xref: /openbmc/linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision a0ae2562c6c4b2721d9fddba63b7286c13517d9f)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright(c) 1999 - 2018 Intel Corporation. */
3 
4 #include <linux/types.h>
5 #include <linux/module.h>
6 #include <linux/pci.h>
7 #include <linux/netdevice.h>
8 #include <linux/vmalloc.h>
9 #include <linux/string.h>
10 #include <linux/in.h>
11 #include <linux/interrupt.h>
12 #include <linux/ip.h>
13 #include <linux/tcp.h>
14 #include <linux/sctp.h>
15 #include <linux/pkt_sched.h>
16 #include <linux/ipv6.h>
17 #include <linux/slab.h>
18 #include <net/checksum.h>
19 #include <net/ip6_checksum.h>
20 #include <linux/etherdevice.h>
21 #include <linux/ethtool.h>
22 #include <linux/if.h>
23 #include <linux/if_vlan.h>
24 #include <linux/if_macvlan.h>
25 #include <linux/if_bridge.h>
26 #include <linux/prefetch.h>
27 #include <linux/bpf.h>
28 #include <linux/bpf_trace.h>
29 #include <linux/atomic.h>
30 #include <scsi/fc/fc_fcoe.h>
31 #include <net/udp_tunnel.h>
32 #include <net/pkt_cls.h>
33 #include <net/tc_act/tc_gact.h>
34 #include <net/tc_act/tc_mirred.h>
35 #include <net/vxlan.h>
36 #include <net/mpls.h>
37 
38 #include "ixgbe.h"
39 #include "ixgbe_common.h"
40 #include "ixgbe_dcb_82599.h"
41 #include "ixgbe_sriov.h"
42 #include "ixgbe_model.h"
43 
44 char ixgbe_driver_name[] = "ixgbe";
45 static const char ixgbe_driver_string[] =
46 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
47 #ifdef IXGBE_FCOE
48 char ixgbe_default_device_descr[] =
49 			      "Intel(R) 10 Gigabit Network Connection";
50 #else
51 static char ixgbe_default_device_descr[] =
52 			      "Intel(R) 10 Gigabit Network Connection";
53 #endif
54 #define DRV_VERSION "5.1.0-k"
55 const char ixgbe_driver_version[] = DRV_VERSION;
56 static const char ixgbe_copyright[] =
57 				"Copyright (c) 1999-2016 Intel Corporation.";
58 
59 static const char ixgbe_overheat_msg[] = "Network adapter has been stopped because it has over heated. Restart the computer. If the problem persists, power off the system and replace the adapter";
60 
61 static const struct ixgbe_info *ixgbe_info_tbl[] = {
62 	[board_82598]		= &ixgbe_82598_info,
63 	[board_82599]		= &ixgbe_82599_info,
64 	[board_X540]		= &ixgbe_X540_info,
65 	[board_X550]		= &ixgbe_X550_info,
66 	[board_X550EM_x]	= &ixgbe_X550EM_x_info,
67 	[board_x550em_x_fw]	= &ixgbe_x550em_x_fw_info,
68 	[board_x550em_a]	= &ixgbe_x550em_a_info,
69 	[board_x550em_a_fw]	= &ixgbe_x550em_a_fw_info,
70 };
71 
72 /* ixgbe_pci_tbl - PCI Device ID Table
73  *
74  * Wildcard entries (PCI_ANY_ID) should come last
75  * Last entry must be all 0s
76  *
77  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
78  *   Class, Class Mask, private data (not used) }
79  */
80 static const struct pci_device_id ixgbe_pci_tbl[] = {
81 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
82 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
83 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
84 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
85 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
86 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
87 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
88 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
89 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
90 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
91 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
92 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
93 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
94 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
95 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
96 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
97 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
98 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
99 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
100 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
101 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
102 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
103 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
104 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
105 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T1), board_X550},
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_XFI), board_X550EM_x},
115 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
116 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_10G_T), board_X550EM_x},
117 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_SFP), board_X550EM_x},
118 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_1G_T), board_x550em_x_fw},
119 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR), board_x550em_a },
120 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR_L), board_x550em_a },
121 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP_N), board_x550em_a },
122 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII), board_x550em_a },
123 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII_L), board_x550em_a },
124 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_10G_T), board_x550em_a},
125 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP), board_x550em_a },
126 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T), board_x550em_a_fw },
127 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T_L), board_x550em_a_fw },
128 	/* required last entry */
129 	{0, }
130 };
131 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
132 
133 #ifdef CONFIG_IXGBE_DCA
134 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
135 			    void *p);
136 static struct notifier_block dca_notifier = {
137 	.notifier_call = ixgbe_notify_dca,
138 	.next          = NULL,
139 	.priority      = 0
140 };
141 #endif
142 
143 #ifdef CONFIG_PCI_IOV
144 static unsigned int max_vfs;
145 module_param(max_vfs, uint, 0);
146 MODULE_PARM_DESC(max_vfs,
147 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
148 #endif /* CONFIG_PCI_IOV */
149 
150 static unsigned int allow_unsupported_sfp;
151 module_param(allow_unsupported_sfp, uint, 0);
152 MODULE_PARM_DESC(allow_unsupported_sfp,
153 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
154 
155 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
156 static int debug = -1;
157 module_param(debug, int, 0);
158 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
159 
160 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
161 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
162 MODULE_LICENSE("GPL");
163 MODULE_VERSION(DRV_VERSION);
164 
165 static struct workqueue_struct *ixgbe_wq;
166 
167 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
168 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *);
169 
170 static const struct net_device_ops ixgbe_netdev_ops;
171 
172 static bool netif_is_ixgbe(struct net_device *dev)
173 {
174 	return dev && (dev->netdev_ops == &ixgbe_netdev_ops);
175 }
176 
177 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
178 					  u32 reg, u16 *value)
179 {
180 	struct pci_dev *parent_dev;
181 	struct pci_bus *parent_bus;
182 
183 	parent_bus = adapter->pdev->bus->parent;
184 	if (!parent_bus)
185 		return -1;
186 
187 	parent_dev = parent_bus->self;
188 	if (!parent_dev)
189 		return -1;
190 
191 	if (!pci_is_pcie(parent_dev))
192 		return -1;
193 
194 	pcie_capability_read_word(parent_dev, reg, value);
195 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
196 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
197 		return -1;
198 	return 0;
199 }
200 
201 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
202 {
203 	struct ixgbe_hw *hw = &adapter->hw;
204 	u16 link_status = 0;
205 	int err;
206 
207 	hw->bus.type = ixgbe_bus_type_pci_express;
208 
209 	/* Get the negotiated link width and speed from PCI config space of the
210 	 * parent, as this device is behind a switch
211 	 */
212 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
213 
214 	/* assume caller will handle error case */
215 	if (err)
216 		return err;
217 
218 	hw->bus.width = ixgbe_convert_bus_width(link_status);
219 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
220 
221 	return 0;
222 }
223 
224 /**
225  * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
226  * @hw: hw specific details
227  *
228  * This function is used by probe to determine whether a device's PCI-Express
229  * bandwidth details should be gathered from the parent bus instead of from the
230  * device. Used to ensure that various locations all have the correct device ID
231  * checks.
232  */
233 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
234 {
235 	switch (hw->device_id) {
236 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
237 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
238 		return true;
239 	default:
240 		return false;
241 	}
242 }
243 
244 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
245 				     int expected_gts)
246 {
247 	struct ixgbe_hw *hw = &adapter->hw;
248 	struct pci_dev *pdev;
249 
250 	/* Some devices are not connected over PCIe and thus do not negotiate
251 	 * speed. These devices do not have valid bus info, and thus any report
252 	 * we generate may not be correct.
253 	 */
254 	if (hw->bus.type == ixgbe_bus_type_internal)
255 		return;
256 
257 	/* determine whether to use the parent device */
258 	if (ixgbe_pcie_from_parent(&adapter->hw))
259 		pdev = adapter->pdev->bus->parent->self;
260 	else
261 		pdev = adapter->pdev;
262 
263 	pcie_print_link_status(pdev);
264 }
265 
266 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
267 {
268 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
269 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
270 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
271 		queue_work(ixgbe_wq, &adapter->service_task);
272 }
273 
274 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
275 {
276 	struct ixgbe_adapter *adapter = hw->back;
277 
278 	if (!hw->hw_addr)
279 		return;
280 	hw->hw_addr = NULL;
281 	e_dev_err("Adapter removed\n");
282 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
283 		ixgbe_service_event_schedule(adapter);
284 }
285 
286 static u32 ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
287 {
288 	u8 __iomem *reg_addr;
289 	u32 value;
290 	int i;
291 
292 	reg_addr = READ_ONCE(hw->hw_addr);
293 	if (ixgbe_removed(reg_addr))
294 		return IXGBE_FAILED_READ_REG;
295 
296 	/* Register read of 0xFFFFFFF can indicate the adapter has been removed,
297 	 * so perform several status register reads to determine if the adapter
298 	 * has been removed.
299 	 */
300 	for (i = 0; i < IXGBE_FAILED_READ_RETRIES; i++) {
301 		value = readl(reg_addr + IXGBE_STATUS);
302 		if (value != IXGBE_FAILED_READ_REG)
303 			break;
304 		mdelay(3);
305 	}
306 
307 	if (value == IXGBE_FAILED_READ_REG)
308 		ixgbe_remove_adapter(hw);
309 	else
310 		value = readl(reg_addr + reg);
311 	return value;
312 }
313 
314 /**
315  * ixgbe_read_reg - Read from device register
316  * @hw: hw specific details
317  * @reg: offset of register to read
318  *
319  * Returns : value read or IXGBE_FAILED_READ_REG if removed
320  *
321  * This function is used to read device registers. It checks for device
322  * removal by confirming any read that returns all ones by checking the
323  * status register value for all ones. This function avoids reading from
324  * the hardware if a removal was previously detected in which case it
325  * returns IXGBE_FAILED_READ_REG (all ones).
326  */
327 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
328 {
329 	u8 __iomem *reg_addr = READ_ONCE(hw->hw_addr);
330 	u32 value;
331 
332 	if (ixgbe_removed(reg_addr))
333 		return IXGBE_FAILED_READ_REG;
334 	if (unlikely(hw->phy.nw_mng_if_sel &
335 		     IXGBE_NW_MNG_IF_SEL_SGMII_ENABLE)) {
336 		struct ixgbe_adapter *adapter;
337 		int i;
338 
339 		for (i = 0; i < 200; ++i) {
340 			value = readl(reg_addr + IXGBE_MAC_SGMII_BUSY);
341 			if (likely(!value))
342 				goto writes_completed;
343 			if (value == IXGBE_FAILED_READ_REG) {
344 				ixgbe_remove_adapter(hw);
345 				return IXGBE_FAILED_READ_REG;
346 			}
347 			udelay(5);
348 		}
349 
350 		adapter = hw->back;
351 		e_warn(hw, "register writes incomplete %08x\n", value);
352 	}
353 
354 writes_completed:
355 	value = readl(reg_addr + reg);
356 	if (unlikely(value == IXGBE_FAILED_READ_REG))
357 		value = ixgbe_check_remove(hw, reg);
358 	return value;
359 }
360 
361 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
362 {
363 	u16 value;
364 
365 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
366 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
367 		ixgbe_remove_adapter(hw);
368 		return true;
369 	}
370 	return false;
371 }
372 
373 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
374 {
375 	struct ixgbe_adapter *adapter = hw->back;
376 	u16 value;
377 
378 	if (ixgbe_removed(hw->hw_addr))
379 		return IXGBE_FAILED_READ_CFG_WORD;
380 	pci_read_config_word(adapter->pdev, reg, &value);
381 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
382 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
383 		return IXGBE_FAILED_READ_CFG_WORD;
384 	return value;
385 }
386 
387 #ifdef CONFIG_PCI_IOV
388 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
389 {
390 	struct ixgbe_adapter *adapter = hw->back;
391 	u32 value;
392 
393 	if (ixgbe_removed(hw->hw_addr))
394 		return IXGBE_FAILED_READ_CFG_DWORD;
395 	pci_read_config_dword(adapter->pdev, reg, &value);
396 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
397 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
398 		return IXGBE_FAILED_READ_CFG_DWORD;
399 	return value;
400 }
401 #endif /* CONFIG_PCI_IOV */
402 
403 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
404 {
405 	struct ixgbe_adapter *adapter = hw->back;
406 
407 	if (ixgbe_removed(hw->hw_addr))
408 		return;
409 	pci_write_config_word(adapter->pdev, reg, value);
410 }
411 
412 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
413 {
414 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
415 
416 	/* flush memory to make sure state is correct before next watchdog */
417 	smp_mb__before_atomic();
418 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
419 }
420 
421 struct ixgbe_reg_info {
422 	u32 ofs;
423 	char *name;
424 };
425 
426 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
427 
428 	/* General Registers */
429 	{IXGBE_CTRL, "CTRL"},
430 	{IXGBE_STATUS, "STATUS"},
431 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
432 
433 	/* Interrupt Registers */
434 	{IXGBE_EICR, "EICR"},
435 
436 	/* RX Registers */
437 	{IXGBE_SRRCTL(0), "SRRCTL"},
438 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
439 	{IXGBE_RDLEN(0), "RDLEN"},
440 	{IXGBE_RDH(0), "RDH"},
441 	{IXGBE_RDT(0), "RDT"},
442 	{IXGBE_RXDCTL(0), "RXDCTL"},
443 	{IXGBE_RDBAL(0), "RDBAL"},
444 	{IXGBE_RDBAH(0), "RDBAH"},
445 
446 	/* TX Registers */
447 	{IXGBE_TDBAL(0), "TDBAL"},
448 	{IXGBE_TDBAH(0), "TDBAH"},
449 	{IXGBE_TDLEN(0), "TDLEN"},
450 	{IXGBE_TDH(0), "TDH"},
451 	{IXGBE_TDT(0), "TDT"},
452 	{IXGBE_TXDCTL(0), "TXDCTL"},
453 
454 	/* List Terminator */
455 	{ .name = NULL }
456 };
457 
458 
459 /*
460  * ixgbe_regdump - register printout routine
461  */
462 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
463 {
464 	int i;
465 	char rname[16];
466 	u32 regs[64];
467 
468 	switch (reginfo->ofs) {
469 	case IXGBE_SRRCTL(0):
470 		for (i = 0; i < 64; i++)
471 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
472 		break;
473 	case IXGBE_DCA_RXCTRL(0):
474 		for (i = 0; i < 64; i++)
475 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
476 		break;
477 	case IXGBE_RDLEN(0):
478 		for (i = 0; i < 64; i++)
479 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
480 		break;
481 	case IXGBE_RDH(0):
482 		for (i = 0; i < 64; i++)
483 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
484 		break;
485 	case IXGBE_RDT(0):
486 		for (i = 0; i < 64; i++)
487 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
488 		break;
489 	case IXGBE_RXDCTL(0):
490 		for (i = 0; i < 64; i++)
491 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
492 		break;
493 	case IXGBE_RDBAL(0):
494 		for (i = 0; i < 64; i++)
495 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
496 		break;
497 	case IXGBE_RDBAH(0):
498 		for (i = 0; i < 64; i++)
499 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
500 		break;
501 	case IXGBE_TDBAL(0):
502 		for (i = 0; i < 64; i++)
503 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
504 		break;
505 	case IXGBE_TDBAH(0):
506 		for (i = 0; i < 64; i++)
507 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
508 		break;
509 	case IXGBE_TDLEN(0):
510 		for (i = 0; i < 64; i++)
511 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
512 		break;
513 	case IXGBE_TDH(0):
514 		for (i = 0; i < 64; i++)
515 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
516 		break;
517 	case IXGBE_TDT(0):
518 		for (i = 0; i < 64; i++)
519 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
520 		break;
521 	case IXGBE_TXDCTL(0):
522 		for (i = 0; i < 64; i++)
523 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
524 		break;
525 	default:
526 		pr_info("%-15s %08x\n",
527 			reginfo->name, IXGBE_READ_REG(hw, reginfo->ofs));
528 		return;
529 	}
530 
531 	i = 0;
532 	while (i < 64) {
533 		int j;
534 		char buf[9 * 8 + 1];
535 		char *p = buf;
536 
537 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i, i + 7);
538 		for (j = 0; j < 8; j++)
539 			p += sprintf(p, " %08x", regs[i++]);
540 		pr_err("%-15s%s\n", rname, buf);
541 	}
542 
543 }
544 
545 static void ixgbe_print_buffer(struct ixgbe_ring *ring, int n)
546 {
547 	struct ixgbe_tx_buffer *tx_buffer;
548 
549 	tx_buffer = &ring->tx_buffer_info[ring->next_to_clean];
550 	pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
551 		n, ring->next_to_use, ring->next_to_clean,
552 		(u64)dma_unmap_addr(tx_buffer, dma),
553 		dma_unmap_len(tx_buffer, len),
554 		tx_buffer->next_to_watch,
555 		(u64)tx_buffer->time_stamp);
556 }
557 
558 /*
559  * ixgbe_dump - Print registers, tx-rings and rx-rings
560  */
561 static void ixgbe_dump(struct ixgbe_adapter *adapter)
562 {
563 	struct net_device *netdev = adapter->netdev;
564 	struct ixgbe_hw *hw = &adapter->hw;
565 	struct ixgbe_reg_info *reginfo;
566 	int n = 0;
567 	struct ixgbe_ring *ring;
568 	struct ixgbe_tx_buffer *tx_buffer;
569 	union ixgbe_adv_tx_desc *tx_desc;
570 	struct my_u0 { u64 a; u64 b; } *u0;
571 	struct ixgbe_ring *rx_ring;
572 	union ixgbe_adv_rx_desc *rx_desc;
573 	struct ixgbe_rx_buffer *rx_buffer_info;
574 	int i = 0;
575 
576 	if (!netif_msg_hw(adapter))
577 		return;
578 
579 	/* Print netdevice Info */
580 	if (netdev) {
581 		dev_info(&adapter->pdev->dev, "Net device Info\n");
582 		pr_info("Device Name     state            "
583 			"trans_start\n");
584 		pr_info("%-15s %016lX %016lX\n",
585 			netdev->name,
586 			netdev->state,
587 			dev_trans_start(netdev));
588 	}
589 
590 	/* Print Registers */
591 	dev_info(&adapter->pdev->dev, "Register Dump\n");
592 	pr_info(" Register Name   Value\n");
593 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
594 	     reginfo->name; reginfo++) {
595 		ixgbe_regdump(hw, reginfo);
596 	}
597 
598 	/* Print TX Ring Summary */
599 	if (!netdev || !netif_running(netdev))
600 		return;
601 
602 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
603 	pr_info(" %s     %s              %s        %s\n",
604 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
605 		"leng", "ntw", "timestamp");
606 	for (n = 0; n < adapter->num_tx_queues; n++) {
607 		ring = adapter->tx_ring[n];
608 		ixgbe_print_buffer(ring, n);
609 	}
610 
611 	for (n = 0; n < adapter->num_xdp_queues; n++) {
612 		ring = adapter->xdp_ring[n];
613 		ixgbe_print_buffer(ring, n);
614 	}
615 
616 	/* Print TX Rings */
617 	if (!netif_msg_tx_done(adapter))
618 		goto rx_ring_summary;
619 
620 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
621 
622 	/* Transmit Descriptor Formats
623 	 *
624 	 * 82598 Advanced Transmit Descriptor
625 	 *   +--------------------------------------------------------------+
626 	 * 0 |         Buffer Address [63:0]                                |
627 	 *   +--------------------------------------------------------------+
628 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
629 	 *   +--------------------------------------------------------------+
630 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
631 	 *
632 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
633 	 *   +--------------------------------------------------------------+
634 	 * 0 |                          RSV [63:0]                          |
635 	 *   +--------------------------------------------------------------+
636 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
637 	 *   +--------------------------------------------------------------+
638 	 *   63                       36 35   32 31                         0
639 	 *
640 	 * 82599+ Advanced Transmit Descriptor
641 	 *   +--------------------------------------------------------------+
642 	 * 0 |         Buffer Address [63:0]                                |
643 	 *   +--------------------------------------------------------------+
644 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
645 	 *   +--------------------------------------------------------------+
646 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
647 	 *
648 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
649 	 *   +--------------------------------------------------------------+
650 	 * 0 |                          RSV [63:0]                          |
651 	 *   +--------------------------------------------------------------+
652 	 * 8 |            RSV           |  STA  |           RSV             |
653 	 *   +--------------------------------------------------------------+
654 	 *   63                       36 35   32 31                         0
655 	 */
656 
657 	for (n = 0; n < adapter->num_tx_queues; n++) {
658 		ring = adapter->tx_ring[n];
659 		pr_info("------------------------------------\n");
660 		pr_info("TX QUEUE INDEX = %d\n", ring->queue_index);
661 		pr_info("------------------------------------\n");
662 		pr_info("%s%s    %s              %s        %s          %s\n",
663 			"T [desc]     [address 63:0  ] ",
664 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
665 			"leng", "ntw", "timestamp", "bi->skb");
666 
667 		for (i = 0; ring->desc && (i < ring->count); i++) {
668 			tx_desc = IXGBE_TX_DESC(ring, i);
669 			tx_buffer = &ring->tx_buffer_info[i];
670 			u0 = (struct my_u0 *)tx_desc;
671 			if (dma_unmap_len(tx_buffer, len) > 0) {
672 				const char *ring_desc;
673 
674 				if (i == ring->next_to_use &&
675 				    i == ring->next_to_clean)
676 					ring_desc = " NTC/U";
677 				else if (i == ring->next_to_use)
678 					ring_desc = " NTU";
679 				else if (i == ring->next_to_clean)
680 					ring_desc = " NTC";
681 				else
682 					ring_desc = "";
683 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p%s",
684 					i,
685 					le64_to_cpu((__force __le64)u0->a),
686 					le64_to_cpu((__force __le64)u0->b),
687 					(u64)dma_unmap_addr(tx_buffer, dma),
688 					dma_unmap_len(tx_buffer, len),
689 					tx_buffer->next_to_watch,
690 					(u64)tx_buffer->time_stamp,
691 					tx_buffer->skb,
692 					ring_desc);
693 
694 				if (netif_msg_pktdata(adapter) &&
695 				    tx_buffer->skb)
696 					print_hex_dump(KERN_INFO, "",
697 						DUMP_PREFIX_ADDRESS, 16, 1,
698 						tx_buffer->skb->data,
699 						dma_unmap_len(tx_buffer, len),
700 						true);
701 			}
702 		}
703 	}
704 
705 	/* Print RX Rings Summary */
706 rx_ring_summary:
707 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
708 	pr_info("Queue [NTU] [NTC]\n");
709 	for (n = 0; n < adapter->num_rx_queues; n++) {
710 		rx_ring = adapter->rx_ring[n];
711 		pr_info("%5d %5X %5X\n",
712 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
713 	}
714 
715 	/* Print RX Rings */
716 	if (!netif_msg_rx_status(adapter))
717 		return;
718 
719 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
720 
721 	/* Receive Descriptor Formats
722 	 *
723 	 * 82598 Advanced Receive Descriptor (Read) Format
724 	 *    63                                           1        0
725 	 *    +-----------------------------------------------------+
726 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
727 	 *    +----------------------------------------------+------+
728 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
729 	 *    +-----------------------------------------------------+
730 	 *
731 	 *
732 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
733 	 *
734 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
735 	 *   +------------------------------------------------------+
736 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
737 	 *   | Packet   | IP     |   |          |     | Type | Type |
738 	 *   | Checksum | Ident  |   |          |     |      |      |
739 	 *   +------------------------------------------------------+
740 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
741 	 *   +------------------------------------------------------+
742 	 *   63       48 47    32 31            20 19               0
743 	 *
744 	 * 82599+ Advanced Receive Descriptor (Read) Format
745 	 *    63                                           1        0
746 	 *    +-----------------------------------------------------+
747 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
748 	 *    +----------------------------------------------+------+
749 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
750 	 *    +-----------------------------------------------------+
751 	 *
752 	 *
753 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
754 	 *
755 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
756 	 *   +------------------------------------------------------+
757 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
758 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
759 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
760 	 *   +------------------------------------------------------+
761 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
762 	 *   +------------------------------------------------------+
763 	 *   63       48 47    32 31          20 19                 0
764 	 */
765 
766 	for (n = 0; n < adapter->num_rx_queues; n++) {
767 		rx_ring = adapter->rx_ring[n];
768 		pr_info("------------------------------------\n");
769 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
770 		pr_info("------------------------------------\n");
771 		pr_info("%s%s%s\n",
772 			"R  [desc]      [ PktBuf     A0] ",
773 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
774 			"<-- Adv Rx Read format");
775 		pr_info("%s%s%s\n",
776 			"RWB[desc]      [PcsmIpSHl PtRs] ",
777 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
778 			"<-- Adv Rx Write-Back format");
779 
780 		for (i = 0; i < rx_ring->count; i++) {
781 			const char *ring_desc;
782 
783 			if (i == rx_ring->next_to_use)
784 				ring_desc = " NTU";
785 			else if (i == rx_ring->next_to_clean)
786 				ring_desc = " NTC";
787 			else
788 				ring_desc = "";
789 
790 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
791 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
792 			u0 = (struct my_u0 *)rx_desc;
793 			if (rx_desc->wb.upper.length) {
794 				/* Descriptor Done */
795 				pr_info("RWB[0x%03X]     %016llX %016llX ---------------- %p%s\n",
796 					i,
797 					le64_to_cpu((__force __le64)u0->a),
798 					le64_to_cpu((__force __le64)u0->b),
799 					rx_buffer_info->skb,
800 					ring_desc);
801 			} else {
802 				pr_info("R  [0x%03X]     %016llX %016llX %016llX %p%s\n",
803 					i,
804 					le64_to_cpu((__force __le64)u0->a),
805 					le64_to_cpu((__force __le64)u0->b),
806 					(u64)rx_buffer_info->dma,
807 					rx_buffer_info->skb,
808 					ring_desc);
809 
810 				if (netif_msg_pktdata(adapter) &&
811 				    rx_buffer_info->dma) {
812 					print_hex_dump(KERN_INFO, "",
813 					   DUMP_PREFIX_ADDRESS, 16, 1,
814 					   page_address(rx_buffer_info->page) +
815 						    rx_buffer_info->page_offset,
816 					   ixgbe_rx_bufsz(rx_ring), true);
817 				}
818 			}
819 		}
820 	}
821 }
822 
823 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
824 {
825 	u32 ctrl_ext;
826 
827 	/* Let firmware take over control of h/w */
828 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
829 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
830 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
831 }
832 
833 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
834 {
835 	u32 ctrl_ext;
836 
837 	/* Let firmware know the driver has taken over */
838 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
839 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
840 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
841 }
842 
843 /**
844  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
845  * @adapter: pointer to adapter struct
846  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
847  * @queue: queue to map the corresponding interrupt to
848  * @msix_vector: the vector to map to the corresponding queue
849  *
850  */
851 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
852 			   u8 queue, u8 msix_vector)
853 {
854 	u32 ivar, index;
855 	struct ixgbe_hw *hw = &adapter->hw;
856 	switch (hw->mac.type) {
857 	case ixgbe_mac_82598EB:
858 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
859 		if (direction == -1)
860 			direction = 0;
861 		index = (((direction * 64) + queue) >> 2) & 0x1F;
862 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
863 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
864 		ivar |= (msix_vector << (8 * (queue & 0x3)));
865 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
866 		break;
867 	case ixgbe_mac_82599EB:
868 	case ixgbe_mac_X540:
869 	case ixgbe_mac_X550:
870 	case ixgbe_mac_X550EM_x:
871 	case ixgbe_mac_x550em_a:
872 		if (direction == -1) {
873 			/* other causes */
874 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
875 			index = ((queue & 1) * 8);
876 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
877 			ivar &= ~(0xFF << index);
878 			ivar |= (msix_vector << index);
879 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
880 			break;
881 		} else {
882 			/* tx or rx causes */
883 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
884 			index = ((16 * (queue & 1)) + (8 * direction));
885 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
886 			ivar &= ~(0xFF << index);
887 			ivar |= (msix_vector << index);
888 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
889 			break;
890 		}
891 	default:
892 		break;
893 	}
894 }
895 
896 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
897 					  u64 qmask)
898 {
899 	u32 mask;
900 
901 	switch (adapter->hw.mac.type) {
902 	case ixgbe_mac_82598EB:
903 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
904 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
905 		break;
906 	case ixgbe_mac_82599EB:
907 	case ixgbe_mac_X540:
908 	case ixgbe_mac_X550:
909 	case ixgbe_mac_X550EM_x:
910 	case ixgbe_mac_x550em_a:
911 		mask = (qmask & 0xFFFFFFFF);
912 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
913 		mask = (qmask >> 32);
914 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
915 		break;
916 	default:
917 		break;
918 	}
919 }
920 
921 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
922 {
923 	struct ixgbe_hw *hw = &adapter->hw;
924 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
925 	int i;
926 	u32 data;
927 
928 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
929 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
930 		return;
931 
932 	switch (hw->mac.type) {
933 	case ixgbe_mac_82598EB:
934 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
935 		break;
936 	default:
937 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
938 	}
939 	hwstats->lxoffrxc += data;
940 
941 	/* refill credits (no tx hang) if we received xoff */
942 	if (!data)
943 		return;
944 
945 	for (i = 0; i < adapter->num_tx_queues; i++)
946 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
947 			  &adapter->tx_ring[i]->state);
948 
949 	for (i = 0; i < adapter->num_xdp_queues; i++)
950 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
951 			  &adapter->xdp_ring[i]->state);
952 }
953 
954 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
955 {
956 	struct ixgbe_hw *hw = &adapter->hw;
957 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
958 	u32 xoff[8] = {0};
959 	u8 tc;
960 	int i;
961 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
962 
963 	if (adapter->ixgbe_ieee_pfc)
964 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
965 
966 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
967 		ixgbe_update_xoff_rx_lfc(adapter);
968 		return;
969 	}
970 
971 	/* update stats for each tc, only valid with PFC enabled */
972 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
973 		u32 pxoffrxc;
974 
975 		switch (hw->mac.type) {
976 		case ixgbe_mac_82598EB:
977 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
978 			break;
979 		default:
980 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
981 		}
982 		hwstats->pxoffrxc[i] += pxoffrxc;
983 		/* Get the TC for given UP */
984 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
985 		xoff[tc] += pxoffrxc;
986 	}
987 
988 	/* disarm tx queues that have received xoff frames */
989 	for (i = 0; i < adapter->num_tx_queues; i++) {
990 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
991 
992 		tc = tx_ring->dcb_tc;
993 		if (xoff[tc])
994 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
995 	}
996 
997 	for (i = 0; i < adapter->num_xdp_queues; i++) {
998 		struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
999 
1000 		tc = xdp_ring->dcb_tc;
1001 		if (xoff[tc])
1002 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &xdp_ring->state);
1003 	}
1004 }
1005 
1006 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1007 {
1008 	return ring->stats.packets;
1009 }
1010 
1011 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1012 {
1013 	unsigned int head, tail;
1014 
1015 	head = ring->next_to_clean;
1016 	tail = ring->next_to_use;
1017 
1018 	return ((head <= tail) ? tail : tail + ring->count) - head;
1019 }
1020 
1021 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1022 {
1023 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1024 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1025 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1026 
1027 	clear_check_for_tx_hang(tx_ring);
1028 
1029 	/*
1030 	 * Check for a hung queue, but be thorough. This verifies
1031 	 * that a transmit has been completed since the previous
1032 	 * check AND there is at least one packet pending. The
1033 	 * ARMED bit is set to indicate a potential hang. The
1034 	 * bit is cleared if a pause frame is received to remove
1035 	 * false hang detection due to PFC or 802.3x frames. By
1036 	 * requiring this to fail twice we avoid races with
1037 	 * pfc clearing the ARMED bit and conditions where we
1038 	 * run the check_tx_hang logic with a transmit completion
1039 	 * pending but without time to complete it yet.
1040 	 */
1041 	if (tx_done_old == tx_done && tx_pending)
1042 		/* make sure it is true for two checks in a row */
1043 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1044 					&tx_ring->state);
1045 	/* update completed stats and continue */
1046 	tx_ring->tx_stats.tx_done_old = tx_done;
1047 	/* reset the countdown */
1048 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1049 
1050 	return false;
1051 }
1052 
1053 /**
1054  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1055  * @adapter: driver private struct
1056  **/
1057 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1058 {
1059 
1060 	/* Do the reset outside of interrupt context */
1061 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1062 		set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
1063 		e_warn(drv, "initiating reset due to tx timeout\n");
1064 		ixgbe_service_event_schedule(adapter);
1065 	}
1066 }
1067 
1068 /**
1069  * ixgbe_tx_maxrate - callback to set the maximum per-queue bitrate
1070  * @netdev: network interface device structure
1071  * @queue_index: Tx queue to set
1072  * @maxrate: desired maximum transmit bitrate
1073  **/
1074 static int ixgbe_tx_maxrate(struct net_device *netdev,
1075 			    int queue_index, u32 maxrate)
1076 {
1077 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1078 	struct ixgbe_hw *hw = &adapter->hw;
1079 	u32 bcnrc_val = ixgbe_link_mbps(adapter);
1080 
1081 	if (!maxrate)
1082 		return 0;
1083 
1084 	/* Calculate the rate factor values to set */
1085 	bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1086 	bcnrc_val /= maxrate;
1087 
1088 	/* clear everything but the rate factor */
1089 	bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1090 	IXGBE_RTTBCNRC_RF_DEC_MASK;
1091 
1092 	/* enable the rate scheduler */
1093 	bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1094 
1095 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_index);
1096 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1097 
1098 	return 0;
1099 }
1100 
1101 /**
1102  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1103  * @q_vector: structure containing interrupt and ring information
1104  * @tx_ring: tx ring to clean
1105  * @napi_budget: Used to determine if we are in netpoll
1106  **/
1107 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1108 			       struct ixgbe_ring *tx_ring, int napi_budget)
1109 {
1110 	struct ixgbe_adapter *adapter = q_vector->adapter;
1111 	struct ixgbe_tx_buffer *tx_buffer;
1112 	union ixgbe_adv_tx_desc *tx_desc;
1113 	unsigned int total_bytes = 0, total_packets = 0, total_ipsec = 0;
1114 	unsigned int budget = q_vector->tx.work_limit;
1115 	unsigned int i = tx_ring->next_to_clean;
1116 
1117 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1118 		return true;
1119 
1120 	tx_buffer = &tx_ring->tx_buffer_info[i];
1121 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1122 	i -= tx_ring->count;
1123 
1124 	do {
1125 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1126 
1127 		/* if next_to_watch is not set then there is no work pending */
1128 		if (!eop_desc)
1129 			break;
1130 
1131 		/* prevent any other reads prior to eop_desc */
1132 		smp_rmb();
1133 
1134 		/* if DD is not set pending work has not been completed */
1135 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1136 			break;
1137 
1138 		/* clear next_to_watch to prevent false hangs */
1139 		tx_buffer->next_to_watch = NULL;
1140 
1141 		/* update the statistics for this packet */
1142 		total_bytes += tx_buffer->bytecount;
1143 		total_packets += tx_buffer->gso_segs;
1144 		if (tx_buffer->tx_flags & IXGBE_TX_FLAGS_IPSEC)
1145 			total_ipsec++;
1146 
1147 		/* free the skb */
1148 		if (ring_is_xdp(tx_ring))
1149 			xdp_return_frame(tx_buffer->xdpf);
1150 		else
1151 			napi_consume_skb(tx_buffer->skb, napi_budget);
1152 
1153 		/* unmap skb header data */
1154 		dma_unmap_single(tx_ring->dev,
1155 				 dma_unmap_addr(tx_buffer, dma),
1156 				 dma_unmap_len(tx_buffer, len),
1157 				 DMA_TO_DEVICE);
1158 
1159 		/* clear tx_buffer data */
1160 		dma_unmap_len_set(tx_buffer, len, 0);
1161 
1162 		/* unmap remaining buffers */
1163 		while (tx_desc != eop_desc) {
1164 			tx_buffer++;
1165 			tx_desc++;
1166 			i++;
1167 			if (unlikely(!i)) {
1168 				i -= tx_ring->count;
1169 				tx_buffer = tx_ring->tx_buffer_info;
1170 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1171 			}
1172 
1173 			/* unmap any remaining paged data */
1174 			if (dma_unmap_len(tx_buffer, len)) {
1175 				dma_unmap_page(tx_ring->dev,
1176 					       dma_unmap_addr(tx_buffer, dma),
1177 					       dma_unmap_len(tx_buffer, len),
1178 					       DMA_TO_DEVICE);
1179 				dma_unmap_len_set(tx_buffer, len, 0);
1180 			}
1181 		}
1182 
1183 		/* move us one more past the eop_desc for start of next pkt */
1184 		tx_buffer++;
1185 		tx_desc++;
1186 		i++;
1187 		if (unlikely(!i)) {
1188 			i -= tx_ring->count;
1189 			tx_buffer = tx_ring->tx_buffer_info;
1190 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1191 		}
1192 
1193 		/* issue prefetch for next Tx descriptor */
1194 		prefetch(tx_desc);
1195 
1196 		/* update budget accounting */
1197 		budget--;
1198 	} while (likely(budget));
1199 
1200 	i += tx_ring->count;
1201 	tx_ring->next_to_clean = i;
1202 	u64_stats_update_begin(&tx_ring->syncp);
1203 	tx_ring->stats.bytes += total_bytes;
1204 	tx_ring->stats.packets += total_packets;
1205 	u64_stats_update_end(&tx_ring->syncp);
1206 	q_vector->tx.total_bytes += total_bytes;
1207 	q_vector->tx.total_packets += total_packets;
1208 	adapter->tx_ipsec += total_ipsec;
1209 
1210 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1211 		/* schedule immediate reset if we believe we hung */
1212 		struct ixgbe_hw *hw = &adapter->hw;
1213 		e_err(drv, "Detected Tx Unit Hang %s\n"
1214 			"  Tx Queue             <%d>\n"
1215 			"  TDH, TDT             <%x>, <%x>\n"
1216 			"  next_to_use          <%x>\n"
1217 			"  next_to_clean        <%x>\n"
1218 			"tx_buffer_info[next_to_clean]\n"
1219 			"  time_stamp           <%lx>\n"
1220 			"  jiffies              <%lx>\n",
1221 			ring_is_xdp(tx_ring) ? "(XDP)" : "",
1222 			tx_ring->queue_index,
1223 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1224 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1225 			tx_ring->next_to_use, i,
1226 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1227 
1228 		if (!ring_is_xdp(tx_ring))
1229 			netif_stop_subqueue(tx_ring->netdev,
1230 					    tx_ring->queue_index);
1231 
1232 		e_info(probe,
1233 		       "tx hang %d detected on queue %d, resetting adapter\n",
1234 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1235 
1236 		/* schedule immediate reset if we believe we hung */
1237 		ixgbe_tx_timeout_reset(adapter);
1238 
1239 		/* the adapter is about to reset, no point in enabling stuff */
1240 		return true;
1241 	}
1242 
1243 	if (ring_is_xdp(tx_ring))
1244 		return !!budget;
1245 
1246 	netdev_tx_completed_queue(txring_txq(tx_ring),
1247 				  total_packets, total_bytes);
1248 
1249 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1250 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1251 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1252 		/* Make sure that anybody stopping the queue after this
1253 		 * sees the new next_to_clean.
1254 		 */
1255 		smp_mb();
1256 		if (__netif_subqueue_stopped(tx_ring->netdev,
1257 					     tx_ring->queue_index)
1258 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1259 			netif_wake_subqueue(tx_ring->netdev,
1260 					    tx_ring->queue_index);
1261 			++tx_ring->tx_stats.restart_queue;
1262 		}
1263 	}
1264 
1265 	return !!budget;
1266 }
1267 
1268 #ifdef CONFIG_IXGBE_DCA
1269 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1270 				struct ixgbe_ring *tx_ring,
1271 				int cpu)
1272 {
1273 	struct ixgbe_hw *hw = &adapter->hw;
1274 	u32 txctrl = 0;
1275 	u16 reg_offset;
1276 
1277 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1278 		txctrl = dca3_get_tag(tx_ring->dev, cpu);
1279 
1280 	switch (hw->mac.type) {
1281 	case ixgbe_mac_82598EB:
1282 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1283 		break;
1284 	case ixgbe_mac_82599EB:
1285 	case ixgbe_mac_X540:
1286 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1287 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1288 		break;
1289 	default:
1290 		/* for unknown hardware do not write register */
1291 		return;
1292 	}
1293 
1294 	/*
1295 	 * We can enable relaxed ordering for reads, but not writes when
1296 	 * DCA is enabled.  This is due to a known issue in some chipsets
1297 	 * which will cause the DCA tag to be cleared.
1298 	 */
1299 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1300 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1301 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1302 
1303 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1304 }
1305 
1306 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1307 				struct ixgbe_ring *rx_ring,
1308 				int cpu)
1309 {
1310 	struct ixgbe_hw *hw = &adapter->hw;
1311 	u32 rxctrl = 0;
1312 	u8 reg_idx = rx_ring->reg_idx;
1313 
1314 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1315 		rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1316 
1317 	switch (hw->mac.type) {
1318 	case ixgbe_mac_82599EB:
1319 	case ixgbe_mac_X540:
1320 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1321 		break;
1322 	default:
1323 		break;
1324 	}
1325 
1326 	/*
1327 	 * We can enable relaxed ordering for reads, but not writes when
1328 	 * DCA is enabled.  This is due to a known issue in some chipsets
1329 	 * which will cause the DCA tag to be cleared.
1330 	 */
1331 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1332 		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1333 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1334 
1335 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1336 }
1337 
1338 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1339 {
1340 	struct ixgbe_adapter *adapter = q_vector->adapter;
1341 	struct ixgbe_ring *ring;
1342 	int cpu = get_cpu();
1343 
1344 	if (q_vector->cpu == cpu)
1345 		goto out_no_update;
1346 
1347 	ixgbe_for_each_ring(ring, q_vector->tx)
1348 		ixgbe_update_tx_dca(adapter, ring, cpu);
1349 
1350 	ixgbe_for_each_ring(ring, q_vector->rx)
1351 		ixgbe_update_rx_dca(adapter, ring, cpu);
1352 
1353 	q_vector->cpu = cpu;
1354 out_no_update:
1355 	put_cpu();
1356 }
1357 
1358 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1359 {
1360 	int i;
1361 
1362 	/* always use CB2 mode, difference is masked in the CB driver */
1363 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1364 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1365 				IXGBE_DCA_CTRL_DCA_MODE_CB2);
1366 	else
1367 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1368 				IXGBE_DCA_CTRL_DCA_DISABLE);
1369 
1370 	for (i = 0; i < adapter->num_q_vectors; i++) {
1371 		adapter->q_vector[i]->cpu = -1;
1372 		ixgbe_update_dca(adapter->q_vector[i]);
1373 	}
1374 }
1375 
1376 static int __ixgbe_notify_dca(struct device *dev, void *data)
1377 {
1378 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1379 	unsigned long event = *(unsigned long *)data;
1380 
1381 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1382 		return 0;
1383 
1384 	switch (event) {
1385 	case DCA_PROVIDER_ADD:
1386 		/* if we're already enabled, don't do it again */
1387 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1388 			break;
1389 		if (dca_add_requester(dev) == 0) {
1390 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1391 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1392 					IXGBE_DCA_CTRL_DCA_MODE_CB2);
1393 			break;
1394 		}
1395 		/* fall through - DCA is disabled. */
1396 	case DCA_PROVIDER_REMOVE:
1397 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1398 			dca_remove_requester(dev);
1399 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1400 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1401 					IXGBE_DCA_CTRL_DCA_DISABLE);
1402 		}
1403 		break;
1404 	}
1405 
1406 	return 0;
1407 }
1408 
1409 #endif /* CONFIG_IXGBE_DCA */
1410 
1411 #define IXGBE_RSS_L4_TYPES_MASK \
1412 	((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1413 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1414 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1415 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1416 
1417 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1418 				 union ixgbe_adv_rx_desc *rx_desc,
1419 				 struct sk_buff *skb)
1420 {
1421 	u16 rss_type;
1422 
1423 	if (!(ring->netdev->features & NETIF_F_RXHASH))
1424 		return;
1425 
1426 	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1427 		   IXGBE_RXDADV_RSSTYPE_MASK;
1428 
1429 	if (!rss_type)
1430 		return;
1431 
1432 	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1433 		     (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1434 		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1435 }
1436 
1437 #ifdef IXGBE_FCOE
1438 /**
1439  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1440  * @ring: structure containing ring specific data
1441  * @rx_desc: advanced rx descriptor
1442  *
1443  * Returns : true if it is FCoE pkt
1444  */
1445 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1446 				    union ixgbe_adv_rx_desc *rx_desc)
1447 {
1448 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1449 
1450 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1451 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1452 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1453 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1454 }
1455 
1456 #endif /* IXGBE_FCOE */
1457 /**
1458  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1459  * @ring: structure containing ring specific data
1460  * @rx_desc: current Rx descriptor being processed
1461  * @skb: skb currently being received and modified
1462  **/
1463 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1464 				     union ixgbe_adv_rx_desc *rx_desc,
1465 				     struct sk_buff *skb)
1466 {
1467 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1468 	bool encap_pkt = false;
1469 
1470 	skb_checksum_none_assert(skb);
1471 
1472 	/* Rx csum disabled */
1473 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1474 		return;
1475 
1476 	/* check for VXLAN and Geneve packets */
1477 	if (pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) {
1478 		encap_pkt = true;
1479 		skb->encapsulation = 1;
1480 	}
1481 
1482 	/* if IP and error */
1483 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1484 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1485 		ring->rx_stats.csum_err++;
1486 		return;
1487 	}
1488 
1489 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1490 		return;
1491 
1492 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1493 		/*
1494 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1495 		 * checksum errors.
1496 		 */
1497 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1498 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1499 			return;
1500 
1501 		ring->rx_stats.csum_err++;
1502 		return;
1503 	}
1504 
1505 	/* It must be a TCP or UDP packet with a valid checksum */
1506 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1507 	if (encap_pkt) {
1508 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1509 			return;
1510 
1511 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1512 			skb->ip_summed = CHECKSUM_NONE;
1513 			return;
1514 		}
1515 		/* If we checked the outer header let the stack know */
1516 		skb->csum_level = 1;
1517 	}
1518 }
1519 
1520 static inline unsigned int ixgbe_rx_offset(struct ixgbe_ring *rx_ring)
1521 {
1522 	return ring_uses_build_skb(rx_ring) ? IXGBE_SKB_PAD : 0;
1523 }
1524 
1525 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1526 				    struct ixgbe_rx_buffer *bi)
1527 {
1528 	struct page *page = bi->page;
1529 	dma_addr_t dma;
1530 
1531 	/* since we are recycling buffers we should seldom need to alloc */
1532 	if (likely(page))
1533 		return true;
1534 
1535 	/* alloc new page for storage */
1536 	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1537 	if (unlikely(!page)) {
1538 		rx_ring->rx_stats.alloc_rx_page_failed++;
1539 		return false;
1540 	}
1541 
1542 	/* map page for use */
1543 	dma = dma_map_page_attrs(rx_ring->dev, page, 0,
1544 				 ixgbe_rx_pg_size(rx_ring),
1545 				 DMA_FROM_DEVICE,
1546 				 IXGBE_RX_DMA_ATTR);
1547 
1548 	/*
1549 	 * if mapping failed free memory back to system since
1550 	 * there isn't much point in holding memory we can't use
1551 	 */
1552 	if (dma_mapping_error(rx_ring->dev, dma)) {
1553 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1554 
1555 		rx_ring->rx_stats.alloc_rx_page_failed++;
1556 		return false;
1557 	}
1558 
1559 	bi->dma = dma;
1560 	bi->page = page;
1561 	bi->page_offset = ixgbe_rx_offset(rx_ring);
1562 	page_ref_add(page, USHRT_MAX - 1);
1563 	bi->pagecnt_bias = USHRT_MAX;
1564 	rx_ring->rx_stats.alloc_rx_page++;
1565 
1566 	return true;
1567 }
1568 
1569 /**
1570  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1571  * @rx_ring: ring to place buffers on
1572  * @cleaned_count: number of buffers to replace
1573  **/
1574 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1575 {
1576 	union ixgbe_adv_rx_desc *rx_desc;
1577 	struct ixgbe_rx_buffer *bi;
1578 	u16 i = rx_ring->next_to_use;
1579 	u16 bufsz;
1580 
1581 	/* nothing to do */
1582 	if (!cleaned_count)
1583 		return;
1584 
1585 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1586 	bi = &rx_ring->rx_buffer_info[i];
1587 	i -= rx_ring->count;
1588 
1589 	bufsz = ixgbe_rx_bufsz(rx_ring);
1590 
1591 	do {
1592 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1593 			break;
1594 
1595 		/* sync the buffer for use by the device */
1596 		dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
1597 						 bi->page_offset, bufsz,
1598 						 DMA_FROM_DEVICE);
1599 
1600 		/*
1601 		 * Refresh the desc even if buffer_addrs didn't change
1602 		 * because each write-back erases this info.
1603 		 */
1604 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1605 
1606 		rx_desc++;
1607 		bi++;
1608 		i++;
1609 		if (unlikely(!i)) {
1610 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1611 			bi = rx_ring->rx_buffer_info;
1612 			i -= rx_ring->count;
1613 		}
1614 
1615 		/* clear the length for the next_to_use descriptor */
1616 		rx_desc->wb.upper.length = 0;
1617 
1618 		cleaned_count--;
1619 	} while (cleaned_count);
1620 
1621 	i += rx_ring->count;
1622 
1623 	if (rx_ring->next_to_use != i) {
1624 		rx_ring->next_to_use = i;
1625 
1626 		/* update next to alloc since we have filled the ring */
1627 		rx_ring->next_to_alloc = i;
1628 
1629 		/* Force memory writes to complete before letting h/w
1630 		 * know there are new descriptors to fetch.  (Only
1631 		 * applicable for weak-ordered memory model archs,
1632 		 * such as IA-64).
1633 		 */
1634 		wmb();
1635 		writel(i, rx_ring->tail);
1636 	}
1637 }
1638 
1639 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1640 				   struct sk_buff *skb)
1641 {
1642 	u16 hdr_len = skb_headlen(skb);
1643 
1644 	/* set gso_size to avoid messing up TCP MSS */
1645 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1646 						 IXGBE_CB(skb)->append_cnt);
1647 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1648 }
1649 
1650 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1651 				   struct sk_buff *skb)
1652 {
1653 	/* if append_cnt is 0 then frame is not RSC */
1654 	if (!IXGBE_CB(skb)->append_cnt)
1655 		return;
1656 
1657 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1658 	rx_ring->rx_stats.rsc_flush++;
1659 
1660 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1661 
1662 	/* gso_size is computed using append_cnt so always clear it last */
1663 	IXGBE_CB(skb)->append_cnt = 0;
1664 }
1665 
1666 /**
1667  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1668  * @rx_ring: rx descriptor ring packet is being transacted on
1669  * @rx_desc: pointer to the EOP Rx descriptor
1670  * @skb: pointer to current skb being populated
1671  *
1672  * This function checks the ring, descriptor, and packet information in
1673  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1674  * other fields within the skb.
1675  **/
1676 static void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1677 				     union ixgbe_adv_rx_desc *rx_desc,
1678 				     struct sk_buff *skb)
1679 {
1680 	struct net_device *dev = rx_ring->netdev;
1681 	u32 flags = rx_ring->q_vector->adapter->flags;
1682 
1683 	ixgbe_update_rsc_stats(rx_ring, skb);
1684 
1685 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1686 
1687 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1688 
1689 	if (unlikely(flags & IXGBE_FLAG_RX_HWTSTAMP_ENABLED))
1690 		ixgbe_ptp_rx_hwtstamp(rx_ring, rx_desc, skb);
1691 
1692 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1693 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1694 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1695 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1696 	}
1697 
1698 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_SECP))
1699 		ixgbe_ipsec_rx(rx_ring, rx_desc, skb);
1700 
1701 	/* record Rx queue, or update MACVLAN statistics */
1702 	if (netif_is_ixgbe(dev))
1703 		skb_record_rx_queue(skb, rx_ring->queue_index);
1704 	else
1705 		macvlan_count_rx(netdev_priv(dev), skb->len + ETH_HLEN, true,
1706 				 false);
1707 
1708 	skb->protocol = eth_type_trans(skb, dev);
1709 }
1710 
1711 static void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1712 			 struct sk_buff *skb)
1713 {
1714 	napi_gro_receive(&q_vector->napi, skb);
1715 }
1716 
1717 /**
1718  * ixgbe_is_non_eop - process handling of non-EOP buffers
1719  * @rx_ring: Rx ring being processed
1720  * @rx_desc: Rx descriptor for current buffer
1721  * @skb: Current socket buffer containing buffer in progress
1722  *
1723  * This function updates next to clean.  If the buffer is an EOP buffer
1724  * this function exits returning false, otherwise it will place the
1725  * sk_buff in the next buffer to be chained and return true indicating
1726  * that this is in fact a non-EOP buffer.
1727  **/
1728 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1729 			     union ixgbe_adv_rx_desc *rx_desc,
1730 			     struct sk_buff *skb)
1731 {
1732 	u32 ntc = rx_ring->next_to_clean + 1;
1733 
1734 	/* fetch, update, and store next to clean */
1735 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1736 	rx_ring->next_to_clean = ntc;
1737 
1738 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1739 
1740 	/* update RSC append count if present */
1741 	if (ring_is_rsc_enabled(rx_ring)) {
1742 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1743 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1744 
1745 		if (unlikely(rsc_enabled)) {
1746 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1747 
1748 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1749 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1750 
1751 			/* update ntc based on RSC value */
1752 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1753 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1754 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1755 		}
1756 	}
1757 
1758 	/* if we are the last buffer then there is nothing else to do */
1759 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1760 		return false;
1761 
1762 	/* place skb in next buffer to be received */
1763 	rx_ring->rx_buffer_info[ntc].skb = skb;
1764 	rx_ring->rx_stats.non_eop_descs++;
1765 
1766 	return true;
1767 }
1768 
1769 /**
1770  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1771  * @rx_ring: rx descriptor ring packet is being transacted on
1772  * @skb: pointer to current skb being adjusted
1773  *
1774  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1775  * main difference between this version and the original function is that
1776  * this function can make several assumptions about the state of things
1777  * that allow for significant optimizations versus the standard function.
1778  * As a result we can do things like drop a frag and maintain an accurate
1779  * truesize for the skb.
1780  */
1781 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1782 			    struct sk_buff *skb)
1783 {
1784 	struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1785 	unsigned char *va;
1786 	unsigned int pull_len;
1787 
1788 	/*
1789 	 * it is valid to use page_address instead of kmap since we are
1790 	 * working with pages allocated out of the lomem pool per
1791 	 * alloc_page(GFP_ATOMIC)
1792 	 */
1793 	va = skb_frag_address(frag);
1794 
1795 	/*
1796 	 * we need the header to contain the greater of either ETH_HLEN or
1797 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1798 	 */
1799 	pull_len = eth_get_headlen(va, IXGBE_RX_HDR_SIZE);
1800 
1801 	/* align pull length to size of long to optimize memcpy performance */
1802 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1803 
1804 	/* update all of the pointers */
1805 	skb_frag_size_sub(frag, pull_len);
1806 	frag->page_offset += pull_len;
1807 	skb->data_len -= pull_len;
1808 	skb->tail += pull_len;
1809 }
1810 
1811 /**
1812  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1813  * @rx_ring: rx descriptor ring packet is being transacted on
1814  * @skb: pointer to current skb being updated
1815  *
1816  * This function provides a basic DMA sync up for the first fragment of an
1817  * skb.  The reason for doing this is that the first fragment cannot be
1818  * unmapped until we have reached the end of packet descriptor for a buffer
1819  * chain.
1820  */
1821 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1822 				struct sk_buff *skb)
1823 {
1824 	/* if the page was released unmap it, else just sync our portion */
1825 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1826 		dma_unmap_page_attrs(rx_ring->dev, IXGBE_CB(skb)->dma,
1827 				     ixgbe_rx_pg_size(rx_ring),
1828 				     DMA_FROM_DEVICE,
1829 				     IXGBE_RX_DMA_ATTR);
1830 	} else if (ring_uses_build_skb(rx_ring)) {
1831 		unsigned long offset = (unsigned long)(skb->data) & ~PAGE_MASK;
1832 
1833 		dma_sync_single_range_for_cpu(rx_ring->dev,
1834 					      IXGBE_CB(skb)->dma,
1835 					      offset,
1836 					      skb_headlen(skb),
1837 					      DMA_FROM_DEVICE);
1838 	} else {
1839 		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
1840 
1841 		dma_sync_single_range_for_cpu(rx_ring->dev,
1842 					      IXGBE_CB(skb)->dma,
1843 					      frag->page_offset,
1844 					      skb_frag_size(frag),
1845 					      DMA_FROM_DEVICE);
1846 	}
1847 }
1848 
1849 /**
1850  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1851  * @rx_ring: rx descriptor ring packet is being transacted on
1852  * @rx_desc: pointer to the EOP Rx descriptor
1853  * @skb: pointer to current skb being fixed
1854  *
1855  * Check if the skb is valid in the XDP case it will be an error pointer.
1856  * Return true in this case to abort processing and advance to next
1857  * descriptor.
1858  *
1859  * Check for corrupted packet headers caused by senders on the local L2
1860  * embedded NIC switch not setting up their Tx Descriptors right.  These
1861  * should be very rare.
1862  *
1863  * Also address the case where we are pulling data in on pages only
1864  * and as such no data is present in the skb header.
1865  *
1866  * In addition if skb is not at least 60 bytes we need to pad it so that
1867  * it is large enough to qualify as a valid Ethernet frame.
1868  *
1869  * Returns true if an error was encountered and skb was freed.
1870  **/
1871 static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1872 				  union ixgbe_adv_rx_desc *rx_desc,
1873 				  struct sk_buff *skb)
1874 {
1875 	struct net_device *netdev = rx_ring->netdev;
1876 
1877 	/* XDP packets use error pointer so abort at this point */
1878 	if (IS_ERR(skb))
1879 		return true;
1880 
1881 	/* Verify netdev is present, and that packet does not have any
1882 	 * errors that would be unacceptable to the netdev.
1883 	 */
1884 	if (!netdev ||
1885 	    (unlikely(ixgbe_test_staterr(rx_desc,
1886 					 IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1887 	     !(netdev->features & NETIF_F_RXALL)))) {
1888 		dev_kfree_skb_any(skb);
1889 		return true;
1890 	}
1891 
1892 	/* place header in linear portion of buffer */
1893 	if (!skb_headlen(skb))
1894 		ixgbe_pull_tail(rx_ring, skb);
1895 
1896 #ifdef IXGBE_FCOE
1897 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1898 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1899 		return false;
1900 
1901 #endif
1902 	/* if eth_skb_pad returns an error the skb was freed */
1903 	if (eth_skb_pad(skb))
1904 		return true;
1905 
1906 	return false;
1907 }
1908 
1909 /**
1910  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1911  * @rx_ring: rx descriptor ring to store buffers on
1912  * @old_buff: donor buffer to have page reused
1913  *
1914  * Synchronizes page for reuse by the adapter
1915  **/
1916 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1917 				struct ixgbe_rx_buffer *old_buff)
1918 {
1919 	struct ixgbe_rx_buffer *new_buff;
1920 	u16 nta = rx_ring->next_to_alloc;
1921 
1922 	new_buff = &rx_ring->rx_buffer_info[nta];
1923 
1924 	/* update, and store next to alloc */
1925 	nta++;
1926 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1927 
1928 	/* Transfer page from old buffer to new buffer.
1929 	 * Move each member individually to avoid possible store
1930 	 * forwarding stalls and unnecessary copy of skb.
1931 	 */
1932 	new_buff->dma		= old_buff->dma;
1933 	new_buff->page		= old_buff->page;
1934 	new_buff->page_offset	= old_buff->page_offset;
1935 	new_buff->pagecnt_bias	= old_buff->pagecnt_bias;
1936 }
1937 
1938 static inline bool ixgbe_page_is_reserved(struct page *page)
1939 {
1940 	return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
1941 }
1942 
1943 static bool ixgbe_can_reuse_rx_page(struct ixgbe_rx_buffer *rx_buffer)
1944 {
1945 	unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
1946 	struct page *page = rx_buffer->page;
1947 
1948 	/* avoid re-using remote pages */
1949 	if (unlikely(ixgbe_page_is_reserved(page)))
1950 		return false;
1951 
1952 #if (PAGE_SIZE < 8192)
1953 	/* if we are only owner of page we can reuse it */
1954 	if (unlikely((page_ref_count(page) - pagecnt_bias) > 1))
1955 		return false;
1956 #else
1957 	/* The last offset is a bit aggressive in that we assume the
1958 	 * worst case of FCoE being enabled and using a 3K buffer.
1959 	 * However this should have minimal impact as the 1K extra is
1960 	 * still less than one buffer in size.
1961 	 */
1962 #define IXGBE_LAST_OFFSET \
1963 	(SKB_WITH_OVERHEAD(PAGE_SIZE) - IXGBE_RXBUFFER_3K)
1964 	if (rx_buffer->page_offset > IXGBE_LAST_OFFSET)
1965 		return false;
1966 #endif
1967 
1968 	/* If we have drained the page fragment pool we need to update
1969 	 * the pagecnt_bias and page count so that we fully restock the
1970 	 * number of references the driver holds.
1971 	 */
1972 	if (unlikely(pagecnt_bias == 1)) {
1973 		page_ref_add(page, USHRT_MAX - 1);
1974 		rx_buffer->pagecnt_bias = USHRT_MAX;
1975 	}
1976 
1977 	return true;
1978 }
1979 
1980 /**
1981  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1982  * @rx_ring: rx descriptor ring to transact packets on
1983  * @rx_buffer: buffer containing page to add
1984  * @skb: sk_buff to place the data into
1985  * @size: size of data in rx_buffer
1986  *
1987  * This function will add the data contained in rx_buffer->page to the skb.
1988  * This is done either through a direct copy if the data in the buffer is
1989  * less than the skb header size, otherwise it will just attach the page as
1990  * a frag to the skb.
1991  *
1992  * The function will then update the page offset if necessary and return
1993  * true if the buffer can be reused by the adapter.
1994  **/
1995 static void ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1996 			      struct ixgbe_rx_buffer *rx_buffer,
1997 			      struct sk_buff *skb,
1998 			      unsigned int size)
1999 {
2000 #if (PAGE_SIZE < 8192)
2001 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2002 #else
2003 	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
2004 				SKB_DATA_ALIGN(IXGBE_SKB_PAD + size) :
2005 				SKB_DATA_ALIGN(size);
2006 #endif
2007 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
2008 			rx_buffer->page_offset, size, truesize);
2009 #if (PAGE_SIZE < 8192)
2010 	rx_buffer->page_offset ^= truesize;
2011 #else
2012 	rx_buffer->page_offset += truesize;
2013 #endif
2014 }
2015 
2016 static struct ixgbe_rx_buffer *ixgbe_get_rx_buffer(struct ixgbe_ring *rx_ring,
2017 						   union ixgbe_adv_rx_desc *rx_desc,
2018 						   struct sk_buff **skb,
2019 						   const unsigned int size)
2020 {
2021 	struct ixgbe_rx_buffer *rx_buffer;
2022 
2023 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
2024 	prefetchw(rx_buffer->page);
2025 	*skb = rx_buffer->skb;
2026 
2027 	/* Delay unmapping of the first packet. It carries the header
2028 	 * information, HW may still access the header after the writeback.
2029 	 * Only unmap it when EOP is reached
2030 	 */
2031 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)) {
2032 		if (!*skb)
2033 			goto skip_sync;
2034 	} else {
2035 		if (*skb)
2036 			ixgbe_dma_sync_frag(rx_ring, *skb);
2037 	}
2038 
2039 	/* we are reusing so sync this buffer for CPU use */
2040 	dma_sync_single_range_for_cpu(rx_ring->dev,
2041 				      rx_buffer->dma,
2042 				      rx_buffer->page_offset,
2043 				      size,
2044 				      DMA_FROM_DEVICE);
2045 skip_sync:
2046 	rx_buffer->pagecnt_bias--;
2047 
2048 	return rx_buffer;
2049 }
2050 
2051 static void ixgbe_put_rx_buffer(struct ixgbe_ring *rx_ring,
2052 				struct ixgbe_rx_buffer *rx_buffer,
2053 				struct sk_buff *skb)
2054 {
2055 	if (ixgbe_can_reuse_rx_page(rx_buffer)) {
2056 		/* hand second half of page back to the ring */
2057 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2058 	} else {
2059 		if (!IS_ERR(skb) && IXGBE_CB(skb)->dma == rx_buffer->dma) {
2060 			/* the page has been released from the ring */
2061 			IXGBE_CB(skb)->page_released = true;
2062 		} else {
2063 			/* we are not reusing the buffer so unmap it */
2064 			dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
2065 					     ixgbe_rx_pg_size(rx_ring),
2066 					     DMA_FROM_DEVICE,
2067 					     IXGBE_RX_DMA_ATTR);
2068 		}
2069 		__page_frag_cache_drain(rx_buffer->page,
2070 					rx_buffer->pagecnt_bias);
2071 	}
2072 
2073 	/* clear contents of rx_buffer */
2074 	rx_buffer->page = NULL;
2075 	rx_buffer->skb = NULL;
2076 }
2077 
2078 static struct sk_buff *ixgbe_construct_skb(struct ixgbe_ring *rx_ring,
2079 					   struct ixgbe_rx_buffer *rx_buffer,
2080 					   struct xdp_buff *xdp,
2081 					   union ixgbe_adv_rx_desc *rx_desc)
2082 {
2083 	unsigned int size = xdp->data_end - xdp->data;
2084 #if (PAGE_SIZE < 8192)
2085 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2086 #else
2087 	unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end -
2088 					       xdp->data_hard_start);
2089 #endif
2090 	struct sk_buff *skb;
2091 
2092 	/* prefetch first cache line of first page */
2093 	prefetch(xdp->data);
2094 #if L1_CACHE_BYTES < 128
2095 	prefetch(xdp->data + L1_CACHE_BYTES);
2096 #endif
2097 	/* Note, we get here by enabling legacy-rx via:
2098 	 *
2099 	 *    ethtool --set-priv-flags <dev> legacy-rx on
2100 	 *
2101 	 * In this mode, we currently get 0 extra XDP headroom as
2102 	 * opposed to having legacy-rx off, where we process XDP
2103 	 * packets going to stack via ixgbe_build_skb(). The latter
2104 	 * provides us currently with 192 bytes of headroom.
2105 	 *
2106 	 * For ixgbe_construct_skb() mode it means that the
2107 	 * xdp->data_meta will always point to xdp->data, since
2108 	 * the helper cannot expand the head. Should this ever
2109 	 * change in future for legacy-rx mode on, then lets also
2110 	 * add xdp->data_meta handling here.
2111 	 */
2112 
2113 	/* allocate a skb to store the frags */
2114 	skb = napi_alloc_skb(&rx_ring->q_vector->napi, IXGBE_RX_HDR_SIZE);
2115 	if (unlikely(!skb))
2116 		return NULL;
2117 
2118 	if (size > IXGBE_RX_HDR_SIZE) {
2119 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2120 			IXGBE_CB(skb)->dma = rx_buffer->dma;
2121 
2122 		skb_add_rx_frag(skb, 0, rx_buffer->page,
2123 				xdp->data - page_address(rx_buffer->page),
2124 				size, truesize);
2125 #if (PAGE_SIZE < 8192)
2126 		rx_buffer->page_offset ^= truesize;
2127 #else
2128 		rx_buffer->page_offset += truesize;
2129 #endif
2130 	} else {
2131 		memcpy(__skb_put(skb, size),
2132 		       xdp->data, ALIGN(size, sizeof(long)));
2133 		rx_buffer->pagecnt_bias++;
2134 	}
2135 
2136 	return skb;
2137 }
2138 
2139 static struct sk_buff *ixgbe_build_skb(struct ixgbe_ring *rx_ring,
2140 				       struct ixgbe_rx_buffer *rx_buffer,
2141 				       struct xdp_buff *xdp,
2142 				       union ixgbe_adv_rx_desc *rx_desc)
2143 {
2144 	unsigned int metasize = xdp->data - xdp->data_meta;
2145 #if (PAGE_SIZE < 8192)
2146 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2147 #else
2148 	unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
2149 				SKB_DATA_ALIGN(xdp->data_end -
2150 					       xdp->data_hard_start);
2151 #endif
2152 	struct sk_buff *skb;
2153 
2154 	/* Prefetch first cache line of first page. If xdp->data_meta
2155 	 * is unused, this points extactly as xdp->data, otherwise we
2156 	 * likely have a consumer accessing first few bytes of meta
2157 	 * data, and then actual data.
2158 	 */
2159 	prefetch(xdp->data_meta);
2160 #if L1_CACHE_BYTES < 128
2161 	prefetch(xdp->data_meta + L1_CACHE_BYTES);
2162 #endif
2163 
2164 	/* build an skb to around the page buffer */
2165 	skb = build_skb(xdp->data_hard_start, truesize);
2166 	if (unlikely(!skb))
2167 		return NULL;
2168 
2169 	/* update pointers within the skb to store the data */
2170 	skb_reserve(skb, xdp->data - xdp->data_hard_start);
2171 	__skb_put(skb, xdp->data_end - xdp->data);
2172 	if (metasize)
2173 		skb_metadata_set(skb, metasize);
2174 
2175 	/* record DMA address if this is the start of a chain of buffers */
2176 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2177 		IXGBE_CB(skb)->dma = rx_buffer->dma;
2178 
2179 	/* update buffer offset */
2180 #if (PAGE_SIZE < 8192)
2181 	rx_buffer->page_offset ^= truesize;
2182 #else
2183 	rx_buffer->page_offset += truesize;
2184 #endif
2185 
2186 	return skb;
2187 }
2188 
2189 #define IXGBE_XDP_PASS		0
2190 #define IXGBE_XDP_CONSUMED	BIT(0)
2191 #define IXGBE_XDP_TX		BIT(1)
2192 #define IXGBE_XDP_REDIR		BIT(2)
2193 
2194 static int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
2195 			       struct xdp_frame *xdpf);
2196 
2197 static struct sk_buff *ixgbe_run_xdp(struct ixgbe_adapter *adapter,
2198 				     struct ixgbe_ring *rx_ring,
2199 				     struct xdp_buff *xdp)
2200 {
2201 	int err, result = IXGBE_XDP_PASS;
2202 	struct bpf_prog *xdp_prog;
2203 	struct xdp_frame *xdpf;
2204 	u32 act;
2205 
2206 	rcu_read_lock();
2207 	xdp_prog = READ_ONCE(rx_ring->xdp_prog);
2208 
2209 	if (!xdp_prog)
2210 		goto xdp_out;
2211 
2212 	prefetchw(xdp->data_hard_start); /* xdp_frame write */
2213 
2214 	act = bpf_prog_run_xdp(xdp_prog, xdp);
2215 	switch (act) {
2216 	case XDP_PASS:
2217 		break;
2218 	case XDP_TX:
2219 		xdpf = convert_to_xdp_frame(xdp);
2220 		if (unlikely(!xdpf)) {
2221 			result = IXGBE_XDP_CONSUMED;
2222 			break;
2223 		}
2224 		result = ixgbe_xmit_xdp_ring(adapter, xdpf);
2225 		break;
2226 	case XDP_REDIRECT:
2227 		err = xdp_do_redirect(adapter->netdev, xdp, xdp_prog);
2228 		if (!err)
2229 			result = IXGBE_XDP_REDIR;
2230 		else
2231 			result = IXGBE_XDP_CONSUMED;
2232 		break;
2233 	default:
2234 		bpf_warn_invalid_xdp_action(act);
2235 		/* fallthrough */
2236 	case XDP_ABORTED:
2237 		trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
2238 		/* fallthrough -- handle aborts by dropping packet */
2239 	case XDP_DROP:
2240 		result = IXGBE_XDP_CONSUMED;
2241 		break;
2242 	}
2243 xdp_out:
2244 	rcu_read_unlock();
2245 	return ERR_PTR(-result);
2246 }
2247 
2248 static void ixgbe_rx_buffer_flip(struct ixgbe_ring *rx_ring,
2249 				 struct ixgbe_rx_buffer *rx_buffer,
2250 				 unsigned int size)
2251 {
2252 #if (PAGE_SIZE < 8192)
2253 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2254 
2255 	rx_buffer->page_offset ^= truesize;
2256 #else
2257 	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
2258 				SKB_DATA_ALIGN(IXGBE_SKB_PAD + size) :
2259 				SKB_DATA_ALIGN(size);
2260 
2261 	rx_buffer->page_offset += truesize;
2262 #endif
2263 }
2264 
2265 /**
2266  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2267  * @q_vector: structure containing interrupt and ring information
2268  * @rx_ring: rx descriptor ring to transact packets on
2269  * @budget: Total limit on number of packets to process
2270  *
2271  * This function provides a "bounce buffer" approach to Rx interrupt
2272  * processing.  The advantage to this is that on systems that have
2273  * expensive overhead for IOMMU access this provides a means of avoiding
2274  * it by maintaining the mapping of the page to the syste.
2275  *
2276  * Returns amount of work completed
2277  **/
2278 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2279 			       struct ixgbe_ring *rx_ring,
2280 			       const int budget)
2281 {
2282 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
2283 	struct ixgbe_adapter *adapter = q_vector->adapter;
2284 #ifdef IXGBE_FCOE
2285 	int ddp_bytes;
2286 	unsigned int mss = 0;
2287 #endif /* IXGBE_FCOE */
2288 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2289 	unsigned int xdp_xmit = 0;
2290 	struct xdp_buff xdp;
2291 
2292 	xdp.rxq = &rx_ring->xdp_rxq;
2293 
2294 	while (likely(total_rx_packets < budget)) {
2295 		union ixgbe_adv_rx_desc *rx_desc;
2296 		struct ixgbe_rx_buffer *rx_buffer;
2297 		struct sk_buff *skb;
2298 		unsigned int size;
2299 
2300 		/* return some buffers to hardware, one at a time is too slow */
2301 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2302 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2303 			cleaned_count = 0;
2304 		}
2305 
2306 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2307 		size = le16_to_cpu(rx_desc->wb.upper.length);
2308 		if (!size)
2309 			break;
2310 
2311 		/* This memory barrier is needed to keep us from reading
2312 		 * any other fields out of the rx_desc until we know the
2313 		 * descriptor has been written back
2314 		 */
2315 		dma_rmb();
2316 
2317 		rx_buffer = ixgbe_get_rx_buffer(rx_ring, rx_desc, &skb, size);
2318 
2319 		/* retrieve a buffer from the ring */
2320 		if (!skb) {
2321 			xdp.data = page_address(rx_buffer->page) +
2322 				   rx_buffer->page_offset;
2323 			xdp.data_meta = xdp.data;
2324 			xdp.data_hard_start = xdp.data -
2325 					      ixgbe_rx_offset(rx_ring);
2326 			xdp.data_end = xdp.data + size;
2327 
2328 			skb = ixgbe_run_xdp(adapter, rx_ring, &xdp);
2329 		}
2330 
2331 		if (IS_ERR(skb)) {
2332 			unsigned int xdp_res = -PTR_ERR(skb);
2333 
2334 			if (xdp_res & (IXGBE_XDP_TX | IXGBE_XDP_REDIR)) {
2335 				xdp_xmit |= xdp_res;
2336 				ixgbe_rx_buffer_flip(rx_ring, rx_buffer, size);
2337 			} else {
2338 				rx_buffer->pagecnt_bias++;
2339 			}
2340 			total_rx_packets++;
2341 			total_rx_bytes += size;
2342 		} else if (skb) {
2343 			ixgbe_add_rx_frag(rx_ring, rx_buffer, skb, size);
2344 		} else if (ring_uses_build_skb(rx_ring)) {
2345 			skb = ixgbe_build_skb(rx_ring, rx_buffer,
2346 					      &xdp, rx_desc);
2347 		} else {
2348 			skb = ixgbe_construct_skb(rx_ring, rx_buffer,
2349 						  &xdp, rx_desc);
2350 		}
2351 
2352 		/* exit if we failed to retrieve a buffer */
2353 		if (!skb) {
2354 			rx_ring->rx_stats.alloc_rx_buff_failed++;
2355 			rx_buffer->pagecnt_bias++;
2356 			break;
2357 		}
2358 
2359 		ixgbe_put_rx_buffer(rx_ring, rx_buffer, skb);
2360 		cleaned_count++;
2361 
2362 		/* place incomplete frames back on ring for completion */
2363 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2364 			continue;
2365 
2366 		/* verify the packet layout is correct */
2367 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2368 			continue;
2369 
2370 		/* probably a little skewed due to removing CRC */
2371 		total_rx_bytes += skb->len;
2372 
2373 		/* populate checksum, timestamp, VLAN, and protocol */
2374 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2375 
2376 #ifdef IXGBE_FCOE
2377 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2378 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2379 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2380 			/* include DDPed FCoE data */
2381 			if (ddp_bytes > 0) {
2382 				if (!mss) {
2383 					mss = rx_ring->netdev->mtu -
2384 						sizeof(struct fcoe_hdr) -
2385 						sizeof(struct fc_frame_header) -
2386 						sizeof(struct fcoe_crc_eof);
2387 					if (mss > 512)
2388 						mss &= ~511;
2389 				}
2390 				total_rx_bytes += ddp_bytes;
2391 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2392 								 mss);
2393 			}
2394 			if (!ddp_bytes) {
2395 				dev_kfree_skb_any(skb);
2396 				continue;
2397 			}
2398 		}
2399 
2400 #endif /* IXGBE_FCOE */
2401 		ixgbe_rx_skb(q_vector, skb);
2402 
2403 		/* update budget accounting */
2404 		total_rx_packets++;
2405 	}
2406 
2407 	if (xdp_xmit & IXGBE_XDP_REDIR)
2408 		xdp_do_flush_map();
2409 
2410 	if (xdp_xmit & IXGBE_XDP_TX) {
2411 		struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
2412 
2413 		/* Force memory writes to complete before letting h/w
2414 		 * know there are new descriptors to fetch.
2415 		 */
2416 		wmb();
2417 		writel(ring->next_to_use, ring->tail);
2418 	}
2419 
2420 	u64_stats_update_begin(&rx_ring->syncp);
2421 	rx_ring->stats.packets += total_rx_packets;
2422 	rx_ring->stats.bytes += total_rx_bytes;
2423 	u64_stats_update_end(&rx_ring->syncp);
2424 	q_vector->rx.total_packets += total_rx_packets;
2425 	q_vector->rx.total_bytes += total_rx_bytes;
2426 
2427 	return total_rx_packets;
2428 }
2429 
2430 /**
2431  * ixgbe_configure_msix - Configure MSI-X hardware
2432  * @adapter: board private structure
2433  *
2434  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2435  * interrupts.
2436  **/
2437 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2438 {
2439 	struct ixgbe_q_vector *q_vector;
2440 	int v_idx;
2441 	u32 mask;
2442 
2443 	/* Populate MSIX to EITR Select */
2444 	if (adapter->num_vfs > 32) {
2445 		u32 eitrsel = BIT(adapter->num_vfs - 32) - 1;
2446 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2447 	}
2448 
2449 	/*
2450 	 * Populate the IVAR table and set the ITR values to the
2451 	 * corresponding register.
2452 	 */
2453 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2454 		struct ixgbe_ring *ring;
2455 		q_vector = adapter->q_vector[v_idx];
2456 
2457 		ixgbe_for_each_ring(ring, q_vector->rx)
2458 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2459 
2460 		ixgbe_for_each_ring(ring, q_vector->tx)
2461 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2462 
2463 		ixgbe_write_eitr(q_vector);
2464 	}
2465 
2466 	switch (adapter->hw.mac.type) {
2467 	case ixgbe_mac_82598EB:
2468 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2469 			       v_idx);
2470 		break;
2471 	case ixgbe_mac_82599EB:
2472 	case ixgbe_mac_X540:
2473 	case ixgbe_mac_X550:
2474 	case ixgbe_mac_X550EM_x:
2475 	case ixgbe_mac_x550em_a:
2476 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2477 		break;
2478 	default:
2479 		break;
2480 	}
2481 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2482 
2483 	/* set up to autoclear timer, and the vectors */
2484 	mask = IXGBE_EIMS_ENABLE_MASK;
2485 	mask &= ~(IXGBE_EIMS_OTHER |
2486 		  IXGBE_EIMS_MAILBOX |
2487 		  IXGBE_EIMS_LSC);
2488 
2489 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2490 }
2491 
2492 /**
2493  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2494  * @q_vector: structure containing interrupt and ring information
2495  * @ring_container: structure containing ring performance data
2496  *
2497  *      Stores a new ITR value based on packets and byte
2498  *      counts during the last interrupt.  The advantage of per interrupt
2499  *      computation is faster updates and more accurate ITR for the current
2500  *      traffic pattern.  Constants in this function were computed
2501  *      based on theoretical maximum wire speed and thresholds were set based
2502  *      on testing data as well as attempting to minimize response time
2503  *      while increasing bulk throughput.
2504  **/
2505 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2506 			     struct ixgbe_ring_container *ring_container)
2507 {
2508 	unsigned int itr = IXGBE_ITR_ADAPTIVE_MIN_USECS |
2509 			   IXGBE_ITR_ADAPTIVE_LATENCY;
2510 	unsigned int avg_wire_size, packets, bytes;
2511 	unsigned long next_update = jiffies;
2512 
2513 	/* If we don't have any rings just leave ourselves set for maximum
2514 	 * possible latency so we take ourselves out of the equation.
2515 	 */
2516 	if (!ring_container->ring)
2517 		return;
2518 
2519 	/* If we didn't update within up to 1 - 2 jiffies we can assume
2520 	 * that either packets are coming in so slow there hasn't been
2521 	 * any work, or that there is so much work that NAPI is dealing
2522 	 * with interrupt moderation and we don't need to do anything.
2523 	 */
2524 	if (time_after(next_update, ring_container->next_update))
2525 		goto clear_counts;
2526 
2527 	packets = ring_container->total_packets;
2528 
2529 	/* We have no packets to actually measure against. This means
2530 	 * either one of the other queues on this vector is active or
2531 	 * we are a Tx queue doing TSO with too high of an interrupt rate.
2532 	 *
2533 	 * When this occurs just tick up our delay by the minimum value
2534 	 * and hope that this extra delay will prevent us from being called
2535 	 * without any work on our queue.
2536 	 */
2537 	if (!packets) {
2538 		itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2539 		if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2540 			itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2541 		itr += ring_container->itr & IXGBE_ITR_ADAPTIVE_LATENCY;
2542 		goto clear_counts;
2543 	}
2544 
2545 	bytes = ring_container->total_bytes;
2546 
2547 	/* If packets are less than 4 or bytes are less than 9000 assume
2548 	 * insufficient data to use bulk rate limiting approach. We are
2549 	 * likely latency driven.
2550 	 */
2551 	if (packets < 4 && bytes < 9000) {
2552 		itr = IXGBE_ITR_ADAPTIVE_LATENCY;
2553 		goto adjust_by_size;
2554 	}
2555 
2556 	/* Between 4 and 48 we can assume that our current interrupt delay
2557 	 * is only slightly too low. As such we should increase it by a small
2558 	 * fixed amount.
2559 	 */
2560 	if (packets < 48) {
2561 		itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2562 		if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2563 			itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2564 		goto clear_counts;
2565 	}
2566 
2567 	/* Between 48 and 96 is our "goldilocks" zone where we are working
2568 	 * out "just right". Just report that our current ITR is good for us.
2569 	 */
2570 	if (packets < 96) {
2571 		itr = q_vector->itr >> 2;
2572 		goto clear_counts;
2573 	}
2574 
2575 	/* If packet count is 96 or greater we are likely looking at a slight
2576 	 * overrun of the delay we want. Try halving our delay to see if that
2577 	 * will cut the number of packets in half per interrupt.
2578 	 */
2579 	if (packets < 256) {
2580 		itr = q_vector->itr >> 3;
2581 		if (itr < IXGBE_ITR_ADAPTIVE_MIN_USECS)
2582 			itr = IXGBE_ITR_ADAPTIVE_MIN_USECS;
2583 		goto clear_counts;
2584 	}
2585 
2586 	/* The paths below assume we are dealing with a bulk ITR since number
2587 	 * of packets is 256 or greater. We are just going to have to compute
2588 	 * a value and try to bring the count under control, though for smaller
2589 	 * packet sizes there isn't much we can do as NAPI polling will likely
2590 	 * be kicking in sooner rather than later.
2591 	 */
2592 	itr = IXGBE_ITR_ADAPTIVE_BULK;
2593 
2594 adjust_by_size:
2595 	/* If packet counts are 256 or greater we can assume we have a gross
2596 	 * overestimation of what the rate should be. Instead of trying to fine
2597 	 * tune it just use the formula below to try and dial in an exact value
2598 	 * give the current packet size of the frame.
2599 	 */
2600 	avg_wire_size = bytes / packets;
2601 
2602 	/* The following is a crude approximation of:
2603 	 *  wmem_default / (size + overhead) = desired_pkts_per_int
2604 	 *  rate / bits_per_byte / (size + ethernet overhead) = pkt_rate
2605 	 *  (desired_pkt_rate / pkt_rate) * usecs_per_sec = ITR value
2606 	 *
2607 	 * Assuming wmem_default is 212992 and overhead is 640 bytes per
2608 	 * packet, (256 skb, 64 headroom, 320 shared info), we can reduce the
2609 	 * formula down to
2610 	 *
2611 	 *  (170 * (size + 24)) / (size + 640) = ITR
2612 	 *
2613 	 * We first do some math on the packet size and then finally bitshift
2614 	 * by 8 after rounding up. We also have to account for PCIe link speed
2615 	 * difference as ITR scales based on this.
2616 	 */
2617 	if (avg_wire_size <= 60) {
2618 		/* Start at 50k ints/sec */
2619 		avg_wire_size = 5120;
2620 	} else if (avg_wire_size <= 316) {
2621 		/* 50K ints/sec to 16K ints/sec */
2622 		avg_wire_size *= 40;
2623 		avg_wire_size += 2720;
2624 	} else if (avg_wire_size <= 1084) {
2625 		/* 16K ints/sec to 9.2K ints/sec */
2626 		avg_wire_size *= 15;
2627 		avg_wire_size += 11452;
2628 	} else if (avg_wire_size <= 1980) {
2629 		/* 9.2K ints/sec to 8K ints/sec */
2630 		avg_wire_size *= 5;
2631 		avg_wire_size += 22420;
2632 	} else {
2633 		/* plateau at a limit of 8K ints/sec */
2634 		avg_wire_size = 32256;
2635 	}
2636 
2637 	/* If we are in low latency mode half our delay which doubles the rate
2638 	 * to somewhere between 100K to 16K ints/sec
2639 	 */
2640 	if (itr & IXGBE_ITR_ADAPTIVE_LATENCY)
2641 		avg_wire_size >>= 1;
2642 
2643 	/* Resultant value is 256 times larger than it needs to be. This
2644 	 * gives us room to adjust the value as needed to either increase
2645 	 * or decrease the value based on link speeds of 10G, 2.5G, 1G, etc.
2646 	 *
2647 	 * Use addition as we have already recorded the new latency flag
2648 	 * for the ITR value.
2649 	 */
2650 	switch (q_vector->adapter->link_speed) {
2651 	case IXGBE_LINK_SPEED_10GB_FULL:
2652 	case IXGBE_LINK_SPEED_100_FULL:
2653 	default:
2654 		itr += DIV_ROUND_UP(avg_wire_size,
2655 				    IXGBE_ITR_ADAPTIVE_MIN_INC * 256) *
2656 		       IXGBE_ITR_ADAPTIVE_MIN_INC;
2657 		break;
2658 	case IXGBE_LINK_SPEED_2_5GB_FULL:
2659 	case IXGBE_LINK_SPEED_1GB_FULL:
2660 	case IXGBE_LINK_SPEED_10_FULL:
2661 		itr += DIV_ROUND_UP(avg_wire_size,
2662 				    IXGBE_ITR_ADAPTIVE_MIN_INC * 64) *
2663 		       IXGBE_ITR_ADAPTIVE_MIN_INC;
2664 		break;
2665 	}
2666 
2667 clear_counts:
2668 	/* write back value */
2669 	ring_container->itr = itr;
2670 
2671 	/* next update should occur within next jiffy */
2672 	ring_container->next_update = next_update + 1;
2673 
2674 	ring_container->total_bytes = 0;
2675 	ring_container->total_packets = 0;
2676 }
2677 
2678 /**
2679  * ixgbe_write_eitr - write EITR register in hardware specific way
2680  * @q_vector: structure containing interrupt and ring information
2681  *
2682  * This function is made to be called by ethtool and by the driver
2683  * when it needs to update EITR registers at runtime.  Hardware
2684  * specific quirks/differences are taken care of here.
2685  */
2686 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2687 {
2688 	struct ixgbe_adapter *adapter = q_vector->adapter;
2689 	struct ixgbe_hw *hw = &adapter->hw;
2690 	int v_idx = q_vector->v_idx;
2691 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2692 
2693 	switch (adapter->hw.mac.type) {
2694 	case ixgbe_mac_82598EB:
2695 		/* must write high and low 16 bits to reset counter */
2696 		itr_reg |= (itr_reg << 16);
2697 		break;
2698 	case ixgbe_mac_82599EB:
2699 	case ixgbe_mac_X540:
2700 	case ixgbe_mac_X550:
2701 	case ixgbe_mac_X550EM_x:
2702 	case ixgbe_mac_x550em_a:
2703 		/*
2704 		 * set the WDIS bit to not clear the timer bits and cause an
2705 		 * immediate assertion of the interrupt
2706 		 */
2707 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2708 		break;
2709 	default:
2710 		break;
2711 	}
2712 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2713 }
2714 
2715 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2716 {
2717 	u32 new_itr;
2718 
2719 	ixgbe_update_itr(q_vector, &q_vector->tx);
2720 	ixgbe_update_itr(q_vector, &q_vector->rx);
2721 
2722 	/* use the smallest value of new ITR delay calculations */
2723 	new_itr = min(q_vector->rx.itr, q_vector->tx.itr);
2724 
2725 	/* Clear latency flag if set, shift into correct position */
2726 	new_itr &= ~IXGBE_ITR_ADAPTIVE_LATENCY;
2727 	new_itr <<= 2;
2728 
2729 	if (new_itr != q_vector->itr) {
2730 		/* save the algorithm value here */
2731 		q_vector->itr = new_itr;
2732 
2733 		ixgbe_write_eitr(q_vector);
2734 	}
2735 }
2736 
2737 /**
2738  * ixgbe_check_overtemp_subtask - check for over temperature
2739  * @adapter: pointer to adapter
2740  **/
2741 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2742 {
2743 	struct ixgbe_hw *hw = &adapter->hw;
2744 	u32 eicr = adapter->interrupt_event;
2745 	s32 rc;
2746 
2747 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2748 		return;
2749 
2750 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2751 		return;
2752 
2753 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2754 
2755 	switch (hw->device_id) {
2756 	case IXGBE_DEV_ID_82599_T3_LOM:
2757 		/*
2758 		 * Since the warning interrupt is for both ports
2759 		 * we don't have to check if:
2760 		 *  - This interrupt wasn't for our port.
2761 		 *  - We may have missed the interrupt so always have to
2762 		 *    check if we  got a LSC
2763 		 */
2764 		if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2765 		    !(eicr & IXGBE_EICR_LSC))
2766 			return;
2767 
2768 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2769 			u32 speed;
2770 			bool link_up = false;
2771 
2772 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2773 
2774 			if (link_up)
2775 				return;
2776 		}
2777 
2778 		/* Check if this is not due to overtemp */
2779 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2780 			return;
2781 
2782 		break;
2783 	case IXGBE_DEV_ID_X550EM_A_1G_T:
2784 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
2785 		rc = hw->phy.ops.check_overtemp(hw);
2786 		if (rc != IXGBE_ERR_OVERTEMP)
2787 			return;
2788 		break;
2789 	default:
2790 		if (adapter->hw.mac.type >= ixgbe_mac_X540)
2791 			return;
2792 		if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
2793 			return;
2794 		break;
2795 	}
2796 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2797 
2798 	adapter->interrupt_event = 0;
2799 }
2800 
2801 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2802 {
2803 	struct ixgbe_hw *hw = &adapter->hw;
2804 
2805 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2806 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2807 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2808 		/* write to clear the interrupt */
2809 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2810 	}
2811 }
2812 
2813 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2814 {
2815 	struct ixgbe_hw *hw = &adapter->hw;
2816 
2817 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2818 		return;
2819 
2820 	switch (adapter->hw.mac.type) {
2821 	case ixgbe_mac_82599EB:
2822 		/*
2823 		 * Need to check link state so complete overtemp check
2824 		 * on service task
2825 		 */
2826 		if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
2827 		     (eicr & IXGBE_EICR_LSC)) &&
2828 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2829 			adapter->interrupt_event = eicr;
2830 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2831 			ixgbe_service_event_schedule(adapter);
2832 			return;
2833 		}
2834 		return;
2835 	case ixgbe_mac_x550em_a:
2836 		if (eicr & IXGBE_EICR_GPI_SDP0_X550EM_a) {
2837 			adapter->interrupt_event = eicr;
2838 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2839 			ixgbe_service_event_schedule(adapter);
2840 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC,
2841 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2842 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICR,
2843 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2844 		}
2845 		return;
2846 	case ixgbe_mac_X550:
2847 	case ixgbe_mac_X540:
2848 		if (!(eicr & IXGBE_EICR_TS))
2849 			return;
2850 		break;
2851 	default:
2852 		return;
2853 	}
2854 
2855 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2856 }
2857 
2858 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
2859 {
2860 	switch (hw->mac.type) {
2861 	case ixgbe_mac_82598EB:
2862 		if (hw->phy.type == ixgbe_phy_nl)
2863 			return true;
2864 		return false;
2865 	case ixgbe_mac_82599EB:
2866 	case ixgbe_mac_X550EM_x:
2867 	case ixgbe_mac_x550em_a:
2868 		switch (hw->mac.ops.get_media_type(hw)) {
2869 		case ixgbe_media_type_fiber:
2870 		case ixgbe_media_type_fiber_qsfp:
2871 			return true;
2872 		default:
2873 			return false;
2874 		}
2875 	default:
2876 		return false;
2877 	}
2878 }
2879 
2880 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2881 {
2882 	struct ixgbe_hw *hw = &adapter->hw;
2883 	u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
2884 
2885 	if (!ixgbe_is_sfp(hw))
2886 		return;
2887 
2888 	/* Later MAC's use different SDP */
2889 	if (hw->mac.type >= ixgbe_mac_X540)
2890 		eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2891 
2892 	if (eicr & eicr_mask) {
2893 		/* Clear the interrupt */
2894 		IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2895 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2896 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2897 			adapter->sfp_poll_time = 0;
2898 			ixgbe_service_event_schedule(adapter);
2899 		}
2900 	}
2901 
2902 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
2903 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2904 		/* Clear the interrupt */
2905 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2906 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2907 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2908 			ixgbe_service_event_schedule(adapter);
2909 		}
2910 	}
2911 }
2912 
2913 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2914 {
2915 	struct ixgbe_hw *hw = &adapter->hw;
2916 
2917 	adapter->lsc_int++;
2918 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2919 	adapter->link_check_timeout = jiffies;
2920 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2921 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2922 		IXGBE_WRITE_FLUSH(hw);
2923 		ixgbe_service_event_schedule(adapter);
2924 	}
2925 }
2926 
2927 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2928 					   u64 qmask)
2929 {
2930 	u32 mask;
2931 	struct ixgbe_hw *hw = &adapter->hw;
2932 
2933 	switch (hw->mac.type) {
2934 	case ixgbe_mac_82598EB:
2935 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2936 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2937 		break;
2938 	case ixgbe_mac_82599EB:
2939 	case ixgbe_mac_X540:
2940 	case ixgbe_mac_X550:
2941 	case ixgbe_mac_X550EM_x:
2942 	case ixgbe_mac_x550em_a:
2943 		mask = (qmask & 0xFFFFFFFF);
2944 		if (mask)
2945 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2946 		mask = (qmask >> 32);
2947 		if (mask)
2948 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2949 		break;
2950 	default:
2951 		break;
2952 	}
2953 	/* skip the flush */
2954 }
2955 
2956 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2957 					    u64 qmask)
2958 {
2959 	u32 mask;
2960 	struct ixgbe_hw *hw = &adapter->hw;
2961 
2962 	switch (hw->mac.type) {
2963 	case ixgbe_mac_82598EB:
2964 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2965 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2966 		break;
2967 	case ixgbe_mac_82599EB:
2968 	case ixgbe_mac_X540:
2969 	case ixgbe_mac_X550:
2970 	case ixgbe_mac_X550EM_x:
2971 	case ixgbe_mac_x550em_a:
2972 		mask = (qmask & 0xFFFFFFFF);
2973 		if (mask)
2974 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2975 		mask = (qmask >> 32);
2976 		if (mask)
2977 			IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2978 		break;
2979 	default:
2980 		break;
2981 	}
2982 	/* skip the flush */
2983 }
2984 
2985 /**
2986  * ixgbe_irq_enable - Enable default interrupt generation settings
2987  * @adapter: board private structure
2988  * @queues: enable irqs for queues
2989  * @flush: flush register write
2990  **/
2991 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2992 				    bool flush)
2993 {
2994 	struct ixgbe_hw *hw = &adapter->hw;
2995 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2996 
2997 	/* don't reenable LSC while waiting for link */
2998 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2999 		mask &= ~IXGBE_EIMS_LSC;
3000 
3001 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
3002 		switch (adapter->hw.mac.type) {
3003 		case ixgbe_mac_82599EB:
3004 			mask |= IXGBE_EIMS_GPI_SDP0(hw);
3005 			break;
3006 		case ixgbe_mac_X540:
3007 		case ixgbe_mac_X550:
3008 		case ixgbe_mac_X550EM_x:
3009 		case ixgbe_mac_x550em_a:
3010 			mask |= IXGBE_EIMS_TS;
3011 			break;
3012 		default:
3013 			break;
3014 		}
3015 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
3016 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
3017 	switch (adapter->hw.mac.type) {
3018 	case ixgbe_mac_82599EB:
3019 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
3020 		mask |= IXGBE_EIMS_GPI_SDP2(hw);
3021 		/* fall through */
3022 	case ixgbe_mac_X540:
3023 	case ixgbe_mac_X550:
3024 	case ixgbe_mac_X550EM_x:
3025 	case ixgbe_mac_x550em_a:
3026 		if (adapter->hw.device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
3027 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
3028 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP_N)
3029 			mask |= IXGBE_EIMS_GPI_SDP0(&adapter->hw);
3030 		if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
3031 			mask |= IXGBE_EICR_GPI_SDP0_X540;
3032 		mask |= IXGBE_EIMS_ECC;
3033 		mask |= IXGBE_EIMS_MAILBOX;
3034 		break;
3035 	default:
3036 		break;
3037 	}
3038 
3039 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
3040 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
3041 		mask |= IXGBE_EIMS_FLOW_DIR;
3042 
3043 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
3044 	if (queues)
3045 		ixgbe_irq_enable_queues(adapter, ~0);
3046 	if (flush)
3047 		IXGBE_WRITE_FLUSH(&adapter->hw);
3048 }
3049 
3050 static irqreturn_t ixgbe_msix_other(int irq, void *data)
3051 {
3052 	struct ixgbe_adapter *adapter = data;
3053 	struct ixgbe_hw *hw = &adapter->hw;
3054 	u32 eicr;
3055 
3056 	/*
3057 	 * Workaround for Silicon errata.  Use clear-by-write instead
3058 	 * of clear-by-read.  Reading with EICS will return the
3059 	 * interrupt causes without clearing, which later be done
3060 	 * with the write to EICR.
3061 	 */
3062 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
3063 
3064 	/* The lower 16bits of the EICR register are for the queue interrupts
3065 	 * which should be masked here in order to not accidentally clear them if
3066 	 * the bits are high when ixgbe_msix_other is called. There is a race
3067 	 * condition otherwise which results in possible performance loss
3068 	 * especially if the ixgbe_msix_other interrupt is triggering
3069 	 * consistently (as it would when PPS is turned on for the X540 device)
3070 	 */
3071 	eicr &= 0xFFFF0000;
3072 
3073 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
3074 
3075 	if (eicr & IXGBE_EICR_LSC)
3076 		ixgbe_check_lsc(adapter);
3077 
3078 	if (eicr & IXGBE_EICR_MAILBOX)
3079 		ixgbe_msg_task(adapter);
3080 
3081 	switch (hw->mac.type) {
3082 	case ixgbe_mac_82599EB:
3083 	case ixgbe_mac_X540:
3084 	case ixgbe_mac_X550:
3085 	case ixgbe_mac_X550EM_x:
3086 	case ixgbe_mac_x550em_a:
3087 		if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
3088 		    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
3089 			adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
3090 			ixgbe_service_event_schedule(adapter);
3091 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
3092 					IXGBE_EICR_GPI_SDP0_X540);
3093 		}
3094 		if (eicr & IXGBE_EICR_ECC) {
3095 			e_info(link, "Received ECC Err, initiating reset\n");
3096 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3097 			ixgbe_service_event_schedule(adapter);
3098 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3099 		}
3100 		/* Handle Flow Director Full threshold interrupt */
3101 		if (eicr & IXGBE_EICR_FLOW_DIR) {
3102 			int reinit_count = 0;
3103 			int i;
3104 			for (i = 0; i < adapter->num_tx_queues; i++) {
3105 				struct ixgbe_ring *ring = adapter->tx_ring[i];
3106 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
3107 						       &ring->state))
3108 					reinit_count++;
3109 			}
3110 			if (reinit_count) {
3111 				/* no more flow director interrupts until after init */
3112 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
3113 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
3114 				ixgbe_service_event_schedule(adapter);
3115 			}
3116 		}
3117 		ixgbe_check_sfp_event(adapter, eicr);
3118 		ixgbe_check_overtemp_event(adapter, eicr);
3119 		break;
3120 	default:
3121 		break;
3122 	}
3123 
3124 	ixgbe_check_fan_failure(adapter, eicr);
3125 
3126 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3127 		ixgbe_ptp_check_pps_event(adapter);
3128 
3129 	/* re-enable the original interrupt state, no lsc, no queues */
3130 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3131 		ixgbe_irq_enable(adapter, false, false);
3132 
3133 	return IRQ_HANDLED;
3134 }
3135 
3136 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
3137 {
3138 	struct ixgbe_q_vector *q_vector = data;
3139 
3140 	/* EIAM disabled interrupts (on this vector) for us */
3141 
3142 	if (q_vector->rx.ring || q_vector->tx.ring)
3143 		napi_schedule_irqoff(&q_vector->napi);
3144 
3145 	return IRQ_HANDLED;
3146 }
3147 
3148 /**
3149  * ixgbe_poll - NAPI Rx polling callback
3150  * @napi: structure for representing this polling device
3151  * @budget: how many packets driver is allowed to clean
3152  *
3153  * This function is used for legacy and MSI, NAPI mode
3154  **/
3155 int ixgbe_poll(struct napi_struct *napi, int budget)
3156 {
3157 	struct ixgbe_q_vector *q_vector =
3158 				container_of(napi, struct ixgbe_q_vector, napi);
3159 	struct ixgbe_adapter *adapter = q_vector->adapter;
3160 	struct ixgbe_ring *ring;
3161 	int per_ring_budget, work_done = 0;
3162 	bool clean_complete = true;
3163 
3164 #ifdef CONFIG_IXGBE_DCA
3165 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
3166 		ixgbe_update_dca(q_vector);
3167 #endif
3168 
3169 	ixgbe_for_each_ring(ring, q_vector->tx) {
3170 		if (!ixgbe_clean_tx_irq(q_vector, ring, budget))
3171 			clean_complete = false;
3172 	}
3173 
3174 	/* Exit if we are called by netpoll */
3175 	if (budget <= 0)
3176 		return budget;
3177 
3178 	/* attempt to distribute budget to each queue fairly, but don't allow
3179 	 * the budget to go below 1 because we'll exit polling */
3180 	if (q_vector->rx.count > 1)
3181 		per_ring_budget = max(budget/q_vector->rx.count, 1);
3182 	else
3183 		per_ring_budget = budget;
3184 
3185 	ixgbe_for_each_ring(ring, q_vector->rx) {
3186 		int cleaned = ixgbe_clean_rx_irq(q_vector, ring,
3187 						 per_ring_budget);
3188 
3189 		work_done += cleaned;
3190 		if (cleaned >= per_ring_budget)
3191 			clean_complete = false;
3192 	}
3193 
3194 	/* If all work not completed, return budget and keep polling */
3195 	if (!clean_complete)
3196 		return budget;
3197 
3198 	/* all work done, exit the polling mode */
3199 	napi_complete_done(napi, work_done);
3200 	if (adapter->rx_itr_setting & 1)
3201 		ixgbe_set_itr(q_vector);
3202 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3203 		ixgbe_irq_enable_queues(adapter, BIT_ULL(q_vector->v_idx));
3204 
3205 	return min(work_done, budget - 1);
3206 }
3207 
3208 /**
3209  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
3210  * @adapter: board private structure
3211  *
3212  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
3213  * interrupts from the kernel.
3214  **/
3215 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
3216 {
3217 	struct net_device *netdev = adapter->netdev;
3218 	unsigned int ri = 0, ti = 0;
3219 	int vector, err;
3220 
3221 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3222 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3223 		struct msix_entry *entry = &adapter->msix_entries[vector];
3224 
3225 		if (q_vector->tx.ring && q_vector->rx.ring) {
3226 			snprintf(q_vector->name, sizeof(q_vector->name),
3227 				 "%s-TxRx-%u", netdev->name, ri++);
3228 			ti++;
3229 		} else if (q_vector->rx.ring) {
3230 			snprintf(q_vector->name, sizeof(q_vector->name),
3231 				 "%s-rx-%u", netdev->name, ri++);
3232 		} else if (q_vector->tx.ring) {
3233 			snprintf(q_vector->name, sizeof(q_vector->name),
3234 				 "%s-tx-%u", netdev->name, ti++);
3235 		} else {
3236 			/* skip this unused q_vector */
3237 			continue;
3238 		}
3239 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
3240 				  q_vector->name, q_vector);
3241 		if (err) {
3242 			e_err(probe, "request_irq failed for MSIX interrupt "
3243 			      "Error: %d\n", err);
3244 			goto free_queue_irqs;
3245 		}
3246 		/* If Flow Director is enabled, set interrupt affinity */
3247 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3248 			/* assign the mask for this irq */
3249 			irq_set_affinity_hint(entry->vector,
3250 					      &q_vector->affinity_mask);
3251 		}
3252 	}
3253 
3254 	err = request_irq(adapter->msix_entries[vector].vector,
3255 			  ixgbe_msix_other, 0, netdev->name, adapter);
3256 	if (err) {
3257 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
3258 		goto free_queue_irqs;
3259 	}
3260 
3261 	return 0;
3262 
3263 free_queue_irqs:
3264 	while (vector) {
3265 		vector--;
3266 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
3267 				      NULL);
3268 		free_irq(adapter->msix_entries[vector].vector,
3269 			 adapter->q_vector[vector]);
3270 	}
3271 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
3272 	pci_disable_msix(adapter->pdev);
3273 	kfree(adapter->msix_entries);
3274 	adapter->msix_entries = NULL;
3275 	return err;
3276 }
3277 
3278 /**
3279  * ixgbe_intr - legacy mode Interrupt Handler
3280  * @irq: interrupt number
3281  * @data: pointer to a network interface device structure
3282  **/
3283 static irqreturn_t ixgbe_intr(int irq, void *data)
3284 {
3285 	struct ixgbe_adapter *adapter = data;
3286 	struct ixgbe_hw *hw = &adapter->hw;
3287 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3288 	u32 eicr;
3289 
3290 	/*
3291 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
3292 	 * before the read of EICR.
3293 	 */
3294 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
3295 
3296 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
3297 	 * therefore no explicit interrupt disable is necessary */
3298 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3299 	if (!eicr) {
3300 		/*
3301 		 * shared interrupt alert!
3302 		 * make sure interrupts are enabled because the read will
3303 		 * have disabled interrupts due to EIAM
3304 		 * finish the workaround of silicon errata on 82598.  Unmask
3305 		 * the interrupt that we masked before the EICR read.
3306 		 */
3307 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
3308 			ixgbe_irq_enable(adapter, true, true);
3309 		return IRQ_NONE;	/* Not our interrupt */
3310 	}
3311 
3312 	if (eicr & IXGBE_EICR_LSC)
3313 		ixgbe_check_lsc(adapter);
3314 
3315 	switch (hw->mac.type) {
3316 	case ixgbe_mac_82599EB:
3317 		ixgbe_check_sfp_event(adapter, eicr);
3318 		/* Fall through */
3319 	case ixgbe_mac_X540:
3320 	case ixgbe_mac_X550:
3321 	case ixgbe_mac_X550EM_x:
3322 	case ixgbe_mac_x550em_a:
3323 		if (eicr & IXGBE_EICR_ECC) {
3324 			e_info(link, "Received ECC Err, initiating reset\n");
3325 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3326 			ixgbe_service_event_schedule(adapter);
3327 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3328 		}
3329 		ixgbe_check_overtemp_event(adapter, eicr);
3330 		break;
3331 	default:
3332 		break;
3333 	}
3334 
3335 	ixgbe_check_fan_failure(adapter, eicr);
3336 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3337 		ixgbe_ptp_check_pps_event(adapter);
3338 
3339 	/* would disable interrupts here but EIAM disabled it */
3340 	napi_schedule_irqoff(&q_vector->napi);
3341 
3342 	/*
3343 	 * re-enable link(maybe) and non-queue interrupts, no flush.
3344 	 * ixgbe_poll will re-enable the queue interrupts
3345 	 */
3346 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3347 		ixgbe_irq_enable(adapter, false, false);
3348 
3349 	return IRQ_HANDLED;
3350 }
3351 
3352 /**
3353  * ixgbe_request_irq - initialize interrupts
3354  * @adapter: board private structure
3355  *
3356  * Attempts to configure interrupts using the best available
3357  * capabilities of the hardware and kernel.
3358  **/
3359 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
3360 {
3361 	struct net_device *netdev = adapter->netdev;
3362 	int err;
3363 
3364 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
3365 		err = ixgbe_request_msix_irqs(adapter);
3366 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
3367 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
3368 				  netdev->name, adapter);
3369 	else
3370 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
3371 				  netdev->name, adapter);
3372 
3373 	if (err)
3374 		e_err(probe, "request_irq failed, Error %d\n", err);
3375 
3376 	return err;
3377 }
3378 
3379 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
3380 {
3381 	int vector;
3382 
3383 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
3384 		free_irq(adapter->pdev->irq, adapter);
3385 		return;
3386 	}
3387 
3388 	if (!adapter->msix_entries)
3389 		return;
3390 
3391 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3392 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3393 		struct msix_entry *entry = &adapter->msix_entries[vector];
3394 
3395 		/* free only the irqs that were actually requested */
3396 		if (!q_vector->rx.ring && !q_vector->tx.ring)
3397 			continue;
3398 
3399 		/* clear the affinity_mask in the IRQ descriptor */
3400 		irq_set_affinity_hint(entry->vector, NULL);
3401 
3402 		free_irq(entry->vector, q_vector);
3403 	}
3404 
3405 	free_irq(adapter->msix_entries[vector].vector, adapter);
3406 }
3407 
3408 /**
3409  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3410  * @adapter: board private structure
3411  **/
3412 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3413 {
3414 	switch (adapter->hw.mac.type) {
3415 	case ixgbe_mac_82598EB:
3416 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3417 		break;
3418 	case ixgbe_mac_82599EB:
3419 	case ixgbe_mac_X540:
3420 	case ixgbe_mac_X550:
3421 	case ixgbe_mac_X550EM_x:
3422 	case ixgbe_mac_x550em_a:
3423 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3424 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3425 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3426 		break;
3427 	default:
3428 		break;
3429 	}
3430 	IXGBE_WRITE_FLUSH(&adapter->hw);
3431 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3432 		int vector;
3433 
3434 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
3435 			synchronize_irq(adapter->msix_entries[vector].vector);
3436 
3437 		synchronize_irq(adapter->msix_entries[vector++].vector);
3438 	} else {
3439 		synchronize_irq(adapter->pdev->irq);
3440 	}
3441 }
3442 
3443 /**
3444  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3445  * @adapter: board private structure
3446  *
3447  **/
3448 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3449 {
3450 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3451 
3452 	ixgbe_write_eitr(q_vector);
3453 
3454 	ixgbe_set_ivar(adapter, 0, 0, 0);
3455 	ixgbe_set_ivar(adapter, 1, 0, 0);
3456 
3457 	e_info(hw, "Legacy interrupt IVAR setup done\n");
3458 }
3459 
3460 /**
3461  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3462  * @adapter: board private structure
3463  * @ring: structure containing ring specific data
3464  *
3465  * Configure the Tx descriptor ring after a reset.
3466  **/
3467 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3468 			     struct ixgbe_ring *ring)
3469 {
3470 	struct ixgbe_hw *hw = &adapter->hw;
3471 	u64 tdba = ring->dma;
3472 	int wait_loop = 10;
3473 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
3474 	u8 reg_idx = ring->reg_idx;
3475 
3476 	/* disable queue to avoid issues while updating state */
3477 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3478 	IXGBE_WRITE_FLUSH(hw);
3479 
3480 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3481 			(tdba & DMA_BIT_MASK(32)));
3482 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3483 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3484 			ring->count * sizeof(union ixgbe_adv_tx_desc));
3485 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3486 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3487 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3488 
3489 	/*
3490 	 * set WTHRESH to encourage burst writeback, it should not be set
3491 	 * higher than 1 when:
3492 	 * - ITR is 0 as it could cause false TX hangs
3493 	 * - ITR is set to > 100k int/sec and BQL is enabled
3494 	 *
3495 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3496 	 * to or less than the number of on chip descriptors, which is
3497 	 * currently 40.
3498 	 */
3499 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3500 		txdctl |= 1u << 16;	/* WTHRESH = 1 */
3501 	else
3502 		txdctl |= 8u << 16;	/* WTHRESH = 8 */
3503 
3504 	/*
3505 	 * Setting PTHRESH to 32 both improves performance
3506 	 * and avoids a TX hang with DFP enabled
3507 	 */
3508 	txdctl |= (1u << 8) |	/* HTHRESH = 1 */
3509 		   32;		/* PTHRESH = 32 */
3510 
3511 	/* reinitialize flowdirector state */
3512 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3513 		ring->atr_sample_rate = adapter->atr_sample_rate;
3514 		ring->atr_count = 0;
3515 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3516 	} else {
3517 		ring->atr_sample_rate = 0;
3518 	}
3519 
3520 	/* initialize XPS */
3521 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3522 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3523 
3524 		if (q_vector)
3525 			netif_set_xps_queue(ring->netdev,
3526 					    &q_vector->affinity_mask,
3527 					    ring->queue_index);
3528 	}
3529 
3530 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3531 
3532 	/* reinitialize tx_buffer_info */
3533 	memset(ring->tx_buffer_info, 0,
3534 	       sizeof(struct ixgbe_tx_buffer) * ring->count);
3535 
3536 	/* enable queue */
3537 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3538 
3539 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3540 	if (hw->mac.type == ixgbe_mac_82598EB &&
3541 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3542 		return;
3543 
3544 	/* poll to verify queue is enabled */
3545 	do {
3546 		usleep_range(1000, 2000);
3547 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3548 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3549 	if (!wait_loop)
3550 		hw_dbg(hw, "Could not enable Tx Queue %d\n", reg_idx);
3551 }
3552 
3553 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3554 {
3555 	struct ixgbe_hw *hw = &adapter->hw;
3556 	u32 rttdcs, mtqc;
3557 	u8 tcs = adapter->hw_tcs;
3558 
3559 	if (hw->mac.type == ixgbe_mac_82598EB)
3560 		return;
3561 
3562 	/* disable the arbiter while setting MTQC */
3563 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3564 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3565 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3566 
3567 	/* set transmit pool layout */
3568 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3569 		mtqc = IXGBE_MTQC_VT_ENA;
3570 		if (tcs > 4)
3571 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3572 		else if (tcs > 1)
3573 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3574 		else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3575 			 IXGBE_82599_VMDQ_4Q_MASK)
3576 			mtqc |= IXGBE_MTQC_32VF;
3577 		else
3578 			mtqc |= IXGBE_MTQC_64VF;
3579 	} else {
3580 		if (tcs > 4)
3581 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3582 		else if (tcs > 1)
3583 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3584 		else
3585 			mtqc = IXGBE_MTQC_64Q_1PB;
3586 	}
3587 
3588 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3589 
3590 	/* Enable Security TX Buffer IFG for multiple pb */
3591 	if (tcs) {
3592 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3593 		sectx |= IXGBE_SECTX_DCB;
3594 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3595 	}
3596 
3597 	/* re-enable the arbiter */
3598 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3599 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3600 }
3601 
3602 /**
3603  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3604  * @adapter: board private structure
3605  *
3606  * Configure the Tx unit of the MAC after a reset.
3607  **/
3608 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3609 {
3610 	struct ixgbe_hw *hw = &adapter->hw;
3611 	u32 dmatxctl;
3612 	u32 i;
3613 
3614 	ixgbe_setup_mtqc(adapter);
3615 
3616 	if (hw->mac.type != ixgbe_mac_82598EB) {
3617 		/* DMATXCTL.EN must be before Tx queues are enabled */
3618 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3619 		dmatxctl |= IXGBE_DMATXCTL_TE;
3620 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3621 	}
3622 
3623 	/* Setup the HW Tx Head and Tail descriptor pointers */
3624 	for (i = 0; i < adapter->num_tx_queues; i++)
3625 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3626 	for (i = 0; i < adapter->num_xdp_queues; i++)
3627 		ixgbe_configure_tx_ring(adapter, adapter->xdp_ring[i]);
3628 }
3629 
3630 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3631 				 struct ixgbe_ring *ring)
3632 {
3633 	struct ixgbe_hw *hw = &adapter->hw;
3634 	u8 reg_idx = ring->reg_idx;
3635 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3636 
3637 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3638 
3639 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3640 }
3641 
3642 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3643 				  struct ixgbe_ring *ring)
3644 {
3645 	struct ixgbe_hw *hw = &adapter->hw;
3646 	u8 reg_idx = ring->reg_idx;
3647 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3648 
3649 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3650 
3651 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3652 }
3653 
3654 #ifdef CONFIG_IXGBE_DCB
3655 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3656 #else
3657 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3658 #endif
3659 {
3660 	int i;
3661 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3662 
3663 	if (adapter->ixgbe_ieee_pfc)
3664 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3665 
3666 	/*
3667 	 * We should set the drop enable bit if:
3668 	 *  SR-IOV is enabled
3669 	 *   or
3670 	 *  Number of Rx queues > 1 and flow control is disabled
3671 	 *
3672 	 *  This allows us to avoid head of line blocking for security
3673 	 *  and performance reasons.
3674 	 */
3675 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3676 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3677 		for (i = 0; i < adapter->num_rx_queues; i++)
3678 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3679 	} else {
3680 		for (i = 0; i < adapter->num_rx_queues; i++)
3681 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3682 	}
3683 }
3684 
3685 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3686 
3687 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3688 				   struct ixgbe_ring *rx_ring)
3689 {
3690 	struct ixgbe_hw *hw = &adapter->hw;
3691 	u32 srrctl;
3692 	u8 reg_idx = rx_ring->reg_idx;
3693 
3694 	if (hw->mac.type == ixgbe_mac_82598EB) {
3695 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3696 
3697 		/*
3698 		 * if VMDq is not active we must program one srrctl register
3699 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3700 		 */
3701 		reg_idx &= mask;
3702 	}
3703 
3704 	/* configure header buffer length, needed for RSC */
3705 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3706 
3707 	/* configure the packet buffer length */
3708 	if (test_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state))
3709 		srrctl |= IXGBE_RXBUFFER_3K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3710 	else
3711 		srrctl |= IXGBE_RXBUFFER_2K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3712 
3713 	/* configure descriptor type */
3714 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3715 
3716 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3717 }
3718 
3719 /**
3720  * ixgbe_rss_indir_tbl_entries - Return RSS indirection table entries
3721  * @adapter: device handle
3722  *
3723  *  - 82598/82599/X540:     128
3724  *  - X550(non-SRIOV mode): 512
3725  *  - X550(SRIOV mode):     64
3726  */
3727 u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3728 {
3729 	if (adapter->hw.mac.type < ixgbe_mac_X550)
3730 		return 128;
3731 	else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3732 		return 64;
3733 	else
3734 		return 512;
3735 }
3736 
3737 /**
3738  * ixgbe_store_key - Write the RSS key to HW
3739  * @adapter: device handle
3740  *
3741  * Write the RSS key stored in adapter.rss_key to HW.
3742  */
3743 void ixgbe_store_key(struct ixgbe_adapter *adapter)
3744 {
3745 	struct ixgbe_hw *hw = &adapter->hw;
3746 	int i;
3747 
3748 	for (i = 0; i < 10; i++)
3749 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3750 }
3751 
3752 /**
3753  * ixgbe_init_rss_key - Initialize adapter RSS key
3754  * @adapter: device handle
3755  *
3756  * Allocates and initializes the RSS key if it is not allocated.
3757  **/
3758 static inline int ixgbe_init_rss_key(struct ixgbe_adapter *adapter)
3759 {
3760 	u32 *rss_key;
3761 
3762 	if (!adapter->rss_key) {
3763 		rss_key = kzalloc(IXGBE_RSS_KEY_SIZE, GFP_KERNEL);
3764 		if (unlikely(!rss_key))
3765 			return -ENOMEM;
3766 
3767 		netdev_rss_key_fill(rss_key, IXGBE_RSS_KEY_SIZE);
3768 		adapter->rss_key = rss_key;
3769 	}
3770 
3771 	return 0;
3772 }
3773 
3774 /**
3775  * ixgbe_store_reta - Write the RETA table to HW
3776  * @adapter: device handle
3777  *
3778  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3779  */
3780 void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3781 {
3782 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3783 	struct ixgbe_hw *hw = &adapter->hw;
3784 	u32 reta = 0;
3785 	u32 indices_multi;
3786 	u8 *indir_tbl = adapter->rss_indir_tbl;
3787 
3788 	/* Fill out the redirection table as follows:
3789 	 *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3790 	 *    indices.
3791 	 *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3792 	 *  - X550:       8 bit wide entries containing 6 bit RSS index
3793 	 */
3794 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3795 		indices_multi = 0x11;
3796 	else
3797 		indices_multi = 0x1;
3798 
3799 	/* Write redirection table to HW */
3800 	for (i = 0; i < reta_entries; i++) {
3801 		reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3802 		if ((i & 3) == 3) {
3803 			if (i < 128)
3804 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3805 			else
3806 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3807 						reta);
3808 			reta = 0;
3809 		}
3810 	}
3811 }
3812 
3813 /**
3814  * ixgbe_store_vfreta - Write the RETA table to HW (x550 devices in SRIOV mode)
3815  * @adapter: device handle
3816  *
3817  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3818  */
3819 static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3820 {
3821 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3822 	struct ixgbe_hw *hw = &adapter->hw;
3823 	u32 vfreta = 0;
3824 
3825 	/* Write redirection table to HW */
3826 	for (i = 0; i < reta_entries; i++) {
3827 		u16 pool = adapter->num_rx_pools;
3828 
3829 		vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3830 		if ((i & 3) != 3)
3831 			continue;
3832 
3833 		while (pool--)
3834 			IXGBE_WRITE_REG(hw,
3835 					IXGBE_PFVFRETA(i >> 2, VMDQ_P(pool)),
3836 					vfreta);
3837 		vfreta = 0;
3838 	}
3839 }
3840 
3841 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3842 {
3843 	u32 i, j;
3844 	u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3845 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3846 
3847 	/* Program table for at least 4 queues w/ SR-IOV so that VFs can
3848 	 * make full use of any rings they may have.  We will use the
3849 	 * PSRTYPE register to control how many rings we use within the PF.
3850 	 */
3851 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 4))
3852 		rss_i = 4;
3853 
3854 	/* Fill out hash function seeds */
3855 	ixgbe_store_key(adapter);
3856 
3857 	/* Fill out redirection table */
3858 	memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3859 
3860 	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3861 		if (j == rss_i)
3862 			j = 0;
3863 
3864 		adapter->rss_indir_tbl[i] = j;
3865 	}
3866 
3867 	ixgbe_store_reta(adapter);
3868 }
3869 
3870 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3871 {
3872 	struct ixgbe_hw *hw = &adapter->hw;
3873 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3874 	int i, j;
3875 
3876 	/* Fill out hash function seeds */
3877 	for (i = 0; i < 10; i++) {
3878 		u16 pool = adapter->num_rx_pools;
3879 
3880 		while (pool--)
3881 			IXGBE_WRITE_REG(hw,
3882 					IXGBE_PFVFRSSRK(i, VMDQ_P(pool)),
3883 					*(adapter->rss_key + i));
3884 	}
3885 
3886 	/* Fill out the redirection table */
3887 	for (i = 0, j = 0; i < 64; i++, j++) {
3888 		if (j == rss_i)
3889 			j = 0;
3890 
3891 		adapter->rss_indir_tbl[i] = j;
3892 	}
3893 
3894 	ixgbe_store_vfreta(adapter);
3895 }
3896 
3897 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3898 {
3899 	struct ixgbe_hw *hw = &adapter->hw;
3900 	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3901 	u32 rxcsum;
3902 
3903 	/* Disable indicating checksum in descriptor, enables RSS hash */
3904 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3905 	rxcsum |= IXGBE_RXCSUM_PCSD;
3906 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3907 
3908 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3909 		if (adapter->ring_feature[RING_F_RSS].mask)
3910 			mrqc = IXGBE_MRQC_RSSEN;
3911 	} else {
3912 		u8 tcs = adapter->hw_tcs;
3913 
3914 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3915 			if (tcs > 4)
3916 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3917 			else if (tcs > 1)
3918 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3919 			else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3920 				 IXGBE_82599_VMDQ_4Q_MASK)
3921 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3922 			else
3923 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3924 
3925 			/* Enable L3/L4 for Tx Switched packets */
3926 			mrqc |= IXGBE_MRQC_L3L4TXSWEN;
3927 		} else {
3928 			if (tcs > 4)
3929 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3930 			else if (tcs > 1)
3931 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3932 			else
3933 				mrqc = IXGBE_MRQC_RSSEN;
3934 		}
3935 	}
3936 
3937 	/* Perform hash on these packet types */
3938 	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3939 		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3940 		     IXGBE_MRQC_RSS_FIELD_IPV6 |
3941 		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3942 
3943 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3944 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3945 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3946 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3947 
3948 	if ((hw->mac.type >= ixgbe_mac_X550) &&
3949 	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
3950 		u16 pool = adapter->num_rx_pools;
3951 
3952 		/* Enable VF RSS mode */
3953 		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
3954 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3955 
3956 		/* Setup RSS through the VF registers */
3957 		ixgbe_setup_vfreta(adapter);
3958 		vfmrqc = IXGBE_MRQC_RSSEN;
3959 		vfmrqc |= rss_field;
3960 
3961 		while (pool--)
3962 			IXGBE_WRITE_REG(hw,
3963 					IXGBE_PFVFMRQC(VMDQ_P(pool)),
3964 					vfmrqc);
3965 	} else {
3966 		ixgbe_setup_reta(adapter);
3967 		mrqc |= rss_field;
3968 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3969 	}
3970 }
3971 
3972 /**
3973  * ixgbe_configure_rscctl - enable RSC for the indicated ring
3974  * @adapter: address of board private structure
3975  * @ring: structure containing ring specific data
3976  **/
3977 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
3978 				   struct ixgbe_ring *ring)
3979 {
3980 	struct ixgbe_hw *hw = &adapter->hw;
3981 	u32 rscctrl;
3982 	u8 reg_idx = ring->reg_idx;
3983 
3984 	if (!ring_is_rsc_enabled(ring))
3985 		return;
3986 
3987 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
3988 	rscctrl |= IXGBE_RSCCTL_RSCEN;
3989 	/*
3990 	 * we must limit the number of descriptors so that the
3991 	 * total size of max desc * buf_len is not greater
3992 	 * than 65536
3993 	 */
3994 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
3995 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
3996 }
3997 
3998 #define IXGBE_MAX_RX_DESC_POLL 10
3999 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
4000 				       struct ixgbe_ring *ring)
4001 {
4002 	struct ixgbe_hw *hw = &adapter->hw;
4003 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
4004 	u32 rxdctl;
4005 	u8 reg_idx = ring->reg_idx;
4006 
4007 	if (ixgbe_removed(hw->hw_addr))
4008 		return;
4009 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
4010 	if (hw->mac.type == ixgbe_mac_82598EB &&
4011 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
4012 		return;
4013 
4014 	do {
4015 		usleep_range(1000, 2000);
4016 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4017 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
4018 
4019 	if (!wait_loop) {
4020 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
4021 		      "the polling period\n", reg_idx);
4022 	}
4023 }
4024 
4025 void ixgbe_disable_rx_queue(struct ixgbe_adapter *adapter,
4026 			    struct ixgbe_ring *ring)
4027 {
4028 	struct ixgbe_hw *hw = &adapter->hw;
4029 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
4030 	u32 rxdctl;
4031 	u8 reg_idx = ring->reg_idx;
4032 
4033 	if (ixgbe_removed(hw->hw_addr))
4034 		return;
4035 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4036 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
4037 
4038 	/* write value back with RXDCTL.ENABLE bit cleared */
4039 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4040 
4041 	if (hw->mac.type == ixgbe_mac_82598EB &&
4042 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
4043 		return;
4044 
4045 	/* the hardware may take up to 100us to really disable the rx queue */
4046 	do {
4047 		udelay(10);
4048 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4049 	} while (--wait_loop && (rxdctl & IXGBE_RXDCTL_ENABLE));
4050 
4051 	if (!wait_loop) {
4052 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not cleared within "
4053 		      "the polling period\n", reg_idx);
4054 	}
4055 }
4056 
4057 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
4058 			     struct ixgbe_ring *ring)
4059 {
4060 	struct ixgbe_hw *hw = &adapter->hw;
4061 	union ixgbe_adv_rx_desc *rx_desc;
4062 	u64 rdba = ring->dma;
4063 	u32 rxdctl;
4064 	u8 reg_idx = ring->reg_idx;
4065 
4066 	/* disable queue to avoid issues while updating state */
4067 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4068 	ixgbe_disable_rx_queue(adapter, ring);
4069 
4070 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
4071 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
4072 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
4073 			ring->count * sizeof(union ixgbe_adv_rx_desc));
4074 	/* Force flushing of IXGBE_RDLEN to prevent MDD */
4075 	IXGBE_WRITE_FLUSH(hw);
4076 
4077 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
4078 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
4079 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
4080 
4081 	ixgbe_configure_srrctl(adapter, ring);
4082 	ixgbe_configure_rscctl(adapter, ring);
4083 
4084 	if (hw->mac.type == ixgbe_mac_82598EB) {
4085 		/*
4086 		 * enable cache line friendly hardware writes:
4087 		 * PTHRESH=32 descriptors (half the internal cache),
4088 		 * this also removes ugly rx_no_buffer_count increment
4089 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
4090 		 * WTHRESH=8 burst writeback up to two cache lines
4091 		 */
4092 		rxdctl &= ~0x3FFFFF;
4093 		rxdctl |=  0x080420;
4094 #if (PAGE_SIZE < 8192)
4095 	/* RXDCTL.RLPML does not work on 82599 */
4096 	} else if (hw->mac.type != ixgbe_mac_82599EB) {
4097 		rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
4098 			    IXGBE_RXDCTL_RLPML_EN);
4099 
4100 		/* Limit the maximum frame size so we don't overrun the skb.
4101 		 * This can happen in SRIOV mode when the MTU of the VF is
4102 		 * higher than the MTU of the PF.
4103 		 */
4104 		if (ring_uses_build_skb(ring) &&
4105 		    !test_bit(__IXGBE_RX_3K_BUFFER, &ring->state))
4106 			rxdctl |= IXGBE_MAX_2K_FRAME_BUILD_SKB |
4107 				  IXGBE_RXDCTL_RLPML_EN;
4108 #endif
4109 	}
4110 
4111 	/* initialize rx_buffer_info */
4112 	memset(ring->rx_buffer_info, 0,
4113 	       sizeof(struct ixgbe_rx_buffer) * ring->count);
4114 
4115 	/* initialize Rx descriptor 0 */
4116 	rx_desc = IXGBE_RX_DESC(ring, 0);
4117 	rx_desc->wb.upper.length = 0;
4118 
4119 	/* enable receive descriptor ring */
4120 	rxdctl |= IXGBE_RXDCTL_ENABLE;
4121 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4122 
4123 	ixgbe_rx_desc_queue_enable(adapter, ring);
4124 	ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
4125 }
4126 
4127 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
4128 {
4129 	struct ixgbe_hw *hw = &adapter->hw;
4130 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
4131 	u16 pool = adapter->num_rx_pools;
4132 
4133 	/* PSRTYPE must be initialized in non 82598 adapters */
4134 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4135 		      IXGBE_PSRTYPE_UDPHDR |
4136 		      IXGBE_PSRTYPE_IPV4HDR |
4137 		      IXGBE_PSRTYPE_L2HDR |
4138 		      IXGBE_PSRTYPE_IPV6HDR;
4139 
4140 	if (hw->mac.type == ixgbe_mac_82598EB)
4141 		return;
4142 
4143 	if (rss_i > 3)
4144 		psrtype |= 2u << 29;
4145 	else if (rss_i > 1)
4146 		psrtype |= 1u << 29;
4147 
4148 	while (pool--)
4149 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4150 }
4151 
4152 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
4153 {
4154 	struct ixgbe_hw *hw = &adapter->hw;
4155 	u16 pool = adapter->num_rx_pools;
4156 	u32 reg_offset, vf_shift, vmolr;
4157 	u32 gcr_ext, vmdctl;
4158 	int i;
4159 
4160 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
4161 		return;
4162 
4163 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
4164 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
4165 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
4166 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
4167 	vmdctl |= IXGBE_VT_CTL_REPLEN;
4168 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
4169 
4170 	/* accept untagged packets until a vlan tag is
4171 	 * specifically set for the VMDQ queue/pool
4172 	 */
4173 	vmolr = IXGBE_VMOLR_AUPE;
4174 	while (pool--)
4175 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(pool)), vmolr);
4176 
4177 	vf_shift = VMDQ_P(0) % 32;
4178 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
4179 
4180 	/* Enable only the PF's pool for Tx/Rx */
4181 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), GENMASK(31, vf_shift));
4182 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
4183 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), GENMASK(31, vf_shift));
4184 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
4185 	if (adapter->bridge_mode == BRIDGE_MODE_VEB)
4186 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
4187 
4188 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
4189 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
4190 
4191 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
4192 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4193 
4194 	/*
4195 	 * Set up VF register offsets for selected VT Mode,
4196 	 * i.e. 32 or 64 VFs for SR-IOV
4197 	 */
4198 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4199 	case IXGBE_82599_VMDQ_8Q_MASK:
4200 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
4201 		break;
4202 	case IXGBE_82599_VMDQ_4Q_MASK:
4203 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
4204 		break;
4205 	default:
4206 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
4207 		break;
4208 	}
4209 
4210 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
4211 
4212 	for (i = 0; i < adapter->num_vfs; i++) {
4213 		/* configure spoof checking */
4214 		ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i,
4215 					  adapter->vfinfo[i].spoofchk_enabled);
4216 
4217 		/* Enable/Disable RSS query feature  */
4218 		ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
4219 					  adapter->vfinfo[i].rss_query_enabled);
4220 	}
4221 }
4222 
4223 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
4224 {
4225 	struct ixgbe_hw *hw = &adapter->hw;
4226 	struct net_device *netdev = adapter->netdev;
4227 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4228 	struct ixgbe_ring *rx_ring;
4229 	int i;
4230 	u32 mhadd, hlreg0;
4231 
4232 #ifdef IXGBE_FCOE
4233 	/* adjust max frame to be able to do baby jumbo for FCoE */
4234 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
4235 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
4236 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4237 
4238 #endif /* IXGBE_FCOE */
4239 
4240 	/* adjust max frame to be at least the size of a standard frame */
4241 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
4242 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
4243 
4244 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
4245 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
4246 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
4247 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
4248 
4249 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
4250 	}
4251 
4252 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4253 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
4254 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4255 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4256 
4257 	/*
4258 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4259 	 * the Base and Length of the Rx Descriptor Ring
4260 	 */
4261 	for (i = 0; i < adapter->num_rx_queues; i++) {
4262 		rx_ring = adapter->rx_ring[i];
4263 
4264 		clear_ring_rsc_enabled(rx_ring);
4265 		clear_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4266 		clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4267 
4268 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4269 			set_ring_rsc_enabled(rx_ring);
4270 
4271 		if (test_bit(__IXGBE_RX_FCOE, &rx_ring->state))
4272 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4273 
4274 		clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4275 		if (adapter->flags2 & IXGBE_FLAG2_RX_LEGACY)
4276 			continue;
4277 
4278 		set_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4279 
4280 #if (PAGE_SIZE < 8192)
4281 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4282 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4283 
4284 		if (IXGBE_2K_TOO_SMALL_WITH_PADDING ||
4285 		    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
4286 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4287 #endif
4288 	}
4289 }
4290 
4291 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
4292 {
4293 	struct ixgbe_hw *hw = &adapter->hw;
4294 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
4295 
4296 	switch (hw->mac.type) {
4297 	case ixgbe_mac_82598EB:
4298 		/*
4299 		 * For VMDq support of different descriptor types or
4300 		 * buffer sizes through the use of multiple SRRCTL
4301 		 * registers, RDRXCTL.MVMEN must be set to 1
4302 		 *
4303 		 * also, the manual doesn't mention it clearly but DCA hints
4304 		 * will only use queue 0's tags unless this bit is set.  Side
4305 		 * effects of setting this bit are only that SRRCTL must be
4306 		 * fully programmed [0..15]
4307 		 */
4308 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
4309 		break;
4310 	case ixgbe_mac_X550:
4311 	case ixgbe_mac_X550EM_x:
4312 	case ixgbe_mac_x550em_a:
4313 		if (adapter->num_vfs)
4314 			rdrxctl |= IXGBE_RDRXCTL_PSP;
4315 		/* fall through */
4316 	case ixgbe_mac_82599EB:
4317 	case ixgbe_mac_X540:
4318 		/* Disable RSC for ACK packets */
4319 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
4320 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
4321 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
4322 		/* hardware requires some bits to be set by default */
4323 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
4324 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
4325 		break;
4326 	default:
4327 		/* We should do nothing since we don't know this hardware */
4328 		return;
4329 	}
4330 
4331 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
4332 }
4333 
4334 /**
4335  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
4336  * @adapter: board private structure
4337  *
4338  * Configure the Rx unit of the MAC after a reset.
4339  **/
4340 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
4341 {
4342 	struct ixgbe_hw *hw = &adapter->hw;
4343 	int i;
4344 	u32 rxctrl, rfctl;
4345 
4346 	/* disable receives while setting up the descriptors */
4347 	hw->mac.ops.disable_rx(hw);
4348 
4349 	ixgbe_setup_psrtype(adapter);
4350 	ixgbe_setup_rdrxctl(adapter);
4351 
4352 	/* RSC Setup */
4353 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
4354 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
4355 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
4356 		rfctl |= IXGBE_RFCTL_RSC_DIS;
4357 
4358 	/* disable NFS filtering */
4359 	rfctl |= (IXGBE_RFCTL_NFSW_DIS | IXGBE_RFCTL_NFSR_DIS);
4360 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
4361 
4362 	/* Program registers for the distribution of queues */
4363 	ixgbe_setup_mrqc(adapter);
4364 
4365 	/* set_rx_buffer_len must be called before ring initialization */
4366 	ixgbe_set_rx_buffer_len(adapter);
4367 
4368 	/*
4369 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4370 	 * the Base and Length of the Rx Descriptor Ring
4371 	 */
4372 	for (i = 0; i < adapter->num_rx_queues; i++)
4373 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
4374 
4375 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4376 	/* disable drop enable for 82598 parts */
4377 	if (hw->mac.type == ixgbe_mac_82598EB)
4378 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
4379 
4380 	/* enable all receives */
4381 	rxctrl |= IXGBE_RXCTRL_RXEN;
4382 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
4383 }
4384 
4385 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
4386 				 __be16 proto, u16 vid)
4387 {
4388 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4389 	struct ixgbe_hw *hw = &adapter->hw;
4390 
4391 	/* add VID to filter table */
4392 	if (!vid || !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4393 		hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true, !!vid);
4394 
4395 	set_bit(vid, adapter->active_vlans);
4396 
4397 	return 0;
4398 }
4399 
4400 static int ixgbe_find_vlvf_entry(struct ixgbe_hw *hw, u32 vlan)
4401 {
4402 	u32 vlvf;
4403 	int idx;
4404 
4405 	/* short cut the special case */
4406 	if (vlan == 0)
4407 		return 0;
4408 
4409 	/* Search for the vlan id in the VLVF entries */
4410 	for (idx = IXGBE_VLVF_ENTRIES; --idx;) {
4411 		vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(idx));
4412 		if ((vlvf & VLAN_VID_MASK) == vlan)
4413 			break;
4414 	}
4415 
4416 	return idx;
4417 }
4418 
4419 void ixgbe_update_pf_promisc_vlvf(struct ixgbe_adapter *adapter, u32 vid)
4420 {
4421 	struct ixgbe_hw *hw = &adapter->hw;
4422 	u32 bits, word;
4423 	int idx;
4424 
4425 	idx = ixgbe_find_vlvf_entry(hw, vid);
4426 	if (!idx)
4427 		return;
4428 
4429 	/* See if any other pools are set for this VLAN filter
4430 	 * entry other than the PF.
4431 	 */
4432 	word = idx * 2 + (VMDQ_P(0) / 32);
4433 	bits = ~BIT(VMDQ_P(0) % 32);
4434 	bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4435 
4436 	/* Disable the filter so this falls into the default pool. */
4437 	if (!bits && !IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1))) {
4438 		if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4439 			IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), 0);
4440 		IXGBE_WRITE_REG(hw, IXGBE_VLVF(idx), 0);
4441 	}
4442 }
4443 
4444 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
4445 				  __be16 proto, u16 vid)
4446 {
4447 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4448 	struct ixgbe_hw *hw = &adapter->hw;
4449 
4450 	/* remove VID from filter table */
4451 	if (vid && !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4452 		hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), false, true);
4453 
4454 	clear_bit(vid, adapter->active_vlans);
4455 
4456 	return 0;
4457 }
4458 
4459 /**
4460  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
4461  * @adapter: driver data
4462  */
4463 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
4464 {
4465 	struct ixgbe_hw *hw = &adapter->hw;
4466 	u32 vlnctrl;
4467 	int i, j;
4468 
4469 	switch (hw->mac.type) {
4470 	case ixgbe_mac_82598EB:
4471 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4472 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
4473 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4474 		break;
4475 	case ixgbe_mac_82599EB:
4476 	case ixgbe_mac_X540:
4477 	case ixgbe_mac_X550:
4478 	case ixgbe_mac_X550EM_x:
4479 	case ixgbe_mac_x550em_a:
4480 		for (i = 0; i < adapter->num_rx_queues; i++) {
4481 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4482 
4483 			if (!netif_is_ixgbe(ring->netdev))
4484 				continue;
4485 
4486 			j = ring->reg_idx;
4487 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4488 			vlnctrl &= ~IXGBE_RXDCTL_VME;
4489 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4490 		}
4491 		break;
4492 	default:
4493 		break;
4494 	}
4495 }
4496 
4497 /**
4498  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
4499  * @adapter: driver data
4500  */
4501 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
4502 {
4503 	struct ixgbe_hw *hw = &adapter->hw;
4504 	u32 vlnctrl;
4505 	int i, j;
4506 
4507 	switch (hw->mac.type) {
4508 	case ixgbe_mac_82598EB:
4509 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4510 		vlnctrl |= IXGBE_VLNCTRL_VME;
4511 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4512 		break;
4513 	case ixgbe_mac_82599EB:
4514 	case ixgbe_mac_X540:
4515 	case ixgbe_mac_X550:
4516 	case ixgbe_mac_X550EM_x:
4517 	case ixgbe_mac_x550em_a:
4518 		for (i = 0; i < adapter->num_rx_queues; i++) {
4519 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4520 
4521 			if (!netif_is_ixgbe(ring->netdev))
4522 				continue;
4523 
4524 			j = ring->reg_idx;
4525 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4526 			vlnctrl |= IXGBE_RXDCTL_VME;
4527 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4528 		}
4529 		break;
4530 	default:
4531 		break;
4532 	}
4533 }
4534 
4535 static void ixgbe_vlan_promisc_enable(struct ixgbe_adapter *adapter)
4536 {
4537 	struct ixgbe_hw *hw = &adapter->hw;
4538 	u32 vlnctrl, i;
4539 
4540 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4541 
4542 	if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) {
4543 	/* For VMDq and SR-IOV we must leave VLAN filtering enabled */
4544 		vlnctrl |= IXGBE_VLNCTRL_VFE;
4545 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4546 	} else {
4547 		vlnctrl &= ~IXGBE_VLNCTRL_VFE;
4548 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4549 		return;
4550 	}
4551 
4552 	/* Nothing to do for 82598 */
4553 	if (hw->mac.type == ixgbe_mac_82598EB)
4554 		return;
4555 
4556 	/* We are already in VLAN promisc, nothing to do */
4557 	if (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC)
4558 		return;
4559 
4560 	/* Set flag so we don't redo unnecessary work */
4561 	adapter->flags2 |= IXGBE_FLAG2_VLAN_PROMISC;
4562 
4563 	/* Add PF to all active pools */
4564 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4565 		u32 reg_offset = IXGBE_VLVFB(i * 2 + VMDQ_P(0) / 32);
4566 		u32 vlvfb = IXGBE_READ_REG(hw, reg_offset);
4567 
4568 		vlvfb |= BIT(VMDQ_P(0) % 32);
4569 		IXGBE_WRITE_REG(hw, reg_offset, vlvfb);
4570 	}
4571 
4572 	/* Set all bits in the VLAN filter table array */
4573 	for (i = hw->mac.vft_size; i--;)
4574 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), ~0U);
4575 }
4576 
4577 #define VFTA_BLOCK_SIZE 8
4578 static void ixgbe_scrub_vfta(struct ixgbe_adapter *adapter, u32 vfta_offset)
4579 {
4580 	struct ixgbe_hw *hw = &adapter->hw;
4581 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4582 	u32 vid_start = vfta_offset * 32;
4583 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4584 	u32 i, vid, word, bits;
4585 
4586 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4587 		u32 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
4588 
4589 		/* pull VLAN ID from VLVF */
4590 		vid = vlvf & VLAN_VID_MASK;
4591 
4592 		/* only concern outselves with a certain range */
4593 		if (vid < vid_start || vid >= vid_end)
4594 			continue;
4595 
4596 		if (vlvf) {
4597 			/* record VLAN ID in VFTA */
4598 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4599 
4600 			/* if PF is part of this then continue */
4601 			if (test_bit(vid, adapter->active_vlans))
4602 				continue;
4603 		}
4604 
4605 		/* remove PF from the pool */
4606 		word = i * 2 + VMDQ_P(0) / 32;
4607 		bits = ~BIT(VMDQ_P(0) % 32);
4608 		bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4609 		IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), bits);
4610 	}
4611 
4612 	/* extract values from active_vlans and write back to VFTA */
4613 	for (i = VFTA_BLOCK_SIZE; i--;) {
4614 		vid = (vfta_offset + i) * 32;
4615 		word = vid / BITS_PER_LONG;
4616 		bits = vid % BITS_PER_LONG;
4617 
4618 		vfta[i] |= adapter->active_vlans[word] >> bits;
4619 
4620 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(vfta_offset + i), vfta[i]);
4621 	}
4622 }
4623 
4624 static void ixgbe_vlan_promisc_disable(struct ixgbe_adapter *adapter)
4625 {
4626 	struct ixgbe_hw *hw = &adapter->hw;
4627 	u32 vlnctrl, i;
4628 
4629 	/* Set VLAN filtering to enabled */
4630 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4631 	vlnctrl |= IXGBE_VLNCTRL_VFE;
4632 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4633 
4634 	if (!(adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) ||
4635 	    hw->mac.type == ixgbe_mac_82598EB)
4636 		return;
4637 
4638 	/* We are not in VLAN promisc, nothing to do */
4639 	if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4640 		return;
4641 
4642 	/* Set flag so we don't redo unnecessary work */
4643 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4644 
4645 	for (i = 0; i < hw->mac.vft_size; i += VFTA_BLOCK_SIZE)
4646 		ixgbe_scrub_vfta(adapter, i);
4647 }
4648 
4649 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
4650 {
4651 	u16 vid = 1;
4652 
4653 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
4654 
4655 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
4656 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
4657 }
4658 
4659 /**
4660  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4661  * @netdev: network interface device structure
4662  *
4663  * Writes multicast address list to the MTA hash table.
4664  * Returns: -ENOMEM on failure
4665  *                0 on no addresses written
4666  *                X on writing X addresses to MTA
4667  **/
4668 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
4669 {
4670 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4671 	struct ixgbe_hw *hw = &adapter->hw;
4672 
4673 	if (!netif_running(netdev))
4674 		return 0;
4675 
4676 	if (hw->mac.ops.update_mc_addr_list)
4677 		hw->mac.ops.update_mc_addr_list(hw, netdev);
4678 	else
4679 		return -ENOMEM;
4680 
4681 #ifdef CONFIG_PCI_IOV
4682 	ixgbe_restore_vf_multicasts(adapter);
4683 #endif
4684 
4685 	return netdev_mc_count(netdev);
4686 }
4687 
4688 #ifdef CONFIG_PCI_IOV
4689 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4690 {
4691 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4692 	struct ixgbe_hw *hw = &adapter->hw;
4693 	int i;
4694 
4695 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4696 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4697 
4698 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4699 			hw->mac.ops.set_rar(hw, i,
4700 					    mac_table->addr,
4701 					    mac_table->pool,
4702 					    IXGBE_RAH_AV);
4703 		else
4704 			hw->mac.ops.clear_rar(hw, i);
4705 	}
4706 }
4707 
4708 #endif
4709 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4710 {
4711 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4712 	struct ixgbe_hw *hw = &adapter->hw;
4713 	int i;
4714 
4715 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4716 		if (!(mac_table->state & IXGBE_MAC_STATE_MODIFIED))
4717 			continue;
4718 
4719 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4720 
4721 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4722 			hw->mac.ops.set_rar(hw, i,
4723 					    mac_table->addr,
4724 					    mac_table->pool,
4725 					    IXGBE_RAH_AV);
4726 		else
4727 			hw->mac.ops.clear_rar(hw, i);
4728 	}
4729 }
4730 
4731 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4732 {
4733 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4734 	struct ixgbe_hw *hw = &adapter->hw;
4735 	int i;
4736 
4737 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4738 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4739 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4740 	}
4741 
4742 	ixgbe_sync_mac_table(adapter);
4743 }
4744 
4745 static int ixgbe_available_rars(struct ixgbe_adapter *adapter, u16 pool)
4746 {
4747 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4748 	struct ixgbe_hw *hw = &adapter->hw;
4749 	int i, count = 0;
4750 
4751 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4752 		/* do not count default RAR as available */
4753 		if (mac_table->state & IXGBE_MAC_STATE_DEFAULT)
4754 			continue;
4755 
4756 		/* only count unused and addresses that belong to us */
4757 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE) {
4758 			if (mac_table->pool != pool)
4759 				continue;
4760 		}
4761 
4762 		count++;
4763 	}
4764 
4765 	return count;
4766 }
4767 
4768 /* this function destroys the first RAR entry */
4769 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter)
4770 {
4771 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4772 	struct ixgbe_hw *hw = &adapter->hw;
4773 
4774 	memcpy(&mac_table->addr, hw->mac.addr, ETH_ALEN);
4775 	mac_table->pool = VMDQ_P(0);
4776 
4777 	mac_table->state = IXGBE_MAC_STATE_DEFAULT | IXGBE_MAC_STATE_IN_USE;
4778 
4779 	hw->mac.ops.set_rar(hw, 0, mac_table->addr, mac_table->pool,
4780 			    IXGBE_RAH_AV);
4781 }
4782 
4783 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter,
4784 			 const u8 *addr, u16 pool)
4785 {
4786 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4787 	struct ixgbe_hw *hw = &adapter->hw;
4788 	int i;
4789 
4790 	if (is_zero_ether_addr(addr))
4791 		return -EINVAL;
4792 
4793 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4794 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4795 			continue;
4796 
4797 		ether_addr_copy(mac_table->addr, addr);
4798 		mac_table->pool = pool;
4799 
4800 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED |
4801 				    IXGBE_MAC_STATE_IN_USE;
4802 
4803 		ixgbe_sync_mac_table(adapter);
4804 
4805 		return i;
4806 	}
4807 
4808 	return -ENOMEM;
4809 }
4810 
4811 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter,
4812 			 const u8 *addr, u16 pool)
4813 {
4814 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4815 	struct ixgbe_hw *hw = &adapter->hw;
4816 	int i;
4817 
4818 	if (is_zero_ether_addr(addr))
4819 		return -EINVAL;
4820 
4821 	/* search table for addr, if found clear IN_USE flag and sync */
4822 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4823 		/* we can only delete an entry if it is in use */
4824 		if (!(mac_table->state & IXGBE_MAC_STATE_IN_USE))
4825 			continue;
4826 		/* we only care about entries that belong to the given pool */
4827 		if (mac_table->pool != pool)
4828 			continue;
4829 		/* we only care about a specific MAC address */
4830 		if (!ether_addr_equal(addr, mac_table->addr))
4831 			continue;
4832 
4833 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4834 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4835 
4836 		ixgbe_sync_mac_table(adapter);
4837 
4838 		return 0;
4839 	}
4840 
4841 	return -ENOMEM;
4842 }
4843 
4844 static int ixgbe_uc_sync(struct net_device *netdev, const unsigned char *addr)
4845 {
4846 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4847 	int ret;
4848 
4849 	ret = ixgbe_add_mac_filter(adapter, addr, VMDQ_P(0));
4850 
4851 	return min_t(int, ret, 0);
4852 }
4853 
4854 static int ixgbe_uc_unsync(struct net_device *netdev, const unsigned char *addr)
4855 {
4856 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4857 
4858 	ixgbe_del_mac_filter(adapter, addr, VMDQ_P(0));
4859 
4860 	return 0;
4861 }
4862 
4863 /**
4864  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4865  * @netdev: network interface device structure
4866  *
4867  * The set_rx_method entry point is called whenever the unicast/multicast
4868  * address list or the network interface flags are updated.  This routine is
4869  * responsible for configuring the hardware for proper unicast, multicast and
4870  * promiscuous mode.
4871  **/
4872 void ixgbe_set_rx_mode(struct net_device *netdev)
4873 {
4874 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4875 	struct ixgbe_hw *hw = &adapter->hw;
4876 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4877 	netdev_features_t features = netdev->features;
4878 	int count;
4879 
4880 	/* Check for Promiscuous and All Multicast modes */
4881 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4882 
4883 	/* set all bits that we expect to always be set */
4884 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4885 	fctrl |= IXGBE_FCTRL_BAM;
4886 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4887 	fctrl |= IXGBE_FCTRL_PMCF;
4888 
4889 	/* clear the bits we are changing the status of */
4890 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4891 	if (netdev->flags & IFF_PROMISC) {
4892 		hw->addr_ctrl.user_set_promisc = true;
4893 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4894 		vmolr |= IXGBE_VMOLR_MPE;
4895 		features &= ~NETIF_F_HW_VLAN_CTAG_FILTER;
4896 	} else {
4897 		if (netdev->flags & IFF_ALLMULTI) {
4898 			fctrl |= IXGBE_FCTRL_MPE;
4899 			vmolr |= IXGBE_VMOLR_MPE;
4900 		}
4901 		hw->addr_ctrl.user_set_promisc = false;
4902 	}
4903 
4904 	/*
4905 	 * Write addresses to available RAR registers, if there is not
4906 	 * sufficient space to store all the addresses then enable
4907 	 * unicast promiscuous mode
4908 	 */
4909 	if (__dev_uc_sync(netdev, ixgbe_uc_sync, ixgbe_uc_unsync)) {
4910 		fctrl |= IXGBE_FCTRL_UPE;
4911 		vmolr |= IXGBE_VMOLR_ROPE;
4912 	}
4913 
4914 	/* Write addresses to the MTA, if the attempt fails
4915 	 * then we should just turn on promiscuous mode so
4916 	 * that we can at least receive multicast traffic
4917 	 */
4918 	count = ixgbe_write_mc_addr_list(netdev);
4919 	if (count < 0) {
4920 		fctrl |= IXGBE_FCTRL_MPE;
4921 		vmolr |= IXGBE_VMOLR_MPE;
4922 	} else if (count) {
4923 		vmolr |= IXGBE_VMOLR_ROMPE;
4924 	}
4925 
4926 	if (hw->mac.type != ixgbe_mac_82598EB) {
4927 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4928 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4929 			   IXGBE_VMOLR_ROPE);
4930 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4931 	}
4932 
4933 	/* This is useful for sniffing bad packets. */
4934 	if (features & NETIF_F_RXALL) {
4935 		/* UPE and MPE will be handled by normal PROMISC logic
4936 		 * in e1000e_set_rx_mode */
4937 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4938 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4939 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4940 
4941 		fctrl &= ~(IXGBE_FCTRL_DPF);
4942 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
4943 	}
4944 
4945 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4946 
4947 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
4948 		ixgbe_vlan_strip_enable(adapter);
4949 	else
4950 		ixgbe_vlan_strip_disable(adapter);
4951 
4952 	if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
4953 		ixgbe_vlan_promisc_disable(adapter);
4954 	else
4955 		ixgbe_vlan_promisc_enable(adapter);
4956 }
4957 
4958 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4959 {
4960 	int q_idx;
4961 
4962 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
4963 		napi_enable(&adapter->q_vector[q_idx]->napi);
4964 }
4965 
4966 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
4967 {
4968 	int q_idx;
4969 
4970 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
4971 		napi_disable(&adapter->q_vector[q_idx]->napi);
4972 }
4973 
4974 static void ixgbe_clear_udp_tunnel_port(struct ixgbe_adapter *adapter, u32 mask)
4975 {
4976 	struct ixgbe_hw *hw = &adapter->hw;
4977 	u32 vxlanctrl;
4978 
4979 	if (!(adapter->flags & (IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE |
4980 				IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)))
4981 		return;
4982 
4983 	vxlanctrl = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) & ~mask;
4984 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, vxlanctrl);
4985 
4986 	if (mask & IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK)
4987 		adapter->vxlan_port = 0;
4988 
4989 	if (mask & IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK)
4990 		adapter->geneve_port = 0;
4991 }
4992 
4993 #ifdef CONFIG_IXGBE_DCB
4994 /**
4995  * ixgbe_configure_dcb - Configure DCB hardware
4996  * @adapter: ixgbe adapter struct
4997  *
4998  * This is called by the driver on open to configure the DCB hardware.
4999  * This is also called by the gennetlink interface when reconfiguring
5000  * the DCB state.
5001  */
5002 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
5003 {
5004 	struct ixgbe_hw *hw = &adapter->hw;
5005 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
5006 
5007 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
5008 		if (hw->mac.type == ixgbe_mac_82598EB)
5009 			netif_set_gso_max_size(adapter->netdev, 65536);
5010 		return;
5011 	}
5012 
5013 	if (hw->mac.type == ixgbe_mac_82598EB)
5014 		netif_set_gso_max_size(adapter->netdev, 32768);
5015 
5016 #ifdef IXGBE_FCOE
5017 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
5018 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
5019 #endif
5020 
5021 	/* reconfigure the hardware */
5022 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
5023 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5024 						DCB_TX_CONFIG);
5025 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5026 						DCB_RX_CONFIG);
5027 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
5028 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
5029 		ixgbe_dcb_hw_ets(&adapter->hw,
5030 				 adapter->ixgbe_ieee_ets,
5031 				 max_frame);
5032 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
5033 					adapter->ixgbe_ieee_pfc->pfc_en,
5034 					adapter->ixgbe_ieee_ets->prio_tc);
5035 	}
5036 
5037 	/* Enable RSS Hash per TC */
5038 	if (hw->mac.type != ixgbe_mac_82598EB) {
5039 		u32 msb = 0;
5040 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
5041 
5042 		while (rss_i) {
5043 			msb++;
5044 			rss_i >>= 1;
5045 		}
5046 
5047 		/* write msb to all 8 TCs in one write */
5048 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
5049 	}
5050 }
5051 #endif
5052 
5053 /* Additional bittime to account for IXGBE framing */
5054 #define IXGBE_ETH_FRAMING 20
5055 
5056 /**
5057  * ixgbe_hpbthresh - calculate high water mark for flow control
5058  *
5059  * @adapter: board private structure to calculate for
5060  * @pb: packet buffer to calculate
5061  */
5062 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
5063 {
5064 	struct ixgbe_hw *hw = &adapter->hw;
5065 	struct net_device *dev = adapter->netdev;
5066 	int link, tc, kb, marker;
5067 	u32 dv_id, rx_pba;
5068 
5069 	/* Calculate max LAN frame size */
5070 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
5071 
5072 #ifdef IXGBE_FCOE
5073 	/* FCoE traffic class uses FCOE jumbo frames */
5074 	if ((dev->features & NETIF_F_FCOE_MTU) &&
5075 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5076 	    (pb == ixgbe_fcoe_get_tc(adapter)))
5077 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5078 #endif
5079 
5080 	/* Calculate delay value for device */
5081 	switch (hw->mac.type) {
5082 	case ixgbe_mac_X540:
5083 	case ixgbe_mac_X550:
5084 	case ixgbe_mac_X550EM_x:
5085 	case ixgbe_mac_x550em_a:
5086 		dv_id = IXGBE_DV_X540(link, tc);
5087 		break;
5088 	default:
5089 		dv_id = IXGBE_DV(link, tc);
5090 		break;
5091 	}
5092 
5093 	/* Loopback switch introduces additional latency */
5094 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5095 		dv_id += IXGBE_B2BT(tc);
5096 
5097 	/* Delay value is calculated in bit times convert to KB */
5098 	kb = IXGBE_BT2KB(dv_id);
5099 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
5100 
5101 	marker = rx_pba - kb;
5102 
5103 	/* It is possible that the packet buffer is not large enough
5104 	 * to provide required headroom. In this case throw an error
5105 	 * to user and a do the best we can.
5106 	 */
5107 	if (marker < 0) {
5108 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
5109 			    "headroom to support flow control."
5110 			    "Decrease MTU or number of traffic classes\n", pb);
5111 		marker = tc + 1;
5112 	}
5113 
5114 	return marker;
5115 }
5116 
5117 /**
5118  * ixgbe_lpbthresh - calculate low water mark for for flow control
5119  *
5120  * @adapter: board private structure to calculate for
5121  * @pb: packet buffer to calculate
5122  */
5123 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
5124 {
5125 	struct ixgbe_hw *hw = &adapter->hw;
5126 	struct net_device *dev = adapter->netdev;
5127 	int tc;
5128 	u32 dv_id;
5129 
5130 	/* Calculate max LAN frame size */
5131 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
5132 
5133 #ifdef IXGBE_FCOE
5134 	/* FCoE traffic class uses FCOE jumbo frames */
5135 	if ((dev->features & NETIF_F_FCOE_MTU) &&
5136 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5137 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
5138 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5139 #endif
5140 
5141 	/* Calculate delay value for device */
5142 	switch (hw->mac.type) {
5143 	case ixgbe_mac_X540:
5144 	case ixgbe_mac_X550:
5145 	case ixgbe_mac_X550EM_x:
5146 	case ixgbe_mac_x550em_a:
5147 		dv_id = IXGBE_LOW_DV_X540(tc);
5148 		break;
5149 	default:
5150 		dv_id = IXGBE_LOW_DV(tc);
5151 		break;
5152 	}
5153 
5154 	/* Delay value is calculated in bit times convert to KB */
5155 	return IXGBE_BT2KB(dv_id);
5156 }
5157 
5158 /*
5159  * ixgbe_pbthresh_setup - calculate and setup high low water marks
5160  */
5161 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
5162 {
5163 	struct ixgbe_hw *hw = &adapter->hw;
5164 	int num_tc = adapter->hw_tcs;
5165 	int i;
5166 
5167 	if (!num_tc)
5168 		num_tc = 1;
5169 
5170 	for (i = 0; i < num_tc; i++) {
5171 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
5172 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
5173 
5174 		/* Low water marks must not be larger than high water marks */
5175 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
5176 			hw->fc.low_water[i] = 0;
5177 	}
5178 
5179 	for (; i < MAX_TRAFFIC_CLASS; i++)
5180 		hw->fc.high_water[i] = 0;
5181 }
5182 
5183 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
5184 {
5185 	struct ixgbe_hw *hw = &adapter->hw;
5186 	int hdrm;
5187 	u8 tc = adapter->hw_tcs;
5188 
5189 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
5190 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
5191 		hdrm = 32 << adapter->fdir_pballoc;
5192 	else
5193 		hdrm = 0;
5194 
5195 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
5196 	ixgbe_pbthresh_setup(adapter);
5197 }
5198 
5199 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
5200 {
5201 	struct ixgbe_hw *hw = &adapter->hw;
5202 	struct hlist_node *node2;
5203 	struct ixgbe_fdir_filter *filter;
5204 
5205 	spin_lock(&adapter->fdir_perfect_lock);
5206 
5207 	if (!hlist_empty(&adapter->fdir_filter_list))
5208 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
5209 
5210 	hlist_for_each_entry_safe(filter, node2,
5211 				  &adapter->fdir_filter_list, fdir_node) {
5212 		ixgbe_fdir_write_perfect_filter_82599(hw,
5213 				&filter->filter,
5214 				filter->sw_idx,
5215 				(filter->action == IXGBE_FDIR_DROP_QUEUE) ?
5216 				IXGBE_FDIR_DROP_QUEUE :
5217 				adapter->rx_ring[filter->action]->reg_idx);
5218 	}
5219 
5220 	spin_unlock(&adapter->fdir_perfect_lock);
5221 }
5222 
5223 /**
5224  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
5225  * @rx_ring: ring to free buffers from
5226  **/
5227 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
5228 {
5229 	u16 i = rx_ring->next_to_clean;
5230 	struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
5231 
5232 	/* Free all the Rx ring sk_buffs */
5233 	while (i != rx_ring->next_to_alloc) {
5234 		if (rx_buffer->skb) {
5235 			struct sk_buff *skb = rx_buffer->skb;
5236 			if (IXGBE_CB(skb)->page_released)
5237 				dma_unmap_page_attrs(rx_ring->dev,
5238 						     IXGBE_CB(skb)->dma,
5239 						     ixgbe_rx_pg_size(rx_ring),
5240 						     DMA_FROM_DEVICE,
5241 						     IXGBE_RX_DMA_ATTR);
5242 			dev_kfree_skb(skb);
5243 		}
5244 
5245 		/* Invalidate cache lines that may have been written to by
5246 		 * device so that we avoid corrupting memory.
5247 		 */
5248 		dma_sync_single_range_for_cpu(rx_ring->dev,
5249 					      rx_buffer->dma,
5250 					      rx_buffer->page_offset,
5251 					      ixgbe_rx_bufsz(rx_ring),
5252 					      DMA_FROM_DEVICE);
5253 
5254 		/* free resources associated with mapping */
5255 		dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
5256 				     ixgbe_rx_pg_size(rx_ring),
5257 				     DMA_FROM_DEVICE,
5258 				     IXGBE_RX_DMA_ATTR);
5259 		__page_frag_cache_drain(rx_buffer->page,
5260 					rx_buffer->pagecnt_bias);
5261 
5262 		i++;
5263 		rx_buffer++;
5264 		if (i == rx_ring->count) {
5265 			i = 0;
5266 			rx_buffer = rx_ring->rx_buffer_info;
5267 		}
5268 	}
5269 
5270 	rx_ring->next_to_alloc = 0;
5271 	rx_ring->next_to_clean = 0;
5272 	rx_ring->next_to_use = 0;
5273 }
5274 
5275 static int ixgbe_fwd_ring_up(struct ixgbe_adapter *adapter,
5276 			     struct ixgbe_fwd_adapter *accel)
5277 {
5278 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
5279 	int num_tc = netdev_get_num_tc(adapter->netdev);
5280 	struct net_device *vdev = accel->netdev;
5281 	int i, baseq, err;
5282 
5283 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
5284 	netdev_dbg(vdev, "pool %i:%i queues %i:%i\n",
5285 		   accel->pool, adapter->num_rx_pools,
5286 		   baseq, baseq + adapter->num_rx_queues_per_pool);
5287 
5288 	accel->rx_base_queue = baseq;
5289 	accel->tx_base_queue = baseq;
5290 
5291 	/* record configuration for macvlan interface in vdev */
5292 	for (i = 0; i < num_tc; i++)
5293 		netdev_bind_sb_channel_queue(adapter->netdev, vdev,
5294 					     i, rss_i, baseq + (rss_i * i));
5295 
5296 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5297 		adapter->rx_ring[baseq + i]->netdev = vdev;
5298 
5299 	/* Guarantee all rings are updated before we update the
5300 	 * MAC address filter.
5301 	 */
5302 	wmb();
5303 
5304 	/* ixgbe_add_mac_filter will return an index if it succeeds, so we
5305 	 * need to only treat it as an error value if it is negative.
5306 	 */
5307 	err = ixgbe_add_mac_filter(adapter, vdev->dev_addr,
5308 				   VMDQ_P(accel->pool));
5309 	if (err >= 0)
5310 		return 0;
5311 
5312 	/* if we cannot add the MAC rule then disable the offload */
5313 	macvlan_release_l2fw_offload(vdev);
5314 
5315 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5316 		adapter->rx_ring[baseq + i]->netdev = NULL;
5317 
5318 	netdev_err(vdev, "L2FW offload disabled due to L2 filter error\n");
5319 
5320 	/* unbind the queues and drop the subordinate channel config */
5321 	netdev_unbind_sb_channel(adapter->netdev, vdev);
5322 	netdev_set_sb_channel(vdev, 0);
5323 
5324 	clear_bit(accel->pool, adapter->fwd_bitmask);
5325 	kfree(accel);
5326 
5327 	return err;
5328 }
5329 
5330 static int ixgbe_macvlan_up(struct net_device *vdev, void *data)
5331 {
5332 	struct ixgbe_adapter *adapter = data;
5333 	struct ixgbe_fwd_adapter *accel;
5334 
5335 	if (!netif_is_macvlan(vdev))
5336 		return 0;
5337 
5338 	accel = macvlan_accel_priv(vdev);
5339 	if (!accel)
5340 		return 0;
5341 
5342 	ixgbe_fwd_ring_up(adapter, accel);
5343 
5344 	return 0;
5345 }
5346 
5347 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
5348 {
5349 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
5350 				      ixgbe_macvlan_up, adapter);
5351 }
5352 
5353 static void ixgbe_configure(struct ixgbe_adapter *adapter)
5354 {
5355 	struct ixgbe_hw *hw = &adapter->hw;
5356 
5357 	ixgbe_configure_pb(adapter);
5358 #ifdef CONFIG_IXGBE_DCB
5359 	ixgbe_configure_dcb(adapter);
5360 #endif
5361 	/*
5362 	 * We must restore virtualization before VLANs or else
5363 	 * the VLVF registers will not be populated
5364 	 */
5365 	ixgbe_configure_virtualization(adapter);
5366 
5367 	ixgbe_set_rx_mode(adapter->netdev);
5368 	ixgbe_restore_vlan(adapter);
5369 	ixgbe_ipsec_restore(adapter);
5370 
5371 	switch (hw->mac.type) {
5372 	case ixgbe_mac_82599EB:
5373 	case ixgbe_mac_X540:
5374 		hw->mac.ops.disable_rx_buff(hw);
5375 		break;
5376 	default:
5377 		break;
5378 	}
5379 
5380 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
5381 		ixgbe_init_fdir_signature_82599(&adapter->hw,
5382 						adapter->fdir_pballoc);
5383 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
5384 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
5385 					      adapter->fdir_pballoc);
5386 		ixgbe_fdir_filter_restore(adapter);
5387 	}
5388 
5389 	switch (hw->mac.type) {
5390 	case ixgbe_mac_82599EB:
5391 	case ixgbe_mac_X540:
5392 		hw->mac.ops.enable_rx_buff(hw);
5393 		break;
5394 	default:
5395 		break;
5396 	}
5397 
5398 #ifdef CONFIG_IXGBE_DCA
5399 	/* configure DCA */
5400 	if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
5401 		ixgbe_setup_dca(adapter);
5402 #endif /* CONFIG_IXGBE_DCA */
5403 
5404 #ifdef IXGBE_FCOE
5405 	/* configure FCoE L2 filters, redirection table, and Rx control */
5406 	ixgbe_configure_fcoe(adapter);
5407 
5408 #endif /* IXGBE_FCOE */
5409 	ixgbe_configure_tx(adapter);
5410 	ixgbe_configure_rx(adapter);
5411 	ixgbe_configure_dfwd(adapter);
5412 }
5413 
5414 /**
5415  * ixgbe_sfp_link_config - set up SFP+ link
5416  * @adapter: pointer to private adapter struct
5417  **/
5418 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
5419 {
5420 	/*
5421 	 * We are assuming the worst case scenario here, and that
5422 	 * is that an SFP was inserted/removed after the reset
5423 	 * but before SFP detection was enabled.  As such the best
5424 	 * solution is to just start searching as soon as we start
5425 	 */
5426 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5427 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5428 
5429 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5430 	adapter->sfp_poll_time = 0;
5431 }
5432 
5433 /**
5434  * ixgbe_non_sfp_link_config - set up non-SFP+ link
5435  * @hw: pointer to private hardware struct
5436  *
5437  * Returns 0 on success, negative on failure
5438  **/
5439 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
5440 {
5441 	u32 speed;
5442 	bool autoneg, link_up = false;
5443 	int ret = IXGBE_ERR_LINK_SETUP;
5444 
5445 	if (hw->mac.ops.check_link)
5446 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
5447 
5448 	if (ret)
5449 		return ret;
5450 
5451 	speed = hw->phy.autoneg_advertised;
5452 	if ((!speed) && (hw->mac.ops.get_link_capabilities))
5453 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
5454 							&autoneg);
5455 	if (ret)
5456 		return ret;
5457 
5458 	if (hw->mac.ops.setup_link)
5459 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
5460 
5461 	return ret;
5462 }
5463 
5464 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
5465 {
5466 	struct ixgbe_hw *hw = &adapter->hw;
5467 	u32 gpie = 0;
5468 
5469 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
5470 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5471 		       IXGBE_GPIE_OCD;
5472 		gpie |= IXGBE_GPIE_EIAME;
5473 		/*
5474 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
5475 		 * this saves a register write for every interrupt
5476 		 */
5477 		switch (hw->mac.type) {
5478 		case ixgbe_mac_82598EB:
5479 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5480 			break;
5481 		case ixgbe_mac_82599EB:
5482 		case ixgbe_mac_X540:
5483 		case ixgbe_mac_X550:
5484 		case ixgbe_mac_X550EM_x:
5485 		case ixgbe_mac_x550em_a:
5486 		default:
5487 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5488 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5489 			break;
5490 		}
5491 	} else {
5492 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
5493 		 * specifically only auto mask tx and rx interrupts */
5494 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5495 	}
5496 
5497 	/* XXX: to interrupt immediately for EICS writes, enable this */
5498 	/* gpie |= IXGBE_GPIE_EIMEN; */
5499 
5500 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
5501 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
5502 
5503 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
5504 		case IXGBE_82599_VMDQ_8Q_MASK:
5505 			gpie |= IXGBE_GPIE_VTMODE_16;
5506 			break;
5507 		case IXGBE_82599_VMDQ_4Q_MASK:
5508 			gpie |= IXGBE_GPIE_VTMODE_32;
5509 			break;
5510 		default:
5511 			gpie |= IXGBE_GPIE_VTMODE_64;
5512 			break;
5513 		}
5514 	}
5515 
5516 	/* Enable Thermal over heat sensor interrupt */
5517 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
5518 		switch (adapter->hw.mac.type) {
5519 		case ixgbe_mac_82599EB:
5520 			gpie |= IXGBE_SDP0_GPIEN_8259X;
5521 			break;
5522 		default:
5523 			break;
5524 		}
5525 	}
5526 
5527 	/* Enable fan failure interrupt */
5528 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
5529 		gpie |= IXGBE_SDP1_GPIEN(hw);
5530 
5531 	switch (hw->mac.type) {
5532 	case ixgbe_mac_82599EB:
5533 		gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
5534 		break;
5535 	case ixgbe_mac_X550EM_x:
5536 	case ixgbe_mac_x550em_a:
5537 		gpie |= IXGBE_SDP0_GPIEN_X540;
5538 		break;
5539 	default:
5540 		break;
5541 	}
5542 
5543 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5544 }
5545 
5546 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
5547 {
5548 	struct ixgbe_hw *hw = &adapter->hw;
5549 	int err;
5550 	u32 ctrl_ext;
5551 
5552 	ixgbe_get_hw_control(adapter);
5553 	ixgbe_setup_gpie(adapter);
5554 
5555 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
5556 		ixgbe_configure_msix(adapter);
5557 	else
5558 		ixgbe_configure_msi_and_legacy(adapter);
5559 
5560 	/* enable the optics for 82599 SFP+ fiber */
5561 	if (hw->mac.ops.enable_tx_laser)
5562 		hw->mac.ops.enable_tx_laser(hw);
5563 
5564 	if (hw->phy.ops.set_phy_power)
5565 		hw->phy.ops.set_phy_power(hw, true);
5566 
5567 	smp_mb__before_atomic();
5568 	clear_bit(__IXGBE_DOWN, &adapter->state);
5569 	ixgbe_napi_enable_all(adapter);
5570 
5571 	if (ixgbe_is_sfp(hw)) {
5572 		ixgbe_sfp_link_config(adapter);
5573 	} else {
5574 		err = ixgbe_non_sfp_link_config(hw);
5575 		if (err)
5576 			e_err(probe, "link_config FAILED %d\n", err);
5577 	}
5578 
5579 	/* clear any pending interrupts, may auto mask */
5580 	IXGBE_READ_REG(hw, IXGBE_EICR);
5581 	ixgbe_irq_enable(adapter, true, true);
5582 
5583 	/*
5584 	 * If this adapter has a fan, check to see if we had a failure
5585 	 * before we enabled the interrupt.
5586 	 */
5587 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
5588 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
5589 		if (esdp & IXGBE_ESDP_SDP1)
5590 			e_crit(drv, "Fan has stopped, replace the adapter\n");
5591 	}
5592 
5593 	/* bring the link up in the watchdog, this could race with our first
5594 	 * link up interrupt but shouldn't be a problem */
5595 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5596 	adapter->link_check_timeout = jiffies;
5597 	mod_timer(&adapter->service_timer, jiffies);
5598 
5599 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
5600 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
5601 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
5602 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
5603 }
5604 
5605 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
5606 {
5607 	WARN_ON(in_interrupt());
5608 	/* put off any impending NetWatchDogTimeout */
5609 	netif_trans_update(adapter->netdev);
5610 
5611 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
5612 		usleep_range(1000, 2000);
5613 	if (adapter->hw.phy.type == ixgbe_phy_fw)
5614 		ixgbe_watchdog_link_is_down(adapter);
5615 	ixgbe_down(adapter);
5616 	/*
5617 	 * If SR-IOV enabled then wait a bit before bringing the adapter
5618 	 * back up to give the VFs time to respond to the reset.  The
5619 	 * two second wait is based upon the watchdog timer cycle in
5620 	 * the VF driver.
5621 	 */
5622 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5623 		msleep(2000);
5624 	ixgbe_up(adapter);
5625 	clear_bit(__IXGBE_RESETTING, &adapter->state);
5626 }
5627 
5628 void ixgbe_up(struct ixgbe_adapter *adapter)
5629 {
5630 	/* hardware has been reset, we need to reload some things */
5631 	ixgbe_configure(adapter);
5632 
5633 	ixgbe_up_complete(adapter);
5634 }
5635 
5636 void ixgbe_reset(struct ixgbe_adapter *adapter)
5637 {
5638 	struct ixgbe_hw *hw = &adapter->hw;
5639 	struct net_device *netdev = adapter->netdev;
5640 	int err;
5641 
5642 	if (ixgbe_removed(hw->hw_addr))
5643 		return;
5644 	/* lock SFP init bit to prevent race conditions with the watchdog */
5645 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5646 		usleep_range(1000, 2000);
5647 
5648 	/* clear all SFP and link config related flags while holding SFP_INIT */
5649 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
5650 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
5651 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5652 
5653 	err = hw->mac.ops.init_hw(hw);
5654 	switch (err) {
5655 	case 0:
5656 	case IXGBE_ERR_SFP_NOT_PRESENT:
5657 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
5658 		break;
5659 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
5660 		e_dev_err("master disable timed out\n");
5661 		break;
5662 	case IXGBE_ERR_EEPROM_VERSION:
5663 		/* We are running on a pre-production device, log a warning */
5664 		e_dev_warn("This device is a pre-production adapter/LOM. "
5665 			   "Please be aware there may be issues associated with "
5666 			   "your hardware.  If you are experiencing problems "
5667 			   "please contact your Intel or hardware "
5668 			   "representative who provided you with this "
5669 			   "hardware.\n");
5670 		break;
5671 	default:
5672 		e_dev_err("Hardware Error: %d\n", err);
5673 	}
5674 
5675 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5676 
5677 	/* flush entries out of MAC table */
5678 	ixgbe_flush_sw_mac_table(adapter);
5679 	__dev_uc_unsync(netdev, NULL);
5680 
5681 	/* do not flush user set addresses */
5682 	ixgbe_mac_set_default_filter(adapter);
5683 
5684 	/* update SAN MAC vmdq pool selection */
5685 	if (hw->mac.san_mac_rar_index)
5686 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
5687 
5688 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
5689 		ixgbe_ptp_reset(adapter);
5690 
5691 	if (hw->phy.ops.set_phy_power) {
5692 		if (!netif_running(adapter->netdev) && !adapter->wol)
5693 			hw->phy.ops.set_phy_power(hw, false);
5694 		else
5695 			hw->phy.ops.set_phy_power(hw, true);
5696 	}
5697 }
5698 
5699 /**
5700  * ixgbe_clean_tx_ring - Free Tx Buffers
5701  * @tx_ring: ring to be cleaned
5702  **/
5703 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
5704 {
5705 	u16 i = tx_ring->next_to_clean;
5706 	struct ixgbe_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
5707 
5708 	while (i != tx_ring->next_to_use) {
5709 		union ixgbe_adv_tx_desc *eop_desc, *tx_desc;
5710 
5711 		/* Free all the Tx ring sk_buffs */
5712 		if (ring_is_xdp(tx_ring))
5713 			xdp_return_frame(tx_buffer->xdpf);
5714 		else
5715 			dev_kfree_skb_any(tx_buffer->skb);
5716 
5717 		/* unmap skb header data */
5718 		dma_unmap_single(tx_ring->dev,
5719 				 dma_unmap_addr(tx_buffer, dma),
5720 				 dma_unmap_len(tx_buffer, len),
5721 				 DMA_TO_DEVICE);
5722 
5723 		/* check for eop_desc to determine the end of the packet */
5724 		eop_desc = tx_buffer->next_to_watch;
5725 		tx_desc = IXGBE_TX_DESC(tx_ring, i);
5726 
5727 		/* unmap remaining buffers */
5728 		while (tx_desc != eop_desc) {
5729 			tx_buffer++;
5730 			tx_desc++;
5731 			i++;
5732 			if (unlikely(i == tx_ring->count)) {
5733 				i = 0;
5734 				tx_buffer = tx_ring->tx_buffer_info;
5735 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
5736 			}
5737 
5738 			/* unmap any remaining paged data */
5739 			if (dma_unmap_len(tx_buffer, len))
5740 				dma_unmap_page(tx_ring->dev,
5741 					       dma_unmap_addr(tx_buffer, dma),
5742 					       dma_unmap_len(tx_buffer, len),
5743 					       DMA_TO_DEVICE);
5744 		}
5745 
5746 		/* move us one more past the eop_desc for start of next pkt */
5747 		tx_buffer++;
5748 		i++;
5749 		if (unlikely(i == tx_ring->count)) {
5750 			i = 0;
5751 			tx_buffer = tx_ring->tx_buffer_info;
5752 		}
5753 	}
5754 
5755 	/* reset BQL for queue */
5756 	if (!ring_is_xdp(tx_ring))
5757 		netdev_tx_reset_queue(txring_txq(tx_ring));
5758 
5759 	/* reset next_to_use and next_to_clean */
5760 	tx_ring->next_to_use = 0;
5761 	tx_ring->next_to_clean = 0;
5762 }
5763 
5764 /**
5765  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
5766  * @adapter: board private structure
5767  **/
5768 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
5769 {
5770 	int i;
5771 
5772 	for (i = 0; i < adapter->num_rx_queues; i++)
5773 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
5774 }
5775 
5776 /**
5777  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
5778  * @adapter: board private structure
5779  **/
5780 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
5781 {
5782 	int i;
5783 
5784 	for (i = 0; i < adapter->num_tx_queues; i++)
5785 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
5786 	for (i = 0; i < adapter->num_xdp_queues; i++)
5787 		ixgbe_clean_tx_ring(adapter->xdp_ring[i]);
5788 }
5789 
5790 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
5791 {
5792 	struct hlist_node *node2;
5793 	struct ixgbe_fdir_filter *filter;
5794 
5795 	spin_lock(&adapter->fdir_perfect_lock);
5796 
5797 	hlist_for_each_entry_safe(filter, node2,
5798 				  &adapter->fdir_filter_list, fdir_node) {
5799 		hlist_del(&filter->fdir_node);
5800 		kfree(filter);
5801 	}
5802 	adapter->fdir_filter_count = 0;
5803 
5804 	spin_unlock(&adapter->fdir_perfect_lock);
5805 }
5806 
5807 void ixgbe_down(struct ixgbe_adapter *adapter)
5808 {
5809 	struct net_device *netdev = adapter->netdev;
5810 	struct ixgbe_hw *hw = &adapter->hw;
5811 	int i;
5812 
5813 	/* signal that we are down to the interrupt handler */
5814 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
5815 		return; /* do nothing if already down */
5816 
5817 	/* disable receives */
5818 	hw->mac.ops.disable_rx(hw);
5819 
5820 	/* disable all enabled rx queues */
5821 	for (i = 0; i < adapter->num_rx_queues; i++)
5822 		/* this call also flushes the previous write */
5823 		ixgbe_disable_rx_queue(adapter, adapter->rx_ring[i]);
5824 
5825 	usleep_range(10000, 20000);
5826 
5827 	/* synchronize_sched() needed for pending XDP buffers to drain */
5828 	if (adapter->xdp_ring[0])
5829 		synchronize_sched();
5830 	netif_tx_stop_all_queues(netdev);
5831 
5832 	/* call carrier off first to avoid false dev_watchdog timeouts */
5833 	netif_carrier_off(netdev);
5834 	netif_tx_disable(netdev);
5835 
5836 	ixgbe_irq_disable(adapter);
5837 
5838 	ixgbe_napi_disable_all(adapter);
5839 
5840 	clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
5841 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
5842 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
5843 
5844 	del_timer_sync(&adapter->service_timer);
5845 
5846 	if (adapter->num_vfs) {
5847 		/* Clear EITR Select mapping */
5848 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
5849 
5850 		/* Mark all the VFs as inactive */
5851 		for (i = 0 ; i < adapter->num_vfs; i++)
5852 			adapter->vfinfo[i].clear_to_send = false;
5853 
5854 		/* ping all the active vfs to let them know we are going down */
5855 		ixgbe_ping_all_vfs(adapter);
5856 
5857 		/* Disable all VFTE/VFRE TX/RX */
5858 		ixgbe_disable_tx_rx(adapter);
5859 	}
5860 
5861 	/* disable transmits in the hardware now that interrupts are off */
5862 	for (i = 0; i < adapter->num_tx_queues; i++) {
5863 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
5864 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5865 	}
5866 	for (i = 0; i < adapter->num_xdp_queues; i++) {
5867 		u8 reg_idx = adapter->xdp_ring[i]->reg_idx;
5868 
5869 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5870 	}
5871 
5872 	/* Disable the Tx DMA engine on 82599 and later MAC */
5873 	switch (hw->mac.type) {
5874 	case ixgbe_mac_82599EB:
5875 	case ixgbe_mac_X540:
5876 	case ixgbe_mac_X550:
5877 	case ixgbe_mac_X550EM_x:
5878 	case ixgbe_mac_x550em_a:
5879 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5880 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5881 				 ~IXGBE_DMATXCTL_TE));
5882 		break;
5883 	default:
5884 		break;
5885 	}
5886 
5887 	if (!pci_channel_offline(adapter->pdev))
5888 		ixgbe_reset(adapter);
5889 
5890 	/* power down the optics for 82599 SFP+ fiber */
5891 	if (hw->mac.ops.disable_tx_laser)
5892 		hw->mac.ops.disable_tx_laser(hw);
5893 
5894 	ixgbe_clean_all_tx_rings(adapter);
5895 	ixgbe_clean_all_rx_rings(adapter);
5896 }
5897 
5898 /**
5899  * ixgbe_eee_capable - helper function to determine EEE support on X550
5900  * @adapter: board private structure
5901  */
5902 static void ixgbe_set_eee_capable(struct ixgbe_adapter *adapter)
5903 {
5904 	struct ixgbe_hw *hw = &adapter->hw;
5905 
5906 	switch (hw->device_id) {
5907 	case IXGBE_DEV_ID_X550EM_A_1G_T:
5908 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
5909 		if (!hw->phy.eee_speeds_supported)
5910 			break;
5911 		adapter->flags2 |= IXGBE_FLAG2_EEE_CAPABLE;
5912 		if (!hw->phy.eee_speeds_advertised)
5913 			break;
5914 		adapter->flags2 |= IXGBE_FLAG2_EEE_ENABLED;
5915 		break;
5916 	default:
5917 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_CAPABLE;
5918 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_ENABLED;
5919 		break;
5920 	}
5921 }
5922 
5923 /**
5924  * ixgbe_tx_timeout - Respond to a Tx Hang
5925  * @netdev: network interface device structure
5926  **/
5927 static void ixgbe_tx_timeout(struct net_device *netdev)
5928 {
5929 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
5930 
5931 	/* Do the reset outside of interrupt context */
5932 	ixgbe_tx_timeout_reset(adapter);
5933 }
5934 
5935 #ifdef CONFIG_IXGBE_DCB
5936 static void ixgbe_init_dcb(struct ixgbe_adapter *adapter)
5937 {
5938 	struct ixgbe_hw *hw = &adapter->hw;
5939 	struct tc_configuration *tc;
5940 	int j;
5941 
5942 	switch (hw->mac.type) {
5943 	case ixgbe_mac_82598EB:
5944 	case ixgbe_mac_82599EB:
5945 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
5946 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
5947 		break;
5948 	case ixgbe_mac_X540:
5949 	case ixgbe_mac_X550:
5950 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
5951 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
5952 		break;
5953 	case ixgbe_mac_X550EM_x:
5954 	case ixgbe_mac_x550em_a:
5955 	default:
5956 		adapter->dcb_cfg.num_tcs.pg_tcs = DEF_TRAFFIC_CLASS;
5957 		adapter->dcb_cfg.num_tcs.pfc_tcs = DEF_TRAFFIC_CLASS;
5958 		break;
5959 	}
5960 
5961 	/* Configure DCB traffic classes */
5962 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
5963 		tc = &adapter->dcb_cfg.tc_config[j];
5964 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
5965 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
5966 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
5967 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
5968 		tc->dcb_pfc = pfc_disabled;
5969 	}
5970 
5971 	/* Initialize default user to priority mapping, UPx->TC0 */
5972 	tc = &adapter->dcb_cfg.tc_config[0];
5973 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
5974 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
5975 
5976 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
5977 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
5978 	adapter->dcb_cfg.pfc_mode_enable = false;
5979 	adapter->dcb_set_bitmap = 0x00;
5980 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
5981 		adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
5982 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
5983 	       sizeof(adapter->temp_dcb_cfg));
5984 }
5985 #endif
5986 
5987 /**
5988  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
5989  * @adapter: board private structure to initialize
5990  * @ii: pointer to ixgbe_info for device
5991  *
5992  * ixgbe_sw_init initializes the Adapter private data structure.
5993  * Fields are initialized based on PCI device information and
5994  * OS network device settings (MTU size).
5995  **/
5996 static int ixgbe_sw_init(struct ixgbe_adapter *adapter,
5997 			 const struct ixgbe_info *ii)
5998 {
5999 	struct ixgbe_hw *hw = &adapter->hw;
6000 	struct pci_dev *pdev = adapter->pdev;
6001 	unsigned int rss, fdir;
6002 	u32 fwsm;
6003 	int i;
6004 
6005 	/* PCI config space info */
6006 
6007 	hw->vendor_id = pdev->vendor;
6008 	hw->device_id = pdev->device;
6009 	hw->revision_id = pdev->revision;
6010 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
6011 	hw->subsystem_device_id = pdev->subsystem_device;
6012 
6013 	/* get_invariants needs the device IDs */
6014 	ii->get_invariants(hw);
6015 
6016 	/* Set common capability flags and settings */
6017 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
6018 	adapter->ring_feature[RING_F_RSS].limit = rss;
6019 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
6020 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
6021 	adapter->atr_sample_rate = 20;
6022 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
6023 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
6024 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
6025 	adapter->ring_feature[RING_F_VMDQ].limit = 1;
6026 #ifdef CONFIG_IXGBE_DCA
6027 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
6028 #endif
6029 #ifdef CONFIG_IXGBE_DCB
6030 	adapter->flags |= IXGBE_FLAG_DCB_CAPABLE;
6031 	adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
6032 #endif
6033 #ifdef IXGBE_FCOE
6034 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
6035 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6036 #ifdef CONFIG_IXGBE_DCB
6037 	/* Default traffic class to use for FCoE */
6038 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
6039 #endif /* CONFIG_IXGBE_DCB */
6040 #endif /* IXGBE_FCOE */
6041 
6042 	/* initialize static ixgbe jump table entries */
6043 	adapter->jump_tables[0] = kzalloc(sizeof(*adapter->jump_tables[0]),
6044 					  GFP_KERNEL);
6045 	if (!adapter->jump_tables[0])
6046 		return -ENOMEM;
6047 	adapter->jump_tables[0]->mat = ixgbe_ipv4_fields;
6048 
6049 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++)
6050 		adapter->jump_tables[i] = NULL;
6051 
6052 	adapter->mac_table = kcalloc(hw->mac.num_rar_entries,
6053 				     sizeof(struct ixgbe_mac_addr),
6054 				     GFP_ATOMIC);
6055 	if (!adapter->mac_table)
6056 		return -ENOMEM;
6057 
6058 	if (ixgbe_init_rss_key(adapter))
6059 		return -ENOMEM;
6060 
6061 	/* Set MAC specific capability flags and exceptions */
6062 	switch (hw->mac.type) {
6063 	case ixgbe_mac_82598EB:
6064 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
6065 
6066 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
6067 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
6068 
6069 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
6070 		adapter->ring_feature[RING_F_FDIR].limit = 0;
6071 		adapter->atr_sample_rate = 0;
6072 		adapter->fdir_pballoc = 0;
6073 #ifdef IXGBE_FCOE
6074 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6075 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6076 #ifdef CONFIG_IXGBE_DCB
6077 		adapter->fcoe.up = 0;
6078 #endif /* IXGBE_DCB */
6079 #endif /* IXGBE_FCOE */
6080 		break;
6081 	case ixgbe_mac_82599EB:
6082 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
6083 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6084 		break;
6085 	case ixgbe_mac_X540:
6086 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
6087 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
6088 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6089 		break;
6090 	case ixgbe_mac_x550em_a:
6091 		adapter->flags |= IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE;
6092 		switch (hw->device_id) {
6093 		case IXGBE_DEV_ID_X550EM_A_1G_T:
6094 		case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6095 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6096 			break;
6097 		default:
6098 			break;
6099 		}
6100 	/* fall through */
6101 	case ixgbe_mac_X550EM_x:
6102 #ifdef CONFIG_IXGBE_DCB
6103 		adapter->flags &= ~IXGBE_FLAG_DCB_CAPABLE;
6104 #endif
6105 #ifdef IXGBE_FCOE
6106 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6107 #ifdef CONFIG_IXGBE_DCB
6108 		adapter->fcoe.up = 0;
6109 #endif /* IXGBE_DCB */
6110 #endif /* IXGBE_FCOE */
6111 	/* Fall Through */
6112 	case ixgbe_mac_X550:
6113 		if (hw->mac.type == ixgbe_mac_X550)
6114 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6115 #ifdef CONFIG_IXGBE_DCA
6116 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
6117 #endif
6118 		adapter->flags |= IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE;
6119 		break;
6120 	default:
6121 		break;
6122 	}
6123 
6124 #ifdef IXGBE_FCOE
6125 	/* FCoE support exists, always init the FCoE lock */
6126 	spin_lock_init(&adapter->fcoe.lock);
6127 
6128 #endif
6129 	/* n-tuple support exists, always init our spinlock */
6130 	spin_lock_init(&adapter->fdir_perfect_lock);
6131 
6132 #ifdef CONFIG_IXGBE_DCB
6133 	ixgbe_init_dcb(adapter);
6134 #endif
6135 	ixgbe_init_ipsec_offload(adapter);
6136 
6137 	/* default flow control settings */
6138 	hw->fc.requested_mode = ixgbe_fc_full;
6139 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
6140 	ixgbe_pbthresh_setup(adapter);
6141 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
6142 	hw->fc.send_xon = true;
6143 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
6144 
6145 #ifdef CONFIG_PCI_IOV
6146 	if (max_vfs > 0)
6147 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
6148 
6149 	/* assign number of SR-IOV VFs */
6150 	if (hw->mac.type != ixgbe_mac_82598EB) {
6151 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
6152 			max_vfs = 0;
6153 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
6154 		}
6155 	}
6156 #endif /* CONFIG_PCI_IOV */
6157 
6158 	/* enable itr by default in dynamic mode */
6159 	adapter->rx_itr_setting = 1;
6160 	adapter->tx_itr_setting = 1;
6161 
6162 	/* set default ring sizes */
6163 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
6164 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
6165 
6166 	/* set default work limits */
6167 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
6168 
6169 	/* initialize eeprom parameters */
6170 	if (ixgbe_init_eeprom_params_generic(hw)) {
6171 		e_dev_err("EEPROM initialization failed\n");
6172 		return -EIO;
6173 	}
6174 
6175 	/* PF holds first pool slot */
6176 	set_bit(0, adapter->fwd_bitmask);
6177 	set_bit(__IXGBE_DOWN, &adapter->state);
6178 
6179 	return 0;
6180 }
6181 
6182 /**
6183  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
6184  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
6185  *
6186  * Return 0 on success, negative on failure
6187  **/
6188 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
6189 {
6190 	struct device *dev = tx_ring->dev;
6191 	int orig_node = dev_to_node(dev);
6192 	int ring_node = -1;
6193 	int size;
6194 
6195 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
6196 
6197 	if (tx_ring->q_vector)
6198 		ring_node = tx_ring->q_vector->numa_node;
6199 
6200 	tx_ring->tx_buffer_info = vmalloc_node(size, ring_node);
6201 	if (!tx_ring->tx_buffer_info)
6202 		tx_ring->tx_buffer_info = vmalloc(size);
6203 	if (!tx_ring->tx_buffer_info)
6204 		goto err;
6205 
6206 	/* round up to nearest 4K */
6207 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
6208 	tx_ring->size = ALIGN(tx_ring->size, 4096);
6209 
6210 	set_dev_node(dev, ring_node);
6211 	tx_ring->desc = dma_alloc_coherent(dev,
6212 					   tx_ring->size,
6213 					   &tx_ring->dma,
6214 					   GFP_KERNEL);
6215 	set_dev_node(dev, orig_node);
6216 	if (!tx_ring->desc)
6217 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
6218 						   &tx_ring->dma, GFP_KERNEL);
6219 	if (!tx_ring->desc)
6220 		goto err;
6221 
6222 	tx_ring->next_to_use = 0;
6223 	tx_ring->next_to_clean = 0;
6224 	return 0;
6225 
6226 err:
6227 	vfree(tx_ring->tx_buffer_info);
6228 	tx_ring->tx_buffer_info = NULL;
6229 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
6230 	return -ENOMEM;
6231 }
6232 
6233 /**
6234  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
6235  * @adapter: board private structure
6236  *
6237  * If this function returns with an error, then it's possible one or
6238  * more of the rings is populated (while the rest are not).  It is the
6239  * callers duty to clean those orphaned rings.
6240  *
6241  * Return 0 on success, negative on failure
6242  **/
6243 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
6244 {
6245 	int i, j = 0, err = 0;
6246 
6247 	for (i = 0; i < adapter->num_tx_queues; i++) {
6248 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
6249 		if (!err)
6250 			continue;
6251 
6252 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
6253 		goto err_setup_tx;
6254 	}
6255 	for (j = 0; j < adapter->num_xdp_queues; j++) {
6256 		err = ixgbe_setup_tx_resources(adapter->xdp_ring[j]);
6257 		if (!err)
6258 			continue;
6259 
6260 		e_err(probe, "Allocation for Tx Queue %u failed\n", j);
6261 		goto err_setup_tx;
6262 	}
6263 
6264 	return 0;
6265 err_setup_tx:
6266 	/* rewind the index freeing the rings as we go */
6267 	while (j--)
6268 		ixgbe_free_tx_resources(adapter->xdp_ring[j]);
6269 	while (i--)
6270 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
6271 	return err;
6272 }
6273 
6274 /**
6275  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
6276  * @adapter: pointer to ixgbe_adapter
6277  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
6278  *
6279  * Returns 0 on success, negative on failure
6280  **/
6281 int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
6282 			     struct ixgbe_ring *rx_ring)
6283 {
6284 	struct device *dev = rx_ring->dev;
6285 	int orig_node = dev_to_node(dev);
6286 	int ring_node = -1;
6287 	int size, err;
6288 
6289 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
6290 
6291 	if (rx_ring->q_vector)
6292 		ring_node = rx_ring->q_vector->numa_node;
6293 
6294 	rx_ring->rx_buffer_info = vmalloc_node(size, ring_node);
6295 	if (!rx_ring->rx_buffer_info)
6296 		rx_ring->rx_buffer_info = vmalloc(size);
6297 	if (!rx_ring->rx_buffer_info)
6298 		goto err;
6299 
6300 	/* Round up to nearest 4K */
6301 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
6302 	rx_ring->size = ALIGN(rx_ring->size, 4096);
6303 
6304 	set_dev_node(dev, ring_node);
6305 	rx_ring->desc = dma_alloc_coherent(dev,
6306 					   rx_ring->size,
6307 					   &rx_ring->dma,
6308 					   GFP_KERNEL);
6309 	set_dev_node(dev, orig_node);
6310 	if (!rx_ring->desc)
6311 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
6312 						   &rx_ring->dma, GFP_KERNEL);
6313 	if (!rx_ring->desc)
6314 		goto err;
6315 
6316 	rx_ring->next_to_clean = 0;
6317 	rx_ring->next_to_use = 0;
6318 
6319 	/* XDP RX-queue info */
6320 	if (xdp_rxq_info_reg(&rx_ring->xdp_rxq, adapter->netdev,
6321 			     rx_ring->queue_index) < 0)
6322 		goto err;
6323 
6324 	err = xdp_rxq_info_reg_mem_model(&rx_ring->xdp_rxq,
6325 					 MEM_TYPE_PAGE_SHARED, NULL);
6326 	if (err) {
6327 		xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
6328 		goto err;
6329 	}
6330 
6331 	rx_ring->xdp_prog = adapter->xdp_prog;
6332 
6333 	return 0;
6334 err:
6335 	vfree(rx_ring->rx_buffer_info);
6336 	rx_ring->rx_buffer_info = NULL;
6337 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
6338 	return -ENOMEM;
6339 }
6340 
6341 /**
6342  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
6343  * @adapter: board private structure
6344  *
6345  * If this function returns with an error, then it's possible one or
6346  * more of the rings is populated (while the rest are not).  It is the
6347  * callers duty to clean those orphaned rings.
6348  *
6349  * Return 0 on success, negative on failure
6350  **/
6351 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
6352 {
6353 	int i, err = 0;
6354 
6355 	for (i = 0; i < adapter->num_rx_queues; i++) {
6356 		err = ixgbe_setup_rx_resources(adapter, adapter->rx_ring[i]);
6357 		if (!err)
6358 			continue;
6359 
6360 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
6361 		goto err_setup_rx;
6362 	}
6363 
6364 #ifdef IXGBE_FCOE
6365 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
6366 	if (!err)
6367 #endif
6368 		return 0;
6369 err_setup_rx:
6370 	/* rewind the index freeing the rings as we go */
6371 	while (i--)
6372 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
6373 	return err;
6374 }
6375 
6376 /**
6377  * ixgbe_free_tx_resources - Free Tx Resources per Queue
6378  * @tx_ring: Tx descriptor ring for a specific queue
6379  *
6380  * Free all transmit software resources
6381  **/
6382 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
6383 {
6384 	ixgbe_clean_tx_ring(tx_ring);
6385 
6386 	vfree(tx_ring->tx_buffer_info);
6387 	tx_ring->tx_buffer_info = NULL;
6388 
6389 	/* if not set, then don't free */
6390 	if (!tx_ring->desc)
6391 		return;
6392 
6393 	dma_free_coherent(tx_ring->dev, tx_ring->size,
6394 			  tx_ring->desc, tx_ring->dma);
6395 
6396 	tx_ring->desc = NULL;
6397 }
6398 
6399 /**
6400  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
6401  * @adapter: board private structure
6402  *
6403  * Free all transmit software resources
6404  **/
6405 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
6406 {
6407 	int i;
6408 
6409 	for (i = 0; i < adapter->num_tx_queues; i++)
6410 		if (adapter->tx_ring[i]->desc)
6411 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
6412 	for (i = 0; i < adapter->num_xdp_queues; i++)
6413 		if (adapter->xdp_ring[i]->desc)
6414 			ixgbe_free_tx_resources(adapter->xdp_ring[i]);
6415 }
6416 
6417 /**
6418  * ixgbe_free_rx_resources - Free Rx Resources
6419  * @rx_ring: ring to clean the resources from
6420  *
6421  * Free all receive software resources
6422  **/
6423 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
6424 {
6425 	ixgbe_clean_rx_ring(rx_ring);
6426 
6427 	rx_ring->xdp_prog = NULL;
6428 	xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
6429 	vfree(rx_ring->rx_buffer_info);
6430 	rx_ring->rx_buffer_info = NULL;
6431 
6432 	/* if not set, then don't free */
6433 	if (!rx_ring->desc)
6434 		return;
6435 
6436 	dma_free_coherent(rx_ring->dev, rx_ring->size,
6437 			  rx_ring->desc, rx_ring->dma);
6438 
6439 	rx_ring->desc = NULL;
6440 }
6441 
6442 /**
6443  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
6444  * @adapter: board private structure
6445  *
6446  * Free all receive software resources
6447  **/
6448 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
6449 {
6450 	int i;
6451 
6452 #ifdef IXGBE_FCOE
6453 	ixgbe_free_fcoe_ddp_resources(adapter);
6454 
6455 #endif
6456 	for (i = 0; i < adapter->num_rx_queues; i++)
6457 		if (adapter->rx_ring[i]->desc)
6458 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
6459 }
6460 
6461 /**
6462  * ixgbe_change_mtu - Change the Maximum Transfer Unit
6463  * @netdev: network interface device structure
6464  * @new_mtu: new value for maximum frame size
6465  *
6466  * Returns 0 on success, negative on failure
6467  **/
6468 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
6469 {
6470 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6471 
6472 	/*
6473 	 * For 82599EB we cannot allow legacy VFs to enable their receive
6474 	 * paths when MTU greater than 1500 is configured.  So display a
6475 	 * warning that legacy VFs will be disabled.
6476 	 */
6477 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
6478 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
6479 	    (new_mtu > ETH_DATA_LEN))
6480 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
6481 
6482 	e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
6483 
6484 	/* must set new MTU before calling down or up */
6485 	netdev->mtu = new_mtu;
6486 
6487 	if (netif_running(netdev))
6488 		ixgbe_reinit_locked(adapter);
6489 
6490 	return 0;
6491 }
6492 
6493 /**
6494  * ixgbe_open - Called when a network interface is made active
6495  * @netdev: network interface device structure
6496  *
6497  * Returns 0 on success, negative value on failure
6498  *
6499  * The open entry point is called when a network interface is made
6500  * active by the system (IFF_UP).  At this point all resources needed
6501  * for transmit and receive operations are allocated, the interrupt
6502  * handler is registered with the OS, the watchdog timer is started,
6503  * and the stack is notified that the interface is ready.
6504  **/
6505 int ixgbe_open(struct net_device *netdev)
6506 {
6507 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6508 	struct ixgbe_hw *hw = &adapter->hw;
6509 	int err, queues;
6510 
6511 	/* disallow open during test */
6512 	if (test_bit(__IXGBE_TESTING, &adapter->state))
6513 		return -EBUSY;
6514 
6515 	netif_carrier_off(netdev);
6516 
6517 	/* allocate transmit descriptors */
6518 	err = ixgbe_setup_all_tx_resources(adapter);
6519 	if (err)
6520 		goto err_setup_tx;
6521 
6522 	/* allocate receive descriptors */
6523 	err = ixgbe_setup_all_rx_resources(adapter);
6524 	if (err)
6525 		goto err_setup_rx;
6526 
6527 	ixgbe_configure(adapter);
6528 
6529 	err = ixgbe_request_irq(adapter);
6530 	if (err)
6531 		goto err_req_irq;
6532 
6533 	/* Notify the stack of the actual queue counts. */
6534 	queues = adapter->num_tx_queues;
6535 	err = netif_set_real_num_tx_queues(netdev, queues);
6536 	if (err)
6537 		goto err_set_queues;
6538 
6539 	queues = adapter->num_rx_queues;
6540 	err = netif_set_real_num_rx_queues(netdev, queues);
6541 	if (err)
6542 		goto err_set_queues;
6543 
6544 	ixgbe_ptp_init(adapter);
6545 
6546 	ixgbe_up_complete(adapter);
6547 
6548 	ixgbe_clear_udp_tunnel_port(adapter, IXGBE_VXLANCTRL_ALL_UDPPORT_MASK);
6549 	udp_tunnel_get_rx_info(netdev);
6550 
6551 	return 0;
6552 
6553 err_set_queues:
6554 	ixgbe_free_irq(adapter);
6555 err_req_irq:
6556 	ixgbe_free_all_rx_resources(adapter);
6557 	if (hw->phy.ops.set_phy_power && !adapter->wol)
6558 		hw->phy.ops.set_phy_power(&adapter->hw, false);
6559 err_setup_rx:
6560 	ixgbe_free_all_tx_resources(adapter);
6561 err_setup_tx:
6562 	ixgbe_reset(adapter);
6563 
6564 	return err;
6565 }
6566 
6567 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
6568 {
6569 	ixgbe_ptp_suspend(adapter);
6570 
6571 	if (adapter->hw.phy.ops.enter_lplu) {
6572 		adapter->hw.phy.reset_disable = true;
6573 		ixgbe_down(adapter);
6574 		adapter->hw.phy.ops.enter_lplu(&adapter->hw);
6575 		adapter->hw.phy.reset_disable = false;
6576 	} else {
6577 		ixgbe_down(adapter);
6578 	}
6579 
6580 	ixgbe_free_irq(adapter);
6581 
6582 	ixgbe_free_all_tx_resources(adapter);
6583 	ixgbe_free_all_rx_resources(adapter);
6584 }
6585 
6586 /**
6587  * ixgbe_close - Disables a network interface
6588  * @netdev: network interface device structure
6589  *
6590  * Returns 0, this is not allowed to fail
6591  *
6592  * The close entry point is called when an interface is de-activated
6593  * by the OS.  The hardware is still under the drivers control, but
6594  * needs to be disabled.  A global MAC reset is issued to stop the
6595  * hardware, and all transmit and receive resources are freed.
6596  **/
6597 int ixgbe_close(struct net_device *netdev)
6598 {
6599 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6600 
6601 	ixgbe_ptp_stop(adapter);
6602 
6603 	if (netif_device_present(netdev))
6604 		ixgbe_close_suspend(adapter);
6605 
6606 	ixgbe_fdir_filter_exit(adapter);
6607 
6608 	ixgbe_release_hw_control(adapter);
6609 
6610 	return 0;
6611 }
6612 
6613 #ifdef CONFIG_PM
6614 static int ixgbe_resume(struct pci_dev *pdev)
6615 {
6616 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6617 	struct net_device *netdev = adapter->netdev;
6618 	u32 err;
6619 
6620 	adapter->hw.hw_addr = adapter->io_addr;
6621 	pci_set_power_state(pdev, PCI_D0);
6622 	pci_restore_state(pdev);
6623 	/*
6624 	 * pci_restore_state clears dev->state_saved so call
6625 	 * pci_save_state to restore it.
6626 	 */
6627 	pci_save_state(pdev);
6628 
6629 	err = pci_enable_device_mem(pdev);
6630 	if (err) {
6631 		e_dev_err("Cannot enable PCI device from suspend\n");
6632 		return err;
6633 	}
6634 	smp_mb__before_atomic();
6635 	clear_bit(__IXGBE_DISABLED, &adapter->state);
6636 	pci_set_master(pdev);
6637 
6638 	pci_wake_from_d3(pdev, false);
6639 
6640 	ixgbe_reset(adapter);
6641 
6642 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
6643 
6644 	rtnl_lock();
6645 	err = ixgbe_init_interrupt_scheme(adapter);
6646 	if (!err && netif_running(netdev))
6647 		err = ixgbe_open(netdev);
6648 
6649 
6650 	if (!err)
6651 		netif_device_attach(netdev);
6652 	rtnl_unlock();
6653 
6654 	return err;
6655 }
6656 #endif /* CONFIG_PM */
6657 
6658 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
6659 {
6660 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6661 	struct net_device *netdev = adapter->netdev;
6662 	struct ixgbe_hw *hw = &adapter->hw;
6663 	u32 ctrl;
6664 	u32 wufc = adapter->wol;
6665 #ifdef CONFIG_PM
6666 	int retval = 0;
6667 #endif
6668 
6669 	rtnl_lock();
6670 	netif_device_detach(netdev);
6671 
6672 	if (netif_running(netdev))
6673 		ixgbe_close_suspend(adapter);
6674 
6675 	ixgbe_clear_interrupt_scheme(adapter);
6676 	rtnl_unlock();
6677 
6678 #ifdef CONFIG_PM
6679 	retval = pci_save_state(pdev);
6680 	if (retval)
6681 		return retval;
6682 
6683 #endif
6684 	if (hw->mac.ops.stop_link_on_d3)
6685 		hw->mac.ops.stop_link_on_d3(hw);
6686 
6687 	if (wufc) {
6688 		u32 fctrl;
6689 
6690 		ixgbe_set_rx_mode(netdev);
6691 
6692 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
6693 		if (hw->mac.ops.enable_tx_laser)
6694 			hw->mac.ops.enable_tx_laser(hw);
6695 
6696 		/* enable the reception of multicast packets */
6697 		fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6698 		fctrl |= IXGBE_FCTRL_MPE;
6699 		IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
6700 
6701 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
6702 		ctrl |= IXGBE_CTRL_GIO_DIS;
6703 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
6704 
6705 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
6706 	} else {
6707 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
6708 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
6709 	}
6710 
6711 	switch (hw->mac.type) {
6712 	case ixgbe_mac_82598EB:
6713 		pci_wake_from_d3(pdev, false);
6714 		break;
6715 	case ixgbe_mac_82599EB:
6716 	case ixgbe_mac_X540:
6717 	case ixgbe_mac_X550:
6718 	case ixgbe_mac_X550EM_x:
6719 	case ixgbe_mac_x550em_a:
6720 		pci_wake_from_d3(pdev, !!wufc);
6721 		break;
6722 	default:
6723 		break;
6724 	}
6725 
6726 	*enable_wake = !!wufc;
6727 	if (hw->phy.ops.set_phy_power && !*enable_wake)
6728 		hw->phy.ops.set_phy_power(hw, false);
6729 
6730 	ixgbe_release_hw_control(adapter);
6731 
6732 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
6733 		pci_disable_device(pdev);
6734 
6735 	return 0;
6736 }
6737 
6738 #ifdef CONFIG_PM
6739 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
6740 {
6741 	int retval;
6742 	bool wake;
6743 
6744 	retval = __ixgbe_shutdown(pdev, &wake);
6745 	if (retval)
6746 		return retval;
6747 
6748 	if (wake) {
6749 		pci_prepare_to_sleep(pdev);
6750 	} else {
6751 		pci_wake_from_d3(pdev, false);
6752 		pci_set_power_state(pdev, PCI_D3hot);
6753 	}
6754 
6755 	return 0;
6756 }
6757 #endif /* CONFIG_PM */
6758 
6759 static void ixgbe_shutdown(struct pci_dev *pdev)
6760 {
6761 	bool wake;
6762 
6763 	__ixgbe_shutdown(pdev, &wake);
6764 
6765 	if (system_state == SYSTEM_POWER_OFF) {
6766 		pci_wake_from_d3(pdev, wake);
6767 		pci_set_power_state(pdev, PCI_D3hot);
6768 	}
6769 }
6770 
6771 /**
6772  * ixgbe_update_stats - Update the board statistics counters.
6773  * @adapter: board private structure
6774  **/
6775 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
6776 {
6777 	struct net_device *netdev = adapter->netdev;
6778 	struct ixgbe_hw *hw = &adapter->hw;
6779 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
6780 	u64 total_mpc = 0;
6781 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
6782 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
6783 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
6784 	u64 alloc_rx_page = 0;
6785 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
6786 
6787 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
6788 	    test_bit(__IXGBE_RESETTING, &adapter->state))
6789 		return;
6790 
6791 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
6792 		u64 rsc_count = 0;
6793 		u64 rsc_flush = 0;
6794 		for (i = 0; i < adapter->num_rx_queues; i++) {
6795 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
6796 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
6797 		}
6798 		adapter->rsc_total_count = rsc_count;
6799 		adapter->rsc_total_flush = rsc_flush;
6800 	}
6801 
6802 	for (i = 0; i < adapter->num_rx_queues; i++) {
6803 		struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
6804 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
6805 		alloc_rx_page += rx_ring->rx_stats.alloc_rx_page;
6806 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
6807 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
6808 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
6809 		bytes += rx_ring->stats.bytes;
6810 		packets += rx_ring->stats.packets;
6811 	}
6812 	adapter->non_eop_descs = non_eop_descs;
6813 	adapter->alloc_rx_page = alloc_rx_page;
6814 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
6815 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
6816 	adapter->hw_csum_rx_error = hw_csum_rx_error;
6817 	netdev->stats.rx_bytes = bytes;
6818 	netdev->stats.rx_packets = packets;
6819 
6820 	bytes = 0;
6821 	packets = 0;
6822 	/* gather some stats to the adapter struct that are per queue */
6823 	for (i = 0; i < adapter->num_tx_queues; i++) {
6824 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
6825 		restart_queue += tx_ring->tx_stats.restart_queue;
6826 		tx_busy += tx_ring->tx_stats.tx_busy;
6827 		bytes += tx_ring->stats.bytes;
6828 		packets += tx_ring->stats.packets;
6829 	}
6830 	for (i = 0; i < adapter->num_xdp_queues; i++) {
6831 		struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
6832 
6833 		restart_queue += xdp_ring->tx_stats.restart_queue;
6834 		tx_busy += xdp_ring->tx_stats.tx_busy;
6835 		bytes += xdp_ring->stats.bytes;
6836 		packets += xdp_ring->stats.packets;
6837 	}
6838 	adapter->restart_queue = restart_queue;
6839 	adapter->tx_busy = tx_busy;
6840 	netdev->stats.tx_bytes = bytes;
6841 	netdev->stats.tx_packets = packets;
6842 
6843 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
6844 
6845 	/* 8 register reads */
6846 	for (i = 0; i < 8; i++) {
6847 		/* for packet buffers not used, the register should read 0 */
6848 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
6849 		missed_rx += mpc;
6850 		hwstats->mpc[i] += mpc;
6851 		total_mpc += hwstats->mpc[i];
6852 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
6853 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
6854 		switch (hw->mac.type) {
6855 		case ixgbe_mac_82598EB:
6856 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
6857 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
6858 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
6859 			hwstats->pxonrxc[i] +=
6860 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
6861 			break;
6862 		case ixgbe_mac_82599EB:
6863 		case ixgbe_mac_X540:
6864 		case ixgbe_mac_X550:
6865 		case ixgbe_mac_X550EM_x:
6866 		case ixgbe_mac_x550em_a:
6867 			hwstats->pxonrxc[i] +=
6868 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
6869 			break;
6870 		default:
6871 			break;
6872 		}
6873 	}
6874 
6875 	/*16 register reads */
6876 	for (i = 0; i < 16; i++) {
6877 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
6878 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
6879 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
6880 		    (hw->mac.type == ixgbe_mac_X540) ||
6881 		    (hw->mac.type == ixgbe_mac_X550) ||
6882 		    (hw->mac.type == ixgbe_mac_X550EM_x) ||
6883 		    (hw->mac.type == ixgbe_mac_x550em_a)) {
6884 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
6885 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
6886 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
6887 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
6888 		}
6889 	}
6890 
6891 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
6892 	/* work around hardware counting issue */
6893 	hwstats->gprc -= missed_rx;
6894 
6895 	ixgbe_update_xoff_received(adapter);
6896 
6897 	/* 82598 hardware only has a 32 bit counter in the high register */
6898 	switch (hw->mac.type) {
6899 	case ixgbe_mac_82598EB:
6900 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
6901 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
6902 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
6903 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
6904 		break;
6905 	case ixgbe_mac_X540:
6906 	case ixgbe_mac_X550:
6907 	case ixgbe_mac_X550EM_x:
6908 	case ixgbe_mac_x550em_a:
6909 		/* OS2BMC stats are X540 and later */
6910 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
6911 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
6912 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
6913 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
6914 		/* fall through */
6915 	case ixgbe_mac_82599EB:
6916 		for (i = 0; i < 16; i++)
6917 			adapter->hw_rx_no_dma_resources +=
6918 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
6919 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
6920 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
6921 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
6922 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
6923 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
6924 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
6925 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
6926 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
6927 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
6928 #ifdef IXGBE_FCOE
6929 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
6930 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
6931 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
6932 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
6933 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
6934 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
6935 		/* Add up per cpu counters for total ddp aloc fail */
6936 		if (adapter->fcoe.ddp_pool) {
6937 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
6938 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
6939 			unsigned int cpu;
6940 			u64 noddp = 0, noddp_ext_buff = 0;
6941 			for_each_possible_cpu(cpu) {
6942 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
6943 				noddp += ddp_pool->noddp;
6944 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
6945 			}
6946 			hwstats->fcoe_noddp = noddp;
6947 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
6948 		}
6949 #endif /* IXGBE_FCOE */
6950 		break;
6951 	default:
6952 		break;
6953 	}
6954 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
6955 	hwstats->bprc += bprc;
6956 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
6957 	if (hw->mac.type == ixgbe_mac_82598EB)
6958 		hwstats->mprc -= bprc;
6959 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
6960 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
6961 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
6962 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
6963 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
6964 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
6965 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
6966 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
6967 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
6968 	hwstats->lxontxc += lxon;
6969 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
6970 	hwstats->lxofftxc += lxoff;
6971 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
6972 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
6973 	/*
6974 	 * 82598 errata - tx of flow control packets is included in tx counters
6975 	 */
6976 	xon_off_tot = lxon + lxoff;
6977 	hwstats->gptc -= xon_off_tot;
6978 	hwstats->mptc -= xon_off_tot;
6979 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
6980 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
6981 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
6982 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
6983 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
6984 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
6985 	hwstats->ptc64 -= xon_off_tot;
6986 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
6987 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
6988 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
6989 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
6990 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
6991 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
6992 
6993 	/* Fill out the OS statistics structure */
6994 	netdev->stats.multicast = hwstats->mprc;
6995 
6996 	/* Rx Errors */
6997 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
6998 	netdev->stats.rx_dropped = 0;
6999 	netdev->stats.rx_length_errors = hwstats->rlec;
7000 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
7001 	netdev->stats.rx_missed_errors = total_mpc;
7002 }
7003 
7004 /**
7005  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
7006  * @adapter: pointer to the device adapter structure
7007  **/
7008 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
7009 {
7010 	struct ixgbe_hw *hw = &adapter->hw;
7011 	int i;
7012 
7013 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
7014 		return;
7015 
7016 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
7017 
7018 	/* if interface is down do nothing */
7019 	if (test_bit(__IXGBE_DOWN, &adapter->state))
7020 		return;
7021 
7022 	/* do nothing if we are not using signature filters */
7023 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
7024 		return;
7025 
7026 	adapter->fdir_overflow++;
7027 
7028 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
7029 		for (i = 0; i < adapter->num_tx_queues; i++)
7030 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7031 				&(adapter->tx_ring[i]->state));
7032 		for (i = 0; i < adapter->num_xdp_queues; i++)
7033 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7034 				&adapter->xdp_ring[i]->state);
7035 		/* re-enable flow director interrupts */
7036 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
7037 	} else {
7038 		e_err(probe, "failed to finish FDIR re-initialization, "
7039 		      "ignored adding FDIR ATR filters\n");
7040 	}
7041 }
7042 
7043 /**
7044  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
7045  * @adapter: pointer to the device adapter structure
7046  *
7047  * This function serves two purposes.  First it strobes the interrupt lines
7048  * in order to make certain interrupts are occurring.  Secondly it sets the
7049  * bits needed to check for TX hangs.  As a result we should immediately
7050  * determine if a hang has occurred.
7051  */
7052 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
7053 {
7054 	struct ixgbe_hw *hw = &adapter->hw;
7055 	u64 eics = 0;
7056 	int i;
7057 
7058 	/* If we're down, removing or resetting, just bail */
7059 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7060 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7061 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7062 		return;
7063 
7064 	/* Force detection of hung controller */
7065 	if (netif_carrier_ok(adapter->netdev)) {
7066 		for (i = 0; i < adapter->num_tx_queues; i++)
7067 			set_check_for_tx_hang(adapter->tx_ring[i]);
7068 		for (i = 0; i < adapter->num_xdp_queues; i++)
7069 			set_check_for_tx_hang(adapter->xdp_ring[i]);
7070 	}
7071 
7072 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
7073 		/*
7074 		 * for legacy and MSI interrupts don't set any bits
7075 		 * that are enabled for EIAM, because this operation
7076 		 * would set *both* EIMS and EICS for any bit in EIAM
7077 		 */
7078 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
7079 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
7080 	} else {
7081 		/* get one bit for every active tx/rx interrupt vector */
7082 		for (i = 0; i < adapter->num_q_vectors; i++) {
7083 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
7084 			if (qv->rx.ring || qv->tx.ring)
7085 				eics |= BIT_ULL(i);
7086 		}
7087 	}
7088 
7089 	/* Cause software interrupt to ensure rings are cleaned */
7090 	ixgbe_irq_rearm_queues(adapter, eics);
7091 }
7092 
7093 /**
7094  * ixgbe_watchdog_update_link - update the link status
7095  * @adapter: pointer to the device adapter structure
7096  **/
7097 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
7098 {
7099 	struct ixgbe_hw *hw = &adapter->hw;
7100 	u32 link_speed = adapter->link_speed;
7101 	bool link_up = adapter->link_up;
7102 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
7103 
7104 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
7105 		return;
7106 
7107 	if (hw->mac.ops.check_link) {
7108 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
7109 	} else {
7110 		/* always assume link is up, if no check link function */
7111 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
7112 		link_up = true;
7113 	}
7114 
7115 	if (adapter->ixgbe_ieee_pfc)
7116 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
7117 
7118 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
7119 		hw->mac.ops.fc_enable(hw);
7120 		ixgbe_set_rx_drop_en(adapter);
7121 	}
7122 
7123 	if (link_up ||
7124 	    time_after(jiffies, (adapter->link_check_timeout +
7125 				 IXGBE_TRY_LINK_TIMEOUT))) {
7126 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
7127 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
7128 		IXGBE_WRITE_FLUSH(hw);
7129 	}
7130 
7131 	adapter->link_up = link_up;
7132 	adapter->link_speed = link_speed;
7133 }
7134 
7135 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
7136 {
7137 #ifdef CONFIG_IXGBE_DCB
7138 	struct net_device *netdev = adapter->netdev;
7139 	struct dcb_app app = {
7140 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
7141 			      .protocol = 0,
7142 			     };
7143 	u8 up = 0;
7144 
7145 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
7146 		up = dcb_ieee_getapp_mask(netdev, &app);
7147 
7148 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
7149 #endif
7150 }
7151 
7152 /**
7153  * ixgbe_watchdog_link_is_up - update netif_carrier status and
7154  *                             print link up message
7155  * @adapter: pointer to the device adapter structure
7156  **/
7157 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
7158 {
7159 	struct net_device *netdev = adapter->netdev;
7160 	struct ixgbe_hw *hw = &adapter->hw;
7161 	u32 link_speed = adapter->link_speed;
7162 	const char *speed_str;
7163 	bool flow_rx, flow_tx;
7164 
7165 	/* only continue if link was previously down */
7166 	if (netif_carrier_ok(netdev))
7167 		return;
7168 
7169 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
7170 
7171 	switch (hw->mac.type) {
7172 	case ixgbe_mac_82598EB: {
7173 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
7174 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
7175 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
7176 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
7177 	}
7178 		break;
7179 	case ixgbe_mac_X540:
7180 	case ixgbe_mac_X550:
7181 	case ixgbe_mac_X550EM_x:
7182 	case ixgbe_mac_x550em_a:
7183 	case ixgbe_mac_82599EB: {
7184 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
7185 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
7186 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
7187 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
7188 	}
7189 		break;
7190 	default:
7191 		flow_tx = false;
7192 		flow_rx = false;
7193 		break;
7194 	}
7195 
7196 	adapter->last_rx_ptp_check = jiffies;
7197 
7198 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7199 		ixgbe_ptp_start_cyclecounter(adapter);
7200 
7201 	switch (link_speed) {
7202 	case IXGBE_LINK_SPEED_10GB_FULL:
7203 		speed_str = "10 Gbps";
7204 		break;
7205 	case IXGBE_LINK_SPEED_5GB_FULL:
7206 		speed_str = "5 Gbps";
7207 		break;
7208 	case IXGBE_LINK_SPEED_2_5GB_FULL:
7209 		speed_str = "2.5 Gbps";
7210 		break;
7211 	case IXGBE_LINK_SPEED_1GB_FULL:
7212 		speed_str = "1 Gbps";
7213 		break;
7214 	case IXGBE_LINK_SPEED_100_FULL:
7215 		speed_str = "100 Mbps";
7216 		break;
7217 	case IXGBE_LINK_SPEED_10_FULL:
7218 		speed_str = "10 Mbps";
7219 		break;
7220 	default:
7221 		speed_str = "unknown speed";
7222 		break;
7223 	}
7224 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
7225 	       ((flow_rx && flow_tx) ? "RX/TX" :
7226 	       (flow_rx ? "RX" :
7227 	       (flow_tx ? "TX" : "None"))));
7228 
7229 	netif_carrier_on(netdev);
7230 	ixgbe_check_vf_rate_limit(adapter);
7231 
7232 	/* enable transmits */
7233 	netif_tx_wake_all_queues(adapter->netdev);
7234 
7235 	/* update the default user priority for VFs */
7236 	ixgbe_update_default_up(adapter);
7237 
7238 	/* ping all the active vfs to let them know link has changed */
7239 	ixgbe_ping_all_vfs(adapter);
7240 }
7241 
7242 /**
7243  * ixgbe_watchdog_link_is_down - update netif_carrier status and
7244  *                               print link down message
7245  * @adapter: pointer to the adapter structure
7246  **/
7247 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
7248 {
7249 	struct net_device *netdev = adapter->netdev;
7250 	struct ixgbe_hw *hw = &adapter->hw;
7251 
7252 	adapter->link_up = false;
7253 	adapter->link_speed = 0;
7254 
7255 	/* only continue if link was up previously */
7256 	if (!netif_carrier_ok(netdev))
7257 		return;
7258 
7259 	/* poll for SFP+ cable when link is down */
7260 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
7261 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
7262 
7263 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7264 		ixgbe_ptp_start_cyclecounter(adapter);
7265 
7266 	e_info(drv, "NIC Link is Down\n");
7267 	netif_carrier_off(netdev);
7268 
7269 	/* ping all the active vfs to let them know link has changed */
7270 	ixgbe_ping_all_vfs(adapter);
7271 }
7272 
7273 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
7274 {
7275 	int i;
7276 
7277 	for (i = 0; i < adapter->num_tx_queues; i++) {
7278 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
7279 
7280 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
7281 			return true;
7282 	}
7283 
7284 	for (i = 0; i < adapter->num_xdp_queues; i++) {
7285 		struct ixgbe_ring *ring = adapter->xdp_ring[i];
7286 
7287 		if (ring->next_to_use != ring->next_to_clean)
7288 			return true;
7289 	}
7290 
7291 	return false;
7292 }
7293 
7294 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
7295 {
7296 	struct ixgbe_hw *hw = &adapter->hw;
7297 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
7298 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
7299 
7300 	int i, j;
7301 
7302 	if (!adapter->num_vfs)
7303 		return false;
7304 
7305 	/* resetting the PF is only needed for MAC before X550 */
7306 	if (hw->mac.type >= ixgbe_mac_X550)
7307 		return false;
7308 
7309 	for (i = 0; i < adapter->num_vfs; i++) {
7310 		for (j = 0; j < q_per_pool; j++) {
7311 			u32 h, t;
7312 
7313 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
7314 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
7315 
7316 			if (h != t)
7317 				return true;
7318 		}
7319 	}
7320 
7321 	return false;
7322 }
7323 
7324 /**
7325  * ixgbe_watchdog_flush_tx - flush queues on link down
7326  * @adapter: pointer to the device adapter structure
7327  **/
7328 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
7329 {
7330 	if (!netif_carrier_ok(adapter->netdev)) {
7331 		if (ixgbe_ring_tx_pending(adapter) ||
7332 		    ixgbe_vf_tx_pending(adapter)) {
7333 			/* We've lost link, so the controller stops DMA,
7334 			 * but we've got queued Tx work that's never going
7335 			 * to get done, so reset controller to flush Tx.
7336 			 * (Do the reset outside of interrupt context).
7337 			 */
7338 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
7339 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
7340 		}
7341 	}
7342 }
7343 
7344 #ifdef CONFIG_PCI_IOV
7345 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
7346 {
7347 	struct ixgbe_hw *hw = &adapter->hw;
7348 	struct pci_dev *pdev = adapter->pdev;
7349 	unsigned int vf;
7350 	u32 gpc;
7351 
7352 	if (!(netif_carrier_ok(adapter->netdev)))
7353 		return;
7354 
7355 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
7356 	if (gpc) /* If incrementing then no need for the check below */
7357 		return;
7358 	/* Check to see if a bad DMA write target from an errant or
7359 	 * malicious VF has caused a PCIe error.  If so then we can
7360 	 * issue a VFLR to the offending VF(s) and then resume without
7361 	 * requesting a full slot reset.
7362 	 */
7363 
7364 	if (!pdev)
7365 		return;
7366 
7367 	/* check status reg for all VFs owned by this PF */
7368 	for (vf = 0; vf < adapter->num_vfs; ++vf) {
7369 		struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
7370 		u16 status_reg;
7371 
7372 		if (!vfdev)
7373 			continue;
7374 		pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
7375 		if (status_reg != IXGBE_FAILED_READ_CFG_WORD &&
7376 		    status_reg & PCI_STATUS_REC_MASTER_ABORT)
7377 			pcie_flr(vfdev);
7378 	}
7379 }
7380 
7381 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
7382 {
7383 	u32 ssvpc;
7384 
7385 	/* Do not perform spoof check for 82598 or if not in IOV mode */
7386 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7387 	    adapter->num_vfs == 0)
7388 		return;
7389 
7390 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
7391 
7392 	/*
7393 	 * ssvpc register is cleared on read, if zero then no
7394 	 * spoofed packets in the last interval.
7395 	 */
7396 	if (!ssvpc)
7397 		return;
7398 
7399 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
7400 }
7401 #else
7402 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
7403 {
7404 }
7405 
7406 static void
7407 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
7408 {
7409 }
7410 #endif /* CONFIG_PCI_IOV */
7411 
7412 
7413 /**
7414  * ixgbe_watchdog_subtask - check and bring link up
7415  * @adapter: pointer to the device adapter structure
7416  **/
7417 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
7418 {
7419 	/* if interface is down, removing or resetting, do nothing */
7420 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7421 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7422 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7423 		return;
7424 
7425 	ixgbe_watchdog_update_link(adapter);
7426 
7427 	if (adapter->link_up)
7428 		ixgbe_watchdog_link_is_up(adapter);
7429 	else
7430 		ixgbe_watchdog_link_is_down(adapter);
7431 
7432 	ixgbe_check_for_bad_vf(adapter);
7433 	ixgbe_spoof_check(adapter);
7434 	ixgbe_update_stats(adapter);
7435 
7436 	ixgbe_watchdog_flush_tx(adapter);
7437 }
7438 
7439 /**
7440  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
7441  * @adapter: the ixgbe adapter structure
7442  **/
7443 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
7444 {
7445 	struct ixgbe_hw *hw = &adapter->hw;
7446 	s32 err;
7447 
7448 	/* not searching for SFP so there is nothing to do here */
7449 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
7450 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7451 		return;
7452 
7453 	if (adapter->sfp_poll_time &&
7454 	    time_after(adapter->sfp_poll_time, jiffies))
7455 		return; /* If not yet time to poll for SFP */
7456 
7457 	/* someone else is in init, wait until next service event */
7458 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7459 		return;
7460 
7461 	adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
7462 
7463 	err = hw->phy.ops.identify_sfp(hw);
7464 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7465 		goto sfp_out;
7466 
7467 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
7468 		/* If no cable is present, then we need to reset
7469 		 * the next time we find a good cable. */
7470 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
7471 	}
7472 
7473 	/* exit on error */
7474 	if (err)
7475 		goto sfp_out;
7476 
7477 	/* exit if reset not needed */
7478 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7479 		goto sfp_out;
7480 
7481 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
7482 
7483 	/*
7484 	 * A module may be identified correctly, but the EEPROM may not have
7485 	 * support for that module.  setup_sfp() will fail in that case, so
7486 	 * we should not allow that module to load.
7487 	 */
7488 	if (hw->mac.type == ixgbe_mac_82598EB)
7489 		err = hw->phy.ops.reset(hw);
7490 	else
7491 		err = hw->mac.ops.setup_sfp(hw);
7492 
7493 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7494 		goto sfp_out;
7495 
7496 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
7497 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
7498 
7499 sfp_out:
7500 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7501 
7502 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
7503 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
7504 		e_dev_err("failed to initialize because an unsupported "
7505 			  "SFP+ module type was detected.\n");
7506 		e_dev_err("Reload the driver after installing a "
7507 			  "supported module.\n");
7508 		unregister_netdev(adapter->netdev);
7509 	}
7510 }
7511 
7512 /**
7513  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
7514  * @adapter: the ixgbe adapter structure
7515  **/
7516 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
7517 {
7518 	struct ixgbe_hw *hw = &adapter->hw;
7519 	u32 cap_speed;
7520 	u32 speed;
7521 	bool autoneg = false;
7522 
7523 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
7524 		return;
7525 
7526 	/* someone else is in init, wait until next service event */
7527 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7528 		return;
7529 
7530 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
7531 
7532 	hw->mac.ops.get_link_capabilities(hw, &cap_speed, &autoneg);
7533 
7534 	/* advertise highest capable link speed */
7535 	if (!autoneg && (cap_speed & IXGBE_LINK_SPEED_10GB_FULL))
7536 		speed = IXGBE_LINK_SPEED_10GB_FULL;
7537 	else
7538 		speed = cap_speed & (IXGBE_LINK_SPEED_10GB_FULL |
7539 				     IXGBE_LINK_SPEED_1GB_FULL);
7540 
7541 	if (hw->mac.ops.setup_link)
7542 		hw->mac.ops.setup_link(hw, speed, true);
7543 
7544 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
7545 	adapter->link_check_timeout = jiffies;
7546 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7547 }
7548 
7549 /**
7550  * ixgbe_service_timer - Timer Call-back
7551  * @t: pointer to timer_list structure
7552  **/
7553 static void ixgbe_service_timer(struct timer_list *t)
7554 {
7555 	struct ixgbe_adapter *adapter = from_timer(adapter, t, service_timer);
7556 	unsigned long next_event_offset;
7557 
7558 	/* poll faster when waiting for link */
7559 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
7560 		next_event_offset = HZ / 10;
7561 	else
7562 		next_event_offset = HZ * 2;
7563 
7564 	/* Reset the timer */
7565 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
7566 
7567 	ixgbe_service_event_schedule(adapter);
7568 }
7569 
7570 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
7571 {
7572 	struct ixgbe_hw *hw = &adapter->hw;
7573 	u32 status;
7574 
7575 	if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
7576 		return;
7577 
7578 	adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
7579 
7580 	if (!hw->phy.ops.handle_lasi)
7581 		return;
7582 
7583 	status = hw->phy.ops.handle_lasi(&adapter->hw);
7584 	if (status != IXGBE_ERR_OVERTEMP)
7585 		return;
7586 
7587 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
7588 }
7589 
7590 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
7591 {
7592 	if (!test_and_clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state))
7593 		return;
7594 
7595 	rtnl_lock();
7596 	/* If we're already down, removing or resetting, just bail */
7597 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7598 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7599 	    test_bit(__IXGBE_RESETTING, &adapter->state)) {
7600 		rtnl_unlock();
7601 		return;
7602 	}
7603 
7604 	ixgbe_dump(adapter);
7605 	netdev_err(adapter->netdev, "Reset adapter\n");
7606 	adapter->tx_timeout_count++;
7607 
7608 	ixgbe_reinit_locked(adapter);
7609 	rtnl_unlock();
7610 }
7611 
7612 /**
7613  * ixgbe_service_task - manages and runs subtasks
7614  * @work: pointer to work_struct containing our data
7615  **/
7616 static void ixgbe_service_task(struct work_struct *work)
7617 {
7618 	struct ixgbe_adapter *adapter = container_of(work,
7619 						     struct ixgbe_adapter,
7620 						     service_task);
7621 	if (ixgbe_removed(adapter->hw.hw_addr)) {
7622 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
7623 			rtnl_lock();
7624 			ixgbe_down(adapter);
7625 			rtnl_unlock();
7626 		}
7627 		ixgbe_service_event_complete(adapter);
7628 		return;
7629 	}
7630 	if (adapter->flags2 & IXGBE_FLAG2_UDP_TUN_REREG_NEEDED) {
7631 		rtnl_lock();
7632 		adapter->flags2 &= ~IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
7633 		udp_tunnel_get_rx_info(adapter->netdev);
7634 		rtnl_unlock();
7635 	}
7636 	ixgbe_reset_subtask(adapter);
7637 	ixgbe_phy_interrupt_subtask(adapter);
7638 	ixgbe_sfp_detection_subtask(adapter);
7639 	ixgbe_sfp_link_config_subtask(adapter);
7640 	ixgbe_check_overtemp_subtask(adapter);
7641 	ixgbe_watchdog_subtask(adapter);
7642 	ixgbe_fdir_reinit_subtask(adapter);
7643 	ixgbe_check_hang_subtask(adapter);
7644 
7645 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
7646 		ixgbe_ptp_overflow_check(adapter);
7647 		if (adapter->flags & IXGBE_FLAG_RX_HWTSTAMP_IN_REGISTER)
7648 			ixgbe_ptp_rx_hang(adapter);
7649 		ixgbe_ptp_tx_hang(adapter);
7650 	}
7651 
7652 	ixgbe_service_event_complete(adapter);
7653 }
7654 
7655 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
7656 		     struct ixgbe_tx_buffer *first,
7657 		     u8 *hdr_len,
7658 		     struct ixgbe_ipsec_tx_data *itd)
7659 {
7660 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
7661 	struct sk_buff *skb = first->skb;
7662 	union {
7663 		struct iphdr *v4;
7664 		struct ipv6hdr *v6;
7665 		unsigned char *hdr;
7666 	} ip;
7667 	union {
7668 		struct tcphdr *tcp;
7669 		unsigned char *hdr;
7670 	} l4;
7671 	u32 paylen, l4_offset;
7672 	u32 fceof_saidx = 0;
7673 	int err;
7674 
7675 	if (skb->ip_summed != CHECKSUM_PARTIAL)
7676 		return 0;
7677 
7678 	if (!skb_is_gso(skb))
7679 		return 0;
7680 
7681 	err = skb_cow_head(skb, 0);
7682 	if (err < 0)
7683 		return err;
7684 
7685 	if (eth_p_mpls(first->protocol))
7686 		ip.hdr = skb_inner_network_header(skb);
7687 	else
7688 		ip.hdr = skb_network_header(skb);
7689 	l4.hdr = skb_checksum_start(skb);
7690 
7691 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
7692 	type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7693 
7694 	/* initialize outer IP header fields */
7695 	if (ip.v4->version == 4) {
7696 		unsigned char *csum_start = skb_checksum_start(skb);
7697 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
7698 		int len = csum_start - trans_start;
7699 
7700 		/* IP header will have to cancel out any data that
7701 		 * is not a part of the outer IP header, so set to
7702 		 * a reverse csum if needed, else init check to 0.
7703 		 */
7704 		ip.v4->check = (skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL) ?
7705 					   csum_fold(csum_partial(trans_start,
7706 								  len, 0)) : 0;
7707 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
7708 
7709 		ip.v4->tot_len = 0;
7710 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7711 				   IXGBE_TX_FLAGS_CSUM |
7712 				   IXGBE_TX_FLAGS_IPV4;
7713 	} else {
7714 		ip.v6->payload_len = 0;
7715 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
7716 				   IXGBE_TX_FLAGS_CSUM;
7717 	}
7718 
7719 	/* determine offset of inner transport header */
7720 	l4_offset = l4.hdr - skb->data;
7721 
7722 	/* compute length of segmentation header */
7723 	*hdr_len = (l4.tcp->doff * 4) + l4_offset;
7724 
7725 	/* remove payload length from inner checksum */
7726 	paylen = skb->len - l4_offset;
7727 	csum_replace_by_diff(&l4.tcp->check, (__force __wsum)htonl(paylen));
7728 
7729 	/* update gso size and bytecount with header size */
7730 	first->gso_segs = skb_shinfo(skb)->gso_segs;
7731 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
7732 
7733 	/* mss_l4len_id: use 0 as index for TSO */
7734 	mss_l4len_idx = (*hdr_len - l4_offset) << IXGBE_ADVTXD_L4LEN_SHIFT;
7735 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
7736 
7737 	fceof_saidx |= itd->sa_idx;
7738 	type_tucmd |= itd->flags | itd->trailer_len;
7739 
7740 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
7741 	vlan_macip_lens = l4.hdr - ip.hdr;
7742 	vlan_macip_lens |= (ip.hdr - skb->data) << IXGBE_ADVTXD_MACLEN_SHIFT;
7743 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7744 
7745 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd,
7746 			  mss_l4len_idx);
7747 
7748 	return 1;
7749 }
7750 
7751 static inline bool ixgbe_ipv6_csum_is_sctp(struct sk_buff *skb)
7752 {
7753 	unsigned int offset = 0;
7754 
7755 	ipv6_find_hdr(skb, &offset, IPPROTO_SCTP, NULL, NULL);
7756 
7757 	return offset == skb_checksum_start_offset(skb);
7758 }
7759 
7760 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
7761 			  struct ixgbe_tx_buffer *first,
7762 			  struct ixgbe_ipsec_tx_data *itd)
7763 {
7764 	struct sk_buff *skb = first->skb;
7765 	u32 vlan_macip_lens = 0;
7766 	u32 fceof_saidx = 0;
7767 	u32 type_tucmd = 0;
7768 
7769 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
7770 csum_failed:
7771 		if (!(first->tx_flags & (IXGBE_TX_FLAGS_HW_VLAN |
7772 					 IXGBE_TX_FLAGS_CC)))
7773 			return;
7774 		goto no_csum;
7775 	}
7776 
7777 	switch (skb->csum_offset) {
7778 	case offsetof(struct tcphdr, check):
7779 		type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7780 		/* fall through */
7781 	case offsetof(struct udphdr, check):
7782 		break;
7783 	case offsetof(struct sctphdr, checksum):
7784 		/* validate that this is actually an SCTP request */
7785 		if (((first->protocol == htons(ETH_P_IP)) &&
7786 		     (ip_hdr(skb)->protocol == IPPROTO_SCTP)) ||
7787 		    ((first->protocol == htons(ETH_P_IPV6)) &&
7788 		     ixgbe_ipv6_csum_is_sctp(skb))) {
7789 			type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_SCTP;
7790 			break;
7791 		}
7792 		/* fall through */
7793 	default:
7794 		skb_checksum_help(skb);
7795 		goto csum_failed;
7796 	}
7797 
7798 	/* update TX checksum flag */
7799 	first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
7800 	vlan_macip_lens = skb_checksum_start_offset(skb) -
7801 			  skb_network_offset(skb);
7802 no_csum:
7803 	/* vlan_macip_lens: MACLEN, VLAN tag */
7804 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
7805 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
7806 
7807 	fceof_saidx |= itd->sa_idx;
7808 	type_tucmd |= itd->flags | itd->trailer_len;
7809 
7810 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd, 0);
7811 }
7812 
7813 #define IXGBE_SET_FLAG(_input, _flag, _result) \
7814 	((_flag <= _result) ? \
7815 	 ((u32)(_input & _flag) * (_result / _flag)) : \
7816 	 ((u32)(_input & _flag) / (_flag / _result)))
7817 
7818 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
7819 {
7820 	/* set type for advanced descriptor with frame checksum insertion */
7821 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
7822 		       IXGBE_ADVTXD_DCMD_DEXT |
7823 		       IXGBE_ADVTXD_DCMD_IFCS;
7824 
7825 	/* set HW vlan bit if vlan is present */
7826 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
7827 				   IXGBE_ADVTXD_DCMD_VLE);
7828 
7829 	/* set segmentation enable bits for TSO/FSO */
7830 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
7831 				   IXGBE_ADVTXD_DCMD_TSE);
7832 
7833 	/* set timestamp bit if present */
7834 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
7835 				   IXGBE_ADVTXD_MAC_TSTAMP);
7836 
7837 	/* insert frame checksum */
7838 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
7839 
7840 	return cmd_type;
7841 }
7842 
7843 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
7844 				   u32 tx_flags, unsigned int paylen)
7845 {
7846 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
7847 
7848 	/* enable L4 checksum for TSO and TX checksum offload */
7849 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7850 					IXGBE_TX_FLAGS_CSUM,
7851 					IXGBE_ADVTXD_POPTS_TXSM);
7852 
7853 	/* enable IPv4 checksum for TSO */
7854 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7855 					IXGBE_TX_FLAGS_IPV4,
7856 					IXGBE_ADVTXD_POPTS_IXSM);
7857 
7858 	/* enable IPsec */
7859 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7860 					IXGBE_TX_FLAGS_IPSEC,
7861 					IXGBE_ADVTXD_POPTS_IPSEC);
7862 
7863 	/*
7864 	 * Check Context must be set if Tx switch is enabled, which it
7865 	 * always is for case where virtual functions are running
7866 	 */
7867 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
7868 					IXGBE_TX_FLAGS_CC,
7869 					IXGBE_ADVTXD_CC);
7870 
7871 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
7872 }
7873 
7874 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7875 {
7876 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
7877 
7878 	/* Herbert's original patch had:
7879 	 *  smp_mb__after_netif_stop_queue();
7880 	 * but since that doesn't exist yet, just open code it.
7881 	 */
7882 	smp_mb();
7883 
7884 	/* We need to check again in a case another CPU has just
7885 	 * made room available.
7886 	 */
7887 	if (likely(ixgbe_desc_unused(tx_ring) < size))
7888 		return -EBUSY;
7889 
7890 	/* A reprieve! - use start_queue because it doesn't call schedule */
7891 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
7892 	++tx_ring->tx_stats.restart_queue;
7893 	return 0;
7894 }
7895 
7896 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
7897 {
7898 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
7899 		return 0;
7900 
7901 	return __ixgbe_maybe_stop_tx(tx_ring, size);
7902 }
7903 
7904 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
7905 		       IXGBE_TXD_CMD_RS)
7906 
7907 static int ixgbe_tx_map(struct ixgbe_ring *tx_ring,
7908 			struct ixgbe_tx_buffer *first,
7909 			const u8 hdr_len)
7910 {
7911 	struct sk_buff *skb = first->skb;
7912 	struct ixgbe_tx_buffer *tx_buffer;
7913 	union ixgbe_adv_tx_desc *tx_desc;
7914 	struct skb_frag_struct *frag;
7915 	dma_addr_t dma;
7916 	unsigned int data_len, size;
7917 	u32 tx_flags = first->tx_flags;
7918 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
7919 	u16 i = tx_ring->next_to_use;
7920 
7921 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
7922 
7923 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
7924 
7925 	size = skb_headlen(skb);
7926 	data_len = skb->data_len;
7927 
7928 #ifdef IXGBE_FCOE
7929 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
7930 		if (data_len < sizeof(struct fcoe_crc_eof)) {
7931 			size -= sizeof(struct fcoe_crc_eof) - data_len;
7932 			data_len = 0;
7933 		} else {
7934 			data_len -= sizeof(struct fcoe_crc_eof);
7935 		}
7936 	}
7937 
7938 #endif
7939 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
7940 
7941 	tx_buffer = first;
7942 
7943 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
7944 		if (dma_mapping_error(tx_ring->dev, dma))
7945 			goto dma_error;
7946 
7947 		/* record length, and DMA address */
7948 		dma_unmap_len_set(tx_buffer, len, size);
7949 		dma_unmap_addr_set(tx_buffer, dma, dma);
7950 
7951 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
7952 
7953 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
7954 			tx_desc->read.cmd_type_len =
7955 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
7956 
7957 			i++;
7958 			tx_desc++;
7959 			if (i == tx_ring->count) {
7960 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7961 				i = 0;
7962 			}
7963 			tx_desc->read.olinfo_status = 0;
7964 
7965 			dma += IXGBE_MAX_DATA_PER_TXD;
7966 			size -= IXGBE_MAX_DATA_PER_TXD;
7967 
7968 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
7969 		}
7970 
7971 		if (likely(!data_len))
7972 			break;
7973 
7974 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
7975 
7976 		i++;
7977 		tx_desc++;
7978 		if (i == tx_ring->count) {
7979 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
7980 			i = 0;
7981 		}
7982 		tx_desc->read.olinfo_status = 0;
7983 
7984 #ifdef IXGBE_FCOE
7985 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
7986 #else
7987 		size = skb_frag_size(frag);
7988 #endif
7989 		data_len -= size;
7990 
7991 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
7992 				       DMA_TO_DEVICE);
7993 
7994 		tx_buffer = &tx_ring->tx_buffer_info[i];
7995 	}
7996 
7997 	/* write last descriptor with RS and EOP bits */
7998 	cmd_type |= size | IXGBE_TXD_CMD;
7999 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8000 
8001 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
8002 
8003 	/* set the timestamp */
8004 	first->time_stamp = jiffies;
8005 
8006 	/*
8007 	 * Force memory writes to complete before letting h/w know there
8008 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
8009 	 * memory model archs, such as IA-64).
8010 	 *
8011 	 * We also need this memory barrier to make certain all of the
8012 	 * status bits have been updated before next_to_watch is written.
8013 	 */
8014 	wmb();
8015 
8016 	/* set next_to_watch value indicating a packet is present */
8017 	first->next_to_watch = tx_desc;
8018 
8019 	i++;
8020 	if (i == tx_ring->count)
8021 		i = 0;
8022 
8023 	tx_ring->next_to_use = i;
8024 
8025 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
8026 
8027 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
8028 		writel(i, tx_ring->tail);
8029 
8030 		/* we need this if more than one processor can write to our tail
8031 		 * at a time, it synchronizes IO on IA64/Altix systems
8032 		 */
8033 		mmiowb();
8034 	}
8035 
8036 	return 0;
8037 dma_error:
8038 	dev_err(tx_ring->dev, "TX DMA map failed\n");
8039 
8040 	/* clear dma mappings for failed tx_buffer_info map */
8041 	for (;;) {
8042 		tx_buffer = &tx_ring->tx_buffer_info[i];
8043 		if (dma_unmap_len(tx_buffer, len))
8044 			dma_unmap_page(tx_ring->dev,
8045 				       dma_unmap_addr(tx_buffer, dma),
8046 				       dma_unmap_len(tx_buffer, len),
8047 				       DMA_TO_DEVICE);
8048 		dma_unmap_len_set(tx_buffer, len, 0);
8049 		if (tx_buffer == first)
8050 			break;
8051 		if (i == 0)
8052 			i += tx_ring->count;
8053 		i--;
8054 	}
8055 
8056 	dev_kfree_skb_any(first->skb);
8057 	first->skb = NULL;
8058 
8059 	tx_ring->next_to_use = i;
8060 
8061 	return -1;
8062 }
8063 
8064 static void ixgbe_atr(struct ixgbe_ring *ring,
8065 		      struct ixgbe_tx_buffer *first)
8066 {
8067 	struct ixgbe_q_vector *q_vector = ring->q_vector;
8068 	union ixgbe_atr_hash_dword input = { .dword = 0 };
8069 	union ixgbe_atr_hash_dword common = { .dword = 0 };
8070 	union {
8071 		unsigned char *network;
8072 		struct iphdr *ipv4;
8073 		struct ipv6hdr *ipv6;
8074 	} hdr;
8075 	struct tcphdr *th;
8076 	unsigned int hlen;
8077 	struct sk_buff *skb;
8078 	__be16 vlan_id;
8079 	int l4_proto;
8080 
8081 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
8082 	if (!q_vector)
8083 		return;
8084 
8085 	/* do nothing if sampling is disabled */
8086 	if (!ring->atr_sample_rate)
8087 		return;
8088 
8089 	ring->atr_count++;
8090 
8091 	/* currently only IPv4/IPv6 with TCP is supported */
8092 	if ((first->protocol != htons(ETH_P_IP)) &&
8093 	    (first->protocol != htons(ETH_P_IPV6)))
8094 		return;
8095 
8096 	/* snag network header to get L4 type and address */
8097 	skb = first->skb;
8098 	hdr.network = skb_network_header(skb);
8099 	if (unlikely(hdr.network <= skb->data))
8100 		return;
8101 	if (skb->encapsulation &&
8102 	    first->protocol == htons(ETH_P_IP) &&
8103 	    hdr.ipv4->protocol == IPPROTO_UDP) {
8104 		struct ixgbe_adapter *adapter = q_vector->adapter;
8105 
8106 		if (unlikely(skb_tail_pointer(skb) < hdr.network +
8107 			     VXLAN_HEADROOM))
8108 			return;
8109 
8110 		/* verify the port is recognized as VXLAN */
8111 		if (adapter->vxlan_port &&
8112 		    udp_hdr(skb)->dest == adapter->vxlan_port)
8113 			hdr.network = skb_inner_network_header(skb);
8114 
8115 		if (adapter->geneve_port &&
8116 		    udp_hdr(skb)->dest == adapter->geneve_port)
8117 			hdr.network = skb_inner_network_header(skb);
8118 	}
8119 
8120 	/* Make sure we have at least [minimum IPv4 header + TCP]
8121 	 * or [IPv6 header] bytes
8122 	 */
8123 	if (unlikely(skb_tail_pointer(skb) < hdr.network + 40))
8124 		return;
8125 
8126 	/* Currently only IPv4/IPv6 with TCP is supported */
8127 	switch (hdr.ipv4->version) {
8128 	case IPVERSION:
8129 		/* access ihl as u8 to avoid unaligned access on ia64 */
8130 		hlen = (hdr.network[0] & 0x0F) << 2;
8131 		l4_proto = hdr.ipv4->protocol;
8132 		break;
8133 	case 6:
8134 		hlen = hdr.network - skb->data;
8135 		l4_proto = ipv6_find_hdr(skb, &hlen, IPPROTO_TCP, NULL, NULL);
8136 		hlen -= hdr.network - skb->data;
8137 		break;
8138 	default:
8139 		return;
8140 	}
8141 
8142 	if (l4_proto != IPPROTO_TCP)
8143 		return;
8144 
8145 	if (unlikely(skb_tail_pointer(skb) < hdr.network +
8146 		     hlen + sizeof(struct tcphdr)))
8147 		return;
8148 
8149 	th = (struct tcphdr *)(hdr.network + hlen);
8150 
8151 	/* skip this packet since the socket is closing */
8152 	if (th->fin)
8153 		return;
8154 
8155 	/* sample on all syn packets or once every atr sample count */
8156 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
8157 		return;
8158 
8159 	/* reset sample count */
8160 	ring->atr_count = 0;
8161 
8162 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
8163 
8164 	/*
8165 	 * src and dst are inverted, think how the receiver sees them
8166 	 *
8167 	 * The input is broken into two sections, a non-compressed section
8168 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
8169 	 * is XORed together and stored in the compressed dword.
8170 	 */
8171 	input.formatted.vlan_id = vlan_id;
8172 
8173 	/*
8174 	 * since src port and flex bytes occupy the same word XOR them together
8175 	 * and write the value to source port portion of compressed dword
8176 	 */
8177 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
8178 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
8179 	else
8180 		common.port.src ^= th->dest ^ first->protocol;
8181 	common.port.dst ^= th->source;
8182 
8183 	switch (hdr.ipv4->version) {
8184 	case IPVERSION:
8185 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
8186 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
8187 		break;
8188 	case 6:
8189 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
8190 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
8191 			     hdr.ipv6->saddr.s6_addr32[1] ^
8192 			     hdr.ipv6->saddr.s6_addr32[2] ^
8193 			     hdr.ipv6->saddr.s6_addr32[3] ^
8194 			     hdr.ipv6->daddr.s6_addr32[0] ^
8195 			     hdr.ipv6->daddr.s6_addr32[1] ^
8196 			     hdr.ipv6->daddr.s6_addr32[2] ^
8197 			     hdr.ipv6->daddr.s6_addr32[3];
8198 		break;
8199 	default:
8200 		break;
8201 	}
8202 
8203 	if (hdr.network != skb_network_header(skb))
8204 		input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
8205 
8206 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
8207 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
8208 					      input, common, ring->queue_index);
8209 }
8210 
8211 #ifdef IXGBE_FCOE
8212 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
8213 			      struct net_device *sb_dev,
8214 			      select_queue_fallback_t fallback)
8215 {
8216 	struct ixgbe_adapter *adapter;
8217 	struct ixgbe_ring_feature *f;
8218 	int txq;
8219 
8220 	if (sb_dev) {
8221 		u8 tc = netdev_get_prio_tc_map(dev, skb->priority);
8222 		struct net_device *vdev = sb_dev;
8223 
8224 		txq = vdev->tc_to_txq[tc].offset;
8225 		txq += reciprocal_scale(skb_get_hash(skb),
8226 					vdev->tc_to_txq[tc].count);
8227 
8228 		return txq;
8229 	}
8230 
8231 	/*
8232 	 * only execute the code below if protocol is FCoE
8233 	 * or FIP and we have FCoE enabled on the adapter
8234 	 */
8235 	switch (vlan_get_protocol(skb)) {
8236 	case htons(ETH_P_FCOE):
8237 	case htons(ETH_P_FIP):
8238 		adapter = netdev_priv(dev);
8239 
8240 		if (!sb_dev && (adapter->flags & IXGBE_FLAG_FCOE_ENABLED))
8241 			break;
8242 		/* fall through */
8243 	default:
8244 		return fallback(dev, skb, sb_dev);
8245 	}
8246 
8247 	f = &adapter->ring_feature[RING_F_FCOE];
8248 
8249 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
8250 					   smp_processor_id();
8251 
8252 	while (txq >= f->indices)
8253 		txq -= f->indices;
8254 
8255 	return txq + f->offset;
8256 }
8257 
8258 #endif
8259 static int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
8260 			       struct xdp_frame *xdpf)
8261 {
8262 	struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
8263 	struct ixgbe_tx_buffer *tx_buffer;
8264 	union ixgbe_adv_tx_desc *tx_desc;
8265 	u32 len, cmd_type;
8266 	dma_addr_t dma;
8267 	u16 i;
8268 
8269 	len = xdpf->len;
8270 
8271 	if (unlikely(!ixgbe_desc_unused(ring)))
8272 		return IXGBE_XDP_CONSUMED;
8273 
8274 	dma = dma_map_single(ring->dev, xdpf->data, len, DMA_TO_DEVICE);
8275 	if (dma_mapping_error(ring->dev, dma))
8276 		return IXGBE_XDP_CONSUMED;
8277 
8278 	/* record the location of the first descriptor for this packet */
8279 	tx_buffer = &ring->tx_buffer_info[ring->next_to_use];
8280 	tx_buffer->bytecount = len;
8281 	tx_buffer->gso_segs = 1;
8282 	tx_buffer->protocol = 0;
8283 
8284 	i = ring->next_to_use;
8285 	tx_desc = IXGBE_TX_DESC(ring, i);
8286 
8287 	dma_unmap_len_set(tx_buffer, len, len);
8288 	dma_unmap_addr_set(tx_buffer, dma, dma);
8289 	tx_buffer->xdpf = xdpf;
8290 
8291 	tx_desc->read.buffer_addr = cpu_to_le64(dma);
8292 
8293 	/* put descriptor type bits */
8294 	cmd_type = IXGBE_ADVTXD_DTYP_DATA |
8295 		   IXGBE_ADVTXD_DCMD_DEXT |
8296 		   IXGBE_ADVTXD_DCMD_IFCS;
8297 	cmd_type |= len | IXGBE_TXD_CMD;
8298 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8299 	tx_desc->read.olinfo_status =
8300 		cpu_to_le32(len << IXGBE_ADVTXD_PAYLEN_SHIFT);
8301 
8302 	/* Avoid any potential race with xdp_xmit and cleanup */
8303 	smp_wmb();
8304 
8305 	/* set next_to_watch value indicating a packet is present */
8306 	i++;
8307 	if (i == ring->count)
8308 		i = 0;
8309 
8310 	tx_buffer->next_to_watch = tx_desc;
8311 	ring->next_to_use = i;
8312 
8313 	return IXGBE_XDP_TX;
8314 }
8315 
8316 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
8317 			  struct ixgbe_adapter *adapter,
8318 			  struct ixgbe_ring *tx_ring)
8319 {
8320 	struct ixgbe_tx_buffer *first;
8321 	int tso;
8322 	u32 tx_flags = 0;
8323 	unsigned short f;
8324 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
8325 	struct ixgbe_ipsec_tx_data ipsec_tx = { 0 };
8326 	__be16 protocol = skb->protocol;
8327 	u8 hdr_len = 0;
8328 
8329 	/*
8330 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
8331 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
8332 	 *       + 2 desc gap to keep tail from touching head,
8333 	 *       + 1 desc for context descriptor,
8334 	 * otherwise try next time
8335 	 */
8336 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
8337 		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
8338 
8339 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
8340 		tx_ring->tx_stats.tx_busy++;
8341 		return NETDEV_TX_BUSY;
8342 	}
8343 
8344 	/* record the location of the first descriptor for this packet */
8345 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
8346 	first->skb = skb;
8347 	first->bytecount = skb->len;
8348 	first->gso_segs = 1;
8349 
8350 	/* if we have a HW VLAN tag being added default to the HW one */
8351 	if (skb_vlan_tag_present(skb)) {
8352 		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
8353 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8354 	/* else if it is a SW VLAN check the next protocol and store the tag */
8355 	} else if (protocol == htons(ETH_P_8021Q)) {
8356 		struct vlan_hdr *vhdr, _vhdr;
8357 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
8358 		if (!vhdr)
8359 			goto out_drop;
8360 
8361 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
8362 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
8363 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
8364 	}
8365 	protocol = vlan_get_protocol(skb);
8366 
8367 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
8368 	    adapter->ptp_clock) {
8369 		if (!test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
8370 					   &adapter->state)) {
8371 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
8372 			tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
8373 
8374 			/* schedule check for Tx timestamp */
8375 			adapter->ptp_tx_skb = skb_get(skb);
8376 			adapter->ptp_tx_start = jiffies;
8377 			schedule_work(&adapter->ptp_tx_work);
8378 		} else {
8379 			adapter->tx_hwtstamp_skipped++;
8380 		}
8381 	}
8382 
8383 	skb_tx_timestamp(skb);
8384 
8385 #ifdef CONFIG_PCI_IOV
8386 	/*
8387 	 * Use the l2switch_enable flag - would be false if the DMA
8388 	 * Tx switch had been disabled.
8389 	 */
8390 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
8391 		tx_flags |= IXGBE_TX_FLAGS_CC;
8392 
8393 #endif
8394 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
8395 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
8396 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
8397 	     (skb->priority != TC_PRIO_CONTROL))) {
8398 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
8399 		tx_flags |= (skb->priority & 0x7) <<
8400 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
8401 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
8402 			struct vlan_ethhdr *vhdr;
8403 
8404 			if (skb_cow_head(skb, 0))
8405 				goto out_drop;
8406 			vhdr = (struct vlan_ethhdr *)skb->data;
8407 			vhdr->h_vlan_TCI = htons(tx_flags >>
8408 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
8409 		} else {
8410 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8411 		}
8412 	}
8413 
8414 	/* record initial flags and protocol */
8415 	first->tx_flags = tx_flags;
8416 	first->protocol = protocol;
8417 
8418 #ifdef IXGBE_FCOE
8419 	/* setup tx offload for FCoE */
8420 	if ((protocol == htons(ETH_P_FCOE)) &&
8421 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
8422 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
8423 		if (tso < 0)
8424 			goto out_drop;
8425 
8426 		goto xmit_fcoe;
8427 	}
8428 
8429 #endif /* IXGBE_FCOE */
8430 
8431 #ifdef CONFIG_XFRM_OFFLOAD
8432 	if (skb->sp && !ixgbe_ipsec_tx(tx_ring, first, &ipsec_tx))
8433 		goto out_drop;
8434 #endif
8435 	tso = ixgbe_tso(tx_ring, first, &hdr_len, &ipsec_tx);
8436 	if (tso < 0)
8437 		goto out_drop;
8438 	else if (!tso)
8439 		ixgbe_tx_csum(tx_ring, first, &ipsec_tx);
8440 
8441 	/* add the ATR filter if ATR is on */
8442 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
8443 		ixgbe_atr(tx_ring, first);
8444 
8445 #ifdef IXGBE_FCOE
8446 xmit_fcoe:
8447 #endif /* IXGBE_FCOE */
8448 	if (ixgbe_tx_map(tx_ring, first, hdr_len))
8449 		goto cleanup_tx_timestamp;
8450 
8451 	return NETDEV_TX_OK;
8452 
8453 out_drop:
8454 	dev_kfree_skb_any(first->skb);
8455 	first->skb = NULL;
8456 cleanup_tx_timestamp:
8457 	if (unlikely(tx_flags & IXGBE_TX_FLAGS_TSTAMP)) {
8458 		dev_kfree_skb_any(adapter->ptp_tx_skb);
8459 		adapter->ptp_tx_skb = NULL;
8460 		cancel_work_sync(&adapter->ptp_tx_work);
8461 		clear_bit_unlock(__IXGBE_PTP_TX_IN_PROGRESS, &adapter->state);
8462 	}
8463 
8464 	return NETDEV_TX_OK;
8465 }
8466 
8467 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
8468 				      struct net_device *netdev,
8469 				      struct ixgbe_ring *ring)
8470 {
8471 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8472 	struct ixgbe_ring *tx_ring;
8473 
8474 	/*
8475 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
8476 	 * in order to meet this minimum size requirement.
8477 	 */
8478 	if (skb_put_padto(skb, 17))
8479 		return NETDEV_TX_OK;
8480 
8481 	tx_ring = ring ? ring : adapter->tx_ring[skb->queue_mapping];
8482 
8483 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
8484 }
8485 
8486 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
8487 				    struct net_device *netdev)
8488 {
8489 	return __ixgbe_xmit_frame(skb, netdev, NULL);
8490 }
8491 
8492 /**
8493  * ixgbe_set_mac - Change the Ethernet Address of the NIC
8494  * @netdev: network interface device structure
8495  * @p: pointer to an address structure
8496  *
8497  * Returns 0 on success, negative on failure
8498  **/
8499 static int ixgbe_set_mac(struct net_device *netdev, void *p)
8500 {
8501 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8502 	struct ixgbe_hw *hw = &adapter->hw;
8503 	struct sockaddr *addr = p;
8504 
8505 	if (!is_valid_ether_addr(addr->sa_data))
8506 		return -EADDRNOTAVAIL;
8507 
8508 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
8509 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
8510 
8511 	ixgbe_mac_set_default_filter(adapter);
8512 
8513 	return 0;
8514 }
8515 
8516 static int
8517 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
8518 {
8519 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8520 	struct ixgbe_hw *hw = &adapter->hw;
8521 	u16 value;
8522 	int rc;
8523 
8524 	if (prtad != hw->phy.mdio.prtad)
8525 		return -EINVAL;
8526 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
8527 	if (!rc)
8528 		rc = value;
8529 	return rc;
8530 }
8531 
8532 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
8533 			    u16 addr, u16 value)
8534 {
8535 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8536 	struct ixgbe_hw *hw = &adapter->hw;
8537 
8538 	if (prtad != hw->phy.mdio.prtad)
8539 		return -EINVAL;
8540 	return hw->phy.ops.write_reg(hw, addr, devad, value);
8541 }
8542 
8543 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
8544 {
8545 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8546 
8547 	switch (cmd) {
8548 	case SIOCSHWTSTAMP:
8549 		return ixgbe_ptp_set_ts_config(adapter, req);
8550 	case SIOCGHWTSTAMP:
8551 		return ixgbe_ptp_get_ts_config(adapter, req);
8552 	case SIOCGMIIPHY:
8553 		if (!adapter->hw.phy.ops.read_reg)
8554 			return -EOPNOTSUPP;
8555 		/* fall through */
8556 	default:
8557 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
8558 	}
8559 }
8560 
8561 /**
8562  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
8563  * netdev->dev_addrs
8564  * @dev: network interface device structure
8565  *
8566  * Returns non-zero on failure
8567  **/
8568 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
8569 {
8570 	int err = 0;
8571 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8572 	struct ixgbe_hw *hw = &adapter->hw;
8573 
8574 	if (is_valid_ether_addr(hw->mac.san_addr)) {
8575 		rtnl_lock();
8576 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
8577 		rtnl_unlock();
8578 
8579 		/* update SAN MAC vmdq pool selection */
8580 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
8581 	}
8582 	return err;
8583 }
8584 
8585 /**
8586  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
8587  * netdev->dev_addrs
8588  * @dev: network interface device structure
8589  *
8590  * Returns non-zero on failure
8591  **/
8592 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
8593 {
8594 	int err = 0;
8595 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8596 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
8597 
8598 	if (is_valid_ether_addr(mac->san_addr)) {
8599 		rtnl_lock();
8600 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
8601 		rtnl_unlock();
8602 	}
8603 	return err;
8604 }
8605 
8606 #ifdef CONFIG_NET_POLL_CONTROLLER
8607 /*
8608  * Polling 'interrupt' - used by things like netconsole to send skbs
8609  * without having to re-enable interrupts. It's not called while
8610  * the interrupt routine is executing.
8611  */
8612 static void ixgbe_netpoll(struct net_device *netdev)
8613 {
8614 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8615 	int i;
8616 
8617 	/* if interface is down do nothing */
8618 	if (test_bit(__IXGBE_DOWN, &adapter->state))
8619 		return;
8620 
8621 	/* loop through and schedule all active queues */
8622 	for (i = 0; i < adapter->num_q_vectors; i++)
8623 		ixgbe_msix_clean_rings(0, adapter->q_vector[i]);
8624 }
8625 
8626 #endif
8627 
8628 static void ixgbe_get_ring_stats64(struct rtnl_link_stats64 *stats,
8629 				   struct ixgbe_ring *ring)
8630 {
8631 	u64 bytes, packets;
8632 	unsigned int start;
8633 
8634 	if (ring) {
8635 		do {
8636 			start = u64_stats_fetch_begin_irq(&ring->syncp);
8637 			packets = ring->stats.packets;
8638 			bytes   = ring->stats.bytes;
8639 		} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8640 		stats->tx_packets += packets;
8641 		stats->tx_bytes   += bytes;
8642 	}
8643 }
8644 
8645 static void ixgbe_get_stats64(struct net_device *netdev,
8646 			      struct rtnl_link_stats64 *stats)
8647 {
8648 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8649 	int i;
8650 
8651 	rcu_read_lock();
8652 	for (i = 0; i < adapter->num_rx_queues; i++) {
8653 		struct ixgbe_ring *ring = READ_ONCE(adapter->rx_ring[i]);
8654 		u64 bytes, packets;
8655 		unsigned int start;
8656 
8657 		if (ring) {
8658 			do {
8659 				start = u64_stats_fetch_begin_irq(&ring->syncp);
8660 				packets = ring->stats.packets;
8661 				bytes   = ring->stats.bytes;
8662 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8663 			stats->rx_packets += packets;
8664 			stats->rx_bytes   += bytes;
8665 		}
8666 	}
8667 
8668 	for (i = 0; i < adapter->num_tx_queues; i++) {
8669 		struct ixgbe_ring *ring = READ_ONCE(adapter->tx_ring[i]);
8670 
8671 		ixgbe_get_ring_stats64(stats, ring);
8672 	}
8673 	for (i = 0; i < adapter->num_xdp_queues; i++) {
8674 		struct ixgbe_ring *ring = READ_ONCE(adapter->xdp_ring[i]);
8675 
8676 		ixgbe_get_ring_stats64(stats, ring);
8677 	}
8678 	rcu_read_unlock();
8679 
8680 	/* following stats updated by ixgbe_watchdog_task() */
8681 	stats->multicast	= netdev->stats.multicast;
8682 	stats->rx_errors	= netdev->stats.rx_errors;
8683 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
8684 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
8685 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
8686 }
8687 
8688 #ifdef CONFIG_IXGBE_DCB
8689 /**
8690  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
8691  * @adapter: pointer to ixgbe_adapter
8692  * @tc: number of traffic classes currently enabled
8693  *
8694  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
8695  * 802.1Q priority maps to a packet buffer that exists.
8696  */
8697 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
8698 {
8699 	struct ixgbe_hw *hw = &adapter->hw;
8700 	u32 reg, rsave;
8701 	int i;
8702 
8703 	/* 82598 have a static priority to TC mapping that can not
8704 	 * be changed so no validation is needed.
8705 	 */
8706 	if (hw->mac.type == ixgbe_mac_82598EB)
8707 		return;
8708 
8709 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
8710 	rsave = reg;
8711 
8712 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
8713 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
8714 
8715 		/* If up2tc is out of bounds default to zero */
8716 		if (up2tc > tc)
8717 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
8718 	}
8719 
8720 	if (reg != rsave)
8721 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
8722 
8723 	return;
8724 }
8725 
8726 /**
8727  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
8728  * @adapter: Pointer to adapter struct
8729  *
8730  * Populate the netdev user priority to tc map
8731  */
8732 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
8733 {
8734 	struct net_device *dev = adapter->netdev;
8735 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
8736 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
8737 	u8 prio;
8738 
8739 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
8740 		u8 tc = 0;
8741 
8742 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
8743 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
8744 		else if (ets)
8745 			tc = ets->prio_tc[prio];
8746 
8747 		netdev_set_prio_tc_map(dev, prio, tc);
8748 	}
8749 }
8750 
8751 #endif /* CONFIG_IXGBE_DCB */
8752 static int ixgbe_reassign_macvlan_pool(struct net_device *vdev, void *data)
8753 {
8754 	struct ixgbe_adapter *adapter = data;
8755 	struct ixgbe_fwd_adapter *accel;
8756 	int pool;
8757 
8758 	/* we only care about macvlans... */
8759 	if (!netif_is_macvlan(vdev))
8760 		return 0;
8761 
8762 	/* that have hardware offload enabled... */
8763 	accel = macvlan_accel_priv(vdev);
8764 	if (!accel)
8765 		return 0;
8766 
8767 	/* If we can relocate to a different bit do so */
8768 	pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
8769 	if (pool < adapter->num_rx_pools) {
8770 		set_bit(pool, adapter->fwd_bitmask);
8771 		accel->pool = pool;
8772 		return 0;
8773 	}
8774 
8775 	/* if we cannot find a free pool then disable the offload */
8776 	netdev_err(vdev, "L2FW offload disabled due to lack of queue resources\n");
8777 	macvlan_release_l2fw_offload(vdev);
8778 
8779 	/* unbind the queues and drop the subordinate channel config */
8780 	netdev_unbind_sb_channel(adapter->netdev, vdev);
8781 	netdev_set_sb_channel(vdev, 0);
8782 
8783 	kfree(accel);
8784 
8785 	return 0;
8786 }
8787 
8788 static void ixgbe_defrag_macvlan_pools(struct net_device *dev)
8789 {
8790 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8791 
8792 	/* flush any stale bits out of the fwd bitmask */
8793 	bitmap_clear(adapter->fwd_bitmask, 1, 63);
8794 
8795 	/* walk through upper devices reassigning pools */
8796 	netdev_walk_all_upper_dev_rcu(dev, ixgbe_reassign_macvlan_pool,
8797 				      adapter);
8798 }
8799 
8800 /**
8801  * ixgbe_setup_tc - configure net_device for multiple traffic classes
8802  *
8803  * @dev: net device to configure
8804  * @tc: number of traffic classes to enable
8805  */
8806 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
8807 {
8808 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8809 	struct ixgbe_hw *hw = &adapter->hw;
8810 
8811 	/* Hardware supports up to 8 traffic classes */
8812 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
8813 		return -EINVAL;
8814 
8815 	if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
8816 		return -EINVAL;
8817 
8818 	/* Hardware has to reinitialize queues and interrupts to
8819 	 * match packet buffer alignment. Unfortunately, the
8820 	 * hardware is not flexible enough to do this dynamically.
8821 	 */
8822 	if (netif_running(dev))
8823 		ixgbe_close(dev);
8824 	else
8825 		ixgbe_reset(adapter);
8826 
8827 	ixgbe_clear_interrupt_scheme(adapter);
8828 
8829 #ifdef CONFIG_IXGBE_DCB
8830 	if (tc) {
8831 		netdev_set_num_tc(dev, tc);
8832 		ixgbe_set_prio_tc_map(adapter);
8833 
8834 		adapter->hw_tcs = tc;
8835 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
8836 
8837 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
8838 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
8839 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
8840 		}
8841 	} else {
8842 		netdev_reset_tc(dev);
8843 
8844 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
8845 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
8846 
8847 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
8848 		adapter->hw_tcs = tc;
8849 
8850 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
8851 		adapter->dcb_cfg.pfc_mode_enable = false;
8852 	}
8853 
8854 	ixgbe_validate_rtr(adapter, tc);
8855 
8856 #endif /* CONFIG_IXGBE_DCB */
8857 	ixgbe_init_interrupt_scheme(adapter);
8858 
8859 	ixgbe_defrag_macvlan_pools(dev);
8860 
8861 	if (netif_running(dev))
8862 		return ixgbe_open(dev);
8863 
8864 	return 0;
8865 }
8866 
8867 static int ixgbe_delete_clsu32(struct ixgbe_adapter *adapter,
8868 			       struct tc_cls_u32_offload *cls)
8869 {
8870 	u32 hdl = cls->knode.handle;
8871 	u32 uhtid = TC_U32_USERHTID(cls->knode.handle);
8872 	u32 loc = cls->knode.handle & 0xfffff;
8873 	int err = 0, i, j;
8874 	struct ixgbe_jump_table *jump = NULL;
8875 
8876 	if (loc > IXGBE_MAX_HW_ENTRIES)
8877 		return -EINVAL;
8878 
8879 	if ((uhtid != 0x800) && (uhtid >= IXGBE_MAX_LINK_HANDLE))
8880 		return -EINVAL;
8881 
8882 	/* Clear this filter in the link data it is associated with */
8883 	if (uhtid != 0x800) {
8884 		jump = adapter->jump_tables[uhtid];
8885 		if (!jump)
8886 			return -EINVAL;
8887 		if (!test_bit(loc - 1, jump->child_loc_map))
8888 			return -EINVAL;
8889 		clear_bit(loc - 1, jump->child_loc_map);
8890 	}
8891 
8892 	/* Check if the filter being deleted is a link */
8893 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
8894 		jump = adapter->jump_tables[i];
8895 		if (jump && jump->link_hdl == hdl) {
8896 			/* Delete filters in the hardware in the child hash
8897 			 * table associated with this link
8898 			 */
8899 			for (j = 0; j < IXGBE_MAX_HW_ENTRIES; j++) {
8900 				if (!test_bit(j, jump->child_loc_map))
8901 					continue;
8902 				spin_lock(&adapter->fdir_perfect_lock);
8903 				err = ixgbe_update_ethtool_fdir_entry(adapter,
8904 								      NULL,
8905 								      j + 1);
8906 				spin_unlock(&adapter->fdir_perfect_lock);
8907 				clear_bit(j, jump->child_loc_map);
8908 			}
8909 			/* Remove resources for this link */
8910 			kfree(jump->input);
8911 			kfree(jump->mask);
8912 			kfree(jump);
8913 			adapter->jump_tables[i] = NULL;
8914 			return err;
8915 		}
8916 	}
8917 
8918 	spin_lock(&adapter->fdir_perfect_lock);
8919 	err = ixgbe_update_ethtool_fdir_entry(adapter, NULL, loc);
8920 	spin_unlock(&adapter->fdir_perfect_lock);
8921 	return err;
8922 }
8923 
8924 static int ixgbe_configure_clsu32_add_hnode(struct ixgbe_adapter *adapter,
8925 					    struct tc_cls_u32_offload *cls)
8926 {
8927 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8928 
8929 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8930 		return -EINVAL;
8931 
8932 	/* This ixgbe devices do not support hash tables at the moment
8933 	 * so abort when given hash tables.
8934 	 */
8935 	if (cls->hnode.divisor > 0)
8936 		return -EINVAL;
8937 
8938 	set_bit(uhtid - 1, &adapter->tables);
8939 	return 0;
8940 }
8941 
8942 static int ixgbe_configure_clsu32_del_hnode(struct ixgbe_adapter *adapter,
8943 					    struct tc_cls_u32_offload *cls)
8944 {
8945 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
8946 
8947 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
8948 		return -EINVAL;
8949 
8950 	clear_bit(uhtid - 1, &adapter->tables);
8951 	return 0;
8952 }
8953 
8954 #ifdef CONFIG_NET_CLS_ACT
8955 struct upper_walk_data {
8956 	struct ixgbe_adapter *adapter;
8957 	u64 action;
8958 	int ifindex;
8959 	u8 queue;
8960 };
8961 
8962 static int get_macvlan_queue(struct net_device *upper, void *_data)
8963 {
8964 	if (netif_is_macvlan(upper)) {
8965 		struct ixgbe_fwd_adapter *vadapter = macvlan_accel_priv(upper);
8966 		struct upper_walk_data *data = _data;
8967 		struct ixgbe_adapter *adapter = data->adapter;
8968 		int ifindex = data->ifindex;
8969 
8970 		if (vadapter && upper->ifindex == ifindex) {
8971 			data->queue = adapter->rx_ring[vadapter->rx_base_queue]->reg_idx;
8972 			data->action = data->queue;
8973 			return 1;
8974 		}
8975 	}
8976 
8977 	return 0;
8978 }
8979 
8980 static int handle_redirect_action(struct ixgbe_adapter *adapter, int ifindex,
8981 				  u8 *queue, u64 *action)
8982 {
8983 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
8984 	unsigned int num_vfs = adapter->num_vfs, vf;
8985 	struct upper_walk_data data;
8986 	struct net_device *upper;
8987 
8988 	/* redirect to a SRIOV VF */
8989 	for (vf = 0; vf < num_vfs; ++vf) {
8990 		upper = pci_get_drvdata(adapter->vfinfo[vf].vfdev);
8991 		if (upper->ifindex == ifindex) {
8992 			*queue = vf * __ALIGN_MASK(1, ~vmdq->mask);
8993 			*action = vf + 1;
8994 			*action <<= ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF;
8995 			return 0;
8996 		}
8997 	}
8998 
8999 	/* redirect to a offloaded macvlan netdev */
9000 	data.adapter = adapter;
9001 	data.ifindex = ifindex;
9002 	data.action = 0;
9003 	data.queue = 0;
9004 	if (netdev_walk_all_upper_dev_rcu(adapter->netdev,
9005 					  get_macvlan_queue, &data)) {
9006 		*action = data.action;
9007 		*queue = data.queue;
9008 
9009 		return 0;
9010 	}
9011 
9012 	return -EINVAL;
9013 }
9014 
9015 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9016 			    struct tcf_exts *exts, u64 *action, u8 *queue)
9017 {
9018 	const struct tc_action *a;
9019 	LIST_HEAD(actions);
9020 
9021 	if (!tcf_exts_has_actions(exts))
9022 		return -EINVAL;
9023 
9024 	tcf_exts_to_list(exts, &actions);
9025 	list_for_each_entry(a, &actions, list) {
9026 
9027 		/* Drop action */
9028 		if (is_tcf_gact_shot(a)) {
9029 			*action = IXGBE_FDIR_DROP_QUEUE;
9030 			*queue = IXGBE_FDIR_DROP_QUEUE;
9031 			return 0;
9032 		}
9033 
9034 		/* Redirect to a VF or a offloaded macvlan */
9035 		if (is_tcf_mirred_egress_redirect(a)) {
9036 			struct net_device *dev = tcf_mirred_dev(a);
9037 
9038 			if (!dev)
9039 				return -EINVAL;
9040 			return handle_redirect_action(adapter, dev->ifindex,
9041 						      queue, action);
9042 		}
9043 
9044 		return -EINVAL;
9045 	}
9046 
9047 	return -EINVAL;
9048 }
9049 #else
9050 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9051 			    struct tcf_exts *exts, u64 *action, u8 *queue)
9052 {
9053 	return -EINVAL;
9054 }
9055 #endif /* CONFIG_NET_CLS_ACT */
9056 
9057 static int ixgbe_clsu32_build_input(struct ixgbe_fdir_filter *input,
9058 				    union ixgbe_atr_input *mask,
9059 				    struct tc_cls_u32_offload *cls,
9060 				    struct ixgbe_mat_field *field_ptr,
9061 				    struct ixgbe_nexthdr *nexthdr)
9062 {
9063 	int i, j, off;
9064 	__be32 val, m;
9065 	bool found_entry = false, found_jump_field = false;
9066 
9067 	for (i = 0; i < cls->knode.sel->nkeys; i++) {
9068 		off = cls->knode.sel->keys[i].off;
9069 		val = cls->knode.sel->keys[i].val;
9070 		m = cls->knode.sel->keys[i].mask;
9071 
9072 		for (j = 0; field_ptr[j].val; j++) {
9073 			if (field_ptr[j].off == off) {
9074 				field_ptr[j].val(input, mask, (__force u32)val,
9075 						 (__force u32)m);
9076 				input->filter.formatted.flow_type |=
9077 					field_ptr[j].type;
9078 				found_entry = true;
9079 				break;
9080 			}
9081 		}
9082 		if (nexthdr) {
9083 			if (nexthdr->off == cls->knode.sel->keys[i].off &&
9084 			    nexthdr->val ==
9085 			    (__force u32)cls->knode.sel->keys[i].val &&
9086 			    nexthdr->mask ==
9087 			    (__force u32)cls->knode.sel->keys[i].mask)
9088 				found_jump_field = true;
9089 			else
9090 				continue;
9091 		}
9092 	}
9093 
9094 	if (nexthdr && !found_jump_field)
9095 		return -EINVAL;
9096 
9097 	if (!found_entry)
9098 		return 0;
9099 
9100 	mask->formatted.flow_type = IXGBE_ATR_L4TYPE_IPV6_MASK |
9101 				    IXGBE_ATR_L4TYPE_MASK;
9102 
9103 	if (input->filter.formatted.flow_type == IXGBE_ATR_FLOW_TYPE_IPV4)
9104 		mask->formatted.flow_type &= IXGBE_ATR_L4TYPE_IPV6_MASK;
9105 
9106 	return 0;
9107 }
9108 
9109 static int ixgbe_configure_clsu32(struct ixgbe_adapter *adapter,
9110 				  struct tc_cls_u32_offload *cls)
9111 {
9112 	__be16 protocol = cls->common.protocol;
9113 	u32 loc = cls->knode.handle & 0xfffff;
9114 	struct ixgbe_hw *hw = &adapter->hw;
9115 	struct ixgbe_mat_field *field_ptr;
9116 	struct ixgbe_fdir_filter *input = NULL;
9117 	union ixgbe_atr_input *mask = NULL;
9118 	struct ixgbe_jump_table *jump = NULL;
9119 	int i, err = -EINVAL;
9120 	u8 queue;
9121 	u32 uhtid, link_uhtid;
9122 
9123 	uhtid = TC_U32_USERHTID(cls->knode.handle);
9124 	link_uhtid = TC_U32_USERHTID(cls->knode.link_handle);
9125 
9126 	/* At the moment cls_u32 jumps to network layer and skips past
9127 	 * L2 headers. The canonical method to match L2 frames is to use
9128 	 * negative values. However this is error prone at best but really
9129 	 * just broken because there is no way to "know" what sort of hdr
9130 	 * is in front of the network layer. Fix cls_u32 to support L2
9131 	 * headers when needed.
9132 	 */
9133 	if (protocol != htons(ETH_P_IP))
9134 		return err;
9135 
9136 	if (loc >= ((1024 << adapter->fdir_pballoc) - 2)) {
9137 		e_err(drv, "Location out of range\n");
9138 		return err;
9139 	}
9140 
9141 	/* cls u32 is a graph starting at root node 0x800. The driver tracks
9142 	 * links and also the fields used to advance the parser across each
9143 	 * link (e.g. nexthdr/eat parameters from 'tc'). This way we can map
9144 	 * the u32 graph onto the hardware parse graph denoted in ixgbe_model.h
9145 	 * To add support for new nodes update ixgbe_model.h parse structures
9146 	 * this function _should_ be generic try not to hardcode values here.
9147 	 */
9148 	if (uhtid == 0x800) {
9149 		field_ptr = (adapter->jump_tables[0])->mat;
9150 	} else {
9151 		if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9152 			return err;
9153 		if (!adapter->jump_tables[uhtid])
9154 			return err;
9155 		field_ptr = (adapter->jump_tables[uhtid])->mat;
9156 	}
9157 
9158 	if (!field_ptr)
9159 		return err;
9160 
9161 	/* At this point we know the field_ptr is valid and need to either
9162 	 * build cls_u32 link or attach filter. Because adding a link to
9163 	 * a handle that does not exist is invalid and the same for adding
9164 	 * rules to handles that don't exist.
9165 	 */
9166 
9167 	if (link_uhtid) {
9168 		struct ixgbe_nexthdr *nexthdr = ixgbe_ipv4_jumps;
9169 
9170 		if (link_uhtid >= IXGBE_MAX_LINK_HANDLE)
9171 			return err;
9172 
9173 		if (!test_bit(link_uhtid - 1, &adapter->tables))
9174 			return err;
9175 
9176 		/* Multiple filters as links to the same hash table are not
9177 		 * supported. To add a new filter with the same next header
9178 		 * but different match/jump conditions, create a new hash table
9179 		 * and link to it.
9180 		 */
9181 		if (adapter->jump_tables[link_uhtid] &&
9182 		    (adapter->jump_tables[link_uhtid])->link_hdl) {
9183 			e_err(drv, "Link filter exists for link: %x\n",
9184 			      link_uhtid);
9185 			return err;
9186 		}
9187 
9188 		for (i = 0; nexthdr[i].jump; i++) {
9189 			if (nexthdr[i].o != cls->knode.sel->offoff ||
9190 			    nexthdr[i].s != cls->knode.sel->offshift ||
9191 			    nexthdr[i].m !=
9192 			    (__force u32)cls->knode.sel->offmask)
9193 				return err;
9194 
9195 			jump = kzalloc(sizeof(*jump), GFP_KERNEL);
9196 			if (!jump)
9197 				return -ENOMEM;
9198 			input = kzalloc(sizeof(*input), GFP_KERNEL);
9199 			if (!input) {
9200 				err = -ENOMEM;
9201 				goto free_jump;
9202 			}
9203 			mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9204 			if (!mask) {
9205 				err = -ENOMEM;
9206 				goto free_input;
9207 			}
9208 			jump->input = input;
9209 			jump->mask = mask;
9210 			jump->link_hdl = cls->knode.handle;
9211 
9212 			err = ixgbe_clsu32_build_input(input, mask, cls,
9213 						       field_ptr, &nexthdr[i]);
9214 			if (!err) {
9215 				jump->mat = nexthdr[i].jump;
9216 				adapter->jump_tables[link_uhtid] = jump;
9217 				break;
9218 			}
9219 		}
9220 		return 0;
9221 	}
9222 
9223 	input = kzalloc(sizeof(*input), GFP_KERNEL);
9224 	if (!input)
9225 		return -ENOMEM;
9226 	mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9227 	if (!mask) {
9228 		err = -ENOMEM;
9229 		goto free_input;
9230 	}
9231 
9232 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid])) {
9233 		if ((adapter->jump_tables[uhtid])->input)
9234 			memcpy(input, (adapter->jump_tables[uhtid])->input,
9235 			       sizeof(*input));
9236 		if ((adapter->jump_tables[uhtid])->mask)
9237 			memcpy(mask, (adapter->jump_tables[uhtid])->mask,
9238 			       sizeof(*mask));
9239 
9240 		/* Lookup in all child hash tables if this location is already
9241 		 * filled with a filter
9242 		 */
9243 		for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
9244 			struct ixgbe_jump_table *link = adapter->jump_tables[i];
9245 
9246 			if (link && (test_bit(loc - 1, link->child_loc_map))) {
9247 				e_err(drv, "Filter exists in location: %x\n",
9248 				      loc);
9249 				err = -EINVAL;
9250 				goto err_out;
9251 			}
9252 		}
9253 	}
9254 	err = ixgbe_clsu32_build_input(input, mask, cls, field_ptr, NULL);
9255 	if (err)
9256 		goto err_out;
9257 
9258 	err = parse_tc_actions(adapter, cls->knode.exts, &input->action,
9259 			       &queue);
9260 	if (err < 0)
9261 		goto err_out;
9262 
9263 	input->sw_idx = loc;
9264 
9265 	spin_lock(&adapter->fdir_perfect_lock);
9266 
9267 	if (hlist_empty(&adapter->fdir_filter_list)) {
9268 		memcpy(&adapter->fdir_mask, mask, sizeof(*mask));
9269 		err = ixgbe_fdir_set_input_mask_82599(hw, mask);
9270 		if (err)
9271 			goto err_out_w_lock;
9272 	} else if (memcmp(&adapter->fdir_mask, mask, sizeof(*mask))) {
9273 		err = -EINVAL;
9274 		goto err_out_w_lock;
9275 	}
9276 
9277 	ixgbe_atr_compute_perfect_hash_82599(&input->filter, mask);
9278 	err = ixgbe_fdir_write_perfect_filter_82599(hw, &input->filter,
9279 						    input->sw_idx, queue);
9280 	if (!err)
9281 		ixgbe_update_ethtool_fdir_entry(adapter, input, input->sw_idx);
9282 	spin_unlock(&adapter->fdir_perfect_lock);
9283 
9284 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid]))
9285 		set_bit(loc - 1, (adapter->jump_tables[uhtid])->child_loc_map);
9286 
9287 	kfree(mask);
9288 	return err;
9289 err_out_w_lock:
9290 	spin_unlock(&adapter->fdir_perfect_lock);
9291 err_out:
9292 	kfree(mask);
9293 free_input:
9294 	kfree(input);
9295 free_jump:
9296 	kfree(jump);
9297 	return err;
9298 }
9299 
9300 static int ixgbe_setup_tc_cls_u32(struct ixgbe_adapter *adapter,
9301 				  struct tc_cls_u32_offload *cls_u32)
9302 {
9303 	switch (cls_u32->command) {
9304 	case TC_CLSU32_NEW_KNODE:
9305 	case TC_CLSU32_REPLACE_KNODE:
9306 		return ixgbe_configure_clsu32(adapter, cls_u32);
9307 	case TC_CLSU32_DELETE_KNODE:
9308 		return ixgbe_delete_clsu32(adapter, cls_u32);
9309 	case TC_CLSU32_NEW_HNODE:
9310 	case TC_CLSU32_REPLACE_HNODE:
9311 		return ixgbe_configure_clsu32_add_hnode(adapter, cls_u32);
9312 	case TC_CLSU32_DELETE_HNODE:
9313 		return ixgbe_configure_clsu32_del_hnode(adapter, cls_u32);
9314 	default:
9315 		return -EOPNOTSUPP;
9316 	}
9317 }
9318 
9319 static int ixgbe_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
9320 				   void *cb_priv)
9321 {
9322 	struct ixgbe_adapter *adapter = cb_priv;
9323 
9324 	if (!tc_cls_can_offload_and_chain0(adapter->netdev, type_data))
9325 		return -EOPNOTSUPP;
9326 
9327 	switch (type) {
9328 	case TC_SETUP_CLSU32:
9329 		return ixgbe_setup_tc_cls_u32(adapter, type_data);
9330 	default:
9331 		return -EOPNOTSUPP;
9332 	}
9333 }
9334 
9335 static int ixgbe_setup_tc_block(struct net_device *dev,
9336 				struct tc_block_offload *f)
9337 {
9338 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9339 
9340 	if (f->binder_type != TCF_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
9341 		return -EOPNOTSUPP;
9342 
9343 	switch (f->command) {
9344 	case TC_BLOCK_BIND:
9345 		return tcf_block_cb_register(f->block, ixgbe_setup_tc_block_cb,
9346 					     adapter, adapter, f->extack);
9347 	case TC_BLOCK_UNBIND:
9348 		tcf_block_cb_unregister(f->block, ixgbe_setup_tc_block_cb,
9349 					adapter);
9350 		return 0;
9351 	default:
9352 		return -EOPNOTSUPP;
9353 	}
9354 }
9355 
9356 static int ixgbe_setup_tc_mqprio(struct net_device *dev,
9357 				 struct tc_mqprio_qopt *mqprio)
9358 {
9359 	mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
9360 	return ixgbe_setup_tc(dev, mqprio->num_tc);
9361 }
9362 
9363 static int __ixgbe_setup_tc(struct net_device *dev, enum tc_setup_type type,
9364 			    void *type_data)
9365 {
9366 	switch (type) {
9367 	case TC_SETUP_BLOCK:
9368 		return ixgbe_setup_tc_block(dev, type_data);
9369 	case TC_SETUP_QDISC_MQPRIO:
9370 		return ixgbe_setup_tc_mqprio(dev, type_data);
9371 	default:
9372 		return -EOPNOTSUPP;
9373 	}
9374 }
9375 
9376 #ifdef CONFIG_PCI_IOV
9377 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
9378 {
9379 	struct net_device *netdev = adapter->netdev;
9380 
9381 	rtnl_lock();
9382 	ixgbe_setup_tc(netdev, adapter->hw_tcs);
9383 	rtnl_unlock();
9384 }
9385 
9386 #endif
9387 void ixgbe_do_reset(struct net_device *netdev)
9388 {
9389 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9390 
9391 	if (netif_running(netdev))
9392 		ixgbe_reinit_locked(adapter);
9393 	else
9394 		ixgbe_reset(adapter);
9395 }
9396 
9397 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
9398 					    netdev_features_t features)
9399 {
9400 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9401 
9402 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
9403 	if (!(features & NETIF_F_RXCSUM))
9404 		features &= ~NETIF_F_LRO;
9405 
9406 	/* Turn off LRO if not RSC capable */
9407 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
9408 		features &= ~NETIF_F_LRO;
9409 
9410 	return features;
9411 }
9412 
9413 static void ixgbe_reset_l2fw_offload(struct ixgbe_adapter *adapter)
9414 {
9415 	int rss = min_t(int, ixgbe_max_rss_indices(adapter),
9416 			num_online_cpus());
9417 
9418 	/* go back to full RSS if we're not running SR-IOV */
9419 	if (!adapter->ring_feature[RING_F_VMDQ].offset)
9420 		adapter->flags &= ~(IXGBE_FLAG_VMDQ_ENABLED |
9421 				    IXGBE_FLAG_SRIOV_ENABLED);
9422 
9423 	adapter->ring_feature[RING_F_RSS].limit = rss;
9424 	adapter->ring_feature[RING_F_VMDQ].limit = 1;
9425 
9426 	ixgbe_setup_tc(adapter->netdev, adapter->hw_tcs);
9427 }
9428 
9429 static int ixgbe_set_features(struct net_device *netdev,
9430 			      netdev_features_t features)
9431 {
9432 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9433 	netdev_features_t changed = netdev->features ^ features;
9434 	bool need_reset = false;
9435 
9436 	/* Make sure RSC matches LRO, reset if change */
9437 	if (!(features & NETIF_F_LRO)) {
9438 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
9439 			need_reset = true;
9440 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
9441 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
9442 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
9443 		if (adapter->rx_itr_setting == 1 ||
9444 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
9445 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
9446 			need_reset = true;
9447 		} else if ((changed ^ features) & NETIF_F_LRO) {
9448 			e_info(probe, "rx-usecs set too low, "
9449 			       "disabling RSC\n");
9450 		}
9451 	}
9452 
9453 	/*
9454 	 * Check if Flow Director n-tuple support or hw_tc support was
9455 	 * enabled or disabled.  If the state changed, we need to reset.
9456 	 */
9457 	if ((features & NETIF_F_NTUPLE) || (features & NETIF_F_HW_TC)) {
9458 		/* turn off ATR, enable perfect filters and reset */
9459 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
9460 			need_reset = true;
9461 
9462 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
9463 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9464 	} else {
9465 		/* turn off perfect filters, enable ATR and reset */
9466 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
9467 			need_reset = true;
9468 
9469 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9470 
9471 		/* We cannot enable ATR if SR-IOV is enabled */
9472 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED ||
9473 		    /* We cannot enable ATR if we have 2 or more tcs */
9474 		    (adapter->hw_tcs > 1) ||
9475 		    /* We cannot enable ATR if RSS is disabled */
9476 		    (adapter->ring_feature[RING_F_RSS].limit <= 1) ||
9477 		    /* A sample rate of 0 indicates ATR disabled */
9478 		    (!adapter->atr_sample_rate))
9479 			; /* do nothing not supported */
9480 		else /* otherwise supported and set the flag */
9481 			adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
9482 	}
9483 
9484 	if (changed & NETIF_F_RXALL)
9485 		need_reset = true;
9486 
9487 	netdev->features = features;
9488 
9489 	if ((adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE)) {
9490 		if (features & NETIF_F_RXCSUM) {
9491 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9492 		} else {
9493 			u32 port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
9494 
9495 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9496 		}
9497 	}
9498 
9499 	if ((adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)) {
9500 		if (features & NETIF_F_RXCSUM) {
9501 			adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9502 		} else {
9503 			u32 port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
9504 
9505 			ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9506 		}
9507 	}
9508 
9509 	if ((changed & NETIF_F_HW_L2FW_DOFFLOAD) && adapter->num_rx_pools > 1)
9510 		ixgbe_reset_l2fw_offload(adapter);
9511 	else if (need_reset)
9512 		ixgbe_do_reset(netdev);
9513 	else if (changed & (NETIF_F_HW_VLAN_CTAG_RX |
9514 			    NETIF_F_HW_VLAN_CTAG_FILTER))
9515 		ixgbe_set_rx_mode(netdev);
9516 
9517 	return 0;
9518 }
9519 
9520 /**
9521  * ixgbe_add_udp_tunnel_port - Get notifications about adding UDP tunnel ports
9522  * @dev: The port's netdev
9523  * @ti: Tunnel endpoint information
9524  **/
9525 static void ixgbe_add_udp_tunnel_port(struct net_device *dev,
9526 				      struct udp_tunnel_info *ti)
9527 {
9528 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9529 	struct ixgbe_hw *hw = &adapter->hw;
9530 	__be16 port = ti->port;
9531 	u32 port_shift = 0;
9532 	u32 reg;
9533 
9534 	if (ti->sa_family != AF_INET)
9535 		return;
9536 
9537 	switch (ti->type) {
9538 	case UDP_TUNNEL_TYPE_VXLAN:
9539 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
9540 			return;
9541 
9542 		if (adapter->vxlan_port == port)
9543 			return;
9544 
9545 		if (adapter->vxlan_port) {
9546 			netdev_info(dev,
9547 				    "VXLAN port %d set, not adding port %d\n",
9548 				    ntohs(adapter->vxlan_port),
9549 				    ntohs(port));
9550 			return;
9551 		}
9552 
9553 		adapter->vxlan_port = port;
9554 		break;
9555 	case UDP_TUNNEL_TYPE_GENEVE:
9556 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
9557 			return;
9558 
9559 		if (adapter->geneve_port == port)
9560 			return;
9561 
9562 		if (adapter->geneve_port) {
9563 			netdev_info(dev,
9564 				    "GENEVE port %d set, not adding port %d\n",
9565 				    ntohs(adapter->geneve_port),
9566 				    ntohs(port));
9567 			return;
9568 		}
9569 
9570 		port_shift = IXGBE_VXLANCTRL_GENEVE_UDPPORT_SHIFT;
9571 		adapter->geneve_port = port;
9572 		break;
9573 	default:
9574 		return;
9575 	}
9576 
9577 	reg = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) | ntohs(port) << port_shift;
9578 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, reg);
9579 }
9580 
9581 /**
9582  * ixgbe_del_udp_tunnel_port - Get notifications about removing UDP tunnel ports
9583  * @dev: The port's netdev
9584  * @ti: Tunnel endpoint information
9585  **/
9586 static void ixgbe_del_udp_tunnel_port(struct net_device *dev,
9587 				      struct udp_tunnel_info *ti)
9588 {
9589 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9590 	u32 port_mask;
9591 
9592 	if (ti->type != UDP_TUNNEL_TYPE_VXLAN &&
9593 	    ti->type != UDP_TUNNEL_TYPE_GENEVE)
9594 		return;
9595 
9596 	if (ti->sa_family != AF_INET)
9597 		return;
9598 
9599 	switch (ti->type) {
9600 	case UDP_TUNNEL_TYPE_VXLAN:
9601 		if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
9602 			return;
9603 
9604 		if (adapter->vxlan_port != ti->port) {
9605 			netdev_info(dev, "VXLAN port %d not found\n",
9606 				    ntohs(ti->port));
9607 			return;
9608 		}
9609 
9610 		port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
9611 		break;
9612 	case UDP_TUNNEL_TYPE_GENEVE:
9613 		if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
9614 			return;
9615 
9616 		if (adapter->geneve_port != ti->port) {
9617 			netdev_info(dev, "GENEVE port %d not found\n",
9618 				    ntohs(ti->port));
9619 			return;
9620 		}
9621 
9622 		port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
9623 		break;
9624 	default:
9625 		return;
9626 	}
9627 
9628 	ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9629 	adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9630 }
9631 
9632 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
9633 			     struct net_device *dev,
9634 			     const unsigned char *addr, u16 vid,
9635 			     u16 flags)
9636 {
9637 	/* guarantee we can provide a unique filter for the unicast address */
9638 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
9639 		struct ixgbe_adapter *adapter = netdev_priv(dev);
9640 		u16 pool = VMDQ_P(0);
9641 
9642 		if (netdev_uc_count(dev) >= ixgbe_available_rars(adapter, pool))
9643 			return -ENOMEM;
9644 	}
9645 
9646 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
9647 }
9648 
9649 /**
9650  * ixgbe_configure_bridge_mode - set various bridge modes
9651  * @adapter: the private structure
9652  * @mode: requested bridge mode
9653  *
9654  * Configure some settings require for various bridge modes.
9655  **/
9656 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
9657 				       __u16 mode)
9658 {
9659 	struct ixgbe_hw *hw = &adapter->hw;
9660 	unsigned int p, num_pools;
9661 	u32 vmdctl;
9662 
9663 	switch (mode) {
9664 	case BRIDGE_MODE_VEPA:
9665 		/* disable Tx loopback, rely on switch hairpin mode */
9666 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
9667 
9668 		/* must enable Rx switching replication to allow multicast
9669 		 * packet reception on all VFs, and to enable source address
9670 		 * pruning.
9671 		 */
9672 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9673 		vmdctl |= IXGBE_VT_CTL_REPLEN;
9674 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9675 
9676 		/* enable Rx source address pruning. Note, this requires
9677 		 * replication to be enabled or else it does nothing.
9678 		 */
9679 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9680 		for (p = 0; p < num_pools; p++) {
9681 			if (hw->mac.ops.set_source_address_pruning)
9682 				hw->mac.ops.set_source_address_pruning(hw,
9683 								       true,
9684 								       p);
9685 		}
9686 		break;
9687 	case BRIDGE_MODE_VEB:
9688 		/* enable Tx loopback for internal VF/PF communication */
9689 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
9690 				IXGBE_PFDTXGSWC_VT_LBEN);
9691 
9692 		/* disable Rx switching replication unless we have SR-IOV
9693 		 * virtual functions
9694 		 */
9695 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9696 		if (!adapter->num_vfs)
9697 			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
9698 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9699 
9700 		/* disable Rx source address pruning, since we don't expect to
9701 		 * be receiving external loopback of our transmitted frames.
9702 		 */
9703 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9704 		for (p = 0; p < num_pools; p++) {
9705 			if (hw->mac.ops.set_source_address_pruning)
9706 				hw->mac.ops.set_source_address_pruning(hw,
9707 								       false,
9708 								       p);
9709 		}
9710 		break;
9711 	default:
9712 		return -EINVAL;
9713 	}
9714 
9715 	adapter->bridge_mode = mode;
9716 
9717 	e_info(drv, "enabling bridge mode: %s\n",
9718 	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
9719 
9720 	return 0;
9721 }
9722 
9723 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
9724 				    struct nlmsghdr *nlh, u16 flags)
9725 {
9726 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9727 	struct nlattr *attr, *br_spec;
9728 	int rem;
9729 
9730 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9731 		return -EOPNOTSUPP;
9732 
9733 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
9734 	if (!br_spec)
9735 		return -EINVAL;
9736 
9737 	nla_for_each_nested(attr, br_spec, rem) {
9738 		int status;
9739 		__u16 mode;
9740 
9741 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
9742 			continue;
9743 
9744 		if (nla_len(attr) < sizeof(mode))
9745 			return -EINVAL;
9746 
9747 		mode = nla_get_u16(attr);
9748 		status = ixgbe_configure_bridge_mode(adapter, mode);
9749 		if (status)
9750 			return status;
9751 
9752 		break;
9753 	}
9754 
9755 	return 0;
9756 }
9757 
9758 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9759 				    struct net_device *dev,
9760 				    u32 filter_mask, int nlflags)
9761 {
9762 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9763 
9764 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9765 		return 0;
9766 
9767 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
9768 				       adapter->bridge_mode, 0, 0, nlflags,
9769 				       filter_mask, NULL);
9770 }
9771 
9772 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
9773 {
9774 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
9775 	struct ixgbe_fwd_adapter *accel;
9776 	int tcs = adapter->hw_tcs ? : 1;
9777 	int pool, err;
9778 
9779 	/* The hardware supported by ixgbe only filters on the destination MAC
9780 	 * address. In order to avoid issues we only support offloading modes
9781 	 * where the hardware can actually provide the functionality.
9782 	 */
9783 	if (!macvlan_supports_dest_filter(vdev))
9784 		return ERR_PTR(-EMEDIUMTYPE);
9785 
9786 	/* We need to lock down the macvlan to be a single queue device so that
9787 	 * we can reuse the tc_to_txq field in the macvlan netdev to represent
9788 	 * the queue mapping to our netdev.
9789 	 */
9790 	if (netif_is_multiqueue(vdev))
9791 		return ERR_PTR(-ERANGE);
9792 
9793 	pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
9794 	if (pool == adapter->num_rx_pools) {
9795 		u16 used_pools = adapter->num_vfs + adapter->num_rx_pools;
9796 		u16 reserved_pools;
9797 
9798 		if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
9799 		     adapter->num_rx_pools >= (MAX_TX_QUEUES / tcs)) ||
9800 		    adapter->num_rx_pools > IXGBE_MAX_MACVLANS)
9801 			return ERR_PTR(-EBUSY);
9802 
9803 		/* Hardware has a limited number of available pools. Each VF,
9804 		 * and the PF require a pool. Check to ensure we don't
9805 		 * attempt to use more then the available number of pools.
9806 		 */
9807 		if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
9808 			return ERR_PTR(-EBUSY);
9809 
9810 		/* Enable VMDq flag so device will be set in VM mode */
9811 		adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED |
9812 				  IXGBE_FLAG_SRIOV_ENABLED;
9813 
9814 		/* Try to reserve as many queues per pool as possible,
9815 		 * we start with the configurations that support 4 queues
9816 		 * per pools, followed by 2, and then by just 1 per pool.
9817 		 */
9818 		if (used_pools < 32 && adapter->num_rx_pools < 16)
9819 			reserved_pools = min_t(u16,
9820 					       32 - used_pools,
9821 					       16 - adapter->num_rx_pools);
9822 		else if (adapter->num_rx_pools < 32)
9823 			reserved_pools = min_t(u16,
9824 					       64 - used_pools,
9825 					       32 - adapter->num_rx_pools);
9826 		else
9827 			reserved_pools = 64 - used_pools;
9828 
9829 
9830 		if (!reserved_pools)
9831 			return ERR_PTR(-EBUSY);
9832 
9833 		adapter->ring_feature[RING_F_VMDQ].limit += reserved_pools;
9834 
9835 		/* Force reinit of ring allocation with VMDQ enabled */
9836 		err = ixgbe_setup_tc(pdev, adapter->hw_tcs);
9837 		if (err)
9838 			return ERR_PTR(err);
9839 
9840 		if (pool >= adapter->num_rx_pools)
9841 			return ERR_PTR(-ENOMEM);
9842 	}
9843 
9844 	accel = kzalloc(sizeof(*accel), GFP_KERNEL);
9845 	if (!accel)
9846 		return ERR_PTR(-ENOMEM);
9847 
9848 	set_bit(pool, adapter->fwd_bitmask);
9849 	netdev_set_sb_channel(vdev, pool);
9850 	accel->pool = pool;
9851 	accel->netdev = vdev;
9852 
9853 	if (!netif_running(pdev))
9854 		return accel;
9855 
9856 	err = ixgbe_fwd_ring_up(adapter, accel);
9857 	if (err)
9858 		return ERR_PTR(err);
9859 
9860 	return accel;
9861 }
9862 
9863 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
9864 {
9865 	struct ixgbe_fwd_adapter *accel = priv;
9866 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
9867 	unsigned int rxbase = accel->rx_base_queue;
9868 	unsigned int i;
9869 
9870 	/* delete unicast filter associated with offloaded interface */
9871 	ixgbe_del_mac_filter(adapter, accel->netdev->dev_addr,
9872 			     VMDQ_P(accel->pool));
9873 
9874 	/* Allow remaining Rx packets to get flushed out of the
9875 	 * Rx FIFO before we drop the netdev for the ring.
9876 	 */
9877 	usleep_range(10000, 20000);
9878 
9879 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
9880 		struct ixgbe_ring *ring = adapter->rx_ring[rxbase + i];
9881 		struct ixgbe_q_vector *qv = ring->q_vector;
9882 
9883 		/* Make sure we aren't processing any packets and clear
9884 		 * netdev to shut down the ring.
9885 		 */
9886 		if (netif_running(adapter->netdev))
9887 			napi_synchronize(&qv->napi);
9888 		ring->netdev = NULL;
9889 	}
9890 
9891 	/* unbind the queues and drop the subordinate channel config */
9892 	netdev_unbind_sb_channel(pdev, accel->netdev);
9893 	netdev_set_sb_channel(accel->netdev, 0);
9894 
9895 	clear_bit(accel->pool, adapter->fwd_bitmask);
9896 	kfree(accel);
9897 }
9898 
9899 #define IXGBE_MAX_MAC_HDR_LEN		127
9900 #define IXGBE_MAX_NETWORK_HDR_LEN	511
9901 
9902 static netdev_features_t
9903 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
9904 		     netdev_features_t features)
9905 {
9906 	unsigned int network_hdr_len, mac_hdr_len;
9907 
9908 	/* Make certain the headers can be described by a context descriptor */
9909 	mac_hdr_len = skb_network_header(skb) - skb->data;
9910 	if (unlikely(mac_hdr_len > IXGBE_MAX_MAC_HDR_LEN))
9911 		return features & ~(NETIF_F_HW_CSUM |
9912 				    NETIF_F_SCTP_CRC |
9913 				    NETIF_F_HW_VLAN_CTAG_TX |
9914 				    NETIF_F_TSO |
9915 				    NETIF_F_TSO6);
9916 
9917 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
9918 	if (unlikely(network_hdr_len >  IXGBE_MAX_NETWORK_HDR_LEN))
9919 		return features & ~(NETIF_F_HW_CSUM |
9920 				    NETIF_F_SCTP_CRC |
9921 				    NETIF_F_TSO |
9922 				    NETIF_F_TSO6);
9923 
9924 	/* We can only support IPV4 TSO in tunnels if we can mangle the
9925 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
9926 	 * IPsec offoad sets skb->encapsulation but still can handle
9927 	 * the TSO, so it's the exception.
9928 	 */
9929 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID)) {
9930 #ifdef CONFIG_XFRM_OFFLOAD
9931 		if (!skb->sp)
9932 #endif
9933 			features &= ~NETIF_F_TSO;
9934 	}
9935 
9936 	return features;
9937 }
9938 
9939 static int ixgbe_xdp_setup(struct net_device *dev, struct bpf_prog *prog)
9940 {
9941 	int i, frame_size = dev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
9942 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9943 	struct bpf_prog *old_prog;
9944 
9945 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
9946 		return -EINVAL;
9947 
9948 	if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
9949 		return -EINVAL;
9950 
9951 	/* verify ixgbe ring attributes are sufficient for XDP */
9952 	for (i = 0; i < adapter->num_rx_queues; i++) {
9953 		struct ixgbe_ring *ring = adapter->rx_ring[i];
9954 
9955 		if (ring_is_rsc_enabled(ring))
9956 			return -EINVAL;
9957 
9958 		if (frame_size > ixgbe_rx_bufsz(ring))
9959 			return -EINVAL;
9960 	}
9961 
9962 	if (nr_cpu_ids > MAX_XDP_QUEUES)
9963 		return -ENOMEM;
9964 
9965 	old_prog = xchg(&adapter->xdp_prog, prog);
9966 
9967 	/* If transitioning XDP modes reconfigure rings */
9968 	if (!!prog != !!old_prog) {
9969 		int err = ixgbe_setup_tc(dev, adapter->hw_tcs);
9970 
9971 		if (err) {
9972 			rcu_assign_pointer(adapter->xdp_prog, old_prog);
9973 			return -EINVAL;
9974 		}
9975 	} else {
9976 		for (i = 0; i < adapter->num_rx_queues; i++)
9977 			(void)xchg(&adapter->rx_ring[i]->xdp_prog,
9978 			    adapter->xdp_prog);
9979 	}
9980 
9981 	if (old_prog)
9982 		bpf_prog_put(old_prog);
9983 
9984 	return 0;
9985 }
9986 
9987 static int ixgbe_xdp(struct net_device *dev, struct netdev_bpf *xdp)
9988 {
9989 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9990 
9991 	switch (xdp->command) {
9992 	case XDP_SETUP_PROG:
9993 		return ixgbe_xdp_setup(dev, xdp->prog);
9994 	case XDP_QUERY_PROG:
9995 		xdp->prog_id = adapter->xdp_prog ?
9996 			adapter->xdp_prog->aux->id : 0;
9997 		return 0;
9998 	default:
9999 		return -EINVAL;
10000 	}
10001 }
10002 
10003 static void ixgbe_xdp_ring_update_tail(struct ixgbe_ring *ring)
10004 {
10005 	/* Force memory writes to complete before letting h/w know there
10006 	 * are new descriptors to fetch.
10007 	 */
10008 	wmb();
10009 	writel(ring->next_to_use, ring->tail);
10010 }
10011 
10012 static int ixgbe_xdp_xmit(struct net_device *dev, int n,
10013 			  struct xdp_frame **frames, u32 flags)
10014 {
10015 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10016 	struct ixgbe_ring *ring;
10017 	int drops = 0;
10018 	int i;
10019 
10020 	if (unlikely(test_bit(__IXGBE_DOWN, &adapter->state)))
10021 		return -ENETDOWN;
10022 
10023 	if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
10024 		return -EINVAL;
10025 
10026 	/* During program transitions its possible adapter->xdp_prog is assigned
10027 	 * but ring has not been configured yet. In this case simply abort xmit.
10028 	 */
10029 	ring = adapter->xdp_prog ? adapter->xdp_ring[smp_processor_id()] : NULL;
10030 	if (unlikely(!ring))
10031 		return -ENXIO;
10032 
10033 	for (i = 0; i < n; i++) {
10034 		struct xdp_frame *xdpf = frames[i];
10035 		int err;
10036 
10037 		err = ixgbe_xmit_xdp_ring(adapter, xdpf);
10038 		if (err != IXGBE_XDP_TX) {
10039 			xdp_return_frame_rx_napi(xdpf);
10040 			drops++;
10041 		}
10042 	}
10043 
10044 	if (unlikely(flags & XDP_XMIT_FLUSH))
10045 		ixgbe_xdp_ring_update_tail(ring);
10046 
10047 	return n - drops;
10048 }
10049 
10050 static const struct net_device_ops ixgbe_netdev_ops = {
10051 	.ndo_open		= ixgbe_open,
10052 	.ndo_stop		= ixgbe_close,
10053 	.ndo_start_xmit		= ixgbe_xmit_frame,
10054 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
10055 	.ndo_validate_addr	= eth_validate_addr,
10056 	.ndo_set_mac_address	= ixgbe_set_mac,
10057 	.ndo_change_mtu		= ixgbe_change_mtu,
10058 	.ndo_tx_timeout		= ixgbe_tx_timeout,
10059 	.ndo_set_tx_maxrate	= ixgbe_tx_maxrate,
10060 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
10061 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
10062 	.ndo_do_ioctl		= ixgbe_ioctl,
10063 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
10064 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
10065 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
10066 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
10067 	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
10068 	.ndo_set_vf_trust	= ixgbe_ndo_set_vf_trust,
10069 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
10070 	.ndo_get_stats64	= ixgbe_get_stats64,
10071 	.ndo_setup_tc		= __ixgbe_setup_tc,
10072 #ifdef CONFIG_NET_POLL_CONTROLLER
10073 	.ndo_poll_controller	= ixgbe_netpoll,
10074 #endif
10075 #ifdef IXGBE_FCOE
10076 	.ndo_select_queue	= ixgbe_select_queue,
10077 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
10078 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
10079 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
10080 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
10081 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
10082 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
10083 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
10084 #endif /* IXGBE_FCOE */
10085 	.ndo_set_features = ixgbe_set_features,
10086 	.ndo_fix_features = ixgbe_fix_features,
10087 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
10088 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
10089 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
10090 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
10091 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
10092 	.ndo_udp_tunnel_add	= ixgbe_add_udp_tunnel_port,
10093 	.ndo_udp_tunnel_del	= ixgbe_del_udp_tunnel_port,
10094 	.ndo_features_check	= ixgbe_features_check,
10095 	.ndo_bpf		= ixgbe_xdp,
10096 	.ndo_xdp_xmit		= ixgbe_xdp_xmit,
10097 };
10098 
10099 /**
10100  * ixgbe_enumerate_functions - Get the number of ports this device has
10101  * @adapter: adapter structure
10102  *
10103  * This function enumerates the phsyical functions co-located on a single slot,
10104  * in order to determine how many ports a device has. This is most useful in
10105  * determining the required GT/s of PCIe bandwidth necessary for optimal
10106  * performance.
10107  **/
10108 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
10109 {
10110 	struct pci_dev *entry, *pdev = adapter->pdev;
10111 	int physfns = 0;
10112 
10113 	/* Some cards can not use the generic count PCIe functions method,
10114 	 * because they are behind a parent switch, so we hardcode these with
10115 	 * the correct number of functions.
10116 	 */
10117 	if (ixgbe_pcie_from_parent(&adapter->hw))
10118 		physfns = 4;
10119 
10120 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
10121 		/* don't count virtual functions */
10122 		if (entry->is_virtfn)
10123 			continue;
10124 
10125 		/* When the devices on the bus don't all match our device ID,
10126 		 * we can't reliably determine the correct number of
10127 		 * functions. This can occur if a function has been direct
10128 		 * attached to a virtual machine using VT-d, for example. In
10129 		 * this case, simply return -1 to indicate this.
10130 		 */
10131 		if ((entry->vendor != pdev->vendor) ||
10132 		    (entry->device != pdev->device))
10133 			return -1;
10134 
10135 		physfns++;
10136 	}
10137 
10138 	return physfns;
10139 }
10140 
10141 /**
10142  * ixgbe_wol_supported - Check whether device supports WoL
10143  * @adapter: the adapter private structure
10144  * @device_id: the device ID
10145  * @subdevice_id: the subsystem device ID
10146  *
10147  * This function is used by probe and ethtool to determine
10148  * which devices have WoL support
10149  *
10150  **/
10151 bool ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
10152 			 u16 subdevice_id)
10153 {
10154 	struct ixgbe_hw *hw = &adapter->hw;
10155 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
10156 
10157 	/* WOL not supported on 82598 */
10158 	if (hw->mac.type == ixgbe_mac_82598EB)
10159 		return false;
10160 
10161 	/* check eeprom to see if WOL is enabled for X540 and newer */
10162 	if (hw->mac.type >= ixgbe_mac_X540) {
10163 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
10164 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
10165 		     (hw->bus.func == 0)))
10166 			return true;
10167 	}
10168 
10169 	/* WOL is determined based on device IDs for 82599 MACs */
10170 	switch (device_id) {
10171 	case IXGBE_DEV_ID_82599_SFP:
10172 		/* Only these subdevices could supports WOL */
10173 		switch (subdevice_id) {
10174 		case IXGBE_SUBDEV_ID_82599_560FLR:
10175 		case IXGBE_SUBDEV_ID_82599_LOM_SNAP6:
10176 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
10177 		case IXGBE_SUBDEV_ID_82599_SFP_2OCP:
10178 			/* only support first port */
10179 			if (hw->bus.func != 0)
10180 				break;
10181 			/* fall through */
10182 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
10183 		case IXGBE_SUBDEV_ID_82599_SFP:
10184 		case IXGBE_SUBDEV_ID_82599_RNDC:
10185 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
10186 		case IXGBE_SUBDEV_ID_82599_SFP_1OCP:
10187 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM1:
10188 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM2:
10189 			return true;
10190 		}
10191 		break;
10192 	case IXGBE_DEV_ID_82599EN_SFP:
10193 		/* Only these subdevices support WOL */
10194 		switch (subdevice_id) {
10195 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
10196 			return true;
10197 		}
10198 		break;
10199 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
10200 		/* All except this subdevice support WOL */
10201 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
10202 			return true;
10203 		break;
10204 	case IXGBE_DEV_ID_82599_KX4:
10205 		return  true;
10206 	default:
10207 		break;
10208 	}
10209 
10210 	return false;
10211 }
10212 
10213 /**
10214  * ixgbe_set_fw_version - Set FW version
10215  * @adapter: the adapter private structure
10216  *
10217  * This function is used by probe and ethtool to determine the FW version to
10218  * format to display. The FW version is taken from the EEPROM/NVM.
10219  */
10220 static void ixgbe_set_fw_version(struct ixgbe_adapter *adapter)
10221 {
10222 	struct ixgbe_hw *hw = &adapter->hw;
10223 	struct ixgbe_nvm_version nvm_ver;
10224 
10225 	ixgbe_get_oem_prod_version(hw, &nvm_ver);
10226 	if (nvm_ver.oem_valid) {
10227 		snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10228 			 "%x.%x.%x", nvm_ver.oem_major, nvm_ver.oem_minor,
10229 			 nvm_ver.oem_release);
10230 		return;
10231 	}
10232 
10233 	ixgbe_get_etk_id(hw, &nvm_ver);
10234 	ixgbe_get_orom_version(hw, &nvm_ver);
10235 
10236 	if (nvm_ver.or_valid) {
10237 		snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10238 			 "0x%08x, %d.%d.%d", nvm_ver.etk_id, nvm_ver.or_major,
10239 			 nvm_ver.or_build, nvm_ver.or_patch);
10240 		return;
10241 	}
10242 
10243 	/* Set ETrack ID format */
10244 	snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10245 		 "0x%08x", nvm_ver.etk_id);
10246 }
10247 
10248 /**
10249  * ixgbe_probe - Device Initialization Routine
10250  * @pdev: PCI device information struct
10251  * @ent: entry in ixgbe_pci_tbl
10252  *
10253  * Returns 0 on success, negative on failure
10254  *
10255  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
10256  * The OS initialization, configuring of the adapter private structure,
10257  * and a hardware reset occur.
10258  **/
10259 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10260 {
10261 	struct net_device *netdev;
10262 	struct ixgbe_adapter *adapter = NULL;
10263 	struct ixgbe_hw *hw;
10264 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
10265 	int i, err, pci_using_dac, expected_gts;
10266 	unsigned int indices = MAX_TX_QUEUES;
10267 	u8 part_str[IXGBE_PBANUM_LENGTH];
10268 	bool disable_dev = false;
10269 #ifdef IXGBE_FCOE
10270 	u16 device_caps;
10271 #endif
10272 	u32 eec;
10273 
10274 	/* Catch broken hardware that put the wrong VF device ID in
10275 	 * the PCIe SR-IOV capability.
10276 	 */
10277 	if (pdev->is_virtfn) {
10278 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
10279 		     pci_name(pdev), pdev->vendor, pdev->device);
10280 		return -EINVAL;
10281 	}
10282 
10283 	err = pci_enable_device_mem(pdev);
10284 	if (err)
10285 		return err;
10286 
10287 	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
10288 		pci_using_dac = 1;
10289 	} else {
10290 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10291 		if (err) {
10292 			dev_err(&pdev->dev,
10293 				"No usable DMA configuration, aborting\n");
10294 			goto err_dma;
10295 		}
10296 		pci_using_dac = 0;
10297 	}
10298 
10299 	err = pci_request_mem_regions(pdev, ixgbe_driver_name);
10300 	if (err) {
10301 		dev_err(&pdev->dev,
10302 			"pci_request_selected_regions failed 0x%x\n", err);
10303 		goto err_pci_reg;
10304 	}
10305 
10306 	pci_enable_pcie_error_reporting(pdev);
10307 
10308 	pci_set_master(pdev);
10309 	pci_save_state(pdev);
10310 
10311 	if (ii->mac == ixgbe_mac_82598EB) {
10312 #ifdef CONFIG_IXGBE_DCB
10313 		/* 8 TC w/ 4 queues per TC */
10314 		indices = 4 * MAX_TRAFFIC_CLASS;
10315 #else
10316 		indices = IXGBE_MAX_RSS_INDICES;
10317 #endif
10318 	}
10319 
10320 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
10321 	if (!netdev) {
10322 		err = -ENOMEM;
10323 		goto err_alloc_etherdev;
10324 	}
10325 
10326 	SET_NETDEV_DEV(netdev, &pdev->dev);
10327 
10328 	adapter = netdev_priv(netdev);
10329 
10330 	adapter->netdev = netdev;
10331 	adapter->pdev = pdev;
10332 	hw = &adapter->hw;
10333 	hw->back = adapter;
10334 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
10335 
10336 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
10337 			      pci_resource_len(pdev, 0));
10338 	adapter->io_addr = hw->hw_addr;
10339 	if (!hw->hw_addr) {
10340 		err = -EIO;
10341 		goto err_ioremap;
10342 	}
10343 
10344 	netdev->netdev_ops = &ixgbe_netdev_ops;
10345 	ixgbe_set_ethtool_ops(netdev);
10346 	netdev->watchdog_timeo = 5 * HZ;
10347 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
10348 
10349 	/* Setup hw api */
10350 	hw->mac.ops   = *ii->mac_ops;
10351 	hw->mac.type  = ii->mac;
10352 	hw->mvals     = ii->mvals;
10353 	if (ii->link_ops)
10354 		hw->link.ops  = *ii->link_ops;
10355 
10356 	/* EEPROM */
10357 	hw->eeprom.ops = *ii->eeprom_ops;
10358 	eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
10359 	if (ixgbe_removed(hw->hw_addr)) {
10360 		err = -EIO;
10361 		goto err_ioremap;
10362 	}
10363 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
10364 	if (!(eec & BIT(8)))
10365 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
10366 
10367 	/* PHY */
10368 	hw->phy.ops = *ii->phy_ops;
10369 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
10370 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
10371 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
10372 	hw->phy.mdio.mmds = 0;
10373 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
10374 	hw->phy.mdio.dev = netdev;
10375 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
10376 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
10377 
10378 	/* setup the private structure */
10379 	err = ixgbe_sw_init(adapter, ii);
10380 	if (err)
10381 		goto err_sw_init;
10382 
10383 	/* Make sure the SWFW semaphore is in a valid state */
10384 	if (hw->mac.ops.init_swfw_sync)
10385 		hw->mac.ops.init_swfw_sync(hw);
10386 
10387 	/* Make it possible the adapter to be woken up via WOL */
10388 	switch (adapter->hw.mac.type) {
10389 	case ixgbe_mac_82599EB:
10390 	case ixgbe_mac_X540:
10391 	case ixgbe_mac_X550:
10392 	case ixgbe_mac_X550EM_x:
10393 	case ixgbe_mac_x550em_a:
10394 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10395 		break;
10396 	default:
10397 		break;
10398 	}
10399 
10400 	/*
10401 	 * If there is a fan on this device and it has failed log the
10402 	 * failure.
10403 	 */
10404 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
10405 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
10406 		if (esdp & IXGBE_ESDP_SDP1)
10407 			e_crit(probe, "Fan has stopped, replace the adapter\n");
10408 	}
10409 
10410 	if (allow_unsupported_sfp)
10411 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
10412 
10413 	/* reset_hw fills in the perm_addr as well */
10414 	hw->phy.reset_if_overtemp = true;
10415 	err = hw->mac.ops.reset_hw(hw);
10416 	hw->phy.reset_if_overtemp = false;
10417 	ixgbe_set_eee_capable(adapter);
10418 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
10419 		err = 0;
10420 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
10421 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
10422 		e_dev_err("Reload the driver after installing a supported module.\n");
10423 		goto err_sw_init;
10424 	} else if (err) {
10425 		e_dev_err("HW Init failed: %d\n", err);
10426 		goto err_sw_init;
10427 	}
10428 
10429 #ifdef CONFIG_PCI_IOV
10430 	/* SR-IOV not supported on the 82598 */
10431 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
10432 		goto skip_sriov;
10433 	/* Mailbox */
10434 	ixgbe_init_mbx_params_pf(hw);
10435 	hw->mbx.ops = ii->mbx_ops;
10436 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
10437 	ixgbe_enable_sriov(adapter, max_vfs);
10438 skip_sriov:
10439 
10440 #endif
10441 	netdev->features = NETIF_F_SG |
10442 			   NETIF_F_TSO |
10443 			   NETIF_F_TSO6 |
10444 			   NETIF_F_RXHASH |
10445 			   NETIF_F_RXCSUM |
10446 			   NETIF_F_HW_CSUM;
10447 
10448 #define IXGBE_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
10449 				    NETIF_F_GSO_GRE_CSUM | \
10450 				    NETIF_F_GSO_IPXIP4 | \
10451 				    NETIF_F_GSO_IPXIP6 | \
10452 				    NETIF_F_GSO_UDP_TUNNEL | \
10453 				    NETIF_F_GSO_UDP_TUNNEL_CSUM)
10454 
10455 	netdev->gso_partial_features = IXGBE_GSO_PARTIAL_FEATURES;
10456 	netdev->features |= NETIF_F_GSO_PARTIAL |
10457 			    IXGBE_GSO_PARTIAL_FEATURES;
10458 
10459 	if (hw->mac.type >= ixgbe_mac_82599EB)
10460 		netdev->features |= NETIF_F_SCTP_CRC;
10461 
10462 #ifdef CONFIG_XFRM_OFFLOAD
10463 #define IXGBE_ESP_FEATURES	(NETIF_F_HW_ESP | \
10464 				 NETIF_F_HW_ESP_TX_CSUM | \
10465 				 NETIF_F_GSO_ESP)
10466 
10467 	if (adapter->ipsec)
10468 		netdev->features |= IXGBE_ESP_FEATURES;
10469 #endif
10470 	/* copy netdev features into list of user selectable features */
10471 	netdev->hw_features |= netdev->features |
10472 			       NETIF_F_HW_VLAN_CTAG_FILTER |
10473 			       NETIF_F_HW_VLAN_CTAG_RX |
10474 			       NETIF_F_HW_VLAN_CTAG_TX |
10475 			       NETIF_F_RXALL |
10476 			       NETIF_F_HW_L2FW_DOFFLOAD;
10477 
10478 	if (hw->mac.type >= ixgbe_mac_82599EB)
10479 		netdev->hw_features |= NETIF_F_NTUPLE |
10480 				       NETIF_F_HW_TC;
10481 
10482 	if (pci_using_dac)
10483 		netdev->features |= NETIF_F_HIGHDMA;
10484 
10485 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
10486 	netdev->hw_enc_features |= netdev->vlan_features;
10487 	netdev->mpls_features |= NETIF_F_SG |
10488 				 NETIF_F_TSO |
10489 				 NETIF_F_TSO6 |
10490 				 NETIF_F_HW_CSUM;
10491 	netdev->mpls_features |= IXGBE_GSO_PARTIAL_FEATURES;
10492 
10493 	/* set this bit last since it cannot be part of vlan_features */
10494 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
10495 			    NETIF_F_HW_VLAN_CTAG_RX |
10496 			    NETIF_F_HW_VLAN_CTAG_TX;
10497 
10498 	netdev->priv_flags |= IFF_UNICAST_FLT;
10499 	netdev->priv_flags |= IFF_SUPP_NOFCS;
10500 
10501 	/* MTU range: 68 - 9710 */
10502 	netdev->min_mtu = ETH_MIN_MTU;
10503 	netdev->max_mtu = IXGBE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN);
10504 
10505 #ifdef CONFIG_IXGBE_DCB
10506 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
10507 		netdev->dcbnl_ops = &ixgbe_dcbnl_ops;
10508 #endif
10509 
10510 #ifdef IXGBE_FCOE
10511 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
10512 		unsigned int fcoe_l;
10513 
10514 		if (hw->mac.ops.get_device_caps) {
10515 			hw->mac.ops.get_device_caps(hw, &device_caps);
10516 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
10517 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
10518 		}
10519 
10520 
10521 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
10522 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
10523 
10524 		netdev->features |= NETIF_F_FSO |
10525 				    NETIF_F_FCOE_CRC;
10526 
10527 		netdev->vlan_features |= NETIF_F_FSO |
10528 					 NETIF_F_FCOE_CRC |
10529 					 NETIF_F_FCOE_MTU;
10530 	}
10531 #endif /* IXGBE_FCOE */
10532 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
10533 		netdev->hw_features |= NETIF_F_LRO;
10534 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
10535 		netdev->features |= NETIF_F_LRO;
10536 
10537 	/* make sure the EEPROM is good */
10538 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
10539 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
10540 		err = -EIO;
10541 		goto err_sw_init;
10542 	}
10543 
10544 	eth_platform_get_mac_address(&adapter->pdev->dev,
10545 				     adapter->hw.mac.perm_addr);
10546 
10547 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
10548 
10549 	if (!is_valid_ether_addr(netdev->dev_addr)) {
10550 		e_dev_err("invalid MAC address\n");
10551 		err = -EIO;
10552 		goto err_sw_init;
10553 	}
10554 
10555 	/* Set hw->mac.addr to permanent MAC address */
10556 	ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
10557 	ixgbe_mac_set_default_filter(adapter);
10558 
10559 	timer_setup(&adapter->service_timer, ixgbe_service_timer, 0);
10560 
10561 	if (ixgbe_removed(hw->hw_addr)) {
10562 		err = -EIO;
10563 		goto err_sw_init;
10564 	}
10565 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
10566 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
10567 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
10568 
10569 	err = ixgbe_init_interrupt_scheme(adapter);
10570 	if (err)
10571 		goto err_sw_init;
10572 
10573 	for (i = 0; i < adapter->num_rx_queues; i++)
10574 		u64_stats_init(&adapter->rx_ring[i]->syncp);
10575 	for (i = 0; i < adapter->num_tx_queues; i++)
10576 		u64_stats_init(&adapter->tx_ring[i]->syncp);
10577 	for (i = 0; i < adapter->num_xdp_queues; i++)
10578 		u64_stats_init(&adapter->xdp_ring[i]->syncp);
10579 
10580 	/* WOL not supported for all devices */
10581 	adapter->wol = 0;
10582 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
10583 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
10584 						pdev->subsystem_device);
10585 	if (hw->wol_enabled)
10586 		adapter->wol = IXGBE_WUFC_MAG;
10587 
10588 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
10589 
10590 	/* save off EEPROM version number */
10591 	ixgbe_set_fw_version(adapter);
10592 
10593 	/* pick up the PCI bus settings for reporting later */
10594 	if (ixgbe_pcie_from_parent(hw))
10595 		ixgbe_get_parent_bus_info(adapter);
10596 	else
10597 		 hw->mac.ops.get_bus_info(hw);
10598 
10599 	/* calculate the expected PCIe bandwidth required for optimal
10600 	 * performance. Note that some older parts will never have enough
10601 	 * bandwidth due to being older generation PCIe parts. We clamp these
10602 	 * parts to ensure no warning is displayed if it can't be fixed.
10603 	 */
10604 	switch (hw->mac.type) {
10605 	case ixgbe_mac_82598EB:
10606 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
10607 		break;
10608 	default:
10609 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
10610 		break;
10611 	}
10612 
10613 	/* don't check link if we failed to enumerate functions */
10614 	if (expected_gts > 0)
10615 		ixgbe_check_minimum_link(adapter, expected_gts);
10616 
10617 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
10618 	if (err)
10619 		strlcpy(part_str, "Unknown", sizeof(part_str));
10620 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
10621 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
10622 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
10623 			   part_str);
10624 	else
10625 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
10626 			   hw->mac.type, hw->phy.type, part_str);
10627 
10628 	e_dev_info("%pM\n", netdev->dev_addr);
10629 
10630 	/* reset the hardware with the new settings */
10631 	err = hw->mac.ops.start_hw(hw);
10632 	if (err == IXGBE_ERR_EEPROM_VERSION) {
10633 		/* We are running on a pre-production device, log a warning */
10634 		e_dev_warn("This device is a pre-production adapter/LOM. "
10635 			   "Please be aware there may be issues associated "
10636 			   "with your hardware.  If you are experiencing "
10637 			   "problems please contact your Intel or hardware "
10638 			   "representative who provided you with this "
10639 			   "hardware.\n");
10640 	}
10641 	strcpy(netdev->name, "eth%d");
10642 	pci_set_drvdata(pdev, adapter);
10643 	err = register_netdev(netdev);
10644 	if (err)
10645 		goto err_register;
10646 
10647 
10648 	/* power down the optics for 82599 SFP+ fiber */
10649 	if (hw->mac.ops.disable_tx_laser)
10650 		hw->mac.ops.disable_tx_laser(hw);
10651 
10652 	/* carrier off reporting is important to ethtool even BEFORE open */
10653 	netif_carrier_off(netdev);
10654 
10655 #ifdef CONFIG_IXGBE_DCA
10656 	if (dca_add_requester(&pdev->dev) == 0) {
10657 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
10658 		ixgbe_setup_dca(adapter);
10659 	}
10660 #endif
10661 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
10662 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
10663 		for (i = 0; i < adapter->num_vfs; i++)
10664 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
10665 	}
10666 
10667 	/* firmware requires driver version to be 0xFFFFFFFF
10668 	 * since os does not support feature
10669 	 */
10670 	if (hw->mac.ops.set_fw_drv_ver)
10671 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF, 0xFF,
10672 					   sizeof(ixgbe_driver_version) - 1,
10673 					   ixgbe_driver_version);
10674 
10675 	/* add san mac addr to netdev */
10676 	ixgbe_add_sanmac_netdev(netdev);
10677 
10678 	e_dev_info("%s\n", ixgbe_default_device_descr);
10679 
10680 #ifdef CONFIG_IXGBE_HWMON
10681 	if (ixgbe_sysfs_init(adapter))
10682 		e_err(probe, "failed to allocate sysfs resources\n");
10683 #endif /* CONFIG_IXGBE_HWMON */
10684 
10685 	ixgbe_dbg_adapter_init(adapter);
10686 
10687 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
10688 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
10689 		hw->mac.ops.setup_link(hw,
10690 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
10691 			true);
10692 
10693 	return 0;
10694 
10695 err_register:
10696 	ixgbe_release_hw_control(adapter);
10697 	ixgbe_clear_interrupt_scheme(adapter);
10698 err_sw_init:
10699 	ixgbe_disable_sriov(adapter);
10700 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
10701 	iounmap(adapter->io_addr);
10702 	kfree(adapter->jump_tables[0]);
10703 	kfree(adapter->mac_table);
10704 	kfree(adapter->rss_key);
10705 err_ioremap:
10706 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
10707 	free_netdev(netdev);
10708 err_alloc_etherdev:
10709 	pci_release_mem_regions(pdev);
10710 err_pci_reg:
10711 err_dma:
10712 	if (!adapter || disable_dev)
10713 		pci_disable_device(pdev);
10714 	return err;
10715 }
10716 
10717 /**
10718  * ixgbe_remove - Device Removal Routine
10719  * @pdev: PCI device information struct
10720  *
10721  * ixgbe_remove is called by the PCI subsystem to alert the driver
10722  * that it should release a PCI device.  The could be caused by a
10723  * Hot-Plug event, or because the driver is going to be removed from
10724  * memory.
10725  **/
10726 static void ixgbe_remove(struct pci_dev *pdev)
10727 {
10728 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10729 	struct net_device *netdev;
10730 	bool disable_dev;
10731 	int i;
10732 
10733 	/* if !adapter then we already cleaned up in probe */
10734 	if (!adapter)
10735 		return;
10736 
10737 	netdev  = adapter->netdev;
10738 	ixgbe_dbg_adapter_exit(adapter);
10739 
10740 	set_bit(__IXGBE_REMOVING, &adapter->state);
10741 	cancel_work_sync(&adapter->service_task);
10742 
10743 
10744 #ifdef CONFIG_IXGBE_DCA
10745 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
10746 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
10747 		dca_remove_requester(&pdev->dev);
10748 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
10749 				IXGBE_DCA_CTRL_DCA_DISABLE);
10750 	}
10751 
10752 #endif
10753 #ifdef CONFIG_IXGBE_HWMON
10754 	ixgbe_sysfs_exit(adapter);
10755 #endif /* CONFIG_IXGBE_HWMON */
10756 
10757 	/* remove the added san mac */
10758 	ixgbe_del_sanmac_netdev(netdev);
10759 
10760 #ifdef CONFIG_PCI_IOV
10761 	ixgbe_disable_sriov(adapter);
10762 #endif
10763 	if (netdev->reg_state == NETREG_REGISTERED)
10764 		unregister_netdev(netdev);
10765 
10766 	ixgbe_stop_ipsec_offload(adapter);
10767 	ixgbe_clear_interrupt_scheme(adapter);
10768 
10769 	ixgbe_release_hw_control(adapter);
10770 
10771 #ifdef CONFIG_DCB
10772 	kfree(adapter->ixgbe_ieee_pfc);
10773 	kfree(adapter->ixgbe_ieee_ets);
10774 
10775 #endif
10776 	iounmap(adapter->io_addr);
10777 	pci_release_mem_regions(pdev);
10778 
10779 	e_dev_info("complete\n");
10780 
10781 	for (i = 0; i < IXGBE_MAX_LINK_HANDLE; i++) {
10782 		if (adapter->jump_tables[i]) {
10783 			kfree(adapter->jump_tables[i]->input);
10784 			kfree(adapter->jump_tables[i]->mask);
10785 		}
10786 		kfree(adapter->jump_tables[i]);
10787 	}
10788 
10789 	kfree(adapter->mac_table);
10790 	kfree(adapter->rss_key);
10791 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
10792 	free_netdev(netdev);
10793 
10794 	pci_disable_pcie_error_reporting(pdev);
10795 
10796 	if (disable_dev)
10797 		pci_disable_device(pdev);
10798 }
10799 
10800 /**
10801  * ixgbe_io_error_detected - called when PCI error is detected
10802  * @pdev: Pointer to PCI device
10803  * @state: The current pci connection state
10804  *
10805  * This function is called after a PCI bus error affecting
10806  * this device has been detected.
10807  */
10808 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
10809 						pci_channel_state_t state)
10810 {
10811 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10812 	struct net_device *netdev = adapter->netdev;
10813 
10814 #ifdef CONFIG_PCI_IOV
10815 	struct ixgbe_hw *hw = &adapter->hw;
10816 	struct pci_dev *bdev, *vfdev;
10817 	u32 dw0, dw1, dw2, dw3;
10818 	int vf, pos;
10819 	u16 req_id, pf_func;
10820 
10821 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
10822 	    adapter->num_vfs == 0)
10823 		goto skip_bad_vf_detection;
10824 
10825 	bdev = pdev->bus->self;
10826 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
10827 		bdev = bdev->bus->self;
10828 
10829 	if (!bdev)
10830 		goto skip_bad_vf_detection;
10831 
10832 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
10833 	if (!pos)
10834 		goto skip_bad_vf_detection;
10835 
10836 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
10837 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
10838 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
10839 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
10840 	if (ixgbe_removed(hw->hw_addr))
10841 		goto skip_bad_vf_detection;
10842 
10843 	req_id = dw1 >> 16;
10844 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
10845 	if (!(req_id & 0x0080))
10846 		goto skip_bad_vf_detection;
10847 
10848 	pf_func = req_id & 0x01;
10849 	if ((pf_func & 1) == (pdev->devfn & 1)) {
10850 		unsigned int device_id;
10851 
10852 		vf = (req_id & 0x7F) >> 1;
10853 		e_dev_err("VF %d has caused a PCIe error\n", vf);
10854 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
10855 				"%8.8x\tdw3: %8.8x\n",
10856 		dw0, dw1, dw2, dw3);
10857 		switch (adapter->hw.mac.type) {
10858 		case ixgbe_mac_82599EB:
10859 			device_id = IXGBE_82599_VF_DEVICE_ID;
10860 			break;
10861 		case ixgbe_mac_X540:
10862 			device_id = IXGBE_X540_VF_DEVICE_ID;
10863 			break;
10864 		case ixgbe_mac_X550:
10865 			device_id = IXGBE_DEV_ID_X550_VF;
10866 			break;
10867 		case ixgbe_mac_X550EM_x:
10868 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
10869 			break;
10870 		case ixgbe_mac_x550em_a:
10871 			device_id = IXGBE_DEV_ID_X550EM_A_VF;
10872 			break;
10873 		default:
10874 			device_id = 0;
10875 			break;
10876 		}
10877 
10878 		/* Find the pci device of the offending VF */
10879 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
10880 		while (vfdev) {
10881 			if (vfdev->devfn == (req_id & 0xFF))
10882 				break;
10883 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
10884 					       device_id, vfdev);
10885 		}
10886 		/*
10887 		 * There's a slim chance the VF could have been hot plugged,
10888 		 * so if it is no longer present we don't need to issue the
10889 		 * VFLR.  Just clean up the AER in that case.
10890 		 */
10891 		if (vfdev) {
10892 			pcie_flr(vfdev);
10893 			/* Free device reference count */
10894 			pci_dev_put(vfdev);
10895 		}
10896 
10897 		pci_cleanup_aer_uncorrect_error_status(pdev);
10898 	}
10899 
10900 	/*
10901 	 * Even though the error may have occurred on the other port
10902 	 * we still need to increment the vf error reference count for
10903 	 * both ports because the I/O resume function will be called
10904 	 * for both of them.
10905 	 */
10906 	adapter->vferr_refcount++;
10907 
10908 	return PCI_ERS_RESULT_RECOVERED;
10909 
10910 skip_bad_vf_detection:
10911 #endif /* CONFIG_PCI_IOV */
10912 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
10913 		return PCI_ERS_RESULT_DISCONNECT;
10914 
10915 	if (!netif_device_present(netdev))
10916 		return PCI_ERS_RESULT_DISCONNECT;
10917 
10918 	rtnl_lock();
10919 	netif_device_detach(netdev);
10920 
10921 	if (netif_running(netdev))
10922 		ixgbe_close_suspend(adapter);
10923 
10924 	if (state == pci_channel_io_perm_failure) {
10925 		rtnl_unlock();
10926 		return PCI_ERS_RESULT_DISCONNECT;
10927 	}
10928 
10929 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
10930 		pci_disable_device(pdev);
10931 	rtnl_unlock();
10932 
10933 	/* Request a slot reset. */
10934 	return PCI_ERS_RESULT_NEED_RESET;
10935 }
10936 
10937 /**
10938  * ixgbe_io_slot_reset - called after the pci bus has been reset.
10939  * @pdev: Pointer to PCI device
10940  *
10941  * Restart the card from scratch, as if from a cold-boot.
10942  */
10943 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
10944 {
10945 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10946 	pci_ers_result_t result;
10947 	int err;
10948 
10949 	if (pci_enable_device_mem(pdev)) {
10950 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
10951 		result = PCI_ERS_RESULT_DISCONNECT;
10952 	} else {
10953 		smp_mb__before_atomic();
10954 		clear_bit(__IXGBE_DISABLED, &adapter->state);
10955 		adapter->hw.hw_addr = adapter->io_addr;
10956 		pci_set_master(pdev);
10957 		pci_restore_state(pdev);
10958 		pci_save_state(pdev);
10959 
10960 		pci_wake_from_d3(pdev, false);
10961 
10962 		ixgbe_reset(adapter);
10963 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10964 		result = PCI_ERS_RESULT_RECOVERED;
10965 	}
10966 
10967 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
10968 	if (err) {
10969 		e_dev_err("pci_cleanup_aer_uncorrect_error_status "
10970 			  "failed 0x%0x\n", err);
10971 		/* non-fatal, continue */
10972 	}
10973 
10974 	return result;
10975 }
10976 
10977 /**
10978  * ixgbe_io_resume - called when traffic can start flowing again.
10979  * @pdev: Pointer to PCI device
10980  *
10981  * This callback is called when the error recovery driver tells us that
10982  * its OK to resume normal operation.
10983  */
10984 static void ixgbe_io_resume(struct pci_dev *pdev)
10985 {
10986 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
10987 	struct net_device *netdev = adapter->netdev;
10988 
10989 #ifdef CONFIG_PCI_IOV
10990 	if (adapter->vferr_refcount) {
10991 		e_info(drv, "Resuming after VF err\n");
10992 		adapter->vferr_refcount--;
10993 		return;
10994 	}
10995 
10996 #endif
10997 	rtnl_lock();
10998 	if (netif_running(netdev))
10999 		ixgbe_open(netdev);
11000 
11001 	netif_device_attach(netdev);
11002 	rtnl_unlock();
11003 }
11004 
11005 static const struct pci_error_handlers ixgbe_err_handler = {
11006 	.error_detected = ixgbe_io_error_detected,
11007 	.slot_reset = ixgbe_io_slot_reset,
11008 	.resume = ixgbe_io_resume,
11009 };
11010 
11011 static struct pci_driver ixgbe_driver = {
11012 	.name     = ixgbe_driver_name,
11013 	.id_table = ixgbe_pci_tbl,
11014 	.probe    = ixgbe_probe,
11015 	.remove   = ixgbe_remove,
11016 #ifdef CONFIG_PM
11017 	.suspend  = ixgbe_suspend,
11018 	.resume   = ixgbe_resume,
11019 #endif
11020 	.shutdown = ixgbe_shutdown,
11021 	.sriov_configure = ixgbe_pci_sriov_configure,
11022 	.err_handler = &ixgbe_err_handler
11023 };
11024 
11025 /**
11026  * ixgbe_init_module - Driver Registration Routine
11027  *
11028  * ixgbe_init_module is the first routine called when the driver is
11029  * loaded. All it does is register with the PCI subsystem.
11030  **/
11031 static int __init ixgbe_init_module(void)
11032 {
11033 	int ret;
11034 	pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
11035 	pr_info("%s\n", ixgbe_copyright);
11036 
11037 	ixgbe_wq = create_singlethread_workqueue(ixgbe_driver_name);
11038 	if (!ixgbe_wq) {
11039 		pr_err("%s: Failed to create workqueue\n", ixgbe_driver_name);
11040 		return -ENOMEM;
11041 	}
11042 
11043 	ixgbe_dbg_init();
11044 
11045 	ret = pci_register_driver(&ixgbe_driver);
11046 	if (ret) {
11047 		destroy_workqueue(ixgbe_wq);
11048 		ixgbe_dbg_exit();
11049 		return ret;
11050 	}
11051 
11052 #ifdef CONFIG_IXGBE_DCA
11053 	dca_register_notify(&dca_notifier);
11054 #endif
11055 
11056 	return 0;
11057 }
11058 
11059 module_init(ixgbe_init_module);
11060 
11061 /**
11062  * ixgbe_exit_module - Driver Exit Cleanup Routine
11063  *
11064  * ixgbe_exit_module is called just before the driver is removed
11065  * from memory.
11066  **/
11067 static void __exit ixgbe_exit_module(void)
11068 {
11069 #ifdef CONFIG_IXGBE_DCA
11070 	dca_unregister_notify(&dca_notifier);
11071 #endif
11072 	pci_unregister_driver(&ixgbe_driver);
11073 
11074 	ixgbe_dbg_exit();
11075 	if (ixgbe_wq) {
11076 		destroy_workqueue(ixgbe_wq);
11077 		ixgbe_wq = NULL;
11078 	}
11079 }
11080 
11081 #ifdef CONFIG_IXGBE_DCA
11082 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
11083 			    void *p)
11084 {
11085 	int ret_val;
11086 
11087 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
11088 					 __ixgbe_notify_dca);
11089 
11090 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
11091 }
11092 
11093 #endif /* CONFIG_IXGBE_DCA */
11094 
11095 module_exit(ixgbe_exit_module);
11096 
11097 /* ixgbe_main.c */
11098