xref: /linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision 3932b9ca55b0be314a36d3e84faff3e823c081f5)
1 /*******************************************************************************
2 
3   Intel 10 Gigabit PCI Express Linux driver
4   Copyright(c) 1999 - 2014 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/ethtool.h>
46 #include <linux/if.h>
47 #include <linux/if_vlan.h>
48 #include <linux/if_macvlan.h>
49 #include <linux/if_bridge.h>
50 #include <linux/prefetch.h>
51 #include <scsi/fc/fc_fcoe.h>
52 
53 #include "ixgbe.h"
54 #include "ixgbe_common.h"
55 #include "ixgbe_dcb_82599.h"
56 #include "ixgbe_sriov.h"
57 
58 char ixgbe_driver_name[] = "ixgbe";
59 static const char ixgbe_driver_string[] =
60 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
61 #ifdef IXGBE_FCOE
62 char ixgbe_default_device_descr[] =
63 			      "Intel(R) 10 Gigabit Network Connection";
64 #else
65 static char ixgbe_default_device_descr[] =
66 			      "Intel(R) 10 Gigabit Network Connection";
67 #endif
68 #define DRV_VERSION "3.19.1-k"
69 const char ixgbe_driver_version[] = DRV_VERSION;
70 static const char ixgbe_copyright[] =
71 				"Copyright (c) 1999-2014 Intel Corporation.";
72 
73 static const struct ixgbe_info *ixgbe_info_tbl[] = {
74 	[board_82598] = &ixgbe_82598_info,
75 	[board_82599] = &ixgbe_82599_info,
76 	[board_X540] = &ixgbe_X540_info,
77 };
78 
79 /* ixgbe_pci_tbl - PCI Device ID Table
80  *
81  * Wildcard entries (PCI_ANY_ID) should come last
82  * Last entry must be all 0s
83  *
84  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
85  *   Class, Class Mask, private data (not used) }
86  */
87 static const struct pci_device_id ixgbe_pci_tbl[] = {
88 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
89 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
90 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
91 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
92 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
93 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
94 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
95 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
96 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
97 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
98 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
99 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
100 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
101 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
102 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
103 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
104 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
105 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
115 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
116 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
117 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
118 	/* required last entry */
119 	{0, }
120 };
121 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
122 
123 #ifdef CONFIG_IXGBE_DCA
124 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
125 			    void *p);
126 static struct notifier_block dca_notifier = {
127 	.notifier_call = ixgbe_notify_dca,
128 	.next          = NULL,
129 	.priority      = 0
130 };
131 #endif
132 
133 #ifdef CONFIG_PCI_IOV
134 static unsigned int max_vfs;
135 module_param(max_vfs, uint, 0);
136 MODULE_PARM_DESC(max_vfs,
137 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
138 #endif /* CONFIG_PCI_IOV */
139 
140 static unsigned int allow_unsupported_sfp;
141 module_param(allow_unsupported_sfp, uint, 0);
142 MODULE_PARM_DESC(allow_unsupported_sfp,
143 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
144 
145 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
146 static int debug = -1;
147 module_param(debug, int, 0);
148 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
149 
150 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
151 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
152 MODULE_LICENSE("GPL");
153 MODULE_VERSION(DRV_VERSION);
154 
155 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
156 
157 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
158 					  u32 reg, u16 *value)
159 {
160 	struct pci_dev *parent_dev;
161 	struct pci_bus *parent_bus;
162 
163 	parent_bus = adapter->pdev->bus->parent;
164 	if (!parent_bus)
165 		return -1;
166 
167 	parent_dev = parent_bus->self;
168 	if (!parent_dev)
169 		return -1;
170 
171 	if (!pci_is_pcie(parent_dev))
172 		return -1;
173 
174 	pcie_capability_read_word(parent_dev, reg, value);
175 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
176 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
177 		return -1;
178 	return 0;
179 }
180 
181 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
182 {
183 	struct ixgbe_hw *hw = &adapter->hw;
184 	u16 link_status = 0;
185 	int err;
186 
187 	hw->bus.type = ixgbe_bus_type_pci_express;
188 
189 	/* Get the negotiated link width and speed from PCI config space of the
190 	 * parent, as this device is behind a switch
191 	 */
192 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
193 
194 	/* assume caller will handle error case */
195 	if (err)
196 		return err;
197 
198 	hw->bus.width = ixgbe_convert_bus_width(link_status);
199 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
200 
201 	return 0;
202 }
203 
204 /**
205  * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
206  * @hw: hw specific details
207  *
208  * This function is used by probe to determine whether a device's PCI-Express
209  * bandwidth details should be gathered from the parent bus instead of from the
210  * device. Used to ensure that various locations all have the correct device ID
211  * checks.
212  */
213 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
214 {
215 	switch (hw->device_id) {
216 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
217 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
218 		return true;
219 	default:
220 		return false;
221 	}
222 }
223 
224 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
225 				     int expected_gts)
226 {
227 	int max_gts = 0;
228 	enum pci_bus_speed speed = PCI_SPEED_UNKNOWN;
229 	enum pcie_link_width width = PCIE_LNK_WIDTH_UNKNOWN;
230 	struct pci_dev *pdev;
231 
232 	/* determine whether to use the the parent device
233 	 */
234 	if (ixgbe_pcie_from_parent(&adapter->hw))
235 		pdev = adapter->pdev->bus->parent->self;
236 	else
237 		pdev = adapter->pdev;
238 
239 	if (pcie_get_minimum_link(pdev, &speed, &width) ||
240 	    speed == PCI_SPEED_UNKNOWN || width == PCIE_LNK_WIDTH_UNKNOWN) {
241 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
242 		return;
243 	}
244 
245 	switch (speed) {
246 	case PCIE_SPEED_2_5GT:
247 		/* 8b/10b encoding reduces max throughput by 20% */
248 		max_gts = 2 * width;
249 		break;
250 	case PCIE_SPEED_5_0GT:
251 		/* 8b/10b encoding reduces max throughput by 20% */
252 		max_gts = 4 * width;
253 		break;
254 	case PCIE_SPEED_8_0GT:
255 		/* 128b/130b encoding reduces throughput by less than 2% */
256 		max_gts = 8 * width;
257 		break;
258 	default:
259 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
260 		return;
261 	}
262 
263 	e_dev_info("PCI Express bandwidth of %dGT/s available\n",
264 		   max_gts);
265 	e_dev_info("(Speed:%s, Width: x%d, Encoding Loss:%s)\n",
266 		   (speed == PCIE_SPEED_8_0GT ? "8.0GT/s" :
267 		    speed == PCIE_SPEED_5_0GT ? "5.0GT/s" :
268 		    speed == PCIE_SPEED_2_5GT ? "2.5GT/s" :
269 		    "Unknown"),
270 		   width,
271 		   (speed == PCIE_SPEED_2_5GT ? "20%" :
272 		    speed == PCIE_SPEED_5_0GT ? "20%" :
273 		    speed == PCIE_SPEED_8_0GT ? "<2%" :
274 		    "Unknown"));
275 
276 	if (max_gts < expected_gts) {
277 		e_dev_warn("This is not sufficient for optimal performance of this card.\n");
278 		e_dev_warn("For optimal performance, at least %dGT/s of bandwidth is required.\n",
279 			expected_gts);
280 		e_dev_warn("A slot with more lanes and/or higher speed is suggested.\n");
281 	}
282 }
283 
284 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
285 {
286 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
287 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
288 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
289 		schedule_work(&adapter->service_task);
290 }
291 
292 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
293 {
294 	struct ixgbe_adapter *adapter = hw->back;
295 
296 	if (!hw->hw_addr)
297 		return;
298 	hw->hw_addr = NULL;
299 	e_dev_err("Adapter removed\n");
300 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
301 		ixgbe_service_event_schedule(adapter);
302 }
303 
304 static void ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
305 {
306 	u32 value;
307 
308 	/* The following check not only optimizes a bit by not
309 	 * performing a read on the status register when the
310 	 * register just read was a status register read that
311 	 * returned IXGBE_FAILED_READ_REG. It also blocks any
312 	 * potential recursion.
313 	 */
314 	if (reg == IXGBE_STATUS) {
315 		ixgbe_remove_adapter(hw);
316 		return;
317 	}
318 	value = ixgbe_read_reg(hw, IXGBE_STATUS);
319 	if (value == IXGBE_FAILED_READ_REG)
320 		ixgbe_remove_adapter(hw);
321 }
322 
323 /**
324  * ixgbe_read_reg - Read from device register
325  * @hw: hw specific details
326  * @reg: offset of register to read
327  *
328  * Returns : value read or IXGBE_FAILED_READ_REG if removed
329  *
330  * This function is used to read device registers. It checks for device
331  * removal by confirming any read that returns all ones by checking the
332  * status register value for all ones. This function avoids reading from
333  * the hardware if a removal was previously detected in which case it
334  * returns IXGBE_FAILED_READ_REG (all ones).
335  */
336 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
337 {
338 	u8 __iomem *reg_addr = ACCESS_ONCE(hw->hw_addr);
339 	u32 value;
340 
341 	if (ixgbe_removed(reg_addr))
342 		return IXGBE_FAILED_READ_REG;
343 	value = readl(reg_addr + reg);
344 	if (unlikely(value == IXGBE_FAILED_READ_REG))
345 		ixgbe_check_remove(hw, reg);
346 	return value;
347 }
348 
349 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
350 {
351 	u16 value;
352 
353 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
354 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
355 		ixgbe_remove_adapter(hw);
356 		return true;
357 	}
358 	return false;
359 }
360 
361 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
362 {
363 	struct ixgbe_adapter *adapter = hw->back;
364 	u16 value;
365 
366 	if (ixgbe_removed(hw->hw_addr))
367 		return IXGBE_FAILED_READ_CFG_WORD;
368 	pci_read_config_word(adapter->pdev, reg, &value);
369 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
370 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
371 		return IXGBE_FAILED_READ_CFG_WORD;
372 	return value;
373 }
374 
375 #ifdef CONFIG_PCI_IOV
376 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
377 {
378 	struct ixgbe_adapter *adapter = hw->back;
379 	u32 value;
380 
381 	if (ixgbe_removed(hw->hw_addr))
382 		return IXGBE_FAILED_READ_CFG_DWORD;
383 	pci_read_config_dword(adapter->pdev, reg, &value);
384 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
385 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
386 		return IXGBE_FAILED_READ_CFG_DWORD;
387 	return value;
388 }
389 #endif /* CONFIG_PCI_IOV */
390 
391 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
392 {
393 	struct ixgbe_adapter *adapter = hw->back;
394 
395 	if (ixgbe_removed(hw->hw_addr))
396 		return;
397 	pci_write_config_word(adapter->pdev, reg, value);
398 }
399 
400 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
401 {
402 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
403 
404 	/* flush memory to make sure state is correct before next watchdog */
405 	smp_mb__before_atomic();
406 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
407 }
408 
409 struct ixgbe_reg_info {
410 	u32 ofs;
411 	char *name;
412 };
413 
414 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
415 
416 	/* General Registers */
417 	{IXGBE_CTRL, "CTRL"},
418 	{IXGBE_STATUS, "STATUS"},
419 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
420 
421 	/* Interrupt Registers */
422 	{IXGBE_EICR, "EICR"},
423 
424 	/* RX Registers */
425 	{IXGBE_SRRCTL(0), "SRRCTL"},
426 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
427 	{IXGBE_RDLEN(0), "RDLEN"},
428 	{IXGBE_RDH(0), "RDH"},
429 	{IXGBE_RDT(0), "RDT"},
430 	{IXGBE_RXDCTL(0), "RXDCTL"},
431 	{IXGBE_RDBAL(0), "RDBAL"},
432 	{IXGBE_RDBAH(0), "RDBAH"},
433 
434 	/* TX Registers */
435 	{IXGBE_TDBAL(0), "TDBAL"},
436 	{IXGBE_TDBAH(0), "TDBAH"},
437 	{IXGBE_TDLEN(0), "TDLEN"},
438 	{IXGBE_TDH(0), "TDH"},
439 	{IXGBE_TDT(0), "TDT"},
440 	{IXGBE_TXDCTL(0), "TXDCTL"},
441 
442 	/* List Terminator */
443 	{}
444 };
445 
446 
447 /*
448  * ixgbe_regdump - register printout routine
449  */
450 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
451 {
452 	int i = 0, j = 0;
453 	char rname[16];
454 	u32 regs[64];
455 
456 	switch (reginfo->ofs) {
457 	case IXGBE_SRRCTL(0):
458 		for (i = 0; i < 64; i++)
459 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
460 		break;
461 	case IXGBE_DCA_RXCTRL(0):
462 		for (i = 0; i < 64; i++)
463 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
464 		break;
465 	case IXGBE_RDLEN(0):
466 		for (i = 0; i < 64; i++)
467 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
468 		break;
469 	case IXGBE_RDH(0):
470 		for (i = 0; i < 64; i++)
471 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
472 		break;
473 	case IXGBE_RDT(0):
474 		for (i = 0; i < 64; i++)
475 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
476 		break;
477 	case IXGBE_RXDCTL(0):
478 		for (i = 0; i < 64; i++)
479 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
480 		break;
481 	case IXGBE_RDBAL(0):
482 		for (i = 0; i < 64; i++)
483 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
484 		break;
485 	case IXGBE_RDBAH(0):
486 		for (i = 0; i < 64; i++)
487 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
488 		break;
489 	case IXGBE_TDBAL(0):
490 		for (i = 0; i < 64; i++)
491 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
492 		break;
493 	case IXGBE_TDBAH(0):
494 		for (i = 0; i < 64; i++)
495 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
496 		break;
497 	case IXGBE_TDLEN(0):
498 		for (i = 0; i < 64; i++)
499 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
500 		break;
501 	case IXGBE_TDH(0):
502 		for (i = 0; i < 64; i++)
503 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
504 		break;
505 	case IXGBE_TDT(0):
506 		for (i = 0; i < 64; i++)
507 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
508 		break;
509 	case IXGBE_TXDCTL(0):
510 		for (i = 0; i < 64; i++)
511 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
512 		break;
513 	default:
514 		pr_info("%-15s %08x\n", reginfo->name,
515 			IXGBE_READ_REG(hw, reginfo->ofs));
516 		return;
517 	}
518 
519 	for (i = 0; i < 8; i++) {
520 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i*8, i*8+7);
521 		pr_err("%-15s", rname);
522 		for (j = 0; j < 8; j++)
523 			pr_cont(" %08x", regs[i*8+j]);
524 		pr_cont("\n");
525 	}
526 
527 }
528 
529 /*
530  * ixgbe_dump - Print registers, tx-rings and rx-rings
531  */
532 static void ixgbe_dump(struct ixgbe_adapter *adapter)
533 {
534 	struct net_device *netdev = adapter->netdev;
535 	struct ixgbe_hw *hw = &adapter->hw;
536 	struct ixgbe_reg_info *reginfo;
537 	int n = 0;
538 	struct ixgbe_ring *tx_ring;
539 	struct ixgbe_tx_buffer *tx_buffer;
540 	union ixgbe_adv_tx_desc *tx_desc;
541 	struct my_u0 { u64 a; u64 b; } *u0;
542 	struct ixgbe_ring *rx_ring;
543 	union ixgbe_adv_rx_desc *rx_desc;
544 	struct ixgbe_rx_buffer *rx_buffer_info;
545 	u32 staterr;
546 	int i = 0;
547 
548 	if (!netif_msg_hw(adapter))
549 		return;
550 
551 	/* Print netdevice Info */
552 	if (netdev) {
553 		dev_info(&adapter->pdev->dev, "Net device Info\n");
554 		pr_info("Device Name     state            "
555 			"trans_start      last_rx\n");
556 		pr_info("%-15s %016lX %016lX %016lX\n",
557 			netdev->name,
558 			netdev->state,
559 			netdev->trans_start,
560 			netdev->last_rx);
561 	}
562 
563 	/* Print Registers */
564 	dev_info(&adapter->pdev->dev, "Register Dump\n");
565 	pr_info(" Register Name   Value\n");
566 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
567 	     reginfo->name; reginfo++) {
568 		ixgbe_regdump(hw, reginfo);
569 	}
570 
571 	/* Print TX Ring Summary */
572 	if (!netdev || !netif_running(netdev))
573 		return;
574 
575 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
576 	pr_info(" %s     %s              %s        %s\n",
577 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
578 		"leng", "ntw", "timestamp");
579 	for (n = 0; n < adapter->num_tx_queues; n++) {
580 		tx_ring = adapter->tx_ring[n];
581 		tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
582 		pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
583 			   n, tx_ring->next_to_use, tx_ring->next_to_clean,
584 			   (u64)dma_unmap_addr(tx_buffer, dma),
585 			   dma_unmap_len(tx_buffer, len),
586 			   tx_buffer->next_to_watch,
587 			   (u64)tx_buffer->time_stamp);
588 	}
589 
590 	/* Print TX Rings */
591 	if (!netif_msg_tx_done(adapter))
592 		goto rx_ring_summary;
593 
594 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
595 
596 	/* Transmit Descriptor Formats
597 	 *
598 	 * 82598 Advanced Transmit Descriptor
599 	 *   +--------------------------------------------------------------+
600 	 * 0 |         Buffer Address [63:0]                                |
601 	 *   +--------------------------------------------------------------+
602 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
603 	 *   +--------------------------------------------------------------+
604 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
605 	 *
606 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
607 	 *   +--------------------------------------------------------------+
608 	 * 0 |                          RSV [63:0]                          |
609 	 *   +--------------------------------------------------------------+
610 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
611 	 *   +--------------------------------------------------------------+
612 	 *   63                       36 35   32 31                         0
613 	 *
614 	 * 82599+ Advanced Transmit Descriptor
615 	 *   +--------------------------------------------------------------+
616 	 * 0 |         Buffer Address [63:0]                                |
617 	 *   +--------------------------------------------------------------+
618 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
619 	 *   +--------------------------------------------------------------+
620 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
621 	 *
622 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
623 	 *   +--------------------------------------------------------------+
624 	 * 0 |                          RSV [63:0]                          |
625 	 *   +--------------------------------------------------------------+
626 	 * 8 |            RSV           |  STA  |           RSV             |
627 	 *   +--------------------------------------------------------------+
628 	 *   63                       36 35   32 31                         0
629 	 */
630 
631 	for (n = 0; n < adapter->num_tx_queues; n++) {
632 		tx_ring = adapter->tx_ring[n];
633 		pr_info("------------------------------------\n");
634 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
635 		pr_info("------------------------------------\n");
636 		pr_info("%s%s    %s              %s        %s          %s\n",
637 			"T [desc]     [address 63:0  ] ",
638 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
639 			"leng", "ntw", "timestamp", "bi->skb");
640 
641 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
642 			tx_desc = IXGBE_TX_DESC(tx_ring, i);
643 			tx_buffer = &tx_ring->tx_buffer_info[i];
644 			u0 = (struct my_u0 *)tx_desc;
645 			if (dma_unmap_len(tx_buffer, len) > 0) {
646 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p",
647 					i,
648 					le64_to_cpu(u0->a),
649 					le64_to_cpu(u0->b),
650 					(u64)dma_unmap_addr(tx_buffer, dma),
651 					dma_unmap_len(tx_buffer, len),
652 					tx_buffer->next_to_watch,
653 					(u64)tx_buffer->time_stamp,
654 					tx_buffer->skb);
655 				if (i == tx_ring->next_to_use &&
656 					i == tx_ring->next_to_clean)
657 					pr_cont(" NTC/U\n");
658 				else if (i == tx_ring->next_to_use)
659 					pr_cont(" NTU\n");
660 				else if (i == tx_ring->next_to_clean)
661 					pr_cont(" NTC\n");
662 				else
663 					pr_cont("\n");
664 
665 				if (netif_msg_pktdata(adapter) &&
666 				    tx_buffer->skb)
667 					print_hex_dump(KERN_INFO, "",
668 						DUMP_PREFIX_ADDRESS, 16, 1,
669 						tx_buffer->skb->data,
670 						dma_unmap_len(tx_buffer, len),
671 						true);
672 			}
673 		}
674 	}
675 
676 	/* Print RX Rings Summary */
677 rx_ring_summary:
678 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
679 	pr_info("Queue [NTU] [NTC]\n");
680 	for (n = 0; n < adapter->num_rx_queues; n++) {
681 		rx_ring = adapter->rx_ring[n];
682 		pr_info("%5d %5X %5X\n",
683 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
684 	}
685 
686 	/* Print RX Rings */
687 	if (!netif_msg_rx_status(adapter))
688 		return;
689 
690 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
691 
692 	/* Receive Descriptor Formats
693 	 *
694 	 * 82598 Advanced Receive Descriptor (Read) Format
695 	 *    63                                           1        0
696 	 *    +-----------------------------------------------------+
697 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
698 	 *    +----------------------------------------------+------+
699 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
700 	 *    +-----------------------------------------------------+
701 	 *
702 	 *
703 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
704 	 *
705 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
706 	 *   +------------------------------------------------------+
707 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
708 	 *   | Packet   | IP     |   |          |     | Type | Type |
709 	 *   | Checksum | Ident  |   |          |     |      |      |
710 	 *   +------------------------------------------------------+
711 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
712 	 *   +------------------------------------------------------+
713 	 *   63       48 47    32 31            20 19               0
714 	 *
715 	 * 82599+ Advanced Receive Descriptor (Read) Format
716 	 *    63                                           1        0
717 	 *    +-----------------------------------------------------+
718 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
719 	 *    +----------------------------------------------+------+
720 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
721 	 *    +-----------------------------------------------------+
722 	 *
723 	 *
724 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
725 	 *
726 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
727 	 *   +------------------------------------------------------+
728 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
729 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
730 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
731 	 *   +------------------------------------------------------+
732 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
733 	 *   +------------------------------------------------------+
734 	 *   63       48 47    32 31          20 19                 0
735 	 */
736 
737 	for (n = 0; n < adapter->num_rx_queues; n++) {
738 		rx_ring = adapter->rx_ring[n];
739 		pr_info("------------------------------------\n");
740 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
741 		pr_info("------------------------------------\n");
742 		pr_info("%s%s%s",
743 			"R  [desc]      [ PktBuf     A0] ",
744 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
745 			"<-- Adv Rx Read format\n");
746 		pr_info("%s%s%s",
747 			"RWB[desc]      [PcsmIpSHl PtRs] ",
748 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
749 			"<-- Adv Rx Write-Back format\n");
750 
751 		for (i = 0; i < rx_ring->count; i++) {
752 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
753 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
754 			u0 = (struct my_u0 *)rx_desc;
755 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
756 			if (staterr & IXGBE_RXD_STAT_DD) {
757 				/* Descriptor Done */
758 				pr_info("RWB[0x%03X]     %016llX "
759 					"%016llX ---------------- %p", i,
760 					le64_to_cpu(u0->a),
761 					le64_to_cpu(u0->b),
762 					rx_buffer_info->skb);
763 			} else {
764 				pr_info("R  [0x%03X]     %016llX "
765 					"%016llX %016llX %p", i,
766 					le64_to_cpu(u0->a),
767 					le64_to_cpu(u0->b),
768 					(u64)rx_buffer_info->dma,
769 					rx_buffer_info->skb);
770 
771 				if (netif_msg_pktdata(adapter) &&
772 				    rx_buffer_info->dma) {
773 					print_hex_dump(KERN_INFO, "",
774 					   DUMP_PREFIX_ADDRESS, 16, 1,
775 					   page_address(rx_buffer_info->page) +
776 						    rx_buffer_info->page_offset,
777 					   ixgbe_rx_bufsz(rx_ring), true);
778 				}
779 			}
780 
781 			if (i == rx_ring->next_to_use)
782 				pr_cont(" NTU\n");
783 			else if (i == rx_ring->next_to_clean)
784 				pr_cont(" NTC\n");
785 			else
786 				pr_cont("\n");
787 
788 		}
789 	}
790 }
791 
792 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
793 {
794 	u32 ctrl_ext;
795 
796 	/* Let firmware take over control of h/w */
797 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
798 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
799 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
800 }
801 
802 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
803 {
804 	u32 ctrl_ext;
805 
806 	/* Let firmware know the driver has taken over */
807 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
808 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
809 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
810 }
811 
812 /**
813  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
814  * @adapter: pointer to adapter struct
815  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
816  * @queue: queue to map the corresponding interrupt to
817  * @msix_vector: the vector to map to the corresponding queue
818  *
819  */
820 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
821 			   u8 queue, u8 msix_vector)
822 {
823 	u32 ivar, index;
824 	struct ixgbe_hw *hw = &adapter->hw;
825 	switch (hw->mac.type) {
826 	case ixgbe_mac_82598EB:
827 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
828 		if (direction == -1)
829 			direction = 0;
830 		index = (((direction * 64) + queue) >> 2) & 0x1F;
831 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
832 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
833 		ivar |= (msix_vector << (8 * (queue & 0x3)));
834 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
835 		break;
836 	case ixgbe_mac_82599EB:
837 	case ixgbe_mac_X540:
838 		if (direction == -1) {
839 			/* other causes */
840 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
841 			index = ((queue & 1) * 8);
842 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
843 			ivar &= ~(0xFF << index);
844 			ivar |= (msix_vector << index);
845 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
846 			break;
847 		} else {
848 			/* tx or rx causes */
849 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
850 			index = ((16 * (queue & 1)) + (8 * direction));
851 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
852 			ivar &= ~(0xFF << index);
853 			ivar |= (msix_vector << index);
854 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
855 			break;
856 		}
857 	default:
858 		break;
859 	}
860 }
861 
862 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
863 					  u64 qmask)
864 {
865 	u32 mask;
866 
867 	switch (adapter->hw.mac.type) {
868 	case ixgbe_mac_82598EB:
869 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
870 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
871 		break;
872 	case ixgbe_mac_82599EB:
873 	case ixgbe_mac_X540:
874 		mask = (qmask & 0xFFFFFFFF);
875 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
876 		mask = (qmask >> 32);
877 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
878 		break;
879 	default:
880 		break;
881 	}
882 }
883 
884 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring *ring,
885 				      struct ixgbe_tx_buffer *tx_buffer)
886 {
887 	if (tx_buffer->skb) {
888 		dev_kfree_skb_any(tx_buffer->skb);
889 		if (dma_unmap_len(tx_buffer, len))
890 			dma_unmap_single(ring->dev,
891 					 dma_unmap_addr(tx_buffer, dma),
892 					 dma_unmap_len(tx_buffer, len),
893 					 DMA_TO_DEVICE);
894 	} else if (dma_unmap_len(tx_buffer, len)) {
895 		dma_unmap_page(ring->dev,
896 			       dma_unmap_addr(tx_buffer, dma),
897 			       dma_unmap_len(tx_buffer, len),
898 			       DMA_TO_DEVICE);
899 	}
900 	tx_buffer->next_to_watch = NULL;
901 	tx_buffer->skb = NULL;
902 	dma_unmap_len_set(tx_buffer, len, 0);
903 	/* tx_buffer must be completely set up in the transmit path */
904 }
905 
906 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
907 {
908 	struct ixgbe_hw *hw = &adapter->hw;
909 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
910 	int i;
911 	u32 data;
912 
913 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
914 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
915 		return;
916 
917 	switch (hw->mac.type) {
918 	case ixgbe_mac_82598EB:
919 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
920 		break;
921 	default:
922 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
923 	}
924 	hwstats->lxoffrxc += data;
925 
926 	/* refill credits (no tx hang) if we received xoff */
927 	if (!data)
928 		return;
929 
930 	for (i = 0; i < adapter->num_tx_queues; i++)
931 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
932 			  &adapter->tx_ring[i]->state);
933 }
934 
935 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
936 {
937 	struct ixgbe_hw *hw = &adapter->hw;
938 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
939 	u32 xoff[8] = {0};
940 	u8 tc;
941 	int i;
942 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
943 
944 	if (adapter->ixgbe_ieee_pfc)
945 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
946 
947 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
948 		ixgbe_update_xoff_rx_lfc(adapter);
949 		return;
950 	}
951 
952 	/* update stats for each tc, only valid with PFC enabled */
953 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
954 		u32 pxoffrxc;
955 
956 		switch (hw->mac.type) {
957 		case ixgbe_mac_82598EB:
958 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
959 			break;
960 		default:
961 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
962 		}
963 		hwstats->pxoffrxc[i] += pxoffrxc;
964 		/* Get the TC for given UP */
965 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
966 		xoff[tc] += pxoffrxc;
967 	}
968 
969 	/* disarm tx queues that have received xoff frames */
970 	for (i = 0; i < adapter->num_tx_queues; i++) {
971 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
972 
973 		tc = tx_ring->dcb_tc;
974 		if (xoff[tc])
975 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
976 	}
977 }
978 
979 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
980 {
981 	return ring->stats.packets;
982 }
983 
984 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
985 {
986 	struct ixgbe_adapter *adapter;
987 	struct ixgbe_hw *hw;
988 	u32 head, tail;
989 
990 	if (ring->l2_accel_priv)
991 		adapter = ring->l2_accel_priv->real_adapter;
992 	else
993 		adapter = netdev_priv(ring->netdev);
994 
995 	hw = &adapter->hw;
996 	head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
997 	tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
998 
999 	if (head != tail)
1000 		return (head < tail) ?
1001 			tail - head : (tail + ring->count - head);
1002 
1003 	return 0;
1004 }
1005 
1006 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1007 {
1008 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1009 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1010 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1011 
1012 	clear_check_for_tx_hang(tx_ring);
1013 
1014 	/*
1015 	 * Check for a hung queue, but be thorough. This verifies
1016 	 * that a transmit has been completed since the previous
1017 	 * check AND there is at least one packet pending. The
1018 	 * ARMED bit is set to indicate a potential hang. The
1019 	 * bit is cleared if a pause frame is received to remove
1020 	 * false hang detection due to PFC or 802.3x frames. By
1021 	 * requiring this to fail twice we avoid races with
1022 	 * pfc clearing the ARMED bit and conditions where we
1023 	 * run the check_tx_hang logic with a transmit completion
1024 	 * pending but without time to complete it yet.
1025 	 */
1026 	if (tx_done_old == tx_done && tx_pending)
1027 		/* make sure it is true for two checks in a row */
1028 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1029 					&tx_ring->state);
1030 	/* update completed stats and continue */
1031 	tx_ring->tx_stats.tx_done_old = tx_done;
1032 	/* reset the countdown */
1033 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1034 
1035 	return false;
1036 }
1037 
1038 /**
1039  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1040  * @adapter: driver private struct
1041  **/
1042 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1043 {
1044 
1045 	/* Do the reset outside of interrupt context */
1046 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1047 		adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
1048 		e_warn(drv, "initiating reset due to tx timeout\n");
1049 		ixgbe_service_event_schedule(adapter);
1050 	}
1051 }
1052 
1053 /**
1054  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1055  * @q_vector: structure containing interrupt and ring information
1056  * @tx_ring: tx ring to clean
1057  **/
1058 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1059 			       struct ixgbe_ring *tx_ring)
1060 {
1061 	struct ixgbe_adapter *adapter = q_vector->adapter;
1062 	struct ixgbe_tx_buffer *tx_buffer;
1063 	union ixgbe_adv_tx_desc *tx_desc;
1064 	unsigned int total_bytes = 0, total_packets = 0;
1065 	unsigned int budget = q_vector->tx.work_limit;
1066 	unsigned int i = tx_ring->next_to_clean;
1067 
1068 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1069 		return true;
1070 
1071 	tx_buffer = &tx_ring->tx_buffer_info[i];
1072 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1073 	i -= tx_ring->count;
1074 
1075 	do {
1076 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1077 
1078 		/* if next_to_watch is not set then there is no work pending */
1079 		if (!eop_desc)
1080 			break;
1081 
1082 		/* prevent any other reads prior to eop_desc */
1083 		read_barrier_depends();
1084 
1085 		/* if DD is not set pending work has not been completed */
1086 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1087 			break;
1088 
1089 		/* clear next_to_watch to prevent false hangs */
1090 		tx_buffer->next_to_watch = NULL;
1091 
1092 		/* update the statistics for this packet */
1093 		total_bytes += tx_buffer->bytecount;
1094 		total_packets += tx_buffer->gso_segs;
1095 
1096 		/* free the skb */
1097 		dev_kfree_skb_any(tx_buffer->skb);
1098 
1099 		/* unmap skb header data */
1100 		dma_unmap_single(tx_ring->dev,
1101 				 dma_unmap_addr(tx_buffer, dma),
1102 				 dma_unmap_len(tx_buffer, len),
1103 				 DMA_TO_DEVICE);
1104 
1105 		/* clear tx_buffer data */
1106 		tx_buffer->skb = NULL;
1107 		dma_unmap_len_set(tx_buffer, len, 0);
1108 
1109 		/* unmap remaining buffers */
1110 		while (tx_desc != eop_desc) {
1111 			tx_buffer++;
1112 			tx_desc++;
1113 			i++;
1114 			if (unlikely(!i)) {
1115 				i -= tx_ring->count;
1116 				tx_buffer = tx_ring->tx_buffer_info;
1117 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1118 			}
1119 
1120 			/* unmap any remaining paged data */
1121 			if (dma_unmap_len(tx_buffer, len)) {
1122 				dma_unmap_page(tx_ring->dev,
1123 					       dma_unmap_addr(tx_buffer, dma),
1124 					       dma_unmap_len(tx_buffer, len),
1125 					       DMA_TO_DEVICE);
1126 				dma_unmap_len_set(tx_buffer, len, 0);
1127 			}
1128 		}
1129 
1130 		/* move us one more past the eop_desc for start of next pkt */
1131 		tx_buffer++;
1132 		tx_desc++;
1133 		i++;
1134 		if (unlikely(!i)) {
1135 			i -= tx_ring->count;
1136 			tx_buffer = tx_ring->tx_buffer_info;
1137 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1138 		}
1139 
1140 		/* issue prefetch for next Tx descriptor */
1141 		prefetch(tx_desc);
1142 
1143 		/* update budget accounting */
1144 		budget--;
1145 	} while (likely(budget));
1146 
1147 	i += tx_ring->count;
1148 	tx_ring->next_to_clean = i;
1149 	u64_stats_update_begin(&tx_ring->syncp);
1150 	tx_ring->stats.bytes += total_bytes;
1151 	tx_ring->stats.packets += total_packets;
1152 	u64_stats_update_end(&tx_ring->syncp);
1153 	q_vector->tx.total_bytes += total_bytes;
1154 	q_vector->tx.total_packets += total_packets;
1155 
1156 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1157 		/* schedule immediate reset if we believe we hung */
1158 		struct ixgbe_hw *hw = &adapter->hw;
1159 		e_err(drv, "Detected Tx Unit Hang\n"
1160 			"  Tx Queue             <%d>\n"
1161 			"  TDH, TDT             <%x>, <%x>\n"
1162 			"  next_to_use          <%x>\n"
1163 			"  next_to_clean        <%x>\n"
1164 			"tx_buffer_info[next_to_clean]\n"
1165 			"  time_stamp           <%lx>\n"
1166 			"  jiffies              <%lx>\n",
1167 			tx_ring->queue_index,
1168 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1169 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1170 			tx_ring->next_to_use, i,
1171 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1172 
1173 		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
1174 
1175 		e_info(probe,
1176 		       "tx hang %d detected on queue %d, resetting adapter\n",
1177 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1178 
1179 		/* schedule immediate reset if we believe we hung */
1180 		ixgbe_tx_timeout_reset(adapter);
1181 
1182 		/* the adapter is about to reset, no point in enabling stuff */
1183 		return true;
1184 	}
1185 
1186 	netdev_tx_completed_queue(txring_txq(tx_ring),
1187 				  total_packets, total_bytes);
1188 
1189 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1190 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1191 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1192 		/* Make sure that anybody stopping the queue after this
1193 		 * sees the new next_to_clean.
1194 		 */
1195 		smp_mb();
1196 		if (__netif_subqueue_stopped(tx_ring->netdev,
1197 					     tx_ring->queue_index)
1198 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1199 			netif_wake_subqueue(tx_ring->netdev,
1200 					    tx_ring->queue_index);
1201 			++tx_ring->tx_stats.restart_queue;
1202 		}
1203 	}
1204 
1205 	return !!budget;
1206 }
1207 
1208 #ifdef CONFIG_IXGBE_DCA
1209 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1210 				struct ixgbe_ring *tx_ring,
1211 				int cpu)
1212 {
1213 	struct ixgbe_hw *hw = &adapter->hw;
1214 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
1215 	u16 reg_offset;
1216 
1217 	switch (hw->mac.type) {
1218 	case ixgbe_mac_82598EB:
1219 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1220 		break;
1221 	case ixgbe_mac_82599EB:
1222 	case ixgbe_mac_X540:
1223 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1224 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1225 		break;
1226 	default:
1227 		/* for unknown hardware do not write register */
1228 		return;
1229 	}
1230 
1231 	/*
1232 	 * We can enable relaxed ordering for reads, but not writes when
1233 	 * DCA is enabled.  This is due to a known issue in some chipsets
1234 	 * which will cause the DCA tag to be cleared.
1235 	 */
1236 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1237 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1238 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1239 
1240 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1241 }
1242 
1243 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1244 				struct ixgbe_ring *rx_ring,
1245 				int cpu)
1246 {
1247 	struct ixgbe_hw *hw = &adapter->hw;
1248 	u32 rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1249 	u8 reg_idx = rx_ring->reg_idx;
1250 
1251 
1252 	switch (hw->mac.type) {
1253 	case ixgbe_mac_82599EB:
1254 	case ixgbe_mac_X540:
1255 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1256 		break;
1257 	default:
1258 		break;
1259 	}
1260 
1261 	/*
1262 	 * We can enable relaxed ordering for reads, but not writes when
1263 	 * DCA is enabled.  This is due to a known issue in some chipsets
1264 	 * which will cause the DCA tag to be cleared.
1265 	 */
1266 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1267 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1268 
1269 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1270 }
1271 
1272 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1273 {
1274 	struct ixgbe_adapter *adapter = q_vector->adapter;
1275 	struct ixgbe_ring *ring;
1276 	int cpu = get_cpu();
1277 
1278 	if (q_vector->cpu == cpu)
1279 		goto out_no_update;
1280 
1281 	ixgbe_for_each_ring(ring, q_vector->tx)
1282 		ixgbe_update_tx_dca(adapter, ring, cpu);
1283 
1284 	ixgbe_for_each_ring(ring, q_vector->rx)
1285 		ixgbe_update_rx_dca(adapter, ring, cpu);
1286 
1287 	q_vector->cpu = cpu;
1288 out_no_update:
1289 	put_cpu();
1290 }
1291 
1292 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1293 {
1294 	int i;
1295 
1296 	if (!(adapter->flags & IXGBE_FLAG_DCA_ENABLED))
1297 		return;
1298 
1299 	/* always use CB2 mode, difference is masked in the CB driver */
1300 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 2);
1301 
1302 	for (i = 0; i < adapter->num_q_vectors; i++) {
1303 		adapter->q_vector[i]->cpu = -1;
1304 		ixgbe_update_dca(adapter->q_vector[i]);
1305 	}
1306 }
1307 
1308 static int __ixgbe_notify_dca(struct device *dev, void *data)
1309 {
1310 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1311 	unsigned long event = *(unsigned long *)data;
1312 
1313 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1314 		return 0;
1315 
1316 	switch (event) {
1317 	case DCA_PROVIDER_ADD:
1318 		/* if we're already enabled, don't do it again */
1319 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1320 			break;
1321 		if (dca_add_requester(dev) == 0) {
1322 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1323 			ixgbe_setup_dca(adapter);
1324 			break;
1325 		}
1326 		/* Fall Through since DCA is disabled. */
1327 	case DCA_PROVIDER_REMOVE:
1328 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1329 			dca_remove_requester(dev);
1330 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1331 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
1332 		}
1333 		break;
1334 	}
1335 
1336 	return 0;
1337 }
1338 
1339 #endif /* CONFIG_IXGBE_DCA */
1340 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1341 				 union ixgbe_adv_rx_desc *rx_desc,
1342 				 struct sk_buff *skb)
1343 {
1344 	if (ring->netdev->features & NETIF_F_RXHASH)
1345 		skb_set_hash(skb,
1346 			     le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1347 			     PKT_HASH_TYPE_L3);
1348 }
1349 
1350 #ifdef IXGBE_FCOE
1351 /**
1352  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1353  * @ring: structure containing ring specific data
1354  * @rx_desc: advanced rx descriptor
1355  *
1356  * Returns : true if it is FCoE pkt
1357  */
1358 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1359 				    union ixgbe_adv_rx_desc *rx_desc)
1360 {
1361 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1362 
1363 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1364 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1365 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1366 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1367 }
1368 
1369 #endif /* IXGBE_FCOE */
1370 /**
1371  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1372  * @ring: structure containing ring specific data
1373  * @rx_desc: current Rx descriptor being processed
1374  * @skb: skb currently being received and modified
1375  **/
1376 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1377 				     union ixgbe_adv_rx_desc *rx_desc,
1378 				     struct sk_buff *skb)
1379 {
1380 	skb_checksum_none_assert(skb);
1381 
1382 	/* Rx csum disabled */
1383 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1384 		return;
1385 
1386 	/* if IP and error */
1387 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1388 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1389 		ring->rx_stats.csum_err++;
1390 		return;
1391 	}
1392 
1393 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1394 		return;
1395 
1396 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1397 		__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1398 
1399 		/*
1400 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1401 		 * checksum errors.
1402 		 */
1403 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1404 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1405 			return;
1406 
1407 		ring->rx_stats.csum_err++;
1408 		return;
1409 	}
1410 
1411 	/* It must be a TCP or UDP packet with a valid checksum */
1412 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1413 }
1414 
1415 static inline void ixgbe_release_rx_desc(struct ixgbe_ring *rx_ring, u32 val)
1416 {
1417 	rx_ring->next_to_use = val;
1418 
1419 	/* update next to alloc since we have filled the ring */
1420 	rx_ring->next_to_alloc = val;
1421 	/*
1422 	 * Force memory writes to complete before letting h/w
1423 	 * know there are new descriptors to fetch.  (Only
1424 	 * applicable for weak-ordered memory model archs,
1425 	 * such as IA-64).
1426 	 */
1427 	wmb();
1428 	ixgbe_write_tail(rx_ring, val);
1429 }
1430 
1431 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1432 				    struct ixgbe_rx_buffer *bi)
1433 {
1434 	struct page *page = bi->page;
1435 	dma_addr_t dma = bi->dma;
1436 
1437 	/* since we are recycling buffers we should seldom need to alloc */
1438 	if (likely(dma))
1439 		return true;
1440 
1441 	/* alloc new page for storage */
1442 	if (likely(!page)) {
1443 		page = __skb_alloc_pages(GFP_ATOMIC | __GFP_COLD | __GFP_COMP,
1444 					 bi->skb, ixgbe_rx_pg_order(rx_ring));
1445 		if (unlikely(!page)) {
1446 			rx_ring->rx_stats.alloc_rx_page_failed++;
1447 			return false;
1448 		}
1449 		bi->page = page;
1450 	}
1451 
1452 	/* map page for use */
1453 	dma = dma_map_page(rx_ring->dev, page, 0,
1454 			   ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1455 
1456 	/*
1457 	 * if mapping failed free memory back to system since
1458 	 * there isn't much point in holding memory we can't use
1459 	 */
1460 	if (dma_mapping_error(rx_ring->dev, dma)) {
1461 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1462 		bi->page = NULL;
1463 
1464 		rx_ring->rx_stats.alloc_rx_page_failed++;
1465 		return false;
1466 	}
1467 
1468 	bi->dma = dma;
1469 	bi->page_offset = 0;
1470 
1471 	return true;
1472 }
1473 
1474 /**
1475  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1476  * @rx_ring: ring to place buffers on
1477  * @cleaned_count: number of buffers to replace
1478  **/
1479 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1480 {
1481 	union ixgbe_adv_rx_desc *rx_desc;
1482 	struct ixgbe_rx_buffer *bi;
1483 	u16 i = rx_ring->next_to_use;
1484 
1485 	/* nothing to do */
1486 	if (!cleaned_count)
1487 		return;
1488 
1489 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1490 	bi = &rx_ring->rx_buffer_info[i];
1491 	i -= rx_ring->count;
1492 
1493 	do {
1494 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1495 			break;
1496 
1497 		/*
1498 		 * Refresh the desc even if buffer_addrs didn't change
1499 		 * because each write-back erases this info.
1500 		 */
1501 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1502 
1503 		rx_desc++;
1504 		bi++;
1505 		i++;
1506 		if (unlikely(!i)) {
1507 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1508 			bi = rx_ring->rx_buffer_info;
1509 			i -= rx_ring->count;
1510 		}
1511 
1512 		/* clear the hdr_addr for the next_to_use descriptor */
1513 		rx_desc->read.hdr_addr = 0;
1514 
1515 		cleaned_count--;
1516 	} while (cleaned_count);
1517 
1518 	i += rx_ring->count;
1519 
1520 	if (rx_ring->next_to_use != i)
1521 		ixgbe_release_rx_desc(rx_ring, i);
1522 }
1523 
1524 /**
1525  * ixgbe_get_headlen - determine size of header for RSC/LRO/GRO/FCOE
1526  * @data: pointer to the start of the headers
1527  * @max_len: total length of section to find headers in
1528  *
1529  * This function is meant to determine the length of headers that will
1530  * be recognized by hardware for LRO, GRO, and RSC offloads.  The main
1531  * motivation of doing this is to only perform one pull for IPv4 TCP
1532  * packets so that we can do basic things like calculating the gso_size
1533  * based on the average data per packet.
1534  **/
1535 static unsigned int ixgbe_get_headlen(unsigned char *data,
1536 				      unsigned int max_len)
1537 {
1538 	union {
1539 		unsigned char *network;
1540 		/* l2 headers */
1541 		struct ethhdr *eth;
1542 		struct vlan_hdr *vlan;
1543 		/* l3 headers */
1544 		struct iphdr *ipv4;
1545 		struct ipv6hdr *ipv6;
1546 	} hdr;
1547 	__be16 protocol;
1548 	u8 nexthdr = 0;	/* default to not TCP */
1549 	u8 hlen;
1550 
1551 	/* this should never happen, but better safe than sorry */
1552 	if (max_len < ETH_HLEN)
1553 		return max_len;
1554 
1555 	/* initialize network frame pointer */
1556 	hdr.network = data;
1557 
1558 	/* set first protocol and move network header forward */
1559 	protocol = hdr.eth->h_proto;
1560 	hdr.network += ETH_HLEN;
1561 
1562 	/* handle any vlan tag if present */
1563 	if (protocol == htons(ETH_P_8021Q)) {
1564 		if ((hdr.network - data) > (max_len - VLAN_HLEN))
1565 			return max_len;
1566 
1567 		protocol = hdr.vlan->h_vlan_encapsulated_proto;
1568 		hdr.network += VLAN_HLEN;
1569 	}
1570 
1571 	/* handle L3 protocols */
1572 	if (protocol == htons(ETH_P_IP)) {
1573 		if ((hdr.network - data) > (max_len - sizeof(struct iphdr)))
1574 			return max_len;
1575 
1576 		/* access ihl as a u8 to avoid unaligned access on ia64 */
1577 		hlen = (hdr.network[0] & 0x0F) << 2;
1578 
1579 		/* verify hlen meets minimum size requirements */
1580 		if (hlen < sizeof(struct iphdr))
1581 			return hdr.network - data;
1582 
1583 		/* record next protocol if header is present */
1584 		if (!(hdr.ipv4->frag_off & htons(IP_OFFSET)))
1585 			nexthdr = hdr.ipv4->protocol;
1586 	} else if (protocol == htons(ETH_P_IPV6)) {
1587 		if ((hdr.network - data) > (max_len - sizeof(struct ipv6hdr)))
1588 			return max_len;
1589 
1590 		/* record next protocol */
1591 		nexthdr = hdr.ipv6->nexthdr;
1592 		hlen = sizeof(struct ipv6hdr);
1593 #ifdef IXGBE_FCOE
1594 	} else if (protocol == htons(ETH_P_FCOE)) {
1595 		if ((hdr.network - data) > (max_len - FCOE_HEADER_LEN))
1596 			return max_len;
1597 		hlen = FCOE_HEADER_LEN;
1598 #endif
1599 	} else {
1600 		return hdr.network - data;
1601 	}
1602 
1603 	/* relocate pointer to start of L4 header */
1604 	hdr.network += hlen;
1605 
1606 	/* finally sort out TCP/UDP */
1607 	if (nexthdr == IPPROTO_TCP) {
1608 		if ((hdr.network - data) > (max_len - sizeof(struct tcphdr)))
1609 			return max_len;
1610 
1611 		/* access doff as a u8 to avoid unaligned access on ia64 */
1612 		hlen = (hdr.network[12] & 0xF0) >> 2;
1613 
1614 		/* verify hlen meets minimum size requirements */
1615 		if (hlen < sizeof(struct tcphdr))
1616 			return hdr.network - data;
1617 
1618 		hdr.network += hlen;
1619 	} else if (nexthdr == IPPROTO_UDP) {
1620 		if ((hdr.network - data) > (max_len - sizeof(struct udphdr)))
1621 			return max_len;
1622 
1623 		hdr.network += sizeof(struct udphdr);
1624 	}
1625 
1626 	/*
1627 	 * If everything has gone correctly hdr.network should be the
1628 	 * data section of the packet and will be the end of the header.
1629 	 * If not then it probably represents the end of the last recognized
1630 	 * header.
1631 	 */
1632 	if ((hdr.network - data) < max_len)
1633 		return hdr.network - data;
1634 	else
1635 		return max_len;
1636 }
1637 
1638 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1639 				   struct sk_buff *skb)
1640 {
1641 	u16 hdr_len = skb_headlen(skb);
1642 
1643 	/* set gso_size to avoid messing up TCP MSS */
1644 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1645 						 IXGBE_CB(skb)->append_cnt);
1646 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1647 }
1648 
1649 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1650 				   struct sk_buff *skb)
1651 {
1652 	/* if append_cnt is 0 then frame is not RSC */
1653 	if (!IXGBE_CB(skb)->append_cnt)
1654 		return;
1655 
1656 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1657 	rx_ring->rx_stats.rsc_flush++;
1658 
1659 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1660 
1661 	/* gso_size is computed using append_cnt so always clear it last */
1662 	IXGBE_CB(skb)->append_cnt = 0;
1663 }
1664 
1665 /**
1666  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1667  * @rx_ring: rx descriptor ring packet is being transacted on
1668  * @rx_desc: pointer to the EOP Rx descriptor
1669  * @skb: pointer to current skb being populated
1670  *
1671  * This function checks the ring, descriptor, and packet information in
1672  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1673  * other fields within the skb.
1674  **/
1675 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1676 				     union ixgbe_adv_rx_desc *rx_desc,
1677 				     struct sk_buff *skb)
1678 {
1679 	struct net_device *dev = rx_ring->netdev;
1680 
1681 	ixgbe_update_rsc_stats(rx_ring, skb);
1682 
1683 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1684 
1685 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1686 
1687 	if (unlikely(ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_TS)))
1688 		ixgbe_ptp_rx_hwtstamp(rx_ring->q_vector->adapter, skb);
1689 
1690 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1691 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1692 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1693 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1694 	}
1695 
1696 	skb_record_rx_queue(skb, rx_ring->queue_index);
1697 
1698 	skb->protocol = eth_type_trans(skb, dev);
1699 }
1700 
1701 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1702 			 struct sk_buff *skb)
1703 {
1704 	struct ixgbe_adapter *adapter = q_vector->adapter;
1705 
1706 	if (ixgbe_qv_busy_polling(q_vector))
1707 		netif_receive_skb(skb);
1708 	else if (!(adapter->flags & IXGBE_FLAG_IN_NETPOLL))
1709 		napi_gro_receive(&q_vector->napi, skb);
1710 	else
1711 		netif_rx(skb);
1712 }
1713 
1714 /**
1715  * ixgbe_is_non_eop - process handling of non-EOP buffers
1716  * @rx_ring: Rx ring being processed
1717  * @rx_desc: Rx descriptor for current buffer
1718  * @skb: Current socket buffer containing buffer in progress
1719  *
1720  * This function updates next to clean.  If the buffer is an EOP buffer
1721  * this function exits returning false, otherwise it will place the
1722  * sk_buff in the next buffer to be chained and return true indicating
1723  * that this is in fact a non-EOP buffer.
1724  **/
1725 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1726 			     union ixgbe_adv_rx_desc *rx_desc,
1727 			     struct sk_buff *skb)
1728 {
1729 	u32 ntc = rx_ring->next_to_clean + 1;
1730 
1731 	/* fetch, update, and store next to clean */
1732 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1733 	rx_ring->next_to_clean = ntc;
1734 
1735 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1736 
1737 	/* update RSC append count if present */
1738 	if (ring_is_rsc_enabled(rx_ring)) {
1739 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1740 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1741 
1742 		if (unlikely(rsc_enabled)) {
1743 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1744 
1745 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1746 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1747 
1748 			/* update ntc based on RSC value */
1749 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1750 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1751 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1752 		}
1753 	}
1754 
1755 	/* if we are the last buffer then there is nothing else to do */
1756 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1757 		return false;
1758 
1759 	/* place skb in next buffer to be received */
1760 	rx_ring->rx_buffer_info[ntc].skb = skb;
1761 	rx_ring->rx_stats.non_eop_descs++;
1762 
1763 	return true;
1764 }
1765 
1766 /**
1767  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1768  * @rx_ring: rx descriptor ring packet is being transacted on
1769  * @skb: pointer to current skb being adjusted
1770  *
1771  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1772  * main difference between this version and the original function is that
1773  * this function can make several assumptions about the state of things
1774  * that allow for significant optimizations versus the standard function.
1775  * As a result we can do things like drop a frag and maintain an accurate
1776  * truesize for the skb.
1777  */
1778 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1779 			    struct sk_buff *skb)
1780 {
1781 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1782 	unsigned char *va;
1783 	unsigned int pull_len;
1784 
1785 	/*
1786 	 * it is valid to use page_address instead of kmap since we are
1787 	 * working with pages allocated out of the lomem pool per
1788 	 * alloc_page(GFP_ATOMIC)
1789 	 */
1790 	va = skb_frag_address(frag);
1791 
1792 	/*
1793 	 * we need the header to contain the greater of either ETH_HLEN or
1794 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1795 	 */
1796 	pull_len = ixgbe_get_headlen(va, IXGBE_RX_HDR_SIZE);
1797 
1798 	/* align pull length to size of long to optimize memcpy performance */
1799 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1800 
1801 	/* update all of the pointers */
1802 	skb_frag_size_sub(frag, pull_len);
1803 	frag->page_offset += pull_len;
1804 	skb->data_len -= pull_len;
1805 	skb->tail += pull_len;
1806 }
1807 
1808 /**
1809  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1810  * @rx_ring: rx descriptor ring packet is being transacted on
1811  * @skb: pointer to current skb being updated
1812  *
1813  * This function provides a basic DMA sync up for the first fragment of an
1814  * skb.  The reason for doing this is that the first fragment cannot be
1815  * unmapped until we have reached the end of packet descriptor for a buffer
1816  * chain.
1817  */
1818 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1819 				struct sk_buff *skb)
1820 {
1821 	/* if the page was released unmap it, else just sync our portion */
1822 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1823 		dma_unmap_page(rx_ring->dev, IXGBE_CB(skb)->dma,
1824 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1825 		IXGBE_CB(skb)->page_released = false;
1826 	} else {
1827 		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1828 
1829 		dma_sync_single_range_for_cpu(rx_ring->dev,
1830 					      IXGBE_CB(skb)->dma,
1831 					      frag->page_offset,
1832 					      ixgbe_rx_bufsz(rx_ring),
1833 					      DMA_FROM_DEVICE);
1834 	}
1835 	IXGBE_CB(skb)->dma = 0;
1836 }
1837 
1838 /**
1839  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1840  * @rx_ring: rx descriptor ring packet is being transacted on
1841  * @rx_desc: pointer to the EOP Rx descriptor
1842  * @skb: pointer to current skb being fixed
1843  *
1844  * Check for corrupted packet headers caused by senders on the local L2
1845  * embedded NIC switch not setting up their Tx Descriptors right.  These
1846  * should be very rare.
1847  *
1848  * Also address the case where we are pulling data in on pages only
1849  * and as such no data is present in the skb header.
1850  *
1851  * In addition if skb is not at least 60 bytes we need to pad it so that
1852  * it is large enough to qualify as a valid Ethernet frame.
1853  *
1854  * Returns true if an error was encountered and skb was freed.
1855  **/
1856 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1857 				  union ixgbe_adv_rx_desc *rx_desc,
1858 				  struct sk_buff *skb)
1859 {
1860 	struct net_device *netdev = rx_ring->netdev;
1861 
1862 	/* verify that the packet does not have any known errors */
1863 	if (unlikely(ixgbe_test_staterr(rx_desc,
1864 					IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1865 	    !(netdev->features & NETIF_F_RXALL))) {
1866 		dev_kfree_skb_any(skb);
1867 		return true;
1868 	}
1869 
1870 	/* place header in linear portion of buffer */
1871 	if (skb_is_nonlinear(skb))
1872 		ixgbe_pull_tail(rx_ring, skb);
1873 
1874 #ifdef IXGBE_FCOE
1875 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1876 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1877 		return false;
1878 
1879 #endif
1880 	/* if skb_pad returns an error the skb was freed */
1881 	if (unlikely(skb->len < 60)) {
1882 		int pad_len = 60 - skb->len;
1883 
1884 		if (skb_pad(skb, pad_len))
1885 			return true;
1886 		__skb_put(skb, pad_len);
1887 	}
1888 
1889 	return false;
1890 }
1891 
1892 /**
1893  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1894  * @rx_ring: rx descriptor ring to store buffers on
1895  * @old_buff: donor buffer to have page reused
1896  *
1897  * Synchronizes page for reuse by the adapter
1898  **/
1899 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1900 				struct ixgbe_rx_buffer *old_buff)
1901 {
1902 	struct ixgbe_rx_buffer *new_buff;
1903 	u16 nta = rx_ring->next_to_alloc;
1904 
1905 	new_buff = &rx_ring->rx_buffer_info[nta];
1906 
1907 	/* update, and store next to alloc */
1908 	nta++;
1909 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1910 
1911 	/* transfer page from old buffer to new buffer */
1912 	new_buff->page = old_buff->page;
1913 	new_buff->dma = old_buff->dma;
1914 	new_buff->page_offset = old_buff->page_offset;
1915 
1916 	/* sync the buffer for use by the device */
1917 	dma_sync_single_range_for_device(rx_ring->dev, new_buff->dma,
1918 					 new_buff->page_offset,
1919 					 ixgbe_rx_bufsz(rx_ring),
1920 					 DMA_FROM_DEVICE);
1921 }
1922 
1923 /**
1924  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1925  * @rx_ring: rx descriptor ring to transact packets on
1926  * @rx_buffer: buffer containing page to add
1927  * @rx_desc: descriptor containing length of buffer written by hardware
1928  * @skb: sk_buff to place the data into
1929  *
1930  * This function will add the data contained in rx_buffer->page to the skb.
1931  * This is done either through a direct copy if the data in the buffer is
1932  * less than the skb header size, otherwise it will just attach the page as
1933  * a frag to the skb.
1934  *
1935  * The function will then update the page offset if necessary and return
1936  * true if the buffer can be reused by the adapter.
1937  **/
1938 static bool ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1939 			      struct ixgbe_rx_buffer *rx_buffer,
1940 			      union ixgbe_adv_rx_desc *rx_desc,
1941 			      struct sk_buff *skb)
1942 {
1943 	struct page *page = rx_buffer->page;
1944 	unsigned int size = le16_to_cpu(rx_desc->wb.upper.length);
1945 #if (PAGE_SIZE < 8192)
1946 	unsigned int truesize = ixgbe_rx_bufsz(rx_ring);
1947 #else
1948 	unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
1949 	unsigned int last_offset = ixgbe_rx_pg_size(rx_ring) -
1950 				   ixgbe_rx_bufsz(rx_ring);
1951 #endif
1952 
1953 	if ((size <= IXGBE_RX_HDR_SIZE) && !skb_is_nonlinear(skb)) {
1954 		unsigned char *va = page_address(page) + rx_buffer->page_offset;
1955 
1956 		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
1957 
1958 		/* we can reuse buffer as-is, just make sure it is local */
1959 		if (likely(page_to_nid(page) == numa_node_id()))
1960 			return true;
1961 
1962 		/* this page cannot be reused so discard it */
1963 		put_page(page);
1964 		return false;
1965 	}
1966 
1967 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
1968 			rx_buffer->page_offset, size, truesize);
1969 
1970 	/* avoid re-using remote pages */
1971 	if (unlikely(page_to_nid(page) != numa_node_id()))
1972 		return false;
1973 
1974 #if (PAGE_SIZE < 8192)
1975 	/* if we are only owner of page we can reuse it */
1976 	if (unlikely(page_count(page) != 1))
1977 		return false;
1978 
1979 	/* flip page offset to other buffer */
1980 	rx_buffer->page_offset ^= truesize;
1981 
1982 	/*
1983 	 * since we are the only owner of the page and we need to
1984 	 * increment it, just set the value to 2 in order to avoid
1985 	 * an unecessary locked operation
1986 	 */
1987 	atomic_set(&page->_count, 2);
1988 #else
1989 	/* move offset up to the next cache line */
1990 	rx_buffer->page_offset += truesize;
1991 
1992 	if (rx_buffer->page_offset > last_offset)
1993 		return false;
1994 
1995 	/* bump ref count on page before it is given to the stack */
1996 	get_page(page);
1997 #endif
1998 
1999 	return true;
2000 }
2001 
2002 static struct sk_buff *ixgbe_fetch_rx_buffer(struct ixgbe_ring *rx_ring,
2003 					     union ixgbe_adv_rx_desc *rx_desc)
2004 {
2005 	struct ixgbe_rx_buffer *rx_buffer;
2006 	struct sk_buff *skb;
2007 	struct page *page;
2008 
2009 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
2010 	page = rx_buffer->page;
2011 	prefetchw(page);
2012 
2013 	skb = rx_buffer->skb;
2014 
2015 	if (likely(!skb)) {
2016 		void *page_addr = page_address(page) +
2017 				  rx_buffer->page_offset;
2018 
2019 		/* prefetch first cache line of first page */
2020 		prefetch(page_addr);
2021 #if L1_CACHE_BYTES < 128
2022 		prefetch(page_addr + L1_CACHE_BYTES);
2023 #endif
2024 
2025 		/* allocate a skb to store the frags */
2026 		skb = netdev_alloc_skb_ip_align(rx_ring->netdev,
2027 						IXGBE_RX_HDR_SIZE);
2028 		if (unlikely(!skb)) {
2029 			rx_ring->rx_stats.alloc_rx_buff_failed++;
2030 			return NULL;
2031 		}
2032 
2033 		/*
2034 		 * we will be copying header into skb->data in
2035 		 * pskb_may_pull so it is in our interest to prefetch
2036 		 * it now to avoid a possible cache miss
2037 		 */
2038 		prefetchw(skb->data);
2039 
2040 		/*
2041 		 * Delay unmapping of the first packet. It carries the
2042 		 * header information, HW may still access the header
2043 		 * after the writeback.  Only unmap it when EOP is
2044 		 * reached
2045 		 */
2046 		if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
2047 			goto dma_sync;
2048 
2049 		IXGBE_CB(skb)->dma = rx_buffer->dma;
2050 	} else {
2051 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2052 			ixgbe_dma_sync_frag(rx_ring, skb);
2053 
2054 dma_sync:
2055 		/* we are reusing so sync this buffer for CPU use */
2056 		dma_sync_single_range_for_cpu(rx_ring->dev,
2057 					      rx_buffer->dma,
2058 					      rx_buffer->page_offset,
2059 					      ixgbe_rx_bufsz(rx_ring),
2060 					      DMA_FROM_DEVICE);
2061 	}
2062 
2063 	/* pull page into skb */
2064 	if (ixgbe_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) {
2065 		/* hand second half of page back to the ring */
2066 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2067 	} else if (IXGBE_CB(skb)->dma == rx_buffer->dma) {
2068 		/* the page has been released from the ring */
2069 		IXGBE_CB(skb)->page_released = true;
2070 	} else {
2071 		/* we are not reusing the buffer so unmap it */
2072 		dma_unmap_page(rx_ring->dev, rx_buffer->dma,
2073 			       ixgbe_rx_pg_size(rx_ring),
2074 			       DMA_FROM_DEVICE);
2075 	}
2076 
2077 	/* clear contents of buffer_info */
2078 	rx_buffer->skb = NULL;
2079 	rx_buffer->dma = 0;
2080 	rx_buffer->page = NULL;
2081 
2082 	return skb;
2083 }
2084 
2085 /**
2086  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2087  * @q_vector: structure containing interrupt and ring information
2088  * @rx_ring: rx descriptor ring to transact packets on
2089  * @budget: Total limit on number of packets to process
2090  *
2091  * This function provides a "bounce buffer" approach to Rx interrupt
2092  * processing.  The advantage to this is that on systems that have
2093  * expensive overhead for IOMMU access this provides a means of avoiding
2094  * it by maintaining the mapping of the page to the syste.
2095  *
2096  * Returns amount of work completed
2097  **/
2098 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2099 			       struct ixgbe_ring *rx_ring,
2100 			       const int budget)
2101 {
2102 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
2103 #ifdef IXGBE_FCOE
2104 	struct ixgbe_adapter *adapter = q_vector->adapter;
2105 	int ddp_bytes;
2106 	unsigned int mss = 0;
2107 #endif /* IXGBE_FCOE */
2108 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2109 
2110 	while (likely(total_rx_packets < budget)) {
2111 		union ixgbe_adv_rx_desc *rx_desc;
2112 		struct sk_buff *skb;
2113 
2114 		/* return some buffers to hardware, one at a time is too slow */
2115 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2116 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2117 			cleaned_count = 0;
2118 		}
2119 
2120 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2121 
2122 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_DD))
2123 			break;
2124 
2125 		/*
2126 		 * This memory barrier is needed to keep us from reading
2127 		 * any other fields out of the rx_desc until we know the
2128 		 * RXD_STAT_DD bit is set
2129 		 */
2130 		rmb();
2131 
2132 		/* retrieve a buffer from the ring */
2133 		skb = ixgbe_fetch_rx_buffer(rx_ring, rx_desc);
2134 
2135 		/* exit if we failed to retrieve a buffer */
2136 		if (!skb)
2137 			break;
2138 
2139 		cleaned_count++;
2140 
2141 		/* place incomplete frames back on ring for completion */
2142 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2143 			continue;
2144 
2145 		/* verify the packet layout is correct */
2146 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2147 			continue;
2148 
2149 		/* probably a little skewed due to removing CRC */
2150 		total_rx_bytes += skb->len;
2151 
2152 		/* populate checksum, timestamp, VLAN, and protocol */
2153 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2154 
2155 #ifdef IXGBE_FCOE
2156 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2157 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2158 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2159 			/* include DDPed FCoE data */
2160 			if (ddp_bytes > 0) {
2161 				if (!mss) {
2162 					mss = rx_ring->netdev->mtu -
2163 						sizeof(struct fcoe_hdr) -
2164 						sizeof(struct fc_frame_header) -
2165 						sizeof(struct fcoe_crc_eof);
2166 					if (mss > 512)
2167 						mss &= ~511;
2168 				}
2169 				total_rx_bytes += ddp_bytes;
2170 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2171 								 mss);
2172 			}
2173 			if (!ddp_bytes) {
2174 				dev_kfree_skb_any(skb);
2175 				continue;
2176 			}
2177 		}
2178 
2179 #endif /* IXGBE_FCOE */
2180 		skb_mark_napi_id(skb, &q_vector->napi);
2181 		ixgbe_rx_skb(q_vector, skb);
2182 
2183 		/* update budget accounting */
2184 		total_rx_packets++;
2185 	}
2186 
2187 	u64_stats_update_begin(&rx_ring->syncp);
2188 	rx_ring->stats.packets += total_rx_packets;
2189 	rx_ring->stats.bytes += total_rx_bytes;
2190 	u64_stats_update_end(&rx_ring->syncp);
2191 	q_vector->rx.total_packets += total_rx_packets;
2192 	q_vector->rx.total_bytes += total_rx_bytes;
2193 
2194 	if (cleaned_count)
2195 		ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2196 
2197 	return total_rx_packets;
2198 }
2199 
2200 #ifdef CONFIG_NET_RX_BUSY_POLL
2201 /* must be called with local_bh_disable()d */
2202 static int ixgbe_low_latency_recv(struct napi_struct *napi)
2203 {
2204 	struct ixgbe_q_vector *q_vector =
2205 			container_of(napi, struct ixgbe_q_vector, napi);
2206 	struct ixgbe_adapter *adapter = q_vector->adapter;
2207 	struct ixgbe_ring  *ring;
2208 	int found = 0;
2209 
2210 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2211 		return LL_FLUSH_FAILED;
2212 
2213 	if (!ixgbe_qv_lock_poll(q_vector))
2214 		return LL_FLUSH_BUSY;
2215 
2216 	ixgbe_for_each_ring(ring, q_vector->rx) {
2217 		found = ixgbe_clean_rx_irq(q_vector, ring, 4);
2218 #ifdef BP_EXTENDED_STATS
2219 		if (found)
2220 			ring->stats.cleaned += found;
2221 		else
2222 			ring->stats.misses++;
2223 #endif
2224 		if (found)
2225 			break;
2226 	}
2227 
2228 	ixgbe_qv_unlock_poll(q_vector);
2229 
2230 	return found;
2231 }
2232 #endif	/* CONFIG_NET_RX_BUSY_POLL */
2233 
2234 /**
2235  * ixgbe_configure_msix - Configure MSI-X hardware
2236  * @adapter: board private structure
2237  *
2238  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2239  * interrupts.
2240  **/
2241 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2242 {
2243 	struct ixgbe_q_vector *q_vector;
2244 	int v_idx;
2245 	u32 mask;
2246 
2247 	/* Populate MSIX to EITR Select */
2248 	if (adapter->num_vfs > 32) {
2249 		u32 eitrsel = (1 << (adapter->num_vfs - 32)) - 1;
2250 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2251 	}
2252 
2253 	/*
2254 	 * Populate the IVAR table and set the ITR values to the
2255 	 * corresponding register.
2256 	 */
2257 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2258 		struct ixgbe_ring *ring;
2259 		q_vector = adapter->q_vector[v_idx];
2260 
2261 		ixgbe_for_each_ring(ring, q_vector->rx)
2262 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2263 
2264 		ixgbe_for_each_ring(ring, q_vector->tx)
2265 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2266 
2267 		ixgbe_write_eitr(q_vector);
2268 	}
2269 
2270 	switch (adapter->hw.mac.type) {
2271 	case ixgbe_mac_82598EB:
2272 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2273 			       v_idx);
2274 		break;
2275 	case ixgbe_mac_82599EB:
2276 	case ixgbe_mac_X540:
2277 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2278 		break;
2279 	default:
2280 		break;
2281 	}
2282 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2283 
2284 	/* set up to autoclear timer, and the vectors */
2285 	mask = IXGBE_EIMS_ENABLE_MASK;
2286 	mask &= ~(IXGBE_EIMS_OTHER |
2287 		  IXGBE_EIMS_MAILBOX |
2288 		  IXGBE_EIMS_LSC);
2289 
2290 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2291 }
2292 
2293 enum latency_range {
2294 	lowest_latency = 0,
2295 	low_latency = 1,
2296 	bulk_latency = 2,
2297 	latency_invalid = 255
2298 };
2299 
2300 /**
2301  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2302  * @q_vector: structure containing interrupt and ring information
2303  * @ring_container: structure containing ring performance data
2304  *
2305  *      Stores a new ITR value based on packets and byte
2306  *      counts during the last interrupt.  The advantage of per interrupt
2307  *      computation is faster updates and more accurate ITR for the current
2308  *      traffic pattern.  Constants in this function were computed
2309  *      based on theoretical maximum wire speed and thresholds were set based
2310  *      on testing data as well as attempting to minimize response time
2311  *      while increasing bulk throughput.
2312  *      this functionality is controlled by the InterruptThrottleRate module
2313  *      parameter (see ixgbe_param.c)
2314  **/
2315 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2316 			     struct ixgbe_ring_container *ring_container)
2317 {
2318 	int bytes = ring_container->total_bytes;
2319 	int packets = ring_container->total_packets;
2320 	u32 timepassed_us;
2321 	u64 bytes_perint;
2322 	u8 itr_setting = ring_container->itr;
2323 
2324 	if (packets == 0)
2325 		return;
2326 
2327 	/* simple throttlerate management
2328 	 *   0-10MB/s   lowest (100000 ints/s)
2329 	 *  10-20MB/s   low    (20000 ints/s)
2330 	 *  20-1249MB/s bulk   (8000 ints/s)
2331 	 */
2332 	/* what was last interrupt timeslice? */
2333 	timepassed_us = q_vector->itr >> 2;
2334 	if (timepassed_us == 0)
2335 		return;
2336 
2337 	bytes_perint = bytes / timepassed_us; /* bytes/usec */
2338 
2339 	switch (itr_setting) {
2340 	case lowest_latency:
2341 		if (bytes_perint > 10)
2342 			itr_setting = low_latency;
2343 		break;
2344 	case low_latency:
2345 		if (bytes_perint > 20)
2346 			itr_setting = bulk_latency;
2347 		else if (bytes_perint <= 10)
2348 			itr_setting = lowest_latency;
2349 		break;
2350 	case bulk_latency:
2351 		if (bytes_perint <= 20)
2352 			itr_setting = low_latency;
2353 		break;
2354 	}
2355 
2356 	/* clear work counters since we have the values we need */
2357 	ring_container->total_bytes = 0;
2358 	ring_container->total_packets = 0;
2359 
2360 	/* write updated itr to ring container */
2361 	ring_container->itr = itr_setting;
2362 }
2363 
2364 /**
2365  * ixgbe_write_eitr - write EITR register in hardware specific way
2366  * @q_vector: structure containing interrupt and ring information
2367  *
2368  * This function is made to be called by ethtool and by the driver
2369  * when it needs to update EITR registers at runtime.  Hardware
2370  * specific quirks/differences are taken care of here.
2371  */
2372 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2373 {
2374 	struct ixgbe_adapter *adapter = q_vector->adapter;
2375 	struct ixgbe_hw *hw = &adapter->hw;
2376 	int v_idx = q_vector->v_idx;
2377 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2378 
2379 	switch (adapter->hw.mac.type) {
2380 	case ixgbe_mac_82598EB:
2381 		/* must write high and low 16 bits to reset counter */
2382 		itr_reg |= (itr_reg << 16);
2383 		break;
2384 	case ixgbe_mac_82599EB:
2385 	case ixgbe_mac_X540:
2386 		/*
2387 		 * set the WDIS bit to not clear the timer bits and cause an
2388 		 * immediate assertion of the interrupt
2389 		 */
2390 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2391 		break;
2392 	default:
2393 		break;
2394 	}
2395 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2396 }
2397 
2398 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2399 {
2400 	u32 new_itr = q_vector->itr;
2401 	u8 current_itr;
2402 
2403 	ixgbe_update_itr(q_vector, &q_vector->tx);
2404 	ixgbe_update_itr(q_vector, &q_vector->rx);
2405 
2406 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
2407 
2408 	switch (current_itr) {
2409 	/* counts and packets in update_itr are dependent on these numbers */
2410 	case lowest_latency:
2411 		new_itr = IXGBE_100K_ITR;
2412 		break;
2413 	case low_latency:
2414 		new_itr = IXGBE_20K_ITR;
2415 		break;
2416 	case bulk_latency:
2417 		new_itr = IXGBE_8K_ITR;
2418 		break;
2419 	default:
2420 		break;
2421 	}
2422 
2423 	if (new_itr != q_vector->itr) {
2424 		/* do an exponential smoothing */
2425 		new_itr = (10 * new_itr * q_vector->itr) /
2426 			  ((9 * new_itr) + q_vector->itr);
2427 
2428 		/* save the algorithm value here */
2429 		q_vector->itr = new_itr;
2430 
2431 		ixgbe_write_eitr(q_vector);
2432 	}
2433 }
2434 
2435 /**
2436  * ixgbe_check_overtemp_subtask - check for over temperature
2437  * @adapter: pointer to adapter
2438  **/
2439 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2440 {
2441 	struct ixgbe_hw *hw = &adapter->hw;
2442 	u32 eicr = adapter->interrupt_event;
2443 
2444 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2445 		return;
2446 
2447 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) &&
2448 	    !(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2449 		return;
2450 
2451 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2452 
2453 	switch (hw->device_id) {
2454 	case IXGBE_DEV_ID_82599_T3_LOM:
2455 		/*
2456 		 * Since the warning interrupt is for both ports
2457 		 * we don't have to check if:
2458 		 *  - This interrupt wasn't for our port.
2459 		 *  - We may have missed the interrupt so always have to
2460 		 *    check if we  got a LSC
2461 		 */
2462 		if (!(eicr & IXGBE_EICR_GPI_SDP0) &&
2463 		    !(eicr & IXGBE_EICR_LSC))
2464 			return;
2465 
2466 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2467 			u32 speed;
2468 			bool link_up = false;
2469 
2470 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2471 
2472 			if (link_up)
2473 				return;
2474 		}
2475 
2476 		/* Check if this is not due to overtemp */
2477 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2478 			return;
2479 
2480 		break;
2481 	default:
2482 		if (!(eicr & IXGBE_EICR_GPI_SDP0))
2483 			return;
2484 		break;
2485 	}
2486 	e_crit(drv,
2487 	       "Network adapter has been stopped because it has over heated. "
2488 	       "Restart the computer. If the problem persists, "
2489 	       "power off the system and replace the adapter\n");
2490 
2491 	adapter->interrupt_event = 0;
2492 }
2493 
2494 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2495 {
2496 	struct ixgbe_hw *hw = &adapter->hw;
2497 
2498 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2499 	    (eicr & IXGBE_EICR_GPI_SDP1)) {
2500 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2501 		/* write to clear the interrupt */
2502 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2503 	}
2504 }
2505 
2506 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2507 {
2508 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2509 		return;
2510 
2511 	switch (adapter->hw.mac.type) {
2512 	case ixgbe_mac_82599EB:
2513 		/*
2514 		 * Need to check link state so complete overtemp check
2515 		 * on service task
2516 		 */
2517 		if (((eicr & IXGBE_EICR_GPI_SDP0) || (eicr & IXGBE_EICR_LSC)) &&
2518 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2519 			adapter->interrupt_event = eicr;
2520 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2521 			ixgbe_service_event_schedule(adapter);
2522 			return;
2523 		}
2524 		return;
2525 	case ixgbe_mac_X540:
2526 		if (!(eicr & IXGBE_EICR_TS))
2527 			return;
2528 		break;
2529 	default:
2530 		return;
2531 	}
2532 
2533 	e_crit(drv,
2534 	       "Network adapter has been stopped because it has over heated. "
2535 	       "Restart the computer. If the problem persists, "
2536 	       "power off the system and replace the adapter\n");
2537 }
2538 
2539 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2540 {
2541 	struct ixgbe_hw *hw = &adapter->hw;
2542 
2543 	if (eicr & IXGBE_EICR_GPI_SDP2) {
2544 		/* Clear the interrupt */
2545 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP2);
2546 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2547 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2548 			ixgbe_service_event_schedule(adapter);
2549 		}
2550 	}
2551 
2552 	if (eicr & IXGBE_EICR_GPI_SDP1) {
2553 		/* Clear the interrupt */
2554 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2555 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2556 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2557 			ixgbe_service_event_schedule(adapter);
2558 		}
2559 	}
2560 }
2561 
2562 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2563 {
2564 	struct ixgbe_hw *hw = &adapter->hw;
2565 
2566 	adapter->lsc_int++;
2567 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2568 	adapter->link_check_timeout = jiffies;
2569 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2570 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2571 		IXGBE_WRITE_FLUSH(hw);
2572 		ixgbe_service_event_schedule(adapter);
2573 	}
2574 }
2575 
2576 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2577 					   u64 qmask)
2578 {
2579 	u32 mask;
2580 	struct ixgbe_hw *hw = &adapter->hw;
2581 
2582 	switch (hw->mac.type) {
2583 	case ixgbe_mac_82598EB:
2584 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2585 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2586 		break;
2587 	case ixgbe_mac_82599EB:
2588 	case ixgbe_mac_X540:
2589 		mask = (qmask & 0xFFFFFFFF);
2590 		if (mask)
2591 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2592 		mask = (qmask >> 32);
2593 		if (mask)
2594 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2595 		break;
2596 	default:
2597 		break;
2598 	}
2599 	/* skip the flush */
2600 }
2601 
2602 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2603 					    u64 qmask)
2604 {
2605 	u32 mask;
2606 	struct ixgbe_hw *hw = &adapter->hw;
2607 
2608 	switch (hw->mac.type) {
2609 	case ixgbe_mac_82598EB:
2610 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2611 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2612 		break;
2613 	case ixgbe_mac_82599EB:
2614 	case ixgbe_mac_X540:
2615 		mask = (qmask & 0xFFFFFFFF);
2616 		if (mask)
2617 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2618 		mask = (qmask >> 32);
2619 		if (mask)
2620 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2621 		break;
2622 	default:
2623 		break;
2624 	}
2625 	/* skip the flush */
2626 }
2627 
2628 /**
2629  * ixgbe_irq_enable - Enable default interrupt generation settings
2630  * @adapter: board private structure
2631  **/
2632 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2633 				    bool flush)
2634 {
2635 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2636 
2637 	/* don't reenable LSC while waiting for link */
2638 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2639 		mask &= ~IXGBE_EIMS_LSC;
2640 
2641 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2642 		switch (adapter->hw.mac.type) {
2643 		case ixgbe_mac_82599EB:
2644 			mask |= IXGBE_EIMS_GPI_SDP0;
2645 			break;
2646 		case ixgbe_mac_X540:
2647 			mask |= IXGBE_EIMS_TS;
2648 			break;
2649 		default:
2650 			break;
2651 		}
2652 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2653 		mask |= IXGBE_EIMS_GPI_SDP1;
2654 	switch (adapter->hw.mac.type) {
2655 	case ixgbe_mac_82599EB:
2656 		mask |= IXGBE_EIMS_GPI_SDP1;
2657 		mask |= IXGBE_EIMS_GPI_SDP2;
2658 	case ixgbe_mac_X540:
2659 		mask |= IXGBE_EIMS_ECC;
2660 		mask |= IXGBE_EIMS_MAILBOX;
2661 		break;
2662 	default:
2663 		break;
2664 	}
2665 
2666 	if (adapter->hw.mac.type == ixgbe_mac_X540)
2667 		mask |= IXGBE_EIMS_TIMESYNC;
2668 
2669 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2670 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
2671 		mask |= IXGBE_EIMS_FLOW_DIR;
2672 
2673 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
2674 	if (queues)
2675 		ixgbe_irq_enable_queues(adapter, ~0);
2676 	if (flush)
2677 		IXGBE_WRITE_FLUSH(&adapter->hw);
2678 }
2679 
2680 static irqreturn_t ixgbe_msix_other(int irq, void *data)
2681 {
2682 	struct ixgbe_adapter *adapter = data;
2683 	struct ixgbe_hw *hw = &adapter->hw;
2684 	u32 eicr;
2685 
2686 	/*
2687 	 * Workaround for Silicon errata.  Use clear-by-write instead
2688 	 * of clear-by-read.  Reading with EICS will return the
2689 	 * interrupt causes without clearing, which later be done
2690 	 * with the write to EICR.
2691 	 */
2692 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2693 
2694 	/* The lower 16bits of the EICR register are for the queue interrupts
2695 	 * which should be masked here in order to not accidently clear them if
2696 	 * the bits are high when ixgbe_msix_other is called. There is a race
2697 	 * condition otherwise which results in possible performance loss
2698 	 * especially if the ixgbe_msix_other interrupt is triggering
2699 	 * consistently (as it would when PPS is turned on for the X540 device)
2700 	 */
2701 	eicr &= 0xFFFF0000;
2702 
2703 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2704 
2705 	if (eicr & IXGBE_EICR_LSC)
2706 		ixgbe_check_lsc(adapter);
2707 
2708 	if (eicr & IXGBE_EICR_MAILBOX)
2709 		ixgbe_msg_task(adapter);
2710 
2711 	switch (hw->mac.type) {
2712 	case ixgbe_mac_82599EB:
2713 	case ixgbe_mac_X540:
2714 		if (eicr & IXGBE_EICR_ECC) {
2715 			e_info(link, "Received ECC Err, initiating reset\n");
2716 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
2717 			ixgbe_service_event_schedule(adapter);
2718 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2719 		}
2720 		/* Handle Flow Director Full threshold interrupt */
2721 		if (eicr & IXGBE_EICR_FLOW_DIR) {
2722 			int reinit_count = 0;
2723 			int i;
2724 			for (i = 0; i < adapter->num_tx_queues; i++) {
2725 				struct ixgbe_ring *ring = adapter->tx_ring[i];
2726 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
2727 						       &ring->state))
2728 					reinit_count++;
2729 			}
2730 			if (reinit_count) {
2731 				/* no more flow director interrupts until after init */
2732 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
2733 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
2734 				ixgbe_service_event_schedule(adapter);
2735 			}
2736 		}
2737 		ixgbe_check_sfp_event(adapter, eicr);
2738 		ixgbe_check_overtemp_event(adapter, eicr);
2739 		break;
2740 	default:
2741 		break;
2742 	}
2743 
2744 	ixgbe_check_fan_failure(adapter, eicr);
2745 
2746 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2747 		ixgbe_ptp_check_pps_event(adapter, eicr);
2748 
2749 	/* re-enable the original interrupt state, no lsc, no queues */
2750 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2751 		ixgbe_irq_enable(adapter, false, false);
2752 
2753 	return IRQ_HANDLED;
2754 }
2755 
2756 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
2757 {
2758 	struct ixgbe_q_vector *q_vector = data;
2759 
2760 	/* EIAM disabled interrupts (on this vector) for us */
2761 
2762 	if (q_vector->rx.ring || q_vector->tx.ring)
2763 		napi_schedule(&q_vector->napi);
2764 
2765 	return IRQ_HANDLED;
2766 }
2767 
2768 /**
2769  * ixgbe_poll - NAPI Rx polling callback
2770  * @napi: structure for representing this polling device
2771  * @budget: how many packets driver is allowed to clean
2772  *
2773  * This function is used for legacy and MSI, NAPI mode
2774  **/
2775 int ixgbe_poll(struct napi_struct *napi, int budget)
2776 {
2777 	struct ixgbe_q_vector *q_vector =
2778 				container_of(napi, struct ixgbe_q_vector, napi);
2779 	struct ixgbe_adapter *adapter = q_vector->adapter;
2780 	struct ixgbe_ring *ring;
2781 	int per_ring_budget;
2782 	bool clean_complete = true;
2783 
2784 #ifdef CONFIG_IXGBE_DCA
2785 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
2786 		ixgbe_update_dca(q_vector);
2787 #endif
2788 
2789 	ixgbe_for_each_ring(ring, q_vector->tx)
2790 		clean_complete &= !!ixgbe_clean_tx_irq(q_vector, ring);
2791 
2792 	if (!ixgbe_qv_lock_napi(q_vector))
2793 		return budget;
2794 
2795 	/* attempt to distribute budget to each queue fairly, but don't allow
2796 	 * the budget to go below 1 because we'll exit polling */
2797 	if (q_vector->rx.count > 1)
2798 		per_ring_budget = max(budget/q_vector->rx.count, 1);
2799 	else
2800 		per_ring_budget = budget;
2801 
2802 	ixgbe_for_each_ring(ring, q_vector->rx)
2803 		clean_complete &= (ixgbe_clean_rx_irq(q_vector, ring,
2804 				   per_ring_budget) < per_ring_budget);
2805 
2806 	ixgbe_qv_unlock_napi(q_vector);
2807 	/* If all work not completed, return budget and keep polling */
2808 	if (!clean_complete)
2809 		return budget;
2810 
2811 	/* all work done, exit the polling mode */
2812 	napi_complete(napi);
2813 	if (adapter->rx_itr_setting & 1)
2814 		ixgbe_set_itr(q_vector);
2815 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2816 		ixgbe_irq_enable_queues(adapter, ((u64)1 << q_vector->v_idx));
2817 
2818 	return 0;
2819 }
2820 
2821 /**
2822  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2823  * @adapter: board private structure
2824  *
2825  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2826  * interrupts from the kernel.
2827  **/
2828 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
2829 {
2830 	struct net_device *netdev = adapter->netdev;
2831 	int vector, err;
2832 	int ri = 0, ti = 0;
2833 
2834 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2835 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2836 		struct msix_entry *entry = &adapter->msix_entries[vector];
2837 
2838 		if (q_vector->tx.ring && q_vector->rx.ring) {
2839 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2840 				 "%s-%s-%d", netdev->name, "TxRx", ri++);
2841 			ti++;
2842 		} else if (q_vector->rx.ring) {
2843 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2844 				 "%s-%s-%d", netdev->name, "rx", ri++);
2845 		} else if (q_vector->tx.ring) {
2846 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2847 				 "%s-%s-%d", netdev->name, "tx", ti++);
2848 		} else {
2849 			/* skip this unused q_vector */
2850 			continue;
2851 		}
2852 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
2853 				  q_vector->name, q_vector);
2854 		if (err) {
2855 			e_err(probe, "request_irq failed for MSIX interrupt "
2856 			      "Error: %d\n", err);
2857 			goto free_queue_irqs;
2858 		}
2859 		/* If Flow Director is enabled, set interrupt affinity */
2860 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2861 			/* assign the mask for this irq */
2862 			irq_set_affinity_hint(entry->vector,
2863 					      &q_vector->affinity_mask);
2864 		}
2865 	}
2866 
2867 	err = request_irq(adapter->msix_entries[vector].vector,
2868 			  ixgbe_msix_other, 0, netdev->name, adapter);
2869 	if (err) {
2870 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
2871 		goto free_queue_irqs;
2872 	}
2873 
2874 	return 0;
2875 
2876 free_queue_irqs:
2877 	while (vector) {
2878 		vector--;
2879 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
2880 				      NULL);
2881 		free_irq(adapter->msix_entries[vector].vector,
2882 			 adapter->q_vector[vector]);
2883 	}
2884 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
2885 	pci_disable_msix(adapter->pdev);
2886 	kfree(adapter->msix_entries);
2887 	adapter->msix_entries = NULL;
2888 	return err;
2889 }
2890 
2891 /**
2892  * ixgbe_intr - legacy mode Interrupt Handler
2893  * @irq: interrupt number
2894  * @data: pointer to a network interface device structure
2895  **/
2896 static irqreturn_t ixgbe_intr(int irq, void *data)
2897 {
2898 	struct ixgbe_adapter *adapter = data;
2899 	struct ixgbe_hw *hw = &adapter->hw;
2900 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2901 	u32 eicr;
2902 
2903 	/*
2904 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
2905 	 * before the read of EICR.
2906 	 */
2907 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
2908 
2909 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2910 	 * therefore no explicit interrupt disable is necessary */
2911 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2912 	if (!eicr) {
2913 		/*
2914 		 * shared interrupt alert!
2915 		 * make sure interrupts are enabled because the read will
2916 		 * have disabled interrupts due to EIAM
2917 		 * finish the workaround of silicon errata on 82598.  Unmask
2918 		 * the interrupt that we masked before the EICR read.
2919 		 */
2920 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
2921 			ixgbe_irq_enable(adapter, true, true);
2922 		return IRQ_NONE;	/* Not our interrupt */
2923 	}
2924 
2925 	if (eicr & IXGBE_EICR_LSC)
2926 		ixgbe_check_lsc(adapter);
2927 
2928 	switch (hw->mac.type) {
2929 	case ixgbe_mac_82599EB:
2930 		ixgbe_check_sfp_event(adapter, eicr);
2931 		/* Fall through */
2932 	case ixgbe_mac_X540:
2933 		if (eicr & IXGBE_EICR_ECC) {
2934 			e_info(link, "Received ECC Err, initiating reset\n");
2935 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
2936 			ixgbe_service_event_schedule(adapter);
2937 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2938 		}
2939 		ixgbe_check_overtemp_event(adapter, eicr);
2940 		break;
2941 	default:
2942 		break;
2943 	}
2944 
2945 	ixgbe_check_fan_failure(adapter, eicr);
2946 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2947 		ixgbe_ptp_check_pps_event(adapter, eicr);
2948 
2949 	/* would disable interrupts here but EIAM disabled it */
2950 	napi_schedule(&q_vector->napi);
2951 
2952 	/*
2953 	 * re-enable link(maybe) and non-queue interrupts, no flush.
2954 	 * ixgbe_poll will re-enable the queue interrupts
2955 	 */
2956 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2957 		ixgbe_irq_enable(adapter, false, false);
2958 
2959 	return IRQ_HANDLED;
2960 }
2961 
2962 /**
2963  * ixgbe_request_irq - initialize interrupts
2964  * @adapter: board private structure
2965  *
2966  * Attempts to configure interrupts using the best available
2967  * capabilities of the hardware and kernel.
2968  **/
2969 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
2970 {
2971 	struct net_device *netdev = adapter->netdev;
2972 	int err;
2973 
2974 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
2975 		err = ixgbe_request_msix_irqs(adapter);
2976 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
2977 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
2978 				  netdev->name, adapter);
2979 	else
2980 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
2981 				  netdev->name, adapter);
2982 
2983 	if (err)
2984 		e_err(probe, "request_irq failed, Error %d\n", err);
2985 
2986 	return err;
2987 }
2988 
2989 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
2990 {
2991 	int vector;
2992 
2993 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
2994 		free_irq(adapter->pdev->irq, adapter);
2995 		return;
2996 	}
2997 
2998 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2999 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3000 		struct msix_entry *entry = &adapter->msix_entries[vector];
3001 
3002 		/* free only the irqs that were actually requested */
3003 		if (!q_vector->rx.ring && !q_vector->tx.ring)
3004 			continue;
3005 
3006 		/* clear the affinity_mask in the IRQ descriptor */
3007 		irq_set_affinity_hint(entry->vector, NULL);
3008 
3009 		free_irq(entry->vector, q_vector);
3010 	}
3011 
3012 	free_irq(adapter->msix_entries[vector++].vector, adapter);
3013 }
3014 
3015 /**
3016  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3017  * @adapter: board private structure
3018  **/
3019 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3020 {
3021 	switch (adapter->hw.mac.type) {
3022 	case ixgbe_mac_82598EB:
3023 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3024 		break;
3025 	case ixgbe_mac_82599EB:
3026 	case ixgbe_mac_X540:
3027 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3028 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3029 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3030 		break;
3031 	default:
3032 		break;
3033 	}
3034 	IXGBE_WRITE_FLUSH(&adapter->hw);
3035 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3036 		int vector;
3037 
3038 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
3039 			synchronize_irq(adapter->msix_entries[vector].vector);
3040 
3041 		synchronize_irq(adapter->msix_entries[vector++].vector);
3042 	} else {
3043 		synchronize_irq(adapter->pdev->irq);
3044 	}
3045 }
3046 
3047 /**
3048  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3049  *
3050  **/
3051 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3052 {
3053 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3054 
3055 	ixgbe_write_eitr(q_vector);
3056 
3057 	ixgbe_set_ivar(adapter, 0, 0, 0);
3058 	ixgbe_set_ivar(adapter, 1, 0, 0);
3059 
3060 	e_info(hw, "Legacy interrupt IVAR setup done\n");
3061 }
3062 
3063 /**
3064  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3065  * @adapter: board private structure
3066  * @ring: structure containing ring specific data
3067  *
3068  * Configure the Tx descriptor ring after a reset.
3069  **/
3070 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3071 			     struct ixgbe_ring *ring)
3072 {
3073 	struct ixgbe_hw *hw = &adapter->hw;
3074 	u64 tdba = ring->dma;
3075 	int wait_loop = 10;
3076 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
3077 	u8 reg_idx = ring->reg_idx;
3078 
3079 	/* disable queue to avoid issues while updating state */
3080 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3081 	IXGBE_WRITE_FLUSH(hw);
3082 
3083 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3084 			(tdba & DMA_BIT_MASK(32)));
3085 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3086 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3087 			ring->count * sizeof(union ixgbe_adv_tx_desc));
3088 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3089 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3090 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3091 
3092 	/*
3093 	 * set WTHRESH to encourage burst writeback, it should not be set
3094 	 * higher than 1 when:
3095 	 * - ITR is 0 as it could cause false TX hangs
3096 	 * - ITR is set to > 100k int/sec and BQL is enabled
3097 	 *
3098 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3099 	 * to or less than the number of on chip descriptors, which is
3100 	 * currently 40.
3101 	 */
3102 #if IS_ENABLED(CONFIG_BQL)
3103 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3104 #else
3105 	if (!ring->q_vector || (ring->q_vector->itr < 8))
3106 #endif
3107 		txdctl |= (1 << 16);	/* WTHRESH = 1 */
3108 	else
3109 		txdctl |= (8 << 16);	/* WTHRESH = 8 */
3110 
3111 	/*
3112 	 * Setting PTHRESH to 32 both improves performance
3113 	 * and avoids a TX hang with DFP enabled
3114 	 */
3115 	txdctl |= (1 << 8) |	/* HTHRESH = 1 */
3116 		   32;		/* PTHRESH = 32 */
3117 
3118 	/* reinitialize flowdirector state */
3119 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3120 		ring->atr_sample_rate = adapter->atr_sample_rate;
3121 		ring->atr_count = 0;
3122 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3123 	} else {
3124 		ring->atr_sample_rate = 0;
3125 	}
3126 
3127 	/* initialize XPS */
3128 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3129 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3130 
3131 		if (q_vector)
3132 			netif_set_xps_queue(ring->netdev,
3133 					    &q_vector->affinity_mask,
3134 					    ring->queue_index);
3135 	}
3136 
3137 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3138 
3139 	/* enable queue */
3140 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3141 
3142 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3143 	if (hw->mac.type == ixgbe_mac_82598EB &&
3144 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3145 		return;
3146 
3147 	/* poll to verify queue is enabled */
3148 	do {
3149 		usleep_range(1000, 2000);
3150 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3151 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3152 	if (!wait_loop)
3153 		e_err(drv, "Could not enable Tx Queue %d\n", reg_idx);
3154 }
3155 
3156 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3157 {
3158 	struct ixgbe_hw *hw = &adapter->hw;
3159 	u32 rttdcs, mtqc;
3160 	u8 tcs = netdev_get_num_tc(adapter->netdev);
3161 
3162 	if (hw->mac.type == ixgbe_mac_82598EB)
3163 		return;
3164 
3165 	/* disable the arbiter while setting MTQC */
3166 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3167 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3168 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3169 
3170 	/* set transmit pool layout */
3171 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3172 		mtqc = IXGBE_MTQC_VT_ENA;
3173 		if (tcs > 4)
3174 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3175 		else if (tcs > 1)
3176 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3177 		else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3178 			mtqc |= IXGBE_MTQC_32VF;
3179 		else
3180 			mtqc |= IXGBE_MTQC_64VF;
3181 	} else {
3182 		if (tcs > 4)
3183 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3184 		else if (tcs > 1)
3185 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3186 		else
3187 			mtqc = IXGBE_MTQC_64Q_1PB;
3188 	}
3189 
3190 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3191 
3192 	/* Enable Security TX Buffer IFG for multiple pb */
3193 	if (tcs) {
3194 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3195 		sectx |= IXGBE_SECTX_DCB;
3196 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3197 	}
3198 
3199 	/* re-enable the arbiter */
3200 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3201 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3202 }
3203 
3204 /**
3205  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3206  * @adapter: board private structure
3207  *
3208  * Configure the Tx unit of the MAC after a reset.
3209  **/
3210 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3211 {
3212 	struct ixgbe_hw *hw = &adapter->hw;
3213 	u32 dmatxctl;
3214 	u32 i;
3215 
3216 	ixgbe_setup_mtqc(adapter);
3217 
3218 	if (hw->mac.type != ixgbe_mac_82598EB) {
3219 		/* DMATXCTL.EN must be before Tx queues are enabled */
3220 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3221 		dmatxctl |= IXGBE_DMATXCTL_TE;
3222 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3223 	}
3224 
3225 	/* Setup the HW Tx Head and Tail descriptor pointers */
3226 	for (i = 0; i < adapter->num_tx_queues; i++)
3227 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3228 }
3229 
3230 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3231 				 struct ixgbe_ring *ring)
3232 {
3233 	struct ixgbe_hw *hw = &adapter->hw;
3234 	u8 reg_idx = ring->reg_idx;
3235 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3236 
3237 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3238 
3239 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3240 }
3241 
3242 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3243 				  struct ixgbe_ring *ring)
3244 {
3245 	struct ixgbe_hw *hw = &adapter->hw;
3246 	u8 reg_idx = ring->reg_idx;
3247 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3248 
3249 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3250 
3251 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3252 }
3253 
3254 #ifdef CONFIG_IXGBE_DCB
3255 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3256 #else
3257 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3258 #endif
3259 {
3260 	int i;
3261 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3262 
3263 	if (adapter->ixgbe_ieee_pfc)
3264 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3265 
3266 	/*
3267 	 * We should set the drop enable bit if:
3268 	 *  SR-IOV is enabled
3269 	 *   or
3270 	 *  Number of Rx queues > 1 and flow control is disabled
3271 	 *
3272 	 *  This allows us to avoid head of line blocking for security
3273 	 *  and performance reasons.
3274 	 */
3275 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3276 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3277 		for (i = 0; i < adapter->num_rx_queues; i++)
3278 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3279 	} else {
3280 		for (i = 0; i < adapter->num_rx_queues; i++)
3281 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3282 	}
3283 }
3284 
3285 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3286 
3287 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3288 				   struct ixgbe_ring *rx_ring)
3289 {
3290 	struct ixgbe_hw *hw = &adapter->hw;
3291 	u32 srrctl;
3292 	u8 reg_idx = rx_ring->reg_idx;
3293 
3294 	if (hw->mac.type == ixgbe_mac_82598EB) {
3295 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3296 
3297 		/*
3298 		 * if VMDq is not active we must program one srrctl register
3299 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3300 		 */
3301 		reg_idx &= mask;
3302 	}
3303 
3304 	/* configure header buffer length, needed for RSC */
3305 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3306 
3307 	/* configure the packet buffer length */
3308 	srrctl |= ixgbe_rx_bufsz(rx_ring) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3309 
3310 	/* configure descriptor type */
3311 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3312 
3313 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3314 }
3315 
3316 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3317 {
3318 	struct ixgbe_hw *hw = &adapter->hw;
3319 	static const u32 seed[10] = { 0xE291D73D, 0x1805EC6C, 0x2A94B30D,
3320 			  0xA54F2BEC, 0xEA49AF7C, 0xE214AD3D, 0xB855AABE,
3321 			  0x6A3E67EA, 0x14364D17, 0x3BED200D};
3322 	u32 mrqc = 0, reta = 0;
3323 	u32 rxcsum;
3324 	int i, j;
3325 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3326 
3327 	/*
3328 	 * Program table for at least 2 queues w/ SR-IOV so that VFs can
3329 	 * make full use of any rings they may have.  We will use the
3330 	 * PSRTYPE register to control how many rings we use within the PF.
3331 	 */
3332 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 2))
3333 		rss_i = 2;
3334 
3335 	/* Fill out hash function seeds */
3336 	for (i = 0; i < 10; i++)
3337 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), seed[i]);
3338 
3339 	/* Fill out redirection table */
3340 	for (i = 0, j = 0; i < 128; i++, j++) {
3341 		if (j == rss_i)
3342 			j = 0;
3343 		/* reta = 4-byte sliding window of
3344 		 * 0x00..(indices-1)(indices-1)00..etc. */
3345 		reta = (reta << 8) | (j * 0x11);
3346 		if ((i & 3) == 3)
3347 			IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3348 	}
3349 
3350 	/* Disable indicating checksum in descriptor, enables RSS hash */
3351 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3352 	rxcsum |= IXGBE_RXCSUM_PCSD;
3353 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3354 
3355 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3356 		if (adapter->ring_feature[RING_F_RSS].mask)
3357 			mrqc = IXGBE_MRQC_RSSEN;
3358 	} else {
3359 		u8 tcs = netdev_get_num_tc(adapter->netdev);
3360 
3361 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3362 			if (tcs > 4)
3363 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3364 			else if (tcs > 1)
3365 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3366 			else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3367 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3368 			else
3369 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3370 		} else {
3371 			if (tcs > 4)
3372 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3373 			else if (tcs > 1)
3374 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3375 			else
3376 				mrqc = IXGBE_MRQC_RSSEN;
3377 		}
3378 	}
3379 
3380 	/* Perform hash on these packet types */
3381 	mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3382 		IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3383 		IXGBE_MRQC_RSS_FIELD_IPV6 |
3384 		IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3385 
3386 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3387 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3388 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3389 		mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3390 
3391 	IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3392 }
3393 
3394 /**
3395  * ixgbe_configure_rscctl - enable RSC for the indicated ring
3396  * @adapter:    address of board private structure
3397  * @index:      index of ring to set
3398  **/
3399 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
3400 				   struct ixgbe_ring *ring)
3401 {
3402 	struct ixgbe_hw *hw = &adapter->hw;
3403 	u32 rscctrl;
3404 	u8 reg_idx = ring->reg_idx;
3405 
3406 	if (!ring_is_rsc_enabled(ring))
3407 		return;
3408 
3409 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
3410 	rscctrl |= IXGBE_RSCCTL_RSCEN;
3411 	/*
3412 	 * we must limit the number of descriptors so that the
3413 	 * total size of max desc * buf_len is not greater
3414 	 * than 65536
3415 	 */
3416 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
3417 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
3418 }
3419 
3420 #define IXGBE_MAX_RX_DESC_POLL 10
3421 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
3422 				       struct ixgbe_ring *ring)
3423 {
3424 	struct ixgbe_hw *hw = &adapter->hw;
3425 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3426 	u32 rxdctl;
3427 	u8 reg_idx = ring->reg_idx;
3428 
3429 	if (ixgbe_removed(hw->hw_addr))
3430 		return;
3431 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3432 	if (hw->mac.type == ixgbe_mac_82598EB &&
3433 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3434 		return;
3435 
3436 	do {
3437 		usleep_range(1000, 2000);
3438 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3439 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
3440 
3441 	if (!wait_loop) {
3442 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
3443 		      "the polling period\n", reg_idx);
3444 	}
3445 }
3446 
3447 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
3448 			    struct ixgbe_ring *ring)
3449 {
3450 	struct ixgbe_hw *hw = &adapter->hw;
3451 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3452 	u32 rxdctl;
3453 	u8 reg_idx = ring->reg_idx;
3454 
3455 	if (ixgbe_removed(hw->hw_addr))
3456 		return;
3457 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3458 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
3459 
3460 	/* write value back with RXDCTL.ENABLE bit cleared */
3461 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3462 
3463 	if (hw->mac.type == ixgbe_mac_82598EB &&
3464 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3465 		return;
3466 
3467 	/* the hardware may take up to 100us to really disable the rx queue */
3468 	do {
3469 		udelay(10);
3470 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3471 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
3472 
3473 	if (!wait_loop) {
3474 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3475 		      "the polling period\n", reg_idx);
3476 	}
3477 }
3478 
3479 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
3480 			     struct ixgbe_ring *ring)
3481 {
3482 	struct ixgbe_hw *hw = &adapter->hw;
3483 	u64 rdba = ring->dma;
3484 	u32 rxdctl;
3485 	u8 reg_idx = ring->reg_idx;
3486 
3487 	/* disable queue to avoid issues while updating state */
3488 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3489 	ixgbe_disable_rx_queue(adapter, ring);
3490 
3491 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
3492 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
3493 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
3494 			ring->count * sizeof(union ixgbe_adv_rx_desc));
3495 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
3496 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
3497 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
3498 
3499 	ixgbe_configure_srrctl(adapter, ring);
3500 	ixgbe_configure_rscctl(adapter, ring);
3501 
3502 	if (hw->mac.type == ixgbe_mac_82598EB) {
3503 		/*
3504 		 * enable cache line friendly hardware writes:
3505 		 * PTHRESH=32 descriptors (half the internal cache),
3506 		 * this also removes ugly rx_no_buffer_count increment
3507 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
3508 		 * WTHRESH=8 burst writeback up to two cache lines
3509 		 */
3510 		rxdctl &= ~0x3FFFFF;
3511 		rxdctl |=  0x080420;
3512 	}
3513 
3514 	/* enable receive descriptor ring */
3515 	rxdctl |= IXGBE_RXDCTL_ENABLE;
3516 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3517 
3518 	ixgbe_rx_desc_queue_enable(adapter, ring);
3519 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
3520 }
3521 
3522 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
3523 {
3524 	struct ixgbe_hw *hw = &adapter->hw;
3525 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
3526 	u16 pool;
3527 
3528 	/* PSRTYPE must be initialized in non 82598 adapters */
3529 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
3530 		      IXGBE_PSRTYPE_UDPHDR |
3531 		      IXGBE_PSRTYPE_IPV4HDR |
3532 		      IXGBE_PSRTYPE_L2HDR |
3533 		      IXGBE_PSRTYPE_IPV6HDR;
3534 
3535 	if (hw->mac.type == ixgbe_mac_82598EB)
3536 		return;
3537 
3538 	if (rss_i > 3)
3539 		psrtype |= 2 << 29;
3540 	else if (rss_i > 1)
3541 		psrtype |= 1 << 29;
3542 
3543 	for_each_set_bit(pool, &adapter->fwd_bitmask, 32)
3544 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
3545 }
3546 
3547 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
3548 {
3549 	struct ixgbe_hw *hw = &adapter->hw;
3550 	u32 reg_offset, vf_shift;
3551 	u32 gcr_ext, vmdctl;
3552 	int i;
3553 
3554 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3555 		return;
3556 
3557 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3558 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
3559 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
3560 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
3561 	vmdctl |= IXGBE_VT_CTL_REPLEN;
3562 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
3563 
3564 	vf_shift = VMDQ_P(0) % 32;
3565 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
3566 
3567 	/* Enable only the PF's pool for Tx/Rx */
3568 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), (~0) << vf_shift);
3569 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
3570 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), (~0) << vf_shift);
3571 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
3572 	if (adapter->flags2 & IXGBE_FLAG2_BRIDGE_MODE_VEB)
3573 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3574 
3575 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3576 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
3577 
3578 	/*
3579 	 * Set up VF register offsets for selected VT Mode,
3580 	 * i.e. 32 or 64 VFs for SR-IOV
3581 	 */
3582 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
3583 	case IXGBE_82599_VMDQ_8Q_MASK:
3584 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
3585 		break;
3586 	case IXGBE_82599_VMDQ_4Q_MASK:
3587 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
3588 		break;
3589 	default:
3590 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
3591 		break;
3592 	}
3593 
3594 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
3595 
3596 
3597 	/* Enable MAC Anti-Spoofing */
3598 	hw->mac.ops.set_mac_anti_spoofing(hw, (adapter->num_vfs != 0),
3599 					  adapter->num_vfs);
3600 	/* For VFs that have spoof checking turned off */
3601 	for (i = 0; i < adapter->num_vfs; i++) {
3602 		if (!adapter->vfinfo[i].spoofchk_enabled)
3603 			ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i, false);
3604 	}
3605 }
3606 
3607 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
3608 {
3609 	struct ixgbe_hw *hw = &adapter->hw;
3610 	struct net_device *netdev = adapter->netdev;
3611 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3612 	struct ixgbe_ring *rx_ring;
3613 	int i;
3614 	u32 mhadd, hlreg0;
3615 
3616 #ifdef IXGBE_FCOE
3617 	/* adjust max frame to be able to do baby jumbo for FCoE */
3618 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
3619 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
3620 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3621 
3622 #endif /* IXGBE_FCOE */
3623 
3624 	/* adjust max frame to be at least the size of a standard frame */
3625 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
3626 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
3627 
3628 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3629 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
3630 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3631 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
3632 
3633 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3634 	}
3635 
3636 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3637 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3638 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3639 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3640 
3641 	/*
3642 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3643 	 * the Base and Length of the Rx Descriptor Ring
3644 	 */
3645 	for (i = 0; i < adapter->num_rx_queues; i++) {
3646 		rx_ring = adapter->rx_ring[i];
3647 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
3648 			set_ring_rsc_enabled(rx_ring);
3649 		else
3650 			clear_ring_rsc_enabled(rx_ring);
3651 	}
3652 }
3653 
3654 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
3655 {
3656 	struct ixgbe_hw *hw = &adapter->hw;
3657 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
3658 
3659 	switch (hw->mac.type) {
3660 	case ixgbe_mac_82598EB:
3661 		/*
3662 		 * For VMDq support of different descriptor types or
3663 		 * buffer sizes through the use of multiple SRRCTL
3664 		 * registers, RDRXCTL.MVMEN must be set to 1
3665 		 *
3666 		 * also, the manual doesn't mention it clearly but DCA hints
3667 		 * will only use queue 0's tags unless this bit is set.  Side
3668 		 * effects of setting this bit are only that SRRCTL must be
3669 		 * fully programmed [0..15]
3670 		 */
3671 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
3672 		break;
3673 	case ixgbe_mac_82599EB:
3674 	case ixgbe_mac_X540:
3675 		/* Disable RSC for ACK packets */
3676 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
3677 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
3678 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
3679 		/* hardware requires some bits to be set by default */
3680 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
3681 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
3682 		break;
3683 	default:
3684 		/* We should do nothing since we don't know this hardware */
3685 		return;
3686 	}
3687 
3688 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
3689 }
3690 
3691 /**
3692  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3693  * @adapter: board private structure
3694  *
3695  * Configure the Rx unit of the MAC after a reset.
3696  **/
3697 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
3698 {
3699 	struct ixgbe_hw *hw = &adapter->hw;
3700 	int i;
3701 	u32 rxctrl, rfctl;
3702 
3703 	/* disable receives while setting up the descriptors */
3704 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3705 	IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxctrl & ~IXGBE_RXCTRL_RXEN);
3706 
3707 	ixgbe_setup_psrtype(adapter);
3708 	ixgbe_setup_rdrxctl(adapter);
3709 
3710 	/* RSC Setup */
3711 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
3712 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
3713 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
3714 		rfctl |= IXGBE_RFCTL_RSC_DIS;
3715 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
3716 
3717 	/* Program registers for the distribution of queues */
3718 	ixgbe_setup_mrqc(adapter);
3719 
3720 	/* set_rx_buffer_len must be called before ring initialization */
3721 	ixgbe_set_rx_buffer_len(adapter);
3722 
3723 	/*
3724 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3725 	 * the Base and Length of the Rx Descriptor Ring
3726 	 */
3727 	for (i = 0; i < adapter->num_rx_queues; i++)
3728 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
3729 
3730 	/* disable drop enable for 82598 parts */
3731 	if (hw->mac.type == ixgbe_mac_82598EB)
3732 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3733 
3734 	/* enable all receives */
3735 	rxctrl |= IXGBE_RXCTRL_RXEN;
3736 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
3737 }
3738 
3739 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
3740 				 __be16 proto, u16 vid)
3741 {
3742 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3743 	struct ixgbe_hw *hw = &adapter->hw;
3744 
3745 	/* add VID to filter table */
3746 	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true);
3747 	set_bit(vid, adapter->active_vlans);
3748 
3749 	return 0;
3750 }
3751 
3752 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
3753 				  __be16 proto, u16 vid)
3754 {
3755 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3756 	struct ixgbe_hw *hw = &adapter->hw;
3757 
3758 	/* remove VID from filter table */
3759 	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), false);
3760 	clear_bit(vid, adapter->active_vlans);
3761 
3762 	return 0;
3763 }
3764 
3765 /**
3766  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
3767  * @adapter: driver data
3768  */
3769 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
3770 {
3771 	struct ixgbe_hw *hw = &adapter->hw;
3772 	u32 vlnctrl;
3773 	int i, j;
3774 
3775 	switch (hw->mac.type) {
3776 	case ixgbe_mac_82598EB:
3777 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3778 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
3779 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3780 		break;
3781 	case ixgbe_mac_82599EB:
3782 	case ixgbe_mac_X540:
3783 		for (i = 0; i < adapter->num_rx_queues; i++) {
3784 			struct ixgbe_ring *ring = adapter->rx_ring[i];
3785 
3786 			if (ring->l2_accel_priv)
3787 				continue;
3788 			j = ring->reg_idx;
3789 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3790 			vlnctrl &= ~IXGBE_RXDCTL_VME;
3791 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3792 		}
3793 		break;
3794 	default:
3795 		break;
3796 	}
3797 }
3798 
3799 /**
3800  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
3801  * @adapter: driver data
3802  */
3803 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
3804 {
3805 	struct ixgbe_hw *hw = &adapter->hw;
3806 	u32 vlnctrl;
3807 	int i, j;
3808 
3809 	switch (hw->mac.type) {
3810 	case ixgbe_mac_82598EB:
3811 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3812 		vlnctrl |= IXGBE_VLNCTRL_VME;
3813 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3814 		break;
3815 	case ixgbe_mac_82599EB:
3816 	case ixgbe_mac_X540:
3817 		for (i = 0; i < adapter->num_rx_queues; i++) {
3818 			struct ixgbe_ring *ring = adapter->rx_ring[i];
3819 
3820 			if (ring->l2_accel_priv)
3821 				continue;
3822 			j = ring->reg_idx;
3823 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3824 			vlnctrl |= IXGBE_RXDCTL_VME;
3825 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3826 		}
3827 		break;
3828 	default:
3829 		break;
3830 	}
3831 }
3832 
3833 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
3834 {
3835 	u16 vid;
3836 
3837 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
3838 
3839 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
3840 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
3841 }
3842 
3843 /**
3844  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
3845  * @netdev: network interface device structure
3846  *
3847  * Writes multicast address list to the MTA hash table.
3848  * Returns: -ENOMEM on failure
3849  *                0 on no addresses written
3850  *                X on writing X addresses to MTA
3851  **/
3852 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
3853 {
3854 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3855 	struct ixgbe_hw *hw = &adapter->hw;
3856 
3857 	if (!netif_running(netdev))
3858 		return 0;
3859 
3860 	if (hw->mac.ops.update_mc_addr_list)
3861 		hw->mac.ops.update_mc_addr_list(hw, netdev);
3862 	else
3863 		return -ENOMEM;
3864 
3865 #ifdef CONFIG_PCI_IOV
3866 	ixgbe_restore_vf_multicasts(adapter);
3867 #endif
3868 
3869 	return netdev_mc_count(netdev);
3870 }
3871 
3872 #ifdef CONFIG_PCI_IOV
3873 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
3874 {
3875 	struct ixgbe_hw *hw = &adapter->hw;
3876 	int i;
3877 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3878 		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_IN_USE)
3879 			hw->mac.ops.set_rar(hw, i, adapter->mac_table[i].addr,
3880 					    adapter->mac_table[i].queue,
3881 					    IXGBE_RAH_AV);
3882 		else
3883 			hw->mac.ops.clear_rar(hw, i);
3884 
3885 		adapter->mac_table[i].state &= ~(IXGBE_MAC_STATE_MODIFIED);
3886 	}
3887 }
3888 #endif
3889 
3890 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
3891 {
3892 	struct ixgbe_hw *hw = &adapter->hw;
3893 	int i;
3894 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3895 		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_MODIFIED) {
3896 			if (adapter->mac_table[i].state &
3897 			    IXGBE_MAC_STATE_IN_USE)
3898 				hw->mac.ops.set_rar(hw, i,
3899 						adapter->mac_table[i].addr,
3900 						adapter->mac_table[i].queue,
3901 						IXGBE_RAH_AV);
3902 			else
3903 				hw->mac.ops.clear_rar(hw, i);
3904 
3905 			adapter->mac_table[i].state &=
3906 						~(IXGBE_MAC_STATE_MODIFIED);
3907 		}
3908 	}
3909 }
3910 
3911 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
3912 {
3913 	int i;
3914 	struct ixgbe_hw *hw = &adapter->hw;
3915 
3916 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3917 		adapter->mac_table[i].state |= IXGBE_MAC_STATE_MODIFIED;
3918 		adapter->mac_table[i].state &= ~IXGBE_MAC_STATE_IN_USE;
3919 		memset(adapter->mac_table[i].addr, 0, ETH_ALEN);
3920 		adapter->mac_table[i].queue = 0;
3921 	}
3922 	ixgbe_sync_mac_table(adapter);
3923 }
3924 
3925 static int ixgbe_available_rars(struct ixgbe_adapter *adapter)
3926 {
3927 	struct ixgbe_hw *hw = &adapter->hw;
3928 	int i, count = 0;
3929 
3930 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3931 		if (adapter->mac_table[i].state == 0)
3932 			count++;
3933 	}
3934 	return count;
3935 }
3936 
3937 /* this function destroys the first RAR entry */
3938 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter,
3939 					 u8 *addr)
3940 {
3941 	struct ixgbe_hw *hw = &adapter->hw;
3942 
3943 	memcpy(&adapter->mac_table[0].addr, addr, ETH_ALEN);
3944 	adapter->mac_table[0].queue = VMDQ_P(0);
3945 	adapter->mac_table[0].state = (IXGBE_MAC_STATE_DEFAULT |
3946 				       IXGBE_MAC_STATE_IN_USE);
3947 	hw->mac.ops.set_rar(hw, 0, adapter->mac_table[0].addr,
3948 			    adapter->mac_table[0].queue,
3949 			    IXGBE_RAH_AV);
3950 }
3951 
3952 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter, u8 *addr, u16 queue)
3953 {
3954 	struct ixgbe_hw *hw = &adapter->hw;
3955 	int i;
3956 
3957 	if (is_zero_ether_addr(addr))
3958 		return -EINVAL;
3959 
3960 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3961 		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_IN_USE)
3962 			continue;
3963 		adapter->mac_table[i].state |= (IXGBE_MAC_STATE_MODIFIED |
3964 						IXGBE_MAC_STATE_IN_USE);
3965 		ether_addr_copy(adapter->mac_table[i].addr, addr);
3966 		adapter->mac_table[i].queue = queue;
3967 		ixgbe_sync_mac_table(adapter);
3968 		return i;
3969 	}
3970 	return -ENOMEM;
3971 }
3972 
3973 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter, u8 *addr, u16 queue)
3974 {
3975 	/* search table for addr, if found, set to 0 and sync */
3976 	int i;
3977 	struct ixgbe_hw *hw = &adapter->hw;
3978 
3979 	if (is_zero_ether_addr(addr))
3980 		return -EINVAL;
3981 
3982 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3983 		if (ether_addr_equal(addr, adapter->mac_table[i].addr) &&
3984 		    adapter->mac_table[i].queue == queue) {
3985 			adapter->mac_table[i].state |= IXGBE_MAC_STATE_MODIFIED;
3986 			adapter->mac_table[i].state &= ~IXGBE_MAC_STATE_IN_USE;
3987 			memset(adapter->mac_table[i].addr, 0, ETH_ALEN);
3988 			adapter->mac_table[i].queue = 0;
3989 			ixgbe_sync_mac_table(adapter);
3990 			return 0;
3991 		}
3992 	}
3993 	return -ENOMEM;
3994 }
3995 /**
3996  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
3997  * @netdev: network interface device structure
3998  *
3999  * Writes unicast address list to the RAR table.
4000  * Returns: -ENOMEM on failure/insufficient address space
4001  *                0 on no addresses written
4002  *                X on writing X addresses to the RAR table
4003  **/
4004 static int ixgbe_write_uc_addr_list(struct net_device *netdev, int vfn)
4005 {
4006 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4007 	int count = 0;
4008 
4009 	/* return ENOMEM indicating insufficient memory for addresses */
4010 	if (netdev_uc_count(netdev) > ixgbe_available_rars(adapter))
4011 		return -ENOMEM;
4012 
4013 	if (!netdev_uc_empty(netdev)) {
4014 		struct netdev_hw_addr *ha;
4015 		netdev_for_each_uc_addr(ha, netdev) {
4016 			ixgbe_del_mac_filter(adapter, ha->addr, vfn);
4017 			ixgbe_add_mac_filter(adapter, ha->addr, vfn);
4018 			count++;
4019 		}
4020 	}
4021 	return count;
4022 }
4023 
4024 /**
4025  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4026  * @netdev: network interface device structure
4027  *
4028  * The set_rx_method entry point is called whenever the unicast/multicast
4029  * address list or the network interface flags are updated.  This routine is
4030  * responsible for configuring the hardware for proper unicast, multicast and
4031  * promiscuous mode.
4032  **/
4033 void ixgbe_set_rx_mode(struct net_device *netdev)
4034 {
4035 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4036 	struct ixgbe_hw *hw = &adapter->hw;
4037 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4038 	u32 vlnctrl;
4039 	int count;
4040 
4041 	/* Check for Promiscuous and All Multicast modes */
4042 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4043 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4044 
4045 	/* set all bits that we expect to always be set */
4046 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4047 	fctrl |= IXGBE_FCTRL_BAM;
4048 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4049 	fctrl |= IXGBE_FCTRL_PMCF;
4050 
4051 	/* clear the bits we are changing the status of */
4052 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4053 	vlnctrl &= ~(IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
4054 	if (netdev->flags & IFF_PROMISC) {
4055 		hw->addr_ctrl.user_set_promisc = true;
4056 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4057 		vmolr |= IXGBE_VMOLR_MPE;
4058 		/* Only disable hardware filter vlans in promiscuous mode
4059 		 * if SR-IOV and VMDQ are disabled - otherwise ensure
4060 		 * that hardware VLAN filters remain enabled.
4061 		 */
4062 		if (!(adapter->flags & (IXGBE_FLAG_VMDQ_ENABLED |
4063 					IXGBE_FLAG_SRIOV_ENABLED)))
4064 			vlnctrl |= (IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
4065 	} else {
4066 		if (netdev->flags & IFF_ALLMULTI) {
4067 			fctrl |= IXGBE_FCTRL_MPE;
4068 			vmolr |= IXGBE_VMOLR_MPE;
4069 		}
4070 		vlnctrl |= IXGBE_VLNCTRL_VFE;
4071 		hw->addr_ctrl.user_set_promisc = false;
4072 	}
4073 
4074 	/*
4075 	 * Write addresses to available RAR registers, if there is not
4076 	 * sufficient space to store all the addresses then enable
4077 	 * unicast promiscuous mode
4078 	 */
4079 	count = ixgbe_write_uc_addr_list(netdev, VMDQ_P(0));
4080 	if (count < 0) {
4081 		fctrl |= IXGBE_FCTRL_UPE;
4082 		vmolr |= IXGBE_VMOLR_ROPE;
4083 	}
4084 
4085 	/* Write addresses to the MTA, if the attempt fails
4086 	 * then we should just turn on promiscuous mode so
4087 	 * that we can at least receive multicast traffic
4088 	 */
4089 	count = ixgbe_write_mc_addr_list(netdev);
4090 	if (count < 0) {
4091 		fctrl |= IXGBE_FCTRL_MPE;
4092 		vmolr |= IXGBE_VMOLR_MPE;
4093 	} else if (count) {
4094 		vmolr |= IXGBE_VMOLR_ROMPE;
4095 	}
4096 
4097 	if (hw->mac.type != ixgbe_mac_82598EB) {
4098 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4099 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4100 			   IXGBE_VMOLR_ROPE);
4101 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4102 	}
4103 
4104 	/* This is useful for sniffing bad packets. */
4105 	if (adapter->netdev->features & NETIF_F_RXALL) {
4106 		/* UPE and MPE will be handled by normal PROMISC logic
4107 		 * in e1000e_set_rx_mode */
4108 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4109 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4110 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4111 
4112 		fctrl &= ~(IXGBE_FCTRL_DPF);
4113 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
4114 	}
4115 
4116 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4117 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4118 
4119 	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
4120 		ixgbe_vlan_strip_enable(adapter);
4121 	else
4122 		ixgbe_vlan_strip_disable(adapter);
4123 }
4124 
4125 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4126 {
4127 	int q_idx;
4128 
4129 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4130 		ixgbe_qv_init_lock(adapter->q_vector[q_idx]);
4131 		napi_enable(&adapter->q_vector[q_idx]->napi);
4132 	}
4133 }
4134 
4135 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
4136 {
4137 	int q_idx;
4138 
4139 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4140 		napi_disable(&adapter->q_vector[q_idx]->napi);
4141 		while (!ixgbe_qv_disable(adapter->q_vector[q_idx])) {
4142 			pr_info("QV %d locked\n", q_idx);
4143 			usleep_range(1000, 20000);
4144 		}
4145 	}
4146 }
4147 
4148 #ifdef CONFIG_IXGBE_DCB
4149 /**
4150  * ixgbe_configure_dcb - Configure DCB hardware
4151  * @adapter: ixgbe adapter struct
4152  *
4153  * This is called by the driver on open to configure the DCB hardware.
4154  * This is also called by the gennetlink interface when reconfiguring
4155  * the DCB state.
4156  */
4157 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
4158 {
4159 	struct ixgbe_hw *hw = &adapter->hw;
4160 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4161 
4162 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
4163 		if (hw->mac.type == ixgbe_mac_82598EB)
4164 			netif_set_gso_max_size(adapter->netdev, 65536);
4165 		return;
4166 	}
4167 
4168 	if (hw->mac.type == ixgbe_mac_82598EB)
4169 		netif_set_gso_max_size(adapter->netdev, 32768);
4170 
4171 #ifdef IXGBE_FCOE
4172 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
4173 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
4174 #endif
4175 
4176 	/* reconfigure the hardware */
4177 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
4178 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4179 						DCB_TX_CONFIG);
4180 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4181 						DCB_RX_CONFIG);
4182 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
4183 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
4184 		ixgbe_dcb_hw_ets(&adapter->hw,
4185 				 adapter->ixgbe_ieee_ets,
4186 				 max_frame);
4187 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
4188 					adapter->ixgbe_ieee_pfc->pfc_en,
4189 					adapter->ixgbe_ieee_ets->prio_tc);
4190 	}
4191 
4192 	/* Enable RSS Hash per TC */
4193 	if (hw->mac.type != ixgbe_mac_82598EB) {
4194 		u32 msb = 0;
4195 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
4196 
4197 		while (rss_i) {
4198 			msb++;
4199 			rss_i >>= 1;
4200 		}
4201 
4202 		/* write msb to all 8 TCs in one write */
4203 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
4204 	}
4205 }
4206 #endif
4207 
4208 /* Additional bittime to account for IXGBE framing */
4209 #define IXGBE_ETH_FRAMING 20
4210 
4211 /**
4212  * ixgbe_hpbthresh - calculate high water mark for flow control
4213  *
4214  * @adapter: board private structure to calculate for
4215  * @pb: packet buffer to calculate
4216  */
4217 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
4218 {
4219 	struct ixgbe_hw *hw = &adapter->hw;
4220 	struct net_device *dev = adapter->netdev;
4221 	int link, tc, kb, marker;
4222 	u32 dv_id, rx_pba;
4223 
4224 	/* Calculate max LAN frame size */
4225 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
4226 
4227 #ifdef IXGBE_FCOE
4228 	/* FCoE traffic class uses FCOE jumbo frames */
4229 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4230 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4231 	    (pb == ixgbe_fcoe_get_tc(adapter)))
4232 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4233 #endif
4234 
4235 	/* Calculate delay value for device */
4236 	switch (hw->mac.type) {
4237 	case ixgbe_mac_X540:
4238 		dv_id = IXGBE_DV_X540(link, tc);
4239 		break;
4240 	default:
4241 		dv_id = IXGBE_DV(link, tc);
4242 		break;
4243 	}
4244 
4245 	/* Loopback switch introduces additional latency */
4246 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4247 		dv_id += IXGBE_B2BT(tc);
4248 
4249 	/* Delay value is calculated in bit times convert to KB */
4250 	kb = IXGBE_BT2KB(dv_id);
4251 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
4252 
4253 	marker = rx_pba - kb;
4254 
4255 	/* It is possible that the packet buffer is not large enough
4256 	 * to provide required headroom. In this case throw an error
4257 	 * to user and a do the best we can.
4258 	 */
4259 	if (marker < 0) {
4260 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
4261 			    "headroom to support flow control."
4262 			    "Decrease MTU or number of traffic classes\n", pb);
4263 		marker = tc + 1;
4264 	}
4265 
4266 	return marker;
4267 }
4268 
4269 /**
4270  * ixgbe_lpbthresh - calculate low water mark for for flow control
4271  *
4272  * @adapter: board private structure to calculate for
4273  * @pb: packet buffer to calculate
4274  */
4275 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
4276 {
4277 	struct ixgbe_hw *hw = &adapter->hw;
4278 	struct net_device *dev = adapter->netdev;
4279 	int tc;
4280 	u32 dv_id;
4281 
4282 	/* Calculate max LAN frame size */
4283 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
4284 
4285 #ifdef IXGBE_FCOE
4286 	/* FCoE traffic class uses FCOE jumbo frames */
4287 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4288 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4289 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
4290 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4291 #endif
4292 
4293 	/* Calculate delay value for device */
4294 	switch (hw->mac.type) {
4295 	case ixgbe_mac_X540:
4296 		dv_id = IXGBE_LOW_DV_X540(tc);
4297 		break;
4298 	default:
4299 		dv_id = IXGBE_LOW_DV(tc);
4300 		break;
4301 	}
4302 
4303 	/* Delay value is calculated in bit times convert to KB */
4304 	return IXGBE_BT2KB(dv_id);
4305 }
4306 
4307 /*
4308  * ixgbe_pbthresh_setup - calculate and setup high low water marks
4309  */
4310 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
4311 {
4312 	struct ixgbe_hw *hw = &adapter->hw;
4313 	int num_tc = netdev_get_num_tc(adapter->netdev);
4314 	int i;
4315 
4316 	if (!num_tc)
4317 		num_tc = 1;
4318 
4319 	for (i = 0; i < num_tc; i++) {
4320 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
4321 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
4322 
4323 		/* Low water marks must not be larger than high water marks */
4324 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
4325 			hw->fc.low_water[i] = 0;
4326 	}
4327 
4328 	for (; i < MAX_TRAFFIC_CLASS; i++)
4329 		hw->fc.high_water[i] = 0;
4330 }
4331 
4332 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
4333 {
4334 	struct ixgbe_hw *hw = &adapter->hw;
4335 	int hdrm;
4336 	u8 tc = netdev_get_num_tc(adapter->netdev);
4337 
4338 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
4339 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
4340 		hdrm = 32 << adapter->fdir_pballoc;
4341 	else
4342 		hdrm = 0;
4343 
4344 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
4345 	ixgbe_pbthresh_setup(adapter);
4346 }
4347 
4348 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
4349 {
4350 	struct ixgbe_hw *hw = &adapter->hw;
4351 	struct hlist_node *node2;
4352 	struct ixgbe_fdir_filter *filter;
4353 
4354 	spin_lock(&adapter->fdir_perfect_lock);
4355 
4356 	if (!hlist_empty(&adapter->fdir_filter_list))
4357 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
4358 
4359 	hlist_for_each_entry_safe(filter, node2,
4360 				  &adapter->fdir_filter_list, fdir_node) {
4361 		ixgbe_fdir_write_perfect_filter_82599(hw,
4362 				&filter->filter,
4363 				filter->sw_idx,
4364 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
4365 				IXGBE_FDIR_DROP_QUEUE :
4366 				adapter->rx_ring[filter->action]->reg_idx);
4367 	}
4368 
4369 	spin_unlock(&adapter->fdir_perfect_lock);
4370 }
4371 
4372 static void ixgbe_macvlan_set_rx_mode(struct net_device *dev, unsigned int pool,
4373 				      struct ixgbe_adapter *adapter)
4374 {
4375 	struct ixgbe_hw *hw = &adapter->hw;
4376 	u32 vmolr;
4377 
4378 	/* No unicast promiscuous support for VMDQ devices. */
4379 	vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
4380 	vmolr |= (IXGBE_VMOLR_ROMPE | IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE);
4381 
4382 	/* clear the affected bit */
4383 	vmolr &= ~IXGBE_VMOLR_MPE;
4384 
4385 	if (dev->flags & IFF_ALLMULTI) {
4386 		vmolr |= IXGBE_VMOLR_MPE;
4387 	} else {
4388 		vmolr |= IXGBE_VMOLR_ROMPE;
4389 		hw->mac.ops.update_mc_addr_list(hw, dev);
4390 	}
4391 	ixgbe_write_uc_addr_list(adapter->netdev, pool);
4392 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
4393 }
4394 
4395 static void ixgbe_fwd_psrtype(struct ixgbe_fwd_adapter *vadapter)
4396 {
4397 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4398 	int rss_i = adapter->num_rx_queues_per_pool;
4399 	struct ixgbe_hw *hw = &adapter->hw;
4400 	u16 pool = vadapter->pool;
4401 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4402 		      IXGBE_PSRTYPE_UDPHDR |
4403 		      IXGBE_PSRTYPE_IPV4HDR |
4404 		      IXGBE_PSRTYPE_L2HDR |
4405 		      IXGBE_PSRTYPE_IPV6HDR;
4406 
4407 	if (hw->mac.type == ixgbe_mac_82598EB)
4408 		return;
4409 
4410 	if (rss_i > 3)
4411 		psrtype |= 2 << 29;
4412 	else if (rss_i > 1)
4413 		psrtype |= 1 << 29;
4414 
4415 	IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4416 }
4417 
4418 /**
4419  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4420  * @rx_ring: ring to free buffers from
4421  **/
4422 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
4423 {
4424 	struct device *dev = rx_ring->dev;
4425 	unsigned long size;
4426 	u16 i;
4427 
4428 	/* ring already cleared, nothing to do */
4429 	if (!rx_ring->rx_buffer_info)
4430 		return;
4431 
4432 	/* Free all the Rx ring sk_buffs */
4433 	for (i = 0; i < rx_ring->count; i++) {
4434 		struct ixgbe_rx_buffer *rx_buffer;
4435 
4436 		rx_buffer = &rx_ring->rx_buffer_info[i];
4437 		if (rx_buffer->skb) {
4438 			struct sk_buff *skb = rx_buffer->skb;
4439 			if (IXGBE_CB(skb)->page_released) {
4440 				dma_unmap_page(dev,
4441 					       IXGBE_CB(skb)->dma,
4442 					       ixgbe_rx_bufsz(rx_ring),
4443 					       DMA_FROM_DEVICE);
4444 				IXGBE_CB(skb)->page_released = false;
4445 			}
4446 			dev_kfree_skb(skb);
4447 		}
4448 		rx_buffer->skb = NULL;
4449 		if (rx_buffer->dma)
4450 			dma_unmap_page(dev, rx_buffer->dma,
4451 				       ixgbe_rx_pg_size(rx_ring),
4452 				       DMA_FROM_DEVICE);
4453 		rx_buffer->dma = 0;
4454 		if (rx_buffer->page)
4455 			__free_pages(rx_buffer->page,
4456 				     ixgbe_rx_pg_order(rx_ring));
4457 		rx_buffer->page = NULL;
4458 	}
4459 
4460 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4461 	memset(rx_ring->rx_buffer_info, 0, size);
4462 
4463 	/* Zero out the descriptor ring */
4464 	memset(rx_ring->desc, 0, rx_ring->size);
4465 
4466 	rx_ring->next_to_alloc = 0;
4467 	rx_ring->next_to_clean = 0;
4468 	rx_ring->next_to_use = 0;
4469 }
4470 
4471 static void ixgbe_disable_fwd_ring(struct ixgbe_fwd_adapter *vadapter,
4472 				   struct ixgbe_ring *rx_ring)
4473 {
4474 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4475 	int index = rx_ring->queue_index + vadapter->rx_base_queue;
4476 
4477 	/* shutdown specific queue receive and wait for dma to settle */
4478 	ixgbe_disable_rx_queue(adapter, rx_ring);
4479 	usleep_range(10000, 20000);
4480 	ixgbe_irq_disable_queues(adapter, ((u64)1 << index));
4481 	ixgbe_clean_rx_ring(rx_ring);
4482 	rx_ring->l2_accel_priv = NULL;
4483 }
4484 
4485 static int ixgbe_fwd_ring_down(struct net_device *vdev,
4486 			       struct ixgbe_fwd_adapter *accel)
4487 {
4488 	struct ixgbe_adapter *adapter = accel->real_adapter;
4489 	unsigned int rxbase = accel->rx_base_queue;
4490 	unsigned int txbase = accel->tx_base_queue;
4491 	int i;
4492 
4493 	netif_tx_stop_all_queues(vdev);
4494 
4495 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4496 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4497 		adapter->rx_ring[rxbase + i]->netdev = adapter->netdev;
4498 	}
4499 
4500 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4501 		adapter->tx_ring[txbase + i]->l2_accel_priv = NULL;
4502 		adapter->tx_ring[txbase + i]->netdev = adapter->netdev;
4503 	}
4504 
4505 
4506 	return 0;
4507 }
4508 
4509 static int ixgbe_fwd_ring_up(struct net_device *vdev,
4510 			     struct ixgbe_fwd_adapter *accel)
4511 {
4512 	struct ixgbe_adapter *adapter = accel->real_adapter;
4513 	unsigned int rxbase, txbase, queues;
4514 	int i, baseq, err = 0;
4515 
4516 	if (!test_bit(accel->pool, &adapter->fwd_bitmask))
4517 		return 0;
4518 
4519 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
4520 	netdev_dbg(vdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
4521 		   accel->pool, adapter->num_rx_pools,
4522 		   baseq, baseq + adapter->num_rx_queues_per_pool,
4523 		   adapter->fwd_bitmask);
4524 
4525 	accel->netdev = vdev;
4526 	accel->rx_base_queue = rxbase = baseq;
4527 	accel->tx_base_queue = txbase = baseq;
4528 
4529 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
4530 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4531 
4532 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4533 		adapter->rx_ring[rxbase + i]->netdev = vdev;
4534 		adapter->rx_ring[rxbase + i]->l2_accel_priv = accel;
4535 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[rxbase + i]);
4536 	}
4537 
4538 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4539 		adapter->tx_ring[txbase + i]->netdev = vdev;
4540 		adapter->tx_ring[txbase + i]->l2_accel_priv = accel;
4541 	}
4542 
4543 	queues = min_t(unsigned int,
4544 		       adapter->num_rx_queues_per_pool, vdev->num_tx_queues);
4545 	err = netif_set_real_num_tx_queues(vdev, queues);
4546 	if (err)
4547 		goto fwd_queue_err;
4548 
4549 	err = netif_set_real_num_rx_queues(vdev, queues);
4550 	if (err)
4551 		goto fwd_queue_err;
4552 
4553 	if (is_valid_ether_addr(vdev->dev_addr))
4554 		ixgbe_add_mac_filter(adapter, vdev->dev_addr, accel->pool);
4555 
4556 	ixgbe_fwd_psrtype(accel);
4557 	ixgbe_macvlan_set_rx_mode(vdev, accel->pool, adapter);
4558 	return err;
4559 fwd_queue_err:
4560 	ixgbe_fwd_ring_down(vdev, accel);
4561 	return err;
4562 }
4563 
4564 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
4565 {
4566 	struct net_device *upper;
4567 	struct list_head *iter;
4568 	int err;
4569 
4570 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
4571 		if (netif_is_macvlan(upper)) {
4572 			struct macvlan_dev *dfwd = netdev_priv(upper);
4573 			struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
4574 
4575 			if (dfwd->fwd_priv) {
4576 				err = ixgbe_fwd_ring_up(upper, vadapter);
4577 				if (err)
4578 					continue;
4579 			}
4580 		}
4581 	}
4582 }
4583 
4584 static void ixgbe_configure(struct ixgbe_adapter *adapter)
4585 {
4586 	struct ixgbe_hw *hw = &adapter->hw;
4587 
4588 	ixgbe_configure_pb(adapter);
4589 #ifdef CONFIG_IXGBE_DCB
4590 	ixgbe_configure_dcb(adapter);
4591 #endif
4592 	/*
4593 	 * We must restore virtualization before VLANs or else
4594 	 * the VLVF registers will not be populated
4595 	 */
4596 	ixgbe_configure_virtualization(adapter);
4597 
4598 	ixgbe_set_rx_mode(adapter->netdev);
4599 	ixgbe_restore_vlan(adapter);
4600 
4601 	switch (hw->mac.type) {
4602 	case ixgbe_mac_82599EB:
4603 	case ixgbe_mac_X540:
4604 		hw->mac.ops.disable_rx_buff(hw);
4605 		break;
4606 	default:
4607 		break;
4608 	}
4609 
4610 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
4611 		ixgbe_init_fdir_signature_82599(&adapter->hw,
4612 						adapter->fdir_pballoc);
4613 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
4614 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
4615 					      adapter->fdir_pballoc);
4616 		ixgbe_fdir_filter_restore(adapter);
4617 	}
4618 
4619 	switch (hw->mac.type) {
4620 	case ixgbe_mac_82599EB:
4621 	case ixgbe_mac_X540:
4622 		hw->mac.ops.enable_rx_buff(hw);
4623 		break;
4624 	default:
4625 		break;
4626 	}
4627 
4628 #ifdef IXGBE_FCOE
4629 	/* configure FCoE L2 filters, redirection table, and Rx control */
4630 	ixgbe_configure_fcoe(adapter);
4631 
4632 #endif /* IXGBE_FCOE */
4633 	ixgbe_configure_tx(adapter);
4634 	ixgbe_configure_rx(adapter);
4635 	ixgbe_configure_dfwd(adapter);
4636 }
4637 
4638 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
4639 {
4640 	switch (hw->phy.type) {
4641 	case ixgbe_phy_sfp_avago:
4642 	case ixgbe_phy_sfp_ftl:
4643 	case ixgbe_phy_sfp_intel:
4644 	case ixgbe_phy_sfp_unknown:
4645 	case ixgbe_phy_sfp_passive_tyco:
4646 	case ixgbe_phy_sfp_passive_unknown:
4647 	case ixgbe_phy_sfp_active_unknown:
4648 	case ixgbe_phy_sfp_ftl_active:
4649 	case ixgbe_phy_qsfp_passive_unknown:
4650 	case ixgbe_phy_qsfp_active_unknown:
4651 	case ixgbe_phy_qsfp_intel:
4652 	case ixgbe_phy_qsfp_unknown:
4653 	/* ixgbe_phy_none is set when no SFP module is present */
4654 	case ixgbe_phy_none:
4655 		return true;
4656 	case ixgbe_phy_nl:
4657 		if (hw->mac.type == ixgbe_mac_82598EB)
4658 			return true;
4659 	default:
4660 		return false;
4661 	}
4662 }
4663 
4664 /**
4665  * ixgbe_sfp_link_config - set up SFP+ link
4666  * @adapter: pointer to private adapter struct
4667  **/
4668 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
4669 {
4670 	/*
4671 	 * We are assuming the worst case scenario here, and that
4672 	 * is that an SFP was inserted/removed after the reset
4673 	 * but before SFP detection was enabled.  As such the best
4674 	 * solution is to just start searching as soon as we start
4675 	 */
4676 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
4677 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
4678 
4679 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
4680 }
4681 
4682 /**
4683  * ixgbe_non_sfp_link_config - set up non-SFP+ link
4684  * @hw: pointer to private hardware struct
4685  *
4686  * Returns 0 on success, negative on failure
4687  **/
4688 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
4689 {
4690 	u32 speed;
4691 	bool autoneg, link_up = false;
4692 	u32 ret = IXGBE_ERR_LINK_SETUP;
4693 
4694 	if (hw->mac.ops.check_link)
4695 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
4696 
4697 	if (ret)
4698 		return ret;
4699 
4700 	speed = hw->phy.autoneg_advertised;
4701 	if ((!speed) && (hw->mac.ops.get_link_capabilities))
4702 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
4703 							&autoneg);
4704 	if (ret)
4705 		return ret;
4706 
4707 	if (hw->mac.ops.setup_link)
4708 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
4709 
4710 	return ret;
4711 }
4712 
4713 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
4714 {
4715 	struct ixgbe_hw *hw = &adapter->hw;
4716 	u32 gpie = 0;
4717 
4718 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
4719 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
4720 		       IXGBE_GPIE_OCD;
4721 		gpie |= IXGBE_GPIE_EIAME;
4722 		/*
4723 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
4724 		 * this saves a register write for every interrupt
4725 		 */
4726 		switch (hw->mac.type) {
4727 		case ixgbe_mac_82598EB:
4728 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4729 			break;
4730 		case ixgbe_mac_82599EB:
4731 		case ixgbe_mac_X540:
4732 		default:
4733 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
4734 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
4735 			break;
4736 		}
4737 	} else {
4738 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
4739 		 * specifically only auto mask tx and rx interrupts */
4740 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4741 	}
4742 
4743 	/* XXX: to interrupt immediately for EICS writes, enable this */
4744 	/* gpie |= IXGBE_GPIE_EIMEN; */
4745 
4746 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
4747 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
4748 
4749 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4750 		case IXGBE_82599_VMDQ_8Q_MASK:
4751 			gpie |= IXGBE_GPIE_VTMODE_16;
4752 			break;
4753 		case IXGBE_82599_VMDQ_4Q_MASK:
4754 			gpie |= IXGBE_GPIE_VTMODE_32;
4755 			break;
4756 		default:
4757 			gpie |= IXGBE_GPIE_VTMODE_64;
4758 			break;
4759 		}
4760 	}
4761 
4762 	/* Enable Thermal over heat sensor interrupt */
4763 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
4764 		switch (adapter->hw.mac.type) {
4765 		case ixgbe_mac_82599EB:
4766 			gpie |= IXGBE_SDP0_GPIEN;
4767 			break;
4768 		case ixgbe_mac_X540:
4769 			gpie |= IXGBE_EIMS_TS;
4770 			break;
4771 		default:
4772 			break;
4773 		}
4774 	}
4775 
4776 	/* Enable fan failure interrupt */
4777 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
4778 		gpie |= IXGBE_SDP1_GPIEN;
4779 
4780 	if (hw->mac.type == ixgbe_mac_82599EB) {
4781 		gpie |= IXGBE_SDP1_GPIEN;
4782 		gpie |= IXGBE_SDP2_GPIEN;
4783 	}
4784 
4785 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
4786 }
4787 
4788 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
4789 {
4790 	struct ixgbe_hw *hw = &adapter->hw;
4791 	int err;
4792 	u32 ctrl_ext;
4793 
4794 	ixgbe_get_hw_control(adapter);
4795 	ixgbe_setup_gpie(adapter);
4796 
4797 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
4798 		ixgbe_configure_msix(adapter);
4799 	else
4800 		ixgbe_configure_msi_and_legacy(adapter);
4801 
4802 	/* enable the optics for 82599 SFP+ fiber */
4803 	if (hw->mac.ops.enable_tx_laser)
4804 		hw->mac.ops.enable_tx_laser(hw);
4805 
4806 	smp_mb__before_atomic();
4807 	clear_bit(__IXGBE_DOWN, &adapter->state);
4808 	ixgbe_napi_enable_all(adapter);
4809 
4810 	if (ixgbe_is_sfp(hw)) {
4811 		ixgbe_sfp_link_config(adapter);
4812 	} else {
4813 		err = ixgbe_non_sfp_link_config(hw);
4814 		if (err)
4815 			e_err(probe, "link_config FAILED %d\n", err);
4816 	}
4817 
4818 	/* clear any pending interrupts, may auto mask */
4819 	IXGBE_READ_REG(hw, IXGBE_EICR);
4820 	ixgbe_irq_enable(adapter, true, true);
4821 
4822 	/*
4823 	 * If this adapter has a fan, check to see if we had a failure
4824 	 * before we enabled the interrupt.
4825 	 */
4826 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
4827 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
4828 		if (esdp & IXGBE_ESDP_SDP1)
4829 			e_crit(drv, "Fan has stopped, replace the adapter\n");
4830 	}
4831 
4832 	/* bring the link up in the watchdog, this could race with our first
4833 	 * link up interrupt but shouldn't be a problem */
4834 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4835 	adapter->link_check_timeout = jiffies;
4836 	mod_timer(&adapter->service_timer, jiffies);
4837 
4838 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
4839 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
4840 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
4841 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
4842 }
4843 
4844 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
4845 {
4846 	WARN_ON(in_interrupt());
4847 	/* put off any impending NetWatchDogTimeout */
4848 	adapter->netdev->trans_start = jiffies;
4849 
4850 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
4851 		usleep_range(1000, 2000);
4852 	ixgbe_down(adapter);
4853 	/*
4854 	 * If SR-IOV enabled then wait a bit before bringing the adapter
4855 	 * back up to give the VFs time to respond to the reset.  The
4856 	 * two second wait is based upon the watchdog timer cycle in
4857 	 * the VF driver.
4858 	 */
4859 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4860 		msleep(2000);
4861 	ixgbe_up(adapter);
4862 	clear_bit(__IXGBE_RESETTING, &adapter->state);
4863 }
4864 
4865 void ixgbe_up(struct ixgbe_adapter *adapter)
4866 {
4867 	/* hardware has been reset, we need to reload some things */
4868 	ixgbe_configure(adapter);
4869 
4870 	ixgbe_up_complete(adapter);
4871 }
4872 
4873 void ixgbe_reset(struct ixgbe_adapter *adapter)
4874 {
4875 	struct ixgbe_hw *hw = &adapter->hw;
4876 	struct net_device *netdev = adapter->netdev;
4877 	int err;
4878 	u8 old_addr[ETH_ALEN];
4879 
4880 	if (ixgbe_removed(hw->hw_addr))
4881 		return;
4882 	/* lock SFP init bit to prevent race conditions with the watchdog */
4883 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
4884 		usleep_range(1000, 2000);
4885 
4886 	/* clear all SFP and link config related flags while holding SFP_INIT */
4887 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
4888 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
4889 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4890 
4891 	err = hw->mac.ops.init_hw(hw);
4892 	switch (err) {
4893 	case 0:
4894 	case IXGBE_ERR_SFP_NOT_PRESENT:
4895 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
4896 		break;
4897 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
4898 		e_dev_err("master disable timed out\n");
4899 		break;
4900 	case IXGBE_ERR_EEPROM_VERSION:
4901 		/* We are running on a pre-production device, log a warning */
4902 		e_dev_warn("This device is a pre-production adapter/LOM. "
4903 			   "Please be aware there may be issues associated with "
4904 			   "your hardware.  If you are experiencing problems "
4905 			   "please contact your Intel or hardware "
4906 			   "representative who provided you with this "
4907 			   "hardware.\n");
4908 		break;
4909 	default:
4910 		e_dev_err("Hardware Error: %d\n", err);
4911 	}
4912 
4913 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
4914 	/* do not flush user set addresses */
4915 	memcpy(old_addr, &adapter->mac_table[0].addr, netdev->addr_len);
4916 	ixgbe_flush_sw_mac_table(adapter);
4917 	ixgbe_mac_set_default_filter(adapter, old_addr);
4918 
4919 	/* update SAN MAC vmdq pool selection */
4920 	if (hw->mac.san_mac_rar_index)
4921 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
4922 
4923 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
4924 		ixgbe_ptp_reset(adapter);
4925 }
4926 
4927 /**
4928  * ixgbe_clean_tx_ring - Free Tx Buffers
4929  * @tx_ring: ring to be cleaned
4930  **/
4931 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
4932 {
4933 	struct ixgbe_tx_buffer *tx_buffer_info;
4934 	unsigned long size;
4935 	u16 i;
4936 
4937 	/* ring already cleared, nothing to do */
4938 	if (!tx_ring->tx_buffer_info)
4939 		return;
4940 
4941 	/* Free all the Tx ring sk_buffs */
4942 	for (i = 0; i < tx_ring->count; i++) {
4943 		tx_buffer_info = &tx_ring->tx_buffer_info[i];
4944 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer_info);
4945 	}
4946 
4947 	netdev_tx_reset_queue(txring_txq(tx_ring));
4948 
4949 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
4950 	memset(tx_ring->tx_buffer_info, 0, size);
4951 
4952 	/* Zero out the descriptor ring */
4953 	memset(tx_ring->desc, 0, tx_ring->size);
4954 
4955 	tx_ring->next_to_use = 0;
4956 	tx_ring->next_to_clean = 0;
4957 }
4958 
4959 /**
4960  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
4961  * @adapter: board private structure
4962  **/
4963 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
4964 {
4965 	int i;
4966 
4967 	for (i = 0; i < adapter->num_rx_queues; i++)
4968 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
4969 }
4970 
4971 /**
4972  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
4973  * @adapter: board private structure
4974  **/
4975 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
4976 {
4977 	int i;
4978 
4979 	for (i = 0; i < adapter->num_tx_queues; i++)
4980 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
4981 }
4982 
4983 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
4984 {
4985 	struct hlist_node *node2;
4986 	struct ixgbe_fdir_filter *filter;
4987 
4988 	spin_lock(&adapter->fdir_perfect_lock);
4989 
4990 	hlist_for_each_entry_safe(filter, node2,
4991 				  &adapter->fdir_filter_list, fdir_node) {
4992 		hlist_del(&filter->fdir_node);
4993 		kfree(filter);
4994 	}
4995 	adapter->fdir_filter_count = 0;
4996 
4997 	spin_unlock(&adapter->fdir_perfect_lock);
4998 }
4999 
5000 void ixgbe_down(struct ixgbe_adapter *adapter)
5001 {
5002 	struct net_device *netdev = adapter->netdev;
5003 	struct ixgbe_hw *hw = &adapter->hw;
5004 	struct net_device *upper;
5005 	struct list_head *iter;
5006 	u32 rxctrl;
5007 	int i;
5008 
5009 	/* signal that we are down to the interrupt handler */
5010 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
5011 		return; /* do nothing if already down */
5012 
5013 	/* disable receives */
5014 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
5015 	IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxctrl & ~IXGBE_RXCTRL_RXEN);
5016 
5017 	/* disable all enabled rx queues */
5018 	for (i = 0; i < adapter->num_rx_queues; i++)
5019 		/* this call also flushes the previous write */
5020 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
5021 
5022 	usleep_range(10000, 20000);
5023 
5024 	netif_tx_stop_all_queues(netdev);
5025 
5026 	/* call carrier off first to avoid false dev_watchdog timeouts */
5027 	netif_carrier_off(netdev);
5028 	netif_tx_disable(netdev);
5029 
5030 	/* disable any upper devices */
5031 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
5032 		if (netif_is_macvlan(upper)) {
5033 			struct macvlan_dev *vlan = netdev_priv(upper);
5034 
5035 			if (vlan->fwd_priv) {
5036 				netif_tx_stop_all_queues(upper);
5037 				netif_carrier_off(upper);
5038 				netif_tx_disable(upper);
5039 			}
5040 		}
5041 	}
5042 
5043 	ixgbe_irq_disable(adapter);
5044 
5045 	ixgbe_napi_disable_all(adapter);
5046 
5047 	adapter->flags2 &= ~(IXGBE_FLAG2_FDIR_REQUIRES_REINIT |
5048 			     IXGBE_FLAG2_RESET_REQUESTED);
5049 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5050 
5051 	del_timer_sync(&adapter->service_timer);
5052 
5053 	if (adapter->num_vfs) {
5054 		/* Clear EITR Select mapping */
5055 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
5056 
5057 		/* Mark all the VFs as inactive */
5058 		for (i = 0 ; i < adapter->num_vfs; i++)
5059 			adapter->vfinfo[i].clear_to_send = false;
5060 
5061 		/* ping all the active vfs to let them know we are going down */
5062 		ixgbe_ping_all_vfs(adapter);
5063 
5064 		/* Disable all VFTE/VFRE TX/RX */
5065 		ixgbe_disable_tx_rx(adapter);
5066 	}
5067 
5068 	/* disable transmits in the hardware now that interrupts are off */
5069 	for (i = 0; i < adapter->num_tx_queues; i++) {
5070 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
5071 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5072 	}
5073 
5074 	/* Disable the Tx DMA engine on 82599 and X540 */
5075 	switch (hw->mac.type) {
5076 	case ixgbe_mac_82599EB:
5077 	case ixgbe_mac_X540:
5078 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5079 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5080 				 ~IXGBE_DMATXCTL_TE));
5081 		break;
5082 	default:
5083 		break;
5084 	}
5085 
5086 	if (!pci_channel_offline(adapter->pdev))
5087 		ixgbe_reset(adapter);
5088 
5089 	/* power down the optics for 82599 SFP+ fiber */
5090 	if (hw->mac.ops.disable_tx_laser)
5091 		hw->mac.ops.disable_tx_laser(hw);
5092 
5093 	ixgbe_clean_all_tx_rings(adapter);
5094 	ixgbe_clean_all_rx_rings(adapter);
5095 
5096 #ifdef CONFIG_IXGBE_DCA
5097 	/* since we reset the hardware DCA settings were cleared */
5098 	ixgbe_setup_dca(adapter);
5099 #endif
5100 }
5101 
5102 /**
5103  * ixgbe_tx_timeout - Respond to a Tx Hang
5104  * @netdev: network interface device structure
5105  **/
5106 static void ixgbe_tx_timeout(struct net_device *netdev)
5107 {
5108 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5109 
5110 	/* Do the reset outside of interrupt context */
5111 	ixgbe_tx_timeout_reset(adapter);
5112 }
5113 
5114 /**
5115  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
5116  * @adapter: board private structure to initialize
5117  *
5118  * ixgbe_sw_init initializes the Adapter private data structure.
5119  * Fields are initialized based on PCI device information and
5120  * OS network device settings (MTU size).
5121  **/
5122 static int ixgbe_sw_init(struct ixgbe_adapter *adapter)
5123 {
5124 	struct ixgbe_hw *hw = &adapter->hw;
5125 	struct pci_dev *pdev = adapter->pdev;
5126 	unsigned int rss, fdir;
5127 	u32 fwsm;
5128 #ifdef CONFIG_IXGBE_DCB
5129 	int j;
5130 	struct tc_configuration *tc;
5131 #endif
5132 
5133 	/* PCI config space info */
5134 
5135 	hw->vendor_id = pdev->vendor;
5136 	hw->device_id = pdev->device;
5137 	hw->revision_id = pdev->revision;
5138 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
5139 	hw->subsystem_device_id = pdev->subsystem_device;
5140 
5141 	/* Set common capability flags and settings */
5142 	rss = min_t(int, IXGBE_MAX_RSS_INDICES, num_online_cpus());
5143 	adapter->ring_feature[RING_F_RSS].limit = rss;
5144 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
5145 	adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
5146 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
5147 	adapter->atr_sample_rate = 20;
5148 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
5149 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
5150 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
5151 #ifdef CONFIG_IXGBE_DCA
5152 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
5153 #endif
5154 #ifdef IXGBE_FCOE
5155 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
5156 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5157 #ifdef CONFIG_IXGBE_DCB
5158 	/* Default traffic class to use for FCoE */
5159 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
5160 #endif /* CONFIG_IXGBE_DCB */
5161 #endif /* IXGBE_FCOE */
5162 
5163 	adapter->mac_table = kzalloc(sizeof(struct ixgbe_mac_addr) *
5164 				     hw->mac.num_rar_entries,
5165 				     GFP_ATOMIC);
5166 
5167 	/* Set MAC specific capability flags and exceptions */
5168 	switch (hw->mac.type) {
5169 	case ixgbe_mac_82598EB:
5170 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
5171 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
5172 
5173 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
5174 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
5175 
5176 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
5177 		adapter->ring_feature[RING_F_FDIR].limit = 0;
5178 		adapter->atr_sample_rate = 0;
5179 		adapter->fdir_pballoc = 0;
5180 #ifdef IXGBE_FCOE
5181 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
5182 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5183 #ifdef CONFIG_IXGBE_DCB
5184 		adapter->fcoe.up = 0;
5185 #endif /* IXGBE_DCB */
5186 #endif /* IXGBE_FCOE */
5187 		break;
5188 	case ixgbe_mac_82599EB:
5189 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
5190 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5191 		break;
5192 	case ixgbe_mac_X540:
5193 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM);
5194 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
5195 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5196 		break;
5197 	default:
5198 		break;
5199 	}
5200 
5201 #ifdef IXGBE_FCOE
5202 	/* FCoE support exists, always init the FCoE lock */
5203 	spin_lock_init(&adapter->fcoe.lock);
5204 
5205 #endif
5206 	/* n-tuple support exists, always init our spinlock */
5207 	spin_lock_init(&adapter->fdir_perfect_lock);
5208 
5209 #ifdef CONFIG_IXGBE_DCB
5210 	switch (hw->mac.type) {
5211 	case ixgbe_mac_X540:
5212 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
5213 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
5214 		break;
5215 	default:
5216 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
5217 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
5218 		break;
5219 	}
5220 
5221 	/* Configure DCB traffic classes */
5222 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
5223 		tc = &adapter->dcb_cfg.tc_config[j];
5224 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
5225 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
5226 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
5227 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
5228 		tc->dcb_pfc = pfc_disabled;
5229 	}
5230 
5231 	/* Initialize default user to priority mapping, UPx->TC0 */
5232 	tc = &adapter->dcb_cfg.tc_config[0];
5233 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
5234 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
5235 
5236 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
5237 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
5238 	adapter->dcb_cfg.pfc_mode_enable = false;
5239 	adapter->dcb_set_bitmap = 0x00;
5240 	adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
5241 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
5242 	       sizeof(adapter->temp_dcb_cfg));
5243 
5244 #endif
5245 
5246 	/* default flow control settings */
5247 	hw->fc.requested_mode = ixgbe_fc_full;
5248 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
5249 	ixgbe_pbthresh_setup(adapter);
5250 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
5251 	hw->fc.send_xon = true;
5252 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
5253 
5254 #ifdef CONFIG_PCI_IOV
5255 	if (max_vfs > 0)
5256 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
5257 
5258 	/* assign number of SR-IOV VFs */
5259 	if (hw->mac.type != ixgbe_mac_82598EB) {
5260 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
5261 			adapter->num_vfs = 0;
5262 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
5263 		} else {
5264 			adapter->num_vfs = max_vfs;
5265 		}
5266 	}
5267 #endif /* CONFIG_PCI_IOV */
5268 
5269 	/* enable itr by default in dynamic mode */
5270 	adapter->rx_itr_setting = 1;
5271 	adapter->tx_itr_setting = 1;
5272 
5273 	/* set default ring sizes */
5274 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
5275 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
5276 
5277 	/* set default work limits */
5278 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
5279 
5280 	/* initialize eeprom parameters */
5281 	if (ixgbe_init_eeprom_params_generic(hw)) {
5282 		e_dev_err("EEPROM initialization failed\n");
5283 		return -EIO;
5284 	}
5285 
5286 	/* PF holds first pool slot */
5287 	set_bit(0, &adapter->fwd_bitmask);
5288 	set_bit(__IXGBE_DOWN, &adapter->state);
5289 
5290 	return 0;
5291 }
5292 
5293 /**
5294  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
5295  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
5296  *
5297  * Return 0 on success, negative on failure
5298  **/
5299 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
5300 {
5301 	struct device *dev = tx_ring->dev;
5302 	int orig_node = dev_to_node(dev);
5303 	int numa_node = -1;
5304 	int size;
5305 
5306 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5307 
5308 	if (tx_ring->q_vector)
5309 		numa_node = tx_ring->q_vector->numa_node;
5310 
5311 	tx_ring->tx_buffer_info = vzalloc_node(size, numa_node);
5312 	if (!tx_ring->tx_buffer_info)
5313 		tx_ring->tx_buffer_info = vzalloc(size);
5314 	if (!tx_ring->tx_buffer_info)
5315 		goto err;
5316 
5317 	u64_stats_init(&tx_ring->syncp);
5318 
5319 	/* round up to nearest 4K */
5320 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
5321 	tx_ring->size = ALIGN(tx_ring->size, 4096);
5322 
5323 	set_dev_node(dev, numa_node);
5324 	tx_ring->desc = dma_alloc_coherent(dev,
5325 					   tx_ring->size,
5326 					   &tx_ring->dma,
5327 					   GFP_KERNEL);
5328 	set_dev_node(dev, orig_node);
5329 	if (!tx_ring->desc)
5330 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
5331 						   &tx_ring->dma, GFP_KERNEL);
5332 	if (!tx_ring->desc)
5333 		goto err;
5334 
5335 	tx_ring->next_to_use = 0;
5336 	tx_ring->next_to_clean = 0;
5337 	return 0;
5338 
5339 err:
5340 	vfree(tx_ring->tx_buffer_info);
5341 	tx_ring->tx_buffer_info = NULL;
5342 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
5343 	return -ENOMEM;
5344 }
5345 
5346 /**
5347  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
5348  * @adapter: board private structure
5349  *
5350  * If this function returns with an error, then it's possible one or
5351  * more of the rings is populated (while the rest are not).  It is the
5352  * callers duty to clean those orphaned rings.
5353  *
5354  * Return 0 on success, negative on failure
5355  **/
5356 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
5357 {
5358 	int i, err = 0;
5359 
5360 	for (i = 0; i < adapter->num_tx_queues; i++) {
5361 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
5362 		if (!err)
5363 			continue;
5364 
5365 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
5366 		goto err_setup_tx;
5367 	}
5368 
5369 	return 0;
5370 err_setup_tx:
5371 	/* rewind the index freeing the rings as we go */
5372 	while (i--)
5373 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
5374 	return err;
5375 }
5376 
5377 /**
5378  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
5379  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
5380  *
5381  * Returns 0 on success, negative on failure
5382  **/
5383 int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
5384 {
5385 	struct device *dev = rx_ring->dev;
5386 	int orig_node = dev_to_node(dev);
5387 	int numa_node = -1;
5388 	int size;
5389 
5390 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
5391 
5392 	if (rx_ring->q_vector)
5393 		numa_node = rx_ring->q_vector->numa_node;
5394 
5395 	rx_ring->rx_buffer_info = vzalloc_node(size, numa_node);
5396 	if (!rx_ring->rx_buffer_info)
5397 		rx_ring->rx_buffer_info = vzalloc(size);
5398 	if (!rx_ring->rx_buffer_info)
5399 		goto err;
5400 
5401 	u64_stats_init(&rx_ring->syncp);
5402 
5403 	/* Round up to nearest 4K */
5404 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
5405 	rx_ring->size = ALIGN(rx_ring->size, 4096);
5406 
5407 	set_dev_node(dev, numa_node);
5408 	rx_ring->desc = dma_alloc_coherent(dev,
5409 					   rx_ring->size,
5410 					   &rx_ring->dma,
5411 					   GFP_KERNEL);
5412 	set_dev_node(dev, orig_node);
5413 	if (!rx_ring->desc)
5414 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
5415 						   &rx_ring->dma, GFP_KERNEL);
5416 	if (!rx_ring->desc)
5417 		goto err;
5418 
5419 	rx_ring->next_to_clean = 0;
5420 	rx_ring->next_to_use = 0;
5421 
5422 	return 0;
5423 err:
5424 	vfree(rx_ring->rx_buffer_info);
5425 	rx_ring->rx_buffer_info = NULL;
5426 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
5427 	return -ENOMEM;
5428 }
5429 
5430 /**
5431  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
5432  * @adapter: board private structure
5433  *
5434  * If this function returns with an error, then it's possible one or
5435  * more of the rings is populated (while the rest are not).  It is the
5436  * callers duty to clean those orphaned rings.
5437  *
5438  * Return 0 on success, negative on failure
5439  **/
5440 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
5441 {
5442 	int i, err = 0;
5443 
5444 	for (i = 0; i < adapter->num_rx_queues; i++) {
5445 		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
5446 		if (!err)
5447 			continue;
5448 
5449 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
5450 		goto err_setup_rx;
5451 	}
5452 
5453 #ifdef IXGBE_FCOE
5454 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
5455 	if (!err)
5456 #endif
5457 		return 0;
5458 err_setup_rx:
5459 	/* rewind the index freeing the rings as we go */
5460 	while (i--)
5461 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
5462 	return err;
5463 }
5464 
5465 /**
5466  * ixgbe_free_tx_resources - Free Tx Resources per Queue
5467  * @tx_ring: Tx descriptor ring for a specific queue
5468  *
5469  * Free all transmit software resources
5470  **/
5471 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
5472 {
5473 	ixgbe_clean_tx_ring(tx_ring);
5474 
5475 	vfree(tx_ring->tx_buffer_info);
5476 	tx_ring->tx_buffer_info = NULL;
5477 
5478 	/* if not set, then don't free */
5479 	if (!tx_ring->desc)
5480 		return;
5481 
5482 	dma_free_coherent(tx_ring->dev, tx_ring->size,
5483 			  tx_ring->desc, tx_ring->dma);
5484 
5485 	tx_ring->desc = NULL;
5486 }
5487 
5488 /**
5489  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
5490  * @adapter: board private structure
5491  *
5492  * Free all transmit software resources
5493  **/
5494 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
5495 {
5496 	int i;
5497 
5498 	for (i = 0; i < adapter->num_tx_queues; i++)
5499 		if (adapter->tx_ring[i]->desc)
5500 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
5501 }
5502 
5503 /**
5504  * ixgbe_free_rx_resources - Free Rx Resources
5505  * @rx_ring: ring to clean the resources from
5506  *
5507  * Free all receive software resources
5508  **/
5509 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
5510 {
5511 	ixgbe_clean_rx_ring(rx_ring);
5512 
5513 	vfree(rx_ring->rx_buffer_info);
5514 	rx_ring->rx_buffer_info = NULL;
5515 
5516 	/* if not set, then don't free */
5517 	if (!rx_ring->desc)
5518 		return;
5519 
5520 	dma_free_coherent(rx_ring->dev, rx_ring->size,
5521 			  rx_ring->desc, rx_ring->dma);
5522 
5523 	rx_ring->desc = NULL;
5524 }
5525 
5526 /**
5527  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
5528  * @adapter: board private structure
5529  *
5530  * Free all receive software resources
5531  **/
5532 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
5533 {
5534 	int i;
5535 
5536 #ifdef IXGBE_FCOE
5537 	ixgbe_free_fcoe_ddp_resources(adapter);
5538 
5539 #endif
5540 	for (i = 0; i < adapter->num_rx_queues; i++)
5541 		if (adapter->rx_ring[i]->desc)
5542 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
5543 }
5544 
5545 /**
5546  * ixgbe_change_mtu - Change the Maximum Transfer Unit
5547  * @netdev: network interface device structure
5548  * @new_mtu: new value for maximum frame size
5549  *
5550  * Returns 0 on success, negative on failure
5551  **/
5552 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
5553 {
5554 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5555 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
5556 
5557 	/* MTU < 68 is an error and causes problems on some kernels */
5558 	if ((new_mtu < 68) || (max_frame > IXGBE_MAX_JUMBO_FRAME_SIZE))
5559 		return -EINVAL;
5560 
5561 	/*
5562 	 * For 82599EB we cannot allow legacy VFs to enable their receive
5563 	 * paths when MTU greater than 1500 is configured.  So display a
5564 	 * warning that legacy VFs will be disabled.
5565 	 */
5566 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
5567 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
5568 	    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
5569 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
5570 
5571 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
5572 
5573 	/* must set new MTU before calling down or up */
5574 	netdev->mtu = new_mtu;
5575 
5576 	if (netif_running(netdev))
5577 		ixgbe_reinit_locked(adapter);
5578 
5579 	return 0;
5580 }
5581 
5582 /**
5583  * ixgbe_open - Called when a network interface is made active
5584  * @netdev: network interface device structure
5585  *
5586  * Returns 0 on success, negative value on failure
5587  *
5588  * The open entry point is called when a network interface is made
5589  * active by the system (IFF_UP).  At this point all resources needed
5590  * for transmit and receive operations are allocated, the interrupt
5591  * handler is registered with the OS, the watchdog timer is started,
5592  * and the stack is notified that the interface is ready.
5593  **/
5594 static int ixgbe_open(struct net_device *netdev)
5595 {
5596 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5597 	int err, queues;
5598 
5599 	/* disallow open during test */
5600 	if (test_bit(__IXGBE_TESTING, &adapter->state))
5601 		return -EBUSY;
5602 
5603 	netif_carrier_off(netdev);
5604 
5605 	/* allocate transmit descriptors */
5606 	err = ixgbe_setup_all_tx_resources(adapter);
5607 	if (err)
5608 		goto err_setup_tx;
5609 
5610 	/* allocate receive descriptors */
5611 	err = ixgbe_setup_all_rx_resources(adapter);
5612 	if (err)
5613 		goto err_setup_rx;
5614 
5615 	ixgbe_configure(adapter);
5616 
5617 	err = ixgbe_request_irq(adapter);
5618 	if (err)
5619 		goto err_req_irq;
5620 
5621 	/* Notify the stack of the actual queue counts. */
5622 	if (adapter->num_rx_pools > 1)
5623 		queues = adapter->num_rx_queues_per_pool;
5624 	else
5625 		queues = adapter->num_tx_queues;
5626 
5627 	err = netif_set_real_num_tx_queues(netdev, queues);
5628 	if (err)
5629 		goto err_set_queues;
5630 
5631 	if (adapter->num_rx_pools > 1 &&
5632 	    adapter->num_rx_queues > IXGBE_MAX_L2A_QUEUES)
5633 		queues = IXGBE_MAX_L2A_QUEUES;
5634 	else
5635 		queues = adapter->num_rx_queues;
5636 	err = netif_set_real_num_rx_queues(netdev, queues);
5637 	if (err)
5638 		goto err_set_queues;
5639 
5640 	ixgbe_ptp_init(adapter);
5641 
5642 	ixgbe_up_complete(adapter);
5643 
5644 	return 0;
5645 
5646 err_set_queues:
5647 	ixgbe_free_irq(adapter);
5648 err_req_irq:
5649 	ixgbe_free_all_rx_resources(adapter);
5650 err_setup_rx:
5651 	ixgbe_free_all_tx_resources(adapter);
5652 err_setup_tx:
5653 	ixgbe_reset(adapter);
5654 
5655 	return err;
5656 }
5657 
5658 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
5659 {
5660 	ixgbe_ptp_suspend(adapter);
5661 
5662 	ixgbe_down(adapter);
5663 	ixgbe_free_irq(adapter);
5664 
5665 	ixgbe_free_all_tx_resources(adapter);
5666 	ixgbe_free_all_rx_resources(adapter);
5667 }
5668 
5669 /**
5670  * ixgbe_close - Disables a network interface
5671  * @netdev: network interface device structure
5672  *
5673  * Returns 0, this is not allowed to fail
5674  *
5675  * The close entry point is called when an interface is de-activated
5676  * by the OS.  The hardware is still under the drivers control, but
5677  * needs to be disabled.  A global MAC reset is issued to stop the
5678  * hardware, and all transmit and receive resources are freed.
5679  **/
5680 static int ixgbe_close(struct net_device *netdev)
5681 {
5682 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5683 
5684 	ixgbe_ptp_stop(adapter);
5685 
5686 	ixgbe_close_suspend(adapter);
5687 
5688 	ixgbe_fdir_filter_exit(adapter);
5689 
5690 	ixgbe_release_hw_control(adapter);
5691 
5692 	return 0;
5693 }
5694 
5695 #ifdef CONFIG_PM
5696 static int ixgbe_resume(struct pci_dev *pdev)
5697 {
5698 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
5699 	struct net_device *netdev = adapter->netdev;
5700 	u32 err;
5701 
5702 	adapter->hw.hw_addr = adapter->io_addr;
5703 	pci_set_power_state(pdev, PCI_D0);
5704 	pci_restore_state(pdev);
5705 	/*
5706 	 * pci_restore_state clears dev->state_saved so call
5707 	 * pci_save_state to restore it.
5708 	 */
5709 	pci_save_state(pdev);
5710 
5711 	err = pci_enable_device_mem(pdev);
5712 	if (err) {
5713 		e_dev_err("Cannot enable PCI device from suspend\n");
5714 		return err;
5715 	}
5716 	smp_mb__before_atomic();
5717 	clear_bit(__IXGBE_DISABLED, &adapter->state);
5718 	pci_set_master(pdev);
5719 
5720 	pci_wake_from_d3(pdev, false);
5721 
5722 	ixgbe_reset(adapter);
5723 
5724 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
5725 
5726 	rtnl_lock();
5727 	err = ixgbe_init_interrupt_scheme(adapter);
5728 	if (!err && netif_running(netdev))
5729 		err = ixgbe_open(netdev);
5730 
5731 	rtnl_unlock();
5732 
5733 	if (err)
5734 		return err;
5735 
5736 	netif_device_attach(netdev);
5737 
5738 	return 0;
5739 }
5740 #endif /* CONFIG_PM */
5741 
5742 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
5743 {
5744 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
5745 	struct net_device *netdev = adapter->netdev;
5746 	struct ixgbe_hw *hw = &adapter->hw;
5747 	u32 ctrl, fctrl;
5748 	u32 wufc = adapter->wol;
5749 #ifdef CONFIG_PM
5750 	int retval = 0;
5751 #endif
5752 
5753 	netif_device_detach(netdev);
5754 
5755 	rtnl_lock();
5756 	if (netif_running(netdev))
5757 		ixgbe_close_suspend(adapter);
5758 	rtnl_unlock();
5759 
5760 	ixgbe_clear_interrupt_scheme(adapter);
5761 
5762 #ifdef CONFIG_PM
5763 	retval = pci_save_state(pdev);
5764 	if (retval)
5765 		return retval;
5766 
5767 #endif
5768 	if (hw->mac.ops.stop_link_on_d3)
5769 		hw->mac.ops.stop_link_on_d3(hw);
5770 
5771 	if (wufc) {
5772 		ixgbe_set_rx_mode(netdev);
5773 
5774 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
5775 		if (hw->mac.ops.enable_tx_laser)
5776 			hw->mac.ops.enable_tx_laser(hw);
5777 
5778 		/* turn on all-multi mode if wake on multicast is enabled */
5779 		if (wufc & IXGBE_WUFC_MC) {
5780 			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
5781 			fctrl |= IXGBE_FCTRL_MPE;
5782 			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
5783 		}
5784 
5785 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
5786 		ctrl |= IXGBE_CTRL_GIO_DIS;
5787 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
5788 
5789 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
5790 	} else {
5791 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
5792 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
5793 	}
5794 
5795 	switch (hw->mac.type) {
5796 	case ixgbe_mac_82598EB:
5797 		pci_wake_from_d3(pdev, false);
5798 		break;
5799 	case ixgbe_mac_82599EB:
5800 	case ixgbe_mac_X540:
5801 		pci_wake_from_d3(pdev, !!wufc);
5802 		break;
5803 	default:
5804 		break;
5805 	}
5806 
5807 	*enable_wake = !!wufc;
5808 
5809 	ixgbe_release_hw_control(adapter);
5810 
5811 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
5812 		pci_disable_device(pdev);
5813 
5814 	return 0;
5815 }
5816 
5817 #ifdef CONFIG_PM
5818 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
5819 {
5820 	int retval;
5821 	bool wake;
5822 
5823 	retval = __ixgbe_shutdown(pdev, &wake);
5824 	if (retval)
5825 		return retval;
5826 
5827 	if (wake) {
5828 		pci_prepare_to_sleep(pdev);
5829 	} else {
5830 		pci_wake_from_d3(pdev, false);
5831 		pci_set_power_state(pdev, PCI_D3hot);
5832 	}
5833 
5834 	return 0;
5835 }
5836 #endif /* CONFIG_PM */
5837 
5838 static void ixgbe_shutdown(struct pci_dev *pdev)
5839 {
5840 	bool wake;
5841 
5842 	__ixgbe_shutdown(pdev, &wake);
5843 
5844 	if (system_state == SYSTEM_POWER_OFF) {
5845 		pci_wake_from_d3(pdev, wake);
5846 		pci_set_power_state(pdev, PCI_D3hot);
5847 	}
5848 }
5849 
5850 /**
5851  * ixgbe_update_stats - Update the board statistics counters.
5852  * @adapter: board private structure
5853  **/
5854 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
5855 {
5856 	struct net_device *netdev = adapter->netdev;
5857 	struct ixgbe_hw *hw = &adapter->hw;
5858 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
5859 	u64 total_mpc = 0;
5860 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
5861 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
5862 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
5863 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
5864 
5865 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5866 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5867 		return;
5868 
5869 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
5870 		u64 rsc_count = 0;
5871 		u64 rsc_flush = 0;
5872 		for (i = 0; i < adapter->num_rx_queues; i++) {
5873 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
5874 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
5875 		}
5876 		adapter->rsc_total_count = rsc_count;
5877 		adapter->rsc_total_flush = rsc_flush;
5878 	}
5879 
5880 	for (i = 0; i < adapter->num_rx_queues; i++) {
5881 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
5882 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
5883 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
5884 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
5885 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
5886 		bytes += rx_ring->stats.bytes;
5887 		packets += rx_ring->stats.packets;
5888 	}
5889 	adapter->non_eop_descs = non_eop_descs;
5890 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
5891 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
5892 	adapter->hw_csum_rx_error = hw_csum_rx_error;
5893 	netdev->stats.rx_bytes = bytes;
5894 	netdev->stats.rx_packets = packets;
5895 
5896 	bytes = 0;
5897 	packets = 0;
5898 	/* gather some stats to the adapter struct that are per queue */
5899 	for (i = 0; i < adapter->num_tx_queues; i++) {
5900 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
5901 		restart_queue += tx_ring->tx_stats.restart_queue;
5902 		tx_busy += tx_ring->tx_stats.tx_busy;
5903 		bytes += tx_ring->stats.bytes;
5904 		packets += tx_ring->stats.packets;
5905 	}
5906 	adapter->restart_queue = restart_queue;
5907 	adapter->tx_busy = tx_busy;
5908 	netdev->stats.tx_bytes = bytes;
5909 	netdev->stats.tx_packets = packets;
5910 
5911 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
5912 
5913 	/* 8 register reads */
5914 	for (i = 0; i < 8; i++) {
5915 		/* for packet buffers not used, the register should read 0 */
5916 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
5917 		missed_rx += mpc;
5918 		hwstats->mpc[i] += mpc;
5919 		total_mpc += hwstats->mpc[i];
5920 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
5921 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
5922 		switch (hw->mac.type) {
5923 		case ixgbe_mac_82598EB:
5924 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
5925 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
5926 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
5927 			hwstats->pxonrxc[i] +=
5928 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
5929 			break;
5930 		case ixgbe_mac_82599EB:
5931 		case ixgbe_mac_X540:
5932 			hwstats->pxonrxc[i] +=
5933 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
5934 			break;
5935 		default:
5936 			break;
5937 		}
5938 	}
5939 
5940 	/*16 register reads */
5941 	for (i = 0; i < 16; i++) {
5942 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
5943 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
5944 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
5945 		    (hw->mac.type == ixgbe_mac_X540)) {
5946 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
5947 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
5948 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
5949 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
5950 		}
5951 	}
5952 
5953 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
5954 	/* work around hardware counting issue */
5955 	hwstats->gprc -= missed_rx;
5956 
5957 	ixgbe_update_xoff_received(adapter);
5958 
5959 	/* 82598 hardware only has a 32 bit counter in the high register */
5960 	switch (hw->mac.type) {
5961 	case ixgbe_mac_82598EB:
5962 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
5963 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
5964 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
5965 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
5966 		break;
5967 	case ixgbe_mac_X540:
5968 		/* OS2BMC stats are X540 only*/
5969 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
5970 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
5971 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
5972 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
5973 	case ixgbe_mac_82599EB:
5974 		for (i = 0; i < 16; i++)
5975 			adapter->hw_rx_no_dma_resources +=
5976 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
5977 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
5978 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
5979 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
5980 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
5981 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
5982 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
5983 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
5984 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
5985 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
5986 #ifdef IXGBE_FCOE
5987 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
5988 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
5989 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
5990 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
5991 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
5992 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
5993 		/* Add up per cpu counters for total ddp aloc fail */
5994 		if (adapter->fcoe.ddp_pool) {
5995 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
5996 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
5997 			unsigned int cpu;
5998 			u64 noddp = 0, noddp_ext_buff = 0;
5999 			for_each_possible_cpu(cpu) {
6000 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
6001 				noddp += ddp_pool->noddp;
6002 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
6003 			}
6004 			hwstats->fcoe_noddp = noddp;
6005 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
6006 		}
6007 #endif /* IXGBE_FCOE */
6008 		break;
6009 	default:
6010 		break;
6011 	}
6012 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
6013 	hwstats->bprc += bprc;
6014 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
6015 	if (hw->mac.type == ixgbe_mac_82598EB)
6016 		hwstats->mprc -= bprc;
6017 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
6018 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
6019 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
6020 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
6021 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
6022 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
6023 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
6024 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
6025 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
6026 	hwstats->lxontxc += lxon;
6027 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
6028 	hwstats->lxofftxc += lxoff;
6029 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
6030 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
6031 	/*
6032 	 * 82598 errata - tx of flow control packets is included in tx counters
6033 	 */
6034 	xon_off_tot = lxon + lxoff;
6035 	hwstats->gptc -= xon_off_tot;
6036 	hwstats->mptc -= xon_off_tot;
6037 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
6038 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
6039 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
6040 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
6041 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
6042 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
6043 	hwstats->ptc64 -= xon_off_tot;
6044 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
6045 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
6046 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
6047 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
6048 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
6049 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
6050 
6051 	/* Fill out the OS statistics structure */
6052 	netdev->stats.multicast = hwstats->mprc;
6053 
6054 	/* Rx Errors */
6055 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
6056 	netdev->stats.rx_dropped = 0;
6057 	netdev->stats.rx_length_errors = hwstats->rlec;
6058 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
6059 	netdev->stats.rx_missed_errors = total_mpc;
6060 }
6061 
6062 /**
6063  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
6064  * @adapter: pointer to the device adapter structure
6065  **/
6066 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
6067 {
6068 	struct ixgbe_hw *hw = &adapter->hw;
6069 	int i;
6070 
6071 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
6072 		return;
6073 
6074 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6075 
6076 	/* if interface is down do nothing */
6077 	if (test_bit(__IXGBE_DOWN, &adapter->state))
6078 		return;
6079 
6080 	/* do nothing if we are not using signature filters */
6081 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
6082 		return;
6083 
6084 	adapter->fdir_overflow++;
6085 
6086 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
6087 		for (i = 0; i < adapter->num_tx_queues; i++)
6088 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
6089 				&(adapter->tx_ring[i]->state));
6090 		/* re-enable flow director interrupts */
6091 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
6092 	} else {
6093 		e_err(probe, "failed to finish FDIR re-initialization, "
6094 		      "ignored adding FDIR ATR filters\n");
6095 	}
6096 }
6097 
6098 /**
6099  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
6100  * @adapter: pointer to the device adapter structure
6101  *
6102  * This function serves two purposes.  First it strobes the interrupt lines
6103  * in order to make certain interrupts are occurring.  Secondly it sets the
6104  * bits needed to check for TX hangs.  As a result we should immediately
6105  * determine if a hang has occurred.
6106  */
6107 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
6108 {
6109 	struct ixgbe_hw *hw = &adapter->hw;
6110 	u64 eics = 0;
6111 	int i;
6112 
6113 	/* If we're down, removing or resetting, just bail */
6114 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6115 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6116 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6117 		return;
6118 
6119 	/* Force detection of hung controller */
6120 	if (netif_carrier_ok(adapter->netdev)) {
6121 		for (i = 0; i < adapter->num_tx_queues; i++)
6122 			set_check_for_tx_hang(adapter->tx_ring[i]);
6123 	}
6124 
6125 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
6126 		/*
6127 		 * for legacy and MSI interrupts don't set any bits
6128 		 * that are enabled for EIAM, because this operation
6129 		 * would set *both* EIMS and EICS for any bit in EIAM
6130 		 */
6131 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
6132 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
6133 	} else {
6134 		/* get one bit for every active tx/rx interrupt vector */
6135 		for (i = 0; i < adapter->num_q_vectors; i++) {
6136 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
6137 			if (qv->rx.ring || qv->tx.ring)
6138 				eics |= ((u64)1 << i);
6139 		}
6140 	}
6141 
6142 	/* Cause software interrupt to ensure rings are cleaned */
6143 	ixgbe_irq_rearm_queues(adapter, eics);
6144 
6145 }
6146 
6147 /**
6148  * ixgbe_watchdog_update_link - update the link status
6149  * @adapter: pointer to the device adapter structure
6150  * @link_speed: pointer to a u32 to store the link_speed
6151  **/
6152 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
6153 {
6154 	struct ixgbe_hw *hw = &adapter->hw;
6155 	u32 link_speed = adapter->link_speed;
6156 	bool link_up = adapter->link_up;
6157 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
6158 
6159 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
6160 		return;
6161 
6162 	if (hw->mac.ops.check_link) {
6163 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
6164 	} else {
6165 		/* always assume link is up, if no check link function */
6166 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
6167 		link_up = true;
6168 	}
6169 
6170 	if (adapter->ixgbe_ieee_pfc)
6171 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
6172 
6173 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
6174 		hw->mac.ops.fc_enable(hw);
6175 		ixgbe_set_rx_drop_en(adapter);
6176 	}
6177 
6178 	if (link_up ||
6179 	    time_after(jiffies, (adapter->link_check_timeout +
6180 				 IXGBE_TRY_LINK_TIMEOUT))) {
6181 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6182 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
6183 		IXGBE_WRITE_FLUSH(hw);
6184 	}
6185 
6186 	adapter->link_up = link_up;
6187 	adapter->link_speed = link_speed;
6188 }
6189 
6190 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
6191 {
6192 #ifdef CONFIG_IXGBE_DCB
6193 	struct net_device *netdev = adapter->netdev;
6194 	struct dcb_app app = {
6195 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
6196 			      .protocol = 0,
6197 			     };
6198 	u8 up = 0;
6199 
6200 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
6201 		up = dcb_ieee_getapp_mask(netdev, &app);
6202 
6203 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
6204 #endif
6205 }
6206 
6207 /**
6208  * ixgbe_watchdog_link_is_up - update netif_carrier status and
6209  *                             print link up message
6210  * @adapter: pointer to the device adapter structure
6211  **/
6212 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
6213 {
6214 	struct net_device *netdev = adapter->netdev;
6215 	struct ixgbe_hw *hw = &adapter->hw;
6216 	struct net_device *upper;
6217 	struct list_head *iter;
6218 	u32 link_speed = adapter->link_speed;
6219 	bool flow_rx, flow_tx;
6220 
6221 	/* only continue if link was previously down */
6222 	if (netif_carrier_ok(netdev))
6223 		return;
6224 
6225 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
6226 
6227 	switch (hw->mac.type) {
6228 	case ixgbe_mac_82598EB: {
6229 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6230 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
6231 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
6232 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
6233 	}
6234 		break;
6235 	case ixgbe_mac_X540:
6236 	case ixgbe_mac_82599EB: {
6237 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
6238 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
6239 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
6240 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
6241 	}
6242 		break;
6243 	default:
6244 		flow_tx = false;
6245 		flow_rx = false;
6246 		break;
6247 	}
6248 
6249 	adapter->last_rx_ptp_check = jiffies;
6250 
6251 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6252 		ixgbe_ptp_start_cyclecounter(adapter);
6253 
6254 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n",
6255 	       (link_speed == IXGBE_LINK_SPEED_10GB_FULL ?
6256 	       "10 Gbps" :
6257 	       (link_speed == IXGBE_LINK_SPEED_1GB_FULL ?
6258 	       "1 Gbps" :
6259 	       (link_speed == IXGBE_LINK_SPEED_100_FULL ?
6260 	       "100 Mbps" :
6261 	       "unknown speed"))),
6262 	       ((flow_rx && flow_tx) ? "RX/TX" :
6263 	       (flow_rx ? "RX" :
6264 	       (flow_tx ? "TX" : "None"))));
6265 
6266 	netif_carrier_on(netdev);
6267 	ixgbe_check_vf_rate_limit(adapter);
6268 
6269 	/* enable transmits */
6270 	netif_tx_wake_all_queues(adapter->netdev);
6271 
6272 	/* enable any upper devices */
6273 	rtnl_lock();
6274 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
6275 		if (netif_is_macvlan(upper)) {
6276 			struct macvlan_dev *vlan = netdev_priv(upper);
6277 
6278 			if (vlan->fwd_priv)
6279 				netif_tx_wake_all_queues(upper);
6280 		}
6281 	}
6282 	rtnl_unlock();
6283 
6284 	/* update the default user priority for VFs */
6285 	ixgbe_update_default_up(adapter);
6286 
6287 	/* ping all the active vfs to let them know link has changed */
6288 	ixgbe_ping_all_vfs(adapter);
6289 }
6290 
6291 /**
6292  * ixgbe_watchdog_link_is_down - update netif_carrier status and
6293  *                               print link down message
6294  * @adapter: pointer to the adapter structure
6295  **/
6296 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
6297 {
6298 	struct net_device *netdev = adapter->netdev;
6299 	struct ixgbe_hw *hw = &adapter->hw;
6300 
6301 	adapter->link_up = false;
6302 	adapter->link_speed = 0;
6303 
6304 	/* only continue if link was up previously */
6305 	if (!netif_carrier_ok(netdev))
6306 		return;
6307 
6308 	/* poll for SFP+ cable when link is down */
6309 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
6310 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
6311 
6312 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6313 		ixgbe_ptp_start_cyclecounter(adapter);
6314 
6315 	e_info(drv, "NIC Link is Down\n");
6316 	netif_carrier_off(netdev);
6317 
6318 	/* ping all the active vfs to let them know link has changed */
6319 	ixgbe_ping_all_vfs(adapter);
6320 }
6321 
6322 /**
6323  * ixgbe_watchdog_flush_tx - flush queues on link down
6324  * @adapter: pointer to the device adapter structure
6325  **/
6326 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
6327 {
6328 	int i;
6329 	int some_tx_pending = 0;
6330 
6331 	if (!netif_carrier_ok(adapter->netdev)) {
6332 		for (i = 0; i < adapter->num_tx_queues; i++) {
6333 			struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6334 			if (tx_ring->next_to_use != tx_ring->next_to_clean) {
6335 				some_tx_pending = 1;
6336 				break;
6337 			}
6338 		}
6339 
6340 		if (some_tx_pending) {
6341 			/* We've lost link, so the controller stops DMA,
6342 			 * but we've got queued Tx work that's never going
6343 			 * to get done, so reset controller to flush Tx.
6344 			 * (Do the reset outside of interrupt context).
6345 			 */
6346 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
6347 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
6348 		}
6349 	}
6350 }
6351 
6352 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
6353 {
6354 	u32 ssvpc;
6355 
6356 	/* Do not perform spoof check for 82598 or if not in IOV mode */
6357 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
6358 	    adapter->num_vfs == 0)
6359 		return;
6360 
6361 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
6362 
6363 	/*
6364 	 * ssvpc register is cleared on read, if zero then no
6365 	 * spoofed packets in the last interval.
6366 	 */
6367 	if (!ssvpc)
6368 		return;
6369 
6370 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
6371 }
6372 
6373 /**
6374  * ixgbe_watchdog_subtask - check and bring link up
6375  * @adapter: pointer to the device adapter structure
6376  **/
6377 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
6378 {
6379 	/* if interface is down, removing or resetting, do nothing */
6380 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6381 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6382 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6383 		return;
6384 
6385 	ixgbe_watchdog_update_link(adapter);
6386 
6387 	if (adapter->link_up)
6388 		ixgbe_watchdog_link_is_up(adapter);
6389 	else
6390 		ixgbe_watchdog_link_is_down(adapter);
6391 
6392 	ixgbe_spoof_check(adapter);
6393 	ixgbe_update_stats(adapter);
6394 
6395 	ixgbe_watchdog_flush_tx(adapter);
6396 }
6397 
6398 /**
6399  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
6400  * @adapter: the ixgbe adapter structure
6401  **/
6402 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
6403 {
6404 	struct ixgbe_hw *hw = &adapter->hw;
6405 	s32 err;
6406 
6407 	/* not searching for SFP so there is nothing to do here */
6408 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
6409 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
6410 		return;
6411 
6412 	/* someone else is in init, wait until next service event */
6413 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
6414 		return;
6415 
6416 	err = hw->phy.ops.identify_sfp(hw);
6417 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
6418 		goto sfp_out;
6419 
6420 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
6421 		/* If no cable is present, then we need to reset
6422 		 * the next time we find a good cable. */
6423 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
6424 	}
6425 
6426 	/* exit on error */
6427 	if (err)
6428 		goto sfp_out;
6429 
6430 	/* exit if reset not needed */
6431 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
6432 		goto sfp_out;
6433 
6434 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
6435 
6436 	/*
6437 	 * A module may be identified correctly, but the EEPROM may not have
6438 	 * support for that module.  setup_sfp() will fail in that case, so
6439 	 * we should not allow that module to load.
6440 	 */
6441 	if (hw->mac.type == ixgbe_mac_82598EB)
6442 		err = hw->phy.ops.reset(hw);
6443 	else
6444 		err = hw->mac.ops.setup_sfp(hw);
6445 
6446 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
6447 		goto sfp_out;
6448 
6449 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
6450 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
6451 
6452 sfp_out:
6453 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
6454 
6455 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
6456 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
6457 		e_dev_err("failed to initialize because an unsupported "
6458 			  "SFP+ module type was detected.\n");
6459 		e_dev_err("Reload the driver after installing a "
6460 			  "supported module.\n");
6461 		unregister_netdev(adapter->netdev);
6462 	}
6463 }
6464 
6465 /**
6466  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
6467  * @adapter: the ixgbe adapter structure
6468  **/
6469 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
6470 {
6471 	struct ixgbe_hw *hw = &adapter->hw;
6472 	u32 speed;
6473 	bool autoneg = false;
6474 
6475 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
6476 		return;
6477 
6478 	/* someone else is in init, wait until next service event */
6479 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
6480 		return;
6481 
6482 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
6483 
6484 	speed = hw->phy.autoneg_advertised;
6485 	if ((!speed) && (hw->mac.ops.get_link_capabilities)) {
6486 		hw->mac.ops.get_link_capabilities(hw, &speed, &autoneg);
6487 
6488 		/* setup the highest link when no autoneg */
6489 		if (!autoneg) {
6490 			if (speed & IXGBE_LINK_SPEED_10GB_FULL)
6491 				speed = IXGBE_LINK_SPEED_10GB_FULL;
6492 		}
6493 	}
6494 
6495 	if (hw->mac.ops.setup_link)
6496 		hw->mac.ops.setup_link(hw, speed, true);
6497 
6498 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
6499 	adapter->link_check_timeout = jiffies;
6500 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
6501 }
6502 
6503 #ifdef CONFIG_PCI_IOV
6504 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
6505 {
6506 	int vf;
6507 	struct ixgbe_hw *hw = &adapter->hw;
6508 	struct net_device *netdev = adapter->netdev;
6509 	u32 gpc;
6510 	u32 ciaa, ciad;
6511 
6512 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
6513 	if (gpc) /* If incrementing then no need for the check below */
6514 		return;
6515 	/*
6516 	 * Check to see if a bad DMA write target from an errant or
6517 	 * malicious VF has caused a PCIe error.  If so then we can
6518 	 * issue a VFLR to the offending VF(s) and then resume without
6519 	 * requesting a full slot reset.
6520 	 */
6521 
6522 	for (vf = 0; vf < adapter->num_vfs; vf++) {
6523 		ciaa = (vf << 16) | 0x80000000;
6524 		/* 32 bit read so align, we really want status at offset 6 */
6525 		ciaa |= PCI_COMMAND;
6526 		IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
6527 		ciad = IXGBE_READ_REG(hw, IXGBE_CIAD_82599);
6528 		ciaa &= 0x7FFFFFFF;
6529 		/* disable debug mode asap after reading data */
6530 		IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
6531 		/* Get the upper 16 bits which will be the PCI status reg */
6532 		ciad >>= 16;
6533 		if (ciad & PCI_STATUS_REC_MASTER_ABORT) {
6534 			netdev_err(netdev, "VF %d Hung DMA\n", vf);
6535 			/* Issue VFLR */
6536 			ciaa = (vf << 16) | 0x80000000;
6537 			ciaa |= 0xA8;
6538 			IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
6539 			ciad = 0x00008000;  /* VFLR */
6540 			IXGBE_WRITE_REG(hw, IXGBE_CIAD_82599, ciad);
6541 			ciaa &= 0x7FFFFFFF;
6542 			IXGBE_WRITE_REG(hw, IXGBE_CIAA_82599, ciaa);
6543 		}
6544 	}
6545 }
6546 
6547 #endif
6548 /**
6549  * ixgbe_service_timer - Timer Call-back
6550  * @data: pointer to adapter cast into an unsigned long
6551  **/
6552 static void ixgbe_service_timer(unsigned long data)
6553 {
6554 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
6555 	unsigned long next_event_offset;
6556 	bool ready = true;
6557 
6558 	/* poll faster when waiting for link */
6559 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
6560 		next_event_offset = HZ / 10;
6561 	else
6562 		next_event_offset = HZ * 2;
6563 
6564 #ifdef CONFIG_PCI_IOV
6565 	/*
6566 	 * don't bother with SR-IOV VF DMA hang check if there are
6567 	 * no VFs or the link is down
6568 	 */
6569 	if (!adapter->num_vfs ||
6570 	    (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
6571 		goto normal_timer_service;
6572 
6573 	/* If we have VFs allocated then we must check for DMA hangs */
6574 	ixgbe_check_for_bad_vf(adapter);
6575 	next_event_offset = HZ / 50;
6576 	adapter->timer_event_accumulator++;
6577 
6578 	if (adapter->timer_event_accumulator >= 100)
6579 		adapter->timer_event_accumulator = 0;
6580 	else
6581 		ready = false;
6582 
6583 normal_timer_service:
6584 #endif
6585 	/* Reset the timer */
6586 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
6587 
6588 	if (ready)
6589 		ixgbe_service_event_schedule(adapter);
6590 }
6591 
6592 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
6593 {
6594 	if (!(adapter->flags2 & IXGBE_FLAG2_RESET_REQUESTED))
6595 		return;
6596 
6597 	adapter->flags2 &= ~IXGBE_FLAG2_RESET_REQUESTED;
6598 
6599 	/* If we're already down, removing or resetting, just bail */
6600 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6601 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6602 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6603 		return;
6604 
6605 	ixgbe_dump(adapter);
6606 	netdev_err(adapter->netdev, "Reset adapter\n");
6607 	adapter->tx_timeout_count++;
6608 
6609 	rtnl_lock();
6610 	ixgbe_reinit_locked(adapter);
6611 	rtnl_unlock();
6612 }
6613 
6614 /**
6615  * ixgbe_service_task - manages and runs subtasks
6616  * @work: pointer to work_struct containing our data
6617  **/
6618 static void ixgbe_service_task(struct work_struct *work)
6619 {
6620 	struct ixgbe_adapter *adapter = container_of(work,
6621 						     struct ixgbe_adapter,
6622 						     service_task);
6623 	if (ixgbe_removed(adapter->hw.hw_addr)) {
6624 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
6625 			rtnl_lock();
6626 			ixgbe_down(adapter);
6627 			rtnl_unlock();
6628 		}
6629 		ixgbe_service_event_complete(adapter);
6630 		return;
6631 	}
6632 	ixgbe_reset_subtask(adapter);
6633 	ixgbe_sfp_detection_subtask(adapter);
6634 	ixgbe_sfp_link_config_subtask(adapter);
6635 	ixgbe_check_overtemp_subtask(adapter);
6636 	ixgbe_watchdog_subtask(adapter);
6637 	ixgbe_fdir_reinit_subtask(adapter);
6638 	ixgbe_check_hang_subtask(adapter);
6639 
6640 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
6641 		ixgbe_ptp_overflow_check(adapter);
6642 		ixgbe_ptp_rx_hang(adapter);
6643 	}
6644 
6645 	ixgbe_service_event_complete(adapter);
6646 }
6647 
6648 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
6649 		     struct ixgbe_tx_buffer *first,
6650 		     u8 *hdr_len)
6651 {
6652 	struct sk_buff *skb = first->skb;
6653 	u32 vlan_macip_lens, type_tucmd;
6654 	u32 mss_l4len_idx, l4len;
6655 	int err;
6656 
6657 	if (skb->ip_summed != CHECKSUM_PARTIAL)
6658 		return 0;
6659 
6660 	if (!skb_is_gso(skb))
6661 		return 0;
6662 
6663 	err = skb_cow_head(skb, 0);
6664 	if (err < 0)
6665 		return err;
6666 
6667 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
6668 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
6669 
6670 	if (first->protocol == htons(ETH_P_IP)) {
6671 		struct iphdr *iph = ip_hdr(skb);
6672 		iph->tot_len = 0;
6673 		iph->check = 0;
6674 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
6675 							 iph->daddr, 0,
6676 							 IPPROTO_TCP,
6677 							 0);
6678 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
6679 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
6680 				   IXGBE_TX_FLAGS_CSUM |
6681 				   IXGBE_TX_FLAGS_IPV4;
6682 	} else if (skb_is_gso_v6(skb)) {
6683 		ipv6_hdr(skb)->payload_len = 0;
6684 		tcp_hdr(skb)->check =
6685 		    ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
6686 				     &ipv6_hdr(skb)->daddr,
6687 				     0, IPPROTO_TCP, 0);
6688 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
6689 				   IXGBE_TX_FLAGS_CSUM;
6690 	}
6691 
6692 	/* compute header lengths */
6693 	l4len = tcp_hdrlen(skb);
6694 	*hdr_len = skb_transport_offset(skb) + l4len;
6695 
6696 	/* update gso size and bytecount with header size */
6697 	first->gso_segs = skb_shinfo(skb)->gso_segs;
6698 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
6699 
6700 	/* mss_l4len_id: use 0 as index for TSO */
6701 	mss_l4len_idx = l4len << IXGBE_ADVTXD_L4LEN_SHIFT;
6702 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
6703 
6704 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
6705 	vlan_macip_lens = skb_network_header_len(skb);
6706 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
6707 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
6708 
6709 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
6710 			  mss_l4len_idx);
6711 
6712 	return 1;
6713 }
6714 
6715 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
6716 			  struct ixgbe_tx_buffer *first)
6717 {
6718 	struct sk_buff *skb = first->skb;
6719 	u32 vlan_macip_lens = 0;
6720 	u32 mss_l4len_idx = 0;
6721 	u32 type_tucmd = 0;
6722 
6723 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
6724 		if (!(first->tx_flags & IXGBE_TX_FLAGS_HW_VLAN) &&
6725 		    !(first->tx_flags & IXGBE_TX_FLAGS_CC))
6726 			return;
6727 	} else {
6728 		u8 l4_hdr = 0;
6729 		switch (first->protocol) {
6730 		case htons(ETH_P_IP):
6731 			vlan_macip_lens |= skb_network_header_len(skb);
6732 			type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
6733 			l4_hdr = ip_hdr(skb)->protocol;
6734 			break;
6735 		case htons(ETH_P_IPV6):
6736 			vlan_macip_lens |= skb_network_header_len(skb);
6737 			l4_hdr = ipv6_hdr(skb)->nexthdr;
6738 			break;
6739 		default:
6740 			if (unlikely(net_ratelimit())) {
6741 				dev_warn(tx_ring->dev,
6742 				 "partial checksum but proto=%x!\n",
6743 				 first->protocol);
6744 			}
6745 			break;
6746 		}
6747 
6748 		switch (l4_hdr) {
6749 		case IPPROTO_TCP:
6750 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_TCP;
6751 			mss_l4len_idx = tcp_hdrlen(skb) <<
6752 					IXGBE_ADVTXD_L4LEN_SHIFT;
6753 			break;
6754 		case IPPROTO_SCTP:
6755 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_SCTP;
6756 			mss_l4len_idx = sizeof(struct sctphdr) <<
6757 					IXGBE_ADVTXD_L4LEN_SHIFT;
6758 			break;
6759 		case IPPROTO_UDP:
6760 			mss_l4len_idx = sizeof(struct udphdr) <<
6761 					IXGBE_ADVTXD_L4LEN_SHIFT;
6762 			break;
6763 		default:
6764 			if (unlikely(net_ratelimit())) {
6765 				dev_warn(tx_ring->dev,
6766 				 "partial checksum but l4 proto=%x!\n",
6767 				 l4_hdr);
6768 			}
6769 			break;
6770 		}
6771 
6772 		/* update TX checksum flag */
6773 		first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
6774 	}
6775 
6776 	/* vlan_macip_lens: MACLEN, VLAN tag */
6777 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
6778 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
6779 
6780 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0,
6781 			  type_tucmd, mss_l4len_idx);
6782 }
6783 
6784 #define IXGBE_SET_FLAG(_input, _flag, _result) \
6785 	((_flag <= _result) ? \
6786 	 ((u32)(_input & _flag) * (_result / _flag)) : \
6787 	 ((u32)(_input & _flag) / (_flag / _result)))
6788 
6789 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
6790 {
6791 	/* set type for advanced descriptor with frame checksum insertion */
6792 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
6793 		       IXGBE_ADVTXD_DCMD_DEXT |
6794 		       IXGBE_ADVTXD_DCMD_IFCS;
6795 
6796 	/* set HW vlan bit if vlan is present */
6797 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
6798 				   IXGBE_ADVTXD_DCMD_VLE);
6799 
6800 	/* set segmentation enable bits for TSO/FSO */
6801 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
6802 				   IXGBE_ADVTXD_DCMD_TSE);
6803 
6804 	/* set timestamp bit if present */
6805 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
6806 				   IXGBE_ADVTXD_MAC_TSTAMP);
6807 
6808 	/* insert frame checksum */
6809 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
6810 
6811 	return cmd_type;
6812 }
6813 
6814 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
6815 				   u32 tx_flags, unsigned int paylen)
6816 {
6817 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
6818 
6819 	/* enable L4 checksum for TSO and TX checksum offload */
6820 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
6821 					IXGBE_TX_FLAGS_CSUM,
6822 					IXGBE_ADVTXD_POPTS_TXSM);
6823 
6824 	/* enble IPv4 checksum for TSO */
6825 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
6826 					IXGBE_TX_FLAGS_IPV4,
6827 					IXGBE_ADVTXD_POPTS_IXSM);
6828 
6829 	/*
6830 	 * Check Context must be set if Tx switch is enabled, which it
6831 	 * always is for case where virtual functions are running
6832 	 */
6833 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
6834 					IXGBE_TX_FLAGS_CC,
6835 					IXGBE_ADVTXD_CC);
6836 
6837 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
6838 }
6839 
6840 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
6841 		       IXGBE_TXD_CMD_RS)
6842 
6843 static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
6844 			 struct ixgbe_tx_buffer *first,
6845 			 const u8 hdr_len)
6846 {
6847 	struct sk_buff *skb = first->skb;
6848 	struct ixgbe_tx_buffer *tx_buffer;
6849 	union ixgbe_adv_tx_desc *tx_desc;
6850 	struct skb_frag_struct *frag;
6851 	dma_addr_t dma;
6852 	unsigned int data_len, size;
6853 	u32 tx_flags = first->tx_flags;
6854 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
6855 	u16 i = tx_ring->next_to_use;
6856 
6857 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
6858 
6859 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
6860 
6861 	size = skb_headlen(skb);
6862 	data_len = skb->data_len;
6863 
6864 #ifdef IXGBE_FCOE
6865 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
6866 		if (data_len < sizeof(struct fcoe_crc_eof)) {
6867 			size -= sizeof(struct fcoe_crc_eof) - data_len;
6868 			data_len = 0;
6869 		} else {
6870 			data_len -= sizeof(struct fcoe_crc_eof);
6871 		}
6872 	}
6873 
6874 #endif
6875 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
6876 
6877 	tx_buffer = first;
6878 
6879 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
6880 		if (dma_mapping_error(tx_ring->dev, dma))
6881 			goto dma_error;
6882 
6883 		/* record length, and DMA address */
6884 		dma_unmap_len_set(tx_buffer, len, size);
6885 		dma_unmap_addr_set(tx_buffer, dma, dma);
6886 
6887 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
6888 
6889 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
6890 			tx_desc->read.cmd_type_len =
6891 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
6892 
6893 			i++;
6894 			tx_desc++;
6895 			if (i == tx_ring->count) {
6896 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6897 				i = 0;
6898 			}
6899 			tx_desc->read.olinfo_status = 0;
6900 
6901 			dma += IXGBE_MAX_DATA_PER_TXD;
6902 			size -= IXGBE_MAX_DATA_PER_TXD;
6903 
6904 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
6905 		}
6906 
6907 		if (likely(!data_len))
6908 			break;
6909 
6910 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
6911 
6912 		i++;
6913 		tx_desc++;
6914 		if (i == tx_ring->count) {
6915 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6916 			i = 0;
6917 		}
6918 		tx_desc->read.olinfo_status = 0;
6919 
6920 #ifdef IXGBE_FCOE
6921 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
6922 #else
6923 		size = skb_frag_size(frag);
6924 #endif
6925 		data_len -= size;
6926 
6927 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
6928 				       DMA_TO_DEVICE);
6929 
6930 		tx_buffer = &tx_ring->tx_buffer_info[i];
6931 	}
6932 
6933 	/* write last descriptor with RS and EOP bits */
6934 	cmd_type |= size | IXGBE_TXD_CMD;
6935 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
6936 
6937 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
6938 
6939 	/* set the timestamp */
6940 	first->time_stamp = jiffies;
6941 
6942 	/*
6943 	 * Force memory writes to complete before letting h/w know there
6944 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
6945 	 * memory model archs, such as IA-64).
6946 	 *
6947 	 * We also need this memory barrier to make certain all of the
6948 	 * status bits have been updated before next_to_watch is written.
6949 	 */
6950 	wmb();
6951 
6952 	/* set next_to_watch value indicating a packet is present */
6953 	first->next_to_watch = tx_desc;
6954 
6955 	i++;
6956 	if (i == tx_ring->count)
6957 		i = 0;
6958 
6959 	tx_ring->next_to_use = i;
6960 
6961 	/* notify HW of packet */
6962 	ixgbe_write_tail(tx_ring, i);
6963 
6964 	return;
6965 dma_error:
6966 	dev_err(tx_ring->dev, "TX DMA map failed\n");
6967 
6968 	/* clear dma mappings for failed tx_buffer_info map */
6969 	for (;;) {
6970 		tx_buffer = &tx_ring->tx_buffer_info[i];
6971 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer);
6972 		if (tx_buffer == first)
6973 			break;
6974 		if (i == 0)
6975 			i = tx_ring->count;
6976 		i--;
6977 	}
6978 
6979 	tx_ring->next_to_use = i;
6980 }
6981 
6982 static void ixgbe_atr(struct ixgbe_ring *ring,
6983 		      struct ixgbe_tx_buffer *first)
6984 {
6985 	struct ixgbe_q_vector *q_vector = ring->q_vector;
6986 	union ixgbe_atr_hash_dword input = { .dword = 0 };
6987 	union ixgbe_atr_hash_dword common = { .dword = 0 };
6988 	union {
6989 		unsigned char *network;
6990 		struct iphdr *ipv4;
6991 		struct ipv6hdr *ipv6;
6992 	} hdr;
6993 	struct tcphdr *th;
6994 	__be16 vlan_id;
6995 
6996 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
6997 	if (!q_vector)
6998 		return;
6999 
7000 	/* do nothing if sampling is disabled */
7001 	if (!ring->atr_sample_rate)
7002 		return;
7003 
7004 	ring->atr_count++;
7005 
7006 	/* snag network header to get L4 type and address */
7007 	hdr.network = skb_network_header(first->skb);
7008 
7009 	/* Currently only IPv4/IPv6 with TCP is supported */
7010 	if ((first->protocol != htons(ETH_P_IPV6) ||
7011 	     hdr.ipv6->nexthdr != IPPROTO_TCP) &&
7012 	    (first->protocol != htons(ETH_P_IP) ||
7013 	     hdr.ipv4->protocol != IPPROTO_TCP))
7014 		return;
7015 
7016 	th = tcp_hdr(first->skb);
7017 
7018 	/* skip this packet since it is invalid or the socket is closing */
7019 	if (!th || th->fin)
7020 		return;
7021 
7022 	/* sample on all syn packets or once every atr sample count */
7023 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
7024 		return;
7025 
7026 	/* reset sample count */
7027 	ring->atr_count = 0;
7028 
7029 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
7030 
7031 	/*
7032 	 * src and dst are inverted, think how the receiver sees them
7033 	 *
7034 	 * The input is broken into two sections, a non-compressed section
7035 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
7036 	 * is XORed together and stored in the compressed dword.
7037 	 */
7038 	input.formatted.vlan_id = vlan_id;
7039 
7040 	/*
7041 	 * since src port and flex bytes occupy the same word XOR them together
7042 	 * and write the value to source port portion of compressed dword
7043 	 */
7044 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
7045 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
7046 	else
7047 		common.port.src ^= th->dest ^ first->protocol;
7048 	common.port.dst ^= th->source;
7049 
7050 	if (first->protocol == htons(ETH_P_IP)) {
7051 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
7052 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
7053 	} else {
7054 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
7055 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
7056 			     hdr.ipv6->saddr.s6_addr32[1] ^
7057 			     hdr.ipv6->saddr.s6_addr32[2] ^
7058 			     hdr.ipv6->saddr.s6_addr32[3] ^
7059 			     hdr.ipv6->daddr.s6_addr32[0] ^
7060 			     hdr.ipv6->daddr.s6_addr32[1] ^
7061 			     hdr.ipv6->daddr.s6_addr32[2] ^
7062 			     hdr.ipv6->daddr.s6_addr32[3];
7063 	}
7064 
7065 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
7066 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
7067 					      input, common, ring->queue_index);
7068 }
7069 
7070 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7071 {
7072 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
7073 	/* Herbert's original patch had:
7074 	 *  smp_mb__after_netif_stop_queue();
7075 	 * but since that doesn't exist yet, just open code it. */
7076 	smp_mb();
7077 
7078 	/* We need to check again in a case another CPU has just
7079 	 * made room available. */
7080 	if (likely(ixgbe_desc_unused(tx_ring) < size))
7081 		return -EBUSY;
7082 
7083 	/* A reprieve! - use start_queue because it doesn't call schedule */
7084 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
7085 	++tx_ring->tx_stats.restart_queue;
7086 	return 0;
7087 }
7088 
7089 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7090 {
7091 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
7092 		return 0;
7093 	return __ixgbe_maybe_stop_tx(tx_ring, size);
7094 }
7095 
7096 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
7097 			      void *accel_priv, select_queue_fallback_t fallback)
7098 {
7099 	struct ixgbe_fwd_adapter *fwd_adapter = accel_priv;
7100 #ifdef IXGBE_FCOE
7101 	struct ixgbe_adapter *adapter;
7102 	struct ixgbe_ring_feature *f;
7103 	int txq;
7104 #endif
7105 
7106 	if (fwd_adapter)
7107 		return skb->queue_mapping + fwd_adapter->tx_base_queue;
7108 
7109 #ifdef IXGBE_FCOE
7110 
7111 	/*
7112 	 * only execute the code below if protocol is FCoE
7113 	 * or FIP and we have FCoE enabled on the adapter
7114 	 */
7115 	switch (vlan_get_protocol(skb)) {
7116 	case htons(ETH_P_FCOE):
7117 	case htons(ETH_P_FIP):
7118 		adapter = netdev_priv(dev);
7119 
7120 		if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
7121 			break;
7122 	default:
7123 		return fallback(dev, skb);
7124 	}
7125 
7126 	f = &adapter->ring_feature[RING_F_FCOE];
7127 
7128 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
7129 					   smp_processor_id();
7130 
7131 	while (txq >= f->indices)
7132 		txq -= f->indices;
7133 
7134 	return txq + f->offset;
7135 #else
7136 	return fallback(dev, skb);
7137 #endif
7138 }
7139 
7140 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
7141 			  struct ixgbe_adapter *adapter,
7142 			  struct ixgbe_ring *tx_ring)
7143 {
7144 	struct ixgbe_tx_buffer *first;
7145 	int tso;
7146 	u32 tx_flags = 0;
7147 	unsigned short f;
7148 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
7149 	__be16 protocol = skb->protocol;
7150 	u8 hdr_len = 0;
7151 
7152 	/*
7153 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
7154 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
7155 	 *       + 2 desc gap to keep tail from touching head,
7156 	 *       + 1 desc for context descriptor,
7157 	 * otherwise try next time
7158 	 */
7159 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
7160 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
7161 
7162 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
7163 		tx_ring->tx_stats.tx_busy++;
7164 		return NETDEV_TX_BUSY;
7165 	}
7166 
7167 	/* record the location of the first descriptor for this packet */
7168 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
7169 	first->skb = skb;
7170 	first->bytecount = skb->len;
7171 	first->gso_segs = 1;
7172 
7173 	/* if we have a HW VLAN tag being added default to the HW one */
7174 	if (vlan_tx_tag_present(skb)) {
7175 		tx_flags |= vlan_tx_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
7176 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7177 	/* else if it is a SW VLAN check the next protocol and store the tag */
7178 	} else if (protocol == htons(ETH_P_8021Q)) {
7179 		struct vlan_hdr *vhdr, _vhdr;
7180 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
7181 		if (!vhdr)
7182 			goto out_drop;
7183 
7184 		protocol = vhdr->h_vlan_encapsulated_proto;
7185 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
7186 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
7187 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
7188 	}
7189 
7190 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP &&
7191 		     !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
7192 					    &adapter->state))) {
7193 		skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
7194 		tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
7195 
7196 		/* schedule check for Tx timestamp */
7197 		adapter->ptp_tx_skb = skb_get(skb);
7198 		adapter->ptp_tx_start = jiffies;
7199 		schedule_work(&adapter->ptp_tx_work);
7200 	}
7201 
7202 	skb_tx_timestamp(skb);
7203 
7204 #ifdef CONFIG_PCI_IOV
7205 	/*
7206 	 * Use the l2switch_enable flag - would be false if the DMA
7207 	 * Tx switch had been disabled.
7208 	 */
7209 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7210 		tx_flags |= IXGBE_TX_FLAGS_CC;
7211 
7212 #endif
7213 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
7214 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
7215 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
7216 	     (skb->priority != TC_PRIO_CONTROL))) {
7217 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
7218 		tx_flags |= (skb->priority & 0x7) <<
7219 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
7220 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
7221 			struct vlan_ethhdr *vhdr;
7222 
7223 			if (skb_cow_head(skb, 0))
7224 				goto out_drop;
7225 			vhdr = (struct vlan_ethhdr *)skb->data;
7226 			vhdr->h_vlan_TCI = htons(tx_flags >>
7227 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
7228 		} else {
7229 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7230 		}
7231 	}
7232 
7233 	/* record initial flags and protocol */
7234 	first->tx_flags = tx_flags;
7235 	first->protocol = protocol;
7236 
7237 #ifdef IXGBE_FCOE
7238 	/* setup tx offload for FCoE */
7239 	if ((protocol == htons(ETH_P_FCOE)) &&
7240 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
7241 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
7242 		if (tso < 0)
7243 			goto out_drop;
7244 
7245 		goto xmit_fcoe;
7246 	}
7247 
7248 #endif /* IXGBE_FCOE */
7249 	tso = ixgbe_tso(tx_ring, first, &hdr_len);
7250 	if (tso < 0)
7251 		goto out_drop;
7252 	else if (!tso)
7253 		ixgbe_tx_csum(tx_ring, first);
7254 
7255 	/* add the ATR filter if ATR is on */
7256 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
7257 		ixgbe_atr(tx_ring, first);
7258 
7259 #ifdef IXGBE_FCOE
7260 xmit_fcoe:
7261 #endif /* IXGBE_FCOE */
7262 	ixgbe_tx_map(tx_ring, first, hdr_len);
7263 
7264 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
7265 
7266 	return NETDEV_TX_OK;
7267 
7268 out_drop:
7269 	dev_kfree_skb_any(first->skb);
7270 	first->skb = NULL;
7271 
7272 	return NETDEV_TX_OK;
7273 }
7274 
7275 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
7276 				      struct net_device *netdev,
7277 				      struct ixgbe_ring *ring)
7278 {
7279 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7280 	struct ixgbe_ring *tx_ring;
7281 
7282 	/*
7283 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
7284 	 * in order to meet this minimum size requirement.
7285 	 */
7286 	if (unlikely(skb->len < 17)) {
7287 		if (skb_pad(skb, 17 - skb->len))
7288 			return NETDEV_TX_OK;
7289 		skb->len = 17;
7290 		skb_set_tail_pointer(skb, 17);
7291 	}
7292 
7293 	tx_ring = ring ? ring : adapter->tx_ring[skb->queue_mapping];
7294 
7295 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
7296 }
7297 
7298 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
7299 				    struct net_device *netdev)
7300 {
7301 	return __ixgbe_xmit_frame(skb, netdev, NULL);
7302 }
7303 
7304 /**
7305  * ixgbe_set_mac - Change the Ethernet Address of the NIC
7306  * @netdev: network interface device structure
7307  * @p: pointer to an address structure
7308  *
7309  * Returns 0 on success, negative on failure
7310  **/
7311 static int ixgbe_set_mac(struct net_device *netdev, void *p)
7312 {
7313 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7314 	struct ixgbe_hw *hw = &adapter->hw;
7315 	struct sockaddr *addr = p;
7316 	int ret;
7317 
7318 	if (!is_valid_ether_addr(addr->sa_data))
7319 		return -EADDRNOTAVAIL;
7320 
7321 	ixgbe_del_mac_filter(adapter, hw->mac.addr, VMDQ_P(0));
7322 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
7323 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
7324 
7325 	ret = ixgbe_add_mac_filter(adapter, hw->mac.addr, VMDQ_P(0));
7326 	return ret > 0 ? 0 : ret;
7327 }
7328 
7329 static int
7330 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
7331 {
7332 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7333 	struct ixgbe_hw *hw = &adapter->hw;
7334 	u16 value;
7335 	int rc;
7336 
7337 	if (prtad != hw->phy.mdio.prtad)
7338 		return -EINVAL;
7339 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
7340 	if (!rc)
7341 		rc = value;
7342 	return rc;
7343 }
7344 
7345 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
7346 			    u16 addr, u16 value)
7347 {
7348 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7349 	struct ixgbe_hw *hw = &adapter->hw;
7350 
7351 	if (prtad != hw->phy.mdio.prtad)
7352 		return -EINVAL;
7353 	return hw->phy.ops.write_reg(hw, addr, devad, value);
7354 }
7355 
7356 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
7357 {
7358 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7359 
7360 	switch (cmd) {
7361 	case SIOCSHWTSTAMP:
7362 		return ixgbe_ptp_set_ts_config(adapter, req);
7363 	case SIOCGHWTSTAMP:
7364 		return ixgbe_ptp_get_ts_config(adapter, req);
7365 	default:
7366 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
7367 	}
7368 }
7369 
7370 /**
7371  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
7372  * netdev->dev_addrs
7373  * @netdev: network interface device structure
7374  *
7375  * Returns non-zero on failure
7376  **/
7377 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
7378 {
7379 	int err = 0;
7380 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7381 	struct ixgbe_hw *hw = &adapter->hw;
7382 
7383 	if (is_valid_ether_addr(hw->mac.san_addr)) {
7384 		rtnl_lock();
7385 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
7386 		rtnl_unlock();
7387 
7388 		/* update SAN MAC vmdq pool selection */
7389 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
7390 	}
7391 	return err;
7392 }
7393 
7394 /**
7395  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
7396  * netdev->dev_addrs
7397  * @netdev: network interface device structure
7398  *
7399  * Returns non-zero on failure
7400  **/
7401 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
7402 {
7403 	int err = 0;
7404 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7405 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
7406 
7407 	if (is_valid_ether_addr(mac->san_addr)) {
7408 		rtnl_lock();
7409 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
7410 		rtnl_unlock();
7411 	}
7412 	return err;
7413 }
7414 
7415 #ifdef CONFIG_NET_POLL_CONTROLLER
7416 /*
7417  * Polling 'interrupt' - used by things like netconsole to send skbs
7418  * without having to re-enable interrupts. It's not called while
7419  * the interrupt routine is executing.
7420  */
7421 static void ixgbe_netpoll(struct net_device *netdev)
7422 {
7423 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7424 	int i;
7425 
7426 	/* if interface is down do nothing */
7427 	if (test_bit(__IXGBE_DOWN, &adapter->state))
7428 		return;
7429 
7430 	adapter->flags |= IXGBE_FLAG_IN_NETPOLL;
7431 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
7432 		for (i = 0; i < adapter->num_q_vectors; i++)
7433 			ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
7434 	} else {
7435 		ixgbe_intr(adapter->pdev->irq, netdev);
7436 	}
7437 	adapter->flags &= ~IXGBE_FLAG_IN_NETPOLL;
7438 }
7439 
7440 #endif
7441 static struct rtnl_link_stats64 *ixgbe_get_stats64(struct net_device *netdev,
7442 						   struct rtnl_link_stats64 *stats)
7443 {
7444 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7445 	int i;
7446 
7447 	rcu_read_lock();
7448 	for (i = 0; i < adapter->num_rx_queues; i++) {
7449 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
7450 		u64 bytes, packets;
7451 		unsigned int start;
7452 
7453 		if (ring) {
7454 			do {
7455 				start = u64_stats_fetch_begin_irq(&ring->syncp);
7456 				packets = ring->stats.packets;
7457 				bytes   = ring->stats.bytes;
7458 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
7459 			stats->rx_packets += packets;
7460 			stats->rx_bytes   += bytes;
7461 		}
7462 	}
7463 
7464 	for (i = 0; i < adapter->num_tx_queues; i++) {
7465 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
7466 		u64 bytes, packets;
7467 		unsigned int start;
7468 
7469 		if (ring) {
7470 			do {
7471 				start = u64_stats_fetch_begin_irq(&ring->syncp);
7472 				packets = ring->stats.packets;
7473 				bytes   = ring->stats.bytes;
7474 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
7475 			stats->tx_packets += packets;
7476 			stats->tx_bytes   += bytes;
7477 		}
7478 	}
7479 	rcu_read_unlock();
7480 	/* following stats updated by ixgbe_watchdog_task() */
7481 	stats->multicast	= netdev->stats.multicast;
7482 	stats->rx_errors	= netdev->stats.rx_errors;
7483 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
7484 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
7485 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
7486 	return stats;
7487 }
7488 
7489 #ifdef CONFIG_IXGBE_DCB
7490 /**
7491  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
7492  * @adapter: pointer to ixgbe_adapter
7493  * @tc: number of traffic classes currently enabled
7494  *
7495  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
7496  * 802.1Q priority maps to a packet buffer that exists.
7497  */
7498 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
7499 {
7500 	struct ixgbe_hw *hw = &adapter->hw;
7501 	u32 reg, rsave;
7502 	int i;
7503 
7504 	/* 82598 have a static priority to TC mapping that can not
7505 	 * be changed so no validation is needed.
7506 	 */
7507 	if (hw->mac.type == ixgbe_mac_82598EB)
7508 		return;
7509 
7510 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
7511 	rsave = reg;
7512 
7513 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
7514 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
7515 
7516 		/* If up2tc is out of bounds default to zero */
7517 		if (up2tc > tc)
7518 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
7519 	}
7520 
7521 	if (reg != rsave)
7522 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
7523 
7524 	return;
7525 }
7526 
7527 /**
7528  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
7529  * @adapter: Pointer to adapter struct
7530  *
7531  * Populate the netdev user priority to tc map
7532  */
7533 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
7534 {
7535 	struct net_device *dev = adapter->netdev;
7536 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
7537 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
7538 	u8 prio;
7539 
7540 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
7541 		u8 tc = 0;
7542 
7543 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
7544 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
7545 		else if (ets)
7546 			tc = ets->prio_tc[prio];
7547 
7548 		netdev_set_prio_tc_map(dev, prio, tc);
7549 	}
7550 }
7551 
7552 #endif /* CONFIG_IXGBE_DCB */
7553 /**
7554  * ixgbe_setup_tc - configure net_device for multiple traffic classes
7555  *
7556  * @netdev: net device to configure
7557  * @tc: number of traffic classes to enable
7558  */
7559 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
7560 {
7561 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7562 	struct ixgbe_hw *hw = &adapter->hw;
7563 	bool pools;
7564 
7565 	/* Hardware supports up to 8 traffic classes */
7566 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs ||
7567 	    (hw->mac.type == ixgbe_mac_82598EB &&
7568 	     tc < MAX_TRAFFIC_CLASS))
7569 		return -EINVAL;
7570 
7571 	pools = (find_first_zero_bit(&adapter->fwd_bitmask, 32) > 1);
7572 	if (tc && pools && adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS)
7573 		return -EBUSY;
7574 
7575 	/* Hardware has to reinitialize queues and interrupts to
7576 	 * match packet buffer alignment. Unfortunately, the
7577 	 * hardware is not flexible enough to do this dynamically.
7578 	 */
7579 	if (netif_running(dev))
7580 		ixgbe_close(dev);
7581 	ixgbe_clear_interrupt_scheme(adapter);
7582 
7583 #ifdef CONFIG_IXGBE_DCB
7584 	if (tc) {
7585 		netdev_set_num_tc(dev, tc);
7586 		ixgbe_set_prio_tc_map(adapter);
7587 
7588 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
7589 
7590 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
7591 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
7592 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
7593 		}
7594 	} else {
7595 		netdev_reset_tc(dev);
7596 
7597 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
7598 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
7599 
7600 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
7601 
7602 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
7603 		adapter->dcb_cfg.pfc_mode_enable = false;
7604 	}
7605 
7606 	ixgbe_validate_rtr(adapter, tc);
7607 
7608 #endif /* CONFIG_IXGBE_DCB */
7609 	ixgbe_init_interrupt_scheme(adapter);
7610 
7611 	if (netif_running(dev))
7612 		return ixgbe_open(dev);
7613 
7614 	return 0;
7615 }
7616 
7617 #ifdef CONFIG_PCI_IOV
7618 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
7619 {
7620 	struct net_device *netdev = adapter->netdev;
7621 
7622 	rtnl_lock();
7623 	ixgbe_setup_tc(netdev, netdev_get_num_tc(netdev));
7624 	rtnl_unlock();
7625 }
7626 
7627 #endif
7628 void ixgbe_do_reset(struct net_device *netdev)
7629 {
7630 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7631 
7632 	if (netif_running(netdev))
7633 		ixgbe_reinit_locked(adapter);
7634 	else
7635 		ixgbe_reset(adapter);
7636 }
7637 
7638 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
7639 					    netdev_features_t features)
7640 {
7641 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7642 
7643 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
7644 	if (!(features & NETIF_F_RXCSUM))
7645 		features &= ~NETIF_F_LRO;
7646 
7647 	/* Turn off LRO if not RSC capable */
7648 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
7649 		features &= ~NETIF_F_LRO;
7650 
7651 	return features;
7652 }
7653 
7654 static int ixgbe_set_features(struct net_device *netdev,
7655 			      netdev_features_t features)
7656 {
7657 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7658 	netdev_features_t changed = netdev->features ^ features;
7659 	bool need_reset = false;
7660 
7661 	/* Make sure RSC matches LRO, reset if change */
7662 	if (!(features & NETIF_F_LRO)) {
7663 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
7664 			need_reset = true;
7665 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
7666 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
7667 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
7668 		if (adapter->rx_itr_setting == 1 ||
7669 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
7670 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
7671 			need_reset = true;
7672 		} else if ((changed ^ features) & NETIF_F_LRO) {
7673 			e_info(probe, "rx-usecs set too low, "
7674 			       "disabling RSC\n");
7675 		}
7676 	}
7677 
7678 	/*
7679 	 * Check if Flow Director n-tuple support was enabled or disabled.  If
7680 	 * the state changed, we need to reset.
7681 	 */
7682 	switch (features & NETIF_F_NTUPLE) {
7683 	case NETIF_F_NTUPLE:
7684 		/* turn off ATR, enable perfect filters and reset */
7685 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
7686 			need_reset = true;
7687 
7688 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
7689 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
7690 		break;
7691 	default:
7692 		/* turn off perfect filters, enable ATR and reset */
7693 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
7694 			need_reset = true;
7695 
7696 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
7697 
7698 		/* We cannot enable ATR if SR-IOV is enabled */
7699 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7700 			break;
7701 
7702 		/* We cannot enable ATR if we have 2 or more traffic classes */
7703 		if (netdev_get_num_tc(netdev) > 1)
7704 			break;
7705 
7706 		/* We cannot enable ATR if RSS is disabled */
7707 		if (adapter->ring_feature[RING_F_RSS].limit <= 1)
7708 			break;
7709 
7710 		/* A sample rate of 0 indicates ATR disabled */
7711 		if (!adapter->atr_sample_rate)
7712 			break;
7713 
7714 		adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
7715 		break;
7716 	}
7717 
7718 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
7719 		ixgbe_vlan_strip_enable(adapter);
7720 	else
7721 		ixgbe_vlan_strip_disable(adapter);
7722 
7723 	if (changed & NETIF_F_RXALL)
7724 		need_reset = true;
7725 
7726 	netdev->features = features;
7727 	if (need_reset)
7728 		ixgbe_do_reset(netdev);
7729 
7730 	return 0;
7731 }
7732 
7733 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
7734 			     struct net_device *dev,
7735 			     const unsigned char *addr,
7736 			     u16 flags)
7737 {
7738 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7739 	int err;
7740 
7741 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
7742 		return ndo_dflt_fdb_add(ndm, tb, dev, addr, flags);
7743 
7744 	/* Hardware does not support aging addresses so if a
7745 	 * ndm_state is given only allow permanent addresses
7746 	 */
7747 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
7748 		pr_info("%s: FDB only supports static addresses\n",
7749 			ixgbe_driver_name);
7750 		return -EINVAL;
7751 	}
7752 
7753 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
7754 		u32 rar_uc_entries = IXGBE_MAX_PF_MACVLANS;
7755 
7756 		if (netdev_uc_count(dev) < rar_uc_entries)
7757 			err = dev_uc_add_excl(dev, addr);
7758 		else
7759 			err = -ENOMEM;
7760 	} else if (is_multicast_ether_addr(addr)) {
7761 		err = dev_mc_add_excl(dev, addr);
7762 	} else {
7763 		err = -EINVAL;
7764 	}
7765 
7766 	/* Only return duplicate errors if NLM_F_EXCL is set */
7767 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
7768 		err = 0;
7769 
7770 	return err;
7771 }
7772 
7773 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
7774 				    struct nlmsghdr *nlh)
7775 {
7776 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7777 	struct nlattr *attr, *br_spec;
7778 	int rem;
7779 
7780 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
7781 		return -EOPNOTSUPP;
7782 
7783 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
7784 
7785 	nla_for_each_nested(attr, br_spec, rem) {
7786 		__u16 mode;
7787 		u32 reg = 0;
7788 
7789 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
7790 			continue;
7791 
7792 		mode = nla_get_u16(attr);
7793 		if (mode == BRIDGE_MODE_VEPA) {
7794 			reg = 0;
7795 			adapter->flags2 &= ~IXGBE_FLAG2_BRIDGE_MODE_VEB;
7796 		} else if (mode == BRIDGE_MODE_VEB) {
7797 			reg = IXGBE_PFDTXGSWC_VT_LBEN;
7798 			adapter->flags2 |= IXGBE_FLAG2_BRIDGE_MODE_VEB;
7799 		} else
7800 			return -EINVAL;
7801 
7802 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, reg);
7803 
7804 		e_info(drv, "enabling bridge mode: %s\n",
7805 			mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
7806 	}
7807 
7808 	return 0;
7809 }
7810 
7811 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
7812 				    struct net_device *dev,
7813 				    u32 filter_mask)
7814 {
7815 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7816 	u16 mode;
7817 
7818 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
7819 		return 0;
7820 
7821 	if (adapter->flags2 & IXGBE_FLAG2_BRIDGE_MODE_VEB)
7822 		mode = BRIDGE_MODE_VEB;
7823 	else
7824 		mode = BRIDGE_MODE_VEPA;
7825 
7826 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev, mode);
7827 }
7828 
7829 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
7830 {
7831 	struct ixgbe_fwd_adapter *fwd_adapter = NULL;
7832 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
7833 	unsigned int limit;
7834 	int pool, err;
7835 
7836 #ifdef CONFIG_RPS
7837 	if (vdev->num_rx_queues != vdev->num_tx_queues) {
7838 		netdev_info(pdev, "%s: Only supports a single queue count for TX and RX\n",
7839 			    vdev->name);
7840 		return ERR_PTR(-EINVAL);
7841 	}
7842 #endif
7843 	/* Check for hardware restriction on number of rx/tx queues */
7844 	if (vdev->num_tx_queues > IXGBE_MAX_L2A_QUEUES ||
7845 	    vdev->num_tx_queues == IXGBE_BAD_L2A_QUEUE) {
7846 		netdev_info(pdev,
7847 			    "%s: Supports RX/TX Queue counts 1,2, and 4\n",
7848 			    pdev->name);
7849 		return ERR_PTR(-EINVAL);
7850 	}
7851 
7852 	if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
7853 	      adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS - 1) ||
7854 	    (adapter->num_rx_pools > IXGBE_MAX_MACVLANS))
7855 		return ERR_PTR(-EBUSY);
7856 
7857 	fwd_adapter = kcalloc(1, sizeof(struct ixgbe_fwd_adapter), GFP_KERNEL);
7858 	if (!fwd_adapter)
7859 		return ERR_PTR(-ENOMEM);
7860 
7861 	pool = find_first_zero_bit(&adapter->fwd_bitmask, 32);
7862 	adapter->num_rx_pools++;
7863 	set_bit(pool, &adapter->fwd_bitmask);
7864 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
7865 
7866 	/* Enable VMDq flag so device will be set in VM mode */
7867 	adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED | IXGBE_FLAG_SRIOV_ENABLED;
7868 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
7869 	adapter->ring_feature[RING_F_RSS].limit = vdev->num_tx_queues;
7870 
7871 	/* Force reinit of ring allocation with VMDQ enabled */
7872 	err = ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
7873 	if (err)
7874 		goto fwd_add_err;
7875 	fwd_adapter->pool = pool;
7876 	fwd_adapter->real_adapter = adapter;
7877 	err = ixgbe_fwd_ring_up(vdev, fwd_adapter);
7878 	if (err)
7879 		goto fwd_add_err;
7880 	netif_tx_start_all_queues(vdev);
7881 	return fwd_adapter;
7882 fwd_add_err:
7883 	/* unwind counter and free adapter struct */
7884 	netdev_info(pdev,
7885 		    "%s: dfwd hardware acceleration failed\n", vdev->name);
7886 	clear_bit(pool, &adapter->fwd_bitmask);
7887 	adapter->num_rx_pools--;
7888 	kfree(fwd_adapter);
7889 	return ERR_PTR(err);
7890 }
7891 
7892 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
7893 {
7894 	struct ixgbe_fwd_adapter *fwd_adapter = priv;
7895 	struct ixgbe_adapter *adapter = fwd_adapter->real_adapter;
7896 	unsigned int limit;
7897 
7898 	clear_bit(fwd_adapter->pool, &adapter->fwd_bitmask);
7899 	adapter->num_rx_pools--;
7900 
7901 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
7902 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
7903 	ixgbe_fwd_ring_down(fwd_adapter->netdev, fwd_adapter);
7904 	ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
7905 	netdev_dbg(pdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
7906 		   fwd_adapter->pool, adapter->num_rx_pools,
7907 		   fwd_adapter->rx_base_queue,
7908 		   fwd_adapter->rx_base_queue + adapter->num_rx_queues_per_pool,
7909 		   adapter->fwd_bitmask);
7910 	kfree(fwd_adapter);
7911 }
7912 
7913 static const struct net_device_ops ixgbe_netdev_ops = {
7914 	.ndo_open		= ixgbe_open,
7915 	.ndo_stop		= ixgbe_close,
7916 	.ndo_start_xmit		= ixgbe_xmit_frame,
7917 	.ndo_select_queue	= ixgbe_select_queue,
7918 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
7919 	.ndo_validate_addr	= eth_validate_addr,
7920 	.ndo_set_mac_address	= ixgbe_set_mac,
7921 	.ndo_change_mtu		= ixgbe_change_mtu,
7922 	.ndo_tx_timeout		= ixgbe_tx_timeout,
7923 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
7924 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
7925 	.ndo_do_ioctl		= ixgbe_ioctl,
7926 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
7927 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
7928 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
7929 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
7930 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
7931 	.ndo_get_stats64	= ixgbe_get_stats64,
7932 #ifdef CONFIG_IXGBE_DCB
7933 	.ndo_setup_tc		= ixgbe_setup_tc,
7934 #endif
7935 #ifdef CONFIG_NET_POLL_CONTROLLER
7936 	.ndo_poll_controller	= ixgbe_netpoll,
7937 #endif
7938 #ifdef CONFIG_NET_RX_BUSY_POLL
7939 	.ndo_busy_poll		= ixgbe_low_latency_recv,
7940 #endif
7941 #ifdef IXGBE_FCOE
7942 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
7943 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
7944 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
7945 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
7946 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
7947 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
7948 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
7949 #endif /* IXGBE_FCOE */
7950 	.ndo_set_features = ixgbe_set_features,
7951 	.ndo_fix_features = ixgbe_fix_features,
7952 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
7953 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
7954 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
7955 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
7956 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
7957 };
7958 
7959 /**
7960  * ixgbe_enumerate_functions - Get the number of ports this device has
7961  * @adapter: adapter structure
7962  *
7963  * This function enumerates the phsyical functions co-located on a single slot,
7964  * in order to determine how many ports a device has. This is most useful in
7965  * determining the required GT/s of PCIe bandwidth necessary for optimal
7966  * performance.
7967  **/
7968 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
7969 {
7970 	struct pci_dev *entry, *pdev = adapter->pdev;
7971 	int physfns = 0;
7972 
7973 	/* Some cards can not use the generic count PCIe functions method,
7974 	 * because they are behind a parent switch, so we hardcode these with
7975 	 * the correct number of functions.
7976 	 */
7977 	if (ixgbe_pcie_from_parent(&adapter->hw))
7978 		physfns = 4;
7979 
7980 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
7981 		/* don't count virtual functions */
7982 		if (entry->is_virtfn)
7983 			continue;
7984 
7985 		/* When the devices on the bus don't all match our device ID,
7986 		 * we can't reliably determine the correct number of
7987 		 * functions. This can occur if a function has been direct
7988 		 * attached to a virtual machine using VT-d, for example. In
7989 		 * this case, simply return -1 to indicate this.
7990 		 */
7991 		if ((entry->vendor != pdev->vendor) ||
7992 		    (entry->device != pdev->device))
7993 			return -1;
7994 
7995 		physfns++;
7996 	}
7997 
7998 	return physfns;
7999 }
8000 
8001 /**
8002  * ixgbe_wol_supported - Check whether device supports WoL
8003  * @hw: hw specific details
8004  * @device_id: the device ID
8005  * @subdev_id: the subsystem device ID
8006  *
8007  * This function is used by probe and ethtool to determine
8008  * which devices have WoL support
8009  *
8010  **/
8011 int ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
8012 			u16 subdevice_id)
8013 {
8014 	struct ixgbe_hw *hw = &adapter->hw;
8015 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
8016 	int is_wol_supported = 0;
8017 
8018 	switch (device_id) {
8019 	case IXGBE_DEV_ID_82599_SFP:
8020 		/* Only these subdevices could supports WOL */
8021 		switch (subdevice_id) {
8022 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
8023 		case IXGBE_SUBDEV_ID_82599_560FLR:
8024 			/* only support first port */
8025 			if (hw->bus.func != 0)
8026 				break;
8027 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
8028 		case IXGBE_SUBDEV_ID_82599_SFP:
8029 		case IXGBE_SUBDEV_ID_82599_RNDC:
8030 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
8031 		case IXGBE_SUBDEV_ID_82599_LOM_SFP:
8032 			is_wol_supported = 1;
8033 			break;
8034 		}
8035 		break;
8036 	case IXGBE_DEV_ID_82599EN_SFP:
8037 		/* Only this subdevice supports WOL */
8038 		switch (subdevice_id) {
8039 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
8040 			is_wol_supported = 1;
8041 			break;
8042 		}
8043 		break;
8044 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
8045 		/* All except this subdevice support WOL */
8046 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
8047 			is_wol_supported = 1;
8048 		break;
8049 	case IXGBE_DEV_ID_82599_KX4:
8050 		is_wol_supported = 1;
8051 		break;
8052 	case IXGBE_DEV_ID_X540T:
8053 	case IXGBE_DEV_ID_X540T1:
8054 		/* check eeprom to see if enabled wol */
8055 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
8056 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
8057 		     (hw->bus.func == 0))) {
8058 			is_wol_supported = 1;
8059 		}
8060 		break;
8061 	}
8062 
8063 	return is_wol_supported;
8064 }
8065 
8066 /**
8067  * ixgbe_probe - Device Initialization Routine
8068  * @pdev: PCI device information struct
8069  * @ent: entry in ixgbe_pci_tbl
8070  *
8071  * Returns 0 on success, negative on failure
8072  *
8073  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
8074  * The OS initialization, configuring of the adapter private structure,
8075  * and a hardware reset occur.
8076  **/
8077 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
8078 {
8079 	struct net_device *netdev;
8080 	struct ixgbe_adapter *adapter = NULL;
8081 	struct ixgbe_hw *hw;
8082 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
8083 	static int cards_found;
8084 	int i, err, pci_using_dac, expected_gts;
8085 	unsigned int indices = MAX_TX_QUEUES;
8086 	u8 part_str[IXGBE_PBANUM_LENGTH];
8087 #ifdef IXGBE_FCOE
8088 	u16 device_caps;
8089 #endif
8090 	u32 eec;
8091 
8092 	/* Catch broken hardware that put the wrong VF device ID in
8093 	 * the PCIe SR-IOV capability.
8094 	 */
8095 	if (pdev->is_virtfn) {
8096 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
8097 		     pci_name(pdev), pdev->vendor, pdev->device);
8098 		return -EINVAL;
8099 	}
8100 
8101 	err = pci_enable_device_mem(pdev);
8102 	if (err)
8103 		return err;
8104 
8105 	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
8106 		pci_using_dac = 1;
8107 	} else {
8108 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
8109 		if (err) {
8110 			dev_err(&pdev->dev,
8111 				"No usable DMA configuration, aborting\n");
8112 			goto err_dma;
8113 		}
8114 		pci_using_dac = 0;
8115 	}
8116 
8117 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
8118 					   IORESOURCE_MEM), ixgbe_driver_name);
8119 	if (err) {
8120 		dev_err(&pdev->dev,
8121 			"pci_request_selected_regions failed 0x%x\n", err);
8122 		goto err_pci_reg;
8123 	}
8124 
8125 	pci_enable_pcie_error_reporting(pdev);
8126 
8127 	pci_set_master(pdev);
8128 	pci_save_state(pdev);
8129 
8130 	if (ii->mac == ixgbe_mac_82598EB) {
8131 #ifdef CONFIG_IXGBE_DCB
8132 		/* 8 TC w/ 4 queues per TC */
8133 		indices = 4 * MAX_TRAFFIC_CLASS;
8134 #else
8135 		indices = IXGBE_MAX_RSS_INDICES;
8136 #endif
8137 	}
8138 
8139 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
8140 	if (!netdev) {
8141 		err = -ENOMEM;
8142 		goto err_alloc_etherdev;
8143 	}
8144 
8145 	SET_NETDEV_DEV(netdev, &pdev->dev);
8146 
8147 	adapter = netdev_priv(netdev);
8148 	pci_set_drvdata(pdev, adapter);
8149 
8150 	adapter->netdev = netdev;
8151 	adapter->pdev = pdev;
8152 	hw = &adapter->hw;
8153 	hw->back = adapter;
8154 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
8155 
8156 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
8157 			      pci_resource_len(pdev, 0));
8158 	adapter->io_addr = hw->hw_addr;
8159 	if (!hw->hw_addr) {
8160 		err = -EIO;
8161 		goto err_ioremap;
8162 	}
8163 
8164 	netdev->netdev_ops = &ixgbe_netdev_ops;
8165 	ixgbe_set_ethtool_ops(netdev);
8166 	netdev->watchdog_timeo = 5 * HZ;
8167 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
8168 
8169 	adapter->bd_number = cards_found;
8170 
8171 	/* Setup hw api */
8172 	memcpy(&hw->mac.ops, ii->mac_ops, sizeof(hw->mac.ops));
8173 	hw->mac.type  = ii->mac;
8174 
8175 	/* EEPROM */
8176 	memcpy(&hw->eeprom.ops, ii->eeprom_ops, sizeof(hw->eeprom.ops));
8177 	eec = IXGBE_READ_REG(hw, IXGBE_EEC);
8178 	if (ixgbe_removed(hw->hw_addr)) {
8179 		err = -EIO;
8180 		goto err_ioremap;
8181 	}
8182 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
8183 	if (!(eec & (1 << 8)))
8184 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
8185 
8186 	/* PHY */
8187 	memcpy(&hw->phy.ops, ii->phy_ops, sizeof(hw->phy.ops));
8188 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
8189 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
8190 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
8191 	hw->phy.mdio.mmds = 0;
8192 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
8193 	hw->phy.mdio.dev = netdev;
8194 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
8195 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
8196 
8197 	ii->get_invariants(hw);
8198 
8199 	/* setup the private structure */
8200 	err = ixgbe_sw_init(adapter);
8201 	if (err)
8202 		goto err_sw_init;
8203 
8204 	/* Make it possible the adapter to be woken up via WOL */
8205 	switch (adapter->hw.mac.type) {
8206 	case ixgbe_mac_82599EB:
8207 	case ixgbe_mac_X540:
8208 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
8209 		break;
8210 	default:
8211 		break;
8212 	}
8213 
8214 	/*
8215 	 * If there is a fan on this device and it has failed log the
8216 	 * failure.
8217 	 */
8218 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
8219 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
8220 		if (esdp & IXGBE_ESDP_SDP1)
8221 			e_crit(probe, "Fan has stopped, replace the adapter\n");
8222 	}
8223 
8224 	if (allow_unsupported_sfp)
8225 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
8226 
8227 	/* reset_hw fills in the perm_addr as well */
8228 	hw->phy.reset_if_overtemp = true;
8229 	err = hw->mac.ops.reset_hw(hw);
8230 	hw->phy.reset_if_overtemp = false;
8231 	if (err == IXGBE_ERR_SFP_NOT_PRESENT &&
8232 	    hw->mac.type == ixgbe_mac_82598EB) {
8233 		err = 0;
8234 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
8235 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
8236 		e_dev_err("Reload the driver after installing a supported module.\n");
8237 		goto err_sw_init;
8238 	} else if (err) {
8239 		e_dev_err("HW Init failed: %d\n", err);
8240 		goto err_sw_init;
8241 	}
8242 
8243 #ifdef CONFIG_PCI_IOV
8244 	/* SR-IOV not supported on the 82598 */
8245 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
8246 		goto skip_sriov;
8247 	/* Mailbox */
8248 	ixgbe_init_mbx_params_pf(hw);
8249 	memcpy(&hw->mbx.ops, ii->mbx_ops, sizeof(hw->mbx.ops));
8250 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
8251 	ixgbe_enable_sriov(adapter);
8252 skip_sriov:
8253 
8254 #endif
8255 	netdev->features = NETIF_F_SG |
8256 			   NETIF_F_IP_CSUM |
8257 			   NETIF_F_IPV6_CSUM |
8258 			   NETIF_F_HW_VLAN_CTAG_TX |
8259 			   NETIF_F_HW_VLAN_CTAG_RX |
8260 			   NETIF_F_HW_VLAN_CTAG_FILTER |
8261 			   NETIF_F_TSO |
8262 			   NETIF_F_TSO6 |
8263 			   NETIF_F_RXHASH |
8264 			   NETIF_F_RXCSUM;
8265 
8266 	netdev->hw_features = netdev->features | NETIF_F_HW_L2FW_DOFFLOAD;
8267 
8268 	switch (adapter->hw.mac.type) {
8269 	case ixgbe_mac_82599EB:
8270 	case ixgbe_mac_X540:
8271 		netdev->features |= NETIF_F_SCTP_CSUM;
8272 		netdev->hw_features |= NETIF_F_SCTP_CSUM |
8273 				       NETIF_F_NTUPLE;
8274 		break;
8275 	default:
8276 		break;
8277 	}
8278 
8279 	netdev->hw_features |= NETIF_F_RXALL;
8280 
8281 	netdev->vlan_features |= NETIF_F_TSO;
8282 	netdev->vlan_features |= NETIF_F_TSO6;
8283 	netdev->vlan_features |= NETIF_F_IP_CSUM;
8284 	netdev->vlan_features |= NETIF_F_IPV6_CSUM;
8285 	netdev->vlan_features |= NETIF_F_SG;
8286 
8287 	netdev->priv_flags |= IFF_UNICAST_FLT;
8288 	netdev->priv_flags |= IFF_SUPP_NOFCS;
8289 
8290 #ifdef CONFIG_IXGBE_DCB
8291 	netdev->dcbnl_ops = &dcbnl_ops;
8292 #endif
8293 
8294 #ifdef IXGBE_FCOE
8295 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
8296 		unsigned int fcoe_l;
8297 
8298 		if (hw->mac.ops.get_device_caps) {
8299 			hw->mac.ops.get_device_caps(hw, &device_caps);
8300 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
8301 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
8302 		}
8303 
8304 
8305 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
8306 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
8307 
8308 		netdev->features |= NETIF_F_FSO |
8309 				    NETIF_F_FCOE_CRC;
8310 
8311 		netdev->vlan_features |= NETIF_F_FSO |
8312 					 NETIF_F_FCOE_CRC |
8313 					 NETIF_F_FCOE_MTU;
8314 	}
8315 #endif /* IXGBE_FCOE */
8316 	if (pci_using_dac) {
8317 		netdev->features |= NETIF_F_HIGHDMA;
8318 		netdev->vlan_features |= NETIF_F_HIGHDMA;
8319 	}
8320 
8321 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
8322 		netdev->hw_features |= NETIF_F_LRO;
8323 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
8324 		netdev->features |= NETIF_F_LRO;
8325 
8326 	/* make sure the EEPROM is good */
8327 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
8328 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
8329 		err = -EIO;
8330 		goto err_sw_init;
8331 	}
8332 
8333 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
8334 
8335 	if (!is_valid_ether_addr(netdev->dev_addr)) {
8336 		e_dev_err("invalid MAC address\n");
8337 		err = -EIO;
8338 		goto err_sw_init;
8339 	}
8340 
8341 	ixgbe_mac_set_default_filter(adapter, hw->mac.perm_addr);
8342 
8343 	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
8344 		    (unsigned long) adapter);
8345 
8346 	if (ixgbe_removed(hw->hw_addr)) {
8347 		err = -EIO;
8348 		goto err_sw_init;
8349 	}
8350 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
8351 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
8352 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
8353 
8354 	err = ixgbe_init_interrupt_scheme(adapter);
8355 	if (err)
8356 		goto err_sw_init;
8357 
8358 	/* WOL not supported for all devices */
8359 	adapter->wol = 0;
8360 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
8361 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
8362 						pdev->subsystem_device);
8363 	if (hw->wol_enabled)
8364 		adapter->wol = IXGBE_WUFC_MAG;
8365 
8366 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
8367 
8368 	/* save off EEPROM version number */
8369 	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
8370 	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
8371 
8372 	/* pick up the PCI bus settings for reporting later */
8373 	hw->mac.ops.get_bus_info(hw);
8374 	if (ixgbe_pcie_from_parent(hw))
8375 		ixgbe_get_parent_bus_info(adapter);
8376 
8377 	/* calculate the expected PCIe bandwidth required for optimal
8378 	 * performance. Note that some older parts will never have enough
8379 	 * bandwidth due to being older generation PCIe parts. We clamp these
8380 	 * parts to ensure no warning is displayed if it can't be fixed.
8381 	 */
8382 	switch (hw->mac.type) {
8383 	case ixgbe_mac_82598EB:
8384 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
8385 		break;
8386 	default:
8387 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
8388 		break;
8389 	}
8390 
8391 	/* don't check link if we failed to enumerate functions */
8392 	if (expected_gts > 0)
8393 		ixgbe_check_minimum_link(adapter, expected_gts);
8394 
8395 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
8396 	if (err)
8397 		strlcpy(part_str, "Unknown", sizeof(part_str));
8398 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
8399 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
8400 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
8401 			   part_str);
8402 	else
8403 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
8404 			   hw->mac.type, hw->phy.type, part_str);
8405 
8406 	e_dev_info("%pM\n", netdev->dev_addr);
8407 
8408 	/* reset the hardware with the new settings */
8409 	err = hw->mac.ops.start_hw(hw);
8410 	if (err == IXGBE_ERR_EEPROM_VERSION) {
8411 		/* We are running on a pre-production device, log a warning */
8412 		e_dev_warn("This device is a pre-production adapter/LOM. "
8413 			   "Please be aware there may be issues associated "
8414 			   "with your hardware.  If you are experiencing "
8415 			   "problems please contact your Intel or hardware "
8416 			   "representative who provided you with this "
8417 			   "hardware.\n");
8418 	}
8419 	strcpy(netdev->name, "eth%d");
8420 	err = register_netdev(netdev);
8421 	if (err)
8422 		goto err_register;
8423 
8424 	/* power down the optics for 82599 SFP+ fiber */
8425 	if (hw->mac.ops.disable_tx_laser)
8426 		hw->mac.ops.disable_tx_laser(hw);
8427 
8428 	/* carrier off reporting is important to ethtool even BEFORE open */
8429 	netif_carrier_off(netdev);
8430 
8431 #ifdef CONFIG_IXGBE_DCA
8432 	if (dca_add_requester(&pdev->dev) == 0) {
8433 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
8434 		ixgbe_setup_dca(adapter);
8435 	}
8436 #endif
8437 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
8438 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
8439 		for (i = 0; i < adapter->num_vfs; i++)
8440 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
8441 	}
8442 
8443 	/* firmware requires driver version to be 0xFFFFFFFF
8444 	 * since os does not support feature
8445 	 */
8446 	if (hw->mac.ops.set_fw_drv_ver)
8447 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF,
8448 					   0xFF);
8449 
8450 	/* add san mac addr to netdev */
8451 	ixgbe_add_sanmac_netdev(netdev);
8452 
8453 	e_dev_info("%s\n", ixgbe_default_device_descr);
8454 	cards_found++;
8455 
8456 #ifdef CONFIG_IXGBE_HWMON
8457 	if (ixgbe_sysfs_init(adapter))
8458 		e_err(probe, "failed to allocate sysfs resources\n");
8459 #endif /* CONFIG_IXGBE_HWMON */
8460 
8461 	ixgbe_dbg_adapter_init(adapter);
8462 
8463 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
8464 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
8465 		hw->mac.ops.setup_link(hw,
8466 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
8467 			true);
8468 
8469 	return 0;
8470 
8471 err_register:
8472 	ixgbe_release_hw_control(adapter);
8473 	ixgbe_clear_interrupt_scheme(adapter);
8474 err_sw_init:
8475 	ixgbe_disable_sriov(adapter);
8476 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
8477 	iounmap(adapter->io_addr);
8478 	kfree(adapter->mac_table);
8479 err_ioremap:
8480 	free_netdev(netdev);
8481 err_alloc_etherdev:
8482 	pci_release_selected_regions(pdev,
8483 				     pci_select_bars(pdev, IORESOURCE_MEM));
8484 err_pci_reg:
8485 err_dma:
8486 	if (!adapter || !test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
8487 		pci_disable_device(pdev);
8488 	return err;
8489 }
8490 
8491 /**
8492  * ixgbe_remove - Device Removal Routine
8493  * @pdev: PCI device information struct
8494  *
8495  * ixgbe_remove is called by the PCI subsystem to alert the driver
8496  * that it should release a PCI device.  The could be caused by a
8497  * Hot-Plug event, or because the driver is going to be removed from
8498  * memory.
8499  **/
8500 static void ixgbe_remove(struct pci_dev *pdev)
8501 {
8502 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
8503 	struct net_device *netdev = adapter->netdev;
8504 
8505 	ixgbe_dbg_adapter_exit(adapter);
8506 
8507 	set_bit(__IXGBE_REMOVING, &adapter->state);
8508 	cancel_work_sync(&adapter->service_task);
8509 
8510 
8511 #ifdef CONFIG_IXGBE_DCA
8512 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
8513 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
8514 		dca_remove_requester(&pdev->dev);
8515 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
8516 	}
8517 
8518 #endif
8519 #ifdef CONFIG_IXGBE_HWMON
8520 	ixgbe_sysfs_exit(adapter);
8521 #endif /* CONFIG_IXGBE_HWMON */
8522 
8523 	/* remove the added san mac */
8524 	ixgbe_del_sanmac_netdev(netdev);
8525 
8526 	if (netdev->reg_state == NETREG_REGISTERED)
8527 		unregister_netdev(netdev);
8528 
8529 #ifdef CONFIG_PCI_IOV
8530 	/*
8531 	 * Only disable SR-IOV on unload if the user specified the now
8532 	 * deprecated max_vfs module parameter.
8533 	 */
8534 	if (max_vfs)
8535 		ixgbe_disable_sriov(adapter);
8536 #endif
8537 	ixgbe_clear_interrupt_scheme(adapter);
8538 
8539 	ixgbe_release_hw_control(adapter);
8540 
8541 #ifdef CONFIG_DCB
8542 	kfree(adapter->ixgbe_ieee_pfc);
8543 	kfree(adapter->ixgbe_ieee_ets);
8544 
8545 #endif
8546 	iounmap(adapter->io_addr);
8547 	pci_release_selected_regions(pdev, pci_select_bars(pdev,
8548 				     IORESOURCE_MEM));
8549 
8550 	e_dev_info("complete\n");
8551 
8552 	kfree(adapter->mac_table);
8553 	free_netdev(netdev);
8554 
8555 	pci_disable_pcie_error_reporting(pdev);
8556 
8557 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
8558 		pci_disable_device(pdev);
8559 }
8560 
8561 /**
8562  * ixgbe_io_error_detected - called when PCI error is detected
8563  * @pdev: Pointer to PCI device
8564  * @state: The current pci connection state
8565  *
8566  * This function is called after a PCI bus error affecting
8567  * this device has been detected.
8568  */
8569 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
8570 						pci_channel_state_t state)
8571 {
8572 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
8573 	struct net_device *netdev = adapter->netdev;
8574 
8575 #ifdef CONFIG_PCI_IOV
8576 	struct ixgbe_hw *hw = &adapter->hw;
8577 	struct pci_dev *bdev, *vfdev;
8578 	u32 dw0, dw1, dw2, dw3;
8579 	int vf, pos;
8580 	u16 req_id, pf_func;
8581 
8582 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
8583 	    adapter->num_vfs == 0)
8584 		goto skip_bad_vf_detection;
8585 
8586 	bdev = pdev->bus->self;
8587 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
8588 		bdev = bdev->bus->self;
8589 
8590 	if (!bdev)
8591 		goto skip_bad_vf_detection;
8592 
8593 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
8594 	if (!pos)
8595 		goto skip_bad_vf_detection;
8596 
8597 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
8598 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
8599 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
8600 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
8601 	if (ixgbe_removed(hw->hw_addr))
8602 		goto skip_bad_vf_detection;
8603 
8604 	req_id = dw1 >> 16;
8605 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
8606 	if (!(req_id & 0x0080))
8607 		goto skip_bad_vf_detection;
8608 
8609 	pf_func = req_id & 0x01;
8610 	if ((pf_func & 1) == (pdev->devfn & 1)) {
8611 		unsigned int device_id;
8612 
8613 		vf = (req_id & 0x7F) >> 1;
8614 		e_dev_err("VF %d has caused a PCIe error\n", vf);
8615 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
8616 				"%8.8x\tdw3: %8.8x\n",
8617 		dw0, dw1, dw2, dw3);
8618 		switch (adapter->hw.mac.type) {
8619 		case ixgbe_mac_82599EB:
8620 			device_id = IXGBE_82599_VF_DEVICE_ID;
8621 			break;
8622 		case ixgbe_mac_X540:
8623 			device_id = IXGBE_X540_VF_DEVICE_ID;
8624 			break;
8625 		default:
8626 			device_id = 0;
8627 			break;
8628 		}
8629 
8630 		/* Find the pci device of the offending VF */
8631 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
8632 		while (vfdev) {
8633 			if (vfdev->devfn == (req_id & 0xFF))
8634 				break;
8635 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
8636 					       device_id, vfdev);
8637 		}
8638 		/*
8639 		 * There's a slim chance the VF could have been hot plugged,
8640 		 * so if it is no longer present we don't need to issue the
8641 		 * VFLR.  Just clean up the AER in that case.
8642 		 */
8643 		if (vfdev) {
8644 			e_dev_err("Issuing VFLR to VF %d\n", vf);
8645 			pci_write_config_dword(vfdev, 0xA8, 0x00008000);
8646 			/* Free device reference count */
8647 			pci_dev_put(vfdev);
8648 		}
8649 
8650 		pci_cleanup_aer_uncorrect_error_status(pdev);
8651 	}
8652 
8653 	/*
8654 	 * Even though the error may have occurred on the other port
8655 	 * we still need to increment the vf error reference count for
8656 	 * both ports because the I/O resume function will be called
8657 	 * for both of them.
8658 	 */
8659 	adapter->vferr_refcount++;
8660 
8661 	return PCI_ERS_RESULT_RECOVERED;
8662 
8663 skip_bad_vf_detection:
8664 #endif /* CONFIG_PCI_IOV */
8665 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
8666 		return PCI_ERS_RESULT_DISCONNECT;
8667 
8668 	rtnl_lock();
8669 	netif_device_detach(netdev);
8670 
8671 	if (state == pci_channel_io_perm_failure) {
8672 		rtnl_unlock();
8673 		return PCI_ERS_RESULT_DISCONNECT;
8674 	}
8675 
8676 	if (netif_running(netdev))
8677 		ixgbe_down(adapter);
8678 
8679 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
8680 		pci_disable_device(pdev);
8681 	rtnl_unlock();
8682 
8683 	/* Request a slot reset. */
8684 	return PCI_ERS_RESULT_NEED_RESET;
8685 }
8686 
8687 /**
8688  * ixgbe_io_slot_reset - called after the pci bus has been reset.
8689  * @pdev: Pointer to PCI device
8690  *
8691  * Restart the card from scratch, as if from a cold-boot.
8692  */
8693 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
8694 {
8695 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
8696 	pci_ers_result_t result;
8697 	int err;
8698 
8699 	if (pci_enable_device_mem(pdev)) {
8700 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
8701 		result = PCI_ERS_RESULT_DISCONNECT;
8702 	} else {
8703 		smp_mb__before_atomic();
8704 		clear_bit(__IXGBE_DISABLED, &adapter->state);
8705 		adapter->hw.hw_addr = adapter->io_addr;
8706 		pci_set_master(pdev);
8707 		pci_restore_state(pdev);
8708 		pci_save_state(pdev);
8709 
8710 		pci_wake_from_d3(pdev, false);
8711 
8712 		ixgbe_reset(adapter);
8713 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
8714 		result = PCI_ERS_RESULT_RECOVERED;
8715 	}
8716 
8717 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
8718 	if (err) {
8719 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
8720 			  "failed 0x%0x\n", err);
8721 		/* non-fatal, continue */
8722 	}
8723 
8724 	return result;
8725 }
8726 
8727 /**
8728  * ixgbe_io_resume - called when traffic can start flowing again.
8729  * @pdev: Pointer to PCI device
8730  *
8731  * This callback is called when the error recovery driver tells us that
8732  * its OK to resume normal operation.
8733  */
8734 static void ixgbe_io_resume(struct pci_dev *pdev)
8735 {
8736 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
8737 	struct net_device *netdev = adapter->netdev;
8738 
8739 #ifdef CONFIG_PCI_IOV
8740 	if (adapter->vferr_refcount) {
8741 		e_info(drv, "Resuming after VF err\n");
8742 		adapter->vferr_refcount--;
8743 		return;
8744 	}
8745 
8746 #endif
8747 	if (netif_running(netdev))
8748 		ixgbe_up(adapter);
8749 
8750 	netif_device_attach(netdev);
8751 }
8752 
8753 static const struct pci_error_handlers ixgbe_err_handler = {
8754 	.error_detected = ixgbe_io_error_detected,
8755 	.slot_reset = ixgbe_io_slot_reset,
8756 	.resume = ixgbe_io_resume,
8757 };
8758 
8759 static struct pci_driver ixgbe_driver = {
8760 	.name     = ixgbe_driver_name,
8761 	.id_table = ixgbe_pci_tbl,
8762 	.probe    = ixgbe_probe,
8763 	.remove   = ixgbe_remove,
8764 #ifdef CONFIG_PM
8765 	.suspend  = ixgbe_suspend,
8766 	.resume   = ixgbe_resume,
8767 #endif
8768 	.shutdown = ixgbe_shutdown,
8769 	.sriov_configure = ixgbe_pci_sriov_configure,
8770 	.err_handler = &ixgbe_err_handler
8771 };
8772 
8773 /**
8774  * ixgbe_init_module - Driver Registration Routine
8775  *
8776  * ixgbe_init_module is the first routine called when the driver is
8777  * loaded. All it does is register with the PCI subsystem.
8778  **/
8779 static int __init ixgbe_init_module(void)
8780 {
8781 	int ret;
8782 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
8783 	pr_info("%s\n", ixgbe_copyright);
8784 
8785 	ixgbe_dbg_init();
8786 
8787 	ret = pci_register_driver(&ixgbe_driver);
8788 	if (ret) {
8789 		ixgbe_dbg_exit();
8790 		return ret;
8791 	}
8792 
8793 #ifdef CONFIG_IXGBE_DCA
8794 	dca_register_notify(&dca_notifier);
8795 #endif
8796 
8797 	return 0;
8798 }
8799 
8800 module_init(ixgbe_init_module);
8801 
8802 /**
8803  * ixgbe_exit_module - Driver Exit Cleanup Routine
8804  *
8805  * ixgbe_exit_module is called just before the driver is removed
8806  * from memory.
8807  **/
8808 static void __exit ixgbe_exit_module(void)
8809 {
8810 #ifdef CONFIG_IXGBE_DCA
8811 	dca_unregister_notify(&dca_notifier);
8812 #endif
8813 	pci_unregister_driver(&ixgbe_driver);
8814 
8815 	ixgbe_dbg_exit();
8816 
8817 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
8818 }
8819 
8820 #ifdef CONFIG_IXGBE_DCA
8821 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
8822 			    void *p)
8823 {
8824 	int ret_val;
8825 
8826 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
8827 					 __ixgbe_notify_dca);
8828 
8829 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
8830 }
8831 
8832 #endif /* CONFIG_IXGBE_DCA */
8833 
8834 module_exit(ixgbe_exit_module);
8835 
8836 /* ixgbe_main.c */
8837