xref: /linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision be239684b18e1cdcafcf8c7face4a2f562c745ad)
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 <linux/numa.h>
31 #include <generated/utsrelease.h>
32 #include <scsi/fc/fc_fcoe.h>
33 #include <net/udp_tunnel.h>
34 #include <net/pkt_cls.h>
35 #include <net/tc_act/tc_gact.h>
36 #include <net/tc_act/tc_mirred.h>
37 #include <net/vxlan.h>
38 #include <net/mpls.h>
39 #include <net/netdev_queues.h>
40 #include <net/xdp_sock_drv.h>
41 #include <net/xfrm.h>
42 
43 #include "ixgbe.h"
44 #include "ixgbe_common.h"
45 #include "ixgbe_dcb_82599.h"
46 #include "ixgbe_phy.h"
47 #include "ixgbe_sriov.h"
48 #include "ixgbe_model.h"
49 #include "ixgbe_txrx_common.h"
50 
51 char ixgbe_driver_name[] = "ixgbe";
52 static const char ixgbe_driver_string[] =
53 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
54 #ifdef IXGBE_FCOE
55 char ixgbe_default_device_descr[] =
56 			      "Intel(R) 10 Gigabit Network Connection";
57 #else
58 static char ixgbe_default_device_descr[] =
59 			      "Intel(R) 10 Gigabit Network Connection";
60 #endif
61 static const char ixgbe_copyright[] =
62 				"Copyright (c) 1999-2016 Intel Corporation.";
63 
64 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";
65 
66 static const struct ixgbe_info *ixgbe_info_tbl[] = {
67 	[board_82598]		= &ixgbe_82598_info,
68 	[board_82599]		= &ixgbe_82599_info,
69 	[board_X540]		= &ixgbe_X540_info,
70 	[board_X550]		= &ixgbe_X550_info,
71 	[board_X550EM_x]	= &ixgbe_X550EM_x_info,
72 	[board_x550em_x_fw]	= &ixgbe_x550em_x_fw_info,
73 	[board_x550em_a]	= &ixgbe_x550em_a_info,
74 	[board_x550em_a_fw]	= &ixgbe_x550em_a_fw_info,
75 };
76 
77 /* ixgbe_pci_tbl - PCI Device ID Table
78  *
79  * Wildcard entries (PCI_ANY_ID) should come last
80  * Last entry must be all 0s
81  *
82  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
83  *   Class, Class Mask, private data (not used) }
84  */
85 static const struct pci_device_id ixgbe_pci_tbl[] = {
86 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
87 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
88 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
89 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
90 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
91 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
92 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
93 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
94 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
95 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
96 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
97 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
98 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
99 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
100 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
101 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
102 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
103 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
104 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
105 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
115 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
116 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
117 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T1), board_X550},
118 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
119 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_XFI), board_X550EM_x},
120 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
121 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_10G_T), board_X550EM_x},
122 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_SFP), board_X550EM_x},
123 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_1G_T), board_x550em_x_fw},
124 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR), board_x550em_a },
125 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR_L), board_x550em_a },
126 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP_N), board_x550em_a },
127 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII), board_x550em_a },
128 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII_L), board_x550em_a },
129 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_10G_T), board_x550em_a},
130 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP), board_x550em_a },
131 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T), board_x550em_a_fw },
132 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T_L), board_x550em_a_fw },
133 	/* required last entry */
134 	{0, }
135 };
136 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
137 
138 #ifdef CONFIG_IXGBE_DCA
139 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
140 			    void *p);
141 static struct notifier_block dca_notifier = {
142 	.notifier_call = ixgbe_notify_dca,
143 	.next          = NULL,
144 	.priority      = 0
145 };
146 #endif
147 
148 #ifdef CONFIG_PCI_IOV
149 static unsigned int max_vfs;
150 module_param(max_vfs, uint, 0);
151 MODULE_PARM_DESC(max_vfs,
152 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
153 #endif /* CONFIG_PCI_IOV */
154 
155 static bool allow_unsupported_sfp;
156 module_param(allow_unsupported_sfp, bool, 0);
157 MODULE_PARM_DESC(allow_unsupported_sfp,
158 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
159 
160 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
161 static int debug = -1;
162 module_param(debug, int, 0);
163 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
164 
165 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
166 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
167 MODULE_LICENSE("GPL v2");
168 
169 DEFINE_STATIC_KEY_FALSE(ixgbe_xdp_locking_key);
170 EXPORT_SYMBOL(ixgbe_xdp_locking_key);
171 
172 static struct workqueue_struct *ixgbe_wq;
173 
174 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
175 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *);
176 
177 static const struct net_device_ops ixgbe_netdev_ops;
178 
179 static bool netif_is_ixgbe(struct net_device *dev)
180 {
181 	return dev && (dev->netdev_ops == &ixgbe_netdev_ops);
182 }
183 
184 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
185 					  u32 reg, u16 *value)
186 {
187 	struct pci_dev *parent_dev;
188 	struct pci_bus *parent_bus;
189 
190 	parent_bus = adapter->pdev->bus->parent;
191 	if (!parent_bus)
192 		return -1;
193 
194 	parent_dev = parent_bus->self;
195 	if (!parent_dev)
196 		return -1;
197 
198 	if (!pci_is_pcie(parent_dev))
199 		return -1;
200 
201 	pcie_capability_read_word(parent_dev, reg, value);
202 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
203 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
204 		return -1;
205 	return 0;
206 }
207 
208 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
209 {
210 	struct ixgbe_hw *hw = &adapter->hw;
211 	u16 link_status = 0;
212 	int err;
213 
214 	hw->bus.type = ixgbe_bus_type_pci_express;
215 
216 	/* Get the negotiated link width and speed from PCI config space of the
217 	 * parent, as this device is behind a switch
218 	 */
219 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
220 
221 	/* assume caller will handle error case */
222 	if (err)
223 		return err;
224 
225 	hw->bus.width = ixgbe_convert_bus_width(link_status);
226 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
227 
228 	return 0;
229 }
230 
231 /**
232  * ixgbe_pcie_from_parent - Determine whether PCIe info should come from parent
233  * @hw: hw specific details
234  *
235  * This function is used by probe to determine whether a device's PCI-Express
236  * bandwidth details should be gathered from the parent bus instead of from the
237  * device. Used to ensure that various locations all have the correct device ID
238  * checks.
239  */
240 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
241 {
242 	switch (hw->device_id) {
243 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
244 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
245 		return true;
246 	default:
247 		return false;
248 	}
249 }
250 
251 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
252 				     int expected_gts)
253 {
254 	struct ixgbe_hw *hw = &adapter->hw;
255 	struct pci_dev *pdev;
256 
257 	/* Some devices are not connected over PCIe and thus do not negotiate
258 	 * speed. These devices do not have valid bus info, and thus any report
259 	 * we generate may not be correct.
260 	 */
261 	if (hw->bus.type == ixgbe_bus_type_internal)
262 		return;
263 
264 	/* determine whether to use the parent device */
265 	if (ixgbe_pcie_from_parent(&adapter->hw))
266 		pdev = adapter->pdev->bus->parent->self;
267 	else
268 		pdev = adapter->pdev;
269 
270 	pcie_print_link_status(pdev);
271 }
272 
273 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
274 {
275 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
276 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
277 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
278 		queue_work(ixgbe_wq, &adapter->service_task);
279 }
280 
281 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
282 {
283 	struct ixgbe_adapter *adapter = hw->back;
284 
285 	if (!hw->hw_addr)
286 		return;
287 	hw->hw_addr = NULL;
288 	e_dev_err("Adapter removed\n");
289 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
290 		ixgbe_service_event_schedule(adapter);
291 }
292 
293 static u32 ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
294 {
295 	u8 __iomem *reg_addr;
296 	u32 value;
297 	int i;
298 
299 	reg_addr = READ_ONCE(hw->hw_addr);
300 	if (ixgbe_removed(reg_addr))
301 		return IXGBE_FAILED_READ_REG;
302 
303 	/* Register read of 0xFFFFFFF can indicate the adapter has been removed,
304 	 * so perform several status register reads to determine if the adapter
305 	 * has been removed.
306 	 */
307 	for (i = 0; i < IXGBE_FAILED_READ_RETRIES; i++) {
308 		value = readl(reg_addr + IXGBE_STATUS);
309 		if (value != IXGBE_FAILED_READ_REG)
310 			break;
311 		mdelay(3);
312 	}
313 
314 	if (value == IXGBE_FAILED_READ_REG)
315 		ixgbe_remove_adapter(hw);
316 	else
317 		value = readl(reg_addr + reg);
318 	return value;
319 }
320 
321 /**
322  * ixgbe_read_reg - Read from device register
323  * @hw: hw specific details
324  * @reg: offset of register to read
325  *
326  * Returns : value read or IXGBE_FAILED_READ_REG if removed
327  *
328  * This function is used to read device registers. It checks for device
329  * removal by confirming any read that returns all ones by checking the
330  * status register value for all ones. This function avoids reading from
331  * the hardware if a removal was previously detected in which case it
332  * returns IXGBE_FAILED_READ_REG (all ones).
333  */
334 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
335 {
336 	u8 __iomem *reg_addr = READ_ONCE(hw->hw_addr);
337 	u32 value;
338 
339 	if (ixgbe_removed(reg_addr))
340 		return IXGBE_FAILED_READ_REG;
341 	if (unlikely(hw->phy.nw_mng_if_sel &
342 		     IXGBE_NW_MNG_IF_SEL_SGMII_ENABLE)) {
343 		struct ixgbe_adapter *adapter;
344 		int i;
345 
346 		for (i = 0; i < 200; ++i) {
347 			value = readl(reg_addr + IXGBE_MAC_SGMII_BUSY);
348 			if (likely(!value))
349 				goto writes_completed;
350 			if (value == IXGBE_FAILED_READ_REG) {
351 				ixgbe_remove_adapter(hw);
352 				return IXGBE_FAILED_READ_REG;
353 			}
354 			udelay(5);
355 		}
356 
357 		adapter = hw->back;
358 		e_warn(hw, "register writes incomplete %08x\n", value);
359 	}
360 
361 writes_completed:
362 	value = readl(reg_addr + reg);
363 	if (unlikely(value == IXGBE_FAILED_READ_REG))
364 		value = ixgbe_check_remove(hw, reg);
365 	return value;
366 }
367 
368 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
369 {
370 	u16 value;
371 
372 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
373 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
374 		ixgbe_remove_adapter(hw);
375 		return true;
376 	}
377 	return false;
378 }
379 
380 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
381 {
382 	struct ixgbe_adapter *adapter = hw->back;
383 	u16 value;
384 
385 	if (ixgbe_removed(hw->hw_addr))
386 		return IXGBE_FAILED_READ_CFG_WORD;
387 	pci_read_config_word(adapter->pdev, reg, &value);
388 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
389 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
390 		return IXGBE_FAILED_READ_CFG_WORD;
391 	return value;
392 }
393 
394 #ifdef CONFIG_PCI_IOV
395 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
396 {
397 	struct ixgbe_adapter *adapter = hw->back;
398 	u32 value;
399 
400 	if (ixgbe_removed(hw->hw_addr))
401 		return IXGBE_FAILED_READ_CFG_DWORD;
402 	pci_read_config_dword(adapter->pdev, reg, &value);
403 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
404 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
405 		return IXGBE_FAILED_READ_CFG_DWORD;
406 	return value;
407 }
408 #endif /* CONFIG_PCI_IOV */
409 
410 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
411 {
412 	struct ixgbe_adapter *adapter = hw->back;
413 
414 	if (ixgbe_removed(hw->hw_addr))
415 		return;
416 	pci_write_config_word(adapter->pdev, reg, value);
417 }
418 
419 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
420 {
421 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
422 
423 	/* flush memory to make sure state is correct before next watchdog */
424 	smp_mb__before_atomic();
425 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
426 }
427 
428 struct ixgbe_reg_info {
429 	u32 ofs;
430 	char *name;
431 };
432 
433 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
434 
435 	/* General Registers */
436 	{IXGBE_CTRL, "CTRL"},
437 	{IXGBE_STATUS, "STATUS"},
438 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
439 
440 	/* Interrupt Registers */
441 	{IXGBE_EICR, "EICR"},
442 
443 	/* RX Registers */
444 	{IXGBE_SRRCTL(0), "SRRCTL"},
445 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
446 	{IXGBE_RDLEN(0), "RDLEN"},
447 	{IXGBE_RDH(0), "RDH"},
448 	{IXGBE_RDT(0), "RDT"},
449 	{IXGBE_RXDCTL(0), "RXDCTL"},
450 	{IXGBE_RDBAL(0), "RDBAL"},
451 	{IXGBE_RDBAH(0), "RDBAH"},
452 
453 	/* TX Registers */
454 	{IXGBE_TDBAL(0), "TDBAL"},
455 	{IXGBE_TDBAH(0), "TDBAH"},
456 	{IXGBE_TDLEN(0), "TDLEN"},
457 	{IXGBE_TDH(0), "TDH"},
458 	{IXGBE_TDT(0), "TDT"},
459 	{IXGBE_TXDCTL(0), "TXDCTL"},
460 
461 	/* List Terminator */
462 	{ .name = NULL }
463 };
464 
465 
466 /*
467  * ixgbe_regdump - register printout routine
468  */
469 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
470 {
471 	int i;
472 	char rname[16];
473 	u32 regs[64];
474 
475 	switch (reginfo->ofs) {
476 	case IXGBE_SRRCTL(0):
477 		for (i = 0; i < 64; i++)
478 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
479 		break;
480 	case IXGBE_DCA_RXCTRL(0):
481 		for (i = 0; i < 64; i++)
482 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
483 		break;
484 	case IXGBE_RDLEN(0):
485 		for (i = 0; i < 64; i++)
486 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
487 		break;
488 	case IXGBE_RDH(0):
489 		for (i = 0; i < 64; i++)
490 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
491 		break;
492 	case IXGBE_RDT(0):
493 		for (i = 0; i < 64; i++)
494 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
495 		break;
496 	case IXGBE_RXDCTL(0):
497 		for (i = 0; i < 64; i++)
498 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
499 		break;
500 	case IXGBE_RDBAL(0):
501 		for (i = 0; i < 64; i++)
502 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
503 		break;
504 	case IXGBE_RDBAH(0):
505 		for (i = 0; i < 64; i++)
506 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
507 		break;
508 	case IXGBE_TDBAL(0):
509 		for (i = 0; i < 64; i++)
510 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
511 		break;
512 	case IXGBE_TDBAH(0):
513 		for (i = 0; i < 64; i++)
514 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
515 		break;
516 	case IXGBE_TDLEN(0):
517 		for (i = 0; i < 64; i++)
518 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
519 		break;
520 	case IXGBE_TDH(0):
521 		for (i = 0; i < 64; i++)
522 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
523 		break;
524 	case IXGBE_TDT(0):
525 		for (i = 0; i < 64; i++)
526 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
527 		break;
528 	case IXGBE_TXDCTL(0):
529 		for (i = 0; i < 64; i++)
530 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
531 		break;
532 	default:
533 		pr_info("%-15s %08x\n",
534 			reginfo->name, IXGBE_READ_REG(hw, reginfo->ofs));
535 		return;
536 	}
537 
538 	i = 0;
539 	while (i < 64) {
540 		int j;
541 		char buf[9 * 8 + 1];
542 		char *p = buf;
543 
544 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i, i + 7);
545 		for (j = 0; j < 8; j++)
546 			p += sprintf(p, " %08x", regs[i++]);
547 		pr_err("%-15s%s\n", rname, buf);
548 	}
549 
550 }
551 
552 static void ixgbe_print_buffer(struct ixgbe_ring *ring, int n)
553 {
554 	struct ixgbe_tx_buffer *tx_buffer;
555 
556 	tx_buffer = &ring->tx_buffer_info[ring->next_to_clean];
557 	pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
558 		n, ring->next_to_use, ring->next_to_clean,
559 		(u64)dma_unmap_addr(tx_buffer, dma),
560 		dma_unmap_len(tx_buffer, len),
561 		tx_buffer->next_to_watch,
562 		(u64)tx_buffer->time_stamp);
563 }
564 
565 /*
566  * ixgbe_dump - Print registers, tx-rings and rx-rings
567  */
568 static void ixgbe_dump(struct ixgbe_adapter *adapter)
569 {
570 	struct net_device *netdev = adapter->netdev;
571 	struct ixgbe_hw *hw = &adapter->hw;
572 	struct ixgbe_reg_info *reginfo;
573 	int n = 0;
574 	struct ixgbe_ring *ring;
575 	struct ixgbe_tx_buffer *tx_buffer;
576 	union ixgbe_adv_tx_desc *tx_desc;
577 	struct my_u0 { u64 a; u64 b; } *u0;
578 	struct ixgbe_ring *rx_ring;
579 	union ixgbe_adv_rx_desc *rx_desc;
580 	struct ixgbe_rx_buffer *rx_buffer_info;
581 	int i = 0;
582 
583 	if (!netif_msg_hw(adapter))
584 		return;
585 
586 	/* Print netdevice Info */
587 	if (netdev) {
588 		dev_info(&adapter->pdev->dev, "Net device Info\n");
589 		pr_info("Device Name     state            "
590 			"trans_start\n");
591 		pr_info("%-15s %016lX %016lX\n",
592 			netdev->name,
593 			netdev->state,
594 			dev_trans_start(netdev));
595 	}
596 
597 	/* Print Registers */
598 	dev_info(&adapter->pdev->dev, "Register Dump\n");
599 	pr_info(" Register Name   Value\n");
600 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
601 	     reginfo->name; reginfo++) {
602 		ixgbe_regdump(hw, reginfo);
603 	}
604 
605 	/* Print TX Ring Summary */
606 	if (!netdev || !netif_running(netdev))
607 		return;
608 
609 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
610 	pr_info(" %s     %s              %s        %s\n",
611 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
612 		"leng", "ntw", "timestamp");
613 	for (n = 0; n < adapter->num_tx_queues; n++) {
614 		ring = adapter->tx_ring[n];
615 		ixgbe_print_buffer(ring, n);
616 	}
617 
618 	for (n = 0; n < adapter->num_xdp_queues; n++) {
619 		ring = adapter->xdp_ring[n];
620 		ixgbe_print_buffer(ring, n);
621 	}
622 
623 	/* Print TX Rings */
624 	if (!netif_msg_tx_done(adapter))
625 		goto rx_ring_summary;
626 
627 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
628 
629 	/* Transmit Descriptor Formats
630 	 *
631 	 * 82598 Advanced Transmit Descriptor
632 	 *   +--------------------------------------------------------------+
633 	 * 0 |         Buffer Address [63:0]                                |
634 	 *   +--------------------------------------------------------------+
635 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
636 	 *   +--------------------------------------------------------------+
637 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
638 	 *
639 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
640 	 *   +--------------------------------------------------------------+
641 	 * 0 |                          RSV [63:0]                          |
642 	 *   +--------------------------------------------------------------+
643 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
644 	 *   +--------------------------------------------------------------+
645 	 *   63                       36 35   32 31                         0
646 	 *
647 	 * 82599+ Advanced Transmit Descriptor
648 	 *   +--------------------------------------------------------------+
649 	 * 0 |         Buffer Address [63:0]                                |
650 	 *   +--------------------------------------------------------------+
651 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
652 	 *   +--------------------------------------------------------------+
653 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
654 	 *
655 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
656 	 *   +--------------------------------------------------------------+
657 	 * 0 |                          RSV [63:0]                          |
658 	 *   +--------------------------------------------------------------+
659 	 * 8 |            RSV           |  STA  |           RSV             |
660 	 *   +--------------------------------------------------------------+
661 	 *   63                       36 35   32 31                         0
662 	 */
663 
664 	for (n = 0; n < adapter->num_tx_queues; n++) {
665 		ring = adapter->tx_ring[n];
666 		pr_info("------------------------------------\n");
667 		pr_info("TX QUEUE INDEX = %d\n", ring->queue_index);
668 		pr_info("------------------------------------\n");
669 		pr_info("%s%s    %s              %s        %s          %s\n",
670 			"T [desc]     [address 63:0  ] ",
671 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
672 			"leng", "ntw", "timestamp", "bi->skb");
673 
674 		for (i = 0; ring->desc && (i < ring->count); i++) {
675 			tx_desc = IXGBE_TX_DESC(ring, i);
676 			tx_buffer = &ring->tx_buffer_info[i];
677 			u0 = (struct my_u0 *)tx_desc;
678 			if (dma_unmap_len(tx_buffer, len) > 0) {
679 				const char *ring_desc;
680 
681 				if (i == ring->next_to_use &&
682 				    i == ring->next_to_clean)
683 					ring_desc = " NTC/U";
684 				else if (i == ring->next_to_use)
685 					ring_desc = " NTU";
686 				else if (i == ring->next_to_clean)
687 					ring_desc = " NTC";
688 				else
689 					ring_desc = "";
690 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p%s",
691 					i,
692 					le64_to_cpu((__force __le64)u0->a),
693 					le64_to_cpu((__force __le64)u0->b),
694 					(u64)dma_unmap_addr(tx_buffer, dma),
695 					dma_unmap_len(tx_buffer, len),
696 					tx_buffer->next_to_watch,
697 					(u64)tx_buffer->time_stamp,
698 					tx_buffer->skb,
699 					ring_desc);
700 
701 				if (netif_msg_pktdata(adapter) &&
702 				    tx_buffer->skb)
703 					print_hex_dump(KERN_INFO, "",
704 						DUMP_PREFIX_ADDRESS, 16, 1,
705 						tx_buffer->skb->data,
706 						dma_unmap_len(tx_buffer, len),
707 						true);
708 			}
709 		}
710 	}
711 
712 	/* Print RX Rings Summary */
713 rx_ring_summary:
714 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
715 	pr_info("Queue [NTU] [NTC]\n");
716 	for (n = 0; n < adapter->num_rx_queues; n++) {
717 		rx_ring = adapter->rx_ring[n];
718 		pr_info("%5d %5X %5X\n",
719 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
720 	}
721 
722 	/* Print RX Rings */
723 	if (!netif_msg_rx_status(adapter))
724 		return;
725 
726 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
727 
728 	/* Receive Descriptor Formats
729 	 *
730 	 * 82598 Advanced Receive Descriptor (Read) Format
731 	 *    63                                           1        0
732 	 *    +-----------------------------------------------------+
733 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
734 	 *    +----------------------------------------------+------+
735 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
736 	 *    +-----------------------------------------------------+
737 	 *
738 	 *
739 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
740 	 *
741 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
742 	 *   +------------------------------------------------------+
743 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
744 	 *   | Packet   | IP     |   |          |     | Type | Type |
745 	 *   | Checksum | Ident  |   |          |     |      |      |
746 	 *   +------------------------------------------------------+
747 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
748 	 *   +------------------------------------------------------+
749 	 *   63       48 47    32 31            20 19               0
750 	 *
751 	 * 82599+ Advanced Receive Descriptor (Read) Format
752 	 *    63                                           1        0
753 	 *    +-----------------------------------------------------+
754 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
755 	 *    +----------------------------------------------+------+
756 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
757 	 *    +-----------------------------------------------------+
758 	 *
759 	 *
760 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
761 	 *
762 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
763 	 *   +------------------------------------------------------+
764 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
765 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
766 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
767 	 *   +------------------------------------------------------+
768 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
769 	 *   +------------------------------------------------------+
770 	 *   63       48 47    32 31          20 19                 0
771 	 */
772 
773 	for (n = 0; n < adapter->num_rx_queues; n++) {
774 		rx_ring = adapter->rx_ring[n];
775 		pr_info("------------------------------------\n");
776 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
777 		pr_info("------------------------------------\n");
778 		pr_info("%s%s%s\n",
779 			"R  [desc]      [ PktBuf     A0] ",
780 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
781 			"<-- Adv Rx Read format");
782 		pr_info("%s%s%s\n",
783 			"RWB[desc]      [PcsmIpSHl PtRs] ",
784 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
785 			"<-- Adv Rx Write-Back format");
786 
787 		for (i = 0; i < rx_ring->count; i++) {
788 			const char *ring_desc;
789 
790 			if (i == rx_ring->next_to_use)
791 				ring_desc = " NTU";
792 			else if (i == rx_ring->next_to_clean)
793 				ring_desc = " NTC";
794 			else
795 				ring_desc = "";
796 
797 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
798 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
799 			u0 = (struct my_u0 *)rx_desc;
800 			if (rx_desc->wb.upper.length) {
801 				/* Descriptor Done */
802 				pr_info("RWB[0x%03X]     %016llX %016llX ---------------- %p%s\n",
803 					i,
804 					le64_to_cpu((__force __le64)u0->a),
805 					le64_to_cpu((__force __le64)u0->b),
806 					rx_buffer_info->skb,
807 					ring_desc);
808 			} else {
809 				pr_info("R  [0x%03X]     %016llX %016llX %016llX %p%s\n",
810 					i,
811 					le64_to_cpu((__force __le64)u0->a),
812 					le64_to_cpu((__force __le64)u0->b),
813 					(u64)rx_buffer_info->dma,
814 					rx_buffer_info->skb,
815 					ring_desc);
816 
817 				if (netif_msg_pktdata(adapter) &&
818 				    rx_buffer_info->dma) {
819 					print_hex_dump(KERN_INFO, "",
820 					   DUMP_PREFIX_ADDRESS, 16, 1,
821 					   page_address(rx_buffer_info->page) +
822 						    rx_buffer_info->page_offset,
823 					   ixgbe_rx_bufsz(rx_ring), true);
824 				}
825 			}
826 		}
827 	}
828 }
829 
830 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
831 {
832 	u32 ctrl_ext;
833 
834 	/* Let firmware take over control of h/w */
835 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
836 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
837 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
838 }
839 
840 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
841 {
842 	u32 ctrl_ext;
843 
844 	/* Let firmware know the driver has taken over */
845 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
846 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
847 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
848 }
849 
850 /**
851  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
852  * @adapter: pointer to adapter struct
853  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
854  * @queue: queue to map the corresponding interrupt to
855  * @msix_vector: the vector to map to the corresponding queue
856  *
857  */
858 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
859 			   u8 queue, u8 msix_vector)
860 {
861 	u32 ivar, index;
862 	struct ixgbe_hw *hw = &adapter->hw;
863 	switch (hw->mac.type) {
864 	case ixgbe_mac_82598EB:
865 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
866 		if (direction == -1)
867 			direction = 0;
868 		index = (((direction * 64) + queue) >> 2) & 0x1F;
869 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
870 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
871 		ivar |= (msix_vector << (8 * (queue & 0x3)));
872 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
873 		break;
874 	case ixgbe_mac_82599EB:
875 	case ixgbe_mac_X540:
876 	case ixgbe_mac_X550:
877 	case ixgbe_mac_X550EM_x:
878 	case ixgbe_mac_x550em_a:
879 		if (direction == -1) {
880 			/* other causes */
881 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
882 			index = ((queue & 1) * 8);
883 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
884 			ivar &= ~(0xFF << index);
885 			ivar |= (msix_vector << index);
886 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
887 			break;
888 		} else {
889 			/* tx or rx causes */
890 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
891 			index = ((16 * (queue & 1)) + (8 * direction));
892 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
893 			ivar &= ~(0xFF << index);
894 			ivar |= (msix_vector << index);
895 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
896 			break;
897 		}
898 	default:
899 		break;
900 	}
901 }
902 
903 void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
904 			    u64 qmask)
905 {
906 	u32 mask;
907 
908 	switch (adapter->hw.mac.type) {
909 	case ixgbe_mac_82598EB:
910 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
911 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
912 		break;
913 	case ixgbe_mac_82599EB:
914 	case ixgbe_mac_X540:
915 	case ixgbe_mac_X550:
916 	case ixgbe_mac_X550EM_x:
917 	case ixgbe_mac_x550em_a:
918 		mask = (qmask & 0xFFFFFFFF);
919 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
920 		mask = (qmask >> 32);
921 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
922 		break;
923 	default:
924 		break;
925 	}
926 }
927 
928 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
929 {
930 	struct ixgbe_hw *hw = &adapter->hw;
931 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
932 	int i;
933 	u32 data;
934 
935 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
936 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
937 		return;
938 
939 	switch (hw->mac.type) {
940 	case ixgbe_mac_82598EB:
941 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
942 		break;
943 	default:
944 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
945 	}
946 	hwstats->lxoffrxc += data;
947 
948 	/* refill credits (no tx hang) if we received xoff */
949 	if (!data)
950 		return;
951 
952 	for (i = 0; i < adapter->num_tx_queues; i++)
953 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
954 			  &adapter->tx_ring[i]->state);
955 
956 	for (i = 0; i < adapter->num_xdp_queues; i++)
957 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
958 			  &adapter->xdp_ring[i]->state);
959 }
960 
961 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
962 {
963 	struct ixgbe_hw *hw = &adapter->hw;
964 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
965 	u32 xoff[8] = {0};
966 	u8 tc;
967 	int i;
968 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
969 
970 	if (adapter->ixgbe_ieee_pfc)
971 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
972 
973 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
974 		ixgbe_update_xoff_rx_lfc(adapter);
975 		return;
976 	}
977 
978 	/* update stats for each tc, only valid with PFC enabled */
979 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
980 		u32 pxoffrxc;
981 
982 		switch (hw->mac.type) {
983 		case ixgbe_mac_82598EB:
984 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
985 			break;
986 		default:
987 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
988 		}
989 		hwstats->pxoffrxc[i] += pxoffrxc;
990 		/* Get the TC for given UP */
991 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
992 		xoff[tc] += pxoffrxc;
993 	}
994 
995 	/* disarm tx queues that have received xoff frames */
996 	for (i = 0; i < adapter->num_tx_queues; i++) {
997 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
998 
999 		tc = tx_ring->dcb_tc;
1000 		if (xoff[tc])
1001 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1002 	}
1003 
1004 	for (i = 0; i < adapter->num_xdp_queues; i++) {
1005 		struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
1006 
1007 		tc = xdp_ring->dcb_tc;
1008 		if (xoff[tc])
1009 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &xdp_ring->state);
1010 	}
1011 }
1012 
1013 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1014 {
1015 	return ring->stats.packets;
1016 }
1017 
1018 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1019 {
1020 	unsigned int head, tail;
1021 
1022 	head = ring->next_to_clean;
1023 	tail = ring->next_to_use;
1024 
1025 	return ((head <= tail) ? tail : tail + ring->count) - head;
1026 }
1027 
1028 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1029 {
1030 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1031 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1032 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1033 
1034 	clear_check_for_tx_hang(tx_ring);
1035 
1036 	/*
1037 	 * Check for a hung queue, but be thorough. This verifies
1038 	 * that a transmit has been completed since the previous
1039 	 * check AND there is at least one packet pending. The
1040 	 * ARMED bit is set to indicate a potential hang. The
1041 	 * bit is cleared if a pause frame is received to remove
1042 	 * false hang detection due to PFC or 802.3x frames. By
1043 	 * requiring this to fail twice we avoid races with
1044 	 * pfc clearing the ARMED bit and conditions where we
1045 	 * run the check_tx_hang logic with a transmit completion
1046 	 * pending but without time to complete it yet.
1047 	 */
1048 	if (tx_done_old == tx_done && tx_pending)
1049 		/* make sure it is true for two checks in a row */
1050 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1051 					&tx_ring->state);
1052 	/* update completed stats and continue */
1053 	tx_ring->tx_stats.tx_done_old = tx_done;
1054 	/* reset the countdown */
1055 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1056 
1057 	return false;
1058 }
1059 
1060 /**
1061  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1062  * @adapter: driver private struct
1063  **/
1064 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1065 {
1066 
1067 	/* Do the reset outside of interrupt context */
1068 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1069 		set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
1070 		e_warn(drv, "initiating reset due to tx timeout\n");
1071 		ixgbe_service_event_schedule(adapter);
1072 	}
1073 }
1074 
1075 /**
1076  * ixgbe_tx_maxrate - callback to set the maximum per-queue bitrate
1077  * @netdev: network interface device structure
1078  * @queue_index: Tx queue to set
1079  * @maxrate: desired maximum transmit bitrate
1080  **/
1081 static int ixgbe_tx_maxrate(struct net_device *netdev,
1082 			    int queue_index, u32 maxrate)
1083 {
1084 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1085 	struct ixgbe_hw *hw = &adapter->hw;
1086 	u32 bcnrc_val = ixgbe_link_mbps(adapter);
1087 
1088 	if (!maxrate)
1089 		return 0;
1090 
1091 	/* Calculate the rate factor values to set */
1092 	bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1093 	bcnrc_val /= maxrate;
1094 
1095 	/* clear everything but the rate factor */
1096 	bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1097 	IXGBE_RTTBCNRC_RF_DEC_MASK;
1098 
1099 	/* enable the rate scheduler */
1100 	bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1101 
1102 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_index);
1103 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1104 
1105 	return 0;
1106 }
1107 
1108 /**
1109  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1110  * @q_vector: structure containing interrupt and ring information
1111  * @tx_ring: tx ring to clean
1112  * @napi_budget: Used to determine if we are in netpoll
1113  **/
1114 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1115 			       struct ixgbe_ring *tx_ring, int napi_budget)
1116 {
1117 	struct ixgbe_adapter *adapter = q_vector->adapter;
1118 	struct ixgbe_tx_buffer *tx_buffer;
1119 	union ixgbe_adv_tx_desc *tx_desc;
1120 	unsigned int total_bytes = 0, total_packets = 0, total_ipsec = 0;
1121 	unsigned int budget = q_vector->tx.work_limit;
1122 	unsigned int i = tx_ring->next_to_clean;
1123 	struct netdev_queue *txq;
1124 
1125 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1126 		return true;
1127 
1128 	tx_buffer = &tx_ring->tx_buffer_info[i];
1129 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1130 	i -= tx_ring->count;
1131 
1132 	do {
1133 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1134 
1135 		/* if next_to_watch is not set then there is no work pending */
1136 		if (!eop_desc)
1137 			break;
1138 
1139 		/* prevent any other reads prior to eop_desc */
1140 		smp_rmb();
1141 
1142 		/* if DD is not set pending work has not been completed */
1143 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1144 			break;
1145 
1146 		/* clear next_to_watch to prevent false hangs */
1147 		tx_buffer->next_to_watch = NULL;
1148 
1149 		/* update the statistics for this packet */
1150 		total_bytes += tx_buffer->bytecount;
1151 		total_packets += tx_buffer->gso_segs;
1152 		if (tx_buffer->tx_flags & IXGBE_TX_FLAGS_IPSEC)
1153 			total_ipsec++;
1154 
1155 		/* free the skb */
1156 		if (ring_is_xdp(tx_ring))
1157 			xdp_return_frame(tx_buffer->xdpf);
1158 		else
1159 			napi_consume_skb(tx_buffer->skb, napi_budget);
1160 
1161 		/* unmap skb header data */
1162 		dma_unmap_single(tx_ring->dev,
1163 				 dma_unmap_addr(tx_buffer, dma),
1164 				 dma_unmap_len(tx_buffer, len),
1165 				 DMA_TO_DEVICE);
1166 
1167 		/* clear tx_buffer data */
1168 		dma_unmap_len_set(tx_buffer, len, 0);
1169 
1170 		/* unmap remaining buffers */
1171 		while (tx_desc != eop_desc) {
1172 			tx_buffer++;
1173 			tx_desc++;
1174 			i++;
1175 			if (unlikely(!i)) {
1176 				i -= tx_ring->count;
1177 				tx_buffer = tx_ring->tx_buffer_info;
1178 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1179 			}
1180 
1181 			/* unmap any remaining paged data */
1182 			if (dma_unmap_len(tx_buffer, len)) {
1183 				dma_unmap_page(tx_ring->dev,
1184 					       dma_unmap_addr(tx_buffer, dma),
1185 					       dma_unmap_len(tx_buffer, len),
1186 					       DMA_TO_DEVICE);
1187 				dma_unmap_len_set(tx_buffer, len, 0);
1188 			}
1189 		}
1190 
1191 		/* move us one more past the eop_desc for start of next pkt */
1192 		tx_buffer++;
1193 		tx_desc++;
1194 		i++;
1195 		if (unlikely(!i)) {
1196 			i -= tx_ring->count;
1197 			tx_buffer = tx_ring->tx_buffer_info;
1198 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1199 		}
1200 
1201 		/* issue prefetch for next Tx descriptor */
1202 		prefetch(tx_desc);
1203 
1204 		/* update budget accounting */
1205 		budget--;
1206 	} while (likely(budget));
1207 
1208 	i += tx_ring->count;
1209 	tx_ring->next_to_clean = i;
1210 	u64_stats_update_begin(&tx_ring->syncp);
1211 	tx_ring->stats.bytes += total_bytes;
1212 	tx_ring->stats.packets += total_packets;
1213 	u64_stats_update_end(&tx_ring->syncp);
1214 	q_vector->tx.total_bytes += total_bytes;
1215 	q_vector->tx.total_packets += total_packets;
1216 	adapter->tx_ipsec += total_ipsec;
1217 
1218 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1219 		/* schedule immediate reset if we believe we hung */
1220 		struct ixgbe_hw *hw = &adapter->hw;
1221 		e_err(drv, "Detected Tx Unit Hang %s\n"
1222 			"  Tx Queue             <%d>\n"
1223 			"  TDH, TDT             <%x>, <%x>\n"
1224 			"  next_to_use          <%x>\n"
1225 			"  next_to_clean        <%x>\n"
1226 			"tx_buffer_info[next_to_clean]\n"
1227 			"  time_stamp           <%lx>\n"
1228 			"  jiffies              <%lx>\n",
1229 			ring_is_xdp(tx_ring) ? "(XDP)" : "",
1230 			tx_ring->queue_index,
1231 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1232 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1233 			tx_ring->next_to_use, i,
1234 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1235 
1236 		if (!ring_is_xdp(tx_ring))
1237 			netif_stop_subqueue(tx_ring->netdev,
1238 					    tx_ring->queue_index);
1239 
1240 		e_info(probe,
1241 		       "tx hang %d detected on queue %d, resetting adapter\n",
1242 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1243 
1244 		/* schedule immediate reset if we believe we hung */
1245 		ixgbe_tx_timeout_reset(adapter);
1246 
1247 		/* the adapter is about to reset, no point in enabling stuff */
1248 		return true;
1249 	}
1250 
1251 	if (ring_is_xdp(tx_ring))
1252 		return !!budget;
1253 
1254 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1255 	txq = netdev_get_tx_queue(tx_ring->netdev, tx_ring->queue_index);
1256 	if (!__netif_txq_completed_wake(txq, total_packets, total_bytes,
1257 					ixgbe_desc_unused(tx_ring),
1258 					TX_WAKE_THRESHOLD,
1259 					!netif_carrier_ok(tx_ring->netdev) ||
1260 					test_bit(__IXGBE_DOWN, &adapter->state)))
1261 		++tx_ring->tx_stats.restart_queue;
1262 
1263 	return !!budget;
1264 }
1265 
1266 #ifdef CONFIG_IXGBE_DCA
1267 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1268 				struct ixgbe_ring *tx_ring,
1269 				int cpu)
1270 {
1271 	struct ixgbe_hw *hw = &adapter->hw;
1272 	u32 txctrl = 0;
1273 	u16 reg_offset;
1274 
1275 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1276 		txctrl = dca3_get_tag(tx_ring->dev, cpu);
1277 
1278 	switch (hw->mac.type) {
1279 	case ixgbe_mac_82598EB:
1280 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1281 		break;
1282 	case ixgbe_mac_82599EB:
1283 	case ixgbe_mac_X540:
1284 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1285 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1286 		break;
1287 	default:
1288 		/* for unknown hardware do not write register */
1289 		return;
1290 	}
1291 
1292 	/*
1293 	 * We can enable relaxed ordering for reads, but not writes when
1294 	 * DCA is enabled.  This is due to a known issue in some chipsets
1295 	 * which will cause the DCA tag to be cleared.
1296 	 */
1297 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1298 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1299 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1300 
1301 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1302 }
1303 
1304 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1305 				struct ixgbe_ring *rx_ring,
1306 				int cpu)
1307 {
1308 	struct ixgbe_hw *hw = &adapter->hw;
1309 	u32 rxctrl = 0;
1310 	u8 reg_idx = rx_ring->reg_idx;
1311 
1312 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1313 		rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1314 
1315 	switch (hw->mac.type) {
1316 	case ixgbe_mac_82599EB:
1317 	case ixgbe_mac_X540:
1318 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1319 		break;
1320 	default:
1321 		break;
1322 	}
1323 
1324 	/*
1325 	 * We can enable relaxed ordering for reads, but not writes when
1326 	 * DCA is enabled.  This is due to a known issue in some chipsets
1327 	 * which will cause the DCA tag to be cleared.
1328 	 */
1329 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1330 		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1331 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1332 
1333 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1334 }
1335 
1336 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1337 {
1338 	struct ixgbe_adapter *adapter = q_vector->adapter;
1339 	struct ixgbe_ring *ring;
1340 	int cpu = get_cpu();
1341 
1342 	if (q_vector->cpu == cpu)
1343 		goto out_no_update;
1344 
1345 	ixgbe_for_each_ring(ring, q_vector->tx)
1346 		ixgbe_update_tx_dca(adapter, ring, cpu);
1347 
1348 	ixgbe_for_each_ring(ring, q_vector->rx)
1349 		ixgbe_update_rx_dca(adapter, ring, cpu);
1350 
1351 	q_vector->cpu = cpu;
1352 out_no_update:
1353 	put_cpu();
1354 }
1355 
1356 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1357 {
1358 	int i;
1359 
1360 	/* always use CB2 mode, difference is masked in the CB driver */
1361 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1362 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1363 				IXGBE_DCA_CTRL_DCA_MODE_CB2);
1364 	else
1365 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1366 				IXGBE_DCA_CTRL_DCA_DISABLE);
1367 
1368 	for (i = 0; i < adapter->num_q_vectors; i++) {
1369 		adapter->q_vector[i]->cpu = -1;
1370 		ixgbe_update_dca(adapter->q_vector[i]);
1371 	}
1372 }
1373 
1374 static int __ixgbe_notify_dca(struct device *dev, void *data)
1375 {
1376 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1377 	unsigned long event = *(unsigned long *)data;
1378 
1379 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1380 		return 0;
1381 
1382 	switch (event) {
1383 	case DCA_PROVIDER_ADD:
1384 		/* if we're already enabled, don't do it again */
1385 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1386 			break;
1387 		if (dca_add_requester(dev) == 0) {
1388 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1389 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1390 					IXGBE_DCA_CTRL_DCA_MODE_CB2);
1391 			break;
1392 		}
1393 		fallthrough; /* DCA is disabled. */
1394 	case DCA_PROVIDER_REMOVE:
1395 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1396 			dca_remove_requester(dev);
1397 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1398 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1399 					IXGBE_DCA_CTRL_DCA_DISABLE);
1400 		}
1401 		break;
1402 	}
1403 
1404 	return 0;
1405 }
1406 
1407 #endif /* CONFIG_IXGBE_DCA */
1408 
1409 #define IXGBE_RSS_L4_TYPES_MASK \
1410 	((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1411 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1412 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1413 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1414 
1415 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1416 				 union ixgbe_adv_rx_desc *rx_desc,
1417 				 struct sk_buff *skb)
1418 {
1419 	u16 rss_type;
1420 
1421 	if (!(ring->netdev->features & NETIF_F_RXHASH))
1422 		return;
1423 
1424 	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1425 		   IXGBE_RXDADV_RSSTYPE_MASK;
1426 
1427 	if (!rss_type)
1428 		return;
1429 
1430 	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1431 		     (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1432 		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1433 }
1434 
1435 #ifdef IXGBE_FCOE
1436 /**
1437  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1438  * @ring: structure containing ring specific data
1439  * @rx_desc: advanced rx descriptor
1440  *
1441  * Returns : true if it is FCoE pkt
1442  */
1443 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1444 				    union ixgbe_adv_rx_desc *rx_desc)
1445 {
1446 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1447 
1448 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1449 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1450 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1451 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1452 }
1453 
1454 #endif /* IXGBE_FCOE */
1455 /**
1456  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1457  * @ring: structure containing ring specific data
1458  * @rx_desc: current Rx descriptor being processed
1459  * @skb: skb currently being received and modified
1460  **/
1461 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1462 				     union ixgbe_adv_rx_desc *rx_desc,
1463 				     struct sk_buff *skb)
1464 {
1465 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1466 	bool encap_pkt = false;
1467 
1468 	skb_checksum_none_assert(skb);
1469 
1470 	/* Rx csum disabled */
1471 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1472 		return;
1473 
1474 	/* check for VXLAN and Geneve packets */
1475 	if (pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) {
1476 		encap_pkt = true;
1477 		skb->encapsulation = 1;
1478 	}
1479 
1480 	/* if IP and error */
1481 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1482 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1483 		ring->rx_stats.csum_err++;
1484 		return;
1485 	}
1486 
1487 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1488 		return;
1489 
1490 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1491 		/*
1492 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1493 		 * checksum errors.
1494 		 */
1495 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1496 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1497 			return;
1498 
1499 		ring->rx_stats.csum_err++;
1500 		return;
1501 	}
1502 
1503 	/* It must be a TCP or UDP packet with a valid checksum */
1504 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1505 	if (encap_pkt) {
1506 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1507 			return;
1508 
1509 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1510 			skb->ip_summed = CHECKSUM_NONE;
1511 			return;
1512 		}
1513 		/* If we checked the outer header let the stack know */
1514 		skb->csum_level = 1;
1515 	}
1516 }
1517 
1518 static unsigned int ixgbe_rx_offset(struct ixgbe_ring *rx_ring)
1519 {
1520 	return ring_uses_build_skb(rx_ring) ? IXGBE_SKB_PAD : 0;
1521 }
1522 
1523 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1524 				    struct ixgbe_rx_buffer *bi)
1525 {
1526 	struct page *page = bi->page;
1527 	dma_addr_t dma;
1528 
1529 	/* since we are recycling buffers we should seldom need to alloc */
1530 	if (likely(page))
1531 		return true;
1532 
1533 	/* alloc new page for storage */
1534 	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1535 	if (unlikely(!page)) {
1536 		rx_ring->rx_stats.alloc_rx_page_failed++;
1537 		return false;
1538 	}
1539 
1540 	/* map page for use */
1541 	dma = dma_map_page_attrs(rx_ring->dev, page, 0,
1542 				 ixgbe_rx_pg_size(rx_ring),
1543 				 DMA_FROM_DEVICE,
1544 				 IXGBE_RX_DMA_ATTR);
1545 
1546 	/*
1547 	 * if mapping failed free memory back to system since
1548 	 * there isn't much point in holding memory we can't use
1549 	 */
1550 	if (dma_mapping_error(rx_ring->dev, dma)) {
1551 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1552 
1553 		rx_ring->rx_stats.alloc_rx_page_failed++;
1554 		return false;
1555 	}
1556 
1557 	bi->dma = dma;
1558 	bi->page = page;
1559 	bi->page_offset = rx_ring->rx_offset;
1560 	page_ref_add(page, USHRT_MAX - 1);
1561 	bi->pagecnt_bias = USHRT_MAX;
1562 	rx_ring->rx_stats.alloc_rx_page++;
1563 
1564 	return true;
1565 }
1566 
1567 /**
1568  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1569  * @rx_ring: ring to place buffers on
1570  * @cleaned_count: number of buffers to replace
1571  **/
1572 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1573 {
1574 	union ixgbe_adv_rx_desc *rx_desc;
1575 	struct ixgbe_rx_buffer *bi;
1576 	u16 i = rx_ring->next_to_use;
1577 	u16 bufsz;
1578 
1579 	/* nothing to do */
1580 	if (!cleaned_count)
1581 		return;
1582 
1583 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1584 	bi = &rx_ring->rx_buffer_info[i];
1585 	i -= rx_ring->count;
1586 
1587 	bufsz = ixgbe_rx_bufsz(rx_ring);
1588 
1589 	do {
1590 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1591 			break;
1592 
1593 		/* sync the buffer for use by the device */
1594 		dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
1595 						 bi->page_offset, bufsz,
1596 						 DMA_FROM_DEVICE);
1597 
1598 		/*
1599 		 * Refresh the desc even if buffer_addrs didn't change
1600 		 * because each write-back erases this info.
1601 		 */
1602 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1603 
1604 		rx_desc++;
1605 		bi++;
1606 		i++;
1607 		if (unlikely(!i)) {
1608 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1609 			bi = rx_ring->rx_buffer_info;
1610 			i -= rx_ring->count;
1611 		}
1612 
1613 		/* clear the length for the next_to_use descriptor */
1614 		rx_desc->wb.upper.length = 0;
1615 
1616 		cleaned_count--;
1617 	} while (cleaned_count);
1618 
1619 	i += rx_ring->count;
1620 
1621 	if (rx_ring->next_to_use != i) {
1622 		rx_ring->next_to_use = i;
1623 
1624 		/* update next to alloc since we have filled the ring */
1625 		rx_ring->next_to_alloc = i;
1626 
1627 		/* Force memory writes to complete before letting h/w
1628 		 * know there are new descriptors to fetch.  (Only
1629 		 * applicable for weak-ordered memory model archs,
1630 		 * such as IA-64).
1631 		 */
1632 		wmb();
1633 		writel(i, rx_ring->tail);
1634 	}
1635 }
1636 
1637 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1638 				   struct sk_buff *skb)
1639 {
1640 	u16 hdr_len = skb_headlen(skb);
1641 
1642 	/* set gso_size to avoid messing up TCP MSS */
1643 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1644 						 IXGBE_CB(skb)->append_cnt);
1645 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1646 }
1647 
1648 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1649 				   struct sk_buff *skb)
1650 {
1651 	/* if append_cnt is 0 then frame is not RSC */
1652 	if (!IXGBE_CB(skb)->append_cnt)
1653 		return;
1654 
1655 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1656 	rx_ring->rx_stats.rsc_flush++;
1657 
1658 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1659 
1660 	/* gso_size is computed using append_cnt so always clear it last */
1661 	IXGBE_CB(skb)->append_cnt = 0;
1662 }
1663 
1664 /**
1665  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1666  * @rx_ring: rx descriptor ring packet is being transacted on
1667  * @rx_desc: pointer to the EOP Rx descriptor
1668  * @skb: pointer to current skb being populated
1669  *
1670  * This function checks the ring, descriptor, and packet information in
1671  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1672  * other fields within the skb.
1673  **/
1674 void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1675 			      union ixgbe_adv_rx_desc *rx_desc,
1676 			      struct sk_buff *skb)
1677 {
1678 	struct net_device *dev = rx_ring->netdev;
1679 	u32 flags = rx_ring->q_vector->adapter->flags;
1680 
1681 	ixgbe_update_rsc_stats(rx_ring, skb);
1682 
1683 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1684 
1685 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1686 
1687 	if (unlikely(flags & IXGBE_FLAG_RX_HWTSTAMP_ENABLED))
1688 		ixgbe_ptp_rx_hwtstamp(rx_ring, rx_desc, skb);
1689 
1690 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1691 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1692 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1693 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1694 	}
1695 
1696 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_SECP))
1697 		ixgbe_ipsec_rx(rx_ring, rx_desc, skb);
1698 
1699 	/* record Rx queue, or update MACVLAN statistics */
1700 	if (netif_is_ixgbe(dev))
1701 		skb_record_rx_queue(skb, rx_ring->queue_index);
1702 	else
1703 		macvlan_count_rx(netdev_priv(dev), skb->len + ETH_HLEN, true,
1704 				 false);
1705 
1706 	skb->protocol = eth_type_trans(skb, dev);
1707 }
1708 
1709 void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1710 		  struct sk_buff *skb)
1711 {
1712 	napi_gro_receive(&q_vector->napi, skb);
1713 }
1714 
1715 /**
1716  * ixgbe_is_non_eop - process handling of non-EOP buffers
1717  * @rx_ring: Rx ring being processed
1718  * @rx_desc: Rx descriptor for current buffer
1719  * @skb: Current socket buffer containing buffer in progress
1720  *
1721  * This function updates next to clean.  If the buffer is an EOP buffer
1722  * this function exits returning false, otherwise it will place the
1723  * sk_buff in the next buffer to be chained and return true indicating
1724  * that this is in fact a non-EOP buffer.
1725  **/
1726 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1727 			     union ixgbe_adv_rx_desc *rx_desc,
1728 			     struct sk_buff *skb)
1729 {
1730 	u32 ntc = rx_ring->next_to_clean + 1;
1731 
1732 	/* fetch, update, and store next to clean */
1733 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1734 	rx_ring->next_to_clean = ntc;
1735 
1736 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1737 
1738 	/* update RSC append count if present */
1739 	if (ring_is_rsc_enabled(rx_ring)) {
1740 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1741 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1742 
1743 		if (unlikely(rsc_enabled)) {
1744 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1745 
1746 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1747 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1748 
1749 			/* update ntc based on RSC value */
1750 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1751 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1752 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1753 		}
1754 	}
1755 
1756 	/* if we are the last buffer then there is nothing else to do */
1757 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1758 		return false;
1759 
1760 	/* place skb in next buffer to be received */
1761 	rx_ring->rx_buffer_info[ntc].skb = skb;
1762 	rx_ring->rx_stats.non_eop_descs++;
1763 
1764 	return true;
1765 }
1766 
1767 /**
1768  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1769  * @rx_ring: rx descriptor ring packet is being transacted on
1770  * @skb: pointer to current skb being adjusted
1771  *
1772  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1773  * main difference between this version and the original function is that
1774  * this function can make several assumptions about the state of things
1775  * that allow for significant optimizations versus the standard function.
1776  * As a result we can do things like drop a frag and maintain an accurate
1777  * truesize for the skb.
1778  */
1779 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1780 			    struct sk_buff *skb)
1781 {
1782 	skb_frag_t *frag = &skb_shinfo(skb)->frags[0];
1783 	unsigned char *va;
1784 	unsigned int pull_len;
1785 
1786 	/*
1787 	 * it is valid to use page_address instead of kmap since we are
1788 	 * working with pages allocated out of the lomem pool per
1789 	 * alloc_page(GFP_ATOMIC)
1790 	 */
1791 	va = skb_frag_address(frag);
1792 
1793 	/*
1794 	 * we need the header to contain the greater of either ETH_HLEN or
1795 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1796 	 */
1797 	pull_len = eth_get_headlen(skb->dev, va, IXGBE_RX_HDR_SIZE);
1798 
1799 	/* align pull length to size of long to optimize memcpy performance */
1800 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1801 
1802 	/* update all of the pointers */
1803 	skb_frag_size_sub(frag, pull_len);
1804 	skb_frag_off_add(frag, pull_len);
1805 	skb->data_len -= pull_len;
1806 	skb->tail += pull_len;
1807 }
1808 
1809 /**
1810  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1811  * @rx_ring: rx descriptor ring packet is being transacted on
1812  * @skb: pointer to current skb being updated
1813  *
1814  * This function provides a basic DMA sync up for the first fragment of an
1815  * skb.  The reason for doing this is that the first fragment cannot be
1816  * unmapped until we have reached the end of packet descriptor for a buffer
1817  * chain.
1818  */
1819 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1820 				struct sk_buff *skb)
1821 {
1822 	if (ring_uses_build_skb(rx_ring)) {
1823 		unsigned long mask = (unsigned long)ixgbe_rx_pg_size(rx_ring) - 1;
1824 		unsigned long offset = (unsigned long)(skb->data) & mask;
1825 
1826 		dma_sync_single_range_for_cpu(rx_ring->dev,
1827 					      IXGBE_CB(skb)->dma,
1828 					      offset,
1829 					      skb_headlen(skb),
1830 					      DMA_FROM_DEVICE);
1831 	} else {
1832 		skb_frag_t *frag = &skb_shinfo(skb)->frags[0];
1833 
1834 		dma_sync_single_range_for_cpu(rx_ring->dev,
1835 					      IXGBE_CB(skb)->dma,
1836 					      skb_frag_off(frag),
1837 					      skb_frag_size(frag),
1838 					      DMA_FROM_DEVICE);
1839 	}
1840 
1841 	/* If the page was released, just unmap it. */
1842 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1843 		dma_unmap_page_attrs(rx_ring->dev, IXGBE_CB(skb)->dma,
1844 				     ixgbe_rx_pg_size(rx_ring),
1845 				     DMA_FROM_DEVICE,
1846 				     IXGBE_RX_DMA_ATTR);
1847 	}
1848 }
1849 
1850 /**
1851  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1852  * @rx_ring: rx descriptor ring packet is being transacted on
1853  * @rx_desc: pointer to the EOP Rx descriptor
1854  * @skb: pointer to current skb being fixed
1855  *
1856  * Check if the skb is valid in the XDP case it will be an error pointer.
1857  * Return true in this case to abort processing and advance to next
1858  * descriptor.
1859  *
1860  * Check for corrupted packet headers caused by senders on the local L2
1861  * embedded NIC switch not setting up their Tx Descriptors right.  These
1862  * should be very rare.
1863  *
1864  * Also address the case where we are pulling data in on pages only
1865  * and as such no data is present in the skb header.
1866  *
1867  * In addition if skb is not at least 60 bytes we need to pad it so that
1868  * it is large enough to qualify as a valid Ethernet frame.
1869  *
1870  * Returns true if an error was encountered and skb was freed.
1871  **/
1872 bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1873 			   union ixgbe_adv_rx_desc *rx_desc,
1874 			   struct sk_buff *skb)
1875 {
1876 	struct net_device *netdev = rx_ring->netdev;
1877 
1878 	/* XDP packets use error pointer so abort at this point */
1879 	if (IS_ERR(skb))
1880 		return true;
1881 
1882 	/* Verify netdev is present, and that packet does not have any
1883 	 * errors that would be unacceptable to the netdev.
1884 	 */
1885 	if (!netdev ||
1886 	    (unlikely(ixgbe_test_staterr(rx_desc,
1887 					 IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1888 	     !(netdev->features & NETIF_F_RXALL)))) {
1889 		dev_kfree_skb_any(skb);
1890 		return true;
1891 	}
1892 
1893 	/* place header in linear portion of buffer */
1894 	if (!skb_headlen(skb))
1895 		ixgbe_pull_tail(rx_ring, skb);
1896 
1897 #ifdef IXGBE_FCOE
1898 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1899 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1900 		return false;
1901 
1902 #endif
1903 	/* if eth_skb_pad returns an error the skb was freed */
1904 	if (eth_skb_pad(skb))
1905 		return true;
1906 
1907 	return false;
1908 }
1909 
1910 /**
1911  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1912  * @rx_ring: rx descriptor ring to store buffers on
1913  * @old_buff: donor buffer to have page reused
1914  *
1915  * Synchronizes page for reuse by the adapter
1916  **/
1917 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1918 				struct ixgbe_rx_buffer *old_buff)
1919 {
1920 	struct ixgbe_rx_buffer *new_buff;
1921 	u16 nta = rx_ring->next_to_alloc;
1922 
1923 	new_buff = &rx_ring->rx_buffer_info[nta];
1924 
1925 	/* update, and store next to alloc */
1926 	nta++;
1927 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1928 
1929 	/* Transfer page from old buffer to new buffer.
1930 	 * Move each member individually to avoid possible store
1931 	 * forwarding stalls and unnecessary copy of skb.
1932 	 */
1933 	new_buff->dma		= old_buff->dma;
1934 	new_buff->page		= old_buff->page;
1935 	new_buff->page_offset	= old_buff->page_offset;
1936 	new_buff->pagecnt_bias	= old_buff->pagecnt_bias;
1937 }
1938 
1939 static bool ixgbe_can_reuse_rx_page(struct ixgbe_rx_buffer *rx_buffer,
1940 				    int rx_buffer_pgcnt)
1941 {
1942 	unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
1943 	struct page *page = rx_buffer->page;
1944 
1945 	/* avoid re-using remote and pfmemalloc pages */
1946 	if (!dev_page_is_reusable(page))
1947 		return false;
1948 
1949 #if (PAGE_SIZE < 8192)
1950 	/* if we are only owner of page we can reuse it */
1951 	if (unlikely((rx_buffer_pgcnt - pagecnt_bias) > 1))
1952 		return false;
1953 #else
1954 	/* The last offset is a bit aggressive in that we assume the
1955 	 * worst case of FCoE being enabled and using a 3K buffer.
1956 	 * However this should have minimal impact as the 1K extra is
1957 	 * still less than one buffer in size.
1958 	 */
1959 #define IXGBE_LAST_OFFSET \
1960 	(SKB_WITH_OVERHEAD(PAGE_SIZE) - IXGBE_RXBUFFER_3K)
1961 	if (rx_buffer->page_offset > IXGBE_LAST_OFFSET)
1962 		return false;
1963 #endif
1964 
1965 	/* If we have drained the page fragment pool we need to update
1966 	 * the pagecnt_bias and page count so that we fully restock the
1967 	 * number of references the driver holds.
1968 	 */
1969 	if (unlikely(pagecnt_bias == 1)) {
1970 		page_ref_add(page, USHRT_MAX - 1);
1971 		rx_buffer->pagecnt_bias = USHRT_MAX;
1972 	}
1973 
1974 	return true;
1975 }
1976 
1977 /**
1978  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1979  * @rx_ring: rx descriptor ring to transact packets on
1980  * @rx_buffer: buffer containing page to add
1981  * @skb: sk_buff to place the data into
1982  * @size: size of data in rx_buffer
1983  *
1984  * This function will add the data contained in rx_buffer->page to the skb.
1985  * This is done either through a direct copy if the data in the buffer is
1986  * less than the skb header size, otherwise it will just attach the page as
1987  * a frag to the skb.
1988  *
1989  * The function will then update the page offset if necessary and return
1990  * true if the buffer can be reused by the adapter.
1991  **/
1992 static void ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1993 			      struct ixgbe_rx_buffer *rx_buffer,
1994 			      struct sk_buff *skb,
1995 			      unsigned int size)
1996 {
1997 #if (PAGE_SIZE < 8192)
1998 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
1999 #else
2000 	unsigned int truesize = rx_ring->rx_offset ?
2001 				SKB_DATA_ALIGN(rx_ring->rx_offset + size) :
2002 				SKB_DATA_ALIGN(size);
2003 #endif
2004 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
2005 			rx_buffer->page_offset, size, truesize);
2006 #if (PAGE_SIZE < 8192)
2007 	rx_buffer->page_offset ^= truesize;
2008 #else
2009 	rx_buffer->page_offset += truesize;
2010 #endif
2011 }
2012 
2013 static struct ixgbe_rx_buffer *ixgbe_get_rx_buffer(struct ixgbe_ring *rx_ring,
2014 						   union ixgbe_adv_rx_desc *rx_desc,
2015 						   struct sk_buff **skb,
2016 						   const unsigned int size,
2017 						   int *rx_buffer_pgcnt)
2018 {
2019 	struct ixgbe_rx_buffer *rx_buffer;
2020 
2021 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
2022 	*rx_buffer_pgcnt =
2023 #if (PAGE_SIZE < 8192)
2024 		page_count(rx_buffer->page);
2025 #else
2026 		0;
2027 #endif
2028 	prefetchw(rx_buffer->page);
2029 	*skb = rx_buffer->skb;
2030 
2031 	/* Delay unmapping of the first packet. It carries the header
2032 	 * information, HW may still access the header after the writeback.
2033 	 * Only unmap it when EOP is reached
2034 	 */
2035 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)) {
2036 		if (!*skb)
2037 			goto skip_sync;
2038 	} else {
2039 		if (*skb)
2040 			ixgbe_dma_sync_frag(rx_ring, *skb);
2041 	}
2042 
2043 	/* we are reusing so sync this buffer for CPU use */
2044 	dma_sync_single_range_for_cpu(rx_ring->dev,
2045 				      rx_buffer->dma,
2046 				      rx_buffer->page_offset,
2047 				      size,
2048 				      DMA_FROM_DEVICE);
2049 skip_sync:
2050 	rx_buffer->pagecnt_bias--;
2051 
2052 	return rx_buffer;
2053 }
2054 
2055 static void ixgbe_put_rx_buffer(struct ixgbe_ring *rx_ring,
2056 				struct ixgbe_rx_buffer *rx_buffer,
2057 				struct sk_buff *skb,
2058 				int rx_buffer_pgcnt)
2059 {
2060 	if (ixgbe_can_reuse_rx_page(rx_buffer, rx_buffer_pgcnt)) {
2061 		/* hand second half of page back to the ring */
2062 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2063 	} else {
2064 		if (!IS_ERR(skb) && IXGBE_CB(skb)->dma == rx_buffer->dma) {
2065 			/* the page has been released from the ring */
2066 			IXGBE_CB(skb)->page_released = true;
2067 		} else {
2068 			/* we are not reusing the buffer so unmap it */
2069 			dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
2070 					     ixgbe_rx_pg_size(rx_ring),
2071 					     DMA_FROM_DEVICE,
2072 					     IXGBE_RX_DMA_ATTR);
2073 		}
2074 		__page_frag_cache_drain(rx_buffer->page,
2075 					rx_buffer->pagecnt_bias);
2076 	}
2077 
2078 	/* clear contents of rx_buffer */
2079 	rx_buffer->page = NULL;
2080 	rx_buffer->skb = NULL;
2081 }
2082 
2083 static struct sk_buff *ixgbe_construct_skb(struct ixgbe_ring *rx_ring,
2084 					   struct ixgbe_rx_buffer *rx_buffer,
2085 					   struct xdp_buff *xdp,
2086 					   union ixgbe_adv_rx_desc *rx_desc)
2087 {
2088 	unsigned int size = xdp->data_end - xdp->data;
2089 #if (PAGE_SIZE < 8192)
2090 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2091 #else
2092 	unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end -
2093 					       xdp->data_hard_start);
2094 #endif
2095 	struct sk_buff *skb;
2096 
2097 	/* prefetch first cache line of first page */
2098 	net_prefetch(xdp->data);
2099 
2100 	/* Note, we get here by enabling legacy-rx via:
2101 	 *
2102 	 *    ethtool --set-priv-flags <dev> legacy-rx on
2103 	 *
2104 	 * In this mode, we currently get 0 extra XDP headroom as
2105 	 * opposed to having legacy-rx off, where we process XDP
2106 	 * packets going to stack via ixgbe_build_skb(). The latter
2107 	 * provides us currently with 192 bytes of headroom.
2108 	 *
2109 	 * For ixgbe_construct_skb() mode it means that the
2110 	 * xdp->data_meta will always point to xdp->data, since
2111 	 * the helper cannot expand the head. Should this ever
2112 	 * change in future for legacy-rx mode on, then lets also
2113 	 * add xdp->data_meta handling here.
2114 	 */
2115 
2116 	/* allocate a skb to store the frags */
2117 	skb = napi_alloc_skb(&rx_ring->q_vector->napi, IXGBE_RX_HDR_SIZE);
2118 	if (unlikely(!skb))
2119 		return NULL;
2120 
2121 	if (size > IXGBE_RX_HDR_SIZE) {
2122 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2123 			IXGBE_CB(skb)->dma = rx_buffer->dma;
2124 
2125 		skb_add_rx_frag(skb, 0, rx_buffer->page,
2126 				xdp->data - page_address(rx_buffer->page),
2127 				size, truesize);
2128 #if (PAGE_SIZE < 8192)
2129 		rx_buffer->page_offset ^= truesize;
2130 #else
2131 		rx_buffer->page_offset += truesize;
2132 #endif
2133 	} else {
2134 		memcpy(__skb_put(skb, size),
2135 		       xdp->data, ALIGN(size, sizeof(long)));
2136 		rx_buffer->pagecnt_bias++;
2137 	}
2138 
2139 	return skb;
2140 }
2141 
2142 static struct sk_buff *ixgbe_build_skb(struct ixgbe_ring *rx_ring,
2143 				       struct ixgbe_rx_buffer *rx_buffer,
2144 				       struct xdp_buff *xdp,
2145 				       union ixgbe_adv_rx_desc *rx_desc)
2146 {
2147 	unsigned int metasize = xdp->data - xdp->data_meta;
2148 #if (PAGE_SIZE < 8192)
2149 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2150 #else
2151 	unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
2152 				SKB_DATA_ALIGN(xdp->data_end -
2153 					       xdp->data_hard_start);
2154 #endif
2155 	struct sk_buff *skb;
2156 
2157 	/* Prefetch first cache line of first page. If xdp->data_meta
2158 	 * is unused, this points extactly as xdp->data, otherwise we
2159 	 * likely have a consumer accessing first few bytes of meta
2160 	 * data, and then actual data.
2161 	 */
2162 	net_prefetch(xdp->data_meta);
2163 
2164 	/* build an skb to around the page buffer */
2165 	skb = napi_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 static struct sk_buff *ixgbe_run_xdp(struct ixgbe_adapter *adapter,
2190 				     struct ixgbe_ring *rx_ring,
2191 				     struct xdp_buff *xdp)
2192 {
2193 	int err, result = IXGBE_XDP_PASS;
2194 	struct bpf_prog *xdp_prog;
2195 	struct ixgbe_ring *ring;
2196 	struct xdp_frame *xdpf;
2197 	u32 act;
2198 
2199 	xdp_prog = READ_ONCE(rx_ring->xdp_prog);
2200 
2201 	if (!xdp_prog)
2202 		goto xdp_out;
2203 
2204 	prefetchw(xdp->data_hard_start); /* xdp_frame write */
2205 
2206 	act = bpf_prog_run_xdp(xdp_prog, xdp);
2207 	switch (act) {
2208 	case XDP_PASS:
2209 		break;
2210 	case XDP_TX:
2211 		xdpf = xdp_convert_buff_to_frame(xdp);
2212 		if (unlikely(!xdpf))
2213 			goto out_failure;
2214 		ring = ixgbe_determine_xdp_ring(adapter);
2215 		if (static_branch_unlikely(&ixgbe_xdp_locking_key))
2216 			spin_lock(&ring->tx_lock);
2217 		result = ixgbe_xmit_xdp_ring(ring, xdpf);
2218 		if (static_branch_unlikely(&ixgbe_xdp_locking_key))
2219 			spin_unlock(&ring->tx_lock);
2220 		if (result == IXGBE_XDP_CONSUMED)
2221 			goto out_failure;
2222 		break;
2223 	case XDP_REDIRECT:
2224 		err = xdp_do_redirect(adapter->netdev, xdp, xdp_prog);
2225 		if (err)
2226 			goto out_failure;
2227 		result = IXGBE_XDP_REDIR;
2228 		break;
2229 	default:
2230 		bpf_warn_invalid_xdp_action(rx_ring->netdev, xdp_prog, act);
2231 		fallthrough;
2232 	case XDP_ABORTED:
2233 out_failure:
2234 		trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
2235 		fallthrough; /* handle aborts by dropping packet */
2236 	case XDP_DROP:
2237 		result = IXGBE_XDP_CONSUMED;
2238 		break;
2239 	}
2240 xdp_out:
2241 	return ERR_PTR(-result);
2242 }
2243 
2244 static unsigned int ixgbe_rx_frame_truesize(struct ixgbe_ring *rx_ring,
2245 					    unsigned int size)
2246 {
2247 	unsigned int truesize;
2248 
2249 #if (PAGE_SIZE < 8192)
2250 	truesize = ixgbe_rx_pg_size(rx_ring) / 2; /* Must be power-of-2 */
2251 #else
2252 	truesize = rx_ring->rx_offset ?
2253 		SKB_DATA_ALIGN(rx_ring->rx_offset + size) +
2254 		SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) :
2255 		SKB_DATA_ALIGN(size);
2256 #endif
2257 	return truesize;
2258 }
2259 
2260 static void ixgbe_rx_buffer_flip(struct ixgbe_ring *rx_ring,
2261 				 struct ixgbe_rx_buffer *rx_buffer,
2262 				 unsigned int size)
2263 {
2264 	unsigned int truesize = ixgbe_rx_frame_truesize(rx_ring, size);
2265 #if (PAGE_SIZE < 8192)
2266 	rx_buffer->page_offset ^= truesize;
2267 #else
2268 	rx_buffer->page_offset += truesize;
2269 #endif
2270 }
2271 
2272 /**
2273  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2274  * @q_vector: structure containing interrupt and ring information
2275  * @rx_ring: rx descriptor ring to transact packets on
2276  * @budget: Total limit on number of packets to process
2277  *
2278  * This function provides a "bounce buffer" approach to Rx interrupt
2279  * processing.  The advantage to this is that on systems that have
2280  * expensive overhead for IOMMU access this provides a means of avoiding
2281  * it by maintaining the mapping of the page to the syste.
2282  *
2283  * Returns amount of work completed
2284  **/
2285 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2286 			       struct ixgbe_ring *rx_ring,
2287 			       const int budget)
2288 {
2289 	unsigned int total_rx_bytes = 0, total_rx_packets = 0, frame_sz = 0;
2290 	struct ixgbe_adapter *adapter = q_vector->adapter;
2291 #ifdef IXGBE_FCOE
2292 	int ddp_bytes;
2293 	unsigned int mss = 0;
2294 #endif /* IXGBE_FCOE */
2295 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2296 	unsigned int offset = rx_ring->rx_offset;
2297 	unsigned int xdp_xmit = 0;
2298 	struct xdp_buff xdp;
2299 
2300 	/* Frame size depend on rx_ring setup when PAGE_SIZE=4K */
2301 #if (PAGE_SIZE < 8192)
2302 	frame_sz = ixgbe_rx_frame_truesize(rx_ring, 0);
2303 #endif
2304 	xdp_init_buff(&xdp, frame_sz, &rx_ring->xdp_rxq);
2305 
2306 	while (likely(total_rx_packets < budget)) {
2307 		union ixgbe_adv_rx_desc *rx_desc;
2308 		struct ixgbe_rx_buffer *rx_buffer;
2309 		struct sk_buff *skb;
2310 		int rx_buffer_pgcnt;
2311 		unsigned int size;
2312 
2313 		/* return some buffers to hardware, one at a time is too slow */
2314 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2315 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2316 			cleaned_count = 0;
2317 		}
2318 
2319 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2320 		size = le16_to_cpu(rx_desc->wb.upper.length);
2321 		if (!size)
2322 			break;
2323 
2324 		/* This memory barrier is needed to keep us from reading
2325 		 * any other fields out of the rx_desc until we know the
2326 		 * descriptor has been written back
2327 		 */
2328 		dma_rmb();
2329 
2330 		rx_buffer = ixgbe_get_rx_buffer(rx_ring, rx_desc, &skb, size, &rx_buffer_pgcnt);
2331 
2332 		/* retrieve a buffer from the ring */
2333 		if (!skb) {
2334 			unsigned char *hard_start;
2335 
2336 			hard_start = page_address(rx_buffer->page) +
2337 				     rx_buffer->page_offset - offset;
2338 			xdp_prepare_buff(&xdp, hard_start, offset, size, true);
2339 			xdp_buff_clear_frags_flag(&xdp);
2340 #if (PAGE_SIZE > 4096)
2341 			/* At larger PAGE_SIZE, frame_sz depend on len size */
2342 			xdp.frame_sz = ixgbe_rx_frame_truesize(rx_ring, size);
2343 #endif
2344 			skb = ixgbe_run_xdp(adapter, rx_ring, &xdp);
2345 		}
2346 
2347 		if (IS_ERR(skb)) {
2348 			unsigned int xdp_res = -PTR_ERR(skb);
2349 
2350 			if (xdp_res & (IXGBE_XDP_TX | IXGBE_XDP_REDIR)) {
2351 				xdp_xmit |= xdp_res;
2352 				ixgbe_rx_buffer_flip(rx_ring, rx_buffer, size);
2353 			} else {
2354 				rx_buffer->pagecnt_bias++;
2355 			}
2356 			total_rx_packets++;
2357 			total_rx_bytes += size;
2358 		} else if (skb) {
2359 			ixgbe_add_rx_frag(rx_ring, rx_buffer, skb, size);
2360 		} else if (ring_uses_build_skb(rx_ring)) {
2361 			skb = ixgbe_build_skb(rx_ring, rx_buffer,
2362 					      &xdp, rx_desc);
2363 		} else {
2364 			skb = ixgbe_construct_skb(rx_ring, rx_buffer,
2365 						  &xdp, rx_desc);
2366 		}
2367 
2368 		/* exit if we failed to retrieve a buffer */
2369 		if (!skb) {
2370 			rx_ring->rx_stats.alloc_rx_buff_failed++;
2371 			rx_buffer->pagecnt_bias++;
2372 			break;
2373 		}
2374 
2375 		ixgbe_put_rx_buffer(rx_ring, rx_buffer, skb, rx_buffer_pgcnt);
2376 		cleaned_count++;
2377 
2378 		/* place incomplete frames back on ring for completion */
2379 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2380 			continue;
2381 
2382 		/* verify the packet layout is correct */
2383 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2384 			continue;
2385 
2386 		/* probably a little skewed due to removing CRC */
2387 		total_rx_bytes += skb->len;
2388 
2389 		/* populate checksum, timestamp, VLAN, and protocol */
2390 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2391 
2392 #ifdef IXGBE_FCOE
2393 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2394 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2395 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2396 			/* include DDPed FCoE data */
2397 			if (ddp_bytes > 0) {
2398 				if (!mss) {
2399 					mss = rx_ring->netdev->mtu -
2400 						sizeof(struct fcoe_hdr) -
2401 						sizeof(struct fc_frame_header) -
2402 						sizeof(struct fcoe_crc_eof);
2403 					if (mss > 512)
2404 						mss &= ~511;
2405 				}
2406 				total_rx_bytes += ddp_bytes;
2407 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2408 								 mss);
2409 			}
2410 			if (!ddp_bytes) {
2411 				dev_kfree_skb_any(skb);
2412 				continue;
2413 			}
2414 		}
2415 
2416 #endif /* IXGBE_FCOE */
2417 		ixgbe_rx_skb(q_vector, skb);
2418 
2419 		/* update budget accounting */
2420 		total_rx_packets++;
2421 	}
2422 
2423 	if (xdp_xmit & IXGBE_XDP_REDIR)
2424 		xdp_do_flush();
2425 
2426 	if (xdp_xmit & IXGBE_XDP_TX) {
2427 		struct ixgbe_ring *ring = ixgbe_determine_xdp_ring(adapter);
2428 
2429 		ixgbe_xdp_ring_update_tail_locked(ring);
2430 	}
2431 
2432 	u64_stats_update_begin(&rx_ring->syncp);
2433 	rx_ring->stats.packets += total_rx_packets;
2434 	rx_ring->stats.bytes += total_rx_bytes;
2435 	u64_stats_update_end(&rx_ring->syncp);
2436 	q_vector->rx.total_packets += total_rx_packets;
2437 	q_vector->rx.total_bytes += total_rx_bytes;
2438 
2439 	return total_rx_packets;
2440 }
2441 
2442 /**
2443  * ixgbe_configure_msix - Configure MSI-X hardware
2444  * @adapter: board private structure
2445  *
2446  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2447  * interrupts.
2448  **/
2449 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2450 {
2451 	struct ixgbe_q_vector *q_vector;
2452 	int v_idx;
2453 	u32 mask;
2454 
2455 	/* Populate MSIX to EITR Select */
2456 	if (adapter->num_vfs > 32) {
2457 		u32 eitrsel = BIT(adapter->num_vfs - 32) - 1;
2458 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2459 	}
2460 
2461 	/*
2462 	 * Populate the IVAR table and set the ITR values to the
2463 	 * corresponding register.
2464 	 */
2465 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2466 		struct ixgbe_ring *ring;
2467 		q_vector = adapter->q_vector[v_idx];
2468 
2469 		ixgbe_for_each_ring(ring, q_vector->rx)
2470 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2471 
2472 		ixgbe_for_each_ring(ring, q_vector->tx)
2473 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2474 
2475 		ixgbe_write_eitr(q_vector);
2476 	}
2477 
2478 	switch (adapter->hw.mac.type) {
2479 	case ixgbe_mac_82598EB:
2480 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2481 			       v_idx);
2482 		break;
2483 	case ixgbe_mac_82599EB:
2484 	case ixgbe_mac_X540:
2485 	case ixgbe_mac_X550:
2486 	case ixgbe_mac_X550EM_x:
2487 	case ixgbe_mac_x550em_a:
2488 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2489 		break;
2490 	default:
2491 		break;
2492 	}
2493 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2494 
2495 	/* set up to autoclear timer, and the vectors */
2496 	mask = IXGBE_EIMS_ENABLE_MASK;
2497 	mask &= ~(IXGBE_EIMS_OTHER |
2498 		  IXGBE_EIMS_MAILBOX |
2499 		  IXGBE_EIMS_LSC);
2500 
2501 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2502 }
2503 
2504 /**
2505  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2506  * @q_vector: structure containing interrupt and ring information
2507  * @ring_container: structure containing ring performance data
2508  *
2509  *      Stores a new ITR value based on packets and byte
2510  *      counts during the last interrupt.  The advantage of per interrupt
2511  *      computation is faster updates and more accurate ITR for the current
2512  *      traffic pattern.  Constants in this function were computed
2513  *      based on theoretical maximum wire speed and thresholds were set based
2514  *      on testing data as well as attempting to minimize response time
2515  *      while increasing bulk throughput.
2516  **/
2517 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2518 			     struct ixgbe_ring_container *ring_container)
2519 {
2520 	unsigned int itr = IXGBE_ITR_ADAPTIVE_MIN_USECS |
2521 			   IXGBE_ITR_ADAPTIVE_LATENCY;
2522 	unsigned int avg_wire_size, packets, bytes;
2523 	unsigned long next_update = jiffies;
2524 
2525 	/* If we don't have any rings just leave ourselves set for maximum
2526 	 * possible latency so we take ourselves out of the equation.
2527 	 */
2528 	if (!ring_container->ring)
2529 		return;
2530 
2531 	/* If we didn't update within up to 1 - 2 jiffies we can assume
2532 	 * that either packets are coming in so slow there hasn't been
2533 	 * any work, or that there is so much work that NAPI is dealing
2534 	 * with interrupt moderation and we don't need to do anything.
2535 	 */
2536 	if (time_after(next_update, ring_container->next_update))
2537 		goto clear_counts;
2538 
2539 	packets = ring_container->total_packets;
2540 
2541 	/* We have no packets to actually measure against. This means
2542 	 * either one of the other queues on this vector is active or
2543 	 * we are a Tx queue doing TSO with too high of an interrupt rate.
2544 	 *
2545 	 * When this occurs just tick up our delay by the minimum value
2546 	 * and hope that this extra delay will prevent us from being called
2547 	 * without any work on our queue.
2548 	 */
2549 	if (!packets) {
2550 		itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2551 		if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2552 			itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2553 		itr += ring_container->itr & IXGBE_ITR_ADAPTIVE_LATENCY;
2554 		goto clear_counts;
2555 	}
2556 
2557 	bytes = ring_container->total_bytes;
2558 
2559 	/* If packets are less than 4 or bytes are less than 9000 assume
2560 	 * insufficient data to use bulk rate limiting approach. We are
2561 	 * likely latency driven.
2562 	 */
2563 	if (packets < 4 && bytes < 9000) {
2564 		itr = IXGBE_ITR_ADAPTIVE_LATENCY;
2565 		goto adjust_by_size;
2566 	}
2567 
2568 	/* Between 4 and 48 we can assume that our current interrupt delay
2569 	 * is only slightly too low. As such we should increase it by a small
2570 	 * fixed amount.
2571 	 */
2572 	if (packets < 48) {
2573 		itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2574 		if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2575 			itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2576 		goto clear_counts;
2577 	}
2578 
2579 	/* Between 48 and 96 is our "goldilocks" zone where we are working
2580 	 * out "just right". Just report that our current ITR is good for us.
2581 	 */
2582 	if (packets < 96) {
2583 		itr = q_vector->itr >> 2;
2584 		goto clear_counts;
2585 	}
2586 
2587 	/* If packet count is 96 or greater we are likely looking at a slight
2588 	 * overrun of the delay we want. Try halving our delay to see if that
2589 	 * will cut the number of packets in half per interrupt.
2590 	 */
2591 	if (packets < 256) {
2592 		itr = q_vector->itr >> 3;
2593 		if (itr < IXGBE_ITR_ADAPTIVE_MIN_USECS)
2594 			itr = IXGBE_ITR_ADAPTIVE_MIN_USECS;
2595 		goto clear_counts;
2596 	}
2597 
2598 	/* The paths below assume we are dealing with a bulk ITR since number
2599 	 * of packets is 256 or greater. We are just going to have to compute
2600 	 * a value and try to bring the count under control, though for smaller
2601 	 * packet sizes there isn't much we can do as NAPI polling will likely
2602 	 * be kicking in sooner rather than later.
2603 	 */
2604 	itr = IXGBE_ITR_ADAPTIVE_BULK;
2605 
2606 adjust_by_size:
2607 	/* If packet counts are 256 or greater we can assume we have a gross
2608 	 * overestimation of what the rate should be. Instead of trying to fine
2609 	 * tune it just use the formula below to try and dial in an exact value
2610 	 * give the current packet size of the frame.
2611 	 */
2612 	avg_wire_size = bytes / packets;
2613 
2614 	/* The following is a crude approximation of:
2615 	 *  wmem_default / (size + overhead) = desired_pkts_per_int
2616 	 *  rate / bits_per_byte / (size + ethernet overhead) = pkt_rate
2617 	 *  (desired_pkt_rate / pkt_rate) * usecs_per_sec = ITR value
2618 	 *
2619 	 * Assuming wmem_default is 212992 and overhead is 640 bytes per
2620 	 * packet, (256 skb, 64 headroom, 320 shared info), we can reduce the
2621 	 * formula down to
2622 	 *
2623 	 *  (170 * (size + 24)) / (size + 640) = ITR
2624 	 *
2625 	 * We first do some math on the packet size and then finally bitshift
2626 	 * by 8 after rounding up. We also have to account for PCIe link speed
2627 	 * difference as ITR scales based on this.
2628 	 */
2629 	if (avg_wire_size <= 60) {
2630 		/* Start at 50k ints/sec */
2631 		avg_wire_size = 5120;
2632 	} else if (avg_wire_size <= 316) {
2633 		/* 50K ints/sec to 16K ints/sec */
2634 		avg_wire_size *= 40;
2635 		avg_wire_size += 2720;
2636 	} else if (avg_wire_size <= 1084) {
2637 		/* 16K ints/sec to 9.2K ints/sec */
2638 		avg_wire_size *= 15;
2639 		avg_wire_size += 11452;
2640 	} else if (avg_wire_size < 1968) {
2641 		/* 9.2K ints/sec to 8K ints/sec */
2642 		avg_wire_size *= 5;
2643 		avg_wire_size += 22420;
2644 	} else {
2645 		/* plateau at a limit of 8K ints/sec */
2646 		avg_wire_size = 32256;
2647 	}
2648 
2649 	/* If we are in low latency mode half our delay which doubles the rate
2650 	 * to somewhere between 100K to 16K ints/sec
2651 	 */
2652 	if (itr & IXGBE_ITR_ADAPTIVE_LATENCY)
2653 		avg_wire_size >>= 1;
2654 
2655 	/* Resultant value is 256 times larger than it needs to be. This
2656 	 * gives us room to adjust the value as needed to either increase
2657 	 * or decrease the value based on link speeds of 10G, 2.5G, 1G, etc.
2658 	 *
2659 	 * Use addition as we have already recorded the new latency flag
2660 	 * for the ITR value.
2661 	 */
2662 	switch (q_vector->adapter->link_speed) {
2663 	case IXGBE_LINK_SPEED_10GB_FULL:
2664 	case IXGBE_LINK_SPEED_100_FULL:
2665 	default:
2666 		itr += DIV_ROUND_UP(avg_wire_size,
2667 				    IXGBE_ITR_ADAPTIVE_MIN_INC * 256) *
2668 		       IXGBE_ITR_ADAPTIVE_MIN_INC;
2669 		break;
2670 	case IXGBE_LINK_SPEED_2_5GB_FULL:
2671 	case IXGBE_LINK_SPEED_1GB_FULL:
2672 	case IXGBE_LINK_SPEED_10_FULL:
2673 		if (avg_wire_size > 8064)
2674 			avg_wire_size = 8064;
2675 		itr += DIV_ROUND_UP(avg_wire_size,
2676 				    IXGBE_ITR_ADAPTIVE_MIN_INC * 64) *
2677 		       IXGBE_ITR_ADAPTIVE_MIN_INC;
2678 		break;
2679 	}
2680 
2681 clear_counts:
2682 	/* write back value */
2683 	ring_container->itr = itr;
2684 
2685 	/* next update should occur within next jiffy */
2686 	ring_container->next_update = next_update + 1;
2687 
2688 	ring_container->total_bytes = 0;
2689 	ring_container->total_packets = 0;
2690 }
2691 
2692 /**
2693  * ixgbe_write_eitr - write EITR register in hardware specific way
2694  * @q_vector: structure containing interrupt and ring information
2695  *
2696  * This function is made to be called by ethtool and by the driver
2697  * when it needs to update EITR registers at runtime.  Hardware
2698  * specific quirks/differences are taken care of here.
2699  */
2700 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2701 {
2702 	struct ixgbe_adapter *adapter = q_vector->adapter;
2703 	struct ixgbe_hw *hw = &adapter->hw;
2704 	int v_idx = q_vector->v_idx;
2705 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2706 
2707 	switch (adapter->hw.mac.type) {
2708 	case ixgbe_mac_82598EB:
2709 		/* must write high and low 16 bits to reset counter */
2710 		itr_reg |= (itr_reg << 16);
2711 		break;
2712 	case ixgbe_mac_82599EB:
2713 	case ixgbe_mac_X540:
2714 	case ixgbe_mac_X550:
2715 	case ixgbe_mac_X550EM_x:
2716 	case ixgbe_mac_x550em_a:
2717 		/*
2718 		 * set the WDIS bit to not clear the timer bits and cause an
2719 		 * immediate assertion of the interrupt
2720 		 */
2721 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2722 		break;
2723 	default:
2724 		break;
2725 	}
2726 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2727 }
2728 
2729 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2730 {
2731 	u32 new_itr;
2732 
2733 	ixgbe_update_itr(q_vector, &q_vector->tx);
2734 	ixgbe_update_itr(q_vector, &q_vector->rx);
2735 
2736 	/* use the smallest value of new ITR delay calculations */
2737 	new_itr = min(q_vector->rx.itr, q_vector->tx.itr);
2738 
2739 	/* Clear latency flag if set, shift into correct position */
2740 	new_itr &= ~IXGBE_ITR_ADAPTIVE_LATENCY;
2741 	new_itr <<= 2;
2742 
2743 	if (new_itr != q_vector->itr) {
2744 		/* save the algorithm value here */
2745 		q_vector->itr = new_itr;
2746 
2747 		ixgbe_write_eitr(q_vector);
2748 	}
2749 }
2750 
2751 /**
2752  * ixgbe_check_overtemp_subtask - check for over temperature
2753  * @adapter: pointer to adapter
2754  **/
2755 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2756 {
2757 	struct ixgbe_hw *hw = &adapter->hw;
2758 	u32 eicr = adapter->interrupt_event;
2759 
2760 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2761 		return;
2762 
2763 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2764 		return;
2765 
2766 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2767 
2768 	switch (hw->device_id) {
2769 	case IXGBE_DEV_ID_82599_T3_LOM:
2770 		/*
2771 		 * Since the warning interrupt is for both ports
2772 		 * we don't have to check if:
2773 		 *  - This interrupt wasn't for our port.
2774 		 *  - We may have missed the interrupt so always have to
2775 		 *    check if we  got a LSC
2776 		 */
2777 		if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2778 		    !(eicr & IXGBE_EICR_LSC))
2779 			return;
2780 
2781 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2782 			u32 speed;
2783 			bool link_up = false;
2784 
2785 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2786 
2787 			if (link_up)
2788 				return;
2789 		}
2790 
2791 		/* Check if this is not due to overtemp */
2792 		if (!hw->phy.ops.check_overtemp(hw))
2793 			return;
2794 
2795 		break;
2796 	case IXGBE_DEV_ID_X550EM_A_1G_T:
2797 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
2798 		if (!hw->phy.ops.check_overtemp(hw))
2799 			return;
2800 		break;
2801 	default:
2802 		if (adapter->hw.mac.type >= ixgbe_mac_X540)
2803 			return;
2804 		if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
2805 			return;
2806 		break;
2807 	}
2808 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2809 
2810 	adapter->interrupt_event = 0;
2811 }
2812 
2813 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2814 {
2815 	struct ixgbe_hw *hw = &adapter->hw;
2816 
2817 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2818 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2819 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2820 		/* write to clear the interrupt */
2821 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2822 	}
2823 }
2824 
2825 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2826 {
2827 	struct ixgbe_hw *hw = &adapter->hw;
2828 
2829 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2830 		return;
2831 
2832 	switch (adapter->hw.mac.type) {
2833 	case ixgbe_mac_82599EB:
2834 		/*
2835 		 * Need to check link state so complete overtemp check
2836 		 * on service task
2837 		 */
2838 		if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
2839 		     (eicr & IXGBE_EICR_LSC)) &&
2840 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2841 			adapter->interrupt_event = eicr;
2842 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2843 			ixgbe_service_event_schedule(adapter);
2844 			return;
2845 		}
2846 		return;
2847 	case ixgbe_mac_x550em_a:
2848 		if (eicr & IXGBE_EICR_GPI_SDP0_X550EM_a) {
2849 			adapter->interrupt_event = eicr;
2850 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2851 			ixgbe_service_event_schedule(adapter);
2852 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC,
2853 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2854 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICR,
2855 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2856 		}
2857 		return;
2858 	case ixgbe_mac_X550:
2859 	case ixgbe_mac_X540:
2860 		if (!(eicr & IXGBE_EICR_TS))
2861 			return;
2862 		break;
2863 	default:
2864 		return;
2865 	}
2866 
2867 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2868 }
2869 
2870 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
2871 {
2872 	switch (hw->mac.type) {
2873 	case ixgbe_mac_82598EB:
2874 		if (hw->phy.type == ixgbe_phy_nl)
2875 			return true;
2876 		return false;
2877 	case ixgbe_mac_82599EB:
2878 	case ixgbe_mac_X550EM_x:
2879 	case ixgbe_mac_x550em_a:
2880 		switch (hw->mac.ops.get_media_type(hw)) {
2881 		case ixgbe_media_type_fiber:
2882 		case ixgbe_media_type_fiber_qsfp:
2883 			return true;
2884 		default:
2885 			return false;
2886 		}
2887 	default:
2888 		return false;
2889 	}
2890 }
2891 
2892 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2893 {
2894 	struct ixgbe_hw *hw = &adapter->hw;
2895 	u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
2896 
2897 	if (!ixgbe_is_sfp(hw))
2898 		return;
2899 
2900 	/* Later MAC's use different SDP */
2901 	if (hw->mac.type >= ixgbe_mac_X540)
2902 		eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2903 
2904 	if (eicr & eicr_mask) {
2905 		/* Clear the interrupt */
2906 		IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2907 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2908 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2909 			adapter->sfp_poll_time = 0;
2910 			ixgbe_service_event_schedule(adapter);
2911 		}
2912 	}
2913 
2914 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
2915 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2916 		/* Clear the interrupt */
2917 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2918 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2919 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2920 			ixgbe_service_event_schedule(adapter);
2921 		}
2922 	}
2923 }
2924 
2925 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2926 {
2927 	struct ixgbe_hw *hw = &adapter->hw;
2928 
2929 	adapter->lsc_int++;
2930 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2931 	adapter->link_check_timeout = jiffies;
2932 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2933 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2934 		IXGBE_WRITE_FLUSH(hw);
2935 		ixgbe_service_event_schedule(adapter);
2936 	}
2937 }
2938 
2939 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2940 					   u64 qmask)
2941 {
2942 	u32 mask;
2943 	struct ixgbe_hw *hw = &adapter->hw;
2944 
2945 	switch (hw->mac.type) {
2946 	case ixgbe_mac_82598EB:
2947 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2948 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2949 		break;
2950 	case ixgbe_mac_82599EB:
2951 	case ixgbe_mac_X540:
2952 	case ixgbe_mac_X550:
2953 	case ixgbe_mac_X550EM_x:
2954 	case ixgbe_mac_x550em_a:
2955 		mask = (qmask & 0xFFFFFFFF);
2956 		if (mask)
2957 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2958 		mask = (qmask >> 32);
2959 		if (mask)
2960 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2961 		break;
2962 	default:
2963 		break;
2964 	}
2965 	/* skip the flush */
2966 }
2967 
2968 /**
2969  * ixgbe_irq_enable - Enable default interrupt generation settings
2970  * @adapter: board private structure
2971  * @queues: enable irqs for queues
2972  * @flush: flush register write
2973  **/
2974 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2975 				    bool flush)
2976 {
2977 	struct ixgbe_hw *hw = &adapter->hw;
2978 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2979 
2980 	/* don't reenable LSC while waiting for link */
2981 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2982 		mask &= ~IXGBE_EIMS_LSC;
2983 
2984 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2985 		switch (adapter->hw.mac.type) {
2986 		case ixgbe_mac_82599EB:
2987 			mask |= IXGBE_EIMS_GPI_SDP0(hw);
2988 			break;
2989 		case ixgbe_mac_X540:
2990 		case ixgbe_mac_X550:
2991 		case ixgbe_mac_X550EM_x:
2992 		case ixgbe_mac_x550em_a:
2993 			mask |= IXGBE_EIMS_TS;
2994 			break;
2995 		default:
2996 			break;
2997 		}
2998 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
2999 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
3000 	switch (adapter->hw.mac.type) {
3001 	case ixgbe_mac_82599EB:
3002 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
3003 		mask |= IXGBE_EIMS_GPI_SDP2(hw);
3004 		fallthrough;
3005 	case ixgbe_mac_X540:
3006 	case ixgbe_mac_X550:
3007 	case ixgbe_mac_X550EM_x:
3008 	case ixgbe_mac_x550em_a:
3009 		if (adapter->hw.device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
3010 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
3011 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP_N)
3012 			mask |= IXGBE_EIMS_GPI_SDP0(&adapter->hw);
3013 		if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
3014 			mask |= IXGBE_EICR_GPI_SDP0_X540;
3015 		mask |= IXGBE_EIMS_ECC;
3016 		mask |= IXGBE_EIMS_MAILBOX;
3017 		break;
3018 	default:
3019 		break;
3020 	}
3021 
3022 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
3023 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
3024 		mask |= IXGBE_EIMS_FLOW_DIR;
3025 
3026 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
3027 	if (queues)
3028 		ixgbe_irq_enable_queues(adapter, ~0);
3029 	if (flush)
3030 		IXGBE_WRITE_FLUSH(&adapter->hw);
3031 }
3032 
3033 static irqreturn_t ixgbe_msix_other(int irq, void *data)
3034 {
3035 	struct ixgbe_adapter *adapter = data;
3036 	struct ixgbe_hw *hw = &adapter->hw;
3037 	u32 eicr;
3038 
3039 	/*
3040 	 * Workaround for Silicon errata.  Use clear-by-write instead
3041 	 * of clear-by-read.  Reading with EICS will return the
3042 	 * interrupt causes without clearing, which later be done
3043 	 * with the write to EICR.
3044 	 */
3045 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
3046 
3047 	/* The lower 16bits of the EICR register are for the queue interrupts
3048 	 * which should be masked here in order to not accidentally clear them if
3049 	 * the bits are high when ixgbe_msix_other is called. There is a race
3050 	 * condition otherwise which results in possible performance loss
3051 	 * especially if the ixgbe_msix_other interrupt is triggering
3052 	 * consistently (as it would when PPS is turned on for the X540 device)
3053 	 */
3054 	eicr &= 0xFFFF0000;
3055 
3056 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
3057 
3058 	if (eicr & IXGBE_EICR_LSC)
3059 		ixgbe_check_lsc(adapter);
3060 
3061 	if (eicr & IXGBE_EICR_MAILBOX)
3062 		ixgbe_msg_task(adapter);
3063 
3064 	switch (hw->mac.type) {
3065 	case ixgbe_mac_82599EB:
3066 	case ixgbe_mac_X540:
3067 	case ixgbe_mac_X550:
3068 	case ixgbe_mac_X550EM_x:
3069 	case ixgbe_mac_x550em_a:
3070 		if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
3071 		    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
3072 			adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
3073 			ixgbe_service_event_schedule(adapter);
3074 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
3075 					IXGBE_EICR_GPI_SDP0_X540);
3076 		}
3077 		if (eicr & IXGBE_EICR_ECC) {
3078 			e_info(link, "Received ECC Err, initiating reset\n");
3079 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3080 			ixgbe_service_event_schedule(adapter);
3081 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3082 		}
3083 		/* Handle Flow Director Full threshold interrupt */
3084 		if (eicr & IXGBE_EICR_FLOW_DIR) {
3085 			int reinit_count = 0;
3086 			int i;
3087 			for (i = 0; i < adapter->num_tx_queues; i++) {
3088 				struct ixgbe_ring *ring = adapter->tx_ring[i];
3089 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
3090 						       &ring->state))
3091 					reinit_count++;
3092 			}
3093 			if (reinit_count) {
3094 				/* no more flow director interrupts until after init */
3095 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
3096 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
3097 				ixgbe_service_event_schedule(adapter);
3098 			}
3099 		}
3100 		ixgbe_check_sfp_event(adapter, eicr);
3101 		ixgbe_check_overtemp_event(adapter, eicr);
3102 		break;
3103 	default:
3104 		break;
3105 	}
3106 
3107 	ixgbe_check_fan_failure(adapter, eicr);
3108 
3109 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3110 		ixgbe_ptp_check_pps_event(adapter);
3111 
3112 	/* re-enable the original interrupt state, no lsc, no queues */
3113 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3114 		ixgbe_irq_enable(adapter, false, false);
3115 
3116 	return IRQ_HANDLED;
3117 }
3118 
3119 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
3120 {
3121 	struct ixgbe_q_vector *q_vector = data;
3122 
3123 	/* EIAM disabled interrupts (on this vector) for us */
3124 
3125 	if (q_vector->rx.ring || q_vector->tx.ring)
3126 		napi_schedule_irqoff(&q_vector->napi);
3127 
3128 	return IRQ_HANDLED;
3129 }
3130 
3131 /**
3132  * ixgbe_poll - NAPI Rx polling callback
3133  * @napi: structure for representing this polling device
3134  * @budget: how many packets driver is allowed to clean
3135  *
3136  * This function is used for legacy and MSI, NAPI mode
3137  **/
3138 int ixgbe_poll(struct napi_struct *napi, int budget)
3139 {
3140 	struct ixgbe_q_vector *q_vector =
3141 				container_of(napi, struct ixgbe_q_vector, napi);
3142 	struct ixgbe_adapter *adapter = q_vector->adapter;
3143 	struct ixgbe_ring *ring;
3144 	int per_ring_budget, work_done = 0;
3145 	bool clean_complete = true;
3146 
3147 #ifdef CONFIG_IXGBE_DCA
3148 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
3149 		ixgbe_update_dca(q_vector);
3150 #endif
3151 
3152 	ixgbe_for_each_ring(ring, q_vector->tx) {
3153 		bool wd = ring->xsk_pool ?
3154 			  ixgbe_clean_xdp_tx_irq(q_vector, ring, budget) :
3155 			  ixgbe_clean_tx_irq(q_vector, ring, budget);
3156 
3157 		if (!wd)
3158 			clean_complete = false;
3159 	}
3160 
3161 	/* Exit if we are called by netpoll */
3162 	if (budget <= 0)
3163 		return budget;
3164 
3165 	/* attempt to distribute budget to each queue fairly, but don't allow
3166 	 * the budget to go below 1 because we'll exit polling */
3167 	if (q_vector->rx.count > 1)
3168 		per_ring_budget = max(budget/q_vector->rx.count, 1);
3169 	else
3170 		per_ring_budget = budget;
3171 
3172 	ixgbe_for_each_ring(ring, q_vector->rx) {
3173 		int cleaned = ring->xsk_pool ?
3174 			      ixgbe_clean_rx_irq_zc(q_vector, ring,
3175 						    per_ring_budget) :
3176 			      ixgbe_clean_rx_irq(q_vector, ring,
3177 						 per_ring_budget);
3178 
3179 		work_done += cleaned;
3180 		if (cleaned >= per_ring_budget)
3181 			clean_complete = false;
3182 	}
3183 
3184 	/* If all work not completed, return budget and keep polling */
3185 	if (!clean_complete)
3186 		return budget;
3187 
3188 	/* all work done, exit the polling mode */
3189 	if (likely(napi_complete_done(napi, work_done))) {
3190 		if (adapter->rx_itr_setting & 1)
3191 			ixgbe_set_itr(q_vector);
3192 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
3193 			ixgbe_irq_enable_queues(adapter,
3194 						BIT_ULL(q_vector->v_idx));
3195 	}
3196 
3197 	return min(work_done, budget - 1);
3198 }
3199 
3200 /**
3201  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
3202  * @adapter: board private structure
3203  *
3204  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
3205  * interrupts from the kernel.
3206  **/
3207 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
3208 {
3209 	struct net_device *netdev = adapter->netdev;
3210 	unsigned int ri = 0, ti = 0;
3211 	int vector, err;
3212 
3213 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3214 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3215 		struct msix_entry *entry = &adapter->msix_entries[vector];
3216 
3217 		if (q_vector->tx.ring && q_vector->rx.ring) {
3218 			snprintf(q_vector->name, sizeof(q_vector->name),
3219 				 "%s-TxRx-%u", netdev->name, ri++);
3220 			ti++;
3221 		} else if (q_vector->rx.ring) {
3222 			snprintf(q_vector->name, sizeof(q_vector->name),
3223 				 "%s-rx-%u", netdev->name, ri++);
3224 		} else if (q_vector->tx.ring) {
3225 			snprintf(q_vector->name, sizeof(q_vector->name),
3226 				 "%s-tx-%u", netdev->name, ti++);
3227 		} else {
3228 			/* skip this unused q_vector */
3229 			continue;
3230 		}
3231 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
3232 				  q_vector->name, q_vector);
3233 		if (err) {
3234 			e_err(probe, "request_irq failed for MSIX interrupt "
3235 			      "Error: %d\n", err);
3236 			goto free_queue_irqs;
3237 		}
3238 		/* If Flow Director is enabled, set interrupt affinity */
3239 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3240 			/* assign the mask for this irq */
3241 			irq_update_affinity_hint(entry->vector,
3242 						 &q_vector->affinity_mask);
3243 		}
3244 	}
3245 
3246 	err = request_irq(adapter->msix_entries[vector].vector,
3247 			  ixgbe_msix_other, 0, netdev->name, adapter);
3248 	if (err) {
3249 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
3250 		goto free_queue_irqs;
3251 	}
3252 
3253 	return 0;
3254 
3255 free_queue_irqs:
3256 	while (vector) {
3257 		vector--;
3258 		irq_update_affinity_hint(adapter->msix_entries[vector].vector,
3259 					 NULL);
3260 		free_irq(adapter->msix_entries[vector].vector,
3261 			 adapter->q_vector[vector]);
3262 	}
3263 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
3264 	pci_disable_msix(adapter->pdev);
3265 	kfree(adapter->msix_entries);
3266 	adapter->msix_entries = NULL;
3267 	return err;
3268 }
3269 
3270 /**
3271  * ixgbe_intr - legacy mode Interrupt Handler
3272  * @irq: interrupt number
3273  * @data: pointer to a network interface device structure
3274  **/
3275 static irqreturn_t ixgbe_intr(int irq, void *data)
3276 {
3277 	struct ixgbe_adapter *adapter = data;
3278 	struct ixgbe_hw *hw = &adapter->hw;
3279 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3280 	u32 eicr;
3281 
3282 	/*
3283 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
3284 	 * before the read of EICR.
3285 	 */
3286 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
3287 
3288 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
3289 	 * therefore no explicit interrupt disable is necessary */
3290 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3291 	if (!eicr) {
3292 		/*
3293 		 * shared interrupt alert!
3294 		 * make sure interrupts are enabled because the read will
3295 		 * have disabled interrupts due to EIAM
3296 		 * finish the workaround of silicon errata on 82598.  Unmask
3297 		 * the interrupt that we masked before the EICR read.
3298 		 */
3299 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
3300 			ixgbe_irq_enable(adapter, true, true);
3301 		return IRQ_NONE;	/* Not our interrupt */
3302 	}
3303 
3304 	if (eicr & IXGBE_EICR_LSC)
3305 		ixgbe_check_lsc(adapter);
3306 
3307 	switch (hw->mac.type) {
3308 	case ixgbe_mac_82599EB:
3309 		ixgbe_check_sfp_event(adapter, eicr);
3310 		fallthrough;
3311 	case ixgbe_mac_X540:
3312 	case ixgbe_mac_X550:
3313 	case ixgbe_mac_X550EM_x:
3314 	case ixgbe_mac_x550em_a:
3315 		if (eicr & IXGBE_EICR_ECC) {
3316 			e_info(link, "Received ECC Err, initiating reset\n");
3317 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3318 			ixgbe_service_event_schedule(adapter);
3319 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3320 		}
3321 		ixgbe_check_overtemp_event(adapter, eicr);
3322 		break;
3323 	default:
3324 		break;
3325 	}
3326 
3327 	ixgbe_check_fan_failure(adapter, eicr);
3328 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3329 		ixgbe_ptp_check_pps_event(adapter);
3330 
3331 	/* would disable interrupts here but EIAM disabled it */
3332 	napi_schedule_irqoff(&q_vector->napi);
3333 
3334 	/*
3335 	 * re-enable link(maybe) and non-queue interrupts, no flush.
3336 	 * ixgbe_poll will re-enable the queue interrupts
3337 	 */
3338 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3339 		ixgbe_irq_enable(adapter, false, false);
3340 
3341 	return IRQ_HANDLED;
3342 }
3343 
3344 /**
3345  * ixgbe_request_irq - initialize interrupts
3346  * @adapter: board private structure
3347  *
3348  * Attempts to configure interrupts using the best available
3349  * capabilities of the hardware and kernel.
3350  **/
3351 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
3352 {
3353 	struct net_device *netdev = adapter->netdev;
3354 	int err;
3355 
3356 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
3357 		err = ixgbe_request_msix_irqs(adapter);
3358 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
3359 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
3360 				  netdev->name, adapter);
3361 	else
3362 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
3363 				  netdev->name, adapter);
3364 
3365 	if (err)
3366 		e_err(probe, "request_irq failed, Error %d\n", err);
3367 
3368 	return err;
3369 }
3370 
3371 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
3372 {
3373 	int vector;
3374 
3375 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
3376 		free_irq(adapter->pdev->irq, adapter);
3377 		return;
3378 	}
3379 
3380 	if (!adapter->msix_entries)
3381 		return;
3382 
3383 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3384 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3385 		struct msix_entry *entry = &adapter->msix_entries[vector];
3386 
3387 		/* free only the irqs that were actually requested */
3388 		if (!q_vector->rx.ring && !q_vector->tx.ring)
3389 			continue;
3390 
3391 		/* clear the affinity_mask in the IRQ descriptor */
3392 		irq_update_affinity_hint(entry->vector, NULL);
3393 
3394 		free_irq(entry->vector, q_vector);
3395 	}
3396 
3397 	free_irq(adapter->msix_entries[vector].vector, adapter);
3398 }
3399 
3400 /**
3401  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3402  * @adapter: board private structure
3403  **/
3404 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3405 {
3406 	switch (adapter->hw.mac.type) {
3407 	case ixgbe_mac_82598EB:
3408 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3409 		break;
3410 	case ixgbe_mac_82599EB:
3411 	case ixgbe_mac_X540:
3412 	case ixgbe_mac_X550:
3413 	case ixgbe_mac_X550EM_x:
3414 	case ixgbe_mac_x550em_a:
3415 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3416 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3417 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3418 		break;
3419 	default:
3420 		break;
3421 	}
3422 	IXGBE_WRITE_FLUSH(&adapter->hw);
3423 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3424 		int vector;
3425 
3426 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
3427 			synchronize_irq(adapter->msix_entries[vector].vector);
3428 
3429 		synchronize_irq(adapter->msix_entries[vector++].vector);
3430 	} else {
3431 		synchronize_irq(adapter->pdev->irq);
3432 	}
3433 }
3434 
3435 /**
3436  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3437  * @adapter: board private structure
3438  *
3439  **/
3440 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3441 {
3442 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3443 
3444 	ixgbe_write_eitr(q_vector);
3445 
3446 	ixgbe_set_ivar(adapter, 0, 0, 0);
3447 	ixgbe_set_ivar(adapter, 1, 0, 0);
3448 
3449 	e_info(hw, "Legacy interrupt IVAR setup done\n");
3450 }
3451 
3452 /**
3453  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3454  * @adapter: board private structure
3455  * @ring: structure containing ring specific data
3456  *
3457  * Configure the Tx descriptor ring after a reset.
3458  **/
3459 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3460 			     struct ixgbe_ring *ring)
3461 {
3462 	struct ixgbe_hw *hw = &adapter->hw;
3463 	u64 tdba = ring->dma;
3464 	int wait_loop = 10;
3465 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
3466 	u8 reg_idx = ring->reg_idx;
3467 
3468 	ring->xsk_pool = NULL;
3469 	if (ring_is_xdp(ring))
3470 		ring->xsk_pool = ixgbe_xsk_pool(adapter, ring);
3471 
3472 	/* disable queue to avoid issues while updating state */
3473 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3474 	IXGBE_WRITE_FLUSH(hw);
3475 
3476 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3477 			(tdba & DMA_BIT_MASK(32)));
3478 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3479 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3480 			ring->count * sizeof(union ixgbe_adv_tx_desc));
3481 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3482 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3483 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3484 
3485 	/*
3486 	 * set WTHRESH to encourage burst writeback, it should not be set
3487 	 * higher than 1 when:
3488 	 * - ITR is 0 as it could cause false TX hangs
3489 	 * - ITR is set to > 100k int/sec and BQL is enabled
3490 	 *
3491 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3492 	 * to or less than the number of on chip descriptors, which is
3493 	 * currently 40.
3494 	 */
3495 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3496 		txdctl |= 1u << 16;	/* WTHRESH = 1 */
3497 	else
3498 		txdctl |= 8u << 16;	/* WTHRESH = 8 */
3499 
3500 	/*
3501 	 * Setting PTHRESH to 32 both improves performance
3502 	 * and avoids a TX hang with DFP enabled
3503 	 */
3504 	txdctl |= (1u << 8) |	/* HTHRESH = 1 */
3505 		   32;		/* PTHRESH = 32 */
3506 
3507 	/* reinitialize flowdirector state */
3508 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3509 		ring->atr_sample_rate = adapter->atr_sample_rate;
3510 		ring->atr_count = 0;
3511 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3512 	} else {
3513 		ring->atr_sample_rate = 0;
3514 	}
3515 
3516 	/* initialize XPS */
3517 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3518 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3519 
3520 		if (q_vector)
3521 			netif_set_xps_queue(ring->netdev,
3522 					    &q_vector->affinity_mask,
3523 					    ring->queue_index);
3524 	}
3525 
3526 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3527 
3528 	/* reinitialize tx_buffer_info */
3529 	memset(ring->tx_buffer_info, 0,
3530 	       sizeof(struct ixgbe_tx_buffer) * ring->count);
3531 
3532 	/* enable queue */
3533 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3534 
3535 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3536 	if (hw->mac.type == ixgbe_mac_82598EB &&
3537 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3538 		return;
3539 
3540 	/* poll to verify queue is enabled */
3541 	do {
3542 		usleep_range(1000, 2000);
3543 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3544 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3545 	if (!wait_loop)
3546 		hw_dbg(hw, "Could not enable Tx Queue %d\n", reg_idx);
3547 }
3548 
3549 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3550 {
3551 	struct ixgbe_hw *hw = &adapter->hw;
3552 	u32 rttdcs, mtqc;
3553 	u8 tcs = adapter->hw_tcs;
3554 
3555 	if (hw->mac.type == ixgbe_mac_82598EB)
3556 		return;
3557 
3558 	/* disable the arbiter while setting MTQC */
3559 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3560 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3561 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3562 
3563 	/* set transmit pool layout */
3564 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3565 		mtqc = IXGBE_MTQC_VT_ENA;
3566 		if (tcs > 4)
3567 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3568 		else if (tcs > 1)
3569 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3570 		else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3571 			 IXGBE_82599_VMDQ_4Q_MASK)
3572 			mtqc |= IXGBE_MTQC_32VF;
3573 		else
3574 			mtqc |= IXGBE_MTQC_64VF;
3575 	} else {
3576 		if (tcs > 4) {
3577 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3578 		} else if (tcs > 1) {
3579 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3580 		} else {
3581 			u8 max_txq = adapter->num_tx_queues +
3582 				adapter->num_xdp_queues;
3583 			if (max_txq > 63)
3584 				mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3585 			else
3586 				mtqc = IXGBE_MTQC_64Q_1PB;
3587 		}
3588 	}
3589 
3590 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3591 
3592 	/* Enable Security TX Buffer IFG for multiple pb */
3593 	if (tcs) {
3594 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3595 		sectx |= IXGBE_SECTX_DCB;
3596 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3597 	}
3598 
3599 	/* re-enable the arbiter */
3600 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3601 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3602 }
3603 
3604 /**
3605  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3606  * @adapter: board private structure
3607  *
3608  * Configure the Tx unit of the MAC after a reset.
3609  **/
3610 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3611 {
3612 	struct ixgbe_hw *hw = &adapter->hw;
3613 	u32 dmatxctl;
3614 	u32 i;
3615 
3616 	ixgbe_setup_mtqc(adapter);
3617 
3618 	if (hw->mac.type != ixgbe_mac_82598EB) {
3619 		/* DMATXCTL.EN must be before Tx queues are enabled */
3620 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3621 		dmatxctl |= IXGBE_DMATXCTL_TE;
3622 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3623 	}
3624 
3625 	/* Setup the HW Tx Head and Tail descriptor pointers */
3626 	for (i = 0; i < adapter->num_tx_queues; i++)
3627 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3628 	for (i = 0; i < adapter->num_xdp_queues; i++)
3629 		ixgbe_configure_tx_ring(adapter, adapter->xdp_ring[i]);
3630 }
3631 
3632 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3633 				 struct ixgbe_ring *ring)
3634 {
3635 	struct ixgbe_hw *hw = &adapter->hw;
3636 	u8 reg_idx = ring->reg_idx;
3637 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3638 
3639 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3640 
3641 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3642 }
3643 
3644 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3645 				  struct ixgbe_ring *ring)
3646 {
3647 	struct ixgbe_hw *hw = &adapter->hw;
3648 	u8 reg_idx = ring->reg_idx;
3649 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3650 
3651 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3652 
3653 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3654 }
3655 
3656 #ifdef CONFIG_IXGBE_DCB
3657 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3658 #else
3659 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3660 #endif
3661 {
3662 	int i;
3663 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3664 
3665 	if (adapter->ixgbe_ieee_pfc)
3666 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3667 
3668 	/*
3669 	 * We should set the drop enable bit if:
3670 	 *  SR-IOV is enabled
3671 	 *   or
3672 	 *  Number of Rx queues > 1 and flow control is disabled
3673 	 *
3674 	 *  This allows us to avoid head of line blocking for security
3675 	 *  and performance reasons.
3676 	 */
3677 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3678 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3679 		for (i = 0; i < adapter->num_rx_queues; i++)
3680 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3681 	} else {
3682 		for (i = 0; i < adapter->num_rx_queues; i++)
3683 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3684 	}
3685 }
3686 
3687 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3688 
3689 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3690 				   struct ixgbe_ring *rx_ring)
3691 {
3692 	struct ixgbe_hw *hw = &adapter->hw;
3693 	u32 srrctl;
3694 	u8 reg_idx = rx_ring->reg_idx;
3695 
3696 	if (hw->mac.type == ixgbe_mac_82598EB) {
3697 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3698 
3699 		/*
3700 		 * if VMDq is not active we must program one srrctl register
3701 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3702 		 */
3703 		reg_idx &= mask;
3704 	}
3705 
3706 	/* configure header buffer length, needed for RSC */
3707 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3708 
3709 	/* configure the packet buffer length */
3710 	if (rx_ring->xsk_pool) {
3711 		u32 xsk_buf_len = xsk_pool_get_rx_frame_size(rx_ring->xsk_pool);
3712 
3713 		/* If the MAC support setting RXDCTL.RLPML, the
3714 		 * SRRCTL[n].BSIZEPKT is set to PAGE_SIZE and
3715 		 * RXDCTL.RLPML is set to the actual UMEM buffer
3716 		 * size. If not, then we are stuck with a 1k buffer
3717 		 * size resolution. In this case frames larger than
3718 		 * the UMEM buffer size viewed in a 1k resolution will
3719 		 * be dropped.
3720 		 */
3721 		if (hw->mac.type != ixgbe_mac_82599EB)
3722 			srrctl |= PAGE_SIZE >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3723 		else
3724 			srrctl |= xsk_buf_len >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3725 	} else if (test_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state)) {
3726 		srrctl |= IXGBE_RXBUFFER_3K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3727 	} else {
3728 		srrctl |= IXGBE_RXBUFFER_2K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3729 	}
3730 
3731 	/* configure descriptor type */
3732 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3733 
3734 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3735 }
3736 
3737 /**
3738  * ixgbe_rss_indir_tbl_entries - Return RSS indirection table entries
3739  * @adapter: device handle
3740  *
3741  *  - 82598/82599/X540:     128
3742  *  - X550(non-SRIOV mode): 512
3743  *  - X550(SRIOV mode):     64
3744  */
3745 u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3746 {
3747 	if (adapter->hw.mac.type < ixgbe_mac_X550)
3748 		return 128;
3749 	else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3750 		return 64;
3751 	else
3752 		return 512;
3753 }
3754 
3755 /**
3756  * ixgbe_store_key - Write the RSS key to HW
3757  * @adapter: device handle
3758  *
3759  * Write the RSS key stored in adapter.rss_key to HW.
3760  */
3761 void ixgbe_store_key(struct ixgbe_adapter *adapter)
3762 {
3763 	struct ixgbe_hw *hw = &adapter->hw;
3764 	int i;
3765 
3766 	for (i = 0; i < 10; i++)
3767 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3768 }
3769 
3770 /**
3771  * ixgbe_init_rss_key - Initialize adapter RSS key
3772  * @adapter: device handle
3773  *
3774  * Allocates and initializes the RSS key if it is not allocated.
3775  **/
3776 static inline int ixgbe_init_rss_key(struct ixgbe_adapter *adapter)
3777 {
3778 	u32 *rss_key;
3779 
3780 	if (!adapter->rss_key) {
3781 		rss_key = kzalloc(IXGBE_RSS_KEY_SIZE, GFP_KERNEL);
3782 		if (unlikely(!rss_key))
3783 			return -ENOMEM;
3784 
3785 		netdev_rss_key_fill(rss_key, IXGBE_RSS_KEY_SIZE);
3786 		adapter->rss_key = rss_key;
3787 	}
3788 
3789 	return 0;
3790 }
3791 
3792 /**
3793  * ixgbe_store_reta - Write the RETA table to HW
3794  * @adapter: device handle
3795  *
3796  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3797  */
3798 void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3799 {
3800 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3801 	struct ixgbe_hw *hw = &adapter->hw;
3802 	u32 reta = 0;
3803 	u32 indices_multi;
3804 	u8 *indir_tbl = adapter->rss_indir_tbl;
3805 
3806 	/* Fill out the redirection table as follows:
3807 	 *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3808 	 *    indices.
3809 	 *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3810 	 *  - X550:       8 bit wide entries containing 6 bit RSS index
3811 	 */
3812 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3813 		indices_multi = 0x11;
3814 	else
3815 		indices_multi = 0x1;
3816 
3817 	/* Write redirection table to HW */
3818 	for (i = 0; i < reta_entries; i++) {
3819 		reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3820 		if ((i & 3) == 3) {
3821 			if (i < 128)
3822 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3823 			else
3824 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3825 						reta);
3826 			reta = 0;
3827 		}
3828 	}
3829 }
3830 
3831 /**
3832  * ixgbe_store_vfreta - Write the RETA table to HW (x550 devices in SRIOV mode)
3833  * @adapter: device handle
3834  *
3835  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3836  */
3837 static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3838 {
3839 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3840 	struct ixgbe_hw *hw = &adapter->hw;
3841 	u32 vfreta = 0;
3842 
3843 	/* Write redirection table to HW */
3844 	for (i = 0; i < reta_entries; i++) {
3845 		u16 pool = adapter->num_rx_pools;
3846 
3847 		vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3848 		if ((i & 3) != 3)
3849 			continue;
3850 
3851 		while (pool--)
3852 			IXGBE_WRITE_REG(hw,
3853 					IXGBE_PFVFRETA(i >> 2, VMDQ_P(pool)),
3854 					vfreta);
3855 		vfreta = 0;
3856 	}
3857 }
3858 
3859 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3860 {
3861 	u32 i, j;
3862 	u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3863 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3864 
3865 	/* Program table for at least 4 queues w/ SR-IOV so that VFs can
3866 	 * make full use of any rings they may have.  We will use the
3867 	 * PSRTYPE register to control how many rings we use within the PF.
3868 	 */
3869 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 4))
3870 		rss_i = 4;
3871 
3872 	/* Fill out hash function seeds */
3873 	ixgbe_store_key(adapter);
3874 
3875 	/* Fill out redirection table */
3876 	memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3877 
3878 	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3879 		if (j == rss_i)
3880 			j = 0;
3881 
3882 		adapter->rss_indir_tbl[i] = j;
3883 	}
3884 
3885 	ixgbe_store_reta(adapter);
3886 }
3887 
3888 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3889 {
3890 	struct ixgbe_hw *hw = &adapter->hw;
3891 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3892 	int i, j;
3893 
3894 	/* Fill out hash function seeds */
3895 	for (i = 0; i < 10; i++) {
3896 		u16 pool = adapter->num_rx_pools;
3897 
3898 		while (pool--)
3899 			IXGBE_WRITE_REG(hw,
3900 					IXGBE_PFVFRSSRK(i, VMDQ_P(pool)),
3901 					*(adapter->rss_key + i));
3902 	}
3903 
3904 	/* Fill out the redirection table */
3905 	for (i = 0, j = 0; i < 64; i++, j++) {
3906 		if (j == rss_i)
3907 			j = 0;
3908 
3909 		adapter->rss_indir_tbl[i] = j;
3910 	}
3911 
3912 	ixgbe_store_vfreta(adapter);
3913 }
3914 
3915 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3916 {
3917 	struct ixgbe_hw *hw = &adapter->hw;
3918 	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3919 	u32 rxcsum;
3920 
3921 	/* Disable indicating checksum in descriptor, enables RSS hash */
3922 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3923 	rxcsum |= IXGBE_RXCSUM_PCSD;
3924 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3925 
3926 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3927 		if (adapter->ring_feature[RING_F_RSS].mask)
3928 			mrqc = IXGBE_MRQC_RSSEN;
3929 	} else {
3930 		u8 tcs = adapter->hw_tcs;
3931 
3932 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3933 			if (tcs > 4)
3934 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3935 			else if (tcs > 1)
3936 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3937 			else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3938 				 IXGBE_82599_VMDQ_4Q_MASK)
3939 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3940 			else
3941 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3942 
3943 			/* Enable L3/L4 for Tx Switched packets only for X550,
3944 			 * older devices do not support this feature
3945 			 */
3946 			if (hw->mac.type >= ixgbe_mac_X550)
3947 				mrqc |= IXGBE_MRQC_L3L4TXSWEN;
3948 		} else {
3949 			if (tcs > 4)
3950 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3951 			else if (tcs > 1)
3952 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3953 			else
3954 				mrqc = IXGBE_MRQC_RSSEN;
3955 		}
3956 	}
3957 
3958 	/* Perform hash on these packet types */
3959 	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3960 		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3961 		     IXGBE_MRQC_RSS_FIELD_IPV6 |
3962 		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3963 
3964 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3965 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3966 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3967 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3968 
3969 	if ((hw->mac.type >= ixgbe_mac_X550) &&
3970 	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
3971 		u16 pool = adapter->num_rx_pools;
3972 
3973 		/* Enable VF RSS mode */
3974 		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
3975 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3976 
3977 		/* Setup RSS through the VF registers */
3978 		ixgbe_setup_vfreta(adapter);
3979 		vfmrqc = IXGBE_MRQC_RSSEN;
3980 		vfmrqc |= rss_field;
3981 
3982 		while (pool--)
3983 			IXGBE_WRITE_REG(hw,
3984 					IXGBE_PFVFMRQC(VMDQ_P(pool)),
3985 					vfmrqc);
3986 	} else {
3987 		ixgbe_setup_reta(adapter);
3988 		mrqc |= rss_field;
3989 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3990 	}
3991 }
3992 
3993 /**
3994  * ixgbe_configure_rscctl - enable RSC for the indicated ring
3995  * @adapter: address of board private structure
3996  * @ring: structure containing ring specific data
3997  **/
3998 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
3999 				   struct ixgbe_ring *ring)
4000 {
4001 	struct ixgbe_hw *hw = &adapter->hw;
4002 	u32 rscctrl;
4003 	u8 reg_idx = ring->reg_idx;
4004 
4005 	if (!ring_is_rsc_enabled(ring))
4006 		return;
4007 
4008 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
4009 	rscctrl |= IXGBE_RSCCTL_RSCEN;
4010 	/*
4011 	 * we must limit the number of descriptors so that the
4012 	 * total size of max desc * buf_len is not greater
4013 	 * than 65536
4014 	 */
4015 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
4016 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
4017 }
4018 
4019 #define IXGBE_MAX_RX_DESC_POLL 10
4020 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
4021 				       struct ixgbe_ring *ring)
4022 {
4023 	struct ixgbe_hw *hw = &adapter->hw;
4024 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
4025 	u32 rxdctl;
4026 	u8 reg_idx = ring->reg_idx;
4027 
4028 	if (ixgbe_removed(hw->hw_addr))
4029 		return;
4030 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
4031 	if (hw->mac.type == ixgbe_mac_82598EB &&
4032 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
4033 		return;
4034 
4035 	do {
4036 		usleep_range(1000, 2000);
4037 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4038 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
4039 
4040 	if (!wait_loop) {
4041 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
4042 		      "the polling period\n", reg_idx);
4043 	}
4044 }
4045 
4046 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
4047 			     struct ixgbe_ring *ring)
4048 {
4049 	struct ixgbe_hw *hw = &adapter->hw;
4050 	union ixgbe_adv_rx_desc *rx_desc;
4051 	u64 rdba = ring->dma;
4052 	u32 rxdctl;
4053 	u8 reg_idx = ring->reg_idx;
4054 
4055 	xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq);
4056 	ring->xsk_pool = ixgbe_xsk_pool(adapter, ring);
4057 	if (ring->xsk_pool) {
4058 		WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
4059 						   MEM_TYPE_XSK_BUFF_POOL,
4060 						   NULL));
4061 		xsk_pool_set_rxq_info(ring->xsk_pool, &ring->xdp_rxq);
4062 	} else {
4063 		WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
4064 						   MEM_TYPE_PAGE_SHARED, NULL));
4065 	}
4066 
4067 	/* disable queue to avoid use of these values while updating state */
4068 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4069 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
4070 
4071 	/* write value back with RXDCTL.ENABLE bit cleared */
4072 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4073 	IXGBE_WRITE_FLUSH(hw);
4074 
4075 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
4076 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
4077 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
4078 			ring->count * sizeof(union ixgbe_adv_rx_desc));
4079 	/* Force flushing of IXGBE_RDLEN to prevent MDD */
4080 	IXGBE_WRITE_FLUSH(hw);
4081 
4082 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
4083 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
4084 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
4085 
4086 	ixgbe_configure_srrctl(adapter, ring);
4087 	ixgbe_configure_rscctl(adapter, ring);
4088 
4089 	if (hw->mac.type == ixgbe_mac_82598EB) {
4090 		/*
4091 		 * enable cache line friendly hardware writes:
4092 		 * PTHRESH=32 descriptors (half the internal cache),
4093 		 * this also removes ugly rx_no_buffer_count increment
4094 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
4095 		 * WTHRESH=8 burst writeback up to two cache lines
4096 		 */
4097 		rxdctl &= ~0x3FFFFF;
4098 		rxdctl |=  0x080420;
4099 #if (PAGE_SIZE < 8192)
4100 	/* RXDCTL.RLPML does not work on 82599 */
4101 	} else if (hw->mac.type != ixgbe_mac_82599EB) {
4102 		rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
4103 			    IXGBE_RXDCTL_RLPML_EN);
4104 
4105 		/* Limit the maximum frame size so we don't overrun the skb.
4106 		 * This can happen in SRIOV mode when the MTU of the VF is
4107 		 * higher than the MTU of the PF.
4108 		 */
4109 		if (ring_uses_build_skb(ring) &&
4110 		    !test_bit(__IXGBE_RX_3K_BUFFER, &ring->state))
4111 			rxdctl |= IXGBE_MAX_2K_FRAME_BUILD_SKB |
4112 				  IXGBE_RXDCTL_RLPML_EN;
4113 #endif
4114 	}
4115 
4116 	ring->rx_offset = ixgbe_rx_offset(ring);
4117 
4118 	if (ring->xsk_pool && hw->mac.type != ixgbe_mac_82599EB) {
4119 		u32 xsk_buf_len = xsk_pool_get_rx_frame_size(ring->xsk_pool);
4120 
4121 		rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
4122 			    IXGBE_RXDCTL_RLPML_EN);
4123 		rxdctl |= xsk_buf_len | IXGBE_RXDCTL_RLPML_EN;
4124 
4125 		ring->rx_buf_len = xsk_buf_len;
4126 	}
4127 
4128 	/* initialize rx_buffer_info */
4129 	memset(ring->rx_buffer_info, 0,
4130 	       sizeof(struct ixgbe_rx_buffer) * ring->count);
4131 
4132 	/* initialize Rx descriptor 0 */
4133 	rx_desc = IXGBE_RX_DESC(ring, 0);
4134 	rx_desc->wb.upper.length = 0;
4135 
4136 	/* enable receive descriptor ring */
4137 	rxdctl |= IXGBE_RXDCTL_ENABLE;
4138 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4139 
4140 	ixgbe_rx_desc_queue_enable(adapter, ring);
4141 	if (ring->xsk_pool)
4142 		ixgbe_alloc_rx_buffers_zc(ring, ixgbe_desc_unused(ring));
4143 	else
4144 		ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
4145 }
4146 
4147 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
4148 {
4149 	struct ixgbe_hw *hw = &adapter->hw;
4150 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
4151 	u16 pool = adapter->num_rx_pools;
4152 
4153 	/* PSRTYPE must be initialized in non 82598 adapters */
4154 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4155 		      IXGBE_PSRTYPE_UDPHDR |
4156 		      IXGBE_PSRTYPE_IPV4HDR |
4157 		      IXGBE_PSRTYPE_L2HDR |
4158 		      IXGBE_PSRTYPE_IPV6HDR;
4159 
4160 	if (hw->mac.type == ixgbe_mac_82598EB)
4161 		return;
4162 
4163 	if (rss_i > 3)
4164 		psrtype |= 2u << 29;
4165 	else if (rss_i > 1)
4166 		psrtype |= 1u << 29;
4167 
4168 	while (pool--)
4169 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4170 }
4171 
4172 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
4173 {
4174 	struct ixgbe_hw *hw = &adapter->hw;
4175 	u16 pool = adapter->num_rx_pools;
4176 	u32 reg_offset, vf_shift, vmolr;
4177 	u32 gcr_ext, vmdctl;
4178 	int i;
4179 
4180 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
4181 		return;
4182 
4183 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
4184 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
4185 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
4186 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
4187 	vmdctl |= IXGBE_VT_CTL_REPLEN;
4188 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
4189 
4190 	/* accept untagged packets until a vlan tag is
4191 	 * specifically set for the VMDQ queue/pool
4192 	 */
4193 	vmolr = IXGBE_VMOLR_AUPE;
4194 	while (pool--)
4195 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(pool)), vmolr);
4196 
4197 	vf_shift = VMDQ_P(0) % 32;
4198 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
4199 
4200 	/* Enable only the PF's pool for Tx/Rx */
4201 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), GENMASK(31, vf_shift));
4202 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
4203 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), GENMASK(31, vf_shift));
4204 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
4205 	if (adapter->bridge_mode == BRIDGE_MODE_VEB)
4206 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
4207 
4208 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
4209 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
4210 
4211 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
4212 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4213 
4214 	/*
4215 	 * Set up VF register offsets for selected VT Mode,
4216 	 * i.e. 32 or 64 VFs for SR-IOV
4217 	 */
4218 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4219 	case IXGBE_82599_VMDQ_8Q_MASK:
4220 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
4221 		break;
4222 	case IXGBE_82599_VMDQ_4Q_MASK:
4223 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
4224 		break;
4225 	default:
4226 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
4227 		break;
4228 	}
4229 
4230 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
4231 
4232 	for (i = 0; i < adapter->num_vfs; i++) {
4233 		/* configure spoof checking */
4234 		ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i,
4235 					  adapter->vfinfo[i].spoofchk_enabled);
4236 
4237 		/* Enable/Disable RSS query feature  */
4238 		ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
4239 					  adapter->vfinfo[i].rss_query_enabled);
4240 	}
4241 }
4242 
4243 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
4244 {
4245 	struct ixgbe_hw *hw = &adapter->hw;
4246 	struct net_device *netdev = adapter->netdev;
4247 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4248 	struct ixgbe_ring *rx_ring;
4249 	int i;
4250 	u32 mhadd, hlreg0;
4251 
4252 #ifdef IXGBE_FCOE
4253 	/* adjust max frame to be able to do baby jumbo for FCoE */
4254 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
4255 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
4256 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4257 
4258 #endif /* IXGBE_FCOE */
4259 
4260 	/* adjust max frame to be at least the size of a standard frame */
4261 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
4262 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
4263 
4264 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
4265 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
4266 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
4267 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
4268 
4269 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
4270 	}
4271 
4272 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4273 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
4274 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4275 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4276 
4277 	/*
4278 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4279 	 * the Base and Length of the Rx Descriptor Ring
4280 	 */
4281 	for (i = 0; i < adapter->num_rx_queues; i++) {
4282 		rx_ring = adapter->rx_ring[i];
4283 
4284 		clear_ring_rsc_enabled(rx_ring);
4285 		clear_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4286 		clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4287 
4288 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4289 			set_ring_rsc_enabled(rx_ring);
4290 
4291 		if (test_bit(__IXGBE_RX_FCOE, &rx_ring->state))
4292 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4293 
4294 		if (adapter->flags2 & IXGBE_FLAG2_RX_LEGACY)
4295 			continue;
4296 
4297 		set_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4298 
4299 #if (PAGE_SIZE < 8192)
4300 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4301 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4302 
4303 		if (IXGBE_2K_TOO_SMALL_WITH_PADDING ||
4304 		    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
4305 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4306 #endif
4307 	}
4308 }
4309 
4310 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
4311 {
4312 	struct ixgbe_hw *hw = &adapter->hw;
4313 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
4314 
4315 	switch (hw->mac.type) {
4316 	case ixgbe_mac_82598EB:
4317 		/*
4318 		 * For VMDq support of different descriptor types or
4319 		 * buffer sizes through the use of multiple SRRCTL
4320 		 * registers, RDRXCTL.MVMEN must be set to 1
4321 		 *
4322 		 * also, the manual doesn't mention it clearly but DCA hints
4323 		 * will only use queue 0's tags unless this bit is set.  Side
4324 		 * effects of setting this bit are only that SRRCTL must be
4325 		 * fully programmed [0..15]
4326 		 */
4327 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
4328 		break;
4329 	case ixgbe_mac_X550:
4330 	case ixgbe_mac_X550EM_x:
4331 	case ixgbe_mac_x550em_a:
4332 		if (adapter->num_vfs)
4333 			rdrxctl |= IXGBE_RDRXCTL_PSP;
4334 		fallthrough;
4335 	case ixgbe_mac_82599EB:
4336 	case ixgbe_mac_X540:
4337 		/* Disable RSC for ACK packets */
4338 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
4339 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
4340 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
4341 		/* hardware requires some bits to be set by default */
4342 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
4343 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
4344 		break;
4345 	default:
4346 		/* We should do nothing since we don't know this hardware */
4347 		return;
4348 	}
4349 
4350 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
4351 }
4352 
4353 /**
4354  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
4355  * @adapter: board private structure
4356  *
4357  * Configure the Rx unit of the MAC after a reset.
4358  **/
4359 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
4360 {
4361 	struct ixgbe_hw *hw = &adapter->hw;
4362 	int i;
4363 	u32 rxctrl, rfctl;
4364 
4365 	/* disable receives while setting up the descriptors */
4366 	hw->mac.ops.disable_rx(hw);
4367 
4368 	ixgbe_setup_psrtype(adapter);
4369 	ixgbe_setup_rdrxctl(adapter);
4370 
4371 	/* RSC Setup */
4372 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
4373 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
4374 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
4375 		rfctl |= IXGBE_RFCTL_RSC_DIS;
4376 
4377 	/* disable NFS filtering */
4378 	rfctl |= (IXGBE_RFCTL_NFSW_DIS | IXGBE_RFCTL_NFSR_DIS);
4379 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
4380 
4381 	/* Program registers for the distribution of queues */
4382 	ixgbe_setup_mrqc(adapter);
4383 
4384 	/* set_rx_buffer_len must be called before ring initialization */
4385 	ixgbe_set_rx_buffer_len(adapter);
4386 
4387 	/*
4388 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4389 	 * the Base and Length of the Rx Descriptor Ring
4390 	 */
4391 	for (i = 0; i < adapter->num_rx_queues; i++)
4392 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
4393 
4394 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4395 	/* disable drop enable for 82598 parts */
4396 	if (hw->mac.type == ixgbe_mac_82598EB)
4397 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
4398 
4399 	/* enable all receives */
4400 	rxctrl |= IXGBE_RXCTRL_RXEN;
4401 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
4402 }
4403 
4404 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
4405 				 __be16 proto, u16 vid)
4406 {
4407 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4408 	struct ixgbe_hw *hw = &adapter->hw;
4409 
4410 	/* add VID to filter table */
4411 	if (!vid || !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4412 		hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true, !!vid);
4413 
4414 	set_bit(vid, adapter->active_vlans);
4415 
4416 	return 0;
4417 }
4418 
4419 static int ixgbe_find_vlvf_entry(struct ixgbe_hw *hw, u32 vlan)
4420 {
4421 	u32 vlvf;
4422 	int idx;
4423 
4424 	/* short cut the special case */
4425 	if (vlan == 0)
4426 		return 0;
4427 
4428 	/* Search for the vlan id in the VLVF entries */
4429 	for (idx = IXGBE_VLVF_ENTRIES; --idx;) {
4430 		vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(idx));
4431 		if ((vlvf & VLAN_VID_MASK) == vlan)
4432 			break;
4433 	}
4434 
4435 	return idx;
4436 }
4437 
4438 void ixgbe_update_pf_promisc_vlvf(struct ixgbe_adapter *adapter, u32 vid)
4439 {
4440 	struct ixgbe_hw *hw = &adapter->hw;
4441 	u32 bits, word;
4442 	int idx;
4443 
4444 	idx = ixgbe_find_vlvf_entry(hw, vid);
4445 	if (!idx)
4446 		return;
4447 
4448 	/* See if any other pools are set for this VLAN filter
4449 	 * entry other than the PF.
4450 	 */
4451 	word = idx * 2 + (VMDQ_P(0) / 32);
4452 	bits = ~BIT(VMDQ_P(0) % 32);
4453 	bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4454 
4455 	/* Disable the filter so this falls into the default pool. */
4456 	if (!bits && !IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1))) {
4457 		if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4458 			IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), 0);
4459 		IXGBE_WRITE_REG(hw, IXGBE_VLVF(idx), 0);
4460 	}
4461 }
4462 
4463 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
4464 				  __be16 proto, u16 vid)
4465 {
4466 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4467 	struct ixgbe_hw *hw = &adapter->hw;
4468 
4469 	/* remove VID from filter table */
4470 	if (vid && !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4471 		hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), false, true);
4472 
4473 	clear_bit(vid, adapter->active_vlans);
4474 
4475 	return 0;
4476 }
4477 
4478 /**
4479  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
4480  * @adapter: driver data
4481  */
4482 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
4483 {
4484 	struct ixgbe_hw *hw = &adapter->hw;
4485 	u32 vlnctrl;
4486 	int i, j;
4487 
4488 	switch (hw->mac.type) {
4489 	case ixgbe_mac_82598EB:
4490 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4491 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
4492 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4493 		break;
4494 	case ixgbe_mac_82599EB:
4495 	case ixgbe_mac_X540:
4496 	case ixgbe_mac_X550:
4497 	case ixgbe_mac_X550EM_x:
4498 	case ixgbe_mac_x550em_a:
4499 		for (i = 0; i < adapter->num_rx_queues; i++) {
4500 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4501 
4502 			if (!netif_is_ixgbe(ring->netdev))
4503 				continue;
4504 
4505 			j = ring->reg_idx;
4506 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4507 			vlnctrl &= ~IXGBE_RXDCTL_VME;
4508 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4509 		}
4510 		break;
4511 	default:
4512 		break;
4513 	}
4514 }
4515 
4516 /**
4517  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
4518  * @adapter: driver data
4519  */
4520 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
4521 {
4522 	struct ixgbe_hw *hw = &adapter->hw;
4523 	u32 vlnctrl;
4524 	int i, j;
4525 
4526 	switch (hw->mac.type) {
4527 	case ixgbe_mac_82598EB:
4528 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4529 		vlnctrl |= IXGBE_VLNCTRL_VME;
4530 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4531 		break;
4532 	case ixgbe_mac_82599EB:
4533 	case ixgbe_mac_X540:
4534 	case ixgbe_mac_X550:
4535 	case ixgbe_mac_X550EM_x:
4536 	case ixgbe_mac_x550em_a:
4537 		for (i = 0; i < adapter->num_rx_queues; i++) {
4538 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4539 
4540 			if (!netif_is_ixgbe(ring->netdev))
4541 				continue;
4542 
4543 			j = ring->reg_idx;
4544 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4545 			vlnctrl |= IXGBE_RXDCTL_VME;
4546 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4547 		}
4548 		break;
4549 	default:
4550 		break;
4551 	}
4552 }
4553 
4554 static void ixgbe_vlan_promisc_enable(struct ixgbe_adapter *adapter)
4555 {
4556 	struct ixgbe_hw *hw = &adapter->hw;
4557 	u32 vlnctrl, i;
4558 
4559 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4560 
4561 	if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) {
4562 	/* For VMDq and SR-IOV we must leave VLAN filtering enabled */
4563 		vlnctrl |= IXGBE_VLNCTRL_VFE;
4564 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4565 	} else {
4566 		vlnctrl &= ~IXGBE_VLNCTRL_VFE;
4567 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4568 		return;
4569 	}
4570 
4571 	/* Nothing to do for 82598 */
4572 	if (hw->mac.type == ixgbe_mac_82598EB)
4573 		return;
4574 
4575 	/* We are already in VLAN promisc, nothing to do */
4576 	if (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC)
4577 		return;
4578 
4579 	/* Set flag so we don't redo unnecessary work */
4580 	adapter->flags2 |= IXGBE_FLAG2_VLAN_PROMISC;
4581 
4582 	/* Add PF to all active pools */
4583 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4584 		u32 reg_offset = IXGBE_VLVFB(i * 2 + VMDQ_P(0) / 32);
4585 		u32 vlvfb = IXGBE_READ_REG(hw, reg_offset);
4586 
4587 		vlvfb |= BIT(VMDQ_P(0) % 32);
4588 		IXGBE_WRITE_REG(hw, reg_offset, vlvfb);
4589 	}
4590 
4591 	/* Set all bits in the VLAN filter table array */
4592 	for (i = hw->mac.vft_size; i--;)
4593 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), ~0U);
4594 }
4595 
4596 #define VFTA_BLOCK_SIZE 8
4597 static void ixgbe_scrub_vfta(struct ixgbe_adapter *adapter, u32 vfta_offset)
4598 {
4599 	struct ixgbe_hw *hw = &adapter->hw;
4600 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4601 	u32 vid_start = vfta_offset * 32;
4602 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4603 	u32 i, vid, word, bits;
4604 
4605 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4606 		u32 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
4607 
4608 		/* pull VLAN ID from VLVF */
4609 		vid = vlvf & VLAN_VID_MASK;
4610 
4611 		/* only concern outselves with a certain range */
4612 		if (vid < vid_start || vid >= vid_end)
4613 			continue;
4614 
4615 		if (vlvf) {
4616 			/* record VLAN ID in VFTA */
4617 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4618 
4619 			/* if PF is part of this then continue */
4620 			if (test_bit(vid, adapter->active_vlans))
4621 				continue;
4622 		}
4623 
4624 		/* remove PF from the pool */
4625 		word = i * 2 + VMDQ_P(0) / 32;
4626 		bits = ~BIT(VMDQ_P(0) % 32);
4627 		bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4628 		IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), bits);
4629 	}
4630 
4631 	/* extract values from active_vlans and write back to VFTA */
4632 	for (i = VFTA_BLOCK_SIZE; i--;) {
4633 		vid = (vfta_offset + i) * 32;
4634 		word = vid / BITS_PER_LONG;
4635 		bits = vid % BITS_PER_LONG;
4636 
4637 		vfta[i] |= adapter->active_vlans[word] >> bits;
4638 
4639 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(vfta_offset + i), vfta[i]);
4640 	}
4641 }
4642 
4643 static void ixgbe_vlan_promisc_disable(struct ixgbe_adapter *adapter)
4644 {
4645 	struct ixgbe_hw *hw = &adapter->hw;
4646 	u32 vlnctrl, i;
4647 
4648 	/* Set VLAN filtering to enabled */
4649 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4650 	vlnctrl |= IXGBE_VLNCTRL_VFE;
4651 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4652 
4653 	if (!(adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) ||
4654 	    hw->mac.type == ixgbe_mac_82598EB)
4655 		return;
4656 
4657 	/* We are not in VLAN promisc, nothing to do */
4658 	if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4659 		return;
4660 
4661 	/* Set flag so we don't redo unnecessary work */
4662 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4663 
4664 	for (i = 0; i < hw->mac.vft_size; i += VFTA_BLOCK_SIZE)
4665 		ixgbe_scrub_vfta(adapter, i);
4666 }
4667 
4668 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
4669 {
4670 	u16 vid = 1;
4671 
4672 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
4673 
4674 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
4675 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
4676 }
4677 
4678 /**
4679  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4680  * @netdev: network interface device structure
4681  *
4682  * Writes multicast address list to the MTA hash table.
4683  * Returns: -ENOMEM on failure
4684  *                0 on no addresses written
4685  *                X on writing X addresses to MTA
4686  **/
4687 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
4688 {
4689 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4690 	struct ixgbe_hw *hw = &adapter->hw;
4691 
4692 	if (!netif_running(netdev))
4693 		return 0;
4694 
4695 	if (hw->mac.ops.update_mc_addr_list)
4696 		hw->mac.ops.update_mc_addr_list(hw, netdev);
4697 	else
4698 		return -ENOMEM;
4699 
4700 #ifdef CONFIG_PCI_IOV
4701 	ixgbe_restore_vf_multicasts(adapter);
4702 #endif
4703 
4704 	return netdev_mc_count(netdev);
4705 }
4706 
4707 #ifdef CONFIG_PCI_IOV
4708 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4709 {
4710 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4711 	struct ixgbe_hw *hw = &adapter->hw;
4712 	int i;
4713 
4714 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4715 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4716 
4717 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4718 			hw->mac.ops.set_rar(hw, i,
4719 					    mac_table->addr,
4720 					    mac_table->pool,
4721 					    IXGBE_RAH_AV);
4722 		else
4723 			hw->mac.ops.clear_rar(hw, i);
4724 	}
4725 }
4726 
4727 #endif
4728 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4729 {
4730 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4731 	struct ixgbe_hw *hw = &adapter->hw;
4732 	int i;
4733 
4734 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4735 		if (!(mac_table->state & IXGBE_MAC_STATE_MODIFIED))
4736 			continue;
4737 
4738 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4739 
4740 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4741 			hw->mac.ops.set_rar(hw, i,
4742 					    mac_table->addr,
4743 					    mac_table->pool,
4744 					    IXGBE_RAH_AV);
4745 		else
4746 			hw->mac.ops.clear_rar(hw, i);
4747 	}
4748 }
4749 
4750 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4751 {
4752 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4753 	struct ixgbe_hw *hw = &adapter->hw;
4754 	int i;
4755 
4756 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4757 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4758 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4759 	}
4760 
4761 	ixgbe_sync_mac_table(adapter);
4762 }
4763 
4764 static int ixgbe_available_rars(struct ixgbe_adapter *adapter, u16 pool)
4765 {
4766 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4767 	struct ixgbe_hw *hw = &adapter->hw;
4768 	int i, count = 0;
4769 
4770 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4771 		/* do not count default RAR as available */
4772 		if (mac_table->state & IXGBE_MAC_STATE_DEFAULT)
4773 			continue;
4774 
4775 		/* only count unused and addresses that belong to us */
4776 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE) {
4777 			if (mac_table->pool != pool)
4778 				continue;
4779 		}
4780 
4781 		count++;
4782 	}
4783 
4784 	return count;
4785 }
4786 
4787 /* this function destroys the first RAR entry */
4788 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter)
4789 {
4790 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4791 	struct ixgbe_hw *hw = &adapter->hw;
4792 
4793 	memcpy(&mac_table->addr, hw->mac.addr, ETH_ALEN);
4794 	mac_table->pool = VMDQ_P(0);
4795 
4796 	mac_table->state = IXGBE_MAC_STATE_DEFAULT | IXGBE_MAC_STATE_IN_USE;
4797 
4798 	hw->mac.ops.set_rar(hw, 0, mac_table->addr, mac_table->pool,
4799 			    IXGBE_RAH_AV);
4800 }
4801 
4802 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter,
4803 			 const u8 *addr, u16 pool)
4804 {
4805 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4806 	struct ixgbe_hw *hw = &adapter->hw;
4807 	int i;
4808 
4809 	if (is_zero_ether_addr(addr))
4810 		return -EINVAL;
4811 
4812 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4813 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4814 			continue;
4815 
4816 		ether_addr_copy(mac_table->addr, addr);
4817 		mac_table->pool = pool;
4818 
4819 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED |
4820 				    IXGBE_MAC_STATE_IN_USE;
4821 
4822 		ixgbe_sync_mac_table(adapter);
4823 
4824 		return i;
4825 	}
4826 
4827 	return -ENOMEM;
4828 }
4829 
4830 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter,
4831 			 const u8 *addr, u16 pool)
4832 {
4833 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4834 	struct ixgbe_hw *hw = &adapter->hw;
4835 	int i;
4836 
4837 	if (is_zero_ether_addr(addr))
4838 		return -EINVAL;
4839 
4840 	/* search table for addr, if found clear IN_USE flag and sync */
4841 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4842 		/* we can only delete an entry if it is in use */
4843 		if (!(mac_table->state & IXGBE_MAC_STATE_IN_USE))
4844 			continue;
4845 		/* we only care about entries that belong to the given pool */
4846 		if (mac_table->pool != pool)
4847 			continue;
4848 		/* we only care about a specific MAC address */
4849 		if (!ether_addr_equal(addr, mac_table->addr))
4850 			continue;
4851 
4852 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4853 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4854 
4855 		ixgbe_sync_mac_table(adapter);
4856 
4857 		return 0;
4858 	}
4859 
4860 	return -ENOMEM;
4861 }
4862 
4863 static int ixgbe_uc_sync(struct net_device *netdev, const unsigned char *addr)
4864 {
4865 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4866 	int ret;
4867 
4868 	ret = ixgbe_add_mac_filter(adapter, addr, VMDQ_P(0));
4869 
4870 	return min_t(int, ret, 0);
4871 }
4872 
4873 static int ixgbe_uc_unsync(struct net_device *netdev, const unsigned char *addr)
4874 {
4875 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4876 
4877 	ixgbe_del_mac_filter(adapter, addr, VMDQ_P(0));
4878 
4879 	return 0;
4880 }
4881 
4882 /**
4883  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4884  * @netdev: network interface device structure
4885  *
4886  * The set_rx_method entry point is called whenever the unicast/multicast
4887  * address list or the network interface flags are updated.  This routine is
4888  * responsible for configuring the hardware for proper unicast, multicast and
4889  * promiscuous mode.
4890  **/
4891 void ixgbe_set_rx_mode(struct net_device *netdev)
4892 {
4893 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4894 	struct ixgbe_hw *hw = &adapter->hw;
4895 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4896 	netdev_features_t features = netdev->features;
4897 	int count;
4898 
4899 	/* Check for Promiscuous and All Multicast modes */
4900 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4901 
4902 	/* set all bits that we expect to always be set */
4903 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4904 	fctrl |= IXGBE_FCTRL_BAM;
4905 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4906 	fctrl |= IXGBE_FCTRL_PMCF;
4907 
4908 	/* clear the bits we are changing the status of */
4909 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4910 	if (netdev->flags & IFF_PROMISC) {
4911 		hw->addr_ctrl.user_set_promisc = true;
4912 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4913 		vmolr |= IXGBE_VMOLR_MPE;
4914 		features &= ~NETIF_F_HW_VLAN_CTAG_FILTER;
4915 	} else {
4916 		if (netdev->flags & IFF_ALLMULTI) {
4917 			fctrl |= IXGBE_FCTRL_MPE;
4918 			vmolr |= IXGBE_VMOLR_MPE;
4919 		}
4920 		hw->addr_ctrl.user_set_promisc = false;
4921 	}
4922 
4923 	/*
4924 	 * Write addresses to available RAR registers, if there is not
4925 	 * sufficient space to store all the addresses then enable
4926 	 * unicast promiscuous mode
4927 	 */
4928 	if (__dev_uc_sync(netdev, ixgbe_uc_sync, ixgbe_uc_unsync)) {
4929 		fctrl |= IXGBE_FCTRL_UPE;
4930 		vmolr |= IXGBE_VMOLR_ROPE;
4931 	}
4932 
4933 	/* Write addresses to the MTA, if the attempt fails
4934 	 * then we should just turn on promiscuous mode so
4935 	 * that we can at least receive multicast traffic
4936 	 */
4937 	count = ixgbe_write_mc_addr_list(netdev);
4938 	if (count < 0) {
4939 		fctrl |= IXGBE_FCTRL_MPE;
4940 		vmolr |= IXGBE_VMOLR_MPE;
4941 	} else if (count) {
4942 		vmolr |= IXGBE_VMOLR_ROMPE;
4943 	}
4944 
4945 	if (hw->mac.type != ixgbe_mac_82598EB) {
4946 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4947 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4948 			   IXGBE_VMOLR_ROPE);
4949 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4950 	}
4951 
4952 	/* This is useful for sniffing bad packets. */
4953 	if (features & NETIF_F_RXALL) {
4954 		/* UPE and MPE will be handled by normal PROMISC logic
4955 		 * in e1000e_set_rx_mode */
4956 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4957 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4958 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4959 
4960 		fctrl &= ~(IXGBE_FCTRL_DPF);
4961 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
4962 	}
4963 
4964 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4965 
4966 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
4967 		ixgbe_vlan_strip_enable(adapter);
4968 	else
4969 		ixgbe_vlan_strip_disable(adapter);
4970 
4971 	if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
4972 		ixgbe_vlan_promisc_disable(adapter);
4973 	else
4974 		ixgbe_vlan_promisc_enable(adapter);
4975 }
4976 
4977 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4978 {
4979 	int q_idx;
4980 
4981 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
4982 		napi_enable(&adapter->q_vector[q_idx]->napi);
4983 }
4984 
4985 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
4986 {
4987 	int q_idx;
4988 
4989 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
4990 		napi_disable(&adapter->q_vector[q_idx]->napi);
4991 }
4992 
4993 static int ixgbe_udp_tunnel_sync(struct net_device *dev, unsigned int table)
4994 {
4995 	struct ixgbe_adapter *adapter = netdev_priv(dev);
4996 	struct ixgbe_hw *hw = &adapter->hw;
4997 	struct udp_tunnel_info ti;
4998 
4999 	udp_tunnel_nic_get_port(dev, table, 0, &ti);
5000 	if (ti.type == UDP_TUNNEL_TYPE_VXLAN)
5001 		adapter->vxlan_port = ti.port;
5002 	else
5003 		adapter->geneve_port = ti.port;
5004 
5005 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL,
5006 			ntohs(adapter->vxlan_port) |
5007 			ntohs(adapter->geneve_port) <<
5008 				IXGBE_VXLANCTRL_GENEVE_UDPPORT_SHIFT);
5009 	return 0;
5010 }
5011 
5012 static const struct udp_tunnel_nic_info ixgbe_udp_tunnels_x550 = {
5013 	.sync_table	= ixgbe_udp_tunnel_sync,
5014 	.flags		= UDP_TUNNEL_NIC_INFO_IPV4_ONLY,
5015 	.tables		= {
5016 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN,  },
5017 	},
5018 };
5019 
5020 static const struct udp_tunnel_nic_info ixgbe_udp_tunnels_x550em_a = {
5021 	.sync_table	= ixgbe_udp_tunnel_sync,
5022 	.flags		= UDP_TUNNEL_NIC_INFO_IPV4_ONLY,
5023 	.tables		= {
5024 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN,  },
5025 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_GENEVE, },
5026 	},
5027 };
5028 
5029 #ifdef CONFIG_IXGBE_DCB
5030 /**
5031  * ixgbe_configure_dcb - Configure DCB hardware
5032  * @adapter: ixgbe adapter struct
5033  *
5034  * This is called by the driver on open to configure the DCB hardware.
5035  * This is also called by the gennetlink interface when reconfiguring
5036  * the DCB state.
5037  */
5038 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
5039 {
5040 	struct ixgbe_hw *hw = &adapter->hw;
5041 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
5042 
5043 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
5044 		if (hw->mac.type == ixgbe_mac_82598EB)
5045 			netif_set_tso_max_size(adapter->netdev, 65536);
5046 		return;
5047 	}
5048 
5049 	if (hw->mac.type == ixgbe_mac_82598EB)
5050 		netif_set_tso_max_size(adapter->netdev, 32768);
5051 
5052 #ifdef IXGBE_FCOE
5053 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
5054 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
5055 #endif
5056 
5057 	/* reconfigure the hardware */
5058 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
5059 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5060 						DCB_TX_CONFIG);
5061 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5062 						DCB_RX_CONFIG);
5063 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
5064 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
5065 		ixgbe_dcb_hw_ets(&adapter->hw,
5066 				 adapter->ixgbe_ieee_ets,
5067 				 max_frame);
5068 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
5069 					adapter->ixgbe_ieee_pfc->pfc_en,
5070 					adapter->ixgbe_ieee_ets->prio_tc);
5071 	}
5072 
5073 	/* Enable RSS Hash per TC */
5074 	if (hw->mac.type != ixgbe_mac_82598EB) {
5075 		u32 msb = 0;
5076 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
5077 
5078 		while (rss_i) {
5079 			msb++;
5080 			rss_i >>= 1;
5081 		}
5082 
5083 		/* write msb to all 8 TCs in one write */
5084 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
5085 	}
5086 }
5087 #endif
5088 
5089 /* Additional bittime to account for IXGBE framing */
5090 #define IXGBE_ETH_FRAMING 20
5091 
5092 /**
5093  * ixgbe_hpbthresh - calculate high water mark for flow control
5094  *
5095  * @adapter: board private structure to calculate for
5096  * @pb: packet buffer to calculate
5097  */
5098 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
5099 {
5100 	struct ixgbe_hw *hw = &adapter->hw;
5101 	struct net_device *dev = adapter->netdev;
5102 	int link, tc, kb, marker;
5103 	u32 dv_id, rx_pba;
5104 
5105 	/* Calculate max LAN frame size */
5106 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
5107 
5108 #ifdef IXGBE_FCOE
5109 	/* FCoE traffic class uses FCOE jumbo frames */
5110 	if ((dev->features & NETIF_F_FCOE_MTU) &&
5111 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5112 	    (pb == ixgbe_fcoe_get_tc(adapter)))
5113 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5114 #endif
5115 
5116 	/* Calculate delay value for device */
5117 	switch (hw->mac.type) {
5118 	case ixgbe_mac_X540:
5119 	case ixgbe_mac_X550:
5120 	case ixgbe_mac_X550EM_x:
5121 	case ixgbe_mac_x550em_a:
5122 		dv_id = IXGBE_DV_X540(link, tc);
5123 		break;
5124 	default:
5125 		dv_id = IXGBE_DV(link, tc);
5126 		break;
5127 	}
5128 
5129 	/* Loopback switch introduces additional latency */
5130 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5131 		dv_id += IXGBE_B2BT(tc);
5132 
5133 	/* Delay value is calculated in bit times convert to KB */
5134 	kb = IXGBE_BT2KB(dv_id);
5135 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
5136 
5137 	marker = rx_pba - kb;
5138 
5139 	/* It is possible that the packet buffer is not large enough
5140 	 * to provide required headroom. In this case throw an error
5141 	 * to user and a do the best we can.
5142 	 */
5143 	if (marker < 0) {
5144 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
5145 			    "headroom to support flow control."
5146 			    "Decrease MTU or number of traffic classes\n", pb);
5147 		marker = tc + 1;
5148 	}
5149 
5150 	return marker;
5151 }
5152 
5153 /**
5154  * ixgbe_lpbthresh - calculate low water mark for flow control
5155  *
5156  * @adapter: board private structure to calculate for
5157  * @pb: packet buffer to calculate
5158  */
5159 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
5160 {
5161 	struct ixgbe_hw *hw = &adapter->hw;
5162 	struct net_device *dev = adapter->netdev;
5163 	int tc;
5164 	u32 dv_id;
5165 
5166 	/* Calculate max LAN frame size */
5167 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
5168 
5169 #ifdef IXGBE_FCOE
5170 	/* FCoE traffic class uses FCOE jumbo frames */
5171 	if ((dev->features & NETIF_F_FCOE_MTU) &&
5172 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5173 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
5174 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5175 #endif
5176 
5177 	/* Calculate delay value for device */
5178 	switch (hw->mac.type) {
5179 	case ixgbe_mac_X540:
5180 	case ixgbe_mac_X550:
5181 	case ixgbe_mac_X550EM_x:
5182 	case ixgbe_mac_x550em_a:
5183 		dv_id = IXGBE_LOW_DV_X540(tc);
5184 		break;
5185 	default:
5186 		dv_id = IXGBE_LOW_DV(tc);
5187 		break;
5188 	}
5189 
5190 	/* Delay value is calculated in bit times convert to KB */
5191 	return IXGBE_BT2KB(dv_id);
5192 }
5193 
5194 /*
5195  * ixgbe_pbthresh_setup - calculate and setup high low water marks
5196  */
5197 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
5198 {
5199 	struct ixgbe_hw *hw = &adapter->hw;
5200 	int num_tc = adapter->hw_tcs;
5201 	int i;
5202 
5203 	if (!num_tc)
5204 		num_tc = 1;
5205 
5206 	for (i = 0; i < num_tc; i++) {
5207 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
5208 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
5209 
5210 		/* Low water marks must not be larger than high water marks */
5211 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
5212 			hw->fc.low_water[i] = 0;
5213 	}
5214 
5215 	for (; i < MAX_TRAFFIC_CLASS; i++)
5216 		hw->fc.high_water[i] = 0;
5217 }
5218 
5219 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
5220 {
5221 	struct ixgbe_hw *hw = &adapter->hw;
5222 	int hdrm;
5223 	u8 tc = adapter->hw_tcs;
5224 
5225 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
5226 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
5227 		hdrm = 32 << adapter->fdir_pballoc;
5228 	else
5229 		hdrm = 0;
5230 
5231 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
5232 	ixgbe_pbthresh_setup(adapter);
5233 }
5234 
5235 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
5236 {
5237 	struct ixgbe_hw *hw = &adapter->hw;
5238 	struct hlist_node *node2;
5239 	struct ixgbe_fdir_filter *filter;
5240 	u8 queue;
5241 
5242 	spin_lock(&adapter->fdir_perfect_lock);
5243 
5244 	if (!hlist_empty(&adapter->fdir_filter_list))
5245 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
5246 
5247 	hlist_for_each_entry_safe(filter, node2,
5248 				  &adapter->fdir_filter_list, fdir_node) {
5249 		if (filter->action == IXGBE_FDIR_DROP_QUEUE) {
5250 			queue = IXGBE_FDIR_DROP_QUEUE;
5251 		} else {
5252 			u32 ring = ethtool_get_flow_spec_ring(filter->action);
5253 			u8 vf = ethtool_get_flow_spec_ring_vf(filter->action);
5254 
5255 			if (!vf && (ring >= adapter->num_rx_queues)) {
5256 				e_err(drv, "FDIR restore failed without VF, ring: %u\n",
5257 				      ring);
5258 				continue;
5259 			} else if (vf &&
5260 				   ((vf > adapter->num_vfs) ||
5261 				     ring >= adapter->num_rx_queues_per_pool)) {
5262 				e_err(drv, "FDIR restore failed with VF, vf: %hhu, ring: %u\n",
5263 				      vf, ring);
5264 				continue;
5265 			}
5266 
5267 			/* Map the ring onto the absolute queue index */
5268 			if (!vf)
5269 				queue = adapter->rx_ring[ring]->reg_idx;
5270 			else
5271 				queue = ((vf - 1) *
5272 					adapter->num_rx_queues_per_pool) + ring;
5273 		}
5274 
5275 		ixgbe_fdir_write_perfect_filter_82599(hw,
5276 				&filter->filter, filter->sw_idx, queue);
5277 	}
5278 
5279 	spin_unlock(&adapter->fdir_perfect_lock);
5280 }
5281 
5282 /**
5283  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
5284  * @rx_ring: ring to free buffers from
5285  **/
5286 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
5287 {
5288 	u16 i = rx_ring->next_to_clean;
5289 	struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
5290 
5291 	if (rx_ring->xsk_pool) {
5292 		ixgbe_xsk_clean_rx_ring(rx_ring);
5293 		goto skip_free;
5294 	}
5295 
5296 	/* Free all the Rx ring sk_buffs */
5297 	while (i != rx_ring->next_to_alloc) {
5298 		if (rx_buffer->skb) {
5299 			struct sk_buff *skb = rx_buffer->skb;
5300 			if (IXGBE_CB(skb)->page_released)
5301 				dma_unmap_page_attrs(rx_ring->dev,
5302 						     IXGBE_CB(skb)->dma,
5303 						     ixgbe_rx_pg_size(rx_ring),
5304 						     DMA_FROM_DEVICE,
5305 						     IXGBE_RX_DMA_ATTR);
5306 			dev_kfree_skb(skb);
5307 		}
5308 
5309 		/* Invalidate cache lines that may have been written to by
5310 		 * device so that we avoid corrupting memory.
5311 		 */
5312 		dma_sync_single_range_for_cpu(rx_ring->dev,
5313 					      rx_buffer->dma,
5314 					      rx_buffer->page_offset,
5315 					      ixgbe_rx_bufsz(rx_ring),
5316 					      DMA_FROM_DEVICE);
5317 
5318 		/* free resources associated with mapping */
5319 		dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
5320 				     ixgbe_rx_pg_size(rx_ring),
5321 				     DMA_FROM_DEVICE,
5322 				     IXGBE_RX_DMA_ATTR);
5323 		__page_frag_cache_drain(rx_buffer->page,
5324 					rx_buffer->pagecnt_bias);
5325 
5326 		i++;
5327 		rx_buffer++;
5328 		if (i == rx_ring->count) {
5329 			i = 0;
5330 			rx_buffer = rx_ring->rx_buffer_info;
5331 		}
5332 	}
5333 
5334 skip_free:
5335 	rx_ring->next_to_alloc = 0;
5336 	rx_ring->next_to_clean = 0;
5337 	rx_ring->next_to_use = 0;
5338 }
5339 
5340 static int ixgbe_fwd_ring_up(struct ixgbe_adapter *adapter,
5341 			     struct ixgbe_fwd_adapter *accel)
5342 {
5343 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
5344 	int num_tc = netdev_get_num_tc(adapter->netdev);
5345 	struct net_device *vdev = accel->netdev;
5346 	int i, baseq, err;
5347 
5348 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
5349 	netdev_dbg(vdev, "pool %i:%i queues %i:%i\n",
5350 		   accel->pool, adapter->num_rx_pools,
5351 		   baseq, baseq + adapter->num_rx_queues_per_pool);
5352 
5353 	accel->rx_base_queue = baseq;
5354 	accel->tx_base_queue = baseq;
5355 
5356 	/* record configuration for macvlan interface in vdev */
5357 	for (i = 0; i < num_tc; i++)
5358 		netdev_bind_sb_channel_queue(adapter->netdev, vdev,
5359 					     i, rss_i, baseq + (rss_i * i));
5360 
5361 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5362 		adapter->rx_ring[baseq + i]->netdev = vdev;
5363 
5364 	/* Guarantee all rings are updated before we update the
5365 	 * MAC address filter.
5366 	 */
5367 	wmb();
5368 
5369 	/* ixgbe_add_mac_filter will return an index if it succeeds, so we
5370 	 * need to only treat it as an error value if it is negative.
5371 	 */
5372 	err = ixgbe_add_mac_filter(adapter, vdev->dev_addr,
5373 				   VMDQ_P(accel->pool));
5374 	if (err >= 0)
5375 		return 0;
5376 
5377 	/* if we cannot add the MAC rule then disable the offload */
5378 	macvlan_release_l2fw_offload(vdev);
5379 
5380 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5381 		adapter->rx_ring[baseq + i]->netdev = NULL;
5382 
5383 	netdev_err(vdev, "L2FW offload disabled due to L2 filter error\n");
5384 
5385 	/* unbind the queues and drop the subordinate channel config */
5386 	netdev_unbind_sb_channel(adapter->netdev, vdev);
5387 	netdev_set_sb_channel(vdev, 0);
5388 
5389 	clear_bit(accel->pool, adapter->fwd_bitmask);
5390 	kfree(accel);
5391 
5392 	return err;
5393 }
5394 
5395 static int ixgbe_macvlan_up(struct net_device *vdev,
5396 			    struct netdev_nested_priv *priv)
5397 {
5398 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)priv->data;
5399 	struct ixgbe_fwd_adapter *accel;
5400 
5401 	if (!netif_is_macvlan(vdev))
5402 		return 0;
5403 
5404 	accel = macvlan_accel_priv(vdev);
5405 	if (!accel)
5406 		return 0;
5407 
5408 	ixgbe_fwd_ring_up(adapter, accel);
5409 
5410 	return 0;
5411 }
5412 
5413 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
5414 {
5415 	struct netdev_nested_priv priv = {
5416 		.data = (void *)adapter,
5417 	};
5418 
5419 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
5420 				      ixgbe_macvlan_up, &priv);
5421 }
5422 
5423 static void ixgbe_configure(struct ixgbe_adapter *adapter)
5424 {
5425 	struct ixgbe_hw *hw = &adapter->hw;
5426 
5427 	ixgbe_configure_pb(adapter);
5428 #ifdef CONFIG_IXGBE_DCB
5429 	ixgbe_configure_dcb(adapter);
5430 #endif
5431 	/*
5432 	 * We must restore virtualization before VLANs or else
5433 	 * the VLVF registers will not be populated
5434 	 */
5435 	ixgbe_configure_virtualization(adapter);
5436 
5437 	ixgbe_set_rx_mode(adapter->netdev);
5438 	ixgbe_restore_vlan(adapter);
5439 	ixgbe_ipsec_restore(adapter);
5440 
5441 	switch (hw->mac.type) {
5442 	case ixgbe_mac_82599EB:
5443 	case ixgbe_mac_X540:
5444 		hw->mac.ops.disable_rx_buff(hw);
5445 		break;
5446 	default:
5447 		break;
5448 	}
5449 
5450 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
5451 		ixgbe_init_fdir_signature_82599(&adapter->hw,
5452 						adapter->fdir_pballoc);
5453 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
5454 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
5455 					      adapter->fdir_pballoc);
5456 		ixgbe_fdir_filter_restore(adapter);
5457 	}
5458 
5459 	switch (hw->mac.type) {
5460 	case ixgbe_mac_82599EB:
5461 	case ixgbe_mac_X540:
5462 		hw->mac.ops.enable_rx_buff(hw);
5463 		break;
5464 	default:
5465 		break;
5466 	}
5467 
5468 #ifdef CONFIG_IXGBE_DCA
5469 	/* configure DCA */
5470 	if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
5471 		ixgbe_setup_dca(adapter);
5472 #endif /* CONFIG_IXGBE_DCA */
5473 
5474 #ifdef IXGBE_FCOE
5475 	/* configure FCoE L2 filters, redirection table, and Rx control */
5476 	ixgbe_configure_fcoe(adapter);
5477 
5478 #endif /* IXGBE_FCOE */
5479 	ixgbe_configure_tx(adapter);
5480 	ixgbe_configure_rx(adapter);
5481 	ixgbe_configure_dfwd(adapter);
5482 }
5483 
5484 /**
5485  * ixgbe_sfp_link_config - set up SFP+ link
5486  * @adapter: pointer to private adapter struct
5487  **/
5488 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
5489 {
5490 	/*
5491 	 * We are assuming the worst case scenario here, and that
5492 	 * is that an SFP was inserted/removed after the reset
5493 	 * but before SFP detection was enabled.  As such the best
5494 	 * solution is to just start searching as soon as we start
5495 	 */
5496 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5497 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5498 
5499 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5500 	adapter->sfp_poll_time = 0;
5501 }
5502 
5503 /**
5504  * ixgbe_non_sfp_link_config - set up non-SFP+ link
5505  * @hw: pointer to private hardware struct
5506  *
5507  * Returns 0 on success, negative on failure
5508  **/
5509 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
5510 {
5511 	u32 speed;
5512 	bool autoneg, link_up = false;
5513 	int ret = -EIO;
5514 
5515 	if (hw->mac.ops.check_link)
5516 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
5517 
5518 	if (ret)
5519 		return ret;
5520 
5521 	speed = hw->phy.autoneg_advertised;
5522 	if (!speed && hw->mac.ops.get_link_capabilities) {
5523 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
5524 							&autoneg);
5525 		/* remove NBASE-T speeds from default autonegotiation
5526 		 * to accommodate broken network switches in the field
5527 		 * which cannot cope with advertised NBASE-T speeds
5528 		 */
5529 		speed &= ~(IXGBE_LINK_SPEED_5GB_FULL |
5530 			   IXGBE_LINK_SPEED_2_5GB_FULL);
5531 	}
5532 
5533 	if (ret)
5534 		return ret;
5535 
5536 	if (hw->mac.ops.setup_link)
5537 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
5538 
5539 	return ret;
5540 }
5541 
5542 /**
5543  * ixgbe_clear_vf_stats_counters - Clear out VF stats after reset
5544  * @adapter: board private structure
5545  *
5546  * On a reset we need to clear out the VF stats or accounting gets
5547  * messed up because they're not clear on read.
5548  **/
5549 static void ixgbe_clear_vf_stats_counters(struct ixgbe_adapter *adapter)
5550 {
5551 	struct ixgbe_hw *hw = &adapter->hw;
5552 	int i;
5553 
5554 	for (i = 0; i < adapter->num_vfs; i++) {
5555 		adapter->vfinfo[i].last_vfstats.gprc =
5556 			IXGBE_READ_REG(hw, IXGBE_PVFGPRC(i));
5557 		adapter->vfinfo[i].saved_rst_vfstats.gprc +=
5558 			adapter->vfinfo[i].vfstats.gprc;
5559 		adapter->vfinfo[i].vfstats.gprc = 0;
5560 		adapter->vfinfo[i].last_vfstats.gptc =
5561 			IXGBE_READ_REG(hw, IXGBE_PVFGPTC(i));
5562 		adapter->vfinfo[i].saved_rst_vfstats.gptc +=
5563 			adapter->vfinfo[i].vfstats.gptc;
5564 		adapter->vfinfo[i].vfstats.gptc = 0;
5565 		adapter->vfinfo[i].last_vfstats.gorc =
5566 			IXGBE_READ_REG(hw, IXGBE_PVFGORC_LSB(i));
5567 		adapter->vfinfo[i].saved_rst_vfstats.gorc +=
5568 			adapter->vfinfo[i].vfstats.gorc;
5569 		adapter->vfinfo[i].vfstats.gorc = 0;
5570 		adapter->vfinfo[i].last_vfstats.gotc =
5571 			IXGBE_READ_REG(hw, IXGBE_PVFGOTC_LSB(i));
5572 		adapter->vfinfo[i].saved_rst_vfstats.gotc +=
5573 			adapter->vfinfo[i].vfstats.gotc;
5574 		adapter->vfinfo[i].vfstats.gotc = 0;
5575 		adapter->vfinfo[i].last_vfstats.mprc =
5576 			IXGBE_READ_REG(hw, IXGBE_PVFMPRC(i));
5577 		adapter->vfinfo[i].saved_rst_vfstats.mprc +=
5578 			adapter->vfinfo[i].vfstats.mprc;
5579 		adapter->vfinfo[i].vfstats.mprc = 0;
5580 	}
5581 }
5582 
5583 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
5584 {
5585 	struct ixgbe_hw *hw = &adapter->hw;
5586 	u32 gpie = 0;
5587 
5588 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
5589 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5590 		       IXGBE_GPIE_OCD;
5591 		gpie |= IXGBE_GPIE_EIAME;
5592 		/*
5593 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
5594 		 * this saves a register write for every interrupt
5595 		 */
5596 		switch (hw->mac.type) {
5597 		case ixgbe_mac_82598EB:
5598 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5599 			break;
5600 		case ixgbe_mac_82599EB:
5601 		case ixgbe_mac_X540:
5602 		case ixgbe_mac_X550:
5603 		case ixgbe_mac_X550EM_x:
5604 		case ixgbe_mac_x550em_a:
5605 		default:
5606 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5607 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5608 			break;
5609 		}
5610 	} else {
5611 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
5612 		 * specifically only auto mask tx and rx interrupts */
5613 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5614 	}
5615 
5616 	/* XXX: to interrupt immediately for EICS writes, enable this */
5617 	/* gpie |= IXGBE_GPIE_EIMEN; */
5618 
5619 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
5620 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
5621 
5622 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
5623 		case IXGBE_82599_VMDQ_8Q_MASK:
5624 			gpie |= IXGBE_GPIE_VTMODE_16;
5625 			break;
5626 		case IXGBE_82599_VMDQ_4Q_MASK:
5627 			gpie |= IXGBE_GPIE_VTMODE_32;
5628 			break;
5629 		default:
5630 			gpie |= IXGBE_GPIE_VTMODE_64;
5631 			break;
5632 		}
5633 	}
5634 
5635 	/* Enable Thermal over heat sensor interrupt */
5636 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
5637 		switch (adapter->hw.mac.type) {
5638 		case ixgbe_mac_82599EB:
5639 			gpie |= IXGBE_SDP0_GPIEN_8259X;
5640 			break;
5641 		default:
5642 			break;
5643 		}
5644 	}
5645 
5646 	/* Enable fan failure interrupt */
5647 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
5648 		gpie |= IXGBE_SDP1_GPIEN(hw);
5649 
5650 	switch (hw->mac.type) {
5651 	case ixgbe_mac_82599EB:
5652 		gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
5653 		break;
5654 	case ixgbe_mac_X550EM_x:
5655 	case ixgbe_mac_x550em_a:
5656 		gpie |= IXGBE_SDP0_GPIEN_X540;
5657 		break;
5658 	default:
5659 		break;
5660 	}
5661 
5662 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5663 }
5664 
5665 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
5666 {
5667 	struct ixgbe_hw *hw = &adapter->hw;
5668 	int err;
5669 	u32 ctrl_ext;
5670 
5671 	ixgbe_get_hw_control(adapter);
5672 	ixgbe_setup_gpie(adapter);
5673 
5674 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
5675 		ixgbe_configure_msix(adapter);
5676 	else
5677 		ixgbe_configure_msi_and_legacy(adapter);
5678 
5679 	/* enable the optics for 82599 SFP+ fiber */
5680 	if (hw->mac.ops.enable_tx_laser)
5681 		hw->mac.ops.enable_tx_laser(hw);
5682 
5683 	if (hw->phy.ops.set_phy_power)
5684 		hw->phy.ops.set_phy_power(hw, true);
5685 
5686 	smp_mb__before_atomic();
5687 	clear_bit(__IXGBE_DOWN, &adapter->state);
5688 	ixgbe_napi_enable_all(adapter);
5689 
5690 	if (ixgbe_is_sfp(hw)) {
5691 		ixgbe_sfp_link_config(adapter);
5692 	} else {
5693 		err = ixgbe_non_sfp_link_config(hw);
5694 		if (err)
5695 			e_err(probe, "link_config FAILED %d\n", err);
5696 	}
5697 
5698 	/* clear any pending interrupts, may auto mask */
5699 	IXGBE_READ_REG(hw, IXGBE_EICR);
5700 	ixgbe_irq_enable(adapter, true, true);
5701 
5702 	/*
5703 	 * If this adapter has a fan, check to see if we had a failure
5704 	 * before we enabled the interrupt.
5705 	 */
5706 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
5707 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
5708 		if (esdp & IXGBE_ESDP_SDP1)
5709 			e_crit(drv, "Fan has stopped, replace the adapter\n");
5710 	}
5711 
5712 	/* bring the link up in the watchdog, this could race with our first
5713 	 * link up interrupt but shouldn't be a problem */
5714 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5715 	adapter->link_check_timeout = jiffies;
5716 	mod_timer(&adapter->service_timer, jiffies);
5717 
5718 	ixgbe_clear_vf_stats_counters(adapter);
5719 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
5720 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
5721 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
5722 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
5723 
5724 	/* update setting rx tx for all active vfs */
5725 	ixgbe_set_all_vfs(adapter);
5726 }
5727 
5728 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
5729 {
5730 	/* put off any impending NetWatchDogTimeout */
5731 	netif_trans_update(adapter->netdev);
5732 
5733 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
5734 		usleep_range(1000, 2000);
5735 	if (adapter->hw.phy.type == ixgbe_phy_fw)
5736 		ixgbe_watchdog_link_is_down(adapter);
5737 	ixgbe_down(adapter);
5738 	/*
5739 	 * If SR-IOV enabled then wait a bit before bringing the adapter
5740 	 * back up to give the VFs time to respond to the reset.  The
5741 	 * two second wait is based upon the watchdog timer cycle in
5742 	 * the VF driver.
5743 	 */
5744 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5745 		msleep(2000);
5746 	ixgbe_up(adapter);
5747 	clear_bit(__IXGBE_RESETTING, &adapter->state);
5748 }
5749 
5750 void ixgbe_up(struct ixgbe_adapter *adapter)
5751 {
5752 	/* hardware has been reset, we need to reload some things */
5753 	ixgbe_configure(adapter);
5754 
5755 	ixgbe_up_complete(adapter);
5756 }
5757 
5758 static unsigned long ixgbe_get_completion_timeout(struct ixgbe_adapter *adapter)
5759 {
5760 	u16 devctl2;
5761 
5762 	pcie_capability_read_word(adapter->pdev, PCI_EXP_DEVCTL2, &devctl2);
5763 
5764 	switch (devctl2 & IXGBE_PCIDEVCTRL2_TIMEO_MASK) {
5765 	case IXGBE_PCIDEVCTRL2_17_34s:
5766 	case IXGBE_PCIDEVCTRL2_4_8s:
5767 		/* For now we cap the upper limit on delay to 2 seconds
5768 		 * as we end up going up to 34 seconds of delay in worst
5769 		 * case timeout value.
5770 		 */
5771 	case IXGBE_PCIDEVCTRL2_1_2s:
5772 		return 2000000ul;	/* 2.0 s */
5773 	case IXGBE_PCIDEVCTRL2_260_520ms:
5774 		return 520000ul;	/* 520 ms */
5775 	case IXGBE_PCIDEVCTRL2_65_130ms:
5776 		return 130000ul;	/* 130 ms */
5777 	case IXGBE_PCIDEVCTRL2_16_32ms:
5778 		return 32000ul;		/* 32 ms */
5779 	case IXGBE_PCIDEVCTRL2_1_2ms:
5780 		return 2000ul;		/* 2 ms */
5781 	case IXGBE_PCIDEVCTRL2_50_100us:
5782 		return 100ul;		/* 100 us */
5783 	case IXGBE_PCIDEVCTRL2_16_32ms_def:
5784 		return 32000ul;		/* 32 ms */
5785 	default:
5786 		break;
5787 	}
5788 
5789 	/* We shouldn't need to hit this path, but just in case default as
5790 	 * though completion timeout is not supported and support 32ms.
5791 	 */
5792 	return 32000ul;
5793 }
5794 
5795 void ixgbe_disable_rx(struct ixgbe_adapter *adapter)
5796 {
5797 	unsigned long wait_delay, delay_interval;
5798 	struct ixgbe_hw *hw = &adapter->hw;
5799 	int i, wait_loop;
5800 	u32 rxdctl;
5801 
5802 	/* disable receives */
5803 	hw->mac.ops.disable_rx(hw);
5804 
5805 	if (ixgbe_removed(hw->hw_addr))
5806 		return;
5807 
5808 	/* disable all enabled Rx queues */
5809 	for (i = 0; i < adapter->num_rx_queues; i++) {
5810 		struct ixgbe_ring *ring = adapter->rx_ring[i];
5811 		u8 reg_idx = ring->reg_idx;
5812 
5813 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
5814 		rxdctl &= ~IXGBE_RXDCTL_ENABLE;
5815 		rxdctl |= IXGBE_RXDCTL_SWFLSH;
5816 
5817 		/* write value back with RXDCTL.ENABLE bit cleared */
5818 		IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
5819 	}
5820 
5821 	/* RXDCTL.EN may not change on 82598 if link is down, so skip it */
5822 	if (hw->mac.type == ixgbe_mac_82598EB &&
5823 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
5824 		return;
5825 
5826 	/* Determine our minimum delay interval. We will increase this value
5827 	 * with each subsequent test. This way if the device returns quickly
5828 	 * we should spend as little time as possible waiting, however as
5829 	 * the time increases we will wait for larger periods of time.
5830 	 *
5831 	 * The trick here is that we increase the interval using the
5832 	 * following pattern: 1x 3x 5x 7x 9x 11x 13x 15x 17x 19x. The result
5833 	 * of that wait is that it totals up to 100x whatever interval we
5834 	 * choose. Since our minimum wait is 100us we can just divide the
5835 	 * total timeout by 100 to get our minimum delay interval.
5836 	 */
5837 	delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
5838 
5839 	wait_loop = IXGBE_MAX_RX_DESC_POLL;
5840 	wait_delay = delay_interval;
5841 
5842 	while (wait_loop--) {
5843 		usleep_range(wait_delay, wait_delay + 10);
5844 		wait_delay += delay_interval * 2;
5845 		rxdctl = 0;
5846 
5847 		/* OR together the reading of all the active RXDCTL registers,
5848 		 * and then test the result. We need the disable to complete
5849 		 * before we start freeing the memory and invalidating the
5850 		 * DMA mappings.
5851 		 */
5852 		for (i = 0; i < adapter->num_rx_queues; i++) {
5853 			struct ixgbe_ring *ring = adapter->rx_ring[i];
5854 			u8 reg_idx = ring->reg_idx;
5855 
5856 			rxdctl |= IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
5857 		}
5858 
5859 		if (!(rxdctl & IXGBE_RXDCTL_ENABLE))
5860 			return;
5861 	}
5862 
5863 	e_err(drv,
5864 	      "RXDCTL.ENABLE for one or more queues not cleared within the polling period\n");
5865 }
5866 
5867 void ixgbe_disable_tx(struct ixgbe_adapter *adapter)
5868 {
5869 	unsigned long wait_delay, delay_interval;
5870 	struct ixgbe_hw *hw = &adapter->hw;
5871 	int i, wait_loop;
5872 	u32 txdctl;
5873 
5874 	if (ixgbe_removed(hw->hw_addr))
5875 		return;
5876 
5877 	/* disable all enabled Tx queues */
5878 	for (i = 0; i < adapter->num_tx_queues; i++) {
5879 		struct ixgbe_ring *ring = adapter->tx_ring[i];
5880 		u8 reg_idx = ring->reg_idx;
5881 
5882 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5883 	}
5884 
5885 	/* disable all enabled XDP Tx queues */
5886 	for (i = 0; i < adapter->num_xdp_queues; i++) {
5887 		struct ixgbe_ring *ring = adapter->xdp_ring[i];
5888 		u8 reg_idx = ring->reg_idx;
5889 
5890 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5891 	}
5892 
5893 	/* If the link is not up there shouldn't be much in the way of
5894 	 * pending transactions. Those that are left will be flushed out
5895 	 * when the reset logic goes through the flush sequence to clean out
5896 	 * the pending Tx transactions.
5897 	 */
5898 	if (!(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
5899 		goto dma_engine_disable;
5900 
5901 	/* Determine our minimum delay interval. We will increase this value
5902 	 * with each subsequent test. This way if the device returns quickly
5903 	 * we should spend as little time as possible waiting, however as
5904 	 * the time increases we will wait for larger periods of time.
5905 	 *
5906 	 * The trick here is that we increase the interval using the
5907 	 * following pattern: 1x 3x 5x 7x 9x 11x 13x 15x 17x 19x. The result
5908 	 * of that wait is that it totals up to 100x whatever interval we
5909 	 * choose. Since our minimum wait is 100us we can just divide the
5910 	 * total timeout by 100 to get our minimum delay interval.
5911 	 */
5912 	delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
5913 
5914 	wait_loop = IXGBE_MAX_RX_DESC_POLL;
5915 	wait_delay = delay_interval;
5916 
5917 	while (wait_loop--) {
5918 		usleep_range(wait_delay, wait_delay + 10);
5919 		wait_delay += delay_interval * 2;
5920 		txdctl = 0;
5921 
5922 		/* OR together the reading of all the active TXDCTL registers,
5923 		 * and then test the result. We need the disable to complete
5924 		 * before we start freeing the memory and invalidating the
5925 		 * DMA mappings.
5926 		 */
5927 		for (i = 0; i < adapter->num_tx_queues; i++) {
5928 			struct ixgbe_ring *ring = adapter->tx_ring[i];
5929 			u8 reg_idx = ring->reg_idx;
5930 
5931 			txdctl |= IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
5932 		}
5933 		for (i = 0; i < adapter->num_xdp_queues; i++) {
5934 			struct ixgbe_ring *ring = adapter->xdp_ring[i];
5935 			u8 reg_idx = ring->reg_idx;
5936 
5937 			txdctl |= IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
5938 		}
5939 
5940 		if (!(txdctl & IXGBE_TXDCTL_ENABLE))
5941 			goto dma_engine_disable;
5942 	}
5943 
5944 	e_err(drv,
5945 	      "TXDCTL.ENABLE for one or more queues not cleared within the polling period\n");
5946 
5947 dma_engine_disable:
5948 	/* Disable the Tx DMA engine on 82599 and later MAC */
5949 	switch (hw->mac.type) {
5950 	case ixgbe_mac_82599EB:
5951 	case ixgbe_mac_X540:
5952 	case ixgbe_mac_X550:
5953 	case ixgbe_mac_X550EM_x:
5954 	case ixgbe_mac_x550em_a:
5955 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5956 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5957 				 ~IXGBE_DMATXCTL_TE));
5958 		fallthrough;
5959 	default:
5960 		break;
5961 	}
5962 }
5963 
5964 void ixgbe_reset(struct ixgbe_adapter *adapter)
5965 {
5966 	struct ixgbe_hw *hw = &adapter->hw;
5967 	struct net_device *netdev = adapter->netdev;
5968 	int err;
5969 
5970 	if (ixgbe_removed(hw->hw_addr))
5971 		return;
5972 	/* lock SFP init bit to prevent race conditions with the watchdog */
5973 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5974 		usleep_range(1000, 2000);
5975 
5976 	/* clear all SFP and link config related flags while holding SFP_INIT */
5977 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
5978 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
5979 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5980 
5981 	err = hw->mac.ops.init_hw(hw);
5982 	switch (err) {
5983 	case 0:
5984 	case -ENOENT:
5985 	case -EOPNOTSUPP:
5986 		break;
5987 	case -EALREADY:
5988 		e_dev_err("primary disable timed out\n");
5989 		break;
5990 	case -EACCES:
5991 		/* We are running on a pre-production device, log a warning */
5992 		e_dev_warn("This device is a pre-production adapter/LOM. "
5993 			   "Please be aware there may be issues associated with "
5994 			   "your hardware.  If you are experiencing problems "
5995 			   "please contact your Intel or hardware "
5996 			   "representative who provided you with this "
5997 			   "hardware.\n");
5998 		break;
5999 	default:
6000 		e_dev_err("Hardware Error: %d\n", err);
6001 	}
6002 
6003 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
6004 
6005 	/* flush entries out of MAC table */
6006 	ixgbe_flush_sw_mac_table(adapter);
6007 	__dev_uc_unsync(netdev, NULL);
6008 
6009 	/* do not flush user set addresses */
6010 	ixgbe_mac_set_default_filter(adapter);
6011 
6012 	/* update SAN MAC vmdq pool selection */
6013 	if (hw->mac.san_mac_rar_index)
6014 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
6015 
6016 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6017 		ixgbe_ptp_reset(adapter);
6018 
6019 	if (hw->phy.ops.set_phy_power) {
6020 		if (!netif_running(adapter->netdev) && !adapter->wol)
6021 			hw->phy.ops.set_phy_power(hw, false);
6022 		else
6023 			hw->phy.ops.set_phy_power(hw, true);
6024 	}
6025 }
6026 
6027 /**
6028  * ixgbe_clean_tx_ring - Free Tx Buffers
6029  * @tx_ring: ring to be cleaned
6030  **/
6031 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
6032 {
6033 	u16 i = tx_ring->next_to_clean;
6034 	struct ixgbe_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
6035 
6036 	if (tx_ring->xsk_pool) {
6037 		ixgbe_xsk_clean_tx_ring(tx_ring);
6038 		goto out;
6039 	}
6040 
6041 	while (i != tx_ring->next_to_use) {
6042 		union ixgbe_adv_tx_desc *eop_desc, *tx_desc;
6043 
6044 		/* Free all the Tx ring sk_buffs */
6045 		if (ring_is_xdp(tx_ring))
6046 			xdp_return_frame(tx_buffer->xdpf);
6047 		else
6048 			dev_kfree_skb_any(tx_buffer->skb);
6049 
6050 		/* unmap skb header data */
6051 		dma_unmap_single(tx_ring->dev,
6052 				 dma_unmap_addr(tx_buffer, dma),
6053 				 dma_unmap_len(tx_buffer, len),
6054 				 DMA_TO_DEVICE);
6055 
6056 		/* check for eop_desc to determine the end of the packet */
6057 		eop_desc = tx_buffer->next_to_watch;
6058 		tx_desc = IXGBE_TX_DESC(tx_ring, i);
6059 
6060 		/* unmap remaining buffers */
6061 		while (tx_desc != eop_desc) {
6062 			tx_buffer++;
6063 			tx_desc++;
6064 			i++;
6065 			if (unlikely(i == tx_ring->count)) {
6066 				i = 0;
6067 				tx_buffer = tx_ring->tx_buffer_info;
6068 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6069 			}
6070 
6071 			/* unmap any remaining paged data */
6072 			if (dma_unmap_len(tx_buffer, len))
6073 				dma_unmap_page(tx_ring->dev,
6074 					       dma_unmap_addr(tx_buffer, dma),
6075 					       dma_unmap_len(tx_buffer, len),
6076 					       DMA_TO_DEVICE);
6077 		}
6078 
6079 		/* move us one more past the eop_desc for start of next pkt */
6080 		tx_buffer++;
6081 		i++;
6082 		if (unlikely(i == tx_ring->count)) {
6083 			i = 0;
6084 			tx_buffer = tx_ring->tx_buffer_info;
6085 		}
6086 	}
6087 
6088 	/* reset BQL for queue */
6089 	if (!ring_is_xdp(tx_ring))
6090 		netdev_tx_reset_queue(txring_txq(tx_ring));
6091 
6092 out:
6093 	/* reset next_to_use and next_to_clean */
6094 	tx_ring->next_to_use = 0;
6095 	tx_ring->next_to_clean = 0;
6096 }
6097 
6098 /**
6099  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
6100  * @adapter: board private structure
6101  **/
6102 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
6103 {
6104 	int i;
6105 
6106 	for (i = 0; i < adapter->num_rx_queues; i++)
6107 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
6108 }
6109 
6110 /**
6111  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
6112  * @adapter: board private structure
6113  **/
6114 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
6115 {
6116 	int i;
6117 
6118 	for (i = 0; i < adapter->num_tx_queues; i++)
6119 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
6120 	for (i = 0; i < adapter->num_xdp_queues; i++)
6121 		ixgbe_clean_tx_ring(adapter->xdp_ring[i]);
6122 }
6123 
6124 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
6125 {
6126 	struct hlist_node *node2;
6127 	struct ixgbe_fdir_filter *filter;
6128 
6129 	spin_lock(&adapter->fdir_perfect_lock);
6130 
6131 	hlist_for_each_entry_safe(filter, node2,
6132 				  &adapter->fdir_filter_list, fdir_node) {
6133 		hlist_del(&filter->fdir_node);
6134 		kfree(filter);
6135 	}
6136 	adapter->fdir_filter_count = 0;
6137 
6138 	spin_unlock(&adapter->fdir_perfect_lock);
6139 }
6140 
6141 void ixgbe_down(struct ixgbe_adapter *adapter)
6142 {
6143 	struct net_device *netdev = adapter->netdev;
6144 	struct ixgbe_hw *hw = &adapter->hw;
6145 	int i;
6146 
6147 	/* signal that we are down to the interrupt handler */
6148 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
6149 		return; /* do nothing if already down */
6150 
6151 	/* Shut off incoming Tx traffic */
6152 	netif_tx_stop_all_queues(netdev);
6153 
6154 	/* call carrier off first to avoid false dev_watchdog timeouts */
6155 	netif_carrier_off(netdev);
6156 	netif_tx_disable(netdev);
6157 
6158 	/* Disable Rx */
6159 	ixgbe_disable_rx(adapter);
6160 
6161 	/* synchronize_rcu() needed for pending XDP buffers to drain */
6162 	if (adapter->xdp_ring[0])
6163 		synchronize_rcu();
6164 
6165 	ixgbe_irq_disable(adapter);
6166 
6167 	ixgbe_napi_disable_all(adapter);
6168 
6169 	clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
6170 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6171 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6172 
6173 	del_timer_sync(&adapter->service_timer);
6174 
6175 	if (adapter->num_vfs) {
6176 		/* Clear EITR Select mapping */
6177 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
6178 
6179 		/* Mark all the VFs as inactive */
6180 		for (i = 0 ; i < adapter->num_vfs; i++)
6181 			adapter->vfinfo[i].clear_to_send = false;
6182 
6183 		/* update setting rx tx for all active vfs */
6184 		ixgbe_set_all_vfs(adapter);
6185 	}
6186 
6187 	/* disable transmits in the hardware now that interrupts are off */
6188 	ixgbe_disable_tx(adapter);
6189 
6190 	if (!pci_channel_offline(adapter->pdev))
6191 		ixgbe_reset(adapter);
6192 
6193 	/* power down the optics for 82599 SFP+ fiber */
6194 	if (hw->mac.ops.disable_tx_laser)
6195 		hw->mac.ops.disable_tx_laser(hw);
6196 
6197 	ixgbe_clean_all_tx_rings(adapter);
6198 	ixgbe_clean_all_rx_rings(adapter);
6199 }
6200 
6201 /**
6202  * ixgbe_set_eee_capable - helper function to determine EEE support on X550
6203  * @adapter: board private structure
6204  */
6205 static void ixgbe_set_eee_capable(struct ixgbe_adapter *adapter)
6206 {
6207 	struct ixgbe_hw *hw = &adapter->hw;
6208 
6209 	switch (hw->device_id) {
6210 	case IXGBE_DEV_ID_X550EM_A_1G_T:
6211 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6212 		if (!hw->phy.eee_speeds_supported)
6213 			break;
6214 		adapter->flags2 |= IXGBE_FLAG2_EEE_CAPABLE;
6215 		if (!hw->phy.eee_speeds_advertised)
6216 			break;
6217 		adapter->flags2 |= IXGBE_FLAG2_EEE_ENABLED;
6218 		break;
6219 	default:
6220 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_CAPABLE;
6221 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_ENABLED;
6222 		break;
6223 	}
6224 }
6225 
6226 /**
6227  * ixgbe_tx_timeout - Respond to a Tx Hang
6228  * @netdev: network interface device structure
6229  * @txqueue: queue number that timed out
6230  **/
6231 static void ixgbe_tx_timeout(struct net_device *netdev, unsigned int __always_unused txqueue)
6232 {
6233 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6234 
6235 	/* Do the reset outside of interrupt context */
6236 	ixgbe_tx_timeout_reset(adapter);
6237 }
6238 
6239 #ifdef CONFIG_IXGBE_DCB
6240 static void ixgbe_init_dcb(struct ixgbe_adapter *adapter)
6241 {
6242 	struct ixgbe_hw *hw = &adapter->hw;
6243 	struct tc_configuration *tc;
6244 	int j;
6245 
6246 	switch (hw->mac.type) {
6247 	case ixgbe_mac_82598EB:
6248 	case ixgbe_mac_82599EB:
6249 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
6250 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
6251 		break;
6252 	case ixgbe_mac_X540:
6253 	case ixgbe_mac_X550:
6254 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
6255 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
6256 		break;
6257 	case ixgbe_mac_X550EM_x:
6258 	case ixgbe_mac_x550em_a:
6259 	default:
6260 		adapter->dcb_cfg.num_tcs.pg_tcs = DEF_TRAFFIC_CLASS;
6261 		adapter->dcb_cfg.num_tcs.pfc_tcs = DEF_TRAFFIC_CLASS;
6262 		break;
6263 	}
6264 
6265 	/* Configure DCB traffic classes */
6266 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
6267 		tc = &adapter->dcb_cfg.tc_config[j];
6268 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
6269 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
6270 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
6271 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
6272 		tc->dcb_pfc = pfc_disabled;
6273 	}
6274 
6275 	/* Initialize default user to priority mapping, UPx->TC0 */
6276 	tc = &adapter->dcb_cfg.tc_config[0];
6277 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
6278 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
6279 
6280 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
6281 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
6282 	adapter->dcb_cfg.pfc_mode_enable = false;
6283 	adapter->dcb_set_bitmap = 0x00;
6284 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
6285 		adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
6286 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
6287 	       sizeof(adapter->temp_dcb_cfg));
6288 }
6289 #endif
6290 
6291 /**
6292  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
6293  * @adapter: board private structure to initialize
6294  * @ii: pointer to ixgbe_info for device
6295  *
6296  * ixgbe_sw_init initializes the Adapter private data structure.
6297  * Fields are initialized based on PCI device information and
6298  * OS network device settings (MTU size).
6299  **/
6300 static int ixgbe_sw_init(struct ixgbe_adapter *adapter,
6301 			 const struct ixgbe_info *ii)
6302 {
6303 	struct ixgbe_hw *hw = &adapter->hw;
6304 	struct pci_dev *pdev = adapter->pdev;
6305 	unsigned int rss, fdir;
6306 	u32 fwsm;
6307 	int i;
6308 
6309 	/* PCI config space info */
6310 
6311 	hw->vendor_id = pdev->vendor;
6312 	hw->device_id = pdev->device;
6313 	hw->revision_id = pdev->revision;
6314 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
6315 	hw->subsystem_device_id = pdev->subsystem_device;
6316 
6317 	/* get_invariants needs the device IDs */
6318 	ii->get_invariants(hw);
6319 
6320 	/* Set common capability flags and settings */
6321 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
6322 	adapter->ring_feature[RING_F_RSS].limit = rss;
6323 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
6324 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
6325 	adapter->atr_sample_rate = 20;
6326 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
6327 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
6328 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
6329 	adapter->ring_feature[RING_F_VMDQ].limit = 1;
6330 #ifdef CONFIG_IXGBE_DCA
6331 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
6332 #endif
6333 #ifdef CONFIG_IXGBE_DCB
6334 	adapter->flags |= IXGBE_FLAG_DCB_CAPABLE;
6335 	adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
6336 #endif
6337 #ifdef IXGBE_FCOE
6338 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
6339 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6340 #ifdef CONFIG_IXGBE_DCB
6341 	/* Default traffic class to use for FCoE */
6342 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
6343 #endif /* CONFIG_IXGBE_DCB */
6344 #endif /* IXGBE_FCOE */
6345 
6346 	/* initialize static ixgbe jump table entries */
6347 	adapter->jump_tables[0] = kzalloc(sizeof(*adapter->jump_tables[0]),
6348 					  GFP_KERNEL);
6349 	if (!adapter->jump_tables[0])
6350 		return -ENOMEM;
6351 	adapter->jump_tables[0]->mat = ixgbe_ipv4_fields;
6352 
6353 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++)
6354 		adapter->jump_tables[i] = NULL;
6355 
6356 	adapter->mac_table = kcalloc(hw->mac.num_rar_entries,
6357 				     sizeof(struct ixgbe_mac_addr),
6358 				     GFP_KERNEL);
6359 	if (!adapter->mac_table)
6360 		return -ENOMEM;
6361 
6362 	if (ixgbe_init_rss_key(adapter))
6363 		return -ENOMEM;
6364 
6365 	adapter->af_xdp_zc_qps = bitmap_zalloc(IXGBE_MAX_XDP_QS, GFP_KERNEL);
6366 	if (!adapter->af_xdp_zc_qps)
6367 		return -ENOMEM;
6368 
6369 	/* Set MAC specific capability flags and exceptions */
6370 	switch (hw->mac.type) {
6371 	case ixgbe_mac_82598EB:
6372 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
6373 
6374 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
6375 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
6376 
6377 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
6378 		adapter->ring_feature[RING_F_FDIR].limit = 0;
6379 		adapter->atr_sample_rate = 0;
6380 		adapter->fdir_pballoc = 0;
6381 #ifdef IXGBE_FCOE
6382 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6383 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6384 #ifdef CONFIG_IXGBE_DCB
6385 		adapter->fcoe.up = 0;
6386 #endif /* IXGBE_DCB */
6387 #endif /* IXGBE_FCOE */
6388 		break;
6389 	case ixgbe_mac_82599EB:
6390 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
6391 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6392 		break;
6393 	case ixgbe_mac_X540:
6394 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
6395 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
6396 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6397 		break;
6398 	case ixgbe_mac_x550em_a:
6399 		switch (hw->device_id) {
6400 		case IXGBE_DEV_ID_X550EM_A_1G_T:
6401 		case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6402 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6403 			break;
6404 		default:
6405 			break;
6406 		}
6407 		fallthrough;
6408 	case ixgbe_mac_X550EM_x:
6409 #ifdef CONFIG_IXGBE_DCB
6410 		adapter->flags &= ~IXGBE_FLAG_DCB_CAPABLE;
6411 #endif
6412 #ifdef IXGBE_FCOE
6413 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6414 #ifdef CONFIG_IXGBE_DCB
6415 		adapter->fcoe.up = 0;
6416 #endif /* IXGBE_DCB */
6417 #endif /* IXGBE_FCOE */
6418 		fallthrough;
6419 	case ixgbe_mac_X550:
6420 		if (hw->mac.type == ixgbe_mac_X550)
6421 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6422 #ifdef CONFIG_IXGBE_DCA
6423 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
6424 #endif
6425 		break;
6426 	default:
6427 		break;
6428 	}
6429 
6430 #ifdef IXGBE_FCOE
6431 	/* FCoE support exists, always init the FCoE lock */
6432 	spin_lock_init(&adapter->fcoe.lock);
6433 
6434 #endif
6435 	/* n-tuple support exists, always init our spinlock */
6436 	spin_lock_init(&adapter->fdir_perfect_lock);
6437 
6438 	/* init spinlock to avoid concurrency of VF resources */
6439 	spin_lock_init(&adapter->vfs_lock);
6440 
6441 #ifdef CONFIG_IXGBE_DCB
6442 	ixgbe_init_dcb(adapter);
6443 #endif
6444 	ixgbe_init_ipsec_offload(adapter);
6445 
6446 	/* default flow control settings */
6447 	hw->fc.requested_mode = ixgbe_fc_full;
6448 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
6449 	ixgbe_pbthresh_setup(adapter);
6450 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
6451 	hw->fc.send_xon = true;
6452 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
6453 
6454 #ifdef CONFIG_PCI_IOV
6455 	if (max_vfs > 0)
6456 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
6457 
6458 	/* assign number of SR-IOV VFs */
6459 	if (hw->mac.type != ixgbe_mac_82598EB) {
6460 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
6461 			max_vfs = 0;
6462 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
6463 		}
6464 	}
6465 #endif /* CONFIG_PCI_IOV */
6466 
6467 	/* enable itr by default in dynamic mode */
6468 	adapter->rx_itr_setting = 1;
6469 	adapter->tx_itr_setting = 1;
6470 
6471 	/* set default ring sizes */
6472 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
6473 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
6474 
6475 	/* set default work limits */
6476 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
6477 
6478 	/* initialize eeprom parameters */
6479 	if (ixgbe_init_eeprom_params_generic(hw)) {
6480 		e_dev_err("EEPROM initialization failed\n");
6481 		return -EIO;
6482 	}
6483 
6484 	/* PF holds first pool slot */
6485 	set_bit(0, adapter->fwd_bitmask);
6486 	set_bit(__IXGBE_DOWN, &adapter->state);
6487 
6488 	/* enable locking for XDP_TX if we have more CPUs than queues */
6489 	if (nr_cpu_ids > IXGBE_MAX_XDP_QS)
6490 		static_branch_enable(&ixgbe_xdp_locking_key);
6491 
6492 	return 0;
6493 }
6494 
6495 /**
6496  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
6497  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
6498  *
6499  * Return 0 on success, negative on failure
6500  **/
6501 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
6502 {
6503 	struct device *dev = tx_ring->dev;
6504 	int orig_node = dev_to_node(dev);
6505 	int ring_node = NUMA_NO_NODE;
6506 	int size;
6507 
6508 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
6509 
6510 	if (tx_ring->q_vector)
6511 		ring_node = tx_ring->q_vector->numa_node;
6512 
6513 	tx_ring->tx_buffer_info = vmalloc_node(size, ring_node);
6514 	if (!tx_ring->tx_buffer_info)
6515 		tx_ring->tx_buffer_info = vmalloc(size);
6516 	if (!tx_ring->tx_buffer_info)
6517 		goto err;
6518 
6519 	/* round up to nearest 4K */
6520 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
6521 	tx_ring->size = ALIGN(tx_ring->size, 4096);
6522 
6523 	set_dev_node(dev, ring_node);
6524 	tx_ring->desc = dma_alloc_coherent(dev,
6525 					   tx_ring->size,
6526 					   &tx_ring->dma,
6527 					   GFP_KERNEL);
6528 	set_dev_node(dev, orig_node);
6529 	if (!tx_ring->desc)
6530 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
6531 						   &tx_ring->dma, GFP_KERNEL);
6532 	if (!tx_ring->desc)
6533 		goto err;
6534 
6535 	tx_ring->next_to_use = 0;
6536 	tx_ring->next_to_clean = 0;
6537 	return 0;
6538 
6539 err:
6540 	vfree(tx_ring->tx_buffer_info);
6541 	tx_ring->tx_buffer_info = NULL;
6542 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
6543 	return -ENOMEM;
6544 }
6545 
6546 /**
6547  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
6548  * @adapter: board private structure
6549  *
6550  * If this function returns with an error, then it's possible one or
6551  * more of the rings is populated (while the rest are not).  It is the
6552  * callers duty to clean those orphaned rings.
6553  *
6554  * Return 0 on success, negative on failure
6555  **/
6556 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
6557 {
6558 	int i, j = 0, err = 0;
6559 
6560 	for (i = 0; i < adapter->num_tx_queues; i++) {
6561 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
6562 		if (!err)
6563 			continue;
6564 
6565 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
6566 		goto err_setup_tx;
6567 	}
6568 	for (j = 0; j < adapter->num_xdp_queues; j++) {
6569 		err = ixgbe_setup_tx_resources(adapter->xdp_ring[j]);
6570 		if (!err)
6571 			continue;
6572 
6573 		e_err(probe, "Allocation for Tx Queue %u failed\n", j);
6574 		goto err_setup_tx;
6575 	}
6576 
6577 	return 0;
6578 err_setup_tx:
6579 	/* rewind the index freeing the rings as we go */
6580 	while (j--)
6581 		ixgbe_free_tx_resources(adapter->xdp_ring[j]);
6582 	while (i--)
6583 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
6584 	return err;
6585 }
6586 
6587 static int ixgbe_rx_napi_id(struct ixgbe_ring *rx_ring)
6588 {
6589 	struct ixgbe_q_vector *q_vector = rx_ring->q_vector;
6590 
6591 	return q_vector ? q_vector->napi.napi_id : 0;
6592 }
6593 
6594 /**
6595  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
6596  * @adapter: pointer to ixgbe_adapter
6597  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
6598  *
6599  * Returns 0 on success, negative on failure
6600  **/
6601 int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
6602 			     struct ixgbe_ring *rx_ring)
6603 {
6604 	struct device *dev = rx_ring->dev;
6605 	int orig_node = dev_to_node(dev);
6606 	int ring_node = NUMA_NO_NODE;
6607 	int size;
6608 
6609 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
6610 
6611 	if (rx_ring->q_vector)
6612 		ring_node = rx_ring->q_vector->numa_node;
6613 
6614 	rx_ring->rx_buffer_info = vmalloc_node(size, ring_node);
6615 	if (!rx_ring->rx_buffer_info)
6616 		rx_ring->rx_buffer_info = vmalloc(size);
6617 	if (!rx_ring->rx_buffer_info)
6618 		goto err;
6619 
6620 	/* Round up to nearest 4K */
6621 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
6622 	rx_ring->size = ALIGN(rx_ring->size, 4096);
6623 
6624 	set_dev_node(dev, ring_node);
6625 	rx_ring->desc = dma_alloc_coherent(dev,
6626 					   rx_ring->size,
6627 					   &rx_ring->dma,
6628 					   GFP_KERNEL);
6629 	set_dev_node(dev, orig_node);
6630 	if (!rx_ring->desc)
6631 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
6632 						   &rx_ring->dma, GFP_KERNEL);
6633 	if (!rx_ring->desc)
6634 		goto err;
6635 
6636 	rx_ring->next_to_clean = 0;
6637 	rx_ring->next_to_use = 0;
6638 
6639 	/* XDP RX-queue info */
6640 	if (xdp_rxq_info_reg(&rx_ring->xdp_rxq, adapter->netdev,
6641 			     rx_ring->queue_index, ixgbe_rx_napi_id(rx_ring)) < 0)
6642 		goto err;
6643 
6644 	WRITE_ONCE(rx_ring->xdp_prog, adapter->xdp_prog);
6645 
6646 	return 0;
6647 err:
6648 	vfree(rx_ring->rx_buffer_info);
6649 	rx_ring->rx_buffer_info = NULL;
6650 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
6651 	return -ENOMEM;
6652 }
6653 
6654 /**
6655  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
6656  * @adapter: board private structure
6657  *
6658  * If this function returns with an error, then it's possible one or
6659  * more of the rings is populated (while the rest are not).  It is the
6660  * callers duty to clean those orphaned rings.
6661  *
6662  * Return 0 on success, negative on failure
6663  **/
6664 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
6665 {
6666 	int i, err = 0;
6667 
6668 	for (i = 0; i < adapter->num_rx_queues; i++) {
6669 		err = ixgbe_setup_rx_resources(adapter, adapter->rx_ring[i]);
6670 		if (!err)
6671 			continue;
6672 
6673 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
6674 		goto err_setup_rx;
6675 	}
6676 
6677 #ifdef IXGBE_FCOE
6678 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
6679 	if (!err)
6680 #endif
6681 		return 0;
6682 err_setup_rx:
6683 	/* rewind the index freeing the rings as we go */
6684 	while (i--)
6685 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
6686 	return err;
6687 }
6688 
6689 /**
6690  * ixgbe_free_tx_resources - Free Tx Resources per Queue
6691  * @tx_ring: Tx descriptor ring for a specific queue
6692  *
6693  * Free all transmit software resources
6694  **/
6695 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
6696 {
6697 	ixgbe_clean_tx_ring(tx_ring);
6698 
6699 	vfree(tx_ring->tx_buffer_info);
6700 	tx_ring->tx_buffer_info = NULL;
6701 
6702 	/* if not set, then don't free */
6703 	if (!tx_ring->desc)
6704 		return;
6705 
6706 	dma_free_coherent(tx_ring->dev, tx_ring->size,
6707 			  tx_ring->desc, tx_ring->dma);
6708 
6709 	tx_ring->desc = NULL;
6710 }
6711 
6712 /**
6713  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
6714  * @adapter: board private structure
6715  *
6716  * Free all transmit software resources
6717  **/
6718 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
6719 {
6720 	int i;
6721 
6722 	for (i = 0; i < adapter->num_tx_queues; i++)
6723 		if (adapter->tx_ring[i]->desc)
6724 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
6725 	for (i = 0; i < adapter->num_xdp_queues; i++)
6726 		if (adapter->xdp_ring[i]->desc)
6727 			ixgbe_free_tx_resources(adapter->xdp_ring[i]);
6728 }
6729 
6730 /**
6731  * ixgbe_free_rx_resources - Free Rx Resources
6732  * @rx_ring: ring to clean the resources from
6733  *
6734  * Free all receive software resources
6735  **/
6736 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
6737 {
6738 	ixgbe_clean_rx_ring(rx_ring);
6739 
6740 	rx_ring->xdp_prog = NULL;
6741 	xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
6742 	vfree(rx_ring->rx_buffer_info);
6743 	rx_ring->rx_buffer_info = NULL;
6744 
6745 	/* if not set, then don't free */
6746 	if (!rx_ring->desc)
6747 		return;
6748 
6749 	dma_free_coherent(rx_ring->dev, rx_ring->size,
6750 			  rx_ring->desc, rx_ring->dma);
6751 
6752 	rx_ring->desc = NULL;
6753 }
6754 
6755 /**
6756  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
6757  * @adapter: board private structure
6758  *
6759  * Free all receive software resources
6760  **/
6761 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
6762 {
6763 	int i;
6764 
6765 #ifdef IXGBE_FCOE
6766 	ixgbe_free_fcoe_ddp_resources(adapter);
6767 
6768 #endif
6769 	for (i = 0; i < adapter->num_rx_queues; i++)
6770 		if (adapter->rx_ring[i]->desc)
6771 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
6772 }
6773 
6774 /**
6775  * ixgbe_max_xdp_frame_size - returns the maximum allowed frame size for XDP
6776  * @adapter: device handle, pointer to adapter
6777  */
6778 static int ixgbe_max_xdp_frame_size(struct ixgbe_adapter *adapter)
6779 {
6780 	if (PAGE_SIZE >= 8192 || adapter->flags2 & IXGBE_FLAG2_RX_LEGACY)
6781 		return IXGBE_RXBUFFER_2K;
6782 	else
6783 		return IXGBE_RXBUFFER_3K;
6784 }
6785 
6786 /**
6787  * ixgbe_change_mtu - Change the Maximum Transfer Unit
6788  * @netdev: network interface device structure
6789  * @new_mtu: new value for maximum frame size
6790  *
6791  * Returns 0 on success, negative on failure
6792  **/
6793 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
6794 {
6795 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6796 
6797 	if (ixgbe_enabled_xdp_adapter(adapter)) {
6798 		int new_frame_size = new_mtu + IXGBE_PKT_HDR_PAD;
6799 
6800 		if (new_frame_size > ixgbe_max_xdp_frame_size(adapter)) {
6801 			e_warn(probe, "Requested MTU size is not supported with XDP\n");
6802 			return -EINVAL;
6803 		}
6804 	}
6805 
6806 	/*
6807 	 * For 82599EB we cannot allow legacy VFs to enable their receive
6808 	 * paths when MTU greater than 1500 is configured.  So display a
6809 	 * warning that legacy VFs will be disabled.
6810 	 */
6811 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
6812 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
6813 	    (new_mtu > ETH_DATA_LEN))
6814 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
6815 
6816 	netdev_dbg(netdev, "changing MTU from %d to %d\n",
6817 		   netdev->mtu, new_mtu);
6818 
6819 	/* must set new MTU before calling down or up */
6820 	netdev->mtu = new_mtu;
6821 
6822 	if (netif_running(netdev))
6823 		ixgbe_reinit_locked(adapter);
6824 
6825 	return 0;
6826 }
6827 
6828 /**
6829  * ixgbe_open - Called when a network interface is made active
6830  * @netdev: network interface device structure
6831  *
6832  * Returns 0 on success, negative value on failure
6833  *
6834  * The open entry point is called when a network interface is made
6835  * active by the system (IFF_UP).  At this point all resources needed
6836  * for transmit and receive operations are allocated, the interrupt
6837  * handler is registered with the OS, the watchdog timer is started,
6838  * and the stack is notified that the interface is ready.
6839  **/
6840 int ixgbe_open(struct net_device *netdev)
6841 {
6842 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6843 	struct ixgbe_hw *hw = &adapter->hw;
6844 	int err, queues;
6845 
6846 	/* disallow open during test */
6847 	if (test_bit(__IXGBE_TESTING, &adapter->state))
6848 		return -EBUSY;
6849 
6850 	netif_carrier_off(netdev);
6851 
6852 	/* allocate transmit descriptors */
6853 	err = ixgbe_setup_all_tx_resources(adapter);
6854 	if (err)
6855 		goto err_setup_tx;
6856 
6857 	/* allocate receive descriptors */
6858 	err = ixgbe_setup_all_rx_resources(adapter);
6859 	if (err)
6860 		goto err_setup_rx;
6861 
6862 	ixgbe_configure(adapter);
6863 
6864 	err = ixgbe_request_irq(adapter);
6865 	if (err)
6866 		goto err_req_irq;
6867 
6868 	/* Notify the stack of the actual queue counts. */
6869 	queues = adapter->num_tx_queues;
6870 	err = netif_set_real_num_tx_queues(netdev, queues);
6871 	if (err)
6872 		goto err_set_queues;
6873 
6874 	queues = adapter->num_rx_queues;
6875 	err = netif_set_real_num_rx_queues(netdev, queues);
6876 	if (err)
6877 		goto err_set_queues;
6878 
6879 	ixgbe_ptp_init(adapter);
6880 
6881 	ixgbe_up_complete(adapter);
6882 
6883 	udp_tunnel_nic_reset_ntf(netdev);
6884 
6885 	return 0;
6886 
6887 err_set_queues:
6888 	ixgbe_free_irq(adapter);
6889 err_req_irq:
6890 	ixgbe_free_all_rx_resources(adapter);
6891 	if (hw->phy.ops.set_phy_power && !adapter->wol)
6892 		hw->phy.ops.set_phy_power(&adapter->hw, false);
6893 err_setup_rx:
6894 	ixgbe_free_all_tx_resources(adapter);
6895 err_setup_tx:
6896 	ixgbe_reset(adapter);
6897 
6898 	return err;
6899 }
6900 
6901 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
6902 {
6903 	ixgbe_ptp_suspend(adapter);
6904 
6905 	if (adapter->hw.phy.ops.enter_lplu) {
6906 		adapter->hw.phy.reset_disable = true;
6907 		ixgbe_down(adapter);
6908 		adapter->hw.phy.ops.enter_lplu(&adapter->hw);
6909 		adapter->hw.phy.reset_disable = false;
6910 	} else {
6911 		ixgbe_down(adapter);
6912 	}
6913 
6914 	ixgbe_free_irq(adapter);
6915 
6916 	ixgbe_free_all_tx_resources(adapter);
6917 	ixgbe_free_all_rx_resources(adapter);
6918 }
6919 
6920 /**
6921  * ixgbe_close - Disables a network interface
6922  * @netdev: network interface device structure
6923  *
6924  * Returns 0, this is not allowed to fail
6925  *
6926  * The close entry point is called when an interface is de-activated
6927  * by the OS.  The hardware is still under the drivers control, but
6928  * needs to be disabled.  A global MAC reset is issued to stop the
6929  * hardware, and all transmit and receive resources are freed.
6930  **/
6931 int ixgbe_close(struct net_device *netdev)
6932 {
6933 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6934 
6935 	ixgbe_ptp_stop(adapter);
6936 
6937 	if (netif_device_present(netdev))
6938 		ixgbe_close_suspend(adapter);
6939 
6940 	ixgbe_fdir_filter_exit(adapter);
6941 
6942 	ixgbe_release_hw_control(adapter);
6943 
6944 	return 0;
6945 }
6946 
6947 static int __maybe_unused ixgbe_resume(struct device *dev_d)
6948 {
6949 	struct pci_dev *pdev = to_pci_dev(dev_d);
6950 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6951 	struct net_device *netdev = adapter->netdev;
6952 	u32 err;
6953 
6954 	adapter->hw.hw_addr = adapter->io_addr;
6955 
6956 	err = pci_enable_device_mem(pdev);
6957 	if (err) {
6958 		e_dev_err("Cannot enable PCI device from suspend\n");
6959 		return err;
6960 	}
6961 	smp_mb__before_atomic();
6962 	clear_bit(__IXGBE_DISABLED, &adapter->state);
6963 	pci_set_master(pdev);
6964 
6965 	device_wakeup_disable(dev_d);
6966 
6967 	ixgbe_reset(adapter);
6968 
6969 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
6970 
6971 	rtnl_lock();
6972 	err = ixgbe_init_interrupt_scheme(adapter);
6973 	if (!err && netif_running(netdev))
6974 		err = ixgbe_open(netdev);
6975 
6976 
6977 	if (!err)
6978 		netif_device_attach(netdev);
6979 	rtnl_unlock();
6980 
6981 	return err;
6982 }
6983 
6984 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
6985 {
6986 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6987 	struct net_device *netdev = adapter->netdev;
6988 	struct ixgbe_hw *hw = &adapter->hw;
6989 	u32 ctrl;
6990 	u32 wufc = adapter->wol;
6991 
6992 	rtnl_lock();
6993 	netif_device_detach(netdev);
6994 
6995 	if (netif_running(netdev))
6996 		ixgbe_close_suspend(adapter);
6997 
6998 	ixgbe_clear_interrupt_scheme(adapter);
6999 	rtnl_unlock();
7000 
7001 	if (hw->mac.ops.stop_link_on_d3)
7002 		hw->mac.ops.stop_link_on_d3(hw);
7003 
7004 	if (wufc) {
7005 		u32 fctrl;
7006 
7007 		ixgbe_set_rx_mode(netdev);
7008 
7009 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
7010 		if (hw->mac.ops.enable_tx_laser)
7011 			hw->mac.ops.enable_tx_laser(hw);
7012 
7013 		/* enable the reception of multicast packets */
7014 		fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
7015 		fctrl |= IXGBE_FCTRL_MPE;
7016 		IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
7017 
7018 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
7019 		ctrl |= IXGBE_CTRL_GIO_DIS;
7020 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
7021 
7022 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
7023 	} else {
7024 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
7025 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
7026 	}
7027 
7028 	switch (hw->mac.type) {
7029 	case ixgbe_mac_82598EB:
7030 		pci_wake_from_d3(pdev, false);
7031 		break;
7032 	case ixgbe_mac_82599EB:
7033 	case ixgbe_mac_X540:
7034 	case ixgbe_mac_X550:
7035 	case ixgbe_mac_X550EM_x:
7036 	case ixgbe_mac_x550em_a:
7037 		pci_wake_from_d3(pdev, !!wufc);
7038 		break;
7039 	default:
7040 		break;
7041 	}
7042 
7043 	*enable_wake = !!wufc;
7044 	if (hw->phy.ops.set_phy_power && !*enable_wake)
7045 		hw->phy.ops.set_phy_power(hw, false);
7046 
7047 	ixgbe_release_hw_control(adapter);
7048 
7049 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
7050 		pci_disable_device(pdev);
7051 
7052 	return 0;
7053 }
7054 
7055 static int __maybe_unused ixgbe_suspend(struct device *dev_d)
7056 {
7057 	struct pci_dev *pdev = to_pci_dev(dev_d);
7058 	int retval;
7059 	bool wake;
7060 
7061 	retval = __ixgbe_shutdown(pdev, &wake);
7062 
7063 	device_set_wakeup_enable(dev_d, wake);
7064 
7065 	return retval;
7066 }
7067 
7068 static void ixgbe_shutdown(struct pci_dev *pdev)
7069 {
7070 	bool wake;
7071 
7072 	__ixgbe_shutdown(pdev, &wake);
7073 
7074 	if (system_state == SYSTEM_POWER_OFF) {
7075 		pci_wake_from_d3(pdev, wake);
7076 		pci_set_power_state(pdev, PCI_D3hot);
7077 	}
7078 }
7079 
7080 /**
7081  * ixgbe_update_stats - Update the board statistics counters.
7082  * @adapter: board private structure
7083  **/
7084 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
7085 {
7086 	struct net_device *netdev = adapter->netdev;
7087 	struct ixgbe_hw *hw = &adapter->hw;
7088 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
7089 	u64 total_mpc = 0;
7090 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
7091 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
7092 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
7093 	u64 alloc_rx_page = 0;
7094 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
7095 
7096 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7097 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7098 		return;
7099 
7100 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
7101 		u64 rsc_count = 0;
7102 		u64 rsc_flush = 0;
7103 		for (i = 0; i < adapter->num_rx_queues; i++) {
7104 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
7105 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
7106 		}
7107 		adapter->rsc_total_count = rsc_count;
7108 		adapter->rsc_total_flush = rsc_flush;
7109 	}
7110 
7111 	for (i = 0; i < adapter->num_rx_queues; i++) {
7112 		struct ixgbe_ring *rx_ring = READ_ONCE(adapter->rx_ring[i]);
7113 
7114 		if (!rx_ring)
7115 			continue;
7116 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
7117 		alloc_rx_page += rx_ring->rx_stats.alloc_rx_page;
7118 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
7119 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
7120 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
7121 		bytes += rx_ring->stats.bytes;
7122 		packets += rx_ring->stats.packets;
7123 	}
7124 	adapter->non_eop_descs = non_eop_descs;
7125 	adapter->alloc_rx_page = alloc_rx_page;
7126 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
7127 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
7128 	adapter->hw_csum_rx_error = hw_csum_rx_error;
7129 	netdev->stats.rx_bytes = bytes;
7130 	netdev->stats.rx_packets = packets;
7131 
7132 	bytes = 0;
7133 	packets = 0;
7134 	/* gather some stats to the adapter struct that are per queue */
7135 	for (i = 0; i < adapter->num_tx_queues; i++) {
7136 		struct ixgbe_ring *tx_ring = READ_ONCE(adapter->tx_ring[i]);
7137 
7138 		if (!tx_ring)
7139 			continue;
7140 		restart_queue += tx_ring->tx_stats.restart_queue;
7141 		tx_busy += tx_ring->tx_stats.tx_busy;
7142 		bytes += tx_ring->stats.bytes;
7143 		packets += tx_ring->stats.packets;
7144 	}
7145 	for (i = 0; i < adapter->num_xdp_queues; i++) {
7146 		struct ixgbe_ring *xdp_ring = READ_ONCE(adapter->xdp_ring[i]);
7147 
7148 		if (!xdp_ring)
7149 			continue;
7150 		restart_queue += xdp_ring->tx_stats.restart_queue;
7151 		tx_busy += xdp_ring->tx_stats.tx_busy;
7152 		bytes += xdp_ring->stats.bytes;
7153 		packets += xdp_ring->stats.packets;
7154 	}
7155 	adapter->restart_queue = restart_queue;
7156 	adapter->tx_busy = tx_busy;
7157 	netdev->stats.tx_bytes = bytes;
7158 	netdev->stats.tx_packets = packets;
7159 
7160 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
7161 
7162 	/* 8 register reads */
7163 	for (i = 0; i < 8; i++) {
7164 		/* for packet buffers not used, the register should read 0 */
7165 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
7166 		missed_rx += mpc;
7167 		hwstats->mpc[i] += mpc;
7168 		total_mpc += hwstats->mpc[i];
7169 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
7170 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
7171 		switch (hw->mac.type) {
7172 		case ixgbe_mac_82598EB:
7173 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
7174 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
7175 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
7176 			hwstats->pxonrxc[i] +=
7177 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
7178 			break;
7179 		case ixgbe_mac_82599EB:
7180 		case ixgbe_mac_X540:
7181 		case ixgbe_mac_X550:
7182 		case ixgbe_mac_X550EM_x:
7183 		case ixgbe_mac_x550em_a:
7184 			hwstats->pxonrxc[i] +=
7185 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
7186 			break;
7187 		default:
7188 			break;
7189 		}
7190 	}
7191 
7192 	/*16 register reads */
7193 	for (i = 0; i < 16; i++) {
7194 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
7195 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
7196 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
7197 		    (hw->mac.type == ixgbe_mac_X540) ||
7198 		    (hw->mac.type == ixgbe_mac_X550) ||
7199 		    (hw->mac.type == ixgbe_mac_X550EM_x) ||
7200 		    (hw->mac.type == ixgbe_mac_x550em_a)) {
7201 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
7202 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
7203 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
7204 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
7205 		}
7206 	}
7207 
7208 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
7209 	/* work around hardware counting issue */
7210 	hwstats->gprc -= missed_rx;
7211 
7212 	ixgbe_update_xoff_received(adapter);
7213 
7214 	/* 82598 hardware only has a 32 bit counter in the high register */
7215 	switch (hw->mac.type) {
7216 	case ixgbe_mac_82598EB:
7217 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
7218 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
7219 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
7220 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
7221 		break;
7222 	case ixgbe_mac_X540:
7223 	case ixgbe_mac_X550:
7224 	case ixgbe_mac_X550EM_x:
7225 	case ixgbe_mac_x550em_a:
7226 		/* OS2BMC stats are X540 and later */
7227 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
7228 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
7229 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
7230 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
7231 		fallthrough;
7232 	case ixgbe_mac_82599EB:
7233 		for (i = 0; i < 16; i++)
7234 			adapter->hw_rx_no_dma_resources +=
7235 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
7236 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
7237 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
7238 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
7239 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
7240 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
7241 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
7242 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
7243 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
7244 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
7245 #ifdef IXGBE_FCOE
7246 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
7247 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
7248 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
7249 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
7250 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
7251 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
7252 		/* Add up per cpu counters for total ddp aloc fail */
7253 		if (adapter->fcoe.ddp_pool) {
7254 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
7255 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
7256 			unsigned int cpu;
7257 			u64 noddp = 0, noddp_ext_buff = 0;
7258 			for_each_possible_cpu(cpu) {
7259 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
7260 				noddp += ddp_pool->noddp;
7261 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
7262 			}
7263 			hwstats->fcoe_noddp = noddp;
7264 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
7265 		}
7266 #endif /* IXGBE_FCOE */
7267 		break;
7268 	default:
7269 		break;
7270 	}
7271 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
7272 	hwstats->bprc += bprc;
7273 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
7274 	if (hw->mac.type == ixgbe_mac_82598EB)
7275 		hwstats->mprc -= bprc;
7276 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
7277 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
7278 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
7279 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
7280 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
7281 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
7282 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
7283 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
7284 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
7285 	hwstats->lxontxc += lxon;
7286 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
7287 	hwstats->lxofftxc += lxoff;
7288 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
7289 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
7290 	/*
7291 	 * 82598 errata - tx of flow control packets is included in tx counters
7292 	 */
7293 	xon_off_tot = lxon + lxoff;
7294 	hwstats->gptc -= xon_off_tot;
7295 	hwstats->mptc -= xon_off_tot;
7296 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
7297 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
7298 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
7299 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
7300 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
7301 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
7302 	hwstats->ptc64 -= xon_off_tot;
7303 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
7304 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
7305 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
7306 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
7307 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
7308 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
7309 
7310 	/* Fill out the OS statistics structure */
7311 	netdev->stats.multicast = hwstats->mprc;
7312 
7313 	/* Rx Errors */
7314 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
7315 	netdev->stats.rx_dropped = 0;
7316 	netdev->stats.rx_length_errors = hwstats->rlec;
7317 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
7318 	netdev->stats.rx_missed_errors = total_mpc;
7319 
7320 	/* VF Stats Collection - skip while resetting because these
7321 	 * are not clear on read and otherwise you'll sometimes get
7322 	 * crazy values.
7323 	 */
7324 	if (!test_bit(__IXGBE_RESETTING, &adapter->state)) {
7325 		for (i = 0; i < adapter->num_vfs; i++) {
7326 			UPDATE_VF_COUNTER_32bit(IXGBE_PVFGPRC(i),
7327 						adapter->vfinfo[i].last_vfstats.gprc,
7328 						adapter->vfinfo[i].vfstats.gprc);
7329 			UPDATE_VF_COUNTER_32bit(IXGBE_PVFGPTC(i),
7330 						adapter->vfinfo[i].last_vfstats.gptc,
7331 						adapter->vfinfo[i].vfstats.gptc);
7332 			UPDATE_VF_COUNTER_36bit(IXGBE_PVFGORC_LSB(i),
7333 						IXGBE_PVFGORC_MSB(i),
7334 						adapter->vfinfo[i].last_vfstats.gorc,
7335 						adapter->vfinfo[i].vfstats.gorc);
7336 			UPDATE_VF_COUNTER_36bit(IXGBE_PVFGOTC_LSB(i),
7337 						IXGBE_PVFGOTC_MSB(i),
7338 						adapter->vfinfo[i].last_vfstats.gotc,
7339 						adapter->vfinfo[i].vfstats.gotc);
7340 			UPDATE_VF_COUNTER_32bit(IXGBE_PVFMPRC(i),
7341 						adapter->vfinfo[i].last_vfstats.mprc,
7342 						adapter->vfinfo[i].vfstats.mprc);
7343 		}
7344 	}
7345 }
7346 
7347 /**
7348  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
7349  * @adapter: pointer to the device adapter structure
7350  **/
7351 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
7352 {
7353 	struct ixgbe_hw *hw = &adapter->hw;
7354 	int i;
7355 
7356 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
7357 		return;
7358 
7359 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
7360 
7361 	/* if interface is down do nothing */
7362 	if (test_bit(__IXGBE_DOWN, &adapter->state))
7363 		return;
7364 
7365 	/* do nothing if we are not using signature filters */
7366 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
7367 		return;
7368 
7369 	adapter->fdir_overflow++;
7370 
7371 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
7372 		for (i = 0; i < adapter->num_tx_queues; i++)
7373 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7374 				&(adapter->tx_ring[i]->state));
7375 		for (i = 0; i < adapter->num_xdp_queues; i++)
7376 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7377 				&adapter->xdp_ring[i]->state);
7378 		/* re-enable flow director interrupts */
7379 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
7380 	} else {
7381 		e_err(probe, "failed to finish FDIR re-initialization, "
7382 		      "ignored adding FDIR ATR filters\n");
7383 	}
7384 }
7385 
7386 /**
7387  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
7388  * @adapter: pointer to the device adapter structure
7389  *
7390  * This function serves two purposes.  First it strobes the interrupt lines
7391  * in order to make certain interrupts are occurring.  Secondly it sets the
7392  * bits needed to check for TX hangs.  As a result we should immediately
7393  * determine if a hang has occurred.
7394  */
7395 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
7396 {
7397 	struct ixgbe_hw *hw = &adapter->hw;
7398 	u64 eics = 0;
7399 	int i;
7400 
7401 	/* If we're down, removing or resetting, just bail */
7402 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7403 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7404 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7405 		return;
7406 
7407 	/* Force detection of hung controller */
7408 	if (netif_carrier_ok(adapter->netdev)) {
7409 		for (i = 0; i < adapter->num_tx_queues; i++)
7410 			set_check_for_tx_hang(adapter->tx_ring[i]);
7411 		for (i = 0; i < adapter->num_xdp_queues; i++)
7412 			set_check_for_tx_hang(adapter->xdp_ring[i]);
7413 	}
7414 
7415 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
7416 		/*
7417 		 * for legacy and MSI interrupts don't set any bits
7418 		 * that are enabled for EIAM, because this operation
7419 		 * would set *both* EIMS and EICS for any bit in EIAM
7420 		 */
7421 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
7422 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
7423 	} else {
7424 		/* get one bit for every active tx/rx interrupt vector */
7425 		for (i = 0; i < adapter->num_q_vectors; i++) {
7426 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
7427 			if (qv->rx.ring || qv->tx.ring)
7428 				eics |= BIT_ULL(i);
7429 		}
7430 	}
7431 
7432 	/* Cause software interrupt to ensure rings are cleaned */
7433 	ixgbe_irq_rearm_queues(adapter, eics);
7434 }
7435 
7436 /**
7437  * ixgbe_watchdog_update_link - update the link status
7438  * @adapter: pointer to the device adapter structure
7439  **/
7440 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
7441 {
7442 	struct ixgbe_hw *hw = &adapter->hw;
7443 	u32 link_speed = adapter->link_speed;
7444 	bool link_up = adapter->link_up;
7445 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
7446 
7447 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
7448 		return;
7449 
7450 	if (hw->mac.ops.check_link) {
7451 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
7452 	} else {
7453 		/* always assume link is up, if no check link function */
7454 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
7455 		link_up = true;
7456 	}
7457 
7458 	if (adapter->ixgbe_ieee_pfc)
7459 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
7460 
7461 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
7462 		hw->mac.ops.fc_enable(hw);
7463 		ixgbe_set_rx_drop_en(adapter);
7464 	}
7465 
7466 	if (link_up ||
7467 	    time_after(jiffies, (adapter->link_check_timeout +
7468 				 IXGBE_TRY_LINK_TIMEOUT))) {
7469 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
7470 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
7471 		IXGBE_WRITE_FLUSH(hw);
7472 	}
7473 
7474 	adapter->link_up = link_up;
7475 	adapter->link_speed = link_speed;
7476 }
7477 
7478 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
7479 {
7480 #ifdef CONFIG_IXGBE_DCB
7481 	struct net_device *netdev = adapter->netdev;
7482 	struct dcb_app app = {
7483 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
7484 			      .protocol = 0,
7485 			     };
7486 	u8 up = 0;
7487 
7488 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
7489 		up = dcb_ieee_getapp_mask(netdev, &app);
7490 
7491 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
7492 #endif
7493 }
7494 
7495 /**
7496  * ixgbe_watchdog_link_is_up - update netif_carrier status and
7497  *                             print link up message
7498  * @adapter: pointer to the device adapter structure
7499  **/
7500 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
7501 {
7502 	struct net_device *netdev = adapter->netdev;
7503 	struct ixgbe_hw *hw = &adapter->hw;
7504 	u32 link_speed = adapter->link_speed;
7505 	const char *speed_str;
7506 	bool flow_rx, flow_tx;
7507 
7508 	/* only continue if link was previously down */
7509 	if (netif_carrier_ok(netdev))
7510 		return;
7511 
7512 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
7513 
7514 	switch (hw->mac.type) {
7515 	case ixgbe_mac_82598EB: {
7516 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
7517 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
7518 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
7519 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
7520 	}
7521 		break;
7522 	case ixgbe_mac_X540:
7523 	case ixgbe_mac_X550:
7524 	case ixgbe_mac_X550EM_x:
7525 	case ixgbe_mac_x550em_a:
7526 	case ixgbe_mac_82599EB: {
7527 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
7528 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
7529 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
7530 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
7531 	}
7532 		break;
7533 	default:
7534 		flow_tx = false;
7535 		flow_rx = false;
7536 		break;
7537 	}
7538 
7539 	adapter->last_rx_ptp_check = jiffies;
7540 
7541 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7542 		ixgbe_ptp_start_cyclecounter(adapter);
7543 
7544 	switch (link_speed) {
7545 	case IXGBE_LINK_SPEED_10GB_FULL:
7546 		speed_str = "10 Gbps";
7547 		break;
7548 	case IXGBE_LINK_SPEED_5GB_FULL:
7549 		speed_str = "5 Gbps";
7550 		break;
7551 	case IXGBE_LINK_SPEED_2_5GB_FULL:
7552 		speed_str = "2.5 Gbps";
7553 		break;
7554 	case IXGBE_LINK_SPEED_1GB_FULL:
7555 		speed_str = "1 Gbps";
7556 		break;
7557 	case IXGBE_LINK_SPEED_100_FULL:
7558 		speed_str = "100 Mbps";
7559 		break;
7560 	case IXGBE_LINK_SPEED_10_FULL:
7561 		speed_str = "10 Mbps";
7562 		break;
7563 	default:
7564 		speed_str = "unknown speed";
7565 		break;
7566 	}
7567 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
7568 	       ((flow_rx && flow_tx) ? "RX/TX" :
7569 	       (flow_rx ? "RX" :
7570 	       (flow_tx ? "TX" : "None"))));
7571 
7572 	netif_carrier_on(netdev);
7573 	ixgbe_check_vf_rate_limit(adapter);
7574 
7575 	/* enable transmits */
7576 	netif_tx_wake_all_queues(adapter->netdev);
7577 
7578 	/* update the default user priority for VFs */
7579 	ixgbe_update_default_up(adapter);
7580 
7581 	/* ping all the active vfs to let them know link has changed */
7582 	ixgbe_ping_all_vfs(adapter);
7583 }
7584 
7585 /**
7586  * ixgbe_watchdog_link_is_down - update netif_carrier status and
7587  *                               print link down message
7588  * @adapter: pointer to the adapter structure
7589  **/
7590 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
7591 {
7592 	struct net_device *netdev = adapter->netdev;
7593 	struct ixgbe_hw *hw = &adapter->hw;
7594 
7595 	adapter->link_up = false;
7596 	adapter->link_speed = 0;
7597 
7598 	/* only continue if link was up previously */
7599 	if (!netif_carrier_ok(netdev))
7600 		return;
7601 
7602 	/* poll for SFP+ cable when link is down */
7603 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
7604 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
7605 
7606 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7607 		ixgbe_ptp_start_cyclecounter(adapter);
7608 
7609 	e_info(drv, "NIC Link is Down\n");
7610 	netif_carrier_off(netdev);
7611 
7612 	/* ping all the active vfs to let them know link has changed */
7613 	ixgbe_ping_all_vfs(adapter);
7614 }
7615 
7616 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
7617 {
7618 	int i;
7619 
7620 	for (i = 0; i < adapter->num_tx_queues; i++) {
7621 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
7622 
7623 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
7624 			return true;
7625 	}
7626 
7627 	for (i = 0; i < adapter->num_xdp_queues; i++) {
7628 		struct ixgbe_ring *ring = adapter->xdp_ring[i];
7629 
7630 		if (ring->next_to_use != ring->next_to_clean)
7631 			return true;
7632 	}
7633 
7634 	return false;
7635 }
7636 
7637 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
7638 {
7639 	struct ixgbe_hw *hw = &adapter->hw;
7640 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
7641 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
7642 
7643 	int i, j;
7644 
7645 	if (!adapter->num_vfs)
7646 		return false;
7647 
7648 	/* resetting the PF is only needed for MAC before X550 */
7649 	if (hw->mac.type >= ixgbe_mac_X550)
7650 		return false;
7651 
7652 	for (i = 0; i < adapter->num_vfs; i++) {
7653 		for (j = 0; j < q_per_pool; j++) {
7654 			u32 h, t;
7655 
7656 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
7657 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
7658 
7659 			if (h != t)
7660 				return true;
7661 		}
7662 	}
7663 
7664 	return false;
7665 }
7666 
7667 /**
7668  * ixgbe_watchdog_flush_tx - flush queues on link down
7669  * @adapter: pointer to the device adapter structure
7670  **/
7671 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
7672 {
7673 	if (!netif_carrier_ok(adapter->netdev)) {
7674 		if (ixgbe_ring_tx_pending(adapter) ||
7675 		    ixgbe_vf_tx_pending(adapter)) {
7676 			/* We've lost link, so the controller stops DMA,
7677 			 * but we've got queued Tx work that's never going
7678 			 * to get done, so reset controller to flush Tx.
7679 			 * (Do the reset outside of interrupt context).
7680 			 */
7681 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
7682 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
7683 		}
7684 	}
7685 }
7686 
7687 #ifdef CONFIG_PCI_IOV
7688 static void ixgbe_bad_vf_abort(struct ixgbe_adapter *adapter, u32 vf)
7689 {
7690 	struct ixgbe_hw *hw = &adapter->hw;
7691 
7692 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
7693 	    adapter->flags2 & IXGBE_FLAG2_AUTO_DISABLE_VF) {
7694 		adapter->vfinfo[vf].primary_abort_count++;
7695 		if (adapter->vfinfo[vf].primary_abort_count ==
7696 		    IXGBE_PRIMARY_ABORT_LIMIT) {
7697 			ixgbe_set_vf_link_state(adapter, vf,
7698 						IFLA_VF_LINK_STATE_DISABLE);
7699 			adapter->vfinfo[vf].primary_abort_count = 0;
7700 
7701 			e_info(drv,
7702 			       "Malicious Driver Detection event detected on PF %d VF %d MAC: %pM mdd-disable-vf=on",
7703 			       hw->bus.func, vf,
7704 			       adapter->vfinfo[vf].vf_mac_addresses);
7705 		}
7706 	}
7707 }
7708 
7709 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
7710 {
7711 	struct ixgbe_hw *hw = &adapter->hw;
7712 	struct pci_dev *pdev = adapter->pdev;
7713 	unsigned int vf;
7714 	u32 gpc;
7715 
7716 	if (!(netif_carrier_ok(adapter->netdev)))
7717 		return;
7718 
7719 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
7720 	if (gpc) /* If incrementing then no need for the check below */
7721 		return;
7722 	/* Check to see if a bad DMA write target from an errant or
7723 	 * malicious VF has caused a PCIe error.  If so then we can
7724 	 * issue a VFLR to the offending VF(s) and then resume without
7725 	 * requesting a full slot reset.
7726 	 */
7727 
7728 	if (!pdev)
7729 		return;
7730 
7731 	/* check status reg for all VFs owned by this PF */
7732 	for (vf = 0; vf < adapter->num_vfs; ++vf) {
7733 		struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
7734 		u16 status_reg;
7735 
7736 		if (!vfdev)
7737 			continue;
7738 		pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
7739 		if (status_reg != IXGBE_FAILED_READ_CFG_WORD &&
7740 		    status_reg & PCI_STATUS_REC_MASTER_ABORT) {
7741 			ixgbe_bad_vf_abort(adapter, vf);
7742 			pcie_flr(vfdev);
7743 		}
7744 	}
7745 }
7746 
7747 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
7748 {
7749 	u32 ssvpc;
7750 
7751 	/* Do not perform spoof check for 82598 or if not in IOV mode */
7752 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7753 	    adapter->num_vfs == 0)
7754 		return;
7755 
7756 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
7757 
7758 	/*
7759 	 * ssvpc register is cleared on read, if zero then no
7760 	 * spoofed packets in the last interval.
7761 	 */
7762 	if (!ssvpc)
7763 		return;
7764 
7765 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
7766 }
7767 #else
7768 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
7769 {
7770 }
7771 
7772 static void
7773 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
7774 {
7775 }
7776 #endif /* CONFIG_PCI_IOV */
7777 
7778 
7779 /**
7780  * ixgbe_watchdog_subtask - check and bring link up
7781  * @adapter: pointer to the device adapter structure
7782  **/
7783 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
7784 {
7785 	/* if interface is down, removing or resetting, do nothing */
7786 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7787 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7788 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7789 		return;
7790 
7791 	ixgbe_watchdog_update_link(adapter);
7792 
7793 	if (adapter->link_up)
7794 		ixgbe_watchdog_link_is_up(adapter);
7795 	else
7796 		ixgbe_watchdog_link_is_down(adapter);
7797 
7798 	ixgbe_check_for_bad_vf(adapter);
7799 	ixgbe_spoof_check(adapter);
7800 	ixgbe_update_stats(adapter);
7801 
7802 	ixgbe_watchdog_flush_tx(adapter);
7803 }
7804 
7805 /**
7806  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
7807  * @adapter: the ixgbe adapter structure
7808  **/
7809 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
7810 {
7811 	struct ixgbe_hw *hw = &adapter->hw;
7812 	s32 err;
7813 
7814 	/* not searching for SFP so there is nothing to do here */
7815 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
7816 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7817 		return;
7818 
7819 	if (adapter->sfp_poll_time &&
7820 	    time_after(adapter->sfp_poll_time, jiffies))
7821 		return; /* If not yet time to poll for SFP */
7822 
7823 	/* someone else is in init, wait until next service event */
7824 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7825 		return;
7826 
7827 	adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
7828 
7829 	err = hw->phy.ops.identify_sfp(hw);
7830 	if (err == -EOPNOTSUPP)
7831 		goto sfp_out;
7832 
7833 	if (err == -ENOENT) {
7834 		/* If no cable is present, then we need to reset
7835 		 * the next time we find a good cable. */
7836 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
7837 	}
7838 
7839 	/* exit on error */
7840 	if (err)
7841 		goto sfp_out;
7842 
7843 	/* exit if reset not needed */
7844 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7845 		goto sfp_out;
7846 
7847 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
7848 
7849 	/*
7850 	 * A module may be identified correctly, but the EEPROM may not have
7851 	 * support for that module.  setup_sfp() will fail in that case, so
7852 	 * we should not allow that module to load.
7853 	 */
7854 	if (hw->mac.type == ixgbe_mac_82598EB)
7855 		err = hw->phy.ops.reset(hw);
7856 	else
7857 		err = hw->mac.ops.setup_sfp(hw);
7858 
7859 	if (err == -EOPNOTSUPP)
7860 		goto sfp_out;
7861 
7862 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
7863 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
7864 
7865 sfp_out:
7866 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7867 
7868 	if (err == -EOPNOTSUPP &&
7869 	    adapter->netdev->reg_state == NETREG_REGISTERED) {
7870 		e_dev_err("failed to initialize because an unsupported "
7871 			  "SFP+ module type was detected.\n");
7872 		e_dev_err("Reload the driver after installing a "
7873 			  "supported module.\n");
7874 		unregister_netdev(adapter->netdev);
7875 	}
7876 }
7877 
7878 /**
7879  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
7880  * @adapter: the ixgbe adapter structure
7881  **/
7882 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
7883 {
7884 	struct ixgbe_hw *hw = &adapter->hw;
7885 	u32 cap_speed;
7886 	u32 speed;
7887 	bool autoneg = false;
7888 
7889 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
7890 		return;
7891 
7892 	/* someone else is in init, wait until next service event */
7893 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7894 		return;
7895 
7896 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
7897 
7898 	hw->mac.ops.get_link_capabilities(hw, &cap_speed, &autoneg);
7899 
7900 	/* advertise highest capable link speed */
7901 	if (!autoneg && (cap_speed & IXGBE_LINK_SPEED_10GB_FULL))
7902 		speed = IXGBE_LINK_SPEED_10GB_FULL;
7903 	else
7904 		speed = cap_speed & (IXGBE_LINK_SPEED_10GB_FULL |
7905 				     IXGBE_LINK_SPEED_1GB_FULL);
7906 
7907 	if (hw->mac.ops.setup_link)
7908 		hw->mac.ops.setup_link(hw, speed, true);
7909 
7910 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
7911 	adapter->link_check_timeout = jiffies;
7912 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7913 }
7914 
7915 /**
7916  * ixgbe_service_timer - Timer Call-back
7917  * @t: pointer to timer_list structure
7918  **/
7919 static void ixgbe_service_timer(struct timer_list *t)
7920 {
7921 	struct ixgbe_adapter *adapter = from_timer(adapter, t, service_timer);
7922 	unsigned long next_event_offset;
7923 
7924 	/* poll faster when waiting for link */
7925 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
7926 		next_event_offset = HZ / 10;
7927 	else
7928 		next_event_offset = HZ * 2;
7929 
7930 	/* Reset the timer */
7931 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
7932 
7933 	ixgbe_service_event_schedule(adapter);
7934 }
7935 
7936 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
7937 {
7938 	struct ixgbe_hw *hw = &adapter->hw;
7939 	bool overtemp;
7940 
7941 	if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
7942 		return;
7943 
7944 	adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
7945 
7946 	if (!hw->phy.ops.handle_lasi)
7947 		return;
7948 
7949 	hw->phy.ops.handle_lasi(&adapter->hw, &overtemp);
7950 	if (overtemp)
7951 		e_crit(drv, "%s\n", ixgbe_overheat_msg);
7952 }
7953 
7954 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
7955 {
7956 	if (!test_and_clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state))
7957 		return;
7958 
7959 	rtnl_lock();
7960 	/* If we're already down, removing or resetting, just bail */
7961 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7962 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7963 	    test_bit(__IXGBE_RESETTING, &adapter->state)) {
7964 		rtnl_unlock();
7965 		return;
7966 	}
7967 
7968 	ixgbe_dump(adapter);
7969 	netdev_err(adapter->netdev, "Reset adapter\n");
7970 	adapter->tx_timeout_count++;
7971 
7972 	ixgbe_reinit_locked(adapter);
7973 	rtnl_unlock();
7974 }
7975 
7976 /**
7977  * ixgbe_check_fw_error - Check firmware for errors
7978  * @adapter: the adapter private structure
7979  *
7980  * Check firmware errors in register FWSM
7981  */
7982 static bool ixgbe_check_fw_error(struct ixgbe_adapter *adapter)
7983 {
7984 	struct ixgbe_hw *hw = &adapter->hw;
7985 	u32 fwsm;
7986 
7987 	/* read fwsm.ext_err_ind register and log errors */
7988 	fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
7989 
7990 	if (fwsm & IXGBE_FWSM_EXT_ERR_IND_MASK ||
7991 	    !(fwsm & IXGBE_FWSM_FW_VAL_BIT))
7992 		e_dev_warn("Warning firmware error detected FWSM: 0x%08X\n",
7993 			   fwsm);
7994 
7995 	if (hw->mac.ops.fw_recovery_mode && hw->mac.ops.fw_recovery_mode(hw)) {
7996 		e_dev_err("Firmware recovery mode detected. Limiting functionality. Refer to the Intel(R) Ethernet Adapters and Devices User Guide for details on firmware recovery mode.\n");
7997 		return true;
7998 	}
7999 
8000 	return false;
8001 }
8002 
8003 /**
8004  * ixgbe_service_task - manages and runs subtasks
8005  * @work: pointer to work_struct containing our data
8006  **/
8007 static void ixgbe_service_task(struct work_struct *work)
8008 {
8009 	struct ixgbe_adapter *adapter = container_of(work,
8010 						     struct ixgbe_adapter,
8011 						     service_task);
8012 	if (ixgbe_removed(adapter->hw.hw_addr)) {
8013 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
8014 			rtnl_lock();
8015 			ixgbe_down(adapter);
8016 			rtnl_unlock();
8017 		}
8018 		ixgbe_service_event_complete(adapter);
8019 		return;
8020 	}
8021 	if (ixgbe_check_fw_error(adapter)) {
8022 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
8023 			unregister_netdev(adapter->netdev);
8024 		ixgbe_service_event_complete(adapter);
8025 		return;
8026 	}
8027 	ixgbe_reset_subtask(adapter);
8028 	ixgbe_phy_interrupt_subtask(adapter);
8029 	ixgbe_sfp_detection_subtask(adapter);
8030 	ixgbe_sfp_link_config_subtask(adapter);
8031 	ixgbe_check_overtemp_subtask(adapter);
8032 	ixgbe_watchdog_subtask(adapter);
8033 	ixgbe_fdir_reinit_subtask(adapter);
8034 	ixgbe_check_hang_subtask(adapter);
8035 
8036 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
8037 		ixgbe_ptp_overflow_check(adapter);
8038 		if (adapter->flags & IXGBE_FLAG_RX_HWTSTAMP_IN_REGISTER)
8039 			ixgbe_ptp_rx_hang(adapter);
8040 		ixgbe_ptp_tx_hang(adapter);
8041 	}
8042 
8043 	ixgbe_service_event_complete(adapter);
8044 }
8045 
8046 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
8047 		     struct ixgbe_tx_buffer *first,
8048 		     u8 *hdr_len,
8049 		     struct ixgbe_ipsec_tx_data *itd)
8050 {
8051 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
8052 	struct sk_buff *skb = first->skb;
8053 	union {
8054 		struct iphdr *v4;
8055 		struct ipv6hdr *v6;
8056 		unsigned char *hdr;
8057 	} ip;
8058 	union {
8059 		struct tcphdr *tcp;
8060 		struct udphdr *udp;
8061 		unsigned char *hdr;
8062 	} l4;
8063 	u32 paylen, l4_offset;
8064 	u32 fceof_saidx = 0;
8065 	int err;
8066 
8067 	if (skb->ip_summed != CHECKSUM_PARTIAL)
8068 		return 0;
8069 
8070 	if (!skb_is_gso(skb))
8071 		return 0;
8072 
8073 	err = skb_cow_head(skb, 0);
8074 	if (err < 0)
8075 		return err;
8076 
8077 	if (eth_p_mpls(first->protocol))
8078 		ip.hdr = skb_inner_network_header(skb);
8079 	else
8080 		ip.hdr = skb_network_header(skb);
8081 	l4.hdr = skb_checksum_start(skb);
8082 
8083 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
8084 	type_tucmd = (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_L4) ?
8085 		      IXGBE_ADVTXD_TUCMD_L4T_UDP : IXGBE_ADVTXD_TUCMD_L4T_TCP;
8086 
8087 	/* initialize outer IP header fields */
8088 	if (ip.v4->version == 4) {
8089 		unsigned char *csum_start = skb_checksum_start(skb);
8090 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
8091 		int len = csum_start - trans_start;
8092 
8093 		/* IP header will have to cancel out any data that
8094 		 * is not a part of the outer IP header, so set to
8095 		 * a reverse csum if needed, else init check to 0.
8096 		 */
8097 		ip.v4->check = (skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL) ?
8098 					   csum_fold(csum_partial(trans_start,
8099 								  len, 0)) : 0;
8100 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
8101 
8102 		ip.v4->tot_len = 0;
8103 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
8104 				   IXGBE_TX_FLAGS_CSUM |
8105 				   IXGBE_TX_FLAGS_IPV4;
8106 	} else {
8107 		ip.v6->payload_len = 0;
8108 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
8109 				   IXGBE_TX_FLAGS_CSUM;
8110 	}
8111 
8112 	/* determine offset of inner transport header */
8113 	l4_offset = l4.hdr - skb->data;
8114 
8115 	/* remove payload length from inner checksum */
8116 	paylen = skb->len - l4_offset;
8117 
8118 	if (type_tucmd & IXGBE_ADVTXD_TUCMD_L4T_TCP) {
8119 		/* compute length of segmentation header */
8120 		*hdr_len = (l4.tcp->doff * 4) + l4_offset;
8121 		csum_replace_by_diff(&l4.tcp->check,
8122 				     (__force __wsum)htonl(paylen));
8123 	} else {
8124 		/* compute length of segmentation header */
8125 		*hdr_len = sizeof(*l4.udp) + l4_offset;
8126 		csum_replace_by_diff(&l4.udp->check,
8127 				     (__force __wsum)htonl(paylen));
8128 	}
8129 
8130 	/* update gso size and bytecount with header size */
8131 	first->gso_segs = skb_shinfo(skb)->gso_segs;
8132 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
8133 
8134 	/* mss_l4len_id: use 0 as index for TSO */
8135 	mss_l4len_idx = (*hdr_len - l4_offset) << IXGBE_ADVTXD_L4LEN_SHIFT;
8136 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
8137 
8138 	fceof_saidx |= itd->sa_idx;
8139 	type_tucmd |= itd->flags | itd->trailer_len;
8140 
8141 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
8142 	vlan_macip_lens = l4.hdr - ip.hdr;
8143 	vlan_macip_lens |= (ip.hdr - skb->data) << IXGBE_ADVTXD_MACLEN_SHIFT;
8144 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
8145 
8146 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd,
8147 			  mss_l4len_idx);
8148 
8149 	return 1;
8150 }
8151 
8152 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
8153 			  struct ixgbe_tx_buffer *first,
8154 			  struct ixgbe_ipsec_tx_data *itd)
8155 {
8156 	struct sk_buff *skb = first->skb;
8157 	u32 vlan_macip_lens = 0;
8158 	u32 fceof_saidx = 0;
8159 	u32 type_tucmd = 0;
8160 
8161 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
8162 csum_failed:
8163 		if (!(first->tx_flags & (IXGBE_TX_FLAGS_HW_VLAN |
8164 					 IXGBE_TX_FLAGS_CC)))
8165 			return;
8166 		goto no_csum;
8167 	}
8168 
8169 	switch (skb->csum_offset) {
8170 	case offsetof(struct tcphdr, check):
8171 		type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
8172 		fallthrough;
8173 	case offsetof(struct udphdr, check):
8174 		break;
8175 	case offsetof(struct sctphdr, checksum):
8176 		/* validate that this is actually an SCTP request */
8177 		if (skb_csum_is_sctp(skb)) {
8178 			type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_SCTP;
8179 			break;
8180 		}
8181 		fallthrough;
8182 	default:
8183 		skb_checksum_help(skb);
8184 		goto csum_failed;
8185 	}
8186 
8187 	/* update TX checksum flag */
8188 	first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
8189 	vlan_macip_lens = skb_checksum_start_offset(skb) -
8190 			  skb_network_offset(skb);
8191 no_csum:
8192 	/* vlan_macip_lens: MACLEN, VLAN tag */
8193 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
8194 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
8195 
8196 	fceof_saidx |= itd->sa_idx;
8197 	type_tucmd |= itd->flags | itd->trailer_len;
8198 
8199 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd, 0);
8200 }
8201 
8202 #define IXGBE_SET_FLAG(_input, _flag, _result) \
8203 	((_flag <= _result) ? \
8204 	 ((u32)(_input & _flag) * (_result / _flag)) : \
8205 	 ((u32)(_input & _flag) / (_flag / _result)))
8206 
8207 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
8208 {
8209 	/* set type for advanced descriptor with frame checksum insertion */
8210 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
8211 		       IXGBE_ADVTXD_DCMD_DEXT |
8212 		       IXGBE_ADVTXD_DCMD_IFCS;
8213 
8214 	/* set HW vlan bit if vlan is present */
8215 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
8216 				   IXGBE_ADVTXD_DCMD_VLE);
8217 
8218 	/* set segmentation enable bits for TSO/FSO */
8219 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
8220 				   IXGBE_ADVTXD_DCMD_TSE);
8221 
8222 	/* set timestamp bit if present */
8223 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
8224 				   IXGBE_ADVTXD_MAC_TSTAMP);
8225 
8226 	/* insert frame checksum */
8227 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
8228 
8229 	return cmd_type;
8230 }
8231 
8232 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
8233 				   u32 tx_flags, unsigned int paylen)
8234 {
8235 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
8236 
8237 	/* enable L4 checksum for TSO and TX checksum offload */
8238 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8239 					IXGBE_TX_FLAGS_CSUM,
8240 					IXGBE_ADVTXD_POPTS_TXSM);
8241 
8242 	/* enable IPv4 checksum for TSO */
8243 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8244 					IXGBE_TX_FLAGS_IPV4,
8245 					IXGBE_ADVTXD_POPTS_IXSM);
8246 
8247 	/* enable IPsec */
8248 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8249 					IXGBE_TX_FLAGS_IPSEC,
8250 					IXGBE_ADVTXD_POPTS_IPSEC);
8251 
8252 	/*
8253 	 * Check Context must be set if Tx switch is enabled, which it
8254 	 * always is for case where virtual functions are running
8255 	 */
8256 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8257 					IXGBE_TX_FLAGS_CC,
8258 					IXGBE_ADVTXD_CC);
8259 
8260 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
8261 }
8262 
8263 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
8264 {
8265 	if (!netif_subqueue_try_stop(tx_ring->netdev, tx_ring->queue_index,
8266 				     ixgbe_desc_unused(tx_ring), size))
8267 		return -EBUSY;
8268 
8269 	++tx_ring->tx_stats.restart_queue;
8270 	return 0;
8271 }
8272 
8273 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
8274 {
8275 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
8276 		return 0;
8277 
8278 	return __ixgbe_maybe_stop_tx(tx_ring, size);
8279 }
8280 
8281 static int ixgbe_tx_map(struct ixgbe_ring *tx_ring,
8282 			struct ixgbe_tx_buffer *first,
8283 			const u8 hdr_len)
8284 {
8285 	struct sk_buff *skb = first->skb;
8286 	struct ixgbe_tx_buffer *tx_buffer;
8287 	union ixgbe_adv_tx_desc *tx_desc;
8288 	skb_frag_t *frag;
8289 	dma_addr_t dma;
8290 	unsigned int data_len, size;
8291 	u32 tx_flags = first->tx_flags;
8292 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
8293 	u16 i = tx_ring->next_to_use;
8294 
8295 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
8296 
8297 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
8298 
8299 	size = skb_headlen(skb);
8300 	data_len = skb->data_len;
8301 
8302 #ifdef IXGBE_FCOE
8303 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
8304 		if (data_len < sizeof(struct fcoe_crc_eof)) {
8305 			size -= sizeof(struct fcoe_crc_eof) - data_len;
8306 			data_len = 0;
8307 		} else {
8308 			data_len -= sizeof(struct fcoe_crc_eof);
8309 		}
8310 	}
8311 
8312 #endif
8313 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
8314 
8315 	tx_buffer = first;
8316 
8317 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
8318 		if (dma_mapping_error(tx_ring->dev, dma))
8319 			goto dma_error;
8320 
8321 		/* record length, and DMA address */
8322 		dma_unmap_len_set(tx_buffer, len, size);
8323 		dma_unmap_addr_set(tx_buffer, dma, dma);
8324 
8325 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
8326 
8327 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
8328 			tx_desc->read.cmd_type_len =
8329 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
8330 
8331 			i++;
8332 			tx_desc++;
8333 			if (i == tx_ring->count) {
8334 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8335 				i = 0;
8336 			}
8337 			tx_desc->read.olinfo_status = 0;
8338 
8339 			dma += IXGBE_MAX_DATA_PER_TXD;
8340 			size -= IXGBE_MAX_DATA_PER_TXD;
8341 
8342 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
8343 		}
8344 
8345 		if (likely(!data_len))
8346 			break;
8347 
8348 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
8349 
8350 		i++;
8351 		tx_desc++;
8352 		if (i == tx_ring->count) {
8353 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8354 			i = 0;
8355 		}
8356 		tx_desc->read.olinfo_status = 0;
8357 
8358 #ifdef IXGBE_FCOE
8359 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
8360 #else
8361 		size = skb_frag_size(frag);
8362 #endif
8363 		data_len -= size;
8364 
8365 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
8366 				       DMA_TO_DEVICE);
8367 
8368 		tx_buffer = &tx_ring->tx_buffer_info[i];
8369 	}
8370 
8371 	/* write last descriptor with RS and EOP bits */
8372 	cmd_type |= size | IXGBE_TXD_CMD;
8373 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8374 
8375 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
8376 
8377 	/* set the timestamp */
8378 	first->time_stamp = jiffies;
8379 
8380 	skb_tx_timestamp(skb);
8381 
8382 	/*
8383 	 * Force memory writes to complete before letting h/w know there
8384 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
8385 	 * memory model archs, such as IA-64).
8386 	 *
8387 	 * We also need this memory barrier to make certain all of the
8388 	 * status bits have been updated before next_to_watch is written.
8389 	 */
8390 	wmb();
8391 
8392 	/* set next_to_watch value indicating a packet is present */
8393 	first->next_to_watch = tx_desc;
8394 
8395 	i++;
8396 	if (i == tx_ring->count)
8397 		i = 0;
8398 
8399 	tx_ring->next_to_use = i;
8400 
8401 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
8402 
8403 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more()) {
8404 		writel(i, tx_ring->tail);
8405 	}
8406 
8407 	return 0;
8408 dma_error:
8409 	dev_err(tx_ring->dev, "TX DMA map failed\n");
8410 
8411 	/* clear dma mappings for failed tx_buffer_info map */
8412 	for (;;) {
8413 		tx_buffer = &tx_ring->tx_buffer_info[i];
8414 		if (dma_unmap_len(tx_buffer, len))
8415 			dma_unmap_page(tx_ring->dev,
8416 				       dma_unmap_addr(tx_buffer, dma),
8417 				       dma_unmap_len(tx_buffer, len),
8418 				       DMA_TO_DEVICE);
8419 		dma_unmap_len_set(tx_buffer, len, 0);
8420 		if (tx_buffer == first)
8421 			break;
8422 		if (i == 0)
8423 			i += tx_ring->count;
8424 		i--;
8425 	}
8426 
8427 	dev_kfree_skb_any(first->skb);
8428 	first->skb = NULL;
8429 
8430 	tx_ring->next_to_use = i;
8431 
8432 	return -1;
8433 }
8434 
8435 static void ixgbe_atr(struct ixgbe_ring *ring,
8436 		      struct ixgbe_tx_buffer *first)
8437 {
8438 	struct ixgbe_q_vector *q_vector = ring->q_vector;
8439 	union ixgbe_atr_hash_dword input = { .dword = 0 };
8440 	union ixgbe_atr_hash_dword common = { .dword = 0 };
8441 	union {
8442 		unsigned char *network;
8443 		struct iphdr *ipv4;
8444 		struct ipv6hdr *ipv6;
8445 	} hdr;
8446 	struct tcphdr *th;
8447 	unsigned int hlen;
8448 	struct sk_buff *skb;
8449 	__be16 vlan_id;
8450 	int l4_proto;
8451 
8452 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
8453 	if (!q_vector)
8454 		return;
8455 
8456 	/* do nothing if sampling is disabled */
8457 	if (!ring->atr_sample_rate)
8458 		return;
8459 
8460 	ring->atr_count++;
8461 
8462 	/* currently only IPv4/IPv6 with TCP is supported */
8463 	if ((first->protocol != htons(ETH_P_IP)) &&
8464 	    (first->protocol != htons(ETH_P_IPV6)))
8465 		return;
8466 
8467 	/* snag network header to get L4 type and address */
8468 	skb = first->skb;
8469 	hdr.network = skb_network_header(skb);
8470 	if (unlikely(hdr.network <= skb->data))
8471 		return;
8472 	if (skb->encapsulation &&
8473 	    first->protocol == htons(ETH_P_IP) &&
8474 	    hdr.ipv4->protocol == IPPROTO_UDP) {
8475 		struct ixgbe_adapter *adapter = q_vector->adapter;
8476 
8477 		if (unlikely(skb_tail_pointer(skb) < hdr.network +
8478 			     vxlan_headroom(0)))
8479 			return;
8480 
8481 		/* verify the port is recognized as VXLAN */
8482 		if (adapter->vxlan_port &&
8483 		    udp_hdr(skb)->dest == adapter->vxlan_port)
8484 			hdr.network = skb_inner_network_header(skb);
8485 
8486 		if (adapter->geneve_port &&
8487 		    udp_hdr(skb)->dest == adapter->geneve_port)
8488 			hdr.network = skb_inner_network_header(skb);
8489 	}
8490 
8491 	/* Make sure we have at least [minimum IPv4 header + TCP]
8492 	 * or [IPv6 header] bytes
8493 	 */
8494 	if (unlikely(skb_tail_pointer(skb) < hdr.network + 40))
8495 		return;
8496 
8497 	/* Currently only IPv4/IPv6 with TCP is supported */
8498 	switch (hdr.ipv4->version) {
8499 	case IPVERSION:
8500 		/* access ihl as u8 to avoid unaligned access on ia64 */
8501 		hlen = (hdr.network[0] & 0x0F) << 2;
8502 		l4_proto = hdr.ipv4->protocol;
8503 		break;
8504 	case 6:
8505 		hlen = hdr.network - skb->data;
8506 		l4_proto = ipv6_find_hdr(skb, &hlen, IPPROTO_TCP, NULL, NULL);
8507 		hlen -= hdr.network - skb->data;
8508 		break;
8509 	default:
8510 		return;
8511 	}
8512 
8513 	if (l4_proto != IPPROTO_TCP)
8514 		return;
8515 
8516 	if (unlikely(skb_tail_pointer(skb) < hdr.network +
8517 		     hlen + sizeof(struct tcphdr)))
8518 		return;
8519 
8520 	th = (struct tcphdr *)(hdr.network + hlen);
8521 
8522 	/* skip this packet since the socket is closing */
8523 	if (th->fin)
8524 		return;
8525 
8526 	/* sample on all syn packets or once every atr sample count */
8527 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
8528 		return;
8529 
8530 	/* reset sample count */
8531 	ring->atr_count = 0;
8532 
8533 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
8534 
8535 	/*
8536 	 * src and dst are inverted, think how the receiver sees them
8537 	 *
8538 	 * The input is broken into two sections, a non-compressed section
8539 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
8540 	 * is XORed together and stored in the compressed dword.
8541 	 */
8542 	input.formatted.vlan_id = vlan_id;
8543 
8544 	/*
8545 	 * since src port and flex bytes occupy the same word XOR them together
8546 	 * and write the value to source port portion of compressed dword
8547 	 */
8548 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
8549 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
8550 	else
8551 		common.port.src ^= th->dest ^ first->protocol;
8552 	common.port.dst ^= th->source;
8553 
8554 	switch (hdr.ipv4->version) {
8555 	case IPVERSION:
8556 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
8557 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
8558 		break;
8559 	case 6:
8560 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
8561 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
8562 			     hdr.ipv6->saddr.s6_addr32[1] ^
8563 			     hdr.ipv6->saddr.s6_addr32[2] ^
8564 			     hdr.ipv6->saddr.s6_addr32[3] ^
8565 			     hdr.ipv6->daddr.s6_addr32[0] ^
8566 			     hdr.ipv6->daddr.s6_addr32[1] ^
8567 			     hdr.ipv6->daddr.s6_addr32[2] ^
8568 			     hdr.ipv6->daddr.s6_addr32[3];
8569 		break;
8570 	default:
8571 		break;
8572 	}
8573 
8574 	if (hdr.network != skb_network_header(skb))
8575 		input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
8576 
8577 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
8578 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
8579 					      input, common, ring->queue_index);
8580 }
8581 
8582 #ifdef IXGBE_FCOE
8583 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
8584 			      struct net_device *sb_dev)
8585 {
8586 	struct ixgbe_adapter *adapter;
8587 	struct ixgbe_ring_feature *f;
8588 	int txq;
8589 
8590 	if (sb_dev) {
8591 		u8 tc = netdev_get_prio_tc_map(dev, skb->priority);
8592 		struct net_device *vdev = sb_dev;
8593 
8594 		txq = vdev->tc_to_txq[tc].offset;
8595 		txq += reciprocal_scale(skb_get_hash(skb),
8596 					vdev->tc_to_txq[tc].count);
8597 
8598 		return txq;
8599 	}
8600 
8601 	/*
8602 	 * only execute the code below if protocol is FCoE
8603 	 * or FIP and we have FCoE enabled on the adapter
8604 	 */
8605 	switch (vlan_get_protocol(skb)) {
8606 	case htons(ETH_P_FCOE):
8607 	case htons(ETH_P_FIP):
8608 		adapter = netdev_priv(dev);
8609 
8610 		if (!sb_dev && (adapter->flags & IXGBE_FLAG_FCOE_ENABLED))
8611 			break;
8612 		fallthrough;
8613 	default:
8614 		return netdev_pick_tx(dev, skb, sb_dev);
8615 	}
8616 
8617 	f = &adapter->ring_feature[RING_F_FCOE];
8618 
8619 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
8620 					   smp_processor_id();
8621 
8622 	while (txq >= f->indices)
8623 		txq -= f->indices;
8624 
8625 	return txq + f->offset;
8626 }
8627 
8628 #endif
8629 int ixgbe_xmit_xdp_ring(struct ixgbe_ring *ring,
8630 			struct xdp_frame *xdpf)
8631 {
8632 	struct skb_shared_info *sinfo = xdp_get_shared_info_from_frame(xdpf);
8633 	u8 nr_frags = unlikely(xdp_frame_has_frags(xdpf)) ? sinfo->nr_frags : 0;
8634 	u16 i = 0, index = ring->next_to_use;
8635 	struct ixgbe_tx_buffer *tx_head = &ring->tx_buffer_info[index];
8636 	struct ixgbe_tx_buffer *tx_buff = tx_head;
8637 	union ixgbe_adv_tx_desc *tx_desc = IXGBE_TX_DESC(ring, index);
8638 	u32 cmd_type, len = xdpf->len;
8639 	void *data = xdpf->data;
8640 
8641 	if (unlikely(ixgbe_desc_unused(ring) < 1 + nr_frags))
8642 		return IXGBE_XDP_CONSUMED;
8643 
8644 	tx_head->bytecount = xdp_get_frame_len(xdpf);
8645 	tx_head->gso_segs = 1;
8646 	tx_head->xdpf = xdpf;
8647 
8648 	tx_desc->read.olinfo_status =
8649 		cpu_to_le32(tx_head->bytecount << IXGBE_ADVTXD_PAYLEN_SHIFT);
8650 
8651 	for (;;) {
8652 		dma_addr_t dma;
8653 
8654 		dma = dma_map_single(ring->dev, data, len, DMA_TO_DEVICE);
8655 		if (dma_mapping_error(ring->dev, dma))
8656 			goto unmap;
8657 
8658 		dma_unmap_len_set(tx_buff, len, len);
8659 		dma_unmap_addr_set(tx_buff, dma, dma);
8660 
8661 		cmd_type = IXGBE_ADVTXD_DTYP_DATA | IXGBE_ADVTXD_DCMD_DEXT |
8662 			   IXGBE_ADVTXD_DCMD_IFCS | len;
8663 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8664 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
8665 		tx_buff->protocol = 0;
8666 
8667 		if (++index == ring->count)
8668 			index = 0;
8669 
8670 		if (i == nr_frags)
8671 			break;
8672 
8673 		tx_buff = &ring->tx_buffer_info[index];
8674 		tx_desc = IXGBE_TX_DESC(ring, index);
8675 		tx_desc->read.olinfo_status = 0;
8676 
8677 		data = skb_frag_address(&sinfo->frags[i]);
8678 		len = skb_frag_size(&sinfo->frags[i]);
8679 		i++;
8680 	}
8681 	/* put descriptor type bits */
8682 	tx_desc->read.cmd_type_len |= cpu_to_le32(IXGBE_TXD_CMD);
8683 
8684 	/* Avoid any potential race with xdp_xmit and cleanup */
8685 	smp_wmb();
8686 
8687 	tx_head->next_to_watch = tx_desc;
8688 	ring->next_to_use = index;
8689 
8690 	return IXGBE_XDP_TX;
8691 
8692 unmap:
8693 	for (;;) {
8694 		tx_buff = &ring->tx_buffer_info[index];
8695 		if (dma_unmap_len(tx_buff, len))
8696 			dma_unmap_page(ring->dev, dma_unmap_addr(tx_buff, dma),
8697 				       dma_unmap_len(tx_buff, len),
8698 				       DMA_TO_DEVICE);
8699 		dma_unmap_len_set(tx_buff, len, 0);
8700 		if (tx_buff == tx_head)
8701 			break;
8702 
8703 		if (!index)
8704 			index += ring->count;
8705 		index--;
8706 	}
8707 
8708 	return IXGBE_XDP_CONSUMED;
8709 }
8710 
8711 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
8712 			  struct ixgbe_adapter *adapter,
8713 			  struct ixgbe_ring *tx_ring)
8714 {
8715 	struct ixgbe_tx_buffer *first;
8716 	int tso;
8717 	u32 tx_flags = 0;
8718 	unsigned short f;
8719 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
8720 	struct ixgbe_ipsec_tx_data ipsec_tx = { 0 };
8721 	__be16 protocol = skb->protocol;
8722 	u8 hdr_len = 0;
8723 
8724 	/*
8725 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
8726 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
8727 	 *       + 2 desc gap to keep tail from touching head,
8728 	 *       + 1 desc for context descriptor,
8729 	 * otherwise try next time
8730 	 */
8731 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
8732 		count += TXD_USE_COUNT(skb_frag_size(
8733 						&skb_shinfo(skb)->frags[f]));
8734 
8735 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
8736 		tx_ring->tx_stats.tx_busy++;
8737 		return NETDEV_TX_BUSY;
8738 	}
8739 
8740 	/* record the location of the first descriptor for this packet */
8741 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
8742 	first->skb = skb;
8743 	first->bytecount = skb->len;
8744 	first->gso_segs = 1;
8745 
8746 	/* if we have a HW VLAN tag being added default to the HW one */
8747 	if (skb_vlan_tag_present(skb)) {
8748 		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
8749 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8750 	/* else if it is a SW VLAN check the next protocol and store the tag */
8751 	} else if (protocol == htons(ETH_P_8021Q)) {
8752 		struct vlan_hdr *vhdr, _vhdr;
8753 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
8754 		if (!vhdr)
8755 			goto out_drop;
8756 
8757 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
8758 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
8759 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
8760 	}
8761 	protocol = vlan_get_protocol(skb);
8762 
8763 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
8764 	    adapter->ptp_clock) {
8765 		if (adapter->tstamp_config.tx_type == HWTSTAMP_TX_ON &&
8766 		    !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
8767 					   &adapter->state)) {
8768 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
8769 			tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
8770 
8771 			/* schedule check for Tx timestamp */
8772 			adapter->ptp_tx_skb = skb_get(skb);
8773 			adapter->ptp_tx_start = jiffies;
8774 			schedule_work(&adapter->ptp_tx_work);
8775 		} else {
8776 			adapter->tx_hwtstamp_skipped++;
8777 		}
8778 	}
8779 
8780 #ifdef CONFIG_PCI_IOV
8781 	/*
8782 	 * Use the l2switch_enable flag - would be false if the DMA
8783 	 * Tx switch had been disabled.
8784 	 */
8785 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
8786 		tx_flags |= IXGBE_TX_FLAGS_CC;
8787 
8788 #endif
8789 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
8790 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
8791 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
8792 	     (skb->priority != TC_PRIO_CONTROL))) {
8793 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
8794 		tx_flags |= (skb->priority & 0x7) <<
8795 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
8796 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
8797 			struct vlan_ethhdr *vhdr;
8798 
8799 			if (skb_cow_head(skb, 0))
8800 				goto out_drop;
8801 			vhdr = skb_vlan_eth_hdr(skb);
8802 			vhdr->h_vlan_TCI = htons(tx_flags >>
8803 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
8804 		} else {
8805 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8806 		}
8807 	}
8808 
8809 	/* record initial flags and protocol */
8810 	first->tx_flags = tx_flags;
8811 	first->protocol = protocol;
8812 
8813 #ifdef IXGBE_FCOE
8814 	/* setup tx offload for FCoE */
8815 	if ((protocol == htons(ETH_P_FCOE)) &&
8816 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
8817 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
8818 		if (tso < 0)
8819 			goto out_drop;
8820 
8821 		goto xmit_fcoe;
8822 	}
8823 
8824 #endif /* IXGBE_FCOE */
8825 
8826 #ifdef CONFIG_IXGBE_IPSEC
8827 	if (xfrm_offload(skb) &&
8828 	    !ixgbe_ipsec_tx(tx_ring, first, &ipsec_tx))
8829 		goto out_drop;
8830 #endif
8831 	tso = ixgbe_tso(tx_ring, first, &hdr_len, &ipsec_tx);
8832 	if (tso < 0)
8833 		goto out_drop;
8834 	else if (!tso)
8835 		ixgbe_tx_csum(tx_ring, first, &ipsec_tx);
8836 
8837 	/* add the ATR filter if ATR is on */
8838 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
8839 		ixgbe_atr(tx_ring, first);
8840 
8841 #ifdef IXGBE_FCOE
8842 xmit_fcoe:
8843 #endif /* IXGBE_FCOE */
8844 	if (ixgbe_tx_map(tx_ring, first, hdr_len))
8845 		goto cleanup_tx_timestamp;
8846 
8847 	return NETDEV_TX_OK;
8848 
8849 out_drop:
8850 	dev_kfree_skb_any(first->skb);
8851 	first->skb = NULL;
8852 cleanup_tx_timestamp:
8853 	if (unlikely(tx_flags & IXGBE_TX_FLAGS_TSTAMP)) {
8854 		dev_kfree_skb_any(adapter->ptp_tx_skb);
8855 		adapter->ptp_tx_skb = NULL;
8856 		cancel_work_sync(&adapter->ptp_tx_work);
8857 		clear_bit_unlock(__IXGBE_PTP_TX_IN_PROGRESS, &adapter->state);
8858 	}
8859 
8860 	return NETDEV_TX_OK;
8861 }
8862 
8863 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
8864 				      struct net_device *netdev,
8865 				      struct ixgbe_ring *ring)
8866 {
8867 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8868 	struct ixgbe_ring *tx_ring;
8869 
8870 	/*
8871 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
8872 	 * in order to meet this minimum size requirement.
8873 	 */
8874 	if (skb_put_padto(skb, 17))
8875 		return NETDEV_TX_OK;
8876 
8877 	tx_ring = ring ? ring : adapter->tx_ring[skb_get_queue_mapping(skb)];
8878 	if (unlikely(test_bit(__IXGBE_TX_DISABLED, &tx_ring->state)))
8879 		return NETDEV_TX_BUSY;
8880 
8881 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
8882 }
8883 
8884 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
8885 				    struct net_device *netdev)
8886 {
8887 	return __ixgbe_xmit_frame(skb, netdev, NULL);
8888 }
8889 
8890 /**
8891  * ixgbe_set_mac - Change the Ethernet Address of the NIC
8892  * @netdev: network interface device structure
8893  * @p: pointer to an address structure
8894  *
8895  * Returns 0 on success, negative on failure
8896  **/
8897 static int ixgbe_set_mac(struct net_device *netdev, void *p)
8898 {
8899 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8900 	struct ixgbe_hw *hw = &adapter->hw;
8901 	struct sockaddr *addr = p;
8902 
8903 	if (!is_valid_ether_addr(addr->sa_data))
8904 		return -EADDRNOTAVAIL;
8905 
8906 	eth_hw_addr_set(netdev, addr->sa_data);
8907 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
8908 
8909 	ixgbe_mac_set_default_filter(adapter);
8910 
8911 	return 0;
8912 }
8913 
8914 static int
8915 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
8916 {
8917 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8918 	struct ixgbe_hw *hw = &adapter->hw;
8919 	u16 value;
8920 	int rc;
8921 
8922 	if (adapter->mii_bus) {
8923 		int regnum = addr;
8924 
8925 		if (devad != MDIO_DEVAD_NONE)
8926 			return mdiobus_c45_read(adapter->mii_bus, prtad,
8927 						devad, regnum);
8928 
8929 		return mdiobus_read(adapter->mii_bus, prtad, regnum);
8930 	}
8931 
8932 	if (prtad != hw->phy.mdio.prtad)
8933 		return -EINVAL;
8934 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
8935 	if (!rc)
8936 		rc = value;
8937 	return rc;
8938 }
8939 
8940 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
8941 			    u16 addr, u16 value)
8942 {
8943 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8944 	struct ixgbe_hw *hw = &adapter->hw;
8945 
8946 	if (adapter->mii_bus) {
8947 		int regnum = addr;
8948 
8949 		if (devad != MDIO_DEVAD_NONE)
8950 			return mdiobus_c45_write(adapter->mii_bus, prtad, devad,
8951 						 regnum, value);
8952 
8953 		return mdiobus_write(adapter->mii_bus, prtad, regnum, value);
8954 	}
8955 
8956 	if (prtad != hw->phy.mdio.prtad)
8957 		return -EINVAL;
8958 	return hw->phy.ops.write_reg(hw, addr, devad, value);
8959 }
8960 
8961 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
8962 {
8963 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8964 
8965 	switch (cmd) {
8966 	case SIOCSHWTSTAMP:
8967 		return ixgbe_ptp_set_ts_config(adapter, req);
8968 	case SIOCGHWTSTAMP:
8969 		return ixgbe_ptp_get_ts_config(adapter, req);
8970 	case SIOCGMIIPHY:
8971 		if (!adapter->hw.phy.ops.read_reg)
8972 			return -EOPNOTSUPP;
8973 		fallthrough;
8974 	default:
8975 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
8976 	}
8977 }
8978 
8979 /**
8980  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
8981  * netdev->dev_addrs
8982  * @dev: network interface device structure
8983  *
8984  * Returns non-zero on failure
8985  **/
8986 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
8987 {
8988 	int err = 0;
8989 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8990 	struct ixgbe_hw *hw = &adapter->hw;
8991 
8992 	if (is_valid_ether_addr(hw->mac.san_addr)) {
8993 		rtnl_lock();
8994 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
8995 		rtnl_unlock();
8996 
8997 		/* update SAN MAC vmdq pool selection */
8998 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
8999 	}
9000 	return err;
9001 }
9002 
9003 /**
9004  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
9005  * netdev->dev_addrs
9006  * @dev: network interface device structure
9007  *
9008  * Returns non-zero on failure
9009  **/
9010 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
9011 {
9012 	int err = 0;
9013 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9014 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
9015 
9016 	if (is_valid_ether_addr(mac->san_addr)) {
9017 		rtnl_lock();
9018 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
9019 		rtnl_unlock();
9020 	}
9021 	return err;
9022 }
9023 
9024 static void ixgbe_get_ring_stats64(struct rtnl_link_stats64 *stats,
9025 				   struct ixgbe_ring *ring)
9026 {
9027 	u64 bytes, packets;
9028 	unsigned int start;
9029 
9030 	if (ring) {
9031 		do {
9032 			start = u64_stats_fetch_begin(&ring->syncp);
9033 			packets = ring->stats.packets;
9034 			bytes   = ring->stats.bytes;
9035 		} while (u64_stats_fetch_retry(&ring->syncp, start));
9036 		stats->tx_packets += packets;
9037 		stats->tx_bytes   += bytes;
9038 	}
9039 }
9040 
9041 static void ixgbe_get_stats64(struct net_device *netdev,
9042 			      struct rtnl_link_stats64 *stats)
9043 {
9044 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9045 	int i;
9046 
9047 	rcu_read_lock();
9048 	for (i = 0; i < adapter->num_rx_queues; i++) {
9049 		struct ixgbe_ring *ring = READ_ONCE(adapter->rx_ring[i]);
9050 		u64 bytes, packets;
9051 		unsigned int start;
9052 
9053 		if (ring) {
9054 			do {
9055 				start = u64_stats_fetch_begin(&ring->syncp);
9056 				packets = ring->stats.packets;
9057 				bytes   = ring->stats.bytes;
9058 			} while (u64_stats_fetch_retry(&ring->syncp, start));
9059 			stats->rx_packets += packets;
9060 			stats->rx_bytes   += bytes;
9061 		}
9062 	}
9063 
9064 	for (i = 0; i < adapter->num_tx_queues; i++) {
9065 		struct ixgbe_ring *ring = READ_ONCE(adapter->tx_ring[i]);
9066 
9067 		ixgbe_get_ring_stats64(stats, ring);
9068 	}
9069 	for (i = 0; i < adapter->num_xdp_queues; i++) {
9070 		struct ixgbe_ring *ring = READ_ONCE(adapter->xdp_ring[i]);
9071 
9072 		ixgbe_get_ring_stats64(stats, ring);
9073 	}
9074 	rcu_read_unlock();
9075 
9076 	/* following stats updated by ixgbe_watchdog_task() */
9077 	stats->multicast	= netdev->stats.multicast;
9078 	stats->rx_errors	= netdev->stats.rx_errors;
9079 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
9080 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
9081 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
9082 }
9083 
9084 static int ixgbe_ndo_get_vf_stats(struct net_device *netdev, int vf,
9085 				  struct ifla_vf_stats *vf_stats)
9086 {
9087 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9088 
9089 	if (vf < 0 || vf >= adapter->num_vfs)
9090 		return -EINVAL;
9091 
9092 	vf_stats->rx_packets = adapter->vfinfo[vf].vfstats.gprc;
9093 	vf_stats->rx_bytes   = adapter->vfinfo[vf].vfstats.gorc;
9094 	vf_stats->tx_packets = adapter->vfinfo[vf].vfstats.gptc;
9095 	vf_stats->tx_bytes   = adapter->vfinfo[vf].vfstats.gotc;
9096 	vf_stats->multicast  = adapter->vfinfo[vf].vfstats.mprc;
9097 
9098 	return 0;
9099 }
9100 
9101 #ifdef CONFIG_IXGBE_DCB
9102 /**
9103  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
9104  * @adapter: pointer to ixgbe_adapter
9105  * @tc: number of traffic classes currently enabled
9106  *
9107  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
9108  * 802.1Q priority maps to a packet buffer that exists.
9109  */
9110 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
9111 {
9112 	struct ixgbe_hw *hw = &adapter->hw;
9113 	u32 reg, rsave;
9114 	int i;
9115 
9116 	/* 82598 have a static priority to TC mapping that can not
9117 	 * be changed so no validation is needed.
9118 	 */
9119 	if (hw->mac.type == ixgbe_mac_82598EB)
9120 		return;
9121 
9122 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
9123 	rsave = reg;
9124 
9125 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
9126 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
9127 
9128 		/* If up2tc is out of bounds default to zero */
9129 		if (up2tc > tc)
9130 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
9131 	}
9132 
9133 	if (reg != rsave)
9134 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
9135 
9136 	return;
9137 }
9138 
9139 /**
9140  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
9141  * @adapter: Pointer to adapter struct
9142  *
9143  * Populate the netdev user priority to tc map
9144  */
9145 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
9146 {
9147 	struct net_device *dev = adapter->netdev;
9148 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
9149 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
9150 	u8 prio;
9151 
9152 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
9153 		u8 tc = 0;
9154 
9155 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
9156 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
9157 		else if (ets)
9158 			tc = ets->prio_tc[prio];
9159 
9160 		netdev_set_prio_tc_map(dev, prio, tc);
9161 	}
9162 }
9163 
9164 #endif /* CONFIG_IXGBE_DCB */
9165 static int ixgbe_reassign_macvlan_pool(struct net_device *vdev,
9166 				       struct netdev_nested_priv *priv)
9167 {
9168 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)priv->data;
9169 	struct ixgbe_fwd_adapter *accel;
9170 	int pool;
9171 
9172 	/* we only care about macvlans... */
9173 	if (!netif_is_macvlan(vdev))
9174 		return 0;
9175 
9176 	/* that have hardware offload enabled... */
9177 	accel = macvlan_accel_priv(vdev);
9178 	if (!accel)
9179 		return 0;
9180 
9181 	/* If we can relocate to a different bit do so */
9182 	pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
9183 	if (pool < adapter->num_rx_pools) {
9184 		set_bit(pool, adapter->fwd_bitmask);
9185 		accel->pool = pool;
9186 		return 0;
9187 	}
9188 
9189 	/* if we cannot find a free pool then disable the offload */
9190 	netdev_err(vdev, "L2FW offload disabled due to lack of queue resources\n");
9191 	macvlan_release_l2fw_offload(vdev);
9192 
9193 	/* unbind the queues and drop the subordinate channel config */
9194 	netdev_unbind_sb_channel(adapter->netdev, vdev);
9195 	netdev_set_sb_channel(vdev, 0);
9196 
9197 	kfree(accel);
9198 
9199 	return 0;
9200 }
9201 
9202 static void ixgbe_defrag_macvlan_pools(struct net_device *dev)
9203 {
9204 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9205 	struct netdev_nested_priv priv = {
9206 		.data = (void *)adapter,
9207 	};
9208 
9209 	/* flush any stale bits out of the fwd bitmask */
9210 	bitmap_clear(adapter->fwd_bitmask, 1, 63);
9211 
9212 	/* walk through upper devices reassigning pools */
9213 	netdev_walk_all_upper_dev_rcu(dev, ixgbe_reassign_macvlan_pool,
9214 				      &priv);
9215 }
9216 
9217 /**
9218  * ixgbe_setup_tc - configure net_device for multiple traffic classes
9219  *
9220  * @dev: net device to configure
9221  * @tc: number of traffic classes to enable
9222  */
9223 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
9224 {
9225 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9226 	struct ixgbe_hw *hw = &adapter->hw;
9227 
9228 	/* Hardware supports up to 8 traffic classes */
9229 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
9230 		return -EINVAL;
9231 
9232 	if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
9233 		return -EINVAL;
9234 
9235 	/* Hardware has to reinitialize queues and interrupts to
9236 	 * match packet buffer alignment. Unfortunately, the
9237 	 * hardware is not flexible enough to do this dynamically.
9238 	 */
9239 	if (netif_running(dev))
9240 		ixgbe_close(dev);
9241 	else
9242 		ixgbe_reset(adapter);
9243 
9244 	ixgbe_clear_interrupt_scheme(adapter);
9245 
9246 #ifdef CONFIG_IXGBE_DCB
9247 	if (tc) {
9248 		if (adapter->xdp_prog) {
9249 			e_warn(probe, "DCB is not supported with XDP\n");
9250 
9251 			ixgbe_init_interrupt_scheme(adapter);
9252 			if (netif_running(dev))
9253 				ixgbe_open(dev);
9254 			return -EINVAL;
9255 		}
9256 
9257 		netdev_set_num_tc(dev, tc);
9258 		ixgbe_set_prio_tc_map(adapter);
9259 
9260 		adapter->hw_tcs = tc;
9261 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
9262 
9263 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
9264 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
9265 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
9266 		}
9267 	} else {
9268 		netdev_reset_tc(dev);
9269 
9270 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
9271 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
9272 
9273 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
9274 		adapter->hw_tcs = tc;
9275 
9276 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
9277 		adapter->dcb_cfg.pfc_mode_enable = false;
9278 	}
9279 
9280 	ixgbe_validate_rtr(adapter, tc);
9281 
9282 #endif /* CONFIG_IXGBE_DCB */
9283 	ixgbe_init_interrupt_scheme(adapter);
9284 
9285 	ixgbe_defrag_macvlan_pools(dev);
9286 
9287 	if (netif_running(dev))
9288 		return ixgbe_open(dev);
9289 
9290 	return 0;
9291 }
9292 
9293 static int ixgbe_delete_clsu32(struct ixgbe_adapter *adapter,
9294 			       struct tc_cls_u32_offload *cls)
9295 {
9296 	u32 hdl = cls->knode.handle;
9297 	u32 uhtid = TC_U32_USERHTID(cls->knode.handle);
9298 	u32 loc = cls->knode.handle & 0xfffff;
9299 	int err = 0, i, j;
9300 	struct ixgbe_jump_table *jump = NULL;
9301 
9302 	if (loc > IXGBE_MAX_HW_ENTRIES)
9303 		return -EINVAL;
9304 
9305 	if ((uhtid != 0x800) && (uhtid >= IXGBE_MAX_LINK_HANDLE))
9306 		return -EINVAL;
9307 
9308 	/* Clear this filter in the link data it is associated with */
9309 	if (uhtid != 0x800) {
9310 		jump = adapter->jump_tables[uhtid];
9311 		if (!jump)
9312 			return -EINVAL;
9313 		if (!test_bit(loc - 1, jump->child_loc_map))
9314 			return -EINVAL;
9315 		clear_bit(loc - 1, jump->child_loc_map);
9316 	}
9317 
9318 	/* Check if the filter being deleted is a link */
9319 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
9320 		jump = adapter->jump_tables[i];
9321 		if (jump && jump->link_hdl == hdl) {
9322 			/* Delete filters in the hardware in the child hash
9323 			 * table associated with this link
9324 			 */
9325 			for (j = 0; j < IXGBE_MAX_HW_ENTRIES; j++) {
9326 				if (!test_bit(j, jump->child_loc_map))
9327 					continue;
9328 				spin_lock(&adapter->fdir_perfect_lock);
9329 				err = ixgbe_update_ethtool_fdir_entry(adapter,
9330 								      NULL,
9331 								      j + 1);
9332 				spin_unlock(&adapter->fdir_perfect_lock);
9333 				clear_bit(j, jump->child_loc_map);
9334 			}
9335 			/* Remove resources for this link */
9336 			kfree(jump->input);
9337 			kfree(jump->mask);
9338 			kfree(jump);
9339 			adapter->jump_tables[i] = NULL;
9340 			return err;
9341 		}
9342 	}
9343 
9344 	spin_lock(&adapter->fdir_perfect_lock);
9345 	err = ixgbe_update_ethtool_fdir_entry(adapter, NULL, loc);
9346 	spin_unlock(&adapter->fdir_perfect_lock);
9347 	return err;
9348 }
9349 
9350 static int ixgbe_configure_clsu32_add_hnode(struct ixgbe_adapter *adapter,
9351 					    struct tc_cls_u32_offload *cls)
9352 {
9353 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
9354 
9355 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9356 		return -EINVAL;
9357 
9358 	/* This ixgbe devices do not support hash tables at the moment
9359 	 * so abort when given hash tables.
9360 	 */
9361 	if (cls->hnode.divisor > 0)
9362 		return -EINVAL;
9363 
9364 	set_bit(uhtid - 1, &adapter->tables);
9365 	return 0;
9366 }
9367 
9368 static int ixgbe_configure_clsu32_del_hnode(struct ixgbe_adapter *adapter,
9369 					    struct tc_cls_u32_offload *cls)
9370 {
9371 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
9372 
9373 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9374 		return -EINVAL;
9375 
9376 	clear_bit(uhtid - 1, &adapter->tables);
9377 	return 0;
9378 }
9379 
9380 #ifdef CONFIG_NET_CLS_ACT
9381 struct upper_walk_data {
9382 	struct ixgbe_adapter *adapter;
9383 	u64 action;
9384 	int ifindex;
9385 	u8 queue;
9386 };
9387 
9388 static int get_macvlan_queue(struct net_device *upper,
9389 			     struct netdev_nested_priv *priv)
9390 {
9391 	if (netif_is_macvlan(upper)) {
9392 		struct ixgbe_fwd_adapter *vadapter = macvlan_accel_priv(upper);
9393 		struct ixgbe_adapter *adapter;
9394 		struct upper_walk_data *data;
9395 		int ifindex;
9396 
9397 		data = (struct upper_walk_data *)priv->data;
9398 		ifindex = data->ifindex;
9399 		adapter = data->adapter;
9400 		if (vadapter && upper->ifindex == ifindex) {
9401 			data->queue = adapter->rx_ring[vadapter->rx_base_queue]->reg_idx;
9402 			data->action = data->queue;
9403 			return 1;
9404 		}
9405 	}
9406 
9407 	return 0;
9408 }
9409 
9410 static int handle_redirect_action(struct ixgbe_adapter *adapter, int ifindex,
9411 				  u8 *queue, u64 *action)
9412 {
9413 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
9414 	unsigned int num_vfs = adapter->num_vfs, vf;
9415 	struct netdev_nested_priv priv;
9416 	struct upper_walk_data data;
9417 	struct net_device *upper;
9418 
9419 	/* redirect to a SRIOV VF */
9420 	for (vf = 0; vf < num_vfs; ++vf) {
9421 		upper = pci_get_drvdata(adapter->vfinfo[vf].vfdev);
9422 		if (upper->ifindex == ifindex) {
9423 			*queue = vf * __ALIGN_MASK(1, ~vmdq->mask);
9424 			*action = vf + 1;
9425 			*action <<= ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF;
9426 			return 0;
9427 		}
9428 	}
9429 
9430 	/* redirect to a offloaded macvlan netdev */
9431 	data.adapter = adapter;
9432 	data.ifindex = ifindex;
9433 	data.action = 0;
9434 	data.queue = 0;
9435 	priv.data = (void *)&data;
9436 	if (netdev_walk_all_upper_dev_rcu(adapter->netdev,
9437 					  get_macvlan_queue, &priv)) {
9438 		*action = data.action;
9439 		*queue = data.queue;
9440 
9441 		return 0;
9442 	}
9443 
9444 	return -EINVAL;
9445 }
9446 
9447 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9448 			    struct tcf_exts *exts, u64 *action, u8 *queue)
9449 {
9450 	const struct tc_action *a;
9451 	int i;
9452 
9453 	if (!tcf_exts_has_actions(exts))
9454 		return -EINVAL;
9455 
9456 	tcf_exts_for_each_action(i, a, exts) {
9457 		/* Drop action */
9458 		if (is_tcf_gact_shot(a)) {
9459 			*action = IXGBE_FDIR_DROP_QUEUE;
9460 			*queue = IXGBE_FDIR_DROP_QUEUE;
9461 			return 0;
9462 		}
9463 
9464 		/* Redirect to a VF or a offloaded macvlan */
9465 		if (is_tcf_mirred_egress_redirect(a)) {
9466 			struct net_device *dev = tcf_mirred_dev(a);
9467 
9468 			if (!dev)
9469 				return -EINVAL;
9470 			return handle_redirect_action(adapter, dev->ifindex,
9471 						      queue, action);
9472 		}
9473 
9474 		return -EINVAL;
9475 	}
9476 
9477 	return -EINVAL;
9478 }
9479 #else
9480 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9481 			    struct tcf_exts *exts, u64 *action, u8 *queue)
9482 {
9483 	return -EINVAL;
9484 }
9485 #endif /* CONFIG_NET_CLS_ACT */
9486 
9487 static int ixgbe_clsu32_build_input(struct ixgbe_fdir_filter *input,
9488 				    union ixgbe_atr_input *mask,
9489 				    struct tc_cls_u32_offload *cls,
9490 				    struct ixgbe_mat_field *field_ptr,
9491 				    struct ixgbe_nexthdr *nexthdr)
9492 {
9493 	int i, j, off;
9494 	__be32 val, m;
9495 	bool found_entry = false, found_jump_field = false;
9496 
9497 	for (i = 0; i < cls->knode.sel->nkeys; i++) {
9498 		off = cls->knode.sel->keys[i].off;
9499 		val = cls->knode.sel->keys[i].val;
9500 		m = cls->knode.sel->keys[i].mask;
9501 
9502 		for (j = 0; field_ptr[j].val; j++) {
9503 			if (field_ptr[j].off == off) {
9504 				field_ptr[j].val(input, mask, (__force u32)val,
9505 						 (__force u32)m);
9506 				input->filter.formatted.flow_type |=
9507 					field_ptr[j].type;
9508 				found_entry = true;
9509 				break;
9510 			}
9511 		}
9512 		if (nexthdr) {
9513 			if (nexthdr->off == cls->knode.sel->keys[i].off &&
9514 			    nexthdr->val ==
9515 			    (__force u32)cls->knode.sel->keys[i].val &&
9516 			    nexthdr->mask ==
9517 			    (__force u32)cls->knode.sel->keys[i].mask)
9518 				found_jump_field = true;
9519 			else
9520 				continue;
9521 		}
9522 	}
9523 
9524 	if (nexthdr && !found_jump_field)
9525 		return -EINVAL;
9526 
9527 	if (!found_entry)
9528 		return 0;
9529 
9530 	mask->formatted.flow_type = IXGBE_ATR_L4TYPE_IPV6_MASK |
9531 				    IXGBE_ATR_L4TYPE_MASK;
9532 
9533 	if (input->filter.formatted.flow_type == IXGBE_ATR_FLOW_TYPE_IPV4)
9534 		mask->formatted.flow_type &= IXGBE_ATR_L4TYPE_IPV6_MASK;
9535 
9536 	return 0;
9537 }
9538 
9539 static int ixgbe_configure_clsu32(struct ixgbe_adapter *adapter,
9540 				  struct tc_cls_u32_offload *cls)
9541 {
9542 	__be16 protocol = cls->common.protocol;
9543 	u32 loc = cls->knode.handle & 0xfffff;
9544 	struct ixgbe_hw *hw = &adapter->hw;
9545 	struct ixgbe_mat_field *field_ptr;
9546 	struct ixgbe_fdir_filter *input = NULL;
9547 	union ixgbe_atr_input *mask = NULL;
9548 	struct ixgbe_jump_table *jump = NULL;
9549 	int i, err = -EINVAL;
9550 	u8 queue;
9551 	u32 uhtid, link_uhtid;
9552 
9553 	uhtid = TC_U32_USERHTID(cls->knode.handle);
9554 	link_uhtid = TC_U32_USERHTID(cls->knode.link_handle);
9555 
9556 	/* At the moment cls_u32 jumps to network layer and skips past
9557 	 * L2 headers. The canonical method to match L2 frames is to use
9558 	 * negative values. However this is error prone at best but really
9559 	 * just broken because there is no way to "know" what sort of hdr
9560 	 * is in front of the network layer. Fix cls_u32 to support L2
9561 	 * headers when needed.
9562 	 */
9563 	if (protocol != htons(ETH_P_IP))
9564 		return err;
9565 
9566 	if (loc >= ((1024 << adapter->fdir_pballoc) - 2)) {
9567 		e_err(drv, "Location out of range\n");
9568 		return err;
9569 	}
9570 
9571 	/* cls u32 is a graph starting at root node 0x800. The driver tracks
9572 	 * links and also the fields used to advance the parser across each
9573 	 * link (e.g. nexthdr/eat parameters from 'tc'). This way we can map
9574 	 * the u32 graph onto the hardware parse graph denoted in ixgbe_model.h
9575 	 * To add support for new nodes update ixgbe_model.h parse structures
9576 	 * this function _should_ be generic try not to hardcode values here.
9577 	 */
9578 	if (uhtid == 0x800) {
9579 		field_ptr = (adapter->jump_tables[0])->mat;
9580 	} else {
9581 		if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9582 			return err;
9583 		if (!adapter->jump_tables[uhtid])
9584 			return err;
9585 		field_ptr = (adapter->jump_tables[uhtid])->mat;
9586 	}
9587 
9588 	if (!field_ptr)
9589 		return err;
9590 
9591 	/* At this point we know the field_ptr is valid and need to either
9592 	 * build cls_u32 link or attach filter. Because adding a link to
9593 	 * a handle that does not exist is invalid and the same for adding
9594 	 * rules to handles that don't exist.
9595 	 */
9596 
9597 	if (link_uhtid) {
9598 		struct ixgbe_nexthdr *nexthdr = ixgbe_ipv4_jumps;
9599 
9600 		if (link_uhtid >= IXGBE_MAX_LINK_HANDLE)
9601 			return err;
9602 
9603 		if (!test_bit(link_uhtid - 1, &adapter->tables))
9604 			return err;
9605 
9606 		/* Multiple filters as links to the same hash table are not
9607 		 * supported. To add a new filter with the same next header
9608 		 * but different match/jump conditions, create a new hash table
9609 		 * and link to it.
9610 		 */
9611 		if (adapter->jump_tables[link_uhtid] &&
9612 		    (adapter->jump_tables[link_uhtid])->link_hdl) {
9613 			e_err(drv, "Link filter exists for link: %x\n",
9614 			      link_uhtid);
9615 			return err;
9616 		}
9617 
9618 		for (i = 0; nexthdr[i].jump; i++) {
9619 			if (nexthdr[i].o != cls->knode.sel->offoff ||
9620 			    nexthdr[i].s != cls->knode.sel->offshift ||
9621 			    nexthdr[i].m !=
9622 			    (__force u32)cls->knode.sel->offmask)
9623 				return err;
9624 
9625 			jump = kzalloc(sizeof(*jump), GFP_KERNEL);
9626 			if (!jump)
9627 				return -ENOMEM;
9628 			input = kzalloc(sizeof(*input), GFP_KERNEL);
9629 			if (!input) {
9630 				err = -ENOMEM;
9631 				goto free_jump;
9632 			}
9633 			mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9634 			if (!mask) {
9635 				err = -ENOMEM;
9636 				goto free_input;
9637 			}
9638 			jump->input = input;
9639 			jump->mask = mask;
9640 			jump->link_hdl = cls->knode.handle;
9641 
9642 			err = ixgbe_clsu32_build_input(input, mask, cls,
9643 						       field_ptr, &nexthdr[i]);
9644 			if (!err) {
9645 				jump->mat = nexthdr[i].jump;
9646 				adapter->jump_tables[link_uhtid] = jump;
9647 				break;
9648 			} else {
9649 				kfree(mask);
9650 				kfree(input);
9651 				kfree(jump);
9652 			}
9653 		}
9654 		return 0;
9655 	}
9656 
9657 	input = kzalloc(sizeof(*input), GFP_KERNEL);
9658 	if (!input)
9659 		return -ENOMEM;
9660 	mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9661 	if (!mask) {
9662 		err = -ENOMEM;
9663 		goto free_input;
9664 	}
9665 
9666 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid])) {
9667 		if ((adapter->jump_tables[uhtid])->input)
9668 			memcpy(input, (adapter->jump_tables[uhtid])->input,
9669 			       sizeof(*input));
9670 		if ((adapter->jump_tables[uhtid])->mask)
9671 			memcpy(mask, (adapter->jump_tables[uhtid])->mask,
9672 			       sizeof(*mask));
9673 
9674 		/* Lookup in all child hash tables if this location is already
9675 		 * filled with a filter
9676 		 */
9677 		for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
9678 			struct ixgbe_jump_table *link = adapter->jump_tables[i];
9679 
9680 			if (link && (test_bit(loc - 1, link->child_loc_map))) {
9681 				e_err(drv, "Filter exists in location: %x\n",
9682 				      loc);
9683 				err = -EINVAL;
9684 				goto err_out;
9685 			}
9686 		}
9687 	}
9688 	err = ixgbe_clsu32_build_input(input, mask, cls, field_ptr, NULL);
9689 	if (err)
9690 		goto err_out;
9691 
9692 	err = parse_tc_actions(adapter, cls->knode.exts, &input->action,
9693 			       &queue);
9694 	if (err < 0)
9695 		goto err_out;
9696 
9697 	input->sw_idx = loc;
9698 
9699 	spin_lock(&adapter->fdir_perfect_lock);
9700 
9701 	if (hlist_empty(&adapter->fdir_filter_list)) {
9702 		memcpy(&adapter->fdir_mask, mask, sizeof(*mask));
9703 		err = ixgbe_fdir_set_input_mask_82599(hw, mask);
9704 		if (err)
9705 			goto err_out_w_lock;
9706 	} else if (memcmp(&adapter->fdir_mask, mask, sizeof(*mask))) {
9707 		err = -EINVAL;
9708 		goto err_out_w_lock;
9709 	}
9710 
9711 	ixgbe_atr_compute_perfect_hash_82599(&input->filter, mask);
9712 	err = ixgbe_fdir_write_perfect_filter_82599(hw, &input->filter,
9713 						    input->sw_idx, queue);
9714 	if (err)
9715 		goto err_out_w_lock;
9716 
9717 	ixgbe_update_ethtool_fdir_entry(adapter, input, input->sw_idx);
9718 	spin_unlock(&adapter->fdir_perfect_lock);
9719 
9720 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid]))
9721 		set_bit(loc - 1, (adapter->jump_tables[uhtid])->child_loc_map);
9722 
9723 	kfree(mask);
9724 	return err;
9725 err_out_w_lock:
9726 	spin_unlock(&adapter->fdir_perfect_lock);
9727 err_out:
9728 	kfree(mask);
9729 free_input:
9730 	kfree(input);
9731 free_jump:
9732 	kfree(jump);
9733 	return err;
9734 }
9735 
9736 static int ixgbe_setup_tc_cls_u32(struct ixgbe_adapter *adapter,
9737 				  struct tc_cls_u32_offload *cls_u32)
9738 {
9739 	switch (cls_u32->command) {
9740 	case TC_CLSU32_NEW_KNODE:
9741 	case TC_CLSU32_REPLACE_KNODE:
9742 		return ixgbe_configure_clsu32(adapter, cls_u32);
9743 	case TC_CLSU32_DELETE_KNODE:
9744 		return ixgbe_delete_clsu32(adapter, cls_u32);
9745 	case TC_CLSU32_NEW_HNODE:
9746 	case TC_CLSU32_REPLACE_HNODE:
9747 		return ixgbe_configure_clsu32_add_hnode(adapter, cls_u32);
9748 	case TC_CLSU32_DELETE_HNODE:
9749 		return ixgbe_configure_clsu32_del_hnode(adapter, cls_u32);
9750 	default:
9751 		return -EOPNOTSUPP;
9752 	}
9753 }
9754 
9755 static int ixgbe_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
9756 				   void *cb_priv)
9757 {
9758 	struct ixgbe_adapter *adapter = cb_priv;
9759 
9760 	if (!tc_cls_can_offload_and_chain0(adapter->netdev, type_data))
9761 		return -EOPNOTSUPP;
9762 
9763 	switch (type) {
9764 	case TC_SETUP_CLSU32:
9765 		return ixgbe_setup_tc_cls_u32(adapter, type_data);
9766 	default:
9767 		return -EOPNOTSUPP;
9768 	}
9769 }
9770 
9771 static int ixgbe_setup_tc_mqprio(struct net_device *dev,
9772 				 struct tc_mqprio_qopt *mqprio)
9773 {
9774 	mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
9775 	return ixgbe_setup_tc(dev, mqprio->num_tc);
9776 }
9777 
9778 static LIST_HEAD(ixgbe_block_cb_list);
9779 
9780 static int __ixgbe_setup_tc(struct net_device *dev, enum tc_setup_type type,
9781 			    void *type_data)
9782 {
9783 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9784 
9785 	switch (type) {
9786 	case TC_SETUP_BLOCK:
9787 		return flow_block_cb_setup_simple(type_data,
9788 						  &ixgbe_block_cb_list,
9789 						  ixgbe_setup_tc_block_cb,
9790 						  adapter, adapter, true);
9791 	case TC_SETUP_QDISC_MQPRIO:
9792 		return ixgbe_setup_tc_mqprio(dev, type_data);
9793 	default:
9794 		return -EOPNOTSUPP;
9795 	}
9796 }
9797 
9798 #ifdef CONFIG_PCI_IOV
9799 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
9800 {
9801 	struct net_device *netdev = adapter->netdev;
9802 
9803 	rtnl_lock();
9804 	ixgbe_setup_tc(netdev, adapter->hw_tcs);
9805 	rtnl_unlock();
9806 }
9807 
9808 #endif
9809 void ixgbe_do_reset(struct net_device *netdev)
9810 {
9811 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9812 
9813 	if (netif_running(netdev))
9814 		ixgbe_reinit_locked(adapter);
9815 	else
9816 		ixgbe_reset(adapter);
9817 }
9818 
9819 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
9820 					    netdev_features_t features)
9821 {
9822 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9823 
9824 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
9825 	if (!(features & NETIF_F_RXCSUM))
9826 		features &= ~NETIF_F_LRO;
9827 
9828 	/* Turn off LRO if not RSC capable */
9829 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
9830 		features &= ~NETIF_F_LRO;
9831 
9832 	if (adapter->xdp_prog && (features & NETIF_F_LRO)) {
9833 		e_dev_err("LRO is not supported with XDP\n");
9834 		features &= ~NETIF_F_LRO;
9835 	}
9836 
9837 	return features;
9838 }
9839 
9840 static void ixgbe_reset_l2fw_offload(struct ixgbe_adapter *adapter)
9841 {
9842 	int rss = min_t(int, ixgbe_max_rss_indices(adapter),
9843 			num_online_cpus());
9844 
9845 	/* go back to full RSS if we're not running SR-IOV */
9846 	if (!adapter->ring_feature[RING_F_VMDQ].offset)
9847 		adapter->flags &= ~(IXGBE_FLAG_VMDQ_ENABLED |
9848 				    IXGBE_FLAG_SRIOV_ENABLED);
9849 
9850 	adapter->ring_feature[RING_F_RSS].limit = rss;
9851 	adapter->ring_feature[RING_F_VMDQ].limit = 1;
9852 
9853 	ixgbe_setup_tc(adapter->netdev, adapter->hw_tcs);
9854 }
9855 
9856 static int ixgbe_set_features(struct net_device *netdev,
9857 			      netdev_features_t features)
9858 {
9859 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9860 	netdev_features_t changed = netdev->features ^ features;
9861 	bool need_reset = false;
9862 
9863 	/* Make sure RSC matches LRO, reset if change */
9864 	if (!(features & NETIF_F_LRO)) {
9865 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
9866 			need_reset = true;
9867 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
9868 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
9869 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
9870 		if (adapter->rx_itr_setting == 1 ||
9871 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
9872 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
9873 			need_reset = true;
9874 		} else if ((changed ^ features) & NETIF_F_LRO) {
9875 			e_info(probe, "rx-usecs set too low, "
9876 			       "disabling RSC\n");
9877 		}
9878 	}
9879 
9880 	/*
9881 	 * Check if Flow Director n-tuple support or hw_tc support was
9882 	 * enabled or disabled.  If the state changed, we need to reset.
9883 	 */
9884 	if ((features & NETIF_F_NTUPLE) || (features & NETIF_F_HW_TC)) {
9885 		/* turn off ATR, enable perfect filters and reset */
9886 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
9887 			need_reset = true;
9888 
9889 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
9890 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9891 	} else {
9892 		/* turn off perfect filters, enable ATR and reset */
9893 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
9894 			need_reset = true;
9895 
9896 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9897 
9898 		/* We cannot enable ATR if SR-IOV is enabled */
9899 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED ||
9900 		    /* We cannot enable ATR if we have 2 or more tcs */
9901 		    (adapter->hw_tcs > 1) ||
9902 		    /* We cannot enable ATR if RSS is disabled */
9903 		    (adapter->ring_feature[RING_F_RSS].limit <= 1) ||
9904 		    /* A sample rate of 0 indicates ATR disabled */
9905 		    (!adapter->atr_sample_rate))
9906 			; /* do nothing not supported */
9907 		else /* otherwise supported and set the flag */
9908 			adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
9909 	}
9910 
9911 	if (changed & NETIF_F_RXALL)
9912 		need_reset = true;
9913 
9914 	netdev->features = features;
9915 
9916 	if ((changed & NETIF_F_HW_L2FW_DOFFLOAD) && adapter->num_rx_pools > 1)
9917 		ixgbe_reset_l2fw_offload(adapter);
9918 	else if (need_reset)
9919 		ixgbe_do_reset(netdev);
9920 	else if (changed & (NETIF_F_HW_VLAN_CTAG_RX |
9921 			    NETIF_F_HW_VLAN_CTAG_FILTER))
9922 		ixgbe_set_rx_mode(netdev);
9923 
9924 	return 1;
9925 }
9926 
9927 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
9928 			     struct net_device *dev,
9929 			     const unsigned char *addr, u16 vid,
9930 			     u16 flags,
9931 			     struct netlink_ext_ack *extack)
9932 {
9933 	/* guarantee we can provide a unique filter for the unicast address */
9934 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
9935 		struct ixgbe_adapter *adapter = netdev_priv(dev);
9936 		u16 pool = VMDQ_P(0);
9937 
9938 		if (netdev_uc_count(dev) >= ixgbe_available_rars(adapter, pool))
9939 			return -ENOMEM;
9940 	}
9941 
9942 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
9943 }
9944 
9945 /**
9946  * ixgbe_configure_bridge_mode - set various bridge modes
9947  * @adapter: the private structure
9948  * @mode: requested bridge mode
9949  *
9950  * Configure some settings require for various bridge modes.
9951  **/
9952 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
9953 				       __u16 mode)
9954 {
9955 	struct ixgbe_hw *hw = &adapter->hw;
9956 	unsigned int p, num_pools;
9957 	u32 vmdctl;
9958 
9959 	switch (mode) {
9960 	case BRIDGE_MODE_VEPA:
9961 		/* disable Tx loopback, rely on switch hairpin mode */
9962 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
9963 
9964 		/* must enable Rx switching replication to allow multicast
9965 		 * packet reception on all VFs, and to enable source address
9966 		 * pruning.
9967 		 */
9968 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9969 		vmdctl |= IXGBE_VT_CTL_REPLEN;
9970 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9971 
9972 		/* enable Rx source address pruning. Note, this requires
9973 		 * replication to be enabled or else it does nothing.
9974 		 */
9975 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9976 		for (p = 0; p < num_pools; p++) {
9977 			if (hw->mac.ops.set_source_address_pruning)
9978 				hw->mac.ops.set_source_address_pruning(hw,
9979 								       true,
9980 								       p);
9981 		}
9982 		break;
9983 	case BRIDGE_MODE_VEB:
9984 		/* enable Tx loopback for internal VF/PF communication */
9985 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
9986 				IXGBE_PFDTXGSWC_VT_LBEN);
9987 
9988 		/* disable Rx switching replication unless we have SR-IOV
9989 		 * virtual functions
9990 		 */
9991 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9992 		if (!adapter->num_vfs)
9993 			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
9994 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9995 
9996 		/* disable Rx source address pruning, since we don't expect to
9997 		 * be receiving external loopback of our transmitted frames.
9998 		 */
9999 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
10000 		for (p = 0; p < num_pools; p++) {
10001 			if (hw->mac.ops.set_source_address_pruning)
10002 				hw->mac.ops.set_source_address_pruning(hw,
10003 								       false,
10004 								       p);
10005 		}
10006 		break;
10007 	default:
10008 		return -EINVAL;
10009 	}
10010 
10011 	adapter->bridge_mode = mode;
10012 
10013 	e_info(drv, "enabling bridge mode: %s\n",
10014 	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
10015 
10016 	return 0;
10017 }
10018 
10019 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
10020 				    struct nlmsghdr *nlh, u16 flags,
10021 				    struct netlink_ext_ack *extack)
10022 {
10023 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10024 	struct nlattr *attr, *br_spec;
10025 	int rem;
10026 
10027 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
10028 		return -EOPNOTSUPP;
10029 
10030 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
10031 	if (!br_spec)
10032 		return -EINVAL;
10033 
10034 	nla_for_each_nested(attr, br_spec, rem) {
10035 		int status;
10036 		__u16 mode;
10037 
10038 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
10039 			continue;
10040 
10041 		mode = nla_get_u16(attr);
10042 		status = ixgbe_configure_bridge_mode(adapter, mode);
10043 		if (status)
10044 			return status;
10045 
10046 		break;
10047 	}
10048 
10049 	return 0;
10050 }
10051 
10052 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
10053 				    struct net_device *dev,
10054 				    u32 filter_mask, int nlflags)
10055 {
10056 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10057 
10058 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
10059 		return 0;
10060 
10061 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
10062 				       adapter->bridge_mode, 0, 0, nlflags,
10063 				       filter_mask, NULL);
10064 }
10065 
10066 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
10067 {
10068 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
10069 	struct ixgbe_fwd_adapter *accel;
10070 	int tcs = adapter->hw_tcs ? : 1;
10071 	int pool, err;
10072 
10073 	if (adapter->xdp_prog) {
10074 		e_warn(probe, "L2FW offload is not supported with XDP\n");
10075 		return ERR_PTR(-EINVAL);
10076 	}
10077 
10078 	/* The hardware supported by ixgbe only filters on the destination MAC
10079 	 * address. In order to avoid issues we only support offloading modes
10080 	 * where the hardware can actually provide the functionality.
10081 	 */
10082 	if (!macvlan_supports_dest_filter(vdev))
10083 		return ERR_PTR(-EMEDIUMTYPE);
10084 
10085 	/* We need to lock down the macvlan to be a single queue device so that
10086 	 * we can reuse the tc_to_txq field in the macvlan netdev to represent
10087 	 * the queue mapping to our netdev.
10088 	 */
10089 	if (netif_is_multiqueue(vdev))
10090 		return ERR_PTR(-ERANGE);
10091 
10092 	pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
10093 	if (pool == adapter->num_rx_pools) {
10094 		u16 used_pools = adapter->num_vfs + adapter->num_rx_pools;
10095 		u16 reserved_pools;
10096 
10097 		if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
10098 		     adapter->num_rx_pools >= (MAX_TX_QUEUES / tcs)) ||
10099 		    adapter->num_rx_pools > IXGBE_MAX_MACVLANS)
10100 			return ERR_PTR(-EBUSY);
10101 
10102 		/* Hardware has a limited number of available pools. Each VF,
10103 		 * and the PF require a pool. Check to ensure we don't
10104 		 * attempt to use more then the available number of pools.
10105 		 */
10106 		if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
10107 			return ERR_PTR(-EBUSY);
10108 
10109 		/* Enable VMDq flag so device will be set in VM mode */
10110 		adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED |
10111 				  IXGBE_FLAG_SRIOV_ENABLED;
10112 
10113 		/* Try to reserve as many queues per pool as possible,
10114 		 * we start with the configurations that support 4 queues
10115 		 * per pools, followed by 2, and then by just 1 per pool.
10116 		 */
10117 		if (used_pools < 32 && adapter->num_rx_pools < 16)
10118 			reserved_pools = min_t(u16,
10119 					       32 - used_pools,
10120 					       16 - adapter->num_rx_pools);
10121 		else if (adapter->num_rx_pools < 32)
10122 			reserved_pools = min_t(u16,
10123 					       64 - used_pools,
10124 					       32 - adapter->num_rx_pools);
10125 		else
10126 			reserved_pools = 64 - used_pools;
10127 
10128 
10129 		if (!reserved_pools)
10130 			return ERR_PTR(-EBUSY);
10131 
10132 		adapter->ring_feature[RING_F_VMDQ].limit += reserved_pools;
10133 
10134 		/* Force reinit of ring allocation with VMDQ enabled */
10135 		err = ixgbe_setup_tc(pdev, adapter->hw_tcs);
10136 		if (err)
10137 			return ERR_PTR(err);
10138 
10139 		if (pool >= adapter->num_rx_pools)
10140 			return ERR_PTR(-ENOMEM);
10141 	}
10142 
10143 	accel = kzalloc(sizeof(*accel), GFP_KERNEL);
10144 	if (!accel)
10145 		return ERR_PTR(-ENOMEM);
10146 
10147 	set_bit(pool, adapter->fwd_bitmask);
10148 	netdev_set_sb_channel(vdev, pool);
10149 	accel->pool = pool;
10150 	accel->netdev = vdev;
10151 
10152 	if (!netif_running(pdev))
10153 		return accel;
10154 
10155 	err = ixgbe_fwd_ring_up(adapter, accel);
10156 	if (err)
10157 		return ERR_PTR(err);
10158 
10159 	return accel;
10160 }
10161 
10162 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
10163 {
10164 	struct ixgbe_fwd_adapter *accel = priv;
10165 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
10166 	unsigned int rxbase = accel->rx_base_queue;
10167 	unsigned int i;
10168 
10169 	/* delete unicast filter associated with offloaded interface */
10170 	ixgbe_del_mac_filter(adapter, accel->netdev->dev_addr,
10171 			     VMDQ_P(accel->pool));
10172 
10173 	/* Allow remaining Rx packets to get flushed out of the
10174 	 * Rx FIFO before we drop the netdev for the ring.
10175 	 */
10176 	usleep_range(10000, 20000);
10177 
10178 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
10179 		struct ixgbe_ring *ring = adapter->rx_ring[rxbase + i];
10180 		struct ixgbe_q_vector *qv = ring->q_vector;
10181 
10182 		/* Make sure we aren't processing any packets and clear
10183 		 * netdev to shut down the ring.
10184 		 */
10185 		if (netif_running(adapter->netdev))
10186 			napi_synchronize(&qv->napi);
10187 		ring->netdev = NULL;
10188 	}
10189 
10190 	/* unbind the queues and drop the subordinate channel config */
10191 	netdev_unbind_sb_channel(pdev, accel->netdev);
10192 	netdev_set_sb_channel(accel->netdev, 0);
10193 
10194 	clear_bit(accel->pool, adapter->fwd_bitmask);
10195 	kfree(accel);
10196 }
10197 
10198 #define IXGBE_MAX_MAC_HDR_LEN		127
10199 #define IXGBE_MAX_NETWORK_HDR_LEN	511
10200 
10201 static netdev_features_t
10202 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
10203 		     netdev_features_t features)
10204 {
10205 	unsigned int network_hdr_len, mac_hdr_len;
10206 
10207 	/* Make certain the headers can be described by a context descriptor */
10208 	mac_hdr_len = skb_network_header(skb) - skb->data;
10209 	if (unlikely(mac_hdr_len > IXGBE_MAX_MAC_HDR_LEN))
10210 		return features & ~(NETIF_F_HW_CSUM |
10211 				    NETIF_F_SCTP_CRC |
10212 				    NETIF_F_GSO_UDP_L4 |
10213 				    NETIF_F_HW_VLAN_CTAG_TX |
10214 				    NETIF_F_TSO |
10215 				    NETIF_F_TSO6);
10216 
10217 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
10218 	if (unlikely(network_hdr_len >  IXGBE_MAX_NETWORK_HDR_LEN))
10219 		return features & ~(NETIF_F_HW_CSUM |
10220 				    NETIF_F_SCTP_CRC |
10221 				    NETIF_F_GSO_UDP_L4 |
10222 				    NETIF_F_TSO |
10223 				    NETIF_F_TSO6);
10224 
10225 	/* We can only support IPV4 TSO in tunnels if we can mangle the
10226 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
10227 	 * IPsec offoad sets skb->encapsulation but still can handle
10228 	 * the TSO, so it's the exception.
10229 	 */
10230 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID)) {
10231 #ifdef CONFIG_IXGBE_IPSEC
10232 		if (!secpath_exists(skb))
10233 #endif
10234 			features &= ~NETIF_F_TSO;
10235 	}
10236 
10237 	return features;
10238 }
10239 
10240 static int ixgbe_xdp_setup(struct net_device *dev, struct bpf_prog *prog)
10241 {
10242 	int i, frame_size = dev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
10243 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10244 	struct bpf_prog *old_prog;
10245 	bool need_reset;
10246 	int num_queues;
10247 
10248 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
10249 		return -EINVAL;
10250 
10251 	if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
10252 		return -EINVAL;
10253 
10254 	/* verify ixgbe ring attributes are sufficient for XDP */
10255 	for (i = 0; i < adapter->num_rx_queues; i++) {
10256 		struct ixgbe_ring *ring = adapter->rx_ring[i];
10257 
10258 		if (ring_is_rsc_enabled(ring))
10259 			return -EINVAL;
10260 
10261 		if (frame_size > ixgbe_rx_bufsz(ring))
10262 			return -EINVAL;
10263 	}
10264 
10265 	/* if the number of cpus is much larger than the maximum of queues,
10266 	 * we should stop it and then return with ENOMEM like before.
10267 	 */
10268 	if (nr_cpu_ids > IXGBE_MAX_XDP_QS * 2)
10269 		return -ENOMEM;
10270 
10271 	old_prog = xchg(&adapter->xdp_prog, prog);
10272 	need_reset = (!!prog != !!old_prog);
10273 
10274 	/* If transitioning XDP modes reconfigure rings */
10275 	if (need_reset) {
10276 		int err;
10277 
10278 		if (!prog)
10279 			/* Wait until ndo_xsk_wakeup completes. */
10280 			synchronize_rcu();
10281 		err = ixgbe_setup_tc(dev, adapter->hw_tcs);
10282 
10283 		if (err)
10284 			return -EINVAL;
10285 		if (!prog)
10286 			xdp_features_clear_redirect_target(dev);
10287 	} else {
10288 		for (i = 0; i < adapter->num_rx_queues; i++) {
10289 			WRITE_ONCE(adapter->rx_ring[i]->xdp_prog,
10290 				   adapter->xdp_prog);
10291 		}
10292 	}
10293 
10294 	if (old_prog)
10295 		bpf_prog_put(old_prog);
10296 
10297 	/* Kick start the NAPI context if there is an AF_XDP socket open
10298 	 * on that queue id. This so that receiving will start.
10299 	 */
10300 	if (need_reset && prog) {
10301 		num_queues = min_t(int, adapter->num_rx_queues,
10302 				   adapter->num_xdp_queues);
10303 		for (i = 0; i < num_queues; i++)
10304 			if (adapter->xdp_ring[i]->xsk_pool)
10305 				(void)ixgbe_xsk_wakeup(adapter->netdev, i,
10306 						       XDP_WAKEUP_RX);
10307 		xdp_features_set_redirect_target(dev, true);
10308 	}
10309 
10310 	return 0;
10311 }
10312 
10313 static int ixgbe_xdp(struct net_device *dev, struct netdev_bpf *xdp)
10314 {
10315 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10316 
10317 	switch (xdp->command) {
10318 	case XDP_SETUP_PROG:
10319 		return ixgbe_xdp_setup(dev, xdp->prog);
10320 	case XDP_SETUP_XSK_POOL:
10321 		return ixgbe_xsk_pool_setup(adapter, xdp->xsk.pool,
10322 					    xdp->xsk.queue_id);
10323 
10324 	default:
10325 		return -EINVAL;
10326 	}
10327 }
10328 
10329 void ixgbe_xdp_ring_update_tail(struct ixgbe_ring *ring)
10330 {
10331 	/* Force memory writes to complete before letting h/w know there
10332 	 * are new descriptors to fetch.
10333 	 */
10334 	wmb();
10335 	writel(ring->next_to_use, ring->tail);
10336 }
10337 
10338 void ixgbe_xdp_ring_update_tail_locked(struct ixgbe_ring *ring)
10339 {
10340 	if (static_branch_unlikely(&ixgbe_xdp_locking_key))
10341 		spin_lock(&ring->tx_lock);
10342 	ixgbe_xdp_ring_update_tail(ring);
10343 	if (static_branch_unlikely(&ixgbe_xdp_locking_key))
10344 		spin_unlock(&ring->tx_lock);
10345 }
10346 
10347 static int ixgbe_xdp_xmit(struct net_device *dev, int n,
10348 			  struct xdp_frame **frames, u32 flags)
10349 {
10350 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10351 	struct ixgbe_ring *ring;
10352 	int nxmit = 0;
10353 	int i;
10354 
10355 	if (unlikely(test_bit(__IXGBE_DOWN, &adapter->state)))
10356 		return -ENETDOWN;
10357 
10358 	if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
10359 		return -EINVAL;
10360 
10361 	/* During program transitions its possible adapter->xdp_prog is assigned
10362 	 * but ring has not been configured yet. In this case simply abort xmit.
10363 	 */
10364 	ring = adapter->xdp_prog ? ixgbe_determine_xdp_ring(adapter) : NULL;
10365 	if (unlikely(!ring))
10366 		return -ENXIO;
10367 
10368 	if (unlikely(test_bit(__IXGBE_TX_DISABLED, &ring->state)))
10369 		return -ENXIO;
10370 
10371 	if (static_branch_unlikely(&ixgbe_xdp_locking_key))
10372 		spin_lock(&ring->tx_lock);
10373 
10374 	for (i = 0; i < n; i++) {
10375 		struct xdp_frame *xdpf = frames[i];
10376 		int err;
10377 
10378 		err = ixgbe_xmit_xdp_ring(ring, xdpf);
10379 		if (err != IXGBE_XDP_TX)
10380 			break;
10381 		nxmit++;
10382 	}
10383 
10384 	if (unlikely(flags & XDP_XMIT_FLUSH))
10385 		ixgbe_xdp_ring_update_tail(ring);
10386 
10387 	if (static_branch_unlikely(&ixgbe_xdp_locking_key))
10388 		spin_unlock(&ring->tx_lock);
10389 
10390 	return nxmit;
10391 }
10392 
10393 static const struct net_device_ops ixgbe_netdev_ops = {
10394 	.ndo_open		= ixgbe_open,
10395 	.ndo_stop		= ixgbe_close,
10396 	.ndo_start_xmit		= ixgbe_xmit_frame,
10397 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
10398 	.ndo_validate_addr	= eth_validate_addr,
10399 	.ndo_set_mac_address	= ixgbe_set_mac,
10400 	.ndo_change_mtu		= ixgbe_change_mtu,
10401 	.ndo_tx_timeout		= ixgbe_tx_timeout,
10402 	.ndo_set_tx_maxrate	= ixgbe_tx_maxrate,
10403 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
10404 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
10405 	.ndo_eth_ioctl		= ixgbe_ioctl,
10406 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
10407 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
10408 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
10409 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
10410 	.ndo_set_vf_link_state	= ixgbe_ndo_set_vf_link_state,
10411 	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
10412 	.ndo_set_vf_trust	= ixgbe_ndo_set_vf_trust,
10413 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
10414 	.ndo_get_vf_stats	= ixgbe_ndo_get_vf_stats,
10415 	.ndo_get_stats64	= ixgbe_get_stats64,
10416 	.ndo_setup_tc		= __ixgbe_setup_tc,
10417 #ifdef IXGBE_FCOE
10418 	.ndo_select_queue	= ixgbe_select_queue,
10419 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
10420 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
10421 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
10422 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
10423 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
10424 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
10425 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
10426 #endif /* IXGBE_FCOE */
10427 	.ndo_set_features = ixgbe_set_features,
10428 	.ndo_fix_features = ixgbe_fix_features,
10429 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
10430 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
10431 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
10432 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
10433 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
10434 	.ndo_features_check	= ixgbe_features_check,
10435 	.ndo_bpf		= ixgbe_xdp,
10436 	.ndo_xdp_xmit		= ixgbe_xdp_xmit,
10437 	.ndo_xsk_wakeup         = ixgbe_xsk_wakeup,
10438 };
10439 
10440 static void ixgbe_disable_txr_hw(struct ixgbe_adapter *adapter,
10441 				 struct ixgbe_ring *tx_ring)
10442 {
10443 	unsigned long wait_delay, delay_interval;
10444 	struct ixgbe_hw *hw = &adapter->hw;
10445 	u8 reg_idx = tx_ring->reg_idx;
10446 	int wait_loop;
10447 	u32 txdctl;
10448 
10449 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
10450 
10451 	/* delay mechanism from ixgbe_disable_tx */
10452 	delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
10453 
10454 	wait_loop = IXGBE_MAX_RX_DESC_POLL;
10455 	wait_delay = delay_interval;
10456 
10457 	while (wait_loop--) {
10458 		usleep_range(wait_delay, wait_delay + 10);
10459 		wait_delay += delay_interval * 2;
10460 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
10461 
10462 		if (!(txdctl & IXGBE_TXDCTL_ENABLE))
10463 			return;
10464 	}
10465 
10466 	e_err(drv, "TXDCTL.ENABLE not cleared within the polling period\n");
10467 }
10468 
10469 static void ixgbe_disable_txr(struct ixgbe_adapter *adapter,
10470 			      struct ixgbe_ring *tx_ring)
10471 {
10472 	set_bit(__IXGBE_TX_DISABLED, &tx_ring->state);
10473 	ixgbe_disable_txr_hw(adapter, tx_ring);
10474 }
10475 
10476 static void ixgbe_disable_rxr_hw(struct ixgbe_adapter *adapter,
10477 				 struct ixgbe_ring *rx_ring)
10478 {
10479 	unsigned long wait_delay, delay_interval;
10480 	struct ixgbe_hw *hw = &adapter->hw;
10481 	u8 reg_idx = rx_ring->reg_idx;
10482 	int wait_loop;
10483 	u32 rxdctl;
10484 
10485 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
10486 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
10487 	rxdctl |= IXGBE_RXDCTL_SWFLSH;
10488 
10489 	/* write value back with RXDCTL.ENABLE bit cleared */
10490 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
10491 
10492 	/* RXDCTL.EN may not change on 82598 if link is down, so skip it */
10493 	if (hw->mac.type == ixgbe_mac_82598EB &&
10494 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
10495 		return;
10496 
10497 	/* delay mechanism from ixgbe_disable_rx */
10498 	delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
10499 
10500 	wait_loop = IXGBE_MAX_RX_DESC_POLL;
10501 	wait_delay = delay_interval;
10502 
10503 	while (wait_loop--) {
10504 		usleep_range(wait_delay, wait_delay + 10);
10505 		wait_delay += delay_interval * 2;
10506 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
10507 
10508 		if (!(rxdctl & IXGBE_RXDCTL_ENABLE))
10509 			return;
10510 	}
10511 
10512 	e_err(drv, "RXDCTL.ENABLE not cleared within the polling period\n");
10513 }
10514 
10515 static void ixgbe_reset_txr_stats(struct ixgbe_ring *tx_ring)
10516 {
10517 	memset(&tx_ring->stats, 0, sizeof(tx_ring->stats));
10518 	memset(&tx_ring->tx_stats, 0, sizeof(tx_ring->tx_stats));
10519 }
10520 
10521 static void ixgbe_reset_rxr_stats(struct ixgbe_ring *rx_ring)
10522 {
10523 	memset(&rx_ring->stats, 0, sizeof(rx_ring->stats));
10524 	memset(&rx_ring->rx_stats, 0, sizeof(rx_ring->rx_stats));
10525 }
10526 
10527 /**
10528  * ixgbe_txrx_ring_disable - Disable Rx/Tx/XDP Tx rings
10529  * @adapter: adapter structure
10530  * @ring: ring index
10531  *
10532  * This function disables a certain Rx/Tx/XDP Tx ring. The function
10533  * assumes that the netdev is running.
10534  **/
10535 void ixgbe_txrx_ring_disable(struct ixgbe_adapter *adapter, int ring)
10536 {
10537 	struct ixgbe_ring *rx_ring, *tx_ring, *xdp_ring;
10538 
10539 	rx_ring = adapter->rx_ring[ring];
10540 	tx_ring = adapter->tx_ring[ring];
10541 	xdp_ring = adapter->xdp_ring[ring];
10542 
10543 	ixgbe_disable_txr(adapter, tx_ring);
10544 	if (xdp_ring)
10545 		ixgbe_disable_txr(adapter, xdp_ring);
10546 	ixgbe_disable_rxr_hw(adapter, rx_ring);
10547 
10548 	if (xdp_ring)
10549 		synchronize_rcu();
10550 
10551 	/* Rx/Tx/XDP Tx share the same napi context. */
10552 	napi_disable(&rx_ring->q_vector->napi);
10553 
10554 	ixgbe_clean_tx_ring(tx_ring);
10555 	if (xdp_ring)
10556 		ixgbe_clean_tx_ring(xdp_ring);
10557 	ixgbe_clean_rx_ring(rx_ring);
10558 
10559 	ixgbe_reset_txr_stats(tx_ring);
10560 	if (xdp_ring)
10561 		ixgbe_reset_txr_stats(xdp_ring);
10562 	ixgbe_reset_rxr_stats(rx_ring);
10563 }
10564 
10565 /**
10566  * ixgbe_txrx_ring_enable - Enable Rx/Tx/XDP Tx rings
10567  * @adapter: adapter structure
10568  * @ring: ring index
10569  *
10570  * This function enables a certain Rx/Tx/XDP Tx ring. The function
10571  * assumes that the netdev is running.
10572  **/
10573 void ixgbe_txrx_ring_enable(struct ixgbe_adapter *adapter, int ring)
10574 {
10575 	struct ixgbe_ring *rx_ring, *tx_ring, *xdp_ring;
10576 
10577 	rx_ring = adapter->rx_ring[ring];
10578 	tx_ring = adapter->tx_ring[ring];
10579 	xdp_ring = adapter->xdp_ring[ring];
10580 
10581 	/* Rx/Tx/XDP Tx share the same napi context. */
10582 	napi_enable(&rx_ring->q_vector->napi);
10583 
10584 	ixgbe_configure_tx_ring(adapter, tx_ring);
10585 	if (xdp_ring)
10586 		ixgbe_configure_tx_ring(adapter, xdp_ring);
10587 	ixgbe_configure_rx_ring(adapter, rx_ring);
10588 
10589 	clear_bit(__IXGBE_TX_DISABLED, &tx_ring->state);
10590 	if (xdp_ring)
10591 		clear_bit(__IXGBE_TX_DISABLED, &xdp_ring->state);
10592 }
10593 
10594 /**
10595  * ixgbe_enumerate_functions - Get the number of ports this device has
10596  * @adapter: adapter structure
10597  *
10598  * This function enumerates the phsyical functions co-located on a single slot,
10599  * in order to determine how many ports a device has. This is most useful in
10600  * determining the required GT/s of PCIe bandwidth necessary for optimal
10601  * performance.
10602  **/
10603 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
10604 {
10605 	struct pci_dev *entry, *pdev = adapter->pdev;
10606 	int physfns = 0;
10607 
10608 	/* Some cards can not use the generic count PCIe functions method,
10609 	 * because they are behind a parent switch, so we hardcode these with
10610 	 * the correct number of functions.
10611 	 */
10612 	if (ixgbe_pcie_from_parent(&adapter->hw))
10613 		physfns = 4;
10614 
10615 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
10616 		/* don't count virtual functions */
10617 		if (entry->is_virtfn)
10618 			continue;
10619 
10620 		/* When the devices on the bus don't all match our device ID,
10621 		 * we can't reliably determine the correct number of
10622 		 * functions. This can occur if a function has been direct
10623 		 * attached to a virtual machine using VT-d, for example. In
10624 		 * this case, simply return -1 to indicate this.
10625 		 */
10626 		if ((entry->vendor != pdev->vendor) ||
10627 		    (entry->device != pdev->device))
10628 			return -1;
10629 
10630 		physfns++;
10631 	}
10632 
10633 	return physfns;
10634 }
10635 
10636 /**
10637  * ixgbe_wol_supported - Check whether device supports WoL
10638  * @adapter: the adapter private structure
10639  * @device_id: the device ID
10640  * @subdevice_id: the subsystem device ID
10641  *
10642  * This function is used by probe and ethtool to determine
10643  * which devices have WoL support
10644  *
10645  **/
10646 bool ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
10647 			 u16 subdevice_id)
10648 {
10649 	struct ixgbe_hw *hw = &adapter->hw;
10650 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
10651 
10652 	/* WOL not supported on 82598 */
10653 	if (hw->mac.type == ixgbe_mac_82598EB)
10654 		return false;
10655 
10656 	/* check eeprom to see if WOL is enabled for X540 and newer */
10657 	if (hw->mac.type >= ixgbe_mac_X540) {
10658 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
10659 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
10660 		     (hw->bus.func == 0)))
10661 			return true;
10662 	}
10663 
10664 	/* WOL is determined based on device IDs for 82599 MACs */
10665 	switch (device_id) {
10666 	case IXGBE_DEV_ID_82599_SFP:
10667 		/* Only these subdevices could supports WOL */
10668 		switch (subdevice_id) {
10669 		case IXGBE_SUBDEV_ID_82599_560FLR:
10670 		case IXGBE_SUBDEV_ID_82599_LOM_SNAP6:
10671 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
10672 		case IXGBE_SUBDEV_ID_82599_SFP_2OCP:
10673 			/* only support first port */
10674 			if (hw->bus.func != 0)
10675 				break;
10676 			fallthrough;
10677 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
10678 		case IXGBE_SUBDEV_ID_82599_SFP:
10679 		case IXGBE_SUBDEV_ID_82599_RNDC:
10680 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
10681 		case IXGBE_SUBDEV_ID_82599_SFP_1OCP:
10682 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM1:
10683 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM2:
10684 			return true;
10685 		}
10686 		break;
10687 	case IXGBE_DEV_ID_82599EN_SFP:
10688 		/* Only these subdevices support WOL */
10689 		switch (subdevice_id) {
10690 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
10691 			return true;
10692 		}
10693 		break;
10694 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
10695 		/* All except this subdevice support WOL */
10696 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
10697 			return true;
10698 		break;
10699 	case IXGBE_DEV_ID_82599_KX4:
10700 		return  true;
10701 	default:
10702 		break;
10703 	}
10704 
10705 	return false;
10706 }
10707 
10708 /**
10709  * ixgbe_set_fw_version - Set FW version
10710  * @adapter: the adapter private structure
10711  *
10712  * This function is used by probe and ethtool to determine the FW version to
10713  * format to display. The FW version is taken from the EEPROM/NVM.
10714  */
10715 static void ixgbe_set_fw_version(struct ixgbe_adapter *adapter)
10716 {
10717 	struct ixgbe_hw *hw = &adapter->hw;
10718 	struct ixgbe_nvm_version nvm_ver;
10719 
10720 	ixgbe_get_oem_prod_version(hw, &nvm_ver);
10721 	if (nvm_ver.oem_valid) {
10722 		snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10723 			 "%x.%x.%x", nvm_ver.oem_major, nvm_ver.oem_minor,
10724 			 nvm_ver.oem_release);
10725 		return;
10726 	}
10727 
10728 	ixgbe_get_etk_id(hw, &nvm_ver);
10729 	ixgbe_get_orom_version(hw, &nvm_ver);
10730 
10731 	if (nvm_ver.or_valid) {
10732 		snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10733 			 "0x%08x, %d.%d.%d", nvm_ver.etk_id, nvm_ver.or_major,
10734 			 nvm_ver.or_build, nvm_ver.or_patch);
10735 		return;
10736 	}
10737 
10738 	/* Set ETrack ID format */
10739 	snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10740 		 "0x%08x", nvm_ver.etk_id);
10741 }
10742 
10743 /**
10744  * ixgbe_probe - Device Initialization Routine
10745  * @pdev: PCI device information struct
10746  * @ent: entry in ixgbe_pci_tbl
10747  *
10748  * Returns 0 on success, negative on failure
10749  *
10750  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
10751  * The OS initialization, configuring of the adapter private structure,
10752  * and a hardware reset occur.
10753  **/
10754 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10755 {
10756 	struct net_device *netdev;
10757 	struct ixgbe_adapter *adapter = NULL;
10758 	struct ixgbe_hw *hw;
10759 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
10760 	unsigned int indices = MAX_TX_QUEUES;
10761 	u8 part_str[IXGBE_PBANUM_LENGTH];
10762 	int i, err, expected_gts;
10763 	bool disable_dev = false;
10764 #ifdef IXGBE_FCOE
10765 	u16 device_caps;
10766 #endif
10767 	u32 eec;
10768 
10769 	/* Catch broken hardware that put the wrong VF device ID in
10770 	 * the PCIe SR-IOV capability.
10771 	 */
10772 	if (pdev->is_virtfn) {
10773 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
10774 		     pci_name(pdev), pdev->vendor, pdev->device);
10775 		return -EINVAL;
10776 	}
10777 
10778 	err = pci_enable_device_mem(pdev);
10779 	if (err)
10780 		return err;
10781 
10782 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
10783 	if (err) {
10784 		dev_err(&pdev->dev,
10785 			"No usable DMA configuration, aborting\n");
10786 		goto err_dma;
10787 	}
10788 
10789 	err = pci_request_mem_regions(pdev, ixgbe_driver_name);
10790 	if (err) {
10791 		dev_err(&pdev->dev,
10792 			"pci_request_selected_regions failed 0x%x\n", err);
10793 		goto err_pci_reg;
10794 	}
10795 
10796 	pci_set_master(pdev);
10797 	pci_save_state(pdev);
10798 
10799 	if (ii->mac == ixgbe_mac_82598EB) {
10800 #ifdef CONFIG_IXGBE_DCB
10801 		/* 8 TC w/ 4 queues per TC */
10802 		indices = 4 * MAX_TRAFFIC_CLASS;
10803 #else
10804 		indices = IXGBE_MAX_RSS_INDICES;
10805 #endif
10806 	}
10807 
10808 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
10809 	if (!netdev) {
10810 		err = -ENOMEM;
10811 		goto err_alloc_etherdev;
10812 	}
10813 
10814 	SET_NETDEV_DEV(netdev, &pdev->dev);
10815 
10816 	adapter = netdev_priv(netdev);
10817 
10818 	adapter->netdev = netdev;
10819 	adapter->pdev = pdev;
10820 	hw = &adapter->hw;
10821 	hw->back = adapter;
10822 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
10823 
10824 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
10825 			      pci_resource_len(pdev, 0));
10826 	adapter->io_addr = hw->hw_addr;
10827 	if (!hw->hw_addr) {
10828 		err = -EIO;
10829 		goto err_ioremap;
10830 	}
10831 
10832 	netdev->netdev_ops = &ixgbe_netdev_ops;
10833 	ixgbe_set_ethtool_ops(netdev);
10834 	netdev->watchdog_timeo = 5 * HZ;
10835 	strscpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
10836 
10837 	/* Setup hw api */
10838 	hw->mac.ops   = *ii->mac_ops;
10839 	hw->mac.type  = ii->mac;
10840 	hw->mvals     = ii->mvals;
10841 	if (ii->link_ops)
10842 		hw->link.ops  = *ii->link_ops;
10843 
10844 	/* EEPROM */
10845 	hw->eeprom.ops = *ii->eeprom_ops;
10846 	eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
10847 	if (ixgbe_removed(hw->hw_addr)) {
10848 		err = -EIO;
10849 		goto err_ioremap;
10850 	}
10851 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
10852 	if (!(eec & BIT(8)))
10853 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
10854 
10855 	/* PHY */
10856 	hw->phy.ops = *ii->phy_ops;
10857 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
10858 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
10859 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
10860 	hw->phy.mdio.mmds = 0;
10861 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
10862 	hw->phy.mdio.dev = netdev;
10863 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
10864 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
10865 
10866 	/* setup the private structure */
10867 	err = ixgbe_sw_init(adapter, ii);
10868 	if (err)
10869 		goto err_sw_init;
10870 
10871 	if (adapter->hw.mac.type == ixgbe_mac_82599EB)
10872 		adapter->flags2 |= IXGBE_FLAG2_AUTO_DISABLE_VF;
10873 
10874 	switch (adapter->hw.mac.type) {
10875 	case ixgbe_mac_X550:
10876 	case ixgbe_mac_X550EM_x:
10877 		netdev->udp_tunnel_nic_info = &ixgbe_udp_tunnels_x550;
10878 		break;
10879 	case ixgbe_mac_x550em_a:
10880 		netdev->udp_tunnel_nic_info = &ixgbe_udp_tunnels_x550em_a;
10881 		break;
10882 	default:
10883 		break;
10884 	}
10885 
10886 	/* Make sure the SWFW semaphore is in a valid state */
10887 	if (hw->mac.ops.init_swfw_sync)
10888 		hw->mac.ops.init_swfw_sync(hw);
10889 
10890 	/* Make it possible the adapter to be woken up via WOL */
10891 	switch (adapter->hw.mac.type) {
10892 	case ixgbe_mac_82599EB:
10893 	case ixgbe_mac_X540:
10894 	case ixgbe_mac_X550:
10895 	case ixgbe_mac_X550EM_x:
10896 	case ixgbe_mac_x550em_a:
10897 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10898 		break;
10899 	default:
10900 		break;
10901 	}
10902 
10903 	/*
10904 	 * If there is a fan on this device and it has failed log the
10905 	 * failure.
10906 	 */
10907 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
10908 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
10909 		if (esdp & IXGBE_ESDP_SDP1)
10910 			e_crit(probe, "Fan has stopped, replace the adapter\n");
10911 	}
10912 
10913 	if (allow_unsupported_sfp)
10914 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
10915 
10916 	/* reset_hw fills in the perm_addr as well */
10917 	hw->phy.reset_if_overtemp = true;
10918 	err = hw->mac.ops.reset_hw(hw);
10919 	hw->phy.reset_if_overtemp = false;
10920 	ixgbe_set_eee_capable(adapter);
10921 	if (err == -ENOENT) {
10922 		err = 0;
10923 	} else if (err == -EOPNOTSUPP) {
10924 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
10925 		e_dev_err("Reload the driver after installing a supported module.\n");
10926 		goto err_sw_init;
10927 	} else if (err) {
10928 		e_dev_err("HW Init failed: %d\n", err);
10929 		goto err_sw_init;
10930 	}
10931 
10932 #ifdef CONFIG_PCI_IOV
10933 	/* SR-IOV not supported on the 82598 */
10934 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
10935 		goto skip_sriov;
10936 	/* Mailbox */
10937 	ixgbe_init_mbx_params_pf(hw);
10938 	hw->mbx.ops = ii->mbx_ops;
10939 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
10940 	ixgbe_enable_sriov(adapter, max_vfs);
10941 skip_sriov:
10942 
10943 #endif
10944 	netdev->features = NETIF_F_SG |
10945 			   NETIF_F_TSO |
10946 			   NETIF_F_TSO6 |
10947 			   NETIF_F_RXHASH |
10948 			   NETIF_F_RXCSUM |
10949 			   NETIF_F_HW_CSUM;
10950 
10951 #define IXGBE_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
10952 				    NETIF_F_GSO_GRE_CSUM | \
10953 				    NETIF_F_GSO_IPXIP4 | \
10954 				    NETIF_F_GSO_IPXIP6 | \
10955 				    NETIF_F_GSO_UDP_TUNNEL | \
10956 				    NETIF_F_GSO_UDP_TUNNEL_CSUM)
10957 
10958 	netdev->gso_partial_features = IXGBE_GSO_PARTIAL_FEATURES;
10959 	netdev->features |= NETIF_F_GSO_PARTIAL |
10960 			    IXGBE_GSO_PARTIAL_FEATURES;
10961 
10962 	if (hw->mac.type >= ixgbe_mac_82599EB)
10963 		netdev->features |= NETIF_F_SCTP_CRC | NETIF_F_GSO_UDP_L4;
10964 
10965 #ifdef CONFIG_IXGBE_IPSEC
10966 #define IXGBE_ESP_FEATURES	(NETIF_F_HW_ESP | \
10967 				 NETIF_F_HW_ESP_TX_CSUM | \
10968 				 NETIF_F_GSO_ESP)
10969 
10970 	if (adapter->ipsec)
10971 		netdev->features |= IXGBE_ESP_FEATURES;
10972 #endif
10973 	/* copy netdev features into list of user selectable features */
10974 	netdev->hw_features |= netdev->features |
10975 			       NETIF_F_HW_VLAN_CTAG_FILTER |
10976 			       NETIF_F_HW_VLAN_CTAG_RX |
10977 			       NETIF_F_HW_VLAN_CTAG_TX |
10978 			       NETIF_F_RXALL |
10979 			       NETIF_F_HW_L2FW_DOFFLOAD;
10980 
10981 	if (hw->mac.type >= ixgbe_mac_82599EB)
10982 		netdev->hw_features |= NETIF_F_NTUPLE |
10983 				       NETIF_F_HW_TC;
10984 
10985 	netdev->features |= NETIF_F_HIGHDMA;
10986 
10987 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
10988 	netdev->hw_enc_features |= netdev->vlan_features;
10989 	netdev->mpls_features |= NETIF_F_SG |
10990 				 NETIF_F_TSO |
10991 				 NETIF_F_TSO6 |
10992 				 NETIF_F_HW_CSUM;
10993 	netdev->mpls_features |= IXGBE_GSO_PARTIAL_FEATURES;
10994 
10995 	/* set this bit last since it cannot be part of vlan_features */
10996 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
10997 			    NETIF_F_HW_VLAN_CTAG_RX |
10998 			    NETIF_F_HW_VLAN_CTAG_TX;
10999 
11000 	netdev->priv_flags |= IFF_UNICAST_FLT;
11001 	netdev->priv_flags |= IFF_SUPP_NOFCS;
11002 
11003 	netdev->xdp_features = NETDEV_XDP_ACT_BASIC | NETDEV_XDP_ACT_REDIRECT |
11004 			       NETDEV_XDP_ACT_XSK_ZEROCOPY;
11005 
11006 	/* MTU range: 68 - 9710 */
11007 	netdev->min_mtu = ETH_MIN_MTU;
11008 	netdev->max_mtu = IXGBE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN);
11009 
11010 #ifdef CONFIG_IXGBE_DCB
11011 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
11012 		netdev->dcbnl_ops = &ixgbe_dcbnl_ops;
11013 #endif
11014 
11015 #ifdef IXGBE_FCOE
11016 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
11017 		unsigned int fcoe_l;
11018 
11019 		if (hw->mac.ops.get_device_caps) {
11020 			hw->mac.ops.get_device_caps(hw, &device_caps);
11021 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
11022 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
11023 		}
11024 
11025 
11026 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
11027 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
11028 
11029 		netdev->features |= NETIF_F_FSO |
11030 				    NETIF_F_FCOE_CRC;
11031 
11032 		netdev->vlan_features |= NETIF_F_FSO |
11033 					 NETIF_F_FCOE_CRC |
11034 					 NETIF_F_FCOE_MTU;
11035 	}
11036 #endif /* IXGBE_FCOE */
11037 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
11038 		netdev->hw_features |= NETIF_F_LRO;
11039 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
11040 		netdev->features |= NETIF_F_LRO;
11041 
11042 	if (ixgbe_check_fw_error(adapter)) {
11043 		err = -EIO;
11044 		goto err_sw_init;
11045 	}
11046 
11047 	/* make sure the EEPROM is good */
11048 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
11049 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
11050 		err = -EIO;
11051 		goto err_sw_init;
11052 	}
11053 
11054 	eth_platform_get_mac_address(&adapter->pdev->dev,
11055 				     adapter->hw.mac.perm_addr);
11056 
11057 	eth_hw_addr_set(netdev, hw->mac.perm_addr);
11058 
11059 	if (!is_valid_ether_addr(netdev->dev_addr)) {
11060 		e_dev_err("invalid MAC address\n");
11061 		err = -EIO;
11062 		goto err_sw_init;
11063 	}
11064 
11065 	/* Set hw->mac.addr to permanent MAC address */
11066 	ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
11067 	ixgbe_mac_set_default_filter(adapter);
11068 
11069 	timer_setup(&adapter->service_timer, ixgbe_service_timer, 0);
11070 
11071 	if (ixgbe_removed(hw->hw_addr)) {
11072 		err = -EIO;
11073 		goto err_sw_init;
11074 	}
11075 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
11076 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
11077 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
11078 
11079 	err = ixgbe_init_interrupt_scheme(adapter);
11080 	if (err)
11081 		goto err_sw_init;
11082 
11083 	for (i = 0; i < adapter->num_rx_queues; i++)
11084 		u64_stats_init(&adapter->rx_ring[i]->syncp);
11085 	for (i = 0; i < adapter->num_tx_queues; i++)
11086 		u64_stats_init(&adapter->tx_ring[i]->syncp);
11087 	for (i = 0; i < adapter->num_xdp_queues; i++)
11088 		u64_stats_init(&adapter->xdp_ring[i]->syncp);
11089 
11090 	/* WOL not supported for all devices */
11091 	adapter->wol = 0;
11092 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
11093 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
11094 						pdev->subsystem_device);
11095 	if (hw->wol_enabled)
11096 		adapter->wol = IXGBE_WUFC_MAG;
11097 
11098 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
11099 
11100 	/* save off EEPROM version number */
11101 	ixgbe_set_fw_version(adapter);
11102 
11103 	/* pick up the PCI bus settings for reporting later */
11104 	if (ixgbe_pcie_from_parent(hw))
11105 		ixgbe_get_parent_bus_info(adapter);
11106 	else
11107 		 hw->mac.ops.get_bus_info(hw);
11108 
11109 	/* calculate the expected PCIe bandwidth required for optimal
11110 	 * performance. Note that some older parts will never have enough
11111 	 * bandwidth due to being older generation PCIe parts. We clamp these
11112 	 * parts to ensure no warning is displayed if it can't be fixed.
11113 	 */
11114 	switch (hw->mac.type) {
11115 	case ixgbe_mac_82598EB:
11116 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
11117 		break;
11118 	default:
11119 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
11120 		break;
11121 	}
11122 
11123 	/* don't check link if we failed to enumerate functions */
11124 	if (expected_gts > 0)
11125 		ixgbe_check_minimum_link(adapter, expected_gts);
11126 
11127 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
11128 	if (err)
11129 		strscpy(part_str, "Unknown", sizeof(part_str));
11130 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
11131 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
11132 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
11133 			   part_str);
11134 	else
11135 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
11136 			   hw->mac.type, hw->phy.type, part_str);
11137 
11138 	e_dev_info("%pM\n", netdev->dev_addr);
11139 
11140 	/* reset the hardware with the new settings */
11141 	err = hw->mac.ops.start_hw(hw);
11142 	if (err == -EACCES) {
11143 		/* We are running on a pre-production device, log a warning */
11144 		e_dev_warn("This device is a pre-production adapter/LOM. "
11145 			   "Please be aware there may be issues associated "
11146 			   "with your hardware.  If you are experiencing "
11147 			   "problems please contact your Intel or hardware "
11148 			   "representative who provided you with this "
11149 			   "hardware.\n");
11150 	}
11151 	strcpy(netdev->name, "eth%d");
11152 	pci_set_drvdata(pdev, adapter);
11153 	err = register_netdev(netdev);
11154 	if (err)
11155 		goto err_register;
11156 
11157 
11158 	/* power down the optics for 82599 SFP+ fiber */
11159 	if (hw->mac.ops.disable_tx_laser)
11160 		hw->mac.ops.disable_tx_laser(hw);
11161 
11162 	/* carrier off reporting is important to ethtool even BEFORE open */
11163 	netif_carrier_off(netdev);
11164 
11165 #ifdef CONFIG_IXGBE_DCA
11166 	if (dca_add_requester(&pdev->dev) == 0) {
11167 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
11168 		ixgbe_setup_dca(adapter);
11169 	}
11170 #endif
11171 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
11172 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
11173 		for (i = 0; i < adapter->num_vfs; i++)
11174 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
11175 	}
11176 
11177 	/* firmware requires driver version to be 0xFFFFFFFF
11178 	 * since os does not support feature
11179 	 */
11180 	if (hw->mac.ops.set_fw_drv_ver)
11181 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF, 0xFF,
11182 					   sizeof(UTS_RELEASE) - 1,
11183 					   UTS_RELEASE);
11184 
11185 	/* add san mac addr to netdev */
11186 	ixgbe_add_sanmac_netdev(netdev);
11187 
11188 	e_dev_info("%s\n", ixgbe_default_device_descr);
11189 
11190 #ifdef CONFIG_IXGBE_HWMON
11191 	if (ixgbe_sysfs_init(adapter))
11192 		e_err(probe, "failed to allocate sysfs resources\n");
11193 #endif /* CONFIG_IXGBE_HWMON */
11194 
11195 	ixgbe_dbg_adapter_init(adapter);
11196 
11197 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
11198 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
11199 		hw->mac.ops.setup_link(hw,
11200 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
11201 			true);
11202 
11203 	err = ixgbe_mii_bus_init(hw);
11204 	if (err)
11205 		goto err_netdev;
11206 
11207 	return 0;
11208 
11209 err_netdev:
11210 	unregister_netdev(netdev);
11211 err_register:
11212 	ixgbe_release_hw_control(adapter);
11213 	ixgbe_clear_interrupt_scheme(adapter);
11214 err_sw_init:
11215 	ixgbe_disable_sriov(adapter);
11216 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
11217 	iounmap(adapter->io_addr);
11218 	kfree(adapter->jump_tables[0]);
11219 	kfree(adapter->mac_table);
11220 	kfree(adapter->rss_key);
11221 	bitmap_free(adapter->af_xdp_zc_qps);
11222 err_ioremap:
11223 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
11224 	free_netdev(netdev);
11225 err_alloc_etherdev:
11226 	pci_release_mem_regions(pdev);
11227 err_pci_reg:
11228 err_dma:
11229 	if (!adapter || disable_dev)
11230 		pci_disable_device(pdev);
11231 	return err;
11232 }
11233 
11234 /**
11235  * ixgbe_remove - Device Removal Routine
11236  * @pdev: PCI device information struct
11237  *
11238  * ixgbe_remove is called by the PCI subsystem to alert the driver
11239  * that it should release a PCI device.  The could be caused by a
11240  * Hot-Plug event, or because the driver is going to be removed from
11241  * memory.
11242  **/
11243 static void ixgbe_remove(struct pci_dev *pdev)
11244 {
11245 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11246 	struct net_device *netdev;
11247 	bool disable_dev;
11248 	int i;
11249 
11250 	/* if !adapter then we already cleaned up in probe */
11251 	if (!adapter)
11252 		return;
11253 
11254 	netdev  = adapter->netdev;
11255 	ixgbe_dbg_adapter_exit(adapter);
11256 
11257 	set_bit(__IXGBE_REMOVING, &adapter->state);
11258 	cancel_work_sync(&adapter->service_task);
11259 
11260 	if (adapter->mii_bus)
11261 		mdiobus_unregister(adapter->mii_bus);
11262 
11263 #ifdef CONFIG_IXGBE_DCA
11264 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
11265 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
11266 		dca_remove_requester(&pdev->dev);
11267 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
11268 				IXGBE_DCA_CTRL_DCA_DISABLE);
11269 	}
11270 
11271 #endif
11272 #ifdef CONFIG_IXGBE_HWMON
11273 	ixgbe_sysfs_exit(adapter);
11274 #endif /* CONFIG_IXGBE_HWMON */
11275 
11276 	/* remove the added san mac */
11277 	ixgbe_del_sanmac_netdev(netdev);
11278 
11279 #ifdef CONFIG_PCI_IOV
11280 	ixgbe_disable_sriov(adapter);
11281 #endif
11282 	if (netdev->reg_state == NETREG_REGISTERED)
11283 		unregister_netdev(netdev);
11284 
11285 	ixgbe_stop_ipsec_offload(adapter);
11286 	ixgbe_clear_interrupt_scheme(adapter);
11287 
11288 	ixgbe_release_hw_control(adapter);
11289 
11290 #ifdef CONFIG_DCB
11291 	kfree(adapter->ixgbe_ieee_pfc);
11292 	kfree(adapter->ixgbe_ieee_ets);
11293 
11294 #endif
11295 	iounmap(adapter->io_addr);
11296 	pci_release_mem_regions(pdev);
11297 
11298 	e_dev_info("complete\n");
11299 
11300 	for (i = 0; i < IXGBE_MAX_LINK_HANDLE; i++) {
11301 		if (adapter->jump_tables[i]) {
11302 			kfree(adapter->jump_tables[i]->input);
11303 			kfree(adapter->jump_tables[i]->mask);
11304 		}
11305 		kfree(adapter->jump_tables[i]);
11306 	}
11307 
11308 	kfree(adapter->mac_table);
11309 	kfree(adapter->rss_key);
11310 	bitmap_free(adapter->af_xdp_zc_qps);
11311 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
11312 	free_netdev(netdev);
11313 
11314 	if (disable_dev)
11315 		pci_disable_device(pdev);
11316 }
11317 
11318 /**
11319  * ixgbe_io_error_detected - called when PCI error is detected
11320  * @pdev: Pointer to PCI device
11321  * @state: The current pci connection state
11322  *
11323  * This function is called after a PCI bus error affecting
11324  * this device has been detected.
11325  */
11326 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
11327 						pci_channel_state_t state)
11328 {
11329 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11330 	struct net_device *netdev = adapter->netdev;
11331 
11332 #ifdef CONFIG_PCI_IOV
11333 	struct ixgbe_hw *hw = &adapter->hw;
11334 	struct pci_dev *bdev, *vfdev;
11335 	u32 dw0, dw1, dw2, dw3;
11336 	int vf, pos;
11337 	u16 req_id, pf_func;
11338 
11339 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
11340 	    adapter->num_vfs == 0)
11341 		goto skip_bad_vf_detection;
11342 
11343 	bdev = pdev->bus->self;
11344 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
11345 		bdev = bdev->bus->self;
11346 
11347 	if (!bdev)
11348 		goto skip_bad_vf_detection;
11349 
11350 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
11351 	if (!pos)
11352 		goto skip_bad_vf_detection;
11353 
11354 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
11355 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
11356 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
11357 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
11358 	if (ixgbe_removed(hw->hw_addr))
11359 		goto skip_bad_vf_detection;
11360 
11361 	req_id = dw1 >> 16;
11362 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
11363 	if (!(req_id & 0x0080))
11364 		goto skip_bad_vf_detection;
11365 
11366 	pf_func = req_id & 0x01;
11367 	if ((pf_func & 1) == (pdev->devfn & 1)) {
11368 		unsigned int device_id;
11369 
11370 		vf = FIELD_GET(0x7F, req_id);
11371 		e_dev_err("VF %d has caused a PCIe error\n", vf);
11372 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
11373 				"%8.8x\tdw3: %8.8x\n",
11374 		dw0, dw1, dw2, dw3);
11375 		switch (adapter->hw.mac.type) {
11376 		case ixgbe_mac_82599EB:
11377 			device_id = IXGBE_82599_VF_DEVICE_ID;
11378 			break;
11379 		case ixgbe_mac_X540:
11380 			device_id = IXGBE_X540_VF_DEVICE_ID;
11381 			break;
11382 		case ixgbe_mac_X550:
11383 			device_id = IXGBE_DEV_ID_X550_VF;
11384 			break;
11385 		case ixgbe_mac_X550EM_x:
11386 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
11387 			break;
11388 		case ixgbe_mac_x550em_a:
11389 			device_id = IXGBE_DEV_ID_X550EM_A_VF;
11390 			break;
11391 		default:
11392 			device_id = 0;
11393 			break;
11394 		}
11395 
11396 		/* Find the pci device of the offending VF */
11397 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
11398 		while (vfdev) {
11399 			if (vfdev->devfn == (req_id & 0xFF))
11400 				break;
11401 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
11402 					       device_id, vfdev);
11403 		}
11404 		/*
11405 		 * There's a slim chance the VF could have been hot plugged,
11406 		 * so if it is no longer present we don't need to issue the
11407 		 * VFLR.  Just clean up the AER in that case.
11408 		 */
11409 		if (vfdev) {
11410 			pcie_flr(vfdev);
11411 			/* Free device reference count */
11412 			pci_dev_put(vfdev);
11413 		}
11414 	}
11415 
11416 	/*
11417 	 * Even though the error may have occurred on the other port
11418 	 * we still need to increment the vf error reference count for
11419 	 * both ports because the I/O resume function will be called
11420 	 * for both of them.
11421 	 */
11422 	adapter->vferr_refcount++;
11423 
11424 	return PCI_ERS_RESULT_RECOVERED;
11425 
11426 skip_bad_vf_detection:
11427 #endif /* CONFIG_PCI_IOV */
11428 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
11429 		return PCI_ERS_RESULT_DISCONNECT;
11430 
11431 	if (!netif_device_present(netdev))
11432 		return PCI_ERS_RESULT_DISCONNECT;
11433 
11434 	rtnl_lock();
11435 	netif_device_detach(netdev);
11436 
11437 	if (netif_running(netdev))
11438 		ixgbe_close_suspend(adapter);
11439 
11440 	if (state == pci_channel_io_perm_failure) {
11441 		rtnl_unlock();
11442 		return PCI_ERS_RESULT_DISCONNECT;
11443 	}
11444 
11445 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
11446 		pci_disable_device(pdev);
11447 	rtnl_unlock();
11448 
11449 	/* Request a slot reset. */
11450 	return PCI_ERS_RESULT_NEED_RESET;
11451 }
11452 
11453 /**
11454  * ixgbe_io_slot_reset - called after the pci bus has been reset.
11455  * @pdev: Pointer to PCI device
11456  *
11457  * Restart the card from scratch, as if from a cold-boot.
11458  */
11459 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
11460 {
11461 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11462 	pci_ers_result_t result;
11463 
11464 	if (pci_enable_device_mem(pdev)) {
11465 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
11466 		result = PCI_ERS_RESULT_DISCONNECT;
11467 	} else {
11468 		smp_mb__before_atomic();
11469 		clear_bit(__IXGBE_DISABLED, &adapter->state);
11470 		adapter->hw.hw_addr = adapter->io_addr;
11471 		pci_set_master(pdev);
11472 		pci_restore_state(pdev);
11473 		pci_save_state(pdev);
11474 
11475 		pci_wake_from_d3(pdev, false);
11476 
11477 		ixgbe_reset(adapter);
11478 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
11479 		result = PCI_ERS_RESULT_RECOVERED;
11480 	}
11481 
11482 	return result;
11483 }
11484 
11485 /**
11486  * ixgbe_io_resume - called when traffic can start flowing again.
11487  * @pdev: Pointer to PCI device
11488  *
11489  * This callback is called when the error recovery driver tells us that
11490  * its OK to resume normal operation.
11491  */
11492 static void ixgbe_io_resume(struct pci_dev *pdev)
11493 {
11494 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11495 	struct net_device *netdev = adapter->netdev;
11496 
11497 #ifdef CONFIG_PCI_IOV
11498 	if (adapter->vferr_refcount) {
11499 		e_info(drv, "Resuming after VF err\n");
11500 		adapter->vferr_refcount--;
11501 		return;
11502 	}
11503 
11504 #endif
11505 	rtnl_lock();
11506 	if (netif_running(netdev))
11507 		ixgbe_open(netdev);
11508 
11509 	netif_device_attach(netdev);
11510 	rtnl_unlock();
11511 }
11512 
11513 static const struct pci_error_handlers ixgbe_err_handler = {
11514 	.error_detected = ixgbe_io_error_detected,
11515 	.slot_reset = ixgbe_io_slot_reset,
11516 	.resume = ixgbe_io_resume,
11517 };
11518 
11519 static SIMPLE_DEV_PM_OPS(ixgbe_pm_ops, ixgbe_suspend, ixgbe_resume);
11520 
11521 static struct pci_driver ixgbe_driver = {
11522 	.name      = ixgbe_driver_name,
11523 	.id_table  = ixgbe_pci_tbl,
11524 	.probe     = ixgbe_probe,
11525 	.remove    = ixgbe_remove,
11526 	.driver.pm = &ixgbe_pm_ops,
11527 	.shutdown  = ixgbe_shutdown,
11528 	.sriov_configure = ixgbe_pci_sriov_configure,
11529 	.err_handler = &ixgbe_err_handler
11530 };
11531 
11532 /**
11533  * ixgbe_init_module - Driver Registration Routine
11534  *
11535  * ixgbe_init_module is the first routine called when the driver is
11536  * loaded. All it does is register with the PCI subsystem.
11537  **/
11538 static int __init ixgbe_init_module(void)
11539 {
11540 	int ret;
11541 	pr_info("%s\n", ixgbe_driver_string);
11542 	pr_info("%s\n", ixgbe_copyright);
11543 
11544 	ixgbe_wq = create_singlethread_workqueue(ixgbe_driver_name);
11545 	if (!ixgbe_wq) {
11546 		pr_err("%s: Failed to create workqueue\n", ixgbe_driver_name);
11547 		return -ENOMEM;
11548 	}
11549 
11550 	ixgbe_dbg_init();
11551 
11552 	ret = pci_register_driver(&ixgbe_driver);
11553 	if (ret) {
11554 		destroy_workqueue(ixgbe_wq);
11555 		ixgbe_dbg_exit();
11556 		return ret;
11557 	}
11558 
11559 #ifdef CONFIG_IXGBE_DCA
11560 	dca_register_notify(&dca_notifier);
11561 #endif
11562 
11563 	return 0;
11564 }
11565 
11566 module_init(ixgbe_init_module);
11567 
11568 /**
11569  * ixgbe_exit_module - Driver Exit Cleanup Routine
11570  *
11571  * ixgbe_exit_module is called just before the driver is removed
11572  * from memory.
11573  **/
11574 static void __exit ixgbe_exit_module(void)
11575 {
11576 #ifdef CONFIG_IXGBE_DCA
11577 	dca_unregister_notify(&dca_notifier);
11578 #endif
11579 	pci_unregister_driver(&ixgbe_driver);
11580 
11581 	ixgbe_dbg_exit();
11582 	if (ixgbe_wq) {
11583 		destroy_workqueue(ixgbe_wq);
11584 		ixgbe_wq = NULL;
11585 	}
11586 }
11587 
11588 #ifdef CONFIG_IXGBE_DCA
11589 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
11590 			    void *p)
11591 {
11592 	int ret_val;
11593 
11594 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
11595 					 __ixgbe_notify_dca);
11596 
11597 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
11598 }
11599 
11600 #endif /* CONFIG_IXGBE_DCA */
11601 
11602 module_exit(ixgbe_exit_module);
11603 
11604 /* ixgbe_main.c */
11605