xref: /linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision 46e6acfe3501fa938af9c5bd730f0020235b08a2)
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, 0444);
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_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
166 MODULE_LICENSE("GPL v2");
167 
168 DEFINE_STATIC_KEY_FALSE(ixgbe_xdp_locking_key);
169 EXPORT_SYMBOL(ixgbe_xdp_locking_key);
170 
171 static struct workqueue_struct *ixgbe_wq;
172 
173 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
174 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *);
175 
176 static const struct net_device_ops ixgbe_netdev_ops;
177 
178 static bool netif_is_ixgbe(struct net_device *dev)
179 {
180 	return dev && (dev->netdev_ops == &ixgbe_netdev_ops);
181 }
182 
183 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
184 					  u32 reg, u16 *value)
185 {
186 	struct pci_dev *parent_dev;
187 	struct pci_bus *parent_bus;
188 
189 	parent_bus = adapter->pdev->bus->parent;
190 	if (!parent_bus)
191 		return -1;
192 
193 	parent_dev = parent_bus->self;
194 	if (!parent_dev)
195 		return -1;
196 
197 	if (!pci_is_pcie(parent_dev))
198 		return -1;
199 
200 	pcie_capability_read_word(parent_dev, reg, value);
201 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
202 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
203 		return -1;
204 	return 0;
205 }
206 
207 static int ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
208 {
209 	struct ixgbe_hw *hw = &adapter->hw;
210 	u16 link_status = 0;
211 	int err;
212 
213 	hw->bus.type = ixgbe_bus_type_pci_express;
214 
215 	/* Get the negotiated link width and speed from PCI config space of the
216 	 * parent, as this device is behind a switch
217 	 */
218 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
219 
220 	/* assume caller will handle error case */
221 	if (err)
222 		return err;
223 
224 	hw->bus.width = ixgbe_convert_bus_width(link_status);
225 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
226 
227 	return 0;
228 }
229 
230 /**
231  * ixgbe_pcie_from_parent - Determine whether PCIe info should come from parent
232  * @hw: hw specific details
233  *
234  * This function is used by probe to determine whether a device's PCI-Express
235  * bandwidth details should be gathered from the parent bus instead of from the
236  * device. Used to ensure that various locations all have the correct device ID
237  * checks.
238  */
239 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
240 {
241 	switch (hw->device_id) {
242 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
243 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
244 		return true;
245 	default:
246 		return false;
247 	}
248 }
249 
250 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
251 				     int expected_gts)
252 {
253 	struct ixgbe_hw *hw = &adapter->hw;
254 	struct pci_dev *pdev;
255 
256 	/* Some devices are not connected over PCIe and thus do not negotiate
257 	 * speed. These devices do not have valid bus info, and thus any report
258 	 * we generate may not be correct.
259 	 */
260 	if (hw->bus.type == ixgbe_bus_type_internal)
261 		return;
262 
263 	/* determine whether to use the parent device */
264 	if (ixgbe_pcie_from_parent(&adapter->hw))
265 		pdev = adapter->pdev->bus->parent->self;
266 	else
267 		pdev = adapter->pdev;
268 
269 	pcie_print_link_status(pdev);
270 }
271 
272 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
273 {
274 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
275 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
276 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
277 		queue_work(ixgbe_wq, &adapter->service_task);
278 }
279 
280 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
281 {
282 	struct ixgbe_adapter *adapter = hw->back;
283 
284 	if (!hw->hw_addr)
285 		return;
286 	hw->hw_addr = NULL;
287 	e_dev_err("Adapter removed\n");
288 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
289 		ixgbe_service_event_schedule(adapter);
290 }
291 
292 static u32 ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
293 {
294 	u8 __iomem *reg_addr;
295 	u32 value;
296 	int i;
297 
298 	reg_addr = READ_ONCE(hw->hw_addr);
299 	if (ixgbe_removed(reg_addr))
300 		return IXGBE_FAILED_READ_REG;
301 
302 	/* Register read of 0xFFFFFFF can indicate the adapter has been removed,
303 	 * so perform several status register reads to determine if the adapter
304 	 * has been removed.
305 	 */
306 	for (i = 0; i < IXGBE_FAILED_READ_RETRIES; i++) {
307 		value = readl(reg_addr + IXGBE_STATUS);
308 		if (value != IXGBE_FAILED_READ_REG)
309 			break;
310 		mdelay(3);
311 	}
312 
313 	if (value == IXGBE_FAILED_READ_REG)
314 		ixgbe_remove_adapter(hw);
315 	else
316 		value = readl(reg_addr + reg);
317 	return value;
318 }
319 
320 /**
321  * ixgbe_read_reg - Read from device register
322  * @hw: hw specific details
323  * @reg: offset of register to read
324  *
325  * Returns : value read or IXGBE_FAILED_READ_REG if removed
326  *
327  * This function is used to read device registers. It checks for device
328  * removal by confirming any read that returns all ones by checking the
329  * status register value for all ones. This function avoids reading from
330  * the hardware if a removal was previously detected in which case it
331  * returns IXGBE_FAILED_READ_REG (all ones).
332  */
333 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
334 {
335 	u8 __iomem *reg_addr = READ_ONCE(hw->hw_addr);
336 	u32 value;
337 
338 	if (ixgbe_removed(reg_addr))
339 		return IXGBE_FAILED_READ_REG;
340 	if (unlikely(hw->phy.nw_mng_if_sel &
341 		     IXGBE_NW_MNG_IF_SEL_SGMII_ENABLE)) {
342 		struct ixgbe_adapter *adapter;
343 		int i;
344 
345 		for (i = 0; i < 200; ++i) {
346 			value = readl(reg_addr + IXGBE_MAC_SGMII_BUSY);
347 			if (likely(!value))
348 				goto writes_completed;
349 			if (value == IXGBE_FAILED_READ_REG) {
350 				ixgbe_remove_adapter(hw);
351 				return IXGBE_FAILED_READ_REG;
352 			}
353 			udelay(5);
354 		}
355 
356 		adapter = hw->back;
357 		e_warn(hw, "register writes incomplete %08x\n", value);
358 	}
359 
360 writes_completed:
361 	value = readl(reg_addr + reg);
362 	if (unlikely(value == IXGBE_FAILED_READ_REG))
363 		value = ixgbe_check_remove(hw, reg);
364 	return value;
365 }
366 
367 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
368 {
369 	u16 value;
370 
371 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
372 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
373 		ixgbe_remove_adapter(hw);
374 		return true;
375 	}
376 	return false;
377 }
378 
379 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
380 {
381 	struct ixgbe_adapter *adapter = hw->back;
382 	u16 value;
383 
384 	if (ixgbe_removed(hw->hw_addr))
385 		return IXGBE_FAILED_READ_CFG_WORD;
386 	pci_read_config_word(adapter->pdev, reg, &value);
387 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
388 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
389 		return IXGBE_FAILED_READ_CFG_WORD;
390 	return value;
391 }
392 
393 #ifdef CONFIG_PCI_IOV
394 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
395 {
396 	struct ixgbe_adapter *adapter = hw->back;
397 	u32 value;
398 
399 	if (ixgbe_removed(hw->hw_addr))
400 		return IXGBE_FAILED_READ_CFG_DWORD;
401 	pci_read_config_dword(adapter->pdev, reg, &value);
402 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
403 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
404 		return IXGBE_FAILED_READ_CFG_DWORD;
405 	return value;
406 }
407 #endif /* CONFIG_PCI_IOV */
408 
409 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
410 {
411 	struct ixgbe_adapter *adapter = hw->back;
412 
413 	if (ixgbe_removed(hw->hw_addr))
414 		return;
415 	pci_write_config_word(adapter->pdev, reg, value);
416 }
417 
418 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
419 {
420 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
421 
422 	/* flush memory to make sure state is correct before next watchdog */
423 	smp_mb__before_atomic();
424 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
425 }
426 
427 struct ixgbe_reg_info {
428 	u32 ofs;
429 	char *name;
430 };
431 
432 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
433 
434 	/* General Registers */
435 	{IXGBE_CTRL, "CTRL"},
436 	{IXGBE_STATUS, "STATUS"},
437 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
438 
439 	/* Interrupt Registers */
440 	{IXGBE_EICR, "EICR"},
441 
442 	/* RX Registers */
443 	{IXGBE_SRRCTL(0), "SRRCTL"},
444 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
445 	{IXGBE_RDLEN(0), "RDLEN"},
446 	{IXGBE_RDH(0), "RDH"},
447 	{IXGBE_RDT(0), "RDT"},
448 	{IXGBE_RXDCTL(0), "RXDCTL"},
449 	{IXGBE_RDBAL(0), "RDBAL"},
450 	{IXGBE_RDBAH(0), "RDBAH"},
451 
452 	/* TX Registers */
453 	{IXGBE_TDBAL(0), "TDBAL"},
454 	{IXGBE_TDBAH(0), "TDBAH"},
455 	{IXGBE_TDLEN(0), "TDLEN"},
456 	{IXGBE_TDH(0), "TDH"},
457 	{IXGBE_TDT(0), "TDT"},
458 	{IXGBE_TXDCTL(0), "TXDCTL"},
459 
460 	/* List Terminator */
461 	{ .name = NULL }
462 };
463 
464 
465 /*
466  * ixgbe_regdump - register printout routine
467  */
468 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
469 {
470 	int i;
471 	char rname[16];
472 	u32 regs[64];
473 
474 	switch (reginfo->ofs) {
475 	case IXGBE_SRRCTL(0):
476 		for (i = 0; i < 64; i++)
477 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
478 		break;
479 	case IXGBE_DCA_RXCTRL(0):
480 		for (i = 0; i < 64; i++)
481 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
482 		break;
483 	case IXGBE_RDLEN(0):
484 		for (i = 0; i < 64; i++)
485 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
486 		break;
487 	case IXGBE_RDH(0):
488 		for (i = 0; i < 64; i++)
489 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
490 		break;
491 	case IXGBE_RDT(0):
492 		for (i = 0; i < 64; i++)
493 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
494 		break;
495 	case IXGBE_RXDCTL(0):
496 		for (i = 0; i < 64; i++)
497 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
498 		break;
499 	case IXGBE_RDBAL(0):
500 		for (i = 0; i < 64; i++)
501 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
502 		break;
503 	case IXGBE_RDBAH(0):
504 		for (i = 0; i < 64; i++)
505 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
506 		break;
507 	case IXGBE_TDBAL(0):
508 		for (i = 0; i < 64; i++)
509 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
510 		break;
511 	case IXGBE_TDBAH(0):
512 		for (i = 0; i < 64; i++)
513 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
514 		break;
515 	case IXGBE_TDLEN(0):
516 		for (i = 0; i < 64; i++)
517 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
518 		break;
519 	case IXGBE_TDH(0):
520 		for (i = 0; i < 64; i++)
521 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
522 		break;
523 	case IXGBE_TDT(0):
524 		for (i = 0; i < 64; i++)
525 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
526 		break;
527 	case IXGBE_TXDCTL(0):
528 		for (i = 0; i < 64; i++)
529 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
530 		break;
531 	default:
532 		pr_info("%-15s %08x\n",
533 			reginfo->name, IXGBE_READ_REG(hw, reginfo->ofs));
534 		return;
535 	}
536 
537 	i = 0;
538 	while (i < 64) {
539 		int j;
540 		char buf[9 * 8 + 1];
541 		char *p = buf;
542 
543 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i, i + 7);
544 		for (j = 0; j < 8; j++)
545 			p += sprintf(p, " %08x", regs[i++]);
546 		pr_err("%-15s%s\n", rname, buf);
547 	}
548 
549 }
550 
551 static void ixgbe_print_buffer(struct ixgbe_ring *ring, int n)
552 {
553 	struct ixgbe_tx_buffer *tx_buffer;
554 
555 	tx_buffer = &ring->tx_buffer_info[ring->next_to_clean];
556 	pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
557 		n, ring->next_to_use, ring->next_to_clean,
558 		(u64)dma_unmap_addr(tx_buffer, dma),
559 		dma_unmap_len(tx_buffer, len),
560 		tx_buffer->next_to_watch,
561 		(u64)tx_buffer->time_stamp);
562 }
563 
564 /*
565  * ixgbe_dump - Print registers, tx-rings and rx-rings
566  */
567 static void ixgbe_dump(struct ixgbe_adapter *adapter)
568 {
569 	struct net_device *netdev = adapter->netdev;
570 	struct ixgbe_hw *hw = &adapter->hw;
571 	struct ixgbe_reg_info *reginfo;
572 	int n = 0;
573 	struct ixgbe_ring *ring;
574 	struct ixgbe_tx_buffer *tx_buffer;
575 	union ixgbe_adv_tx_desc *tx_desc;
576 	struct my_u0 { u64 a; u64 b; } *u0;
577 	struct ixgbe_ring *rx_ring;
578 	union ixgbe_adv_rx_desc *rx_desc;
579 	struct ixgbe_rx_buffer *rx_buffer_info;
580 	int i = 0;
581 
582 	if (!netif_msg_hw(adapter))
583 		return;
584 
585 	/* Print netdevice Info */
586 	if (netdev) {
587 		dev_info(&adapter->pdev->dev, "Net device Info\n");
588 		pr_info("Device Name     state            "
589 			"trans_start\n");
590 		pr_info("%-15s %016lX %016lX\n",
591 			netdev->name,
592 			netdev->state,
593 			dev_trans_start(netdev));
594 	}
595 
596 	/* Print Registers */
597 	dev_info(&adapter->pdev->dev, "Register Dump\n");
598 	pr_info(" Register Name   Value\n");
599 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
600 	     reginfo->name; reginfo++) {
601 		ixgbe_regdump(hw, reginfo);
602 	}
603 
604 	/* Print TX Ring Summary */
605 	if (!netdev || !netif_running(netdev))
606 		return;
607 
608 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
609 	pr_info(" %s     %s              %s        %s\n",
610 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
611 		"leng", "ntw", "timestamp");
612 	for (n = 0; n < adapter->num_tx_queues; n++) {
613 		ring = adapter->tx_ring[n];
614 		ixgbe_print_buffer(ring, n);
615 	}
616 
617 	for (n = 0; n < adapter->num_xdp_queues; n++) {
618 		ring = adapter->xdp_ring[n];
619 		ixgbe_print_buffer(ring, n);
620 	}
621 
622 	/* Print TX Rings */
623 	if (!netif_msg_tx_done(adapter))
624 		goto rx_ring_summary;
625 
626 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
627 
628 	/* Transmit Descriptor Formats
629 	 *
630 	 * 82598 Advanced Transmit Descriptor
631 	 *   +--------------------------------------------------------------+
632 	 * 0 |         Buffer Address [63:0]                                |
633 	 *   +--------------------------------------------------------------+
634 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
635 	 *   +--------------------------------------------------------------+
636 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
637 	 *
638 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
639 	 *   +--------------------------------------------------------------+
640 	 * 0 |                          RSV [63:0]                          |
641 	 *   +--------------------------------------------------------------+
642 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
643 	 *   +--------------------------------------------------------------+
644 	 *   63                       36 35   32 31                         0
645 	 *
646 	 * 82599+ Advanced Transmit Descriptor
647 	 *   +--------------------------------------------------------------+
648 	 * 0 |         Buffer Address [63:0]                                |
649 	 *   +--------------------------------------------------------------+
650 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
651 	 *   +--------------------------------------------------------------+
652 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
653 	 *
654 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
655 	 *   +--------------------------------------------------------------+
656 	 * 0 |                          RSV [63:0]                          |
657 	 *   +--------------------------------------------------------------+
658 	 * 8 |            RSV           |  STA  |           RSV             |
659 	 *   +--------------------------------------------------------------+
660 	 *   63                       36 35   32 31                         0
661 	 */
662 
663 	for (n = 0; n < adapter->num_tx_queues; n++) {
664 		ring = adapter->tx_ring[n];
665 		pr_info("------------------------------------\n");
666 		pr_info("TX QUEUE INDEX = %d\n", ring->queue_index);
667 		pr_info("------------------------------------\n");
668 		pr_info("%s%s    %s              %s        %s          %s\n",
669 			"T [desc]     [address 63:0  ] ",
670 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
671 			"leng", "ntw", "timestamp", "bi->skb");
672 
673 		for (i = 0; ring->desc && (i < ring->count); i++) {
674 			tx_desc = IXGBE_TX_DESC(ring, i);
675 			tx_buffer = &ring->tx_buffer_info[i];
676 			u0 = (struct my_u0 *)tx_desc;
677 			if (dma_unmap_len(tx_buffer, len) > 0) {
678 				const char *ring_desc;
679 
680 				if (i == ring->next_to_use &&
681 				    i == ring->next_to_clean)
682 					ring_desc = " NTC/U";
683 				else if (i == ring->next_to_use)
684 					ring_desc = " NTU";
685 				else if (i == ring->next_to_clean)
686 					ring_desc = " NTC";
687 				else
688 					ring_desc = "";
689 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p%s",
690 					i,
691 					le64_to_cpu((__force __le64)u0->a),
692 					le64_to_cpu((__force __le64)u0->b),
693 					(u64)dma_unmap_addr(tx_buffer, dma),
694 					dma_unmap_len(tx_buffer, len),
695 					tx_buffer->next_to_watch,
696 					(u64)tx_buffer->time_stamp,
697 					tx_buffer->skb,
698 					ring_desc);
699 
700 				if (netif_msg_pktdata(adapter) &&
701 				    tx_buffer->skb)
702 					print_hex_dump(KERN_INFO, "",
703 						DUMP_PREFIX_ADDRESS, 16, 1,
704 						tx_buffer->skb->data,
705 						dma_unmap_len(tx_buffer, len),
706 						true);
707 			}
708 		}
709 	}
710 
711 	/* Print RX Rings Summary */
712 rx_ring_summary:
713 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
714 	pr_info("Queue [NTU] [NTC]\n");
715 	for (n = 0; n < adapter->num_rx_queues; n++) {
716 		rx_ring = adapter->rx_ring[n];
717 		pr_info("%5d %5X %5X\n",
718 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
719 	}
720 
721 	/* Print RX Rings */
722 	if (!netif_msg_rx_status(adapter))
723 		return;
724 
725 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
726 
727 	/* Receive Descriptor Formats
728 	 *
729 	 * 82598 Advanced Receive Descriptor (Read) Format
730 	 *    63                                           1        0
731 	 *    +-----------------------------------------------------+
732 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
733 	 *    +----------------------------------------------+------+
734 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
735 	 *    +-----------------------------------------------------+
736 	 *
737 	 *
738 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
739 	 *
740 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
741 	 *   +------------------------------------------------------+
742 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
743 	 *   | Packet   | IP     |   |          |     | Type | Type |
744 	 *   | Checksum | Ident  |   |          |     |      |      |
745 	 *   +------------------------------------------------------+
746 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
747 	 *   +------------------------------------------------------+
748 	 *   63       48 47    32 31            20 19               0
749 	 *
750 	 * 82599+ Advanced Receive Descriptor (Read) Format
751 	 *    63                                           1        0
752 	 *    +-----------------------------------------------------+
753 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
754 	 *    +----------------------------------------------+------+
755 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
756 	 *    +-----------------------------------------------------+
757 	 *
758 	 *
759 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
760 	 *
761 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
762 	 *   +------------------------------------------------------+
763 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
764 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
765 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
766 	 *   +------------------------------------------------------+
767 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
768 	 *   +------------------------------------------------------+
769 	 *   63       48 47    32 31          20 19                 0
770 	 */
771 
772 	for (n = 0; n < adapter->num_rx_queues; n++) {
773 		rx_ring = adapter->rx_ring[n];
774 		pr_info("------------------------------------\n");
775 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
776 		pr_info("------------------------------------\n");
777 		pr_info("%s%s%s\n",
778 			"R  [desc]      [ PktBuf     A0] ",
779 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
780 			"<-- Adv Rx Read format");
781 		pr_info("%s%s%s\n",
782 			"RWB[desc]      [PcsmIpSHl PtRs] ",
783 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
784 			"<-- Adv Rx Write-Back format");
785 
786 		for (i = 0; i < rx_ring->count; i++) {
787 			const char *ring_desc;
788 
789 			if (i == rx_ring->next_to_use)
790 				ring_desc = " NTU";
791 			else if (i == rx_ring->next_to_clean)
792 				ring_desc = " NTC";
793 			else
794 				ring_desc = "";
795 
796 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
797 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
798 			u0 = (struct my_u0 *)rx_desc;
799 			if (rx_desc->wb.upper.length) {
800 				/* Descriptor Done */
801 				pr_info("RWB[0x%03X]     %016llX %016llX ---------------- %p%s\n",
802 					i,
803 					le64_to_cpu((__force __le64)u0->a),
804 					le64_to_cpu((__force __le64)u0->b),
805 					rx_buffer_info->skb,
806 					ring_desc);
807 			} else {
808 				pr_info("R  [0x%03X]     %016llX %016llX %016llX %p%s\n",
809 					i,
810 					le64_to_cpu((__force __le64)u0->a),
811 					le64_to_cpu((__force __le64)u0->b),
812 					(u64)rx_buffer_info->dma,
813 					rx_buffer_info->skb,
814 					ring_desc);
815 
816 				if (netif_msg_pktdata(adapter) &&
817 				    rx_buffer_info->dma) {
818 					print_hex_dump(KERN_INFO, "",
819 					   DUMP_PREFIX_ADDRESS, 16, 1,
820 					   page_address(rx_buffer_info->page) +
821 						    rx_buffer_info->page_offset,
822 					   ixgbe_rx_bufsz(rx_ring), true);
823 				}
824 			}
825 		}
826 	}
827 }
828 
829 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
830 {
831 	u32 ctrl_ext;
832 
833 	/* Let firmware take over control of h/w */
834 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
835 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
836 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
837 }
838 
839 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
840 {
841 	u32 ctrl_ext;
842 
843 	/* Let firmware know the driver has taken over */
844 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
845 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
846 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
847 }
848 
849 /**
850  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
851  * @adapter: pointer to adapter struct
852  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
853  * @queue: queue to map the corresponding interrupt to
854  * @msix_vector: the vector to map to the corresponding queue
855  *
856  */
857 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
858 			   u8 queue, u8 msix_vector)
859 {
860 	u32 ivar, index;
861 	struct ixgbe_hw *hw = &adapter->hw;
862 	switch (hw->mac.type) {
863 	case ixgbe_mac_82598EB:
864 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
865 		if (direction == -1)
866 			direction = 0;
867 		index = (((direction * 64) + queue) >> 2) & 0x1F;
868 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
869 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
870 		ivar |= (msix_vector << (8 * (queue & 0x3)));
871 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
872 		break;
873 	case ixgbe_mac_82599EB:
874 	case ixgbe_mac_X540:
875 	case ixgbe_mac_X550:
876 	case ixgbe_mac_X550EM_x:
877 	case ixgbe_mac_x550em_a:
878 		if (direction == -1) {
879 			/* other causes */
880 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
881 			index = ((queue & 1) * 8);
882 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
883 			ivar &= ~(0xFF << index);
884 			ivar |= (msix_vector << index);
885 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
886 			break;
887 		} else {
888 			/* tx or rx causes */
889 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
890 			index = ((16 * (queue & 1)) + (8 * direction));
891 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
892 			ivar &= ~(0xFF << index);
893 			ivar |= (msix_vector << index);
894 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
895 			break;
896 		}
897 	default:
898 		break;
899 	}
900 }
901 
902 void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
903 			    u64 qmask)
904 {
905 	u32 mask;
906 
907 	switch (adapter->hw.mac.type) {
908 	case ixgbe_mac_82598EB:
909 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
910 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
911 		break;
912 	case ixgbe_mac_82599EB:
913 	case ixgbe_mac_X540:
914 	case ixgbe_mac_X550:
915 	case ixgbe_mac_X550EM_x:
916 	case ixgbe_mac_x550em_a:
917 		mask = (qmask & 0xFFFFFFFF);
918 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
919 		mask = (qmask >> 32);
920 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
921 		break;
922 	default:
923 		break;
924 	}
925 }
926 
927 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
928 {
929 	struct ixgbe_hw *hw = &adapter->hw;
930 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
931 	int i;
932 	u32 data;
933 
934 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
935 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
936 		return;
937 
938 	switch (hw->mac.type) {
939 	case ixgbe_mac_82598EB:
940 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
941 		break;
942 	default:
943 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
944 	}
945 	hwstats->lxoffrxc += data;
946 
947 	/* refill credits (no tx hang) if we received xoff */
948 	if (!data)
949 		return;
950 
951 	for (i = 0; i < adapter->num_tx_queues; i++)
952 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
953 			  &adapter->tx_ring[i]->state);
954 
955 	for (i = 0; i < adapter->num_xdp_queues; i++)
956 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
957 			  &adapter->xdp_ring[i]->state);
958 }
959 
960 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
961 {
962 	struct ixgbe_hw *hw = &adapter->hw;
963 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
964 	u32 xoff[8] = {0};
965 	u8 tc;
966 	int i;
967 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
968 
969 	if (adapter->ixgbe_ieee_pfc)
970 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
971 
972 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
973 		ixgbe_update_xoff_rx_lfc(adapter);
974 		return;
975 	}
976 
977 	/* update stats for each tc, only valid with PFC enabled */
978 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
979 		u32 pxoffrxc;
980 
981 		switch (hw->mac.type) {
982 		case ixgbe_mac_82598EB:
983 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
984 			break;
985 		default:
986 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
987 		}
988 		hwstats->pxoffrxc[i] += pxoffrxc;
989 		/* Get the TC for given UP */
990 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
991 		xoff[tc] += pxoffrxc;
992 	}
993 
994 	/* disarm tx queues that have received xoff frames */
995 	for (i = 0; i < adapter->num_tx_queues; i++) {
996 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
997 
998 		tc = tx_ring->dcb_tc;
999 		if (xoff[tc])
1000 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1001 	}
1002 
1003 	for (i = 0; i < adapter->num_xdp_queues; i++) {
1004 		struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
1005 
1006 		tc = xdp_ring->dcb_tc;
1007 		if (xoff[tc])
1008 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &xdp_ring->state);
1009 	}
1010 }
1011 
1012 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1013 {
1014 	return ring->stats.packets;
1015 }
1016 
1017 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1018 {
1019 	unsigned int head, tail;
1020 
1021 	head = ring->next_to_clean;
1022 	tail = ring->next_to_use;
1023 
1024 	return ((head <= tail) ? tail : tail + ring->count) - head;
1025 }
1026 
1027 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1028 {
1029 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1030 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1031 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1032 
1033 	clear_check_for_tx_hang(tx_ring);
1034 
1035 	/*
1036 	 * Check for a hung queue, but be thorough. This verifies
1037 	 * that a transmit has been completed since the previous
1038 	 * check AND there is at least one packet pending. The
1039 	 * ARMED bit is set to indicate a potential hang. The
1040 	 * bit is cleared if a pause frame is received to remove
1041 	 * false hang detection due to PFC or 802.3x frames. By
1042 	 * requiring this to fail twice we avoid races with
1043 	 * pfc clearing the ARMED bit and conditions where we
1044 	 * run the check_tx_hang logic with a transmit completion
1045 	 * pending but without time to complete it yet.
1046 	 */
1047 	if (tx_done_old == tx_done && tx_pending)
1048 		/* make sure it is true for two checks in a row */
1049 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1050 					&tx_ring->state);
1051 	/* update completed stats and continue */
1052 	tx_ring->tx_stats.tx_done_old = tx_done;
1053 	/* reset the countdown */
1054 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1055 
1056 	return false;
1057 }
1058 
1059 /**
1060  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1061  * @adapter: driver private struct
1062  **/
1063 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1064 {
1065 
1066 	/* Do the reset outside of interrupt context */
1067 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1068 		set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
1069 		e_warn(drv, "initiating reset due to tx timeout\n");
1070 		ixgbe_service_event_schedule(adapter);
1071 	}
1072 }
1073 
1074 /**
1075  * ixgbe_tx_maxrate - callback to set the maximum per-queue bitrate
1076  * @netdev: network interface device structure
1077  * @queue_index: Tx queue to set
1078  * @maxrate: desired maximum transmit bitrate
1079  **/
1080 static int ixgbe_tx_maxrate(struct net_device *netdev,
1081 			    int queue_index, u32 maxrate)
1082 {
1083 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1084 	struct ixgbe_hw *hw = &adapter->hw;
1085 	u32 bcnrc_val = ixgbe_link_mbps(adapter);
1086 
1087 	if (!maxrate)
1088 		return 0;
1089 
1090 	/* Calculate the rate factor values to set */
1091 	bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1092 	bcnrc_val /= maxrate;
1093 
1094 	/* clear everything but the rate factor */
1095 	bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1096 	IXGBE_RTTBCNRC_RF_DEC_MASK;
1097 
1098 	/* enable the rate scheduler */
1099 	bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1100 
1101 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_index);
1102 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1103 
1104 	return 0;
1105 }
1106 
1107 /**
1108  * ixgbe_update_tx_ring_stats - Update Tx ring specific counters
1109  * @tx_ring: ring to update
1110  * @q_vector: queue vector ring belongs to
1111  * @pkts: number of processed packets
1112  * @bytes: number of processed bytes
1113  */
1114 void ixgbe_update_tx_ring_stats(struct ixgbe_ring *tx_ring,
1115 				struct ixgbe_q_vector *q_vector, u64 pkts,
1116 				u64 bytes)
1117 {
1118 	u64_stats_update_begin(&tx_ring->syncp);
1119 	tx_ring->stats.bytes += bytes;
1120 	tx_ring->stats.packets += pkts;
1121 	u64_stats_update_end(&tx_ring->syncp);
1122 	q_vector->tx.total_bytes += bytes;
1123 	q_vector->tx.total_packets += pkts;
1124 }
1125 
1126 /**
1127  * ixgbe_update_rx_ring_stats - Update Rx ring specific counters
1128  * @rx_ring: ring to update
1129  * @q_vector: queue vector ring belongs to
1130  * @pkts: number of processed packets
1131  * @bytes: number of processed bytes
1132  */
1133 void ixgbe_update_rx_ring_stats(struct ixgbe_ring *rx_ring,
1134 				struct ixgbe_q_vector *q_vector, u64 pkts,
1135 				u64 bytes)
1136 {
1137 	u64_stats_update_begin(&rx_ring->syncp);
1138 	rx_ring->stats.bytes += bytes;
1139 	rx_ring->stats.packets += pkts;
1140 	u64_stats_update_end(&rx_ring->syncp);
1141 	q_vector->rx.total_bytes += bytes;
1142 	q_vector->rx.total_packets += pkts;
1143 }
1144 
1145 /**
1146  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1147  * @q_vector: structure containing interrupt and ring information
1148  * @tx_ring: tx ring to clean
1149  * @napi_budget: Used to determine if we are in netpoll
1150  **/
1151 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1152 			       struct ixgbe_ring *tx_ring, int napi_budget)
1153 {
1154 	struct ixgbe_adapter *adapter = q_vector->adapter;
1155 	struct ixgbe_tx_buffer *tx_buffer;
1156 	union ixgbe_adv_tx_desc *tx_desc;
1157 	unsigned int total_bytes = 0, total_packets = 0, total_ipsec = 0;
1158 	unsigned int budget = q_vector->tx.work_limit;
1159 	unsigned int i = tx_ring->next_to_clean;
1160 	struct netdev_queue *txq;
1161 
1162 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1163 		return true;
1164 
1165 	tx_buffer = &tx_ring->tx_buffer_info[i];
1166 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1167 	i -= tx_ring->count;
1168 
1169 	do {
1170 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1171 
1172 		/* if next_to_watch is not set then there is no work pending */
1173 		if (!eop_desc)
1174 			break;
1175 
1176 		/* prevent any other reads prior to eop_desc */
1177 		smp_rmb();
1178 
1179 		/* if DD is not set pending work has not been completed */
1180 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1181 			break;
1182 
1183 		/* clear next_to_watch to prevent false hangs */
1184 		tx_buffer->next_to_watch = NULL;
1185 
1186 		/* update the statistics for this packet */
1187 		total_bytes += tx_buffer->bytecount;
1188 		total_packets += tx_buffer->gso_segs;
1189 		if (tx_buffer->tx_flags & IXGBE_TX_FLAGS_IPSEC)
1190 			total_ipsec++;
1191 
1192 		/* free the skb */
1193 		if (ring_is_xdp(tx_ring))
1194 			xdp_return_frame(tx_buffer->xdpf);
1195 		else
1196 			napi_consume_skb(tx_buffer->skb, napi_budget);
1197 
1198 		/* unmap skb header data */
1199 		dma_unmap_single(tx_ring->dev,
1200 				 dma_unmap_addr(tx_buffer, dma),
1201 				 dma_unmap_len(tx_buffer, len),
1202 				 DMA_TO_DEVICE);
1203 
1204 		/* clear tx_buffer data */
1205 		dma_unmap_len_set(tx_buffer, len, 0);
1206 
1207 		/* unmap remaining buffers */
1208 		while (tx_desc != eop_desc) {
1209 			tx_buffer++;
1210 			tx_desc++;
1211 			i++;
1212 			if (unlikely(!i)) {
1213 				i -= tx_ring->count;
1214 				tx_buffer = tx_ring->tx_buffer_info;
1215 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1216 			}
1217 
1218 			/* unmap any remaining paged data */
1219 			if (dma_unmap_len(tx_buffer, len)) {
1220 				dma_unmap_page(tx_ring->dev,
1221 					       dma_unmap_addr(tx_buffer, dma),
1222 					       dma_unmap_len(tx_buffer, len),
1223 					       DMA_TO_DEVICE);
1224 				dma_unmap_len_set(tx_buffer, len, 0);
1225 			}
1226 		}
1227 
1228 		/* move us one more past the eop_desc for start of next pkt */
1229 		tx_buffer++;
1230 		tx_desc++;
1231 		i++;
1232 		if (unlikely(!i)) {
1233 			i -= tx_ring->count;
1234 			tx_buffer = tx_ring->tx_buffer_info;
1235 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1236 		}
1237 
1238 		/* issue prefetch for next Tx descriptor */
1239 		prefetch(tx_desc);
1240 
1241 		/* update budget accounting */
1242 		budget--;
1243 	} while (likely(budget));
1244 
1245 	i += tx_ring->count;
1246 	tx_ring->next_to_clean = i;
1247 	ixgbe_update_tx_ring_stats(tx_ring, q_vector, total_packets,
1248 				   total_bytes);
1249 	adapter->tx_ipsec += total_ipsec;
1250 
1251 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1252 		/* schedule immediate reset if we believe we hung */
1253 		struct ixgbe_hw *hw = &adapter->hw;
1254 		e_err(drv, "Detected Tx Unit Hang %s\n"
1255 			"  Tx Queue             <%d>\n"
1256 			"  TDH, TDT             <%x>, <%x>\n"
1257 			"  next_to_use          <%x>\n"
1258 			"  next_to_clean        <%x>\n"
1259 			"tx_buffer_info[next_to_clean]\n"
1260 			"  time_stamp           <%lx>\n"
1261 			"  jiffies              <%lx>\n",
1262 			ring_is_xdp(tx_ring) ? "(XDP)" : "",
1263 			tx_ring->queue_index,
1264 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1265 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1266 			tx_ring->next_to_use, i,
1267 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1268 
1269 		if (!ring_is_xdp(tx_ring))
1270 			netif_stop_subqueue(tx_ring->netdev,
1271 					    tx_ring->queue_index);
1272 
1273 		e_info(probe,
1274 		       "tx hang %d detected on queue %d, resetting adapter\n",
1275 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1276 
1277 		/* schedule immediate reset if we believe we hung */
1278 		ixgbe_tx_timeout_reset(adapter);
1279 
1280 		/* the adapter is about to reset, no point in enabling stuff */
1281 		return true;
1282 	}
1283 
1284 	if (ring_is_xdp(tx_ring))
1285 		return !!budget;
1286 
1287 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1288 	txq = netdev_get_tx_queue(tx_ring->netdev, tx_ring->queue_index);
1289 	if (!__netif_txq_completed_wake(txq, total_packets, total_bytes,
1290 					ixgbe_desc_unused(tx_ring),
1291 					TX_WAKE_THRESHOLD,
1292 					!netif_carrier_ok(tx_ring->netdev) ||
1293 					test_bit(__IXGBE_DOWN, &adapter->state)))
1294 		++tx_ring->tx_stats.restart_queue;
1295 
1296 	return !!budget;
1297 }
1298 
1299 #ifdef CONFIG_IXGBE_DCA
1300 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1301 				struct ixgbe_ring *tx_ring,
1302 				int cpu)
1303 {
1304 	struct ixgbe_hw *hw = &adapter->hw;
1305 	u32 txctrl = 0;
1306 	u16 reg_offset;
1307 
1308 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1309 		txctrl = dca3_get_tag(tx_ring->dev, cpu);
1310 
1311 	switch (hw->mac.type) {
1312 	case ixgbe_mac_82598EB:
1313 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1314 		break;
1315 	case ixgbe_mac_82599EB:
1316 	case ixgbe_mac_X540:
1317 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1318 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1319 		break;
1320 	default:
1321 		/* for unknown hardware do not write register */
1322 		return;
1323 	}
1324 
1325 	/*
1326 	 * We can enable relaxed ordering for reads, but not writes when
1327 	 * DCA is enabled.  This is due to a known issue in some chipsets
1328 	 * which will cause the DCA tag to be cleared.
1329 	 */
1330 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1331 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1332 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1333 
1334 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1335 }
1336 
1337 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1338 				struct ixgbe_ring *rx_ring,
1339 				int cpu)
1340 {
1341 	struct ixgbe_hw *hw = &adapter->hw;
1342 	u32 rxctrl = 0;
1343 	u8 reg_idx = rx_ring->reg_idx;
1344 
1345 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1346 		rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1347 
1348 	switch (hw->mac.type) {
1349 	case ixgbe_mac_82599EB:
1350 	case ixgbe_mac_X540:
1351 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1352 		break;
1353 	default:
1354 		break;
1355 	}
1356 
1357 	/*
1358 	 * We can enable relaxed ordering for reads, but not writes when
1359 	 * DCA is enabled.  This is due to a known issue in some chipsets
1360 	 * which will cause the DCA tag to be cleared.
1361 	 */
1362 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1363 		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1364 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1365 
1366 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1367 }
1368 
1369 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1370 {
1371 	struct ixgbe_adapter *adapter = q_vector->adapter;
1372 	struct ixgbe_ring *ring;
1373 	int cpu = get_cpu();
1374 
1375 	if (q_vector->cpu == cpu)
1376 		goto out_no_update;
1377 
1378 	ixgbe_for_each_ring(ring, q_vector->tx)
1379 		ixgbe_update_tx_dca(adapter, ring, cpu);
1380 
1381 	ixgbe_for_each_ring(ring, q_vector->rx)
1382 		ixgbe_update_rx_dca(adapter, ring, cpu);
1383 
1384 	q_vector->cpu = cpu;
1385 out_no_update:
1386 	put_cpu();
1387 }
1388 
1389 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1390 {
1391 	int i;
1392 
1393 	/* always use CB2 mode, difference is masked in the CB driver */
1394 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1395 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1396 				IXGBE_DCA_CTRL_DCA_MODE_CB2);
1397 	else
1398 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1399 				IXGBE_DCA_CTRL_DCA_DISABLE);
1400 
1401 	for (i = 0; i < adapter->num_q_vectors; i++) {
1402 		adapter->q_vector[i]->cpu = -1;
1403 		ixgbe_update_dca(adapter->q_vector[i]);
1404 	}
1405 }
1406 
1407 static int __ixgbe_notify_dca(struct device *dev, void *data)
1408 {
1409 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1410 	unsigned long event = *(unsigned long *)data;
1411 
1412 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1413 		return 0;
1414 
1415 	switch (event) {
1416 	case DCA_PROVIDER_ADD:
1417 		/* if we're already enabled, don't do it again */
1418 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1419 			break;
1420 		if (dca_add_requester(dev) == 0) {
1421 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1422 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1423 					IXGBE_DCA_CTRL_DCA_MODE_CB2);
1424 			break;
1425 		}
1426 		fallthrough; /* DCA is disabled. */
1427 	case DCA_PROVIDER_REMOVE:
1428 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1429 			dca_remove_requester(dev);
1430 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1431 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1432 					IXGBE_DCA_CTRL_DCA_DISABLE);
1433 		}
1434 		break;
1435 	}
1436 
1437 	return 0;
1438 }
1439 
1440 #endif /* CONFIG_IXGBE_DCA */
1441 
1442 #define IXGBE_RSS_L4_TYPES_MASK \
1443 	((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1444 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1445 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1446 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1447 
1448 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1449 				 union ixgbe_adv_rx_desc *rx_desc,
1450 				 struct sk_buff *skb)
1451 {
1452 	u16 rss_type;
1453 
1454 	if (!(ring->netdev->features & NETIF_F_RXHASH))
1455 		return;
1456 
1457 	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1458 		   IXGBE_RXDADV_RSSTYPE_MASK;
1459 
1460 	if (!rss_type)
1461 		return;
1462 
1463 	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1464 		     (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1465 		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1466 }
1467 
1468 #ifdef IXGBE_FCOE
1469 /**
1470  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1471  * @ring: structure containing ring specific data
1472  * @rx_desc: advanced rx descriptor
1473  *
1474  * Returns : true if it is FCoE pkt
1475  */
1476 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1477 				    union ixgbe_adv_rx_desc *rx_desc)
1478 {
1479 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1480 
1481 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1482 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1483 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1484 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1485 }
1486 
1487 #endif /* IXGBE_FCOE */
1488 /**
1489  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1490  * @ring: structure containing ring specific data
1491  * @rx_desc: current Rx descriptor being processed
1492  * @skb: skb currently being received and modified
1493  **/
1494 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1495 				     union ixgbe_adv_rx_desc *rx_desc,
1496 				     struct sk_buff *skb)
1497 {
1498 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1499 	bool encap_pkt = false;
1500 
1501 	skb_checksum_none_assert(skb);
1502 
1503 	/* Rx csum disabled */
1504 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1505 		return;
1506 
1507 	/* check for VXLAN and Geneve packets */
1508 	if (pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) {
1509 		encap_pkt = true;
1510 		skb->encapsulation = 1;
1511 	}
1512 
1513 	/* if IP and error */
1514 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1515 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1516 		ring->rx_stats.csum_err++;
1517 		return;
1518 	}
1519 
1520 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1521 		return;
1522 
1523 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1524 		/*
1525 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1526 		 * checksum errors.
1527 		 */
1528 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1529 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1530 			return;
1531 
1532 		ring->rx_stats.csum_err++;
1533 		return;
1534 	}
1535 
1536 	/* It must be a TCP or UDP packet with a valid checksum */
1537 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1538 	if (encap_pkt) {
1539 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1540 			return;
1541 
1542 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1543 			skb->ip_summed = CHECKSUM_NONE;
1544 			return;
1545 		}
1546 		/* If we checked the outer header let the stack know */
1547 		skb->csum_level = 1;
1548 	}
1549 }
1550 
1551 static unsigned int ixgbe_rx_offset(struct ixgbe_ring *rx_ring)
1552 {
1553 	return ring_uses_build_skb(rx_ring) ? IXGBE_SKB_PAD : 0;
1554 }
1555 
1556 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1557 				    struct ixgbe_rx_buffer *bi)
1558 {
1559 	struct page *page = bi->page;
1560 	dma_addr_t dma;
1561 
1562 	/* since we are recycling buffers we should seldom need to alloc */
1563 	if (likely(page))
1564 		return true;
1565 
1566 	/* alloc new page for storage */
1567 	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1568 	if (unlikely(!page)) {
1569 		rx_ring->rx_stats.alloc_rx_page_failed++;
1570 		return false;
1571 	}
1572 
1573 	/* map page for use */
1574 	dma = dma_map_page_attrs(rx_ring->dev, page, 0,
1575 				 ixgbe_rx_pg_size(rx_ring),
1576 				 DMA_FROM_DEVICE,
1577 				 IXGBE_RX_DMA_ATTR);
1578 
1579 	/*
1580 	 * if mapping failed free memory back to system since
1581 	 * there isn't much point in holding memory we can't use
1582 	 */
1583 	if (dma_mapping_error(rx_ring->dev, dma)) {
1584 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1585 
1586 		rx_ring->rx_stats.alloc_rx_page_failed++;
1587 		return false;
1588 	}
1589 
1590 	bi->dma = dma;
1591 	bi->page = page;
1592 	bi->page_offset = rx_ring->rx_offset;
1593 	page_ref_add(page, USHRT_MAX - 1);
1594 	bi->pagecnt_bias = USHRT_MAX;
1595 	rx_ring->rx_stats.alloc_rx_page++;
1596 
1597 	return true;
1598 }
1599 
1600 /**
1601  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1602  * @rx_ring: ring to place buffers on
1603  * @cleaned_count: number of buffers to replace
1604  **/
1605 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1606 {
1607 	union ixgbe_adv_rx_desc *rx_desc;
1608 	struct ixgbe_rx_buffer *bi;
1609 	u16 i = rx_ring->next_to_use;
1610 	u16 bufsz;
1611 
1612 	/* nothing to do */
1613 	if (!cleaned_count)
1614 		return;
1615 
1616 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1617 	bi = &rx_ring->rx_buffer_info[i];
1618 	i -= rx_ring->count;
1619 
1620 	bufsz = ixgbe_rx_bufsz(rx_ring);
1621 
1622 	do {
1623 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1624 			break;
1625 
1626 		/* sync the buffer for use by the device */
1627 		dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
1628 						 bi->page_offset, bufsz,
1629 						 DMA_FROM_DEVICE);
1630 
1631 		/*
1632 		 * Refresh the desc even if buffer_addrs didn't change
1633 		 * because each write-back erases this info.
1634 		 */
1635 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1636 
1637 		rx_desc++;
1638 		bi++;
1639 		i++;
1640 		if (unlikely(!i)) {
1641 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1642 			bi = rx_ring->rx_buffer_info;
1643 			i -= rx_ring->count;
1644 		}
1645 
1646 		/* clear the length for the next_to_use descriptor */
1647 		rx_desc->wb.upper.length = 0;
1648 
1649 		cleaned_count--;
1650 	} while (cleaned_count);
1651 
1652 	i += rx_ring->count;
1653 
1654 	if (rx_ring->next_to_use != i) {
1655 		rx_ring->next_to_use = i;
1656 
1657 		/* update next to alloc since we have filled the ring */
1658 		rx_ring->next_to_alloc = i;
1659 
1660 		/* Force memory writes to complete before letting h/w
1661 		 * know there are new descriptors to fetch.  (Only
1662 		 * applicable for weak-ordered memory model archs,
1663 		 * such as IA-64).
1664 		 */
1665 		wmb();
1666 		writel(i, rx_ring->tail);
1667 	}
1668 }
1669 
1670 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1671 				   struct sk_buff *skb)
1672 {
1673 	u16 hdr_len = skb_headlen(skb);
1674 
1675 	/* set gso_size to avoid messing up TCP MSS */
1676 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1677 						 IXGBE_CB(skb)->append_cnt);
1678 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1679 }
1680 
1681 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1682 				   struct sk_buff *skb)
1683 {
1684 	/* if append_cnt is 0 then frame is not RSC */
1685 	if (!IXGBE_CB(skb)->append_cnt)
1686 		return;
1687 
1688 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1689 	rx_ring->rx_stats.rsc_flush++;
1690 
1691 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1692 
1693 	/* gso_size is computed using append_cnt so always clear it last */
1694 	IXGBE_CB(skb)->append_cnt = 0;
1695 }
1696 
1697 /**
1698  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1699  * @rx_ring: rx descriptor ring packet is being transacted on
1700  * @rx_desc: pointer to the EOP Rx descriptor
1701  * @skb: pointer to current skb being populated
1702  *
1703  * This function checks the ring, descriptor, and packet information in
1704  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1705  * other fields within the skb.
1706  **/
1707 void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1708 			      union ixgbe_adv_rx_desc *rx_desc,
1709 			      struct sk_buff *skb)
1710 {
1711 	struct net_device *dev = rx_ring->netdev;
1712 	u32 flags = rx_ring->q_vector->adapter->flags;
1713 
1714 	ixgbe_update_rsc_stats(rx_ring, skb);
1715 
1716 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1717 
1718 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1719 
1720 	if (unlikely(flags & IXGBE_FLAG_RX_HWTSTAMP_ENABLED))
1721 		ixgbe_ptp_rx_hwtstamp(rx_ring, rx_desc, skb);
1722 
1723 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1724 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1725 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1726 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1727 	}
1728 
1729 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_SECP))
1730 		ixgbe_ipsec_rx(rx_ring, rx_desc, skb);
1731 
1732 	/* record Rx queue, or update MACVLAN statistics */
1733 	if (netif_is_ixgbe(dev))
1734 		skb_record_rx_queue(skb, rx_ring->queue_index);
1735 	else
1736 		macvlan_count_rx(netdev_priv(dev), skb->len + ETH_HLEN, true,
1737 				 false);
1738 
1739 	skb->protocol = eth_type_trans(skb, dev);
1740 }
1741 
1742 void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1743 		  struct sk_buff *skb)
1744 {
1745 	napi_gro_receive(&q_vector->napi, skb);
1746 }
1747 
1748 /**
1749  * ixgbe_is_non_eop - process handling of non-EOP buffers
1750  * @rx_ring: Rx ring being processed
1751  * @rx_desc: Rx descriptor for current buffer
1752  * @skb: Current socket buffer containing buffer in progress
1753  *
1754  * This function updates next to clean.  If the buffer is an EOP buffer
1755  * this function exits returning false, otherwise it will place the
1756  * sk_buff in the next buffer to be chained and return true indicating
1757  * that this is in fact a non-EOP buffer.
1758  **/
1759 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1760 			     union ixgbe_adv_rx_desc *rx_desc,
1761 			     struct sk_buff *skb)
1762 {
1763 	u32 ntc = rx_ring->next_to_clean + 1;
1764 
1765 	/* fetch, update, and store next to clean */
1766 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1767 	rx_ring->next_to_clean = ntc;
1768 
1769 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1770 
1771 	/* update RSC append count if present */
1772 	if (ring_is_rsc_enabled(rx_ring)) {
1773 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1774 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1775 
1776 		if (unlikely(rsc_enabled)) {
1777 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1778 
1779 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1780 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1781 
1782 			/* update ntc based on RSC value */
1783 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1784 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1785 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1786 		}
1787 	}
1788 
1789 	/* if we are the last buffer then there is nothing else to do */
1790 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1791 		return false;
1792 
1793 	/* place skb in next buffer to be received */
1794 	rx_ring->rx_buffer_info[ntc].skb = skb;
1795 	rx_ring->rx_stats.non_eop_descs++;
1796 
1797 	return true;
1798 }
1799 
1800 /**
1801  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1802  * @rx_ring: rx descriptor ring packet is being transacted on
1803  * @skb: pointer to current skb being adjusted
1804  *
1805  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1806  * main difference between this version and the original function is that
1807  * this function can make several assumptions about the state of things
1808  * that allow for significant optimizations versus the standard function.
1809  * As a result we can do things like drop a frag and maintain an accurate
1810  * truesize for the skb.
1811  */
1812 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1813 			    struct sk_buff *skb)
1814 {
1815 	skb_frag_t *frag = &skb_shinfo(skb)->frags[0];
1816 	unsigned char *va;
1817 	unsigned int pull_len;
1818 
1819 	/*
1820 	 * it is valid to use page_address instead of kmap since we are
1821 	 * working with pages allocated out of the lomem pool per
1822 	 * alloc_page(GFP_ATOMIC)
1823 	 */
1824 	va = skb_frag_address(frag);
1825 
1826 	/*
1827 	 * we need the header to contain the greater of either ETH_HLEN or
1828 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1829 	 */
1830 	pull_len = eth_get_headlen(skb->dev, va, IXGBE_RX_HDR_SIZE);
1831 
1832 	/* align pull length to size of long to optimize memcpy performance */
1833 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1834 
1835 	/* update all of the pointers */
1836 	skb_frag_size_sub(frag, pull_len);
1837 	skb_frag_off_add(frag, pull_len);
1838 	skb->data_len -= pull_len;
1839 	skb->tail += pull_len;
1840 }
1841 
1842 /**
1843  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1844  * @rx_ring: rx descriptor ring packet is being transacted on
1845  * @skb: pointer to current skb being updated
1846  *
1847  * This function provides a basic DMA sync up for the first fragment of an
1848  * skb.  The reason for doing this is that the first fragment cannot be
1849  * unmapped until we have reached the end of packet descriptor for a buffer
1850  * chain.
1851  */
1852 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1853 				struct sk_buff *skb)
1854 {
1855 	if (ring_uses_build_skb(rx_ring)) {
1856 		unsigned long mask = (unsigned long)ixgbe_rx_pg_size(rx_ring) - 1;
1857 		unsigned long offset = (unsigned long)(skb->data) & mask;
1858 
1859 		dma_sync_single_range_for_cpu(rx_ring->dev,
1860 					      IXGBE_CB(skb)->dma,
1861 					      offset,
1862 					      skb_headlen(skb),
1863 					      DMA_FROM_DEVICE);
1864 	} else {
1865 		skb_frag_t *frag = &skb_shinfo(skb)->frags[0];
1866 
1867 		dma_sync_single_range_for_cpu(rx_ring->dev,
1868 					      IXGBE_CB(skb)->dma,
1869 					      skb_frag_off(frag),
1870 					      skb_frag_size(frag),
1871 					      DMA_FROM_DEVICE);
1872 	}
1873 
1874 	/* If the page was released, just unmap it. */
1875 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1876 		dma_unmap_page_attrs(rx_ring->dev, IXGBE_CB(skb)->dma,
1877 				     ixgbe_rx_pg_size(rx_ring),
1878 				     DMA_FROM_DEVICE,
1879 				     IXGBE_RX_DMA_ATTR);
1880 	}
1881 }
1882 
1883 /**
1884  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1885  * @rx_ring: rx descriptor ring packet is being transacted on
1886  * @rx_desc: pointer to the EOP Rx descriptor
1887  * @skb: pointer to current skb being fixed
1888  *
1889  * Check if the skb is valid in the XDP case it will be an error pointer.
1890  * Return true in this case to abort processing and advance to next
1891  * descriptor.
1892  *
1893  * Check for corrupted packet headers caused by senders on the local L2
1894  * embedded NIC switch not setting up their Tx Descriptors right.  These
1895  * should be very rare.
1896  *
1897  * Also address the case where we are pulling data in on pages only
1898  * and as such no data is present in the skb header.
1899  *
1900  * In addition if skb is not at least 60 bytes we need to pad it so that
1901  * it is large enough to qualify as a valid Ethernet frame.
1902  *
1903  * Returns true if an error was encountered and skb was freed.
1904  **/
1905 bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1906 			   union ixgbe_adv_rx_desc *rx_desc,
1907 			   struct sk_buff *skb)
1908 {
1909 	struct net_device *netdev = rx_ring->netdev;
1910 
1911 	/* XDP packets use error pointer so abort at this point */
1912 	if (IS_ERR(skb))
1913 		return true;
1914 
1915 	/* Verify netdev is present, and that packet does not have any
1916 	 * errors that would be unacceptable to the netdev.
1917 	 */
1918 	if (!netdev ||
1919 	    (unlikely(ixgbe_test_staterr(rx_desc,
1920 					 IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1921 	     !(netdev->features & NETIF_F_RXALL)))) {
1922 		dev_kfree_skb_any(skb);
1923 		return true;
1924 	}
1925 
1926 	/* place header in linear portion of buffer */
1927 	if (!skb_headlen(skb))
1928 		ixgbe_pull_tail(rx_ring, skb);
1929 
1930 #ifdef IXGBE_FCOE
1931 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1932 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1933 		return false;
1934 
1935 #endif
1936 	/* if eth_skb_pad returns an error the skb was freed */
1937 	if (eth_skb_pad(skb))
1938 		return true;
1939 
1940 	return false;
1941 }
1942 
1943 /**
1944  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1945  * @rx_ring: rx descriptor ring to store buffers on
1946  * @old_buff: donor buffer to have page reused
1947  *
1948  * Synchronizes page for reuse by the adapter
1949  **/
1950 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1951 				struct ixgbe_rx_buffer *old_buff)
1952 {
1953 	struct ixgbe_rx_buffer *new_buff;
1954 	u16 nta = rx_ring->next_to_alloc;
1955 
1956 	new_buff = &rx_ring->rx_buffer_info[nta];
1957 
1958 	/* update, and store next to alloc */
1959 	nta++;
1960 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1961 
1962 	/* Transfer page from old buffer to new buffer.
1963 	 * Move each member individually to avoid possible store
1964 	 * forwarding stalls and unnecessary copy of skb.
1965 	 */
1966 	new_buff->dma		= old_buff->dma;
1967 	new_buff->page		= old_buff->page;
1968 	new_buff->page_offset	= old_buff->page_offset;
1969 	new_buff->pagecnt_bias	= old_buff->pagecnt_bias;
1970 }
1971 
1972 static bool ixgbe_can_reuse_rx_page(struct ixgbe_rx_buffer *rx_buffer,
1973 				    int rx_buffer_pgcnt)
1974 {
1975 	unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
1976 	struct page *page = rx_buffer->page;
1977 
1978 	/* avoid re-using remote and pfmemalloc pages */
1979 	if (!dev_page_is_reusable(page))
1980 		return false;
1981 
1982 #if (PAGE_SIZE < 8192)
1983 	/* if we are only owner of page we can reuse it */
1984 	if (unlikely((rx_buffer_pgcnt - pagecnt_bias) > 1))
1985 		return false;
1986 #else
1987 	/* The last offset is a bit aggressive in that we assume the
1988 	 * worst case of FCoE being enabled and using a 3K buffer.
1989 	 * However this should have minimal impact as the 1K extra is
1990 	 * still less than one buffer in size.
1991 	 */
1992 #define IXGBE_LAST_OFFSET \
1993 	(SKB_WITH_OVERHEAD(PAGE_SIZE) - IXGBE_RXBUFFER_3K)
1994 	if (rx_buffer->page_offset > IXGBE_LAST_OFFSET)
1995 		return false;
1996 #endif
1997 
1998 	/* If we have drained the page fragment pool we need to update
1999 	 * the pagecnt_bias and page count so that we fully restock the
2000 	 * number of references the driver holds.
2001 	 */
2002 	if (unlikely(pagecnt_bias == 1)) {
2003 		page_ref_add(page, USHRT_MAX - 1);
2004 		rx_buffer->pagecnt_bias = USHRT_MAX;
2005 	}
2006 
2007 	return true;
2008 }
2009 
2010 /**
2011  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
2012  * @rx_ring: rx descriptor ring to transact packets on
2013  * @rx_buffer: buffer containing page to add
2014  * @skb: sk_buff to place the data into
2015  * @size: size of data in rx_buffer
2016  *
2017  * This function will add the data contained in rx_buffer->page to the skb.
2018  * This is done either through a direct copy if the data in the buffer is
2019  * less than the skb header size, otherwise it will just attach the page as
2020  * a frag to the skb.
2021  *
2022  * The function will then update the page offset if necessary and return
2023  * true if the buffer can be reused by the adapter.
2024  **/
2025 static void ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
2026 			      struct ixgbe_rx_buffer *rx_buffer,
2027 			      struct sk_buff *skb,
2028 			      unsigned int size)
2029 {
2030 #if (PAGE_SIZE < 8192)
2031 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2032 #else
2033 	unsigned int truesize = rx_ring->rx_offset ?
2034 				SKB_DATA_ALIGN(rx_ring->rx_offset + size) :
2035 				SKB_DATA_ALIGN(size);
2036 #endif
2037 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
2038 			rx_buffer->page_offset, size, truesize);
2039 #if (PAGE_SIZE < 8192)
2040 	rx_buffer->page_offset ^= truesize;
2041 #else
2042 	rx_buffer->page_offset += truesize;
2043 #endif
2044 }
2045 
2046 static struct ixgbe_rx_buffer *ixgbe_get_rx_buffer(struct ixgbe_ring *rx_ring,
2047 						   union ixgbe_adv_rx_desc *rx_desc,
2048 						   struct sk_buff **skb,
2049 						   const unsigned int size,
2050 						   int *rx_buffer_pgcnt)
2051 {
2052 	struct ixgbe_rx_buffer *rx_buffer;
2053 
2054 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
2055 	*rx_buffer_pgcnt =
2056 #if (PAGE_SIZE < 8192)
2057 		page_count(rx_buffer->page);
2058 #else
2059 		0;
2060 #endif
2061 	prefetchw(rx_buffer->page);
2062 	*skb = rx_buffer->skb;
2063 
2064 	/* Delay unmapping of the first packet. It carries the header
2065 	 * information, HW may still access the header after the writeback.
2066 	 * Only unmap it when EOP is reached
2067 	 */
2068 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)) {
2069 		if (!*skb)
2070 			goto skip_sync;
2071 	} else {
2072 		if (*skb)
2073 			ixgbe_dma_sync_frag(rx_ring, *skb);
2074 	}
2075 
2076 	/* we are reusing so sync this buffer for CPU use */
2077 	dma_sync_single_range_for_cpu(rx_ring->dev,
2078 				      rx_buffer->dma,
2079 				      rx_buffer->page_offset,
2080 				      size,
2081 				      DMA_FROM_DEVICE);
2082 skip_sync:
2083 	rx_buffer->pagecnt_bias--;
2084 
2085 	return rx_buffer;
2086 }
2087 
2088 static void ixgbe_put_rx_buffer(struct ixgbe_ring *rx_ring,
2089 				struct ixgbe_rx_buffer *rx_buffer,
2090 				struct sk_buff *skb,
2091 				int rx_buffer_pgcnt)
2092 {
2093 	if (ixgbe_can_reuse_rx_page(rx_buffer, rx_buffer_pgcnt)) {
2094 		/* hand second half of page back to the ring */
2095 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2096 	} else {
2097 		if (!IS_ERR(skb) && IXGBE_CB(skb)->dma == rx_buffer->dma) {
2098 			/* the page has been released from the ring */
2099 			IXGBE_CB(skb)->page_released = true;
2100 		} else {
2101 			/* we are not reusing the buffer so unmap it */
2102 			dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
2103 					     ixgbe_rx_pg_size(rx_ring),
2104 					     DMA_FROM_DEVICE,
2105 					     IXGBE_RX_DMA_ATTR);
2106 		}
2107 		__page_frag_cache_drain(rx_buffer->page,
2108 					rx_buffer->pagecnt_bias);
2109 	}
2110 
2111 	/* clear contents of rx_buffer */
2112 	rx_buffer->page = NULL;
2113 	rx_buffer->skb = NULL;
2114 }
2115 
2116 static struct sk_buff *ixgbe_construct_skb(struct ixgbe_ring *rx_ring,
2117 					   struct ixgbe_rx_buffer *rx_buffer,
2118 					   struct xdp_buff *xdp,
2119 					   union ixgbe_adv_rx_desc *rx_desc)
2120 {
2121 	unsigned int size = xdp->data_end - xdp->data;
2122 #if (PAGE_SIZE < 8192)
2123 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2124 #else
2125 	unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end -
2126 					       xdp->data_hard_start);
2127 #endif
2128 	struct sk_buff *skb;
2129 
2130 	/* prefetch first cache line of first page */
2131 	net_prefetch(xdp->data);
2132 
2133 	/* Note, we get here by enabling legacy-rx via:
2134 	 *
2135 	 *    ethtool --set-priv-flags <dev> legacy-rx on
2136 	 *
2137 	 * In this mode, we currently get 0 extra XDP headroom as
2138 	 * opposed to having legacy-rx off, where we process XDP
2139 	 * packets going to stack via ixgbe_build_skb(). The latter
2140 	 * provides us currently with 192 bytes of headroom.
2141 	 *
2142 	 * For ixgbe_construct_skb() mode it means that the
2143 	 * xdp->data_meta will always point to xdp->data, since
2144 	 * the helper cannot expand the head. Should this ever
2145 	 * change in future for legacy-rx mode on, then lets also
2146 	 * add xdp->data_meta handling here.
2147 	 */
2148 
2149 	/* allocate a skb to store the frags */
2150 	skb = napi_alloc_skb(&rx_ring->q_vector->napi, IXGBE_RX_HDR_SIZE);
2151 	if (unlikely(!skb))
2152 		return NULL;
2153 
2154 	if (size > IXGBE_RX_HDR_SIZE) {
2155 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2156 			IXGBE_CB(skb)->dma = rx_buffer->dma;
2157 
2158 		skb_add_rx_frag(skb, 0, rx_buffer->page,
2159 				xdp->data - page_address(rx_buffer->page),
2160 				size, truesize);
2161 #if (PAGE_SIZE < 8192)
2162 		rx_buffer->page_offset ^= truesize;
2163 #else
2164 		rx_buffer->page_offset += truesize;
2165 #endif
2166 	} else {
2167 		memcpy(__skb_put(skb, size),
2168 		       xdp->data, ALIGN(size, sizeof(long)));
2169 		rx_buffer->pagecnt_bias++;
2170 	}
2171 
2172 	return skb;
2173 }
2174 
2175 static struct sk_buff *ixgbe_build_skb(struct ixgbe_ring *rx_ring,
2176 				       struct ixgbe_rx_buffer *rx_buffer,
2177 				       struct xdp_buff *xdp,
2178 				       union ixgbe_adv_rx_desc *rx_desc)
2179 {
2180 	unsigned int metasize = xdp->data - xdp->data_meta;
2181 #if (PAGE_SIZE < 8192)
2182 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2183 #else
2184 	unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
2185 				SKB_DATA_ALIGN(xdp->data_end -
2186 					       xdp->data_hard_start);
2187 #endif
2188 	struct sk_buff *skb;
2189 
2190 	/* Prefetch first cache line of first page. If xdp->data_meta
2191 	 * is unused, this points extactly as xdp->data, otherwise we
2192 	 * likely have a consumer accessing first few bytes of meta
2193 	 * data, and then actual data.
2194 	 */
2195 	net_prefetch(xdp->data_meta);
2196 
2197 	/* build an skb to around the page buffer */
2198 	skb = napi_build_skb(xdp->data_hard_start, truesize);
2199 	if (unlikely(!skb))
2200 		return NULL;
2201 
2202 	/* update pointers within the skb to store the data */
2203 	skb_reserve(skb, xdp->data - xdp->data_hard_start);
2204 	__skb_put(skb, xdp->data_end - xdp->data);
2205 	if (metasize)
2206 		skb_metadata_set(skb, metasize);
2207 
2208 	/* record DMA address if this is the start of a chain of buffers */
2209 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2210 		IXGBE_CB(skb)->dma = rx_buffer->dma;
2211 
2212 	/* update buffer offset */
2213 #if (PAGE_SIZE < 8192)
2214 	rx_buffer->page_offset ^= truesize;
2215 #else
2216 	rx_buffer->page_offset += truesize;
2217 #endif
2218 
2219 	return skb;
2220 }
2221 
2222 static struct sk_buff *ixgbe_run_xdp(struct ixgbe_adapter *adapter,
2223 				     struct ixgbe_ring *rx_ring,
2224 				     struct xdp_buff *xdp)
2225 {
2226 	int err, result = IXGBE_XDP_PASS;
2227 	struct bpf_prog *xdp_prog;
2228 	struct ixgbe_ring *ring;
2229 	struct xdp_frame *xdpf;
2230 	u32 act;
2231 
2232 	xdp_prog = READ_ONCE(rx_ring->xdp_prog);
2233 
2234 	if (!xdp_prog)
2235 		goto xdp_out;
2236 
2237 	prefetchw(xdp->data_hard_start); /* xdp_frame write */
2238 
2239 	act = bpf_prog_run_xdp(xdp_prog, xdp);
2240 	switch (act) {
2241 	case XDP_PASS:
2242 		break;
2243 	case XDP_TX:
2244 		xdpf = xdp_convert_buff_to_frame(xdp);
2245 		if (unlikely(!xdpf))
2246 			goto out_failure;
2247 		ring = ixgbe_determine_xdp_ring(adapter);
2248 		if (static_branch_unlikely(&ixgbe_xdp_locking_key))
2249 			spin_lock(&ring->tx_lock);
2250 		result = ixgbe_xmit_xdp_ring(ring, xdpf);
2251 		if (static_branch_unlikely(&ixgbe_xdp_locking_key))
2252 			spin_unlock(&ring->tx_lock);
2253 		if (result == IXGBE_XDP_CONSUMED)
2254 			goto out_failure;
2255 		break;
2256 	case XDP_REDIRECT:
2257 		err = xdp_do_redirect(adapter->netdev, xdp, xdp_prog);
2258 		if (err)
2259 			goto out_failure;
2260 		result = IXGBE_XDP_REDIR;
2261 		break;
2262 	default:
2263 		bpf_warn_invalid_xdp_action(rx_ring->netdev, xdp_prog, act);
2264 		fallthrough;
2265 	case XDP_ABORTED:
2266 out_failure:
2267 		trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
2268 		fallthrough; /* handle aborts by dropping packet */
2269 	case XDP_DROP:
2270 		result = IXGBE_XDP_CONSUMED;
2271 		break;
2272 	}
2273 xdp_out:
2274 	return ERR_PTR(-result);
2275 }
2276 
2277 static unsigned int ixgbe_rx_frame_truesize(struct ixgbe_ring *rx_ring,
2278 					    unsigned int size)
2279 {
2280 	unsigned int truesize;
2281 
2282 #if (PAGE_SIZE < 8192)
2283 	truesize = ixgbe_rx_pg_size(rx_ring) / 2; /* Must be power-of-2 */
2284 #else
2285 	truesize = rx_ring->rx_offset ?
2286 		SKB_DATA_ALIGN(rx_ring->rx_offset + size) +
2287 		SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) :
2288 		SKB_DATA_ALIGN(size);
2289 #endif
2290 	return truesize;
2291 }
2292 
2293 static void ixgbe_rx_buffer_flip(struct ixgbe_ring *rx_ring,
2294 				 struct ixgbe_rx_buffer *rx_buffer,
2295 				 unsigned int size)
2296 {
2297 	unsigned int truesize = ixgbe_rx_frame_truesize(rx_ring, size);
2298 #if (PAGE_SIZE < 8192)
2299 	rx_buffer->page_offset ^= truesize;
2300 #else
2301 	rx_buffer->page_offset += truesize;
2302 #endif
2303 }
2304 
2305 /**
2306  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2307  * @q_vector: structure containing interrupt and ring information
2308  * @rx_ring: rx descriptor ring to transact packets on
2309  * @budget: Total limit on number of packets to process
2310  *
2311  * This function provides a "bounce buffer" approach to Rx interrupt
2312  * processing.  The advantage to this is that on systems that have
2313  * expensive overhead for IOMMU access this provides a means of avoiding
2314  * it by maintaining the mapping of the page to the syste.
2315  *
2316  * Returns amount of work completed
2317  **/
2318 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2319 			       struct ixgbe_ring *rx_ring,
2320 			       const int budget)
2321 {
2322 	unsigned int total_rx_bytes = 0, total_rx_packets = 0, frame_sz = 0;
2323 	struct ixgbe_adapter *adapter = q_vector->adapter;
2324 #ifdef IXGBE_FCOE
2325 	int ddp_bytes;
2326 	unsigned int mss = 0;
2327 #endif /* IXGBE_FCOE */
2328 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2329 	unsigned int offset = rx_ring->rx_offset;
2330 	unsigned int xdp_xmit = 0;
2331 	struct xdp_buff xdp;
2332 
2333 	/* Frame size depend on rx_ring setup when PAGE_SIZE=4K */
2334 #if (PAGE_SIZE < 8192)
2335 	frame_sz = ixgbe_rx_frame_truesize(rx_ring, 0);
2336 #endif
2337 	xdp_init_buff(&xdp, frame_sz, &rx_ring->xdp_rxq);
2338 
2339 	while (likely(total_rx_packets < budget)) {
2340 		union ixgbe_adv_rx_desc *rx_desc;
2341 		struct ixgbe_rx_buffer *rx_buffer;
2342 		struct sk_buff *skb;
2343 		int rx_buffer_pgcnt;
2344 		unsigned int size;
2345 
2346 		/* return some buffers to hardware, one at a time is too slow */
2347 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2348 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2349 			cleaned_count = 0;
2350 		}
2351 
2352 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2353 		size = le16_to_cpu(rx_desc->wb.upper.length);
2354 		if (!size)
2355 			break;
2356 
2357 		/* This memory barrier is needed to keep us from reading
2358 		 * any other fields out of the rx_desc until we know the
2359 		 * descriptor has been written back
2360 		 */
2361 		dma_rmb();
2362 
2363 		rx_buffer = ixgbe_get_rx_buffer(rx_ring, rx_desc, &skb, size, &rx_buffer_pgcnt);
2364 
2365 		/* retrieve a buffer from the ring */
2366 		if (!skb) {
2367 			unsigned char *hard_start;
2368 
2369 			hard_start = page_address(rx_buffer->page) +
2370 				     rx_buffer->page_offset - offset;
2371 			xdp_prepare_buff(&xdp, hard_start, offset, size, true);
2372 			xdp_buff_clear_frags_flag(&xdp);
2373 #if (PAGE_SIZE > 4096)
2374 			/* At larger PAGE_SIZE, frame_sz depend on len size */
2375 			xdp.frame_sz = ixgbe_rx_frame_truesize(rx_ring, size);
2376 #endif
2377 			skb = ixgbe_run_xdp(adapter, rx_ring, &xdp);
2378 		}
2379 
2380 		if (IS_ERR(skb)) {
2381 			unsigned int xdp_res = -PTR_ERR(skb);
2382 
2383 			if (xdp_res & (IXGBE_XDP_TX | IXGBE_XDP_REDIR)) {
2384 				xdp_xmit |= xdp_res;
2385 				ixgbe_rx_buffer_flip(rx_ring, rx_buffer, size);
2386 			} else {
2387 				rx_buffer->pagecnt_bias++;
2388 			}
2389 			total_rx_packets++;
2390 			total_rx_bytes += size;
2391 		} else if (skb) {
2392 			ixgbe_add_rx_frag(rx_ring, rx_buffer, skb, size);
2393 		} else if (ring_uses_build_skb(rx_ring)) {
2394 			skb = ixgbe_build_skb(rx_ring, rx_buffer,
2395 					      &xdp, rx_desc);
2396 		} else {
2397 			skb = ixgbe_construct_skb(rx_ring, rx_buffer,
2398 						  &xdp, rx_desc);
2399 		}
2400 
2401 		/* exit if we failed to retrieve a buffer */
2402 		if (!skb) {
2403 			rx_ring->rx_stats.alloc_rx_buff_failed++;
2404 			rx_buffer->pagecnt_bias++;
2405 			break;
2406 		}
2407 
2408 		ixgbe_put_rx_buffer(rx_ring, rx_buffer, skb, rx_buffer_pgcnt);
2409 		cleaned_count++;
2410 
2411 		/* place incomplete frames back on ring for completion */
2412 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2413 			continue;
2414 
2415 		/* verify the packet layout is correct */
2416 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2417 			continue;
2418 
2419 		/* probably a little skewed due to removing CRC */
2420 		total_rx_bytes += skb->len;
2421 
2422 		/* populate checksum, timestamp, VLAN, and protocol */
2423 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2424 
2425 #ifdef IXGBE_FCOE
2426 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2427 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2428 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2429 			/* include DDPed FCoE data */
2430 			if (ddp_bytes > 0) {
2431 				if (!mss) {
2432 					mss = rx_ring->netdev->mtu -
2433 						sizeof(struct fcoe_hdr) -
2434 						sizeof(struct fc_frame_header) -
2435 						sizeof(struct fcoe_crc_eof);
2436 					if (mss > 512)
2437 						mss &= ~511;
2438 				}
2439 				total_rx_bytes += ddp_bytes;
2440 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2441 								 mss);
2442 			}
2443 			if (!ddp_bytes) {
2444 				dev_kfree_skb_any(skb);
2445 				continue;
2446 			}
2447 		}
2448 
2449 #endif /* IXGBE_FCOE */
2450 		ixgbe_rx_skb(q_vector, skb);
2451 
2452 		/* update budget accounting */
2453 		total_rx_packets++;
2454 	}
2455 
2456 	if (xdp_xmit & IXGBE_XDP_REDIR)
2457 		xdp_do_flush();
2458 
2459 	if (xdp_xmit & IXGBE_XDP_TX) {
2460 		struct ixgbe_ring *ring = ixgbe_determine_xdp_ring(adapter);
2461 
2462 		ixgbe_xdp_ring_update_tail_locked(ring);
2463 	}
2464 
2465 	ixgbe_update_rx_ring_stats(rx_ring, q_vector, total_rx_packets,
2466 				   total_rx_bytes);
2467 
2468 	return total_rx_packets;
2469 }
2470 
2471 /**
2472  * ixgbe_configure_msix - Configure MSI-X hardware
2473  * @adapter: board private structure
2474  *
2475  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2476  * interrupts.
2477  **/
2478 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2479 {
2480 	struct ixgbe_q_vector *q_vector;
2481 	int v_idx;
2482 	u32 mask;
2483 
2484 	/* Populate MSIX to EITR Select */
2485 	if (adapter->num_vfs > 32) {
2486 		u32 eitrsel = BIT(adapter->num_vfs - 32) - 1;
2487 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2488 	}
2489 
2490 	/*
2491 	 * Populate the IVAR table and set the ITR values to the
2492 	 * corresponding register.
2493 	 */
2494 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2495 		struct ixgbe_ring *ring;
2496 		q_vector = adapter->q_vector[v_idx];
2497 
2498 		ixgbe_for_each_ring(ring, q_vector->rx)
2499 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2500 
2501 		ixgbe_for_each_ring(ring, q_vector->tx)
2502 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2503 
2504 		ixgbe_write_eitr(q_vector);
2505 	}
2506 
2507 	switch (adapter->hw.mac.type) {
2508 	case ixgbe_mac_82598EB:
2509 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2510 			       v_idx);
2511 		break;
2512 	case ixgbe_mac_82599EB:
2513 	case ixgbe_mac_X540:
2514 	case ixgbe_mac_X550:
2515 	case ixgbe_mac_X550EM_x:
2516 	case ixgbe_mac_x550em_a:
2517 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2518 		break;
2519 	default:
2520 		break;
2521 	}
2522 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2523 
2524 	/* set up to autoclear timer, and the vectors */
2525 	mask = IXGBE_EIMS_ENABLE_MASK;
2526 	mask &= ~(IXGBE_EIMS_OTHER |
2527 		  IXGBE_EIMS_MAILBOX |
2528 		  IXGBE_EIMS_LSC);
2529 
2530 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2531 }
2532 
2533 /**
2534  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2535  * @q_vector: structure containing interrupt and ring information
2536  * @ring_container: structure containing ring performance data
2537  *
2538  *      Stores a new ITR value based on packets and byte
2539  *      counts during the last interrupt.  The advantage of per interrupt
2540  *      computation is faster updates and more accurate ITR for the current
2541  *      traffic pattern.  Constants in this function were computed
2542  *      based on theoretical maximum wire speed and thresholds were set based
2543  *      on testing data as well as attempting to minimize response time
2544  *      while increasing bulk throughput.
2545  **/
2546 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2547 			     struct ixgbe_ring_container *ring_container)
2548 {
2549 	unsigned int itr = IXGBE_ITR_ADAPTIVE_MIN_USECS |
2550 			   IXGBE_ITR_ADAPTIVE_LATENCY;
2551 	unsigned int avg_wire_size, packets, bytes;
2552 	unsigned long next_update = jiffies;
2553 
2554 	/* If we don't have any rings just leave ourselves set for maximum
2555 	 * possible latency so we take ourselves out of the equation.
2556 	 */
2557 	if (!ring_container->ring)
2558 		return;
2559 
2560 	/* If we didn't update within up to 1 - 2 jiffies we can assume
2561 	 * that either packets are coming in so slow there hasn't been
2562 	 * any work, or that there is so much work that NAPI is dealing
2563 	 * with interrupt moderation and we don't need to do anything.
2564 	 */
2565 	if (time_after(next_update, ring_container->next_update))
2566 		goto clear_counts;
2567 
2568 	packets = ring_container->total_packets;
2569 
2570 	/* We have no packets to actually measure against. This means
2571 	 * either one of the other queues on this vector is active or
2572 	 * we are a Tx queue doing TSO with too high of an interrupt rate.
2573 	 *
2574 	 * When this occurs just tick up our delay by the minimum value
2575 	 * and hope that this extra delay will prevent us from being called
2576 	 * without any work on our queue.
2577 	 */
2578 	if (!packets) {
2579 		itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2580 		if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2581 			itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2582 		itr += ring_container->itr & IXGBE_ITR_ADAPTIVE_LATENCY;
2583 		goto clear_counts;
2584 	}
2585 
2586 	bytes = ring_container->total_bytes;
2587 
2588 	/* If packets are less than 4 or bytes are less than 9000 assume
2589 	 * insufficient data to use bulk rate limiting approach. We are
2590 	 * likely latency driven.
2591 	 */
2592 	if (packets < 4 && bytes < 9000) {
2593 		itr = IXGBE_ITR_ADAPTIVE_LATENCY;
2594 		goto adjust_by_size;
2595 	}
2596 
2597 	/* Between 4 and 48 we can assume that our current interrupt delay
2598 	 * is only slightly too low. As such we should increase it by a small
2599 	 * fixed amount.
2600 	 */
2601 	if (packets < 48) {
2602 		itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2603 		if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2604 			itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2605 		goto clear_counts;
2606 	}
2607 
2608 	/* Between 48 and 96 is our "goldilocks" zone where we are working
2609 	 * out "just right". Just report that our current ITR is good for us.
2610 	 */
2611 	if (packets < 96) {
2612 		itr = q_vector->itr >> 2;
2613 		goto clear_counts;
2614 	}
2615 
2616 	/* If packet count is 96 or greater we are likely looking at a slight
2617 	 * overrun of the delay we want. Try halving our delay to see if that
2618 	 * will cut the number of packets in half per interrupt.
2619 	 */
2620 	if (packets < 256) {
2621 		itr = q_vector->itr >> 3;
2622 		if (itr < IXGBE_ITR_ADAPTIVE_MIN_USECS)
2623 			itr = IXGBE_ITR_ADAPTIVE_MIN_USECS;
2624 		goto clear_counts;
2625 	}
2626 
2627 	/* The paths below assume we are dealing with a bulk ITR since number
2628 	 * of packets is 256 or greater. We are just going to have to compute
2629 	 * a value and try to bring the count under control, though for smaller
2630 	 * packet sizes there isn't much we can do as NAPI polling will likely
2631 	 * be kicking in sooner rather than later.
2632 	 */
2633 	itr = IXGBE_ITR_ADAPTIVE_BULK;
2634 
2635 adjust_by_size:
2636 	/* If packet counts are 256 or greater we can assume we have a gross
2637 	 * overestimation of what the rate should be. Instead of trying to fine
2638 	 * tune it just use the formula below to try and dial in an exact value
2639 	 * give the current packet size of the frame.
2640 	 */
2641 	avg_wire_size = bytes / packets;
2642 
2643 	/* The following is a crude approximation of:
2644 	 *  wmem_default / (size + overhead) = desired_pkts_per_int
2645 	 *  rate / bits_per_byte / (size + ethernet overhead) = pkt_rate
2646 	 *  (desired_pkt_rate / pkt_rate) * usecs_per_sec = ITR value
2647 	 *
2648 	 * Assuming wmem_default is 212992 and overhead is 640 bytes per
2649 	 * packet, (256 skb, 64 headroom, 320 shared info), we can reduce the
2650 	 * formula down to
2651 	 *
2652 	 *  (170 * (size + 24)) / (size + 640) = ITR
2653 	 *
2654 	 * We first do some math on the packet size and then finally bitshift
2655 	 * by 8 after rounding up. We also have to account for PCIe link speed
2656 	 * difference as ITR scales based on this.
2657 	 */
2658 	if (avg_wire_size <= 60) {
2659 		/* Start at 50k ints/sec */
2660 		avg_wire_size = 5120;
2661 	} else if (avg_wire_size <= 316) {
2662 		/* 50K ints/sec to 16K ints/sec */
2663 		avg_wire_size *= 40;
2664 		avg_wire_size += 2720;
2665 	} else if (avg_wire_size <= 1084) {
2666 		/* 16K ints/sec to 9.2K ints/sec */
2667 		avg_wire_size *= 15;
2668 		avg_wire_size += 11452;
2669 	} else if (avg_wire_size < 1968) {
2670 		/* 9.2K ints/sec to 8K ints/sec */
2671 		avg_wire_size *= 5;
2672 		avg_wire_size += 22420;
2673 	} else {
2674 		/* plateau at a limit of 8K ints/sec */
2675 		avg_wire_size = 32256;
2676 	}
2677 
2678 	/* If we are in low latency mode half our delay which doubles the rate
2679 	 * to somewhere between 100K to 16K ints/sec
2680 	 */
2681 	if (itr & IXGBE_ITR_ADAPTIVE_LATENCY)
2682 		avg_wire_size >>= 1;
2683 
2684 	/* Resultant value is 256 times larger than it needs to be. This
2685 	 * gives us room to adjust the value as needed to either increase
2686 	 * or decrease the value based on link speeds of 10G, 2.5G, 1G, etc.
2687 	 *
2688 	 * Use addition as we have already recorded the new latency flag
2689 	 * for the ITR value.
2690 	 */
2691 	switch (q_vector->adapter->link_speed) {
2692 	case IXGBE_LINK_SPEED_10GB_FULL:
2693 	case IXGBE_LINK_SPEED_100_FULL:
2694 	default:
2695 		itr += DIV_ROUND_UP(avg_wire_size,
2696 				    IXGBE_ITR_ADAPTIVE_MIN_INC * 256) *
2697 		       IXGBE_ITR_ADAPTIVE_MIN_INC;
2698 		break;
2699 	case IXGBE_LINK_SPEED_2_5GB_FULL:
2700 	case IXGBE_LINK_SPEED_1GB_FULL:
2701 	case IXGBE_LINK_SPEED_10_FULL:
2702 		if (avg_wire_size > 8064)
2703 			avg_wire_size = 8064;
2704 		itr += DIV_ROUND_UP(avg_wire_size,
2705 				    IXGBE_ITR_ADAPTIVE_MIN_INC * 64) *
2706 		       IXGBE_ITR_ADAPTIVE_MIN_INC;
2707 		break;
2708 	}
2709 
2710 clear_counts:
2711 	/* write back value */
2712 	ring_container->itr = itr;
2713 
2714 	/* next update should occur within next jiffy */
2715 	ring_container->next_update = next_update + 1;
2716 
2717 	ring_container->total_bytes = 0;
2718 	ring_container->total_packets = 0;
2719 }
2720 
2721 /**
2722  * ixgbe_write_eitr - write EITR register in hardware specific way
2723  * @q_vector: structure containing interrupt and ring information
2724  *
2725  * This function is made to be called by ethtool and by the driver
2726  * when it needs to update EITR registers at runtime.  Hardware
2727  * specific quirks/differences are taken care of here.
2728  */
2729 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2730 {
2731 	struct ixgbe_adapter *adapter = q_vector->adapter;
2732 	struct ixgbe_hw *hw = &adapter->hw;
2733 	int v_idx = q_vector->v_idx;
2734 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2735 
2736 	switch (adapter->hw.mac.type) {
2737 	case ixgbe_mac_82598EB:
2738 		/* must write high and low 16 bits to reset counter */
2739 		itr_reg |= (itr_reg << 16);
2740 		break;
2741 	case ixgbe_mac_82599EB:
2742 	case ixgbe_mac_X540:
2743 	case ixgbe_mac_X550:
2744 	case ixgbe_mac_X550EM_x:
2745 	case ixgbe_mac_x550em_a:
2746 		/*
2747 		 * set the WDIS bit to not clear the timer bits and cause an
2748 		 * immediate assertion of the interrupt
2749 		 */
2750 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2751 		break;
2752 	default:
2753 		break;
2754 	}
2755 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2756 }
2757 
2758 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2759 {
2760 	u32 new_itr;
2761 
2762 	ixgbe_update_itr(q_vector, &q_vector->tx);
2763 	ixgbe_update_itr(q_vector, &q_vector->rx);
2764 
2765 	/* use the smallest value of new ITR delay calculations */
2766 	new_itr = min(q_vector->rx.itr, q_vector->tx.itr);
2767 
2768 	/* Clear latency flag if set, shift into correct position */
2769 	new_itr &= ~IXGBE_ITR_ADAPTIVE_LATENCY;
2770 	new_itr <<= 2;
2771 
2772 	if (new_itr != q_vector->itr) {
2773 		/* save the algorithm value here */
2774 		q_vector->itr = new_itr;
2775 
2776 		ixgbe_write_eitr(q_vector);
2777 	}
2778 }
2779 
2780 /**
2781  * ixgbe_check_overtemp_subtask - check for over temperature
2782  * @adapter: pointer to adapter
2783  **/
2784 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2785 {
2786 	struct ixgbe_hw *hw = &adapter->hw;
2787 	u32 eicr = adapter->interrupt_event;
2788 
2789 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2790 		return;
2791 
2792 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2793 		return;
2794 
2795 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2796 
2797 	switch (hw->device_id) {
2798 	case IXGBE_DEV_ID_82599_T3_LOM:
2799 		/*
2800 		 * Since the warning interrupt is for both ports
2801 		 * we don't have to check if:
2802 		 *  - This interrupt wasn't for our port.
2803 		 *  - We may have missed the interrupt so always have to
2804 		 *    check if we  got a LSC
2805 		 */
2806 		if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2807 		    !(eicr & IXGBE_EICR_LSC))
2808 			return;
2809 
2810 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2811 			u32 speed;
2812 			bool link_up = false;
2813 
2814 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2815 
2816 			if (link_up)
2817 				return;
2818 		}
2819 
2820 		/* Check if this is not due to overtemp */
2821 		if (!hw->phy.ops.check_overtemp(hw))
2822 			return;
2823 
2824 		break;
2825 	case IXGBE_DEV_ID_X550EM_A_1G_T:
2826 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
2827 		if (!hw->phy.ops.check_overtemp(hw))
2828 			return;
2829 		break;
2830 	default:
2831 		if (adapter->hw.mac.type >= ixgbe_mac_X540)
2832 			return;
2833 		if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
2834 			return;
2835 		break;
2836 	}
2837 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2838 
2839 	adapter->interrupt_event = 0;
2840 }
2841 
2842 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2843 {
2844 	struct ixgbe_hw *hw = &adapter->hw;
2845 
2846 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2847 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2848 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2849 		/* write to clear the interrupt */
2850 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2851 	}
2852 }
2853 
2854 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2855 {
2856 	struct ixgbe_hw *hw = &adapter->hw;
2857 
2858 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2859 		return;
2860 
2861 	switch (adapter->hw.mac.type) {
2862 	case ixgbe_mac_82599EB:
2863 		/*
2864 		 * Need to check link state so complete overtemp check
2865 		 * on service task
2866 		 */
2867 		if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
2868 		     (eicr & IXGBE_EICR_LSC)) &&
2869 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2870 			adapter->interrupt_event = eicr;
2871 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2872 			ixgbe_service_event_schedule(adapter);
2873 			return;
2874 		}
2875 		return;
2876 	case ixgbe_mac_x550em_a:
2877 		if (eicr & IXGBE_EICR_GPI_SDP0_X550EM_a) {
2878 			adapter->interrupt_event = eicr;
2879 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2880 			ixgbe_service_event_schedule(adapter);
2881 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC,
2882 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2883 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICR,
2884 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2885 		}
2886 		return;
2887 	case ixgbe_mac_X550:
2888 	case ixgbe_mac_X540:
2889 		if (!(eicr & IXGBE_EICR_TS))
2890 			return;
2891 		break;
2892 	default:
2893 		return;
2894 	}
2895 
2896 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2897 }
2898 
2899 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
2900 {
2901 	switch (hw->mac.type) {
2902 	case ixgbe_mac_82598EB:
2903 		if (hw->phy.type == ixgbe_phy_nl)
2904 			return true;
2905 		return false;
2906 	case ixgbe_mac_82599EB:
2907 	case ixgbe_mac_X550EM_x:
2908 	case ixgbe_mac_x550em_a:
2909 		switch (hw->mac.ops.get_media_type(hw)) {
2910 		case ixgbe_media_type_fiber:
2911 		case ixgbe_media_type_fiber_qsfp:
2912 			return true;
2913 		default:
2914 			return false;
2915 		}
2916 	default:
2917 		return false;
2918 	}
2919 }
2920 
2921 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2922 {
2923 	struct ixgbe_hw *hw = &adapter->hw;
2924 	u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
2925 
2926 	if (!ixgbe_is_sfp(hw))
2927 		return;
2928 
2929 	/* Later MAC's use different SDP */
2930 	if (hw->mac.type >= ixgbe_mac_X540)
2931 		eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2932 
2933 	if (eicr & eicr_mask) {
2934 		/* Clear the interrupt */
2935 		IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2936 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2937 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2938 			adapter->sfp_poll_time = 0;
2939 			ixgbe_service_event_schedule(adapter);
2940 		}
2941 	}
2942 
2943 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
2944 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2945 		/* Clear the interrupt */
2946 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2947 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2948 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2949 			ixgbe_service_event_schedule(adapter);
2950 		}
2951 	}
2952 }
2953 
2954 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2955 {
2956 	struct ixgbe_hw *hw = &adapter->hw;
2957 
2958 	adapter->lsc_int++;
2959 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2960 	adapter->link_check_timeout = jiffies;
2961 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2962 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2963 		IXGBE_WRITE_FLUSH(hw);
2964 		ixgbe_service_event_schedule(adapter);
2965 	}
2966 }
2967 
2968 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2969 					   u64 qmask)
2970 {
2971 	struct ixgbe_hw *hw = &adapter->hw;
2972 	u32 mask;
2973 
2974 	switch (hw->mac.type) {
2975 	case ixgbe_mac_82598EB:
2976 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2977 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2978 		break;
2979 	case ixgbe_mac_82599EB:
2980 	case ixgbe_mac_X540:
2981 	case ixgbe_mac_X550:
2982 	case ixgbe_mac_X550EM_x:
2983 	case ixgbe_mac_x550em_a:
2984 		mask = (qmask & 0xFFFFFFFF);
2985 		if (mask)
2986 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2987 		mask = (qmask >> 32);
2988 		if (mask)
2989 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2990 		break;
2991 	default:
2992 		break;
2993 	}
2994 	/* skip the flush */
2995 }
2996 
2997 /**
2998  * ixgbe_irq_enable - Enable default interrupt generation settings
2999  * @adapter: board private structure
3000  * @queues: enable irqs for queues
3001  * @flush: flush register write
3002  **/
3003 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
3004 				    bool flush)
3005 {
3006 	struct ixgbe_hw *hw = &adapter->hw;
3007 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
3008 
3009 	/* don't reenable LSC while waiting for link */
3010 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
3011 		mask &= ~IXGBE_EIMS_LSC;
3012 
3013 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
3014 		switch (adapter->hw.mac.type) {
3015 		case ixgbe_mac_82599EB:
3016 			mask |= IXGBE_EIMS_GPI_SDP0(hw);
3017 			break;
3018 		case ixgbe_mac_X540:
3019 		case ixgbe_mac_X550:
3020 		case ixgbe_mac_X550EM_x:
3021 		case ixgbe_mac_x550em_a:
3022 			mask |= IXGBE_EIMS_TS;
3023 			break;
3024 		default:
3025 			break;
3026 		}
3027 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
3028 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
3029 	switch (adapter->hw.mac.type) {
3030 	case ixgbe_mac_82599EB:
3031 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
3032 		mask |= IXGBE_EIMS_GPI_SDP2(hw);
3033 		fallthrough;
3034 	case ixgbe_mac_X540:
3035 	case ixgbe_mac_X550:
3036 	case ixgbe_mac_X550EM_x:
3037 	case ixgbe_mac_x550em_a:
3038 		if (adapter->hw.device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
3039 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
3040 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP_N)
3041 			mask |= IXGBE_EIMS_GPI_SDP0(&adapter->hw);
3042 		if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
3043 			mask |= IXGBE_EICR_GPI_SDP0_X540;
3044 		mask |= IXGBE_EIMS_ECC;
3045 		mask |= IXGBE_EIMS_MAILBOX;
3046 		break;
3047 	default:
3048 		break;
3049 	}
3050 
3051 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
3052 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
3053 		mask |= IXGBE_EIMS_FLOW_DIR;
3054 
3055 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
3056 	if (queues)
3057 		ixgbe_irq_enable_queues(adapter, ~0);
3058 	if (flush)
3059 		IXGBE_WRITE_FLUSH(&adapter->hw);
3060 }
3061 
3062 static irqreturn_t ixgbe_msix_other(int irq, void *data)
3063 {
3064 	struct ixgbe_adapter *adapter = data;
3065 	struct ixgbe_hw *hw = &adapter->hw;
3066 	u32 eicr;
3067 
3068 	/*
3069 	 * Workaround for Silicon errata.  Use clear-by-write instead
3070 	 * of clear-by-read.  Reading with EICS will return the
3071 	 * interrupt causes without clearing, which later be done
3072 	 * with the write to EICR.
3073 	 */
3074 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
3075 
3076 	/* The lower 16bits of the EICR register are for the queue interrupts
3077 	 * which should be masked here in order to not accidentally clear them if
3078 	 * the bits are high when ixgbe_msix_other is called. There is a race
3079 	 * condition otherwise which results in possible performance loss
3080 	 * especially if the ixgbe_msix_other interrupt is triggering
3081 	 * consistently (as it would when PPS is turned on for the X540 device)
3082 	 */
3083 	eicr &= 0xFFFF0000;
3084 
3085 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
3086 
3087 	if (eicr & IXGBE_EICR_LSC)
3088 		ixgbe_check_lsc(adapter);
3089 
3090 	if (eicr & IXGBE_EICR_MAILBOX)
3091 		ixgbe_msg_task(adapter);
3092 
3093 	switch (hw->mac.type) {
3094 	case ixgbe_mac_82599EB:
3095 	case ixgbe_mac_X540:
3096 	case ixgbe_mac_X550:
3097 	case ixgbe_mac_X550EM_x:
3098 	case ixgbe_mac_x550em_a:
3099 		if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
3100 		    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
3101 			adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
3102 			ixgbe_service_event_schedule(adapter);
3103 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
3104 					IXGBE_EICR_GPI_SDP0_X540);
3105 		}
3106 		if (eicr & IXGBE_EICR_ECC) {
3107 			e_info(link, "Received ECC Err, initiating reset\n");
3108 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3109 			ixgbe_service_event_schedule(adapter);
3110 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3111 		}
3112 		/* Handle Flow Director Full threshold interrupt */
3113 		if (eicr & IXGBE_EICR_FLOW_DIR) {
3114 			int reinit_count = 0;
3115 			int i;
3116 			for (i = 0; i < adapter->num_tx_queues; i++) {
3117 				struct ixgbe_ring *ring = adapter->tx_ring[i];
3118 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
3119 						       &ring->state))
3120 					reinit_count++;
3121 			}
3122 			if (reinit_count) {
3123 				/* no more flow director interrupts until after init */
3124 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
3125 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
3126 				ixgbe_service_event_schedule(adapter);
3127 			}
3128 		}
3129 		ixgbe_check_sfp_event(adapter, eicr);
3130 		ixgbe_check_overtemp_event(adapter, eicr);
3131 		break;
3132 	default:
3133 		break;
3134 	}
3135 
3136 	ixgbe_check_fan_failure(adapter, eicr);
3137 
3138 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3139 		ixgbe_ptp_check_pps_event(adapter);
3140 
3141 	/* re-enable the original interrupt state, no lsc, no queues */
3142 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3143 		ixgbe_irq_enable(adapter, false, false);
3144 
3145 	return IRQ_HANDLED;
3146 }
3147 
3148 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
3149 {
3150 	struct ixgbe_q_vector *q_vector = data;
3151 
3152 	/* EIAM disabled interrupts (on this vector) for us */
3153 
3154 	if (q_vector->rx.ring || q_vector->tx.ring)
3155 		napi_schedule_irqoff(&q_vector->napi);
3156 
3157 	return IRQ_HANDLED;
3158 }
3159 
3160 /**
3161  * ixgbe_poll - NAPI Rx polling callback
3162  * @napi: structure for representing this polling device
3163  * @budget: how many packets driver is allowed to clean
3164  *
3165  * This function is used for legacy and MSI, NAPI mode
3166  **/
3167 int ixgbe_poll(struct napi_struct *napi, int budget)
3168 {
3169 	struct ixgbe_q_vector *q_vector =
3170 				container_of(napi, struct ixgbe_q_vector, napi);
3171 	struct ixgbe_adapter *adapter = q_vector->adapter;
3172 	struct ixgbe_ring *ring;
3173 	int per_ring_budget, work_done = 0;
3174 	bool clean_complete = true;
3175 
3176 #ifdef CONFIG_IXGBE_DCA
3177 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
3178 		ixgbe_update_dca(q_vector);
3179 #endif
3180 
3181 	ixgbe_for_each_ring(ring, q_vector->tx) {
3182 		bool wd = ring->xsk_pool ?
3183 			  ixgbe_clean_xdp_tx_irq(q_vector, ring, budget) :
3184 			  ixgbe_clean_tx_irq(q_vector, ring, budget);
3185 
3186 		if (!wd)
3187 			clean_complete = false;
3188 	}
3189 
3190 	/* Exit if we are called by netpoll */
3191 	if (budget <= 0)
3192 		return budget;
3193 
3194 	/* attempt to distribute budget to each queue fairly, but don't allow
3195 	 * the budget to go below 1 because we'll exit polling */
3196 	if (q_vector->rx.count > 1)
3197 		per_ring_budget = max(budget/q_vector->rx.count, 1);
3198 	else
3199 		per_ring_budget = budget;
3200 
3201 	ixgbe_for_each_ring(ring, q_vector->rx) {
3202 		int cleaned = ring->xsk_pool ?
3203 			      ixgbe_clean_rx_irq_zc(q_vector, ring,
3204 						    per_ring_budget) :
3205 			      ixgbe_clean_rx_irq(q_vector, ring,
3206 						 per_ring_budget);
3207 
3208 		work_done += cleaned;
3209 		if (cleaned >= per_ring_budget)
3210 			clean_complete = false;
3211 	}
3212 
3213 	/* If all work not completed, return budget and keep polling */
3214 	if (!clean_complete)
3215 		return budget;
3216 
3217 	/* all work done, exit the polling mode */
3218 	if (likely(napi_complete_done(napi, work_done))) {
3219 		if (adapter->rx_itr_setting & 1)
3220 			ixgbe_set_itr(q_vector);
3221 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
3222 			ixgbe_irq_enable_queues(adapter,
3223 						BIT_ULL(q_vector->v_idx));
3224 	}
3225 
3226 	return min(work_done, budget - 1);
3227 }
3228 
3229 /**
3230  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
3231  * @adapter: board private structure
3232  *
3233  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
3234  * interrupts from the kernel.
3235  **/
3236 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
3237 {
3238 	struct net_device *netdev = adapter->netdev;
3239 	unsigned int ri = 0, ti = 0;
3240 	int vector, err;
3241 
3242 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3243 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3244 		struct msix_entry *entry = &adapter->msix_entries[vector];
3245 
3246 		if (q_vector->tx.ring && q_vector->rx.ring) {
3247 			snprintf(q_vector->name, sizeof(q_vector->name),
3248 				 "%s-TxRx-%u", netdev->name, ri++);
3249 			ti++;
3250 		} else if (q_vector->rx.ring) {
3251 			snprintf(q_vector->name, sizeof(q_vector->name),
3252 				 "%s-rx-%u", netdev->name, ri++);
3253 		} else if (q_vector->tx.ring) {
3254 			snprintf(q_vector->name, sizeof(q_vector->name),
3255 				 "%s-tx-%u", netdev->name, ti++);
3256 		} else {
3257 			/* skip this unused q_vector */
3258 			continue;
3259 		}
3260 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
3261 				  q_vector->name, q_vector);
3262 		if (err) {
3263 			e_err(probe, "request_irq failed for MSIX interrupt "
3264 			      "Error: %d\n", err);
3265 			goto free_queue_irqs;
3266 		}
3267 		/* If Flow Director is enabled, set interrupt affinity */
3268 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3269 			/* assign the mask for this irq */
3270 			irq_update_affinity_hint(entry->vector,
3271 						 &q_vector->affinity_mask);
3272 		}
3273 	}
3274 
3275 	err = request_irq(adapter->msix_entries[vector].vector,
3276 			  ixgbe_msix_other, 0, netdev->name, adapter);
3277 	if (err) {
3278 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
3279 		goto free_queue_irqs;
3280 	}
3281 
3282 	return 0;
3283 
3284 free_queue_irqs:
3285 	while (vector) {
3286 		vector--;
3287 		irq_update_affinity_hint(adapter->msix_entries[vector].vector,
3288 					 NULL);
3289 		free_irq(adapter->msix_entries[vector].vector,
3290 			 adapter->q_vector[vector]);
3291 	}
3292 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
3293 	pci_disable_msix(adapter->pdev);
3294 	kfree(adapter->msix_entries);
3295 	adapter->msix_entries = NULL;
3296 	return err;
3297 }
3298 
3299 /**
3300  * ixgbe_intr - legacy mode Interrupt Handler
3301  * @irq: interrupt number
3302  * @data: pointer to a network interface device structure
3303  **/
3304 static irqreturn_t ixgbe_intr(int irq, void *data)
3305 {
3306 	struct ixgbe_adapter *adapter = data;
3307 	struct ixgbe_hw *hw = &adapter->hw;
3308 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3309 	u32 eicr;
3310 
3311 	/*
3312 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
3313 	 * before the read of EICR.
3314 	 */
3315 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
3316 
3317 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
3318 	 * therefore no explicit interrupt disable is necessary */
3319 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3320 	if (!eicr) {
3321 		/*
3322 		 * shared interrupt alert!
3323 		 * make sure interrupts are enabled because the read will
3324 		 * have disabled interrupts due to EIAM
3325 		 * finish the workaround of silicon errata on 82598.  Unmask
3326 		 * the interrupt that we masked before the EICR read.
3327 		 */
3328 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
3329 			ixgbe_irq_enable(adapter, true, true);
3330 		return IRQ_NONE;	/* Not our interrupt */
3331 	}
3332 
3333 	if (eicr & IXGBE_EICR_LSC)
3334 		ixgbe_check_lsc(adapter);
3335 
3336 	switch (hw->mac.type) {
3337 	case ixgbe_mac_82599EB:
3338 		ixgbe_check_sfp_event(adapter, eicr);
3339 		fallthrough;
3340 	case ixgbe_mac_X540:
3341 	case ixgbe_mac_X550:
3342 	case ixgbe_mac_X550EM_x:
3343 	case ixgbe_mac_x550em_a:
3344 		if (eicr & IXGBE_EICR_ECC) {
3345 			e_info(link, "Received ECC Err, initiating reset\n");
3346 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3347 			ixgbe_service_event_schedule(adapter);
3348 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3349 		}
3350 		ixgbe_check_overtemp_event(adapter, eicr);
3351 		break;
3352 	default:
3353 		break;
3354 	}
3355 
3356 	ixgbe_check_fan_failure(adapter, eicr);
3357 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3358 		ixgbe_ptp_check_pps_event(adapter);
3359 
3360 	/* would disable interrupts here but EIAM disabled it */
3361 	napi_schedule_irqoff(&q_vector->napi);
3362 
3363 	/*
3364 	 * re-enable link(maybe) and non-queue interrupts, no flush.
3365 	 * ixgbe_poll will re-enable the queue interrupts
3366 	 */
3367 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3368 		ixgbe_irq_enable(adapter, false, false);
3369 
3370 	return IRQ_HANDLED;
3371 }
3372 
3373 /**
3374  * ixgbe_request_irq - initialize interrupts
3375  * @adapter: board private structure
3376  *
3377  * Attempts to configure interrupts using the best available
3378  * capabilities of the hardware and kernel.
3379  **/
3380 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
3381 {
3382 	struct net_device *netdev = adapter->netdev;
3383 	int err;
3384 
3385 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
3386 		err = ixgbe_request_msix_irqs(adapter);
3387 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
3388 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
3389 				  netdev->name, adapter);
3390 	else
3391 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
3392 				  netdev->name, adapter);
3393 
3394 	if (err)
3395 		e_err(probe, "request_irq failed, Error %d\n", err);
3396 
3397 	return err;
3398 }
3399 
3400 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
3401 {
3402 	int vector;
3403 
3404 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
3405 		free_irq(adapter->pdev->irq, adapter);
3406 		return;
3407 	}
3408 
3409 	if (!adapter->msix_entries)
3410 		return;
3411 
3412 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3413 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3414 		struct msix_entry *entry = &adapter->msix_entries[vector];
3415 
3416 		/* free only the irqs that were actually requested */
3417 		if (!q_vector->rx.ring && !q_vector->tx.ring)
3418 			continue;
3419 
3420 		/* clear the affinity_mask in the IRQ descriptor */
3421 		irq_update_affinity_hint(entry->vector, NULL);
3422 
3423 		free_irq(entry->vector, q_vector);
3424 	}
3425 
3426 	free_irq(adapter->msix_entries[vector].vector, adapter);
3427 }
3428 
3429 /**
3430  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3431  * @adapter: board private structure
3432  **/
3433 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3434 {
3435 	switch (adapter->hw.mac.type) {
3436 	case ixgbe_mac_82598EB:
3437 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3438 		break;
3439 	case ixgbe_mac_82599EB:
3440 	case ixgbe_mac_X540:
3441 	case ixgbe_mac_X550:
3442 	case ixgbe_mac_X550EM_x:
3443 	case ixgbe_mac_x550em_a:
3444 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3445 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3446 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3447 		break;
3448 	default:
3449 		break;
3450 	}
3451 	IXGBE_WRITE_FLUSH(&adapter->hw);
3452 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3453 		int vector;
3454 
3455 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
3456 			synchronize_irq(adapter->msix_entries[vector].vector);
3457 
3458 		synchronize_irq(adapter->msix_entries[vector++].vector);
3459 	} else {
3460 		synchronize_irq(adapter->pdev->irq);
3461 	}
3462 }
3463 
3464 /**
3465  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3466  * @adapter: board private structure
3467  *
3468  **/
3469 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3470 {
3471 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3472 
3473 	ixgbe_write_eitr(q_vector);
3474 
3475 	ixgbe_set_ivar(adapter, 0, 0, 0);
3476 	ixgbe_set_ivar(adapter, 1, 0, 0);
3477 
3478 	e_info(hw, "Legacy interrupt IVAR setup done\n");
3479 }
3480 
3481 /**
3482  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3483  * @adapter: board private structure
3484  * @ring: structure containing ring specific data
3485  *
3486  * Configure the Tx descriptor ring after a reset.
3487  **/
3488 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3489 			     struct ixgbe_ring *ring)
3490 {
3491 	struct ixgbe_hw *hw = &adapter->hw;
3492 	u64 tdba = ring->dma;
3493 	int wait_loop = 10;
3494 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
3495 	u8 reg_idx = ring->reg_idx;
3496 
3497 	ring->xsk_pool = NULL;
3498 	if (ring_is_xdp(ring))
3499 		ring->xsk_pool = ixgbe_xsk_pool(adapter, ring);
3500 
3501 	/* disable queue to avoid issues while updating state */
3502 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3503 	IXGBE_WRITE_FLUSH(hw);
3504 
3505 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3506 			(tdba & DMA_BIT_MASK(32)));
3507 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3508 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3509 			ring->count * sizeof(union ixgbe_adv_tx_desc));
3510 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3511 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3512 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3513 
3514 	/*
3515 	 * set WTHRESH to encourage burst writeback, it should not be set
3516 	 * higher than 1 when:
3517 	 * - ITR is 0 as it could cause false TX hangs
3518 	 * - ITR is set to > 100k int/sec and BQL is enabled
3519 	 *
3520 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3521 	 * to or less than the number of on chip descriptors, which is
3522 	 * currently 40.
3523 	 */
3524 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3525 		txdctl |= 1u << 16;	/* WTHRESH = 1 */
3526 	else
3527 		txdctl |= 8u << 16;	/* WTHRESH = 8 */
3528 
3529 	/*
3530 	 * Setting PTHRESH to 32 both improves performance
3531 	 * and avoids a TX hang with DFP enabled
3532 	 */
3533 	txdctl |= (1u << 8) |	/* HTHRESH = 1 */
3534 		   32;		/* PTHRESH = 32 */
3535 
3536 	/* reinitialize flowdirector state */
3537 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3538 		ring->atr_sample_rate = adapter->atr_sample_rate;
3539 		ring->atr_count = 0;
3540 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3541 	} else {
3542 		ring->atr_sample_rate = 0;
3543 	}
3544 
3545 	/* initialize XPS */
3546 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3547 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3548 
3549 		if (q_vector)
3550 			netif_set_xps_queue(ring->netdev,
3551 					    &q_vector->affinity_mask,
3552 					    ring->queue_index);
3553 	}
3554 
3555 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3556 
3557 	/* reinitialize tx_buffer_info */
3558 	memset(ring->tx_buffer_info, 0,
3559 	       sizeof(struct ixgbe_tx_buffer) * ring->count);
3560 
3561 	/* enable queue */
3562 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3563 
3564 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3565 	if (hw->mac.type == ixgbe_mac_82598EB &&
3566 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3567 		return;
3568 
3569 	/* poll to verify queue is enabled */
3570 	do {
3571 		usleep_range(1000, 2000);
3572 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3573 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3574 	if (!wait_loop)
3575 		hw_dbg(hw, "Could not enable Tx Queue %d\n", reg_idx);
3576 }
3577 
3578 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3579 {
3580 	struct ixgbe_hw *hw = &adapter->hw;
3581 	u32 rttdcs, mtqc;
3582 	u8 tcs = adapter->hw_tcs;
3583 
3584 	if (hw->mac.type == ixgbe_mac_82598EB)
3585 		return;
3586 
3587 	/* disable the arbiter while setting MTQC */
3588 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3589 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3590 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3591 
3592 	/* set transmit pool layout */
3593 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3594 		mtqc = IXGBE_MTQC_VT_ENA;
3595 		if (tcs > 4)
3596 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3597 		else if (tcs > 1)
3598 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3599 		else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3600 			 IXGBE_82599_VMDQ_4Q_MASK)
3601 			mtqc |= IXGBE_MTQC_32VF;
3602 		else
3603 			mtqc |= IXGBE_MTQC_64VF;
3604 	} else {
3605 		if (tcs > 4) {
3606 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3607 		} else if (tcs > 1) {
3608 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3609 		} else {
3610 			u8 max_txq = adapter->num_tx_queues +
3611 				adapter->num_xdp_queues;
3612 			if (max_txq > 63)
3613 				mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3614 			else
3615 				mtqc = IXGBE_MTQC_64Q_1PB;
3616 		}
3617 	}
3618 
3619 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3620 
3621 	/* Enable Security TX Buffer IFG for multiple pb */
3622 	if (tcs) {
3623 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3624 		sectx |= IXGBE_SECTX_DCB;
3625 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3626 	}
3627 
3628 	/* re-enable the arbiter */
3629 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3630 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3631 }
3632 
3633 /**
3634  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3635  * @adapter: board private structure
3636  *
3637  * Configure the Tx unit of the MAC after a reset.
3638  **/
3639 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3640 {
3641 	struct ixgbe_hw *hw = &adapter->hw;
3642 	u32 dmatxctl;
3643 	u32 i;
3644 
3645 	ixgbe_setup_mtqc(adapter);
3646 
3647 	if (hw->mac.type != ixgbe_mac_82598EB) {
3648 		/* DMATXCTL.EN must be before Tx queues are enabled */
3649 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3650 		dmatxctl |= IXGBE_DMATXCTL_TE;
3651 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3652 	}
3653 
3654 	/* Setup the HW Tx Head and Tail descriptor pointers */
3655 	for (i = 0; i < adapter->num_tx_queues; i++)
3656 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3657 	for (i = 0; i < adapter->num_xdp_queues; i++)
3658 		ixgbe_configure_tx_ring(adapter, adapter->xdp_ring[i]);
3659 }
3660 
3661 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3662 				 struct ixgbe_ring *ring)
3663 {
3664 	struct ixgbe_hw *hw = &adapter->hw;
3665 	u8 reg_idx = ring->reg_idx;
3666 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3667 
3668 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3669 
3670 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3671 }
3672 
3673 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3674 				  struct ixgbe_ring *ring)
3675 {
3676 	struct ixgbe_hw *hw = &adapter->hw;
3677 	u8 reg_idx = ring->reg_idx;
3678 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3679 
3680 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3681 
3682 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3683 }
3684 
3685 #ifdef CONFIG_IXGBE_DCB
3686 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3687 #else
3688 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3689 #endif
3690 {
3691 	int i;
3692 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3693 
3694 	if (adapter->ixgbe_ieee_pfc)
3695 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3696 
3697 	/*
3698 	 * We should set the drop enable bit if:
3699 	 *  SR-IOV is enabled
3700 	 *   or
3701 	 *  Number of Rx queues > 1 and flow control is disabled
3702 	 *
3703 	 *  This allows us to avoid head of line blocking for security
3704 	 *  and performance reasons.
3705 	 */
3706 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3707 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3708 		for (i = 0; i < adapter->num_rx_queues; i++)
3709 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3710 	} else {
3711 		for (i = 0; i < adapter->num_rx_queues; i++)
3712 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3713 	}
3714 }
3715 
3716 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3717 
3718 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3719 				   struct ixgbe_ring *rx_ring)
3720 {
3721 	struct ixgbe_hw *hw = &adapter->hw;
3722 	u32 srrctl;
3723 	u8 reg_idx = rx_ring->reg_idx;
3724 
3725 	if (hw->mac.type == ixgbe_mac_82598EB) {
3726 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3727 
3728 		/*
3729 		 * if VMDq is not active we must program one srrctl register
3730 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3731 		 */
3732 		reg_idx &= mask;
3733 	}
3734 
3735 	/* configure header buffer length, needed for RSC */
3736 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3737 
3738 	/* configure the packet buffer length */
3739 	if (rx_ring->xsk_pool) {
3740 		u32 xsk_buf_len = xsk_pool_get_rx_frame_size(rx_ring->xsk_pool);
3741 
3742 		/* If the MAC support setting RXDCTL.RLPML, the
3743 		 * SRRCTL[n].BSIZEPKT is set to PAGE_SIZE and
3744 		 * RXDCTL.RLPML is set to the actual UMEM buffer
3745 		 * size. If not, then we are stuck with a 1k buffer
3746 		 * size resolution. In this case frames larger than
3747 		 * the UMEM buffer size viewed in a 1k resolution will
3748 		 * be dropped.
3749 		 */
3750 		if (hw->mac.type != ixgbe_mac_82599EB)
3751 			srrctl |= PAGE_SIZE >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3752 		else
3753 			srrctl |= xsk_buf_len >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3754 	} else if (test_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state)) {
3755 		srrctl |= IXGBE_RXBUFFER_3K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3756 	} else {
3757 		srrctl |= IXGBE_RXBUFFER_2K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3758 	}
3759 
3760 	/* configure descriptor type */
3761 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3762 
3763 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3764 }
3765 
3766 /**
3767  * ixgbe_rss_indir_tbl_entries - Return RSS indirection table entries
3768  * @adapter: device handle
3769  *
3770  *  - 82598/82599/X540:     128
3771  *  - X550(non-SRIOV mode): 512
3772  *  - X550(SRIOV mode):     64
3773  */
3774 u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3775 {
3776 	if (adapter->hw.mac.type < ixgbe_mac_X550)
3777 		return 128;
3778 	else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3779 		return 64;
3780 	else
3781 		return 512;
3782 }
3783 
3784 /**
3785  * ixgbe_store_key - Write the RSS key to HW
3786  * @adapter: device handle
3787  *
3788  * Write the RSS key stored in adapter.rss_key to HW.
3789  */
3790 void ixgbe_store_key(struct ixgbe_adapter *adapter)
3791 {
3792 	struct ixgbe_hw *hw = &adapter->hw;
3793 	int i;
3794 
3795 	for (i = 0; i < 10; i++)
3796 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3797 }
3798 
3799 /**
3800  * ixgbe_init_rss_key - Initialize adapter RSS key
3801  * @adapter: device handle
3802  *
3803  * Allocates and initializes the RSS key if it is not allocated.
3804  **/
3805 static inline int ixgbe_init_rss_key(struct ixgbe_adapter *adapter)
3806 {
3807 	u32 *rss_key;
3808 
3809 	if (!adapter->rss_key) {
3810 		rss_key = kzalloc(IXGBE_RSS_KEY_SIZE, GFP_KERNEL);
3811 		if (unlikely(!rss_key))
3812 			return -ENOMEM;
3813 
3814 		netdev_rss_key_fill(rss_key, IXGBE_RSS_KEY_SIZE);
3815 		adapter->rss_key = rss_key;
3816 	}
3817 
3818 	return 0;
3819 }
3820 
3821 /**
3822  * ixgbe_store_reta - Write the RETA table to HW
3823  * @adapter: device handle
3824  *
3825  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3826  */
3827 void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3828 {
3829 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3830 	struct ixgbe_hw *hw = &adapter->hw;
3831 	u32 reta = 0;
3832 	u32 indices_multi;
3833 	u8 *indir_tbl = adapter->rss_indir_tbl;
3834 
3835 	/* Fill out the redirection table as follows:
3836 	 *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3837 	 *    indices.
3838 	 *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3839 	 *  - X550:       8 bit wide entries containing 6 bit RSS index
3840 	 */
3841 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3842 		indices_multi = 0x11;
3843 	else
3844 		indices_multi = 0x1;
3845 
3846 	/* Write redirection table to HW */
3847 	for (i = 0; i < reta_entries; i++) {
3848 		reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3849 		if ((i & 3) == 3) {
3850 			if (i < 128)
3851 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3852 			else
3853 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3854 						reta);
3855 			reta = 0;
3856 		}
3857 	}
3858 }
3859 
3860 /**
3861  * ixgbe_store_vfreta - Write the RETA table to HW (x550 devices in SRIOV mode)
3862  * @adapter: device handle
3863  *
3864  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3865  */
3866 static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3867 {
3868 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3869 	struct ixgbe_hw *hw = &adapter->hw;
3870 	u32 vfreta = 0;
3871 
3872 	/* Write redirection table to HW */
3873 	for (i = 0; i < reta_entries; i++) {
3874 		u16 pool = adapter->num_rx_pools;
3875 
3876 		vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3877 		if ((i & 3) != 3)
3878 			continue;
3879 
3880 		while (pool--)
3881 			IXGBE_WRITE_REG(hw,
3882 					IXGBE_PFVFRETA(i >> 2, VMDQ_P(pool)),
3883 					vfreta);
3884 		vfreta = 0;
3885 	}
3886 }
3887 
3888 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3889 {
3890 	u32 i, j;
3891 	u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3892 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3893 
3894 	/* Program table for at least 4 queues w/ SR-IOV so that VFs can
3895 	 * make full use of any rings they may have.  We will use the
3896 	 * PSRTYPE register to control how many rings we use within the PF.
3897 	 */
3898 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 4))
3899 		rss_i = 4;
3900 
3901 	/* Fill out hash function seeds */
3902 	ixgbe_store_key(adapter);
3903 
3904 	/* Fill out redirection table */
3905 	memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3906 
3907 	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3908 		if (j == rss_i)
3909 			j = 0;
3910 
3911 		adapter->rss_indir_tbl[i] = j;
3912 	}
3913 
3914 	ixgbe_store_reta(adapter);
3915 }
3916 
3917 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3918 {
3919 	struct ixgbe_hw *hw = &adapter->hw;
3920 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3921 	int i, j;
3922 
3923 	/* Fill out hash function seeds */
3924 	for (i = 0; i < 10; i++) {
3925 		u16 pool = adapter->num_rx_pools;
3926 
3927 		while (pool--)
3928 			IXGBE_WRITE_REG(hw,
3929 					IXGBE_PFVFRSSRK(i, VMDQ_P(pool)),
3930 					*(adapter->rss_key + i));
3931 	}
3932 
3933 	/* Fill out the redirection table */
3934 	for (i = 0, j = 0; i < 64; i++, j++) {
3935 		if (j == rss_i)
3936 			j = 0;
3937 
3938 		adapter->rss_indir_tbl[i] = j;
3939 	}
3940 
3941 	ixgbe_store_vfreta(adapter);
3942 }
3943 
3944 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3945 {
3946 	struct ixgbe_hw *hw = &adapter->hw;
3947 	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3948 	u32 rxcsum;
3949 
3950 	/* Disable indicating checksum in descriptor, enables RSS hash */
3951 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3952 	rxcsum |= IXGBE_RXCSUM_PCSD;
3953 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3954 
3955 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3956 		if (adapter->ring_feature[RING_F_RSS].mask)
3957 			mrqc = IXGBE_MRQC_RSSEN;
3958 	} else {
3959 		u8 tcs = adapter->hw_tcs;
3960 
3961 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3962 			if (tcs > 4)
3963 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3964 			else if (tcs > 1)
3965 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3966 			else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3967 				 IXGBE_82599_VMDQ_4Q_MASK)
3968 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3969 			else
3970 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3971 
3972 			/* Enable L3/L4 for Tx Switched packets only for X550,
3973 			 * older devices do not support this feature
3974 			 */
3975 			if (hw->mac.type >= ixgbe_mac_X550)
3976 				mrqc |= IXGBE_MRQC_L3L4TXSWEN;
3977 		} else {
3978 			if (tcs > 4)
3979 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3980 			else if (tcs > 1)
3981 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3982 			else
3983 				mrqc = IXGBE_MRQC_RSSEN;
3984 		}
3985 	}
3986 
3987 	/* Perform hash on these packet types */
3988 	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3989 		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3990 		     IXGBE_MRQC_RSS_FIELD_IPV6 |
3991 		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3992 
3993 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3994 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3995 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3996 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3997 
3998 	if ((hw->mac.type >= ixgbe_mac_X550) &&
3999 	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
4000 		u16 pool = adapter->num_rx_pools;
4001 
4002 		/* Enable VF RSS mode */
4003 		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
4004 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
4005 
4006 		/* Setup RSS through the VF registers */
4007 		ixgbe_setup_vfreta(adapter);
4008 		vfmrqc = IXGBE_MRQC_RSSEN;
4009 		vfmrqc |= rss_field;
4010 
4011 		while (pool--)
4012 			IXGBE_WRITE_REG(hw,
4013 					IXGBE_PFVFMRQC(VMDQ_P(pool)),
4014 					vfmrqc);
4015 	} else {
4016 		ixgbe_setup_reta(adapter);
4017 		mrqc |= rss_field;
4018 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
4019 	}
4020 }
4021 
4022 /**
4023  * ixgbe_configure_rscctl - enable RSC for the indicated ring
4024  * @adapter: address of board private structure
4025  * @ring: structure containing ring specific data
4026  **/
4027 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
4028 				   struct ixgbe_ring *ring)
4029 {
4030 	struct ixgbe_hw *hw = &adapter->hw;
4031 	u32 rscctrl;
4032 	u8 reg_idx = ring->reg_idx;
4033 
4034 	if (!ring_is_rsc_enabled(ring))
4035 		return;
4036 
4037 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
4038 	rscctrl |= IXGBE_RSCCTL_RSCEN;
4039 	/*
4040 	 * we must limit the number of descriptors so that the
4041 	 * total size of max desc * buf_len is not greater
4042 	 * than 65536
4043 	 */
4044 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
4045 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
4046 }
4047 
4048 #define IXGBE_MAX_RX_DESC_POLL 10
4049 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
4050 				       struct ixgbe_ring *ring)
4051 {
4052 	struct ixgbe_hw *hw = &adapter->hw;
4053 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
4054 	u32 rxdctl;
4055 	u8 reg_idx = ring->reg_idx;
4056 
4057 	if (ixgbe_removed(hw->hw_addr))
4058 		return;
4059 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
4060 	if (hw->mac.type == ixgbe_mac_82598EB &&
4061 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
4062 		return;
4063 
4064 	do {
4065 		usleep_range(1000, 2000);
4066 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4067 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
4068 
4069 	if (!wait_loop) {
4070 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
4071 		      "the polling period\n", reg_idx);
4072 	}
4073 }
4074 
4075 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
4076 			     struct ixgbe_ring *ring)
4077 {
4078 	struct ixgbe_hw *hw = &adapter->hw;
4079 	union ixgbe_adv_rx_desc *rx_desc;
4080 	u64 rdba = ring->dma;
4081 	u32 rxdctl;
4082 	u8 reg_idx = ring->reg_idx;
4083 
4084 	xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq);
4085 	ring->xsk_pool = ixgbe_xsk_pool(adapter, ring);
4086 	if (ring->xsk_pool) {
4087 		WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
4088 						   MEM_TYPE_XSK_BUFF_POOL,
4089 						   NULL));
4090 		xsk_pool_set_rxq_info(ring->xsk_pool, &ring->xdp_rxq);
4091 	} else {
4092 		WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
4093 						   MEM_TYPE_PAGE_SHARED, NULL));
4094 	}
4095 
4096 	/* disable queue to avoid use of these values while updating state */
4097 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4098 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
4099 
4100 	/* write value back with RXDCTL.ENABLE bit cleared */
4101 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4102 	IXGBE_WRITE_FLUSH(hw);
4103 
4104 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
4105 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
4106 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
4107 			ring->count * sizeof(union ixgbe_adv_rx_desc));
4108 	/* Force flushing of IXGBE_RDLEN to prevent MDD */
4109 	IXGBE_WRITE_FLUSH(hw);
4110 
4111 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
4112 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
4113 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
4114 
4115 	ixgbe_configure_srrctl(adapter, ring);
4116 	ixgbe_configure_rscctl(adapter, ring);
4117 
4118 	if (hw->mac.type == ixgbe_mac_82598EB) {
4119 		/*
4120 		 * enable cache line friendly hardware writes:
4121 		 * PTHRESH=32 descriptors (half the internal cache),
4122 		 * this also removes ugly rx_no_buffer_count increment
4123 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
4124 		 * WTHRESH=8 burst writeback up to two cache lines
4125 		 */
4126 		rxdctl &= ~0x3FFFFF;
4127 		rxdctl |=  0x080420;
4128 #if (PAGE_SIZE < 8192)
4129 	/* RXDCTL.RLPML does not work on 82599 */
4130 	} else if (hw->mac.type != ixgbe_mac_82599EB) {
4131 		rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
4132 			    IXGBE_RXDCTL_RLPML_EN);
4133 
4134 		/* Limit the maximum frame size so we don't overrun the skb.
4135 		 * This can happen in SRIOV mode when the MTU of the VF is
4136 		 * higher than the MTU of the PF.
4137 		 */
4138 		if (ring_uses_build_skb(ring) &&
4139 		    !test_bit(__IXGBE_RX_3K_BUFFER, &ring->state))
4140 			rxdctl |= IXGBE_MAX_2K_FRAME_BUILD_SKB |
4141 				  IXGBE_RXDCTL_RLPML_EN;
4142 #endif
4143 	}
4144 
4145 	ring->rx_offset = ixgbe_rx_offset(ring);
4146 
4147 	if (ring->xsk_pool && hw->mac.type != ixgbe_mac_82599EB) {
4148 		u32 xsk_buf_len = xsk_pool_get_rx_frame_size(ring->xsk_pool);
4149 
4150 		rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
4151 			    IXGBE_RXDCTL_RLPML_EN);
4152 		rxdctl |= xsk_buf_len | IXGBE_RXDCTL_RLPML_EN;
4153 
4154 		ring->rx_buf_len = xsk_buf_len;
4155 	}
4156 
4157 	/* initialize rx_buffer_info */
4158 	memset(ring->rx_buffer_info, 0,
4159 	       sizeof(struct ixgbe_rx_buffer) * ring->count);
4160 
4161 	/* initialize Rx descriptor 0 */
4162 	rx_desc = IXGBE_RX_DESC(ring, 0);
4163 	rx_desc->wb.upper.length = 0;
4164 
4165 	/* enable receive descriptor ring */
4166 	rxdctl |= IXGBE_RXDCTL_ENABLE;
4167 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4168 
4169 	ixgbe_rx_desc_queue_enable(adapter, ring);
4170 	if (ring->xsk_pool)
4171 		ixgbe_alloc_rx_buffers_zc(ring, ixgbe_desc_unused(ring));
4172 	else
4173 		ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
4174 }
4175 
4176 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
4177 {
4178 	struct ixgbe_hw *hw = &adapter->hw;
4179 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
4180 	u16 pool = adapter->num_rx_pools;
4181 
4182 	/* PSRTYPE must be initialized in non 82598 adapters */
4183 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4184 		      IXGBE_PSRTYPE_UDPHDR |
4185 		      IXGBE_PSRTYPE_IPV4HDR |
4186 		      IXGBE_PSRTYPE_L2HDR |
4187 		      IXGBE_PSRTYPE_IPV6HDR;
4188 
4189 	if (hw->mac.type == ixgbe_mac_82598EB)
4190 		return;
4191 
4192 	if (rss_i > 3)
4193 		psrtype |= 2u << 29;
4194 	else if (rss_i > 1)
4195 		psrtype |= 1u << 29;
4196 
4197 	while (pool--)
4198 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4199 }
4200 
4201 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
4202 {
4203 	struct ixgbe_hw *hw = &adapter->hw;
4204 	u16 pool = adapter->num_rx_pools;
4205 	u32 reg_offset, vf_shift, vmolr;
4206 	u32 gcr_ext, vmdctl;
4207 	int i;
4208 
4209 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
4210 		return;
4211 
4212 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
4213 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
4214 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
4215 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
4216 	vmdctl |= IXGBE_VT_CTL_REPLEN;
4217 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
4218 
4219 	/* accept untagged packets until a vlan tag is
4220 	 * specifically set for the VMDQ queue/pool
4221 	 */
4222 	vmolr = IXGBE_VMOLR_AUPE;
4223 	while (pool--)
4224 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(pool)), vmolr);
4225 
4226 	vf_shift = VMDQ_P(0) % 32;
4227 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
4228 
4229 	/* Enable only the PF's pool for Tx/Rx */
4230 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), GENMASK(31, vf_shift));
4231 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
4232 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), GENMASK(31, vf_shift));
4233 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
4234 	if (adapter->bridge_mode == BRIDGE_MODE_VEB)
4235 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
4236 
4237 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
4238 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
4239 
4240 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
4241 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4242 
4243 	/*
4244 	 * Set up VF register offsets for selected VT Mode,
4245 	 * i.e. 32 or 64 VFs for SR-IOV
4246 	 */
4247 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4248 	case IXGBE_82599_VMDQ_8Q_MASK:
4249 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
4250 		break;
4251 	case IXGBE_82599_VMDQ_4Q_MASK:
4252 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
4253 		break;
4254 	default:
4255 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
4256 		break;
4257 	}
4258 
4259 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
4260 
4261 	for (i = 0; i < adapter->num_vfs; i++) {
4262 		/* configure spoof checking */
4263 		ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i,
4264 					  adapter->vfinfo[i].spoofchk_enabled);
4265 
4266 		/* Enable/Disable RSS query feature  */
4267 		ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
4268 					  adapter->vfinfo[i].rss_query_enabled);
4269 	}
4270 }
4271 
4272 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
4273 {
4274 	struct ixgbe_hw *hw = &adapter->hw;
4275 	struct net_device *netdev = adapter->netdev;
4276 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4277 	struct ixgbe_ring *rx_ring;
4278 	int i;
4279 	u32 mhadd, hlreg0;
4280 
4281 #ifdef IXGBE_FCOE
4282 	/* adjust max frame to be able to do baby jumbo for FCoE */
4283 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
4284 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
4285 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4286 
4287 #endif /* IXGBE_FCOE */
4288 
4289 	/* adjust max frame to be at least the size of a standard frame */
4290 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
4291 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
4292 
4293 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
4294 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
4295 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
4296 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
4297 
4298 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
4299 	}
4300 
4301 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4302 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
4303 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4304 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4305 
4306 	/*
4307 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4308 	 * the Base and Length of the Rx Descriptor Ring
4309 	 */
4310 	for (i = 0; i < adapter->num_rx_queues; i++) {
4311 		rx_ring = adapter->rx_ring[i];
4312 
4313 		clear_ring_rsc_enabled(rx_ring);
4314 		clear_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4315 		clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4316 
4317 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4318 			set_ring_rsc_enabled(rx_ring);
4319 
4320 		if (test_bit(__IXGBE_RX_FCOE, &rx_ring->state))
4321 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4322 
4323 		if (adapter->flags2 & IXGBE_FLAG2_RX_LEGACY)
4324 			continue;
4325 
4326 		set_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4327 
4328 #if (PAGE_SIZE < 8192)
4329 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4330 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4331 
4332 		if (IXGBE_2K_TOO_SMALL_WITH_PADDING ||
4333 		    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
4334 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4335 #endif
4336 	}
4337 }
4338 
4339 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
4340 {
4341 	struct ixgbe_hw *hw = &adapter->hw;
4342 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
4343 
4344 	switch (hw->mac.type) {
4345 	case ixgbe_mac_82598EB:
4346 		/*
4347 		 * For VMDq support of different descriptor types or
4348 		 * buffer sizes through the use of multiple SRRCTL
4349 		 * registers, RDRXCTL.MVMEN must be set to 1
4350 		 *
4351 		 * also, the manual doesn't mention it clearly but DCA hints
4352 		 * will only use queue 0's tags unless this bit is set.  Side
4353 		 * effects of setting this bit are only that SRRCTL must be
4354 		 * fully programmed [0..15]
4355 		 */
4356 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
4357 		break;
4358 	case ixgbe_mac_X550:
4359 	case ixgbe_mac_X550EM_x:
4360 	case ixgbe_mac_x550em_a:
4361 		if (adapter->num_vfs)
4362 			rdrxctl |= IXGBE_RDRXCTL_PSP;
4363 		fallthrough;
4364 	case ixgbe_mac_82599EB:
4365 	case ixgbe_mac_X540:
4366 		/* Disable RSC for ACK packets */
4367 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
4368 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
4369 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
4370 		/* hardware requires some bits to be set by default */
4371 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
4372 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
4373 		break;
4374 	default:
4375 		/* We should do nothing since we don't know this hardware */
4376 		return;
4377 	}
4378 
4379 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
4380 }
4381 
4382 /**
4383  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
4384  * @adapter: board private structure
4385  *
4386  * Configure the Rx unit of the MAC after a reset.
4387  **/
4388 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
4389 {
4390 	struct ixgbe_hw *hw = &adapter->hw;
4391 	int i;
4392 	u32 rxctrl, rfctl;
4393 
4394 	/* disable receives while setting up the descriptors */
4395 	hw->mac.ops.disable_rx(hw);
4396 
4397 	ixgbe_setup_psrtype(adapter);
4398 	ixgbe_setup_rdrxctl(adapter);
4399 
4400 	/* RSC Setup */
4401 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
4402 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
4403 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
4404 		rfctl |= IXGBE_RFCTL_RSC_DIS;
4405 
4406 	/* disable NFS filtering */
4407 	rfctl |= (IXGBE_RFCTL_NFSW_DIS | IXGBE_RFCTL_NFSR_DIS);
4408 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
4409 
4410 	/* Program registers for the distribution of queues */
4411 	ixgbe_setup_mrqc(adapter);
4412 
4413 	/* set_rx_buffer_len must be called before ring initialization */
4414 	ixgbe_set_rx_buffer_len(adapter);
4415 
4416 	/*
4417 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4418 	 * the Base and Length of the Rx Descriptor Ring
4419 	 */
4420 	for (i = 0; i < adapter->num_rx_queues; i++)
4421 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
4422 
4423 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4424 	/* disable drop enable for 82598 parts */
4425 	if (hw->mac.type == ixgbe_mac_82598EB)
4426 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
4427 
4428 	/* enable all receives */
4429 	rxctrl |= IXGBE_RXCTRL_RXEN;
4430 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
4431 }
4432 
4433 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
4434 				 __be16 proto, u16 vid)
4435 {
4436 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4437 	struct ixgbe_hw *hw = &adapter->hw;
4438 
4439 	/* add VID to filter table */
4440 	if (!vid || !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4441 		hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true, !!vid);
4442 
4443 	set_bit(vid, adapter->active_vlans);
4444 
4445 	return 0;
4446 }
4447 
4448 static int ixgbe_find_vlvf_entry(struct ixgbe_hw *hw, u32 vlan)
4449 {
4450 	u32 vlvf;
4451 	int idx;
4452 
4453 	/* short cut the special case */
4454 	if (vlan == 0)
4455 		return 0;
4456 
4457 	/* Search for the vlan id in the VLVF entries */
4458 	for (idx = IXGBE_VLVF_ENTRIES; --idx;) {
4459 		vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(idx));
4460 		if ((vlvf & VLAN_VID_MASK) == vlan)
4461 			break;
4462 	}
4463 
4464 	return idx;
4465 }
4466 
4467 void ixgbe_update_pf_promisc_vlvf(struct ixgbe_adapter *adapter, u32 vid)
4468 {
4469 	struct ixgbe_hw *hw = &adapter->hw;
4470 	u32 bits, word;
4471 	int idx;
4472 
4473 	idx = ixgbe_find_vlvf_entry(hw, vid);
4474 	if (!idx)
4475 		return;
4476 
4477 	/* See if any other pools are set for this VLAN filter
4478 	 * entry other than the PF.
4479 	 */
4480 	word = idx * 2 + (VMDQ_P(0) / 32);
4481 	bits = ~BIT(VMDQ_P(0) % 32);
4482 	bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4483 
4484 	/* Disable the filter so this falls into the default pool. */
4485 	if (!bits && !IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1))) {
4486 		if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4487 			IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), 0);
4488 		IXGBE_WRITE_REG(hw, IXGBE_VLVF(idx), 0);
4489 	}
4490 }
4491 
4492 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
4493 				  __be16 proto, u16 vid)
4494 {
4495 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4496 	struct ixgbe_hw *hw = &adapter->hw;
4497 
4498 	/* remove VID from filter table */
4499 	if (vid && !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4500 		hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), false, true);
4501 
4502 	clear_bit(vid, adapter->active_vlans);
4503 
4504 	return 0;
4505 }
4506 
4507 /**
4508  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
4509  * @adapter: driver data
4510  */
4511 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
4512 {
4513 	struct ixgbe_hw *hw = &adapter->hw;
4514 	u32 vlnctrl;
4515 	int i, j;
4516 
4517 	switch (hw->mac.type) {
4518 	case ixgbe_mac_82598EB:
4519 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4520 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
4521 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4522 		break;
4523 	case ixgbe_mac_82599EB:
4524 	case ixgbe_mac_X540:
4525 	case ixgbe_mac_X550:
4526 	case ixgbe_mac_X550EM_x:
4527 	case ixgbe_mac_x550em_a:
4528 		for (i = 0; i < adapter->num_rx_queues; i++) {
4529 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4530 
4531 			if (!netif_is_ixgbe(ring->netdev))
4532 				continue;
4533 
4534 			j = ring->reg_idx;
4535 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4536 			vlnctrl &= ~IXGBE_RXDCTL_VME;
4537 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4538 		}
4539 		break;
4540 	default:
4541 		break;
4542 	}
4543 }
4544 
4545 /**
4546  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
4547  * @adapter: driver data
4548  */
4549 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
4550 {
4551 	struct ixgbe_hw *hw = &adapter->hw;
4552 	u32 vlnctrl;
4553 	int i, j;
4554 
4555 	switch (hw->mac.type) {
4556 	case ixgbe_mac_82598EB:
4557 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4558 		vlnctrl |= IXGBE_VLNCTRL_VME;
4559 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4560 		break;
4561 	case ixgbe_mac_82599EB:
4562 	case ixgbe_mac_X540:
4563 	case ixgbe_mac_X550:
4564 	case ixgbe_mac_X550EM_x:
4565 	case ixgbe_mac_x550em_a:
4566 		for (i = 0; i < adapter->num_rx_queues; i++) {
4567 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4568 
4569 			if (!netif_is_ixgbe(ring->netdev))
4570 				continue;
4571 
4572 			j = ring->reg_idx;
4573 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4574 			vlnctrl |= IXGBE_RXDCTL_VME;
4575 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4576 		}
4577 		break;
4578 	default:
4579 		break;
4580 	}
4581 }
4582 
4583 static void ixgbe_vlan_promisc_enable(struct ixgbe_adapter *adapter)
4584 {
4585 	struct ixgbe_hw *hw = &adapter->hw;
4586 	u32 vlnctrl, i;
4587 
4588 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4589 
4590 	if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) {
4591 	/* For VMDq and SR-IOV we must leave VLAN filtering enabled */
4592 		vlnctrl |= IXGBE_VLNCTRL_VFE;
4593 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4594 	} else {
4595 		vlnctrl &= ~IXGBE_VLNCTRL_VFE;
4596 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4597 		return;
4598 	}
4599 
4600 	/* Nothing to do for 82598 */
4601 	if (hw->mac.type == ixgbe_mac_82598EB)
4602 		return;
4603 
4604 	/* We are already in VLAN promisc, nothing to do */
4605 	if (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC)
4606 		return;
4607 
4608 	/* Set flag so we don't redo unnecessary work */
4609 	adapter->flags2 |= IXGBE_FLAG2_VLAN_PROMISC;
4610 
4611 	/* Add PF to all active pools */
4612 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4613 		u32 reg_offset = IXGBE_VLVFB(i * 2 + VMDQ_P(0) / 32);
4614 		u32 vlvfb = IXGBE_READ_REG(hw, reg_offset);
4615 
4616 		vlvfb |= BIT(VMDQ_P(0) % 32);
4617 		IXGBE_WRITE_REG(hw, reg_offset, vlvfb);
4618 	}
4619 
4620 	/* Set all bits in the VLAN filter table array */
4621 	for (i = hw->mac.vft_size; i--;)
4622 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), ~0U);
4623 }
4624 
4625 #define VFTA_BLOCK_SIZE 8
4626 static void ixgbe_scrub_vfta(struct ixgbe_adapter *adapter, u32 vfta_offset)
4627 {
4628 	struct ixgbe_hw *hw = &adapter->hw;
4629 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4630 	u32 vid_start = vfta_offset * 32;
4631 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4632 	u32 i, vid, word, bits;
4633 
4634 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4635 		u32 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
4636 
4637 		/* pull VLAN ID from VLVF */
4638 		vid = vlvf & VLAN_VID_MASK;
4639 
4640 		/* only concern outselves with a certain range */
4641 		if (vid < vid_start || vid >= vid_end)
4642 			continue;
4643 
4644 		if (vlvf) {
4645 			/* record VLAN ID in VFTA */
4646 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4647 
4648 			/* if PF is part of this then continue */
4649 			if (test_bit(vid, adapter->active_vlans))
4650 				continue;
4651 		}
4652 
4653 		/* remove PF from the pool */
4654 		word = i * 2 + VMDQ_P(0) / 32;
4655 		bits = ~BIT(VMDQ_P(0) % 32);
4656 		bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4657 		IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), bits);
4658 	}
4659 
4660 	/* extract values from active_vlans and write back to VFTA */
4661 	for (i = VFTA_BLOCK_SIZE; i--;) {
4662 		vid = (vfta_offset + i) * 32;
4663 		word = vid / BITS_PER_LONG;
4664 		bits = vid % BITS_PER_LONG;
4665 
4666 		vfta[i] |= adapter->active_vlans[word] >> bits;
4667 
4668 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(vfta_offset + i), vfta[i]);
4669 	}
4670 }
4671 
4672 static void ixgbe_vlan_promisc_disable(struct ixgbe_adapter *adapter)
4673 {
4674 	struct ixgbe_hw *hw = &adapter->hw;
4675 	u32 vlnctrl, i;
4676 
4677 	/* Set VLAN filtering to enabled */
4678 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4679 	vlnctrl |= IXGBE_VLNCTRL_VFE;
4680 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4681 
4682 	if (!(adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) ||
4683 	    hw->mac.type == ixgbe_mac_82598EB)
4684 		return;
4685 
4686 	/* We are not in VLAN promisc, nothing to do */
4687 	if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4688 		return;
4689 
4690 	/* Set flag so we don't redo unnecessary work */
4691 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4692 
4693 	for (i = 0; i < hw->mac.vft_size; i += VFTA_BLOCK_SIZE)
4694 		ixgbe_scrub_vfta(adapter, i);
4695 }
4696 
4697 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
4698 {
4699 	u16 vid = 1;
4700 
4701 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
4702 
4703 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
4704 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
4705 }
4706 
4707 /**
4708  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4709  * @netdev: network interface device structure
4710  *
4711  * Writes multicast address list to the MTA hash table.
4712  * Returns: -ENOMEM on failure
4713  *                0 on no addresses written
4714  *                X on writing X addresses to MTA
4715  **/
4716 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
4717 {
4718 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4719 	struct ixgbe_hw *hw = &adapter->hw;
4720 
4721 	if (!netif_running(netdev))
4722 		return 0;
4723 
4724 	if (hw->mac.ops.update_mc_addr_list)
4725 		hw->mac.ops.update_mc_addr_list(hw, netdev);
4726 	else
4727 		return -ENOMEM;
4728 
4729 #ifdef CONFIG_PCI_IOV
4730 	ixgbe_restore_vf_multicasts(adapter);
4731 #endif
4732 
4733 	return netdev_mc_count(netdev);
4734 }
4735 
4736 #ifdef CONFIG_PCI_IOV
4737 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4738 {
4739 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4740 	struct ixgbe_hw *hw = &adapter->hw;
4741 	int i;
4742 
4743 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4744 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4745 
4746 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4747 			hw->mac.ops.set_rar(hw, i,
4748 					    mac_table->addr,
4749 					    mac_table->pool,
4750 					    IXGBE_RAH_AV);
4751 		else
4752 			hw->mac.ops.clear_rar(hw, i);
4753 	}
4754 }
4755 
4756 #endif
4757 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4758 {
4759 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4760 	struct ixgbe_hw *hw = &adapter->hw;
4761 	int i;
4762 
4763 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4764 		if (!(mac_table->state & IXGBE_MAC_STATE_MODIFIED))
4765 			continue;
4766 
4767 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4768 
4769 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4770 			hw->mac.ops.set_rar(hw, i,
4771 					    mac_table->addr,
4772 					    mac_table->pool,
4773 					    IXGBE_RAH_AV);
4774 		else
4775 			hw->mac.ops.clear_rar(hw, i);
4776 	}
4777 }
4778 
4779 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4780 {
4781 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4782 	struct ixgbe_hw *hw = &adapter->hw;
4783 	int i;
4784 
4785 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4786 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4787 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4788 	}
4789 
4790 	ixgbe_sync_mac_table(adapter);
4791 }
4792 
4793 static int ixgbe_available_rars(struct ixgbe_adapter *adapter, u16 pool)
4794 {
4795 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4796 	struct ixgbe_hw *hw = &adapter->hw;
4797 	int i, count = 0;
4798 
4799 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4800 		/* do not count default RAR as available */
4801 		if (mac_table->state & IXGBE_MAC_STATE_DEFAULT)
4802 			continue;
4803 
4804 		/* only count unused and addresses that belong to us */
4805 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE) {
4806 			if (mac_table->pool != pool)
4807 				continue;
4808 		}
4809 
4810 		count++;
4811 	}
4812 
4813 	return count;
4814 }
4815 
4816 /* this function destroys the first RAR entry */
4817 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter)
4818 {
4819 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4820 	struct ixgbe_hw *hw = &adapter->hw;
4821 
4822 	memcpy(&mac_table->addr, hw->mac.addr, ETH_ALEN);
4823 	mac_table->pool = VMDQ_P(0);
4824 
4825 	mac_table->state = IXGBE_MAC_STATE_DEFAULT | IXGBE_MAC_STATE_IN_USE;
4826 
4827 	hw->mac.ops.set_rar(hw, 0, mac_table->addr, mac_table->pool,
4828 			    IXGBE_RAH_AV);
4829 }
4830 
4831 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter,
4832 			 const u8 *addr, u16 pool)
4833 {
4834 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4835 	struct ixgbe_hw *hw = &adapter->hw;
4836 	int i;
4837 
4838 	if (is_zero_ether_addr(addr))
4839 		return -EINVAL;
4840 
4841 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4842 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4843 			continue;
4844 
4845 		ether_addr_copy(mac_table->addr, addr);
4846 		mac_table->pool = pool;
4847 
4848 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED |
4849 				    IXGBE_MAC_STATE_IN_USE;
4850 
4851 		ixgbe_sync_mac_table(adapter);
4852 
4853 		return i;
4854 	}
4855 
4856 	return -ENOMEM;
4857 }
4858 
4859 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter,
4860 			 const u8 *addr, u16 pool)
4861 {
4862 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4863 	struct ixgbe_hw *hw = &adapter->hw;
4864 	int i;
4865 
4866 	if (is_zero_ether_addr(addr))
4867 		return -EINVAL;
4868 
4869 	/* search table for addr, if found clear IN_USE flag and sync */
4870 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4871 		/* we can only delete an entry if it is in use */
4872 		if (!(mac_table->state & IXGBE_MAC_STATE_IN_USE))
4873 			continue;
4874 		/* we only care about entries that belong to the given pool */
4875 		if (mac_table->pool != pool)
4876 			continue;
4877 		/* we only care about a specific MAC address */
4878 		if (!ether_addr_equal(addr, mac_table->addr))
4879 			continue;
4880 
4881 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4882 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4883 
4884 		ixgbe_sync_mac_table(adapter);
4885 
4886 		return 0;
4887 	}
4888 
4889 	return -ENOMEM;
4890 }
4891 
4892 static int ixgbe_uc_sync(struct net_device *netdev, const unsigned char *addr)
4893 {
4894 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4895 	int ret;
4896 
4897 	ret = ixgbe_add_mac_filter(adapter, addr, VMDQ_P(0));
4898 
4899 	return min_t(int, ret, 0);
4900 }
4901 
4902 static int ixgbe_uc_unsync(struct net_device *netdev, const unsigned char *addr)
4903 {
4904 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4905 
4906 	ixgbe_del_mac_filter(adapter, addr, VMDQ_P(0));
4907 
4908 	return 0;
4909 }
4910 
4911 /**
4912  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4913  * @netdev: network interface device structure
4914  *
4915  * The set_rx_method entry point is called whenever the unicast/multicast
4916  * address list or the network interface flags are updated.  This routine is
4917  * responsible for configuring the hardware for proper unicast, multicast and
4918  * promiscuous mode.
4919  **/
4920 void ixgbe_set_rx_mode(struct net_device *netdev)
4921 {
4922 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4923 	struct ixgbe_hw *hw = &adapter->hw;
4924 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4925 	netdev_features_t features = netdev->features;
4926 	int count;
4927 
4928 	/* Check for Promiscuous and All Multicast modes */
4929 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4930 
4931 	/* set all bits that we expect to always be set */
4932 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4933 	fctrl |= IXGBE_FCTRL_BAM;
4934 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4935 	fctrl |= IXGBE_FCTRL_PMCF;
4936 
4937 	/* clear the bits we are changing the status of */
4938 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4939 	if (netdev->flags & IFF_PROMISC) {
4940 		hw->addr_ctrl.user_set_promisc = true;
4941 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4942 		vmolr |= IXGBE_VMOLR_MPE;
4943 		features &= ~NETIF_F_HW_VLAN_CTAG_FILTER;
4944 	} else {
4945 		if (netdev->flags & IFF_ALLMULTI) {
4946 			fctrl |= IXGBE_FCTRL_MPE;
4947 			vmolr |= IXGBE_VMOLR_MPE;
4948 		}
4949 		hw->addr_ctrl.user_set_promisc = false;
4950 	}
4951 
4952 	/*
4953 	 * Write addresses to available RAR registers, if there is not
4954 	 * sufficient space to store all the addresses then enable
4955 	 * unicast promiscuous mode
4956 	 */
4957 	if (__dev_uc_sync(netdev, ixgbe_uc_sync, ixgbe_uc_unsync)) {
4958 		fctrl |= IXGBE_FCTRL_UPE;
4959 		vmolr |= IXGBE_VMOLR_ROPE;
4960 	}
4961 
4962 	/* Write addresses to the MTA, if the attempt fails
4963 	 * then we should just turn on promiscuous mode so
4964 	 * that we can at least receive multicast traffic
4965 	 */
4966 	count = ixgbe_write_mc_addr_list(netdev);
4967 	if (count < 0) {
4968 		fctrl |= IXGBE_FCTRL_MPE;
4969 		vmolr |= IXGBE_VMOLR_MPE;
4970 	} else if (count) {
4971 		vmolr |= IXGBE_VMOLR_ROMPE;
4972 	}
4973 
4974 	if (hw->mac.type != ixgbe_mac_82598EB) {
4975 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4976 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4977 			   IXGBE_VMOLR_ROPE);
4978 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4979 	}
4980 
4981 	/* This is useful for sniffing bad packets. */
4982 	if (features & NETIF_F_RXALL) {
4983 		/* UPE and MPE will be handled by normal PROMISC logic
4984 		 * in e1000e_set_rx_mode */
4985 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4986 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4987 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4988 
4989 		fctrl &= ~(IXGBE_FCTRL_DPF);
4990 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
4991 	}
4992 
4993 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4994 
4995 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
4996 		ixgbe_vlan_strip_enable(adapter);
4997 	else
4998 		ixgbe_vlan_strip_disable(adapter);
4999 
5000 	if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
5001 		ixgbe_vlan_promisc_disable(adapter);
5002 	else
5003 		ixgbe_vlan_promisc_enable(adapter);
5004 }
5005 
5006 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
5007 {
5008 	int q_idx;
5009 
5010 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
5011 		napi_enable(&adapter->q_vector[q_idx]->napi);
5012 }
5013 
5014 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
5015 {
5016 	int q_idx;
5017 
5018 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
5019 		napi_disable(&adapter->q_vector[q_idx]->napi);
5020 }
5021 
5022 static int ixgbe_udp_tunnel_sync(struct net_device *dev, unsigned int table)
5023 {
5024 	struct ixgbe_adapter *adapter = netdev_priv(dev);
5025 	struct ixgbe_hw *hw = &adapter->hw;
5026 	struct udp_tunnel_info ti;
5027 
5028 	udp_tunnel_nic_get_port(dev, table, 0, &ti);
5029 	if (ti.type == UDP_TUNNEL_TYPE_VXLAN)
5030 		adapter->vxlan_port = ti.port;
5031 	else
5032 		adapter->geneve_port = ti.port;
5033 
5034 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL,
5035 			ntohs(adapter->vxlan_port) |
5036 			ntohs(adapter->geneve_port) <<
5037 				IXGBE_VXLANCTRL_GENEVE_UDPPORT_SHIFT);
5038 	return 0;
5039 }
5040 
5041 static const struct udp_tunnel_nic_info ixgbe_udp_tunnels_x550 = {
5042 	.sync_table	= ixgbe_udp_tunnel_sync,
5043 	.flags		= UDP_TUNNEL_NIC_INFO_IPV4_ONLY,
5044 	.tables		= {
5045 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN,  },
5046 	},
5047 };
5048 
5049 static const struct udp_tunnel_nic_info ixgbe_udp_tunnels_x550em_a = {
5050 	.sync_table	= ixgbe_udp_tunnel_sync,
5051 	.flags		= UDP_TUNNEL_NIC_INFO_IPV4_ONLY,
5052 	.tables		= {
5053 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN,  },
5054 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_GENEVE, },
5055 	},
5056 };
5057 
5058 #ifdef CONFIG_IXGBE_DCB
5059 /**
5060  * ixgbe_configure_dcb - Configure DCB hardware
5061  * @adapter: ixgbe adapter struct
5062  *
5063  * This is called by the driver on open to configure the DCB hardware.
5064  * This is also called by the gennetlink interface when reconfiguring
5065  * the DCB state.
5066  */
5067 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
5068 {
5069 	struct ixgbe_hw *hw = &adapter->hw;
5070 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
5071 
5072 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
5073 		if (hw->mac.type == ixgbe_mac_82598EB)
5074 			netif_set_tso_max_size(adapter->netdev, 65536);
5075 		return;
5076 	}
5077 
5078 	if (hw->mac.type == ixgbe_mac_82598EB)
5079 		netif_set_tso_max_size(adapter->netdev, 32768);
5080 
5081 #ifdef IXGBE_FCOE
5082 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
5083 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
5084 #endif
5085 
5086 	/* reconfigure the hardware */
5087 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
5088 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5089 						DCB_TX_CONFIG);
5090 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5091 						DCB_RX_CONFIG);
5092 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
5093 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
5094 		ixgbe_dcb_hw_ets(&adapter->hw,
5095 				 adapter->ixgbe_ieee_ets,
5096 				 max_frame);
5097 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
5098 					adapter->ixgbe_ieee_pfc->pfc_en,
5099 					adapter->ixgbe_ieee_ets->prio_tc);
5100 	}
5101 
5102 	/* Enable RSS Hash per TC */
5103 	if (hw->mac.type != ixgbe_mac_82598EB) {
5104 		u32 msb = 0;
5105 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
5106 
5107 		while (rss_i) {
5108 			msb++;
5109 			rss_i >>= 1;
5110 		}
5111 
5112 		/* write msb to all 8 TCs in one write */
5113 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
5114 	}
5115 }
5116 #endif
5117 
5118 /* Additional bittime to account for IXGBE framing */
5119 #define IXGBE_ETH_FRAMING 20
5120 
5121 /**
5122  * ixgbe_hpbthresh - calculate high water mark for flow control
5123  *
5124  * @adapter: board private structure to calculate for
5125  * @pb: packet buffer to calculate
5126  */
5127 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
5128 {
5129 	struct ixgbe_hw *hw = &adapter->hw;
5130 	struct net_device *dev = adapter->netdev;
5131 	int link, tc, kb, marker;
5132 	u32 dv_id, rx_pba;
5133 
5134 	/* Calculate max LAN frame size */
5135 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
5136 
5137 #ifdef IXGBE_FCOE
5138 	/* FCoE traffic class uses FCOE jumbo frames */
5139 	if ((dev->features & NETIF_F_FCOE_MTU) &&
5140 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5141 	    (pb == ixgbe_fcoe_get_tc(adapter)))
5142 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5143 #endif
5144 
5145 	/* Calculate delay value for device */
5146 	switch (hw->mac.type) {
5147 	case ixgbe_mac_X540:
5148 	case ixgbe_mac_X550:
5149 	case ixgbe_mac_X550EM_x:
5150 	case ixgbe_mac_x550em_a:
5151 		dv_id = IXGBE_DV_X540(link, tc);
5152 		break;
5153 	default:
5154 		dv_id = IXGBE_DV(link, tc);
5155 		break;
5156 	}
5157 
5158 	/* Loopback switch introduces additional latency */
5159 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5160 		dv_id += IXGBE_B2BT(tc);
5161 
5162 	/* Delay value is calculated in bit times convert to KB */
5163 	kb = IXGBE_BT2KB(dv_id);
5164 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
5165 
5166 	marker = rx_pba - kb;
5167 
5168 	/* It is possible that the packet buffer is not large enough
5169 	 * to provide required headroom. In this case throw an error
5170 	 * to user and a do the best we can.
5171 	 */
5172 	if (marker < 0) {
5173 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
5174 			    "headroom to support flow control."
5175 			    "Decrease MTU or number of traffic classes\n", pb);
5176 		marker = tc + 1;
5177 	}
5178 
5179 	return marker;
5180 }
5181 
5182 /**
5183  * ixgbe_lpbthresh - calculate low water mark for flow control
5184  *
5185  * @adapter: board private structure to calculate for
5186  * @pb: packet buffer to calculate
5187  */
5188 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
5189 {
5190 	struct ixgbe_hw *hw = &adapter->hw;
5191 	struct net_device *dev = adapter->netdev;
5192 	int tc;
5193 	u32 dv_id;
5194 
5195 	/* Calculate max LAN frame size */
5196 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
5197 
5198 #ifdef IXGBE_FCOE
5199 	/* FCoE traffic class uses FCOE jumbo frames */
5200 	if ((dev->features & NETIF_F_FCOE_MTU) &&
5201 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5202 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
5203 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5204 #endif
5205 
5206 	/* Calculate delay value for device */
5207 	switch (hw->mac.type) {
5208 	case ixgbe_mac_X540:
5209 	case ixgbe_mac_X550:
5210 	case ixgbe_mac_X550EM_x:
5211 	case ixgbe_mac_x550em_a:
5212 		dv_id = IXGBE_LOW_DV_X540(tc);
5213 		break;
5214 	default:
5215 		dv_id = IXGBE_LOW_DV(tc);
5216 		break;
5217 	}
5218 
5219 	/* Delay value is calculated in bit times convert to KB */
5220 	return IXGBE_BT2KB(dv_id);
5221 }
5222 
5223 /*
5224  * ixgbe_pbthresh_setup - calculate and setup high low water marks
5225  */
5226 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
5227 {
5228 	struct ixgbe_hw *hw = &adapter->hw;
5229 	int num_tc = adapter->hw_tcs;
5230 	int i;
5231 
5232 	if (!num_tc)
5233 		num_tc = 1;
5234 
5235 	for (i = 0; i < num_tc; i++) {
5236 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
5237 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
5238 
5239 		/* Low water marks must not be larger than high water marks */
5240 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
5241 			hw->fc.low_water[i] = 0;
5242 	}
5243 
5244 	for (; i < MAX_TRAFFIC_CLASS; i++)
5245 		hw->fc.high_water[i] = 0;
5246 }
5247 
5248 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
5249 {
5250 	struct ixgbe_hw *hw = &adapter->hw;
5251 	int hdrm;
5252 	u8 tc = adapter->hw_tcs;
5253 
5254 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
5255 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
5256 		hdrm = 32 << adapter->fdir_pballoc;
5257 	else
5258 		hdrm = 0;
5259 
5260 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
5261 	ixgbe_pbthresh_setup(adapter);
5262 }
5263 
5264 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
5265 {
5266 	struct ixgbe_hw *hw = &adapter->hw;
5267 	struct hlist_node *node2;
5268 	struct ixgbe_fdir_filter *filter;
5269 	u8 queue;
5270 
5271 	spin_lock(&adapter->fdir_perfect_lock);
5272 
5273 	if (!hlist_empty(&adapter->fdir_filter_list))
5274 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
5275 
5276 	hlist_for_each_entry_safe(filter, node2,
5277 				  &adapter->fdir_filter_list, fdir_node) {
5278 		if (filter->action == IXGBE_FDIR_DROP_QUEUE) {
5279 			queue = IXGBE_FDIR_DROP_QUEUE;
5280 		} else {
5281 			u32 ring = ethtool_get_flow_spec_ring(filter->action);
5282 			u8 vf = ethtool_get_flow_spec_ring_vf(filter->action);
5283 
5284 			if (!vf && (ring >= adapter->num_rx_queues)) {
5285 				e_err(drv, "FDIR restore failed without VF, ring: %u\n",
5286 				      ring);
5287 				continue;
5288 			} else if (vf &&
5289 				   ((vf > adapter->num_vfs) ||
5290 				     ring >= adapter->num_rx_queues_per_pool)) {
5291 				e_err(drv, "FDIR restore failed with VF, vf: %hhu, ring: %u\n",
5292 				      vf, ring);
5293 				continue;
5294 			}
5295 
5296 			/* Map the ring onto the absolute queue index */
5297 			if (!vf)
5298 				queue = adapter->rx_ring[ring]->reg_idx;
5299 			else
5300 				queue = ((vf - 1) *
5301 					adapter->num_rx_queues_per_pool) + ring;
5302 		}
5303 
5304 		ixgbe_fdir_write_perfect_filter_82599(hw,
5305 				&filter->filter, filter->sw_idx, queue);
5306 	}
5307 
5308 	spin_unlock(&adapter->fdir_perfect_lock);
5309 }
5310 
5311 /**
5312  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
5313  * @rx_ring: ring to free buffers from
5314  **/
5315 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
5316 {
5317 	u16 i = rx_ring->next_to_clean;
5318 	struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
5319 
5320 	if (rx_ring->xsk_pool) {
5321 		ixgbe_xsk_clean_rx_ring(rx_ring);
5322 		goto skip_free;
5323 	}
5324 
5325 	/* Free all the Rx ring sk_buffs */
5326 	while (i != rx_ring->next_to_alloc) {
5327 		if (rx_buffer->skb) {
5328 			struct sk_buff *skb = rx_buffer->skb;
5329 			if (IXGBE_CB(skb)->page_released)
5330 				dma_unmap_page_attrs(rx_ring->dev,
5331 						     IXGBE_CB(skb)->dma,
5332 						     ixgbe_rx_pg_size(rx_ring),
5333 						     DMA_FROM_DEVICE,
5334 						     IXGBE_RX_DMA_ATTR);
5335 			dev_kfree_skb(skb);
5336 		}
5337 
5338 		/* Invalidate cache lines that may have been written to by
5339 		 * device so that we avoid corrupting memory.
5340 		 */
5341 		dma_sync_single_range_for_cpu(rx_ring->dev,
5342 					      rx_buffer->dma,
5343 					      rx_buffer->page_offset,
5344 					      ixgbe_rx_bufsz(rx_ring),
5345 					      DMA_FROM_DEVICE);
5346 
5347 		/* free resources associated with mapping */
5348 		dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
5349 				     ixgbe_rx_pg_size(rx_ring),
5350 				     DMA_FROM_DEVICE,
5351 				     IXGBE_RX_DMA_ATTR);
5352 		__page_frag_cache_drain(rx_buffer->page,
5353 					rx_buffer->pagecnt_bias);
5354 
5355 		i++;
5356 		rx_buffer++;
5357 		if (i == rx_ring->count) {
5358 			i = 0;
5359 			rx_buffer = rx_ring->rx_buffer_info;
5360 		}
5361 	}
5362 
5363 skip_free:
5364 	rx_ring->next_to_alloc = 0;
5365 	rx_ring->next_to_clean = 0;
5366 	rx_ring->next_to_use = 0;
5367 }
5368 
5369 static int ixgbe_fwd_ring_up(struct ixgbe_adapter *adapter,
5370 			     struct ixgbe_fwd_adapter *accel)
5371 {
5372 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
5373 	int num_tc = netdev_get_num_tc(adapter->netdev);
5374 	struct net_device *vdev = accel->netdev;
5375 	int i, baseq, err;
5376 
5377 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
5378 	netdev_dbg(vdev, "pool %i:%i queues %i:%i\n",
5379 		   accel->pool, adapter->num_rx_pools,
5380 		   baseq, baseq + adapter->num_rx_queues_per_pool);
5381 
5382 	accel->rx_base_queue = baseq;
5383 	accel->tx_base_queue = baseq;
5384 
5385 	/* record configuration for macvlan interface in vdev */
5386 	for (i = 0; i < num_tc; i++)
5387 		netdev_bind_sb_channel_queue(adapter->netdev, vdev,
5388 					     i, rss_i, baseq + (rss_i * i));
5389 
5390 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5391 		adapter->rx_ring[baseq + i]->netdev = vdev;
5392 
5393 	/* Guarantee all rings are updated before we update the
5394 	 * MAC address filter.
5395 	 */
5396 	wmb();
5397 
5398 	/* ixgbe_add_mac_filter will return an index if it succeeds, so we
5399 	 * need to only treat it as an error value if it is negative.
5400 	 */
5401 	err = ixgbe_add_mac_filter(adapter, vdev->dev_addr,
5402 				   VMDQ_P(accel->pool));
5403 	if (err >= 0)
5404 		return 0;
5405 
5406 	/* if we cannot add the MAC rule then disable the offload */
5407 	macvlan_release_l2fw_offload(vdev);
5408 
5409 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5410 		adapter->rx_ring[baseq + i]->netdev = NULL;
5411 
5412 	netdev_err(vdev, "L2FW offload disabled due to L2 filter error\n");
5413 
5414 	/* unbind the queues and drop the subordinate channel config */
5415 	netdev_unbind_sb_channel(adapter->netdev, vdev);
5416 	netdev_set_sb_channel(vdev, 0);
5417 
5418 	clear_bit(accel->pool, adapter->fwd_bitmask);
5419 	kfree(accel);
5420 
5421 	return err;
5422 }
5423 
5424 static int ixgbe_macvlan_up(struct net_device *vdev,
5425 			    struct netdev_nested_priv *priv)
5426 {
5427 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)priv->data;
5428 	struct ixgbe_fwd_adapter *accel;
5429 
5430 	if (!netif_is_macvlan(vdev))
5431 		return 0;
5432 
5433 	accel = macvlan_accel_priv(vdev);
5434 	if (!accel)
5435 		return 0;
5436 
5437 	ixgbe_fwd_ring_up(adapter, accel);
5438 
5439 	return 0;
5440 }
5441 
5442 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
5443 {
5444 	struct netdev_nested_priv priv = {
5445 		.data = (void *)adapter,
5446 	};
5447 
5448 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
5449 				      ixgbe_macvlan_up, &priv);
5450 }
5451 
5452 static void ixgbe_configure(struct ixgbe_adapter *adapter)
5453 {
5454 	struct ixgbe_hw *hw = &adapter->hw;
5455 
5456 	ixgbe_configure_pb(adapter);
5457 #ifdef CONFIG_IXGBE_DCB
5458 	ixgbe_configure_dcb(adapter);
5459 #endif
5460 	/*
5461 	 * We must restore virtualization before VLANs or else
5462 	 * the VLVF registers will not be populated
5463 	 */
5464 	ixgbe_configure_virtualization(adapter);
5465 
5466 	ixgbe_set_rx_mode(adapter->netdev);
5467 	ixgbe_restore_vlan(adapter);
5468 	ixgbe_ipsec_restore(adapter);
5469 
5470 	switch (hw->mac.type) {
5471 	case ixgbe_mac_82599EB:
5472 	case ixgbe_mac_X540:
5473 		hw->mac.ops.disable_rx_buff(hw);
5474 		break;
5475 	default:
5476 		break;
5477 	}
5478 
5479 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
5480 		ixgbe_init_fdir_signature_82599(&adapter->hw,
5481 						adapter->fdir_pballoc);
5482 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
5483 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
5484 					      adapter->fdir_pballoc);
5485 		ixgbe_fdir_filter_restore(adapter);
5486 	}
5487 
5488 	switch (hw->mac.type) {
5489 	case ixgbe_mac_82599EB:
5490 	case ixgbe_mac_X540:
5491 		hw->mac.ops.enable_rx_buff(hw);
5492 		break;
5493 	default:
5494 		break;
5495 	}
5496 
5497 #ifdef CONFIG_IXGBE_DCA
5498 	/* configure DCA */
5499 	if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
5500 		ixgbe_setup_dca(adapter);
5501 #endif /* CONFIG_IXGBE_DCA */
5502 
5503 #ifdef IXGBE_FCOE
5504 	/* configure FCoE L2 filters, redirection table, and Rx control */
5505 	ixgbe_configure_fcoe(adapter);
5506 
5507 #endif /* IXGBE_FCOE */
5508 	ixgbe_configure_tx(adapter);
5509 	ixgbe_configure_rx(adapter);
5510 	ixgbe_configure_dfwd(adapter);
5511 }
5512 
5513 /**
5514  * ixgbe_sfp_link_config - set up SFP+ link
5515  * @adapter: pointer to private adapter struct
5516  **/
5517 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
5518 {
5519 	/*
5520 	 * We are assuming the worst case scenario here, and that
5521 	 * is that an SFP was inserted/removed after the reset
5522 	 * but before SFP detection was enabled.  As such the best
5523 	 * solution is to just start searching as soon as we start
5524 	 */
5525 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5526 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5527 
5528 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5529 	adapter->sfp_poll_time = 0;
5530 }
5531 
5532 /**
5533  * ixgbe_non_sfp_link_config - set up non-SFP+ link
5534  * @hw: pointer to private hardware struct
5535  *
5536  * Returns 0 on success, negative on failure
5537  **/
5538 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
5539 {
5540 	u32 speed;
5541 	bool autoneg, link_up = false;
5542 	int ret = -EIO;
5543 
5544 	if (hw->mac.ops.check_link)
5545 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
5546 
5547 	if (ret)
5548 		return ret;
5549 
5550 	speed = hw->phy.autoneg_advertised;
5551 	if (!speed && hw->mac.ops.get_link_capabilities) {
5552 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
5553 							&autoneg);
5554 		/* remove NBASE-T speeds from default autonegotiation
5555 		 * to accommodate broken network switches in the field
5556 		 * which cannot cope with advertised NBASE-T speeds
5557 		 */
5558 		speed &= ~(IXGBE_LINK_SPEED_5GB_FULL |
5559 			   IXGBE_LINK_SPEED_2_5GB_FULL);
5560 	}
5561 
5562 	if (ret)
5563 		return ret;
5564 
5565 	if (hw->mac.ops.setup_link)
5566 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
5567 
5568 	return ret;
5569 }
5570 
5571 /**
5572  * ixgbe_clear_vf_stats_counters - Clear out VF stats after reset
5573  * @adapter: board private structure
5574  *
5575  * On a reset we need to clear out the VF stats or accounting gets
5576  * messed up because they're not clear on read.
5577  **/
5578 static void ixgbe_clear_vf_stats_counters(struct ixgbe_adapter *adapter)
5579 {
5580 	struct ixgbe_hw *hw = &adapter->hw;
5581 	int i;
5582 
5583 	for (i = 0; i < adapter->num_vfs; i++) {
5584 		adapter->vfinfo[i].last_vfstats.gprc =
5585 			IXGBE_READ_REG(hw, IXGBE_PVFGPRC(i));
5586 		adapter->vfinfo[i].saved_rst_vfstats.gprc +=
5587 			adapter->vfinfo[i].vfstats.gprc;
5588 		adapter->vfinfo[i].vfstats.gprc = 0;
5589 		adapter->vfinfo[i].last_vfstats.gptc =
5590 			IXGBE_READ_REG(hw, IXGBE_PVFGPTC(i));
5591 		adapter->vfinfo[i].saved_rst_vfstats.gptc +=
5592 			adapter->vfinfo[i].vfstats.gptc;
5593 		adapter->vfinfo[i].vfstats.gptc = 0;
5594 		adapter->vfinfo[i].last_vfstats.gorc =
5595 			IXGBE_READ_REG(hw, IXGBE_PVFGORC_LSB(i));
5596 		adapter->vfinfo[i].saved_rst_vfstats.gorc +=
5597 			adapter->vfinfo[i].vfstats.gorc;
5598 		adapter->vfinfo[i].vfstats.gorc = 0;
5599 		adapter->vfinfo[i].last_vfstats.gotc =
5600 			IXGBE_READ_REG(hw, IXGBE_PVFGOTC_LSB(i));
5601 		adapter->vfinfo[i].saved_rst_vfstats.gotc +=
5602 			adapter->vfinfo[i].vfstats.gotc;
5603 		adapter->vfinfo[i].vfstats.gotc = 0;
5604 		adapter->vfinfo[i].last_vfstats.mprc =
5605 			IXGBE_READ_REG(hw, IXGBE_PVFMPRC(i));
5606 		adapter->vfinfo[i].saved_rst_vfstats.mprc +=
5607 			adapter->vfinfo[i].vfstats.mprc;
5608 		adapter->vfinfo[i].vfstats.mprc = 0;
5609 	}
5610 }
5611 
5612 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
5613 {
5614 	struct ixgbe_hw *hw = &adapter->hw;
5615 	u32 gpie = 0;
5616 
5617 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
5618 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5619 		       IXGBE_GPIE_OCD;
5620 		gpie |= IXGBE_GPIE_EIAME;
5621 		/*
5622 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
5623 		 * this saves a register write for every interrupt
5624 		 */
5625 		switch (hw->mac.type) {
5626 		case ixgbe_mac_82598EB:
5627 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5628 			break;
5629 		case ixgbe_mac_82599EB:
5630 		case ixgbe_mac_X540:
5631 		case ixgbe_mac_X550:
5632 		case ixgbe_mac_X550EM_x:
5633 		case ixgbe_mac_x550em_a:
5634 		default:
5635 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5636 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5637 			break;
5638 		}
5639 	} else {
5640 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
5641 		 * specifically only auto mask tx and rx interrupts */
5642 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5643 	}
5644 
5645 	/* XXX: to interrupt immediately for EICS writes, enable this */
5646 	/* gpie |= IXGBE_GPIE_EIMEN; */
5647 
5648 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
5649 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
5650 
5651 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
5652 		case IXGBE_82599_VMDQ_8Q_MASK:
5653 			gpie |= IXGBE_GPIE_VTMODE_16;
5654 			break;
5655 		case IXGBE_82599_VMDQ_4Q_MASK:
5656 			gpie |= IXGBE_GPIE_VTMODE_32;
5657 			break;
5658 		default:
5659 			gpie |= IXGBE_GPIE_VTMODE_64;
5660 			break;
5661 		}
5662 	}
5663 
5664 	/* Enable Thermal over heat sensor interrupt */
5665 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
5666 		switch (adapter->hw.mac.type) {
5667 		case ixgbe_mac_82599EB:
5668 			gpie |= IXGBE_SDP0_GPIEN_8259X;
5669 			break;
5670 		default:
5671 			break;
5672 		}
5673 	}
5674 
5675 	/* Enable fan failure interrupt */
5676 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
5677 		gpie |= IXGBE_SDP1_GPIEN(hw);
5678 
5679 	switch (hw->mac.type) {
5680 	case ixgbe_mac_82599EB:
5681 		gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
5682 		break;
5683 	case ixgbe_mac_X550EM_x:
5684 	case ixgbe_mac_x550em_a:
5685 		gpie |= IXGBE_SDP0_GPIEN_X540;
5686 		break;
5687 	default:
5688 		break;
5689 	}
5690 
5691 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5692 }
5693 
5694 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
5695 {
5696 	struct ixgbe_hw *hw = &adapter->hw;
5697 	int err;
5698 	u32 ctrl_ext;
5699 
5700 	ixgbe_get_hw_control(adapter);
5701 	ixgbe_setup_gpie(adapter);
5702 
5703 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
5704 		ixgbe_configure_msix(adapter);
5705 	else
5706 		ixgbe_configure_msi_and_legacy(adapter);
5707 
5708 	/* enable the optics for 82599 SFP+ fiber */
5709 	if (hw->mac.ops.enable_tx_laser)
5710 		hw->mac.ops.enable_tx_laser(hw);
5711 
5712 	if (hw->phy.ops.set_phy_power)
5713 		hw->phy.ops.set_phy_power(hw, true);
5714 
5715 	smp_mb__before_atomic();
5716 	clear_bit(__IXGBE_DOWN, &adapter->state);
5717 	ixgbe_napi_enable_all(adapter);
5718 
5719 	if (ixgbe_is_sfp(hw)) {
5720 		ixgbe_sfp_link_config(adapter);
5721 	} else {
5722 		err = ixgbe_non_sfp_link_config(hw);
5723 		if (err)
5724 			e_err(probe, "link_config FAILED %d\n", err);
5725 	}
5726 
5727 	/* clear any pending interrupts, may auto mask */
5728 	IXGBE_READ_REG(hw, IXGBE_EICR);
5729 	ixgbe_irq_enable(adapter, true, true);
5730 
5731 	/*
5732 	 * If this adapter has a fan, check to see if we had a failure
5733 	 * before we enabled the interrupt.
5734 	 */
5735 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
5736 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
5737 		if (esdp & IXGBE_ESDP_SDP1)
5738 			e_crit(drv, "Fan has stopped, replace the adapter\n");
5739 	}
5740 
5741 	/* bring the link up in the watchdog, this could race with our first
5742 	 * link up interrupt but shouldn't be a problem */
5743 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5744 	adapter->link_check_timeout = jiffies;
5745 	mod_timer(&adapter->service_timer, jiffies);
5746 
5747 	ixgbe_clear_vf_stats_counters(adapter);
5748 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
5749 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
5750 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
5751 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
5752 
5753 	/* update setting rx tx for all active vfs */
5754 	ixgbe_set_all_vfs(adapter);
5755 }
5756 
5757 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
5758 {
5759 	/* put off any impending NetWatchDogTimeout */
5760 	netif_trans_update(adapter->netdev);
5761 
5762 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
5763 		usleep_range(1000, 2000);
5764 	if (adapter->hw.phy.type == ixgbe_phy_fw)
5765 		ixgbe_watchdog_link_is_down(adapter);
5766 	ixgbe_down(adapter);
5767 	/*
5768 	 * If SR-IOV enabled then wait a bit before bringing the adapter
5769 	 * back up to give the VFs time to respond to the reset.  The
5770 	 * two second wait is based upon the watchdog timer cycle in
5771 	 * the VF driver.
5772 	 */
5773 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5774 		msleep(2000);
5775 	ixgbe_up(adapter);
5776 	clear_bit(__IXGBE_RESETTING, &adapter->state);
5777 }
5778 
5779 void ixgbe_up(struct ixgbe_adapter *adapter)
5780 {
5781 	/* hardware has been reset, we need to reload some things */
5782 	ixgbe_configure(adapter);
5783 
5784 	ixgbe_up_complete(adapter);
5785 }
5786 
5787 static unsigned long ixgbe_get_completion_timeout(struct ixgbe_adapter *adapter)
5788 {
5789 	u16 devctl2;
5790 
5791 	pcie_capability_read_word(adapter->pdev, PCI_EXP_DEVCTL2, &devctl2);
5792 
5793 	switch (devctl2 & IXGBE_PCIDEVCTRL2_TIMEO_MASK) {
5794 	case IXGBE_PCIDEVCTRL2_17_34s:
5795 	case IXGBE_PCIDEVCTRL2_4_8s:
5796 		/* For now we cap the upper limit on delay to 2 seconds
5797 		 * as we end up going up to 34 seconds of delay in worst
5798 		 * case timeout value.
5799 		 */
5800 	case IXGBE_PCIDEVCTRL2_1_2s:
5801 		return 2000000ul;	/* 2.0 s */
5802 	case IXGBE_PCIDEVCTRL2_260_520ms:
5803 		return 520000ul;	/* 520 ms */
5804 	case IXGBE_PCIDEVCTRL2_65_130ms:
5805 		return 130000ul;	/* 130 ms */
5806 	case IXGBE_PCIDEVCTRL2_16_32ms:
5807 		return 32000ul;		/* 32 ms */
5808 	case IXGBE_PCIDEVCTRL2_1_2ms:
5809 		return 2000ul;		/* 2 ms */
5810 	case IXGBE_PCIDEVCTRL2_50_100us:
5811 		return 100ul;		/* 100 us */
5812 	case IXGBE_PCIDEVCTRL2_16_32ms_def:
5813 		return 32000ul;		/* 32 ms */
5814 	default:
5815 		break;
5816 	}
5817 
5818 	/* We shouldn't need to hit this path, but just in case default as
5819 	 * though completion timeout is not supported and support 32ms.
5820 	 */
5821 	return 32000ul;
5822 }
5823 
5824 void ixgbe_disable_rx(struct ixgbe_adapter *adapter)
5825 {
5826 	unsigned long wait_delay, delay_interval;
5827 	struct ixgbe_hw *hw = &adapter->hw;
5828 	int i, wait_loop;
5829 	u32 rxdctl;
5830 
5831 	/* disable receives */
5832 	hw->mac.ops.disable_rx(hw);
5833 
5834 	if (ixgbe_removed(hw->hw_addr))
5835 		return;
5836 
5837 	/* disable all enabled Rx queues */
5838 	for (i = 0; i < adapter->num_rx_queues; i++) {
5839 		struct ixgbe_ring *ring = adapter->rx_ring[i];
5840 		u8 reg_idx = ring->reg_idx;
5841 
5842 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
5843 		rxdctl &= ~IXGBE_RXDCTL_ENABLE;
5844 		rxdctl |= IXGBE_RXDCTL_SWFLSH;
5845 
5846 		/* write value back with RXDCTL.ENABLE bit cleared */
5847 		IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
5848 	}
5849 
5850 	/* RXDCTL.EN may not change on 82598 if link is down, so skip it */
5851 	if (hw->mac.type == ixgbe_mac_82598EB &&
5852 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
5853 		return;
5854 
5855 	/* Determine our minimum delay interval. We will increase this value
5856 	 * with each subsequent test. This way if the device returns quickly
5857 	 * we should spend as little time as possible waiting, however as
5858 	 * the time increases we will wait for larger periods of time.
5859 	 *
5860 	 * The trick here is that we increase the interval using the
5861 	 * following pattern: 1x 3x 5x 7x 9x 11x 13x 15x 17x 19x. The result
5862 	 * of that wait is that it totals up to 100x whatever interval we
5863 	 * choose. Since our minimum wait is 100us we can just divide the
5864 	 * total timeout by 100 to get our minimum delay interval.
5865 	 */
5866 	delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
5867 
5868 	wait_loop = IXGBE_MAX_RX_DESC_POLL;
5869 	wait_delay = delay_interval;
5870 
5871 	while (wait_loop--) {
5872 		usleep_range(wait_delay, wait_delay + 10);
5873 		wait_delay += delay_interval * 2;
5874 		rxdctl = 0;
5875 
5876 		/* OR together the reading of all the active RXDCTL registers,
5877 		 * and then test the result. We need the disable to complete
5878 		 * before we start freeing the memory and invalidating the
5879 		 * DMA mappings.
5880 		 */
5881 		for (i = 0; i < adapter->num_rx_queues; i++) {
5882 			struct ixgbe_ring *ring = adapter->rx_ring[i];
5883 			u8 reg_idx = ring->reg_idx;
5884 
5885 			rxdctl |= IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
5886 		}
5887 
5888 		if (!(rxdctl & IXGBE_RXDCTL_ENABLE))
5889 			return;
5890 	}
5891 
5892 	e_err(drv,
5893 	      "RXDCTL.ENABLE for one or more queues not cleared within the polling period\n");
5894 }
5895 
5896 void ixgbe_disable_tx(struct ixgbe_adapter *adapter)
5897 {
5898 	unsigned long wait_delay, delay_interval;
5899 	struct ixgbe_hw *hw = &adapter->hw;
5900 	int i, wait_loop;
5901 	u32 txdctl;
5902 
5903 	if (ixgbe_removed(hw->hw_addr))
5904 		return;
5905 
5906 	/* disable all enabled Tx queues */
5907 	for (i = 0; i < adapter->num_tx_queues; i++) {
5908 		struct ixgbe_ring *ring = adapter->tx_ring[i];
5909 		u8 reg_idx = ring->reg_idx;
5910 
5911 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5912 	}
5913 
5914 	/* disable all enabled XDP Tx queues */
5915 	for (i = 0; i < adapter->num_xdp_queues; i++) {
5916 		struct ixgbe_ring *ring = adapter->xdp_ring[i];
5917 		u8 reg_idx = ring->reg_idx;
5918 
5919 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5920 	}
5921 
5922 	/* If the link is not up there shouldn't be much in the way of
5923 	 * pending transactions. Those that are left will be flushed out
5924 	 * when the reset logic goes through the flush sequence to clean out
5925 	 * the pending Tx transactions.
5926 	 */
5927 	if (!(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
5928 		goto dma_engine_disable;
5929 
5930 	/* Determine our minimum delay interval. We will increase this value
5931 	 * with each subsequent test. This way if the device returns quickly
5932 	 * we should spend as little time as possible waiting, however as
5933 	 * the time increases we will wait for larger periods of time.
5934 	 *
5935 	 * The trick here is that we increase the interval using the
5936 	 * following pattern: 1x 3x 5x 7x 9x 11x 13x 15x 17x 19x. The result
5937 	 * of that wait is that it totals up to 100x whatever interval we
5938 	 * choose. Since our minimum wait is 100us we can just divide the
5939 	 * total timeout by 100 to get our minimum delay interval.
5940 	 */
5941 	delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
5942 
5943 	wait_loop = IXGBE_MAX_RX_DESC_POLL;
5944 	wait_delay = delay_interval;
5945 
5946 	while (wait_loop--) {
5947 		usleep_range(wait_delay, wait_delay + 10);
5948 		wait_delay += delay_interval * 2;
5949 		txdctl = 0;
5950 
5951 		/* OR together the reading of all the active TXDCTL registers,
5952 		 * and then test the result. We need the disable to complete
5953 		 * before we start freeing the memory and invalidating the
5954 		 * DMA mappings.
5955 		 */
5956 		for (i = 0; i < adapter->num_tx_queues; i++) {
5957 			struct ixgbe_ring *ring = adapter->tx_ring[i];
5958 			u8 reg_idx = ring->reg_idx;
5959 
5960 			txdctl |= IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
5961 		}
5962 		for (i = 0; i < adapter->num_xdp_queues; i++) {
5963 			struct ixgbe_ring *ring = adapter->xdp_ring[i];
5964 			u8 reg_idx = ring->reg_idx;
5965 
5966 			txdctl |= IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
5967 		}
5968 
5969 		if (!(txdctl & IXGBE_TXDCTL_ENABLE))
5970 			goto dma_engine_disable;
5971 	}
5972 
5973 	e_err(drv,
5974 	      "TXDCTL.ENABLE for one or more queues not cleared within the polling period\n");
5975 
5976 dma_engine_disable:
5977 	/* Disable the Tx DMA engine on 82599 and later MAC */
5978 	switch (hw->mac.type) {
5979 	case ixgbe_mac_82599EB:
5980 	case ixgbe_mac_X540:
5981 	case ixgbe_mac_X550:
5982 	case ixgbe_mac_X550EM_x:
5983 	case ixgbe_mac_x550em_a:
5984 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5985 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5986 				 ~IXGBE_DMATXCTL_TE));
5987 		fallthrough;
5988 	default:
5989 		break;
5990 	}
5991 }
5992 
5993 void ixgbe_reset(struct ixgbe_adapter *adapter)
5994 {
5995 	struct ixgbe_hw *hw = &adapter->hw;
5996 	struct net_device *netdev = adapter->netdev;
5997 	int err;
5998 
5999 	if (ixgbe_removed(hw->hw_addr))
6000 		return;
6001 	/* lock SFP init bit to prevent race conditions with the watchdog */
6002 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
6003 		usleep_range(1000, 2000);
6004 
6005 	/* clear all SFP and link config related flags while holding SFP_INIT */
6006 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
6007 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
6008 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
6009 
6010 	err = hw->mac.ops.init_hw(hw);
6011 	switch (err) {
6012 	case 0:
6013 	case -ENOENT:
6014 	case -EOPNOTSUPP:
6015 		break;
6016 	case -EALREADY:
6017 		e_dev_err("primary disable timed out\n");
6018 		break;
6019 	case -EACCES:
6020 		/* We are running on a pre-production device, log a warning */
6021 		e_dev_warn("This device is a pre-production adapter/LOM. "
6022 			   "Please be aware there may be issues associated with "
6023 			   "your hardware.  If you are experiencing problems "
6024 			   "please contact your Intel or hardware "
6025 			   "representative who provided you with this "
6026 			   "hardware.\n");
6027 		break;
6028 	default:
6029 		e_dev_err("Hardware Error: %d\n", err);
6030 	}
6031 
6032 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
6033 
6034 	/* flush entries out of MAC table */
6035 	ixgbe_flush_sw_mac_table(adapter);
6036 	__dev_uc_unsync(netdev, NULL);
6037 
6038 	/* do not flush user set addresses */
6039 	ixgbe_mac_set_default_filter(adapter);
6040 
6041 	/* update SAN MAC vmdq pool selection */
6042 	if (hw->mac.san_mac_rar_index)
6043 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
6044 
6045 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6046 		ixgbe_ptp_reset(adapter);
6047 
6048 	if (hw->phy.ops.set_phy_power) {
6049 		if (!netif_running(adapter->netdev) && !adapter->wol)
6050 			hw->phy.ops.set_phy_power(hw, false);
6051 		else
6052 			hw->phy.ops.set_phy_power(hw, true);
6053 	}
6054 }
6055 
6056 /**
6057  * ixgbe_clean_tx_ring - Free Tx Buffers
6058  * @tx_ring: ring to be cleaned
6059  **/
6060 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
6061 {
6062 	u16 i = tx_ring->next_to_clean;
6063 	struct ixgbe_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
6064 
6065 	if (tx_ring->xsk_pool) {
6066 		ixgbe_xsk_clean_tx_ring(tx_ring);
6067 		goto out;
6068 	}
6069 
6070 	while (i != tx_ring->next_to_use) {
6071 		union ixgbe_adv_tx_desc *eop_desc, *tx_desc;
6072 
6073 		/* Free all the Tx ring sk_buffs */
6074 		if (ring_is_xdp(tx_ring))
6075 			xdp_return_frame(tx_buffer->xdpf);
6076 		else
6077 			dev_kfree_skb_any(tx_buffer->skb);
6078 
6079 		/* unmap skb header data */
6080 		dma_unmap_single(tx_ring->dev,
6081 				 dma_unmap_addr(tx_buffer, dma),
6082 				 dma_unmap_len(tx_buffer, len),
6083 				 DMA_TO_DEVICE);
6084 
6085 		/* check for eop_desc to determine the end of the packet */
6086 		eop_desc = tx_buffer->next_to_watch;
6087 		tx_desc = IXGBE_TX_DESC(tx_ring, i);
6088 
6089 		/* unmap remaining buffers */
6090 		while (tx_desc != eop_desc) {
6091 			tx_buffer++;
6092 			tx_desc++;
6093 			i++;
6094 			if (unlikely(i == tx_ring->count)) {
6095 				i = 0;
6096 				tx_buffer = tx_ring->tx_buffer_info;
6097 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6098 			}
6099 
6100 			/* unmap any remaining paged data */
6101 			if (dma_unmap_len(tx_buffer, len))
6102 				dma_unmap_page(tx_ring->dev,
6103 					       dma_unmap_addr(tx_buffer, dma),
6104 					       dma_unmap_len(tx_buffer, len),
6105 					       DMA_TO_DEVICE);
6106 		}
6107 
6108 		/* move us one more past the eop_desc for start of next pkt */
6109 		tx_buffer++;
6110 		i++;
6111 		if (unlikely(i == tx_ring->count)) {
6112 			i = 0;
6113 			tx_buffer = tx_ring->tx_buffer_info;
6114 		}
6115 	}
6116 
6117 	/* reset BQL for queue */
6118 	if (!ring_is_xdp(tx_ring))
6119 		netdev_tx_reset_queue(txring_txq(tx_ring));
6120 
6121 out:
6122 	/* reset next_to_use and next_to_clean */
6123 	tx_ring->next_to_use = 0;
6124 	tx_ring->next_to_clean = 0;
6125 }
6126 
6127 /**
6128  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
6129  * @adapter: board private structure
6130  **/
6131 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
6132 {
6133 	int i;
6134 
6135 	for (i = 0; i < adapter->num_rx_queues; i++)
6136 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
6137 }
6138 
6139 /**
6140  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
6141  * @adapter: board private structure
6142  **/
6143 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
6144 {
6145 	int i;
6146 
6147 	for (i = 0; i < adapter->num_tx_queues; i++)
6148 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
6149 	for (i = 0; i < adapter->num_xdp_queues; i++)
6150 		ixgbe_clean_tx_ring(adapter->xdp_ring[i]);
6151 }
6152 
6153 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
6154 {
6155 	struct hlist_node *node2;
6156 	struct ixgbe_fdir_filter *filter;
6157 
6158 	spin_lock(&adapter->fdir_perfect_lock);
6159 
6160 	hlist_for_each_entry_safe(filter, node2,
6161 				  &adapter->fdir_filter_list, fdir_node) {
6162 		hlist_del(&filter->fdir_node);
6163 		kfree(filter);
6164 	}
6165 	adapter->fdir_filter_count = 0;
6166 
6167 	spin_unlock(&adapter->fdir_perfect_lock);
6168 }
6169 
6170 void ixgbe_down(struct ixgbe_adapter *adapter)
6171 {
6172 	struct net_device *netdev = adapter->netdev;
6173 	struct ixgbe_hw *hw = &adapter->hw;
6174 	int i;
6175 
6176 	/* signal that we are down to the interrupt handler */
6177 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
6178 		return; /* do nothing if already down */
6179 
6180 	/* Shut off incoming Tx traffic */
6181 	netif_tx_stop_all_queues(netdev);
6182 
6183 	/* call carrier off first to avoid false dev_watchdog timeouts */
6184 	netif_carrier_off(netdev);
6185 	netif_tx_disable(netdev);
6186 
6187 	/* Disable Rx */
6188 	ixgbe_disable_rx(adapter);
6189 
6190 	/* synchronize_rcu() needed for pending XDP buffers to drain */
6191 	if (adapter->xdp_ring[0])
6192 		synchronize_rcu();
6193 
6194 	ixgbe_irq_disable(adapter);
6195 
6196 	ixgbe_napi_disable_all(adapter);
6197 
6198 	clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
6199 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6200 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6201 
6202 	del_timer_sync(&adapter->service_timer);
6203 
6204 	if (adapter->num_vfs) {
6205 		/* Clear EITR Select mapping */
6206 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
6207 
6208 		/* Mark all the VFs as inactive */
6209 		for (i = 0 ; i < adapter->num_vfs; i++)
6210 			adapter->vfinfo[i].clear_to_send = false;
6211 
6212 		/* update setting rx tx for all active vfs */
6213 		ixgbe_set_all_vfs(adapter);
6214 	}
6215 
6216 	/* disable transmits in the hardware now that interrupts are off */
6217 	ixgbe_disable_tx(adapter);
6218 
6219 	if (!pci_channel_offline(adapter->pdev))
6220 		ixgbe_reset(adapter);
6221 
6222 	/* power down the optics for 82599 SFP+ fiber */
6223 	if (hw->mac.ops.disable_tx_laser)
6224 		hw->mac.ops.disable_tx_laser(hw);
6225 
6226 	ixgbe_clean_all_tx_rings(adapter);
6227 	ixgbe_clean_all_rx_rings(adapter);
6228 }
6229 
6230 /**
6231  * ixgbe_set_eee_capable - helper function to determine EEE support on X550
6232  * @adapter: board private structure
6233  */
6234 static void ixgbe_set_eee_capable(struct ixgbe_adapter *adapter)
6235 {
6236 	struct ixgbe_hw *hw = &adapter->hw;
6237 
6238 	switch (hw->device_id) {
6239 	case IXGBE_DEV_ID_X550EM_A_1G_T:
6240 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6241 		if (!hw->phy.eee_speeds_supported)
6242 			break;
6243 		adapter->flags2 |= IXGBE_FLAG2_EEE_CAPABLE;
6244 		if (!hw->phy.eee_speeds_advertised)
6245 			break;
6246 		adapter->flags2 |= IXGBE_FLAG2_EEE_ENABLED;
6247 		break;
6248 	default:
6249 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_CAPABLE;
6250 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_ENABLED;
6251 		break;
6252 	}
6253 }
6254 
6255 /**
6256  * ixgbe_tx_timeout - Respond to a Tx Hang
6257  * @netdev: network interface device structure
6258  * @txqueue: queue number that timed out
6259  **/
6260 static void ixgbe_tx_timeout(struct net_device *netdev, unsigned int __always_unused txqueue)
6261 {
6262 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6263 
6264 	/* Do the reset outside of interrupt context */
6265 	ixgbe_tx_timeout_reset(adapter);
6266 }
6267 
6268 #ifdef CONFIG_IXGBE_DCB
6269 static void ixgbe_init_dcb(struct ixgbe_adapter *adapter)
6270 {
6271 	struct ixgbe_hw *hw = &adapter->hw;
6272 	struct tc_configuration *tc;
6273 	int j;
6274 
6275 	switch (hw->mac.type) {
6276 	case ixgbe_mac_82598EB:
6277 	case ixgbe_mac_82599EB:
6278 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
6279 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
6280 		break;
6281 	case ixgbe_mac_X540:
6282 	case ixgbe_mac_X550:
6283 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
6284 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
6285 		break;
6286 	case ixgbe_mac_X550EM_x:
6287 	case ixgbe_mac_x550em_a:
6288 	default:
6289 		adapter->dcb_cfg.num_tcs.pg_tcs = DEF_TRAFFIC_CLASS;
6290 		adapter->dcb_cfg.num_tcs.pfc_tcs = DEF_TRAFFIC_CLASS;
6291 		break;
6292 	}
6293 
6294 	/* Configure DCB traffic classes */
6295 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
6296 		tc = &adapter->dcb_cfg.tc_config[j];
6297 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
6298 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
6299 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
6300 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
6301 		tc->dcb_pfc = pfc_disabled;
6302 	}
6303 
6304 	/* Initialize default user to priority mapping, UPx->TC0 */
6305 	tc = &adapter->dcb_cfg.tc_config[0];
6306 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
6307 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
6308 
6309 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
6310 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
6311 	adapter->dcb_cfg.pfc_mode_enable = false;
6312 	adapter->dcb_set_bitmap = 0x00;
6313 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
6314 		adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
6315 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
6316 	       sizeof(adapter->temp_dcb_cfg));
6317 }
6318 #endif
6319 
6320 /**
6321  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
6322  * @adapter: board private structure to initialize
6323  * @ii: pointer to ixgbe_info for device
6324  *
6325  * ixgbe_sw_init initializes the Adapter private data structure.
6326  * Fields are initialized based on PCI device information and
6327  * OS network device settings (MTU size).
6328  **/
6329 static int ixgbe_sw_init(struct ixgbe_adapter *adapter,
6330 			 const struct ixgbe_info *ii)
6331 {
6332 	struct ixgbe_hw *hw = &adapter->hw;
6333 	struct pci_dev *pdev = adapter->pdev;
6334 	unsigned int rss, fdir;
6335 	u32 fwsm;
6336 	int i;
6337 
6338 	/* PCI config space info */
6339 
6340 	hw->vendor_id = pdev->vendor;
6341 	hw->device_id = pdev->device;
6342 	hw->revision_id = pdev->revision;
6343 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
6344 	hw->subsystem_device_id = pdev->subsystem_device;
6345 
6346 	/* get_invariants needs the device IDs */
6347 	ii->get_invariants(hw);
6348 
6349 	/* Set common capability flags and settings */
6350 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
6351 	adapter->ring_feature[RING_F_RSS].limit = rss;
6352 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
6353 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
6354 	adapter->atr_sample_rate = 20;
6355 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
6356 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
6357 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
6358 	adapter->ring_feature[RING_F_VMDQ].limit = 1;
6359 #ifdef CONFIG_IXGBE_DCA
6360 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
6361 #endif
6362 #ifdef CONFIG_IXGBE_DCB
6363 	adapter->flags |= IXGBE_FLAG_DCB_CAPABLE;
6364 	adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
6365 #endif
6366 #ifdef IXGBE_FCOE
6367 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
6368 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6369 #ifdef CONFIG_IXGBE_DCB
6370 	/* Default traffic class to use for FCoE */
6371 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
6372 #endif /* CONFIG_IXGBE_DCB */
6373 #endif /* IXGBE_FCOE */
6374 
6375 	/* initialize static ixgbe jump table entries */
6376 	adapter->jump_tables[0] = kzalloc(sizeof(*adapter->jump_tables[0]),
6377 					  GFP_KERNEL);
6378 	if (!adapter->jump_tables[0])
6379 		return -ENOMEM;
6380 	adapter->jump_tables[0]->mat = ixgbe_ipv4_fields;
6381 
6382 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++)
6383 		adapter->jump_tables[i] = NULL;
6384 
6385 	adapter->mac_table = kcalloc(hw->mac.num_rar_entries,
6386 				     sizeof(struct ixgbe_mac_addr),
6387 				     GFP_KERNEL);
6388 	if (!adapter->mac_table)
6389 		return -ENOMEM;
6390 
6391 	if (ixgbe_init_rss_key(adapter))
6392 		return -ENOMEM;
6393 
6394 	adapter->af_xdp_zc_qps = bitmap_zalloc(IXGBE_MAX_XDP_QS, GFP_KERNEL);
6395 	if (!adapter->af_xdp_zc_qps)
6396 		return -ENOMEM;
6397 
6398 	/* Set MAC specific capability flags and exceptions */
6399 	switch (hw->mac.type) {
6400 	case ixgbe_mac_82598EB:
6401 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
6402 
6403 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
6404 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
6405 
6406 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
6407 		adapter->ring_feature[RING_F_FDIR].limit = 0;
6408 		adapter->atr_sample_rate = 0;
6409 		adapter->fdir_pballoc = 0;
6410 #ifdef IXGBE_FCOE
6411 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6412 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6413 #ifdef CONFIG_IXGBE_DCB
6414 		adapter->fcoe.up = 0;
6415 #endif /* IXGBE_DCB */
6416 #endif /* IXGBE_FCOE */
6417 		break;
6418 	case ixgbe_mac_82599EB:
6419 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
6420 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6421 		break;
6422 	case ixgbe_mac_X540:
6423 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
6424 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
6425 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6426 		break;
6427 	case ixgbe_mac_x550em_a:
6428 		switch (hw->device_id) {
6429 		case IXGBE_DEV_ID_X550EM_A_1G_T:
6430 		case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6431 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6432 			break;
6433 		default:
6434 			break;
6435 		}
6436 		fallthrough;
6437 	case ixgbe_mac_X550EM_x:
6438 #ifdef CONFIG_IXGBE_DCB
6439 		adapter->flags &= ~IXGBE_FLAG_DCB_CAPABLE;
6440 #endif
6441 #ifdef IXGBE_FCOE
6442 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6443 #ifdef CONFIG_IXGBE_DCB
6444 		adapter->fcoe.up = 0;
6445 #endif /* IXGBE_DCB */
6446 #endif /* IXGBE_FCOE */
6447 		fallthrough;
6448 	case ixgbe_mac_X550:
6449 		if (hw->mac.type == ixgbe_mac_X550)
6450 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6451 #ifdef CONFIG_IXGBE_DCA
6452 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
6453 #endif
6454 		break;
6455 	default:
6456 		break;
6457 	}
6458 
6459 #ifdef IXGBE_FCOE
6460 	/* FCoE support exists, always init the FCoE lock */
6461 	spin_lock_init(&adapter->fcoe.lock);
6462 
6463 #endif
6464 	/* n-tuple support exists, always init our spinlock */
6465 	spin_lock_init(&adapter->fdir_perfect_lock);
6466 
6467 	/* init spinlock to avoid concurrency of VF resources */
6468 	spin_lock_init(&adapter->vfs_lock);
6469 
6470 #ifdef CONFIG_IXGBE_DCB
6471 	ixgbe_init_dcb(adapter);
6472 #endif
6473 	ixgbe_init_ipsec_offload(adapter);
6474 
6475 	/* default flow control settings */
6476 	hw->fc.requested_mode = ixgbe_fc_full;
6477 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
6478 	ixgbe_pbthresh_setup(adapter);
6479 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
6480 	hw->fc.send_xon = true;
6481 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
6482 
6483 #ifdef CONFIG_PCI_IOV
6484 	if (max_vfs > 0)
6485 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
6486 
6487 	/* assign number of SR-IOV VFs */
6488 	if (hw->mac.type != ixgbe_mac_82598EB) {
6489 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
6490 			max_vfs = 0;
6491 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
6492 		}
6493 	}
6494 #endif /* CONFIG_PCI_IOV */
6495 
6496 	/* enable itr by default in dynamic mode */
6497 	adapter->rx_itr_setting = 1;
6498 	adapter->tx_itr_setting = 1;
6499 
6500 	/* set default ring sizes */
6501 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
6502 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
6503 
6504 	/* set default work limits */
6505 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
6506 
6507 	/* initialize eeprom parameters */
6508 	if (ixgbe_init_eeprom_params_generic(hw)) {
6509 		e_dev_err("EEPROM initialization failed\n");
6510 		return -EIO;
6511 	}
6512 
6513 	/* PF holds first pool slot */
6514 	set_bit(0, adapter->fwd_bitmask);
6515 	set_bit(__IXGBE_DOWN, &adapter->state);
6516 
6517 	/* enable locking for XDP_TX if we have more CPUs than queues */
6518 	if (nr_cpu_ids > IXGBE_MAX_XDP_QS)
6519 		static_branch_enable(&ixgbe_xdp_locking_key);
6520 
6521 	return 0;
6522 }
6523 
6524 /**
6525  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
6526  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
6527  *
6528  * Return 0 on success, negative on failure
6529  **/
6530 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
6531 {
6532 	struct device *dev = tx_ring->dev;
6533 	int orig_node = dev_to_node(dev);
6534 	int ring_node = NUMA_NO_NODE;
6535 	int size;
6536 
6537 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
6538 
6539 	if (tx_ring->q_vector)
6540 		ring_node = tx_ring->q_vector->numa_node;
6541 
6542 	tx_ring->tx_buffer_info = vmalloc_node(size, ring_node);
6543 	if (!tx_ring->tx_buffer_info)
6544 		tx_ring->tx_buffer_info = vmalloc(size);
6545 	if (!tx_ring->tx_buffer_info)
6546 		goto err;
6547 
6548 	/* round up to nearest 4K */
6549 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
6550 	tx_ring->size = ALIGN(tx_ring->size, 4096);
6551 
6552 	set_dev_node(dev, ring_node);
6553 	tx_ring->desc = dma_alloc_coherent(dev,
6554 					   tx_ring->size,
6555 					   &tx_ring->dma,
6556 					   GFP_KERNEL);
6557 	set_dev_node(dev, orig_node);
6558 	if (!tx_ring->desc)
6559 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
6560 						   &tx_ring->dma, GFP_KERNEL);
6561 	if (!tx_ring->desc)
6562 		goto err;
6563 
6564 	tx_ring->next_to_use = 0;
6565 	tx_ring->next_to_clean = 0;
6566 	return 0;
6567 
6568 err:
6569 	vfree(tx_ring->tx_buffer_info);
6570 	tx_ring->tx_buffer_info = NULL;
6571 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
6572 	return -ENOMEM;
6573 }
6574 
6575 /**
6576  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
6577  * @adapter: board private structure
6578  *
6579  * If this function returns with an error, then it's possible one or
6580  * more of the rings is populated (while the rest are not).  It is the
6581  * callers duty to clean those orphaned rings.
6582  *
6583  * Return 0 on success, negative on failure
6584  **/
6585 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
6586 {
6587 	int i, j = 0, err = 0;
6588 
6589 	for (i = 0; i < adapter->num_tx_queues; i++) {
6590 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
6591 		if (!err)
6592 			continue;
6593 
6594 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
6595 		goto err_setup_tx;
6596 	}
6597 	for (j = 0; j < adapter->num_xdp_queues; j++) {
6598 		err = ixgbe_setup_tx_resources(adapter->xdp_ring[j]);
6599 		if (!err)
6600 			continue;
6601 
6602 		e_err(probe, "Allocation for Tx Queue %u failed\n", j);
6603 		goto err_setup_tx;
6604 	}
6605 
6606 	return 0;
6607 err_setup_tx:
6608 	/* rewind the index freeing the rings as we go */
6609 	while (j--)
6610 		ixgbe_free_tx_resources(adapter->xdp_ring[j]);
6611 	while (i--)
6612 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
6613 	return err;
6614 }
6615 
6616 static int ixgbe_rx_napi_id(struct ixgbe_ring *rx_ring)
6617 {
6618 	struct ixgbe_q_vector *q_vector = rx_ring->q_vector;
6619 
6620 	return q_vector ? q_vector->napi.napi_id : 0;
6621 }
6622 
6623 /**
6624  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
6625  * @adapter: pointer to ixgbe_adapter
6626  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
6627  *
6628  * Returns 0 on success, negative on failure
6629  **/
6630 int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
6631 			     struct ixgbe_ring *rx_ring)
6632 {
6633 	struct device *dev = rx_ring->dev;
6634 	int orig_node = dev_to_node(dev);
6635 	int ring_node = NUMA_NO_NODE;
6636 	int size;
6637 
6638 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
6639 
6640 	if (rx_ring->q_vector)
6641 		ring_node = rx_ring->q_vector->numa_node;
6642 
6643 	rx_ring->rx_buffer_info = vmalloc_node(size, ring_node);
6644 	if (!rx_ring->rx_buffer_info)
6645 		rx_ring->rx_buffer_info = vmalloc(size);
6646 	if (!rx_ring->rx_buffer_info)
6647 		goto err;
6648 
6649 	/* Round up to nearest 4K */
6650 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
6651 	rx_ring->size = ALIGN(rx_ring->size, 4096);
6652 
6653 	set_dev_node(dev, ring_node);
6654 	rx_ring->desc = dma_alloc_coherent(dev,
6655 					   rx_ring->size,
6656 					   &rx_ring->dma,
6657 					   GFP_KERNEL);
6658 	set_dev_node(dev, orig_node);
6659 	if (!rx_ring->desc)
6660 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
6661 						   &rx_ring->dma, GFP_KERNEL);
6662 	if (!rx_ring->desc)
6663 		goto err;
6664 
6665 	rx_ring->next_to_clean = 0;
6666 	rx_ring->next_to_use = 0;
6667 
6668 	/* XDP RX-queue info */
6669 	if (xdp_rxq_info_reg(&rx_ring->xdp_rxq, adapter->netdev,
6670 			     rx_ring->queue_index, ixgbe_rx_napi_id(rx_ring)) < 0)
6671 		goto err;
6672 
6673 	WRITE_ONCE(rx_ring->xdp_prog, adapter->xdp_prog);
6674 
6675 	return 0;
6676 err:
6677 	vfree(rx_ring->rx_buffer_info);
6678 	rx_ring->rx_buffer_info = NULL;
6679 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
6680 	return -ENOMEM;
6681 }
6682 
6683 /**
6684  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
6685  * @adapter: board private structure
6686  *
6687  * If this function returns with an error, then it's possible one or
6688  * more of the rings is populated (while the rest are not).  It is the
6689  * callers duty to clean those orphaned rings.
6690  *
6691  * Return 0 on success, negative on failure
6692  **/
6693 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
6694 {
6695 	int i, err = 0;
6696 
6697 	for (i = 0; i < adapter->num_rx_queues; i++) {
6698 		err = ixgbe_setup_rx_resources(adapter, adapter->rx_ring[i]);
6699 		if (!err)
6700 			continue;
6701 
6702 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
6703 		goto err_setup_rx;
6704 	}
6705 
6706 #ifdef IXGBE_FCOE
6707 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
6708 	if (!err)
6709 #endif
6710 		return 0;
6711 err_setup_rx:
6712 	/* rewind the index freeing the rings as we go */
6713 	while (i--)
6714 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
6715 	return err;
6716 }
6717 
6718 /**
6719  * ixgbe_free_tx_resources - Free Tx Resources per Queue
6720  * @tx_ring: Tx descriptor ring for a specific queue
6721  *
6722  * Free all transmit software resources
6723  **/
6724 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
6725 {
6726 	ixgbe_clean_tx_ring(tx_ring);
6727 
6728 	vfree(tx_ring->tx_buffer_info);
6729 	tx_ring->tx_buffer_info = NULL;
6730 
6731 	/* if not set, then don't free */
6732 	if (!tx_ring->desc)
6733 		return;
6734 
6735 	dma_free_coherent(tx_ring->dev, tx_ring->size,
6736 			  tx_ring->desc, tx_ring->dma);
6737 
6738 	tx_ring->desc = NULL;
6739 }
6740 
6741 /**
6742  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
6743  * @adapter: board private structure
6744  *
6745  * Free all transmit software resources
6746  **/
6747 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
6748 {
6749 	int i;
6750 
6751 	for (i = 0; i < adapter->num_tx_queues; i++)
6752 		if (adapter->tx_ring[i]->desc)
6753 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
6754 	for (i = 0; i < adapter->num_xdp_queues; i++)
6755 		if (adapter->xdp_ring[i]->desc)
6756 			ixgbe_free_tx_resources(adapter->xdp_ring[i]);
6757 }
6758 
6759 /**
6760  * ixgbe_free_rx_resources - Free Rx Resources
6761  * @rx_ring: ring to clean the resources from
6762  *
6763  * Free all receive software resources
6764  **/
6765 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
6766 {
6767 	ixgbe_clean_rx_ring(rx_ring);
6768 
6769 	rx_ring->xdp_prog = NULL;
6770 	xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
6771 	vfree(rx_ring->rx_buffer_info);
6772 	rx_ring->rx_buffer_info = NULL;
6773 
6774 	/* if not set, then don't free */
6775 	if (!rx_ring->desc)
6776 		return;
6777 
6778 	dma_free_coherent(rx_ring->dev, rx_ring->size,
6779 			  rx_ring->desc, rx_ring->dma);
6780 
6781 	rx_ring->desc = NULL;
6782 }
6783 
6784 /**
6785  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
6786  * @adapter: board private structure
6787  *
6788  * Free all receive software resources
6789  **/
6790 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
6791 {
6792 	int i;
6793 
6794 #ifdef IXGBE_FCOE
6795 	ixgbe_free_fcoe_ddp_resources(adapter);
6796 
6797 #endif
6798 	for (i = 0; i < adapter->num_rx_queues; i++)
6799 		if (adapter->rx_ring[i]->desc)
6800 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
6801 }
6802 
6803 /**
6804  * ixgbe_max_xdp_frame_size - returns the maximum allowed frame size for XDP
6805  * @adapter: device handle, pointer to adapter
6806  */
6807 static int ixgbe_max_xdp_frame_size(struct ixgbe_adapter *adapter)
6808 {
6809 	if (PAGE_SIZE >= 8192 || adapter->flags2 & IXGBE_FLAG2_RX_LEGACY)
6810 		return IXGBE_RXBUFFER_2K;
6811 	else
6812 		return IXGBE_RXBUFFER_3K;
6813 }
6814 
6815 /**
6816  * ixgbe_change_mtu - Change the Maximum Transfer Unit
6817  * @netdev: network interface device structure
6818  * @new_mtu: new value for maximum frame size
6819  *
6820  * Returns 0 on success, negative on failure
6821  **/
6822 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
6823 {
6824 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6825 
6826 	if (ixgbe_enabled_xdp_adapter(adapter)) {
6827 		int new_frame_size = new_mtu + IXGBE_PKT_HDR_PAD;
6828 
6829 		if (new_frame_size > ixgbe_max_xdp_frame_size(adapter)) {
6830 			e_warn(probe, "Requested MTU size is not supported with XDP\n");
6831 			return -EINVAL;
6832 		}
6833 	}
6834 
6835 	/*
6836 	 * For 82599EB we cannot allow legacy VFs to enable their receive
6837 	 * paths when MTU greater than 1500 is configured.  So display a
6838 	 * warning that legacy VFs will be disabled.
6839 	 */
6840 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
6841 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
6842 	    (new_mtu > ETH_DATA_LEN))
6843 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
6844 
6845 	netdev_dbg(netdev, "changing MTU from %d to %d\n",
6846 		   netdev->mtu, new_mtu);
6847 
6848 	/* must set new MTU before calling down or up */
6849 	WRITE_ONCE(netdev->mtu, new_mtu);
6850 
6851 	if (netif_running(netdev))
6852 		ixgbe_reinit_locked(adapter);
6853 
6854 	return 0;
6855 }
6856 
6857 /**
6858  * ixgbe_open - Called when a network interface is made active
6859  * @netdev: network interface device structure
6860  *
6861  * Returns 0 on success, negative value on failure
6862  *
6863  * The open entry point is called when a network interface is made
6864  * active by the system (IFF_UP).  At this point all resources needed
6865  * for transmit and receive operations are allocated, the interrupt
6866  * handler is registered with the OS, the watchdog timer is started,
6867  * and the stack is notified that the interface is ready.
6868  **/
6869 int ixgbe_open(struct net_device *netdev)
6870 {
6871 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6872 	struct ixgbe_hw *hw = &adapter->hw;
6873 	int err, queues;
6874 
6875 	/* disallow open during test */
6876 	if (test_bit(__IXGBE_TESTING, &adapter->state))
6877 		return -EBUSY;
6878 
6879 	netif_carrier_off(netdev);
6880 
6881 	/* allocate transmit descriptors */
6882 	err = ixgbe_setup_all_tx_resources(adapter);
6883 	if (err)
6884 		goto err_setup_tx;
6885 
6886 	/* allocate receive descriptors */
6887 	err = ixgbe_setup_all_rx_resources(adapter);
6888 	if (err)
6889 		goto err_setup_rx;
6890 
6891 	ixgbe_configure(adapter);
6892 
6893 	err = ixgbe_request_irq(adapter);
6894 	if (err)
6895 		goto err_req_irq;
6896 
6897 	/* Notify the stack of the actual queue counts. */
6898 	queues = adapter->num_tx_queues;
6899 	err = netif_set_real_num_tx_queues(netdev, queues);
6900 	if (err)
6901 		goto err_set_queues;
6902 
6903 	queues = adapter->num_rx_queues;
6904 	err = netif_set_real_num_rx_queues(netdev, queues);
6905 	if (err)
6906 		goto err_set_queues;
6907 
6908 	ixgbe_ptp_init(adapter);
6909 
6910 	ixgbe_up_complete(adapter);
6911 
6912 	udp_tunnel_nic_reset_ntf(netdev);
6913 
6914 	return 0;
6915 
6916 err_set_queues:
6917 	ixgbe_free_irq(adapter);
6918 err_req_irq:
6919 	ixgbe_free_all_rx_resources(adapter);
6920 	if (hw->phy.ops.set_phy_power && !adapter->wol)
6921 		hw->phy.ops.set_phy_power(&adapter->hw, false);
6922 err_setup_rx:
6923 	ixgbe_free_all_tx_resources(adapter);
6924 err_setup_tx:
6925 	ixgbe_reset(adapter);
6926 
6927 	return err;
6928 }
6929 
6930 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
6931 {
6932 	ixgbe_ptp_suspend(adapter);
6933 
6934 	if (adapter->hw.phy.ops.enter_lplu) {
6935 		adapter->hw.phy.reset_disable = true;
6936 		ixgbe_down(adapter);
6937 		adapter->hw.phy.ops.enter_lplu(&adapter->hw);
6938 		adapter->hw.phy.reset_disable = false;
6939 	} else {
6940 		ixgbe_down(adapter);
6941 	}
6942 
6943 	ixgbe_free_irq(adapter);
6944 
6945 	ixgbe_free_all_tx_resources(adapter);
6946 	ixgbe_free_all_rx_resources(adapter);
6947 }
6948 
6949 /**
6950  * ixgbe_close - Disables a network interface
6951  * @netdev: network interface device structure
6952  *
6953  * Returns 0, this is not allowed to fail
6954  *
6955  * The close entry point is called when an interface is de-activated
6956  * by the OS.  The hardware is still under the drivers control, but
6957  * needs to be disabled.  A global MAC reset is issued to stop the
6958  * hardware, and all transmit and receive resources are freed.
6959  **/
6960 int ixgbe_close(struct net_device *netdev)
6961 {
6962 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6963 
6964 	ixgbe_ptp_stop(adapter);
6965 
6966 	if (netif_device_present(netdev))
6967 		ixgbe_close_suspend(adapter);
6968 
6969 	ixgbe_fdir_filter_exit(adapter);
6970 
6971 	ixgbe_release_hw_control(adapter);
6972 
6973 	return 0;
6974 }
6975 
6976 static int ixgbe_resume(struct device *dev_d)
6977 {
6978 	struct pci_dev *pdev = to_pci_dev(dev_d);
6979 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6980 	struct net_device *netdev = adapter->netdev;
6981 	u32 err;
6982 
6983 	adapter->hw.hw_addr = adapter->io_addr;
6984 
6985 	err = pci_enable_device_mem(pdev);
6986 	if (err) {
6987 		e_dev_err("Cannot enable PCI device from suspend\n");
6988 		return err;
6989 	}
6990 	smp_mb__before_atomic();
6991 	clear_bit(__IXGBE_DISABLED, &adapter->state);
6992 	pci_set_master(pdev);
6993 
6994 	device_wakeup_disable(dev_d);
6995 
6996 	ixgbe_reset(adapter);
6997 
6998 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
6999 
7000 	rtnl_lock();
7001 	err = ixgbe_init_interrupt_scheme(adapter);
7002 	if (!err && netif_running(netdev))
7003 		err = ixgbe_open(netdev);
7004 
7005 
7006 	if (!err)
7007 		netif_device_attach(netdev);
7008 	rtnl_unlock();
7009 
7010 	return err;
7011 }
7012 
7013 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
7014 {
7015 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7016 	struct net_device *netdev = adapter->netdev;
7017 	struct ixgbe_hw *hw = &adapter->hw;
7018 	u32 ctrl;
7019 	u32 wufc = adapter->wol;
7020 
7021 	rtnl_lock();
7022 	netif_device_detach(netdev);
7023 
7024 	if (netif_running(netdev))
7025 		ixgbe_close_suspend(adapter);
7026 
7027 	ixgbe_clear_interrupt_scheme(adapter);
7028 	rtnl_unlock();
7029 
7030 	if (hw->mac.ops.stop_link_on_d3)
7031 		hw->mac.ops.stop_link_on_d3(hw);
7032 
7033 	if (wufc) {
7034 		u32 fctrl;
7035 
7036 		ixgbe_set_rx_mode(netdev);
7037 
7038 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
7039 		if (hw->mac.ops.enable_tx_laser)
7040 			hw->mac.ops.enable_tx_laser(hw);
7041 
7042 		/* enable the reception of multicast packets */
7043 		fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
7044 		fctrl |= IXGBE_FCTRL_MPE;
7045 		IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
7046 
7047 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
7048 		ctrl |= IXGBE_CTRL_GIO_DIS;
7049 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
7050 
7051 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
7052 	} else {
7053 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
7054 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
7055 	}
7056 
7057 	switch (hw->mac.type) {
7058 	case ixgbe_mac_82598EB:
7059 		pci_wake_from_d3(pdev, false);
7060 		break;
7061 	case ixgbe_mac_82599EB:
7062 	case ixgbe_mac_X540:
7063 	case ixgbe_mac_X550:
7064 	case ixgbe_mac_X550EM_x:
7065 	case ixgbe_mac_x550em_a:
7066 		pci_wake_from_d3(pdev, !!wufc);
7067 		break;
7068 	default:
7069 		break;
7070 	}
7071 
7072 	*enable_wake = !!wufc;
7073 	if (hw->phy.ops.set_phy_power && !*enable_wake)
7074 		hw->phy.ops.set_phy_power(hw, false);
7075 
7076 	ixgbe_release_hw_control(adapter);
7077 
7078 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
7079 		pci_disable_device(pdev);
7080 
7081 	return 0;
7082 }
7083 
7084 static int ixgbe_suspend(struct device *dev_d)
7085 {
7086 	struct pci_dev *pdev = to_pci_dev(dev_d);
7087 	int retval;
7088 	bool wake;
7089 
7090 	retval = __ixgbe_shutdown(pdev, &wake);
7091 
7092 	device_set_wakeup_enable(dev_d, wake);
7093 
7094 	return retval;
7095 }
7096 
7097 static void ixgbe_shutdown(struct pci_dev *pdev)
7098 {
7099 	bool wake;
7100 
7101 	__ixgbe_shutdown(pdev, &wake);
7102 
7103 	if (system_state == SYSTEM_POWER_OFF) {
7104 		pci_wake_from_d3(pdev, wake);
7105 		pci_set_power_state(pdev, PCI_D3hot);
7106 	}
7107 }
7108 
7109 /**
7110  * ixgbe_update_stats - Update the board statistics counters.
7111  * @adapter: board private structure
7112  **/
7113 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
7114 {
7115 	struct net_device *netdev = adapter->netdev;
7116 	struct ixgbe_hw *hw = &adapter->hw;
7117 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
7118 	u64 total_mpc = 0;
7119 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
7120 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
7121 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
7122 	u64 alloc_rx_page = 0;
7123 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
7124 
7125 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7126 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7127 		return;
7128 
7129 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
7130 		u64 rsc_count = 0;
7131 		u64 rsc_flush = 0;
7132 		for (i = 0; i < adapter->num_rx_queues; i++) {
7133 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
7134 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
7135 		}
7136 		adapter->rsc_total_count = rsc_count;
7137 		adapter->rsc_total_flush = rsc_flush;
7138 	}
7139 
7140 	for (i = 0; i < adapter->num_rx_queues; i++) {
7141 		struct ixgbe_ring *rx_ring = READ_ONCE(adapter->rx_ring[i]);
7142 
7143 		if (!rx_ring)
7144 			continue;
7145 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
7146 		alloc_rx_page += rx_ring->rx_stats.alloc_rx_page;
7147 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
7148 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
7149 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
7150 		bytes += rx_ring->stats.bytes;
7151 		packets += rx_ring->stats.packets;
7152 	}
7153 	adapter->non_eop_descs = non_eop_descs;
7154 	adapter->alloc_rx_page = alloc_rx_page;
7155 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
7156 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
7157 	adapter->hw_csum_rx_error = hw_csum_rx_error;
7158 	netdev->stats.rx_bytes = bytes;
7159 	netdev->stats.rx_packets = packets;
7160 
7161 	bytes = 0;
7162 	packets = 0;
7163 	/* gather some stats to the adapter struct that are per queue */
7164 	for (i = 0; i < adapter->num_tx_queues; i++) {
7165 		struct ixgbe_ring *tx_ring = READ_ONCE(adapter->tx_ring[i]);
7166 
7167 		if (!tx_ring)
7168 			continue;
7169 		restart_queue += tx_ring->tx_stats.restart_queue;
7170 		tx_busy += tx_ring->tx_stats.tx_busy;
7171 		bytes += tx_ring->stats.bytes;
7172 		packets += tx_ring->stats.packets;
7173 	}
7174 	for (i = 0; i < adapter->num_xdp_queues; i++) {
7175 		struct ixgbe_ring *xdp_ring = READ_ONCE(adapter->xdp_ring[i]);
7176 
7177 		if (!xdp_ring)
7178 			continue;
7179 		restart_queue += xdp_ring->tx_stats.restart_queue;
7180 		tx_busy += xdp_ring->tx_stats.tx_busy;
7181 		bytes += xdp_ring->stats.bytes;
7182 		packets += xdp_ring->stats.packets;
7183 	}
7184 	adapter->restart_queue = restart_queue;
7185 	adapter->tx_busy = tx_busy;
7186 	netdev->stats.tx_bytes = bytes;
7187 	netdev->stats.tx_packets = packets;
7188 
7189 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
7190 
7191 	/* 8 register reads */
7192 	for (i = 0; i < 8; i++) {
7193 		/* for packet buffers not used, the register should read 0 */
7194 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
7195 		missed_rx += mpc;
7196 		hwstats->mpc[i] += mpc;
7197 		total_mpc += hwstats->mpc[i];
7198 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
7199 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
7200 		switch (hw->mac.type) {
7201 		case ixgbe_mac_82598EB:
7202 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
7203 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
7204 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
7205 			hwstats->pxonrxc[i] +=
7206 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
7207 			break;
7208 		case ixgbe_mac_82599EB:
7209 		case ixgbe_mac_X540:
7210 		case ixgbe_mac_X550:
7211 		case ixgbe_mac_X550EM_x:
7212 		case ixgbe_mac_x550em_a:
7213 			hwstats->pxonrxc[i] +=
7214 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
7215 			break;
7216 		default:
7217 			break;
7218 		}
7219 	}
7220 
7221 	/*16 register reads */
7222 	for (i = 0; i < 16; i++) {
7223 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
7224 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
7225 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
7226 		    (hw->mac.type == ixgbe_mac_X540) ||
7227 		    (hw->mac.type == ixgbe_mac_X550) ||
7228 		    (hw->mac.type == ixgbe_mac_X550EM_x) ||
7229 		    (hw->mac.type == ixgbe_mac_x550em_a)) {
7230 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
7231 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
7232 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
7233 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
7234 		}
7235 	}
7236 
7237 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
7238 	/* work around hardware counting issue */
7239 	hwstats->gprc -= missed_rx;
7240 
7241 	ixgbe_update_xoff_received(adapter);
7242 
7243 	/* 82598 hardware only has a 32 bit counter in the high register */
7244 	switch (hw->mac.type) {
7245 	case ixgbe_mac_82598EB:
7246 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
7247 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
7248 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
7249 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
7250 		break;
7251 	case ixgbe_mac_X540:
7252 	case ixgbe_mac_X550:
7253 	case ixgbe_mac_X550EM_x:
7254 	case ixgbe_mac_x550em_a:
7255 		/* OS2BMC stats are X540 and later */
7256 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
7257 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
7258 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
7259 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
7260 		fallthrough;
7261 	case ixgbe_mac_82599EB:
7262 		for (i = 0; i < 16; i++)
7263 			adapter->hw_rx_no_dma_resources +=
7264 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
7265 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
7266 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
7267 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
7268 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
7269 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
7270 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
7271 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
7272 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
7273 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
7274 #ifdef IXGBE_FCOE
7275 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
7276 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
7277 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
7278 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
7279 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
7280 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
7281 		/* Add up per cpu counters for total ddp aloc fail */
7282 		if (adapter->fcoe.ddp_pool) {
7283 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
7284 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
7285 			unsigned int cpu;
7286 			u64 noddp = 0, noddp_ext_buff = 0;
7287 			for_each_possible_cpu(cpu) {
7288 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
7289 				noddp += ddp_pool->noddp;
7290 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
7291 			}
7292 			hwstats->fcoe_noddp = noddp;
7293 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
7294 		}
7295 #endif /* IXGBE_FCOE */
7296 		break;
7297 	default:
7298 		break;
7299 	}
7300 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
7301 	hwstats->bprc += bprc;
7302 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
7303 	if (hw->mac.type == ixgbe_mac_82598EB)
7304 		hwstats->mprc -= bprc;
7305 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
7306 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
7307 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
7308 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
7309 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
7310 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
7311 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
7312 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
7313 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
7314 	hwstats->lxontxc += lxon;
7315 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
7316 	hwstats->lxofftxc += lxoff;
7317 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
7318 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
7319 	/*
7320 	 * 82598 errata - tx of flow control packets is included in tx counters
7321 	 */
7322 	xon_off_tot = lxon + lxoff;
7323 	hwstats->gptc -= xon_off_tot;
7324 	hwstats->mptc -= xon_off_tot;
7325 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
7326 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
7327 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
7328 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
7329 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
7330 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
7331 	hwstats->ptc64 -= xon_off_tot;
7332 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
7333 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
7334 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
7335 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
7336 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
7337 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
7338 
7339 	/* Fill out the OS statistics structure */
7340 	netdev->stats.multicast = hwstats->mprc;
7341 
7342 	/* Rx Errors */
7343 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
7344 	netdev->stats.rx_dropped = 0;
7345 	netdev->stats.rx_length_errors = hwstats->rlec;
7346 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
7347 	netdev->stats.rx_missed_errors = total_mpc;
7348 
7349 	/* VF Stats Collection - skip while resetting because these
7350 	 * are not clear on read and otherwise you'll sometimes get
7351 	 * crazy values.
7352 	 */
7353 	if (!test_bit(__IXGBE_RESETTING, &adapter->state)) {
7354 		for (i = 0; i < adapter->num_vfs; i++) {
7355 			UPDATE_VF_COUNTER_32bit(IXGBE_PVFGPRC(i),
7356 						adapter->vfinfo[i].last_vfstats.gprc,
7357 						adapter->vfinfo[i].vfstats.gprc);
7358 			UPDATE_VF_COUNTER_32bit(IXGBE_PVFGPTC(i),
7359 						adapter->vfinfo[i].last_vfstats.gptc,
7360 						adapter->vfinfo[i].vfstats.gptc);
7361 			UPDATE_VF_COUNTER_36bit(IXGBE_PVFGORC_LSB(i),
7362 						IXGBE_PVFGORC_MSB(i),
7363 						adapter->vfinfo[i].last_vfstats.gorc,
7364 						adapter->vfinfo[i].vfstats.gorc);
7365 			UPDATE_VF_COUNTER_36bit(IXGBE_PVFGOTC_LSB(i),
7366 						IXGBE_PVFGOTC_MSB(i),
7367 						adapter->vfinfo[i].last_vfstats.gotc,
7368 						adapter->vfinfo[i].vfstats.gotc);
7369 			UPDATE_VF_COUNTER_32bit(IXGBE_PVFMPRC(i),
7370 						adapter->vfinfo[i].last_vfstats.mprc,
7371 						adapter->vfinfo[i].vfstats.mprc);
7372 		}
7373 	}
7374 }
7375 
7376 /**
7377  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
7378  * @adapter: pointer to the device adapter structure
7379  **/
7380 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
7381 {
7382 	struct ixgbe_hw *hw = &adapter->hw;
7383 	int i;
7384 
7385 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
7386 		return;
7387 
7388 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
7389 
7390 	/* if interface is down do nothing */
7391 	if (test_bit(__IXGBE_DOWN, &adapter->state))
7392 		return;
7393 
7394 	/* do nothing if we are not using signature filters */
7395 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
7396 		return;
7397 
7398 	adapter->fdir_overflow++;
7399 
7400 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
7401 		for (i = 0; i < adapter->num_tx_queues; i++)
7402 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7403 				&(adapter->tx_ring[i]->state));
7404 		for (i = 0; i < adapter->num_xdp_queues; i++)
7405 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7406 				&adapter->xdp_ring[i]->state);
7407 		/* re-enable flow director interrupts */
7408 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
7409 	} else {
7410 		e_err(probe, "failed to finish FDIR re-initialization, "
7411 		      "ignored adding FDIR ATR filters\n");
7412 	}
7413 }
7414 
7415 /**
7416  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
7417  * @adapter: pointer to the device adapter structure
7418  *
7419  * This function serves two purposes.  First it strobes the interrupt lines
7420  * in order to make certain interrupts are occurring.  Secondly it sets the
7421  * bits needed to check for TX hangs.  As a result we should immediately
7422  * determine if a hang has occurred.
7423  */
7424 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
7425 {
7426 	struct ixgbe_hw *hw = &adapter->hw;
7427 	u64 eics = 0;
7428 	int i;
7429 
7430 	/* If we're down, removing or resetting, just bail */
7431 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7432 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7433 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7434 		return;
7435 
7436 	/* Force detection of hung controller */
7437 	if (netif_carrier_ok(adapter->netdev)) {
7438 		for (i = 0; i < adapter->num_tx_queues; i++)
7439 			set_check_for_tx_hang(adapter->tx_ring[i]);
7440 		for (i = 0; i < adapter->num_xdp_queues; i++)
7441 			set_check_for_tx_hang(adapter->xdp_ring[i]);
7442 	}
7443 
7444 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
7445 		/*
7446 		 * for legacy and MSI interrupts don't set any bits
7447 		 * that are enabled for EIAM, because this operation
7448 		 * would set *both* EIMS and EICS for any bit in EIAM
7449 		 */
7450 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
7451 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
7452 	} else {
7453 		/* get one bit for every active tx/rx interrupt vector */
7454 		for (i = 0; i < adapter->num_q_vectors; i++) {
7455 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
7456 			if (qv->rx.ring || qv->tx.ring)
7457 				eics |= BIT_ULL(i);
7458 		}
7459 	}
7460 
7461 	/* Cause software interrupt to ensure rings are cleaned */
7462 	ixgbe_irq_rearm_queues(adapter, eics);
7463 }
7464 
7465 /**
7466  * ixgbe_watchdog_update_link - update the link status
7467  * @adapter: pointer to the device adapter structure
7468  **/
7469 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
7470 {
7471 	struct ixgbe_hw *hw = &adapter->hw;
7472 	u32 link_speed = adapter->link_speed;
7473 	bool link_up = adapter->link_up;
7474 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
7475 
7476 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
7477 		return;
7478 
7479 	if (hw->mac.ops.check_link) {
7480 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
7481 	} else {
7482 		/* always assume link is up, if no check link function */
7483 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
7484 		link_up = true;
7485 	}
7486 
7487 	if (adapter->ixgbe_ieee_pfc)
7488 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
7489 
7490 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
7491 		hw->mac.ops.fc_enable(hw);
7492 		ixgbe_set_rx_drop_en(adapter);
7493 	}
7494 
7495 	if (link_up ||
7496 	    time_after(jiffies, (adapter->link_check_timeout +
7497 				 IXGBE_TRY_LINK_TIMEOUT))) {
7498 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
7499 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
7500 		IXGBE_WRITE_FLUSH(hw);
7501 	}
7502 
7503 	adapter->link_up = link_up;
7504 	adapter->link_speed = link_speed;
7505 }
7506 
7507 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
7508 {
7509 #ifdef CONFIG_IXGBE_DCB
7510 	struct net_device *netdev = adapter->netdev;
7511 	struct dcb_app app = {
7512 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
7513 			      .protocol = 0,
7514 			     };
7515 	u8 up = 0;
7516 
7517 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
7518 		up = dcb_ieee_getapp_mask(netdev, &app);
7519 
7520 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
7521 #endif
7522 }
7523 
7524 /**
7525  * ixgbe_watchdog_link_is_up - update netif_carrier status and
7526  *                             print link up message
7527  * @adapter: pointer to the device adapter structure
7528  **/
7529 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
7530 {
7531 	struct net_device *netdev = adapter->netdev;
7532 	struct ixgbe_hw *hw = &adapter->hw;
7533 	u32 link_speed = adapter->link_speed;
7534 	const char *speed_str;
7535 	bool flow_rx, flow_tx;
7536 
7537 	/* only continue if link was previously down */
7538 	if (netif_carrier_ok(netdev))
7539 		return;
7540 
7541 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
7542 
7543 	switch (hw->mac.type) {
7544 	case ixgbe_mac_82598EB: {
7545 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
7546 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
7547 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
7548 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
7549 	}
7550 		break;
7551 	case ixgbe_mac_X540:
7552 	case ixgbe_mac_X550:
7553 	case ixgbe_mac_X550EM_x:
7554 	case ixgbe_mac_x550em_a:
7555 	case ixgbe_mac_82599EB: {
7556 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
7557 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
7558 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
7559 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
7560 	}
7561 		break;
7562 	default:
7563 		flow_tx = false;
7564 		flow_rx = false;
7565 		break;
7566 	}
7567 
7568 	adapter->last_rx_ptp_check = jiffies;
7569 
7570 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7571 		ixgbe_ptp_start_cyclecounter(adapter);
7572 
7573 	switch (link_speed) {
7574 	case IXGBE_LINK_SPEED_10GB_FULL:
7575 		speed_str = "10 Gbps";
7576 		break;
7577 	case IXGBE_LINK_SPEED_5GB_FULL:
7578 		speed_str = "5 Gbps";
7579 		break;
7580 	case IXGBE_LINK_SPEED_2_5GB_FULL:
7581 		speed_str = "2.5 Gbps";
7582 		break;
7583 	case IXGBE_LINK_SPEED_1GB_FULL:
7584 		speed_str = "1 Gbps";
7585 		break;
7586 	case IXGBE_LINK_SPEED_100_FULL:
7587 		speed_str = "100 Mbps";
7588 		break;
7589 	case IXGBE_LINK_SPEED_10_FULL:
7590 		speed_str = "10 Mbps";
7591 		break;
7592 	default:
7593 		speed_str = "unknown speed";
7594 		break;
7595 	}
7596 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
7597 	       ((flow_rx && flow_tx) ? "RX/TX" :
7598 	       (flow_rx ? "RX" :
7599 	       (flow_tx ? "TX" : "None"))));
7600 
7601 	netif_carrier_on(netdev);
7602 	ixgbe_check_vf_rate_limit(adapter);
7603 
7604 	/* enable transmits */
7605 	netif_tx_wake_all_queues(adapter->netdev);
7606 
7607 	/* update the default user priority for VFs */
7608 	ixgbe_update_default_up(adapter);
7609 
7610 	/* ping all the active vfs to let them know link has changed */
7611 	ixgbe_ping_all_vfs(adapter);
7612 }
7613 
7614 /**
7615  * ixgbe_watchdog_link_is_down - update netif_carrier status and
7616  *                               print link down message
7617  * @adapter: pointer to the adapter structure
7618  **/
7619 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
7620 {
7621 	struct net_device *netdev = adapter->netdev;
7622 	struct ixgbe_hw *hw = &adapter->hw;
7623 
7624 	adapter->link_up = false;
7625 	adapter->link_speed = 0;
7626 
7627 	/* only continue if link was up previously */
7628 	if (!netif_carrier_ok(netdev))
7629 		return;
7630 
7631 	/* poll for SFP+ cable when link is down */
7632 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
7633 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
7634 
7635 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7636 		ixgbe_ptp_start_cyclecounter(adapter);
7637 
7638 	e_info(drv, "NIC Link is Down\n");
7639 	netif_carrier_off(netdev);
7640 
7641 	/* ping all the active vfs to let them know link has changed */
7642 	ixgbe_ping_all_vfs(adapter);
7643 }
7644 
7645 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
7646 {
7647 	int i;
7648 
7649 	for (i = 0; i < adapter->num_tx_queues; i++) {
7650 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
7651 
7652 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
7653 			return true;
7654 	}
7655 
7656 	for (i = 0; i < adapter->num_xdp_queues; i++) {
7657 		struct ixgbe_ring *ring = adapter->xdp_ring[i];
7658 
7659 		if (ring->next_to_use != ring->next_to_clean)
7660 			return true;
7661 	}
7662 
7663 	return false;
7664 }
7665 
7666 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
7667 {
7668 	struct ixgbe_hw *hw = &adapter->hw;
7669 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
7670 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
7671 
7672 	int i, j;
7673 
7674 	if (!adapter->num_vfs)
7675 		return false;
7676 
7677 	/* resetting the PF is only needed for MAC before X550 */
7678 	if (hw->mac.type >= ixgbe_mac_X550)
7679 		return false;
7680 
7681 	for (i = 0; i < adapter->num_vfs; i++) {
7682 		for (j = 0; j < q_per_pool; j++) {
7683 			u32 h, t;
7684 
7685 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
7686 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
7687 
7688 			if (h != t)
7689 				return true;
7690 		}
7691 	}
7692 
7693 	return false;
7694 }
7695 
7696 /**
7697  * ixgbe_watchdog_flush_tx - flush queues on link down
7698  * @adapter: pointer to the device adapter structure
7699  **/
7700 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
7701 {
7702 	if (!netif_carrier_ok(adapter->netdev)) {
7703 		if (ixgbe_ring_tx_pending(adapter) ||
7704 		    ixgbe_vf_tx_pending(adapter)) {
7705 			/* We've lost link, so the controller stops DMA,
7706 			 * but we've got queued Tx work that's never going
7707 			 * to get done, so reset controller to flush Tx.
7708 			 * (Do the reset outside of interrupt context).
7709 			 */
7710 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
7711 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
7712 		}
7713 	}
7714 }
7715 
7716 #ifdef CONFIG_PCI_IOV
7717 static void ixgbe_bad_vf_abort(struct ixgbe_adapter *adapter, u32 vf)
7718 {
7719 	struct ixgbe_hw *hw = &adapter->hw;
7720 
7721 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
7722 	    adapter->flags2 & IXGBE_FLAG2_AUTO_DISABLE_VF) {
7723 		adapter->vfinfo[vf].primary_abort_count++;
7724 		if (adapter->vfinfo[vf].primary_abort_count ==
7725 		    IXGBE_PRIMARY_ABORT_LIMIT) {
7726 			ixgbe_set_vf_link_state(adapter, vf,
7727 						IFLA_VF_LINK_STATE_DISABLE);
7728 			adapter->vfinfo[vf].primary_abort_count = 0;
7729 
7730 			e_info(drv,
7731 			       "Malicious Driver Detection event detected on PF %d VF %d MAC: %pM mdd-disable-vf=on",
7732 			       hw->bus.func, vf,
7733 			       adapter->vfinfo[vf].vf_mac_addresses);
7734 		}
7735 	}
7736 }
7737 
7738 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
7739 {
7740 	struct ixgbe_hw *hw = &adapter->hw;
7741 	struct pci_dev *pdev = adapter->pdev;
7742 	unsigned int vf;
7743 	u32 gpc;
7744 
7745 	if (!(netif_carrier_ok(adapter->netdev)))
7746 		return;
7747 
7748 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
7749 	if (gpc) /* If incrementing then no need for the check below */
7750 		return;
7751 	/* Check to see if a bad DMA write target from an errant or
7752 	 * malicious VF has caused a PCIe error.  If so then we can
7753 	 * issue a VFLR to the offending VF(s) and then resume without
7754 	 * requesting a full slot reset.
7755 	 */
7756 
7757 	if (!pdev)
7758 		return;
7759 
7760 	/* check status reg for all VFs owned by this PF */
7761 	for (vf = 0; vf < adapter->num_vfs; ++vf) {
7762 		struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
7763 		u16 status_reg;
7764 
7765 		if (!vfdev)
7766 			continue;
7767 		pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
7768 		if (status_reg != IXGBE_FAILED_READ_CFG_WORD &&
7769 		    status_reg & PCI_STATUS_REC_MASTER_ABORT) {
7770 			ixgbe_bad_vf_abort(adapter, vf);
7771 			pcie_flr(vfdev);
7772 		}
7773 	}
7774 }
7775 
7776 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
7777 {
7778 	u32 ssvpc;
7779 
7780 	/* Do not perform spoof check for 82598 or if not in IOV mode */
7781 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7782 	    adapter->num_vfs == 0)
7783 		return;
7784 
7785 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
7786 
7787 	/*
7788 	 * ssvpc register is cleared on read, if zero then no
7789 	 * spoofed packets in the last interval.
7790 	 */
7791 	if (!ssvpc)
7792 		return;
7793 
7794 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
7795 }
7796 #else
7797 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
7798 {
7799 }
7800 
7801 static void
7802 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
7803 {
7804 }
7805 #endif /* CONFIG_PCI_IOV */
7806 
7807 
7808 /**
7809  * ixgbe_watchdog_subtask - check and bring link up
7810  * @adapter: pointer to the device adapter structure
7811  **/
7812 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
7813 {
7814 	/* if interface is down, removing or resetting, do nothing */
7815 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7816 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7817 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7818 		return;
7819 
7820 	ixgbe_watchdog_update_link(adapter);
7821 
7822 	if (adapter->link_up)
7823 		ixgbe_watchdog_link_is_up(adapter);
7824 	else
7825 		ixgbe_watchdog_link_is_down(adapter);
7826 
7827 	ixgbe_check_for_bad_vf(adapter);
7828 	ixgbe_spoof_check(adapter);
7829 	ixgbe_update_stats(adapter);
7830 
7831 	ixgbe_watchdog_flush_tx(adapter);
7832 }
7833 
7834 /**
7835  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
7836  * @adapter: the ixgbe adapter structure
7837  **/
7838 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
7839 {
7840 	struct ixgbe_hw *hw = &adapter->hw;
7841 	int err;
7842 
7843 	/* not searching for SFP so there is nothing to do here */
7844 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
7845 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7846 		return;
7847 
7848 	if (adapter->sfp_poll_time &&
7849 	    time_after(adapter->sfp_poll_time, jiffies))
7850 		return; /* If not yet time to poll for SFP */
7851 
7852 	/* someone else is in init, wait until next service event */
7853 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7854 		return;
7855 
7856 	adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
7857 
7858 	err = hw->phy.ops.identify_sfp(hw);
7859 	if (err == -EOPNOTSUPP)
7860 		goto sfp_out;
7861 
7862 	if (err == -ENOENT) {
7863 		/* If no cable is present, then we need to reset
7864 		 * the next time we find a good cable. */
7865 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
7866 	}
7867 
7868 	/* exit on error */
7869 	if (err)
7870 		goto sfp_out;
7871 
7872 	/* exit if reset not needed */
7873 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7874 		goto sfp_out;
7875 
7876 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
7877 
7878 	/*
7879 	 * A module may be identified correctly, but the EEPROM may not have
7880 	 * support for that module.  setup_sfp() will fail in that case, so
7881 	 * we should not allow that module to load.
7882 	 */
7883 	if (hw->mac.type == ixgbe_mac_82598EB)
7884 		err = hw->phy.ops.reset(hw);
7885 	else
7886 		err = hw->mac.ops.setup_sfp(hw);
7887 
7888 	if (err == -EOPNOTSUPP)
7889 		goto sfp_out;
7890 
7891 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
7892 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
7893 
7894 sfp_out:
7895 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7896 
7897 	if (err == -EOPNOTSUPP &&
7898 	    adapter->netdev->reg_state == NETREG_REGISTERED) {
7899 		e_dev_err("failed to initialize because an unsupported "
7900 			  "SFP+ module type was detected.\n");
7901 		e_dev_err("Reload the driver after installing a "
7902 			  "supported module.\n");
7903 		unregister_netdev(adapter->netdev);
7904 	}
7905 }
7906 
7907 /**
7908  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
7909  * @adapter: the ixgbe adapter structure
7910  **/
7911 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
7912 {
7913 	struct ixgbe_hw *hw = &adapter->hw;
7914 	u32 cap_speed;
7915 	u32 speed;
7916 	bool autoneg = false;
7917 
7918 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
7919 		return;
7920 
7921 	/* someone else is in init, wait until next service event */
7922 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7923 		return;
7924 
7925 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
7926 
7927 	hw->mac.ops.get_link_capabilities(hw, &cap_speed, &autoneg);
7928 
7929 	/* advertise highest capable link speed */
7930 	if (!autoneg && (cap_speed & IXGBE_LINK_SPEED_10GB_FULL))
7931 		speed = IXGBE_LINK_SPEED_10GB_FULL;
7932 	else
7933 		speed = cap_speed & (IXGBE_LINK_SPEED_10GB_FULL |
7934 				     IXGBE_LINK_SPEED_1GB_FULL);
7935 
7936 	if (hw->mac.ops.setup_link)
7937 		hw->mac.ops.setup_link(hw, speed, true);
7938 
7939 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
7940 	adapter->link_check_timeout = jiffies;
7941 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7942 }
7943 
7944 /**
7945  * ixgbe_service_timer - Timer Call-back
7946  * @t: pointer to timer_list structure
7947  **/
7948 static void ixgbe_service_timer(struct timer_list *t)
7949 {
7950 	struct ixgbe_adapter *adapter = from_timer(adapter, t, service_timer);
7951 	unsigned long next_event_offset;
7952 
7953 	/* poll faster when waiting for link */
7954 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
7955 		next_event_offset = HZ / 10;
7956 	else
7957 		next_event_offset = HZ * 2;
7958 
7959 	/* Reset the timer */
7960 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
7961 
7962 	ixgbe_service_event_schedule(adapter);
7963 }
7964 
7965 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
7966 {
7967 	struct ixgbe_hw *hw = &adapter->hw;
7968 	bool overtemp;
7969 
7970 	if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
7971 		return;
7972 
7973 	adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
7974 
7975 	if (!hw->phy.ops.handle_lasi)
7976 		return;
7977 
7978 	hw->phy.ops.handle_lasi(&adapter->hw, &overtemp);
7979 	if (overtemp)
7980 		e_crit(drv, "%s\n", ixgbe_overheat_msg);
7981 }
7982 
7983 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
7984 {
7985 	if (!test_and_clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state))
7986 		return;
7987 
7988 	rtnl_lock();
7989 	/* If we're already down, removing or resetting, just bail */
7990 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7991 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7992 	    test_bit(__IXGBE_RESETTING, &adapter->state)) {
7993 		rtnl_unlock();
7994 		return;
7995 	}
7996 
7997 	ixgbe_dump(adapter);
7998 	netdev_err(adapter->netdev, "Reset adapter\n");
7999 	adapter->tx_timeout_count++;
8000 
8001 	ixgbe_reinit_locked(adapter);
8002 	rtnl_unlock();
8003 }
8004 
8005 /**
8006  * ixgbe_check_fw_error - Check firmware for errors
8007  * @adapter: the adapter private structure
8008  *
8009  * Check firmware errors in register FWSM
8010  */
8011 static bool ixgbe_check_fw_error(struct ixgbe_adapter *adapter)
8012 {
8013 	struct ixgbe_hw *hw = &adapter->hw;
8014 	u32 fwsm;
8015 
8016 	/* read fwsm.ext_err_ind register and log errors */
8017 	fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
8018 
8019 	if (fwsm & IXGBE_FWSM_EXT_ERR_IND_MASK ||
8020 	    !(fwsm & IXGBE_FWSM_FW_VAL_BIT))
8021 		e_dev_warn("Warning firmware error detected FWSM: 0x%08X\n",
8022 			   fwsm);
8023 
8024 	if (hw->mac.ops.fw_recovery_mode && hw->mac.ops.fw_recovery_mode(hw)) {
8025 		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");
8026 		return true;
8027 	}
8028 
8029 	return false;
8030 }
8031 
8032 /**
8033  * ixgbe_service_task - manages and runs subtasks
8034  * @work: pointer to work_struct containing our data
8035  **/
8036 static void ixgbe_service_task(struct work_struct *work)
8037 {
8038 	struct ixgbe_adapter *adapter = container_of(work,
8039 						     struct ixgbe_adapter,
8040 						     service_task);
8041 	if (ixgbe_removed(adapter->hw.hw_addr)) {
8042 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
8043 			rtnl_lock();
8044 			ixgbe_down(adapter);
8045 			rtnl_unlock();
8046 		}
8047 		ixgbe_service_event_complete(adapter);
8048 		return;
8049 	}
8050 	if (ixgbe_check_fw_error(adapter)) {
8051 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
8052 			unregister_netdev(adapter->netdev);
8053 		ixgbe_service_event_complete(adapter);
8054 		return;
8055 	}
8056 	ixgbe_reset_subtask(adapter);
8057 	ixgbe_phy_interrupt_subtask(adapter);
8058 	ixgbe_sfp_detection_subtask(adapter);
8059 	ixgbe_sfp_link_config_subtask(adapter);
8060 	ixgbe_check_overtemp_subtask(adapter);
8061 	ixgbe_watchdog_subtask(adapter);
8062 	ixgbe_fdir_reinit_subtask(adapter);
8063 	ixgbe_check_hang_subtask(adapter);
8064 
8065 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
8066 		ixgbe_ptp_overflow_check(adapter);
8067 		if (adapter->flags & IXGBE_FLAG_RX_HWTSTAMP_IN_REGISTER)
8068 			ixgbe_ptp_rx_hang(adapter);
8069 		ixgbe_ptp_tx_hang(adapter);
8070 	}
8071 
8072 	ixgbe_service_event_complete(adapter);
8073 }
8074 
8075 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
8076 		     struct ixgbe_tx_buffer *first,
8077 		     u8 *hdr_len,
8078 		     struct ixgbe_ipsec_tx_data *itd)
8079 {
8080 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
8081 	struct sk_buff *skb = first->skb;
8082 	union {
8083 		struct iphdr *v4;
8084 		struct ipv6hdr *v6;
8085 		unsigned char *hdr;
8086 	} ip;
8087 	union {
8088 		struct tcphdr *tcp;
8089 		struct udphdr *udp;
8090 		unsigned char *hdr;
8091 	} l4;
8092 	u32 paylen, l4_offset;
8093 	u32 fceof_saidx = 0;
8094 	int err;
8095 
8096 	if (skb->ip_summed != CHECKSUM_PARTIAL)
8097 		return 0;
8098 
8099 	if (!skb_is_gso(skb))
8100 		return 0;
8101 
8102 	err = skb_cow_head(skb, 0);
8103 	if (err < 0)
8104 		return err;
8105 
8106 	if (eth_p_mpls(first->protocol))
8107 		ip.hdr = skb_inner_network_header(skb);
8108 	else
8109 		ip.hdr = skb_network_header(skb);
8110 	l4.hdr = skb_checksum_start(skb);
8111 
8112 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
8113 	type_tucmd = (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_L4) ?
8114 		      IXGBE_ADVTXD_TUCMD_L4T_UDP : IXGBE_ADVTXD_TUCMD_L4T_TCP;
8115 
8116 	/* initialize outer IP header fields */
8117 	if (ip.v4->version == 4) {
8118 		unsigned char *csum_start = skb_checksum_start(skb);
8119 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
8120 		int len = csum_start - trans_start;
8121 
8122 		/* IP header will have to cancel out any data that
8123 		 * is not a part of the outer IP header, so set to
8124 		 * a reverse csum if needed, else init check to 0.
8125 		 */
8126 		ip.v4->check = (skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL) ?
8127 					   csum_fold(csum_partial(trans_start,
8128 								  len, 0)) : 0;
8129 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
8130 
8131 		ip.v4->tot_len = 0;
8132 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
8133 				   IXGBE_TX_FLAGS_CSUM |
8134 				   IXGBE_TX_FLAGS_IPV4;
8135 	} else {
8136 		ip.v6->payload_len = 0;
8137 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
8138 				   IXGBE_TX_FLAGS_CSUM;
8139 	}
8140 
8141 	/* determine offset of inner transport header */
8142 	l4_offset = l4.hdr - skb->data;
8143 
8144 	/* remove payload length from inner checksum */
8145 	paylen = skb->len - l4_offset;
8146 
8147 	if (type_tucmd & IXGBE_ADVTXD_TUCMD_L4T_TCP) {
8148 		/* compute length of segmentation header */
8149 		*hdr_len = (l4.tcp->doff * 4) + l4_offset;
8150 		csum_replace_by_diff(&l4.tcp->check,
8151 				     (__force __wsum)htonl(paylen));
8152 	} else {
8153 		/* compute length of segmentation header */
8154 		*hdr_len = sizeof(*l4.udp) + l4_offset;
8155 		csum_replace_by_diff(&l4.udp->check,
8156 				     (__force __wsum)htonl(paylen));
8157 	}
8158 
8159 	/* update gso size and bytecount with header size */
8160 	first->gso_segs = skb_shinfo(skb)->gso_segs;
8161 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
8162 
8163 	/* mss_l4len_id: use 0 as index for TSO */
8164 	mss_l4len_idx = (*hdr_len - l4_offset) << IXGBE_ADVTXD_L4LEN_SHIFT;
8165 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
8166 
8167 	fceof_saidx |= itd->sa_idx;
8168 	type_tucmd |= itd->flags | itd->trailer_len;
8169 
8170 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
8171 	vlan_macip_lens = l4.hdr - ip.hdr;
8172 	vlan_macip_lens |= (ip.hdr - skb->data) << IXGBE_ADVTXD_MACLEN_SHIFT;
8173 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
8174 
8175 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd,
8176 			  mss_l4len_idx);
8177 
8178 	return 1;
8179 }
8180 
8181 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
8182 			  struct ixgbe_tx_buffer *first,
8183 			  struct ixgbe_ipsec_tx_data *itd)
8184 {
8185 	struct sk_buff *skb = first->skb;
8186 	u32 vlan_macip_lens = 0;
8187 	u32 fceof_saidx = 0;
8188 	u32 type_tucmd = 0;
8189 
8190 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
8191 csum_failed:
8192 		if (!(first->tx_flags & (IXGBE_TX_FLAGS_HW_VLAN |
8193 					 IXGBE_TX_FLAGS_CC)))
8194 			return;
8195 		goto no_csum;
8196 	}
8197 
8198 	switch (skb->csum_offset) {
8199 	case offsetof(struct tcphdr, check):
8200 		type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
8201 		fallthrough;
8202 	case offsetof(struct udphdr, check):
8203 		break;
8204 	case offsetof(struct sctphdr, checksum):
8205 		/* validate that this is actually an SCTP request */
8206 		if (skb_csum_is_sctp(skb)) {
8207 			type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_SCTP;
8208 			break;
8209 		}
8210 		fallthrough;
8211 	default:
8212 		skb_checksum_help(skb);
8213 		goto csum_failed;
8214 	}
8215 
8216 	/* update TX checksum flag */
8217 	first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
8218 	vlan_macip_lens = skb_checksum_start_offset(skb) -
8219 			  skb_network_offset(skb);
8220 no_csum:
8221 	/* vlan_macip_lens: MACLEN, VLAN tag */
8222 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
8223 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
8224 
8225 	fceof_saidx |= itd->sa_idx;
8226 	type_tucmd |= itd->flags | itd->trailer_len;
8227 
8228 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd, 0);
8229 }
8230 
8231 #define IXGBE_SET_FLAG(_input, _flag, _result) \
8232 	((_flag <= _result) ? \
8233 	 ((u32)(_input & _flag) * (_result / _flag)) : \
8234 	 ((u32)(_input & _flag) / (_flag / _result)))
8235 
8236 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
8237 {
8238 	/* set type for advanced descriptor with frame checksum insertion */
8239 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
8240 		       IXGBE_ADVTXD_DCMD_DEXT |
8241 		       IXGBE_ADVTXD_DCMD_IFCS;
8242 
8243 	/* set HW vlan bit if vlan is present */
8244 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
8245 				   IXGBE_ADVTXD_DCMD_VLE);
8246 
8247 	/* set segmentation enable bits for TSO/FSO */
8248 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
8249 				   IXGBE_ADVTXD_DCMD_TSE);
8250 
8251 	/* set timestamp bit if present */
8252 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
8253 				   IXGBE_ADVTXD_MAC_TSTAMP);
8254 
8255 	/* insert frame checksum */
8256 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
8257 
8258 	return cmd_type;
8259 }
8260 
8261 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
8262 				   u32 tx_flags, unsigned int paylen)
8263 {
8264 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
8265 
8266 	/* enable L4 checksum for TSO and TX checksum offload */
8267 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8268 					IXGBE_TX_FLAGS_CSUM,
8269 					IXGBE_ADVTXD_POPTS_TXSM);
8270 
8271 	/* enable IPv4 checksum for TSO */
8272 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8273 					IXGBE_TX_FLAGS_IPV4,
8274 					IXGBE_ADVTXD_POPTS_IXSM);
8275 
8276 	/* enable IPsec */
8277 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8278 					IXGBE_TX_FLAGS_IPSEC,
8279 					IXGBE_ADVTXD_POPTS_IPSEC);
8280 
8281 	/*
8282 	 * Check Context must be set if Tx switch is enabled, which it
8283 	 * always is for case where virtual functions are running
8284 	 */
8285 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8286 					IXGBE_TX_FLAGS_CC,
8287 					IXGBE_ADVTXD_CC);
8288 
8289 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
8290 }
8291 
8292 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
8293 {
8294 	if (!netif_subqueue_try_stop(tx_ring->netdev, tx_ring->queue_index,
8295 				     ixgbe_desc_unused(tx_ring), size))
8296 		return -EBUSY;
8297 
8298 	++tx_ring->tx_stats.restart_queue;
8299 	return 0;
8300 }
8301 
8302 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
8303 {
8304 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
8305 		return 0;
8306 
8307 	return __ixgbe_maybe_stop_tx(tx_ring, size);
8308 }
8309 
8310 static int ixgbe_tx_map(struct ixgbe_ring *tx_ring,
8311 			struct ixgbe_tx_buffer *first,
8312 			const u8 hdr_len)
8313 {
8314 	struct sk_buff *skb = first->skb;
8315 	struct ixgbe_tx_buffer *tx_buffer;
8316 	union ixgbe_adv_tx_desc *tx_desc;
8317 	skb_frag_t *frag;
8318 	dma_addr_t dma;
8319 	unsigned int data_len, size;
8320 	u32 tx_flags = first->tx_flags;
8321 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
8322 	u16 i = tx_ring->next_to_use;
8323 
8324 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
8325 
8326 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
8327 
8328 	size = skb_headlen(skb);
8329 	data_len = skb->data_len;
8330 
8331 #ifdef IXGBE_FCOE
8332 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
8333 		if (data_len < sizeof(struct fcoe_crc_eof)) {
8334 			size -= sizeof(struct fcoe_crc_eof) - data_len;
8335 			data_len = 0;
8336 		} else {
8337 			data_len -= sizeof(struct fcoe_crc_eof);
8338 		}
8339 	}
8340 
8341 #endif
8342 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
8343 
8344 	tx_buffer = first;
8345 
8346 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
8347 		if (dma_mapping_error(tx_ring->dev, dma))
8348 			goto dma_error;
8349 
8350 		/* record length, and DMA address */
8351 		dma_unmap_len_set(tx_buffer, len, size);
8352 		dma_unmap_addr_set(tx_buffer, dma, dma);
8353 
8354 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
8355 
8356 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
8357 			tx_desc->read.cmd_type_len =
8358 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
8359 
8360 			i++;
8361 			tx_desc++;
8362 			if (i == tx_ring->count) {
8363 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8364 				i = 0;
8365 			}
8366 			tx_desc->read.olinfo_status = 0;
8367 
8368 			dma += IXGBE_MAX_DATA_PER_TXD;
8369 			size -= IXGBE_MAX_DATA_PER_TXD;
8370 
8371 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
8372 		}
8373 
8374 		if (likely(!data_len))
8375 			break;
8376 
8377 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
8378 
8379 		i++;
8380 		tx_desc++;
8381 		if (i == tx_ring->count) {
8382 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8383 			i = 0;
8384 		}
8385 		tx_desc->read.olinfo_status = 0;
8386 
8387 #ifdef IXGBE_FCOE
8388 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
8389 #else
8390 		size = skb_frag_size(frag);
8391 #endif
8392 		data_len -= size;
8393 
8394 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
8395 				       DMA_TO_DEVICE);
8396 
8397 		tx_buffer = &tx_ring->tx_buffer_info[i];
8398 	}
8399 
8400 	/* write last descriptor with RS and EOP bits */
8401 	cmd_type |= size | IXGBE_TXD_CMD;
8402 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8403 
8404 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
8405 
8406 	/* set the timestamp */
8407 	first->time_stamp = jiffies;
8408 
8409 	skb_tx_timestamp(skb);
8410 
8411 	/*
8412 	 * Force memory writes to complete before letting h/w know there
8413 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
8414 	 * memory model archs, such as IA-64).
8415 	 *
8416 	 * We also need this memory barrier to make certain all of the
8417 	 * status bits have been updated before next_to_watch is written.
8418 	 */
8419 	wmb();
8420 
8421 	/* set next_to_watch value indicating a packet is present */
8422 	first->next_to_watch = tx_desc;
8423 
8424 	i++;
8425 	if (i == tx_ring->count)
8426 		i = 0;
8427 
8428 	tx_ring->next_to_use = i;
8429 
8430 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
8431 
8432 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more()) {
8433 		writel(i, tx_ring->tail);
8434 	}
8435 
8436 	return 0;
8437 dma_error:
8438 	dev_err(tx_ring->dev, "TX DMA map failed\n");
8439 
8440 	/* clear dma mappings for failed tx_buffer_info map */
8441 	for (;;) {
8442 		tx_buffer = &tx_ring->tx_buffer_info[i];
8443 		if (dma_unmap_len(tx_buffer, len))
8444 			dma_unmap_page(tx_ring->dev,
8445 				       dma_unmap_addr(tx_buffer, dma),
8446 				       dma_unmap_len(tx_buffer, len),
8447 				       DMA_TO_DEVICE);
8448 		dma_unmap_len_set(tx_buffer, len, 0);
8449 		if (tx_buffer == first)
8450 			break;
8451 		if (i == 0)
8452 			i += tx_ring->count;
8453 		i--;
8454 	}
8455 
8456 	dev_kfree_skb_any(first->skb);
8457 	first->skb = NULL;
8458 
8459 	tx_ring->next_to_use = i;
8460 
8461 	return -1;
8462 }
8463 
8464 static void ixgbe_atr(struct ixgbe_ring *ring,
8465 		      struct ixgbe_tx_buffer *first)
8466 {
8467 	struct ixgbe_q_vector *q_vector = ring->q_vector;
8468 	union ixgbe_atr_hash_dword input = { .dword = 0 };
8469 	union ixgbe_atr_hash_dword common = { .dword = 0 };
8470 	union {
8471 		unsigned char *network;
8472 		struct iphdr *ipv4;
8473 		struct ipv6hdr *ipv6;
8474 	} hdr;
8475 	struct tcphdr *th;
8476 	unsigned int hlen;
8477 	struct sk_buff *skb;
8478 	__be16 vlan_id;
8479 	int l4_proto;
8480 
8481 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
8482 	if (!q_vector)
8483 		return;
8484 
8485 	/* do nothing if sampling is disabled */
8486 	if (!ring->atr_sample_rate)
8487 		return;
8488 
8489 	ring->atr_count++;
8490 
8491 	/* currently only IPv4/IPv6 with TCP is supported */
8492 	if ((first->protocol != htons(ETH_P_IP)) &&
8493 	    (first->protocol != htons(ETH_P_IPV6)))
8494 		return;
8495 
8496 	/* snag network header to get L4 type and address */
8497 	skb = first->skb;
8498 	hdr.network = skb_network_header(skb);
8499 	if (unlikely(hdr.network <= skb->data))
8500 		return;
8501 	if (skb->encapsulation &&
8502 	    first->protocol == htons(ETH_P_IP) &&
8503 	    hdr.ipv4->protocol == IPPROTO_UDP) {
8504 		struct ixgbe_adapter *adapter = q_vector->adapter;
8505 
8506 		if (unlikely(skb_tail_pointer(skb) < hdr.network +
8507 			     vxlan_headroom(0)))
8508 			return;
8509 
8510 		/* verify the port is recognized as VXLAN */
8511 		if (adapter->vxlan_port &&
8512 		    udp_hdr(skb)->dest == adapter->vxlan_port)
8513 			hdr.network = skb_inner_network_header(skb);
8514 
8515 		if (adapter->geneve_port &&
8516 		    udp_hdr(skb)->dest == adapter->geneve_port)
8517 			hdr.network = skb_inner_network_header(skb);
8518 	}
8519 
8520 	/* Make sure we have at least [minimum IPv4 header + TCP]
8521 	 * or [IPv6 header] bytes
8522 	 */
8523 	if (unlikely(skb_tail_pointer(skb) < hdr.network + 40))
8524 		return;
8525 
8526 	/* Currently only IPv4/IPv6 with TCP is supported */
8527 	switch (hdr.ipv4->version) {
8528 	case IPVERSION:
8529 		/* access ihl as u8 to avoid unaligned access on ia64 */
8530 		hlen = (hdr.network[0] & 0x0F) << 2;
8531 		l4_proto = hdr.ipv4->protocol;
8532 		break;
8533 	case 6:
8534 		hlen = hdr.network - skb->data;
8535 		l4_proto = ipv6_find_hdr(skb, &hlen, IPPROTO_TCP, NULL, NULL);
8536 		hlen -= hdr.network - skb->data;
8537 		break;
8538 	default:
8539 		return;
8540 	}
8541 
8542 	if (l4_proto != IPPROTO_TCP)
8543 		return;
8544 
8545 	if (unlikely(skb_tail_pointer(skb) < hdr.network +
8546 		     hlen + sizeof(struct tcphdr)))
8547 		return;
8548 
8549 	th = (struct tcphdr *)(hdr.network + hlen);
8550 
8551 	/* skip this packet since the socket is closing */
8552 	if (th->fin)
8553 		return;
8554 
8555 	/* sample on all syn packets or once every atr sample count */
8556 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
8557 		return;
8558 
8559 	/* reset sample count */
8560 	ring->atr_count = 0;
8561 
8562 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
8563 
8564 	/*
8565 	 * src and dst are inverted, think how the receiver sees them
8566 	 *
8567 	 * The input is broken into two sections, a non-compressed section
8568 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
8569 	 * is XORed together and stored in the compressed dword.
8570 	 */
8571 	input.formatted.vlan_id = vlan_id;
8572 
8573 	/*
8574 	 * since src port and flex bytes occupy the same word XOR them together
8575 	 * and write the value to source port portion of compressed dword
8576 	 */
8577 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
8578 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
8579 	else
8580 		common.port.src ^= th->dest ^ first->protocol;
8581 	common.port.dst ^= th->source;
8582 
8583 	switch (hdr.ipv4->version) {
8584 	case IPVERSION:
8585 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
8586 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
8587 		break;
8588 	case 6:
8589 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
8590 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
8591 			     hdr.ipv6->saddr.s6_addr32[1] ^
8592 			     hdr.ipv6->saddr.s6_addr32[2] ^
8593 			     hdr.ipv6->saddr.s6_addr32[3] ^
8594 			     hdr.ipv6->daddr.s6_addr32[0] ^
8595 			     hdr.ipv6->daddr.s6_addr32[1] ^
8596 			     hdr.ipv6->daddr.s6_addr32[2] ^
8597 			     hdr.ipv6->daddr.s6_addr32[3];
8598 		break;
8599 	default:
8600 		break;
8601 	}
8602 
8603 	if (hdr.network != skb_network_header(skb))
8604 		input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
8605 
8606 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
8607 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
8608 					      input, common, ring->queue_index);
8609 }
8610 
8611 #ifdef IXGBE_FCOE
8612 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
8613 			      struct net_device *sb_dev)
8614 {
8615 	struct ixgbe_adapter *adapter;
8616 	struct ixgbe_ring_feature *f;
8617 	int txq;
8618 
8619 	if (sb_dev) {
8620 		u8 tc = netdev_get_prio_tc_map(dev, skb->priority);
8621 		struct net_device *vdev = sb_dev;
8622 
8623 		txq = vdev->tc_to_txq[tc].offset;
8624 		txq += reciprocal_scale(skb_get_hash(skb),
8625 					vdev->tc_to_txq[tc].count);
8626 
8627 		return txq;
8628 	}
8629 
8630 	/*
8631 	 * only execute the code below if protocol is FCoE
8632 	 * or FIP and we have FCoE enabled on the adapter
8633 	 */
8634 	switch (vlan_get_protocol(skb)) {
8635 	case htons(ETH_P_FCOE):
8636 	case htons(ETH_P_FIP):
8637 		adapter = netdev_priv(dev);
8638 
8639 		if (!sb_dev && (adapter->flags & IXGBE_FLAG_FCOE_ENABLED))
8640 			break;
8641 		fallthrough;
8642 	default:
8643 		return netdev_pick_tx(dev, skb, sb_dev);
8644 	}
8645 
8646 	f = &adapter->ring_feature[RING_F_FCOE];
8647 
8648 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
8649 					   smp_processor_id();
8650 
8651 	while (txq >= f->indices)
8652 		txq -= f->indices;
8653 
8654 	return txq + f->offset;
8655 }
8656 
8657 #endif
8658 int ixgbe_xmit_xdp_ring(struct ixgbe_ring *ring,
8659 			struct xdp_frame *xdpf)
8660 {
8661 	struct skb_shared_info *sinfo = xdp_get_shared_info_from_frame(xdpf);
8662 	u8 nr_frags = unlikely(xdp_frame_has_frags(xdpf)) ? sinfo->nr_frags : 0;
8663 	u16 i = 0, index = ring->next_to_use;
8664 	struct ixgbe_tx_buffer *tx_head = &ring->tx_buffer_info[index];
8665 	struct ixgbe_tx_buffer *tx_buff = tx_head;
8666 	union ixgbe_adv_tx_desc *tx_desc = IXGBE_TX_DESC(ring, index);
8667 	u32 cmd_type, len = xdpf->len;
8668 	void *data = xdpf->data;
8669 
8670 	if (unlikely(ixgbe_desc_unused(ring) < 1 + nr_frags))
8671 		return IXGBE_XDP_CONSUMED;
8672 
8673 	tx_head->bytecount = xdp_get_frame_len(xdpf);
8674 	tx_head->gso_segs = 1;
8675 	tx_head->xdpf = xdpf;
8676 
8677 	tx_desc->read.olinfo_status =
8678 		cpu_to_le32(tx_head->bytecount << IXGBE_ADVTXD_PAYLEN_SHIFT);
8679 
8680 	for (;;) {
8681 		dma_addr_t dma;
8682 
8683 		dma = dma_map_single(ring->dev, data, len, DMA_TO_DEVICE);
8684 		if (dma_mapping_error(ring->dev, dma))
8685 			goto unmap;
8686 
8687 		dma_unmap_len_set(tx_buff, len, len);
8688 		dma_unmap_addr_set(tx_buff, dma, dma);
8689 
8690 		cmd_type = IXGBE_ADVTXD_DTYP_DATA | IXGBE_ADVTXD_DCMD_DEXT |
8691 			   IXGBE_ADVTXD_DCMD_IFCS | len;
8692 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8693 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
8694 		tx_buff->protocol = 0;
8695 
8696 		if (++index == ring->count)
8697 			index = 0;
8698 
8699 		if (i == nr_frags)
8700 			break;
8701 
8702 		tx_buff = &ring->tx_buffer_info[index];
8703 		tx_desc = IXGBE_TX_DESC(ring, index);
8704 		tx_desc->read.olinfo_status = 0;
8705 
8706 		data = skb_frag_address(&sinfo->frags[i]);
8707 		len = skb_frag_size(&sinfo->frags[i]);
8708 		i++;
8709 	}
8710 	/* put descriptor type bits */
8711 	tx_desc->read.cmd_type_len |= cpu_to_le32(IXGBE_TXD_CMD);
8712 
8713 	/* Avoid any potential race with xdp_xmit and cleanup */
8714 	smp_wmb();
8715 
8716 	tx_head->next_to_watch = tx_desc;
8717 	ring->next_to_use = index;
8718 
8719 	return IXGBE_XDP_TX;
8720 
8721 unmap:
8722 	for (;;) {
8723 		tx_buff = &ring->tx_buffer_info[index];
8724 		if (dma_unmap_len(tx_buff, len))
8725 			dma_unmap_page(ring->dev, dma_unmap_addr(tx_buff, dma),
8726 				       dma_unmap_len(tx_buff, len),
8727 				       DMA_TO_DEVICE);
8728 		dma_unmap_len_set(tx_buff, len, 0);
8729 		if (tx_buff == tx_head)
8730 			break;
8731 
8732 		if (!index)
8733 			index += ring->count;
8734 		index--;
8735 	}
8736 
8737 	return IXGBE_XDP_CONSUMED;
8738 }
8739 
8740 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
8741 			  struct ixgbe_adapter *adapter,
8742 			  struct ixgbe_ring *tx_ring)
8743 {
8744 	struct ixgbe_tx_buffer *first;
8745 	int tso;
8746 	u32 tx_flags = 0;
8747 	unsigned short f;
8748 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
8749 	struct ixgbe_ipsec_tx_data ipsec_tx = { 0 };
8750 	__be16 protocol = skb->protocol;
8751 	u8 hdr_len = 0;
8752 
8753 	/*
8754 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
8755 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
8756 	 *       + 2 desc gap to keep tail from touching head,
8757 	 *       + 1 desc for context descriptor,
8758 	 * otherwise try next time
8759 	 */
8760 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
8761 		count += TXD_USE_COUNT(skb_frag_size(
8762 						&skb_shinfo(skb)->frags[f]));
8763 
8764 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
8765 		tx_ring->tx_stats.tx_busy++;
8766 		return NETDEV_TX_BUSY;
8767 	}
8768 
8769 	/* record the location of the first descriptor for this packet */
8770 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
8771 	first->skb = skb;
8772 	first->bytecount = skb->len;
8773 	first->gso_segs = 1;
8774 
8775 	/* if we have a HW VLAN tag being added default to the HW one */
8776 	if (skb_vlan_tag_present(skb)) {
8777 		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
8778 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8779 	/* else if it is a SW VLAN check the next protocol and store the tag */
8780 	} else if (protocol == htons(ETH_P_8021Q)) {
8781 		struct vlan_hdr *vhdr, _vhdr;
8782 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
8783 		if (!vhdr)
8784 			goto out_drop;
8785 
8786 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
8787 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
8788 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
8789 	}
8790 	protocol = vlan_get_protocol(skb);
8791 
8792 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
8793 	    adapter->ptp_clock) {
8794 		if (adapter->tstamp_config.tx_type == HWTSTAMP_TX_ON &&
8795 		    !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
8796 					   &adapter->state)) {
8797 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
8798 			tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
8799 
8800 			/* schedule check for Tx timestamp */
8801 			adapter->ptp_tx_skb = skb_get(skb);
8802 			adapter->ptp_tx_start = jiffies;
8803 			schedule_work(&adapter->ptp_tx_work);
8804 		} else {
8805 			adapter->tx_hwtstamp_skipped++;
8806 		}
8807 	}
8808 
8809 #ifdef CONFIG_PCI_IOV
8810 	/*
8811 	 * Use the l2switch_enable flag - would be false if the DMA
8812 	 * Tx switch had been disabled.
8813 	 */
8814 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
8815 		tx_flags |= IXGBE_TX_FLAGS_CC;
8816 
8817 #endif
8818 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
8819 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
8820 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
8821 	     (skb->priority != TC_PRIO_CONTROL))) {
8822 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
8823 		tx_flags |= (skb->priority & 0x7) <<
8824 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
8825 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
8826 			struct vlan_ethhdr *vhdr;
8827 
8828 			if (skb_cow_head(skb, 0))
8829 				goto out_drop;
8830 			vhdr = skb_vlan_eth_hdr(skb);
8831 			vhdr->h_vlan_TCI = htons(tx_flags >>
8832 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
8833 		} else {
8834 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8835 		}
8836 	}
8837 
8838 	/* record initial flags and protocol */
8839 	first->tx_flags = tx_flags;
8840 	first->protocol = protocol;
8841 
8842 #ifdef IXGBE_FCOE
8843 	/* setup tx offload for FCoE */
8844 	if ((protocol == htons(ETH_P_FCOE)) &&
8845 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
8846 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
8847 		if (tso < 0)
8848 			goto out_drop;
8849 
8850 		goto xmit_fcoe;
8851 	}
8852 
8853 #endif /* IXGBE_FCOE */
8854 
8855 #ifdef CONFIG_IXGBE_IPSEC
8856 	if (xfrm_offload(skb) &&
8857 	    !ixgbe_ipsec_tx(tx_ring, first, &ipsec_tx))
8858 		goto out_drop;
8859 #endif
8860 	tso = ixgbe_tso(tx_ring, first, &hdr_len, &ipsec_tx);
8861 	if (tso < 0)
8862 		goto out_drop;
8863 	else if (!tso)
8864 		ixgbe_tx_csum(tx_ring, first, &ipsec_tx);
8865 
8866 	/* add the ATR filter if ATR is on */
8867 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
8868 		ixgbe_atr(tx_ring, first);
8869 
8870 #ifdef IXGBE_FCOE
8871 xmit_fcoe:
8872 #endif /* IXGBE_FCOE */
8873 	if (ixgbe_tx_map(tx_ring, first, hdr_len))
8874 		goto cleanup_tx_timestamp;
8875 
8876 	return NETDEV_TX_OK;
8877 
8878 out_drop:
8879 	dev_kfree_skb_any(first->skb);
8880 	first->skb = NULL;
8881 cleanup_tx_timestamp:
8882 	if (unlikely(tx_flags & IXGBE_TX_FLAGS_TSTAMP)) {
8883 		dev_kfree_skb_any(adapter->ptp_tx_skb);
8884 		adapter->ptp_tx_skb = NULL;
8885 		cancel_work_sync(&adapter->ptp_tx_work);
8886 		clear_bit_unlock(__IXGBE_PTP_TX_IN_PROGRESS, &adapter->state);
8887 	}
8888 
8889 	return NETDEV_TX_OK;
8890 }
8891 
8892 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
8893 				      struct net_device *netdev,
8894 				      struct ixgbe_ring *ring)
8895 {
8896 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8897 	struct ixgbe_ring *tx_ring;
8898 
8899 	/*
8900 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
8901 	 * in order to meet this minimum size requirement.
8902 	 */
8903 	if (skb_put_padto(skb, 17))
8904 		return NETDEV_TX_OK;
8905 
8906 	tx_ring = ring ? ring : adapter->tx_ring[skb_get_queue_mapping(skb)];
8907 	if (unlikely(test_bit(__IXGBE_TX_DISABLED, &tx_ring->state)))
8908 		return NETDEV_TX_BUSY;
8909 
8910 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
8911 }
8912 
8913 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
8914 				    struct net_device *netdev)
8915 {
8916 	return __ixgbe_xmit_frame(skb, netdev, NULL);
8917 }
8918 
8919 /**
8920  * ixgbe_set_mac - Change the Ethernet Address of the NIC
8921  * @netdev: network interface device structure
8922  * @p: pointer to an address structure
8923  *
8924  * Returns 0 on success, negative on failure
8925  **/
8926 static int ixgbe_set_mac(struct net_device *netdev, void *p)
8927 {
8928 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8929 	struct ixgbe_hw *hw = &adapter->hw;
8930 	struct sockaddr *addr = p;
8931 
8932 	if (!is_valid_ether_addr(addr->sa_data))
8933 		return -EADDRNOTAVAIL;
8934 
8935 	eth_hw_addr_set(netdev, addr->sa_data);
8936 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
8937 
8938 	ixgbe_mac_set_default_filter(adapter);
8939 
8940 	return 0;
8941 }
8942 
8943 static int
8944 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
8945 {
8946 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8947 	struct ixgbe_hw *hw = &adapter->hw;
8948 	u16 value;
8949 	int rc;
8950 
8951 	if (adapter->mii_bus) {
8952 		int regnum = addr;
8953 
8954 		if (devad != MDIO_DEVAD_NONE)
8955 			return mdiobus_c45_read(adapter->mii_bus, prtad,
8956 						devad, regnum);
8957 
8958 		return mdiobus_read(adapter->mii_bus, prtad, regnum);
8959 	}
8960 
8961 	if (prtad != hw->phy.mdio.prtad)
8962 		return -EINVAL;
8963 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
8964 	if (!rc)
8965 		rc = value;
8966 	return rc;
8967 }
8968 
8969 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
8970 			    u16 addr, u16 value)
8971 {
8972 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8973 	struct ixgbe_hw *hw = &adapter->hw;
8974 
8975 	if (adapter->mii_bus) {
8976 		int regnum = addr;
8977 
8978 		if (devad != MDIO_DEVAD_NONE)
8979 			return mdiobus_c45_write(adapter->mii_bus, prtad, devad,
8980 						 regnum, value);
8981 
8982 		return mdiobus_write(adapter->mii_bus, prtad, regnum, value);
8983 	}
8984 
8985 	if (prtad != hw->phy.mdio.prtad)
8986 		return -EINVAL;
8987 	return hw->phy.ops.write_reg(hw, addr, devad, value);
8988 }
8989 
8990 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
8991 {
8992 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8993 
8994 	switch (cmd) {
8995 	case SIOCSHWTSTAMP:
8996 		return ixgbe_ptp_set_ts_config(adapter, req);
8997 	case SIOCGHWTSTAMP:
8998 		return ixgbe_ptp_get_ts_config(adapter, req);
8999 	case SIOCGMIIPHY:
9000 		if (!adapter->hw.phy.ops.read_reg)
9001 			return -EOPNOTSUPP;
9002 		fallthrough;
9003 	default:
9004 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
9005 	}
9006 }
9007 
9008 /**
9009  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
9010  * netdev->dev_addrs
9011  * @dev: network interface device structure
9012  *
9013  * Returns non-zero on failure
9014  **/
9015 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
9016 {
9017 	int err = 0;
9018 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9019 	struct ixgbe_hw *hw = &adapter->hw;
9020 
9021 	if (is_valid_ether_addr(hw->mac.san_addr)) {
9022 		rtnl_lock();
9023 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
9024 		rtnl_unlock();
9025 
9026 		/* update SAN MAC vmdq pool selection */
9027 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
9028 	}
9029 	return err;
9030 }
9031 
9032 /**
9033  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
9034  * netdev->dev_addrs
9035  * @dev: network interface device structure
9036  *
9037  * Returns non-zero on failure
9038  **/
9039 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
9040 {
9041 	int err = 0;
9042 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9043 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
9044 
9045 	if (is_valid_ether_addr(mac->san_addr)) {
9046 		rtnl_lock();
9047 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
9048 		rtnl_unlock();
9049 	}
9050 	return err;
9051 }
9052 
9053 static void ixgbe_get_ring_stats64(struct rtnl_link_stats64 *stats,
9054 				   struct ixgbe_ring *ring)
9055 {
9056 	u64 bytes, packets;
9057 	unsigned int start;
9058 
9059 	if (ring) {
9060 		do {
9061 			start = u64_stats_fetch_begin(&ring->syncp);
9062 			packets = ring->stats.packets;
9063 			bytes   = ring->stats.bytes;
9064 		} while (u64_stats_fetch_retry(&ring->syncp, start));
9065 		stats->tx_packets += packets;
9066 		stats->tx_bytes   += bytes;
9067 	}
9068 }
9069 
9070 static void ixgbe_get_stats64(struct net_device *netdev,
9071 			      struct rtnl_link_stats64 *stats)
9072 {
9073 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9074 	int i;
9075 
9076 	rcu_read_lock();
9077 	for (i = 0; i < adapter->num_rx_queues; i++) {
9078 		struct ixgbe_ring *ring = READ_ONCE(adapter->rx_ring[i]);
9079 		u64 bytes, packets;
9080 		unsigned int start;
9081 
9082 		if (ring) {
9083 			do {
9084 				start = u64_stats_fetch_begin(&ring->syncp);
9085 				packets = ring->stats.packets;
9086 				bytes   = ring->stats.bytes;
9087 			} while (u64_stats_fetch_retry(&ring->syncp, start));
9088 			stats->rx_packets += packets;
9089 			stats->rx_bytes   += bytes;
9090 		}
9091 	}
9092 
9093 	for (i = 0; i < adapter->num_tx_queues; i++) {
9094 		struct ixgbe_ring *ring = READ_ONCE(adapter->tx_ring[i]);
9095 
9096 		ixgbe_get_ring_stats64(stats, ring);
9097 	}
9098 	for (i = 0; i < adapter->num_xdp_queues; i++) {
9099 		struct ixgbe_ring *ring = READ_ONCE(adapter->xdp_ring[i]);
9100 
9101 		ixgbe_get_ring_stats64(stats, ring);
9102 	}
9103 	rcu_read_unlock();
9104 
9105 	/* following stats updated by ixgbe_watchdog_task() */
9106 	stats->multicast	= netdev->stats.multicast;
9107 	stats->rx_errors	= netdev->stats.rx_errors;
9108 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
9109 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
9110 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
9111 }
9112 
9113 static int ixgbe_ndo_get_vf_stats(struct net_device *netdev, int vf,
9114 				  struct ifla_vf_stats *vf_stats)
9115 {
9116 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9117 
9118 	if (vf < 0 || vf >= adapter->num_vfs)
9119 		return -EINVAL;
9120 
9121 	vf_stats->rx_packets = adapter->vfinfo[vf].vfstats.gprc;
9122 	vf_stats->rx_bytes   = adapter->vfinfo[vf].vfstats.gorc;
9123 	vf_stats->tx_packets = adapter->vfinfo[vf].vfstats.gptc;
9124 	vf_stats->tx_bytes   = adapter->vfinfo[vf].vfstats.gotc;
9125 	vf_stats->multicast  = adapter->vfinfo[vf].vfstats.mprc;
9126 
9127 	return 0;
9128 }
9129 
9130 #ifdef CONFIG_IXGBE_DCB
9131 /**
9132  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
9133  * @adapter: pointer to ixgbe_adapter
9134  * @tc: number of traffic classes currently enabled
9135  *
9136  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
9137  * 802.1Q priority maps to a packet buffer that exists.
9138  */
9139 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
9140 {
9141 	struct ixgbe_hw *hw = &adapter->hw;
9142 	u32 reg, rsave;
9143 	int i;
9144 
9145 	/* 82598 have a static priority to TC mapping that can not
9146 	 * be changed so no validation is needed.
9147 	 */
9148 	if (hw->mac.type == ixgbe_mac_82598EB)
9149 		return;
9150 
9151 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
9152 	rsave = reg;
9153 
9154 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
9155 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
9156 
9157 		/* If up2tc is out of bounds default to zero */
9158 		if (up2tc > tc)
9159 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
9160 	}
9161 
9162 	if (reg != rsave)
9163 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
9164 
9165 	return;
9166 }
9167 
9168 /**
9169  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
9170  * @adapter: Pointer to adapter struct
9171  *
9172  * Populate the netdev user priority to tc map
9173  */
9174 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
9175 {
9176 	struct net_device *dev = adapter->netdev;
9177 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
9178 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
9179 	u8 prio;
9180 
9181 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
9182 		u8 tc = 0;
9183 
9184 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
9185 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
9186 		else if (ets)
9187 			tc = ets->prio_tc[prio];
9188 
9189 		netdev_set_prio_tc_map(dev, prio, tc);
9190 	}
9191 }
9192 
9193 #endif /* CONFIG_IXGBE_DCB */
9194 static int ixgbe_reassign_macvlan_pool(struct net_device *vdev,
9195 				       struct netdev_nested_priv *priv)
9196 {
9197 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)priv->data;
9198 	struct ixgbe_fwd_adapter *accel;
9199 	int pool;
9200 
9201 	/* we only care about macvlans... */
9202 	if (!netif_is_macvlan(vdev))
9203 		return 0;
9204 
9205 	/* that have hardware offload enabled... */
9206 	accel = macvlan_accel_priv(vdev);
9207 	if (!accel)
9208 		return 0;
9209 
9210 	/* If we can relocate to a different bit do so */
9211 	pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
9212 	if (pool < adapter->num_rx_pools) {
9213 		set_bit(pool, adapter->fwd_bitmask);
9214 		accel->pool = pool;
9215 		return 0;
9216 	}
9217 
9218 	/* if we cannot find a free pool then disable the offload */
9219 	netdev_err(vdev, "L2FW offload disabled due to lack of queue resources\n");
9220 	macvlan_release_l2fw_offload(vdev);
9221 
9222 	/* unbind the queues and drop the subordinate channel config */
9223 	netdev_unbind_sb_channel(adapter->netdev, vdev);
9224 	netdev_set_sb_channel(vdev, 0);
9225 
9226 	kfree(accel);
9227 
9228 	return 0;
9229 }
9230 
9231 static void ixgbe_defrag_macvlan_pools(struct net_device *dev)
9232 {
9233 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9234 	struct netdev_nested_priv priv = {
9235 		.data = (void *)adapter,
9236 	};
9237 
9238 	/* flush any stale bits out of the fwd bitmask */
9239 	bitmap_clear(adapter->fwd_bitmask, 1, 63);
9240 
9241 	/* walk through upper devices reassigning pools */
9242 	netdev_walk_all_upper_dev_rcu(dev, ixgbe_reassign_macvlan_pool,
9243 				      &priv);
9244 }
9245 
9246 /**
9247  * ixgbe_setup_tc - configure net_device for multiple traffic classes
9248  *
9249  * @dev: net device to configure
9250  * @tc: number of traffic classes to enable
9251  */
9252 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
9253 {
9254 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9255 	struct ixgbe_hw *hw = &adapter->hw;
9256 
9257 	/* Hardware supports up to 8 traffic classes */
9258 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
9259 		return -EINVAL;
9260 
9261 	if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
9262 		return -EINVAL;
9263 
9264 	/* Hardware has to reinitialize queues and interrupts to
9265 	 * match packet buffer alignment. Unfortunately, the
9266 	 * hardware is not flexible enough to do this dynamically.
9267 	 */
9268 	if (netif_running(dev))
9269 		ixgbe_close(dev);
9270 	else
9271 		ixgbe_reset(adapter);
9272 
9273 	ixgbe_clear_interrupt_scheme(adapter);
9274 
9275 #ifdef CONFIG_IXGBE_DCB
9276 	if (tc) {
9277 		if (adapter->xdp_prog) {
9278 			e_warn(probe, "DCB is not supported with XDP\n");
9279 
9280 			ixgbe_init_interrupt_scheme(adapter);
9281 			if (netif_running(dev))
9282 				ixgbe_open(dev);
9283 			return -EINVAL;
9284 		}
9285 
9286 		netdev_set_num_tc(dev, tc);
9287 		ixgbe_set_prio_tc_map(adapter);
9288 
9289 		adapter->hw_tcs = tc;
9290 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
9291 
9292 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
9293 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
9294 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
9295 		}
9296 	} else {
9297 		netdev_reset_tc(dev);
9298 
9299 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
9300 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
9301 
9302 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
9303 		adapter->hw_tcs = tc;
9304 
9305 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
9306 		adapter->dcb_cfg.pfc_mode_enable = false;
9307 	}
9308 
9309 	ixgbe_validate_rtr(adapter, tc);
9310 
9311 #endif /* CONFIG_IXGBE_DCB */
9312 	ixgbe_init_interrupt_scheme(adapter);
9313 
9314 	ixgbe_defrag_macvlan_pools(dev);
9315 
9316 	if (netif_running(dev))
9317 		return ixgbe_open(dev);
9318 
9319 	return 0;
9320 }
9321 
9322 static int ixgbe_delete_clsu32(struct ixgbe_adapter *adapter,
9323 			       struct tc_cls_u32_offload *cls)
9324 {
9325 	u32 hdl = cls->knode.handle;
9326 	u32 uhtid = TC_U32_USERHTID(cls->knode.handle);
9327 	u32 loc = cls->knode.handle & 0xfffff;
9328 	int err = 0, i, j;
9329 	struct ixgbe_jump_table *jump = NULL;
9330 
9331 	if (loc > IXGBE_MAX_HW_ENTRIES)
9332 		return -EINVAL;
9333 
9334 	if ((uhtid != 0x800) && (uhtid >= IXGBE_MAX_LINK_HANDLE))
9335 		return -EINVAL;
9336 
9337 	/* Clear this filter in the link data it is associated with */
9338 	if (uhtid != 0x800) {
9339 		jump = adapter->jump_tables[uhtid];
9340 		if (!jump)
9341 			return -EINVAL;
9342 		if (!test_bit(loc - 1, jump->child_loc_map))
9343 			return -EINVAL;
9344 		clear_bit(loc - 1, jump->child_loc_map);
9345 	}
9346 
9347 	/* Check if the filter being deleted is a link */
9348 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
9349 		jump = adapter->jump_tables[i];
9350 		if (jump && jump->link_hdl == hdl) {
9351 			/* Delete filters in the hardware in the child hash
9352 			 * table associated with this link
9353 			 */
9354 			for (j = 0; j < IXGBE_MAX_HW_ENTRIES; j++) {
9355 				if (!test_bit(j, jump->child_loc_map))
9356 					continue;
9357 				spin_lock(&adapter->fdir_perfect_lock);
9358 				err = ixgbe_update_ethtool_fdir_entry(adapter,
9359 								      NULL,
9360 								      j + 1);
9361 				spin_unlock(&adapter->fdir_perfect_lock);
9362 				clear_bit(j, jump->child_loc_map);
9363 			}
9364 			/* Remove resources for this link */
9365 			kfree(jump->input);
9366 			kfree(jump->mask);
9367 			kfree(jump);
9368 			adapter->jump_tables[i] = NULL;
9369 			return err;
9370 		}
9371 	}
9372 
9373 	spin_lock(&adapter->fdir_perfect_lock);
9374 	err = ixgbe_update_ethtool_fdir_entry(adapter, NULL, loc);
9375 	spin_unlock(&adapter->fdir_perfect_lock);
9376 	return err;
9377 }
9378 
9379 static int ixgbe_configure_clsu32_add_hnode(struct ixgbe_adapter *adapter,
9380 					    struct tc_cls_u32_offload *cls)
9381 {
9382 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
9383 
9384 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9385 		return -EINVAL;
9386 
9387 	/* This ixgbe devices do not support hash tables at the moment
9388 	 * so abort when given hash tables.
9389 	 */
9390 	if (cls->hnode.divisor > 0)
9391 		return -EINVAL;
9392 
9393 	set_bit(uhtid - 1, &adapter->tables);
9394 	return 0;
9395 }
9396 
9397 static int ixgbe_configure_clsu32_del_hnode(struct ixgbe_adapter *adapter,
9398 					    struct tc_cls_u32_offload *cls)
9399 {
9400 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
9401 
9402 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9403 		return -EINVAL;
9404 
9405 	clear_bit(uhtid - 1, &adapter->tables);
9406 	return 0;
9407 }
9408 
9409 #ifdef CONFIG_NET_CLS_ACT
9410 struct upper_walk_data {
9411 	struct ixgbe_adapter *adapter;
9412 	u64 action;
9413 	int ifindex;
9414 	u8 queue;
9415 };
9416 
9417 static int get_macvlan_queue(struct net_device *upper,
9418 			     struct netdev_nested_priv *priv)
9419 {
9420 	if (netif_is_macvlan(upper)) {
9421 		struct ixgbe_fwd_adapter *vadapter = macvlan_accel_priv(upper);
9422 		struct ixgbe_adapter *adapter;
9423 		struct upper_walk_data *data;
9424 		int ifindex;
9425 
9426 		data = (struct upper_walk_data *)priv->data;
9427 		ifindex = data->ifindex;
9428 		adapter = data->adapter;
9429 		if (vadapter && upper->ifindex == ifindex) {
9430 			data->queue = adapter->rx_ring[vadapter->rx_base_queue]->reg_idx;
9431 			data->action = data->queue;
9432 			return 1;
9433 		}
9434 	}
9435 
9436 	return 0;
9437 }
9438 
9439 static int handle_redirect_action(struct ixgbe_adapter *adapter, int ifindex,
9440 				  u8 *queue, u64 *action)
9441 {
9442 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
9443 	unsigned int num_vfs = adapter->num_vfs, vf;
9444 	struct netdev_nested_priv priv;
9445 	struct upper_walk_data data;
9446 	struct net_device *upper;
9447 
9448 	/* redirect to a SRIOV VF */
9449 	for (vf = 0; vf < num_vfs; ++vf) {
9450 		upper = pci_get_drvdata(adapter->vfinfo[vf].vfdev);
9451 		if (upper->ifindex == ifindex) {
9452 			*queue = vf * __ALIGN_MASK(1, ~vmdq->mask);
9453 			*action = vf + 1;
9454 			*action <<= ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF;
9455 			return 0;
9456 		}
9457 	}
9458 
9459 	/* redirect to a offloaded macvlan netdev */
9460 	data.adapter = adapter;
9461 	data.ifindex = ifindex;
9462 	data.action = 0;
9463 	data.queue = 0;
9464 	priv.data = (void *)&data;
9465 	if (netdev_walk_all_upper_dev_rcu(adapter->netdev,
9466 					  get_macvlan_queue, &priv)) {
9467 		*action = data.action;
9468 		*queue = data.queue;
9469 
9470 		return 0;
9471 	}
9472 
9473 	return -EINVAL;
9474 }
9475 
9476 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9477 			    struct tcf_exts *exts, u64 *action, u8 *queue)
9478 {
9479 	const struct tc_action *a;
9480 	int i;
9481 
9482 	if (!tcf_exts_has_actions(exts))
9483 		return -EINVAL;
9484 
9485 	tcf_exts_for_each_action(i, a, exts) {
9486 		/* Drop action */
9487 		if (is_tcf_gact_shot(a)) {
9488 			*action = IXGBE_FDIR_DROP_QUEUE;
9489 			*queue = IXGBE_FDIR_DROP_QUEUE;
9490 			return 0;
9491 		}
9492 
9493 		/* Redirect to a VF or a offloaded macvlan */
9494 		if (is_tcf_mirred_egress_redirect(a)) {
9495 			struct net_device *dev = tcf_mirred_dev(a);
9496 
9497 			if (!dev)
9498 				return -EINVAL;
9499 			return handle_redirect_action(adapter, dev->ifindex,
9500 						      queue, action);
9501 		}
9502 
9503 		return -EINVAL;
9504 	}
9505 
9506 	return -EINVAL;
9507 }
9508 #else
9509 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9510 			    struct tcf_exts *exts, u64 *action, u8 *queue)
9511 {
9512 	return -EINVAL;
9513 }
9514 #endif /* CONFIG_NET_CLS_ACT */
9515 
9516 static int ixgbe_clsu32_build_input(struct ixgbe_fdir_filter *input,
9517 				    union ixgbe_atr_input *mask,
9518 				    struct tc_cls_u32_offload *cls,
9519 				    struct ixgbe_mat_field *field_ptr,
9520 				    struct ixgbe_nexthdr *nexthdr)
9521 {
9522 	int i, j, off;
9523 	__be32 val, m;
9524 	bool found_entry = false, found_jump_field = false;
9525 
9526 	for (i = 0; i < cls->knode.sel->nkeys; i++) {
9527 		off = cls->knode.sel->keys[i].off;
9528 		val = cls->knode.sel->keys[i].val;
9529 		m = cls->knode.sel->keys[i].mask;
9530 
9531 		for (j = 0; field_ptr[j].val; j++) {
9532 			if (field_ptr[j].off == off) {
9533 				field_ptr[j].val(input, mask, (__force u32)val,
9534 						 (__force u32)m);
9535 				input->filter.formatted.flow_type |=
9536 					field_ptr[j].type;
9537 				found_entry = true;
9538 				break;
9539 			}
9540 		}
9541 		if (nexthdr) {
9542 			if (nexthdr->off == cls->knode.sel->keys[i].off &&
9543 			    nexthdr->val ==
9544 			    (__force u32)cls->knode.sel->keys[i].val &&
9545 			    nexthdr->mask ==
9546 			    (__force u32)cls->knode.sel->keys[i].mask)
9547 				found_jump_field = true;
9548 			else
9549 				continue;
9550 		}
9551 	}
9552 
9553 	if (nexthdr && !found_jump_field)
9554 		return -EINVAL;
9555 
9556 	if (!found_entry)
9557 		return 0;
9558 
9559 	mask->formatted.flow_type = IXGBE_ATR_L4TYPE_IPV6_MASK |
9560 				    IXGBE_ATR_L4TYPE_MASK;
9561 
9562 	if (input->filter.formatted.flow_type == IXGBE_ATR_FLOW_TYPE_IPV4)
9563 		mask->formatted.flow_type &= IXGBE_ATR_L4TYPE_IPV6_MASK;
9564 
9565 	return 0;
9566 }
9567 
9568 static int ixgbe_configure_clsu32(struct ixgbe_adapter *adapter,
9569 				  struct tc_cls_u32_offload *cls)
9570 {
9571 	__be16 protocol = cls->common.protocol;
9572 	u32 loc = cls->knode.handle & 0xfffff;
9573 	struct ixgbe_hw *hw = &adapter->hw;
9574 	struct ixgbe_mat_field *field_ptr;
9575 	struct ixgbe_fdir_filter *input = NULL;
9576 	union ixgbe_atr_input *mask = NULL;
9577 	struct ixgbe_jump_table *jump = NULL;
9578 	int i, err = -EINVAL;
9579 	u8 queue;
9580 	u32 uhtid, link_uhtid;
9581 
9582 	uhtid = TC_U32_USERHTID(cls->knode.handle);
9583 	link_uhtid = TC_U32_USERHTID(cls->knode.link_handle);
9584 
9585 	/* At the moment cls_u32 jumps to network layer and skips past
9586 	 * L2 headers. The canonical method to match L2 frames is to use
9587 	 * negative values. However this is error prone at best but really
9588 	 * just broken because there is no way to "know" what sort of hdr
9589 	 * is in front of the network layer. Fix cls_u32 to support L2
9590 	 * headers when needed.
9591 	 */
9592 	if (protocol != htons(ETH_P_IP))
9593 		return err;
9594 
9595 	if (loc >= ((1024 << adapter->fdir_pballoc) - 2)) {
9596 		e_err(drv, "Location out of range\n");
9597 		return err;
9598 	}
9599 
9600 	/* cls u32 is a graph starting at root node 0x800. The driver tracks
9601 	 * links and also the fields used to advance the parser across each
9602 	 * link (e.g. nexthdr/eat parameters from 'tc'). This way we can map
9603 	 * the u32 graph onto the hardware parse graph denoted in ixgbe_model.h
9604 	 * To add support for new nodes update ixgbe_model.h parse structures
9605 	 * this function _should_ be generic try not to hardcode values here.
9606 	 */
9607 	if (uhtid == 0x800) {
9608 		field_ptr = (adapter->jump_tables[0])->mat;
9609 	} else {
9610 		if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9611 			return err;
9612 		if (!adapter->jump_tables[uhtid])
9613 			return err;
9614 		field_ptr = (adapter->jump_tables[uhtid])->mat;
9615 	}
9616 
9617 	if (!field_ptr)
9618 		return err;
9619 
9620 	/* At this point we know the field_ptr is valid and need to either
9621 	 * build cls_u32 link or attach filter. Because adding a link to
9622 	 * a handle that does not exist is invalid and the same for adding
9623 	 * rules to handles that don't exist.
9624 	 */
9625 
9626 	if (link_uhtid) {
9627 		struct ixgbe_nexthdr *nexthdr = ixgbe_ipv4_jumps;
9628 
9629 		if (link_uhtid >= IXGBE_MAX_LINK_HANDLE)
9630 			return err;
9631 
9632 		if (!test_bit(link_uhtid - 1, &adapter->tables))
9633 			return err;
9634 
9635 		/* Multiple filters as links to the same hash table are not
9636 		 * supported. To add a new filter with the same next header
9637 		 * but different match/jump conditions, create a new hash table
9638 		 * and link to it.
9639 		 */
9640 		if (adapter->jump_tables[link_uhtid] &&
9641 		    (adapter->jump_tables[link_uhtid])->link_hdl) {
9642 			e_err(drv, "Link filter exists for link: %x\n",
9643 			      link_uhtid);
9644 			return err;
9645 		}
9646 
9647 		for (i = 0; nexthdr[i].jump; i++) {
9648 			if (nexthdr[i].o != cls->knode.sel->offoff ||
9649 			    nexthdr[i].s != cls->knode.sel->offshift ||
9650 			    nexthdr[i].m !=
9651 			    (__force u32)cls->knode.sel->offmask)
9652 				return err;
9653 
9654 			jump = kzalloc(sizeof(*jump), GFP_KERNEL);
9655 			if (!jump)
9656 				return -ENOMEM;
9657 			input = kzalloc(sizeof(*input), GFP_KERNEL);
9658 			if (!input) {
9659 				err = -ENOMEM;
9660 				goto free_jump;
9661 			}
9662 			mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9663 			if (!mask) {
9664 				err = -ENOMEM;
9665 				goto free_input;
9666 			}
9667 			jump->input = input;
9668 			jump->mask = mask;
9669 			jump->link_hdl = cls->knode.handle;
9670 
9671 			err = ixgbe_clsu32_build_input(input, mask, cls,
9672 						       field_ptr, &nexthdr[i]);
9673 			if (!err) {
9674 				jump->mat = nexthdr[i].jump;
9675 				adapter->jump_tables[link_uhtid] = jump;
9676 				break;
9677 			} else {
9678 				kfree(mask);
9679 				kfree(input);
9680 				kfree(jump);
9681 			}
9682 		}
9683 		return 0;
9684 	}
9685 
9686 	input = kzalloc(sizeof(*input), GFP_KERNEL);
9687 	if (!input)
9688 		return -ENOMEM;
9689 	mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9690 	if (!mask) {
9691 		err = -ENOMEM;
9692 		goto free_input;
9693 	}
9694 
9695 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid])) {
9696 		if ((adapter->jump_tables[uhtid])->input)
9697 			memcpy(input, (adapter->jump_tables[uhtid])->input,
9698 			       sizeof(*input));
9699 		if ((adapter->jump_tables[uhtid])->mask)
9700 			memcpy(mask, (adapter->jump_tables[uhtid])->mask,
9701 			       sizeof(*mask));
9702 
9703 		/* Lookup in all child hash tables if this location is already
9704 		 * filled with a filter
9705 		 */
9706 		for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
9707 			struct ixgbe_jump_table *link = adapter->jump_tables[i];
9708 
9709 			if (link && (test_bit(loc - 1, link->child_loc_map))) {
9710 				e_err(drv, "Filter exists in location: %x\n",
9711 				      loc);
9712 				err = -EINVAL;
9713 				goto err_out;
9714 			}
9715 		}
9716 	}
9717 	err = ixgbe_clsu32_build_input(input, mask, cls, field_ptr, NULL);
9718 	if (err)
9719 		goto err_out;
9720 
9721 	err = parse_tc_actions(adapter, cls->knode.exts, &input->action,
9722 			       &queue);
9723 	if (err < 0)
9724 		goto err_out;
9725 
9726 	input->sw_idx = loc;
9727 
9728 	spin_lock(&adapter->fdir_perfect_lock);
9729 
9730 	if (hlist_empty(&adapter->fdir_filter_list)) {
9731 		memcpy(&adapter->fdir_mask, mask, sizeof(*mask));
9732 		err = ixgbe_fdir_set_input_mask_82599(hw, mask);
9733 		if (err)
9734 			goto err_out_w_lock;
9735 	} else if (memcmp(&adapter->fdir_mask, mask, sizeof(*mask))) {
9736 		err = -EINVAL;
9737 		goto err_out_w_lock;
9738 	}
9739 
9740 	ixgbe_atr_compute_perfect_hash_82599(&input->filter, mask);
9741 	err = ixgbe_fdir_write_perfect_filter_82599(hw, &input->filter,
9742 						    input->sw_idx, queue);
9743 	if (err)
9744 		goto err_out_w_lock;
9745 
9746 	ixgbe_update_ethtool_fdir_entry(adapter, input, input->sw_idx);
9747 	spin_unlock(&adapter->fdir_perfect_lock);
9748 
9749 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid]))
9750 		set_bit(loc - 1, (adapter->jump_tables[uhtid])->child_loc_map);
9751 
9752 	kfree(mask);
9753 	return err;
9754 err_out_w_lock:
9755 	spin_unlock(&adapter->fdir_perfect_lock);
9756 err_out:
9757 	kfree(mask);
9758 free_input:
9759 	kfree(input);
9760 free_jump:
9761 	kfree(jump);
9762 	return err;
9763 }
9764 
9765 static int ixgbe_setup_tc_cls_u32(struct ixgbe_adapter *adapter,
9766 				  struct tc_cls_u32_offload *cls_u32)
9767 {
9768 	switch (cls_u32->command) {
9769 	case TC_CLSU32_NEW_KNODE:
9770 	case TC_CLSU32_REPLACE_KNODE:
9771 		return ixgbe_configure_clsu32(adapter, cls_u32);
9772 	case TC_CLSU32_DELETE_KNODE:
9773 		return ixgbe_delete_clsu32(adapter, cls_u32);
9774 	case TC_CLSU32_NEW_HNODE:
9775 	case TC_CLSU32_REPLACE_HNODE:
9776 		return ixgbe_configure_clsu32_add_hnode(adapter, cls_u32);
9777 	case TC_CLSU32_DELETE_HNODE:
9778 		return ixgbe_configure_clsu32_del_hnode(adapter, cls_u32);
9779 	default:
9780 		return -EOPNOTSUPP;
9781 	}
9782 }
9783 
9784 static int ixgbe_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
9785 				   void *cb_priv)
9786 {
9787 	struct ixgbe_adapter *adapter = cb_priv;
9788 
9789 	if (!tc_cls_can_offload_and_chain0(adapter->netdev, type_data))
9790 		return -EOPNOTSUPP;
9791 
9792 	switch (type) {
9793 	case TC_SETUP_CLSU32:
9794 		return ixgbe_setup_tc_cls_u32(adapter, type_data);
9795 	default:
9796 		return -EOPNOTSUPP;
9797 	}
9798 }
9799 
9800 static int ixgbe_setup_tc_mqprio(struct net_device *dev,
9801 				 struct tc_mqprio_qopt *mqprio)
9802 {
9803 	mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
9804 	return ixgbe_setup_tc(dev, mqprio->num_tc);
9805 }
9806 
9807 static LIST_HEAD(ixgbe_block_cb_list);
9808 
9809 static int __ixgbe_setup_tc(struct net_device *dev, enum tc_setup_type type,
9810 			    void *type_data)
9811 {
9812 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9813 
9814 	switch (type) {
9815 	case TC_SETUP_BLOCK:
9816 		return flow_block_cb_setup_simple(type_data,
9817 						  &ixgbe_block_cb_list,
9818 						  ixgbe_setup_tc_block_cb,
9819 						  adapter, adapter, true);
9820 	case TC_SETUP_QDISC_MQPRIO:
9821 		return ixgbe_setup_tc_mqprio(dev, type_data);
9822 	default:
9823 		return -EOPNOTSUPP;
9824 	}
9825 }
9826 
9827 #ifdef CONFIG_PCI_IOV
9828 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
9829 {
9830 	struct net_device *netdev = adapter->netdev;
9831 
9832 	rtnl_lock();
9833 	ixgbe_setup_tc(netdev, adapter->hw_tcs);
9834 	rtnl_unlock();
9835 }
9836 
9837 #endif
9838 void ixgbe_do_reset(struct net_device *netdev)
9839 {
9840 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9841 
9842 	if (netif_running(netdev))
9843 		ixgbe_reinit_locked(adapter);
9844 	else
9845 		ixgbe_reset(adapter);
9846 }
9847 
9848 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
9849 					    netdev_features_t features)
9850 {
9851 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9852 
9853 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
9854 	if (!(features & NETIF_F_RXCSUM))
9855 		features &= ~NETIF_F_LRO;
9856 
9857 	/* Turn off LRO if not RSC capable */
9858 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
9859 		features &= ~NETIF_F_LRO;
9860 
9861 	if (adapter->xdp_prog && (features & NETIF_F_LRO)) {
9862 		e_dev_err("LRO is not supported with XDP\n");
9863 		features &= ~NETIF_F_LRO;
9864 	}
9865 
9866 	return features;
9867 }
9868 
9869 static void ixgbe_reset_l2fw_offload(struct ixgbe_adapter *adapter)
9870 {
9871 	int rss = min_t(int, ixgbe_max_rss_indices(adapter),
9872 			num_online_cpus());
9873 
9874 	/* go back to full RSS if we're not running SR-IOV */
9875 	if (!adapter->ring_feature[RING_F_VMDQ].offset)
9876 		adapter->flags &= ~(IXGBE_FLAG_VMDQ_ENABLED |
9877 				    IXGBE_FLAG_SRIOV_ENABLED);
9878 
9879 	adapter->ring_feature[RING_F_RSS].limit = rss;
9880 	adapter->ring_feature[RING_F_VMDQ].limit = 1;
9881 
9882 	ixgbe_setup_tc(adapter->netdev, adapter->hw_tcs);
9883 }
9884 
9885 static int ixgbe_set_features(struct net_device *netdev,
9886 			      netdev_features_t features)
9887 {
9888 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9889 	netdev_features_t changed = netdev->features ^ features;
9890 	bool need_reset = false;
9891 
9892 	/* Make sure RSC matches LRO, reset if change */
9893 	if (!(features & NETIF_F_LRO)) {
9894 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
9895 			need_reset = true;
9896 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
9897 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
9898 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
9899 		if (adapter->rx_itr_setting == 1 ||
9900 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
9901 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
9902 			need_reset = true;
9903 		} else if ((changed ^ features) & NETIF_F_LRO) {
9904 			e_info(probe, "rx-usecs set too low, "
9905 			       "disabling RSC\n");
9906 		}
9907 	}
9908 
9909 	/*
9910 	 * Check if Flow Director n-tuple support or hw_tc support was
9911 	 * enabled or disabled.  If the state changed, we need to reset.
9912 	 */
9913 	if ((features & NETIF_F_NTUPLE) || (features & NETIF_F_HW_TC)) {
9914 		/* turn off ATR, enable perfect filters and reset */
9915 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
9916 			need_reset = true;
9917 
9918 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
9919 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9920 	} else {
9921 		/* turn off perfect filters, enable ATR and reset */
9922 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
9923 			need_reset = true;
9924 
9925 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9926 
9927 		/* We cannot enable ATR if SR-IOV is enabled */
9928 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED ||
9929 		    /* We cannot enable ATR if we have 2 or more tcs */
9930 		    (adapter->hw_tcs > 1) ||
9931 		    /* We cannot enable ATR if RSS is disabled */
9932 		    (adapter->ring_feature[RING_F_RSS].limit <= 1) ||
9933 		    /* A sample rate of 0 indicates ATR disabled */
9934 		    (!adapter->atr_sample_rate))
9935 			; /* do nothing not supported */
9936 		else /* otherwise supported and set the flag */
9937 			adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
9938 	}
9939 
9940 	if (changed & NETIF_F_RXALL)
9941 		need_reset = true;
9942 
9943 	netdev->features = features;
9944 
9945 	if ((changed & NETIF_F_HW_L2FW_DOFFLOAD) && adapter->num_rx_pools > 1)
9946 		ixgbe_reset_l2fw_offload(adapter);
9947 	else if (need_reset)
9948 		ixgbe_do_reset(netdev);
9949 	else if (changed & (NETIF_F_HW_VLAN_CTAG_RX |
9950 			    NETIF_F_HW_VLAN_CTAG_FILTER))
9951 		ixgbe_set_rx_mode(netdev);
9952 
9953 	return 1;
9954 }
9955 
9956 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
9957 			     struct net_device *dev,
9958 			     const unsigned char *addr, u16 vid,
9959 			     u16 flags,
9960 			     struct netlink_ext_ack *extack)
9961 {
9962 	/* guarantee we can provide a unique filter for the unicast address */
9963 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
9964 		struct ixgbe_adapter *adapter = netdev_priv(dev);
9965 		u16 pool = VMDQ_P(0);
9966 
9967 		if (netdev_uc_count(dev) >= ixgbe_available_rars(adapter, pool))
9968 			return -ENOMEM;
9969 	}
9970 
9971 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
9972 }
9973 
9974 /**
9975  * ixgbe_configure_bridge_mode - set various bridge modes
9976  * @adapter: the private structure
9977  * @mode: requested bridge mode
9978  *
9979  * Configure some settings require for various bridge modes.
9980  **/
9981 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
9982 				       __u16 mode)
9983 {
9984 	struct ixgbe_hw *hw = &adapter->hw;
9985 	unsigned int p, num_pools;
9986 	u32 vmdctl;
9987 
9988 	switch (mode) {
9989 	case BRIDGE_MODE_VEPA:
9990 		/* disable Tx loopback, rely on switch hairpin mode */
9991 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
9992 
9993 		/* must enable Rx switching replication to allow multicast
9994 		 * packet reception on all VFs, and to enable source address
9995 		 * pruning.
9996 		 */
9997 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9998 		vmdctl |= IXGBE_VT_CTL_REPLEN;
9999 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
10000 
10001 		/* enable Rx source address pruning. Note, this requires
10002 		 * replication to be enabled or else it does nothing.
10003 		 */
10004 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
10005 		for (p = 0; p < num_pools; p++) {
10006 			if (hw->mac.ops.set_source_address_pruning)
10007 				hw->mac.ops.set_source_address_pruning(hw,
10008 								       true,
10009 								       p);
10010 		}
10011 		break;
10012 	case BRIDGE_MODE_VEB:
10013 		/* enable Tx loopback for internal VF/PF communication */
10014 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
10015 				IXGBE_PFDTXGSWC_VT_LBEN);
10016 
10017 		/* disable Rx switching replication unless we have SR-IOV
10018 		 * virtual functions
10019 		 */
10020 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
10021 		if (!adapter->num_vfs)
10022 			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
10023 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
10024 
10025 		/* disable Rx source address pruning, since we don't expect to
10026 		 * be receiving external loopback of our transmitted frames.
10027 		 */
10028 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
10029 		for (p = 0; p < num_pools; p++) {
10030 			if (hw->mac.ops.set_source_address_pruning)
10031 				hw->mac.ops.set_source_address_pruning(hw,
10032 								       false,
10033 								       p);
10034 		}
10035 		break;
10036 	default:
10037 		return -EINVAL;
10038 	}
10039 
10040 	adapter->bridge_mode = mode;
10041 
10042 	e_info(drv, "enabling bridge mode: %s\n",
10043 	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
10044 
10045 	return 0;
10046 }
10047 
10048 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
10049 				    struct nlmsghdr *nlh, u16 flags,
10050 				    struct netlink_ext_ack *extack)
10051 {
10052 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10053 	struct nlattr *attr, *br_spec;
10054 	int rem;
10055 
10056 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
10057 		return -EOPNOTSUPP;
10058 
10059 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
10060 	if (!br_spec)
10061 		return -EINVAL;
10062 
10063 	nla_for_each_nested_type(attr, IFLA_BRIDGE_MODE, br_spec, rem) {
10064 		__u16 mode = nla_get_u16(attr);
10065 		int status = ixgbe_configure_bridge_mode(adapter, mode);
10066 
10067 		if (status)
10068 			return status;
10069 
10070 		break;
10071 	}
10072 
10073 	return 0;
10074 }
10075 
10076 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
10077 				    struct net_device *dev,
10078 				    u32 filter_mask, int nlflags)
10079 {
10080 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10081 
10082 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
10083 		return 0;
10084 
10085 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
10086 				       adapter->bridge_mode, 0, 0, nlflags,
10087 				       filter_mask, NULL);
10088 }
10089 
10090 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
10091 {
10092 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
10093 	struct ixgbe_fwd_adapter *accel;
10094 	int tcs = adapter->hw_tcs ? : 1;
10095 	int pool, err;
10096 
10097 	if (adapter->xdp_prog) {
10098 		e_warn(probe, "L2FW offload is not supported with XDP\n");
10099 		return ERR_PTR(-EINVAL);
10100 	}
10101 
10102 	/* The hardware supported by ixgbe only filters on the destination MAC
10103 	 * address. In order to avoid issues we only support offloading modes
10104 	 * where the hardware can actually provide the functionality.
10105 	 */
10106 	if (!macvlan_supports_dest_filter(vdev))
10107 		return ERR_PTR(-EMEDIUMTYPE);
10108 
10109 	/* We need to lock down the macvlan to be a single queue device so that
10110 	 * we can reuse the tc_to_txq field in the macvlan netdev to represent
10111 	 * the queue mapping to our netdev.
10112 	 */
10113 	if (netif_is_multiqueue(vdev))
10114 		return ERR_PTR(-ERANGE);
10115 
10116 	pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
10117 	if (pool == adapter->num_rx_pools) {
10118 		u16 used_pools = adapter->num_vfs + adapter->num_rx_pools;
10119 		u16 reserved_pools;
10120 
10121 		if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
10122 		     adapter->num_rx_pools >= (MAX_TX_QUEUES / tcs)) ||
10123 		    adapter->num_rx_pools > IXGBE_MAX_MACVLANS)
10124 			return ERR_PTR(-EBUSY);
10125 
10126 		/* Hardware has a limited number of available pools. Each VF,
10127 		 * and the PF require a pool. Check to ensure we don't
10128 		 * attempt to use more then the available number of pools.
10129 		 */
10130 		if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
10131 			return ERR_PTR(-EBUSY);
10132 
10133 		/* Enable VMDq flag so device will be set in VM mode */
10134 		adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED |
10135 				  IXGBE_FLAG_SRIOV_ENABLED;
10136 
10137 		/* Try to reserve as many queues per pool as possible,
10138 		 * we start with the configurations that support 4 queues
10139 		 * per pools, followed by 2, and then by just 1 per pool.
10140 		 */
10141 		if (used_pools < 32 && adapter->num_rx_pools < 16)
10142 			reserved_pools = min_t(u16,
10143 					       32 - used_pools,
10144 					       16 - adapter->num_rx_pools);
10145 		else if (adapter->num_rx_pools < 32)
10146 			reserved_pools = min_t(u16,
10147 					       64 - used_pools,
10148 					       32 - adapter->num_rx_pools);
10149 		else
10150 			reserved_pools = 64 - used_pools;
10151 
10152 
10153 		if (!reserved_pools)
10154 			return ERR_PTR(-EBUSY);
10155 
10156 		adapter->ring_feature[RING_F_VMDQ].limit += reserved_pools;
10157 
10158 		/* Force reinit of ring allocation with VMDQ enabled */
10159 		err = ixgbe_setup_tc(pdev, adapter->hw_tcs);
10160 		if (err)
10161 			return ERR_PTR(err);
10162 
10163 		if (pool >= adapter->num_rx_pools)
10164 			return ERR_PTR(-ENOMEM);
10165 	}
10166 
10167 	accel = kzalloc(sizeof(*accel), GFP_KERNEL);
10168 	if (!accel)
10169 		return ERR_PTR(-ENOMEM);
10170 
10171 	set_bit(pool, adapter->fwd_bitmask);
10172 	netdev_set_sb_channel(vdev, pool);
10173 	accel->pool = pool;
10174 	accel->netdev = vdev;
10175 
10176 	if (!netif_running(pdev))
10177 		return accel;
10178 
10179 	err = ixgbe_fwd_ring_up(adapter, accel);
10180 	if (err)
10181 		return ERR_PTR(err);
10182 
10183 	return accel;
10184 }
10185 
10186 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
10187 {
10188 	struct ixgbe_fwd_adapter *accel = priv;
10189 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
10190 	unsigned int rxbase = accel->rx_base_queue;
10191 	unsigned int i;
10192 
10193 	/* delete unicast filter associated with offloaded interface */
10194 	ixgbe_del_mac_filter(adapter, accel->netdev->dev_addr,
10195 			     VMDQ_P(accel->pool));
10196 
10197 	/* Allow remaining Rx packets to get flushed out of the
10198 	 * Rx FIFO before we drop the netdev for the ring.
10199 	 */
10200 	usleep_range(10000, 20000);
10201 
10202 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
10203 		struct ixgbe_ring *ring = adapter->rx_ring[rxbase + i];
10204 		struct ixgbe_q_vector *qv = ring->q_vector;
10205 
10206 		/* Make sure we aren't processing any packets and clear
10207 		 * netdev to shut down the ring.
10208 		 */
10209 		if (netif_running(adapter->netdev))
10210 			napi_synchronize(&qv->napi);
10211 		ring->netdev = NULL;
10212 	}
10213 
10214 	/* unbind the queues and drop the subordinate channel config */
10215 	netdev_unbind_sb_channel(pdev, accel->netdev);
10216 	netdev_set_sb_channel(accel->netdev, 0);
10217 
10218 	clear_bit(accel->pool, adapter->fwd_bitmask);
10219 	kfree(accel);
10220 }
10221 
10222 #define IXGBE_MAX_MAC_HDR_LEN		127
10223 #define IXGBE_MAX_NETWORK_HDR_LEN	511
10224 
10225 static netdev_features_t
10226 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
10227 		     netdev_features_t features)
10228 {
10229 	unsigned int network_hdr_len, mac_hdr_len;
10230 
10231 	/* Make certain the headers can be described by a context descriptor */
10232 	mac_hdr_len = skb_network_offset(skb);
10233 	if (unlikely(mac_hdr_len > IXGBE_MAX_MAC_HDR_LEN))
10234 		return features & ~(NETIF_F_HW_CSUM |
10235 				    NETIF_F_SCTP_CRC |
10236 				    NETIF_F_GSO_UDP_L4 |
10237 				    NETIF_F_HW_VLAN_CTAG_TX |
10238 				    NETIF_F_TSO |
10239 				    NETIF_F_TSO6);
10240 
10241 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
10242 	if (unlikely(network_hdr_len >  IXGBE_MAX_NETWORK_HDR_LEN))
10243 		return features & ~(NETIF_F_HW_CSUM |
10244 				    NETIF_F_SCTP_CRC |
10245 				    NETIF_F_GSO_UDP_L4 |
10246 				    NETIF_F_TSO |
10247 				    NETIF_F_TSO6);
10248 
10249 	/* We can only support IPV4 TSO in tunnels if we can mangle the
10250 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
10251 	 * IPsec offoad sets skb->encapsulation but still can handle
10252 	 * the TSO, so it's the exception.
10253 	 */
10254 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID)) {
10255 #ifdef CONFIG_IXGBE_IPSEC
10256 		if (!secpath_exists(skb))
10257 #endif
10258 			features &= ~NETIF_F_TSO;
10259 	}
10260 
10261 	return features;
10262 }
10263 
10264 static int ixgbe_xdp_setup(struct net_device *dev, struct bpf_prog *prog)
10265 {
10266 	int i, frame_size = dev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
10267 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10268 	struct bpf_prog *old_prog;
10269 	bool need_reset;
10270 	int num_queues;
10271 
10272 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
10273 		return -EINVAL;
10274 
10275 	if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
10276 		return -EINVAL;
10277 
10278 	/* verify ixgbe ring attributes are sufficient for XDP */
10279 	for (i = 0; i < adapter->num_rx_queues; i++) {
10280 		struct ixgbe_ring *ring = adapter->rx_ring[i];
10281 
10282 		if (ring_is_rsc_enabled(ring))
10283 			return -EINVAL;
10284 
10285 		if (frame_size > ixgbe_rx_bufsz(ring))
10286 			return -EINVAL;
10287 	}
10288 
10289 	/* if the number of cpus is much larger than the maximum of queues,
10290 	 * we should stop it and then return with ENOMEM like before.
10291 	 */
10292 	if (nr_cpu_ids > IXGBE_MAX_XDP_QS * 2)
10293 		return -ENOMEM;
10294 
10295 	old_prog = xchg(&adapter->xdp_prog, prog);
10296 	need_reset = (!!prog != !!old_prog);
10297 
10298 	/* If transitioning XDP modes reconfigure rings */
10299 	if (need_reset) {
10300 		int err;
10301 
10302 		if (!prog)
10303 			/* Wait until ndo_xsk_wakeup completes. */
10304 			synchronize_rcu();
10305 		err = ixgbe_setup_tc(dev, adapter->hw_tcs);
10306 
10307 		if (err)
10308 			return -EINVAL;
10309 		if (!prog)
10310 			xdp_features_clear_redirect_target(dev);
10311 	} else {
10312 		for (i = 0; i < adapter->num_rx_queues; i++) {
10313 			WRITE_ONCE(adapter->rx_ring[i]->xdp_prog,
10314 				   adapter->xdp_prog);
10315 		}
10316 	}
10317 
10318 	if (old_prog)
10319 		bpf_prog_put(old_prog);
10320 
10321 	/* Kick start the NAPI context if there is an AF_XDP socket open
10322 	 * on that queue id. This so that receiving will start.
10323 	 */
10324 	if (need_reset && prog) {
10325 		num_queues = min_t(int, adapter->num_rx_queues,
10326 				   adapter->num_xdp_queues);
10327 		for (i = 0; i < num_queues; i++)
10328 			if (adapter->xdp_ring[i]->xsk_pool)
10329 				(void)ixgbe_xsk_wakeup(adapter->netdev, i,
10330 						       XDP_WAKEUP_RX);
10331 		xdp_features_set_redirect_target(dev, true);
10332 	}
10333 
10334 	return 0;
10335 }
10336 
10337 static int ixgbe_xdp(struct net_device *dev, struct netdev_bpf *xdp)
10338 {
10339 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10340 
10341 	switch (xdp->command) {
10342 	case XDP_SETUP_PROG:
10343 		return ixgbe_xdp_setup(dev, xdp->prog);
10344 	case XDP_SETUP_XSK_POOL:
10345 		return ixgbe_xsk_pool_setup(adapter, xdp->xsk.pool,
10346 					    xdp->xsk.queue_id);
10347 
10348 	default:
10349 		return -EINVAL;
10350 	}
10351 }
10352 
10353 void ixgbe_xdp_ring_update_tail(struct ixgbe_ring *ring)
10354 {
10355 	/* Force memory writes to complete before letting h/w know there
10356 	 * are new descriptors to fetch.
10357 	 */
10358 	wmb();
10359 	writel(ring->next_to_use, ring->tail);
10360 }
10361 
10362 void ixgbe_xdp_ring_update_tail_locked(struct ixgbe_ring *ring)
10363 {
10364 	if (static_branch_unlikely(&ixgbe_xdp_locking_key))
10365 		spin_lock(&ring->tx_lock);
10366 	ixgbe_xdp_ring_update_tail(ring);
10367 	if (static_branch_unlikely(&ixgbe_xdp_locking_key))
10368 		spin_unlock(&ring->tx_lock);
10369 }
10370 
10371 static int ixgbe_xdp_xmit(struct net_device *dev, int n,
10372 			  struct xdp_frame **frames, u32 flags)
10373 {
10374 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10375 	struct ixgbe_ring *ring;
10376 	int nxmit = 0;
10377 	int i;
10378 
10379 	if (unlikely(test_bit(__IXGBE_DOWN, &adapter->state)))
10380 		return -ENETDOWN;
10381 
10382 	if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
10383 		return -EINVAL;
10384 
10385 	/* During program transitions its possible adapter->xdp_prog is assigned
10386 	 * but ring has not been configured yet. In this case simply abort xmit.
10387 	 */
10388 	ring = adapter->xdp_prog ? ixgbe_determine_xdp_ring(adapter) : NULL;
10389 	if (unlikely(!ring))
10390 		return -ENXIO;
10391 
10392 	if (unlikely(test_bit(__IXGBE_TX_DISABLED, &ring->state)))
10393 		return -ENXIO;
10394 
10395 	if (static_branch_unlikely(&ixgbe_xdp_locking_key))
10396 		spin_lock(&ring->tx_lock);
10397 
10398 	for (i = 0; i < n; i++) {
10399 		struct xdp_frame *xdpf = frames[i];
10400 		int err;
10401 
10402 		err = ixgbe_xmit_xdp_ring(ring, xdpf);
10403 		if (err != IXGBE_XDP_TX)
10404 			break;
10405 		nxmit++;
10406 	}
10407 
10408 	if (unlikely(flags & XDP_XMIT_FLUSH))
10409 		ixgbe_xdp_ring_update_tail(ring);
10410 
10411 	if (static_branch_unlikely(&ixgbe_xdp_locking_key))
10412 		spin_unlock(&ring->tx_lock);
10413 
10414 	return nxmit;
10415 }
10416 
10417 static const struct net_device_ops ixgbe_netdev_ops = {
10418 	.ndo_open		= ixgbe_open,
10419 	.ndo_stop		= ixgbe_close,
10420 	.ndo_start_xmit		= ixgbe_xmit_frame,
10421 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
10422 	.ndo_validate_addr	= eth_validate_addr,
10423 	.ndo_set_mac_address	= ixgbe_set_mac,
10424 	.ndo_change_mtu		= ixgbe_change_mtu,
10425 	.ndo_tx_timeout		= ixgbe_tx_timeout,
10426 	.ndo_set_tx_maxrate	= ixgbe_tx_maxrate,
10427 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
10428 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
10429 	.ndo_eth_ioctl		= ixgbe_ioctl,
10430 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
10431 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
10432 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
10433 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
10434 	.ndo_set_vf_link_state	= ixgbe_ndo_set_vf_link_state,
10435 	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
10436 	.ndo_set_vf_trust	= ixgbe_ndo_set_vf_trust,
10437 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
10438 	.ndo_get_vf_stats	= ixgbe_ndo_get_vf_stats,
10439 	.ndo_get_stats64	= ixgbe_get_stats64,
10440 	.ndo_setup_tc		= __ixgbe_setup_tc,
10441 #ifdef IXGBE_FCOE
10442 	.ndo_select_queue	= ixgbe_select_queue,
10443 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
10444 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
10445 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
10446 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
10447 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
10448 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
10449 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
10450 #endif /* IXGBE_FCOE */
10451 	.ndo_set_features = ixgbe_set_features,
10452 	.ndo_fix_features = ixgbe_fix_features,
10453 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
10454 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
10455 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
10456 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
10457 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
10458 	.ndo_features_check	= ixgbe_features_check,
10459 	.ndo_bpf		= ixgbe_xdp,
10460 	.ndo_xdp_xmit		= ixgbe_xdp_xmit,
10461 	.ndo_xsk_wakeup         = ixgbe_xsk_wakeup,
10462 };
10463 
10464 static void ixgbe_disable_txr_hw(struct ixgbe_adapter *adapter,
10465 				 struct ixgbe_ring *tx_ring)
10466 {
10467 	unsigned long wait_delay, delay_interval;
10468 	struct ixgbe_hw *hw = &adapter->hw;
10469 	u8 reg_idx = tx_ring->reg_idx;
10470 	int wait_loop;
10471 	u32 txdctl;
10472 
10473 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
10474 
10475 	/* delay mechanism from ixgbe_disable_tx */
10476 	delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
10477 
10478 	wait_loop = IXGBE_MAX_RX_DESC_POLL;
10479 	wait_delay = delay_interval;
10480 
10481 	while (wait_loop--) {
10482 		usleep_range(wait_delay, wait_delay + 10);
10483 		wait_delay += delay_interval * 2;
10484 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
10485 
10486 		if (!(txdctl & IXGBE_TXDCTL_ENABLE))
10487 			return;
10488 	}
10489 
10490 	e_err(drv, "TXDCTL.ENABLE not cleared within the polling period\n");
10491 }
10492 
10493 static void ixgbe_disable_txr(struct ixgbe_adapter *adapter,
10494 			      struct ixgbe_ring *tx_ring)
10495 {
10496 	set_bit(__IXGBE_TX_DISABLED, &tx_ring->state);
10497 	ixgbe_disable_txr_hw(adapter, tx_ring);
10498 }
10499 
10500 static void ixgbe_disable_rxr_hw(struct ixgbe_adapter *adapter,
10501 				 struct ixgbe_ring *rx_ring)
10502 {
10503 	unsigned long wait_delay, delay_interval;
10504 	struct ixgbe_hw *hw = &adapter->hw;
10505 	u8 reg_idx = rx_ring->reg_idx;
10506 	int wait_loop;
10507 	u32 rxdctl;
10508 
10509 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
10510 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
10511 	rxdctl |= IXGBE_RXDCTL_SWFLSH;
10512 
10513 	/* write value back with RXDCTL.ENABLE bit cleared */
10514 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
10515 
10516 	/* RXDCTL.EN may not change on 82598 if link is down, so skip it */
10517 	if (hw->mac.type == ixgbe_mac_82598EB &&
10518 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
10519 		return;
10520 
10521 	/* delay mechanism from ixgbe_disable_rx */
10522 	delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
10523 
10524 	wait_loop = IXGBE_MAX_RX_DESC_POLL;
10525 	wait_delay = delay_interval;
10526 
10527 	while (wait_loop--) {
10528 		usleep_range(wait_delay, wait_delay + 10);
10529 		wait_delay += delay_interval * 2;
10530 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
10531 
10532 		if (!(rxdctl & IXGBE_RXDCTL_ENABLE))
10533 			return;
10534 	}
10535 
10536 	e_err(drv, "RXDCTL.ENABLE not cleared within the polling period\n");
10537 }
10538 
10539 static void ixgbe_reset_txr_stats(struct ixgbe_ring *tx_ring)
10540 {
10541 	memset(&tx_ring->stats, 0, sizeof(tx_ring->stats));
10542 	memset(&tx_ring->tx_stats, 0, sizeof(tx_ring->tx_stats));
10543 }
10544 
10545 static void ixgbe_reset_rxr_stats(struct ixgbe_ring *rx_ring)
10546 {
10547 	memset(&rx_ring->stats, 0, sizeof(rx_ring->stats));
10548 	memset(&rx_ring->rx_stats, 0, sizeof(rx_ring->rx_stats));
10549 }
10550 
10551 /**
10552  * ixgbe_irq_disable_single - Disable single IRQ vector
10553  * @adapter: adapter structure
10554  * @ring: ring index
10555  **/
10556 static void ixgbe_irq_disable_single(struct ixgbe_adapter *adapter, u32 ring)
10557 {
10558 	struct ixgbe_hw *hw = &adapter->hw;
10559 	u64 qmask = BIT_ULL(ring);
10560 	u32 mask;
10561 
10562 	switch (adapter->hw.mac.type) {
10563 	case ixgbe_mac_82598EB:
10564 		mask = qmask & IXGBE_EIMC_RTX_QUEUE;
10565 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, mask);
10566 		break;
10567 	case ixgbe_mac_82599EB:
10568 	case ixgbe_mac_X540:
10569 	case ixgbe_mac_X550:
10570 	case ixgbe_mac_X550EM_x:
10571 	case ixgbe_mac_x550em_a:
10572 		mask = (qmask & 0xFFFFFFFF);
10573 		if (mask)
10574 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
10575 		mask = (qmask >> 32);
10576 		if (mask)
10577 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
10578 		break;
10579 	default:
10580 		break;
10581 	}
10582 	IXGBE_WRITE_FLUSH(&adapter->hw);
10583 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
10584 		synchronize_irq(adapter->msix_entries[ring].vector);
10585 	else
10586 		synchronize_irq(adapter->pdev->irq);
10587 }
10588 
10589 /**
10590  * ixgbe_txrx_ring_disable - Disable Rx/Tx/XDP Tx rings
10591  * @adapter: adapter structure
10592  * @ring: ring index
10593  *
10594  * This function disables a certain Rx/Tx/XDP Tx ring. The function
10595  * assumes that the netdev is running.
10596  **/
10597 void ixgbe_txrx_ring_disable(struct ixgbe_adapter *adapter, int ring)
10598 {
10599 	struct ixgbe_ring *rx_ring, *tx_ring, *xdp_ring;
10600 
10601 	rx_ring = adapter->rx_ring[ring];
10602 	tx_ring = adapter->tx_ring[ring];
10603 	xdp_ring = adapter->xdp_ring[ring];
10604 
10605 	ixgbe_irq_disable_single(adapter, ring);
10606 
10607 	/* Rx/Tx/XDP Tx share the same napi context. */
10608 	napi_disable(&rx_ring->q_vector->napi);
10609 
10610 	ixgbe_disable_txr(adapter, tx_ring);
10611 	if (xdp_ring)
10612 		ixgbe_disable_txr(adapter, xdp_ring);
10613 	ixgbe_disable_rxr_hw(adapter, rx_ring);
10614 
10615 	if (xdp_ring)
10616 		synchronize_rcu();
10617 
10618 	ixgbe_clean_tx_ring(tx_ring);
10619 	if (xdp_ring)
10620 		ixgbe_clean_tx_ring(xdp_ring);
10621 	ixgbe_clean_rx_ring(rx_ring);
10622 
10623 	ixgbe_reset_txr_stats(tx_ring);
10624 	if (xdp_ring)
10625 		ixgbe_reset_txr_stats(xdp_ring);
10626 	ixgbe_reset_rxr_stats(rx_ring);
10627 }
10628 
10629 /**
10630  * ixgbe_txrx_ring_enable - Enable Rx/Tx/XDP Tx rings
10631  * @adapter: adapter structure
10632  * @ring: ring index
10633  *
10634  * This function enables a certain Rx/Tx/XDP Tx ring. The function
10635  * assumes that the netdev is running.
10636  **/
10637 void ixgbe_txrx_ring_enable(struct ixgbe_adapter *adapter, int ring)
10638 {
10639 	struct ixgbe_ring *rx_ring, *tx_ring, *xdp_ring;
10640 
10641 	rx_ring = adapter->rx_ring[ring];
10642 	tx_ring = adapter->tx_ring[ring];
10643 	xdp_ring = adapter->xdp_ring[ring];
10644 
10645 	ixgbe_configure_tx_ring(adapter, tx_ring);
10646 	if (xdp_ring)
10647 		ixgbe_configure_tx_ring(adapter, xdp_ring);
10648 	ixgbe_configure_rx_ring(adapter, rx_ring);
10649 
10650 	clear_bit(__IXGBE_TX_DISABLED, &tx_ring->state);
10651 	if (xdp_ring)
10652 		clear_bit(__IXGBE_TX_DISABLED, &xdp_ring->state);
10653 
10654 	/* Rx/Tx/XDP Tx share the same napi context. */
10655 	napi_enable(&rx_ring->q_vector->napi);
10656 	ixgbe_irq_enable_queues(adapter, BIT_ULL(ring));
10657 	IXGBE_WRITE_FLUSH(&adapter->hw);
10658 }
10659 
10660 /**
10661  * ixgbe_enumerate_functions - Get the number of ports this device has
10662  * @adapter: adapter structure
10663  *
10664  * This function enumerates the phsyical functions co-located on a single slot,
10665  * in order to determine how many ports a device has. This is most useful in
10666  * determining the required GT/s of PCIe bandwidth necessary for optimal
10667  * performance.
10668  **/
10669 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
10670 {
10671 	struct pci_dev *entry, *pdev = adapter->pdev;
10672 	int physfns = 0;
10673 
10674 	/* Some cards can not use the generic count PCIe functions method,
10675 	 * because they are behind a parent switch, so we hardcode these with
10676 	 * the correct number of functions.
10677 	 */
10678 	if (ixgbe_pcie_from_parent(&adapter->hw))
10679 		physfns = 4;
10680 
10681 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
10682 		/* don't count virtual functions */
10683 		if (entry->is_virtfn)
10684 			continue;
10685 
10686 		/* When the devices on the bus don't all match our device ID,
10687 		 * we can't reliably determine the correct number of
10688 		 * functions. This can occur if a function has been direct
10689 		 * attached to a virtual machine using VT-d, for example. In
10690 		 * this case, simply return -1 to indicate this.
10691 		 */
10692 		if ((entry->vendor != pdev->vendor) ||
10693 		    (entry->device != pdev->device))
10694 			return -1;
10695 
10696 		physfns++;
10697 	}
10698 
10699 	return physfns;
10700 }
10701 
10702 /**
10703  * ixgbe_wol_supported - Check whether device supports WoL
10704  * @adapter: the adapter private structure
10705  * @device_id: the device ID
10706  * @subdevice_id: the subsystem device ID
10707  *
10708  * This function is used by probe and ethtool to determine
10709  * which devices have WoL support
10710  *
10711  **/
10712 bool ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
10713 			 u16 subdevice_id)
10714 {
10715 	struct ixgbe_hw *hw = &adapter->hw;
10716 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
10717 
10718 	/* WOL not supported on 82598 */
10719 	if (hw->mac.type == ixgbe_mac_82598EB)
10720 		return false;
10721 
10722 	/* check eeprom to see if WOL is enabled for X540 and newer */
10723 	if (hw->mac.type >= ixgbe_mac_X540) {
10724 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
10725 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
10726 		     (hw->bus.func == 0)))
10727 			return true;
10728 	}
10729 
10730 	/* WOL is determined based on device IDs for 82599 MACs */
10731 	switch (device_id) {
10732 	case IXGBE_DEV_ID_82599_SFP:
10733 		/* Only these subdevices could supports WOL */
10734 		switch (subdevice_id) {
10735 		case IXGBE_SUBDEV_ID_82599_560FLR:
10736 		case IXGBE_SUBDEV_ID_82599_LOM_SNAP6:
10737 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
10738 		case IXGBE_SUBDEV_ID_82599_SFP_2OCP:
10739 			/* only support first port */
10740 			if (hw->bus.func != 0)
10741 				break;
10742 			fallthrough;
10743 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
10744 		case IXGBE_SUBDEV_ID_82599_SFP:
10745 		case IXGBE_SUBDEV_ID_82599_RNDC:
10746 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
10747 		case IXGBE_SUBDEV_ID_82599_SFP_1OCP:
10748 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM1:
10749 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM2:
10750 			return true;
10751 		}
10752 		break;
10753 	case IXGBE_DEV_ID_82599EN_SFP:
10754 		/* Only these subdevices support WOL */
10755 		switch (subdevice_id) {
10756 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
10757 			return true;
10758 		}
10759 		break;
10760 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
10761 		/* All except this subdevice support WOL */
10762 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
10763 			return true;
10764 		break;
10765 	case IXGBE_DEV_ID_82599_KX4:
10766 		return  true;
10767 	default:
10768 		break;
10769 	}
10770 
10771 	return false;
10772 }
10773 
10774 /**
10775  * ixgbe_set_fw_version - Set FW version
10776  * @adapter: the adapter private structure
10777  *
10778  * This function is used by probe and ethtool to determine the FW version to
10779  * format to display. The FW version is taken from the EEPROM/NVM.
10780  */
10781 static void ixgbe_set_fw_version(struct ixgbe_adapter *adapter)
10782 {
10783 	struct ixgbe_hw *hw = &adapter->hw;
10784 	struct ixgbe_nvm_version nvm_ver;
10785 
10786 	ixgbe_get_oem_prod_version(hw, &nvm_ver);
10787 	if (nvm_ver.oem_valid) {
10788 		snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10789 			 "%x.%x.%x", nvm_ver.oem_major, nvm_ver.oem_minor,
10790 			 nvm_ver.oem_release);
10791 		return;
10792 	}
10793 
10794 	ixgbe_get_etk_id(hw, &nvm_ver);
10795 	ixgbe_get_orom_version(hw, &nvm_ver);
10796 
10797 	if (nvm_ver.or_valid) {
10798 		snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10799 			 "0x%08x, %d.%d.%d", nvm_ver.etk_id, nvm_ver.or_major,
10800 			 nvm_ver.or_build, nvm_ver.or_patch);
10801 		return;
10802 	}
10803 
10804 	/* Set ETrack ID format */
10805 	snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10806 		 "0x%08x", nvm_ver.etk_id);
10807 }
10808 
10809 /**
10810  * ixgbe_probe - Device Initialization Routine
10811  * @pdev: PCI device information struct
10812  * @ent: entry in ixgbe_pci_tbl
10813  *
10814  * Returns 0 on success, negative on failure
10815  *
10816  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
10817  * The OS initialization, configuring of the adapter private structure,
10818  * and a hardware reset occur.
10819  **/
10820 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10821 {
10822 	struct net_device *netdev;
10823 	struct ixgbe_adapter *adapter = NULL;
10824 	struct ixgbe_hw *hw;
10825 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
10826 	unsigned int indices = MAX_TX_QUEUES;
10827 	u8 part_str[IXGBE_PBANUM_LENGTH];
10828 	int i, err, expected_gts;
10829 	bool disable_dev = false;
10830 #ifdef IXGBE_FCOE
10831 	u16 device_caps;
10832 #endif
10833 	u32 eec;
10834 
10835 	/* Catch broken hardware that put the wrong VF device ID in
10836 	 * the PCIe SR-IOV capability.
10837 	 */
10838 	if (pdev->is_virtfn) {
10839 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
10840 		     pci_name(pdev), pdev->vendor, pdev->device);
10841 		return -EINVAL;
10842 	}
10843 
10844 	err = pci_enable_device_mem(pdev);
10845 	if (err)
10846 		return err;
10847 
10848 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
10849 	if (err) {
10850 		dev_err(&pdev->dev,
10851 			"No usable DMA configuration, aborting\n");
10852 		goto err_dma;
10853 	}
10854 
10855 	err = pci_request_mem_regions(pdev, ixgbe_driver_name);
10856 	if (err) {
10857 		dev_err(&pdev->dev,
10858 			"pci_request_selected_regions failed 0x%x\n", err);
10859 		goto err_pci_reg;
10860 	}
10861 
10862 	pci_set_master(pdev);
10863 	pci_save_state(pdev);
10864 
10865 	if (ii->mac == ixgbe_mac_82598EB) {
10866 #ifdef CONFIG_IXGBE_DCB
10867 		/* 8 TC w/ 4 queues per TC */
10868 		indices = 4 * MAX_TRAFFIC_CLASS;
10869 #else
10870 		indices = IXGBE_MAX_RSS_INDICES;
10871 #endif
10872 	}
10873 
10874 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
10875 	if (!netdev) {
10876 		err = -ENOMEM;
10877 		goto err_alloc_etherdev;
10878 	}
10879 
10880 	SET_NETDEV_DEV(netdev, &pdev->dev);
10881 
10882 	adapter = netdev_priv(netdev);
10883 
10884 	adapter->netdev = netdev;
10885 	adapter->pdev = pdev;
10886 	hw = &adapter->hw;
10887 	hw->back = adapter;
10888 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
10889 
10890 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
10891 			      pci_resource_len(pdev, 0));
10892 	adapter->io_addr = hw->hw_addr;
10893 	if (!hw->hw_addr) {
10894 		err = -EIO;
10895 		goto err_ioremap;
10896 	}
10897 
10898 	netdev->netdev_ops = &ixgbe_netdev_ops;
10899 	ixgbe_set_ethtool_ops(netdev);
10900 	netdev->watchdog_timeo = 5 * HZ;
10901 	strscpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
10902 
10903 	/* Setup hw api */
10904 	hw->mac.ops   = *ii->mac_ops;
10905 	hw->mac.type  = ii->mac;
10906 	hw->mvals     = ii->mvals;
10907 	if (ii->link_ops)
10908 		hw->link.ops  = *ii->link_ops;
10909 
10910 	/* EEPROM */
10911 	hw->eeprom.ops = *ii->eeprom_ops;
10912 	eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
10913 	if (ixgbe_removed(hw->hw_addr)) {
10914 		err = -EIO;
10915 		goto err_ioremap;
10916 	}
10917 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
10918 	if (!(eec & BIT(8)))
10919 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
10920 
10921 	/* PHY */
10922 	hw->phy.ops = *ii->phy_ops;
10923 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
10924 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
10925 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
10926 	hw->phy.mdio.mmds = 0;
10927 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
10928 	hw->phy.mdio.dev = netdev;
10929 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
10930 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
10931 
10932 	/* setup the private structure */
10933 	err = ixgbe_sw_init(adapter, ii);
10934 	if (err)
10935 		goto err_sw_init;
10936 
10937 	if (adapter->hw.mac.type == ixgbe_mac_82599EB)
10938 		adapter->flags2 |= IXGBE_FLAG2_AUTO_DISABLE_VF;
10939 
10940 	switch (adapter->hw.mac.type) {
10941 	case ixgbe_mac_X550:
10942 	case ixgbe_mac_X550EM_x:
10943 		netdev->udp_tunnel_nic_info = &ixgbe_udp_tunnels_x550;
10944 		break;
10945 	case ixgbe_mac_x550em_a:
10946 		netdev->udp_tunnel_nic_info = &ixgbe_udp_tunnels_x550em_a;
10947 		break;
10948 	default:
10949 		break;
10950 	}
10951 
10952 	/* Make sure the SWFW semaphore is in a valid state */
10953 	if (hw->mac.ops.init_swfw_sync)
10954 		hw->mac.ops.init_swfw_sync(hw);
10955 
10956 	/* Make it possible the adapter to be woken up via WOL */
10957 	switch (adapter->hw.mac.type) {
10958 	case ixgbe_mac_82599EB:
10959 	case ixgbe_mac_X540:
10960 	case ixgbe_mac_X550:
10961 	case ixgbe_mac_X550EM_x:
10962 	case ixgbe_mac_x550em_a:
10963 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10964 		break;
10965 	default:
10966 		break;
10967 	}
10968 
10969 	/*
10970 	 * If there is a fan on this device and it has failed log the
10971 	 * failure.
10972 	 */
10973 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
10974 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
10975 		if (esdp & IXGBE_ESDP_SDP1)
10976 			e_crit(probe, "Fan has stopped, replace the adapter\n");
10977 	}
10978 
10979 	if (allow_unsupported_sfp)
10980 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
10981 
10982 	/* reset_hw fills in the perm_addr as well */
10983 	hw->phy.reset_if_overtemp = true;
10984 	err = hw->mac.ops.reset_hw(hw);
10985 	hw->phy.reset_if_overtemp = false;
10986 	ixgbe_set_eee_capable(adapter);
10987 	if (err == -ENOENT) {
10988 		err = 0;
10989 	} else if (err == -EOPNOTSUPP) {
10990 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
10991 		e_dev_err("Reload the driver after installing a supported module.\n");
10992 		goto err_sw_init;
10993 	} else if (err) {
10994 		e_dev_err("HW Init failed: %d\n", err);
10995 		goto err_sw_init;
10996 	}
10997 
10998 #ifdef CONFIG_PCI_IOV
10999 	/* SR-IOV not supported on the 82598 */
11000 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
11001 		goto skip_sriov;
11002 	/* Mailbox */
11003 	ixgbe_init_mbx_params_pf(hw);
11004 	hw->mbx.ops = ii->mbx_ops;
11005 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
11006 	ixgbe_enable_sriov(adapter, max_vfs);
11007 skip_sriov:
11008 
11009 #endif
11010 	netdev->features = NETIF_F_SG |
11011 			   NETIF_F_TSO |
11012 			   NETIF_F_TSO6 |
11013 			   NETIF_F_RXHASH |
11014 			   NETIF_F_RXCSUM |
11015 			   NETIF_F_HW_CSUM;
11016 
11017 #define IXGBE_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
11018 				    NETIF_F_GSO_GRE_CSUM | \
11019 				    NETIF_F_GSO_IPXIP4 | \
11020 				    NETIF_F_GSO_IPXIP6 | \
11021 				    NETIF_F_GSO_UDP_TUNNEL | \
11022 				    NETIF_F_GSO_UDP_TUNNEL_CSUM)
11023 
11024 	netdev->gso_partial_features = IXGBE_GSO_PARTIAL_FEATURES;
11025 	netdev->features |= NETIF_F_GSO_PARTIAL |
11026 			    IXGBE_GSO_PARTIAL_FEATURES;
11027 
11028 	if (hw->mac.type >= ixgbe_mac_82599EB)
11029 		netdev->features |= NETIF_F_SCTP_CRC | NETIF_F_GSO_UDP_L4;
11030 
11031 #ifdef CONFIG_IXGBE_IPSEC
11032 #define IXGBE_ESP_FEATURES	(NETIF_F_HW_ESP | \
11033 				 NETIF_F_HW_ESP_TX_CSUM | \
11034 				 NETIF_F_GSO_ESP)
11035 
11036 	if (adapter->ipsec)
11037 		netdev->features |= IXGBE_ESP_FEATURES;
11038 #endif
11039 	/* copy netdev features into list of user selectable features */
11040 	netdev->hw_features |= netdev->features |
11041 			       NETIF_F_HW_VLAN_CTAG_FILTER |
11042 			       NETIF_F_HW_VLAN_CTAG_RX |
11043 			       NETIF_F_HW_VLAN_CTAG_TX |
11044 			       NETIF_F_RXALL |
11045 			       NETIF_F_HW_L2FW_DOFFLOAD;
11046 
11047 	if (hw->mac.type >= ixgbe_mac_82599EB)
11048 		netdev->hw_features |= NETIF_F_NTUPLE |
11049 				       NETIF_F_HW_TC;
11050 
11051 	netdev->features |= NETIF_F_HIGHDMA;
11052 
11053 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
11054 	netdev->hw_enc_features |= netdev->vlan_features;
11055 	netdev->mpls_features |= NETIF_F_SG |
11056 				 NETIF_F_TSO |
11057 				 NETIF_F_TSO6 |
11058 				 NETIF_F_HW_CSUM;
11059 	netdev->mpls_features |= IXGBE_GSO_PARTIAL_FEATURES;
11060 
11061 	/* set this bit last since it cannot be part of vlan_features */
11062 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
11063 			    NETIF_F_HW_VLAN_CTAG_RX |
11064 			    NETIF_F_HW_VLAN_CTAG_TX;
11065 
11066 	netdev->priv_flags |= IFF_UNICAST_FLT;
11067 	netdev->priv_flags |= IFF_SUPP_NOFCS;
11068 
11069 	netdev->xdp_features = NETDEV_XDP_ACT_BASIC | NETDEV_XDP_ACT_REDIRECT |
11070 			       NETDEV_XDP_ACT_XSK_ZEROCOPY;
11071 
11072 	/* MTU range: 68 - 9710 */
11073 	netdev->min_mtu = ETH_MIN_MTU;
11074 	netdev->max_mtu = IXGBE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN);
11075 
11076 #ifdef CONFIG_IXGBE_DCB
11077 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
11078 		netdev->dcbnl_ops = &ixgbe_dcbnl_ops;
11079 #endif
11080 
11081 #ifdef IXGBE_FCOE
11082 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
11083 		unsigned int fcoe_l;
11084 
11085 		if (hw->mac.ops.get_device_caps) {
11086 			hw->mac.ops.get_device_caps(hw, &device_caps);
11087 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
11088 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
11089 		}
11090 
11091 
11092 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
11093 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
11094 
11095 		netdev->features |= NETIF_F_FSO |
11096 				    NETIF_F_FCOE_CRC;
11097 
11098 		netdev->vlan_features |= NETIF_F_FSO |
11099 					 NETIF_F_FCOE_CRC |
11100 					 NETIF_F_FCOE_MTU;
11101 	}
11102 #endif /* IXGBE_FCOE */
11103 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
11104 		netdev->hw_features |= NETIF_F_LRO;
11105 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
11106 		netdev->features |= NETIF_F_LRO;
11107 
11108 	if (ixgbe_check_fw_error(adapter)) {
11109 		err = -EIO;
11110 		goto err_sw_init;
11111 	}
11112 
11113 	/* make sure the EEPROM is good */
11114 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
11115 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
11116 		err = -EIO;
11117 		goto err_sw_init;
11118 	}
11119 
11120 	eth_platform_get_mac_address(&adapter->pdev->dev,
11121 				     adapter->hw.mac.perm_addr);
11122 
11123 	eth_hw_addr_set(netdev, hw->mac.perm_addr);
11124 
11125 	if (!is_valid_ether_addr(netdev->dev_addr)) {
11126 		e_dev_err("invalid MAC address\n");
11127 		err = -EIO;
11128 		goto err_sw_init;
11129 	}
11130 
11131 	/* Set hw->mac.addr to permanent MAC address */
11132 	ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
11133 	ixgbe_mac_set_default_filter(adapter);
11134 
11135 	timer_setup(&adapter->service_timer, ixgbe_service_timer, 0);
11136 
11137 	if (ixgbe_removed(hw->hw_addr)) {
11138 		err = -EIO;
11139 		goto err_sw_init;
11140 	}
11141 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
11142 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
11143 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
11144 
11145 	err = ixgbe_init_interrupt_scheme(adapter);
11146 	if (err)
11147 		goto err_sw_init;
11148 
11149 	for (i = 0; i < adapter->num_rx_queues; i++)
11150 		u64_stats_init(&adapter->rx_ring[i]->syncp);
11151 	for (i = 0; i < adapter->num_tx_queues; i++)
11152 		u64_stats_init(&adapter->tx_ring[i]->syncp);
11153 	for (i = 0; i < adapter->num_xdp_queues; i++)
11154 		u64_stats_init(&adapter->xdp_ring[i]->syncp);
11155 
11156 	/* WOL not supported for all devices */
11157 	adapter->wol = 0;
11158 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
11159 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
11160 						pdev->subsystem_device);
11161 	if (hw->wol_enabled)
11162 		adapter->wol = IXGBE_WUFC_MAG;
11163 
11164 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
11165 
11166 	/* save off EEPROM version number */
11167 	ixgbe_set_fw_version(adapter);
11168 
11169 	/* pick up the PCI bus settings for reporting later */
11170 	if (ixgbe_pcie_from_parent(hw))
11171 		ixgbe_get_parent_bus_info(adapter);
11172 	else
11173 		 hw->mac.ops.get_bus_info(hw);
11174 
11175 	/* calculate the expected PCIe bandwidth required for optimal
11176 	 * performance. Note that some older parts will never have enough
11177 	 * bandwidth due to being older generation PCIe parts. We clamp these
11178 	 * parts to ensure no warning is displayed if it can't be fixed.
11179 	 */
11180 	switch (hw->mac.type) {
11181 	case ixgbe_mac_82598EB:
11182 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
11183 		break;
11184 	default:
11185 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
11186 		break;
11187 	}
11188 
11189 	/* don't check link if we failed to enumerate functions */
11190 	if (expected_gts > 0)
11191 		ixgbe_check_minimum_link(adapter, expected_gts);
11192 
11193 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
11194 	if (err)
11195 		strscpy(part_str, "Unknown", sizeof(part_str));
11196 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
11197 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
11198 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
11199 			   part_str);
11200 	else
11201 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
11202 			   hw->mac.type, hw->phy.type, part_str);
11203 
11204 	e_dev_info("%pM\n", netdev->dev_addr);
11205 
11206 	/* reset the hardware with the new settings */
11207 	err = hw->mac.ops.start_hw(hw);
11208 	if (err == -EACCES) {
11209 		/* We are running on a pre-production device, log a warning */
11210 		e_dev_warn("This device is a pre-production adapter/LOM. "
11211 			   "Please be aware there may be issues associated "
11212 			   "with your hardware.  If you are experiencing "
11213 			   "problems please contact your Intel or hardware "
11214 			   "representative who provided you with this "
11215 			   "hardware.\n");
11216 	}
11217 	strcpy(netdev->name, "eth%d");
11218 	pci_set_drvdata(pdev, adapter);
11219 	err = register_netdev(netdev);
11220 	if (err)
11221 		goto err_register;
11222 
11223 
11224 	/* power down the optics for 82599 SFP+ fiber */
11225 	if (hw->mac.ops.disable_tx_laser)
11226 		hw->mac.ops.disable_tx_laser(hw);
11227 
11228 	/* carrier off reporting is important to ethtool even BEFORE open */
11229 	netif_carrier_off(netdev);
11230 
11231 #ifdef CONFIG_IXGBE_DCA
11232 	if (dca_add_requester(&pdev->dev) == 0) {
11233 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
11234 		ixgbe_setup_dca(adapter);
11235 	}
11236 #endif
11237 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
11238 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
11239 		for (i = 0; i < adapter->num_vfs; i++)
11240 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
11241 	}
11242 
11243 	/* firmware requires driver version to be 0xFFFFFFFF
11244 	 * since os does not support feature
11245 	 */
11246 	if (hw->mac.ops.set_fw_drv_ver)
11247 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF, 0xFF,
11248 					   sizeof(UTS_RELEASE) - 1,
11249 					   UTS_RELEASE);
11250 
11251 	/* add san mac addr to netdev */
11252 	ixgbe_add_sanmac_netdev(netdev);
11253 
11254 	e_dev_info("%s\n", ixgbe_default_device_descr);
11255 
11256 #ifdef CONFIG_IXGBE_HWMON
11257 	if (ixgbe_sysfs_init(adapter))
11258 		e_err(probe, "failed to allocate sysfs resources\n");
11259 #endif /* CONFIG_IXGBE_HWMON */
11260 
11261 	ixgbe_dbg_adapter_init(adapter);
11262 
11263 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
11264 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
11265 		hw->mac.ops.setup_link(hw,
11266 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
11267 			true);
11268 
11269 	err = ixgbe_mii_bus_init(hw);
11270 	if (err)
11271 		goto err_netdev;
11272 
11273 	return 0;
11274 
11275 err_netdev:
11276 	unregister_netdev(netdev);
11277 err_register:
11278 	ixgbe_release_hw_control(adapter);
11279 	ixgbe_clear_interrupt_scheme(adapter);
11280 err_sw_init:
11281 	ixgbe_disable_sriov(adapter);
11282 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
11283 	iounmap(adapter->io_addr);
11284 	kfree(adapter->jump_tables[0]);
11285 	kfree(adapter->mac_table);
11286 	kfree(adapter->rss_key);
11287 	bitmap_free(adapter->af_xdp_zc_qps);
11288 err_ioremap:
11289 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
11290 	free_netdev(netdev);
11291 err_alloc_etherdev:
11292 	pci_release_mem_regions(pdev);
11293 err_pci_reg:
11294 err_dma:
11295 	if (!adapter || disable_dev)
11296 		pci_disable_device(pdev);
11297 	return err;
11298 }
11299 
11300 /**
11301  * ixgbe_remove - Device Removal Routine
11302  * @pdev: PCI device information struct
11303  *
11304  * ixgbe_remove is called by the PCI subsystem to alert the driver
11305  * that it should release a PCI device.  The could be caused by a
11306  * Hot-Plug event, or because the driver is going to be removed from
11307  * memory.
11308  **/
11309 static void ixgbe_remove(struct pci_dev *pdev)
11310 {
11311 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11312 	struct net_device *netdev;
11313 	bool disable_dev;
11314 	int i;
11315 
11316 	/* if !adapter then we already cleaned up in probe */
11317 	if (!adapter)
11318 		return;
11319 
11320 	netdev  = adapter->netdev;
11321 	ixgbe_dbg_adapter_exit(adapter);
11322 
11323 	set_bit(__IXGBE_REMOVING, &adapter->state);
11324 	cancel_work_sync(&adapter->service_task);
11325 
11326 	if (adapter->mii_bus)
11327 		mdiobus_unregister(adapter->mii_bus);
11328 
11329 #ifdef CONFIG_IXGBE_DCA
11330 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
11331 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
11332 		dca_remove_requester(&pdev->dev);
11333 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
11334 				IXGBE_DCA_CTRL_DCA_DISABLE);
11335 	}
11336 
11337 #endif
11338 #ifdef CONFIG_IXGBE_HWMON
11339 	ixgbe_sysfs_exit(adapter);
11340 #endif /* CONFIG_IXGBE_HWMON */
11341 
11342 	/* remove the added san mac */
11343 	ixgbe_del_sanmac_netdev(netdev);
11344 
11345 #ifdef CONFIG_PCI_IOV
11346 	ixgbe_disable_sriov(adapter);
11347 #endif
11348 	if (netdev->reg_state == NETREG_REGISTERED)
11349 		unregister_netdev(netdev);
11350 
11351 	ixgbe_stop_ipsec_offload(adapter);
11352 	ixgbe_clear_interrupt_scheme(adapter);
11353 
11354 	ixgbe_release_hw_control(adapter);
11355 
11356 #ifdef CONFIG_DCB
11357 	kfree(adapter->ixgbe_ieee_pfc);
11358 	kfree(adapter->ixgbe_ieee_ets);
11359 
11360 #endif
11361 	iounmap(adapter->io_addr);
11362 	pci_release_mem_regions(pdev);
11363 
11364 	e_dev_info("complete\n");
11365 
11366 	for (i = 0; i < IXGBE_MAX_LINK_HANDLE; i++) {
11367 		if (adapter->jump_tables[i]) {
11368 			kfree(adapter->jump_tables[i]->input);
11369 			kfree(adapter->jump_tables[i]->mask);
11370 		}
11371 		kfree(adapter->jump_tables[i]);
11372 	}
11373 
11374 	kfree(adapter->mac_table);
11375 	kfree(adapter->rss_key);
11376 	bitmap_free(adapter->af_xdp_zc_qps);
11377 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
11378 	free_netdev(netdev);
11379 
11380 	if (disable_dev)
11381 		pci_disable_device(pdev);
11382 }
11383 
11384 /**
11385  * ixgbe_io_error_detected - called when PCI error is detected
11386  * @pdev: Pointer to PCI device
11387  * @state: The current pci connection state
11388  *
11389  * This function is called after a PCI bus error affecting
11390  * this device has been detected.
11391  */
11392 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
11393 						pci_channel_state_t state)
11394 {
11395 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11396 	struct net_device *netdev = adapter->netdev;
11397 
11398 #ifdef CONFIG_PCI_IOV
11399 	struct ixgbe_hw *hw = &adapter->hw;
11400 	struct pci_dev *bdev, *vfdev;
11401 	u32 dw0, dw1, dw2, dw3;
11402 	int vf, pos;
11403 	u16 req_id, pf_func;
11404 
11405 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
11406 	    adapter->num_vfs == 0)
11407 		goto skip_bad_vf_detection;
11408 
11409 	bdev = pdev->bus->self;
11410 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
11411 		bdev = bdev->bus->self;
11412 
11413 	if (!bdev)
11414 		goto skip_bad_vf_detection;
11415 
11416 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
11417 	if (!pos)
11418 		goto skip_bad_vf_detection;
11419 
11420 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
11421 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
11422 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
11423 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
11424 	if (ixgbe_removed(hw->hw_addr))
11425 		goto skip_bad_vf_detection;
11426 
11427 	req_id = dw1 >> 16;
11428 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
11429 	if (!(req_id & 0x0080))
11430 		goto skip_bad_vf_detection;
11431 
11432 	pf_func = req_id & 0x01;
11433 	if ((pf_func & 1) == (pdev->devfn & 1)) {
11434 		unsigned int device_id;
11435 
11436 		vf = FIELD_GET(0x7F, req_id);
11437 		e_dev_err("VF %d has caused a PCIe error\n", vf);
11438 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
11439 				"%8.8x\tdw3: %8.8x\n",
11440 		dw0, dw1, dw2, dw3);
11441 		switch (adapter->hw.mac.type) {
11442 		case ixgbe_mac_82599EB:
11443 			device_id = IXGBE_82599_VF_DEVICE_ID;
11444 			break;
11445 		case ixgbe_mac_X540:
11446 			device_id = IXGBE_X540_VF_DEVICE_ID;
11447 			break;
11448 		case ixgbe_mac_X550:
11449 			device_id = IXGBE_DEV_ID_X550_VF;
11450 			break;
11451 		case ixgbe_mac_X550EM_x:
11452 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
11453 			break;
11454 		case ixgbe_mac_x550em_a:
11455 			device_id = IXGBE_DEV_ID_X550EM_A_VF;
11456 			break;
11457 		default:
11458 			device_id = 0;
11459 			break;
11460 		}
11461 
11462 		/* Find the pci device of the offending VF */
11463 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
11464 		while (vfdev) {
11465 			if (vfdev->devfn == (req_id & 0xFF))
11466 				break;
11467 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
11468 					       device_id, vfdev);
11469 		}
11470 		/*
11471 		 * There's a slim chance the VF could have been hot plugged,
11472 		 * so if it is no longer present we don't need to issue the
11473 		 * VFLR.  Just clean up the AER in that case.
11474 		 */
11475 		if (vfdev) {
11476 			pcie_flr(vfdev);
11477 			/* Free device reference count */
11478 			pci_dev_put(vfdev);
11479 		}
11480 	}
11481 
11482 	/*
11483 	 * Even though the error may have occurred on the other port
11484 	 * we still need to increment the vf error reference count for
11485 	 * both ports because the I/O resume function will be called
11486 	 * for both of them.
11487 	 */
11488 	adapter->vferr_refcount++;
11489 
11490 	return PCI_ERS_RESULT_RECOVERED;
11491 
11492 skip_bad_vf_detection:
11493 #endif /* CONFIG_PCI_IOV */
11494 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
11495 		return PCI_ERS_RESULT_DISCONNECT;
11496 
11497 	if (!netif_device_present(netdev))
11498 		return PCI_ERS_RESULT_DISCONNECT;
11499 
11500 	rtnl_lock();
11501 	netif_device_detach(netdev);
11502 
11503 	if (netif_running(netdev))
11504 		ixgbe_close_suspend(adapter);
11505 
11506 	if (state == pci_channel_io_perm_failure) {
11507 		rtnl_unlock();
11508 		return PCI_ERS_RESULT_DISCONNECT;
11509 	}
11510 
11511 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
11512 		pci_disable_device(pdev);
11513 	rtnl_unlock();
11514 
11515 	/* Request a slot reset. */
11516 	return PCI_ERS_RESULT_NEED_RESET;
11517 }
11518 
11519 /**
11520  * ixgbe_io_slot_reset - called after the pci bus has been reset.
11521  * @pdev: Pointer to PCI device
11522  *
11523  * Restart the card from scratch, as if from a cold-boot.
11524  */
11525 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
11526 {
11527 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11528 	pci_ers_result_t result;
11529 
11530 	if (pci_enable_device_mem(pdev)) {
11531 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
11532 		result = PCI_ERS_RESULT_DISCONNECT;
11533 	} else {
11534 		smp_mb__before_atomic();
11535 		clear_bit(__IXGBE_DISABLED, &adapter->state);
11536 		adapter->hw.hw_addr = adapter->io_addr;
11537 		pci_set_master(pdev);
11538 		pci_restore_state(pdev);
11539 		pci_save_state(pdev);
11540 
11541 		pci_wake_from_d3(pdev, false);
11542 
11543 		ixgbe_reset(adapter);
11544 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
11545 		result = PCI_ERS_RESULT_RECOVERED;
11546 	}
11547 
11548 	return result;
11549 }
11550 
11551 /**
11552  * ixgbe_io_resume - called when traffic can start flowing again.
11553  * @pdev: Pointer to PCI device
11554  *
11555  * This callback is called when the error recovery driver tells us that
11556  * its OK to resume normal operation.
11557  */
11558 static void ixgbe_io_resume(struct pci_dev *pdev)
11559 {
11560 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11561 	struct net_device *netdev = adapter->netdev;
11562 
11563 #ifdef CONFIG_PCI_IOV
11564 	if (adapter->vferr_refcount) {
11565 		e_info(drv, "Resuming after VF err\n");
11566 		adapter->vferr_refcount--;
11567 		return;
11568 	}
11569 
11570 #endif
11571 	rtnl_lock();
11572 	if (netif_running(netdev))
11573 		ixgbe_open(netdev);
11574 
11575 	netif_device_attach(netdev);
11576 	rtnl_unlock();
11577 }
11578 
11579 static const struct pci_error_handlers ixgbe_err_handler = {
11580 	.error_detected = ixgbe_io_error_detected,
11581 	.slot_reset = ixgbe_io_slot_reset,
11582 	.resume = ixgbe_io_resume,
11583 };
11584 
11585 static DEFINE_SIMPLE_DEV_PM_OPS(ixgbe_pm_ops, ixgbe_suspend, ixgbe_resume);
11586 
11587 static struct pci_driver ixgbe_driver = {
11588 	.name      = ixgbe_driver_name,
11589 	.id_table  = ixgbe_pci_tbl,
11590 	.probe     = ixgbe_probe,
11591 	.remove    = ixgbe_remove,
11592 	.driver.pm = pm_sleep_ptr(&ixgbe_pm_ops),
11593 	.shutdown  = ixgbe_shutdown,
11594 	.sriov_configure = ixgbe_pci_sriov_configure,
11595 	.err_handler = &ixgbe_err_handler
11596 };
11597 
11598 /**
11599  * ixgbe_init_module - Driver Registration Routine
11600  *
11601  * ixgbe_init_module is the first routine called when the driver is
11602  * loaded. All it does is register with the PCI subsystem.
11603  **/
11604 static int __init ixgbe_init_module(void)
11605 {
11606 	int ret;
11607 	pr_info("%s\n", ixgbe_driver_string);
11608 	pr_info("%s\n", ixgbe_copyright);
11609 
11610 	ixgbe_wq = create_singlethread_workqueue(ixgbe_driver_name);
11611 	if (!ixgbe_wq) {
11612 		pr_err("%s: Failed to create workqueue\n", ixgbe_driver_name);
11613 		return -ENOMEM;
11614 	}
11615 
11616 	ixgbe_dbg_init();
11617 
11618 	ret = pci_register_driver(&ixgbe_driver);
11619 	if (ret) {
11620 		destroy_workqueue(ixgbe_wq);
11621 		ixgbe_dbg_exit();
11622 		return ret;
11623 	}
11624 
11625 #ifdef CONFIG_IXGBE_DCA
11626 	dca_register_notify(&dca_notifier);
11627 #endif
11628 
11629 	return 0;
11630 }
11631 
11632 module_init(ixgbe_init_module);
11633 
11634 /**
11635  * ixgbe_exit_module - Driver Exit Cleanup Routine
11636  *
11637  * ixgbe_exit_module is called just before the driver is removed
11638  * from memory.
11639  **/
11640 static void __exit ixgbe_exit_module(void)
11641 {
11642 #ifdef CONFIG_IXGBE_DCA
11643 	dca_unregister_notify(&dca_notifier);
11644 #endif
11645 	pci_unregister_driver(&ixgbe_driver);
11646 
11647 	ixgbe_dbg_exit();
11648 	if (ixgbe_wq) {
11649 		destroy_workqueue(ixgbe_wq);
11650 		ixgbe_wq = NULL;
11651 	}
11652 }
11653 
11654 #ifdef CONFIG_IXGBE_DCA
11655 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
11656 			    void *p)
11657 {
11658 	int ret_val;
11659 
11660 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
11661 					 __ixgbe_notify_dca);
11662 
11663 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
11664 }
11665 
11666 #endif /* CONFIG_IXGBE_DCA */
11667 
11668 module_exit(ixgbe_exit_module);
11669 
11670 /* ixgbe_main.c */
11671