xref: /linux/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c (revision 07fdad3a93756b872da7b53647715c48d0f4a2d0)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright(c) 1999 - 2024 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/iopoll.h>
13 #include <linux/ip.h>
14 #include <linux/tcp.h>
15 #include <linux/sctp.h>
16 #include <linux/pkt_sched.h>
17 #include <linux/ipv6.h>
18 #include <linux/slab.h>
19 #include <net/checksum.h>
20 #include <net/ip6_checksum.h>
21 #include <linux/etherdevice.h>
22 #include <linux/ethtool.h>
23 #include <linux/if.h>
24 #include <linux/if_vlan.h>
25 #include <linux/if_macvlan.h>
26 #include <linux/if_bridge.h>
27 #include <linux/prefetch.h>
28 #include <linux/bpf.h>
29 #include <linux/bpf_trace.h>
30 #include <linux/atomic.h>
31 #include <linux/numa.h>
32 #include <generated/utsrelease.h>
33 #include <scsi/fc/fc_fcoe.h>
34 #include <net/udp_tunnel.h>
35 #include <net/pkt_cls.h>
36 #include <net/tc_act/tc_gact.h>
37 #include <net/tc_act/tc_mirred.h>
38 #include <net/vxlan.h>
39 #include <net/mpls.h>
40 #include <net/netdev_queues.h>
41 #include <net/xdp_sock_drv.h>
42 #include <net/xfrm.h>
43 
44 #include "ixgbe.h"
45 #include "ixgbe_common.h"
46 #include "ixgbe_e610.h"
47 #include "ixgbe_dcb_82599.h"
48 #include "ixgbe_mbx.h"
49 #include "ixgbe_phy.h"
50 #include "ixgbe_sriov.h"
51 #include "ixgbe_model.h"
52 #include "ixgbe_txrx_common.h"
53 #include "devlink/devlink.h"
54 
55 char ixgbe_driver_name[] = "ixgbe";
56 static const char ixgbe_driver_string[] =
57 			      "Intel(R) 10 Gigabit PCI Express Network Driver";
58 #ifdef IXGBE_FCOE
59 char ixgbe_default_device_descr[] =
60 			      "Intel(R) 10 Gigabit Network Connection";
61 #else
62 static char ixgbe_default_device_descr[] =
63 			      "Intel(R) 10 Gigabit Network Connection";
64 #endif
65 static const char ixgbe_copyright[] =
66 				"Copyright (c) 1999-2016 Intel Corporation.";
67 
68 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";
69 
70 static const struct ixgbe_info *ixgbe_info_tbl[] = {
71 	[board_82598]		= &ixgbe_82598_info,
72 	[board_82599]		= &ixgbe_82599_info,
73 	[board_X540]		= &ixgbe_X540_info,
74 	[board_X550]		= &ixgbe_X550_info,
75 	[board_X550EM_x]	= &ixgbe_X550EM_x_info,
76 	[board_x550em_x_fw]	= &ixgbe_x550em_x_fw_info,
77 	[board_x550em_a]	= &ixgbe_x550em_a_info,
78 	[board_x550em_a_fw]	= &ixgbe_x550em_a_fw_info,
79 	[board_e610]		= &ixgbe_e610_info,
80 };
81 
82 /* ixgbe_pci_tbl - PCI Device ID Table
83  *
84  * Wildcard entries (PCI_ANY_ID) should come last
85  * Last entry must be all 0s
86  *
87  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
88  *   Class, Class Mask, private data (not used) }
89  */
90 static const struct pci_device_id ixgbe_pci_tbl[] = {
91 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
92 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
93 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
94 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
95 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
96 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
97 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
98 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
99 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
100 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
101 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
102 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
103 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
104 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
105 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
106 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
107 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
108 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
109 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
110 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
111 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
112 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
113 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
114 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
115 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
116 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
117 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
118 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
119 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
120 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
121 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
122 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T1), board_X550},
123 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
124 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_XFI), board_X550EM_x},
125 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
126 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_10G_T), board_X550EM_x},
127 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_SFP), board_X550EM_x},
128 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_1G_T), board_x550em_x_fw},
129 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR), board_x550em_a },
130 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR_L), board_x550em_a },
131 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP_N), board_x550em_a },
132 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII), board_x550em_a },
133 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII_L), board_x550em_a },
134 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_10G_T), board_x550em_a},
135 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP), board_x550em_a },
136 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T), board_x550em_a_fw },
137 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T_L), board_x550em_a_fw },
138 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_E610_BACKPLANE), board_e610},
139 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_E610_SFP), board_e610},
140 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_E610_10G_T), board_e610},
141 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_E610_2_5G_T), board_e610},
142 	{PCI_VDEVICE(INTEL, IXGBE_DEV_ID_E610_SGMII), board_e610},
143 	/* required last entry */
144 	{0, }
145 };
146 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
147 
148 #ifdef CONFIG_IXGBE_DCA
149 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
150 			    void *p);
151 static struct notifier_block dca_notifier = {
152 	.notifier_call = ixgbe_notify_dca,
153 	.next          = NULL,
154 	.priority      = 0
155 };
156 #endif
157 
158 #ifdef CONFIG_PCI_IOV
159 static unsigned int max_vfs;
160 module_param(max_vfs, uint, 0);
161 MODULE_PARM_DESC(max_vfs,
162 		 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
163 #endif /* CONFIG_PCI_IOV */
164 
165 static bool allow_unsupported_sfp;
166 module_param(allow_unsupported_sfp, bool, 0444);
167 MODULE_PARM_DESC(allow_unsupported_sfp,
168 		 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
169 
170 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
171 static int debug = -1;
172 module_param(debug, int, 0);
173 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
174 
175 MODULE_IMPORT_NS("LIBIE_FWLOG");
176 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
177 MODULE_LICENSE("GPL v2");
178 
179 DEFINE_STATIC_KEY_FALSE(ixgbe_xdp_locking_key);
180 EXPORT_SYMBOL(ixgbe_xdp_locking_key);
181 
182 static struct workqueue_struct *ixgbe_wq;
183 
184 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
185 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *);
186 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *);
187 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *);
188 
189 static const struct net_device_ops ixgbe_netdev_ops;
190 
191 static bool netif_is_ixgbe(struct net_device *dev)
192 {
193 	return dev && (dev->netdev_ops == &ixgbe_netdev_ops);
194 }
195 
196 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
197 					  u32 reg, u16 *value)
198 {
199 	struct pci_dev *parent_dev;
200 	struct pci_bus *parent_bus;
201 
202 	parent_bus = adapter->pdev->bus->parent;
203 	if (!parent_bus)
204 		return -1;
205 
206 	parent_dev = parent_bus->self;
207 	if (!parent_dev)
208 		return -1;
209 
210 	if (!pci_is_pcie(parent_dev))
211 		return -1;
212 
213 	pcie_capability_read_word(parent_dev, reg, value);
214 	if (*value == IXGBE_FAILED_READ_CFG_WORD &&
215 	    ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
216 		return -1;
217 	return 0;
218 }
219 
220 static int ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
221 {
222 	struct ixgbe_hw *hw = &adapter->hw;
223 	u16 link_status = 0;
224 	int err;
225 
226 	hw->bus.type = ixgbe_bus_type_pci_express;
227 
228 	/* Get the negotiated link width and speed from PCI config space of the
229 	 * parent, as this device is behind a switch
230 	 */
231 	err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
232 
233 	/* assume caller will handle error case */
234 	if (err)
235 		return err;
236 
237 	hw->bus.width = ixgbe_convert_bus_width(link_status);
238 	hw->bus.speed = ixgbe_convert_bus_speed(link_status);
239 
240 	return 0;
241 }
242 
243 /**
244  * ixgbe_pcie_from_parent - Determine whether PCIe info should come from parent
245  * @hw: hw specific details
246  *
247  * This function is used by probe to determine whether a device's PCI-Express
248  * bandwidth details should be gathered from the parent bus instead of from the
249  * device. Used to ensure that various locations all have the correct device ID
250  * checks.
251  *
252  * Return: true if information should be collected from the parent bus, false
253  *         otherwise
254  */
255 static bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
256 {
257 	switch (hw->device_id) {
258 	case IXGBE_DEV_ID_82599_SFP_SF_QP:
259 	case IXGBE_DEV_ID_82599_QSFP_SF_QP:
260 		return true;
261 	default:
262 		return false;
263 	}
264 }
265 
266 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
267 				     int expected_gts)
268 {
269 	struct ixgbe_hw *hw = &adapter->hw;
270 	struct pci_dev *pdev;
271 
272 	/* Some devices are not connected over PCIe and thus do not negotiate
273 	 * speed. These devices do not have valid bus info, and thus any report
274 	 * we generate may not be correct.
275 	 */
276 	if (hw->bus.type == ixgbe_bus_type_internal)
277 		return;
278 
279 	/* determine whether to use the parent device */
280 	if (ixgbe_pcie_from_parent(&adapter->hw))
281 		pdev = adapter->pdev->bus->parent->self;
282 	else
283 		pdev = adapter->pdev;
284 
285 	pcie_print_link_status(pdev);
286 }
287 
288 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
289 {
290 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
291 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
292 	    !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
293 		queue_work(ixgbe_wq, &adapter->service_task);
294 }
295 
296 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
297 {
298 	struct ixgbe_adapter *adapter = hw->back;
299 
300 	if (!hw->hw_addr)
301 		return;
302 	hw->hw_addr = NULL;
303 	e_dev_err("Adapter removed\n");
304 	if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
305 		ixgbe_service_event_schedule(adapter);
306 }
307 
308 static u32 ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
309 {
310 	u8 __iomem *reg_addr;
311 	u32 value;
312 	int i;
313 
314 	reg_addr = READ_ONCE(hw->hw_addr);
315 	if (ixgbe_removed(reg_addr))
316 		return IXGBE_FAILED_READ_REG;
317 
318 	/* Register read of 0xFFFFFFF can indicate the adapter has been removed,
319 	 * so perform several status register reads to determine if the adapter
320 	 * has been removed.
321 	 */
322 	for (i = 0; i < IXGBE_FAILED_READ_RETRIES; i++) {
323 		value = readl(reg_addr + IXGBE_STATUS);
324 		if (value != IXGBE_FAILED_READ_REG)
325 			break;
326 		mdelay(3);
327 	}
328 
329 	if (value == IXGBE_FAILED_READ_REG)
330 		ixgbe_remove_adapter(hw);
331 	else
332 		value = readl(reg_addr + reg);
333 	return value;
334 }
335 
336 /**
337  * ixgbe_read_reg - Read from device register
338  * @hw: hw specific details
339  * @reg: offset of register to read
340  *
341  * Returns : value read or IXGBE_FAILED_READ_REG if removed
342  *
343  * This function is used to read device registers. It checks for device
344  * removal by confirming any read that returns all ones by checking the
345  * status register value for all ones. This function avoids reading from
346  * the hardware if a removal was previously detected in which case it
347  * returns IXGBE_FAILED_READ_REG (all ones).
348  */
349 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
350 {
351 	u8 __iomem *reg_addr = READ_ONCE(hw->hw_addr);
352 	u32 value;
353 
354 	if (ixgbe_removed(reg_addr))
355 		return IXGBE_FAILED_READ_REG;
356 	if (unlikely(hw->phy.nw_mng_if_sel &
357 		     IXGBE_NW_MNG_IF_SEL_SGMII_ENABLE)) {
358 		struct ixgbe_adapter *adapter;
359 		int i;
360 
361 		for (i = 0; i < 200; ++i) {
362 			value = readl(reg_addr + IXGBE_MAC_SGMII_BUSY);
363 			if (likely(!value))
364 				goto writes_completed;
365 			if (value == IXGBE_FAILED_READ_REG) {
366 				ixgbe_remove_adapter(hw);
367 				return IXGBE_FAILED_READ_REG;
368 			}
369 			udelay(5);
370 		}
371 
372 		adapter = hw->back;
373 		e_warn(hw, "register writes incomplete %08x\n", value);
374 	}
375 
376 writes_completed:
377 	value = readl(reg_addr + reg);
378 	if (unlikely(value == IXGBE_FAILED_READ_REG))
379 		value = ixgbe_check_remove(hw, reg);
380 	return value;
381 }
382 
383 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
384 {
385 	u16 value;
386 
387 	pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
388 	if (value == IXGBE_FAILED_READ_CFG_WORD) {
389 		ixgbe_remove_adapter(hw);
390 		return true;
391 	}
392 	return false;
393 }
394 
395 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
396 {
397 	struct ixgbe_adapter *adapter = hw->back;
398 	u16 value;
399 
400 	if (ixgbe_removed(hw->hw_addr))
401 		return IXGBE_FAILED_READ_CFG_WORD;
402 	pci_read_config_word(adapter->pdev, reg, &value);
403 	if (value == IXGBE_FAILED_READ_CFG_WORD &&
404 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
405 		return IXGBE_FAILED_READ_CFG_WORD;
406 	return value;
407 }
408 
409 #ifdef CONFIG_PCI_IOV
410 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
411 {
412 	struct ixgbe_adapter *adapter = hw->back;
413 	u32 value;
414 
415 	if (ixgbe_removed(hw->hw_addr))
416 		return IXGBE_FAILED_READ_CFG_DWORD;
417 	pci_read_config_dword(adapter->pdev, reg, &value);
418 	if (value == IXGBE_FAILED_READ_CFG_DWORD &&
419 	    ixgbe_check_cfg_remove(hw, adapter->pdev))
420 		return IXGBE_FAILED_READ_CFG_DWORD;
421 	return value;
422 }
423 #endif /* CONFIG_PCI_IOV */
424 
425 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
426 {
427 	struct ixgbe_adapter *adapter = hw->back;
428 
429 	if (ixgbe_removed(hw->hw_addr))
430 		return;
431 	pci_write_config_word(adapter->pdev, reg, value);
432 }
433 
434 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
435 {
436 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
437 
438 	/* flush memory to make sure state is correct before next watchdog */
439 	smp_mb__before_atomic();
440 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
441 }
442 
443 struct ixgbe_reg_info {
444 	u32 ofs;
445 	char *name;
446 };
447 
448 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
449 
450 	/* General Registers */
451 	{IXGBE_CTRL, "CTRL"},
452 	{IXGBE_STATUS, "STATUS"},
453 	{IXGBE_CTRL_EXT, "CTRL_EXT"},
454 
455 	/* Interrupt Registers */
456 	{IXGBE_EICR, "EICR"},
457 
458 	/* RX Registers */
459 	{IXGBE_SRRCTL(0), "SRRCTL"},
460 	{IXGBE_DCA_RXCTRL(0), "DRXCTL"},
461 	{IXGBE_RDLEN(0), "RDLEN"},
462 	{IXGBE_RDH(0), "RDH"},
463 	{IXGBE_RDT(0), "RDT"},
464 	{IXGBE_RXDCTL(0), "RXDCTL"},
465 	{IXGBE_RDBAL(0), "RDBAL"},
466 	{IXGBE_RDBAH(0), "RDBAH"},
467 
468 	/* TX Registers */
469 	{IXGBE_TDBAL(0), "TDBAL"},
470 	{IXGBE_TDBAH(0), "TDBAH"},
471 	{IXGBE_TDLEN(0), "TDLEN"},
472 	{IXGBE_TDH(0), "TDH"},
473 	{IXGBE_TDT(0), "TDT"},
474 	{IXGBE_TXDCTL(0), "TXDCTL"},
475 
476 	/* List Terminator */
477 	{ .name = NULL }
478 };
479 
480 
481 /*
482  * ixgbe_regdump - register printout routine
483  */
484 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
485 {
486 	int i;
487 	char rname[16];
488 	u32 regs[64];
489 
490 	switch (reginfo->ofs) {
491 	case IXGBE_SRRCTL(0):
492 		for (i = 0; i < 64; i++)
493 			regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
494 		break;
495 	case IXGBE_DCA_RXCTRL(0):
496 		for (i = 0; i < 64; i++)
497 			regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
498 		break;
499 	case IXGBE_RDLEN(0):
500 		for (i = 0; i < 64; i++)
501 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
502 		break;
503 	case IXGBE_RDH(0):
504 		for (i = 0; i < 64; i++)
505 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
506 		break;
507 	case IXGBE_RDT(0):
508 		for (i = 0; i < 64; i++)
509 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
510 		break;
511 	case IXGBE_RXDCTL(0):
512 		for (i = 0; i < 64; i++)
513 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
514 		break;
515 	case IXGBE_RDBAL(0):
516 		for (i = 0; i < 64; i++)
517 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
518 		break;
519 	case IXGBE_RDBAH(0):
520 		for (i = 0; i < 64; i++)
521 			regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
522 		break;
523 	case IXGBE_TDBAL(0):
524 		for (i = 0; i < 64; i++)
525 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
526 		break;
527 	case IXGBE_TDBAH(0):
528 		for (i = 0; i < 64; i++)
529 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
530 		break;
531 	case IXGBE_TDLEN(0):
532 		for (i = 0; i < 64; i++)
533 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
534 		break;
535 	case IXGBE_TDH(0):
536 		for (i = 0; i < 64; i++)
537 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
538 		break;
539 	case IXGBE_TDT(0):
540 		for (i = 0; i < 64; i++)
541 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
542 		break;
543 	case IXGBE_TXDCTL(0):
544 		for (i = 0; i < 64; i++)
545 			regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
546 		break;
547 	default:
548 		pr_info("%-15s %08x\n",
549 			reginfo->name, IXGBE_READ_REG(hw, reginfo->ofs));
550 		return;
551 	}
552 
553 	i = 0;
554 	while (i < 64) {
555 		int j;
556 		char buf[9 * 8 + 1];
557 		char *p = buf;
558 
559 		snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i, i + 7);
560 		for (j = 0; j < 8; j++)
561 			p += sprintf(p, " %08x", regs[i++]);
562 		pr_err("%-15s%s\n", rname, buf);
563 	}
564 
565 }
566 
567 static void ixgbe_print_buffer(struct ixgbe_ring *ring, int n)
568 {
569 	struct ixgbe_tx_buffer *tx_buffer;
570 
571 	tx_buffer = &ring->tx_buffer_info[ring->next_to_clean];
572 	pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
573 		n, ring->next_to_use, ring->next_to_clean,
574 		(u64)dma_unmap_addr(tx_buffer, dma),
575 		dma_unmap_len(tx_buffer, len),
576 		tx_buffer->next_to_watch,
577 		(u64)tx_buffer->time_stamp);
578 }
579 
580 /*
581  * ixgbe_dump - Print registers, tx-rings and rx-rings
582  */
583 static void ixgbe_dump(struct ixgbe_adapter *adapter)
584 {
585 	struct net_device *netdev = adapter->netdev;
586 	struct ixgbe_hw *hw = &adapter->hw;
587 	struct ixgbe_reg_info *reginfo;
588 	int n = 0;
589 	struct ixgbe_ring *ring;
590 	struct ixgbe_tx_buffer *tx_buffer;
591 	union ixgbe_adv_tx_desc *tx_desc;
592 	struct my_u0 { u64 a; u64 b; } *u0;
593 	struct ixgbe_ring *rx_ring;
594 	union ixgbe_adv_rx_desc *rx_desc;
595 	struct ixgbe_rx_buffer *rx_buffer_info;
596 	int i = 0;
597 
598 	if (!netif_msg_hw(adapter))
599 		return;
600 
601 	/* Print netdevice Info */
602 	if (netdev) {
603 		dev_info(&adapter->pdev->dev, "Net device Info\n");
604 		pr_info("Device Name     state            "
605 			"trans_start\n");
606 		pr_info("%-15s %016lX %016lX\n",
607 			netdev->name,
608 			netdev->state,
609 			dev_trans_start(netdev));
610 	}
611 
612 	/* Print Registers */
613 	dev_info(&adapter->pdev->dev, "Register Dump\n");
614 	pr_info(" Register Name   Value\n");
615 	for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
616 	     reginfo->name; reginfo++) {
617 		ixgbe_regdump(hw, reginfo);
618 	}
619 
620 	/* Print TX Ring Summary */
621 	if (!netdev || !netif_running(netdev))
622 		return;
623 
624 	dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
625 	pr_info(" %s     %s              %s        %s\n",
626 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
627 		"leng", "ntw", "timestamp");
628 	for (n = 0; n < adapter->num_tx_queues; n++) {
629 		ring = adapter->tx_ring[n];
630 		ixgbe_print_buffer(ring, n);
631 	}
632 
633 	for (n = 0; n < adapter->num_xdp_queues; n++) {
634 		ring = adapter->xdp_ring[n];
635 		ixgbe_print_buffer(ring, n);
636 	}
637 
638 	/* Print TX Rings */
639 	if (!netif_msg_tx_done(adapter))
640 		goto rx_ring_summary;
641 
642 	dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
643 
644 	/* Transmit Descriptor Formats
645 	 *
646 	 * 82598 Advanced Transmit Descriptor
647 	 *   +--------------------------------------------------------------+
648 	 * 0 |         Buffer Address [63:0]                                |
649 	 *   +--------------------------------------------------------------+
650 	 * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
651 	 *   +--------------------------------------------------------------+
652 	 *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
653 	 *
654 	 * 82598 Advanced Transmit Descriptor (Write-Back Format)
655 	 *   +--------------------------------------------------------------+
656 	 * 0 |                          RSV [63:0]                          |
657 	 *   +--------------------------------------------------------------+
658 	 * 8 |            RSV           |  STA  |          NXTSEQ           |
659 	 *   +--------------------------------------------------------------+
660 	 *   63                       36 35   32 31                         0
661 	 *
662 	 * 82599+ Advanced Transmit Descriptor
663 	 *   +--------------------------------------------------------------+
664 	 * 0 |         Buffer Address [63:0]                                |
665 	 *   +--------------------------------------------------------------+
666 	 * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
667 	 *   +--------------------------------------------------------------+
668 	 *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
669 	 *
670 	 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
671 	 *   +--------------------------------------------------------------+
672 	 * 0 |                          RSV [63:0]                          |
673 	 *   +--------------------------------------------------------------+
674 	 * 8 |            RSV           |  STA  |           RSV             |
675 	 *   +--------------------------------------------------------------+
676 	 *   63                       36 35   32 31                         0
677 	 */
678 
679 	for (n = 0; n < adapter->num_tx_queues; n++) {
680 		ring = adapter->tx_ring[n];
681 		pr_info("------------------------------------\n");
682 		pr_info("TX QUEUE INDEX = %d\n", ring->queue_index);
683 		pr_info("------------------------------------\n");
684 		pr_info("%s%s    %s              %s        %s          %s\n",
685 			"T [desc]     [address 63:0  ] ",
686 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
687 			"leng", "ntw", "timestamp", "bi->skb");
688 
689 		for (i = 0; ring->desc && (i < ring->count); i++) {
690 			tx_desc = IXGBE_TX_DESC(ring, i);
691 			tx_buffer = &ring->tx_buffer_info[i];
692 			u0 = (struct my_u0 *)tx_desc;
693 			if (dma_unmap_len(tx_buffer, len) > 0) {
694 				const char *ring_desc;
695 
696 				if (i == ring->next_to_use &&
697 				    i == ring->next_to_clean)
698 					ring_desc = " NTC/U";
699 				else if (i == ring->next_to_use)
700 					ring_desc = " NTU";
701 				else if (i == ring->next_to_clean)
702 					ring_desc = " NTC";
703 				else
704 					ring_desc = "";
705 				pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p%s",
706 					i,
707 					le64_to_cpu((__force __le64)u0->a),
708 					le64_to_cpu((__force __le64)u0->b),
709 					(u64)dma_unmap_addr(tx_buffer, dma),
710 					dma_unmap_len(tx_buffer, len),
711 					tx_buffer->next_to_watch,
712 					(u64)tx_buffer->time_stamp,
713 					tx_buffer->skb,
714 					ring_desc);
715 
716 				if (netif_msg_pktdata(adapter) &&
717 				    tx_buffer->skb)
718 					print_hex_dump(KERN_INFO, "",
719 						DUMP_PREFIX_ADDRESS, 16, 1,
720 						tx_buffer->skb->data,
721 						dma_unmap_len(tx_buffer, len),
722 						true);
723 			}
724 		}
725 	}
726 
727 	/* Print RX Rings Summary */
728 rx_ring_summary:
729 	dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
730 	pr_info("Queue [NTU] [NTC]\n");
731 	for (n = 0; n < adapter->num_rx_queues; n++) {
732 		rx_ring = adapter->rx_ring[n];
733 		pr_info("%5d %5X %5X\n",
734 			n, rx_ring->next_to_use, rx_ring->next_to_clean);
735 	}
736 
737 	/* Print RX Rings */
738 	if (!netif_msg_rx_status(adapter))
739 		return;
740 
741 	dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
742 
743 	/* Receive Descriptor Formats
744 	 *
745 	 * 82598 Advanced Receive Descriptor (Read) Format
746 	 *    63                                           1        0
747 	 *    +-----------------------------------------------------+
748 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
749 	 *    +----------------------------------------------+------+
750 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
751 	 *    +-----------------------------------------------------+
752 	 *
753 	 *
754 	 * 82598 Advanced Receive Descriptor (Write-Back) Format
755 	 *
756 	 *   63       48 47    32 31  30      21 20 16 15   4 3     0
757 	 *   +------------------------------------------------------+
758 	 * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
759 	 *   | Packet   | IP     |   |          |     | Type | Type |
760 	 *   | Checksum | Ident  |   |          |     |      |      |
761 	 *   +------------------------------------------------------+
762 	 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
763 	 *   +------------------------------------------------------+
764 	 *   63       48 47    32 31            20 19               0
765 	 *
766 	 * 82599+ Advanced Receive Descriptor (Read) Format
767 	 *    63                                           1        0
768 	 *    +-----------------------------------------------------+
769 	 *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
770 	 *    +----------------------------------------------+------+
771 	 *  8 |       Header Buffer Address [63:1]           |  DD  |
772 	 *    +-----------------------------------------------------+
773 	 *
774 	 *
775 	 * 82599+ Advanced Receive Descriptor (Write-Back) Format
776 	 *
777 	 *   63       48 47    32 31  30      21 20 17 16   4 3     0
778 	 *   +------------------------------------------------------+
779 	 * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
780 	 *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
781 	 *   |/ Flow Dir Flt ID  |   |          |     |      |      |
782 	 *   +------------------------------------------------------+
783 	 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
784 	 *   +------------------------------------------------------+
785 	 *   63       48 47    32 31          20 19                 0
786 	 */
787 
788 	for (n = 0; n < adapter->num_rx_queues; n++) {
789 		rx_ring = adapter->rx_ring[n];
790 		pr_info("------------------------------------\n");
791 		pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
792 		pr_info("------------------------------------\n");
793 		pr_info("%s%s%s\n",
794 			"R  [desc]      [ PktBuf     A0] ",
795 			"[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
796 			"<-- Adv Rx Read format");
797 		pr_info("%s%s%s\n",
798 			"RWB[desc]      [PcsmIpSHl PtRs] ",
799 			"[vl er S cks ln] ---------------- [bi->skb       ] ",
800 			"<-- Adv Rx Write-Back format");
801 
802 		for (i = 0; i < rx_ring->count; i++) {
803 			const char *ring_desc;
804 
805 			if (i == rx_ring->next_to_use)
806 				ring_desc = " NTU";
807 			else if (i == rx_ring->next_to_clean)
808 				ring_desc = " NTC";
809 			else
810 				ring_desc = "";
811 
812 			rx_buffer_info = &rx_ring->rx_buffer_info[i];
813 			rx_desc = IXGBE_RX_DESC(rx_ring, i);
814 			u0 = (struct my_u0 *)rx_desc;
815 			if (rx_desc->wb.upper.length) {
816 				/* Descriptor Done */
817 				pr_info("RWB[0x%03X]     %016llX %016llX ---------------- %p%s\n",
818 					i,
819 					le64_to_cpu((__force __le64)u0->a),
820 					le64_to_cpu((__force __le64)u0->b),
821 					rx_buffer_info->skb,
822 					ring_desc);
823 			} else {
824 				pr_info("R  [0x%03X]     %016llX %016llX %016llX %p%s\n",
825 					i,
826 					le64_to_cpu((__force __le64)u0->a),
827 					le64_to_cpu((__force __le64)u0->b),
828 					(u64)rx_buffer_info->dma,
829 					rx_buffer_info->skb,
830 					ring_desc);
831 
832 				if (netif_msg_pktdata(adapter) &&
833 				    rx_buffer_info->dma) {
834 					print_hex_dump(KERN_INFO, "",
835 					   DUMP_PREFIX_ADDRESS, 16, 1,
836 					   page_address(rx_buffer_info->page) +
837 						    rx_buffer_info->page_offset,
838 					   ixgbe_rx_bufsz(rx_ring), true);
839 				}
840 			}
841 		}
842 	}
843 }
844 
845 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
846 {
847 	u32 ctrl_ext;
848 
849 	/* Let firmware take over control of h/w */
850 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
851 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
852 			ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
853 }
854 
855 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
856 {
857 	u32 ctrl_ext;
858 
859 	/* Let firmware know the driver has taken over */
860 	ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
861 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
862 			ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
863 }
864 
865 /**
866  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
867  * @adapter: pointer to adapter struct
868  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
869  * @queue: queue to map the corresponding interrupt to
870  * @msix_vector: the vector to map to the corresponding queue
871  *
872  */
873 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
874 			   u8 queue, u8 msix_vector)
875 {
876 	u32 ivar, index;
877 	struct ixgbe_hw *hw = &adapter->hw;
878 	switch (hw->mac.type) {
879 	case ixgbe_mac_82598EB:
880 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
881 		if (direction == -1)
882 			direction = 0;
883 		index = (((direction * 64) + queue) >> 2) & 0x1F;
884 		ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
885 		ivar &= ~(0xFF << (8 * (queue & 0x3)));
886 		ivar |= (msix_vector << (8 * (queue & 0x3)));
887 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
888 		break;
889 	case ixgbe_mac_82599EB:
890 	case ixgbe_mac_X540:
891 	case ixgbe_mac_X550:
892 	case ixgbe_mac_X550EM_x:
893 	case ixgbe_mac_x550em_a:
894 	case ixgbe_mac_e610:
895 		if (direction == -1) {
896 			/* other causes */
897 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
898 			index = ((queue & 1) * 8);
899 			ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
900 			ivar &= ~(0xFF << index);
901 			ivar |= (msix_vector << index);
902 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
903 			break;
904 		} else {
905 			/* tx or rx causes */
906 			msix_vector |= IXGBE_IVAR_ALLOC_VAL;
907 			index = ((16 * (queue & 1)) + (8 * direction));
908 			ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
909 			ivar &= ~(0xFF << index);
910 			ivar |= (msix_vector << index);
911 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
912 			break;
913 		}
914 	default:
915 		break;
916 	}
917 }
918 
919 void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
920 			    u64 qmask)
921 {
922 	u32 mask;
923 
924 	switch (adapter->hw.mac.type) {
925 	case ixgbe_mac_82598EB:
926 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
927 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
928 		break;
929 	case ixgbe_mac_82599EB:
930 	case ixgbe_mac_X540:
931 	case ixgbe_mac_X550:
932 	case ixgbe_mac_X550EM_x:
933 	case ixgbe_mac_x550em_a:
934 	case ixgbe_mac_e610:
935 		mask = (qmask & 0xFFFFFFFF);
936 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
937 		mask = (qmask >> 32);
938 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
939 		break;
940 	default:
941 		break;
942 	}
943 }
944 
945 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
946 {
947 	struct ixgbe_hw *hw = &adapter->hw;
948 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
949 	int i;
950 	u32 data;
951 
952 	if ((hw->fc.current_mode != ixgbe_fc_full) &&
953 	    (hw->fc.current_mode != ixgbe_fc_rx_pause))
954 		return;
955 
956 	switch (hw->mac.type) {
957 	case ixgbe_mac_82598EB:
958 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
959 		break;
960 	default:
961 		data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
962 	}
963 	hwstats->lxoffrxc += data;
964 
965 	/* refill credits (no tx hang) if we received xoff */
966 	if (!data)
967 		return;
968 
969 	for (i = 0; i < adapter->num_tx_queues; i++)
970 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
971 			  &adapter->tx_ring[i]->state);
972 }
973 
974 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
975 {
976 	struct ixgbe_hw *hw = &adapter->hw;
977 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
978 	u32 xoff[8] = {0};
979 	u8 tc;
980 	int i;
981 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
982 
983 	if (adapter->ixgbe_ieee_pfc)
984 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
985 
986 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
987 		ixgbe_update_xoff_rx_lfc(adapter);
988 		return;
989 	}
990 
991 	/* update stats for each tc, only valid with PFC enabled */
992 	for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
993 		u32 pxoffrxc;
994 
995 		switch (hw->mac.type) {
996 		case ixgbe_mac_82598EB:
997 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
998 			break;
999 		default:
1000 			pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
1001 		}
1002 		hwstats->pxoffrxc[i] += pxoffrxc;
1003 		/* Get the TC for given UP */
1004 		tc = netdev_get_prio_tc_map(adapter->netdev, i);
1005 		xoff[tc] += pxoffrxc;
1006 	}
1007 
1008 	/* disarm tx queues that have received xoff frames */
1009 	for (i = 0; i < adapter->num_tx_queues; i++) {
1010 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
1011 
1012 		tc = tx_ring->dcb_tc;
1013 		if (xoff[tc])
1014 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1015 	}
1016 
1017 	for (i = 0; i < adapter->num_xdp_queues; i++) {
1018 		struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
1019 
1020 		tc = xdp_ring->dcb_tc;
1021 		if (xoff[tc])
1022 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &xdp_ring->state);
1023 	}
1024 }
1025 
1026 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1027 {
1028 	return ring->stats.packets;
1029 }
1030 
1031 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1032 {
1033 	unsigned int head, tail;
1034 
1035 	head = ring->next_to_clean;
1036 	tail = ring->next_to_use;
1037 
1038 	return ((head <= tail) ? tail : tail + ring->count) - head;
1039 }
1040 
1041 /**
1042  * ixgbe_get_vf_idx - provide VF index number based on queue index
1043  * @adapter: pointer to the adapter struct
1044  * @queue: Tx queue identifier
1045  * @vf: output VF index
1046  *
1047  * Provide VF index number associated to the input queue.
1048  *
1049  * Returns: 0 if VF provided or error number.
1050  */
1051 static int ixgbe_get_vf_idx(struct ixgbe_adapter *adapter, u16 queue, u16 *vf)
1052 {
1053 	struct ixgbe_hw *hw = &adapter->hw;
1054 	u8 queue_count;
1055 	u32 reg;
1056 
1057 	if (queue >= adapter->num_tx_queues)
1058 		return -EINVAL;
1059 
1060 	/* Determine number of queues by checking
1061 	 * number of virtual functions
1062 	 */
1063 	reg = IXGBE_READ_REG(hw, IXGBE_GCR_EXT);
1064 	switch (reg & IXGBE_GCR_EXT_VT_MODE_MASK) {
1065 	case IXGBE_GCR_EXT_VT_MODE_64:
1066 		queue_count = IXGBE_64VFS_QUEUES;
1067 		break;
1068 	case IXGBE_GCR_EXT_VT_MODE_32:
1069 		queue_count = IXGBE_32VFS_QUEUES;
1070 		break;
1071 	case IXGBE_GCR_EXT_VT_MODE_16:
1072 		queue_count = IXGBE_16VFS_QUEUES;
1073 		break;
1074 	default:
1075 		return -EINVAL;
1076 	}
1077 
1078 	*vf = queue / queue_count;
1079 
1080 	return 0;
1081 }
1082 
1083 static bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1084 {
1085 	u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1086 	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1087 	u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1088 
1089 	clear_check_for_tx_hang(tx_ring);
1090 
1091 	/*
1092 	 * Check for a hung queue, but be thorough. This verifies
1093 	 * that a transmit has been completed since the previous
1094 	 * check AND there is at least one packet pending. The
1095 	 * ARMED bit is set to indicate a potential hang. The
1096 	 * bit is cleared if a pause frame is received to remove
1097 	 * false hang detection due to PFC or 802.3x frames. By
1098 	 * requiring this to fail twice we avoid races with
1099 	 * pfc clearing the ARMED bit and conditions where we
1100 	 * run the check_tx_hang logic with a transmit completion
1101 	 * pending but without time to complete it yet.
1102 	 */
1103 	if (tx_done_old == tx_done && tx_pending)
1104 		/* make sure it is true for two checks in a row */
1105 		return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1106 					&tx_ring->state);
1107 	/* update completed stats and continue */
1108 	tx_ring->tx_stats.tx_done_old = tx_done;
1109 	/* reset the countdown */
1110 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1111 
1112 	return false;
1113 }
1114 
1115 /**
1116  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1117  * @adapter: driver private struct
1118  **/
1119 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1120 {
1121 
1122 	/* Do the reset outside of interrupt context */
1123 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1124 		set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
1125 		e_warn(drv, "initiating reset due to tx timeout\n");
1126 		ixgbe_service_event_schedule(adapter);
1127 	}
1128 }
1129 
1130 /**
1131  * ixgbe_tx_maxrate - callback to set the maximum per-queue bitrate
1132  * @netdev: network interface device structure
1133  * @queue_index: Tx queue to set
1134  * @maxrate: desired maximum transmit bitrate
1135  **/
1136 static int ixgbe_tx_maxrate(struct net_device *netdev,
1137 			    int queue_index, u32 maxrate)
1138 {
1139 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
1140 	struct ixgbe_hw *hw = &adapter->hw;
1141 	u32 bcnrc_val = ixgbe_link_mbps(adapter);
1142 
1143 	if (!maxrate)
1144 		return 0;
1145 
1146 	/* Calculate the rate factor values to set */
1147 	bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1148 	bcnrc_val /= maxrate;
1149 
1150 	/* clear everything but the rate factor */
1151 	bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1152 	IXGBE_RTTBCNRC_RF_DEC_MASK;
1153 
1154 	/* enable the rate scheduler */
1155 	bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1156 
1157 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_index);
1158 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1159 
1160 	return 0;
1161 }
1162 
1163 /**
1164  * ixgbe_update_tx_ring_stats - Update Tx ring specific counters
1165  * @tx_ring: ring to update
1166  * @q_vector: queue vector ring belongs to
1167  * @pkts: number of processed packets
1168  * @bytes: number of processed bytes
1169  */
1170 void ixgbe_update_tx_ring_stats(struct ixgbe_ring *tx_ring,
1171 				struct ixgbe_q_vector *q_vector, u64 pkts,
1172 				u64 bytes)
1173 {
1174 	u64_stats_update_begin(&tx_ring->syncp);
1175 	tx_ring->stats.bytes += bytes;
1176 	tx_ring->stats.packets += pkts;
1177 	u64_stats_update_end(&tx_ring->syncp);
1178 	q_vector->tx.total_bytes += bytes;
1179 	q_vector->tx.total_packets += pkts;
1180 }
1181 
1182 /**
1183  * ixgbe_update_rx_ring_stats - Update Rx ring specific counters
1184  * @rx_ring: ring to update
1185  * @q_vector: queue vector ring belongs to
1186  * @pkts: number of processed packets
1187  * @bytes: number of processed bytes
1188  */
1189 void ixgbe_update_rx_ring_stats(struct ixgbe_ring *rx_ring,
1190 				struct ixgbe_q_vector *q_vector, u64 pkts,
1191 				u64 bytes)
1192 {
1193 	u64_stats_update_begin(&rx_ring->syncp);
1194 	rx_ring->stats.bytes += bytes;
1195 	rx_ring->stats.packets += pkts;
1196 	u64_stats_update_end(&rx_ring->syncp);
1197 	q_vector->rx.total_bytes += bytes;
1198 	q_vector->rx.total_packets += pkts;
1199 }
1200 
1201 /**
1202  * ixgbe_pf_handle_tx_hang - handle Tx hang on PF
1203  * @tx_ring: tx ring number
1204  * @next: next ring
1205  *
1206  * Prints a message containing details about the tx hang.
1207  */
1208 static void ixgbe_pf_handle_tx_hang(struct ixgbe_ring *tx_ring,
1209 				    unsigned int next)
1210 {
1211 	struct ixgbe_adapter *adapter = netdev_priv(tx_ring->netdev);
1212 	struct ixgbe_hw *hw = &adapter->hw;
1213 
1214 	e_err(drv, "Detected Tx Unit Hang\n"
1215 		   "  Tx Queue             <%d>\n"
1216 		   "  TDH, TDT             <%x>, <%x>\n"
1217 		   "  next_to_use          <%x>\n"
1218 		   "  next_to_clean        <%x>\n"
1219 		   "tx_buffer_info[next_to_clean]\n"
1220 		   "  time_stamp           <%lx>\n"
1221 		   "  jiffies              <%lx>\n",
1222 	      tx_ring->queue_index,
1223 	      IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1224 	      IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1225 	      tx_ring->next_to_use, next,
1226 	      tx_ring->tx_buffer_info[next].time_stamp, jiffies);
1227 
1228 	netif_stop_subqueue(tx_ring->netdev,
1229 			    tx_ring->queue_index);
1230 }
1231 
1232 /**
1233  * ixgbe_vf_handle_tx_hang - handle Tx hang on VF
1234  * @adapter: structure containing ring specific data
1235  * @vf: VF index
1236  *
1237  * Print a message containing details about malicious driver detection.
1238  * Set malicious VF link down if the detection happened several times.
1239  */
1240 static void ixgbe_vf_handle_tx_hang(struct ixgbe_adapter *adapter, u16 vf)
1241 {
1242 	struct ixgbe_hw *hw = &adapter->hw;
1243 
1244 	if (adapter->hw.mac.type != ixgbe_mac_e610)
1245 		return;
1246 
1247 	e_warn(drv,
1248 	       "Malicious Driver Detection tx hang detected on PF %d VF %d MAC: %pM",
1249 	       hw->bus.func, vf, adapter->vfinfo[vf].vf_mac_addresses);
1250 
1251 	adapter->tx_hang_count[vf]++;
1252 	if (adapter->tx_hang_count[vf] == IXGBE_MAX_TX_VF_HANGS) {
1253 		ixgbe_set_vf_link_state(adapter, vf,
1254 					IFLA_VF_LINK_STATE_DISABLE);
1255 		adapter->tx_hang_count[vf] = 0;
1256 	}
1257 }
1258 
1259 static u32 ixgbe_poll_tx_icache(struct ixgbe_hw *hw, u16 queue, u16 idx)
1260 {
1261 	IXGBE_WRITE_REG(hw, IXGBE_TXDESCIC, queue * idx);
1262 	return IXGBE_READ_REG(hw, IXGBE_TXDESCIC);
1263 }
1264 
1265 /**
1266  * ixgbe_check_illegal_queue - search for queue with illegal packet
1267  * @adapter: structure containing ring specific data
1268  * @queue: queue index
1269  *
1270  * Check if tx descriptor connected with input queue
1271  * contains illegal packet.
1272  *
1273  * Returns: true if queue contain illegal packet.
1274  */
1275 static bool ixgbe_check_illegal_queue(struct ixgbe_adapter *adapter,
1276 				      u16 queue)
1277 {
1278 	u32 hdr_len_reg, mss_len_reg, type_reg;
1279 	struct ixgbe_hw *hw = &adapter->hw;
1280 	u32 mss_len, header_len, reg;
1281 
1282 	for (u16 i = 0; i < IXGBE_MAX_TX_DESCRIPTORS; i++) {
1283 		/* HW will clear bit IXGBE_TXDESCIC_READY when address
1284 		 * is written to address field. HW will set this bit
1285 		 * when iCache read is done, and data is ready at TIC_DWx.
1286 		 * Set descriptor address.
1287 		 */
1288 		read_poll_timeout(ixgbe_poll_tx_icache, reg,
1289 				  !(reg & IXGBE_TXDESCIC_READY), 0, 0, false,
1290 				  hw, queue, i);
1291 
1292 		/* read tx descriptor access registers */
1293 		hdr_len_reg = IXGBE_READ_REG(hw, IXGBE_TIC_DW2(IXGBE_VLAN_MACIP_LENS_REG));
1294 		type_reg = IXGBE_READ_REG(hw, IXGBE_TIC_DW2(IXGBE_TYPE_TUCMD_MLHL));
1295 		mss_len_reg = IXGBE_READ_REG(hw, IXGBE_TIC_DW2(IXGBE_MSS_L4LEN_IDX));
1296 
1297 		/* check if Advanced Context Descriptor */
1298 		if (FIELD_GET(IXGBE_ADVTXD_DTYP_MASK, type_reg) !=
1299 		    IXGBE_ADVTXD_DTYP_CTXT)
1300 			continue;
1301 
1302 		/* check for illegal MSS and Header length */
1303 		mss_len = FIELD_GET(IXGBE_ADVTXD_MSS_MASK, mss_len_reg);
1304 		header_len = FIELD_GET(IXGBE_ADVTXD_HEADER_LEN_MASK,
1305 				       hdr_len_reg);
1306 		if ((mss_len + header_len) > SZ_16K) {
1307 			e_warn(probe, "mss len + header len too long\n");
1308 			return true;
1309 		}
1310 	}
1311 
1312 	return false;
1313 }
1314 
1315 /**
1316  * ixgbe_handle_mdd_event - handle mdd event
1317  * @adapter: structure containing ring specific data
1318  * @tx_ring: tx descriptor ring to handle
1319  *
1320  * Reset VF driver if malicious vf detected or
1321  * illegal packet in an any queue detected.
1322  */
1323 static void ixgbe_handle_mdd_event(struct ixgbe_adapter *adapter,
1324 				   struct ixgbe_ring *tx_ring)
1325 {
1326 	u16 vf, q;
1327 
1328 	if (adapter->vfinfo && ixgbe_check_mdd_event(adapter)) {
1329 		/* vf mdd info and malicious vf detected */
1330 		if (!ixgbe_get_vf_idx(adapter, tx_ring->queue_index, &vf))
1331 			ixgbe_vf_handle_tx_hang(adapter, vf);
1332 	} else {
1333 		/* malicious vf not detected */
1334 		for (q = 0; q < IXGBE_MAX_TX_QUEUES; q++) {
1335 			if (ixgbe_check_illegal_queue(adapter, q) &&
1336 			    !ixgbe_get_vf_idx(adapter, q, &vf))
1337 				/* illegal queue detected */
1338 				ixgbe_vf_handle_tx_hang(adapter, vf);
1339 		}
1340 	}
1341 }
1342 
1343 /**
1344  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1345  * @q_vector: structure containing interrupt and ring information
1346  * @tx_ring: tx ring to clean
1347  * @napi_budget: Used to determine if we are in netpoll
1348  **/
1349 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1350 			       struct ixgbe_ring *tx_ring, int napi_budget)
1351 {
1352 	struct ixgbe_adapter *adapter = q_vector->adapter;
1353 	struct ixgbe_tx_buffer *tx_buffer;
1354 	union ixgbe_adv_tx_desc *tx_desc;
1355 	unsigned int total_bytes = 0, total_packets = 0, total_ipsec = 0;
1356 	unsigned int budget = q_vector->tx.work_limit;
1357 	unsigned int i = tx_ring->next_to_clean;
1358 	struct netdev_queue *txq;
1359 
1360 	if (test_bit(__IXGBE_DOWN, &adapter->state))
1361 		return true;
1362 
1363 	tx_buffer = &tx_ring->tx_buffer_info[i];
1364 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
1365 	i -= tx_ring->count;
1366 
1367 	do {
1368 		union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1369 
1370 		/* if next_to_watch is not set then there is no work pending */
1371 		if (!eop_desc)
1372 			break;
1373 
1374 		/* prevent any other reads prior to eop_desc */
1375 		smp_rmb();
1376 
1377 		/* if DD is not set pending work has not been completed */
1378 		if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1379 			break;
1380 
1381 		/* clear next_to_watch to prevent false hangs */
1382 		tx_buffer->next_to_watch = NULL;
1383 
1384 		/* update the statistics for this packet */
1385 		total_bytes += tx_buffer->bytecount;
1386 		total_packets += tx_buffer->gso_segs;
1387 		if (tx_buffer->tx_flags & IXGBE_TX_FLAGS_IPSEC)
1388 			total_ipsec++;
1389 
1390 		/* free the skb */
1391 		if (ring_is_xdp(tx_ring))
1392 			xdp_return_frame(tx_buffer->xdpf);
1393 		else
1394 			napi_consume_skb(tx_buffer->skb, napi_budget);
1395 
1396 		/* unmap skb header data */
1397 		dma_unmap_single(tx_ring->dev,
1398 				 dma_unmap_addr(tx_buffer, dma),
1399 				 dma_unmap_len(tx_buffer, len),
1400 				 DMA_TO_DEVICE);
1401 
1402 		/* clear tx_buffer data */
1403 		dma_unmap_len_set(tx_buffer, len, 0);
1404 
1405 		/* unmap remaining buffers */
1406 		while (tx_desc != eop_desc) {
1407 			tx_buffer++;
1408 			tx_desc++;
1409 			i++;
1410 			if (unlikely(!i)) {
1411 				i -= tx_ring->count;
1412 				tx_buffer = tx_ring->tx_buffer_info;
1413 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1414 			}
1415 
1416 			/* unmap any remaining paged data */
1417 			if (dma_unmap_len(tx_buffer, len)) {
1418 				dma_unmap_page(tx_ring->dev,
1419 					       dma_unmap_addr(tx_buffer, dma),
1420 					       dma_unmap_len(tx_buffer, len),
1421 					       DMA_TO_DEVICE);
1422 				dma_unmap_len_set(tx_buffer, len, 0);
1423 			}
1424 		}
1425 
1426 		/* move us one more past the eop_desc for start of next pkt */
1427 		tx_buffer++;
1428 		tx_desc++;
1429 		i++;
1430 		if (unlikely(!i)) {
1431 			i -= tx_ring->count;
1432 			tx_buffer = tx_ring->tx_buffer_info;
1433 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1434 		}
1435 
1436 		/* issue prefetch for next Tx descriptor */
1437 		prefetch(tx_desc);
1438 
1439 		/* update budget accounting */
1440 		budget--;
1441 	} while (likely(budget));
1442 
1443 	i += tx_ring->count;
1444 	tx_ring->next_to_clean = i;
1445 	ixgbe_update_tx_ring_stats(tx_ring, q_vector, total_packets,
1446 				   total_bytes);
1447 	adapter->tx_ipsec += total_ipsec;
1448 
1449 	if (ring_is_xdp(tx_ring))
1450 		return !!budget;
1451 
1452 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1453 		if (adapter->hw.mac.type == ixgbe_mac_e610)
1454 			ixgbe_handle_mdd_event(adapter, tx_ring);
1455 
1456 		ixgbe_pf_handle_tx_hang(tx_ring, i);
1457 
1458 		e_info(probe,
1459 		       "tx hang %d detected on queue %d, resetting adapter\n",
1460 			adapter->tx_timeout_count + 1, tx_ring->queue_index);
1461 
1462 		/* schedule immediate reset if we believe we hung */
1463 		ixgbe_tx_timeout_reset(adapter);
1464 
1465 		/* the adapter is about to reset, no point in enabling stuff */
1466 		return true;
1467 	}
1468 
1469 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1470 	txq = netdev_get_tx_queue(tx_ring->netdev, tx_ring->queue_index);
1471 	if (!__netif_txq_completed_wake(txq, total_packets, total_bytes,
1472 					ixgbe_desc_unused(tx_ring),
1473 					TX_WAKE_THRESHOLD,
1474 					!netif_carrier_ok(tx_ring->netdev) ||
1475 					test_bit(__IXGBE_DOWN, &adapter->state)))
1476 		++tx_ring->tx_stats.restart_queue;
1477 
1478 	return !!budget;
1479 }
1480 
1481 #ifdef CONFIG_IXGBE_DCA
1482 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1483 				struct ixgbe_ring *tx_ring,
1484 				int cpu)
1485 {
1486 	struct ixgbe_hw *hw = &adapter->hw;
1487 	u32 txctrl = 0;
1488 	u16 reg_offset;
1489 
1490 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1491 		txctrl = dca3_get_tag(tx_ring->dev, cpu);
1492 
1493 	switch (hw->mac.type) {
1494 	case ixgbe_mac_82598EB:
1495 		reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1496 		break;
1497 	case ixgbe_mac_82599EB:
1498 	case ixgbe_mac_X540:
1499 		reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1500 		txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1501 		break;
1502 	default:
1503 		/* for unknown hardware do not write register */
1504 		return;
1505 	}
1506 
1507 	/*
1508 	 * We can enable relaxed ordering for reads, but not writes when
1509 	 * DCA is enabled.  This is due to a known issue in some chipsets
1510 	 * which will cause the DCA tag to be cleared.
1511 	 */
1512 	txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1513 		  IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1514 		  IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1515 
1516 	IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1517 }
1518 
1519 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1520 				struct ixgbe_ring *rx_ring,
1521 				int cpu)
1522 {
1523 	struct ixgbe_hw *hw = &adapter->hw;
1524 	u32 rxctrl = 0;
1525 	u8 reg_idx = rx_ring->reg_idx;
1526 
1527 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1528 		rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1529 
1530 	switch (hw->mac.type) {
1531 	case ixgbe_mac_82599EB:
1532 	case ixgbe_mac_X540:
1533 		rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1534 		break;
1535 	default:
1536 		break;
1537 	}
1538 
1539 	/*
1540 	 * We can enable relaxed ordering for reads, but not writes when
1541 	 * DCA is enabled.  This is due to a known issue in some chipsets
1542 	 * which will cause the DCA tag to be cleared.
1543 	 */
1544 	rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1545 		  IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1546 		  IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1547 
1548 	IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1549 }
1550 
1551 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1552 {
1553 	struct ixgbe_adapter *adapter = q_vector->adapter;
1554 	struct ixgbe_ring *ring;
1555 	int cpu = get_cpu();
1556 
1557 	if (q_vector->cpu == cpu)
1558 		goto out_no_update;
1559 
1560 	ixgbe_for_each_ring(ring, q_vector->tx)
1561 		ixgbe_update_tx_dca(adapter, ring, cpu);
1562 
1563 	ixgbe_for_each_ring(ring, q_vector->rx)
1564 		ixgbe_update_rx_dca(adapter, ring, cpu);
1565 
1566 	q_vector->cpu = cpu;
1567 out_no_update:
1568 	put_cpu();
1569 }
1570 
1571 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1572 {
1573 	int i;
1574 
1575 	/* always use CB2 mode, difference is masked in the CB driver */
1576 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1577 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1578 				IXGBE_DCA_CTRL_DCA_MODE_CB2);
1579 	else
1580 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1581 				IXGBE_DCA_CTRL_DCA_DISABLE);
1582 
1583 	for (i = 0; i < adapter->num_q_vectors; i++) {
1584 		adapter->q_vector[i]->cpu = -1;
1585 		ixgbe_update_dca(adapter->q_vector[i]);
1586 	}
1587 }
1588 
1589 static int __ixgbe_notify_dca(struct device *dev, void *data)
1590 {
1591 	struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1592 	unsigned long event = *(unsigned long *)data;
1593 
1594 	if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1595 		return 0;
1596 
1597 	switch (event) {
1598 	case DCA_PROVIDER_ADD:
1599 		/* if we're already enabled, don't do it again */
1600 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1601 			break;
1602 		if (dca_add_requester(dev) == 0) {
1603 			adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1604 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1605 					IXGBE_DCA_CTRL_DCA_MODE_CB2);
1606 			break;
1607 		}
1608 		fallthrough; /* DCA is disabled. */
1609 	case DCA_PROVIDER_REMOVE:
1610 		if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1611 			dca_remove_requester(dev);
1612 			adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1613 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1614 					IXGBE_DCA_CTRL_DCA_DISABLE);
1615 		}
1616 		break;
1617 	}
1618 
1619 	return 0;
1620 }
1621 
1622 #endif /* CONFIG_IXGBE_DCA */
1623 
1624 #define IXGBE_RSS_L4_TYPES_MASK \
1625 	((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1626 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1627 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1628 	 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1629 
1630 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1631 				 union ixgbe_adv_rx_desc *rx_desc,
1632 				 struct sk_buff *skb)
1633 {
1634 	u16 rss_type;
1635 
1636 	if (!(ring->netdev->features & NETIF_F_RXHASH))
1637 		return;
1638 
1639 	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1640 		   IXGBE_RXDADV_RSSTYPE_MASK;
1641 
1642 	if (!rss_type)
1643 		return;
1644 
1645 	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1646 		     (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1647 		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1648 }
1649 
1650 #ifdef IXGBE_FCOE
1651 /**
1652  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1653  * @ring: structure containing ring specific data
1654  * @rx_desc: advanced rx descriptor
1655  *
1656  * Returns : true if it is FCoE pkt
1657  */
1658 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1659 				    union ixgbe_adv_rx_desc *rx_desc)
1660 {
1661 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1662 
1663 	return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1664 	       ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1665 		(cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1666 			     IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1667 }
1668 
1669 #endif /* IXGBE_FCOE */
1670 /**
1671  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1672  * @ring: structure containing ring specific data
1673  * @rx_desc: current Rx descriptor being processed
1674  * @skb: skb currently being received and modified
1675  **/
1676 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1677 				     union ixgbe_adv_rx_desc *rx_desc,
1678 				     struct sk_buff *skb)
1679 {
1680 	__le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1681 	bool encap_pkt = false;
1682 
1683 	skb_checksum_none_assert(skb);
1684 
1685 	/* Rx csum disabled */
1686 	if (!(ring->netdev->features & NETIF_F_RXCSUM))
1687 		return;
1688 
1689 	/* check for VXLAN and Geneve packets */
1690 	if (pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) {
1691 		encap_pkt = true;
1692 		skb->encapsulation = 1;
1693 	}
1694 
1695 	/* if IP and error */
1696 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1697 	    ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1698 		ring->rx_stats.csum_err++;
1699 		return;
1700 	}
1701 
1702 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1703 		return;
1704 
1705 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1706 		/*
1707 		 * 82599 errata, UDP frames with a 0 checksum can be marked as
1708 		 * checksum errors.
1709 		 */
1710 		if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1711 		    test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1712 			return;
1713 
1714 		ring->rx_stats.csum_err++;
1715 		return;
1716 	}
1717 
1718 	/* It must be a TCP or UDP packet with a valid checksum */
1719 	skb->ip_summed = CHECKSUM_UNNECESSARY;
1720 	if (encap_pkt) {
1721 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1722 			return;
1723 
1724 		if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1725 			skb->ip_summed = CHECKSUM_NONE;
1726 			return;
1727 		}
1728 		/* If we checked the outer header let the stack know */
1729 		skb->csum_level = 1;
1730 	}
1731 }
1732 
1733 static unsigned int ixgbe_rx_offset(struct ixgbe_ring *rx_ring)
1734 {
1735 	return ring_uses_build_skb(rx_ring) ? IXGBE_SKB_PAD : 0;
1736 }
1737 
1738 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1739 				    struct ixgbe_rx_buffer *bi)
1740 {
1741 	struct page *page = bi->page;
1742 	dma_addr_t dma;
1743 
1744 	/* since we are recycling buffers we should seldom need to alloc */
1745 	if (likely(page))
1746 		return true;
1747 
1748 	/* alloc new page for storage */
1749 	page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1750 	if (unlikely(!page)) {
1751 		rx_ring->rx_stats.alloc_rx_page_failed++;
1752 		return false;
1753 	}
1754 
1755 	/* map page for use */
1756 	dma = dma_map_page_attrs(rx_ring->dev, page, 0,
1757 				 ixgbe_rx_pg_size(rx_ring),
1758 				 DMA_FROM_DEVICE,
1759 				 IXGBE_RX_DMA_ATTR);
1760 
1761 	/*
1762 	 * if mapping failed free memory back to system since
1763 	 * there isn't much point in holding memory we can't use
1764 	 */
1765 	if (dma_mapping_error(rx_ring->dev, dma)) {
1766 		__free_pages(page, ixgbe_rx_pg_order(rx_ring));
1767 
1768 		rx_ring->rx_stats.alloc_rx_page_failed++;
1769 		return false;
1770 	}
1771 
1772 	bi->dma = dma;
1773 	bi->page = page;
1774 	bi->page_offset = rx_ring->rx_offset;
1775 	page_ref_add(page, USHRT_MAX - 1);
1776 	bi->pagecnt_bias = USHRT_MAX;
1777 	rx_ring->rx_stats.alloc_rx_page++;
1778 
1779 	return true;
1780 }
1781 
1782 /**
1783  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1784  * @rx_ring: ring to place buffers on
1785  * @cleaned_count: number of buffers to replace
1786  **/
1787 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1788 {
1789 	union ixgbe_adv_rx_desc *rx_desc;
1790 	struct ixgbe_rx_buffer *bi;
1791 	u16 i = rx_ring->next_to_use;
1792 	u16 bufsz;
1793 
1794 	/* nothing to do */
1795 	if (!cleaned_count)
1796 		return;
1797 
1798 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
1799 	bi = &rx_ring->rx_buffer_info[i];
1800 	i -= rx_ring->count;
1801 
1802 	bufsz = ixgbe_rx_bufsz(rx_ring);
1803 
1804 	do {
1805 		if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1806 			break;
1807 
1808 		/* sync the buffer for use by the device */
1809 		dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
1810 						 bi->page_offset, bufsz,
1811 						 DMA_FROM_DEVICE);
1812 
1813 		/*
1814 		 * Refresh the desc even if buffer_addrs didn't change
1815 		 * because each write-back erases this info.
1816 		 */
1817 		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1818 
1819 		rx_desc++;
1820 		bi++;
1821 		i++;
1822 		if (unlikely(!i)) {
1823 			rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1824 			bi = rx_ring->rx_buffer_info;
1825 			i -= rx_ring->count;
1826 		}
1827 
1828 		/* clear the length for the next_to_use descriptor */
1829 		rx_desc->wb.upper.length = 0;
1830 
1831 		cleaned_count--;
1832 	} while (cleaned_count);
1833 
1834 	i += rx_ring->count;
1835 
1836 	if (rx_ring->next_to_use != i) {
1837 		rx_ring->next_to_use = i;
1838 
1839 		/* update next to alloc since we have filled the ring */
1840 		rx_ring->next_to_alloc = i;
1841 
1842 		/* Force memory writes to complete before letting h/w
1843 		 * know there are new descriptors to fetch.  (Only
1844 		 * applicable for weak-ordered memory model archs,
1845 		 * such as IA-64).
1846 		 */
1847 		wmb();
1848 		writel(i, rx_ring->tail);
1849 	}
1850 }
1851 
1852 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1853 				   struct sk_buff *skb)
1854 {
1855 	u16 hdr_len = skb_headlen(skb);
1856 
1857 	/* set gso_size to avoid messing up TCP MSS */
1858 	skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1859 						 IXGBE_CB(skb)->append_cnt);
1860 	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1861 }
1862 
1863 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1864 				   struct sk_buff *skb)
1865 {
1866 	/* if append_cnt is 0 then frame is not RSC */
1867 	if (!IXGBE_CB(skb)->append_cnt)
1868 		return;
1869 
1870 	rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1871 	rx_ring->rx_stats.rsc_flush++;
1872 
1873 	ixgbe_set_rsc_gso_size(rx_ring, skb);
1874 
1875 	/* gso_size is computed using append_cnt so always clear it last */
1876 	IXGBE_CB(skb)->append_cnt = 0;
1877 }
1878 
1879 /**
1880  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1881  * @rx_ring: rx descriptor ring packet is being transacted on
1882  * @rx_desc: pointer to the EOP Rx descriptor
1883  * @skb: pointer to current skb being populated
1884  *
1885  * This function checks the ring, descriptor, and packet information in
1886  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1887  * other fields within the skb.
1888  **/
1889 void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1890 			      union ixgbe_adv_rx_desc *rx_desc,
1891 			      struct sk_buff *skb)
1892 {
1893 	struct net_device *dev = rx_ring->netdev;
1894 	u32 flags = rx_ring->q_vector->adapter->flags;
1895 
1896 	ixgbe_update_rsc_stats(rx_ring, skb);
1897 
1898 	ixgbe_rx_hash(rx_ring, rx_desc, skb);
1899 
1900 	ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1901 
1902 	if (unlikely(flags & IXGBE_FLAG_RX_HWTSTAMP_ENABLED))
1903 		ixgbe_ptp_rx_hwtstamp(rx_ring, rx_desc, skb);
1904 
1905 	if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1906 	    ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1907 		u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1908 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1909 	}
1910 
1911 	if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_SECP))
1912 		ixgbe_ipsec_rx(rx_ring, rx_desc, skb);
1913 
1914 	/* record Rx queue, or update MACVLAN statistics */
1915 	if (netif_is_ixgbe(dev))
1916 		skb_record_rx_queue(skb, rx_ring->queue_index);
1917 	else
1918 		macvlan_count_rx(netdev_priv(dev), skb->len + ETH_HLEN, true,
1919 				 false);
1920 
1921 	skb->protocol = eth_type_trans(skb, dev);
1922 }
1923 
1924 void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1925 		  struct sk_buff *skb)
1926 {
1927 	napi_gro_receive(&q_vector->napi, skb);
1928 }
1929 
1930 /**
1931  * ixgbe_is_non_eop - process handling of non-EOP buffers
1932  * @rx_ring: Rx ring being processed
1933  * @rx_desc: Rx descriptor for current buffer
1934  * @skb: Current socket buffer containing buffer in progress
1935  *
1936  * This function updates next to clean.  If the buffer is an EOP buffer
1937  * this function exits returning false, otherwise it will place the
1938  * sk_buff in the next buffer to be chained and return true indicating
1939  * that this is in fact a non-EOP buffer.
1940  **/
1941 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1942 			     union ixgbe_adv_rx_desc *rx_desc,
1943 			     struct sk_buff *skb)
1944 {
1945 	u32 ntc = rx_ring->next_to_clean + 1;
1946 
1947 	/* fetch, update, and store next to clean */
1948 	ntc = (ntc < rx_ring->count) ? ntc : 0;
1949 	rx_ring->next_to_clean = ntc;
1950 
1951 	prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1952 
1953 	/* update RSC append count if present */
1954 	if (ring_is_rsc_enabled(rx_ring)) {
1955 		__le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1956 				     cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1957 
1958 		if (unlikely(rsc_enabled)) {
1959 			u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1960 
1961 			rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1962 			IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1963 
1964 			/* update ntc based on RSC value */
1965 			ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1966 			ntc &= IXGBE_RXDADV_NEXTP_MASK;
1967 			ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1968 		}
1969 	}
1970 
1971 	/* if we are the last buffer then there is nothing else to do */
1972 	if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1973 		return false;
1974 
1975 	/* place skb in next buffer to be received */
1976 	rx_ring->rx_buffer_info[ntc].skb = skb;
1977 	rx_ring->rx_stats.non_eop_descs++;
1978 
1979 	return true;
1980 }
1981 
1982 /**
1983  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1984  * @rx_ring: rx descriptor ring packet is being transacted on
1985  * @skb: pointer to current skb being adjusted
1986  *
1987  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1988  * main difference between this version and the original function is that
1989  * this function can make several assumptions about the state of things
1990  * that allow for significant optimizations versus the standard function.
1991  * As a result we can do things like drop a frag and maintain an accurate
1992  * truesize for the skb.
1993  */
1994 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1995 			    struct sk_buff *skb)
1996 {
1997 	skb_frag_t *frag = &skb_shinfo(skb)->frags[0];
1998 	unsigned char *va;
1999 	unsigned int pull_len;
2000 
2001 	/*
2002 	 * it is valid to use page_address instead of kmap since we are
2003 	 * working with pages allocated out of the lomem pool per
2004 	 * alloc_page(GFP_ATOMIC)
2005 	 */
2006 	va = skb_frag_address(frag);
2007 
2008 	/*
2009 	 * we need the header to contain the greater of either ETH_HLEN or
2010 	 * 60 bytes if the skb->len is less than 60 for skb_pad.
2011 	 */
2012 	pull_len = eth_get_headlen(skb->dev, va, IXGBE_RX_HDR_SIZE);
2013 
2014 	/* align pull length to size of long to optimize memcpy performance */
2015 	skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
2016 
2017 	/* update all of the pointers */
2018 	skb_frag_size_sub(frag, pull_len);
2019 	skb_frag_off_add(frag, pull_len);
2020 	skb->data_len -= pull_len;
2021 	skb->tail += pull_len;
2022 }
2023 
2024 /**
2025  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
2026  * @rx_ring: rx descriptor ring packet is being transacted on
2027  * @skb: pointer to current skb being updated
2028  *
2029  * This function provides a basic DMA sync up for the first fragment of an
2030  * skb.  The reason for doing this is that the first fragment cannot be
2031  * unmapped until we have reached the end of packet descriptor for a buffer
2032  * chain.
2033  */
2034 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
2035 				struct sk_buff *skb)
2036 {
2037 	if (ring_uses_build_skb(rx_ring)) {
2038 		unsigned long mask = (unsigned long)ixgbe_rx_pg_size(rx_ring) - 1;
2039 		unsigned long offset = (unsigned long)(skb->data) & mask;
2040 
2041 		dma_sync_single_range_for_cpu(rx_ring->dev,
2042 					      IXGBE_CB(skb)->dma,
2043 					      offset,
2044 					      skb_headlen(skb),
2045 					      DMA_FROM_DEVICE);
2046 	} else {
2047 		skb_frag_t *frag = &skb_shinfo(skb)->frags[0];
2048 
2049 		dma_sync_single_range_for_cpu(rx_ring->dev,
2050 					      IXGBE_CB(skb)->dma,
2051 					      skb_frag_off(frag),
2052 					      skb_frag_size(frag),
2053 					      DMA_FROM_DEVICE);
2054 	}
2055 
2056 	/* If the page was released, just unmap it. */
2057 	if (unlikely(IXGBE_CB(skb)->page_released)) {
2058 		dma_unmap_page_attrs(rx_ring->dev, IXGBE_CB(skb)->dma,
2059 				     ixgbe_rx_pg_size(rx_ring),
2060 				     DMA_FROM_DEVICE,
2061 				     IXGBE_RX_DMA_ATTR);
2062 	}
2063 }
2064 
2065 /**
2066  * ixgbe_cleanup_headers - Correct corrupted or empty headers
2067  * @rx_ring: rx descriptor ring packet is being transacted on
2068  * @rx_desc: pointer to the EOP Rx descriptor
2069  * @skb: pointer to current skb being fixed
2070  *
2071  * Check if the skb is valid in the XDP case it will be an error pointer.
2072  * Return true in this case to abort processing and advance to next
2073  * descriptor.
2074  *
2075  * Check for corrupted packet headers caused by senders on the local L2
2076  * embedded NIC switch not setting up their Tx Descriptors right.  These
2077  * should be very rare.
2078  *
2079  * Also address the case where we are pulling data in on pages only
2080  * and as such no data is present in the skb header.
2081  *
2082  * In addition if skb is not at least 60 bytes we need to pad it so that
2083  * it is large enough to qualify as a valid Ethernet frame.
2084  *
2085  * Returns true if an error was encountered and skb was freed.
2086  **/
2087 bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
2088 			   union ixgbe_adv_rx_desc *rx_desc,
2089 			   struct sk_buff *skb)
2090 {
2091 	struct net_device *netdev = rx_ring->netdev;
2092 
2093 	/* Verify netdev is present, and that packet does not have any
2094 	 * errors that would be unacceptable to the netdev.
2095 	 */
2096 	if (!netdev ||
2097 	    (unlikely(ixgbe_test_staterr(rx_desc,
2098 					 IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
2099 	     !(netdev->features & NETIF_F_RXALL)))) {
2100 		dev_kfree_skb_any(skb);
2101 		return true;
2102 	}
2103 
2104 	/* place header in linear portion of buffer */
2105 	if (!skb_headlen(skb))
2106 		ixgbe_pull_tail(rx_ring, skb);
2107 
2108 #ifdef IXGBE_FCOE
2109 	/* do not attempt to pad FCoE Frames as this will disrupt DDP */
2110 	if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
2111 		return false;
2112 
2113 #endif
2114 	/* if eth_skb_pad returns an error the skb was freed */
2115 	if (eth_skb_pad(skb))
2116 		return true;
2117 
2118 	return false;
2119 }
2120 
2121 /**
2122  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
2123  * @rx_ring: rx descriptor ring to store buffers on
2124  * @old_buff: donor buffer to have page reused
2125  *
2126  * Synchronizes page for reuse by the adapter
2127  **/
2128 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
2129 				struct ixgbe_rx_buffer *old_buff)
2130 {
2131 	struct ixgbe_rx_buffer *new_buff;
2132 	u16 nta = rx_ring->next_to_alloc;
2133 
2134 	new_buff = &rx_ring->rx_buffer_info[nta];
2135 
2136 	/* update, and store next to alloc */
2137 	nta++;
2138 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
2139 
2140 	/* Transfer page from old buffer to new buffer.
2141 	 * Move each member individually to avoid possible store
2142 	 * forwarding stalls and unnecessary copy of skb.
2143 	 */
2144 	new_buff->dma		= old_buff->dma;
2145 	new_buff->page		= old_buff->page;
2146 	new_buff->page_offset	= old_buff->page_offset;
2147 	new_buff->pagecnt_bias	= old_buff->pagecnt_bias;
2148 }
2149 
2150 static bool ixgbe_can_reuse_rx_page(struct ixgbe_rx_buffer *rx_buffer,
2151 				    int rx_buffer_pgcnt)
2152 {
2153 	unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
2154 	struct page *page = rx_buffer->page;
2155 
2156 	/* avoid re-using remote and pfmemalloc pages */
2157 	if (!dev_page_is_reusable(page))
2158 		return false;
2159 
2160 #if (PAGE_SIZE < 8192)
2161 	/* if we are only owner of page we can reuse it */
2162 	if (unlikely((rx_buffer_pgcnt - pagecnt_bias) > 1))
2163 		return false;
2164 #else
2165 	/* The last offset is a bit aggressive in that we assume the
2166 	 * worst case of FCoE being enabled and using a 3K buffer.
2167 	 * However this should have minimal impact as the 1K extra is
2168 	 * still less than one buffer in size.
2169 	 */
2170 #define IXGBE_LAST_OFFSET \
2171 	(SKB_WITH_OVERHEAD(PAGE_SIZE) - IXGBE_RXBUFFER_3K)
2172 	if (rx_buffer->page_offset > IXGBE_LAST_OFFSET)
2173 		return false;
2174 #endif
2175 
2176 	/* If we have drained the page fragment pool we need to update
2177 	 * the pagecnt_bias and page count so that we fully restock the
2178 	 * number of references the driver holds.
2179 	 */
2180 	if (unlikely(pagecnt_bias == 1)) {
2181 		page_ref_add(page, USHRT_MAX - 1);
2182 		rx_buffer->pagecnt_bias = USHRT_MAX;
2183 	}
2184 
2185 	return true;
2186 }
2187 
2188 /**
2189  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
2190  * @rx_ring: rx descriptor ring to transact packets on
2191  * @rx_buffer: buffer containing page to add
2192  * @skb: sk_buff to place the data into
2193  * @size: size of data in rx_buffer
2194  *
2195  * This function will add the data contained in rx_buffer->page to the skb.
2196  * This is done either through a direct copy if the data in the buffer is
2197  * less than the skb header size, otherwise it will just attach the page as
2198  * a frag to the skb.
2199  *
2200  * The function will then update the page offset if necessary and return
2201  * true if the buffer can be reused by the adapter.
2202  **/
2203 static void ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
2204 			      struct ixgbe_rx_buffer *rx_buffer,
2205 			      struct sk_buff *skb,
2206 			      unsigned int size)
2207 {
2208 #if (PAGE_SIZE < 8192)
2209 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2210 #else
2211 	unsigned int truesize = rx_ring->rx_offset ?
2212 				SKB_DATA_ALIGN(rx_ring->rx_offset + size) :
2213 				SKB_DATA_ALIGN(size);
2214 #endif
2215 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
2216 			rx_buffer->page_offset, size, truesize);
2217 #if (PAGE_SIZE < 8192)
2218 	rx_buffer->page_offset ^= truesize;
2219 #else
2220 	rx_buffer->page_offset += truesize;
2221 #endif
2222 }
2223 
2224 static struct ixgbe_rx_buffer *ixgbe_get_rx_buffer(struct ixgbe_ring *rx_ring,
2225 						   union ixgbe_adv_rx_desc *rx_desc,
2226 						   struct sk_buff **skb,
2227 						   const unsigned int size,
2228 						   int *rx_buffer_pgcnt)
2229 {
2230 	struct ixgbe_rx_buffer *rx_buffer;
2231 
2232 	rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
2233 	*rx_buffer_pgcnt =
2234 #if (PAGE_SIZE < 8192)
2235 		page_count(rx_buffer->page);
2236 #else
2237 		0;
2238 #endif
2239 	prefetchw(rx_buffer->page);
2240 	*skb = rx_buffer->skb;
2241 
2242 	/* Delay unmapping of the first packet. It carries the header
2243 	 * information, HW may still access the header after the writeback.
2244 	 * Only unmap it when EOP is reached
2245 	 */
2246 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)) {
2247 		if (!*skb)
2248 			goto skip_sync;
2249 	} else {
2250 		if (*skb)
2251 			ixgbe_dma_sync_frag(rx_ring, *skb);
2252 	}
2253 
2254 	/* we are reusing so sync this buffer for CPU use */
2255 	dma_sync_single_range_for_cpu(rx_ring->dev,
2256 				      rx_buffer->dma,
2257 				      rx_buffer->page_offset,
2258 				      size,
2259 				      DMA_FROM_DEVICE);
2260 skip_sync:
2261 	rx_buffer->pagecnt_bias--;
2262 
2263 	return rx_buffer;
2264 }
2265 
2266 static void ixgbe_put_rx_buffer(struct ixgbe_ring *rx_ring,
2267 				struct ixgbe_rx_buffer *rx_buffer,
2268 				struct sk_buff *skb,
2269 				int rx_buffer_pgcnt)
2270 {
2271 	if (ixgbe_can_reuse_rx_page(rx_buffer, rx_buffer_pgcnt)) {
2272 		/* hand second half of page back to the ring */
2273 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2274 	} else {
2275 		if (skb && IXGBE_CB(skb)->dma == rx_buffer->dma) {
2276 			/* the page has been released from the ring */
2277 			IXGBE_CB(skb)->page_released = true;
2278 		} else {
2279 			/* we are not reusing the buffer so unmap it */
2280 			dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
2281 					     ixgbe_rx_pg_size(rx_ring),
2282 					     DMA_FROM_DEVICE,
2283 					     IXGBE_RX_DMA_ATTR);
2284 		}
2285 		__page_frag_cache_drain(rx_buffer->page,
2286 					rx_buffer->pagecnt_bias);
2287 	}
2288 
2289 	/* clear contents of rx_buffer */
2290 	rx_buffer->page = NULL;
2291 	rx_buffer->skb = NULL;
2292 }
2293 
2294 static struct sk_buff *ixgbe_construct_skb(struct ixgbe_ring *rx_ring,
2295 					   struct ixgbe_rx_buffer *rx_buffer,
2296 					   struct xdp_buff *xdp,
2297 					   union ixgbe_adv_rx_desc *rx_desc)
2298 {
2299 	unsigned int size = xdp->data_end - xdp->data;
2300 #if (PAGE_SIZE < 8192)
2301 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2302 #else
2303 	unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end -
2304 					       xdp->data_hard_start);
2305 #endif
2306 	struct sk_buff *skb;
2307 
2308 	/* prefetch first cache line of first page */
2309 	net_prefetch(xdp->data);
2310 
2311 	/* Note, we get here by enabling legacy-rx via:
2312 	 *
2313 	 *    ethtool --set-priv-flags <dev> legacy-rx on
2314 	 *
2315 	 * In this mode, we currently get 0 extra XDP headroom as
2316 	 * opposed to having legacy-rx off, where we process XDP
2317 	 * packets going to stack via ixgbe_build_skb(). The latter
2318 	 * provides us currently with 192 bytes of headroom.
2319 	 *
2320 	 * For ixgbe_construct_skb() mode it means that the
2321 	 * xdp->data_meta will always point to xdp->data, since
2322 	 * the helper cannot expand the head. Should this ever
2323 	 * change in future for legacy-rx mode on, then lets also
2324 	 * add xdp->data_meta handling here.
2325 	 */
2326 
2327 	/* allocate a skb to store the frags */
2328 	skb = napi_alloc_skb(&rx_ring->q_vector->napi, IXGBE_RX_HDR_SIZE);
2329 	if (unlikely(!skb))
2330 		return NULL;
2331 
2332 	if (size > IXGBE_RX_HDR_SIZE) {
2333 		if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2334 			IXGBE_CB(skb)->dma = rx_buffer->dma;
2335 
2336 		skb_add_rx_frag(skb, 0, rx_buffer->page,
2337 				xdp->data - page_address(rx_buffer->page),
2338 				size, truesize);
2339 #if (PAGE_SIZE < 8192)
2340 		rx_buffer->page_offset ^= truesize;
2341 #else
2342 		rx_buffer->page_offset += truesize;
2343 #endif
2344 	} else {
2345 		memcpy(__skb_put(skb, size),
2346 		       xdp->data, ALIGN(size, sizeof(long)));
2347 		rx_buffer->pagecnt_bias++;
2348 	}
2349 
2350 	return skb;
2351 }
2352 
2353 static struct sk_buff *ixgbe_build_skb(struct ixgbe_ring *rx_ring,
2354 				       struct ixgbe_rx_buffer *rx_buffer,
2355 				       struct xdp_buff *xdp,
2356 				       union ixgbe_adv_rx_desc *rx_desc)
2357 {
2358 	unsigned int metasize = xdp->data - xdp->data_meta;
2359 #if (PAGE_SIZE < 8192)
2360 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2361 #else
2362 	unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
2363 				SKB_DATA_ALIGN(xdp->data_end -
2364 					       xdp->data_hard_start);
2365 #endif
2366 	struct sk_buff *skb;
2367 
2368 	/* Prefetch first cache line of first page. If xdp->data_meta
2369 	 * is unused, this points exactly as xdp->data, otherwise we
2370 	 * likely have a consumer accessing first few bytes of meta
2371 	 * data, and then actual data.
2372 	 */
2373 	net_prefetch(xdp->data_meta);
2374 
2375 	/* build an skb to around the page buffer */
2376 	skb = napi_build_skb(xdp->data_hard_start, truesize);
2377 	if (unlikely(!skb))
2378 		return NULL;
2379 
2380 	/* update pointers within the skb to store the data */
2381 	skb_reserve(skb, xdp->data - xdp->data_hard_start);
2382 	__skb_put(skb, xdp->data_end - xdp->data);
2383 	if (metasize)
2384 		skb_metadata_set(skb, metasize);
2385 
2386 	/* record DMA address if this is the start of a chain of buffers */
2387 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2388 		IXGBE_CB(skb)->dma = rx_buffer->dma;
2389 
2390 	/* update buffer offset */
2391 #if (PAGE_SIZE < 8192)
2392 	rx_buffer->page_offset ^= truesize;
2393 #else
2394 	rx_buffer->page_offset += truesize;
2395 #endif
2396 
2397 	return skb;
2398 }
2399 
2400 static int ixgbe_run_xdp(struct ixgbe_adapter *adapter,
2401 			 struct ixgbe_ring *rx_ring,
2402 			 struct xdp_buff *xdp)
2403 {
2404 	int err, result = IXGBE_XDP_PASS;
2405 	struct bpf_prog *xdp_prog;
2406 	struct ixgbe_ring *ring;
2407 	struct xdp_frame *xdpf;
2408 	u32 act;
2409 
2410 	xdp_prog = READ_ONCE(rx_ring->xdp_prog);
2411 
2412 	if (!xdp_prog)
2413 		goto xdp_out;
2414 
2415 	prefetchw(xdp->data_hard_start); /* xdp_frame write */
2416 
2417 	act = bpf_prog_run_xdp(xdp_prog, xdp);
2418 	switch (act) {
2419 	case XDP_PASS:
2420 		break;
2421 	case XDP_TX:
2422 		xdpf = xdp_convert_buff_to_frame(xdp);
2423 		if (unlikely(!xdpf))
2424 			goto out_failure;
2425 		ring = ixgbe_determine_xdp_ring(adapter);
2426 		if (static_branch_unlikely(&ixgbe_xdp_locking_key))
2427 			spin_lock(&ring->tx_lock);
2428 		result = ixgbe_xmit_xdp_ring(ring, xdpf);
2429 		if (static_branch_unlikely(&ixgbe_xdp_locking_key))
2430 			spin_unlock(&ring->tx_lock);
2431 		if (result == IXGBE_XDP_CONSUMED)
2432 			goto out_failure;
2433 		break;
2434 	case XDP_REDIRECT:
2435 		err = xdp_do_redirect(adapter->netdev, xdp, xdp_prog);
2436 		if (err)
2437 			goto out_failure;
2438 		result = IXGBE_XDP_REDIR;
2439 		break;
2440 	default:
2441 		bpf_warn_invalid_xdp_action(rx_ring->netdev, xdp_prog, act);
2442 		fallthrough;
2443 	case XDP_ABORTED:
2444 out_failure:
2445 		trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
2446 		fallthrough; /* handle aborts by dropping packet */
2447 	case XDP_DROP:
2448 		result = IXGBE_XDP_CONSUMED;
2449 		break;
2450 	}
2451 xdp_out:
2452 	return result;
2453 }
2454 
2455 static unsigned int ixgbe_rx_frame_truesize(struct ixgbe_ring *rx_ring,
2456 					    unsigned int size)
2457 {
2458 	unsigned int truesize;
2459 
2460 #if (PAGE_SIZE < 8192)
2461 	truesize = ixgbe_rx_pg_size(rx_ring) / 2; /* Must be power-of-2 */
2462 #else
2463 	truesize = rx_ring->rx_offset ?
2464 		SKB_DATA_ALIGN(rx_ring->rx_offset + size) +
2465 		SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) :
2466 		SKB_DATA_ALIGN(size);
2467 #endif
2468 	return truesize;
2469 }
2470 
2471 static void ixgbe_rx_buffer_flip(struct ixgbe_ring *rx_ring,
2472 				 struct ixgbe_rx_buffer *rx_buffer,
2473 				 unsigned int size)
2474 {
2475 	unsigned int truesize = ixgbe_rx_frame_truesize(rx_ring, size);
2476 #if (PAGE_SIZE < 8192)
2477 	rx_buffer->page_offset ^= truesize;
2478 #else
2479 	rx_buffer->page_offset += truesize;
2480 #endif
2481 }
2482 
2483 /**
2484  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2485  * @q_vector: structure containing interrupt and ring information
2486  * @rx_ring: rx descriptor ring to transact packets on
2487  * @budget: Total limit on number of packets to process
2488  *
2489  * This function provides a "bounce buffer" approach to Rx interrupt
2490  * processing.  The advantage to this is that on systems that have
2491  * expensive overhead for IOMMU access this provides a means of avoiding
2492  * it by maintaining the mapping of the page to the system.
2493  *
2494  * Returns amount of work completed
2495  **/
2496 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2497 			       struct ixgbe_ring *rx_ring,
2498 			       const int budget)
2499 {
2500 	unsigned int total_rx_bytes = 0, total_rx_packets = 0, frame_sz = 0;
2501 	struct ixgbe_adapter *adapter = q_vector->adapter;
2502 #ifdef IXGBE_FCOE
2503 	int ddp_bytes;
2504 	unsigned int mss = 0;
2505 #endif /* IXGBE_FCOE */
2506 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2507 	unsigned int offset = rx_ring->rx_offset;
2508 	unsigned int xdp_xmit = 0;
2509 	struct xdp_buff xdp;
2510 	int xdp_res = 0;
2511 
2512 	/* Frame size depend on rx_ring setup when PAGE_SIZE=4K */
2513 #if (PAGE_SIZE < 8192)
2514 	frame_sz = ixgbe_rx_frame_truesize(rx_ring, 0);
2515 #endif
2516 	xdp_init_buff(&xdp, frame_sz, &rx_ring->xdp_rxq);
2517 
2518 	while (likely(total_rx_packets < budget)) {
2519 		union ixgbe_adv_rx_desc *rx_desc;
2520 		struct ixgbe_rx_buffer *rx_buffer;
2521 		struct sk_buff *skb;
2522 		int rx_buffer_pgcnt;
2523 		unsigned int size;
2524 
2525 		/* return some buffers to hardware, one at a time is too slow */
2526 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2527 			ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2528 			cleaned_count = 0;
2529 		}
2530 
2531 		rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2532 		size = le16_to_cpu(rx_desc->wb.upper.length);
2533 		if (!size)
2534 			break;
2535 
2536 		/* This memory barrier is needed to keep us from reading
2537 		 * any other fields out of the rx_desc until we know the
2538 		 * descriptor has been written back
2539 		 */
2540 		dma_rmb();
2541 
2542 		rx_buffer = ixgbe_get_rx_buffer(rx_ring, rx_desc, &skb, size, &rx_buffer_pgcnt);
2543 
2544 		/* retrieve a buffer from the ring */
2545 		if (!skb) {
2546 			unsigned char *hard_start;
2547 
2548 			hard_start = page_address(rx_buffer->page) +
2549 				     rx_buffer->page_offset - offset;
2550 			xdp_prepare_buff(&xdp, hard_start, offset, size, true);
2551 			xdp_buff_clear_frags_flag(&xdp);
2552 #if (PAGE_SIZE > 4096)
2553 			/* At larger PAGE_SIZE, frame_sz depend on len size */
2554 			xdp.frame_sz = ixgbe_rx_frame_truesize(rx_ring, size);
2555 #endif
2556 			xdp_res = ixgbe_run_xdp(adapter, rx_ring, &xdp);
2557 		}
2558 
2559 		if (xdp_res) {
2560 			if (xdp_res & (IXGBE_XDP_TX | IXGBE_XDP_REDIR)) {
2561 				xdp_xmit |= xdp_res;
2562 				ixgbe_rx_buffer_flip(rx_ring, rx_buffer, size);
2563 			} else {
2564 				rx_buffer->pagecnt_bias++;
2565 			}
2566 			total_rx_packets++;
2567 			total_rx_bytes += size;
2568 		} else if (skb) {
2569 			ixgbe_add_rx_frag(rx_ring, rx_buffer, skb, size);
2570 		} else if (ring_uses_build_skb(rx_ring)) {
2571 			skb = ixgbe_build_skb(rx_ring, rx_buffer,
2572 					      &xdp, rx_desc);
2573 		} else {
2574 			skb = ixgbe_construct_skb(rx_ring, rx_buffer,
2575 						  &xdp, rx_desc);
2576 		}
2577 
2578 		/* exit if we failed to retrieve a buffer */
2579 		if (!xdp_res && !skb) {
2580 			rx_ring->rx_stats.alloc_rx_buff_failed++;
2581 			rx_buffer->pagecnt_bias++;
2582 			break;
2583 		}
2584 
2585 		ixgbe_put_rx_buffer(rx_ring, rx_buffer, skb, rx_buffer_pgcnt);
2586 		cleaned_count++;
2587 
2588 		/* place incomplete frames back on ring for completion */
2589 		if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2590 			continue;
2591 
2592 		/* verify the packet layout is correct */
2593 		if (xdp_res || ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2594 			continue;
2595 
2596 		/* probably a little skewed due to removing CRC */
2597 		total_rx_bytes += skb->len;
2598 
2599 		/* populate checksum, timestamp, VLAN, and protocol */
2600 		ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2601 
2602 #ifdef IXGBE_FCOE
2603 		/* if ddp, not passing to ULD unless for FCP_RSP or error */
2604 		if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2605 			ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2606 			/* include DDPed FCoE data */
2607 			if (ddp_bytes > 0) {
2608 				if (!mss) {
2609 					mss = rx_ring->netdev->mtu -
2610 						sizeof(struct fcoe_hdr) -
2611 						sizeof(struct fc_frame_header) -
2612 						sizeof(struct fcoe_crc_eof);
2613 					if (mss > 512)
2614 						mss &= ~511;
2615 				}
2616 				total_rx_bytes += ddp_bytes;
2617 				total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2618 								 mss);
2619 			}
2620 			if (!ddp_bytes) {
2621 				dev_kfree_skb_any(skb);
2622 				continue;
2623 			}
2624 		}
2625 
2626 #endif /* IXGBE_FCOE */
2627 		ixgbe_rx_skb(q_vector, skb);
2628 
2629 		/* update budget accounting */
2630 		total_rx_packets++;
2631 	}
2632 
2633 	if (xdp_xmit & IXGBE_XDP_REDIR)
2634 		xdp_do_flush();
2635 
2636 	if (xdp_xmit & IXGBE_XDP_TX) {
2637 		struct ixgbe_ring *ring = ixgbe_determine_xdp_ring(adapter);
2638 
2639 		ixgbe_xdp_ring_update_tail_locked(ring);
2640 	}
2641 
2642 	ixgbe_update_rx_ring_stats(rx_ring, q_vector, total_rx_packets,
2643 				   total_rx_bytes);
2644 
2645 	return total_rx_packets;
2646 }
2647 
2648 /**
2649  * ixgbe_configure_msix - Configure MSI-X hardware
2650  * @adapter: board private structure
2651  *
2652  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2653  * interrupts.
2654  **/
2655 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2656 {
2657 	struct ixgbe_q_vector *q_vector;
2658 	int v_idx;
2659 	u32 mask;
2660 
2661 	/* Populate MSIX to EITR Select */
2662 	if (adapter->num_vfs > 32) {
2663 		u32 eitrsel = BIT(adapter->num_vfs - 32) - 1;
2664 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2665 	}
2666 
2667 	/*
2668 	 * Populate the IVAR table and set the ITR values to the
2669 	 * corresponding register.
2670 	 */
2671 	for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2672 		struct ixgbe_ring *ring;
2673 		q_vector = adapter->q_vector[v_idx];
2674 
2675 		ixgbe_for_each_ring(ring, q_vector->rx)
2676 			ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2677 
2678 		ixgbe_for_each_ring(ring, q_vector->tx)
2679 			ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2680 
2681 		ixgbe_write_eitr(q_vector);
2682 	}
2683 
2684 	switch (adapter->hw.mac.type) {
2685 	case ixgbe_mac_82598EB:
2686 		ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2687 			       v_idx);
2688 		break;
2689 	case ixgbe_mac_82599EB:
2690 	case ixgbe_mac_X540:
2691 	case ixgbe_mac_X550:
2692 	case ixgbe_mac_X550EM_x:
2693 	case ixgbe_mac_x550em_a:
2694 	case ixgbe_mac_e610:
2695 		ixgbe_set_ivar(adapter, -1, 1, v_idx);
2696 		break;
2697 	default:
2698 		break;
2699 	}
2700 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2701 
2702 	/* set up to autoclear timer, and the vectors */
2703 	mask = IXGBE_EIMS_ENABLE_MASK;
2704 	mask &= ~(IXGBE_EIMS_OTHER |
2705 		  IXGBE_EIMS_MAILBOX |
2706 		  IXGBE_EIMS_LSC);
2707 
2708 	if (adapter->hw.mac.type == ixgbe_mac_e610)
2709 		mask &= ~IXGBE_EIMS_FW_EVENT;
2710 
2711 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2712 }
2713 
2714 /**
2715  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2716  * @q_vector: structure containing interrupt and ring information
2717  * @ring_container: structure containing ring performance data
2718  *
2719  *      Stores a new ITR value based on packets and byte
2720  *      counts during the last interrupt.  The advantage of per interrupt
2721  *      computation is faster updates and more accurate ITR for the current
2722  *      traffic pattern.  Constants in this function were computed
2723  *      based on theoretical maximum wire speed and thresholds were set based
2724  *      on testing data as well as attempting to minimize response time
2725  *      while increasing bulk throughput.
2726  **/
2727 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2728 			     struct ixgbe_ring_container *ring_container)
2729 {
2730 	unsigned int itr = IXGBE_ITR_ADAPTIVE_MIN_USECS |
2731 			   IXGBE_ITR_ADAPTIVE_LATENCY;
2732 	unsigned int avg_wire_size, packets, bytes;
2733 	unsigned long next_update = jiffies;
2734 
2735 	/* If we don't have any rings just leave ourselves set for maximum
2736 	 * possible latency so we take ourselves out of the equation.
2737 	 */
2738 	if (!ring_container->ring)
2739 		return;
2740 
2741 	/* If we didn't update within up to 1 - 2 jiffies we can assume
2742 	 * that either packets are coming in so slow there hasn't been
2743 	 * any work, or that there is so much work that NAPI is dealing
2744 	 * with interrupt moderation and we don't need to do anything.
2745 	 */
2746 	if (time_after(next_update, ring_container->next_update))
2747 		goto clear_counts;
2748 
2749 	packets = ring_container->total_packets;
2750 
2751 	/* We have no packets to actually measure against. This means
2752 	 * either one of the other queues on this vector is active or
2753 	 * we are a Tx queue doing TSO with too high of an interrupt rate.
2754 	 *
2755 	 * When this occurs just tick up our delay by the minimum value
2756 	 * and hope that this extra delay will prevent us from being called
2757 	 * without any work on our queue.
2758 	 */
2759 	if (!packets) {
2760 		itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2761 		if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2762 			itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2763 		itr += ring_container->itr & IXGBE_ITR_ADAPTIVE_LATENCY;
2764 		goto clear_counts;
2765 	}
2766 
2767 	bytes = ring_container->total_bytes;
2768 
2769 	/* If packets are less than 4 or bytes are less than 9000 assume
2770 	 * insufficient data to use bulk rate limiting approach. We are
2771 	 * likely latency driven.
2772 	 */
2773 	if (packets < 4 && bytes < 9000) {
2774 		itr = IXGBE_ITR_ADAPTIVE_LATENCY;
2775 		goto adjust_by_size;
2776 	}
2777 
2778 	/* Between 4 and 48 we can assume that our current interrupt delay
2779 	 * is only slightly too low. As such we should increase it by a small
2780 	 * fixed amount.
2781 	 */
2782 	if (packets < 48) {
2783 		itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2784 		if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2785 			itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2786 		goto clear_counts;
2787 	}
2788 
2789 	/* Between 48 and 96 is our "goldilocks" zone where we are working
2790 	 * out "just right". Just report that our current ITR is good for us.
2791 	 */
2792 	if (packets < 96) {
2793 		itr = q_vector->itr >> 2;
2794 		goto clear_counts;
2795 	}
2796 
2797 	/* If packet count is 96 or greater we are likely looking at a slight
2798 	 * overrun of the delay we want. Try halving our delay to see if that
2799 	 * will cut the number of packets in half per interrupt.
2800 	 */
2801 	if (packets < 256) {
2802 		itr = q_vector->itr >> 3;
2803 		if (itr < IXGBE_ITR_ADAPTIVE_MIN_USECS)
2804 			itr = IXGBE_ITR_ADAPTIVE_MIN_USECS;
2805 		goto clear_counts;
2806 	}
2807 
2808 	/* The paths below assume we are dealing with a bulk ITR since number
2809 	 * of packets is 256 or greater. We are just going to have to compute
2810 	 * a value and try to bring the count under control, though for smaller
2811 	 * packet sizes there isn't much we can do as NAPI polling will likely
2812 	 * be kicking in sooner rather than later.
2813 	 */
2814 	itr = IXGBE_ITR_ADAPTIVE_BULK;
2815 
2816 adjust_by_size:
2817 	/* If packet counts are 256 or greater we can assume we have a gross
2818 	 * overestimation of what the rate should be. Instead of trying to fine
2819 	 * tune it just use the formula below to try and dial in an exact value
2820 	 * give the current packet size of the frame.
2821 	 */
2822 	avg_wire_size = bytes / packets;
2823 
2824 	/* The following is a crude approximation of:
2825 	 *  wmem_default / (size + overhead) = desired_pkts_per_int
2826 	 *  rate / bits_per_byte / (size + ethernet overhead) = pkt_rate
2827 	 *  (desired_pkt_rate / pkt_rate) * usecs_per_sec = ITR value
2828 	 *
2829 	 * Assuming wmem_default is 212992 and overhead is 640 bytes per
2830 	 * packet, (256 skb, 64 headroom, 320 shared info), we can reduce the
2831 	 * formula down to
2832 	 *
2833 	 *  (170 * (size + 24)) / (size + 640) = ITR
2834 	 *
2835 	 * We first do some math on the packet size and then finally bitshift
2836 	 * by 8 after rounding up. We also have to account for PCIe link speed
2837 	 * difference as ITR scales based on this.
2838 	 */
2839 	if (avg_wire_size <= 60) {
2840 		/* Start at 50k ints/sec */
2841 		avg_wire_size = 5120;
2842 	} else if (avg_wire_size <= 316) {
2843 		/* 50K ints/sec to 16K ints/sec */
2844 		avg_wire_size *= 40;
2845 		avg_wire_size += 2720;
2846 	} else if (avg_wire_size <= 1084) {
2847 		/* 16K ints/sec to 9.2K ints/sec */
2848 		avg_wire_size *= 15;
2849 		avg_wire_size += 11452;
2850 	} else if (avg_wire_size < 1968) {
2851 		/* 9.2K ints/sec to 8K ints/sec */
2852 		avg_wire_size *= 5;
2853 		avg_wire_size += 22420;
2854 	} else {
2855 		/* plateau at a limit of 8K ints/sec */
2856 		avg_wire_size = 32256;
2857 	}
2858 
2859 	/* If we are in low latency mode half our delay which doubles the rate
2860 	 * to somewhere between 100K to 16K ints/sec
2861 	 */
2862 	if (itr & IXGBE_ITR_ADAPTIVE_LATENCY)
2863 		avg_wire_size >>= 1;
2864 
2865 	/* Resultant value is 256 times larger than it needs to be. This
2866 	 * gives us room to adjust the value as needed to either increase
2867 	 * or decrease the value based on link speeds of 10G, 2.5G, 1G, etc.
2868 	 *
2869 	 * Use addition as we have already recorded the new latency flag
2870 	 * for the ITR value.
2871 	 */
2872 	switch (q_vector->adapter->link_speed) {
2873 	case IXGBE_LINK_SPEED_10GB_FULL:
2874 	case IXGBE_LINK_SPEED_100_FULL:
2875 	default:
2876 		itr += DIV_ROUND_UP(avg_wire_size,
2877 				    IXGBE_ITR_ADAPTIVE_MIN_INC * 256) *
2878 		       IXGBE_ITR_ADAPTIVE_MIN_INC;
2879 		break;
2880 	case IXGBE_LINK_SPEED_2_5GB_FULL:
2881 	case IXGBE_LINK_SPEED_1GB_FULL:
2882 	case IXGBE_LINK_SPEED_10_FULL:
2883 		if (avg_wire_size > 8064)
2884 			avg_wire_size = 8064;
2885 		itr += DIV_ROUND_UP(avg_wire_size,
2886 				    IXGBE_ITR_ADAPTIVE_MIN_INC * 64) *
2887 		       IXGBE_ITR_ADAPTIVE_MIN_INC;
2888 		break;
2889 	}
2890 
2891 clear_counts:
2892 	/* write back value */
2893 	ring_container->itr = itr;
2894 
2895 	/* next update should occur within next jiffy */
2896 	ring_container->next_update = next_update + 1;
2897 
2898 	ring_container->total_bytes = 0;
2899 	ring_container->total_packets = 0;
2900 }
2901 
2902 /**
2903  * ixgbe_write_eitr - write EITR register in hardware specific way
2904  * @q_vector: structure containing interrupt and ring information
2905  *
2906  * This function is made to be called by ethtool and by the driver
2907  * when it needs to update EITR registers at runtime.  Hardware
2908  * specific quirks/differences are taken care of here.
2909  */
2910 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2911 {
2912 	struct ixgbe_adapter *adapter = q_vector->adapter;
2913 	struct ixgbe_hw *hw = &adapter->hw;
2914 	int v_idx = q_vector->v_idx;
2915 	u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2916 
2917 	switch (adapter->hw.mac.type) {
2918 	case ixgbe_mac_82598EB:
2919 		/* must write high and low 16 bits to reset counter */
2920 		itr_reg |= (itr_reg << 16);
2921 		break;
2922 	case ixgbe_mac_82599EB:
2923 	case ixgbe_mac_X540:
2924 	case ixgbe_mac_X550:
2925 	case ixgbe_mac_X550EM_x:
2926 	case ixgbe_mac_x550em_a:
2927 	case ixgbe_mac_e610:
2928 		/*
2929 		 * set the WDIS bit to not clear the timer bits and cause an
2930 		 * immediate assertion of the interrupt
2931 		 */
2932 		itr_reg |= IXGBE_EITR_CNT_WDIS;
2933 		break;
2934 	default:
2935 		break;
2936 	}
2937 	IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2938 }
2939 
2940 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2941 {
2942 	u32 new_itr;
2943 
2944 	ixgbe_update_itr(q_vector, &q_vector->tx);
2945 	ixgbe_update_itr(q_vector, &q_vector->rx);
2946 
2947 	/* use the smallest value of new ITR delay calculations */
2948 	new_itr = min(q_vector->rx.itr, q_vector->tx.itr);
2949 
2950 	/* Clear latency flag if set, shift into correct position */
2951 	new_itr &= ~IXGBE_ITR_ADAPTIVE_LATENCY;
2952 	new_itr <<= 2;
2953 
2954 	if (new_itr != q_vector->itr) {
2955 		/* save the algorithm value here */
2956 		q_vector->itr = new_itr;
2957 
2958 		ixgbe_write_eitr(q_vector);
2959 	}
2960 }
2961 
2962 /**
2963  * ixgbe_check_overtemp_subtask - check for over temperature
2964  * @adapter: pointer to adapter
2965  **/
2966 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2967 {
2968 	struct ixgbe_hw *hw = &adapter->hw;
2969 	u32 eicr = adapter->interrupt_event;
2970 
2971 	if (test_bit(__IXGBE_DOWN, &adapter->state))
2972 		return;
2973 
2974 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2975 		return;
2976 
2977 	adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2978 
2979 	switch (hw->device_id) {
2980 	case IXGBE_DEV_ID_82599_T3_LOM:
2981 		/*
2982 		 * Since the warning interrupt is for both ports
2983 		 * we don't have to check if:
2984 		 *  - This interrupt wasn't for our port.
2985 		 *  - We may have missed the interrupt so always have to
2986 		 *    check if we  got a LSC
2987 		 */
2988 		if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2989 		    !(eicr & IXGBE_EICR_LSC))
2990 			return;
2991 
2992 		if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2993 			u32 speed;
2994 			bool link_up = false;
2995 
2996 			hw->mac.ops.check_link(hw, &speed, &link_up, false);
2997 
2998 			if (link_up)
2999 				return;
3000 		}
3001 
3002 		/* Check if this is not due to overtemp */
3003 		if (!hw->phy.ops.check_overtemp(hw))
3004 			return;
3005 
3006 		break;
3007 	case IXGBE_DEV_ID_X550EM_A_1G_T:
3008 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
3009 		if (!hw->phy.ops.check_overtemp(hw))
3010 			return;
3011 		break;
3012 	default:
3013 		if (adapter->hw.mac.type >= ixgbe_mac_X540)
3014 			return;
3015 		if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
3016 			return;
3017 		break;
3018 	}
3019 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
3020 
3021 	adapter->interrupt_event = 0;
3022 }
3023 
3024 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
3025 {
3026 	struct ixgbe_hw *hw = &adapter->hw;
3027 
3028 	if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
3029 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
3030 		e_crit(probe, "Fan has stopped, replace the adapter\n");
3031 		/* write to clear the interrupt */
3032 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
3033 	}
3034 }
3035 
3036 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
3037 {
3038 	struct ixgbe_hw *hw = &adapter->hw;
3039 
3040 	if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
3041 		return;
3042 
3043 	switch (adapter->hw.mac.type) {
3044 	case ixgbe_mac_82599EB:
3045 		/*
3046 		 * Need to check link state so complete overtemp check
3047 		 * on service task
3048 		 */
3049 		if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
3050 		     (eicr & IXGBE_EICR_LSC)) &&
3051 		    (!test_bit(__IXGBE_DOWN, &adapter->state))) {
3052 			adapter->interrupt_event = eicr;
3053 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
3054 			ixgbe_service_event_schedule(adapter);
3055 			return;
3056 		}
3057 		return;
3058 	case ixgbe_mac_x550em_a:
3059 		if (eicr & IXGBE_EICR_GPI_SDP0_X550EM_a) {
3060 			adapter->interrupt_event = eicr;
3061 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
3062 			ixgbe_service_event_schedule(adapter);
3063 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC,
3064 					IXGBE_EICR_GPI_SDP0_X550EM_a);
3065 			IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICR,
3066 					IXGBE_EICR_GPI_SDP0_X550EM_a);
3067 		}
3068 		return;
3069 	case ixgbe_mac_X550:
3070 	case ixgbe_mac_X540:
3071 		if (!(eicr & IXGBE_EICR_TS))
3072 			return;
3073 		break;
3074 	default:
3075 		return;
3076 	}
3077 
3078 	e_crit(drv, "%s\n", ixgbe_overheat_msg);
3079 }
3080 
3081 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
3082 {
3083 	switch (hw->mac.type) {
3084 	case ixgbe_mac_82598EB:
3085 		if (hw->phy.type == ixgbe_phy_nl)
3086 			return true;
3087 		return false;
3088 	case ixgbe_mac_82599EB:
3089 	case ixgbe_mac_X550EM_x:
3090 	case ixgbe_mac_x550em_a:
3091 		switch (hw->mac.ops.get_media_type(hw)) {
3092 		case ixgbe_media_type_fiber:
3093 		case ixgbe_media_type_fiber_qsfp:
3094 			return true;
3095 		default:
3096 			return false;
3097 		}
3098 	default:
3099 		return false;
3100 	}
3101 }
3102 
3103 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
3104 {
3105 	struct ixgbe_hw *hw = &adapter->hw;
3106 	u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
3107 
3108 	if (!ixgbe_is_sfp(hw))
3109 		return;
3110 
3111 	/* Later MAC's use different SDP */
3112 	if (hw->mac.type >= ixgbe_mac_X540)
3113 		eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
3114 
3115 	if (eicr & eicr_mask) {
3116 		/* Clear the interrupt */
3117 		IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
3118 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
3119 			adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
3120 			adapter->sfp_poll_time = 0;
3121 			ixgbe_service_event_schedule(adapter);
3122 		}
3123 	}
3124 
3125 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
3126 	    (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
3127 		/* Clear the interrupt */
3128 		IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
3129 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
3130 			adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
3131 			ixgbe_service_event_schedule(adapter);
3132 		}
3133 	}
3134 }
3135 
3136 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
3137 {
3138 	struct ixgbe_hw *hw = &adapter->hw;
3139 
3140 	adapter->lsc_int++;
3141 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
3142 	adapter->link_check_timeout = jiffies;
3143 	if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
3144 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
3145 		IXGBE_WRITE_FLUSH(hw);
3146 		ixgbe_service_event_schedule(adapter);
3147 	}
3148 }
3149 
3150 /**
3151  * ixgbe_check_phy_fw_load - check if PHY FW load failed
3152  * @adapter: pointer to adapter structure
3153  * @link_cfg_err: bitmap from the link info structure
3154  *
3155  * Check if external PHY FW load failed and print an error message if it did.
3156  */
3157 static void ixgbe_check_phy_fw_load(struct ixgbe_adapter *adapter,
3158 				    u8 link_cfg_err)
3159 {
3160 	if (!(link_cfg_err & IXGBE_ACI_LINK_EXTERNAL_PHY_LOAD_FAILURE)) {
3161 		adapter->flags2 &= ~IXGBE_FLAG2_PHY_FW_LOAD_FAILED;
3162 		return;
3163 	}
3164 
3165 	if (adapter->flags2 & IXGBE_FLAG2_PHY_FW_LOAD_FAILED)
3166 		return;
3167 
3168 	if (link_cfg_err & IXGBE_ACI_LINK_EXTERNAL_PHY_LOAD_FAILURE) {
3169 		netdev_err(adapter->netdev, "Device failed to load the FW for the external PHY. Please download and install the latest NVM for your device and try again\n");
3170 		adapter->flags2 |= IXGBE_FLAG2_PHY_FW_LOAD_FAILED;
3171 	}
3172 }
3173 
3174 /**
3175  * ixgbe_check_module_power - check module power level
3176  * @adapter: pointer to adapter structure
3177  * @link_cfg_err: bitmap from the link info structure
3178  *
3179  * Check module power level returned by a previous call to aci_get_link_info
3180  * and print error messages if module power level is not supported.
3181  */
3182 static void ixgbe_check_module_power(struct ixgbe_adapter *adapter,
3183 				     u8 link_cfg_err)
3184 {
3185 	/* If module power level is supported, clear the flag. */
3186 	if (!(link_cfg_err & (IXGBE_ACI_LINK_INVAL_MAX_POWER_LIMIT |
3187 			      IXGBE_ACI_LINK_MODULE_POWER_UNSUPPORTED))) {
3188 		adapter->flags2 &= ~IXGBE_FLAG2_MOD_POWER_UNSUPPORTED;
3189 		return;
3190 	}
3191 
3192 	/* If IXGBE_FLAG2_MOD_POWER_UNSUPPORTED was previously set and the
3193 	 * above block didn't clear this bit, there's nothing to do.
3194 	 */
3195 	if (adapter->flags2 & IXGBE_FLAG2_MOD_POWER_UNSUPPORTED)
3196 		return;
3197 
3198 	if (link_cfg_err & IXGBE_ACI_LINK_INVAL_MAX_POWER_LIMIT) {
3199 		netdev_err(adapter->netdev, "The installed module is incompatible with the device's NVM image. Cannot start link.\n");
3200 		adapter->flags2 |= IXGBE_FLAG2_MOD_POWER_UNSUPPORTED;
3201 	} else if (link_cfg_err & IXGBE_ACI_LINK_MODULE_POWER_UNSUPPORTED) {
3202 		netdev_err(adapter->netdev, "The module's power requirements exceed the device's power supply. Cannot start link.\n");
3203 		adapter->flags2 |= IXGBE_FLAG2_MOD_POWER_UNSUPPORTED;
3204 	}
3205 }
3206 
3207 /**
3208  * ixgbe_check_link_cfg_err - check if link configuration failed
3209  * @adapter: pointer to adapter structure
3210  * @link_cfg_err: bitmap from the link info structure
3211  *
3212  * Print if any link configuration failure happens due to the value in the
3213  * link_cfg_err parameter in the link info structure.
3214  */
3215 static void ixgbe_check_link_cfg_err(struct ixgbe_adapter *adapter,
3216 				     u8 link_cfg_err)
3217 {
3218 	ixgbe_check_module_power(adapter, link_cfg_err);
3219 	ixgbe_check_phy_fw_load(adapter, link_cfg_err);
3220 }
3221 
3222 /**
3223  * ixgbe_process_link_status_event - process the link event
3224  * @adapter: pointer to adapter structure
3225  * @link_up: true if the physical link is up and false if it is down
3226  * @link_speed: current link speed received from the link event
3227  *
3228  * Return: 0 on success or negative value on failure.
3229  */
3230 static int
3231 ixgbe_process_link_status_event(struct ixgbe_adapter *adapter, bool link_up,
3232 				u16 link_speed)
3233 {
3234 	struct ixgbe_hw *hw = &adapter->hw;
3235 	int status;
3236 
3237 	/* Update the link info structures and re-enable link events,
3238 	 * don't bail on failure due to other book keeping needed.
3239 	 */
3240 	status = ixgbe_update_link_info(hw);
3241 	if (status)
3242 		e_dev_err("Failed to update link status, err %d aq_err %d\n",
3243 			  status, hw->aci.last_status);
3244 
3245 	ixgbe_check_link_cfg_err(adapter, hw->link.link_info.link_cfg_err);
3246 
3247 	/* Check if the link state is up after updating link info, and treat
3248 	 * this event as an UP event since the link is actually UP now.
3249 	 */
3250 	if (hw->link.link_info.link_info & IXGBE_ACI_LINK_UP)
3251 		link_up = true;
3252 
3253 	/* Turn off PHY if media was removed. */
3254 	if (!(adapter->flags2 & IXGBE_FLAG2_NO_MEDIA) &&
3255 	    !(hw->link.link_info.link_info & IXGBE_ACI_MEDIA_AVAILABLE))
3256 		adapter->flags2 |= IXGBE_FLAG2_NO_MEDIA;
3257 
3258 	if (link_up == adapter->link_up &&
3259 	    link_up == netif_carrier_ok(adapter->netdev) &&
3260 	    link_speed == adapter->link_speed)
3261 		return 0;
3262 
3263 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
3264 	adapter->link_check_timeout = jiffies;
3265 	ixgbe_watchdog_update_link(adapter);
3266 
3267 	if (link_up)
3268 		ixgbe_watchdog_link_is_up(adapter);
3269 	else
3270 		ixgbe_watchdog_link_is_down(adapter);
3271 
3272 	return 0;
3273 }
3274 
3275 /**
3276  * ixgbe_handle_link_status_event - handle link status event via ACI
3277  * @adapter: pointer to adapter structure
3278  * @e: event structure containing link status info
3279  */
3280 static void
3281 ixgbe_handle_link_status_event(struct ixgbe_adapter *adapter,
3282 			       struct ixgbe_aci_event *e)
3283 {
3284 	struct ixgbe_aci_cmd_get_link_status_data *link_data;
3285 	u16 link_speed;
3286 	bool link_up;
3287 
3288 	link_data = (struct ixgbe_aci_cmd_get_link_status_data *)e->msg_buf;
3289 
3290 	link_up = !!(link_data->link_info & IXGBE_ACI_LINK_UP);
3291 	link_speed = le16_to_cpu(link_data->link_speed);
3292 
3293 	if (ixgbe_process_link_status_event(adapter, link_up, link_speed))
3294 		e_dev_warn("Could not process link status event");
3295 }
3296 
3297 /**
3298  * ixgbe_schedule_fw_event - schedule Firmware event
3299  * @adapter: pointer to the adapter structure
3300  *
3301  * If the adapter is not in down, removing or resetting state,
3302  * an event is scheduled.
3303  */
3304 static void ixgbe_schedule_fw_event(struct ixgbe_adapter *adapter)
3305 {
3306 	if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
3307 	    !test_bit(__IXGBE_REMOVING, &adapter->state) &&
3308 	    !test_bit(__IXGBE_RESETTING, &adapter->state)) {
3309 		adapter->flags2 |= IXGBE_FLAG2_FW_ASYNC_EVENT;
3310 		ixgbe_service_event_schedule(adapter);
3311 	}
3312 }
3313 
3314 /**
3315  * ixgbe_aci_event_cleanup - release msg_buf memory
3316  * @event: pointer to the event holding msg_buf to be released
3317  *
3318  * Clean memory allocated for event's msg_buf. Implements auto memory cleanup.
3319  */
3320 static void ixgbe_aci_event_cleanup(struct ixgbe_aci_event *event)
3321 {
3322 	kfree(event->msg_buf);
3323 }
3324 
3325 /**
3326  * ixgbe_handle_fw_event - handle Firmware event
3327  * @adapter: pointer to the adapter structure
3328  *
3329  * Obtain an event from the ACI and then and then process it according to the
3330  * type of the event and the opcode.
3331  */
3332 static void ixgbe_handle_fw_event(struct ixgbe_adapter *adapter)
3333 {
3334 	struct ixgbe_aci_event event __cleanup(ixgbe_aci_event_cleanup);
3335 	struct ixgbe_hw *hw = &adapter->hw;
3336 	bool pending = false;
3337 	int err;
3338 
3339 	if (adapter->flags2 & IXGBE_FLAG2_FW_ASYNC_EVENT)
3340 		adapter->flags2 &= ~IXGBE_FLAG2_FW_ASYNC_EVENT;
3341 	event.buf_len = IXGBE_ACI_MAX_BUFFER_SIZE;
3342 	event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
3343 	if (!event.msg_buf)
3344 		return;
3345 
3346 	do {
3347 		err = ixgbe_aci_get_event(hw, &event, &pending);
3348 		if (err)
3349 			break;
3350 
3351 		switch (le16_to_cpu(event.desc.opcode)) {
3352 		case ixgbe_aci_opc_get_link_status:
3353 			ixgbe_handle_link_status_event(adapter, &event);
3354 			break;
3355 		case ixgbe_aci_opc_temp_tca_event:
3356 			e_crit(drv, "%s\n", ixgbe_overheat_msg);
3357 			ixgbe_down(adapter);
3358 			break;
3359 		case libie_aqc_opc_fw_logs_event:
3360 			libie_get_fwlog_data(&hw->fwlog, event.msg_buf,
3361 					     le16_to_cpu(event.desc.datalen));
3362 			break;
3363 		default:
3364 			e_warn(hw, "unknown FW async event captured\n");
3365 			break;
3366 		}
3367 	} while (pending);
3368 }
3369 
3370 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
3371 					   u64 qmask)
3372 {
3373 	struct ixgbe_hw *hw = &adapter->hw;
3374 	u32 mask;
3375 
3376 	switch (hw->mac.type) {
3377 	case ixgbe_mac_82598EB:
3378 		mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
3379 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
3380 		break;
3381 	case ixgbe_mac_82599EB:
3382 	case ixgbe_mac_X540:
3383 	case ixgbe_mac_X550:
3384 	case ixgbe_mac_X550EM_x:
3385 	case ixgbe_mac_x550em_a:
3386 	case ixgbe_mac_e610:
3387 		mask = (qmask & 0xFFFFFFFF);
3388 		if (mask)
3389 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
3390 		mask = (qmask >> 32);
3391 		if (mask)
3392 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
3393 		break;
3394 	default:
3395 		break;
3396 	}
3397 	/* skip the flush */
3398 }
3399 
3400 /**
3401  * ixgbe_irq_enable - Enable default interrupt generation settings
3402  * @adapter: board private structure
3403  * @queues: enable irqs for queues
3404  * @flush: flush register write
3405  **/
3406 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
3407 				    bool flush)
3408 {
3409 	struct ixgbe_hw *hw = &adapter->hw;
3410 	u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
3411 
3412 	/* don't reenable LSC while waiting for link */
3413 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
3414 		mask &= ~IXGBE_EIMS_LSC;
3415 
3416 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
3417 		switch (adapter->hw.mac.type) {
3418 		case ixgbe_mac_82599EB:
3419 			mask |= IXGBE_EIMS_GPI_SDP0(hw);
3420 			break;
3421 		case ixgbe_mac_X540:
3422 		case ixgbe_mac_X550:
3423 		case ixgbe_mac_X550EM_x:
3424 		case ixgbe_mac_x550em_a:
3425 			mask |= IXGBE_EIMS_TS;
3426 			break;
3427 		default:
3428 			break;
3429 		}
3430 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
3431 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
3432 	switch (adapter->hw.mac.type) {
3433 	case ixgbe_mac_82599EB:
3434 		mask |= IXGBE_EIMS_GPI_SDP1(hw);
3435 		mask |= IXGBE_EIMS_GPI_SDP2(hw);
3436 		fallthrough;
3437 	case ixgbe_mac_X540:
3438 	case ixgbe_mac_X550:
3439 	case ixgbe_mac_X550EM_x:
3440 	case ixgbe_mac_e610:
3441 		mask |= IXGBE_EIMS_FW_EVENT;
3442 		fallthrough;
3443 	case ixgbe_mac_x550em_a:
3444 		if (adapter->hw.device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
3445 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
3446 		    adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP_N)
3447 			mask |= IXGBE_EIMS_GPI_SDP0(&adapter->hw);
3448 		if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
3449 			mask |= IXGBE_EICR_GPI_SDP0_X540;
3450 		mask |= IXGBE_EIMS_ECC;
3451 		mask |= IXGBE_EIMS_MAILBOX;
3452 		break;
3453 	default:
3454 		break;
3455 	}
3456 
3457 	if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
3458 	    !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
3459 		mask |= IXGBE_EIMS_FLOW_DIR;
3460 
3461 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
3462 	if (queues)
3463 		ixgbe_irq_enable_queues(adapter, ~0);
3464 	if (flush)
3465 		IXGBE_WRITE_FLUSH(&adapter->hw);
3466 }
3467 
3468 static irqreturn_t ixgbe_msix_other(int irq, void *data)
3469 {
3470 	struct ixgbe_adapter *adapter = data;
3471 	struct ixgbe_hw *hw = &adapter->hw;
3472 	u32 eicr;
3473 
3474 	/*
3475 	 * Workaround for Silicon errata.  Use clear-by-write instead
3476 	 * of clear-by-read.  Reading with EICS will return the
3477 	 * interrupt causes without clearing, which later be done
3478 	 * with the write to EICR.
3479 	 */
3480 	eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
3481 
3482 	/* The lower 16bits of the EICR register are for the queue interrupts
3483 	 * which should be masked here in order to not accidentally clear them if
3484 	 * the bits are high when ixgbe_msix_other is called. There is a race
3485 	 * condition otherwise which results in possible performance loss
3486 	 * especially if the ixgbe_msix_other interrupt is triggering
3487 	 * consistently (as it would when PPS is turned on for the X540 device)
3488 	 */
3489 	eicr &= 0xFFFF0000;
3490 
3491 	IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
3492 
3493 	if (eicr & IXGBE_EICR_LSC)
3494 		ixgbe_check_lsc(adapter);
3495 
3496 	if (eicr & IXGBE_EICR_MAILBOX)
3497 		ixgbe_msg_task(adapter);
3498 
3499 	if (eicr & IXGBE_EICR_FW_EVENT)
3500 		ixgbe_schedule_fw_event(adapter);
3501 
3502 	switch (hw->mac.type) {
3503 	case ixgbe_mac_82599EB:
3504 	case ixgbe_mac_X540:
3505 	case ixgbe_mac_X550:
3506 	case ixgbe_mac_X550EM_x:
3507 	case ixgbe_mac_x550em_a:
3508 	case ixgbe_mac_e610:
3509 		if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
3510 		    (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
3511 			adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
3512 			ixgbe_service_event_schedule(adapter);
3513 			IXGBE_WRITE_REG(hw, IXGBE_EICR,
3514 					IXGBE_EICR_GPI_SDP0_X540);
3515 		}
3516 		if (eicr & IXGBE_EICR_ECC) {
3517 			e_info(link, "Received ECC Err, initiating reset\n");
3518 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3519 			ixgbe_service_event_schedule(adapter);
3520 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3521 		}
3522 		/* Handle Flow Director Full threshold interrupt */
3523 		if (eicr & IXGBE_EICR_FLOW_DIR) {
3524 			int reinit_count = 0;
3525 			int i;
3526 			for (i = 0; i < adapter->num_tx_queues; i++) {
3527 				struct ixgbe_ring *ring = adapter->tx_ring[i];
3528 				if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
3529 						       &ring->state))
3530 					reinit_count++;
3531 			}
3532 			if (reinit_count) {
3533 				/* no more flow director interrupts until after init */
3534 				IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
3535 				adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
3536 				ixgbe_service_event_schedule(adapter);
3537 			}
3538 		}
3539 		ixgbe_check_sfp_event(adapter, eicr);
3540 		ixgbe_check_overtemp_event(adapter, eicr);
3541 		break;
3542 	default:
3543 		break;
3544 	}
3545 
3546 	ixgbe_check_fan_failure(adapter, eicr);
3547 
3548 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3549 		ixgbe_ptp_check_pps_event(adapter);
3550 
3551 	/* re-enable the original interrupt state, no lsc, no queues */
3552 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3553 		ixgbe_irq_enable(adapter, false, false);
3554 
3555 	return IRQ_HANDLED;
3556 }
3557 
3558 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
3559 {
3560 	struct ixgbe_q_vector *q_vector = data;
3561 
3562 	/* EIAM disabled interrupts (on this vector) for us */
3563 
3564 	if (q_vector->rx.ring || q_vector->tx.ring)
3565 		napi_schedule_irqoff(&q_vector->napi);
3566 
3567 	return IRQ_HANDLED;
3568 }
3569 
3570 /**
3571  * ixgbe_poll - NAPI Rx polling callback
3572  * @napi: structure for representing this polling device
3573  * @budget: how many packets driver is allowed to clean
3574  *
3575  * This function is used for legacy and MSI, NAPI mode
3576  **/
3577 int ixgbe_poll(struct napi_struct *napi, int budget)
3578 {
3579 	struct ixgbe_q_vector *q_vector =
3580 				container_of(napi, struct ixgbe_q_vector, napi);
3581 	struct ixgbe_adapter *adapter = q_vector->adapter;
3582 	struct ixgbe_ring *ring;
3583 	int per_ring_budget, work_done = 0;
3584 	bool clean_complete = true;
3585 
3586 #ifdef CONFIG_IXGBE_DCA
3587 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
3588 		ixgbe_update_dca(q_vector);
3589 #endif
3590 
3591 	ixgbe_for_each_ring(ring, q_vector->tx) {
3592 		bool wd = ring->xsk_pool ?
3593 			  ixgbe_clean_xdp_tx_irq(q_vector, ring, budget) :
3594 			  ixgbe_clean_tx_irq(q_vector, ring, budget);
3595 
3596 		if (!wd)
3597 			clean_complete = false;
3598 	}
3599 
3600 	/* Exit if we are called by netpoll */
3601 	if (budget <= 0)
3602 		return budget;
3603 
3604 	/* attempt to distribute budget to each queue fairly, but don't allow
3605 	 * the budget to go below 1 because we'll exit polling */
3606 	if (q_vector->rx.count > 1)
3607 		per_ring_budget = max(budget/q_vector->rx.count, 1);
3608 	else
3609 		per_ring_budget = budget;
3610 
3611 	ixgbe_for_each_ring(ring, q_vector->rx) {
3612 		int cleaned = ring->xsk_pool ?
3613 			      ixgbe_clean_rx_irq_zc(q_vector, ring,
3614 						    per_ring_budget) :
3615 			      ixgbe_clean_rx_irq(q_vector, ring,
3616 						 per_ring_budget);
3617 
3618 		work_done += cleaned;
3619 		if (cleaned >= per_ring_budget)
3620 			clean_complete = false;
3621 	}
3622 
3623 	/* If all work not completed, return budget and keep polling */
3624 	if (!clean_complete)
3625 		return budget;
3626 
3627 	/* all work done, exit the polling mode */
3628 	if (likely(napi_complete_done(napi, work_done))) {
3629 		if (adapter->rx_itr_setting & 1)
3630 			ixgbe_set_itr(q_vector);
3631 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
3632 			ixgbe_irq_enable_queues(adapter,
3633 						BIT_ULL(q_vector->v_idx));
3634 	}
3635 
3636 	return min(work_done, budget - 1);
3637 }
3638 
3639 /**
3640  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
3641  * @adapter: board private structure
3642  *
3643  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
3644  * interrupts from the kernel.
3645  **/
3646 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
3647 {
3648 	struct net_device *netdev = adapter->netdev;
3649 	unsigned int ri = 0, ti = 0;
3650 	int vector, err;
3651 
3652 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3653 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3654 		struct msix_entry *entry = &adapter->msix_entries[vector];
3655 
3656 		if (q_vector->tx.ring && q_vector->rx.ring) {
3657 			snprintf(q_vector->name, sizeof(q_vector->name),
3658 				 "%s-TxRx-%u", netdev->name, ri++);
3659 			ti++;
3660 		} else if (q_vector->rx.ring) {
3661 			snprintf(q_vector->name, sizeof(q_vector->name),
3662 				 "%s-rx-%u", netdev->name, ri++);
3663 		} else if (q_vector->tx.ring) {
3664 			snprintf(q_vector->name, sizeof(q_vector->name),
3665 				 "%s-tx-%u", netdev->name, ti++);
3666 		} else {
3667 			/* skip this unused q_vector */
3668 			continue;
3669 		}
3670 		err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
3671 				  q_vector->name, q_vector);
3672 		if (err) {
3673 			e_err(probe, "request_irq failed for MSIX interrupt "
3674 			      "Error: %d\n", err);
3675 			goto free_queue_irqs;
3676 		}
3677 		/* If Flow Director is enabled, set interrupt affinity */
3678 		if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3679 			/* assign the mask for this irq */
3680 			irq_update_affinity_hint(entry->vector,
3681 						 &q_vector->affinity_mask);
3682 		}
3683 	}
3684 
3685 	err = request_irq(adapter->msix_entries[vector].vector,
3686 			  ixgbe_msix_other, 0, netdev->name, adapter);
3687 	if (err) {
3688 		e_err(probe, "request_irq for msix_other failed: %d\n", err);
3689 		goto free_queue_irqs;
3690 	}
3691 
3692 	return 0;
3693 
3694 free_queue_irqs:
3695 	while (vector) {
3696 		vector--;
3697 		irq_update_affinity_hint(adapter->msix_entries[vector].vector,
3698 					 NULL);
3699 		free_irq(adapter->msix_entries[vector].vector,
3700 			 adapter->q_vector[vector]);
3701 	}
3702 	adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
3703 	pci_disable_msix(adapter->pdev);
3704 	kfree(adapter->msix_entries);
3705 	adapter->msix_entries = NULL;
3706 	return err;
3707 }
3708 
3709 /**
3710  * ixgbe_intr - legacy mode Interrupt Handler
3711  * @irq: interrupt number
3712  * @data: pointer to a network interface device structure
3713  **/
3714 static irqreturn_t ixgbe_intr(int irq, void *data)
3715 {
3716 	struct ixgbe_adapter *adapter = data;
3717 	struct ixgbe_hw *hw = &adapter->hw;
3718 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3719 	u32 eicr;
3720 
3721 	/*
3722 	 * Workaround for silicon errata #26 on 82598.  Mask the interrupt
3723 	 * before the read of EICR.
3724 	 */
3725 	IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
3726 
3727 	/* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
3728 	 * therefore no explicit interrupt disable is necessary */
3729 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3730 	if (!eicr) {
3731 		/*
3732 		 * shared interrupt alert!
3733 		 * make sure interrupts are enabled because the read will
3734 		 * have disabled interrupts due to EIAM
3735 		 * finish the workaround of silicon errata on 82598.  Unmask
3736 		 * the interrupt that we masked before the EICR read.
3737 		 */
3738 		if (!test_bit(__IXGBE_DOWN, &adapter->state))
3739 			ixgbe_irq_enable(adapter, true, true);
3740 		return IRQ_NONE;	/* Not our interrupt */
3741 	}
3742 
3743 	if (eicr & IXGBE_EICR_LSC)
3744 		ixgbe_check_lsc(adapter);
3745 
3746 	if (eicr & IXGBE_EICR_FW_EVENT)
3747 		ixgbe_schedule_fw_event(adapter);
3748 
3749 	switch (hw->mac.type) {
3750 	case ixgbe_mac_82599EB:
3751 		ixgbe_check_sfp_event(adapter, eicr);
3752 		fallthrough;
3753 	case ixgbe_mac_X540:
3754 	case ixgbe_mac_X550:
3755 	case ixgbe_mac_X550EM_x:
3756 	case ixgbe_mac_x550em_a:
3757 	case ixgbe_mac_e610:
3758 		if (eicr & IXGBE_EICR_ECC) {
3759 			e_info(link, "Received ECC Err, initiating reset\n");
3760 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3761 			ixgbe_service_event_schedule(adapter);
3762 			IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3763 		}
3764 		ixgbe_check_overtemp_event(adapter, eicr);
3765 		break;
3766 	default:
3767 		break;
3768 	}
3769 
3770 	ixgbe_check_fan_failure(adapter, eicr);
3771 	if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3772 		ixgbe_ptp_check_pps_event(adapter);
3773 
3774 	/* would disable interrupts here but EIAM disabled it */
3775 	napi_schedule_irqoff(&q_vector->napi);
3776 
3777 	/*
3778 	 * re-enable link(maybe) and non-queue interrupts, no flush.
3779 	 * ixgbe_poll will re-enable the queue interrupts
3780 	 */
3781 	if (!test_bit(__IXGBE_DOWN, &adapter->state))
3782 		ixgbe_irq_enable(adapter, false, false);
3783 
3784 	return IRQ_HANDLED;
3785 }
3786 
3787 /**
3788  * ixgbe_request_irq - initialize interrupts
3789  * @adapter: board private structure
3790  *
3791  * Attempts to configure interrupts using the best available
3792  * capabilities of the hardware and kernel.
3793  **/
3794 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
3795 {
3796 	struct net_device *netdev = adapter->netdev;
3797 	int err;
3798 
3799 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
3800 		err = ixgbe_request_msix_irqs(adapter);
3801 	else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
3802 		err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
3803 				  netdev->name, adapter);
3804 	else
3805 		err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
3806 				  netdev->name, adapter);
3807 
3808 	if (err)
3809 		e_err(probe, "request_irq failed, Error %d\n", err);
3810 
3811 	return err;
3812 }
3813 
3814 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
3815 {
3816 	int vector;
3817 
3818 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
3819 		free_irq(adapter->pdev->irq, adapter);
3820 		return;
3821 	}
3822 
3823 	if (!adapter->msix_entries)
3824 		return;
3825 
3826 	for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3827 		struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3828 		struct msix_entry *entry = &adapter->msix_entries[vector];
3829 
3830 		/* free only the irqs that were actually requested */
3831 		if (!q_vector->rx.ring && !q_vector->tx.ring)
3832 			continue;
3833 
3834 		/* clear the affinity_mask in the IRQ descriptor */
3835 		irq_update_affinity_hint(entry->vector, NULL);
3836 
3837 		free_irq(entry->vector, q_vector);
3838 	}
3839 
3840 	free_irq(adapter->msix_entries[vector].vector, adapter);
3841 }
3842 
3843 /**
3844  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3845  * @adapter: board private structure
3846  **/
3847 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3848 {
3849 	switch (adapter->hw.mac.type) {
3850 	case ixgbe_mac_82598EB:
3851 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3852 		break;
3853 	case ixgbe_mac_82599EB:
3854 	case ixgbe_mac_X540:
3855 	case ixgbe_mac_X550:
3856 	case ixgbe_mac_X550EM_x:
3857 	case ixgbe_mac_x550em_a:
3858 	case ixgbe_mac_e610:
3859 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3860 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3861 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3862 		break;
3863 	default:
3864 		break;
3865 	}
3866 	IXGBE_WRITE_FLUSH(&adapter->hw);
3867 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3868 		int vector;
3869 
3870 		for (vector = 0; vector < adapter->num_q_vectors; vector++)
3871 			synchronize_irq(adapter->msix_entries[vector].vector);
3872 
3873 		synchronize_irq(adapter->msix_entries[vector++].vector);
3874 	} else {
3875 		synchronize_irq(adapter->pdev->irq);
3876 	}
3877 }
3878 
3879 /**
3880  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3881  * @adapter: board private structure
3882  *
3883  **/
3884 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3885 {
3886 	struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3887 
3888 	ixgbe_write_eitr(q_vector);
3889 
3890 	ixgbe_set_ivar(adapter, 0, 0, 0);
3891 	ixgbe_set_ivar(adapter, 1, 0, 0);
3892 
3893 	e_info(hw, "Legacy interrupt IVAR setup done\n");
3894 }
3895 
3896 /**
3897  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3898  * @adapter: board private structure
3899  * @ring: structure containing ring specific data
3900  *
3901  * Configure the Tx descriptor ring after a reset.
3902  **/
3903 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3904 			     struct ixgbe_ring *ring)
3905 {
3906 	struct ixgbe_hw *hw = &adapter->hw;
3907 	u64 tdba = ring->dma;
3908 	int wait_loop = 10;
3909 	u32 txdctl = IXGBE_TXDCTL_ENABLE;
3910 	u8 reg_idx = ring->reg_idx;
3911 
3912 	ring->xsk_pool = NULL;
3913 	if (ring_is_xdp(ring))
3914 		ring->xsk_pool = ixgbe_xsk_pool(adapter, ring);
3915 
3916 	/* disable queue to avoid issues while updating state */
3917 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3918 	IXGBE_WRITE_FLUSH(hw);
3919 
3920 	IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3921 			(tdba & DMA_BIT_MASK(32)));
3922 	IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3923 	IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3924 			ring->count * sizeof(union ixgbe_adv_tx_desc));
3925 	IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3926 	IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3927 	ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3928 
3929 	/*
3930 	 * set WTHRESH to encourage burst writeback, it should not be set
3931 	 * higher than 1 when:
3932 	 * - ITR is 0 as it could cause false TX hangs
3933 	 * - ITR is set to > 100k int/sec and BQL is enabled
3934 	 *
3935 	 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3936 	 * to or less than the number of on chip descriptors, which is
3937 	 * currently 40.
3938 	 */
3939 	if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3940 		txdctl |= 1u << 16;	/* WTHRESH = 1 */
3941 	else
3942 		txdctl |= 8u << 16;	/* WTHRESH = 8 */
3943 
3944 	/*
3945 	 * Setting PTHRESH to 32 both improves performance
3946 	 * and avoids a TX hang with DFP enabled
3947 	 */
3948 	txdctl |= (1u << 8) |	/* HTHRESH = 1 */
3949 		   32;		/* PTHRESH = 32 */
3950 
3951 	/* reinitialize flowdirector state */
3952 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3953 		ring->atr_sample_rate = adapter->atr_sample_rate;
3954 		ring->atr_count = 0;
3955 		set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3956 	} else {
3957 		ring->atr_sample_rate = 0;
3958 	}
3959 
3960 	/* initialize XPS */
3961 	if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3962 		struct ixgbe_q_vector *q_vector = ring->q_vector;
3963 
3964 		if (q_vector)
3965 			netif_set_xps_queue(ring->netdev,
3966 					    &q_vector->affinity_mask,
3967 					    ring->queue_index);
3968 	}
3969 
3970 	clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3971 
3972 	/* reinitialize tx_buffer_info */
3973 	memset(ring->tx_buffer_info, 0,
3974 	       sizeof(struct ixgbe_tx_buffer) * ring->count);
3975 
3976 	/* enable queue */
3977 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3978 
3979 	/* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3980 	if (hw->mac.type == ixgbe_mac_82598EB &&
3981 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3982 		return;
3983 
3984 	/* poll to verify queue is enabled */
3985 	do {
3986 		usleep_range(1000, 2000);
3987 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3988 	} while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3989 	if (!wait_loop)
3990 		hw_dbg(hw, "Could not enable Tx Queue %d\n", reg_idx);
3991 }
3992 
3993 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3994 {
3995 	struct ixgbe_hw *hw = &adapter->hw;
3996 	u32 rttdcs, mtqc;
3997 	u8 tcs = adapter->hw_tcs;
3998 
3999 	if (hw->mac.type == ixgbe_mac_82598EB)
4000 		return;
4001 
4002 	/* disable the arbiter while setting MTQC */
4003 	rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
4004 	rttdcs |= IXGBE_RTTDCS_ARBDIS;
4005 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
4006 
4007 	/* set transmit pool layout */
4008 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
4009 		mtqc = IXGBE_MTQC_VT_ENA;
4010 		if (tcs > 4)
4011 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
4012 		else if (tcs > 1)
4013 			mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
4014 		else if (adapter->ring_feature[RING_F_VMDQ].mask ==
4015 			 IXGBE_82599_VMDQ_4Q_MASK)
4016 			mtqc |= IXGBE_MTQC_32VF;
4017 		else
4018 			mtqc |= IXGBE_MTQC_64VF;
4019 	} else {
4020 		if (tcs > 4) {
4021 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
4022 		} else if (tcs > 1) {
4023 			mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
4024 		} else {
4025 			u8 max_txq = adapter->num_tx_queues +
4026 				adapter->num_xdp_queues;
4027 			if (max_txq > 63)
4028 				mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
4029 			else
4030 				mtqc = IXGBE_MTQC_64Q_1PB;
4031 		}
4032 	}
4033 
4034 	IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
4035 
4036 	/* Enable Security TX Buffer IFG for multiple pb */
4037 	if (tcs) {
4038 		u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
4039 		sectx |= IXGBE_SECTX_DCB;
4040 		IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
4041 	}
4042 
4043 	/* re-enable the arbiter */
4044 	rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
4045 	IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
4046 }
4047 
4048 /**
4049  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
4050  * @adapter: board private structure
4051  *
4052  * Configure the Tx unit of the MAC after a reset.
4053  **/
4054 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
4055 {
4056 	struct ixgbe_hw *hw = &adapter->hw;
4057 	u32 dmatxctl;
4058 	u32 i;
4059 
4060 	ixgbe_setup_mtqc(adapter);
4061 
4062 	if (hw->mac.type != ixgbe_mac_82598EB) {
4063 		/* DMATXCTL.EN must be before Tx queues are enabled */
4064 		dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
4065 		dmatxctl |= IXGBE_DMATXCTL_TE;
4066 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
4067 	}
4068 
4069 	/* Setup the HW Tx Head and Tail descriptor pointers */
4070 	for (i = 0; i < adapter->num_tx_queues; i++)
4071 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
4072 	for (i = 0; i < adapter->num_xdp_queues; i++)
4073 		ixgbe_configure_tx_ring(adapter, adapter->xdp_ring[i]);
4074 }
4075 
4076 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
4077 				 struct ixgbe_ring *ring)
4078 {
4079 	struct ixgbe_hw *hw = &adapter->hw;
4080 	u8 reg_idx = ring->reg_idx;
4081 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
4082 
4083 	srrctl |= IXGBE_SRRCTL_DROP_EN;
4084 
4085 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
4086 }
4087 
4088 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
4089 				  struct ixgbe_ring *ring)
4090 {
4091 	struct ixgbe_hw *hw = &adapter->hw;
4092 	u8 reg_idx = ring->reg_idx;
4093 	u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
4094 
4095 	srrctl &= ~IXGBE_SRRCTL_DROP_EN;
4096 
4097 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
4098 }
4099 
4100 #ifdef CONFIG_IXGBE_DCB
4101 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
4102 #else
4103 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
4104 #endif
4105 {
4106 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
4107 	struct ixgbe_hw *hw = &adapter->hw;
4108 	int i;
4109 
4110 	if (hw->mac.ops.disable_mdd)
4111 		hw->mac.ops.disable_mdd(hw);
4112 
4113 	if (adapter->ixgbe_ieee_pfc)
4114 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
4115 
4116 	/*
4117 	 * We should set the drop enable bit if:
4118 	 *  SR-IOV is enabled
4119 	 *   or
4120 	 *  Number of Rx queues > 1 and flow control is disabled
4121 	 *
4122 	 *  This allows us to avoid head of line blocking for security
4123 	 *  and performance reasons.
4124 	 */
4125 	if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
4126 	    !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
4127 		for (i = 0; i < adapter->num_rx_queues; i++)
4128 			ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
4129 	} else {
4130 		for (i = 0; i < adapter->num_rx_queues; i++)
4131 			ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
4132 	}
4133 
4134 	if (hw->mac.ops.enable_mdd)
4135 		hw->mac.ops.enable_mdd(hw);
4136 }
4137 
4138 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
4139 
4140 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
4141 				   struct ixgbe_ring *rx_ring)
4142 {
4143 	struct ixgbe_hw *hw = &adapter->hw;
4144 	u32 srrctl;
4145 	u8 reg_idx = rx_ring->reg_idx;
4146 
4147 	if (hw->mac.type == ixgbe_mac_82598EB) {
4148 		u16 mask = adapter->ring_feature[RING_F_RSS].mask;
4149 
4150 		/*
4151 		 * if VMDq is not active we must program one srrctl register
4152 		 * per RSS queue since we have enabled RDRXCTL.MVMEN
4153 		 */
4154 		reg_idx &= mask;
4155 	}
4156 
4157 	/* configure header buffer length, needed for RSC */
4158 	srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
4159 
4160 	/* configure the packet buffer length */
4161 	if (rx_ring->xsk_pool) {
4162 		u32 xsk_buf_len = xsk_pool_get_rx_frame_size(rx_ring->xsk_pool);
4163 
4164 		/* If the MAC support setting RXDCTL.RLPML, the
4165 		 * SRRCTL[n].BSIZEPKT is set to PAGE_SIZE and
4166 		 * RXDCTL.RLPML is set to the actual UMEM buffer
4167 		 * size. If not, then we are stuck with a 1k buffer
4168 		 * size resolution. In this case frames larger than
4169 		 * the UMEM buffer size viewed in a 1k resolution will
4170 		 * be dropped.
4171 		 */
4172 		if (hw->mac.type != ixgbe_mac_82599EB)
4173 			srrctl |= PAGE_SIZE >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
4174 		else
4175 			srrctl |= xsk_buf_len >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
4176 	} else if (test_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state)) {
4177 		srrctl |= IXGBE_RXBUFFER_3K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
4178 	} else {
4179 		srrctl |= IXGBE_RXBUFFER_2K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
4180 	}
4181 
4182 	/* configure descriptor type */
4183 	srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
4184 
4185 	IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
4186 }
4187 
4188 /**
4189  * ixgbe_rss_indir_tbl_entries - Return RSS indirection table entries
4190  * @adapter: device handle
4191  *
4192  *  - 82598/82599/X540:     128
4193  *  - X550(non-SRIOV mode): 512
4194  *  - X550(SRIOV mode):     64
4195  */
4196 u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
4197 {
4198 	if (adapter->hw.mac.type < ixgbe_mac_X550)
4199 		return 128;
4200 	else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
4201 		return 64;
4202 	else
4203 		return 512;
4204 }
4205 
4206 /**
4207  * ixgbe_store_key - Write the RSS key to HW
4208  * @adapter: device handle
4209  *
4210  * Write the RSS key stored in adapter.rss_key to HW.
4211  */
4212 void ixgbe_store_key(struct ixgbe_adapter *adapter)
4213 {
4214 	struct ixgbe_hw *hw = &adapter->hw;
4215 	int i;
4216 
4217 	for (i = 0; i < 10; i++)
4218 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
4219 }
4220 
4221 /**
4222  * ixgbe_init_rss_key - Initialize adapter RSS key
4223  * @adapter: device handle
4224  *
4225  * Allocates and initializes the RSS key if it is not allocated.
4226  **/
4227 static inline int ixgbe_init_rss_key(struct ixgbe_adapter *adapter)
4228 {
4229 	u32 *rss_key;
4230 
4231 	if (!adapter->rss_key) {
4232 		rss_key = kzalloc(IXGBE_RSS_KEY_SIZE, GFP_KERNEL);
4233 		if (unlikely(!rss_key))
4234 			return -ENOMEM;
4235 
4236 		netdev_rss_key_fill(rss_key, IXGBE_RSS_KEY_SIZE);
4237 		adapter->rss_key = rss_key;
4238 	}
4239 
4240 	return 0;
4241 }
4242 
4243 /**
4244  * ixgbe_store_reta - Write the RETA table to HW
4245  * @adapter: device handle
4246  *
4247  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
4248  */
4249 void ixgbe_store_reta(struct ixgbe_adapter *adapter)
4250 {
4251 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
4252 	struct ixgbe_hw *hw = &adapter->hw;
4253 	u32 reta = 0;
4254 	u32 indices_multi;
4255 	u8 *indir_tbl = adapter->rss_indir_tbl;
4256 
4257 	/* Fill out the redirection table as follows:
4258 	 *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
4259 	 *    indices.
4260 	 *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
4261 	 *  - X550:       8 bit wide entries containing 6 bit RSS index
4262 	 */
4263 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
4264 		indices_multi = 0x11;
4265 	else
4266 		indices_multi = 0x1;
4267 
4268 	/* Write redirection table to HW */
4269 	for (i = 0; i < reta_entries; i++) {
4270 		reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
4271 		if ((i & 3) == 3) {
4272 			if (i < 128)
4273 				IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
4274 			else
4275 				IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
4276 						reta);
4277 			reta = 0;
4278 		}
4279 	}
4280 }
4281 
4282 /**
4283  * ixgbe_store_vfreta - Write the RETA table to HW (x550 devices in SRIOV mode)
4284  * @adapter: device handle
4285  *
4286  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
4287  */
4288 static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
4289 {
4290 	u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
4291 	struct ixgbe_hw *hw = &adapter->hw;
4292 	u32 vfreta = 0;
4293 
4294 	/* Write redirection table to HW */
4295 	for (i = 0; i < reta_entries; i++) {
4296 		u16 pool = adapter->num_rx_pools;
4297 
4298 		vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
4299 		if ((i & 3) != 3)
4300 			continue;
4301 
4302 		while (pool--)
4303 			IXGBE_WRITE_REG(hw,
4304 					IXGBE_PFVFRETA(i >> 2, VMDQ_P(pool)),
4305 					vfreta);
4306 		vfreta = 0;
4307 	}
4308 }
4309 
4310 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
4311 {
4312 	u32 i, j;
4313 	u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
4314 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
4315 
4316 	/* Program table for at least 4 queues w/ SR-IOV so that VFs can
4317 	 * make full use of any rings they may have.  We will use the
4318 	 * PSRTYPE register to control how many rings we use within the PF.
4319 	 */
4320 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 4))
4321 		rss_i = 4;
4322 
4323 	/* Fill out hash function seeds */
4324 	ixgbe_store_key(adapter);
4325 
4326 	/* Fill out redirection table */
4327 	memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
4328 
4329 	for (i = 0, j = 0; i < reta_entries; i++, j++) {
4330 		if (j == rss_i)
4331 			j = 0;
4332 
4333 		adapter->rss_indir_tbl[i] = j;
4334 	}
4335 
4336 	ixgbe_store_reta(adapter);
4337 }
4338 
4339 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
4340 {
4341 	struct ixgbe_hw *hw = &adapter->hw;
4342 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
4343 	int i, j;
4344 
4345 	/* Fill out hash function seeds */
4346 	for (i = 0; i < 10; i++) {
4347 		u16 pool = adapter->num_rx_pools;
4348 
4349 		while (pool--)
4350 			IXGBE_WRITE_REG(hw,
4351 					IXGBE_PFVFRSSRK(i, VMDQ_P(pool)),
4352 					*(adapter->rss_key + i));
4353 	}
4354 
4355 	/* Fill out the redirection table */
4356 	for (i = 0, j = 0; i < 64; i++, j++) {
4357 		if (j == rss_i)
4358 			j = 0;
4359 
4360 		adapter->rss_indir_tbl[i] = j;
4361 	}
4362 
4363 	ixgbe_store_vfreta(adapter);
4364 }
4365 
4366 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
4367 {
4368 	struct ixgbe_hw *hw = &adapter->hw;
4369 	u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
4370 	u32 rxcsum;
4371 
4372 	/* Disable indicating checksum in descriptor, enables RSS hash */
4373 	rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
4374 	rxcsum |= IXGBE_RXCSUM_PCSD;
4375 	IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
4376 
4377 	if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
4378 		if (adapter->ring_feature[RING_F_RSS].mask)
4379 			mrqc = IXGBE_MRQC_RSSEN;
4380 	} else {
4381 		u8 tcs = adapter->hw_tcs;
4382 
4383 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
4384 			if (tcs > 4)
4385 				mrqc = IXGBE_MRQC_VMDQRT8TCEN;	/* 8 TCs */
4386 			else if (tcs > 1)
4387 				mrqc = IXGBE_MRQC_VMDQRT4TCEN;	/* 4 TCs */
4388 			else if (adapter->ring_feature[RING_F_VMDQ].mask ==
4389 				 IXGBE_82599_VMDQ_4Q_MASK)
4390 				mrqc = IXGBE_MRQC_VMDQRSS32EN;
4391 			else
4392 				mrqc = IXGBE_MRQC_VMDQRSS64EN;
4393 
4394 			/* Enable L3/L4 for Tx Switched packets only for X550,
4395 			 * older devices do not support this feature
4396 			 */
4397 			if (hw->mac.type >= ixgbe_mac_X550)
4398 				mrqc |= IXGBE_MRQC_L3L4TXSWEN;
4399 		} else {
4400 			if (tcs > 4)
4401 				mrqc = IXGBE_MRQC_RTRSS8TCEN;
4402 			else if (tcs > 1)
4403 				mrqc = IXGBE_MRQC_RTRSS4TCEN;
4404 			else
4405 				mrqc = IXGBE_MRQC_RSSEN;
4406 		}
4407 	}
4408 
4409 	/* Perform hash on these packet types */
4410 	rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
4411 		     IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
4412 		     IXGBE_MRQC_RSS_FIELD_IPV6 |
4413 		     IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
4414 
4415 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
4416 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
4417 	if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
4418 		rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
4419 
4420 	if ((hw->mac.type >= ixgbe_mac_X550) &&
4421 	    (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
4422 		u16 pool = adapter->num_rx_pools;
4423 
4424 		/* Enable VF RSS mode */
4425 		mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
4426 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
4427 
4428 		/* Setup RSS through the VF registers */
4429 		ixgbe_setup_vfreta(adapter);
4430 		vfmrqc = IXGBE_MRQC_RSSEN;
4431 		vfmrqc |= rss_field;
4432 
4433 		while (pool--)
4434 			IXGBE_WRITE_REG(hw,
4435 					IXGBE_PFVFMRQC(VMDQ_P(pool)),
4436 					vfmrqc);
4437 	} else {
4438 		ixgbe_setup_reta(adapter);
4439 		mrqc |= rss_field;
4440 		IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
4441 	}
4442 }
4443 
4444 /**
4445  * ixgbe_configure_rscctl - enable RSC for the indicated ring
4446  * @adapter: address of board private structure
4447  * @ring: structure containing ring specific data
4448  **/
4449 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
4450 				   struct ixgbe_ring *ring)
4451 {
4452 	struct ixgbe_hw *hw = &adapter->hw;
4453 	u32 rscctrl;
4454 	u8 reg_idx = ring->reg_idx;
4455 
4456 	if (!ring_is_rsc_enabled(ring))
4457 		return;
4458 
4459 	rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
4460 	rscctrl |= IXGBE_RSCCTL_RSCEN;
4461 	/*
4462 	 * we must limit the number of descriptors so that the
4463 	 * total size of max desc * buf_len is not greater
4464 	 * than 65536
4465 	 */
4466 	rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
4467 	IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
4468 }
4469 
4470 #define IXGBE_MAX_RX_DESC_POLL 10
4471 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
4472 				       struct ixgbe_ring *ring)
4473 {
4474 	struct ixgbe_hw *hw = &adapter->hw;
4475 	int wait_loop = IXGBE_MAX_RX_DESC_POLL;
4476 	u32 rxdctl;
4477 	u8 reg_idx = ring->reg_idx;
4478 
4479 	if (ixgbe_removed(hw->hw_addr))
4480 		return;
4481 	/* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
4482 	if (hw->mac.type == ixgbe_mac_82598EB &&
4483 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
4484 		return;
4485 
4486 	do {
4487 		usleep_range(1000, 2000);
4488 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4489 	} while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
4490 
4491 	if (!wait_loop) {
4492 		e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
4493 		      "the polling period\n", reg_idx);
4494 	}
4495 }
4496 
4497 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
4498 			     struct ixgbe_ring *ring)
4499 {
4500 	struct ixgbe_hw *hw = &adapter->hw;
4501 	union ixgbe_adv_rx_desc *rx_desc;
4502 	u64 rdba = ring->dma;
4503 	u32 rxdctl;
4504 	u8 reg_idx = ring->reg_idx;
4505 
4506 	xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq);
4507 	ring->xsk_pool = ixgbe_xsk_pool(adapter, ring);
4508 	if (ring->xsk_pool) {
4509 		WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
4510 						   MEM_TYPE_XSK_BUFF_POOL,
4511 						   NULL));
4512 		xsk_pool_set_rxq_info(ring->xsk_pool, &ring->xdp_rxq);
4513 	} else {
4514 		WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
4515 						   MEM_TYPE_PAGE_SHARED, NULL));
4516 	}
4517 
4518 	/* disable queue to avoid use of these values while updating state */
4519 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4520 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
4521 
4522 	/* write value back with RXDCTL.ENABLE bit cleared */
4523 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4524 	IXGBE_WRITE_FLUSH(hw);
4525 
4526 	IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
4527 	IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
4528 	IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
4529 			ring->count * sizeof(union ixgbe_adv_rx_desc));
4530 	/* Force flushing of IXGBE_RDLEN to prevent MDD */
4531 	IXGBE_WRITE_FLUSH(hw);
4532 
4533 	IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
4534 	IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
4535 	ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
4536 
4537 	ixgbe_configure_srrctl(adapter, ring);
4538 	ixgbe_configure_rscctl(adapter, ring);
4539 
4540 	if (hw->mac.type == ixgbe_mac_82598EB) {
4541 		/*
4542 		 * enable cache line friendly hardware writes:
4543 		 * PTHRESH=32 descriptors (half the internal cache),
4544 		 * this also removes ugly rx_no_buffer_count increment
4545 		 * HTHRESH=4 descriptors (to minimize latency on fetch)
4546 		 * WTHRESH=8 burst writeback up to two cache lines
4547 		 */
4548 		rxdctl &= ~0x3FFFFF;
4549 		rxdctl |=  0x080420;
4550 #if (PAGE_SIZE < 8192)
4551 	/* RXDCTL.RLPML does not work on 82599 */
4552 	} else if (hw->mac.type != ixgbe_mac_82599EB) {
4553 		rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
4554 			    IXGBE_RXDCTL_RLPML_EN);
4555 
4556 		/* Limit the maximum frame size so we don't overrun the skb.
4557 		 * This can happen in SRIOV mode when the MTU of the VF is
4558 		 * higher than the MTU of the PF.
4559 		 */
4560 		if (ring_uses_build_skb(ring) &&
4561 		    !test_bit(__IXGBE_RX_3K_BUFFER, &ring->state))
4562 			rxdctl |= IXGBE_MAX_2K_FRAME_BUILD_SKB |
4563 				  IXGBE_RXDCTL_RLPML_EN;
4564 #endif
4565 	}
4566 
4567 	ring->rx_offset = ixgbe_rx_offset(ring);
4568 
4569 	if (ring->xsk_pool && hw->mac.type != ixgbe_mac_82599EB) {
4570 		u32 xsk_buf_len = xsk_pool_get_rx_frame_size(ring->xsk_pool);
4571 
4572 		rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
4573 			    IXGBE_RXDCTL_RLPML_EN);
4574 		rxdctl |= xsk_buf_len | IXGBE_RXDCTL_RLPML_EN;
4575 
4576 		ring->rx_buf_len = xsk_buf_len;
4577 	}
4578 
4579 	/* initialize rx_buffer_info */
4580 	memset(ring->rx_buffer_info, 0,
4581 	       sizeof(struct ixgbe_rx_buffer) * ring->count);
4582 
4583 	/* initialize Rx descriptor 0 */
4584 	rx_desc = IXGBE_RX_DESC(ring, 0);
4585 	rx_desc->wb.upper.length = 0;
4586 
4587 	/* enable receive descriptor ring */
4588 	rxdctl |= IXGBE_RXDCTL_ENABLE;
4589 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4590 
4591 	ixgbe_rx_desc_queue_enable(adapter, ring);
4592 	if (ring->xsk_pool)
4593 		ixgbe_alloc_rx_buffers_zc(ring, ixgbe_desc_unused(ring));
4594 	else
4595 		ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
4596 }
4597 
4598 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
4599 {
4600 	struct ixgbe_hw *hw = &adapter->hw;
4601 	int rss_i = adapter->ring_feature[RING_F_RSS].indices;
4602 	u16 pool = adapter->num_rx_pools;
4603 
4604 	/* PSRTYPE must be initialized in non 82598 adapters */
4605 	u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4606 		      IXGBE_PSRTYPE_UDPHDR |
4607 		      IXGBE_PSRTYPE_IPV4HDR |
4608 		      IXGBE_PSRTYPE_L2HDR |
4609 		      IXGBE_PSRTYPE_IPV6HDR;
4610 
4611 	if (hw->mac.type == ixgbe_mac_82598EB)
4612 		return;
4613 
4614 	if (rss_i > 3)
4615 		psrtype |= 2u << 29;
4616 	else if (rss_i > 1)
4617 		psrtype |= 1u << 29;
4618 
4619 	while (pool--)
4620 		IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4621 }
4622 
4623 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
4624 {
4625 	struct ixgbe_hw *hw = &adapter->hw;
4626 	u16 pool = adapter->num_rx_pools;
4627 	u32 reg_offset, vf_shift, vmolr;
4628 	u32 gcr_ext, vmdctl;
4629 	int i;
4630 
4631 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
4632 		return;
4633 
4634 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
4635 	vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
4636 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
4637 	vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
4638 	vmdctl |= IXGBE_VT_CTL_REPLEN;
4639 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
4640 
4641 	/* accept untagged packets until a vlan tag is
4642 	 * specifically set for the VMDQ queue/pool
4643 	 */
4644 	vmolr = IXGBE_VMOLR_AUPE;
4645 	while (pool--)
4646 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(pool)), vmolr);
4647 
4648 	vf_shift = VMDQ_P(0) % 32;
4649 	reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
4650 
4651 	/* Enable only the PF's pool for Tx/Rx */
4652 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), GENMASK(31, vf_shift));
4653 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
4654 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), GENMASK(31, vf_shift));
4655 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
4656 	if (adapter->bridge_mode == BRIDGE_MODE_VEB)
4657 		IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
4658 
4659 	/* Map PF MAC address in RAR Entry 0 to first pool following VFs */
4660 	hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
4661 
4662 	/* clear VLAN promisc flag so VFTA will be updated if necessary */
4663 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4664 
4665 	/*
4666 	 * Set up VF register offsets for selected VT Mode,
4667 	 * i.e. 32 or 64 VFs for SR-IOV
4668 	 */
4669 	switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4670 	case IXGBE_82599_VMDQ_8Q_MASK:
4671 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
4672 		break;
4673 	case IXGBE_82599_VMDQ_4Q_MASK:
4674 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
4675 		break;
4676 	default:
4677 		gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
4678 		break;
4679 	}
4680 
4681 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
4682 
4683 	for (i = 0; i < adapter->num_vfs; i++) {
4684 		/* configure spoof checking */
4685 		ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i,
4686 					  adapter->vfinfo[i].spoofchk_enabled);
4687 
4688 		/* Enable/Disable RSS query feature  */
4689 		ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
4690 					  adapter->vfinfo[i].rss_query_enabled);
4691 	}
4692 }
4693 
4694 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
4695 {
4696 	struct ixgbe_hw *hw = &adapter->hw;
4697 	struct net_device *netdev = adapter->netdev;
4698 	int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4699 	struct ixgbe_ring *rx_ring;
4700 	int i;
4701 	u32 mhadd, hlreg0;
4702 
4703 #ifdef IXGBE_FCOE
4704 	/* adjust max frame to be able to do baby jumbo for FCoE */
4705 	if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
4706 	    (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
4707 		max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4708 
4709 #endif /* IXGBE_FCOE */
4710 
4711 	/* adjust max frame to be at least the size of a standard frame */
4712 	if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
4713 		max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
4714 
4715 	mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
4716 	if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
4717 		mhadd &= ~IXGBE_MHADD_MFS_MASK;
4718 		mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
4719 
4720 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
4721 	}
4722 
4723 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4724 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
4725 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4726 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4727 
4728 	/*
4729 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4730 	 * the Base and Length of the Rx Descriptor Ring
4731 	 */
4732 	for (i = 0; i < adapter->num_rx_queues; i++) {
4733 		rx_ring = adapter->rx_ring[i];
4734 
4735 		clear_ring_rsc_enabled(rx_ring);
4736 		clear_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4737 		clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4738 
4739 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4740 			set_ring_rsc_enabled(rx_ring);
4741 
4742 		if (test_bit(__IXGBE_RX_FCOE, &rx_ring->state))
4743 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4744 
4745 		if (adapter->flags2 & IXGBE_FLAG2_RX_LEGACY)
4746 			continue;
4747 
4748 		set_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4749 
4750 #if (PAGE_SIZE < 8192)
4751 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4752 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4753 
4754 		if (IXGBE_2K_TOO_SMALL_WITH_PADDING ||
4755 		    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
4756 			set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4757 #endif
4758 	}
4759 }
4760 
4761 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
4762 {
4763 	struct ixgbe_hw *hw = &adapter->hw;
4764 	u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
4765 
4766 	switch (hw->mac.type) {
4767 	case ixgbe_mac_82598EB:
4768 		/*
4769 		 * For VMDq support of different descriptor types or
4770 		 * buffer sizes through the use of multiple SRRCTL
4771 		 * registers, RDRXCTL.MVMEN must be set to 1
4772 		 *
4773 		 * also, the manual doesn't mention it clearly but DCA hints
4774 		 * will only use queue 0's tags unless this bit is set.  Side
4775 		 * effects of setting this bit are only that SRRCTL must be
4776 		 * fully programmed [0..15]
4777 		 */
4778 		rdrxctl |= IXGBE_RDRXCTL_MVMEN;
4779 		break;
4780 	case ixgbe_mac_X550:
4781 	case ixgbe_mac_X550EM_x:
4782 	case ixgbe_mac_x550em_a:
4783 	case ixgbe_mac_e610:
4784 		if (adapter->num_vfs)
4785 			rdrxctl |= IXGBE_RDRXCTL_PSP;
4786 		fallthrough;
4787 	case ixgbe_mac_82599EB:
4788 	case ixgbe_mac_X540:
4789 		/* Disable RSC for ACK packets */
4790 		IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
4791 		   (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
4792 		rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
4793 		/* hardware requires some bits to be set by default */
4794 		rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
4795 		rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
4796 		break;
4797 	default:
4798 		/* We should do nothing since we don't know this hardware */
4799 		return;
4800 	}
4801 
4802 	IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
4803 }
4804 
4805 /**
4806  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
4807  * @adapter: board private structure
4808  *
4809  * Configure the Rx unit of the MAC after a reset.
4810  **/
4811 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
4812 {
4813 	struct ixgbe_hw *hw = &adapter->hw;
4814 	int i;
4815 	u32 rxctrl, rfctl;
4816 
4817 	/* disable receives while setting up the descriptors */
4818 	hw->mac.ops.disable_rx(hw);
4819 
4820 	ixgbe_setup_psrtype(adapter);
4821 	ixgbe_setup_rdrxctl(adapter);
4822 
4823 	/* RSC Setup */
4824 	rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
4825 	rfctl &= ~IXGBE_RFCTL_RSC_DIS;
4826 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
4827 		rfctl |= IXGBE_RFCTL_RSC_DIS;
4828 
4829 	/* disable NFS filtering */
4830 	rfctl |= (IXGBE_RFCTL_NFSW_DIS | IXGBE_RFCTL_NFSR_DIS);
4831 	IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
4832 
4833 	/* Program registers for the distribution of queues */
4834 	ixgbe_setup_mrqc(adapter);
4835 
4836 	/* set_rx_buffer_len must be called before ring initialization */
4837 	ixgbe_set_rx_buffer_len(adapter);
4838 
4839 	/*
4840 	 * Setup the HW Rx Head and Tail Descriptor Pointers and
4841 	 * the Base and Length of the Rx Descriptor Ring
4842 	 */
4843 	for (i = 0; i < adapter->num_rx_queues; i++)
4844 		ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
4845 
4846 	rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4847 	/* disable drop enable for 82598 parts */
4848 	if (hw->mac.type == ixgbe_mac_82598EB)
4849 		rxctrl |= IXGBE_RXCTRL_DMBYPS;
4850 
4851 	/* enable all receives */
4852 	rxctrl |= IXGBE_RXCTRL_RXEN;
4853 	hw->mac.ops.enable_rx_dma(hw, rxctrl);
4854 }
4855 
4856 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
4857 				 __be16 proto, u16 vid)
4858 {
4859 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
4860 	struct ixgbe_hw *hw = &adapter->hw;
4861 
4862 	/* add VID to filter table */
4863 	if (!vid || !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4864 		hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true, !!vid);
4865 
4866 	set_bit(vid, adapter->active_vlans);
4867 
4868 	return 0;
4869 }
4870 
4871 static int ixgbe_find_vlvf_entry(struct ixgbe_hw *hw, u32 vlan)
4872 {
4873 	u32 vlvf;
4874 	int idx;
4875 
4876 	/* short cut the special case */
4877 	if (vlan == 0)
4878 		return 0;
4879 
4880 	/* Search for the vlan id in the VLVF entries */
4881 	for (idx = IXGBE_VLVF_ENTRIES; --idx;) {
4882 		vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(idx));
4883 		if ((vlvf & VLAN_VID_MASK) == vlan)
4884 			break;
4885 	}
4886 
4887 	return idx;
4888 }
4889 
4890 void ixgbe_update_pf_promisc_vlvf(struct ixgbe_adapter *adapter, u32 vid)
4891 {
4892 	struct ixgbe_hw *hw = &adapter->hw;
4893 	u32 bits, word;
4894 	int idx;
4895 
4896 	idx = ixgbe_find_vlvf_entry(hw, vid);
4897 	if (!idx)
4898 		return;
4899 
4900 	/* See if any other pools are set for this VLAN filter
4901 	 * entry other than the PF.
4902 	 */
4903 	word = idx * 2 + (VMDQ_P(0) / 32);
4904 	bits = ~BIT(VMDQ_P(0) % 32);
4905 	bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4906 
4907 	/* Disable the filter so this falls into the default pool. */
4908 	if (!bits && !IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1))) {
4909 		if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4910 			IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), 0);
4911 		IXGBE_WRITE_REG(hw, IXGBE_VLVF(idx), 0);
4912 	}
4913 }
4914 
4915 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
4916 				  __be16 proto, u16 vid)
4917 {
4918 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
4919 	struct ixgbe_hw *hw = &adapter->hw;
4920 
4921 	/* remove VID from filter table */
4922 	if (vid && !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4923 		hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), false, true);
4924 
4925 	clear_bit(vid, adapter->active_vlans);
4926 
4927 	return 0;
4928 }
4929 
4930 /**
4931  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
4932  * @adapter: driver data
4933  */
4934 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
4935 {
4936 	struct ixgbe_hw *hw = &adapter->hw;
4937 	u32 vlnctrl;
4938 	int i, j;
4939 
4940 	switch (hw->mac.type) {
4941 	case ixgbe_mac_82598EB:
4942 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4943 		vlnctrl &= ~IXGBE_VLNCTRL_VME;
4944 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4945 		break;
4946 	case ixgbe_mac_82599EB:
4947 	case ixgbe_mac_X540:
4948 	case ixgbe_mac_X550:
4949 	case ixgbe_mac_X550EM_x:
4950 	case ixgbe_mac_x550em_a:
4951 	case ixgbe_mac_e610:
4952 		for (i = 0; i < adapter->num_rx_queues; i++) {
4953 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4954 
4955 			if (!netif_is_ixgbe(ring->netdev))
4956 				continue;
4957 
4958 			j = ring->reg_idx;
4959 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4960 			vlnctrl &= ~IXGBE_RXDCTL_VME;
4961 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4962 		}
4963 		break;
4964 	default:
4965 		break;
4966 	}
4967 }
4968 
4969 /**
4970  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
4971  * @adapter: driver data
4972  */
4973 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
4974 {
4975 	struct ixgbe_hw *hw = &adapter->hw;
4976 	u32 vlnctrl;
4977 	int i, j;
4978 
4979 	switch (hw->mac.type) {
4980 	case ixgbe_mac_82598EB:
4981 		vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4982 		vlnctrl |= IXGBE_VLNCTRL_VME;
4983 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4984 		break;
4985 	case ixgbe_mac_82599EB:
4986 	case ixgbe_mac_X540:
4987 	case ixgbe_mac_X550:
4988 	case ixgbe_mac_X550EM_x:
4989 	case ixgbe_mac_x550em_a:
4990 	case ixgbe_mac_e610:
4991 		for (i = 0; i < adapter->num_rx_queues; i++) {
4992 			struct ixgbe_ring *ring = adapter->rx_ring[i];
4993 
4994 			if (!netif_is_ixgbe(ring->netdev))
4995 				continue;
4996 
4997 			j = ring->reg_idx;
4998 			vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4999 			vlnctrl |= IXGBE_RXDCTL_VME;
5000 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
5001 		}
5002 		break;
5003 	default:
5004 		break;
5005 	}
5006 }
5007 
5008 static void ixgbe_vlan_promisc_enable(struct ixgbe_adapter *adapter)
5009 {
5010 	struct ixgbe_hw *hw = &adapter->hw;
5011 	u32 vlnctrl, i;
5012 
5013 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
5014 
5015 	if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) {
5016 	/* For VMDq and SR-IOV we must leave VLAN filtering enabled */
5017 		vlnctrl |= IXGBE_VLNCTRL_VFE;
5018 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
5019 	} else {
5020 		vlnctrl &= ~IXGBE_VLNCTRL_VFE;
5021 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
5022 		return;
5023 	}
5024 
5025 	/* Nothing to do for 82598 */
5026 	if (hw->mac.type == ixgbe_mac_82598EB)
5027 		return;
5028 
5029 	/* We are already in VLAN promisc, nothing to do */
5030 	if (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC)
5031 		return;
5032 
5033 	/* Set flag so we don't redo unnecessary work */
5034 	adapter->flags2 |= IXGBE_FLAG2_VLAN_PROMISC;
5035 
5036 	/* Add PF to all active pools */
5037 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
5038 		u32 reg_offset = IXGBE_VLVFB(i * 2 + VMDQ_P(0) / 32);
5039 		u32 vlvfb = IXGBE_READ_REG(hw, reg_offset);
5040 
5041 		vlvfb |= BIT(VMDQ_P(0) % 32);
5042 		IXGBE_WRITE_REG(hw, reg_offset, vlvfb);
5043 	}
5044 
5045 	/* Set all bits in the VLAN filter table array */
5046 	for (i = hw->mac.vft_size; i--;)
5047 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), ~0U);
5048 }
5049 
5050 #define VFTA_BLOCK_SIZE 8
5051 static void ixgbe_scrub_vfta(struct ixgbe_adapter *adapter, u32 vfta_offset)
5052 {
5053 	struct ixgbe_hw *hw = &adapter->hw;
5054 	u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
5055 	u32 vid_start = vfta_offset * 32;
5056 	u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
5057 	u32 i, vid, word, bits;
5058 
5059 	for (i = IXGBE_VLVF_ENTRIES; --i;) {
5060 		u32 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
5061 
5062 		/* pull VLAN ID from VLVF */
5063 		vid = vlvf & VLAN_VID_MASK;
5064 
5065 		/* only concern ourselves with a certain range */
5066 		if (vid < vid_start || vid >= vid_end)
5067 			continue;
5068 
5069 		if (vlvf) {
5070 			/* record VLAN ID in VFTA */
5071 			vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
5072 
5073 			/* if PF is part of this then continue */
5074 			if (test_bit(vid, adapter->active_vlans))
5075 				continue;
5076 		}
5077 
5078 		/* remove PF from the pool */
5079 		word = i * 2 + VMDQ_P(0) / 32;
5080 		bits = ~BIT(VMDQ_P(0) % 32);
5081 		bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
5082 		IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), bits);
5083 	}
5084 
5085 	/* extract values from active_vlans and write back to VFTA */
5086 	for (i = VFTA_BLOCK_SIZE; i--;) {
5087 		vid = (vfta_offset + i) * 32;
5088 		word = vid / BITS_PER_LONG;
5089 		bits = vid % BITS_PER_LONG;
5090 
5091 		vfta[i] |= adapter->active_vlans[word] >> bits;
5092 
5093 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(vfta_offset + i), vfta[i]);
5094 	}
5095 }
5096 
5097 static void ixgbe_vlan_promisc_disable(struct ixgbe_adapter *adapter)
5098 {
5099 	struct ixgbe_hw *hw = &adapter->hw;
5100 	u32 vlnctrl, i;
5101 
5102 	/* Set VLAN filtering to enabled */
5103 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
5104 	vlnctrl |= IXGBE_VLNCTRL_VFE;
5105 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
5106 
5107 	if (!(adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) ||
5108 	    hw->mac.type == ixgbe_mac_82598EB)
5109 		return;
5110 
5111 	/* We are not in VLAN promisc, nothing to do */
5112 	if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
5113 		return;
5114 
5115 	/* Set flag so we don't redo unnecessary work */
5116 	adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
5117 
5118 	for (i = 0; i < hw->mac.vft_size; i += VFTA_BLOCK_SIZE)
5119 		ixgbe_scrub_vfta(adapter, i);
5120 }
5121 
5122 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
5123 {
5124 	u16 vid = 1;
5125 
5126 	ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
5127 
5128 	for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
5129 		ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
5130 }
5131 
5132 /**
5133  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
5134  * @netdev: network interface device structure
5135  *
5136  * Writes multicast address list to the MTA hash table.
5137  * Returns: -ENOMEM on failure
5138  *                0 on no addresses written
5139  *                X on writing X addresses to MTA
5140  **/
5141 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
5142 {
5143 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
5144 	struct ixgbe_hw *hw = &adapter->hw;
5145 
5146 	if (!netif_running(netdev))
5147 		return 0;
5148 
5149 	if (hw->mac.ops.update_mc_addr_list)
5150 		hw->mac.ops.update_mc_addr_list(hw, netdev);
5151 	else
5152 		return -ENOMEM;
5153 
5154 #ifdef CONFIG_PCI_IOV
5155 	ixgbe_restore_vf_multicasts(adapter);
5156 #endif
5157 
5158 	return netdev_mc_count(netdev);
5159 }
5160 
5161 #ifdef CONFIG_PCI_IOV
5162 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
5163 {
5164 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
5165 	struct ixgbe_hw *hw = &adapter->hw;
5166 	int i;
5167 
5168 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
5169 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
5170 
5171 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
5172 			hw->mac.ops.set_rar(hw, i,
5173 					    mac_table->addr,
5174 					    mac_table->pool,
5175 					    IXGBE_RAH_AV);
5176 		else
5177 			hw->mac.ops.clear_rar(hw, i);
5178 	}
5179 }
5180 
5181 #endif
5182 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
5183 {
5184 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
5185 	struct ixgbe_hw *hw = &adapter->hw;
5186 	int i;
5187 
5188 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
5189 		if (!(mac_table->state & IXGBE_MAC_STATE_MODIFIED))
5190 			continue;
5191 
5192 		mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
5193 
5194 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
5195 			hw->mac.ops.set_rar(hw, i,
5196 					    mac_table->addr,
5197 					    mac_table->pool,
5198 					    IXGBE_RAH_AV);
5199 		else
5200 			hw->mac.ops.clear_rar(hw, i);
5201 	}
5202 }
5203 
5204 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
5205 {
5206 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
5207 	struct ixgbe_hw *hw = &adapter->hw;
5208 	int i;
5209 
5210 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
5211 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
5212 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
5213 	}
5214 
5215 	ixgbe_sync_mac_table(adapter);
5216 }
5217 
5218 static int ixgbe_available_rars(struct ixgbe_adapter *adapter, u16 pool)
5219 {
5220 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
5221 	struct ixgbe_hw *hw = &adapter->hw;
5222 	int i, count = 0;
5223 
5224 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
5225 		/* do not count default RAR as available */
5226 		if (mac_table->state & IXGBE_MAC_STATE_DEFAULT)
5227 			continue;
5228 
5229 		/* only count unused and addresses that belong to us */
5230 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE) {
5231 			if (mac_table->pool != pool)
5232 				continue;
5233 		}
5234 
5235 		count++;
5236 	}
5237 
5238 	return count;
5239 }
5240 
5241 /* this function destroys the first RAR entry */
5242 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter)
5243 {
5244 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
5245 	struct ixgbe_hw *hw = &adapter->hw;
5246 
5247 	memcpy(&mac_table->addr, hw->mac.addr, ETH_ALEN);
5248 	mac_table->pool = VMDQ_P(0);
5249 
5250 	mac_table->state = IXGBE_MAC_STATE_DEFAULT | IXGBE_MAC_STATE_IN_USE;
5251 
5252 	hw->mac.ops.set_rar(hw, 0, mac_table->addr, mac_table->pool,
5253 			    IXGBE_RAH_AV);
5254 }
5255 
5256 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter,
5257 			 const u8 *addr, u16 pool)
5258 {
5259 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
5260 	struct ixgbe_hw *hw = &adapter->hw;
5261 	int i;
5262 
5263 	if (is_zero_ether_addr(addr))
5264 		return -EINVAL;
5265 
5266 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
5267 		if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
5268 			continue;
5269 
5270 		ether_addr_copy(mac_table->addr, addr);
5271 		mac_table->pool = pool;
5272 
5273 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED |
5274 				    IXGBE_MAC_STATE_IN_USE;
5275 
5276 		ixgbe_sync_mac_table(adapter);
5277 
5278 		return i;
5279 	}
5280 
5281 	return -ENOMEM;
5282 }
5283 
5284 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter,
5285 			 const u8 *addr, u16 pool)
5286 {
5287 	struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
5288 	struct ixgbe_hw *hw = &adapter->hw;
5289 	int i;
5290 
5291 	if (is_zero_ether_addr(addr))
5292 		return -EINVAL;
5293 
5294 	/* search table for addr, if found clear IN_USE flag and sync */
5295 	for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
5296 		/* we can only delete an entry if it is in use */
5297 		if (!(mac_table->state & IXGBE_MAC_STATE_IN_USE))
5298 			continue;
5299 		/* we only care about entries that belong to the given pool */
5300 		if (mac_table->pool != pool)
5301 			continue;
5302 		/* we only care about a specific MAC address */
5303 		if (!ether_addr_equal(addr, mac_table->addr))
5304 			continue;
5305 
5306 		mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
5307 		mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
5308 
5309 		ixgbe_sync_mac_table(adapter);
5310 
5311 		return 0;
5312 	}
5313 
5314 	return -ENOMEM;
5315 }
5316 
5317 static int ixgbe_uc_sync(struct net_device *netdev, const unsigned char *addr)
5318 {
5319 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
5320 	int ret;
5321 
5322 	ret = ixgbe_add_mac_filter(adapter, addr, VMDQ_P(0));
5323 
5324 	return min_t(int, ret, 0);
5325 }
5326 
5327 static int ixgbe_uc_unsync(struct net_device *netdev, const unsigned char *addr)
5328 {
5329 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
5330 
5331 	ixgbe_del_mac_filter(adapter, addr, VMDQ_P(0));
5332 
5333 	return 0;
5334 }
5335 
5336 /**
5337  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
5338  * @netdev: network interface device structure
5339  *
5340  * The set_rx_method entry point is called whenever the unicast/multicast
5341  * address list or the network interface flags are updated.  This routine is
5342  * responsible for configuring the hardware for proper unicast, multicast and
5343  * promiscuous mode.
5344  **/
5345 void ixgbe_set_rx_mode(struct net_device *netdev)
5346 {
5347 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
5348 	struct ixgbe_hw *hw = &adapter->hw;
5349 	u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
5350 	netdev_features_t features = netdev->features;
5351 	int count;
5352 
5353 	/* Check for Promiscuous and All Multicast modes */
5354 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
5355 
5356 	/* set all bits that we expect to always be set */
5357 	fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
5358 	fctrl |= IXGBE_FCTRL_BAM;
5359 	fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
5360 	fctrl |= IXGBE_FCTRL_PMCF;
5361 
5362 	/* clear the bits we are changing the status of */
5363 	fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
5364 	if (netdev->flags & IFF_PROMISC) {
5365 		hw->addr_ctrl.user_set_promisc = true;
5366 		fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
5367 		vmolr |= IXGBE_VMOLR_MPE;
5368 		features &= ~NETIF_F_HW_VLAN_CTAG_FILTER;
5369 	} else {
5370 		if (netdev->flags & IFF_ALLMULTI) {
5371 			fctrl |= IXGBE_FCTRL_MPE;
5372 			vmolr |= IXGBE_VMOLR_MPE;
5373 		}
5374 		hw->addr_ctrl.user_set_promisc = false;
5375 	}
5376 
5377 	/*
5378 	 * Write addresses to available RAR registers, if there is not
5379 	 * sufficient space to store all the addresses then enable
5380 	 * unicast promiscuous mode
5381 	 */
5382 	if (__dev_uc_sync(netdev, ixgbe_uc_sync, ixgbe_uc_unsync)) {
5383 		fctrl |= IXGBE_FCTRL_UPE;
5384 		vmolr |= IXGBE_VMOLR_ROPE;
5385 	}
5386 
5387 	/* Write addresses to the MTA, if the attempt fails
5388 	 * then we should just turn on promiscuous mode so
5389 	 * that we can at least receive multicast traffic
5390 	 */
5391 	count = ixgbe_write_mc_addr_list(netdev);
5392 	if (count < 0) {
5393 		fctrl |= IXGBE_FCTRL_MPE;
5394 		vmolr |= IXGBE_VMOLR_MPE;
5395 	} else if (count) {
5396 		vmolr |= IXGBE_VMOLR_ROMPE;
5397 	}
5398 
5399 	if (hw->mac.type != ixgbe_mac_82598EB) {
5400 		vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
5401 			 ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
5402 			   IXGBE_VMOLR_ROPE);
5403 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
5404 	}
5405 
5406 	/* This is useful for sniffing bad packets. */
5407 	if (features & NETIF_F_RXALL) {
5408 		/* UPE and MPE will be handled by normal PROMISC logic
5409 		 * in e1000e_set_rx_mode */
5410 		fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
5411 			  IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
5412 			  IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
5413 
5414 		fctrl &= ~(IXGBE_FCTRL_DPF);
5415 		/* NOTE:  VLAN filtering is disabled by setting PROMISC */
5416 	}
5417 
5418 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
5419 
5420 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
5421 		ixgbe_vlan_strip_enable(adapter);
5422 	else
5423 		ixgbe_vlan_strip_disable(adapter);
5424 
5425 	if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
5426 		ixgbe_vlan_promisc_disable(adapter);
5427 	else
5428 		ixgbe_vlan_promisc_enable(adapter);
5429 }
5430 
5431 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
5432 {
5433 	int q_idx;
5434 
5435 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
5436 		napi_enable(&adapter->q_vector[q_idx]->napi);
5437 }
5438 
5439 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
5440 {
5441 	int q_idx;
5442 
5443 	for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
5444 		napi_disable(&adapter->q_vector[q_idx]->napi);
5445 }
5446 
5447 static int ixgbe_udp_tunnel_sync(struct net_device *dev, unsigned int table)
5448 {
5449 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(dev);
5450 	struct ixgbe_hw *hw = &adapter->hw;
5451 	struct udp_tunnel_info ti;
5452 
5453 	udp_tunnel_nic_get_port(dev, table, 0, &ti);
5454 	if (ti.type == UDP_TUNNEL_TYPE_VXLAN)
5455 		adapter->vxlan_port = ti.port;
5456 	else
5457 		adapter->geneve_port = ti.port;
5458 
5459 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL,
5460 			ntohs(adapter->vxlan_port) |
5461 			ntohs(adapter->geneve_port) <<
5462 				IXGBE_VXLANCTRL_GENEVE_UDPPORT_SHIFT);
5463 	return 0;
5464 }
5465 
5466 static const struct udp_tunnel_nic_info ixgbe_udp_tunnels_x550 = {
5467 	.sync_table	= ixgbe_udp_tunnel_sync,
5468 	.flags		= UDP_TUNNEL_NIC_INFO_IPV4_ONLY,
5469 	.tables		= {
5470 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN,  },
5471 	},
5472 };
5473 
5474 static const struct udp_tunnel_nic_info ixgbe_udp_tunnels_x550em_a = {
5475 	.sync_table	= ixgbe_udp_tunnel_sync,
5476 	.flags		= UDP_TUNNEL_NIC_INFO_IPV4_ONLY,
5477 	.tables		= {
5478 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN,  },
5479 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_GENEVE, },
5480 	},
5481 };
5482 
5483 #ifdef CONFIG_IXGBE_DCB
5484 /**
5485  * ixgbe_configure_dcb - Configure DCB hardware
5486  * @adapter: ixgbe adapter struct
5487  *
5488  * This is called by the driver on open to configure the DCB hardware.
5489  * This is also called by the gennetlink interface when reconfiguring
5490  * the DCB state.
5491  */
5492 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
5493 {
5494 	struct ixgbe_hw *hw = &adapter->hw;
5495 	int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
5496 
5497 	if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
5498 		if (hw->mac.type == ixgbe_mac_82598EB)
5499 			netif_set_tso_max_size(adapter->netdev, 65536);
5500 		return;
5501 	}
5502 
5503 	if (hw->mac.type == ixgbe_mac_82598EB)
5504 		netif_set_tso_max_size(adapter->netdev, 32768);
5505 
5506 #ifdef IXGBE_FCOE
5507 	if (adapter->netdev->fcoe_mtu)
5508 		max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
5509 #endif
5510 
5511 	/* reconfigure the hardware */
5512 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
5513 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5514 						DCB_TX_CONFIG);
5515 		ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5516 						DCB_RX_CONFIG);
5517 		ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
5518 	} else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
5519 		ixgbe_dcb_hw_ets(&adapter->hw,
5520 				 adapter->ixgbe_ieee_ets,
5521 				 max_frame);
5522 		ixgbe_dcb_hw_pfc_config(&adapter->hw,
5523 					adapter->ixgbe_ieee_pfc->pfc_en,
5524 					adapter->ixgbe_ieee_ets->prio_tc);
5525 	}
5526 
5527 	/* Enable RSS Hash per TC */
5528 	if (hw->mac.type != ixgbe_mac_82598EB) {
5529 		u32 msb = 0;
5530 		u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
5531 
5532 		while (rss_i) {
5533 			msb++;
5534 			rss_i >>= 1;
5535 		}
5536 
5537 		/* write msb to all 8 TCs in one write */
5538 		IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
5539 	}
5540 }
5541 #endif
5542 
5543 /* Additional bittime to account for IXGBE framing */
5544 #define IXGBE_ETH_FRAMING 20
5545 
5546 /**
5547  * ixgbe_hpbthresh - calculate high water mark for flow control
5548  *
5549  * @adapter: board private structure to calculate for
5550  * @pb: packet buffer to calculate
5551  */
5552 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
5553 {
5554 	struct ixgbe_hw *hw = &adapter->hw;
5555 	struct net_device *dev = adapter->netdev;
5556 	int link, tc, kb, marker;
5557 	u32 dv_id, rx_pba;
5558 
5559 	/* Calculate max LAN frame size */
5560 	tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
5561 
5562 #ifdef IXGBE_FCOE
5563 	/* FCoE traffic class uses FCOE jumbo frames */
5564 	if (dev->fcoe_mtu && tc < IXGBE_FCOE_JUMBO_FRAME_SIZE &&
5565 	    (pb == ixgbe_fcoe_get_tc(adapter)))
5566 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5567 #endif
5568 
5569 	/* Calculate delay value for device */
5570 	switch (hw->mac.type) {
5571 	case ixgbe_mac_X540:
5572 	case ixgbe_mac_X550:
5573 	case ixgbe_mac_X550EM_x:
5574 	case ixgbe_mac_x550em_a:
5575 	case ixgbe_mac_e610:
5576 		dv_id = IXGBE_DV_X540(link, tc);
5577 		break;
5578 	default:
5579 		dv_id = IXGBE_DV(link, tc);
5580 		break;
5581 	}
5582 
5583 	/* Loopback switch introduces additional latency */
5584 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5585 		dv_id += IXGBE_B2BT(tc);
5586 
5587 	/* Delay value is calculated in bit times convert to KB */
5588 	kb = IXGBE_BT2KB(dv_id);
5589 	rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
5590 
5591 	marker = rx_pba - kb;
5592 
5593 	/* It is possible that the packet buffer is not large enough
5594 	 * to provide required headroom. In this case throw an error
5595 	 * to user and a do the best we can.
5596 	 */
5597 	if (marker < 0) {
5598 		e_warn(drv, "Packet Buffer(%i) can not provide enough"
5599 			    "headroom to support flow control."
5600 			    "Decrease MTU or number of traffic classes\n", pb);
5601 		marker = tc + 1;
5602 	}
5603 
5604 	return marker;
5605 }
5606 
5607 /**
5608  * ixgbe_lpbthresh - calculate low water mark for flow control
5609  *
5610  * @adapter: board private structure to calculate for
5611  * @pb: packet buffer to calculate
5612  */
5613 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
5614 {
5615 	struct ixgbe_hw *hw = &adapter->hw;
5616 	struct net_device *dev = adapter->netdev;
5617 	int tc;
5618 	u32 dv_id;
5619 
5620 	/* Calculate max LAN frame size */
5621 	tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
5622 
5623 #ifdef IXGBE_FCOE
5624 	/* FCoE traffic class uses FCOE jumbo frames */
5625 	if (dev->fcoe_mtu && tc < IXGBE_FCOE_JUMBO_FRAME_SIZE &&
5626 	    (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
5627 		tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5628 #endif
5629 
5630 	/* Calculate delay value for device */
5631 	switch (hw->mac.type) {
5632 	case ixgbe_mac_X540:
5633 	case ixgbe_mac_X550:
5634 	case ixgbe_mac_X550EM_x:
5635 	case ixgbe_mac_x550em_a:
5636 	case ixgbe_mac_e610:
5637 		dv_id = IXGBE_LOW_DV_X540(tc);
5638 		break;
5639 	default:
5640 		dv_id = IXGBE_LOW_DV(tc);
5641 		break;
5642 	}
5643 
5644 	/* Delay value is calculated in bit times convert to KB */
5645 	return IXGBE_BT2KB(dv_id);
5646 }
5647 
5648 /*
5649  * ixgbe_pbthresh_setup - calculate and setup high low water marks
5650  */
5651 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
5652 {
5653 	struct ixgbe_hw *hw = &adapter->hw;
5654 	int num_tc = adapter->hw_tcs;
5655 	int i;
5656 
5657 	if (!num_tc)
5658 		num_tc = 1;
5659 
5660 	for (i = 0; i < num_tc; i++) {
5661 		hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
5662 		hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
5663 
5664 		/* Low water marks must not be larger than high water marks */
5665 		if (hw->fc.low_water[i] > hw->fc.high_water[i])
5666 			hw->fc.low_water[i] = 0;
5667 	}
5668 
5669 	for (; i < MAX_TRAFFIC_CLASS; i++)
5670 		hw->fc.high_water[i] = 0;
5671 }
5672 
5673 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
5674 {
5675 	struct ixgbe_hw *hw = &adapter->hw;
5676 	int hdrm;
5677 	u8 tc = adapter->hw_tcs;
5678 
5679 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
5680 	    adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
5681 		hdrm = 32 << adapter->fdir_pballoc;
5682 	else
5683 		hdrm = 0;
5684 
5685 	hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
5686 	ixgbe_pbthresh_setup(adapter);
5687 }
5688 
5689 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
5690 {
5691 	struct ixgbe_hw *hw = &adapter->hw;
5692 	struct hlist_node *node2;
5693 	struct ixgbe_fdir_filter *filter;
5694 	u8 queue;
5695 
5696 	spin_lock(&adapter->fdir_perfect_lock);
5697 
5698 	if (!hlist_empty(&adapter->fdir_filter_list))
5699 		ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
5700 
5701 	hlist_for_each_entry_safe(filter, node2,
5702 				  &adapter->fdir_filter_list, fdir_node) {
5703 		if (filter->action == IXGBE_FDIR_DROP_QUEUE) {
5704 			queue = IXGBE_FDIR_DROP_QUEUE;
5705 		} else {
5706 			u32 ring = ethtool_get_flow_spec_ring(filter->action);
5707 			u8 vf = ethtool_get_flow_spec_ring_vf(filter->action);
5708 
5709 			if (!vf && (ring >= adapter->num_rx_queues)) {
5710 				e_err(drv, "FDIR restore failed without VF, ring: %u\n",
5711 				      ring);
5712 				continue;
5713 			} else if (vf &&
5714 				   ((vf > adapter->num_vfs) ||
5715 				     ring >= adapter->num_rx_queues_per_pool)) {
5716 				e_err(drv, "FDIR restore failed with VF, vf: %hhu, ring: %u\n",
5717 				      vf, ring);
5718 				continue;
5719 			}
5720 
5721 			/* Map the ring onto the absolute queue index */
5722 			if (!vf)
5723 				queue = adapter->rx_ring[ring]->reg_idx;
5724 			else
5725 				queue = ((vf - 1) *
5726 					adapter->num_rx_queues_per_pool) + ring;
5727 		}
5728 
5729 		ixgbe_fdir_write_perfect_filter_82599(hw,
5730 				&filter->filter, filter->sw_idx, queue);
5731 	}
5732 
5733 	spin_unlock(&adapter->fdir_perfect_lock);
5734 }
5735 
5736 /**
5737  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
5738  * @rx_ring: ring to free buffers from
5739  **/
5740 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
5741 {
5742 	u16 i = rx_ring->next_to_clean;
5743 	struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
5744 
5745 	if (rx_ring->xsk_pool) {
5746 		ixgbe_xsk_clean_rx_ring(rx_ring);
5747 		goto skip_free;
5748 	}
5749 
5750 	/* Free all the Rx ring sk_buffs */
5751 	while (i != rx_ring->next_to_alloc) {
5752 		if (rx_buffer->skb) {
5753 			struct sk_buff *skb = rx_buffer->skb;
5754 			if (IXGBE_CB(skb)->page_released)
5755 				dma_unmap_page_attrs(rx_ring->dev,
5756 						     IXGBE_CB(skb)->dma,
5757 						     ixgbe_rx_pg_size(rx_ring),
5758 						     DMA_FROM_DEVICE,
5759 						     IXGBE_RX_DMA_ATTR);
5760 			dev_kfree_skb(skb);
5761 		}
5762 
5763 		/* Invalidate cache lines that may have been written to by
5764 		 * device so that we avoid corrupting memory.
5765 		 */
5766 		dma_sync_single_range_for_cpu(rx_ring->dev,
5767 					      rx_buffer->dma,
5768 					      rx_buffer->page_offset,
5769 					      ixgbe_rx_bufsz(rx_ring),
5770 					      DMA_FROM_DEVICE);
5771 
5772 		/* free resources associated with mapping */
5773 		dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
5774 				     ixgbe_rx_pg_size(rx_ring),
5775 				     DMA_FROM_DEVICE,
5776 				     IXGBE_RX_DMA_ATTR);
5777 		__page_frag_cache_drain(rx_buffer->page,
5778 					rx_buffer->pagecnt_bias);
5779 
5780 		i++;
5781 		rx_buffer++;
5782 		if (i == rx_ring->count) {
5783 			i = 0;
5784 			rx_buffer = rx_ring->rx_buffer_info;
5785 		}
5786 	}
5787 
5788 skip_free:
5789 	rx_ring->next_to_alloc = 0;
5790 	rx_ring->next_to_clean = 0;
5791 	rx_ring->next_to_use = 0;
5792 }
5793 
5794 static int ixgbe_fwd_ring_up(struct ixgbe_adapter *adapter,
5795 			     struct ixgbe_fwd_adapter *accel)
5796 {
5797 	u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
5798 	int num_tc = netdev_get_num_tc(adapter->netdev);
5799 	struct net_device *vdev = accel->netdev;
5800 	int i, baseq, err;
5801 
5802 	baseq = accel->pool * adapter->num_rx_queues_per_pool;
5803 	netdev_dbg(vdev, "pool %i:%i queues %i:%i\n",
5804 		   accel->pool, adapter->num_rx_pools,
5805 		   baseq, baseq + adapter->num_rx_queues_per_pool);
5806 
5807 	accel->rx_base_queue = baseq;
5808 	accel->tx_base_queue = baseq;
5809 
5810 	/* record configuration for macvlan interface in vdev */
5811 	for (i = 0; i < num_tc; i++)
5812 		netdev_bind_sb_channel_queue(adapter->netdev, vdev,
5813 					     i, rss_i, baseq + (rss_i * i));
5814 
5815 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5816 		adapter->rx_ring[baseq + i]->netdev = vdev;
5817 
5818 	/* Guarantee all rings are updated before we update the
5819 	 * MAC address filter.
5820 	 */
5821 	wmb();
5822 
5823 	/* ixgbe_add_mac_filter will return an index if it succeeds, so we
5824 	 * need to only treat it as an error value if it is negative.
5825 	 */
5826 	err = ixgbe_add_mac_filter(adapter, vdev->dev_addr,
5827 				   VMDQ_P(accel->pool));
5828 	if (err >= 0)
5829 		return 0;
5830 
5831 	/* if we cannot add the MAC rule then disable the offload */
5832 	macvlan_release_l2fw_offload(vdev);
5833 
5834 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5835 		adapter->rx_ring[baseq + i]->netdev = NULL;
5836 
5837 	netdev_err(vdev, "L2FW offload disabled due to L2 filter error\n");
5838 
5839 	/* unbind the queues and drop the subordinate channel config */
5840 	netdev_unbind_sb_channel(adapter->netdev, vdev);
5841 	netdev_set_sb_channel(vdev, 0);
5842 
5843 	clear_bit(accel->pool, adapter->fwd_bitmask);
5844 	kfree(accel);
5845 
5846 	return err;
5847 }
5848 
5849 static int ixgbe_macvlan_up(struct net_device *vdev,
5850 			    struct netdev_nested_priv *priv)
5851 {
5852 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)priv->data;
5853 	struct ixgbe_fwd_adapter *accel;
5854 
5855 	if (!netif_is_macvlan(vdev))
5856 		return 0;
5857 
5858 	accel = macvlan_accel_priv(vdev);
5859 	if (!accel)
5860 		return 0;
5861 
5862 	ixgbe_fwd_ring_up(adapter, accel);
5863 
5864 	return 0;
5865 }
5866 
5867 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
5868 {
5869 	struct netdev_nested_priv priv = {
5870 		.data = (void *)adapter,
5871 	};
5872 
5873 	netdev_walk_all_upper_dev_rcu(adapter->netdev,
5874 				      ixgbe_macvlan_up, &priv);
5875 }
5876 
5877 static void ixgbe_configure(struct ixgbe_adapter *adapter)
5878 {
5879 	struct ixgbe_hw *hw = &adapter->hw;
5880 
5881 	ixgbe_configure_pb(adapter);
5882 #ifdef CONFIG_IXGBE_DCB
5883 	ixgbe_configure_dcb(adapter);
5884 #endif
5885 	/*
5886 	 * We must restore virtualization before VLANs or else
5887 	 * the VLVF registers will not be populated
5888 	 */
5889 	ixgbe_configure_virtualization(adapter);
5890 
5891 	ixgbe_set_rx_mode(adapter->netdev);
5892 	ixgbe_restore_vlan(adapter);
5893 	ixgbe_ipsec_restore(adapter);
5894 
5895 	switch (hw->mac.type) {
5896 	case ixgbe_mac_82599EB:
5897 	case ixgbe_mac_X540:
5898 		hw->mac.ops.disable_rx_buff(hw);
5899 		break;
5900 	default:
5901 		break;
5902 	}
5903 
5904 	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
5905 		ixgbe_init_fdir_signature_82599(&adapter->hw,
5906 						adapter->fdir_pballoc);
5907 	} else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
5908 		ixgbe_init_fdir_perfect_82599(&adapter->hw,
5909 					      adapter->fdir_pballoc);
5910 		ixgbe_fdir_filter_restore(adapter);
5911 	}
5912 
5913 	switch (hw->mac.type) {
5914 	case ixgbe_mac_82599EB:
5915 	case ixgbe_mac_X540:
5916 		hw->mac.ops.enable_rx_buff(hw);
5917 		break;
5918 	default:
5919 		break;
5920 	}
5921 
5922 #ifdef CONFIG_IXGBE_DCA
5923 	/* configure DCA */
5924 	if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
5925 		ixgbe_setup_dca(adapter);
5926 #endif /* CONFIG_IXGBE_DCA */
5927 
5928 #ifdef IXGBE_FCOE
5929 	/* configure FCoE L2 filters, redirection table, and Rx control */
5930 	ixgbe_configure_fcoe(adapter);
5931 
5932 #endif /* IXGBE_FCOE */
5933 	ixgbe_configure_tx(adapter);
5934 	ixgbe_configure_rx(adapter);
5935 	ixgbe_configure_dfwd(adapter);
5936 }
5937 
5938 /**
5939  * ixgbe_enable_link_status_events - enable link status events
5940  * @adapter: pointer to the adapter structure
5941  * @mask: event mask to be set
5942  *
5943  * Enables link status events by invoking ixgbe_configure_lse()
5944  *
5945  * Return: the exit code of the operation.
5946  */
5947 static int ixgbe_enable_link_status_events(struct ixgbe_adapter *adapter,
5948 					   u16 mask)
5949 {
5950 	int err;
5951 
5952 	err = ixgbe_configure_lse(&adapter->hw, true, mask);
5953 	if (err)
5954 		return err;
5955 
5956 	adapter->lse_mask = mask;
5957 	return 0;
5958 }
5959 
5960 /**
5961  * ixgbe_disable_link_status_events - disable link status events
5962  * @adapter: pointer to the adapter structure
5963  *
5964  * Disables link status events by invoking ixgbe_configure_lse()
5965  *
5966  * Return: the exit code of the operation.
5967  */
5968 static int ixgbe_disable_link_status_events(struct ixgbe_adapter *adapter)
5969 {
5970 	int err;
5971 
5972 	err = ixgbe_configure_lse(&adapter->hw, false, adapter->lse_mask);
5973 	if (err)
5974 		return err;
5975 
5976 	adapter->lse_mask = 0;
5977 	return 0;
5978 }
5979 
5980 /**
5981  * ixgbe_sfp_link_config - set up SFP+ link
5982  * @adapter: pointer to private adapter struct
5983  **/
5984 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
5985 {
5986 	/*
5987 	 * We are assuming the worst case scenario here, and that
5988 	 * is that an SFP was inserted/removed after the reset
5989 	 * but before SFP detection was enabled.  As such the best
5990 	 * solution is to just start searching as soon as we start
5991 	 */
5992 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5993 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5994 
5995 	adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5996 	adapter->sfp_poll_time = 0;
5997 }
5998 
5999 /**
6000  * ixgbe_non_sfp_link_config - set up non-SFP+ link
6001  * @hw: pointer to private hardware struct
6002  *
6003  * Configure non-SFP link.
6004  *
6005  * Return: 0 on success, negative on failure
6006  **/
6007 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
6008 {
6009 	struct ixgbe_adapter *adapter = container_of(hw, struct ixgbe_adapter,
6010 						     hw);
6011 	u16 mask = ~((u16)(IXGBE_ACI_LINK_EVENT_UPDOWN |
6012 			   IXGBE_ACI_LINK_EVENT_MEDIA_NA |
6013 			   IXGBE_ACI_LINK_EVENT_MODULE_QUAL_FAIL |
6014 			   IXGBE_ACI_LINK_EVENT_PHY_FW_LOAD_FAIL));
6015 	bool autoneg, link_up = false;
6016 	int ret = -EIO;
6017 	u32 speed;
6018 
6019 	if (hw->mac.ops.check_link)
6020 		ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
6021 
6022 	if (ret)
6023 		return ret;
6024 
6025 	speed = hw->phy.autoneg_advertised;
6026 	if (!speed && hw->mac.ops.get_link_capabilities) {
6027 		ret = hw->mac.ops.get_link_capabilities(hw, &speed,
6028 							&autoneg);
6029 		/* remove NBASE-T speeds from default autonegotiation
6030 		 * to accommodate broken network switches in the field
6031 		 * which cannot cope with advertised NBASE-T speeds
6032 		 */
6033 		speed &= ~(IXGBE_LINK_SPEED_5GB_FULL |
6034 			   IXGBE_LINK_SPEED_2_5GB_FULL);
6035 	}
6036 
6037 	if (ret)
6038 		return ret;
6039 
6040 	if (hw->mac.ops.setup_link) {
6041 		if (adapter->hw.mac.type == ixgbe_mac_e610) {
6042 			ret = ixgbe_enable_link_status_events(adapter, mask);
6043 			if (ret)
6044 				return ret;
6045 		}
6046 		ret = hw->mac.ops.setup_link(hw, speed, link_up);
6047 	}
6048 
6049 	return ret;
6050 }
6051 
6052 /**
6053  * ixgbe_check_media_subtask - check for media
6054  * @adapter: pointer to adapter structure
6055  *
6056  * If media is available then initialize PHY user configuration. Configure the
6057  * PHY if the interface is up.
6058  */
6059 static void ixgbe_check_media_subtask(struct ixgbe_adapter *adapter)
6060 {
6061 	struct ixgbe_hw *hw = &adapter->hw;
6062 
6063 	/* No need to check for media if it's already present */
6064 	if (!(adapter->flags2 & IXGBE_FLAG2_NO_MEDIA))
6065 		return;
6066 
6067 	/* Refresh link info and check if media is present */
6068 	if (ixgbe_update_link_info(hw))
6069 		return;
6070 
6071 	ixgbe_check_link_cfg_err(adapter, hw->link.link_info.link_cfg_err);
6072 
6073 	if (hw->link.link_info.link_info & IXGBE_ACI_MEDIA_AVAILABLE) {
6074 		/* PHY settings are reset on media insertion, reconfigure
6075 		 * PHY to preserve settings.
6076 		 */
6077 		if (!(ixgbe_non_sfp_link_config(&adapter->hw)))
6078 			adapter->flags2 &= ~IXGBE_FLAG2_NO_MEDIA;
6079 
6080 		/* A Link Status Event will be generated; the event handler
6081 		 * will complete bringing the interface up
6082 		 */
6083 	}
6084 }
6085 
6086 /**
6087  * ixgbe_clear_vf_stats_counters - Clear out VF stats after reset
6088  * @adapter: board private structure
6089  *
6090  * On a reset we need to clear out the VF stats or accounting gets
6091  * messed up because they're not clear on read.
6092  **/
6093 static void ixgbe_clear_vf_stats_counters(struct ixgbe_adapter *adapter)
6094 {
6095 	struct ixgbe_hw *hw = &adapter->hw;
6096 	int i;
6097 
6098 	for (i = 0; i < adapter->num_vfs; i++) {
6099 		adapter->vfinfo[i].last_vfstats.gprc =
6100 			IXGBE_READ_REG(hw, IXGBE_PVFGPRC(i));
6101 		adapter->vfinfo[i].saved_rst_vfstats.gprc +=
6102 			adapter->vfinfo[i].vfstats.gprc;
6103 		adapter->vfinfo[i].vfstats.gprc = 0;
6104 		adapter->vfinfo[i].last_vfstats.gptc =
6105 			IXGBE_READ_REG(hw, IXGBE_PVFGPTC(i));
6106 		adapter->vfinfo[i].saved_rst_vfstats.gptc +=
6107 			adapter->vfinfo[i].vfstats.gptc;
6108 		adapter->vfinfo[i].vfstats.gptc = 0;
6109 		adapter->vfinfo[i].last_vfstats.gorc =
6110 			IXGBE_READ_REG(hw, IXGBE_PVFGORC_LSB(i));
6111 		adapter->vfinfo[i].saved_rst_vfstats.gorc +=
6112 			adapter->vfinfo[i].vfstats.gorc;
6113 		adapter->vfinfo[i].vfstats.gorc = 0;
6114 		adapter->vfinfo[i].last_vfstats.gotc =
6115 			IXGBE_READ_REG(hw, IXGBE_PVFGOTC_LSB(i));
6116 		adapter->vfinfo[i].saved_rst_vfstats.gotc +=
6117 			adapter->vfinfo[i].vfstats.gotc;
6118 		adapter->vfinfo[i].vfstats.gotc = 0;
6119 		adapter->vfinfo[i].last_vfstats.mprc =
6120 			IXGBE_READ_REG(hw, IXGBE_PVFMPRC(i));
6121 		adapter->vfinfo[i].saved_rst_vfstats.mprc +=
6122 			adapter->vfinfo[i].vfstats.mprc;
6123 		adapter->vfinfo[i].vfstats.mprc = 0;
6124 	}
6125 }
6126 
6127 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
6128 {
6129 	struct ixgbe_hw *hw = &adapter->hw;
6130 	u32 gpie = 0;
6131 
6132 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
6133 		gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
6134 		       IXGBE_GPIE_OCD;
6135 		gpie |= IXGBE_GPIE_EIAME;
6136 		/*
6137 		 * use EIAM to auto-mask when MSI-X interrupt is asserted
6138 		 * this saves a register write for every interrupt
6139 		 */
6140 		switch (hw->mac.type) {
6141 		case ixgbe_mac_82598EB:
6142 			IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
6143 			break;
6144 		case ixgbe_mac_82599EB:
6145 		case ixgbe_mac_X540:
6146 		case ixgbe_mac_X550:
6147 		case ixgbe_mac_X550EM_x:
6148 		case ixgbe_mac_x550em_a:
6149 		case ixgbe_mac_e610:
6150 		default:
6151 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
6152 			IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
6153 			break;
6154 		}
6155 	} else {
6156 		/* legacy interrupts, use EIAM to auto-mask when reading EICR,
6157 		 * specifically only auto mask tx and rx interrupts */
6158 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
6159 	}
6160 
6161 	/* XXX: to interrupt immediately for EICS writes, enable this */
6162 	/* gpie |= IXGBE_GPIE_EIMEN; */
6163 
6164 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
6165 		gpie &= ~IXGBE_GPIE_VTMODE_MASK;
6166 
6167 		switch (adapter->ring_feature[RING_F_VMDQ].mask) {
6168 		case IXGBE_82599_VMDQ_8Q_MASK:
6169 			gpie |= IXGBE_GPIE_VTMODE_16;
6170 			break;
6171 		case IXGBE_82599_VMDQ_4Q_MASK:
6172 			gpie |= IXGBE_GPIE_VTMODE_32;
6173 			break;
6174 		default:
6175 			gpie |= IXGBE_GPIE_VTMODE_64;
6176 			break;
6177 		}
6178 	}
6179 
6180 	/* Enable Thermal over heat sensor interrupt */
6181 	if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
6182 		switch (adapter->hw.mac.type) {
6183 		case ixgbe_mac_82599EB:
6184 			gpie |= IXGBE_SDP0_GPIEN_8259X;
6185 			break;
6186 		default:
6187 			break;
6188 		}
6189 	}
6190 
6191 	/* Enable fan failure interrupt */
6192 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
6193 		gpie |= IXGBE_SDP1_GPIEN(hw);
6194 
6195 	switch (hw->mac.type) {
6196 	case ixgbe_mac_82599EB:
6197 		gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
6198 		break;
6199 	case ixgbe_mac_X550EM_x:
6200 	case ixgbe_mac_x550em_a:
6201 		gpie |= IXGBE_SDP0_GPIEN_X540;
6202 		break;
6203 	default:
6204 		break;
6205 	}
6206 
6207 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
6208 }
6209 
6210 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
6211 {
6212 	struct ixgbe_hw *hw = &adapter->hw;
6213 	int err;
6214 	u32 ctrl_ext;
6215 
6216 	ixgbe_get_hw_control(adapter);
6217 	ixgbe_setup_gpie(adapter);
6218 
6219 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
6220 		ixgbe_configure_msix(adapter);
6221 	else
6222 		ixgbe_configure_msi_and_legacy(adapter);
6223 
6224 	/* enable the optics for 82599 SFP+ fiber */
6225 	if (hw->mac.ops.enable_tx_laser)
6226 		hw->mac.ops.enable_tx_laser(hw);
6227 
6228 	if (hw->phy.ops.set_phy_power)
6229 		hw->phy.ops.set_phy_power(hw, true);
6230 
6231 	smp_mb__before_atomic();
6232 	clear_bit(__IXGBE_DOWN, &adapter->state);
6233 	ixgbe_napi_enable_all(adapter);
6234 
6235 	if (ixgbe_is_sfp(hw)) {
6236 		ixgbe_sfp_link_config(adapter);
6237 	} else {
6238 		err = ixgbe_non_sfp_link_config(hw);
6239 		if (err)
6240 			e_err(probe, "link_config FAILED %d\n", err);
6241 	}
6242 
6243 	/* clear any pending interrupts, may auto mask */
6244 	IXGBE_READ_REG(hw, IXGBE_EICR);
6245 	ixgbe_irq_enable(adapter, true, true);
6246 
6247 	/*
6248 	 * If this adapter has a fan, check to see if we had a failure
6249 	 * before we enabled the interrupt.
6250 	 */
6251 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
6252 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
6253 		if (esdp & IXGBE_ESDP_SDP1)
6254 			e_crit(drv, "Fan has stopped, replace the adapter\n");
6255 	}
6256 
6257 	/* bring the link up in the watchdog, this could race with our first
6258 	 * link up interrupt but shouldn't be a problem */
6259 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
6260 	adapter->link_check_timeout = jiffies;
6261 	mod_timer(&adapter->service_timer, jiffies);
6262 
6263 	ixgbe_clear_vf_stats_counters(adapter);
6264 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
6265 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
6266 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
6267 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
6268 
6269 	/* update setting rx tx for all active vfs */
6270 	ixgbe_set_all_vfs(adapter);
6271 }
6272 
6273 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
6274 {
6275 	/* put off any impending NetWatchDogTimeout */
6276 	netif_trans_update(adapter->netdev);
6277 
6278 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
6279 		usleep_range(1000, 2000);
6280 	if (adapter->hw.phy.type == ixgbe_phy_fw)
6281 		ixgbe_watchdog_link_is_down(adapter);
6282 	ixgbe_down(adapter);
6283 	/*
6284 	 * If SR-IOV enabled then wait a bit before bringing the adapter
6285 	 * back up to give the VFs time to respond to the reset.  The
6286 	 * two second wait is based upon the watchdog timer cycle in
6287 	 * the VF driver.
6288 	 */
6289 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
6290 		msleep(2000);
6291 	ixgbe_up(adapter);
6292 	clear_bit(__IXGBE_RESETTING, &adapter->state);
6293 }
6294 
6295 void ixgbe_up(struct ixgbe_adapter *adapter)
6296 {
6297 	/* hardware has been reset, we need to reload some things */
6298 	ixgbe_configure(adapter);
6299 
6300 	ixgbe_up_complete(adapter);
6301 }
6302 
6303 static unsigned long ixgbe_get_completion_timeout(struct ixgbe_adapter *adapter)
6304 {
6305 	u16 devctl2;
6306 
6307 	pcie_capability_read_word(adapter->pdev, PCI_EXP_DEVCTL2, &devctl2);
6308 
6309 	switch (devctl2 & IXGBE_PCIDEVCTRL2_TIMEO_MASK) {
6310 	case IXGBE_PCIDEVCTRL2_17_34s:
6311 	case IXGBE_PCIDEVCTRL2_4_8s:
6312 		/* For now we cap the upper limit on delay to 2 seconds
6313 		 * as we end up going up to 34 seconds of delay in worst
6314 		 * case timeout value.
6315 		 */
6316 	case IXGBE_PCIDEVCTRL2_1_2s:
6317 		return 2000000ul;	/* 2.0 s */
6318 	case IXGBE_PCIDEVCTRL2_260_520ms:
6319 		return 520000ul;	/* 520 ms */
6320 	case IXGBE_PCIDEVCTRL2_65_130ms:
6321 		return 130000ul;	/* 130 ms */
6322 	case IXGBE_PCIDEVCTRL2_16_32ms:
6323 		return 32000ul;		/* 32 ms */
6324 	case IXGBE_PCIDEVCTRL2_1_2ms:
6325 		return 2000ul;		/* 2 ms */
6326 	case IXGBE_PCIDEVCTRL2_50_100us:
6327 		return 100ul;		/* 100 us */
6328 	case IXGBE_PCIDEVCTRL2_16_32ms_def:
6329 		return 32000ul;		/* 32 ms */
6330 	default:
6331 		break;
6332 	}
6333 
6334 	/* We shouldn't need to hit this path, but just in case default as
6335 	 * though completion timeout is not supported and support 32ms.
6336 	 */
6337 	return 32000ul;
6338 }
6339 
6340 void ixgbe_disable_rx(struct ixgbe_adapter *adapter)
6341 {
6342 	unsigned long wait_delay, delay_interval;
6343 	struct ixgbe_hw *hw = &adapter->hw;
6344 	int i, wait_loop;
6345 	u32 rxdctl;
6346 
6347 	/* disable receives */
6348 	hw->mac.ops.disable_rx(hw);
6349 
6350 	if (ixgbe_removed(hw->hw_addr))
6351 		return;
6352 
6353 	/* disable all enabled Rx queues */
6354 	for (i = 0; i < adapter->num_rx_queues; i++) {
6355 		struct ixgbe_ring *ring = adapter->rx_ring[i];
6356 		u8 reg_idx = ring->reg_idx;
6357 
6358 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
6359 		rxdctl &= ~IXGBE_RXDCTL_ENABLE;
6360 		rxdctl |= IXGBE_RXDCTL_SWFLSH;
6361 
6362 		/* write value back with RXDCTL.ENABLE bit cleared */
6363 		IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
6364 	}
6365 
6366 	/* RXDCTL.EN may not change on 82598 if link is down, so skip it */
6367 	if (hw->mac.type == ixgbe_mac_82598EB &&
6368 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
6369 		return;
6370 
6371 	/* Determine our minimum delay interval. We will increase this value
6372 	 * with each subsequent test. This way if the device returns quickly
6373 	 * we should spend as little time as possible waiting, however as
6374 	 * the time increases we will wait for larger periods of time.
6375 	 *
6376 	 * The trick here is that we increase the interval using the
6377 	 * following pattern: 1x 3x 5x 7x 9x 11x 13x 15x 17x 19x. The result
6378 	 * of that wait is that it totals up to 100x whatever interval we
6379 	 * choose. Since our minimum wait is 100us we can just divide the
6380 	 * total timeout by 100 to get our minimum delay interval.
6381 	 */
6382 	delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
6383 
6384 	wait_loop = IXGBE_MAX_RX_DESC_POLL;
6385 	wait_delay = delay_interval;
6386 
6387 	while (wait_loop--) {
6388 		usleep_range(wait_delay, wait_delay + 10);
6389 		wait_delay += delay_interval * 2;
6390 		rxdctl = 0;
6391 
6392 		/* OR together the reading of all the active RXDCTL registers,
6393 		 * and then test the result. We need the disable to complete
6394 		 * before we start freeing the memory and invalidating the
6395 		 * DMA mappings.
6396 		 */
6397 		for (i = 0; i < adapter->num_rx_queues; i++) {
6398 			struct ixgbe_ring *ring = adapter->rx_ring[i];
6399 			u8 reg_idx = ring->reg_idx;
6400 
6401 			rxdctl |= IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
6402 		}
6403 
6404 		if (!(rxdctl & IXGBE_RXDCTL_ENABLE))
6405 			return;
6406 	}
6407 
6408 	e_err(drv,
6409 	      "RXDCTL.ENABLE for one or more queues not cleared within the polling period\n");
6410 }
6411 
6412 void ixgbe_disable_tx(struct ixgbe_adapter *adapter)
6413 {
6414 	unsigned long wait_delay, delay_interval;
6415 	struct ixgbe_hw *hw = &adapter->hw;
6416 	int i, wait_loop;
6417 	u32 txdctl;
6418 
6419 	if (ixgbe_removed(hw->hw_addr))
6420 		return;
6421 
6422 	/* disable all enabled Tx queues */
6423 	for (i = 0; i < adapter->num_tx_queues; i++) {
6424 		struct ixgbe_ring *ring = adapter->tx_ring[i];
6425 		u8 reg_idx = ring->reg_idx;
6426 
6427 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
6428 	}
6429 
6430 	/* disable all enabled XDP Tx queues */
6431 	for (i = 0; i < adapter->num_xdp_queues; i++) {
6432 		struct ixgbe_ring *ring = adapter->xdp_ring[i];
6433 		u8 reg_idx = ring->reg_idx;
6434 
6435 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
6436 	}
6437 
6438 	/* If the link is not up there shouldn't be much in the way of
6439 	 * pending transactions. Those that are left will be flushed out
6440 	 * when the reset logic goes through the flush sequence to clean out
6441 	 * the pending Tx transactions.
6442 	 */
6443 	if (!(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
6444 		goto dma_engine_disable;
6445 
6446 	/* Determine our minimum delay interval. We will increase this value
6447 	 * with each subsequent test. This way if the device returns quickly
6448 	 * we should spend as little time as possible waiting, however as
6449 	 * the time increases we will wait for larger periods of time.
6450 	 *
6451 	 * The trick here is that we increase the interval using the
6452 	 * following pattern: 1x 3x 5x 7x 9x 11x 13x 15x 17x 19x. The result
6453 	 * of that wait is that it totals up to 100x whatever interval we
6454 	 * choose. Since our minimum wait is 100us we can just divide the
6455 	 * total timeout by 100 to get our minimum delay interval.
6456 	 */
6457 	delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
6458 
6459 	wait_loop = IXGBE_MAX_RX_DESC_POLL;
6460 	wait_delay = delay_interval;
6461 
6462 	while (wait_loop--) {
6463 		usleep_range(wait_delay, wait_delay + 10);
6464 		wait_delay += delay_interval * 2;
6465 		txdctl = 0;
6466 
6467 		/* OR together the reading of all the active TXDCTL registers,
6468 		 * and then test the result. We need the disable to complete
6469 		 * before we start freeing the memory and invalidating the
6470 		 * DMA mappings.
6471 		 */
6472 		for (i = 0; i < adapter->num_tx_queues; i++) {
6473 			struct ixgbe_ring *ring = adapter->tx_ring[i];
6474 			u8 reg_idx = ring->reg_idx;
6475 
6476 			txdctl |= IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
6477 		}
6478 		for (i = 0; i < adapter->num_xdp_queues; i++) {
6479 			struct ixgbe_ring *ring = adapter->xdp_ring[i];
6480 			u8 reg_idx = ring->reg_idx;
6481 
6482 			txdctl |= IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
6483 		}
6484 
6485 		if (!(txdctl & IXGBE_TXDCTL_ENABLE))
6486 			goto dma_engine_disable;
6487 	}
6488 
6489 	e_err(drv,
6490 	      "TXDCTL.ENABLE for one or more queues not cleared within the polling period\n");
6491 
6492 dma_engine_disable:
6493 	/* Disable the Tx DMA engine on 82599 and later MAC */
6494 	switch (hw->mac.type) {
6495 	case ixgbe_mac_82599EB:
6496 	case ixgbe_mac_X540:
6497 	case ixgbe_mac_X550:
6498 	case ixgbe_mac_X550EM_x:
6499 	case ixgbe_mac_x550em_a:
6500 	case ixgbe_mac_e610:
6501 		IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
6502 				(IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
6503 				 ~IXGBE_DMATXCTL_TE));
6504 		fallthrough;
6505 	default:
6506 		break;
6507 	}
6508 }
6509 
6510 void ixgbe_reset(struct ixgbe_adapter *adapter)
6511 {
6512 	struct ixgbe_hw *hw = &adapter->hw;
6513 	struct net_device *netdev = adapter->netdev;
6514 	int err;
6515 
6516 	if (ixgbe_removed(hw->hw_addr))
6517 		return;
6518 	/* lock SFP init bit to prevent race conditions with the watchdog */
6519 	while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
6520 		usleep_range(1000, 2000);
6521 
6522 	/* clear all SFP and link config related flags while holding SFP_INIT */
6523 	adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
6524 			     IXGBE_FLAG2_SFP_NEEDS_RESET);
6525 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
6526 
6527 	err = hw->mac.ops.init_hw(hw);
6528 	switch (err) {
6529 	case 0:
6530 	case -ENOENT:
6531 	case -EOPNOTSUPP:
6532 		break;
6533 	case -EALREADY:
6534 		e_dev_err("primary disable timed out\n");
6535 		break;
6536 	case -EACCES:
6537 		/* We are running on a pre-production device, log a warning */
6538 		e_dev_warn("This device is a pre-production adapter/LOM. "
6539 			   "Please be aware there may be issues associated with "
6540 			   "your hardware.  If you are experiencing problems "
6541 			   "please contact your Intel or hardware "
6542 			   "representative who provided you with this "
6543 			   "hardware.\n");
6544 		break;
6545 	default:
6546 		e_dev_err("Hardware Error: %d\n", err);
6547 	}
6548 
6549 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
6550 
6551 	/* flush entries out of MAC table */
6552 	ixgbe_flush_sw_mac_table(adapter);
6553 	__dev_uc_unsync(netdev, NULL);
6554 
6555 	/* do not flush user set addresses */
6556 	ixgbe_mac_set_default_filter(adapter);
6557 
6558 	/* update SAN MAC vmdq pool selection */
6559 	if (hw->mac.san_mac_rar_index)
6560 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
6561 
6562 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6563 		ixgbe_ptp_reset(adapter);
6564 
6565 	if (hw->phy.ops.set_phy_power) {
6566 		if (!netif_running(adapter->netdev) && !adapter->wol)
6567 			hw->phy.ops.set_phy_power(hw, false);
6568 		else
6569 			hw->phy.ops.set_phy_power(hw, true);
6570 	}
6571 }
6572 
6573 /**
6574  * ixgbe_clean_tx_ring - Free Tx Buffers
6575  * @tx_ring: ring to be cleaned
6576  **/
6577 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
6578 {
6579 	u16 i = tx_ring->next_to_clean;
6580 	struct ixgbe_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
6581 
6582 	if (tx_ring->xsk_pool) {
6583 		ixgbe_xsk_clean_tx_ring(tx_ring);
6584 		goto out;
6585 	}
6586 
6587 	while (i != tx_ring->next_to_use) {
6588 		union ixgbe_adv_tx_desc *eop_desc, *tx_desc;
6589 
6590 		/* Free all the Tx ring sk_buffs */
6591 		if (ring_is_xdp(tx_ring))
6592 			xdp_return_frame(tx_buffer->xdpf);
6593 		else
6594 			dev_kfree_skb_any(tx_buffer->skb);
6595 
6596 		/* unmap skb header data */
6597 		dma_unmap_single(tx_ring->dev,
6598 				 dma_unmap_addr(tx_buffer, dma),
6599 				 dma_unmap_len(tx_buffer, len),
6600 				 DMA_TO_DEVICE);
6601 
6602 		/* check for eop_desc to determine the end of the packet */
6603 		eop_desc = tx_buffer->next_to_watch;
6604 		tx_desc = IXGBE_TX_DESC(tx_ring, i);
6605 
6606 		/* unmap remaining buffers */
6607 		while (tx_desc != eop_desc) {
6608 			tx_buffer++;
6609 			tx_desc++;
6610 			i++;
6611 			if (unlikely(i == tx_ring->count)) {
6612 				i = 0;
6613 				tx_buffer = tx_ring->tx_buffer_info;
6614 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6615 			}
6616 
6617 			/* unmap any remaining paged data */
6618 			if (dma_unmap_len(tx_buffer, len))
6619 				dma_unmap_page(tx_ring->dev,
6620 					       dma_unmap_addr(tx_buffer, dma),
6621 					       dma_unmap_len(tx_buffer, len),
6622 					       DMA_TO_DEVICE);
6623 		}
6624 
6625 		/* move us one more past the eop_desc for start of next pkt */
6626 		tx_buffer++;
6627 		i++;
6628 		if (unlikely(i == tx_ring->count)) {
6629 			i = 0;
6630 			tx_buffer = tx_ring->tx_buffer_info;
6631 		}
6632 	}
6633 
6634 	/* reset BQL for queue */
6635 	if (!ring_is_xdp(tx_ring))
6636 		netdev_tx_reset_queue(txring_txq(tx_ring));
6637 
6638 out:
6639 	/* reset next_to_use and next_to_clean */
6640 	tx_ring->next_to_use = 0;
6641 	tx_ring->next_to_clean = 0;
6642 }
6643 
6644 /**
6645  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
6646  * @adapter: board private structure
6647  **/
6648 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
6649 {
6650 	int i;
6651 
6652 	for (i = 0; i < adapter->num_rx_queues; i++)
6653 		ixgbe_clean_rx_ring(adapter->rx_ring[i]);
6654 }
6655 
6656 /**
6657  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
6658  * @adapter: board private structure
6659  **/
6660 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
6661 {
6662 	int i;
6663 
6664 	for (i = 0; i < adapter->num_tx_queues; i++)
6665 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
6666 	for (i = 0; i < adapter->num_xdp_queues; i++)
6667 		ixgbe_clean_tx_ring(adapter->xdp_ring[i]);
6668 }
6669 
6670 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
6671 {
6672 	struct hlist_node *node2;
6673 	struct ixgbe_fdir_filter *filter;
6674 
6675 	spin_lock(&adapter->fdir_perfect_lock);
6676 
6677 	hlist_for_each_entry_safe(filter, node2,
6678 				  &adapter->fdir_filter_list, fdir_node) {
6679 		hlist_del(&filter->fdir_node);
6680 		kfree(filter);
6681 	}
6682 	adapter->fdir_filter_count = 0;
6683 
6684 	spin_unlock(&adapter->fdir_perfect_lock);
6685 }
6686 
6687 void ixgbe_down(struct ixgbe_adapter *adapter)
6688 {
6689 	struct net_device *netdev = adapter->netdev;
6690 	struct ixgbe_hw *hw = &adapter->hw;
6691 	int i;
6692 
6693 	/* signal that we are down to the interrupt handler */
6694 	if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
6695 		return; /* do nothing if already down */
6696 
6697 	/* Shut off incoming Tx traffic */
6698 	netif_tx_stop_all_queues(netdev);
6699 
6700 	/* call carrier off first to avoid false dev_watchdog timeouts */
6701 	netif_carrier_off(netdev);
6702 	netif_tx_disable(netdev);
6703 
6704 	/* Disable Rx */
6705 	ixgbe_disable_rx(adapter);
6706 
6707 	/* synchronize_rcu() needed for pending XDP buffers to drain */
6708 	if (adapter->xdp_ring[0])
6709 		synchronize_rcu();
6710 
6711 	ixgbe_irq_disable(adapter);
6712 
6713 	ixgbe_napi_disable_all(adapter);
6714 
6715 	clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
6716 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6717 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6718 
6719 	timer_delete_sync(&adapter->service_timer);
6720 
6721 	if (adapter->num_vfs) {
6722 		/* Clear EITR Select mapping */
6723 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
6724 
6725 		/* Mark all the VFs as inactive */
6726 		for (i = 0 ; i < adapter->num_vfs; i++)
6727 			adapter->vfinfo[i].clear_to_send = false;
6728 
6729 		/* update setting rx tx for all active vfs */
6730 		ixgbe_set_all_vfs(adapter);
6731 	}
6732 
6733 	/* disable transmits in the hardware now that interrupts are off */
6734 	ixgbe_disable_tx(adapter);
6735 
6736 	if (!pci_channel_offline(adapter->pdev))
6737 		ixgbe_reset(adapter);
6738 
6739 	/* power down the optics for 82599 SFP+ fiber */
6740 	if (hw->mac.ops.disable_tx_laser)
6741 		hw->mac.ops.disable_tx_laser(hw);
6742 
6743 	ixgbe_clean_all_tx_rings(adapter);
6744 	ixgbe_clean_all_rx_rings(adapter);
6745 	if (adapter->hw.mac.type == ixgbe_mac_e610)
6746 		ixgbe_disable_link_status_events(adapter);
6747 }
6748 
6749 /**
6750  * ixgbe_set_eee_capable - helper function to determine EEE support on X550
6751  * @adapter: board private structure
6752  */
6753 static void ixgbe_set_eee_capable(struct ixgbe_adapter *adapter)
6754 {
6755 	struct ixgbe_hw *hw = &adapter->hw;
6756 
6757 	switch (hw->device_id) {
6758 	case IXGBE_DEV_ID_X550EM_A_1G_T:
6759 	case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6760 		if (!hw->phy.eee_speeds_supported)
6761 			break;
6762 		adapter->flags2 |= IXGBE_FLAG2_EEE_CAPABLE;
6763 		if (!hw->phy.eee_speeds_advertised)
6764 			break;
6765 		adapter->flags2 |= IXGBE_FLAG2_EEE_ENABLED;
6766 		break;
6767 	default:
6768 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_CAPABLE;
6769 		adapter->flags2 &= ~IXGBE_FLAG2_EEE_ENABLED;
6770 		break;
6771 	}
6772 }
6773 
6774 /**
6775  * ixgbe_tx_timeout - Respond to a Tx Hang
6776  * @netdev: network interface device structure
6777  * @txqueue: queue number that timed out
6778  **/
6779 static void ixgbe_tx_timeout(struct net_device *netdev, unsigned int __always_unused txqueue)
6780 {
6781 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
6782 
6783 	/* Do the reset outside of interrupt context */
6784 	ixgbe_tx_timeout_reset(adapter);
6785 }
6786 
6787 #ifdef CONFIG_IXGBE_DCB
6788 static void ixgbe_init_dcb(struct ixgbe_adapter *adapter)
6789 {
6790 	struct ixgbe_hw *hw = &adapter->hw;
6791 	struct tc_configuration *tc;
6792 	int j;
6793 
6794 	switch (hw->mac.type) {
6795 	case ixgbe_mac_82598EB:
6796 	case ixgbe_mac_82599EB:
6797 		adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
6798 		adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
6799 		break;
6800 	case ixgbe_mac_X540:
6801 	case ixgbe_mac_X550:
6802 	case ixgbe_mac_e610:
6803 		adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
6804 		adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
6805 		break;
6806 	case ixgbe_mac_X550EM_x:
6807 	case ixgbe_mac_x550em_a:
6808 	default:
6809 		adapter->dcb_cfg.num_tcs.pg_tcs = DEF_TRAFFIC_CLASS;
6810 		adapter->dcb_cfg.num_tcs.pfc_tcs = DEF_TRAFFIC_CLASS;
6811 		break;
6812 	}
6813 
6814 	/* Configure DCB traffic classes */
6815 	for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
6816 		tc = &adapter->dcb_cfg.tc_config[j];
6817 		tc->path[DCB_TX_CONFIG].bwg_id = 0;
6818 		tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
6819 		tc->path[DCB_RX_CONFIG].bwg_id = 0;
6820 		tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
6821 		tc->dcb_pfc = pfc_disabled;
6822 	}
6823 
6824 	/* Initialize default user to priority mapping, UPx->TC0 */
6825 	tc = &adapter->dcb_cfg.tc_config[0];
6826 	tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
6827 	tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
6828 
6829 	adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
6830 	adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
6831 	adapter->dcb_cfg.pfc_mode_enable = false;
6832 	adapter->dcb_set_bitmap = 0x00;
6833 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
6834 		adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
6835 	memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
6836 	       sizeof(adapter->temp_dcb_cfg));
6837 }
6838 #endif
6839 
6840 /**
6841  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
6842  * @adapter: board private structure to initialize
6843  * @ii: pointer to ixgbe_info for device
6844  *
6845  * ixgbe_sw_init initializes the Adapter private data structure.
6846  * Fields are initialized based on PCI device information and
6847  * OS network device settings (MTU size).
6848  **/
6849 static int ixgbe_sw_init(struct ixgbe_adapter *adapter,
6850 			 const struct ixgbe_info *ii)
6851 {
6852 	struct ixgbe_hw *hw = &adapter->hw;
6853 	struct pci_dev *pdev = adapter->pdev;
6854 	unsigned int rss, fdir;
6855 	u32 fwsm;
6856 	int i;
6857 
6858 	/* PCI config space info */
6859 
6860 	hw->vendor_id = pdev->vendor;
6861 	hw->device_id = pdev->device;
6862 	hw->revision_id = pdev->revision;
6863 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
6864 	hw->subsystem_device_id = pdev->subsystem_device;
6865 
6866 	hw->mac.max_link_up_time = IXGBE_LINK_UP_TIME;
6867 
6868 	/* get_invariants needs the device IDs */
6869 	ii->get_invariants(hw);
6870 
6871 	/* Set common capability flags and settings */
6872 	rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
6873 	adapter->ring_feature[RING_F_RSS].limit = rss;
6874 	adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
6875 	adapter->max_q_vectors = MAX_Q_VECTORS_82599;
6876 	adapter->atr_sample_rate = 20;
6877 	fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
6878 	adapter->ring_feature[RING_F_FDIR].limit = fdir;
6879 	adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
6880 	adapter->ring_feature[RING_F_VMDQ].limit = 1;
6881 #ifdef CONFIG_IXGBE_DCA
6882 	adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
6883 #endif
6884 #ifdef CONFIG_IXGBE_DCB
6885 	adapter->flags |= IXGBE_FLAG_DCB_CAPABLE;
6886 	adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
6887 #endif
6888 #ifdef IXGBE_FCOE
6889 	adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
6890 	adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6891 #ifdef CONFIG_IXGBE_DCB
6892 	/* Default traffic class to use for FCoE */
6893 	adapter->fcoe.up = IXGBE_FCOE_DEFTC;
6894 #endif /* CONFIG_IXGBE_DCB */
6895 #endif /* IXGBE_FCOE */
6896 
6897 	/* initialize static ixgbe jump table entries */
6898 	adapter->jump_tables[0] = kzalloc(sizeof(*adapter->jump_tables[0]),
6899 					  GFP_KERNEL);
6900 	if (!adapter->jump_tables[0])
6901 		return -ENOMEM;
6902 	adapter->jump_tables[0]->mat = ixgbe_ipv4_fields;
6903 
6904 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++)
6905 		adapter->jump_tables[i] = NULL;
6906 
6907 	adapter->mac_table = kcalloc(hw->mac.num_rar_entries,
6908 				     sizeof(struct ixgbe_mac_addr),
6909 				     GFP_KERNEL);
6910 	if (!adapter->mac_table)
6911 		return -ENOMEM;
6912 
6913 	if (ixgbe_init_rss_key(adapter))
6914 		return -ENOMEM;
6915 
6916 	adapter->af_xdp_zc_qps = bitmap_zalloc(IXGBE_MAX_XDP_QS, GFP_KERNEL);
6917 	if (!adapter->af_xdp_zc_qps)
6918 		return -ENOMEM;
6919 
6920 	/* Set MAC specific capability flags and exceptions */
6921 	switch (hw->mac.type) {
6922 	case ixgbe_mac_82598EB:
6923 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
6924 
6925 		if (hw->device_id == IXGBE_DEV_ID_82598AT)
6926 			adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
6927 
6928 		adapter->max_q_vectors = MAX_Q_VECTORS_82598;
6929 		adapter->ring_feature[RING_F_FDIR].limit = 0;
6930 		adapter->atr_sample_rate = 0;
6931 		adapter->fdir_pballoc = 0;
6932 #ifdef IXGBE_FCOE
6933 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6934 		adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6935 #ifdef CONFIG_IXGBE_DCB
6936 		adapter->fcoe.up = 0;
6937 #endif /* IXGBE_DCB */
6938 #endif /* IXGBE_FCOE */
6939 		break;
6940 	case ixgbe_mac_82599EB:
6941 		if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
6942 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6943 		break;
6944 	case ixgbe_mac_X540:
6945 		fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
6946 		if (fwsm & IXGBE_FWSM_TS_ENABLED)
6947 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6948 		break;
6949 	case ixgbe_mac_x550em_a:
6950 		switch (hw->device_id) {
6951 		case IXGBE_DEV_ID_X550EM_A_1G_T:
6952 		case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6953 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6954 			break;
6955 		default:
6956 			break;
6957 		}
6958 		fallthrough;
6959 	case ixgbe_mac_X550EM_x:
6960 #ifdef CONFIG_IXGBE_DCB
6961 		adapter->flags &= ~IXGBE_FLAG_DCB_CAPABLE;
6962 #endif
6963 #ifdef IXGBE_FCOE
6964 		adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6965 #ifdef CONFIG_IXGBE_DCB
6966 		adapter->fcoe.up = 0;
6967 #endif /* IXGBE_DCB */
6968 #endif /* IXGBE_FCOE */
6969 		fallthrough;
6970 	case ixgbe_mac_X550:
6971 		if (hw->mac.type == ixgbe_mac_X550)
6972 			adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6973 #ifdef CONFIG_IXGBE_DCA
6974 		adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
6975 #endif
6976 		break;
6977 	default:
6978 		break;
6979 	}
6980 
6981 	/* Make sure the SWFW semaphore is in a valid state */
6982 	if (hw->mac.ops.init_swfw_sync)
6983 		hw->mac.ops.init_swfw_sync(hw);
6984 
6985 	if (hw->mac.type == ixgbe_mac_e610)
6986 		mutex_init(&hw->aci.lock);
6987 
6988 #ifdef IXGBE_FCOE
6989 	/* FCoE support exists, always init the FCoE lock */
6990 	spin_lock_init(&adapter->fcoe.lock);
6991 
6992 #endif
6993 	/* n-tuple support exists, always init our spinlock */
6994 	spin_lock_init(&adapter->fdir_perfect_lock);
6995 
6996 	/* init spinlock to avoid concurrency of VF resources */
6997 	spin_lock_init(&adapter->vfs_lock);
6998 
6999 #ifdef CONFIG_IXGBE_DCB
7000 	ixgbe_init_dcb(adapter);
7001 #endif
7002 	ixgbe_init_ipsec_offload(adapter);
7003 
7004 	/* default flow control settings */
7005 	hw->fc.requested_mode = ixgbe_fc_full;
7006 	hw->fc.current_mode = ixgbe_fc_full;	/* init for ethtool output */
7007 	ixgbe_pbthresh_setup(adapter);
7008 	hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
7009 	hw->fc.send_xon = true;
7010 	hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
7011 
7012 #ifdef CONFIG_PCI_IOV
7013 	if (max_vfs > 0)
7014 		e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
7015 
7016 	/* assign number of SR-IOV VFs */
7017 	if (hw->mac.type != ixgbe_mac_82598EB) {
7018 		if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
7019 			max_vfs = 0;
7020 			e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
7021 		}
7022 	}
7023 #endif /* CONFIG_PCI_IOV */
7024 
7025 	/* enable itr by default in dynamic mode */
7026 	adapter->rx_itr_setting = 1;
7027 	adapter->tx_itr_setting = 1;
7028 
7029 	/* set default ring sizes */
7030 	adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
7031 	adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
7032 
7033 	/* set default work limits */
7034 	adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
7035 
7036 	/* initialize eeprom parameters */
7037 	if (hw->eeprom.ops.init_params(hw)) {
7038 		e_dev_err("EEPROM initialization failed\n");
7039 		return -EIO;
7040 	}
7041 
7042 	/* PF holds first pool slot */
7043 	set_bit(0, adapter->fwd_bitmask);
7044 	set_bit(__IXGBE_DOWN, &adapter->state);
7045 
7046 	/* enable locking for XDP_TX if we have more CPUs than queues */
7047 	if (nr_cpu_ids > IXGBE_MAX_XDP_QS)
7048 		static_branch_enable(&ixgbe_xdp_locking_key);
7049 
7050 	return 0;
7051 }
7052 
7053 /**
7054  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
7055  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
7056  *
7057  * Return 0 on success, negative on failure
7058  **/
7059 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
7060 {
7061 	struct device *dev = tx_ring->dev;
7062 	int orig_node = dev_to_node(dev);
7063 	int ring_node = NUMA_NO_NODE;
7064 	int size;
7065 
7066 	size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
7067 
7068 	if (tx_ring->q_vector)
7069 		ring_node = tx_ring->q_vector->numa_node;
7070 
7071 	tx_ring->tx_buffer_info = vmalloc_node(size, ring_node);
7072 	if (!tx_ring->tx_buffer_info)
7073 		tx_ring->tx_buffer_info = vmalloc(size);
7074 	if (!tx_ring->tx_buffer_info)
7075 		goto err;
7076 
7077 	/* round up to nearest 4K */
7078 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
7079 	tx_ring->size = ALIGN(tx_ring->size, 4096);
7080 
7081 	set_dev_node(dev, ring_node);
7082 	tx_ring->desc = dma_alloc_coherent(dev,
7083 					   tx_ring->size,
7084 					   &tx_ring->dma,
7085 					   GFP_KERNEL);
7086 	set_dev_node(dev, orig_node);
7087 	if (!tx_ring->desc)
7088 		tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
7089 						   &tx_ring->dma, GFP_KERNEL);
7090 	if (!tx_ring->desc)
7091 		goto err;
7092 
7093 	tx_ring->next_to_use = 0;
7094 	tx_ring->next_to_clean = 0;
7095 	return 0;
7096 
7097 err:
7098 	vfree(tx_ring->tx_buffer_info);
7099 	tx_ring->tx_buffer_info = NULL;
7100 	dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
7101 	return -ENOMEM;
7102 }
7103 
7104 /**
7105  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
7106  * @adapter: board private structure
7107  *
7108  * If this function returns with an error, then it's possible one or
7109  * more of the rings is populated (while the rest are not).  It is the
7110  * callers duty to clean those orphaned rings.
7111  *
7112  * Return 0 on success, negative on failure
7113  **/
7114 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
7115 {
7116 	int i, j = 0, err = 0;
7117 
7118 	for (i = 0; i < adapter->num_tx_queues; i++) {
7119 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
7120 		if (!err)
7121 			continue;
7122 
7123 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
7124 		goto err_setup_tx;
7125 	}
7126 	for (j = 0; j < adapter->num_xdp_queues; j++) {
7127 		err = ixgbe_setup_tx_resources(adapter->xdp_ring[j]);
7128 		if (!err)
7129 			continue;
7130 
7131 		e_err(probe, "Allocation for Tx Queue %u failed\n", j);
7132 		goto err_setup_tx;
7133 	}
7134 
7135 	return 0;
7136 err_setup_tx:
7137 	/* rewind the index freeing the rings as we go */
7138 	while (j--)
7139 		ixgbe_free_tx_resources(adapter->xdp_ring[j]);
7140 	while (i--)
7141 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
7142 	return err;
7143 }
7144 
7145 static int ixgbe_rx_napi_id(struct ixgbe_ring *rx_ring)
7146 {
7147 	struct ixgbe_q_vector *q_vector = rx_ring->q_vector;
7148 
7149 	return q_vector ? q_vector->napi.napi_id : 0;
7150 }
7151 
7152 /**
7153  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
7154  * @adapter: pointer to ixgbe_adapter
7155  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
7156  *
7157  * Returns 0 on success, negative on failure
7158  **/
7159 int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
7160 			     struct ixgbe_ring *rx_ring)
7161 {
7162 	struct device *dev = rx_ring->dev;
7163 	int orig_node = dev_to_node(dev);
7164 	int ring_node = NUMA_NO_NODE;
7165 	int size;
7166 
7167 	size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
7168 
7169 	if (rx_ring->q_vector)
7170 		ring_node = rx_ring->q_vector->numa_node;
7171 
7172 	rx_ring->rx_buffer_info = vmalloc_node(size, ring_node);
7173 	if (!rx_ring->rx_buffer_info)
7174 		rx_ring->rx_buffer_info = vmalloc(size);
7175 	if (!rx_ring->rx_buffer_info)
7176 		goto err;
7177 
7178 	/* Round up to nearest 4K */
7179 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
7180 	rx_ring->size = ALIGN(rx_ring->size, 4096);
7181 
7182 	set_dev_node(dev, ring_node);
7183 	rx_ring->desc = dma_alloc_coherent(dev,
7184 					   rx_ring->size,
7185 					   &rx_ring->dma,
7186 					   GFP_KERNEL);
7187 	set_dev_node(dev, orig_node);
7188 	if (!rx_ring->desc)
7189 		rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
7190 						   &rx_ring->dma, GFP_KERNEL);
7191 	if (!rx_ring->desc)
7192 		goto err;
7193 
7194 	rx_ring->next_to_clean = 0;
7195 	rx_ring->next_to_use = 0;
7196 
7197 	/* XDP RX-queue info */
7198 	if (xdp_rxq_info_reg(&rx_ring->xdp_rxq, adapter->netdev,
7199 			     rx_ring->queue_index, ixgbe_rx_napi_id(rx_ring)) < 0)
7200 		goto err;
7201 
7202 	WRITE_ONCE(rx_ring->xdp_prog, adapter->xdp_prog);
7203 
7204 	return 0;
7205 err:
7206 	vfree(rx_ring->rx_buffer_info);
7207 	rx_ring->rx_buffer_info = NULL;
7208 	dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
7209 	return -ENOMEM;
7210 }
7211 
7212 /**
7213  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
7214  * @adapter: board private structure
7215  *
7216  * If this function returns with an error, then it's possible one or
7217  * more of the rings is populated (while the rest are not).  It is the
7218  * callers duty to clean those orphaned rings.
7219  *
7220  * Return 0 on success, negative on failure
7221  **/
7222 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
7223 {
7224 	int i, err = 0;
7225 
7226 	for (i = 0; i < adapter->num_rx_queues; i++) {
7227 		err = ixgbe_setup_rx_resources(adapter, adapter->rx_ring[i]);
7228 		if (!err)
7229 			continue;
7230 
7231 		e_err(probe, "Allocation for Rx Queue %u failed\n", i);
7232 		goto err_setup_rx;
7233 	}
7234 
7235 #ifdef IXGBE_FCOE
7236 	err = ixgbe_setup_fcoe_ddp_resources(adapter);
7237 	if (!err)
7238 #endif
7239 		return 0;
7240 err_setup_rx:
7241 	/* rewind the index freeing the rings as we go */
7242 	while (i--)
7243 		ixgbe_free_rx_resources(adapter->rx_ring[i]);
7244 	return err;
7245 }
7246 
7247 /**
7248  * ixgbe_free_tx_resources - Free Tx Resources per Queue
7249  * @tx_ring: Tx descriptor ring for a specific queue
7250  *
7251  * Free all transmit software resources
7252  **/
7253 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
7254 {
7255 	ixgbe_clean_tx_ring(tx_ring);
7256 
7257 	vfree(tx_ring->tx_buffer_info);
7258 	tx_ring->tx_buffer_info = NULL;
7259 
7260 	/* if not set, then don't free */
7261 	if (!tx_ring->desc)
7262 		return;
7263 
7264 	dma_free_coherent(tx_ring->dev, tx_ring->size,
7265 			  tx_ring->desc, tx_ring->dma);
7266 
7267 	tx_ring->desc = NULL;
7268 }
7269 
7270 /**
7271  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
7272  * @adapter: board private structure
7273  *
7274  * Free all transmit software resources
7275  **/
7276 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
7277 {
7278 	int i;
7279 
7280 	for (i = 0; i < adapter->num_tx_queues; i++)
7281 		if (adapter->tx_ring[i]->desc)
7282 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
7283 	for (i = 0; i < adapter->num_xdp_queues; i++)
7284 		if (adapter->xdp_ring[i]->desc)
7285 			ixgbe_free_tx_resources(adapter->xdp_ring[i]);
7286 }
7287 
7288 /**
7289  * ixgbe_free_rx_resources - Free Rx Resources
7290  * @rx_ring: ring to clean the resources from
7291  *
7292  * Free all receive software resources
7293  **/
7294 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
7295 {
7296 	ixgbe_clean_rx_ring(rx_ring);
7297 
7298 	rx_ring->xdp_prog = NULL;
7299 	xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
7300 	vfree(rx_ring->rx_buffer_info);
7301 	rx_ring->rx_buffer_info = NULL;
7302 
7303 	/* if not set, then don't free */
7304 	if (!rx_ring->desc)
7305 		return;
7306 
7307 	dma_free_coherent(rx_ring->dev, rx_ring->size,
7308 			  rx_ring->desc, rx_ring->dma);
7309 
7310 	rx_ring->desc = NULL;
7311 }
7312 
7313 /**
7314  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
7315  * @adapter: board private structure
7316  *
7317  * Free all receive software resources
7318  **/
7319 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
7320 {
7321 	int i;
7322 
7323 #ifdef IXGBE_FCOE
7324 	ixgbe_free_fcoe_ddp_resources(adapter);
7325 
7326 #endif
7327 	for (i = 0; i < adapter->num_rx_queues; i++)
7328 		if (adapter->rx_ring[i]->desc)
7329 			ixgbe_free_rx_resources(adapter->rx_ring[i]);
7330 }
7331 
7332 /**
7333  * ixgbe_max_xdp_frame_size - returns the maximum allowed frame size for XDP
7334  * @adapter: device handle, pointer to adapter
7335  */
7336 static int ixgbe_max_xdp_frame_size(struct ixgbe_adapter *adapter)
7337 {
7338 	if (PAGE_SIZE >= 8192 || adapter->flags2 & IXGBE_FLAG2_RX_LEGACY)
7339 		return IXGBE_RXBUFFER_2K;
7340 	else
7341 		return IXGBE_RXBUFFER_3K;
7342 }
7343 
7344 /**
7345  * ixgbe_change_mtu - Change the Maximum Transfer Unit
7346  * @netdev: network interface device structure
7347  * @new_mtu: new value for maximum frame size
7348  *
7349  * Returns 0 on success, negative on failure
7350  **/
7351 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
7352 {
7353 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
7354 
7355 	if (ixgbe_enabled_xdp_adapter(adapter)) {
7356 		int new_frame_size = new_mtu + IXGBE_PKT_HDR_PAD;
7357 
7358 		if (new_frame_size > ixgbe_max_xdp_frame_size(adapter)) {
7359 			e_warn(probe, "Requested MTU size is not supported with XDP\n");
7360 			return -EINVAL;
7361 		}
7362 	}
7363 
7364 	/*
7365 	 * For 82599EB we cannot allow legacy VFs to enable their receive
7366 	 * paths when MTU greater than 1500 is configured.  So display a
7367 	 * warning that legacy VFs will be disabled.
7368 	 */
7369 	if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
7370 	    (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
7371 	    (new_mtu > ETH_DATA_LEN))
7372 		e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
7373 
7374 	netdev_dbg(netdev, "changing MTU from %d to %d\n",
7375 		   netdev->mtu, new_mtu);
7376 
7377 	/* must set new MTU before calling down or up */
7378 	WRITE_ONCE(netdev->mtu, new_mtu);
7379 
7380 	if (netif_running(netdev))
7381 		ixgbe_reinit_locked(adapter);
7382 
7383 	return 0;
7384 }
7385 
7386 /**
7387  * ixgbe_open - Called when a network interface is made active
7388  * @netdev: network interface device structure
7389  *
7390  * Returns 0 on success, negative value on failure
7391  *
7392  * The open entry point is called when a network interface is made
7393  * active by the system (IFF_UP).  At this point all resources needed
7394  * for transmit and receive operations are allocated, the interrupt
7395  * handler is registered with the OS, the watchdog timer is started,
7396  * and the stack is notified that the interface is ready.
7397  **/
7398 int ixgbe_open(struct net_device *netdev)
7399 {
7400 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
7401 	struct ixgbe_hw *hw = &adapter->hw;
7402 	int err, queues;
7403 
7404 	/* disallow open during test */
7405 	if (test_bit(__IXGBE_TESTING, &adapter->state))
7406 		return -EBUSY;
7407 
7408 	netif_carrier_off(netdev);
7409 
7410 	/* allocate transmit descriptors */
7411 	err = ixgbe_setup_all_tx_resources(adapter);
7412 	if (err)
7413 		goto err_setup_tx;
7414 
7415 	/* allocate receive descriptors */
7416 	err = ixgbe_setup_all_rx_resources(adapter);
7417 	if (err)
7418 		goto err_setup_rx;
7419 
7420 	ixgbe_configure(adapter);
7421 
7422 	err = ixgbe_request_irq(adapter);
7423 	if (err)
7424 		goto err_req_irq;
7425 
7426 	/* Notify the stack of the actual queue counts. */
7427 	queues = adapter->num_tx_queues;
7428 	err = netif_set_real_num_tx_queues(netdev, queues);
7429 	if (err)
7430 		goto err_set_queues;
7431 
7432 	queues = adapter->num_rx_queues;
7433 	err = netif_set_real_num_rx_queues(netdev, queues);
7434 	if (err)
7435 		goto err_set_queues;
7436 
7437 	ixgbe_ptp_init(adapter);
7438 
7439 	ixgbe_up_complete(adapter);
7440 
7441 	udp_tunnel_nic_reset_ntf(netdev);
7442 	if (adapter->hw.mac.type == ixgbe_mac_e610) {
7443 		int err = ixgbe_update_link_info(&adapter->hw);
7444 
7445 		if (err)
7446 			e_dev_err("Failed to update link info, err %d.\n", err);
7447 
7448 		ixgbe_check_link_cfg_err(adapter,
7449 					 adapter->hw.link.link_info.link_cfg_err);
7450 
7451 		err = ixgbe_non_sfp_link_config(&adapter->hw);
7452 		if (ixgbe_non_sfp_link_config(&adapter->hw))
7453 			e_dev_err("Link setup failed, err %d.\n", err);
7454 	}
7455 
7456 	return 0;
7457 
7458 err_set_queues:
7459 	ixgbe_free_irq(adapter);
7460 err_req_irq:
7461 	ixgbe_free_all_rx_resources(adapter);
7462 	if (hw->phy.ops.set_phy_power && !adapter->wol)
7463 		hw->phy.ops.set_phy_power(&adapter->hw, false);
7464 err_setup_rx:
7465 	ixgbe_free_all_tx_resources(adapter);
7466 err_setup_tx:
7467 	ixgbe_reset(adapter);
7468 
7469 	return err;
7470 }
7471 
7472 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
7473 {
7474 	ixgbe_ptp_suspend(adapter);
7475 
7476 	if (adapter->hw.phy.ops.enter_lplu) {
7477 		adapter->hw.phy.reset_disable = true;
7478 		ixgbe_down(adapter);
7479 		adapter->hw.phy.ops.enter_lplu(&adapter->hw);
7480 		adapter->hw.phy.reset_disable = false;
7481 	} else {
7482 		ixgbe_down(adapter);
7483 	}
7484 
7485 	ixgbe_free_irq(adapter);
7486 
7487 	ixgbe_free_all_tx_resources(adapter);
7488 	ixgbe_free_all_rx_resources(adapter);
7489 }
7490 
7491 /**
7492  * ixgbe_close - Disables a network interface
7493  * @netdev: network interface device structure
7494  *
7495  * Returns 0, this is not allowed to fail
7496  *
7497  * The close entry point is called when an interface is de-activated
7498  * by the OS.  The hardware is still under the drivers control, but
7499  * needs to be disabled.  A global MAC reset is issued to stop the
7500  * hardware, and all transmit and receive resources are freed.
7501  **/
7502 int ixgbe_close(struct net_device *netdev)
7503 {
7504 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
7505 
7506 	ixgbe_ptp_stop(adapter);
7507 
7508 	if (netif_device_present(netdev))
7509 		ixgbe_close_suspend(adapter);
7510 
7511 	ixgbe_fdir_filter_exit(adapter);
7512 
7513 	ixgbe_release_hw_control(adapter);
7514 
7515 	return 0;
7516 }
7517 
7518 static int ixgbe_resume(struct device *dev_d)
7519 {
7520 	struct pci_dev *pdev = to_pci_dev(dev_d);
7521 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7522 	struct net_device *netdev = adapter->netdev;
7523 	u32 err;
7524 
7525 	adapter->hw.hw_addr = adapter->io_addr;
7526 
7527 	err = pci_enable_device_mem(pdev);
7528 	if (err) {
7529 		e_dev_err("Cannot enable PCI device from suspend\n");
7530 		return err;
7531 	}
7532 	smp_mb__before_atomic();
7533 	clear_bit(__IXGBE_DISABLED, &adapter->state);
7534 	pci_set_master(pdev);
7535 
7536 	device_wakeup_disable(dev_d);
7537 
7538 	ixgbe_reset(adapter);
7539 
7540 	IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
7541 
7542 	rtnl_lock();
7543 	err = ixgbe_init_interrupt_scheme(adapter);
7544 	if (!err && netif_running(netdev))
7545 		err = ixgbe_open(netdev);
7546 
7547 
7548 	if (!err)
7549 		netif_device_attach(netdev);
7550 	rtnl_unlock();
7551 
7552 	return err;
7553 }
7554 
7555 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
7556 {
7557 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
7558 	struct net_device *netdev = adapter->netdev;
7559 	struct ixgbe_hw *hw = &adapter->hw;
7560 	u32 ctrl;
7561 	u32 wufc = adapter->wol;
7562 
7563 	rtnl_lock();
7564 	netif_device_detach(netdev);
7565 
7566 	if (netif_running(netdev))
7567 		ixgbe_close_suspend(adapter);
7568 
7569 	ixgbe_clear_interrupt_scheme(adapter);
7570 	rtnl_unlock();
7571 
7572 	if (hw->mac.ops.stop_link_on_d3)
7573 		hw->mac.ops.stop_link_on_d3(hw);
7574 
7575 	if (wufc) {
7576 		u32 fctrl;
7577 
7578 		ixgbe_set_rx_mode(netdev);
7579 
7580 		/* enable the optics for 82599 SFP+ fiber as we can WoL */
7581 		if (hw->mac.ops.enable_tx_laser)
7582 			hw->mac.ops.enable_tx_laser(hw);
7583 
7584 		/* enable the reception of multicast packets */
7585 		fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
7586 		fctrl |= IXGBE_FCTRL_MPE;
7587 		IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
7588 
7589 		ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
7590 		ctrl |= IXGBE_CTRL_GIO_DIS;
7591 		IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
7592 
7593 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
7594 	} else {
7595 		IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
7596 		IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
7597 	}
7598 
7599 	switch (hw->mac.type) {
7600 	case ixgbe_mac_82598EB:
7601 		pci_wake_from_d3(pdev, false);
7602 		break;
7603 	case ixgbe_mac_82599EB:
7604 	case ixgbe_mac_X540:
7605 	case ixgbe_mac_X550:
7606 	case ixgbe_mac_X550EM_x:
7607 	case ixgbe_mac_x550em_a:
7608 	case ixgbe_mac_e610:
7609 		pci_wake_from_d3(pdev, !!wufc);
7610 		break;
7611 	default:
7612 		break;
7613 	}
7614 
7615 	*enable_wake = !!wufc;
7616 	if (hw->phy.ops.set_phy_power && !*enable_wake)
7617 		hw->phy.ops.set_phy_power(hw, false);
7618 
7619 	ixgbe_release_hw_control(adapter);
7620 
7621 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
7622 		pci_disable_device(pdev);
7623 
7624 	return 0;
7625 }
7626 
7627 static int ixgbe_suspend(struct device *dev_d)
7628 {
7629 	struct pci_dev *pdev = to_pci_dev(dev_d);
7630 	int retval;
7631 	bool wake;
7632 
7633 	retval = __ixgbe_shutdown(pdev, &wake);
7634 
7635 	device_set_wakeup_enable(dev_d, wake);
7636 
7637 	return retval;
7638 }
7639 
7640 static void ixgbe_shutdown(struct pci_dev *pdev)
7641 {
7642 	bool wake;
7643 
7644 	__ixgbe_shutdown(pdev, &wake);
7645 
7646 	if (system_state == SYSTEM_POWER_OFF) {
7647 		pci_wake_from_d3(pdev, wake);
7648 		pci_set_power_state(pdev, PCI_D3hot);
7649 	}
7650 }
7651 
7652 /**
7653  * ixgbe_update_stats - Update the board statistics counters.
7654  * @adapter: board private structure
7655  **/
7656 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
7657 {
7658 	struct net_device *netdev = adapter->netdev;
7659 	struct ixgbe_hw *hw = &adapter->hw;
7660 	struct ixgbe_hw_stats *hwstats = &adapter->stats;
7661 	u64 total_mpc = 0;
7662 	u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
7663 	u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
7664 	u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
7665 	u64 alloc_rx_page = 0;
7666 	u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
7667 
7668 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7669 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7670 		return;
7671 
7672 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
7673 		u64 rsc_count = 0;
7674 		u64 rsc_flush = 0;
7675 		for (i = 0; i < adapter->num_rx_queues; i++) {
7676 			rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
7677 			rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
7678 		}
7679 		adapter->rsc_total_count = rsc_count;
7680 		adapter->rsc_total_flush = rsc_flush;
7681 	}
7682 
7683 	for (i = 0; i < adapter->num_rx_queues; i++) {
7684 		struct ixgbe_ring *rx_ring = READ_ONCE(adapter->rx_ring[i]);
7685 
7686 		if (!rx_ring)
7687 			continue;
7688 		non_eop_descs += rx_ring->rx_stats.non_eop_descs;
7689 		alloc_rx_page += rx_ring->rx_stats.alloc_rx_page;
7690 		alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
7691 		alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
7692 		hw_csum_rx_error += rx_ring->rx_stats.csum_err;
7693 		bytes += rx_ring->stats.bytes;
7694 		packets += rx_ring->stats.packets;
7695 	}
7696 	adapter->non_eop_descs = non_eop_descs;
7697 	adapter->alloc_rx_page = alloc_rx_page;
7698 	adapter->alloc_rx_page_failed = alloc_rx_page_failed;
7699 	adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
7700 	adapter->hw_csum_rx_error = hw_csum_rx_error;
7701 	netdev->stats.rx_bytes = bytes;
7702 	netdev->stats.rx_packets = packets;
7703 
7704 	bytes = 0;
7705 	packets = 0;
7706 	/* gather some stats to the adapter struct that are per queue */
7707 	for (i = 0; i < adapter->num_tx_queues; i++) {
7708 		struct ixgbe_ring *tx_ring = READ_ONCE(adapter->tx_ring[i]);
7709 
7710 		if (!tx_ring)
7711 			continue;
7712 		restart_queue += tx_ring->tx_stats.restart_queue;
7713 		tx_busy += tx_ring->tx_stats.tx_busy;
7714 		bytes += tx_ring->stats.bytes;
7715 		packets += tx_ring->stats.packets;
7716 	}
7717 	for (i = 0; i < adapter->num_xdp_queues; i++) {
7718 		struct ixgbe_ring *xdp_ring = READ_ONCE(adapter->xdp_ring[i]);
7719 
7720 		if (!xdp_ring)
7721 			continue;
7722 		restart_queue += xdp_ring->tx_stats.restart_queue;
7723 		tx_busy += xdp_ring->tx_stats.tx_busy;
7724 		bytes += xdp_ring->stats.bytes;
7725 		packets += xdp_ring->stats.packets;
7726 	}
7727 	adapter->restart_queue = restart_queue;
7728 	adapter->tx_busy = tx_busy;
7729 	netdev->stats.tx_bytes = bytes;
7730 	netdev->stats.tx_packets = packets;
7731 
7732 	hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
7733 
7734 	/* 8 register reads */
7735 	for (i = 0; i < 8; i++) {
7736 		/* for packet buffers not used, the register should read 0 */
7737 		mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
7738 		missed_rx += mpc;
7739 		hwstats->mpc[i] += mpc;
7740 		total_mpc += hwstats->mpc[i];
7741 		hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
7742 		hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
7743 		switch (hw->mac.type) {
7744 		case ixgbe_mac_82598EB:
7745 			hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
7746 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
7747 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
7748 			hwstats->pxonrxc[i] +=
7749 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
7750 			break;
7751 		case ixgbe_mac_82599EB:
7752 		case ixgbe_mac_X540:
7753 		case ixgbe_mac_X550:
7754 		case ixgbe_mac_X550EM_x:
7755 		case ixgbe_mac_x550em_a:
7756 		case ixgbe_mac_e610:
7757 			hwstats->pxonrxc[i] +=
7758 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
7759 			break;
7760 		default:
7761 			break;
7762 		}
7763 	}
7764 
7765 	/*16 register reads */
7766 	for (i = 0; i < 16; i++) {
7767 		hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
7768 		hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
7769 		if (hw->mac.type == ixgbe_mac_82599EB ||
7770 		    hw->mac.type == ixgbe_mac_X540 ||
7771 		    hw->mac.type == ixgbe_mac_X550 ||
7772 		    hw->mac.type == ixgbe_mac_X550EM_x ||
7773 		    hw->mac.type == ixgbe_mac_x550em_a ||
7774 		    hw->mac.type == ixgbe_mac_e610) {
7775 			hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
7776 			IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
7777 			hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
7778 			IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
7779 		}
7780 	}
7781 
7782 	hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
7783 	/* work around hardware counting issue */
7784 	hwstats->gprc -= missed_rx;
7785 
7786 	ixgbe_update_xoff_received(adapter);
7787 
7788 	/* 82598 hardware only has a 32 bit counter in the high register */
7789 	switch (hw->mac.type) {
7790 	case ixgbe_mac_82598EB:
7791 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
7792 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
7793 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
7794 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
7795 		break;
7796 	case ixgbe_mac_X540:
7797 	case ixgbe_mac_X550:
7798 	case ixgbe_mac_X550EM_x:
7799 	case ixgbe_mac_x550em_a:
7800 	case ixgbe_mac_e610:
7801 		/* OS2BMC stats are X540 and later */
7802 		hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
7803 		hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
7804 		hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
7805 		hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
7806 		fallthrough;
7807 	case ixgbe_mac_82599EB:
7808 		for (i = 0; i < 16; i++)
7809 			adapter->hw_rx_no_dma_resources +=
7810 					     IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
7811 		hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
7812 		IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
7813 		hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
7814 		IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
7815 		hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
7816 		IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
7817 		hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
7818 		hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
7819 		hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
7820 #ifdef IXGBE_FCOE
7821 		hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
7822 		hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
7823 		hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
7824 		hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
7825 		hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
7826 		hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
7827 		/* Add up per cpu counters for total ddp aloc fail */
7828 		if (adapter->fcoe.ddp_pool) {
7829 			struct ixgbe_fcoe *fcoe = &adapter->fcoe;
7830 			struct ixgbe_fcoe_ddp_pool *ddp_pool;
7831 			unsigned int cpu;
7832 			u64 noddp = 0, noddp_ext_buff = 0;
7833 			for_each_possible_cpu(cpu) {
7834 				ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
7835 				noddp += ddp_pool->noddp;
7836 				noddp_ext_buff += ddp_pool->noddp_ext_buff;
7837 			}
7838 			hwstats->fcoe_noddp = noddp;
7839 			hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
7840 		}
7841 #endif /* IXGBE_FCOE */
7842 		break;
7843 	default:
7844 		break;
7845 	}
7846 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
7847 	hwstats->bprc += bprc;
7848 	hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
7849 	if (hw->mac.type == ixgbe_mac_82598EB)
7850 		hwstats->mprc -= bprc;
7851 	hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
7852 	hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
7853 	hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
7854 	hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
7855 	hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
7856 	hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
7857 	hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
7858 	hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
7859 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
7860 	hwstats->lxontxc += lxon;
7861 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
7862 	hwstats->lxofftxc += lxoff;
7863 	hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
7864 	hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
7865 	/*
7866 	 * 82598 errata - tx of flow control packets is included in tx counters
7867 	 */
7868 	xon_off_tot = lxon + lxoff;
7869 	hwstats->gptc -= xon_off_tot;
7870 	hwstats->mptc -= xon_off_tot;
7871 	hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
7872 	hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
7873 	hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
7874 	hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
7875 	hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
7876 	hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
7877 	hwstats->ptc64 -= xon_off_tot;
7878 	hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
7879 	hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
7880 	hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
7881 	hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
7882 	hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
7883 	hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
7884 
7885 	/* Fill out the OS statistics structure */
7886 	netdev->stats.multicast = hwstats->mprc;
7887 
7888 	/* Rx Errors */
7889 	netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
7890 	netdev->stats.rx_dropped = 0;
7891 	netdev->stats.rx_length_errors = hwstats->rlec;
7892 	netdev->stats.rx_crc_errors = hwstats->crcerrs;
7893 	netdev->stats.rx_missed_errors = total_mpc;
7894 
7895 	/* VF Stats Collection - skip while resetting because these
7896 	 * are not clear on read and otherwise you'll sometimes get
7897 	 * crazy values.
7898 	 */
7899 	if (!test_bit(__IXGBE_RESETTING, &adapter->state)) {
7900 		for (i = 0; i < adapter->num_vfs; i++) {
7901 			UPDATE_VF_COUNTER_32bit(IXGBE_PVFGPRC(i),
7902 						adapter->vfinfo[i].last_vfstats.gprc,
7903 						adapter->vfinfo[i].vfstats.gprc);
7904 			UPDATE_VF_COUNTER_32bit(IXGBE_PVFGPTC(i),
7905 						adapter->vfinfo[i].last_vfstats.gptc,
7906 						adapter->vfinfo[i].vfstats.gptc);
7907 			UPDATE_VF_COUNTER_36bit(IXGBE_PVFGORC_LSB(i),
7908 						IXGBE_PVFGORC_MSB(i),
7909 						adapter->vfinfo[i].last_vfstats.gorc,
7910 						adapter->vfinfo[i].vfstats.gorc);
7911 			UPDATE_VF_COUNTER_36bit(IXGBE_PVFGOTC_LSB(i),
7912 						IXGBE_PVFGOTC_MSB(i),
7913 						adapter->vfinfo[i].last_vfstats.gotc,
7914 						adapter->vfinfo[i].vfstats.gotc);
7915 			UPDATE_VF_COUNTER_32bit(IXGBE_PVFMPRC(i),
7916 						adapter->vfinfo[i].last_vfstats.mprc,
7917 						adapter->vfinfo[i].vfstats.mprc);
7918 		}
7919 	}
7920 }
7921 
7922 /**
7923  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
7924  * @adapter: pointer to the device adapter structure
7925  **/
7926 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
7927 {
7928 	struct ixgbe_hw *hw = &adapter->hw;
7929 	int i;
7930 
7931 	if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
7932 		return;
7933 
7934 	adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
7935 
7936 	/* if interface is down do nothing */
7937 	if (test_bit(__IXGBE_DOWN, &adapter->state))
7938 		return;
7939 
7940 	/* do nothing if we are not using signature filters */
7941 	if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
7942 		return;
7943 
7944 	adapter->fdir_overflow++;
7945 
7946 	if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
7947 		for (i = 0; i < adapter->num_tx_queues; i++)
7948 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7949 				&(adapter->tx_ring[i]->state));
7950 		for (i = 0; i < adapter->num_xdp_queues; i++)
7951 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7952 				&adapter->xdp_ring[i]->state);
7953 		/* re-enable flow director interrupts */
7954 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
7955 	} else {
7956 		e_err(probe, "failed to finish FDIR re-initialization, "
7957 		      "ignored adding FDIR ATR filters\n");
7958 	}
7959 }
7960 
7961 /**
7962  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
7963  * @adapter: pointer to the device adapter structure
7964  *
7965  * This function serves two purposes.  First it strobes the interrupt lines
7966  * in order to make certain interrupts are occurring.  Secondly it sets the
7967  * bits needed to check for TX hangs.  As a result we should immediately
7968  * determine if a hang has occurred.
7969  */
7970 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
7971 {
7972 	struct ixgbe_hw *hw = &adapter->hw;
7973 	u64 eics = 0;
7974 	int i;
7975 
7976 	/* If we're down, removing or resetting, just bail */
7977 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7978 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
7979 	    test_bit(__IXGBE_RESETTING, &adapter->state))
7980 		return;
7981 
7982 	/* Force detection of hung controller */
7983 	if (netif_carrier_ok(adapter->netdev))
7984 		for (i = 0; i < adapter->num_tx_queues; i++)
7985 			set_check_for_tx_hang(adapter->tx_ring[i]);
7986 
7987 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
7988 		/*
7989 		 * for legacy and MSI interrupts don't set any bits
7990 		 * that are enabled for EIAM, because this operation
7991 		 * would set *both* EIMS and EICS for any bit in EIAM
7992 		 */
7993 		IXGBE_WRITE_REG(hw, IXGBE_EICS,
7994 			(IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
7995 	} else {
7996 		/* get one bit for every active tx/rx interrupt vector */
7997 		for (i = 0; i < adapter->num_q_vectors; i++) {
7998 			struct ixgbe_q_vector *qv = adapter->q_vector[i];
7999 			if (qv->rx.ring || qv->tx.ring)
8000 				eics |= BIT_ULL(i);
8001 		}
8002 	}
8003 
8004 	/* Cause software interrupt to ensure rings are cleaned */
8005 	ixgbe_irq_rearm_queues(adapter, eics);
8006 }
8007 
8008 /**
8009  * ixgbe_watchdog_update_link - update the link status
8010  * @adapter: pointer to the device adapter structure
8011  **/
8012 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
8013 {
8014 	struct ixgbe_hw *hw = &adapter->hw;
8015 	u32 link_speed = adapter->link_speed;
8016 	bool link_up = adapter->link_up;
8017 	bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
8018 
8019 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
8020 		return;
8021 
8022 	if (hw->mac.ops.check_link) {
8023 		hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
8024 	} else {
8025 		/* always assume link is up, if no check link function */
8026 		link_speed = IXGBE_LINK_SPEED_10GB_FULL;
8027 		link_up = true;
8028 	}
8029 
8030 	if (adapter->ixgbe_ieee_pfc)
8031 		pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
8032 
8033 	if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
8034 		hw->mac.ops.fc_enable(hw);
8035 		ixgbe_set_rx_drop_en(adapter);
8036 	}
8037 
8038 	if (link_up ||
8039 	    time_after(jiffies, (adapter->link_check_timeout +
8040 				 IXGBE_TRY_LINK_TIMEOUT))) {
8041 		adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
8042 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
8043 		IXGBE_WRITE_FLUSH(hw);
8044 	}
8045 
8046 	adapter->link_up = link_up;
8047 	adapter->link_speed = link_speed;
8048 }
8049 
8050 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
8051 {
8052 #ifdef CONFIG_IXGBE_DCB
8053 	struct net_device *netdev = adapter->netdev;
8054 	struct dcb_app app = {
8055 			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
8056 			      .protocol = 0,
8057 			     };
8058 	u8 up = 0;
8059 
8060 	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
8061 		up = dcb_ieee_getapp_mask(netdev, &app);
8062 
8063 	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
8064 #endif
8065 }
8066 
8067 /**
8068  * ixgbe_watchdog_link_is_up - update netif_carrier status and
8069  *                             print link up message
8070  * @adapter: pointer to the device adapter structure
8071  **/
8072 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
8073 {
8074 	struct net_device *netdev = adapter->netdev;
8075 	struct ixgbe_hw *hw = &adapter->hw;
8076 	u32 link_speed = adapter->link_speed;
8077 	const char *speed_str;
8078 	bool flow_rx, flow_tx;
8079 
8080 	/* only continue if link was previously down */
8081 	if (netif_carrier_ok(netdev))
8082 		return;
8083 
8084 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
8085 
8086 	switch (hw->mac.type) {
8087 	case ixgbe_mac_82598EB: {
8088 		u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
8089 		u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
8090 		flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
8091 		flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
8092 	}
8093 		break;
8094 	case ixgbe_mac_X540:
8095 	case ixgbe_mac_X550:
8096 	case ixgbe_mac_X550EM_x:
8097 	case ixgbe_mac_x550em_a:
8098 	case ixgbe_mac_e610:
8099 	case ixgbe_mac_82599EB: {
8100 		u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
8101 		u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
8102 		flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
8103 		flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
8104 	}
8105 		break;
8106 	default:
8107 		flow_tx = false;
8108 		flow_rx = false;
8109 		break;
8110 	}
8111 
8112 	adapter->last_rx_ptp_check = jiffies;
8113 
8114 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
8115 		ixgbe_ptp_start_cyclecounter(adapter);
8116 
8117 	switch (link_speed) {
8118 	case IXGBE_LINK_SPEED_10GB_FULL:
8119 		speed_str = "10 Gbps";
8120 		break;
8121 	case IXGBE_LINK_SPEED_5GB_FULL:
8122 		speed_str = "5 Gbps";
8123 		break;
8124 	case IXGBE_LINK_SPEED_2_5GB_FULL:
8125 		speed_str = "2.5 Gbps";
8126 		break;
8127 	case IXGBE_LINK_SPEED_1GB_FULL:
8128 		speed_str = "1 Gbps";
8129 		break;
8130 	case IXGBE_LINK_SPEED_100_FULL:
8131 		speed_str = "100 Mbps";
8132 		break;
8133 	case IXGBE_LINK_SPEED_10_FULL:
8134 		speed_str = "10 Mbps";
8135 		break;
8136 	default:
8137 		speed_str = "unknown speed";
8138 		break;
8139 	}
8140 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
8141 	       ((flow_rx && flow_tx) ? "RX/TX" :
8142 	       (flow_rx ? "RX" :
8143 	       (flow_tx ? "TX" : "None"))));
8144 
8145 	netif_carrier_on(netdev);
8146 	ixgbe_check_vf_rate_limit(adapter);
8147 
8148 	if (adapter->num_vfs && hw->mac.ops.enable_mdd)
8149 		hw->mac.ops.enable_mdd(hw);
8150 
8151 	/* enable transmits */
8152 	netif_tx_wake_all_queues(adapter->netdev);
8153 
8154 	/* update the default user priority for VFs */
8155 	ixgbe_update_default_up(adapter);
8156 
8157 	/* ping all the active vfs to let them know link has changed */
8158 	ixgbe_ping_all_vfs(adapter);
8159 }
8160 
8161 /**
8162  * ixgbe_watchdog_link_is_down - update netif_carrier status and
8163  *                               print link down message
8164  * @adapter: pointer to the adapter structure
8165  **/
8166 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
8167 {
8168 	struct net_device *netdev = adapter->netdev;
8169 	struct ixgbe_hw *hw = &adapter->hw;
8170 
8171 	adapter->link_up = false;
8172 	adapter->link_speed = 0;
8173 
8174 	/* only continue if link was up previously */
8175 	if (!netif_carrier_ok(netdev))
8176 		return;
8177 
8178 	adapter->link_down_events++;
8179 
8180 	/* poll for SFP+ cable when link is down */
8181 	if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
8182 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
8183 
8184 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
8185 		ixgbe_ptp_start_cyclecounter(adapter);
8186 
8187 	e_info(drv, "NIC Link is Down\n");
8188 	netif_carrier_off(netdev);
8189 
8190 	/* ping all the active vfs to let them know link has changed */
8191 	ixgbe_ping_all_vfs(adapter);
8192 }
8193 
8194 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
8195 {
8196 	int i;
8197 
8198 	for (i = 0; i < adapter->num_tx_queues; i++) {
8199 		struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
8200 
8201 		if (tx_ring->next_to_use != tx_ring->next_to_clean)
8202 			return true;
8203 	}
8204 
8205 	return false;
8206 }
8207 
8208 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
8209 {
8210 	struct ixgbe_hw *hw = &adapter->hw;
8211 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
8212 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
8213 
8214 	int i, j;
8215 
8216 	if (!adapter->num_vfs)
8217 		return false;
8218 
8219 	/* resetting the PF is only needed for MAC before X550 */
8220 	if (hw->mac.type >= ixgbe_mac_X550)
8221 		return false;
8222 
8223 	for (i = 0; i < adapter->num_vfs; i++) {
8224 		for (j = 0; j < q_per_pool; j++) {
8225 			u32 h, t;
8226 
8227 			h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
8228 			t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
8229 
8230 			if (h != t)
8231 				return true;
8232 		}
8233 	}
8234 
8235 	return false;
8236 }
8237 
8238 /**
8239  * ixgbe_watchdog_flush_tx - flush queues on link down
8240  * @adapter: pointer to the device adapter structure
8241  **/
8242 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
8243 {
8244 	if (!netif_carrier_ok(adapter->netdev)) {
8245 		if (ixgbe_ring_tx_pending(adapter) ||
8246 		    ixgbe_vf_tx_pending(adapter)) {
8247 			/* We've lost link, so the controller stops DMA,
8248 			 * but we've got queued Tx work that's never going
8249 			 * to get done, so reset controller to flush Tx.
8250 			 * (Do the reset outside of interrupt context).
8251 			 */
8252 			e_warn(drv, "initiating reset to clear Tx work after link loss\n");
8253 			set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
8254 		}
8255 	}
8256 }
8257 
8258 #ifdef CONFIG_PCI_IOV
8259 static void ixgbe_bad_vf_abort(struct ixgbe_adapter *adapter, u32 vf)
8260 {
8261 	struct ixgbe_hw *hw = &adapter->hw;
8262 
8263 	if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
8264 	    adapter->flags2 & IXGBE_FLAG2_AUTO_DISABLE_VF) {
8265 		adapter->vfinfo[vf].primary_abort_count++;
8266 		if (adapter->vfinfo[vf].primary_abort_count ==
8267 		    IXGBE_PRIMARY_ABORT_LIMIT) {
8268 			ixgbe_set_vf_link_state(adapter, vf,
8269 						IFLA_VF_LINK_STATE_DISABLE);
8270 			adapter->vfinfo[vf].primary_abort_count = 0;
8271 
8272 			e_info(drv,
8273 			       "Malicious Driver Detection event detected on PF %d VF %d MAC: %pM mdd-disable-vf=on",
8274 			       hw->bus.func, vf,
8275 			       adapter->vfinfo[vf].vf_mac_addresses);
8276 		}
8277 	}
8278 }
8279 
8280 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
8281 {
8282 	struct ixgbe_hw *hw = &adapter->hw;
8283 	struct pci_dev *pdev = adapter->pdev;
8284 	unsigned int vf;
8285 	u32 gpc;
8286 
8287 	if (!(netif_carrier_ok(adapter->netdev)))
8288 		return;
8289 
8290 	gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
8291 	if (gpc) /* If incrementing then no need for the check below */
8292 		return;
8293 	/* Check to see if a bad DMA write target from an errant or
8294 	 * malicious VF has caused a PCIe error.  If so then we can
8295 	 * issue a VFLR to the offending VF(s) and then resume without
8296 	 * requesting a full slot reset.
8297 	 */
8298 
8299 	if (!pdev)
8300 		return;
8301 
8302 	/* check status reg for all VFs owned by this PF */
8303 	for (vf = 0; vf < adapter->num_vfs; ++vf) {
8304 		struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
8305 		u16 status_reg;
8306 
8307 		if (!vfdev)
8308 			continue;
8309 		pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
8310 		if (status_reg != IXGBE_FAILED_READ_CFG_WORD &&
8311 		    status_reg & PCI_STATUS_REC_MASTER_ABORT) {
8312 			ixgbe_bad_vf_abort(adapter, vf);
8313 			pcie_flr(vfdev);
8314 		}
8315 	}
8316 }
8317 
8318 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
8319 {
8320 	u32 ssvpc;
8321 
8322 	/* Do not perform spoof check for 82598 or if not in IOV mode */
8323 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
8324 	    adapter->num_vfs == 0)
8325 		return;
8326 
8327 	ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
8328 
8329 	/*
8330 	 * ssvpc register is cleared on read, if zero then no
8331 	 * spoofed packets in the last interval.
8332 	 */
8333 	if (!ssvpc)
8334 		return;
8335 
8336 	e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
8337 }
8338 #else
8339 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
8340 {
8341 }
8342 
8343 static void
8344 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
8345 {
8346 }
8347 #endif /* CONFIG_PCI_IOV */
8348 
8349 
8350 /**
8351  * ixgbe_watchdog_subtask - check and bring link up
8352  * @adapter: pointer to the device adapter structure
8353  **/
8354 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
8355 {
8356 	/* if interface is down, removing or resetting, do nothing */
8357 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
8358 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
8359 	    test_bit(__IXGBE_RESETTING, &adapter->state))
8360 		return;
8361 
8362 	ixgbe_watchdog_update_link(adapter);
8363 
8364 	if (adapter->link_up)
8365 		ixgbe_watchdog_link_is_up(adapter);
8366 	else
8367 		ixgbe_watchdog_link_is_down(adapter);
8368 
8369 	ixgbe_check_for_bad_vf(adapter);
8370 	ixgbe_spoof_check(adapter);
8371 	ixgbe_update_stats(adapter);
8372 
8373 	ixgbe_watchdog_flush_tx(adapter);
8374 }
8375 
8376 /**
8377  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
8378  * @adapter: the ixgbe adapter structure
8379  **/
8380 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
8381 {
8382 	struct ixgbe_hw *hw = &adapter->hw;
8383 	int err;
8384 
8385 	/* not searching for SFP so there is nothing to do here */
8386 	if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
8387 	    !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
8388 		return;
8389 
8390 	if (adapter->sfp_poll_time &&
8391 	    time_after(adapter->sfp_poll_time, jiffies))
8392 		return; /* If not yet time to poll for SFP */
8393 
8394 	/* someone else is in init, wait until next service event */
8395 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
8396 		return;
8397 
8398 	adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
8399 
8400 	err = hw->phy.ops.identify_sfp(hw);
8401 	if (err == -EOPNOTSUPP)
8402 		goto sfp_out;
8403 
8404 	if (err == -ENOENT) {
8405 		/* If no cable is present, then we need to reset
8406 		 * the next time we find a good cable. */
8407 		adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
8408 	}
8409 
8410 	/* exit on error */
8411 	if (err)
8412 		goto sfp_out;
8413 
8414 	/* exit if reset not needed */
8415 	if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
8416 		goto sfp_out;
8417 
8418 	adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
8419 
8420 	/*
8421 	 * A module may be identified correctly, but the EEPROM may not have
8422 	 * support for that module.  setup_sfp() will fail in that case, so
8423 	 * we should not allow that module to load.
8424 	 */
8425 	if (hw->mac.type == ixgbe_mac_82598EB)
8426 		err = hw->phy.ops.reset(hw);
8427 	else
8428 		err = hw->mac.ops.setup_sfp(hw);
8429 
8430 	if (err == -EOPNOTSUPP)
8431 		goto sfp_out;
8432 
8433 	adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
8434 	e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
8435 
8436 sfp_out:
8437 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
8438 
8439 	if (err == -EOPNOTSUPP &&
8440 	    adapter->netdev->reg_state == NETREG_REGISTERED) {
8441 		e_dev_err("failed to initialize because an unsupported "
8442 			  "SFP+ module type was detected.\n");
8443 		e_dev_err("Reload the driver after installing a "
8444 			  "supported module.\n");
8445 		unregister_netdev(adapter->netdev);
8446 	}
8447 }
8448 
8449 /**
8450  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
8451  * @adapter: the ixgbe adapter structure
8452  **/
8453 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
8454 {
8455 	struct ixgbe_hw *hw = &adapter->hw;
8456 	u32 cap_speed;
8457 	u32 speed;
8458 	bool autoneg = false;
8459 
8460 	if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
8461 		return;
8462 
8463 	/* someone else is in init, wait until next service event */
8464 	if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
8465 		return;
8466 
8467 	adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
8468 
8469 	hw->mac.ops.get_link_capabilities(hw, &cap_speed, &autoneg);
8470 
8471 	/* advertise highest capable link speed */
8472 	if (!autoneg && (cap_speed & IXGBE_LINK_SPEED_10GB_FULL))
8473 		speed = IXGBE_LINK_SPEED_10GB_FULL;
8474 	else
8475 		speed = cap_speed & (IXGBE_LINK_SPEED_10GB_FULL |
8476 				     IXGBE_LINK_SPEED_1GB_FULL);
8477 
8478 	if (hw->mac.ops.setup_link)
8479 		hw->mac.ops.setup_link(hw, speed, true);
8480 
8481 	adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
8482 	adapter->link_check_timeout = jiffies;
8483 	clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
8484 }
8485 
8486 /**
8487  * ixgbe_service_timer - Timer Call-back
8488  * @t: pointer to timer_list structure
8489  **/
8490 static void ixgbe_service_timer(struct timer_list *t)
8491 {
8492 	struct ixgbe_adapter *adapter = timer_container_of(adapter, t,
8493 							   service_timer);
8494 	unsigned long next_event_offset;
8495 
8496 	/* poll faster when waiting for link */
8497 	if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
8498 		next_event_offset = HZ / 10;
8499 	else
8500 		next_event_offset = HZ * 2;
8501 
8502 	/* Reset the timer */
8503 	mod_timer(&adapter->service_timer, next_event_offset + jiffies);
8504 
8505 	ixgbe_service_event_schedule(adapter);
8506 }
8507 
8508 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
8509 {
8510 	struct ixgbe_hw *hw = &adapter->hw;
8511 	bool overtemp;
8512 
8513 	if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
8514 		return;
8515 
8516 	adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
8517 
8518 	if (!hw->phy.ops.handle_lasi)
8519 		return;
8520 
8521 	hw->phy.ops.handle_lasi(&adapter->hw, &overtemp);
8522 	if (overtemp)
8523 		e_crit(drv, "%s\n", ixgbe_overheat_msg);
8524 }
8525 
8526 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
8527 {
8528 	if (!test_and_clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state))
8529 		return;
8530 
8531 	rtnl_lock();
8532 	/* If we're already down, removing or resetting, just bail */
8533 	if (test_bit(__IXGBE_DOWN, &adapter->state) ||
8534 	    test_bit(__IXGBE_REMOVING, &adapter->state) ||
8535 	    test_bit(__IXGBE_RESETTING, &adapter->state)) {
8536 		rtnl_unlock();
8537 		return;
8538 	}
8539 
8540 	ixgbe_dump(adapter);
8541 	netdev_err(adapter->netdev, "Reset adapter\n");
8542 	adapter->tx_timeout_count++;
8543 
8544 	ixgbe_reinit_locked(adapter);
8545 	rtnl_unlock();
8546 }
8547 
8548 static int ixgbe_check_fw_api_mismatch(struct ixgbe_adapter *adapter)
8549 {
8550 	struct ixgbe_hw *hw = &adapter->hw;
8551 
8552 	if (hw->mac.type != ixgbe_mac_e610)
8553 		return 0;
8554 
8555 	if (hw->mac.ops.get_fw_ver && hw->mac.ops.get_fw_ver(hw))
8556 		return 0;
8557 
8558 	if (hw->api_maj_ver > IXGBE_FW_API_VER_MAJOR) {
8559 		e_dev_err("The driver for the device stopped because the NVM image is newer than expected. You must install the most recent version of the network driver.\n");
8560 
8561 		adapter->flags2 |= IXGBE_FLAG2_API_MISMATCH;
8562 		return -EOPNOTSUPP;
8563 	} else if (hw->api_maj_ver == IXGBE_FW_API_VER_MAJOR &&
8564 		   hw->api_min_ver > IXGBE_FW_API_VER_MINOR + IXGBE_FW_API_VER_DIFF_ALLOWED) {
8565 		e_dev_info("The driver for the device detected a newer version of the NVM image than expected. Please install the most recent version of the network driver.\n");
8566 		adapter->flags2 |= IXGBE_FLAG2_API_MISMATCH;
8567 	} else if (hw->api_maj_ver < IXGBE_FW_API_VER_MAJOR ||
8568 		   hw->api_min_ver < IXGBE_FW_API_VER_MINOR - IXGBE_FW_API_VER_DIFF_ALLOWED) {
8569 		e_dev_info("The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
8570 		adapter->flags2 |= IXGBE_FLAG2_API_MISMATCH;
8571 	}
8572 
8573 	return 0;
8574 }
8575 
8576 /**
8577  * ixgbe_check_fw_error - Check firmware for errors
8578  * @adapter: the adapter private structure
8579  *
8580  * Check firmware errors in register FWSM
8581  */
8582 static bool ixgbe_check_fw_error(struct ixgbe_adapter *adapter)
8583 {
8584 	struct ixgbe_hw *hw = &adapter->hw;
8585 	u32 fwsm;
8586 	int err;
8587 
8588 	/* read fwsm.ext_err_ind register and log errors */
8589 	fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
8590 
8591 	/* skip if E610's FW is reloading, warning in that case may be misleading */
8592 	if (fwsm & IXGBE_FWSM_EXT_ERR_IND_MASK ||
8593 	    (!(fwsm & IXGBE_FWSM_FW_VAL_BIT) && !(hw->mac.type == ixgbe_mac_e610)))
8594 		e_dev_warn("Warning firmware error detected FWSM: 0x%08X\n",
8595 			   fwsm);
8596 
8597 	if (hw->mac.ops.fw_recovery_mode && hw->mac.ops.fw_recovery_mode(hw)) {
8598 		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");
8599 		return true;
8600 	}
8601 	if (!(adapter->flags2 & IXGBE_FLAG2_API_MISMATCH)) {
8602 		err = ixgbe_check_fw_api_mismatch(adapter);
8603 		if (err)
8604 			return true;
8605 	}
8606 
8607 	/* return here if FW rollback mode has been already detected */
8608 	if (adapter->flags2 & IXGBE_FLAG2_FW_ROLLBACK)
8609 		return false;
8610 
8611 	if (hw->mac.ops.fw_rollback_mode && hw->mac.ops.fw_rollback_mode(hw)) {
8612 		struct ixgbe_nvm_info *nvm_info = &adapter->hw.flash.nvm;
8613 		char ver_buff[64] = "";
8614 
8615 		if (hw->mac.ops.get_fw_ver && hw->mac.ops.get_fw_ver(hw))
8616 			goto no_version;
8617 
8618 		if (hw->mac.ops.get_nvm_ver &&
8619 		    hw->mac.ops.get_nvm_ver(hw, nvm_info))
8620 			goto no_version;
8621 
8622 		snprintf(ver_buff, sizeof(ver_buff),
8623 			 "Current version is NVM:%x.%x.%x, FW:%d.%d. ",
8624 			 nvm_info->major, nvm_info->minor, nvm_info->eetrack,
8625 			 hw->fw_maj_ver, hw->fw_maj_ver);
8626 no_version:
8627 		e_dev_warn("Firmware rollback mode detected. %sDevice may exhibit limited functionality. Refer to the Intel(R) Ethernet Adapters and Devices User Guide for details on firmware rollback mode.",
8628 			   ver_buff);
8629 
8630 		adapter->flags2 |= IXGBE_FLAG2_FW_ROLLBACK;
8631 	}
8632 
8633 	return false;
8634 }
8635 
8636 static void ixgbe_recovery_service_task(struct work_struct *work)
8637 {
8638 	struct ixgbe_adapter *adapter = container_of(work,
8639 						     struct ixgbe_adapter,
8640 						     service_task);
8641 
8642 	ixgbe_handle_fw_event(adapter);
8643 	ixgbe_service_event_complete(adapter);
8644 
8645 	mod_timer(&adapter->service_timer, jiffies + msecs_to_jiffies(100));
8646 }
8647 
8648 /**
8649  * ixgbe_service_task - manages and runs subtasks
8650  * @work: pointer to work_struct containing our data
8651  **/
8652 static void ixgbe_service_task(struct work_struct *work)
8653 {
8654 	struct ixgbe_adapter *adapter = container_of(work,
8655 						     struct ixgbe_adapter,
8656 						     service_task);
8657 	if (ixgbe_removed(adapter->hw.hw_addr)) {
8658 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
8659 			rtnl_lock();
8660 			ixgbe_down(adapter);
8661 			rtnl_unlock();
8662 		}
8663 		ixgbe_service_event_complete(adapter);
8664 		return;
8665 	}
8666 	if (ixgbe_check_fw_error(adapter)) {
8667 		if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
8668 			if (adapter->mii_bus) {
8669 				mdiobus_unregister(adapter->mii_bus);
8670 				adapter->mii_bus = NULL;
8671 			}
8672 			unregister_netdev(adapter->netdev);
8673 		}
8674 		ixgbe_service_event_complete(adapter);
8675 		return;
8676 	}
8677 	if (adapter->hw.mac.type == ixgbe_mac_e610) {
8678 		if (adapter->flags2 & IXGBE_FLAG2_FW_ASYNC_EVENT)
8679 			ixgbe_handle_fw_event(adapter);
8680 		ixgbe_check_media_subtask(adapter);
8681 	}
8682 	ixgbe_reset_subtask(adapter);
8683 	ixgbe_phy_interrupt_subtask(adapter);
8684 	ixgbe_sfp_detection_subtask(adapter);
8685 	ixgbe_sfp_link_config_subtask(adapter);
8686 	ixgbe_check_overtemp_subtask(adapter);
8687 	ixgbe_watchdog_subtask(adapter);
8688 	ixgbe_fdir_reinit_subtask(adapter);
8689 	ixgbe_check_hang_subtask(adapter);
8690 
8691 	if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
8692 		ixgbe_ptp_overflow_check(adapter);
8693 		if (adapter->flags & IXGBE_FLAG_RX_HWTSTAMP_IN_REGISTER)
8694 			ixgbe_ptp_rx_hang(adapter);
8695 		ixgbe_ptp_tx_hang(adapter);
8696 	}
8697 
8698 	ixgbe_service_event_complete(adapter);
8699 }
8700 
8701 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
8702 		     struct ixgbe_tx_buffer *first,
8703 		     u8 *hdr_len,
8704 		     struct ixgbe_ipsec_tx_data *itd)
8705 {
8706 	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
8707 	struct sk_buff *skb = first->skb;
8708 	union {
8709 		struct iphdr *v4;
8710 		struct ipv6hdr *v6;
8711 		unsigned char *hdr;
8712 	} ip;
8713 	union {
8714 		struct tcphdr *tcp;
8715 		struct udphdr *udp;
8716 		unsigned char *hdr;
8717 	} l4;
8718 	u32 paylen, l4_offset;
8719 	u32 fceof_saidx = 0;
8720 	int err;
8721 
8722 	if (skb->ip_summed != CHECKSUM_PARTIAL)
8723 		return 0;
8724 
8725 	if (!skb_is_gso(skb))
8726 		return 0;
8727 
8728 	err = skb_cow_head(skb, 0);
8729 	if (err < 0)
8730 		return err;
8731 
8732 	if (eth_p_mpls(first->protocol))
8733 		ip.hdr = skb_inner_network_header(skb);
8734 	else
8735 		ip.hdr = skb_network_header(skb);
8736 	l4.hdr = skb_checksum_start(skb);
8737 
8738 	/* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
8739 	type_tucmd = (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_L4) ?
8740 		      IXGBE_ADVTXD_TUCMD_L4T_UDP : IXGBE_ADVTXD_TUCMD_L4T_TCP;
8741 
8742 	/* initialize outer IP header fields */
8743 	if (ip.v4->version == 4) {
8744 		unsigned char *csum_start = skb_checksum_start(skb);
8745 		unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
8746 		int len = csum_start - trans_start;
8747 
8748 		/* IP header will have to cancel out any data that
8749 		 * is not a part of the outer IP header, so set to
8750 		 * a reverse csum if needed, else init check to 0.
8751 		 */
8752 		ip.v4->check = (skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL) ?
8753 					   csum_fold(csum_partial(trans_start,
8754 								  len, 0)) : 0;
8755 		type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
8756 
8757 		ip.v4->tot_len = 0;
8758 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
8759 				   IXGBE_TX_FLAGS_CSUM |
8760 				   IXGBE_TX_FLAGS_IPV4;
8761 	} else {
8762 		ip.v6->payload_len = 0;
8763 		first->tx_flags |= IXGBE_TX_FLAGS_TSO |
8764 				   IXGBE_TX_FLAGS_CSUM;
8765 	}
8766 
8767 	/* determine offset of inner transport header */
8768 	l4_offset = l4.hdr - skb->data;
8769 
8770 	/* remove payload length from inner checksum */
8771 	paylen = skb->len - l4_offset;
8772 
8773 	if (type_tucmd & IXGBE_ADVTXD_TUCMD_L4T_TCP) {
8774 		/* compute length of segmentation header */
8775 		*hdr_len = (l4.tcp->doff * 4) + l4_offset;
8776 		csum_replace_by_diff(&l4.tcp->check,
8777 				     (__force __wsum)htonl(paylen));
8778 	} else {
8779 		/* compute length of segmentation header */
8780 		*hdr_len = sizeof(*l4.udp) + l4_offset;
8781 		csum_replace_by_diff(&l4.udp->check,
8782 				     (__force __wsum)htonl(paylen));
8783 	}
8784 
8785 	/* update gso size and bytecount with header size */
8786 	first->gso_segs = skb_shinfo(skb)->gso_segs;
8787 	first->bytecount += (first->gso_segs - 1) * *hdr_len;
8788 
8789 	/* mss_l4len_id: use 0 as index for TSO */
8790 	mss_l4len_idx = (*hdr_len - l4_offset) << IXGBE_ADVTXD_L4LEN_SHIFT;
8791 	mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
8792 
8793 	fceof_saidx |= itd->sa_idx;
8794 	type_tucmd |= itd->flags | itd->trailer_len;
8795 
8796 	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
8797 	vlan_macip_lens = l4.hdr - ip.hdr;
8798 	vlan_macip_lens |= (ip.hdr - skb->data) << IXGBE_ADVTXD_MACLEN_SHIFT;
8799 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
8800 
8801 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd,
8802 			  mss_l4len_idx);
8803 
8804 	return 1;
8805 }
8806 
8807 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
8808 			  struct ixgbe_tx_buffer *first,
8809 			  struct ixgbe_ipsec_tx_data *itd)
8810 {
8811 	struct sk_buff *skb = first->skb;
8812 	u32 vlan_macip_lens = 0;
8813 	u32 fceof_saidx = 0;
8814 	u32 type_tucmd = 0;
8815 
8816 	if (skb->ip_summed != CHECKSUM_PARTIAL) {
8817 csum_failed:
8818 		if (!(first->tx_flags & (IXGBE_TX_FLAGS_HW_VLAN |
8819 					 IXGBE_TX_FLAGS_CC)))
8820 			return;
8821 		goto no_csum;
8822 	}
8823 
8824 	switch (skb->csum_offset) {
8825 	case offsetof(struct tcphdr, check):
8826 		type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
8827 		fallthrough;
8828 	case offsetof(struct udphdr, check):
8829 		break;
8830 	case offsetof(struct sctphdr, checksum):
8831 		/* validate that this is actually an SCTP request */
8832 		if (skb_csum_is_sctp(skb)) {
8833 			type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_SCTP;
8834 			break;
8835 		}
8836 		fallthrough;
8837 	default:
8838 		skb_checksum_help(skb);
8839 		goto csum_failed;
8840 	}
8841 
8842 	/* update TX checksum flag */
8843 	first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
8844 	vlan_macip_lens = skb_checksum_start_offset(skb) -
8845 			  skb_network_offset(skb);
8846 no_csum:
8847 	/* vlan_macip_lens: MACLEN, VLAN tag */
8848 	vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
8849 	vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
8850 
8851 	fceof_saidx |= itd->sa_idx;
8852 	type_tucmd |= itd->flags | itd->trailer_len;
8853 
8854 	ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd, 0);
8855 }
8856 
8857 #define IXGBE_SET_FLAG(_input, _flag, _result) \
8858 	((_flag <= _result) ? \
8859 	 ((u32)(_input & _flag) * (_result / _flag)) : \
8860 	 ((u32)(_input & _flag) / (_flag / _result)))
8861 
8862 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
8863 {
8864 	/* set type for advanced descriptor with frame checksum insertion */
8865 	u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
8866 		       IXGBE_ADVTXD_DCMD_DEXT |
8867 		       IXGBE_ADVTXD_DCMD_IFCS;
8868 
8869 	/* set HW vlan bit if vlan is present */
8870 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
8871 				   IXGBE_ADVTXD_DCMD_VLE);
8872 
8873 	/* set segmentation enable bits for TSO/FSO */
8874 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
8875 				   IXGBE_ADVTXD_DCMD_TSE);
8876 
8877 	/* set timestamp bit if present */
8878 	cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
8879 				   IXGBE_ADVTXD_MAC_TSTAMP);
8880 
8881 	/* insert frame checksum */
8882 	cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
8883 
8884 	return cmd_type;
8885 }
8886 
8887 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
8888 				   u32 tx_flags, unsigned int paylen)
8889 {
8890 	u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
8891 
8892 	/* enable L4 checksum for TSO and TX checksum offload */
8893 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8894 					IXGBE_TX_FLAGS_CSUM,
8895 					IXGBE_ADVTXD_POPTS_TXSM);
8896 
8897 	/* enable IPv4 checksum for TSO */
8898 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8899 					IXGBE_TX_FLAGS_IPV4,
8900 					IXGBE_ADVTXD_POPTS_IXSM);
8901 
8902 	/* enable IPsec */
8903 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8904 					IXGBE_TX_FLAGS_IPSEC,
8905 					IXGBE_ADVTXD_POPTS_IPSEC);
8906 
8907 	/*
8908 	 * Check Context must be set if Tx switch is enabled, which it
8909 	 * always is for case where virtual functions are running
8910 	 */
8911 	olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8912 					IXGBE_TX_FLAGS_CC,
8913 					IXGBE_ADVTXD_CC);
8914 
8915 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
8916 }
8917 
8918 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
8919 {
8920 	if (!netif_subqueue_try_stop(tx_ring->netdev, tx_ring->queue_index,
8921 				     ixgbe_desc_unused(tx_ring), size))
8922 		return -EBUSY;
8923 
8924 	++tx_ring->tx_stats.restart_queue;
8925 	return 0;
8926 }
8927 
8928 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
8929 {
8930 	if (likely(ixgbe_desc_unused(tx_ring) >= size))
8931 		return 0;
8932 
8933 	return __ixgbe_maybe_stop_tx(tx_ring, size);
8934 }
8935 
8936 static int ixgbe_tx_map(struct ixgbe_ring *tx_ring,
8937 			struct ixgbe_tx_buffer *first,
8938 			const u8 hdr_len)
8939 {
8940 	struct sk_buff *skb = first->skb;
8941 	struct ixgbe_tx_buffer *tx_buffer;
8942 	union ixgbe_adv_tx_desc *tx_desc;
8943 	skb_frag_t *frag;
8944 	dma_addr_t dma;
8945 	unsigned int data_len, size;
8946 	u32 tx_flags = first->tx_flags;
8947 	u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
8948 	u16 i = tx_ring->next_to_use;
8949 
8950 	tx_desc = IXGBE_TX_DESC(tx_ring, i);
8951 
8952 	ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
8953 
8954 	size = skb_headlen(skb);
8955 	data_len = skb->data_len;
8956 
8957 #ifdef IXGBE_FCOE
8958 	if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
8959 		if (data_len < sizeof(struct fcoe_crc_eof)) {
8960 			size -= sizeof(struct fcoe_crc_eof) - data_len;
8961 			data_len = 0;
8962 		} else {
8963 			data_len -= sizeof(struct fcoe_crc_eof);
8964 		}
8965 	}
8966 
8967 #endif
8968 	dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
8969 
8970 	tx_buffer = first;
8971 
8972 	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
8973 		if (dma_mapping_error(tx_ring->dev, dma))
8974 			goto dma_error;
8975 
8976 		/* record length, and DMA address */
8977 		dma_unmap_len_set(tx_buffer, len, size);
8978 		dma_unmap_addr_set(tx_buffer, dma, dma);
8979 
8980 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
8981 
8982 		while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
8983 			tx_desc->read.cmd_type_len =
8984 				cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
8985 
8986 			i++;
8987 			tx_desc++;
8988 			if (i == tx_ring->count) {
8989 				tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8990 				i = 0;
8991 			}
8992 			tx_desc->read.olinfo_status = 0;
8993 
8994 			dma += IXGBE_MAX_DATA_PER_TXD;
8995 			size -= IXGBE_MAX_DATA_PER_TXD;
8996 
8997 			tx_desc->read.buffer_addr = cpu_to_le64(dma);
8998 		}
8999 
9000 		if (likely(!data_len))
9001 			break;
9002 
9003 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
9004 
9005 		i++;
9006 		tx_desc++;
9007 		if (i == tx_ring->count) {
9008 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
9009 			i = 0;
9010 		}
9011 		tx_desc->read.olinfo_status = 0;
9012 
9013 #ifdef IXGBE_FCOE
9014 		size = min_t(unsigned int, data_len, skb_frag_size(frag));
9015 #else
9016 		size = skb_frag_size(frag);
9017 #endif
9018 		data_len -= size;
9019 
9020 		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
9021 				       DMA_TO_DEVICE);
9022 
9023 		tx_buffer = &tx_ring->tx_buffer_info[i];
9024 	}
9025 
9026 	/* write last descriptor with RS and EOP bits */
9027 	cmd_type |= size | IXGBE_TXD_CMD;
9028 	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
9029 
9030 	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
9031 
9032 	/* set the timestamp */
9033 	first->time_stamp = jiffies;
9034 
9035 	skb_tx_timestamp(skb);
9036 
9037 	/*
9038 	 * Force memory writes to complete before letting h/w know there
9039 	 * are new descriptors to fetch.  (Only applicable for weak-ordered
9040 	 * memory model archs, such as IA-64).
9041 	 *
9042 	 * We also need this memory barrier to make certain all of the
9043 	 * status bits have been updated before next_to_watch is written.
9044 	 */
9045 	wmb();
9046 
9047 	/* set next_to_watch value indicating a packet is present */
9048 	first->next_to_watch = tx_desc;
9049 
9050 	i++;
9051 	if (i == tx_ring->count)
9052 		i = 0;
9053 
9054 	tx_ring->next_to_use = i;
9055 
9056 	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
9057 
9058 	if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more()) {
9059 		writel(i, tx_ring->tail);
9060 	}
9061 
9062 	return 0;
9063 dma_error:
9064 	dev_err(tx_ring->dev, "TX DMA map failed\n");
9065 
9066 	/* clear dma mappings for failed tx_buffer_info map */
9067 	for (;;) {
9068 		tx_buffer = &tx_ring->tx_buffer_info[i];
9069 		if (dma_unmap_len(tx_buffer, len))
9070 			dma_unmap_page(tx_ring->dev,
9071 				       dma_unmap_addr(tx_buffer, dma),
9072 				       dma_unmap_len(tx_buffer, len),
9073 				       DMA_TO_DEVICE);
9074 		dma_unmap_len_set(tx_buffer, len, 0);
9075 		if (tx_buffer == first)
9076 			break;
9077 		if (i == 0)
9078 			i += tx_ring->count;
9079 		i--;
9080 	}
9081 
9082 	dev_kfree_skb_any(first->skb);
9083 	first->skb = NULL;
9084 
9085 	tx_ring->next_to_use = i;
9086 
9087 	return -1;
9088 }
9089 
9090 static void ixgbe_atr(struct ixgbe_ring *ring,
9091 		      struct ixgbe_tx_buffer *first)
9092 {
9093 	struct ixgbe_q_vector *q_vector = ring->q_vector;
9094 	union ixgbe_atr_hash_dword input = { .dword = 0 };
9095 	union ixgbe_atr_hash_dword common = { .dword = 0 };
9096 	union {
9097 		unsigned char *network;
9098 		struct iphdr *ipv4;
9099 		struct ipv6hdr *ipv6;
9100 	} hdr;
9101 	struct tcphdr *th;
9102 	unsigned int hlen;
9103 	struct sk_buff *skb;
9104 	__be16 vlan_id;
9105 	int l4_proto;
9106 
9107 	/* if ring doesn't have a interrupt vector, cannot perform ATR */
9108 	if (!q_vector)
9109 		return;
9110 
9111 	/* do nothing if sampling is disabled */
9112 	if (!ring->atr_sample_rate)
9113 		return;
9114 
9115 	ring->atr_count++;
9116 
9117 	/* currently only IPv4/IPv6 with TCP is supported */
9118 	if ((first->protocol != htons(ETH_P_IP)) &&
9119 	    (first->protocol != htons(ETH_P_IPV6)))
9120 		return;
9121 
9122 	/* snag network header to get L4 type and address */
9123 	skb = first->skb;
9124 	hdr.network = skb_network_header(skb);
9125 	if (unlikely(hdr.network <= skb->data))
9126 		return;
9127 	if (skb->encapsulation &&
9128 	    first->protocol == htons(ETH_P_IP) &&
9129 	    hdr.ipv4->protocol == IPPROTO_UDP) {
9130 		struct ixgbe_adapter *adapter = q_vector->adapter;
9131 
9132 		if (unlikely(skb_tail_pointer(skb) < hdr.network +
9133 			     vxlan_headroom(0)))
9134 			return;
9135 
9136 		/* verify the port is recognized as VXLAN */
9137 		if (adapter->vxlan_port &&
9138 		    udp_hdr(skb)->dest == adapter->vxlan_port)
9139 			hdr.network = skb_inner_network_header(skb);
9140 
9141 		if (adapter->geneve_port &&
9142 		    udp_hdr(skb)->dest == adapter->geneve_port)
9143 			hdr.network = skb_inner_network_header(skb);
9144 	}
9145 
9146 	/* Make sure we have at least [minimum IPv4 header + TCP]
9147 	 * or [IPv6 header] bytes
9148 	 */
9149 	if (unlikely(skb_tail_pointer(skb) < hdr.network + 40))
9150 		return;
9151 
9152 	/* Currently only IPv4/IPv6 with TCP is supported */
9153 	switch (hdr.ipv4->version) {
9154 	case IPVERSION:
9155 		/* access ihl as u8 to avoid unaligned access on ia64 */
9156 		hlen = (hdr.network[0] & 0x0F) << 2;
9157 		l4_proto = hdr.ipv4->protocol;
9158 		break;
9159 	case 6:
9160 		hlen = hdr.network - skb->data;
9161 		l4_proto = ipv6_find_hdr(skb, &hlen, IPPROTO_TCP, NULL, NULL);
9162 		hlen -= hdr.network - skb->data;
9163 		break;
9164 	default:
9165 		return;
9166 	}
9167 
9168 	if (l4_proto != IPPROTO_TCP)
9169 		return;
9170 
9171 	if (unlikely(skb_tail_pointer(skb) < hdr.network +
9172 		     hlen + sizeof(struct tcphdr)))
9173 		return;
9174 
9175 	th = (struct tcphdr *)(hdr.network + hlen);
9176 
9177 	/* skip this packet since the socket is closing */
9178 	if (th->fin)
9179 		return;
9180 
9181 	/* sample on all syn packets or once every atr sample count */
9182 	if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
9183 		return;
9184 
9185 	/* reset sample count */
9186 	ring->atr_count = 0;
9187 
9188 	vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
9189 
9190 	/*
9191 	 * src and dst are inverted, think how the receiver sees them
9192 	 *
9193 	 * The input is broken into two sections, a non-compressed section
9194 	 * containing vm_pool, vlan_id, and flow_type.  The rest of the data
9195 	 * is XORed together and stored in the compressed dword.
9196 	 */
9197 	input.formatted.vlan_id = vlan_id;
9198 
9199 	/*
9200 	 * since src port and flex bytes occupy the same word XOR them together
9201 	 * and write the value to source port portion of compressed dword
9202 	 */
9203 	if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
9204 		common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
9205 	else
9206 		common.port.src ^= th->dest ^ first->protocol;
9207 	common.port.dst ^= th->source;
9208 
9209 	switch (hdr.ipv4->version) {
9210 	case IPVERSION:
9211 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
9212 		common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
9213 		break;
9214 	case 6:
9215 		input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
9216 		common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
9217 			     hdr.ipv6->saddr.s6_addr32[1] ^
9218 			     hdr.ipv6->saddr.s6_addr32[2] ^
9219 			     hdr.ipv6->saddr.s6_addr32[3] ^
9220 			     hdr.ipv6->daddr.s6_addr32[0] ^
9221 			     hdr.ipv6->daddr.s6_addr32[1] ^
9222 			     hdr.ipv6->daddr.s6_addr32[2] ^
9223 			     hdr.ipv6->daddr.s6_addr32[3];
9224 		break;
9225 	default:
9226 		break;
9227 	}
9228 
9229 	if (hdr.network != skb_network_header(skb))
9230 		input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
9231 
9232 	/* This assumes the Rx queue and Tx queue are bound to the same CPU */
9233 	ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
9234 					      input, common, ring->queue_index);
9235 }
9236 
9237 #ifdef IXGBE_FCOE
9238 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
9239 			      struct net_device *sb_dev)
9240 {
9241 	struct ixgbe_adapter *adapter;
9242 	struct ixgbe_ring_feature *f;
9243 	int txq;
9244 
9245 	if (sb_dev) {
9246 		u8 tc = netdev_get_prio_tc_map(dev, skb->priority);
9247 		struct net_device *vdev = sb_dev;
9248 
9249 		txq = vdev->tc_to_txq[tc].offset;
9250 		txq += reciprocal_scale(skb_get_hash(skb),
9251 					vdev->tc_to_txq[tc].count);
9252 
9253 		return txq;
9254 	}
9255 
9256 	/*
9257 	 * only execute the code below if protocol is FCoE
9258 	 * or FIP and we have FCoE enabled on the adapter
9259 	 */
9260 	switch (vlan_get_protocol(skb)) {
9261 	case htons(ETH_P_FCOE):
9262 	case htons(ETH_P_FIP):
9263 		adapter = ixgbe_from_netdev(dev);
9264 
9265 		if (!sb_dev && (adapter->flags & IXGBE_FLAG_FCOE_ENABLED))
9266 			break;
9267 		fallthrough;
9268 	default:
9269 		return netdev_pick_tx(dev, skb, sb_dev);
9270 	}
9271 
9272 	f = &adapter->ring_feature[RING_F_FCOE];
9273 
9274 	txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
9275 					   smp_processor_id();
9276 
9277 	while (txq >= f->indices)
9278 		txq -= f->indices;
9279 
9280 	return txq + f->offset;
9281 }
9282 
9283 #endif
9284 int ixgbe_xmit_xdp_ring(struct ixgbe_ring *ring,
9285 			struct xdp_frame *xdpf)
9286 {
9287 	struct skb_shared_info *sinfo = xdp_get_shared_info_from_frame(xdpf);
9288 	u8 nr_frags = unlikely(xdp_frame_has_frags(xdpf)) ? sinfo->nr_frags : 0;
9289 	u16 i = 0, index = ring->next_to_use;
9290 	struct ixgbe_tx_buffer *tx_head = &ring->tx_buffer_info[index];
9291 	struct ixgbe_tx_buffer *tx_buff = tx_head;
9292 	union ixgbe_adv_tx_desc *tx_desc = IXGBE_TX_DESC(ring, index);
9293 	u32 cmd_type, len = xdpf->len;
9294 	void *data = xdpf->data;
9295 
9296 	if (unlikely(ixgbe_desc_unused(ring) < 1 + nr_frags))
9297 		return IXGBE_XDP_CONSUMED;
9298 
9299 	tx_head->bytecount = xdp_get_frame_len(xdpf);
9300 	tx_head->gso_segs = 1;
9301 	tx_head->xdpf = xdpf;
9302 
9303 	tx_desc->read.olinfo_status =
9304 		cpu_to_le32(tx_head->bytecount << IXGBE_ADVTXD_PAYLEN_SHIFT);
9305 
9306 	for (;;) {
9307 		dma_addr_t dma;
9308 
9309 		dma = dma_map_single(ring->dev, data, len, DMA_TO_DEVICE);
9310 		if (dma_mapping_error(ring->dev, dma))
9311 			goto unmap;
9312 
9313 		dma_unmap_len_set(tx_buff, len, len);
9314 		dma_unmap_addr_set(tx_buff, dma, dma);
9315 
9316 		cmd_type = IXGBE_ADVTXD_DTYP_DATA | IXGBE_ADVTXD_DCMD_DEXT |
9317 			   IXGBE_ADVTXD_DCMD_IFCS | len;
9318 		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
9319 		tx_desc->read.buffer_addr = cpu_to_le64(dma);
9320 		tx_buff->protocol = 0;
9321 
9322 		if (++index == ring->count)
9323 			index = 0;
9324 
9325 		if (i == nr_frags)
9326 			break;
9327 
9328 		tx_buff = &ring->tx_buffer_info[index];
9329 		tx_desc = IXGBE_TX_DESC(ring, index);
9330 		tx_desc->read.olinfo_status = 0;
9331 
9332 		data = skb_frag_address(&sinfo->frags[i]);
9333 		len = skb_frag_size(&sinfo->frags[i]);
9334 		i++;
9335 	}
9336 	/* put descriptor type bits */
9337 	tx_desc->read.cmd_type_len |= cpu_to_le32(IXGBE_TXD_CMD);
9338 
9339 	/* Avoid any potential race with xdp_xmit and cleanup */
9340 	smp_wmb();
9341 
9342 	tx_head->next_to_watch = tx_desc;
9343 	ring->next_to_use = index;
9344 
9345 	return IXGBE_XDP_TX;
9346 
9347 unmap:
9348 	for (;;) {
9349 		tx_buff = &ring->tx_buffer_info[index];
9350 		if (dma_unmap_len(tx_buff, len))
9351 			dma_unmap_page(ring->dev, dma_unmap_addr(tx_buff, dma),
9352 				       dma_unmap_len(tx_buff, len),
9353 				       DMA_TO_DEVICE);
9354 		dma_unmap_len_set(tx_buff, len, 0);
9355 		if (tx_buff == tx_head)
9356 			break;
9357 
9358 		if (!index)
9359 			index += ring->count;
9360 		index--;
9361 	}
9362 
9363 	return IXGBE_XDP_CONSUMED;
9364 }
9365 
9366 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
9367 			  struct ixgbe_adapter *adapter,
9368 			  struct ixgbe_ring *tx_ring)
9369 {
9370 	struct ixgbe_tx_buffer *first;
9371 	int tso;
9372 	u32 tx_flags = 0;
9373 	unsigned short f;
9374 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
9375 	struct ixgbe_ipsec_tx_data ipsec_tx = { 0 };
9376 	__be16 protocol = skb->protocol;
9377 	u8 hdr_len = 0;
9378 
9379 	/*
9380 	 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
9381 	 *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
9382 	 *       + 2 desc gap to keep tail from touching head,
9383 	 *       + 1 desc for context descriptor,
9384 	 * otherwise try next time
9385 	 */
9386 	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
9387 		count += TXD_USE_COUNT(skb_frag_size(
9388 						&skb_shinfo(skb)->frags[f]));
9389 
9390 	if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
9391 		tx_ring->tx_stats.tx_busy++;
9392 		return NETDEV_TX_BUSY;
9393 	}
9394 
9395 	/* record the location of the first descriptor for this packet */
9396 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
9397 	first->skb = skb;
9398 	first->bytecount = skb->len;
9399 	first->gso_segs = 1;
9400 
9401 	/* if we have a HW VLAN tag being added default to the HW one */
9402 	if (skb_vlan_tag_present(skb)) {
9403 		tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
9404 		tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
9405 	/* else if it is a SW VLAN check the next protocol and store the tag */
9406 	} else if (protocol == htons(ETH_P_8021Q)) {
9407 		struct vlan_hdr *vhdr, _vhdr;
9408 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
9409 		if (!vhdr)
9410 			goto out_drop;
9411 
9412 		tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
9413 				  IXGBE_TX_FLAGS_VLAN_SHIFT;
9414 		tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
9415 	}
9416 	protocol = vlan_get_protocol(skb);
9417 
9418 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
9419 	    adapter->ptp_clock) {
9420 		if (adapter->tstamp_config.tx_type == HWTSTAMP_TX_ON &&
9421 		    !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
9422 					   &adapter->state)) {
9423 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
9424 			tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
9425 
9426 			/* schedule check for Tx timestamp */
9427 			adapter->ptp_tx_skb = skb_get(skb);
9428 			adapter->ptp_tx_start = jiffies;
9429 			schedule_work(&adapter->ptp_tx_work);
9430 		} else {
9431 			adapter->tx_hwtstamp_skipped++;
9432 		}
9433 	}
9434 
9435 #ifdef CONFIG_PCI_IOV
9436 	/*
9437 	 * Use the l2switch_enable flag - would be false if the DMA
9438 	 * Tx switch had been disabled.
9439 	 */
9440 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
9441 		tx_flags |= IXGBE_TX_FLAGS_CC;
9442 
9443 #endif
9444 	/* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
9445 	if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
9446 	    ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
9447 	     (skb->priority != TC_PRIO_CONTROL))) {
9448 		tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
9449 		tx_flags |= (skb->priority & 0x7) <<
9450 					IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
9451 		if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
9452 			struct vlan_ethhdr *vhdr;
9453 
9454 			if (skb_cow_head(skb, 0))
9455 				goto out_drop;
9456 			vhdr = skb_vlan_eth_hdr(skb);
9457 			vhdr->h_vlan_TCI = htons(tx_flags >>
9458 						 IXGBE_TX_FLAGS_VLAN_SHIFT);
9459 		} else {
9460 			tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
9461 		}
9462 	}
9463 
9464 	/* record initial flags and protocol */
9465 	first->tx_flags = tx_flags;
9466 	first->protocol = protocol;
9467 
9468 #ifdef IXGBE_FCOE
9469 	/* setup tx offload for FCoE */
9470 	if ((protocol == htons(ETH_P_FCOE)) &&
9471 	    (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
9472 		tso = ixgbe_fso(tx_ring, first, &hdr_len);
9473 		if (tso < 0)
9474 			goto out_drop;
9475 
9476 		goto xmit_fcoe;
9477 	}
9478 
9479 #endif /* IXGBE_FCOE */
9480 
9481 #ifdef CONFIG_IXGBE_IPSEC
9482 	if (xfrm_offload(skb) &&
9483 	    !ixgbe_ipsec_tx(tx_ring, first, &ipsec_tx))
9484 		goto out_drop;
9485 #endif
9486 	tso = ixgbe_tso(tx_ring, first, &hdr_len, &ipsec_tx);
9487 	if (tso < 0)
9488 		goto out_drop;
9489 	else if (!tso)
9490 		ixgbe_tx_csum(tx_ring, first, &ipsec_tx);
9491 
9492 	/* add the ATR filter if ATR is on */
9493 	if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
9494 		ixgbe_atr(tx_ring, first);
9495 
9496 #ifdef IXGBE_FCOE
9497 xmit_fcoe:
9498 #endif /* IXGBE_FCOE */
9499 	if (ixgbe_tx_map(tx_ring, first, hdr_len))
9500 		goto cleanup_tx_timestamp;
9501 
9502 	return NETDEV_TX_OK;
9503 
9504 out_drop:
9505 	dev_kfree_skb_any(first->skb);
9506 	first->skb = NULL;
9507 cleanup_tx_timestamp:
9508 	if (unlikely(tx_flags & IXGBE_TX_FLAGS_TSTAMP)) {
9509 		dev_kfree_skb_any(adapter->ptp_tx_skb);
9510 		adapter->ptp_tx_skb = NULL;
9511 		cancel_work_sync(&adapter->ptp_tx_work);
9512 		clear_bit_unlock(__IXGBE_PTP_TX_IN_PROGRESS, &adapter->state);
9513 	}
9514 
9515 	return NETDEV_TX_OK;
9516 }
9517 
9518 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
9519 				      struct net_device *netdev,
9520 				      struct ixgbe_ring *ring)
9521 {
9522 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
9523 	struct ixgbe_ring *tx_ring;
9524 
9525 	/*
9526 	 * The minimum packet size for olinfo paylen is 17 so pad the skb
9527 	 * in order to meet this minimum size requirement.
9528 	 */
9529 	if (skb_put_padto(skb, 17))
9530 		return NETDEV_TX_OK;
9531 
9532 	tx_ring = ring ? ring : adapter->tx_ring[skb_get_queue_mapping(skb)];
9533 	if (unlikely(test_bit(__IXGBE_TX_DISABLED, &tx_ring->state)))
9534 		return NETDEV_TX_BUSY;
9535 
9536 	return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
9537 }
9538 
9539 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
9540 				    struct net_device *netdev)
9541 {
9542 	return __ixgbe_xmit_frame(skb, netdev, NULL);
9543 }
9544 
9545 /**
9546  * ixgbe_set_mac - Change the Ethernet Address of the NIC
9547  * @netdev: network interface device structure
9548  * @p: pointer to an address structure
9549  *
9550  * Returns 0 on success, negative on failure
9551  **/
9552 static int ixgbe_set_mac(struct net_device *netdev, void *p)
9553 {
9554 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
9555 	struct ixgbe_hw *hw = &adapter->hw;
9556 	struct sockaddr *addr = p;
9557 
9558 	if (!is_valid_ether_addr(addr->sa_data))
9559 		return -EADDRNOTAVAIL;
9560 
9561 	eth_hw_addr_set(netdev, addr->sa_data);
9562 	memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
9563 
9564 	ixgbe_mac_set_default_filter(adapter);
9565 
9566 	return 0;
9567 }
9568 
9569 static int
9570 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
9571 {
9572 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
9573 	struct ixgbe_hw *hw = &adapter->hw;
9574 	u16 value;
9575 	int rc;
9576 
9577 	if (adapter->mii_bus) {
9578 		int regnum = addr;
9579 
9580 		if (devad != MDIO_DEVAD_NONE)
9581 			return mdiobus_c45_read(adapter->mii_bus, prtad,
9582 						devad, regnum);
9583 
9584 		return mdiobus_read(adapter->mii_bus, prtad, regnum);
9585 	}
9586 
9587 	if (prtad != hw->phy.mdio.prtad)
9588 		return -EINVAL;
9589 	rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
9590 	if (!rc)
9591 		rc = value;
9592 	return rc;
9593 }
9594 
9595 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
9596 			    u16 addr, u16 value)
9597 {
9598 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
9599 	struct ixgbe_hw *hw = &adapter->hw;
9600 
9601 	if (adapter->mii_bus) {
9602 		int regnum = addr;
9603 
9604 		if (devad != MDIO_DEVAD_NONE)
9605 			return mdiobus_c45_write(adapter->mii_bus, prtad, devad,
9606 						 regnum, value);
9607 
9608 		return mdiobus_write(adapter->mii_bus, prtad, regnum, value);
9609 	}
9610 
9611 	if (prtad != hw->phy.mdio.prtad)
9612 		return -EINVAL;
9613 	return hw->phy.ops.write_reg(hw, addr, devad, value);
9614 }
9615 
9616 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
9617 {
9618 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
9619 
9620 	switch (cmd) {
9621 	case SIOCGMIIPHY:
9622 		if (!adapter->hw.phy.ops.read_reg)
9623 			return -EOPNOTSUPP;
9624 		fallthrough;
9625 	default:
9626 		return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
9627 	}
9628 }
9629 
9630 /**
9631  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
9632  * netdev->dev_addrs
9633  * @dev: network interface device structure
9634  *
9635  * Returns non-zero on failure
9636  **/
9637 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
9638 {
9639 	int err = 0;
9640 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(dev);
9641 	struct ixgbe_hw *hw = &adapter->hw;
9642 
9643 	if (is_valid_ether_addr(hw->mac.san_addr)) {
9644 		rtnl_lock();
9645 		err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
9646 		rtnl_unlock();
9647 
9648 		/* update SAN MAC vmdq pool selection */
9649 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
9650 	}
9651 	return err;
9652 }
9653 
9654 /**
9655  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
9656  * netdev->dev_addrs
9657  * @dev: network interface device structure
9658  *
9659  * Returns non-zero on failure
9660  **/
9661 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
9662 {
9663 	int err = 0;
9664 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(dev);
9665 	struct ixgbe_mac_info *mac = &adapter->hw.mac;
9666 
9667 	if (is_valid_ether_addr(mac->san_addr)) {
9668 		rtnl_lock();
9669 		err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
9670 		rtnl_unlock();
9671 	}
9672 	return err;
9673 }
9674 
9675 static void ixgbe_get_ring_stats64(struct rtnl_link_stats64 *stats,
9676 				   struct ixgbe_ring *ring)
9677 {
9678 	u64 bytes, packets;
9679 	unsigned int start;
9680 
9681 	if (ring) {
9682 		do {
9683 			start = u64_stats_fetch_begin(&ring->syncp);
9684 			packets = ring->stats.packets;
9685 			bytes   = ring->stats.bytes;
9686 		} while (u64_stats_fetch_retry(&ring->syncp, start));
9687 		stats->tx_packets += packets;
9688 		stats->tx_bytes   += bytes;
9689 	}
9690 }
9691 
9692 static void ixgbe_get_stats64(struct net_device *netdev,
9693 			      struct rtnl_link_stats64 *stats)
9694 {
9695 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
9696 	int i;
9697 
9698 	rcu_read_lock();
9699 	for (i = 0; i < adapter->num_rx_queues; i++) {
9700 		struct ixgbe_ring *ring = READ_ONCE(adapter->rx_ring[i]);
9701 		u64 bytes, packets;
9702 		unsigned int start;
9703 
9704 		if (ring) {
9705 			do {
9706 				start = u64_stats_fetch_begin(&ring->syncp);
9707 				packets = ring->stats.packets;
9708 				bytes   = ring->stats.bytes;
9709 			} while (u64_stats_fetch_retry(&ring->syncp, start));
9710 			stats->rx_packets += packets;
9711 			stats->rx_bytes   += bytes;
9712 		}
9713 	}
9714 
9715 	for (i = 0; i < adapter->num_tx_queues; i++) {
9716 		struct ixgbe_ring *ring = READ_ONCE(adapter->tx_ring[i]);
9717 
9718 		ixgbe_get_ring_stats64(stats, ring);
9719 	}
9720 	for (i = 0; i < adapter->num_xdp_queues; i++) {
9721 		struct ixgbe_ring *ring = READ_ONCE(adapter->xdp_ring[i]);
9722 
9723 		ixgbe_get_ring_stats64(stats, ring);
9724 	}
9725 	rcu_read_unlock();
9726 
9727 	/* following stats updated by ixgbe_watchdog_task() */
9728 	stats->multicast	= netdev->stats.multicast;
9729 	stats->rx_errors	= netdev->stats.rx_errors;
9730 	stats->rx_length_errors	= netdev->stats.rx_length_errors;
9731 	stats->rx_crc_errors	= netdev->stats.rx_crc_errors;
9732 	stats->rx_missed_errors	= netdev->stats.rx_missed_errors;
9733 }
9734 
9735 static int ixgbe_ndo_get_vf_stats(struct net_device *netdev, int vf,
9736 				  struct ifla_vf_stats *vf_stats)
9737 {
9738 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
9739 
9740 	if (vf < 0 || vf >= adapter->num_vfs)
9741 		return -EINVAL;
9742 
9743 	vf_stats->rx_packets = adapter->vfinfo[vf].vfstats.gprc;
9744 	vf_stats->rx_bytes   = adapter->vfinfo[vf].vfstats.gorc;
9745 	vf_stats->tx_packets = adapter->vfinfo[vf].vfstats.gptc;
9746 	vf_stats->tx_bytes   = adapter->vfinfo[vf].vfstats.gotc;
9747 	vf_stats->multicast  = adapter->vfinfo[vf].vfstats.mprc;
9748 
9749 	return 0;
9750 }
9751 
9752 #ifdef CONFIG_IXGBE_DCB
9753 /**
9754  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
9755  * @adapter: pointer to ixgbe_adapter
9756  * @tc: number of traffic classes currently enabled
9757  *
9758  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
9759  * 802.1Q priority maps to a packet buffer that exists.
9760  */
9761 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
9762 {
9763 	struct ixgbe_hw *hw = &adapter->hw;
9764 	u32 reg, rsave;
9765 	int i;
9766 
9767 	/* 82598 have a static priority to TC mapping that can not
9768 	 * be changed so no validation is needed.
9769 	 */
9770 	if (hw->mac.type == ixgbe_mac_82598EB)
9771 		return;
9772 
9773 	reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
9774 	rsave = reg;
9775 
9776 	for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
9777 		u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
9778 
9779 		/* If up2tc is out of bounds default to zero */
9780 		if (up2tc > tc)
9781 			reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
9782 	}
9783 
9784 	if (reg != rsave)
9785 		IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
9786 
9787 	return;
9788 }
9789 
9790 /**
9791  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
9792  * @adapter: Pointer to adapter struct
9793  *
9794  * Populate the netdev user priority to tc map
9795  */
9796 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
9797 {
9798 	struct net_device *dev = adapter->netdev;
9799 	struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
9800 	struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
9801 	u8 prio;
9802 
9803 	for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
9804 		u8 tc = 0;
9805 
9806 		if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
9807 			tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
9808 		else if (ets)
9809 			tc = ets->prio_tc[prio];
9810 
9811 		netdev_set_prio_tc_map(dev, prio, tc);
9812 	}
9813 }
9814 
9815 #endif /* CONFIG_IXGBE_DCB */
9816 static int ixgbe_reassign_macvlan_pool(struct net_device *vdev,
9817 				       struct netdev_nested_priv *priv)
9818 {
9819 	struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)priv->data;
9820 	struct ixgbe_fwd_adapter *accel;
9821 	int pool;
9822 
9823 	/* we only care about macvlans... */
9824 	if (!netif_is_macvlan(vdev))
9825 		return 0;
9826 
9827 	/* that have hardware offload enabled... */
9828 	accel = macvlan_accel_priv(vdev);
9829 	if (!accel)
9830 		return 0;
9831 
9832 	/* If we can relocate to a different bit do so */
9833 	pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
9834 	if (pool < adapter->num_rx_pools) {
9835 		set_bit(pool, adapter->fwd_bitmask);
9836 		accel->pool = pool;
9837 		return 0;
9838 	}
9839 
9840 	/* if we cannot find a free pool then disable the offload */
9841 	netdev_err(vdev, "L2FW offload disabled due to lack of queue resources\n");
9842 	macvlan_release_l2fw_offload(vdev);
9843 
9844 	/* unbind the queues and drop the subordinate channel config */
9845 	netdev_unbind_sb_channel(adapter->netdev, vdev);
9846 	netdev_set_sb_channel(vdev, 0);
9847 
9848 	kfree(accel);
9849 
9850 	return 0;
9851 }
9852 
9853 static void ixgbe_defrag_macvlan_pools(struct net_device *dev)
9854 {
9855 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(dev);
9856 	struct netdev_nested_priv priv = {
9857 		.data = (void *)adapter,
9858 	};
9859 
9860 	/* flush any stale bits out of the fwd bitmask */
9861 	bitmap_clear(adapter->fwd_bitmask, 1, 63);
9862 
9863 	/* walk through upper devices reassigning pools */
9864 	netdev_walk_all_upper_dev_rcu(dev, ixgbe_reassign_macvlan_pool,
9865 				      &priv);
9866 }
9867 
9868 /**
9869  * ixgbe_setup_tc - configure net_device for multiple traffic classes
9870  *
9871  * @dev: net device to configure
9872  * @tc: number of traffic classes to enable
9873  */
9874 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
9875 {
9876 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(dev);
9877 	struct ixgbe_hw *hw = &adapter->hw;
9878 
9879 	/* Hardware supports up to 8 traffic classes */
9880 	if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
9881 		return -EINVAL;
9882 
9883 	if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
9884 		return -EINVAL;
9885 
9886 	/* Hardware has to reinitialize queues and interrupts to
9887 	 * match packet buffer alignment. Unfortunately, the
9888 	 * hardware is not flexible enough to do this dynamically.
9889 	 */
9890 	if (netif_running(dev))
9891 		ixgbe_close(dev);
9892 	else
9893 		ixgbe_reset(adapter);
9894 
9895 	ixgbe_clear_interrupt_scheme(adapter);
9896 
9897 #ifdef CONFIG_IXGBE_DCB
9898 	if (tc) {
9899 		if (adapter->xdp_prog) {
9900 			e_warn(probe, "DCB is not supported with XDP\n");
9901 
9902 			ixgbe_init_interrupt_scheme(adapter);
9903 			if (netif_running(dev))
9904 				ixgbe_open(dev);
9905 			return -EINVAL;
9906 		}
9907 
9908 		netdev_set_num_tc(dev, tc);
9909 		ixgbe_set_prio_tc_map(adapter);
9910 
9911 		adapter->hw_tcs = tc;
9912 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
9913 
9914 		if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
9915 			adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
9916 			adapter->hw.fc.requested_mode = ixgbe_fc_none;
9917 		}
9918 	} else {
9919 		netdev_reset_tc(dev);
9920 
9921 		if (adapter->hw.mac.type == ixgbe_mac_82598EB)
9922 			adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
9923 
9924 		adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
9925 		adapter->hw_tcs = tc;
9926 
9927 		adapter->temp_dcb_cfg.pfc_mode_enable = false;
9928 		adapter->dcb_cfg.pfc_mode_enable = false;
9929 	}
9930 
9931 	ixgbe_validate_rtr(adapter, tc);
9932 
9933 #endif /* CONFIG_IXGBE_DCB */
9934 	ixgbe_init_interrupt_scheme(adapter);
9935 
9936 	ixgbe_defrag_macvlan_pools(dev);
9937 
9938 	if (netif_running(dev))
9939 		return ixgbe_open(dev);
9940 
9941 	return 0;
9942 }
9943 
9944 static int ixgbe_delete_clsu32(struct ixgbe_adapter *adapter,
9945 			       struct tc_cls_u32_offload *cls)
9946 {
9947 	u32 hdl = cls->knode.handle;
9948 	u32 uhtid = TC_U32_USERHTID(cls->knode.handle);
9949 	u32 loc = cls->knode.handle & 0xfffff;
9950 	int err = 0, i, j;
9951 	struct ixgbe_jump_table *jump = NULL;
9952 
9953 	if (loc > IXGBE_MAX_HW_ENTRIES)
9954 		return -EINVAL;
9955 
9956 	if ((uhtid != 0x800) && (uhtid >= IXGBE_MAX_LINK_HANDLE))
9957 		return -EINVAL;
9958 
9959 	/* Clear this filter in the link data it is associated with */
9960 	if (uhtid != 0x800) {
9961 		jump = adapter->jump_tables[uhtid];
9962 		if (!jump)
9963 			return -EINVAL;
9964 		if (!test_bit(loc - 1, jump->child_loc_map))
9965 			return -EINVAL;
9966 		clear_bit(loc - 1, jump->child_loc_map);
9967 	}
9968 
9969 	/* Check if the filter being deleted is a link */
9970 	for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
9971 		jump = adapter->jump_tables[i];
9972 		if (jump && jump->link_hdl == hdl) {
9973 			/* Delete filters in the hardware in the child hash
9974 			 * table associated with this link
9975 			 */
9976 			for (j = 0; j < IXGBE_MAX_HW_ENTRIES; j++) {
9977 				if (!test_bit(j, jump->child_loc_map))
9978 					continue;
9979 				spin_lock(&adapter->fdir_perfect_lock);
9980 				err = ixgbe_update_ethtool_fdir_entry(adapter,
9981 								      NULL,
9982 								      j + 1);
9983 				spin_unlock(&adapter->fdir_perfect_lock);
9984 				clear_bit(j, jump->child_loc_map);
9985 			}
9986 			/* Remove resources for this link */
9987 			kfree(jump->input);
9988 			kfree(jump->mask);
9989 			kfree(jump);
9990 			adapter->jump_tables[i] = NULL;
9991 			return err;
9992 		}
9993 	}
9994 
9995 	spin_lock(&adapter->fdir_perfect_lock);
9996 	err = ixgbe_update_ethtool_fdir_entry(adapter, NULL, loc);
9997 	spin_unlock(&adapter->fdir_perfect_lock);
9998 	return err;
9999 }
10000 
10001 static int ixgbe_configure_clsu32_add_hnode(struct ixgbe_adapter *adapter,
10002 					    struct tc_cls_u32_offload *cls)
10003 {
10004 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
10005 
10006 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
10007 		return -EINVAL;
10008 
10009 	/* This ixgbe devices do not support hash tables at the moment
10010 	 * so abort when given hash tables.
10011 	 */
10012 	if (cls->hnode.divisor > 0)
10013 		return -EINVAL;
10014 
10015 	set_bit(uhtid - 1, &adapter->tables);
10016 	return 0;
10017 }
10018 
10019 static int ixgbe_configure_clsu32_del_hnode(struct ixgbe_adapter *adapter,
10020 					    struct tc_cls_u32_offload *cls)
10021 {
10022 	u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
10023 
10024 	if (uhtid >= IXGBE_MAX_LINK_HANDLE)
10025 		return -EINVAL;
10026 
10027 	clear_bit(uhtid - 1, &adapter->tables);
10028 	return 0;
10029 }
10030 
10031 #ifdef CONFIG_NET_CLS_ACT
10032 struct upper_walk_data {
10033 	struct ixgbe_adapter *adapter;
10034 	u64 action;
10035 	int ifindex;
10036 	u8 queue;
10037 };
10038 
10039 static int get_macvlan_queue(struct net_device *upper,
10040 			     struct netdev_nested_priv *priv)
10041 {
10042 	if (netif_is_macvlan(upper)) {
10043 		struct ixgbe_fwd_adapter *vadapter = macvlan_accel_priv(upper);
10044 		struct ixgbe_adapter *adapter;
10045 		struct upper_walk_data *data;
10046 		int ifindex;
10047 
10048 		data = (struct upper_walk_data *)priv->data;
10049 		ifindex = data->ifindex;
10050 		adapter = data->adapter;
10051 		if (vadapter && upper->ifindex == ifindex) {
10052 			data->queue = adapter->rx_ring[vadapter->rx_base_queue]->reg_idx;
10053 			data->action = data->queue;
10054 			return 1;
10055 		}
10056 	}
10057 
10058 	return 0;
10059 }
10060 
10061 static int handle_redirect_action(struct ixgbe_adapter *adapter, int ifindex,
10062 				  u8 *queue, u64 *action)
10063 {
10064 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
10065 	unsigned int num_vfs = adapter->num_vfs, vf;
10066 	struct netdev_nested_priv priv;
10067 	struct upper_walk_data data;
10068 	struct net_device *upper;
10069 
10070 	/* redirect to a SRIOV VF */
10071 	for (vf = 0; vf < num_vfs; ++vf) {
10072 		upper = pci_get_drvdata(adapter->vfinfo[vf].vfdev);
10073 		if (upper->ifindex == ifindex) {
10074 			*queue = vf * __ALIGN_MASK(1, ~vmdq->mask);
10075 			*action = vf + 1;
10076 			*action <<= ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF;
10077 			return 0;
10078 		}
10079 	}
10080 
10081 	/* redirect to a offloaded macvlan netdev */
10082 	data.adapter = adapter;
10083 	data.ifindex = ifindex;
10084 	data.action = 0;
10085 	data.queue = 0;
10086 	priv.data = (void *)&data;
10087 	if (netdev_walk_all_upper_dev_rcu(adapter->netdev,
10088 					  get_macvlan_queue, &priv)) {
10089 		*action = data.action;
10090 		*queue = data.queue;
10091 
10092 		return 0;
10093 	}
10094 
10095 	return -EINVAL;
10096 }
10097 
10098 static int parse_tc_actions(struct ixgbe_adapter *adapter,
10099 			    struct tcf_exts *exts, u64 *action, u8 *queue)
10100 {
10101 	const struct tc_action *a;
10102 	int i;
10103 
10104 	if (!tcf_exts_has_actions(exts))
10105 		return -EINVAL;
10106 
10107 	tcf_exts_for_each_action(i, a, exts) {
10108 		/* Drop action */
10109 		if (is_tcf_gact_shot(a)) {
10110 			*action = IXGBE_FDIR_DROP_QUEUE;
10111 			*queue = IXGBE_FDIR_DROP_QUEUE;
10112 			return 0;
10113 		}
10114 
10115 		/* Redirect to a VF or a offloaded macvlan */
10116 		if (is_tcf_mirred_egress_redirect(a)) {
10117 			struct net_device *dev = tcf_mirred_dev(a);
10118 
10119 			if (!dev)
10120 				return -EINVAL;
10121 			return handle_redirect_action(adapter, dev->ifindex,
10122 						      queue, action);
10123 		}
10124 
10125 		return -EINVAL;
10126 	}
10127 
10128 	return -EINVAL;
10129 }
10130 #else
10131 static int parse_tc_actions(struct ixgbe_adapter *adapter,
10132 			    struct tcf_exts *exts, u64 *action, u8 *queue)
10133 {
10134 	return -EINVAL;
10135 }
10136 #endif /* CONFIG_NET_CLS_ACT */
10137 
10138 static int ixgbe_clsu32_build_input(struct ixgbe_fdir_filter *input,
10139 				    union ixgbe_atr_input *mask,
10140 				    struct tc_cls_u32_offload *cls,
10141 				    struct ixgbe_mat_field *field_ptr,
10142 				    struct ixgbe_nexthdr *nexthdr)
10143 {
10144 	int i, j, off;
10145 	__be32 val, m;
10146 	bool found_entry = false, found_jump_field = false;
10147 
10148 	for (i = 0; i < cls->knode.sel->nkeys; i++) {
10149 		off = cls->knode.sel->keys[i].off;
10150 		val = cls->knode.sel->keys[i].val;
10151 		m = cls->knode.sel->keys[i].mask;
10152 
10153 		for (j = 0; field_ptr[j].val; j++) {
10154 			if (field_ptr[j].off == off) {
10155 				field_ptr[j].val(input, mask, (__force u32)val,
10156 						 (__force u32)m);
10157 				input->filter.formatted.flow_type |=
10158 					field_ptr[j].type;
10159 				found_entry = true;
10160 				break;
10161 			}
10162 		}
10163 		if (nexthdr) {
10164 			if (nexthdr->off == cls->knode.sel->keys[i].off &&
10165 			    nexthdr->val ==
10166 			    (__force u32)cls->knode.sel->keys[i].val &&
10167 			    nexthdr->mask ==
10168 			    (__force u32)cls->knode.sel->keys[i].mask)
10169 				found_jump_field = true;
10170 			else
10171 				continue;
10172 		}
10173 	}
10174 
10175 	if (nexthdr && !found_jump_field)
10176 		return -EINVAL;
10177 
10178 	if (!found_entry)
10179 		return 0;
10180 
10181 	mask->formatted.flow_type = IXGBE_ATR_L4TYPE_IPV6_MASK |
10182 				    IXGBE_ATR_L4TYPE_MASK;
10183 
10184 	if (input->filter.formatted.flow_type == IXGBE_ATR_FLOW_TYPE_IPV4)
10185 		mask->formatted.flow_type &= IXGBE_ATR_L4TYPE_IPV6_MASK;
10186 
10187 	return 0;
10188 }
10189 
10190 static int ixgbe_configure_clsu32(struct ixgbe_adapter *adapter,
10191 				  struct tc_cls_u32_offload *cls)
10192 {
10193 	__be16 protocol = cls->common.protocol;
10194 	u32 loc = cls->knode.handle & 0xfffff;
10195 	struct ixgbe_hw *hw = &adapter->hw;
10196 	struct ixgbe_mat_field *field_ptr;
10197 	struct ixgbe_fdir_filter *input = NULL;
10198 	union ixgbe_atr_input *mask = NULL;
10199 	struct ixgbe_jump_table *jump = NULL;
10200 	int i, err = -EINVAL;
10201 	u8 queue;
10202 	u32 uhtid, link_uhtid;
10203 
10204 	uhtid = TC_U32_USERHTID(cls->knode.handle);
10205 	link_uhtid = TC_U32_USERHTID(cls->knode.link_handle);
10206 
10207 	/* At the moment cls_u32 jumps to network layer and skips past
10208 	 * L2 headers. The canonical method to match L2 frames is to use
10209 	 * negative values. However this is error prone at best but really
10210 	 * just broken because there is no way to "know" what sort of hdr
10211 	 * is in front of the network layer. Fix cls_u32 to support L2
10212 	 * headers when needed.
10213 	 */
10214 	if (protocol != htons(ETH_P_IP))
10215 		return err;
10216 
10217 	if (loc >= ((1024 << adapter->fdir_pballoc) - 2)) {
10218 		e_err(drv, "Location out of range\n");
10219 		return err;
10220 	}
10221 
10222 	/* cls u32 is a graph starting at root node 0x800. The driver tracks
10223 	 * links and also the fields used to advance the parser across each
10224 	 * link (e.g. nexthdr/eat parameters from 'tc'). This way we can map
10225 	 * the u32 graph onto the hardware parse graph denoted in ixgbe_model.h
10226 	 * To add support for new nodes update ixgbe_model.h parse structures
10227 	 * this function _should_ be generic try not to hardcode values here.
10228 	 */
10229 	if (uhtid == 0x800) {
10230 		field_ptr = (adapter->jump_tables[0])->mat;
10231 	} else {
10232 		if (uhtid >= IXGBE_MAX_LINK_HANDLE)
10233 			return err;
10234 		if (!adapter->jump_tables[uhtid])
10235 			return err;
10236 		field_ptr = (adapter->jump_tables[uhtid])->mat;
10237 	}
10238 
10239 	if (!field_ptr)
10240 		return err;
10241 
10242 	/* At this point we know the field_ptr is valid and need to either
10243 	 * build cls_u32 link or attach filter. Because adding a link to
10244 	 * a handle that does not exist is invalid and the same for adding
10245 	 * rules to handles that don't exist.
10246 	 */
10247 
10248 	if (link_uhtid) {
10249 		struct ixgbe_nexthdr *nexthdr = ixgbe_ipv4_jumps;
10250 
10251 		if (link_uhtid >= IXGBE_MAX_LINK_HANDLE)
10252 			return err;
10253 
10254 		if (!test_bit(link_uhtid - 1, &adapter->tables))
10255 			return err;
10256 
10257 		/* Multiple filters as links to the same hash table are not
10258 		 * supported. To add a new filter with the same next header
10259 		 * but different match/jump conditions, create a new hash table
10260 		 * and link to it.
10261 		 */
10262 		if (adapter->jump_tables[link_uhtid] &&
10263 		    (adapter->jump_tables[link_uhtid])->link_hdl) {
10264 			e_err(drv, "Link filter exists for link: %x\n",
10265 			      link_uhtid);
10266 			return err;
10267 		}
10268 
10269 		for (i = 0; nexthdr[i].jump; i++) {
10270 			if (nexthdr[i].o != cls->knode.sel->offoff ||
10271 			    nexthdr[i].s != cls->knode.sel->offshift ||
10272 			    nexthdr[i].m !=
10273 			    (__force u32)cls->knode.sel->offmask)
10274 				return err;
10275 
10276 			jump = kzalloc(sizeof(*jump), GFP_KERNEL);
10277 			if (!jump)
10278 				return -ENOMEM;
10279 			input = kzalloc(sizeof(*input), GFP_KERNEL);
10280 			if (!input) {
10281 				err = -ENOMEM;
10282 				goto free_jump;
10283 			}
10284 			mask = kzalloc(sizeof(*mask), GFP_KERNEL);
10285 			if (!mask) {
10286 				err = -ENOMEM;
10287 				goto free_input;
10288 			}
10289 			jump->input = input;
10290 			jump->mask = mask;
10291 			jump->link_hdl = cls->knode.handle;
10292 
10293 			err = ixgbe_clsu32_build_input(input, mask, cls,
10294 						       field_ptr, &nexthdr[i]);
10295 			if (!err) {
10296 				jump->mat = nexthdr[i].jump;
10297 				adapter->jump_tables[link_uhtid] = jump;
10298 				break;
10299 			} else {
10300 				kfree(mask);
10301 				kfree(input);
10302 				kfree(jump);
10303 			}
10304 		}
10305 		return 0;
10306 	}
10307 
10308 	input = kzalloc(sizeof(*input), GFP_KERNEL);
10309 	if (!input)
10310 		return -ENOMEM;
10311 	mask = kzalloc(sizeof(*mask), GFP_KERNEL);
10312 	if (!mask) {
10313 		err = -ENOMEM;
10314 		goto free_input;
10315 	}
10316 
10317 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid])) {
10318 		if ((adapter->jump_tables[uhtid])->input)
10319 			memcpy(input, (adapter->jump_tables[uhtid])->input,
10320 			       sizeof(*input));
10321 		if ((adapter->jump_tables[uhtid])->mask)
10322 			memcpy(mask, (adapter->jump_tables[uhtid])->mask,
10323 			       sizeof(*mask));
10324 
10325 		/* Lookup in all child hash tables if this location is already
10326 		 * filled with a filter
10327 		 */
10328 		for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
10329 			struct ixgbe_jump_table *link = adapter->jump_tables[i];
10330 
10331 			if (link && (test_bit(loc - 1, link->child_loc_map))) {
10332 				e_err(drv, "Filter exists in location: %x\n",
10333 				      loc);
10334 				err = -EINVAL;
10335 				goto err_out;
10336 			}
10337 		}
10338 	}
10339 	err = ixgbe_clsu32_build_input(input, mask, cls, field_ptr, NULL);
10340 	if (err)
10341 		goto err_out;
10342 
10343 	err = parse_tc_actions(adapter, cls->knode.exts, &input->action,
10344 			       &queue);
10345 	if (err < 0)
10346 		goto err_out;
10347 
10348 	input->sw_idx = loc;
10349 
10350 	spin_lock(&adapter->fdir_perfect_lock);
10351 
10352 	if (hlist_empty(&adapter->fdir_filter_list)) {
10353 		memcpy(&adapter->fdir_mask, mask, sizeof(*mask));
10354 		err = ixgbe_fdir_set_input_mask_82599(hw, mask);
10355 		if (err)
10356 			goto err_out_w_lock;
10357 	} else if (memcmp(&adapter->fdir_mask, mask, sizeof(*mask))) {
10358 		err = -EINVAL;
10359 		goto err_out_w_lock;
10360 	}
10361 
10362 	ixgbe_atr_compute_perfect_hash_82599(&input->filter, mask);
10363 	err = ixgbe_fdir_write_perfect_filter_82599(hw, &input->filter,
10364 						    input->sw_idx, queue);
10365 	if (err)
10366 		goto err_out_w_lock;
10367 
10368 	ixgbe_update_ethtool_fdir_entry(adapter, input, input->sw_idx);
10369 	spin_unlock(&adapter->fdir_perfect_lock);
10370 
10371 	if ((uhtid != 0x800) && (adapter->jump_tables[uhtid]))
10372 		set_bit(loc - 1, (adapter->jump_tables[uhtid])->child_loc_map);
10373 
10374 	kfree(mask);
10375 	return err;
10376 err_out_w_lock:
10377 	spin_unlock(&adapter->fdir_perfect_lock);
10378 err_out:
10379 	kfree(mask);
10380 free_input:
10381 	kfree(input);
10382 free_jump:
10383 	kfree(jump);
10384 	return err;
10385 }
10386 
10387 static int ixgbe_setup_tc_cls_u32(struct ixgbe_adapter *adapter,
10388 				  struct tc_cls_u32_offload *cls_u32)
10389 {
10390 	switch (cls_u32->command) {
10391 	case TC_CLSU32_NEW_KNODE:
10392 	case TC_CLSU32_REPLACE_KNODE:
10393 		return ixgbe_configure_clsu32(adapter, cls_u32);
10394 	case TC_CLSU32_DELETE_KNODE:
10395 		return ixgbe_delete_clsu32(adapter, cls_u32);
10396 	case TC_CLSU32_NEW_HNODE:
10397 	case TC_CLSU32_REPLACE_HNODE:
10398 		return ixgbe_configure_clsu32_add_hnode(adapter, cls_u32);
10399 	case TC_CLSU32_DELETE_HNODE:
10400 		return ixgbe_configure_clsu32_del_hnode(adapter, cls_u32);
10401 	default:
10402 		return -EOPNOTSUPP;
10403 	}
10404 }
10405 
10406 static int ixgbe_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
10407 				   void *cb_priv)
10408 {
10409 	struct ixgbe_adapter *adapter = cb_priv;
10410 
10411 	if (!tc_cls_can_offload_and_chain0(adapter->netdev, type_data))
10412 		return -EOPNOTSUPP;
10413 
10414 	switch (type) {
10415 	case TC_SETUP_CLSU32:
10416 		return ixgbe_setup_tc_cls_u32(adapter, type_data);
10417 	default:
10418 		return -EOPNOTSUPP;
10419 	}
10420 }
10421 
10422 static int ixgbe_setup_tc_mqprio(struct net_device *dev,
10423 				 struct tc_mqprio_qopt *mqprio)
10424 {
10425 	mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
10426 	return ixgbe_setup_tc(dev, mqprio->num_tc);
10427 }
10428 
10429 static LIST_HEAD(ixgbe_block_cb_list);
10430 
10431 static int __ixgbe_setup_tc(struct net_device *dev, enum tc_setup_type type,
10432 			    void *type_data)
10433 {
10434 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(dev);
10435 
10436 	switch (type) {
10437 	case TC_SETUP_BLOCK:
10438 		return flow_block_cb_setup_simple(type_data,
10439 						  &ixgbe_block_cb_list,
10440 						  ixgbe_setup_tc_block_cb,
10441 						  adapter, adapter, true);
10442 	case TC_SETUP_QDISC_MQPRIO:
10443 		return ixgbe_setup_tc_mqprio(dev, type_data);
10444 	default:
10445 		return -EOPNOTSUPP;
10446 	}
10447 }
10448 
10449 #ifdef CONFIG_PCI_IOV
10450 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
10451 {
10452 	struct net_device *netdev = adapter->netdev;
10453 
10454 	rtnl_lock();
10455 	ixgbe_setup_tc(netdev, adapter->hw_tcs);
10456 	rtnl_unlock();
10457 }
10458 
10459 #endif
10460 void ixgbe_do_reset(struct net_device *netdev)
10461 {
10462 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
10463 
10464 	if (netif_running(netdev))
10465 		ixgbe_reinit_locked(adapter);
10466 	else
10467 		ixgbe_reset(adapter);
10468 }
10469 
10470 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
10471 					    netdev_features_t features)
10472 {
10473 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
10474 
10475 	/* If Rx checksum is disabled, then RSC/LRO should also be disabled */
10476 	if (!(features & NETIF_F_RXCSUM))
10477 		features &= ~NETIF_F_LRO;
10478 
10479 	/* Turn off LRO if not RSC capable */
10480 	if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
10481 		features &= ~NETIF_F_LRO;
10482 
10483 	if (adapter->xdp_prog && (features & NETIF_F_LRO)) {
10484 		e_dev_err("LRO is not supported with XDP\n");
10485 		features &= ~NETIF_F_LRO;
10486 	}
10487 
10488 	return features;
10489 }
10490 
10491 static void ixgbe_reset_l2fw_offload(struct ixgbe_adapter *adapter)
10492 {
10493 	int rss = min_t(int, ixgbe_max_rss_indices(adapter),
10494 			num_online_cpus());
10495 
10496 	/* go back to full RSS if we're not running SR-IOV */
10497 	if (!adapter->ring_feature[RING_F_VMDQ].offset)
10498 		adapter->flags &= ~(IXGBE_FLAG_VMDQ_ENABLED |
10499 				    IXGBE_FLAG_SRIOV_ENABLED);
10500 
10501 	adapter->ring_feature[RING_F_RSS].limit = rss;
10502 	adapter->ring_feature[RING_F_VMDQ].limit = 1;
10503 
10504 	ixgbe_setup_tc(adapter->netdev, adapter->hw_tcs);
10505 }
10506 
10507 static int ixgbe_set_features(struct net_device *netdev,
10508 			      netdev_features_t features)
10509 {
10510 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(netdev);
10511 	netdev_features_t changed = netdev->features ^ features;
10512 	bool need_reset = false;
10513 
10514 	/* Make sure RSC matches LRO, reset if change */
10515 	if (!(features & NETIF_F_LRO)) {
10516 		if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
10517 			need_reset = true;
10518 		adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
10519 	} else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
10520 		   !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
10521 		if (adapter->rx_itr_setting == 1 ||
10522 		    adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
10523 			adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
10524 			need_reset = true;
10525 		} else if ((changed ^ features) & NETIF_F_LRO) {
10526 			e_info(probe, "rx-usecs set too low, "
10527 			       "disabling RSC\n");
10528 		}
10529 	}
10530 
10531 	/*
10532 	 * Check if Flow Director n-tuple support or hw_tc support was
10533 	 * enabled or disabled.  If the state changed, we need to reset.
10534 	 */
10535 	if ((features & NETIF_F_NTUPLE) || (features & NETIF_F_HW_TC)) {
10536 		/* turn off ATR, enable perfect filters and reset */
10537 		if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
10538 			need_reset = true;
10539 
10540 		adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
10541 		adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
10542 	} else {
10543 		/* turn off perfect filters, enable ATR and reset */
10544 		if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
10545 			need_reset = true;
10546 
10547 		adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
10548 
10549 		/* We cannot enable ATR if SR-IOV is enabled */
10550 		if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED ||
10551 		    /* We cannot enable ATR if we have 2 or more tcs */
10552 		    (adapter->hw_tcs > 1) ||
10553 		    /* We cannot enable ATR if RSS is disabled */
10554 		    (adapter->ring_feature[RING_F_RSS].limit <= 1) ||
10555 		    /* A sample rate of 0 indicates ATR disabled */
10556 		    (!adapter->atr_sample_rate))
10557 			; /* do nothing not supported */
10558 		else /* otherwise supported and set the flag */
10559 			adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
10560 	}
10561 
10562 	if (changed & NETIF_F_RXALL)
10563 		need_reset = true;
10564 
10565 	netdev->features = features;
10566 
10567 	if ((changed & NETIF_F_HW_L2FW_DOFFLOAD) && adapter->num_rx_pools > 1)
10568 		ixgbe_reset_l2fw_offload(adapter);
10569 	else if (need_reset)
10570 		ixgbe_do_reset(netdev);
10571 	else if (changed & (NETIF_F_HW_VLAN_CTAG_RX |
10572 			    NETIF_F_HW_VLAN_CTAG_FILTER))
10573 		ixgbe_set_rx_mode(netdev);
10574 
10575 	return 1;
10576 }
10577 
10578 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
10579 			     struct net_device *dev,
10580 			     const unsigned char *addr, u16 vid,
10581 			     u16 flags, bool *notified,
10582 			     struct netlink_ext_ack *extack)
10583 {
10584 	/* guarantee we can provide a unique filter for the unicast address */
10585 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
10586 		struct ixgbe_adapter *adapter = ixgbe_from_netdev(dev);
10587 		u16 pool = VMDQ_P(0);
10588 
10589 		if (netdev_uc_count(dev) >= ixgbe_available_rars(adapter, pool))
10590 			return -ENOMEM;
10591 	}
10592 
10593 	return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
10594 }
10595 
10596 /**
10597  * ixgbe_configure_bridge_mode - set various bridge modes
10598  * @adapter: the private structure
10599  * @mode: requested bridge mode
10600  *
10601  * Configure some settings require for various bridge modes.
10602  **/
10603 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
10604 				       __u16 mode)
10605 {
10606 	struct ixgbe_hw *hw = &adapter->hw;
10607 	unsigned int p, num_pools;
10608 	u32 vmdctl;
10609 
10610 	switch (mode) {
10611 	case BRIDGE_MODE_VEPA:
10612 		/* disable Tx loopback, rely on switch hairpin mode */
10613 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
10614 
10615 		/* must enable Rx switching replication to allow multicast
10616 		 * packet reception on all VFs, and to enable source address
10617 		 * pruning.
10618 		 */
10619 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
10620 		vmdctl |= IXGBE_VT_CTL_REPLEN;
10621 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
10622 
10623 		/* enable Rx source address pruning. Note, this requires
10624 		 * replication to be enabled or else it does nothing.
10625 		 */
10626 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
10627 		for (p = 0; p < num_pools; p++) {
10628 			if (hw->mac.ops.set_source_address_pruning)
10629 				hw->mac.ops.set_source_address_pruning(hw,
10630 								       true,
10631 								       p);
10632 		}
10633 		break;
10634 	case BRIDGE_MODE_VEB:
10635 		/* enable Tx loopback for internal VF/PF communication */
10636 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
10637 				IXGBE_PFDTXGSWC_VT_LBEN);
10638 
10639 		/* disable Rx switching replication unless we have SR-IOV
10640 		 * virtual functions
10641 		 */
10642 		vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
10643 		if (!adapter->num_vfs)
10644 			vmdctl &= ~IXGBE_VT_CTL_REPLEN;
10645 		IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
10646 
10647 		/* disable Rx source address pruning, since we don't expect to
10648 		 * be receiving external loopback of our transmitted frames.
10649 		 */
10650 		num_pools = adapter->num_vfs + adapter->num_rx_pools;
10651 		for (p = 0; p < num_pools; p++) {
10652 			if (hw->mac.ops.set_source_address_pruning)
10653 				hw->mac.ops.set_source_address_pruning(hw,
10654 								       false,
10655 								       p);
10656 		}
10657 		break;
10658 	default:
10659 		return -EINVAL;
10660 	}
10661 
10662 	adapter->bridge_mode = mode;
10663 
10664 	e_info(drv, "enabling bridge mode: %s\n",
10665 	       mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
10666 
10667 	return 0;
10668 }
10669 
10670 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
10671 				    struct nlmsghdr *nlh, u16 flags,
10672 				    struct netlink_ext_ack *extack)
10673 {
10674 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(dev);
10675 	struct nlattr *attr, *br_spec;
10676 	int rem;
10677 
10678 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
10679 		return -EOPNOTSUPP;
10680 
10681 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
10682 	if (!br_spec)
10683 		return -EINVAL;
10684 
10685 	nla_for_each_nested_type(attr, IFLA_BRIDGE_MODE, br_spec, rem) {
10686 		__u16 mode = nla_get_u16(attr);
10687 		int status = ixgbe_configure_bridge_mode(adapter, mode);
10688 
10689 		if (status)
10690 			return status;
10691 
10692 		break;
10693 	}
10694 
10695 	return 0;
10696 }
10697 
10698 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
10699 				    struct net_device *dev,
10700 				    u32 filter_mask, int nlflags)
10701 {
10702 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(dev);
10703 
10704 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
10705 		return 0;
10706 
10707 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
10708 				       adapter->bridge_mode, 0, 0, nlflags,
10709 				       filter_mask, NULL);
10710 }
10711 
10712 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
10713 {
10714 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(pdev);
10715 	struct ixgbe_fwd_adapter *accel;
10716 	int tcs = adapter->hw_tcs ? : 1;
10717 	int pool, err;
10718 
10719 	if (adapter->xdp_prog) {
10720 		e_warn(probe, "L2FW offload is not supported with XDP\n");
10721 		return ERR_PTR(-EINVAL);
10722 	}
10723 
10724 	/* The hardware supported by ixgbe only filters on the destination MAC
10725 	 * address. In order to avoid issues we only support offloading modes
10726 	 * where the hardware can actually provide the functionality.
10727 	 */
10728 	if (!macvlan_supports_dest_filter(vdev))
10729 		return ERR_PTR(-EMEDIUMTYPE);
10730 
10731 	/* We need to lock down the macvlan to be a single queue device so that
10732 	 * we can reuse the tc_to_txq field in the macvlan netdev to represent
10733 	 * the queue mapping to our netdev.
10734 	 */
10735 	if (netif_is_multiqueue(vdev))
10736 		return ERR_PTR(-ERANGE);
10737 
10738 	pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
10739 	if (pool == adapter->num_rx_pools) {
10740 		u16 used_pools = adapter->num_vfs + adapter->num_rx_pools;
10741 		u16 reserved_pools;
10742 
10743 		if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
10744 		     adapter->num_rx_pools >= (MAX_TX_QUEUES / tcs)) ||
10745 		    adapter->num_rx_pools > IXGBE_MAX_MACVLANS)
10746 			return ERR_PTR(-EBUSY);
10747 
10748 		/* Hardware has a limited number of available pools. Each VF,
10749 		 * and the PF require a pool. Check to ensure we don't
10750 		 * attempt to use more then the available number of pools.
10751 		 */
10752 		if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
10753 			return ERR_PTR(-EBUSY);
10754 
10755 		/* Enable VMDq flag so device will be set in VM mode */
10756 		adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED |
10757 				  IXGBE_FLAG_SRIOV_ENABLED;
10758 
10759 		/* Try to reserve as many queues per pool as possible,
10760 		 * we start with the configurations that support 4 queues
10761 		 * per pools, followed by 2, and then by just 1 per pool.
10762 		 */
10763 		if (used_pools < 32 && adapter->num_rx_pools < 16)
10764 			reserved_pools = min_t(u16,
10765 					       32 - used_pools,
10766 					       16 - adapter->num_rx_pools);
10767 		else if (adapter->num_rx_pools < 32)
10768 			reserved_pools = min_t(u16,
10769 					       64 - used_pools,
10770 					       32 - adapter->num_rx_pools);
10771 		else
10772 			reserved_pools = 64 - used_pools;
10773 
10774 
10775 		if (!reserved_pools)
10776 			return ERR_PTR(-EBUSY);
10777 
10778 		adapter->ring_feature[RING_F_VMDQ].limit += reserved_pools;
10779 
10780 		/* Force reinit of ring allocation with VMDQ enabled */
10781 		err = ixgbe_setup_tc(pdev, adapter->hw_tcs);
10782 		if (err)
10783 			return ERR_PTR(err);
10784 
10785 		if (pool >= adapter->num_rx_pools)
10786 			return ERR_PTR(-ENOMEM);
10787 	}
10788 
10789 	accel = kzalloc(sizeof(*accel), GFP_KERNEL);
10790 	if (!accel)
10791 		return ERR_PTR(-ENOMEM);
10792 
10793 	set_bit(pool, adapter->fwd_bitmask);
10794 	netdev_set_sb_channel(vdev, pool);
10795 	accel->pool = pool;
10796 	accel->netdev = vdev;
10797 
10798 	if (!netif_running(pdev))
10799 		return accel;
10800 
10801 	err = ixgbe_fwd_ring_up(adapter, accel);
10802 	if (err)
10803 		return ERR_PTR(err);
10804 
10805 	return accel;
10806 }
10807 
10808 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
10809 {
10810 	struct ixgbe_fwd_adapter *accel = priv;
10811 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(pdev);
10812 	unsigned int rxbase = accel->rx_base_queue;
10813 	unsigned int i;
10814 
10815 	/* delete unicast filter associated with offloaded interface */
10816 	ixgbe_del_mac_filter(adapter, accel->netdev->dev_addr,
10817 			     VMDQ_P(accel->pool));
10818 
10819 	/* Allow remaining Rx packets to get flushed out of the
10820 	 * Rx FIFO before we drop the netdev for the ring.
10821 	 */
10822 	usleep_range(10000, 20000);
10823 
10824 	for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
10825 		struct ixgbe_ring *ring = adapter->rx_ring[rxbase + i];
10826 		struct ixgbe_q_vector *qv = ring->q_vector;
10827 
10828 		/* Make sure we aren't processing any packets and clear
10829 		 * netdev to shut down the ring.
10830 		 */
10831 		if (netif_running(adapter->netdev))
10832 			napi_synchronize(&qv->napi);
10833 		ring->netdev = NULL;
10834 	}
10835 
10836 	/* unbind the queues and drop the subordinate channel config */
10837 	netdev_unbind_sb_channel(pdev, accel->netdev);
10838 	netdev_set_sb_channel(accel->netdev, 0);
10839 
10840 	clear_bit(accel->pool, adapter->fwd_bitmask);
10841 	kfree(accel);
10842 }
10843 
10844 #define IXGBE_MAX_MAC_HDR_LEN		127
10845 #define IXGBE_MAX_NETWORK_HDR_LEN	511
10846 
10847 static netdev_features_t
10848 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
10849 		     netdev_features_t features)
10850 {
10851 	unsigned int network_hdr_len, mac_hdr_len;
10852 
10853 	/* Make certain the headers can be described by a context descriptor */
10854 	mac_hdr_len = skb_network_offset(skb);
10855 	if (unlikely(mac_hdr_len > IXGBE_MAX_MAC_HDR_LEN))
10856 		return features & ~(NETIF_F_HW_CSUM |
10857 				    NETIF_F_SCTP_CRC |
10858 				    NETIF_F_GSO_UDP_L4 |
10859 				    NETIF_F_HW_VLAN_CTAG_TX |
10860 				    NETIF_F_TSO |
10861 				    NETIF_F_TSO6);
10862 
10863 	network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
10864 	if (unlikely(network_hdr_len >  IXGBE_MAX_NETWORK_HDR_LEN))
10865 		return features & ~(NETIF_F_HW_CSUM |
10866 				    NETIF_F_SCTP_CRC |
10867 				    NETIF_F_GSO_UDP_L4 |
10868 				    NETIF_F_TSO |
10869 				    NETIF_F_TSO6);
10870 
10871 	/* We can only support IPV4 TSO in tunnels if we can mangle the
10872 	 * inner IP ID field, so strip TSO if MANGLEID is not supported.
10873 	 * IPsec offoad sets skb->encapsulation but still can handle
10874 	 * the TSO, so it's the exception.
10875 	 */
10876 	if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID)) {
10877 #ifdef CONFIG_IXGBE_IPSEC
10878 		if (!secpath_exists(skb))
10879 #endif
10880 			features &= ~NETIF_F_TSO;
10881 	}
10882 
10883 	return features;
10884 }
10885 
10886 static int ixgbe_xdp_setup(struct net_device *dev, struct bpf_prog *prog)
10887 {
10888 	int i, frame_size = dev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
10889 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(dev);
10890 	struct bpf_prog *old_prog;
10891 	bool need_reset;
10892 	int num_queues;
10893 
10894 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
10895 		return -EINVAL;
10896 
10897 	if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
10898 		return -EINVAL;
10899 
10900 	/* verify ixgbe ring attributes are sufficient for XDP */
10901 	for (i = 0; i < adapter->num_rx_queues; i++) {
10902 		struct ixgbe_ring *ring = adapter->rx_ring[i];
10903 
10904 		if (ring_is_rsc_enabled(ring))
10905 			return -EINVAL;
10906 
10907 		if (frame_size > ixgbe_rx_bufsz(ring))
10908 			return -EINVAL;
10909 	}
10910 
10911 	/* if the number of cpus is much larger than the maximum of queues,
10912 	 * we should stop it and then return with ENOMEM like before.
10913 	 */
10914 	if (nr_cpu_ids > IXGBE_MAX_XDP_QS * 2)
10915 		return -ENOMEM;
10916 
10917 	old_prog = xchg(&adapter->xdp_prog, prog);
10918 	need_reset = (!!prog != !!old_prog);
10919 
10920 	/* If transitioning XDP modes reconfigure rings */
10921 	if (need_reset) {
10922 		int err;
10923 
10924 		if (!prog)
10925 			/* Wait until ndo_xsk_wakeup completes. */
10926 			synchronize_rcu();
10927 		err = ixgbe_setup_tc(dev, adapter->hw_tcs);
10928 
10929 		if (err)
10930 			return -EINVAL;
10931 		if (!prog)
10932 			xdp_features_clear_redirect_target(dev);
10933 	} else {
10934 		for (i = 0; i < adapter->num_rx_queues; i++) {
10935 			WRITE_ONCE(adapter->rx_ring[i]->xdp_prog,
10936 				   adapter->xdp_prog);
10937 		}
10938 	}
10939 
10940 	if (old_prog)
10941 		bpf_prog_put(old_prog);
10942 
10943 	/* Kick start the NAPI context if there is an AF_XDP socket open
10944 	 * on that queue id. This so that receiving will start.
10945 	 */
10946 	if (need_reset && prog) {
10947 		num_queues = min_t(int, adapter->num_rx_queues,
10948 				   adapter->num_xdp_queues);
10949 		for (i = 0; i < num_queues; i++)
10950 			if (adapter->xdp_ring[i]->xsk_pool)
10951 				(void)ixgbe_xsk_wakeup(adapter->netdev, i,
10952 						       XDP_WAKEUP_RX);
10953 		xdp_features_set_redirect_target(dev, true);
10954 	}
10955 
10956 	return 0;
10957 }
10958 
10959 static int ixgbe_xdp(struct net_device *dev, struct netdev_bpf *xdp)
10960 {
10961 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(dev);
10962 
10963 	switch (xdp->command) {
10964 	case XDP_SETUP_PROG:
10965 		return ixgbe_xdp_setup(dev, xdp->prog);
10966 	case XDP_SETUP_XSK_POOL:
10967 		return ixgbe_xsk_pool_setup(adapter, xdp->xsk.pool,
10968 					    xdp->xsk.queue_id);
10969 
10970 	default:
10971 		return -EINVAL;
10972 	}
10973 }
10974 
10975 void ixgbe_xdp_ring_update_tail(struct ixgbe_ring *ring)
10976 {
10977 	/* Force memory writes to complete before letting h/w know there
10978 	 * are new descriptors to fetch.
10979 	 */
10980 	wmb();
10981 	writel(ring->next_to_use, ring->tail);
10982 }
10983 
10984 void ixgbe_xdp_ring_update_tail_locked(struct ixgbe_ring *ring)
10985 {
10986 	if (static_branch_unlikely(&ixgbe_xdp_locking_key))
10987 		spin_lock(&ring->tx_lock);
10988 	ixgbe_xdp_ring_update_tail(ring);
10989 	if (static_branch_unlikely(&ixgbe_xdp_locking_key))
10990 		spin_unlock(&ring->tx_lock);
10991 }
10992 
10993 static int ixgbe_xdp_xmit(struct net_device *dev, int n,
10994 			  struct xdp_frame **frames, u32 flags)
10995 {
10996 	struct ixgbe_adapter *adapter = ixgbe_from_netdev(dev);
10997 	struct ixgbe_ring *ring;
10998 	int nxmit = 0;
10999 	int i;
11000 
11001 	if (unlikely(test_bit(__IXGBE_DOWN, &adapter->state)))
11002 		return -ENETDOWN;
11003 
11004 	if (!netif_carrier_ok(adapter->netdev) ||
11005 	    !netif_running(adapter->netdev))
11006 		return -ENETDOWN;
11007 
11008 	if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
11009 		return -EINVAL;
11010 
11011 	/* During program transitions its possible adapter->xdp_prog is assigned
11012 	 * but ring has not been configured yet. In this case simply abort xmit.
11013 	 */
11014 	ring = adapter->xdp_prog ? ixgbe_determine_xdp_ring(adapter) : NULL;
11015 	if (unlikely(!ring))
11016 		return -ENXIO;
11017 
11018 	if (unlikely(test_bit(__IXGBE_TX_DISABLED, &ring->state)))
11019 		return -ENXIO;
11020 
11021 	if (static_branch_unlikely(&ixgbe_xdp_locking_key))
11022 		spin_lock(&ring->tx_lock);
11023 
11024 	for (i = 0; i < n; i++) {
11025 		struct xdp_frame *xdpf = frames[i];
11026 		int err;
11027 
11028 		err = ixgbe_xmit_xdp_ring(ring, xdpf);
11029 		if (err != IXGBE_XDP_TX)
11030 			break;
11031 		nxmit++;
11032 	}
11033 
11034 	if (unlikely(flags & XDP_XMIT_FLUSH))
11035 		ixgbe_xdp_ring_update_tail(ring);
11036 
11037 	if (static_branch_unlikely(&ixgbe_xdp_locking_key))
11038 		spin_unlock(&ring->tx_lock);
11039 
11040 	return nxmit;
11041 }
11042 
11043 static const struct net_device_ops ixgbe_netdev_ops = {
11044 	.ndo_open		= ixgbe_open,
11045 	.ndo_stop		= ixgbe_close,
11046 	.ndo_start_xmit		= ixgbe_xmit_frame,
11047 	.ndo_set_rx_mode	= ixgbe_set_rx_mode,
11048 	.ndo_validate_addr	= eth_validate_addr,
11049 	.ndo_set_mac_address	= ixgbe_set_mac,
11050 	.ndo_change_mtu		= ixgbe_change_mtu,
11051 	.ndo_tx_timeout		= ixgbe_tx_timeout,
11052 	.ndo_set_tx_maxrate	= ixgbe_tx_maxrate,
11053 	.ndo_vlan_rx_add_vid	= ixgbe_vlan_rx_add_vid,
11054 	.ndo_vlan_rx_kill_vid	= ixgbe_vlan_rx_kill_vid,
11055 	.ndo_eth_ioctl		= ixgbe_ioctl,
11056 	.ndo_set_vf_mac		= ixgbe_ndo_set_vf_mac,
11057 	.ndo_set_vf_vlan	= ixgbe_ndo_set_vf_vlan,
11058 	.ndo_set_vf_rate	= ixgbe_ndo_set_vf_bw,
11059 	.ndo_set_vf_spoofchk	= ixgbe_ndo_set_vf_spoofchk,
11060 	.ndo_set_vf_link_state	= ixgbe_ndo_set_vf_link_state,
11061 	.ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
11062 	.ndo_set_vf_trust	= ixgbe_ndo_set_vf_trust,
11063 	.ndo_get_vf_config	= ixgbe_ndo_get_vf_config,
11064 	.ndo_get_vf_stats	= ixgbe_ndo_get_vf_stats,
11065 	.ndo_get_stats64	= ixgbe_get_stats64,
11066 	.ndo_setup_tc		= __ixgbe_setup_tc,
11067 #ifdef IXGBE_FCOE
11068 	.ndo_select_queue	= ixgbe_select_queue,
11069 	.ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
11070 	.ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
11071 	.ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
11072 	.ndo_fcoe_enable = ixgbe_fcoe_enable,
11073 	.ndo_fcoe_disable = ixgbe_fcoe_disable,
11074 	.ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
11075 	.ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
11076 #endif /* IXGBE_FCOE */
11077 	.ndo_set_features = ixgbe_set_features,
11078 	.ndo_fix_features = ixgbe_fix_features,
11079 	.ndo_fdb_add		= ixgbe_ndo_fdb_add,
11080 	.ndo_bridge_setlink	= ixgbe_ndo_bridge_setlink,
11081 	.ndo_bridge_getlink	= ixgbe_ndo_bridge_getlink,
11082 	.ndo_dfwd_add_station	= ixgbe_fwd_add,
11083 	.ndo_dfwd_del_station	= ixgbe_fwd_del,
11084 	.ndo_features_check	= ixgbe_features_check,
11085 	.ndo_bpf		= ixgbe_xdp,
11086 	.ndo_xdp_xmit		= ixgbe_xdp_xmit,
11087 	.ndo_xsk_wakeup         = ixgbe_xsk_wakeup,
11088 	.ndo_hwtstamp_get	= ixgbe_ptp_hwtstamp_get,
11089 	.ndo_hwtstamp_set	= ixgbe_ptp_hwtstamp_set,
11090 };
11091 
11092 static void ixgbe_disable_txr_hw(struct ixgbe_adapter *adapter,
11093 				 struct ixgbe_ring *tx_ring)
11094 {
11095 	unsigned long wait_delay, delay_interval;
11096 	struct ixgbe_hw *hw = &adapter->hw;
11097 	u8 reg_idx = tx_ring->reg_idx;
11098 	int wait_loop;
11099 	u32 txdctl;
11100 
11101 	IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
11102 
11103 	/* delay mechanism from ixgbe_disable_tx */
11104 	delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
11105 
11106 	wait_loop = IXGBE_MAX_RX_DESC_POLL;
11107 	wait_delay = delay_interval;
11108 
11109 	while (wait_loop--) {
11110 		usleep_range(wait_delay, wait_delay + 10);
11111 		wait_delay += delay_interval * 2;
11112 		txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
11113 
11114 		if (!(txdctl & IXGBE_TXDCTL_ENABLE))
11115 			return;
11116 	}
11117 
11118 	e_err(drv, "TXDCTL.ENABLE not cleared within the polling period\n");
11119 }
11120 
11121 static void ixgbe_disable_txr(struct ixgbe_adapter *adapter,
11122 			      struct ixgbe_ring *tx_ring)
11123 {
11124 	set_bit(__IXGBE_TX_DISABLED, &tx_ring->state);
11125 	ixgbe_disable_txr_hw(adapter, tx_ring);
11126 }
11127 
11128 static void ixgbe_disable_rxr_hw(struct ixgbe_adapter *adapter,
11129 				 struct ixgbe_ring *rx_ring)
11130 {
11131 	unsigned long wait_delay, delay_interval;
11132 	struct ixgbe_hw *hw = &adapter->hw;
11133 	u8 reg_idx = rx_ring->reg_idx;
11134 	int wait_loop;
11135 	u32 rxdctl;
11136 
11137 	rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
11138 	rxdctl &= ~IXGBE_RXDCTL_ENABLE;
11139 	rxdctl |= IXGBE_RXDCTL_SWFLSH;
11140 
11141 	/* write value back with RXDCTL.ENABLE bit cleared */
11142 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
11143 
11144 	/* RXDCTL.EN may not change on 82598 if link is down, so skip it */
11145 	if (hw->mac.type == ixgbe_mac_82598EB &&
11146 	    !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
11147 		return;
11148 
11149 	/* delay mechanism from ixgbe_disable_rx */
11150 	delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
11151 
11152 	wait_loop = IXGBE_MAX_RX_DESC_POLL;
11153 	wait_delay = delay_interval;
11154 
11155 	while (wait_loop--) {
11156 		usleep_range(wait_delay, wait_delay + 10);
11157 		wait_delay += delay_interval * 2;
11158 		rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
11159 
11160 		if (!(rxdctl & IXGBE_RXDCTL_ENABLE))
11161 			return;
11162 	}
11163 
11164 	e_err(drv, "RXDCTL.ENABLE not cleared within the polling period\n");
11165 }
11166 
11167 static void ixgbe_reset_txr_stats(struct ixgbe_ring *tx_ring)
11168 {
11169 	memset(&tx_ring->stats, 0, sizeof(tx_ring->stats));
11170 	memset(&tx_ring->tx_stats, 0, sizeof(tx_ring->tx_stats));
11171 }
11172 
11173 static void ixgbe_reset_rxr_stats(struct ixgbe_ring *rx_ring)
11174 {
11175 	memset(&rx_ring->stats, 0, sizeof(rx_ring->stats));
11176 	memset(&rx_ring->rx_stats, 0, sizeof(rx_ring->rx_stats));
11177 }
11178 
11179 /**
11180  * ixgbe_irq_disable_single - Disable single IRQ vector
11181  * @adapter: adapter structure
11182  * @ring: ring index
11183  **/
11184 static void ixgbe_irq_disable_single(struct ixgbe_adapter *adapter, u32 ring)
11185 {
11186 	struct ixgbe_hw *hw = &adapter->hw;
11187 	u64 qmask = BIT_ULL(ring);
11188 	u32 mask;
11189 
11190 	switch (adapter->hw.mac.type) {
11191 	case ixgbe_mac_82598EB:
11192 		mask = qmask & IXGBE_EIMC_RTX_QUEUE;
11193 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, mask);
11194 		break;
11195 	case ixgbe_mac_82599EB:
11196 	case ixgbe_mac_X540:
11197 	case ixgbe_mac_X550:
11198 	case ixgbe_mac_X550EM_x:
11199 	case ixgbe_mac_x550em_a:
11200 		mask = (qmask & 0xFFFFFFFF);
11201 		if (mask)
11202 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
11203 		mask = (qmask >> 32);
11204 		if (mask)
11205 			IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
11206 		break;
11207 	default:
11208 		break;
11209 	}
11210 	IXGBE_WRITE_FLUSH(&adapter->hw);
11211 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
11212 		synchronize_irq(adapter->msix_entries[ring].vector);
11213 	else
11214 		synchronize_irq(adapter->pdev->irq);
11215 }
11216 
11217 /**
11218  * ixgbe_txrx_ring_disable - Disable Rx/Tx/XDP Tx rings
11219  * @adapter: adapter structure
11220  * @ring: ring index
11221  *
11222  * This function disables a certain Rx/Tx/XDP Tx ring. The function
11223  * assumes that the netdev is running.
11224  **/
11225 void ixgbe_txrx_ring_disable(struct ixgbe_adapter *adapter, int ring)
11226 {
11227 	struct ixgbe_ring *rx_ring, *tx_ring, *xdp_ring;
11228 
11229 	rx_ring = adapter->rx_ring[ring];
11230 	tx_ring = adapter->tx_ring[ring];
11231 	xdp_ring = adapter->xdp_ring[ring];
11232 
11233 	ixgbe_irq_disable_single(adapter, ring);
11234 
11235 	/* Rx/Tx/XDP Tx share the same napi context. */
11236 	napi_disable(&rx_ring->q_vector->napi);
11237 
11238 	ixgbe_disable_txr(adapter, tx_ring);
11239 	if (xdp_ring)
11240 		ixgbe_disable_txr(adapter, xdp_ring);
11241 	ixgbe_disable_rxr_hw(adapter, rx_ring);
11242 
11243 	if (xdp_ring)
11244 		synchronize_rcu();
11245 
11246 	ixgbe_clean_tx_ring(tx_ring);
11247 	if (xdp_ring)
11248 		ixgbe_clean_tx_ring(xdp_ring);
11249 	ixgbe_clean_rx_ring(rx_ring);
11250 
11251 	ixgbe_reset_txr_stats(tx_ring);
11252 	if (xdp_ring)
11253 		ixgbe_reset_txr_stats(xdp_ring);
11254 	ixgbe_reset_rxr_stats(rx_ring);
11255 }
11256 
11257 /**
11258  * ixgbe_txrx_ring_enable - Enable Rx/Tx/XDP Tx rings
11259  * @adapter: adapter structure
11260  * @ring: ring index
11261  *
11262  * This function enables a certain Rx/Tx/XDP Tx ring. The function
11263  * assumes that the netdev is running.
11264  **/
11265 void ixgbe_txrx_ring_enable(struct ixgbe_adapter *adapter, int ring)
11266 {
11267 	struct ixgbe_ring *rx_ring, *tx_ring, *xdp_ring;
11268 
11269 	rx_ring = adapter->rx_ring[ring];
11270 	tx_ring = adapter->tx_ring[ring];
11271 	xdp_ring = adapter->xdp_ring[ring];
11272 
11273 	ixgbe_configure_tx_ring(adapter, tx_ring);
11274 	if (xdp_ring)
11275 		ixgbe_configure_tx_ring(adapter, xdp_ring);
11276 	ixgbe_configure_rx_ring(adapter, rx_ring);
11277 
11278 	clear_bit(__IXGBE_TX_DISABLED, &tx_ring->state);
11279 	if (xdp_ring)
11280 		clear_bit(__IXGBE_TX_DISABLED, &xdp_ring->state);
11281 
11282 	/* Rx/Tx/XDP Tx share the same napi context. */
11283 	napi_enable(&rx_ring->q_vector->napi);
11284 	ixgbe_irq_enable_queues(adapter, BIT_ULL(ring));
11285 	IXGBE_WRITE_FLUSH(&adapter->hw);
11286 }
11287 
11288 /**
11289  * ixgbe_enumerate_functions - Get the number of ports this device has
11290  * @adapter: adapter structure
11291  *
11292  * This function enumerates the physical functions co-located on a single slot,
11293  * in order to determine how many ports a device has. This is most useful in
11294  * determining the required GT/s of PCIe bandwidth necessary for optimal
11295  * performance.
11296  **/
11297 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
11298 {
11299 	struct pci_dev *entry, *pdev = adapter->pdev;
11300 	int physfns = 0;
11301 
11302 	/* Some cards can not use the generic count PCIe functions method,
11303 	 * because they are behind a parent switch, so we hardcode these with
11304 	 * the correct number of functions.
11305 	 */
11306 	if (ixgbe_pcie_from_parent(&adapter->hw))
11307 		physfns = 4;
11308 
11309 	list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
11310 		/* don't count virtual functions */
11311 		if (entry->is_virtfn)
11312 			continue;
11313 
11314 		/* When the devices on the bus don't all match our device ID,
11315 		 * we can't reliably determine the correct number of
11316 		 * functions. This can occur if a function has been direct
11317 		 * attached to a virtual machine using VT-d, for example. In
11318 		 * this case, simply return -1 to indicate this.
11319 		 */
11320 		if ((entry->vendor != pdev->vendor) ||
11321 		    (entry->device != pdev->device))
11322 			return -1;
11323 
11324 		physfns++;
11325 	}
11326 
11327 	return physfns;
11328 }
11329 
11330 /**
11331  * ixgbe_wol_supported - Check whether device supports WoL
11332  * @adapter: the adapter private structure
11333  * @device_id: the device ID
11334  * @subdevice_id: the subsystem device ID
11335  *
11336  * This function is used by probe and ethtool to determine
11337  * which devices have WoL support
11338  *
11339  **/
11340 bool ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
11341 			 u16 subdevice_id)
11342 {
11343 	struct ixgbe_hw *hw = &adapter->hw;
11344 	u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
11345 
11346 	/* WOL not supported on 82598 */
11347 	if (hw->mac.type == ixgbe_mac_82598EB)
11348 		return false;
11349 
11350 	/* check eeprom to see if WOL is enabled for X540 and newer */
11351 	if (hw->mac.type >= ixgbe_mac_X540) {
11352 		if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
11353 		    ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
11354 		     (hw->bus.func == 0)))
11355 			return true;
11356 	}
11357 
11358 	/* WOL is determined based on device IDs for 82599 MACs */
11359 	switch (device_id) {
11360 	case IXGBE_DEV_ID_82599_SFP:
11361 		/* Only these subdevices could supports WOL */
11362 		switch (subdevice_id) {
11363 		case IXGBE_SUBDEV_ID_82599_560FLR:
11364 		case IXGBE_SUBDEV_ID_82599_LOM_SNAP6:
11365 		case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
11366 		case IXGBE_SUBDEV_ID_82599_SFP_2OCP:
11367 			/* only support first port */
11368 			if (hw->bus.func != 0)
11369 				break;
11370 			fallthrough;
11371 		case IXGBE_SUBDEV_ID_82599_SP_560FLR:
11372 		case IXGBE_SUBDEV_ID_82599_SFP:
11373 		case IXGBE_SUBDEV_ID_82599_RNDC:
11374 		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
11375 		case IXGBE_SUBDEV_ID_82599_SFP_1OCP:
11376 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM1:
11377 		case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM2:
11378 			return true;
11379 		}
11380 		break;
11381 	case IXGBE_DEV_ID_82599EN_SFP:
11382 		/* Only these subdevices support WOL */
11383 		switch (subdevice_id) {
11384 		case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
11385 			return true;
11386 		}
11387 		break;
11388 	case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
11389 		/* All except this subdevice support WOL */
11390 		if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
11391 			return true;
11392 		break;
11393 	case IXGBE_DEV_ID_82599_KX4:
11394 		return  true;
11395 	default:
11396 		break;
11397 	}
11398 
11399 	return false;
11400 }
11401 
11402 /**
11403  * ixgbe_set_fw_version_e610 - Set FW version specifically on E610 adapters
11404  * @adapter: the adapter private structure
11405  *
11406  * This function is used by probe and ethtool to determine the FW version to
11407  * format to display. The FW version is taken from the EEPROM/NVM.
11408  *
11409  */
11410 void ixgbe_set_fw_version_e610(struct ixgbe_adapter *adapter)
11411 {
11412 	struct ixgbe_orom_info *orom = &adapter->hw.flash.orom;
11413 	struct ixgbe_nvm_info *nvm = &adapter->hw.flash.nvm;
11414 
11415 	snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
11416 		 "%x.%02x 0x%x %d.%d.%d", nvm->major, nvm->minor,
11417 		 nvm->eetrack, orom->major, orom->build, orom->patch);
11418 }
11419 
11420 /**
11421  * ixgbe_set_fw_version - Set FW version
11422  * @adapter: the adapter private structure
11423  *
11424  * This function is used by probe and ethtool to determine the FW version to
11425  * format to display. The FW version is taken from the EEPROM/NVM.
11426  */
11427 static void ixgbe_set_fw_version(struct ixgbe_adapter *adapter)
11428 {
11429 	struct ixgbe_hw *hw = &adapter->hw;
11430 	struct ixgbe_nvm_version nvm_ver;
11431 
11432 	if (adapter->hw.mac.type == ixgbe_mac_e610) {
11433 		ixgbe_set_fw_version_e610(adapter);
11434 		return;
11435 	}
11436 
11437 	ixgbe_get_oem_prod_version(hw, &nvm_ver);
11438 	if (nvm_ver.oem_valid) {
11439 		snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
11440 			 "%x.%x.%x", nvm_ver.oem_major, nvm_ver.oem_minor,
11441 			 nvm_ver.oem_release);
11442 		return;
11443 	}
11444 
11445 	ixgbe_get_etk_id(hw, &nvm_ver);
11446 	ixgbe_get_orom_version(hw, &nvm_ver);
11447 
11448 	if (nvm_ver.or_valid) {
11449 		snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
11450 			 "0x%08x, %d.%d.%d", nvm_ver.etk_id, nvm_ver.or_major,
11451 			 nvm_ver.or_build, nvm_ver.or_patch);
11452 		return;
11453 	}
11454 
11455 	/* Set ETrack ID format */
11456 	snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
11457 		 "0x%08x", nvm_ver.etk_id);
11458 }
11459 
11460 /**
11461  * ixgbe_recovery_probe - Handle FW recovery mode during probe
11462  * @adapter: the adapter private structure
11463  *
11464  * Perform limited driver initialization when FW error is detected.
11465  *
11466  * Return: 0 on successful probe for E610, -EIO if recovery mode is detected
11467  * for non-E610 adapter, error status code on any other case.
11468  */
11469 static int ixgbe_recovery_probe(struct ixgbe_adapter *adapter)
11470 {
11471 	struct net_device *netdev = adapter->netdev;
11472 	struct pci_dev *pdev = adapter->pdev;
11473 	struct ixgbe_hw *hw = &adapter->hw;
11474 	bool disable_dev;
11475 	int err = -EIO;
11476 
11477 	if (hw->mac.type != ixgbe_mac_e610)
11478 		goto clean_up_probe;
11479 
11480 	ixgbe_get_hw_control(adapter);
11481 	mutex_init(&hw->aci.lock);
11482 	err = ixgbe_get_flash_data(&adapter->hw);
11483 	if (err)
11484 		goto shutdown_aci;
11485 
11486 	timer_setup(&adapter->service_timer, ixgbe_service_timer, 0);
11487 	INIT_WORK(&adapter->service_task, ixgbe_recovery_service_task);
11488 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
11489 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
11490 
11491 	if (hw->mac.ops.get_bus_info)
11492 		hw->mac.ops.get_bus_info(hw);
11493 
11494 	pci_set_drvdata(pdev, adapter);
11495 	/* We are creating devlink interface so NIC can be managed,
11496 	 * e.g. new NVM image loaded
11497 	 */
11498 	devl_lock(adapter->devlink);
11499 	ixgbe_devlink_register_port(adapter);
11500 	SET_NETDEV_DEVLINK_PORT(adapter->netdev,
11501 				&adapter->devlink_port);
11502 	ixgbe_devlink_init_regions(adapter);
11503 	devl_register(adapter->devlink);
11504 	devl_unlock(adapter->devlink);
11505 
11506 	return 0;
11507 shutdown_aci:
11508 	mutex_destroy(&adapter->hw.aci.lock);
11509 	ixgbe_release_hw_control(adapter);
11510 	devlink_free(adapter->devlink);
11511 clean_up_probe:
11512 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
11513 	free_netdev(netdev);
11514 	pci_release_mem_regions(pdev);
11515 	if (disable_dev)
11516 		pci_disable_device(pdev);
11517 	return err;
11518 }
11519 
11520 /**
11521  * ixgbe_probe - Device Initialization Routine
11522  * @pdev: PCI device information struct
11523  * @ent: entry in ixgbe_pci_tbl
11524  *
11525  * Returns 0 on success, negative on failure
11526  *
11527  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
11528  * The OS initialization, configuring of the adapter private structure,
11529  * and a hardware reset occur.
11530  **/
11531 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
11532 {
11533 	struct net_device *netdev;
11534 	struct ixgbe_netdevice_priv *netdev_priv_wrapper;
11535 	struct ixgbe_adapter *adapter = NULL;
11536 	struct ixgbe_hw *hw;
11537 	const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
11538 	unsigned int indices = MAX_TX_QUEUES;
11539 	u8 part_str[IXGBE_PBANUM_LENGTH];
11540 	int i, err, expected_gts;
11541 	bool disable_dev = false;
11542 #ifdef IXGBE_FCOE
11543 	u16 device_caps;
11544 #endif
11545 	u32 eec;
11546 
11547 	/* Catch broken hardware that put the wrong VF device ID in
11548 	 * the PCIe SR-IOV capability.
11549 	 */
11550 	if (pdev->is_virtfn) {
11551 		WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
11552 		     pci_name(pdev), pdev->vendor, pdev->device);
11553 		return -EINVAL;
11554 	}
11555 
11556 	err = pci_enable_device_mem(pdev);
11557 	if (err)
11558 		return err;
11559 
11560 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
11561 	if (err) {
11562 		dev_err(&pdev->dev,
11563 			"No usable DMA configuration, aborting\n");
11564 		goto err_dma;
11565 	}
11566 
11567 	err = pci_request_mem_regions(pdev, ixgbe_driver_name);
11568 	if (err) {
11569 		dev_err(&pdev->dev,
11570 			"pci_request_selected_regions failed 0x%x\n", err);
11571 		goto err_pci_reg;
11572 	}
11573 
11574 	pci_set_master(pdev);
11575 	pci_save_state(pdev);
11576 
11577 	if (ii->mac == ixgbe_mac_82598EB) {
11578 #ifdef CONFIG_IXGBE_DCB
11579 		/* 8 TC w/ 4 queues per TC */
11580 		indices = 4 * MAX_TRAFFIC_CLASS;
11581 #else
11582 		indices = IXGBE_MAX_RSS_INDICES;
11583 #endif
11584 	} else if (ii->mac == ixgbe_mac_e610) {
11585 		indices = IXGBE_MAX_RSS_INDICES_X550;
11586 	}
11587 
11588 	adapter = ixgbe_allocate_devlink(&pdev->dev);
11589 	if (IS_ERR(adapter)) {
11590 		err = PTR_ERR(adapter);
11591 		goto err_devlink;
11592 	}
11593 
11594 	netdev = alloc_etherdev_mq(sizeof(*netdev_priv_wrapper), indices);
11595 	if (!netdev) {
11596 		err = -ENOMEM;
11597 		goto err_alloc_etherdev;
11598 	}
11599 
11600 	SET_NETDEV_DEV(netdev, &pdev->dev);
11601 
11602 	netdev_priv_wrapper = netdev_priv(netdev);
11603 	netdev_priv_wrapper->adapter = adapter;
11604 
11605 	adapter->netdev = netdev;
11606 	adapter->pdev = pdev;
11607 	hw = &adapter->hw;
11608 	hw->back = adapter;
11609 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
11610 
11611 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
11612 			      pci_resource_len(pdev, 0));
11613 	adapter->io_addr = hw->hw_addr;
11614 	if (!hw->hw_addr) {
11615 		err = -EIO;
11616 		goto err_ioremap;
11617 	}
11618 
11619 	/* Setup hw api */
11620 	hw->mac.ops   = *ii->mac_ops;
11621 	hw->mac.type  = ii->mac;
11622 	hw->mvals     = ii->mvals;
11623 	if (ii->link_ops)
11624 		hw->link.ops  = *ii->link_ops;
11625 
11626 	/* EEPROM */
11627 	hw->eeprom.ops = *ii->eeprom_ops;
11628 	eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
11629 	if (ixgbe_removed(hw->hw_addr)) {
11630 		err = -EIO;
11631 		goto err_ioremap;
11632 	}
11633 	/* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
11634 	if (!(eec & BIT(8)))
11635 		hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
11636 
11637 	/* PHY */
11638 	hw->phy.ops = *ii->phy_ops;
11639 	hw->phy.sfp_type = ixgbe_sfp_type_unknown;
11640 	/* ixgbe_identify_phy_generic will set prtad and mmds properly */
11641 	hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
11642 	hw->phy.mdio.mmds = 0;
11643 	hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
11644 	hw->phy.mdio.dev = netdev;
11645 	hw->phy.mdio.mdio_read = ixgbe_mdio_read;
11646 	hw->phy.mdio.mdio_write = ixgbe_mdio_write;
11647 
11648 	netdev->netdev_ops = &ixgbe_netdev_ops;
11649 	ixgbe_set_ethtool_ops(netdev);
11650 	netdev->watchdog_timeo = 5 * HZ;
11651 	strscpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
11652 
11653 	/* setup the private structure */
11654 	err = ixgbe_sw_init(adapter, ii);
11655 	if (err)
11656 		goto err_sw_init;
11657 
11658 	if (ixgbe_check_fw_error(adapter))
11659 		return ixgbe_recovery_probe(adapter);
11660 
11661 	if (adapter->hw.mac.type == ixgbe_mac_e610) {
11662 		err = ixgbe_get_caps(&adapter->hw);
11663 		if (err)
11664 			dev_err(&pdev->dev, "ixgbe_get_caps failed %d\n", err);
11665 
11666 		err = ixgbe_get_flash_data(&adapter->hw);
11667 		if (err)
11668 			goto err_sw_init;
11669 	}
11670 
11671 	if (adapter->hw.mac.type == ixgbe_mac_82599EB)
11672 		adapter->flags2 |= IXGBE_FLAG2_AUTO_DISABLE_VF;
11673 
11674 	switch (adapter->hw.mac.type) {
11675 	case ixgbe_mac_X550:
11676 	case ixgbe_mac_X550EM_x:
11677 	case ixgbe_mac_e610:
11678 		netdev->udp_tunnel_nic_info = &ixgbe_udp_tunnels_x550;
11679 		break;
11680 	case ixgbe_mac_x550em_a:
11681 		netdev->udp_tunnel_nic_info = &ixgbe_udp_tunnels_x550em_a;
11682 		break;
11683 	default:
11684 		break;
11685 	}
11686 
11687 	/* Make it possible the adapter to be woken up via WOL */
11688 	switch (adapter->hw.mac.type) {
11689 	case ixgbe_mac_82599EB:
11690 	case ixgbe_mac_X540:
11691 	case ixgbe_mac_X550:
11692 	case ixgbe_mac_X550EM_x:
11693 	case ixgbe_mac_x550em_a:
11694 	case ixgbe_mac_e610:
11695 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
11696 		break;
11697 	default:
11698 		break;
11699 	}
11700 
11701 	/*
11702 	 * If there is a fan on this device and it has failed log the
11703 	 * failure.
11704 	 */
11705 	if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
11706 		u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
11707 		if (esdp & IXGBE_ESDP_SDP1)
11708 			e_crit(probe, "Fan has stopped, replace the adapter\n");
11709 	}
11710 
11711 	if (allow_unsupported_sfp)
11712 		hw->allow_unsupported_sfp = allow_unsupported_sfp;
11713 
11714 	/* reset_hw fills in the perm_addr as well */
11715 	hw->phy.reset_if_overtemp = true;
11716 	err = hw->mac.ops.reset_hw(hw);
11717 	hw->phy.reset_if_overtemp = false;
11718 	ixgbe_set_eee_capable(adapter);
11719 	if (err == -ENOENT) {
11720 		err = 0;
11721 	} else if (err == -EOPNOTSUPP) {
11722 		e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
11723 		e_dev_err("Reload the driver after installing a supported module.\n");
11724 		goto err_sw_init;
11725 	} else if (err) {
11726 		e_dev_err("HW Init failed: %d\n", err);
11727 		goto err_sw_init;
11728 	}
11729 
11730 #ifdef CONFIG_PCI_IOV
11731 	/* SR-IOV not supported on the 82598 */
11732 	if (adapter->hw.mac.type == ixgbe_mac_82598EB)
11733 		goto skip_sriov;
11734 	/* Mailbox */
11735 	ixgbe_init_mbx_params_pf(hw);
11736 	hw->mbx.ops = ii->mbx_ops;
11737 	pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
11738 	ixgbe_enable_sriov(adapter, max_vfs);
11739 skip_sriov:
11740 
11741 #endif
11742 	netdev->features = NETIF_F_SG |
11743 			   NETIF_F_TSO |
11744 			   NETIF_F_TSO6 |
11745 			   NETIF_F_RXHASH |
11746 			   NETIF_F_RXCSUM |
11747 			   NETIF_F_HW_CSUM;
11748 
11749 #define IXGBE_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
11750 				    NETIF_F_GSO_GRE_CSUM | \
11751 				    NETIF_F_GSO_IPXIP4 | \
11752 				    NETIF_F_GSO_IPXIP6 | \
11753 				    NETIF_F_GSO_UDP_TUNNEL | \
11754 				    NETIF_F_GSO_UDP_TUNNEL_CSUM)
11755 
11756 	netdev->gso_partial_features = IXGBE_GSO_PARTIAL_FEATURES;
11757 	netdev->features |= NETIF_F_GSO_PARTIAL |
11758 			    IXGBE_GSO_PARTIAL_FEATURES;
11759 
11760 	if (hw->mac.type >= ixgbe_mac_82599EB)
11761 		netdev->features |= NETIF_F_SCTP_CRC | NETIF_F_GSO_UDP_L4;
11762 
11763 #ifdef CONFIG_IXGBE_IPSEC
11764 #define IXGBE_ESP_FEATURES	(NETIF_F_HW_ESP | \
11765 				 NETIF_F_HW_ESP_TX_CSUM | \
11766 				 NETIF_F_GSO_ESP)
11767 
11768 	if (adapter->ipsec)
11769 		netdev->features |= IXGBE_ESP_FEATURES;
11770 #endif
11771 	/* copy netdev features into list of user selectable features */
11772 	netdev->hw_features |= netdev->features |
11773 			       NETIF_F_HW_VLAN_CTAG_FILTER |
11774 			       NETIF_F_HW_VLAN_CTAG_RX |
11775 			       NETIF_F_HW_VLAN_CTAG_TX |
11776 			       NETIF_F_RXALL |
11777 			       NETIF_F_HW_L2FW_DOFFLOAD;
11778 
11779 	if (hw->mac.type >= ixgbe_mac_82599EB)
11780 		netdev->hw_features |= NETIF_F_NTUPLE |
11781 				       NETIF_F_HW_TC;
11782 
11783 	netdev->features |= NETIF_F_HIGHDMA;
11784 
11785 	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
11786 	netdev->hw_enc_features |= netdev->vlan_features;
11787 	netdev->mpls_features |= NETIF_F_SG |
11788 				 NETIF_F_TSO |
11789 				 NETIF_F_TSO6 |
11790 				 NETIF_F_HW_CSUM;
11791 	netdev->mpls_features |= IXGBE_GSO_PARTIAL_FEATURES;
11792 
11793 	/* set this bit last since it cannot be part of vlan_features */
11794 	netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
11795 			    NETIF_F_HW_VLAN_CTAG_RX |
11796 			    NETIF_F_HW_VLAN_CTAG_TX;
11797 
11798 	netdev->priv_flags |= IFF_UNICAST_FLT;
11799 	netdev->priv_flags |= IFF_SUPP_NOFCS;
11800 
11801 	netdev->xdp_features = NETDEV_XDP_ACT_BASIC | NETDEV_XDP_ACT_REDIRECT |
11802 			       NETDEV_XDP_ACT_XSK_ZEROCOPY;
11803 
11804 	/* MTU range: 68 - 9710 */
11805 	netdev->min_mtu = ETH_MIN_MTU;
11806 	netdev->max_mtu = IXGBE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN);
11807 
11808 #ifdef CONFIG_IXGBE_DCB
11809 	if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
11810 		netdev->dcbnl_ops = &ixgbe_dcbnl_ops;
11811 #endif
11812 
11813 #ifdef IXGBE_FCOE
11814 	if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
11815 		unsigned int fcoe_l;
11816 
11817 		if (hw->mac.ops.get_device_caps) {
11818 			hw->mac.ops.get_device_caps(hw, &device_caps);
11819 			if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
11820 				adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
11821 		}
11822 
11823 
11824 		fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
11825 		adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
11826 
11827 		netdev->features |= NETIF_F_FSO |
11828 				    NETIF_F_FCOE_CRC;
11829 
11830 		netdev->vlan_features |= NETIF_F_FSO |
11831 					 NETIF_F_FCOE_CRC;
11832 	}
11833 #endif /* IXGBE_FCOE */
11834 	if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
11835 		netdev->hw_features |= NETIF_F_LRO;
11836 	if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
11837 		netdev->features |= NETIF_F_LRO;
11838 
11839 	/* make sure the EEPROM is good */
11840 	if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
11841 		e_dev_err("The EEPROM Checksum Is Not Valid\n");
11842 		err = -EIO;
11843 		goto err_sw_init;
11844 	}
11845 
11846 	eth_platform_get_mac_address(&adapter->pdev->dev,
11847 				     adapter->hw.mac.perm_addr);
11848 
11849 	eth_hw_addr_set(netdev, hw->mac.perm_addr);
11850 
11851 	if (!is_valid_ether_addr(netdev->dev_addr)) {
11852 		e_dev_err("invalid MAC address\n");
11853 		err = -EIO;
11854 		goto err_sw_init;
11855 	}
11856 
11857 	/* Set hw->mac.addr to permanent MAC address */
11858 	ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
11859 	ixgbe_mac_set_default_filter(adapter);
11860 
11861 	timer_setup(&adapter->service_timer, ixgbe_service_timer, 0);
11862 
11863 	if (ixgbe_removed(hw->hw_addr)) {
11864 		err = -EIO;
11865 		goto err_sw_init;
11866 	}
11867 	INIT_WORK(&adapter->service_task, ixgbe_service_task);
11868 	set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
11869 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
11870 
11871 	err = ixgbe_init_interrupt_scheme(adapter);
11872 	if (err)
11873 		goto err_sw_init;
11874 
11875 	for (i = 0; i < adapter->num_rx_queues; i++)
11876 		u64_stats_init(&adapter->rx_ring[i]->syncp);
11877 	for (i = 0; i < adapter->num_tx_queues; i++)
11878 		u64_stats_init(&adapter->tx_ring[i]->syncp);
11879 	for (i = 0; i < adapter->num_xdp_queues; i++)
11880 		u64_stats_init(&adapter->xdp_ring[i]->syncp);
11881 
11882 	/* WOL not supported for all devices */
11883 	adapter->wol = 0;
11884 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
11885 	hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
11886 						pdev->subsystem_device);
11887 	if (hw->wol_enabled)
11888 		adapter->wol = IXGBE_WUFC_MAG;
11889 
11890 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
11891 
11892 	/* save off EEPROM version number */
11893 	ixgbe_set_fw_version(adapter);
11894 
11895 	/* pick up the PCI bus settings for reporting later */
11896 	if (ixgbe_pcie_from_parent(hw))
11897 		ixgbe_get_parent_bus_info(adapter);
11898 	else
11899 		 hw->mac.ops.get_bus_info(hw);
11900 
11901 	/* calculate the expected PCIe bandwidth required for optimal
11902 	 * performance. Note that some older parts will never have enough
11903 	 * bandwidth due to being older generation PCIe parts. We clamp these
11904 	 * parts to ensure no warning is displayed if it can't be fixed.
11905 	 */
11906 	switch (hw->mac.type) {
11907 	case ixgbe_mac_82598EB:
11908 		expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
11909 		break;
11910 	default:
11911 		expected_gts = ixgbe_enumerate_functions(adapter) * 10;
11912 		break;
11913 	}
11914 
11915 	/* don't check link if we failed to enumerate functions */
11916 	if (expected_gts > 0)
11917 		ixgbe_check_minimum_link(adapter, expected_gts);
11918 
11919 	err = hw->eeprom.ops.read_pba_string(hw, part_str, sizeof(part_str));
11920 	if (err)
11921 		strscpy(part_str, "Unknown", sizeof(part_str));
11922 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
11923 		e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
11924 			   hw->mac.type, hw->phy.type, hw->phy.sfp_type,
11925 			   part_str);
11926 	else
11927 		e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
11928 			   hw->mac.type, hw->phy.type, part_str);
11929 
11930 	e_dev_info("%pM\n", netdev->dev_addr);
11931 
11932 	/* reset the hardware with the new settings */
11933 	err = hw->mac.ops.start_hw(hw);
11934 	if (err == -EACCES) {
11935 		/* We are running on a pre-production device, log a warning */
11936 		e_dev_warn("This device is a pre-production adapter/LOM. "
11937 			   "Please be aware there may be issues associated "
11938 			   "with your hardware.  If you are experiencing "
11939 			   "problems please contact your Intel or hardware "
11940 			   "representative who provided you with this "
11941 			   "hardware.\n");
11942 	}
11943 	strcpy(netdev->name, "eth%d");
11944 	pci_set_drvdata(pdev, adapter);
11945 
11946 	devl_lock(adapter->devlink);
11947 	ixgbe_devlink_register_port(adapter);
11948 	SET_NETDEV_DEVLINK_PORT(adapter->netdev, &adapter->devlink_port);
11949 
11950 	err = register_netdev(netdev);
11951 	if (err)
11952 		goto err_register;
11953 
11954 
11955 	/* power down the optics for 82599 SFP+ fiber */
11956 	if (hw->mac.ops.disable_tx_laser)
11957 		hw->mac.ops.disable_tx_laser(hw);
11958 
11959 	/* carrier off reporting is important to ethtool even BEFORE open */
11960 	netif_carrier_off(netdev);
11961 
11962 #ifdef CONFIG_IXGBE_DCA
11963 	if (dca_add_requester(&pdev->dev) == 0) {
11964 		adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
11965 		ixgbe_setup_dca(adapter);
11966 	}
11967 #endif
11968 	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
11969 		e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
11970 		for (i = 0; i < adapter->num_vfs; i++)
11971 			ixgbe_vf_configuration(pdev, (i | 0x10000000));
11972 	}
11973 
11974 	/* firmware requires driver version to be 0xFFFFFFFF
11975 	 * since os does not support feature
11976 	 */
11977 	if (hw->mac.ops.set_fw_drv_ver)
11978 		hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF, 0xFF,
11979 					   sizeof(UTS_RELEASE) - 1,
11980 					   UTS_RELEASE);
11981 
11982 	/* add san mac addr to netdev */
11983 	ixgbe_add_sanmac_netdev(netdev);
11984 
11985 	e_dev_info("%s\n", ixgbe_default_device_descr);
11986 
11987 #ifdef CONFIG_IXGBE_HWMON
11988 	if (ixgbe_sysfs_init(adapter))
11989 		e_err(probe, "failed to allocate sysfs resources\n");
11990 #endif /* CONFIG_IXGBE_HWMON */
11991 
11992 	ixgbe_dbg_adapter_init(adapter);
11993 
11994 	/* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
11995 	if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
11996 		hw->mac.ops.setup_link(hw,
11997 			IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
11998 			true);
11999 
12000 	err = ixgbe_mii_bus_init(hw);
12001 	if (err)
12002 		goto err_netdev;
12003 
12004 	ixgbe_devlink_init_regions(adapter);
12005 	devl_register(adapter->devlink);
12006 	devl_unlock(adapter->devlink);
12007 
12008 	if (ixgbe_fwlog_init(hw))
12009 		e_dev_info("Firmware logging not supported\n");
12010 
12011 	return 0;
12012 
12013 err_netdev:
12014 	unregister_netdev(netdev);
12015 err_register:
12016 	devl_port_unregister(&adapter->devlink_port);
12017 	devl_unlock(adapter->devlink);
12018 	ixgbe_release_hw_control(adapter);
12019 	ixgbe_clear_interrupt_scheme(adapter);
12020 err_sw_init:
12021 	if (hw->mac.type == ixgbe_mac_e610)
12022 		mutex_destroy(&adapter->hw.aci.lock);
12023 	ixgbe_disable_sriov(adapter);
12024 	adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
12025 	iounmap(adapter->io_addr);
12026 	kfree(adapter->jump_tables[0]);
12027 	kfree(adapter->mac_table);
12028 	kfree(adapter->rss_key);
12029 	bitmap_free(adapter->af_xdp_zc_qps);
12030 err_ioremap:
12031 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
12032 	free_netdev(netdev);
12033 err_alloc_etherdev:
12034 	devlink_free(adapter->devlink);
12035 	pci_release_mem_regions(pdev);
12036 err_devlink:
12037 err_pci_reg:
12038 err_dma:
12039 	if (!adapter || disable_dev)
12040 		pci_disable_device(pdev);
12041 	return err;
12042 }
12043 
12044 /**
12045  * ixgbe_remove - Device Removal Routine
12046  * @pdev: PCI device information struct
12047  *
12048  * ixgbe_remove is called by the PCI subsystem to alert the driver
12049  * that it should release a PCI device.  The could be caused by a
12050  * Hot-Plug event, or because the driver is going to be removed from
12051  * memory.
12052  **/
12053 static void ixgbe_remove(struct pci_dev *pdev)
12054 {
12055 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
12056 	struct net_device *netdev;
12057 	bool disable_dev;
12058 	int i;
12059 
12060 	/* if !adapter then we already cleaned up in probe */
12061 	if (!adapter)
12062 		return;
12063 
12064 	netdev  = adapter->netdev;
12065 	devl_lock(adapter->devlink);
12066 	devl_unregister(adapter->devlink);
12067 	ixgbe_devlink_destroy_regions(adapter);
12068 	ixgbe_fwlog_deinit(&adapter->hw);
12069 	ixgbe_dbg_adapter_exit(adapter);
12070 
12071 	set_bit(__IXGBE_REMOVING, &adapter->state);
12072 	cancel_work_sync(&adapter->service_task);
12073 
12074 	if (adapter->hw.mac.type == ixgbe_mac_e610)
12075 		ixgbe_disable_link_status_events(adapter);
12076 
12077 	if (adapter->mii_bus)
12078 		mdiobus_unregister(adapter->mii_bus);
12079 
12080 #ifdef CONFIG_IXGBE_DCA
12081 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
12082 		adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
12083 		dca_remove_requester(&pdev->dev);
12084 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
12085 				IXGBE_DCA_CTRL_DCA_DISABLE);
12086 	}
12087 
12088 #endif
12089 #ifdef CONFIG_IXGBE_HWMON
12090 	ixgbe_sysfs_exit(adapter);
12091 #endif /* CONFIG_IXGBE_HWMON */
12092 
12093 	/* remove the added san mac */
12094 	ixgbe_del_sanmac_netdev(netdev);
12095 
12096 #ifdef CONFIG_PCI_IOV
12097 	ixgbe_disable_sriov(adapter);
12098 #endif
12099 	if (netdev->reg_state == NETREG_REGISTERED)
12100 		unregister_netdev(netdev);
12101 
12102 	devl_port_unregister(&adapter->devlink_port);
12103 	devl_unlock(adapter->devlink);
12104 	devlink_free(adapter->devlink);
12105 
12106 	ixgbe_stop_ipsec_offload(adapter);
12107 	ixgbe_clear_interrupt_scheme(adapter);
12108 
12109 	ixgbe_release_hw_control(adapter);
12110 
12111 #ifdef CONFIG_DCB
12112 	kfree(adapter->ixgbe_ieee_pfc);
12113 	kfree(adapter->ixgbe_ieee_ets);
12114 
12115 #endif
12116 	iounmap(adapter->io_addr);
12117 	pci_release_mem_regions(pdev);
12118 
12119 	e_dev_info("complete\n");
12120 
12121 	for (i = 0; i < IXGBE_MAX_LINK_HANDLE; i++) {
12122 		if (adapter->jump_tables[i]) {
12123 			kfree(adapter->jump_tables[i]->input);
12124 			kfree(adapter->jump_tables[i]->mask);
12125 		}
12126 		kfree(adapter->jump_tables[i]);
12127 	}
12128 
12129 	kfree(adapter->mac_table);
12130 	kfree(adapter->rss_key);
12131 	bitmap_free(adapter->af_xdp_zc_qps);
12132 	disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
12133 	free_netdev(netdev);
12134 
12135 	if (adapter->hw.mac.type == ixgbe_mac_e610)
12136 		mutex_destroy(&adapter->hw.aci.lock);
12137 
12138 	if (disable_dev)
12139 		pci_disable_device(pdev);
12140 }
12141 
12142 /**
12143  * ixgbe_io_error_detected - called when PCI error is detected
12144  * @pdev: Pointer to PCI device
12145  * @state: The current pci connection state
12146  *
12147  * This function is called after a PCI bus error affecting
12148  * this device has been detected.
12149  */
12150 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
12151 						pci_channel_state_t state)
12152 {
12153 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
12154 	struct net_device *netdev = adapter->netdev;
12155 
12156 #ifdef CONFIG_PCI_IOV
12157 	struct ixgbe_hw *hw = &adapter->hw;
12158 	struct pci_dev *bdev, *vfdev;
12159 	u32 dw0, dw1, dw2, dw3;
12160 	int vf, pos;
12161 	u16 req_id, pf_func;
12162 
12163 	if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
12164 	    adapter->num_vfs == 0)
12165 		goto skip_bad_vf_detection;
12166 
12167 	bdev = pdev->bus->self;
12168 	while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
12169 		bdev = bdev->bus->self;
12170 
12171 	if (!bdev)
12172 		goto skip_bad_vf_detection;
12173 
12174 	pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
12175 	if (!pos)
12176 		goto skip_bad_vf_detection;
12177 
12178 	dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
12179 	dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
12180 	dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
12181 	dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
12182 	if (ixgbe_removed(hw->hw_addr))
12183 		goto skip_bad_vf_detection;
12184 
12185 	req_id = dw1 >> 16;
12186 	/* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
12187 	if (!(req_id & 0x0080))
12188 		goto skip_bad_vf_detection;
12189 
12190 	pf_func = req_id & 0x01;
12191 	if ((pf_func & 1) == (pdev->devfn & 1)) {
12192 		unsigned int device_id;
12193 
12194 		vf = FIELD_GET(0x7F, req_id);
12195 		e_dev_err("VF %d has caused a PCIe error\n", vf);
12196 		e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
12197 				"%8.8x\tdw3: %8.8x\n",
12198 		dw0, dw1, dw2, dw3);
12199 		switch (adapter->hw.mac.type) {
12200 		case ixgbe_mac_82599EB:
12201 			device_id = IXGBE_82599_VF_DEVICE_ID;
12202 			break;
12203 		case ixgbe_mac_X540:
12204 			device_id = IXGBE_X540_VF_DEVICE_ID;
12205 			break;
12206 		case ixgbe_mac_X550:
12207 			device_id = IXGBE_DEV_ID_X550_VF;
12208 			break;
12209 		case ixgbe_mac_X550EM_x:
12210 			device_id = IXGBE_DEV_ID_X550EM_X_VF;
12211 			break;
12212 		case ixgbe_mac_x550em_a:
12213 			device_id = IXGBE_DEV_ID_X550EM_A_VF;
12214 			break;
12215 		case ixgbe_mac_e610:
12216 			device_id = IXGBE_DEV_ID_E610_VF;
12217 			break;
12218 		default:
12219 			device_id = 0;
12220 			break;
12221 		}
12222 
12223 		/* Find the pci device of the offending VF */
12224 		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
12225 		while (vfdev) {
12226 			if (vfdev->devfn == (req_id & 0xFF))
12227 				break;
12228 			vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
12229 					       device_id, vfdev);
12230 		}
12231 		/*
12232 		 * There's a slim chance the VF could have been hot plugged,
12233 		 * so if it is no longer present we don't need to issue the
12234 		 * VFLR.  Just clean up the AER in that case.
12235 		 */
12236 		if (vfdev) {
12237 			pcie_flr(vfdev);
12238 			/* Free device reference count */
12239 			pci_dev_put(vfdev);
12240 		}
12241 	}
12242 
12243 	/*
12244 	 * Even though the error may have occurred on the other port
12245 	 * we still need to increment the vf error reference count for
12246 	 * both ports because the I/O resume function will be called
12247 	 * for both of them.
12248 	 */
12249 	adapter->vferr_refcount++;
12250 
12251 	return PCI_ERS_RESULT_RECOVERED;
12252 
12253 skip_bad_vf_detection:
12254 #endif /* CONFIG_PCI_IOV */
12255 	if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
12256 		return PCI_ERS_RESULT_DISCONNECT;
12257 
12258 	if (!netif_device_present(netdev))
12259 		return PCI_ERS_RESULT_DISCONNECT;
12260 
12261 	rtnl_lock();
12262 	netif_device_detach(netdev);
12263 
12264 	if (netif_running(netdev))
12265 		ixgbe_close_suspend(adapter);
12266 
12267 	if (state == pci_channel_io_perm_failure) {
12268 		rtnl_unlock();
12269 		return PCI_ERS_RESULT_DISCONNECT;
12270 	}
12271 
12272 	if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
12273 		pci_disable_device(pdev);
12274 	rtnl_unlock();
12275 
12276 	/* Request a slot reset. */
12277 	return PCI_ERS_RESULT_NEED_RESET;
12278 }
12279 
12280 /**
12281  * ixgbe_io_slot_reset - called after the pci bus has been reset.
12282  * @pdev: Pointer to PCI device
12283  *
12284  * Restart the card from scratch, as if from a cold-boot.
12285  */
12286 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
12287 {
12288 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
12289 	pci_ers_result_t result;
12290 
12291 	if (pci_enable_device_mem(pdev)) {
12292 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
12293 		result = PCI_ERS_RESULT_DISCONNECT;
12294 	} else {
12295 		smp_mb__before_atomic();
12296 		clear_bit(__IXGBE_DISABLED, &adapter->state);
12297 		adapter->hw.hw_addr = adapter->io_addr;
12298 		pci_set_master(pdev);
12299 		pci_restore_state(pdev);
12300 		pci_save_state(pdev);
12301 
12302 		pci_wake_from_d3(pdev, false);
12303 
12304 		ixgbe_reset(adapter);
12305 		IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
12306 		result = PCI_ERS_RESULT_RECOVERED;
12307 	}
12308 
12309 	return result;
12310 }
12311 
12312 /**
12313  * ixgbe_io_resume - called when traffic can start flowing again.
12314  * @pdev: Pointer to PCI device
12315  *
12316  * This callback is called when the error recovery driver tells us that
12317  * its OK to resume normal operation.
12318  */
12319 static void ixgbe_io_resume(struct pci_dev *pdev)
12320 {
12321 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
12322 	struct net_device *netdev = adapter->netdev;
12323 
12324 #ifdef CONFIG_PCI_IOV
12325 	if (adapter->vferr_refcount) {
12326 		e_info(drv, "Resuming after VF err\n");
12327 		adapter->vferr_refcount--;
12328 		return;
12329 	}
12330 
12331 #endif
12332 	rtnl_lock();
12333 	if (netif_running(netdev))
12334 		ixgbe_open(netdev);
12335 
12336 	netif_device_attach(netdev);
12337 	rtnl_unlock();
12338 }
12339 
12340 static const struct pci_error_handlers ixgbe_err_handler = {
12341 	.error_detected = ixgbe_io_error_detected,
12342 	.slot_reset = ixgbe_io_slot_reset,
12343 	.resume = ixgbe_io_resume,
12344 };
12345 
12346 static DEFINE_SIMPLE_DEV_PM_OPS(ixgbe_pm_ops, ixgbe_suspend, ixgbe_resume);
12347 
12348 static struct pci_driver ixgbe_driver = {
12349 	.name      = ixgbe_driver_name,
12350 	.id_table  = ixgbe_pci_tbl,
12351 	.probe     = ixgbe_probe,
12352 	.remove    = ixgbe_remove,
12353 	.driver.pm = pm_sleep_ptr(&ixgbe_pm_ops),
12354 	.shutdown  = ixgbe_shutdown,
12355 	.sriov_configure = ixgbe_pci_sriov_configure,
12356 	.err_handler = &ixgbe_err_handler
12357 };
12358 
12359 /**
12360  * ixgbe_init_module - Driver Registration Routine
12361  *
12362  * ixgbe_init_module is the first routine called when the driver is
12363  * loaded. All it does is register with the PCI subsystem.
12364  **/
12365 static int __init ixgbe_init_module(void)
12366 {
12367 	int ret;
12368 	pr_info("%s\n", ixgbe_driver_string);
12369 	pr_info("%s\n", ixgbe_copyright);
12370 
12371 	ixgbe_wq = create_singlethread_workqueue(ixgbe_driver_name);
12372 	if (!ixgbe_wq) {
12373 		pr_err("%s: Failed to create workqueue\n", ixgbe_driver_name);
12374 		return -ENOMEM;
12375 	}
12376 
12377 	ixgbe_dbg_init();
12378 
12379 	ret = pci_register_driver(&ixgbe_driver);
12380 	if (ret) {
12381 		destroy_workqueue(ixgbe_wq);
12382 		ixgbe_dbg_exit();
12383 		return ret;
12384 	}
12385 
12386 #ifdef CONFIG_IXGBE_DCA
12387 	dca_register_notify(&dca_notifier);
12388 #endif
12389 
12390 	return 0;
12391 }
12392 
12393 module_init(ixgbe_init_module);
12394 
12395 /**
12396  * ixgbe_exit_module - Driver Exit Cleanup Routine
12397  *
12398  * ixgbe_exit_module is called just before the driver is removed
12399  * from memory.
12400  **/
12401 static void __exit ixgbe_exit_module(void)
12402 {
12403 #ifdef CONFIG_IXGBE_DCA
12404 	dca_unregister_notify(&dca_notifier);
12405 #endif
12406 	pci_unregister_driver(&ixgbe_driver);
12407 
12408 	ixgbe_dbg_exit();
12409 	if (ixgbe_wq) {
12410 		destroy_workqueue(ixgbe_wq);
12411 		ixgbe_wq = NULL;
12412 	}
12413 }
12414 
12415 #ifdef CONFIG_IXGBE_DCA
12416 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
12417 			    void *p)
12418 {
12419 	int ret_val;
12420 
12421 	ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
12422 					 __ixgbe_notify_dca);
12423 
12424 	return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
12425 }
12426 
12427 #endif /* CONFIG_IXGBE_DCA */
12428 
12429 module_exit(ixgbe_exit_module);
12430 
12431 /* ixgbe_main.c */
12432