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/xdp_sock_drv.h>
40 #include <net/xfrm.h>
41 
42 #include "ixgbe.h"
43 #include "ixgbe_common.h"
44 #include "ixgbe_dcb_82599.h"
45 #include "ixgbe_phy.h"
46 #include "ixgbe_sriov.h"
47 #include "ixgbe_model.h"
48 #include "ixgbe_txrx_common.h"
49 
50 char ixgbe_driver_name[] = "ixgbe";
51 static const char ixgbe_driver_string[] =
52 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
53 #ifdef IXGBE_FCOE
54 char ixgbe_default_device_descr[] =
55 			      "Intel(R) 10 Gigabit Network Connection";
56 #else
57 static char ixgbe_default_device_descr[] =
58 			      "Intel(R) 10 Gigabit Network Connection";
59 #endif
60 static const char ixgbe_copyright[] =
61 				"Copyright (c) 1999-2016 Intel Corporation.";
62 
63 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";
64 
65 static const struct ixgbe_info *ixgbe_info_tbl[] = {
66 	[board_82598]		= &ixgbe_82598_info,
67 	[board_82599]		= &ixgbe_82599_info,
68 	[board_X540]		= &ixgbe_X540_info,
69 	[board_X550]		= &ixgbe_X550_info,
70 	[board_X550EM_x]	= &ixgbe_X550EM_x_info,
71 	[board_x550em_x_fw]	= &ixgbe_x550em_x_fw_info,
72 	[board_x550em_a]	= &ixgbe_x550em_a_info,
73 	[board_x550em_a_fw]	= &ixgbe_x550em_a_fw_info,
74 };
75 
76 /* ixgbe_pci_tbl - PCI Device ID Table
77  *
78  * Wildcard entries (PCI_ANY_ID) should come last
79  * Last entry must be all 0s
80  *
81  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
82  *   Class, Class Mask, private data (not used) }
83  */
84 static const struct pci_device_id ixgbe_pci_tbl[] = {
85 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
86 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
87 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
88 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
89 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
90 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
91 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
92 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
93 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
94 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
95 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
96 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
97 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
98 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
99 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
100 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
101 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
102 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
103 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
104 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
105 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
115 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
116 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T1), board_X550},
117 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
118 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_XFI), board_X550EM_x},
119 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
120 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_10G_T), board_X550EM_x},
121 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_SFP), board_X550EM_x},
122 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_1G_T), board_x550em_x_fw},
123 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR), board_x550em_a },
124 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR_L), board_x550em_a },
125 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP_N), board_x550em_a },
126 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII), board_x550em_a },
127 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII_L), board_x550em_a },
128 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_10G_T), board_x550em_a},
129 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP), board_x550em_a },
130 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T), board_x550em_a_fw },
131 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T_L), board_x550em_a_fw },
132 	/* required last entry */
133 	{0, }
134 };
135 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
136 
137 #ifdef CONFIG_IXGBE_DCA
138 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
139 			    void *p);
140 static struct notifier_block dca_notifier = {
141 	.notifier_call = ixgbe_notify_dca,
142 	.next          = NULL,
143 	.priority      = 0
144 };
145 #endif
146 
147 #ifdef CONFIG_PCI_IOV
148 static unsigned int max_vfs;
149 module_param(max_vfs, uint, 0);
150 MODULE_PARM_DESC(max_vfs,
151 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
152 #endif /* CONFIG_PCI_IOV */
153 
154 static unsigned int allow_unsupported_sfp;
155 module_param(allow_unsupported_sfp, uint, 0);
156 MODULE_PARM_DESC(allow_unsupported_sfp,
157 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
158 
159 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
160 static int debug = -1;
161 module_param(debug, int, 0);
162 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
163 
164 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
165 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
166 MODULE_LICENSE("GPL v2");
167 
168 static struct workqueue_struct *ixgbe_wq;
169 
170 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
171 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *);
172 
173 static const struct net_device_ops ixgbe_netdev_ops;
174 
175 static bool netif_is_ixgbe(struct net_device *dev)
176 {
177 	return dev && (dev->netdev_ops == &ixgbe_netdev_ops);
178 }
179 
180 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
181 					  u32 reg, u16 *value)
182 {
183 	struct pci_dev *parent_dev;
184 	struct pci_bus *parent_bus;
185 
186 	parent_bus = adapter->pdev->bus->parent;
187 	if (!parent_bus)
188 		return -1;
189 
190 	parent_dev = parent_bus->self;
191 	if (!parent_dev)
192 		return -1;
193 
194 	if (!pci_is_pcie(parent_dev))
195 		return -1;
196 
197 	pcie_capability_read_word(parent_dev, reg, value);
198 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
199 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
200 		return -1;
201 	return 0;
202 }
203 
204 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
205 {
206 	struct ixgbe_hw *hw = &adapter->hw;
207 	u16 link_status = 0;
208 	int err;
209 
210 	hw->bus.type = ixgbe_bus_type_pci_express;
211 
212 	/* Get the negotiated link width and speed from PCI config space of the
213 	 * parent, as this device is behind a switch
214 	 */
215 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
216 
217 	/* assume caller will handle error case */
218 	if (err)
219 		return err;
220 
221 	hw->bus.width = ixgbe_convert_bus_width(link_status);
222 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
223 
224 	return 0;
225 }
226 
227 /**
228  * ixgbe_pcie_from_parent - Determine whether PCIe info should come from parent
229  * @hw: hw specific details
230  *
231  * This function is used by probe to determine whether a device's PCI-Express
232  * bandwidth details should be gathered from the parent bus instead of from the
233  * device. Used to ensure that various locations all have the correct device ID
234  * checks.
235  */
236 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
237 {
238 	switch (hw->device_id) {
239 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
240 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
241 		return true;
242 	default:
243 		return false;
244 	}
245 }
246 
247 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
248 				     int expected_gts)
249 {
250 	struct ixgbe_hw *hw = &adapter->hw;
251 	struct pci_dev *pdev;
252 
253 	/* Some devices are not connected over PCIe and thus do not negotiate
254 	 * speed. These devices do not have valid bus info, and thus any report
255 	 * we generate may not be correct.
256 	 */
257 	if (hw->bus.type == ixgbe_bus_type_internal)
258 		return;
259 
260 	/* determine whether to use the parent device */
261 	if (ixgbe_pcie_from_parent(&adapter->hw))
262 		pdev = adapter->pdev->bus->parent->self;
263 	else
264 		pdev = adapter->pdev;
265 
266 	pcie_print_link_status(pdev);
267 }
268 
269 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
270 {
271 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
272 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
273 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
274 		queue_work(ixgbe_wq, &adapter->service_task);
275 }
276 
277 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
278 {
279 	struct ixgbe_adapter *adapter = hw->back;
280 
281 	if (!hw->hw_addr)
282 		return;
283 	hw->hw_addr = NULL;
284 	e_dev_err("Adapter removed\n");
285 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
286 		ixgbe_service_event_schedule(adapter);
287 }
288 
289 static u32 ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
290 {
291 	u8 __iomem *reg_addr;
292 	u32 value;
293 	int i;
294 
295 	reg_addr = READ_ONCE(hw->hw_addr);
296 	if (ixgbe_removed(reg_addr))
297 		return IXGBE_FAILED_READ_REG;
298 
299 	/* Register read of 0xFFFFFFF can indicate the adapter has been removed,
300 	 * so perform several status register reads to determine if the adapter
301 	 * has been removed.
302 	 */
303 	for (i = 0; i < IXGBE_FAILED_READ_RETRIES; i++) {
304 		value = readl(reg_addr + IXGBE_STATUS);
305 		if (value != IXGBE_FAILED_READ_REG)
306 			break;
307 		mdelay(3);
308 	}
309 
310 	if (value == IXGBE_FAILED_READ_REG)
311 		ixgbe_remove_adapter(hw);
312 	else
313 		value = readl(reg_addr + reg);
314 	return value;
315 }
316 
317 /**
318  * ixgbe_read_reg - Read from device register
319  * @hw: hw specific details
320  * @reg: offset of register to read
321  *
322  * Returns : value read or IXGBE_FAILED_READ_REG if removed
323  *
324  * This function is used to read device registers. It checks for device
325  * removal by confirming any read that returns all ones by checking the
326  * status register value for all ones. This function avoids reading from
327  * the hardware if a removal was previously detected in which case it
328  * returns IXGBE_FAILED_READ_REG (all ones).
329  */
330 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
331 {
332 	u8 __iomem *reg_addr = READ_ONCE(hw->hw_addr);
333 	u32 value;
334 
335 	if (ixgbe_removed(reg_addr))
336 		return IXGBE_FAILED_READ_REG;
337 	if (unlikely(hw->phy.nw_mng_if_sel &
338 		     IXGBE_NW_MNG_IF_SEL_SGMII_ENABLE)) {
339 		struct ixgbe_adapter *adapter;
340 		int i;
341 
342 		for (i = 0; i < 200; ++i) {
343 			value = readl(reg_addr + IXGBE_MAC_SGMII_BUSY);
344 			if (likely(!value))
345 				goto writes_completed;
346 			if (value == IXGBE_FAILED_READ_REG) {
347 				ixgbe_remove_adapter(hw);
348 				return IXGBE_FAILED_READ_REG;
349 			}
350 			udelay(5);
351 		}
352 
353 		adapter = hw->back;
354 		e_warn(hw, "register writes incomplete %08x\n", value);
355 	}
356 
357 writes_completed:
358 	value = readl(reg_addr + reg);
359 	if (unlikely(value == IXGBE_FAILED_READ_REG))
360 		value = ixgbe_check_remove(hw, reg);
361 	return value;
362 }
363 
364 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
365 {
366 	u16 value;
367 
368 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
369 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
370 		ixgbe_remove_adapter(hw);
371 		return true;
372 	}
373 	return false;
374 }
375 
376 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
377 {
378 	struct ixgbe_adapter *adapter = hw->back;
379 	u16 value;
380 
381 	if (ixgbe_removed(hw->hw_addr))
382 		return IXGBE_FAILED_READ_CFG_WORD;
383 	pci_read_config_word(adapter->pdev, reg, &value);
384 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
385 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
386 		return IXGBE_FAILED_READ_CFG_WORD;
387 	return value;
388 }
389 
390 #ifdef CONFIG_PCI_IOV
391 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
392 {
393 	struct ixgbe_adapter *adapter = hw->back;
394 	u32 value;
395 
396 	if (ixgbe_removed(hw->hw_addr))
397 		return IXGBE_FAILED_READ_CFG_DWORD;
398 	pci_read_config_dword(adapter->pdev, reg, &value);
399 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
400 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
401 		return IXGBE_FAILED_READ_CFG_DWORD;
402 	return value;
403 }
404 #endif /* CONFIG_PCI_IOV */
405 
406 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
407 {
408 	struct ixgbe_adapter *adapter = hw->back;
409 
410 	if (ixgbe_removed(hw->hw_addr))
411 		return;
412 	pci_write_config_word(adapter->pdev, reg, value);
413 }
414 
415 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
416 {
417 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
418 
419 	/* flush memory to make sure state is correct before next watchdog */
420 	smp_mb__before_atomic();
421 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
422 }
423 
424 struct ixgbe_reg_info {
425 	u32 ofs;
426 	char *name;
427 };
428 
429 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
430 
431 	/* General Registers */
432 	{IXGBE_CTRL, "CTRL"},
433 	{IXGBE_STATUS, "STATUS"},
434 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
435 
436 	/* Interrupt Registers */
437 	{IXGBE_EICR, "EICR"},
438 
439 	/* RX Registers */
440 	{IXGBE_SRRCTL(0), "SRRCTL"},
441 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
442 	{IXGBE_RDLEN(0), "RDLEN"},
443 	{IXGBE_RDH(0), "RDH"},
444 	{IXGBE_RDT(0), "RDT"},
445 	{IXGBE_RXDCTL(0), "RXDCTL"},
446 	{IXGBE_RDBAL(0), "RDBAL"},
447 	{IXGBE_RDBAH(0), "RDBAH"},
448 
449 	/* TX Registers */
450 	{IXGBE_TDBAL(0), "TDBAL"},
451 	{IXGBE_TDBAH(0), "TDBAH"},
452 	{IXGBE_TDLEN(0), "TDLEN"},
453 	{IXGBE_TDH(0), "TDH"},
454 	{IXGBE_TDT(0), "TDT"},
455 	{IXGBE_TXDCTL(0), "TXDCTL"},
456 
457 	/* List Terminator */
458 	{ .name = NULL }
459 };
460 
461 
462 /*
463  * ixgbe_regdump - register printout routine
464  */
465 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
466 {
467 	int i;
468 	char rname[16];
469 	u32 regs[64];
470 
471 	switch (reginfo->ofs) {
472 	case IXGBE_SRRCTL(0):
473 		for (i = 0; i < 64; i++)
474 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
475 		break;
476 	case IXGBE_DCA_RXCTRL(0):
477 		for (i = 0; i < 64; i++)
478 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
479 		break;
480 	case IXGBE_RDLEN(0):
481 		for (i = 0; i < 64; i++)
482 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
483 		break;
484 	case IXGBE_RDH(0):
485 		for (i = 0; i < 64; i++)
486 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
487 		break;
488 	case IXGBE_RDT(0):
489 		for (i = 0; i < 64; i++)
490 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
491 		break;
492 	case IXGBE_RXDCTL(0):
493 		for (i = 0; i < 64; i++)
494 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
495 		break;
496 	case IXGBE_RDBAL(0):
497 		for (i = 0; i < 64; i++)
498 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
499 		break;
500 	case IXGBE_RDBAH(0):
501 		for (i = 0; i < 64; i++)
502 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
503 		break;
504 	case IXGBE_TDBAL(0):
505 		for (i = 0; i < 64; i++)
506 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
507 		break;
508 	case IXGBE_TDBAH(0):
509 		for (i = 0; i < 64; i++)
510 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
511 		break;
512 	case IXGBE_TDLEN(0):
513 		for (i = 0; i < 64; i++)
514 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
515 		break;
516 	case IXGBE_TDH(0):
517 		for (i = 0; i < 64; i++)
518 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
519 		break;
520 	case IXGBE_TDT(0):
521 		for (i = 0; i < 64; i++)
522 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
523 		break;
524 	case IXGBE_TXDCTL(0):
525 		for (i = 0; i < 64; i++)
526 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
527 		break;
528 	default:
529 		pr_info("%-15s %08x\n",
530 			reginfo->name, IXGBE_READ_REG(hw, reginfo->ofs));
531 		return;
532 	}
533 
534 	i = 0;
535 	while (i < 64) {
536 		int j;
537 		char buf[9 * 8 + 1];
538 		char *p = buf;
539 
540 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i, i + 7);
541 		for (j = 0; j < 8; j++)
542 			p += sprintf(p, " %08x", regs[i++]);
543 		pr_err("%-15s%s\n", rname, buf);
544 	}
545 
546 }
547 
548 static void ixgbe_print_buffer(struct ixgbe_ring *ring, int n)
549 {
550 	struct ixgbe_tx_buffer *tx_buffer;
551 
552 	tx_buffer = &ring->tx_buffer_info[ring->next_to_clean];
553 	pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
554 		n, ring->next_to_use, ring->next_to_clean,
555 		(u64)dma_unmap_addr(tx_buffer, dma),
556 		dma_unmap_len(tx_buffer, len),
557 		tx_buffer->next_to_watch,
558 		(u64)tx_buffer->time_stamp);
559 }
560 
561 /*
562  * ixgbe_dump - Print registers, tx-rings and rx-rings
563  */
564 static void ixgbe_dump(struct ixgbe_adapter *adapter)
565 {
566 	struct net_device *netdev = adapter->netdev;
567 	struct ixgbe_hw *hw = &adapter->hw;
568 	struct ixgbe_reg_info *reginfo;
569 	int n = 0;
570 	struct ixgbe_ring *ring;
571 	struct ixgbe_tx_buffer *tx_buffer;
572 	union ixgbe_adv_tx_desc *tx_desc;
573 	struct my_u0 { u64 a; u64 b; } *u0;
574 	struct ixgbe_ring *rx_ring;
575 	union ixgbe_adv_rx_desc *rx_desc;
576 	struct ixgbe_rx_buffer *rx_buffer_info;
577 	int i = 0;
578 
579 	if (!netif_msg_hw(adapter))
580 		return;
581 
582 	/* Print netdevice Info */
583 	if (netdev) {
584 		dev_info(&adapter->pdev->dev, "Net device Info\n");
585 		pr_info("Device Name     state            "
586 			"trans_start\n");
587 		pr_info("%-15s %016lX %016lX\n",
588 			netdev->name,
589 			netdev->state,
590 			dev_trans_start(netdev));
591 	}
592 
593 	/* Print Registers */
594 	dev_info(&adapter->pdev->dev, "Register Dump\n");
595 	pr_info(" Register Name   Value\n");
596 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
597 	     reginfo->name; reginfo++) {
598 		ixgbe_regdump(hw, reginfo);
599 	}
600 
601 	/* Print TX Ring Summary */
602 	if (!netdev || !netif_running(netdev))
603 		return;
604 
605 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
606 	pr_info(" %s     %s              %s        %s\n",
607 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
608 		"leng", "ntw", "timestamp");
609 	for (n = 0; n < adapter->num_tx_queues; n++) {
610 		ring = adapter->tx_ring[n];
611 		ixgbe_print_buffer(ring, n);
612 	}
613 
614 	for (n = 0; n < adapter->num_xdp_queues; n++) {
615 		ring = adapter->xdp_ring[n];
616 		ixgbe_print_buffer(ring, n);
617 	}
618 
619 	/* Print TX Rings */
620 	if (!netif_msg_tx_done(adapter))
621 		goto rx_ring_summary;
622 
623 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
624 
625 	/* Transmit Descriptor Formats
626 	 *
627 	 * 82598 Advanced Transmit Descriptor
628 	 *   +--------------------------------------------------------------+
629 	 * 0 |         Buffer Address [63:0]                                |
630 	 *   +--------------------------------------------------------------+
631 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
632 	 *   +--------------------------------------------------------------+
633 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
634 	 *
635 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
636 	 *   +--------------------------------------------------------------+
637 	 * 0 |                          RSV [63:0]                          |
638 	 *   +--------------------------------------------------------------+
639 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
640 	 *   +--------------------------------------------------------------+
641 	 *   63                       36 35   32 31                         0
642 	 *
643 	 * 82599+ Advanced Transmit Descriptor
644 	 *   +--------------------------------------------------------------+
645 	 * 0 |         Buffer Address [63:0]                                |
646 	 *   +--------------------------------------------------------------+
647 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
648 	 *   +--------------------------------------------------------------+
649 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
650 	 *
651 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
652 	 *   +--------------------------------------------------------------+
653 	 * 0 |                          RSV [63:0]                          |
654 	 *   +--------------------------------------------------------------+
655 	 * 8 |            RSV           |  STA  |           RSV             |
656 	 *   +--------------------------------------------------------------+
657 	 *   63                       36 35   32 31                         0
658 	 */
659 
660 	for (n = 0; n < adapter->num_tx_queues; n++) {
661 		ring = adapter->tx_ring[n];
662 		pr_info("------------------------------------\n");
663 		pr_info("TX QUEUE INDEX = %d\n", ring->queue_index);
664 		pr_info("------------------------------------\n");
665 		pr_info("%s%s    %s              %s        %s          %s\n",
666 			"T [desc]     [address 63:0  ] ",
667 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
668 			"leng", "ntw", "timestamp", "bi->skb");
669 
670 		for (i = 0; ring->desc && (i < ring->count); i++) {
671 			tx_desc = IXGBE_TX_DESC(ring, i);
672 			tx_buffer = &ring->tx_buffer_info[i];
673 			u0 = (struct my_u0 *)tx_desc;
674 			if (dma_unmap_len(tx_buffer, len) > 0) {
675 				const char *ring_desc;
676 
677 				if (i == ring->next_to_use &&
678 				    i == ring->next_to_clean)
679 					ring_desc = " NTC/U";
680 				else if (i == ring->next_to_use)
681 					ring_desc = " NTU";
682 				else if (i == ring->next_to_clean)
683 					ring_desc = " NTC";
684 				else
685 					ring_desc = "";
686 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p%s",
687 					i,
688 					le64_to_cpu((__force __le64)u0->a),
689 					le64_to_cpu((__force __le64)u0->b),
690 					(u64)dma_unmap_addr(tx_buffer, dma),
691 					dma_unmap_len(tx_buffer, len),
692 					tx_buffer->next_to_watch,
693 					(u64)tx_buffer->time_stamp,
694 					tx_buffer->skb,
695 					ring_desc);
696 
697 				if (netif_msg_pktdata(adapter) &&
698 				    tx_buffer->skb)
699 					print_hex_dump(KERN_INFO, "",
700 						DUMP_PREFIX_ADDRESS, 16, 1,
701 						tx_buffer->skb->data,
702 						dma_unmap_len(tx_buffer, len),
703 						true);
704 			}
705 		}
706 	}
707 
708 	/* Print RX Rings Summary */
709 rx_ring_summary:
710 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
711 	pr_info("Queue [NTU] [NTC]\n");
712 	for (n = 0; n < adapter->num_rx_queues; n++) {
713 		rx_ring = adapter->rx_ring[n];
714 		pr_info("%5d %5X %5X\n",
715 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
716 	}
717 
718 	/* Print RX Rings */
719 	if (!netif_msg_rx_status(adapter))
720 		return;
721 
722 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
723 
724 	/* Receive Descriptor Formats
725 	 *
726 	 * 82598 Advanced Receive Descriptor (Read) Format
727 	 *    63                                           1        0
728 	 *    +-----------------------------------------------------+
729 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
730 	 *    +----------------------------------------------+------+
731 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
732 	 *    +-----------------------------------------------------+
733 	 *
734 	 *
735 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
736 	 *
737 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
738 	 *   +------------------------------------------------------+
739 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
740 	 *   | Packet   | IP     |   |          |     | Type | Type |
741 	 *   | Checksum | Ident  |   |          |     |      |      |
742 	 *   +------------------------------------------------------+
743 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
744 	 *   +------------------------------------------------------+
745 	 *   63       48 47    32 31            20 19               0
746 	 *
747 	 * 82599+ Advanced Receive Descriptor (Read) Format
748 	 *    63                                           1        0
749 	 *    +-----------------------------------------------------+
750 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
751 	 *    +----------------------------------------------+------+
752 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
753 	 *    +-----------------------------------------------------+
754 	 *
755 	 *
756 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
757 	 *
758 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
759 	 *   +------------------------------------------------------+
760 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
761 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
762 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
763 	 *   +------------------------------------------------------+
764 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
765 	 *   +------------------------------------------------------+
766 	 *   63       48 47    32 31          20 19                 0
767 	 */
768 
769 	for (n = 0; n < adapter->num_rx_queues; n++) {
770 		rx_ring = adapter->rx_ring[n];
771 		pr_info("------------------------------------\n");
772 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
773 		pr_info("------------------------------------\n");
774 		pr_info("%s%s%s\n",
775 			"R  [desc]      [ PktBuf     A0] ",
776 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
777 			"<-- Adv Rx Read format");
778 		pr_info("%s%s%s\n",
779 			"RWB[desc]      [PcsmIpSHl PtRs] ",
780 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
781 			"<-- Adv Rx Write-Back format");
782 
783 		for (i = 0; i < rx_ring->count; i++) {
784 			const char *ring_desc;
785 
786 			if (i == rx_ring->next_to_use)
787 				ring_desc = " NTU";
788 			else if (i == rx_ring->next_to_clean)
789 				ring_desc = " NTC";
790 			else
791 				ring_desc = "";
792 
793 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
794 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
795 			u0 = (struct my_u0 *)rx_desc;
796 			if (rx_desc->wb.upper.length) {
797 				/* Descriptor Done */
798 				pr_info("RWB[0x%03X]     %016llX %016llX ---------------- %p%s\n",
799 					i,
800 					le64_to_cpu((__force __le64)u0->a),
801 					le64_to_cpu((__force __le64)u0->b),
802 					rx_buffer_info->skb,
803 					ring_desc);
804 			} else {
805 				pr_info("R  [0x%03X]     %016llX %016llX %016llX %p%s\n",
806 					i,
807 					le64_to_cpu((__force __le64)u0->a),
808 					le64_to_cpu((__force __le64)u0->b),
809 					(u64)rx_buffer_info->dma,
810 					rx_buffer_info->skb,
811 					ring_desc);
812 
813 				if (netif_msg_pktdata(adapter) &&
814 				    rx_buffer_info->dma) {
815 					print_hex_dump(KERN_INFO, "",
816 					   DUMP_PREFIX_ADDRESS, 16, 1,
817 					   page_address(rx_buffer_info->page) +
818 						    rx_buffer_info->page_offset,
819 					   ixgbe_rx_bufsz(rx_ring), true);
820 				}
821 			}
822 		}
823 	}
824 }
825 
826 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
827 {
828 	u32 ctrl_ext;
829 
830 	/* Let firmware take over control of h/w */
831 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
832 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
833 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
834 }
835 
836 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
837 {
838 	u32 ctrl_ext;
839 
840 	/* Let firmware know the driver has taken over */
841 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
842 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
843 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
844 }
845 
846 /**
847  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
848  * @adapter: pointer to adapter struct
849  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
850  * @queue: queue to map the corresponding interrupt to
851  * @msix_vector: the vector to map to the corresponding queue
852  *
853  */
854 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
855 			   u8 queue, u8 msix_vector)
856 {
857 	u32 ivar, index;
858 	struct ixgbe_hw *hw = &adapter->hw;
859 	switch (hw->mac.type) {
860 	case ixgbe_mac_82598EB:
861 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
862 		if (direction == -1)
863 			direction = 0;
864 		index = (((direction * 64) + queue) >> 2) & 0x1F;
865 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
866 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
867 		ivar |= (msix_vector << (8 * (queue & 0x3)));
868 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
869 		break;
870 	case ixgbe_mac_82599EB:
871 	case ixgbe_mac_X540:
872 	case ixgbe_mac_X550:
873 	case ixgbe_mac_X550EM_x:
874 	case ixgbe_mac_x550em_a:
875 		if (direction == -1) {
876 			/* other causes */
877 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
878 			index = ((queue & 1) * 8);
879 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
880 			ivar &= ~(0xFF << index);
881 			ivar |= (msix_vector << index);
882 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
883 			break;
884 		} else {
885 			/* tx or rx causes */
886 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
887 			index = ((16 * (queue & 1)) + (8 * direction));
888 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
889 			ivar &= ~(0xFF << index);
890 			ivar |= (msix_vector << index);
891 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
892 			break;
893 		}
894 	default:
895 		break;
896 	}
897 }
898 
899 void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
900 			    u64 qmask)
901 {
902 	u32 mask;
903 
904 	switch (adapter->hw.mac.type) {
905 	case ixgbe_mac_82598EB:
906 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
907 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
908 		break;
909 	case ixgbe_mac_82599EB:
910 	case ixgbe_mac_X540:
911 	case ixgbe_mac_X550:
912 	case ixgbe_mac_X550EM_x:
913 	case ixgbe_mac_x550em_a:
914 		mask = (qmask & 0xFFFFFFFF);
915 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
916 		mask = (qmask >> 32);
917 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
918 		break;
919 	default:
920 		break;
921 	}
922 }
923 
924 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
925 {
926 	struct ixgbe_hw *hw = &adapter->hw;
927 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
928 	int i;
929 	u32 data;
930 
931 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
932 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
933 		return;
934 
935 	switch (hw->mac.type) {
936 	case ixgbe_mac_82598EB:
937 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
938 		break;
939 	default:
940 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
941 	}
942 	hwstats->lxoffrxc += data;
943 
944 	/* refill credits (no tx hang) if we received xoff */
945 	if (!data)
946 		return;
947 
948 	for (i = 0; i < adapter->num_tx_queues; i++)
949 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
950 			  &adapter->tx_ring[i]->state);
951 
952 	for (i = 0; i < adapter->num_xdp_queues; i++)
953 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
954 			  &adapter->xdp_ring[i]->state);
955 }
956 
957 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
958 {
959 	struct ixgbe_hw *hw = &adapter->hw;
960 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
961 	u32 xoff[8] = {0};
962 	u8 tc;
963 	int i;
964 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
965 
966 	if (adapter->ixgbe_ieee_pfc)
967 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
968 
969 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
970 		ixgbe_update_xoff_rx_lfc(adapter);
971 		return;
972 	}
973 
974 	/* update stats for each tc, only valid with PFC enabled */
975 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
976 		u32 pxoffrxc;
977 
978 		switch (hw->mac.type) {
979 		case ixgbe_mac_82598EB:
980 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
981 			break;
982 		default:
983 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
984 		}
985 		hwstats->pxoffrxc[i] += pxoffrxc;
986 		/* Get the TC for given UP */
987 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
988 		xoff[tc] += pxoffrxc;
989 	}
990 
991 	/* disarm tx queues that have received xoff frames */
992 	for (i = 0; i < adapter->num_tx_queues; i++) {
993 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
994 
995 		tc = tx_ring->dcb_tc;
996 		if (xoff[tc])
997 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
998 	}
999 
1000 	for (i = 0; i < adapter->num_xdp_queues; i++) {
1001 		struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
1002 
1003 		tc = xdp_ring->dcb_tc;
1004 		if (xoff[tc])
1005 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &xdp_ring->state);
1006 	}
1007 }
1008 
1009 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1010 {
1011 	return ring->stats.packets;
1012 }
1013 
1014 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1015 {
1016 	unsigned int head, tail;
1017 
1018 	head = ring->next_to_clean;
1019 	tail = ring->next_to_use;
1020 
1021 	return ((head <= tail) ? tail : tail + ring->count) - head;
1022 }
1023 
1024 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1025 {
1026 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1027 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1028 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1029 
1030 	clear_check_for_tx_hang(tx_ring);
1031 
1032 	/*
1033 	 * Check for a hung queue, but be thorough. This verifies
1034 	 * that a transmit has been completed since the previous
1035 	 * check AND there is at least one packet pending. The
1036 	 * ARMED bit is set to indicate a potential hang. The
1037 	 * bit is cleared if a pause frame is received to remove
1038 	 * false hang detection due to PFC or 802.3x frames. By
1039 	 * requiring this to fail twice we avoid races with
1040 	 * pfc clearing the ARMED bit and conditions where we
1041 	 * run the check_tx_hang logic with a transmit completion
1042 	 * pending but without time to complete it yet.
1043 	 */
1044 	if (tx_done_old == tx_done && tx_pending)
1045 		/* make sure it is true for two checks in a row */
1046 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1047 					&tx_ring->state);
1048 	/* update completed stats and continue */
1049 	tx_ring->tx_stats.tx_done_old = tx_done;
1050 	/* reset the countdown */
1051 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1052 
1053 	return false;
1054 }
1055 
1056 /**
1057  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1058  * @adapter: driver private struct
1059  **/
1060 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1061 {
1062 
1063 	/* Do the reset outside of interrupt context */
1064 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1065 		set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
1066 		e_warn(drv, "initiating reset due to tx timeout\n");
1067 		ixgbe_service_event_schedule(adapter);
1068 	}
1069 }
1070 
1071 /**
1072  * ixgbe_tx_maxrate - callback to set the maximum per-queue bitrate
1073  * @netdev: network interface device structure
1074  * @queue_index: Tx queue to set
1075  * @maxrate: desired maximum transmit bitrate
1076  **/
1077 static int ixgbe_tx_maxrate(struct net_device *netdev,
1078 			    int queue_index, u32 maxrate)
1079 {
1080 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1081 	struct ixgbe_hw *hw = &adapter->hw;
1082 	u32 bcnrc_val = ixgbe_link_mbps(adapter);
1083 
1084 	if (!maxrate)
1085 		return 0;
1086 
1087 	/* Calculate the rate factor values to set */
1088 	bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1089 	bcnrc_val /= maxrate;
1090 
1091 	/* clear everything but the rate factor */
1092 	bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1093 	IXGBE_RTTBCNRC_RF_DEC_MASK;
1094 
1095 	/* enable the rate scheduler */
1096 	bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1097 
1098 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_index);
1099 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1100 
1101 	return 0;
1102 }
1103 
1104 /**
1105  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1106  * @q_vector: structure containing interrupt and ring information
1107  * @tx_ring: tx ring to clean
1108  * @napi_budget: Used to determine if we are in netpoll
1109  **/
1110 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1111 			       struct ixgbe_ring *tx_ring, int napi_budget)
1112 {
1113 	struct ixgbe_adapter *adapter = q_vector->adapter;
1114 	struct ixgbe_tx_buffer *tx_buffer;
1115 	union ixgbe_adv_tx_desc *tx_desc;
1116 	unsigned int total_bytes = 0, total_packets = 0, total_ipsec = 0;
1117 	unsigned int budget = q_vector->tx.work_limit;
1118 	unsigned int i = tx_ring->next_to_clean;
1119 
1120 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1121 		return true;
1122 
1123 	tx_buffer = &tx_ring->tx_buffer_info[i];
1124 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1125 	i -= tx_ring->count;
1126 
1127 	do {
1128 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1129 
1130 		/* if next_to_watch is not set then there is no work pending */
1131 		if (!eop_desc)
1132 			break;
1133 
1134 		/* prevent any other reads prior to eop_desc */
1135 		smp_rmb();
1136 
1137 		/* if DD is not set pending work has not been completed */
1138 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1139 			break;
1140 
1141 		/* clear next_to_watch to prevent false hangs */
1142 		tx_buffer->next_to_watch = NULL;
1143 
1144 		/* update the statistics for this packet */
1145 		total_bytes += tx_buffer->bytecount;
1146 		total_packets += tx_buffer->gso_segs;
1147 		if (tx_buffer->tx_flags & IXGBE_TX_FLAGS_IPSEC)
1148 			total_ipsec++;
1149 
1150 		/* free the skb */
1151 		if (ring_is_xdp(tx_ring))
1152 			xdp_return_frame(tx_buffer->xdpf);
1153 		else
1154 			napi_consume_skb(tx_buffer->skb, napi_budget);
1155 
1156 		/* unmap skb header data */
1157 		dma_unmap_single(tx_ring->dev,
1158 				 dma_unmap_addr(tx_buffer, dma),
1159 				 dma_unmap_len(tx_buffer, len),
1160 				 DMA_TO_DEVICE);
1161 
1162 		/* clear tx_buffer data */
1163 		dma_unmap_len_set(tx_buffer, len, 0);
1164 
1165 		/* unmap remaining buffers */
1166 		while (tx_desc != eop_desc) {
1167 			tx_buffer++;
1168 			tx_desc++;
1169 			i++;
1170 			if (unlikely(!i)) {
1171 				i -= tx_ring->count;
1172 				tx_buffer = tx_ring->tx_buffer_info;
1173 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1174 			}
1175 
1176 			/* unmap any remaining paged data */
1177 			if (dma_unmap_len(tx_buffer, len)) {
1178 				dma_unmap_page(tx_ring->dev,
1179 					       dma_unmap_addr(tx_buffer, dma),
1180 					       dma_unmap_len(tx_buffer, len),
1181 					       DMA_TO_DEVICE);
1182 				dma_unmap_len_set(tx_buffer, len, 0);
1183 			}
1184 		}
1185 
1186 		/* move us one more past the eop_desc for start of next pkt */
1187 		tx_buffer++;
1188 		tx_desc++;
1189 		i++;
1190 		if (unlikely(!i)) {
1191 			i -= tx_ring->count;
1192 			tx_buffer = tx_ring->tx_buffer_info;
1193 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1194 		}
1195 
1196 		/* issue prefetch for next Tx descriptor */
1197 		prefetch(tx_desc);
1198 
1199 		/* update budget accounting */
1200 		budget--;
1201 	} while (likely(budget));
1202 
1203 	i += tx_ring->count;
1204 	tx_ring->next_to_clean = i;
1205 	u64_stats_update_begin(&tx_ring->syncp);
1206 	tx_ring->stats.bytes += total_bytes;
1207 	tx_ring->stats.packets += total_packets;
1208 	u64_stats_update_end(&tx_ring->syncp);
1209 	q_vector->tx.total_bytes += total_bytes;
1210 	q_vector->tx.total_packets += total_packets;
1211 	adapter->tx_ipsec += total_ipsec;
1212 
1213 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1214 		/* schedule immediate reset if we believe we hung */
1215 		struct ixgbe_hw *hw = &adapter->hw;
1216 		e_err(drv, "Detected Tx Unit Hang %s\n"
1217 			"  Tx Queue             <%d>\n"
1218 			"  TDH, TDT             <%x>, <%x>\n"
1219 			"  next_to_use          <%x>\n"
1220 			"  next_to_clean        <%x>\n"
1221 			"tx_buffer_info[next_to_clean]\n"
1222 			"  time_stamp           <%lx>\n"
1223 			"  jiffies              <%lx>\n",
1224 			ring_is_xdp(tx_ring) ? "(XDP)" : "",
1225 			tx_ring->queue_index,
1226 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1227 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1228 			tx_ring->next_to_use, i,
1229 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1230 
1231 		if (!ring_is_xdp(tx_ring))
1232 			netif_stop_subqueue(tx_ring->netdev,
1233 					    tx_ring->queue_index);
1234 
1235 		e_info(probe,
1236 		       "tx hang %d detected on queue %d, resetting adapter\n",
1237 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1238 
1239 		/* schedule immediate reset if we believe we hung */
1240 		ixgbe_tx_timeout_reset(adapter);
1241 
1242 		/* the adapter is about to reset, no point in enabling stuff */
1243 		return true;
1244 	}
1245 
1246 	if (ring_is_xdp(tx_ring))
1247 		return !!budget;
1248 
1249 	netdev_tx_completed_queue(txring_txq(tx_ring),
1250 				  total_packets, total_bytes);
1251 
1252 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1253 	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1254 		     (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1255 		/* Make sure that anybody stopping the queue after this
1256 		 * sees the new next_to_clean.
1257 		 */
1258 		smp_mb();
1259 		if (__netif_subqueue_stopped(tx_ring->netdev,
1260 					     tx_ring->queue_index)
1261 		    && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1262 			netif_wake_subqueue(tx_ring->netdev,
1263 					    tx_ring->queue_index);
1264 			++tx_ring->tx_stats.restart_queue;
1265 		}
1266 	}
1267 
1268 	return !!budget;
1269 }
1270 
1271 #ifdef CONFIG_IXGBE_DCA
1272 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1273 				struct ixgbe_ring *tx_ring,
1274 				int cpu)
1275 {
1276 	struct ixgbe_hw *hw = &adapter->hw;
1277 	u32 txctrl = 0;
1278 	u16 reg_offset;
1279 
1280 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1281 		txctrl = dca3_get_tag(tx_ring->dev, cpu);
1282 
1283 	switch (hw->mac.type) {
1284 	case ixgbe_mac_82598EB:
1285 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1286 		break;
1287 	case ixgbe_mac_82599EB:
1288 	case ixgbe_mac_X540:
1289 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1290 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1291 		break;
1292 	default:
1293 		/* for unknown hardware do not write register */
1294 		return;
1295 	}
1296 
1297 	/*
1298 	 * We can enable relaxed ordering for reads, but not writes when
1299 	 * DCA is enabled.  This is due to a known issue in some chipsets
1300 	 * which will cause the DCA tag to be cleared.
1301 	 */
1302 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1303 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1304 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1305 
1306 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1307 }
1308 
1309 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1310 				struct ixgbe_ring *rx_ring,
1311 				int cpu)
1312 {
1313 	struct ixgbe_hw *hw = &adapter->hw;
1314 	u32 rxctrl = 0;
1315 	u8 reg_idx = rx_ring->reg_idx;
1316 
1317 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1318 		rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1319 
1320 	switch (hw->mac.type) {
1321 	case ixgbe_mac_82599EB:
1322 	case ixgbe_mac_X540:
1323 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1324 		break;
1325 	default:
1326 		break;
1327 	}
1328 
1329 	/*
1330 	 * We can enable relaxed ordering for reads, but not writes when
1331 	 * DCA is enabled.  This is due to a known issue in some chipsets
1332 	 * which will cause the DCA tag to be cleared.
1333 	 */
1334 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1335 		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1336 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1337 
1338 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1339 }
1340 
1341 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1342 {
1343 	struct ixgbe_adapter *adapter = q_vector->adapter;
1344 	struct ixgbe_ring *ring;
1345 	int cpu = get_cpu();
1346 
1347 	if (q_vector->cpu == cpu)
1348 		goto out_no_update;
1349 
1350 	ixgbe_for_each_ring(ring, q_vector->tx)
1351 		ixgbe_update_tx_dca(adapter, ring, cpu);
1352 
1353 	ixgbe_for_each_ring(ring, q_vector->rx)
1354 		ixgbe_update_rx_dca(adapter, ring, cpu);
1355 
1356 	q_vector->cpu = cpu;
1357 out_no_update:
1358 	put_cpu();
1359 }
1360 
1361 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1362 {
1363 	int i;
1364 
1365 	/* always use CB2 mode, difference is masked in the CB driver */
1366 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1367 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1368 				IXGBE_DCA_CTRL_DCA_MODE_CB2);
1369 	else
1370 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1371 				IXGBE_DCA_CTRL_DCA_DISABLE);
1372 
1373 	for (i = 0; i < adapter->num_q_vectors; i++) {
1374 		adapter->q_vector[i]->cpu = -1;
1375 		ixgbe_update_dca(adapter->q_vector[i]);
1376 	}
1377 }
1378 
1379 static int __ixgbe_notify_dca(struct device *dev, void *data)
1380 {
1381 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1382 	unsigned long event = *(unsigned long *)data;
1383 
1384 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1385 		return 0;
1386 
1387 	switch (event) {
1388 	case DCA_PROVIDER_ADD:
1389 		/* if we're already enabled, don't do it again */
1390 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1391 			break;
1392 		if (dca_add_requester(dev) == 0) {
1393 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1394 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1395 					IXGBE_DCA_CTRL_DCA_MODE_CB2);
1396 			break;
1397 		}
1398 		fallthrough; /* DCA is disabled. */
1399 	case DCA_PROVIDER_REMOVE:
1400 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1401 			dca_remove_requester(dev);
1402 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1403 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1404 					IXGBE_DCA_CTRL_DCA_DISABLE);
1405 		}
1406 		break;
1407 	}
1408 
1409 	return 0;
1410 }
1411 
1412 #endif /* CONFIG_IXGBE_DCA */
1413 
1414 #define IXGBE_RSS_L4_TYPES_MASK \
1415 	((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1416 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1417 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1418 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1419 
1420 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1421 				 union ixgbe_adv_rx_desc *rx_desc,
1422 				 struct sk_buff *skb)
1423 {
1424 	u16 rss_type;
1425 
1426 	if (!(ring->netdev->features & NETIF_F_RXHASH))
1427 		return;
1428 
1429 	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1430 		   IXGBE_RXDADV_RSSTYPE_MASK;
1431 
1432 	if (!rss_type)
1433 		return;
1434 
1435 	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1436 		     (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1437 		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1438 }
1439 
1440 #ifdef IXGBE_FCOE
1441 /**
1442  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1443  * @ring: structure containing ring specific data
1444  * @rx_desc: advanced rx descriptor
1445  *
1446  * Returns : true if it is FCoE pkt
1447  */
1448 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1449 				    union ixgbe_adv_rx_desc *rx_desc)
1450 {
1451 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1452 
1453 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1454 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1455 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1456 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1457 }
1458 
1459 #endif /* IXGBE_FCOE */
1460 /**
1461  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1462  * @ring: structure containing ring specific data
1463  * @rx_desc: current Rx descriptor being processed
1464  * @skb: skb currently being received and modified
1465  **/
1466 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1467 				     union ixgbe_adv_rx_desc *rx_desc,
1468 				     struct sk_buff *skb)
1469 {
1470 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1471 	bool encap_pkt = false;
1472 
1473 	skb_checksum_none_assert(skb);
1474 
1475 	/* Rx csum disabled */
1476 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1477 		return;
1478 
1479 	/* check for VXLAN and Geneve packets */
1480 	if (pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) {
1481 		encap_pkt = true;
1482 		skb->encapsulation = 1;
1483 	}
1484 
1485 	/* if IP and error */
1486 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1487 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1488 		ring->rx_stats.csum_err++;
1489 		return;
1490 	}
1491 
1492 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1493 		return;
1494 
1495 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1496 		/*
1497 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1498 		 * checksum errors.
1499 		 */
1500 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1501 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1502 			return;
1503 
1504 		ring->rx_stats.csum_err++;
1505 		return;
1506 	}
1507 
1508 	/* It must be a TCP or UDP packet with a valid checksum */
1509 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1510 	if (encap_pkt) {
1511 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1512 			return;
1513 
1514 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1515 			skb->ip_summed = CHECKSUM_NONE;
1516 			return;
1517 		}
1518 		/* If we checked the outer header let the stack know */
1519 		skb->csum_level = 1;
1520 	}
1521 }
1522 
1523 static unsigned int ixgbe_rx_offset(struct ixgbe_ring *rx_ring)
1524 {
1525 	return ring_uses_build_skb(rx_ring) ? IXGBE_SKB_PAD : 0;
1526 }
1527 
1528 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1529 				    struct ixgbe_rx_buffer *bi)
1530 {
1531 	struct page *page = bi->page;
1532 	dma_addr_t dma;
1533 
1534 	/* since we are recycling buffers we should seldom need to alloc */
1535 	if (likely(page))
1536 		return true;
1537 
1538 	/* alloc new page for storage */
1539 	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1540 	if (unlikely(!page)) {
1541 		rx_ring->rx_stats.alloc_rx_page_failed++;
1542 		return false;
1543 	}
1544 
1545 	/* map page for use */
1546 	dma = dma_map_page_attrs(rx_ring->dev, page, 0,
1547 				 ixgbe_rx_pg_size(rx_ring),
1548 				 DMA_FROM_DEVICE,
1549 				 IXGBE_RX_DMA_ATTR);
1550 
1551 	/*
1552 	 * if mapping failed free memory back to system since
1553 	 * there isn't much point in holding memory we can't use
1554 	 */
1555 	if (dma_mapping_error(rx_ring->dev, dma)) {
1556 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1557 
1558 		rx_ring->rx_stats.alloc_rx_page_failed++;
1559 		return false;
1560 	}
1561 
1562 	bi->dma = dma;
1563 	bi->page = page;
1564 	bi->page_offset = rx_ring->rx_offset;
1565 	page_ref_add(page, USHRT_MAX - 1);
1566 	bi->pagecnt_bias = USHRT_MAX;
1567 	rx_ring->rx_stats.alloc_rx_page++;
1568 
1569 	return true;
1570 }
1571 
1572 /**
1573  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1574  * @rx_ring: ring to place buffers on
1575  * @cleaned_count: number of buffers to replace
1576  **/
1577 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1578 {
1579 	union ixgbe_adv_rx_desc *rx_desc;
1580 	struct ixgbe_rx_buffer *bi;
1581 	u16 i = rx_ring->next_to_use;
1582 	u16 bufsz;
1583 
1584 	/* nothing to do */
1585 	if (!cleaned_count)
1586 		return;
1587 
1588 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1589 	bi = &rx_ring->rx_buffer_info[i];
1590 	i -= rx_ring->count;
1591 
1592 	bufsz = ixgbe_rx_bufsz(rx_ring);
1593 
1594 	do {
1595 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1596 			break;
1597 
1598 		/* sync the buffer for use by the device */
1599 		dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
1600 						 bi->page_offset, bufsz,
1601 						 DMA_FROM_DEVICE);
1602 
1603 		/*
1604 		 * Refresh the desc even if buffer_addrs didn't change
1605 		 * because each write-back erases this info.
1606 		 */
1607 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1608 
1609 		rx_desc++;
1610 		bi++;
1611 		i++;
1612 		if (unlikely(!i)) {
1613 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1614 			bi = rx_ring->rx_buffer_info;
1615 			i -= rx_ring->count;
1616 		}
1617 
1618 		/* clear the length for the next_to_use descriptor */
1619 		rx_desc->wb.upper.length = 0;
1620 
1621 		cleaned_count--;
1622 	} while (cleaned_count);
1623 
1624 	i += rx_ring->count;
1625 
1626 	if (rx_ring->next_to_use != i) {
1627 		rx_ring->next_to_use = i;
1628 
1629 		/* update next to alloc since we have filled the ring */
1630 		rx_ring->next_to_alloc = i;
1631 
1632 		/* Force memory writes to complete before letting h/w
1633 		 * know there are new descriptors to fetch.  (Only
1634 		 * applicable for weak-ordered memory model archs,
1635 		 * such as IA-64).
1636 		 */
1637 		wmb();
1638 		writel(i, rx_ring->tail);
1639 	}
1640 }
1641 
1642 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1643 				   struct sk_buff *skb)
1644 {
1645 	u16 hdr_len = skb_headlen(skb);
1646 
1647 	/* set gso_size to avoid messing up TCP MSS */
1648 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1649 						 IXGBE_CB(skb)->append_cnt);
1650 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1651 }
1652 
1653 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1654 				   struct sk_buff *skb)
1655 {
1656 	/* if append_cnt is 0 then frame is not RSC */
1657 	if (!IXGBE_CB(skb)->append_cnt)
1658 		return;
1659 
1660 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1661 	rx_ring->rx_stats.rsc_flush++;
1662 
1663 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1664 
1665 	/* gso_size is computed using append_cnt so always clear it last */
1666 	IXGBE_CB(skb)->append_cnt = 0;
1667 }
1668 
1669 /**
1670  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1671  * @rx_ring: rx descriptor ring packet is being transacted on
1672  * @rx_desc: pointer to the EOP Rx descriptor
1673  * @skb: pointer to current skb being populated
1674  *
1675  * This function checks the ring, descriptor, and packet information in
1676  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1677  * other fields within the skb.
1678  **/
1679 void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1680 			      union ixgbe_adv_rx_desc *rx_desc,
1681 			      struct sk_buff *skb)
1682 {
1683 	struct net_device *dev = rx_ring->netdev;
1684 	u32 flags = rx_ring->q_vector->adapter->flags;
1685 
1686 	ixgbe_update_rsc_stats(rx_ring, skb);
1687 
1688 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1689 
1690 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1691 
1692 	if (unlikely(flags & IXGBE_FLAG_RX_HWTSTAMP_ENABLED))
1693 		ixgbe_ptp_rx_hwtstamp(rx_ring, rx_desc, skb);
1694 
1695 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1696 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1697 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1698 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1699 	}
1700 
1701 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_SECP))
1702 		ixgbe_ipsec_rx(rx_ring, rx_desc, skb);
1703 
1704 	/* record Rx queue, or update MACVLAN statistics */
1705 	if (netif_is_ixgbe(dev))
1706 		skb_record_rx_queue(skb, rx_ring->queue_index);
1707 	else
1708 		macvlan_count_rx(netdev_priv(dev), skb->len + ETH_HLEN, true,
1709 				 false);
1710 
1711 	skb->protocol = eth_type_trans(skb, dev);
1712 }
1713 
1714 void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1715 		  struct sk_buff *skb)
1716 {
1717 	napi_gro_receive(&q_vector->napi, skb);
1718 }
1719 
1720 /**
1721  * ixgbe_is_non_eop - process handling of non-EOP buffers
1722  * @rx_ring: Rx ring being processed
1723  * @rx_desc: Rx descriptor for current buffer
1724  * @skb: Current socket buffer containing buffer in progress
1725  *
1726  * This function updates next to clean.  If the buffer is an EOP buffer
1727  * this function exits returning false, otherwise it will place the
1728  * sk_buff in the next buffer to be chained and return true indicating
1729  * that this is in fact a non-EOP buffer.
1730  **/
1731 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1732 			     union ixgbe_adv_rx_desc *rx_desc,
1733 			     struct sk_buff *skb)
1734 {
1735 	u32 ntc = rx_ring->next_to_clean + 1;
1736 
1737 	/* fetch, update, and store next to clean */
1738 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1739 	rx_ring->next_to_clean = ntc;
1740 
1741 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1742 
1743 	/* update RSC append count if present */
1744 	if (ring_is_rsc_enabled(rx_ring)) {
1745 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1746 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1747 
1748 		if (unlikely(rsc_enabled)) {
1749 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1750 
1751 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1752 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1753 
1754 			/* update ntc based on RSC value */
1755 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1756 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1757 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1758 		}
1759 	}
1760 
1761 	/* if we are the last buffer then there is nothing else to do */
1762 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1763 		return false;
1764 
1765 	/* place skb in next buffer to be received */
1766 	rx_ring->rx_buffer_info[ntc].skb = skb;
1767 	rx_ring->rx_stats.non_eop_descs++;
1768 
1769 	return true;
1770 }
1771 
1772 /**
1773  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1774  * @rx_ring: rx descriptor ring packet is being transacted on
1775  * @skb: pointer to current skb being adjusted
1776  *
1777  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1778  * main difference between this version and the original function is that
1779  * this function can make several assumptions about the state of things
1780  * that allow for significant optimizations versus the standard function.
1781  * As a result we can do things like drop a frag and maintain an accurate
1782  * truesize for the skb.
1783  */
1784 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1785 			    struct sk_buff *skb)
1786 {
1787 	skb_frag_t *frag = &skb_shinfo(skb)->frags[0];
1788 	unsigned char *va;
1789 	unsigned int pull_len;
1790 
1791 	/*
1792 	 * it is valid to use page_address instead of kmap since we are
1793 	 * working with pages allocated out of the lomem pool per
1794 	 * alloc_page(GFP_ATOMIC)
1795 	 */
1796 	va = skb_frag_address(frag);
1797 
1798 	/*
1799 	 * we need the header to contain the greater of either ETH_HLEN or
1800 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
1801 	 */
1802 	pull_len = eth_get_headlen(skb->dev, va, IXGBE_RX_HDR_SIZE);
1803 
1804 	/* align pull length to size of long to optimize memcpy performance */
1805 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1806 
1807 	/* update all of the pointers */
1808 	skb_frag_size_sub(frag, pull_len);
1809 	skb_frag_off_add(frag, pull_len);
1810 	skb->data_len -= pull_len;
1811 	skb->tail += pull_len;
1812 }
1813 
1814 /**
1815  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1816  * @rx_ring: rx descriptor ring packet is being transacted on
1817  * @skb: pointer to current skb being updated
1818  *
1819  * This function provides a basic DMA sync up for the first fragment of an
1820  * skb.  The reason for doing this is that the first fragment cannot be
1821  * unmapped until we have reached the end of packet descriptor for a buffer
1822  * chain.
1823  */
1824 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1825 				struct sk_buff *skb)
1826 {
1827 	if (ring_uses_build_skb(rx_ring)) {
1828 		unsigned long offset = (unsigned long)(skb->data) & ~PAGE_MASK;
1829 
1830 		dma_sync_single_range_for_cpu(rx_ring->dev,
1831 					      IXGBE_CB(skb)->dma,
1832 					      offset,
1833 					      skb_headlen(skb),
1834 					      DMA_FROM_DEVICE);
1835 	} else {
1836 		skb_frag_t *frag = &skb_shinfo(skb)->frags[0];
1837 
1838 		dma_sync_single_range_for_cpu(rx_ring->dev,
1839 					      IXGBE_CB(skb)->dma,
1840 					      skb_frag_off(frag),
1841 					      skb_frag_size(frag),
1842 					      DMA_FROM_DEVICE);
1843 	}
1844 
1845 	/* If the page was released, just unmap it. */
1846 	if (unlikely(IXGBE_CB(skb)->page_released)) {
1847 		dma_unmap_page_attrs(rx_ring->dev, IXGBE_CB(skb)->dma,
1848 				     ixgbe_rx_pg_size(rx_ring),
1849 				     DMA_FROM_DEVICE,
1850 				     IXGBE_RX_DMA_ATTR);
1851 	}
1852 }
1853 
1854 /**
1855  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1856  * @rx_ring: rx descriptor ring packet is being transacted on
1857  * @rx_desc: pointer to the EOP Rx descriptor
1858  * @skb: pointer to current skb being fixed
1859  *
1860  * Check if the skb is valid in the XDP case it will be an error pointer.
1861  * Return true in this case to abort processing and advance to next
1862  * descriptor.
1863  *
1864  * Check for corrupted packet headers caused by senders on the local L2
1865  * embedded NIC switch not setting up their Tx Descriptors right.  These
1866  * should be very rare.
1867  *
1868  * Also address the case where we are pulling data in on pages only
1869  * and as such no data is present in the skb header.
1870  *
1871  * In addition if skb is not at least 60 bytes we need to pad it so that
1872  * it is large enough to qualify as a valid Ethernet frame.
1873  *
1874  * Returns true if an error was encountered and skb was freed.
1875  **/
1876 bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1877 			   union ixgbe_adv_rx_desc *rx_desc,
1878 			   struct sk_buff *skb)
1879 {
1880 	struct net_device *netdev = rx_ring->netdev;
1881 
1882 	/* XDP packets use error pointer so abort at this point */
1883 	if (IS_ERR(skb))
1884 		return true;
1885 
1886 	/* Verify netdev is present, and that packet does not have any
1887 	 * errors that would be unacceptable to the netdev.
1888 	 */
1889 	if (!netdev ||
1890 	    (unlikely(ixgbe_test_staterr(rx_desc,
1891 					 IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1892 	     !(netdev->features & NETIF_F_RXALL)))) {
1893 		dev_kfree_skb_any(skb);
1894 		return true;
1895 	}
1896 
1897 	/* place header in linear portion of buffer */
1898 	if (!skb_headlen(skb))
1899 		ixgbe_pull_tail(rx_ring, skb);
1900 
1901 #ifdef IXGBE_FCOE
1902 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
1903 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1904 		return false;
1905 
1906 #endif
1907 	/* if eth_skb_pad returns an error the skb was freed */
1908 	if (eth_skb_pad(skb))
1909 		return true;
1910 
1911 	return false;
1912 }
1913 
1914 /**
1915  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1916  * @rx_ring: rx descriptor ring to store buffers on
1917  * @old_buff: donor buffer to have page reused
1918  *
1919  * Synchronizes page for reuse by the adapter
1920  **/
1921 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1922 				struct ixgbe_rx_buffer *old_buff)
1923 {
1924 	struct ixgbe_rx_buffer *new_buff;
1925 	u16 nta = rx_ring->next_to_alloc;
1926 
1927 	new_buff = &rx_ring->rx_buffer_info[nta];
1928 
1929 	/* update, and store next to alloc */
1930 	nta++;
1931 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1932 
1933 	/* Transfer page from old buffer to new buffer.
1934 	 * Move each member individually to avoid possible store
1935 	 * forwarding stalls and unnecessary copy of skb.
1936 	 */
1937 	new_buff->dma		= old_buff->dma;
1938 	new_buff->page		= old_buff->page;
1939 	new_buff->page_offset	= old_buff->page_offset;
1940 	new_buff->pagecnt_bias	= old_buff->pagecnt_bias;
1941 }
1942 
1943 static bool ixgbe_can_reuse_rx_page(struct ixgbe_rx_buffer *rx_buffer,
1944 				    int rx_buffer_pgcnt)
1945 {
1946 	unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
1947 	struct page *page = rx_buffer->page;
1948 
1949 	/* avoid re-using remote and pfmemalloc pages */
1950 	if (!dev_page_is_reusable(page))
1951 		return false;
1952 
1953 #if (PAGE_SIZE < 8192)
1954 	/* if we are only owner of page we can reuse it */
1955 	if (unlikely((rx_buffer_pgcnt - pagecnt_bias) > 1))
1956 		return false;
1957 #else
1958 	/* The last offset is a bit aggressive in that we assume the
1959 	 * worst case of FCoE being enabled and using a 3K buffer.
1960 	 * However this should have minimal impact as the 1K extra is
1961 	 * still less than one buffer in size.
1962 	 */
1963 #define IXGBE_LAST_OFFSET \
1964 	(SKB_WITH_OVERHEAD(PAGE_SIZE) - IXGBE_RXBUFFER_3K)
1965 	if (rx_buffer->page_offset > IXGBE_LAST_OFFSET)
1966 		return false;
1967 #endif
1968 
1969 	/* If we have drained the page fragment pool we need to update
1970 	 * the pagecnt_bias and page count so that we fully restock the
1971 	 * number of references the driver holds.
1972 	 */
1973 	if (unlikely(pagecnt_bias == 1)) {
1974 		page_ref_add(page, USHRT_MAX - 1);
1975 		rx_buffer->pagecnt_bias = USHRT_MAX;
1976 	}
1977 
1978 	return true;
1979 }
1980 
1981 /**
1982  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1983  * @rx_ring: rx descriptor ring to transact packets on
1984  * @rx_buffer: buffer containing page to add
1985  * @skb: sk_buff to place the data into
1986  * @size: size of data in rx_buffer
1987  *
1988  * This function will add the data contained in rx_buffer->page to the skb.
1989  * This is done either through a direct copy if the data in the buffer is
1990  * less than the skb header size, otherwise it will just attach the page as
1991  * a frag to the skb.
1992  *
1993  * The function will then update the page offset if necessary and return
1994  * true if the buffer can be reused by the adapter.
1995  **/
1996 static void ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
1997 			      struct ixgbe_rx_buffer *rx_buffer,
1998 			      struct sk_buff *skb,
1999 			      unsigned int size)
2000 {
2001 #if (PAGE_SIZE < 8192)
2002 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2003 #else
2004 	unsigned int truesize = rx_ring->rx_offset ?
2005 				SKB_DATA_ALIGN(rx_ring->rx_offset + size) :
2006 				SKB_DATA_ALIGN(size);
2007 #endif
2008 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
2009 			rx_buffer->page_offset, size, truesize);
2010 #if (PAGE_SIZE < 8192)
2011 	rx_buffer->page_offset ^= truesize;
2012 #else
2013 	rx_buffer->page_offset += truesize;
2014 #endif
2015 }
2016 
2017 static struct ixgbe_rx_buffer *ixgbe_get_rx_buffer(struct ixgbe_ring *rx_ring,
2018 						   union ixgbe_adv_rx_desc *rx_desc,
2019 						   struct sk_buff **skb,
2020 						   const unsigned int size,
2021 						   int *rx_buffer_pgcnt)
2022 {
2023 	struct ixgbe_rx_buffer *rx_buffer;
2024 
2025 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
2026 	*rx_buffer_pgcnt =
2027 #if (PAGE_SIZE < 8192)
2028 		page_count(rx_buffer->page);
2029 #else
2030 		0;
2031 #endif
2032 	prefetchw(rx_buffer->page);
2033 	*skb = rx_buffer->skb;
2034 
2035 	/* Delay unmapping of the first packet. It carries the header
2036 	 * information, HW may still access the header after the writeback.
2037 	 * Only unmap it when EOP is reached
2038 	 */
2039 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)) {
2040 		if (!*skb)
2041 			goto skip_sync;
2042 	} else {
2043 		if (*skb)
2044 			ixgbe_dma_sync_frag(rx_ring, *skb);
2045 	}
2046 
2047 	/* we are reusing so sync this buffer for CPU use */
2048 	dma_sync_single_range_for_cpu(rx_ring->dev,
2049 				      rx_buffer->dma,
2050 				      rx_buffer->page_offset,
2051 				      size,
2052 				      DMA_FROM_DEVICE);
2053 skip_sync:
2054 	rx_buffer->pagecnt_bias--;
2055 
2056 	return rx_buffer;
2057 }
2058 
2059 static void ixgbe_put_rx_buffer(struct ixgbe_ring *rx_ring,
2060 				struct ixgbe_rx_buffer *rx_buffer,
2061 				struct sk_buff *skb,
2062 				int rx_buffer_pgcnt)
2063 {
2064 	if (ixgbe_can_reuse_rx_page(rx_buffer, rx_buffer_pgcnt)) {
2065 		/* hand second half of page back to the ring */
2066 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2067 	} else {
2068 		if (!IS_ERR(skb) && IXGBE_CB(skb)->dma == rx_buffer->dma) {
2069 			/* the page has been released from the ring */
2070 			IXGBE_CB(skb)->page_released = true;
2071 		} else {
2072 			/* we are not reusing the buffer so unmap it */
2073 			dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
2074 					     ixgbe_rx_pg_size(rx_ring),
2075 					     DMA_FROM_DEVICE,
2076 					     IXGBE_RX_DMA_ATTR);
2077 		}
2078 		__page_frag_cache_drain(rx_buffer->page,
2079 					rx_buffer->pagecnt_bias);
2080 	}
2081 
2082 	/* clear contents of rx_buffer */
2083 	rx_buffer->page = NULL;
2084 	rx_buffer->skb = NULL;
2085 }
2086 
2087 static struct sk_buff *ixgbe_construct_skb(struct ixgbe_ring *rx_ring,
2088 					   struct ixgbe_rx_buffer *rx_buffer,
2089 					   struct xdp_buff *xdp,
2090 					   union ixgbe_adv_rx_desc *rx_desc)
2091 {
2092 	unsigned int size = xdp->data_end - xdp->data;
2093 #if (PAGE_SIZE < 8192)
2094 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2095 #else
2096 	unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end -
2097 					       xdp->data_hard_start);
2098 #endif
2099 	struct sk_buff *skb;
2100 
2101 	/* prefetch first cache line of first page */
2102 	net_prefetch(xdp->data);
2103 
2104 	/* Note, we get here by enabling legacy-rx via:
2105 	 *
2106 	 *    ethtool --set-priv-flags <dev> legacy-rx on
2107 	 *
2108 	 * In this mode, we currently get 0 extra XDP headroom as
2109 	 * opposed to having legacy-rx off, where we process XDP
2110 	 * packets going to stack via ixgbe_build_skb(). The latter
2111 	 * provides us currently with 192 bytes of headroom.
2112 	 *
2113 	 * For ixgbe_construct_skb() mode it means that the
2114 	 * xdp->data_meta will always point to xdp->data, since
2115 	 * the helper cannot expand the head. Should this ever
2116 	 * change in future for legacy-rx mode on, then lets also
2117 	 * add xdp->data_meta handling here.
2118 	 */
2119 
2120 	/* allocate a skb to store the frags */
2121 	skb = napi_alloc_skb(&rx_ring->q_vector->napi, IXGBE_RX_HDR_SIZE);
2122 	if (unlikely(!skb))
2123 		return NULL;
2124 
2125 	if (size > IXGBE_RX_HDR_SIZE) {
2126 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2127 			IXGBE_CB(skb)->dma = rx_buffer->dma;
2128 
2129 		skb_add_rx_frag(skb, 0, rx_buffer->page,
2130 				xdp->data - page_address(rx_buffer->page),
2131 				size, truesize);
2132 #if (PAGE_SIZE < 8192)
2133 		rx_buffer->page_offset ^= truesize;
2134 #else
2135 		rx_buffer->page_offset += truesize;
2136 #endif
2137 	} else {
2138 		memcpy(__skb_put(skb, size),
2139 		       xdp->data, ALIGN(size, sizeof(long)));
2140 		rx_buffer->pagecnt_bias++;
2141 	}
2142 
2143 	return skb;
2144 }
2145 
2146 static struct sk_buff *ixgbe_build_skb(struct ixgbe_ring *rx_ring,
2147 				       struct ixgbe_rx_buffer *rx_buffer,
2148 				       struct xdp_buff *xdp,
2149 				       union ixgbe_adv_rx_desc *rx_desc)
2150 {
2151 	unsigned int metasize = xdp->data - xdp->data_meta;
2152 #if (PAGE_SIZE < 8192)
2153 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2154 #else
2155 	unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
2156 				SKB_DATA_ALIGN(xdp->data_end -
2157 					       xdp->data_hard_start);
2158 #endif
2159 	struct sk_buff *skb;
2160 
2161 	/* Prefetch first cache line of first page. If xdp->data_meta
2162 	 * is unused, this points extactly as xdp->data, otherwise we
2163 	 * likely have a consumer accessing first few bytes of meta
2164 	 * data, and then actual data.
2165 	 */
2166 	net_prefetch(xdp->data_meta);
2167 
2168 	/* build an skb to around the page buffer */
2169 	skb = build_skb(xdp->data_hard_start, truesize);
2170 	if (unlikely(!skb))
2171 		return NULL;
2172 
2173 	/* update pointers within the skb to store the data */
2174 	skb_reserve(skb, xdp->data - xdp->data_hard_start);
2175 	__skb_put(skb, xdp->data_end - xdp->data);
2176 	if (metasize)
2177 		skb_metadata_set(skb, metasize);
2178 
2179 	/* record DMA address if this is the start of a chain of buffers */
2180 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2181 		IXGBE_CB(skb)->dma = rx_buffer->dma;
2182 
2183 	/* update buffer offset */
2184 #if (PAGE_SIZE < 8192)
2185 	rx_buffer->page_offset ^= truesize;
2186 #else
2187 	rx_buffer->page_offset += truesize;
2188 #endif
2189 
2190 	return skb;
2191 }
2192 
2193 static struct sk_buff *ixgbe_run_xdp(struct ixgbe_adapter *adapter,
2194 				     struct ixgbe_ring *rx_ring,
2195 				     struct xdp_buff *xdp)
2196 {
2197 	int err, result = IXGBE_XDP_PASS;
2198 	struct bpf_prog *xdp_prog;
2199 	struct xdp_frame *xdpf;
2200 	u32 act;
2201 
2202 	rcu_read_lock();
2203 	xdp_prog = READ_ONCE(rx_ring->xdp_prog);
2204 
2205 	if (!xdp_prog)
2206 		goto xdp_out;
2207 
2208 	prefetchw(xdp->data_hard_start); /* xdp_frame write */
2209 
2210 	act = bpf_prog_run_xdp(xdp_prog, xdp);
2211 	switch (act) {
2212 	case XDP_PASS:
2213 		break;
2214 	case XDP_TX:
2215 		xdpf = xdp_convert_buff_to_frame(xdp);
2216 		if (unlikely(!xdpf)) {
2217 			result = IXGBE_XDP_CONSUMED;
2218 			break;
2219 		}
2220 		result = ixgbe_xmit_xdp_ring(adapter, xdpf);
2221 		break;
2222 	case XDP_REDIRECT:
2223 		err = xdp_do_redirect(adapter->netdev, xdp, xdp_prog);
2224 		if (!err)
2225 			result = IXGBE_XDP_REDIR;
2226 		else
2227 			result = IXGBE_XDP_CONSUMED;
2228 		break;
2229 	default:
2230 		bpf_warn_invalid_xdp_action(act);
2231 		fallthrough;
2232 	case XDP_ABORTED:
2233 		trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
2234 		fallthrough; /* handle aborts by dropping packet */
2235 	case XDP_DROP:
2236 		result = IXGBE_XDP_CONSUMED;
2237 		break;
2238 	}
2239 xdp_out:
2240 	rcu_read_unlock();
2241 	return ERR_PTR(-result);
2242 }
2243 
2244 static unsigned int ixgbe_rx_frame_truesize(struct ixgbe_ring *rx_ring,
2245 					    unsigned int size)
2246 {
2247 	unsigned int truesize;
2248 
2249 #if (PAGE_SIZE < 8192)
2250 	truesize = ixgbe_rx_pg_size(rx_ring) / 2; /* Must be power-of-2 */
2251 #else
2252 	truesize = rx_ring->rx_offset ?
2253 		SKB_DATA_ALIGN(rx_ring->rx_offset + size) +
2254 		SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) :
2255 		SKB_DATA_ALIGN(size);
2256 #endif
2257 	return truesize;
2258 }
2259 
2260 static void ixgbe_rx_buffer_flip(struct ixgbe_ring *rx_ring,
2261 				 struct ixgbe_rx_buffer *rx_buffer,
2262 				 unsigned int size)
2263 {
2264 	unsigned int truesize = ixgbe_rx_frame_truesize(rx_ring, size);
2265 #if (PAGE_SIZE < 8192)
2266 	rx_buffer->page_offset ^= truesize;
2267 #else
2268 	rx_buffer->page_offset += truesize;
2269 #endif
2270 }
2271 
2272 /**
2273  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2274  * @q_vector: structure containing interrupt and ring information
2275  * @rx_ring: rx descriptor ring to transact packets on
2276  * @budget: Total limit on number of packets to process
2277  *
2278  * This function provides a "bounce buffer" approach to Rx interrupt
2279  * processing.  The advantage to this is that on systems that have
2280  * expensive overhead for IOMMU access this provides a means of avoiding
2281  * it by maintaining the mapping of the page to the syste.
2282  *
2283  * Returns amount of work completed
2284  **/
2285 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2286 			       struct ixgbe_ring *rx_ring,
2287 			       const int budget)
2288 {
2289 	unsigned int total_rx_bytes = 0, total_rx_packets = 0, frame_sz = 0;
2290 	struct ixgbe_adapter *adapter = q_vector->adapter;
2291 #ifdef IXGBE_FCOE
2292 	int ddp_bytes;
2293 	unsigned int mss = 0;
2294 #endif /* IXGBE_FCOE */
2295 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2296 	unsigned int offset = rx_ring->rx_offset;
2297 	unsigned int xdp_xmit = 0;
2298 	struct xdp_buff xdp;
2299 
2300 	/* Frame size depend on rx_ring setup when PAGE_SIZE=4K */
2301 #if (PAGE_SIZE < 8192)
2302 	frame_sz = ixgbe_rx_frame_truesize(rx_ring, 0);
2303 #endif
2304 	xdp_init_buff(&xdp, frame_sz, &rx_ring->xdp_rxq);
2305 
2306 	while (likely(total_rx_packets < budget)) {
2307 		union ixgbe_adv_rx_desc *rx_desc;
2308 		struct ixgbe_rx_buffer *rx_buffer;
2309 		struct sk_buff *skb;
2310 		int rx_buffer_pgcnt;
2311 		unsigned int size;
2312 
2313 		/* return some buffers to hardware, one at a time is too slow */
2314 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2315 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2316 			cleaned_count = 0;
2317 		}
2318 
2319 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2320 		size = le16_to_cpu(rx_desc->wb.upper.length);
2321 		if (!size)
2322 			break;
2323 
2324 		/* This memory barrier is needed to keep us from reading
2325 		 * any other fields out of the rx_desc until we know the
2326 		 * descriptor has been written back
2327 		 */
2328 		dma_rmb();
2329 
2330 		rx_buffer = ixgbe_get_rx_buffer(rx_ring, rx_desc, &skb, size, &rx_buffer_pgcnt);
2331 
2332 		/* retrieve a buffer from the ring */
2333 		if (!skb) {
2334 			unsigned char *hard_start;
2335 
2336 			hard_start = page_address(rx_buffer->page) +
2337 				     rx_buffer->page_offset - offset;
2338 			xdp_prepare_buff(&xdp, hard_start, offset, size, true);
2339 #if (PAGE_SIZE > 4096)
2340 			/* At larger PAGE_SIZE, frame_sz depend on len size */
2341 			xdp.frame_sz = ixgbe_rx_frame_truesize(rx_ring, size);
2342 #endif
2343 			skb = ixgbe_run_xdp(adapter, rx_ring, &xdp);
2344 		}
2345 
2346 		if (IS_ERR(skb)) {
2347 			unsigned int xdp_res = -PTR_ERR(skb);
2348 
2349 			if (xdp_res & (IXGBE_XDP_TX | IXGBE_XDP_REDIR)) {
2350 				xdp_xmit |= xdp_res;
2351 				ixgbe_rx_buffer_flip(rx_ring, rx_buffer, size);
2352 			} else {
2353 				rx_buffer->pagecnt_bias++;
2354 			}
2355 			total_rx_packets++;
2356 			total_rx_bytes += size;
2357 		} else if (skb) {
2358 			ixgbe_add_rx_frag(rx_ring, rx_buffer, skb, size);
2359 		} else if (ring_uses_build_skb(rx_ring)) {
2360 			skb = ixgbe_build_skb(rx_ring, rx_buffer,
2361 					      &xdp, rx_desc);
2362 		} else {
2363 			skb = ixgbe_construct_skb(rx_ring, rx_buffer,
2364 						  &xdp, rx_desc);
2365 		}
2366 
2367 		/* exit if we failed to retrieve a buffer */
2368 		if (!skb) {
2369 			rx_ring->rx_stats.alloc_rx_buff_failed++;
2370 			rx_buffer->pagecnt_bias++;
2371 			break;
2372 		}
2373 
2374 		ixgbe_put_rx_buffer(rx_ring, rx_buffer, skb, rx_buffer_pgcnt);
2375 		cleaned_count++;
2376 
2377 		/* place incomplete frames back on ring for completion */
2378 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2379 			continue;
2380 
2381 		/* verify the packet layout is correct */
2382 		if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2383 			continue;
2384 
2385 		/* probably a little skewed due to removing CRC */
2386 		total_rx_bytes += skb->len;
2387 
2388 		/* populate checksum, timestamp, VLAN, and protocol */
2389 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2390 
2391 #ifdef IXGBE_FCOE
2392 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2393 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2394 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2395 			/* include DDPed FCoE data */
2396 			if (ddp_bytes > 0) {
2397 				if (!mss) {
2398 					mss = rx_ring->netdev->mtu -
2399 						sizeof(struct fcoe_hdr) -
2400 						sizeof(struct fc_frame_header) -
2401 						sizeof(struct fcoe_crc_eof);
2402 					if (mss > 512)
2403 						mss &= ~511;
2404 				}
2405 				total_rx_bytes += ddp_bytes;
2406 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2407 								 mss);
2408 			}
2409 			if (!ddp_bytes) {
2410 				dev_kfree_skb_any(skb);
2411 				continue;
2412 			}
2413 		}
2414 
2415 #endif /* IXGBE_FCOE */
2416 		ixgbe_rx_skb(q_vector, skb);
2417 
2418 		/* update budget accounting */
2419 		total_rx_packets++;
2420 	}
2421 
2422 	if (xdp_xmit & IXGBE_XDP_REDIR)
2423 		xdp_do_flush_map();
2424 
2425 	if (xdp_xmit & IXGBE_XDP_TX) {
2426 		struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
2427 
2428 		/* Force memory writes to complete before letting h/w
2429 		 * know there are new descriptors to fetch.
2430 		 */
2431 		wmb();
2432 		writel(ring->next_to_use, ring->tail);
2433 	}
2434 
2435 	u64_stats_update_begin(&rx_ring->syncp);
2436 	rx_ring->stats.packets += total_rx_packets;
2437 	rx_ring->stats.bytes += total_rx_bytes;
2438 	u64_stats_update_end(&rx_ring->syncp);
2439 	q_vector->rx.total_packets += total_rx_packets;
2440 	q_vector->rx.total_bytes += total_rx_bytes;
2441 
2442 	return total_rx_packets;
2443 }
2444 
2445 /**
2446  * ixgbe_configure_msix - Configure MSI-X hardware
2447  * @adapter: board private structure
2448  *
2449  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2450  * interrupts.
2451  **/
2452 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2453 {
2454 	struct ixgbe_q_vector *q_vector;
2455 	int v_idx;
2456 	u32 mask;
2457 
2458 	/* Populate MSIX to EITR Select */
2459 	if (adapter->num_vfs > 32) {
2460 		u32 eitrsel = BIT(adapter->num_vfs - 32) - 1;
2461 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2462 	}
2463 
2464 	/*
2465 	 * Populate the IVAR table and set the ITR values to the
2466 	 * corresponding register.
2467 	 */
2468 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2469 		struct ixgbe_ring *ring;
2470 		q_vector = adapter->q_vector[v_idx];
2471 
2472 		ixgbe_for_each_ring(ring, q_vector->rx)
2473 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2474 
2475 		ixgbe_for_each_ring(ring, q_vector->tx)
2476 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2477 
2478 		ixgbe_write_eitr(q_vector);
2479 	}
2480 
2481 	switch (adapter->hw.mac.type) {
2482 	case ixgbe_mac_82598EB:
2483 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2484 			       v_idx);
2485 		break;
2486 	case ixgbe_mac_82599EB:
2487 	case ixgbe_mac_X540:
2488 	case ixgbe_mac_X550:
2489 	case ixgbe_mac_X550EM_x:
2490 	case ixgbe_mac_x550em_a:
2491 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2492 		break;
2493 	default:
2494 		break;
2495 	}
2496 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2497 
2498 	/* set up to autoclear timer, and the vectors */
2499 	mask = IXGBE_EIMS_ENABLE_MASK;
2500 	mask &= ~(IXGBE_EIMS_OTHER |
2501 		  IXGBE_EIMS_MAILBOX |
2502 		  IXGBE_EIMS_LSC);
2503 
2504 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2505 }
2506 
2507 /**
2508  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2509  * @q_vector: structure containing interrupt and ring information
2510  * @ring_container: structure containing ring performance data
2511  *
2512  *      Stores a new ITR value based on packets and byte
2513  *      counts during the last interrupt.  The advantage of per interrupt
2514  *      computation is faster updates and more accurate ITR for the current
2515  *      traffic pattern.  Constants in this function were computed
2516  *      based on theoretical maximum wire speed and thresholds were set based
2517  *      on testing data as well as attempting to minimize response time
2518  *      while increasing bulk throughput.
2519  **/
2520 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2521 			     struct ixgbe_ring_container *ring_container)
2522 {
2523 	unsigned int itr = IXGBE_ITR_ADAPTIVE_MIN_USECS |
2524 			   IXGBE_ITR_ADAPTIVE_LATENCY;
2525 	unsigned int avg_wire_size, packets, bytes;
2526 	unsigned long next_update = jiffies;
2527 
2528 	/* If we don't have any rings just leave ourselves set for maximum
2529 	 * possible latency so we take ourselves out of the equation.
2530 	 */
2531 	if (!ring_container->ring)
2532 		return;
2533 
2534 	/* If we didn't update within up to 1 - 2 jiffies we can assume
2535 	 * that either packets are coming in so slow there hasn't been
2536 	 * any work, or that there is so much work that NAPI is dealing
2537 	 * with interrupt moderation and we don't need to do anything.
2538 	 */
2539 	if (time_after(next_update, ring_container->next_update))
2540 		goto clear_counts;
2541 
2542 	packets = ring_container->total_packets;
2543 
2544 	/* We have no packets to actually measure against. This means
2545 	 * either one of the other queues on this vector is active or
2546 	 * we are a Tx queue doing TSO with too high of an interrupt rate.
2547 	 *
2548 	 * When this occurs just tick up our delay by the minimum value
2549 	 * and hope that this extra delay will prevent us from being called
2550 	 * without any work on our queue.
2551 	 */
2552 	if (!packets) {
2553 		itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2554 		if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2555 			itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2556 		itr += ring_container->itr & IXGBE_ITR_ADAPTIVE_LATENCY;
2557 		goto clear_counts;
2558 	}
2559 
2560 	bytes = ring_container->total_bytes;
2561 
2562 	/* If packets are less than 4 or bytes are less than 9000 assume
2563 	 * insufficient data to use bulk rate limiting approach. We are
2564 	 * likely latency driven.
2565 	 */
2566 	if (packets < 4 && bytes < 9000) {
2567 		itr = IXGBE_ITR_ADAPTIVE_LATENCY;
2568 		goto adjust_by_size;
2569 	}
2570 
2571 	/* Between 4 and 48 we can assume that our current interrupt delay
2572 	 * is only slightly too low. As such we should increase it by a small
2573 	 * fixed amount.
2574 	 */
2575 	if (packets < 48) {
2576 		itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2577 		if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2578 			itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2579 		goto clear_counts;
2580 	}
2581 
2582 	/* Between 48 and 96 is our "goldilocks" zone where we are working
2583 	 * out "just right". Just report that our current ITR is good for us.
2584 	 */
2585 	if (packets < 96) {
2586 		itr = q_vector->itr >> 2;
2587 		goto clear_counts;
2588 	}
2589 
2590 	/* If packet count is 96 or greater we are likely looking at a slight
2591 	 * overrun of the delay we want. Try halving our delay to see if that
2592 	 * will cut the number of packets in half per interrupt.
2593 	 */
2594 	if (packets < 256) {
2595 		itr = q_vector->itr >> 3;
2596 		if (itr < IXGBE_ITR_ADAPTIVE_MIN_USECS)
2597 			itr = IXGBE_ITR_ADAPTIVE_MIN_USECS;
2598 		goto clear_counts;
2599 	}
2600 
2601 	/* The paths below assume we are dealing with a bulk ITR since number
2602 	 * of packets is 256 or greater. We are just going to have to compute
2603 	 * a value and try to bring the count under control, though for smaller
2604 	 * packet sizes there isn't much we can do as NAPI polling will likely
2605 	 * be kicking in sooner rather than later.
2606 	 */
2607 	itr = IXGBE_ITR_ADAPTIVE_BULK;
2608 
2609 adjust_by_size:
2610 	/* If packet counts are 256 or greater we can assume we have a gross
2611 	 * overestimation of what the rate should be. Instead of trying to fine
2612 	 * tune it just use the formula below to try and dial in an exact value
2613 	 * give the current packet size of the frame.
2614 	 */
2615 	avg_wire_size = bytes / packets;
2616 
2617 	/* The following is a crude approximation of:
2618 	 *  wmem_default / (size + overhead) = desired_pkts_per_int
2619 	 *  rate / bits_per_byte / (size + ethernet overhead) = pkt_rate
2620 	 *  (desired_pkt_rate / pkt_rate) * usecs_per_sec = ITR value
2621 	 *
2622 	 * Assuming wmem_default is 212992 and overhead is 640 bytes per
2623 	 * packet, (256 skb, 64 headroom, 320 shared info), we can reduce the
2624 	 * formula down to
2625 	 *
2626 	 *  (170 * (size + 24)) / (size + 640) = ITR
2627 	 *
2628 	 * We first do some math on the packet size and then finally bitshift
2629 	 * by 8 after rounding up. We also have to account for PCIe link speed
2630 	 * difference as ITR scales based on this.
2631 	 */
2632 	if (avg_wire_size <= 60) {
2633 		/* Start at 50k ints/sec */
2634 		avg_wire_size = 5120;
2635 	} else if (avg_wire_size <= 316) {
2636 		/* 50K ints/sec to 16K ints/sec */
2637 		avg_wire_size *= 40;
2638 		avg_wire_size += 2720;
2639 	} else if (avg_wire_size <= 1084) {
2640 		/* 16K ints/sec to 9.2K ints/sec */
2641 		avg_wire_size *= 15;
2642 		avg_wire_size += 11452;
2643 	} else if (avg_wire_size < 1968) {
2644 		/* 9.2K ints/sec to 8K ints/sec */
2645 		avg_wire_size *= 5;
2646 		avg_wire_size += 22420;
2647 	} else {
2648 		/* plateau at a limit of 8K ints/sec */
2649 		avg_wire_size = 32256;
2650 	}
2651 
2652 	/* If we are in low latency mode half our delay which doubles the rate
2653 	 * to somewhere between 100K to 16K ints/sec
2654 	 */
2655 	if (itr & IXGBE_ITR_ADAPTIVE_LATENCY)
2656 		avg_wire_size >>= 1;
2657 
2658 	/* Resultant value is 256 times larger than it needs to be. This
2659 	 * gives us room to adjust the value as needed to either increase
2660 	 * or decrease the value based on link speeds of 10G, 2.5G, 1G, etc.
2661 	 *
2662 	 * Use addition as we have already recorded the new latency flag
2663 	 * for the ITR value.
2664 	 */
2665 	switch (q_vector->adapter->link_speed) {
2666 	case IXGBE_LINK_SPEED_10GB_FULL:
2667 	case IXGBE_LINK_SPEED_100_FULL:
2668 	default:
2669 		itr += DIV_ROUND_UP(avg_wire_size,
2670 				    IXGBE_ITR_ADAPTIVE_MIN_INC * 256) *
2671 		       IXGBE_ITR_ADAPTIVE_MIN_INC;
2672 		break;
2673 	case IXGBE_LINK_SPEED_2_5GB_FULL:
2674 	case IXGBE_LINK_SPEED_1GB_FULL:
2675 	case IXGBE_LINK_SPEED_10_FULL:
2676 		if (avg_wire_size > 8064)
2677 			avg_wire_size = 8064;
2678 		itr += DIV_ROUND_UP(avg_wire_size,
2679 				    IXGBE_ITR_ADAPTIVE_MIN_INC * 64) *
2680 		       IXGBE_ITR_ADAPTIVE_MIN_INC;
2681 		break;
2682 	}
2683 
2684 clear_counts:
2685 	/* write back value */
2686 	ring_container->itr = itr;
2687 
2688 	/* next update should occur within next jiffy */
2689 	ring_container->next_update = next_update + 1;
2690 
2691 	ring_container->total_bytes = 0;
2692 	ring_container->total_packets = 0;
2693 }
2694 
2695 /**
2696  * ixgbe_write_eitr - write EITR register in hardware specific way
2697  * @q_vector: structure containing interrupt and ring information
2698  *
2699  * This function is made to be called by ethtool and by the driver
2700  * when it needs to update EITR registers at runtime.  Hardware
2701  * specific quirks/differences are taken care of here.
2702  */
2703 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2704 {
2705 	struct ixgbe_adapter *adapter = q_vector->adapter;
2706 	struct ixgbe_hw *hw = &adapter->hw;
2707 	int v_idx = q_vector->v_idx;
2708 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2709 
2710 	switch (adapter->hw.mac.type) {
2711 	case ixgbe_mac_82598EB:
2712 		/* must write high and low 16 bits to reset counter */
2713 		itr_reg |= (itr_reg << 16);
2714 		break;
2715 	case ixgbe_mac_82599EB:
2716 	case ixgbe_mac_X540:
2717 	case ixgbe_mac_X550:
2718 	case ixgbe_mac_X550EM_x:
2719 	case ixgbe_mac_x550em_a:
2720 		/*
2721 		 * set the WDIS bit to not clear the timer bits and cause an
2722 		 * immediate assertion of the interrupt
2723 		 */
2724 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2725 		break;
2726 	default:
2727 		break;
2728 	}
2729 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2730 }
2731 
2732 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2733 {
2734 	u32 new_itr;
2735 
2736 	ixgbe_update_itr(q_vector, &q_vector->tx);
2737 	ixgbe_update_itr(q_vector, &q_vector->rx);
2738 
2739 	/* use the smallest value of new ITR delay calculations */
2740 	new_itr = min(q_vector->rx.itr, q_vector->tx.itr);
2741 
2742 	/* Clear latency flag if set, shift into correct position */
2743 	new_itr &= ~IXGBE_ITR_ADAPTIVE_LATENCY;
2744 	new_itr <<= 2;
2745 
2746 	if (new_itr != q_vector->itr) {
2747 		/* save the algorithm value here */
2748 		q_vector->itr = new_itr;
2749 
2750 		ixgbe_write_eitr(q_vector);
2751 	}
2752 }
2753 
2754 /**
2755  * ixgbe_check_overtemp_subtask - check for over temperature
2756  * @adapter: pointer to adapter
2757  **/
2758 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2759 {
2760 	struct ixgbe_hw *hw = &adapter->hw;
2761 	u32 eicr = adapter->interrupt_event;
2762 	s32 rc;
2763 
2764 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2765 		return;
2766 
2767 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2768 		return;
2769 
2770 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2771 
2772 	switch (hw->device_id) {
2773 	case IXGBE_DEV_ID_82599_T3_LOM:
2774 		/*
2775 		 * Since the warning interrupt is for both ports
2776 		 * we don't have to check if:
2777 		 *  - This interrupt wasn't for our port.
2778 		 *  - We may have missed the interrupt so always have to
2779 		 *    check if we  got a LSC
2780 		 */
2781 		if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2782 		    !(eicr & IXGBE_EICR_LSC))
2783 			return;
2784 
2785 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2786 			u32 speed;
2787 			bool link_up = false;
2788 
2789 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2790 
2791 			if (link_up)
2792 				return;
2793 		}
2794 
2795 		/* Check if this is not due to overtemp */
2796 		if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2797 			return;
2798 
2799 		break;
2800 	case IXGBE_DEV_ID_X550EM_A_1G_T:
2801 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
2802 		rc = hw->phy.ops.check_overtemp(hw);
2803 		if (rc != IXGBE_ERR_OVERTEMP)
2804 			return;
2805 		break;
2806 	default:
2807 		if (adapter->hw.mac.type >= ixgbe_mac_X540)
2808 			return;
2809 		if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
2810 			return;
2811 		break;
2812 	}
2813 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2814 
2815 	adapter->interrupt_event = 0;
2816 }
2817 
2818 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2819 {
2820 	struct ixgbe_hw *hw = &adapter->hw;
2821 
2822 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2823 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2824 		e_crit(probe, "Fan has stopped, replace the adapter\n");
2825 		/* write to clear the interrupt */
2826 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2827 	}
2828 }
2829 
2830 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2831 {
2832 	struct ixgbe_hw *hw = &adapter->hw;
2833 
2834 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2835 		return;
2836 
2837 	switch (adapter->hw.mac.type) {
2838 	case ixgbe_mac_82599EB:
2839 		/*
2840 		 * Need to check link state so complete overtemp check
2841 		 * on service task
2842 		 */
2843 		if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
2844 		     (eicr & IXGBE_EICR_LSC)) &&
2845 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2846 			adapter->interrupt_event = eicr;
2847 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2848 			ixgbe_service_event_schedule(adapter);
2849 			return;
2850 		}
2851 		return;
2852 	case ixgbe_mac_x550em_a:
2853 		if (eicr & IXGBE_EICR_GPI_SDP0_X550EM_a) {
2854 			adapter->interrupt_event = eicr;
2855 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2856 			ixgbe_service_event_schedule(adapter);
2857 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC,
2858 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2859 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICR,
2860 					IXGBE_EICR_GPI_SDP0_X550EM_a);
2861 		}
2862 		return;
2863 	case ixgbe_mac_X550:
2864 	case ixgbe_mac_X540:
2865 		if (!(eicr & IXGBE_EICR_TS))
2866 			return;
2867 		break;
2868 	default:
2869 		return;
2870 	}
2871 
2872 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
2873 }
2874 
2875 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
2876 {
2877 	switch (hw->mac.type) {
2878 	case ixgbe_mac_82598EB:
2879 		if (hw->phy.type == ixgbe_phy_nl)
2880 			return true;
2881 		return false;
2882 	case ixgbe_mac_82599EB:
2883 	case ixgbe_mac_X550EM_x:
2884 	case ixgbe_mac_x550em_a:
2885 		switch (hw->mac.ops.get_media_type(hw)) {
2886 		case ixgbe_media_type_fiber:
2887 		case ixgbe_media_type_fiber_qsfp:
2888 			return true;
2889 		default:
2890 			return false;
2891 		}
2892 	default:
2893 		return false;
2894 	}
2895 }
2896 
2897 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2898 {
2899 	struct ixgbe_hw *hw = &adapter->hw;
2900 	u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
2901 
2902 	if (!ixgbe_is_sfp(hw))
2903 		return;
2904 
2905 	/* Later MAC's use different SDP */
2906 	if (hw->mac.type >= ixgbe_mac_X540)
2907 		eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2908 
2909 	if (eicr & eicr_mask) {
2910 		/* Clear the interrupt */
2911 		IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2912 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2913 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2914 			adapter->sfp_poll_time = 0;
2915 			ixgbe_service_event_schedule(adapter);
2916 		}
2917 	}
2918 
2919 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
2920 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2921 		/* Clear the interrupt */
2922 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2923 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2924 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2925 			ixgbe_service_event_schedule(adapter);
2926 		}
2927 	}
2928 }
2929 
2930 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2931 {
2932 	struct ixgbe_hw *hw = &adapter->hw;
2933 
2934 	adapter->lsc_int++;
2935 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2936 	adapter->link_check_timeout = jiffies;
2937 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2938 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2939 		IXGBE_WRITE_FLUSH(hw);
2940 		ixgbe_service_event_schedule(adapter);
2941 	}
2942 }
2943 
2944 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2945 					   u64 qmask)
2946 {
2947 	u32 mask;
2948 	struct ixgbe_hw *hw = &adapter->hw;
2949 
2950 	switch (hw->mac.type) {
2951 	case ixgbe_mac_82598EB:
2952 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2953 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2954 		break;
2955 	case ixgbe_mac_82599EB:
2956 	case ixgbe_mac_X540:
2957 	case ixgbe_mac_X550:
2958 	case ixgbe_mac_X550EM_x:
2959 	case ixgbe_mac_x550em_a:
2960 		mask = (qmask & 0xFFFFFFFF);
2961 		if (mask)
2962 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2963 		mask = (qmask >> 32);
2964 		if (mask)
2965 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2966 		break;
2967 	default:
2968 		break;
2969 	}
2970 	/* skip the flush */
2971 }
2972 
2973 /**
2974  * ixgbe_irq_enable - Enable default interrupt generation settings
2975  * @adapter: board private structure
2976  * @queues: enable irqs for queues
2977  * @flush: flush register write
2978  **/
2979 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2980 				    bool flush)
2981 {
2982 	struct ixgbe_hw *hw = &adapter->hw;
2983 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2984 
2985 	/* don't reenable LSC while waiting for link */
2986 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2987 		mask &= ~IXGBE_EIMS_LSC;
2988 
2989 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2990 		switch (adapter->hw.mac.type) {
2991 		case ixgbe_mac_82599EB:
2992 			mask |= IXGBE_EIMS_GPI_SDP0(hw);
2993 			break;
2994 		case ixgbe_mac_X540:
2995 		case ixgbe_mac_X550:
2996 		case ixgbe_mac_X550EM_x:
2997 		case ixgbe_mac_x550em_a:
2998 			mask |= IXGBE_EIMS_TS;
2999 			break;
3000 		default:
3001 			break;
3002 		}
3003 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
3004 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
3005 	switch (adapter->hw.mac.type) {
3006 	case ixgbe_mac_82599EB:
3007 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
3008 		mask |= IXGBE_EIMS_GPI_SDP2(hw);
3009 		fallthrough;
3010 	case ixgbe_mac_X540:
3011 	case ixgbe_mac_X550:
3012 	case ixgbe_mac_X550EM_x:
3013 	case ixgbe_mac_x550em_a:
3014 		if (adapter->hw.device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
3015 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
3016 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP_N)
3017 			mask |= IXGBE_EIMS_GPI_SDP0(&adapter->hw);
3018 		if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
3019 			mask |= IXGBE_EICR_GPI_SDP0_X540;
3020 		mask |= IXGBE_EIMS_ECC;
3021 		mask |= IXGBE_EIMS_MAILBOX;
3022 		break;
3023 	default:
3024 		break;
3025 	}
3026 
3027 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
3028 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
3029 		mask |= IXGBE_EIMS_FLOW_DIR;
3030 
3031 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
3032 	if (queues)
3033 		ixgbe_irq_enable_queues(adapter, ~0);
3034 	if (flush)
3035 		IXGBE_WRITE_FLUSH(&adapter->hw);
3036 }
3037 
3038 static irqreturn_t ixgbe_msix_other(int irq, void *data)
3039 {
3040 	struct ixgbe_adapter *adapter = data;
3041 	struct ixgbe_hw *hw = &adapter->hw;
3042 	u32 eicr;
3043 
3044 	/*
3045 	 * Workaround for Silicon errata.  Use clear-by-write instead
3046 	 * of clear-by-read.  Reading with EICS will return the
3047 	 * interrupt causes without clearing, which later be done
3048 	 * with the write to EICR.
3049 	 */
3050 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
3051 
3052 	/* The lower 16bits of the EICR register are for the queue interrupts
3053 	 * which should be masked here in order to not accidentally clear them if
3054 	 * the bits are high when ixgbe_msix_other is called. There is a race
3055 	 * condition otherwise which results in possible performance loss
3056 	 * especially if the ixgbe_msix_other interrupt is triggering
3057 	 * consistently (as it would when PPS is turned on for the X540 device)
3058 	 */
3059 	eicr &= 0xFFFF0000;
3060 
3061 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
3062 
3063 	if (eicr & IXGBE_EICR_LSC)
3064 		ixgbe_check_lsc(adapter);
3065 
3066 	if (eicr & IXGBE_EICR_MAILBOX)
3067 		ixgbe_msg_task(adapter);
3068 
3069 	switch (hw->mac.type) {
3070 	case ixgbe_mac_82599EB:
3071 	case ixgbe_mac_X540:
3072 	case ixgbe_mac_X550:
3073 	case ixgbe_mac_X550EM_x:
3074 	case ixgbe_mac_x550em_a:
3075 		if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
3076 		    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
3077 			adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
3078 			ixgbe_service_event_schedule(adapter);
3079 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
3080 					IXGBE_EICR_GPI_SDP0_X540);
3081 		}
3082 		if (eicr & IXGBE_EICR_ECC) {
3083 			e_info(link, "Received ECC Err, initiating reset\n");
3084 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3085 			ixgbe_service_event_schedule(adapter);
3086 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3087 		}
3088 		/* Handle Flow Director Full threshold interrupt */
3089 		if (eicr & IXGBE_EICR_FLOW_DIR) {
3090 			int reinit_count = 0;
3091 			int i;
3092 			for (i = 0; i < adapter->num_tx_queues; i++) {
3093 				struct ixgbe_ring *ring = adapter->tx_ring[i];
3094 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
3095 						       &ring->state))
3096 					reinit_count++;
3097 			}
3098 			if (reinit_count) {
3099 				/* no more flow director interrupts until after init */
3100 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
3101 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
3102 				ixgbe_service_event_schedule(adapter);
3103 			}
3104 		}
3105 		ixgbe_check_sfp_event(adapter, eicr);
3106 		ixgbe_check_overtemp_event(adapter, eicr);
3107 		break;
3108 	default:
3109 		break;
3110 	}
3111 
3112 	ixgbe_check_fan_failure(adapter, eicr);
3113 
3114 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3115 		ixgbe_ptp_check_pps_event(adapter);
3116 
3117 	/* re-enable the original interrupt state, no lsc, no queues */
3118 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3119 		ixgbe_irq_enable(adapter, false, false);
3120 
3121 	return IRQ_HANDLED;
3122 }
3123 
3124 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
3125 {
3126 	struct ixgbe_q_vector *q_vector = data;
3127 
3128 	/* EIAM disabled interrupts (on this vector) for us */
3129 
3130 	if (q_vector->rx.ring || q_vector->tx.ring)
3131 		napi_schedule_irqoff(&q_vector->napi);
3132 
3133 	return IRQ_HANDLED;
3134 }
3135 
3136 /**
3137  * ixgbe_poll - NAPI Rx polling callback
3138  * @napi: structure for representing this polling device
3139  * @budget: how many packets driver is allowed to clean
3140  *
3141  * This function is used for legacy and MSI, NAPI mode
3142  **/
3143 int ixgbe_poll(struct napi_struct *napi, int budget)
3144 {
3145 	struct ixgbe_q_vector *q_vector =
3146 				container_of(napi, struct ixgbe_q_vector, napi);
3147 	struct ixgbe_adapter *adapter = q_vector->adapter;
3148 	struct ixgbe_ring *ring;
3149 	int per_ring_budget, work_done = 0;
3150 	bool clean_complete = true;
3151 
3152 #ifdef CONFIG_IXGBE_DCA
3153 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
3154 		ixgbe_update_dca(q_vector);
3155 #endif
3156 
3157 	ixgbe_for_each_ring(ring, q_vector->tx) {
3158 		bool wd = ring->xsk_pool ?
3159 			  ixgbe_clean_xdp_tx_irq(q_vector, ring, budget) :
3160 			  ixgbe_clean_tx_irq(q_vector, ring, budget);
3161 
3162 		if (!wd)
3163 			clean_complete = false;
3164 	}
3165 
3166 	/* Exit if we are called by netpoll */
3167 	if (budget <= 0)
3168 		return budget;
3169 
3170 	/* attempt to distribute budget to each queue fairly, but don't allow
3171 	 * the budget to go below 1 because we'll exit polling */
3172 	if (q_vector->rx.count > 1)
3173 		per_ring_budget = max(budget/q_vector->rx.count, 1);
3174 	else
3175 		per_ring_budget = budget;
3176 
3177 	ixgbe_for_each_ring(ring, q_vector->rx) {
3178 		int cleaned = ring->xsk_pool ?
3179 			      ixgbe_clean_rx_irq_zc(q_vector, ring,
3180 						    per_ring_budget) :
3181 			      ixgbe_clean_rx_irq(q_vector, ring,
3182 						 per_ring_budget);
3183 
3184 		work_done += cleaned;
3185 		if (cleaned >= per_ring_budget)
3186 			clean_complete = false;
3187 	}
3188 
3189 	/* If all work not completed, return budget and keep polling */
3190 	if (!clean_complete)
3191 		return budget;
3192 
3193 	/* all work done, exit the polling mode */
3194 	if (likely(napi_complete_done(napi, work_done))) {
3195 		if (adapter->rx_itr_setting & 1)
3196 			ixgbe_set_itr(q_vector);
3197 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
3198 			ixgbe_irq_enable_queues(adapter,
3199 						BIT_ULL(q_vector->v_idx));
3200 	}
3201 
3202 	return min(work_done, budget - 1);
3203 }
3204 
3205 /**
3206  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
3207  * @adapter: board private structure
3208  *
3209  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
3210  * interrupts from the kernel.
3211  **/
3212 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
3213 {
3214 	struct net_device *netdev = adapter->netdev;
3215 	unsigned int ri = 0, ti = 0;
3216 	int vector, err;
3217 
3218 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3219 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3220 		struct msix_entry *entry = &adapter->msix_entries[vector];
3221 
3222 		if (q_vector->tx.ring && q_vector->rx.ring) {
3223 			snprintf(q_vector->name, sizeof(q_vector->name),
3224 				 "%s-TxRx-%u", netdev->name, ri++);
3225 			ti++;
3226 		} else if (q_vector->rx.ring) {
3227 			snprintf(q_vector->name, sizeof(q_vector->name),
3228 				 "%s-rx-%u", netdev->name, ri++);
3229 		} else if (q_vector->tx.ring) {
3230 			snprintf(q_vector->name, sizeof(q_vector->name),
3231 				 "%s-tx-%u", netdev->name, ti++);
3232 		} else {
3233 			/* skip this unused q_vector */
3234 			continue;
3235 		}
3236 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
3237 				  q_vector->name, q_vector);
3238 		if (err) {
3239 			e_err(probe, "request_irq failed for MSIX interrupt "
3240 			      "Error: %d\n", err);
3241 			goto free_queue_irqs;
3242 		}
3243 		/* If Flow Director is enabled, set interrupt affinity */
3244 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3245 			/* assign the mask for this irq */
3246 			irq_set_affinity_hint(entry->vector,
3247 					      &q_vector->affinity_mask);
3248 		}
3249 	}
3250 
3251 	err = request_irq(adapter->msix_entries[vector].vector,
3252 			  ixgbe_msix_other, 0, netdev->name, adapter);
3253 	if (err) {
3254 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
3255 		goto free_queue_irqs;
3256 	}
3257 
3258 	return 0;
3259 
3260 free_queue_irqs:
3261 	while (vector) {
3262 		vector--;
3263 		irq_set_affinity_hint(adapter->msix_entries[vector].vector,
3264 				      NULL);
3265 		free_irq(adapter->msix_entries[vector].vector,
3266 			 adapter->q_vector[vector]);
3267 	}
3268 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
3269 	pci_disable_msix(adapter->pdev);
3270 	kfree(adapter->msix_entries);
3271 	adapter->msix_entries = NULL;
3272 	return err;
3273 }
3274 
3275 /**
3276  * ixgbe_intr - legacy mode Interrupt Handler
3277  * @irq: interrupt number
3278  * @data: pointer to a network interface device structure
3279  **/
3280 static irqreturn_t ixgbe_intr(int irq, void *data)
3281 {
3282 	struct ixgbe_adapter *adapter = data;
3283 	struct ixgbe_hw *hw = &adapter->hw;
3284 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3285 	u32 eicr;
3286 
3287 	/*
3288 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
3289 	 * before the read of EICR.
3290 	 */
3291 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
3292 
3293 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
3294 	 * therefore no explicit interrupt disable is necessary */
3295 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3296 	if (!eicr) {
3297 		/*
3298 		 * shared interrupt alert!
3299 		 * make sure interrupts are enabled because the read will
3300 		 * have disabled interrupts due to EIAM
3301 		 * finish the workaround of silicon errata on 82598.  Unmask
3302 		 * the interrupt that we masked before the EICR read.
3303 		 */
3304 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
3305 			ixgbe_irq_enable(adapter, true, true);
3306 		return IRQ_NONE;	/* Not our interrupt */
3307 	}
3308 
3309 	if (eicr & IXGBE_EICR_LSC)
3310 		ixgbe_check_lsc(adapter);
3311 
3312 	switch (hw->mac.type) {
3313 	case ixgbe_mac_82599EB:
3314 		ixgbe_check_sfp_event(adapter, eicr);
3315 		fallthrough;
3316 	case ixgbe_mac_X540:
3317 	case ixgbe_mac_X550:
3318 	case ixgbe_mac_X550EM_x:
3319 	case ixgbe_mac_x550em_a:
3320 		if (eicr & IXGBE_EICR_ECC) {
3321 			e_info(link, "Received ECC Err, initiating reset\n");
3322 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3323 			ixgbe_service_event_schedule(adapter);
3324 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3325 		}
3326 		ixgbe_check_overtemp_event(adapter, eicr);
3327 		break;
3328 	default:
3329 		break;
3330 	}
3331 
3332 	ixgbe_check_fan_failure(adapter, eicr);
3333 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3334 		ixgbe_ptp_check_pps_event(adapter);
3335 
3336 	/* would disable interrupts here but EIAM disabled it */
3337 	napi_schedule_irqoff(&q_vector->napi);
3338 
3339 	/*
3340 	 * re-enable link(maybe) and non-queue interrupts, no flush.
3341 	 * ixgbe_poll will re-enable the queue interrupts
3342 	 */
3343 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3344 		ixgbe_irq_enable(adapter, false, false);
3345 
3346 	return IRQ_HANDLED;
3347 }
3348 
3349 /**
3350  * ixgbe_request_irq - initialize interrupts
3351  * @adapter: board private structure
3352  *
3353  * Attempts to configure interrupts using the best available
3354  * capabilities of the hardware and kernel.
3355  **/
3356 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
3357 {
3358 	struct net_device *netdev = adapter->netdev;
3359 	int err;
3360 
3361 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
3362 		err = ixgbe_request_msix_irqs(adapter);
3363 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
3364 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
3365 				  netdev->name, adapter);
3366 	else
3367 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
3368 				  netdev->name, adapter);
3369 
3370 	if (err)
3371 		e_err(probe, "request_irq failed, Error %d\n", err);
3372 
3373 	return err;
3374 }
3375 
3376 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
3377 {
3378 	int vector;
3379 
3380 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
3381 		free_irq(adapter->pdev->irq, adapter);
3382 		return;
3383 	}
3384 
3385 	if (!adapter->msix_entries)
3386 		return;
3387 
3388 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3389 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3390 		struct msix_entry *entry = &adapter->msix_entries[vector];
3391 
3392 		/* free only the irqs that were actually requested */
3393 		if (!q_vector->rx.ring && !q_vector->tx.ring)
3394 			continue;
3395 
3396 		/* clear the affinity_mask in the IRQ descriptor */
3397 		irq_set_affinity_hint(entry->vector, NULL);
3398 
3399 		free_irq(entry->vector, q_vector);
3400 	}
3401 
3402 	free_irq(adapter->msix_entries[vector].vector, adapter);
3403 }
3404 
3405 /**
3406  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3407  * @adapter: board private structure
3408  **/
3409 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3410 {
3411 	switch (adapter->hw.mac.type) {
3412 	case ixgbe_mac_82598EB:
3413 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3414 		break;
3415 	case ixgbe_mac_82599EB:
3416 	case ixgbe_mac_X540:
3417 	case ixgbe_mac_X550:
3418 	case ixgbe_mac_X550EM_x:
3419 	case ixgbe_mac_x550em_a:
3420 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3421 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3422 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3423 		break;
3424 	default:
3425 		break;
3426 	}
3427 	IXGBE_WRITE_FLUSH(&adapter->hw);
3428 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3429 		int vector;
3430 
3431 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
3432 			synchronize_irq(adapter->msix_entries[vector].vector);
3433 
3434 		synchronize_irq(adapter->msix_entries[vector++].vector);
3435 	} else {
3436 		synchronize_irq(adapter->pdev->irq);
3437 	}
3438 }
3439 
3440 /**
3441  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3442  * @adapter: board private structure
3443  *
3444  **/
3445 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3446 {
3447 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3448 
3449 	ixgbe_write_eitr(q_vector);
3450 
3451 	ixgbe_set_ivar(adapter, 0, 0, 0);
3452 	ixgbe_set_ivar(adapter, 1, 0, 0);
3453 
3454 	e_info(hw, "Legacy interrupt IVAR setup done\n");
3455 }
3456 
3457 /**
3458  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3459  * @adapter: board private structure
3460  * @ring: structure containing ring specific data
3461  *
3462  * Configure the Tx descriptor ring after a reset.
3463  **/
3464 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3465 			     struct ixgbe_ring *ring)
3466 {
3467 	struct ixgbe_hw *hw = &adapter->hw;
3468 	u64 tdba = ring->dma;
3469 	int wait_loop = 10;
3470 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
3471 	u8 reg_idx = ring->reg_idx;
3472 
3473 	ring->xsk_pool = NULL;
3474 	if (ring_is_xdp(ring))
3475 		ring->xsk_pool = ixgbe_xsk_pool(adapter, ring);
3476 
3477 	/* disable queue to avoid issues while updating state */
3478 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3479 	IXGBE_WRITE_FLUSH(hw);
3480 
3481 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3482 			(tdba & DMA_BIT_MASK(32)));
3483 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3484 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3485 			ring->count * sizeof(union ixgbe_adv_tx_desc));
3486 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3487 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3488 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3489 
3490 	/*
3491 	 * set WTHRESH to encourage burst writeback, it should not be set
3492 	 * higher than 1 when:
3493 	 * - ITR is 0 as it could cause false TX hangs
3494 	 * - ITR is set to > 100k int/sec and BQL is enabled
3495 	 *
3496 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3497 	 * to or less than the number of on chip descriptors, which is
3498 	 * currently 40.
3499 	 */
3500 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3501 		txdctl |= 1u << 16;	/* WTHRESH = 1 */
3502 	else
3503 		txdctl |= 8u << 16;	/* WTHRESH = 8 */
3504 
3505 	/*
3506 	 * Setting PTHRESH to 32 both improves performance
3507 	 * and avoids a TX hang with DFP enabled
3508 	 */
3509 	txdctl |= (1u << 8) |	/* HTHRESH = 1 */
3510 		   32;		/* PTHRESH = 32 */
3511 
3512 	/* reinitialize flowdirector state */
3513 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3514 		ring->atr_sample_rate = adapter->atr_sample_rate;
3515 		ring->atr_count = 0;
3516 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3517 	} else {
3518 		ring->atr_sample_rate = 0;
3519 	}
3520 
3521 	/* initialize XPS */
3522 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3523 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3524 
3525 		if (q_vector)
3526 			netif_set_xps_queue(ring->netdev,
3527 					    &q_vector->affinity_mask,
3528 					    ring->queue_index);
3529 	}
3530 
3531 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3532 
3533 	/* reinitialize tx_buffer_info */
3534 	memset(ring->tx_buffer_info, 0,
3535 	       sizeof(struct ixgbe_tx_buffer) * ring->count);
3536 
3537 	/* enable queue */
3538 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3539 
3540 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3541 	if (hw->mac.type == ixgbe_mac_82598EB &&
3542 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3543 		return;
3544 
3545 	/* poll to verify queue is enabled */
3546 	do {
3547 		usleep_range(1000, 2000);
3548 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3549 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3550 	if (!wait_loop)
3551 		hw_dbg(hw, "Could not enable Tx Queue %d\n", reg_idx);
3552 }
3553 
3554 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3555 {
3556 	struct ixgbe_hw *hw = &adapter->hw;
3557 	u32 rttdcs, mtqc;
3558 	u8 tcs = adapter->hw_tcs;
3559 
3560 	if (hw->mac.type == ixgbe_mac_82598EB)
3561 		return;
3562 
3563 	/* disable the arbiter while setting MTQC */
3564 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3565 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
3566 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3567 
3568 	/* set transmit pool layout */
3569 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3570 		mtqc = IXGBE_MTQC_VT_ENA;
3571 		if (tcs > 4)
3572 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3573 		else if (tcs > 1)
3574 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3575 		else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3576 			 IXGBE_82599_VMDQ_4Q_MASK)
3577 			mtqc |= IXGBE_MTQC_32VF;
3578 		else
3579 			mtqc |= IXGBE_MTQC_64VF;
3580 	} else {
3581 		if (tcs > 4) {
3582 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3583 		} else if (tcs > 1) {
3584 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3585 		} else {
3586 			u8 max_txq = adapter->num_tx_queues +
3587 				adapter->num_xdp_queues;
3588 			if (max_txq > 63)
3589 				mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3590 			else
3591 				mtqc = IXGBE_MTQC_64Q_1PB;
3592 		}
3593 	}
3594 
3595 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3596 
3597 	/* Enable Security TX Buffer IFG for multiple pb */
3598 	if (tcs) {
3599 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3600 		sectx |= IXGBE_SECTX_DCB;
3601 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3602 	}
3603 
3604 	/* re-enable the arbiter */
3605 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3606 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3607 }
3608 
3609 /**
3610  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3611  * @adapter: board private structure
3612  *
3613  * Configure the Tx unit of the MAC after a reset.
3614  **/
3615 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3616 {
3617 	struct ixgbe_hw *hw = &adapter->hw;
3618 	u32 dmatxctl;
3619 	u32 i;
3620 
3621 	ixgbe_setup_mtqc(adapter);
3622 
3623 	if (hw->mac.type != ixgbe_mac_82598EB) {
3624 		/* DMATXCTL.EN must be before Tx queues are enabled */
3625 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3626 		dmatxctl |= IXGBE_DMATXCTL_TE;
3627 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3628 	}
3629 
3630 	/* Setup the HW Tx Head and Tail descriptor pointers */
3631 	for (i = 0; i < adapter->num_tx_queues; i++)
3632 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3633 	for (i = 0; i < adapter->num_xdp_queues; i++)
3634 		ixgbe_configure_tx_ring(adapter, adapter->xdp_ring[i]);
3635 }
3636 
3637 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3638 				 struct ixgbe_ring *ring)
3639 {
3640 	struct ixgbe_hw *hw = &adapter->hw;
3641 	u8 reg_idx = ring->reg_idx;
3642 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3643 
3644 	srrctl |= IXGBE_SRRCTL_DROP_EN;
3645 
3646 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3647 }
3648 
3649 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3650 				  struct ixgbe_ring *ring)
3651 {
3652 	struct ixgbe_hw *hw = &adapter->hw;
3653 	u8 reg_idx = ring->reg_idx;
3654 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3655 
3656 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3657 
3658 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3659 }
3660 
3661 #ifdef CONFIG_IXGBE_DCB
3662 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3663 #else
3664 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3665 #endif
3666 {
3667 	int i;
3668 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3669 
3670 	if (adapter->ixgbe_ieee_pfc)
3671 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3672 
3673 	/*
3674 	 * We should set the drop enable bit if:
3675 	 *  SR-IOV is enabled
3676 	 *   or
3677 	 *  Number of Rx queues > 1 and flow control is disabled
3678 	 *
3679 	 *  This allows us to avoid head of line blocking for security
3680 	 *  and performance reasons.
3681 	 */
3682 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3683 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3684 		for (i = 0; i < adapter->num_rx_queues; i++)
3685 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3686 	} else {
3687 		for (i = 0; i < adapter->num_rx_queues; i++)
3688 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3689 	}
3690 }
3691 
3692 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3693 
3694 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3695 				   struct ixgbe_ring *rx_ring)
3696 {
3697 	struct ixgbe_hw *hw = &adapter->hw;
3698 	u32 srrctl;
3699 	u8 reg_idx = rx_ring->reg_idx;
3700 
3701 	if (hw->mac.type == ixgbe_mac_82598EB) {
3702 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3703 
3704 		/*
3705 		 * if VMDq is not active we must program one srrctl register
3706 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
3707 		 */
3708 		reg_idx &= mask;
3709 	}
3710 
3711 	/* configure header buffer length, needed for RSC */
3712 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3713 
3714 	/* configure the packet buffer length */
3715 	if (rx_ring->xsk_pool) {
3716 		u32 xsk_buf_len = xsk_pool_get_rx_frame_size(rx_ring->xsk_pool);
3717 
3718 		/* If the MAC support setting RXDCTL.RLPML, the
3719 		 * SRRCTL[n].BSIZEPKT is set to PAGE_SIZE and
3720 		 * RXDCTL.RLPML is set to the actual UMEM buffer
3721 		 * size. If not, then we are stuck with a 1k buffer
3722 		 * size resolution. In this case frames larger than
3723 		 * the UMEM buffer size viewed in a 1k resolution will
3724 		 * be dropped.
3725 		 */
3726 		if (hw->mac.type != ixgbe_mac_82599EB)
3727 			srrctl |= PAGE_SIZE >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3728 		else
3729 			srrctl |= xsk_buf_len >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3730 	} else if (test_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state)) {
3731 		srrctl |= IXGBE_RXBUFFER_3K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3732 	} else {
3733 		srrctl |= IXGBE_RXBUFFER_2K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3734 	}
3735 
3736 	/* configure descriptor type */
3737 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3738 
3739 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3740 }
3741 
3742 /**
3743  * ixgbe_rss_indir_tbl_entries - Return RSS indirection table entries
3744  * @adapter: device handle
3745  *
3746  *  - 82598/82599/X540:     128
3747  *  - X550(non-SRIOV mode): 512
3748  *  - X550(SRIOV mode):     64
3749  */
3750 u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3751 {
3752 	if (adapter->hw.mac.type < ixgbe_mac_X550)
3753 		return 128;
3754 	else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3755 		return 64;
3756 	else
3757 		return 512;
3758 }
3759 
3760 /**
3761  * ixgbe_store_key - Write the RSS key to HW
3762  * @adapter: device handle
3763  *
3764  * Write the RSS key stored in adapter.rss_key to HW.
3765  */
3766 void ixgbe_store_key(struct ixgbe_adapter *adapter)
3767 {
3768 	struct ixgbe_hw *hw = &adapter->hw;
3769 	int i;
3770 
3771 	for (i = 0; i < 10; i++)
3772 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3773 }
3774 
3775 /**
3776  * ixgbe_init_rss_key - Initialize adapter RSS key
3777  * @adapter: device handle
3778  *
3779  * Allocates and initializes the RSS key if it is not allocated.
3780  **/
3781 static inline int ixgbe_init_rss_key(struct ixgbe_adapter *adapter)
3782 {
3783 	u32 *rss_key;
3784 
3785 	if (!adapter->rss_key) {
3786 		rss_key = kzalloc(IXGBE_RSS_KEY_SIZE, GFP_KERNEL);
3787 		if (unlikely(!rss_key))
3788 			return -ENOMEM;
3789 
3790 		netdev_rss_key_fill(rss_key, IXGBE_RSS_KEY_SIZE);
3791 		adapter->rss_key = rss_key;
3792 	}
3793 
3794 	return 0;
3795 }
3796 
3797 /**
3798  * ixgbe_store_reta - Write the RETA table to HW
3799  * @adapter: device handle
3800  *
3801  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3802  */
3803 void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3804 {
3805 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3806 	struct ixgbe_hw *hw = &adapter->hw;
3807 	u32 reta = 0;
3808 	u32 indices_multi;
3809 	u8 *indir_tbl = adapter->rss_indir_tbl;
3810 
3811 	/* Fill out the redirection table as follows:
3812 	 *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3813 	 *    indices.
3814 	 *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3815 	 *  - X550:       8 bit wide entries containing 6 bit RSS index
3816 	 */
3817 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3818 		indices_multi = 0x11;
3819 	else
3820 		indices_multi = 0x1;
3821 
3822 	/* Write redirection table to HW */
3823 	for (i = 0; i < reta_entries; i++) {
3824 		reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3825 		if ((i & 3) == 3) {
3826 			if (i < 128)
3827 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3828 			else
3829 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3830 						reta);
3831 			reta = 0;
3832 		}
3833 	}
3834 }
3835 
3836 /**
3837  * ixgbe_store_vfreta - Write the RETA table to HW (x550 devices in SRIOV mode)
3838  * @adapter: device handle
3839  *
3840  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3841  */
3842 static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3843 {
3844 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3845 	struct ixgbe_hw *hw = &adapter->hw;
3846 	u32 vfreta = 0;
3847 
3848 	/* Write redirection table to HW */
3849 	for (i = 0; i < reta_entries; i++) {
3850 		u16 pool = adapter->num_rx_pools;
3851 
3852 		vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3853 		if ((i & 3) != 3)
3854 			continue;
3855 
3856 		while (pool--)
3857 			IXGBE_WRITE_REG(hw,
3858 					IXGBE_PFVFRETA(i >> 2, VMDQ_P(pool)),
3859 					vfreta);
3860 		vfreta = 0;
3861 	}
3862 }
3863 
3864 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3865 {
3866 	u32 i, j;
3867 	u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3868 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3869 
3870 	/* Program table for at least 4 queues w/ SR-IOV so that VFs can
3871 	 * make full use of any rings they may have.  We will use the
3872 	 * PSRTYPE register to control how many rings we use within the PF.
3873 	 */
3874 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 4))
3875 		rss_i = 4;
3876 
3877 	/* Fill out hash function seeds */
3878 	ixgbe_store_key(adapter);
3879 
3880 	/* Fill out redirection table */
3881 	memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3882 
3883 	for (i = 0, j = 0; i < reta_entries; i++, j++) {
3884 		if (j == rss_i)
3885 			j = 0;
3886 
3887 		adapter->rss_indir_tbl[i] = j;
3888 	}
3889 
3890 	ixgbe_store_reta(adapter);
3891 }
3892 
3893 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3894 {
3895 	struct ixgbe_hw *hw = &adapter->hw;
3896 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3897 	int i, j;
3898 
3899 	/* Fill out hash function seeds */
3900 	for (i = 0; i < 10; i++) {
3901 		u16 pool = adapter->num_rx_pools;
3902 
3903 		while (pool--)
3904 			IXGBE_WRITE_REG(hw,
3905 					IXGBE_PFVFRSSRK(i, VMDQ_P(pool)),
3906 					*(adapter->rss_key + i));
3907 	}
3908 
3909 	/* Fill out the redirection table */
3910 	for (i = 0, j = 0; i < 64; i++, j++) {
3911 		if (j == rss_i)
3912 			j = 0;
3913 
3914 		adapter->rss_indir_tbl[i] = j;
3915 	}
3916 
3917 	ixgbe_store_vfreta(adapter);
3918 }
3919 
3920 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3921 {
3922 	struct ixgbe_hw *hw = &adapter->hw;
3923 	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3924 	u32 rxcsum;
3925 
3926 	/* Disable indicating checksum in descriptor, enables RSS hash */
3927 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3928 	rxcsum |= IXGBE_RXCSUM_PCSD;
3929 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3930 
3931 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3932 		if (adapter->ring_feature[RING_F_RSS].mask)
3933 			mrqc = IXGBE_MRQC_RSSEN;
3934 	} else {
3935 		u8 tcs = adapter->hw_tcs;
3936 
3937 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3938 			if (tcs > 4)
3939 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
3940 			else if (tcs > 1)
3941 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
3942 			else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3943 				 IXGBE_82599_VMDQ_4Q_MASK)
3944 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
3945 			else
3946 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
3947 
3948 			/* Enable L3/L4 for Tx Switched packets only for X550,
3949 			 * older devices do not support this feature
3950 			 */
3951 			if (hw->mac.type >= ixgbe_mac_X550)
3952 				mrqc |= IXGBE_MRQC_L3L4TXSWEN;
3953 		} else {
3954 			if (tcs > 4)
3955 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
3956 			else if (tcs > 1)
3957 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
3958 			else
3959 				mrqc = IXGBE_MRQC_RSSEN;
3960 		}
3961 	}
3962 
3963 	/* Perform hash on these packet types */
3964 	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3965 		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3966 		     IXGBE_MRQC_RSS_FIELD_IPV6 |
3967 		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3968 
3969 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3970 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3971 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3972 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3973 
3974 	if ((hw->mac.type >= ixgbe_mac_X550) &&
3975 	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
3976 		u16 pool = adapter->num_rx_pools;
3977 
3978 		/* Enable VF RSS mode */
3979 		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
3980 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3981 
3982 		/* Setup RSS through the VF registers */
3983 		ixgbe_setup_vfreta(adapter);
3984 		vfmrqc = IXGBE_MRQC_RSSEN;
3985 		vfmrqc |= rss_field;
3986 
3987 		while (pool--)
3988 			IXGBE_WRITE_REG(hw,
3989 					IXGBE_PFVFMRQC(VMDQ_P(pool)),
3990 					vfmrqc);
3991 	} else {
3992 		ixgbe_setup_reta(adapter);
3993 		mrqc |= rss_field;
3994 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3995 	}
3996 }
3997 
3998 /**
3999  * ixgbe_configure_rscctl - enable RSC for the indicated ring
4000  * @adapter: address of board private structure
4001  * @ring: structure containing ring specific data
4002  **/
4003 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
4004 				   struct ixgbe_ring *ring)
4005 {
4006 	struct ixgbe_hw *hw = &adapter->hw;
4007 	u32 rscctrl;
4008 	u8 reg_idx = ring->reg_idx;
4009 
4010 	if (!ring_is_rsc_enabled(ring))
4011 		return;
4012 
4013 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
4014 	rscctrl |= IXGBE_RSCCTL_RSCEN;
4015 	/*
4016 	 * we must limit the number of descriptors so that the
4017 	 * total size of max desc * buf_len is not greater
4018 	 * than 65536
4019 	 */
4020 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
4021 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
4022 }
4023 
4024 #define IXGBE_MAX_RX_DESC_POLL 10
4025 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
4026 				       struct ixgbe_ring *ring)
4027 {
4028 	struct ixgbe_hw *hw = &adapter->hw;
4029 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
4030 	u32 rxdctl;
4031 	u8 reg_idx = ring->reg_idx;
4032 
4033 	if (ixgbe_removed(hw->hw_addr))
4034 		return;
4035 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
4036 	if (hw->mac.type == ixgbe_mac_82598EB &&
4037 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
4038 		return;
4039 
4040 	do {
4041 		usleep_range(1000, 2000);
4042 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4043 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
4044 
4045 	if (!wait_loop) {
4046 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
4047 		      "the polling period\n", reg_idx);
4048 	}
4049 }
4050 
4051 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
4052 			     struct ixgbe_ring *ring)
4053 {
4054 	struct ixgbe_hw *hw = &adapter->hw;
4055 	union ixgbe_adv_rx_desc *rx_desc;
4056 	u64 rdba = ring->dma;
4057 	u32 rxdctl;
4058 	u8 reg_idx = ring->reg_idx;
4059 
4060 	xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq);
4061 	ring->xsk_pool = ixgbe_xsk_pool(adapter, ring);
4062 	if (ring->xsk_pool) {
4063 		WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
4064 						   MEM_TYPE_XSK_BUFF_POOL,
4065 						   NULL));
4066 		xsk_pool_set_rxq_info(ring->xsk_pool, &ring->xdp_rxq);
4067 	} else {
4068 		WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
4069 						   MEM_TYPE_PAGE_SHARED, NULL));
4070 	}
4071 
4072 	/* disable queue to avoid use of these values while updating state */
4073 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4074 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
4075 
4076 	/* write value back with RXDCTL.ENABLE bit cleared */
4077 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4078 	IXGBE_WRITE_FLUSH(hw);
4079 
4080 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
4081 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
4082 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
4083 			ring->count * sizeof(union ixgbe_adv_rx_desc));
4084 	/* Force flushing of IXGBE_RDLEN to prevent MDD */
4085 	IXGBE_WRITE_FLUSH(hw);
4086 
4087 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
4088 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
4089 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
4090 
4091 	ixgbe_configure_srrctl(adapter, ring);
4092 	ixgbe_configure_rscctl(adapter, ring);
4093 
4094 	if (hw->mac.type == ixgbe_mac_82598EB) {
4095 		/*
4096 		 * enable cache line friendly hardware writes:
4097 		 * PTHRESH=32 descriptors (half the internal cache),
4098 		 * this also removes ugly rx_no_buffer_count increment
4099 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
4100 		 * WTHRESH=8 burst writeback up to two cache lines
4101 		 */
4102 		rxdctl &= ~0x3FFFFF;
4103 		rxdctl |=  0x080420;
4104 #if (PAGE_SIZE < 8192)
4105 	/* RXDCTL.RLPML does not work on 82599 */
4106 	} else if (hw->mac.type != ixgbe_mac_82599EB) {
4107 		rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
4108 			    IXGBE_RXDCTL_RLPML_EN);
4109 
4110 		/* Limit the maximum frame size so we don't overrun the skb.
4111 		 * This can happen in SRIOV mode when the MTU of the VF is
4112 		 * higher than the MTU of the PF.
4113 		 */
4114 		if (ring_uses_build_skb(ring) &&
4115 		    !test_bit(__IXGBE_RX_3K_BUFFER, &ring->state))
4116 			rxdctl |= IXGBE_MAX_2K_FRAME_BUILD_SKB |
4117 				  IXGBE_RXDCTL_RLPML_EN;
4118 #endif
4119 	}
4120 
4121 	ring->rx_offset = ixgbe_rx_offset(ring);
4122 
4123 	if (ring->xsk_pool && hw->mac.type != ixgbe_mac_82599EB) {
4124 		u32 xsk_buf_len = xsk_pool_get_rx_frame_size(ring->xsk_pool);
4125 
4126 		rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
4127 			    IXGBE_RXDCTL_RLPML_EN);
4128 		rxdctl |= xsk_buf_len | IXGBE_RXDCTL_RLPML_EN;
4129 
4130 		ring->rx_buf_len = xsk_buf_len;
4131 	}
4132 
4133 	/* initialize rx_buffer_info */
4134 	memset(ring->rx_buffer_info, 0,
4135 	       sizeof(struct ixgbe_rx_buffer) * ring->count);
4136 
4137 	/* initialize Rx descriptor 0 */
4138 	rx_desc = IXGBE_RX_DESC(ring, 0);
4139 	rx_desc->wb.upper.length = 0;
4140 
4141 	/* enable receive descriptor ring */
4142 	rxdctl |= IXGBE_RXDCTL_ENABLE;
4143 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4144 
4145 	ixgbe_rx_desc_queue_enable(adapter, ring);
4146 	if (ring->xsk_pool)
4147 		ixgbe_alloc_rx_buffers_zc(ring, ixgbe_desc_unused(ring));
4148 	else
4149 		ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
4150 }
4151 
4152 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
4153 {
4154 	struct ixgbe_hw *hw = &adapter->hw;
4155 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
4156 	u16 pool = adapter->num_rx_pools;
4157 
4158 	/* PSRTYPE must be initialized in non 82598 adapters */
4159 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4160 		      IXGBE_PSRTYPE_UDPHDR |
4161 		      IXGBE_PSRTYPE_IPV4HDR |
4162 		      IXGBE_PSRTYPE_L2HDR |
4163 		      IXGBE_PSRTYPE_IPV6HDR;
4164 
4165 	if (hw->mac.type == ixgbe_mac_82598EB)
4166 		return;
4167 
4168 	if (rss_i > 3)
4169 		psrtype |= 2u << 29;
4170 	else if (rss_i > 1)
4171 		psrtype |= 1u << 29;
4172 
4173 	while (pool--)
4174 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4175 }
4176 
4177 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
4178 {
4179 	struct ixgbe_hw *hw = &adapter->hw;
4180 	u16 pool = adapter->num_rx_pools;
4181 	u32 reg_offset, vf_shift, vmolr;
4182 	u32 gcr_ext, vmdctl;
4183 	int i;
4184 
4185 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
4186 		return;
4187 
4188 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
4189 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
4190 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
4191 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
4192 	vmdctl |= IXGBE_VT_CTL_REPLEN;
4193 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
4194 
4195 	/* accept untagged packets until a vlan tag is
4196 	 * specifically set for the VMDQ queue/pool
4197 	 */
4198 	vmolr = IXGBE_VMOLR_AUPE;
4199 	while (pool--)
4200 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(pool)), vmolr);
4201 
4202 	vf_shift = VMDQ_P(0) % 32;
4203 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
4204 
4205 	/* Enable only the PF's pool for Tx/Rx */
4206 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), GENMASK(31, vf_shift));
4207 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
4208 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), GENMASK(31, vf_shift));
4209 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
4210 	if (adapter->bridge_mode == BRIDGE_MODE_VEB)
4211 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
4212 
4213 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
4214 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
4215 
4216 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
4217 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4218 
4219 	/*
4220 	 * Set up VF register offsets for selected VT Mode,
4221 	 * i.e. 32 or 64 VFs for SR-IOV
4222 	 */
4223 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4224 	case IXGBE_82599_VMDQ_8Q_MASK:
4225 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
4226 		break;
4227 	case IXGBE_82599_VMDQ_4Q_MASK:
4228 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
4229 		break;
4230 	default:
4231 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
4232 		break;
4233 	}
4234 
4235 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
4236 
4237 	for (i = 0; i < adapter->num_vfs; i++) {
4238 		/* configure spoof checking */
4239 		ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i,
4240 					  adapter->vfinfo[i].spoofchk_enabled);
4241 
4242 		/* Enable/Disable RSS query feature  */
4243 		ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
4244 					  adapter->vfinfo[i].rss_query_enabled);
4245 	}
4246 }
4247 
4248 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
4249 {
4250 	struct ixgbe_hw *hw = &adapter->hw;
4251 	struct net_device *netdev = adapter->netdev;
4252 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4253 	struct ixgbe_ring *rx_ring;
4254 	int i;
4255 	u32 mhadd, hlreg0;
4256 
4257 #ifdef IXGBE_FCOE
4258 	/* adjust max frame to be able to do baby jumbo for FCoE */
4259 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
4260 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
4261 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4262 
4263 #endif /* IXGBE_FCOE */
4264 
4265 	/* adjust max frame to be at least the size of a standard frame */
4266 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
4267 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
4268 
4269 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
4270 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
4271 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
4272 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
4273 
4274 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
4275 	}
4276 
4277 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4278 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
4279 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4280 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4281 
4282 	/*
4283 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4284 	 * the Base and Length of the Rx Descriptor Ring
4285 	 */
4286 	for (i = 0; i < adapter->num_rx_queues; i++) {
4287 		rx_ring = adapter->rx_ring[i];
4288 
4289 		clear_ring_rsc_enabled(rx_ring);
4290 		clear_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4291 		clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4292 
4293 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4294 			set_ring_rsc_enabled(rx_ring);
4295 
4296 		if (test_bit(__IXGBE_RX_FCOE, &rx_ring->state))
4297 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4298 
4299 		if (adapter->flags2 & IXGBE_FLAG2_RX_LEGACY)
4300 			continue;
4301 
4302 		set_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4303 
4304 #if (PAGE_SIZE < 8192)
4305 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4306 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4307 
4308 		if (IXGBE_2K_TOO_SMALL_WITH_PADDING ||
4309 		    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
4310 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4311 #endif
4312 	}
4313 }
4314 
4315 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
4316 {
4317 	struct ixgbe_hw *hw = &adapter->hw;
4318 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
4319 
4320 	switch (hw->mac.type) {
4321 	case ixgbe_mac_82598EB:
4322 		/*
4323 		 * For VMDq support of different descriptor types or
4324 		 * buffer sizes through the use of multiple SRRCTL
4325 		 * registers, RDRXCTL.MVMEN must be set to 1
4326 		 *
4327 		 * also, the manual doesn't mention it clearly but DCA hints
4328 		 * will only use queue 0's tags unless this bit is set.  Side
4329 		 * effects of setting this bit are only that SRRCTL must be
4330 		 * fully programmed [0..15]
4331 		 */
4332 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
4333 		break;
4334 	case ixgbe_mac_X550:
4335 	case ixgbe_mac_X550EM_x:
4336 	case ixgbe_mac_x550em_a:
4337 		if (adapter->num_vfs)
4338 			rdrxctl |= IXGBE_RDRXCTL_PSP;
4339 		fallthrough;
4340 	case ixgbe_mac_82599EB:
4341 	case ixgbe_mac_X540:
4342 		/* Disable RSC for ACK packets */
4343 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
4344 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
4345 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
4346 		/* hardware requires some bits to be set by default */
4347 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
4348 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
4349 		break;
4350 	default:
4351 		/* We should do nothing since we don't know this hardware */
4352 		return;
4353 	}
4354 
4355 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
4356 }
4357 
4358 /**
4359  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
4360  * @adapter: board private structure
4361  *
4362  * Configure the Rx unit of the MAC after a reset.
4363  **/
4364 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
4365 {
4366 	struct ixgbe_hw *hw = &adapter->hw;
4367 	int i;
4368 	u32 rxctrl, rfctl;
4369 
4370 	/* disable receives while setting up the descriptors */
4371 	hw->mac.ops.disable_rx(hw);
4372 
4373 	ixgbe_setup_psrtype(adapter);
4374 	ixgbe_setup_rdrxctl(adapter);
4375 
4376 	/* RSC Setup */
4377 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
4378 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
4379 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
4380 		rfctl |= IXGBE_RFCTL_RSC_DIS;
4381 
4382 	/* disable NFS filtering */
4383 	rfctl |= (IXGBE_RFCTL_NFSW_DIS | IXGBE_RFCTL_NFSR_DIS);
4384 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
4385 
4386 	/* Program registers for the distribution of queues */
4387 	ixgbe_setup_mrqc(adapter);
4388 
4389 	/* set_rx_buffer_len must be called before ring initialization */
4390 	ixgbe_set_rx_buffer_len(adapter);
4391 
4392 	/*
4393 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4394 	 * the Base and Length of the Rx Descriptor Ring
4395 	 */
4396 	for (i = 0; i < adapter->num_rx_queues; i++)
4397 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
4398 
4399 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4400 	/* disable drop enable for 82598 parts */
4401 	if (hw->mac.type == ixgbe_mac_82598EB)
4402 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
4403 
4404 	/* enable all receives */
4405 	rxctrl |= IXGBE_RXCTRL_RXEN;
4406 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
4407 }
4408 
4409 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
4410 				 __be16 proto, u16 vid)
4411 {
4412 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4413 	struct ixgbe_hw *hw = &adapter->hw;
4414 
4415 	/* add VID to filter table */
4416 	if (!vid || !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4417 		hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true, !!vid);
4418 
4419 	set_bit(vid, adapter->active_vlans);
4420 
4421 	return 0;
4422 }
4423 
4424 static int ixgbe_find_vlvf_entry(struct ixgbe_hw *hw, u32 vlan)
4425 {
4426 	u32 vlvf;
4427 	int idx;
4428 
4429 	/* short cut the special case */
4430 	if (vlan == 0)
4431 		return 0;
4432 
4433 	/* Search for the vlan id in the VLVF entries */
4434 	for (idx = IXGBE_VLVF_ENTRIES; --idx;) {
4435 		vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(idx));
4436 		if ((vlvf & VLAN_VID_MASK) == vlan)
4437 			break;
4438 	}
4439 
4440 	return idx;
4441 }
4442 
4443 void ixgbe_update_pf_promisc_vlvf(struct ixgbe_adapter *adapter, u32 vid)
4444 {
4445 	struct ixgbe_hw *hw = &adapter->hw;
4446 	u32 bits, word;
4447 	int idx;
4448 
4449 	idx = ixgbe_find_vlvf_entry(hw, vid);
4450 	if (!idx)
4451 		return;
4452 
4453 	/* See if any other pools are set for this VLAN filter
4454 	 * entry other than the PF.
4455 	 */
4456 	word = idx * 2 + (VMDQ_P(0) / 32);
4457 	bits = ~BIT(VMDQ_P(0) % 32);
4458 	bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4459 
4460 	/* Disable the filter so this falls into the default pool. */
4461 	if (!bits && !IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1))) {
4462 		if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4463 			IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), 0);
4464 		IXGBE_WRITE_REG(hw, IXGBE_VLVF(idx), 0);
4465 	}
4466 }
4467 
4468 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
4469 				  __be16 proto, u16 vid)
4470 {
4471 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4472 	struct ixgbe_hw *hw = &adapter->hw;
4473 
4474 	/* remove VID from filter table */
4475 	if (vid && !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4476 		hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), false, true);
4477 
4478 	clear_bit(vid, adapter->active_vlans);
4479 
4480 	return 0;
4481 }
4482 
4483 /**
4484  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
4485  * @adapter: driver data
4486  */
4487 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
4488 {
4489 	struct ixgbe_hw *hw = &adapter->hw;
4490 	u32 vlnctrl;
4491 	int i, j;
4492 
4493 	switch (hw->mac.type) {
4494 	case ixgbe_mac_82598EB:
4495 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4496 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
4497 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4498 		break;
4499 	case ixgbe_mac_82599EB:
4500 	case ixgbe_mac_X540:
4501 	case ixgbe_mac_X550:
4502 	case ixgbe_mac_X550EM_x:
4503 	case ixgbe_mac_x550em_a:
4504 		for (i = 0; i < adapter->num_rx_queues; i++) {
4505 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4506 
4507 			if (!netif_is_ixgbe(ring->netdev))
4508 				continue;
4509 
4510 			j = ring->reg_idx;
4511 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4512 			vlnctrl &= ~IXGBE_RXDCTL_VME;
4513 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4514 		}
4515 		break;
4516 	default:
4517 		break;
4518 	}
4519 }
4520 
4521 /**
4522  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
4523  * @adapter: driver data
4524  */
4525 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
4526 {
4527 	struct ixgbe_hw *hw = &adapter->hw;
4528 	u32 vlnctrl;
4529 	int i, j;
4530 
4531 	switch (hw->mac.type) {
4532 	case ixgbe_mac_82598EB:
4533 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4534 		vlnctrl |= IXGBE_VLNCTRL_VME;
4535 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4536 		break;
4537 	case ixgbe_mac_82599EB:
4538 	case ixgbe_mac_X540:
4539 	case ixgbe_mac_X550:
4540 	case ixgbe_mac_X550EM_x:
4541 	case ixgbe_mac_x550em_a:
4542 		for (i = 0; i < adapter->num_rx_queues; i++) {
4543 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4544 
4545 			if (!netif_is_ixgbe(ring->netdev))
4546 				continue;
4547 
4548 			j = ring->reg_idx;
4549 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4550 			vlnctrl |= IXGBE_RXDCTL_VME;
4551 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4552 		}
4553 		break;
4554 	default:
4555 		break;
4556 	}
4557 }
4558 
4559 static void ixgbe_vlan_promisc_enable(struct ixgbe_adapter *adapter)
4560 {
4561 	struct ixgbe_hw *hw = &adapter->hw;
4562 	u32 vlnctrl, i;
4563 
4564 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4565 
4566 	if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) {
4567 	/* For VMDq and SR-IOV we must leave VLAN filtering enabled */
4568 		vlnctrl |= IXGBE_VLNCTRL_VFE;
4569 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4570 	} else {
4571 		vlnctrl &= ~IXGBE_VLNCTRL_VFE;
4572 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4573 		return;
4574 	}
4575 
4576 	/* Nothing to do for 82598 */
4577 	if (hw->mac.type == ixgbe_mac_82598EB)
4578 		return;
4579 
4580 	/* We are already in VLAN promisc, nothing to do */
4581 	if (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC)
4582 		return;
4583 
4584 	/* Set flag so we don't redo unnecessary work */
4585 	adapter->flags2 |= IXGBE_FLAG2_VLAN_PROMISC;
4586 
4587 	/* Add PF to all active pools */
4588 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4589 		u32 reg_offset = IXGBE_VLVFB(i * 2 + VMDQ_P(0) / 32);
4590 		u32 vlvfb = IXGBE_READ_REG(hw, reg_offset);
4591 
4592 		vlvfb |= BIT(VMDQ_P(0) % 32);
4593 		IXGBE_WRITE_REG(hw, reg_offset, vlvfb);
4594 	}
4595 
4596 	/* Set all bits in the VLAN filter table array */
4597 	for (i = hw->mac.vft_size; i--;)
4598 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), ~0U);
4599 }
4600 
4601 #define VFTA_BLOCK_SIZE 8
4602 static void ixgbe_scrub_vfta(struct ixgbe_adapter *adapter, u32 vfta_offset)
4603 {
4604 	struct ixgbe_hw *hw = &adapter->hw;
4605 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4606 	u32 vid_start = vfta_offset * 32;
4607 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4608 	u32 i, vid, word, bits;
4609 
4610 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
4611 		u32 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
4612 
4613 		/* pull VLAN ID from VLVF */
4614 		vid = vlvf & VLAN_VID_MASK;
4615 
4616 		/* only concern outselves with a certain range */
4617 		if (vid < vid_start || vid >= vid_end)
4618 			continue;
4619 
4620 		if (vlvf) {
4621 			/* record VLAN ID in VFTA */
4622 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4623 
4624 			/* if PF is part of this then continue */
4625 			if (test_bit(vid, adapter->active_vlans))
4626 				continue;
4627 		}
4628 
4629 		/* remove PF from the pool */
4630 		word = i * 2 + VMDQ_P(0) / 32;
4631 		bits = ~BIT(VMDQ_P(0) % 32);
4632 		bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4633 		IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), bits);
4634 	}
4635 
4636 	/* extract values from active_vlans and write back to VFTA */
4637 	for (i = VFTA_BLOCK_SIZE; i--;) {
4638 		vid = (vfta_offset + i) * 32;
4639 		word = vid / BITS_PER_LONG;
4640 		bits = vid % BITS_PER_LONG;
4641 
4642 		vfta[i] |= adapter->active_vlans[word] >> bits;
4643 
4644 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(vfta_offset + i), vfta[i]);
4645 	}
4646 }
4647 
4648 static void ixgbe_vlan_promisc_disable(struct ixgbe_adapter *adapter)
4649 {
4650 	struct ixgbe_hw *hw = &adapter->hw;
4651 	u32 vlnctrl, i;
4652 
4653 	/* Set VLAN filtering to enabled */
4654 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4655 	vlnctrl |= IXGBE_VLNCTRL_VFE;
4656 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4657 
4658 	if (!(adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) ||
4659 	    hw->mac.type == ixgbe_mac_82598EB)
4660 		return;
4661 
4662 	/* We are not in VLAN promisc, nothing to do */
4663 	if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4664 		return;
4665 
4666 	/* Set flag so we don't redo unnecessary work */
4667 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4668 
4669 	for (i = 0; i < hw->mac.vft_size; i += VFTA_BLOCK_SIZE)
4670 		ixgbe_scrub_vfta(adapter, i);
4671 }
4672 
4673 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
4674 {
4675 	u16 vid = 1;
4676 
4677 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
4678 
4679 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
4680 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
4681 }
4682 
4683 /**
4684  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4685  * @netdev: network interface device structure
4686  *
4687  * Writes multicast address list to the MTA hash table.
4688  * Returns: -ENOMEM on failure
4689  *                0 on no addresses written
4690  *                X on writing X addresses to MTA
4691  **/
4692 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
4693 {
4694 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4695 	struct ixgbe_hw *hw = &adapter->hw;
4696 
4697 	if (!netif_running(netdev))
4698 		return 0;
4699 
4700 	if (hw->mac.ops.update_mc_addr_list)
4701 		hw->mac.ops.update_mc_addr_list(hw, netdev);
4702 	else
4703 		return -ENOMEM;
4704 
4705 #ifdef CONFIG_PCI_IOV
4706 	ixgbe_restore_vf_multicasts(adapter);
4707 #endif
4708 
4709 	return netdev_mc_count(netdev);
4710 }
4711 
4712 #ifdef CONFIG_PCI_IOV
4713 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4714 {
4715 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4716 	struct ixgbe_hw *hw = &adapter->hw;
4717 	int i;
4718 
4719 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4720 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4721 
4722 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4723 			hw->mac.ops.set_rar(hw, i,
4724 					    mac_table->addr,
4725 					    mac_table->pool,
4726 					    IXGBE_RAH_AV);
4727 		else
4728 			hw->mac.ops.clear_rar(hw, i);
4729 	}
4730 }
4731 
4732 #endif
4733 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4734 {
4735 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4736 	struct ixgbe_hw *hw = &adapter->hw;
4737 	int i;
4738 
4739 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4740 		if (!(mac_table->state & IXGBE_MAC_STATE_MODIFIED))
4741 			continue;
4742 
4743 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4744 
4745 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4746 			hw->mac.ops.set_rar(hw, i,
4747 					    mac_table->addr,
4748 					    mac_table->pool,
4749 					    IXGBE_RAH_AV);
4750 		else
4751 			hw->mac.ops.clear_rar(hw, i);
4752 	}
4753 }
4754 
4755 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4756 {
4757 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4758 	struct ixgbe_hw *hw = &adapter->hw;
4759 	int i;
4760 
4761 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4762 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4763 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4764 	}
4765 
4766 	ixgbe_sync_mac_table(adapter);
4767 }
4768 
4769 static int ixgbe_available_rars(struct ixgbe_adapter *adapter, u16 pool)
4770 {
4771 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4772 	struct ixgbe_hw *hw = &adapter->hw;
4773 	int i, count = 0;
4774 
4775 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4776 		/* do not count default RAR as available */
4777 		if (mac_table->state & IXGBE_MAC_STATE_DEFAULT)
4778 			continue;
4779 
4780 		/* only count unused and addresses that belong to us */
4781 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE) {
4782 			if (mac_table->pool != pool)
4783 				continue;
4784 		}
4785 
4786 		count++;
4787 	}
4788 
4789 	return count;
4790 }
4791 
4792 /* this function destroys the first RAR entry */
4793 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter)
4794 {
4795 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4796 	struct ixgbe_hw *hw = &adapter->hw;
4797 
4798 	memcpy(&mac_table->addr, hw->mac.addr, ETH_ALEN);
4799 	mac_table->pool = VMDQ_P(0);
4800 
4801 	mac_table->state = IXGBE_MAC_STATE_DEFAULT | IXGBE_MAC_STATE_IN_USE;
4802 
4803 	hw->mac.ops.set_rar(hw, 0, mac_table->addr, mac_table->pool,
4804 			    IXGBE_RAH_AV);
4805 }
4806 
4807 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter,
4808 			 const u8 *addr, u16 pool)
4809 {
4810 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4811 	struct ixgbe_hw *hw = &adapter->hw;
4812 	int i;
4813 
4814 	if (is_zero_ether_addr(addr))
4815 		return -EINVAL;
4816 
4817 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4818 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4819 			continue;
4820 
4821 		ether_addr_copy(mac_table->addr, addr);
4822 		mac_table->pool = pool;
4823 
4824 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED |
4825 				    IXGBE_MAC_STATE_IN_USE;
4826 
4827 		ixgbe_sync_mac_table(adapter);
4828 
4829 		return i;
4830 	}
4831 
4832 	return -ENOMEM;
4833 }
4834 
4835 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter,
4836 			 const u8 *addr, u16 pool)
4837 {
4838 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4839 	struct ixgbe_hw *hw = &adapter->hw;
4840 	int i;
4841 
4842 	if (is_zero_ether_addr(addr))
4843 		return -EINVAL;
4844 
4845 	/* search table for addr, if found clear IN_USE flag and sync */
4846 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4847 		/* we can only delete an entry if it is in use */
4848 		if (!(mac_table->state & IXGBE_MAC_STATE_IN_USE))
4849 			continue;
4850 		/* we only care about entries that belong to the given pool */
4851 		if (mac_table->pool != pool)
4852 			continue;
4853 		/* we only care about a specific MAC address */
4854 		if (!ether_addr_equal(addr, mac_table->addr))
4855 			continue;
4856 
4857 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4858 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4859 
4860 		ixgbe_sync_mac_table(adapter);
4861 
4862 		return 0;
4863 	}
4864 
4865 	return -ENOMEM;
4866 }
4867 
4868 static int ixgbe_uc_sync(struct net_device *netdev, const unsigned char *addr)
4869 {
4870 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4871 	int ret;
4872 
4873 	ret = ixgbe_add_mac_filter(adapter, addr, VMDQ_P(0));
4874 
4875 	return min_t(int, ret, 0);
4876 }
4877 
4878 static int ixgbe_uc_unsync(struct net_device *netdev, const unsigned char *addr)
4879 {
4880 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4881 
4882 	ixgbe_del_mac_filter(adapter, addr, VMDQ_P(0));
4883 
4884 	return 0;
4885 }
4886 
4887 /**
4888  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4889  * @netdev: network interface device structure
4890  *
4891  * The set_rx_method entry point is called whenever the unicast/multicast
4892  * address list or the network interface flags are updated.  This routine is
4893  * responsible for configuring the hardware for proper unicast, multicast and
4894  * promiscuous mode.
4895  **/
4896 void ixgbe_set_rx_mode(struct net_device *netdev)
4897 {
4898 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
4899 	struct ixgbe_hw *hw = &adapter->hw;
4900 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4901 	netdev_features_t features = netdev->features;
4902 	int count;
4903 
4904 	/* Check for Promiscuous and All Multicast modes */
4905 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4906 
4907 	/* set all bits that we expect to always be set */
4908 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4909 	fctrl |= IXGBE_FCTRL_BAM;
4910 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4911 	fctrl |= IXGBE_FCTRL_PMCF;
4912 
4913 	/* clear the bits we are changing the status of */
4914 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4915 	if (netdev->flags & IFF_PROMISC) {
4916 		hw->addr_ctrl.user_set_promisc = true;
4917 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4918 		vmolr |= IXGBE_VMOLR_MPE;
4919 		features &= ~NETIF_F_HW_VLAN_CTAG_FILTER;
4920 	} else {
4921 		if (netdev->flags & IFF_ALLMULTI) {
4922 			fctrl |= IXGBE_FCTRL_MPE;
4923 			vmolr |= IXGBE_VMOLR_MPE;
4924 		}
4925 		hw->addr_ctrl.user_set_promisc = false;
4926 	}
4927 
4928 	/*
4929 	 * Write addresses to available RAR registers, if there is not
4930 	 * sufficient space to store all the addresses then enable
4931 	 * unicast promiscuous mode
4932 	 */
4933 	if (__dev_uc_sync(netdev, ixgbe_uc_sync, ixgbe_uc_unsync)) {
4934 		fctrl |= IXGBE_FCTRL_UPE;
4935 		vmolr |= IXGBE_VMOLR_ROPE;
4936 	}
4937 
4938 	/* Write addresses to the MTA, if the attempt fails
4939 	 * then we should just turn on promiscuous mode so
4940 	 * that we can at least receive multicast traffic
4941 	 */
4942 	count = ixgbe_write_mc_addr_list(netdev);
4943 	if (count < 0) {
4944 		fctrl |= IXGBE_FCTRL_MPE;
4945 		vmolr |= IXGBE_VMOLR_MPE;
4946 	} else if (count) {
4947 		vmolr |= IXGBE_VMOLR_ROMPE;
4948 	}
4949 
4950 	if (hw->mac.type != ixgbe_mac_82598EB) {
4951 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4952 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4953 			   IXGBE_VMOLR_ROPE);
4954 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4955 	}
4956 
4957 	/* This is useful for sniffing bad packets. */
4958 	if (features & NETIF_F_RXALL) {
4959 		/* UPE and MPE will be handled by normal PROMISC logic
4960 		 * in e1000e_set_rx_mode */
4961 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4962 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4963 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4964 
4965 		fctrl &= ~(IXGBE_FCTRL_DPF);
4966 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
4967 	}
4968 
4969 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4970 
4971 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
4972 		ixgbe_vlan_strip_enable(adapter);
4973 	else
4974 		ixgbe_vlan_strip_disable(adapter);
4975 
4976 	if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
4977 		ixgbe_vlan_promisc_disable(adapter);
4978 	else
4979 		ixgbe_vlan_promisc_enable(adapter);
4980 }
4981 
4982 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4983 {
4984 	int q_idx;
4985 
4986 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
4987 		napi_enable(&adapter->q_vector[q_idx]->napi);
4988 }
4989 
4990 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
4991 {
4992 	int q_idx;
4993 
4994 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
4995 		napi_disable(&adapter->q_vector[q_idx]->napi);
4996 }
4997 
4998 static int ixgbe_udp_tunnel_sync(struct net_device *dev, unsigned int table)
4999 {
5000 	struct ixgbe_adapter *adapter = netdev_priv(dev);
5001 	struct ixgbe_hw *hw = &adapter->hw;
5002 	struct udp_tunnel_info ti;
5003 
5004 	udp_tunnel_nic_get_port(dev, table, 0, &ti);
5005 	if (ti.type == UDP_TUNNEL_TYPE_VXLAN)
5006 		adapter->vxlan_port = ti.port;
5007 	else
5008 		adapter->geneve_port = ti.port;
5009 
5010 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL,
5011 			ntohs(adapter->vxlan_port) |
5012 			ntohs(adapter->geneve_port) <<
5013 				IXGBE_VXLANCTRL_GENEVE_UDPPORT_SHIFT);
5014 	return 0;
5015 }
5016 
5017 static const struct udp_tunnel_nic_info ixgbe_udp_tunnels_x550 = {
5018 	.sync_table	= ixgbe_udp_tunnel_sync,
5019 	.flags		= UDP_TUNNEL_NIC_INFO_IPV4_ONLY,
5020 	.tables		= {
5021 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN,  },
5022 	},
5023 };
5024 
5025 static const struct udp_tunnel_nic_info ixgbe_udp_tunnels_x550em_a = {
5026 	.sync_table	= ixgbe_udp_tunnel_sync,
5027 	.flags		= UDP_TUNNEL_NIC_INFO_IPV4_ONLY,
5028 	.tables		= {
5029 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN,  },
5030 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_GENEVE, },
5031 	},
5032 };
5033 
5034 #ifdef CONFIG_IXGBE_DCB
5035 /**
5036  * ixgbe_configure_dcb - Configure DCB hardware
5037  * @adapter: ixgbe adapter struct
5038  *
5039  * This is called by the driver on open to configure the DCB hardware.
5040  * This is also called by the gennetlink interface when reconfiguring
5041  * the DCB state.
5042  */
5043 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
5044 {
5045 	struct ixgbe_hw *hw = &adapter->hw;
5046 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
5047 
5048 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
5049 		if (hw->mac.type == ixgbe_mac_82598EB)
5050 			netif_set_gso_max_size(adapter->netdev, 65536);
5051 		return;
5052 	}
5053 
5054 	if (hw->mac.type == ixgbe_mac_82598EB)
5055 		netif_set_gso_max_size(adapter->netdev, 32768);
5056 
5057 #ifdef IXGBE_FCOE
5058 	if (adapter->netdev->features & NETIF_F_FCOE_MTU)
5059 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
5060 #endif
5061 
5062 	/* reconfigure the hardware */
5063 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
5064 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5065 						DCB_TX_CONFIG);
5066 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5067 						DCB_RX_CONFIG);
5068 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
5069 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
5070 		ixgbe_dcb_hw_ets(&adapter->hw,
5071 				 adapter->ixgbe_ieee_ets,
5072 				 max_frame);
5073 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
5074 					adapter->ixgbe_ieee_pfc->pfc_en,
5075 					adapter->ixgbe_ieee_ets->prio_tc);
5076 	}
5077 
5078 	/* Enable RSS Hash per TC */
5079 	if (hw->mac.type != ixgbe_mac_82598EB) {
5080 		u32 msb = 0;
5081 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
5082 
5083 		while (rss_i) {
5084 			msb++;
5085 			rss_i >>= 1;
5086 		}
5087 
5088 		/* write msb to all 8 TCs in one write */
5089 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
5090 	}
5091 }
5092 #endif
5093 
5094 /* Additional bittime to account for IXGBE framing */
5095 #define IXGBE_ETH_FRAMING 20
5096 
5097 /**
5098  * ixgbe_hpbthresh - calculate high water mark for flow control
5099  *
5100  * @adapter: board private structure to calculate for
5101  * @pb: packet buffer to calculate
5102  */
5103 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
5104 {
5105 	struct ixgbe_hw *hw = &adapter->hw;
5106 	struct net_device *dev = adapter->netdev;
5107 	int link, tc, kb, marker;
5108 	u32 dv_id, rx_pba;
5109 
5110 	/* Calculate max LAN frame size */
5111 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
5112 
5113 #ifdef IXGBE_FCOE
5114 	/* FCoE traffic class uses FCOE jumbo frames */
5115 	if ((dev->features & NETIF_F_FCOE_MTU) &&
5116 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5117 	    (pb == ixgbe_fcoe_get_tc(adapter)))
5118 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5119 #endif
5120 
5121 	/* Calculate delay value for device */
5122 	switch (hw->mac.type) {
5123 	case ixgbe_mac_X540:
5124 	case ixgbe_mac_X550:
5125 	case ixgbe_mac_X550EM_x:
5126 	case ixgbe_mac_x550em_a:
5127 		dv_id = IXGBE_DV_X540(link, tc);
5128 		break;
5129 	default:
5130 		dv_id = IXGBE_DV(link, tc);
5131 		break;
5132 	}
5133 
5134 	/* Loopback switch introduces additional latency */
5135 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5136 		dv_id += IXGBE_B2BT(tc);
5137 
5138 	/* Delay value is calculated in bit times convert to KB */
5139 	kb = IXGBE_BT2KB(dv_id);
5140 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
5141 
5142 	marker = rx_pba - kb;
5143 
5144 	/* It is possible that the packet buffer is not large enough
5145 	 * to provide required headroom. In this case throw an error
5146 	 * to user and a do the best we can.
5147 	 */
5148 	if (marker < 0) {
5149 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
5150 			    "headroom to support flow control."
5151 			    "Decrease MTU or number of traffic classes\n", pb);
5152 		marker = tc + 1;
5153 	}
5154 
5155 	return marker;
5156 }
5157 
5158 /**
5159  * ixgbe_lpbthresh - calculate low water mark for for flow control
5160  *
5161  * @adapter: board private structure to calculate for
5162  * @pb: packet buffer to calculate
5163  */
5164 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
5165 {
5166 	struct ixgbe_hw *hw = &adapter->hw;
5167 	struct net_device *dev = adapter->netdev;
5168 	int tc;
5169 	u32 dv_id;
5170 
5171 	/* Calculate max LAN frame size */
5172 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
5173 
5174 #ifdef IXGBE_FCOE
5175 	/* FCoE traffic class uses FCOE jumbo frames */
5176 	if ((dev->features & NETIF_F_FCOE_MTU) &&
5177 	    (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5178 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
5179 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5180 #endif
5181 
5182 	/* Calculate delay value for device */
5183 	switch (hw->mac.type) {
5184 	case ixgbe_mac_X540:
5185 	case ixgbe_mac_X550:
5186 	case ixgbe_mac_X550EM_x:
5187 	case ixgbe_mac_x550em_a:
5188 		dv_id = IXGBE_LOW_DV_X540(tc);
5189 		break;
5190 	default:
5191 		dv_id = IXGBE_LOW_DV(tc);
5192 		break;
5193 	}
5194 
5195 	/* Delay value is calculated in bit times convert to KB */
5196 	return IXGBE_BT2KB(dv_id);
5197 }
5198 
5199 /*
5200  * ixgbe_pbthresh_setup - calculate and setup high low water marks
5201  */
5202 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
5203 {
5204 	struct ixgbe_hw *hw = &adapter->hw;
5205 	int num_tc = adapter->hw_tcs;
5206 	int i;
5207 
5208 	if (!num_tc)
5209 		num_tc = 1;
5210 
5211 	for (i = 0; i < num_tc; i++) {
5212 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
5213 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
5214 
5215 		/* Low water marks must not be larger than high water marks */
5216 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
5217 			hw->fc.low_water[i] = 0;
5218 	}
5219 
5220 	for (; i < MAX_TRAFFIC_CLASS; i++)
5221 		hw->fc.high_water[i] = 0;
5222 }
5223 
5224 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
5225 {
5226 	struct ixgbe_hw *hw = &adapter->hw;
5227 	int hdrm;
5228 	u8 tc = adapter->hw_tcs;
5229 
5230 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
5231 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
5232 		hdrm = 32 << adapter->fdir_pballoc;
5233 	else
5234 		hdrm = 0;
5235 
5236 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
5237 	ixgbe_pbthresh_setup(adapter);
5238 }
5239 
5240 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
5241 {
5242 	struct ixgbe_hw *hw = &adapter->hw;
5243 	struct hlist_node *node2;
5244 	struct ixgbe_fdir_filter *filter;
5245 	u8 queue;
5246 
5247 	spin_lock(&adapter->fdir_perfect_lock);
5248 
5249 	if (!hlist_empty(&adapter->fdir_filter_list))
5250 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
5251 
5252 	hlist_for_each_entry_safe(filter, node2,
5253 				  &adapter->fdir_filter_list, fdir_node) {
5254 		if (filter->action == IXGBE_FDIR_DROP_QUEUE) {
5255 			queue = IXGBE_FDIR_DROP_QUEUE;
5256 		} else {
5257 			u32 ring = ethtool_get_flow_spec_ring(filter->action);
5258 			u8 vf = ethtool_get_flow_spec_ring_vf(filter->action);
5259 
5260 			if (!vf && (ring >= adapter->num_rx_queues)) {
5261 				e_err(drv, "FDIR restore failed without VF, ring: %u\n",
5262 				      ring);
5263 				continue;
5264 			} else if (vf &&
5265 				   ((vf > adapter->num_vfs) ||
5266 				     ring >= adapter->num_rx_queues_per_pool)) {
5267 				e_err(drv, "FDIR restore failed with VF, vf: %hhu, ring: %u\n",
5268 				      vf, ring);
5269 				continue;
5270 			}
5271 
5272 			/* Map the ring onto the absolute queue index */
5273 			if (!vf)
5274 				queue = adapter->rx_ring[ring]->reg_idx;
5275 			else
5276 				queue = ((vf - 1) *
5277 					adapter->num_rx_queues_per_pool) + ring;
5278 		}
5279 
5280 		ixgbe_fdir_write_perfect_filter_82599(hw,
5281 				&filter->filter, filter->sw_idx, queue);
5282 	}
5283 
5284 	spin_unlock(&adapter->fdir_perfect_lock);
5285 }
5286 
5287 /**
5288  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
5289  * @rx_ring: ring to free buffers from
5290  **/
5291 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
5292 {
5293 	u16 i = rx_ring->next_to_clean;
5294 	struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
5295 
5296 	if (rx_ring->xsk_pool) {
5297 		ixgbe_xsk_clean_rx_ring(rx_ring);
5298 		goto skip_free;
5299 	}
5300 
5301 	/* Free all the Rx ring sk_buffs */
5302 	while (i != rx_ring->next_to_alloc) {
5303 		if (rx_buffer->skb) {
5304 			struct sk_buff *skb = rx_buffer->skb;
5305 			if (IXGBE_CB(skb)->page_released)
5306 				dma_unmap_page_attrs(rx_ring->dev,
5307 						     IXGBE_CB(skb)->dma,
5308 						     ixgbe_rx_pg_size(rx_ring),
5309 						     DMA_FROM_DEVICE,
5310 						     IXGBE_RX_DMA_ATTR);
5311 			dev_kfree_skb(skb);
5312 		}
5313 
5314 		/* Invalidate cache lines that may have been written to by
5315 		 * device so that we avoid corrupting memory.
5316 		 */
5317 		dma_sync_single_range_for_cpu(rx_ring->dev,
5318 					      rx_buffer->dma,
5319 					      rx_buffer->page_offset,
5320 					      ixgbe_rx_bufsz(rx_ring),
5321 					      DMA_FROM_DEVICE);
5322 
5323 		/* free resources associated with mapping */
5324 		dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
5325 				     ixgbe_rx_pg_size(rx_ring),
5326 				     DMA_FROM_DEVICE,
5327 				     IXGBE_RX_DMA_ATTR);
5328 		__page_frag_cache_drain(rx_buffer->page,
5329 					rx_buffer->pagecnt_bias);
5330 
5331 		i++;
5332 		rx_buffer++;
5333 		if (i == rx_ring->count) {
5334 			i = 0;
5335 			rx_buffer = rx_ring->rx_buffer_info;
5336 		}
5337 	}
5338 
5339 skip_free:
5340 	rx_ring->next_to_alloc = 0;
5341 	rx_ring->next_to_clean = 0;
5342 	rx_ring->next_to_use = 0;
5343 }
5344 
5345 static int ixgbe_fwd_ring_up(struct ixgbe_adapter *adapter,
5346 			     struct ixgbe_fwd_adapter *accel)
5347 {
5348 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
5349 	int num_tc = netdev_get_num_tc(adapter->netdev);
5350 	struct net_device *vdev = accel->netdev;
5351 	int i, baseq, err;
5352 
5353 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
5354 	netdev_dbg(vdev, "pool %i:%i queues %i:%i\n",
5355 		   accel->pool, adapter->num_rx_pools,
5356 		   baseq, baseq + adapter->num_rx_queues_per_pool);
5357 
5358 	accel->rx_base_queue = baseq;
5359 	accel->tx_base_queue = baseq;
5360 
5361 	/* record configuration for macvlan interface in vdev */
5362 	for (i = 0; i < num_tc; i++)
5363 		netdev_bind_sb_channel_queue(adapter->netdev, vdev,
5364 					     i, rss_i, baseq + (rss_i * i));
5365 
5366 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5367 		adapter->rx_ring[baseq + i]->netdev = vdev;
5368 
5369 	/* Guarantee all rings are updated before we update the
5370 	 * MAC address filter.
5371 	 */
5372 	wmb();
5373 
5374 	/* ixgbe_add_mac_filter will return an index if it succeeds, so we
5375 	 * need to only treat it as an error value if it is negative.
5376 	 */
5377 	err = ixgbe_add_mac_filter(adapter, vdev->dev_addr,
5378 				   VMDQ_P(accel->pool));
5379 	if (err >= 0)
5380 		return 0;
5381 
5382 	/* if we cannot add the MAC rule then disable the offload */
5383 	macvlan_release_l2fw_offload(vdev);
5384 
5385 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5386 		adapter->rx_ring[baseq + i]->netdev = NULL;
5387 
5388 	netdev_err(vdev, "L2FW offload disabled due to L2 filter error\n");
5389 
5390 	/* unbind the queues and drop the subordinate channel config */
5391 	netdev_unbind_sb_channel(adapter->netdev, vdev);
5392 	netdev_set_sb_channel(vdev, 0);
5393 
5394 	clear_bit(accel->pool, adapter->fwd_bitmask);
5395 	kfree(accel);
5396 
5397 	return err;
5398 }
5399 
5400 static int ixgbe_macvlan_up(struct net_device *vdev,
5401 			    struct netdev_nested_priv *priv)
5402 {
5403 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)priv->data;
5404 	struct ixgbe_fwd_adapter *accel;
5405 
5406 	if (!netif_is_macvlan(vdev))
5407 		return 0;
5408 
5409 	accel = macvlan_accel_priv(vdev);
5410 	if (!accel)
5411 		return 0;
5412 
5413 	ixgbe_fwd_ring_up(adapter, accel);
5414 
5415 	return 0;
5416 }
5417 
5418 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
5419 {
5420 	struct netdev_nested_priv priv = {
5421 		.data = (void *)adapter,
5422 	};
5423 
5424 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
5425 				      ixgbe_macvlan_up, &priv);
5426 }
5427 
5428 static void ixgbe_configure(struct ixgbe_adapter *adapter)
5429 {
5430 	struct ixgbe_hw *hw = &adapter->hw;
5431 
5432 	ixgbe_configure_pb(adapter);
5433 #ifdef CONFIG_IXGBE_DCB
5434 	ixgbe_configure_dcb(adapter);
5435 #endif
5436 	/*
5437 	 * We must restore virtualization before VLANs or else
5438 	 * the VLVF registers will not be populated
5439 	 */
5440 	ixgbe_configure_virtualization(adapter);
5441 
5442 	ixgbe_set_rx_mode(adapter->netdev);
5443 	ixgbe_restore_vlan(adapter);
5444 	ixgbe_ipsec_restore(adapter);
5445 
5446 	switch (hw->mac.type) {
5447 	case ixgbe_mac_82599EB:
5448 	case ixgbe_mac_X540:
5449 		hw->mac.ops.disable_rx_buff(hw);
5450 		break;
5451 	default:
5452 		break;
5453 	}
5454 
5455 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
5456 		ixgbe_init_fdir_signature_82599(&adapter->hw,
5457 						adapter->fdir_pballoc);
5458 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
5459 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
5460 					      adapter->fdir_pballoc);
5461 		ixgbe_fdir_filter_restore(adapter);
5462 	}
5463 
5464 	switch (hw->mac.type) {
5465 	case ixgbe_mac_82599EB:
5466 	case ixgbe_mac_X540:
5467 		hw->mac.ops.enable_rx_buff(hw);
5468 		break;
5469 	default:
5470 		break;
5471 	}
5472 
5473 #ifdef CONFIG_IXGBE_DCA
5474 	/* configure DCA */
5475 	if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
5476 		ixgbe_setup_dca(adapter);
5477 #endif /* CONFIG_IXGBE_DCA */
5478 
5479 #ifdef IXGBE_FCOE
5480 	/* configure FCoE L2 filters, redirection table, and Rx control */
5481 	ixgbe_configure_fcoe(adapter);
5482 
5483 #endif /* IXGBE_FCOE */
5484 	ixgbe_configure_tx(adapter);
5485 	ixgbe_configure_rx(adapter);
5486 	ixgbe_configure_dfwd(adapter);
5487 }
5488 
5489 /**
5490  * ixgbe_sfp_link_config - set up SFP+ link
5491  * @adapter: pointer to private adapter struct
5492  **/
5493 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
5494 {
5495 	/*
5496 	 * We are assuming the worst case scenario here, and that
5497 	 * is that an SFP was inserted/removed after the reset
5498 	 * but before SFP detection was enabled.  As such the best
5499 	 * solution is to just start searching as soon as we start
5500 	 */
5501 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5502 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5503 
5504 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5505 	adapter->sfp_poll_time = 0;
5506 }
5507 
5508 /**
5509  * ixgbe_non_sfp_link_config - set up non-SFP+ link
5510  * @hw: pointer to private hardware struct
5511  *
5512  * Returns 0 on success, negative on failure
5513  **/
5514 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
5515 {
5516 	u32 speed;
5517 	bool autoneg, link_up = false;
5518 	int ret = IXGBE_ERR_LINK_SETUP;
5519 
5520 	if (hw->mac.ops.check_link)
5521 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
5522 
5523 	if (ret)
5524 		return ret;
5525 
5526 	speed = hw->phy.autoneg_advertised;
5527 	if (!speed && hw->mac.ops.get_link_capabilities) {
5528 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
5529 							&autoneg);
5530 		speed &= ~(IXGBE_LINK_SPEED_5GB_FULL |
5531 			   IXGBE_LINK_SPEED_2_5GB_FULL);
5532 	}
5533 
5534 	if (ret)
5535 		return ret;
5536 
5537 	if (hw->mac.ops.setup_link)
5538 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
5539 
5540 	return ret;
5541 }
5542 
5543 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
5544 {
5545 	struct ixgbe_hw *hw = &adapter->hw;
5546 	u32 gpie = 0;
5547 
5548 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
5549 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5550 		       IXGBE_GPIE_OCD;
5551 		gpie |= IXGBE_GPIE_EIAME;
5552 		/*
5553 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
5554 		 * this saves a register write for every interrupt
5555 		 */
5556 		switch (hw->mac.type) {
5557 		case ixgbe_mac_82598EB:
5558 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5559 			break;
5560 		case ixgbe_mac_82599EB:
5561 		case ixgbe_mac_X540:
5562 		case ixgbe_mac_X550:
5563 		case ixgbe_mac_X550EM_x:
5564 		case ixgbe_mac_x550em_a:
5565 		default:
5566 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5567 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5568 			break;
5569 		}
5570 	} else {
5571 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
5572 		 * specifically only auto mask tx and rx interrupts */
5573 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5574 	}
5575 
5576 	/* XXX: to interrupt immediately for EICS writes, enable this */
5577 	/* gpie |= IXGBE_GPIE_EIMEN; */
5578 
5579 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
5580 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
5581 
5582 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
5583 		case IXGBE_82599_VMDQ_8Q_MASK:
5584 			gpie |= IXGBE_GPIE_VTMODE_16;
5585 			break;
5586 		case IXGBE_82599_VMDQ_4Q_MASK:
5587 			gpie |= IXGBE_GPIE_VTMODE_32;
5588 			break;
5589 		default:
5590 			gpie |= IXGBE_GPIE_VTMODE_64;
5591 			break;
5592 		}
5593 	}
5594 
5595 	/* Enable Thermal over heat sensor interrupt */
5596 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
5597 		switch (adapter->hw.mac.type) {
5598 		case ixgbe_mac_82599EB:
5599 			gpie |= IXGBE_SDP0_GPIEN_8259X;
5600 			break;
5601 		default:
5602 			break;
5603 		}
5604 	}
5605 
5606 	/* Enable fan failure interrupt */
5607 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
5608 		gpie |= IXGBE_SDP1_GPIEN(hw);
5609 
5610 	switch (hw->mac.type) {
5611 	case ixgbe_mac_82599EB:
5612 		gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
5613 		break;
5614 	case ixgbe_mac_X550EM_x:
5615 	case ixgbe_mac_x550em_a:
5616 		gpie |= IXGBE_SDP0_GPIEN_X540;
5617 		break;
5618 	default:
5619 		break;
5620 	}
5621 
5622 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5623 }
5624 
5625 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
5626 {
5627 	struct ixgbe_hw *hw = &adapter->hw;
5628 	int err;
5629 	u32 ctrl_ext;
5630 
5631 	ixgbe_get_hw_control(adapter);
5632 	ixgbe_setup_gpie(adapter);
5633 
5634 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
5635 		ixgbe_configure_msix(adapter);
5636 	else
5637 		ixgbe_configure_msi_and_legacy(adapter);
5638 
5639 	/* enable the optics for 82599 SFP+ fiber */
5640 	if (hw->mac.ops.enable_tx_laser)
5641 		hw->mac.ops.enable_tx_laser(hw);
5642 
5643 	if (hw->phy.ops.set_phy_power)
5644 		hw->phy.ops.set_phy_power(hw, true);
5645 
5646 	smp_mb__before_atomic();
5647 	clear_bit(__IXGBE_DOWN, &adapter->state);
5648 	ixgbe_napi_enable_all(adapter);
5649 
5650 	if (ixgbe_is_sfp(hw)) {
5651 		ixgbe_sfp_link_config(adapter);
5652 	} else {
5653 		err = ixgbe_non_sfp_link_config(hw);
5654 		if (err)
5655 			e_err(probe, "link_config FAILED %d\n", err);
5656 	}
5657 
5658 	/* clear any pending interrupts, may auto mask */
5659 	IXGBE_READ_REG(hw, IXGBE_EICR);
5660 	ixgbe_irq_enable(adapter, true, true);
5661 
5662 	/*
5663 	 * If this adapter has a fan, check to see if we had a failure
5664 	 * before we enabled the interrupt.
5665 	 */
5666 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
5667 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
5668 		if (esdp & IXGBE_ESDP_SDP1)
5669 			e_crit(drv, "Fan has stopped, replace the adapter\n");
5670 	}
5671 
5672 	/* bring the link up in the watchdog, this could race with our first
5673 	 * link up interrupt but shouldn't be a problem */
5674 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5675 	adapter->link_check_timeout = jiffies;
5676 	mod_timer(&adapter->service_timer, jiffies);
5677 
5678 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
5679 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
5680 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
5681 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
5682 }
5683 
5684 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
5685 {
5686 	/* put off any impending NetWatchDogTimeout */
5687 	netif_trans_update(adapter->netdev);
5688 
5689 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
5690 		usleep_range(1000, 2000);
5691 	if (adapter->hw.phy.type == ixgbe_phy_fw)
5692 		ixgbe_watchdog_link_is_down(adapter);
5693 	ixgbe_down(adapter);
5694 	/*
5695 	 * If SR-IOV enabled then wait a bit before bringing the adapter
5696 	 * back up to give the VFs time to respond to the reset.  The
5697 	 * two second wait is based upon the watchdog timer cycle in
5698 	 * the VF driver.
5699 	 */
5700 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5701 		msleep(2000);
5702 	ixgbe_up(adapter);
5703 	clear_bit(__IXGBE_RESETTING, &adapter->state);
5704 }
5705 
5706 void ixgbe_up(struct ixgbe_adapter *adapter)
5707 {
5708 	/* hardware has been reset, we need to reload some things */
5709 	ixgbe_configure(adapter);
5710 
5711 	ixgbe_up_complete(adapter);
5712 }
5713 
5714 static unsigned long ixgbe_get_completion_timeout(struct ixgbe_adapter *adapter)
5715 {
5716 	u16 devctl2;
5717 
5718 	pcie_capability_read_word(adapter->pdev, PCI_EXP_DEVCTL2, &devctl2);
5719 
5720 	switch (devctl2 & IXGBE_PCIDEVCTRL2_TIMEO_MASK) {
5721 	case IXGBE_PCIDEVCTRL2_17_34s:
5722 	case IXGBE_PCIDEVCTRL2_4_8s:
5723 		/* For now we cap the upper limit on delay to 2 seconds
5724 		 * as we end up going up to 34 seconds of delay in worst
5725 		 * case timeout value.
5726 		 */
5727 	case IXGBE_PCIDEVCTRL2_1_2s:
5728 		return 2000000ul;	/* 2.0 s */
5729 	case IXGBE_PCIDEVCTRL2_260_520ms:
5730 		return 520000ul;	/* 520 ms */
5731 	case IXGBE_PCIDEVCTRL2_65_130ms:
5732 		return 130000ul;	/* 130 ms */
5733 	case IXGBE_PCIDEVCTRL2_16_32ms:
5734 		return 32000ul;		/* 32 ms */
5735 	case IXGBE_PCIDEVCTRL2_1_2ms:
5736 		return 2000ul;		/* 2 ms */
5737 	case IXGBE_PCIDEVCTRL2_50_100us:
5738 		return 100ul;		/* 100 us */
5739 	case IXGBE_PCIDEVCTRL2_16_32ms_def:
5740 		return 32000ul;		/* 32 ms */
5741 	default:
5742 		break;
5743 	}
5744 
5745 	/* We shouldn't need to hit this path, but just in case default as
5746 	 * though completion timeout is not supported and support 32ms.
5747 	 */
5748 	return 32000ul;
5749 }
5750 
5751 void ixgbe_disable_rx(struct ixgbe_adapter *adapter)
5752 {
5753 	unsigned long wait_delay, delay_interval;
5754 	struct ixgbe_hw *hw = &adapter->hw;
5755 	int i, wait_loop;
5756 	u32 rxdctl;
5757 
5758 	/* disable receives */
5759 	hw->mac.ops.disable_rx(hw);
5760 
5761 	if (ixgbe_removed(hw->hw_addr))
5762 		return;
5763 
5764 	/* disable all enabled Rx queues */
5765 	for (i = 0; i < adapter->num_rx_queues; i++) {
5766 		struct ixgbe_ring *ring = adapter->rx_ring[i];
5767 		u8 reg_idx = ring->reg_idx;
5768 
5769 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
5770 		rxdctl &= ~IXGBE_RXDCTL_ENABLE;
5771 		rxdctl |= IXGBE_RXDCTL_SWFLSH;
5772 
5773 		/* write value back with RXDCTL.ENABLE bit cleared */
5774 		IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
5775 	}
5776 
5777 	/* RXDCTL.EN may not change on 82598 if link is down, so skip it */
5778 	if (hw->mac.type == ixgbe_mac_82598EB &&
5779 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
5780 		return;
5781 
5782 	/* Determine our minimum delay interval. We will increase this value
5783 	 * with each subsequent test. This way if the device returns quickly
5784 	 * we should spend as little time as possible waiting, however as
5785 	 * the time increases we will wait for larger periods of time.
5786 	 *
5787 	 * The trick here is that we increase the interval using the
5788 	 * following pattern: 1x 3x 5x 7x 9x 11x 13x 15x 17x 19x. The result
5789 	 * of that wait is that it totals up to 100x whatever interval we
5790 	 * choose. Since our minimum wait is 100us we can just divide the
5791 	 * total timeout by 100 to get our minimum delay interval.
5792 	 */
5793 	delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
5794 
5795 	wait_loop = IXGBE_MAX_RX_DESC_POLL;
5796 	wait_delay = delay_interval;
5797 
5798 	while (wait_loop--) {
5799 		usleep_range(wait_delay, wait_delay + 10);
5800 		wait_delay += delay_interval * 2;
5801 		rxdctl = 0;
5802 
5803 		/* OR together the reading of all the active RXDCTL registers,
5804 		 * and then test the result. We need the disable to complete
5805 		 * before we start freeing the memory and invalidating the
5806 		 * DMA mappings.
5807 		 */
5808 		for (i = 0; i < adapter->num_rx_queues; i++) {
5809 			struct ixgbe_ring *ring = adapter->rx_ring[i];
5810 			u8 reg_idx = ring->reg_idx;
5811 
5812 			rxdctl |= IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
5813 		}
5814 
5815 		if (!(rxdctl & IXGBE_RXDCTL_ENABLE))
5816 			return;
5817 	}
5818 
5819 	e_err(drv,
5820 	      "RXDCTL.ENABLE for one or more queues not cleared within the polling period\n");
5821 }
5822 
5823 void ixgbe_disable_tx(struct ixgbe_adapter *adapter)
5824 {
5825 	unsigned long wait_delay, delay_interval;
5826 	struct ixgbe_hw *hw = &adapter->hw;
5827 	int i, wait_loop;
5828 	u32 txdctl;
5829 
5830 	if (ixgbe_removed(hw->hw_addr))
5831 		return;
5832 
5833 	/* disable all enabled Tx queues */
5834 	for (i = 0; i < adapter->num_tx_queues; i++) {
5835 		struct ixgbe_ring *ring = adapter->tx_ring[i];
5836 		u8 reg_idx = ring->reg_idx;
5837 
5838 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5839 	}
5840 
5841 	/* disable all enabled XDP Tx queues */
5842 	for (i = 0; i < adapter->num_xdp_queues; i++) {
5843 		struct ixgbe_ring *ring = adapter->xdp_ring[i];
5844 		u8 reg_idx = ring->reg_idx;
5845 
5846 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5847 	}
5848 
5849 	/* If the link is not up there shouldn't be much in the way of
5850 	 * pending transactions. Those that are left will be flushed out
5851 	 * when the reset logic goes through the flush sequence to clean out
5852 	 * the pending Tx transactions.
5853 	 */
5854 	if (!(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
5855 		goto dma_engine_disable;
5856 
5857 	/* Determine our minimum delay interval. We will increase this value
5858 	 * with each subsequent test. This way if the device returns quickly
5859 	 * we should spend as little time as possible waiting, however as
5860 	 * the time increases we will wait for larger periods of time.
5861 	 *
5862 	 * The trick here is that we increase the interval using the
5863 	 * following pattern: 1x 3x 5x 7x 9x 11x 13x 15x 17x 19x. The result
5864 	 * of that wait is that it totals up to 100x whatever interval we
5865 	 * choose. Since our minimum wait is 100us we can just divide the
5866 	 * total timeout by 100 to get our minimum delay interval.
5867 	 */
5868 	delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
5869 
5870 	wait_loop = IXGBE_MAX_RX_DESC_POLL;
5871 	wait_delay = delay_interval;
5872 
5873 	while (wait_loop--) {
5874 		usleep_range(wait_delay, wait_delay + 10);
5875 		wait_delay += delay_interval * 2;
5876 		txdctl = 0;
5877 
5878 		/* OR together the reading of all the active TXDCTL registers,
5879 		 * and then test the result. We need the disable to complete
5880 		 * before we start freeing the memory and invalidating the
5881 		 * DMA mappings.
5882 		 */
5883 		for (i = 0; i < adapter->num_tx_queues; i++) {
5884 			struct ixgbe_ring *ring = adapter->tx_ring[i];
5885 			u8 reg_idx = ring->reg_idx;
5886 
5887 			txdctl |= IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
5888 		}
5889 		for (i = 0; i < adapter->num_xdp_queues; i++) {
5890 			struct ixgbe_ring *ring = adapter->xdp_ring[i];
5891 			u8 reg_idx = ring->reg_idx;
5892 
5893 			txdctl |= IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
5894 		}
5895 
5896 		if (!(txdctl & IXGBE_TXDCTL_ENABLE))
5897 			goto dma_engine_disable;
5898 	}
5899 
5900 	e_err(drv,
5901 	      "TXDCTL.ENABLE for one or more queues not cleared within the polling period\n");
5902 
5903 dma_engine_disable:
5904 	/* Disable the Tx DMA engine on 82599 and later MAC */
5905 	switch (hw->mac.type) {
5906 	case ixgbe_mac_82599EB:
5907 	case ixgbe_mac_X540:
5908 	case ixgbe_mac_X550:
5909 	case ixgbe_mac_X550EM_x:
5910 	case ixgbe_mac_x550em_a:
5911 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5912 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5913 				 ~IXGBE_DMATXCTL_TE));
5914 		fallthrough;
5915 	default:
5916 		break;
5917 	}
5918 }
5919 
5920 void ixgbe_reset(struct ixgbe_adapter *adapter)
5921 {
5922 	struct ixgbe_hw *hw = &adapter->hw;
5923 	struct net_device *netdev = adapter->netdev;
5924 	int err;
5925 
5926 	if (ixgbe_removed(hw->hw_addr))
5927 		return;
5928 	/* lock SFP init bit to prevent race conditions with the watchdog */
5929 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5930 		usleep_range(1000, 2000);
5931 
5932 	/* clear all SFP and link config related flags while holding SFP_INIT */
5933 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
5934 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
5935 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5936 
5937 	err = hw->mac.ops.init_hw(hw);
5938 	switch (err) {
5939 	case 0:
5940 	case IXGBE_ERR_SFP_NOT_PRESENT:
5941 	case IXGBE_ERR_SFP_NOT_SUPPORTED:
5942 		break;
5943 	case IXGBE_ERR_MASTER_REQUESTS_PENDING:
5944 		e_dev_err("master disable timed out\n");
5945 		break;
5946 	case IXGBE_ERR_EEPROM_VERSION:
5947 		/* We are running on a pre-production device, log a warning */
5948 		e_dev_warn("This device is a pre-production adapter/LOM. "
5949 			   "Please be aware there may be issues associated with "
5950 			   "your hardware.  If you are experiencing problems "
5951 			   "please contact your Intel or hardware "
5952 			   "representative who provided you with this "
5953 			   "hardware.\n");
5954 		break;
5955 	default:
5956 		e_dev_err("Hardware Error: %d\n", err);
5957 	}
5958 
5959 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5960 
5961 	/* flush entries out of MAC table */
5962 	ixgbe_flush_sw_mac_table(adapter);
5963 	__dev_uc_unsync(netdev, NULL);
5964 
5965 	/* do not flush user set addresses */
5966 	ixgbe_mac_set_default_filter(adapter);
5967 
5968 	/* update SAN MAC vmdq pool selection */
5969 	if (hw->mac.san_mac_rar_index)
5970 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
5971 
5972 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
5973 		ixgbe_ptp_reset(adapter);
5974 
5975 	if (hw->phy.ops.set_phy_power) {
5976 		if (!netif_running(adapter->netdev) && !adapter->wol)
5977 			hw->phy.ops.set_phy_power(hw, false);
5978 		else
5979 			hw->phy.ops.set_phy_power(hw, true);
5980 	}
5981 }
5982 
5983 /**
5984  * ixgbe_clean_tx_ring - Free Tx Buffers
5985  * @tx_ring: ring to be cleaned
5986  **/
5987 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
5988 {
5989 	u16 i = tx_ring->next_to_clean;
5990 	struct ixgbe_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
5991 
5992 	if (tx_ring->xsk_pool) {
5993 		ixgbe_xsk_clean_tx_ring(tx_ring);
5994 		goto out;
5995 	}
5996 
5997 	while (i != tx_ring->next_to_use) {
5998 		union ixgbe_adv_tx_desc *eop_desc, *tx_desc;
5999 
6000 		/* Free all the Tx ring sk_buffs */
6001 		if (ring_is_xdp(tx_ring))
6002 			xdp_return_frame(tx_buffer->xdpf);
6003 		else
6004 			dev_kfree_skb_any(tx_buffer->skb);
6005 
6006 		/* unmap skb header data */
6007 		dma_unmap_single(tx_ring->dev,
6008 				 dma_unmap_addr(tx_buffer, dma),
6009 				 dma_unmap_len(tx_buffer, len),
6010 				 DMA_TO_DEVICE);
6011 
6012 		/* check for eop_desc to determine the end of the packet */
6013 		eop_desc = tx_buffer->next_to_watch;
6014 		tx_desc = IXGBE_TX_DESC(tx_ring, i);
6015 
6016 		/* unmap remaining buffers */
6017 		while (tx_desc != eop_desc) {
6018 			tx_buffer++;
6019 			tx_desc++;
6020 			i++;
6021 			if (unlikely(i == tx_ring->count)) {
6022 				i = 0;
6023 				tx_buffer = tx_ring->tx_buffer_info;
6024 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6025 			}
6026 
6027 			/* unmap any remaining paged data */
6028 			if (dma_unmap_len(tx_buffer, len))
6029 				dma_unmap_page(tx_ring->dev,
6030 					       dma_unmap_addr(tx_buffer, dma),
6031 					       dma_unmap_len(tx_buffer, len),
6032 					       DMA_TO_DEVICE);
6033 		}
6034 
6035 		/* move us one more past the eop_desc for start of next pkt */
6036 		tx_buffer++;
6037 		i++;
6038 		if (unlikely(i == tx_ring->count)) {
6039 			i = 0;
6040 			tx_buffer = tx_ring->tx_buffer_info;
6041 		}
6042 	}
6043 
6044 	/* reset BQL for queue */
6045 	if (!ring_is_xdp(tx_ring))
6046 		netdev_tx_reset_queue(txring_txq(tx_ring));
6047 
6048 out:
6049 	/* reset next_to_use and next_to_clean */
6050 	tx_ring->next_to_use = 0;
6051 	tx_ring->next_to_clean = 0;
6052 }
6053 
6054 /**
6055  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
6056  * @adapter: board private structure
6057  **/
6058 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
6059 {
6060 	int i;
6061 
6062 	for (i = 0; i < adapter->num_rx_queues; i++)
6063 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
6064 }
6065 
6066 /**
6067  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
6068  * @adapter: board private structure
6069  **/
6070 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
6071 {
6072 	int i;
6073 
6074 	for (i = 0; i < adapter->num_tx_queues; i++)
6075 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
6076 	for (i = 0; i < adapter->num_xdp_queues; i++)
6077 		ixgbe_clean_tx_ring(adapter->xdp_ring[i]);
6078 }
6079 
6080 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
6081 {
6082 	struct hlist_node *node2;
6083 	struct ixgbe_fdir_filter *filter;
6084 
6085 	spin_lock(&adapter->fdir_perfect_lock);
6086 
6087 	hlist_for_each_entry_safe(filter, node2,
6088 				  &adapter->fdir_filter_list, fdir_node) {
6089 		hlist_del(&filter->fdir_node);
6090 		kfree(filter);
6091 	}
6092 	adapter->fdir_filter_count = 0;
6093 
6094 	spin_unlock(&adapter->fdir_perfect_lock);
6095 }
6096 
6097 void ixgbe_down(struct ixgbe_adapter *adapter)
6098 {
6099 	struct net_device *netdev = adapter->netdev;
6100 	struct ixgbe_hw *hw = &adapter->hw;
6101 	int i;
6102 
6103 	/* signal that we are down to the interrupt handler */
6104 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
6105 		return; /* do nothing if already down */
6106 
6107 	/* Shut off incoming Tx traffic */
6108 	netif_tx_stop_all_queues(netdev);
6109 
6110 	/* call carrier off first to avoid false dev_watchdog timeouts */
6111 	netif_carrier_off(netdev);
6112 	netif_tx_disable(netdev);
6113 
6114 	/* Disable Rx */
6115 	ixgbe_disable_rx(adapter);
6116 
6117 	/* synchronize_rcu() needed for pending XDP buffers to drain */
6118 	if (adapter->xdp_ring[0])
6119 		synchronize_rcu();
6120 
6121 	ixgbe_irq_disable(adapter);
6122 
6123 	ixgbe_napi_disable_all(adapter);
6124 
6125 	clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
6126 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6127 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6128 
6129 	del_timer_sync(&adapter->service_timer);
6130 
6131 	if (adapter->num_vfs) {
6132 		/* Clear EITR Select mapping */
6133 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
6134 
6135 		/* Mark all the VFs as inactive */
6136 		for (i = 0 ; i < adapter->num_vfs; i++)
6137 			adapter->vfinfo[i].clear_to_send = false;
6138 
6139 		/* ping all the active vfs to let them know we are going down */
6140 		ixgbe_ping_all_vfs(adapter);
6141 
6142 		/* Disable all VFTE/VFRE TX/RX */
6143 		ixgbe_disable_tx_rx(adapter);
6144 	}
6145 
6146 	/* disable transmits in the hardware now that interrupts are off */
6147 	ixgbe_disable_tx(adapter);
6148 
6149 	if (!pci_channel_offline(adapter->pdev))
6150 		ixgbe_reset(adapter);
6151 
6152 	/* power down the optics for 82599 SFP+ fiber */
6153 	if (hw->mac.ops.disable_tx_laser)
6154 		hw->mac.ops.disable_tx_laser(hw);
6155 
6156 	ixgbe_clean_all_tx_rings(adapter);
6157 	ixgbe_clean_all_rx_rings(adapter);
6158 }
6159 
6160 /**
6161  * ixgbe_set_eee_capable - helper function to determine EEE support on X550
6162  * @adapter: board private structure
6163  */
6164 static void ixgbe_set_eee_capable(struct ixgbe_adapter *adapter)
6165 {
6166 	struct ixgbe_hw *hw = &adapter->hw;
6167 
6168 	switch (hw->device_id) {
6169 	case IXGBE_DEV_ID_X550EM_A_1G_T:
6170 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6171 		if (!hw->phy.eee_speeds_supported)
6172 			break;
6173 		adapter->flags2 |= IXGBE_FLAG2_EEE_CAPABLE;
6174 		if (!hw->phy.eee_speeds_advertised)
6175 			break;
6176 		adapter->flags2 |= IXGBE_FLAG2_EEE_ENABLED;
6177 		break;
6178 	default:
6179 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_CAPABLE;
6180 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_ENABLED;
6181 		break;
6182 	}
6183 }
6184 
6185 /**
6186  * ixgbe_tx_timeout - Respond to a Tx Hang
6187  * @netdev: network interface device structure
6188  * @txqueue: queue number that timed out
6189  **/
6190 static void ixgbe_tx_timeout(struct net_device *netdev, unsigned int __always_unused txqueue)
6191 {
6192 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6193 
6194 	/* Do the reset outside of interrupt context */
6195 	ixgbe_tx_timeout_reset(adapter);
6196 }
6197 
6198 #ifdef CONFIG_IXGBE_DCB
6199 static void ixgbe_init_dcb(struct ixgbe_adapter *adapter)
6200 {
6201 	struct ixgbe_hw *hw = &adapter->hw;
6202 	struct tc_configuration *tc;
6203 	int j;
6204 
6205 	switch (hw->mac.type) {
6206 	case ixgbe_mac_82598EB:
6207 	case ixgbe_mac_82599EB:
6208 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
6209 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
6210 		break;
6211 	case ixgbe_mac_X540:
6212 	case ixgbe_mac_X550:
6213 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
6214 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
6215 		break;
6216 	case ixgbe_mac_X550EM_x:
6217 	case ixgbe_mac_x550em_a:
6218 	default:
6219 		adapter->dcb_cfg.num_tcs.pg_tcs = DEF_TRAFFIC_CLASS;
6220 		adapter->dcb_cfg.num_tcs.pfc_tcs = DEF_TRAFFIC_CLASS;
6221 		break;
6222 	}
6223 
6224 	/* Configure DCB traffic classes */
6225 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
6226 		tc = &adapter->dcb_cfg.tc_config[j];
6227 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
6228 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
6229 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
6230 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
6231 		tc->dcb_pfc = pfc_disabled;
6232 	}
6233 
6234 	/* Initialize default user to priority mapping, UPx->TC0 */
6235 	tc = &adapter->dcb_cfg.tc_config[0];
6236 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
6237 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
6238 
6239 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
6240 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
6241 	adapter->dcb_cfg.pfc_mode_enable = false;
6242 	adapter->dcb_set_bitmap = 0x00;
6243 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
6244 		adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
6245 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
6246 	       sizeof(adapter->temp_dcb_cfg));
6247 }
6248 #endif
6249 
6250 /**
6251  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
6252  * @adapter: board private structure to initialize
6253  * @ii: pointer to ixgbe_info for device
6254  *
6255  * ixgbe_sw_init initializes the Adapter private data structure.
6256  * Fields are initialized based on PCI device information and
6257  * OS network device settings (MTU size).
6258  **/
6259 static int ixgbe_sw_init(struct ixgbe_adapter *adapter,
6260 			 const struct ixgbe_info *ii)
6261 {
6262 	struct ixgbe_hw *hw = &adapter->hw;
6263 	struct pci_dev *pdev = adapter->pdev;
6264 	unsigned int rss, fdir;
6265 	u32 fwsm;
6266 	int i;
6267 
6268 	/* PCI config space info */
6269 
6270 	hw->vendor_id = pdev->vendor;
6271 	hw->device_id = pdev->device;
6272 	hw->revision_id = pdev->revision;
6273 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
6274 	hw->subsystem_device_id = pdev->subsystem_device;
6275 
6276 	/* get_invariants needs the device IDs */
6277 	ii->get_invariants(hw);
6278 
6279 	/* Set common capability flags and settings */
6280 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
6281 	adapter->ring_feature[RING_F_RSS].limit = rss;
6282 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
6283 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
6284 	adapter->atr_sample_rate = 20;
6285 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
6286 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
6287 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
6288 	adapter->ring_feature[RING_F_VMDQ].limit = 1;
6289 #ifdef CONFIG_IXGBE_DCA
6290 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
6291 #endif
6292 #ifdef CONFIG_IXGBE_DCB
6293 	adapter->flags |= IXGBE_FLAG_DCB_CAPABLE;
6294 	adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
6295 #endif
6296 #ifdef IXGBE_FCOE
6297 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
6298 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6299 #ifdef CONFIG_IXGBE_DCB
6300 	/* Default traffic class to use for FCoE */
6301 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
6302 #endif /* CONFIG_IXGBE_DCB */
6303 #endif /* IXGBE_FCOE */
6304 
6305 	/* initialize static ixgbe jump table entries */
6306 	adapter->jump_tables[0] = kzalloc(sizeof(*adapter->jump_tables[0]),
6307 					  GFP_KERNEL);
6308 	if (!adapter->jump_tables[0])
6309 		return -ENOMEM;
6310 	adapter->jump_tables[0]->mat = ixgbe_ipv4_fields;
6311 
6312 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++)
6313 		adapter->jump_tables[i] = NULL;
6314 
6315 	adapter->mac_table = kcalloc(hw->mac.num_rar_entries,
6316 				     sizeof(struct ixgbe_mac_addr),
6317 				     GFP_KERNEL);
6318 	if (!adapter->mac_table)
6319 		return -ENOMEM;
6320 
6321 	if (ixgbe_init_rss_key(adapter))
6322 		return -ENOMEM;
6323 
6324 	adapter->af_xdp_zc_qps = bitmap_zalloc(MAX_XDP_QUEUES, GFP_KERNEL);
6325 	if (!adapter->af_xdp_zc_qps)
6326 		return -ENOMEM;
6327 
6328 	/* Set MAC specific capability flags and exceptions */
6329 	switch (hw->mac.type) {
6330 	case ixgbe_mac_82598EB:
6331 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
6332 
6333 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
6334 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
6335 
6336 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
6337 		adapter->ring_feature[RING_F_FDIR].limit = 0;
6338 		adapter->atr_sample_rate = 0;
6339 		adapter->fdir_pballoc = 0;
6340 #ifdef IXGBE_FCOE
6341 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6342 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6343 #ifdef CONFIG_IXGBE_DCB
6344 		adapter->fcoe.up = 0;
6345 #endif /* IXGBE_DCB */
6346 #endif /* IXGBE_FCOE */
6347 		break;
6348 	case ixgbe_mac_82599EB:
6349 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
6350 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6351 		break;
6352 	case ixgbe_mac_X540:
6353 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
6354 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
6355 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6356 		break;
6357 	case ixgbe_mac_x550em_a:
6358 		switch (hw->device_id) {
6359 		case IXGBE_DEV_ID_X550EM_A_1G_T:
6360 		case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6361 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6362 			break;
6363 		default:
6364 			break;
6365 		}
6366 		fallthrough;
6367 	case ixgbe_mac_X550EM_x:
6368 #ifdef CONFIG_IXGBE_DCB
6369 		adapter->flags &= ~IXGBE_FLAG_DCB_CAPABLE;
6370 #endif
6371 #ifdef IXGBE_FCOE
6372 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6373 #ifdef CONFIG_IXGBE_DCB
6374 		adapter->fcoe.up = 0;
6375 #endif /* IXGBE_DCB */
6376 #endif /* IXGBE_FCOE */
6377 		fallthrough;
6378 	case ixgbe_mac_X550:
6379 		if (hw->mac.type == ixgbe_mac_X550)
6380 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6381 #ifdef CONFIG_IXGBE_DCA
6382 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
6383 #endif
6384 		break;
6385 	default:
6386 		break;
6387 	}
6388 
6389 #ifdef IXGBE_FCOE
6390 	/* FCoE support exists, always init the FCoE lock */
6391 	spin_lock_init(&adapter->fcoe.lock);
6392 
6393 #endif
6394 	/* n-tuple support exists, always init our spinlock */
6395 	spin_lock_init(&adapter->fdir_perfect_lock);
6396 
6397 #ifdef CONFIG_IXGBE_DCB
6398 	ixgbe_init_dcb(adapter);
6399 #endif
6400 	ixgbe_init_ipsec_offload(adapter);
6401 
6402 	/* default flow control settings */
6403 	hw->fc.requested_mode = ixgbe_fc_full;
6404 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
6405 	ixgbe_pbthresh_setup(adapter);
6406 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
6407 	hw->fc.send_xon = true;
6408 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
6409 
6410 #ifdef CONFIG_PCI_IOV
6411 	if (max_vfs > 0)
6412 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
6413 
6414 	/* assign number of SR-IOV VFs */
6415 	if (hw->mac.type != ixgbe_mac_82598EB) {
6416 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
6417 			max_vfs = 0;
6418 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
6419 		}
6420 	}
6421 #endif /* CONFIG_PCI_IOV */
6422 
6423 	/* enable itr by default in dynamic mode */
6424 	adapter->rx_itr_setting = 1;
6425 	adapter->tx_itr_setting = 1;
6426 
6427 	/* set default ring sizes */
6428 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
6429 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
6430 
6431 	/* set default work limits */
6432 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
6433 
6434 	/* initialize eeprom parameters */
6435 	if (ixgbe_init_eeprom_params_generic(hw)) {
6436 		e_dev_err("EEPROM initialization failed\n");
6437 		return -EIO;
6438 	}
6439 
6440 	/* PF holds first pool slot */
6441 	set_bit(0, adapter->fwd_bitmask);
6442 	set_bit(__IXGBE_DOWN, &adapter->state);
6443 
6444 	return 0;
6445 }
6446 
6447 /**
6448  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
6449  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
6450  *
6451  * Return 0 on success, negative on failure
6452  **/
6453 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
6454 {
6455 	struct device *dev = tx_ring->dev;
6456 	int orig_node = dev_to_node(dev);
6457 	int ring_node = NUMA_NO_NODE;
6458 	int size;
6459 
6460 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
6461 
6462 	if (tx_ring->q_vector)
6463 		ring_node = tx_ring->q_vector->numa_node;
6464 
6465 	tx_ring->tx_buffer_info = vmalloc_node(size, ring_node);
6466 	if (!tx_ring->tx_buffer_info)
6467 		tx_ring->tx_buffer_info = vmalloc(size);
6468 	if (!tx_ring->tx_buffer_info)
6469 		goto err;
6470 
6471 	/* round up to nearest 4K */
6472 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
6473 	tx_ring->size = ALIGN(tx_ring->size, 4096);
6474 
6475 	set_dev_node(dev, ring_node);
6476 	tx_ring->desc = dma_alloc_coherent(dev,
6477 					   tx_ring->size,
6478 					   &tx_ring->dma,
6479 					   GFP_KERNEL);
6480 	set_dev_node(dev, orig_node);
6481 	if (!tx_ring->desc)
6482 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
6483 						   &tx_ring->dma, GFP_KERNEL);
6484 	if (!tx_ring->desc)
6485 		goto err;
6486 
6487 	tx_ring->next_to_use = 0;
6488 	tx_ring->next_to_clean = 0;
6489 	return 0;
6490 
6491 err:
6492 	vfree(tx_ring->tx_buffer_info);
6493 	tx_ring->tx_buffer_info = NULL;
6494 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
6495 	return -ENOMEM;
6496 }
6497 
6498 /**
6499  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
6500  * @adapter: board private structure
6501  *
6502  * If this function returns with an error, then it's possible one or
6503  * more of the rings is populated (while the rest are not).  It is the
6504  * callers duty to clean those orphaned rings.
6505  *
6506  * Return 0 on success, negative on failure
6507  **/
6508 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
6509 {
6510 	int i, j = 0, err = 0;
6511 
6512 	for (i = 0; i < adapter->num_tx_queues; i++) {
6513 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
6514 		if (!err)
6515 			continue;
6516 
6517 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
6518 		goto err_setup_tx;
6519 	}
6520 	for (j = 0; j < adapter->num_xdp_queues; j++) {
6521 		err = ixgbe_setup_tx_resources(adapter->xdp_ring[j]);
6522 		if (!err)
6523 			continue;
6524 
6525 		e_err(probe, "Allocation for Tx Queue %u failed\n", j);
6526 		goto err_setup_tx;
6527 	}
6528 
6529 	return 0;
6530 err_setup_tx:
6531 	/* rewind the index freeing the rings as we go */
6532 	while (j--)
6533 		ixgbe_free_tx_resources(adapter->xdp_ring[j]);
6534 	while (i--)
6535 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
6536 	return err;
6537 }
6538 
6539 static int ixgbe_rx_napi_id(struct ixgbe_ring *rx_ring)
6540 {
6541 	struct ixgbe_q_vector *q_vector = rx_ring->q_vector;
6542 
6543 	return q_vector ? q_vector->napi.napi_id : 0;
6544 }
6545 
6546 /**
6547  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
6548  * @adapter: pointer to ixgbe_adapter
6549  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
6550  *
6551  * Returns 0 on success, negative on failure
6552  **/
6553 int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
6554 			     struct ixgbe_ring *rx_ring)
6555 {
6556 	struct device *dev = rx_ring->dev;
6557 	int orig_node = dev_to_node(dev);
6558 	int ring_node = NUMA_NO_NODE;
6559 	int size;
6560 
6561 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
6562 
6563 	if (rx_ring->q_vector)
6564 		ring_node = rx_ring->q_vector->numa_node;
6565 
6566 	rx_ring->rx_buffer_info = vmalloc_node(size, ring_node);
6567 	if (!rx_ring->rx_buffer_info)
6568 		rx_ring->rx_buffer_info = vmalloc(size);
6569 	if (!rx_ring->rx_buffer_info)
6570 		goto err;
6571 
6572 	/* Round up to nearest 4K */
6573 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
6574 	rx_ring->size = ALIGN(rx_ring->size, 4096);
6575 
6576 	set_dev_node(dev, ring_node);
6577 	rx_ring->desc = dma_alloc_coherent(dev,
6578 					   rx_ring->size,
6579 					   &rx_ring->dma,
6580 					   GFP_KERNEL);
6581 	set_dev_node(dev, orig_node);
6582 	if (!rx_ring->desc)
6583 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
6584 						   &rx_ring->dma, GFP_KERNEL);
6585 	if (!rx_ring->desc)
6586 		goto err;
6587 
6588 	rx_ring->next_to_clean = 0;
6589 	rx_ring->next_to_use = 0;
6590 
6591 	/* XDP RX-queue info */
6592 	if (xdp_rxq_info_reg(&rx_ring->xdp_rxq, adapter->netdev,
6593 			     rx_ring->queue_index, ixgbe_rx_napi_id(rx_ring)) < 0)
6594 		goto err;
6595 
6596 	rx_ring->xdp_prog = adapter->xdp_prog;
6597 
6598 	return 0;
6599 err:
6600 	vfree(rx_ring->rx_buffer_info);
6601 	rx_ring->rx_buffer_info = NULL;
6602 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
6603 	return -ENOMEM;
6604 }
6605 
6606 /**
6607  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
6608  * @adapter: board private structure
6609  *
6610  * If this function returns with an error, then it's possible one or
6611  * more of the rings is populated (while the rest are not).  It is the
6612  * callers duty to clean those orphaned rings.
6613  *
6614  * Return 0 on success, negative on failure
6615  **/
6616 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
6617 {
6618 	int i, err = 0;
6619 
6620 	for (i = 0; i < adapter->num_rx_queues; i++) {
6621 		err = ixgbe_setup_rx_resources(adapter, adapter->rx_ring[i]);
6622 		if (!err)
6623 			continue;
6624 
6625 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
6626 		goto err_setup_rx;
6627 	}
6628 
6629 #ifdef IXGBE_FCOE
6630 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
6631 	if (!err)
6632 #endif
6633 		return 0;
6634 err_setup_rx:
6635 	/* rewind the index freeing the rings as we go */
6636 	while (i--)
6637 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
6638 	return err;
6639 }
6640 
6641 /**
6642  * ixgbe_free_tx_resources - Free Tx Resources per Queue
6643  * @tx_ring: Tx descriptor ring for a specific queue
6644  *
6645  * Free all transmit software resources
6646  **/
6647 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
6648 {
6649 	ixgbe_clean_tx_ring(tx_ring);
6650 
6651 	vfree(tx_ring->tx_buffer_info);
6652 	tx_ring->tx_buffer_info = NULL;
6653 
6654 	/* if not set, then don't free */
6655 	if (!tx_ring->desc)
6656 		return;
6657 
6658 	dma_free_coherent(tx_ring->dev, tx_ring->size,
6659 			  tx_ring->desc, tx_ring->dma);
6660 
6661 	tx_ring->desc = NULL;
6662 }
6663 
6664 /**
6665  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
6666  * @adapter: board private structure
6667  *
6668  * Free all transmit software resources
6669  **/
6670 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
6671 {
6672 	int i;
6673 
6674 	for (i = 0; i < adapter->num_tx_queues; i++)
6675 		if (adapter->tx_ring[i]->desc)
6676 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
6677 	for (i = 0; i < adapter->num_xdp_queues; i++)
6678 		if (adapter->xdp_ring[i]->desc)
6679 			ixgbe_free_tx_resources(adapter->xdp_ring[i]);
6680 }
6681 
6682 /**
6683  * ixgbe_free_rx_resources - Free Rx Resources
6684  * @rx_ring: ring to clean the resources from
6685  *
6686  * Free all receive software resources
6687  **/
6688 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
6689 {
6690 	ixgbe_clean_rx_ring(rx_ring);
6691 
6692 	rx_ring->xdp_prog = NULL;
6693 	xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
6694 	vfree(rx_ring->rx_buffer_info);
6695 	rx_ring->rx_buffer_info = NULL;
6696 
6697 	/* if not set, then don't free */
6698 	if (!rx_ring->desc)
6699 		return;
6700 
6701 	dma_free_coherent(rx_ring->dev, rx_ring->size,
6702 			  rx_ring->desc, rx_ring->dma);
6703 
6704 	rx_ring->desc = NULL;
6705 }
6706 
6707 /**
6708  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
6709  * @adapter: board private structure
6710  *
6711  * Free all receive software resources
6712  **/
6713 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
6714 {
6715 	int i;
6716 
6717 #ifdef IXGBE_FCOE
6718 	ixgbe_free_fcoe_ddp_resources(adapter);
6719 
6720 #endif
6721 	for (i = 0; i < adapter->num_rx_queues; i++)
6722 		if (adapter->rx_ring[i]->desc)
6723 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
6724 }
6725 
6726 /**
6727  * ixgbe_change_mtu - Change the Maximum Transfer Unit
6728  * @netdev: network interface device structure
6729  * @new_mtu: new value for maximum frame size
6730  *
6731  * Returns 0 on success, negative on failure
6732  **/
6733 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
6734 {
6735 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6736 
6737 	if (adapter->xdp_prog) {
6738 		int new_frame_size = new_mtu + ETH_HLEN + ETH_FCS_LEN +
6739 				     VLAN_HLEN;
6740 		int i;
6741 
6742 		for (i = 0; i < adapter->num_rx_queues; i++) {
6743 			struct ixgbe_ring *ring = adapter->rx_ring[i];
6744 
6745 			if (new_frame_size > ixgbe_rx_bufsz(ring)) {
6746 				e_warn(probe, "Requested MTU size is not supported with XDP\n");
6747 				return -EINVAL;
6748 			}
6749 		}
6750 	}
6751 
6752 	/*
6753 	 * For 82599EB we cannot allow legacy VFs to enable their receive
6754 	 * paths when MTU greater than 1500 is configured.  So display a
6755 	 * warning that legacy VFs will be disabled.
6756 	 */
6757 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
6758 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
6759 	    (new_mtu > ETH_DATA_LEN))
6760 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
6761 
6762 	netdev_dbg(netdev, "changing MTU from %d to %d\n",
6763 		   netdev->mtu, new_mtu);
6764 
6765 	/* must set new MTU before calling down or up */
6766 	netdev->mtu = new_mtu;
6767 
6768 	if (netif_running(netdev))
6769 		ixgbe_reinit_locked(adapter);
6770 
6771 	return 0;
6772 }
6773 
6774 /**
6775  * ixgbe_open - Called when a network interface is made active
6776  * @netdev: network interface device structure
6777  *
6778  * Returns 0 on success, negative value on failure
6779  *
6780  * The open entry point is called when a network interface is made
6781  * active by the system (IFF_UP).  At this point all resources needed
6782  * for transmit and receive operations are allocated, the interrupt
6783  * handler is registered with the OS, the watchdog timer is started,
6784  * and the stack is notified that the interface is ready.
6785  **/
6786 int ixgbe_open(struct net_device *netdev)
6787 {
6788 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6789 	struct ixgbe_hw *hw = &adapter->hw;
6790 	int err, queues;
6791 
6792 	/* disallow open during test */
6793 	if (test_bit(__IXGBE_TESTING, &adapter->state))
6794 		return -EBUSY;
6795 
6796 	netif_carrier_off(netdev);
6797 
6798 	/* allocate transmit descriptors */
6799 	err = ixgbe_setup_all_tx_resources(adapter);
6800 	if (err)
6801 		goto err_setup_tx;
6802 
6803 	/* allocate receive descriptors */
6804 	err = ixgbe_setup_all_rx_resources(adapter);
6805 	if (err)
6806 		goto err_setup_rx;
6807 
6808 	ixgbe_configure(adapter);
6809 
6810 	err = ixgbe_request_irq(adapter);
6811 	if (err)
6812 		goto err_req_irq;
6813 
6814 	/* Notify the stack of the actual queue counts. */
6815 	queues = adapter->num_tx_queues;
6816 	err = netif_set_real_num_tx_queues(netdev, queues);
6817 	if (err)
6818 		goto err_set_queues;
6819 
6820 	queues = adapter->num_rx_queues;
6821 	err = netif_set_real_num_rx_queues(netdev, queues);
6822 	if (err)
6823 		goto err_set_queues;
6824 
6825 	ixgbe_ptp_init(adapter);
6826 
6827 	ixgbe_up_complete(adapter);
6828 
6829 	udp_tunnel_nic_reset_ntf(netdev);
6830 
6831 	return 0;
6832 
6833 err_set_queues:
6834 	ixgbe_free_irq(adapter);
6835 err_req_irq:
6836 	ixgbe_free_all_rx_resources(adapter);
6837 	if (hw->phy.ops.set_phy_power && !adapter->wol)
6838 		hw->phy.ops.set_phy_power(&adapter->hw, false);
6839 err_setup_rx:
6840 	ixgbe_free_all_tx_resources(adapter);
6841 err_setup_tx:
6842 	ixgbe_reset(adapter);
6843 
6844 	return err;
6845 }
6846 
6847 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
6848 {
6849 	ixgbe_ptp_suspend(adapter);
6850 
6851 	if (adapter->hw.phy.ops.enter_lplu) {
6852 		adapter->hw.phy.reset_disable = true;
6853 		ixgbe_down(adapter);
6854 		adapter->hw.phy.ops.enter_lplu(&adapter->hw);
6855 		adapter->hw.phy.reset_disable = false;
6856 	} else {
6857 		ixgbe_down(adapter);
6858 	}
6859 
6860 	ixgbe_free_irq(adapter);
6861 
6862 	ixgbe_free_all_tx_resources(adapter);
6863 	ixgbe_free_all_rx_resources(adapter);
6864 }
6865 
6866 /**
6867  * ixgbe_close - Disables a network interface
6868  * @netdev: network interface device structure
6869  *
6870  * Returns 0, this is not allowed to fail
6871  *
6872  * The close entry point is called when an interface is de-activated
6873  * by the OS.  The hardware is still under the drivers control, but
6874  * needs to be disabled.  A global MAC reset is issued to stop the
6875  * hardware, and all transmit and receive resources are freed.
6876  **/
6877 int ixgbe_close(struct net_device *netdev)
6878 {
6879 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
6880 
6881 	ixgbe_ptp_stop(adapter);
6882 
6883 	if (netif_device_present(netdev))
6884 		ixgbe_close_suspend(adapter);
6885 
6886 	ixgbe_fdir_filter_exit(adapter);
6887 
6888 	ixgbe_release_hw_control(adapter);
6889 
6890 	return 0;
6891 }
6892 
6893 static int __maybe_unused ixgbe_resume(struct device *dev_d)
6894 {
6895 	struct pci_dev *pdev = to_pci_dev(dev_d);
6896 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6897 	struct net_device *netdev = adapter->netdev;
6898 	u32 err;
6899 
6900 	adapter->hw.hw_addr = adapter->io_addr;
6901 
6902 	err = pci_enable_device_mem(pdev);
6903 	if (err) {
6904 		e_dev_err("Cannot enable PCI device from suspend\n");
6905 		return err;
6906 	}
6907 	smp_mb__before_atomic();
6908 	clear_bit(__IXGBE_DISABLED, &adapter->state);
6909 	pci_set_master(pdev);
6910 
6911 	device_wakeup_disable(dev_d);
6912 
6913 	ixgbe_reset(adapter);
6914 
6915 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
6916 
6917 	rtnl_lock();
6918 	err = ixgbe_init_interrupt_scheme(adapter);
6919 	if (!err && netif_running(netdev))
6920 		err = ixgbe_open(netdev);
6921 
6922 
6923 	if (!err)
6924 		netif_device_attach(netdev);
6925 	rtnl_unlock();
6926 
6927 	return err;
6928 }
6929 
6930 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
6931 {
6932 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6933 	struct net_device *netdev = adapter->netdev;
6934 	struct ixgbe_hw *hw = &adapter->hw;
6935 	u32 ctrl;
6936 	u32 wufc = adapter->wol;
6937 
6938 	rtnl_lock();
6939 	netif_device_detach(netdev);
6940 
6941 	if (netif_running(netdev))
6942 		ixgbe_close_suspend(adapter);
6943 
6944 	ixgbe_clear_interrupt_scheme(adapter);
6945 	rtnl_unlock();
6946 
6947 	if (hw->mac.ops.stop_link_on_d3)
6948 		hw->mac.ops.stop_link_on_d3(hw);
6949 
6950 	if (wufc) {
6951 		u32 fctrl;
6952 
6953 		ixgbe_set_rx_mode(netdev);
6954 
6955 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
6956 		if (hw->mac.ops.enable_tx_laser)
6957 			hw->mac.ops.enable_tx_laser(hw);
6958 
6959 		/* enable the reception of multicast packets */
6960 		fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6961 		fctrl |= IXGBE_FCTRL_MPE;
6962 		IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
6963 
6964 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
6965 		ctrl |= IXGBE_CTRL_GIO_DIS;
6966 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
6967 
6968 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
6969 	} else {
6970 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
6971 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
6972 	}
6973 
6974 	switch (hw->mac.type) {
6975 	case ixgbe_mac_82598EB:
6976 		pci_wake_from_d3(pdev, false);
6977 		break;
6978 	case ixgbe_mac_82599EB:
6979 	case ixgbe_mac_X540:
6980 	case ixgbe_mac_X550:
6981 	case ixgbe_mac_X550EM_x:
6982 	case ixgbe_mac_x550em_a:
6983 		pci_wake_from_d3(pdev, !!wufc);
6984 		break;
6985 	default:
6986 		break;
6987 	}
6988 
6989 	*enable_wake = !!wufc;
6990 	if (hw->phy.ops.set_phy_power && !*enable_wake)
6991 		hw->phy.ops.set_phy_power(hw, false);
6992 
6993 	ixgbe_release_hw_control(adapter);
6994 
6995 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
6996 		pci_disable_device(pdev);
6997 
6998 	return 0;
6999 }
7000 
7001 static int __maybe_unused ixgbe_suspend(struct device *dev_d)
7002 {
7003 	struct pci_dev *pdev = to_pci_dev(dev_d);
7004 	int retval;
7005 	bool wake;
7006 
7007 	retval = __ixgbe_shutdown(pdev, &wake);
7008 
7009 	device_set_wakeup_enable(dev_d, wake);
7010 
7011 	return retval;
7012 }
7013 
7014 static void ixgbe_shutdown(struct pci_dev *pdev)
7015 {
7016 	bool wake;
7017 
7018 	__ixgbe_shutdown(pdev, &wake);
7019 
7020 	if (system_state == SYSTEM_POWER_OFF) {
7021 		pci_wake_from_d3(pdev, wake);
7022 		pci_set_power_state(pdev, PCI_D3hot);
7023 	}
7024 }
7025 
7026 /**
7027  * ixgbe_update_stats - Update the board statistics counters.
7028  * @adapter: board private structure
7029  **/
7030 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
7031 {
7032 	struct net_device *netdev = adapter->netdev;
7033 	struct ixgbe_hw *hw = &adapter->hw;
7034 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
7035 	u64 total_mpc = 0;
7036 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
7037 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
7038 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
7039 	u64 alloc_rx_page = 0;
7040 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
7041 
7042 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7043 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7044 		return;
7045 
7046 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
7047 		u64 rsc_count = 0;
7048 		u64 rsc_flush = 0;
7049 		for (i = 0; i < adapter->num_rx_queues; i++) {
7050 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
7051 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
7052 		}
7053 		adapter->rsc_total_count = rsc_count;
7054 		adapter->rsc_total_flush = rsc_flush;
7055 	}
7056 
7057 	for (i = 0; i < adapter->num_rx_queues; i++) {
7058 		struct ixgbe_ring *rx_ring = READ_ONCE(adapter->rx_ring[i]);
7059 
7060 		if (!rx_ring)
7061 			continue;
7062 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
7063 		alloc_rx_page += rx_ring->rx_stats.alloc_rx_page;
7064 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
7065 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
7066 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
7067 		bytes += rx_ring->stats.bytes;
7068 		packets += rx_ring->stats.packets;
7069 	}
7070 	adapter->non_eop_descs = non_eop_descs;
7071 	adapter->alloc_rx_page = alloc_rx_page;
7072 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
7073 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
7074 	adapter->hw_csum_rx_error = hw_csum_rx_error;
7075 	netdev->stats.rx_bytes = bytes;
7076 	netdev->stats.rx_packets = packets;
7077 
7078 	bytes = 0;
7079 	packets = 0;
7080 	/* gather some stats to the adapter struct that are per queue */
7081 	for (i = 0; i < adapter->num_tx_queues; i++) {
7082 		struct ixgbe_ring *tx_ring = READ_ONCE(adapter->tx_ring[i]);
7083 
7084 		if (!tx_ring)
7085 			continue;
7086 		restart_queue += tx_ring->tx_stats.restart_queue;
7087 		tx_busy += tx_ring->tx_stats.tx_busy;
7088 		bytes += tx_ring->stats.bytes;
7089 		packets += tx_ring->stats.packets;
7090 	}
7091 	for (i = 0; i < adapter->num_xdp_queues; i++) {
7092 		struct ixgbe_ring *xdp_ring = READ_ONCE(adapter->xdp_ring[i]);
7093 
7094 		if (!xdp_ring)
7095 			continue;
7096 		restart_queue += xdp_ring->tx_stats.restart_queue;
7097 		tx_busy += xdp_ring->tx_stats.tx_busy;
7098 		bytes += xdp_ring->stats.bytes;
7099 		packets += xdp_ring->stats.packets;
7100 	}
7101 	adapter->restart_queue = restart_queue;
7102 	adapter->tx_busy = tx_busy;
7103 	netdev->stats.tx_bytes = bytes;
7104 	netdev->stats.tx_packets = packets;
7105 
7106 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
7107 
7108 	/* 8 register reads */
7109 	for (i = 0; i < 8; i++) {
7110 		/* for packet buffers not used, the register should read 0 */
7111 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
7112 		missed_rx += mpc;
7113 		hwstats->mpc[i] += mpc;
7114 		total_mpc += hwstats->mpc[i];
7115 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
7116 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
7117 		switch (hw->mac.type) {
7118 		case ixgbe_mac_82598EB:
7119 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
7120 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
7121 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
7122 			hwstats->pxonrxc[i] +=
7123 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
7124 			break;
7125 		case ixgbe_mac_82599EB:
7126 		case ixgbe_mac_X540:
7127 		case ixgbe_mac_X550:
7128 		case ixgbe_mac_X550EM_x:
7129 		case ixgbe_mac_x550em_a:
7130 			hwstats->pxonrxc[i] +=
7131 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
7132 			break;
7133 		default:
7134 			break;
7135 		}
7136 	}
7137 
7138 	/*16 register reads */
7139 	for (i = 0; i < 16; i++) {
7140 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
7141 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
7142 		if ((hw->mac.type == ixgbe_mac_82599EB) ||
7143 		    (hw->mac.type == ixgbe_mac_X540) ||
7144 		    (hw->mac.type == ixgbe_mac_X550) ||
7145 		    (hw->mac.type == ixgbe_mac_X550EM_x) ||
7146 		    (hw->mac.type == ixgbe_mac_x550em_a)) {
7147 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
7148 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
7149 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
7150 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
7151 		}
7152 	}
7153 
7154 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
7155 	/* work around hardware counting issue */
7156 	hwstats->gprc -= missed_rx;
7157 
7158 	ixgbe_update_xoff_received(adapter);
7159 
7160 	/* 82598 hardware only has a 32 bit counter in the high register */
7161 	switch (hw->mac.type) {
7162 	case ixgbe_mac_82598EB:
7163 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
7164 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
7165 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
7166 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
7167 		break;
7168 	case ixgbe_mac_X540:
7169 	case ixgbe_mac_X550:
7170 	case ixgbe_mac_X550EM_x:
7171 	case ixgbe_mac_x550em_a:
7172 		/* OS2BMC stats are X540 and later */
7173 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
7174 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
7175 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
7176 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
7177 		fallthrough;
7178 	case ixgbe_mac_82599EB:
7179 		for (i = 0; i < 16; i++)
7180 			adapter->hw_rx_no_dma_resources +=
7181 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
7182 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
7183 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
7184 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
7185 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
7186 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
7187 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
7188 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
7189 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
7190 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
7191 #ifdef IXGBE_FCOE
7192 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
7193 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
7194 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
7195 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
7196 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
7197 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
7198 		/* Add up per cpu counters for total ddp aloc fail */
7199 		if (adapter->fcoe.ddp_pool) {
7200 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
7201 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
7202 			unsigned int cpu;
7203 			u64 noddp = 0, noddp_ext_buff = 0;
7204 			for_each_possible_cpu(cpu) {
7205 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
7206 				noddp += ddp_pool->noddp;
7207 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
7208 			}
7209 			hwstats->fcoe_noddp = noddp;
7210 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
7211 		}
7212 #endif /* IXGBE_FCOE */
7213 		break;
7214 	default:
7215 		break;
7216 	}
7217 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
7218 	hwstats->bprc += bprc;
7219 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
7220 	if (hw->mac.type == ixgbe_mac_82598EB)
7221 		hwstats->mprc -= bprc;
7222 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
7223 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
7224 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
7225 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
7226 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
7227 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
7228 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
7229 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
7230 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
7231 	hwstats->lxontxc += lxon;
7232 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
7233 	hwstats->lxofftxc += lxoff;
7234 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
7235 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
7236 	/*
7237 	 * 82598 errata - tx of flow control packets is included in tx counters
7238 	 */
7239 	xon_off_tot = lxon + lxoff;
7240 	hwstats->gptc -= xon_off_tot;
7241 	hwstats->mptc -= xon_off_tot;
7242 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
7243 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
7244 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
7245 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
7246 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
7247 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
7248 	hwstats->ptc64 -= xon_off_tot;
7249 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
7250 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
7251 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
7252 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
7253 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
7254 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
7255 
7256 	/* Fill out the OS statistics structure */
7257 	netdev->stats.multicast = hwstats->mprc;
7258 
7259 	/* Rx Errors */
7260 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
7261 	netdev->stats.rx_dropped = 0;
7262 	netdev->stats.rx_length_errors = hwstats->rlec;
7263 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
7264 	netdev->stats.rx_missed_errors = total_mpc;
7265 }
7266 
7267 /**
7268  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
7269  * @adapter: pointer to the device adapter structure
7270  **/
7271 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
7272 {
7273 	struct ixgbe_hw *hw = &adapter->hw;
7274 	int i;
7275 
7276 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
7277 		return;
7278 
7279 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
7280 
7281 	/* if interface is down do nothing */
7282 	if (test_bit(__IXGBE_DOWN, &adapter->state))
7283 		return;
7284 
7285 	/* do nothing if we are not using signature filters */
7286 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
7287 		return;
7288 
7289 	adapter->fdir_overflow++;
7290 
7291 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
7292 		for (i = 0; i < adapter->num_tx_queues; i++)
7293 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7294 				&(adapter->tx_ring[i]->state));
7295 		for (i = 0; i < adapter->num_xdp_queues; i++)
7296 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7297 				&adapter->xdp_ring[i]->state);
7298 		/* re-enable flow director interrupts */
7299 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
7300 	} else {
7301 		e_err(probe, "failed to finish FDIR re-initialization, "
7302 		      "ignored adding FDIR ATR filters\n");
7303 	}
7304 }
7305 
7306 /**
7307  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
7308  * @adapter: pointer to the device adapter structure
7309  *
7310  * This function serves two purposes.  First it strobes the interrupt lines
7311  * in order to make certain interrupts are occurring.  Secondly it sets the
7312  * bits needed to check for TX hangs.  As a result we should immediately
7313  * determine if a hang has occurred.
7314  */
7315 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
7316 {
7317 	struct ixgbe_hw *hw = &adapter->hw;
7318 	u64 eics = 0;
7319 	int i;
7320 
7321 	/* If we're down, removing or resetting, just bail */
7322 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7323 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7324 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7325 		return;
7326 
7327 	/* Force detection of hung controller */
7328 	if (netif_carrier_ok(adapter->netdev)) {
7329 		for (i = 0; i < adapter->num_tx_queues; i++)
7330 			set_check_for_tx_hang(adapter->tx_ring[i]);
7331 		for (i = 0; i < adapter->num_xdp_queues; i++)
7332 			set_check_for_tx_hang(adapter->xdp_ring[i]);
7333 	}
7334 
7335 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
7336 		/*
7337 		 * for legacy and MSI interrupts don't set any bits
7338 		 * that are enabled for EIAM, because this operation
7339 		 * would set *both* EIMS and EICS for any bit in EIAM
7340 		 */
7341 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
7342 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
7343 	} else {
7344 		/* get one bit for every active tx/rx interrupt vector */
7345 		for (i = 0; i < adapter->num_q_vectors; i++) {
7346 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
7347 			if (qv->rx.ring || qv->tx.ring)
7348 				eics |= BIT_ULL(i);
7349 		}
7350 	}
7351 
7352 	/* Cause software interrupt to ensure rings are cleaned */
7353 	ixgbe_irq_rearm_queues(adapter, eics);
7354 }
7355 
7356 /**
7357  * ixgbe_watchdog_update_link - update the link status
7358  * @adapter: pointer to the device adapter structure
7359  **/
7360 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
7361 {
7362 	struct ixgbe_hw *hw = &adapter->hw;
7363 	u32 link_speed = adapter->link_speed;
7364 	bool link_up = adapter->link_up;
7365 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
7366 
7367 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
7368 		return;
7369 
7370 	if (hw->mac.ops.check_link) {
7371 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
7372 	} else {
7373 		/* always assume link is up, if no check link function */
7374 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
7375 		link_up = true;
7376 	}
7377 
7378 	if (adapter->ixgbe_ieee_pfc)
7379 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
7380 
7381 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
7382 		hw->mac.ops.fc_enable(hw);
7383 		ixgbe_set_rx_drop_en(adapter);
7384 	}
7385 
7386 	if (link_up ||
7387 	    time_after(jiffies, (adapter->link_check_timeout +
7388 				 IXGBE_TRY_LINK_TIMEOUT))) {
7389 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
7390 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
7391 		IXGBE_WRITE_FLUSH(hw);
7392 	}
7393 
7394 	adapter->link_up = link_up;
7395 	adapter->link_speed = link_speed;
7396 }
7397 
7398 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
7399 {
7400 #ifdef CONFIG_IXGBE_DCB
7401 	struct net_device *netdev = adapter->netdev;
7402 	struct dcb_app app = {
7403 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
7404 			      .protocol = 0,
7405 			     };
7406 	u8 up = 0;
7407 
7408 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
7409 		up = dcb_ieee_getapp_mask(netdev, &app);
7410 
7411 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
7412 #endif
7413 }
7414 
7415 /**
7416  * ixgbe_watchdog_link_is_up - update netif_carrier status and
7417  *                             print link up message
7418  * @adapter: pointer to the device adapter structure
7419  **/
7420 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
7421 {
7422 	struct net_device *netdev = adapter->netdev;
7423 	struct ixgbe_hw *hw = &adapter->hw;
7424 	u32 link_speed = adapter->link_speed;
7425 	const char *speed_str;
7426 	bool flow_rx, flow_tx;
7427 
7428 	/* only continue if link was previously down */
7429 	if (netif_carrier_ok(netdev))
7430 		return;
7431 
7432 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
7433 
7434 	switch (hw->mac.type) {
7435 	case ixgbe_mac_82598EB: {
7436 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
7437 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
7438 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
7439 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
7440 	}
7441 		break;
7442 	case ixgbe_mac_X540:
7443 	case ixgbe_mac_X550:
7444 	case ixgbe_mac_X550EM_x:
7445 	case ixgbe_mac_x550em_a:
7446 	case ixgbe_mac_82599EB: {
7447 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
7448 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
7449 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
7450 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
7451 	}
7452 		break;
7453 	default:
7454 		flow_tx = false;
7455 		flow_rx = false;
7456 		break;
7457 	}
7458 
7459 	adapter->last_rx_ptp_check = jiffies;
7460 
7461 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7462 		ixgbe_ptp_start_cyclecounter(adapter);
7463 
7464 	switch (link_speed) {
7465 	case IXGBE_LINK_SPEED_10GB_FULL:
7466 		speed_str = "10 Gbps";
7467 		break;
7468 	case IXGBE_LINK_SPEED_5GB_FULL:
7469 		speed_str = "5 Gbps";
7470 		break;
7471 	case IXGBE_LINK_SPEED_2_5GB_FULL:
7472 		speed_str = "2.5 Gbps";
7473 		break;
7474 	case IXGBE_LINK_SPEED_1GB_FULL:
7475 		speed_str = "1 Gbps";
7476 		break;
7477 	case IXGBE_LINK_SPEED_100_FULL:
7478 		speed_str = "100 Mbps";
7479 		break;
7480 	case IXGBE_LINK_SPEED_10_FULL:
7481 		speed_str = "10 Mbps";
7482 		break;
7483 	default:
7484 		speed_str = "unknown speed";
7485 		break;
7486 	}
7487 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
7488 	       ((flow_rx && flow_tx) ? "RX/TX" :
7489 	       (flow_rx ? "RX" :
7490 	       (flow_tx ? "TX" : "None"))));
7491 
7492 	netif_carrier_on(netdev);
7493 	ixgbe_check_vf_rate_limit(adapter);
7494 
7495 	/* enable transmits */
7496 	netif_tx_wake_all_queues(adapter->netdev);
7497 
7498 	/* update the default user priority for VFs */
7499 	ixgbe_update_default_up(adapter);
7500 
7501 	/* ping all the active vfs to let them know link has changed */
7502 	ixgbe_ping_all_vfs(adapter);
7503 }
7504 
7505 /**
7506  * ixgbe_watchdog_link_is_down - update netif_carrier status and
7507  *                               print link down message
7508  * @adapter: pointer to the adapter structure
7509  **/
7510 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
7511 {
7512 	struct net_device *netdev = adapter->netdev;
7513 	struct ixgbe_hw *hw = &adapter->hw;
7514 
7515 	adapter->link_up = false;
7516 	adapter->link_speed = 0;
7517 
7518 	/* only continue if link was up previously */
7519 	if (!netif_carrier_ok(netdev))
7520 		return;
7521 
7522 	/* poll for SFP+ cable when link is down */
7523 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
7524 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
7525 
7526 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7527 		ixgbe_ptp_start_cyclecounter(adapter);
7528 
7529 	e_info(drv, "NIC Link is Down\n");
7530 	netif_carrier_off(netdev);
7531 
7532 	/* ping all the active vfs to let them know link has changed */
7533 	ixgbe_ping_all_vfs(adapter);
7534 }
7535 
7536 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
7537 {
7538 	int i;
7539 
7540 	for (i = 0; i < adapter->num_tx_queues; i++) {
7541 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
7542 
7543 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
7544 			return true;
7545 	}
7546 
7547 	for (i = 0; i < adapter->num_xdp_queues; i++) {
7548 		struct ixgbe_ring *ring = adapter->xdp_ring[i];
7549 
7550 		if (ring->next_to_use != ring->next_to_clean)
7551 			return true;
7552 	}
7553 
7554 	return false;
7555 }
7556 
7557 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
7558 {
7559 	struct ixgbe_hw *hw = &adapter->hw;
7560 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
7561 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
7562 
7563 	int i, j;
7564 
7565 	if (!adapter->num_vfs)
7566 		return false;
7567 
7568 	/* resetting the PF is only needed for MAC before X550 */
7569 	if (hw->mac.type >= ixgbe_mac_X550)
7570 		return false;
7571 
7572 	for (i = 0; i < adapter->num_vfs; i++) {
7573 		for (j = 0; j < q_per_pool; j++) {
7574 			u32 h, t;
7575 
7576 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
7577 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
7578 
7579 			if (h != t)
7580 				return true;
7581 		}
7582 	}
7583 
7584 	return false;
7585 }
7586 
7587 /**
7588  * ixgbe_watchdog_flush_tx - flush queues on link down
7589  * @adapter: pointer to the device adapter structure
7590  **/
7591 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
7592 {
7593 	if (!netif_carrier_ok(adapter->netdev)) {
7594 		if (ixgbe_ring_tx_pending(adapter) ||
7595 		    ixgbe_vf_tx_pending(adapter)) {
7596 			/* We've lost link, so the controller stops DMA,
7597 			 * but we've got queued Tx work that's never going
7598 			 * to get done, so reset controller to flush Tx.
7599 			 * (Do the reset outside of interrupt context).
7600 			 */
7601 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
7602 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
7603 		}
7604 	}
7605 }
7606 
7607 #ifdef CONFIG_PCI_IOV
7608 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
7609 {
7610 	struct ixgbe_hw *hw = &adapter->hw;
7611 	struct pci_dev *pdev = adapter->pdev;
7612 	unsigned int vf;
7613 	u32 gpc;
7614 
7615 	if (!(netif_carrier_ok(adapter->netdev)))
7616 		return;
7617 
7618 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
7619 	if (gpc) /* If incrementing then no need for the check below */
7620 		return;
7621 	/* Check to see if a bad DMA write target from an errant or
7622 	 * malicious VF has caused a PCIe error.  If so then we can
7623 	 * issue a VFLR to the offending VF(s) and then resume without
7624 	 * requesting a full slot reset.
7625 	 */
7626 
7627 	if (!pdev)
7628 		return;
7629 
7630 	/* check status reg for all VFs owned by this PF */
7631 	for (vf = 0; vf < adapter->num_vfs; ++vf) {
7632 		struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
7633 		u16 status_reg;
7634 
7635 		if (!vfdev)
7636 			continue;
7637 		pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
7638 		if (status_reg != IXGBE_FAILED_READ_CFG_WORD &&
7639 		    status_reg & PCI_STATUS_REC_MASTER_ABORT)
7640 			pcie_flr(vfdev);
7641 	}
7642 }
7643 
7644 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
7645 {
7646 	u32 ssvpc;
7647 
7648 	/* Do not perform spoof check for 82598 or if not in IOV mode */
7649 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7650 	    adapter->num_vfs == 0)
7651 		return;
7652 
7653 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
7654 
7655 	/*
7656 	 * ssvpc register is cleared on read, if zero then no
7657 	 * spoofed packets in the last interval.
7658 	 */
7659 	if (!ssvpc)
7660 		return;
7661 
7662 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
7663 }
7664 #else
7665 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
7666 {
7667 }
7668 
7669 static void
7670 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
7671 {
7672 }
7673 #endif /* CONFIG_PCI_IOV */
7674 
7675 
7676 /**
7677  * ixgbe_watchdog_subtask - check and bring link up
7678  * @adapter: pointer to the device adapter structure
7679  **/
7680 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
7681 {
7682 	/* if interface is down, removing or resetting, do nothing */
7683 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7684 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7685 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7686 		return;
7687 
7688 	ixgbe_watchdog_update_link(adapter);
7689 
7690 	if (adapter->link_up)
7691 		ixgbe_watchdog_link_is_up(adapter);
7692 	else
7693 		ixgbe_watchdog_link_is_down(adapter);
7694 
7695 	ixgbe_check_for_bad_vf(adapter);
7696 	ixgbe_spoof_check(adapter);
7697 	ixgbe_update_stats(adapter);
7698 
7699 	ixgbe_watchdog_flush_tx(adapter);
7700 }
7701 
7702 /**
7703  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
7704  * @adapter: the ixgbe adapter structure
7705  **/
7706 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
7707 {
7708 	struct ixgbe_hw *hw = &adapter->hw;
7709 	s32 err;
7710 
7711 	/* not searching for SFP so there is nothing to do here */
7712 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
7713 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7714 		return;
7715 
7716 	if (adapter->sfp_poll_time &&
7717 	    time_after(adapter->sfp_poll_time, jiffies))
7718 		return; /* If not yet time to poll for SFP */
7719 
7720 	/* someone else is in init, wait until next service event */
7721 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7722 		return;
7723 
7724 	adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
7725 
7726 	err = hw->phy.ops.identify_sfp(hw);
7727 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7728 		goto sfp_out;
7729 
7730 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
7731 		/* If no cable is present, then we need to reset
7732 		 * the next time we find a good cable. */
7733 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
7734 	}
7735 
7736 	/* exit on error */
7737 	if (err)
7738 		goto sfp_out;
7739 
7740 	/* exit if reset not needed */
7741 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7742 		goto sfp_out;
7743 
7744 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
7745 
7746 	/*
7747 	 * A module may be identified correctly, but the EEPROM may not have
7748 	 * support for that module.  setup_sfp() will fail in that case, so
7749 	 * we should not allow that module to load.
7750 	 */
7751 	if (hw->mac.type == ixgbe_mac_82598EB)
7752 		err = hw->phy.ops.reset(hw);
7753 	else
7754 		err = hw->mac.ops.setup_sfp(hw);
7755 
7756 	if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7757 		goto sfp_out;
7758 
7759 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
7760 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
7761 
7762 sfp_out:
7763 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7764 
7765 	if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
7766 	    (adapter->netdev->reg_state == NETREG_REGISTERED)) {
7767 		e_dev_err("failed to initialize because an unsupported "
7768 			  "SFP+ module type was detected.\n");
7769 		e_dev_err("Reload the driver after installing a "
7770 			  "supported module.\n");
7771 		unregister_netdev(adapter->netdev);
7772 	}
7773 }
7774 
7775 /**
7776  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
7777  * @adapter: the ixgbe adapter structure
7778  **/
7779 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
7780 {
7781 	struct ixgbe_hw *hw = &adapter->hw;
7782 	u32 cap_speed;
7783 	u32 speed;
7784 	bool autoneg = false;
7785 
7786 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
7787 		return;
7788 
7789 	/* someone else is in init, wait until next service event */
7790 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7791 		return;
7792 
7793 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
7794 
7795 	hw->mac.ops.get_link_capabilities(hw, &cap_speed, &autoneg);
7796 
7797 	/* advertise highest capable link speed */
7798 	if (!autoneg && (cap_speed & IXGBE_LINK_SPEED_10GB_FULL))
7799 		speed = IXGBE_LINK_SPEED_10GB_FULL;
7800 	else
7801 		speed = cap_speed & (IXGBE_LINK_SPEED_10GB_FULL |
7802 				     IXGBE_LINK_SPEED_1GB_FULL);
7803 
7804 	if (hw->mac.ops.setup_link)
7805 		hw->mac.ops.setup_link(hw, speed, true);
7806 
7807 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
7808 	adapter->link_check_timeout = jiffies;
7809 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7810 }
7811 
7812 /**
7813  * ixgbe_service_timer - Timer Call-back
7814  * @t: pointer to timer_list structure
7815  **/
7816 static void ixgbe_service_timer(struct timer_list *t)
7817 {
7818 	struct ixgbe_adapter *adapter = from_timer(adapter, t, service_timer);
7819 	unsigned long next_event_offset;
7820 
7821 	/* poll faster when waiting for link */
7822 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
7823 		next_event_offset = HZ / 10;
7824 	else
7825 		next_event_offset = HZ * 2;
7826 
7827 	/* Reset the timer */
7828 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
7829 
7830 	ixgbe_service_event_schedule(adapter);
7831 }
7832 
7833 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
7834 {
7835 	struct ixgbe_hw *hw = &adapter->hw;
7836 	u32 status;
7837 
7838 	if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
7839 		return;
7840 
7841 	adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
7842 
7843 	if (!hw->phy.ops.handle_lasi)
7844 		return;
7845 
7846 	status = hw->phy.ops.handle_lasi(&adapter->hw);
7847 	if (status != IXGBE_ERR_OVERTEMP)
7848 		return;
7849 
7850 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
7851 }
7852 
7853 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
7854 {
7855 	if (!test_and_clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state))
7856 		return;
7857 
7858 	rtnl_lock();
7859 	/* If we're already down, removing or resetting, just bail */
7860 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7861 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7862 	    test_bit(__IXGBE_RESETTING, &adapter->state)) {
7863 		rtnl_unlock();
7864 		return;
7865 	}
7866 
7867 	ixgbe_dump(adapter);
7868 	netdev_err(adapter->netdev, "Reset adapter\n");
7869 	adapter->tx_timeout_count++;
7870 
7871 	ixgbe_reinit_locked(adapter);
7872 	rtnl_unlock();
7873 }
7874 
7875 /**
7876  * ixgbe_check_fw_error - Check firmware for errors
7877  * @adapter: the adapter private structure
7878  *
7879  * Check firmware errors in register FWSM
7880  */
7881 static bool ixgbe_check_fw_error(struct ixgbe_adapter *adapter)
7882 {
7883 	struct ixgbe_hw *hw = &adapter->hw;
7884 	u32 fwsm;
7885 
7886 	/* read fwsm.ext_err_ind register and log errors */
7887 	fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
7888 
7889 	if (fwsm & IXGBE_FWSM_EXT_ERR_IND_MASK ||
7890 	    !(fwsm & IXGBE_FWSM_FW_VAL_BIT))
7891 		e_dev_warn("Warning firmware error detected FWSM: 0x%08X\n",
7892 			   fwsm);
7893 
7894 	if (hw->mac.ops.fw_recovery_mode && hw->mac.ops.fw_recovery_mode(hw)) {
7895 		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");
7896 		return true;
7897 	}
7898 
7899 	return false;
7900 }
7901 
7902 /**
7903  * ixgbe_service_task - manages and runs subtasks
7904  * @work: pointer to work_struct containing our data
7905  **/
7906 static void ixgbe_service_task(struct work_struct *work)
7907 {
7908 	struct ixgbe_adapter *adapter = container_of(work,
7909 						     struct ixgbe_adapter,
7910 						     service_task);
7911 	if (ixgbe_removed(adapter->hw.hw_addr)) {
7912 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
7913 			rtnl_lock();
7914 			ixgbe_down(adapter);
7915 			rtnl_unlock();
7916 		}
7917 		ixgbe_service_event_complete(adapter);
7918 		return;
7919 	}
7920 	if (ixgbe_check_fw_error(adapter)) {
7921 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
7922 			unregister_netdev(adapter->netdev);
7923 		ixgbe_service_event_complete(adapter);
7924 		return;
7925 	}
7926 	ixgbe_reset_subtask(adapter);
7927 	ixgbe_phy_interrupt_subtask(adapter);
7928 	ixgbe_sfp_detection_subtask(adapter);
7929 	ixgbe_sfp_link_config_subtask(adapter);
7930 	ixgbe_check_overtemp_subtask(adapter);
7931 	ixgbe_watchdog_subtask(adapter);
7932 	ixgbe_fdir_reinit_subtask(adapter);
7933 	ixgbe_check_hang_subtask(adapter);
7934 
7935 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
7936 		ixgbe_ptp_overflow_check(adapter);
7937 		if (adapter->flags & IXGBE_FLAG_RX_HWTSTAMP_IN_REGISTER)
7938 			ixgbe_ptp_rx_hang(adapter);
7939 		ixgbe_ptp_tx_hang(adapter);
7940 	}
7941 
7942 	ixgbe_service_event_complete(adapter);
7943 }
7944 
7945 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
7946 		     struct ixgbe_tx_buffer *first,
7947 		     u8 *hdr_len,
7948 		     struct ixgbe_ipsec_tx_data *itd)
7949 {
7950 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
7951 	struct sk_buff *skb = first->skb;
7952 	union {
7953 		struct iphdr *v4;
7954 		struct ipv6hdr *v6;
7955 		unsigned char *hdr;
7956 	} ip;
7957 	union {
7958 		struct tcphdr *tcp;
7959 		struct udphdr *udp;
7960 		unsigned char *hdr;
7961 	} l4;
7962 	u32 paylen, l4_offset;
7963 	u32 fceof_saidx = 0;
7964 	int err;
7965 
7966 	if (skb->ip_summed != CHECKSUM_PARTIAL)
7967 		return 0;
7968 
7969 	if (!skb_is_gso(skb))
7970 		return 0;
7971 
7972 	err = skb_cow_head(skb, 0);
7973 	if (err < 0)
7974 		return err;
7975 
7976 	if (eth_p_mpls(first->protocol))
7977 		ip.hdr = skb_inner_network_header(skb);
7978 	else
7979 		ip.hdr = skb_network_header(skb);
7980 	l4.hdr = skb_checksum_start(skb);
7981 
7982 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
7983 	type_tucmd = (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_L4) ?
7984 		      IXGBE_ADVTXD_TUCMD_L4T_UDP : IXGBE_ADVTXD_TUCMD_L4T_TCP;
7985 
7986 	/* initialize outer IP header fields */
7987 	if (ip.v4->version == 4) {
7988 		unsigned char *csum_start = skb_checksum_start(skb);
7989 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
7990 		int len = csum_start - trans_start;
7991 
7992 		/* IP header will have to cancel out any data that
7993 		 * is not a part of the outer IP header, so set to
7994 		 * a reverse csum if needed, else init check to 0.
7995 		 */
7996 		ip.v4->check = (skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL) ?
7997 					   csum_fold(csum_partial(trans_start,
7998 								  len, 0)) : 0;
7999 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
8000 
8001 		ip.v4->tot_len = 0;
8002 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
8003 				   IXGBE_TX_FLAGS_CSUM |
8004 				   IXGBE_TX_FLAGS_IPV4;
8005 	} else {
8006 		ip.v6->payload_len = 0;
8007 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
8008 				   IXGBE_TX_FLAGS_CSUM;
8009 	}
8010 
8011 	/* determine offset of inner transport header */
8012 	l4_offset = l4.hdr - skb->data;
8013 
8014 	/* remove payload length from inner checksum */
8015 	paylen = skb->len - l4_offset;
8016 
8017 	if (type_tucmd & IXGBE_ADVTXD_TUCMD_L4T_TCP) {
8018 		/* compute length of segmentation header */
8019 		*hdr_len = (l4.tcp->doff * 4) + l4_offset;
8020 		csum_replace_by_diff(&l4.tcp->check,
8021 				     (__force __wsum)htonl(paylen));
8022 	} else {
8023 		/* compute length of segmentation header */
8024 		*hdr_len = sizeof(*l4.udp) + l4_offset;
8025 		csum_replace_by_diff(&l4.udp->check,
8026 				     (__force __wsum)htonl(paylen));
8027 	}
8028 
8029 	/* update gso size and bytecount with header size */
8030 	first->gso_segs = skb_shinfo(skb)->gso_segs;
8031 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
8032 
8033 	/* mss_l4len_id: use 0 as index for TSO */
8034 	mss_l4len_idx = (*hdr_len - l4_offset) << IXGBE_ADVTXD_L4LEN_SHIFT;
8035 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
8036 
8037 	fceof_saidx |= itd->sa_idx;
8038 	type_tucmd |= itd->flags | itd->trailer_len;
8039 
8040 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
8041 	vlan_macip_lens = l4.hdr - ip.hdr;
8042 	vlan_macip_lens |= (ip.hdr - skb->data) << IXGBE_ADVTXD_MACLEN_SHIFT;
8043 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
8044 
8045 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd,
8046 			  mss_l4len_idx);
8047 
8048 	return 1;
8049 }
8050 
8051 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
8052 			  struct ixgbe_tx_buffer *first,
8053 			  struct ixgbe_ipsec_tx_data *itd)
8054 {
8055 	struct sk_buff *skb = first->skb;
8056 	u32 vlan_macip_lens = 0;
8057 	u32 fceof_saidx = 0;
8058 	u32 type_tucmd = 0;
8059 
8060 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
8061 csum_failed:
8062 		if (!(first->tx_flags & (IXGBE_TX_FLAGS_HW_VLAN |
8063 					 IXGBE_TX_FLAGS_CC)))
8064 			return;
8065 		goto no_csum;
8066 	}
8067 
8068 	switch (skb->csum_offset) {
8069 	case offsetof(struct tcphdr, check):
8070 		type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
8071 		fallthrough;
8072 	case offsetof(struct udphdr, check):
8073 		break;
8074 	case offsetof(struct sctphdr, checksum):
8075 		/* validate that this is actually an SCTP request */
8076 		if (skb_csum_is_sctp(skb)) {
8077 			type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_SCTP;
8078 			break;
8079 		}
8080 		fallthrough;
8081 	default:
8082 		skb_checksum_help(skb);
8083 		goto csum_failed;
8084 	}
8085 
8086 	/* update TX checksum flag */
8087 	first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
8088 	vlan_macip_lens = skb_checksum_start_offset(skb) -
8089 			  skb_network_offset(skb);
8090 no_csum:
8091 	/* vlan_macip_lens: MACLEN, VLAN tag */
8092 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
8093 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
8094 
8095 	fceof_saidx |= itd->sa_idx;
8096 	type_tucmd |= itd->flags | itd->trailer_len;
8097 
8098 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd, 0);
8099 }
8100 
8101 #define IXGBE_SET_FLAG(_input, _flag, _result) \
8102 	((_flag <= _result) ? \
8103 	 ((u32)(_input & _flag) * (_result / _flag)) : \
8104 	 ((u32)(_input & _flag) / (_flag / _result)))
8105 
8106 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
8107 {
8108 	/* set type for advanced descriptor with frame checksum insertion */
8109 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
8110 		       IXGBE_ADVTXD_DCMD_DEXT |
8111 		       IXGBE_ADVTXD_DCMD_IFCS;
8112 
8113 	/* set HW vlan bit if vlan is present */
8114 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
8115 				   IXGBE_ADVTXD_DCMD_VLE);
8116 
8117 	/* set segmentation enable bits for TSO/FSO */
8118 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
8119 				   IXGBE_ADVTXD_DCMD_TSE);
8120 
8121 	/* set timestamp bit if present */
8122 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
8123 				   IXGBE_ADVTXD_MAC_TSTAMP);
8124 
8125 	/* insert frame checksum */
8126 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
8127 
8128 	return cmd_type;
8129 }
8130 
8131 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
8132 				   u32 tx_flags, unsigned int paylen)
8133 {
8134 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
8135 
8136 	/* enable L4 checksum for TSO and TX checksum offload */
8137 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8138 					IXGBE_TX_FLAGS_CSUM,
8139 					IXGBE_ADVTXD_POPTS_TXSM);
8140 
8141 	/* enable IPv4 checksum for TSO */
8142 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8143 					IXGBE_TX_FLAGS_IPV4,
8144 					IXGBE_ADVTXD_POPTS_IXSM);
8145 
8146 	/* enable IPsec */
8147 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8148 					IXGBE_TX_FLAGS_IPSEC,
8149 					IXGBE_ADVTXD_POPTS_IPSEC);
8150 
8151 	/*
8152 	 * Check Context must be set if Tx switch is enabled, which it
8153 	 * always is for case where virtual functions are running
8154 	 */
8155 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8156 					IXGBE_TX_FLAGS_CC,
8157 					IXGBE_ADVTXD_CC);
8158 
8159 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
8160 }
8161 
8162 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
8163 {
8164 	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
8165 
8166 	/* Herbert's original patch had:
8167 	 *  smp_mb__after_netif_stop_queue();
8168 	 * but since that doesn't exist yet, just open code it.
8169 	 */
8170 	smp_mb();
8171 
8172 	/* We need to check again in a case another CPU has just
8173 	 * made room available.
8174 	 */
8175 	if (likely(ixgbe_desc_unused(tx_ring) < size))
8176 		return -EBUSY;
8177 
8178 	/* A reprieve! - use start_queue because it doesn't call schedule */
8179 	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
8180 	++tx_ring->tx_stats.restart_queue;
8181 	return 0;
8182 }
8183 
8184 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
8185 {
8186 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
8187 		return 0;
8188 
8189 	return __ixgbe_maybe_stop_tx(tx_ring, size);
8190 }
8191 
8192 static int ixgbe_tx_map(struct ixgbe_ring *tx_ring,
8193 			struct ixgbe_tx_buffer *first,
8194 			const u8 hdr_len)
8195 {
8196 	struct sk_buff *skb = first->skb;
8197 	struct ixgbe_tx_buffer *tx_buffer;
8198 	union ixgbe_adv_tx_desc *tx_desc;
8199 	skb_frag_t *frag;
8200 	dma_addr_t dma;
8201 	unsigned int data_len, size;
8202 	u32 tx_flags = first->tx_flags;
8203 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
8204 	u16 i = tx_ring->next_to_use;
8205 
8206 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
8207 
8208 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
8209 
8210 	size = skb_headlen(skb);
8211 	data_len = skb->data_len;
8212 
8213 #ifdef IXGBE_FCOE
8214 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
8215 		if (data_len < sizeof(struct fcoe_crc_eof)) {
8216 			size -= sizeof(struct fcoe_crc_eof) - data_len;
8217 			data_len = 0;
8218 		} else {
8219 			data_len -= sizeof(struct fcoe_crc_eof);
8220 		}
8221 	}
8222 
8223 #endif
8224 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
8225 
8226 	tx_buffer = first;
8227 
8228 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
8229 		if (dma_mapping_error(tx_ring->dev, dma))
8230 			goto dma_error;
8231 
8232 		/* record length, and DMA address */
8233 		dma_unmap_len_set(tx_buffer, len, size);
8234 		dma_unmap_addr_set(tx_buffer, dma, dma);
8235 
8236 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
8237 
8238 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
8239 			tx_desc->read.cmd_type_len =
8240 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
8241 
8242 			i++;
8243 			tx_desc++;
8244 			if (i == tx_ring->count) {
8245 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8246 				i = 0;
8247 			}
8248 			tx_desc->read.olinfo_status = 0;
8249 
8250 			dma += IXGBE_MAX_DATA_PER_TXD;
8251 			size -= IXGBE_MAX_DATA_PER_TXD;
8252 
8253 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
8254 		}
8255 
8256 		if (likely(!data_len))
8257 			break;
8258 
8259 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
8260 
8261 		i++;
8262 		tx_desc++;
8263 		if (i == tx_ring->count) {
8264 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8265 			i = 0;
8266 		}
8267 		tx_desc->read.olinfo_status = 0;
8268 
8269 #ifdef IXGBE_FCOE
8270 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
8271 #else
8272 		size = skb_frag_size(frag);
8273 #endif
8274 		data_len -= size;
8275 
8276 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
8277 				       DMA_TO_DEVICE);
8278 
8279 		tx_buffer = &tx_ring->tx_buffer_info[i];
8280 	}
8281 
8282 	/* write last descriptor with RS and EOP bits */
8283 	cmd_type |= size | IXGBE_TXD_CMD;
8284 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8285 
8286 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
8287 
8288 	/* set the timestamp */
8289 	first->time_stamp = jiffies;
8290 
8291 	skb_tx_timestamp(skb);
8292 
8293 	/*
8294 	 * Force memory writes to complete before letting h/w know there
8295 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
8296 	 * memory model archs, such as IA-64).
8297 	 *
8298 	 * We also need this memory barrier to make certain all of the
8299 	 * status bits have been updated before next_to_watch is written.
8300 	 */
8301 	wmb();
8302 
8303 	/* set next_to_watch value indicating a packet is present */
8304 	first->next_to_watch = tx_desc;
8305 
8306 	i++;
8307 	if (i == tx_ring->count)
8308 		i = 0;
8309 
8310 	tx_ring->next_to_use = i;
8311 
8312 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
8313 
8314 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more()) {
8315 		writel(i, tx_ring->tail);
8316 	}
8317 
8318 	return 0;
8319 dma_error:
8320 	dev_err(tx_ring->dev, "TX DMA map failed\n");
8321 
8322 	/* clear dma mappings for failed tx_buffer_info map */
8323 	for (;;) {
8324 		tx_buffer = &tx_ring->tx_buffer_info[i];
8325 		if (dma_unmap_len(tx_buffer, len))
8326 			dma_unmap_page(tx_ring->dev,
8327 				       dma_unmap_addr(tx_buffer, dma),
8328 				       dma_unmap_len(tx_buffer, len),
8329 				       DMA_TO_DEVICE);
8330 		dma_unmap_len_set(tx_buffer, len, 0);
8331 		if (tx_buffer == first)
8332 			break;
8333 		if (i == 0)
8334 			i += tx_ring->count;
8335 		i--;
8336 	}
8337 
8338 	dev_kfree_skb_any(first->skb);
8339 	first->skb = NULL;
8340 
8341 	tx_ring->next_to_use = i;
8342 
8343 	return -1;
8344 }
8345 
8346 static void ixgbe_atr(struct ixgbe_ring *ring,
8347 		      struct ixgbe_tx_buffer *first)
8348 {
8349 	struct ixgbe_q_vector *q_vector = ring->q_vector;
8350 	union ixgbe_atr_hash_dword input = { .dword = 0 };
8351 	union ixgbe_atr_hash_dword common = { .dword = 0 };
8352 	union {
8353 		unsigned char *network;
8354 		struct iphdr *ipv4;
8355 		struct ipv6hdr *ipv6;
8356 	} hdr;
8357 	struct tcphdr *th;
8358 	unsigned int hlen;
8359 	struct sk_buff *skb;
8360 	__be16 vlan_id;
8361 	int l4_proto;
8362 
8363 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
8364 	if (!q_vector)
8365 		return;
8366 
8367 	/* do nothing if sampling is disabled */
8368 	if (!ring->atr_sample_rate)
8369 		return;
8370 
8371 	ring->atr_count++;
8372 
8373 	/* currently only IPv4/IPv6 with TCP is supported */
8374 	if ((first->protocol != htons(ETH_P_IP)) &&
8375 	    (first->protocol != htons(ETH_P_IPV6)))
8376 		return;
8377 
8378 	/* snag network header to get L4 type and address */
8379 	skb = first->skb;
8380 	hdr.network = skb_network_header(skb);
8381 	if (unlikely(hdr.network <= skb->data))
8382 		return;
8383 	if (skb->encapsulation &&
8384 	    first->protocol == htons(ETH_P_IP) &&
8385 	    hdr.ipv4->protocol == IPPROTO_UDP) {
8386 		struct ixgbe_adapter *adapter = q_vector->adapter;
8387 
8388 		if (unlikely(skb_tail_pointer(skb) < hdr.network +
8389 			     VXLAN_HEADROOM))
8390 			return;
8391 
8392 		/* verify the port is recognized as VXLAN */
8393 		if (adapter->vxlan_port &&
8394 		    udp_hdr(skb)->dest == adapter->vxlan_port)
8395 			hdr.network = skb_inner_network_header(skb);
8396 
8397 		if (adapter->geneve_port &&
8398 		    udp_hdr(skb)->dest == adapter->geneve_port)
8399 			hdr.network = skb_inner_network_header(skb);
8400 	}
8401 
8402 	/* Make sure we have at least [minimum IPv4 header + TCP]
8403 	 * or [IPv6 header] bytes
8404 	 */
8405 	if (unlikely(skb_tail_pointer(skb) < hdr.network + 40))
8406 		return;
8407 
8408 	/* Currently only IPv4/IPv6 with TCP is supported */
8409 	switch (hdr.ipv4->version) {
8410 	case IPVERSION:
8411 		/* access ihl as u8 to avoid unaligned access on ia64 */
8412 		hlen = (hdr.network[0] & 0x0F) << 2;
8413 		l4_proto = hdr.ipv4->protocol;
8414 		break;
8415 	case 6:
8416 		hlen = hdr.network - skb->data;
8417 		l4_proto = ipv6_find_hdr(skb, &hlen, IPPROTO_TCP, NULL, NULL);
8418 		hlen -= hdr.network - skb->data;
8419 		break;
8420 	default:
8421 		return;
8422 	}
8423 
8424 	if (l4_proto != IPPROTO_TCP)
8425 		return;
8426 
8427 	if (unlikely(skb_tail_pointer(skb) < hdr.network +
8428 		     hlen + sizeof(struct tcphdr)))
8429 		return;
8430 
8431 	th = (struct tcphdr *)(hdr.network + hlen);
8432 
8433 	/* skip this packet since the socket is closing */
8434 	if (th->fin)
8435 		return;
8436 
8437 	/* sample on all syn packets or once every atr sample count */
8438 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
8439 		return;
8440 
8441 	/* reset sample count */
8442 	ring->atr_count = 0;
8443 
8444 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
8445 
8446 	/*
8447 	 * src and dst are inverted, think how the receiver sees them
8448 	 *
8449 	 * The input is broken into two sections, a non-compressed section
8450 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
8451 	 * is XORed together and stored in the compressed dword.
8452 	 */
8453 	input.formatted.vlan_id = vlan_id;
8454 
8455 	/*
8456 	 * since src port and flex bytes occupy the same word XOR them together
8457 	 * and write the value to source port portion of compressed dword
8458 	 */
8459 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
8460 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
8461 	else
8462 		common.port.src ^= th->dest ^ first->protocol;
8463 	common.port.dst ^= th->source;
8464 
8465 	switch (hdr.ipv4->version) {
8466 	case IPVERSION:
8467 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
8468 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
8469 		break;
8470 	case 6:
8471 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
8472 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
8473 			     hdr.ipv6->saddr.s6_addr32[1] ^
8474 			     hdr.ipv6->saddr.s6_addr32[2] ^
8475 			     hdr.ipv6->saddr.s6_addr32[3] ^
8476 			     hdr.ipv6->daddr.s6_addr32[0] ^
8477 			     hdr.ipv6->daddr.s6_addr32[1] ^
8478 			     hdr.ipv6->daddr.s6_addr32[2] ^
8479 			     hdr.ipv6->daddr.s6_addr32[3];
8480 		break;
8481 	default:
8482 		break;
8483 	}
8484 
8485 	if (hdr.network != skb_network_header(skb))
8486 		input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
8487 
8488 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
8489 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
8490 					      input, common, ring->queue_index);
8491 }
8492 
8493 #ifdef IXGBE_FCOE
8494 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
8495 			      struct net_device *sb_dev)
8496 {
8497 	struct ixgbe_adapter *adapter;
8498 	struct ixgbe_ring_feature *f;
8499 	int txq;
8500 
8501 	if (sb_dev) {
8502 		u8 tc = netdev_get_prio_tc_map(dev, skb->priority);
8503 		struct net_device *vdev = sb_dev;
8504 
8505 		txq = vdev->tc_to_txq[tc].offset;
8506 		txq += reciprocal_scale(skb_get_hash(skb),
8507 					vdev->tc_to_txq[tc].count);
8508 
8509 		return txq;
8510 	}
8511 
8512 	/*
8513 	 * only execute the code below if protocol is FCoE
8514 	 * or FIP and we have FCoE enabled on the adapter
8515 	 */
8516 	switch (vlan_get_protocol(skb)) {
8517 	case htons(ETH_P_FCOE):
8518 	case htons(ETH_P_FIP):
8519 		adapter = netdev_priv(dev);
8520 
8521 		if (!sb_dev && (adapter->flags & IXGBE_FLAG_FCOE_ENABLED))
8522 			break;
8523 		fallthrough;
8524 	default:
8525 		return netdev_pick_tx(dev, skb, sb_dev);
8526 	}
8527 
8528 	f = &adapter->ring_feature[RING_F_FCOE];
8529 
8530 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
8531 					   smp_processor_id();
8532 
8533 	while (txq >= f->indices)
8534 		txq -= f->indices;
8535 
8536 	return txq + f->offset;
8537 }
8538 
8539 #endif
8540 int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
8541 			struct xdp_frame *xdpf)
8542 {
8543 	struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
8544 	struct ixgbe_tx_buffer *tx_buffer;
8545 	union ixgbe_adv_tx_desc *tx_desc;
8546 	u32 len, cmd_type;
8547 	dma_addr_t dma;
8548 	u16 i;
8549 
8550 	len = xdpf->len;
8551 
8552 	if (unlikely(!ixgbe_desc_unused(ring)))
8553 		return IXGBE_XDP_CONSUMED;
8554 
8555 	dma = dma_map_single(ring->dev, xdpf->data, len, DMA_TO_DEVICE);
8556 	if (dma_mapping_error(ring->dev, dma))
8557 		return IXGBE_XDP_CONSUMED;
8558 
8559 	/* record the location of the first descriptor for this packet */
8560 	tx_buffer = &ring->tx_buffer_info[ring->next_to_use];
8561 	tx_buffer->bytecount = len;
8562 	tx_buffer->gso_segs = 1;
8563 	tx_buffer->protocol = 0;
8564 
8565 	i = ring->next_to_use;
8566 	tx_desc = IXGBE_TX_DESC(ring, i);
8567 
8568 	dma_unmap_len_set(tx_buffer, len, len);
8569 	dma_unmap_addr_set(tx_buffer, dma, dma);
8570 	tx_buffer->xdpf = xdpf;
8571 
8572 	tx_desc->read.buffer_addr = cpu_to_le64(dma);
8573 
8574 	/* put descriptor type bits */
8575 	cmd_type = IXGBE_ADVTXD_DTYP_DATA |
8576 		   IXGBE_ADVTXD_DCMD_DEXT |
8577 		   IXGBE_ADVTXD_DCMD_IFCS;
8578 	cmd_type |= len | IXGBE_TXD_CMD;
8579 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8580 	tx_desc->read.olinfo_status =
8581 		cpu_to_le32(len << IXGBE_ADVTXD_PAYLEN_SHIFT);
8582 
8583 	/* Avoid any potential race with xdp_xmit and cleanup */
8584 	smp_wmb();
8585 
8586 	/* set next_to_watch value indicating a packet is present */
8587 	i++;
8588 	if (i == ring->count)
8589 		i = 0;
8590 
8591 	tx_buffer->next_to_watch = tx_desc;
8592 	ring->next_to_use = i;
8593 
8594 	return IXGBE_XDP_TX;
8595 }
8596 
8597 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
8598 			  struct ixgbe_adapter *adapter,
8599 			  struct ixgbe_ring *tx_ring)
8600 {
8601 	struct ixgbe_tx_buffer *first;
8602 	int tso;
8603 	u32 tx_flags = 0;
8604 	unsigned short f;
8605 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
8606 	struct ixgbe_ipsec_tx_data ipsec_tx = { 0 };
8607 	__be16 protocol = skb->protocol;
8608 	u8 hdr_len = 0;
8609 
8610 	/*
8611 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
8612 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
8613 	 *       + 2 desc gap to keep tail from touching head,
8614 	 *       + 1 desc for context descriptor,
8615 	 * otherwise try next time
8616 	 */
8617 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
8618 		count += TXD_USE_COUNT(skb_frag_size(
8619 						&skb_shinfo(skb)->frags[f]));
8620 
8621 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
8622 		tx_ring->tx_stats.tx_busy++;
8623 		return NETDEV_TX_BUSY;
8624 	}
8625 
8626 	/* record the location of the first descriptor for this packet */
8627 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
8628 	first->skb = skb;
8629 	first->bytecount = skb->len;
8630 	first->gso_segs = 1;
8631 
8632 	/* if we have a HW VLAN tag being added default to the HW one */
8633 	if (skb_vlan_tag_present(skb)) {
8634 		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
8635 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8636 	/* else if it is a SW VLAN check the next protocol and store the tag */
8637 	} else if (protocol == htons(ETH_P_8021Q)) {
8638 		struct vlan_hdr *vhdr, _vhdr;
8639 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
8640 		if (!vhdr)
8641 			goto out_drop;
8642 
8643 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
8644 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
8645 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
8646 	}
8647 	protocol = vlan_get_protocol(skb);
8648 
8649 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
8650 	    adapter->ptp_clock) {
8651 		if (adapter->tstamp_config.tx_type == HWTSTAMP_TX_ON &&
8652 		    !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
8653 					   &adapter->state)) {
8654 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
8655 			tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
8656 
8657 			/* schedule check for Tx timestamp */
8658 			adapter->ptp_tx_skb = skb_get(skb);
8659 			adapter->ptp_tx_start = jiffies;
8660 			schedule_work(&adapter->ptp_tx_work);
8661 		} else {
8662 			adapter->tx_hwtstamp_skipped++;
8663 		}
8664 	}
8665 
8666 #ifdef CONFIG_PCI_IOV
8667 	/*
8668 	 * Use the l2switch_enable flag - would be false if the DMA
8669 	 * Tx switch had been disabled.
8670 	 */
8671 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
8672 		tx_flags |= IXGBE_TX_FLAGS_CC;
8673 
8674 #endif
8675 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
8676 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
8677 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
8678 	     (skb->priority != TC_PRIO_CONTROL))) {
8679 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
8680 		tx_flags |= (skb->priority & 0x7) <<
8681 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
8682 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
8683 			struct vlan_ethhdr *vhdr;
8684 
8685 			if (skb_cow_head(skb, 0))
8686 				goto out_drop;
8687 			vhdr = (struct vlan_ethhdr *)skb->data;
8688 			vhdr->h_vlan_TCI = htons(tx_flags >>
8689 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
8690 		} else {
8691 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8692 		}
8693 	}
8694 
8695 	/* record initial flags and protocol */
8696 	first->tx_flags = tx_flags;
8697 	first->protocol = protocol;
8698 
8699 #ifdef IXGBE_FCOE
8700 	/* setup tx offload for FCoE */
8701 	if ((protocol == htons(ETH_P_FCOE)) &&
8702 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
8703 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
8704 		if (tso < 0)
8705 			goto out_drop;
8706 
8707 		goto xmit_fcoe;
8708 	}
8709 
8710 #endif /* IXGBE_FCOE */
8711 
8712 #ifdef CONFIG_IXGBE_IPSEC
8713 	if (xfrm_offload(skb) &&
8714 	    !ixgbe_ipsec_tx(tx_ring, first, &ipsec_tx))
8715 		goto out_drop;
8716 #endif
8717 	tso = ixgbe_tso(tx_ring, first, &hdr_len, &ipsec_tx);
8718 	if (tso < 0)
8719 		goto out_drop;
8720 	else if (!tso)
8721 		ixgbe_tx_csum(tx_ring, first, &ipsec_tx);
8722 
8723 	/* add the ATR filter if ATR is on */
8724 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
8725 		ixgbe_atr(tx_ring, first);
8726 
8727 #ifdef IXGBE_FCOE
8728 xmit_fcoe:
8729 #endif /* IXGBE_FCOE */
8730 	if (ixgbe_tx_map(tx_ring, first, hdr_len))
8731 		goto cleanup_tx_timestamp;
8732 
8733 	return NETDEV_TX_OK;
8734 
8735 out_drop:
8736 	dev_kfree_skb_any(first->skb);
8737 	first->skb = NULL;
8738 cleanup_tx_timestamp:
8739 	if (unlikely(tx_flags & IXGBE_TX_FLAGS_TSTAMP)) {
8740 		dev_kfree_skb_any(adapter->ptp_tx_skb);
8741 		adapter->ptp_tx_skb = NULL;
8742 		cancel_work_sync(&adapter->ptp_tx_work);
8743 		clear_bit_unlock(__IXGBE_PTP_TX_IN_PROGRESS, &adapter->state);
8744 	}
8745 
8746 	return NETDEV_TX_OK;
8747 }
8748 
8749 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
8750 				      struct net_device *netdev,
8751 				      struct ixgbe_ring *ring)
8752 {
8753 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8754 	struct ixgbe_ring *tx_ring;
8755 
8756 	/*
8757 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
8758 	 * in order to meet this minimum size requirement.
8759 	 */
8760 	if (skb_put_padto(skb, 17))
8761 		return NETDEV_TX_OK;
8762 
8763 	tx_ring = ring ? ring : adapter->tx_ring[skb_get_queue_mapping(skb)];
8764 	if (unlikely(test_bit(__IXGBE_TX_DISABLED, &tx_ring->state)))
8765 		return NETDEV_TX_BUSY;
8766 
8767 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
8768 }
8769 
8770 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
8771 				    struct net_device *netdev)
8772 {
8773 	return __ixgbe_xmit_frame(skb, netdev, NULL);
8774 }
8775 
8776 /**
8777  * ixgbe_set_mac - Change the Ethernet Address of the NIC
8778  * @netdev: network interface device structure
8779  * @p: pointer to an address structure
8780  *
8781  * Returns 0 on success, negative on failure
8782  **/
8783 static int ixgbe_set_mac(struct net_device *netdev, void *p)
8784 {
8785 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8786 	struct ixgbe_hw *hw = &adapter->hw;
8787 	struct sockaddr *addr = p;
8788 
8789 	if (!is_valid_ether_addr(addr->sa_data))
8790 		return -EADDRNOTAVAIL;
8791 
8792 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
8793 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
8794 
8795 	ixgbe_mac_set_default_filter(adapter);
8796 
8797 	return 0;
8798 }
8799 
8800 static int
8801 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
8802 {
8803 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8804 	struct ixgbe_hw *hw = &adapter->hw;
8805 	u16 value;
8806 	int rc;
8807 
8808 	if (adapter->mii_bus) {
8809 		int regnum = addr;
8810 
8811 		if (devad != MDIO_DEVAD_NONE)
8812 			regnum |= (devad << 16) | MII_ADDR_C45;
8813 
8814 		return mdiobus_read(adapter->mii_bus, prtad, regnum);
8815 	}
8816 
8817 	if (prtad != hw->phy.mdio.prtad)
8818 		return -EINVAL;
8819 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
8820 	if (!rc)
8821 		rc = value;
8822 	return rc;
8823 }
8824 
8825 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
8826 			    u16 addr, u16 value)
8827 {
8828 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8829 	struct ixgbe_hw *hw = &adapter->hw;
8830 
8831 	if (adapter->mii_bus) {
8832 		int regnum = addr;
8833 
8834 		if (devad != MDIO_DEVAD_NONE)
8835 			regnum |= (devad << 16) | MII_ADDR_C45;
8836 
8837 		return mdiobus_write(adapter->mii_bus, prtad, regnum, value);
8838 	}
8839 
8840 	if (prtad != hw->phy.mdio.prtad)
8841 		return -EINVAL;
8842 	return hw->phy.ops.write_reg(hw, addr, devad, value);
8843 }
8844 
8845 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
8846 {
8847 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8848 
8849 	switch (cmd) {
8850 	case SIOCSHWTSTAMP:
8851 		return ixgbe_ptp_set_ts_config(adapter, req);
8852 	case SIOCGHWTSTAMP:
8853 		return ixgbe_ptp_get_ts_config(adapter, req);
8854 	case SIOCGMIIPHY:
8855 		if (!adapter->hw.phy.ops.read_reg)
8856 			return -EOPNOTSUPP;
8857 		fallthrough;
8858 	default:
8859 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
8860 	}
8861 }
8862 
8863 /**
8864  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
8865  * netdev->dev_addrs
8866  * @dev: network interface device structure
8867  *
8868  * Returns non-zero on failure
8869  **/
8870 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
8871 {
8872 	int err = 0;
8873 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8874 	struct ixgbe_hw *hw = &adapter->hw;
8875 
8876 	if (is_valid_ether_addr(hw->mac.san_addr)) {
8877 		rtnl_lock();
8878 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
8879 		rtnl_unlock();
8880 
8881 		/* update SAN MAC vmdq pool selection */
8882 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
8883 	}
8884 	return err;
8885 }
8886 
8887 /**
8888  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
8889  * netdev->dev_addrs
8890  * @dev: network interface device structure
8891  *
8892  * Returns non-zero on failure
8893  **/
8894 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
8895 {
8896 	int err = 0;
8897 	struct ixgbe_adapter *adapter = netdev_priv(dev);
8898 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
8899 
8900 	if (is_valid_ether_addr(mac->san_addr)) {
8901 		rtnl_lock();
8902 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
8903 		rtnl_unlock();
8904 	}
8905 	return err;
8906 }
8907 
8908 static void ixgbe_get_ring_stats64(struct rtnl_link_stats64 *stats,
8909 				   struct ixgbe_ring *ring)
8910 {
8911 	u64 bytes, packets;
8912 	unsigned int start;
8913 
8914 	if (ring) {
8915 		do {
8916 			start = u64_stats_fetch_begin_irq(&ring->syncp);
8917 			packets = ring->stats.packets;
8918 			bytes   = ring->stats.bytes;
8919 		} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8920 		stats->tx_packets += packets;
8921 		stats->tx_bytes   += bytes;
8922 	}
8923 }
8924 
8925 static void ixgbe_get_stats64(struct net_device *netdev,
8926 			      struct rtnl_link_stats64 *stats)
8927 {
8928 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
8929 	int i;
8930 
8931 	rcu_read_lock();
8932 	for (i = 0; i < adapter->num_rx_queues; i++) {
8933 		struct ixgbe_ring *ring = READ_ONCE(adapter->rx_ring[i]);
8934 		u64 bytes, packets;
8935 		unsigned int start;
8936 
8937 		if (ring) {
8938 			do {
8939 				start = u64_stats_fetch_begin_irq(&ring->syncp);
8940 				packets = ring->stats.packets;
8941 				bytes   = ring->stats.bytes;
8942 			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8943 			stats->rx_packets += packets;
8944 			stats->rx_bytes   += bytes;
8945 		}
8946 	}
8947 
8948 	for (i = 0; i < adapter->num_tx_queues; i++) {
8949 		struct ixgbe_ring *ring = READ_ONCE(adapter->tx_ring[i]);
8950 
8951 		ixgbe_get_ring_stats64(stats, ring);
8952 	}
8953 	for (i = 0; i < adapter->num_xdp_queues; i++) {
8954 		struct ixgbe_ring *ring = READ_ONCE(adapter->xdp_ring[i]);
8955 
8956 		ixgbe_get_ring_stats64(stats, ring);
8957 	}
8958 	rcu_read_unlock();
8959 
8960 	/* following stats updated by ixgbe_watchdog_task() */
8961 	stats->multicast	= netdev->stats.multicast;
8962 	stats->rx_errors	= netdev->stats.rx_errors;
8963 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
8964 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
8965 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
8966 }
8967 
8968 #ifdef CONFIG_IXGBE_DCB
8969 /**
8970  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
8971  * @adapter: pointer to ixgbe_adapter
8972  * @tc: number of traffic classes currently enabled
8973  *
8974  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
8975  * 802.1Q priority maps to a packet buffer that exists.
8976  */
8977 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
8978 {
8979 	struct ixgbe_hw *hw = &adapter->hw;
8980 	u32 reg, rsave;
8981 	int i;
8982 
8983 	/* 82598 have a static priority to TC mapping that can not
8984 	 * be changed so no validation is needed.
8985 	 */
8986 	if (hw->mac.type == ixgbe_mac_82598EB)
8987 		return;
8988 
8989 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
8990 	rsave = reg;
8991 
8992 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
8993 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
8994 
8995 		/* If up2tc is out of bounds default to zero */
8996 		if (up2tc > tc)
8997 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
8998 	}
8999 
9000 	if (reg != rsave)
9001 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
9002 
9003 	return;
9004 }
9005 
9006 /**
9007  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
9008  * @adapter: Pointer to adapter struct
9009  *
9010  * Populate the netdev user priority to tc map
9011  */
9012 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
9013 {
9014 	struct net_device *dev = adapter->netdev;
9015 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
9016 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
9017 	u8 prio;
9018 
9019 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
9020 		u8 tc = 0;
9021 
9022 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
9023 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
9024 		else if (ets)
9025 			tc = ets->prio_tc[prio];
9026 
9027 		netdev_set_prio_tc_map(dev, prio, tc);
9028 	}
9029 }
9030 
9031 #endif /* CONFIG_IXGBE_DCB */
9032 static int ixgbe_reassign_macvlan_pool(struct net_device *vdev,
9033 				       struct netdev_nested_priv *priv)
9034 {
9035 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)priv->data;
9036 	struct ixgbe_fwd_adapter *accel;
9037 	int pool;
9038 
9039 	/* we only care about macvlans... */
9040 	if (!netif_is_macvlan(vdev))
9041 		return 0;
9042 
9043 	/* that have hardware offload enabled... */
9044 	accel = macvlan_accel_priv(vdev);
9045 	if (!accel)
9046 		return 0;
9047 
9048 	/* If we can relocate to a different bit do so */
9049 	pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
9050 	if (pool < adapter->num_rx_pools) {
9051 		set_bit(pool, adapter->fwd_bitmask);
9052 		accel->pool = pool;
9053 		return 0;
9054 	}
9055 
9056 	/* if we cannot find a free pool then disable the offload */
9057 	netdev_err(vdev, "L2FW offload disabled due to lack of queue resources\n");
9058 	macvlan_release_l2fw_offload(vdev);
9059 
9060 	/* unbind the queues and drop the subordinate channel config */
9061 	netdev_unbind_sb_channel(adapter->netdev, vdev);
9062 	netdev_set_sb_channel(vdev, 0);
9063 
9064 	kfree(accel);
9065 
9066 	return 0;
9067 }
9068 
9069 static void ixgbe_defrag_macvlan_pools(struct net_device *dev)
9070 {
9071 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9072 	struct netdev_nested_priv priv = {
9073 		.data = (void *)adapter,
9074 	};
9075 
9076 	/* flush any stale bits out of the fwd bitmask */
9077 	bitmap_clear(adapter->fwd_bitmask, 1, 63);
9078 
9079 	/* walk through upper devices reassigning pools */
9080 	netdev_walk_all_upper_dev_rcu(dev, ixgbe_reassign_macvlan_pool,
9081 				      &priv);
9082 }
9083 
9084 /**
9085  * ixgbe_setup_tc - configure net_device for multiple traffic classes
9086  *
9087  * @dev: net device to configure
9088  * @tc: number of traffic classes to enable
9089  */
9090 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
9091 {
9092 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9093 	struct ixgbe_hw *hw = &adapter->hw;
9094 
9095 	/* Hardware supports up to 8 traffic classes */
9096 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
9097 		return -EINVAL;
9098 
9099 	if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
9100 		return -EINVAL;
9101 
9102 	/* Hardware has to reinitialize queues and interrupts to
9103 	 * match packet buffer alignment. Unfortunately, the
9104 	 * hardware is not flexible enough to do this dynamically.
9105 	 */
9106 	if (netif_running(dev))
9107 		ixgbe_close(dev);
9108 	else
9109 		ixgbe_reset(adapter);
9110 
9111 	ixgbe_clear_interrupt_scheme(adapter);
9112 
9113 #ifdef CONFIG_IXGBE_DCB
9114 	if (tc) {
9115 		if (adapter->xdp_prog) {
9116 			e_warn(probe, "DCB is not supported with XDP\n");
9117 
9118 			ixgbe_init_interrupt_scheme(adapter);
9119 			if (netif_running(dev))
9120 				ixgbe_open(dev);
9121 			return -EINVAL;
9122 		}
9123 
9124 		netdev_set_num_tc(dev, tc);
9125 		ixgbe_set_prio_tc_map(adapter);
9126 
9127 		adapter->hw_tcs = tc;
9128 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
9129 
9130 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
9131 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
9132 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
9133 		}
9134 	} else {
9135 		netdev_reset_tc(dev);
9136 
9137 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
9138 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
9139 
9140 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
9141 		adapter->hw_tcs = tc;
9142 
9143 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
9144 		adapter->dcb_cfg.pfc_mode_enable = false;
9145 	}
9146 
9147 	ixgbe_validate_rtr(adapter, tc);
9148 
9149 #endif /* CONFIG_IXGBE_DCB */
9150 	ixgbe_init_interrupt_scheme(adapter);
9151 
9152 	ixgbe_defrag_macvlan_pools(dev);
9153 
9154 	if (netif_running(dev))
9155 		return ixgbe_open(dev);
9156 
9157 	return 0;
9158 }
9159 
9160 static int ixgbe_delete_clsu32(struct ixgbe_adapter *adapter,
9161 			       struct tc_cls_u32_offload *cls)
9162 {
9163 	u32 hdl = cls->knode.handle;
9164 	u32 uhtid = TC_U32_USERHTID(cls->knode.handle);
9165 	u32 loc = cls->knode.handle & 0xfffff;
9166 	int err = 0, i, j;
9167 	struct ixgbe_jump_table *jump = NULL;
9168 
9169 	if (loc > IXGBE_MAX_HW_ENTRIES)
9170 		return -EINVAL;
9171 
9172 	if ((uhtid != 0x800) && (uhtid >= IXGBE_MAX_LINK_HANDLE))
9173 		return -EINVAL;
9174 
9175 	/* Clear this filter in the link data it is associated with */
9176 	if (uhtid != 0x800) {
9177 		jump = adapter->jump_tables[uhtid];
9178 		if (!jump)
9179 			return -EINVAL;
9180 		if (!test_bit(loc - 1, jump->child_loc_map))
9181 			return -EINVAL;
9182 		clear_bit(loc - 1, jump->child_loc_map);
9183 	}
9184 
9185 	/* Check if the filter being deleted is a link */
9186 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
9187 		jump = adapter->jump_tables[i];
9188 		if (jump && jump->link_hdl == hdl) {
9189 			/* Delete filters in the hardware in the child hash
9190 			 * table associated with this link
9191 			 */
9192 			for (j = 0; j < IXGBE_MAX_HW_ENTRIES; j++) {
9193 				if (!test_bit(j, jump->child_loc_map))
9194 					continue;
9195 				spin_lock(&adapter->fdir_perfect_lock);
9196 				err = ixgbe_update_ethtool_fdir_entry(adapter,
9197 								      NULL,
9198 								      j + 1);
9199 				spin_unlock(&adapter->fdir_perfect_lock);
9200 				clear_bit(j, jump->child_loc_map);
9201 			}
9202 			/* Remove resources for this link */
9203 			kfree(jump->input);
9204 			kfree(jump->mask);
9205 			kfree(jump);
9206 			adapter->jump_tables[i] = NULL;
9207 			return err;
9208 		}
9209 	}
9210 
9211 	spin_lock(&adapter->fdir_perfect_lock);
9212 	err = ixgbe_update_ethtool_fdir_entry(adapter, NULL, loc);
9213 	spin_unlock(&adapter->fdir_perfect_lock);
9214 	return err;
9215 }
9216 
9217 static int ixgbe_configure_clsu32_add_hnode(struct ixgbe_adapter *adapter,
9218 					    struct tc_cls_u32_offload *cls)
9219 {
9220 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
9221 
9222 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9223 		return -EINVAL;
9224 
9225 	/* This ixgbe devices do not support hash tables at the moment
9226 	 * so abort when given hash tables.
9227 	 */
9228 	if (cls->hnode.divisor > 0)
9229 		return -EINVAL;
9230 
9231 	set_bit(uhtid - 1, &adapter->tables);
9232 	return 0;
9233 }
9234 
9235 static int ixgbe_configure_clsu32_del_hnode(struct ixgbe_adapter *adapter,
9236 					    struct tc_cls_u32_offload *cls)
9237 {
9238 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
9239 
9240 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9241 		return -EINVAL;
9242 
9243 	clear_bit(uhtid - 1, &adapter->tables);
9244 	return 0;
9245 }
9246 
9247 #ifdef CONFIG_NET_CLS_ACT
9248 struct upper_walk_data {
9249 	struct ixgbe_adapter *adapter;
9250 	u64 action;
9251 	int ifindex;
9252 	u8 queue;
9253 };
9254 
9255 static int get_macvlan_queue(struct net_device *upper,
9256 			     struct netdev_nested_priv *priv)
9257 {
9258 	if (netif_is_macvlan(upper)) {
9259 		struct ixgbe_fwd_adapter *vadapter = macvlan_accel_priv(upper);
9260 		struct ixgbe_adapter *adapter;
9261 		struct upper_walk_data *data;
9262 		int ifindex;
9263 
9264 		data = (struct upper_walk_data *)priv->data;
9265 		ifindex = data->ifindex;
9266 		adapter = data->adapter;
9267 		if (vadapter && upper->ifindex == ifindex) {
9268 			data->queue = adapter->rx_ring[vadapter->rx_base_queue]->reg_idx;
9269 			data->action = data->queue;
9270 			return 1;
9271 		}
9272 	}
9273 
9274 	return 0;
9275 }
9276 
9277 static int handle_redirect_action(struct ixgbe_adapter *adapter, int ifindex,
9278 				  u8 *queue, u64 *action)
9279 {
9280 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
9281 	unsigned int num_vfs = adapter->num_vfs, vf;
9282 	struct netdev_nested_priv priv;
9283 	struct upper_walk_data data;
9284 	struct net_device *upper;
9285 
9286 	/* redirect to a SRIOV VF */
9287 	for (vf = 0; vf < num_vfs; ++vf) {
9288 		upper = pci_get_drvdata(adapter->vfinfo[vf].vfdev);
9289 		if (upper->ifindex == ifindex) {
9290 			*queue = vf * __ALIGN_MASK(1, ~vmdq->mask);
9291 			*action = vf + 1;
9292 			*action <<= ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF;
9293 			return 0;
9294 		}
9295 	}
9296 
9297 	/* redirect to a offloaded macvlan netdev */
9298 	data.adapter = adapter;
9299 	data.ifindex = ifindex;
9300 	data.action = 0;
9301 	data.queue = 0;
9302 	priv.data = (void *)&data;
9303 	if (netdev_walk_all_upper_dev_rcu(adapter->netdev,
9304 					  get_macvlan_queue, &priv)) {
9305 		*action = data.action;
9306 		*queue = data.queue;
9307 
9308 		return 0;
9309 	}
9310 
9311 	return -EINVAL;
9312 }
9313 
9314 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9315 			    struct tcf_exts *exts, u64 *action, u8 *queue)
9316 {
9317 	const struct tc_action *a;
9318 	int i;
9319 
9320 	if (!tcf_exts_has_actions(exts))
9321 		return -EINVAL;
9322 
9323 	tcf_exts_for_each_action(i, a, exts) {
9324 		/* Drop action */
9325 		if (is_tcf_gact_shot(a)) {
9326 			*action = IXGBE_FDIR_DROP_QUEUE;
9327 			*queue = IXGBE_FDIR_DROP_QUEUE;
9328 			return 0;
9329 		}
9330 
9331 		/* Redirect to a VF or a offloaded macvlan */
9332 		if (is_tcf_mirred_egress_redirect(a)) {
9333 			struct net_device *dev = tcf_mirred_dev(a);
9334 
9335 			if (!dev)
9336 				return -EINVAL;
9337 			return handle_redirect_action(adapter, dev->ifindex,
9338 						      queue, action);
9339 		}
9340 
9341 		return -EINVAL;
9342 	}
9343 
9344 	return -EINVAL;
9345 }
9346 #else
9347 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9348 			    struct tcf_exts *exts, u64 *action, u8 *queue)
9349 {
9350 	return -EINVAL;
9351 }
9352 #endif /* CONFIG_NET_CLS_ACT */
9353 
9354 static int ixgbe_clsu32_build_input(struct ixgbe_fdir_filter *input,
9355 				    union ixgbe_atr_input *mask,
9356 				    struct tc_cls_u32_offload *cls,
9357 				    struct ixgbe_mat_field *field_ptr,
9358 				    struct ixgbe_nexthdr *nexthdr)
9359 {
9360 	int i, j, off;
9361 	__be32 val, m;
9362 	bool found_entry = false, found_jump_field = false;
9363 
9364 	for (i = 0; i < cls->knode.sel->nkeys; i++) {
9365 		off = cls->knode.sel->keys[i].off;
9366 		val = cls->knode.sel->keys[i].val;
9367 		m = cls->knode.sel->keys[i].mask;
9368 
9369 		for (j = 0; field_ptr[j].val; j++) {
9370 			if (field_ptr[j].off == off) {
9371 				field_ptr[j].val(input, mask, (__force u32)val,
9372 						 (__force u32)m);
9373 				input->filter.formatted.flow_type |=
9374 					field_ptr[j].type;
9375 				found_entry = true;
9376 				break;
9377 			}
9378 		}
9379 		if (nexthdr) {
9380 			if (nexthdr->off == cls->knode.sel->keys[i].off &&
9381 			    nexthdr->val ==
9382 			    (__force u32)cls->knode.sel->keys[i].val &&
9383 			    nexthdr->mask ==
9384 			    (__force u32)cls->knode.sel->keys[i].mask)
9385 				found_jump_field = true;
9386 			else
9387 				continue;
9388 		}
9389 	}
9390 
9391 	if (nexthdr && !found_jump_field)
9392 		return -EINVAL;
9393 
9394 	if (!found_entry)
9395 		return 0;
9396 
9397 	mask->formatted.flow_type = IXGBE_ATR_L4TYPE_IPV6_MASK |
9398 				    IXGBE_ATR_L4TYPE_MASK;
9399 
9400 	if (input->filter.formatted.flow_type == IXGBE_ATR_FLOW_TYPE_IPV4)
9401 		mask->formatted.flow_type &= IXGBE_ATR_L4TYPE_IPV6_MASK;
9402 
9403 	return 0;
9404 }
9405 
9406 static int ixgbe_configure_clsu32(struct ixgbe_adapter *adapter,
9407 				  struct tc_cls_u32_offload *cls)
9408 {
9409 	__be16 protocol = cls->common.protocol;
9410 	u32 loc = cls->knode.handle & 0xfffff;
9411 	struct ixgbe_hw *hw = &adapter->hw;
9412 	struct ixgbe_mat_field *field_ptr;
9413 	struct ixgbe_fdir_filter *input = NULL;
9414 	union ixgbe_atr_input *mask = NULL;
9415 	struct ixgbe_jump_table *jump = NULL;
9416 	int i, err = -EINVAL;
9417 	u8 queue;
9418 	u32 uhtid, link_uhtid;
9419 
9420 	uhtid = TC_U32_USERHTID(cls->knode.handle);
9421 	link_uhtid = TC_U32_USERHTID(cls->knode.link_handle);
9422 
9423 	/* At the moment cls_u32 jumps to network layer and skips past
9424 	 * L2 headers. The canonical method to match L2 frames is to use
9425 	 * negative values. However this is error prone at best but really
9426 	 * just broken because there is no way to "know" what sort of hdr
9427 	 * is in front of the network layer. Fix cls_u32 to support L2
9428 	 * headers when needed.
9429 	 */
9430 	if (protocol != htons(ETH_P_IP))
9431 		return err;
9432 
9433 	if (loc >= ((1024 << adapter->fdir_pballoc) - 2)) {
9434 		e_err(drv, "Location out of range\n");
9435 		return err;
9436 	}
9437 
9438 	/* cls u32 is a graph starting at root node 0x800. The driver tracks
9439 	 * links and also the fields used to advance the parser across each
9440 	 * link (e.g. nexthdr/eat parameters from 'tc'). This way we can map
9441 	 * the u32 graph onto the hardware parse graph denoted in ixgbe_model.h
9442 	 * To add support for new nodes update ixgbe_model.h parse structures
9443 	 * this function _should_ be generic try not to hardcode values here.
9444 	 */
9445 	if (uhtid == 0x800) {
9446 		field_ptr = (adapter->jump_tables[0])->mat;
9447 	} else {
9448 		if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9449 			return err;
9450 		if (!adapter->jump_tables[uhtid])
9451 			return err;
9452 		field_ptr = (adapter->jump_tables[uhtid])->mat;
9453 	}
9454 
9455 	if (!field_ptr)
9456 		return err;
9457 
9458 	/* At this point we know the field_ptr is valid and need to either
9459 	 * build cls_u32 link or attach filter. Because adding a link to
9460 	 * a handle that does not exist is invalid and the same for adding
9461 	 * rules to handles that don't exist.
9462 	 */
9463 
9464 	if (link_uhtid) {
9465 		struct ixgbe_nexthdr *nexthdr = ixgbe_ipv4_jumps;
9466 
9467 		if (link_uhtid >= IXGBE_MAX_LINK_HANDLE)
9468 			return err;
9469 
9470 		if (!test_bit(link_uhtid - 1, &adapter->tables))
9471 			return err;
9472 
9473 		/* Multiple filters as links to the same hash table are not
9474 		 * supported. To add a new filter with the same next header
9475 		 * but different match/jump conditions, create a new hash table
9476 		 * and link to it.
9477 		 */
9478 		if (adapter->jump_tables[link_uhtid] &&
9479 		    (adapter->jump_tables[link_uhtid])->link_hdl) {
9480 			e_err(drv, "Link filter exists for link: %x\n",
9481 			      link_uhtid);
9482 			return err;
9483 		}
9484 
9485 		for (i = 0; nexthdr[i].jump; i++) {
9486 			if (nexthdr[i].o != cls->knode.sel->offoff ||
9487 			    nexthdr[i].s != cls->knode.sel->offshift ||
9488 			    nexthdr[i].m !=
9489 			    (__force u32)cls->knode.sel->offmask)
9490 				return err;
9491 
9492 			jump = kzalloc(sizeof(*jump), GFP_KERNEL);
9493 			if (!jump)
9494 				return -ENOMEM;
9495 			input = kzalloc(sizeof(*input), GFP_KERNEL);
9496 			if (!input) {
9497 				err = -ENOMEM;
9498 				goto free_jump;
9499 			}
9500 			mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9501 			if (!mask) {
9502 				err = -ENOMEM;
9503 				goto free_input;
9504 			}
9505 			jump->input = input;
9506 			jump->mask = mask;
9507 			jump->link_hdl = cls->knode.handle;
9508 
9509 			err = ixgbe_clsu32_build_input(input, mask, cls,
9510 						       field_ptr, &nexthdr[i]);
9511 			if (!err) {
9512 				jump->mat = nexthdr[i].jump;
9513 				adapter->jump_tables[link_uhtid] = jump;
9514 				break;
9515 			} else {
9516 				kfree(mask);
9517 				kfree(input);
9518 				kfree(jump);
9519 			}
9520 		}
9521 		return 0;
9522 	}
9523 
9524 	input = kzalloc(sizeof(*input), GFP_KERNEL);
9525 	if (!input)
9526 		return -ENOMEM;
9527 	mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9528 	if (!mask) {
9529 		err = -ENOMEM;
9530 		goto free_input;
9531 	}
9532 
9533 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid])) {
9534 		if ((adapter->jump_tables[uhtid])->input)
9535 			memcpy(input, (adapter->jump_tables[uhtid])->input,
9536 			       sizeof(*input));
9537 		if ((adapter->jump_tables[uhtid])->mask)
9538 			memcpy(mask, (adapter->jump_tables[uhtid])->mask,
9539 			       sizeof(*mask));
9540 
9541 		/* Lookup in all child hash tables if this location is already
9542 		 * filled with a filter
9543 		 */
9544 		for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
9545 			struct ixgbe_jump_table *link = adapter->jump_tables[i];
9546 
9547 			if (link && (test_bit(loc - 1, link->child_loc_map))) {
9548 				e_err(drv, "Filter exists in location: %x\n",
9549 				      loc);
9550 				err = -EINVAL;
9551 				goto err_out;
9552 			}
9553 		}
9554 	}
9555 	err = ixgbe_clsu32_build_input(input, mask, cls, field_ptr, NULL);
9556 	if (err)
9557 		goto err_out;
9558 
9559 	err = parse_tc_actions(adapter, cls->knode.exts, &input->action,
9560 			       &queue);
9561 	if (err < 0)
9562 		goto err_out;
9563 
9564 	input->sw_idx = loc;
9565 
9566 	spin_lock(&adapter->fdir_perfect_lock);
9567 
9568 	if (hlist_empty(&adapter->fdir_filter_list)) {
9569 		memcpy(&adapter->fdir_mask, mask, sizeof(*mask));
9570 		err = ixgbe_fdir_set_input_mask_82599(hw, mask);
9571 		if (err)
9572 			goto err_out_w_lock;
9573 	} else if (memcmp(&adapter->fdir_mask, mask, sizeof(*mask))) {
9574 		err = -EINVAL;
9575 		goto err_out_w_lock;
9576 	}
9577 
9578 	ixgbe_atr_compute_perfect_hash_82599(&input->filter, mask);
9579 	err = ixgbe_fdir_write_perfect_filter_82599(hw, &input->filter,
9580 						    input->sw_idx, queue);
9581 	if (err)
9582 		goto err_out_w_lock;
9583 
9584 	ixgbe_update_ethtool_fdir_entry(adapter, input, input->sw_idx);
9585 	spin_unlock(&adapter->fdir_perfect_lock);
9586 
9587 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid]))
9588 		set_bit(loc - 1, (adapter->jump_tables[uhtid])->child_loc_map);
9589 
9590 	kfree(mask);
9591 	return err;
9592 err_out_w_lock:
9593 	spin_unlock(&adapter->fdir_perfect_lock);
9594 err_out:
9595 	kfree(mask);
9596 free_input:
9597 	kfree(input);
9598 free_jump:
9599 	kfree(jump);
9600 	return err;
9601 }
9602 
9603 static int ixgbe_setup_tc_cls_u32(struct ixgbe_adapter *adapter,
9604 				  struct tc_cls_u32_offload *cls_u32)
9605 {
9606 	switch (cls_u32->command) {
9607 	case TC_CLSU32_NEW_KNODE:
9608 	case TC_CLSU32_REPLACE_KNODE:
9609 		return ixgbe_configure_clsu32(adapter, cls_u32);
9610 	case TC_CLSU32_DELETE_KNODE:
9611 		return ixgbe_delete_clsu32(adapter, cls_u32);
9612 	case TC_CLSU32_NEW_HNODE:
9613 	case TC_CLSU32_REPLACE_HNODE:
9614 		return ixgbe_configure_clsu32_add_hnode(adapter, cls_u32);
9615 	case TC_CLSU32_DELETE_HNODE:
9616 		return ixgbe_configure_clsu32_del_hnode(adapter, cls_u32);
9617 	default:
9618 		return -EOPNOTSUPP;
9619 	}
9620 }
9621 
9622 static int ixgbe_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
9623 				   void *cb_priv)
9624 {
9625 	struct ixgbe_adapter *adapter = cb_priv;
9626 
9627 	if (!tc_cls_can_offload_and_chain0(adapter->netdev, type_data))
9628 		return -EOPNOTSUPP;
9629 
9630 	switch (type) {
9631 	case TC_SETUP_CLSU32:
9632 		return ixgbe_setup_tc_cls_u32(adapter, type_data);
9633 	default:
9634 		return -EOPNOTSUPP;
9635 	}
9636 }
9637 
9638 static int ixgbe_setup_tc_mqprio(struct net_device *dev,
9639 				 struct tc_mqprio_qopt *mqprio)
9640 {
9641 	mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
9642 	return ixgbe_setup_tc(dev, mqprio->num_tc);
9643 }
9644 
9645 static LIST_HEAD(ixgbe_block_cb_list);
9646 
9647 static int __ixgbe_setup_tc(struct net_device *dev, enum tc_setup_type type,
9648 			    void *type_data)
9649 {
9650 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9651 
9652 	switch (type) {
9653 	case TC_SETUP_BLOCK:
9654 		return flow_block_cb_setup_simple(type_data,
9655 						  &ixgbe_block_cb_list,
9656 						  ixgbe_setup_tc_block_cb,
9657 						  adapter, adapter, true);
9658 	case TC_SETUP_QDISC_MQPRIO:
9659 		return ixgbe_setup_tc_mqprio(dev, type_data);
9660 	default:
9661 		return -EOPNOTSUPP;
9662 	}
9663 }
9664 
9665 #ifdef CONFIG_PCI_IOV
9666 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
9667 {
9668 	struct net_device *netdev = adapter->netdev;
9669 
9670 	rtnl_lock();
9671 	ixgbe_setup_tc(netdev, adapter->hw_tcs);
9672 	rtnl_unlock();
9673 }
9674 
9675 #endif
9676 void ixgbe_do_reset(struct net_device *netdev)
9677 {
9678 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9679 
9680 	if (netif_running(netdev))
9681 		ixgbe_reinit_locked(adapter);
9682 	else
9683 		ixgbe_reset(adapter);
9684 }
9685 
9686 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
9687 					    netdev_features_t features)
9688 {
9689 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9690 
9691 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
9692 	if (!(features & NETIF_F_RXCSUM))
9693 		features &= ~NETIF_F_LRO;
9694 
9695 	/* Turn off LRO if not RSC capable */
9696 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
9697 		features &= ~NETIF_F_LRO;
9698 
9699 	if (adapter->xdp_prog && (features & NETIF_F_LRO)) {
9700 		e_dev_err("LRO is not supported with XDP\n");
9701 		features &= ~NETIF_F_LRO;
9702 	}
9703 
9704 	return features;
9705 }
9706 
9707 static void ixgbe_reset_l2fw_offload(struct ixgbe_adapter *adapter)
9708 {
9709 	int rss = min_t(int, ixgbe_max_rss_indices(adapter),
9710 			num_online_cpus());
9711 
9712 	/* go back to full RSS if we're not running SR-IOV */
9713 	if (!adapter->ring_feature[RING_F_VMDQ].offset)
9714 		adapter->flags &= ~(IXGBE_FLAG_VMDQ_ENABLED |
9715 				    IXGBE_FLAG_SRIOV_ENABLED);
9716 
9717 	adapter->ring_feature[RING_F_RSS].limit = rss;
9718 	adapter->ring_feature[RING_F_VMDQ].limit = 1;
9719 
9720 	ixgbe_setup_tc(adapter->netdev, adapter->hw_tcs);
9721 }
9722 
9723 static int ixgbe_set_features(struct net_device *netdev,
9724 			      netdev_features_t features)
9725 {
9726 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
9727 	netdev_features_t changed = netdev->features ^ features;
9728 	bool need_reset = false;
9729 
9730 	/* Make sure RSC matches LRO, reset if change */
9731 	if (!(features & NETIF_F_LRO)) {
9732 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
9733 			need_reset = true;
9734 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
9735 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
9736 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
9737 		if (adapter->rx_itr_setting == 1 ||
9738 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
9739 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
9740 			need_reset = true;
9741 		} else if ((changed ^ features) & NETIF_F_LRO) {
9742 			e_info(probe, "rx-usecs set too low, "
9743 			       "disabling RSC\n");
9744 		}
9745 	}
9746 
9747 	/*
9748 	 * Check if Flow Director n-tuple support or hw_tc support was
9749 	 * enabled or disabled.  If the state changed, we need to reset.
9750 	 */
9751 	if ((features & NETIF_F_NTUPLE) || (features & NETIF_F_HW_TC)) {
9752 		/* turn off ATR, enable perfect filters and reset */
9753 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
9754 			need_reset = true;
9755 
9756 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
9757 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9758 	} else {
9759 		/* turn off perfect filters, enable ATR and reset */
9760 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
9761 			need_reset = true;
9762 
9763 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9764 
9765 		/* We cannot enable ATR if SR-IOV is enabled */
9766 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED ||
9767 		    /* We cannot enable ATR if we have 2 or more tcs */
9768 		    (adapter->hw_tcs > 1) ||
9769 		    /* We cannot enable ATR if RSS is disabled */
9770 		    (adapter->ring_feature[RING_F_RSS].limit <= 1) ||
9771 		    /* A sample rate of 0 indicates ATR disabled */
9772 		    (!adapter->atr_sample_rate))
9773 			; /* do nothing not supported */
9774 		else /* otherwise supported and set the flag */
9775 			adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
9776 	}
9777 
9778 	if (changed & NETIF_F_RXALL)
9779 		need_reset = true;
9780 
9781 	netdev->features = features;
9782 
9783 	if ((changed & NETIF_F_HW_L2FW_DOFFLOAD) && adapter->num_rx_pools > 1)
9784 		ixgbe_reset_l2fw_offload(adapter);
9785 	else if (need_reset)
9786 		ixgbe_do_reset(netdev);
9787 	else if (changed & (NETIF_F_HW_VLAN_CTAG_RX |
9788 			    NETIF_F_HW_VLAN_CTAG_FILTER))
9789 		ixgbe_set_rx_mode(netdev);
9790 
9791 	return 1;
9792 }
9793 
9794 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
9795 			     struct net_device *dev,
9796 			     const unsigned char *addr, u16 vid,
9797 			     u16 flags,
9798 			     struct netlink_ext_ack *extack)
9799 {
9800 	/* guarantee we can provide a unique filter for the unicast address */
9801 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
9802 		struct ixgbe_adapter *adapter = netdev_priv(dev);
9803 		u16 pool = VMDQ_P(0);
9804 
9805 		if (netdev_uc_count(dev) >= ixgbe_available_rars(adapter, pool))
9806 			return -ENOMEM;
9807 	}
9808 
9809 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
9810 }
9811 
9812 /**
9813  * ixgbe_configure_bridge_mode - set various bridge modes
9814  * @adapter: the private structure
9815  * @mode: requested bridge mode
9816  *
9817  * Configure some settings require for various bridge modes.
9818  **/
9819 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
9820 				       __u16 mode)
9821 {
9822 	struct ixgbe_hw *hw = &adapter->hw;
9823 	unsigned int p, num_pools;
9824 	u32 vmdctl;
9825 
9826 	switch (mode) {
9827 	case BRIDGE_MODE_VEPA:
9828 		/* disable Tx loopback, rely on switch hairpin mode */
9829 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
9830 
9831 		/* must enable Rx switching replication to allow multicast
9832 		 * packet reception on all VFs, and to enable source address
9833 		 * pruning.
9834 		 */
9835 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9836 		vmdctl |= IXGBE_VT_CTL_REPLEN;
9837 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9838 
9839 		/* enable Rx source address pruning. Note, this requires
9840 		 * replication to be enabled or else it does nothing.
9841 		 */
9842 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9843 		for (p = 0; p < num_pools; p++) {
9844 			if (hw->mac.ops.set_source_address_pruning)
9845 				hw->mac.ops.set_source_address_pruning(hw,
9846 								       true,
9847 								       p);
9848 		}
9849 		break;
9850 	case BRIDGE_MODE_VEB:
9851 		/* enable Tx loopback for internal VF/PF communication */
9852 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
9853 				IXGBE_PFDTXGSWC_VT_LBEN);
9854 
9855 		/* disable Rx switching replication unless we have SR-IOV
9856 		 * virtual functions
9857 		 */
9858 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9859 		if (!adapter->num_vfs)
9860 			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
9861 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9862 
9863 		/* disable Rx source address pruning, since we don't expect to
9864 		 * be receiving external loopback of our transmitted frames.
9865 		 */
9866 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
9867 		for (p = 0; p < num_pools; p++) {
9868 			if (hw->mac.ops.set_source_address_pruning)
9869 				hw->mac.ops.set_source_address_pruning(hw,
9870 								       false,
9871 								       p);
9872 		}
9873 		break;
9874 	default:
9875 		return -EINVAL;
9876 	}
9877 
9878 	adapter->bridge_mode = mode;
9879 
9880 	e_info(drv, "enabling bridge mode: %s\n",
9881 	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
9882 
9883 	return 0;
9884 }
9885 
9886 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
9887 				    struct nlmsghdr *nlh, u16 flags,
9888 				    struct netlink_ext_ack *extack)
9889 {
9890 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9891 	struct nlattr *attr, *br_spec;
9892 	int rem;
9893 
9894 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9895 		return -EOPNOTSUPP;
9896 
9897 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
9898 	if (!br_spec)
9899 		return -EINVAL;
9900 
9901 	nla_for_each_nested(attr, br_spec, rem) {
9902 		int status;
9903 		__u16 mode;
9904 
9905 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
9906 			continue;
9907 
9908 		if (nla_len(attr) < sizeof(mode))
9909 			return -EINVAL;
9910 
9911 		mode = nla_get_u16(attr);
9912 		status = ixgbe_configure_bridge_mode(adapter, mode);
9913 		if (status)
9914 			return status;
9915 
9916 		break;
9917 	}
9918 
9919 	return 0;
9920 }
9921 
9922 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9923 				    struct net_device *dev,
9924 				    u32 filter_mask, int nlflags)
9925 {
9926 	struct ixgbe_adapter *adapter = netdev_priv(dev);
9927 
9928 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
9929 		return 0;
9930 
9931 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
9932 				       adapter->bridge_mode, 0, 0, nlflags,
9933 				       filter_mask, NULL);
9934 }
9935 
9936 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
9937 {
9938 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
9939 	struct ixgbe_fwd_adapter *accel;
9940 	int tcs = adapter->hw_tcs ? : 1;
9941 	int pool, err;
9942 
9943 	if (adapter->xdp_prog) {
9944 		e_warn(probe, "L2FW offload is not supported with XDP\n");
9945 		return ERR_PTR(-EINVAL);
9946 	}
9947 
9948 	/* The hardware supported by ixgbe only filters on the destination MAC
9949 	 * address. In order to avoid issues we only support offloading modes
9950 	 * where the hardware can actually provide the functionality.
9951 	 */
9952 	if (!macvlan_supports_dest_filter(vdev))
9953 		return ERR_PTR(-EMEDIUMTYPE);
9954 
9955 	/* We need to lock down the macvlan to be a single queue device so that
9956 	 * we can reuse the tc_to_txq field in the macvlan netdev to represent
9957 	 * the queue mapping to our netdev.
9958 	 */
9959 	if (netif_is_multiqueue(vdev))
9960 		return ERR_PTR(-ERANGE);
9961 
9962 	pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
9963 	if (pool == adapter->num_rx_pools) {
9964 		u16 used_pools = adapter->num_vfs + adapter->num_rx_pools;
9965 		u16 reserved_pools;
9966 
9967 		if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
9968 		     adapter->num_rx_pools >= (MAX_TX_QUEUES / tcs)) ||
9969 		    adapter->num_rx_pools > IXGBE_MAX_MACVLANS)
9970 			return ERR_PTR(-EBUSY);
9971 
9972 		/* Hardware has a limited number of available pools. Each VF,
9973 		 * and the PF require a pool. Check to ensure we don't
9974 		 * attempt to use more then the available number of pools.
9975 		 */
9976 		if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
9977 			return ERR_PTR(-EBUSY);
9978 
9979 		/* Enable VMDq flag so device will be set in VM mode */
9980 		adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED |
9981 				  IXGBE_FLAG_SRIOV_ENABLED;
9982 
9983 		/* Try to reserve as many queues per pool as possible,
9984 		 * we start with the configurations that support 4 queues
9985 		 * per pools, followed by 2, and then by just 1 per pool.
9986 		 */
9987 		if (used_pools < 32 && adapter->num_rx_pools < 16)
9988 			reserved_pools = min_t(u16,
9989 					       32 - used_pools,
9990 					       16 - adapter->num_rx_pools);
9991 		else if (adapter->num_rx_pools < 32)
9992 			reserved_pools = min_t(u16,
9993 					       64 - used_pools,
9994 					       32 - adapter->num_rx_pools);
9995 		else
9996 			reserved_pools = 64 - used_pools;
9997 
9998 
9999 		if (!reserved_pools)
10000 			return ERR_PTR(-EBUSY);
10001 
10002 		adapter->ring_feature[RING_F_VMDQ].limit += reserved_pools;
10003 
10004 		/* Force reinit of ring allocation with VMDQ enabled */
10005 		err = ixgbe_setup_tc(pdev, adapter->hw_tcs);
10006 		if (err)
10007 			return ERR_PTR(err);
10008 
10009 		if (pool >= adapter->num_rx_pools)
10010 			return ERR_PTR(-ENOMEM);
10011 	}
10012 
10013 	accel = kzalloc(sizeof(*accel), GFP_KERNEL);
10014 	if (!accel)
10015 		return ERR_PTR(-ENOMEM);
10016 
10017 	set_bit(pool, adapter->fwd_bitmask);
10018 	netdev_set_sb_channel(vdev, pool);
10019 	accel->pool = pool;
10020 	accel->netdev = vdev;
10021 
10022 	if (!netif_running(pdev))
10023 		return accel;
10024 
10025 	err = ixgbe_fwd_ring_up(adapter, accel);
10026 	if (err)
10027 		return ERR_PTR(err);
10028 
10029 	return accel;
10030 }
10031 
10032 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
10033 {
10034 	struct ixgbe_fwd_adapter *accel = priv;
10035 	struct ixgbe_adapter *adapter = netdev_priv(pdev);
10036 	unsigned int rxbase = accel->rx_base_queue;
10037 	unsigned int i;
10038 
10039 	/* delete unicast filter associated with offloaded interface */
10040 	ixgbe_del_mac_filter(adapter, accel->netdev->dev_addr,
10041 			     VMDQ_P(accel->pool));
10042 
10043 	/* Allow remaining Rx packets to get flushed out of the
10044 	 * Rx FIFO before we drop the netdev for the ring.
10045 	 */
10046 	usleep_range(10000, 20000);
10047 
10048 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
10049 		struct ixgbe_ring *ring = adapter->rx_ring[rxbase + i];
10050 		struct ixgbe_q_vector *qv = ring->q_vector;
10051 
10052 		/* Make sure we aren't processing any packets and clear
10053 		 * netdev to shut down the ring.
10054 		 */
10055 		if (netif_running(adapter->netdev))
10056 			napi_synchronize(&qv->napi);
10057 		ring->netdev = NULL;
10058 	}
10059 
10060 	/* unbind the queues and drop the subordinate channel config */
10061 	netdev_unbind_sb_channel(pdev, accel->netdev);
10062 	netdev_set_sb_channel(accel->netdev, 0);
10063 
10064 	clear_bit(accel->pool, adapter->fwd_bitmask);
10065 	kfree(accel);
10066 }
10067 
10068 #define IXGBE_MAX_MAC_HDR_LEN		127
10069 #define IXGBE_MAX_NETWORK_HDR_LEN	511
10070 
10071 static netdev_features_t
10072 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
10073 		     netdev_features_t features)
10074 {
10075 	unsigned int network_hdr_len, mac_hdr_len;
10076 
10077 	/* Make certain the headers can be described by a context descriptor */
10078 	mac_hdr_len = skb_network_header(skb) - skb->data;
10079 	if (unlikely(mac_hdr_len > IXGBE_MAX_MAC_HDR_LEN))
10080 		return features & ~(NETIF_F_HW_CSUM |
10081 				    NETIF_F_SCTP_CRC |
10082 				    NETIF_F_GSO_UDP_L4 |
10083 				    NETIF_F_HW_VLAN_CTAG_TX |
10084 				    NETIF_F_TSO |
10085 				    NETIF_F_TSO6);
10086 
10087 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
10088 	if (unlikely(network_hdr_len >  IXGBE_MAX_NETWORK_HDR_LEN))
10089 		return features & ~(NETIF_F_HW_CSUM |
10090 				    NETIF_F_SCTP_CRC |
10091 				    NETIF_F_GSO_UDP_L4 |
10092 				    NETIF_F_TSO |
10093 				    NETIF_F_TSO6);
10094 
10095 	/* We can only support IPV4 TSO in tunnels if we can mangle the
10096 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
10097 	 * IPsec offoad sets skb->encapsulation but still can handle
10098 	 * the TSO, so it's the exception.
10099 	 */
10100 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID)) {
10101 #ifdef CONFIG_IXGBE_IPSEC
10102 		if (!secpath_exists(skb))
10103 #endif
10104 			features &= ~NETIF_F_TSO;
10105 	}
10106 
10107 	return features;
10108 }
10109 
10110 static int ixgbe_xdp_setup(struct net_device *dev, struct bpf_prog *prog)
10111 {
10112 	int i, frame_size = dev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
10113 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10114 	struct bpf_prog *old_prog;
10115 	bool need_reset;
10116 
10117 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
10118 		return -EINVAL;
10119 
10120 	if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
10121 		return -EINVAL;
10122 
10123 	/* verify ixgbe ring attributes are sufficient for XDP */
10124 	for (i = 0; i < adapter->num_rx_queues; i++) {
10125 		struct ixgbe_ring *ring = adapter->rx_ring[i];
10126 
10127 		if (ring_is_rsc_enabled(ring))
10128 			return -EINVAL;
10129 
10130 		if (frame_size > ixgbe_rx_bufsz(ring))
10131 			return -EINVAL;
10132 	}
10133 
10134 	if (nr_cpu_ids > MAX_XDP_QUEUES)
10135 		return -ENOMEM;
10136 
10137 	old_prog = xchg(&adapter->xdp_prog, prog);
10138 	need_reset = (!!prog != !!old_prog);
10139 
10140 	/* If transitioning XDP modes reconfigure rings */
10141 	if (need_reset) {
10142 		int err;
10143 
10144 		if (!prog)
10145 			/* Wait until ndo_xsk_wakeup completes. */
10146 			synchronize_rcu();
10147 		err = ixgbe_setup_tc(dev, adapter->hw_tcs);
10148 
10149 		if (err) {
10150 			rcu_assign_pointer(adapter->xdp_prog, old_prog);
10151 			return -EINVAL;
10152 		}
10153 	} else {
10154 		for (i = 0; i < adapter->num_rx_queues; i++)
10155 			(void)xchg(&adapter->rx_ring[i]->xdp_prog,
10156 			    adapter->xdp_prog);
10157 	}
10158 
10159 	if (old_prog)
10160 		bpf_prog_put(old_prog);
10161 
10162 	/* Kick start the NAPI context if there is an AF_XDP socket open
10163 	 * on that queue id. This so that receiving will start.
10164 	 */
10165 	if (need_reset && prog)
10166 		for (i = 0; i < adapter->num_rx_queues; i++)
10167 			if (adapter->xdp_ring[i]->xsk_pool)
10168 				(void)ixgbe_xsk_wakeup(adapter->netdev, i,
10169 						       XDP_WAKEUP_RX);
10170 
10171 	return 0;
10172 }
10173 
10174 static int ixgbe_xdp(struct net_device *dev, struct netdev_bpf *xdp)
10175 {
10176 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10177 
10178 	switch (xdp->command) {
10179 	case XDP_SETUP_PROG:
10180 		return ixgbe_xdp_setup(dev, xdp->prog);
10181 	case XDP_SETUP_XSK_POOL:
10182 		return ixgbe_xsk_pool_setup(adapter, xdp->xsk.pool,
10183 					    xdp->xsk.queue_id);
10184 
10185 	default:
10186 		return -EINVAL;
10187 	}
10188 }
10189 
10190 void ixgbe_xdp_ring_update_tail(struct ixgbe_ring *ring)
10191 {
10192 	/* Force memory writes to complete before letting h/w know there
10193 	 * are new descriptors to fetch.
10194 	 */
10195 	wmb();
10196 	writel(ring->next_to_use, ring->tail);
10197 }
10198 
10199 static int ixgbe_xdp_xmit(struct net_device *dev, int n,
10200 			  struct xdp_frame **frames, u32 flags)
10201 {
10202 	struct ixgbe_adapter *adapter = netdev_priv(dev);
10203 	struct ixgbe_ring *ring;
10204 	int nxmit = 0;
10205 	int i;
10206 
10207 	if (unlikely(test_bit(__IXGBE_DOWN, &adapter->state)))
10208 		return -ENETDOWN;
10209 
10210 	if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
10211 		return -EINVAL;
10212 
10213 	/* During program transitions its possible adapter->xdp_prog is assigned
10214 	 * but ring has not been configured yet. In this case simply abort xmit.
10215 	 */
10216 	ring = adapter->xdp_prog ? adapter->xdp_ring[smp_processor_id()] : NULL;
10217 	if (unlikely(!ring))
10218 		return -ENXIO;
10219 
10220 	if (unlikely(test_bit(__IXGBE_TX_DISABLED, &ring->state)))
10221 		return -ENXIO;
10222 
10223 	for (i = 0; i < n; i++) {
10224 		struct xdp_frame *xdpf = frames[i];
10225 		int err;
10226 
10227 		err = ixgbe_xmit_xdp_ring(adapter, xdpf);
10228 		if (err != IXGBE_XDP_TX)
10229 			break;
10230 		nxmit++;
10231 	}
10232 
10233 	if (unlikely(flags & XDP_XMIT_FLUSH))
10234 		ixgbe_xdp_ring_update_tail(ring);
10235 
10236 	return nxmit;
10237 }
10238 
10239 static const struct net_device_ops ixgbe_netdev_ops = {
10240 	.ndo_open		= ixgbe_open,
10241 	.ndo_stop		= ixgbe_close,
10242 	.ndo_start_xmit		= ixgbe_xmit_frame,
10243 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
10244 	.ndo_validate_addr	= eth_validate_addr,
10245 	.ndo_set_mac_address	= ixgbe_set_mac,
10246 	.ndo_change_mtu		= ixgbe_change_mtu,
10247 	.ndo_tx_timeout		= ixgbe_tx_timeout,
10248 	.ndo_set_tx_maxrate	= ixgbe_tx_maxrate,
10249 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
10250 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
10251 	.ndo_do_ioctl		= ixgbe_ioctl,
10252 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
10253 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
10254 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
10255 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
10256 	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
10257 	.ndo_set_vf_trust	= ixgbe_ndo_set_vf_trust,
10258 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
10259 	.ndo_get_stats64	= ixgbe_get_stats64,
10260 	.ndo_setup_tc		= __ixgbe_setup_tc,
10261 #ifdef IXGBE_FCOE
10262 	.ndo_select_queue	= ixgbe_select_queue,
10263 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
10264 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
10265 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
10266 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
10267 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
10268 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
10269 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
10270 #endif /* IXGBE_FCOE */
10271 	.ndo_set_features = ixgbe_set_features,
10272 	.ndo_fix_features = ixgbe_fix_features,
10273 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
10274 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
10275 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
10276 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
10277 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
10278 	.ndo_features_check	= ixgbe_features_check,
10279 	.ndo_bpf		= ixgbe_xdp,
10280 	.ndo_xdp_xmit		= ixgbe_xdp_xmit,
10281 	.ndo_xsk_wakeup         = ixgbe_xsk_wakeup,
10282 };
10283 
10284 static void ixgbe_disable_txr_hw(struct ixgbe_adapter *adapter,
10285 				 struct ixgbe_ring *tx_ring)
10286 {
10287 	unsigned long wait_delay, delay_interval;
10288 	struct ixgbe_hw *hw = &adapter->hw;
10289 	u8 reg_idx = tx_ring->reg_idx;
10290 	int wait_loop;
10291 	u32 txdctl;
10292 
10293 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
10294 
10295 	/* delay mechanism from ixgbe_disable_tx */
10296 	delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
10297 
10298 	wait_loop = IXGBE_MAX_RX_DESC_POLL;
10299 	wait_delay = delay_interval;
10300 
10301 	while (wait_loop--) {
10302 		usleep_range(wait_delay, wait_delay + 10);
10303 		wait_delay += delay_interval * 2;
10304 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
10305 
10306 		if (!(txdctl & IXGBE_TXDCTL_ENABLE))
10307 			return;
10308 	}
10309 
10310 	e_err(drv, "TXDCTL.ENABLE not cleared within the polling period\n");
10311 }
10312 
10313 static void ixgbe_disable_txr(struct ixgbe_adapter *adapter,
10314 			      struct ixgbe_ring *tx_ring)
10315 {
10316 	set_bit(__IXGBE_TX_DISABLED, &tx_ring->state);
10317 	ixgbe_disable_txr_hw(adapter, tx_ring);
10318 }
10319 
10320 static void ixgbe_disable_rxr_hw(struct ixgbe_adapter *adapter,
10321 				 struct ixgbe_ring *rx_ring)
10322 {
10323 	unsigned long wait_delay, delay_interval;
10324 	struct ixgbe_hw *hw = &adapter->hw;
10325 	u8 reg_idx = rx_ring->reg_idx;
10326 	int wait_loop;
10327 	u32 rxdctl;
10328 
10329 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
10330 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
10331 	rxdctl |= IXGBE_RXDCTL_SWFLSH;
10332 
10333 	/* write value back with RXDCTL.ENABLE bit cleared */
10334 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
10335 
10336 	/* RXDCTL.EN may not change on 82598 if link is down, so skip it */
10337 	if (hw->mac.type == ixgbe_mac_82598EB &&
10338 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
10339 		return;
10340 
10341 	/* delay mechanism from ixgbe_disable_rx */
10342 	delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
10343 
10344 	wait_loop = IXGBE_MAX_RX_DESC_POLL;
10345 	wait_delay = delay_interval;
10346 
10347 	while (wait_loop--) {
10348 		usleep_range(wait_delay, wait_delay + 10);
10349 		wait_delay += delay_interval * 2;
10350 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
10351 
10352 		if (!(rxdctl & IXGBE_RXDCTL_ENABLE))
10353 			return;
10354 	}
10355 
10356 	e_err(drv, "RXDCTL.ENABLE not cleared within the polling period\n");
10357 }
10358 
10359 static void ixgbe_reset_txr_stats(struct ixgbe_ring *tx_ring)
10360 {
10361 	memset(&tx_ring->stats, 0, sizeof(tx_ring->stats));
10362 	memset(&tx_ring->tx_stats, 0, sizeof(tx_ring->tx_stats));
10363 }
10364 
10365 static void ixgbe_reset_rxr_stats(struct ixgbe_ring *rx_ring)
10366 {
10367 	memset(&rx_ring->stats, 0, sizeof(rx_ring->stats));
10368 	memset(&rx_ring->rx_stats, 0, sizeof(rx_ring->rx_stats));
10369 }
10370 
10371 /**
10372  * ixgbe_txrx_ring_disable - Disable Rx/Tx/XDP Tx rings
10373  * @adapter: adapter structure
10374  * @ring: ring index
10375  *
10376  * This function disables a certain Rx/Tx/XDP Tx ring. The function
10377  * assumes that the netdev is running.
10378  **/
10379 void ixgbe_txrx_ring_disable(struct ixgbe_adapter *adapter, int ring)
10380 {
10381 	struct ixgbe_ring *rx_ring, *tx_ring, *xdp_ring;
10382 
10383 	rx_ring = adapter->rx_ring[ring];
10384 	tx_ring = adapter->tx_ring[ring];
10385 	xdp_ring = adapter->xdp_ring[ring];
10386 
10387 	ixgbe_disable_txr(adapter, tx_ring);
10388 	if (xdp_ring)
10389 		ixgbe_disable_txr(adapter, xdp_ring);
10390 	ixgbe_disable_rxr_hw(adapter, rx_ring);
10391 
10392 	if (xdp_ring)
10393 		synchronize_rcu();
10394 
10395 	/* Rx/Tx/XDP Tx share the same napi context. */
10396 	napi_disable(&rx_ring->q_vector->napi);
10397 
10398 	ixgbe_clean_tx_ring(tx_ring);
10399 	if (xdp_ring)
10400 		ixgbe_clean_tx_ring(xdp_ring);
10401 	ixgbe_clean_rx_ring(rx_ring);
10402 
10403 	ixgbe_reset_txr_stats(tx_ring);
10404 	if (xdp_ring)
10405 		ixgbe_reset_txr_stats(xdp_ring);
10406 	ixgbe_reset_rxr_stats(rx_ring);
10407 }
10408 
10409 /**
10410  * ixgbe_txrx_ring_enable - Enable Rx/Tx/XDP Tx rings
10411  * @adapter: adapter structure
10412  * @ring: ring index
10413  *
10414  * This function enables a certain Rx/Tx/XDP Tx ring. The function
10415  * assumes that the netdev is running.
10416  **/
10417 void ixgbe_txrx_ring_enable(struct ixgbe_adapter *adapter, int ring)
10418 {
10419 	struct ixgbe_ring *rx_ring, *tx_ring, *xdp_ring;
10420 
10421 	rx_ring = adapter->rx_ring[ring];
10422 	tx_ring = adapter->tx_ring[ring];
10423 	xdp_ring = adapter->xdp_ring[ring];
10424 
10425 	/* Rx/Tx/XDP Tx share the same napi context. */
10426 	napi_enable(&rx_ring->q_vector->napi);
10427 
10428 	ixgbe_configure_tx_ring(adapter, tx_ring);
10429 	if (xdp_ring)
10430 		ixgbe_configure_tx_ring(adapter, xdp_ring);
10431 	ixgbe_configure_rx_ring(adapter, rx_ring);
10432 
10433 	clear_bit(__IXGBE_TX_DISABLED, &tx_ring->state);
10434 	if (xdp_ring)
10435 		clear_bit(__IXGBE_TX_DISABLED, &xdp_ring->state);
10436 }
10437 
10438 /**
10439  * ixgbe_enumerate_functions - Get the number of ports this device has
10440  * @adapter: adapter structure
10441  *
10442  * This function enumerates the phsyical functions co-located on a single slot,
10443  * in order to determine how many ports a device has. This is most useful in
10444  * determining the required GT/s of PCIe bandwidth necessary for optimal
10445  * performance.
10446  **/
10447 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
10448 {
10449 	struct pci_dev *entry, *pdev = adapter->pdev;
10450 	int physfns = 0;
10451 
10452 	/* Some cards can not use the generic count PCIe functions method,
10453 	 * because they are behind a parent switch, so we hardcode these with
10454 	 * the correct number of functions.
10455 	 */
10456 	if (ixgbe_pcie_from_parent(&adapter->hw))
10457 		physfns = 4;
10458 
10459 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
10460 		/* don't count virtual functions */
10461 		if (entry->is_virtfn)
10462 			continue;
10463 
10464 		/* When the devices on the bus don't all match our device ID,
10465 		 * we can't reliably determine the correct number of
10466 		 * functions. This can occur if a function has been direct
10467 		 * attached to a virtual machine using VT-d, for example. In
10468 		 * this case, simply return -1 to indicate this.
10469 		 */
10470 		if ((entry->vendor != pdev->vendor) ||
10471 		    (entry->device != pdev->device))
10472 			return -1;
10473 
10474 		physfns++;
10475 	}
10476 
10477 	return physfns;
10478 }
10479 
10480 /**
10481  * ixgbe_wol_supported - Check whether device supports WoL
10482  * @adapter: the adapter private structure
10483  * @device_id: the device ID
10484  * @subdevice_id: the subsystem device ID
10485  *
10486  * This function is used by probe and ethtool to determine
10487  * which devices have WoL support
10488  *
10489  **/
10490 bool ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
10491 			 u16 subdevice_id)
10492 {
10493 	struct ixgbe_hw *hw = &adapter->hw;
10494 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
10495 
10496 	/* WOL not supported on 82598 */
10497 	if (hw->mac.type == ixgbe_mac_82598EB)
10498 		return false;
10499 
10500 	/* check eeprom to see if WOL is enabled for X540 and newer */
10501 	if (hw->mac.type >= ixgbe_mac_X540) {
10502 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
10503 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
10504 		     (hw->bus.func == 0)))
10505 			return true;
10506 	}
10507 
10508 	/* WOL is determined based on device IDs for 82599 MACs */
10509 	switch (device_id) {
10510 	case IXGBE_DEV_ID_82599_SFP:
10511 		/* Only these subdevices could supports WOL */
10512 		switch (subdevice_id) {
10513 		case IXGBE_SUBDEV_ID_82599_560FLR:
10514 		case IXGBE_SUBDEV_ID_82599_LOM_SNAP6:
10515 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
10516 		case IXGBE_SUBDEV_ID_82599_SFP_2OCP:
10517 			/* only support first port */
10518 			if (hw->bus.func != 0)
10519 				break;
10520 			fallthrough;
10521 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
10522 		case IXGBE_SUBDEV_ID_82599_SFP:
10523 		case IXGBE_SUBDEV_ID_82599_RNDC:
10524 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
10525 		case IXGBE_SUBDEV_ID_82599_SFP_1OCP:
10526 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM1:
10527 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM2:
10528 			return true;
10529 		}
10530 		break;
10531 	case IXGBE_DEV_ID_82599EN_SFP:
10532 		/* Only these subdevices support WOL */
10533 		switch (subdevice_id) {
10534 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
10535 			return true;
10536 		}
10537 		break;
10538 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
10539 		/* All except this subdevice support WOL */
10540 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
10541 			return true;
10542 		break;
10543 	case IXGBE_DEV_ID_82599_KX4:
10544 		return  true;
10545 	default:
10546 		break;
10547 	}
10548 
10549 	return false;
10550 }
10551 
10552 /**
10553  * ixgbe_set_fw_version - Set FW version
10554  * @adapter: the adapter private structure
10555  *
10556  * This function is used by probe and ethtool to determine the FW version to
10557  * format to display. The FW version is taken from the EEPROM/NVM.
10558  */
10559 static void ixgbe_set_fw_version(struct ixgbe_adapter *adapter)
10560 {
10561 	struct ixgbe_hw *hw = &adapter->hw;
10562 	struct ixgbe_nvm_version nvm_ver;
10563 
10564 	ixgbe_get_oem_prod_version(hw, &nvm_ver);
10565 	if (nvm_ver.oem_valid) {
10566 		snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10567 			 "%x.%x.%x", nvm_ver.oem_major, nvm_ver.oem_minor,
10568 			 nvm_ver.oem_release);
10569 		return;
10570 	}
10571 
10572 	ixgbe_get_etk_id(hw, &nvm_ver);
10573 	ixgbe_get_orom_version(hw, &nvm_ver);
10574 
10575 	if (nvm_ver.or_valid) {
10576 		snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10577 			 "0x%08x, %d.%d.%d", nvm_ver.etk_id, nvm_ver.or_major,
10578 			 nvm_ver.or_build, nvm_ver.or_patch);
10579 		return;
10580 	}
10581 
10582 	/* Set ETrack ID format */
10583 	snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10584 		 "0x%08x", nvm_ver.etk_id);
10585 }
10586 
10587 /**
10588  * ixgbe_probe - Device Initialization Routine
10589  * @pdev: PCI device information struct
10590  * @ent: entry in ixgbe_pci_tbl
10591  *
10592  * Returns 0 on success, negative on failure
10593  *
10594  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
10595  * The OS initialization, configuring of the adapter private structure,
10596  * and a hardware reset occur.
10597  **/
10598 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10599 {
10600 	struct net_device *netdev;
10601 	struct ixgbe_adapter *adapter = NULL;
10602 	struct ixgbe_hw *hw;
10603 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
10604 	int i, err, pci_using_dac, expected_gts;
10605 	unsigned int indices = MAX_TX_QUEUES;
10606 	u8 part_str[IXGBE_PBANUM_LENGTH];
10607 	bool disable_dev = false;
10608 #ifdef IXGBE_FCOE
10609 	u16 device_caps;
10610 #endif
10611 	u32 eec;
10612 
10613 	/* Catch broken hardware that put the wrong VF device ID in
10614 	 * the PCIe SR-IOV capability.
10615 	 */
10616 	if (pdev->is_virtfn) {
10617 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
10618 		     pci_name(pdev), pdev->vendor, pdev->device);
10619 		return -EINVAL;
10620 	}
10621 
10622 	err = pci_enable_device_mem(pdev);
10623 	if (err)
10624 		return err;
10625 
10626 	if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
10627 		pci_using_dac = 1;
10628 	} else {
10629 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10630 		if (err) {
10631 			dev_err(&pdev->dev,
10632 				"No usable DMA configuration, aborting\n");
10633 			goto err_dma;
10634 		}
10635 		pci_using_dac = 0;
10636 	}
10637 
10638 	err = pci_request_mem_regions(pdev, ixgbe_driver_name);
10639 	if (err) {
10640 		dev_err(&pdev->dev,
10641 			"pci_request_selected_regions failed 0x%x\n", err);
10642 		goto err_pci_reg;
10643 	}
10644 
10645 	pci_enable_pcie_error_reporting(pdev);
10646 
10647 	pci_set_master(pdev);
10648 	pci_save_state(pdev);
10649 
10650 	if (ii->mac == ixgbe_mac_82598EB) {
10651 #ifdef CONFIG_IXGBE_DCB
10652 		/* 8 TC w/ 4 queues per TC */
10653 		indices = 4 * MAX_TRAFFIC_CLASS;
10654 #else
10655 		indices = IXGBE_MAX_RSS_INDICES;
10656 #endif
10657 	}
10658 
10659 	netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
10660 	if (!netdev) {
10661 		err = -ENOMEM;
10662 		goto err_alloc_etherdev;
10663 	}
10664 
10665 	SET_NETDEV_DEV(netdev, &pdev->dev);
10666 
10667 	adapter = netdev_priv(netdev);
10668 
10669 	adapter->netdev = netdev;
10670 	adapter->pdev = pdev;
10671 	hw = &adapter->hw;
10672 	hw->back = adapter;
10673 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
10674 
10675 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
10676 			      pci_resource_len(pdev, 0));
10677 	adapter->io_addr = hw->hw_addr;
10678 	if (!hw->hw_addr) {
10679 		err = -EIO;
10680 		goto err_ioremap;
10681 	}
10682 
10683 	netdev->netdev_ops = &ixgbe_netdev_ops;
10684 	ixgbe_set_ethtool_ops(netdev);
10685 	netdev->watchdog_timeo = 5 * HZ;
10686 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
10687 
10688 	/* Setup hw api */
10689 	hw->mac.ops   = *ii->mac_ops;
10690 	hw->mac.type  = ii->mac;
10691 	hw->mvals     = ii->mvals;
10692 	if (ii->link_ops)
10693 		hw->link.ops  = *ii->link_ops;
10694 
10695 	/* EEPROM */
10696 	hw->eeprom.ops = *ii->eeprom_ops;
10697 	eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
10698 	if (ixgbe_removed(hw->hw_addr)) {
10699 		err = -EIO;
10700 		goto err_ioremap;
10701 	}
10702 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
10703 	if (!(eec & BIT(8)))
10704 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
10705 
10706 	/* PHY */
10707 	hw->phy.ops = *ii->phy_ops;
10708 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
10709 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
10710 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
10711 	hw->phy.mdio.mmds = 0;
10712 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
10713 	hw->phy.mdio.dev = netdev;
10714 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
10715 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
10716 
10717 	/* setup the private structure */
10718 	err = ixgbe_sw_init(adapter, ii);
10719 	if (err)
10720 		goto err_sw_init;
10721 
10722 	switch (adapter->hw.mac.type) {
10723 	case ixgbe_mac_X550:
10724 	case ixgbe_mac_X550EM_x:
10725 		netdev->udp_tunnel_nic_info = &ixgbe_udp_tunnels_x550;
10726 		break;
10727 	case ixgbe_mac_x550em_a:
10728 		netdev->udp_tunnel_nic_info = &ixgbe_udp_tunnels_x550em_a;
10729 		break;
10730 	default:
10731 		break;
10732 	}
10733 
10734 	/* Make sure the SWFW semaphore is in a valid state */
10735 	if (hw->mac.ops.init_swfw_sync)
10736 		hw->mac.ops.init_swfw_sync(hw);
10737 
10738 	/* Make it possible the adapter to be woken up via WOL */
10739 	switch (adapter->hw.mac.type) {
10740 	case ixgbe_mac_82599EB:
10741 	case ixgbe_mac_X540:
10742 	case ixgbe_mac_X550:
10743 	case ixgbe_mac_X550EM_x:
10744 	case ixgbe_mac_x550em_a:
10745 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10746 		break;
10747 	default:
10748 		break;
10749 	}
10750 
10751 	/*
10752 	 * If there is a fan on this device and it has failed log the
10753 	 * failure.
10754 	 */
10755 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
10756 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
10757 		if (esdp & IXGBE_ESDP_SDP1)
10758 			e_crit(probe, "Fan has stopped, replace the adapter\n");
10759 	}
10760 
10761 	if (allow_unsupported_sfp)
10762 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
10763 
10764 	/* reset_hw fills in the perm_addr as well */
10765 	hw->phy.reset_if_overtemp = true;
10766 	err = hw->mac.ops.reset_hw(hw);
10767 	hw->phy.reset_if_overtemp = false;
10768 	ixgbe_set_eee_capable(adapter);
10769 	if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
10770 		err = 0;
10771 	} else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
10772 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
10773 		e_dev_err("Reload the driver after installing a supported module.\n");
10774 		goto err_sw_init;
10775 	} else if (err) {
10776 		e_dev_err("HW Init failed: %d\n", err);
10777 		goto err_sw_init;
10778 	}
10779 
10780 #ifdef CONFIG_PCI_IOV
10781 	/* SR-IOV not supported on the 82598 */
10782 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
10783 		goto skip_sriov;
10784 	/* Mailbox */
10785 	ixgbe_init_mbx_params_pf(hw);
10786 	hw->mbx.ops = ii->mbx_ops;
10787 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
10788 	ixgbe_enable_sriov(adapter, max_vfs);
10789 skip_sriov:
10790 
10791 #endif
10792 	netdev->features = NETIF_F_SG |
10793 			   NETIF_F_TSO |
10794 			   NETIF_F_TSO6 |
10795 			   NETIF_F_RXHASH |
10796 			   NETIF_F_RXCSUM |
10797 			   NETIF_F_HW_CSUM;
10798 
10799 #define IXGBE_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
10800 				    NETIF_F_GSO_GRE_CSUM | \
10801 				    NETIF_F_GSO_IPXIP4 | \
10802 				    NETIF_F_GSO_IPXIP6 | \
10803 				    NETIF_F_GSO_UDP_TUNNEL | \
10804 				    NETIF_F_GSO_UDP_TUNNEL_CSUM)
10805 
10806 	netdev->gso_partial_features = IXGBE_GSO_PARTIAL_FEATURES;
10807 	netdev->features |= NETIF_F_GSO_PARTIAL |
10808 			    IXGBE_GSO_PARTIAL_FEATURES;
10809 
10810 	if (hw->mac.type >= ixgbe_mac_82599EB)
10811 		netdev->features |= NETIF_F_SCTP_CRC | NETIF_F_GSO_UDP_L4;
10812 
10813 #ifdef CONFIG_IXGBE_IPSEC
10814 #define IXGBE_ESP_FEATURES	(NETIF_F_HW_ESP | \
10815 				 NETIF_F_HW_ESP_TX_CSUM | \
10816 				 NETIF_F_GSO_ESP)
10817 
10818 	if (adapter->ipsec)
10819 		netdev->features |= IXGBE_ESP_FEATURES;
10820 #endif
10821 	/* copy netdev features into list of user selectable features */
10822 	netdev->hw_features |= netdev->features |
10823 			       NETIF_F_HW_VLAN_CTAG_FILTER |
10824 			       NETIF_F_HW_VLAN_CTAG_RX |
10825 			       NETIF_F_HW_VLAN_CTAG_TX |
10826 			       NETIF_F_RXALL |
10827 			       NETIF_F_HW_L2FW_DOFFLOAD;
10828 
10829 	if (hw->mac.type >= ixgbe_mac_82599EB)
10830 		netdev->hw_features |= NETIF_F_NTUPLE |
10831 				       NETIF_F_HW_TC;
10832 
10833 	if (pci_using_dac)
10834 		netdev->features |= NETIF_F_HIGHDMA;
10835 
10836 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
10837 	netdev->hw_enc_features |= netdev->vlan_features;
10838 	netdev->mpls_features |= NETIF_F_SG |
10839 				 NETIF_F_TSO |
10840 				 NETIF_F_TSO6 |
10841 				 NETIF_F_HW_CSUM;
10842 	netdev->mpls_features |= IXGBE_GSO_PARTIAL_FEATURES;
10843 
10844 	/* set this bit last since it cannot be part of vlan_features */
10845 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
10846 			    NETIF_F_HW_VLAN_CTAG_RX |
10847 			    NETIF_F_HW_VLAN_CTAG_TX;
10848 
10849 	netdev->priv_flags |= IFF_UNICAST_FLT;
10850 	netdev->priv_flags |= IFF_SUPP_NOFCS;
10851 
10852 	/* MTU range: 68 - 9710 */
10853 	netdev->min_mtu = ETH_MIN_MTU;
10854 	netdev->max_mtu = IXGBE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN);
10855 
10856 #ifdef CONFIG_IXGBE_DCB
10857 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
10858 		netdev->dcbnl_ops = &ixgbe_dcbnl_ops;
10859 #endif
10860 
10861 #ifdef IXGBE_FCOE
10862 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
10863 		unsigned int fcoe_l;
10864 
10865 		if (hw->mac.ops.get_device_caps) {
10866 			hw->mac.ops.get_device_caps(hw, &device_caps);
10867 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
10868 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
10869 		}
10870 
10871 
10872 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
10873 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
10874 
10875 		netdev->features |= NETIF_F_FSO |
10876 				    NETIF_F_FCOE_CRC;
10877 
10878 		netdev->vlan_features |= NETIF_F_FSO |
10879 					 NETIF_F_FCOE_CRC |
10880 					 NETIF_F_FCOE_MTU;
10881 	}
10882 #endif /* IXGBE_FCOE */
10883 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
10884 		netdev->hw_features |= NETIF_F_LRO;
10885 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
10886 		netdev->features |= NETIF_F_LRO;
10887 
10888 	if (ixgbe_check_fw_error(adapter)) {
10889 		err = -EIO;
10890 		goto err_sw_init;
10891 	}
10892 
10893 	/* make sure the EEPROM is good */
10894 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
10895 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
10896 		err = -EIO;
10897 		goto err_sw_init;
10898 	}
10899 
10900 	eth_platform_get_mac_address(&adapter->pdev->dev,
10901 				     adapter->hw.mac.perm_addr);
10902 
10903 	memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
10904 
10905 	if (!is_valid_ether_addr(netdev->dev_addr)) {
10906 		e_dev_err("invalid MAC address\n");
10907 		err = -EIO;
10908 		goto err_sw_init;
10909 	}
10910 
10911 	/* Set hw->mac.addr to permanent MAC address */
10912 	ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
10913 	ixgbe_mac_set_default_filter(adapter);
10914 
10915 	timer_setup(&adapter->service_timer, ixgbe_service_timer, 0);
10916 
10917 	if (ixgbe_removed(hw->hw_addr)) {
10918 		err = -EIO;
10919 		goto err_sw_init;
10920 	}
10921 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
10922 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
10923 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
10924 
10925 	err = ixgbe_init_interrupt_scheme(adapter);
10926 	if (err)
10927 		goto err_sw_init;
10928 
10929 	for (i = 0; i < adapter->num_rx_queues; i++)
10930 		u64_stats_init(&adapter->rx_ring[i]->syncp);
10931 	for (i = 0; i < adapter->num_tx_queues; i++)
10932 		u64_stats_init(&adapter->tx_ring[i]->syncp);
10933 	for (i = 0; i < adapter->num_xdp_queues; i++)
10934 		u64_stats_init(&adapter->xdp_ring[i]->syncp);
10935 
10936 	/* WOL not supported for all devices */
10937 	adapter->wol = 0;
10938 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
10939 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
10940 						pdev->subsystem_device);
10941 	if (hw->wol_enabled)
10942 		adapter->wol = IXGBE_WUFC_MAG;
10943 
10944 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
10945 
10946 	/* save off EEPROM version number */
10947 	ixgbe_set_fw_version(adapter);
10948 
10949 	/* pick up the PCI bus settings for reporting later */
10950 	if (ixgbe_pcie_from_parent(hw))
10951 		ixgbe_get_parent_bus_info(adapter);
10952 	else
10953 		 hw->mac.ops.get_bus_info(hw);
10954 
10955 	/* calculate the expected PCIe bandwidth required for optimal
10956 	 * performance. Note that some older parts will never have enough
10957 	 * bandwidth due to being older generation PCIe parts. We clamp these
10958 	 * parts to ensure no warning is displayed if it can't be fixed.
10959 	 */
10960 	switch (hw->mac.type) {
10961 	case ixgbe_mac_82598EB:
10962 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
10963 		break;
10964 	default:
10965 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
10966 		break;
10967 	}
10968 
10969 	/* don't check link if we failed to enumerate functions */
10970 	if (expected_gts > 0)
10971 		ixgbe_check_minimum_link(adapter, expected_gts);
10972 
10973 	err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
10974 	if (err)
10975 		strlcpy(part_str, "Unknown", sizeof(part_str));
10976 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
10977 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
10978 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
10979 			   part_str);
10980 	else
10981 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
10982 			   hw->mac.type, hw->phy.type, part_str);
10983 
10984 	e_dev_info("%pM\n", netdev->dev_addr);
10985 
10986 	/* reset the hardware with the new settings */
10987 	err = hw->mac.ops.start_hw(hw);
10988 	if (err == IXGBE_ERR_EEPROM_VERSION) {
10989 		/* We are running on a pre-production device, log a warning */
10990 		e_dev_warn("This device is a pre-production adapter/LOM. "
10991 			   "Please be aware there may be issues associated "
10992 			   "with your hardware.  If you are experiencing "
10993 			   "problems please contact your Intel or hardware "
10994 			   "representative who provided you with this "
10995 			   "hardware.\n");
10996 	}
10997 	strcpy(netdev->name, "eth%d");
10998 	pci_set_drvdata(pdev, adapter);
10999 	err = register_netdev(netdev);
11000 	if (err)
11001 		goto err_register;
11002 
11003 
11004 	/* power down the optics for 82599 SFP+ fiber */
11005 	if (hw->mac.ops.disable_tx_laser)
11006 		hw->mac.ops.disable_tx_laser(hw);
11007 
11008 	/* carrier off reporting is important to ethtool even BEFORE open */
11009 	netif_carrier_off(netdev);
11010 
11011 #ifdef CONFIG_IXGBE_DCA
11012 	if (dca_add_requester(&pdev->dev) == 0) {
11013 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
11014 		ixgbe_setup_dca(adapter);
11015 	}
11016 #endif
11017 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
11018 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
11019 		for (i = 0; i < adapter->num_vfs; i++)
11020 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
11021 	}
11022 
11023 	/* firmware requires driver version to be 0xFFFFFFFF
11024 	 * since os does not support feature
11025 	 */
11026 	if (hw->mac.ops.set_fw_drv_ver)
11027 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF, 0xFF,
11028 					   sizeof(UTS_RELEASE) - 1,
11029 					   UTS_RELEASE);
11030 
11031 	/* add san mac addr to netdev */
11032 	ixgbe_add_sanmac_netdev(netdev);
11033 
11034 	e_dev_info("%s\n", ixgbe_default_device_descr);
11035 
11036 #ifdef CONFIG_IXGBE_HWMON
11037 	if (ixgbe_sysfs_init(adapter))
11038 		e_err(probe, "failed to allocate sysfs resources\n");
11039 #endif /* CONFIG_IXGBE_HWMON */
11040 
11041 	ixgbe_dbg_adapter_init(adapter);
11042 
11043 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
11044 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
11045 		hw->mac.ops.setup_link(hw,
11046 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
11047 			true);
11048 
11049 	err = ixgbe_mii_bus_init(hw);
11050 	if (err)
11051 		goto err_netdev;
11052 
11053 	return 0;
11054 
11055 err_netdev:
11056 	unregister_netdev(netdev);
11057 err_register:
11058 	ixgbe_release_hw_control(adapter);
11059 	ixgbe_clear_interrupt_scheme(adapter);
11060 err_sw_init:
11061 	ixgbe_disable_sriov(adapter);
11062 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
11063 	iounmap(adapter->io_addr);
11064 	kfree(adapter->jump_tables[0]);
11065 	kfree(adapter->mac_table);
11066 	kfree(adapter->rss_key);
11067 	bitmap_free(adapter->af_xdp_zc_qps);
11068 err_ioremap:
11069 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
11070 	free_netdev(netdev);
11071 err_alloc_etherdev:
11072 	pci_release_mem_regions(pdev);
11073 err_pci_reg:
11074 err_dma:
11075 	if (!adapter || disable_dev)
11076 		pci_disable_device(pdev);
11077 	return err;
11078 }
11079 
11080 /**
11081  * ixgbe_remove - Device Removal Routine
11082  * @pdev: PCI device information struct
11083  *
11084  * ixgbe_remove is called by the PCI subsystem to alert the driver
11085  * that it should release a PCI device.  The could be caused by a
11086  * Hot-Plug event, or because the driver is going to be removed from
11087  * memory.
11088  **/
11089 static void ixgbe_remove(struct pci_dev *pdev)
11090 {
11091 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11092 	struct net_device *netdev;
11093 	bool disable_dev;
11094 	int i;
11095 
11096 	/* if !adapter then we already cleaned up in probe */
11097 	if (!adapter)
11098 		return;
11099 
11100 	netdev  = adapter->netdev;
11101 	ixgbe_dbg_adapter_exit(adapter);
11102 
11103 	set_bit(__IXGBE_REMOVING, &adapter->state);
11104 	cancel_work_sync(&adapter->service_task);
11105 
11106 	if (adapter->mii_bus)
11107 		mdiobus_unregister(adapter->mii_bus);
11108 
11109 #ifdef CONFIG_IXGBE_DCA
11110 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
11111 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
11112 		dca_remove_requester(&pdev->dev);
11113 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
11114 				IXGBE_DCA_CTRL_DCA_DISABLE);
11115 	}
11116 
11117 #endif
11118 #ifdef CONFIG_IXGBE_HWMON
11119 	ixgbe_sysfs_exit(adapter);
11120 #endif /* CONFIG_IXGBE_HWMON */
11121 
11122 	/* remove the added san mac */
11123 	ixgbe_del_sanmac_netdev(netdev);
11124 
11125 #ifdef CONFIG_PCI_IOV
11126 	ixgbe_disable_sriov(adapter);
11127 #endif
11128 	if (netdev->reg_state == NETREG_REGISTERED)
11129 		unregister_netdev(netdev);
11130 
11131 	ixgbe_stop_ipsec_offload(adapter);
11132 	ixgbe_clear_interrupt_scheme(adapter);
11133 
11134 	ixgbe_release_hw_control(adapter);
11135 
11136 #ifdef CONFIG_DCB
11137 	kfree(adapter->ixgbe_ieee_pfc);
11138 	kfree(adapter->ixgbe_ieee_ets);
11139 
11140 #endif
11141 	iounmap(adapter->io_addr);
11142 	pci_release_mem_regions(pdev);
11143 
11144 	e_dev_info("complete\n");
11145 
11146 	for (i = 0; i < IXGBE_MAX_LINK_HANDLE; i++) {
11147 		if (adapter->jump_tables[i]) {
11148 			kfree(adapter->jump_tables[i]->input);
11149 			kfree(adapter->jump_tables[i]->mask);
11150 		}
11151 		kfree(adapter->jump_tables[i]);
11152 	}
11153 
11154 	kfree(adapter->mac_table);
11155 	kfree(adapter->rss_key);
11156 	bitmap_free(adapter->af_xdp_zc_qps);
11157 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
11158 	free_netdev(netdev);
11159 
11160 	pci_disable_pcie_error_reporting(pdev);
11161 
11162 	if (disable_dev)
11163 		pci_disable_device(pdev);
11164 }
11165 
11166 /**
11167  * ixgbe_io_error_detected - called when PCI error is detected
11168  * @pdev: Pointer to PCI device
11169  * @state: The current pci connection state
11170  *
11171  * This function is called after a PCI bus error affecting
11172  * this device has been detected.
11173  */
11174 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
11175 						pci_channel_state_t state)
11176 {
11177 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11178 	struct net_device *netdev = adapter->netdev;
11179 
11180 #ifdef CONFIG_PCI_IOV
11181 	struct ixgbe_hw *hw = &adapter->hw;
11182 	struct pci_dev *bdev, *vfdev;
11183 	u32 dw0, dw1, dw2, dw3;
11184 	int vf, pos;
11185 	u16 req_id, pf_func;
11186 
11187 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
11188 	    adapter->num_vfs == 0)
11189 		goto skip_bad_vf_detection;
11190 
11191 	bdev = pdev->bus->self;
11192 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
11193 		bdev = bdev->bus->self;
11194 
11195 	if (!bdev)
11196 		goto skip_bad_vf_detection;
11197 
11198 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
11199 	if (!pos)
11200 		goto skip_bad_vf_detection;
11201 
11202 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
11203 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
11204 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
11205 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
11206 	if (ixgbe_removed(hw->hw_addr))
11207 		goto skip_bad_vf_detection;
11208 
11209 	req_id = dw1 >> 16;
11210 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
11211 	if (!(req_id & 0x0080))
11212 		goto skip_bad_vf_detection;
11213 
11214 	pf_func = req_id & 0x01;
11215 	if ((pf_func & 1) == (pdev->devfn & 1)) {
11216 		unsigned int device_id;
11217 
11218 		vf = (req_id & 0x7F) >> 1;
11219 		e_dev_err("VF %d has caused a PCIe error\n", vf);
11220 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
11221 				"%8.8x\tdw3: %8.8x\n",
11222 		dw0, dw1, dw2, dw3);
11223 		switch (adapter->hw.mac.type) {
11224 		case ixgbe_mac_82599EB:
11225 			device_id = IXGBE_82599_VF_DEVICE_ID;
11226 			break;
11227 		case ixgbe_mac_X540:
11228 			device_id = IXGBE_X540_VF_DEVICE_ID;
11229 			break;
11230 		case ixgbe_mac_X550:
11231 			device_id = IXGBE_DEV_ID_X550_VF;
11232 			break;
11233 		case ixgbe_mac_X550EM_x:
11234 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
11235 			break;
11236 		case ixgbe_mac_x550em_a:
11237 			device_id = IXGBE_DEV_ID_X550EM_A_VF;
11238 			break;
11239 		default:
11240 			device_id = 0;
11241 			break;
11242 		}
11243 
11244 		/* Find the pci device of the offending VF */
11245 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
11246 		while (vfdev) {
11247 			if (vfdev->devfn == (req_id & 0xFF))
11248 				break;
11249 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
11250 					       device_id, vfdev);
11251 		}
11252 		/*
11253 		 * There's a slim chance the VF could have been hot plugged,
11254 		 * so if it is no longer present we don't need to issue the
11255 		 * VFLR.  Just clean up the AER in that case.
11256 		 */
11257 		if (vfdev) {
11258 			pcie_flr(vfdev);
11259 			/* Free device reference count */
11260 			pci_dev_put(vfdev);
11261 		}
11262 	}
11263 
11264 	/*
11265 	 * Even though the error may have occurred on the other port
11266 	 * we still need to increment the vf error reference count for
11267 	 * both ports because the I/O resume function will be called
11268 	 * for both of them.
11269 	 */
11270 	adapter->vferr_refcount++;
11271 
11272 	return PCI_ERS_RESULT_RECOVERED;
11273 
11274 skip_bad_vf_detection:
11275 #endif /* CONFIG_PCI_IOV */
11276 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
11277 		return PCI_ERS_RESULT_DISCONNECT;
11278 
11279 	if (!netif_device_present(netdev))
11280 		return PCI_ERS_RESULT_DISCONNECT;
11281 
11282 	rtnl_lock();
11283 	netif_device_detach(netdev);
11284 
11285 	if (netif_running(netdev))
11286 		ixgbe_close_suspend(adapter);
11287 
11288 	if (state == pci_channel_io_perm_failure) {
11289 		rtnl_unlock();
11290 		return PCI_ERS_RESULT_DISCONNECT;
11291 	}
11292 
11293 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
11294 		pci_disable_device(pdev);
11295 	rtnl_unlock();
11296 
11297 	/* Request a slot reset. */
11298 	return PCI_ERS_RESULT_NEED_RESET;
11299 }
11300 
11301 /**
11302  * ixgbe_io_slot_reset - called after the pci bus has been reset.
11303  * @pdev: Pointer to PCI device
11304  *
11305  * Restart the card from scratch, as if from a cold-boot.
11306  */
11307 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
11308 {
11309 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11310 	pci_ers_result_t result;
11311 
11312 	if (pci_enable_device_mem(pdev)) {
11313 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
11314 		result = PCI_ERS_RESULT_DISCONNECT;
11315 	} else {
11316 		smp_mb__before_atomic();
11317 		clear_bit(__IXGBE_DISABLED, &adapter->state);
11318 		adapter->hw.hw_addr = adapter->io_addr;
11319 		pci_set_master(pdev);
11320 		pci_restore_state(pdev);
11321 		pci_save_state(pdev);
11322 
11323 		pci_wake_from_d3(pdev, false);
11324 
11325 		ixgbe_reset(adapter);
11326 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
11327 		result = PCI_ERS_RESULT_RECOVERED;
11328 	}
11329 
11330 	return result;
11331 }
11332 
11333 /**
11334  * ixgbe_io_resume - called when traffic can start flowing again.
11335  * @pdev: Pointer to PCI device
11336  *
11337  * This callback is called when the error recovery driver tells us that
11338  * its OK to resume normal operation.
11339  */
11340 static void ixgbe_io_resume(struct pci_dev *pdev)
11341 {
11342 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11343 	struct net_device *netdev = adapter->netdev;
11344 
11345 #ifdef CONFIG_PCI_IOV
11346 	if (adapter->vferr_refcount) {
11347 		e_info(drv, "Resuming after VF err\n");
11348 		adapter->vferr_refcount--;
11349 		return;
11350 	}
11351 
11352 #endif
11353 	rtnl_lock();
11354 	if (netif_running(netdev))
11355 		ixgbe_open(netdev);
11356 
11357 	netif_device_attach(netdev);
11358 	rtnl_unlock();
11359 }
11360 
11361 static const struct pci_error_handlers ixgbe_err_handler = {
11362 	.error_detected = ixgbe_io_error_detected,
11363 	.slot_reset = ixgbe_io_slot_reset,
11364 	.resume = ixgbe_io_resume,
11365 };
11366 
11367 static SIMPLE_DEV_PM_OPS(ixgbe_pm_ops, ixgbe_suspend, ixgbe_resume);
11368 
11369 static struct pci_driver ixgbe_driver = {
11370 	.name      = ixgbe_driver_name,
11371 	.id_table  = ixgbe_pci_tbl,
11372 	.probe     = ixgbe_probe,
11373 	.remove    = ixgbe_remove,
11374 	.driver.pm = &ixgbe_pm_ops,
11375 	.shutdown  = ixgbe_shutdown,
11376 	.sriov_configure = ixgbe_pci_sriov_configure,
11377 	.err_handler = &ixgbe_err_handler
11378 };
11379 
11380 /**
11381  * ixgbe_init_module - Driver Registration Routine
11382  *
11383  * ixgbe_init_module is the first routine called when the driver is
11384  * loaded. All it does is register with the PCI subsystem.
11385  **/
11386 static int __init ixgbe_init_module(void)
11387 {
11388 	int ret;
11389 	pr_info("%s\n", ixgbe_driver_string);
11390 	pr_info("%s\n", ixgbe_copyright);
11391 
11392 	ixgbe_wq = create_singlethread_workqueue(ixgbe_driver_name);
11393 	if (!ixgbe_wq) {
11394 		pr_err("%s: Failed to create workqueue\n", ixgbe_driver_name);
11395 		return -ENOMEM;
11396 	}
11397 
11398 	ixgbe_dbg_init();
11399 
11400 	ret = pci_register_driver(&ixgbe_driver);
11401 	if (ret) {
11402 		destroy_workqueue(ixgbe_wq);
11403 		ixgbe_dbg_exit();
11404 		return ret;
11405 	}
11406 
11407 #ifdef CONFIG_IXGBE_DCA
11408 	dca_register_notify(&dca_notifier);
11409 #endif
11410 
11411 	return 0;
11412 }
11413 
11414 module_init(ixgbe_init_module);
11415 
11416 /**
11417  * ixgbe_exit_module - Driver Exit Cleanup Routine
11418  *
11419  * ixgbe_exit_module is called just before the driver is removed
11420  * from memory.
11421  **/
11422 static void __exit ixgbe_exit_module(void)
11423 {
11424 #ifdef CONFIG_IXGBE_DCA
11425 	dca_unregister_notify(&dca_notifier);
11426 #endif
11427 	pci_unregister_driver(&ixgbe_driver);
11428 
11429 	ixgbe_dbg_exit();
11430 	if (ixgbe_wq) {
11431 		destroy_workqueue(ixgbe_wq);
11432 		ixgbe_wq = NULL;
11433 	}
11434 }
11435 
11436 #ifdef CONFIG_IXGBE_DCA
11437 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
11438 			    void *p)
11439 {
11440 	int ret_val;
11441 
11442 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
11443 					 __ixgbe_notify_dca);
11444 
11445 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
11446 }
11447 
11448 #endif /* CONFIG_IXGBE_DCA */
11449 
11450 module_exit(ixgbe_exit_module);
11451 
11452 /* ixgbe_main.c */
11453