5072 	u16 i;
5073 
5074 	/* ring already cleared, nothing to do */
5075 	if (!tx_ring->tx_buffer_info)
5076 		return;
5077 
5078 	/* Free all the Tx ring sk_buffs */
5079 	for (i = 0; i < tx_ring->count; i++) {
5080 		tx_buffer_info = &tx_ring->tx_buffer_info[i];
5081 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer_info);
5082 	}
5083 
5084 	netdev_tx_reset_queue(txring_txq(tx_ring));
5085 
5086 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5087 	memset(tx_ring->tx_buffer_info, 0, size);
5088 
5089 	/* Zero out the descriptor ring */
5090 	memset(tx_ring->desc, 0, tx_ring->size);
5091 
5092 	tx_ring->next_to_use = 0;
5093 	tx_ring->next_to_clean = 0;
5094 }
5095 
5096 /**
5097  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
5098  * @adapter: board private structure
5099  **/
5100 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
5101 {
5102 	int i;
5103 
5104 	for (i = 0; i < adapter->num_rx_queues; i++)
5105 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
5106 }
5107 
5108 /**
5109  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
5110  * @adapter: board private structure
5111  **/
5112 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
5113 {
5114 	int i;
5115 
5116 	for (i = 0; i < adapter->num_tx_queues; i++)
5117 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
5118 }
5119 
5120 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
5121 {
5122 	struct hlist_node *node2;
5123 	struct ixgbe_fdir_filter *filter;
5124 
5125 	spin_lock(&adapter->fdir_perfect_lock);
5126 
5127 	hlist_for_each_entry_safe(filter, node2,
5128 				  &adapter->fdir_filter_list, fdir_node) {
5129 		hlist_del(&filter->fdir_node);
5130 		kfree(filter);
5131 	}
5132 	adapter->fdir_filter_count = 0;
5133 
5134 	spin_unlock(&adapter->fdir_perfect_lock);
5135 }
5136 
5137 void ixgbe_down(struct ixgbe_adapter *adapter)
5138 {
5139 	struct net_device *netdev = adapter->netdev;
5140 	struct ixgbe_hw *hw = &adapter->hw;
5141 	struct net_device *upper;
5142 	struct list_head *iter;
5143 	int i;
5144 
5145 	/* signal that we are down to the interrupt handler */
5146 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
5147 		return; /* do nothing if already down */
5148 
5149 	/* disable receives */
5150 	hw->mac.ops.disable_rx(hw);
5151 
5152 	/* disable all enabled rx queues */
5153 	for (i = 0; i < adapter->num_rx_queues; i++)
5154 		/* this call also flushes the previous write */
5155 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
5156 
5157 	usleep_range(10000, 20000);
5158 
5159 	netif_tx_stop_all_queues(netdev);
5160 
5161 	/* call carrier off first to avoid false dev_watchdog timeouts */
5162 	netif_carrier_off(netdev);
5163 	netif_tx_disable(netdev);
5164 
5165 	/* disable any upper devices */
5166 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
5167 		if (netif_is_macvlan(upper)) {
5168 			struct macvlan_dev *vlan = netdev_priv(upper);
5169 
5170 			if (vlan->fwd_priv) {
5171 				netif_tx_stop_all_queues(upper);
5172 				netif_carrier_off(upper);
5173 				netif_tx_disable(upper);
5174 			}
5175 		}
5176 	}
5177 
5178 	ixgbe_irq_disable(adapter);
5179 
5180 	ixgbe_napi_disable_all(adapter);
5181 
5182 	adapter->flags2 &= ~(IXGBE_FLAG2_FDIR_REQUIRES_REINIT |
5183 			     IXGBE_FLAG2_RESET_REQUESTED);
5184 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5185 
5186 	del_timer_sync(&adapter->service_timer);
5187 
5188 	if (adapter->num_vfs) {
5189 		/* Clear EITR Select mapping */
5190 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
5191 
5192 		/* Mark all the VFs as inactive */
5193 		for (i = 0 ; i < adapter->num_vfs; i++)
5194 			adapter->vfinfo[i].clear_to_send = false;
5195 
5196 		/* ping all the active vfs to let them know we are going down */
5197 		ixgbe_ping_all_vfs(adapter);
5198 
5199 		/* Disable all VFTE/VFRE TX/RX */
5200 		ixgbe_disable_tx_rx(adapter);
5201 	}
5202 
5203 	/* disable transmits in the hardware now that interrupts are off */
5204 	for (i = 0; i < adapter->num_tx_queues; i++) {
5205 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
5206 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5207 	}
5208 
5209 	/* Disable the Tx DMA engine on 82599 and later MAC */
5210 	switch (hw->mac.type) {
5211 	case ixgbe_mac_82599EB:
5212 	case ixgbe_mac_X540:
5213 	case ixgbe_mac_X550:
5214 	case ixgbe_mac_X550EM_x:
5215 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5216 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5217 				 ~IXGBE_DMATXCTL_TE));
5218 		break;
5219 	default:
5220 		break;
5221 	}
5222 
5223 	if (!pci_channel_offline(adapter->pdev))
5224 		ixgbe_reset(adapter);
5225 
5226 	/* power down the optics for 82599 SFP+ fiber */
5227 	if (hw->mac.ops.disable_tx_laser)
5228 		hw->mac.ops.disable_tx_laser(hw);
5229 
5230 	ixgbe_clean_all_tx_rings(adapter);
5231 	ixgbe_clean_all_rx_rings(adapter);
5232 
5233 #ifdef CONFIG_IXGBE_DCA
5234 	/* since we reset the hardware DCA settings were cleared */
5235 	ixgbe_setup_dca(adapter);
5236 #endif
5237 }
5238 
5239 /**
5240  * ixgbe_tx_timeout - Respond to a Tx Hang
5241  * @netdev: network interface device structure
5242  **/
5243 static void ixgbe_tx_timeout(struct net_device *netdev)
5244 {
5245 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5246 
5247 	/* Do the reset outside of interrupt context */
5248 	ixgbe_tx_timeout_reset(adapter);
5249 }
5250 
5251 /**
5252  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
5253  * @adapter: board private structure to initialize
5254  *
5255  * ixgbe_sw_init initializes the Adapter private data structure.
5256  * Fields are initialized based on PCI device information and
5257  * OS network device settings (MTU size).
5258  **/
5259 static int ixgbe_sw_init(struct ixgbe_adapter *adapter)
5260 {
5261 	struct ixgbe_hw *hw = &adapter->hw;
5262 	struct pci_dev *pdev = adapter->pdev;
5263 	unsigned int rss, fdir;
5264 	u32 fwsm;
5265 #ifdef CONFIG_IXGBE_DCB
5266 	int j;
5267 	struct tc_configuration *tc;
5268 #endif
5269 
5270 	/* PCI config space info */
5271 
5272 	hw->vendor_id = pdev->vendor;
5273 	hw->device_id = pdev->device;
5274 	hw->revision_id = pdev->revision;
5275 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
5276 	hw->subsystem_device_id = pdev->subsystem_device;
5277 
5278 	/* Set common capability flags and settings */
5279 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
5280 	adapter->ring_feature[RING_F_RSS].limit = rss;
5281 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
5282 	adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
5283 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
5284 	adapter->atr_sample_rate = 20;
5285 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
5286 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
5287 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
5288 #ifdef CONFIG_IXGBE_DCA
5289 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
5290 #endif
5291 #ifdef IXGBE_FCOE
5292 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
5293 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5294 #ifdef CONFIG_IXGBE_DCB
5295 	/* Default traffic class to use for FCoE */
5296 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
5297 #endif /* CONFIG_IXGBE_DCB */
5298 #endif /* IXGBE_FCOE */
5299 
5300 	adapter->mac_table = kzalloc(sizeof(struct ixgbe_mac_addr) *
5301 				     hw->mac.num_rar_entries,
5302 				     GFP_ATOMIC);
5303 
5304 	/* Set MAC specific capability flags and exceptions */
5305 	switch (hw->mac.type) {
5306 	case ixgbe_mac_82598EB:
5307 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
5308 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
5309 
5310 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
5311 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
5312 
5313 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
5314 		adapter->ring_feature[RING_F_FDIR].limit = 0;
5315 		adapter->atr_sample_rate = 0;
5316 		adapter->fdir_pballoc = 0;
5317 #ifdef IXGBE_FCOE
5318 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
5319 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5320 #ifdef CONFIG_IXGBE_DCB
5321 		adapter->fcoe.up = 0;
5322 #endif /* IXGBE_DCB */
5323 #endif /* IXGBE_FCOE */
5324 		break;
5325 	case ixgbe_mac_82599EB:
5326 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
5327 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5328 		break;
5329 	case ixgbe_mac_X540:
5330 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
5331 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
5332 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5333 		break;
5334 	case ixgbe_mac_X550EM_x:
5335 	case ixgbe_mac_X550:
5336 #ifdef CONFIG_IXGBE_DCA
5337 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
5338 #endif
5339 #ifdef CONFIG_IXGBE_VXLAN
5340 		adapter->flags |= IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE;
5341 #endif
5342 		break;
5343 	default:
5344 		break;
5345 	}
5346 
5347 #ifdef IXGBE_FCOE
5348 	/* FCoE support exists, always init the FCoE lock */
5349 	spin_lock_init(&adapter->fcoe.lock);
5350 
5351 #endif
5352 	/* n-tuple support exists, always init our spinlock */
5353 	spin_lock_init(&adapter->fdir_perfect_lock);
5354 
5355 #ifdef CONFIG_IXGBE_DCB
5356 	switch (hw->mac.type) {
5357 	case ixgbe_mac_X540:
5358 	case ixgbe_mac_X550:
5359 	case ixgbe_mac_X550EM_x:
5360 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
5361 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
5362 		break;
5363 	default:
5364 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
5365 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
5366 		break;
5367 	}
5368 
5369 	/* Configure DCB traffic classes */
5370 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
5371 		tc = &adapter->dcb_cfg.tc_config[j];
5372 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
5373 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
5374 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
5375 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
5376 		tc->dcb_pfc = pfc_disabled;
5377 	}
5378 
5379 	/* Initialize default user to priority mapping, UPx->TC0 */
5380 	tc = &adapter->dcb_cfg.tc_config[0];
5381 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
5382 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
5383 
5384 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
5385 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
5386 	adapter->dcb_cfg.pfc_mode_enable = false;
5387 	adapter->dcb_set_bitmap = 0x00;
5388 	adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
5389 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
5390 	       sizeof(adapter->temp_dcb_cfg));
5391 
5392 #endif
5393 
5394 	/* default flow control settings */
5395 	hw->fc.requested_mode = ixgbe_fc_full;
5396 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
5397 	ixgbe_pbthresh_setup(adapter);
5398 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
5399 	hw->fc.send_xon = true;
5400 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
5401 
5402 #ifdef CONFIG_PCI_IOV
5403 	if (max_vfs > 0)
5404 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
5405 
5406 	/* assign number of SR-IOV VFs */
5407 	if (hw->mac.type != ixgbe_mac_82598EB) {
5408 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
5409 			adapter->num_vfs = 0;
5410 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
5411 		} else {
5412 			adapter->num_vfs = max_vfs;
5413 		}
5414 	}
5415 #endif /* CONFIG_PCI_IOV */
5416 
5417 	/* enable itr by default in dynamic mode */
5418 	adapter->rx_itr_setting = 1;
5419 	adapter->tx_itr_setting = 1;
5420 
5421 	/* set default ring sizes */
5422 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
5423 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
5424 
5425 	/* set default work limits */
5426 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
5427 
5428 	/* initialize eeprom parameters */
5429 	if (ixgbe_init_eeprom_params_generic(hw)) {
5430 		e_dev_err("EEPROM initialization failed\n");
5431 		return -EIO;
5432 	}
5433 
5434 	/* PF holds first pool slot */
5435 	set_bit(0, &adapter->fwd_bitmask);
5436 	set_bit(__IXGBE_DOWN, &adapter->state);
5437 
5438 	return 0;
5439 }
5440 
5441 /**
5442  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
5443  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
5444  *
5445  * Return 0 on success, negative on failure
5446  **/
5447 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
5448 {
5449 	struct device *dev = tx_ring->dev;
5450 	int orig_node = dev_to_node(dev);
5451 	int ring_node = -1;
5452 	int size;
5453 
5454 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5455 
5456 	if (tx_ring->q_vector)
5457 		ring_node = tx_ring->q_vector->numa_node;
5458 
5459 	tx_ring->tx_buffer_info = vzalloc_node(size, ring_node);
5460 	if (!tx_ring->tx_buffer_info)
5461 		tx_ring->tx_buffer_info = vzalloc(size);
5462 	if (!tx_ring->tx_buffer_info)
5463 		goto err;
5464 
5465 	u64_stats_init(&tx_ring->syncp);
5466 
5467 	/* round up to nearest 4K */
5468 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
5469 	tx_ring->size = ALIGN(tx_ring->size, 4096);
5470 
5471 	set_dev_node(dev, ring_node);
5472 	tx_ring->desc = dma_alloc_coherent(dev,
5473 					   tx_ring->size,
5474 					   &tx_ring->dma,
5475 					   GFP_KERNEL);
5476 	set_dev_node(dev, orig_node);
5477 	if (!tx_ring->desc)
5478 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
5479 						   &tx_ring->dma, GFP_KERNEL);
5480 	if (!tx_ring->desc)
5481 		goto err;
5482 
5483 	tx_ring->next_to_use = 0;
5484 	tx_ring->next_to_clean = 0;
5485 	return 0;
5486 
5487 err:
5488 	vfree(tx_ring->tx_buffer_info);
5489 	tx_ring->tx_buffer_info = NULL;
5490 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
5491 	return -ENOMEM;
5492 }
5493 
5494 /**
5495  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
5496  * @adapter: board private structure
5497  *
5498  * If this function returns with an error, then it's possible one or
5499  * more of the rings is populated (while the rest are not).  It is the
5500  * callers duty to clean those orphaned rings.
5501  *
5502  * Return 0 on success, negative on failure
5503  **/
5504 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
5505 {
5506 	int i, err = 0;
5507 
5508 	for (i = 0; i < adapter->num_tx_queues; i++) {
5509 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
5510 		if (!err)
5511 			continue;
5512 
5513 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
5514 		goto err_setup_tx;
5515 	}
5516 
5517 	return 0;
5518 err_setup_tx:
5519 	/* rewind the index freeing the rings as we go */
5520 	while (i--)
5521 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
5522 	return err;
5523 }
5524 
5525 /**
5526  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
5527  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
5528  *
5529  * Returns 0 on success, negative on failure
5530  **/
5531 int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
5532 {
5533 	struct device *dev = rx_ring->dev;
5534 	int orig_node = dev_to_node(dev);
5535 	int ring_node = -1;
5536 	int size;
5537 
5538 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
5539 
5540 	if (rx_ring->q_vector)
5541 		ring_node = rx_ring->q_vector->numa_node;
5542 
5543 	rx_ring->rx_buffer_info = vzalloc_node(size, ring_node);
5544 	if (!rx_ring->rx_buffer_info)
5545 		rx_ring->rx_buffer_info = vzalloc(size);
5546 	if (!rx_ring->rx_buffer_info)
5547 		goto err;
5548 
5549 	u64_stats_init(&rx_ring->syncp);
5550 
5551 	/* Round up to nearest 4K */
5552 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
5553 	rx_ring->size = ALIGN(rx_ring->size, 4096);
5554 
5555 	set_dev_node(dev, ring_node);
5556 	rx_ring->desc = dma_alloc_coherent(dev,
5557 					   rx_ring->size,
5558 					   &rx_ring->dma,
5559 					   GFP_KERNEL);
5560 	set_dev_node(dev, orig_node);
5561 	if (!rx_ring->desc)
5562 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
5563 						   &rx_ring->dma, GFP_KERNEL);
5564 	if (!rx_ring->desc)
5565 		goto err;
5566 
5567 	rx_ring->next_to_clean = 0;
5568 	rx_ring->next_to_use = 0;
5569 
5570 	return 0;
5571 err:
5572 	vfree(rx_ring->rx_buffer_info);
5573 	rx_ring->rx_buffer_info = NULL;
5574 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
5575 	return -ENOMEM;
5576 }
5577 
5578 /**
5579  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
5580  * @adapter: board private structure
5581  *
5582  * If this function returns with an error, then it's possible one or
5583  * more of the rings is populated (while the rest are not).  It is the
5584  * callers duty to clean those orphaned rings.
5585  *
5586  * Return 0 on success, negative on failure
5587  **/
5588 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
5589 {
5590 	int i, err = 0;
5591 
5592 	for (i = 0; i < adapter->num_rx_queues; i++) {
5593 		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
5594 		if (!err)
5595 			continue;
5596 
5597 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
5598 		goto err_setup_rx;
5599 	}
5600 
5601 #ifdef IXGBE_FCOE
5602 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
5603 	if (!err)
5604 #endif
5605 		return 0;
5606 err_setup_rx:
5607 	/* rewind the index freeing the rings as we go */
5608 	while (i--)
5609 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
5610 	return err;
5611 }
5612 
5613 /**
5614  * ixgbe_free_tx_resources - Free Tx Resources per Queue
5615  * @tx_ring: Tx descriptor ring for a specific queue
5616  *
5617  * Free all transmit software resources
5618  **/
5619 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
5620 {
5621 	ixgbe_clean_tx_ring(tx_ring);
5622 
5623 	vfree(tx_ring->tx_buffer_info);
5624 	tx_ring->tx_buffer_info = NULL;
5625 
5626 	/* if not set, then don't free */
5627 	if (!tx_ring->desc)
5628 		return;
5629 
5630 	dma_free_coherent(tx_ring->dev, tx_ring->size,
5631 			  tx_ring->desc, tx_ring->dma);
5632 
5633 	tx_ring->desc = NULL;
5634 }
5635 
5636 /**
5637  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
5638  * @adapter: board private structure
5639  *
5640  * Free all transmit software resources
5641  **/
5642 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
5643 {
5644 	int i;
5645 
5646 	for (i = 0; i < adapter->num_tx_queues; i++)
5647 		if (adapter->tx_ring[i]->desc)
5648 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
5649 }
5650 
5651 /**
5652  * ixgbe_free_rx_resources - Free Rx Resources
5653  * @rx_ring: ring to clean the resources from
5654  *
5655  * Free all receive software resources
5656  **/
5657 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
5658 {
5659 	ixgbe_clean_rx_ring(rx_ring);
5660 
5661 	vfree(rx_ring->rx_buffer_info);
5662 	rx_ring->rx_buffer_info = NULL;
5663 
5664 	/* if not set, then don't free */
5665 	if (!rx_ring->desc)
5666 		return;
5667 
5668 	dma_free_coherent(rx_ring->dev, rx_ring->size,
5669 			  rx_ring->desc, rx_ring->dma);
5670 
5671 	rx_ring->desc = NULL;
5672 }
5673 
5674 /**
5675  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
5676  * @adapter: board private structure
5677  *
5678  * Free all receive software resources
5679  **/
5680 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
5681 {
5682 	int i;
5683 
5684 #ifdef IXGBE_FCOE
5685 	ixgbe_free_fcoe_ddp_resources(adapter);
5686 
5687 #endif
5688 	for (i = 0; i < adapter->num_rx_queues; i++)
5689 		if (adapter->rx_ring[i]->desc)
5690 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
5691 }
5692 
5693 /**
5694  * ixgbe_change_mtu - Change the Maximum Transfer Unit
5695  * @netdev: network interface device structure
5696  * @new_mtu: new value for maximum frame size
5697  *
5698  * Returns 0 on success, negative on failure
5699  **/
5700 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
5701 {
5702 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5703 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
5704 
5705 	/* MTU < 68 is an error and causes problems on some kernels */
5706 	if ((new_mtu < 68) || (max_frame > IXGBE_MAX_JUMBO_FRAME_SIZE))
5707 		return -EINVAL;
5708 
5709 	/*
5710 	 * For 82599EB we cannot allow legacy VFs to enable their receive
5711 	 * paths when MTU greater than 1500 is configured.  So display a
5712 	 * warning that legacy VFs will be disabled.
5713 	 */
5714 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
5715 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
5716 	    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
5717 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
5718 
5719 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
5720 
5721 	/* must set new MTU before calling down or up */
5722 	netdev->mtu = new_mtu;
5723 
5724 	if (netif_running(netdev))
5725 		ixgbe_reinit_locked(adapter);
5726 
5727 	return 0;
5728 }
5729 
5730 /**
5731  * ixgbe_open - Called when a network interface is made active
5732  * @netdev: network interface device structure
5733  *
5734  * Returns 0 on success, negative value on failure
5735  *
5736  * The open entry point is called when a network interface is made
5737  * active by the system (IFF_UP).  At this point all resources needed
5738  * for transmit and receive operations are allocated, the interrupt
5739  * handler is registered with the OS, the watchdog timer is started,
5740  * and the stack is notified that the interface is ready.
5741  **/
5742 static int ixgbe_open(struct net_device *netdev)
5743 {
5744 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5745 	struct ixgbe_hw *hw = &adapter->hw;
5746 	int err, queues;
5747 
5748 	/* disallow open during test */
5749 	if (test_bit(__IXGBE_TESTING, &adapter->state))
5750 		return -EBUSY;
5751 
5752 	netif_carrier_off(netdev);
5753 
5754 	/* allocate transmit descriptors */
5755 	err = ixgbe_setup_all_tx_resources(adapter);
5756 	if (err)
5757 		goto err_setup_tx;
5758 
5759 	/* allocate receive descriptors */
5760 	err = ixgbe_setup_all_rx_resources(adapter);
5761 	if (err)
5762 		goto err_setup_rx;
5763 
5764 	ixgbe_configure(adapter);
5765 
5766 	err = ixgbe_request_irq(adapter);
5767 	if (err)
5768 		goto err_req_irq;
5769 
5770 	/* Notify the stack of the actual queue counts. */
5771 	if (adapter->num_rx_pools > 1)
5772 		queues = adapter->num_rx_queues_per_pool;
5773 	else
5774 		queues = adapter->num_tx_queues;
5775 
5776 	err = netif_set_real_num_tx_queues(netdev, queues);
5777 	if (err)
5778 		goto err_set_queues;
5779 
5780 	if (adapter->num_rx_pools > 1 &&
5781 	    adapter->num_rx_queues > IXGBE_MAX_L2A_QUEUES)
5782 		queues = IXGBE_MAX_L2A_QUEUES;
5783 	else
5784 		queues = adapter->num_rx_queues;
5785 	err = netif_set_real_num_rx_queues(netdev, queues);
5786 	if (err)
5787 		goto err_set_queues;
5788 
5789 	ixgbe_ptp_init(adapter);
5790 
5791 	ixgbe_up_complete(adapter);
5792 
5793 	ixgbe_clear_vxlan_port(adapter);
5794 #ifdef CONFIG_IXGBE_VXLAN
5795 	vxlan_get_rx_port(netdev);
5796 #endif
5797 
5798 	return 0;
5799 
5800 err_set_queues:
5801 	ixgbe_free_irq(adapter);
5802 err_req_irq:
5803 	ixgbe_free_all_rx_resources(adapter);
5804 	if (hw->phy.ops.set_phy_power && !adapter->wol)
5805 		hw->phy.ops.set_phy_power(&adapter->hw, false);
5806 err_setup_rx:
5807 	ixgbe_free_all_tx_resources(adapter);
5808 err_setup_tx:
5809 	ixgbe_reset(adapter);
5810 
5811 	return err;
5812 }
5813 
5814 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
5815 {
5816 	ixgbe_ptp_suspend(adapter);
5817 
5818 	if (adapter->hw.phy.ops.enter_lplu) {
5819 		adapter->hw.phy.reset_disable = true;
5820 		ixgbe_down(adapter);
5821 		adapter->hw.phy.ops.enter_lplu(&adapter->hw);
5822 		adapter->hw.phy.reset_disable = false;
5823 	} else {
5824 		ixgbe_down(adapter);
5825 	}
5826 
5827 	ixgbe_free_irq(adapter);
5828 
5829 	ixgbe_free_all_tx_resources(adapter);
5830 	ixgbe_free_all_rx_resources(adapter);
5831 }
5832 
5833 /**
5834  * ixgbe_close - Disables a network interface
5835  * @netdev: network interface device structure
5836  *
5837  * Returns 0, this is not allowed to fail
5838  *
5839  * The close entry point is called when an interface is de-activated
5840  * by the OS.  The hardware is still under the drivers control, but
5841  * needs to be disabled.  A global MAC reset is issued to stop the
5842  * hardware, and all transmit and receive resources are freed.
5843  **/
5844 static int ixgbe_close(struct net_device *netdev)
5845 {
5846 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5847 
5848 	ixgbe_ptp_stop(adapter);
5849 
5850 	ixgbe_close_suspend(adapter);
5851 
5852 	ixgbe_fdir_filter_exit(adapter);
5853 
5854 	ixgbe_release_hw_control(adapter);
5855 
5856 	return 0;
5857 }
5858 
5859 #ifdef CONFIG_PM
5860 static int ixgbe_resume(struct pci_dev *pdev)
5861 {
5862 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
5863 	struct net_device *netdev = adapter->netdev;
5864 	u32 err;
5865 
5866 	adapter->hw.hw_addr = adapter->io_addr;
5867 	pci_set_power_state(pdev, PCI_D0);
5868 	pci_restore_state(pdev);
5869 	/*
5870 	 * pci_restore_state clears dev->state_saved so call
5871 	 * pci_save_state to restore it.
5872 	 */
5873 	pci_save_state(pdev);
5874 
5875 	err = pci_enable_device_mem(pdev);
5876 	if (err) {
5877 		e_dev_err("Cannot enable PCI device from suspend\n");
5878 		return err;
5879 	}
5880 	smp_mb__before_atomic();
5881 	clear_bit(__IXGBE_DISABLED, &adapter->state);
5882 	pci_set_master(pdev);
5883 
5884 	pci_wake_from_d3(pdev, false);
5885 
5886 	ixgbe_reset(adapter);
5887 
5888 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
5889 
5890 	rtnl_lock();
5891 	err = ixgbe_init_interrupt_scheme(adapter);
5892 	if (!err && netif_running(netdev))
5893 		err = ixgbe_open(netdev);
5894 
5895 	rtnl_unlock();
5896 
5897 	if (err)
5898 		return err;
5899 
5900 	netif_device_attach(netdev);
5901 
5902 	return 0;
5903 }
5904 #endif /* CONFIG_PM */
5905 
5906 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
5907 {
5908 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
5909 	struct net_device *netdev = adapter->netdev;
5910 	struct ixgbe_hw *hw = &adapter->hw;
5911 	u32 ctrl, fctrl;
5912 	u32 wufc = adapter->wol;
5913 #ifdef CONFIG_PM
5914 	int retval = 0;
5915 #endif
5916 
5917 	netif_device_detach(netdev);
5918 
5919 	rtnl_lock();
5920 	if (netif_running(netdev))
5921 		ixgbe_close_suspend(adapter);
5922 	rtnl_unlock();
5923 
5924 	ixgbe_clear_interrupt_scheme(adapter);
5925 
5926 #ifdef CONFIG_PM
5927 	retval = pci_save_state(pdev);
5928 	if (retval)
5929 		return retval;
5930 
5931 #endif
5932 	if (hw->mac.ops.stop_link_on_d3)
5933 		hw->mac.ops.stop_link_on_d3(hw);
5934 
5935 	if (wufc) {
5936 		ixgbe_set_rx_mode(netdev);
5937 
5938 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
5939 		if (hw->mac.ops.enable_tx_laser)
5940 			hw->mac.ops.enable_tx_laser(hw);
5941 
5942 		/* turn on all-multi mode if wake on multicast is enabled */
5943 		if (wufc & IXGBE_WUFC_MC) {
5944 			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
5945 			fctrl |= IXGBE_FCTRL_MPE;
5946 			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
5947 		}
5948 
5949 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
5950 		ctrl |= IXGBE_CTRL_GIO_DIS;
5951 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
5952 
5953 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
5954 	} else {
5955 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
5956 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
5957 	}
5958 
5959 	switch (hw->mac.type) {
5960 	case ixgbe_mac_82598EB:
5961 		pci_wake_from_d3(pdev, false);
5962 		break;
5963 	case ixgbe_mac_82599EB:
5964 	case ixgbe_mac_X540:
5965 	case ixgbe_mac_X550:
5966 	case ixgbe_mac_X550EM_x:
5967 		pci_wake_from_d3(pdev, !!wufc);
5968 		break;
5969 	default:
5970 		break;
5971 	}
5972 
5973 	*enable_wake = !!wufc;
5974 	if (hw->phy.ops.set_phy_power && !*enable_wake)
5975 		hw->phy.ops.set_phy_power(hw, false);
5976 
5977 	ixgbe_release_hw_control(adapter);
5978 
5979 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
5980 		pci_disable_device(pdev);
5981 
5982 	return 0;
5983 }
5984 
5985 #ifdef CONFIG_PM
5986 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
5987 {
5988 	int retval;
5989 	bool wake;
5990 
5991 	retval = __ixgbe_shutdown(pdev, &wake);
5992 	if (retval)
5993 		return retval;
5994 
5995 	if (wake) {
5996 		pci_prepare_to_sleep(pdev);
5997 	} else {
5998 		pci_wake_from_d3(pdev, false);
5999 		pci_set_power_state(pdev, PCI_D3hot);
6000 	}
6001 
6002 	return 0;
6003 }
6004 #endif /* CONFIG_PM */
6005 
6006 static void ixgbe_shutdown(struct pci_dev *pdev)
6007 {
6008 	bool wake;
6009 
6010 	__ixgbe_shutdown(pdev, &wake);
6011 
6012 	if (system_state == SYSTEM_POWER_OFF) {
6013 		pci_wake_from_d3(pdev, wake);
6014 		pci_set_power_state(pdev, PCI_D3hot);
6015 	}
6016 }
6017 
6018 /**
6019  * ixgbe_update_stats - Update the board statistics counters.
6020  * @adapter: board private structure
6021  **/
6022 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
6023 {
6024 	struct net_device *netdev = adapter->netdev;
6025 	struct ixgbe_hw *hw = &adapter->hw;
6026 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
6027 	u64 total_mpc = 0;
6028 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
6029 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
6030 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
6031 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
6032 
6033 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6034 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6035 		return;
6036 
6037 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
6038 		u64 rsc_count = 0;
6039 		u64 rsc_flush = 0;
6040 		for (i = 0; i < adapter->num_rx_queues; i++) {
6041 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
6042 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
6043 		}
6044 		adapter->rsc_total_count = rsc_count;
6045 		adapter->rsc_total_flush = rsc_flush;
6046 	}
6047 
6048 	for (i = 0; i < adapter->num_rx_queues; i++) {
6049 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
6050 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
6051 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
6052 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
6053 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
6054 		bytes += rx_ring->stats.bytes;
6055 		packets += rx_ring->stats.packets;
6056 	}
6057 	adapter->non_eop_descs = non_eop_descs;
6058 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
6059 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
6060 	adapter->hw_csum_rx_error = hw_csum_rx_error;
6061 	netdev->stats.rx_bytes = bytes;
6062 	netdev->stats.rx_packets = packets;
6063 
6064 	bytes = 0;
6065 	packets = 0;
6066 	/* gather some stats to the adapter struct that are per queue */
6067 	for (i = 0; i < adapter->num_tx_queues; i++) {
6068 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6069 		restart_queue += tx_ring->tx_stats.restart_queue;
6070 		tx_busy += tx_ring->tx_stats.tx_busy;
6071 		bytes += tx_ring->stats.bytes;
6072 		packets += tx_ring->stats.packets;
6073 	}
6074 	adapter->restart_queue = restart_queue;
6075 	adapter->tx_busy = tx_busy;
6076 	netdev->stats.tx_bytes = bytes;
6077 	netdev->stats.tx_packets = packets;
6078 
6079 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
6080 
6081 	/* 8 register reads */
6082 	for (i = 0; i < 8; i++) {
6083 		/* for packet buffers not used, the register should read 0 */
6084 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
6085 		missed_rx += mpc;
6086 		hwstats->mpc[i] += mpc;
6087 		total_mpc += hwstats->mpc[i];
6088 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
6089 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
6090 		switch (hw->mac.type) {
6091 		case ixgbe_mac_82598EB:
6092 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
6093 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
6094 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
6095 			hwstats->pxonrxc[i] +=
6096 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
6097 			break;
6098 		case ixgbe_mac_82599EB:
6099 		case ixgbe_mac_X540:
6100 		case ixgbe_mac_X550:
6101 		case ixgbe_mac_X550EM_x:
6102 			hwstats->pxonrxc[i] +=
6103 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
6104 			break;
6105 		default:
6106 			break;
6107 		}
6108 	}
6109 
6110 	/*16 register reads */
6111 	for (i = 0; i < 16; i++) {
6112 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
6113 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
6114 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
6115 		    (hw->mac.type == ixgbe_mac_X540) ||
6116 		    (hw->mac.type == ixgbe_mac_X550) ||
6117 		    (hw->mac.type == ixgbe_mac_X550EM_x)) {
6118 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
6119 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
6120 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
6121 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
6122 		}
6123 	}
6124 
6125 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
6126 	/* work around hardware counting issue */
6127 	hwstats->gprc -= missed_rx;
6128 
6129 	ixgbe_update_xoff_received(adapter);
6130 
6131 	/* 82598 hardware only has a 32 bit counter in the high register */
6132 	switch (hw->mac.type) {
6133 	case ixgbe_mac_82598EB:
6134 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
6135 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
6136 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
6137 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
6138 		break;
6139 	case ixgbe_mac_X540:
6140 	case ixgbe_mac_X550:
6141 	case ixgbe_mac_X550EM_x:
6142 		/* OS2BMC stats are X540 and later */
6143 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
6144 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
6145 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
6146 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
6147 	case ixgbe_mac_82599EB:
6148 		for (i = 0; i < 16; i++)
6149 			adapter->hw_rx_no_dma_resources +=
6150 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
6151 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
6152 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
6153 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
6154 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
6155 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
6156 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
6157 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
6158 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
6159 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
6160 #ifdef IXGBE_FCOE
6161 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
6162 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
6163 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
6164 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
6165 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
6166 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
6167 		/* Add up per cpu counters for total ddp aloc fail */
6168 		if (adapter->fcoe.ddp_pool) {
6169 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
6170 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
6171 			unsigned int cpu;
6172 			u64 noddp = 0, noddp_ext_buff = 0;
6173 			for_each_possible_cpu(cpu) {
6174 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
6175 				noddp += ddp_pool->noddp;
6176 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
6177 			}
6178 			hwstats->fcoe_noddp = noddp;
6179 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
6180 		}
6181 #endif /* IXGBE_FCOE */
6182 		break;
6183 	default:
6184 		break;
6185 	}
6186 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
6187 	hwstats->bprc += bprc;
6188 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
6189 	if (hw->mac.type == ixgbe_mac_82598EB)
6190 		hwstats->mprc -= bprc;
6191 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
6192 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
6193 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
6194 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
6195 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
6196 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
6197 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
6198 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
6199 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
6200 	hwstats->lxontxc += lxon;
6201 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
6202 	hwstats->lxofftxc += lxoff;
6203 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
6204 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
6205 	/*
6206 	 * 82598 errata - tx of flow control packets is included in tx counters
6207 	 */
6208 	xon_off_tot = lxon + lxoff;
6209 	hwstats->gptc -= xon_off_tot;
6210 	hwstats->mptc -= xon_off_tot;
6211 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
6212 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
6213 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
6214 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
6215 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
6216 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
6217 	hwstats->ptc64 -= xon_off_tot;
6218 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
6219 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
6220 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
6221 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
6222 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
6223 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
6224 
6225 	/* Fill out the OS statistics structure */
6226 	netdev->stats.multicast = hwstats->mprc;
6227 
6228 	/* Rx Errors */
6229 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
6230 	netdev->stats.rx_dropped = 0;
6231 	netdev->stats.rx_length_errors = hwstats->rlec;
6232 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
6233 	netdev->stats.rx_missed_errors = total_mpc;
6234 }
6235 
6236 /**
6237  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
6238  * @adapter: pointer to the device adapter structure
6239  **/
6240 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
6241 {
6242 	struct ixgbe_hw *hw = &adapter->hw;
6243 	int i;
6244 
6245 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
6246 		return;
6247 
6248 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6249 
6250 	/* if interface is down do nothing */
6251 	if (test_bit(__IXGBE_DOWN, &adapter->state))
6252 		return;
6253 
6254 	/* do nothing if we are not using signature filters */
6255 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
6256 		return;
6257 
6258 	adapter->fdir_overflow++;
6259 
6260 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
6261 		for (i = 0; i < adapter->num_tx_queues; i++)
6262 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
6263 				&(adapter->tx_ring[i]->state));
6264 		/* re-enable flow director interrupts */
6265 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
6266 	} else {
6267 		e_err(probe, "failed to finish FDIR re-initialization, "
6268 		      "ignored adding FDIR ATR filters\n");
6269 	}
6270 }
6271 
6272 /**
6273  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
6274  * @adapter: pointer to the device adapter structure
6275  *
6276  * This function serves two purposes.  First it strobes the interrupt lines
6277  * in order to make certain interrupts are occurring.  Secondly it sets the
6278  * bits needed to check for TX hangs.  As a result we should immediately
6279  * determine if a hang has occurred.
6280  */
6281 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
6282 {
6283 	struct ixgbe_hw *hw = &adapter->hw;
6284 	u64 eics = 0;
6285 	int i;
6286 
6287 	/* If we're down, removing or resetting, just bail */
6288 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6289 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6290 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6291 		return;
6292 
6293 	/* Force detection of hung controller */
6294 	if (netif_carrier_ok(adapter->netdev)) {
6295 		for (i = 0; i < adapter->num_tx_queues; i++)
6296 			set_check_for_tx_hang(adapter->tx_ring[i]);
6297 	}
6298 
6299 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
6300 		/*
6301 		 * for legacy and MSI interrupts don't set any bits
6302 		 * that are enabled for EIAM, because this operation
6303 		 * would set *both* EIMS and EICS for any bit in EIAM
6304 		 */
6305 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
6306 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
6307 	} else {
6308 		/* get one bit for every active tx/rx interrupt vector */
6309 		for (i = 0; i < adapter->num_q_vectors; i++) {
6310 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
6311 			if (qv->rx.ring || qv->tx.ring)
6312 				eics |= ((u64)1 << i);
6313 		}
6314 	}
6315 
6316 	/* Cause software interrupt to ensure rings are cleaned */
6317 	ixgbe_irq_rearm_queues(adapter, eics);
6318 }
6319 
6320 /**
6321  * ixgbe_watchdog_update_link - update the link status
6322  * @adapter: pointer to the device adapter structure
6323  * @link_speed: pointer to a u32 to store the link_speed
6324  **/
6325 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
6326 {
6327 	struct ixgbe_hw *hw = &adapter->hw;
6328 	u32 link_speed = adapter->link_speed;
6329 	bool link_up = adapter->link_up;
6330 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
6331 
6332 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
6333 		return;
6334 
6335 	if (hw->mac.ops.check_link) {
6336 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
6337 	} else {
6338 		/* always assume link is up, if no check link function */
6339 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
6340 		link_up = true;
6341 	}
6342 
6343 	if (adapter->ixgbe_ieee_pfc)
6344 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
6345 
6346 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
6347 		hw->mac.ops.fc_enable(hw);
6348 		ixgbe_set_rx_drop_en(adapter);
6349 	}
6350 
6351 	if (link_up ||
6352 	    time_after(jiffies, (adapter->link_check_timeout +
6353 				 IXGBE_TRY_LINK_TIMEOUT))) {
6354 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6355 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
6356 		IXGBE_WRITE_FLUSH(hw);
6357 	}
6358 
6359 	adapter->link_up = link_up;
6360 	adapter->link_speed = link_speed;
6361 }
6362 
6363 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
6364 {
6365 #ifdef CONFIG_IXGBE_DCB
6366 	struct net_device *netdev = adapter->netdev;
6367 	struct dcb_app app = {
6368 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
6369 			      .protocol = 0,
6370 			     };
6371 	u8 up = 0;
6372 
6373 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
6374 		up = dcb_ieee_getapp_mask(netdev, &app);
6375 
6376 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
6377 #endif
6378 }
6379 
6380 /**
6381  * ixgbe_watchdog_link_is_up - update netif_carrier status and
6382  *                             print link up message
6383  * @adapter: pointer to the device adapter structure
6384  **/
6385 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
6386 {
6387 	struct net_device *netdev = adapter->netdev;
6388 	struct ixgbe_hw *hw = &adapter->hw;
6389 	struct net_device *upper;
6390 	struct list_head *iter;
6391 	u32 link_speed = adapter->link_speed;
6392 	const char *speed_str;
6393 	bool flow_rx, flow_tx;
6394 
6395 	/* only continue if link was previously down */
6396 	if (netif_carrier_ok(netdev))
6397 		return;
6398 
6399 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
6400 
6401 	switch (hw->mac.type) {
6402 	case ixgbe_mac_82598EB: {
6403 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6404 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
6405 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
6406 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
6407 	}
6408 		break;
6409 	case ixgbe_mac_X540:
6410 	case ixgbe_mac_X550:
6411 	case ixgbe_mac_X550EM_x:
6412 	case ixgbe_mac_82599EB: {
6413 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
6414 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
6415 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
6416 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
6417 	}
6418 		break;
6419 	default:
6420 		flow_tx = false;
6421 		flow_rx = false;
6422 		break;
6423 	}
6424 
6425 	adapter->last_rx_ptp_check = jiffies;
6426 
6427 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6428 		ixgbe_ptp_start_cyclecounter(adapter);
6429 
6430 	switch (link_speed) {
6431 	case IXGBE_LINK_SPEED_10GB_FULL:
6432 		speed_str = "10 Gbps";
6433 		break;
6434 	case IXGBE_LINK_SPEED_2_5GB_FULL:
6435 		speed_str = "2.5 Gbps";
6436 		break;
6437 	case IXGBE_LINK_SPEED_1GB_FULL:
6438 		speed_str = "1 Gbps";
6439 		break;
6440 	case IXGBE_LINK_SPEED_100_FULL:
6441 		speed_str = "100 Mbps";
6442 		break;
6443 	default:
6444 		speed_str = "unknown speed";
6445 		break;
6446 	}
6447 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
6448 	       ((flow_rx && flow_tx) ? "RX/TX" :
6449 	       (flow_rx ? "RX" :
6450 	       (flow_tx ? "TX" : "None"))));
6451 
6452 	netif_carrier_on(netdev);
6453 	ixgbe_check_vf_rate_limit(adapter);
6454 
6455 	/* enable transmits */
6456 	netif_tx_wake_all_queues(adapter->netdev);
6457 
6458 	/* enable any upper devices */
6459 	rtnl_lock();
6460 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
6461 		if (netif_is_macvlan(upper)) {
6462 			struct macvlan_dev *vlan = netdev_priv(upper);
6463 
6464 			if (vlan->fwd_priv)
6465 				netif_tx_wake_all_queues(upper);
6466 		}
6467 	}
6468 	rtnl_unlock();
6469 
6470 	/* update the default user priority for VFs */
6471 	ixgbe_update_default_up(adapter);
6472 
6473 	/* ping all the active vfs to let them know link has changed */
6474 	ixgbe_ping_all_vfs(adapter);
6475 }
6476 
6477 /**
6478  * ixgbe_watchdog_link_is_down - update netif_carrier status and
6479  *                               print link down message
6480  * @adapter: pointer to the adapter structure
6481  **/
6482 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
6483 {
6484 	struct net_device *netdev = adapter->netdev;
6485 	struct ixgbe_hw *hw = &adapter->hw;
6486 
6487 	adapter->link_up = false;
6488 	adapter->link_speed = 0;
6489 
6490 	/* only continue if link was up previously */
6491 	if (!netif_carrier_ok(netdev))
6492 		return;
6493 
6494 	/* poll for SFP+ cable when link is down */
6495 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
6496 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
6497 
6498 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6499 		ixgbe_ptp_start_cyclecounter(adapter);
6500 
6501 	e_info(drv, "NIC Link is Down\n");
6502 	netif_carrier_off(netdev);
6503 
6504 	/* ping all the active vfs to let them know link has changed */
6505 	ixgbe_ping_all_vfs(adapter);
6506 }
6507 
6508 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
6509 {
6510 	int i;
6511 
6512 	for (i = 0; i < adapter->num_tx_queues; i++) {
6513 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6514 
6515 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
6516 			return true;
6517 	}
6518 
6519 	return false;
6520 }
6521 
6522 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
6523 {
6524 	struct ixgbe_hw *hw = &adapter->hw;
6525 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
6526 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
6527 
6528 	int i, j;
6529 
6530 	if (!adapter->num_vfs)
6531 		return false;
6532 
6533 	/* resetting the PF is only needed for MAC before X550 */
6534 	if (hw->mac.type >= ixgbe_mac_X550)
6535 		return false;
6536 
6537 	for (i = 0; i < adapter->num_vfs; i++) {
6538 		for (j = 0; j < q_per_pool; j++) {
6539 			u32 h, t;
6540 
6541 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
6542 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
6543 
6544 			if (h != t)
6545 				return true;
6546 		}
6547 	}
6548 
6549 	return false;
6550 }
6551 
6552 /**
6553  * ixgbe_watchdog_flush_tx - flush queues on link down
6554  * @adapter: pointer to the device adapter structure
6555  **/
6556 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
6557 {
6558 	if (!netif_carrier_ok(adapter->netdev)) {
6559 		if (ixgbe_ring_tx_pending(adapter) ||
6560 		    ixgbe_vf_tx_pending(adapter)) {
6561 			/* We've lost link, so the controller stops DMA,
6562 			 * but we've got queued Tx work that's never going
6563 			 * to get done, so reset controller to flush Tx.
6564 			 * (Do the reset outside of interrupt context).
6565 			 */
6566 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
6567 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
6568 		}
6569 	}
6570 }
6571 
6572 #ifdef CONFIG_PCI_IOV
6573 static inline void ixgbe_issue_vf_flr(struct ixgbe_adapter *adapter,
6574 				      struct pci_dev *vfdev)
6575 {
6576 	if (!pci_wait_for_pending_transaction(vfdev))
6577 		e_dev_warn("Issuing VFLR with pending transactions\n");
6578 
6579 	e_dev_err("Issuing VFLR for VF %s\n", pci_name(vfdev));
6580 	pcie_capability_set_word(vfdev, PCI_EXP_DEVCTL, PCI_EXP_DEVCTL_BCR_FLR);
6581 
6582 	msleep(100);
6583 }
6584 
6585 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
6586 {
6587 	struct ixgbe_hw *hw = &adapter->hw;
6588 	struct pci_dev *pdev = adapter->pdev;
6589 	struct pci_dev *vfdev;
6590 	u32 gpc;
6591 	int pos;
6592 	unsigned short vf_id;
6593 
6594 	if (!(netif_carrier_ok(adapter->netdev)))
6595 		return;
6596 
6597 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
6598 	if (gpc) /* If incrementing then no need for the check below */
6599 		return;
6600 	/* Check to see if a bad DMA write target from an errant or
6601 	 * malicious VF has caused a PCIe error.  If so then we can
6602 	 * issue a VFLR to the offending VF(s) and then resume without
6603 	 * requesting a full slot reset.
6604 	 */
6605 
6606 	if (!pdev)
6607 		return;
6608 
6609 	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_SRIOV);
6610 	if (!pos)
6611 		return;
6612 
6613 	/* get the device ID for the VF */
6614 	pci_read_config_word(pdev, pos + PCI_SRIOV_VF_DID, &vf_id);
6615 
6616 	/* check status reg for all VFs owned by this PF */
6617 	vfdev = pci_get_device(pdev->vendor, vf_id, NULL);
6618 	while (vfdev) {
6619 		if (vfdev->is_virtfn && (vfdev->physfn == pdev)) {
6620 			u16 status_reg;
6621 
6622 			pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
6623 			if (status_reg & PCI_STATUS_REC_MASTER_ABORT)
6624 				/* issue VFLR */
6625 				ixgbe_issue_vf_flr(adapter, vfdev);
6626 		}
6627 
6628 		vfdev = pci_get_device(pdev->vendor, vf_id, vfdev);
6629 	}
6630 }
6631 
6632 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
6633 {
6634 	u32 ssvpc;
6635 
6636 	/* Do not perform spoof check for 82598 or if not in IOV mode */
6637 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
6638 	    adapter->num_vfs == 0)
6639 		return;
6640 
6641 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
6642 
6643 	/*
6644 	 * ssvpc register is cleared on read, if zero then no
6645 	 * spoofed packets in the last interval.
6646 	 */
6647 	if (!ssvpc)
6648 		return;
6649 
6650 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
6651 }
6652 #else
6653 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
6654 {
6655 }
6656 
6657 static void
6658 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
6659 {
6660 }
6661 #endif /* CONFIG_PCI_IOV */
6662 
6663 
6664 /**
6665  * ixgbe_watchdog_subtask - check and bring link up
6666  * @adapter: pointer to the device adapter structure
6667  **/
6668 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
6669 {
6670 	/* if interface is down, removing or resetting, do nothing */
6671 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6672 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6673 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6674 		return;
6675 
6676 	ixgbe_watchdog_update_link(adapter);
6677 
6678 	if (adapter->link_up)
6679 		ixgbe_watchdog_link_is_up(adapter);
6680 	else
6681 		ixgbe_watchdog_link_is_down(adapter);
6682 
6683 	ixgbe_check_for_bad_vf(adapter);
6684 	ixgbe_spoof_check(adapter);
6685 	ixgbe_update_stats(adapter);
6686 
6687 	ixgbe_watchdog_flush_tx(adapter);
6688 }
6689 
6690 /**
6691  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
6692  * @adapter: the ixgbe adapter structure
6693  **/
6694 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
6695 {
6696 	struct ixgbe_hw *hw = &adapter->hw;
6697 	s32 err;
6698 
6699 	/* not searching for SFP so there is nothing to do here */
6700 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
6701 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
6702 		return;
6703 
6704 	/* someone else is in init, wait until next service event */
6705 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
6706 		return;
6707 
6708 	err = hw->phy.ops.identify_sfp(hw);
6709 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
6710 		goto sfp_out;
6711 
6712 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
6713 		/* If no cable is present, then we need to reset
6714 		 * the next time we find a good cable. */
6715 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
6716 	}
6717 
6718 	/* exit on error */
6719 	if (err)
6720 		goto sfp_out;
6721 
6722 	/* exit if reset not needed */
6723 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
6724 		goto sfp_out;
6725 
6726 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
6727 
6728 	/*
6729 	 * A module may be identified correctly, but the EEPROM may not have
6730 	 * support for that module.  setup_sfp() will fail in that case, so
6731 	 * we should not allow that module to load.
6732 	 */
6733 	if (hw->mac.type == ixgbe_mac_82598EB)
6734 		err = hw->phy.ops.reset(hw);
6735 	else
6736 		err = hw->mac.ops.setup_sfp(hw);
6737 
6738 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
6739 		goto sfp_out;
6740 
6741 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
6742 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
6743 
6744 sfp_out:
6745 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
6746 
6747 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
6748 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
6749 		e_dev_err("failed to initialize because an unsupported "
6750 			  "SFP+ module type was detected.\n");
6751 		e_dev_err("Reload the driver after installing a "
6752 			  "supported module.\n");
6753 		unregister_netdev(adapter->netdev);
6754 	}
6755 }
6756 
6757 /**
6758  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
6759  * @adapter: the ixgbe adapter structure
6760  **/
6761 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
6762 {
6763 	struct ixgbe_hw *hw = &adapter->hw;
6764 	u32 speed;
6765 	bool autoneg = false;
6766 
6767 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
6768 		return;
6769 
6770 	/* someone else is in init, wait until next service event */
6771 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
6772 		return;
6773 
6774 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
6775 
6776 	speed = hw->phy.autoneg_advertised;
6777 	if ((!speed) && (hw->mac.ops.get_link_capabilities)) {
6778 		hw->mac.ops.get_link_capabilities(hw, &speed, &autoneg);
6779 
6780 		/* setup the highest link when no autoneg */
6781 		if (!autoneg) {
6782 			if (speed & IXGBE_LINK_SPEED_10GB_FULL)
6783 				speed = IXGBE_LINK_SPEED_10GB_FULL;
6784 		}
6785 	}
6786 
6787 	if (hw->mac.ops.setup_link)
6788 		hw->mac.ops.setup_link(hw, speed, true);
6789 
6790 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
6791 	adapter->link_check_timeout = jiffies;
6792 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
6793 }
6794 
6795 /**
6796  * ixgbe_service_timer - Timer Call-back
6797  * @data: pointer to adapter cast into an unsigned long
6798  **/
6799 static void ixgbe_service_timer(unsigned long data)
6800 {
6801 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
6802 	unsigned long next_event_offset;
6803 
6804 	/* poll faster when waiting for link */
6805 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
6806 		next_event_offset = HZ / 10;
6807 	else
6808 		next_event_offset = HZ * 2;
6809 
6810 	/* Reset the timer */
6811 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
6812 
6813 	ixgbe_service_event_schedule(adapter);
6814 }
6815 
6816 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
6817 {
6818 	struct ixgbe_hw *hw = &adapter->hw;
6819 	u32 status;
6820 
6821 	if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
6822 		return;
6823 
6824 	adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
6825 
6826 	if (!hw->phy.ops.handle_lasi)
6827 		return;
6828 
6829 	status = hw->phy.ops.handle_lasi(&adapter->hw);
6830 	if (status != IXGBE_ERR_OVERTEMP)
6831 		return;
6832 
6833 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
6834 }
6835 
6836 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
6837 {
6838 	if (!(adapter->flags2 & IXGBE_FLAG2_RESET_REQUESTED))
6839 		return;
6840 
6841 	adapter->flags2 &= ~IXGBE_FLAG2_RESET_REQUESTED;
6842 
6843 	/* If we're already down, removing or resetting, just bail */
6844 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6845 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6846 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6847 		return;
6848 
6849 	ixgbe_dump(adapter);
6850 	netdev_err(adapter->netdev, "Reset adapter\n");
6851 	adapter->tx_timeout_count++;
6852 
6853 	rtnl_lock();
6854 	ixgbe_reinit_locked(adapter);
6855 	rtnl_unlock();
6856 }
6857 
6858 /**
6859  * ixgbe_service_task - manages and runs subtasks
6860  * @work: pointer to work_struct containing our data
6861  **/
6862 static void ixgbe_service_task(struct work_struct *work)
6863 {
6864 	struct ixgbe_adapter *adapter = container_of(work,
6865 						     struct ixgbe_adapter,
6866 						     service_task);
6867 	if (ixgbe_removed(adapter->hw.hw_addr)) {
6868 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
6869 			rtnl_lock();
6870 			ixgbe_down(adapter);
6871 			rtnl_unlock();
6872 		}
6873 		ixgbe_service_event_complete(adapter);
6874 		return;
6875 	}
6876 #ifdef CONFIG_IXGBE_VXLAN
6877 	if (adapter->flags2 & IXGBE_FLAG2_VXLAN_REREG_NEEDED) {
6878 		adapter->flags2 &= ~IXGBE_FLAG2_VXLAN_REREG_NEEDED;
6879 		vxlan_get_rx_port(adapter->netdev);
6880 	}
6881 #endif /* CONFIG_IXGBE_VXLAN */
6882 	ixgbe_reset_subtask(adapter);
6883 	ixgbe_phy_interrupt_subtask(adapter);
6884 	ixgbe_sfp_detection_subtask(adapter);
6885 	ixgbe_sfp_link_config_subtask(adapter);
6886 	ixgbe_check_overtemp_subtask(adapter);
6887 	ixgbe_watchdog_subtask(adapter);
6888 	ixgbe_fdir_reinit_subtask(adapter);
6889 	ixgbe_check_hang_subtask(adapter);
6890 
6891 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
6892 		ixgbe_ptp_overflow_check(adapter);
6893 		ixgbe_ptp_rx_hang(adapter);
6894 	}
6895 
6896 	ixgbe_service_event_complete(adapter);
6897 }
6898 
6899 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
6900 		     struct ixgbe_tx_buffer *first,
6901 		     u8 *hdr_len)
6902 {
6903 	struct sk_buff *skb = first->skb;
6904 	u32 vlan_macip_lens, type_tucmd;
6905 	u32 mss_l4len_idx, l4len;
6906 	int err;
6907 
6908 	if (skb->ip_summed != CHECKSUM_PARTIAL)
6909 		return 0;
6910 
6911 	if (!skb_is_gso(skb))
6912 		return 0;
6913 
6914 	err = skb_cow_head(skb, 0);
6915 	if (err < 0)
6916 		return err;
6917 
6918 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
6919 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
6920 
6921 	if (first->protocol == htons(ETH_P_IP)) {
6922 		struct iphdr *iph = ip_hdr(skb);
6923 		iph->tot_len = 0;
6924 		iph->check = 0;
6925 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
6926 							 iph->daddr, 0,
6927 							 IPPROTO_TCP,
6928 							 0);
6929 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
6930 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
6931 				   IXGBE_TX_FLAGS_CSUM |
6932 				   IXGBE_TX_FLAGS_IPV4;
6933 	} else if (skb_is_gso_v6(skb)) {
6934 		ipv6_hdr(skb)->payload_len = 0;
6935 		tcp_hdr(skb)->check =
6936 		    ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
6937 				     &ipv6_hdr(skb)->daddr,
6938 				     0, IPPROTO_TCP, 0);
6939 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
6940 				   IXGBE_TX_FLAGS_CSUM;
6941 	}
6942 
6943 	/* compute header lengths */
6944 	l4len = tcp_hdrlen(skb);
6945 	*hdr_len = skb_transport_offset(skb) + l4len;
6946 
6947 	/* update gso size and bytecount with header size */
6948 	first->gso_segs = skb_shinfo(skb)->gso_segs;
6949 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
6950 
6951 	/* mss_l4len_id: use 0 as index for TSO */
6952 	mss_l4len_idx = l4len << IXGBE_ADVTXD_L4LEN_SHIFT;
6953 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
6954 
6955 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
6956 	vlan_macip_lens = skb_network_header_len(skb);
6957 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
6958 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
6959 
6960 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
6961 			  mss_l4len_idx);
6962 
6963 	return 1;
6964 }
6965 
6966 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
6967 			  struct ixgbe_tx_buffer *first)
6968 {
6969 	struct sk_buff *skb = first->skb;
6970 	u32 vlan_macip_lens = 0;
6971 	u32 mss_l4len_idx = 0;
6972 	u32 type_tucmd = 0;
6973 
6974 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
6975 		if (!(first->tx_flags & IXGBE_TX_FLAGS_HW_VLAN) &&
6976 		    !(first->tx_flags & IXGBE_TX_FLAGS_CC))
6977 			return;
6978 		vlan_macip_lens = skb_network_offset(skb) <<
6979 				  IXGBE_ADVTXD_MACLEN_SHIFT;
6980 	} else {
6981 		u8 l4_hdr = 0;
6982 		union {
6983 			struct iphdr *ipv4;
6984 			struct ipv6hdr *ipv6;
6985 			u8 *raw;
6986 		} network_hdr;
6987 		union {
6988 			struct tcphdr *tcphdr;
6989 			u8 *raw;
6990 		} transport_hdr;
6991 
6992 		if (skb->encapsulation) {
6993 			network_hdr.raw = skb_inner_network_header(skb);
6994 			transport_hdr.raw = skb_inner_transport_header(skb);
6995 			vlan_macip_lens = skb_inner_network_offset(skb) <<
6996 					  IXGBE_ADVTXD_MACLEN_SHIFT;
6997 		} else {
6998 			network_hdr.raw = skb_network_header(skb);
6999 			transport_hdr.raw = skb_transport_header(skb);
7000 			vlan_macip_lens = skb_network_offset(skb) <<
7001 					  IXGBE_ADVTXD_MACLEN_SHIFT;
7002 		}
7003 
7004 		/* use first 4 bits to determine IP version */
7005 		switch (network_hdr.ipv4->version) {
7006 		case IPVERSION:
7007 			vlan_macip_lens |= transport_hdr.raw - network_hdr.raw;
7008 			type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
7009 			l4_hdr = network_hdr.ipv4->protocol;
7010 			break;
7011 		case 6:
7012 			vlan_macip_lens |= transport_hdr.raw - network_hdr.raw;
7013 			l4_hdr = network_hdr.ipv6->nexthdr;
7014 			break;
7015 		default:
7016 			if (unlikely(net_ratelimit())) {
7017 				dev_warn(tx_ring->dev,
7018 					 "partial checksum but version=%d\n",
7019 					 network_hdr.ipv4->version);
7020 			}
7021 		}
7022 
7023 		switch (l4_hdr) {
7024 		case IPPROTO_TCP:
7025 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_TCP;
7026 			mss_l4len_idx = (transport_hdr.tcphdr->doff * 4) <<
7027 					IXGBE_ADVTXD_L4LEN_SHIFT;
7028 			break;
7029 		case IPPROTO_SCTP:
7030 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_SCTP;
7031 			mss_l4len_idx = sizeof(struct sctphdr) <<
7032 					IXGBE_ADVTXD_L4LEN_SHIFT;
7033 			break;
7034 		case IPPROTO_UDP:
7035 			mss_l4len_idx = sizeof(struct udphdr) <<
7036 					IXGBE_ADVTXD_L4LEN_SHIFT;
7037 			break;
7038 		default:
7039 			if (unlikely(net_ratelimit())) {
7040 				dev_warn(tx_ring->dev,
7041 				 "partial checksum but l4 proto=%x!\n",
7042 				 l4_hdr);
7043 			}
7044 			break;
7045 		}
7046 
7047 		/* update TX checksum flag */
7048 		first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
7049 	}
7050 
7051 	/* vlan_macip_lens: MACLEN, VLAN tag */
7052 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7053 
7054 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0,
7055 			  type_tucmd, mss_l4len_idx);
7056 }
7057 
7058 #define IXGBE_SET_FLAG(_input, _flag, _result) \
7059 	((_flag <= _result) ? \
7060 	 ((u32)(_input & _flag) * (_result / _flag)) : \
7061 	 ((u32)(_input & _flag) / (_flag / _result)))
7062 
7063 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
7064 {
7065 	/* set type for advanced descriptor with frame checksum insertion */
7066 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
7067 		       IXGBE_ADVTXD_DCMD_DEXT |
7068 		       IXGBE_ADVTXD_DCMD_IFCS;
7069 
7070 	/* set HW vlan bit if vlan is present */
7071 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
7072 				   IXGBE_ADVTXD_DCMD_VLE);
7073 
7074 	/* set segmentation enable bits for TSO/FSO */
7075 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
7076 				   IXGBE_ADVTXD_DCMD_TSE);
7077 
7078 	/* set timestamp bit if present */
7079 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
7080 				   IXGBE_ADVTXD_MAC_TSTAMP);
7081 
7082 	/* insert frame checksum */
7083 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
7084 
7085 	return cmd_type;
7086 }
7087 
7088 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
7089 				   u32 tx_flags, unsigned int paylen)
7090 {
7091 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
7092 
7093 	/* enable L4 checksum for TSO and TX checksum offload */
7094 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7095 					IXGBE_TX_FLAGS_CSUM,
7096 					IXGBE_ADVTXD_POPTS_TXSM);
7097 
7098 	/* enble IPv4 checksum for TSO */
7099 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7100 					IXGBE_TX_FLAGS_IPV4,
7101 					IXGBE_ADVTXD_POPTS_IXSM);
7102 
7103 	/*
7104 	 * Check Context must be set if Tx switch is enabled, which it
7105 	 * always is for case where virtual functions are running
7106 	 */
7107 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7108 					IXGBE_TX_FLAGS_CC,
7109 					IXGBE_ADVTXD_CC);
7110 
7111 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
7112 }
7113 
7114 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7115 {
7116 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
7117 
7118 	/* Herbert's original patch had:
7119 	 *  smp_mb__after_netif_stop_queue();
7120 	 * but since that doesn't exist yet, just open code it.
7121 	 */
7122 	smp_mb();
7123 
7124 	/* We need to check again in a case another CPU has just
7125 	 * made room available.
7126 	 */
7127 	if (likely(ixgbe_desc_unused(tx_ring) < size))
7128 		return -EBUSY;
7129 
7130 	/* A reprieve! - use start_queue because it doesn't call schedule */
7131 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
7132 	++tx_ring->tx_stats.restart_queue;
7133 	return 0;
7134 }
7135 
7136 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7137 {
7138 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
7139 		return 0;
7140 
7141 	return __ixgbe_maybe_stop_tx(tx_ring, size);
7142 }
7143 
7144 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
7145 		       IXGBE_TXD_CMD_RS)
7146 
7147 static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
7148 			 struct ixgbe_tx_buffer *first,
7149 			 const u8 hdr_len)
7150 {
7151 	struct sk_buff *skb = first->skb;
7152 	struct ixgbe_tx_buffer *tx_buffer;
7153 	union ixgbe_adv_tx_desc *tx_desc;
7154 	struct skb_frag_struct *frag;
7155 	dma_addr_t dma;
7156 	unsigned int data_len, size;
7157 	u32 tx_flags = first->tx_flags;
7158 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
7159 	u16 i = tx_ring->next_to_use;
7160 
7161 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
7162 
7163 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
7164 
7165 	size = skb_headlen(skb);
7166 	data_len = skb->data_len;
7167 
7168 #ifdef IXGBE_FCOE
7169 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
7170 		if (data_len < sizeof(struct fcoe_crc_eof)) {
7171 			size -= sizeof(struct fcoe_crc_eof) - data_len;
7172 			data_len = 0;
7173 		} else {
7174 			data_len -= sizeof(struct fcoe_crc_eof);
7175 		}
7176 	}
7177 
7178 #endif
7179 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
7180 
7181 	tx_buffer = first;
7182 
7183 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
7184 		if (dma_mapping_error(tx_ring->dev, dma))
7185 			goto dma_error;
7186 
7187 		/* record length, and DMA address */
7188 		dma_unmap_len_set(tx_buffer, len, size);
7189 		dma_unmap_addr_set(tx_buffer, dma, dma);
7190 
7191 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
7192 
7193 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
7194 			tx_desc->read.cmd_type_len =
7195 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
7196 
7197 			i++;
7198 			tx_desc++;
7199 			if (i == tx_ring->count) {
7200 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7201 				i = 0;
7202 			}
7203 			tx_desc->read.olinfo_status = 0;
7204 
7205 			dma += IXGBE_MAX_DATA_PER_TXD;
7206 			size -= IXGBE_MAX_DATA_PER_TXD;
7207 
7208 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
7209 		}
7210 
7211 		if (likely(!data_len))
7212 			break;
7213 
7214 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
7215 
7216 		i++;
7217 		tx_desc++;
7218 		if (i == tx_ring->count) {
7219 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7220 			i = 0;
7221 		}
7222 		tx_desc->read.olinfo_status = 0;
7223 
7224 #ifdef IXGBE_FCOE
7225 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
7226 #else
7227 		size = skb_frag_size(frag);
7228 #endif
7229 		data_len -= size;
7230 
7231 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
7232 				       DMA_TO_DEVICE);
7233 
7234 		tx_buffer = &tx_ring->tx_buffer_info[i];
7235 	}
7236 
7237 	/* write last descriptor with RS and EOP bits */
7238 	cmd_type |= size | IXGBE_TXD_CMD;
7239 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
7240 
7241 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
7242 
7243 	/* set the timestamp */
7244 	first->time_stamp = jiffies;
7245 
7246 	/*
7247 	 * Force memory writes to complete before letting h/w know there
7248 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
7249 	 * memory model archs, such as IA-64).
7250 	 *
7251 	 * We also need this memory barrier to make certain all of the
7252 	 * status bits have been updated before next_to_watch is written.
7253 	 */
7254 	wmb();
7255 
7256 	/* set next_to_watch value indicating a packet is present */
7257 	first->next_to_watch = tx_desc;
7258 
7259 	i++;
7260 	if (i == tx_ring->count)
7261 		i = 0;
7262 
7263 	tx_ring->next_to_use = i;
7264 
7265 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
7266 
7267 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
7268 		writel(i, tx_ring->tail);
7269 
7270 		/* we need this if more than one processor can write to our tail
7271 		 * at a time, it synchronizes IO on IA64/Altix systems
7272 		 */
7273 		mmiowb();
7274 	}
7275 
7276 	return;
7277 dma_error:
7278 	dev_err(tx_ring->dev, "TX DMA map failed\n");
7279 
7280 	/* clear dma mappings for failed tx_buffer_info map */
7281 	for (;;) {
7282 		tx_buffer = &tx_ring->tx_buffer_info[i];
7283 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer);
7284 		if (tx_buffer == first)
7285 			break;
7286 		if (i == 0)
7287 			i = tx_ring->count;
7288 		i--;
7289 	}
7290 
7291 	tx_ring->next_to_use = i;
7292 }
7293 
7294 static void ixgbe_atr(struct ixgbe_ring *ring,
7295 		      struct ixgbe_tx_buffer *first)
7296 {
7297 	struct ixgbe_q_vector *q_vector = ring->q_vector;
7298 	union ixgbe_atr_hash_dword input = { .dword = 0 };
7299 	union ixgbe_atr_hash_dword common = { .dword = 0 };
7300 	union {
7301 		unsigned char *network;
7302 		struct iphdr *ipv4;
7303 		struct ipv6hdr *ipv6;
7304 	} hdr;
7305 	struct tcphdr *th;
7306 	struct sk_buff *skb;
7307 #ifdef CONFIG_IXGBE_VXLAN
7308 	u8 encap = false;
7309 #endif /* CONFIG_IXGBE_VXLAN */
7310 	__be16 vlan_id;
7311 
7312 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
7313 	if (!q_vector)
7314 		return;
7315 
7316 	/* do nothing if sampling is disabled */
7317 	if (!ring->atr_sample_rate)
7318 		return;
7319 
7320 	ring->atr_count++;
7321 
7322 	/* snag network header to get L4 type and address */
7323 	skb = first->skb;
7324 	hdr.network = skb_network_header(skb);
7325 	if (skb->encapsulation) {
7326 #ifdef CONFIG_IXGBE_VXLAN
7327 		struct ixgbe_adapter *adapter = q_vector->adapter;
7328 
7329 		if (!adapter->vxlan_port)
7330 			return;
7331 		if (first->protocol != htons(ETH_P_IP) ||
7332 		    hdr.ipv4->version != IPVERSION ||
7333 		    hdr.ipv4->protocol != IPPROTO_UDP) {
7334 			return;
7335 		}
7336 		if (ntohs(udp_hdr(skb)->dest) != adapter->vxlan_port)
7337 			return;
7338 		encap = true;
7339 		hdr.network = skb_inner_network_header(skb);
7340 		th = inner_tcp_hdr(skb);
7341 #else
7342 		return;
7343 #endif /* CONFIG_IXGBE_VXLAN */
7344 	} else {
7345 		/* Currently only IPv4/IPv6 with TCP is supported */
7346 		if ((first->protocol != htons(ETH_P_IPV6) ||
7347 		     hdr.ipv6->nexthdr != IPPROTO_TCP) &&
7348 		    (first->protocol != htons(ETH_P_IP) ||
7349 		     hdr.ipv4->protocol != IPPROTO_TCP))
7350 			return;
7351 		th = tcp_hdr(skb);
7352 	}
7353 
7354 	/* skip this packet since it is invalid or the socket is closing */
7355 	if (!th || th->fin)
7356 		return;
7357 
7358 	/* sample on all syn packets or once every atr sample count */
7359 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
7360 		return;
7361 
7362 	/* reset sample count */
7363 	ring->atr_count = 0;
7364 
7365 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
7366 
7367 	/*
7368 	 * src and dst are inverted, think how the receiver sees them
7369 	 *
7370 	 * The input is broken into two sections, a non-compressed section
7371 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
7372 	 * is XORed together and stored in the compressed dword.
7373 	 */
7374 	input.formatted.vlan_id = vlan_id;
7375 
7376 	/*
7377 	 * since src port and flex bytes occupy the same word XOR them together
7378 	 * and write the value to source port portion of compressed dword
7379 	 */
7380 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
7381 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
7382 	else
7383 		common.port.src ^= th->dest ^ first->protocol;
7384 	common.port.dst ^= th->source;
7385 
7386 	if (first->protocol == htons(ETH_P_IP)) {
7387 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
7388 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
7389 	} else {
7390 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
7391 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
7392 			     hdr.ipv6->saddr.s6_addr32[1] ^
7393 			     hdr.ipv6->saddr.s6_addr32[2] ^
7394 			     hdr.ipv6->saddr.s6_addr32[3] ^
7395 			     hdr.ipv6->daddr.s6_addr32[0] ^
7396 			     hdr.ipv6->daddr.s6_addr32[1] ^
7397 			     hdr.ipv6->daddr.s6_addr32[2] ^
7398 			     hdr.ipv6->daddr.s6_addr32[3];
7399 	}
7400 
7401 #ifdef CONFIG_IXGBE_VXLAN
7402 	if (encap)
7403 		input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
7404 #endif /* CONFIG_IXGBE_VXLAN */
7405 
7406 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
7407 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
7408 					      input, common, ring->queue_index);
7409 }
7410 
7411 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
7412 			      void *accel_priv, select_queue_fallback_t fallback)
7413 {
7414 	struct ixgbe_fwd_adapter *fwd_adapter = accel_priv;
7415 #ifdef IXGBE_FCOE
7416 	struct ixgbe_adapter *adapter;
7417 	struct ixgbe_ring_feature *f;
7418 	int txq;
7419 #endif
7420 
7421 	if (fwd_adapter)
7422 		return skb->queue_mapping + fwd_adapter->tx_base_queue;
7423 
7424 #ifdef IXGBE_FCOE
7425 
7426 	/*
7427 	 * only execute the code below if protocol is FCoE
7428 	 * or FIP and we have FCoE enabled on the adapter
7429 	 */
7430 	switch (vlan_get_protocol(skb)) {
7431 	case htons(ETH_P_FCOE):
7432 	case htons(ETH_P_FIP):
7433 		adapter = netdev_priv(dev);
7434 
7435 		if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
7436 			break;
7437 	default:
7438 		return fallback(dev, skb);
7439 	}
7440 
7441 	f = &adapter->ring_feature[RING_F_FCOE];
7442 
7443 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
7444 					   smp_processor_id();
7445 
7446 	while (txq >= f->indices)
7447 		txq -= f->indices;
7448 
7449 	return txq + f->offset;
7450 #else
7451 	return fallback(dev, skb);
7452 #endif
7453 }
7454 
7455 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
7456 			  struct ixgbe_adapter *adapter,
7457 			  struct ixgbe_ring *tx_ring)
7458 {
7459 	struct ixgbe_tx_buffer *first;
7460 	int tso;
7461 	u32 tx_flags = 0;
7462 	unsigned short f;
7463 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
7464 	__be16 protocol = skb->protocol;
7465 	u8 hdr_len = 0;
7466 
7467 	/*
7468 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
7469 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
7470 	 *       + 2 desc gap to keep tail from touching head,
7471 	 *       + 1 desc for context descriptor,
7472 	 * otherwise try next time
7473 	 */
7474 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
7475 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
7476 
7477 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
7478 		tx_ring->tx_stats.tx_busy++;
7479 		return NETDEV_TX_BUSY;
7480 	}
7481 
7482 	/* record the location of the first descriptor for this packet */
7483 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
7484 	first->skb = skb;
7485 	first->bytecount = skb->len;
7486 	first->gso_segs = 1;
7487 
7488 	/* if we have a HW VLAN tag being added default to the HW one */
7489 	if (skb_vlan_tag_present(skb)) {
7490 		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
7491 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7492 	/* else if it is a SW VLAN check the next protocol and store the tag */
7493 	} else if (protocol == htons(ETH_P_8021Q)) {
7494 		struct vlan_hdr *vhdr, _vhdr;
7495 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
7496 		if (!vhdr)
7497 			goto out_drop;
7498 
7499 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
7500 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
7501 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
7502 	}
7503 	protocol = vlan_get_protocol(skb);
7504 
7505 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
7506 	    adapter->ptp_clock &&
7507 	    !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
7508 				   &adapter->state)) {
7509 		skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
7510 		tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
7511 
7512 		/* schedule check for Tx timestamp */
7513 		adapter->ptp_tx_skb = skb_get(skb);
7514 		adapter->ptp_tx_start = jiffies;
7515 		schedule_work(&adapter->ptp_tx_work);
7516 	}
7517 
7518 	skb_tx_timestamp(skb);
7519 
7520 #ifdef CONFIG_PCI_IOV
7521 	/*
7522 	 * Use the l2switch_enable flag - would be false if the DMA
7523 	 * Tx switch had been disabled.
7524 	 */
7525 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7526 		tx_flags |= IXGBE_TX_FLAGS_CC;
7527 
7528 #endif
7529 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
7530 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
7531 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
7532 	     (skb->priority != TC_PRIO_CONTROL))) {
7533 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
7534 		tx_flags |= (skb->priority & 0x7) <<
7535 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
7536 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
7537 			struct vlan_ethhdr *vhdr;
7538 
7539 			if (skb_cow_head(skb, 0))
7540 				goto out_drop;
7541 			vhdr = (struct vlan_ethhdr *)skb->data;
7542 			vhdr->h_vlan_TCI = htons(tx_flags >>
7543 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
7544 		} else {
7545 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7546 		}
7547 	}
7548 
7549 	/* record initial flags and protocol */
7550 	first->tx_flags = tx_flags;
7551 	first->protocol = protocol;
7552 
7553 #ifdef IXGBE_FCOE
7554 	/* setup tx offload for FCoE */
7555 	if ((protocol == htons(ETH_P_FCOE)) &&
7556 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
7557 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
7558 		if (tso < 0)
7559 			goto out_drop;
7560 
7561 		goto xmit_fcoe;
7562 	}
7563 
7564 #endif /* IXGBE_FCOE */
7565 	tso = ixgbe_tso(tx_ring, first, &hdr_len);
7566 	if (tso < 0)
7567 		goto out_drop;
7568 	else if (!tso)
7569 		ixgbe_tx_csum(tx_ring, first);
7570 
7571 	/* add the ATR filter if ATR is on */
7572 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
7573 		ixgbe_atr(tx_ring, first);
7574 
7575 #ifdef IXGBE_FCOE
7576 xmit_fcoe:
7577 #endif /* IXGBE_FCOE */
7578 	ixgbe_tx_map(tx_ring, first, hdr_len);
7579 
7580 	return NETDEV_TX_OK;
7581 
7582 out_drop:
7583 	dev_kfree_skb_any(first->skb);
7584 	first->skb = NULL;
7585 
7586 	return NETDEV_TX_OK;
7587 }
7588 
7589 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
7590 				      struct net_device *netdev,
7591 				      struct ixgbe_ring *ring)
7592 {
7593 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7594 	struct ixgbe_ring *tx_ring;
7595 
7596 	/*
7597 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
7598 	 * in order to meet this minimum size requirement.
7599 	 */
7600 	if (skb_put_padto(skb, 17))
7601 		return NETDEV_TX_OK;
7602 
7603 	tx_ring = ring ? ring : adapter->tx_ring[skb->queue_mapping];
7604 
7605 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
7606 }
7607 
7608 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
7609 				    struct net_device *netdev)
7610 {
7611 	return __ixgbe_xmit_frame(skb, netdev, NULL);
7612 }
7613 
7614 /**
7615  * ixgbe_set_mac - Change the Ethernet Address of the NIC
7616  * @netdev: network interface device structure
7617  * @p: pointer to an address structure
7618  *
7619  * Returns 0 on success, negative on failure
7620  **/
7621 static int ixgbe_set_mac(struct net_device *netdev, void *p)
7622 {
7623 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7624 	struct ixgbe_hw *hw = &adapter->hw;
7625 	struct sockaddr *addr = p;
7626 	int ret;
7627 
7628 	if (!is_valid_ether_addr(addr->sa_data))
7629 		return -EADDRNOTAVAIL;
7630 
7631 	ixgbe_del_mac_filter(adapter, hw->mac.addr, VMDQ_P(0));
7632 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
7633 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
7634 
7635 	ret = ixgbe_add_mac_filter(adapter, hw->mac.addr, VMDQ_P(0));
7636 	return ret > 0 ? 0 : ret;
7637 }
7638 
7639 static int
7640 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
7641 {
7642 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7643 	struct ixgbe_hw *hw = &adapter->hw;
7644 	u16 value;
7645 	int rc;
7646 
7647 	if (prtad != hw->phy.mdio.prtad)
7648 		return -EINVAL;
7649 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
7650 	if (!rc)
7651 		rc = value;
7652 	return rc;
7653 }
7654 
7655 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
7656 			    u16 addr, u16 value)
7657 {
7658 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7659 	struct ixgbe_hw *hw = &adapter->hw;
7660 
7661 	if (prtad != hw->phy.mdio.prtad)
7662 		return -EINVAL;
7663 	return hw->phy.ops.write_reg(hw, addr, devad, value);
7664 }
7665 
7666 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
7667 {
7668 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7669 
7670 	switch (cmd) {
7671 	case SIOCSHWTSTAMP:
7672 		return ixgbe_ptp_set_ts_config(adapter, req);
7673 	case SIOCGHWTSTAMP:
7674 		return ixgbe_ptp_get_ts_config(adapter, req);
7675 	default:
7676 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
7677 	}
7678 }
7679 
7680 /**
7681  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
7682  * netdev->dev_addrs
7683  * @netdev: network interface device structure
7684  *
7685  * Returns non-zero on failure
7686  **/
7687 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
7688 {
7689 	int err = 0;
7690 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7691 	struct ixgbe_hw *hw = &adapter->hw;
7692 
7693 	if (is_valid_ether_addr(hw->mac.san_addr)) {
7694 		rtnl_lock();
7695 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
7696 		rtnl_unlock();
7697 
7698 		/* update SAN MAC vmdq pool selection */
7699 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
7700 	}
7701 	return err;
7702 }
7703 
7704 /**
7705  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
7706  * netdev->dev_addrs
7707  * @netdev: network interface device structure
7708  *
7709  * Returns non-zero on failure
7710  **/
7711 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
7712 {
7713 	int err = 0;
7714 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7715 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
7716 
7717 	if (is_valid_ether_addr(mac->san_addr)) {
7718 		rtnl_lock();
7719 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
7720 		rtnl_unlock();
7721 	}
7722 	return err;
7723 }
7724 
7725 #ifdef CONFIG_NET_POLL_CONTROLLER
7726 /*
7727  * Polling 'interrupt' - used by things like netconsole to send skbs
7728  * without having to re-enable interrupts. It's not called while
7729  * the interrupt routine is executing.
7730  */
7731 static void ixgbe_netpoll(struct net_device *netdev)
7732 {
7733 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7734 	int i;
7735 
7736 	/* if interface is down do nothing */
7737 	if (test_bit(__IXGBE_DOWN, &adapter->state))
7738 		return;
7739 
7740 	/* loop through and schedule all active queues */
7741 	for (i = 0; i < adapter->num_q_vectors; i++)
7742 		ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
7743 }
7744 
7745 #endif
7746 static struct rtnl_link_stats64 *ixgbe_get_stats64(struct net_device *netdev,
7747 						   struct rtnl_link_stats64 *stats)
7748 {
7749 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7750 	int i;
7751 
7752 	rcu_read_lock();
7753 	for (i = 0; i < adapter->num_rx_queues; i++) {
7754 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
7755 		u64 bytes, packets;
7756 		unsigned int start;
7757 
7758 		if (ring) {
7759 			do {
7760 				start = u64_stats_fetch_begin_irq(&ring->syncp);
7761 				packets = ring->stats.packets;
7762 				bytes   = ring->stats.bytes;
7763 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
7764 			stats->rx_packets += packets;
7765 			stats->rx_bytes   += bytes;
7766 		}
7767 	}
7768 
7769 	for (i = 0; i < adapter->num_tx_queues; i++) {
7770 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
7771 		u64 bytes, packets;
7772 		unsigned int start;
7773 
7774 		if (ring) {
7775 			do {
7776 				start = u64_stats_fetch_begin_irq(&ring->syncp);
7777 				packets = ring->stats.packets;
7778 				bytes   = ring->stats.bytes;
7779 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
7780 			stats->tx_packets += packets;
7781 			stats->tx_bytes   += bytes;
7782 		}
7783 	}
7784 	rcu_read_unlock();
7785 	/* following stats updated by ixgbe_watchdog_task() */
7786 	stats->multicast	= netdev->stats.multicast;
7787 	stats->rx_errors	= netdev->stats.rx_errors;
7788 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
7789 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
7790 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
7791 	return stats;
7792 }
7793 
7794 #ifdef CONFIG_IXGBE_DCB
7795 /**
7796  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
7797  * @adapter: pointer to ixgbe_adapter
7798  * @tc: number of traffic classes currently enabled
7799  *
7800  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
7801  * 802.1Q priority maps to a packet buffer that exists.
7802  */
7803 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
7804 {
7805 	struct ixgbe_hw *hw = &adapter->hw;
7806 	u32 reg, rsave;
7807 	int i;
7808 
7809 	/* 82598 have a static priority to TC mapping that can not
7810 	 * be changed so no validation is needed.
7811 	 */
7812 	if (hw->mac.type == ixgbe_mac_82598EB)
7813 		return;
7814 
7815 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
7816 	rsave = reg;
7817 
7818 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
7819 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
7820 
7821 		/* If up2tc is out of bounds default to zero */
7822 		if (up2tc > tc)
7823 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
7824 	}
7825 
7826 	if (reg != rsave)
7827 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
7828 
7829 	return;
7830 }
7831 
7832 /**
7833  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
7834  * @adapter: Pointer to adapter struct
7835  *
7836  * Populate the netdev user priority to tc map
7837  */
7838 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
7839 {
7840 	struct net_device *dev = adapter->netdev;
7841 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
7842 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
7843 	u8 prio;
7844 
7845 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
7846 		u8 tc = 0;
7847 
7848 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
7849 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
7850 		else if (ets)
7851 			tc = ets->prio_tc[prio];
7852 
7853 		netdev_set_prio_tc_map(dev, prio, tc);
7854 	}
7855 }
7856 
7857 #endif /* CONFIG_IXGBE_DCB */
7858 /**
7859  * ixgbe_setup_tc - configure net_device for multiple traffic classes
7860  *
7861  * @netdev: net device to configure
7862  * @tc: number of traffic classes to enable
7863  */
7864 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
7865 {
7866 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7867 	struct ixgbe_hw *hw = &adapter->hw;
7868 	bool pools;
7869 
7870 	/* Hardware supports up to 8 traffic classes */
7871 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
7872 		return -EINVAL;
7873 
7874 	if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
7875 		return -EINVAL;
7876 
7877 	pools = (find_first_zero_bit(&adapter->fwd_bitmask, 32) > 1);
7878 	if (tc && pools && adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS)
7879 		return -EBUSY;
7880 
7881 	/* Hardware has to reinitialize queues and interrupts to
7882 	 * match packet buffer alignment. Unfortunately, the
7883 	 * hardware is not flexible enough to do this dynamically.
7884 	 */
7885 	if (netif_running(dev))
7886 		ixgbe_close(dev);
7887 	ixgbe_clear_interrupt_scheme(adapter);
7888 
7889 #ifdef CONFIG_IXGBE_DCB
7890 	if (tc) {
7891 		netdev_set_num_tc(dev, tc);
7892 		ixgbe_set_prio_tc_map(adapter);
7893 
7894 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
7895 
7896 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
7897 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
7898 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
7899 		}
7900 	} else {
7901 		netdev_reset_tc(dev);
7902 
7903 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
7904 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
7905 
7906 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
7907 
7908 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
7909 		adapter->dcb_cfg.pfc_mode_enable = false;
7910 	}
7911 
7912 	ixgbe_validate_rtr(adapter, tc);
7913 
7914 #endif /* CONFIG_IXGBE_DCB */
7915 	ixgbe_init_interrupt_scheme(adapter);
7916 
7917 	if (netif_running(dev))
7918 		return ixgbe_open(dev);
7919 
7920 	return 0;
7921 }
7922 
7923 #ifdef CONFIG_PCI_IOV
7924 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
7925 {
7926 	struct net_device *netdev = adapter->netdev;
7927 
7928 	rtnl_lock();
7929 	ixgbe_setup_tc(netdev, netdev_get_num_tc(netdev));
7930 	rtnl_unlock();
7931 }
7932 
7933 #endif
7934 void ixgbe_do_reset(struct net_device *netdev)
7935 {
7936 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7937 
7938 	if (netif_running(netdev))
7939 		ixgbe_reinit_locked(adapter);
7940 	else
7941 		ixgbe_reset(adapter);
7942 }
7943 
7944 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
7945 					    netdev_features_t features)
7946 {
7947 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7948 
7949 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
7950 	if (!(features & NETIF_F_RXCSUM))
7951 		features &= ~NETIF_F_LRO;
7952 
7953 	/* Turn off LRO if not RSC capable */
7954 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
7955 		features &= ~NETIF_F_LRO;
7956 
7957 	return features;
7958 }
7959 
7960 static int ixgbe_set_features(struct net_device *netdev,
7961 			      netdev_features_t features)
7962 {
7963 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7964 	netdev_features_t changed = netdev->features ^ features;
7965 	bool need_reset = false;
7966 
7967 	/* Make sure RSC matches LRO, reset if change */
7968 	if (!(features & NETIF_F_LRO)) {
7969 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
7970 			need_reset = true;
7971 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
7972 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
7973 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
7974 		if (adapter->rx_itr_setting == 1 ||
7975 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
7976 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
7977 			need_reset = true;
7978 		} else if ((changed ^ features) & NETIF_F_LRO) {
7979 			e_info(probe, "rx-usecs set too low, "
7980 			       "disabling RSC\n");
7981 		}
7982 	}
7983 
7984 	/*
7985 	 * Check if Flow Director n-tuple support was enabled or disabled.  If
7986 	 * the state changed, we need to reset.
7987 	 */
7988 	switch (features & NETIF_F_NTUPLE) {
7989 	case NETIF_F_NTUPLE:
7990 		/* turn off ATR, enable perfect filters and reset */
7991 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
7992 			need_reset = true;
7993 
7994 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
7995 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
7996 		break;
7997 	default:
7998 		/* turn off perfect filters, enable ATR and reset */
7999 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
8000 			need_reset = true;
8001 
8002 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
8003 
8004 		/* We cannot enable ATR if SR-IOV is enabled */
8005 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
8006 			break;
8007 
8008 		/* We cannot enable ATR if we have 2 or more traffic classes */
8009 		if (netdev_get_num_tc(netdev) > 1)
8010 			break;
8011 
8012 		/* We cannot enable ATR if RSS is disabled */
8013 		if (adapter->ring_feature[RING_F_RSS].limit <= 1)
8014 			break;
8015 
8016 		/* A sample rate of 0 indicates ATR disabled */
8017 		if (!adapter->atr_sample_rate)
8018 			break;
8019 
8020 		adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
8021 		break;
8022 	}
8023 
8024 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
8025 		ixgbe_vlan_strip_enable(adapter);
8026 	else
8027 		ixgbe_vlan_strip_disable(adapter);
8028 
8029 	if (changed & NETIF_F_RXALL)
8030 		need_reset = true;
8031 
8032 	netdev->features = features;
8033 
8034 #ifdef CONFIG_IXGBE_VXLAN
8035 	if ((adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE)) {
8036 		if (features & NETIF_F_RXCSUM)
8037 			adapter->flags2 |= IXGBE_FLAG2_VXLAN_REREG_NEEDED;
8038 		else
8039 			ixgbe_clear_vxlan_port(adapter);
8040 	}
8041 #endif /* CONFIG_IXGBE_VXLAN */
8042 
8043 	if (need_reset)
8044 		ixgbe_do_reset(netdev);
8045 
8046 	return 0;
8047 }
8048 
8049 #ifdef CONFIG_IXGBE_VXLAN
8050 /**
8051  * ixgbe_add_vxlan_port - Get notifications about VXLAN ports that come up
8052  * @dev: The port's netdev
8053  * @sa_family: Socket Family that VXLAN is notifiying us about
8054  * @port: New UDP port number that VXLAN started listening to
8055  **/
8056 static void ixgbe_add_vxlan_port(struct net_device *dev, sa_family_t sa_family,
8057 				 __be16 port)
8058 {
8059 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8060 	struct ixgbe_hw *hw = &adapter->hw;
8061 	u16 new_port = ntohs(port);
8062 
8063 	if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
8064 		return;
8065 
8066 	if (sa_family == AF_INET6)
8067 		return;
8068 
8069 	if (adapter->vxlan_port == new_port)
8070 		return;
8071 
8072 	if (adapter->vxlan_port) {
8073 		netdev_info(dev,
8074 			    "Hit Max num of VXLAN ports, not adding port %d\n",
8075 			    new_port);
8076 		return;
8077 	}
8078 
8079 	adapter->vxlan_port = new_port;
8080 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, new_port);
8081 }
8082 
8083 /**
8084  * ixgbe_del_vxlan_port - Get notifications about VXLAN ports that go away
8085  * @dev: The port's netdev
8086  * @sa_family: Socket Family that VXLAN is notifying us about
8087  * @port: UDP port number that VXLAN stopped listening to
8088  **/
8089 static void ixgbe_del_vxlan_port(struct net_device *dev, sa_family_t sa_family,
8090 				 __be16 port)
8091 {
8092 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8093 	u16 new_port = ntohs(port);
8094 
8095 	if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
8096 		return;
8097 
8098 	if (sa_family == AF_INET6)
8099 		return;
8100 
8101 	if (adapter->vxlan_port != new_port) {
8102 		netdev_info(dev, "Port %d was not found, not deleting\n",
8103 			    new_port);
8104 		return;
8105 	}
8106 
8107 	ixgbe_clear_vxlan_port(adapter);
8108 	adapter->flags2 |= IXGBE_FLAG2_VXLAN_REREG_NEEDED;
8109 }
8110 #endif /* CONFIG_IXGBE_VXLAN */
8111 
8112 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
8113 			     struct net_device *dev,
8114 			     const unsigned char *addr, u16 vid,
8115 			     u16 flags)
8116 {
8117 	/* guarantee we can provide a unique filter for the unicast address */
8118 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
8119 		if (IXGBE_MAX_PF_MACVLANS <= netdev_uc_count(dev))
8120 			return -ENOMEM;
8121 	}
8122 
8123 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
8124 }
8125 
8126 /**
8127  * ixgbe_configure_bridge_mode - set various bridge modes
8128  * @adapter - the private structure
8129  * @mode - requested bridge mode
8130  *
8131  * Configure some settings require for various bridge modes.
8132  **/
8133 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
8134 				       __u16 mode)
8135 {
8136 	struct ixgbe_hw *hw = &adapter->hw;
8137 	unsigned int p, num_pools;
8138 	u32 vmdctl;
8139 
8140 	switch (mode) {
8141 	case BRIDGE_MODE_VEPA:
8142 		/* disable Tx loopback, rely on switch hairpin mode */
8143 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
8144 
8145 		/* must enable Rx switching replication to allow multicast
8146 		 * packet reception on all VFs, and to enable source address
8147 		 * pruning.
8148 		 */
8149 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
8150 		vmdctl |= IXGBE_VT_CTL_REPLEN;
8151 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
8152 
8153 		/* enable Rx source address pruning. Note, this requires
8154 		 * replication to be enabled or else it does nothing.
8155 		 */
8156 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
8157 		for (p = 0; p < num_pools; p++) {
8158 			if (hw->mac.ops.set_source_address_pruning)
8159 				hw->mac.ops.set_source_address_pruning(hw,
8160 								       true,
8161 								       p);
8162 		}
8163 		break;
8164 	case BRIDGE_MODE_VEB:
8165 		/* enable Tx loopback for internal VF/PF communication */
8166 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
8167 				IXGBE_PFDTXGSWC_VT_LBEN);
8168 
8169 		/* disable Rx switching replication unless we have SR-IOV
8170 		 * virtual functions
8171 		 */
8172 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
8173 		if (!adapter->num_vfs)
8174 			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
8175 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
8176 
8177 		/* disable Rx source address pruning, since we don't expect to
8178 		 * be receiving external loopback of our transmitted frames.
8179 		 */
8180 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
8181 		for (p = 0; p < num_pools; p++) {
8182 			if (hw->mac.ops.set_source_address_pruning)
8183 				hw->mac.ops.set_source_address_pruning(hw,
8184 								       false,
8185 								       p);
8186 		}
8187 		break;
8188 	default:
8189 		return -EINVAL;
8190 	}
8191 
8192 	adapter->bridge_mode = mode;
8193 
8194 	e_info(drv, "enabling bridge mode: %s\n",
8195 	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
8196 
8197 	return 0;
8198 }
8199 
8200 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
8201 				    struct nlmsghdr *nlh, u16 flags)
8202 {
8203 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8204 	struct nlattr *attr, *br_spec;
8205 	int rem;
8206 
8207 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
8208 		return -EOPNOTSUPP;
8209 
8210 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8211 	if (!br_spec)
8212 		return -EINVAL;
8213 
8214 	nla_for_each_nested(attr, br_spec, rem) {
8215 		int status;
8216 		__u16 mode;
8217 
8218 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
8219 			continue;
8220 
8221 		if (nla_len(attr) < sizeof(mode))
8222 			return -EINVAL;
8223 
8224 		mode = nla_get_u16(attr);
8225 		status = ixgbe_configure_bridge_mode(adapter, mode);
8226 		if (status)
8227 			return status;
8228 
8229 		break;
8230 	}
8231 
8232 	return 0;
8233 }
8234 
8235 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
8236 				    struct net_device *dev,
8237 				    u32 filter_mask, int nlflags)
8238 {
8239 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8240 
8241 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
8242 		return 0;
8243 
8244 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
8245 				       adapter->bridge_mode, 0, 0, nlflags,
8246 				       filter_mask, NULL);
8247 }
8248 
8249 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
8250 {
8251 	struct ixgbe_fwd_adapter *fwd_adapter = NULL;
8252 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
8253 	int used_pools = adapter->num_vfs + adapter->num_rx_pools;
8254 	unsigned int limit;
8255 	int pool, err;
8256 
8257 	/* Hardware has a limited number of available pools. Each VF, and the
8258 	 * PF require a pool. Check to ensure we don't attempt to use more
8259 	 * then the available number of pools.
8260 	 */
8261 	if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
8262 		return ERR_PTR(-EINVAL);
8263 
8264 #ifdef CONFIG_RPS
8265 	if (vdev->num_rx_queues != vdev->num_tx_queues) {
8266 		netdev_info(pdev, "%s: Only supports a single queue count for TX and RX\n",
8267 			    vdev->name);
8268 		return ERR_PTR(-EINVAL);
8269 	}
8270 #endif
8271 	/* Check for hardware restriction on number of rx/tx queues */
8272 	if (vdev->num_tx_queues > IXGBE_MAX_L2A_QUEUES ||
8273 	    vdev->num_tx_queues == IXGBE_BAD_L2A_QUEUE) {
8274 		netdev_info(pdev,
8275 			    "%s: Supports RX/TX Queue counts 1,2, and 4\n",
8276 			    pdev->name);
8277 		return ERR_PTR(-EINVAL);
8278 	}
8279 
8280 	if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
8281 	      adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS - 1) ||
8282 	    (adapter->num_rx_pools > IXGBE_MAX_MACVLANS))
8283 		return ERR_PTR(-EBUSY);
8284 
8285 	fwd_adapter = kzalloc(sizeof(*fwd_adapter), GFP_KERNEL);
8286 	if (!fwd_adapter)
8287 		return ERR_PTR(-ENOMEM);
8288 
8289 	pool = find_first_zero_bit(&adapter->fwd_bitmask, 32);
8290 	adapter->num_rx_pools++;
8291 	set_bit(pool, &adapter->fwd_bitmask);
8292 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
8293 
8294 	/* Enable VMDq flag so device will be set in VM mode */
8295 	adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED | IXGBE_FLAG_SRIOV_ENABLED;
8296 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
8297 	adapter->ring_feature[RING_F_RSS].limit = vdev->num_tx_queues;
8298 
8299 	/* Force reinit of ring allocation with VMDQ enabled */
8300 	err = ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
8301 	if (err)
8302 		goto fwd_add_err;
8303 	fwd_adapter->pool = pool;
8304 	fwd_adapter->real_adapter = adapter;
8305 	err = ixgbe_fwd_ring_up(vdev, fwd_adapter);
8306 	if (err)
8307 		goto fwd_add_err;
8308 	netif_tx_start_all_queues(vdev);
8309 	return fwd_adapter;
8310 fwd_add_err:
8311 	/* unwind counter and free adapter struct */
8312 	netdev_info(pdev,
8313 		    "%s: dfwd hardware acceleration failed\n", vdev->name);
8314 	clear_bit(pool, &adapter->fwd_bitmask);
8315 	adapter->num_rx_pools--;
8316 	kfree(fwd_adapter);
8317 	return ERR_PTR(err);
8318 }
8319 
8320 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
8321 {
8322 	struct ixgbe_fwd_adapter *fwd_adapter = priv;
8323 	struct ixgbe_adapter *adapter = fwd_adapter->real_adapter;
8324 	unsigned int limit;
8325 
8326 	clear_bit(fwd_adapter->pool, &adapter->fwd_bitmask);
8327 	adapter->num_rx_pools--;
8328 
8329 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
8330 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
8331 	ixgbe_fwd_ring_down(fwd_adapter->netdev, fwd_adapter);
8332 	ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
8333 	netdev_dbg(pdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
8334 		   fwd_adapter->pool, adapter->num_rx_pools,
8335 		   fwd_adapter->rx_base_queue,
8336 		   fwd_adapter->rx_base_queue + adapter->num_rx_queues_per_pool,
8337 		   adapter->fwd_bitmask);
8338 	kfree(fwd_adapter);
8339 }
8340 
8341 #define IXGBE_MAX_TUNNEL_HDR_LEN 80
8342 static netdev_features_t
8343 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
8344 		     netdev_features_t features)
8345 {
8346 	if (!skb->encapsulation)
8347 		return features;
8348 
8349 	if (unlikely(skb_inner_mac_header(skb) - skb_transport_header(skb) >
8350 		     IXGBE_MAX_TUNNEL_HDR_LEN))
8351 		return features & ~NETIF_F_ALL_CSUM;
8352 
8353 	return features;
8354 }
8355 
8356 static const struct net_device_ops ixgbe_netdev_ops = {
8357 	.ndo_open		= ixgbe_open,
8358 	.ndo_stop		= ixgbe_close,
8359 	.ndo_start_xmit		= ixgbe_xmit_frame,
8360 	.ndo_select_queue	= ixgbe_select_queue,
8361 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
8362 	.ndo_validate_addr	= eth_validate_addr,
8363 	.ndo_set_mac_address	= ixgbe_set_mac,
8364 	.ndo_change_mtu		= ixgbe_change_mtu,
8365 	.ndo_tx_timeout		= ixgbe_tx_timeout,
8366 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
8367 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
8368 	.ndo_do_ioctl		= ixgbe_ioctl,
8369 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
8370 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
8371 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
8372 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
8373 	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
8374 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
8375 	.ndo_get_stats64	= ixgbe_get_stats64,
8376 #ifdef CONFIG_IXGBE_DCB
8377 	.ndo_setup_tc		= ixgbe_setup_tc,
8378 #endif
8379 #ifdef CONFIG_NET_POLL_CONTROLLER
8380 	.ndo_poll_controller	= ixgbe_netpoll,
8381 #endif
8382 #ifdef CONFIG_NET_RX_BUSY_POLL
8383 	.ndo_busy_poll		= ixgbe_low_latency_recv,
8384 #endif
8385 #ifdef IXGBE_FCOE
8386 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
8387 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
8388 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
8389 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
8390 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
8391 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
8392 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
8393 #endif /* IXGBE_FCOE */
8394 	.ndo_set_features = ixgbe_set_features,
8395 	.ndo_fix_features = ixgbe_fix_features,
8396 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
8397 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
8398 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
8399 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
8400 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
8401 #ifdef CONFIG_IXGBE_VXLAN
8402 	.ndo_add_vxlan_port	= ixgbe_add_vxlan_port,
8403 	.ndo_del_vxlan_port	= ixgbe_del_vxlan_port,
8404 #endif /* CONFIG_IXGBE_VXLAN */
8405 	.ndo_features_check	= ixgbe_features_check,
8406 };
8407 
8408 /**
8409  * ixgbe_enumerate_functions - Get the number of ports this device has
8410  * @adapter: adapter structure
8411  *
8412  * This function enumerates the phsyical functions co-located on a single slot,
8413  * in order to determine how many ports a device has. This is most useful in
8414  * determining the required GT/s of PCIe bandwidth necessary for optimal
8415  * performance.
8416  **/
8417 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
8418 {
8419 	struct pci_dev *entry, *pdev = adapter->pdev;
8420 	int physfns = 0;
8421 
8422 	/* Some cards can not use the generic count PCIe functions method,
8423 	 * because they are behind a parent switch, so we hardcode these with
8424 	 * the correct number of functions.
8425 	 */
8426 	if (ixgbe_pcie_from_parent(&adapter->hw))
8427 		physfns = 4;
8428 
8429 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
8430 		/* don't count virtual functions */
8431 		if (entry->is_virtfn)
8432 			continue;
8433 
8434 		/* When the devices on the bus don't all match our device ID,
8435 		 * we can't reliably determine the correct number of
8436 		 * functions. This can occur if a function has been direct
8437 		 * attached to a virtual machine using VT-d, for example. In
8438 		 * this case, simply return -1 to indicate this.
8439 		 */
8440 		if ((entry->vendor != pdev->vendor) ||
8441 		    (entry->device != pdev->device))
8442 			return -1;
8443 
8444 		physfns++;
8445 	}
8446 
8447 	return physfns;
8448 }
8449 
8450 /**
8451  * ixgbe_wol_supported - Check whether device supports WoL
8452  * @hw: hw specific details
8453  * @device_id: the device ID
8454  * @subdev_id: the subsystem device ID
8455  *
8456  * This function is used by probe and ethtool to determine
8457  * which devices have WoL support
8458  *
8459  **/
8460 int ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
8461 			u16 subdevice_id)
8462 {
8463 	struct ixgbe_hw *hw = &adapter->hw;
8464 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
8465 	int is_wol_supported = 0;
8466 
8467 	switch (device_id) {
8468 	case IXGBE_DEV_ID_82599_SFP:
8469 		/* Only these subdevices could supports WOL */
8470 		switch (subdevice_id) {
8471 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
8472 		case IXGBE_SUBDEV_ID_82599_560FLR:
8473 			/* only support first port */
8474 			if (hw->bus.func != 0)
8475 				break;
8476 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
8477 		case IXGBE_SUBDEV_ID_82599_SFP:
8478 		case IXGBE_SUBDEV_ID_82599_RNDC:
8479 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
8480 		case IXGBE_SUBDEV_ID_82599_LOM_SFP:
8481 			is_wol_supported = 1;
8482 			break;
8483 		}
8484 		break;
8485 	case IXGBE_DEV_ID_82599EN_SFP:
8486 		/* Only this subdevice supports WOL */
8487 		switch (subdevice_id) {
8488 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
8489 			is_wol_supported = 1;
8490 			break;
8491 		}
8492 		break;
8493 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
8494 		/* All except this subdevice support WOL */
8495 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
8496 			is_wol_supported = 1;
8497 		break;
8498 	case IXGBE_DEV_ID_82599_KX4:
8499 		is_wol_supported = 1;
8500 		break;
8501 	case IXGBE_DEV_ID_X540T:
8502 	case IXGBE_DEV_ID_X540T1:
8503 	case IXGBE_DEV_ID_X550T:
8504 	case IXGBE_DEV_ID_X550EM_X_KX4:
8505 	case IXGBE_DEV_ID_X550EM_X_KR:
8506 	case IXGBE_DEV_ID_X550EM_X_10G_T:
8507 		/* check eeprom to see if enabled wol */
8508 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
8509 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
8510 		     (hw->bus.func == 0))) {
8511 			is_wol_supported = 1;
8512 		}
8513 		break;
8514 	}
8515 
8516 	return is_wol_supported;
8517 }
8518 
8519 /**
8520  * ixgbe_get_platform_mac_addr - Look up MAC address in Open Firmware / IDPROM
8521  * @adapter: Pointer to adapter struct
8522  */
8523 static void ixgbe_get_platform_mac_addr(struct ixgbe_adapter *adapter)
8524 {
8525 #ifdef CONFIG_OF
8526 	struct device_node *dp = pci_device_to_OF_node(adapter->pdev);
8527 	struct ixgbe_hw *hw = &adapter->hw;
8528 	const unsigned char *addr;
8529 
8530 	addr = of_get_mac_address(dp);
8531 	if (addr) {
8532 		ether_addr_copy(hw->mac.perm_addr, addr);
8533 		return;
8534 	}
8535 #endif /* CONFIG_OF */
8536 
8537 #ifdef CONFIG_SPARC
8538 	ether_addr_copy(hw->mac.perm_addr, idprom->id_ethaddr);
8539 #endif /* CONFIG_SPARC */
8540 }
8541 
8542 /**
8543  * ixgbe_probe - Device Initialization Routine
8544  * @pdev: PCI device information struct
8545  * @ent: entry in ixgbe_pci_tbl
8546  *
8547  * Returns 0 on success, negative on failure
8548  *
8549  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
8550  * The OS initialization, configuring of the adapter private structure,
8551  * and a hardware reset occur.
8552  **/
8553 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
8554 {
8555 	struct net_device *netdev;
8556 	struct ixgbe_adapter *adapter = NULL;
8557 	struct ixgbe_hw *hw;
8558 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
8559 	int i, err, pci_using_dac, expected_gts;
8560 	unsigned int indices = MAX_TX_QUEUES;
8561 	u8 part_str[IXGBE_PBANUM_LENGTH];
8562 	bool disable_dev = false;
8563 #ifdef IXGBE_FCOE
8564 	u16 device_caps;
8565 #endif
8566 	u32 eec;
8567 
8568 	/* Catch broken hardware that put the wrong VF device ID in
8569 	 * the PCIe SR-IOV capability.
8570 	 */
8571 	if (pdev->is_virtfn) {
8572 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
8573 		     pci_name(pdev), pdev->vendor, pdev->device);
8574 		return -EINVAL;
8575 	}
8576 
8577 	err = pci_enable_device_mem(pdev);
8578 	if (err)
8579 		return err;
8580 
8581 	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
8582 		pci_using_dac = 1;
8583 	} else {
8584 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
8585 		if (err) {
8586 			dev_err(&pdev->dev,
8587 				"No usable DMA configuration, aborting\n");
8588 			goto err_dma;
8589 		}
8590 		pci_using_dac = 0;
8591 	}
8592 
8593 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
8594 					   IORESOURCE_MEM), ixgbe_driver_name);
8595 	if (err) {
8596 		dev_err(&pdev->dev,
8597 			"pci_request_selected_regions failed 0x%x\n", err);
8598 		goto err_pci_reg;
8599 	}
8600 
8601 	pci_enable_pcie_error_reporting(pdev);
8602 
8603 	pci_set_master(pdev);
8604 	pci_save_state(pdev);
8605 
8606 	if (ii->mac == ixgbe_mac_82598EB) {
8607 #ifdef CONFIG_IXGBE_DCB
8608 		/* 8 TC w/ 4 queues per TC */
8609 		indices = 4 * MAX_TRAFFIC_CLASS;
8610 #else
8611 		indices = IXGBE_MAX_RSS_INDICES;
8612 #endif
8613 	}
8614 
8615 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
8616 	if (!netdev) {
8617 		err = -ENOMEM;
8618 		goto err_alloc_etherdev;
8619 	}
8620 
8621 	SET_NETDEV_DEV(netdev, &pdev->dev);
8622 
8623 	adapter = netdev_priv(netdev);
8624 
8625 	adapter->netdev = netdev;
8626 	adapter->pdev = pdev;
8627 	hw = &adapter->hw;
8628 	hw->back = adapter;
8629 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
8630 
8631 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
8632 			      pci_resource_len(pdev, 0));
8633 	adapter->io_addr = hw->hw_addr;
8634 	if (!hw->hw_addr) {
8635 		err = -EIO;
8636 		goto err_ioremap;
8637 	}
8638 
8639 	netdev->netdev_ops = &ixgbe_netdev_ops;
8640 	ixgbe_set_ethtool_ops(netdev);
8641 	netdev->watchdog_timeo = 5 * HZ;
8642 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
8643 
8644 	/* Setup hw api */
8645 	memcpy(&hw->mac.ops, ii->mac_ops, sizeof(hw->mac.ops));
8646 	hw->mac.type  = ii->mac;
8647 	hw->mvals     = ii->mvals;
8648 
8649 	/* EEPROM */
8650 	memcpy(&hw->eeprom.ops, ii->eeprom_ops, sizeof(hw->eeprom.ops));
8651 	eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
8652 	if (ixgbe_removed(hw->hw_addr)) {
8653 		err = -EIO;
8654 		goto err_ioremap;
8655 	}
8656 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
8657 	if (!(eec & (1 << 8)))
8658 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
8659 
8660 	/* PHY */
8661 	memcpy(&hw->phy.ops, ii->phy_ops, sizeof(hw->phy.ops));
8662 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
8663 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
8664 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
8665 	hw->phy.mdio.mmds = 0;
8666 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
8667 	hw->phy.mdio.dev = netdev;
8668 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
8669 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
8670 
8671 	ii->get_invariants(hw);
8672 
8673 	/* setup the private structure */
8674 	err = ixgbe_sw_init(adapter);
8675 	if (err)
8676 		goto err_sw_init;
8677 
8678 	/* Make it possible the adapter to be woken up via WOL */
8679 	switch (adapter->hw.mac.type) {
8680 	case ixgbe_mac_82599EB:
8681 	case ixgbe_mac_X540:
8682 	case ixgbe_mac_X550:
8683 	case ixgbe_mac_X550EM_x:
8684 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
8685 		break;
8686 	default:
8687 		break;
8688 	}
8689 
8690 	/*
8691 	 * If there is a fan on this device and it has failed log the
8692 	 * failure.
8693 	 */
8694 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
8695 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
8696 		if (esdp & IXGBE_ESDP_SDP1)
8697 			e_crit(probe, "Fan has stopped, replace the adapter\n");
8698 	}
8699 
8700 	if (allow_unsupported_sfp)
8701 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
8702 
8703 	/* reset_hw fills in the perm_addr as well */
8704 	hw->phy.reset_if_overtemp = true;
8705 	err = hw->mac.ops.reset_hw(hw);
8706 	hw->phy.reset_if_overtemp = false;
8707 	if (err == IXGBE_ERR_SFP_NOT_PRESENT &&
8708 	    hw->mac.type == ixgbe_mac_82598EB) {
8709 		err = 0;
8710 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
8711 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
8712 		e_dev_err("Reload the driver after installing a supported module.\n");
8713 		goto err_sw_init;
8714 	} else if (err) {
8715 		e_dev_err("HW Init failed: %d\n", err);
8716 		goto err_sw_init;
8717 	}
8718 
8719 #ifdef CONFIG_PCI_IOV
8720 	/* SR-IOV not supported on the 82598 */
8721 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
8722 		goto skip_sriov;
8723 	/* Mailbox */
8724 	ixgbe_init_mbx_params_pf(hw);
8725 	memcpy(&hw->mbx.ops, ii->mbx_ops, sizeof(hw->mbx.ops));
8726 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
8727 	ixgbe_enable_sriov(adapter);
8728 skip_sriov:
8729 
8730 #endif
8731 	netdev->features = NETIF_F_SG |
8732 			   NETIF_F_IP_CSUM |
8733 			   NETIF_F_IPV6_CSUM |
8734 			   NETIF_F_HW_VLAN_CTAG_TX |
8735 			   NETIF_F_HW_VLAN_CTAG_RX |
8736 			   NETIF_F_TSO |
8737 			   NETIF_F_TSO6 |
8738 			   NETIF_F_RXHASH |
8739 			   NETIF_F_RXCSUM;
8740 
8741 	netdev->hw_features = netdev->features | NETIF_F_HW_L2FW_DOFFLOAD;
8742 
8743 	switch (adapter->hw.mac.type) {
8744 	case ixgbe_mac_82599EB:
8745 	case ixgbe_mac_X540:
8746 	case ixgbe_mac_X550:
8747 	case ixgbe_mac_X550EM_x:
8748 		netdev->features |= NETIF_F_SCTP_CSUM;
8749 		netdev->hw_features |= NETIF_F_SCTP_CSUM |
8750 				       NETIF_F_NTUPLE;
8751 		break;
8752 	default:
8753 		break;
8754 	}
8755 
8756 	netdev->hw_features |= NETIF_F_RXALL;
8757 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
8758 
8759 	netdev->vlan_features |= NETIF_F_TSO;
8760 	netdev->vlan_features |= NETIF_F_TSO6;
8761 	netdev->vlan_features |= NETIF_F_IP_CSUM;
8762 	netdev->vlan_features |= NETIF_F_IPV6_CSUM;
8763 	netdev->vlan_features |= NETIF_F_SG;
8764 
8765 	netdev->hw_enc_features |= NETIF_F_SG | NETIF_F_IP_CSUM |
8766 				   NETIF_F_IPV6_CSUM;
8767 
8768 	netdev->priv_flags |= IFF_UNICAST_FLT;
8769 	netdev->priv_flags |= IFF_SUPP_NOFCS;
8770 
8771 #ifdef CONFIG_IXGBE_VXLAN
8772 	switch (adapter->hw.mac.type) {
8773 	case ixgbe_mac_X550:
8774 	case ixgbe_mac_X550EM_x:
8775 		netdev->hw_enc_features |= NETIF_F_RXCSUM |
8776 					   NETIF_F_IP_CSUM |
8777 					   NETIF_F_IPV6_CSUM;
8778 		break;
8779 	default:
8780 		break;
8781 	}
8782 #endif /* CONFIG_IXGBE_VXLAN */
8783 
8784 #ifdef CONFIG_IXGBE_DCB
8785 	netdev->dcbnl_ops = &dcbnl_ops;
8786 #endif
8787 
8788 #ifdef IXGBE_FCOE
8789 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
8790 		unsigned int fcoe_l;
8791 
8792 		if (hw->mac.ops.get_device_caps) {
8793 			hw->mac.ops.get_device_caps(hw, &device_caps);
8794 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
8795 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
8796 		}
8797 
8798 
8799 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
8800 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
8801 
8802 		netdev->features |= NETIF_F_FSO |
8803 				    NETIF_F_FCOE_CRC;
8804 
8805 		netdev->vlan_features |= NETIF_F_FSO |
8806 					 NETIF_F_FCOE_CRC |
8807 					 NETIF_F_FCOE_MTU;
8808 	}
8809 #endif /* IXGBE_FCOE */
8810 	if (pci_using_dac) {
8811 		netdev->features |= NETIF_F_HIGHDMA;
8812 		netdev->vlan_features |= NETIF_F_HIGHDMA;
8813 	}
8814 
8815 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
8816 		netdev->hw_features |= NETIF_F_LRO;
8817 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
8818 		netdev->features |= NETIF_F_LRO;
8819 
8820 	/* make sure the EEPROM is good */
8821 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
8822 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
8823 		err = -EIO;
8824 		goto err_sw_init;
8825 	}
8826 
8827 	ixgbe_get_platform_mac_addr(adapter);
8828 
8829 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
8830 
8831 	if (!is_valid_ether_addr(netdev->dev_addr)) {
8832 		e_dev_err("invalid MAC address\n");
8833 		err = -EIO;
8834 		goto err_sw_init;
8835 	}
8836 
8837 	ixgbe_mac_set_default_filter(adapter, hw->mac.perm_addr);
8838 
8839 	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
8840 		    (unsigned long) adapter);
8841 
8842 	if (ixgbe_removed(hw->hw_addr)) {
8843 		err = -EIO;
8844 		goto err_sw_init;
8845 	}
8846 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
8847 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
8848 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
8849 
8850 	err = ixgbe_init_interrupt_scheme(adapter);
8851 	if (err)
8852 		goto err_sw_init;
8853 
8854 	/* WOL not supported for all devices */
8855 	adapter->wol = 0;
8856 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
8857 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
8858 						pdev->subsystem_device);
8859 	if (hw->wol_enabled)
8860 		adapter->wol = IXGBE_WUFC_MAG;
8861 
8862 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
8863 
8864 	/* save off EEPROM version number */
8865 	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
8866 	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
8867 
8868 	/* pick up the PCI bus settings for reporting later */
8869 	if (ixgbe_pcie_from_parent(hw))
8870 		ixgbe_get_parent_bus_info(adapter);
8871 	else
8872 		 hw->mac.ops.get_bus_info(hw);
8873 
8874 	/* calculate the expected PCIe bandwidth required for optimal
8875 	 * performance. Note that some older parts will never have enough
8876 	 * bandwidth due to being older generation PCIe parts. We clamp these
8877 	 * parts to ensure no warning is displayed if it can't be fixed.
8878 	 */
8879 	switch (hw->mac.type) {
8880 	case ixgbe_mac_82598EB:
8881 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
8882 		break;
8883 	default:
8884 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
8885 		break;
8886 	}
8887 
8888 	/* don't check link if we failed to enumerate functions */
8889 	if (expected_gts > 0)
8890 		ixgbe_check_minimum_link(adapter, expected_gts);
8891 
8892 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
8893 	if (err)
8894 		strlcpy(part_str, "Unknown", sizeof(part_str));
8895 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
8896 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
8897 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
8898 			   part_str);
8899 	else
8900 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
8901 			   hw->mac.type, hw->phy.type, part_str);
8902 
8903 	e_dev_info("%pM\n", netdev->dev_addr);
8904 
8905 	/* reset the hardware with the new settings */
8906 	err = hw->mac.ops.start_hw(hw);
8907 	if (err == IXGBE_ERR_EEPROM_VERSION) {
8908 		/* We are running on a pre-production device, log a warning */
8909 		e_dev_warn("This device is a pre-production adapter/LOM. "
8910 			   "Please be aware there may be issues associated "
8911 			   "with your hardware.  If you are experiencing "
8912 			   "problems please contact your Intel or hardware "
8913 			   "representative who provided you with this "
8914 			   "hardware.\n");
8915 	}
8916 	strcpy(netdev->name, "eth%d");
8917 	err = register_netdev(netdev);
8918 	if (err)
8919 		goto err_register;
8920 
8921 	pci_set_drvdata(pdev, adapter);
8922 
8923 	/* power down the optics for 82599 SFP+ fiber */
8924 	if (hw->mac.ops.disable_tx_laser)
8925 		hw->mac.ops.disable_tx_laser(hw);
8926 
8927 	/* carrier off reporting is important to ethtool even BEFORE open */
8928 	netif_carrier_off(netdev);
8929 
8930 #ifdef CONFIG_IXGBE_DCA
8931 	if (dca_add_requester(&pdev->dev) == 0) {
8932 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
8933 		ixgbe_setup_dca(adapter);
8934 	}
8935 #endif
8936 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
8937 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
8938 		for (i = 0; i < adapter->num_vfs; i++)
8939 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
8940 	}
8941 
8942 	/* firmware requires driver version to be 0xFFFFFFFF
8943 	 * since os does not support feature
8944 	 */
8945 	if (hw->mac.ops.set_fw_drv_ver)
8946 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF,
8947 					   0xFF);
8948 
8949 	/* add san mac addr to netdev */
8950 	ixgbe_add_sanmac_netdev(netdev);
8951 
8952 	e_dev_info("%s\n", ixgbe_default_device_descr);
8953 
8954 #ifdef CONFIG_IXGBE_HWMON
8955 	if (ixgbe_sysfs_init(adapter))
8956 		e_err(probe, "failed to allocate sysfs resources\n");
8957 #endif /* CONFIG_IXGBE_HWMON */
8958 
8959 	ixgbe_dbg_adapter_init(adapter);
8960 
8961 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
8962 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
8963 		hw->mac.ops.setup_link(hw,
8964 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
8965 			true);
8966 
8967 	return 0;
8968 
8969 err_register:
8970 	ixgbe_release_hw_control(adapter);
8971 	ixgbe_clear_interrupt_scheme(adapter);
8972 err_sw_init:
8973 	ixgbe_disable_sriov(adapter);
8974 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
8975 	iounmap(adapter->io_addr);
8976 	kfree(adapter->mac_table);
8977 err_ioremap:
8978 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
8979 	free_netdev(netdev);
8980 err_alloc_etherdev:
8981 	pci_release_selected_regions(pdev,
8982 				     pci_select_bars(pdev, IORESOURCE_MEM));
8983 err_pci_reg:
8984 err_dma:
8985 	if (!adapter || disable_dev)
8986 		pci_disable_device(pdev);
8987 	return err;
8988 }
8989 
8990 /**
8991  * ixgbe_remove - Device Removal Routine
8992  * @pdev: PCI device information struct
8993  *
8994  * ixgbe_remove is called by the PCI subsystem to alert the driver
8995  * that it should release a PCI device.  The could be caused by a
8996  * Hot-Plug event, or because the driver is going to be removed from
8997  * memory.
8998  **/
8999 static void ixgbe_remove(struct pci_dev *pdev)
9000 {
9001 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9002 	struct net_device *netdev;
9003 	bool disable_dev;
9004 
9005 	/* if !adapter then we already cleaned up in probe */
9006 	if (!adapter)
9007 		return;
9008 
9009 	netdev  = adapter->netdev;
9010 	ixgbe_dbg_adapter_exit(adapter);
9011 
9012 	set_bit(__IXGBE_REMOVING, &adapter->state);
9013 	cancel_work_sync(&adapter->service_task);
9014 
9015 
9016 #ifdef CONFIG_IXGBE_DCA
9017 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
9018 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
9019 		dca_remove_requester(&pdev->dev);
9020 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
9021 	}
9022 
9023 #endif
9024 #ifdef CONFIG_IXGBE_HWMON
9025 	ixgbe_sysfs_exit(adapter);
9026 #endif /* CONFIG_IXGBE_HWMON */
9027 
9028 	/* remove the added san mac */
9029 	ixgbe_del_sanmac_netdev(netdev);
9030 
9031 #ifdef CONFIG_PCI_IOV
9032 	ixgbe_disable_sriov(adapter);
9033 #endif
9034 	if (netdev->reg_state == NETREG_REGISTERED)
9035 		unregister_netdev(netdev);
9036 
9037 	ixgbe_clear_interrupt_scheme(adapter);
9038 
9039 	ixgbe_release_hw_control(adapter);
9040 
9041 #ifdef CONFIG_DCB
9042 	kfree(adapter->ixgbe_ieee_pfc);
9043 	kfree(adapter->ixgbe_ieee_ets);
9044 
9045 #endif
9046 	iounmap(adapter->io_addr);
9047 	pci_release_selected_regions(pdev, pci_select_bars(pdev,
9048 				     IORESOURCE_MEM));
9049 
9050 	e_dev_info("complete\n");
9051 
9052 	kfree(adapter->mac_table);
9053 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
9054 	free_netdev(netdev);
9055 
9056 	pci_disable_pcie_error_reporting(pdev);
9057 
9058 	if (disable_dev)
9059 		pci_disable_device(pdev);
9060 }
9061 
9062 /**
9063  * ixgbe_io_error_detected - called when PCI error is detected
9064  * @pdev: Pointer to PCI device
9065  * @state: The current pci connection state
9066  *
9067  * This function is called after a PCI bus error affecting
9068  * this device has been detected.
9069  */
9070 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
9071 						pci_channel_state_t state)
9072 {
9073 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9074 	struct net_device *netdev = adapter->netdev;
9075 
9076 #ifdef CONFIG_PCI_IOV
9077 	struct ixgbe_hw *hw = &adapter->hw;
9078 	struct pci_dev *bdev, *vfdev;
9079 	u32 dw0, dw1, dw2, dw3;
9080 	int vf, pos;
9081 	u16 req_id, pf_func;
9082 
9083 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
9084 	    adapter->num_vfs == 0)
9085 		goto skip_bad_vf_detection;
9086 
9087 	bdev = pdev->bus->self;
9088 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
9089 		bdev = bdev->bus->self;
9090 
9091 	if (!bdev)
9092 		goto skip_bad_vf_detection;
9093 
9094 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
9095 	if (!pos)
9096 		goto skip_bad_vf_detection;
9097 
9098 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
9099 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
9100 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
9101 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
9102 	if (ixgbe_removed(hw->hw_addr))
9103 		goto skip_bad_vf_detection;
9104 
9105 	req_id = dw1 >> 16;
9106 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
9107 	if (!(req_id & 0x0080))
9108 		goto skip_bad_vf_detection;
9109 
9110 	pf_func = req_id & 0x01;
9111 	if ((pf_func & 1) == (pdev->devfn & 1)) {
9112 		unsigned int device_id;
9113 
9114 		vf = (req_id & 0x7F) >> 1;
9115 		e_dev_err("VF %d has caused a PCIe error\n", vf);
9116 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
9117 				"%8.8x\tdw3: %8.8x\n",
9118 		dw0, dw1, dw2, dw3);
9119 		switch (adapter->hw.mac.type) {
9120 		case ixgbe_mac_82599EB:
9121 			device_id = IXGBE_82599_VF_DEVICE_ID;
9122 			break;
9123 		case ixgbe_mac_X540:
9124 			device_id = IXGBE_X540_VF_DEVICE_ID;
9125 			break;
9126 		case ixgbe_mac_X550:
9127 			device_id = IXGBE_DEV_ID_X550_VF;
9128 			break;
9129 		case ixgbe_mac_X550EM_x:
9130 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
9131 			break;
9132 		default:
9133 			device_id = 0;
9134 			break;
9135 		}
9136 
9137 		/* Find the pci device of the offending VF */
9138 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
9139 		while (vfdev) {
9140 			if (vfdev->devfn == (req_id & 0xFF))
9141 				break;
9142 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
9143 					       device_id, vfdev);
9144 		}
9145 		/*
9146 		 * There's a slim chance the VF could have been hot plugged,
9147 		 * so if it is no longer present we don't need to issue the
9148 		 * VFLR.  Just clean up the AER in that case.
9149 		 */
9150 		if (vfdev) {
9151 			ixgbe_issue_vf_flr(adapter, vfdev);
9152 			/* Free device reference count */
9153 			pci_dev_put(vfdev);
9154 		}
9155 
9156 		pci_cleanup_aer_uncorrect_error_status(pdev);
9157 	}
9158 
9159 	/*
9160 	 * Even though the error may have occurred on the other port
9161 	 * we still need to increment the vf error reference count for
9162 	 * both ports because the I/O resume function will be called
9163 	 * for both of them.
9164 	 */
9165 	adapter->vferr_refcount++;
9166 
9167 	return PCI_ERS_RESULT_RECOVERED;
9168 
9169 skip_bad_vf_detection:
9170 #endif /* CONFIG_PCI_IOV */
9171 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
9172 		return PCI_ERS_RESULT_DISCONNECT;
9173 
9174 	rtnl_lock();
9175 	netif_device_detach(netdev);
9176 
9177 	if (state == pci_channel_io_perm_failure) {
9178 		rtnl_unlock();
9179 		return PCI_ERS_RESULT_DISCONNECT;
9180 	}
9181 
9182 	if (netif_running(netdev))
9183 		ixgbe_down(adapter);
9184 
9185 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
9186 		pci_disable_device(pdev);
9187 	rtnl_unlock();
9188 
9189 	/* Request a slot reset. */
9190 	return PCI_ERS_RESULT_NEED_RESET;
9191 }
9192 
9193 /**
9194  * ixgbe_io_slot_reset - called after the pci bus has been reset.
9195  * @pdev: Pointer to PCI device
9196  *
9197  * Restart the card from scratch, as if from a cold-boot.
9198  */
9199 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
9200 {
9201 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9202 	pci_ers_result_t result;
9203 	int err;
9204 
9205 	if (pci_enable_device_mem(pdev)) {
9206 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
9207 		result = PCI_ERS_RESULT_DISCONNECT;
9208 	} else {
9209 		smp_mb__before_atomic();
9210 		clear_bit(__IXGBE_DISABLED, &adapter->state);
9211 		adapter->hw.hw_addr = adapter->io_addr;
9212 		pci_set_master(pdev);
9213 		pci_restore_state(pdev);
9214 		pci_save_state(pdev);
9215 
9216 		pci_wake_from_d3(pdev, false);
9217 
9218 		ixgbe_reset(adapter);
9219 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
9220 		result = PCI_ERS_RESULT_RECOVERED;
9221 	}
9222 
9223 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
9224 	if (err) {
9225 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
9226 			  "failed 0x%0x\n", err);
9227 		/* non-fatal, continue */
9228 	}
9229 
9230 	return result;
9231 }
9232 
9233 /**
9234  * ixgbe_io_resume - called when traffic can start flowing again.
9235  * @pdev: Pointer to PCI device
9236  *
9237  * This callback is called when the error recovery driver tells us that
9238  * its OK to resume normal operation.
9239  */
9240 static void ixgbe_io_resume(struct pci_dev *pdev)
9241 {
9242 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
9243 	struct net_device *netdev = adapter->netdev;
9244 
9245 #ifdef CONFIG_PCI_IOV
9246 	if (adapter->vferr_refcount) {
9247 		e_info(drv, "Resuming after VF err\n");
9248 		adapter->vferr_refcount--;
9249 		return;
9250 	}
9251 
9252 #endif
9253 	if (netif_running(netdev))
9254 		ixgbe_up(adapter);
9255 
9256 	netif_device_attach(netdev);
9257 }
9258 
9259 static const struct pci_error_handlers ixgbe_err_handler = {
9260 	.error_detected = ixgbe_io_error_detected,
9261 	.slot_reset = ixgbe_io_slot_reset,
9262 	.resume = ixgbe_io_resume,
9263 };
9264 
9265 static struct pci_driver ixgbe_driver = {
9266 	.name     = ixgbe_driver_name,
9267 	.id_table = ixgbe_pci_tbl,
9268 	.probe    = ixgbe_probe,
9269 	.remove   = ixgbe_remove,
9270 #ifdef CONFIG_PM
9271 	.suspend  = ixgbe_suspend,
9272 	.resume   = ixgbe_resume,
9273 #endif
9274 	.shutdown = ixgbe_shutdown,
9275 	.sriov_configure = ixgbe_pci_sriov_configure,
9276 	.err_handler = &ixgbe_err_handler
9277 };
9278 
9279 /**
9280  * ixgbe_init_module - Driver Registration Routine
9281  *
9282  * ixgbe_init_module is the first routine called when the driver is
9283  * loaded. All it does is register with the PCI subsystem.
9284  **/
9285 static int __init ixgbe_init_module(void)
9286 {
9287 	int ret;
9288 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
9289 	pr_info("%s\n", ixgbe_copyright);
9290 
9291 	ixgbe_dbg_init();
9292 
9293 	ret = pci_register_driver(&ixgbe_driver);
9294 	if (ret) {
9295 		ixgbe_dbg_exit();
9296 		return ret;
9297 	}
9298 
9299 #ifdef CONFIG_IXGBE_DCA
9300 	dca_register_notify(&dca_notifier);
9301 #endif
9302 
9303 	return 0;
9304 }
9305 
9306 module_init(ixgbe_init_module);
9307 
9308 /**
9309  * ixgbe_exit_module - Driver Exit Cleanup Routine
9310  *
9311  * ixgbe_exit_module is called just before the driver is removed
9312  * from memory.
9313  **/
9314 static void __exit ixgbe_exit_module(void)
9315 {
9316 #ifdef CONFIG_IXGBE_DCA
9317 	dca_unregister_notify(&dca_notifier);
9318 #endif
9319 	pci_unregister_driver(&ixgbe_driver);
9320 
9321 	ixgbe_dbg_exit();
9322 }
9323 
9324 #ifdef CONFIG_IXGBE_DCA
9325 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
9326 			    void *p)
9327 {
9328 	int ret_val;
9329 
9330 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
9331 					 __ixgbe_notify_dca);
9332 
9333 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
9334 }
9335 
9336 #endif /* CONFIG_IXGBE_DCA */
9337 
9338 module_exit(ixgbe_exit_module);
9339 
9340 /* ixgbe_main.c */
9341