xref: /openbmc/linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision 2d33394e23d63b750dcba40e5feaeba425427b52)
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 #include <net/vxlan.h>
54 
55 #ifdef CONFIG_OF
56 #include <linux/of_net.h>
57 #endif
58 
59 #ifdef CONFIG_SPARC
60 #include <asm/idprom.h>
61 #include <asm/prom.h>
62 #endif
63 
64 #include "ixgbe.h"
65 #include "ixgbe_common.h"
66 #include "ixgbe_dcb_82599.h"
67 #include "ixgbe_sriov.h"
68 
69 char ixgbe_driver_name[] = "ixgbe";
70 static const char ixgbe_driver_string[] =
71 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
72 #ifdef IXGBE_FCOE
73 char ixgbe_default_device_descr[] =
74 			      "Intel(R) 10 Gigabit Network Connection";
75 #else
76 static char ixgbe_default_device_descr[] =
77 			      "Intel(R) 10 Gigabit Network Connection";
78 #endif
79 #define DRV_VERSION "4.0.1-k"
80 const char ixgbe_driver_version[] = DRV_VERSION;
81 static const char ixgbe_copyright[] =
82 				"Copyright (c) 1999-2014 Intel Corporation.";
83 
84 static const struct ixgbe_info *ixgbe_info_tbl[] = {
85 	[board_82598]		= &ixgbe_82598_info,
86 	[board_82599]		= &ixgbe_82599_info,
87 	[board_X540]		= &ixgbe_X540_info,
88 	[board_X550]		= &ixgbe_X550_info,
89 	[board_X550EM_x]	= &ixgbe_X550EM_x_info,
90 };
91 
92 /* ixgbe_pci_tbl - PCI Device ID Table
93  *
94  * Wildcard entries (PCI_ANY_ID) should come last
95  * Last entry must be all 0s
96  *
97  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
98  *   Class, Class Mask, private data (not used) }
99  */
100 static const struct pci_device_id ixgbe_pci_tbl[] = {
101 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
102 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
103 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
104 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
105 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
115 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
116 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
117 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
118 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
119 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
120 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
121 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
122 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
123 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
124 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
125 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
126 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
127 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
128 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
129 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
130 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
131 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
132 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
133 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
134 	/* required last entry */
135 	{0, }
136 };
137 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
138 
139 #ifdef CONFIG_IXGBE_DCA
140 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
141 			    void *p);
142 static struct notifier_block dca_notifier = {
143 	.notifier_call = ixgbe_notify_dca,
144 	.next          = NULL,
145 	.priority      = 0
146 };
147 #endif
148 
149 #ifdef CONFIG_PCI_IOV
150 static unsigned int max_vfs;
151 module_param(max_vfs, uint, 0);
152 MODULE_PARM_DESC(max_vfs,
153 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
154 #endif /* CONFIG_PCI_IOV */
155 
156 static unsigned int allow_unsupported_sfp;
157 module_param(allow_unsupported_sfp, uint, 0);
158 MODULE_PARM_DESC(allow_unsupported_sfp,
159 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
160 
161 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
162 static int debug = -1;
163 module_param(debug, int, 0);
164 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
165 
166 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
167 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
168 MODULE_LICENSE("GPL");
169 MODULE_VERSION(DRV_VERSION);
170 
171 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
172 
173 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
174 					  u32 reg, u16 *value)
175 {
176 	struct pci_dev *parent_dev;
177 	struct pci_bus *parent_bus;
178 
179 	parent_bus = adapter->pdev->bus->parent;
180 	if (!parent_bus)
181 		return -1;
182 
183 	parent_dev = parent_bus->self;
184 	if (!parent_dev)
185 		return -1;
186 
187 	if (!pci_is_pcie(parent_dev))
188 		return -1;
189 
190 	pcie_capability_read_word(parent_dev, reg, value);
191 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
192 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
193 		return -1;
194 	return 0;
195 }
196 
197 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
198 {
199 	struct ixgbe_hw *hw = &adapter->hw;
200 	u16 link_status = 0;
201 	int err;
202 
203 	hw->bus.type = ixgbe_bus_type_pci_express;
204 
205 	/* Get the negotiated link width and speed from PCI config space of the
206 	 * parent, as this device is behind a switch
207 	 */
208 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
209 
210 	/* assume caller will handle error case */
211 	if (err)
212 		return err;
213 
214 	hw->bus.width = ixgbe_convert_bus_width(link_status);
215 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
216 
217 	return 0;
218 }
219 
220 /**
221  * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
222  * @hw: hw specific details
223  *
224  * This function is used by probe to determine whether a device's PCI-Express
225  * bandwidth details should be gathered from the parent bus instead of from the
226  * device. Used to ensure that various locations all have the correct device ID
227  * checks.
228  */
229 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
230 {
231 	switch (hw->device_id) {
232 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
233 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
234 		return true;
235 	default:
236 		return false;
237 	}
238 }
239 
240 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
241 				     int expected_gts)
242 {
243 	int max_gts = 0;
244 	enum pci_bus_speed speed = PCI_SPEED_UNKNOWN;
245 	enum pcie_link_width width = PCIE_LNK_WIDTH_UNKNOWN;
246 	struct pci_dev *pdev;
247 
248 	/* determine whether to use the the parent device
249 	 */
250 	if (ixgbe_pcie_from_parent(&adapter->hw))
251 		pdev = adapter->pdev->bus->parent->self;
252 	else
253 		pdev = adapter->pdev;
254 
255 	if (pcie_get_minimum_link(pdev, &speed, &width) ||
256 	    speed == PCI_SPEED_UNKNOWN || width == PCIE_LNK_WIDTH_UNKNOWN) {
257 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
258 		return;
259 	}
260 
261 	switch (speed) {
262 	case PCIE_SPEED_2_5GT:
263 		/* 8b/10b encoding reduces max throughput by 20% */
264 		max_gts = 2 * width;
265 		break;
266 	case PCIE_SPEED_5_0GT:
267 		/* 8b/10b encoding reduces max throughput by 20% */
268 		max_gts = 4 * width;
269 		break;
270 	case PCIE_SPEED_8_0GT:
271 		/* 128b/130b encoding reduces throughput by less than 2% */
272 		max_gts = 8 * width;
273 		break;
274 	default:
275 		e_dev_warn("Unable to determine PCI Express bandwidth.\n");
276 		return;
277 	}
278 
279 	e_dev_info("PCI Express bandwidth of %dGT/s available\n",
280 		   max_gts);
281 	e_dev_info("(Speed:%s, Width: x%d, Encoding Loss:%s)\n",
282 		   (speed == PCIE_SPEED_8_0GT ? "8.0GT/s" :
283 		    speed == PCIE_SPEED_5_0GT ? "5.0GT/s" :
284 		    speed == PCIE_SPEED_2_5GT ? "2.5GT/s" :
285 		    "Unknown"),
286 		   width,
287 		   (speed == PCIE_SPEED_2_5GT ? "20%" :
288 		    speed == PCIE_SPEED_5_0GT ? "20%" :
289 		    speed == PCIE_SPEED_8_0GT ? "<2%" :
290 		    "Unknown"));
291 
292 	if (max_gts < expected_gts) {
293 		e_dev_warn("This is not sufficient for optimal performance of this card.\n");
294 		e_dev_warn("For optimal performance, at least %dGT/s of bandwidth is required.\n",
295 			expected_gts);
296 		e_dev_warn("A slot with more lanes and/or higher speed is suggested.\n");
297 	}
298 }
299 
300 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
301 {
302 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
303 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
304 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
305 		schedule_work(&adapter->service_task);
306 }
307 
308 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
309 {
310 	struct ixgbe_adapter *adapter = hw->back;
311 
312 	if (!hw->hw_addr)
313 		return;
314 	hw->hw_addr = NULL;
315 	e_dev_err("Adapter removed\n");
316 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
317 		ixgbe_service_event_schedule(adapter);
318 }
319 
320 static void ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
321 {
322 	u32 value;
323 
324 	/* The following check not only optimizes a bit by not
325 	 * performing a read on the status register when the
326 	 * register just read was a status register read that
327 	 * returned IXGBE_FAILED_READ_REG. It also blocks any
328 	 * potential recursion.
329 	 */
330 	if (reg == IXGBE_STATUS) {
331 		ixgbe_remove_adapter(hw);
332 		return;
333 	}
334 	value = ixgbe_read_reg(hw, IXGBE_STATUS);
335 	if (value == IXGBE_FAILED_READ_REG)
336 		ixgbe_remove_adapter(hw);
337 }
338 
339 /**
340  * ixgbe_read_reg - Read from device register
341  * @hw: hw specific details
342  * @reg: offset of register to read
343  *
344  * Returns : value read or IXGBE_FAILED_READ_REG if removed
345  *
346  * This function is used to read device registers. It checks for device
347  * removal by confirming any read that returns all ones by checking the
348  * status register value for all ones. This function avoids reading from
349  * the hardware if a removal was previously detected in which case it
350  * returns IXGBE_FAILED_READ_REG (all ones).
351  */
352 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
353 {
354 	u8 __iomem *reg_addr = ACCESS_ONCE(hw->hw_addr);
355 	u32 value;
356 
357 	if (ixgbe_removed(reg_addr))
358 		return IXGBE_FAILED_READ_REG;
359 	value = readl(reg_addr + reg);
360 	if (unlikely(value == IXGBE_FAILED_READ_REG))
361 		ixgbe_check_remove(hw, reg);
362 	return value;
363 }
364 
365 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
366 {
367 	u16 value;
368 
369 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
370 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
371 		ixgbe_remove_adapter(hw);
372 		return true;
373 	}
374 	return false;
375 }
376 
377 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
378 {
379 	struct ixgbe_adapter *adapter = hw->back;
380 	u16 value;
381 
382 	if (ixgbe_removed(hw->hw_addr))
383 		return IXGBE_FAILED_READ_CFG_WORD;
384 	pci_read_config_word(adapter->pdev, reg, &value);
385 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
386 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
387 		return IXGBE_FAILED_READ_CFG_WORD;
388 	return value;
389 }
390 
391 #ifdef CONFIG_PCI_IOV
392 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
393 {
394 	struct ixgbe_adapter *adapter = hw->back;
395 	u32 value;
396 
397 	if (ixgbe_removed(hw->hw_addr))
398 		return IXGBE_FAILED_READ_CFG_DWORD;
399 	pci_read_config_dword(adapter->pdev, reg, &value);
400 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
401 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
402 		return IXGBE_FAILED_READ_CFG_DWORD;
403 	return value;
404 }
405 #endif /* CONFIG_PCI_IOV */
406 
407 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
408 {
409 	struct ixgbe_adapter *adapter = hw->back;
410 
411 	if (ixgbe_removed(hw->hw_addr))
412 		return;
413 	pci_write_config_word(adapter->pdev, reg, value);
414 }
415 
416 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
417 {
418 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
419 
420 	/* flush memory to make sure state is correct before next watchdog */
421 	smp_mb__before_atomic();
422 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
423 }
424 
425 struct ixgbe_reg_info {
426 	u32 ofs;
427 	char *name;
428 };
429 
430 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
431 
432 	/* General Registers */
433 	{IXGBE_CTRL, "CTRL"},
434 	{IXGBE_STATUS, "STATUS"},
435 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
436 
437 	/* Interrupt Registers */
438 	{IXGBE_EICR, "EICR"},
439 
440 	/* RX Registers */
441 	{IXGBE_SRRCTL(0), "SRRCTL"},
442 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
443 	{IXGBE_RDLEN(0), "RDLEN"},
444 	{IXGBE_RDH(0), "RDH"},
445 	{IXGBE_RDT(0), "RDT"},
446 	{IXGBE_RXDCTL(0), "RXDCTL"},
447 	{IXGBE_RDBAL(0), "RDBAL"},
448 	{IXGBE_RDBAH(0), "RDBAH"},
449 
450 	/* TX Registers */
451 	{IXGBE_TDBAL(0), "TDBAL"},
452 	{IXGBE_TDBAH(0), "TDBAH"},
453 	{IXGBE_TDLEN(0), "TDLEN"},
454 	{IXGBE_TDH(0), "TDH"},
455 	{IXGBE_TDT(0), "TDT"},
456 	{IXGBE_TXDCTL(0), "TXDCTL"},
457 
458 	/* List Terminator */
459 	{ .name = NULL }
460 };
461 
462 
463 /*
464  * ixgbe_regdump - register printout routine
465  */
466 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
467 {
468 	int i = 0, j = 0;
469 	char rname[16];
470 	u32 regs[64];
471 
472 	switch (reginfo->ofs) {
473 	case IXGBE_SRRCTL(0):
474 		for (i = 0; i < 64; i++)
475 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
476 		break;
477 	case IXGBE_DCA_RXCTRL(0):
478 		for (i = 0; i < 64; i++)
479 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
480 		break;
481 	case IXGBE_RDLEN(0):
482 		for (i = 0; i < 64; i++)
483 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
484 		break;
485 	case IXGBE_RDH(0):
486 		for (i = 0; i < 64; i++)
487 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
488 		break;
489 	case IXGBE_RDT(0):
490 		for (i = 0; i < 64; i++)
491 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
492 		break;
493 	case IXGBE_RXDCTL(0):
494 		for (i = 0; i < 64; i++)
495 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
496 		break;
497 	case IXGBE_RDBAL(0):
498 		for (i = 0; i < 64; i++)
499 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
500 		break;
501 	case IXGBE_RDBAH(0):
502 		for (i = 0; i < 64; i++)
503 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
504 		break;
505 	case IXGBE_TDBAL(0):
506 		for (i = 0; i < 64; i++)
507 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
508 		break;
509 	case IXGBE_TDBAH(0):
510 		for (i = 0; i < 64; i++)
511 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
512 		break;
513 	case IXGBE_TDLEN(0):
514 		for (i = 0; i < 64; i++)
515 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
516 		break;
517 	case IXGBE_TDH(0):
518 		for (i = 0; i < 64; i++)
519 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
520 		break;
521 	case IXGBE_TDT(0):
522 		for (i = 0; i < 64; i++)
523 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
524 		break;
525 	case IXGBE_TXDCTL(0):
526 		for (i = 0; i < 64; i++)
527 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
528 		break;
529 	default:
530 		pr_info("%-15s %08x\n", reginfo->name,
531 			IXGBE_READ_REG(hw, reginfo->ofs));
532 		return;
533 	}
534 
535 	for (i = 0; i < 8; i++) {
536 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i*8, i*8+7);
537 		pr_err("%-15s", rname);
538 		for (j = 0; j < 8; j++)
539 			pr_cont(" %08x", regs[i*8+j]);
540 		pr_cont("\n");
541 	}
542 
543 }
544 
545 /*
546  * ixgbe_dump - Print registers, tx-rings and rx-rings
547  */
548 static void ixgbe_dump(struct ixgbe_adapter *adapter)
549 {
550 	struct net_device *netdev = adapter->netdev;
551 	struct ixgbe_hw *hw = &adapter->hw;
552 	struct ixgbe_reg_info *reginfo;
553 	int n = 0;
554 	struct ixgbe_ring *tx_ring;
555 	struct ixgbe_tx_buffer *tx_buffer;
556 	union ixgbe_adv_tx_desc *tx_desc;
557 	struct my_u0 { u64 a; u64 b; } *u0;
558 	struct ixgbe_ring *rx_ring;
559 	union ixgbe_adv_rx_desc *rx_desc;
560 	struct ixgbe_rx_buffer *rx_buffer_info;
561 	u32 staterr;
562 	int i = 0;
563 
564 	if (!netif_msg_hw(adapter))
565 		return;
566 
567 	/* Print netdevice Info */
568 	if (netdev) {
569 		dev_info(&adapter->pdev->dev, "Net device Info\n");
570 		pr_info("Device Name     state            "
571 			"trans_start      last_rx\n");
572 		pr_info("%-15s %016lX %016lX %016lX\n",
573 			netdev->name,
574 			netdev->state,
575 			netdev->trans_start,
576 			netdev->last_rx);
577 	}
578 
579 	/* Print Registers */
580 	dev_info(&adapter->pdev->dev, "Register Dump\n");
581 	pr_info(" Register Name   Value\n");
582 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
583 	     reginfo->name; reginfo++) {
584 		ixgbe_regdump(hw, reginfo);
585 	}
586 
587 	/* Print TX Ring Summary */
588 	if (!netdev || !netif_running(netdev))
589 		return;
590 
591 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
592 	pr_info(" %s     %s              %s        %s\n",
593 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
594 		"leng", "ntw", "timestamp");
595 	for (n = 0; n < adapter->num_tx_queues; n++) {
596 		tx_ring = adapter->tx_ring[n];
597 		tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
598 		pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
599 			   n, tx_ring->next_to_use, tx_ring->next_to_clean,
600 			   (u64)dma_unmap_addr(tx_buffer, dma),
601 			   dma_unmap_len(tx_buffer, len),
602 			   tx_buffer->next_to_watch,
603 			   (u64)tx_buffer->time_stamp);
604 	}
605 
606 	/* Print TX Rings */
607 	if (!netif_msg_tx_done(adapter))
608 		goto rx_ring_summary;
609 
610 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
611 
612 	/* Transmit Descriptor Formats
613 	 *
614 	 * 82598 Advanced Transmit Descriptor
615 	 *   +--------------------------------------------------------------+
616 	 * 0 |         Buffer Address [63:0]                                |
617 	 *   +--------------------------------------------------------------+
618 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
619 	 *   +--------------------------------------------------------------+
620 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
621 	 *
622 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
623 	 *   +--------------------------------------------------------------+
624 	 * 0 |                          RSV [63:0]                          |
625 	 *   +--------------------------------------------------------------+
626 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
627 	 *   +--------------------------------------------------------------+
628 	 *   63                       36 35   32 31                         0
629 	 *
630 	 * 82599+ Advanced Transmit Descriptor
631 	 *   +--------------------------------------------------------------+
632 	 * 0 |         Buffer Address [63:0]                                |
633 	 *   +--------------------------------------------------------------+
634 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
635 	 *   +--------------------------------------------------------------+
636 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
637 	 *
638 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
639 	 *   +--------------------------------------------------------------+
640 	 * 0 |                          RSV [63:0]                          |
641 	 *   +--------------------------------------------------------------+
642 	 * 8 |            RSV           |  STA  |           RSV             |
643 	 *   +--------------------------------------------------------------+
644 	 *   63                       36 35   32 31                         0
645 	 */
646 
647 	for (n = 0; n < adapter->num_tx_queues; n++) {
648 		tx_ring = adapter->tx_ring[n];
649 		pr_info("------------------------------------\n");
650 		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
651 		pr_info("------------------------------------\n");
652 		pr_info("%s%s    %s              %s        %s          %s\n",
653 			"T [desc]     [address 63:0  ] ",
654 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
655 			"leng", "ntw", "timestamp", "bi->skb");
656 
657 		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
658 			tx_desc = IXGBE_TX_DESC(tx_ring, i);
659 			tx_buffer = &tx_ring->tx_buffer_info[i];
660 			u0 = (struct my_u0 *)tx_desc;
661 			if (dma_unmap_len(tx_buffer, len) > 0) {
662 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p",
663 					i,
664 					le64_to_cpu(u0->a),
665 					le64_to_cpu(u0->b),
666 					(u64)dma_unmap_addr(tx_buffer, dma),
667 					dma_unmap_len(tx_buffer, len),
668 					tx_buffer->next_to_watch,
669 					(u64)tx_buffer->time_stamp,
670 					tx_buffer->skb);
671 				if (i == tx_ring->next_to_use &&
672 					i == tx_ring->next_to_clean)
673 					pr_cont(" NTC/U\n");
674 				else if (i == tx_ring->next_to_use)
675 					pr_cont(" NTU\n");
676 				else if (i == tx_ring->next_to_clean)
677 					pr_cont(" NTC\n");
678 				else
679 					pr_cont("\n");
680 
681 				if (netif_msg_pktdata(adapter) &&
682 				    tx_buffer->skb)
683 					print_hex_dump(KERN_INFO, "",
684 						DUMP_PREFIX_ADDRESS, 16, 1,
685 						tx_buffer->skb->data,
686 						dma_unmap_len(tx_buffer, len),
687 						true);
688 			}
689 		}
690 	}
691 
692 	/* Print RX Rings Summary */
693 rx_ring_summary:
694 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
695 	pr_info("Queue [NTU] [NTC]\n");
696 	for (n = 0; n < adapter->num_rx_queues; n++) {
697 		rx_ring = adapter->rx_ring[n];
698 		pr_info("%5d %5X %5X\n",
699 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
700 	}
701 
702 	/* Print RX Rings */
703 	if (!netif_msg_rx_status(adapter))
704 		return;
705 
706 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
707 
708 	/* Receive Descriptor Formats
709 	 *
710 	 * 82598 Advanced Receive Descriptor (Read) Format
711 	 *    63                                           1        0
712 	 *    +-----------------------------------------------------+
713 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
714 	 *    +----------------------------------------------+------+
715 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
716 	 *    +-----------------------------------------------------+
717 	 *
718 	 *
719 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
720 	 *
721 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
722 	 *   +------------------------------------------------------+
723 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
724 	 *   | Packet   | IP     |   |          |     | Type | Type |
725 	 *   | Checksum | Ident  |   |          |     |      |      |
726 	 *   +------------------------------------------------------+
727 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
728 	 *   +------------------------------------------------------+
729 	 *   63       48 47    32 31            20 19               0
730 	 *
731 	 * 82599+ Advanced Receive Descriptor (Read) Format
732 	 *    63                                           1        0
733 	 *    +-----------------------------------------------------+
734 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
735 	 *    +----------------------------------------------+------+
736 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
737 	 *    +-----------------------------------------------------+
738 	 *
739 	 *
740 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
741 	 *
742 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
743 	 *   +------------------------------------------------------+
744 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
745 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
746 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
747 	 *   +------------------------------------------------------+
748 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
749 	 *   +------------------------------------------------------+
750 	 *   63       48 47    32 31          20 19                 0
751 	 */
752 
753 	for (n = 0; n < adapter->num_rx_queues; n++) {
754 		rx_ring = adapter->rx_ring[n];
755 		pr_info("------------------------------------\n");
756 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
757 		pr_info("------------------------------------\n");
758 		pr_info("%s%s%s",
759 			"R  [desc]      [ PktBuf     A0] ",
760 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
761 			"<-- Adv Rx Read format\n");
762 		pr_info("%s%s%s",
763 			"RWB[desc]      [PcsmIpSHl PtRs] ",
764 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
765 			"<-- Adv Rx Write-Back format\n");
766 
767 		for (i = 0; i < rx_ring->count; i++) {
768 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
769 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
770 			u0 = (struct my_u0 *)rx_desc;
771 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
772 			if (staterr & IXGBE_RXD_STAT_DD) {
773 				/* Descriptor Done */
774 				pr_info("RWB[0x%03X]     %016llX "
775 					"%016llX ---------------- %p", i,
776 					le64_to_cpu(u0->a),
777 					le64_to_cpu(u0->b),
778 					rx_buffer_info->skb);
779 			} else {
780 				pr_info("R  [0x%03X]     %016llX "
781 					"%016llX %016llX %p", i,
782 					le64_to_cpu(u0->a),
783 					le64_to_cpu(u0->b),
784 					(u64)rx_buffer_info->dma,
785 					rx_buffer_info->skb);
786 
787 				if (netif_msg_pktdata(adapter) &&
788 				    rx_buffer_info->dma) {
789 					print_hex_dump(KERN_INFO, "",
790 					   DUMP_PREFIX_ADDRESS, 16, 1,
791 					   page_address(rx_buffer_info->page) +
792 						    rx_buffer_info->page_offset,
793 					   ixgbe_rx_bufsz(rx_ring), true);
794 				}
795 			}
796 
797 			if (i == rx_ring->next_to_use)
798 				pr_cont(" NTU\n");
799 			else if (i == rx_ring->next_to_clean)
800 				pr_cont(" NTC\n");
801 			else
802 				pr_cont("\n");
803 
804 		}
805 	}
806 }
807 
808 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
809 {
810 	u32 ctrl_ext;
811 
812 	/* Let firmware take over control of h/w */
813 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
814 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
815 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
816 }
817 
818 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
819 {
820 	u32 ctrl_ext;
821 
822 	/* Let firmware know the driver has taken over */
823 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
824 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
825 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
826 }
827 
828 /**
829  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
830  * @adapter: pointer to adapter struct
831  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
832  * @queue: queue to map the corresponding interrupt to
833  * @msix_vector: the vector to map to the corresponding queue
834  *
835  */
836 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
837 			   u8 queue, u8 msix_vector)
838 {
839 	u32 ivar, index;
840 	struct ixgbe_hw *hw = &adapter->hw;
841 	switch (hw->mac.type) {
842 	case ixgbe_mac_82598EB:
843 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
844 		if (direction == -1)
845 			direction = 0;
846 		index = (((direction * 64) + queue) >> 2) & 0x1F;
847 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
848 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
849 		ivar |= (msix_vector << (8 * (queue & 0x3)));
850 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
851 		break;
852 	case ixgbe_mac_82599EB:
853 	case ixgbe_mac_X540:
854 	case ixgbe_mac_X550:
855 	case ixgbe_mac_X550EM_x:
856 		if (direction == -1) {
857 			/* other causes */
858 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
859 			index = ((queue & 1) * 8);
860 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
861 			ivar &= ~(0xFF << index);
862 			ivar |= (msix_vector << index);
863 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
864 			break;
865 		} else {
866 			/* tx or rx causes */
867 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
868 			index = ((16 * (queue & 1)) + (8 * direction));
869 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
870 			ivar &= ~(0xFF << index);
871 			ivar |= (msix_vector << index);
872 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
873 			break;
874 		}
875 	default:
876 		break;
877 	}
878 }
879 
880 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
881 					  u64 qmask)
882 {
883 	u32 mask;
884 
885 	switch (adapter->hw.mac.type) {
886 	case ixgbe_mac_82598EB:
887 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
888 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
889 		break;
890 	case ixgbe_mac_82599EB:
891 	case ixgbe_mac_X540:
892 	case ixgbe_mac_X550:
893 	case ixgbe_mac_X550EM_x:
894 		mask = (qmask & 0xFFFFFFFF);
895 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
896 		mask = (qmask >> 32);
897 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
898 		break;
899 	default:
900 		break;
901 	}
902 }
903 
904 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring *ring,
905 				      struct ixgbe_tx_buffer *tx_buffer)
906 {
907 	if (tx_buffer->skb) {
908 		dev_kfree_skb_any(tx_buffer->skb);
909 		if (dma_unmap_len(tx_buffer, len))
910 			dma_unmap_single(ring->dev,
911 					 dma_unmap_addr(tx_buffer, dma),
912 					 dma_unmap_len(tx_buffer, len),
913 					 DMA_TO_DEVICE);
914 	} else if (dma_unmap_len(tx_buffer, len)) {
915 		dma_unmap_page(ring->dev,
916 			       dma_unmap_addr(tx_buffer, dma),
917 			       dma_unmap_len(tx_buffer, len),
918 			       DMA_TO_DEVICE);
919 	}
920 	tx_buffer->next_to_watch = NULL;
921 	tx_buffer->skb = NULL;
922 	dma_unmap_len_set(tx_buffer, len, 0);
923 	/* tx_buffer must be completely set up in the transmit path */
924 }
925 
926 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
927 {
928 	struct ixgbe_hw *hw = &adapter->hw;
929 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
930 	int i;
931 	u32 data;
932 
933 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
934 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
935 		return;
936 
937 	switch (hw->mac.type) {
938 	case ixgbe_mac_82598EB:
939 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
940 		break;
941 	default:
942 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
943 	}
944 	hwstats->lxoffrxc += data;
945 
946 	/* refill credits (no tx hang) if we received xoff */
947 	if (!data)
948 		return;
949 
950 	for (i = 0; i < adapter->num_tx_queues; i++)
951 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
952 			  &adapter->tx_ring[i]->state);
953 }
954 
955 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
956 {
957 	struct ixgbe_hw *hw = &adapter->hw;
958 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
959 	u32 xoff[8] = {0};
960 	u8 tc;
961 	int i;
962 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
963 
964 	if (adapter->ixgbe_ieee_pfc)
965 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
966 
967 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
968 		ixgbe_update_xoff_rx_lfc(adapter);
969 		return;
970 	}
971 
972 	/* update stats for each tc, only valid with PFC enabled */
973 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
974 		u32 pxoffrxc;
975 
976 		switch (hw->mac.type) {
977 		case ixgbe_mac_82598EB:
978 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
979 			break;
980 		default:
981 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
982 		}
983 		hwstats->pxoffrxc[i] += pxoffrxc;
984 		/* Get the TC for given UP */
985 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
986 		xoff[tc] += pxoffrxc;
987 	}
988 
989 	/* disarm tx queues that have received xoff frames */
990 	for (i = 0; i < adapter->num_tx_queues; i++) {
991 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
992 
993 		tc = tx_ring->dcb_tc;
994 		if (xoff[tc])
995 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
996 	}
997 }
998 
999 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1000 {
1001 	return ring->stats.packets;
1002 }
1003 
1004 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1005 {
1006 	struct ixgbe_adapter *adapter;
1007 	struct ixgbe_hw *hw;
1008 	u32 head, tail;
1009 
1010 	if (ring->l2_accel_priv)
1011 		adapter = ring->l2_accel_priv->real_adapter;
1012 	else
1013 		adapter = netdev_priv(ring->netdev);
1014 
1015 	hw = &adapter->hw;
1016 	head = IXGBE_READ_REG(hw, IXGBE_TDH(ring->reg_idx));
1017 	tail = IXGBE_READ_REG(hw, IXGBE_TDT(ring->reg_idx));
1018 
1019 	if (head != tail)
1020 		return (head < tail) ?
1021 			tail - head : (tail + ring->count - head);
1022 
1023 	return 0;
1024 }
1025 
1026 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1027 {
1028 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1029 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1030 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1031 
1032 	clear_check_for_tx_hang(tx_ring);
1033 
1034 	/*
1035 	 * Check for a hung queue, but be thorough. This verifies
1036 	 * that a transmit has been completed since the previous
1037 	 * check AND there is at least one packet pending. The
1038 	 * ARMED bit is set to indicate a potential hang. The
1039 	 * bit is cleared if a pause frame is received to remove
1040 	 * false hang detection due to PFC or 802.3x frames. By
1041 	 * requiring this to fail twice we avoid races with
1042 	 * pfc clearing the ARMED bit and conditions where we
1043 	 * run the check_tx_hang logic with a transmit completion
1044 	 * pending but without time to complete it yet.
1045 	 */
1046 	if (tx_done_old == tx_done && tx_pending)
1047 		/* make sure it is true for two checks in a row */
1048 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1049 					&tx_ring->state);
1050 	/* update completed stats and continue */
1051 	tx_ring->tx_stats.tx_done_old = tx_done;
1052 	/* reset the countdown */
1053 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1054 
1055 	return false;
1056 }
1057 
1058 /**
1059  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1060  * @adapter: driver private struct
1061  **/
1062 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1063 {
1064 
1065 	/* Do the reset outside of interrupt context */
1066 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1067 		adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
1068 		e_warn(drv, "initiating reset due to tx timeout\n");
1069 		ixgbe_service_event_schedule(adapter);
1070 	}
1071 }
1072 
1073 /**
1074  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1075  * @q_vector: structure containing interrupt and ring information
1076  * @tx_ring: tx ring to clean
1077  **/
1078 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1079 			       struct ixgbe_ring *tx_ring)
1080 {
1081 	struct ixgbe_adapter *adapter = q_vector->adapter;
1082 	struct ixgbe_tx_buffer *tx_buffer;
1083 	union ixgbe_adv_tx_desc *tx_desc;
1084 	unsigned int total_bytes = 0, total_packets = 0;
1085 	unsigned int budget = q_vector->tx.work_limit;
1086 	unsigned int i = tx_ring->next_to_clean;
1087 
1088 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1089 		return true;
1090 
1091 	tx_buffer = &tx_ring->tx_buffer_info[i];
1092 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1093 	i -= tx_ring->count;
1094 
1095 	do {
1096 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1097 
1098 		/* if next_to_watch is not set then there is no work pending */
1099 		if (!eop_desc)
1100 			break;
1101 
1102 		/* prevent any other reads prior to eop_desc */
1103 		read_barrier_depends();
1104 
1105 		/* if DD is not set pending work has not been completed */
1106 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1107 			break;
1108 
1109 		/* clear next_to_watch to prevent false hangs */
1110 		tx_buffer->next_to_watch = NULL;
1111 
1112 		/* update the statistics for this packet */
1113 		total_bytes += tx_buffer->bytecount;
1114 		total_packets += tx_buffer->gso_segs;
1115 
1116 		/* free the skb */
1117 		dev_consume_skb_any(tx_buffer->skb);
1118 
1119 		/* unmap skb header data */
1120 		dma_unmap_single(tx_ring->dev,
1121 				 dma_unmap_addr(tx_buffer, dma),
1122 				 dma_unmap_len(tx_buffer, len),
1123 				 DMA_TO_DEVICE);
1124 
1125 		/* clear tx_buffer data */
1126 		tx_buffer->skb = NULL;
1127 		dma_unmap_len_set(tx_buffer, len, 0);
1128 
1129 		/* unmap remaining buffers */
1130 		while (tx_desc != eop_desc) {
1131 			tx_buffer++;
1132 			tx_desc++;
1133 			i++;
1134 			if (unlikely(!i)) {
1135 				i -= tx_ring->count;
1136 				tx_buffer = tx_ring->tx_buffer_info;
1137 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1138 			}
1139 
1140 			/* unmap any remaining paged data */
1141 			if (dma_unmap_len(tx_buffer, len)) {
1142 				dma_unmap_page(tx_ring->dev,
1143 					       dma_unmap_addr(tx_buffer, dma),
1144 					       dma_unmap_len(tx_buffer, len),
1145 					       DMA_TO_DEVICE);
1146 				dma_unmap_len_set(tx_buffer, len, 0);
1147 			}
1148 		}
1149 
1150 		/* move us one more past the eop_desc for start of next pkt */
1151 		tx_buffer++;
1152 		tx_desc++;
1153 		i++;
1154 		if (unlikely(!i)) {
1155 			i -= tx_ring->count;
1156 			tx_buffer = tx_ring->tx_buffer_info;
1157 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1158 		}
1159 
1160 		/* issue prefetch for next Tx descriptor */
1161 		prefetch(tx_desc);
1162 
1163 		/* update budget accounting */
1164 		budget--;
1165 	} while (likely(budget));
1166 
1167 	i += tx_ring->count;
1168 	tx_ring->next_to_clean = i;
1169 	u64_stats_update_begin(&tx_ring->syncp);
1170 	tx_ring->stats.bytes += total_bytes;
1171 	tx_ring->stats.packets += total_packets;
1172 	u64_stats_update_end(&tx_ring->syncp);
1173 	q_vector->tx.total_bytes += total_bytes;
1174 	q_vector->tx.total_packets += total_packets;
1175 
1176 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1177 		/* schedule immediate reset if we believe we hung */
1178 		struct ixgbe_hw *hw = &adapter->hw;
1179 		e_err(drv, "Detected Tx Unit Hang\n"
1180 			"  Tx Queue             <%d>\n"
1181 			"  TDH, TDT             <%x>, <%x>\n"
1182 			"  next_to_use          <%x>\n"
1183 			"  next_to_clean        <%x>\n"
1184 			"tx_buffer_info[next_to_clean]\n"
1185 			"  time_stamp           <%lx>\n"
1186 			"  jiffies              <%lx>\n",
1187 			tx_ring->queue_index,
1188 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1189 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1190 			tx_ring->next_to_use, i,
1191 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1192 
1193 		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
1194 
1195 		e_info(probe,
1196 		       "tx hang %d detected on queue %d, resetting adapter\n",
1197 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1198 
1199 		/* schedule immediate reset if we believe we hung */
1200 		ixgbe_tx_timeout_reset(adapter);
1201 
1202 		/* the adapter is about to reset, no point in enabling stuff */
1203 		return true;
1204 	}
1205 
1206 	netdev_tx_completed_queue(txring_txq(tx_ring),
1207 				  total_packets, total_bytes);
1208 
1209 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1210 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1211 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1212 		/* Make sure that anybody stopping the queue after this
1213 		 * sees the new next_to_clean.
1214 		 */
1215 		smp_mb();
1216 		if (__netif_subqueue_stopped(tx_ring->netdev,
1217 					     tx_ring->queue_index)
1218 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1219 			netif_wake_subqueue(tx_ring->netdev,
1220 					    tx_ring->queue_index);
1221 			++tx_ring->tx_stats.restart_queue;
1222 		}
1223 	}
1224 
1225 	return !!budget;
1226 }
1227 
1228 #ifdef CONFIG_IXGBE_DCA
1229 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1230 				struct ixgbe_ring *tx_ring,
1231 				int cpu)
1232 {
1233 	struct ixgbe_hw *hw = &adapter->hw;
1234 	u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
1235 	u16 reg_offset;
1236 
1237 	switch (hw->mac.type) {
1238 	case ixgbe_mac_82598EB:
1239 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1240 		break;
1241 	case ixgbe_mac_82599EB:
1242 	case ixgbe_mac_X540:
1243 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1244 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1245 		break;
1246 	default:
1247 		/* for unknown hardware do not write register */
1248 		return;
1249 	}
1250 
1251 	/*
1252 	 * We can enable relaxed ordering for reads, but not writes when
1253 	 * DCA is enabled.  This is due to a known issue in some chipsets
1254 	 * which will cause the DCA tag to be cleared.
1255 	 */
1256 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1257 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1258 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1259 
1260 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1261 }
1262 
1263 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1264 				struct ixgbe_ring *rx_ring,
1265 				int cpu)
1266 {
1267 	struct ixgbe_hw *hw = &adapter->hw;
1268 	u32 rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1269 	u8 reg_idx = rx_ring->reg_idx;
1270 
1271 
1272 	switch (hw->mac.type) {
1273 	case ixgbe_mac_82599EB:
1274 	case ixgbe_mac_X540:
1275 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1276 		break;
1277 	default:
1278 		break;
1279 	}
1280 
1281 	/*
1282 	 * We can enable relaxed ordering for reads, but not writes when
1283 	 * DCA is enabled.  This is due to a known issue in some chipsets
1284 	 * which will cause the DCA tag to be cleared.
1285 	 */
1286 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1287 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1288 
1289 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1290 }
1291 
1292 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1293 {
1294 	struct ixgbe_adapter *adapter = q_vector->adapter;
1295 	struct ixgbe_ring *ring;
1296 	int cpu = get_cpu();
1297 
1298 	if (q_vector->cpu == cpu)
1299 		goto out_no_update;
1300 
1301 	ixgbe_for_each_ring(ring, q_vector->tx)
1302 		ixgbe_update_tx_dca(adapter, ring, cpu);
1303 
1304 	ixgbe_for_each_ring(ring, q_vector->rx)
1305 		ixgbe_update_rx_dca(adapter, ring, cpu);
1306 
1307 	q_vector->cpu = cpu;
1308 out_no_update:
1309 	put_cpu();
1310 }
1311 
1312 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1313 {
1314 	int i;
1315 
1316 	if (!(adapter->flags & IXGBE_FLAG_DCA_ENABLED))
1317 		return;
1318 
1319 	/* always use CB2 mode, difference is masked in the CB driver */
1320 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 2);
1321 
1322 	for (i = 0; i < adapter->num_q_vectors; i++) {
1323 		adapter->q_vector[i]->cpu = -1;
1324 		ixgbe_update_dca(adapter->q_vector[i]);
1325 	}
1326 }
1327 
1328 static int __ixgbe_notify_dca(struct device *dev, void *data)
1329 {
1330 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1331 	unsigned long event = *(unsigned long *)data;
1332 
1333 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1334 		return 0;
1335 
1336 	switch (event) {
1337 	case DCA_PROVIDER_ADD:
1338 		/* if we're already enabled, don't do it again */
1339 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1340 			break;
1341 		if (dca_add_requester(dev) == 0) {
1342 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1343 			ixgbe_setup_dca(adapter);
1344 			break;
1345 		}
1346 		/* Fall Through since DCA is disabled. */
1347 	case DCA_PROVIDER_REMOVE:
1348 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1349 			dca_remove_requester(dev);
1350 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1351 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
1352 		}
1353 		break;
1354 	}
1355 
1356 	return 0;
1357 }
1358 
1359 #endif /* CONFIG_IXGBE_DCA */
1360 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1361 				 union ixgbe_adv_rx_desc *rx_desc,
1362 				 struct sk_buff *skb)
1363 {
1364 	if (ring->netdev->features & NETIF_F_RXHASH)
1365 		skb_set_hash(skb,
1366 			     le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1367 			     PKT_HASH_TYPE_L3);
1368 }
1369 
1370 #ifdef IXGBE_FCOE
1371 /**
1372  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1373  * @ring: structure containing ring specific data
1374  * @rx_desc: advanced rx descriptor
1375  *
1376  * Returns : true if it is FCoE pkt
1377  */
1378 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1379 				    union ixgbe_adv_rx_desc *rx_desc)
1380 {
1381 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1382 
1383 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1384 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1385 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1386 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1387 }
1388 
1389 #endif /* IXGBE_FCOE */
1390 /**
1391  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1392  * @ring: structure containing ring specific data
1393  * @rx_desc: current Rx descriptor being processed
1394  * @skb: skb currently being received and modified
1395  **/
1396 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1397 				     union ixgbe_adv_rx_desc *rx_desc,
1398 				     struct sk_buff *skb)
1399 {
1400 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1401 	__le16 hdr_info = rx_desc->wb.lower.lo_dword.hs_rss.hdr_info;
1402 	bool encap_pkt = false;
1403 
1404 	skb_checksum_none_assert(skb);
1405 
1406 	/* Rx csum disabled */
1407 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1408 		return;
1409 
1410 	if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) &&
1411 	    (hdr_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_TUNNEL >> 16))) {
1412 		encap_pkt = true;
1413 		skb->encapsulation = 1;
1414 		skb->ip_summed = CHECKSUM_NONE;
1415 	}
1416 
1417 	/* if IP and error */
1418 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1419 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1420 		ring->rx_stats.csum_err++;
1421 		return;
1422 	}
1423 
1424 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1425 		return;
1426 
1427 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1428 		/*
1429 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1430 		 * checksum errors.
1431 		 */
1432 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1433 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1434 			return;
1435 
1436 		ring->rx_stats.csum_err++;
1437 		return;
1438 	}
1439 
1440 	/* It must be a TCP or UDP packet with a valid checksum */
1441 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1442 	if (encap_pkt) {
1443 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1444 			return;
1445 
1446 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1447 			ring->rx_stats.csum_err++;
1448 			return;
1449 		}
1450 		/* If we checked the outer header let the stack know */
1451 		skb->csum_level = 1;
1452 	}
1453 }
1454 
1455 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1456 				    struct ixgbe_rx_buffer *bi)
1457 {
1458 	struct page *page = bi->page;
1459 	dma_addr_t dma;
1460 
1461 	/* since we are recycling buffers we should seldom need to alloc */
1462 	if (likely(page))
1463 		return true;
1464 
1465 	/* alloc new page for storage */
1466 	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1467 	if (unlikely(!page)) {
1468 		rx_ring->rx_stats.alloc_rx_page_failed++;
1469 		return false;
1470 	}
1471 
1472 	/* map page for use */
1473 	dma = dma_map_page(rx_ring->dev, page, 0,
1474 			   ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1475 
1476 	/*
1477 	 * if mapping failed free memory back to system since
1478 	 * there isn't much point in holding memory we can't use
1479 	 */
1480 	if (dma_mapping_error(rx_ring->dev, dma)) {
1481 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1482 
1483 		rx_ring->rx_stats.alloc_rx_page_failed++;
1484 		return false;
1485 	}
1486 
1487 	bi->dma = dma;
1488 	bi->page = page;
1489 	bi->page_offset = 0;
1490 
1491 	return true;
1492 }
1493 
1494 /**
1495  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1496  * @rx_ring: ring to place buffers on
1497  * @cleaned_count: number of buffers to replace
1498  **/
1499 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1500 {
1501 	union ixgbe_adv_rx_desc *rx_desc;
1502 	struct ixgbe_rx_buffer *bi;
1503 	u16 i = rx_ring->next_to_use;
1504 
1505 	/* nothing to do */
1506 	if (!cleaned_count)
1507 		return;
1508 
1509 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1510 	bi = &rx_ring->rx_buffer_info[i];
1511 	i -= rx_ring->count;
1512 
1513 	do {
1514 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1515 			break;
1516 
1517 		/*
1518 		 * Refresh the desc even if buffer_addrs didn't change
1519 		 * because each write-back erases this info.
1520 		 */
1521 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1522 
1523 		rx_desc++;
1524 		bi++;
1525 		i++;
1526 		if (unlikely(!i)) {
1527 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1528 			bi = rx_ring->rx_buffer_info;
1529 			i -= rx_ring->count;
1530 		}
1531 
1532 		/* clear the status bits for the next_to_use descriptor */
1533 		rx_desc->wb.upper.status_error = 0;
1534 
1535 		cleaned_count--;
1536 	} while (cleaned_count);
1537 
1538 	i += rx_ring->count;
1539 
1540 	if (rx_ring->next_to_use != i) {
1541 		rx_ring->next_to_use = i;
1542 
1543 		/* update next to alloc since we have filled the ring */
1544 		rx_ring->next_to_alloc = i;
1545 
1546 		/* Force memory writes to complete before letting h/w
1547 		 * know there are new descriptors to fetch.  (Only
1548 		 * applicable for weak-ordered memory model archs,
1549 		 * such as IA-64).
1550 		 */
1551 		wmb();
1552 		writel(i, rx_ring->tail);
1553 	}
1554 }
1555 
1556 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1557 				   struct sk_buff *skb)
1558 {
1559 	u16 hdr_len = skb_headlen(skb);
1560 
1561 	/* set gso_size to avoid messing up TCP MSS */
1562 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1563 						 IXGBE_CB(skb)->append_cnt);
1564 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1565 }
1566 
1567 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1568 				   struct sk_buff *skb)
1569 {
1570 	/* if append_cnt is 0 then frame is not RSC */
1571 	if (!IXGBE_CB(skb)->append_cnt)
1572 		return;
1573 
1574 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1575 	rx_ring->rx_stats.rsc_flush++;
1576 
1577 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1578 
1579 	/* gso_size is computed using append_cnt so always clear it last */
1580 	IXGBE_CB(skb)->append_cnt = 0;
1581 }
1582 
1583 /**
1584  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1585  * @rx_ring: rx descriptor ring packet is being transacted on
1586  * @rx_desc: pointer to the EOP Rx descriptor
1587  * @skb: pointer to current skb being populated
1588  *
1589  * This function checks the ring, descriptor, and packet information in
1590  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1591  * other fields within the skb.
1592  **/
1593 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1594 				     union ixgbe_adv_rx_desc *rx_desc,
1595 				     struct sk_buff *skb)
1596 {
1597 	struct net_device *dev = rx_ring->netdev;
1598 
1599 	ixgbe_update_rsc_stats(rx_ring, skb);
1600 
1601 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1602 
1603 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1604 
1605 	if (unlikely(ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_TS)))
1606 		ixgbe_ptp_rx_hwtstamp(rx_ring->q_vector->adapter, skb);
1607 
1608 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1609 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1610 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1611 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1612 	}
1613 
1614 	skb_record_rx_queue(skb, rx_ring->queue_index);
1615 
1616 	skb->protocol = eth_type_trans(skb, dev);
1617 }
1618 
1619 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1620 			 struct sk_buff *skb)
1621 {
1622 	if (ixgbe_qv_busy_polling(q_vector))
1623 		netif_receive_skb(skb);
1624 	else
1625 		napi_gro_receive(&q_vector->napi, skb);
1626 }
1627 
1628 /**
1629  * ixgbe_is_non_eop - process handling of non-EOP buffers
1630  * @rx_ring: Rx ring being processed
1631  * @rx_desc: Rx descriptor for current buffer
1632  * @skb: Current socket buffer containing buffer in progress
1633  *
1634  * This function updates next to clean.  If the buffer is an EOP buffer
1635  * this function exits returning false, otherwise it will place the
1636  * sk_buff in the next buffer to be chained and return true indicating
1637  * that this is in fact a non-EOP buffer.
1638  **/
1639 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1640 			     union ixgbe_adv_rx_desc *rx_desc,
1641 			     struct sk_buff *skb)
1642 {
1643 	u32 ntc = rx_ring->next_to_clean + 1;
1644 
1645 	/* fetch, update, and store next to clean */
1646 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1647 	rx_ring->next_to_clean = ntc;
1648 
1649 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1650 
1651 	/* update RSC append count if present */
1652 	if (ring_is_rsc_enabled(rx_ring)) {
1653 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1654 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1655 
1656 		if (unlikely(rsc_enabled)) {
1657 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1658 
1659 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1660 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1661 
1662 			/* update ntc based on RSC value */
1663 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1664 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1665 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1666 		}
1667 	}
1668 
1669 	/* if we are the last buffer then there is nothing else to do */
1670 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1671 		return false;
1672 
1673 	/* place skb in next buffer to be received */
1674 	rx_ring->rx_buffer_info[ntc].skb = skb;
1675 	rx_ring->rx_stats.non_eop_descs++;
1676 
1677 	return true;
1678 }
1679 
1680 /**
1681  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1682  * @rx_ring: rx descriptor ring packet is being transacted on
1683  * @skb: pointer to current skb being adjusted
1684  *
1685  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1686  * main difference between this version and the original function is that
1687  * this function can make several assumptions about the state of things
1688  * that allow for significant optimizations versus the standard function.
1689  * As a result we can do things like drop a frag and maintain an accurate
1690  * truesize for the skb.
1691  */
1692 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1693 			    struct sk_buff *skb)
1694 {
1695 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1696 	unsigned char *va;
1697 	unsigned int pull_len;
1698 
1699 	/*
1700 	 * it is valid to use page_address instead of kmap since we are
1701 	 * working with pages allocated out of the lomem pool per
1702 	 * alloc_page(GFP_ATOMIC)
1703 	 */
1704 	va = skb_frag_address(frag);
1705 
1706 	/*
1707 	 * we need the header to contain the greater of either ETH_HLEN or
1708 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1709 	 */
1710 	pull_len = eth_get_headlen(va, IXGBE_RX_HDR_SIZE);
1711 
1712 	/* align pull length to size of long to optimize memcpy performance */
1713 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1714 
1715 	/* update all of the pointers */
1716 	skb_frag_size_sub(frag, pull_len);
1717 	frag->page_offset += pull_len;
1718 	skb->data_len -= pull_len;
1719 	skb->tail += pull_len;
1720 }
1721 
1722 /**
1723  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1724  * @rx_ring: rx descriptor ring packet is being transacted on
1725  * @skb: pointer to current skb being updated
1726  *
1727  * This function provides a basic DMA sync up for the first fragment of an
1728  * skb.  The reason for doing this is that the first fragment cannot be
1729  * unmapped until we have reached the end of packet descriptor for a buffer
1730  * chain.
1731  */
1732 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1733 				struct sk_buff *skb)
1734 {
1735 	/* if the page was released unmap it, else just sync our portion */
1736 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1737 		dma_unmap_page(rx_ring->dev, IXGBE_CB(skb)->dma,
1738 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
1739 		IXGBE_CB(skb)->page_released = false;
1740 	} else {
1741 		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1742 
1743 		dma_sync_single_range_for_cpu(rx_ring->dev,
1744 					      IXGBE_CB(skb)->dma,
1745 					      frag->page_offset,
1746 					      ixgbe_rx_bufsz(rx_ring),
1747 					      DMA_FROM_DEVICE);
1748 	}
1749 	IXGBE_CB(skb)->dma = 0;
1750 }
1751 
1752 /**
1753  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1754  * @rx_ring: rx descriptor ring packet is being transacted on
1755  * @rx_desc: pointer to the EOP Rx descriptor
1756  * @skb: pointer to current skb being fixed
1757  *
1758  * Check for corrupted packet headers caused by senders on the local L2
1759  * embedded NIC switch not setting up their Tx Descriptors right.  These
1760  * should be very rare.
1761  *
1762  * Also address the case where we are pulling data in on pages only
1763  * and as such no data is present in the skb header.
1764  *
1765  * In addition if skb is not at least 60 bytes we need to pad it so that
1766  * it is large enough to qualify as a valid Ethernet frame.
1767  *
1768  * Returns true if an error was encountered and skb was freed.
1769  **/
1770 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1771 				  union ixgbe_adv_rx_desc *rx_desc,
1772 				  struct sk_buff *skb)
1773 {
1774 	struct net_device *netdev = rx_ring->netdev;
1775 
1776 	/* verify that the packet does not have any known errors */
1777 	if (unlikely(ixgbe_test_staterr(rx_desc,
1778 					IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1779 	    !(netdev->features & NETIF_F_RXALL))) {
1780 		dev_kfree_skb_any(skb);
1781 		return true;
1782 	}
1783 
1784 	/* place header in linear portion of buffer */
1785 	if (skb_is_nonlinear(skb))
1786 		ixgbe_pull_tail(rx_ring, skb);
1787 
1788 #ifdef IXGBE_FCOE
1789 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1790 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1791 		return false;
1792 
1793 #endif
1794 	/* if eth_skb_pad returns an error the skb was freed */
1795 	if (eth_skb_pad(skb))
1796 		return true;
1797 
1798 	return false;
1799 }
1800 
1801 /**
1802  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1803  * @rx_ring: rx descriptor ring to store buffers on
1804  * @old_buff: donor buffer to have page reused
1805  *
1806  * Synchronizes page for reuse by the adapter
1807  **/
1808 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1809 				struct ixgbe_rx_buffer *old_buff)
1810 {
1811 	struct ixgbe_rx_buffer *new_buff;
1812 	u16 nta = rx_ring->next_to_alloc;
1813 
1814 	new_buff = &rx_ring->rx_buffer_info[nta];
1815 
1816 	/* update, and store next to alloc */
1817 	nta++;
1818 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1819 
1820 	/* transfer page from old buffer to new buffer */
1821 	*new_buff = *old_buff;
1822 
1823 	/* sync the buffer for use by the device */
1824 	dma_sync_single_range_for_device(rx_ring->dev, new_buff->dma,
1825 					 new_buff->page_offset,
1826 					 ixgbe_rx_bufsz(rx_ring),
1827 					 DMA_FROM_DEVICE);
1828 }
1829 
1830 static inline bool ixgbe_page_is_reserved(struct page *page)
1831 {
1832 	return (page_to_nid(page) != numa_mem_id()) || page->pfmemalloc;
1833 }
1834 
1835 /**
1836  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1837  * @rx_ring: rx descriptor ring to transact packets on
1838  * @rx_buffer: buffer containing page to add
1839  * @rx_desc: descriptor containing length of buffer written by hardware
1840  * @skb: sk_buff to place the data into
1841  *
1842  * This function will add the data contained in rx_buffer->page to the skb.
1843  * This is done either through a direct copy if the data in the buffer is
1844  * less than the skb header size, otherwise it will just attach the page as
1845  * a frag to the skb.
1846  *
1847  * The function will then update the page offset if necessary and return
1848  * true if the buffer can be reused by the adapter.
1849  **/
1850 static bool ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1851 			      struct ixgbe_rx_buffer *rx_buffer,
1852 			      union ixgbe_adv_rx_desc *rx_desc,
1853 			      struct sk_buff *skb)
1854 {
1855 	struct page *page = rx_buffer->page;
1856 	unsigned int size = le16_to_cpu(rx_desc->wb.upper.length);
1857 #if (PAGE_SIZE < 8192)
1858 	unsigned int truesize = ixgbe_rx_bufsz(rx_ring);
1859 #else
1860 	unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
1861 	unsigned int last_offset = ixgbe_rx_pg_size(rx_ring) -
1862 				   ixgbe_rx_bufsz(rx_ring);
1863 #endif
1864 
1865 	if ((size <= IXGBE_RX_HDR_SIZE) && !skb_is_nonlinear(skb)) {
1866 		unsigned char *va = page_address(page) + rx_buffer->page_offset;
1867 
1868 		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
1869 
1870 		/* page is not reserved, we can reuse buffer as-is */
1871 		if (likely(!ixgbe_page_is_reserved(page)))
1872 			return true;
1873 
1874 		/* this page cannot be reused so discard it */
1875 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1876 		return false;
1877 	}
1878 
1879 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
1880 			rx_buffer->page_offset, size, truesize);
1881 
1882 	/* avoid re-using remote pages */
1883 	if (unlikely(ixgbe_page_is_reserved(page)))
1884 		return false;
1885 
1886 #if (PAGE_SIZE < 8192)
1887 	/* if we are only owner of page we can reuse it */
1888 	if (unlikely(page_count(page) != 1))
1889 		return false;
1890 
1891 	/* flip page offset to other buffer */
1892 	rx_buffer->page_offset ^= truesize;
1893 #else
1894 	/* move offset up to the next cache line */
1895 	rx_buffer->page_offset += truesize;
1896 
1897 	if (rx_buffer->page_offset > last_offset)
1898 		return false;
1899 #endif
1900 
1901 	/* Even if we own the page, we are not allowed to use atomic_set()
1902 	 * This would break get_page_unless_zero() users.
1903 	 */
1904 	atomic_inc(&page->_count);
1905 
1906 	return true;
1907 }
1908 
1909 static struct sk_buff *ixgbe_fetch_rx_buffer(struct ixgbe_ring *rx_ring,
1910 					     union ixgbe_adv_rx_desc *rx_desc)
1911 {
1912 	struct ixgbe_rx_buffer *rx_buffer;
1913 	struct sk_buff *skb;
1914 	struct page *page;
1915 
1916 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
1917 	page = rx_buffer->page;
1918 	prefetchw(page);
1919 
1920 	skb = rx_buffer->skb;
1921 
1922 	if (likely(!skb)) {
1923 		void *page_addr = page_address(page) +
1924 				  rx_buffer->page_offset;
1925 
1926 		/* prefetch first cache line of first page */
1927 		prefetch(page_addr);
1928 #if L1_CACHE_BYTES < 128
1929 		prefetch(page_addr + L1_CACHE_BYTES);
1930 #endif
1931 
1932 		/* allocate a skb to store the frags */
1933 		skb = napi_alloc_skb(&rx_ring->q_vector->napi,
1934 				     IXGBE_RX_HDR_SIZE);
1935 		if (unlikely(!skb)) {
1936 			rx_ring->rx_stats.alloc_rx_buff_failed++;
1937 			return NULL;
1938 		}
1939 
1940 		/*
1941 		 * we will be copying header into skb->data in
1942 		 * pskb_may_pull so it is in our interest to prefetch
1943 		 * it now to avoid a possible cache miss
1944 		 */
1945 		prefetchw(skb->data);
1946 
1947 		/*
1948 		 * Delay unmapping of the first packet. It carries the
1949 		 * header information, HW may still access the header
1950 		 * after the writeback.  Only unmap it when EOP is
1951 		 * reached
1952 		 */
1953 		if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1954 			goto dma_sync;
1955 
1956 		IXGBE_CB(skb)->dma = rx_buffer->dma;
1957 	} else {
1958 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
1959 			ixgbe_dma_sync_frag(rx_ring, skb);
1960 
1961 dma_sync:
1962 		/* we are reusing so sync this buffer for CPU use */
1963 		dma_sync_single_range_for_cpu(rx_ring->dev,
1964 					      rx_buffer->dma,
1965 					      rx_buffer->page_offset,
1966 					      ixgbe_rx_bufsz(rx_ring),
1967 					      DMA_FROM_DEVICE);
1968 
1969 		rx_buffer->skb = NULL;
1970 	}
1971 
1972 	/* pull page into skb */
1973 	if (ixgbe_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) {
1974 		/* hand second half of page back to the ring */
1975 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
1976 	} else if (IXGBE_CB(skb)->dma == rx_buffer->dma) {
1977 		/* the page has been released from the ring */
1978 		IXGBE_CB(skb)->page_released = true;
1979 	} else {
1980 		/* we are not reusing the buffer so unmap it */
1981 		dma_unmap_page(rx_ring->dev, rx_buffer->dma,
1982 			       ixgbe_rx_pg_size(rx_ring),
1983 			       DMA_FROM_DEVICE);
1984 	}
1985 
1986 	/* clear contents of buffer_info */
1987 	rx_buffer->page = NULL;
1988 
1989 	return skb;
1990 }
1991 
1992 /**
1993  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
1994  * @q_vector: structure containing interrupt and ring information
1995  * @rx_ring: rx descriptor ring to transact packets on
1996  * @budget: Total limit on number of packets to process
1997  *
1998  * This function provides a "bounce buffer" approach to Rx interrupt
1999  * processing.  The advantage to this is that on systems that have
2000  * expensive overhead for IOMMU access this provides a means of avoiding
2001  * it by maintaining the mapping of the page to the syste.
2002  *
2003  * Returns amount of work completed
2004  **/
2005 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2006 			       struct ixgbe_ring *rx_ring,
2007 			       const int budget)
2008 {
2009 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
2010 #ifdef IXGBE_FCOE
2011 	struct ixgbe_adapter *adapter = q_vector->adapter;
2012 	int ddp_bytes;
2013 	unsigned int mss = 0;
2014 #endif /* IXGBE_FCOE */
2015 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2016 
2017 	while (likely(total_rx_packets < budget)) {
2018 		union ixgbe_adv_rx_desc *rx_desc;
2019 		struct sk_buff *skb;
2020 
2021 		/* return some buffers to hardware, one at a time is too slow */
2022 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2023 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2024 			cleaned_count = 0;
2025 		}
2026 
2027 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2028 
2029 		if (!rx_desc->wb.upper.status_error)
2030 			break;
2031 
2032 		/* This memory barrier is needed to keep us from reading
2033 		 * any other fields out of the rx_desc until we know the
2034 		 * descriptor has been written back
2035 		 */
2036 		dma_rmb();
2037 
2038 		/* retrieve a buffer from the ring */
2039 		skb = ixgbe_fetch_rx_buffer(rx_ring, rx_desc);
2040 
2041 		/* exit if we failed to retrieve a buffer */
2042 		if (!skb)
2043 			break;
2044 
2045 		cleaned_count++;
2046 
2047 		/* place incomplete frames back on ring for completion */
2048 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2049 			continue;
2050 
2051 		/* verify the packet layout is correct */
2052 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2053 			continue;
2054 
2055 		/* probably a little skewed due to removing CRC */
2056 		total_rx_bytes += skb->len;
2057 
2058 		/* populate checksum, timestamp, VLAN, and protocol */
2059 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2060 
2061 #ifdef IXGBE_FCOE
2062 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2063 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2064 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2065 			/* include DDPed FCoE data */
2066 			if (ddp_bytes > 0) {
2067 				if (!mss) {
2068 					mss = rx_ring->netdev->mtu -
2069 						sizeof(struct fcoe_hdr) -
2070 						sizeof(struct fc_frame_header) -
2071 						sizeof(struct fcoe_crc_eof);
2072 					if (mss > 512)
2073 						mss &= ~511;
2074 				}
2075 				total_rx_bytes += ddp_bytes;
2076 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2077 								 mss);
2078 			}
2079 			if (!ddp_bytes) {
2080 				dev_kfree_skb_any(skb);
2081 				continue;
2082 			}
2083 		}
2084 
2085 #endif /* IXGBE_FCOE */
2086 		skb_mark_napi_id(skb, &q_vector->napi);
2087 		ixgbe_rx_skb(q_vector, skb);
2088 
2089 		/* update budget accounting */
2090 		total_rx_packets++;
2091 	}
2092 
2093 	u64_stats_update_begin(&rx_ring->syncp);
2094 	rx_ring->stats.packets += total_rx_packets;
2095 	rx_ring->stats.bytes += total_rx_bytes;
2096 	u64_stats_update_end(&rx_ring->syncp);
2097 	q_vector->rx.total_packets += total_rx_packets;
2098 	q_vector->rx.total_bytes += total_rx_bytes;
2099 
2100 	return total_rx_packets;
2101 }
2102 
2103 #ifdef CONFIG_NET_RX_BUSY_POLL
2104 /* must be called with local_bh_disable()d */
2105 static int ixgbe_low_latency_recv(struct napi_struct *napi)
2106 {
2107 	struct ixgbe_q_vector *q_vector =
2108 			container_of(napi, struct ixgbe_q_vector, napi);
2109 	struct ixgbe_adapter *adapter = q_vector->adapter;
2110 	struct ixgbe_ring  *ring;
2111 	int found = 0;
2112 
2113 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2114 		return LL_FLUSH_FAILED;
2115 
2116 	if (!ixgbe_qv_lock_poll(q_vector))
2117 		return LL_FLUSH_BUSY;
2118 
2119 	ixgbe_for_each_ring(ring, q_vector->rx) {
2120 		found = ixgbe_clean_rx_irq(q_vector, ring, 4);
2121 #ifdef BP_EXTENDED_STATS
2122 		if (found)
2123 			ring->stats.cleaned += found;
2124 		else
2125 			ring->stats.misses++;
2126 #endif
2127 		if (found)
2128 			break;
2129 	}
2130 
2131 	ixgbe_qv_unlock_poll(q_vector);
2132 
2133 	return found;
2134 }
2135 #endif	/* CONFIG_NET_RX_BUSY_POLL */
2136 
2137 /**
2138  * ixgbe_configure_msix - Configure MSI-X hardware
2139  * @adapter: board private structure
2140  *
2141  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2142  * interrupts.
2143  **/
2144 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2145 {
2146 	struct ixgbe_q_vector *q_vector;
2147 	int v_idx;
2148 	u32 mask;
2149 
2150 	/* Populate MSIX to EITR Select */
2151 	if (adapter->num_vfs > 32) {
2152 		u32 eitrsel = (1 << (adapter->num_vfs - 32)) - 1;
2153 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2154 	}
2155 
2156 	/*
2157 	 * Populate the IVAR table and set the ITR values to the
2158 	 * corresponding register.
2159 	 */
2160 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2161 		struct ixgbe_ring *ring;
2162 		q_vector = adapter->q_vector[v_idx];
2163 
2164 		ixgbe_for_each_ring(ring, q_vector->rx)
2165 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2166 
2167 		ixgbe_for_each_ring(ring, q_vector->tx)
2168 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2169 
2170 		ixgbe_write_eitr(q_vector);
2171 	}
2172 
2173 	switch (adapter->hw.mac.type) {
2174 	case ixgbe_mac_82598EB:
2175 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2176 			       v_idx);
2177 		break;
2178 	case ixgbe_mac_82599EB:
2179 	case ixgbe_mac_X540:
2180 	case ixgbe_mac_X550:
2181 	case ixgbe_mac_X550EM_x:
2182 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2183 		break;
2184 	default:
2185 		break;
2186 	}
2187 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2188 
2189 	/* set up to autoclear timer, and the vectors */
2190 	mask = IXGBE_EIMS_ENABLE_MASK;
2191 	mask &= ~(IXGBE_EIMS_OTHER |
2192 		  IXGBE_EIMS_MAILBOX |
2193 		  IXGBE_EIMS_LSC);
2194 
2195 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2196 }
2197 
2198 enum latency_range {
2199 	lowest_latency = 0,
2200 	low_latency = 1,
2201 	bulk_latency = 2,
2202 	latency_invalid = 255
2203 };
2204 
2205 /**
2206  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2207  * @q_vector: structure containing interrupt and ring information
2208  * @ring_container: structure containing ring performance data
2209  *
2210  *      Stores a new ITR value based on packets and byte
2211  *      counts during the last interrupt.  The advantage of per interrupt
2212  *      computation is faster updates and more accurate ITR for the current
2213  *      traffic pattern.  Constants in this function were computed
2214  *      based on theoretical maximum wire speed and thresholds were set based
2215  *      on testing data as well as attempting to minimize response time
2216  *      while increasing bulk throughput.
2217  *      this functionality is controlled by the InterruptThrottleRate module
2218  *      parameter (see ixgbe_param.c)
2219  **/
2220 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2221 			     struct ixgbe_ring_container *ring_container)
2222 {
2223 	int bytes = ring_container->total_bytes;
2224 	int packets = ring_container->total_packets;
2225 	u32 timepassed_us;
2226 	u64 bytes_perint;
2227 	u8 itr_setting = ring_container->itr;
2228 
2229 	if (packets == 0)
2230 		return;
2231 
2232 	/* simple throttlerate management
2233 	 *   0-10MB/s   lowest (100000 ints/s)
2234 	 *  10-20MB/s   low    (20000 ints/s)
2235 	 *  20-1249MB/s bulk   (8000 ints/s)
2236 	 */
2237 	/* what was last interrupt timeslice? */
2238 	timepassed_us = q_vector->itr >> 2;
2239 	if (timepassed_us == 0)
2240 		return;
2241 
2242 	bytes_perint = bytes / timepassed_us; /* bytes/usec */
2243 
2244 	switch (itr_setting) {
2245 	case lowest_latency:
2246 		if (bytes_perint > 10)
2247 			itr_setting = low_latency;
2248 		break;
2249 	case low_latency:
2250 		if (bytes_perint > 20)
2251 			itr_setting = bulk_latency;
2252 		else if (bytes_perint <= 10)
2253 			itr_setting = lowest_latency;
2254 		break;
2255 	case bulk_latency:
2256 		if (bytes_perint <= 20)
2257 			itr_setting = low_latency;
2258 		break;
2259 	}
2260 
2261 	/* clear work counters since we have the values we need */
2262 	ring_container->total_bytes = 0;
2263 	ring_container->total_packets = 0;
2264 
2265 	/* write updated itr to ring container */
2266 	ring_container->itr = itr_setting;
2267 }
2268 
2269 /**
2270  * ixgbe_write_eitr - write EITR register in hardware specific way
2271  * @q_vector: structure containing interrupt and ring information
2272  *
2273  * This function is made to be called by ethtool and by the driver
2274  * when it needs to update EITR registers at runtime.  Hardware
2275  * specific quirks/differences are taken care of here.
2276  */
2277 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2278 {
2279 	struct ixgbe_adapter *adapter = q_vector->adapter;
2280 	struct ixgbe_hw *hw = &adapter->hw;
2281 	int v_idx = q_vector->v_idx;
2282 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2283 
2284 	switch (adapter->hw.mac.type) {
2285 	case ixgbe_mac_82598EB:
2286 		/* must write high and low 16 bits to reset counter */
2287 		itr_reg |= (itr_reg << 16);
2288 		break;
2289 	case ixgbe_mac_82599EB:
2290 	case ixgbe_mac_X540:
2291 	case ixgbe_mac_X550:
2292 	case ixgbe_mac_X550EM_x:
2293 		/*
2294 		 * set the WDIS bit to not clear the timer bits and cause an
2295 		 * immediate assertion of the interrupt
2296 		 */
2297 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2298 		break;
2299 	default:
2300 		break;
2301 	}
2302 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2303 }
2304 
2305 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2306 {
2307 	u32 new_itr = q_vector->itr;
2308 	u8 current_itr;
2309 
2310 	ixgbe_update_itr(q_vector, &q_vector->tx);
2311 	ixgbe_update_itr(q_vector, &q_vector->rx);
2312 
2313 	current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
2314 
2315 	switch (current_itr) {
2316 	/* counts and packets in update_itr are dependent on these numbers */
2317 	case lowest_latency:
2318 		new_itr = IXGBE_100K_ITR;
2319 		break;
2320 	case low_latency:
2321 		new_itr = IXGBE_20K_ITR;
2322 		break;
2323 	case bulk_latency:
2324 		new_itr = IXGBE_8K_ITR;
2325 		break;
2326 	default:
2327 		break;
2328 	}
2329 
2330 	if (new_itr != q_vector->itr) {
2331 		/* do an exponential smoothing */
2332 		new_itr = (10 * new_itr * q_vector->itr) /
2333 			  ((9 * new_itr) + q_vector->itr);
2334 
2335 		/* save the algorithm value here */
2336 		q_vector->itr = new_itr;
2337 
2338 		ixgbe_write_eitr(q_vector);
2339 	}
2340 }
2341 
2342 /**
2343  * ixgbe_check_overtemp_subtask - check for over temperature
2344  * @adapter: pointer to adapter
2345  **/
2346 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2347 {
2348 	struct ixgbe_hw *hw = &adapter->hw;
2349 	u32 eicr = adapter->interrupt_event;
2350 
2351 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2352 		return;
2353 
2354 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) &&
2355 	    !(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2356 		return;
2357 
2358 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2359 
2360 	switch (hw->device_id) {
2361 	case IXGBE_DEV_ID_82599_T3_LOM:
2362 		/*
2363 		 * Since the warning interrupt is for both ports
2364 		 * we don't have to check if:
2365 		 *  - This interrupt wasn't for our port.
2366 		 *  - We may have missed the interrupt so always have to
2367 		 *    check if we  got a LSC
2368 		 */
2369 		if (!(eicr & IXGBE_EICR_GPI_SDP0) &&
2370 		    !(eicr & IXGBE_EICR_LSC))
2371 			return;
2372 
2373 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2374 			u32 speed;
2375 			bool link_up = false;
2376 
2377 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2378 
2379 			if (link_up)
2380 				return;
2381 		}
2382 
2383 		/* Check if this is not due to overtemp */
2384 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2385 			return;
2386 
2387 		break;
2388 	default:
2389 		if (!(eicr & IXGBE_EICR_GPI_SDP0))
2390 			return;
2391 		break;
2392 	}
2393 	e_crit(drv,
2394 	       "Network adapter has been stopped because it has over heated. "
2395 	       "Restart the computer. If the problem persists, "
2396 	       "power off the system and replace the adapter\n");
2397 
2398 	adapter->interrupt_event = 0;
2399 }
2400 
2401 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2402 {
2403 	struct ixgbe_hw *hw = &adapter->hw;
2404 
2405 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2406 	    (eicr & IXGBE_EICR_GPI_SDP1)) {
2407 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2408 		/* write to clear the interrupt */
2409 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2410 	}
2411 }
2412 
2413 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2414 {
2415 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2416 		return;
2417 
2418 	switch (adapter->hw.mac.type) {
2419 	case ixgbe_mac_82599EB:
2420 		/*
2421 		 * Need to check link state so complete overtemp check
2422 		 * on service task
2423 		 */
2424 		if (((eicr & IXGBE_EICR_GPI_SDP0) || (eicr & IXGBE_EICR_LSC)) &&
2425 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2426 			adapter->interrupt_event = eicr;
2427 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2428 			ixgbe_service_event_schedule(adapter);
2429 			return;
2430 		}
2431 		return;
2432 	case ixgbe_mac_X540:
2433 		if (!(eicr & IXGBE_EICR_TS))
2434 			return;
2435 		break;
2436 	default:
2437 		return;
2438 	}
2439 
2440 	e_crit(drv,
2441 	       "Network adapter has been stopped because it has over heated. "
2442 	       "Restart the computer. If the problem persists, "
2443 	       "power off the system and replace the adapter\n");
2444 }
2445 
2446 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2447 {
2448 	struct ixgbe_hw *hw = &adapter->hw;
2449 
2450 	if (eicr & IXGBE_EICR_GPI_SDP2) {
2451 		/* Clear the interrupt */
2452 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP2);
2453 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2454 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2455 			ixgbe_service_event_schedule(adapter);
2456 		}
2457 	}
2458 
2459 	if (eicr & IXGBE_EICR_GPI_SDP1) {
2460 		/* Clear the interrupt */
2461 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
2462 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2463 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2464 			ixgbe_service_event_schedule(adapter);
2465 		}
2466 	}
2467 }
2468 
2469 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2470 {
2471 	struct ixgbe_hw *hw = &adapter->hw;
2472 
2473 	adapter->lsc_int++;
2474 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2475 	adapter->link_check_timeout = jiffies;
2476 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2477 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2478 		IXGBE_WRITE_FLUSH(hw);
2479 		ixgbe_service_event_schedule(adapter);
2480 	}
2481 }
2482 
2483 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2484 					   u64 qmask)
2485 {
2486 	u32 mask;
2487 	struct ixgbe_hw *hw = &adapter->hw;
2488 
2489 	switch (hw->mac.type) {
2490 	case ixgbe_mac_82598EB:
2491 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2492 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2493 		break;
2494 	case ixgbe_mac_82599EB:
2495 	case ixgbe_mac_X540:
2496 	case ixgbe_mac_X550:
2497 	case ixgbe_mac_X550EM_x:
2498 		mask = (qmask & 0xFFFFFFFF);
2499 		if (mask)
2500 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2501 		mask = (qmask >> 32);
2502 		if (mask)
2503 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2504 		break;
2505 	default:
2506 		break;
2507 	}
2508 	/* skip the flush */
2509 }
2510 
2511 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2512 					    u64 qmask)
2513 {
2514 	u32 mask;
2515 	struct ixgbe_hw *hw = &adapter->hw;
2516 
2517 	switch (hw->mac.type) {
2518 	case ixgbe_mac_82598EB:
2519 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2520 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2521 		break;
2522 	case ixgbe_mac_82599EB:
2523 	case ixgbe_mac_X540:
2524 	case ixgbe_mac_X550:
2525 	case ixgbe_mac_X550EM_x:
2526 		mask = (qmask & 0xFFFFFFFF);
2527 		if (mask)
2528 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2529 		mask = (qmask >> 32);
2530 		if (mask)
2531 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2532 		break;
2533 	default:
2534 		break;
2535 	}
2536 	/* skip the flush */
2537 }
2538 
2539 /**
2540  * ixgbe_irq_enable - Enable default interrupt generation settings
2541  * @adapter: board private structure
2542  **/
2543 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2544 				    bool flush)
2545 {
2546 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2547 
2548 	/* don't reenable LSC while waiting for link */
2549 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2550 		mask &= ~IXGBE_EIMS_LSC;
2551 
2552 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2553 		switch (adapter->hw.mac.type) {
2554 		case ixgbe_mac_82599EB:
2555 			mask |= IXGBE_EIMS_GPI_SDP0;
2556 			break;
2557 		case ixgbe_mac_X540:
2558 		case ixgbe_mac_X550:
2559 		case ixgbe_mac_X550EM_x:
2560 			mask |= IXGBE_EIMS_TS;
2561 			break;
2562 		default:
2563 			break;
2564 		}
2565 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2566 		mask |= IXGBE_EIMS_GPI_SDP1;
2567 	switch (adapter->hw.mac.type) {
2568 	case ixgbe_mac_82599EB:
2569 		mask |= IXGBE_EIMS_GPI_SDP1;
2570 		mask |= IXGBE_EIMS_GPI_SDP2;
2571 		/* fall through */
2572 	case ixgbe_mac_X540:
2573 	case ixgbe_mac_X550:
2574 	case ixgbe_mac_X550EM_x:
2575 		mask |= IXGBE_EIMS_ECC;
2576 		mask |= IXGBE_EIMS_MAILBOX;
2577 		break;
2578 	default:
2579 		break;
2580 	}
2581 
2582 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
2583 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
2584 		mask |= IXGBE_EIMS_FLOW_DIR;
2585 
2586 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
2587 	if (queues)
2588 		ixgbe_irq_enable_queues(adapter, ~0);
2589 	if (flush)
2590 		IXGBE_WRITE_FLUSH(&adapter->hw);
2591 }
2592 
2593 static irqreturn_t ixgbe_msix_other(int irq, void *data)
2594 {
2595 	struct ixgbe_adapter *adapter = data;
2596 	struct ixgbe_hw *hw = &adapter->hw;
2597 	u32 eicr;
2598 
2599 	/*
2600 	 * Workaround for Silicon errata.  Use clear-by-write instead
2601 	 * of clear-by-read.  Reading with EICS will return the
2602 	 * interrupt causes without clearing, which later be done
2603 	 * with the write to EICR.
2604 	 */
2605 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
2606 
2607 	/* The lower 16bits of the EICR register are for the queue interrupts
2608 	 * which should be masked here in order to not accidentally clear them if
2609 	 * the bits are high when ixgbe_msix_other is called. There is a race
2610 	 * condition otherwise which results in possible performance loss
2611 	 * especially if the ixgbe_msix_other interrupt is triggering
2612 	 * consistently (as it would when PPS is turned on for the X540 device)
2613 	 */
2614 	eicr &= 0xFFFF0000;
2615 
2616 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
2617 
2618 	if (eicr & IXGBE_EICR_LSC)
2619 		ixgbe_check_lsc(adapter);
2620 
2621 	if (eicr & IXGBE_EICR_MAILBOX)
2622 		ixgbe_msg_task(adapter);
2623 
2624 	switch (hw->mac.type) {
2625 	case ixgbe_mac_82599EB:
2626 	case ixgbe_mac_X540:
2627 	case ixgbe_mac_X550:
2628 	case ixgbe_mac_X550EM_x:
2629 		if (eicr & IXGBE_EICR_ECC) {
2630 			e_info(link, "Received ECC Err, initiating reset\n");
2631 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
2632 			ixgbe_service_event_schedule(adapter);
2633 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2634 		}
2635 		/* Handle Flow Director Full threshold interrupt */
2636 		if (eicr & IXGBE_EICR_FLOW_DIR) {
2637 			int reinit_count = 0;
2638 			int i;
2639 			for (i = 0; i < adapter->num_tx_queues; i++) {
2640 				struct ixgbe_ring *ring = adapter->tx_ring[i];
2641 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
2642 						       &ring->state))
2643 					reinit_count++;
2644 			}
2645 			if (reinit_count) {
2646 				/* no more flow director interrupts until after init */
2647 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
2648 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
2649 				ixgbe_service_event_schedule(adapter);
2650 			}
2651 		}
2652 		ixgbe_check_sfp_event(adapter, eicr);
2653 		ixgbe_check_overtemp_event(adapter, eicr);
2654 		break;
2655 	default:
2656 		break;
2657 	}
2658 
2659 	ixgbe_check_fan_failure(adapter, eicr);
2660 
2661 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2662 		ixgbe_ptp_check_pps_event(adapter, eicr);
2663 
2664 	/* re-enable the original interrupt state, no lsc, no queues */
2665 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2666 		ixgbe_irq_enable(adapter, false, false);
2667 
2668 	return IRQ_HANDLED;
2669 }
2670 
2671 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
2672 {
2673 	struct ixgbe_q_vector *q_vector = data;
2674 
2675 	/* EIAM disabled interrupts (on this vector) for us */
2676 
2677 	if (q_vector->rx.ring || q_vector->tx.ring)
2678 		napi_schedule(&q_vector->napi);
2679 
2680 	return IRQ_HANDLED;
2681 }
2682 
2683 /**
2684  * ixgbe_poll - NAPI Rx polling callback
2685  * @napi: structure for representing this polling device
2686  * @budget: how many packets driver is allowed to clean
2687  *
2688  * This function is used for legacy and MSI, NAPI mode
2689  **/
2690 int ixgbe_poll(struct napi_struct *napi, int budget)
2691 {
2692 	struct ixgbe_q_vector *q_vector =
2693 				container_of(napi, struct ixgbe_q_vector, napi);
2694 	struct ixgbe_adapter *adapter = q_vector->adapter;
2695 	struct ixgbe_ring *ring;
2696 	int per_ring_budget;
2697 	bool clean_complete = true;
2698 
2699 #ifdef CONFIG_IXGBE_DCA
2700 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
2701 		ixgbe_update_dca(q_vector);
2702 #endif
2703 
2704 	ixgbe_for_each_ring(ring, q_vector->tx)
2705 		clean_complete &= !!ixgbe_clean_tx_irq(q_vector, ring);
2706 
2707 	if (!ixgbe_qv_lock_napi(q_vector))
2708 		return budget;
2709 
2710 	/* attempt to distribute budget to each queue fairly, but don't allow
2711 	 * the budget to go below 1 because we'll exit polling */
2712 	if (q_vector->rx.count > 1)
2713 		per_ring_budget = max(budget/q_vector->rx.count, 1);
2714 	else
2715 		per_ring_budget = budget;
2716 
2717 	ixgbe_for_each_ring(ring, q_vector->rx)
2718 		clean_complete &= (ixgbe_clean_rx_irq(q_vector, ring,
2719 				   per_ring_budget) < per_ring_budget);
2720 
2721 	ixgbe_qv_unlock_napi(q_vector);
2722 	/* If all work not completed, return budget and keep polling */
2723 	if (!clean_complete)
2724 		return budget;
2725 
2726 	/* all work done, exit the polling mode */
2727 	napi_complete(napi);
2728 	if (adapter->rx_itr_setting & 1)
2729 		ixgbe_set_itr(q_vector);
2730 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2731 		ixgbe_irq_enable_queues(adapter, ((u64)1 << q_vector->v_idx));
2732 
2733 	return 0;
2734 }
2735 
2736 /**
2737  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2738  * @adapter: board private structure
2739  *
2740  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2741  * interrupts from the kernel.
2742  **/
2743 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
2744 {
2745 	struct net_device *netdev = adapter->netdev;
2746 	int vector, err;
2747 	int ri = 0, ti = 0;
2748 
2749 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2750 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2751 		struct msix_entry *entry = &adapter->msix_entries[vector];
2752 
2753 		if (q_vector->tx.ring && q_vector->rx.ring) {
2754 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2755 				 "%s-%s-%d", netdev->name, "TxRx", ri++);
2756 			ti++;
2757 		} else if (q_vector->rx.ring) {
2758 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2759 				 "%s-%s-%d", netdev->name, "rx", ri++);
2760 		} else if (q_vector->tx.ring) {
2761 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2762 				 "%s-%s-%d", netdev->name, "tx", ti++);
2763 		} else {
2764 			/* skip this unused q_vector */
2765 			continue;
2766 		}
2767 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
2768 				  q_vector->name, q_vector);
2769 		if (err) {
2770 			e_err(probe, "request_irq failed for MSIX interrupt "
2771 			      "Error: %d\n", err);
2772 			goto free_queue_irqs;
2773 		}
2774 		/* If Flow Director is enabled, set interrupt affinity */
2775 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
2776 			/* assign the mask for this irq */
2777 			irq_set_affinity_hint(entry->vector,
2778 					      &q_vector->affinity_mask);
2779 		}
2780 	}
2781 
2782 	err = request_irq(adapter->msix_entries[vector].vector,
2783 			  ixgbe_msix_other, 0, netdev->name, adapter);
2784 	if (err) {
2785 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
2786 		goto free_queue_irqs;
2787 	}
2788 
2789 	return 0;
2790 
2791 free_queue_irqs:
2792 	while (vector) {
2793 		vector--;
2794 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
2795 				      NULL);
2796 		free_irq(adapter->msix_entries[vector].vector,
2797 			 adapter->q_vector[vector]);
2798 	}
2799 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
2800 	pci_disable_msix(adapter->pdev);
2801 	kfree(adapter->msix_entries);
2802 	adapter->msix_entries = NULL;
2803 	return err;
2804 }
2805 
2806 /**
2807  * ixgbe_intr - legacy mode Interrupt Handler
2808  * @irq: interrupt number
2809  * @data: pointer to a network interface device structure
2810  **/
2811 static irqreturn_t ixgbe_intr(int irq, void *data)
2812 {
2813 	struct ixgbe_adapter *adapter = data;
2814 	struct ixgbe_hw *hw = &adapter->hw;
2815 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2816 	u32 eicr;
2817 
2818 	/*
2819 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
2820 	 * before the read of EICR.
2821 	 */
2822 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
2823 
2824 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2825 	 * therefore no explicit interrupt disable is necessary */
2826 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2827 	if (!eicr) {
2828 		/*
2829 		 * shared interrupt alert!
2830 		 * make sure interrupts are enabled because the read will
2831 		 * have disabled interrupts due to EIAM
2832 		 * finish the workaround of silicon errata on 82598.  Unmask
2833 		 * the interrupt that we masked before the EICR read.
2834 		 */
2835 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
2836 			ixgbe_irq_enable(adapter, true, true);
2837 		return IRQ_NONE;	/* Not our interrupt */
2838 	}
2839 
2840 	if (eicr & IXGBE_EICR_LSC)
2841 		ixgbe_check_lsc(adapter);
2842 
2843 	switch (hw->mac.type) {
2844 	case ixgbe_mac_82599EB:
2845 		ixgbe_check_sfp_event(adapter, eicr);
2846 		/* Fall through */
2847 	case ixgbe_mac_X540:
2848 	case ixgbe_mac_X550:
2849 	case ixgbe_mac_X550EM_x:
2850 		if (eicr & IXGBE_EICR_ECC) {
2851 			e_info(link, "Received ECC Err, initiating reset\n");
2852 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
2853 			ixgbe_service_event_schedule(adapter);
2854 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
2855 		}
2856 		ixgbe_check_overtemp_event(adapter, eicr);
2857 		break;
2858 	default:
2859 		break;
2860 	}
2861 
2862 	ixgbe_check_fan_failure(adapter, eicr);
2863 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
2864 		ixgbe_ptp_check_pps_event(adapter, eicr);
2865 
2866 	/* would disable interrupts here but EIAM disabled it */
2867 	napi_schedule(&q_vector->napi);
2868 
2869 	/*
2870 	 * re-enable link(maybe) and non-queue interrupts, no flush.
2871 	 * ixgbe_poll will re-enable the queue interrupts
2872 	 */
2873 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
2874 		ixgbe_irq_enable(adapter, false, false);
2875 
2876 	return IRQ_HANDLED;
2877 }
2878 
2879 /**
2880  * ixgbe_request_irq - initialize interrupts
2881  * @adapter: board private structure
2882  *
2883  * Attempts to configure interrupts using the best available
2884  * capabilities of the hardware and kernel.
2885  **/
2886 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
2887 {
2888 	struct net_device *netdev = adapter->netdev;
2889 	int err;
2890 
2891 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
2892 		err = ixgbe_request_msix_irqs(adapter);
2893 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
2894 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
2895 				  netdev->name, adapter);
2896 	else
2897 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
2898 				  netdev->name, adapter);
2899 
2900 	if (err)
2901 		e_err(probe, "request_irq failed, Error %d\n", err);
2902 
2903 	return err;
2904 }
2905 
2906 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
2907 {
2908 	int vector;
2909 
2910 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
2911 		free_irq(adapter->pdev->irq, adapter);
2912 		return;
2913 	}
2914 
2915 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
2916 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
2917 		struct msix_entry *entry = &adapter->msix_entries[vector];
2918 
2919 		/* free only the irqs that were actually requested */
2920 		if (!q_vector->rx.ring && !q_vector->tx.ring)
2921 			continue;
2922 
2923 		/* clear the affinity_mask in the IRQ descriptor */
2924 		irq_set_affinity_hint(entry->vector, NULL);
2925 
2926 		free_irq(entry->vector, q_vector);
2927 	}
2928 
2929 	free_irq(adapter->msix_entries[vector++].vector, adapter);
2930 }
2931 
2932 /**
2933  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
2934  * @adapter: board private structure
2935  **/
2936 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
2937 {
2938 	switch (adapter->hw.mac.type) {
2939 	case ixgbe_mac_82598EB:
2940 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
2941 		break;
2942 	case ixgbe_mac_82599EB:
2943 	case ixgbe_mac_X540:
2944 	case ixgbe_mac_X550:
2945 	case ixgbe_mac_X550EM_x:
2946 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
2947 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
2948 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
2949 		break;
2950 	default:
2951 		break;
2952 	}
2953 	IXGBE_WRITE_FLUSH(&adapter->hw);
2954 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
2955 		int vector;
2956 
2957 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
2958 			synchronize_irq(adapter->msix_entries[vector].vector);
2959 
2960 		synchronize_irq(adapter->msix_entries[vector++].vector);
2961 	} else {
2962 		synchronize_irq(adapter->pdev->irq);
2963 	}
2964 }
2965 
2966 /**
2967  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
2968  *
2969  **/
2970 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
2971 {
2972 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
2973 
2974 	ixgbe_write_eitr(q_vector);
2975 
2976 	ixgbe_set_ivar(adapter, 0, 0, 0);
2977 	ixgbe_set_ivar(adapter, 1, 0, 0);
2978 
2979 	e_info(hw, "Legacy interrupt IVAR setup done\n");
2980 }
2981 
2982 /**
2983  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
2984  * @adapter: board private structure
2985  * @ring: structure containing ring specific data
2986  *
2987  * Configure the Tx descriptor ring after a reset.
2988  **/
2989 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
2990 			     struct ixgbe_ring *ring)
2991 {
2992 	struct ixgbe_hw *hw = &adapter->hw;
2993 	u64 tdba = ring->dma;
2994 	int wait_loop = 10;
2995 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
2996 	u8 reg_idx = ring->reg_idx;
2997 
2998 	/* disable queue to avoid issues while updating state */
2999 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3000 	IXGBE_WRITE_FLUSH(hw);
3001 
3002 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3003 			(tdba & DMA_BIT_MASK(32)));
3004 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3005 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3006 			ring->count * sizeof(union ixgbe_adv_tx_desc));
3007 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3008 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3009 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3010 
3011 	/*
3012 	 * set WTHRESH to encourage burst writeback, it should not be set
3013 	 * higher than 1 when:
3014 	 * - ITR is 0 as it could cause false TX hangs
3015 	 * - ITR is set to > 100k int/sec and BQL is enabled
3016 	 *
3017 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3018 	 * to or less than the number of on chip descriptors, which is
3019 	 * currently 40.
3020 	 */
3021 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3022 		txdctl |= (1 << 16);	/* WTHRESH = 1 */
3023 	else
3024 		txdctl |= (8 << 16);	/* WTHRESH = 8 */
3025 
3026 	/*
3027 	 * Setting PTHRESH to 32 both improves performance
3028 	 * and avoids a TX hang with DFP enabled
3029 	 */
3030 	txdctl |= (1 << 8) |	/* HTHRESH = 1 */
3031 		   32;		/* PTHRESH = 32 */
3032 
3033 	/* reinitialize flowdirector state */
3034 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3035 		ring->atr_sample_rate = adapter->atr_sample_rate;
3036 		ring->atr_count = 0;
3037 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3038 	} else {
3039 		ring->atr_sample_rate = 0;
3040 	}
3041 
3042 	/* initialize XPS */
3043 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3044 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3045 
3046 		if (q_vector)
3047 			netif_set_xps_queue(ring->netdev,
3048 					    &q_vector->affinity_mask,
3049 					    ring->queue_index);
3050 	}
3051 
3052 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3053 
3054 	/* enable queue */
3055 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3056 
3057 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3058 	if (hw->mac.type == ixgbe_mac_82598EB &&
3059 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3060 		return;
3061 
3062 	/* poll to verify queue is enabled */
3063 	do {
3064 		usleep_range(1000, 2000);
3065 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3066 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3067 	if (!wait_loop)
3068 		e_err(drv, "Could not enable Tx Queue %d\n", reg_idx);
3069 }
3070 
3071 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3072 {
3073 	struct ixgbe_hw *hw = &adapter->hw;
3074 	u32 rttdcs, mtqc;
3075 	u8 tcs = netdev_get_num_tc(adapter->netdev);
3076 
3077 	if (hw->mac.type == ixgbe_mac_82598EB)
3078 		return;
3079 
3080 	/* disable the arbiter while setting MTQC */
3081 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3082 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3083 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3084 
3085 	/* set transmit pool layout */
3086 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3087 		mtqc = IXGBE_MTQC_VT_ENA;
3088 		if (tcs > 4)
3089 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3090 		else if (tcs > 1)
3091 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3092 		else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3093 			mtqc |= IXGBE_MTQC_32VF;
3094 		else
3095 			mtqc |= IXGBE_MTQC_64VF;
3096 	} else {
3097 		if (tcs > 4)
3098 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3099 		else if (tcs > 1)
3100 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3101 		else
3102 			mtqc = IXGBE_MTQC_64Q_1PB;
3103 	}
3104 
3105 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3106 
3107 	/* Enable Security TX Buffer IFG for multiple pb */
3108 	if (tcs) {
3109 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3110 		sectx |= IXGBE_SECTX_DCB;
3111 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3112 	}
3113 
3114 	/* re-enable the arbiter */
3115 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3116 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3117 }
3118 
3119 /**
3120  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3121  * @adapter: board private structure
3122  *
3123  * Configure the Tx unit of the MAC after a reset.
3124  **/
3125 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3126 {
3127 	struct ixgbe_hw *hw = &adapter->hw;
3128 	u32 dmatxctl;
3129 	u32 i;
3130 
3131 	ixgbe_setup_mtqc(adapter);
3132 
3133 	if (hw->mac.type != ixgbe_mac_82598EB) {
3134 		/* DMATXCTL.EN must be before Tx queues are enabled */
3135 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3136 		dmatxctl |= IXGBE_DMATXCTL_TE;
3137 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3138 	}
3139 
3140 	/* Setup the HW Tx Head and Tail descriptor pointers */
3141 	for (i = 0; i < adapter->num_tx_queues; i++)
3142 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3143 }
3144 
3145 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3146 				 struct ixgbe_ring *ring)
3147 {
3148 	struct ixgbe_hw *hw = &adapter->hw;
3149 	u8 reg_idx = ring->reg_idx;
3150 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3151 
3152 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3153 
3154 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3155 }
3156 
3157 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3158 				  struct ixgbe_ring *ring)
3159 {
3160 	struct ixgbe_hw *hw = &adapter->hw;
3161 	u8 reg_idx = ring->reg_idx;
3162 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3163 
3164 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3165 
3166 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3167 }
3168 
3169 #ifdef CONFIG_IXGBE_DCB
3170 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3171 #else
3172 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3173 #endif
3174 {
3175 	int i;
3176 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3177 
3178 	if (adapter->ixgbe_ieee_pfc)
3179 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3180 
3181 	/*
3182 	 * We should set the drop enable bit if:
3183 	 *  SR-IOV is enabled
3184 	 *   or
3185 	 *  Number of Rx queues > 1 and flow control is disabled
3186 	 *
3187 	 *  This allows us to avoid head of line blocking for security
3188 	 *  and performance reasons.
3189 	 */
3190 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3191 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3192 		for (i = 0; i < adapter->num_rx_queues; i++)
3193 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3194 	} else {
3195 		for (i = 0; i < adapter->num_rx_queues; i++)
3196 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3197 	}
3198 }
3199 
3200 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3201 
3202 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3203 				   struct ixgbe_ring *rx_ring)
3204 {
3205 	struct ixgbe_hw *hw = &adapter->hw;
3206 	u32 srrctl;
3207 	u8 reg_idx = rx_ring->reg_idx;
3208 
3209 	if (hw->mac.type == ixgbe_mac_82598EB) {
3210 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3211 
3212 		/*
3213 		 * if VMDq is not active we must program one srrctl register
3214 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3215 		 */
3216 		reg_idx &= mask;
3217 	}
3218 
3219 	/* configure header buffer length, needed for RSC */
3220 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3221 
3222 	/* configure the packet buffer length */
3223 	srrctl |= ixgbe_rx_bufsz(rx_ring) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3224 
3225 	/* configure descriptor type */
3226 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3227 
3228 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3229 }
3230 
3231 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter, const u32 *seed)
3232 {
3233 	struct ixgbe_hw *hw = &adapter->hw;
3234 	u32 reta = 0;
3235 	int i, j;
3236 	int reta_entries = 128;
3237 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3238 	int indices_multi;
3239 
3240 	/*
3241 	 * Program table for at least 2 queues w/ SR-IOV so that VFs can
3242 	 * make full use of any rings they may have.  We will use the
3243 	 * PSRTYPE register to control how many rings we use within the PF.
3244 	 */
3245 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 2))
3246 		rss_i = 2;
3247 
3248 	/* Fill out hash function seeds */
3249 	for (i = 0; i < 10; i++)
3250 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), seed[i]);
3251 
3252 	/* Fill out the redirection table as follows:
3253 	 * 82598: 128 (8 bit wide) entries containing pair of 4 bit RSS indices
3254 	 * 82599/X540: 128 (8 bit wide) entries containing 4 bit RSS index
3255 	 * X550: 512 (8 bit wide) entries containing 6 bit RSS index
3256 	 */
3257 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3258 		indices_multi = 0x11;
3259 	else
3260 		indices_multi = 0x1;
3261 
3262 	switch (adapter->hw.mac.type) {
3263 	case ixgbe_mac_X550:
3264 	case ixgbe_mac_X550EM_x:
3265 		if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3266 			reta_entries = 512;
3267 	default:
3268 		break;
3269 	}
3270 
3271 	/* Fill out redirection table */
3272 	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3273 		if (j == rss_i)
3274 			j = 0;
3275 		reta = (reta << 8) | (j * indices_multi);
3276 		if ((i & 3) == 3) {
3277 			if (i < 128)
3278 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3279 			else
3280 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3281 						reta);
3282 		}
3283 	}
3284 }
3285 
3286 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter, const u32 *seed)
3287 {
3288 	struct ixgbe_hw *hw = &adapter->hw;
3289 	u32 vfreta = 0;
3290 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3291 	unsigned int pf_pool = adapter->num_vfs;
3292 	int i, j;
3293 
3294 	/* Fill out hash function seeds */
3295 	for (i = 0; i < 10; i++)
3296 		IXGBE_WRITE_REG(hw, IXGBE_PFVFRSSRK(i, pf_pool), seed[i]);
3297 
3298 	/* Fill out the redirection table */
3299 	for (i = 0, j = 0; i < 64; i++, j++) {
3300 		if (j == rss_i)
3301 			j = 0;
3302 		vfreta = (vfreta << 8) | j;
3303 		if ((i & 3) == 3)
3304 			IXGBE_WRITE_REG(hw, IXGBE_PFVFRETA(i >> 2, pf_pool),
3305 					vfreta);
3306 	}
3307 }
3308 
3309 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3310 {
3311 	struct ixgbe_hw *hw = &adapter->hw;
3312 	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3313 	u32 rss_key[10];
3314 	u32 rxcsum;
3315 
3316 	/* Disable indicating checksum in descriptor, enables RSS hash */
3317 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3318 	rxcsum |= IXGBE_RXCSUM_PCSD;
3319 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3320 
3321 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3322 		if (adapter->ring_feature[RING_F_RSS].mask)
3323 			mrqc = IXGBE_MRQC_RSSEN;
3324 	} else {
3325 		u8 tcs = netdev_get_num_tc(adapter->netdev);
3326 
3327 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3328 			if (tcs > 4)
3329 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3330 			else if (tcs > 1)
3331 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3332 			else if (adapter->ring_feature[RING_F_RSS].indices == 4)
3333 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3334 			else
3335 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3336 		} else {
3337 			if (tcs > 4)
3338 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3339 			else if (tcs > 1)
3340 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3341 			else
3342 				mrqc = IXGBE_MRQC_RSSEN;
3343 		}
3344 	}
3345 
3346 	/* Perform hash on these packet types */
3347 	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3348 		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3349 		     IXGBE_MRQC_RSS_FIELD_IPV6 |
3350 		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3351 
3352 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3353 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3354 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3355 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3356 
3357 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3358 	if ((hw->mac.type >= ixgbe_mac_X550) &&
3359 	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
3360 		unsigned int pf_pool = adapter->num_vfs;
3361 
3362 		/* Enable VF RSS mode */
3363 		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
3364 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3365 
3366 		/* Setup RSS through the VF registers */
3367 		ixgbe_setup_vfreta(adapter, rss_key);
3368 		vfmrqc = IXGBE_MRQC_RSSEN;
3369 		vfmrqc |= rss_field;
3370 		IXGBE_WRITE_REG(hw, IXGBE_PFVFMRQC(pf_pool), vfmrqc);
3371 	} else {
3372 		ixgbe_setup_reta(adapter, rss_key);
3373 		mrqc |= rss_field;
3374 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3375 	}
3376 }
3377 
3378 /**
3379  * ixgbe_configure_rscctl - enable RSC for the indicated ring
3380  * @adapter:    address of board private structure
3381  * @index:      index of ring to set
3382  **/
3383 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
3384 				   struct ixgbe_ring *ring)
3385 {
3386 	struct ixgbe_hw *hw = &adapter->hw;
3387 	u32 rscctrl;
3388 	u8 reg_idx = ring->reg_idx;
3389 
3390 	if (!ring_is_rsc_enabled(ring))
3391 		return;
3392 
3393 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
3394 	rscctrl |= IXGBE_RSCCTL_RSCEN;
3395 	/*
3396 	 * we must limit the number of descriptors so that the
3397 	 * total size of max desc * buf_len is not greater
3398 	 * than 65536
3399 	 */
3400 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
3401 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
3402 }
3403 
3404 #define IXGBE_MAX_RX_DESC_POLL 10
3405 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
3406 				       struct ixgbe_ring *ring)
3407 {
3408 	struct ixgbe_hw *hw = &adapter->hw;
3409 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3410 	u32 rxdctl;
3411 	u8 reg_idx = ring->reg_idx;
3412 
3413 	if (ixgbe_removed(hw->hw_addr))
3414 		return;
3415 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3416 	if (hw->mac.type == ixgbe_mac_82598EB &&
3417 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3418 		return;
3419 
3420 	do {
3421 		usleep_range(1000, 2000);
3422 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3423 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
3424 
3425 	if (!wait_loop) {
3426 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
3427 		      "the polling period\n", reg_idx);
3428 	}
3429 }
3430 
3431 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
3432 			    struct ixgbe_ring *ring)
3433 {
3434 	struct ixgbe_hw *hw = &adapter->hw;
3435 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
3436 	u32 rxdctl;
3437 	u8 reg_idx = ring->reg_idx;
3438 
3439 	if (ixgbe_removed(hw->hw_addr))
3440 		return;
3441 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3442 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
3443 
3444 	/* write value back with RXDCTL.ENABLE bit cleared */
3445 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3446 
3447 	if (hw->mac.type == ixgbe_mac_82598EB &&
3448 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3449 		return;
3450 
3451 	/* the hardware may take up to 100us to really disable the rx queue */
3452 	do {
3453 		udelay(10);
3454 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3455 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
3456 
3457 	if (!wait_loop) {
3458 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3459 		      "the polling period\n", reg_idx);
3460 	}
3461 }
3462 
3463 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
3464 			     struct ixgbe_ring *ring)
3465 {
3466 	struct ixgbe_hw *hw = &adapter->hw;
3467 	u64 rdba = ring->dma;
3468 	u32 rxdctl;
3469 	u8 reg_idx = ring->reg_idx;
3470 
3471 	/* disable queue to avoid issues while updating state */
3472 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
3473 	ixgbe_disable_rx_queue(adapter, ring);
3474 
3475 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
3476 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
3477 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
3478 			ring->count * sizeof(union ixgbe_adv_rx_desc));
3479 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
3480 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
3481 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
3482 
3483 	ixgbe_configure_srrctl(adapter, ring);
3484 	ixgbe_configure_rscctl(adapter, ring);
3485 
3486 	if (hw->mac.type == ixgbe_mac_82598EB) {
3487 		/*
3488 		 * enable cache line friendly hardware writes:
3489 		 * PTHRESH=32 descriptors (half the internal cache),
3490 		 * this also removes ugly rx_no_buffer_count increment
3491 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
3492 		 * WTHRESH=8 burst writeback up to two cache lines
3493 		 */
3494 		rxdctl &= ~0x3FFFFF;
3495 		rxdctl |=  0x080420;
3496 	}
3497 
3498 	/* enable receive descriptor ring */
3499 	rxdctl |= IXGBE_RXDCTL_ENABLE;
3500 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
3501 
3502 	ixgbe_rx_desc_queue_enable(adapter, ring);
3503 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
3504 }
3505 
3506 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
3507 {
3508 	struct ixgbe_hw *hw = &adapter->hw;
3509 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
3510 	u16 pool;
3511 
3512 	/* PSRTYPE must be initialized in non 82598 adapters */
3513 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
3514 		      IXGBE_PSRTYPE_UDPHDR |
3515 		      IXGBE_PSRTYPE_IPV4HDR |
3516 		      IXGBE_PSRTYPE_L2HDR |
3517 		      IXGBE_PSRTYPE_IPV6HDR;
3518 
3519 	if (hw->mac.type == ixgbe_mac_82598EB)
3520 		return;
3521 
3522 	if (rss_i > 3)
3523 		psrtype |= 2 << 29;
3524 	else if (rss_i > 1)
3525 		psrtype |= 1 << 29;
3526 
3527 	for_each_set_bit(pool, &adapter->fwd_bitmask, 32)
3528 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
3529 }
3530 
3531 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
3532 {
3533 	struct ixgbe_hw *hw = &adapter->hw;
3534 	u32 reg_offset, vf_shift;
3535 	u32 gcr_ext, vmdctl;
3536 	int i;
3537 
3538 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
3539 		return;
3540 
3541 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3542 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
3543 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
3544 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
3545 	vmdctl |= IXGBE_VT_CTL_REPLEN;
3546 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
3547 
3548 	vf_shift = VMDQ_P(0) % 32;
3549 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
3550 
3551 	/* Enable only the PF's pool for Tx/Rx */
3552 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), (~0) << vf_shift);
3553 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
3554 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), (~0) << vf_shift);
3555 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
3556 	if (adapter->flags2 & IXGBE_FLAG2_BRIDGE_MODE_VEB)
3557 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
3558 
3559 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3560 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
3561 
3562 	/*
3563 	 * Set up VF register offsets for selected VT Mode,
3564 	 * i.e. 32 or 64 VFs for SR-IOV
3565 	 */
3566 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
3567 	case IXGBE_82599_VMDQ_8Q_MASK:
3568 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
3569 		break;
3570 	case IXGBE_82599_VMDQ_4Q_MASK:
3571 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
3572 		break;
3573 	default:
3574 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
3575 		break;
3576 	}
3577 
3578 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
3579 
3580 
3581 	/* Enable MAC Anti-Spoofing */
3582 	hw->mac.ops.set_mac_anti_spoofing(hw, (adapter->num_vfs != 0),
3583 					  adapter->num_vfs);
3584 
3585 	/* Ensure LLDP is set for Ethertype Antispoofing if we will be
3586 	 * calling set_ethertype_anti_spoofing for each VF in loop below
3587 	 */
3588 	if (hw->mac.ops.set_ethertype_anti_spoofing)
3589 		IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_LLDP),
3590 				(IXGBE_ETQF_FILTER_EN    | /* enable filter */
3591 				 IXGBE_ETQF_TX_ANTISPOOF | /* tx antispoof */
3592 				 IXGBE_ETH_P_LLDP));	   /* LLDP eth type */
3593 
3594 	/* For VFs that have spoof checking turned off */
3595 	for (i = 0; i < adapter->num_vfs; i++) {
3596 		if (!adapter->vfinfo[i].spoofchk_enabled)
3597 			ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i, false);
3598 
3599 		/* enable ethertype anti spoofing if hw supports it */
3600 		if (hw->mac.ops.set_ethertype_anti_spoofing)
3601 			hw->mac.ops.set_ethertype_anti_spoofing(hw, true, i);
3602 	}
3603 }
3604 
3605 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
3606 {
3607 	struct ixgbe_hw *hw = &adapter->hw;
3608 	struct net_device *netdev = adapter->netdev;
3609 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
3610 	struct ixgbe_ring *rx_ring;
3611 	int i;
3612 	u32 mhadd, hlreg0;
3613 
3614 #ifdef IXGBE_FCOE
3615 	/* adjust max frame to be able to do baby jumbo for FCoE */
3616 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
3617 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
3618 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
3619 
3620 #endif /* IXGBE_FCOE */
3621 
3622 	/* adjust max frame to be at least the size of a standard frame */
3623 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
3624 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
3625 
3626 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
3627 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
3628 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
3629 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
3630 
3631 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
3632 	}
3633 
3634 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3635 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3636 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3637 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3638 
3639 	/*
3640 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3641 	 * the Base and Length of the Rx Descriptor Ring
3642 	 */
3643 	for (i = 0; i < adapter->num_rx_queues; i++) {
3644 		rx_ring = adapter->rx_ring[i];
3645 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
3646 			set_ring_rsc_enabled(rx_ring);
3647 		else
3648 			clear_ring_rsc_enabled(rx_ring);
3649 	}
3650 }
3651 
3652 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
3653 {
3654 	struct ixgbe_hw *hw = &adapter->hw;
3655 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
3656 
3657 	switch (hw->mac.type) {
3658 	case ixgbe_mac_X550:
3659 	case ixgbe_mac_X550EM_x:
3660 	case ixgbe_mac_82598EB:
3661 		/*
3662 		 * For VMDq support of different descriptor types or
3663 		 * buffer sizes through the use of multiple SRRCTL
3664 		 * registers, RDRXCTL.MVMEN must be set to 1
3665 		 *
3666 		 * also, the manual doesn't mention it clearly but DCA hints
3667 		 * will only use queue 0's tags unless this bit is set.  Side
3668 		 * effects of setting this bit are only that SRRCTL must be
3669 		 * fully programmed [0..15]
3670 		 */
3671 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
3672 		break;
3673 	case ixgbe_mac_82599EB:
3674 	case ixgbe_mac_X540:
3675 		/* Disable RSC for ACK packets */
3676 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
3677 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
3678 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
3679 		/* hardware requires some bits to be set by default */
3680 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
3681 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
3682 		break;
3683 	default:
3684 		/* We should do nothing since we don't know this hardware */
3685 		return;
3686 	}
3687 
3688 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
3689 }
3690 
3691 /**
3692  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3693  * @adapter: board private structure
3694  *
3695  * Configure the Rx unit of the MAC after a reset.
3696  **/
3697 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
3698 {
3699 	struct ixgbe_hw *hw = &adapter->hw;
3700 	int i;
3701 	u32 rxctrl, rfctl;
3702 
3703 	/* disable receives while setting up the descriptors */
3704 	hw->mac.ops.disable_rx(hw);
3705 
3706 	ixgbe_setup_psrtype(adapter);
3707 	ixgbe_setup_rdrxctl(adapter);
3708 
3709 	/* RSC Setup */
3710 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
3711 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
3712 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
3713 		rfctl |= IXGBE_RFCTL_RSC_DIS;
3714 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
3715 
3716 	/* Program registers for the distribution of queues */
3717 	ixgbe_setup_mrqc(adapter);
3718 
3719 	/* set_rx_buffer_len must be called before ring initialization */
3720 	ixgbe_set_rx_buffer_len(adapter);
3721 
3722 	/*
3723 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
3724 	 * the Base and Length of the Rx Descriptor Ring
3725 	 */
3726 	for (i = 0; i < adapter->num_rx_queues; i++)
3727 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
3728 
3729 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
3730 	/* disable drop enable for 82598 parts */
3731 	if (hw->mac.type == ixgbe_mac_82598EB)
3732 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
3733 
3734 	/* enable all receives */
3735 	rxctrl |= IXGBE_RXCTRL_RXEN;
3736 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
3737 }
3738 
3739 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
3740 				 __be16 proto, u16 vid)
3741 {
3742 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3743 	struct ixgbe_hw *hw = &adapter->hw;
3744 
3745 	/* add VID to filter table */
3746 	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true);
3747 	set_bit(vid, adapter->active_vlans);
3748 
3749 	return 0;
3750 }
3751 
3752 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
3753 				  __be16 proto, u16 vid)
3754 {
3755 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3756 	struct ixgbe_hw *hw = &adapter->hw;
3757 
3758 	/* remove VID from filter table */
3759 	hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), false);
3760 	clear_bit(vid, adapter->active_vlans);
3761 
3762 	return 0;
3763 }
3764 
3765 /**
3766  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
3767  * @adapter: driver data
3768  */
3769 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
3770 {
3771 	struct ixgbe_hw *hw = &adapter->hw;
3772 	u32 vlnctrl;
3773 	int i, j;
3774 
3775 	switch (hw->mac.type) {
3776 	case ixgbe_mac_82598EB:
3777 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3778 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
3779 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3780 		break;
3781 	case ixgbe_mac_82599EB:
3782 	case ixgbe_mac_X540:
3783 	case ixgbe_mac_X550:
3784 	case ixgbe_mac_X550EM_x:
3785 		for (i = 0; i < adapter->num_rx_queues; i++) {
3786 			struct ixgbe_ring *ring = adapter->rx_ring[i];
3787 
3788 			if (ring->l2_accel_priv)
3789 				continue;
3790 			j = ring->reg_idx;
3791 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3792 			vlnctrl &= ~IXGBE_RXDCTL_VME;
3793 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3794 		}
3795 		break;
3796 	default:
3797 		break;
3798 	}
3799 }
3800 
3801 /**
3802  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
3803  * @adapter: driver data
3804  */
3805 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
3806 {
3807 	struct ixgbe_hw *hw = &adapter->hw;
3808 	u32 vlnctrl;
3809 	int i, j;
3810 
3811 	switch (hw->mac.type) {
3812 	case ixgbe_mac_82598EB:
3813 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
3814 		vlnctrl |= IXGBE_VLNCTRL_VME;
3815 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
3816 		break;
3817 	case ixgbe_mac_82599EB:
3818 	case ixgbe_mac_X540:
3819 	case ixgbe_mac_X550:
3820 	case ixgbe_mac_X550EM_x:
3821 		for (i = 0; i < adapter->num_rx_queues; i++) {
3822 			struct ixgbe_ring *ring = adapter->rx_ring[i];
3823 
3824 			if (ring->l2_accel_priv)
3825 				continue;
3826 			j = ring->reg_idx;
3827 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
3828 			vlnctrl |= IXGBE_RXDCTL_VME;
3829 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
3830 		}
3831 		break;
3832 	default:
3833 		break;
3834 	}
3835 }
3836 
3837 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
3838 {
3839 	u16 vid;
3840 
3841 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
3842 
3843 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
3844 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
3845 }
3846 
3847 /**
3848  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
3849  * @netdev: network interface device structure
3850  *
3851  * Writes multicast address list to the MTA hash table.
3852  * Returns: -ENOMEM on failure
3853  *                0 on no addresses written
3854  *                X on writing X addresses to MTA
3855  **/
3856 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
3857 {
3858 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
3859 	struct ixgbe_hw *hw = &adapter->hw;
3860 
3861 	if (!netif_running(netdev))
3862 		return 0;
3863 
3864 	if (hw->mac.ops.update_mc_addr_list)
3865 		hw->mac.ops.update_mc_addr_list(hw, netdev);
3866 	else
3867 		return -ENOMEM;
3868 
3869 #ifdef CONFIG_PCI_IOV
3870 	ixgbe_restore_vf_multicasts(adapter);
3871 #endif
3872 
3873 	return netdev_mc_count(netdev);
3874 }
3875 
3876 #ifdef CONFIG_PCI_IOV
3877 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
3878 {
3879 	struct ixgbe_hw *hw = &adapter->hw;
3880 	int i;
3881 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3882 		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_IN_USE)
3883 			hw->mac.ops.set_rar(hw, i, adapter->mac_table[i].addr,
3884 					    adapter->mac_table[i].queue,
3885 					    IXGBE_RAH_AV);
3886 		else
3887 			hw->mac.ops.clear_rar(hw, i);
3888 
3889 		adapter->mac_table[i].state &= ~(IXGBE_MAC_STATE_MODIFIED);
3890 	}
3891 }
3892 #endif
3893 
3894 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
3895 {
3896 	struct ixgbe_hw *hw = &adapter->hw;
3897 	int i;
3898 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3899 		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_MODIFIED) {
3900 			if (adapter->mac_table[i].state &
3901 			    IXGBE_MAC_STATE_IN_USE)
3902 				hw->mac.ops.set_rar(hw, i,
3903 						adapter->mac_table[i].addr,
3904 						adapter->mac_table[i].queue,
3905 						IXGBE_RAH_AV);
3906 			else
3907 				hw->mac.ops.clear_rar(hw, i);
3908 
3909 			adapter->mac_table[i].state &=
3910 						~(IXGBE_MAC_STATE_MODIFIED);
3911 		}
3912 	}
3913 }
3914 
3915 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
3916 {
3917 	int i;
3918 	struct ixgbe_hw *hw = &adapter->hw;
3919 
3920 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3921 		adapter->mac_table[i].state |= IXGBE_MAC_STATE_MODIFIED;
3922 		adapter->mac_table[i].state &= ~IXGBE_MAC_STATE_IN_USE;
3923 		eth_zero_addr(adapter->mac_table[i].addr);
3924 		adapter->mac_table[i].queue = 0;
3925 	}
3926 	ixgbe_sync_mac_table(adapter);
3927 }
3928 
3929 static int ixgbe_available_rars(struct ixgbe_adapter *adapter)
3930 {
3931 	struct ixgbe_hw *hw = &adapter->hw;
3932 	int i, count = 0;
3933 
3934 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3935 		if (adapter->mac_table[i].state == 0)
3936 			count++;
3937 	}
3938 	return count;
3939 }
3940 
3941 /* this function destroys the first RAR entry */
3942 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter,
3943 					 u8 *addr)
3944 {
3945 	struct ixgbe_hw *hw = &adapter->hw;
3946 
3947 	memcpy(&adapter->mac_table[0].addr, addr, ETH_ALEN);
3948 	adapter->mac_table[0].queue = VMDQ_P(0);
3949 	adapter->mac_table[0].state = (IXGBE_MAC_STATE_DEFAULT |
3950 				       IXGBE_MAC_STATE_IN_USE);
3951 	hw->mac.ops.set_rar(hw, 0, adapter->mac_table[0].addr,
3952 			    adapter->mac_table[0].queue,
3953 			    IXGBE_RAH_AV);
3954 }
3955 
3956 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter, u8 *addr, u16 queue)
3957 {
3958 	struct ixgbe_hw *hw = &adapter->hw;
3959 	int i;
3960 
3961 	if (is_zero_ether_addr(addr))
3962 		return -EINVAL;
3963 
3964 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3965 		if (adapter->mac_table[i].state & IXGBE_MAC_STATE_IN_USE)
3966 			continue;
3967 		adapter->mac_table[i].state |= (IXGBE_MAC_STATE_MODIFIED |
3968 						IXGBE_MAC_STATE_IN_USE);
3969 		ether_addr_copy(adapter->mac_table[i].addr, addr);
3970 		adapter->mac_table[i].queue = queue;
3971 		ixgbe_sync_mac_table(adapter);
3972 		return i;
3973 	}
3974 	return -ENOMEM;
3975 }
3976 
3977 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter, u8 *addr, u16 queue)
3978 {
3979 	/* search table for addr, if found, set to 0 and sync */
3980 	int i;
3981 	struct ixgbe_hw *hw = &adapter->hw;
3982 
3983 	if (is_zero_ether_addr(addr))
3984 		return -EINVAL;
3985 
3986 	for (i = 0; i < hw->mac.num_rar_entries; i++) {
3987 		if (ether_addr_equal(addr, adapter->mac_table[i].addr) &&
3988 		    adapter->mac_table[i].queue == queue) {
3989 			adapter->mac_table[i].state |= IXGBE_MAC_STATE_MODIFIED;
3990 			adapter->mac_table[i].state &= ~IXGBE_MAC_STATE_IN_USE;
3991 			eth_zero_addr(adapter->mac_table[i].addr);
3992 			adapter->mac_table[i].queue = 0;
3993 			ixgbe_sync_mac_table(adapter);
3994 			return 0;
3995 		}
3996 	}
3997 	return -ENOMEM;
3998 }
3999 /**
4000  * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
4001  * @netdev: network interface device structure
4002  *
4003  * Writes unicast address list to the RAR table.
4004  * Returns: -ENOMEM on failure/insufficient address space
4005  *                0 on no addresses written
4006  *                X on writing X addresses to the RAR table
4007  **/
4008 static int ixgbe_write_uc_addr_list(struct net_device *netdev, int vfn)
4009 {
4010 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4011 	int count = 0;
4012 
4013 	/* return ENOMEM indicating insufficient memory for addresses */
4014 	if (netdev_uc_count(netdev) > ixgbe_available_rars(adapter))
4015 		return -ENOMEM;
4016 
4017 	if (!netdev_uc_empty(netdev)) {
4018 		struct netdev_hw_addr *ha;
4019 		netdev_for_each_uc_addr(ha, netdev) {
4020 			ixgbe_del_mac_filter(adapter, ha->addr, vfn);
4021 			ixgbe_add_mac_filter(adapter, ha->addr, vfn);
4022 			count++;
4023 		}
4024 	}
4025 	return count;
4026 }
4027 
4028 /**
4029  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4030  * @netdev: network interface device structure
4031  *
4032  * The set_rx_method entry point is called whenever the unicast/multicast
4033  * address list or the network interface flags are updated.  This routine is
4034  * responsible for configuring the hardware for proper unicast, multicast and
4035  * promiscuous mode.
4036  **/
4037 void ixgbe_set_rx_mode(struct net_device *netdev)
4038 {
4039 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4040 	struct ixgbe_hw *hw = &adapter->hw;
4041 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4042 	u32 vlnctrl;
4043 	int count;
4044 
4045 	/* Check for Promiscuous and All Multicast modes */
4046 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4047 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4048 
4049 	/* set all bits that we expect to always be set */
4050 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4051 	fctrl |= IXGBE_FCTRL_BAM;
4052 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4053 	fctrl |= IXGBE_FCTRL_PMCF;
4054 
4055 	/* clear the bits we are changing the status of */
4056 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4057 	vlnctrl &= ~(IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
4058 	if (netdev->flags & IFF_PROMISC) {
4059 		hw->addr_ctrl.user_set_promisc = true;
4060 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4061 		vmolr |= IXGBE_VMOLR_MPE;
4062 		/* Only disable hardware filter vlans in promiscuous mode
4063 		 * if SR-IOV and VMDQ are disabled - otherwise ensure
4064 		 * that hardware VLAN filters remain enabled.
4065 		 */
4066 		if (adapter->flags & (IXGBE_FLAG_VMDQ_ENABLED |
4067 				      IXGBE_FLAG_SRIOV_ENABLED))
4068 			vlnctrl |= (IXGBE_VLNCTRL_VFE | IXGBE_VLNCTRL_CFIEN);
4069 	} else {
4070 		if (netdev->flags & IFF_ALLMULTI) {
4071 			fctrl |= IXGBE_FCTRL_MPE;
4072 			vmolr |= IXGBE_VMOLR_MPE;
4073 		}
4074 		vlnctrl |= IXGBE_VLNCTRL_VFE;
4075 		hw->addr_ctrl.user_set_promisc = false;
4076 	}
4077 
4078 	/*
4079 	 * Write addresses to available RAR registers, if there is not
4080 	 * sufficient space to store all the addresses then enable
4081 	 * unicast promiscuous mode
4082 	 */
4083 	count = ixgbe_write_uc_addr_list(netdev, VMDQ_P(0));
4084 	if (count < 0) {
4085 		fctrl |= IXGBE_FCTRL_UPE;
4086 		vmolr |= IXGBE_VMOLR_ROPE;
4087 	}
4088 
4089 	/* Write addresses to the MTA, if the attempt fails
4090 	 * then we should just turn on promiscuous mode so
4091 	 * that we can at least receive multicast traffic
4092 	 */
4093 	count = ixgbe_write_mc_addr_list(netdev);
4094 	if (count < 0) {
4095 		fctrl |= IXGBE_FCTRL_MPE;
4096 		vmolr |= IXGBE_VMOLR_MPE;
4097 	} else if (count) {
4098 		vmolr |= IXGBE_VMOLR_ROMPE;
4099 	}
4100 
4101 	if (hw->mac.type != ixgbe_mac_82598EB) {
4102 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4103 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4104 			   IXGBE_VMOLR_ROPE);
4105 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4106 	}
4107 
4108 	/* This is useful for sniffing bad packets. */
4109 	if (adapter->netdev->features & NETIF_F_RXALL) {
4110 		/* UPE and MPE will be handled by normal PROMISC logic
4111 		 * in e1000e_set_rx_mode */
4112 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4113 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4114 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4115 
4116 		fctrl &= ~(IXGBE_FCTRL_DPF);
4117 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
4118 	}
4119 
4120 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4121 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4122 
4123 	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
4124 		ixgbe_vlan_strip_enable(adapter);
4125 	else
4126 		ixgbe_vlan_strip_disable(adapter);
4127 }
4128 
4129 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4130 {
4131 	int q_idx;
4132 
4133 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4134 		ixgbe_qv_init_lock(adapter->q_vector[q_idx]);
4135 		napi_enable(&adapter->q_vector[q_idx]->napi);
4136 	}
4137 }
4138 
4139 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
4140 {
4141 	int q_idx;
4142 
4143 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++) {
4144 		napi_disable(&adapter->q_vector[q_idx]->napi);
4145 		while (!ixgbe_qv_disable(adapter->q_vector[q_idx])) {
4146 			pr_info("QV %d locked\n", q_idx);
4147 			usleep_range(1000, 20000);
4148 		}
4149 	}
4150 }
4151 
4152 #ifdef CONFIG_IXGBE_DCB
4153 /**
4154  * ixgbe_configure_dcb - Configure DCB hardware
4155  * @adapter: ixgbe adapter struct
4156  *
4157  * This is called by the driver on open to configure the DCB hardware.
4158  * This is also called by the gennetlink interface when reconfiguring
4159  * the DCB state.
4160  */
4161 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
4162 {
4163 	struct ixgbe_hw *hw = &adapter->hw;
4164 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4165 
4166 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
4167 		if (hw->mac.type == ixgbe_mac_82598EB)
4168 			netif_set_gso_max_size(adapter->netdev, 65536);
4169 		return;
4170 	}
4171 
4172 	if (hw->mac.type == ixgbe_mac_82598EB)
4173 		netif_set_gso_max_size(adapter->netdev, 32768);
4174 
4175 #ifdef IXGBE_FCOE
4176 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
4177 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
4178 #endif
4179 
4180 	/* reconfigure the hardware */
4181 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
4182 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4183 						DCB_TX_CONFIG);
4184 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
4185 						DCB_RX_CONFIG);
4186 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
4187 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
4188 		ixgbe_dcb_hw_ets(&adapter->hw,
4189 				 adapter->ixgbe_ieee_ets,
4190 				 max_frame);
4191 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
4192 					adapter->ixgbe_ieee_pfc->pfc_en,
4193 					adapter->ixgbe_ieee_ets->prio_tc);
4194 	}
4195 
4196 	/* Enable RSS Hash per TC */
4197 	if (hw->mac.type != ixgbe_mac_82598EB) {
4198 		u32 msb = 0;
4199 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
4200 
4201 		while (rss_i) {
4202 			msb++;
4203 			rss_i >>= 1;
4204 		}
4205 
4206 		/* write msb to all 8 TCs in one write */
4207 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
4208 	}
4209 }
4210 #endif
4211 
4212 /* Additional bittime to account for IXGBE framing */
4213 #define IXGBE_ETH_FRAMING 20
4214 
4215 /**
4216  * ixgbe_hpbthresh - calculate high water mark for flow control
4217  *
4218  * @adapter: board private structure to calculate for
4219  * @pb: packet buffer to calculate
4220  */
4221 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
4222 {
4223 	struct ixgbe_hw *hw = &adapter->hw;
4224 	struct net_device *dev = adapter->netdev;
4225 	int link, tc, kb, marker;
4226 	u32 dv_id, rx_pba;
4227 
4228 	/* Calculate max LAN frame size */
4229 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
4230 
4231 #ifdef IXGBE_FCOE
4232 	/* FCoE traffic class uses FCOE jumbo frames */
4233 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4234 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4235 	    (pb == ixgbe_fcoe_get_tc(adapter)))
4236 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4237 #endif
4238 
4239 	/* Calculate delay value for device */
4240 	switch (hw->mac.type) {
4241 	case ixgbe_mac_X540:
4242 	case ixgbe_mac_X550:
4243 	case ixgbe_mac_X550EM_x:
4244 		dv_id = IXGBE_DV_X540(link, tc);
4245 		break;
4246 	default:
4247 		dv_id = IXGBE_DV(link, tc);
4248 		break;
4249 	}
4250 
4251 	/* Loopback switch introduces additional latency */
4252 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4253 		dv_id += IXGBE_B2BT(tc);
4254 
4255 	/* Delay value is calculated in bit times convert to KB */
4256 	kb = IXGBE_BT2KB(dv_id);
4257 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
4258 
4259 	marker = rx_pba - kb;
4260 
4261 	/* It is possible that the packet buffer is not large enough
4262 	 * to provide required headroom. In this case throw an error
4263 	 * to user and a do the best we can.
4264 	 */
4265 	if (marker < 0) {
4266 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
4267 			    "headroom to support flow control."
4268 			    "Decrease MTU or number of traffic classes\n", pb);
4269 		marker = tc + 1;
4270 	}
4271 
4272 	return marker;
4273 }
4274 
4275 /**
4276  * ixgbe_lpbthresh - calculate low water mark for for flow control
4277  *
4278  * @adapter: board private structure to calculate for
4279  * @pb: packet buffer to calculate
4280  */
4281 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
4282 {
4283 	struct ixgbe_hw *hw = &adapter->hw;
4284 	struct net_device *dev = adapter->netdev;
4285 	int tc;
4286 	u32 dv_id;
4287 
4288 	/* Calculate max LAN frame size */
4289 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
4290 
4291 #ifdef IXGBE_FCOE
4292 	/* FCoE traffic class uses FCOE jumbo frames */
4293 	if ((dev->features & NETIF_F_FCOE_MTU) &&
4294 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
4295 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
4296 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4297 #endif
4298 
4299 	/* Calculate delay value for device */
4300 	switch (hw->mac.type) {
4301 	case ixgbe_mac_X540:
4302 	case ixgbe_mac_X550:
4303 	case ixgbe_mac_X550EM_x:
4304 		dv_id = IXGBE_LOW_DV_X540(tc);
4305 		break;
4306 	default:
4307 		dv_id = IXGBE_LOW_DV(tc);
4308 		break;
4309 	}
4310 
4311 	/* Delay value is calculated in bit times convert to KB */
4312 	return IXGBE_BT2KB(dv_id);
4313 }
4314 
4315 /*
4316  * ixgbe_pbthresh_setup - calculate and setup high low water marks
4317  */
4318 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
4319 {
4320 	struct ixgbe_hw *hw = &adapter->hw;
4321 	int num_tc = netdev_get_num_tc(adapter->netdev);
4322 	int i;
4323 
4324 	if (!num_tc)
4325 		num_tc = 1;
4326 
4327 	for (i = 0; i < num_tc; i++) {
4328 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
4329 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
4330 
4331 		/* Low water marks must not be larger than high water marks */
4332 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
4333 			hw->fc.low_water[i] = 0;
4334 	}
4335 
4336 	for (; i < MAX_TRAFFIC_CLASS; i++)
4337 		hw->fc.high_water[i] = 0;
4338 }
4339 
4340 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
4341 {
4342 	struct ixgbe_hw *hw = &adapter->hw;
4343 	int hdrm;
4344 	u8 tc = netdev_get_num_tc(adapter->netdev);
4345 
4346 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
4347 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
4348 		hdrm = 32 << adapter->fdir_pballoc;
4349 	else
4350 		hdrm = 0;
4351 
4352 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
4353 	ixgbe_pbthresh_setup(adapter);
4354 }
4355 
4356 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
4357 {
4358 	struct ixgbe_hw *hw = &adapter->hw;
4359 	struct hlist_node *node2;
4360 	struct ixgbe_fdir_filter *filter;
4361 
4362 	spin_lock(&adapter->fdir_perfect_lock);
4363 
4364 	if (!hlist_empty(&adapter->fdir_filter_list))
4365 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
4366 
4367 	hlist_for_each_entry_safe(filter, node2,
4368 				  &adapter->fdir_filter_list, fdir_node) {
4369 		ixgbe_fdir_write_perfect_filter_82599(hw,
4370 				&filter->filter,
4371 				filter->sw_idx,
4372 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
4373 				IXGBE_FDIR_DROP_QUEUE :
4374 				adapter->rx_ring[filter->action]->reg_idx);
4375 	}
4376 
4377 	spin_unlock(&adapter->fdir_perfect_lock);
4378 }
4379 
4380 static void ixgbe_macvlan_set_rx_mode(struct net_device *dev, unsigned int pool,
4381 				      struct ixgbe_adapter *adapter)
4382 {
4383 	struct ixgbe_hw *hw = &adapter->hw;
4384 	u32 vmolr;
4385 
4386 	/* No unicast promiscuous support for VMDQ devices. */
4387 	vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
4388 	vmolr |= (IXGBE_VMOLR_ROMPE | IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE);
4389 
4390 	/* clear the affected bit */
4391 	vmolr &= ~IXGBE_VMOLR_MPE;
4392 
4393 	if (dev->flags & IFF_ALLMULTI) {
4394 		vmolr |= IXGBE_VMOLR_MPE;
4395 	} else {
4396 		vmolr |= IXGBE_VMOLR_ROMPE;
4397 		hw->mac.ops.update_mc_addr_list(hw, dev);
4398 	}
4399 	ixgbe_write_uc_addr_list(adapter->netdev, pool);
4400 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
4401 }
4402 
4403 static void ixgbe_fwd_psrtype(struct ixgbe_fwd_adapter *vadapter)
4404 {
4405 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4406 	int rss_i = adapter->num_rx_queues_per_pool;
4407 	struct ixgbe_hw *hw = &adapter->hw;
4408 	u16 pool = vadapter->pool;
4409 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4410 		      IXGBE_PSRTYPE_UDPHDR |
4411 		      IXGBE_PSRTYPE_IPV4HDR |
4412 		      IXGBE_PSRTYPE_L2HDR |
4413 		      IXGBE_PSRTYPE_IPV6HDR;
4414 
4415 	if (hw->mac.type == ixgbe_mac_82598EB)
4416 		return;
4417 
4418 	if (rss_i > 3)
4419 		psrtype |= 2 << 29;
4420 	else if (rss_i > 1)
4421 		psrtype |= 1 << 29;
4422 
4423 	IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4424 }
4425 
4426 /**
4427  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4428  * @rx_ring: ring to free buffers from
4429  **/
4430 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
4431 {
4432 	struct device *dev = rx_ring->dev;
4433 	unsigned long size;
4434 	u16 i;
4435 
4436 	/* ring already cleared, nothing to do */
4437 	if (!rx_ring->rx_buffer_info)
4438 		return;
4439 
4440 	/* Free all the Rx ring sk_buffs */
4441 	for (i = 0; i < rx_ring->count; i++) {
4442 		struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
4443 
4444 		if (rx_buffer->skb) {
4445 			struct sk_buff *skb = rx_buffer->skb;
4446 			if (IXGBE_CB(skb)->page_released)
4447 				dma_unmap_page(dev,
4448 					       IXGBE_CB(skb)->dma,
4449 					       ixgbe_rx_bufsz(rx_ring),
4450 					       DMA_FROM_DEVICE);
4451 			dev_kfree_skb(skb);
4452 			rx_buffer->skb = NULL;
4453 		}
4454 
4455 		if (!rx_buffer->page)
4456 			continue;
4457 
4458 		dma_unmap_page(dev, rx_buffer->dma,
4459 			       ixgbe_rx_pg_size(rx_ring), DMA_FROM_DEVICE);
4460 		__free_pages(rx_buffer->page, ixgbe_rx_pg_order(rx_ring));
4461 
4462 		rx_buffer->page = NULL;
4463 	}
4464 
4465 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
4466 	memset(rx_ring->rx_buffer_info, 0, size);
4467 
4468 	/* Zero out the descriptor ring */
4469 	memset(rx_ring->desc, 0, rx_ring->size);
4470 
4471 	rx_ring->next_to_alloc = 0;
4472 	rx_ring->next_to_clean = 0;
4473 	rx_ring->next_to_use = 0;
4474 }
4475 
4476 static void ixgbe_disable_fwd_ring(struct ixgbe_fwd_adapter *vadapter,
4477 				   struct ixgbe_ring *rx_ring)
4478 {
4479 	struct ixgbe_adapter *adapter = vadapter->real_adapter;
4480 	int index = rx_ring->queue_index + vadapter->rx_base_queue;
4481 
4482 	/* shutdown specific queue receive and wait for dma to settle */
4483 	ixgbe_disable_rx_queue(adapter, rx_ring);
4484 	usleep_range(10000, 20000);
4485 	ixgbe_irq_disable_queues(adapter, ((u64)1 << index));
4486 	ixgbe_clean_rx_ring(rx_ring);
4487 	rx_ring->l2_accel_priv = NULL;
4488 }
4489 
4490 static int ixgbe_fwd_ring_down(struct net_device *vdev,
4491 			       struct ixgbe_fwd_adapter *accel)
4492 {
4493 	struct ixgbe_adapter *adapter = accel->real_adapter;
4494 	unsigned int rxbase = accel->rx_base_queue;
4495 	unsigned int txbase = accel->tx_base_queue;
4496 	int i;
4497 
4498 	netif_tx_stop_all_queues(vdev);
4499 
4500 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4501 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4502 		adapter->rx_ring[rxbase + i]->netdev = adapter->netdev;
4503 	}
4504 
4505 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4506 		adapter->tx_ring[txbase + i]->l2_accel_priv = NULL;
4507 		adapter->tx_ring[txbase + i]->netdev = adapter->netdev;
4508 	}
4509 
4510 
4511 	return 0;
4512 }
4513 
4514 static int ixgbe_fwd_ring_up(struct net_device *vdev,
4515 			     struct ixgbe_fwd_adapter *accel)
4516 {
4517 	struct ixgbe_adapter *adapter = accel->real_adapter;
4518 	unsigned int rxbase, txbase, queues;
4519 	int i, baseq, err = 0;
4520 
4521 	if (!test_bit(accel->pool, &adapter->fwd_bitmask))
4522 		return 0;
4523 
4524 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
4525 	netdev_dbg(vdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
4526 		   accel->pool, adapter->num_rx_pools,
4527 		   baseq, baseq + adapter->num_rx_queues_per_pool,
4528 		   adapter->fwd_bitmask);
4529 
4530 	accel->netdev = vdev;
4531 	accel->rx_base_queue = rxbase = baseq;
4532 	accel->tx_base_queue = txbase = baseq;
4533 
4534 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
4535 		ixgbe_disable_fwd_ring(accel, adapter->rx_ring[rxbase + i]);
4536 
4537 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4538 		adapter->rx_ring[rxbase + i]->netdev = vdev;
4539 		adapter->rx_ring[rxbase + i]->l2_accel_priv = accel;
4540 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[rxbase + i]);
4541 	}
4542 
4543 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
4544 		adapter->tx_ring[txbase + i]->netdev = vdev;
4545 		adapter->tx_ring[txbase + i]->l2_accel_priv = accel;
4546 	}
4547 
4548 	queues = min_t(unsigned int,
4549 		       adapter->num_rx_queues_per_pool, vdev->num_tx_queues);
4550 	err = netif_set_real_num_tx_queues(vdev, queues);
4551 	if (err)
4552 		goto fwd_queue_err;
4553 
4554 	err = netif_set_real_num_rx_queues(vdev, queues);
4555 	if (err)
4556 		goto fwd_queue_err;
4557 
4558 	if (is_valid_ether_addr(vdev->dev_addr))
4559 		ixgbe_add_mac_filter(adapter, vdev->dev_addr, accel->pool);
4560 
4561 	ixgbe_fwd_psrtype(accel);
4562 	ixgbe_macvlan_set_rx_mode(vdev, accel->pool, adapter);
4563 	return err;
4564 fwd_queue_err:
4565 	ixgbe_fwd_ring_down(vdev, accel);
4566 	return err;
4567 }
4568 
4569 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
4570 {
4571 	struct net_device *upper;
4572 	struct list_head *iter;
4573 	int err;
4574 
4575 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
4576 		if (netif_is_macvlan(upper)) {
4577 			struct macvlan_dev *dfwd = netdev_priv(upper);
4578 			struct ixgbe_fwd_adapter *vadapter = dfwd->fwd_priv;
4579 
4580 			if (dfwd->fwd_priv) {
4581 				err = ixgbe_fwd_ring_up(upper, vadapter);
4582 				if (err)
4583 					continue;
4584 			}
4585 		}
4586 	}
4587 }
4588 
4589 static void ixgbe_configure(struct ixgbe_adapter *adapter)
4590 {
4591 	struct ixgbe_hw *hw = &adapter->hw;
4592 
4593 	ixgbe_configure_pb(adapter);
4594 #ifdef CONFIG_IXGBE_DCB
4595 	ixgbe_configure_dcb(adapter);
4596 #endif
4597 	/*
4598 	 * We must restore virtualization before VLANs or else
4599 	 * the VLVF registers will not be populated
4600 	 */
4601 	ixgbe_configure_virtualization(adapter);
4602 
4603 	ixgbe_set_rx_mode(adapter->netdev);
4604 	ixgbe_restore_vlan(adapter);
4605 
4606 	switch (hw->mac.type) {
4607 	case ixgbe_mac_82599EB:
4608 	case ixgbe_mac_X540:
4609 		hw->mac.ops.disable_rx_buff(hw);
4610 		break;
4611 	default:
4612 		break;
4613 	}
4614 
4615 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
4616 		ixgbe_init_fdir_signature_82599(&adapter->hw,
4617 						adapter->fdir_pballoc);
4618 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
4619 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
4620 					      adapter->fdir_pballoc);
4621 		ixgbe_fdir_filter_restore(adapter);
4622 	}
4623 
4624 	switch (hw->mac.type) {
4625 	case ixgbe_mac_82599EB:
4626 	case ixgbe_mac_X540:
4627 		hw->mac.ops.enable_rx_buff(hw);
4628 		break;
4629 	default:
4630 		break;
4631 	}
4632 
4633 #ifdef IXGBE_FCOE
4634 	/* configure FCoE L2 filters, redirection table, and Rx control */
4635 	ixgbe_configure_fcoe(adapter);
4636 
4637 #endif /* IXGBE_FCOE */
4638 	ixgbe_configure_tx(adapter);
4639 	ixgbe_configure_rx(adapter);
4640 	ixgbe_configure_dfwd(adapter);
4641 }
4642 
4643 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
4644 {
4645 	switch (hw->phy.type) {
4646 	case ixgbe_phy_sfp_avago:
4647 	case ixgbe_phy_sfp_ftl:
4648 	case ixgbe_phy_sfp_intel:
4649 	case ixgbe_phy_sfp_unknown:
4650 	case ixgbe_phy_sfp_passive_tyco:
4651 	case ixgbe_phy_sfp_passive_unknown:
4652 	case ixgbe_phy_sfp_active_unknown:
4653 	case ixgbe_phy_sfp_ftl_active:
4654 	case ixgbe_phy_qsfp_passive_unknown:
4655 	case ixgbe_phy_qsfp_active_unknown:
4656 	case ixgbe_phy_qsfp_intel:
4657 	case ixgbe_phy_qsfp_unknown:
4658 	/* ixgbe_phy_none is set when no SFP module is present */
4659 	case ixgbe_phy_none:
4660 		return true;
4661 	case ixgbe_phy_nl:
4662 		if (hw->mac.type == ixgbe_mac_82598EB)
4663 			return true;
4664 	default:
4665 		return false;
4666 	}
4667 }
4668 
4669 /**
4670  * ixgbe_sfp_link_config - set up SFP+ link
4671  * @adapter: pointer to private adapter struct
4672  **/
4673 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
4674 {
4675 	/*
4676 	 * We are assuming the worst case scenario here, and that
4677 	 * is that an SFP was inserted/removed after the reset
4678 	 * but before SFP detection was enabled.  As such the best
4679 	 * solution is to just start searching as soon as we start
4680 	 */
4681 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
4682 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
4683 
4684 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
4685 }
4686 
4687 /**
4688  * ixgbe_non_sfp_link_config - set up non-SFP+ link
4689  * @hw: pointer to private hardware struct
4690  *
4691  * Returns 0 on success, negative on failure
4692  **/
4693 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
4694 {
4695 	u32 speed;
4696 	bool autoneg, link_up = false;
4697 	u32 ret = IXGBE_ERR_LINK_SETUP;
4698 
4699 	if (hw->mac.ops.check_link)
4700 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
4701 
4702 	if (ret)
4703 		return ret;
4704 
4705 	speed = hw->phy.autoneg_advertised;
4706 	if ((!speed) && (hw->mac.ops.get_link_capabilities))
4707 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
4708 							&autoneg);
4709 	if (ret)
4710 		return ret;
4711 
4712 	if (hw->mac.ops.setup_link)
4713 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
4714 
4715 	return ret;
4716 }
4717 
4718 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
4719 {
4720 	struct ixgbe_hw *hw = &adapter->hw;
4721 	u32 gpie = 0;
4722 
4723 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
4724 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
4725 		       IXGBE_GPIE_OCD;
4726 		gpie |= IXGBE_GPIE_EIAME;
4727 		/*
4728 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
4729 		 * this saves a register write for every interrupt
4730 		 */
4731 		switch (hw->mac.type) {
4732 		case ixgbe_mac_82598EB:
4733 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4734 			break;
4735 		case ixgbe_mac_82599EB:
4736 		case ixgbe_mac_X540:
4737 		case ixgbe_mac_X550:
4738 		case ixgbe_mac_X550EM_x:
4739 		default:
4740 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
4741 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
4742 			break;
4743 		}
4744 	} else {
4745 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
4746 		 * specifically only auto mask tx and rx interrupts */
4747 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4748 	}
4749 
4750 	/* XXX: to interrupt immediately for EICS writes, enable this */
4751 	/* gpie |= IXGBE_GPIE_EIMEN; */
4752 
4753 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
4754 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
4755 
4756 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4757 		case IXGBE_82599_VMDQ_8Q_MASK:
4758 			gpie |= IXGBE_GPIE_VTMODE_16;
4759 			break;
4760 		case IXGBE_82599_VMDQ_4Q_MASK:
4761 			gpie |= IXGBE_GPIE_VTMODE_32;
4762 			break;
4763 		default:
4764 			gpie |= IXGBE_GPIE_VTMODE_64;
4765 			break;
4766 		}
4767 	}
4768 
4769 	/* Enable Thermal over heat sensor interrupt */
4770 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
4771 		switch (adapter->hw.mac.type) {
4772 		case ixgbe_mac_82599EB:
4773 			gpie |= IXGBE_SDP0_GPIEN;
4774 			break;
4775 		case ixgbe_mac_X540:
4776 			gpie |= IXGBE_EIMS_TS;
4777 			break;
4778 		default:
4779 			break;
4780 		}
4781 	}
4782 
4783 	/* Enable fan failure interrupt */
4784 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
4785 		gpie |= IXGBE_SDP1_GPIEN;
4786 
4787 	if (hw->mac.type == ixgbe_mac_82599EB) {
4788 		gpie |= IXGBE_SDP1_GPIEN;
4789 		gpie |= IXGBE_SDP2_GPIEN;
4790 	}
4791 
4792 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
4793 }
4794 
4795 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
4796 {
4797 	struct ixgbe_hw *hw = &adapter->hw;
4798 	int err;
4799 	u32 ctrl_ext;
4800 
4801 	ixgbe_get_hw_control(adapter);
4802 	ixgbe_setup_gpie(adapter);
4803 
4804 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
4805 		ixgbe_configure_msix(adapter);
4806 	else
4807 		ixgbe_configure_msi_and_legacy(adapter);
4808 
4809 	/* enable the optics for 82599 SFP+ fiber */
4810 	if (hw->mac.ops.enable_tx_laser)
4811 		hw->mac.ops.enable_tx_laser(hw);
4812 
4813 	smp_mb__before_atomic();
4814 	clear_bit(__IXGBE_DOWN, &adapter->state);
4815 	ixgbe_napi_enable_all(adapter);
4816 
4817 	if (ixgbe_is_sfp(hw)) {
4818 		ixgbe_sfp_link_config(adapter);
4819 	} else {
4820 		err = ixgbe_non_sfp_link_config(hw);
4821 		if (err)
4822 			e_err(probe, "link_config FAILED %d\n", err);
4823 	}
4824 
4825 	/* clear any pending interrupts, may auto mask */
4826 	IXGBE_READ_REG(hw, IXGBE_EICR);
4827 	ixgbe_irq_enable(adapter, true, true);
4828 
4829 	/*
4830 	 * If this adapter has a fan, check to see if we had a failure
4831 	 * before we enabled the interrupt.
4832 	 */
4833 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
4834 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
4835 		if (esdp & IXGBE_ESDP_SDP1)
4836 			e_crit(drv, "Fan has stopped, replace the adapter\n");
4837 	}
4838 
4839 	/* bring the link up in the watchdog, this could race with our first
4840 	 * link up interrupt but shouldn't be a problem */
4841 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4842 	adapter->link_check_timeout = jiffies;
4843 	mod_timer(&adapter->service_timer, jiffies);
4844 
4845 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
4846 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
4847 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
4848 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
4849 }
4850 
4851 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
4852 {
4853 	WARN_ON(in_interrupt());
4854 	/* put off any impending NetWatchDogTimeout */
4855 	adapter->netdev->trans_start = jiffies;
4856 
4857 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
4858 		usleep_range(1000, 2000);
4859 	ixgbe_down(adapter);
4860 	/*
4861 	 * If SR-IOV enabled then wait a bit before bringing the adapter
4862 	 * back up to give the VFs time to respond to the reset.  The
4863 	 * two second wait is based upon the watchdog timer cycle in
4864 	 * the VF driver.
4865 	 */
4866 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4867 		msleep(2000);
4868 	ixgbe_up(adapter);
4869 	clear_bit(__IXGBE_RESETTING, &adapter->state);
4870 }
4871 
4872 void ixgbe_up(struct ixgbe_adapter *adapter)
4873 {
4874 	/* hardware has been reset, we need to reload some things */
4875 	ixgbe_configure(adapter);
4876 
4877 	ixgbe_up_complete(adapter);
4878 }
4879 
4880 void ixgbe_reset(struct ixgbe_adapter *adapter)
4881 {
4882 	struct ixgbe_hw *hw = &adapter->hw;
4883 	struct net_device *netdev = adapter->netdev;
4884 	int err;
4885 	u8 old_addr[ETH_ALEN];
4886 
4887 	if (ixgbe_removed(hw->hw_addr))
4888 		return;
4889 	/* lock SFP init bit to prevent race conditions with the watchdog */
4890 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
4891 		usleep_range(1000, 2000);
4892 
4893 	/* clear all SFP and link config related flags while holding SFP_INIT */
4894 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
4895 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
4896 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4897 
4898 	err = hw->mac.ops.init_hw(hw);
4899 	switch (err) {
4900 	case 0:
4901 	case IXGBE_ERR_SFP_NOT_PRESENT:
4902 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
4903 		break;
4904 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
4905 		e_dev_err("master disable timed out\n");
4906 		break;
4907 	case IXGBE_ERR_EEPROM_VERSION:
4908 		/* We are running on a pre-production device, log a warning */
4909 		e_dev_warn("This device is a pre-production adapter/LOM. "
4910 			   "Please be aware there may be issues associated with "
4911 			   "your hardware.  If you are experiencing problems "
4912 			   "please contact your Intel or hardware "
4913 			   "representative who provided you with this "
4914 			   "hardware.\n");
4915 		break;
4916 	default:
4917 		e_dev_err("Hardware Error: %d\n", err);
4918 	}
4919 
4920 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
4921 	/* do not flush user set addresses */
4922 	memcpy(old_addr, &adapter->mac_table[0].addr, netdev->addr_len);
4923 	ixgbe_flush_sw_mac_table(adapter);
4924 	ixgbe_mac_set_default_filter(adapter, old_addr);
4925 
4926 	/* update SAN MAC vmdq pool selection */
4927 	if (hw->mac.san_mac_rar_index)
4928 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
4929 
4930 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
4931 		ixgbe_ptp_reset(adapter);
4932 }
4933 
4934 /**
4935  * ixgbe_clean_tx_ring - Free Tx Buffers
4936  * @tx_ring: ring to be cleaned
4937  **/
4938 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
4939 {
4940 	struct ixgbe_tx_buffer *tx_buffer_info;
4941 	unsigned long size;
4942 	u16 i;
4943 
4944 	/* ring already cleared, nothing to do */
4945 	if (!tx_ring->tx_buffer_info)
4946 		return;
4947 
4948 	/* Free all the Tx ring sk_buffs */
4949 	for (i = 0; i < tx_ring->count; i++) {
4950 		tx_buffer_info = &tx_ring->tx_buffer_info[i];
4951 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer_info);
4952 	}
4953 
4954 	netdev_tx_reset_queue(txring_txq(tx_ring));
4955 
4956 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
4957 	memset(tx_ring->tx_buffer_info, 0, size);
4958 
4959 	/* Zero out the descriptor ring */
4960 	memset(tx_ring->desc, 0, tx_ring->size);
4961 
4962 	tx_ring->next_to_use = 0;
4963 	tx_ring->next_to_clean = 0;
4964 }
4965 
4966 /**
4967  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
4968  * @adapter: board private structure
4969  **/
4970 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
4971 {
4972 	int i;
4973 
4974 	for (i = 0; i < adapter->num_rx_queues; i++)
4975 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
4976 }
4977 
4978 /**
4979  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
4980  * @adapter: board private structure
4981  **/
4982 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
4983 {
4984 	int i;
4985 
4986 	for (i = 0; i < adapter->num_tx_queues; i++)
4987 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
4988 }
4989 
4990 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
4991 {
4992 	struct hlist_node *node2;
4993 	struct ixgbe_fdir_filter *filter;
4994 
4995 	spin_lock(&adapter->fdir_perfect_lock);
4996 
4997 	hlist_for_each_entry_safe(filter, node2,
4998 				  &adapter->fdir_filter_list, fdir_node) {
4999 		hlist_del(&filter->fdir_node);
5000 		kfree(filter);
5001 	}
5002 	adapter->fdir_filter_count = 0;
5003 
5004 	spin_unlock(&adapter->fdir_perfect_lock);
5005 }
5006 
5007 void ixgbe_down(struct ixgbe_adapter *adapter)
5008 {
5009 	struct net_device *netdev = adapter->netdev;
5010 	struct ixgbe_hw *hw = &adapter->hw;
5011 	struct net_device *upper;
5012 	struct list_head *iter;
5013 	int i;
5014 
5015 	/* signal that we are down to the interrupt handler */
5016 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
5017 		return; /* do nothing if already down */
5018 
5019 	/* disable receives */
5020 	hw->mac.ops.disable_rx(hw);
5021 
5022 	/* disable all enabled rx queues */
5023 	for (i = 0; i < adapter->num_rx_queues; i++)
5024 		/* this call also flushes the previous write */
5025 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
5026 
5027 	usleep_range(10000, 20000);
5028 
5029 	netif_tx_stop_all_queues(netdev);
5030 
5031 	/* call carrier off first to avoid false dev_watchdog timeouts */
5032 	netif_carrier_off(netdev);
5033 	netif_tx_disable(netdev);
5034 
5035 	/* disable any upper devices */
5036 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
5037 		if (netif_is_macvlan(upper)) {
5038 			struct macvlan_dev *vlan = netdev_priv(upper);
5039 
5040 			if (vlan->fwd_priv) {
5041 				netif_tx_stop_all_queues(upper);
5042 				netif_carrier_off(upper);
5043 				netif_tx_disable(upper);
5044 			}
5045 		}
5046 	}
5047 
5048 	ixgbe_irq_disable(adapter);
5049 
5050 	ixgbe_napi_disable_all(adapter);
5051 
5052 	adapter->flags2 &= ~(IXGBE_FLAG2_FDIR_REQUIRES_REINIT |
5053 			     IXGBE_FLAG2_RESET_REQUESTED);
5054 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5055 
5056 	del_timer_sync(&adapter->service_timer);
5057 
5058 	if (adapter->num_vfs) {
5059 		/* Clear EITR Select mapping */
5060 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
5061 
5062 		/* Mark all the VFs as inactive */
5063 		for (i = 0 ; i < adapter->num_vfs; i++)
5064 			adapter->vfinfo[i].clear_to_send = false;
5065 
5066 		/* ping all the active vfs to let them know we are going down */
5067 		ixgbe_ping_all_vfs(adapter);
5068 
5069 		/* Disable all VFTE/VFRE TX/RX */
5070 		ixgbe_disable_tx_rx(adapter);
5071 	}
5072 
5073 	/* disable transmits in the hardware now that interrupts are off */
5074 	for (i = 0; i < adapter->num_tx_queues; i++) {
5075 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
5076 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5077 	}
5078 
5079 	/* Disable the Tx DMA engine on 82599 and later MAC */
5080 	switch (hw->mac.type) {
5081 	case ixgbe_mac_82599EB:
5082 	case ixgbe_mac_X540:
5083 	case ixgbe_mac_X550:
5084 	case ixgbe_mac_X550EM_x:
5085 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5086 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5087 				 ~IXGBE_DMATXCTL_TE));
5088 		break;
5089 	default:
5090 		break;
5091 	}
5092 
5093 	if (!pci_channel_offline(adapter->pdev))
5094 		ixgbe_reset(adapter);
5095 
5096 	/* power down the optics for 82599 SFP+ fiber */
5097 	if (hw->mac.ops.disable_tx_laser)
5098 		hw->mac.ops.disable_tx_laser(hw);
5099 
5100 	ixgbe_clean_all_tx_rings(adapter);
5101 	ixgbe_clean_all_rx_rings(adapter);
5102 
5103 #ifdef CONFIG_IXGBE_DCA
5104 	/* since we reset the hardware DCA settings were cleared */
5105 	ixgbe_setup_dca(adapter);
5106 #endif
5107 }
5108 
5109 /**
5110  * ixgbe_tx_timeout - Respond to a Tx Hang
5111  * @netdev: network interface device structure
5112  **/
5113 static void ixgbe_tx_timeout(struct net_device *netdev)
5114 {
5115 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5116 
5117 	/* Do the reset outside of interrupt context */
5118 	ixgbe_tx_timeout_reset(adapter);
5119 }
5120 
5121 /**
5122  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
5123  * @adapter: board private structure to initialize
5124  *
5125  * ixgbe_sw_init initializes the Adapter private data structure.
5126  * Fields are initialized based on PCI device information and
5127  * OS network device settings (MTU size).
5128  **/
5129 static int ixgbe_sw_init(struct ixgbe_adapter *adapter)
5130 {
5131 	struct ixgbe_hw *hw = &adapter->hw;
5132 	struct pci_dev *pdev = adapter->pdev;
5133 	unsigned int rss, fdir;
5134 	u32 fwsm;
5135 #ifdef CONFIG_IXGBE_DCB
5136 	int j;
5137 	struct tc_configuration *tc;
5138 #endif
5139 
5140 	/* PCI config space info */
5141 
5142 	hw->vendor_id = pdev->vendor;
5143 	hw->device_id = pdev->device;
5144 	hw->revision_id = pdev->revision;
5145 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
5146 	hw->subsystem_device_id = pdev->subsystem_device;
5147 
5148 	/* Set common capability flags and settings */
5149 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
5150 	adapter->ring_feature[RING_F_RSS].limit = rss;
5151 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
5152 	adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
5153 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
5154 	adapter->atr_sample_rate = 20;
5155 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
5156 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
5157 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
5158 #ifdef CONFIG_IXGBE_DCA
5159 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
5160 #endif
5161 #ifdef IXGBE_FCOE
5162 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
5163 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5164 #ifdef CONFIG_IXGBE_DCB
5165 	/* Default traffic class to use for FCoE */
5166 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
5167 #endif /* CONFIG_IXGBE_DCB */
5168 #endif /* IXGBE_FCOE */
5169 
5170 	adapter->mac_table = kzalloc(sizeof(struct ixgbe_mac_addr) *
5171 				     hw->mac.num_rar_entries,
5172 				     GFP_ATOMIC);
5173 
5174 	/* Set MAC specific capability flags and exceptions */
5175 	switch (hw->mac.type) {
5176 	case ixgbe_mac_82598EB:
5177 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
5178 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
5179 
5180 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
5181 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
5182 
5183 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
5184 		adapter->ring_feature[RING_F_FDIR].limit = 0;
5185 		adapter->atr_sample_rate = 0;
5186 		adapter->fdir_pballoc = 0;
5187 #ifdef IXGBE_FCOE
5188 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
5189 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
5190 #ifdef CONFIG_IXGBE_DCB
5191 		adapter->fcoe.up = 0;
5192 #endif /* IXGBE_DCB */
5193 #endif /* IXGBE_FCOE */
5194 		break;
5195 	case ixgbe_mac_82599EB:
5196 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
5197 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5198 		break;
5199 	case ixgbe_mac_X540:
5200 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM);
5201 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
5202 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
5203 		break;
5204 	case ixgbe_mac_X550EM_x:
5205 	case ixgbe_mac_X550:
5206 #ifdef CONFIG_IXGBE_DCA
5207 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
5208 #endif
5209 		break;
5210 	default:
5211 		break;
5212 	}
5213 
5214 #ifdef IXGBE_FCOE
5215 	/* FCoE support exists, always init the FCoE lock */
5216 	spin_lock_init(&adapter->fcoe.lock);
5217 
5218 #endif
5219 	/* n-tuple support exists, always init our spinlock */
5220 	spin_lock_init(&adapter->fdir_perfect_lock);
5221 
5222 #ifdef CONFIG_IXGBE_DCB
5223 	switch (hw->mac.type) {
5224 	case ixgbe_mac_X540:
5225 	case ixgbe_mac_X550:
5226 	case ixgbe_mac_X550EM_x:
5227 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
5228 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
5229 		break;
5230 	default:
5231 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
5232 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
5233 		break;
5234 	}
5235 
5236 	/* Configure DCB traffic classes */
5237 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
5238 		tc = &adapter->dcb_cfg.tc_config[j];
5239 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
5240 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
5241 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
5242 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
5243 		tc->dcb_pfc = pfc_disabled;
5244 	}
5245 
5246 	/* Initialize default user to priority mapping, UPx->TC0 */
5247 	tc = &adapter->dcb_cfg.tc_config[0];
5248 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
5249 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
5250 
5251 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
5252 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
5253 	adapter->dcb_cfg.pfc_mode_enable = false;
5254 	adapter->dcb_set_bitmap = 0x00;
5255 	adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
5256 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
5257 	       sizeof(adapter->temp_dcb_cfg));
5258 
5259 #endif
5260 
5261 	/* default flow control settings */
5262 	hw->fc.requested_mode = ixgbe_fc_full;
5263 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
5264 	ixgbe_pbthresh_setup(adapter);
5265 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
5266 	hw->fc.send_xon = true;
5267 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
5268 
5269 #ifdef CONFIG_PCI_IOV
5270 	if (max_vfs > 0)
5271 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
5272 
5273 	/* assign number of SR-IOV VFs */
5274 	if (hw->mac.type != ixgbe_mac_82598EB) {
5275 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
5276 			adapter->num_vfs = 0;
5277 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
5278 		} else {
5279 			adapter->num_vfs = max_vfs;
5280 		}
5281 	}
5282 #endif /* CONFIG_PCI_IOV */
5283 
5284 	/* enable itr by default in dynamic mode */
5285 	adapter->rx_itr_setting = 1;
5286 	adapter->tx_itr_setting = 1;
5287 
5288 	/* set default ring sizes */
5289 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
5290 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
5291 
5292 	/* set default work limits */
5293 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
5294 
5295 	/* initialize eeprom parameters */
5296 	if (ixgbe_init_eeprom_params_generic(hw)) {
5297 		e_dev_err("EEPROM initialization failed\n");
5298 		return -EIO;
5299 	}
5300 
5301 	/* PF holds first pool slot */
5302 	set_bit(0, &adapter->fwd_bitmask);
5303 	set_bit(__IXGBE_DOWN, &adapter->state);
5304 
5305 	return 0;
5306 }
5307 
5308 /**
5309  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
5310  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
5311  *
5312  * Return 0 on success, negative on failure
5313  **/
5314 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
5315 {
5316 	struct device *dev = tx_ring->dev;
5317 	int orig_node = dev_to_node(dev);
5318 	int ring_node = -1;
5319 	int size;
5320 
5321 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
5322 
5323 	if (tx_ring->q_vector)
5324 		ring_node = tx_ring->q_vector->numa_node;
5325 
5326 	tx_ring->tx_buffer_info = vzalloc_node(size, ring_node);
5327 	if (!tx_ring->tx_buffer_info)
5328 		tx_ring->tx_buffer_info = vzalloc(size);
5329 	if (!tx_ring->tx_buffer_info)
5330 		goto err;
5331 
5332 	u64_stats_init(&tx_ring->syncp);
5333 
5334 	/* round up to nearest 4K */
5335 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
5336 	tx_ring->size = ALIGN(tx_ring->size, 4096);
5337 
5338 	set_dev_node(dev, ring_node);
5339 	tx_ring->desc = dma_alloc_coherent(dev,
5340 					   tx_ring->size,
5341 					   &tx_ring->dma,
5342 					   GFP_KERNEL);
5343 	set_dev_node(dev, orig_node);
5344 	if (!tx_ring->desc)
5345 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
5346 						   &tx_ring->dma, GFP_KERNEL);
5347 	if (!tx_ring->desc)
5348 		goto err;
5349 
5350 	tx_ring->next_to_use = 0;
5351 	tx_ring->next_to_clean = 0;
5352 	return 0;
5353 
5354 err:
5355 	vfree(tx_ring->tx_buffer_info);
5356 	tx_ring->tx_buffer_info = NULL;
5357 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
5358 	return -ENOMEM;
5359 }
5360 
5361 /**
5362  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
5363  * @adapter: board private structure
5364  *
5365  * If this function returns with an error, then it's possible one or
5366  * more of the rings is populated (while the rest are not).  It is the
5367  * callers duty to clean those orphaned rings.
5368  *
5369  * Return 0 on success, negative on failure
5370  **/
5371 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
5372 {
5373 	int i, err = 0;
5374 
5375 	for (i = 0; i < adapter->num_tx_queues; i++) {
5376 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
5377 		if (!err)
5378 			continue;
5379 
5380 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
5381 		goto err_setup_tx;
5382 	}
5383 
5384 	return 0;
5385 err_setup_tx:
5386 	/* rewind the index freeing the rings as we go */
5387 	while (i--)
5388 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
5389 	return err;
5390 }
5391 
5392 /**
5393  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
5394  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
5395  *
5396  * Returns 0 on success, negative on failure
5397  **/
5398 int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
5399 {
5400 	struct device *dev = rx_ring->dev;
5401 	int orig_node = dev_to_node(dev);
5402 	int ring_node = -1;
5403 	int size;
5404 
5405 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
5406 
5407 	if (rx_ring->q_vector)
5408 		ring_node = rx_ring->q_vector->numa_node;
5409 
5410 	rx_ring->rx_buffer_info = vzalloc_node(size, ring_node);
5411 	if (!rx_ring->rx_buffer_info)
5412 		rx_ring->rx_buffer_info = vzalloc(size);
5413 	if (!rx_ring->rx_buffer_info)
5414 		goto err;
5415 
5416 	u64_stats_init(&rx_ring->syncp);
5417 
5418 	/* Round up to nearest 4K */
5419 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
5420 	rx_ring->size = ALIGN(rx_ring->size, 4096);
5421 
5422 	set_dev_node(dev, ring_node);
5423 	rx_ring->desc = dma_alloc_coherent(dev,
5424 					   rx_ring->size,
5425 					   &rx_ring->dma,
5426 					   GFP_KERNEL);
5427 	set_dev_node(dev, orig_node);
5428 	if (!rx_ring->desc)
5429 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
5430 						   &rx_ring->dma, GFP_KERNEL);
5431 	if (!rx_ring->desc)
5432 		goto err;
5433 
5434 	rx_ring->next_to_clean = 0;
5435 	rx_ring->next_to_use = 0;
5436 
5437 	return 0;
5438 err:
5439 	vfree(rx_ring->rx_buffer_info);
5440 	rx_ring->rx_buffer_info = NULL;
5441 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
5442 	return -ENOMEM;
5443 }
5444 
5445 /**
5446  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
5447  * @adapter: board private structure
5448  *
5449  * If this function returns with an error, then it's possible one or
5450  * more of the rings is populated (while the rest are not).  It is the
5451  * callers duty to clean those orphaned rings.
5452  *
5453  * Return 0 on success, negative on failure
5454  **/
5455 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
5456 {
5457 	int i, err = 0;
5458 
5459 	for (i = 0; i < adapter->num_rx_queues; i++) {
5460 		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
5461 		if (!err)
5462 			continue;
5463 
5464 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
5465 		goto err_setup_rx;
5466 	}
5467 
5468 #ifdef IXGBE_FCOE
5469 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
5470 	if (!err)
5471 #endif
5472 		return 0;
5473 err_setup_rx:
5474 	/* rewind the index freeing the rings as we go */
5475 	while (i--)
5476 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
5477 	return err;
5478 }
5479 
5480 /**
5481  * ixgbe_free_tx_resources - Free Tx Resources per Queue
5482  * @tx_ring: Tx descriptor ring for a specific queue
5483  *
5484  * Free all transmit software resources
5485  **/
5486 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
5487 {
5488 	ixgbe_clean_tx_ring(tx_ring);
5489 
5490 	vfree(tx_ring->tx_buffer_info);
5491 	tx_ring->tx_buffer_info = NULL;
5492 
5493 	/* if not set, then don't free */
5494 	if (!tx_ring->desc)
5495 		return;
5496 
5497 	dma_free_coherent(tx_ring->dev, tx_ring->size,
5498 			  tx_ring->desc, tx_ring->dma);
5499 
5500 	tx_ring->desc = NULL;
5501 }
5502 
5503 /**
5504  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
5505  * @adapter: board private structure
5506  *
5507  * Free all transmit software resources
5508  **/
5509 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
5510 {
5511 	int i;
5512 
5513 	for (i = 0; i < adapter->num_tx_queues; i++)
5514 		if (adapter->tx_ring[i]->desc)
5515 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
5516 }
5517 
5518 /**
5519  * ixgbe_free_rx_resources - Free Rx Resources
5520  * @rx_ring: ring to clean the resources from
5521  *
5522  * Free all receive software resources
5523  **/
5524 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
5525 {
5526 	ixgbe_clean_rx_ring(rx_ring);
5527 
5528 	vfree(rx_ring->rx_buffer_info);
5529 	rx_ring->rx_buffer_info = NULL;
5530 
5531 	/* if not set, then don't free */
5532 	if (!rx_ring->desc)
5533 		return;
5534 
5535 	dma_free_coherent(rx_ring->dev, rx_ring->size,
5536 			  rx_ring->desc, rx_ring->dma);
5537 
5538 	rx_ring->desc = NULL;
5539 }
5540 
5541 /**
5542  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
5543  * @adapter: board private structure
5544  *
5545  * Free all receive software resources
5546  **/
5547 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
5548 {
5549 	int i;
5550 
5551 #ifdef IXGBE_FCOE
5552 	ixgbe_free_fcoe_ddp_resources(adapter);
5553 
5554 #endif
5555 	for (i = 0; i < adapter->num_rx_queues; i++)
5556 		if (adapter->rx_ring[i]->desc)
5557 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
5558 }
5559 
5560 /**
5561  * ixgbe_change_mtu - Change the Maximum Transfer Unit
5562  * @netdev: network interface device structure
5563  * @new_mtu: new value for maximum frame size
5564  *
5565  * Returns 0 on success, negative on failure
5566  **/
5567 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
5568 {
5569 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5570 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
5571 
5572 	/* MTU < 68 is an error and causes problems on some kernels */
5573 	if ((new_mtu < 68) || (max_frame > IXGBE_MAX_JUMBO_FRAME_SIZE))
5574 		return -EINVAL;
5575 
5576 	/*
5577 	 * For 82599EB we cannot allow legacy VFs to enable their receive
5578 	 * paths when MTU greater than 1500 is configured.  So display a
5579 	 * warning that legacy VFs will be disabled.
5580 	 */
5581 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
5582 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
5583 	    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
5584 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
5585 
5586 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
5587 
5588 	/* must set new MTU before calling down or up */
5589 	netdev->mtu = new_mtu;
5590 
5591 	if (netif_running(netdev))
5592 		ixgbe_reinit_locked(adapter);
5593 
5594 	return 0;
5595 }
5596 
5597 /**
5598  * ixgbe_open - Called when a network interface is made active
5599  * @netdev: network interface device structure
5600  *
5601  * Returns 0 on success, negative value on failure
5602  *
5603  * The open entry point is called when a network interface is made
5604  * active by the system (IFF_UP).  At this point all resources needed
5605  * for transmit and receive operations are allocated, the interrupt
5606  * handler is registered with the OS, the watchdog timer is started,
5607  * and the stack is notified that the interface is ready.
5608  **/
5609 static int ixgbe_open(struct net_device *netdev)
5610 {
5611 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5612 	int err, queues;
5613 
5614 	/* disallow open during test */
5615 	if (test_bit(__IXGBE_TESTING, &adapter->state))
5616 		return -EBUSY;
5617 
5618 	netif_carrier_off(netdev);
5619 
5620 	/* allocate transmit descriptors */
5621 	err = ixgbe_setup_all_tx_resources(adapter);
5622 	if (err)
5623 		goto err_setup_tx;
5624 
5625 	/* allocate receive descriptors */
5626 	err = ixgbe_setup_all_rx_resources(adapter);
5627 	if (err)
5628 		goto err_setup_rx;
5629 
5630 	ixgbe_configure(adapter);
5631 
5632 	err = ixgbe_request_irq(adapter);
5633 	if (err)
5634 		goto err_req_irq;
5635 
5636 	/* Notify the stack of the actual queue counts. */
5637 	if (adapter->num_rx_pools > 1)
5638 		queues = adapter->num_rx_queues_per_pool;
5639 	else
5640 		queues = adapter->num_tx_queues;
5641 
5642 	err = netif_set_real_num_tx_queues(netdev, queues);
5643 	if (err)
5644 		goto err_set_queues;
5645 
5646 	if (adapter->num_rx_pools > 1 &&
5647 	    adapter->num_rx_queues > IXGBE_MAX_L2A_QUEUES)
5648 		queues = IXGBE_MAX_L2A_QUEUES;
5649 	else
5650 		queues = adapter->num_rx_queues;
5651 	err = netif_set_real_num_rx_queues(netdev, queues);
5652 	if (err)
5653 		goto err_set_queues;
5654 
5655 	ixgbe_ptp_init(adapter);
5656 
5657 	ixgbe_up_complete(adapter);
5658 
5659 #if IS_ENABLED(CONFIG_IXGBE_VXLAN)
5660 	vxlan_get_rx_port(netdev);
5661 
5662 #endif
5663 	return 0;
5664 
5665 err_set_queues:
5666 	ixgbe_free_irq(adapter);
5667 err_req_irq:
5668 	ixgbe_free_all_rx_resources(adapter);
5669 err_setup_rx:
5670 	ixgbe_free_all_tx_resources(adapter);
5671 err_setup_tx:
5672 	ixgbe_reset(adapter);
5673 
5674 	return err;
5675 }
5676 
5677 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
5678 {
5679 	ixgbe_ptp_suspend(adapter);
5680 
5681 	ixgbe_down(adapter);
5682 	ixgbe_free_irq(adapter);
5683 
5684 	ixgbe_free_all_tx_resources(adapter);
5685 	ixgbe_free_all_rx_resources(adapter);
5686 }
5687 
5688 /**
5689  * ixgbe_close - Disables a network interface
5690  * @netdev: network interface device structure
5691  *
5692  * Returns 0, this is not allowed to fail
5693  *
5694  * The close entry point is called when an interface is de-activated
5695  * by the OS.  The hardware is still under the drivers control, but
5696  * needs to be disabled.  A global MAC reset is issued to stop the
5697  * hardware, and all transmit and receive resources are freed.
5698  **/
5699 static int ixgbe_close(struct net_device *netdev)
5700 {
5701 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5702 
5703 	ixgbe_ptp_stop(adapter);
5704 
5705 	ixgbe_close_suspend(adapter);
5706 
5707 	ixgbe_fdir_filter_exit(adapter);
5708 
5709 	ixgbe_release_hw_control(adapter);
5710 
5711 	return 0;
5712 }
5713 
5714 #ifdef CONFIG_PM
5715 static int ixgbe_resume(struct pci_dev *pdev)
5716 {
5717 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
5718 	struct net_device *netdev = adapter->netdev;
5719 	u32 err;
5720 
5721 	adapter->hw.hw_addr = adapter->io_addr;
5722 	pci_set_power_state(pdev, PCI_D0);
5723 	pci_restore_state(pdev);
5724 	/*
5725 	 * pci_restore_state clears dev->state_saved so call
5726 	 * pci_save_state to restore it.
5727 	 */
5728 	pci_save_state(pdev);
5729 
5730 	err = pci_enable_device_mem(pdev);
5731 	if (err) {
5732 		e_dev_err("Cannot enable PCI device from suspend\n");
5733 		return err;
5734 	}
5735 	smp_mb__before_atomic();
5736 	clear_bit(__IXGBE_DISABLED, &adapter->state);
5737 	pci_set_master(pdev);
5738 
5739 	pci_wake_from_d3(pdev, false);
5740 
5741 	ixgbe_reset(adapter);
5742 
5743 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
5744 
5745 	rtnl_lock();
5746 	err = ixgbe_init_interrupt_scheme(adapter);
5747 	if (!err && netif_running(netdev))
5748 		err = ixgbe_open(netdev);
5749 
5750 	rtnl_unlock();
5751 
5752 	if (err)
5753 		return err;
5754 
5755 	netif_device_attach(netdev);
5756 
5757 	return 0;
5758 }
5759 #endif /* CONFIG_PM */
5760 
5761 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
5762 {
5763 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
5764 	struct net_device *netdev = adapter->netdev;
5765 	struct ixgbe_hw *hw = &adapter->hw;
5766 	u32 ctrl, fctrl;
5767 	u32 wufc = adapter->wol;
5768 #ifdef CONFIG_PM
5769 	int retval = 0;
5770 #endif
5771 
5772 	netif_device_detach(netdev);
5773 
5774 	rtnl_lock();
5775 	if (netif_running(netdev))
5776 		ixgbe_close_suspend(adapter);
5777 	rtnl_unlock();
5778 
5779 	ixgbe_clear_interrupt_scheme(adapter);
5780 
5781 #ifdef CONFIG_PM
5782 	retval = pci_save_state(pdev);
5783 	if (retval)
5784 		return retval;
5785 
5786 #endif
5787 	if (hw->mac.ops.stop_link_on_d3)
5788 		hw->mac.ops.stop_link_on_d3(hw);
5789 
5790 	if (wufc) {
5791 		ixgbe_set_rx_mode(netdev);
5792 
5793 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
5794 		if (hw->mac.ops.enable_tx_laser)
5795 			hw->mac.ops.enable_tx_laser(hw);
5796 
5797 		/* turn on all-multi mode if wake on multicast is enabled */
5798 		if (wufc & IXGBE_WUFC_MC) {
5799 			fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
5800 			fctrl |= IXGBE_FCTRL_MPE;
5801 			IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
5802 		}
5803 
5804 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
5805 		ctrl |= IXGBE_CTRL_GIO_DIS;
5806 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
5807 
5808 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
5809 	} else {
5810 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
5811 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
5812 	}
5813 
5814 	switch (hw->mac.type) {
5815 	case ixgbe_mac_82598EB:
5816 		pci_wake_from_d3(pdev, false);
5817 		break;
5818 	case ixgbe_mac_82599EB:
5819 	case ixgbe_mac_X540:
5820 	case ixgbe_mac_X550:
5821 	case ixgbe_mac_X550EM_x:
5822 		pci_wake_from_d3(pdev, !!wufc);
5823 		break;
5824 	default:
5825 		break;
5826 	}
5827 
5828 	*enable_wake = !!wufc;
5829 
5830 	ixgbe_release_hw_control(adapter);
5831 
5832 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
5833 		pci_disable_device(pdev);
5834 
5835 	return 0;
5836 }
5837 
5838 #ifdef CONFIG_PM
5839 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
5840 {
5841 	int retval;
5842 	bool wake;
5843 
5844 	retval = __ixgbe_shutdown(pdev, &wake);
5845 	if (retval)
5846 		return retval;
5847 
5848 	if (wake) {
5849 		pci_prepare_to_sleep(pdev);
5850 	} else {
5851 		pci_wake_from_d3(pdev, false);
5852 		pci_set_power_state(pdev, PCI_D3hot);
5853 	}
5854 
5855 	return 0;
5856 }
5857 #endif /* CONFIG_PM */
5858 
5859 static void ixgbe_shutdown(struct pci_dev *pdev)
5860 {
5861 	bool wake;
5862 
5863 	__ixgbe_shutdown(pdev, &wake);
5864 
5865 	if (system_state == SYSTEM_POWER_OFF) {
5866 		pci_wake_from_d3(pdev, wake);
5867 		pci_set_power_state(pdev, PCI_D3hot);
5868 	}
5869 }
5870 
5871 /**
5872  * ixgbe_update_stats - Update the board statistics counters.
5873  * @adapter: board private structure
5874  **/
5875 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
5876 {
5877 	struct net_device *netdev = adapter->netdev;
5878 	struct ixgbe_hw *hw = &adapter->hw;
5879 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
5880 	u64 total_mpc = 0;
5881 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
5882 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
5883 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
5884 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
5885 
5886 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
5887 	    test_bit(__IXGBE_RESETTING, &adapter->state))
5888 		return;
5889 
5890 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
5891 		u64 rsc_count = 0;
5892 		u64 rsc_flush = 0;
5893 		for (i = 0; i < adapter->num_rx_queues; i++) {
5894 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
5895 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
5896 		}
5897 		adapter->rsc_total_count = rsc_count;
5898 		adapter->rsc_total_flush = rsc_flush;
5899 	}
5900 
5901 	for (i = 0; i < adapter->num_rx_queues; i++) {
5902 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
5903 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
5904 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
5905 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
5906 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
5907 		bytes += rx_ring->stats.bytes;
5908 		packets += rx_ring->stats.packets;
5909 	}
5910 	adapter->non_eop_descs = non_eop_descs;
5911 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
5912 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
5913 	adapter->hw_csum_rx_error = hw_csum_rx_error;
5914 	netdev->stats.rx_bytes = bytes;
5915 	netdev->stats.rx_packets = packets;
5916 
5917 	bytes = 0;
5918 	packets = 0;
5919 	/* gather some stats to the adapter struct that are per queue */
5920 	for (i = 0; i < adapter->num_tx_queues; i++) {
5921 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
5922 		restart_queue += tx_ring->tx_stats.restart_queue;
5923 		tx_busy += tx_ring->tx_stats.tx_busy;
5924 		bytes += tx_ring->stats.bytes;
5925 		packets += tx_ring->stats.packets;
5926 	}
5927 	adapter->restart_queue = restart_queue;
5928 	adapter->tx_busy = tx_busy;
5929 	netdev->stats.tx_bytes = bytes;
5930 	netdev->stats.tx_packets = packets;
5931 
5932 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
5933 
5934 	/* 8 register reads */
5935 	for (i = 0; i < 8; i++) {
5936 		/* for packet buffers not used, the register should read 0 */
5937 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
5938 		missed_rx += mpc;
5939 		hwstats->mpc[i] += mpc;
5940 		total_mpc += hwstats->mpc[i];
5941 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
5942 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
5943 		switch (hw->mac.type) {
5944 		case ixgbe_mac_82598EB:
5945 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
5946 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
5947 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
5948 			hwstats->pxonrxc[i] +=
5949 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
5950 			break;
5951 		case ixgbe_mac_82599EB:
5952 		case ixgbe_mac_X540:
5953 		case ixgbe_mac_X550:
5954 		case ixgbe_mac_X550EM_x:
5955 			hwstats->pxonrxc[i] +=
5956 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
5957 			break;
5958 		default:
5959 			break;
5960 		}
5961 	}
5962 
5963 	/*16 register reads */
5964 	for (i = 0; i < 16; i++) {
5965 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
5966 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
5967 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
5968 		    (hw->mac.type == ixgbe_mac_X540) ||
5969 		    (hw->mac.type == ixgbe_mac_X550) ||
5970 		    (hw->mac.type == ixgbe_mac_X550EM_x)) {
5971 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
5972 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
5973 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
5974 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
5975 		}
5976 	}
5977 
5978 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
5979 	/* work around hardware counting issue */
5980 	hwstats->gprc -= missed_rx;
5981 
5982 	ixgbe_update_xoff_received(adapter);
5983 
5984 	/* 82598 hardware only has a 32 bit counter in the high register */
5985 	switch (hw->mac.type) {
5986 	case ixgbe_mac_82598EB:
5987 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
5988 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
5989 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
5990 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
5991 		break;
5992 	case ixgbe_mac_X540:
5993 	case ixgbe_mac_X550:
5994 	case ixgbe_mac_X550EM_x:
5995 		/* OS2BMC stats are X540 and later */
5996 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
5997 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
5998 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
5999 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
6000 	case ixgbe_mac_82599EB:
6001 		for (i = 0; i < 16; i++)
6002 			adapter->hw_rx_no_dma_resources +=
6003 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
6004 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
6005 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
6006 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
6007 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
6008 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
6009 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
6010 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
6011 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
6012 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
6013 #ifdef IXGBE_FCOE
6014 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
6015 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
6016 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
6017 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
6018 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
6019 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
6020 		/* Add up per cpu counters for total ddp aloc fail */
6021 		if (adapter->fcoe.ddp_pool) {
6022 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
6023 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
6024 			unsigned int cpu;
6025 			u64 noddp = 0, noddp_ext_buff = 0;
6026 			for_each_possible_cpu(cpu) {
6027 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
6028 				noddp += ddp_pool->noddp;
6029 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
6030 			}
6031 			hwstats->fcoe_noddp = noddp;
6032 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
6033 		}
6034 #endif /* IXGBE_FCOE */
6035 		break;
6036 	default:
6037 		break;
6038 	}
6039 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
6040 	hwstats->bprc += bprc;
6041 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
6042 	if (hw->mac.type == ixgbe_mac_82598EB)
6043 		hwstats->mprc -= bprc;
6044 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
6045 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
6046 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
6047 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
6048 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
6049 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
6050 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
6051 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
6052 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
6053 	hwstats->lxontxc += lxon;
6054 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
6055 	hwstats->lxofftxc += lxoff;
6056 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
6057 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
6058 	/*
6059 	 * 82598 errata - tx of flow control packets is included in tx counters
6060 	 */
6061 	xon_off_tot = lxon + lxoff;
6062 	hwstats->gptc -= xon_off_tot;
6063 	hwstats->mptc -= xon_off_tot;
6064 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
6065 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
6066 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
6067 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
6068 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
6069 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
6070 	hwstats->ptc64 -= xon_off_tot;
6071 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
6072 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
6073 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
6074 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
6075 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
6076 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
6077 
6078 	/* Fill out the OS statistics structure */
6079 	netdev->stats.multicast = hwstats->mprc;
6080 
6081 	/* Rx Errors */
6082 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
6083 	netdev->stats.rx_dropped = 0;
6084 	netdev->stats.rx_length_errors = hwstats->rlec;
6085 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
6086 	netdev->stats.rx_missed_errors = total_mpc;
6087 }
6088 
6089 /**
6090  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
6091  * @adapter: pointer to the device adapter structure
6092  **/
6093 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
6094 {
6095 	struct ixgbe_hw *hw = &adapter->hw;
6096 	int i;
6097 
6098 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
6099 		return;
6100 
6101 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6102 
6103 	/* if interface is down do nothing */
6104 	if (test_bit(__IXGBE_DOWN, &adapter->state))
6105 		return;
6106 
6107 	/* do nothing if we are not using signature filters */
6108 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
6109 		return;
6110 
6111 	adapter->fdir_overflow++;
6112 
6113 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
6114 		for (i = 0; i < adapter->num_tx_queues; i++)
6115 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
6116 				&(adapter->tx_ring[i]->state));
6117 		/* re-enable flow director interrupts */
6118 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
6119 	} else {
6120 		e_err(probe, "failed to finish FDIR re-initialization, "
6121 		      "ignored adding FDIR ATR filters\n");
6122 	}
6123 }
6124 
6125 /**
6126  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
6127  * @adapter: pointer to the device adapter structure
6128  *
6129  * This function serves two purposes.  First it strobes the interrupt lines
6130  * in order to make certain interrupts are occurring.  Secondly it sets the
6131  * bits needed to check for TX hangs.  As a result we should immediately
6132  * determine if a hang has occurred.
6133  */
6134 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
6135 {
6136 	struct ixgbe_hw *hw = &adapter->hw;
6137 	u64 eics = 0;
6138 	int i;
6139 
6140 	/* If we're down, removing or resetting, just bail */
6141 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6142 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6143 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6144 		return;
6145 
6146 	/* Force detection of hung controller */
6147 	if (netif_carrier_ok(adapter->netdev)) {
6148 		for (i = 0; i < adapter->num_tx_queues; i++)
6149 			set_check_for_tx_hang(adapter->tx_ring[i]);
6150 	}
6151 
6152 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
6153 		/*
6154 		 * for legacy and MSI interrupts don't set any bits
6155 		 * that are enabled for EIAM, because this operation
6156 		 * would set *both* EIMS and EICS for any bit in EIAM
6157 		 */
6158 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
6159 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
6160 	} else {
6161 		/* get one bit for every active tx/rx interrupt vector */
6162 		for (i = 0; i < adapter->num_q_vectors; i++) {
6163 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
6164 			if (qv->rx.ring || qv->tx.ring)
6165 				eics |= ((u64)1 << i);
6166 		}
6167 	}
6168 
6169 	/* Cause software interrupt to ensure rings are cleaned */
6170 	ixgbe_irq_rearm_queues(adapter, eics);
6171 }
6172 
6173 /**
6174  * ixgbe_watchdog_update_link - update the link status
6175  * @adapter: pointer to the device adapter structure
6176  * @link_speed: pointer to a u32 to store the link_speed
6177  **/
6178 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
6179 {
6180 	struct ixgbe_hw *hw = &adapter->hw;
6181 	u32 link_speed = adapter->link_speed;
6182 	bool link_up = adapter->link_up;
6183 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
6184 
6185 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
6186 		return;
6187 
6188 	if (hw->mac.ops.check_link) {
6189 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
6190 	} else {
6191 		/* always assume link is up, if no check link function */
6192 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
6193 		link_up = true;
6194 	}
6195 
6196 	if (adapter->ixgbe_ieee_pfc)
6197 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
6198 
6199 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
6200 		hw->mac.ops.fc_enable(hw);
6201 		ixgbe_set_rx_drop_en(adapter);
6202 	}
6203 
6204 	if (link_up ||
6205 	    time_after(jiffies, (adapter->link_check_timeout +
6206 				 IXGBE_TRY_LINK_TIMEOUT))) {
6207 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6208 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
6209 		IXGBE_WRITE_FLUSH(hw);
6210 	}
6211 
6212 	adapter->link_up = link_up;
6213 	adapter->link_speed = link_speed;
6214 }
6215 
6216 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
6217 {
6218 #ifdef CONFIG_IXGBE_DCB
6219 	struct net_device *netdev = adapter->netdev;
6220 	struct dcb_app app = {
6221 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
6222 			      .protocol = 0,
6223 			     };
6224 	u8 up = 0;
6225 
6226 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
6227 		up = dcb_ieee_getapp_mask(netdev, &app);
6228 
6229 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
6230 #endif
6231 }
6232 
6233 /**
6234  * ixgbe_watchdog_link_is_up - update netif_carrier status and
6235  *                             print link up message
6236  * @adapter: pointer to the device adapter structure
6237  **/
6238 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
6239 {
6240 	struct net_device *netdev = adapter->netdev;
6241 	struct ixgbe_hw *hw = &adapter->hw;
6242 	struct net_device *upper;
6243 	struct list_head *iter;
6244 	u32 link_speed = adapter->link_speed;
6245 	bool flow_rx, flow_tx;
6246 
6247 	/* only continue if link was previously down */
6248 	if (netif_carrier_ok(netdev))
6249 		return;
6250 
6251 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
6252 
6253 	switch (hw->mac.type) {
6254 	case ixgbe_mac_82598EB: {
6255 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6256 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
6257 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
6258 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
6259 	}
6260 		break;
6261 	case ixgbe_mac_X540:
6262 	case ixgbe_mac_X550:
6263 	case ixgbe_mac_X550EM_x:
6264 	case ixgbe_mac_82599EB: {
6265 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
6266 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
6267 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
6268 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
6269 	}
6270 		break;
6271 	default:
6272 		flow_tx = false;
6273 		flow_rx = false;
6274 		break;
6275 	}
6276 
6277 	adapter->last_rx_ptp_check = jiffies;
6278 
6279 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6280 		ixgbe_ptp_start_cyclecounter(adapter);
6281 
6282 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n",
6283 	       (link_speed == IXGBE_LINK_SPEED_10GB_FULL ?
6284 	       "10 Gbps" :
6285 	       (link_speed == IXGBE_LINK_SPEED_1GB_FULL ?
6286 	       "1 Gbps" :
6287 	       (link_speed == IXGBE_LINK_SPEED_100_FULL ?
6288 	       "100 Mbps" :
6289 	       "unknown speed"))),
6290 	       ((flow_rx && flow_tx) ? "RX/TX" :
6291 	       (flow_rx ? "RX" :
6292 	       (flow_tx ? "TX" : "None"))));
6293 
6294 	netif_carrier_on(netdev);
6295 	ixgbe_check_vf_rate_limit(adapter);
6296 
6297 	/* enable transmits */
6298 	netif_tx_wake_all_queues(adapter->netdev);
6299 
6300 	/* enable any upper devices */
6301 	rtnl_lock();
6302 	netdev_for_each_all_upper_dev_rcu(adapter->netdev, upper, iter) {
6303 		if (netif_is_macvlan(upper)) {
6304 			struct macvlan_dev *vlan = netdev_priv(upper);
6305 
6306 			if (vlan->fwd_priv)
6307 				netif_tx_wake_all_queues(upper);
6308 		}
6309 	}
6310 	rtnl_unlock();
6311 
6312 	/* update the default user priority for VFs */
6313 	ixgbe_update_default_up(adapter);
6314 
6315 	/* ping all the active vfs to let them know link has changed */
6316 	ixgbe_ping_all_vfs(adapter);
6317 }
6318 
6319 /**
6320  * ixgbe_watchdog_link_is_down - update netif_carrier status and
6321  *                               print link down message
6322  * @adapter: pointer to the adapter structure
6323  **/
6324 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
6325 {
6326 	struct net_device *netdev = adapter->netdev;
6327 	struct ixgbe_hw *hw = &adapter->hw;
6328 
6329 	adapter->link_up = false;
6330 	adapter->link_speed = 0;
6331 
6332 	/* only continue if link was up previously */
6333 	if (!netif_carrier_ok(netdev))
6334 		return;
6335 
6336 	/* poll for SFP+ cable when link is down */
6337 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
6338 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
6339 
6340 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6341 		ixgbe_ptp_start_cyclecounter(adapter);
6342 
6343 	e_info(drv, "NIC Link is Down\n");
6344 	netif_carrier_off(netdev);
6345 
6346 	/* ping all the active vfs to let them know link has changed */
6347 	ixgbe_ping_all_vfs(adapter);
6348 }
6349 
6350 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
6351 {
6352 	int i;
6353 
6354 	for (i = 0; i < adapter->num_tx_queues; i++) {
6355 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6356 
6357 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
6358 			return true;
6359 	}
6360 
6361 	return false;
6362 }
6363 
6364 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
6365 {
6366 	struct ixgbe_hw *hw = &adapter->hw;
6367 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
6368 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
6369 
6370 	int i, j;
6371 
6372 	if (!adapter->num_vfs)
6373 		return false;
6374 
6375 	/* resetting the PF is only needed for MAC before X550 */
6376 	if (hw->mac.type >= ixgbe_mac_X550)
6377 		return false;
6378 
6379 	for (i = 0; i < adapter->num_vfs; i++) {
6380 		for (j = 0; j < q_per_pool; j++) {
6381 			u32 h, t;
6382 
6383 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
6384 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
6385 
6386 			if (h != t)
6387 				return true;
6388 		}
6389 	}
6390 
6391 	return false;
6392 }
6393 
6394 /**
6395  * ixgbe_watchdog_flush_tx - flush queues on link down
6396  * @adapter: pointer to the device adapter structure
6397  **/
6398 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
6399 {
6400 	if (!netif_carrier_ok(adapter->netdev)) {
6401 		if (ixgbe_ring_tx_pending(adapter) ||
6402 		    ixgbe_vf_tx_pending(adapter)) {
6403 			/* We've lost link, so the controller stops DMA,
6404 			 * but we've got queued Tx work that's never going
6405 			 * to get done, so reset controller to flush Tx.
6406 			 * (Do the reset outside of interrupt context).
6407 			 */
6408 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
6409 			adapter->flags2 |= IXGBE_FLAG2_RESET_REQUESTED;
6410 		}
6411 	}
6412 }
6413 
6414 #ifdef CONFIG_PCI_IOV
6415 static inline void ixgbe_issue_vf_flr(struct ixgbe_adapter *adapter,
6416 				      struct pci_dev *vfdev)
6417 {
6418 	if (!pci_wait_for_pending_transaction(vfdev))
6419 		e_dev_warn("Issuing VFLR with pending transactions\n");
6420 
6421 	e_dev_err("Issuing VFLR for VF %s\n", pci_name(vfdev));
6422 	pcie_capability_set_word(vfdev, PCI_EXP_DEVCTL, PCI_EXP_DEVCTL_BCR_FLR);
6423 
6424 	msleep(100);
6425 }
6426 
6427 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
6428 {
6429 	struct ixgbe_hw *hw = &adapter->hw;
6430 	struct pci_dev *pdev = adapter->pdev;
6431 	struct pci_dev *vfdev;
6432 	u32 gpc;
6433 	int pos;
6434 	unsigned short vf_id;
6435 
6436 	if (!(netif_carrier_ok(adapter->netdev)))
6437 		return;
6438 
6439 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
6440 	if (gpc) /* If incrementing then no need for the check below */
6441 		return;
6442 	/* Check to see if a bad DMA write target from an errant or
6443 	 * malicious VF has caused a PCIe error.  If so then we can
6444 	 * issue a VFLR to the offending VF(s) and then resume without
6445 	 * requesting a full slot reset.
6446 	 */
6447 
6448 	if (!pdev)
6449 		return;
6450 
6451 	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_SRIOV);
6452 	if (!pos)
6453 		return;
6454 
6455 	/* get the device ID for the VF */
6456 	pci_read_config_word(pdev, pos + PCI_SRIOV_VF_DID, &vf_id);
6457 
6458 	/* check status reg for all VFs owned by this PF */
6459 	vfdev = pci_get_device(pdev->vendor, vf_id, NULL);
6460 	while (vfdev) {
6461 		if (vfdev->is_virtfn && (vfdev->physfn == pdev)) {
6462 			u16 status_reg;
6463 
6464 			pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
6465 			if (status_reg & PCI_STATUS_REC_MASTER_ABORT)
6466 				/* issue VFLR */
6467 				ixgbe_issue_vf_flr(adapter, vfdev);
6468 		}
6469 
6470 		vfdev = pci_get_device(pdev->vendor, vf_id, vfdev);
6471 	}
6472 }
6473 
6474 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
6475 {
6476 	u32 ssvpc;
6477 
6478 	/* Do not perform spoof check for 82598 or if not in IOV mode */
6479 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
6480 	    adapter->num_vfs == 0)
6481 		return;
6482 
6483 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
6484 
6485 	/*
6486 	 * ssvpc register is cleared on read, if zero then no
6487 	 * spoofed packets in the last interval.
6488 	 */
6489 	if (!ssvpc)
6490 		return;
6491 
6492 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
6493 }
6494 #else
6495 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
6496 {
6497 }
6498 
6499 static void
6500 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
6501 {
6502 }
6503 #endif /* CONFIG_PCI_IOV */
6504 
6505 
6506 /**
6507  * ixgbe_watchdog_subtask - check and bring link up
6508  * @adapter: pointer to the device adapter structure
6509  **/
6510 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
6511 {
6512 	/* if interface is down, removing or resetting, do nothing */
6513 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6514 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6515 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6516 		return;
6517 
6518 	ixgbe_watchdog_update_link(adapter);
6519 
6520 	if (adapter->link_up)
6521 		ixgbe_watchdog_link_is_up(adapter);
6522 	else
6523 		ixgbe_watchdog_link_is_down(adapter);
6524 
6525 	ixgbe_check_for_bad_vf(adapter);
6526 	ixgbe_spoof_check(adapter);
6527 	ixgbe_update_stats(adapter);
6528 
6529 	ixgbe_watchdog_flush_tx(adapter);
6530 }
6531 
6532 /**
6533  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
6534  * @adapter: the ixgbe adapter structure
6535  **/
6536 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
6537 {
6538 	struct ixgbe_hw *hw = &adapter->hw;
6539 	s32 err;
6540 
6541 	/* not searching for SFP so there is nothing to do here */
6542 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
6543 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
6544 		return;
6545 
6546 	/* someone else is in init, wait until next service event */
6547 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
6548 		return;
6549 
6550 	err = hw->phy.ops.identify_sfp(hw);
6551 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
6552 		goto sfp_out;
6553 
6554 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
6555 		/* If no cable is present, then we need to reset
6556 		 * the next time we find a good cable. */
6557 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
6558 	}
6559 
6560 	/* exit on error */
6561 	if (err)
6562 		goto sfp_out;
6563 
6564 	/* exit if reset not needed */
6565 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
6566 		goto sfp_out;
6567 
6568 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
6569 
6570 	/*
6571 	 * A module may be identified correctly, but the EEPROM may not have
6572 	 * support for that module.  setup_sfp() will fail in that case, so
6573 	 * we should not allow that module to load.
6574 	 */
6575 	if (hw->mac.type == ixgbe_mac_82598EB)
6576 		err = hw->phy.ops.reset(hw);
6577 	else
6578 		err = hw->mac.ops.setup_sfp(hw);
6579 
6580 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
6581 		goto sfp_out;
6582 
6583 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
6584 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
6585 
6586 sfp_out:
6587 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
6588 
6589 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
6590 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
6591 		e_dev_err("failed to initialize because an unsupported "
6592 			  "SFP+ module type was detected.\n");
6593 		e_dev_err("Reload the driver after installing a "
6594 			  "supported module.\n");
6595 		unregister_netdev(adapter->netdev);
6596 	}
6597 }
6598 
6599 /**
6600  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
6601  * @adapter: the ixgbe adapter structure
6602  **/
6603 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
6604 {
6605 	struct ixgbe_hw *hw = &adapter->hw;
6606 	u32 speed;
6607 	bool autoneg = false;
6608 
6609 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
6610 		return;
6611 
6612 	/* someone else is in init, wait until next service event */
6613 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
6614 		return;
6615 
6616 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
6617 
6618 	speed = hw->phy.autoneg_advertised;
6619 	if ((!speed) && (hw->mac.ops.get_link_capabilities)) {
6620 		hw->mac.ops.get_link_capabilities(hw, &speed, &autoneg);
6621 
6622 		/* setup the highest link when no autoneg */
6623 		if (!autoneg) {
6624 			if (speed & IXGBE_LINK_SPEED_10GB_FULL)
6625 				speed = IXGBE_LINK_SPEED_10GB_FULL;
6626 		}
6627 	}
6628 
6629 	if (hw->mac.ops.setup_link)
6630 		hw->mac.ops.setup_link(hw, speed, true);
6631 
6632 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
6633 	adapter->link_check_timeout = jiffies;
6634 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
6635 }
6636 
6637 /**
6638  * ixgbe_service_timer - Timer Call-back
6639  * @data: pointer to adapter cast into an unsigned long
6640  **/
6641 static void ixgbe_service_timer(unsigned long data)
6642 {
6643 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)data;
6644 	unsigned long next_event_offset;
6645 
6646 	/* poll faster when waiting for link */
6647 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
6648 		next_event_offset = HZ / 10;
6649 	else
6650 		next_event_offset = HZ * 2;
6651 
6652 	/* Reset the timer */
6653 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
6654 
6655 	ixgbe_service_event_schedule(adapter);
6656 }
6657 
6658 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
6659 {
6660 	if (!(adapter->flags2 & IXGBE_FLAG2_RESET_REQUESTED))
6661 		return;
6662 
6663 	adapter->flags2 &= ~IXGBE_FLAG2_RESET_REQUESTED;
6664 
6665 	/* If we're already down, removing or resetting, just bail */
6666 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6667 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
6668 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6669 		return;
6670 
6671 	ixgbe_dump(adapter);
6672 	netdev_err(adapter->netdev, "Reset adapter\n");
6673 	adapter->tx_timeout_count++;
6674 
6675 	rtnl_lock();
6676 	ixgbe_reinit_locked(adapter);
6677 	rtnl_unlock();
6678 }
6679 
6680 /**
6681  * ixgbe_service_task - manages and runs subtasks
6682  * @work: pointer to work_struct containing our data
6683  **/
6684 static void ixgbe_service_task(struct work_struct *work)
6685 {
6686 	struct ixgbe_adapter *adapter = container_of(work,
6687 						     struct ixgbe_adapter,
6688 						     service_task);
6689 	if (ixgbe_removed(adapter->hw.hw_addr)) {
6690 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
6691 			rtnl_lock();
6692 			ixgbe_down(adapter);
6693 			rtnl_unlock();
6694 		}
6695 		ixgbe_service_event_complete(adapter);
6696 		return;
6697 	}
6698 	ixgbe_reset_subtask(adapter);
6699 	ixgbe_sfp_detection_subtask(adapter);
6700 	ixgbe_sfp_link_config_subtask(adapter);
6701 	ixgbe_check_overtemp_subtask(adapter);
6702 	ixgbe_watchdog_subtask(adapter);
6703 	ixgbe_fdir_reinit_subtask(adapter);
6704 	ixgbe_check_hang_subtask(adapter);
6705 
6706 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
6707 		ixgbe_ptp_overflow_check(adapter);
6708 		ixgbe_ptp_rx_hang(adapter);
6709 	}
6710 
6711 	ixgbe_service_event_complete(adapter);
6712 }
6713 
6714 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
6715 		     struct ixgbe_tx_buffer *first,
6716 		     u8 *hdr_len)
6717 {
6718 	struct sk_buff *skb = first->skb;
6719 	u32 vlan_macip_lens, type_tucmd;
6720 	u32 mss_l4len_idx, l4len;
6721 	int err;
6722 
6723 	if (skb->ip_summed != CHECKSUM_PARTIAL)
6724 		return 0;
6725 
6726 	if (!skb_is_gso(skb))
6727 		return 0;
6728 
6729 	err = skb_cow_head(skb, 0);
6730 	if (err < 0)
6731 		return err;
6732 
6733 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
6734 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
6735 
6736 	if (first->protocol == htons(ETH_P_IP)) {
6737 		struct iphdr *iph = ip_hdr(skb);
6738 		iph->tot_len = 0;
6739 		iph->check = 0;
6740 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
6741 							 iph->daddr, 0,
6742 							 IPPROTO_TCP,
6743 							 0);
6744 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
6745 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
6746 				   IXGBE_TX_FLAGS_CSUM |
6747 				   IXGBE_TX_FLAGS_IPV4;
6748 	} else if (skb_is_gso_v6(skb)) {
6749 		ipv6_hdr(skb)->payload_len = 0;
6750 		tcp_hdr(skb)->check =
6751 		    ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
6752 				     &ipv6_hdr(skb)->daddr,
6753 				     0, IPPROTO_TCP, 0);
6754 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
6755 				   IXGBE_TX_FLAGS_CSUM;
6756 	}
6757 
6758 	/* compute header lengths */
6759 	l4len = tcp_hdrlen(skb);
6760 	*hdr_len = skb_transport_offset(skb) + l4len;
6761 
6762 	/* update gso size and bytecount with header size */
6763 	first->gso_segs = skb_shinfo(skb)->gso_segs;
6764 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
6765 
6766 	/* mss_l4len_id: use 0 as index for TSO */
6767 	mss_l4len_idx = l4len << IXGBE_ADVTXD_L4LEN_SHIFT;
6768 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
6769 
6770 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
6771 	vlan_macip_lens = skb_network_header_len(skb);
6772 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
6773 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
6774 
6775 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd,
6776 			  mss_l4len_idx);
6777 
6778 	return 1;
6779 }
6780 
6781 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
6782 			  struct ixgbe_tx_buffer *first)
6783 {
6784 	struct sk_buff *skb = first->skb;
6785 	u32 vlan_macip_lens = 0;
6786 	u32 mss_l4len_idx = 0;
6787 	u32 type_tucmd = 0;
6788 
6789 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
6790 		if (!(first->tx_flags & IXGBE_TX_FLAGS_HW_VLAN) &&
6791 		    !(first->tx_flags & IXGBE_TX_FLAGS_CC))
6792 			return;
6793 	} else {
6794 		u8 l4_hdr = 0;
6795 		switch (first->protocol) {
6796 		case htons(ETH_P_IP):
6797 			vlan_macip_lens |= skb_network_header_len(skb);
6798 			type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
6799 			l4_hdr = ip_hdr(skb)->protocol;
6800 			break;
6801 		case htons(ETH_P_IPV6):
6802 			vlan_macip_lens |= skb_network_header_len(skb);
6803 			l4_hdr = ipv6_hdr(skb)->nexthdr;
6804 			break;
6805 		default:
6806 			if (unlikely(net_ratelimit())) {
6807 				dev_warn(tx_ring->dev,
6808 				 "partial checksum but proto=%x!\n",
6809 				 first->protocol);
6810 			}
6811 			break;
6812 		}
6813 
6814 		switch (l4_hdr) {
6815 		case IPPROTO_TCP:
6816 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_TCP;
6817 			mss_l4len_idx = tcp_hdrlen(skb) <<
6818 					IXGBE_ADVTXD_L4LEN_SHIFT;
6819 			break;
6820 		case IPPROTO_SCTP:
6821 			type_tucmd |= IXGBE_ADVTXD_TUCMD_L4T_SCTP;
6822 			mss_l4len_idx = sizeof(struct sctphdr) <<
6823 					IXGBE_ADVTXD_L4LEN_SHIFT;
6824 			break;
6825 		case IPPROTO_UDP:
6826 			mss_l4len_idx = sizeof(struct udphdr) <<
6827 					IXGBE_ADVTXD_L4LEN_SHIFT;
6828 			break;
6829 		default:
6830 			if (unlikely(net_ratelimit())) {
6831 				dev_warn(tx_ring->dev,
6832 				 "partial checksum but l4 proto=%x!\n",
6833 				 l4_hdr);
6834 			}
6835 			break;
6836 		}
6837 
6838 		/* update TX checksum flag */
6839 		first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
6840 	}
6841 
6842 	/* vlan_macip_lens: MACLEN, VLAN tag */
6843 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
6844 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
6845 
6846 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0,
6847 			  type_tucmd, mss_l4len_idx);
6848 }
6849 
6850 #define IXGBE_SET_FLAG(_input, _flag, _result) \
6851 	((_flag <= _result) ? \
6852 	 ((u32)(_input & _flag) * (_result / _flag)) : \
6853 	 ((u32)(_input & _flag) / (_flag / _result)))
6854 
6855 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
6856 {
6857 	/* set type for advanced descriptor with frame checksum insertion */
6858 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
6859 		       IXGBE_ADVTXD_DCMD_DEXT |
6860 		       IXGBE_ADVTXD_DCMD_IFCS;
6861 
6862 	/* set HW vlan bit if vlan is present */
6863 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
6864 				   IXGBE_ADVTXD_DCMD_VLE);
6865 
6866 	/* set segmentation enable bits for TSO/FSO */
6867 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
6868 				   IXGBE_ADVTXD_DCMD_TSE);
6869 
6870 	/* set timestamp bit if present */
6871 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
6872 				   IXGBE_ADVTXD_MAC_TSTAMP);
6873 
6874 	/* insert frame checksum */
6875 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
6876 
6877 	return cmd_type;
6878 }
6879 
6880 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
6881 				   u32 tx_flags, unsigned int paylen)
6882 {
6883 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
6884 
6885 	/* enable L4 checksum for TSO and TX checksum offload */
6886 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
6887 					IXGBE_TX_FLAGS_CSUM,
6888 					IXGBE_ADVTXD_POPTS_TXSM);
6889 
6890 	/* enble IPv4 checksum for TSO */
6891 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
6892 					IXGBE_TX_FLAGS_IPV4,
6893 					IXGBE_ADVTXD_POPTS_IXSM);
6894 
6895 	/*
6896 	 * Check Context must be set if Tx switch is enabled, which it
6897 	 * always is for case where virtual functions are running
6898 	 */
6899 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
6900 					IXGBE_TX_FLAGS_CC,
6901 					IXGBE_ADVTXD_CC);
6902 
6903 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
6904 }
6905 
6906 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
6907 {
6908 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
6909 
6910 	/* Herbert's original patch had:
6911 	 *  smp_mb__after_netif_stop_queue();
6912 	 * but since that doesn't exist yet, just open code it.
6913 	 */
6914 	smp_mb();
6915 
6916 	/* We need to check again in a case another CPU has just
6917 	 * made room available.
6918 	 */
6919 	if (likely(ixgbe_desc_unused(tx_ring) < size))
6920 		return -EBUSY;
6921 
6922 	/* A reprieve! - use start_queue because it doesn't call schedule */
6923 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
6924 	++tx_ring->tx_stats.restart_queue;
6925 	return 0;
6926 }
6927 
6928 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
6929 {
6930 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
6931 		return 0;
6932 
6933 	return __ixgbe_maybe_stop_tx(tx_ring, size);
6934 }
6935 
6936 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
6937 		       IXGBE_TXD_CMD_RS)
6938 
6939 static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
6940 			 struct ixgbe_tx_buffer *first,
6941 			 const u8 hdr_len)
6942 {
6943 	struct sk_buff *skb = first->skb;
6944 	struct ixgbe_tx_buffer *tx_buffer;
6945 	union ixgbe_adv_tx_desc *tx_desc;
6946 	struct skb_frag_struct *frag;
6947 	dma_addr_t dma;
6948 	unsigned int data_len, size;
6949 	u32 tx_flags = first->tx_flags;
6950 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
6951 	u16 i = tx_ring->next_to_use;
6952 
6953 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
6954 
6955 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
6956 
6957 	size = skb_headlen(skb);
6958 	data_len = skb->data_len;
6959 
6960 #ifdef IXGBE_FCOE
6961 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
6962 		if (data_len < sizeof(struct fcoe_crc_eof)) {
6963 			size -= sizeof(struct fcoe_crc_eof) - data_len;
6964 			data_len = 0;
6965 		} else {
6966 			data_len -= sizeof(struct fcoe_crc_eof);
6967 		}
6968 	}
6969 
6970 #endif
6971 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
6972 
6973 	tx_buffer = first;
6974 
6975 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
6976 		if (dma_mapping_error(tx_ring->dev, dma))
6977 			goto dma_error;
6978 
6979 		/* record length, and DMA address */
6980 		dma_unmap_len_set(tx_buffer, len, size);
6981 		dma_unmap_addr_set(tx_buffer, dma, dma);
6982 
6983 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
6984 
6985 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
6986 			tx_desc->read.cmd_type_len =
6987 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
6988 
6989 			i++;
6990 			tx_desc++;
6991 			if (i == tx_ring->count) {
6992 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6993 				i = 0;
6994 			}
6995 			tx_desc->read.olinfo_status = 0;
6996 
6997 			dma += IXGBE_MAX_DATA_PER_TXD;
6998 			size -= IXGBE_MAX_DATA_PER_TXD;
6999 
7000 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
7001 		}
7002 
7003 		if (likely(!data_len))
7004 			break;
7005 
7006 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
7007 
7008 		i++;
7009 		tx_desc++;
7010 		if (i == tx_ring->count) {
7011 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7012 			i = 0;
7013 		}
7014 		tx_desc->read.olinfo_status = 0;
7015 
7016 #ifdef IXGBE_FCOE
7017 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
7018 #else
7019 		size = skb_frag_size(frag);
7020 #endif
7021 		data_len -= size;
7022 
7023 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
7024 				       DMA_TO_DEVICE);
7025 
7026 		tx_buffer = &tx_ring->tx_buffer_info[i];
7027 	}
7028 
7029 	/* write last descriptor with RS and EOP bits */
7030 	cmd_type |= size | IXGBE_TXD_CMD;
7031 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
7032 
7033 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
7034 
7035 	/* set the timestamp */
7036 	first->time_stamp = jiffies;
7037 
7038 	/*
7039 	 * Force memory writes to complete before letting h/w know there
7040 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
7041 	 * memory model archs, such as IA-64).
7042 	 *
7043 	 * We also need this memory barrier to make certain all of the
7044 	 * status bits have been updated before next_to_watch is written.
7045 	 */
7046 	wmb();
7047 
7048 	/* set next_to_watch value indicating a packet is present */
7049 	first->next_to_watch = tx_desc;
7050 
7051 	i++;
7052 	if (i == tx_ring->count)
7053 		i = 0;
7054 
7055 	tx_ring->next_to_use = i;
7056 
7057 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
7058 
7059 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
7060 		writel(i, tx_ring->tail);
7061 
7062 		/* we need this if more than one processor can write to our tail
7063 		 * at a time, it synchronizes IO on IA64/Altix systems
7064 		 */
7065 		mmiowb();
7066 	}
7067 
7068 	return;
7069 dma_error:
7070 	dev_err(tx_ring->dev, "TX DMA map failed\n");
7071 
7072 	/* clear dma mappings for failed tx_buffer_info map */
7073 	for (;;) {
7074 		tx_buffer = &tx_ring->tx_buffer_info[i];
7075 		ixgbe_unmap_and_free_tx_resource(tx_ring, tx_buffer);
7076 		if (tx_buffer == first)
7077 			break;
7078 		if (i == 0)
7079 			i = tx_ring->count;
7080 		i--;
7081 	}
7082 
7083 	tx_ring->next_to_use = i;
7084 }
7085 
7086 static void ixgbe_atr(struct ixgbe_ring *ring,
7087 		      struct ixgbe_tx_buffer *first)
7088 {
7089 	struct ixgbe_q_vector *q_vector = ring->q_vector;
7090 	union ixgbe_atr_hash_dword input = { .dword = 0 };
7091 	union ixgbe_atr_hash_dword common = { .dword = 0 };
7092 	union {
7093 		unsigned char *network;
7094 		struct iphdr *ipv4;
7095 		struct ipv6hdr *ipv6;
7096 	} hdr;
7097 	struct tcphdr *th;
7098 	__be16 vlan_id;
7099 
7100 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
7101 	if (!q_vector)
7102 		return;
7103 
7104 	/* do nothing if sampling is disabled */
7105 	if (!ring->atr_sample_rate)
7106 		return;
7107 
7108 	ring->atr_count++;
7109 
7110 	/* snag network header to get L4 type and address */
7111 	hdr.network = skb_network_header(first->skb);
7112 
7113 	/* Currently only IPv4/IPv6 with TCP is supported */
7114 	if ((first->protocol != htons(ETH_P_IPV6) ||
7115 	     hdr.ipv6->nexthdr != IPPROTO_TCP) &&
7116 	    (first->protocol != htons(ETH_P_IP) ||
7117 	     hdr.ipv4->protocol != IPPROTO_TCP))
7118 		return;
7119 
7120 	th = tcp_hdr(first->skb);
7121 
7122 	/* skip this packet since it is invalid or the socket is closing */
7123 	if (!th || th->fin)
7124 		return;
7125 
7126 	/* sample on all syn packets or once every atr sample count */
7127 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
7128 		return;
7129 
7130 	/* reset sample count */
7131 	ring->atr_count = 0;
7132 
7133 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
7134 
7135 	/*
7136 	 * src and dst are inverted, think how the receiver sees them
7137 	 *
7138 	 * The input is broken into two sections, a non-compressed section
7139 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
7140 	 * is XORed together and stored in the compressed dword.
7141 	 */
7142 	input.formatted.vlan_id = vlan_id;
7143 
7144 	/*
7145 	 * since src port and flex bytes occupy the same word XOR them together
7146 	 * and write the value to source port portion of compressed dword
7147 	 */
7148 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
7149 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
7150 	else
7151 		common.port.src ^= th->dest ^ first->protocol;
7152 	common.port.dst ^= th->source;
7153 
7154 	if (first->protocol == htons(ETH_P_IP)) {
7155 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
7156 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
7157 	} else {
7158 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
7159 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
7160 			     hdr.ipv6->saddr.s6_addr32[1] ^
7161 			     hdr.ipv6->saddr.s6_addr32[2] ^
7162 			     hdr.ipv6->saddr.s6_addr32[3] ^
7163 			     hdr.ipv6->daddr.s6_addr32[0] ^
7164 			     hdr.ipv6->daddr.s6_addr32[1] ^
7165 			     hdr.ipv6->daddr.s6_addr32[2] ^
7166 			     hdr.ipv6->daddr.s6_addr32[3];
7167 	}
7168 
7169 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
7170 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
7171 					      input, common, ring->queue_index);
7172 }
7173 
7174 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
7175 			      void *accel_priv, select_queue_fallback_t fallback)
7176 {
7177 	struct ixgbe_fwd_adapter *fwd_adapter = accel_priv;
7178 #ifdef IXGBE_FCOE
7179 	struct ixgbe_adapter *adapter;
7180 	struct ixgbe_ring_feature *f;
7181 	int txq;
7182 #endif
7183 
7184 	if (fwd_adapter)
7185 		return skb->queue_mapping + fwd_adapter->tx_base_queue;
7186 
7187 #ifdef IXGBE_FCOE
7188 
7189 	/*
7190 	 * only execute the code below if protocol is FCoE
7191 	 * or FIP and we have FCoE enabled on the adapter
7192 	 */
7193 	switch (vlan_get_protocol(skb)) {
7194 	case htons(ETH_P_FCOE):
7195 	case htons(ETH_P_FIP):
7196 		adapter = netdev_priv(dev);
7197 
7198 		if (adapter->flags & IXGBE_FLAG_FCOE_ENABLED)
7199 			break;
7200 	default:
7201 		return fallback(dev, skb);
7202 	}
7203 
7204 	f = &adapter->ring_feature[RING_F_FCOE];
7205 
7206 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
7207 					   smp_processor_id();
7208 
7209 	while (txq >= f->indices)
7210 		txq -= f->indices;
7211 
7212 	return txq + f->offset;
7213 #else
7214 	return fallback(dev, skb);
7215 #endif
7216 }
7217 
7218 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
7219 			  struct ixgbe_adapter *adapter,
7220 			  struct ixgbe_ring *tx_ring)
7221 {
7222 	struct ixgbe_tx_buffer *first;
7223 	int tso;
7224 	u32 tx_flags = 0;
7225 	unsigned short f;
7226 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
7227 	__be16 protocol = skb->protocol;
7228 	u8 hdr_len = 0;
7229 
7230 	/*
7231 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
7232 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
7233 	 *       + 2 desc gap to keep tail from touching head,
7234 	 *       + 1 desc for context descriptor,
7235 	 * otherwise try next time
7236 	 */
7237 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
7238 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
7239 
7240 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
7241 		tx_ring->tx_stats.tx_busy++;
7242 		return NETDEV_TX_BUSY;
7243 	}
7244 
7245 	/* record the location of the first descriptor for this packet */
7246 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
7247 	first->skb = skb;
7248 	first->bytecount = skb->len;
7249 	first->gso_segs = 1;
7250 
7251 	/* if we have a HW VLAN tag being added default to the HW one */
7252 	if (skb_vlan_tag_present(skb)) {
7253 		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
7254 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7255 	/* else if it is a SW VLAN check the next protocol and store the tag */
7256 	} else if (protocol == htons(ETH_P_8021Q)) {
7257 		struct vlan_hdr *vhdr, _vhdr;
7258 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
7259 		if (!vhdr)
7260 			goto out_drop;
7261 
7262 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
7263 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
7264 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
7265 	}
7266 	protocol = vlan_get_protocol(skb);
7267 
7268 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
7269 	    adapter->ptp_clock &&
7270 	    !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
7271 				   &adapter->state)) {
7272 		skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
7273 		tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
7274 
7275 		/* schedule check for Tx timestamp */
7276 		adapter->ptp_tx_skb = skb_get(skb);
7277 		adapter->ptp_tx_start = jiffies;
7278 		schedule_work(&adapter->ptp_tx_work);
7279 	}
7280 
7281 	skb_tx_timestamp(skb);
7282 
7283 #ifdef CONFIG_PCI_IOV
7284 	/*
7285 	 * Use the l2switch_enable flag - would be false if the DMA
7286 	 * Tx switch had been disabled.
7287 	 */
7288 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7289 		tx_flags |= IXGBE_TX_FLAGS_CC;
7290 
7291 #endif
7292 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
7293 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
7294 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
7295 	     (skb->priority != TC_PRIO_CONTROL))) {
7296 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
7297 		tx_flags |= (skb->priority & 0x7) <<
7298 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
7299 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
7300 			struct vlan_ethhdr *vhdr;
7301 
7302 			if (skb_cow_head(skb, 0))
7303 				goto out_drop;
7304 			vhdr = (struct vlan_ethhdr *)skb->data;
7305 			vhdr->h_vlan_TCI = htons(tx_flags >>
7306 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
7307 		} else {
7308 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
7309 		}
7310 	}
7311 
7312 	/* record initial flags and protocol */
7313 	first->tx_flags = tx_flags;
7314 	first->protocol = protocol;
7315 
7316 #ifdef IXGBE_FCOE
7317 	/* setup tx offload for FCoE */
7318 	if ((protocol == htons(ETH_P_FCOE)) &&
7319 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
7320 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
7321 		if (tso < 0)
7322 			goto out_drop;
7323 
7324 		goto xmit_fcoe;
7325 	}
7326 
7327 #endif /* IXGBE_FCOE */
7328 	tso = ixgbe_tso(tx_ring, first, &hdr_len);
7329 	if (tso < 0)
7330 		goto out_drop;
7331 	else if (!tso)
7332 		ixgbe_tx_csum(tx_ring, first);
7333 
7334 	/* add the ATR filter if ATR is on */
7335 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
7336 		ixgbe_atr(tx_ring, first);
7337 
7338 #ifdef IXGBE_FCOE
7339 xmit_fcoe:
7340 #endif /* IXGBE_FCOE */
7341 	ixgbe_tx_map(tx_ring, first, hdr_len);
7342 
7343 	return NETDEV_TX_OK;
7344 
7345 out_drop:
7346 	dev_kfree_skb_any(first->skb);
7347 	first->skb = NULL;
7348 
7349 	return NETDEV_TX_OK;
7350 }
7351 
7352 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
7353 				      struct net_device *netdev,
7354 				      struct ixgbe_ring *ring)
7355 {
7356 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7357 	struct ixgbe_ring *tx_ring;
7358 
7359 	/*
7360 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
7361 	 * in order to meet this minimum size requirement.
7362 	 */
7363 	if (skb_put_padto(skb, 17))
7364 		return NETDEV_TX_OK;
7365 
7366 	tx_ring = ring ? ring : adapter->tx_ring[skb->queue_mapping];
7367 
7368 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
7369 }
7370 
7371 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
7372 				    struct net_device *netdev)
7373 {
7374 	return __ixgbe_xmit_frame(skb, netdev, NULL);
7375 }
7376 
7377 /**
7378  * ixgbe_set_mac - Change the Ethernet Address of the NIC
7379  * @netdev: network interface device structure
7380  * @p: pointer to an address structure
7381  *
7382  * Returns 0 on success, negative on failure
7383  **/
7384 static int ixgbe_set_mac(struct net_device *netdev, void *p)
7385 {
7386 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7387 	struct ixgbe_hw *hw = &adapter->hw;
7388 	struct sockaddr *addr = p;
7389 	int ret;
7390 
7391 	if (!is_valid_ether_addr(addr->sa_data))
7392 		return -EADDRNOTAVAIL;
7393 
7394 	ixgbe_del_mac_filter(adapter, hw->mac.addr, VMDQ_P(0));
7395 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
7396 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
7397 
7398 	ret = ixgbe_add_mac_filter(adapter, hw->mac.addr, VMDQ_P(0));
7399 	return ret > 0 ? 0 : ret;
7400 }
7401 
7402 static int
7403 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
7404 {
7405 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7406 	struct ixgbe_hw *hw = &adapter->hw;
7407 	u16 value;
7408 	int rc;
7409 
7410 	if (prtad != hw->phy.mdio.prtad)
7411 		return -EINVAL;
7412 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
7413 	if (!rc)
7414 		rc = value;
7415 	return rc;
7416 }
7417 
7418 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
7419 			    u16 addr, u16 value)
7420 {
7421 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7422 	struct ixgbe_hw *hw = &adapter->hw;
7423 
7424 	if (prtad != hw->phy.mdio.prtad)
7425 		return -EINVAL;
7426 	return hw->phy.ops.write_reg(hw, addr, devad, value);
7427 }
7428 
7429 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
7430 {
7431 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7432 
7433 	switch (cmd) {
7434 	case SIOCSHWTSTAMP:
7435 		return ixgbe_ptp_set_ts_config(adapter, req);
7436 	case SIOCGHWTSTAMP:
7437 		return ixgbe_ptp_get_ts_config(adapter, req);
7438 	default:
7439 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
7440 	}
7441 }
7442 
7443 /**
7444  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
7445  * netdev->dev_addrs
7446  * @netdev: network interface device structure
7447  *
7448  * Returns non-zero on failure
7449  **/
7450 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
7451 {
7452 	int err = 0;
7453 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7454 	struct ixgbe_hw *hw = &adapter->hw;
7455 
7456 	if (is_valid_ether_addr(hw->mac.san_addr)) {
7457 		rtnl_lock();
7458 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
7459 		rtnl_unlock();
7460 
7461 		/* update SAN MAC vmdq pool selection */
7462 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
7463 	}
7464 	return err;
7465 }
7466 
7467 /**
7468  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
7469  * netdev->dev_addrs
7470  * @netdev: network interface device structure
7471  *
7472  * Returns non-zero on failure
7473  **/
7474 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
7475 {
7476 	int err = 0;
7477 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7478 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
7479 
7480 	if (is_valid_ether_addr(mac->san_addr)) {
7481 		rtnl_lock();
7482 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
7483 		rtnl_unlock();
7484 	}
7485 	return err;
7486 }
7487 
7488 #ifdef CONFIG_NET_POLL_CONTROLLER
7489 /*
7490  * Polling 'interrupt' - used by things like netconsole to send skbs
7491  * without having to re-enable interrupts. It's not called while
7492  * the interrupt routine is executing.
7493  */
7494 static void ixgbe_netpoll(struct net_device *netdev)
7495 {
7496 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7497 	int i;
7498 
7499 	/* if interface is down do nothing */
7500 	if (test_bit(__IXGBE_DOWN, &adapter->state))
7501 		return;
7502 
7503 	/* loop through and schedule all active queues */
7504 	for (i = 0; i < adapter->num_q_vectors; i++)
7505 		ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
7506 }
7507 
7508 #endif
7509 static struct rtnl_link_stats64 *ixgbe_get_stats64(struct net_device *netdev,
7510 						   struct rtnl_link_stats64 *stats)
7511 {
7512 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7513 	int i;
7514 
7515 	rcu_read_lock();
7516 	for (i = 0; i < adapter->num_rx_queues; i++) {
7517 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->rx_ring[i]);
7518 		u64 bytes, packets;
7519 		unsigned int start;
7520 
7521 		if (ring) {
7522 			do {
7523 				start = u64_stats_fetch_begin_irq(&ring->syncp);
7524 				packets = ring->stats.packets;
7525 				bytes   = ring->stats.bytes;
7526 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
7527 			stats->rx_packets += packets;
7528 			stats->rx_bytes   += bytes;
7529 		}
7530 	}
7531 
7532 	for (i = 0; i < adapter->num_tx_queues; i++) {
7533 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
7534 		u64 bytes, packets;
7535 		unsigned int start;
7536 
7537 		if (ring) {
7538 			do {
7539 				start = u64_stats_fetch_begin_irq(&ring->syncp);
7540 				packets = ring->stats.packets;
7541 				bytes   = ring->stats.bytes;
7542 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
7543 			stats->tx_packets += packets;
7544 			stats->tx_bytes   += bytes;
7545 		}
7546 	}
7547 	rcu_read_unlock();
7548 	/* following stats updated by ixgbe_watchdog_task() */
7549 	stats->multicast	= netdev->stats.multicast;
7550 	stats->rx_errors	= netdev->stats.rx_errors;
7551 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
7552 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
7553 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
7554 	return stats;
7555 }
7556 
7557 #ifdef CONFIG_IXGBE_DCB
7558 /**
7559  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
7560  * @adapter: pointer to ixgbe_adapter
7561  * @tc: number of traffic classes currently enabled
7562  *
7563  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
7564  * 802.1Q priority maps to a packet buffer that exists.
7565  */
7566 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
7567 {
7568 	struct ixgbe_hw *hw = &adapter->hw;
7569 	u32 reg, rsave;
7570 	int i;
7571 
7572 	/* 82598 have a static priority to TC mapping that can not
7573 	 * be changed so no validation is needed.
7574 	 */
7575 	if (hw->mac.type == ixgbe_mac_82598EB)
7576 		return;
7577 
7578 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
7579 	rsave = reg;
7580 
7581 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
7582 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
7583 
7584 		/* If up2tc is out of bounds default to zero */
7585 		if (up2tc > tc)
7586 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
7587 	}
7588 
7589 	if (reg != rsave)
7590 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
7591 
7592 	return;
7593 }
7594 
7595 /**
7596  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
7597  * @adapter: Pointer to adapter struct
7598  *
7599  * Populate the netdev user priority to tc map
7600  */
7601 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
7602 {
7603 	struct net_device *dev = adapter->netdev;
7604 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
7605 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
7606 	u8 prio;
7607 
7608 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
7609 		u8 tc = 0;
7610 
7611 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
7612 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
7613 		else if (ets)
7614 			tc = ets->prio_tc[prio];
7615 
7616 		netdev_set_prio_tc_map(dev, prio, tc);
7617 	}
7618 }
7619 
7620 #endif /* CONFIG_IXGBE_DCB */
7621 /**
7622  * ixgbe_setup_tc - configure net_device for multiple traffic classes
7623  *
7624  * @netdev: net device to configure
7625  * @tc: number of traffic classes to enable
7626  */
7627 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
7628 {
7629 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7630 	struct ixgbe_hw *hw = &adapter->hw;
7631 	bool pools;
7632 
7633 	/* Hardware supports up to 8 traffic classes */
7634 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs ||
7635 	    (hw->mac.type == ixgbe_mac_82598EB &&
7636 	     tc < MAX_TRAFFIC_CLASS))
7637 		return -EINVAL;
7638 
7639 	pools = (find_first_zero_bit(&adapter->fwd_bitmask, 32) > 1);
7640 	if (tc && pools && adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS)
7641 		return -EBUSY;
7642 
7643 	/* Hardware has to reinitialize queues and interrupts to
7644 	 * match packet buffer alignment. Unfortunately, the
7645 	 * hardware is not flexible enough to do this dynamically.
7646 	 */
7647 	if (netif_running(dev))
7648 		ixgbe_close(dev);
7649 	ixgbe_clear_interrupt_scheme(adapter);
7650 
7651 #ifdef CONFIG_IXGBE_DCB
7652 	if (tc) {
7653 		netdev_set_num_tc(dev, tc);
7654 		ixgbe_set_prio_tc_map(adapter);
7655 
7656 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
7657 
7658 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
7659 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
7660 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
7661 		}
7662 	} else {
7663 		netdev_reset_tc(dev);
7664 
7665 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
7666 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
7667 
7668 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
7669 
7670 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
7671 		adapter->dcb_cfg.pfc_mode_enable = false;
7672 	}
7673 
7674 	ixgbe_validate_rtr(adapter, tc);
7675 
7676 #endif /* CONFIG_IXGBE_DCB */
7677 	ixgbe_init_interrupt_scheme(adapter);
7678 
7679 	if (netif_running(dev))
7680 		return ixgbe_open(dev);
7681 
7682 	return 0;
7683 }
7684 
7685 #ifdef CONFIG_PCI_IOV
7686 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
7687 {
7688 	struct net_device *netdev = adapter->netdev;
7689 
7690 	rtnl_lock();
7691 	ixgbe_setup_tc(netdev, netdev_get_num_tc(netdev));
7692 	rtnl_unlock();
7693 }
7694 
7695 #endif
7696 void ixgbe_do_reset(struct net_device *netdev)
7697 {
7698 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7699 
7700 	if (netif_running(netdev))
7701 		ixgbe_reinit_locked(adapter);
7702 	else
7703 		ixgbe_reset(adapter);
7704 }
7705 
7706 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
7707 					    netdev_features_t features)
7708 {
7709 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7710 
7711 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
7712 	if (!(features & NETIF_F_RXCSUM))
7713 		features &= ~NETIF_F_LRO;
7714 
7715 	/* Turn off LRO if not RSC capable */
7716 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
7717 		features &= ~NETIF_F_LRO;
7718 
7719 	return features;
7720 }
7721 
7722 static int ixgbe_set_features(struct net_device *netdev,
7723 			      netdev_features_t features)
7724 {
7725 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
7726 	netdev_features_t changed = netdev->features ^ features;
7727 	bool need_reset = false;
7728 
7729 	/* Make sure RSC matches LRO, reset if change */
7730 	if (!(features & NETIF_F_LRO)) {
7731 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
7732 			need_reset = true;
7733 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
7734 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
7735 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
7736 		if (adapter->rx_itr_setting == 1 ||
7737 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
7738 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
7739 			need_reset = true;
7740 		} else if ((changed ^ features) & NETIF_F_LRO) {
7741 			e_info(probe, "rx-usecs set too low, "
7742 			       "disabling RSC\n");
7743 		}
7744 	}
7745 
7746 	/*
7747 	 * Check if Flow Director n-tuple support was enabled or disabled.  If
7748 	 * the state changed, we need to reset.
7749 	 */
7750 	switch (features & NETIF_F_NTUPLE) {
7751 	case NETIF_F_NTUPLE:
7752 		/* turn off ATR, enable perfect filters and reset */
7753 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
7754 			need_reset = true;
7755 
7756 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
7757 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
7758 		break;
7759 	default:
7760 		/* turn off perfect filters, enable ATR and reset */
7761 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
7762 			need_reset = true;
7763 
7764 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
7765 
7766 		/* We cannot enable ATR if SR-IOV is enabled */
7767 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
7768 			break;
7769 
7770 		/* We cannot enable ATR if we have 2 or more traffic classes */
7771 		if (netdev_get_num_tc(netdev) > 1)
7772 			break;
7773 
7774 		/* We cannot enable ATR if RSS is disabled */
7775 		if (adapter->ring_feature[RING_F_RSS].limit <= 1)
7776 			break;
7777 
7778 		/* A sample rate of 0 indicates ATR disabled */
7779 		if (!adapter->atr_sample_rate)
7780 			break;
7781 
7782 		adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
7783 		break;
7784 	}
7785 
7786 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
7787 		ixgbe_vlan_strip_enable(adapter);
7788 	else
7789 		ixgbe_vlan_strip_disable(adapter);
7790 
7791 	if (changed & NETIF_F_RXALL)
7792 		need_reset = true;
7793 
7794 	netdev->features = features;
7795 	if (need_reset)
7796 		ixgbe_do_reset(netdev);
7797 
7798 	return 0;
7799 }
7800 
7801 /**
7802  * ixgbe_add_vxlan_port - Get notifications about VXLAN ports that come up
7803  * @dev: The port's netdev
7804  * @sa_family: Socket Family that VXLAN is notifiying us about
7805  * @port: New UDP port number that VXLAN started listening to
7806  **/
7807 static void ixgbe_add_vxlan_port(struct net_device *dev, sa_family_t sa_family,
7808 				 __be16 port)
7809 {
7810 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7811 	struct ixgbe_hw *hw = &adapter->hw;
7812 	u16 new_port = ntohs(port);
7813 
7814 	if (sa_family == AF_INET6)
7815 		return;
7816 
7817 	if (adapter->vxlan_port == new_port) {
7818 		netdev_info(dev, "Port %d already offloaded\n", new_port);
7819 		return;
7820 	}
7821 
7822 	if (adapter->vxlan_port) {
7823 		netdev_info(dev,
7824 			    "Hit Max num of UDP ports, not adding port %d\n",
7825 			    new_port);
7826 		return;
7827 	}
7828 
7829 	adapter->vxlan_port = new_port;
7830 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, new_port);
7831 }
7832 
7833 /**
7834  * ixgbe_del_vxlan_port - Get notifications about VXLAN ports that go away
7835  * @dev: The port's netdev
7836  * @sa_family: Socket Family that VXLAN is notifying us about
7837  * @port: UDP port number that VXLAN stopped listening to
7838  **/
7839 static void ixgbe_del_vxlan_port(struct net_device *dev, sa_family_t sa_family,
7840 				 __be16 port)
7841 {
7842 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7843 	struct ixgbe_hw *hw = &adapter->hw;
7844 	u16 new_port = ntohs(port);
7845 
7846 	if (sa_family == AF_INET6)
7847 		return;
7848 
7849 	if (adapter->vxlan_port != new_port) {
7850 		netdev_info(dev, "Port %d was not found, not deleting\n",
7851 			    new_port);
7852 		return;
7853 	}
7854 
7855 	adapter->vxlan_port = 0;
7856 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, 0);
7857 }
7858 
7859 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
7860 			     struct net_device *dev,
7861 			     const unsigned char *addr, u16 vid,
7862 			     u16 flags)
7863 {
7864 	/* guarantee we can provide a unique filter for the unicast address */
7865 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
7866 		if (IXGBE_MAX_PF_MACVLANS <= netdev_uc_count(dev))
7867 			return -ENOMEM;
7868 	}
7869 
7870 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
7871 }
7872 
7873 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
7874 				    struct nlmsghdr *nlh, u16 flags)
7875 {
7876 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7877 	struct nlattr *attr, *br_spec;
7878 	int rem;
7879 
7880 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
7881 		return -EOPNOTSUPP;
7882 
7883 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
7884 	if (!br_spec)
7885 		return -EINVAL;
7886 
7887 	nla_for_each_nested(attr, br_spec, rem) {
7888 		__u16 mode;
7889 		u32 reg = 0;
7890 
7891 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
7892 			continue;
7893 
7894 		if (nla_len(attr) < sizeof(mode))
7895 			return -EINVAL;
7896 
7897 		mode = nla_get_u16(attr);
7898 		if (mode == BRIDGE_MODE_VEPA) {
7899 			reg = 0;
7900 			adapter->flags2 &= ~IXGBE_FLAG2_BRIDGE_MODE_VEB;
7901 		} else if (mode == BRIDGE_MODE_VEB) {
7902 			reg = IXGBE_PFDTXGSWC_VT_LBEN;
7903 			adapter->flags2 |= IXGBE_FLAG2_BRIDGE_MODE_VEB;
7904 		} else
7905 			return -EINVAL;
7906 
7907 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, reg);
7908 
7909 		e_info(drv, "enabling bridge mode: %s\n",
7910 			mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
7911 	}
7912 
7913 	return 0;
7914 }
7915 
7916 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
7917 				    struct net_device *dev,
7918 				    u32 filter_mask)
7919 {
7920 	struct ixgbe_adapter *adapter = netdev_priv(dev);
7921 	u16 mode;
7922 
7923 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
7924 		return 0;
7925 
7926 	if (adapter->flags2 & IXGBE_FLAG2_BRIDGE_MODE_VEB)
7927 		mode = BRIDGE_MODE_VEB;
7928 	else
7929 		mode = BRIDGE_MODE_VEPA;
7930 
7931 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev, mode, 0, 0);
7932 }
7933 
7934 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
7935 {
7936 	struct ixgbe_fwd_adapter *fwd_adapter = NULL;
7937 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
7938 	int used_pools = adapter->num_vfs + adapter->num_rx_pools;
7939 	unsigned int limit;
7940 	int pool, err;
7941 
7942 	/* Hardware has a limited number of available pools. Each VF, and the
7943 	 * PF require a pool. Check to ensure we don't attempt to use more
7944 	 * then the available number of pools.
7945 	 */
7946 	if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
7947 		return ERR_PTR(-EINVAL);
7948 
7949 #ifdef CONFIG_RPS
7950 	if (vdev->num_rx_queues != vdev->num_tx_queues) {
7951 		netdev_info(pdev, "%s: Only supports a single queue count for TX and RX\n",
7952 			    vdev->name);
7953 		return ERR_PTR(-EINVAL);
7954 	}
7955 #endif
7956 	/* Check for hardware restriction on number of rx/tx queues */
7957 	if (vdev->num_tx_queues > IXGBE_MAX_L2A_QUEUES ||
7958 	    vdev->num_tx_queues == IXGBE_BAD_L2A_QUEUE) {
7959 		netdev_info(pdev,
7960 			    "%s: Supports RX/TX Queue counts 1,2, and 4\n",
7961 			    pdev->name);
7962 		return ERR_PTR(-EINVAL);
7963 	}
7964 
7965 	if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
7966 	      adapter->num_rx_pools > IXGBE_MAX_DCBMACVLANS - 1) ||
7967 	    (adapter->num_rx_pools > IXGBE_MAX_MACVLANS))
7968 		return ERR_PTR(-EBUSY);
7969 
7970 	fwd_adapter = kcalloc(1, sizeof(struct ixgbe_fwd_adapter), GFP_KERNEL);
7971 	if (!fwd_adapter)
7972 		return ERR_PTR(-ENOMEM);
7973 
7974 	pool = find_first_zero_bit(&adapter->fwd_bitmask, 32);
7975 	adapter->num_rx_pools++;
7976 	set_bit(pool, &adapter->fwd_bitmask);
7977 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
7978 
7979 	/* Enable VMDq flag so device will be set in VM mode */
7980 	adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED | IXGBE_FLAG_SRIOV_ENABLED;
7981 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
7982 	adapter->ring_feature[RING_F_RSS].limit = vdev->num_tx_queues;
7983 
7984 	/* Force reinit of ring allocation with VMDQ enabled */
7985 	err = ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
7986 	if (err)
7987 		goto fwd_add_err;
7988 	fwd_adapter->pool = pool;
7989 	fwd_adapter->real_adapter = adapter;
7990 	err = ixgbe_fwd_ring_up(vdev, fwd_adapter);
7991 	if (err)
7992 		goto fwd_add_err;
7993 	netif_tx_start_all_queues(vdev);
7994 	return fwd_adapter;
7995 fwd_add_err:
7996 	/* unwind counter and free adapter struct */
7997 	netdev_info(pdev,
7998 		    "%s: dfwd hardware acceleration failed\n", vdev->name);
7999 	clear_bit(pool, &adapter->fwd_bitmask);
8000 	adapter->num_rx_pools--;
8001 	kfree(fwd_adapter);
8002 	return ERR_PTR(err);
8003 }
8004 
8005 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
8006 {
8007 	struct ixgbe_fwd_adapter *fwd_adapter = priv;
8008 	struct ixgbe_adapter *adapter = fwd_adapter->real_adapter;
8009 	unsigned int limit;
8010 
8011 	clear_bit(fwd_adapter->pool, &adapter->fwd_bitmask);
8012 	adapter->num_rx_pools--;
8013 
8014 	limit = find_last_bit(&adapter->fwd_bitmask, 32);
8015 	adapter->ring_feature[RING_F_VMDQ].limit = limit + 1;
8016 	ixgbe_fwd_ring_down(fwd_adapter->netdev, fwd_adapter);
8017 	ixgbe_setup_tc(pdev, netdev_get_num_tc(pdev));
8018 	netdev_dbg(pdev, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
8019 		   fwd_adapter->pool, adapter->num_rx_pools,
8020 		   fwd_adapter->rx_base_queue,
8021 		   fwd_adapter->rx_base_queue + adapter->num_rx_queues_per_pool,
8022 		   adapter->fwd_bitmask);
8023 	kfree(fwd_adapter);
8024 }
8025 
8026 static const struct net_device_ops ixgbe_netdev_ops = {
8027 	.ndo_open		= ixgbe_open,
8028 	.ndo_stop		= ixgbe_close,
8029 	.ndo_start_xmit		= ixgbe_xmit_frame,
8030 	.ndo_select_queue	= ixgbe_select_queue,
8031 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
8032 	.ndo_validate_addr	= eth_validate_addr,
8033 	.ndo_set_mac_address	= ixgbe_set_mac,
8034 	.ndo_change_mtu		= ixgbe_change_mtu,
8035 	.ndo_tx_timeout		= ixgbe_tx_timeout,
8036 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
8037 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
8038 	.ndo_do_ioctl		= ixgbe_ioctl,
8039 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
8040 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
8041 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
8042 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
8043 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
8044 	.ndo_get_stats64	= ixgbe_get_stats64,
8045 #ifdef CONFIG_IXGBE_DCB
8046 	.ndo_setup_tc		= ixgbe_setup_tc,
8047 #endif
8048 #ifdef CONFIG_NET_POLL_CONTROLLER
8049 	.ndo_poll_controller	= ixgbe_netpoll,
8050 #endif
8051 #ifdef CONFIG_NET_RX_BUSY_POLL
8052 	.ndo_busy_poll		= ixgbe_low_latency_recv,
8053 #endif
8054 #ifdef IXGBE_FCOE
8055 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
8056 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
8057 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
8058 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
8059 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
8060 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
8061 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
8062 #endif /* IXGBE_FCOE */
8063 	.ndo_set_features = ixgbe_set_features,
8064 	.ndo_fix_features = ixgbe_fix_features,
8065 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
8066 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
8067 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
8068 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
8069 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
8070 	.ndo_add_vxlan_port	= ixgbe_add_vxlan_port,
8071 	.ndo_del_vxlan_port	= ixgbe_del_vxlan_port,
8072 };
8073 
8074 /**
8075  * ixgbe_enumerate_functions - Get the number of ports this device has
8076  * @adapter: adapter structure
8077  *
8078  * This function enumerates the phsyical functions co-located on a single slot,
8079  * in order to determine how many ports a device has. This is most useful in
8080  * determining the required GT/s of PCIe bandwidth necessary for optimal
8081  * performance.
8082  **/
8083 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
8084 {
8085 	struct pci_dev *entry, *pdev = adapter->pdev;
8086 	int physfns = 0;
8087 
8088 	/* Some cards can not use the generic count PCIe functions method,
8089 	 * because they are behind a parent switch, so we hardcode these with
8090 	 * the correct number of functions.
8091 	 */
8092 	if (ixgbe_pcie_from_parent(&adapter->hw))
8093 		physfns = 4;
8094 
8095 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
8096 		/* don't count virtual functions */
8097 		if (entry->is_virtfn)
8098 			continue;
8099 
8100 		/* When the devices on the bus don't all match our device ID,
8101 		 * we can't reliably determine the correct number of
8102 		 * functions. This can occur if a function has been direct
8103 		 * attached to a virtual machine using VT-d, for example. In
8104 		 * this case, simply return -1 to indicate this.
8105 		 */
8106 		if ((entry->vendor != pdev->vendor) ||
8107 		    (entry->device != pdev->device))
8108 			return -1;
8109 
8110 		physfns++;
8111 	}
8112 
8113 	return physfns;
8114 }
8115 
8116 /**
8117  * ixgbe_wol_supported - Check whether device supports WoL
8118  * @hw: hw specific details
8119  * @device_id: the device ID
8120  * @subdev_id: the subsystem device ID
8121  *
8122  * This function is used by probe and ethtool to determine
8123  * which devices have WoL support
8124  *
8125  **/
8126 int ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
8127 			u16 subdevice_id)
8128 {
8129 	struct ixgbe_hw *hw = &adapter->hw;
8130 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
8131 	int is_wol_supported = 0;
8132 
8133 	switch (device_id) {
8134 	case IXGBE_DEV_ID_82599_SFP:
8135 		/* Only these subdevices could supports WOL */
8136 		switch (subdevice_id) {
8137 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
8138 		case IXGBE_SUBDEV_ID_82599_560FLR:
8139 			/* only support first port */
8140 			if (hw->bus.func != 0)
8141 				break;
8142 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
8143 		case IXGBE_SUBDEV_ID_82599_SFP:
8144 		case IXGBE_SUBDEV_ID_82599_RNDC:
8145 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
8146 		case IXGBE_SUBDEV_ID_82599_LOM_SFP:
8147 			is_wol_supported = 1;
8148 			break;
8149 		}
8150 		break;
8151 	case IXGBE_DEV_ID_82599EN_SFP:
8152 		/* Only this subdevice supports WOL */
8153 		switch (subdevice_id) {
8154 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
8155 			is_wol_supported = 1;
8156 			break;
8157 		}
8158 		break;
8159 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
8160 		/* All except this subdevice support WOL */
8161 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
8162 			is_wol_supported = 1;
8163 		break;
8164 	case IXGBE_DEV_ID_82599_KX4:
8165 		is_wol_supported = 1;
8166 		break;
8167 	case IXGBE_DEV_ID_X540T:
8168 	case IXGBE_DEV_ID_X540T1:
8169 		/* check eeprom to see if enabled wol */
8170 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
8171 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
8172 		     (hw->bus.func == 0))) {
8173 			is_wol_supported = 1;
8174 		}
8175 		break;
8176 	}
8177 
8178 	return is_wol_supported;
8179 }
8180 
8181 /**
8182  * ixgbe_get_platform_mac_addr - Look up MAC address in Open Firmware / IDPROM
8183  * @adapter: Pointer to adapter struct
8184  */
8185 static void ixgbe_get_platform_mac_addr(struct ixgbe_adapter *adapter)
8186 {
8187 #ifdef CONFIG_OF
8188 	struct device_node *dp = pci_device_to_OF_node(adapter->pdev);
8189 	struct ixgbe_hw *hw = &adapter->hw;
8190 	const unsigned char *addr;
8191 
8192 	addr = of_get_mac_address(dp);
8193 	if (addr) {
8194 		ether_addr_copy(hw->mac.perm_addr, addr);
8195 		return;
8196 	}
8197 #endif /* CONFIG_OF */
8198 
8199 #ifdef CONFIG_SPARC
8200 	ether_addr_copy(hw->mac.perm_addr, idprom->id_ethaddr);
8201 #endif /* CONFIG_SPARC */
8202 }
8203 
8204 /**
8205  * ixgbe_probe - Device Initialization Routine
8206  * @pdev: PCI device information struct
8207  * @ent: entry in ixgbe_pci_tbl
8208  *
8209  * Returns 0 on success, negative on failure
8210  *
8211  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
8212  * The OS initialization, configuring of the adapter private structure,
8213  * and a hardware reset occur.
8214  **/
8215 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
8216 {
8217 	struct net_device *netdev;
8218 	struct ixgbe_adapter *adapter = NULL;
8219 	struct ixgbe_hw *hw;
8220 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
8221 	int i, err, pci_using_dac, expected_gts;
8222 	unsigned int indices = MAX_TX_QUEUES;
8223 	u8 part_str[IXGBE_PBANUM_LENGTH];
8224 	bool disable_dev = false;
8225 #ifdef IXGBE_FCOE
8226 	u16 device_caps;
8227 #endif
8228 	u32 eec;
8229 
8230 	/* Catch broken hardware that put the wrong VF device ID in
8231 	 * the PCIe SR-IOV capability.
8232 	 */
8233 	if (pdev->is_virtfn) {
8234 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
8235 		     pci_name(pdev), pdev->vendor, pdev->device);
8236 		return -EINVAL;
8237 	}
8238 
8239 	err = pci_enable_device_mem(pdev);
8240 	if (err)
8241 		return err;
8242 
8243 	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
8244 		pci_using_dac = 1;
8245 	} else {
8246 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
8247 		if (err) {
8248 			dev_err(&pdev->dev,
8249 				"No usable DMA configuration, aborting\n");
8250 			goto err_dma;
8251 		}
8252 		pci_using_dac = 0;
8253 	}
8254 
8255 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
8256 					   IORESOURCE_MEM), ixgbe_driver_name);
8257 	if (err) {
8258 		dev_err(&pdev->dev,
8259 			"pci_request_selected_regions failed 0x%x\n", err);
8260 		goto err_pci_reg;
8261 	}
8262 
8263 	pci_enable_pcie_error_reporting(pdev);
8264 
8265 	pci_set_master(pdev);
8266 	pci_save_state(pdev);
8267 
8268 	if (ii->mac == ixgbe_mac_82598EB) {
8269 #ifdef CONFIG_IXGBE_DCB
8270 		/* 8 TC w/ 4 queues per TC */
8271 		indices = 4 * MAX_TRAFFIC_CLASS;
8272 #else
8273 		indices = IXGBE_MAX_RSS_INDICES;
8274 #endif
8275 	}
8276 
8277 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
8278 	if (!netdev) {
8279 		err = -ENOMEM;
8280 		goto err_alloc_etherdev;
8281 	}
8282 
8283 	SET_NETDEV_DEV(netdev, &pdev->dev);
8284 
8285 	adapter = netdev_priv(netdev);
8286 
8287 	adapter->netdev = netdev;
8288 	adapter->pdev = pdev;
8289 	hw = &adapter->hw;
8290 	hw->back = adapter;
8291 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
8292 
8293 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
8294 			      pci_resource_len(pdev, 0));
8295 	adapter->io_addr = hw->hw_addr;
8296 	if (!hw->hw_addr) {
8297 		err = -EIO;
8298 		goto err_ioremap;
8299 	}
8300 
8301 	netdev->netdev_ops = &ixgbe_netdev_ops;
8302 	ixgbe_set_ethtool_ops(netdev);
8303 	netdev->watchdog_timeo = 5 * HZ;
8304 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
8305 
8306 	/* Setup hw api */
8307 	memcpy(&hw->mac.ops, ii->mac_ops, sizeof(hw->mac.ops));
8308 	hw->mac.type  = ii->mac;
8309 
8310 	/* EEPROM */
8311 	memcpy(&hw->eeprom.ops, ii->eeprom_ops, sizeof(hw->eeprom.ops));
8312 	eec = IXGBE_READ_REG(hw, IXGBE_EEC);
8313 	if (ixgbe_removed(hw->hw_addr)) {
8314 		err = -EIO;
8315 		goto err_ioremap;
8316 	}
8317 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
8318 	if (!(eec & (1 << 8)))
8319 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
8320 
8321 	/* PHY */
8322 	memcpy(&hw->phy.ops, ii->phy_ops, sizeof(hw->phy.ops));
8323 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
8324 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
8325 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
8326 	hw->phy.mdio.mmds = 0;
8327 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
8328 	hw->phy.mdio.dev = netdev;
8329 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
8330 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
8331 
8332 	ii->get_invariants(hw);
8333 
8334 	/* setup the private structure */
8335 	err = ixgbe_sw_init(adapter);
8336 	if (err)
8337 		goto err_sw_init;
8338 
8339 	/* Make it possible the adapter to be woken up via WOL */
8340 	switch (adapter->hw.mac.type) {
8341 	case ixgbe_mac_82599EB:
8342 	case ixgbe_mac_X540:
8343 	case ixgbe_mac_X550:
8344 	case ixgbe_mac_X550EM_x:
8345 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
8346 		break;
8347 	default:
8348 		break;
8349 	}
8350 
8351 	/*
8352 	 * If there is a fan on this device and it has failed log the
8353 	 * failure.
8354 	 */
8355 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
8356 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
8357 		if (esdp & IXGBE_ESDP_SDP1)
8358 			e_crit(probe, "Fan has stopped, replace the adapter\n");
8359 	}
8360 
8361 	if (allow_unsupported_sfp)
8362 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
8363 
8364 	/* reset_hw fills in the perm_addr as well */
8365 	hw->phy.reset_if_overtemp = true;
8366 	err = hw->mac.ops.reset_hw(hw);
8367 	hw->phy.reset_if_overtemp = false;
8368 	if (err == IXGBE_ERR_SFP_NOT_PRESENT &&
8369 	    hw->mac.type == ixgbe_mac_82598EB) {
8370 		err = 0;
8371 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
8372 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
8373 		e_dev_err("Reload the driver after installing a supported module.\n");
8374 		goto err_sw_init;
8375 	} else if (err) {
8376 		e_dev_err("HW Init failed: %d\n", err);
8377 		goto err_sw_init;
8378 	}
8379 
8380 #ifdef CONFIG_PCI_IOV
8381 	/* SR-IOV not supported on the 82598 */
8382 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
8383 		goto skip_sriov;
8384 	/* Mailbox */
8385 	ixgbe_init_mbx_params_pf(hw);
8386 	memcpy(&hw->mbx.ops, ii->mbx_ops, sizeof(hw->mbx.ops));
8387 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
8388 	ixgbe_enable_sriov(adapter);
8389 skip_sriov:
8390 
8391 #endif
8392 	netdev->features = NETIF_F_SG |
8393 			   NETIF_F_IP_CSUM |
8394 			   NETIF_F_IPV6_CSUM |
8395 			   NETIF_F_HW_VLAN_CTAG_TX |
8396 			   NETIF_F_HW_VLAN_CTAG_RX |
8397 			   NETIF_F_HW_VLAN_CTAG_FILTER |
8398 			   NETIF_F_TSO |
8399 			   NETIF_F_TSO6 |
8400 			   NETIF_F_RXHASH |
8401 			   NETIF_F_RXCSUM;
8402 
8403 	netdev->hw_features = netdev->features | NETIF_F_HW_L2FW_DOFFLOAD;
8404 
8405 	switch (adapter->hw.mac.type) {
8406 	case ixgbe_mac_82599EB:
8407 	case ixgbe_mac_X540:
8408 	case ixgbe_mac_X550:
8409 	case ixgbe_mac_X550EM_x:
8410 		netdev->features |= NETIF_F_SCTP_CSUM;
8411 		netdev->hw_features |= NETIF_F_SCTP_CSUM |
8412 				       NETIF_F_NTUPLE;
8413 		break;
8414 	default:
8415 		break;
8416 	}
8417 
8418 	netdev->hw_features |= NETIF_F_RXALL;
8419 
8420 	netdev->vlan_features |= NETIF_F_TSO;
8421 	netdev->vlan_features |= NETIF_F_TSO6;
8422 	netdev->vlan_features |= NETIF_F_IP_CSUM;
8423 	netdev->vlan_features |= NETIF_F_IPV6_CSUM;
8424 	netdev->vlan_features |= NETIF_F_SG;
8425 
8426 	netdev->priv_flags |= IFF_UNICAST_FLT;
8427 	netdev->priv_flags |= IFF_SUPP_NOFCS;
8428 
8429 	switch (adapter->hw.mac.type) {
8430 	case ixgbe_mac_X550:
8431 	case ixgbe_mac_X550EM_x:
8432 		netdev->hw_enc_features |= NETIF_F_RXCSUM;
8433 		break;
8434 	default:
8435 		break;
8436 	}
8437 
8438 #ifdef CONFIG_IXGBE_DCB
8439 	netdev->dcbnl_ops = &dcbnl_ops;
8440 #endif
8441 
8442 #ifdef IXGBE_FCOE
8443 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
8444 		unsigned int fcoe_l;
8445 
8446 		if (hw->mac.ops.get_device_caps) {
8447 			hw->mac.ops.get_device_caps(hw, &device_caps);
8448 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
8449 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
8450 		}
8451 
8452 
8453 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
8454 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
8455 
8456 		netdev->features |= NETIF_F_FSO |
8457 				    NETIF_F_FCOE_CRC;
8458 
8459 		netdev->vlan_features |= NETIF_F_FSO |
8460 					 NETIF_F_FCOE_CRC |
8461 					 NETIF_F_FCOE_MTU;
8462 	}
8463 #endif /* IXGBE_FCOE */
8464 	if (pci_using_dac) {
8465 		netdev->features |= NETIF_F_HIGHDMA;
8466 		netdev->vlan_features |= NETIF_F_HIGHDMA;
8467 	}
8468 
8469 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
8470 		netdev->hw_features |= NETIF_F_LRO;
8471 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
8472 		netdev->features |= NETIF_F_LRO;
8473 
8474 	/* make sure the EEPROM is good */
8475 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
8476 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
8477 		err = -EIO;
8478 		goto err_sw_init;
8479 	}
8480 
8481 	ixgbe_get_platform_mac_addr(adapter);
8482 
8483 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
8484 
8485 	if (!is_valid_ether_addr(netdev->dev_addr)) {
8486 		e_dev_err("invalid MAC address\n");
8487 		err = -EIO;
8488 		goto err_sw_init;
8489 	}
8490 
8491 	ixgbe_mac_set_default_filter(adapter, hw->mac.perm_addr);
8492 
8493 	setup_timer(&adapter->service_timer, &ixgbe_service_timer,
8494 		    (unsigned long) adapter);
8495 
8496 	if (ixgbe_removed(hw->hw_addr)) {
8497 		err = -EIO;
8498 		goto err_sw_init;
8499 	}
8500 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
8501 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
8502 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
8503 
8504 	err = ixgbe_init_interrupt_scheme(adapter);
8505 	if (err)
8506 		goto err_sw_init;
8507 
8508 	/* WOL not supported for all devices */
8509 	adapter->wol = 0;
8510 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
8511 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
8512 						pdev->subsystem_device);
8513 	if (hw->wol_enabled)
8514 		adapter->wol = IXGBE_WUFC_MAG;
8515 
8516 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
8517 
8518 	/* save off EEPROM version number */
8519 	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
8520 	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
8521 
8522 	/* pick up the PCI bus settings for reporting later */
8523 	hw->mac.ops.get_bus_info(hw);
8524 	if (ixgbe_pcie_from_parent(hw))
8525 		ixgbe_get_parent_bus_info(adapter);
8526 
8527 	/* calculate the expected PCIe bandwidth required for optimal
8528 	 * performance. Note that some older parts will never have enough
8529 	 * bandwidth due to being older generation PCIe parts. We clamp these
8530 	 * parts to ensure no warning is displayed if it can't be fixed.
8531 	 */
8532 	switch (hw->mac.type) {
8533 	case ixgbe_mac_82598EB:
8534 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
8535 		break;
8536 	default:
8537 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
8538 		break;
8539 	}
8540 
8541 	/* don't check link if we failed to enumerate functions */
8542 	if (expected_gts > 0)
8543 		ixgbe_check_minimum_link(adapter, expected_gts);
8544 
8545 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
8546 	if (err)
8547 		strlcpy(part_str, "Unknown", sizeof(part_str));
8548 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
8549 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
8550 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
8551 			   part_str);
8552 	else
8553 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
8554 			   hw->mac.type, hw->phy.type, part_str);
8555 
8556 	e_dev_info("%pM\n", netdev->dev_addr);
8557 
8558 	/* reset the hardware with the new settings */
8559 	err = hw->mac.ops.start_hw(hw);
8560 	if (err == IXGBE_ERR_EEPROM_VERSION) {
8561 		/* We are running on a pre-production device, log a warning */
8562 		e_dev_warn("This device is a pre-production adapter/LOM. "
8563 			   "Please be aware there may be issues associated "
8564 			   "with your hardware.  If you are experiencing "
8565 			   "problems please contact your Intel or hardware "
8566 			   "representative who provided you with this "
8567 			   "hardware.\n");
8568 	}
8569 	strcpy(netdev->name, "eth%d");
8570 	err = register_netdev(netdev);
8571 	if (err)
8572 		goto err_register;
8573 
8574 	pci_set_drvdata(pdev, adapter);
8575 
8576 	/* power down the optics for 82599 SFP+ fiber */
8577 	if (hw->mac.ops.disable_tx_laser)
8578 		hw->mac.ops.disable_tx_laser(hw);
8579 
8580 	/* carrier off reporting is important to ethtool even BEFORE open */
8581 	netif_carrier_off(netdev);
8582 
8583 #ifdef CONFIG_IXGBE_DCA
8584 	if (dca_add_requester(&pdev->dev) == 0) {
8585 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
8586 		ixgbe_setup_dca(adapter);
8587 	}
8588 #endif
8589 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
8590 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
8591 		for (i = 0; i < adapter->num_vfs; i++)
8592 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
8593 	}
8594 
8595 	/* firmware requires driver version to be 0xFFFFFFFF
8596 	 * since os does not support feature
8597 	 */
8598 	if (hw->mac.ops.set_fw_drv_ver)
8599 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF,
8600 					   0xFF);
8601 
8602 	/* add san mac addr to netdev */
8603 	ixgbe_add_sanmac_netdev(netdev);
8604 
8605 	e_dev_info("%s\n", ixgbe_default_device_descr);
8606 
8607 #ifdef CONFIG_IXGBE_HWMON
8608 	if (ixgbe_sysfs_init(adapter))
8609 		e_err(probe, "failed to allocate sysfs resources\n");
8610 #endif /* CONFIG_IXGBE_HWMON */
8611 
8612 	ixgbe_dbg_adapter_init(adapter);
8613 
8614 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
8615 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
8616 		hw->mac.ops.setup_link(hw,
8617 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
8618 			true);
8619 
8620 	return 0;
8621 
8622 err_register:
8623 	ixgbe_release_hw_control(adapter);
8624 	ixgbe_clear_interrupt_scheme(adapter);
8625 err_sw_init:
8626 	ixgbe_disable_sriov(adapter);
8627 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
8628 	iounmap(adapter->io_addr);
8629 	kfree(adapter->mac_table);
8630 err_ioremap:
8631 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
8632 	free_netdev(netdev);
8633 err_alloc_etherdev:
8634 	pci_release_selected_regions(pdev,
8635 				     pci_select_bars(pdev, IORESOURCE_MEM));
8636 err_pci_reg:
8637 err_dma:
8638 	if (!adapter || disable_dev)
8639 		pci_disable_device(pdev);
8640 	return err;
8641 }
8642 
8643 /**
8644  * ixgbe_remove - Device Removal Routine
8645  * @pdev: PCI device information struct
8646  *
8647  * ixgbe_remove is called by the PCI subsystem to alert the driver
8648  * that it should release a PCI device.  The could be caused by a
8649  * Hot-Plug event, or because the driver is going to be removed from
8650  * memory.
8651  **/
8652 static void ixgbe_remove(struct pci_dev *pdev)
8653 {
8654 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
8655 	struct net_device *netdev;
8656 	bool disable_dev;
8657 
8658 	/* if !adapter then we already cleaned up in probe */
8659 	if (!adapter)
8660 		return;
8661 
8662 	netdev  = adapter->netdev;
8663 	ixgbe_dbg_adapter_exit(adapter);
8664 
8665 	set_bit(__IXGBE_REMOVING, &adapter->state);
8666 	cancel_work_sync(&adapter->service_task);
8667 
8668 
8669 #ifdef CONFIG_IXGBE_DCA
8670 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
8671 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
8672 		dca_remove_requester(&pdev->dev);
8673 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL, 1);
8674 	}
8675 
8676 #endif
8677 #ifdef CONFIG_IXGBE_HWMON
8678 	ixgbe_sysfs_exit(adapter);
8679 #endif /* CONFIG_IXGBE_HWMON */
8680 
8681 	/* remove the added san mac */
8682 	ixgbe_del_sanmac_netdev(netdev);
8683 
8684 	if (netdev->reg_state == NETREG_REGISTERED)
8685 		unregister_netdev(netdev);
8686 
8687 #ifdef CONFIG_PCI_IOV
8688 	/*
8689 	 * Only disable SR-IOV on unload if the user specified the now
8690 	 * deprecated max_vfs module parameter.
8691 	 */
8692 	if (max_vfs)
8693 		ixgbe_disable_sriov(adapter);
8694 #endif
8695 	ixgbe_clear_interrupt_scheme(adapter);
8696 
8697 	ixgbe_release_hw_control(adapter);
8698 
8699 #ifdef CONFIG_DCB
8700 	kfree(adapter->ixgbe_ieee_pfc);
8701 	kfree(adapter->ixgbe_ieee_ets);
8702 
8703 #endif
8704 	iounmap(adapter->io_addr);
8705 	pci_release_selected_regions(pdev, pci_select_bars(pdev,
8706 				     IORESOURCE_MEM));
8707 
8708 	e_dev_info("complete\n");
8709 
8710 	kfree(adapter->mac_table);
8711 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
8712 	free_netdev(netdev);
8713 
8714 	pci_disable_pcie_error_reporting(pdev);
8715 
8716 	if (disable_dev)
8717 		pci_disable_device(pdev);
8718 }
8719 
8720 /**
8721  * ixgbe_io_error_detected - called when PCI error is detected
8722  * @pdev: Pointer to PCI device
8723  * @state: The current pci connection state
8724  *
8725  * This function is called after a PCI bus error affecting
8726  * this device has been detected.
8727  */
8728 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
8729 						pci_channel_state_t state)
8730 {
8731 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
8732 	struct net_device *netdev = adapter->netdev;
8733 
8734 #ifdef CONFIG_PCI_IOV
8735 	struct ixgbe_hw *hw = &adapter->hw;
8736 	struct pci_dev *bdev, *vfdev;
8737 	u32 dw0, dw1, dw2, dw3;
8738 	int vf, pos;
8739 	u16 req_id, pf_func;
8740 
8741 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
8742 	    adapter->num_vfs == 0)
8743 		goto skip_bad_vf_detection;
8744 
8745 	bdev = pdev->bus->self;
8746 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
8747 		bdev = bdev->bus->self;
8748 
8749 	if (!bdev)
8750 		goto skip_bad_vf_detection;
8751 
8752 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
8753 	if (!pos)
8754 		goto skip_bad_vf_detection;
8755 
8756 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
8757 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
8758 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
8759 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
8760 	if (ixgbe_removed(hw->hw_addr))
8761 		goto skip_bad_vf_detection;
8762 
8763 	req_id = dw1 >> 16;
8764 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
8765 	if (!(req_id & 0x0080))
8766 		goto skip_bad_vf_detection;
8767 
8768 	pf_func = req_id & 0x01;
8769 	if ((pf_func & 1) == (pdev->devfn & 1)) {
8770 		unsigned int device_id;
8771 
8772 		vf = (req_id & 0x7F) >> 1;
8773 		e_dev_err("VF %d has caused a PCIe error\n", vf);
8774 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
8775 				"%8.8x\tdw3: %8.8x\n",
8776 		dw0, dw1, dw2, dw3);
8777 		switch (adapter->hw.mac.type) {
8778 		case ixgbe_mac_82599EB:
8779 			device_id = IXGBE_82599_VF_DEVICE_ID;
8780 			break;
8781 		case ixgbe_mac_X540:
8782 			device_id = IXGBE_X540_VF_DEVICE_ID;
8783 			break;
8784 		case ixgbe_mac_X550:
8785 			device_id = IXGBE_DEV_ID_X550_VF;
8786 			break;
8787 		case ixgbe_mac_X550EM_x:
8788 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
8789 			break;
8790 		default:
8791 			device_id = 0;
8792 			break;
8793 		}
8794 
8795 		/* Find the pci device of the offending VF */
8796 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
8797 		while (vfdev) {
8798 			if (vfdev->devfn == (req_id & 0xFF))
8799 				break;
8800 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
8801 					       device_id, vfdev);
8802 		}
8803 		/*
8804 		 * There's a slim chance the VF could have been hot plugged,
8805 		 * so if it is no longer present we don't need to issue the
8806 		 * VFLR.  Just clean up the AER in that case.
8807 		 */
8808 		if (vfdev) {
8809 			ixgbe_issue_vf_flr(adapter, vfdev);
8810 			/* Free device reference count */
8811 			pci_dev_put(vfdev);
8812 		}
8813 
8814 		pci_cleanup_aer_uncorrect_error_status(pdev);
8815 	}
8816 
8817 	/*
8818 	 * Even though the error may have occurred on the other port
8819 	 * we still need to increment the vf error reference count for
8820 	 * both ports because the I/O resume function will be called
8821 	 * for both of them.
8822 	 */
8823 	adapter->vferr_refcount++;
8824 
8825 	return PCI_ERS_RESULT_RECOVERED;
8826 
8827 skip_bad_vf_detection:
8828 #endif /* CONFIG_PCI_IOV */
8829 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
8830 		return PCI_ERS_RESULT_DISCONNECT;
8831 
8832 	rtnl_lock();
8833 	netif_device_detach(netdev);
8834 
8835 	if (state == pci_channel_io_perm_failure) {
8836 		rtnl_unlock();
8837 		return PCI_ERS_RESULT_DISCONNECT;
8838 	}
8839 
8840 	if (netif_running(netdev))
8841 		ixgbe_down(adapter);
8842 
8843 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
8844 		pci_disable_device(pdev);
8845 	rtnl_unlock();
8846 
8847 	/* Request a slot reset. */
8848 	return PCI_ERS_RESULT_NEED_RESET;
8849 }
8850 
8851 /**
8852  * ixgbe_io_slot_reset - called after the pci bus has been reset.
8853  * @pdev: Pointer to PCI device
8854  *
8855  * Restart the card from scratch, as if from a cold-boot.
8856  */
8857 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
8858 {
8859 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
8860 	pci_ers_result_t result;
8861 	int err;
8862 
8863 	if (pci_enable_device_mem(pdev)) {
8864 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
8865 		result = PCI_ERS_RESULT_DISCONNECT;
8866 	} else {
8867 		smp_mb__before_atomic();
8868 		clear_bit(__IXGBE_DISABLED, &adapter->state);
8869 		adapter->hw.hw_addr = adapter->io_addr;
8870 		pci_set_master(pdev);
8871 		pci_restore_state(pdev);
8872 		pci_save_state(pdev);
8873 
8874 		pci_wake_from_d3(pdev, false);
8875 
8876 		ixgbe_reset(adapter);
8877 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
8878 		result = PCI_ERS_RESULT_RECOVERED;
8879 	}
8880 
8881 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
8882 	if (err) {
8883 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
8884 			  "failed 0x%0x\n", err);
8885 		/* non-fatal, continue */
8886 	}
8887 
8888 	return result;
8889 }
8890 
8891 /**
8892  * ixgbe_io_resume - called when traffic can start flowing again.
8893  * @pdev: Pointer to PCI device
8894  *
8895  * This callback is called when the error recovery driver tells us that
8896  * its OK to resume normal operation.
8897  */
8898 static void ixgbe_io_resume(struct pci_dev *pdev)
8899 {
8900 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
8901 	struct net_device *netdev = adapter->netdev;
8902 
8903 #ifdef CONFIG_PCI_IOV
8904 	if (adapter->vferr_refcount) {
8905 		e_info(drv, "Resuming after VF err\n");
8906 		adapter->vferr_refcount--;
8907 		return;
8908 	}
8909 
8910 #endif
8911 	if (netif_running(netdev))
8912 		ixgbe_up(adapter);
8913 
8914 	netif_device_attach(netdev);
8915 }
8916 
8917 static const struct pci_error_handlers ixgbe_err_handler = {
8918 	.error_detected = ixgbe_io_error_detected,
8919 	.slot_reset = ixgbe_io_slot_reset,
8920 	.resume = ixgbe_io_resume,
8921 };
8922 
8923 static struct pci_driver ixgbe_driver = {
8924 	.name     = ixgbe_driver_name,
8925 	.id_table = ixgbe_pci_tbl,
8926 	.probe    = ixgbe_probe,
8927 	.remove   = ixgbe_remove,
8928 #ifdef CONFIG_PM
8929 	.suspend  = ixgbe_suspend,
8930 	.resume   = ixgbe_resume,
8931 #endif
8932 	.shutdown = ixgbe_shutdown,
8933 	.sriov_configure = ixgbe_pci_sriov_configure,
8934 	.err_handler = &ixgbe_err_handler
8935 };
8936 
8937 /**
8938  * ixgbe_init_module - Driver Registration Routine
8939  *
8940  * ixgbe_init_module is the first routine called when the driver is
8941  * loaded. All it does is register with the PCI subsystem.
8942  **/
8943 static int __init ixgbe_init_module(void)
8944 {
8945 	int ret;
8946 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
8947 	pr_info("%s\n", ixgbe_copyright);
8948 
8949 	ixgbe_dbg_init();
8950 
8951 	ret = pci_register_driver(&ixgbe_driver);
8952 	if (ret) {
8953 		ixgbe_dbg_exit();
8954 		return ret;
8955 	}
8956 
8957 #ifdef CONFIG_IXGBE_DCA
8958 	dca_register_notify(&dca_notifier);
8959 #endif
8960 
8961 	return 0;
8962 }
8963 
8964 module_init(ixgbe_init_module);
8965 
8966 /**
8967  * ixgbe_exit_module - Driver Exit Cleanup Routine
8968  *
8969  * ixgbe_exit_module is called just before the driver is removed
8970  * from memory.
8971  **/
8972 static void __exit ixgbe_exit_module(void)
8973 {
8974 #ifdef CONFIG_IXGBE_DCA
8975 	dca_unregister_notify(&dca_notifier);
8976 #endif
8977 	pci_unregister_driver(&ixgbe_driver);
8978 
8979 	ixgbe_dbg_exit();
8980 
8981 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
8982 }
8983 
8984 #ifdef CONFIG_IXGBE_DCA
8985 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
8986 			    void *p)
8987 {
8988 	int ret_val;
8989 
8990 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
8991 					 __ixgbe_notify_dca);
8992 
8993 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
8994 }
8995 
8996 #endif /* CONFIG_IXGBE_DCA */
8997 
8998 module_exit(ixgbe_exit_module);
8999 
9000 /* ixgbe_main.c */
9001