xref: /linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision 4949009eb8d40a441dcddcd96e101e77d31cf1b2)
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/etherdevice.h>
46 #include <linux/ethtool.h>
47 #include <linux/if.h>
48 #include <linux/if_vlan.h>
49 #include <linux/if_macvlan.h>
50 #include <linux/if_bridge.h>
51 #include <linux/prefetch.h>
52 #include <scsi/fc/fc_fcoe.h>
53 
54 #ifdef CONFIG_OF
55 #include <linux/of_net.h>
56 #endif
57 
58 #ifdef CONFIG_SPARC
59 #include <asm/idprom.h>
60 #include <asm/prom.h>
61 #endif
62 
63 #include "ixgbe.h"
64 #include "ixgbe_common.h"
65 #include "ixgbe_dcb_82599.h"
66 #include "ixgbe_sriov.h"
67 
68 char ixgbe_driver_name[] = "ixgbe";
69 static const char ixgbe_driver_string[] =
70 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
71 #ifdef IXGBE_FCOE
72 char ixgbe_default_device_descr[] =
73 			      "Intel(R) 10 Gigabit Network Connection";
74 #else
75 static char ixgbe_default_device_descr[] =
76 			      "Intel(R) 10 Gigabit Network Connection";
77 #endif
78 #define DRV_VERSION "4.0.1-k"
79 const char ixgbe_driver_version[] = DRV_VERSION;
80 static const char ixgbe_copyright[] =
81 				"Copyright (c) 1999-2014 Intel Corporation.";
82 
83 static const struct ixgbe_info *ixgbe_info_tbl[] = {
84 	[board_82598]		= &ixgbe_82598_info,
85 	[board_82599]		= &ixgbe_82599_info,
86 	[board_X540]		= &ixgbe_X540_info,
87 	[board_X550]		= &ixgbe_X550_info,
88 	[board_X550EM_x]	= &ixgbe_X550EM_x_info,
89 };
90 
91 /* ixgbe_pci_tbl - PCI Device ID Table
92  *
93  * Wildcard entries (PCI_ANY_ID) should come last
94  * Last entry must be all 0s
95  *
96  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
97  *   Class, Class Mask, private data (not used) }
98  */
99 static const struct pci_device_id ixgbe_pci_tbl[] = {
100 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
101 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
102 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
103 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
104 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
105 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
115 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
116 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
117 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
118 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
119 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
120 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
121 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
122 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
123 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
124 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
125 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
126 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
127 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
128 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
129 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
130 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
131 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
132 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
133 	/* required last entry */
134 	{0, }
135 };
136 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
137 
138 #ifdef CONFIG_IXGBE_DCA
139 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
140 			    void *p);
141 static struct notifier_block dca_notifier = {
142 	.notifier_call = ixgbe_notify_dca,
143 	.next          = NULL,
144 	.priority      = 0
145 };
146 #endif
147 
148 #ifdef CONFIG_PCI_IOV
149 static unsigned int max_vfs;
150 module_param(max_vfs, uint, 0);
151 MODULE_PARM_DESC(max_vfs,
152 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
153 #endif /* CONFIG_PCI_IOV */
154 
155 static unsigned int allow_unsupported_sfp;
156 module_param(allow_unsupported_sfp, uint, 0);
157 MODULE_PARM_DESC(allow_unsupported_sfp,
158 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
159 
160 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
161 static int debug = -1;
162 module_param(debug, int, 0);
163 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
164 
165 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
166 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
167 MODULE_LICENSE("GPL");
168 MODULE_VERSION(DRV_VERSION);
169 
170 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
171 
172 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
173 					  u32 reg, u16 *value)
174 {
175 	struct pci_dev *parent_dev;
176 	struct pci_bus *parent_bus;
177 
178 	parent_bus = adapter->pdev->bus->parent;
179 	if (!parent_bus)
180 		return -1;
181 
182 	parent_dev = parent_bus->self;
183 	if (!parent_dev)
184 		return -1;
185 
186 	if (!pci_is_pcie(parent_dev))
187 		return -1;
188 
189 	pcie_capability_read_word(parent_dev, reg, value);
190 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
191 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
192 		return -1;
193 	return 0;
194 }
195 
196 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
197 {
198 	struct ixgbe_hw *hw = &adapter->hw;
199 	u16 link_status = 0;
200 	int err;
201 
202 	hw->bus.type = ixgbe_bus_type_pci_express;
203 
204 	/* Get the negotiated link width and speed from PCI config space of the
205 	 * parent, as this device is behind a switch
206 	 */
207 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
208 
209 	/* assume caller will handle error case */
210 	if (err)
211 		return err;
212 
213 	hw->bus.width = ixgbe_convert_bus_width(link_status);
214 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
215 
216 	return 0;
217 }
218 
219 /**
220  * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
221  * @hw: hw specific details
222  *
223  * This function is used by probe to determine whether a device's PCI-Express
224  * bandwidth details should be gathered from the parent bus instead of from the
225  * device. Used to ensure that various locations all have the correct device ID
226  * checks.
227  */
228 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
229 {
230 	switch (hw->device_id) {
231 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
232 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
233 		return true;
234 	default:
235 		return false;
236 	}
237 }
238 
239 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
240 				     int expected_gts)
241 {
242 	int max_gts = 0;
243 	enum pci_bus_speed speed = PCI_SPEED_UNKNOWN;
244 	enum pcie_link_width width = PCIE_LNK_WIDTH_UNKNOWN;
245 	struct pci_dev *pdev;
246 
247 	/* determine whether to use the the parent device
248 	 */
249 	if (ixgbe_pcie_from_parent(&adapter->hw))
250 		pdev = adapter->pdev->bus->parent->self;
251 	else
252 		pdev = adapter->pdev;
253 
254 	if (pcie_get_minimum_link(pdev, &speed, &width) ||
255 	    speed == PCI_SPEED_UNKNOWN || width == PCIE_LNK_WIDTH_UNKNOWN) {
256 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
257 		return;
258 	}
259 
260 	switch (speed) {
261 	case PCIE_SPEED_2_5GT:
262 		/* 8b/10b encoding reduces max throughput by 20% */
263 		max_gts = 2 * width;
264 		break;
265 	case PCIE_SPEED_5_0GT:
266 		/* 8b/10b encoding reduces max throughput by 20% */
267 		max_gts = 4 * width;
268 		break;
269 	case PCIE_SPEED_8_0GT:
270 		/* 128b/130b encoding reduces throughput by less than 2% */
271 		max_gts = 8 * width;
272 		break;
273 	default:
274 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
275 		return;
276 	}
277 
278 	e_dev_info("PCI Express bandwidth of %dGT/s available\n",
279 		   max_gts);
280 	e_dev_info("(Speed:%s, Width: x%d, Encoding Loss:%s)\n",
281 		   (speed == PCIE_SPEED_8_0GT ? "8.0GT/s" :
282 		    speed == PCIE_SPEED_5_0GT ? "5.0GT/s" :
283 		    speed == PCIE_SPEED_2_5GT ? "2.5GT/s" :
284 		    "Unknown"),
285 		   width,
286 		   (speed == PCIE_SPEED_2_5GT ? "20%" :
287 		    speed == PCIE_SPEED_5_0GT ? "20%" :
288 		    speed == PCIE_SPEED_8_0GT ? "<2%" :
289 		    "Unknown"));
290 
291 	if (max_gts < expected_gts) {
292 		e_dev_warn("This is not sufficient for optimal performance of this card.\n");
293 		e_dev_warn("For optimal performance, at least %dGT/s of bandwidth is required.\n",
294 			expected_gts);
295 		e_dev_warn("A slot with more lanes and/or higher speed is suggested.\n");
296 	}
297 }
298 
299 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
300 {
301 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
302 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
303 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
304 		schedule_work(&adapter->service_task);
305 }
306 
307 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
308 {
309 	struct ixgbe_adapter *adapter = hw->back;
310 
311 	if (!hw->hw_addr)
312 		return;
313 	hw->hw_addr = NULL;
314 	e_dev_err("Adapter removed\n");
315 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
316 		ixgbe_service_event_schedule(adapter);
317 }
318 
319 static void ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
320 {
321 	u32 value;
322 
323 	/* The following check not only optimizes a bit by not
324 	 * performing a read on the status register when the
325 	 * register just read was a status register read that
326 	 * returned IXGBE_FAILED_READ_REG. It also blocks any
327 	 * potential recursion.
328 	 */
329 	if (reg == IXGBE_STATUS) {
330 		ixgbe_remove_adapter(hw);
331 		return;
332 	}
333 	value = ixgbe_read_reg(hw, IXGBE_STATUS);
334 	if (value == IXGBE_FAILED_READ_REG)
335 		ixgbe_remove_adapter(hw);
336 }
337 
338 /**
339  * ixgbe_read_reg - Read from device register
340  * @hw: hw specific details
341  * @reg: offset of register to read
342  *
343  * Returns : value read or IXGBE_FAILED_READ_REG if removed
344  *
345  * This function is used to read device registers. It checks for device
346  * removal by confirming any read that returns all ones by checking the
347  * status register value for all ones. This function avoids reading from
348  * the hardware if a removal was previously detected in which case it
349  * returns IXGBE_FAILED_READ_REG (all ones).
350  */
351 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
352 {
353 	u8 __iomem *reg_addr = ACCESS_ONCE(hw->hw_addr);
354 	u32 value;
355 
356 	if (ixgbe_removed(reg_addr))
357 		return IXGBE_FAILED_READ_REG;
358 	value = readl(reg_addr + reg);
359 	if (unlikely(value == IXGBE_FAILED_READ_REG))
360 		ixgbe_check_remove(hw, reg);
361 	return value;
362 }
363 
364 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
365 {
366 	u16 value;
367 
368 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
369 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
370 		ixgbe_remove_adapter(hw);
371 		return true;
372 	}
373 	return false;
374 }
375 
376 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
377 {
378 	struct ixgbe_adapter *adapter = hw->back;
379 	u16 value;
380 
381 	if (ixgbe_removed(hw->hw_addr))
382 		return IXGBE_FAILED_READ_CFG_WORD;
383 	pci_read_config_word(adapter->pdev, reg, &value);
384 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
385 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
386 		return IXGBE_FAILED_READ_CFG_WORD;
387 	return value;
388 }
389 
390 #ifdef CONFIG_PCI_IOV
391 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
392 {
393 	struct ixgbe_adapter *adapter = hw->back;
394 	u32 value;
395 
396 	if (ixgbe_removed(hw->hw_addr))
397 		return IXGBE_FAILED_READ_CFG_DWORD;
398 	pci_read_config_dword(adapter->pdev, reg, &value);
399 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
400 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
401 		return IXGBE_FAILED_READ_CFG_DWORD;
402 	return value;
403 }
404 #endif /* CONFIG_PCI_IOV */
405 
406 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
407 {
408 	struct ixgbe_adapter *adapter = hw->back;
409 
410 	if (ixgbe_removed(hw->hw_addr))
411 		return;
412 	pci_write_config_word(adapter->pdev, reg, value);
413 }
414 
415 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
416 {
417 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
418 
419 	/* flush memory to make sure state is correct before next watchdog */
420 	smp_mb__before_atomic();
421 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
422 }
423 
424 struct ixgbe_reg_info {
425 	u32 ofs;
426 	char *name;
427 };
428 
429 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
430 
431 	/* General Registers */
432 	{IXGBE_CTRL, "CTRL"},
433 	{IXGBE_STATUS, "STATUS"},
434 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
435 
436 	/* Interrupt Registers */
437 	{IXGBE_EICR, "EICR"},
438 
439 	/* RX Registers */
440 	{IXGBE_SRRCTL(0), "SRRCTL"},
441 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
442 	{IXGBE_RDLEN(0), "RDLEN"},
443 	{IXGBE_RDH(0), "RDH"},
444 	{IXGBE_RDT(0), "RDT"},
445 	{IXGBE_RXDCTL(0), "RXDCTL"},
446 	{IXGBE_RDBAL(0), "RDBAL"},
447 	{IXGBE_RDBAH(0), "RDBAH"},
448 
449 	/* TX Registers */
450 	{IXGBE_TDBAL(0), "TDBAL"},
451 	{IXGBE_TDBAH(0), "TDBAH"},
452 	{IXGBE_TDLEN(0), "TDLEN"},
453 	{IXGBE_TDH(0), "TDH"},
454 	{IXGBE_TDT(0), "TDT"},
455 	{IXGBE_TXDCTL(0), "TXDCTL"},
456 
457 	/* List Terminator */
458 	{ .name = NULL }
459 };
460 
461 
462 /*
463  * ixgbe_regdump - register printout routine
464  */
465 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
466 {
467 	int i = 0, j = 0;
468 	char rname[16];
469 	u32 regs[64];
470 
471 	switch (reginfo->ofs) {
472 	case IXGBE_SRRCTL(0):
473 		for (i = 0; i < 64; i++)
474 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
475 		break;
476 	case IXGBE_DCA_RXCTRL(0):
477 		for (i = 0; i < 64; i++)
478 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
479 		break;
480 	case IXGBE_RDLEN(0):
481 		for (i = 0; i < 64; i++)
482 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
483 		break;
484 	case IXGBE_RDH(0):
485 		for (i = 0; i < 64; i++)
486 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
487 		break;
488 	case IXGBE_RDT(0):
489 		for (i = 0; i < 64; i++)
490 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
491 		break;
492 	case IXGBE_RXDCTL(0):
493 		for (i = 0; i < 64; i++)
494 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
495 		break;
496 	case IXGBE_RDBAL(0):
497 		for (i = 0; i < 64; i++)
498 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
499 		break;
500 	case IXGBE_RDBAH(0):
501 		for (i = 0; i < 64; i++)
502 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
503 		break;
504 	case IXGBE_TDBAL(0):
505 		for (i = 0; i < 64; i++)
506 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
507 		break;
508 	case IXGBE_TDBAH(0):
509 		for (i = 0; i < 64; i++)
510 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
511 		break;
512 	case IXGBE_TDLEN(0):
513 		for (i = 0; i < 64; i++)
514 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
515 		break;
516 	case IXGBE_TDH(0):
517 		for (i = 0; i < 64; i++)
518 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
519 		break;
520 	case IXGBE_TDT(0):
521 		for (i = 0; i < 64; i++)
522 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
523 		break;
524 	case IXGBE_TXDCTL(0):
525 		for (i = 0; i < 64; i++)
526 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
527 		break;
528 	default:
529 		pr_info("%-15s %08x\n", reginfo->name,
530 			IXGBE_READ_REG(hw, reginfo->ofs));
531 		return;
532 	}
533 
534 	for (i = 0; i < 8; i++) {
535 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i*8, i*8+7);
536 		pr_err("%-15s", rname);
537 		for (j = 0; j < 8; j++)
538 			pr_cont(" %08x", regs[i*8+j]);
539 		pr_cont("\n");
540 	}
541 
542 }
543 
544 /*
545  * ixgbe_dump - Print registers, tx-rings and rx-rings
546  */
547 static void ixgbe_dump(struct ixgbe_adapter *adapter)
548 {
549 	struct net_device *netdev = adapter->netdev;
550 	struct ixgbe_hw *hw = &adapter->hw;
551 	struct ixgbe_reg_info *reginfo;
552 	int n = 0;
553 	struct ixgbe_ring *tx_ring;
554 	struct ixgbe_tx_buffer *tx_buffer;
555 	union ixgbe_adv_tx_desc *tx_desc;
556 	struct my_u0 { u64 a; u64 b; } *u0;
557 	struct ixgbe_ring *rx_ring;
558 	union ixgbe_adv_rx_desc *rx_desc;
559 	struct ixgbe_rx_buffer *rx_buffer_info;
560 	u32 staterr;
561 	int i = 0;
562 
563 	if (!netif_msg_hw(adapter))
564 		return;
565 
566 	/* Print netdevice Info */
567 	if (netdev) {
568 		dev_info(&adapter->pdev->dev, "Net device Info\n");
569 		pr_info("Device Name     state            "
570 			"trans_start      last_rx\n");
571 		pr_info("%-15s %016lX %016lX %016lX\n",
572 			netdev->name,
573 			netdev->state,
574 			netdev->trans_start,
575 			netdev->last_rx);
576 	}
577 
578 	/* Print Registers */
579 	dev_info(&adapter->pdev->dev, "Register Dump\n");
580 	pr_info(" Register Name   Value\n");
581 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
582 	     reginfo->name; reginfo++) {
583 		ixgbe_regdump(hw, reginfo);
584 	}
585 
586 	/* Print TX Ring Summary */
587 	if (!netdev || !netif_running(netdev))
588 		return;
589 
590 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
591 	pr_info(" %s     %s              %s        %s\n",
592 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
593 		"leng", "ntw", "timestamp");
594 	for (n = 0; n < adapter->num_tx_queues; n++) {
595 		tx_ring = adapter->tx_ring[n];
596 		tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
597 		pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
598 			   n, tx_ring->next_to_use, tx_ring->next_to_clean,
599 			   (u64)dma_unmap_addr(tx_buffer, dma),
600 			   dma_unmap_len(tx_buffer, len),
601 			   tx_buffer->next_to_watch,
602 			   (u64)tx_buffer->time_stamp);
603 	}
604 
605 	/* Print TX Rings */
606 	if (!netif_msg_tx_done(adapter))
607 		goto rx_ring_summary;
608 
609 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
610 
611 	/* Transmit Descriptor Formats
612 	 *
613 	 * 82598 Advanced Transmit Descriptor
614 	 *   +--------------------------------------------------------------+
615 	 * 0 |         Buffer Address [63:0]                                |
616 	 *   +--------------------------------------------------------------+
617 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
618 	 *   +--------------------------------------------------------------+
619 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
620 	 *
621 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
622 	 *   +--------------------------------------------------------------+
623 	 * 0 |                          RSV [63:0]                          |
624 	 *   +--------------------------------------------------------------+
625 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
626 	 *   +--------------------------------------------------------------+
627 	 *   63                       36 35   32 31                         0
628 	 *
629 	 * 82599+ Advanced Transmit Descriptor
630 	 *   +--------------------------------------------------------------+
631 	 * 0 |         Buffer Address [63:0]                                |
632 	 *   +--------------------------------------------------------------+
633 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
634 	 *   +--------------------------------------------------------------+
635 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
636 	 *
637 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
638 	 *   +--------------------------------------------------------------+
639 	 * 0 |                          RSV [63:0]                          |
640 	 *   +--------------------------------------------------------------+
641 	 * 8 |            RSV           |  STA  |           RSV             |
642 	 *   +--------------------------------------------------------------+
643 	 *   63                       36 35   32 31                         0
644 	 */
645 
646 	for (n = 0; n < adapter->num_tx_queues; n++) {
647 		tx_ring = adapter->tx_ring[n];
648 		pr_info("------------------------------------\n");
649 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
650 		pr_info("------------------------------------\n");
651 		pr_info("%s%s    %s              %s        %s          %s\n",
652 			"T [desc]     [address 63:0  ] ",
653 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
654 			"leng", "ntw", "timestamp", "bi->skb");
655 
656 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
657 			tx_desc = IXGBE_TX_DESC(tx_ring, i);
658 			tx_buffer = &tx_ring->tx_buffer_info[i];
659 			u0 = (struct my_u0 *)tx_desc;
660 			if (dma_unmap_len(tx_buffer, len) > 0) {
661 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p",
662 					i,
663 					le64_to_cpu(u0->a),
664 					le64_to_cpu(u0->b),
665 					(u64)dma_unmap_addr(tx_buffer, dma),
666 					dma_unmap_len(tx_buffer, len),
667 					tx_buffer->next_to_watch,
668 					(u64)tx_buffer->time_stamp,
669 					tx_buffer->skb);
670 				if (i == tx_ring->next_to_use &&
671 					i == tx_ring->next_to_clean)
672 					pr_cont(" NTC/U\n");
673 				else if (i == tx_ring->next_to_use)
674 					pr_cont(" NTU\n");
675 				else if (i == tx_ring->next_to_clean)
676 					pr_cont(" NTC\n");
677 				else
678 					pr_cont("\n");
679 
680 				if (netif_msg_pktdata(adapter) &&
681 				    tx_buffer->skb)
682 					print_hex_dump(KERN_INFO, "",
683 						DUMP_PREFIX_ADDRESS, 16, 1,
684 						tx_buffer->skb->data,
685 						dma_unmap_len(tx_buffer, len),
686 						true);
687 			}
688 		}
689 	}
690 
691 	/* Print RX Rings Summary */
692 rx_ring_summary:
693 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
694 	pr_info("Queue [NTU] [NTC]\n");
695 	for (n = 0; n < adapter->num_rx_queues; n++) {
696 		rx_ring = adapter->rx_ring[n];
697 		pr_info("%5d %5X %5X\n",
698 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
699 	}
700 
701 	/* Print RX Rings */
702 	if (!netif_msg_rx_status(adapter))
703 		return;
704 
705 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
706 
707 	/* Receive Descriptor Formats
708 	 *
709 	 * 82598 Advanced Receive Descriptor (Read) Format
710 	 *    63                                           1        0
711 	 *    +-----------------------------------------------------+
712 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
713 	 *    +----------------------------------------------+------+
714 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
715 	 *    +-----------------------------------------------------+
716 	 *
717 	 *
718 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
719 	 *
720 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
721 	 *   +------------------------------------------------------+
722 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
723 	 *   | Packet   | IP     |   |          |     | Type | Type |
724 	 *   | Checksum | Ident  |   |          |     |      |      |
725 	 *   +------------------------------------------------------+
726 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
727 	 *   +------------------------------------------------------+
728 	 *   63       48 47    32 31            20 19               0
729 	 *
730 	 * 82599+ Advanced Receive Descriptor (Read) Format
731 	 *    63                                           1        0
732 	 *    +-----------------------------------------------------+
733 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
734 	 *    +----------------------------------------------+------+
735 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
736 	 *    +-----------------------------------------------------+
737 	 *
738 	 *
739 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
740 	 *
741 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
742 	 *   +------------------------------------------------------+
743 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
744 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
745 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
746 	 *   +------------------------------------------------------+
747 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
748 	 *   +------------------------------------------------------+
749 	 *   63       48 47    32 31          20 19                 0
750 	 */
751 
752 	for (n = 0; n < adapter->num_rx_queues; n++) {
753 		rx_ring = adapter->rx_ring[n];
754 		pr_info("------------------------------------\n");
755 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
756 		pr_info("------------------------------------\n");
757 		pr_info("%s%s%s",
758 			"R  [desc]      [ PktBuf     A0] ",
759 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
760 			"<-- Adv Rx Read format\n");
761 		pr_info("%s%s%s",
762 			"RWB[desc]      [PcsmIpSHl PtRs] ",
763 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
764 			"<-- Adv Rx Write-Back format\n");
765 
766 		for (i = 0; i < rx_ring->count; i++) {
767 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
768 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
769 			u0 = (struct my_u0 *)rx_desc;
770 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
771 			if (staterr & IXGBE_RXD_STAT_DD) {
772 				/* Descriptor Done */
773 				pr_info("RWB[0x%03X]     %016llX "
774 					"%016llX ---------------- %p", i,
775 					le64_to_cpu(u0->a),
776 					le64_to_cpu(u0->b),
777 					rx_buffer_info->skb);
778 			} else {
779 				pr_info("R  [0x%03X]     %016llX "
780 					"%016llX %016llX %p", i,
781 					le64_to_cpu(u0->a),
782 					le64_to_cpu(u0->b),
783 					(u64)rx_buffer_info->dma,
784 					rx_buffer_info->skb);
785 
786 				if (netif_msg_pktdata(adapter) &&
787 				    rx_buffer_info->dma) {
788 					print_hex_dump(KERN_INFO, "",
789 					   DUMP_PREFIX_ADDRESS, 16, 1,
790 					   page_address(rx_buffer_info->page) +
791 						    rx_buffer_info->page_offset,
792 					   ixgbe_rx_bufsz(rx_ring), true);
793 				}
794 			}
795 
796 			if (i == rx_ring->next_to_use)
797 				pr_cont(" NTU\n");
798 			else if (i == rx_ring->next_to_clean)
799 				pr_cont(" NTC\n");
800 			else
801 				pr_cont("\n");
802 
803 		}
804 	}
805 }
806 
807 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
808 {
809 	u32 ctrl_ext;
810 
811 	/* Let firmware take over control of h/w */
812 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
813 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
814 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
815 }
816 
817 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
818 {
819 	u32 ctrl_ext;
820 
821 	/* Let firmware know the driver has taken over */
822 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
823 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
824 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
825 }
826 
827 /**
828  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
829  * @adapter: pointer to adapter struct
830  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
831  * @queue: queue to map the corresponding interrupt to
832  * @msix_vector: the vector to map to the corresponding queue
833  *
834  */
835 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
836 			   u8 queue, u8 msix_vector)
837 {
838 	u32 ivar, index;
839 	struct ixgbe_hw *hw = &adapter->hw;
840 	switch (hw->mac.type) {
841 	case ixgbe_mac_82598EB:
842 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
843 		if (direction == -1)
844 			direction = 0;
845 		index = (((direction * 64) + queue) >> 2) & 0x1F;
846 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
847 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
848 		ivar |= (msix_vector << (8 * (queue & 0x3)));
849 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
850 		break;
851 	case ixgbe_mac_82599EB:
852 	case ixgbe_mac_X540:
853 	case ixgbe_mac_X550:
854 	case ixgbe_mac_X550EM_x:
855 		if (direction == -1) {
856 			/* other causes */
857 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
858 			index = ((queue & 1) * 8);
859 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
860 			ivar &= ~(0xFF << index);
861 			ivar |= (msix_vector << index);
862 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
863 			break;
864 		} else {
865 			/* tx or rx causes */
866 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
867 			index = ((16 * (queue & 1)) + (8 * direction));
868 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
869 			ivar &= ~(0xFF << index);
870 			ivar |= (msix_vector << index);
871 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
872 			break;
873 		}
874 	default:
875 		break;
876 	}
877 }
878 
879 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
880 					  u64 qmask)
881 {
882 	u32 mask;
883 
884 	switch (adapter->hw.mac.type) {
885 	case ixgbe_mac_82598EB:
886 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
887 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
888 		break;
889 	case ixgbe_mac_82599EB:
890 	case ixgbe_mac_X540:
891 	case ixgbe_mac_X550:
892 	case ixgbe_mac_X550EM_x:
893 		mask = (qmask & 0xFFFFFFFF);
894 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
895 		mask = (qmask >> 32);
896 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
897 		break;
898 	default:
899 		break;
900 	}
901 }
902 
903 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring *ring,
904 				      struct ixgbe_tx_buffer *tx_buffer)
905 {
906 	if (tx_buffer->skb) {
907 		dev_kfree_skb_any(tx_buffer->skb);
908 		if (dma_unmap_len(tx_buffer, len))
909 			dma_unmap_single(ring->dev,
910 					 dma_unmap_addr(tx_buffer, dma),
911 					 dma_unmap_len(tx_buffer, len),
912 					 DMA_TO_DEVICE);
913 	} else if (dma_unmap_len(tx_buffer, len)) {
914 		dma_unmap_page(ring->dev,
915 			       dma_unmap_addr(tx_buffer, dma),
916 			       dma_unmap_len(tx_buffer, len),
917 			       DMA_TO_DEVICE);
918 	}
919 	tx_buffer->next_to_watch = NULL;
920 	tx_buffer->skb = NULL;
921 	dma_unmap_len_set(tx_buffer, len, 0);
922 	/* tx_buffer must be completely set up in the transmit path */
923 }
924 
925 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
926 {
927 	struct ixgbe_hw *hw = &adapter->hw;
928 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
929 	int i;
930 	u32 data;
931 
932 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
933 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
934 		return;
935 
936 	switch (hw->mac.type) {
937 	case ixgbe_mac_82598EB:
938 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
939 		break;
940 	default:
941 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
942 	}
943 	hwstats->lxoffrxc += data;
944 
945 	/* refill credits (no tx hang) if we received xoff */
946 	if (!data)
947 		return;
948 
949 	for (i = 0; i < adapter->num_tx_queues; i++)
950 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
951 			  &adapter->tx_ring[i]->state);
952 }
953 
954 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
955 {
956 	struct ixgbe_hw *hw = &adapter->hw;
957 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
958 	u32 xoff[8] = {0};
959 	u8 tc;
960 	int i;
961 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
962 
963 	if (adapter->ixgbe_ieee_pfc)
964 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
965 
966 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
967 		ixgbe_update_xoff_rx_lfc(adapter);
968 		return;
969 	}
970 
971 	/* update stats for each tc, only valid with PFC enabled */
972 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
973 		u32 pxoffrxc;
974 
975 		switch (hw->mac.type) {
976 		case ixgbe_mac_82598EB:
977 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
978 			break;
979 		default:
980 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
981 		}
982 		hwstats->pxoffrxc[i] += pxoffrxc;
983 		/* Get the TC for given UP */
984 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
985 		xoff[tc] += pxoffrxc;
986 	}
987 
988 	/* disarm tx queues that have received xoff frames */
989 	for (i = 0; i < adapter->num_tx_queues; i++) {
990 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
991 
992 		tc = tx_ring->dcb_tc;
993 		if (xoff[tc])
994 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
995 	}
996 }
997 
998 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
999 {
1000 	return ring->stats.packets;
1001 }
1002 
1003 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1004 {
1005 	struct ixgbe_adapter *adapter;
1006 	struct ixgbe_hw *hw;
1007 	u32 head, tail;
1008 
1009 	if (ring->l2_accel_priv)
1010 		adapter = ring->l2_accel_priv->real_adapter;
1011 	else
1012 		adapter = netdev_priv(ring->netdev);
1013 
1014 	hw = &adapter->hw;
1015 	head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
1016 	tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
1017 
1018 	if (head != tail)
1019 		return (head < tail) ?
1020 			tail - head : (tail + ring->count - head);
1021 
1022 	return 0;
1023 }
1024 
1025 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1026 {
1027 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1028 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1029 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1030 
1031 	clear_check_for_tx_hang(tx_ring);
1032 
1033 	/*
1034 	 * Check for a hung queue, but be thorough. This verifies
1035 	 * that a transmit has been completed since the previous
1036 	 * check AND there is at least one packet pending. The
1037 	 * ARMED bit is set to indicate a potential hang. The
1038 	 * bit is cleared if a pause frame is received to remove
1039 	 * false hang detection due to PFC or 802.3x frames. By
1040 	 * requiring this to fail twice we avoid races with
1041 	 * pfc clearing the ARMED bit and conditions where we
1042 	 * run the check_tx_hang logic with a transmit completion
1043 	 * pending but without time to complete it yet.
1044 	 */
1045 	if (tx_done_old == tx_done && tx_pending)
1046 		/* make sure it is true for two checks in a row */
1047 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1048 					&tx_ring->state);
1049 	/* update completed stats and continue */
1050 	tx_ring->tx_stats.tx_done_old = tx_done;
1051 	/* reset the countdown */
1052 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1053 
1054 	return false;
1055 }
1056 
1057 /**
1058  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1059  * @adapter: driver private struct
1060  **/
1061 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1062 {
1063 
1064 	/* Do the reset outside of interrupt context */
1065 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1066 		adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
1067 		e_warn(drv, "initiating reset due to tx timeout\n");
1068 		ixgbe_service_event_schedule(adapter);
1069 	}
1070 }
1071 
1072 /**
1073  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1074  * @q_vector: structure containing interrupt and ring information
1075  * @tx_ring: tx ring to clean
1076  **/
1077 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1078 			       struct ixgbe_ring *tx_ring)
1079 {
1080 	struct ixgbe_adapter *adapter = q_vector->adapter;
1081 	struct ixgbe_tx_buffer *tx_buffer;
1082 	union ixgbe_adv_tx_desc *tx_desc;
1083 	unsigned int total_bytes = 0, total_packets = 0;
1084 	unsigned int budget = q_vector->tx.work_limit;
1085 	unsigned int i = tx_ring->next_to_clean;
1086 
1087 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1088 		return true;
1089 
1090 	tx_buffer = &tx_ring->tx_buffer_info[i];
1091 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1092 	i -= tx_ring->count;
1093 
1094 	do {
1095 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1096 
1097 		/* if next_to_watch is not set then there is no work pending */
1098 		if (!eop_desc)
1099 			break;
1100 
1101 		/* prevent any other reads prior to eop_desc */
1102 		read_barrier_depends();
1103 
1104 		/* if DD is not set pending work has not been completed */
1105 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1106 			break;
1107 
1108 		/* clear next_to_watch to prevent false hangs */
1109 		tx_buffer->next_to_watch = NULL;
1110 
1111 		/* update the statistics for this packet */
1112 		total_bytes += tx_buffer->bytecount;
1113 		total_packets += tx_buffer->gso_segs;
1114 
1115 		/* free the skb */
1116 		dev_consume_skb_any(tx_buffer->skb);
1117 
1118 		/* unmap skb header data */
1119 		dma_unmap_single(tx_ring->dev,
1120 				 dma_unmap_addr(tx_buffer, dma),
1121 				 dma_unmap_len(tx_buffer, len),
1122 				 DMA_TO_DEVICE);
1123 
1124 		/* clear tx_buffer data */
1125 		tx_buffer->skb = NULL;
1126 		dma_unmap_len_set(tx_buffer, len, 0);
1127 
1128 		/* unmap remaining buffers */
1129 		while (tx_desc != eop_desc) {
1130 			tx_buffer++;
1131 			tx_desc++;
1132 			i++;
1133 			if (unlikely(!i)) {
1134 				i -= tx_ring->count;
1135 				tx_buffer = tx_ring->tx_buffer_info;
1136 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1137 			}
1138 
1139 			/* unmap any remaining paged data */
1140 			if (dma_unmap_len(tx_buffer, len)) {
1141 				dma_unmap_page(tx_ring->dev,
1142 					       dma_unmap_addr(tx_buffer, dma),
1143 					       dma_unmap_len(tx_buffer, len),
1144 					       DMA_TO_DEVICE);
1145 				dma_unmap_len_set(tx_buffer, len, 0);
1146 			}
1147 		}
1148 
1149 		/* move us one more past the eop_desc for start of next pkt */
1150 		tx_buffer++;
1151 		tx_desc++;
1152 		i++;
1153 		if (unlikely(!i)) {
1154 			i -= tx_ring->count;
1155 			tx_buffer = tx_ring->tx_buffer_info;
1156 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1157 		}
1158 
1159 		/* issue prefetch for next Tx descriptor */
1160 		prefetch(tx_desc);
1161 
1162 		/* update budget accounting */
1163 		budget--;
1164 	} while (likely(budget));
1165 
1166 	i += tx_ring->count;
1167 	tx_ring->next_to_clean = i;
1168 	u64_stats_update_begin(&tx_ring->syncp);
1169 	tx_ring->stats.bytes += total_bytes;
1170 	tx_ring->stats.packets += total_packets;
1171 	u64_stats_update_end(&tx_ring->syncp);
1172 	q_vector->tx.total_bytes += total_bytes;
1173 	q_vector->tx.total_packets += total_packets;
1174 
1175 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1176 		/* schedule immediate reset if we believe we hung */
1177 		struct ixgbe_hw *hw = &adapter->hw;
1178 		e_err(drv, "Detected Tx Unit Hang\n"
1179 			"  Tx Queue             <%d>\n"
1180 			"  TDH, TDT             <%x>, <%x>\n"
1181 			"  next_to_use          <%x>\n"
1182 			"  next_to_clean        <%x>\n"
1183 			"tx_buffer_info[next_to_clean]\n"
1184 			"  time_stamp           <%lx>\n"
1185 			"  jiffies              <%lx>\n",
1186 			tx_ring->queue_index,
1187 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1188 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1189 			tx_ring->next_to_use, i,
1190 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1191 
1192 		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
1193 
1194 		e_info(probe,
1195 		       "tx hang %d detected on queue %d, resetting adapter\n",
1196 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1197 
1198 		/* schedule immediate reset if we believe we hung */
1199 		ixgbe_tx_timeout_reset(adapter);
1200 
1201 		/* the adapter is about to reset, no point in enabling stuff */
1202 		return true;
1203 	}
1204 
1205 	netdev_tx_completed_queue(txring_txq(tx_ring),
1206 				  total_packets, total_bytes);
1207 
1208 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1209 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1210 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1211 		/* Make sure that anybody stopping the queue after this
1212 		 * sees the new next_to_clean.
1213 		 */
1214 		smp_mb();
1215 		if (__netif_subqueue_stopped(tx_ring->netdev,
1216 					     tx_ring->queue_index)
1217 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1218 			netif_wake_subqueue(tx_ring->netdev,
1219 					    tx_ring->queue_index);
1220 			++tx_ring->tx_stats.restart_queue;
1221 		}
1222 	}
1223 
1224 	return !!budget;
1225 }
1226 
1227 #ifdef CONFIG_IXGBE_DCA
1228 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1229 				struct ixgbe_ring *tx_ring,
1230 				int cpu)
1231 {
1232 	struct ixgbe_hw *hw = &adapter->hw;
1233 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
1234 	u16 reg_offset;
1235 
1236 	switch (hw->mac.type) {
1237 	case ixgbe_mac_82598EB:
1238 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1239 		break;
1240 	case ixgbe_mac_82599EB:
1241 	case ixgbe_mac_X540:
1242 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1243 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1244 		break;
1245 	default:
1246 		/* for unknown hardware do not write register */
1247 		return;
1248 	}
1249 
1250 	/*
1251 	 * We can enable relaxed ordering for reads, but not writes when
1252 	 * DCA is enabled.  This is due to a known issue in some chipsets
1253 	 * which will cause the DCA tag to be cleared.
1254 	 */
1255 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1256 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1257 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1258 
1259 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1260 }
1261 
1262 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1263 				struct ixgbe_ring *rx_ring,
1264 				int cpu)
1265 {
1266 	struct ixgbe_hw *hw = &adapter->hw;
1267 	u32 rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1268 	u8 reg_idx = rx_ring->reg_idx;
1269 
1270 
1271 	switch (hw->mac.type) {
1272 	case ixgbe_mac_82599EB:
1273 	case ixgbe_mac_X540:
1274 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1275 		break;
1276 	default:
1277 		break;
1278 	}
1279 
1280 	/*
1281 	 * We can enable relaxed ordering for reads, but not writes when
1282 	 * DCA is enabled.  This is due to a known issue in some chipsets
1283 	 * which will cause the DCA tag to be cleared.
1284 	 */
1285 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1286 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1287 
1288 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1289 }
1290 
1291 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1292 {
1293 	struct ixgbe_adapter *adapter = q_vector->adapter;
1294 	struct ixgbe_ring *ring;
1295 	int cpu = get_cpu();
1296 
1297 	if (q_vector->cpu == cpu)
1298 		goto out_no_update;
1299 
1300 	ixgbe_for_each_ring(ring, q_vector->tx)
1301 		ixgbe_update_tx_dca(adapter, ring, cpu);
1302 
1303 	ixgbe_for_each_ring(ring, q_vector->rx)
1304 		ixgbe_update_rx_dca(adapter, ring, cpu);
1305 
1306 	q_vector->cpu = cpu;
1307 out_no_update:
1308 	put_cpu();
1309 }
1310 
1311 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1312 {
1313 	int i;
1314 
1315 	if (!(adapter->flags & IXGBE_FLAG_DCA_ENABLED))
1316 		return;
1317 
1318 	/* always use CB2 mode, difference is masked in the CB driver */
1319 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 2);
1320 
1321 	for (i = 0; i < adapter->num_q_vectors; i++) {
1322 		adapter->q_vector[i]->cpu = -1;
1323 		ixgbe_update_dca(adapter->q_vector[i]);
1324 	}
1325 }
1326 
1327 static int __ixgbe_notify_dca(struct device *dev, void *data)
1328 {
1329 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1330 	unsigned long event = *(unsigned long *)data;
1331 
1332 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1333 		return 0;
1334 
1335 	switch (event) {
1336 	case DCA_PROVIDER_ADD:
1337 		/* if we're already enabled, don't do it again */
1338 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1339 			break;
1340 		if (dca_add_requester(dev) == 0) {
1341 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1342 			ixgbe_setup_dca(adapter);
1343 			break;
1344 		}
1345 		/* Fall Through since DCA is disabled. */
1346 	case DCA_PROVIDER_REMOVE:
1347 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1348 			dca_remove_requester(dev);
1349 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1350 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
1351 		}
1352 		break;
1353 	}
1354 
1355 	return 0;
1356 }
1357 
1358 #endif /* CONFIG_IXGBE_DCA */
1359 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1360 				 union ixgbe_adv_rx_desc *rx_desc,
1361 				 struct sk_buff *skb)
1362 {
1363 	if (ring->netdev->features & NETIF_F_RXHASH)
1364 		skb_set_hash(skb,
1365 			     le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1366 			     PKT_HASH_TYPE_L3);
1367 }
1368 
1369 #ifdef IXGBE_FCOE
1370 /**
1371  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1372  * @ring: structure containing ring specific data
1373  * @rx_desc: advanced rx descriptor
1374  *
1375  * Returns : true if it is FCoE pkt
1376  */
1377 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1378 				    union ixgbe_adv_rx_desc *rx_desc)
1379 {
1380 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1381 
1382 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1383 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1384 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1385 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1386 }
1387 
1388 #endif /* IXGBE_FCOE */
1389 /**
1390  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1391  * @ring: structure containing ring specific data
1392  * @rx_desc: current Rx descriptor being processed
1393  * @skb: skb currently being received and modified
1394  **/
1395 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1396 				     union ixgbe_adv_rx_desc *rx_desc,
1397 				     struct sk_buff *skb)
1398 {
1399 	skb_checksum_none_assert(skb);
1400 
1401 	/* Rx csum disabled */
1402 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1403 		return;
1404 
1405 	/* if IP and error */
1406 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1407 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1408 		ring->rx_stats.csum_err++;
1409 		return;
1410 	}
1411 
1412 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1413 		return;
1414 
1415 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1416 		__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1417 
1418 		/*
1419 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1420 		 * checksum errors.
1421 		 */
1422 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1423 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1424 			return;
1425 
1426 		ring->rx_stats.csum_err++;
1427 		return;
1428 	}
1429 
1430 	/* It must be a TCP or UDP packet with a valid checksum */
1431 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1432 }
1433 
1434 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1435 				    struct ixgbe_rx_buffer *bi)
1436 {
1437 	struct page *page = bi->page;
1438 	dma_addr_t dma;
1439 
1440 	/* since we are recycling buffers we should seldom need to alloc */
1441 	if (likely(page))
1442 		return true;
1443 
1444 	/* alloc new page for storage */
1445 	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1446 	if (unlikely(!page)) {
1447 		rx_ring->rx_stats.alloc_rx_page_failed++;
1448 		return false;
1449 	}
1450 
1451 	/* map page for use */
1452 	dma = dma_map_page(rx_ring->dev, page, 0,
1453 			   ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1454 
1455 	/*
1456 	 * if mapping failed free memory back to system since
1457 	 * there isn't much point in holding memory we can't use
1458 	 */
1459 	if (dma_mapping_error(rx_ring->dev, dma)) {
1460 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1461 
1462 		rx_ring->rx_stats.alloc_rx_page_failed++;
1463 		return false;
1464 	}
1465 
1466 	bi->dma = dma;
1467 	bi->page = page;
1468 	bi->page_offset = 0;
1469 
1470 	return true;
1471 }
1472 
1473 /**
1474  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1475  * @rx_ring: ring to place buffers on
1476  * @cleaned_count: number of buffers to replace
1477  **/
1478 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1479 {
1480 	union ixgbe_adv_rx_desc *rx_desc;
1481 	struct ixgbe_rx_buffer *bi;
1482 	u16 i = rx_ring->next_to_use;
1483 
1484 	/* nothing to do */
1485 	if (!cleaned_count)
1486 		return;
1487 
1488 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1489 	bi = &rx_ring->rx_buffer_info[i];
1490 	i -= rx_ring->count;
1491 
1492 	do {
1493 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1494 			break;
1495 
1496 		/*
1497 		 * Refresh the desc even if buffer_addrs didn't change
1498 		 * because each write-back erases this info.
1499 		 */
1500 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1501 
1502 		rx_desc++;
1503 		bi++;
1504 		i++;
1505 		if (unlikely(!i)) {
1506 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1507 			bi = rx_ring->rx_buffer_info;
1508 			i -= rx_ring->count;
1509 		}
1510 
1511 		/* clear the status bits for the next_to_use descriptor */
1512 		rx_desc->wb.upper.status_error = 0;
1513 
1514 		cleaned_count--;
1515 	} while (cleaned_count);
1516 
1517 	i += rx_ring->count;
1518 
1519 	if (rx_ring->next_to_use != i) {
1520 		rx_ring->next_to_use = i;
1521 
1522 		/* update next to alloc since we have filled the ring */
1523 		rx_ring->next_to_alloc = i;
1524 
1525 		/* Force memory writes to complete before letting h/w
1526 		 * know there are new descriptors to fetch.  (Only
1527 		 * applicable for weak-ordered memory model archs,
1528 		 * such as IA-64).
1529 		 */
1530 		wmb();
1531 		writel(i, rx_ring->tail);
1532 	}
1533 }
1534 
1535 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1536 				   struct sk_buff *skb)
1537 {
1538 	u16 hdr_len = skb_headlen(skb);
1539 
1540 	/* set gso_size to avoid messing up TCP MSS */
1541 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1542 						 IXGBE_CB(skb)->append_cnt);
1543 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1544 }
1545 
1546 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1547 				   struct sk_buff *skb)
1548 {
1549 	/* if append_cnt is 0 then frame is not RSC */
1550 	if (!IXGBE_CB(skb)->append_cnt)
1551 		return;
1552 
1553 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1554 	rx_ring->rx_stats.rsc_flush++;
1555 
1556 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1557 
1558 	/* gso_size is computed using append_cnt so always clear it last */
1559 	IXGBE_CB(skb)->append_cnt = 0;
1560 }
1561 
1562 /**
1563  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1564  * @rx_ring: rx descriptor ring packet is being transacted on
1565  * @rx_desc: pointer to the EOP Rx descriptor
1566  * @skb: pointer to current skb being populated
1567  *
1568  * This function checks the ring, descriptor, and packet information in
1569  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1570  * other fields within the skb.
1571  **/
1572 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1573 				     union ixgbe_adv_rx_desc *rx_desc,
1574 				     struct sk_buff *skb)
1575 {
1576 	struct net_device *dev = rx_ring->netdev;
1577 
1578 	ixgbe_update_rsc_stats(rx_ring, skb);
1579 
1580 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1581 
1582 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1583 
1584 	if (unlikely(ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_TS)))
1585 		ixgbe_ptp_rx_hwtstamp(rx_ring->q_vector->adapter, skb);
1586 
1587 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1588 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1589 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1590 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1591 	}
1592 
1593 	skb_record_rx_queue(skb, rx_ring->queue_index);
1594 
1595 	skb->protocol = eth_type_trans(skb, dev);
1596 }
1597 
1598 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1599 			 struct sk_buff *skb)
1600 {
1601 	struct ixgbe_adapter *adapter = q_vector->adapter;
1602 
1603 	if (ixgbe_qv_busy_polling(q_vector))
1604 		netif_receive_skb(skb);
1605 	else if (!(adapter->flags & IXGBE_FLAG_IN_NETPOLL))
1606 		napi_gro_receive(&q_vector->napi, skb);
1607 	else
1608 		netif_rx(skb);
1609 }
1610 
1611 /**
1612  * ixgbe_is_non_eop - process handling of non-EOP buffers
1613  * @rx_ring: Rx ring being processed
1614  * @rx_desc: Rx descriptor for current buffer
1615  * @skb: Current socket buffer containing buffer in progress
1616  *
1617  * This function updates next to clean.  If the buffer is an EOP buffer
1618  * this function exits returning false, otherwise it will place the
1619  * sk_buff in the next buffer to be chained and return true indicating
1620  * that this is in fact a non-EOP buffer.
1621  **/
1622 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1623 			     union ixgbe_adv_rx_desc *rx_desc,
1624 			     struct sk_buff *skb)
1625 {
1626 	u32 ntc = rx_ring->next_to_clean + 1;
1627 
1628 	/* fetch, update, and store next to clean */
1629 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1630 	rx_ring->next_to_clean = ntc;
1631 
1632 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1633 
1634 	/* update RSC append count if present */
1635 	if (ring_is_rsc_enabled(rx_ring)) {
1636 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1637 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1638 
1639 		if (unlikely(rsc_enabled)) {
1640 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1641 
1642 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1643 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1644 
1645 			/* update ntc based on RSC value */
1646 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1647 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1648 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1649 		}
1650 	}
1651 
1652 	/* if we are the last buffer then there is nothing else to do */
1653 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1654 		return false;
1655 
1656 	/* place skb in next buffer to be received */
1657 	rx_ring->rx_buffer_info[ntc].skb = skb;
1658 	rx_ring->rx_stats.non_eop_descs++;
1659 
1660 	return true;
1661 }
1662 
1663 /**
1664  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1665  * @rx_ring: rx descriptor ring packet is being transacted on
1666  * @skb: pointer to current skb being adjusted
1667  *
1668  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1669  * main difference between this version and the original function is that
1670  * this function can make several assumptions about the state of things
1671  * that allow for significant optimizations versus the standard function.
1672  * As a result we can do things like drop a frag and maintain an accurate
1673  * truesize for the skb.
1674  */
1675 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1676 			    struct sk_buff *skb)
1677 {
1678 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1679 	unsigned char *va;
1680 	unsigned int pull_len;
1681 
1682 	/*
1683 	 * it is valid to use page_address instead of kmap since we are
1684 	 * working with pages allocated out of the lomem pool per
1685 	 * alloc_page(GFP_ATOMIC)
1686 	 */
1687 	va = skb_frag_address(frag);
1688 
1689 	/*
1690 	 * we need the header to contain the greater of either ETH_HLEN or
1691 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1692 	 */
1693 	pull_len = eth_get_headlen(va, IXGBE_RX_HDR_SIZE);
1694 
1695 	/* align pull length to size of long to optimize memcpy performance */
1696 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1697 
1698 	/* update all of the pointers */
1699 	skb_frag_size_sub(frag, pull_len);
1700 	frag->page_offset += pull_len;
1701 	skb->data_len -= pull_len;
1702 	skb->tail += pull_len;
1703 }
1704 
1705 /**
1706  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1707  * @rx_ring: rx descriptor ring packet is being transacted on
1708  * @skb: pointer to current skb being updated
1709  *
1710  * This function provides a basic DMA sync up for the first fragment of an
1711  * skb.  The reason for doing this is that the first fragment cannot be
1712  * unmapped until we have reached the end of packet descriptor for a buffer
1713  * chain.
1714  */
1715 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1716 				struct sk_buff *skb)
1717 {
1718 	/* if the page was released unmap it, else just sync our portion */
1719 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1720 		dma_unmap_page(rx_ring->dev, IXGBE_CB(skb)->dma,
1721 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1722 		IXGBE_CB(skb)->page_released = false;
1723 	} else {
1724 		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1725 
1726 		dma_sync_single_range_for_cpu(rx_ring->dev,
1727 					      IXGBE_CB(skb)->dma,
1728 					      frag->page_offset,
1729 					      ixgbe_rx_bufsz(rx_ring),
1730 					      DMA_FROM_DEVICE);
1731 	}
1732 	IXGBE_CB(skb)->dma = 0;
1733 }
1734 
1735 /**
1736  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1737  * @rx_ring: rx descriptor ring packet is being transacted on
1738  * @rx_desc: pointer to the EOP Rx descriptor
1739  * @skb: pointer to current skb being fixed
1740  *
1741  * Check for corrupted packet headers caused by senders on the local L2
1742  * embedded NIC switch not setting up their Tx Descriptors right.  These
1743  * should be very rare.
1744  *
1745  * Also address the case where we are pulling data in on pages only
1746  * and as such no data is present in the skb header.
1747  *
1748  * In addition if skb is not at least 60 bytes we need to pad it so that
1749  * it is large enough to qualify as a valid Ethernet frame.
1750  *
1751  * Returns true if an error was encountered and skb was freed.
1752  **/
1753 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1754 				  union ixgbe_adv_rx_desc *rx_desc,
1755 				  struct sk_buff *skb)
1756 {
1757 	struct net_device *netdev = rx_ring->netdev;
1758 
1759 	/* verify that the packet does not have any known errors */
1760 	if (unlikely(ixgbe_test_staterr(rx_desc,
1761 					IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1762 	    !(netdev->features & NETIF_F_RXALL))) {
1763 		dev_kfree_skb_any(skb);
1764 		return true;
1765 	}
1766 
1767 	/* place header in linear portion of buffer */
1768 	if (skb_is_nonlinear(skb))
1769 		ixgbe_pull_tail(rx_ring, skb);
1770 
1771 #ifdef IXGBE_FCOE
1772 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1773 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1774 		return false;
1775 
1776 #endif
1777 	/* if eth_skb_pad returns an error the skb was freed */
1778 	if (eth_skb_pad(skb))
1779 		return true;
1780 
1781 	return false;
1782 }
1783 
1784 /**
1785  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1786  * @rx_ring: rx descriptor ring to store buffers on
1787  * @old_buff: donor buffer to have page reused
1788  *
1789  * Synchronizes page for reuse by the adapter
1790  **/
1791 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1792 				struct ixgbe_rx_buffer *old_buff)
1793 {
1794 	struct ixgbe_rx_buffer *new_buff;
1795 	u16 nta = rx_ring->next_to_alloc;
1796 
1797 	new_buff = &rx_ring->rx_buffer_info[nta];
1798 
1799 	/* update, and store next to alloc */
1800 	nta++;
1801 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1802 
1803 	/* transfer page from old buffer to new buffer */
1804 	*new_buff = *old_buff;
1805 
1806 	/* sync the buffer for use by the device */
1807 	dma_sync_single_range_for_device(rx_ring->dev, new_buff->dma,
1808 					 new_buff->page_offset,
1809 					 ixgbe_rx_bufsz(rx_ring),
1810 					 DMA_FROM_DEVICE);
1811 }
1812 
1813 static inline bool ixgbe_page_is_reserved(struct page *page)
1814 {
1815 	return (page_to_nid(page) != numa_mem_id()) || page->pfmemalloc;
1816 }
1817 
1818 /**
1819  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1820  * @rx_ring: rx descriptor ring to transact packets on
1821  * @rx_buffer: buffer containing page to add
1822  * @rx_desc: descriptor containing length of buffer written by hardware
1823  * @skb: sk_buff to place the data into
1824  *
1825  * This function will add the data contained in rx_buffer->page to the skb.
1826  * This is done either through a direct copy if the data in the buffer is
1827  * less than the skb header size, otherwise it will just attach the page as
1828  * a frag to the skb.
1829  *
1830  * The function will then update the page offset if necessary and return
1831  * true if the buffer can be reused by the adapter.
1832  **/
1833 static bool ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1834 			      struct ixgbe_rx_buffer *rx_buffer,
1835 			      union ixgbe_adv_rx_desc *rx_desc,
1836 			      struct sk_buff *skb)
1837 {
1838 	struct page *page = rx_buffer->page;
1839 	unsigned int size = le16_to_cpu(rx_desc->wb.upper.length);
1840 #if (PAGE_SIZE < 8192)
1841 	unsigned int truesize = ixgbe_rx_bufsz(rx_ring);
1842 #else
1843 	unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
1844 	unsigned int last_offset = ixgbe_rx_pg_size(rx_ring) -
1845 				   ixgbe_rx_bufsz(rx_ring);
1846 #endif
1847 
1848 	if ((size <= IXGBE_RX_HDR_SIZE) && !skb_is_nonlinear(skb)) {
1849 		unsigned char *va = page_address(page) + rx_buffer->page_offset;
1850 
1851 		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
1852 
1853 		/* page is not reserved, we can reuse buffer as-is */
1854 		if (likely(!ixgbe_page_is_reserved(page)))
1855 			return true;
1856 
1857 		/* this page cannot be reused so discard it */
1858 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1859 		return false;
1860 	}
1861 
1862 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
1863 			rx_buffer->page_offset, size, truesize);
1864 
1865 	/* avoid re-using remote pages */
1866 	if (unlikely(ixgbe_page_is_reserved(page)))
1867 		return false;
1868 
1869 #if (PAGE_SIZE < 8192)
1870 	/* if we are only owner of page we can reuse it */
1871 	if (unlikely(page_count(page) != 1))
1872 		return false;
1873 
1874 	/* flip page offset to other buffer */
1875 	rx_buffer->page_offset ^= truesize;
1876 #else
1877 	/* move offset up to the next cache line */
1878 	rx_buffer->page_offset += truesize;
1879 
1880 	if (rx_buffer->page_offset > last_offset)
1881 		return false;
1882 #endif
1883 
1884 	/* Even if we own the page, we are not allowed to use atomic_set()
1885 	 * This would break get_page_unless_zero() users.
1886 	 */
1887 	atomic_inc(&page->_count);
1888 
1889 	return true;
1890 }
1891 
1892 static struct sk_buff *ixgbe_fetch_rx_buffer(struct ixgbe_ring *rx_ring,
1893 					     union ixgbe_adv_rx_desc *rx_desc)
1894 {
1895 	struct ixgbe_rx_buffer *rx_buffer;
1896 	struct sk_buff *skb;
1897 	struct page *page;
1898 
1899 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
1900 	page = rx_buffer->page;
1901 	prefetchw(page);
1902 
1903 	skb = rx_buffer->skb;
1904 
1905 	if (likely(!skb)) {
1906 		void *page_addr = page_address(page) +
1907 				  rx_buffer->page_offset;
1908 
1909 		/* prefetch first cache line of first page */
1910 		prefetch(page_addr);
1911 #if L1_CACHE_BYTES < 128
1912 		prefetch(page_addr + L1_CACHE_BYTES);
1913 #endif
1914 
1915 		/* allocate a skb to store the frags */
1916 		skb = napi_alloc_skb(&rx_ring->q_vector->napi,
1917 				     IXGBE_RX_HDR_SIZE);
1918 		if (unlikely(!skb)) {
1919 			rx_ring->rx_stats.alloc_rx_buff_failed++;
1920 			return NULL;
1921 		}
1922 
1923 		/*
1924 		 * we will be copying header into skb->data in
1925 		 * pskb_may_pull so it is in our interest to prefetch
1926 		 * it now to avoid a possible cache miss
1927 		 */
1928 		prefetchw(skb->data);
1929 
1930 		/*
1931 		 * Delay unmapping of the first packet. It carries the
1932 		 * header information, HW may still access the header
1933 		 * after the writeback.  Only unmap it when EOP is
1934 		 * reached
1935 		 */
1936 		if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1937 			goto dma_sync;
1938 
1939 		IXGBE_CB(skb)->dma = rx_buffer->dma;
1940 	} else {
1941 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
1942 			ixgbe_dma_sync_frag(rx_ring, skb);
1943 
1944 dma_sync:
1945 		/* we are reusing so sync this buffer for CPU use */
1946 		dma_sync_single_range_for_cpu(rx_ring->dev,
1947 					      rx_buffer->dma,
1948 					      rx_buffer->page_offset,
1949 					      ixgbe_rx_bufsz(rx_ring),
1950 					      DMA_FROM_DEVICE);
1951 
1952 		rx_buffer->skb = NULL;
1953 	}
1954 
1955 	/* pull page into skb */
1956 	if (ixgbe_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) {
1957 		/* hand second half of page back to the ring */
1958 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
1959 	} else if (IXGBE_CB(skb)->dma == rx_buffer->dma) {
1960 		/* the page has been released from the ring */
1961 		IXGBE_CB(skb)->page_released = true;
1962 	} else {
1963 		/* we are not reusing the buffer so unmap it */
1964 		dma_unmap_page(rx_ring->dev, rx_buffer->dma,
1965 			       ixgbe_rx_pg_size(rx_ring),
1966 			       DMA_FROM_DEVICE);
1967 	}
1968 
1969 	/* clear contents of buffer_info */
1970 	rx_buffer->page = NULL;
1971 
1972 	return skb;
1973 }
1974 
1975 /**
1976  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
1977  * @q_vector: structure containing interrupt and ring information
1978  * @rx_ring: rx descriptor ring to transact packets on
1979  * @budget: Total limit on number of packets to process
1980  *
1981  * This function provides a "bounce buffer" approach to Rx interrupt
1982  * processing.  The advantage to this is that on systems that have
1983  * expensive overhead for IOMMU access this provides a means of avoiding
1984  * it by maintaining the mapping of the page to the syste.
1985  *
1986  * Returns amount of work completed
1987  **/
1988 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
1989 			       struct ixgbe_ring *rx_ring,
1990 			       const int budget)
1991 {
1992 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1993 #ifdef IXGBE_FCOE
1994 	struct ixgbe_adapter *adapter = q_vector->adapter;
1995 	int ddp_bytes;
1996 	unsigned int mss = 0;
1997 #endif /* IXGBE_FCOE */
1998 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
1999 
2000 	while (likely(total_rx_packets < budget)) {
2001 		union ixgbe_adv_rx_desc *rx_desc;
2002 		struct sk_buff *skb;
2003 
2004 		/* return some buffers to hardware, one at a time is too slow */
2005 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2006 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2007 			cleaned_count = 0;
2008 		}
2009 
2010 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2011 
2012 		if (!rx_desc->wb.upper.status_error)
2013 			break;
2014 
2015 		/* This memory barrier is needed to keep us from reading
2016 		 * any other fields out of the rx_desc until we know the
2017 		 * descriptor has been written back
2018 		 */
2019 		dma_rmb();
2020 
2021 		/* retrieve a buffer from the ring */
2022 		skb = ixgbe_fetch_rx_buffer(rx_ring, rx_desc);
2023 
2024 		/* exit if we failed to retrieve a buffer */
2025 		if (!skb)
2026 			break;
2027 
2028 		cleaned_count++;
2029 
2030 		/* place incomplete frames back on ring for completion */
2031 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2032 			continue;
2033 
2034 		/* verify the packet layout is correct */
2035 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2036 			continue;
2037 
2038 		/* probably a little skewed due to removing CRC */
2039 		total_rx_bytes += skb->len;
2040 
2041 		/* populate checksum, timestamp, VLAN, and protocol */
2042 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2043 
2044 #ifdef IXGBE_FCOE
2045 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2046 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2047 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2048 			/* include DDPed FCoE data */
2049 			if (ddp_bytes > 0) {
2050 				if (!mss) {
2051 					mss = rx_ring->netdev->mtu -
2052 						sizeof(struct fcoe_hdr) -
2053 						sizeof(struct fc_frame_header) -
2054 						sizeof(struct fcoe_crc_eof);
2055 					if (mss > 512)
2056 						mss &= ~511;
2057 				}
2058 				total_rx_bytes += ddp_bytes;
2059 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2060 								 mss);
2061 			}
2062 			if (!ddp_bytes) {
2063 				dev_kfree_skb_any(skb);
2064 				continue;
2065 			}
2066 		}
2067 
2068 #endif /* IXGBE_FCOE */
2069 		skb_mark_napi_id(skb, &q_vector->napi);
2070 		ixgbe_rx_skb(q_vector, skb);
2071 
2072 		/* update budget accounting */
2073 		total_rx_packets++;
2074 	}
2075 
2076 	u64_stats_update_begin(&rx_ring->syncp);
2077 	rx_ring->stats.packets += total_rx_packets;
2078 	rx_ring->stats.bytes += total_rx_bytes;
2079 	u64_stats_update_end(&rx_ring->syncp);
2080 	q_vector->rx.total_packets += total_rx_packets;
2081 	q_vector->rx.total_bytes += total_rx_bytes;
2082 
2083 	return total_rx_packets;
2084 }
2085 
2086 #ifdef CONFIG_NET_RX_BUSY_POLL
2087 /* must be called with local_bh_disable()d */
2088 static int ixgbe_low_latency_recv(struct napi_struct *napi)
2089 {
2090 	struct ixgbe_q_vector *q_vector =
2091 			container_of(napi, struct ixgbe_q_vector, napi);
2092 	struct ixgbe_adapter *adapter = q_vector->adapter;
2093 	struct ixgbe_ring  *ring;
2094 	int found = 0;
2095 
2096 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2097 		return LL_FLUSH_FAILED;
2098 
2099 	if (!ixgbe_qv_lock_poll(q_vector))
2100 		return LL_FLUSH_BUSY;
2101 
2102 	ixgbe_for_each_ring(ring, q_vector->rx) {
2103 		found = ixgbe_clean_rx_irq(q_vector, ring, 4);
2104 #ifdef BP_EXTENDED_STATS
2105 		if (found)
2106 			ring->stats.cleaned += found;
2107 		else
2108 			ring->stats.misses++;
2109 #endif
2110 		if (found)
2111 			break;
2112 	}
2113 
2114 	ixgbe_qv_unlock_poll(q_vector);
2115 
2116 	return found;
2117 }
2118 #endif	/* CONFIG_NET_RX_BUSY_POLL */
2119 
2120 /**
2121  * ixgbe_configure_msix - Configure MSI-X hardware
2122  * @adapter: board private structure
2123  *
2124  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2125  * interrupts.
2126  **/
2127 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2128 {
2129 	struct ixgbe_q_vector *q_vector;
2130 	int v_idx;
2131 	u32 mask;
2132 
2133 	/* Populate MSIX to EITR Select */
2134 	if (adapter->num_vfs > 32) {
2135 		u32 eitrsel = (1 << (adapter->num_vfs - 32)) - 1;
2136 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2137 	}
2138 
2139 	/*
2140 	 * Populate the IVAR table and set the ITR values to the
2141 	 * corresponding register.
2142 	 */
2143 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2144 		struct ixgbe_ring *ring;
2145 		q_vector = adapter->q_vector[v_idx];
2146 
2147 		ixgbe_for_each_ring(ring, q_vector->rx)
2148 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2149 
2150 		ixgbe_for_each_ring(ring, q_vector->tx)
2151 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2152 
2153 		ixgbe_write_eitr(q_vector);
2154 	}
2155 
2156 	switch (adapter->hw.mac.type) {
2157 	case ixgbe_mac_82598EB:
2158 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2159 			       v_idx);
2160 		break;
2161 	case ixgbe_mac_82599EB:
2162 	case ixgbe_mac_X540:
2163 	case ixgbe_mac_X550:
2164 	case ixgbe_mac_X550EM_x:
2165 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2166 		break;
2167 	default:
2168 		break;
2169 	}
2170 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2171 
2172 	/* set up to autoclear timer, and the vectors */
2173 	mask = IXGBE_EIMS_ENABLE_MASK;
2174 	mask &= ~(IXGBE_EIMS_OTHER |
2175 		  IXGBE_EIMS_MAILBOX |
2176 		  IXGBE_EIMS_LSC);
2177 
2178 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2179 }
2180 
2181 enum latency_range {
2182 	lowest_latency = 0,
2183 	low_latency = 1,
2184 	bulk_latency = 2,
2185 	latency_invalid = 255
2186 };
2187 
2188 /**
2189  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2190  * @q_vector: structure containing interrupt and ring information
2191  * @ring_container: structure containing ring performance data
2192  *
2193  *      Stores a new ITR value based on packets and byte
2194  *      counts during the last interrupt.  The advantage of per interrupt
2195  *      computation is faster updates and more accurate ITR for the current
2196  *      traffic pattern.  Constants in this function were computed
2197  *      based on theoretical maximum wire speed and thresholds were set based
2198  *      on testing data as well as attempting to minimize response time
2199  *      while increasing bulk throughput.
2200  *      this functionality is controlled by the InterruptThrottleRate module
2201  *      parameter (see ixgbe_param.c)
2202  **/
2203 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2204 			     struct ixgbe_ring_container *ring_container)
2205 {
2206 	int bytes = ring_container->total_bytes;
2207 	int packets = ring_container->total_packets;
2208 	u32 timepassed_us;
2209 	u64 bytes_perint;
2210 	u8 itr_setting = ring_container->itr;
2211 
2212 	if (packets == 0)
2213 		return;
2214 
2215 	/* simple throttlerate management
2216 	 *   0-10MB/s   lowest (100000 ints/s)
2217 	 *  10-20MB/s   low    (20000 ints/s)
2218 	 *  20-1249MB/s bulk   (8000 ints/s)
2219 	 */
2220 	/* what was last interrupt timeslice? */
2221 	timepassed_us = q_vector->itr >> 2;
2222 	if (timepassed_us == 0)
2223 		return;
2224 
2225 	bytes_perint = bytes / timepassed_us; /* bytes/usec */
2226 
2227 	switch (itr_setting) {
2228 	case lowest_latency:
2229 		if (bytes_perint > 10)
2230 			itr_setting = low_latency;
2231 		break;
2232 	case low_latency:
2233 		if (bytes_perint > 20)
2234 			itr_setting = bulk_latency;
2235 		else if (bytes_perint <= 10)
2236 			itr_setting = lowest_latency;
2237 		break;
2238 	case bulk_latency:
2239 		if (bytes_perint <= 20)
2240 			itr_setting = low_latency;
2241 		break;
2242 	}
2243 
2244 	/* clear work counters since we have the values we need */
2245 	ring_container->total_bytes = 0;
2246 	ring_container->total_packets = 0;
2247 
2248 	/* write updated itr to ring container */
2249 	ring_container->itr = itr_setting;
2250 }
2251 
2252 /**
2253  * ixgbe_write_eitr - write EITR register in hardware specific way
2254  * @q_vector: structure containing interrupt and ring information
2255  *
2256  * This function is made to be called by ethtool and by the driver
2257  * when it needs to update EITR registers at runtime.  Hardware
2258  * specific quirks/differences are taken care of here.
2259  */
2260 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2261 {
2262 	struct ixgbe_adapter *adapter = q_vector->adapter;
2263 	struct ixgbe_hw *hw = &adapter->hw;
2264 	int v_idx = q_vector->v_idx;
2265 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2266 
2267 	switch (adapter->hw.mac.type) {
2268 	case ixgbe_mac_82598EB:
2269 		/* must write high and low 16 bits to reset counter */
2270 		itr_reg |= (itr_reg << 16);
2271 		break;
2272 	case ixgbe_mac_82599EB:
2273 	case ixgbe_mac_X540:
2274 	case ixgbe_mac_X550:
2275 	case ixgbe_mac_X550EM_x:
2276 		/*
2277 		 * set the WDIS bit to not clear the timer bits and cause an
2278 		 * immediate assertion of the interrupt
2279 		 */
2280 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2281 		break;
2282 	default:
2283 		break;
2284 	}
2285 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2286 }
2287 
2288 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2289 {
2290 	u32 new_itr = q_vector->itr;
2291 	u8 current_itr;
2292 
2293 	ixgbe_update_itr(q_vector, &q_vector->tx);
2294 	ixgbe_update_itr(q_vector, &q_vector->rx);
2295 
2296 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
2297 
2298 	switch (current_itr) {
2299 	/* counts and packets in update_itr are dependent on these numbers */
2300 	case lowest_latency:
2301 		new_itr = IXGBE_100K_ITR;
2302 		break;
2303 	case low_latency:
2304 		new_itr = IXGBE_20K_ITR;
2305 		break;
2306 	case bulk_latency:
2307 		new_itr = IXGBE_8K_ITR;
2308 		break;
2309 	default:
2310 		break;
2311 	}
2312 
2313 	if (new_itr != q_vector->itr) {
2314 		/* do an exponential smoothing */
2315 		new_itr = (10 * new_itr * q_vector->itr) /
2316 			  ((9 * new_itr) + q_vector->itr);
2317 
2318 		/* save the algorithm value here */
2319 		q_vector->itr = new_itr;
2320 
2321 		ixgbe_write_eitr(q_vector);
2322 	}
2323 }
2324 
2325 /**
2326  * ixgbe_check_overtemp_subtask - check for over temperature
2327  * @adapter: pointer to adapter
2328  **/
2329 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2330 {
2331 	struct ixgbe_hw *hw = &adapter->hw;
2332 	u32 eicr = adapter->interrupt_event;
2333 
2334 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2335 		return;
2336 
2337 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) &&
2338 	    !(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2339 		return;
2340 
2341 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2342 
2343 	switch (hw->device_id) {
2344 	case IXGBE_DEV_ID_82599_T3_LOM:
2345 		/*
2346 		 * Since the warning interrupt is for both ports
2347 		 * we don't have to check if:
2348 		 *  - This interrupt wasn't for our port.
2349 		 *  - We may have missed the interrupt so always have to
2350 		 *    check if we  got a LSC
2351 		 */
2352 		if (!(eicr & IXGBE_EICR_GPI_SDP0) &&
2353 		    !(eicr & IXGBE_EICR_LSC))
2354 			return;
2355 
2356 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2357 			u32 speed;
2358 			bool link_up = false;
2359 
2360 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2361 
2362 			if (link_up)
2363 				return;
2364 		}
2365 
2366 		/* Check if this is not due to overtemp */
2367 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2368 			return;
2369 
2370 		break;
2371 	default:
2372 		if (!(eicr & IXGBE_EICR_GPI_SDP0))
2373 			return;
2374 		break;
2375 	}
2376 	e_crit(drv,
2377 	       "Network adapter has been stopped because it has over heated. "
2378 	       "Restart the computer. If the problem persists, "
2379 	       "power off the system and replace the adapter\n");
2380 
2381 	adapter->interrupt_event = 0;
2382 }
2383 
2384 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2385 {
2386 	struct ixgbe_hw *hw = &adapter->hw;
2387 
2388 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2389 	    (eicr & IXGBE_EICR_GPI_SDP1)) {
2390 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2391 		/* write to clear the interrupt */
2392 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2393 	}
2394 }
2395 
2396 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2397 {
2398 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2399 		return;
2400 
2401 	switch (adapter->hw.mac.type) {
2402 	case ixgbe_mac_82599EB:
2403 		/*
2404 		 * Need to check link state so complete overtemp check
2405 		 * on service task
2406 		 */
2407 		if (((eicr & IXGBE_EICR_GPI_SDP0) || (eicr & IXGBE_EICR_LSC)) &&
2408 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2409 			adapter->interrupt_event = eicr;
2410 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2411 			ixgbe_service_event_schedule(adapter);
2412 			return;
2413 		}
2414 		return;
2415 	case ixgbe_mac_X540:
2416 		if (!(eicr & IXGBE_EICR_TS))
2417 			return;
2418 		break;
2419 	default:
2420 		return;
2421 	}
2422 
2423 	e_crit(drv,
2424 	       "Network adapter has been stopped because it has over heated. "
2425 	       "Restart the computer. If the problem persists, "
2426 	       "power off the system and replace the adapter\n");
2427 }
2428 
2429 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2430 {
2431 	struct ixgbe_hw *hw = &adapter->hw;
2432 
2433 	if (eicr & IXGBE_EICR_GPI_SDP2) {
2434 		/* Clear the interrupt */
2435 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP2);
2436 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2437 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2438 			ixgbe_service_event_schedule(adapter);
2439 		}
2440 	}
2441 
2442 	if (eicr & IXGBE_EICR_GPI_SDP1) {
2443 		/* Clear the interrupt */
2444 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2445 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2446 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2447 			ixgbe_service_event_schedule(adapter);
2448 		}
2449 	}
2450 }
2451 
2452 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2453 {
2454 	struct ixgbe_hw *hw = &adapter->hw;
2455 
2456 	adapter->lsc_int++;
2457 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2458 	adapter->link_check_timeout = jiffies;
2459 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2460 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2461 		IXGBE_WRITE_FLUSH(hw);
2462 		ixgbe_service_event_schedule(adapter);
2463 	}
2464 }
2465 
2466 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2467 					   u64 qmask)
2468 {
2469 	u32 mask;
2470 	struct ixgbe_hw *hw = &adapter->hw;
2471 
2472 	switch (hw->mac.type) {
2473 	case ixgbe_mac_82598EB:
2474 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2475 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2476 		break;
2477 	case ixgbe_mac_82599EB:
2478 	case ixgbe_mac_X540:
2479 	case ixgbe_mac_X550:
2480 	case ixgbe_mac_X550EM_x:
2481 		mask = (qmask & 0xFFFFFFFF);
2482 		if (mask)
2483 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2484 		mask = (qmask >> 32);
2485 		if (mask)
2486 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2487 		break;
2488 	default:
2489 		break;
2490 	}
2491 	/* skip the flush */
2492 }
2493 
2494 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2495 					    u64 qmask)
2496 {
2497 	u32 mask;
2498 	struct ixgbe_hw *hw = &adapter->hw;
2499 
2500 	switch (hw->mac.type) {
2501 	case ixgbe_mac_82598EB:
2502 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2503 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2504 		break;
2505 	case ixgbe_mac_82599EB:
2506 	case ixgbe_mac_X540:
2507 	case ixgbe_mac_X550:
2508 	case ixgbe_mac_X550EM_x:
2509 		mask = (qmask & 0xFFFFFFFF);
2510 		if (mask)
2511 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2512 		mask = (qmask >> 32);
2513 		if (mask)
2514 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2515 		break;
2516 	default:
2517 		break;
2518 	}
2519 	/* skip the flush */
2520 }
2521 
2522 /**
2523  * ixgbe_irq_enable - Enable default interrupt generation settings
2524  * @adapter: board private structure
2525  **/
2526 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2527 				    bool flush)
2528 {
2529 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2530 
2531 	/* don't reenable LSC while waiting for link */
2532 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2533 		mask &= ~IXGBE_EIMS_LSC;
2534 
2535 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2536 		switch (adapter->hw.mac.type) {
2537 		case ixgbe_mac_82599EB:
2538 			mask |= IXGBE_EIMS_GPI_SDP0;
2539 			break;
2540 		case ixgbe_mac_X540:
2541 		case ixgbe_mac_X550:
2542 		case ixgbe_mac_X550EM_x:
2543 			mask |= IXGBE_EIMS_TS;
2544 			break;
2545 		default:
2546 			break;
2547 		}
2548 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2549 		mask |= IXGBE_EIMS_GPI_SDP1;
2550 	switch (adapter->hw.mac.type) {
2551 	case ixgbe_mac_82599EB:
2552 		mask |= IXGBE_EIMS_GPI_SDP1;
2553 		mask |= IXGBE_EIMS_GPI_SDP2;
2554 		/* fall through */
2555 	case ixgbe_mac_X540:
2556 	case ixgbe_mac_X550:
2557 	case ixgbe_mac_X550EM_x:
2558 		mask |= IXGBE_EIMS_ECC;
2559 		mask |= IXGBE_EIMS_MAILBOX;
2560 		break;
2561 	default:
2562 		break;
2563 	}
2564 
2565 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2566 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
2567 		mask |= IXGBE_EIMS_FLOW_DIR;
2568 
2569 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
2570 	if (queues)
2571 		ixgbe_irq_enable_queues(adapter, ~0);
2572 	if (flush)
2573 		IXGBE_WRITE_FLUSH(&adapter->hw);
2574 }
2575 
2576 static irqreturn_t ixgbe_msix_other(int irq, void *data)
2577 {
2578 	struct ixgbe_adapter *adapter = data;
2579 	struct ixgbe_hw *hw = &adapter->hw;
2580 	u32 eicr;
2581 
2582 	/*
2583 	 * Workaround for Silicon errata.  Use clear-by-write instead
2584 	 * of clear-by-read.  Reading with EICS will return the
2585 	 * interrupt causes without clearing, which later be done
2586 	 * with the write to EICR.
2587 	 */
2588 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2589 
2590 	/* The lower 16bits of the EICR register are for the queue interrupts
2591 	 * which should be masked here in order to not accidently clear them if
2592 	 * the bits are high when ixgbe_msix_other is called. There is a race
2593 	 * condition otherwise which results in possible performance loss
2594 	 * especially if the ixgbe_msix_other interrupt is triggering
2595 	 * consistently (as it would when PPS is turned on for the X540 device)
2596 	 */
2597 	eicr &= 0xFFFF0000;
2598 
2599 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2600 
2601 	if (eicr & IXGBE_EICR_LSC)
2602 		ixgbe_check_lsc(adapter);
2603 
2604 	if (eicr & IXGBE_EICR_MAILBOX)
2605 		ixgbe_msg_task(adapter);
2606 
2607 	switch (hw->mac.type) {
2608 	case ixgbe_mac_82599EB:
2609 	case ixgbe_mac_X540:
2610 	case ixgbe_mac_X550:
2611 	case ixgbe_mac_X550EM_x:
2612 		if (eicr & IXGBE_EICR_ECC) {
2613 			e_info(link, "Received ECC Err, initiating reset\n");
2614 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
2615 			ixgbe_service_event_schedule(adapter);
2616 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2617 		}
2618 		/* Handle Flow Director Full threshold interrupt */
2619 		if (eicr & IXGBE_EICR_FLOW_DIR) {
2620 			int reinit_count = 0;
2621 			int i;
2622 			for (i = 0; i < adapter->num_tx_queues; i++) {
2623 				struct ixgbe_ring *ring = adapter->tx_ring[i];
2624 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
2625 						       &ring->state))
2626 					reinit_count++;
2627 			}
2628 			if (reinit_count) {
2629 				/* no more flow director interrupts until after init */
2630 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
2631 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
2632 				ixgbe_service_event_schedule(adapter);
2633 			}
2634 		}
2635 		ixgbe_check_sfp_event(adapter, eicr);
2636 		ixgbe_check_overtemp_event(adapter, eicr);
2637 		break;
2638 	default:
2639 		break;
2640 	}
2641 
2642 	ixgbe_check_fan_failure(adapter, eicr);
2643 
2644 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2645 		ixgbe_ptp_check_pps_event(adapter, eicr);
2646 
2647 	/* re-enable the original interrupt state, no lsc, no queues */
2648 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2649 		ixgbe_irq_enable(adapter, false, false);
2650 
2651 	return IRQ_HANDLED;
2652 }
2653 
2654 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
2655 {
2656 	struct ixgbe_q_vector *q_vector = data;
2657 
2658 	/* EIAM disabled interrupts (on this vector) for us */
2659 
2660 	if (q_vector->rx.ring || q_vector->tx.ring)
2661 		napi_schedule(&q_vector->napi);
2662 
2663 	return IRQ_HANDLED;
2664 }
2665 
2666 /**
2667  * ixgbe_poll - NAPI Rx polling callback
2668  * @napi: structure for representing this polling device
2669  * @budget: how many packets driver is allowed to clean
2670  *
2671  * This function is used for legacy and MSI, NAPI mode
2672  **/
2673 int ixgbe_poll(struct napi_struct *napi, int budget)
2674 {
2675 	struct ixgbe_q_vector *q_vector =
2676 				container_of(napi, struct ixgbe_q_vector, napi);
2677 	struct ixgbe_adapter *adapter = q_vector->adapter;
2678 	struct ixgbe_ring *ring;
2679 	int per_ring_budget;
2680 	bool clean_complete = true;
2681 
2682 #ifdef CONFIG_IXGBE_DCA
2683 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
2684 		ixgbe_update_dca(q_vector);
2685 #endif
2686 
2687 	ixgbe_for_each_ring(ring, q_vector->tx)
2688 		clean_complete &= !!ixgbe_clean_tx_irq(q_vector, ring);
2689 
2690 	if (!ixgbe_qv_lock_napi(q_vector))
2691 		return budget;
2692 
2693 	/* attempt to distribute budget to each queue fairly, but don't allow
2694 	 * the budget to go below 1 because we'll exit polling */
2695 	if (q_vector->rx.count > 1)
2696 		per_ring_budget = max(budget/q_vector->rx.count, 1);
2697 	else
2698 		per_ring_budget = budget;
2699 
2700 	ixgbe_for_each_ring(ring, q_vector->rx)
2701 		clean_complete &= (ixgbe_clean_rx_irq(q_vector, ring,
2702 				   per_ring_budget) < per_ring_budget);
2703 
2704 	ixgbe_qv_unlock_napi(q_vector);
2705 	/* If all work not completed, return budget and keep polling */
2706 	if (!clean_complete)
2707 		return budget;
2708 
2709 	/* all work done, exit the polling mode */
2710 	napi_complete(napi);
2711 	if (adapter->rx_itr_setting & 1)
2712 		ixgbe_set_itr(q_vector);
2713 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2714 		ixgbe_irq_enable_queues(adapter, ((u64)1 << q_vector->v_idx));
2715 
2716 	return 0;
2717 }
2718 
2719 /**
2720  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2721  * @adapter: board private structure
2722  *
2723  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2724  * interrupts from the kernel.
2725  **/
2726 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
2727 {
2728 	struct net_device *netdev = adapter->netdev;
2729 	int vector, err;
2730 	int ri = 0, ti = 0;
2731 
2732 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2733 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2734 		struct msix_entry *entry = &adapter->msix_entries[vector];
2735 
2736 		if (q_vector->tx.ring && q_vector->rx.ring) {
2737 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2738 				 "%s-%s-%d", netdev->name, "TxRx", ri++);
2739 			ti++;
2740 		} else if (q_vector->rx.ring) {
2741 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2742 				 "%s-%s-%d", netdev->name, "rx", ri++);
2743 		} else if (q_vector->tx.ring) {
2744 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2745 				 "%s-%s-%d", netdev->name, "tx", ti++);
2746 		} else {
2747 			/* skip this unused q_vector */
2748 			continue;
2749 		}
2750 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
2751 				  q_vector->name, q_vector);
2752 		if (err) {
2753 			e_err(probe, "request_irq failed for MSIX interrupt "
2754 			      "Error: %d\n", err);
2755 			goto free_queue_irqs;
2756 		}
2757 		/* If Flow Director is enabled, set interrupt affinity */
2758 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2759 			/* assign the mask for this irq */
2760 			irq_set_affinity_hint(entry->vector,
2761 					      &q_vector->affinity_mask);
2762 		}
2763 	}
2764 
2765 	err = request_irq(adapter->msix_entries[vector].vector,
2766 			  ixgbe_msix_other, 0, netdev->name, adapter);
2767 	if (err) {
2768 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
2769 		goto free_queue_irqs;
2770 	}
2771 
2772 	return 0;
2773 
2774 free_queue_irqs:
2775 	while (vector) {
2776 		vector--;
2777 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
2778 				      NULL);
2779 		free_irq(adapter->msix_entries[vector].vector,
2780 			 adapter->q_vector[vector]);
2781 	}
2782 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
2783 	pci_disable_msix(adapter->pdev);
2784 	kfree(adapter->msix_entries);
2785 	adapter->msix_entries = NULL;
2786 	return err;
2787 }
2788 
2789 /**
2790  * ixgbe_intr - legacy mode Interrupt Handler
2791  * @irq: interrupt number
2792  * @data: pointer to a network interface device structure
2793  **/
2794 static irqreturn_t ixgbe_intr(int irq, void *data)
2795 {
2796 	struct ixgbe_adapter *adapter = data;
2797 	struct ixgbe_hw *hw = &adapter->hw;
2798 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2799 	u32 eicr;
2800 
2801 	/*
2802 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
2803 	 * before the read of EICR.
2804 	 */
2805 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
2806 
2807 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2808 	 * therefore no explicit interrupt disable is necessary */
2809 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2810 	if (!eicr) {
2811 		/*
2812 		 * shared interrupt alert!
2813 		 * make sure interrupts are enabled because the read will
2814 		 * have disabled interrupts due to EIAM
2815 		 * finish the workaround of silicon errata on 82598.  Unmask
2816 		 * the interrupt that we masked before the EICR read.
2817 		 */
2818 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
2819 			ixgbe_irq_enable(adapter, true, true);
2820 		return IRQ_NONE;	/* Not our interrupt */
2821 	}
2822 
2823 	if (eicr & IXGBE_EICR_LSC)
2824 		ixgbe_check_lsc(adapter);
2825 
2826 	switch (hw->mac.type) {
2827 	case ixgbe_mac_82599EB:
2828 		ixgbe_check_sfp_event(adapter, eicr);
2829 		/* Fall through */
2830 	case ixgbe_mac_X540:
2831 	case ixgbe_mac_X550:
2832 	case ixgbe_mac_X550EM_x:
2833 		if (eicr & IXGBE_EICR_ECC) {
2834 			e_info(link, "Received ECC Err, initiating reset\n");
2835 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
2836 			ixgbe_service_event_schedule(adapter);
2837 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2838 		}
2839 		ixgbe_check_overtemp_event(adapter, eicr);
2840 		break;
2841 	default:
2842 		break;
2843 	}
2844 
2845 	ixgbe_check_fan_failure(adapter, eicr);
2846 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2847 		ixgbe_ptp_check_pps_event(adapter, eicr);
2848 
2849 	/* would disable interrupts here but EIAM disabled it */
2850 	napi_schedule(&q_vector->napi);
2851 
2852 	/*
2853 	 * re-enable link(maybe) and non-queue interrupts, no flush.
2854 	 * ixgbe_poll will re-enable the queue interrupts
2855 	 */
2856 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2857 		ixgbe_irq_enable(adapter, false, false);
2858 
2859 	return IRQ_HANDLED;
2860 }
2861 
2862 /**
2863  * ixgbe_request_irq - initialize interrupts
2864  * @adapter: board private structure
2865  *
2866  * Attempts to configure interrupts using the best available
2867  * capabilities of the hardware and kernel.
2868  **/
2869 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
2870 {
2871 	struct net_device *netdev = adapter->netdev;
2872 	int err;
2873 
2874 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
2875 		err = ixgbe_request_msix_irqs(adapter);
2876 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
2877 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
2878 				  netdev->name, adapter);
2879 	else
2880 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
2881 				  netdev->name, adapter);
2882 
2883 	if (err)
2884 		e_err(probe, "request_irq failed, Error %d\n", err);
2885 
2886 	return err;
2887 }
2888 
2889 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
2890 {
2891 	int vector;
2892 
2893 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
2894 		free_irq(adapter->pdev->irq, adapter);
2895 		return;
2896 	}
2897 
2898 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2899 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2900 		struct msix_entry *entry = &adapter->msix_entries[vector];
2901 
2902 		/* free only the irqs that were actually requested */
2903 		if (!q_vector->rx.ring && !q_vector->tx.ring)
2904 			continue;
2905 
2906 		/* clear the affinity_mask in the IRQ descriptor */
2907 		irq_set_affinity_hint(entry->vector, NULL);
2908 
2909 		free_irq(entry->vector, q_vector);
2910 	}
2911 
2912 	free_irq(adapter->msix_entries[vector++].vector, adapter);
2913 }
2914 
2915 /**
2916  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
2917  * @adapter: board private structure
2918  **/
2919 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
2920 {
2921 	switch (adapter->hw.mac.type) {
2922 	case ixgbe_mac_82598EB:
2923 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
2924 		break;
2925 	case ixgbe_mac_82599EB:
2926 	case ixgbe_mac_X540:
2927 	case ixgbe_mac_X550:
2928 	case ixgbe_mac_X550EM_x:
2929 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
2930 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
2931 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
2932 		break;
2933 	default:
2934 		break;
2935 	}
2936 	IXGBE_WRITE_FLUSH(&adapter->hw);
2937 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
2938 		int vector;
2939 
2940 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
2941 			synchronize_irq(adapter->msix_entries[vector].vector);
2942 
2943 		synchronize_irq(adapter->msix_entries[vector++].vector);
2944 	} else {
2945 		synchronize_irq(adapter->pdev->irq);
2946 	}
2947 }
2948 
2949 /**
2950  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
2951  *
2952  **/
2953 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
2954 {
2955 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2956 
2957 	ixgbe_write_eitr(q_vector);
2958 
2959 	ixgbe_set_ivar(adapter, 0, 0, 0);
2960 	ixgbe_set_ivar(adapter, 1, 0, 0);
2961 
2962 	e_info(hw, "Legacy interrupt IVAR setup done\n");
2963 }
2964 
2965 /**
2966  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
2967  * @adapter: board private structure
2968  * @ring: structure containing ring specific data
2969  *
2970  * Configure the Tx descriptor ring after a reset.
2971  **/
2972 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
2973 			     struct ixgbe_ring *ring)
2974 {
2975 	struct ixgbe_hw *hw = &adapter->hw;
2976 	u64 tdba = ring->dma;
2977 	int wait_loop = 10;
2978 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
2979 	u8 reg_idx = ring->reg_idx;
2980 
2981 	/* disable queue to avoid issues while updating state */
2982 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
2983 	IXGBE_WRITE_FLUSH(hw);
2984 
2985 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
2986 			(tdba & DMA_BIT_MASK(32)));
2987 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
2988 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
2989 			ring->count * sizeof(union ixgbe_adv_tx_desc));
2990 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
2991 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
2992 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
2993 
2994 	/*
2995 	 * set WTHRESH to encourage burst writeback, it should not be set
2996 	 * higher than 1 when:
2997 	 * - ITR is 0 as it could cause false TX hangs
2998 	 * - ITR is set to > 100k int/sec and BQL is enabled
2999 	 *
3000 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3001 	 * to or less than the number of on chip descriptors, which is
3002 	 * currently 40.
3003 	 */
3004 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3005 		txdctl |= (1 << 16);	/* WTHRESH = 1 */
3006 	else
3007 		txdctl |= (8 << 16);	/* WTHRESH = 8 */
3008 
3009 	/*
3010 	 * Setting PTHRESH to 32 both improves performance
3011 	 * and avoids a TX hang with DFP enabled
3012 	 */
3013 	txdctl |= (1 << 8) |	/* HTHRESH = 1 */
3014 		   32;		/* PTHRESH = 32 */
3015 
3016 	/* reinitialize flowdirector state */
3017 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3018 		ring->atr_sample_rate = adapter->atr_sample_rate;
3019 		ring->atr_count = 0;
3020 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3021 	} else {
3022 		ring->atr_sample_rate = 0;
3023 	}
3024 
3025 	/* initialize XPS */
3026 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3027 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3028 
3029 		if (q_vector)
3030 			netif_set_xps_queue(ring->netdev,
3031 					    &q_vector->affinity_mask,
3032 					    ring->queue_index);
3033 	}
3034 
3035 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3036 
3037 	/* enable queue */
3038 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3039 
3040 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3041 	if (hw->mac.type == ixgbe_mac_82598EB &&
3042 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3043 		return;
3044 
3045 	/* poll to verify queue is enabled */
3046 	do {
3047 		usleep_range(1000, 2000);
3048 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3049 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3050 	if (!wait_loop)
3051 		e_err(drv, "Could not enable Tx Queue %d\n", reg_idx);
3052 }
3053 
3054 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3055 {
3056 	struct ixgbe_hw *hw = &adapter->hw;
3057 	u32 rttdcs, mtqc;
3058 	u8 tcs = netdev_get_num_tc(adapter->netdev);
3059 
3060 	if (hw->mac.type == ixgbe_mac_82598EB)
3061 		return;
3062 
3063 	/* disable the arbiter while setting MTQC */
3064 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3065 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3066 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3067 
3068 	/* set transmit pool layout */
3069 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3070 		mtqc = IXGBE_MTQC_VT_ENA;
3071 		if (tcs > 4)
3072 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3073 		else if (tcs > 1)
3074 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3075 		else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3076 			mtqc |= IXGBE_MTQC_32VF;
3077 		else
3078 			mtqc |= IXGBE_MTQC_64VF;
3079 	} else {
3080 		if (tcs > 4)
3081 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3082 		else if (tcs > 1)
3083 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3084 		else
3085 			mtqc = IXGBE_MTQC_64Q_1PB;
3086 	}
3087 
3088 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3089 
3090 	/* Enable Security TX Buffer IFG for multiple pb */
3091 	if (tcs) {
3092 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3093 		sectx |= IXGBE_SECTX_DCB;
3094 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3095 	}
3096 
3097 	/* re-enable the arbiter */
3098 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3099 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3100 }
3101 
3102 /**
3103  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3104  * @adapter: board private structure
3105  *
3106  * Configure the Tx unit of the MAC after a reset.
3107  **/
3108 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3109 {
3110 	struct ixgbe_hw *hw = &adapter->hw;
3111 	u32 dmatxctl;
3112 	u32 i;
3113 
3114 	ixgbe_setup_mtqc(adapter);
3115 
3116 	if (hw->mac.type != ixgbe_mac_82598EB) {
3117 		/* DMATXCTL.EN must be before Tx queues are enabled */
3118 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3119 		dmatxctl |= IXGBE_DMATXCTL_TE;
3120 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3121 	}
3122 
3123 	/* Setup the HW Tx Head and Tail descriptor pointers */
3124 	for (i = 0; i < adapter->num_tx_queues; i++)
3125 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3126 }
3127 
3128 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3129 				 struct ixgbe_ring *ring)
3130 {
3131 	struct ixgbe_hw *hw = &adapter->hw;
3132 	u8 reg_idx = ring->reg_idx;
3133 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3134 
3135 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3136 
3137 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3138 }
3139 
3140 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3141 				  struct ixgbe_ring *ring)
3142 {
3143 	struct ixgbe_hw *hw = &adapter->hw;
3144 	u8 reg_idx = ring->reg_idx;
3145 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3146 
3147 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3148 
3149 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3150 }
3151 
3152 #ifdef CONFIG_IXGBE_DCB
3153 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3154 #else
3155 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3156 #endif
3157 {
3158 	int i;
3159 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3160 
3161 	if (adapter->ixgbe_ieee_pfc)
3162 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3163 
3164 	/*
3165 	 * We should set the drop enable bit if:
3166 	 *  SR-IOV is enabled
3167 	 *   or
3168 	 *  Number of Rx queues > 1 and flow control is disabled
3169 	 *
3170 	 *  This allows us to avoid head of line blocking for security
3171 	 *  and performance reasons.
3172 	 */
3173 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3174 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3175 		for (i = 0; i < adapter->num_rx_queues; i++)
3176 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3177 	} else {
3178 		for (i = 0; i < adapter->num_rx_queues; i++)
3179 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3180 	}
3181 }
3182 
3183 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3184 
3185 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3186 				   struct ixgbe_ring *rx_ring)
3187 {
3188 	struct ixgbe_hw *hw = &adapter->hw;
3189 	u32 srrctl;
3190 	u8 reg_idx = rx_ring->reg_idx;
3191 
3192 	if (hw->mac.type == ixgbe_mac_82598EB) {
3193 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3194 
3195 		/*
3196 		 * if VMDq is not active we must program one srrctl register
3197 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3198 		 */
3199 		reg_idx &= mask;
3200 	}
3201 
3202 	/* configure header buffer length, needed for RSC */
3203 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3204 
3205 	/* configure the packet buffer length */
3206 	srrctl |= ixgbe_rx_bufsz(rx_ring) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3207 
3208 	/* configure descriptor type */
3209 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3210 
3211 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3212 }
3213 
3214 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter, const u32 *seed)
3215 {
3216 	struct ixgbe_hw *hw = &adapter->hw;
3217 	u32 reta = 0;
3218 	int i, j;
3219 	int reta_entries = 128;
3220 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3221 	int indices_multi;
3222 
3223 	/*
3224 	 * Program table for at least 2 queues w/ SR-IOV so that VFs can
3225 	 * make full use of any rings they may have.  We will use the
3226 	 * PSRTYPE register to control how many rings we use within the PF.
3227 	 */
3228 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 2))
3229 		rss_i = 2;
3230 
3231 	/* Fill out hash function seeds */
3232 	for (i = 0; i < 10; i++)
3233 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), seed[i]);
3234 
3235 	/* Fill out the redirection table as follows:
3236 	 * 82598: 128 (8 bit wide) entries containing pair of 4 bit RSS indices
3237 	 * 82599/X540: 128 (8 bit wide) entries containing 4 bit RSS index
3238 	 * X550: 512 (8 bit wide) entries containing 6 bit RSS index
3239 	 */
3240 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3241 		indices_multi = 0x11;
3242 	else
3243 		indices_multi = 0x1;
3244 
3245 	switch (adapter->hw.mac.type) {
3246 	case ixgbe_mac_X550:
3247 	case ixgbe_mac_X550EM_x:
3248 		if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3249 			reta_entries = 512;
3250 	default:
3251 		break;
3252 	}
3253 
3254 	/* Fill out redirection table */
3255 	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3256 		if (j == rss_i)
3257 			j = 0;
3258 		reta = (reta << 8) | (j * indices_multi);
3259 		if ((i & 3) == 3) {
3260 			if (i < 128)
3261 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3262 			else
3263 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3264 						reta);
3265 		}
3266 	}
3267 }
3268 
3269 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter, const u32 *seed)
3270 {
3271 	struct ixgbe_hw *hw = &adapter->hw;
3272 	u32 vfreta = 0;
3273 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3274 	unsigned int pf_pool = adapter->num_vfs;
3275 	int i, j;
3276 
3277 	/* Fill out hash function seeds */
3278 	for (i = 0; i < 10; i++)
3279 		IXGBE_WRITE_REG(hw, IXGBE_PFVFRSSRK(i, pf_pool), seed[i]);
3280 
3281 	/* Fill out the redirection table */
3282 	for (i = 0, j = 0; i < 64; i++, j++) {
3283 		if (j == rss_i)
3284 			j = 0;
3285 		vfreta = (vfreta << 8) | j;
3286 		if ((i & 3) == 3)
3287 			IXGBE_WRITE_REG(hw, IXGBE_PFVFRETA(i >> 2, pf_pool),
3288 					vfreta);
3289 	}
3290 }
3291 
3292 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3293 {
3294 	struct ixgbe_hw *hw = &adapter->hw;
3295 	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3296 	u32 rss_key[10];
3297 	u32 rxcsum;
3298 
3299 	/* Disable indicating checksum in descriptor, enables RSS hash */
3300 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3301 	rxcsum |= IXGBE_RXCSUM_PCSD;
3302 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3303 
3304 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3305 		if (adapter->ring_feature[RING_F_RSS].mask)
3306 			mrqc = IXGBE_MRQC_RSSEN;
3307 	} else {
3308 		u8 tcs = netdev_get_num_tc(adapter->netdev);
3309 
3310 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3311 			if (tcs > 4)
3312 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3313 			else if (tcs > 1)
3314 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3315 			else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3316 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3317 			else
3318 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3319 		} else {
3320 			if (tcs > 4)
3321 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3322 			else if (tcs > 1)
3323 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3324 			else
3325 				mrqc = IXGBE_MRQC_RSSEN;
3326 		}
3327 	}
3328 
3329 	/* Perform hash on these packet types */
3330 	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3331 		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3332 		     IXGBE_MRQC_RSS_FIELD_IPV6 |
3333 		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3334 
3335 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3336 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3337 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3338 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3339 
3340 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3341 	if ((hw->mac.type >= ixgbe_mac_X550) &&
3342 	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
3343 		unsigned int pf_pool = adapter->num_vfs;
3344 
3345 		/* Enable VF RSS mode */
3346 		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
3347 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3348 
3349 		/* Setup RSS through the VF registers */
3350 		ixgbe_setup_vfreta(adapter, rss_key);
3351 		vfmrqc = IXGBE_MRQC_RSSEN;
3352 		vfmrqc |= rss_field;
3353 		IXGBE_WRITE_REG(hw, IXGBE_PFVFMRQC(pf_pool), vfmrqc);
3354 	} else {
3355 		ixgbe_setup_reta(adapter, rss_key);
3356 		mrqc |= rss_field;
3357 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3358 	}
3359 }
3360 
3361 /**
3362  * ixgbe_configure_rscctl - enable RSC for the indicated ring
3363  * @adapter:    address of board private structure
3364  * @index:      index of ring to set
3365  **/
3366 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
3367 				   struct ixgbe_ring *ring)
3368 {
3369 	struct ixgbe_hw *hw = &adapter->hw;
3370 	u32 rscctrl;
3371 	u8 reg_idx = ring->reg_idx;
3372 
3373 	if (!ring_is_rsc_enabled(ring))
3374 		return;
3375 
3376 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
3377 	rscctrl |= IXGBE_RSCCTL_RSCEN;
3378 	/*
3379 	 * we must limit the number of descriptors so that the
3380 	 * total size of max desc * buf_len is not greater
3381 	 * than 65536
3382 	 */
3383 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
3384 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
3385 }
3386 
3387 #define IXGBE_MAX_RX_DESC_POLL 10
3388 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
3389 				       struct ixgbe_ring *ring)
3390 {
3391 	struct ixgbe_hw *hw = &adapter->hw;
3392 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3393 	u32 rxdctl;
3394 	u8 reg_idx = ring->reg_idx;
3395 
3396 	if (ixgbe_removed(hw->hw_addr))
3397 		return;
3398 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3399 	if (hw->mac.type == ixgbe_mac_82598EB &&
3400 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3401 		return;
3402 
3403 	do {
3404 		usleep_range(1000, 2000);
3405 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3406 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
3407 
3408 	if (!wait_loop) {
3409 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
3410 		      "the polling period\n", reg_idx);
3411 	}
3412 }
3413 
3414 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
3415 			    struct ixgbe_ring *ring)
3416 {
3417 	struct ixgbe_hw *hw = &adapter->hw;
3418 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3419 	u32 rxdctl;
3420 	u8 reg_idx = ring->reg_idx;
3421 
3422 	if (ixgbe_removed(hw->hw_addr))
3423 		return;
3424 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3425 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
3426 
3427 	/* write value back with RXDCTL.ENABLE bit cleared */
3428 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3429 
3430 	if (hw->mac.type == ixgbe_mac_82598EB &&
3431 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3432 		return;
3433 
3434 	/* the hardware may take up to 100us to really disable the rx queue */
3435 	do {
3436 		udelay(10);
3437 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3438 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
3439 
3440 	if (!wait_loop) {
3441 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3442 		      "the polling period\n", reg_idx);
3443 	}
3444 }
3445 
3446 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
3447 			     struct ixgbe_ring *ring)
3448 {
3449 	struct ixgbe_hw *hw = &adapter->hw;
3450 	u64 rdba = ring->dma;
3451 	u32 rxdctl;
3452 	u8 reg_idx = ring->reg_idx;
3453 
3454 	/* disable queue to avoid issues while updating state */
3455 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3456 	ixgbe_disable_rx_queue(adapter, ring);
3457 
3458 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
3459 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
3460 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
3461 			ring->count * sizeof(union ixgbe_adv_rx_desc));
3462 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
3463 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
3464 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
3465 
3466 	ixgbe_configure_srrctl(adapter, ring);
3467 	ixgbe_configure_rscctl(adapter, ring);
3468 
3469 	if (hw->mac.type == ixgbe_mac_82598EB) {
3470 		/*
3471 		 * enable cache line friendly hardware writes:
3472 		 * PTHRESH=32 descriptors (half the internal cache),
3473 		 * this also removes ugly rx_no_buffer_count increment
3474 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
3475 		 * WTHRESH=8 burst writeback up to two cache lines
3476 		 */
3477 		rxdctl &= ~0x3FFFFF;
3478 		rxdctl |=  0x080420;
3479 	}
3480 
3481 	/* enable receive descriptor ring */
3482 	rxdctl |= IXGBE_RXDCTL_ENABLE;
3483 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3484 
3485 	ixgbe_rx_desc_queue_enable(adapter, ring);
3486 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
3487 }
3488 
3489 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
3490 {
3491 	struct ixgbe_hw *hw = &adapter->hw;
3492 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
3493 	u16 pool;
3494 
3495 	/* PSRTYPE must be initialized in non 82598 adapters */
3496 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
3497 		      IXGBE_PSRTYPE_UDPHDR |
3498 		      IXGBE_PSRTYPE_IPV4HDR |
3499 		      IXGBE_PSRTYPE_L2HDR |
3500 		      IXGBE_PSRTYPE_IPV6HDR;
3501 
3502 	if (hw->mac.type == ixgbe_mac_82598EB)
3503 		return;
3504 
3505 	if (rss_i > 3)
3506 		psrtype |= 2 << 29;
3507 	else if (rss_i > 1)
3508 		psrtype |= 1 << 29;
3509 
3510 	for_each_set_bit(pool, &adapter->fwd_bitmask, 32)
3511 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
3512 }
3513 
3514 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
3515 {
3516 	struct ixgbe_hw *hw = &adapter->hw;
3517 	u32 reg_offset, vf_shift;
3518 	u32 gcr_ext, vmdctl;
3519 	int i;
3520 
3521 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3522 		return;
3523 
3524 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3525 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
3526 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
3527 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
3528 	vmdctl |= IXGBE_VT_CTL_REPLEN;
3529 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
3530 
3531 	vf_shift = VMDQ_P(0) % 32;
3532 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
3533 
3534 	/* Enable only the PF's pool for Tx/Rx */
3535 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), (~0) << vf_shift);
3536 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
3537 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), (~0) << vf_shift);
3538 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
3539 	if (adapter->flags2 & IXGBE_FLAG2_BRIDGE_MODE_VEB)
3540 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3541 
3542 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3543 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
3544 
3545 	/*
3546 	 * Set up VF register offsets for selected VT Mode,
3547 	 * i.e. 32 or 64 VFs for SR-IOV
3548 	 */
3549 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
3550 	case IXGBE_82599_VMDQ_8Q_MASK:
3551 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
3552 		break;
3553 	case IXGBE_82599_VMDQ_4Q_MASK:
3554 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
3555 		break;
3556 	default:
3557 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
3558 		break;
3559 	}
3560 
3561 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
3562 
3563 
3564 	/* Enable MAC Anti-Spoofing */
3565 	hw->mac.ops.set_mac_anti_spoofing(hw, (adapter->num_vfs != 0),
3566 					  adapter->num_vfs);
3567 	/* For VFs that have spoof checking turned off */
3568 	for (i = 0; i < adapter->num_vfs; i++) {
3569 		if (!adapter->vfinfo[i].spoofchk_enabled)
3570 			ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i, false);
3571 	}
3572 }
3573 
3574 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
3575 {
3576 	struct ixgbe_hw *hw = &adapter->hw;
3577 	struct net_device *netdev = adapter->netdev;
3578 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3579 	struct ixgbe_ring *rx_ring;
3580 	int i;
3581 	u32 mhadd, hlreg0;
3582 
3583 #ifdef IXGBE_FCOE
3584 	/* adjust max frame to be able to do baby jumbo for FCoE */
3585 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
3586 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
3587 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3588 
3589 #endif /* IXGBE_FCOE */
3590 
3591 	/* adjust max frame to be at least the size of a standard frame */
3592 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
3593 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
3594 
3595 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3596 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
3597 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3598 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
3599 
3600 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3601 	}
3602 
3603 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3604 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3605 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3606 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3607 
3608 	/*
3609 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3610 	 * the Base and Length of the Rx Descriptor Ring
3611 	 */
3612 	for (i = 0; i < adapter->num_rx_queues; i++) {
3613 		rx_ring = adapter->rx_ring[i];
3614 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
3615 			set_ring_rsc_enabled(rx_ring);
3616 		else
3617 			clear_ring_rsc_enabled(rx_ring);
3618 	}
3619 }
3620 
3621 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
3622 {
3623 	struct ixgbe_hw *hw = &adapter->hw;
3624 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
3625 
3626 	switch (hw->mac.type) {
3627 	case ixgbe_mac_X550:
3628 	case ixgbe_mac_X550EM_x:
3629 	case ixgbe_mac_82598EB:
3630 		/*
3631 		 * For VMDq support of different descriptor types or
3632 		 * buffer sizes through the use of multiple SRRCTL
3633 		 * registers, RDRXCTL.MVMEN must be set to 1
3634 		 *
3635 		 * also, the manual doesn't mention it clearly but DCA hints
3636 		 * will only use queue 0's tags unless this bit is set.  Side
3637 		 * effects of setting this bit are only that SRRCTL must be
3638 		 * fully programmed [0..15]
3639 		 */
3640 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
3641 		break;
3642 	case ixgbe_mac_82599EB:
3643 	case ixgbe_mac_X540:
3644 		/* Disable RSC for ACK packets */
3645 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
3646 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
3647 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
3648 		/* hardware requires some bits to be set by default */
3649 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
3650 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
3651 		break;
3652 	default:
3653 		/* We should do nothing since we don't know this hardware */
3654 		return;
3655 	}
3656 
3657 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
3658 }
3659 
3660 /**
3661  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3662  * @adapter: board private structure
3663  *
3664  * Configure the Rx unit of the MAC after a reset.
3665  **/
3666 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
3667 {
3668 	struct ixgbe_hw *hw = &adapter->hw;
3669 	int i;
3670 	u32 rxctrl, rfctl;
3671 
3672 	/* disable receives while setting up the descriptors */
3673 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3674 	IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxctrl & ~IXGBE_RXCTRL_RXEN);
3675 
3676 	ixgbe_setup_psrtype(adapter);
3677 	ixgbe_setup_rdrxctl(adapter);
3678 
3679 	/* RSC Setup */
3680 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
3681 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
3682 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
3683 		rfctl |= IXGBE_RFCTL_RSC_DIS;
3684 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
3685 
3686 	/* Program registers for the distribution of queues */
3687 	ixgbe_setup_mrqc(adapter);
3688 
3689 	/* set_rx_buffer_len must be called before ring initialization */
3690 	ixgbe_set_rx_buffer_len(adapter);
3691 
3692 	/*
3693 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3694 	 * the Base and Length of the Rx Descriptor Ring
3695 	 */
3696 	for (i = 0; i < adapter->num_rx_queues; i++)
3697 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
3698 
3699 	/* disable drop enable for 82598 parts */
3700 	if (hw->mac.type == ixgbe_mac_82598EB)
3701 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3702 
3703 	/* enable all receives */
3704 	rxctrl |= IXGBE_RXCTRL_RXEN;
3705 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
3706 }
3707 
3708 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
3709 				 __be16 proto, u16 vid)
3710 {
3711 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3712 	struct ixgbe_hw *hw = &adapter->hw;
3713 
3714 	/* add VID to filter table */
3715 	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true);
3716 	set_bit(vid, adapter->active_vlans);
3717 
3718 	return 0;
3719 }
3720 
3721 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
3722 				  __be16 proto, u16 vid)
3723 {
3724 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3725 	struct ixgbe_hw *hw = &adapter->hw;
3726 
3727 	/* remove VID from filter table */
3728 	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), false);
3729 	clear_bit(vid, adapter->active_vlans);
3730 
3731 	return 0;
3732 }
3733 
3734 /**
3735  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
3736  * @adapter: driver data
3737  */
3738 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
3739 {
3740 	struct ixgbe_hw *hw = &adapter->hw;
3741 	u32 vlnctrl;
3742 	int i, j;
3743 
3744 	switch (hw->mac.type) {
3745 	case ixgbe_mac_82598EB:
3746 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3747 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
3748 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3749 		break;
3750 	case ixgbe_mac_82599EB:
3751 	case ixgbe_mac_X540:
3752 	case ixgbe_mac_X550:
3753 	case ixgbe_mac_X550EM_x:
3754 		for (i = 0; i < adapter->num_rx_queues; i++) {
3755 			struct ixgbe_ring *ring = adapter->rx_ring[i];
3756 
3757 			if (ring->l2_accel_priv)
3758 				continue;
3759 			j = ring->reg_idx;
3760 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3761 			vlnctrl &= ~IXGBE_RXDCTL_VME;
3762 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3763 		}
3764 		break;
3765 	default:
3766 		break;
3767 	}
3768 }
3769 
3770 /**
3771  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
3772  * @adapter: driver data
3773  */
3774 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
3775 {
3776 	struct ixgbe_hw *hw = &adapter->hw;
3777 	u32 vlnctrl;
3778 	int i, j;
3779 
3780 	switch (hw->mac.type) {
3781 	case ixgbe_mac_82598EB:
3782 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3783 		vlnctrl |= IXGBE_VLNCTRL_VME;
3784 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3785 		break;
3786 	case ixgbe_mac_82599EB:
3787 	case ixgbe_mac_X540:
3788 	case ixgbe_mac_X550:
3789 	case ixgbe_mac_X550EM_x:
3790 		for (i = 0; i < adapter->num_rx_queues; i++) {
3791 			struct ixgbe_ring *ring = adapter->rx_ring[i];
3792 
3793 			if (ring->l2_accel_priv)
3794 				continue;
3795 			j = ring->reg_idx;
3796 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3797 			vlnctrl |= IXGBE_RXDCTL_VME;
3798 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3799 		}
3800 		break;
3801 	default:
3802 		break;
3803 	}
3804 }
3805 
3806 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
3807 {
3808 	u16 vid;
3809 
3810 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
3811 
3812 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
3813 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
3814 }
3815 
3816 /**
3817  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
3818  * @netdev: network interface device structure
3819  *
3820  * Writes multicast address list to the MTA hash table.
3821  * Returns: -ENOMEM on failure
3822  *                0 on no addresses written
3823  *                X on writing X addresses to MTA
3824  **/
3825 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
3826 {
3827 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3828 	struct ixgbe_hw *hw = &adapter->hw;
3829 
3830 	if (!netif_running(netdev))
3831 		return 0;
3832 
3833 	if (hw->mac.ops.update_mc_addr_list)
3834 		hw->mac.ops.update_mc_addr_list(hw, netdev);
3835 	else
3836 		return -ENOMEM;
3837 
3838 #ifdef CONFIG_PCI_IOV
3839 	ixgbe_restore_vf_multicasts(adapter);
3840 #endif
3841 
3842 	return netdev_mc_count(netdev);
3843 }
3844 
3845 #ifdef CONFIG_PCI_IOV
3846 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
3847 {
3848 	struct ixgbe_hw *hw = &adapter->hw;
3849 	int i;
3850 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3851 		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_IN_USE)
3852 			hw->mac.ops.set_rar(hw, i, adapter->mac_table[i].addr,
3853 					    adapter->mac_table[i].queue,
3854 					    IXGBE_RAH_AV);
3855 		else
3856 			hw->mac.ops.clear_rar(hw, i);
3857 
3858 		adapter->mac_table[i].state &= ~(IXGBE_MAC_STATE_MODIFIED);
3859 	}
3860 }
3861 #endif
3862 
3863 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
3864 {
3865 	struct ixgbe_hw *hw = &adapter->hw;
3866 	int i;
3867 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3868 		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_MODIFIED) {
3869 			if (adapter->mac_table[i].state &
3870 			    IXGBE_MAC_STATE_IN_USE)
3871 				hw->mac.ops.set_rar(hw, i,
3872 						adapter->mac_table[i].addr,
3873 						adapter->mac_table[i].queue,
3874 						IXGBE_RAH_AV);
3875 			else
3876 				hw->mac.ops.clear_rar(hw, i);
3877 
3878 			adapter->mac_table[i].state &=
3879 						~(IXGBE_MAC_STATE_MODIFIED);
3880 		}
3881 	}
3882 }
3883 
3884 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
3885 {
3886 	int i;
3887 	struct ixgbe_hw *hw = &adapter->hw;
3888 
3889 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3890 		adapter->mac_table[i].state |= IXGBE_MAC_STATE_MODIFIED;
3891 		adapter->mac_table[i].state &= ~IXGBE_MAC_STATE_IN_USE;
3892 		memset(adapter->mac_table[i].addr, 0, ETH_ALEN);
3893 		adapter->mac_table[i].queue = 0;
3894 	}
3895 	ixgbe_sync_mac_table(adapter);
3896 }
3897 
3898 static int ixgbe_available_rars(struct ixgbe_adapter *adapter)
3899 {
3900 	struct ixgbe_hw *hw = &adapter->hw;
3901 	int i, count = 0;
3902 
3903 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3904 		if (adapter->mac_table[i].state == 0)
3905 			count++;
3906 	}
3907 	return count;
3908 }
3909 
3910 /* this function destroys the first RAR entry */
3911 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter,
3912 					 u8 *addr)
3913 {
3914 	struct ixgbe_hw *hw = &adapter->hw;
3915 
3916 	memcpy(&adapter->mac_table[0].addr, addr, ETH_ALEN);
3917 	adapter->mac_table[0].queue = VMDQ_P(0);
3918 	adapter->mac_table[0].state = (IXGBE_MAC_STATE_DEFAULT |
3919 				       IXGBE_MAC_STATE_IN_USE);
3920 	hw->mac.ops.set_rar(hw, 0, adapter->mac_table[0].addr,
3921 			    adapter->mac_table[0].queue,
3922 			    IXGBE_RAH_AV);
3923 }
3924 
3925 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter, u8 *addr, u16 queue)
3926 {
3927 	struct ixgbe_hw *hw = &adapter->hw;
3928 	int i;
3929 
3930 	if (is_zero_ether_addr(addr))
3931 		return -EINVAL;
3932 
3933 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3934 		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_IN_USE)
3935 			continue;
3936 		adapter->mac_table[i].state |= (IXGBE_MAC_STATE_MODIFIED |
3937 						IXGBE_MAC_STATE_IN_USE);
3938 		ether_addr_copy(adapter->mac_table[i].addr, addr);
3939 		adapter->mac_table[i].queue = queue;
3940 		ixgbe_sync_mac_table(adapter);
3941 		return i;
3942 	}
3943 	return -ENOMEM;
3944 }
3945 
3946 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter, u8 *addr, u16 queue)
3947 {
3948 	/* search table for addr, if found, set to 0 and sync */
3949 	int i;
3950 	struct ixgbe_hw *hw = &adapter->hw;
3951 
3952 	if (is_zero_ether_addr(addr))
3953 		return -EINVAL;
3954 
3955 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3956 		if (ether_addr_equal(addr, adapter->mac_table[i].addr) &&
3957 		    adapter->mac_table[i].queue == queue) {
3958 			adapter->mac_table[i].state |= IXGBE_MAC_STATE_MODIFIED;
3959 			adapter->mac_table[i].state &= ~IXGBE_MAC_STATE_IN_USE;
3960 			memset(adapter->mac_table[i].addr, 0, ETH_ALEN);
3961 			adapter->mac_table[i].queue = 0;
3962 			ixgbe_sync_mac_table(adapter);
3963 			return 0;
3964 		}
3965 	}
3966 	return -ENOMEM;
3967 }
3968 /**
3969  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
3970  * @netdev: network interface device structure
3971  *
3972  * Writes unicast address list to the RAR table.
3973  * Returns: -ENOMEM on failure/insufficient address space
3974  *                0 on no addresses written
3975  *                X on writing X addresses to the RAR table
3976  **/
3977 static int ixgbe_write_uc_addr_list(struct net_device *netdev, int vfn)
3978 {
3979 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3980 	int count = 0;
3981 
3982 	/* return ENOMEM indicating insufficient memory for addresses */
3983 	if (netdev_uc_count(netdev) > ixgbe_available_rars(adapter))
3984 		return -ENOMEM;
3985 
3986 	if (!netdev_uc_empty(netdev)) {
3987 		struct netdev_hw_addr *ha;
3988 		netdev_for_each_uc_addr(ha, netdev) {
3989 			ixgbe_del_mac_filter(adapter, ha->addr, vfn);
3990 			ixgbe_add_mac_filter(adapter, ha->addr, vfn);
3991 			count++;
3992 		}
3993 	}
3994 	return count;
3995 }
3996 
3997 /**
3998  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
3999  * @netdev: network interface device structure
4000  *
4001  * The set_rx_method entry point is called whenever the unicast/multicast
4002  * address list or the network interface flags are updated.  This routine is
4003  * responsible for configuring the hardware for proper unicast, multicast and
4004  * promiscuous mode.
4005  **/
4006 void ixgbe_set_rx_mode(struct net_device *netdev)
4007 {
4008 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4009 	struct ixgbe_hw *hw = &adapter->hw;
4010 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4011 	u32 vlnctrl;
4012 	int count;
4013 
4014 	/* Check for Promiscuous and All Multicast modes */
4015 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4016 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4017 
4018 	/* set all bits that we expect to always be set */
4019 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4020 	fctrl |= IXGBE_FCTRL_BAM;
4021 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4022 	fctrl |= IXGBE_FCTRL_PMCF;
4023 
4024 	/* clear the bits we are changing the status of */
4025 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4026 	vlnctrl &= ~(IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
4027 	if (netdev->flags & IFF_PROMISC) {
4028 		hw->addr_ctrl.user_set_promisc = true;
4029 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4030 		vmolr |= IXGBE_VMOLR_MPE;
4031 		/* Only disable hardware filter vlans in promiscuous mode
4032 		 * if SR-IOV and VMDQ are disabled - otherwise ensure
4033 		 * that hardware VLAN filters remain enabled.
4034 		 */
4035 		if (adapter->flags & (IXGBE_FLAG_VMDQ_ENABLED |
4036 				      IXGBE_FLAG_SRIOV_ENABLED))
4037 			vlnctrl |= (IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
4038 	} else {
4039 		if (netdev->flags & IFF_ALLMULTI) {
4040 			fctrl |= IXGBE_FCTRL_MPE;
4041 			vmolr |= IXGBE_VMOLR_MPE;
4042 		}
4043 		vlnctrl |= IXGBE_VLNCTRL_VFE;
4044 		hw->addr_ctrl.user_set_promisc = false;
4045 	}
4046 
4047 	/*
4048 	 * Write addresses to available RAR registers, if there is not
4049 	 * sufficient space to store all the addresses then enable
4050 	 * unicast promiscuous mode
4051 	 */
4052 	count = ixgbe_write_uc_addr_list(netdev, VMDQ_P(0));
4053 	if (count < 0) {
4054 		fctrl |= IXGBE_FCTRL_UPE;
4055 		vmolr |= IXGBE_VMOLR_ROPE;
4056 	}
4057 
4058 	/* Write addresses to the MTA, if the attempt fails
4059 	 * then we should just turn on promiscuous mode so
4060 	 * that we can at least receive multicast traffic
4061 	 */
4062 	count = ixgbe_write_mc_addr_list(netdev);
4063 	if (count < 0) {
4064 		fctrl |= IXGBE_FCTRL_MPE;
4065 		vmolr |= IXGBE_VMOLR_MPE;
4066 	} else if (count) {
4067 		vmolr |= IXGBE_VMOLR_ROMPE;
4068 	}
4069 
4070 	if (hw->mac.type != ixgbe_mac_82598EB) {
4071 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4072 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4073 			   IXGBE_VMOLR_ROPE);
4074 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4075 	}
4076 
4077 	/* This is useful for sniffing bad packets. */
4078 	if (adapter->netdev->features & NETIF_F_RXALL) {
4079 		/* UPE and MPE will be handled by normal PROMISC logic
4080 		 * in e1000e_set_rx_mode */
4081 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4082 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4083 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4084 
4085 		fctrl &= ~(IXGBE_FCTRL_DPF);
4086 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
4087 	}
4088 
4089 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4090 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4091 
4092 	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
4093 		ixgbe_vlan_strip_enable(adapter);
4094 	else
4095 		ixgbe_vlan_strip_disable(adapter);
4096 }
4097 
4098 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4099 {
4100 	int q_idx;
4101 
4102 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4103 		ixgbe_qv_init_lock(adapter->q_vector[q_idx]);
4104 		napi_enable(&adapter->q_vector[q_idx]->napi);
4105 	}
4106 }
4107 
4108 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
4109 {
4110 	int q_idx;
4111 
4112 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4113 		napi_disable(&adapter->q_vector[q_idx]->napi);
4114 		while (!ixgbe_qv_disable(adapter->q_vector[q_idx])) {
4115 			pr_info("QV %d locked\n", q_idx);
4116 			usleep_range(1000, 20000);
4117 		}
4118 	}
4119 }
4120 
4121 #ifdef CONFIG_IXGBE_DCB
4122 /**
4123  * ixgbe_configure_dcb - Configure DCB hardware
4124  * @adapter: ixgbe adapter struct
4125  *
4126  * This is called by the driver on open to configure the DCB hardware.
4127  * This is also called by the gennetlink interface when reconfiguring
4128  * the DCB state.
4129  */
4130 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
4131 {
4132 	struct ixgbe_hw *hw = &adapter->hw;
4133 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4134 
4135 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
4136 		if (hw->mac.type == ixgbe_mac_82598EB)
4137 			netif_set_gso_max_size(adapter->netdev, 65536);
4138 		return;
4139 	}
4140 
4141 	if (hw->mac.type == ixgbe_mac_82598EB)
4142 		netif_set_gso_max_size(adapter->netdev, 32768);
4143 
4144 #ifdef IXGBE_FCOE
4145 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
4146 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
4147 #endif
4148 
4149 	/* reconfigure the hardware */
4150 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
4151 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4152 						DCB_TX_CONFIG);
4153 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4154 						DCB_RX_CONFIG);
4155 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
4156 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
4157 		ixgbe_dcb_hw_ets(&adapter->hw,
4158 				 adapter->ixgbe_ieee_ets,
4159 				 max_frame);
4160 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
4161 					adapter->ixgbe_ieee_pfc->pfc_en,
4162 					adapter->ixgbe_ieee_ets->prio_tc);
4163 	}
4164 
4165 	/* Enable RSS Hash per TC */
4166 	if (hw->mac.type != ixgbe_mac_82598EB) {
4167 		u32 msb = 0;
4168 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
4169 
4170 		while (rss_i) {
4171 			msb++;
4172 			rss_i >>= 1;
4173 		}
4174 
4175 		/* write msb to all 8 TCs in one write */
4176 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
4177 	}
4178 }
4179 #endif
4180 
4181 /* Additional bittime to account for IXGBE framing */
4182 #define IXGBE_ETH_FRAMING 20
4183 
4184 /**
4185  * ixgbe_hpbthresh - calculate high water mark for flow control
4186  *
4187  * @adapter: board private structure to calculate for
4188  * @pb: packet buffer to calculate
4189  */
4190 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
4191 {
4192 	struct ixgbe_hw *hw = &adapter->hw;
4193 	struct net_device *dev = adapter->netdev;
4194 	int link, tc, kb, marker;
4195 	u32 dv_id, rx_pba;
4196 
4197 	/* Calculate max LAN frame size */
4198 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
4199 
4200 #ifdef IXGBE_FCOE
4201 	/* FCoE traffic class uses FCOE jumbo frames */
4202 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4203 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4204 	    (pb == ixgbe_fcoe_get_tc(adapter)))
4205 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4206 #endif
4207 
4208 	/* Calculate delay value for device */
4209 	switch (hw->mac.type) {
4210 	case ixgbe_mac_X540:
4211 	case ixgbe_mac_X550:
4212 	case ixgbe_mac_X550EM_x:
4213 		dv_id = IXGBE_DV_X540(link, tc);
4214 		break;
4215 	default:
4216 		dv_id = IXGBE_DV(link, tc);
4217 		break;
4218 	}
4219 
4220 	/* Loopback switch introduces additional latency */
4221 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4222 		dv_id += IXGBE_B2BT(tc);
4223 
4224 	/* Delay value is calculated in bit times convert to KB */
4225 	kb = IXGBE_BT2KB(dv_id);
4226 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
4227 
4228 	marker = rx_pba - kb;
4229 
4230 	/* It is possible that the packet buffer is not large enough
4231 	 * to provide required headroom. In this case throw an error
4232 	 * to user and a do the best we can.
4233 	 */
4234 	if (marker < 0) {
4235 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
4236 			    "headroom to support flow control."
4237 			    "Decrease MTU or number of traffic classes\n", pb);
4238 		marker = tc + 1;
4239 	}
4240 
4241 	return marker;
4242 }
4243 
4244 /**
4245  * ixgbe_lpbthresh - calculate low water mark for for flow control
4246  *
4247  * @adapter: board private structure to calculate for
4248  * @pb: packet buffer to calculate
4249  */
4250 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
4251 {
4252 	struct ixgbe_hw *hw = &adapter->hw;
4253 	struct net_device *dev = adapter->netdev;
4254 	int tc;
4255 	u32 dv_id;
4256 
4257 	/* Calculate max LAN frame size */
4258 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
4259 
4260 #ifdef IXGBE_FCOE
4261 	/* FCoE traffic class uses FCOE jumbo frames */
4262 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4263 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4264 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
4265 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4266 #endif
4267 
4268 	/* Calculate delay value for device */
4269 	switch (hw->mac.type) {
4270 	case ixgbe_mac_X540:
4271 	case ixgbe_mac_X550:
4272 	case ixgbe_mac_X550EM_x:
4273 		dv_id = IXGBE_LOW_DV_X540(tc);
4274 		break;
4275 	default:
4276 		dv_id = IXGBE_LOW_DV(tc);
4277 		break;
4278 	}
4279 
4280 	/* Delay value is calculated in bit times convert to KB */
4281 	return IXGBE_BT2KB(dv_id);
4282 }
4283 
4284 /*
4285  * ixgbe_pbthresh_setup - calculate and setup high low water marks
4286  */
4287 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
4288 {
4289 	struct ixgbe_hw *hw = &adapter->hw;
4290 	int num_tc = netdev_get_num_tc(adapter->netdev);
4291 	int i;
4292 
4293 	if (!num_tc)
4294 		num_tc = 1;
4295 
4296 	for (i = 0; i < num_tc; i++) {
4297 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
4298 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
4299 
4300 		/* Low water marks must not be larger than high water marks */
4301 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
4302 			hw->fc.low_water[i] = 0;
4303 	}
4304 
4305 	for (; i < MAX_TRAFFIC_CLASS; i++)
4306 		hw->fc.high_water[i] = 0;
4307 }
4308 
4309 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
4310 {
4311 	struct ixgbe_hw *hw = &adapter->hw;
4312 	int hdrm;
4313 	u8 tc = netdev_get_num_tc(adapter->netdev);
4314 
4315 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
4316 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
4317 		hdrm = 32 << adapter->fdir_pballoc;
4318 	else
4319 		hdrm = 0;
4320 
4321 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
4322 	ixgbe_pbthresh_setup(adapter);
4323 }
4324 
4325 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
4326 {
4327 	struct ixgbe_hw *hw = &adapter->hw;
4328 	struct hlist_node *node2;
4329 	struct ixgbe_fdir_filter *filter;
4330 
4331 	spin_lock(&adapter->fdir_perfect_lock);
4332 
4333 	if (!hlist_empty(&adapter->fdir_filter_list))
4334 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
4335 
4336 	hlist_for_each_entry_safe(filter, node2,
4337 				  &adapter->fdir_filter_list, fdir_node) {
4338 		ixgbe_fdir_write_perfect_filter_82599(hw,
4339 				&filter->filter,
4340 				filter->sw_idx,
4341 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
4342 				IXGBE_FDIR_DROP_QUEUE :
4343 				adapter->rx_ring[filter->action]->reg_idx);
4344 	}
4345 
4346 	spin_unlock(&adapter->fdir_perfect_lock);
4347 }
4348 
4349 static void ixgbe_macvlan_set_rx_mode(struct net_device *dev, unsigned int pool,
4350 				      struct ixgbe_adapter *adapter)
4351 {
4352 	struct ixgbe_hw *hw = &adapter->hw;
4353 	u32 vmolr;
4354 
4355 	/* No unicast promiscuous support for VMDQ devices. */
4356 	vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
4357 	vmolr |= (IXGBE_VMOLR_ROMPE | IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE);
4358 
4359 	/* clear the affected bit */
4360 	vmolr &= ~IXGBE_VMOLR_MPE;
4361 
4362 	if (dev->flags & IFF_ALLMULTI) {
4363 		vmolr |= IXGBE_VMOLR_MPE;
4364 	} else {
4365 		vmolr |= IXGBE_VMOLR_ROMPE;
4366 		hw->mac.ops.update_mc_addr_list(hw, dev);
4367 	}
4368 	ixgbe_write_uc_addr_list(adapter->netdev, pool);
4369 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
4370 }
4371 
4372 static void ixgbe_fwd_psrtype(struct ixgbe_fwd_adapter *vadapter)
4373 {
4374 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4375 	int rss_i = adapter->num_rx_queues_per_pool;
4376 	struct ixgbe_hw *hw = &adapter->hw;
4377 	u16 pool = vadapter->pool;
4378 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4379 		      IXGBE_PSRTYPE_UDPHDR |
4380 		      IXGBE_PSRTYPE_IPV4HDR |
4381 		      IXGBE_PSRTYPE_L2HDR |
4382 		      IXGBE_PSRTYPE_IPV6HDR;
4383 
4384 	if (hw->mac.type == ixgbe_mac_82598EB)
4385 		return;
4386 
4387 	if (rss_i > 3)
4388 		psrtype |= 2 << 29;
4389 	else if (rss_i > 1)
4390 		psrtype |= 1 << 29;
4391 
4392 	IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4393 }
4394 
4395 /**
4396  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4397  * @rx_ring: ring to free buffers from
4398  **/
4399 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
4400 {
4401 	struct device *dev = rx_ring->dev;
4402 	unsigned long size;
4403 	u16 i;
4404 
4405 	/* ring already cleared, nothing to do */
4406 	if (!rx_ring->rx_buffer_info)
4407 		return;
4408 
4409 	/* Free all the Rx ring sk_buffs */
4410 	for (i = 0; i < rx_ring->count; i++) {
4411 		struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
4412 
4413 		if (rx_buffer->skb) {
4414 			struct sk_buff *skb = rx_buffer->skb;
4415 			if (IXGBE_CB(skb)->page_released)
4416 				dma_unmap_page(dev,
4417 					       IXGBE_CB(skb)->dma,
4418 					       ixgbe_rx_bufsz(rx_ring),
4419 					       DMA_FROM_DEVICE);
4420 			dev_kfree_skb(skb);
4421 			rx_buffer->skb = NULL;
4422 		}
4423 
4424 		if (!rx_buffer->page)
4425 			continue;
4426 
4427 		dma_unmap_page(dev, rx_buffer->dma,
4428 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
4429 		__free_pages(rx_buffer->page, ixgbe_rx_pg_order(rx_ring));
4430 
4431 		rx_buffer->page = NULL;
4432 	}
4433 
4434 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4435 	memset(rx_ring->rx_buffer_info, 0, size);
4436 
4437 	/* Zero out the descriptor ring */
4438 	memset(rx_ring->desc, 0, rx_ring->size);
4439 
4440 	rx_ring->next_to_alloc = 0;
4441 	rx_ring->next_to_clean = 0;
4442 	rx_ring->next_to_use = 0;
4443 }
4444 
4445 static void ixgbe_disable_fwd_ring(struct ixgbe_fwd_adapter *vadapter,
4446 				   struct ixgbe_ring *rx_ring)
4447 {
4448 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4449 	int index = rx_ring->queue_index + vadapter->rx_base_queue;
4450 
4451 	/* shutdown specific queue receive and wait for dma to settle */
4452 	ixgbe_disable_rx_queue(adapter, rx_ring);
4453 	usleep_range(10000, 20000);
4454 	ixgbe_irq_disable_queues(adapter, ((u64)1 << index));
4455 	ixgbe_clean_rx_ring(rx_ring);
4456 	rx_ring->l2_accel_priv = NULL;
4457 }
4458 
4459 static int ixgbe_fwd_ring_down(struct net_device *vdev,
4460 			       struct ixgbe_fwd_adapter *accel)
4461 {
4462 	struct ixgbe_adapter *adapter = accel->real_adapter;
4463 	unsigned int rxbase = accel->rx_base_queue;
4464 	unsigned int txbase = accel->tx_base_queue;
4465 	int i;
4466 
4467 	netif_tx_stop_all_queues(vdev);
4468 
4469 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4470 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4471 		adapter->rx_ring[rxbase + i]->netdev = adapter->netdev;
4472 	}
4473 
4474 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4475 		adapter->tx_ring[txbase + i]->l2_accel_priv = NULL;
4476 		adapter->tx_ring[txbase + i]->netdev = adapter->netdev;
4477 	}
4478 
4479 
4480 	return 0;
4481 }
4482 
4483 static int ixgbe_fwd_ring_up(struct net_device *vdev,
4484 			     struct ixgbe_fwd_adapter *accel)
4485 {
4486 	struct ixgbe_adapter *adapter = accel->real_adapter;
4487 	unsigned int rxbase, txbase, queues;
4488 	int i, baseq, err = 0;
4489 
4490 	if (!test_bit(accel->pool, &adapter->fwd_bitmask))
4491 		return 0;
4492 
4493 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
4494 	netdev_dbg(vdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
4495 		   accel->pool, adapter->num_rx_pools,
4496 		   baseq, baseq + adapter->num_rx_queues_per_pool,
4497 		   adapter->fwd_bitmask);
4498 
4499 	accel->netdev = vdev;
4500 	accel->rx_base_queue = rxbase = baseq;
4501 	accel->tx_base_queue = txbase = baseq;
4502 
4503 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
4504 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4505 
4506 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4507 		adapter->rx_ring[rxbase + i]->netdev = vdev;
4508 		adapter->rx_ring[rxbase + i]->l2_accel_priv = accel;
4509 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[rxbase + i]);
4510 	}
4511 
4512 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4513 		adapter->tx_ring[txbase + i]->netdev = vdev;
4514 		adapter->tx_ring[txbase + i]->l2_accel_priv = accel;
4515 	}
4516 
4517 	queues = min_t(unsigned int,
4518 		       adapter->num_rx_queues_per_pool, vdev->num_tx_queues);
4519 	err = netif_set_real_num_tx_queues(vdev, queues);
4520 	if (err)
4521 		goto fwd_queue_err;
4522 
4523 	err = netif_set_real_num_rx_queues(vdev, queues);
4524 	if (err)
4525 		goto fwd_queue_err;
4526 
4527 	if (is_valid_ether_addr(vdev->dev_addr))
4528 		ixgbe_add_mac_filter(adapter, vdev->dev_addr, accel->pool);
4529 
4530 	ixgbe_fwd_psrtype(accel);
4531 	ixgbe_macvlan_set_rx_mode(vdev, accel->pool, adapter);
4532 	return err;
4533 fwd_queue_err:
4534 	ixgbe_fwd_ring_down(vdev, accel);
4535 	return err;
4536 }
4537 
4538 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
4539 {
4540 	struct net_device *upper;
4541 	struct list_head *iter;
4542 	int err;
4543 
4544 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
4545 		if (netif_is_macvlan(upper)) {
4546 			struct macvlan_dev *dfwd = netdev_priv(upper);
4547 			struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
4548 
4549 			if (dfwd->fwd_priv) {
4550 				err = ixgbe_fwd_ring_up(upper, vadapter);
4551 				if (err)
4552 					continue;
4553 			}
4554 		}
4555 	}
4556 }
4557 
4558 static void ixgbe_configure(struct ixgbe_adapter *adapter)
4559 {
4560 	struct ixgbe_hw *hw = &adapter->hw;
4561 
4562 	ixgbe_configure_pb(adapter);
4563 #ifdef CONFIG_IXGBE_DCB
4564 	ixgbe_configure_dcb(adapter);
4565 #endif
4566 	/*
4567 	 * We must restore virtualization before VLANs or else
4568 	 * the VLVF registers will not be populated
4569 	 */
4570 	ixgbe_configure_virtualization(adapter);
4571 
4572 	ixgbe_set_rx_mode(adapter->netdev);
4573 	ixgbe_restore_vlan(adapter);
4574 
4575 	switch (hw->mac.type) {
4576 	case ixgbe_mac_82599EB:
4577 	case ixgbe_mac_X540:
4578 		hw->mac.ops.disable_rx_buff(hw);
4579 		break;
4580 	default:
4581 		break;
4582 	}
4583 
4584 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
4585 		ixgbe_init_fdir_signature_82599(&adapter->hw,
4586 						adapter->fdir_pballoc);
4587 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
4588 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
4589 					      adapter->fdir_pballoc);
4590 		ixgbe_fdir_filter_restore(adapter);
4591 	}
4592 
4593 	switch (hw->mac.type) {
4594 	case ixgbe_mac_82599EB:
4595 	case ixgbe_mac_X540:
4596 		hw->mac.ops.enable_rx_buff(hw);
4597 		break;
4598 	default:
4599 		break;
4600 	}
4601 
4602 #ifdef IXGBE_FCOE
4603 	/* configure FCoE L2 filters, redirection table, and Rx control */
4604 	ixgbe_configure_fcoe(adapter);
4605 
4606 #endif /* IXGBE_FCOE */
4607 	ixgbe_configure_tx(adapter);
4608 	ixgbe_configure_rx(adapter);
4609 	ixgbe_configure_dfwd(adapter);
4610 }
4611 
4612 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
4613 {
4614 	switch (hw->phy.type) {
4615 	case ixgbe_phy_sfp_avago:
4616 	case ixgbe_phy_sfp_ftl:
4617 	case ixgbe_phy_sfp_intel:
4618 	case ixgbe_phy_sfp_unknown:
4619 	case ixgbe_phy_sfp_passive_tyco:
4620 	case ixgbe_phy_sfp_passive_unknown:
4621 	case ixgbe_phy_sfp_active_unknown:
4622 	case ixgbe_phy_sfp_ftl_active:
4623 	case ixgbe_phy_qsfp_passive_unknown:
4624 	case ixgbe_phy_qsfp_active_unknown:
4625 	case ixgbe_phy_qsfp_intel:
4626 	case ixgbe_phy_qsfp_unknown:
4627 	/* ixgbe_phy_none is set when no SFP module is present */
4628 	case ixgbe_phy_none:
4629 		return true;
4630 	case ixgbe_phy_nl:
4631 		if (hw->mac.type == ixgbe_mac_82598EB)
4632 			return true;
4633 	default:
4634 		return false;
4635 	}
4636 }
4637 
4638 /**
4639  * ixgbe_sfp_link_config - set up SFP+ link
4640  * @adapter: pointer to private adapter struct
4641  **/
4642 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
4643 {
4644 	/*
4645 	 * We are assuming the worst case scenario here, and that
4646 	 * is that an SFP was inserted/removed after the reset
4647 	 * but before SFP detection was enabled.  As such the best
4648 	 * solution is to just start searching as soon as we start
4649 	 */
4650 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
4651 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
4652 
4653 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
4654 }
4655 
4656 /**
4657  * ixgbe_non_sfp_link_config - set up non-SFP+ link
4658  * @hw: pointer to private hardware struct
4659  *
4660  * Returns 0 on success, negative on failure
4661  **/
4662 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
4663 {
4664 	u32 speed;
4665 	bool autoneg, link_up = false;
4666 	u32 ret = IXGBE_ERR_LINK_SETUP;
4667 
4668 	if (hw->mac.ops.check_link)
4669 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
4670 
4671 	if (ret)
4672 		return ret;
4673 
4674 	speed = hw->phy.autoneg_advertised;
4675 	if ((!speed) && (hw->mac.ops.get_link_capabilities))
4676 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
4677 							&autoneg);
4678 	if (ret)
4679 		return ret;
4680 
4681 	if (hw->mac.ops.setup_link)
4682 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
4683 
4684 	return ret;
4685 }
4686 
4687 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
4688 {
4689 	struct ixgbe_hw *hw = &adapter->hw;
4690 	u32 gpie = 0;
4691 
4692 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
4693 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
4694 		       IXGBE_GPIE_OCD;
4695 		gpie |= IXGBE_GPIE_EIAME;
4696 		/*
4697 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
4698 		 * this saves a register write for every interrupt
4699 		 */
4700 		switch (hw->mac.type) {
4701 		case ixgbe_mac_82598EB:
4702 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4703 			break;
4704 		case ixgbe_mac_82599EB:
4705 		case ixgbe_mac_X540:
4706 		case ixgbe_mac_X550:
4707 		case ixgbe_mac_X550EM_x:
4708 		default:
4709 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
4710 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
4711 			break;
4712 		}
4713 	} else {
4714 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
4715 		 * specifically only auto mask tx and rx interrupts */
4716 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4717 	}
4718 
4719 	/* XXX: to interrupt immediately for EICS writes, enable this */
4720 	/* gpie |= IXGBE_GPIE_EIMEN; */
4721 
4722 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
4723 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
4724 
4725 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4726 		case IXGBE_82599_VMDQ_8Q_MASK:
4727 			gpie |= IXGBE_GPIE_VTMODE_16;
4728 			break;
4729 		case IXGBE_82599_VMDQ_4Q_MASK:
4730 			gpie |= IXGBE_GPIE_VTMODE_32;
4731 			break;
4732 		default:
4733 			gpie |= IXGBE_GPIE_VTMODE_64;
4734 			break;
4735 		}
4736 	}
4737 
4738 	/* Enable Thermal over heat sensor interrupt */
4739 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
4740 		switch (adapter->hw.mac.type) {
4741 		case ixgbe_mac_82599EB:
4742 			gpie |= IXGBE_SDP0_GPIEN;
4743 			break;
4744 		case ixgbe_mac_X540:
4745 			gpie |= IXGBE_EIMS_TS;
4746 			break;
4747 		default:
4748 			break;
4749 		}
4750 	}
4751 
4752 	/* Enable fan failure interrupt */
4753 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
4754 		gpie |= IXGBE_SDP1_GPIEN;
4755 
4756 	if (hw->mac.type == ixgbe_mac_82599EB) {
4757 		gpie |= IXGBE_SDP1_GPIEN;
4758 		gpie |= IXGBE_SDP2_GPIEN;
4759 	}
4760 
4761 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
4762 }
4763 
4764 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
4765 {
4766 	struct ixgbe_hw *hw = &adapter->hw;
4767 	int err;
4768 	u32 ctrl_ext;
4769 
4770 	ixgbe_get_hw_control(adapter);
4771 	ixgbe_setup_gpie(adapter);
4772 
4773 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
4774 		ixgbe_configure_msix(adapter);
4775 	else
4776 		ixgbe_configure_msi_and_legacy(adapter);
4777 
4778 	/* enable the optics for 82599 SFP+ fiber */
4779 	if (hw->mac.ops.enable_tx_laser)
4780 		hw->mac.ops.enable_tx_laser(hw);
4781 
4782 	smp_mb__before_atomic();
4783 	clear_bit(__IXGBE_DOWN, &adapter->state);
4784 	ixgbe_napi_enable_all(adapter);
4785 
4786 	if (ixgbe_is_sfp(hw)) {
4787 		ixgbe_sfp_link_config(adapter);
4788 	} else {
4789 		err = ixgbe_non_sfp_link_config(hw);
4790 		if (err)
4791 			e_err(probe, "link_config FAILED %d\n", err);
4792 	}
4793 
4794 	/* clear any pending interrupts, may auto mask */
4795 	IXGBE_READ_REG(hw, IXGBE_EICR);
4796 	ixgbe_irq_enable(adapter, true, true);
4797 
4798 	/*
4799 	 * If this adapter has a fan, check to see if we had a failure
4800 	 * before we enabled the interrupt.
4801 	 */
4802 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
4803 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
4804 		if (esdp & IXGBE_ESDP_SDP1)
4805 			e_crit(drv, "Fan has stopped, replace the adapter\n");
4806 	}
4807 
4808 	/* bring the link up in the watchdog, this could race with our first
4809 	 * link up interrupt but shouldn't be a problem */
4810 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4811 	adapter->link_check_timeout = jiffies;
4812 	mod_timer(&adapter->service_timer, jiffies);
4813 
4814 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
4815 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
4816 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
4817 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
4818 }
4819 
4820 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
4821 {
4822 	WARN_ON(in_interrupt());
4823 	/* put off any impending NetWatchDogTimeout */
4824 	adapter->netdev->trans_start = jiffies;
4825 
4826 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
4827 		usleep_range(1000, 2000);
4828 	ixgbe_down(adapter);
4829 	/*
4830 	 * If SR-IOV enabled then wait a bit before bringing the adapter
4831 	 * back up to give the VFs time to respond to the reset.  The
4832 	 * two second wait is based upon the watchdog timer cycle in
4833 	 * the VF driver.
4834 	 */
4835 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4836 		msleep(2000);
4837 	ixgbe_up(adapter);
4838 	clear_bit(__IXGBE_RESETTING, &adapter->state);
4839 }
4840 
4841 void ixgbe_up(struct ixgbe_adapter *adapter)
4842 {
4843 	/* hardware has been reset, we need to reload some things */
4844 	ixgbe_configure(adapter);
4845 
4846 	ixgbe_up_complete(adapter);
4847 }
4848 
4849 void ixgbe_reset(struct ixgbe_adapter *adapter)
4850 {
4851 	struct ixgbe_hw *hw = &adapter->hw;
4852 	struct net_device *netdev = adapter->netdev;
4853 	int err;
4854 	u8 old_addr[ETH_ALEN];
4855 
4856 	if (ixgbe_removed(hw->hw_addr))
4857 		return;
4858 	/* lock SFP init bit to prevent race conditions with the watchdog */
4859 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
4860 		usleep_range(1000, 2000);
4861 
4862 	/* clear all SFP and link config related flags while holding SFP_INIT */
4863 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
4864 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
4865 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4866 
4867 	err = hw->mac.ops.init_hw(hw);
4868 	switch (err) {
4869 	case 0:
4870 	case IXGBE_ERR_SFP_NOT_PRESENT:
4871 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
4872 		break;
4873 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
4874 		e_dev_err("master disable timed out\n");
4875 		break;
4876 	case IXGBE_ERR_EEPROM_VERSION:
4877 		/* We are running on a pre-production device, log a warning */
4878 		e_dev_warn("This device is a pre-production adapter/LOM. "
4879 			   "Please be aware there may be issues associated with "
4880 			   "your hardware.  If you are experiencing problems "
4881 			   "please contact your Intel or hardware "
4882 			   "representative who provided you with this "
4883 			   "hardware.\n");
4884 		break;
4885 	default:
4886 		e_dev_err("Hardware Error: %d\n", err);
4887 	}
4888 
4889 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
4890 	/* do not flush user set addresses */
4891 	memcpy(old_addr, &adapter->mac_table[0].addr, netdev->addr_len);
4892 	ixgbe_flush_sw_mac_table(adapter);
4893 	ixgbe_mac_set_default_filter(adapter, old_addr);
4894 
4895 	/* update SAN MAC vmdq pool selection */
4896 	if (hw->mac.san_mac_rar_index)
4897 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
4898 
4899 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
4900 		ixgbe_ptp_reset(adapter);
4901 }
4902 
4903 /**
4904  * ixgbe_clean_tx_ring - Free Tx Buffers
4905  * @tx_ring: ring to be cleaned
4906  **/
4907 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
4908 {
4909 	struct ixgbe_tx_buffer *tx_buffer_info;
4910 	unsigned long size;
4911 	u16 i;
4912 
4913 	/* ring already cleared, nothing to do */
4914 	if (!tx_ring->tx_buffer_info)
4915 		return;
4916 
4917 	/* Free all the Tx ring sk_buffs */
4918 	for (i = 0; i < tx_ring->count; i++) {
4919 		tx_buffer_info = &tx_ring->tx_buffer_info[i];
4920 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer_info);
4921 	}
4922 
4923 	netdev_tx_reset_queue(txring_txq(tx_ring));
4924 
4925 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
4926 	memset(tx_ring->tx_buffer_info, 0, size);
4927 
4928 	/* Zero out the descriptor ring */
4929 	memset(tx_ring->desc, 0, tx_ring->size);
4930 
4931 	tx_ring->next_to_use = 0;
4932 	tx_ring->next_to_clean = 0;
4933 }
4934 
4935 /**
4936  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
4937  * @adapter: board private structure
4938  **/
4939 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
4940 {
4941 	int i;
4942 
4943 	for (i = 0; i < adapter->num_rx_queues; i++)
4944 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
4945 }
4946 
4947 /**
4948  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
4949  * @adapter: board private structure
4950  **/
4951 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
4952 {
4953 	int i;
4954 
4955 	for (i = 0; i < adapter->num_tx_queues; i++)
4956 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
4957 }
4958 
4959 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
4960 {
4961 	struct hlist_node *node2;
4962 	struct ixgbe_fdir_filter *filter;
4963 
4964 	spin_lock(&adapter->fdir_perfect_lock);
4965 
4966 	hlist_for_each_entry_safe(filter, node2,
4967 				  &adapter->fdir_filter_list, fdir_node) {
4968 		hlist_del(&filter->fdir_node);
4969 		kfree(filter);
4970 	}
4971 	adapter->fdir_filter_count = 0;
4972 
4973 	spin_unlock(&adapter->fdir_perfect_lock);
4974 }
4975 
4976 void ixgbe_down(struct ixgbe_adapter *adapter)
4977 {
4978 	struct net_device *netdev = adapter->netdev;
4979 	struct ixgbe_hw *hw = &adapter->hw;
4980 	struct net_device *upper;
4981 	struct list_head *iter;
4982 	u32 rxctrl;
4983 	int i;
4984 
4985 	/* signal that we are down to the interrupt handler */
4986 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
4987 		return; /* do nothing if already down */
4988 
4989 	/* disable receives */
4990 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4991 	IXGBE_WRITE_REG(hw, IXGBE_RXCTRL, rxctrl & ~IXGBE_RXCTRL_RXEN);
4992 
4993 	/* disable all enabled rx queues */
4994 	for (i = 0; i < adapter->num_rx_queues; i++)
4995 		/* this call also flushes the previous write */
4996 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
4997 
4998 	usleep_range(10000, 20000);
4999 
5000 	netif_tx_stop_all_queues(netdev);
5001 
5002 	/* call carrier off first to avoid false dev_watchdog timeouts */
5003 	netif_carrier_off(netdev);
5004 	netif_tx_disable(netdev);
5005 
5006 	/* disable any upper devices */
5007 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
5008 		if (netif_is_macvlan(upper)) {
5009 			struct macvlan_dev *vlan = netdev_priv(upper);
5010 
5011 			if (vlan->fwd_priv) {
5012 				netif_tx_stop_all_queues(upper);
5013 				netif_carrier_off(upper);
5014 				netif_tx_disable(upper);
5015 			}
5016 		}
5017 	}
5018 
5019 	ixgbe_irq_disable(adapter);
5020 
5021 	ixgbe_napi_disable_all(adapter);
5022 
5023 	adapter->flags2 &= ~(IXGBE_FLAG2_FDIR_REQUIRES_REINIT |
5024 			     IXGBE_FLAG2_RESET_REQUESTED);
5025 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5026 
5027 	del_timer_sync(&adapter->service_timer);
5028 
5029 	if (adapter->num_vfs) {
5030 		/* Clear EITR Select mapping */
5031 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
5032 
5033 		/* Mark all the VFs as inactive */
5034 		for (i = 0 ; i < adapter->num_vfs; i++)
5035 			adapter->vfinfo[i].clear_to_send = false;
5036 
5037 		/* ping all the active vfs to let them know we are going down */
5038 		ixgbe_ping_all_vfs(adapter);
5039 
5040 		/* Disable all VFTE/VFRE TX/RX */
5041 		ixgbe_disable_tx_rx(adapter);
5042 	}
5043 
5044 	/* disable transmits in the hardware now that interrupts are off */
5045 	for (i = 0; i < adapter->num_tx_queues; i++) {
5046 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
5047 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5048 	}
5049 
5050 	/* Disable the Tx DMA engine on 82599 and later MAC */
5051 	switch (hw->mac.type) {
5052 	case ixgbe_mac_82599EB:
5053 	case ixgbe_mac_X540:
5054 	case ixgbe_mac_X550:
5055 	case ixgbe_mac_X550EM_x:
5056 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5057 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5058 				 ~IXGBE_DMATXCTL_TE));
5059 		break;
5060 	default:
5061 		break;
5062 	}
5063 
5064 	if (!pci_channel_offline(adapter->pdev))
5065 		ixgbe_reset(adapter);
5066 
5067 	/* power down the optics for 82599 SFP+ fiber */
5068 	if (hw->mac.ops.disable_tx_laser)
5069 		hw->mac.ops.disable_tx_laser(hw);
5070 
5071 	ixgbe_clean_all_tx_rings(adapter);
5072 	ixgbe_clean_all_rx_rings(adapter);
5073 
5074 #ifdef CONFIG_IXGBE_DCA
5075 	/* since we reset the hardware DCA settings were cleared */
5076 	ixgbe_setup_dca(adapter);
5077 #endif
5078 }
5079 
5080 /**
5081  * ixgbe_tx_timeout - Respond to a Tx Hang
5082  * @netdev: network interface device structure
5083  **/
5084 static void ixgbe_tx_timeout(struct net_device *netdev)
5085 {
5086 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5087 
5088 	/* Do the reset outside of interrupt context */
5089 	ixgbe_tx_timeout_reset(adapter);
5090 }
5091 
5092 /**
5093  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
5094  * @adapter: board private structure to initialize
5095  *
5096  * ixgbe_sw_init initializes the Adapter private data structure.
5097  * Fields are initialized based on PCI device information and
5098  * OS network device settings (MTU size).
5099  **/
5100 static int ixgbe_sw_init(struct ixgbe_adapter *adapter)
5101 {
5102 	struct ixgbe_hw *hw = &adapter->hw;
5103 	struct pci_dev *pdev = adapter->pdev;
5104 	unsigned int rss, fdir;
5105 	u32 fwsm;
5106 #ifdef CONFIG_IXGBE_DCB
5107 	int j;
5108 	struct tc_configuration *tc;
5109 #endif
5110 
5111 	/* PCI config space info */
5112 
5113 	hw->vendor_id = pdev->vendor;
5114 	hw->device_id = pdev->device;
5115 	hw->revision_id = pdev->revision;
5116 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
5117 	hw->subsystem_device_id = pdev->subsystem_device;
5118 
5119 	/* Set common capability flags and settings */
5120 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
5121 	adapter->ring_feature[RING_F_RSS].limit = rss;
5122 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
5123 	adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
5124 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
5125 	adapter->atr_sample_rate = 20;
5126 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
5127 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
5128 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
5129 #ifdef CONFIG_IXGBE_DCA
5130 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
5131 #endif
5132 #ifdef IXGBE_FCOE
5133 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
5134 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5135 #ifdef CONFIG_IXGBE_DCB
5136 	/* Default traffic class to use for FCoE */
5137 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
5138 #endif /* CONFIG_IXGBE_DCB */
5139 #endif /* IXGBE_FCOE */
5140 
5141 	adapter->mac_table = kzalloc(sizeof(struct ixgbe_mac_addr) *
5142 				     hw->mac.num_rar_entries,
5143 				     GFP_ATOMIC);
5144 
5145 	/* Set MAC specific capability flags and exceptions */
5146 	switch (hw->mac.type) {
5147 	case ixgbe_mac_82598EB:
5148 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
5149 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
5150 
5151 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
5152 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
5153 
5154 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
5155 		adapter->ring_feature[RING_F_FDIR].limit = 0;
5156 		adapter->atr_sample_rate = 0;
5157 		adapter->fdir_pballoc = 0;
5158 #ifdef IXGBE_FCOE
5159 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
5160 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5161 #ifdef CONFIG_IXGBE_DCB
5162 		adapter->fcoe.up = 0;
5163 #endif /* IXGBE_DCB */
5164 #endif /* IXGBE_FCOE */
5165 		break;
5166 	case ixgbe_mac_82599EB:
5167 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
5168 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5169 		break;
5170 	case ixgbe_mac_X540:
5171 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM);
5172 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
5173 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5174 		break;
5175 	case ixgbe_mac_X550EM_x:
5176 	case ixgbe_mac_X550:
5177 #ifdef CONFIG_IXGBE_DCA
5178 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
5179 #endif
5180 		break;
5181 	default:
5182 		break;
5183 	}
5184 
5185 #ifdef IXGBE_FCOE
5186 	/* FCoE support exists, always init the FCoE lock */
5187 	spin_lock_init(&adapter->fcoe.lock);
5188 
5189 #endif
5190 	/* n-tuple support exists, always init our spinlock */
5191 	spin_lock_init(&adapter->fdir_perfect_lock);
5192 
5193 #ifdef CONFIG_IXGBE_DCB
5194 	switch (hw->mac.type) {
5195 	case ixgbe_mac_X540:
5196 	case ixgbe_mac_X550:
5197 	case ixgbe_mac_X550EM_x:
5198 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
5199 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
5200 		break;
5201 	default:
5202 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
5203 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
5204 		break;
5205 	}
5206 
5207 	/* Configure DCB traffic classes */
5208 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
5209 		tc = &adapter->dcb_cfg.tc_config[j];
5210 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
5211 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
5212 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
5213 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
5214 		tc->dcb_pfc = pfc_disabled;
5215 	}
5216 
5217 	/* Initialize default user to priority mapping, UPx->TC0 */
5218 	tc = &adapter->dcb_cfg.tc_config[0];
5219 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
5220 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
5221 
5222 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
5223 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
5224 	adapter->dcb_cfg.pfc_mode_enable = false;
5225 	adapter->dcb_set_bitmap = 0x00;
5226 	adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
5227 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
5228 	       sizeof(adapter->temp_dcb_cfg));
5229 
5230 #endif
5231 
5232 	/* default flow control settings */
5233 	hw->fc.requested_mode = ixgbe_fc_full;
5234 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
5235 	ixgbe_pbthresh_setup(adapter);
5236 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
5237 	hw->fc.send_xon = true;
5238 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
5239 
5240 #ifdef CONFIG_PCI_IOV
5241 	if (max_vfs > 0)
5242 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
5243 
5244 	/* assign number of SR-IOV VFs */
5245 	if (hw->mac.type != ixgbe_mac_82598EB) {
5246 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
5247 			adapter->num_vfs = 0;
5248 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
5249 		} else {
5250 			adapter->num_vfs = max_vfs;
5251 		}
5252 	}
5253 #endif /* CONFIG_PCI_IOV */
5254 
5255 	/* enable itr by default in dynamic mode */
5256 	adapter->rx_itr_setting = 1;
5257 	adapter->tx_itr_setting = 1;
5258 
5259 	/* set default ring sizes */
5260 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
5261 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
5262 
5263 	/* set default work limits */
5264 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
5265 
5266 	/* initialize eeprom parameters */
5267 	if (ixgbe_init_eeprom_params_generic(hw)) {
5268 		e_dev_err("EEPROM initialization failed\n");
5269 		return -EIO;
5270 	}
5271 
5272 	/* PF holds first pool slot */
5273 	set_bit(0, &adapter->fwd_bitmask);
5274 	set_bit(__IXGBE_DOWN, &adapter->state);
5275 
5276 	return 0;
5277 }
5278 
5279 /**
5280  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
5281  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
5282  *
5283  * Return 0 on success, negative on failure
5284  **/
5285 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
5286 {
5287 	struct device *dev = tx_ring->dev;
5288 	int orig_node = dev_to_node(dev);
5289 	int ring_node = -1;
5290 	int size;
5291 
5292 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5293 
5294 	if (tx_ring->q_vector)
5295 		ring_node = tx_ring->q_vector->numa_node;
5296 
5297 	tx_ring->tx_buffer_info = vzalloc_node(size, ring_node);
5298 	if (!tx_ring->tx_buffer_info)
5299 		tx_ring->tx_buffer_info = vzalloc(size);
5300 	if (!tx_ring->tx_buffer_info)
5301 		goto err;
5302 
5303 	u64_stats_init(&tx_ring->syncp);
5304 
5305 	/* round up to nearest 4K */
5306 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
5307 	tx_ring->size = ALIGN(tx_ring->size, 4096);
5308 
5309 	set_dev_node(dev, ring_node);
5310 	tx_ring->desc = dma_alloc_coherent(dev,
5311 					   tx_ring->size,
5312 					   &tx_ring->dma,
5313 					   GFP_KERNEL);
5314 	set_dev_node(dev, orig_node);
5315 	if (!tx_ring->desc)
5316 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
5317 						   &tx_ring->dma, GFP_KERNEL);
5318 	if (!tx_ring->desc)
5319 		goto err;
5320 
5321 	tx_ring->next_to_use = 0;
5322 	tx_ring->next_to_clean = 0;
5323 	return 0;
5324 
5325 err:
5326 	vfree(tx_ring->tx_buffer_info);
5327 	tx_ring->tx_buffer_info = NULL;
5328 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
5329 	return -ENOMEM;
5330 }
5331 
5332 /**
5333  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
5334  * @adapter: board private structure
5335  *
5336  * If this function returns with an error, then it's possible one or
5337  * more of the rings is populated (while the rest are not).  It is the
5338  * callers duty to clean those orphaned rings.
5339  *
5340  * Return 0 on success, negative on failure
5341  **/
5342 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
5343 {
5344 	int i, err = 0;
5345 
5346 	for (i = 0; i < adapter->num_tx_queues; i++) {
5347 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
5348 		if (!err)
5349 			continue;
5350 
5351 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
5352 		goto err_setup_tx;
5353 	}
5354 
5355 	return 0;
5356 err_setup_tx:
5357 	/* rewind the index freeing the rings as we go */
5358 	while (i--)
5359 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
5360 	return err;
5361 }
5362 
5363 /**
5364  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
5365  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
5366  *
5367  * Returns 0 on success, negative on failure
5368  **/
5369 int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
5370 {
5371 	struct device *dev = rx_ring->dev;
5372 	int orig_node = dev_to_node(dev);
5373 	int ring_node = -1;
5374 	int size;
5375 
5376 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
5377 
5378 	if (rx_ring->q_vector)
5379 		ring_node = rx_ring->q_vector->numa_node;
5380 
5381 	rx_ring->rx_buffer_info = vzalloc_node(size, ring_node);
5382 	if (!rx_ring->rx_buffer_info)
5383 		rx_ring->rx_buffer_info = vzalloc(size);
5384 	if (!rx_ring->rx_buffer_info)
5385 		goto err;
5386 
5387 	u64_stats_init(&rx_ring->syncp);
5388 
5389 	/* Round up to nearest 4K */
5390 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
5391 	rx_ring->size = ALIGN(rx_ring->size, 4096);
5392 
5393 	set_dev_node(dev, ring_node);
5394 	rx_ring->desc = dma_alloc_coherent(dev,
5395 					   rx_ring->size,
5396 					   &rx_ring->dma,
5397 					   GFP_KERNEL);
5398 	set_dev_node(dev, orig_node);
5399 	if (!rx_ring->desc)
5400 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
5401 						   &rx_ring->dma, GFP_KERNEL);
5402 	if (!rx_ring->desc)
5403 		goto err;
5404 
5405 	rx_ring->next_to_clean = 0;
5406 	rx_ring->next_to_use = 0;
5407 
5408 	return 0;
5409 err:
5410 	vfree(rx_ring->rx_buffer_info);
5411 	rx_ring->rx_buffer_info = NULL;
5412 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
5413 	return -ENOMEM;
5414 }
5415 
5416 /**
5417  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
5418  * @adapter: board private structure
5419  *
5420  * If this function returns with an error, then it's possible one or
5421  * more of the rings is populated (while the rest are not).  It is the
5422  * callers duty to clean those orphaned rings.
5423  *
5424  * Return 0 on success, negative on failure
5425  **/
5426 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
5427 {
5428 	int i, err = 0;
5429 
5430 	for (i = 0; i < adapter->num_rx_queues; i++) {
5431 		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
5432 		if (!err)
5433 			continue;
5434 
5435 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
5436 		goto err_setup_rx;
5437 	}
5438 
5439 #ifdef IXGBE_FCOE
5440 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
5441 	if (!err)
5442 #endif
5443 		return 0;
5444 err_setup_rx:
5445 	/* rewind the index freeing the rings as we go */
5446 	while (i--)
5447 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
5448 	return err;
5449 }
5450 
5451 /**
5452  * ixgbe_free_tx_resources - Free Tx Resources per Queue
5453  * @tx_ring: Tx descriptor ring for a specific queue
5454  *
5455  * Free all transmit software resources
5456  **/
5457 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
5458 {
5459 	ixgbe_clean_tx_ring(tx_ring);
5460 
5461 	vfree(tx_ring->tx_buffer_info);
5462 	tx_ring->tx_buffer_info = NULL;
5463 
5464 	/* if not set, then don't free */
5465 	if (!tx_ring->desc)
5466 		return;
5467 
5468 	dma_free_coherent(tx_ring->dev, tx_ring->size,
5469 			  tx_ring->desc, tx_ring->dma);
5470 
5471 	tx_ring->desc = NULL;
5472 }
5473 
5474 /**
5475  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
5476  * @adapter: board private structure
5477  *
5478  * Free all transmit software resources
5479  **/
5480 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
5481 {
5482 	int i;
5483 
5484 	for (i = 0; i < adapter->num_tx_queues; i++)
5485 		if (adapter->tx_ring[i]->desc)
5486 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
5487 }
5488 
5489 /**
5490  * ixgbe_free_rx_resources - Free Rx Resources
5491  * @rx_ring: ring to clean the resources from
5492  *
5493  * Free all receive software resources
5494  **/
5495 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
5496 {
5497 	ixgbe_clean_rx_ring(rx_ring);
5498 
5499 	vfree(rx_ring->rx_buffer_info);
5500 	rx_ring->rx_buffer_info = NULL;
5501 
5502 	/* if not set, then don't free */
5503 	if (!rx_ring->desc)
5504 		return;
5505 
5506 	dma_free_coherent(rx_ring->dev, rx_ring->size,
5507 			  rx_ring->desc, rx_ring->dma);
5508 
5509 	rx_ring->desc = NULL;
5510 }
5511 
5512 /**
5513  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
5514  * @adapter: board private structure
5515  *
5516  * Free all receive software resources
5517  **/
5518 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
5519 {
5520 	int i;
5521 
5522 #ifdef IXGBE_FCOE
5523 	ixgbe_free_fcoe_ddp_resources(adapter);
5524 
5525 #endif
5526 	for (i = 0; i < adapter->num_rx_queues; i++)
5527 		if (adapter->rx_ring[i]->desc)
5528 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
5529 }
5530 
5531 /**
5532  * ixgbe_change_mtu - Change the Maximum Transfer Unit
5533  * @netdev: network interface device structure
5534  * @new_mtu: new value for maximum frame size
5535  *
5536  * Returns 0 on success, negative on failure
5537  **/
5538 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
5539 {
5540 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5541 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
5542 
5543 	/* MTU < 68 is an error and causes problems on some kernels */
5544 	if ((new_mtu < 68) || (max_frame > IXGBE_MAX_JUMBO_FRAME_SIZE))
5545 		return -EINVAL;
5546 
5547 	/*
5548 	 * For 82599EB we cannot allow legacy VFs to enable their receive
5549 	 * paths when MTU greater than 1500 is configured.  So display a
5550 	 * warning that legacy VFs will be disabled.
5551 	 */
5552 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
5553 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
5554 	    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
5555 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
5556 
5557 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
5558 
5559 	/* must set new MTU before calling down or up */
5560 	netdev->mtu = new_mtu;
5561 
5562 	if (netif_running(netdev))
5563 		ixgbe_reinit_locked(adapter);
5564 
5565 	return 0;
5566 }
5567 
5568 /**
5569  * ixgbe_open - Called when a network interface is made active
5570  * @netdev: network interface device structure
5571  *
5572  * Returns 0 on success, negative value on failure
5573  *
5574  * The open entry point is called when a network interface is made
5575  * active by the system (IFF_UP).  At this point all resources needed
5576  * for transmit and receive operations are allocated, the interrupt
5577  * handler is registered with the OS, the watchdog timer is started,
5578  * and the stack is notified that the interface is ready.
5579  **/
5580 static int ixgbe_open(struct net_device *netdev)
5581 {
5582 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5583 	int err, queues;
5584 
5585 	/* disallow open during test */
5586 	if (test_bit(__IXGBE_TESTING, &adapter->state))
5587 		return -EBUSY;
5588 
5589 	netif_carrier_off(netdev);
5590 
5591 	/* allocate transmit descriptors */
5592 	err = ixgbe_setup_all_tx_resources(adapter);
5593 	if (err)
5594 		goto err_setup_tx;
5595 
5596 	/* allocate receive descriptors */
5597 	err = ixgbe_setup_all_rx_resources(adapter);
5598 	if (err)
5599 		goto err_setup_rx;
5600 
5601 	ixgbe_configure(adapter);
5602 
5603 	err = ixgbe_request_irq(adapter);
5604 	if (err)
5605 		goto err_req_irq;
5606 
5607 	/* Notify the stack of the actual queue counts. */
5608 	if (adapter->num_rx_pools > 1)
5609 		queues = adapter->num_rx_queues_per_pool;
5610 	else
5611 		queues = adapter->num_tx_queues;
5612 
5613 	err = netif_set_real_num_tx_queues(netdev, queues);
5614 	if (err)
5615 		goto err_set_queues;
5616 
5617 	if (adapter->num_rx_pools > 1 &&
5618 	    adapter->num_rx_queues > IXGBE_MAX_L2A_QUEUES)
5619 		queues = IXGBE_MAX_L2A_QUEUES;
5620 	else
5621 		queues = adapter->num_rx_queues;
5622 	err = netif_set_real_num_rx_queues(netdev, queues);
5623 	if (err)
5624 		goto err_set_queues;
5625 
5626 	ixgbe_ptp_init(adapter);
5627 
5628 	ixgbe_up_complete(adapter);
5629 
5630 	return 0;
5631 
5632 err_set_queues:
5633 	ixgbe_free_irq(adapter);
5634 err_req_irq:
5635 	ixgbe_free_all_rx_resources(adapter);
5636 err_setup_rx:
5637 	ixgbe_free_all_tx_resources(adapter);
5638 err_setup_tx:
5639 	ixgbe_reset(adapter);
5640 
5641 	return err;
5642 }
5643 
5644 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
5645 {
5646 	ixgbe_ptp_suspend(adapter);
5647 
5648 	ixgbe_down(adapter);
5649 	ixgbe_free_irq(adapter);
5650 
5651 	ixgbe_free_all_tx_resources(adapter);
5652 	ixgbe_free_all_rx_resources(adapter);
5653 }
5654 
5655 /**
5656  * ixgbe_close - Disables a network interface
5657  * @netdev: network interface device structure
5658  *
5659  * Returns 0, this is not allowed to fail
5660  *
5661  * The close entry point is called when an interface is de-activated
5662  * by the OS.  The hardware is still under the drivers control, but
5663  * needs to be disabled.  A global MAC reset is issued to stop the
5664  * hardware, and all transmit and receive resources are freed.
5665  **/
5666 static int ixgbe_close(struct net_device *netdev)
5667 {
5668 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5669 
5670 	ixgbe_ptp_stop(adapter);
5671 
5672 	ixgbe_close_suspend(adapter);
5673 
5674 	ixgbe_fdir_filter_exit(adapter);
5675 
5676 	ixgbe_release_hw_control(adapter);
5677 
5678 	return 0;
5679 }
5680 
5681 #ifdef CONFIG_PM
5682 static int ixgbe_resume(struct pci_dev *pdev)
5683 {
5684 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
5685 	struct net_device *netdev = adapter->netdev;
5686 	u32 err;
5687 
5688 	adapter->hw.hw_addr = adapter->io_addr;
5689 	pci_set_power_state(pdev, PCI_D0);
5690 	pci_restore_state(pdev);
5691 	/*
5692 	 * pci_restore_state clears dev->state_saved so call
5693 	 * pci_save_state to restore it.
5694 	 */
5695 	pci_save_state(pdev);
5696 
5697 	err = pci_enable_device_mem(pdev);
5698 	if (err) {
5699 		e_dev_err("Cannot enable PCI device from suspend\n");
5700 		return err;
5701 	}
5702 	smp_mb__before_atomic();
5703 	clear_bit(__IXGBE_DISABLED, &adapter->state);
5704 	pci_set_master(pdev);
5705 
5706 	pci_wake_from_d3(pdev, false);
5707 
5708 	ixgbe_reset(adapter);
5709 
5710 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
5711 
5712 	rtnl_lock();
5713 	err = ixgbe_init_interrupt_scheme(adapter);
5714 	if (!err && netif_running(netdev))
5715 		err = ixgbe_open(netdev);
5716 
5717 	rtnl_unlock();
5718 
5719 	if (err)
5720 		return err;
5721 
5722 	netif_device_attach(netdev);
5723 
5724 	return 0;
5725 }
5726 #endif /* CONFIG_PM */
5727 
5728 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
5729 {
5730 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
5731 	struct net_device *netdev = adapter->netdev;
5732 	struct ixgbe_hw *hw = &adapter->hw;
5733 	u32 ctrl, fctrl;
5734 	u32 wufc = adapter->wol;
5735 #ifdef CONFIG_PM
5736 	int retval = 0;
5737 #endif
5738 
5739 	netif_device_detach(netdev);
5740 
5741 	rtnl_lock();
5742 	if (netif_running(netdev))
5743 		ixgbe_close_suspend(adapter);
5744 	rtnl_unlock();
5745 
5746 	ixgbe_clear_interrupt_scheme(adapter);
5747 
5748 #ifdef CONFIG_PM
5749 	retval = pci_save_state(pdev);
5750 	if (retval)
5751 		return retval;
5752 
5753 #endif
5754 	if (hw->mac.ops.stop_link_on_d3)
5755 		hw->mac.ops.stop_link_on_d3(hw);
5756 
5757 	if (wufc) {
5758 		ixgbe_set_rx_mode(netdev);
5759 
5760 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
5761 		if (hw->mac.ops.enable_tx_laser)
5762 			hw->mac.ops.enable_tx_laser(hw);
5763 
5764 		/* turn on all-multi mode if wake on multicast is enabled */
5765 		if (wufc & IXGBE_WUFC_MC) {
5766 			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
5767 			fctrl |= IXGBE_FCTRL_MPE;
5768 			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
5769 		}
5770 
5771 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
5772 		ctrl |= IXGBE_CTRL_GIO_DIS;
5773 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
5774 
5775 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
5776 	} else {
5777 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
5778 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
5779 	}
5780 
5781 	switch (hw->mac.type) {
5782 	case ixgbe_mac_82598EB:
5783 		pci_wake_from_d3(pdev, false);
5784 		break;
5785 	case ixgbe_mac_82599EB:
5786 	case ixgbe_mac_X540:
5787 	case ixgbe_mac_X550:
5788 	case ixgbe_mac_X550EM_x:
5789 		pci_wake_from_d3(pdev, !!wufc);
5790 		break;
5791 	default:
5792 		break;
5793 	}
5794 
5795 	*enable_wake = !!wufc;
5796 
5797 	ixgbe_release_hw_control(adapter);
5798 
5799 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
5800 		pci_disable_device(pdev);
5801 
5802 	return 0;
5803 }
5804 
5805 #ifdef CONFIG_PM
5806 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
5807 {
5808 	int retval;
5809 	bool wake;
5810 
5811 	retval = __ixgbe_shutdown(pdev, &wake);
5812 	if (retval)
5813 		return retval;
5814 
5815 	if (wake) {
5816 		pci_prepare_to_sleep(pdev);
5817 	} else {
5818 		pci_wake_from_d3(pdev, false);
5819 		pci_set_power_state(pdev, PCI_D3hot);
5820 	}
5821 
5822 	return 0;
5823 }
5824 #endif /* CONFIG_PM */
5825 
5826 static void ixgbe_shutdown(struct pci_dev *pdev)
5827 {
5828 	bool wake;
5829 
5830 	__ixgbe_shutdown(pdev, &wake);
5831 
5832 	if (system_state == SYSTEM_POWER_OFF) {
5833 		pci_wake_from_d3(pdev, wake);
5834 		pci_set_power_state(pdev, PCI_D3hot);
5835 	}
5836 }
5837 
5838 /**
5839  * ixgbe_update_stats - Update the board statistics counters.
5840  * @adapter: board private structure
5841  **/
5842 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
5843 {
5844 	struct net_device *netdev = adapter->netdev;
5845 	struct ixgbe_hw *hw = &adapter->hw;
5846 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
5847 	u64 total_mpc = 0;
5848 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
5849 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
5850 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
5851 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
5852 
5853 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5854 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5855 		return;
5856 
5857 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
5858 		u64 rsc_count = 0;
5859 		u64 rsc_flush = 0;
5860 		for (i = 0; i < adapter->num_rx_queues; i++) {
5861 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
5862 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
5863 		}
5864 		adapter->rsc_total_count = rsc_count;
5865 		adapter->rsc_total_flush = rsc_flush;
5866 	}
5867 
5868 	for (i = 0; i < adapter->num_rx_queues; i++) {
5869 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
5870 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
5871 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
5872 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
5873 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
5874 		bytes += rx_ring->stats.bytes;
5875 		packets += rx_ring->stats.packets;
5876 	}
5877 	adapter->non_eop_descs = non_eop_descs;
5878 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
5879 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
5880 	adapter->hw_csum_rx_error = hw_csum_rx_error;
5881 	netdev->stats.rx_bytes = bytes;
5882 	netdev->stats.rx_packets = packets;
5883 
5884 	bytes = 0;
5885 	packets = 0;
5886 	/* gather some stats to the adapter struct that are per queue */
5887 	for (i = 0; i < adapter->num_tx_queues; i++) {
5888 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
5889 		restart_queue += tx_ring->tx_stats.restart_queue;
5890 		tx_busy += tx_ring->tx_stats.tx_busy;
5891 		bytes += tx_ring->stats.bytes;
5892 		packets += tx_ring->stats.packets;
5893 	}
5894 	adapter->restart_queue = restart_queue;
5895 	adapter->tx_busy = tx_busy;
5896 	netdev->stats.tx_bytes = bytes;
5897 	netdev->stats.tx_packets = packets;
5898 
5899 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
5900 
5901 	/* 8 register reads */
5902 	for (i = 0; i < 8; i++) {
5903 		/* for packet buffers not used, the register should read 0 */
5904 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
5905 		missed_rx += mpc;
5906 		hwstats->mpc[i] += mpc;
5907 		total_mpc += hwstats->mpc[i];
5908 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
5909 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
5910 		switch (hw->mac.type) {
5911 		case ixgbe_mac_82598EB:
5912 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
5913 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
5914 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
5915 			hwstats->pxonrxc[i] +=
5916 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
5917 			break;
5918 		case ixgbe_mac_82599EB:
5919 		case ixgbe_mac_X540:
5920 		case ixgbe_mac_X550:
5921 		case ixgbe_mac_X550EM_x:
5922 			hwstats->pxonrxc[i] +=
5923 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
5924 			break;
5925 		default:
5926 			break;
5927 		}
5928 	}
5929 
5930 	/*16 register reads */
5931 	for (i = 0; i < 16; i++) {
5932 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
5933 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
5934 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
5935 		    (hw->mac.type == ixgbe_mac_X540) ||
5936 		    (hw->mac.type == ixgbe_mac_X550) ||
5937 		    (hw->mac.type == ixgbe_mac_X550EM_x)) {
5938 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
5939 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
5940 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
5941 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
5942 		}
5943 	}
5944 
5945 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
5946 	/* work around hardware counting issue */
5947 	hwstats->gprc -= missed_rx;
5948 
5949 	ixgbe_update_xoff_received(adapter);
5950 
5951 	/* 82598 hardware only has a 32 bit counter in the high register */
5952 	switch (hw->mac.type) {
5953 	case ixgbe_mac_82598EB:
5954 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
5955 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
5956 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
5957 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
5958 		break;
5959 	case ixgbe_mac_X540:
5960 	case ixgbe_mac_X550:
5961 	case ixgbe_mac_X550EM_x:
5962 		/* OS2BMC stats are X540 and later */
5963 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
5964 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
5965 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
5966 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
5967 	case ixgbe_mac_82599EB:
5968 		for (i = 0; i < 16; i++)
5969 			adapter->hw_rx_no_dma_resources +=
5970 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
5971 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
5972 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
5973 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
5974 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
5975 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
5976 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
5977 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
5978 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
5979 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
5980 #ifdef IXGBE_FCOE
5981 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
5982 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
5983 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
5984 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
5985 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
5986 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
5987 		/* Add up per cpu counters for total ddp aloc fail */
5988 		if (adapter->fcoe.ddp_pool) {
5989 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
5990 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
5991 			unsigned int cpu;
5992 			u64 noddp = 0, noddp_ext_buff = 0;
5993 			for_each_possible_cpu(cpu) {
5994 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
5995 				noddp += ddp_pool->noddp;
5996 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
5997 			}
5998 			hwstats->fcoe_noddp = noddp;
5999 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
6000 		}
6001 #endif /* IXGBE_FCOE */
6002 		break;
6003 	default:
6004 		break;
6005 	}
6006 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
6007 	hwstats->bprc += bprc;
6008 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
6009 	if (hw->mac.type == ixgbe_mac_82598EB)
6010 		hwstats->mprc -= bprc;
6011 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
6012 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
6013 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
6014 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
6015 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
6016 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
6017 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
6018 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
6019 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
6020 	hwstats->lxontxc += lxon;
6021 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
6022 	hwstats->lxofftxc += lxoff;
6023 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
6024 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
6025 	/*
6026 	 * 82598 errata - tx of flow control packets is included in tx counters
6027 	 */
6028 	xon_off_tot = lxon + lxoff;
6029 	hwstats->gptc -= xon_off_tot;
6030 	hwstats->mptc -= xon_off_tot;
6031 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
6032 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
6033 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
6034 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
6035 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
6036 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
6037 	hwstats->ptc64 -= xon_off_tot;
6038 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
6039 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
6040 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
6041 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
6042 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
6043 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
6044 
6045 	/* Fill out the OS statistics structure */
6046 	netdev->stats.multicast = hwstats->mprc;
6047 
6048 	/* Rx Errors */
6049 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
6050 	netdev->stats.rx_dropped = 0;
6051 	netdev->stats.rx_length_errors = hwstats->rlec;
6052 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
6053 	netdev->stats.rx_missed_errors = total_mpc;
6054 }
6055 
6056 /**
6057  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
6058  * @adapter: pointer to the device adapter structure
6059  **/
6060 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
6061 {
6062 	struct ixgbe_hw *hw = &adapter->hw;
6063 	int i;
6064 
6065 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
6066 		return;
6067 
6068 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6069 
6070 	/* if interface is down do nothing */
6071 	if (test_bit(__IXGBE_DOWN, &adapter->state))
6072 		return;
6073 
6074 	/* do nothing if we are not using signature filters */
6075 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
6076 		return;
6077 
6078 	adapter->fdir_overflow++;
6079 
6080 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
6081 		for (i = 0; i < adapter->num_tx_queues; i++)
6082 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
6083 				&(adapter->tx_ring[i]->state));
6084 		/* re-enable flow director interrupts */
6085 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
6086 	} else {
6087 		e_err(probe, "failed to finish FDIR re-initialization, "
6088 		      "ignored adding FDIR ATR filters\n");
6089 	}
6090 }
6091 
6092 /**
6093  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
6094  * @adapter: pointer to the device adapter structure
6095  *
6096  * This function serves two purposes.  First it strobes the interrupt lines
6097  * in order to make certain interrupts are occurring.  Secondly it sets the
6098  * bits needed to check for TX hangs.  As a result we should immediately
6099  * determine if a hang has occurred.
6100  */
6101 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
6102 {
6103 	struct ixgbe_hw *hw = &adapter->hw;
6104 	u64 eics = 0;
6105 	int i;
6106 
6107 	/* If we're down, removing or resetting, just bail */
6108 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6109 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6110 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6111 		return;
6112 
6113 	/* Force detection of hung controller */
6114 	if (netif_carrier_ok(adapter->netdev)) {
6115 		for (i = 0; i < adapter->num_tx_queues; i++)
6116 			set_check_for_tx_hang(adapter->tx_ring[i]);
6117 	}
6118 
6119 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
6120 		/*
6121 		 * for legacy and MSI interrupts don't set any bits
6122 		 * that are enabled for EIAM, because this operation
6123 		 * would set *both* EIMS and EICS for any bit in EIAM
6124 		 */
6125 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
6126 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
6127 	} else {
6128 		/* get one bit for every active tx/rx interrupt vector */
6129 		for (i = 0; i < adapter->num_q_vectors; i++) {
6130 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
6131 			if (qv->rx.ring || qv->tx.ring)
6132 				eics |= ((u64)1 << i);
6133 		}
6134 	}
6135 
6136 	/* Cause software interrupt to ensure rings are cleaned */
6137 	ixgbe_irq_rearm_queues(adapter, eics);
6138 
6139 }
6140 
6141 /**
6142  * ixgbe_watchdog_update_link - update the link status
6143  * @adapter: pointer to the device adapter structure
6144  * @link_speed: pointer to a u32 to store the link_speed
6145  **/
6146 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
6147 {
6148 	struct ixgbe_hw *hw = &adapter->hw;
6149 	u32 link_speed = adapter->link_speed;
6150 	bool link_up = adapter->link_up;
6151 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
6152 
6153 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
6154 		return;
6155 
6156 	if (hw->mac.ops.check_link) {
6157 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
6158 	} else {
6159 		/* always assume link is up, if no check link function */
6160 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
6161 		link_up = true;
6162 	}
6163 
6164 	if (adapter->ixgbe_ieee_pfc)
6165 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
6166 
6167 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
6168 		hw->mac.ops.fc_enable(hw);
6169 		ixgbe_set_rx_drop_en(adapter);
6170 	}
6171 
6172 	if (link_up ||
6173 	    time_after(jiffies, (adapter->link_check_timeout +
6174 				 IXGBE_TRY_LINK_TIMEOUT))) {
6175 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6176 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
6177 		IXGBE_WRITE_FLUSH(hw);
6178 	}
6179 
6180 	adapter->link_up = link_up;
6181 	adapter->link_speed = link_speed;
6182 }
6183 
6184 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
6185 {
6186 #ifdef CONFIG_IXGBE_DCB
6187 	struct net_device *netdev = adapter->netdev;
6188 	struct dcb_app app = {
6189 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
6190 			      .protocol = 0,
6191 			     };
6192 	u8 up = 0;
6193 
6194 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
6195 		up = dcb_ieee_getapp_mask(netdev, &app);
6196 
6197 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
6198 #endif
6199 }
6200 
6201 /**
6202  * ixgbe_watchdog_link_is_up - update netif_carrier status and
6203  *                             print link up message
6204  * @adapter: pointer to the device adapter structure
6205  **/
6206 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
6207 {
6208 	struct net_device *netdev = adapter->netdev;
6209 	struct ixgbe_hw *hw = &adapter->hw;
6210 	struct net_device *upper;
6211 	struct list_head *iter;
6212 	u32 link_speed = adapter->link_speed;
6213 	bool flow_rx, flow_tx;
6214 
6215 	/* only continue if link was previously down */
6216 	if (netif_carrier_ok(netdev))
6217 		return;
6218 
6219 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
6220 
6221 	switch (hw->mac.type) {
6222 	case ixgbe_mac_82598EB: {
6223 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6224 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
6225 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
6226 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
6227 	}
6228 		break;
6229 	case ixgbe_mac_X540:
6230 	case ixgbe_mac_X550:
6231 	case ixgbe_mac_X550EM_x:
6232 	case ixgbe_mac_82599EB: {
6233 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
6234 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
6235 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
6236 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
6237 	}
6238 		break;
6239 	default:
6240 		flow_tx = false;
6241 		flow_rx = false;
6242 		break;
6243 	}
6244 
6245 	adapter->last_rx_ptp_check = jiffies;
6246 
6247 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6248 		ixgbe_ptp_start_cyclecounter(adapter);
6249 
6250 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n",
6251 	       (link_speed == IXGBE_LINK_SPEED_10GB_FULL ?
6252 	       "10 Gbps" :
6253 	       (link_speed == IXGBE_LINK_SPEED_1GB_FULL ?
6254 	       "1 Gbps" :
6255 	       (link_speed == IXGBE_LINK_SPEED_100_FULL ?
6256 	       "100 Mbps" :
6257 	       "unknown speed"))),
6258 	       ((flow_rx && flow_tx) ? "RX/TX" :
6259 	       (flow_rx ? "RX" :
6260 	       (flow_tx ? "TX" : "None"))));
6261 
6262 	netif_carrier_on(netdev);
6263 	ixgbe_check_vf_rate_limit(adapter);
6264 
6265 	/* enable transmits */
6266 	netif_tx_wake_all_queues(adapter->netdev);
6267 
6268 	/* enable any upper devices */
6269 	rtnl_lock();
6270 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
6271 		if (netif_is_macvlan(upper)) {
6272 			struct macvlan_dev *vlan = netdev_priv(upper);
6273 
6274 			if (vlan->fwd_priv)
6275 				netif_tx_wake_all_queues(upper);
6276 		}
6277 	}
6278 	rtnl_unlock();
6279 
6280 	/* update the default user priority for VFs */
6281 	ixgbe_update_default_up(adapter);
6282 
6283 	/* ping all the active vfs to let them know link has changed */
6284 	ixgbe_ping_all_vfs(adapter);
6285 }
6286 
6287 /**
6288  * ixgbe_watchdog_link_is_down - update netif_carrier status and
6289  *                               print link down message
6290  * @adapter: pointer to the adapter structure
6291  **/
6292 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
6293 {
6294 	struct net_device *netdev = adapter->netdev;
6295 	struct ixgbe_hw *hw = &adapter->hw;
6296 
6297 	adapter->link_up = false;
6298 	adapter->link_speed = 0;
6299 
6300 	/* only continue if link was up previously */
6301 	if (!netif_carrier_ok(netdev))
6302 		return;
6303 
6304 	/* poll for SFP+ cable when link is down */
6305 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
6306 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
6307 
6308 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6309 		ixgbe_ptp_start_cyclecounter(adapter);
6310 
6311 	e_info(drv, "NIC Link is Down\n");
6312 	netif_carrier_off(netdev);
6313 
6314 	/* ping all the active vfs to let them know link has changed */
6315 	ixgbe_ping_all_vfs(adapter);
6316 }
6317 
6318 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
6319 {
6320 	int i;
6321 
6322 	for (i = 0; i < adapter->num_tx_queues; i++) {
6323 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6324 
6325 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
6326 			return true;
6327 	}
6328 
6329 	return false;
6330 }
6331 
6332 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
6333 {
6334 	struct ixgbe_hw *hw = &adapter->hw;
6335 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
6336 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
6337 
6338 	int i, j;
6339 
6340 	if (!adapter->num_vfs)
6341 		return false;
6342 
6343 	/* resetting the PF is only needed for MAC before X550 */
6344 	if (hw->mac.type >= ixgbe_mac_X550)
6345 		return false;
6346 
6347 	for (i = 0; i < adapter->num_vfs; i++) {
6348 		for (j = 0; j < q_per_pool; j++) {
6349 			u32 h, t;
6350 
6351 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
6352 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
6353 
6354 			if (h != t)
6355 				return true;
6356 		}
6357 	}
6358 
6359 	return false;
6360 }
6361 
6362 /**
6363  * ixgbe_watchdog_flush_tx - flush queues on link down
6364  * @adapter: pointer to the device adapter structure
6365  **/
6366 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
6367 {
6368 	if (!netif_carrier_ok(adapter->netdev)) {
6369 		if (ixgbe_ring_tx_pending(adapter) ||
6370 		    ixgbe_vf_tx_pending(adapter)) {
6371 			/* We've lost link, so the controller stops DMA,
6372 			 * but we've got queued Tx work that's never going
6373 			 * to get done, so reset controller to flush Tx.
6374 			 * (Do the reset outside of interrupt context).
6375 			 */
6376 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
6377 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
6378 		}
6379 	}
6380 }
6381 
6382 #ifdef CONFIG_PCI_IOV
6383 static inline void ixgbe_issue_vf_flr(struct ixgbe_adapter *adapter,
6384 				      struct pci_dev *vfdev)
6385 {
6386 	if (!pci_wait_for_pending_transaction(vfdev))
6387 		e_dev_warn("Issuing VFLR with pending transactions\n");
6388 
6389 	e_dev_err("Issuing VFLR for VF %s\n", pci_name(vfdev));
6390 	pcie_capability_set_word(vfdev, PCI_EXP_DEVCTL, PCI_EXP_DEVCTL_BCR_FLR);
6391 
6392 	msleep(100);
6393 }
6394 
6395 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
6396 {
6397 	struct ixgbe_hw *hw = &adapter->hw;
6398 	struct pci_dev *pdev = adapter->pdev;
6399 	struct pci_dev *vfdev;
6400 	u32 gpc;
6401 	int pos;
6402 	unsigned short vf_id;
6403 
6404 	if (!(netif_carrier_ok(adapter->netdev)))
6405 		return;
6406 
6407 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
6408 	if (gpc) /* If incrementing then no need for the check below */
6409 		return;
6410 	/* Check to see if a bad DMA write target from an errant or
6411 	 * malicious VF has caused a PCIe error.  If so then we can
6412 	 * issue a VFLR to the offending VF(s) and then resume without
6413 	 * requesting a full slot reset.
6414 	 */
6415 
6416 	if (!pdev)
6417 		return;
6418 
6419 	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_SRIOV);
6420 	if (!pos)
6421 		return;
6422 
6423 	/* get the device ID for the VF */
6424 	pci_read_config_word(pdev, pos + PCI_SRIOV_VF_DID, &vf_id);
6425 
6426 	/* check status reg for all VFs owned by this PF */
6427 	vfdev = pci_get_device(pdev->vendor, vf_id, NULL);
6428 	while (vfdev) {
6429 		if (vfdev->is_virtfn && (vfdev->physfn == pdev)) {
6430 			u16 status_reg;
6431 
6432 			pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
6433 			if (status_reg & PCI_STATUS_REC_MASTER_ABORT)
6434 				/* issue VFLR */
6435 				ixgbe_issue_vf_flr(adapter, vfdev);
6436 		}
6437 
6438 		vfdev = pci_get_device(pdev->vendor, vf_id, vfdev);
6439 	}
6440 }
6441 
6442 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
6443 {
6444 	u32 ssvpc;
6445 
6446 	/* Do not perform spoof check for 82598 or if not in IOV mode */
6447 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
6448 	    adapter->num_vfs == 0)
6449 		return;
6450 
6451 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
6452 
6453 	/*
6454 	 * ssvpc register is cleared on read, if zero then no
6455 	 * spoofed packets in the last interval.
6456 	 */
6457 	if (!ssvpc)
6458 		return;
6459 
6460 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
6461 }
6462 #else
6463 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
6464 {
6465 }
6466 
6467 static void
6468 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
6469 {
6470 }
6471 #endif /* CONFIG_PCI_IOV */
6472 
6473 
6474 /**
6475  * ixgbe_watchdog_subtask - check and bring link up
6476  * @adapter: pointer to the device adapter structure
6477  **/
6478 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
6479 {
6480 	/* if interface is down, removing or resetting, do nothing */
6481 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6482 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6483 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6484 		return;
6485 
6486 	ixgbe_watchdog_update_link(adapter);
6487 
6488 	if (adapter->link_up)
6489 		ixgbe_watchdog_link_is_up(adapter);
6490 	else
6491 		ixgbe_watchdog_link_is_down(adapter);
6492 
6493 	ixgbe_check_for_bad_vf(adapter);
6494 	ixgbe_spoof_check(adapter);
6495 	ixgbe_update_stats(adapter);
6496 
6497 	ixgbe_watchdog_flush_tx(adapter);
6498 }
6499 
6500 /**
6501  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
6502  * @adapter: the ixgbe adapter structure
6503  **/
6504 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
6505 {
6506 	struct ixgbe_hw *hw = &adapter->hw;
6507 	s32 err;
6508 
6509 	/* not searching for SFP so there is nothing to do here */
6510 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
6511 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
6512 		return;
6513 
6514 	/* someone else is in init, wait until next service event */
6515 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
6516 		return;
6517 
6518 	err = hw->phy.ops.identify_sfp(hw);
6519 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
6520 		goto sfp_out;
6521 
6522 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
6523 		/* If no cable is present, then we need to reset
6524 		 * the next time we find a good cable. */
6525 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
6526 	}
6527 
6528 	/* exit on error */
6529 	if (err)
6530 		goto sfp_out;
6531 
6532 	/* exit if reset not needed */
6533 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
6534 		goto sfp_out;
6535 
6536 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
6537 
6538 	/*
6539 	 * A module may be identified correctly, but the EEPROM may not have
6540 	 * support for that module.  setup_sfp() will fail in that case, so
6541 	 * we should not allow that module to load.
6542 	 */
6543 	if (hw->mac.type == ixgbe_mac_82598EB)
6544 		err = hw->phy.ops.reset(hw);
6545 	else
6546 		err = hw->mac.ops.setup_sfp(hw);
6547 
6548 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
6549 		goto sfp_out;
6550 
6551 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
6552 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
6553 
6554 sfp_out:
6555 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
6556 
6557 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
6558 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
6559 		e_dev_err("failed to initialize because an unsupported "
6560 			  "SFP+ module type was detected.\n");
6561 		e_dev_err("Reload the driver after installing a "
6562 			  "supported module.\n");
6563 		unregister_netdev(adapter->netdev);
6564 	}
6565 }
6566 
6567 /**
6568  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
6569  * @adapter: the ixgbe adapter structure
6570  **/
6571 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
6572 {
6573 	struct ixgbe_hw *hw = &adapter->hw;
6574 	u32 speed;
6575 	bool autoneg = false;
6576 
6577 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
6578 		return;
6579 
6580 	/* someone else is in init, wait until next service event */
6581 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
6582 		return;
6583 
6584 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
6585 
6586 	speed = hw->phy.autoneg_advertised;
6587 	if ((!speed) && (hw->mac.ops.get_link_capabilities)) {
6588 		hw->mac.ops.get_link_capabilities(hw, &speed, &autoneg);
6589 
6590 		/* setup the highest link when no autoneg */
6591 		if (!autoneg) {
6592 			if (speed & IXGBE_LINK_SPEED_10GB_FULL)
6593 				speed = IXGBE_LINK_SPEED_10GB_FULL;
6594 		}
6595 	}
6596 
6597 	if (hw->mac.ops.setup_link)
6598 		hw->mac.ops.setup_link(hw, speed, true);
6599 
6600 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
6601 	adapter->link_check_timeout = jiffies;
6602 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
6603 }
6604 
6605 /**
6606  * ixgbe_service_timer - Timer Call-back
6607  * @data: pointer to adapter cast into an unsigned long
6608  **/
6609 static void ixgbe_service_timer(unsigned long data)
6610 {
6611 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
6612 	unsigned long next_event_offset;
6613 
6614 	/* poll faster when waiting for link */
6615 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
6616 		next_event_offset = HZ / 10;
6617 	else
6618 		next_event_offset = HZ * 2;
6619 
6620 	/* Reset the timer */
6621 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
6622 
6623 	ixgbe_service_event_schedule(adapter);
6624 }
6625 
6626 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
6627 {
6628 	if (!(adapter->flags2 & IXGBE_FLAG2_RESET_REQUESTED))
6629 		return;
6630 
6631 	adapter->flags2 &= ~IXGBE_FLAG2_RESET_REQUESTED;
6632 
6633 	/* If we're already down, removing or resetting, just bail */
6634 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6635 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6636 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6637 		return;
6638 
6639 	ixgbe_dump(adapter);
6640 	netdev_err(adapter->netdev, "Reset adapter\n");
6641 	adapter->tx_timeout_count++;
6642 
6643 	rtnl_lock();
6644 	ixgbe_reinit_locked(adapter);
6645 	rtnl_unlock();
6646 }
6647 
6648 /**
6649  * ixgbe_service_task - manages and runs subtasks
6650  * @work: pointer to work_struct containing our data
6651  **/
6652 static void ixgbe_service_task(struct work_struct *work)
6653 {
6654 	struct ixgbe_adapter *adapter = container_of(work,
6655 						     struct ixgbe_adapter,
6656 						     service_task);
6657 	if (ixgbe_removed(adapter->hw.hw_addr)) {
6658 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
6659 			rtnl_lock();
6660 			ixgbe_down(adapter);
6661 			rtnl_unlock();
6662 		}
6663 		ixgbe_service_event_complete(adapter);
6664 		return;
6665 	}
6666 	ixgbe_reset_subtask(adapter);
6667 	ixgbe_sfp_detection_subtask(adapter);
6668 	ixgbe_sfp_link_config_subtask(adapter);
6669 	ixgbe_check_overtemp_subtask(adapter);
6670 	ixgbe_watchdog_subtask(adapter);
6671 	ixgbe_fdir_reinit_subtask(adapter);
6672 	ixgbe_check_hang_subtask(adapter);
6673 
6674 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
6675 		ixgbe_ptp_overflow_check(adapter);
6676 		ixgbe_ptp_rx_hang(adapter);
6677 	}
6678 
6679 	ixgbe_service_event_complete(adapter);
6680 }
6681 
6682 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
6683 		     struct ixgbe_tx_buffer *first,
6684 		     u8 *hdr_len)
6685 {
6686 	struct sk_buff *skb = first->skb;
6687 	u32 vlan_macip_lens, type_tucmd;
6688 	u32 mss_l4len_idx, l4len;
6689 	int err;
6690 
6691 	if (skb->ip_summed != CHECKSUM_PARTIAL)
6692 		return 0;
6693 
6694 	if (!skb_is_gso(skb))
6695 		return 0;
6696 
6697 	err = skb_cow_head(skb, 0);
6698 	if (err < 0)
6699 		return err;
6700 
6701 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
6702 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
6703 
6704 	if (first->protocol == htons(ETH_P_IP)) {
6705 		struct iphdr *iph = ip_hdr(skb);
6706 		iph->tot_len = 0;
6707 		iph->check = 0;
6708 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
6709 							 iph->daddr, 0,
6710 							 IPPROTO_TCP,
6711 							 0);
6712 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
6713 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
6714 				   IXGBE_TX_FLAGS_CSUM |
6715 				   IXGBE_TX_FLAGS_IPV4;
6716 	} else if (skb_is_gso_v6(skb)) {
6717 		ipv6_hdr(skb)->payload_len = 0;
6718 		tcp_hdr(skb)->check =
6719 		    ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
6720 				     &ipv6_hdr(skb)->daddr,
6721 				     0, IPPROTO_TCP, 0);
6722 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
6723 				   IXGBE_TX_FLAGS_CSUM;
6724 	}
6725 
6726 	/* compute header lengths */
6727 	l4len = tcp_hdrlen(skb);
6728 	*hdr_len = skb_transport_offset(skb) + l4len;
6729 
6730 	/* update gso size and bytecount with header size */
6731 	first->gso_segs = skb_shinfo(skb)->gso_segs;
6732 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
6733 
6734 	/* mss_l4len_id: use 0 as index for TSO */
6735 	mss_l4len_idx = l4len << IXGBE_ADVTXD_L4LEN_SHIFT;
6736 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
6737 
6738 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
6739 	vlan_macip_lens = skb_network_header_len(skb);
6740 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
6741 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
6742 
6743 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
6744 			  mss_l4len_idx);
6745 
6746 	return 1;
6747 }
6748 
6749 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
6750 			  struct ixgbe_tx_buffer *first)
6751 {
6752 	struct sk_buff *skb = first->skb;
6753 	u32 vlan_macip_lens = 0;
6754 	u32 mss_l4len_idx = 0;
6755 	u32 type_tucmd = 0;
6756 
6757 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
6758 		if (!(first->tx_flags & IXGBE_TX_FLAGS_HW_VLAN) &&
6759 		    !(first->tx_flags & IXGBE_TX_FLAGS_CC))
6760 			return;
6761 	} else {
6762 		u8 l4_hdr = 0;
6763 		switch (first->protocol) {
6764 		case htons(ETH_P_IP):
6765 			vlan_macip_lens |= skb_network_header_len(skb);
6766 			type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
6767 			l4_hdr = ip_hdr(skb)->protocol;
6768 			break;
6769 		case htons(ETH_P_IPV6):
6770 			vlan_macip_lens |= skb_network_header_len(skb);
6771 			l4_hdr = ipv6_hdr(skb)->nexthdr;
6772 			break;
6773 		default:
6774 			if (unlikely(net_ratelimit())) {
6775 				dev_warn(tx_ring->dev,
6776 				 "partial checksum but proto=%x!\n",
6777 				 first->protocol);
6778 			}
6779 			break;
6780 		}
6781 
6782 		switch (l4_hdr) {
6783 		case IPPROTO_TCP:
6784 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_TCP;
6785 			mss_l4len_idx = tcp_hdrlen(skb) <<
6786 					IXGBE_ADVTXD_L4LEN_SHIFT;
6787 			break;
6788 		case IPPROTO_SCTP:
6789 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_SCTP;
6790 			mss_l4len_idx = sizeof(struct sctphdr) <<
6791 					IXGBE_ADVTXD_L4LEN_SHIFT;
6792 			break;
6793 		case IPPROTO_UDP:
6794 			mss_l4len_idx = sizeof(struct udphdr) <<
6795 					IXGBE_ADVTXD_L4LEN_SHIFT;
6796 			break;
6797 		default:
6798 			if (unlikely(net_ratelimit())) {
6799 				dev_warn(tx_ring->dev,
6800 				 "partial checksum but l4 proto=%x!\n",
6801 				 l4_hdr);
6802 			}
6803 			break;
6804 		}
6805 
6806 		/* update TX checksum flag */
6807 		first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
6808 	}
6809 
6810 	/* vlan_macip_lens: MACLEN, VLAN tag */
6811 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
6812 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
6813 
6814 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0,
6815 			  type_tucmd, mss_l4len_idx);
6816 }
6817 
6818 #define IXGBE_SET_FLAG(_input, _flag, _result) \
6819 	((_flag <= _result) ? \
6820 	 ((u32)(_input & _flag) * (_result / _flag)) : \
6821 	 ((u32)(_input & _flag) / (_flag / _result)))
6822 
6823 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
6824 {
6825 	/* set type for advanced descriptor with frame checksum insertion */
6826 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
6827 		       IXGBE_ADVTXD_DCMD_DEXT |
6828 		       IXGBE_ADVTXD_DCMD_IFCS;
6829 
6830 	/* set HW vlan bit if vlan is present */
6831 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
6832 				   IXGBE_ADVTXD_DCMD_VLE);
6833 
6834 	/* set segmentation enable bits for TSO/FSO */
6835 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
6836 				   IXGBE_ADVTXD_DCMD_TSE);
6837 
6838 	/* set timestamp bit if present */
6839 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
6840 				   IXGBE_ADVTXD_MAC_TSTAMP);
6841 
6842 	/* insert frame checksum */
6843 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
6844 
6845 	return cmd_type;
6846 }
6847 
6848 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
6849 				   u32 tx_flags, unsigned int paylen)
6850 {
6851 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
6852 
6853 	/* enable L4 checksum for TSO and TX checksum offload */
6854 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
6855 					IXGBE_TX_FLAGS_CSUM,
6856 					IXGBE_ADVTXD_POPTS_TXSM);
6857 
6858 	/* enble IPv4 checksum for TSO */
6859 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
6860 					IXGBE_TX_FLAGS_IPV4,
6861 					IXGBE_ADVTXD_POPTS_IXSM);
6862 
6863 	/*
6864 	 * Check Context must be set if Tx switch is enabled, which it
6865 	 * always is for case where virtual functions are running
6866 	 */
6867 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
6868 					IXGBE_TX_FLAGS_CC,
6869 					IXGBE_ADVTXD_CC);
6870 
6871 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
6872 }
6873 
6874 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
6875 {
6876 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
6877 
6878 	/* Herbert's original patch had:
6879 	 *  smp_mb__after_netif_stop_queue();
6880 	 * but since that doesn't exist yet, just open code it.
6881 	 */
6882 	smp_mb();
6883 
6884 	/* We need to check again in a case another CPU has just
6885 	 * made room available.
6886 	 */
6887 	if (likely(ixgbe_desc_unused(tx_ring) < size))
6888 		return -EBUSY;
6889 
6890 	/* A reprieve! - use start_queue because it doesn't call schedule */
6891 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
6892 	++tx_ring->tx_stats.restart_queue;
6893 	return 0;
6894 }
6895 
6896 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
6897 {
6898 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
6899 		return 0;
6900 
6901 	return __ixgbe_maybe_stop_tx(tx_ring, size);
6902 }
6903 
6904 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
6905 		       IXGBE_TXD_CMD_RS)
6906 
6907 static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
6908 			 struct ixgbe_tx_buffer *first,
6909 			 const u8 hdr_len)
6910 {
6911 	struct sk_buff *skb = first->skb;
6912 	struct ixgbe_tx_buffer *tx_buffer;
6913 	union ixgbe_adv_tx_desc *tx_desc;
6914 	struct skb_frag_struct *frag;
6915 	dma_addr_t dma;
6916 	unsigned int data_len, size;
6917 	u32 tx_flags = first->tx_flags;
6918 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
6919 	u16 i = tx_ring->next_to_use;
6920 
6921 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
6922 
6923 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
6924 
6925 	size = skb_headlen(skb);
6926 	data_len = skb->data_len;
6927 
6928 #ifdef IXGBE_FCOE
6929 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
6930 		if (data_len < sizeof(struct fcoe_crc_eof)) {
6931 			size -= sizeof(struct fcoe_crc_eof) - data_len;
6932 			data_len = 0;
6933 		} else {
6934 			data_len -= sizeof(struct fcoe_crc_eof);
6935 		}
6936 	}
6937 
6938 #endif
6939 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
6940 
6941 	tx_buffer = first;
6942 
6943 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
6944 		if (dma_mapping_error(tx_ring->dev, dma))
6945 			goto dma_error;
6946 
6947 		/* record length, and DMA address */
6948 		dma_unmap_len_set(tx_buffer, len, size);
6949 		dma_unmap_addr_set(tx_buffer, dma, dma);
6950 
6951 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
6952 
6953 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
6954 			tx_desc->read.cmd_type_len =
6955 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
6956 
6957 			i++;
6958 			tx_desc++;
6959 			if (i == tx_ring->count) {
6960 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6961 				i = 0;
6962 			}
6963 			tx_desc->read.olinfo_status = 0;
6964 
6965 			dma += IXGBE_MAX_DATA_PER_TXD;
6966 			size -= IXGBE_MAX_DATA_PER_TXD;
6967 
6968 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
6969 		}
6970 
6971 		if (likely(!data_len))
6972 			break;
6973 
6974 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
6975 
6976 		i++;
6977 		tx_desc++;
6978 		if (i == tx_ring->count) {
6979 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6980 			i = 0;
6981 		}
6982 		tx_desc->read.olinfo_status = 0;
6983 
6984 #ifdef IXGBE_FCOE
6985 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
6986 #else
6987 		size = skb_frag_size(frag);
6988 #endif
6989 		data_len -= size;
6990 
6991 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
6992 				       DMA_TO_DEVICE);
6993 
6994 		tx_buffer = &tx_ring->tx_buffer_info[i];
6995 	}
6996 
6997 	/* write last descriptor with RS and EOP bits */
6998 	cmd_type |= size | IXGBE_TXD_CMD;
6999 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
7000 
7001 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
7002 
7003 	/* set the timestamp */
7004 	first->time_stamp = jiffies;
7005 
7006 	/*
7007 	 * Force memory writes to complete before letting h/w know there
7008 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
7009 	 * memory model archs, such as IA-64).
7010 	 *
7011 	 * We also need this memory barrier to make certain all of the
7012 	 * status bits have been updated before next_to_watch is written.
7013 	 */
7014 	wmb();
7015 
7016 	/* set next_to_watch value indicating a packet is present */
7017 	first->next_to_watch = tx_desc;
7018 
7019 	i++;
7020 	if (i == tx_ring->count)
7021 		i = 0;
7022 
7023 	tx_ring->next_to_use = i;
7024 
7025 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
7026 
7027 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
7028 		writel(i, tx_ring->tail);
7029 
7030 		/* we need this if more than one processor can write to our tail
7031 		 * at a time, it synchronizes IO on IA64/Altix systems
7032 		 */
7033 		mmiowb();
7034 	}
7035 
7036 	return;
7037 dma_error:
7038 	dev_err(tx_ring->dev, "TX DMA map failed\n");
7039 
7040 	/* clear dma mappings for failed tx_buffer_info map */
7041 	for (;;) {
7042 		tx_buffer = &tx_ring->tx_buffer_info[i];
7043 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer);
7044 		if (tx_buffer == first)
7045 			break;
7046 		if (i == 0)
7047 			i = tx_ring->count;
7048 		i--;
7049 	}
7050 
7051 	tx_ring->next_to_use = i;
7052 }
7053 
7054 static void ixgbe_atr(struct ixgbe_ring *ring,
7055 		      struct ixgbe_tx_buffer *first)
7056 {
7057 	struct ixgbe_q_vector *q_vector = ring->q_vector;
7058 	union ixgbe_atr_hash_dword input = { .dword = 0 };
7059 	union ixgbe_atr_hash_dword common = { .dword = 0 };
7060 	union {
7061 		unsigned char *network;
7062 		struct iphdr *ipv4;
7063 		struct ipv6hdr *ipv6;
7064 	} hdr;
7065 	struct tcphdr *th;
7066 	__be16 vlan_id;
7067 
7068 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
7069 	if (!q_vector)
7070 		return;
7071 
7072 	/* do nothing if sampling is disabled */
7073 	if (!ring->atr_sample_rate)
7074 		return;
7075 
7076 	ring->atr_count++;
7077 
7078 	/* snag network header to get L4 type and address */
7079 	hdr.network = skb_network_header(first->skb);
7080 
7081 	/* Currently only IPv4/IPv6 with TCP is supported */
7082 	if ((first->protocol != htons(ETH_P_IPV6) ||
7083 	     hdr.ipv6->nexthdr != IPPROTO_TCP) &&
7084 	    (first->protocol != htons(ETH_P_IP) ||
7085 	     hdr.ipv4->protocol != IPPROTO_TCP))
7086 		return;
7087 
7088 	th = tcp_hdr(first->skb);
7089 
7090 	/* skip this packet since it is invalid or the socket is closing */
7091 	if (!th || th->fin)
7092 		return;
7093 
7094 	/* sample on all syn packets or once every atr sample count */
7095 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
7096 		return;
7097 
7098 	/* reset sample count */
7099 	ring->atr_count = 0;
7100 
7101 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
7102 
7103 	/*
7104 	 * src and dst are inverted, think how the receiver sees them
7105 	 *
7106 	 * The input is broken into two sections, a non-compressed section
7107 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
7108 	 * is XORed together and stored in the compressed dword.
7109 	 */
7110 	input.formatted.vlan_id = vlan_id;
7111 
7112 	/*
7113 	 * since src port and flex bytes occupy the same word XOR them together
7114 	 * and write the value to source port portion of compressed dword
7115 	 */
7116 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
7117 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
7118 	else
7119 		common.port.src ^= th->dest ^ first->protocol;
7120 	common.port.dst ^= th->source;
7121 
7122 	if (first->protocol == htons(ETH_P_IP)) {
7123 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
7124 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
7125 	} else {
7126 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
7127 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
7128 			     hdr.ipv6->saddr.s6_addr32[1] ^
7129 			     hdr.ipv6->saddr.s6_addr32[2] ^
7130 			     hdr.ipv6->saddr.s6_addr32[3] ^
7131 			     hdr.ipv6->daddr.s6_addr32[0] ^
7132 			     hdr.ipv6->daddr.s6_addr32[1] ^
7133 			     hdr.ipv6->daddr.s6_addr32[2] ^
7134 			     hdr.ipv6->daddr.s6_addr32[3];
7135 	}
7136 
7137 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
7138 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
7139 					      input, common, ring->queue_index);
7140 }
7141 
7142 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
7143 			      void *accel_priv, select_queue_fallback_t fallback)
7144 {
7145 	struct ixgbe_fwd_adapter *fwd_adapter = accel_priv;
7146 #ifdef IXGBE_FCOE
7147 	struct ixgbe_adapter *adapter;
7148 	struct ixgbe_ring_feature *f;
7149 	int txq;
7150 #endif
7151 
7152 	if (fwd_adapter)
7153 		return skb->queue_mapping + fwd_adapter->tx_base_queue;
7154 
7155 #ifdef IXGBE_FCOE
7156 
7157 	/*
7158 	 * only execute the code below if protocol is FCoE
7159 	 * or FIP and we have FCoE enabled on the adapter
7160 	 */
7161 	switch (vlan_get_protocol(skb)) {
7162 	case htons(ETH_P_FCOE):
7163 	case htons(ETH_P_FIP):
7164 		adapter = netdev_priv(dev);
7165 
7166 		if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
7167 			break;
7168 	default:
7169 		return fallback(dev, skb);
7170 	}
7171 
7172 	f = &adapter->ring_feature[RING_F_FCOE];
7173 
7174 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
7175 					   smp_processor_id();
7176 
7177 	while (txq >= f->indices)
7178 		txq -= f->indices;
7179 
7180 	return txq + f->offset;
7181 #else
7182 	return fallback(dev, skb);
7183 #endif
7184 }
7185 
7186 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
7187 			  struct ixgbe_adapter *adapter,
7188 			  struct ixgbe_ring *tx_ring)
7189 {
7190 	struct ixgbe_tx_buffer *first;
7191 	int tso;
7192 	u32 tx_flags = 0;
7193 	unsigned short f;
7194 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
7195 	__be16 protocol = skb->protocol;
7196 	u8 hdr_len = 0;
7197 
7198 	/*
7199 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
7200 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
7201 	 *       + 2 desc gap to keep tail from touching head,
7202 	 *       + 1 desc for context descriptor,
7203 	 * otherwise try next time
7204 	 */
7205 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
7206 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
7207 
7208 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
7209 		tx_ring->tx_stats.tx_busy++;
7210 		return NETDEV_TX_BUSY;
7211 	}
7212 
7213 	/* record the location of the first descriptor for this packet */
7214 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
7215 	first->skb = skb;
7216 	first->bytecount = skb->len;
7217 	first->gso_segs = 1;
7218 
7219 	/* if we have a HW VLAN tag being added default to the HW one */
7220 	if (vlan_tx_tag_present(skb)) {
7221 		tx_flags |= vlan_tx_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
7222 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7223 	/* else if it is a SW VLAN check the next protocol and store the tag */
7224 	} else if (protocol == htons(ETH_P_8021Q)) {
7225 		struct vlan_hdr *vhdr, _vhdr;
7226 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
7227 		if (!vhdr)
7228 			goto out_drop;
7229 
7230 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
7231 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
7232 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
7233 	}
7234 	protocol = vlan_get_protocol(skb);
7235 
7236 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
7237 	    adapter->ptp_clock &&
7238 	    !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
7239 				   &adapter->state)) {
7240 		skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
7241 		tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
7242 
7243 		/* schedule check for Tx timestamp */
7244 		adapter->ptp_tx_skb = skb_get(skb);
7245 		adapter->ptp_tx_start = jiffies;
7246 		schedule_work(&adapter->ptp_tx_work);
7247 	}
7248 
7249 	skb_tx_timestamp(skb);
7250 
7251 #ifdef CONFIG_PCI_IOV
7252 	/*
7253 	 * Use the l2switch_enable flag - would be false if the DMA
7254 	 * Tx switch had been disabled.
7255 	 */
7256 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7257 		tx_flags |= IXGBE_TX_FLAGS_CC;
7258 
7259 #endif
7260 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
7261 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
7262 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
7263 	     (skb->priority != TC_PRIO_CONTROL))) {
7264 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
7265 		tx_flags |= (skb->priority & 0x7) <<
7266 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
7267 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
7268 			struct vlan_ethhdr *vhdr;
7269 
7270 			if (skb_cow_head(skb, 0))
7271 				goto out_drop;
7272 			vhdr = (struct vlan_ethhdr *)skb->data;
7273 			vhdr->h_vlan_TCI = htons(tx_flags >>
7274 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
7275 		} else {
7276 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7277 		}
7278 	}
7279 
7280 	/* record initial flags and protocol */
7281 	first->tx_flags = tx_flags;
7282 	first->protocol = protocol;
7283 
7284 #ifdef IXGBE_FCOE
7285 	/* setup tx offload for FCoE */
7286 	if ((protocol == htons(ETH_P_FCOE)) &&
7287 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
7288 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
7289 		if (tso < 0)
7290 			goto out_drop;
7291 
7292 		goto xmit_fcoe;
7293 	}
7294 
7295 #endif /* IXGBE_FCOE */
7296 	tso = ixgbe_tso(tx_ring, first, &hdr_len);
7297 	if (tso < 0)
7298 		goto out_drop;
7299 	else if (!tso)
7300 		ixgbe_tx_csum(tx_ring, first);
7301 
7302 	/* add the ATR filter if ATR is on */
7303 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
7304 		ixgbe_atr(tx_ring, first);
7305 
7306 #ifdef IXGBE_FCOE
7307 xmit_fcoe:
7308 #endif /* IXGBE_FCOE */
7309 	ixgbe_tx_map(tx_ring, first, hdr_len);
7310 
7311 	return NETDEV_TX_OK;
7312 
7313 out_drop:
7314 	dev_kfree_skb_any(first->skb);
7315 	first->skb = NULL;
7316 
7317 	return NETDEV_TX_OK;
7318 }
7319 
7320 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
7321 				      struct net_device *netdev,
7322 				      struct ixgbe_ring *ring)
7323 {
7324 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7325 	struct ixgbe_ring *tx_ring;
7326 
7327 	/*
7328 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
7329 	 * in order to meet this minimum size requirement.
7330 	 */
7331 	if (skb_put_padto(skb, 17))
7332 		return NETDEV_TX_OK;
7333 
7334 	tx_ring = ring ? ring : adapter->tx_ring[skb->queue_mapping];
7335 
7336 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
7337 }
7338 
7339 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
7340 				    struct net_device *netdev)
7341 {
7342 	return __ixgbe_xmit_frame(skb, netdev, NULL);
7343 }
7344 
7345 /**
7346  * ixgbe_set_mac - Change the Ethernet Address of the NIC
7347  * @netdev: network interface device structure
7348  * @p: pointer to an address structure
7349  *
7350  * Returns 0 on success, negative on failure
7351  **/
7352 static int ixgbe_set_mac(struct net_device *netdev, void *p)
7353 {
7354 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7355 	struct ixgbe_hw *hw = &adapter->hw;
7356 	struct sockaddr *addr = p;
7357 	int ret;
7358 
7359 	if (!is_valid_ether_addr(addr->sa_data))
7360 		return -EADDRNOTAVAIL;
7361 
7362 	ixgbe_del_mac_filter(adapter, hw->mac.addr, VMDQ_P(0));
7363 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
7364 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
7365 
7366 	ret = ixgbe_add_mac_filter(adapter, hw->mac.addr, VMDQ_P(0));
7367 	return ret > 0 ? 0 : ret;
7368 }
7369 
7370 static int
7371 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
7372 {
7373 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7374 	struct ixgbe_hw *hw = &adapter->hw;
7375 	u16 value;
7376 	int rc;
7377 
7378 	if (prtad != hw->phy.mdio.prtad)
7379 		return -EINVAL;
7380 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
7381 	if (!rc)
7382 		rc = value;
7383 	return rc;
7384 }
7385 
7386 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
7387 			    u16 addr, u16 value)
7388 {
7389 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7390 	struct ixgbe_hw *hw = &adapter->hw;
7391 
7392 	if (prtad != hw->phy.mdio.prtad)
7393 		return -EINVAL;
7394 	return hw->phy.ops.write_reg(hw, addr, devad, value);
7395 }
7396 
7397 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
7398 {
7399 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7400 
7401 	switch (cmd) {
7402 	case SIOCSHWTSTAMP:
7403 		return ixgbe_ptp_set_ts_config(adapter, req);
7404 	case SIOCGHWTSTAMP:
7405 		return ixgbe_ptp_get_ts_config(adapter, req);
7406 	default:
7407 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
7408 	}
7409 }
7410 
7411 /**
7412  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
7413  * netdev->dev_addrs
7414  * @netdev: network interface device structure
7415  *
7416  * Returns non-zero on failure
7417  **/
7418 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
7419 {
7420 	int err = 0;
7421 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7422 	struct ixgbe_hw *hw = &adapter->hw;
7423 
7424 	if (is_valid_ether_addr(hw->mac.san_addr)) {
7425 		rtnl_lock();
7426 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
7427 		rtnl_unlock();
7428 
7429 		/* update SAN MAC vmdq pool selection */
7430 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
7431 	}
7432 	return err;
7433 }
7434 
7435 /**
7436  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
7437  * netdev->dev_addrs
7438  * @netdev: network interface device structure
7439  *
7440  * Returns non-zero on failure
7441  **/
7442 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
7443 {
7444 	int err = 0;
7445 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7446 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
7447 
7448 	if (is_valid_ether_addr(mac->san_addr)) {
7449 		rtnl_lock();
7450 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
7451 		rtnl_unlock();
7452 	}
7453 	return err;
7454 }
7455 
7456 #ifdef CONFIG_NET_POLL_CONTROLLER
7457 /*
7458  * Polling 'interrupt' - used by things like netconsole to send skbs
7459  * without having to re-enable interrupts. It's not called while
7460  * the interrupt routine is executing.
7461  */
7462 static void ixgbe_netpoll(struct net_device *netdev)
7463 {
7464 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7465 	int i;
7466 
7467 	/* if interface is down do nothing */
7468 	if (test_bit(__IXGBE_DOWN, &adapter->state))
7469 		return;
7470 
7471 	adapter->flags |= IXGBE_FLAG_IN_NETPOLL;
7472 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
7473 		for (i = 0; i < adapter->num_q_vectors; i++)
7474 			ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
7475 	} else {
7476 		ixgbe_intr(adapter->pdev->irq, netdev);
7477 	}
7478 	adapter->flags &= ~IXGBE_FLAG_IN_NETPOLL;
7479 }
7480 
7481 #endif
7482 static struct rtnl_link_stats64 *ixgbe_get_stats64(struct net_device *netdev,
7483 						   struct rtnl_link_stats64 *stats)
7484 {
7485 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7486 	int i;
7487 
7488 	rcu_read_lock();
7489 	for (i = 0; i < adapter->num_rx_queues; i++) {
7490 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
7491 		u64 bytes, packets;
7492 		unsigned int start;
7493 
7494 		if (ring) {
7495 			do {
7496 				start = u64_stats_fetch_begin_irq(&ring->syncp);
7497 				packets = ring->stats.packets;
7498 				bytes   = ring->stats.bytes;
7499 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
7500 			stats->rx_packets += packets;
7501 			stats->rx_bytes   += bytes;
7502 		}
7503 	}
7504 
7505 	for (i = 0; i < adapter->num_tx_queues; i++) {
7506 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
7507 		u64 bytes, packets;
7508 		unsigned int start;
7509 
7510 		if (ring) {
7511 			do {
7512 				start = u64_stats_fetch_begin_irq(&ring->syncp);
7513 				packets = ring->stats.packets;
7514 				bytes   = ring->stats.bytes;
7515 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
7516 			stats->tx_packets += packets;
7517 			stats->tx_bytes   += bytes;
7518 		}
7519 	}
7520 	rcu_read_unlock();
7521 	/* following stats updated by ixgbe_watchdog_task() */
7522 	stats->multicast	= netdev->stats.multicast;
7523 	stats->rx_errors	= netdev->stats.rx_errors;
7524 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
7525 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
7526 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
7527 	return stats;
7528 }
7529 
7530 #ifdef CONFIG_IXGBE_DCB
7531 /**
7532  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
7533  * @adapter: pointer to ixgbe_adapter
7534  * @tc: number of traffic classes currently enabled
7535  *
7536  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
7537  * 802.1Q priority maps to a packet buffer that exists.
7538  */
7539 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
7540 {
7541 	struct ixgbe_hw *hw = &adapter->hw;
7542 	u32 reg, rsave;
7543 	int i;
7544 
7545 	/* 82598 have a static priority to TC mapping that can not
7546 	 * be changed so no validation is needed.
7547 	 */
7548 	if (hw->mac.type == ixgbe_mac_82598EB)
7549 		return;
7550 
7551 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
7552 	rsave = reg;
7553 
7554 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
7555 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
7556 
7557 		/* If up2tc is out of bounds default to zero */
7558 		if (up2tc > tc)
7559 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
7560 	}
7561 
7562 	if (reg != rsave)
7563 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
7564 
7565 	return;
7566 }
7567 
7568 /**
7569  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
7570  * @adapter: Pointer to adapter struct
7571  *
7572  * Populate the netdev user priority to tc map
7573  */
7574 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
7575 {
7576 	struct net_device *dev = adapter->netdev;
7577 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
7578 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
7579 	u8 prio;
7580 
7581 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
7582 		u8 tc = 0;
7583 
7584 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
7585 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
7586 		else if (ets)
7587 			tc = ets->prio_tc[prio];
7588 
7589 		netdev_set_prio_tc_map(dev, prio, tc);
7590 	}
7591 }
7592 
7593 #endif /* CONFIG_IXGBE_DCB */
7594 /**
7595  * ixgbe_setup_tc - configure net_device for multiple traffic classes
7596  *
7597  * @netdev: net device to configure
7598  * @tc: number of traffic classes to enable
7599  */
7600 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
7601 {
7602 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7603 	struct ixgbe_hw *hw = &adapter->hw;
7604 	bool pools;
7605 
7606 	/* Hardware supports up to 8 traffic classes */
7607 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs ||
7608 	    (hw->mac.type == ixgbe_mac_82598EB &&
7609 	     tc < MAX_TRAFFIC_CLASS))
7610 		return -EINVAL;
7611 
7612 	pools = (find_first_zero_bit(&adapter->fwd_bitmask, 32) > 1);
7613 	if (tc && pools && adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS)
7614 		return -EBUSY;
7615 
7616 	/* Hardware has to reinitialize queues and interrupts to
7617 	 * match packet buffer alignment. Unfortunately, the
7618 	 * hardware is not flexible enough to do this dynamically.
7619 	 */
7620 	if (netif_running(dev))
7621 		ixgbe_close(dev);
7622 	ixgbe_clear_interrupt_scheme(adapter);
7623 
7624 #ifdef CONFIG_IXGBE_DCB
7625 	if (tc) {
7626 		netdev_set_num_tc(dev, tc);
7627 		ixgbe_set_prio_tc_map(adapter);
7628 
7629 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
7630 
7631 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
7632 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
7633 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
7634 		}
7635 	} else {
7636 		netdev_reset_tc(dev);
7637 
7638 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
7639 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
7640 
7641 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
7642 
7643 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
7644 		adapter->dcb_cfg.pfc_mode_enable = false;
7645 	}
7646 
7647 	ixgbe_validate_rtr(adapter, tc);
7648 
7649 #endif /* CONFIG_IXGBE_DCB */
7650 	ixgbe_init_interrupt_scheme(adapter);
7651 
7652 	if (netif_running(dev))
7653 		return ixgbe_open(dev);
7654 
7655 	return 0;
7656 }
7657 
7658 #ifdef CONFIG_PCI_IOV
7659 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
7660 {
7661 	struct net_device *netdev = adapter->netdev;
7662 
7663 	rtnl_lock();
7664 	ixgbe_setup_tc(netdev, netdev_get_num_tc(netdev));
7665 	rtnl_unlock();
7666 }
7667 
7668 #endif
7669 void ixgbe_do_reset(struct net_device *netdev)
7670 {
7671 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7672 
7673 	if (netif_running(netdev))
7674 		ixgbe_reinit_locked(adapter);
7675 	else
7676 		ixgbe_reset(adapter);
7677 }
7678 
7679 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
7680 					    netdev_features_t features)
7681 {
7682 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7683 
7684 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
7685 	if (!(features & NETIF_F_RXCSUM))
7686 		features &= ~NETIF_F_LRO;
7687 
7688 	/* Turn off LRO if not RSC capable */
7689 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
7690 		features &= ~NETIF_F_LRO;
7691 
7692 	return features;
7693 }
7694 
7695 static int ixgbe_set_features(struct net_device *netdev,
7696 			      netdev_features_t features)
7697 {
7698 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7699 	netdev_features_t changed = netdev->features ^ features;
7700 	bool need_reset = false;
7701 
7702 	/* Make sure RSC matches LRO, reset if change */
7703 	if (!(features & NETIF_F_LRO)) {
7704 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
7705 			need_reset = true;
7706 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
7707 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
7708 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
7709 		if (adapter->rx_itr_setting == 1 ||
7710 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
7711 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
7712 			need_reset = true;
7713 		} else if ((changed ^ features) & NETIF_F_LRO) {
7714 			e_info(probe, "rx-usecs set too low, "
7715 			       "disabling RSC\n");
7716 		}
7717 	}
7718 
7719 	/*
7720 	 * Check if Flow Director n-tuple support was enabled or disabled.  If
7721 	 * the state changed, we need to reset.
7722 	 */
7723 	switch (features & NETIF_F_NTUPLE) {
7724 	case NETIF_F_NTUPLE:
7725 		/* turn off ATR, enable perfect filters and reset */
7726 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
7727 			need_reset = true;
7728 
7729 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
7730 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
7731 		break;
7732 	default:
7733 		/* turn off perfect filters, enable ATR and reset */
7734 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
7735 			need_reset = true;
7736 
7737 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
7738 
7739 		/* We cannot enable ATR if SR-IOV is enabled */
7740 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7741 			break;
7742 
7743 		/* We cannot enable ATR if we have 2 or more traffic classes */
7744 		if (netdev_get_num_tc(netdev) > 1)
7745 			break;
7746 
7747 		/* We cannot enable ATR if RSS is disabled */
7748 		if (adapter->ring_feature[RING_F_RSS].limit <= 1)
7749 			break;
7750 
7751 		/* A sample rate of 0 indicates ATR disabled */
7752 		if (!adapter->atr_sample_rate)
7753 			break;
7754 
7755 		adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
7756 		break;
7757 	}
7758 
7759 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
7760 		ixgbe_vlan_strip_enable(adapter);
7761 	else
7762 		ixgbe_vlan_strip_disable(adapter);
7763 
7764 	if (changed & NETIF_F_RXALL)
7765 		need_reset = true;
7766 
7767 	netdev->features = features;
7768 	if (need_reset)
7769 		ixgbe_do_reset(netdev);
7770 
7771 	return 0;
7772 }
7773 
7774 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
7775 			     struct net_device *dev,
7776 			     const unsigned char *addr, u16 vid,
7777 			     u16 flags)
7778 {
7779 	/* guarantee we can provide a unique filter for the unicast address */
7780 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
7781 		if (IXGBE_MAX_PF_MACVLANS <= netdev_uc_count(dev))
7782 			return -ENOMEM;
7783 	}
7784 
7785 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
7786 }
7787 
7788 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
7789 				    struct nlmsghdr *nlh)
7790 {
7791 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7792 	struct nlattr *attr, *br_spec;
7793 	int rem;
7794 
7795 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
7796 		return -EOPNOTSUPP;
7797 
7798 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
7799 	if (!br_spec)
7800 		return -EINVAL;
7801 
7802 	nla_for_each_nested(attr, br_spec, rem) {
7803 		__u16 mode;
7804 		u32 reg = 0;
7805 
7806 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
7807 			continue;
7808 
7809 		if (nla_len(attr) < sizeof(mode))
7810 			return -EINVAL;
7811 
7812 		mode = nla_get_u16(attr);
7813 		if (mode == BRIDGE_MODE_VEPA) {
7814 			reg = 0;
7815 			adapter->flags2 &= ~IXGBE_FLAG2_BRIDGE_MODE_VEB;
7816 		} else if (mode == BRIDGE_MODE_VEB) {
7817 			reg = IXGBE_PFDTXGSWC_VT_LBEN;
7818 			adapter->flags2 |= IXGBE_FLAG2_BRIDGE_MODE_VEB;
7819 		} else
7820 			return -EINVAL;
7821 
7822 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, reg);
7823 
7824 		e_info(drv, "enabling bridge mode: %s\n",
7825 			mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
7826 	}
7827 
7828 	return 0;
7829 }
7830 
7831 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
7832 				    struct net_device *dev,
7833 				    u32 filter_mask)
7834 {
7835 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7836 	u16 mode;
7837 
7838 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
7839 		return 0;
7840 
7841 	if (adapter->flags2 & IXGBE_FLAG2_BRIDGE_MODE_VEB)
7842 		mode = BRIDGE_MODE_VEB;
7843 	else
7844 		mode = BRIDGE_MODE_VEPA;
7845 
7846 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev, mode, 0, 0);
7847 }
7848 
7849 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
7850 {
7851 	struct ixgbe_fwd_adapter *fwd_adapter = NULL;
7852 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
7853 	int used_pools = adapter->num_vfs + adapter->num_rx_pools;
7854 	unsigned int limit;
7855 	int pool, err;
7856 
7857 	/* Hardware has a limited number of available pools. Each VF, and the
7858 	 * PF require a pool. Check to ensure we don't attempt to use more
7859 	 * then the available number of pools.
7860 	 */
7861 	if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
7862 		return ERR_PTR(-EINVAL);
7863 
7864 #ifdef CONFIG_RPS
7865 	if (vdev->num_rx_queues != vdev->num_tx_queues) {
7866 		netdev_info(pdev, "%s: Only supports a single queue count for TX and RX\n",
7867 			    vdev->name);
7868 		return ERR_PTR(-EINVAL);
7869 	}
7870 #endif
7871 	/* Check for hardware restriction on number of rx/tx queues */
7872 	if (vdev->num_tx_queues > IXGBE_MAX_L2A_QUEUES ||
7873 	    vdev->num_tx_queues == IXGBE_BAD_L2A_QUEUE) {
7874 		netdev_info(pdev,
7875 			    "%s: Supports RX/TX Queue counts 1,2, and 4\n",
7876 			    pdev->name);
7877 		return ERR_PTR(-EINVAL);
7878 	}
7879 
7880 	if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
7881 	      adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS - 1) ||
7882 	    (adapter->num_rx_pools > IXGBE_MAX_MACVLANS))
7883 		return ERR_PTR(-EBUSY);
7884 
7885 	fwd_adapter = kcalloc(1, sizeof(struct ixgbe_fwd_adapter), GFP_KERNEL);
7886 	if (!fwd_adapter)
7887 		return ERR_PTR(-ENOMEM);
7888 
7889 	pool = find_first_zero_bit(&adapter->fwd_bitmask, 32);
7890 	adapter->num_rx_pools++;
7891 	set_bit(pool, &adapter->fwd_bitmask);
7892 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
7893 
7894 	/* Enable VMDq flag so device will be set in VM mode */
7895 	adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED | IXGBE_FLAG_SRIOV_ENABLED;
7896 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
7897 	adapter->ring_feature[RING_F_RSS].limit = vdev->num_tx_queues;
7898 
7899 	/* Force reinit of ring allocation with VMDQ enabled */
7900 	err = ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
7901 	if (err)
7902 		goto fwd_add_err;
7903 	fwd_adapter->pool = pool;
7904 	fwd_adapter->real_adapter = adapter;
7905 	err = ixgbe_fwd_ring_up(vdev, fwd_adapter);
7906 	if (err)
7907 		goto fwd_add_err;
7908 	netif_tx_start_all_queues(vdev);
7909 	return fwd_adapter;
7910 fwd_add_err:
7911 	/* unwind counter and free adapter struct */
7912 	netdev_info(pdev,
7913 		    "%s: dfwd hardware acceleration failed\n", vdev->name);
7914 	clear_bit(pool, &adapter->fwd_bitmask);
7915 	adapter->num_rx_pools--;
7916 	kfree(fwd_adapter);
7917 	return ERR_PTR(err);
7918 }
7919 
7920 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
7921 {
7922 	struct ixgbe_fwd_adapter *fwd_adapter = priv;
7923 	struct ixgbe_adapter *adapter = fwd_adapter->real_adapter;
7924 	unsigned int limit;
7925 
7926 	clear_bit(fwd_adapter->pool, &adapter->fwd_bitmask);
7927 	adapter->num_rx_pools--;
7928 
7929 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
7930 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
7931 	ixgbe_fwd_ring_down(fwd_adapter->netdev, fwd_adapter);
7932 	ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
7933 	netdev_dbg(pdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
7934 		   fwd_adapter->pool, adapter->num_rx_pools,
7935 		   fwd_adapter->rx_base_queue,
7936 		   fwd_adapter->rx_base_queue + adapter->num_rx_queues_per_pool,
7937 		   adapter->fwd_bitmask);
7938 	kfree(fwd_adapter);
7939 }
7940 
7941 static const struct net_device_ops ixgbe_netdev_ops = {
7942 	.ndo_open		= ixgbe_open,
7943 	.ndo_stop		= ixgbe_close,
7944 	.ndo_start_xmit		= ixgbe_xmit_frame,
7945 	.ndo_select_queue	= ixgbe_select_queue,
7946 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
7947 	.ndo_validate_addr	= eth_validate_addr,
7948 	.ndo_set_mac_address	= ixgbe_set_mac,
7949 	.ndo_change_mtu		= ixgbe_change_mtu,
7950 	.ndo_tx_timeout		= ixgbe_tx_timeout,
7951 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
7952 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
7953 	.ndo_do_ioctl		= ixgbe_ioctl,
7954 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
7955 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
7956 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
7957 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
7958 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
7959 	.ndo_get_stats64	= ixgbe_get_stats64,
7960 #ifdef CONFIG_IXGBE_DCB
7961 	.ndo_setup_tc		= ixgbe_setup_tc,
7962 #endif
7963 #ifdef CONFIG_NET_POLL_CONTROLLER
7964 	.ndo_poll_controller	= ixgbe_netpoll,
7965 #endif
7966 #ifdef CONFIG_NET_RX_BUSY_POLL
7967 	.ndo_busy_poll		= ixgbe_low_latency_recv,
7968 #endif
7969 #ifdef IXGBE_FCOE
7970 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
7971 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
7972 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
7973 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
7974 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
7975 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
7976 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
7977 #endif /* IXGBE_FCOE */
7978 	.ndo_set_features = ixgbe_set_features,
7979 	.ndo_fix_features = ixgbe_fix_features,
7980 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
7981 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
7982 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
7983 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
7984 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
7985 };
7986 
7987 /**
7988  * ixgbe_enumerate_functions - Get the number of ports this device has
7989  * @adapter: adapter structure
7990  *
7991  * This function enumerates the phsyical functions co-located on a single slot,
7992  * in order to determine how many ports a device has. This is most useful in
7993  * determining the required GT/s of PCIe bandwidth necessary for optimal
7994  * performance.
7995  **/
7996 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
7997 {
7998 	struct pci_dev *entry, *pdev = adapter->pdev;
7999 	int physfns = 0;
8000 
8001 	/* Some cards can not use the generic count PCIe functions method,
8002 	 * because they are behind a parent switch, so we hardcode these with
8003 	 * the correct number of functions.
8004 	 */
8005 	if (ixgbe_pcie_from_parent(&adapter->hw))
8006 		physfns = 4;
8007 
8008 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
8009 		/* don't count virtual functions */
8010 		if (entry->is_virtfn)
8011 			continue;
8012 
8013 		/* When the devices on the bus don't all match our device ID,
8014 		 * we can't reliably determine the correct number of
8015 		 * functions. This can occur if a function has been direct
8016 		 * attached to a virtual machine using VT-d, for example. In
8017 		 * this case, simply return -1 to indicate this.
8018 		 */
8019 		if ((entry->vendor != pdev->vendor) ||
8020 		    (entry->device != pdev->device))
8021 			return -1;
8022 
8023 		physfns++;
8024 	}
8025 
8026 	return physfns;
8027 }
8028 
8029 /**
8030  * ixgbe_wol_supported - Check whether device supports WoL
8031  * @hw: hw specific details
8032  * @device_id: the device ID
8033  * @subdev_id: the subsystem device ID
8034  *
8035  * This function is used by probe and ethtool to determine
8036  * which devices have WoL support
8037  *
8038  **/
8039 int ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
8040 			u16 subdevice_id)
8041 {
8042 	struct ixgbe_hw *hw = &adapter->hw;
8043 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
8044 	int is_wol_supported = 0;
8045 
8046 	switch (device_id) {
8047 	case IXGBE_DEV_ID_82599_SFP:
8048 		/* Only these subdevices could supports WOL */
8049 		switch (subdevice_id) {
8050 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
8051 		case IXGBE_SUBDEV_ID_82599_560FLR:
8052 			/* only support first port */
8053 			if (hw->bus.func != 0)
8054 				break;
8055 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
8056 		case IXGBE_SUBDEV_ID_82599_SFP:
8057 		case IXGBE_SUBDEV_ID_82599_RNDC:
8058 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
8059 		case IXGBE_SUBDEV_ID_82599_LOM_SFP:
8060 			is_wol_supported = 1;
8061 			break;
8062 		}
8063 		break;
8064 	case IXGBE_DEV_ID_82599EN_SFP:
8065 		/* Only this subdevice supports WOL */
8066 		switch (subdevice_id) {
8067 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
8068 			is_wol_supported = 1;
8069 			break;
8070 		}
8071 		break;
8072 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
8073 		/* All except this subdevice support WOL */
8074 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
8075 			is_wol_supported = 1;
8076 		break;
8077 	case IXGBE_DEV_ID_82599_KX4:
8078 		is_wol_supported = 1;
8079 		break;
8080 	case IXGBE_DEV_ID_X540T:
8081 	case IXGBE_DEV_ID_X540T1:
8082 		/* check eeprom to see if enabled wol */
8083 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
8084 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
8085 		     (hw->bus.func == 0))) {
8086 			is_wol_supported = 1;
8087 		}
8088 		break;
8089 	}
8090 
8091 	return is_wol_supported;
8092 }
8093 
8094 /**
8095  * ixgbe_get_platform_mac_addr - Look up MAC address in Open Firmware / IDPROM
8096  * @adapter: Pointer to adapter struct
8097  */
8098 static void ixgbe_get_platform_mac_addr(struct ixgbe_adapter *adapter)
8099 {
8100 #ifdef CONFIG_OF
8101 	struct device_node *dp = pci_device_to_OF_node(adapter->pdev);
8102 	struct ixgbe_hw *hw = &adapter->hw;
8103 	const unsigned char *addr;
8104 
8105 	addr = of_get_mac_address(dp);
8106 	if (addr) {
8107 		ether_addr_copy(hw->mac.perm_addr, addr);
8108 		return;
8109 	}
8110 #endif /* CONFIG_OF */
8111 
8112 #ifdef CONFIG_SPARC
8113 	ether_addr_copy(hw->mac.perm_addr, idprom->id_ethaddr);
8114 #endif /* CONFIG_SPARC */
8115 }
8116 
8117 /**
8118  * ixgbe_probe - Device Initialization Routine
8119  * @pdev: PCI device information struct
8120  * @ent: entry in ixgbe_pci_tbl
8121  *
8122  * Returns 0 on success, negative on failure
8123  *
8124  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
8125  * The OS initialization, configuring of the adapter private structure,
8126  * and a hardware reset occur.
8127  **/
8128 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
8129 {
8130 	struct net_device *netdev;
8131 	struct ixgbe_adapter *adapter = NULL;
8132 	struct ixgbe_hw *hw;
8133 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
8134 	int i, err, pci_using_dac, expected_gts;
8135 	unsigned int indices = MAX_TX_QUEUES;
8136 	u8 part_str[IXGBE_PBANUM_LENGTH];
8137 	bool disable_dev = false;
8138 #ifdef IXGBE_FCOE
8139 	u16 device_caps;
8140 #endif
8141 	u32 eec;
8142 
8143 	/* Catch broken hardware that put the wrong VF device ID in
8144 	 * the PCIe SR-IOV capability.
8145 	 */
8146 	if (pdev->is_virtfn) {
8147 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
8148 		     pci_name(pdev), pdev->vendor, pdev->device);
8149 		return -EINVAL;
8150 	}
8151 
8152 	err = pci_enable_device_mem(pdev);
8153 	if (err)
8154 		return err;
8155 
8156 	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
8157 		pci_using_dac = 1;
8158 	} else {
8159 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
8160 		if (err) {
8161 			dev_err(&pdev->dev,
8162 				"No usable DMA configuration, aborting\n");
8163 			goto err_dma;
8164 		}
8165 		pci_using_dac = 0;
8166 	}
8167 
8168 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
8169 					   IORESOURCE_MEM), ixgbe_driver_name);
8170 	if (err) {
8171 		dev_err(&pdev->dev,
8172 			"pci_request_selected_regions failed 0x%x\n", err);
8173 		goto err_pci_reg;
8174 	}
8175 
8176 	pci_enable_pcie_error_reporting(pdev);
8177 
8178 	pci_set_master(pdev);
8179 	pci_save_state(pdev);
8180 
8181 	if (ii->mac == ixgbe_mac_82598EB) {
8182 #ifdef CONFIG_IXGBE_DCB
8183 		/* 8 TC w/ 4 queues per TC */
8184 		indices = 4 * MAX_TRAFFIC_CLASS;
8185 #else
8186 		indices = IXGBE_MAX_RSS_INDICES;
8187 #endif
8188 	}
8189 
8190 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
8191 	if (!netdev) {
8192 		err = -ENOMEM;
8193 		goto err_alloc_etherdev;
8194 	}
8195 
8196 	SET_NETDEV_DEV(netdev, &pdev->dev);
8197 
8198 	adapter = netdev_priv(netdev);
8199 
8200 	adapter->netdev = netdev;
8201 	adapter->pdev = pdev;
8202 	hw = &adapter->hw;
8203 	hw->back = adapter;
8204 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
8205 
8206 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
8207 			      pci_resource_len(pdev, 0));
8208 	adapter->io_addr = hw->hw_addr;
8209 	if (!hw->hw_addr) {
8210 		err = -EIO;
8211 		goto err_ioremap;
8212 	}
8213 
8214 	netdev->netdev_ops = &ixgbe_netdev_ops;
8215 	ixgbe_set_ethtool_ops(netdev);
8216 	netdev->watchdog_timeo = 5 * HZ;
8217 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
8218 
8219 	/* Setup hw api */
8220 	memcpy(&hw->mac.ops, ii->mac_ops, sizeof(hw->mac.ops));
8221 	hw->mac.type  = ii->mac;
8222 
8223 	/* EEPROM */
8224 	memcpy(&hw->eeprom.ops, ii->eeprom_ops, sizeof(hw->eeprom.ops));
8225 	eec = IXGBE_READ_REG(hw, IXGBE_EEC);
8226 	if (ixgbe_removed(hw->hw_addr)) {
8227 		err = -EIO;
8228 		goto err_ioremap;
8229 	}
8230 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
8231 	if (!(eec & (1 << 8)))
8232 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
8233 
8234 	/* PHY */
8235 	memcpy(&hw->phy.ops, ii->phy_ops, sizeof(hw->phy.ops));
8236 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
8237 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
8238 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
8239 	hw->phy.mdio.mmds = 0;
8240 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
8241 	hw->phy.mdio.dev = netdev;
8242 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
8243 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
8244 
8245 	ii->get_invariants(hw);
8246 
8247 	/* setup the private structure */
8248 	err = ixgbe_sw_init(adapter);
8249 	if (err)
8250 		goto err_sw_init;
8251 
8252 	/* Make it possible the adapter to be woken up via WOL */
8253 	switch (adapter->hw.mac.type) {
8254 	case ixgbe_mac_82599EB:
8255 	case ixgbe_mac_X540:
8256 	case ixgbe_mac_X550:
8257 	case ixgbe_mac_X550EM_x:
8258 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
8259 		break;
8260 	default:
8261 		break;
8262 	}
8263 
8264 	/*
8265 	 * If there is a fan on this device and it has failed log the
8266 	 * failure.
8267 	 */
8268 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
8269 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
8270 		if (esdp & IXGBE_ESDP_SDP1)
8271 			e_crit(probe, "Fan has stopped, replace the adapter\n");
8272 	}
8273 
8274 	if (allow_unsupported_sfp)
8275 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
8276 
8277 	/* reset_hw fills in the perm_addr as well */
8278 	hw->phy.reset_if_overtemp = true;
8279 	err = hw->mac.ops.reset_hw(hw);
8280 	hw->phy.reset_if_overtemp = false;
8281 	if (err == IXGBE_ERR_SFP_NOT_PRESENT &&
8282 	    hw->mac.type == ixgbe_mac_82598EB) {
8283 		err = 0;
8284 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
8285 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
8286 		e_dev_err("Reload the driver after installing a supported module.\n");
8287 		goto err_sw_init;
8288 	} else if (err) {
8289 		e_dev_err("HW Init failed: %d\n", err);
8290 		goto err_sw_init;
8291 	}
8292 
8293 #ifdef CONFIG_PCI_IOV
8294 	/* SR-IOV not supported on the 82598 */
8295 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
8296 		goto skip_sriov;
8297 	/* Mailbox */
8298 	ixgbe_init_mbx_params_pf(hw);
8299 	memcpy(&hw->mbx.ops, ii->mbx_ops, sizeof(hw->mbx.ops));
8300 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
8301 	ixgbe_enable_sriov(adapter);
8302 skip_sriov:
8303 
8304 #endif
8305 	netdev->features = NETIF_F_SG |
8306 			   NETIF_F_IP_CSUM |
8307 			   NETIF_F_IPV6_CSUM |
8308 			   NETIF_F_HW_VLAN_CTAG_TX |
8309 			   NETIF_F_HW_VLAN_CTAG_RX |
8310 			   NETIF_F_HW_VLAN_CTAG_FILTER |
8311 			   NETIF_F_TSO |
8312 			   NETIF_F_TSO6 |
8313 			   NETIF_F_RXHASH |
8314 			   NETIF_F_RXCSUM;
8315 
8316 	netdev->hw_features = netdev->features | NETIF_F_HW_L2FW_DOFFLOAD;
8317 
8318 	switch (adapter->hw.mac.type) {
8319 	case ixgbe_mac_82599EB:
8320 	case ixgbe_mac_X540:
8321 	case ixgbe_mac_X550:
8322 	case ixgbe_mac_X550EM_x:
8323 		netdev->features |= NETIF_F_SCTP_CSUM;
8324 		netdev->hw_features |= NETIF_F_SCTP_CSUM |
8325 				       NETIF_F_NTUPLE;
8326 		break;
8327 	default:
8328 		break;
8329 	}
8330 
8331 	netdev->hw_features |= NETIF_F_RXALL;
8332 
8333 	netdev->vlan_features |= NETIF_F_TSO;
8334 	netdev->vlan_features |= NETIF_F_TSO6;
8335 	netdev->vlan_features |= NETIF_F_IP_CSUM;
8336 	netdev->vlan_features |= NETIF_F_IPV6_CSUM;
8337 	netdev->vlan_features |= NETIF_F_SG;
8338 
8339 	netdev->priv_flags |= IFF_UNICAST_FLT;
8340 	netdev->priv_flags |= IFF_SUPP_NOFCS;
8341 
8342 #ifdef CONFIG_IXGBE_DCB
8343 	netdev->dcbnl_ops = &dcbnl_ops;
8344 #endif
8345 
8346 #ifdef IXGBE_FCOE
8347 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
8348 		unsigned int fcoe_l;
8349 
8350 		if (hw->mac.ops.get_device_caps) {
8351 			hw->mac.ops.get_device_caps(hw, &device_caps);
8352 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
8353 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
8354 		}
8355 
8356 
8357 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
8358 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
8359 
8360 		netdev->features |= NETIF_F_FSO |
8361 				    NETIF_F_FCOE_CRC;
8362 
8363 		netdev->vlan_features |= NETIF_F_FSO |
8364 					 NETIF_F_FCOE_CRC |
8365 					 NETIF_F_FCOE_MTU;
8366 	}
8367 #endif /* IXGBE_FCOE */
8368 	if (pci_using_dac) {
8369 		netdev->features |= NETIF_F_HIGHDMA;
8370 		netdev->vlan_features |= NETIF_F_HIGHDMA;
8371 	}
8372 
8373 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
8374 		netdev->hw_features |= NETIF_F_LRO;
8375 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
8376 		netdev->features |= NETIF_F_LRO;
8377 
8378 	/* make sure the EEPROM is good */
8379 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
8380 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
8381 		err = -EIO;
8382 		goto err_sw_init;
8383 	}
8384 
8385 	ixgbe_get_platform_mac_addr(adapter);
8386 
8387 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
8388 
8389 	if (!is_valid_ether_addr(netdev->dev_addr)) {
8390 		e_dev_err("invalid MAC address\n");
8391 		err = -EIO;
8392 		goto err_sw_init;
8393 	}
8394 
8395 	ixgbe_mac_set_default_filter(adapter, hw->mac.perm_addr);
8396 
8397 	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
8398 		    (unsigned long) adapter);
8399 
8400 	if (ixgbe_removed(hw->hw_addr)) {
8401 		err = -EIO;
8402 		goto err_sw_init;
8403 	}
8404 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
8405 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
8406 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
8407 
8408 	err = ixgbe_init_interrupt_scheme(adapter);
8409 	if (err)
8410 		goto err_sw_init;
8411 
8412 	/* WOL not supported for all devices */
8413 	adapter->wol = 0;
8414 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
8415 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
8416 						pdev->subsystem_device);
8417 	if (hw->wol_enabled)
8418 		adapter->wol = IXGBE_WUFC_MAG;
8419 
8420 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
8421 
8422 	/* save off EEPROM version number */
8423 	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
8424 	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
8425 
8426 	/* pick up the PCI bus settings for reporting later */
8427 	hw->mac.ops.get_bus_info(hw);
8428 	if (ixgbe_pcie_from_parent(hw))
8429 		ixgbe_get_parent_bus_info(adapter);
8430 
8431 	/* calculate the expected PCIe bandwidth required for optimal
8432 	 * performance. Note that some older parts will never have enough
8433 	 * bandwidth due to being older generation PCIe parts. We clamp these
8434 	 * parts to ensure no warning is displayed if it can't be fixed.
8435 	 */
8436 	switch (hw->mac.type) {
8437 	case ixgbe_mac_82598EB:
8438 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
8439 		break;
8440 	default:
8441 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
8442 		break;
8443 	}
8444 
8445 	/* don't check link if we failed to enumerate functions */
8446 	if (expected_gts > 0)
8447 		ixgbe_check_minimum_link(adapter, expected_gts);
8448 
8449 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
8450 	if (err)
8451 		strlcpy(part_str, "Unknown", sizeof(part_str));
8452 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
8453 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
8454 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
8455 			   part_str);
8456 	else
8457 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
8458 			   hw->mac.type, hw->phy.type, part_str);
8459 
8460 	e_dev_info("%pM\n", netdev->dev_addr);
8461 
8462 	/* reset the hardware with the new settings */
8463 	err = hw->mac.ops.start_hw(hw);
8464 	if (err == IXGBE_ERR_EEPROM_VERSION) {
8465 		/* We are running on a pre-production device, log a warning */
8466 		e_dev_warn("This device is a pre-production adapter/LOM. "
8467 			   "Please be aware there may be issues associated "
8468 			   "with your hardware.  If you are experiencing "
8469 			   "problems please contact your Intel or hardware "
8470 			   "representative who provided you with this "
8471 			   "hardware.\n");
8472 	}
8473 	strcpy(netdev->name, "eth%d");
8474 	err = register_netdev(netdev);
8475 	if (err)
8476 		goto err_register;
8477 
8478 	pci_set_drvdata(pdev, adapter);
8479 
8480 	/* power down the optics for 82599 SFP+ fiber */
8481 	if (hw->mac.ops.disable_tx_laser)
8482 		hw->mac.ops.disable_tx_laser(hw);
8483 
8484 	/* carrier off reporting is important to ethtool even BEFORE open */
8485 	netif_carrier_off(netdev);
8486 
8487 #ifdef CONFIG_IXGBE_DCA
8488 	if (dca_add_requester(&pdev->dev) == 0) {
8489 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
8490 		ixgbe_setup_dca(adapter);
8491 	}
8492 #endif
8493 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
8494 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
8495 		for (i = 0; i < adapter->num_vfs; i++)
8496 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
8497 	}
8498 
8499 	/* firmware requires driver version to be 0xFFFFFFFF
8500 	 * since os does not support feature
8501 	 */
8502 	if (hw->mac.ops.set_fw_drv_ver)
8503 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF,
8504 					   0xFF);
8505 
8506 	/* add san mac addr to netdev */
8507 	ixgbe_add_sanmac_netdev(netdev);
8508 
8509 	e_dev_info("%s\n", ixgbe_default_device_descr);
8510 
8511 #ifdef CONFIG_IXGBE_HWMON
8512 	if (ixgbe_sysfs_init(adapter))
8513 		e_err(probe, "failed to allocate sysfs resources\n");
8514 #endif /* CONFIG_IXGBE_HWMON */
8515 
8516 	ixgbe_dbg_adapter_init(adapter);
8517 
8518 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
8519 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
8520 		hw->mac.ops.setup_link(hw,
8521 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
8522 			true);
8523 
8524 	return 0;
8525 
8526 err_register:
8527 	ixgbe_release_hw_control(adapter);
8528 	ixgbe_clear_interrupt_scheme(adapter);
8529 err_sw_init:
8530 	ixgbe_disable_sriov(adapter);
8531 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
8532 	iounmap(adapter->io_addr);
8533 	kfree(adapter->mac_table);
8534 err_ioremap:
8535 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
8536 	free_netdev(netdev);
8537 err_alloc_etherdev:
8538 	pci_release_selected_regions(pdev,
8539 				     pci_select_bars(pdev, IORESOURCE_MEM));
8540 err_pci_reg:
8541 err_dma:
8542 	if (!adapter || disable_dev)
8543 		pci_disable_device(pdev);
8544 	return err;
8545 }
8546 
8547 /**
8548  * ixgbe_remove - Device Removal Routine
8549  * @pdev: PCI device information struct
8550  *
8551  * ixgbe_remove is called by the PCI subsystem to alert the driver
8552  * that it should release a PCI device.  The could be caused by a
8553  * Hot-Plug event, or because the driver is going to be removed from
8554  * memory.
8555  **/
8556 static void ixgbe_remove(struct pci_dev *pdev)
8557 {
8558 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
8559 	struct net_device *netdev;
8560 	bool disable_dev;
8561 
8562 	/* if !adapter then we already cleaned up in probe */
8563 	if (!adapter)
8564 		return;
8565 
8566 	netdev  = adapter->netdev;
8567 	ixgbe_dbg_adapter_exit(adapter);
8568 
8569 	set_bit(__IXGBE_REMOVING, &adapter->state);
8570 	cancel_work_sync(&adapter->service_task);
8571 
8572 
8573 #ifdef CONFIG_IXGBE_DCA
8574 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
8575 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
8576 		dca_remove_requester(&pdev->dev);
8577 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
8578 	}
8579 
8580 #endif
8581 #ifdef CONFIG_IXGBE_HWMON
8582 	ixgbe_sysfs_exit(adapter);
8583 #endif /* CONFIG_IXGBE_HWMON */
8584 
8585 	/* remove the added san mac */
8586 	ixgbe_del_sanmac_netdev(netdev);
8587 
8588 	if (netdev->reg_state == NETREG_REGISTERED)
8589 		unregister_netdev(netdev);
8590 
8591 #ifdef CONFIG_PCI_IOV
8592 	/*
8593 	 * Only disable SR-IOV on unload if the user specified the now
8594 	 * deprecated max_vfs module parameter.
8595 	 */
8596 	if (max_vfs)
8597 		ixgbe_disable_sriov(adapter);
8598 #endif
8599 	ixgbe_clear_interrupt_scheme(adapter);
8600 
8601 	ixgbe_release_hw_control(adapter);
8602 
8603 #ifdef CONFIG_DCB
8604 	kfree(adapter->ixgbe_ieee_pfc);
8605 	kfree(adapter->ixgbe_ieee_ets);
8606 
8607 #endif
8608 	iounmap(adapter->io_addr);
8609 	pci_release_selected_regions(pdev, pci_select_bars(pdev,
8610 				     IORESOURCE_MEM));
8611 
8612 	e_dev_info("complete\n");
8613 
8614 	kfree(adapter->mac_table);
8615 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
8616 	free_netdev(netdev);
8617 
8618 	pci_disable_pcie_error_reporting(pdev);
8619 
8620 	if (disable_dev)
8621 		pci_disable_device(pdev);
8622 }
8623 
8624 /**
8625  * ixgbe_io_error_detected - called when PCI error is detected
8626  * @pdev: Pointer to PCI device
8627  * @state: The current pci connection state
8628  *
8629  * This function is called after a PCI bus error affecting
8630  * this device has been detected.
8631  */
8632 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
8633 						pci_channel_state_t state)
8634 {
8635 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
8636 	struct net_device *netdev = adapter->netdev;
8637 
8638 #ifdef CONFIG_PCI_IOV
8639 	struct ixgbe_hw *hw = &adapter->hw;
8640 	struct pci_dev *bdev, *vfdev;
8641 	u32 dw0, dw1, dw2, dw3;
8642 	int vf, pos;
8643 	u16 req_id, pf_func;
8644 
8645 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
8646 	    adapter->num_vfs == 0)
8647 		goto skip_bad_vf_detection;
8648 
8649 	bdev = pdev->bus->self;
8650 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
8651 		bdev = bdev->bus->self;
8652 
8653 	if (!bdev)
8654 		goto skip_bad_vf_detection;
8655 
8656 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
8657 	if (!pos)
8658 		goto skip_bad_vf_detection;
8659 
8660 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
8661 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
8662 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
8663 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
8664 	if (ixgbe_removed(hw->hw_addr))
8665 		goto skip_bad_vf_detection;
8666 
8667 	req_id = dw1 >> 16;
8668 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
8669 	if (!(req_id & 0x0080))
8670 		goto skip_bad_vf_detection;
8671 
8672 	pf_func = req_id & 0x01;
8673 	if ((pf_func & 1) == (pdev->devfn & 1)) {
8674 		unsigned int device_id;
8675 
8676 		vf = (req_id & 0x7F) >> 1;
8677 		e_dev_err("VF %d has caused a PCIe error\n", vf);
8678 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
8679 				"%8.8x\tdw3: %8.8x\n",
8680 		dw0, dw1, dw2, dw3);
8681 		switch (adapter->hw.mac.type) {
8682 		case ixgbe_mac_82599EB:
8683 			device_id = IXGBE_82599_VF_DEVICE_ID;
8684 			break;
8685 		case ixgbe_mac_X540:
8686 			device_id = IXGBE_X540_VF_DEVICE_ID;
8687 			break;
8688 		case ixgbe_mac_X550:
8689 			device_id = IXGBE_DEV_ID_X550_VF;
8690 			break;
8691 		case ixgbe_mac_X550EM_x:
8692 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
8693 			break;
8694 		default:
8695 			device_id = 0;
8696 			break;
8697 		}
8698 
8699 		/* Find the pci device of the offending VF */
8700 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
8701 		while (vfdev) {
8702 			if (vfdev->devfn == (req_id & 0xFF))
8703 				break;
8704 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
8705 					       device_id, vfdev);
8706 		}
8707 		/*
8708 		 * There's a slim chance the VF could have been hot plugged,
8709 		 * so if it is no longer present we don't need to issue the
8710 		 * VFLR.  Just clean up the AER in that case.
8711 		 */
8712 		if (vfdev) {
8713 			ixgbe_issue_vf_flr(adapter, vfdev);
8714 			/* Free device reference count */
8715 			pci_dev_put(vfdev);
8716 		}
8717 
8718 		pci_cleanup_aer_uncorrect_error_status(pdev);
8719 	}
8720 
8721 	/*
8722 	 * Even though the error may have occurred on the other port
8723 	 * we still need to increment the vf error reference count for
8724 	 * both ports because the I/O resume function will be called
8725 	 * for both of them.
8726 	 */
8727 	adapter->vferr_refcount++;
8728 
8729 	return PCI_ERS_RESULT_RECOVERED;
8730 
8731 skip_bad_vf_detection:
8732 #endif /* CONFIG_PCI_IOV */
8733 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
8734 		return PCI_ERS_RESULT_DISCONNECT;
8735 
8736 	rtnl_lock();
8737 	netif_device_detach(netdev);
8738 
8739 	if (state == pci_channel_io_perm_failure) {
8740 		rtnl_unlock();
8741 		return PCI_ERS_RESULT_DISCONNECT;
8742 	}
8743 
8744 	if (netif_running(netdev))
8745 		ixgbe_down(adapter);
8746 
8747 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
8748 		pci_disable_device(pdev);
8749 	rtnl_unlock();
8750 
8751 	/* Request a slot reset. */
8752 	return PCI_ERS_RESULT_NEED_RESET;
8753 }
8754 
8755 /**
8756  * ixgbe_io_slot_reset - called after the pci bus has been reset.
8757  * @pdev: Pointer to PCI device
8758  *
8759  * Restart the card from scratch, as if from a cold-boot.
8760  */
8761 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
8762 {
8763 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
8764 	pci_ers_result_t result;
8765 	int err;
8766 
8767 	if (pci_enable_device_mem(pdev)) {
8768 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
8769 		result = PCI_ERS_RESULT_DISCONNECT;
8770 	} else {
8771 		smp_mb__before_atomic();
8772 		clear_bit(__IXGBE_DISABLED, &adapter->state);
8773 		adapter->hw.hw_addr = adapter->io_addr;
8774 		pci_set_master(pdev);
8775 		pci_restore_state(pdev);
8776 		pci_save_state(pdev);
8777 
8778 		pci_wake_from_d3(pdev, false);
8779 
8780 		ixgbe_reset(adapter);
8781 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
8782 		result = PCI_ERS_RESULT_RECOVERED;
8783 	}
8784 
8785 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
8786 	if (err) {
8787 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
8788 			  "failed 0x%0x\n", err);
8789 		/* non-fatal, continue */
8790 	}
8791 
8792 	return result;
8793 }
8794 
8795 /**
8796  * ixgbe_io_resume - called when traffic can start flowing again.
8797  * @pdev: Pointer to PCI device
8798  *
8799  * This callback is called when the error recovery driver tells us that
8800  * its OK to resume normal operation.
8801  */
8802 static void ixgbe_io_resume(struct pci_dev *pdev)
8803 {
8804 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
8805 	struct net_device *netdev = adapter->netdev;
8806 
8807 #ifdef CONFIG_PCI_IOV
8808 	if (adapter->vferr_refcount) {
8809 		e_info(drv, "Resuming after VF err\n");
8810 		adapter->vferr_refcount--;
8811 		return;
8812 	}
8813 
8814 #endif
8815 	if (netif_running(netdev))
8816 		ixgbe_up(adapter);
8817 
8818 	netif_device_attach(netdev);
8819 }
8820 
8821 static const struct pci_error_handlers ixgbe_err_handler = {
8822 	.error_detected = ixgbe_io_error_detected,
8823 	.slot_reset = ixgbe_io_slot_reset,
8824 	.resume = ixgbe_io_resume,
8825 };
8826 
8827 static struct pci_driver ixgbe_driver = {
8828 	.name     = ixgbe_driver_name,
8829 	.id_table = ixgbe_pci_tbl,
8830 	.probe    = ixgbe_probe,
8831 	.remove   = ixgbe_remove,
8832 #ifdef CONFIG_PM
8833 	.suspend  = ixgbe_suspend,
8834 	.resume   = ixgbe_resume,
8835 #endif
8836 	.shutdown = ixgbe_shutdown,
8837 	.sriov_configure = ixgbe_pci_sriov_configure,
8838 	.err_handler = &ixgbe_err_handler
8839 };
8840 
8841 /**
8842  * ixgbe_init_module - Driver Registration Routine
8843  *
8844  * ixgbe_init_module is the first routine called when the driver is
8845  * loaded. All it does is register with the PCI subsystem.
8846  **/
8847 static int __init ixgbe_init_module(void)
8848 {
8849 	int ret;
8850 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
8851 	pr_info("%s\n", ixgbe_copyright);
8852 
8853 	ixgbe_dbg_init();
8854 
8855 	ret = pci_register_driver(&ixgbe_driver);
8856 	if (ret) {
8857 		ixgbe_dbg_exit();
8858 		return ret;
8859 	}
8860 
8861 #ifdef CONFIG_IXGBE_DCA
8862 	dca_register_notify(&dca_notifier);
8863 #endif
8864 
8865 	return 0;
8866 }
8867 
8868 module_init(ixgbe_init_module);
8869 
8870 /**
8871  * ixgbe_exit_module - Driver Exit Cleanup Routine
8872  *
8873  * ixgbe_exit_module is called just before the driver is removed
8874  * from memory.
8875  **/
8876 static void __exit ixgbe_exit_module(void)
8877 {
8878 #ifdef CONFIG_IXGBE_DCA
8879 	dca_unregister_notify(&dca_notifier);
8880 #endif
8881 	pci_unregister_driver(&ixgbe_driver);
8882 
8883 	ixgbe_dbg_exit();
8884 
8885 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
8886 }
8887 
8888 #ifdef CONFIG_IXGBE_DCA
8889 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
8890 			    void *p)
8891 {
8892 	int ret_val;
8893 
8894 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
8895 					 __ixgbe_notify_dca);
8896 
8897 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
8898 }
8899 
8900 #endif /* CONFIG_IXGBE_DCA */
8901 
8902 module_exit(ixgbe_exit_module);
8903 
8904 /* ixgbe_main.c */
8905